Datasets:

ccdv

/

arxiv-summarization

like 67

we consider compact lagrangian correspondences @xmath0 and @xmath1 , where @xmath2 are closed symplectic manifolds , and where @xmath3 . the * geometric composition * of the lagrangian correspondences is @xmath4 , the image under the projection @xmath5 of the fiber product @xmath6 here @xmath7 is the diagonal . if @xmath8 intersects @xmath9 transversely then @xmath10 is a lagrangian immersion ( see @xcite ) , in which case we call @xmath11 an * immersed composition*. in the case of * embedded composition * , where the projection is injective and hence a lagrangian embedding , monotonicity and maslov index assumptions allowed wehrheim woodward @xcite to establish an isomorphism of quilted floer cohomologies ( as defined in @xcite ) @xmath12 the analytic core of the proof was a * strip - shrinking degeneration * , in which a triple of pseudoholomorphic strips coupled by lagrangian seam conditions degenerates to a pair of strips , via the width of the middle strip shrinking to zero . the monotonicity and embeddedness assumptions allowed for an implicit exclusion of all bubbling , which was conjectured to include a novel * figure eight bubbling * that ( unlike disk or sphere bubbling ) could be an algebraic obstruction to . the current author and katrin wehrheim prove in @xcite that a blowup of the gradient in a sequence of pseudoholomorphic quilts with an annulus or strip of shrinking width gives rise to one of the standard bubbling phenomena ( pseudoholomorphic spheres and disks ) or a nontrivial * figure eight bubble * , as depicted in figure [ fig : viewsof8 ] . the latter is a tuple of finite energy pseudoholomorphic maps @xmath13\to m_0 , \qquad w_1\colon { \mathbb{r}}\times[-\tfrac 1 2 , \tfrac 1 2]\to m_1 , \qquad w_2\colon { \mathbb{r}}\times[\tfrac 1 2 , \infty)\to m_2\ ] ] satisfying the * seam conditions * @xmath14 in the current paper we apply this gromov compactness theorem to show that the figure eight singularity can be removed , as @xcite conjectured : * removal of singularity theorem [ thm : remsing ] : * if the composition @xmath15 is * cleanly immersed * ( immersed , and in addition the local branches of @xmath15 intersect one another cleanly ) , then @xmath16 resp . @xmath17 extend to continuous maps on @xmath18 ) \cup \{\infty\}$ ] resp . @xmath19 , and @xmath20 converges to constant paths as @xmath21 . if @xmath15 is embedded , then the latter limits are equal . this theorem is the first step in the program outlined in @xcite , which proposes a collection of composition operations amongst fukaya categories of different symplectic manifolds . in support of @xcite , appendix [ app : remsing ] also proves the analogous removal of singularities for pseudoholomorphic disks with a type of immersed boundary values in @xmath11 , under the assumption that the latter is cleanly - immersed resp . immersed . these results are not necessarily new , see appendix [ app : remsing ] for citations , but provided for the sake of completeness . it is also conceptually useful to recast the ( possibly singular ) disk bubbles with boundary on @xmath11 as * squashed eight bubbles * , that is as triples of finite energy pseudoholomorphic maps @xmath22\to m_0 , \qquad w_1\colon { \mathbb{r}}\to m_1 , \qquad w_2\colon { \mathbb{r}}\times[0 , \infty)\to m_2\end{aligned}\ ] ] satisfying the generalized seam condition @xmath23 there is a further logical dependence between @xcite and the current paper : in lemma [ lem : masterestimate ] we substantially strengthen the strip - shrinking estimates in @xcite in particular , from embedded to immersed geometric composition . these strengthened estimates form the analytic core of theorem [ thm : nonfoldedstripshrink ] , which is the analytic core of the gromov compactness theorem in @xcite . one of the ingredients in lemma [ lem : masterestimate ] is a special connection that allows us to obtain estimates without boundary terms for quilted cauchy riemann operators , with uniform constants for all small widths of a strip . this allows us to strengthen the previously - established uniform @xmath24 estimates to @xmath25 and thus @xmath26 , which is e.g. needed to deduce nontriviality of bubbles with generalized boundary condition in @xmath11 . our estimates allow for nonstandard complex structures on the shrinking strip . this is necessitated by the following analytic formulation for the figure eight singularity : in cylindrical coordinates for a neighborhood of infinity in , the two seams become two pairs of curves approaching each other asymptotically ( see the right figure in figure [ fig : viewsof8 ] ) . on finite cylinders , the standard complex structure on this quilted surface can be pulled back to a quilted surface in which the width of the strips is constant and the complex structures are nonstandard , but converge in @xmath27 and stay within a controlled @xmath28-distance from the standard structure for any @xmath29 . the hypothesis that @xmath2 are closed is not essential : as explained in @xcite , it is enough for the symplectic manifolds to be geometrically bounded and to have a priori @xmath27-bounds on the various pseudoholomorphic curves . in a future paper we will treat the noncompact setting in a more systematic way . the author is grateful to his phd advisor , katrin wehrheim , for suggesting in early 2012 that he study figure eight bubbles , and for generously sharing her knowledge throughout this project . casim abbas and helmut hofer shared their approach to a removal - of - singularity result in their unpublished book @xcite , which led to a crucial part of the argument in [ sec : remsing ] . the author acknowledges support by an nsf graduate research fellowship and a davidson fellowship , and would like to thank the institute for advanced study , princeton university , and the university of california , berkeley for their hospitality . in this section and the next we will be working with symplectic manifolds @xmath30 , almost complex structures @xmath31 , and pseudoholomorphic curves with seam conditions defined by compact lagrangian correspondences @xmath32 with @xmath15 either immersed or cleanly immersed : * @xmath33 and @xmath34 have * immersed composition * if the intersection @xmath35 is transverse . this implies that @xmath36 is a lagrangian immersion , e.g. by ( * ? ? ? * lemma 2.0.5 ) , and in this situation we will denote the image by @xmath37 . * if @xmath38 have immersed composition and furthermore any two local branches of @xmath15 intersect cleanly i.e. at any intersection of two local branches there is a chart for @xmath39 ( as a smooth manifold ) in which each of those two branches is identified with an open subset of a vector subspace of @xmath40 then the composition @xmath15 is * cleanly immersed*. the purpose of [ sec : remsing ] is to prove a removal of singularity theorem for inverted figure eight bubbles . [ def : inverted8 ] an * inverted figure eight bubble between @xmath41 and @xmath42 * is a triple of smooth maps @xmath43 satisfying the cauchy riemann equations @xmath44 for @xmath45 and the seam conditions @xmath46 and which have finite energy @xmath47 where we have endowed @xmath48 with the metric @xmath49 throughout [ sec : remsing ] , the norm of a tangent vector on @xmath48 will always be defined using @xmath50 . in fact , only the arguments in 2.2 require the composition @xmath15 to be cleanly immersed , rather than just immersed , but we assume the stronger hypothesis throughout [ sec : remsing ] for cohesiveness . the following theorem says that the singularity at 0 of a figure eight bubble can be continuously removed , under the hypothesis of cleanly - immersed composition . [ thm : remsing ] the maps @xmath51 continuously extend to 0 , and the limits @xmath52 and @xmath53 both exist . if moreover the immersion @xmath36 is an embedding , then the latter limits are equal so that @xmath54 also extends continuously to @xmath55 . the proof of this theorem draws on the removal of singularity strategies in @xcite and in @xcite . first , we follow @xcite and establish a uniform gradient bound in cylindrical coordinates near the puncture ( lemma [ lem : cylgradbound ] ) , which we use to show that the lengths of the paths @xmath56 converge to zero as @xmath57 ( lemma [ lem : lengthsgotozero ] ) . the substantial modification to the argument of @xcite is that we must use the gromov compactness theorem @xcite in order to prove uniform gradient bounds in lemma [ lem : cylgradbound ] . once we have proven that lengths go to zero , we follow @xcite and prove an isoperimetric inequality for the energy ( lemma [ lem : isoineq ] ) , which we use to show that the energy on disks around the puncture decays exponentially with respect to the logarithm of the radius . here the nontrivial modification is in the quilted nature of our isoperimetric inequality . finally , an argument from @xcite allows us to conclude that @xmath16 and @xmath17 extend continuously to the puncture . the continuous extension of @xmath54 follows from the gradient bound in cylindrical coordinates and the immersed composition of @xmath33 and @xmath34 . the formal proof of theorem [ thm : remsing ] is given in [ ss : isoineq ] . the first step toward the removal of singularity theorem [ thm : remsing ] is to show that the lengths of the paths @xmath56 converge to zero as @xmath57 . this is nontrivial since the conformal structure of the quilted surface near the singularity does not allow us to apply mean value inequalities effectively , as in previous removal of singularity results for pseudoholomorphic curves . hence the finiteness of energy only provides a sequence @xmath58 along which the lengths tend to zero . this allowed bottman wehrheim to deduce a weak removal of singularity in @xcite , but the stronger theorem [ thm : remsing ] will require the full strength of the generalized strip - shrinking analysis developed in [ sec : convergence ] and the resulting gromov compactness theorem in @xcite . we record a consequence of the latter as corollary [ cor : rescale ] below . in this subsection we will work in cylindrical coordinates centered at the singularity , hence we define the reparametrized maps @xmath59 whose domains @xmath60 \times { \mathbb{r}}/{\mathbb{z}}$ ] are given by @xmath61 with @xmath62 now the paths @xmath63 for fixed @xmath64 $ ] correspond to the following paths for fixed @xmath65 : @xmath66 \longrightarrow m_0 , \qquad \gamma_s^2 : = v_2(s , - ) \ , : \ ; [ \theta(s ) , \tfrac 1 2 - \theta(s ) ] \longrightarrow m_2 , \nonumber \\ \gamma_s^1 : = v_1(s , - ) \ , : \ ; [ \tfrac 1 2 - \theta(s ) , \tfrac 1 2 + \theta(s ) ] \cup [ 1 - \theta(s ) , 1 + \theta(s ) ] \longrightarrow m_1 . \label{eq : gamma}\end{gathered}\ ] ] the length of @xmath67 is given by the integral @xmath68 over the respective domain , and will be controlled by the following main result of this subsection . [ lem : lengthsgotozero ] the @xmath69-lengths of the paths @xmath70 defined in converge to zero as @xmath71 : @xmath72 in particular , the length @xmath73 tends to zero as @xmath71 . the proof of lemma [ lem : lengthsgotozero ] will use ideas from @xcite . the novel difficulty due to the conformal structure is to establish the following uniform gradient bound on @xmath74 , the upper semicontinuous function defined by @xmath75 are set to zero where they are not defined . [ lem : cylgradbound ] the gradient @xmath74 defined in is uniformly bounded . we prove lemma [ lem : cylgradbound ] by contradiction : if @xmath76 is not bounded for some @xmath77 , then there is a sequence of points @xmath78 ( necessarily with @xmath79 ) at which @xmath76 diverges . rescaling at these points produces a nonconstant quilted map , as illustrated in figure [ fig : gradbound ] , but this contradicts the finite - energy hypothesis on @xmath80 . the technical input is the gromov compactness theorem in @xcite , a consequence of which we record as theorem [ cor : rescale ] . this theorem is needed to deduce that the rescaled maps actually converge . in order to state it , we need to define the domains of the maps and a controlled fashion in which the strip - width can tend to zero . the following definition is the only instance in [ sec : remsing ] where we allow the almost complex structures to be domain - dependent , so that the notion of a squiggly strip quilt is flexible enough to be used in [ sec : convergence ] . fix @xmath81 , a real - analytic function @xmath82 \to ( 0 , \tfrac 1 2]$ ] , domain - dependent compatible almost complex structures @xmath83 ^ 2 \to { \mathcal{j}}(m_\ell,\om_\ell)$ ] , @xmath84 , and a complex structure @xmath85 on @xmath86 ^ 2 $ ] . * a * @xmath87-holomorphic size-@xmath88 squiggly strip quilt for @xmath89 * is a triple of smooth maps @xmath90 that fulfill the seam conditions @xmath91 satisfy the cauchy riemann equations @xmath92 for @xmath93 in the relevant domains , and have finite energy @xmath94 * a * @xmath95-holomorphic size-@xmath96 degenerate strip quilt for @xmath97 with singularities * is a triple of smooth maps @xmath98 \;{{\smallsetminus}}\ ; s\times\{0\ } \to m_0 \\ v_1\colon & \ , ( -\rho,\rho ) \;{{\smallsetminus}}\ ; s \to m_1 \\ v_2\colon & \ , ( -\rho,\rho)\times[0,\rho ) \;{{\smallsetminus}}\ ; s\times\{0\ } \to m_2 \end{aligned } \right)\end{aligned}\ ] ] defined on the complement of a finite set @xmath99 that fulfill the lifted seam condition @xmath100 satisfy the cauchy riemann equation for @xmath101 and @xmath93 in the relevant domains , and have finite energy @xmath102 when @xmath85 is the standard complex structure @xmath103 , the cauchy riemann equation can be expressed in coordinates as : @xmath104 the novel hypothesis necessary for a sequence of squiggly strip quilts of widths @xmath105 to converge @xmath106 away from the gradient blow - up points is that the widths `` obediently shrink to zero '' : [ def : obedience ] fix @xmath81 . a sequence @xmath107 of real - analytic functions @xmath108\to ( 0,\tfrac \rho2]$ ] * obediently shrinks to zero * , @xmath109 , if @xmath110 } f^\nu(s ) \underset{\nu\to\infty}{\longrightarrow } 0 $ ] and @xmath111 } \bigl| \tfrac{\rd^k}{\rd s^k}f^\nu(s ) \bigr|}{\min_{s\in[-\rho,\rho ] } f^\nu(s ) } = : c_k < \infty \qquad\forall \ : k\in{\mathbb{n}},\end{aligned}\ ] ] and in addition there are holomorphic extensions @xmath112 ^ 2 \to { \mathbb{c}}$ ] of @xmath113 such that @xmath114 converges @xmath115 to zero . the key to the following special case of the gromov compactness theorem from @xcite is a collection of width - independent elliptic estimates proven in [ sec : convergence ] for the linearized cauchy riemann operator . those elliptic estimates allow for a nonstandard domain complex structure , which is necessary in order to allow widths @xmath116 that are not constant in @xmath117 . [ cor : rescale ] fix @xmath81 , a sequence @xmath118 \to ( 0,\tfrac \rho 2])$ ] of real - analytic functions shrinking obediently to zero , and a sequence @xmath119 of @xmath120-holomorphic size-@xmath121 squiggly strip quilts for @xmath122 of bounded energy @xmath123 . if @xmath124 is a sequence of points where the gradient blows up , i.e. @xmath125 then there must be a concentration of energy @xmath126 at @xmath127 , in the sense that there is a sequence of radii @xmath128 such that : @xmath129 we are finally in the position to bound the gradients of the reparametrized maps @xmath130 from . we will prove the equivalent statement that the `` folded maps '' @xmath131 have uniformly - bounded gradients , where the domains @xmath132 are given by @xmath133 these maps are pseudoholomorphic with respect to the almost complex structures @xmath134 and satisfy the following boundary and seam conditions for @xmath135 : @xmath136 ( here @xmath137 as in , and @xmath138 is the image of @xmath139 under the permutation @xmath140 . ) finiteness of the energy of the inverted figure eight @xmath141 translates into convergence of the integral @xmath142 \times [ -1/4 , 1/4 ] } \tfrac 1 2|{{\rm d}}{\underline}u|^2 < \infty$ ] of the energy density @xmath143 are set to zero where they are not already defined ( so @xmath144 is upper semi - continuous ) . this convergence in particular implies @xmath145 \times [ -1/4 , 1/4 ] } \tfrac 1 2|{{\rm d}}{\underline}u|^2 \;\underset{s\to-\infty}{\longrightarrow } \ ; 0 .\ ] ] now assume for a contradiction that there exists a sequence @xmath146 \times \left[-1/4 , 1/4\right]$ ] such that @xmath147 . since the @xmath148 are smooth , this is possible only for @xmath149 ; passing to a further subsequence , we may in fact assume @xmath150 and @xmath151 . depending on whether @xmath152 is @xmath153 or is contained in @xmath154 , we derive a contradiction to : @xmath155 assume @xmath156 ; the @xmath157 case can be treated in analogous fashion . define a sequence @xmath158 by : @xmath159 the map @xmath160 is @xmath161-holomorphic and satisfies the lagrangian boundary condition @xmath162 for @xmath163 . furthermore , @xmath164 , @xmath165 by assumption , and the energy of @xmath160 is bounded by the energy of @xmath80 , so ( * ? ? ? * lemma 4.6.5 ) implies the inequality @xmath166 , which contradicts . @xmath167 define a sequence @xmath168 of @xmath169-holomorphic size-@xmath170 squiggly strip quilts , with @xmath171 \to ( 0 , \tfrac 1 2 ] , \qquad \theta^\nu(s ) : = \tfrac 1 { 2\pi}\arcsin(\tfrac 1 2e^{2\pi ( s + s^\nu)}),\end{aligned}\ ] ] by : @xmath172 the energy @xmath173 is bounded by the energy of @xmath80 , and by assumption , the gradient @xmath174 tends to @xmath175 . in the following sublemma we establish the last hypothesis needed to apply corollary [ cor : rescale ] . [ sublem : obedient ] the functions @xmath176 obediently shrink to zero as @xmath177 . the convergence @xmath79 implies @xmath178 in @xmath27 , so the equality @xmath179 implies the @xmath27-convergence of @xmath180 to zero . to check the second condition for obedient shrinking , fix @xmath29 and note that @xmath181 , with @xmath137 as above . the derivative @xmath182 is a linear combination of the functions @xmath183 for @xmath184 . ( this can be seen by induction starting from @xmath185 . ) this decomposition , the inequality @xmath186 for @xmath187 $ ] , and the convergence @xmath188 allows us to establish the second condition : @xmath189 } |f_\ell(s)| } { \min_{s \in [ -1/4,1/4 ] } \theta^\nu(s ) } & \leq \sup_{\nu \in { \mathbb{n } } } \frac { \exp(4\pi(\ell-\tfrac 1 2)(s^\nu + 1/4 ) ) } { \tfrac 1 { 4\pi}\exp(2\pi(s^\nu-1/4 ) ) } \\ & = \sup_{\nu \in { \mathbb{n } } } 4\pi\exp(4\pi((\ell-1)s^\nu + \tfrac 1 4 ) ) \\ & \leq 4\pi\exp(\pi).\end{aligned}\ ] ] the arcsine function extends to a holomorphic function @xmath190 by the power series @xmath191 , so @xmath116 extends to a holomorphic function @xmath192 from @xmath193 ^ 2 $ ] to @xmath194 . since the functions @xmath195 tend @xmath115 to zero and since @xmath179 , the extensions @xmath192 also tend @xmath115 to zero . part ( 2 ) of corollary [ cor : rescale ] now implies the inequality @xmath196 , which contradicts . first , note that the domain @xmath197 \cup [ 1 - \theta(s ) , 1 + \theta(s)]$ ] of @xmath198 has total length @xmath199 , which converges to @xmath55 as @xmath71 . hence the gradient bounds of lemma [ lem : cylgradbound ] immediately imply that the @xmath69-length of @xmath198 converges to zero as @xmath71 . moreover , these gradient bounds imply that to show the @xmath69-lengths of @xmath200 converge to zero , it suffices to fix an arbitrary @xmath201 and show that the @xmath69-lengths of @xmath202 } , \gamma_s^2|_{[1/2+{\epsilon } , 1-{\epsilon}]}$ ] converge to zero as @xmath71 . fix @xmath203 . we will show that the @xmath69-length of @xmath204}$ ] converges to zero as @xmath71 ; the proof for @xmath205 is similar . choose @xmath206 so that the domain of @xmath207 contains @xmath208 $ ] for all @xmath209 . now the @xmath27-bound on @xmath210 from lemma [ lem : cylgradbound ] induces a @xmath211-bound on @xmath212 \times [ 1/2+{\epsilon } , 1-{\epsilon}]}$ ] for any @xmath213 . indeed , we can apply the interior elliptic estimates ( e.g. @xcite ) on each of the precompactly - nested domains @xmath214 \times [ 1/2+{\epsilon},1-{\epsilon } ] \;\subset\ ; [ s_0 - k - 2 , s_0 - k + 1 ] \times [ 1/2+{\epsilon}/2,1-{\epsilon}/2]\end{aligned}\ ] ] for @xmath215 . since for different @xmath216 these domains are translations of one other , the constants in the elliptic estimates are independent of @xmath216 , and thus yield the desired @xmath211-bounds . for @xmath217 , define @xmath218 then @xmath219\to[0 , \infty)$ ] is nondecreasing with @xmath220 and @xmath221 where in the last quantity we are using the levi - civita connection with respect to the metric @xmath222 defined in . by the previous paragraph , there exists a constant @xmath223 so that @xmath224 for all @xmath225 . now for any fixed @xmath226 we can choose @xmath227 such that @xmath228 . for @xmath229 , we obtain : @xmath230 where the last step uses the bound on @xmath231 to deduce @xmath232 . this inequality can be rearranged to yield @xmath233 for all @xmath229 , and thus proves @xmath234 . since @xmath235 is equal to @xmath236)}$ ] and since @xmath237 is uniformly bounded , we have now shown that the @xmath69-norm of @xmath238 converges to zero as @xmath71 . the cauchy schwarz inequality implies that the @xmath239-norm of @xmath240 i.e. the length @xmath241 also tends to zero as @xmath71 . in this subsection , we prove theorem [ thm : remsing ] . the crucial inputs will be lemma [ lem : lengthsgotozero ] from [ ss : lengths ] together with the following isoperimetric inequality for the energy on @xmath242 \times { \mathbb{r}}/{\mathbb{z}}$ ] , @xmath243 \times { \mathbb{r}}/{\mathbb{z } } } \tfrac 1 2|{{\rm d}}{\underline}v|^2 \,{{\rm d}}s{{\rm d}}t.\end{aligned}\ ] ] [ lem : isoineq ] there exists @xmath244 such that the following inequality holds for all @xmath245 : @xmath246 we defer the proof to later in [ ss : isoineq ] ; now , we turn to the proof of removal of singularity . throughout this subsection we denote @xmath247 there exist @xmath248 such that the inequality @xmath249 holds for all @xmath245 . fix @xmath245 . the following inequality follows from lemma [ lem : isoineq ] : @xmath250 manipulating this inequality and integrating from @xmath117 to @xmath55 , we obtain @xmath251 . the limit @xmath252 exists in @xmath253 , m_0)$ ] . fix a @xmath26 embedding @xmath254 ; we will show that @xmath255})$ ] exists in @xmath27 . we begin by showing that @xmath256 exists in @xmath69 , where @xmath257,{\mathbb{r}}^n)$ ] is defined using the euclidean metric on @xmath258 . fix @xmath259 . schwarz implies the following inequality : @xmath260 ) } & = \left(\int_{5/8}^{7/8 } \left| \int_{s_2}^{s_1 } \partial_s ( i \circ v_0 ) \,{{\rm d}}s \right|^2 \,{{\rm d}}t \right)^{1/2 } \\ & \leq ( s_1 - s_2)^{1/2}\left ( \int_{5/8}^{7/8 } \int_{s_2}^{s_1 } \left|\partial_s(i \circ v_0)\right|_{g_{{{{\operatorname}{euc}}}}}^2 \,{{\rm d}}s { { \rm d}}t\right)^{1/2}. \nonumber\end{aligned}\ ] ] since @xmath261 is compact , there exists a constant of equivalence @xmath262 for the norms induced by @xmath263 and @xmath264 , so yields the following : @xmath265 ) } & \stackrel{\eqref{eq : l2viaminkowski}}{\leq } \mu(s_1 - s_2)^{1/2}\left ( \int_{5/8}^{7/8 } \int_{s_2}^{s_1 } |\partial_sv_0|_{g_{m_0}}^2 \,{{\rm d}}s { { \rm d}}t\right)^{1/2 } \nonumber \\ & \stackrel{\text{step 1}}{\leq } \mu \ : c_1^{1/2}(s_1 - s_2)^{1/2 } \exp(c_2s_1/2 ) \label{eq : anotherremsingcalc } \\ & = : c_3(s_1 - s_2)^{1/2}\exp(c_2s_1/2 ) . \nonumber\end{aligned}\ ] ] write @xmath266 for @xmath267 and @xmath268 . we have : @xmath269 ) } \nonumber \\ & \hspace{1.25in}\leq \|(i \circ v_0)(ms_1 , - ) - ( i \circ v_0)((m+\epsilon ) s_1 , -)\|_{l^2([5/8 , 7/8 ] ) } \nonumber \\ & \hspace{2.25 in } + \sum_{j=1}^{m-1 } \|(i \circ v_0)(js_1 , - ) - ( i \circ v_0)((j+1)s_1 , -)\|_{l^2([5/8 , 7/8 ] ) } \label{eq : yetanotherremsingcalc } \\ & \hspace{1.25in}\stackrel{\eqref{eq : anotherremsingcalc}}{\leq } c_3|s_1|^{1/2}\sum_{j=1}^m \exp(j c_2 s_1/2 ) \nonumber \\ & \hspace{1.25in}\leq \frac { c_3|s_1|^{1/2}\exp(c_2 s_1/2 ) } { 1 - \exp(c_2 s_1/2)}. \nonumber\end{aligned}\ ] ] define @xmath270)}$ ] . this quantity tends to zero as @xmath71 : @xmath271 ) } \stackrel{\text{lem.~\ref{lem : lengthsgotozero}}}{= } 0.\end{aligned}\ ] ] we can now show that @xmath256 exists in @xmath272 : we have @xmath273 which implies the equality @xmath274 } |(i \circ v_0)(s_1 , - ) - ( i \circ v_0)(s_2 , -)|_{w^{1,2}([5/8,7/8 ] ) } = 0.\end{aligned}\ ] ] since @xmath275 , { \mathbb{r}}^n)$ ] is complete , @xmath256 exists in @xmath272 . the sobolev embedding @xmath276 for one - dimensional domains now implies that @xmath256 exists in @xmath27 . we prove theorem [ thm : remsing ] . by lemma [ lem : lengthsgotozero ] , the first claim of theorem [ thm : remsing ] would follow from the existence of the limits @xmath277 it follows from step 2 that @xmath278 exists , and an analogous argument shows that @xmath279 exists . it remains to show that @xmath280 exist . to show that @xmath281 exists , we will show convergence of the path @xmath282 as @xmath71 . this path takes values in @xmath283 and @xmath284 ( by lemma [ lem : cylgradbound ] ) , so the distances @xmath285 converge to zero . hence there exists a path @xmath286 \to l_{01 } \times_{m_1 } l_{12}$ ] satisfying the equality @xmath287 ( indeed , define @xmath288 by choosing a tubular neighborhood @xmath289 of @xmath290 , and compose @xmath291 with the projection @xmath292 . ) we will show that @xmath293 exists by showing that @xmath294 exists . lemma [ lem : lengthsgotozero ] , the existence of @xmath278 and @xmath279 , and imply that @xmath295 exists . since @xmath296 restricts to an immersion of @xmath297 into @xmath298 , there exist finitely many preimages @xmath299 of @xmath300 in @xmath297 . choose @xmath201 small enough that the preimage of @xmath301 under @xmath302 consists of @xmath216 connected components @xmath303 , with @xmath304 contained in @xmath305 . now choose @xmath306 $ ] such that @xmath307))$ ] is contained in @xmath301 . the image @xmath308)$ ] must then be contained in a single @xmath309 . if @xmath310 are sequences with limit @xmath311 such that @xmath312 and @xmath313 exist , then @xmath314 and @xmath315 must be equal ; since @xmath297 is compact , this is enough to conclude that @xmath294 exists . as noted above , this is enough to conclude the first statement of theorem [ thm : remsing ] . the points @xmath316 and @xmath317 are lifts in @xmath297 of @xmath318 , so if the projection from @xmath290 to @xmath298 is injective , then @xmath280 are the same point . our proof of lemma [ lem : isoineq ] is an adaptation to the quilted setting of ( * ? ? ? * lemma 4.5.1 ) , which is an isoperimetric inequality for the energy near an interior point of a @xmath319-holomorphic curve . their argument went like this : restricting the map to an annulus , then reparametrizing , yields a map defined on the curved part of the boundary of a cylinder . by a lengths - go - to - zero result analogous to our lemma [ lem : lengthsgotozero ] , they extend this map to the entire cylinder . their result now follows from stokes theorem , along with the isoperimetric inequality for the symplectic area applied to the top and bottom caps of the cylinder . the difficulty in adapting this result to the quilted setting is in the extension to the cylinder ( see figure [ fig : isoineq ] for an illustration of the setup ) ; the key will be the consequences of cleanly - immersed composition recorded in the following lemma . [ lem : cleanconsequences ] there exist @xmath320 such that : * if @xmath321 have lifts @xmath322 with small distances @xmath323 then there exists a smooth path @xmath324 \to m_{02}$ ] with image in @xmath11 and smooth lifts @xmath325 \to l_{01}\times_{m_1}l_{12}$ ] that have bounded lengths @xmath326 and satisfy @xmath327 , @xmath328 , @xmath329 , and @xmath330 . * for @xmath331 with @xmath332 , there exists a point @xmath333 and preimages @xmath334 such that the following inequality holds : @xmath335 we will give only a brief sketch , since a formal proof is no more enlightening . the key is that the cleanly - immersed hypothesis implies that any two branches of @xmath15 meet like two vector subspaces . * if @xmath336 lie in the same local branch of @xmath11 , then the conclusion is immediate . otherwise , @xmath337 and @xmath338 lie in one branch , and @xmath339 and @xmath340 lie in another . represent these branches as open subsets of vector subspaces @xmath341 . then @xmath342 lie in @xmath343 , and we may define @xmath344 to be a path in @xmath343 from @xmath300 to @xmath345 and @xmath291 ( resp . @xmath346 ) to be the lift to the portion of @xmath297 corresponding to @xmath347 ( resp . to @xmath348 ) . * if @xmath349 lie in the same local branch of @xmath11 , the conclusion is again immediate . otherwise , represent the branches containing @xmath349 as open subsets of @xmath341 . set @xmath345 to be the nearest point in @xmath343 to @xmath337 , and let @xmath338 ( resp . @xmath340 ) be the lift to the portion of @xmath297 corresponding to @xmath347 ( resp . to @xmath348 ) . we prove lemma [ lem : isoineq ] up to an extension result , which we defer to steps 2 and 3 . it suffices to prove the lemma for @xmath350 , where @xmath206 is chosen so that @xmath351 is bounded by a constant @xmath226 to be determined later . as illustrated in figure [ fig : isoineqdomains ] , partition the unit circle @xmath352 into four segments by @xmath353 and set @xmath354 for @xmath355 . given @xmath356 with @xmath357 , define maps @xmath358 \to m_i$ ] , @xmath359 ( where we set @xmath360 ) like so : @xmath361 where we take @xmath362 $ ] . these maps satisfy the seam condition @xmath363 where we set @xmath364 , @xmath365 . in order to apply stokes theorem , we will extend the maps @xmath366 to the following four quadrants of the closed unit disk : @xmath367 choose @xmath368 such that for every @xmath85 , the diameters of the images @xmath369 ) $ ] are bounded by @xmath370 . as long as @xmath370 is small enough , steps 2 and 3 below allow us to extend @xmath366 to a continuous map @xmath371 \to m_i$ ] that is smooth on @xmath372 $ ] , such that the extended maps satisfy the lagrangian seam conditions @xmath373\end{aligned}\ ] ] indeed , use step 2 to define the maps @xmath374 on the slices @xmath375 , then use step 3 to extend @xmath374 to all of @xmath376 $ ] . since @xmath377 are closed , stokes theorem yields the following : @xmath378 \times { \mathbb{r}}/{\mathbb{z } } ) & \leq \sum_{i \in \{1 , 2\ } } \bigl| \sum_{j \in \{0,1,2,3\ } } \tint_{u_j \times \{s_i\ } } \sigma_j^*\om_j \bigr| \leq c \!\ ! \sum_{i \in \{1 , 2\ } } \sum_{j \in \{1,2,3\ } } \ell(\gamma_{s_i}^j)^2,\end{aligned}\ ] ] where in the first inequality we have used the seam conditions , and in the second inequality we have used the isoperimetric inequality for the symplectic area ( * ? ? ? * theorem 4.4.1 ) . taking the limit as @xmath379 goes to @xmath311 and applying lemma [ lem : lengthsgotozero ] yields the conclusion of the lemma . throughout the final two steps , the constants @xmath380 will depend only on the geometry of @xmath38 . there exist @xmath244 , @xmath381 so that if @xmath382 are smooth maps with @xmath383 then there exist extensions @xmath384 of @xmath366 such that : @xmath385 the points @xmath386 lie in @xmath387 . since the intersection @xmath388 defining @xmath297 is transverse , there are points @xmath389 that are close to @xmath390 resp . @xmath391 , @xmath392 for a uniform constant @xmath393 . the triangle inequality bounds the distance between the projections of @xmath394 : @xmath395 as long as @xmath396 is chosen to be small enough , it follows from lemma [ lem : cleanconsequences](ii ) that there exist lifts @xmath397 of a single point @xmath333 with small distances to @xmath390 resp . @xmath391 : @xmath398 where @xmath399 is another constant . we can now define the extensions @xmath374 at the origin : @xmath400 inequalities and and the triangle inequality yield : @xmath401 the local triviality of smooth submanifolds implies that there exists a constant @xmath402 such that after redefining @xmath396 if necessary , we may extend the maps @xmath374 to the set @xmath403 such that the seam conditions hold and the length of the loop @xmath404 is bounded by @xmath405 . once more redefining @xmath396 if necessary , we may extend each map @xmath374 to @xmath406 in such a way that the diameter of @xmath407 is bounded by @xmath408 for @xmath409 another constant . there exists @xmath410 such that the following holds . assume that @xmath382 are smooth maps and @xmath411 are real numbers with : @xmath412 \cup u_i \times \{a , b\ } \to m_i , \qquad \max_{i \in \{0,1,2,3\ } } { { \operatorname}{diam}}{{\operatorname}{im}}\sigma_i \leq \lambda , \\ ( \sigma_i(q),\sigma_{i+1}(q ) ) \in l_{i(i+1 ) } \qquad \forall \ : q \in \bigl(p_{i(i+1 ) } \times [ a , b]\bigr ) \cup \bigl((u_i \cap u_{i+1 } ) \times \{a , b\}\bigr ) .\end{gathered}\ ] ] then each @xmath366 can be extended to a smooth map @xmath413 \to m_i$ ] such that the following seam conditions hold : @xmath414.\end{aligned}\ ] ] define @xmath415 like so : @xmath416 then @xmath417 and @xmath418 , and @xmath337 resp . @xmath339 are close to @xmath338 resp . @xmath340 : @xmath419 it follows from lemma [ lem : cleanconsequences](i ) that as long as @xmath420 is chosen to be small enough , there exists a path @xmath421 \to l_{01 } \circ l_{12}$ ] and lifts @xmath422 \to l_{01 } \times_{m_1 } l_{12}$ ] from @xmath337 to @xmath338 resp . from @xmath339 to @xmath340 of small lengths : @xmath423 for @xmath424 a constant . define @xmath425 on @xmath426 $ ] like so : @xmath427 the diameter of the loop @xmath428)}$ ] is bounded by @xmath429 , so by redefining @xmath420 if necessary , we may extend @xmath430 to a map @xmath431 \to m_0 ^ - \times m_1 $ ] with small diameter : @xmath432 ) ) \leq c_6\lambda\end{aligned}\ ] ] for @xmath433 a constant . extend @xmath434 , @xmath435 , @xmath436 to @xmath437 $ ] , @xmath438 $ ] , @xmath439 $ ] in the same fashion . finally , @xmath440)}$ ] is a map to @xmath441 from a domain homeomorphic to @xmath442 , and its diameter is small : @xmath443 ) ) ) \leq ( 2c_6 + 1)\lambda.\end{aligned}\ ] ] redefining @xmath420 if necessary , we may extend @xmath374 to all of @xmath444 $ ] . the proof of the gromov compactness theorem in @xcite relies on @xmath28-compactness in the presence of a uniform gradient bound . this result is based on a strengthening of the strip - shrinking analysis of @xcite from @xmath24-convergence to @xmath28-convergence ; we also allow the domain to be equipped with nonstandard complex structures and the geometric composition @xmath15 to be immersed , rather than embedded . the purpose of this section is to establish convergence mod bubbling in theorem [ thm : nonfoldedstripshrink ] , deferring the @xmath370-independent sobolev estimate lemma [ lem : masterestimate ] to [ ss : estimate ] . for convenience , we will denote by @xmath445 , @xmath446 the symplectic manifolds @xmath447 and by @xmath448 the transposed lagrangian gotten by permuting the factors of @xmath449 by @xmath450 . the notion of `` symmetric complex structure '' in the following theorem will be defined in [ ss : csandacs ] . [ thm : nonfoldedstripshrink ] there exists @xmath201 such that the following holds : fix @xmath451 , positive reals @xmath452 and @xmath453 , symmetric complex structures @xmath454 on @xmath86 ^ 2 $ ] that converge @xmath115 to @xmath455 with @xmath456 , and @xmath457-bounded sequences of domain - dependent compatible almost complex structures @xmath458 ^ 2 \to { \mathcal{j}}_\ell(m_\ell,\om_\ell)$ ] , @xmath84 such that the @xmath459-limit of each @xmath460 is a compatible @xmath115 almost complex structures @xmath461 ^ 2 \to { \mathcal{j}}(m_\ell,\omega_\ell)$ ] . then if @xmath462 is a sequence of size-@xmath463 @xmath464-holomorphic squiggly strip quilts for @xmath465 with uniformly bounded gradients , @xmath466 ^ 2 } |{{\rm d}}v^\nu|(s , t ) < \infty,\end{aligned}\ ] ] then there is a subsequence in which @xmath467 , @xmath468 , @xmath469 converge @xmath470 to a @xmath471-holomorphic size-@xmath96 degenerate strip quilt @xmath472 for @xmath297 . if the inequality @xmath473 ^ 2 } | { { \rm d}}v^\nu |(s , t ) > 0 $ ] holds , then @xmath474 are not both constant . the analysis in our proof of theorem [ thm : nonfoldedstripshrink ] will be phrased in terms of pairs of smooth maps @xmath475 : @xmath476 \to m_{0211 } , \\ ( w_{02},{\widehat}w)(s , 0 ) \in \delta_{m_{02 } } \times \delta_{m_1 } , \qquad { \widehat}w(s , \delta ) \in ( l_{01}\times l_{12})^t \qquad \forall \ : s \in ( -\rho,\rho ) , \nonumber\end{gathered}\ ] ] where @xmath370 is nonnegative . from now on we denote the domains of @xmath477 and @xmath478 by @xmath479,\end{aligned}\ ] ] and combine them into the notation @xmath480 . we denote the closures in @xmath481 by @xmath482 \times [ 0,\rho-2\delta ] , \qquad { \overline}{{\widehat}q}_{\delta,\rho } : = [ -\rho,\rho ] \times [ 0,\delta].\end{aligned}\ ] ] for @xmath483 ( resp . @xmath484 ) , the setup @xmath485 is equivalent to a triple of smooth maps @xmath486 with the same domain and targets as a size-@xmath487 squiggly strip quilt for @xmath465 @xmath488 ( resp . as a size-@xmath96 degenerate strip quilt for @xmath297 ) and that fulfill the seam conditions @xmath489 ( resp . ) but are not necessarily pseudoholomorphic or of finite energy . indeed , given such @xmath486 , define @xmath490 like so : @xmath491 conversely , for @xmath492 and @xmath490 satisfying @xmath485 , define @xmath486 satisfying @xmath489 , @xmath489 ( for @xmath226 ) or , ( for @xmath493 ) like so : @xmath494 the transformations , are inverse to one another . the following proof of theorem [ thm : nonfoldedstripshrink ] uses several notions that will be defined in [ ss : csandacs][ss : estimate ] . we divide the proof into steps : in step 1 , we show that the squiggly strip quilts converge @xmath495 in a subsequence . in step 2 , we upgrade this convergence to @xmath470 . finally , we prove in step 3 that if the gradient satisfies a lower bound at a sequence of points with limit on the boundary , then at least one of @xmath496 is nonconstant . throughout this proof , @xmath497 will be a constant that may change from line to line . after passing to a subsequence , @xmath467 , @xmath468 , @xmath469 converge @xmath495 to a @xmath498-holomorphic size-@xmath96 degenerate strip quilt @xmath499 for @xmath297 . the arzel ascoli theorem implies that there exist continuous maps @xmath500 \to m_0 , \qquad v_1^\infty\colon ( -\rho,\rho ) \to m_1 , \qquad v_2^\infty\colon ( -\rho,\rho ) \times [ 0,\rho ) \to m_2\end{aligned}\ ] ] such that after passing to a subsequence , @xmath501 , @xmath468 , @xmath502 converge @xmath495 to @xmath503 , @xmath504 , @xmath505 . standard compactness for pseudoholomorphic curves ( e.g. ( * ? ? ? * theorem b.4.2 ) ) implies that this convergence takes place in @xmath470 on the interior ( i.e. away from the line @xmath506 ) ; in particular , @xmath503 resp . @xmath505 are @xmath507- resp . @xmath508-holomorphic on the interior , hence @xmath115 by ( * ? ? ? * theorem b.4.1 ) . in fact , we claim that @xmath503 and @xmath505 are @xmath115 on their full domains , and that they satisfy a generalized lagrangian boundary condition in @xmath297 at @xmath506 . denote by @xmath509 the map @xmath510 to show that @xmath503 , @xmath505 satisfy a generalized lagrangian boundary condition in @xmath15 , we will show that for any @xmath511 , @xmath512 lies in @xmath297 . the containment @xmath513 is clear . to show the containment @xmath514 , we will show that @xmath515 lies in @xmath33 ; the proof that @xmath516 lies in @xmath34 is analogous . since @xmath517 lies in @xmath33 , and since @xmath518 converges @xmath495 to @xmath504 , it suffices to show that the distances @xmath519 converge to zero . this follows from the uniform gradient bound on @xmath520 and the convergence of @xmath521 to zero . let us show that @xmath503 and @xmath505 are @xmath115 . we have already concluded that these maps are @xmath115 on the interior , so it only remains to show that they are @xmath115 at the boundary points , w.l.o.g . at @xmath522 . for that purpose we choose a neighborhood @xmath523 of @xmath524 such that @xmath525 is a smooth embedding . then @xmath526 is a noncompact embedded lagrangian , and since @xmath503 and @xmath505 are continuous we find @xmath201 such that @xmath527 is contained in @xmath289 . hence @xmath528 is contained in @xmath529 , so standard elliptic regularity ( e.g. ( * ? ? ? * theorem b.4.1 ) ) applied to the map @xmath530 shows that @xmath503 and @xmath505 are @xmath115 at @xmath522 . since @xmath531 is a diffeomorphism onto its image , @xmath509 is @xmath115 at @xmath55 and thus we have shown that @xmath532 are @xmath115 . after passing to a further subsequence , the convergence of @xmath501 , @xmath518 , @xmath533 takes place in @xmath470 . in order to establish @xmath470 convergence near @xmath534 , we can not rely on ( * ? ? ? * theorem b.4.2 ) . rather , we will establish uniform sobolev bounds for all three sequences of maps . the compact sobolev embeddings @xmath535 resp . @xmath536 for two - dimensional resp . one - dimensional domains will then provide @xmath470-convergent subsequences . set @xmath537 resp . @xmath538 to be the coherent pair of almost complex structures resp . coherent collection of complex structures resulting from the transformations resp . applied to @xmath539 resp . @xmath454 , and set @xmath540 to be the @xmath541-holomorphic size-@xmath463 folded strip quilt resulting from the transformation applied to @xmath462 . then @xmath542 resp . @xmath543 converge @xmath495 to @xmath544 resp . @xmath545 , where we have used the assumed @xmath26-bounds on @xmath546 , @xmath547 . since @xmath548 resp . @xmath549 resp . @xmath550 converge @xmath459 to @xmath507 resp . @xmath551 resp . @xmath508 , and since @xmath552 are @xmath553-bounded , @xmath554 resp . @xmath555 converge @xmath459 to @xmath556 resp . @xmath557 ; since @xmath454 converges in @xmath106 to the standard complex structure @xmath558 , the components of @xmath538 converge in @xmath106 to the standard coherent collection @xmath559 of complex structures , @xmath560 fix @xmath561 and choose @xmath562 . set @xmath563 to be the restriction and extension to @xmath564 of @xmath565 as defined in . due to the @xmath495-convergence of @xmath566 resp . @xmath567 to @xmath568 resp . @xmath569 and the uniform @xmath26-bounds on @xmath570 , we can express @xmath566 resp . @xmath570 on @xmath571 resp . @xmath572 for sufficiently large @xmath573 in terms of the corrected exponential maps @xmath574 resp . @xmath575 and sections @xmath576 as introduced in [ ss : estimate ] : @xmath577 the sections @xmath578 converge to zero in @xmath27 as @xmath177 , are uniformly bounded in @xmath26 , and satisfy boundary conditions in the linearizations of @xmath579 and @xmath283 . * iteration claim . * _ we bound @xmath580 and @xmath581 for @xmath582 $ ] by induction on @xmath583 , where @xmath584 and @xmath585 are the modified sobolev space and the linear delbar operator defined in [ ss : estimate ] using @xmath537 , @xmath538 , and the pair of connections @xmath586 constructed in lemma [ lem : conn ] . _ the first key fact for this claim is the formula @xmath587 justified in , where @xmath588 is the nonlinear delbar operator defined in . the relevant fact here is that @xmath589 is a pair of smooth maps @xmath590 that preserve fibers but do not necessarily respect their linear structure . furthermore , for any @xmath216 , @xmath589 is uniformly bounded in @xmath28 . the second key fact is lemma [ lem : masterestimate ] , which is a collection of @xmath370-independent elliptic estimates . since @xmath591 is uniformly bounded in @xmath26 , @xmath592 and @xmath593 are uniformly bounded . this establishes the base case of the iteration . next , say that @xmath591 and @xmath594 are uniformly bounded in @xmath595 for some @xmath596 $ ] . lemma [ lem : masterestimate ] yields : @xmath597 it remains to bound @xmath598 . since @xmath591 is uniformly bounded in @xmath599 , it is uniformly bounded in @xmath600 by lemma [ lem : improvedsobolev ] , which allows us to bound @xmath598 : @xmath601 this , together with , establishes the iteration step and completes the iteration claim . the uniform bounds on @xmath602 and the @xmath28-bounds that result from lemma [ lem : improvedsobolev ] yield uniform bounds on @xmath603 , @xmath604 , and @xmath605 . these bounds induce uniform bounds on the @xmath606-norms of @xmath607 on the relevant subdomains of @xmath608 and on the @xmath609-norms of @xmath610 . the compact embeddings @xmath611 resp . @xmath536 for two - dimensional resp . one - dimensional domains implies the desired @xmath470-convergence of @xmath612 resp . @xmath613 resp . @xmath614 to @xmath503 resp . @xmath504 resp . @xmath505 . we show that if for some @xmath84 and @xmath615 the gradient satisfies a lower bound @xmath616 for some @xmath617 , then at least one of @xmath474 is nonconstant . in the notation of step 2 , it suffices to show that if for some @xmath618 and @xmath615 the inequality @xmath619 is satisfied , then @xmath568 is not constant . we prove the contrapositive of this statement : assuming that @xmath568 is constant , we will show that the quantities @xmath620 and @xmath621 } | { { \rm d}}{\widehat}w^\nu(0,t ) |$ ] are both zero . since the convergence of @xmath622 to @xmath568 takes place in @xmath623 , the quantity @xmath624 is zero . to see that the quantity @xmath621 } |{{\rm d}}{\widehat}w^\nu(0,t)|$ ] is also zero , note that by the last paragraph of step 1 , the limit @xmath569 of @xmath625 is also constant , which implies the formula @xmath626 . it follows that to prove the equality @xmath627 } | { \widehat}\nabla{\widehat}w^\nu(0,t)| = 0 $ ] , it suffices to prove the equality @xmath627 } | { \widehat}\nabla { \widehat}\zeta^\nu(0,t)| = 0 $ ] . we can now estimate , using the sobolev inequality @xmath628 for one - dimensional domains whose lengths are bounded away from zero : @xmath629 } |{\widehat}\nabla { \widehat}\zeta^\nu(0,t)| & \leq \lim_{\nu\to\infty}\sup_{t \in [ 0,\delta^\nu ] } |{\widehat}\nabla{\widehat}\zeta^\nu(0,0)| + \lim_{\nu\to\infty } \sup_{t \in [ 0,\delta^\nu ] } |{\widehat}\nabla{\widehat}\zeta(0,t ) - { \widehat}\nabla{\widehat}\zeta(0,0)| \\ & = \lim_{\nu\to\infty } \sup_{t\in[0,\delta^\nu ] } |{\widehat}\nabla { \widehat}\zeta^\nu(0,t ) - { \widehat}\nabla { \widehat}\zeta^\nu(0,0)| \\ & \leq \lim_{\nu\to\infty } c_1\tint_0^{\delta^\nu } |{\widehat}\nabla_t{\widehat}\nabla{\widehat}\zeta^\nu(0,t)| \,{{\rm d}}t \\ & \leq \lim_{\nu\to\infty } c_1(\delta^\nu)^{1/2 } \left(\tint_0^{\delta^\nu } |{\widehat}\nabla_t{\widehat}\nabla{\widehat}\zeta^\nu(0,t)|^2 \,{{\rm d}}t\right)^{1/2 } \\ & { \stackrel}{\text{sobolev}}{\leq } \lim_{\nu\to\infty } c_1(\delta^\nu)^{1/2 } \|{\widehat}\zeta\|_{h^3({\widehat}q_{\delta^\nu,\rho } ) } = 0.\end{aligned}\ ] ] this completes the contrapositive of step 3 , which concludes our proof of theorem [ thm : nonfoldedstripshrink ] . the gromov compactness theorem in @xcite is proved by `` straightening '' the seams of a squiggly strip quilt . pushing forward the standard complex structure from the squiggly strip quilt to the new quilt with horizontal seams produces a nonstandard complex structure , which is symmetric under conjugation . we axiomatize this property in the following definition . fix @xmath453 . a * symmetric complex structure on @xmath630 ^ 2}$ ] * is a complex structure @xmath85 such that the equality @xmath631 holds for any @xmath632 ^ 2 $ ] , where @xmath633 is the conjugation @xmath634 . when a symmetric complex structure , almost complex structures , and a pseudoholomorphic squiggly strip quilt are `` pushed forward '' by the folding operation , the result is a `` coherent system of complex structures '' , a `` coherent pair of almost complex structures '' , and a `` pseudoholomorphic folded strip quilt '' , defined as follows . [ def : coherent ] fix @xmath226 and @xmath453 . * a * coherent collection of complex structures @xmath635 on @xmath636 * is a pair @xmath637 , where @xmath638 ( resp . @xmath639 ) are complex structures on @xmath640 ( resp . on @xmath641 ) such that the following equalities hold for all @xmath511 : @xmath642 * a * coherent pair of almost complex structures @xmath643 on @xmath636 * is a pair @xmath644 , where @xmath645 are almost complex structures @xmath646 satisfying the following compatibility condition : set @xmath647 resp . @xmath648 to be the inclusion resp . projection . then for any @xmath511 , the following equality must hold : @xmath649 * fix a coherent collection @xmath635 of complex structures and a coherent pair @xmath643 of almost complex structures on @xmath650 . a * @xmath651-holomorphic size-@xmath652 folded strip quilt * is a collection of smooth maps @xmath653 satisfying that have finite energy , @xmath654 and satisfy the cauchy riemann equations @xmath655 where @xmath656 is the pair of operators defined by : @xmath657 given a @xmath658-holomorphic squiggly strip quilt @xmath486 with @xmath85 symmetric , we can produce a folded strip quilt like this : define a coherent collection @xmath635 of complex structures by @xmath659 and a coherent pair @xmath643 of almost complex structures by @xmath660 if @xmath490 is defined by applying to @xmath486 , then @xmath661 is a @xmath662-holomorphic size-@xmath487 folded strip quilt . indeed , @xmath661 have the correct domains and codomains and satisfy the seam conditions , as discussed earlier , and the finite - energy hypothesis on @xmath486 implies that @xmath661 has finite energy . the cauchy riemann equation for @xmath663 on @xmath664 $ ] can be rewritten as @xmath665 for @xmath666 as in , so @xmath16 is @xmath667-holomorphic on @xmath668 . five similar calculations complete the check that @xmath661 is @xmath662-holomorphic . finally , we consider the coordinate representation of a coherent collection of complex structures . fix a coherent collection @xmath669 of complex structures on @xmath650 . define @xmath670 by @xmath671 and define @xmath672 for @xmath673 and @xmath674 for @xmath675 in the same way . then and translate into the following conditions on these coefficients : @xmath676 we will use this coordinate representation in [ ss : estimate ] . this subsection is devoted to proving lemma [ lem : masterestimate ] , which is the crucial @xmath370-independent elliptic estimate needed for the proof of theorem [ thm : nonfoldedstripshrink ] . furthermore , we continue to denote by @xmath559 the standard coherent collection of complex structures defined in , and for any @xmath677 $ ] we define a pair @xmath678 of smooth maps satisfying @xmath485 by : @xmath679 our approach is inspired by @xcite , but we deviate from that approach by working with a special connection which allows us to drop boundary terms from the @xmath680-estimate ( * ? ? ? * lemma 3.2.1(b ) ) . this special connection is constructed in the following lemma , which is a generalization to the immersed case of a connection constructed in @xcite . [ lem : conn ] there is an assignment @xmath681 that sends @xmath677 $ ] to a pair of connections @xmath682 resp . @xmath683 on @xmath684 resp . @xmath685 such that the following hold : * parallel transport under @xmath683 preserves @xmath686 and @xmath687 ; * for a section @xmath688 we have @xmath689 , where @xmath690 is the projection ; * for @xmath691 , the restrictions of @xmath692 agree : @xmath693 fix metrics on @xmath694 and @xmath695 so that given a smooth subbundle , we may form its orthogonal complement . for any fixed @xmath511 we denote : @xmath696 the transversality of @xmath697 implies @xmath698 , so the projection from @xmath699 to @xmath700 is injective ( see e.g. ( * ? ? ? * lemma 2.0.5 ) ) . hence the intersection of @xmath699 and @xmath701 is trivial . it follows that if we let @xmath497 denote the complement of @xmath702 in @xmath703 , the diagonal decomposes as @xmath704 . let @xmath705 be the complement of @xmath699 in @xmath706 . transversality implies @xmath707 , so we have deduced the following decomposition : @xmath708 the subspace @xmath706 ( resp . @xmath703 ) is given by the sum of the first two factors ( resp . the sum of the last three factors ) in this decomposition . therefore , if we choose connections on each of these four subbundles and set @xmath709 to be the product connection , then extend @xmath709 to a connection @xmath683 on @xmath685 by defining @xmath710 and defining @xmath711 in terms of the levi - civita connection @xmath712 , @xmath683 satisfies the first bullet . denote by @xmath713 projection and by @xmath714 the inclusion defined by sending @xmath715 to @xmath716 . define a connection @xmath717 on @xmath718 by @xmath719 . extend @xmath717 in any way to a connection @xmath720 on @xmath694 ; for @xmath677 $ ] , define @xmath721 . the second bullet now follows from a computation , in which @xmath722 is an arbitrary section of @xmath687 : @xmath723 the term @xmath724 in the third quantity vanishes since the subbundle @xmath725 is preserved under parallel transport by @xmath726 . we will use the connections @xmath727 just constructed throughout the rest of [ ss : estimate ] . due to the third property in lemma [ lem : conn ] , it is unambiguous to drop the subscript and refer to @xmath727 simply as @xmath728 . note that this pair of connections induce connections on the pullbacks by @xmath729 or @xmath730 of any tensor bundle of @xmath731 or @xmath732 in a canonical way . before we state the elliptic estimate lemma [ lem : masterestimate ] , we need to define our function spaces and delbar operators . [ def : functionspaces ] fix @xmath733 , @xmath226 , and @xmath734 . define the space of sections @xmath735 and the norms @xmath736 as follows . * define @xmath735 by : @xmath737 where denotes the following linearized boundary conditions : @xmath738 * define two norms @xmath739 on @xmath740 by : @xmath741 } \| { \widehat}\nabla^l{\widehat}\xi ( - , t ) \|_{h^1((-r , r),{\widehat}u_\delta(-,t)^*\rt m_{0211})}^2 \bigr ) .\end{aligned}\ ] ] note that @xmath742 is a well - defined norm on @xmath743 due to the embedding @xmath744 for one - dimensional domains . however , the constant in the bound @xmath745 is @xmath370-dependent . in @xcite , wehrheim woodward introduced an exponential map with quadratic corrections , which allowed them to treat the lagrangian boundary conditions as totally geodesic . woodward assumed the composition @xmath15 to be embedded , but their construction of the corrected exponential map only used the immersedness of that composition . we may therefore import their corrected exponential map into our setting : [ def : exp ] given @xmath746 and @xmath226 , define the * corrected exponential map @xmath747 * and its * linearization @xmath748 * and * @xmath749- and @xmath750-derivatives as follows*. * let @xmath751 be the pair of maps defined in ( * ? ? ? * lemma 3.1.2 ) ; @xmath752 sends @xmath753 with @xmath754 sufficiently small to a pair of maps @xmath755 satisfying . * for @xmath756 , @xmath757 is defined by including the fiber @xmath758 into @xmath759 as the vertical vectors , then postcomposing with the tangent map @xmath760 . the linearization @xmath761 is defined analogously . * for @xmath756 , define @xmath762 to be the vector gotten by choosing a flat section @xmath633 of @xmath763 for @xmath764 small , then setting @xmath765 . the derivatives @xmath766 are defined analogously , and each of these derivatives depends smoothly on the argument @xmath767 or @xmath768 . this exponential map will allow us to define fiberwise complex structures in the following , which are parametrized by vector fields rather than by maps . in the following definition of the linear delbar operator , we must go into coordinates . fix @xmath226 and a coherent collection @xmath669 of complex structures on @xmath650 . then @xmath635 induces via two pairs of endomorphisms @xmath769 , @xmath770 of @xmath771 , with @xmath772 defined as follows and @xmath773 defined in analogous fashion : @xmath774 note that the conditions ( which are equivalent to the coherence conditions , ) imply that for any @xmath775 $ ] , the endomorphisms @xmath776 are scalar multiples of the identity ; we will use this fact later in [ ss : estimate ] . [ def : operators ] for @xmath226 , @xmath777 , @xmath734 , a coherent collection @xmath635 of complex structures and a coherent pair of almost complex structures @xmath643 on @xmath778 , and @xmath779 , define the * linear delbar operator @xmath780 * to be the following map from @xmath781 to @xmath782 : @xmath783 where @xmath784 is the * pulled - back complex structure * @xmath785 if @xmath786 is a pair of sections in @xmath787 , we can write @xmath788 and @xmath789 in terms of @xmath790 : @xmath791 this decomposition allows us to relate the delbar operator @xmath792 from with the linear delbar operator @xmath793 just defined : @xmath794 the inhomogeneous term @xmath795 depends smoothly on @xmath796 , which is crucial for the proof of theorem [ thm : nonfoldedstripshrink ] . the following is the main result of [ ss : estimate ] . it generalizes ( * ? ? ? * lemma 3.2.1 ) , which bounds the @xmath797-norm of @xmath796 when the domain complex structure is standard . [ lem : masterestimate ] there is a constant @xmath201 and for every @xmath798 , @xmath799 , and @xmath800 with @xmath801 there is a constant @xmath497 such that the inequality @xmath802 holds for any choice of @xmath803 $ ] , a coherent collection @xmath635 of complex structures on @xmath650 with @xmath804 and @xmath805 , a coherent pair @xmath643 of almost complex structures on @xmath650 which are contained in a @xmath806-ball of radius @xmath807 and which induce by metrics whose pairwise constants of equivalence are bounded above by @xmath807 , and a pair of sections @xmath808 with @xmath809 , @xmath810 , and @xmath811 . we begin by establishing @xmath370-independent sobolev estimates for elements of @xmath743 . [ lem : improvedsobolev ] fix @xmath798 , @xmath799 , and @xmath800 with @xmath801 . then there is a constant @xmath497 and a polynomial @xmath812 such that the inequality @xmath813 ( where the term @xmath814 is to be omitted when @xmath815 ) holds for any choice of @xmath803 $ ] , @xmath816 $ ] , a coherent collection @xmath635 of complex structures on @xmath650 with @xmath817 , a coherent pair @xmath643 of compatible almost complex structures on @xmath818 which are contained in a @xmath28-ball of radius @xmath807 and which induce by metrics whose pairwise constants of equivalence are bounded above by @xmath807 , and pairs of sections @xmath819 with @xmath820 . here is the idea of the proof : ( * ? ? ? * lemma 3.1.4 ) is a uniform sobolev inequality for sections @xmath796 satisfying the linearized boundary conditions . since the special connection constructed in lemma [ lem : conn ] preserves the linearized boundary conditions , ( * ? ? ? * lemma 3.1.4 ) immediately gives a bound on @xmath821 . to derive a bound on @xmath822 for @xmath823 , we trade indices using the operator @xmath793 . we prove this lemma in two steps : first , we prove a slightly different inequality , which has terms of the form @xmath824 on the right - hand side . then , we prove the desired inequality by inductively removing these unwanted terms . throughout this proof , @xmath497 and @xmath812 will denote a @xmath370-independent constant and @xmath370-independent polynomial that may change from line to line . we prove the following inequality : @xmath825 we begin by proving the @xmath826 case of , which is essentially a consequence of ( * ? ? ? * lemma 3.1.4 ) . one modification must be made to that lemma : we must relax the hypothesis that the composition @xmath15 is embedded to the hypothesis that this composition is immersed . to make this modification , change the proof of ( * lemma 3.1.4 ) like so : instead of using ( * ? ? ? * lemma 3.1.3(c ) ) , use the fact that for @xmath827 , @xmath828 where @xmath829 is the projection onto the orthogonal complement of the tangent space of @xmath579 . this inequality follows from the pointwise estimate @xmath830 , which can be proved like ( * ? ? ? * lemma 3.1.3b ) . next , fix @xmath29 ; let us prove for this @xmath216 . let @xmath831 be sections in @xmath832 , and assume that the other hypotheses of the lemma are satisfied . we will show that for every tuple @xmath833 , there is a polynomial @xmath834 so that the following inequality holds : @xmath835 we prove this by induction on @xmath836 \ : | \ : \alpha_m = t \ } $ ] . if @xmath837 , then since the special connection we have constructed preserves the boundary conditions of @xmath838 , the desired inequality follows immediately from the @xmath815 case of the current lemma : @xmath839 . let us prove the inductive step for some @xmath840 $ ] . write @xmath841 . using the assumed bound on @xmath635 , we estimate : @xmath842 let us bound separately the five terms in the last expression . we estimate : @xmath843\nabla_s^{k - m - l-1}\zeta)\|_{{\mathcal{c}}^0h^1}.\end{aligned}\ ] ] let us bound each of the four terms on the right - hand side . the first term on the right - hand side , @xmath844 , is bounded by @xmath845 . due to the assumed bound on @xmath635 , the term @xmath846 is bounded by a constant times @xmath847 . to bound the term @xmath848 , observe that the assumed bound on @xmath643 yields : @xmath849 ( in the last inequality we have used the banach algebra property of @xmath850 . ) finally , the curvature of @xmath728 is a tensor , so the term @xmath851\nabla_s^{k - m - l-1}\zeta)\|_{{\mathcal{c}}^0h^1}$ ] can be bounded by a constant times @xmath852 . by the inductive hypothesis , this term is bounded appropriately : @xmath853 to bound this term , it suffices to bound @xmath854 and @xmath855 separately , where in the second term @xmath288 and @xmath291 are nonnegative integers with @xmath856 . the quantity @xmath854 can be bounded using the banach algebra property of @xmath850 , the assumed @xmath26-bounds on @xmath857 , and the inductive hypothesis . using the banach algebra property of @xmath850 , the quantity @xmath855 can be bounded by @xmath858 . this term is already bounded appropriately . by the banach algebra property of @xmath850 , this term is bounded by @xmath859 . this establishes the inductive step , so we have proven for all @xmath799 . we prove by induction on @xmath216 . as in step 1 , the @xmath826 case follows from ( * ? ? ? * lemma 3.1.4 ) . next , say that holds up to , but not including , some @xmath29 . by , we have : @xmath860 replacing the sum @xmath847 appearing in the last term using the inductive hypothesis finishes the inductive step . we now turn to the proof of lemma [ lem : masterestimate ] . here is our strategy : in lemma [ lem : k=0ellipticineq ] , we bound @xmath861 in terms of @xmath862 and @xmath863 , for @xmath796 supported in @xmath864 . in lemma [ lem : ellipticineq ] , we use lemma [ lem : k=0ellipticineq ] to bound @xmath865 in terms of @xmath866 and @xmath867 , where @xmath868 is supported in @xmath869 and @xmath796 has arbitrary support . finally , we use lemma [ lem : ellipticineq ] to prove lemma [ lem : masterestimate ] . [ lem : k=0ellipticineq ] there is a constant @xmath201 and for every @xmath798 , @xmath799 , and @xmath870 with @xmath871 there is a constant @xmath497 such that the inequality @xmath872 holds for any choice of @xmath803 $ ] , @xmath816 $ ] , a coherent collection @xmath635 of complex structures on @xmath650 with @xmath804 and @xmath873 , a coherent pair @xmath643 of almost complex structures on @xmath650 which are contained in a @xmath26-ball of radius @xmath807 and which induce by metrics whose pairwise constants of equivalence are bounded above by @xmath807 , and sections @xmath874 with @xmath875 , @xmath820 , and @xmath876 , @xmath877 compact subsets of @xmath878 , @xmath879 . throughout this proof , @xmath497 will denote a @xmath370-independent constant that may change from line to line , and @xmath769 , @xmath770 will be the endomorphisms of @xmath880 and @xmath881 defined in . we begin by fixing convenient metrics on @xmath298 and @xmath449 that will be used for the pointwise norms in the definition of the sobolev norms . via , @xmath643 induces fiberwise metrics @xmath882 on @xmath880 and @xmath881 . in this proof , however , we will use the pullback metrics @xmath883 of @xmath882 under @xmath884 , @xmath885 ; note that @xmath886 is @xmath887-invariant . if we pick @xmath201 to be sufficiently small , then @xmath888 is @xmath27-close to the identity , and hence the induced norms @xmath889 on @xmath890 are equivalent to the standard norms @xmath891 . with these metrics we calculate for @xmath892 compactly supported and @xmath893 satisfying @xmath894 and @xmath895 : @xmath896 let us estimate the two integrals on the right - hand side separately . we begin with the first integral : @xmath897 where the last inequality follows from the hypothesis @xmath898 as long as @xmath764 is chosen small enough . to bound the second integral on the right - hand side of , we first derive a convenient formula for its integrand : @xmath899\zeta ) - g_\xi(c\zeta,\bj(\xi)\nabla_s\nabla_t\xi ) \bigr ) \nonumber \\ & \hspace{0.1in}= \bigl ( \partial_s ( g_\xi(c\zeta,\bj(\xi)\nabla_t\zeta ) ) - \partial_t ( g_\xi(c\zeta,\bj(\xi)\nabla_s\zeta ) ) \bigr ) - \bigl ( ( \nabla_sg_\xi)(c\zeta , \bj(\xi)\nabla_t\zeta ) - ( \nabla_tg_\xi)(c\zeta , \bj(\xi)\nabla_s\zeta ) \bigr ) \nonumber \\ & \hspace{0.65 in } - \bigl(g_\xi((\nabla_sc)\zeta,\bj(\xi)\nabla_t\zeta ) - g_\xi((\nabla_tc)\zeta,\bj(\xi)\nabla_s\zeta)\bigr ) - g_\xi(c\zeta , \nabla_s(\bj(\xi))\nabla_t\zeta - \nabla_t(\bj(\xi))\nabla_s\zeta ) \nonumber \\ & \hspace{4.8 in } - g_\xi(c\zeta,\bj(\xi)[\nabla_s,\nabla_t]\zeta ) . \nonumber\end{aligned}\ ] ] we can now use green s formula and the assumed @xmath26-bounds on @xmath635 , @xmath643 , and @xmath857 to bound the second integral on the right - hand side of : @xmath900\zeta ) \ , { { \rm d}}s { { \rm d}}t \nonumber \\ & \hspace{0.3in}\geq -\tint_{\bq_{\delta , r } } c_1|\zeta|_\xi ( |\zeta|_\xi + |\nabla\zeta|_\xi ) \ , { { \rm d}}s { { \rm d}}t \nonumber { \stackrel}{\text{am - gm}}{\geq } -\tfrac 1 2\|\nabla \zeta\|_{\xi , h^0}^2 - c_1\|\zeta\|_{\xi , h^0}^2 , \nonumber\end{aligned}\ ] ] where in the first inequality we have eliminated the integrals over the @xmath506 and @xmath901 boundary via the coherence condition on @xmath635 and the fact that @xmath902 and @xmath903 vanish . indeed , @xmath903 vanishes by the lagrangian boundary condition : @xmath904 where we crucially used the fact that both the exponential map @xmath885 and the connection @xmath905 preserve @xmath906 . the boundary term @xmath902 vanishes due to the facts that @xmath907 preserves @xmath908 , @xmath728 satisfies @xmath909 for @xmath910 the projection , and @xmath911 satisfy @xmath912 : @xmath913 combining , , and yields the following inequality : @xmath914 adding @xmath915 to both sides of this inequality and taking the square root of the result , we obtain : @xmath916 in this estimate , we may replace @xmath917 with @xmath918 by using the @xmath370-independent uniform equivalence of these norms , which yields . [ lem : ellipticineq ] there is a constant @xmath201 and for every @xmath798 , @xmath799 , and @xmath871 there is a constant @xmath497 such that the inequality @xmath919 holds for any choice of @xmath803 $ ] , @xmath816 $ ] , a coherent collection @xmath635 of complex structures on @xmath650 with @xmath804 and @xmath805 , a pair @xmath643 of compatible almost complex structures on @xmath650 which are contained in a @xmath806-ball of radius @xmath807 and which induce by metrics whose pairwise constants of equivalence are bounded above by @xmath807 , a pair of sections @xmath920 with @xmath809 , @xmath810 , and @xmath921 , and a smooth function @xmath922 with @xmath923 and @xmath924 . throughout this proof , @xmath497 will denote a @xmath370-independent constant and @xmath812 will denote a @xmath370-independent polynomial , and both may change from line to line . we break down the proof into several steps : in step 1 , we establish , but with an extra term on the right - hand side . in step 2 , we bound this extra term , using different arguments in the @xmath925 and @xmath926 cases . in step 3 , we establish . we prove the following inequality : @xmath927 for @xmath928 . since the connection @xmath728 preserves the linearized boundary conditions and @xmath868 is supported in @xmath878 , we may estimate @xmath929 using lemma [ lem : k=0ellipticineq ] : @xmath930 \nabla_s^{k - l}\zeta \nonumber \\ & \hspace{3.35 in } - ( \partial_s\eta(a - \bj(\zeta ) ) + c\partial_t\eta)\nabla_s^k\zeta\bigr\|_{h^0 } \bigr ) \nonumber \\ & \leq c_1\bigl ( \|{\mathcal{d}}_\zeta\zeta\|_{h^k } + \|\zeta\|_{h^k } + \sum_{\beta\geq1,\gamma \geq 0 , \atop \beta + \gamma = k } \bigl\| \eta \nabla^\beta(\bj(\zeta))\nabla^\gamma\nabla_s\zeta \bigr\|_{h^0 } \bigr ) . \nonumber\end{aligned}\ ] ] we prove for a general multiindex @xmath931 of length @xmath216 . we establish step 1b by induction on @xmath932 \ : | \ : \alpha_m = t\}$ ] . step 1a is the base case for this induction . for the inductive step , fix @xmath931 with @xmath933 , and write @xmath934 . we estimate : @xmath935 \nabla_s^{k - m - l}\zeta \bigr ) \bigr\|_{h^1 } \\ & \hspace{2.15 in } + \|\eta\nabla_{\alpha'}\nabla_s^{k - m+1}\zeta\|_{h^1 } + \|\eta\nabla_{\alpha'}(\bj(\zeta)\nabla_s^{k - m+1}\zeta)\|_{h^1 } \bigr ) \nonumber \\ & \hspace{-0.5in}\leq c_1 \bigl ( \|{\mathcal{d}}_\zeta\zeta\|_{h^k } + \|\zeta\|_{h^k } + \sum_{\beta\geq1,\gamma\geq0 , \atop \beta+\gamma = k } \| \eta \nabla^\beta(\bj(\zeta ) ) \nabla^\gamma\nabla_s\zeta \|_{h^0 } \bigr),\end{aligned}\ ] ] where in the last inequality we have used the inductive hypothesis to bound @xmath936 . in the @xmath925 case , we prove the following inequality : @xmath937 it follows from the assumption @xmath925 that if @xmath938 satisfy @xmath939 , then @xmath940 . furthermore , the assumption @xmath941 implies the inequality @xmath942 by the embedding of @xmath744 for one - dimensional domains whose lengths are bounded away from zero . this , along with the assumed @xmath26-bound on @xmath796 , yields in the @xmath925 case . in the @xmath943 case , we prove the following inequality : @xmath944 the assumed @xmath26-bound on @xmath796 implies that the only term in the left - hand side of that is not immediately bounded by @xmath945 is @xmath946 . choose smooth maps @xmath947 so that the formula @xmath948 holds , where the maps @xmath949 preserve fibers but may not respect their linear structure . since @xmath643 is bounded in @xmath950 , @xmath949 must be bounded in @xmath26 . we may now use to bound the hat - part of @xmath951 : @xmath952{\widehat}\zeta \|_{h^0 } \bigr ) \nonumber \\ & \leq c_1 ( \|{\widehat}\zeta\|_{h^3 } + \delta^{1/2}\| s(\zeta)(\eta{\widehat}\nabla^2{\widehat}\zeta){\widehat}\nabla{\widehat}\zeta\|_{{\mathcal{c}}^0h^1 } ) \nonumber \\ & \leq c_1 ( \|{\widehat}\zeta\|_{h^3 } + \delta^{1/2}\| s(\zeta)(\eta{\widehat}\nabla^2{\widehat}\zeta)\|_{{\mathcal{c}}^0h^1}\|{\widehat}\nabla{\widehat}\zeta\|_{{\mathcal{c}}^0h^1 } ) , \nonumber\end{aligned}\ ] ] where in the last inequality we have used the @xmath370-independent banach algebra property of @xmath850 . by lemma [ lem : improvedsobolev ] , @xmath953 is bounded by @xmath954 and therefore by @xmath955 ; on the other hand , the @xmath26-bound on @xmath956 and the @xmath26-bound on @xmath796 implies the inequality @xmath957 . substituting these inequalities into , we obtain : @xmath958 next , we use lemma [ lem : improvedsobolev ] to bound @xmath959 : @xmath960 where the last inequality follows from the assumed bound on @xmath961 . substituting into , we obtain : @xmath962 to bound the @xmath963-part of @xmath964 , we use the the fact that the domains @xmath878 satisfy a uniform cone condition : @xmath965 where the second inequality follows from the sobolev embedding @xmath966 for two - dimensional domains satisfying a cone condition . combining and and using the assumed bound on @xmath961 yields the desired bound : @xmath967 we prove lemma [ lem : ellipticineq ] . the @xmath925 case of lemma [ lem : ellipticineq ] is an immediate consequence of steps 1b and 2a . toward the @xmath926 case of lemma [ lem : ellipticineq ] , let us show that there exists @xmath968 $ ] such that holds for @xmath969 $ ] . combining and yields the following inequality : @xmath970 if we set @xmath971 , where @xmath497 is the constant appearing in , then yields the uniform inequality @xmath972 for all @xmath969 $ ] . it remains to establish the @xmath943 case of for @xmath973 $ ] . to do so , we begin by bounding @xmath974 , using the fact that the domains @xmath778 satisfy a uniform cone condition for @xmath975 $ ] : @xmath976 substituting into yields the @xmath943 case of for @xmath975 $ ] : @xmath977 lemma [ lem : masterestimate ] follows immediately from lemmata [ lem : improvedsobolev ] and [ lem : ellipticineq ] . indeed , choose @xmath978 to be a smooth function with @xmath979 and @xmath980 . @xmath497 and @xmath812 will denote a @xmath370-independent constant and a @xmath370-independent polynomial that may change from line to line . lemma [ lem : ellipticineq ] yields a bound on @xmath981 : @xmath982 lemma [ lem : improvedsobolev ] yields a bound on @xmath983 : @xmath984 where in the second inequality we have used the assumed bound on @xmath985 . combining and yields @xmath986 , which can be used to inductively prove the desired inequality . we will not use the following proposition in this paper . however , it will be used in @xcite to show that the linearized cauchy riemann operator defines a fredholm section . [ prop : k=2linearineq ] there is a constant @xmath201 and for every @xmath798 , @xmath799 , and @xmath871 there is a constant @xmath497 such that the inequality @xmath987 holds for any choice of @xmath988 $ ] , a coherent collection @xmath635 of complex structures on @xmath650 with @xmath804 and @xmath989 , a pair @xmath643 of compatible almost complex structures on @xmath650 which are contained in a @xmath990-ball of radius @xmath807 and which induce by whose pairwise constants of equivalence are bounded above by @xmath807 , and two pairs of sections @xmath991 with @xmath875 and @xmath820 . the proof is an easier version of the proof of lemma [ lem : masterestimate ] . in this appendix , we sketch a proof of removal of singularity for a holomorphic curve satisfying a generalized lagrangian boundary condition in an immersed lagrangian with locally - clean self - intersection . we emphasize that this is not a new result , see e.g. @xcite . we have included the following proposition in this paper because our methods allow us to give a short proof . this removal of singularity will be stated for maps @xmath565 with lagrangian boundary conditions lifting to paths @xmath992 : @xmath993 where @xmath994 is a closed symplectic manifold , @xmath995 and @xmath996 are lagrangian immersions with @xmath997 closed , and @xmath319 is an almost complex structure @xmath998 . we will assume that @xmath999 intersect locally cleanly , which means that there are finite covers @xmath1000 , @xmath1001 such that @xmath1002 resp . @xmath1003 restrict to an embedding on each @xmath406 resp . @xmath1004 , and @xmath1005 intersect cleanly for all @xmath1006 . the first part of the proof of ( * ? ? ? * theorem 7.3.1 ) yields a uniform gradient bound on @xmath565 in cylindrical coordinates near the puncture . we must make a minor modification due to the fact that the lagrangians defining our boundary conditions are immersed , not embedded : recall that the uniform gradient bound in cylindrical coordinates is established in @xcite by assuming that there is a sequence @xmath1008 \times [ 0 , \tfrac 1 2]$ ] so that @xmath1009 , which necessarily has @xmath1010 . rescaling at the points @xmath1011 yields a sequence of maps that converges in @xmath106 to a nonconstant map on either @xmath481 or @xmath1012 , which contradicts the finiteness of the energy . to adapt this proof to our situation , let @xmath370 be a lebesgue number for @xmath1013 and @xmath1001 . that is , if @xmath1014 is a subset of @xmath1015 ( resp . of @xmath1016 ) with @xmath1017 , then @xmath1018 ( resp . @xmath1019 ) for some @xmath1020 . now rescale at the points @xmath1011 as in @xcite , but restrict the resulting maps to the intersection of @xmath1021 with their domain . the gradient bound on these rescaled maps and our choice of @xmath370 allows us to pass to a subsequence so that for some @xmath1006 , all the rescaled maps have boundary values in @xmath1022 or @xmath1023 . a further subsequence converges in @xmath106 , so we get a contradiction and therefore a uniform bound on @xmath1024 in cylindrical coordinates . the analogue of lemma [ lem : lengthsgotozero ] holds in this setting ; the proof is the same as for lemma [ lem : lengthsgotozero ] but simpler . as in the first paragraph , some care must be taken with the immersed lagrangians . the analogue of lemma [ lem : isoineq ] holds in this setting , though the proof must be modified . specifically , the domains @xmath1025 used in the proof of that lemma must be replaced by the domain @xmath1026 .

we show that the novel figure eight singularity in a pseudoholomorphic quilt can be continuously removed when composition of lagrangian correspondences is cleanly immersed . the proof of this result requires a collection of width - independent elliptic estimates that allow for nonstandard complex structures on the domain .

it is theoretically argued that the heavy mass of the @xmath0 meson , which is around @xmath3 , originates from the interplay between the @xmath1 anomaly in quantum chromodynamics and the spontaneous breaking of chiral symmetry @xcite . furthermore , when chiral symmetry is partially restored at finite density , the mass is expected to be reduced . for example , @xmath4 mass reduction at the normal nuclear density is estimated in the nambu jona - lasinio model calculation @xcite , and a recent calculation using the linear sigma model predicts @xmath5 reduction @xcite . in the qmc ( quark - meson coupling ) model , a rather small mass reduction of @xmath6 is predicted for the case of an @xmath7-@xmath0 mixing angle of @xmath8 @xcite . the mass reduction at finite density is , in other words , the outcome of an attractive interaction between an @xmath0 meson and a nucleus . if the attraction is strong enough , an @xmath0-nucleus bound state may be formed . for the experimental observation of such a bound state , the decay width is one of the important ingredients . the chiral unitary model including a lagrangian term which couples the singlet meson to the baryons indicates that the real part of the optical potential is in general deeper than the imaginary part , while the coupling constant is an unknown parameter @xcite . from the experimental side , two kinds of measurements , @xmath0 photoproduction off nuclei @xcite and @xmath0 production in proton - proton collision @xcite , have been used to extract the information on @xmath0-nucleus interaction and @xmath0-nucleon interaction , respectively , by taking different approaches . the cbelsa / taps experiment determined both the real part @xcite and the imaginary part @xcite of the @xmath0-nucleus optical potential . the imaginary part , which is half of the absorption width , was obtained to be @xmath9 through a measurement of the transparency ratios for different nuclear targets @xcite . in addition , by comparing the excitation function and the momentum distribution of @xmath0 mesons for the photoproduction off carbon with corresponding theoretical calculations , the real part was derived as @xmath10 @xcite . through these studies , they concluded a possible existence of @xmath0 mesic nuclei @xcite . the near - threshold behavior of the cross section of the @xmath11 reaction supplies information on the @xmath0-nucleon interaction . the cosy-11 experiment determined the @xmath0-proton scattering length in free space , as @xmath12 . it indicates a shallow depth of the @xmath0-nucleus optical potential if one assumes no energy and density dependence @xcite . the investigation of @xmath0 mesic nuclei will be an alternative way to extract direct information of the @xmath0-nucleus interaction . we plan a series of missing - mass spectroscopy experiments with the @xmath13 reaction @xcite at gsi and fair . as the first step , we performed an inclusive measurement at gsi . moreover , a semi - exclusive measurement , with the coincidence of decay particles from @xmath0 mesic nuclei , will be studied in the future at fair . @xmath0-mesic nuclei will be produced by impinging a @xmath14 proton beam onto a @xmath15-thick carbon target . the incident energy is slightly above the @xmath0 production threshold in the @xmath16 reaction . the missing - mass is obtained by analyzing the momentum of the ejectile deuterons with the fragment separator ( frs ) used as a high - resolution spectrometer . the missing - mass resolution is estimated to be less than @xmath17 in @xmath18 , which is dominated by the uncertainty of the reaction vertex inside the target . it would be sufficiently smaller than the decay width of @xmath0 mesic nuclei , and this feature is one of the advantages of our experiment . figure [ missingmass ] shows a simulated spectrum corresponding to 4.5 days of data acquisition for different @xmath0-nucleus optical potentials parameterized as @xmath19 @xcite . in the simulation , the cross section in the @xmath13 reaction is assumed to be the sum of a theoretical calculation @xcite , in which @xmath0 production is involved , and contributions of background processes of multi - pion production , evaluated from past measurements of proton - nucleon cross sections @xcite . the experimental sensitivity of finding peak structures above a huge background will be large enough in case of a large @xmath20 and a small @xmath21 . while the signal - to - noise ratio will be of the order of 1/100 even in such a case , a high - statistics measurement with an intense primary beam will compensate the poor signal - to - noise ratio . the first experiment ( gsi s-437 ) was carried out at gsi in august 2014 . figure [ frs_setup ] shows the experimental setup at frs . two sets of multi - wire drift chambers ( mwdc s ) were installed close to a dispersive focal plane at s4 , and the momentum of ejectile deuterons could be analyzed by measuring their tracks with them . for particle discrimination between deuterons and protons from inelastic scattering , we measured the time - of - flight between s2 and s4 by plastic scintillators labelled with sc2h , sc2v , sc41 , and sc42 . in addition , we prepared erenkov detectors with high refractive - index ( @xmath22 ) silica aerogel radiators , which were developed at chiba university @xcite , ( hirac and mini - hirac ) and a total - reflection erenkov detector with an acrylite radiator ( torch ) . it is worth emphasizing that we could make the main trigger for the @xmath13 reaction with a tight coincidence of the signals of the scintillators in s2 and s4 , and without an anticoincidence of these erenkov detectors . this removes the influence of a possible position - dependence of the overkill rate of each erenkov detector on the missing - mass spectrum . figure [ tof2h41 ] shows the tof distribution between s2 and s4 obtained by an unbiased trigger . the two peaks are @xmath23 distant from each other , which is consistent with the tof difference between protons and deuterons within the momentum acceptance . the deuteron - to - proton ratio is found to be approximately 1/200 . later on , we could achieve a deuteron - to - proton ratio around unity by tuning the coincidence timing in order to remove the protons on the trigger level . the remaining background , which is seen beneath the deuteron peak in fig . [ tof2h41 ] , is due to sequential protons with a short ( @xmath23 ) time interval . the waveforms of the scintillator signals will serve to distinguish one - pulse events from two - pulse events . we have been working to improve the purity without losing the efficiency in particle identification . we have taken data for the following reactions with changing the frs magnetic field to cover a wide range of excitation energies between @xmath24 and @xmath25 relative to the @xmath0 emission threshold . * @xmath26 reaction at @xmath14 incident energy * @xmath27 reaction at @xmath14 incident energy ( for background study ) * @xmath28 reaction at @xmath29 incident energy ( for calibration ) the analysis of the mwdc s is in progress . for example , the overall missing - mass resolution for the @xmath26 reaction has been estimated to be @xmath30 ( preliminary ) from the missing - mass resolution of the proton peak in the calibration process . while we will proceed with an inclusive measurement with higher statistics in a future experiment , another way is to make a coincidence of the forward going deuteron with decay particles from the decay of @xmath0 mesic nuclei ( semi - exclusive measurement ) . we aim to tag a proton from the two - nucleon absorption process ( @xmath31 ) , which is one of the major decay modes of @xmath0 mesic nuclei @xcite . this proton has a kinetic energy around @xmath32 . such a fast proton will be hardly emitted from the main background of multi - pion production processes , even if rescattering takes place , and the coincidence is expected to improve the signal - to - noise ratio significantly . more quantitative evaluation of the proton distribution from the background processes as well as the decay of @xmath0 mesic nuclei , by an intra - nuclear cascade simulation using a microscopic transport model jam @xcite , is under way . we plan to develop a proton counter to be installed surrounding the target , by which protons can be distinguished from charged pions . this experimental setup will be possible at super - frs , which is to be built in the fair complex . we are investigating @xmath0 mesic nuclei by means of missing - mass spectroscopy with the @xmath13 reaction . an inclusive measurement of the @xmath13 reaction on @xmath33 at frs / gsi was carried out in august 2014 . the expected resolution will be around @xmath17 , which is much smaller than the width of @xmath0 mesic nuclei . the analysis is under way . furthermore , we plan to perform a semi - exclusive measurement at super - frs / fair , with tagging decay protons from the two - nucleon absorption process in @xmath0 mesic nuclei . the signal - to - noise ratio , which is of the order of 1/100 at most in the inclusive spectrum , will be improved significantly . a detailed monte carlo simulation with a microscopic transport model jam is in progress . the experiment was performed in the framework of the super - frs collaboration for fair . this work is partly supported by a grant - in - aid for scientific research on innovative areas ( no . 24105705 ) from the ministry of education , culture , sports , science and technology(mext ) , japan , and a grant - in - aid for young scientists ( a ) ( no . 25707018 ) from japan society for the promotion of science ( jsps ) . h. nagahiro , m. takizawa , and s. hirenzaki , phys . c * 74 * , 045203 ( 2006 ) . s. sakai and d. jido , phys . c * 88 * , 064906 ( 2013 ) . s.d . bass and a.w . thomas , phys . b * 634 * , 368 ( 2006 ) . h. nagahiro , s. hirenzaki , e. oset , and a. ramos , phys . lett . * b709 * , 87 ( 2012 ) . v. metag , in these proceedings . p. moskal , in these proceedings . m. nanova _ et al . b * 727 * , 417 ( 2013 ) . m. nanova _ et al . b * 710 * , 600 ( 2012 ) . e. czerwiski _ et al . _ , * 113 * , 062004 ( 2014 ) . k. itahashi et al . , letter of intent for gsi - sis ( 2011 ) ; k. itahashi , h. fujioka _ et al . . phys . * 128 * , 601 ( 2012 ) . h. nagahiro , d. jido , h. fujioka , k. itahashi , and s. hirenzaki , phys . c * 87 * , 045201 ( 2013 ) . h. nagahiro , in these proceedings . m. tabata and h. kawai , arxiv:1410.2439 [ physics.ins-det ] y. nara , n. otuka , a. ohnishi , k. niita , and s. chiba , phys . rev . c * 61 * , 024901 ( 2000 ) .

the possible existence of @xmath0-nucleus bound states has been put forward through theoretical and experimental studies . it is strongly related to the @xmath0 mass at finite density , which is expected to be reduced because of the interplay between the @xmath1 anomaly and partial restoration of chiral symmetry . the investigation of the @xmath2 reaction at gsi and fair , as well as an overview of the experimental program at gsi and future plans at fair are discussed . example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore

the cross section for forward jet production in deep inelastic scattering ( dis ) has been proposed as a particularly sensitive means to investigate the parton dynamics at small @xmath2 @xcite . analytic calculations based on the bfkl equation @xcite in the leading - logarithmic approximation show a strong rise of this cross section with decreasing @xmath2 @xcite and were found in reasonable agreement @xcite with the first data from the h1 collaboration at hera @xcite . more recent measurements of the forward cross section , based on an order of magnitude increased statistics compared to @xcite , have been presented recently by the zeus @xcite and the h1 @xcite collaborations confirming the earlier findings @xcite . monte carlo generators based on direct photon interactions ( dir ) calculated from leading order ( lo ) @xmath3 matrix elements together with leading - logarithm parton showers disagree with the measured jet cross section @xcite by an appreciable factor . also next - to - leading order ( nlo ) , i.e. @xmath4 , calculations predict too small forward cross sections at small @xmath2 as already shown by mirkes and zeppenfeld @xcite using their mepjet program @xcite when comparing to the data in @xcite . this has been confirmed also with the new data in @xcite . a similar deficiency between nlo calculations and measured data occurs for the dijet rate in the region @xmath5 @xcite , where @xmath6 is the transverse energy of the produced jets and @xmath7 is the usual squared lepton momentum transfer . this kinematic range is also relevant for the forward jet production , as will be seen later . the region of small enough @xmath7 is the photoproduction regime where the virtual photon resolves into partons . indeed , introducing a resolved photon contribution , the measured dijet rate and the forward jet cross section can be described satisfactorily @xcite concerning the shape of the cross section as a function of @xmath2 as well as the absolute normalization . this description is based on the monte carlo program rapgap @xcite which includes a resolved photon contribution in addition to the direct process , which both are evaluated with lo matrix elements with additional emissions in the initial and final state generated by parton showers together with subsequent hadronization . the dijet rate has been calculated also in nlo including direct and resolved photon contributions @xcite . in order to avoid double counting in the full nlo calculation , the contribution from the virtual photon splitting into @xmath8 pairs , where either the quark or the antiquark subsequently interacts with a parton originating from the proton , had to be subtracted @xcite , similar as is done in the nlo theory for the photoproduction of jets @xcite . the subtracted terms in the nlo direct contribution are part of the parton distribution functions ( pdf s ) of the virtual photon and appear in the resolved contribution in an evolved form . with this procedure the whole cross section for two - jet production , which is a superposition of the direct and resolved contributions minus the photon splitting piece , becomes to a large extent independent of the factorization scale at the photon vertex . this full nlo calculation of the dijet rate agreed well with the h1 data over the full @xmath7 domain , @xmath9 gev@xmath10 , and the @xmath2 domain , @xmath11 , and for jet transverse momenta @xmath12 @xcite . in this work we want to present the results of a calculation of the forward jet cross section on the basis of the nlo theory used for the dijet rate . although the kinematic constraints , very low @xmath2 and @xmath13 of order one , are rather similar , it is not obvious that the calculated cross sections will agree with the recent zeus @xcite and h1 @xcite experimental results . after some comparisons with the mepjet results @xcite to make sure that our dis jet program , called jetvip @xcite , gives the same results under identical kinematical conditions we shall give our results with the experimental cuts of the zeus @xcite and h1 @xcite analysis . we close with a short summary and an outlook to future studies . before we present our results with the zeus and h1 kinematical constraints for selecting the forward jets we performed a check of our program jetvip with the forward jet kinematics by comparing with the nlo results of mirkes and zeppenfeld @xcite , who have produced their results with the fixed order program mepjet @xcite , which only includes direct photon contributions . we have chosen the same kinematical cuts , which differ somewhat from the cuts used in the zeus @xcite and h1 @xcite analyses . the @xmath3 results are obtained taking the glck , reya and vogt ( grv ) lo proton pdf s @xcite together with the one - loop formula for @xmath14 . for the @xmath4 results we employ the grv higher order pdf s together with the two - loop running @xmath14 formula . we take @xmath15 and match the strong coupling at the charm and bottom thresholds @xmath16 , respectively . jets are defined in the laboratory frame using the cone algorithm with the opening angle @xmath17 in the so - called e - scheme . in this scheme the four - vector of the combined jet is given as the sum of the four - vectors of the two partons . the differences of pseudorapidities and azimuthal angles with respect to the jet direction are @xmath18 and @xmath19 . all jets have to fulfill @xmath20 and @xmath21 gev , where the index b refers to quantities in the breit frame . @xmath22 and @xmath6 are measured in the hera laboratory frame . additional cuts are made for events which contain a forward jet . these requirements are @xmath23 and @xmath24 gev with @xmath25 the @xmath2 variable is restricted to the small-@xmath2 region of @xmath26 . the cuts on the electron variables are @xmath27 gev@xmath10 , @xmath28 , @xmath29 gev and @xmath30 $ ] . the electron and proton energies are @xmath31 gev and @xmath32 gev , respectively . the positive @xmath33-direction is the direction of the incoming proton momentum . the renormalization ( @xmath34 ) and factorization scales ( @xmath35 ) are taken equal and are identified with the sum @xmath36 where @xmath37 and @xmath38 , the parton s transverse momentum , are related in the breit frame by @xmath39 ^ 2 = 2e_i^2(1-\cos\theta_{ip})=\frac{2}{1+\cos\theta_{ip } } [ p_t^b(i)]^2 , \ ] ] where @xmath40 is the angle between the parton and the proton direction in the breit system . in lo one - jet production , i.e. , in the naive parton model limit , @xmath41 . with these constraints we obtain the cross sections in tab . 1 , where our jetvip results are compared to results from @xcite , referenced as mepjet in the table . in addition to the forward jet cross sections we also list the full 2-jet , inclusive 2-jet and exclusive 3-jet cross sections . compared to these cross sections , the kinematic constraints defining the forward cross section lead to considerable reductions . furthermore we notice that the nlo corrections to the forward jet cross sections are large in agreement with @xcite . the o(@xmath42 ) single - jet cross section is not considered since it vanishes if a forward jet is required , due to the kinematical restrictions of the phase space . .jet cross sections in the forward region compared to mepjet results . [ jv vs mep ] [ cols="<,^,^,^",options="header " , ] first , the lo resolved cross section is @xmath43% larger than the subtracted photon splitting term . part of this increase is due to the evolution of the pdf s of the photon . the other part comes from the gluon component of the photon pdf , which amounts to 7.3 pb . second , the nlo corrections to the resolved cross section give a further increase as compared to the lo result of approximately @xmath44 , which originates from the nlo corrections to the resolved matrix elements . we conclude , that the nlo resolved contribution supplies higher order terms in two ways , first through the nlo corrections in the hard scattering cross section and second in the leading logarithmic approximation by evolving the pdf s of the virtual photon to the chosen factorization scale . this way we sum the logarithms in @xmath13 , which , however , in the considered kinematical region is not an important effect numerically , as we have seen in tab . 2 . therefore , the enhancement of the nlo direct cross section through inclusion of resolved processes in nlo is mainly due to the convolution of the point - like term in the photon pdf with the nlo resolved matrix elements , which gives an approximation to the nnlo direct cross section without resolved contributions . one of the dominant contributions to the forward cross section is shown in fig . 1 ( left part ) , where the photon splitting term is convoluted with a matrix element that provides two gluons in the final state . this way one gluon rung is added to the gluon ladder as compared to the corresponding nlo direct cross section , shown on the right of fig . 1 . the additional gluon in the nlo resolved term is in our approach calculated from perturbative qcd , producing an additional term which makes a contribution to the forward jet . in the bfkl approach for the forward jet cross section @xcite this extra gluon is part of the bfkl evolution . in contrast , the nlo direct term in fig . 1 ( right part ) contains also additional gluons in the dglap evolution of the proton pdf , which , however , are not resolved , i.e. , go to the proton remnant . another way to generate a larger forward jet cross section would be to go to the nnlo corrections of the direct cross section , which has not been done yet . we think that including the nlo resolved component produces a reasonable approximation to this nnlo cross section . such a correspondence is present at one order lower . as already remarked above , the superposition of the lo direct and resolved cross section is almost equal to the nlo direct cross section . ( 161,53 ) ( 10,-11 ) ( 80,-11 ) ( 0,6 ) for the comparison with the 1995 zeus @xcite and the 1994 h1 @xcite forward jet cross section data we calculated the nlo cross section with slightly different input and in particular with the exact kinematical constraints for the forward jet selection as used in the two experiments . as proton pdf s we apply now the cteq4 m parametrization @xcite with the two - loop @xmath14 . we take @xmath15 as before and match the value of @xmath14 at the thresholds @xmath16 with a @xmath45 as used in cteq4 m . jets are defined with the cone algorithm in the hera frame as described above , except that the axis of the jet is calculated now as the transverse energy weighted mean of @xmath22 and @xmath46 of the two partons or jets belonging to the combined jet . this kind of jet definition was also applied in the experimental jet analysis . as scales we choose @xmath47 with a fixed @xmath48 gev@xmath10 related to the mean @xmath49 of the forward jet . we take this fixed value of @xmath50 instead of @xmath6 for technical reasons , since the calculations in jetvip start from the hadronic c.m.s .. the choice @xmath51 is mandatory if we want to include a resolved contribution . another choice of scale would be @xmath52 or @xmath53 . in the case @xmath54 only for @xmath55 , which covers only part of the zeus kinematical range . to have a resolved cross section in all @xmath13 bins we consider the choice @xmath56 more appropriate . in the two experiments the forward jet selection criteria are different . in the zeus experiment the kinematical constraints are : @xmath57 gev , @xmath28 , @xmath58 and @xmath24 gev with the forward jet constraints @xmath59 , @xmath60 , @xmath61 and @xmath62 . our results for the forward jet cross section under these zeus kinematical conditions are shown in fig . 2 a , b . in fig . 2 a we plotted the full o(@xmath63 ) inclusive two - jet cross section ( dir ) as a function of @xmath2 for three different scales @xmath64 and @xmath65 and compared them with the measured points from zeus @xcite . as to be expected the calculated nlo direct cross section is by a factor 2 to 4 too small compared to the data . the variation inside the assumed range of scales is small , so that also with a reasonable change of scales we can not get agreement with the data . in fig . 2 b we show the corresponding forward jet cross sections with the nlo resolved contribution included , as described in the previous subsection , again for the three different scales @xmath66 as in fig . 2 a. now we find good agreement with the zeus data . the scale variation of the calculated cross section is larger than in fig . 2 a. in particular , the largest scale gives now the largest cross section opposite to what is observed in fig . 2 a. this different scale variation comes primarily from the scale dependence of the virtual photon pdf . this factorization scale variation is supposed to be compensated between the lo resolved and the nlo direct contribution but not for the nlo resolved contribution . this could only occur if we could include the nnlo direct contribution which , however , is not available . since the nlo resolved contribution for the forward jet is rather large , as discussed above , this scale dependence can not be avoided . on the other hand , the scale dependence is not so large that we must fear our results not to be trustworthy . in fig . 2 b the cross section is labeled dir@xmath67+res , where dir@xmath67 stands for nlo direct minus the photon - quark - antiquark splitting term . we have calculated the forward jet cross section also with the scale @xmath68 . for this choice we obtain a somewhat smaller cross section which is approximately equal to the cross section with the scale @xmath69 . ( 122,70 ) ( -6,-60 ) ( 80,-60 ) ( 0,0 ) ( 122,140 ) ( -6,10 ) ( 80,10 ) ( -6,-50 ) ( 80,-50 ) ( 0,10 ) in the h1 experiment @xcite the forward jets are selected with the kinematical cuts @xmath70 gev , @xmath28 , @xmath71 , @xmath72 ( this corresponds to @xmath73 ) , @xmath74 gev , @xmath75 and @xmath76 . this corresponds approximately to the @xmath7 range @xmath77 gev@xmath10 . the forward jet cross section is measured for various @xmath2 bins ranging from @xmath78 to @xmath79 . otherwise the calculated forward cross section is obtained under the same assumptions as for the zeus selection cuts . in fig . 3 a , b , c , d we show the results compared to the h1 data obtained with two @xmath6 cuts in the hera system , @xmath80 gev ( fig . 3 a , b ) and @xmath81 gev ( fig . 3 c , d ) . in the plots on the left ( fig . 3 a , c ) the data are compared with the pure nlo direct prediction , which turns out to be too small by a similar factor as observed in the comparison with the zeus data . in fig . 3 b , c the forward jet cross section is plotted with the nlo resolved contribution included in the way described above . for both @xmath6 cuts , @xmath80 gev ( in fig . 3b ) and @xmath81 gev ( fig . 3d ) , we find good agreement with the 1994 h1 data inside the scale variation window @xmath82 . compared to the zeus data the h1 measurements extend down to smaller @xmath2 . in the lowest @xmath2 bin , @xmath83 the forward jet cross section has a dip , which is also reproduced in the theoretical calculation and which is due to the kinematical constraints for selecting forward jets . we conclude that the nlo theory with a resolved virtual photon contribution gives a good description of both , the zeus and the h1 forward jet data . it is important , that both components , the direct and the resolved one , are calculated up to nlo . a lo calculation of both components would fall short of the experimental data , as it is clear from the results presented in tab . 2 , where we compared lo and nlo results . we remark that the forward jet cross sections as measured by zeus and h1 are obtained at the hadron level , i.e. , the jets are constructed from measured hadron momenta using the same cone algorithm . in our nlo calculation the jets are combined from partons with the same jet algorithm . the size of the corrections from hadron to parton level has been studied by the zeus collaboration @xcite using several monte carlo simulation programs with the result that for the models which account very well for the zeus forward jet cross sections the correction factors are close to unity for all @xmath2 values considered in the analysis . in the h1 work similar results are reported @xcite . in @xcite the h1 forward jet data are also successfully described with the rapgap model @xcite which includes a direct and a resolved component , both in lo , and with a similar scale @xmath66 as we have used . in addition this model has leading logarithm parton showers in the initial and final state built in . we think that these parton shower contributions produce the higher order effects which we found necessary to account for the correct normalization and the @xmath2 dependence of the forward jet cross section data . we conclude that the measurements of the forward jet cross section presented recently by the zeus and h1 collaborations can be described very well by the nlo theory with direct and resolved virtual photon contributions added in a consistent way . the theory shows good agreement with the data with respect to the normalization and also the functional dependence with decreasing @xmath2 . whereas this variation of the cross section with @xmath2 is also compatible with the nlo predictions based on direct photons , for the correct normalization the resolved component up to nlo is needed . to avoid double counting the @xmath84 splitting term is removed from the nlo direct contribution . in contrast , lo bfkl predictions @xcite yield much larger forward cross sections than the data @xcite . these calculations suffer , however , from several deficiencies . they are asymptotic and do not contain the correct kinematic constraints of the produced jets @xcite . furthermore they do not allow the implementation of a jet algorithm as used in the experimental analysis . also nlo @xmath85 terms in the bfkl kernel @xcite predict large negative corrections which are expected to reduce the forward cross section as well . when all these points are taken into account the bfkl approach may give an equally well description of the forward jet data . even if this is the case , it is clear from this work that the bfkl theory is not the only theoretical approach that describes the forward jet cross sections .

the production of forward jets of transverse energy @xmath0 and large momentum fraction @xmath1 is calculated in next - to - leading order including consistently direct and resolved virtual photon contributions . the predictions are compared to recent zeus and h1 data . good agreement with the data is found .

if supersymmetry ( susy ) as the most attractive extension of the standard model is realized at low energies , the next generation of high energy physics experiments at tevatron , lhc and a future @xmath3 linear collider will discover supersymmetric particles . particularly at a linear collider , it will be possible to perform measurements with high precision @xcite which allows to test the underlying susy model . for instance , at tesla @xcite the precision of the mass determination of charginos and neutralinos , the supersymmetric partners of the gauge and higgs bosons , will be @xmath4 gev . to match this accuracy it is indispensable to include higher order radiative corrections . + one goal of all analyses based on precise measurements of cross sections , decay branching ratios , masses of supersymmetric particles , etc . will be the reconstruction of the fundamental parameters of the underlying supersymmetric model . in particular , this is needed for extrapolating the parameters to the gut point to check the unification of the supersymmetry breaking parameters @xcite . + in the minimal supersymmetric standard model ( mssm ) the chargino and neutralino system depends on the parameters @xmath0 , @xmath1 , @xmath2 and @xmath5 . @xmath0 and @xmath1 are the su(2 ) and u(1 ) gauge mass parameter , @xmath2 is the higgsino mass parameter and @xmath6 with @xmath7 the vacuum expectation values of the two neutral higgs doublet fields . at lowest order , it was shown in @xcite that these parameters can be extracted from the masses and production cross sections in @xmath3 collisions with polarized electron beams . at higher order , this extraction of the parameters is , however , not trivial . it depends on the definition of the mass matrices ( at higher order ) and on the renormalization scheme . this is the subject of this paper . + in the ( scale dependent ) @xmath8 scheme the one loop corrections to the chargino / neutralino mass matrix were calculated in @xcite . in @xcite effective chargino mixing matrices were introduced , which are independent of the renormalization scale . for the on shell renormalization of the chargino and neutralino system which we adopt here , two methods were proposed @xcite . they differ by different counter terms to the parameters @xmath1 , @xmath0 and @xmath2 . although both schemes are equivalent in the sense that the observables ( masses , cross sections , branching ratios , etc . ) are the same , the meaning of the parameters @xmath0 , @xmath1 , @xmath2 extracted are different . in the following we want to analyze in detail the determination of the parameters of the chargino / neutralino mass matrices at one loop level in the scheme @xcite . we will point out that at one loop level the values of the on shell parameters @xmath0 and @xmath2 depend on whether they are determined from the chargino or neutralino system . another interesting issue is how the gut relation @xmath9 ( @xmath10 in su(5 ) , @xmath11 in amsb ) valid in the @xmath8 scheme can be tested if the on shell values of @xmath1 and @xmath0 are extracted from experiment . the one loop corrections will also change the gaugino and higgsino nature of the charginos and neutralinos , in particular also of the lightest neutralino . this is important for the dark matter search @xcite . the whole analysis is based on a full one loop calculation within the mssm . the corresponding formulae are given in the appendix . in the cp conserving mssm the chargino mass matrix has at tree level the form @xmath12 where we take @xmath0 , @xmath2 , @xmath13 and @xmath5 as on shell parameters . @xmath0 and @xmath2 are taken real . with the real rotation matrices @xmath14 and @xmath15 @xmath16 it can be diagonalized , @xmath17 with @xmath18 and the tree level masses @xmath19 . the solutions of eq . ( [ eq : chi+rot ] ) are @xmath20 and @xmath21 as shown in @xcite , the on shell mass matrix @xmath22 at one loop level can be written as a sum of the tree level mass matrix @xmath23 in terms of the on shell parameters as in ( [ eq : chi+mat ] ) and the ultraviolet finite shifts @xmath24 @xmath25 this implies corrections in the mass eigenvalues , @xmath26 , and in the rotation angles of the coupling matrices , @xmath27 and @xmath28 . + in the neutralino sector , we have the symmetric tree level mass matrix @xmath29 using the real matrix @xmath30 , we can rotate from the gauge eigenstate basis of the neutral gauginos and higgsinos @xmath31 to the mass eigenstate basis of the neutralinos @xmath32 , @xmath33 taking the one loop terms into account @xmath34 we again obtain corrections in the masses , @xmath35 , and in the coupling matrix @xmath36 . in supersymmetry one has several mass matrices due to the mixing of interaction states . we define the on shell mass matrix such that all elements which are non zero at tree level have formally the tree level form but give the physical masses and rotation matrices . we always start with a certain set of on - shell input parameters . for these we need fixing conditions . all other on - shell entries in the mass matrices can then be calculated . + the standard model input parameters are the pole masses @xmath37 gev and @xmath38 gev . the weinberg angle @xmath39 is fixed by @xmath40 @xcite . the susy parameter @xmath5 is fixed by the condition that there is no transition from the physical cp odd higgs particle @xmath41 to the vector boson @xmath42 @xcite , @xmath43 which gives the counter term @xmath44 . @xmath45 is the renormalized self energy for the mixing of @xmath41 and @xmath42 . in this study , the physical input for calculating our input on - shell parameters @xmath0 , @xmath2 , and @xmath1 are the two chargino masses and one neutralino mass . for the other susy parameters we use the simplifications @xmath46 for the trilinear couplings and @xmath47 , @xmath48 for the soft breaking sfermion mass parameters . + in the @xmath8 scheme ( at a scale @xmath49 ) the parameters @xmath50 and @xmath51 are the same in the chargino and neutralino sector . however , the on shell parameters @xmath0 and @xmath2 get different one loop corrections and thus have different on shell values due to different thresholds : @xmath52 @xmath53 , @xmath54 are the counter terms to the elements @xmath55 , @xmath56 of the chargino and neutralino mass matrix . the corresponding expressions are given in the appendix . the finite difference can be expressed in terms of the chargino and neutralino mass matrix counter terms @xmath57 therefore , we have the freedom to define the input on shell parameters @xmath0 and @xmath2 in the chargino sector , i.e. @xmath58 , @xmath59 , and obtain corrections in the neutralino sector , or fix @xmath0 and @xmath2 in the neutralino sector , i.e. @xmath60 , @xmath61 , and get corrections in the chargino mass matrix . for a particular physical situation the elements of the one loop mass matrices @xmath22 and @xmath62 ( with on shell parameters plus corrections ) are given by the measured neutralino , chargino masses and other observables , e.g. cross sections . + if @xmath0 and @xmath1 are independent parameters it is convenient to use for the on - shell @xmath1 the definition @xmath63 . if the su(5 ) gut relation , @xmath64 , holds for the @xmath8 parameters @xmath50 and @xmath65 , we obtain a finite shift for the on shell parameters . thus we can write @xmath66 , with @xmath67 the correction @xmath68 is due to the same effect and of the same order as @xmath69 and @xmath70 , eq.([dm ] ) and ( [ dmu ] ) . therefore we include it in our calculations in the cases where gauge unification is explicitly assumed . because @xmath0 depends on the fixing this is also the case for @xmath68 . let @xmath71 be the correction in the case , where @xmath0 is fixed in the chargino sector , and @xmath72 the case , where @xmath0 is fixed in the neutralino sector , it follows @xmath73 in fig . [ fig : fixing1 ] the mass corrections for the lightest neutralino and chargino assuming gauge unification are shown as a function of @xmath0 . if @xmath0 and @xmath2 are fixed in the neutralino sector ( dashed lines ) we have @xmath74 , @xmath75 , @xmath76 , @xmath77 and @xmath78 . if @xmath0 and @xmath2 are fixed in the chargino sector ( full lines ) we get @xmath79 , @xmath80 , @xmath81 , @xmath82 and @xmath83 . the differences between the full and the dashed line are due to @xmath69 and @xmath84 , eqs . ( [ dm ] ) , ( [ dmu ] ) and ( [ dmp ] ) . in fig . [ fig : fixing2 ] the corrections @xmath69 and @xmath70 are given as a function of @xmath0 and @xmath2 . for @xmath85 the corrections are in the range of @xmath86 gev ( white ) and @xmath87 gev ( black ) . the corrections @xmath70 are between @xmath88 gev ( white ) and @xmath89 gev ( dark grey ) . the difference between two lines are @xmath90 gev . with the one loop corrections to the rotation matrices @xmath14 , @xmath15 and @xmath30 , the gaugino and higgsino characters of the individual chargino and neutralino states change . this can have large effects on decay widths of processes where these particles are involved @xcite . the character of the lsp neutralino plays a key role in dark matter theories @xcite . in fig . [ fig : coupling1 ] the correction in the gaugino ( higgsino ) components of the neutralino @xmath91 , @xmath92 ( @xmath93 ) , is presented . in fig . [ fig : coupling1]a we show that the correction for the lightest neutralino is in the range of 5% ( full line ) . in the case of gauge unification ( dashed line ) the additional large correction to @xmath94 ( approx . + 10.8% at @xmath95 gev ) leads to a change in the gaugino component up to 30% . in fig . [ fig : coupling1]b the corrections for all four neutralinos is given for the same parameter set . in the range between @xmath95 gev and @xmath96 gev , @xmath97 and @xmath98 are nearly mass degenerated at tree level . at one loop the mass order and as a consequence the numbering changes . this is just a small effect in the mass spectrum , but the interchanging of the gaugino and higgsino components is in the range of @xmath9930% . the chargino masses @xmath100 and production cross sections can be measured very precisely at a future @xmath3 linear collider @xcite . from these observables the mixing angles @xmath101 and by inverting the relations ( [ treeangle ] ) and ( [ treemasses ] ) the fundamental susy parameters @xmath0 , @xmath2 and @xmath5 can be obtained in lowest order @xcite . if a neutralino mass is known , one can also obtain @xmath1 at tree level . however , high precision experiments will make it necessary to take into account one loop corrections . in the following , we will compare the tree level approximations and the full one loop corrected fundamental susy parameters . + we use as input the two chargino masses , the mass of the lightest neutralino and assume the on shell @xmath5 is known from the higgs sector @xcite . calculating the susy parameters @xmath0 and @xmath2 from the tree level mass matrices as given in ( [ eq : chi+mat ] ) and ( [ neumat1 ] ) leads to a four fold ambiguity . for comparison we choose the same branch as in @xcite . because we use as input the physical masses this set of tree level mass matrices @xmath102 , @xmath103 is different from @xmath23 and @xmath104 , which give the true tree level mass eigenvalues . on the other hand , @xmath102 and @xmath103 are defined to give the right physical ( on shell ) chargino masses and one neutralino mass . to calculate the one loop corrections , values for the other susy parameters ( @xmath105 , @xmath106 , @xmath107 , @xmath108 ) are needed . the following example , calculated for the same set of parameters as in fig . [ fig : parameter1 ] but @xmath109 gev , shows the chargino and neutralino mass matrices in tree level approximation plus the one loop corrections , @xmath110 @xmath111 note that both @xmath102 ( @xmath103 ) and @xmath22 ( @xmath62 ) have the same physical mass eigenvalues of @xmath112 and @xmath113 . we call the parameters used in @xmath102 and @xmath103 effective parameters @xmath114 , @xmath115 and @xmath116 corresponding to the parameters used in @xcite . with @xmath117 ( and the corresponding relation for the chargino mass matrix ) the fundamental on shell parameters can be determined . for instance , @xmath118 . with @xmath0 fixed in the chargino system we get @xmath119 . + in fig . [ fig : parameter1 ] and [ fig : parameter2]a the differences between the effective parameters in @xmath103 , @xmath120 and the properly defined one loop on shell parameters are shown . the effective parameters are obtained applying tree level relations on the measured masses , while the on shell parameters are defined by the elements of the one loop corrected mass matrices . as the effective tree level and the one loop corrected chargino mass matrix have the same eigenvalues , this may imply sizeable corrections in the rotation angles @xmath121 . this can be seen in fig . [ fig : parameter2]b if the two chargino masses are known from experiment the complete neutralino mass spectrum can be predicted by assuming the relation @xmath122 for the @xmath8 parameters or in the tree level approximation for the effective parameters . in fig . [ fig : gutsu5 ] and [ fig : gutamsb ] there are two different cases shown : for a su(5 ) gut ( @xmath123 ) and an amsb model ( @xmath124 ) . we get large corrections for the bino like neutralino due to the correction in @xmath94 . this is for @xmath125 or @xmath98 ( depending on @xmath126 ) in the su(5 ) gut scenario and @xmath127 in the amsb model . the gut relation can be tested by calculating the @xmath8 parameters @xmath128 , @xmath129 and @xmath130 at a scale @xmath49 . assuming such a relation for the on shell or effective parameters is an inaccurate approximation , as shown in fig . [ fig : guttest ] . for the given set of input parameters the ratio @xmath131 ( full line ) fulfills the su(5 ) gut relation at @xmath132 gev . using the effective @xmath114 , @xmath116 ( dotted line ) and the on shell @xmath133 the calculation leads to @xmath134 gev . even for the on shell @xmath0 and @xmath1 the gut point lies at @xmath135 gev . we have presented a detailed discussion of the chargino and neutralino mass parameters at one loop level . the on shell parameters @xmath0 , @xmath2 and @xmath1 are properly defined by the on shell mass matrix elements . we have shown that at one loop level the values @xmath0 and @xmath2 depend on whether they are determined from the chargino or neutralino system . we discussed the difference between the on shell and the so called effective parameters , which are obtained from observables , e.g. on shell masses , inserted into tree level relations . the corrections to the tree level mass matrices in terms of the on shell and effective parameters are discussed in different scenarios . the numerical analysis based on a complete one loop calculation has shown that the corrections to the chargino and neutralino masses can go up to 10% and the change in the gaugino and higgsino components can be in the range of 30% . in addition , we have presented how a possible gut relation for the parameters @xmath0 and @xmath1 can be tested at one loop level . the authors acknowledge support from eu under the hprn - ct-2000 - 00149 network programme . the work was also supported by the fonds zur frderung der wissenschaftlichen forschung of austria , project no . p13139-phy . in the following we present the explicit formulas of all non(s)fermionic self energies for the neutralinos , charginos , @xmath136 and @xmath30 bosons and the @xmath137graphs in the mssm . the ( s)fermionic part can be found in the appendix of @xcite . the two point functions @xmath138 , @xmath139 , @xmath140 and @xmath141 @xcite are given in the convention @xcite . the neutralino and chargino mass matrix counter terms are @xcite : @xmath142 \ , , \nonumber \\ { \delta}y_{ij}&=&\frac{1}{2 } \sum_{l , n=1}^4 z_{li}z_{nj}\,{\rm re } \left [ m_{{{\tilde\chi^0}}_l } \pi^l_{nl } ( m_{{{\tilde\chi^0}}_l}^2 ) + m_{{{\tilde\chi^0}}_n } \pi^r_{ln}(m_{{{\tilde\chi^0}}_n}^2)+\pi^{s , r}_{nl}(m_{{{\tilde\chi^0}}_l}^2)+\pi^{s , l}_{ln}(m_{{{\tilde\chi^0}}_n}^2 ) \right ] \nonumber \ , , \\\end{aligned}\ ] ] with the convention @xmath143 @xmath144,\hspace{0.8 cm } \\ \pi_{ij}^{a_l^0}(k ) \ \ & = & \!\ ! \frac{-g^2}{(4\pi)^2}\sum_{n=1}^4 \sum_{l=3}^4 \left [ f_{nil}^0 f_{jnl}^0 \left ( m_{\tilde{\chi}_n^0 } b_0(k^2,m_{\tilde{\chi}_n^0}^2,m_{a_l^0}^2)+\ksla b_1(k^2,m_{\tilde{\chi}_n^0}^2,m_{a_l^0}^2)\right ) \right ] , \\ \pi_{ij}^{h_k^+}(k ) \hspace{0.1 cm } & = & \sum_{n=1}^2 \sum_{k=1}^2 \left [ \left ( f_{nik}^r f_{njk}^l + f_{nik}^l f_{njk}^r \right ) m_{\tilde{\chi}_n^+ } b_0(k^2,m_{\tilde{\chi}_n^+}^2,m_{h_k^+}^2)- \right . \nonumber \\ & & \hspace{3.5 cm } \left . \left ( f_{nik}^r f_{njk}^r + f_{nik}^l f_{njk}^l \right ) \ksla b_1(k^2,m_{\tilde{\chi}_n^+}^2,m_{h_k^+}^2 ) \right ] \ , . \\ \pi_{ij}^z(k ) \hspace{0.3 cm } & = & \frac{2g_z^2}{(4\pi)^2 } \sum_{n=1}^4 o_{ni}^{''l}o_{jn}^{''l}\left [ 2m_{\tilde{\chi}_n^0 } b_0(k^2,m_{\tilde{\chi}_n^0}^2,m_z^2)-\ksla b_1(k^2,m_{\tilde{\chi}_n^0}^2,m_z^2 ) \right ] \ , , \\ \pi_{ij}^w(k ) \hspace{0.2 cm } & = & \frac{-2g^2}{(4\pi)^2}\sum_{n=1}^2 \left [ 2\left(o_{in}^r o_{jn}^l + o_{in}^l o_{jn}^r \right)m_{\tilde{\chi}_n^+ } b_0(k^2,m_{\tilde{\chi}_n^+}^2,m_w^2)+ \right . \nonumber \\ & & \hspace{3.5 cm } \left . \left ( o_{in}^l o_{jn}^l + o_{in}^r o_{jn}^r \right)\ksla b_1(k^2,m_{\tilde{\chi}_n^+}^2,m_w^2 ) \right ] \ , .\end{aligned}\ ] ] @xmath145\ , , \\ \pi_{ij}^{a_l^0}(k ) & = & -\frac{g^2}{(4\pi)^2 } \sum_{n=1}^2 \sum_{l=3}^4 \left [ \ksla \left ( f_{nil}^+f_{njl}^+ p_r + f_{inl}^+f_{jnl}^+ p_l \right ) b_1(k^2,m_{\tilde \chi_n^+}^2,m_{a_l^0}^2)+ \right . \nonumber \\ & & \hspace{20 mm } \left . m_{\tilde \chi_n^+ } \left ( f_{nil}^+f_{jnl}^+ p_r + f_{inl}^+f_{njl}^+ p_l \right ) b_0(k^2,m_{\tilde \chi_n^+}^2,m_{a_l^0}^2 ) \right ] \ , , \\ \pi_{ij}^{h_k^+}(k ) & = & -\frac{g^2}{(4\pi)^2 } \sum_{n=1}^4 \sum_{k=1}^2 \left [ \ksla \left ( f_{ink}^l f_{jnk}^l p_r + f_{ink}^r f_{jnk}^r p_l \right ) b_1(k^2,m_{\tilde \chi_n^0}^2,m_{h_k^+}^2)- \right . \nonumber \\ & & \hspace{20 mm } \left . m_{\tilde \chi_n^0 } \left ( f_{ink}^l f_{jnk}^r p_r + f_{ink}^r f_{jnk}^l p_l \right ) b_0(k^2,m_{\tilde \chi_n^0}^2,m_{h_k^+}^2 ) \right]\ , , \\ \pi_{ij}^\gamma ( k ) & = & -\frac{2e^2}{(4\pi)^2}\delta_{ij } \left [ \ksla b_1(k^2,m_{\tilde \chi_i^+}^2,0)+2m_{\tilde \chi_i^+}b_0(k^2,m_{\tilde \chi_i^+}^2,0 ) \right]\ , , \\ \pi_{ij}^{z}(k ) & = & -\frac{2g_z^2}{(4\pi)^2 } \sum_{n=1}^2 \left [ \ksla \left ( o_{ni}^{'r}o_{jn}^{'r } p_r + o_{ni}^{'l}o_{jn}^{'l } p_l \right ) b_1(k^2,m_{\tilde \chi_n^+}^2,m_z^2)+ \right . \nonumber \\ & & \hspace{20 mm } \left . 2m_{\tilde \chi_n^+ } \left ( o_{ni}^{'r}o_{jn}^{'l } p_r + o_{ni}^{'l}o_{jn}^{'r } p_l \right ) b_0(k^2,m_{\tilde \chi_n^+}^2,m_z^2 ) \right]\ , , \\ \pi_{ij}^{w}(k ) & = & -\frac{2g^2}{(4\pi)^2 } \sum_{n=1}^4 \left [ \ksla \left ( o_{ni}^{r}o_{nj}^{r } p_r + o_{ni}^{l}o_{nj}^{l } p_l \right ) b_1(k^2,m_{\tilde \chi_n^0}^2,m_w^2)+ \right . \nonumber \\ & & \hspace{20 mm } \left . 2m_{\tilde \chi_n^0 } \left ( o_{ni}^{r}o_{nj}^{l } p_r + o_{ni}^{l}o_{nj}^{r } p_l \right ) b_0(k^2,m_{\tilde \chi_n^0}^2,m_w^2 ) \right]\,.\end{aligned}\ ] ] @xmath146\ , , \\ \pi_t^{\tilde \chi^+ \tilde \chi^+ } ( k^2 ) & = & -\frac{g_z^2}{(4\pi)^2 } \sum_{i , j=1}^2\left[\left ( ( m_{\tilde \chi^+_i}^2+m_{\tilde \chi^+_j}^2-k^2)((o_{ij}^{'l})^2+(o_{ij}^{'r})^2 ) -4 o_{ij}^{'l } o_{ij}^{'r}m_{\tilde \chi^+_i}m_{\tilde \chi^+_j } \right ) \right . \nonumber \\ & & \hspace{-20 mm } \left . b_0(k^2,m_{\tilde \chi^+_i}^2,m_{\tilde \chi^+_j}^2 ) + ( ( o_{ij}^{'l})^2+(o_{ij}^{'r})^2 ) \left ( 2a_0(m_{\tilde \chi^+_i}^2)-4b_{00}(k^2,m_{\tilde \chi^+_i}^2,m_{\tilde \chi^+_j}^2)\right)\right ] \ , , \\ \pi_t^{h_k^0 a_l^0}(k^2 ) & = & -\frac{g_z^2}{(4\pi)^2 } \sum_{k , l=1}^2 c_{kl}^2b_{00}(k^2,m_{a_l^0}^2,m_{h_k^0}^2)\ , , \\ \pi_t^{h_k^+ h_k^+}(k^2 ) & = & -\frac{g_z^2}{(4\pi)^2 } ( 1 - 2s_w^2)^2 \sum_{k=1}^2 b_{00}(k^2,m_{h_k^+}^2,m_{h_k^+}^2)\ , , \\ \pi_t^{h_k^0/a_k^0/h_k^+ } & = & \frac{g_z^2}{4(4\pi)^2}\sum_{k=1}^2 \left ( a_0(m_{h_k^0}^2 ) + a_0(m_{a_k^0}^2)+2(1 - 2s_w^2)^2a_0(m_{h_k^+}^2 ) \right ) \ , , \\ \pi_t^{z h_k^0}(k^2 ) & = & \frac{g_z^2}{(4\pi)^2}m_z^2 \left ( s^2_{{\alpha}{\beta}}b_0(k^2,m_{h^0}^2,m_z^2)+c^2_{{\alpha}{\beta}}b_0(k^2,m_{h^0}^2,m_z^2 ) \right)\ , , \\ \pi_t^{w g^+}(k^2 ) & = & 2\frac{g_z^2}{(4\pi)^2}m_w^2s_w^4 b_0(k^2,m_{g^+}^2,m_w^2)\ , , \\ \pi_t^{ww}(k^2 ) & = & -\frac{g_z^2}{(4\pi)^2}\left [ 10 b_{00}(k^2,m_w^2,m_w^2)+(5k^2 + 2m_w^2)b_0(k^2,m_w^2,m_w^2 ) \right . \nonumber \\ & & \hspace{40 mm } \left . + 2k^2b_1(k^2,m_w^2,m_w^2 ) + 2a_0(m_w^2)\right]\ , , \\ \pi_t^{w/\omega}(k^2 ) & = & \frac{3g^2}{8\pi^2}c_w^2 a_0(m_w^2 ) + 2\frac{g^2}{(4\pi)^2}c_w^2 b_{00}(k^2,m_{\omega^+}^2,m_{\omega^+}^2)\,.\end{aligned}\ ] ] @xmath147 \ , , \nonumber \\ \pi_t^{h h}(k^2 ) & = & -\frac{g^2}{(4\pi)^2}\left ( \sum_{k , l=1}^2 c_{kl}^2 b_{00}(k^2,m_{h_l^+}^2,m_{h_k^0}^2)+\sum_{k=1}^2 b_{00}(k^2,m_{h_k^+}^2,m_{a_k^0}^2 ) \right ) \ , , \\ \pi_t^{h } & = & \frac{1}{(4\pi)^2}\frac{g^2}{2}\sum_{k=1}^2\left [ \frac{1}{2}a_0(m_{h_k^0}^2)+\frac{1}{2}a_0(m_{a_k^0}^2)+a_0(m_{h_k^+}^2 ) \right ] \ , , \\ \pi_t^{h_k^0 w}(k^2)&=&\frac{1}{(4\pi)^2}g^2m_w^2\left(c^2_{{\alpha}{\beta } } b_0(k^2,m_{h^0}^2,m_w^2)+s^2_{{\alpha}{\beta}}b_0(k^2,m_{h^0}^2,m_w^2)\right)\ , , \\ \pi_t^{z g^+}(k^2 ) & = & \frac{1}{(4\pi)^2}m_w^2\left[g_z^2 s_w^4 b_0(k^2,m_{g^+}^2,m_z^2)+e^2b_0(k^2,m_{g^+}^2,0 ) \right]\ , , \\ \pi_t^{wz}(k^2)&= & -\frac{1}{(4\pi)^2}g^2c_w^2\left [ 10 b_{00}(k^2,m_z^2,m_w^2)+(5k^2 + 2m_z^2)b_0(k^2,m_z^2,m_w^2 ) \right . \nonumber \\ & & \hspace{40 mm } \left . + 2k^2b_1(k^2,m_z^2,m_w^2 ) + 2a_0(m_w^2)\right ] \ , , \\ \pi_t^{w\gamma}(k^2 ) & = & -\frac{1}{(4\pi)^2}g^2s_w^2\left [ 10 b_{00}(k^2,0,m_w^2)+5k^2b_0(k^2,0,m_w^2 ) \right . \nonumber \\ & & \hspace{40 mm } \left . + 2k^2b_1(k^2,0,m_w^2 ) + 2a_0(m_w^2)\right]\ , , \\ \pi_t^{z/\gamma / w } & = & \frac{3g^2}{(4\pi)^2}\left [ c_w^2 a_0(m_z^2 ) + s_w^2 a_0(0)+a_0(m_w^2)\right]\ , , \\ \pi_t^{\omega}(k^2 ) & = & 2\frac{g^2}{(4\pi)^2}\left [ c_w^2 b_{00}(k^2,m_{\omega_z}^2,m_{\omega_+}^2 ) + s_w^2 b_{00}(k^2,m_{\omega_\gamma}^2,m_{\omega_+}^2 ) \right]\,.\end{aligned}\ ] ] @xmath148\ , , \\ \pi_{az}^{h}(k^2)&=&\frac{ig_z^2m_z}{4(4\pi)^2}\sum_{l , k=1}^2 c_{kl}c_{kl } ' \left [ b_0(k^2,m_{a_l^0}^2,m_{h_k^0}^2)+ 2b_1(k^2,m_{a_l^0}^2,m_{h_k^0}^2 ) \right]\ , , \\ \pi_{az}^{zh}(k^2)&=&\frac{ig_z^2m_z}{2(4\pi)^2 } c_{{\alpha}{\beta } } s_{{\alpha}{\beta}}\sum_{k=1}^2\left[(-1)^k ( 2b_0+b_1)(k^2,m_z^2,m_{h_k^0}^2 ) \right ] \,.\end{aligned}\ ] ] we used the abbreviations @xmath149 , @xmath150 , @xmath151 , @xmath152 , @xmath153 , with @xmath154 the mixing angle in the \{@xmath155 , @xmath156 } system , and for the higgs fields @xmath157 , @xmath158 , @xmath159 . the coupling matrices are : @xmath160 \\[2 mm ] & & + \frac{d_k}{2 } \big [ z_{l4 } z_{m2 } + z_{m4 } z_{l2 } - \tan\theta_w \left ( z_{l4 } z_{m1 } + z_{m4 } z_{l1 } \right ) \big ] \ = \ f_{mlk}^0\end{aligned}\ ] ] and @xmath161 \ , , \\ f_{ilk}^l & = & -e_{k+2 } \left [ u_{i1 } z_{l3 } - \frac{1}{\sqrt2 } ( z_{l2 } + z_{l1 } \tan\theta_w ) u_{i2 } \right ] \,.\end{aligned}\ ] ] @xmath162 and @xmath163 take the values @xmath164 the other used couplings are @xmath165 @xmath166 @xmath167 @xmath168 @xmath169 g. jungman , m. kamionkowski , and k. griest , phys . rep . * 267 * , 195 ( 1996 ) ; a. b. lahanas , d. v. nanopoulos , and v. c. spanos , mod . a * 16 * , 1229 ( 2001 ) ; phys . b * 518 * , 94 ( 2001 ) ; m. drees , y. g. kim , t. kobayashi , and m. m. nojiri , phys . d * 63 * , 115009 ( 2001 ) .

in the minimal supersymmetric standard model ( mssm ) the masses of the neutralinos and charginos depend on the gaugino and higgsino mass parameters @xmath0 , @xmath1 and @xmath2 . if supersymmetry is realized , the extraction of these parameters from future high energy experiments will be crucial to test the underlying theory . we present a consistent method how on shell parameters can be properly defined at one loop level and how they can be determined from precision measurements . in addition , we show how a gut relation for the parameters @xmath0 and @xmath1 can be tested at one loop level . the numerical analysis is based on a complete one loop calculation . the derived analytic formulae are given in the appendix . hephy - pub 769/03 + hep - ph/0304006 * analysis of the chargino and neutralino mass parameters at one loop level * w. ller , h. eberl , w. majerotto , c. weber [ cols= " < " , ]

the schematic form of the equations ( 3a ) and ( 3b ) , is obtained by following the outline given in the paper . the final forms will be : @xmath104 where @xmath105 is the schematic version of @xmath106 : @xmath107 eqs . ( 7a ) and ( 7b ) contain two `` source '' terms which we present in detail here . to proceed , let us define @xmath108{\mathrm{d}}s.\end{aligned}\ ] ] then the source terms can be written in the form @xmath109 the equations for the three - point correlators are also schematicised in a similar way as stated in the paper . the final schematic forms of equations ( 7a ) and ( 7b ) will be

we construct the equations for the growth kinetics of a structural glass within mode - coupling theory , through a non - stationary variant of the 3-density correlator defined in phys . rev . lett . * 97 * , 195701 ( 2006 ) . we solve a schematic form of the resulting equations to obtain the coarsening of the 3-point correlator @xmath0 as a function of waiting time @xmath1 . for a quench into the glass , we find that @xmath2 attains a peak value @xmath3 at @xmath4 , providing a theoretical basis for the numerical observations of parisi [ j. phys . chem . b * 103 * , 4128 ( 1999 ) ] and kob and barrat [ phys . rev . lett . * 78 * , 4581 ( 1997 ) ] . the aging is not `` simple '' : the @xmath1 dependence can not be attributed to an evolving effective temperature . when a system is quenched below an ordering transition , domains of the ordered phase appear and begin to grow @xcite , with characteristic size given by the decay length of equal - time correlations of the order parameter . the corresponding issue for the glass transition has been examined numerically @xcite , using susceptibilities and correlation lengths that capture the onset of amorphous freezing @xcite , but a quantitative theory of these observations has been lacking @xcite . length - scale information similar to that obtained from the 4-density correlator and related overlap functions @xcite has been shown @xcite to be contained in a certain three - point correlator @xmath5 , whose peak value , and the time at which the peak is attained , diverge @xcite upon approaching the mode - coupling glass transition @xcite . in this work we present a theory of _ the coarsening of glassy order _ , using a non - stationary generalization @xmath0 , whose peak value @xmath6 is the correlation volume as a function of the waiting time @xmath1 since the quench . we formulate our calculation in the framework of the fluctuating hydrodynamics of a dense liquid , and obtain results using mode - coupling theory ( mct ) @xcite , in a schematic approach @xcite . figures [ corrfn ] - [ chi4peakfit ] summarize our results . we find that @xmath6 grows without bound for a quench into the mct glass ( fig . [ largelambda ] ) , as @xmath7 , and the relaxation time as @xmath8 ( fig . [ chi4peakfit ] ) , in agreement respectively with the computer experiments of parisi @xcite and kob and barrat @xcite . as effects beyond mct cut off the transition , the coarsening in experiments , simulations , or a complete theory will cease at long enough times , but typical simulations do not explore these asymptotically long time scales and can therefore be compared usefully to our mct coarsening predictions . the three - point function , fig . [ chi4peakfit ] , shows features incompatible with `` simple aging '' @xcite but qualitatively similar to @xcite . for a quench to a distance @xmath9 from the threshold value on the _ liquid _ side , @xmath10 grows to saturation ( fig . [ lowlambda ] ) , reaching an equilibrium value @xmath11 , with a relaxation time @xmath12 ( figure [ ss_chi3scaling ] ) . as a function of @xmath13 , for various waiting times @xmath1 shown in the legend . the decay with @xmath13 becomes progressively slower with increasing @xmath1 . the final parameter values are @xmath14 and @xmath15 . * inset : * scaling @xmath16 by @xmath17 yields a data collapse in the @xmath18-relaxation regime . such `` simple aging '' , however , is not seen in the three - point function , fig . [ chi4peakfit].,width=287 ] mct is an analytically tractable approximation to equilibrium liquid - state dynamics that yields a glass transition in a homogeneous system , which is why it is so widely used despite its shortcomings @xcite . in order to extend mct to describe non - stationary states such as coarsening we work with a general field - theoretic approach @xcite , taking care not to use results like the kubo formulae and fluctuation - dissipation relations ( fdr ) , which are justified only in equilibrium treatments @xcite . we start with the equations of fluctuating hydrodynamics for the velocity and density fields for an isothermal compressible fluid , extended to large wavenumbers so as to take into account the modes around the structure factor peak @xcite . in order to obtain an equation for the density field alone , we eliminate the velocity while retaining momentum conservation but ignoring inertia . this yields the dynamical equation @xmath19 for the fourier - transformed density fluctuation @xmath20 at wavevector @xmath21 , with @xmath22 $ ] , @xmath23 and @xmath24 . eq . ( [ kdep1 ] ) can be viewed as the no - inertia limit of eq . ( 4.1 ) of @xcite . here @xmath25 is the longitudinal damping , where @xmath26 and @xmath27 are the bare shear and bulk viscosities , @xmath28 is boltzmann s constant times temperature , @xmath29 and @xmath30 are the equilibrium static structure factor and direct correlation function respectively and the noise @xmath31 obeys @xmath32 from diagrammatic perturbation theory @xcite , we construct the equations of motion for the correlation function @xmath33 and response function @xmath34 : [ expliciteq ] @xmath35 with @xmath36 , @xmath37 and @xmath38 . the contribution to ( [ expliciteqa ] ) from the first term in @xmath39 vanishes due to causality . franz and hertz @xcite obtained schematic equations similar to ( [ expliciteqa ] ) and ( [ expliciteqb ] ) for the amit - roginsky model @xcite . how are the input quantities @xmath40 and @xmath41 in equations ( [ expliciteqa ] ) and ( [ expliciteqb ] ) defined for the case of a quench ? a comparison with the treatment of zaccarelli _ et al . @xcite is useful here . the @xmath41 term in ( [ kdep1 ] ) and ( [ expliciteq ] ) involves the `` residual interactions '' in @xcite . we define our quench to be an abrupt increase in the interaction strength , implying that @xmath41 should be evaluated at the final parameter value . to determine @xmath40 , which must now be a time - dependent quantity as we are dealing with a non - stationary state , we insist , as in @xcite , that for @xmath42 eq . ( [ expliciteq ] ) obeys time - translation invariance and the fdr . this leads , after some algebra , to @xmath43{\mathrm{d}}s.\end{aligned}\ ] ] in @xcite the term corresponding to @xmath40 enters through the equal - time density correlator . the latter being time - dependent in a coarsening situation , it is natural that @xmath40 should change in time . grows and shifts to higher @xmath44 with increasing waiting time @xmath1.,width=249 ] to obtain the equation of motion for the growth kinetics of glassy correlations , we look at the behaviour of our nonstationary generalization of the three - density correlation function @xcite mentioned above . we introduce in the free - energy functional a one - body term @xmath45 , coupling the density to an external potential @xmath46 and leading on average to an inhomogeneous shift @xmath47 in the mean density field @xcite . we work in the limit where @xmath46 and hence @xmath47 are uniform , so that the fourier transform @xmath48 has non - zero weight @xmath49 only for wavevector @xmath50 . we will see that this suffices for the purpose of extracting the correlation volume . the resulting generalized langevin equation for @xmath51 , to first order in the background density @xmath49 which encodes the effects of the field , is @xcite @xmath52 let @xmath53 and @xmath54 denote the @xmath49-dependent two - point correlation and response functions implied by ( [ withpotential ] ) . as we are working in a non - stationary state , we must define separate 3-point quantities analogous to @xmath2 in @xcite for @xmath55 and @xmath56 : @xmath57 and @xmath58 , with equations of motion [ chirc ] @xmath59 where @xmath60 , and @xmath61 , @xmath62 and @xmath63 are quantities corresponding to @xmath64 and @xmath65 but evaluated in the presence of @xmath46 . the expressions for the source terms @xmath66 and @xmath67 are given in the supplementary information ( si ) and the straightforward but tedious derivation of ( [ chirca ] ) and ( [ chircb ] ) will be presented in a subsequent paper @xcite . as a function of @xmath13 for various @xmath1 for @xmath68 , corresponding to a quench into the liquid . * inset : * the two - point correlator , for the same parameter values , shows a relaxation time that grows progressively with increasing @xmath1 but approaches a finite value.,width=249 ] as a function of @xmath13 for various @xmath69 . ( b ) when @xmath70 is scaled with @xmath71 and time with @xmath44 , we obtain data collapse for large @xmath13 . ( c ) @xmath72 . ( d ) @xmath73,width=287 ] . [ ss_chi3scaling ] simplified integral equations keeping track of time - dependence alone @xcite have proved invaluable in extracting meaningful results from mct within a manageable calculation . in this spirit , we suppress dependence on wavevector @xmath21 and write the self - energies in ( [ expliciteq ] ) and ( [ chirc ] ) as @xmath74 and @xmath75 , yielding equations for the two- and three - point correlation and response functions which we now denote @xmath76 , @xmath77 , @xmath78 and @xmath79 . we solve the resulting schematic versions of ( [ expliciteq ] ) , ( [ k1oft ] ) and ( [ chirc ] ) , whose detailed forms are given in the si section , using the algorithm developed by kim and latz @xcite , to give the results quoted at the start of this paper . first , aging can clearly be seen in the behaviour of @xmath76 in fig . [ corrfn ] . second and more important is the characteristic non - monotone behavior of @xmath78 , and its dependence on @xmath1 and interaction strength @xmath80 ( fig . [ largelambda ] ) . for a fixed initial condition corresponding in our schematic approach to a liquid with negligible correlations , we examine in particular how @xmath78 as a function of @xmath13 changes with @xmath1 , for values of @xmath80 corresponding to the liquid and the glass phase . recall that @xmath80 defines the point to which the system is quenched . for @xmath80 in the liquid phase but close to the transition we find , as expected , that @xmath78 attains a peak value @xmath71 at a time @xmath44 , with both @xmath71 and @xmath44 growing with @xmath1 but saturating to finite values as shown in fig . [ lowlambda ] . the final peak value of @xmath78 grows as @xmath81 and @xmath73 ( fig . [ ss_chi3scaling ] ) . these final values , obtained at @xmath82 are the equilibrium values of the corresponding quantities . in the notation of @xcite we are working at @xmath83 and our results are consistent with theirs in that limit . a more detailed comparison with @xcite or @xcite , including an estimate of the correlation _ length _ requires a calculation of the sensitivity of two - point functions to a spatially varying potential . by @xmath84 and time by @xmath44 , the three - point function does not follow a master curve : the behaviour at various @xmath1 is not the equilibrium dynamics at an evolving effective temperature . this behaviour of @xmath85 belies the expectation of `` simple aging '' suggested by the two - point function in fig . [ corrfn ] . ( b ) the peak height @xmath86 . ( c ) @xmath87 and @xmath17 both grow as @xmath8 . ( d ) @xmath87 is proportional to @xmath17 . ( e ) if we scale @xmath78 by @xmath88 and plot it as a function of @xmath89 , as in parsaeian _ et al . _ @xcite , no data collapse is found . @xmath1 values are shown in the legend . ( f ) scaling @xmath78 by @xmath71 and plotting it as a function of @xmath89 shows data collapse in the regime of @xmath18-relaxation . waiting times as in ( e).,width=325 ] for a quench into the glassy region , @xmath90 , as shown in fig . [ largelambda ] , @xmath71 grows without bound . in more detail ( fig . [ chi4peakfit ] ) , @xmath91 at a time @xmath92 , with increasing @xmath1 . the exponents @xmath93 in agreement with simulations @xcite and @xmath94 . from the two - point correlator @xmath76 we find a relaxation time @xmath95 close to the transition . thus our result implies @xmath96 , in agreement with the numerical experiment of ref . @xcite ; the relation between @xmath17 and @xmath44 remains to be tested . @xmath71 measures an effective correlation volume , so that its growth is the claimed coarsening of glassy structure , and is consistent with the idea of a growing `` domain size '' . regardless of the precise values obtained , it is significant that our theory and the simulations of @xcite all find a total structure factor @xmath71 growing very sublinearly in time . our scaling laws differ quantitatively ( fig . [ chi4peakfit]e ) from those of @xcite , perhaps because we measure different quantities . however , if we scale @xmath78 by @xmath71 and plot them as a function of @xmath89 , data collapse is obtained in the @xmath18-relaxation regime ( fig . [ chi4peakfit]f ) . we do not claim to understand the origin of this scaling or , for that matter , that of ref . @xcite . a similar calculation @xcite for the three - point correlation function for a @xmath97-spin spin - glass model with @xmath98 finds again a growing @xmath71 , but slower than for the present problem . we emphasize that the @xmath1-dependent properties we extract do not correspond to those of an equilibrium system at an evolving @xmath80 or temperature . had it been so , scaling @xmath78 by @xmath84 and time by @xmath44 would have given data collapse for all @xmath1 as for the equilibrium case ( fig . [ ss_chi3scaling ] ) . fig . [ chi4peakfit ] shows the absence of such collapse even for larger @xmath1 . it would appear that the 3-point correlator is more sensitive to departures from `` simple aging '' , and an interpretation in terms of an evolving effective temperature , than the two - time correlation function @xcite ( see inset of fig . [ corrfn ] ) . perhaps the monotone decay of the latter masks such deviations or , more likely , @xmath85 carries additional , independent information . we close by summarising the achievements of this work . we have shown that mode - coupling theory adapted to describe non - stationary states captures the key features of the emergence and coarsening of glassy order from a liquid . through the evolution of a three - point function we have shown that the glassy correlation volume grows as @xmath7 with waiting time @xmath1 , slower than domain volumes in conventional coarsening , and the relaxation time of the glass grows as @xmath8 . these theoretical growth laws are supported by simulation studies @xcite , and the broad features we observe are similar to those in @xcite . in an experimental realization , if the quench is below the mct transition but above a putative ideal glass transition at , say , the kauzmann temperature @xmath99 , activated processes @xcite outside the scope of mct should cut off the growth . presumably a quench below @xmath99 will give indefinite growth of a different glassy length scale @xcite with a form not predicted by mct . in results to be presented separately @xcite we find further that an imposed shear - rate @xmath100 cuts off aging and coarsening at @xmath101 in the glassy region and @xmath102 in the fluid . since the relaxation time goes as @xmath8 , this should imply that @xmath17 or @xmath44 should vary as @xmath103 . we look forward to experimental tests of our results . we thank c. dasgupta for valuable suggestions and a critical reading of our paper and n. menon for enlightening comments . we also thank s.m . bhattacharyya , b. kim , k. miyazaki , s. sastry , d. sen , s.p . singh and e. zaccarelli for discussions . skn was supported in part by the university grants commission and sr by a j.c . bose fellowship from the department of science and technology . * *

the interaction between intense laser beams and plasmas leads to a variety of different instabilities , including brillouin and raman forward and backward @xcite scattering and modulational instabilities . in multiple dimensions we also have filamentation and side - scattering instabilities . relativistic effects can then play an important role @xcite . when two laser beams interact in the plasma , we have a new set of phenomena . an interesting application is the beat - wave accelerator , in which two crossing beams with somewhat different frequencies can accelerate electrons to ultra - relativistic speeds via the ponderomotive force acting on the electrons . the modulational and filamentation instabilities of multiple co - propagating electromagnetic waves can be described by a system of coupled nonlinear schrdinger equations from which the nonlinear wave coupling and the interaction between localized light wave packets can be easily studied @xcite . two co - propagating narrow laser beams may attract each other and spiral around each other @xcite or merge @xcite . counter - propagating laser beams detuned by twice the plasma frequency can , at relativistic intensities , give rise to fast plasma waves via higher - order nonlinearities @xcite . at relativistic amplitudes , plasma waves can also be excited via beat wave excitation at frequencies different from the electron plasma frequency , with applications to efficient wake - field accelerators @xcite . the relativistic wakefield behind intense laser pulses is periodic in one - dimension @xcite and shows a quasi - periodic behavior in multi - dimensional simulations @xcite . particle - in - cell simulations have demonstrated the generation of large - amplitude plasma wake - fields by colliding laser pulses @xcite or by two co - propagating pulses where a long trailing pulse is modulated efficiently by the periodic plasma wake behind the first short pulse @xcite . in the present paper , we consider the nonlinear interaction between two weakly relativistic crossing laser beams in plasmas . we derive a set of nonlinear mode coupled equations and nonlinear dispersion relations , which we analyze for raman backward and forward scattering instabilities as well as for brillouin and modulation / self - focusing instabilities . we consider the propagation of intense laser light in an electron ion plasma . the slowly varying electron density perturbation is denoted by @xmath0 . thus , our starting point is the maxwell equation @xmath1 the laser field is given in the radiation gauge , @xmath2 and @xmath3 . since @xmath4 , we thus have @xmath5 . moreover , @xmath6 , where @xmath7 is the electron rest mass and @xmath8 is the relativistic gamma factor , so that @xmath9 for weakly relativistic particles , i.e. @xmath10 , we can approximate ( [ eq : e - vel ] ) by @xmath11 with these prerequisites , eq . ( [ eq : evol ] ) becomes @xmath12 where @xmath13 is the electron plasma frequency and we have denoted @xmath14 . next , we divide the vector potential into two parts according to @xmath15 , representing the two laser pulses . we also consider the case @xmath16 . with this , we obtain from ( [ eq : nlin ] ) the two coupled equations @xmath17 and @xmath18 [ eq : nlin - coupled ] assuming that @xmath19 is proportional to @xmath20 where @xmath21 , we obtain in the slowly varying envelope approximation two coupled nonlinear schrdinger equations @xmath22 and @xmath23 [ eq : nlse - coupled ] where @xmath24 is the group velocity and @xmath25 is the electromagnetic wave frequency . in order to close ( [ eq : nlse - coupled ] ) , we next consider the slow plasma response . here we may follow two routes . first , if we assume immobile ions , the slowly varying electron number density and velocity perturbations satisfy the equations @xmath26 and @xmath27 where @xmath28 is the electron temperature , together with the poisson equation @xmath29 thus , combining ( [ eq7])([eq9 ] ) together with the vector potential decomposition , we obtain @xmath30 where the electron thermal velocity is denoted by @xmath31 . second , if the electrons are treated as inertialess , we have in the quasi - neutral limit @xmath32 @xmath33 and @xmath34 adding eqs . ( [ eq : inless1 ] ) and ( [ eq : inless2 ] ) , we obtain @xmath35 which should be combined with @xmath36 to obtain @xmath37 where the sound speed is @xmath38 and @xmath39 is the ion temperature . we shall consider , successively , eqs . ( 6a , b ) combined with ( [ eq : acoustic1 ] ) ( case i : raman scattering ) or with ( [ eq : acoustic2 ] ) ( case ii : brillouin scattering ) . setting @xmath40 and @xmath41 into the equations for the plasma density responses , we obtain @xmath42 where , for case i : @xmath43 and for case ii : @xmath44 [ alpha0 ] the expressions ( [ response ] ) and ( [ alpha0 ] ) derived above provide the slow plasma response for any given pair of fields @xmath45 ( @xmath46 ) . the latter now obey a set of coupled equations , which are obtained by substituting ( [ response ] ) into ( [ eq : nlse - coupled ] ) , @xmath47 and @xmath48 [ cnlse1 ] for convenience , eqs . ( [ cnlse1]a ) and ( [ cnlse1]b ) are cast into the reduced form as @xmath49 has been normalized by @xmath50 and where the nonlinearity / coupling coefficients are @xmath51 and @xmath52 [ qcoeff ] for stimulated raman ( case i ) and brillouin ( case ii ) scattering , respectively . we observe that the expressions ( [ qcoeff]a ) and ( [ qcoeff]b ) may be either positive or negative , depending on the frequency @xmath53 , prescribing either the modulational instability or the raman and brillouin scattering instabilities @xcite . the two nonlinear wave equations are identical upon an index ( @xmath54 ) interchange , and coincide for equal frequencies @xmath55 . we now investigate the parametric instabilities of the system of equations ( [ red1 ] ) and ( [ red2 ] ) . fourier decomposing the system by the ansatz @xmath56\exp(-i\omega_0 t)$ ] , where @xmath57 , and sorting for different powers of @xmath58 , we find the nonlinear frequency shift @xmath59 where @xmath60 denotes the expression for @xmath61 with @xmath62 . for the nonlinear wave couplings , we have from ( [ cnlse ] ) the system of equations @xmath63 [ nonlinsys ] where the unknowns are @xmath64 , @xmath65 , @xmath66 , and @xmath67 . the sidebands are characterized by @xmath68-c^2k^2,\ ] ] where we have used @xmath69 . the solution of the system of equations ( [ nonlinsys ] ) yields the nonlinear dispersion relation @xmath70 which relates the complex - valued frequency @xmath53 to the wavenumber @xmath71 . equation ( [ nonlindisp ] ) covers raman forward and backscattering instabilities , as well as the brillouin backscattering instability or the modulational / self - focusing instability , depending on the two expressions for the coupling constant @xmath61 . if either @xmath72 or @xmath73 is zero , then we recover the usual expressions for a single laser beam in a laboratory plasma , or for a high - frequency radio beam in the ionosphere @xcite . we have solved the nonlinear dispersion relation ( [ nonlindisp ] ) and presented the numerical results in figs . 15 . in all cases , we have used the normalized weakly relativistic pump wave amplitudes @xmath74 with different sets of wavenumbers for the two beams . the nonlinear couplings between the laser beams and the langmuir waves , giving rise to the raman scattering instabilities ( case i ) , are considered in figs . 1 and 2 . the instability essentially obeys the matching conditions @xmath75 and @xmath76 , where @xmath77 and @xmath78 are the frequency and wavenumbers of the pump wave , @xmath79 and @xmath80 are the frequency and wavenumbers for the scattered and frequency downshifted electromagnetic daughter wave , @xmath53 and @xmath81 are the frequencies of the langmuir waves , and where the light waves approximately obey the linear dispersion relation , @xmath82 , @xmath83 and the low - frequency waves obey the langmuir dispersion relation @xmath84 . we thus have the matching condition @xmath85^{1/2}+(\omega_{p0}^2 + 3 k^2 v_{te}^2)^{1/2}$ ] , which in two - dimensions relates the components @xmath86 and @xmath87 of the langmuir waves to each other , and which gives rise to almost circular regions of instability , as seen in figs . 1 and 2 . in the upper left and right panels of fig . 1 , we have assumed that the single beams @xmath88 and @xmath89 propagate in the @xmath90 and @xmath91 direction , respectively , having the wavenumber @xmath92 and @xmath93 , respectively . we can clearly see a backward raman instability , which for the beams @xmath88 and @xmath89 have maximum growth rates at @xmath94 and @xmath95 , respectively . the backward raman instability is connected via the obliquely growing wave modes to the forward raman scattering instability that has a maximum growth rate ( much smaller than that of the backward raman scattering instability ) at the wave number @xmath96 in the same directions as the laser beams . in the lower panels , we consider the two beams propagating simultaneously in the plasma , at a right angle to each other ( lower left panel ) and in opposite directions ( lower right panel ) . we see that the dispersion relation predicts a rather weak interaction between the two laser beams , where the lower left panel shows more or less a superposition of the growth rates in the two upper panels . the case of two counter - propagating laser beams ( lower right panel ) also shows a weak interaction between the two beams . for the case of equal wavelengths of the two pump waves , as shown in fig . 2 , we have a similar scenario as in fig . 1 . the lower left panel of fig . 2 shows that the growth rate of two interacting laser beams propagating at a right angle to each other is almost a superposition of the growth rates of the single laser beams displayed in the upper panels of fig . 2 . only for the counter - propagating laser beams in the lower right panel we see that the instability regions have split into broader and narrower bands of instability , while the magnitude of the instability is the same as for the single beam cases . we next turn to the brillouin scattering scenario ( case ii ) , in which the laser wave is scattered against ion acoustic waves , displayed in figs . 3 and 4 . in the weakly nonlinear case , we have three - wave couplings in the same manner as for the interaction with langmuir waves , and we see in both figs . 3 and 4 that the instability has a maximum growth rate in a narrow , almost circular band in the @xmath97 plane . in the upper two panels , we also see the backscattered brillouin instability with a maximum growth rate at approximately twice the pump wavenumbers , but we do not have the forward scattered instability . instead , we see a broadband weak instability in all directions and also perpendicular to the pump wavenumbers . a careful study shows that the perpendicular waves are purely growing , i.e. there may be density channels created along the propagation direction of the laser beam . in the lower panels of figs . 3 and 4 , we display the cases with interacting laser beams . also in the case of brillouin scattering , the nonlinear dispersion relation predicts a rather weak interaction between the two beams , where the instability regions of the two beams are more or less superimposed without dramatic differences in the growth rates . in order to investigate the nonlinear dynamics of the interacting laser beams in plasmas , we have carried out numerical simulations of the reduced system of equations ( [ eq : nlse - coupled ] ) in two spatial dimensions , and have presented the results in figs . 58 . in these simulations , we have used as an initial condition that either @xmath88 has a constant amplitude of @xmath98 and @xmath89 has a zero amplitude , or that both beams have a constant amplitude of @xmath98 and that they initially have group velocities at a right angle to each other . due to symmetry reasons , it is sufficient to simulate one vector component of @xmath99 , which we will denote @xmath19 ( @xmath100 ) . the background plasma density is slightly perturbed with a low - level noise ( random numbers ) . we first consider stimulated raman scattering , displayed in figs . 5 and 6 . the single beam case in fig . 5 shows a growth of density waves mainly in the direction of the beam , while a standing wave pattern is created in the amplitude of the electromagnetic wave envelope , where maxima in the laser beam amplitude is ( roughly ) correlated with minima in the electron density . this is in line with the standard raman backscattering instability . the simulation is ended when the plasma density fluctuations are large and self - nonlinearities and kinetic effects are likely to become important . in fig . 6 , we show the case with the two beams crossing each other at a right angle . in this case , the wave pattern becomes slightly more complicated with local maxima of the laser beam envelope amplitude correlated with local minima of the electron density . however , this pattern is very regular and there is no clear sign of nonlinear structures in the numerical solution . we next turn to the case of stimulated brillouin scattering , presented in figs . 7 and 8 . in this case , the waves grow not only in the direction of the laser beam but also , with almost the same growth rate , obliquely to the propagation direction of the laser beam . we see in the single beam case , presented in fig . 7 , that the envelope of the ion beam becomes modulated in localized areas both in @xmath90 and @xmath91 directions , and in the nonlinear phase at the end of the simulation , the laser beam envelope has local maxima correlated with local minima of the ion density . for the case of two crossed laser beams , displayed in fig . 8 , we see a more irregular structure of the instability and that at the final stage , local `` hot spots '' are created in which large amplitude laser beam envelopes are correlated with local depletions of the ion density . in summary , we have investigated the instability and dynamics of two nonlinearly interacting intense laser beams in an unmagnetized plasma . our analytical and numerical results reveal that stimulated raman forward and backward scattering instabilities are the dominating nonlinear processes that determine the stability of intense laser beams in plasmas , where relativistic mass increases and the radiation pressure effects play a dominant role . our nonlinear dispersion relation for two interacting laser beams with different wavenumbers predicts a superposition of the instabilities for the single beams . the numerical simulation of the coupled nonlinear schrdinger equations for the laser beams and the governing equations for the slow plasma density perturbations in the presence of the radiation pressures , reveal that in the case of stimulated raman scattering , the nonlinear interaction between the two beams is weaker than for the case of stimulated brillouin scattering . the latter case lead to local density cavities correlated with maxima in the electromagnetic wave envelope . the present results should be useful for understanding the nonlinear propagation of two nonlinearly interacting laser beams in plasmas , as well as for the acceleration of electrons by high gradient electrostatic fields that are created due to stimulated raman scattering instabilities in laser - plasma interactions . 99 p. k. shukla , n. n. rao , m. y. yu , and n. l. tsintsadze , phys . rep . * 135 * , 1 ( 1986 ) . a. sjlund and l. stenflo , appl . 10 * , 201 ( 1967 ) . m. y. yu , k. h. spatschek , and p. k. shukla , z. naturforsch . a * 29 * , 1736 ( 1974 ) . p. k. shukla , m. y. yu , and k. h. spatschek , phys . fluids * 18 * , 265 ( 1975 ) . p. k. shukla and l. stenflo , phys . a * 30 * , 2110 ( 1984 ) . n. l. tsintsadze and l. stenflo , phys . a * 48 * , 399 ( 1974 ) . c. e. max , j. arons , and a. b. langdon , phys . lett . * 33 * , 209 ( 1974 ) . p. k. shukla , phys . scripta * 45 * , 618 ( 1992 ) . l. berg , phys . e * 58 * , 6606 ( 1998 ) . c. ren , b. j. duda , and w. b. mori , phys . e * 64 * , 067401 ( 2001 ) . dong , z .- m . sheng , and j. zhang , phys . e * 66 * , 027402 ( 2002 ) . m. n. rosenbluth and c. s. liu , phys . * 29 * , 701 ( 1972 ) . g. shvets and n. j. fisch , phys . * 86 * , 3328 ( 2001 ) . r. bingham , j. t. mendona and p. k. shukla , plasma phys . fusion * 46 * , r1 ( 2004 ) . g. shvets , phys . lett . * 93 * , 195001 ( 2004 ) . v. i. berezhiani and i. g. murusidze , phys . a * 148 * , 338 ( 1990 ) . f. s. tsung , r. narang , w. b. mori , r. a. fonseca , and l. o. silva , phys . * 93 * , 185002 ( 2004 ) . k. nagashima , j. koga , and m. kando , phys . lett . * 64 * , 066403 ( 2001 ) . sheng , k. mima , y. setoku , k. nishihara , and j. zhang , phys . plasmas * 9 * , 3147 ( 2002 ) . n. l. tsintsadze , d. d. tskhakaya , and l. stenflo , phys . a * 72 * , 115 ( 1979 ) . l. stenflo , phys . scripta * t30 * , 166 ( 1990 ) ; _ ibid _ * t107 * , 262 ( 2004 ) . 1 : the normalized ( by @xmath101 ) growth rates due to stimulated raman scattering ( case i ) for single laser beams ( upper panels ) and for two laser beams ( lower panel ) , as a function of the wave numbers @xmath86 and @xmath87 . the upper left and right panels show the growth rate for beam @xmath102 and @xmath103 , respectively , where the wave vector for @xmath102 is @xmath104 and the one for @xmath103 is @xmath105 , i.e. the two beams are launched in the @xmath90 and @xmath91 directions , respectively . in the lower left panel , @xmath102 and @xmath103 are launched simultaneously at a perpendicular angle to each other , and in the lower right panel , the two beams are counter - propagating . we used the normalized amplitudes @xmath106 and the electron thermal speed @xmath107 . 2 : the normalized ( by @xmath101 ) growth rates due to stimulated raman scattering ( case i ) for single laser beams ( upper panels ) and for two laser beams ( lower panel ) , as a function of the wave numbers @xmath86 and @xmath87 . the upper left and right panels show the growth rate for beam @xmath102 and @xmath103 , respectively , where the wavenumber for @xmath102 is @xmath108 and the one for @xmath103 is @xmath109 . in the lower left panel , two beams are launched at a perpendicular angle to each other , and in the lower right panel , the two beams are counter - propagating . we used the normalized amplitudes @xmath106 and the electron thermal speed @xmath107 . 3 : the normalized ( by @xmath101 ) growth rates due to stimulated brillouin scattering ( case ii ) for single laser beams ( upper panels ) and for two laser beams ( lower panel ) , as a function of the wave numbers @xmath86 and @xmath87 . the upper left and right panels show the growth rate for the beam @xmath102 and @xmath103 , respectively , where the wave number for @xmath102 is @xmath104 and the one for @xmath103 is @xmath105 . in the lower left panel , two beams are launched at a perpendicular angle to each other , and in the lower right panel , the two beams are counter - propagating . we used the normalized amplitudes @xmath106 , the ion to electron mass ratio @xmath110 ( argon ) , and the ion sound speed @xmath111 . fig . 4 : the normalized ( by @xmath101 ) growth rates due to stimulated brillouin scattering ( case ii ) for single laser beams ( upper panels ) and for two laser beams ( lower panel ) , as a function of the wave numbers @xmath86 and @xmath87 . the upper left and right panels show the growth rate for beam @xmath102 and @xmath103 , respectively , where the wavenumber for @xmath102 is @xmath108 and the one for @xmath103 is @xmath109 . in the lower left panel , two beams are launched at a perpendicular angle to each other , and in the lower right panel , the two beams are counter - propagating . we used the normalized amplitudes @xmath106 , the ion to electron mass ratio @xmath110 ( argon ) , and the ion sound speed @xmath111 . fig . 5 : the amplitude of a single laser beam @xmath112 ( left panels ) and the electron density @xmath113 ( right panels ) involving stimulated raman scattering ( case i ) , at times @xmath114 , @xmath115 and @xmath116 ( upper to lower panels ) . the laser beam initially has the amplitude @xmath117 and wavenumber @xmath118 . the electron density is initially perturbed with a small - amplitude noise ( random numbers ) of order @xmath119 . 6 : the amplitude of two crossed laser beams , @xmath120 ( left panels ) and the electron density @xmath113 ( right panels ) involving stimulated raman scattering ( case i ) , at times @xmath114 , @xmath115 and @xmath116 ( upper to lower panels ) . the laser beams initially have the amplitude @xmath121 , and @xmath88 initially has the wavenumber @xmath118 while @xmath89 has the wavenumber @xmath122 . the electron density is initially perturbed with a small - amplitude noise ( random numbers ) of order @xmath119 . fig . 7 : the amplitude of a single laser beam @xmath112 ( left panels ) and the electron density @xmath113 ( right panels ) involving stimulated brillouin scattering ( case ii ) , at times @xmath123 , @xmath124 and @xmath125 ( upper to lower panels ) . the laser beam initially has the amplitude @xmath117 and wavenumber @xmath118 . the ion density is initially perturbed with a small - amplitude noise ( random numbers ) of order @xmath119 . fig . 8 : the amplitude of two crossed laser beams , @xmath120 ( left panels ) and the electron density @xmath113 ( right panels ) involving stimulated brillouin scattering ( case ii ) , at times @xmath114 , @xmath115 and @xmath116 ( upper to lower panels ) . the laser beams initially have the amplitude @xmath121 , and @xmath88 initially has the wavenumber @xmath118 while @xmath89 has the wavenumber @xmath122 . the electron density is initially perturbed with a small - amplitude noise ( random numbers ) of order @xmath119 .

we investigate the nonlinear interaction between two laser beams in a plasma in the weakly nonlinear and relativistic regime . the evolution of the laser beams is governed by two nonlinear schrdinger equations that are coupled with the slow plasma density response . we study the growth rates of the raman forward and backward scattering instabilities as well of the brillouin and self - focusing / modulational instabilities . the nonlinear evolution of the instabilities is investigated by means of direct simulations of the time - dependent system of nonlinear equations .

z.w . , t.y . , and magnetotransport studies in georgia state university have been supported by the u.s . department of energy , office of basic energy sciences , material sciences and engineering division under de - sc0001762 . additional support has been provided by the aro under w911nf-14 - 2 - 0076 . measurements were done by z.w . technical assistance was given by t.y . experimental development and manuscript was written by z.w . and r.g.m . the authors declare no competing financial interests .

this study examines the magnetotransport response observed in flakes of the @xmath0d topological insulator bi@xmath1te@xmath2 , including indium superconducting electrodes , and demonstrates two critical transitions in the magnetoresistive response with decreasing temperatures below @xmath3 k. the first transition is attributed to superconductivity in the indium electrodes , and the second transition , with a critical field exceeding the transition field of indium , is attributed to a proximity effect at the 2d planar interface of this hybrid ti / superconductor structure . topological insulators ( ti ) are 2- and 3-dimensional electronic materials that behave as a normal insulator in the bulk but conduct electricity in edge or surface states , respectively , that arise from a bulk - boundary correspondence , and are topologically protected by time - reversal symmetry.@xcite ti are characterized by a strong spin - orbit coupling that provides a locking between spin and linear momentum , and the helical spin polarization as gapless surface states exhibit a linear dispersion relation , i.e. , `` dirac cones , '' characteristic of relativistic massless particles . the @xmath4 quantum well system was the first theoretically predicted@xcite and experimentally demonstrated@xcite 2d ti , followed by the prediction@xcite and experimental confirmation@xcite of the technological significant @xmath5 quantum well system as another 2d ti . @xmath6 was the first theoretically predicted- and experimentally identified-3d ti.@xcite since then , a large number of other systems with a band - gap @xmath7 at @xmath8 k including @xmath9,@xcite @xmath10,@xcite and @xmath11,@xcite have joined the class of 3d ti . remarkably , among these systems , @xmath9 has exhibited superconductivity with a critical transition temperature of @xmath12 k , when copper is intercalated between the adjacent quintuple layers.@xcite further , the application of pressure in the range of 3 - 6 gpa in @xmath11 has been shown to induce superconductivity with @xmath13 k and @xmath14k@xmath15 t.@xcite the proximity effect from an ordinary s - wave superconductor in the ti leads to exotic @xmath16 superconductivity capable of hosting majorana fermions , and this prediction has heightened interest in experimental studies of superconductor - topological insulator hybrid devices.@xcite thus , many have examined ti nanowire - superconductor devices in search of majorana modes at the ends of one dimensional devices.@xcite nevertheless , a two - dimensional experimental setting for the study of majorana modes is desirable . in ref . @xcite , superconductivity is induced in the surface state of the ti by the proximity effect produced by an ordinary s - wave superconductor , and the majorana mode is a vortex bound state at the interface of the two materials . recently , there has been interest in whether in systems now called `` topological metals , '' which include 3d ti materials such as @xmath11 and @xmath9 that have bulk states coexisting with surface states at the fermi energy it is possible for the proximity effect to extend from the superconductor covered surfaces to uncovered surfaces of the 3d ti@xcite and whether the uncovered surfaces of the 3d ti can thus host majorana modes.@xcite in moving towards exploring such possibilities , we have examined the magnetotransport response in the @xmath0d topological insulator @xmath11 in the presence of ( indium ) topside superconducting electrodes.@xcite remarkably , the experimental results showed two critical transitions in the magnetoresistive response below @xmath17 k. while the first transition is associated with a shunting effect by the contacts , the second anomalous contribution , which indicates @xmath18 k and @xmath19k@xmath20 t , is attributed here to a generalized proximity effect in the ti . @xmath11 flakes ( @xmath21 thick ) were mechanically exfoliated using a scotch tape method from single crystals of @xmath11 and transferred onto @xmath22 substrates . the @xmath11 flakes were approximately shaped like rectangles with the length - to - width ratio @xmath23 . superconductor / topological insulator junctions were realized by directly pressing indium onto the surface of the @xmath11 flake in a hall bar configuration ; see supplementary material for the images of the samples.@xcite here , indium is a s - wave superconductor with the critical temperature @xmath24 k.@xcite electrical measurements were carried out using the standard four - terminal low - frequency lock - in techniques , and all the reported resistances are four - terminal resistances . the _ ac _ was applied along the long axis of the hall bar , through electrodes at the two ends , and the magnetic field was applied perpendicular to the hall bar surface , as usual . since the flakes are not so thin , the applied current is carried by both the specimen bulk and the specimen surfaces . further , since the proximity effect should be restricted to the near surface regions , one expects to observe just resistance corrections due to the proximity effect in the measured four terminal resistances . that is , one does not expect the four terminal resistance to vanish even with the onset of superrconductivity in the contacts and the proximity effect in the surfaces near the contacts . the specimens were immersed in pumped liquid helium , and the temperature was varied in the range @xmath25 k by controlling the vapor pressure of liquid helium within a @xmath26he cryostat . hall effect measurements indicated the carrier concentration @xmath27 in these samples . fig . 1a shows a color plot of the normalized magnetoresistance , @xmath28 as a function of both the magnetic field , @xmath29 , and the temperature , @xmath30 , for sample 1 . here , @xmath31/r(b_n)$ ] , where @xmath32 t. the color plot ( fig . 1(a ) ) shows a set of horizontal and vertical lines , which correspond to constant @xmath29- and @xmath30-cross - sections , which are exhibited in fig . 1(b ) and fig . 1(c ) , respectively . in the @xmath33 k cross - section , shown in light green ( see fig . 1(b ) ) , sample @xmath34 shows a featureless @xmath28 , with a small positive magnetoresistance of @xmath35 t.@xcite however , at @xmath36 k , which is slightly below the superconducting transition temperature of indium , @xmath28 exhibits a sharp and narrow dip near zero magnetic field . as the temperature is reduced further , this dip in @xmath37 rapidly becomes deeper until @xmath38 k , as its width along the @xmath29-axis , denoted here as @xmath39 , increases . a close examination of @xmath28 suggests the emergence of a second @xmath28 contribution for @xmath40 k in this specimen . the width but not the depth of this second contribution to @xmath28 grows rapidly with decreasing @xmath30 . at @xmath41 k , the critical magnetic field is denoted as @xmath42k@xmath43 t. fig.1(c ) depicts the magnitude of @xmath28 vs. @xmath30 at the constant-@xmath29 cross - sections shown in fig . 1(a ) . this figure shows that , with decreasing temperatures , the @xmath28 drops within a small @xmath44 interval and then plateaus . the @xmath44 interval where the drop is observed depends upon @xmath29 . the temperature dependence of the critical fields associated with the two contributions , i.e. , @xmath45 and @xmath46 , are shown in fig.1 ( d ) . here , the critical fields were determined by examining the first derivative of the magnetoresistance data and identifying the magnetic fields where the slope , @xmath47 , exhibits a substantial change . the results of fig . 1(d ) indicate an approximately linear relationship between the critical fields and temperature for the two contributions to @xmath28 . * transport in bi@xmath1te@xmath2 flakes including indium contacts . * ( a ) a 2d color plot of the normalized magnetoresistance , @xmath28 , is shown vs. the magnetic field , @xmath29 , and the temperature , @xmath30 , for a @xmath11 specimen , sample @xmath34 . ( b ) this panel exhibits the normalized magnetoresistance vs. @xmath29 , at the @xmath30-cross - sections indicated in panel ( a ) . ( c ) this panel shows the temperature - dependence of @xmath28 at zero magnetic field . ( d ) the critical fields @xmath45 and @xmath46 are plotted on the left and right , respectively , vs. @xmath30 . data are shown as symbols . lines represent the trend.,width=264 ] the results for a second @xmath11 specimen are shown in fig . 2 . the top panel in fig . 2 shows a color plot of @xmath28 vs. @xmath29 and @xmath30 , with five horizontal- and three vertical - cross sections . the data corresponding to the horizontal - cross sections are shown in fig . 2(b ) . here , @xmath37 shows an initial negative magnetoresistance followed by a weak positive magnetoresistance at @xmath48 k(see fig.2(b ) ) . fig . 2 ( b ) also shows a rapid dip in @xmath28 beginning at @xmath49 k in the vicinity of @xmath50 . a second resistance correction term is readily apparent by @xmath51 k , and this term mainly gains width but not depth with the decreasing temperature . at @xmath52 k , the critical magnetic field is denoted as @xmath53k@xmath54 t in this specimen . the temperature dependences of the critical fields associated with the two contribution , i.e. , @xmath45 and @xmath46 , are shown in fig . 2(d ) . the results indicate an approximately linear increase in the critical fields with the decreasing temperatures . the vertical ( constant magnetic field ) cross sections indicated in fig . 2(a ) are shown in fig . 2(c ) . as in fig . 1(c ) , fig.2(c ) shows that , with decreasing temperatures , the @xmath28 drops within a small @xmath44 interval and then plateaus . the @xmath30-band where the drop is observed depends again upon @xmath29 . * transport in bi@xmath1te@xmath2 flakes including indium contacts . * ( a ) a 2d color plot of the normalized magnetoresistance , @xmath28 , is shown vs. the magnetic field , @xmath29 , and the temperature , @xmath30 , for a @xmath11 specimen , sample @xmath55 . ( b ) this graph exhibits the normalized magnetoresistance vs. @xmath29 , at the @xmath30-cross - sections indicated in the top panel . ( c ) this panel compares the temperature dependence of the normalized resistance at @xmath56 mt , @xmath57 mt , and @xmath58 mt . ( d ) the critical fields @xmath45 and @xmath46 are plotted on the left and right , respectively , vs @xmath30 data are shown as symbols . lines represent the trend.,width=264 ] * transport in bi@xmath1te@xmath2 flakes including indium contacts . * ( a ) this figure shows a 2d color plot of normalized magnetoresistance , @xmath28 , vs. the magnetic field , @xmath29 , and the temperature , @xmath30 , for a @xmath11 specimen , sample @xmath0 . ( b ) the normalized magnetoresistance , @xmath28 , vs. @xmath29 , is shown at the different @xmath30 cross - sections indicated in the top panel . ( c ) this panel compares the @xmath30-dependence of the normalized resistance at the magnetic field cross - sections at @xmath50 mt , @xmath59 mt , and @xmath58 mt . , width=264 ] a color plot of normalized magnetoresistance for a third @xmath11 specimen , sample @xmath0 , is shown in fig . 3(a ) . the data corresponding to the horizontal - cross sections are shown in fig . 3(b ) . here , @xmath28 shows predominantly negative magnetoresistance at the highest temperatures . fig . 3 ( b ) also shows a small sharp dip in @xmath28 in the @xmath60 k data trace in the vicinity of @xmath50 . the dip quickly reaches its maximum depth near @xmath61 k. with a further reduction in @xmath30 , the resistance correction rapidly increases in width . the constant magnetic field cross sections indicated in fig . 3(a ) are shown in fig . 3(c ) . as for specimens 1 and 2 , the @xmath28 drops within a small @xmath44 interval and then plateaus . * possible scenario for observed effects . * a traditional proximity effect based explanation for the observed effects : ( a ) a sketch of a bi@xmath1te@xmath2 flake with superconducting indium contacts . ( b ) a bi@xmath1te@xmath2 flake with a bulk current in the normal state above the contact superconducting transition critical temperature , @xmath62 . ( c ) below the @xmath62 , some additional fraction of the current @xmath63 is shunted through the contacts when the contacts are in the superconducting state . ( d ) below the critical temperature , @xmath64 , associated with the proximity effect , cooper pairs on the ti side of the interface provide an additional shunt for the current . , width=245 ] this magnetotransport study of @xmath11 with indium contacts shows the following features : ( a ) above the superconducting transition temperature of indium , the specimens show either a weak positive ( fig . 1 ( b ) ) , mixed type ( fig . 2(b ) ) , or weak negative ( fig . 3(b ) ) magnetoresistance . this aspect will be examined in greater detail elsewhere . at the moment , we attribute the observed differences in the normal state magnetoresistance characteristics to variations in the concentration of tellurium vacancies in the bulk crystals utilized to realize these specimens . ( b ) below the transition temperature of indium , @xmath65 k,@xcite , there is a sharp drop in the resistance in the vicinity of @xmath50 , which grows rapidly with decreasing temperatures down to @xmath66 k. the critical magnetic field associated with this resistance correction is @xmath67k@xmath68 , consistent with the known critical field for indium , @xmath69k@xmath70 t.@xcite ( c ) . below @xmath71 k , there appears an additional correction in @xmath28 which mainly affects the width of the resistance anomaly with the decreasing temperatures . the critical magnetic field associated with this term @xmath72 t easily exceeds the critical field of indium . ( d ) constant magnetic field cross sections show that the drop in the resistance occurs over a narrow temperature range ( see fig . 1(c ) , 2(c ) , and 3(c ) ) , and this is followed by an apparent plateau in the temperature dependence of @xmath28 . due to the correlation of the observed effects with superconductivity in the contacts , one may , at the outset , rule out weak localization and electron - interactions as the origin of the observed magnetoresistance anomalies.@xcite when a normal metal ( n ) is brought into good contact with a superconductor ( s ) , cooper pairs from the superconductor can leak into the normal metal and help to manifest signs of superconductivity in n , while weakening the superconductivity in s.@xcite the spatial extent of the proximity effect in n depends also on the presence of impurities and disorder dependent dephasing mechanisms in n.@xcite thus , andreev reflection theory encompasses both the ideal ballistic n - s interfaces , as well as real - world n - s systems in the diffusive limit ( see , for example , ref.@xcite ) . one possible image for the observed effects is presented in fig 4 . the voltage and current probes ( see fig . 4(a ) ) used in these experiments consist of the s - wave superconductor indium . fig . 4(b ) illustrates the four terminal transport measurement at @xmath73 . in this high temperature condition , the voltage probes behave as normal metals . further , the applied current is transported by the surface states of the ti as well as by the bulk of @xmath11 . when the temperature is reduced to @xmath74 , the current and voltage probes become superconducting ( see fig . 4(c ) ) , and some additional current next to the surface is shunted through the contacts with the onset of superconductivity in indium , and the contact shunt current path is favored over the current path within the ti . this effect explains the resistance drop associated with @xmath39 in the @xmath28 data . a further decrease in the temperature could result in a leakage of cooper pairs into the topological insulator ( see fig . 4(d ) ) , which can be characterized by a critical temperature @xmath75 . if this proximity superconductivity layer shunts additional current , there will be another correction in the resistance , and this could provide for the observed broad resistance correction at the lowest temperatures in our specimens . there is , however , the observable feature that the critical field @xmath76 is as large as 0.2 t at the lowest temperatures . this @xmath46 exceeds by nearly a factor - of - ten the critical field of indium , @xmath77k@xmath68 . from our understanding , it is not possible to have a proximity effect in the neighboring normal metal above the critical field of the superconductor . note , however , the reported result that @xmath11 exhibits superconductivity under the application of pressure in the range of 36 gpa.@xcite further , ref.@xcite indicates that induced superconductivity in @xmath11 tends to exhibit a @xmath13 k and @xmath78 t.@xcite remarkably , our data also indicate a @xmath79 k and @xmath80 t. this point suggests that the second critical transition associated with @xmath46 could originate from a contact induced surface modification in the @xmath11 flake , which mimics the effect of pressure , but only in the immediate area under the contact . 0 m. z. hasan , and c. l kane , rev . mod . phys . 82 , 3045 ( 2010 ) . x. l. qi , and s. c. zhang , rev . mod . phys . 83 , 1057 ( 2011 ) . y. ando , j. phys . soc . jpn . 82 , 102001 ( 2013 ) . b. a. bernevig , t. l. hughes , and s. c. zhang , science 314 1757 ( 2006 ) . m. konig , s. weidmann , c. brune , a. roth , and h. buhmann , science 318 , 766 ( 2007 ) . l. fu , and c. l. kane , phys . rev . b 76 , 045302 ( 2007 ) . d. hsieh , d. qian , l. wray , y. xia , y. s. hor , r. j. cava , and m. z hasan , nature 452 , 970 ( 2008 ) . c. liu , t. l. hughes , x. l. qi , k. wang , and s. c. zhang , phys . rev . lett . 100 , 236601 ( 2008 ) . i. knez , r. r. du , and g. sullivan , phys . rev . lett . 107 , 136603 ( 2011 ) . y. xia , d. qian , d. hsieh , l. wray , a. pal , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava , and m. z. hasan , nat . phys . 5 , 398 ( 2009 ) . y. jiang , y. wang , m. chen , z. li , c. song , k. he , l. wang , x. chen , x. ma , and q - k . xue , phys . rev . lett . 108 016401 ( 2012 ) . y. l. chen , j. g. analytis , j - h . chu , z. k. liu , s. k. mo , x. l. qi , h. j. zhang , d. h. lu , x. dai , z. fang et al . , science 325 , 178 ( 2009 ) . d. hsieh , y. xia , d. qian , l. wray , f. meier , j. h. dil , j. osterwalder , l. patthey , a. v. fedorov , h. lin , et al . , phys . rev . lett . 103 146401 ( 2009 ) . y. s. hor , a. j. williams , j. g. checkelsky , p. roushan , j. seo , q. xu , h. w. zandbergen , a. yazdani , n. p. ong , and r. j. cava , phys . rev . lett . 104 , 057001 ( 2010 ) . s. sasaki , m. kriener , k. segawa , k. yada , y. tanaka , m. sato , and y. ando , phys . rev . lett . 107 , 217001 ( 2011 ) . j. l zhang , s. j. zhang , h. m. weng , w. zhang , l. x. yang , q. q. liu , s. m. feng , x.c . wang , r. c. yu , l. z. cao et al . , proc . natl . acad . sci . u.s.a . 108 , 24 - 28 ( 2011 ) . l. fu , and c. l. kane , phys . rev . lett . 100 , 096407 ( 2008 ) . c. nayak , s. h. simon , a. stern , m. freedman , and s. das sarma , rev . mod . phys . 80 , 1083 - 1159 ( 2008 ) . v. mourik , k. zuo , s. m. frolov , s. r. plissard , e. p. a. m. bakkers , and l. p. kouwenhoven , science 336 , 1003 ( 2012 ) . a. das , y. ronen , y. most , y. oreg , m. heiblum , and h. shtrikman , nat . phys . 8 , 887 ( 2012 ) . l. p. rokhinson , x. liu , and j. k. furdyna , nat . phys . 8 , 795 ( 2012 ) . m. t. deng , c. l. yu , g. y. huang , m. larsson , p. caroff , and h. q. xu , nano lett . 12 , 6414 ( 2012 ) . w. chang , v. e. manucharyan , t. s. jespersen , j. nygard , and c. m. marcus , phys . rev . lett . 110 , 217005 ( 2013 ) . a. d. k. finck , d. j. van harlingen , p. k. mohseni , k. jung , and x. li , phys . rev . lett . 110 , 126406 ( 2013 ) . m. x. wang , h. ding , a. v. fwdorov , w. yao , z. li , y. f. lv , k. zhao , l. g. zhang , z. xu , j. schneeloch et al . , science 336 , 52 - 55 ( 2011 ) . k. lee , a. vaezi , m. h. fischer , & e. a. kim , phys . rev . b 90 , 214510 ( 2014 ) . r. g. mani , l. ghenim , and t. theis , phys . rev . b 45 , 9877 ( 1992 ) . l. ghenim , and r. g. mani , appl . phys . lett . 60 , 2391 ( 1993 ) . d. zhang , j. wang , a. m. dasilva , j. s. lee , h. r. gutierrez , moses h. w. chan , j. jain , and n. samarth , phys . rev . b 84 , 165120 ( 2011 ) . f. yang , y. ding , f. qu , j. shen , j. chen , z. wei , z. ji , g. liu , j. fan , c. yang et al . , phys . rev . b 85 , 104508 ( 2012 ) . m. veldhorst , m. snelder , m. hoek , t. gang , v. k. guduru , x. l. wang , u. zeitler , w. g. van der wiel , a. a. golubov , h. hilgenkamp , and a. brinkman , nature mater . 11 , 417 ( 2012 ) . f. qu , f. yang , j. shen , j. chen , z. ji , g. liu , j. fan , x. jing , c. yang , and l. lu , scientific reports 2 , 339 ( 2012 ) . see supplementary material at http://dx.doi.org/10.1063/1.4934871 for sample - pictures and magnetotransport data examining the normal states . see http://www.superconductors.org/type1.htm for type 1 superconductor properties and a periodic chart comparison ( 2007 ) . m. tinkham , _ introduction to superconductivity_. 2nd edn ( mcgraw hall , new york , 1996 ) . g. bergmann , phys . rep . 107 , 1 ( 1984 ) . m. liu , j. zhang , c. chang , z. zhang , x. feng , k. li , k. he , l. wang , x. chen , x. dai , et al . , phys . rev . lett . 108 , 036805 ( 2012 ) . r. g. mani , a. kriisa , and w. wegscheider , sci . rep . 3 , 2747 ( 2013 ) . r. g. mani , l. ghenim , and j. b choi , phys . rev . b43 , 12630 ( 1991 ) . g. blonder , m. tinkham , and t. m. klapwijk , phys . rev . b 25 , 4515 ( 1982 ) . t. m. klapwijk , j. supercond : incorp . novel magnetism , 17 , 593 ( 2004 ) . s. gueron , h. pothier , n. o. birge , d. esteve , and m. h. devoret , phys . rev . lett . 77 , 3025 - 3028 ( 1996 ) .

recently , the perovskite - type oxides , which display ferroelectric and magnetic properties , have attracted considerable interest for their potential applications in multifunctional magnetoelectric devices@xcite . it has been shown that @xmath1 is a good candidate which exhibits the coexistence of ferroelectric and antiferromagnetic orders with high curie temperature@xcite and n@xmath2el temperature@xcite . numerous studies have been performed on @xmath1 samples and especially , more recently , on thin films . generally , bottom and top electrodes are fabricated to form a capacitor configuration for ferroelectric measurements . the physical origin of the interfacial layer has been proposed to be process - induced domain evolution or the effect of lattice misfit between the electrode and surface of the film . the interfaces in @xmath1 film capacitors have already been verified by using x - ray diffraction and scanning transmission electron microscope high - angle annular dark - field imaging@xcite , and the effect of the interface on the dielectric behavior of @xmath1 film capacitor was also investigated in experiment@xcite . for epitaxial system , the substrate - induced strains have been proven to have great influence on the ferroelectric properties of ferroelectric material@xcite . the previous studies show that substrate - induced strains can lead to a substantial increase in spontaneous polarization for bfo films@xcite . the mechanical substrate effects and thickness dependence of ferroelectric properties in epitaxial bfo films are investigated within the framework of landau - devonshire theory . although the interface between the electrode and the bfo layer has been reported in experiment@xcite , there is no theoretical investigation of the interfacial effect on the polarization of bfo film . in this paper , the interfaces between electrode and @xmath1 layer are considered with different film thickness . it is found that the interface will have great influence on the polarization and the hysteresis behavior of ultrathin @xmath1 films . due to the interfacial effect , the maximum polarization on the p - e loop of thin film can be lower(or higher ) than that of thickness film , and the maximum polarization can also be thickness - independent through the competition between the interfacial effect and the compressive stress under an adequate applied field . here , we consider a @xmath1 film grown on @xmath3 substrate with a bottom @xmath4 electrode and a top @xmath5 electrode@xcite . it is known that there is a clear and sharp interface between @xmath1 layer and the bottom @xmath4 electrode , and a rough , inter - diffused interface between @xmath1 layer and the top @xmath5 electrode , which can be confirmed by the cross - section hr - tem image@xcite . it is reasonable that the interface between bfo and the bottom electrode can be neglected for simplicity . therefore , we simulate the @xmath1 film by a ferroelectric layer with a top non - switching interface , and investigate the effects of the interface on the p - e loop with different film thickness . the electric displacement across bfo layer is given by @xmath6 , where @xmath7 , @xmath8 and @xmath9 denote the permittivity , the local electric field and the spontaneous polarization of bfo layer , respectively . the electric displacement across the interface is @xmath10 , where @xmath11 and @xmath12 stand for the permittivity and the local electric field of interface . the applied field @xmath13 can be written as @xmath14 where @xmath15 is the thickness ratio of interface . the continuity of the total current density across the film can be expressed as @xmath16 combining eq . ( 1 ) with eq . ( 2 ) , we obtain @xmath17e_{f}\nonumber\\ & & + [ v(\epsilon_{f}+\chi_{f})+\epsilon_{in}(1-v)]\dot{e}_{f},\end{aligned}\ ] ] where @xmath18 and @xmath19 are the conductivities of top interface and @xmath1 layer , respectively , @xmath20 is the electric susceptibility of @xmath1 layer , which equals @xmath21@xcite @xmath22\}^{-1},\\ \delta&=&e_{c}[log(\frac{1+p_{fr}/p_{fs}}{1-p_{fr}/p_{fs}})]^{-1},\nonumber\end{aligned}\ ] ] where @xmath23 , @xmath24 and @xmath25 denote the remanent polarization , saturate polarization and coercive field of @xmath1 layer , respectively , which can be determined through landau - devonshire theory@xcite . the @xmath1 films with different thickness are supposed to grow on ( 001)-oriented @xmath3 substrate . a tetragonal structure of @xmath1 thin film based on cubic perovskite structure with symmetry lowered to p4mm(neglect the small monoclinic distortion of about @xmath26 of c axis ) is suggested with an elongated c axis@xcite . the polarization for such a single phase is along with @xmath27 axis . the thermodynamic potential of pseudo cubic @xmath1 film is expressed as the function of magnetization @xmath28 , polarization @xmath9 , temperature @xmath29 and misfit strain @xmath30 , where @xmath31 is the in - plane lattice parameter for substrate and @xmath32 is the equivalent cubic lattice constant of free standing film . we do not include the magnetic effect and the coupling between magnetic order and ferroelectric order , owing to a relative weak magnetic order compared to the stronger ferroelectric one . the thermodynamic potential of @xmath1 layer can be written as following , @xmath33 where @xmath34 , @xmath35 , @xmath36 is the electrostricitive coefficient , @xmath37 is the elastic compliances , and the dielectric stiffness coefficient @xmath38 . the equilibrium state polarization @xmath9 under a local field @xmath8 , @xmath23 and @xmath24 in eq . ( 4 ) can be numerically resolved from eq . ( 5 ) by @xmath39 , @xmath40 in order to introduce the misfit strain into the epitaxial @xmath1 film , an effective substrate lattice parameter @xmath41 is defined as @xmath42 , which is used to calculate the misfit strain @xmath43 instead of actual substrate lattice parameter @xmath31 , and @xmath44 . @xmath45 is the equilibrium linear misfit dislocation density at the deposition temperature @xmath46 . @xmath47 , where @xmath48 is the critical film thickness corresponding to the generation of misfit dislocation . all the coefficients in landau expansion used in this letter are given in ref . from eq . ( 6 ) , we calculate the spontaneous polarization @xmath9 without considering the interface . the hysteresis behavior of the single @xmath1 layer is shown in fig . 1 , where solid line stands for the hysteresis loop of the 77-nm - thick film , and dotted line for the 960-nm - thick film . the remanent polarization of the @xmath49 nm @xmath1 film is @xmath50 , which is close to the polarization of single crystal@xcite . we can see that the spontaneous polarization increases as the thickness decreases , and our results are qualitatively consistent with previous theoretical works@xcite . it is obviously that the compressive stress between @xmath1 layer and substrate @xmath3 will remarkable increase the polarization , and the spontaneous polarization @xmath9 will increase with the decrease of thickness . however , as observed by afm@xcite and simulated by the capacitor configuration@xcite , there really exists an interface between top electrode and @xmath1 layer . so , the measured polarization is not the polarization of a single ferroelectric layer . in the following , an interface between the top @xmath5 electrode and @xmath1 layer is introduced , and two @xmath1 film samples of thickness @xmath51 nm and @xmath49 nm are selected in our simulation@xcite . the dielectric constant and the conductivity of the @xmath1 layer are selected as 100@xcite and @xmath52@xcite , respectively . due to the interface is formed between the metal material and @xmath1 layer , we assume its conductivity and dielectric constant take the values of @xmath53@xcite and @xmath54 . the thickness of interface is introduced by taking into account the thickness ratio of interface to @xmath1 layer . the ratio @xmath55 will increase as the decrease of film thickness , and the ratios @xmath55 are taken as @xmath56 and @xmath57 for the two samples . for a given sinusoidal field @xmath13 , the effective field @xmath8 can be calculated from eq . ( 3 ) as a function of time @xmath58 . then the polarization @xmath9 at the @xmath59th increment in time is calculated by using @xmath60 . the simulated hysteresis loops corresponding to different applied fields are shown in fig . 2 , the frequency of the applied field in our simulation is @xmath61khz . the solid lines in fig . 2 denote the loops of 77-nm - thick film , and the dotted lines stand for the loops of 960-nm - thick film . from fig . 2(a ) to ( d ) , the maximum field are selected as @xmath62 mv / cm , @xmath63 mv / cm , @xmath64 mv / cm and @xmath65 mv / cm , respectively . by taking into account the interface , the local field @xmath8 will decrease as the decrease of thickness under the same applied field . for the 960-nm - thick film , due to the small proportion of the interface component , the local field @xmath8 is close to the applied field , and the p - e loop is already saturated in fig . 2(a ) , whereas the p - e loop for 77-nm - thick film does not saturated , and the maximum polarization on the p - e loop is smaller than the saturated polarization of 960-nm - thick film . one can see , from fig . 1 and fig . 2(a ) , that the interface will great suppress the polarization of the ultrathin film , and the interfacial effect of an ultrathin film is more important than the internal stress effect under a low applied field . for a selected applied field @xmath66=@xmath63 mv / cm in fig . 2(b ) , the maximum polarization on the p - e loop is independent on the thickness of the film , which is the consequence of the competition between the interfacial effect and the internal compressive stress effect . in fig . 2(c ) , the maximum polarization of the ultrathin film is larger than that of the thick film , which had also been observed in experiment(see fig . 4(d ) in ref . @xcite ) , and it shows that the interfacial effect is less dominant for the ultrathin film under a high applied field . as further increasing the maximum field , the polarizations for the two samples are both saturated , and the saturated polarization of the 77-nm - thick film is larger than that of the 960-nm - thick film . in addition , the p - e loops for materials such as @xmath1 film are often dominated by leakage current , and the saturated p - e curve in fig . 2(d ) for an ultrathin film can only be observed for samples with high resistivity and high breakdown field . in summary , the epitaxial @xmath67 film is simulated by a multilayer model which takes the effect of interface . it is found that the existence of interfacial layer will increase the coercive field , and change the shape of the p - e loop , especially for an ultrathin film . the p - e loop will become more slant for the ultrathin film , which is resulted from the decreasing of the effective field in @xmath1 layer . for different maximum fields , the maximum polarization of the ultrathin film can be smaller(or larger ) than that of the thick film . for an adequate applied field , the maximum polarization on the p - e loop could be thickness - independent , and the origin is caused by the competition between the interfacial effect and the internal compressive stress . it is concluded that the effect of interface between the electrode and @xmath1 layer has great influence on the polarization of @xmath1 film . when measuring the saturated p - e loops for multiferroic ultrathin films with metal contacts , one should apply a field large enough to saturate the polarization . 99 w. eerenstein , n. d. mathur , j. f. scott , _ nature_*442 * , 759(2006 ) . n. venevtsev , g. zhadanov , and s. solovev , _ phys . crystallogr._*4 * , 538(1960 ) . p. fischer , m. polomskya , i. sosnowska , and m. szymanksi , _ c_*13 * , 1931(1980 ) . yi - hsien lee , jenn - ming wu , yu - lun chuen , and li - jen chou , _ appl . phys . lett._*87 * , 172901 ( 2008 ) . guo - zhen liu , can wang , chun - chang wang , jie qiu , meng he , jie xing , kui - juan jin , hui - bin lu , and guo - zhen yang , _ appl . phys . lett._*92 * , 122903(2008 ) . f. tsui , m. c. smoak , t. k. nath . and c. b. eom,_appl . phys . lett._*76 * , 2421(2000 ) . hua ma , lang chen , junling wang , j. ma , and f. boey , _ appl . lett._*92 * , 0182902(2008 ) . q. jiang , and j. h. qiu , _ j. appl . phys . _ * 99 * , 103901(2006 ) . j. x. zhang , y. l. li , y. wang , z. k. liu , l. q. chen , y . h. chu , f. zavaliche , and r. ramesh , _ phys . _ * 101 * , 114105(2007 ) . y. w. cho , s. k. choi , and g. venkata rao , _ appl . _ * 86 * , 202905(2006 ) . w. eerenstein , f .d . morrison , j. dho , m. g. blamire , j. f. scott , n. d. mathur , science 307 , 1203a ( 2005 ) dae ho kim , ho nyung lee , michael d. biegalski , and hans m. christen , _ appl . lett . _ * 92 * , 012911 ( 2008 ) . s. l. miller , r. d. nasby , j. r. schwank , m. s. rodgers , and p. v. dressendorfer , _ j. appl . phys . _ * 68 * , 6463(1990 ) . j. f. li , j. wang , m. wuttig , r. ramesh , n. wang , b. ruette , a. p. pyatakov , a. k. zvezdin , and d. viehland , _ appl . lett . _ * 84 * , 5261(2004 ) . list of parameters in the calculation of spontaneous polarization in eq . ( 6 ) are selected as@xcite . @xmath68 , t=77k , @xmath69 , @xmath70 , @xmath71 , @xmath72 , @xmath73 , @xmath74 , @xmath75 , @xmath76 and @xmath77 for bfo thin film , and @xmath78 for @xmath3 substrate . d. lebeugle et al . , arxiv : 0706.0404 . figure captions + fig1 : polarization of @xmath1 layer as a function of the applied field without considering the interface . + fig2 : p - e loop of @xmath1 film with interface for different applied fields . solid lines correspond to the 77-nm - thick film , and the dotted lines correspond to the 960-nm - thick film .

by considering an interfacial layer between the electrode and the @xmath0(@xmath1 ) layer , the polarization and the hysteresis behavior of @xmath1 film are simulated . it is found that the non - ferroelectric interface will increase the coercive field , and remarkably suppress the polarization of the ultrathin film under low applied fields . due to the competition between the interfacial effect and the internal compressive stress , the maximum polarization on the p - e loop of a @xmath1 film can be independent on the film thickness under an adequate applied field .

hadronic matrix elements of flavor - changing currents are of paramount importance to the phenomenology of weak decays . they appear , e.g. , in the theoretical description of semileptonic @xmath2 decays . these processes give direct access to the elements @xmath4 and @xmath1 of the quark - mixing matrix , provided the hadronic matrix elements are known from some nonperturbative computation . for inclusive semileptonic decays , the matrix elements can be calculated using a short - distance expansion in the large @xmath5-quark mass . improving the theoretical accuracy in the determinations of @xmath1 is currently one of the most important challenges in heavy - flavor theory . relevant to the determination of @xmath1 are hadronic matrix elements of the weak current @xmath6 . in general , local current operators composed of a heavy quark @xmath7 and a light quark @xmath8 exhibit an interesting behavior under renormalization at scales below the heavy - quark mass . large logarithms of the type @xmath9 arise from the exchange of gluons that are hard with respect to the light quark but soft with respect to the heavy quark . since such gluons see the heavy quark as a static color source , their effects can be investigated systematically in the framework of the heavy - quark effective theory ( hqet ) , which provides a systematic expansion of matrix elements in inverse powers and logarithms of @xmath10 @xcite . in the hqet , the four - component heavy - quark field is replaced by a velocity - dependent two - component field @xmath11 satisfying @xmath12 , where @xmath13 is the four - velocity of the hadron containing the heavy quark . operators in the effective theory have a different evolution than in usual qcd . for instance , whereas the vector current @xmath14 is conserved in qcd ( i.e. , its anomalous dimension vanishes ) , the corresponding current @xmath15 in the hqet has a nontrivial anomalous dimension , which governs its renormalization - group ( rg ) evolution for scales below @xmath10 . as an example , consider the heavy - quark expansion of the vector current ( neglecting the light - quark mass @xmath16 ) @xcite @xmath17 where the symbol @xmath18 is used to indicate that this is a relation that holds after taking matrix elements on both sides . ( in a regularization scheme with anticommuting @xmath19 , the expansion of the axial - vector current is obtained by replacing @xmath20 in the hqet operators . ) @xmath21 and @xmath22 are the effective operators entering at leading order in the expansion . the operators @xmath23 appearing at next - to - leading order ( nlo ) form a basis of dimension-4 operators in the hqet that closes under renormalization . the standard choice of these operators is ( using the notation @xmath24 and @xmath25 ) @xmath26 and @xmath27 in addition to the local operators @xmath28 , there appear four bilocal operators @xmath29 built from time - ordered products of the leading - order currents with an insertion of the `` kinetic operator '' @xmath30 or the `` chromo - magnetic operator '' @xmath31 . these operators appear at order @xmath0 in the effective lagrangian of the hqet . part of their effect is to take into account the expansion of the external hadron states in terms of states of the effective theory @xcite . the wilson coefficients of these nonlocal operators are given by the products of the coefficients of their component operators , i.e. , @xmath32 where @xmath33 is the coefficient of the chromo - magnetic operator in the hqet lagrangian , which has been calculated at nlo in @xcite . the kinetic operator is not renormalized @xcite . the calculation of the wilson coefficients @xmath34 and @xmath35 at nlo in rg - improved perturbation theory requires one - loop matching at the scale @xmath36 and two - loop anomalous dimensions in the effective theory . the leading - order coefficients @xmath34 have been computed long ago @xcite . the calculation at order @xmath0 is , however , much more complicated . the leading logarithmic corrections were obtained in @xcite . first steps towards a nlo analysis were made in @xcite , where the one - loop matching calculations were computed , and general arguments based on the reparameterization invariance @xcite of the hqet were used to derive the exact relations ( valid to all orders in perturbation theory ) @xmath37 moreover , it was shown that the @xmath38 anomalous dimension matrix governing the mixing of the operators @xmath23 under rg transformations has the texture @xmath39 where @xmath40 here @xmath41 and @xmath42 are the anomalous dimensions of the leading - order heavy light currents @xmath43 and of the chromo - magnetic operator , respectively . the matrix @xmath44 describing the mixing of the local operators @xmath45 into @xmath46 has been calculated at two - loop order in @xcite . explicit expression for the two - loop anomalous dimensions will be presented in section [ sec : analysis ] . it was observed in @xcite that there are numerically large discrepancies between the one - loop and leading - logarithmic approximations to the wilson coefficients @xmath35 , which could only be resolved at nlo in rg - improved perturbation theory . the missing piece in the nlo analysis is the calculation of the anomalous dimension matrix @xmath47 describing the mixing of the bilocal operators @xmath29 into the local operators @xmath46 . in this paper , we report the calculation of this matrix at two - loop order . it requires the evaluation of two - loop tensor integrals in the hqet , which are infrared ( ir ) singular when one of the external lines is taken on - shell . a general algorithm for computing such integrals has been developed in @xcite . for our analysis we will generalize the above discussion and consider an arbitrary dirac structure @xmath48 of the current @xmath49 in the time - ordered products . at the end we will study the particular case of vector currents in detail . the reader interested only in the final results of our calculation but not its technical details can continue with section [ sec : analysis ] , where we present the final expressions for the wilson coefficients @xmath35 at nlo and briefly discuss some phenomenological applications . our task is to calculate , at two - loop order , the mixing of bilocal operators , consisting of time - ordered products of a local current @xmath49 with either the kinetic operator or the chromo - magnetic operator , into local dimension-4 current operators with a derivative acting on the light - quark field . here @xmath48 can be an arbitrary dirac matrix , depending on the lorentz structure of the original current that is expanded in terms of hqet operators . we start by constructing , in the hqet , a basis of the relevant operators that mix under renormalization . since ultimately our interest is in the matrix elements of these operators between physical hadron states , it is sufficient to consider gauge - invariant operators that do not vanish by the equations of motions . for the case where the bilocal operator contains the kinetic operator , such a basis consists of the two operators @xmath50 if the bilocal operator contains the magnetic operator , the basis consists of the three operators @xmath51 for each of the two cases , we define a matrix @xmath52 of renormalization constants , which absorb the ultraviolet ( uv ) divergences in the matrix elements of the bare operators , by the relation @xmath53 . the matrix @xmath54 of the anomalous dimensions , which govern the scale dependence of the renormalized operators , is given by @xmath55 in a minimal subtraction scheme , the renormalization constants are defined to remove the @xmath56 poles arising in the calculation of the bare green functions with insertions of the operators @xmath57 in dimensional regularization , i.e. in @xmath58 space time dimensions . hence , @xmath59 the requirement that the anomalous dimension matrix be finite in the limit @xmath60 implies the relations @xcite @xmath61 where @xmath62 is the @xmath63 function . the first equation shows that the anomalous dimension matrix can be obtained from the coefficient of the @xmath56 pole in @xmath52 , whereas the second one implies a nontrivial constraint on the coefficient of the @xmath64 pole , arising at two - loop and higher order . in our case , the structure of the two @xmath52 matrices is @xmath65 where the renormalization constants @xmath66 , @xmath67 and @xmath68 determine the mixing of local , dimension-4 heavy light currents and have been calculated at two - loop order in @xcite , whereas @xmath69 is the renormalization constant of the chromo - magnetic operator , which has been computed at two - loop order in @xcite . here we will calculate the remaining entries @xmath70 with the same accuracy . previous authors have calculated the mixing of the bilocal operators into local operators at the one - loop order , finding @xcite @xmath71 using the known expression for the other @xmath72-factors , this information can be used to predict the @xmath64 poles in the two - loop coefficients of the renormalization constants . from the second relation in ( [ magic ] ) , we obtain @xmath73 + o(\alpha_s^3 ) \ , , \nonumber\\ z_{t1}^{{\rm mag}\,(2 ) } & = & \left ( \frac{\alpha_s}{4\pi } \right)^2 \left [ \frac{7}{4}\,c_f^2 + \frac{4}{3}\,c_f c_a - \frac{2}{3}\,c_f t_f\,n_f \right ] + o(\alpha_s^3 ) \ , , \nonumber\\ z_{t2}^{{\rm mag}\,(2 ) } & = & \left ( \frac{\alpha_s}{4\pi } \right)^2 \left [ \frac{3}{4}\,c_f^2 + \frac{4}{3}\,c_f c_a - \frac{2}{3}\,c_f t_f\,n_f \right ] + o(\alpha_s^3 ) \,.\end{aligned}\ ] ] here @xmath74 , @xmath75 and @xmath76 are the color factors for an @xmath77 gauge group , and @xmath78 is the number of light - quark flavors . these relations will provide a check on our two - loop results . to obtain the renormalization constants @xmath70 at order @xmath79 , we calculate the insertions of the bilocal operators @xmath80 and @xmath81 into the amputated green function with a heavy and a light quark to two - loop order . the relevant diagrams are shown in figure [ fig:2loop ] . we use dimensional regularization in @xmath58 space time dimensions and work in the @xmath82 subtraction scheme . the relation between the renormalized and the bare , regularized amputated green functions is @xmath83 to compute the mixing of the bilocal operators into the local operators with a derivative acting on the light - quark field , we only need to keep terms linear in the momentum @xmath84 of the light quark . because the pole parts are polynomial in the external momenta , we can first take a derivative with respect to @xmath84 and then set @xmath85 , so that all integrals are of propagator type and depend on the single variable @xmath86 . however , this method fails for some of the diagrams , for which setting @xmath85 after differentiation leads to infrared ( ir ) divergences . in these cases , we apply a variant of the so - called @xmath87 operation @xcite , which compensates these ir poles by a recursive construction of counterterms for the ir - divergent subgraphs . [ cols="^,^,^,^,^,^,^,^,^ " , ] consider , as an example , the first diagram in figure [ fig:2loop ] . its contribution is proportional to the integral @xmath88 when linearizing this expression in @xmath84 , we encounter an ir divergence from the region @xmath89 , which can be removed by adding and subtracting the ir subtraction term @xmath90 the difference @xmath91 can be evaluated using a naive linearization in @xmath84 , since the behavior of the ir subtraction term for @xmath89 is the same as that of the original integral . however , because the expansion of @xmath92 involves tadpole integrals that vanish in dimensional regularization , the difference @xmath93 coincides with the naive linearization of the original expression for @xmath94 . the contribution @xmath92 , which is necessary to subtract the ir subdivergence of the original diagram , factorizes into an ir counterterm @xmath95 and the original diagram with the lines of the ir - sensitive subgraph removed , and with @xmath96 and @xmath84 set to zero . the remaining two - loop tensor integrals are of the general form @xmath97^{\alpha_5 } } \nonumber\\ & & \hspace{2.2 cm } \equiv -\pi^d ( -2\omega)^{2(d-\alpha_3-\alpha_4-\alpha_5)+n+m}\ , i_{\mu_1\dots\mu_n}^{\nu_1\dots\nu_m}(v;\{\alpha_i\ } ) \,.\end{aligned}\ ] ] using the method of integration by parts @xcite , one obtains the recurrence relation @xcite @xmath98\ , i_{\mu_1\dots\mu_n}^{\nu_1\dots\nu_m}(v;\{\alpha_i\ } ) \nonumber\\ & & \hspace{2.2 cm } = \sum_{j=1}^n\ , i_{\mu_1\dots[\mu_j]\dots\mu_n}^{\mu_j\nu_1\dots\nu_m } ( v;\{\alpha_i\ } ) \,,\end{aligned}\ ] ] which can be used to express any two - loop integral in terms of degenerate integrals , which have @xmath99 , @xmath100 or @xmath101 . here @xmath102 is an operator raising the index @xmath103 by one unit etc . , and @xmath104 $ ] means that this index is omitted . the degenerate integrals can be related in a straightforward way to products of one - loop tensor integrals @xcite . using this technique , we have calculated the pole parts of the two - loop diagrams in the t hooft feynman gauge . the result is a linear combination of terms with the dirac structures corresponding to the local operators @xmath105 and @xmath106 defined in ( [ okin ] ) and ( [ omag ] ) . the results are summarized in the first two columns of table [ tab:2loop ] . in the @xmath82 scheme , the renormalization scale @xmath107 is introduced by the replacement of the bare coupling constant with the renormalized one through the relation @xmath108 with @xmath109 . ' '' '' @xmath110 ' '' '' & @xmath111 & @xmath112 & @xmath113 + @xmath105 ' '' '' & @xmath114 & @xmath115 & @xmath116 & @xmath117 & 0 + & @xmath118 & @xmath119 & @xmath120 & @xmath121 & @xmath122 + & @xmath123 & @xmath124 & 0 & @xmath125 & 0 + @xmath126 ' '' '' & @xmath114 & @xmath127 & @xmath128 & @xmath129 & @xmath130 + & @xmath118 & @xmath131 & 0 & @xmath132 & 0 + & @xmath123 & @xmath133 & 0 & @xmath134 & 0 + @xmath135 ' '' '' & @xmath114 & @xmath136 & @xmath137 & @xmath138 & @xmath130 + & @xmath118 & @xmath139 & 0 & @xmath132 & 0 + & @xmath123 & @xmath133 & 0 & @xmath134 & 0 + the two - loop diagrams in figure [ fig:2loop ] contain subdivergences , which must be subtracted by uv counterterms . in addition to the standard one - loop counterterms for the quark and gluon propagators and vertices , local operator counterterms are required . to find these , we have calculated at one - loop order all insertions of the operators @xmath80 and @xmath81 into the amputated green functions with a non - negative degree of divergence . in our case , those are the two- and three - point functions with field content @xmath140 , @xmath141 , @xmath142 , and @xmath143 . we find that in the t hooft feynman gauge the uv divergences of these functions are removed by the counterterms @xmath144 \nonumber\\ & & \mbox{}- \frac{c_a\alpha_s}{4\pi\ep } \bigg [ - \frac{3}{2}\,\bar q\,\gamma\,v\!\cdot\!a\,h_v - \frac{3}{2}\,i\!\int\!\mbox{d}^4x\,{\rm t}\ { \bar q\,\gamma h_v(0 ) , \bar h_v\,\{iv\!\cdot\!d , v\!\cdot\!a\ } h_v(x ) \ } \nonumber\\ & & \hspace{1.5cm}\mbox{}+ i\!\int\!\mbox{d}^4x\,{\rm t}\ { \bar q\,\gamma h_v(0 ) , \bar h_v\,\{id_\mu , a^\mu\ } h_v(x ) \ } \bigg ] \,,\end{aligned}\ ] ] and @xmath145 \,.\ ] ] since the two - loop calculation was performed off - shell , some gauge - dependent operators must be included in addition to the operators @xmath57 and @xmath146 , as well as operators that vanish by the equations of motion . ( the local operators vanishing by the equations of motion actually yield no contribution and could be dropped ; however , one must keep these operators when they appear inside time - ordered products . ) the results for the counterterm contributions are summarized in the last two columns of table [ tab:2loop ] . when these contributions are added to the result for the sum of the two - loop diagrams , all nonlocal @xmath147 divergences proportional to @xmath148 and @xmath149 cancel . this is a check of our calculation . it follows from ( [ gamrel ] ) that the sum of the two - loop diagrams plus their counterterms determines the two - loop coefficients of the products @xmath150 . to obtain the results for the renormalization constants @xmath70 at order @xmath79 , we have to account for one - loop operator mixing and wave - function renormalization of the external quark fields . our final expressions are @xmath151 \bigg\ } \ , , \nonumber\\ z_{t1}^{\rm mag } & = & \frac{c_f\alpha_s}{4\pi\ep } \bigg\ { - 1 + \frac{\alpha_s}{4\pi } \bigg [ c_f \left ( \frac{7}{4\ep } - \frac{4}{3 } - \frac{10\pi^2}{9 } \right ) + c_a \left ( \frac{4}{3\ep } - \frac{25}{8 } + \frac{5\pi^2}{18 } \right ) \nonumber\\ & & \hspace{3.3cm}\mbox{}+ t_f\,n_f \left ( - \frac{2}{3\ep } + \frac{1}{9 } \right ) \bigg ] \bigg\ } \ , , \nonumber\\ z_{t2}^{\rm mag } & = & \frac{c_f\alpha_s}{4\pi\ep } \bigg\ { - 1 + \frac{\alpha_s}{4\pi } \bigg [ c_f \left ( \frac{3}{4\ep } + \frac{11}{6 } - \frac{2\pi^2}{9 } \right ) + c_a \left ( \frac{4}{3\ep } - \frac{55}{18 } + \frac{\pi^2}{18 } \right ) \nonumber\\ & & \hspace{3.3cm}\mbox{}+ t_f\,n_f \left ( - \frac{2}{3\ep } + \frac{1}{9 } \right ) \bigg ] \bigg\ } \,.\end{aligned}\ ] ] they are the main result of this work . note that the @xmath64 poles in these expressions agree with ( [ ep2terms ] ) . this a nontrivial check of our calculation . the results derived in this work are sufficient to calculate , at the two - loop order , the operator mixing of bilocal operators into local dimension-4 operators for any choice of the dirac structure @xmath48 . however , care must be taken when using the results in ( [ zres ] ) for a particular choice of @xmath48 , if instead of the operators in ( [ okin ] ) and ( [ omag ] ) another set of basis operators is employed . whereas the transformation between one operator basis and another is trivial at the one - loop order , it can be subtle at nlo , because in dimensional regularization the relations between the operators of different bases may depend on @xmath152 . consider the general case where the operators @xmath153 are expressed in terms of some other operators @xmath154 by a linear transformation of the form @xmath155 . depending on the choice of @xmath48 , some of the operators @xmath57 may not be independent , so the new set may contain fewer operators than the original one . we define a left - inverse matrix @xmath156 such that @xmath157 . it then follows that the matrix @xmath158 of renormalization constants in the new basis ( denoted by a tilde ) must be such that @xmath159 next , we expand the transformation matrices as @xmath160 and @xmath161 with @xmath162 , and extract from this result the coefficient of the @xmath56 pole in the matrix @xmath158 , which determines the anomalous dimension matrix in the new basis . we find @xmath163 \nonumber\\ & & - \big [ \bm{l_1 } \bm{z}^{(1 ) } \bm{r_0 } + \bm{l_0 } \bm{z}^{(1 ) } \bm{r_1 } \big ] \bm{l_0 } \bm{z}^{(1 ) } \bm{r_0 } + \dots \,,\end{aligned}\ ] ] where the dots represent terms that do not contribute at two - loop order . armed with this general result , we now discuss the case of the vector current considered in the introduction . then the relevant dirac structures are @xmath164 and @xmath165 , and the operators @xmath57 are conveniently expressed in terms of the operators @xmath46 defined in ( [ vbasis ] ) . for @xmath164 we have @xmath166 , @xmath167 , @xmath168 , and @xmath169 whereas for @xmath165 we find @xmath170 , @xmath171 , @xmath172 , and @xmath173 hence only for the case of insertions of the magnetic operator there appear nontrivial basis transformations . we thus consider the basis transformation from the six operators consisting of two sets of @xmath174 evaluated with @xmath164 and @xmath165 to the new basis spanned by the operators @xmath175 . the corresponding transformation matrices @xmath176 and @xmath156 can easily be obtained from ( [ rel1 ] ) and ( [ rel2 ] ) . it is then straightforward to compute the matrix @xmath177 corresponding to the operator mixing of @xmath29 into @xmath46 , and then using the first relation in ( [ magic ] ) to calculate the two - loop anomalous dimension matrix @xmath47 . we find that @xmath178 where @xmath179 \bigg\ } \ , , \nonumber\\ \gamma_1^{\rm mag } & = & \frac{c_f\alpha_s}{4\pi } \bigg\ { -2 + \frac{\alpha_s}{4\pi } \bigg [ c_f \left ( - \frac{10}{3 } - \frac{40\pi^2}{9 } \right ) + c_a \left ( \frac{46}{9 } + \frac{10\pi^2}{9 } \right ) - \frac{44}{9}\,t_f\,n_f \bigg ] \bigg\ } \ , , \nonumber\\ \gamma_2^{\rm mag } & = & \frac{c_f\alpha_s}{4\pi } \bigg\ { \hspace{0.3 cm } 12 + \frac{\alpha_s}{4\pi } \bigg [ c_f \left ( \frac{38}{3 } + \frac{176\pi^2}{9 } \right ) + c_a \left ( \frac{236}{3 } - \frac{44\pi^2}{9 } \right ) - \frac{56}{3}\,t_f\,n_f \bigg ] \bigg\ } \ , , \nonumber\\ \gamma_3^{\rm mag } & = & \frac{c_f\alpha_s}{4\pi } \bigg\ { -2 + \frac{\alpha_s}{4\pi } \bigg [ c_f \left ( \frac{4}{3 } - \frac{8\pi^2}{9 } \right ) + c_a \left ( - \frac{206}{9 } + \frac{2\pi^2}{9 } \right ) + \frac{52}{9}\,t_f\,n_f \bigg ] \bigg\ } \,.\end{aligned}\ ] ] this completes the calculation of the anomalous dimension matrix ( [ gammamatr ] ) at two - loop order . the wilson coefficients @xmath35 in the heavy - quark expansion of the vector current in ( [ vector ] ) obey the renormalization - group equation @xmath180 where the vector @xmath181 contains the ten coefficients @xmath35 . the anomalous dimension matrix @xmath54 has been given in ( [ gammamatr ] ) , ( [ entries ] ) , ( [ gammaa ] ) , and ( [ gammac ] ) . besides the entries of the submatrix @xmath47 computed in this paper and shown in ( [ results ] ) , we need the two - loop anomalous dimension @xmath182 \bigg\}\ ] ] of the chromo - magnetic operator calculated in @xcite , as well as the two - loop anomalous dimensions of local dimension-4 currents , @xmath183 \bigg\ } \ , , \nonumber\\ \gamma_2 & = & \frac{c_f\alpha_s}{4\pi } \bigg\ { \hspace{0.53 cm } 3 + \frac{\alpha_s}{4\pi } \bigg [ c_f \left ( -5 + \frac{4\pi^2}{3 } \right ) + c_a \left ( \frac{41}{3 } - \frac{\pi^2}{3 } \right ) - \frac{10}{3}\,t_f\,n_f \bigg ] \bigg\ } \ , , \nonumber\\ \gamma_3 & = & \frac{c_f\alpha_s}{4\pi } \bigg\ { -2 + \frac{\alpha_s}{4\pi } \bigg [ c_f \left ( \frac{13}{3 } - \frac{8\pi^2}{9 } \right ) + c_a \left ( - \frac{104}{9 } + \frac{2\pi^2}{9 } \right ) + \frac{28}{9}\,t_f\,n_f \bigg ] \bigg\ } \ , , \end{aligned}\ ] ] which were computed in @xcite . the coefficient @xmath41 coincides with the universal hybrid anomalous dimension of the leading - order heavy light currents first obtained in @xcite . in addition to the anomalous dimensions , we need the one - loop matching conditions for the wilson coefficients at the scale @xmath36 . in the @xmath82 subtraction scheme , they are @xcite @xmath184 the solution of the renormalization - group equation at next - to - leading order is standard and described in detail in @xcite . a subtlety complicating the solution is that the one - loop anomalous dimension matrix can not be diagonalized . we circumvent this problem by adding a small diagonal contribution @xmath185 to the matrix @xmath186 in ( [ gammamatr ] ) , taking the limit @xmath187 at the end of the calculation . we quote results for @xmath188 colors and @xmath189 light quark flavors . we first reproduce the known nlo results @xmath190 \ , , \nonumber\\ c_2(\mu ) & = & b_2(\mu ) = \frac{1}{2}\,b_3(\mu ) = \frac{8}{3}\,x^{6/25}\,\frac{\alpha_s(m_q)}{4\pi } \ , , \nonumber\\ c_{\rm mag}(\mu ) & = & x^{-9/25 } \left [ 1 + 8.449\,\frac{\alpha_s(m_q)}{4\pi } + ( 0.218 + 3\delta_{\rm rs})\,\frac{\alpha_s(\mu)}{4\pi } \right ] \,,\end{aligned}\ ] ] where @xmath191 . according to ( [ b7to10 ] ) , products of these coefficients determine the coefficients @xmath192 . in addition , we obtain the new expressions @xmath193 \nonumber\\ & + & x^{6/25 } \ln x \left [ \frac{16}{25 } - 3.996\,\frac{\alpha_s(m_q)}{4\pi } + \left ( 0.582 - \frac{32}{25}\,\delta_{\rm rs } \right ) \frac{\alpha_s(\mu)}{4\pi } \right ] \nonumber\\ & + & x^{-3/25 } \left [ - \frac{4}{27 } - 0.327\,\frac{\alpha_s(m_q)}{4\pi } + \left ( 2.013 - \frac{4}{27}\,\delta_{\rm rs } \right ) \frac{\alpha_s(\mu)}{4\pi } \right ] \nonumber\\ & - & \frac{10}{9 } + 0.280\,\frac{\alpha_s(m_q)}{4\pi } - 0.993\,\frac{\alpha_s(\mu)}{4\pi } \ , , \nonumber\\ b_5(\mu ) & = & x^{6/25 } \left [ - \frac{28}{27 } + ( 32.097 + 1.707\ln x)\,\frac{\alpha_s(m_q)}{4\pi } - \left ( 0.944 - \frac{56}{27}\,\delta_{\rm rs } \right ) \frac{\alpha_s(\mu)}{4\pi } \right ] \nonumber\\ & + & x^{-3/25 } \left [ \frac{88}{27 } + 11.930\,\frac{\alpha_s(m_q)}{4\pi } - \left ( 30.101 - \frac{88}{27}\,\delta_{\rm rs } \right ) \frac{\alpha_s(\mu)}{4\pi } \right ] \nonumber\\ & - & \frac{20}{9 } + 0.560\,\frac{\alpha_s(m_q)}{4\pi } + 2.458\,\frac{\alpha_s(\mu)}{4\pi } \ , , \nonumber\\ b_6(\mu ) & = & x^{6/25 } \left [ - 2 + 7.153\,\frac{\alpha_s(m_q)}{4\pi } - ( 1.820 - 4\delta_{\rm rs})\,\frac{\alpha_s(\mu)}{4\pi } \right ] \nonumber\\ & + & x^{-3/25 } \left [ - \frac{4}{3 } - 2.941\,\frac{\alpha_s(m_q)}{4\pi } + \left ( 5.246 - \frac{4}{3}\,\delta_{\rm rs } \right ) \frac{\alpha_s(\mu)}{4\pi } \right ] \nonumber\\ & + & \frac{10}{3 } - 0.841\,\frac{\alpha_s(m_q)}{4\pi } - 1.464\,\frac{\alpha_s(\mu)}{4\pi } \,.\end{aligned}\ ] ] all terms proportional to the coupling @xmath194 in these expressions , as well as the leading - logarithmic rescaling factors , are independent of the renormalization scheme . the terms proportional to @xmath195 are scheme dependent , however . here we consider a class of schemes parameterized by the quantity @xmath196 these include all `` minimal - subtraction - like '' schemes , which are related to each other by a change in the renormalization scale . for instance , we have @xmath197 and @xmath198 by definition , i.e. , the above results evaluated with @xmath199 refer to the @xmath82 scheme . it follows from ( [ rge ] ) that at nlo the scheme - dependent terms in the wilson coefficients are given by @xmath200 in a complete nlo calculation , these terms combine with scheme - dependent terms in the hadronic matrix elements of the hqet operators to give a renormalization - group invariant answer . it is evident from the magnitude of some of the coefficients in ( [ bi ] ) that the next - to - leading corrections are , in some cases , quite large . to illustrate this point , we compare our nlo results with the naive one - loop expressions for the wilson coefficients calculated in @xcite , and with the results obtained in leading - logarithmic approximation @xcite , where all terms proportional to @xmath201 are omitted . the different results for the wilson coefficients are shown in figure [ fig:2 ] as a function of the renormalization scale @xmath107 , setting @xmath202gev corresponding to the @xmath5-quark mass . for the one - loop approximation we show a band obtained by varying the coupling between @xmath195 and @xmath194 . we note two important observations : ( i ) the one - loop and leading - logarithmic approximations give rather different results , which sometimes do not even have the same sign . this shows that the one - loop matching corrections at the scale @xmath36 are important . ( ii ) for low renormalization scales @xmath2032gev , some of the coefficients , in particular @xmath204 and @xmath205 , are numerically large despite the fact that they vanish at tree level . this shows that the effects of running are important , and thus a rg summation of the large logarithms @xmath9 becomes mandatory . the effects of matching and running can only be combined in a consistent , scheme - independent manner by going beyond the leading order . our nlo results ( shown by the solid lines ) typically lie between the two approximations , except for @xmath206 , where they agree with the naive one - loop results . we have calculated at two - loop order the mixing of bilocal operators into local dimension-4 current operators containing a heavy and a light quark field . when combined with results obtained by previous authors , this completes the calculation of the @xmath38 two - loop anomalous dimension matrix governing the renormalization - scale dependence of the operators appearing at order @xmath0 in the heavy - quark expansion of the weak current @xmath207 . we have presented expressions for the wilson coefficients in this expansion at next - to - leading order in renormalization - group improved perturbation theory . we find that the next - to - leading corrections are numerically large and must be included for a meaningful determination of these coefficients . this makes our results relevant for phenomenological applications of the heavy - quark expansion . the matrix elements of dimension-4 heavy light current operators play an important role in heavy - flavor phenomenology . they appear , e.g. , at order @xmath0 in the heavy - quark expansion of meson decay constants @xcite , and of the semileptonic form factors describing the exclusive @xmath208 and @xmath209 transitions @xcite . the primary motivation for our calculation , however , was the recent proposal for extracting the element @xmath1 of the quark mixing matrix from the lepton invariant mass spectrum in inclusive @xmath210 decays @xcite . this spectrum can be obtained using a two - step operator product expansion ( called `` hybrid expansion '' ) of a time - ordered product of two heavy light currents @xcite . our results will help to reduce the perturbative uncertainty in this expansion , which will ultimately lead to an improved accuracy in the determination of @xmath1 . at nlo in the hybrid expansion , one needs the expression for the combination of wilson coefficients @xmath211 derived in the present work . we now present an exact result for this combination , evaluating the coefficients with @xmath188 colors , but for an arbitrary number of light quark flavors . we find @xmath212 \nonumber\\ & - & \frac{16}{9}\,x^{-1/\beta_0 } \left [ 1 + \frac{\alpha_s(m_q)}{4\pi}\,k_3 - \frac{\alpha_s(\mu)}{4\pi } \left ( k_3-k_2 - \frac{34}{3 } - \delta_{\rm rs } \right ) \right ] \nonumber\\ & + & \frac{16}{\beta_0}\,x^{2/\beta_0 } \ln x \left [ 1 + \frac{\alpha_s(m_q)}{4\pi}\,k_4 - \frac{\alpha_s(\mu)}{4\pi } \left ( k_4 + \frac{16}{3 } + 2\delta_{\rm rs } \right ) \right ] \nonumber\\ & - & \frac{40}{9}\,\frac{\alpha_s(\mu)}{4\pi } \,,\end{aligned}\ ] ] where @xmath213 and @xmath214 is the first coefficient of the @xmath63-function . the numerical values in ( [ ki ] ) refer to @xmath189 . in @xcite , we combine this result with an explicit calculation of the hadronic matrix elements in the hybrid expansion , finding that the scheme dependence parameterized by @xmath215 disappears from the final result for the semileptonic @xmath210 decay rate . _ acknowledgements : _ one of us ( m.n . ) would like to thank gabriel amors for earlier collaboration on a similar subject . this work was supported in part by the national science foundation . t.b . is supported by the swiss national science foundation .

we present exact results , at next - to - leading order in renormalization - group improved perturbation theory , for the wilson coefficients appearing at order @xmath0 in the heavy - quark expansion of heavy light current operators . to this end , we complete the calculation of the corresponding two - loop anomalous dimension matrix . our results are important for determinations of @xmath1 using exclusive and inclusive semileptonic @xmath2 decays . they are also relevant to computations of the decay constant @xmath3 based on a heavy - quark expansion . clns 00/1713 + hep - ph/0012183 * two - loop renormalization of heavy light currents + at order @xmath0 in the heavy - quark expansion * thomas becher , matthias neubert and alexey a. petrov + _ newman laboratory of nuclear studies , cornell university + ithaca , ny 14853 , usa _ december 2000

in the semileptonic @xmath6 decay with a meson @xmath7 and a charged lepton @xmath8 , the @xmath9 pair involves no direct qcd interaction so that the theoretical description of the amplitude can be reduced to a simple form with the @xmath10 transition . for example , the rate for @xmath11 is proportional to @xmath12 , where the form factor @xmath13 for the @xmath14 transition depends on the momentum transfer squared , @xmath15 . this benefits the precision measurement of @xmath16 , where @xmath16 is one of the least known cabibbo - kobayashi - maskawa ( ckm ) matrix elements @xcite in the standard model ( sm ) . as long as we choose a point @xmath17 in the decay spectrum , the corresponding data point with other parameters can be fixed to extract the value of @xmath16 . however , @xmath13 relies on the calculations in the qcd models , such as quark models @xcite , lattice qcd @xcite , and light cone sum rules @xcite . starting with @xmath18 and @xmath16 , one is allowed to inversely extract the @xmath15 dependence of @xmath13 in different @xmath15 intervals from the measured data @xcite . the extraction compared with various theoretical models hence improves the knowledge of @xmath13 . moreover , such extraction also provides crosschecks for the @xmath19 and @xmath20 transition form factors @xcite . in particular , the size of the gluonic singlet contribution @xcite to the @xmath21 transition to explain the unexpectedly large two - body hadronic @xmath22 decay rate has been constrained by measuring @xmath23 decays @xcite . similar to the mesonic cases , it should be interesting to extend the study to baryonic decay modes , such as @xmath24 with @xmath25 being a baryon pair , to investigate the @xmath26 transition form factors , which have been used as theoretical inputs in the three - body @xmath27 decays . the factorizable amplitudes for the three - body baryonic @xmath28 decays are normally classified into current and transition parts , given by @xmath29 respectively , where @xmath30 and @xmath31 stand for the weak currents . the matrix elements of @xmath32 in @xmath33 are presented as the timelike baryonic form factors , for which the theoretical calculations are available , such as the approach of the pqcd counting rules @xcite . consequently , the observed branching ratios for @xmath34 @xcite , @xmath35 @xcite and @xmath36 @xcite can be explained due to the @xmath33-like amplitudes @xcite . on the other hand , the measured decays of @xmath37 , @xmath38 @xcite , and @xmath39 , shown in fig . [ fig1 ] , are considered to have @xmath40 as their amplitudes @xcite . to explain the data , the transition matrix elements of @xmath41 are parameterized in terms of various form factors @xcite . decays with the @xmath42 transition , where ( a ) @xmath43 , ( b ) @xmath44 , ( c ) @xmath45 , and ( d ) @xmath46 . , title="fig:",width=144 ] decays with the @xmath42 transition , where ( a ) @xmath43 , ( b ) @xmath44 , ( c ) @xmath45 , and ( d ) @xmath46 . , title="fig:",width=144 ] decays with the @xmath42 transition , where ( a ) @xmath43 , ( b ) @xmath44 , ( c ) @xmath45 , and ( d ) @xmath46 . , title="fig:",width=172 ] decays with the @xmath42 transition , where ( a ) @xmath43 , ( b ) @xmath44 , ( c ) @xmath45 , and ( d ) @xmath46 . , title="fig:",width=144 ] as seen in fig . [ fig1 ] , the decay of @xmath47 is also a @xmath40 type as @xmath48 . it is clear that the observation of @xmath49 shall directly determine the transition form factors , which have been widely used to explain the @xmath27 data as theoretical inputs . in analogy with the timelike baryonic form factors , similar momentum dependences of the transition form factors may be chosen , which can be justified by investigating the shape of the invariant mass spectrum for the @xmath50 decay . moreover , we expect that the measurements of the angular distributions in @xmath51 will provide some information to understand the unexpectedly large angular distribution asymmetries of @xmath52 @xcite . at present , the cleo collaboration has given an experimental upper limit : @xcite @xmath53 while the theoretical estimation has only been done for the inclusive @xmath5 decays with charmless dibaryons , given by @xcite @xmath54 in this paper , we concentrate on the exclusive four - body semileptonic baryonic decay of @xmath55 ( @xmath56 , or @xmath57 ) . in particular , we will study its decay branching ratio in the sm . the paper is organized as follows . in sec . ii , we provide the formalism , in which we show the decay amplitude and rate of @xmath55 along with the definitions of the transition form factors of @xmath41 . we give our numerical results and discussions in sec . iii . in sec . iv , we present the conclusions . we then need the kinematics for the four - body @xmath75 decay to integrate over the phase space . as the formalisms in @xmath76 , @xmath77 , and @xmath78 decays given in refs . @xcite , we use five kinematic variables , @xmath79 , @xmath80 , @xmath81 , @xmath82 , and @xmath83 to describe the decay . the three angles @xmath81 , @xmath82 , and @xmath83 are depicted in fig . [ 3angles ] , where the angle @xmath84 is between @xmath85 ( @xmath86 ) in the @xmath87 ( @xmath88 ) rest frame and the line of flight of the @xmath87 ( @xmath88 ) system in the rest frame of the @xmath0 meson , while the angle @xmath83 is from the @xmath87 plane defined by the momenta of the @xmath87 pair to the @xmath88 plane defined by the momenta of the @xmath88 pair in the rest frame of @xmath0 . , @xmath82 , and @xmath83 in @xmath89.,width=240 ] the partial decay width reads @xmath90 where @xmath91 , @xmath92 , and @xmath93 are given by @xmath94^{1/2}\,,\nonumber\\ \beta_{\bf b}&=&\frac{1}{t}\lambda^{1/2}(t , m_{\bf b}^2,m_{\bf \bar b'}^2)\,,\nonumber\\ \beta_{\bf l}&=&\frac{1}{s}\lambda^{1/2}(s , m_{\ell}^2,m_{\bar \nu}^2)\,,\end{aligned}\ ] ] respectively , with @xmath95 . the regions for the five variables of the phase space are given by @xmath96 in our numerical analysis , we take @xmath97 from the pdg @xcite . to deal with @xmath73 and @xmath72 in eq . ( [ figi ] ) , it is helpful to use the approach of the pqcd counting rules again , where with @xmath98 flavor and @xmath99 spin symmetries the vector and axial - vector currents are incorporated as two chiral currents in the large t limit @xcite . consequently , @xmath73 and @xmath72 from the vector currents are related by another set of constants @xmath100 and @xmath101 from the chiral currents . explicitly , for the @xmath102 transition form factors we have @xcite @xmath103 with @xmath104 , where their values are determined by fitting the data of the total branching ratios , invariant mass spectra , and angular distributions measured in the @xmath27 decays . to illustrate our results , we adopt the values in ref . @xcite : @xmath105 thus , the total branching ratios of @xmath49 are found to be @xmath106 where the two errors in eq . ( [ pred ] ) are from those in eq . ( [ inputd ] ) and @xmath16 , respectively . the invariant mass spectra and angular distributions for @xmath107 are shown in fig . [ fig3 ] . and @xmath108 and angular distributions as functions of @xmath109 for @xmath110 , respectively.,title="fig:",width=184 ] and @xmath108 and angular distributions as functions of @xmath109 for @xmath110 , respectively.,title="fig:",width=192 ] and @xmath108 and angular distributions as functions of @xmath109 for @xmath110 , respectively.,title="fig:",width=192 ] the integrated angular distribution asymmetries , defined by @xmath111 are obtained to be @xmath112 where the errors are from those in eq . ( [ inputd ] ) . since our result on @xmath113 in eq . ( [ pred ] ) is around @xmath114 , which is the same order of magnitude as those of the well measured mesonic @xmath115 decays at belle and babar , such as @xmath116 and @xmath117 , this four - body mode should be observed at these b - factories @xcite . moreover , as seen from fig . [ fig3]a , the @xmath107 decay inherits the same threshold enhancement as those in the three - body baryonic @xmath0 decays , resulting from the adoption of @xmath118 for the momentum dependence in the @xmath102 transition form factors . the spectrum in fig . [ fig3]b reflects the fact that in the helicity structure the amplitude of the @xmath119 pair is proportional to @xmath120 . it is interesting to note that our study of @xmath49 is similar to that of @xmath121 @xcite . the terms related to @xmath122 and @xmath123 in the @xmath102 transition form factors give the main contributions to @xmath1 . since the pair of the left - handed electron and the right - handed anti - neutrino in the helicity structure behaves as one of the polarization vector @xmath124 with @xmath125 , leading to @xmath126 , the contributions from @xmath127 and @xmath128 disappear . those from @xmath129 and @xmath130 are effectively small due to the tiny @xmath131 . as the branching ratio receives the most contribution near the threshold area , the @xmath132(@xmath133)-accompanied term @xmath134 is suppressed . moreover , since the terms of @xmath122 and @xmath123 contain @xmath135 , we have the relation @xmath136 @xmath137 @xmath138 , which explains why @xmath122 and @xmath123 prevail over other terms in the @xmath49 decay . finally , it is interesting to point out that the angular distribution asymmetries in eq . ( [ as ] ) in the @xmath49 decay are sensitive to new physics , such as the currents of @xmath139 and @xmath140 beyond the sm . note that @xmath141 in eq . ( [ angbl ] ) ( see also fig . [ fig3]c ) is caused by the @xmath142 pair of @xmath143 in the sm , which forms a polarization vector @xmath124 to couple to the left - handed helicity state of the virtual weak boson @xmath144 . therefore , a new physics with the @xmath139 current , which lets the @xmath145 pair to be another polarization state @xmath146 , must result in the deviation of @xmath147 in eq . ( [ angbl ] ) . apart from a direct @xmath148 violation @xcite , @xmath49 can easily create @xmath149-odd triple product correlations ( tpc s ) to test direct @xmath149 violation effects . since the three - momenta of @xmath68 and those of @xmath9 are not in the same plane , @xmath150 can be a nonzero tpc observable . like the case of @xmath151 in @xmath152 @xcite with @xmath153 denoting the @xmath154 spin , there are other tpc observables @xmath155 and @xmath156 in @xmath49 . these rich tpc observables are expected to be useful to test new physics in the advantage of @xmath157 of order @xmath158 much larger than the sensitivity of @xmath159 in the @xmath115 factories . we have examined the four - body semileptonic baryonic @xmath0 decay of @xmath49 in the sm , which proceeds via @xmath160 at the quark level . the transition form factors of @xmath102 , which are well studied in the three - body baryonic @xmath27 decays , play the key role in the theoretical calculation . we have found that @xmath161 for @xmath162 , respectively , which are just a little below the cleo s upper limit of @xmath4 for @xmath163 but much larger than the previous estimations of @xmath164 for the inclusive modes of @xmath5 . it is clear that the four - body decays of @xmath49 , in particular the light charged leoton modes , should be observed by the b - factories of belle and babar as well as future b - factories , such as super - belle and lhcb . 99 n. cabibbo , phys . * 10 * , 531 ( 1963 ) . m. kobayashi and t. maskawa , prog . phys . * 49 * , 652 ( 1973 ) . d. scora and n. isgur , phys . rev . d * 52 * , 2783 ( 1995 ) . e. dalgic , a. gray , m. wingate , c. t. h. davies , g. p. lepage and j. shigemitsu , phys . d * 73 * , 074502 ( 2006 ) [ erratum - ibid . d * 75 * , 119906 ( 2007 ) ] . p. ball and r. zwicky , phys . d * 71 * , 014015 ( 2005 ) . b. aubert _ et al . _ [ babar collaboration ] , phys . lett . * 98 * , 091801 ( 2007 ) . h. ha _ et al . _ [ belle collaboration ] , phys . d * 83 * , 071101 ( 2011 ) . n. e. adam _ et al . _ [ cleo collaboration ] , phys . lett . * 99 * , 041802 ( 2007 ) . r. gray _ et al . _ [ cleo collaboration ] , phys . d * 76 * , 012007 ( 2007 ) . del amo sanchez _ et al . _ [ babar collaboration ] , phys . rev . d * 83 * , 052011 ( 2011 ) . m. beneke and m. neubert , nucl . b * 651 * , 225 ( 2003 ) . kim , s. oh , and c. yu , phys . b * 590 * , 223 ( 2004 ) . c.h . chen and c.q . geng , phys . b * 645 * , 197 ( 2007 ) . lepage and s.j . brodsky , phys . lett . * 43 * , 545(1979 ) [ erratum - ibid . * 43 * , 1625 ( 1979 ) ] . lepage and s.j . brodsky , phys . d * 22 * , 2157 ( 1980 ) . brodsky , g.p . lepage and s. a. a. zaidi , phys . d * 23 * , 1152 ( 1981 ) . s. anderson _ [ cleo collaboration ] , phys . lett . * 86 * , 2732 ( 2001 ) . wang _ et al . _ [ belle collaboration ] , phys . rev . lett . * 90 * , 201802 ( 2003 ) ; m.z . wang _ et al . _ [ belle collaboration ] , phys . d * 76 * , 052004 ( 2007 ) . wang _ et al . _ [ belle collaboration ] , phys . lett . b * 617 * , 141 ( 2005 ) . b. aubert _ et al . _ [ babar collaboration ] , phys . d * 79 * , 112009 ( 2009 ) . y.j . lee _ et al . _ [ belle collaboration ] , phys . lett . * 93 * , 211801 ( 2004 ) . y.w . chang _ et al . _ [ belle collaboration ] , phys . d * 79 * , 052006 ( 2009 ) . cheng and k.c . yang , phys . d * 66 * , 014020 ( 2002 ) . cheng and k.c . yang , phys . d * 66 * , 094009 ( 2002 ) . chua , w.s . hou and s.y . tsai , phys . d * 65 * , 034003 ( 2002 ) . chua , w.s . hou and s. y. tsai , phys . d * 66 * , 054004 ( 2002 ) . chua and w.s . hou , eur . j. * c29 * , 27 ( 2003 ) . geng and y.k . hsiao , phys . d * 75 * , 094005 ( 2007 ) . geng and y.k . hsiao , phys . b * 619 * , 305 ( 2005 ) . chen , h.y . cheng , c.q . geng and y. k. hsiao , phys . d * 78 * , 054016 ( 2008 ) . y.k . hsiao , int . j. mod . a * 24 * , 3638 ( 2009 ) . et al_. [ belle collaboration ] , phys . lett . * 92 * , 131801 ( 2004 ) . b. aubert _ et al . _ [ babar collaboration ] , phys . d * 72 * , 051101 ( 2005 ) . wei _ et al . _ [ belle collaboration ] , phys . b * 659 * , 80 ( 2008 ) . b. aubert _ et al . _ [ babar collaboration ] , phys . d * 76 * , 092004 ( 2007 ) . chen _ et al . _ [ belle collaboration ] , phys . lett . * 100 * , 251801 ( 2008 ) . k. abe _ et al . _ , [ belle collaboration ] , phys . rev * 89 * , 151802 ( 2002 ) . b. aubert _ et al . _ [ babar collaboration ] , phys . d * 74 * , 051101 ( 2006 ) . geng and y.k . hsiao , phys . d * 74 * , 094023 ( 2006 ) . geng , y.k . hsiao and j. n. ng , phys . lett . * 98 * , 011801 ( 2007 ) . geng , y.k . hsiao and j. n. ng , phys . rev . d * 75 * , 094013 ( 2007 ) . adam _ et al . _ [ cleo collaboration ] , phys . d * 68 * , 012004 ( 2003 ) . hou and a. soni , phys . rev . lett . * 86 * , 4247 ( 2001 ) . a. pais and s.b . treiman , phys . rev . * 168 * , 1858 ( 1968 ) . lee , m. lu and m.b . wise , phys . d * 46 * , 5040 ( 1992 ) . cheng , c.y . cheung , w. dimm , g.l . lin , y.c . lin , t.m . yan and h.l . yu , phys . d * 48 * , 3204 ( 1993 ) . k. nakamura _ et al . _ [ particle data group ] , j. phys . g * 37 * , 075021 ( 2010 ) . wang , private communication . geng and y.k . hsiao , phys . d * 72 * , 037901 ( 2005 ) ; int . j. mod . a * 21 * , 897 ( 2006 ) ; int . j. mod . a * 23 * , 3290 ( 2008 ) .

we study the four - body exclusive semileptonic baryonic @xmath0 decays of @xmath1 ( @xmath2 ) in the standard model . we find that their decay branching ratios are about @xmath3 , respectively . in particular , the electron mode is close to the corresponding cleo s upper limit of @xmath4 , while all results are about one or two orders of magnitude larger than the previous estimated values for the inclusive modes of @xmath5 . clearly , both b - factories of belle and babar should be able to observe these exclusive four - body modes .

up to the summer of 1996 , there had been a number of spectacular successes in mirror symmetry . after the pioneering initial work of candelas , de la ossa , greene and parkes @xcite which calculated the instanton predictions for the quintic three - fold , batyrev @xcite gave a powerful mirror symmetry construction for calabi - yau hypersurfaces in toric varieties , later generalized by batyrev and borisov @xcite to complete intersections in toric varieties . kontsevich @xcite introduced his notion of stable maps of curves and studied their moduli , setting the stage for a new flowering of enumerative geometry . eventually , this led to the mathematical calculation of gromov - witten invariants for the quintic in varying forms , @xcite . in between , a great deal of the structure of mirror symmetry was elucidated by many researchers in both string theory and algebraic geometry . on the other hand , at that time i had been primarily interested in the geometry of calabi - yau manifolds . many of the results about mirror symmetry seemed to rely primarily on information about the ambient toric varieties ; in particular , givental s work @xcite and succeeding work by lian , liu and yau @xcite , bertram @xcite and gathmann @xcite always performed calculations on the moduli space of stable maps into @xmath0 rather than the quintic , and in the batyrev - borisov type constructions , while there was a clear combinatorial relationship between the ambient toric varieties , there was no apparent geometric relationship between the calabi - yaus themselves . as a result , i tended to avoid thinking about mirror symmetry precisely because of this lack of geometric understanding . this changed dramatically in june of 1996 , when strominger , yau and zaslow @xcite released their paper `` mirror symmetry is @xmath1-duality . '' they made a remarkable proposal , based on recent ideas in string theory , that for the first time gave a geometric interpretation for mirror symmetry . let me summarize , very roughly , the physical argument here . developments in string theory in the mid-1990s had introduced the notion of _ dirichlet branes _ , or @xmath2-branes . these are submanifolds of space - time , with some additional data , which should serve as a boundary condition for open strings , i.e. we allow open strings to propagate with their endpoints constrained to lie on a @xmath2-brane . remembering that space - time , according to string theory , looks like @xmath3 , where @xmath4 is ordinary space - time and @xmath5 is a calabi - yau three - fold , we can split a @xmath2-brane into a product of a submanifold of @xmath4 and one on @xmath5 . it turned out , simplifying a great deal , that there were two particular types of submanifolds on @xmath5 of interest : _ holomorphic _ @xmath2-branes , i.e. holomorphic submanifolds with a holomorphic line bundle , and _ special lagrangian _ @xmath2-branes , which are _ special lagrangian submanifolds _ with flat @xmath6 bundle : let @xmath5 be an @xmath7-dimensional calabi - yau manifold with @xmath8 the khler form of a ricci - flat metric on @xmath5 and @xmath9 a nowhere vanishing holomorphic @xmath7-form . then a submanifold @xmath10 is _ special lagrangian _ if it is lagrangian , i.e. @xmath11 and @xmath12 , and in addition @xmath13 . the origins of mirror symmetry in physics suggest that if @xmath5 and @xmath14 are a mirror pair of calabi - yau manifolds , then string theory on a compactification of space - time using @xmath5 should be the same as that using @xmath14 , but with certain data interchanged . in the case of @xmath2-branes , the suggestion is that the moduli space of holomorphic @xmath2-branes on @xmath5 should be isomorphic to the moduli space of special lagrangian @xmath2-branes on @xmath14 . now @xmath5 itself is the moduli space of points on @xmath5 . so each point on @xmath5 should correspond to a pair @xmath15 , where @xmath16 is a special lagrangian submanifold and @xmath17 is a flat connection on @xmath18 . a theorem of mclean @xcite tells us that the tangent space to the moduli space of special lagrangian deformations of a special lagrangian submanifold @xmath19 is @xmath20 . of course , the moduli space of flat @xmath6-connections modulo gauge equivalence on @xmath18 is the torus @xmath21 . in order for this moduli space to be of the correct dimension , we need @xmath22 , the complex dimension of @xmath5 . this suggests that @xmath5 consists of a family of tori which are dual to a family of special lagrangian tori on @xmath14 . an elaboration of this argument yields the following conjecture : _ the strominger - yau - zaslow conjecture_. if @xmath5 and @xmath14 are a mirror pair of calabi - yau @xmath7-folds , then there exists fibrations @xmath23 and @xmath24 whose fibres are special lagrangian , with general fibre an @xmath7-torus . furthermore , these fibrations are dual , in the sense that canonically @xmath25 and @xmath26 whenever @xmath27 and @xmath28 are non - singular tori . i will clarify this statement as we review the work of the past ten years ; however , as i have stated this conjecture , it is likely to be false . on the other hand , there are weaker versions of the conjecture which probably are true . even better , these weaker statements are probably within reach of modern - day technology ( with a lot of hard work ) . nevertheless , there has been a lot of good progress on precise versions of the above conjecture at the topological and symplectic level . in addition , the conjecture has been successful at explaining many features of mirror symmetry , some of which are still heuristic and some of which are rigorous . it is my belief that a final satisfactory understanding of mirror symmetry will flow from the syz conjecture , even if results do not take the form initially suggested by it . my main goal here is to explain the journey taken over the last ten years . i want to focus on explaining the evolution and development of the ideas , rather than focus on precise statements . except in the first few sections , i will give few precise statements . in those first sections , i will clarify the above statement of the conjecture , and show how it gives a satisfactory explanation of mirror symmetry in the so - called semi - flat case , i.e. the case when the metric along the special lagrangian fibres is flat . this leads naturally to a discussion of affine manifolds , metrics on them , and the legendre transform . these now appear to be the key structures underlying mirror symmetry . we next take a look at the case when singular fibres appear . in this case we need to abandon the precise form of duality we developed in the semi - flat case and restrict our attention to topological duality . in the realm of purely topological duality , the syz conjecture has been entirely successful at explaining topological features of mirror symmetry for a large range of calabi - yau manifolds , including those produced by the batyrev - borisov construction for complete intersections in toric varieties . moving on , we take a look at dominic joyce s arguments demonstrating the problems with the strong form of the syz conjecture stated above . this forces us to recast the syz conjecture as a limiting statement . mirror symmetry is always about the behaviour of calabi - yau manifolds near maximally unipotent degenerations . a limiting form of the syz conjecture suggests that one can find special lagrangian tori on calabi - yau manifolds near a maximally unipotent degeneration , and as we approach the limit point in complex moduli space , we expect to see a larger portion of the calabi - yau manifold filled out by special lagrangian tori . unlike the original syz conjecture , though still difficult , this one looks likely to be accessible by current techniques . this form of the conjecture then motivates a new round of questions . in this limiting picture , we expect the base @xmath29 of the hypothetical special lagrangian fibration to be the so - called gromov - hausdorff limit of a sequence of calabi - yau manifolds approaching the maximally unipotent degeneration . gromov - hausdorff convergence is a metric space concept , while maximally unipotent degeneration is an algebro - geometric , hodge - theoretic concept . how do these two concepts relate ? in the summer of 2000 , kontsevich suggested that the gromov - hausdorff limit will be , roughly , the dual intersection complex of the algebro - geometric degeneration , at least on a topological level . we explore this idea in 6 . on the other hand , how does this help us with mirror symmetry ? parallel to these developments on limiting forms of the syz conjecture , my coauthor bernd siebert had been studying degenerations of calabi - yau manifolds using logarithmic geometry with stefan schrer . siebert noticed that mirror symmetry seemed to coincide with a combinatorial exchange of logarithmic data on the one side and polarizations on the other . together , we realised that this approach to mirror symmetry meshed well with the limiting picture predicted by syz . synthesizing these two approaches , we discovered an algebro - geometric version of the syz approach , which i will describe here . the basic idea is to forget about special lagrangian fibrations , and only keep track of the base of the fibration , which is an affine manifold . we show how polyhedral decompositions of affine manifolds give rise to degenerate calabi - yau varieties , and conversely how certain sorts of degenerations of calabi - yau varieties , which we call _ toric degenerations _ , give rise to affine manifolds as their dual intersection complex . mirror symmetry is again explained by a discrete version of the legendre transform much as in the semi - flat case . we end with a discussion of the connection of tropical curves with this approach . the use of tropical curves in curve counting in two - dimensional toric varieties has been pioneered in work of mihkalkin @xcite ; nishinou and siebert @xcite generalized this work to higher dimensions using an approach directly inspired by the approach i discuss here . on the other hand , tropical curves have not yet been used for counting curves in calabi - yau manifolds , so i will end the paper by discussing how tropical curves arise naturally in our picture . i would like to thank the organizers of the seattle conference for running an excellent conference , and my coauthors pelham wilson and bernd siebert on syz related results ; much of the work mentioned here came out of work with them . before doing anything else , let s ask a very basic question : why should dualizing torus fibrations interchange hodge numbers of calabi - yau threefolds ? if you havent seen this , it s the first thing one should look at as it is particularly easy to see , if we make a few assumptions . suppose we are given a pair of calabi - yau threefolds @xmath5 and @xmath14 with fibrations @xmath23 , @xmath24 with the property that there is a dense open set @xmath30 such that @xmath31 and @xmath32 are torus fibre bundles . so all the singular fibres of @xmath33 and @xmath34 lie over @xmath35 . ( note that unless @xmath36 , there must be some singular fibres . ) finally , assume @xmath37 and @xmath38 are dual torus fibrations , i.e. @xmath37 can be identified with the torus fibration @xmath39 and @xmath38 can be identified with the torus fibration @xmath40 . ( this is a slight abuse of notation : by @xmath41 we really mean the torus bundle obtained by taking the vector bundle associated to the local system @xmath42 and dividing out by the family of lattices @xmath43 . ) if @xmath44 is a single torus with @xmath45 an @xmath7-dimensional vector space and @xmath46 a lattice in @xmath45 , then @xmath47 , while the dual torus , @xmath48 , has @xmath49 . if we choose an isomorphism @xmath50 , then we get an isomorphism @xmath51 . similarly , in the relative setting for @xmath37 and @xmath38 , if we have an isomorphism @xmath52 , we obtain isomorphisms @xmath53 we now make a simplifying assumption . let @xmath54 be the inclusion . we will say @xmath33 is _ @xmath55-simple _ if @xmath56 for all @xmath57 . ( we can in general replace @xmath55 by any abelian group @xmath58 , and then we say @xmath33 is @xmath58-simple . ) of course , not all torus fibrations are @xmath55-simple , but it turns out that the most interesting ones which occur in the topological form of syz are . so let s assume @xmath33 and @xmath34 are @xmath55-simple . with this assumption , we obtain isomorphisms @xmath59 we can now use this to study the leray spectral sequence for @xmath33 and @xmath34 . let s make an additional assumption that @xmath5 and @xmath14 are simply connected . so in particular @xmath29 is simply connected . let s assume @xmath29 is a three - manifold . so we have the @xmath60 terms in the leray spectral sequence for @xmath33 : @xmath61 since @xmath5 is simply connected , @xmath62 , from which we conclude that @xmath63 . the same argument works for @xmath34 , and then ( [ dualityiso ] ) gives @xmath64 and similarly @xmath65 . finally , consider the possible non - zero maps for the spectral sequence : @xmath66^{d_1}&0&0&{\mathbb{r}}\\ 0&h^1(b , r^2f_*{\mathbb{r}})&h^2(b , r^2f_*{\mathbb{r}})&0\\ 0&h^1(b , r^1f_*{\mathbb{r}})\ar[rrd]^{d_2}&h^2(b , r^1f_*{\mathbb{r}})&0\\ { \mathbb{r}}&0&0&{\mathbb{r}}}\ ] ] we need one further assumption , which is again natural given the duality relationship of @xmath37 and @xmath38 : both @xmath33 and @xmath34 possess sections . ( actually , working over @xmath55 , we just need the existence of cohomology classes on @xmath5 and @xmath14 which evaluate to something non - zero on a fibre of @xmath33 or @xmath34 . ) a section intersects each fibre non - trivially , and hence gives a section of @xmath67 . since such a section also represents a cohomology class on @xmath5 , @xmath68 must be the zero map . similarly , a fibre of @xmath33 can not be homologically trivial because it intersects the section non - trivially , and thus the map @xmath69 must be zero . so the spectral sequence degenerates at @xmath60 . in particular , we get @xmath70 and @xmath71 , where @xmath72 are the hodge numbers of @xmath5 . thus the third betti number of @xmath14 is @xmath73 , so we see @xmath74 and @xmath75 . so modulo some assumptions which would of course eventually have to be justified , it is clear , at least in the three - dimensional case , why the hodge numbers are interchanged by duality . more generally , in any dimension , one might hope that @xmath55-simplicity implies @xmath76 , and then a more general exchange of hodge numbers becomes clear . see theorem [ hodgedecomp ] for a related result . this argument can be refined over @xmath77 to make new predictions about the behaviour of _ integral _ cohomology under mirror symmetry . in @xcite , theorem 3.10 , it was shown , again in the three - dimensional case , that if @xmath33 and @xmath34 are @xmath77-simple and @xmath78-simple , @xmath33 and @xmath34 have sections , and @xmath79 , then @xmath80)&\cong&h^{odd}(\check x,{\mathbb{z}}[1/2])\\ h^{odd}(x,{\mathbb{z}}[1/2])&\cong&h^{even}(\check x,{\mathbb{z}}[1/2]).\end{aligned}\ ] ] there are problems in the argument with two - torsion , but it is likely the above isomorphisms hold over @xmath77 . see @xcite for evidence for this latter conjecture . enough speculation . now let s get serious about the structure of special lagrangian fibrations . the first step in really understanding the syz conjecture is to examine the structures which arise on the base of a special lagrangian fibration . these structures arise from mclean s theorem on the moduli space of special lagrangian submanifolds @xcite , and these structures and their relationships were explained by hitchin in @xcite . we outline some of these ideas here . mclean s theorem says that the moduli space of deformations of a compact special lagrangian submanifold of a compact calabi - yau manifold @xmath5 is unobstructed , with tangent space at @xmath10 special lagrangian canonically isomorphic to the space of harmonic @xmath81-forms on @xmath18 . this isomorphism is seen explicitly as follows . let @xmath82 be a normal vector field to @xmath18 in @xmath5 . then @xmath83 and @xmath84 are both seen to be well - defined forms on @xmath18 : one needs to lift @xmath85 to a vector field but the choice is irrelevant because @xmath8 and @xmath86 restrict to zero on @xmath18 . mclean shows that if @xmath18 is special lagrangian then @xmath87 where @xmath88 denotes the hodge star operator on @xmath18 , and furthermore , @xmath85 corresponds to an infinitesimal deformation preserving the special lagrangian condition if and only if @xmath89 . this gives the correspondence between harmonic @xmath81-forms and infinitesimal special lagrangian deformations . let @xmath23 be a special lagrangian fibration with torus fibres , and assume for now that all fibres of @xmath33 are non - singular . then we obtain three structures on @xmath29 : two affine structures and a metric , as we shall now see . [ affine ] let @xmath29 be an @xmath7-dimensional manifold . an _ affine structure _ on @xmath29 is given by an atlas @xmath90 of coordinate charts @xmath91 , whose transition functions @xmath92 lie in @xmath93 . we say the affine structure is _ tropical _ if the transition functions lie in @xmath94 , i.e. have integral linear part . we say the affine structure is _ integral _ if the transition functions lie in @xmath95 . if an affine manifold @xmath29 carries a riemannian metric @xmath96 , then we say the metric is _ affine khler _ or _ hessian _ if @xmath96 is locally given by @xmath97 for some convex function @xmath98 and @xmath99 affine coordinates . then we obtain the three structures as follows : _ affine structure 1 . _ for a normal vector field @xmath85 to a fibre @xmath27 of @xmath33 , @xmath100 is a well - defined @xmath81-form on @xmath27 , and we can compute its periods as follows . let @xmath101 be a small open set , and suppose we have submanifolds @xmath102 which are families of 1-cycles over @xmath103 and such that @xmath104 form a basis for @xmath105 for each @xmath106 . consider the @xmath81-forms @xmath107 on @xmath103 defined by fibrewise integration : @xmath108 for @xmath85 a tangent vector on @xmath29 at @xmath109 , which we can lift to a normal vector field of @xmath27 . we have @xmath110 , and since @xmath8 is closed , so is @xmath111 . thus there are locally defined functions @xmath99 on @xmath103 with @xmath112 . furthermore , these functions are well - defined up to the choice of basis of @xmath105 and constants . finally , they give well - defined coordinates , as follows from the fact that @xmath113 yields an isomorphism of @xmath114 with @xmath115 by mclean s theorem . thus @xmath99 define local coordinates of a tropical affine structure on @xmath29 . _ affine structure 2 . _ we can play the same trick with @xmath86 : choose submanifolds @xmath116 which are families of @xmath117-cycles over @xmath103 and such that @xmath118 form a basis for @xmath119 . we define @xmath120 by @xmath121 , or equivalently , @xmath122 again @xmath123 are closed @xmath81-forms , with @xmath124 locally , and again @xmath125 are affine coordinates for a tropical affine structure on @xmath29 . _ the mclean metric . _ the hodge metric on @xmath115 is given by @xmath126 for @xmath127 , @xmath128 harmonic @xmath81-forms , and hence induces a metric on @xmath29 , which can be written as @xmath129 a crucial observation of hitchin @xcite is that these structures are related by the legendre transform : [ hessianmetric ] let @xmath99 be local affine coordinates on @xmath29 with respect to the affine structure induced by @xmath8 . then locally there is a function @xmath98 on @xmath29 such that @xmath130 furthermore , @xmath131 form a system of affine coordinates with respect to the affine structure induced by @xmath86 , and if @xmath132 is the legendre transform of @xmath98 , then @xmath133 and @xmath134 take families @xmath135 as above over an open neighbourhood @xmath103 with the two bases being poincar dual , i.e. @xmath136 for @xmath106 . let @xmath137 and @xmath138 be the dual bases for @xmath139 and @xmath140 respectively . from the choice of @xmath141 s , we get local coordinates @xmath99 with @xmath112 , so in particular @xmath142 so @xmath143 defines the cohomology class @xmath144 in @xmath115 . similarly , let @xmath145 then @xmath146 defines the cohomology class @xmath147 in @xmath148 , and @xmath149 . thus @xmath150 on the other hand , let @xmath151 be coordinates with @xmath152 . then @xmath153 so @xmath154 is a closed 1-form . thus there exists locally a function @xmath98 such that @xmath155 and @xmath156 . a simple calculation then confirms that @xmath157 . on the other hand , @xmath158 thus we introduce the notion of _ legendre transform _ of an affine manifold with a multi - valued convex function . [ multivaluedconvex ] let @xmath29 be an affine manifold . multi - valued _ function @xmath98 on @xmath29 is a collection of functions on an open cover @xmath159 such that on @xmath160 , @xmath161 is affine linear . we say @xmath98 is _ convex _ if the hessian @xmath162 is positive definite for all @xmath163 , in any , or equivalently all , affine coordinate systems @xmath164 . given a pair @xmath165 of affine manifold and convex multi - valued function , the _ legendre transform _ of @xmath165 is a pair @xmath166 where @xmath167 is an affine structure on the underlying manifold of @xmath29 with coordinates given locally by @xmath168 , and @xmath169 is defined by @xmath170 check that @xmath169 is also convex , and that the legendre transform of @xmath171 is @xmath165 . now let s forget about special lagrangian fibrations for the moment . instead , we see how the structures found on @xmath29 give a toy version of mirror symmetry . let @xmath29 be a tropical affine manifold . 1 . define @xmath172 to be the local system of lattices generated locally by @xmath173 , where @xmath99 are local affine coordinates . this is well - defined because transition maps are in @xmath174 . set @xmath175 this is a torus bundle over @xmath29 . in addition , @xmath176 carries a complex structure defined locally as follows . let @xmath101 be an open set with affine coordinates @xmath99 , so @xmath177 has coordinate functions @xmath99 , @xmath178 . then @xmath179 gives a system of holomorphic coordinates on @xmath180 , and the induced complex structure is independent of the choice of affine coordinates . + later we will need a variant of this : for @xmath181 , set @xmath182 this has a complex structure with coordinates given by @xmath183 ( as we shall see later , the limit @xmath184 corresponds to a large complex structure limit . ) 2 . define @xmath185 to be the local system of lattices generated locally by @xmath186 , with @xmath99 local affine coordinates . set @xmath187 of course @xmath188 carries a canonical symplectic structure , and this symplectic structure descends to @xmath189 . we write @xmath190 and @xmath191 for these torus fibrations ; these are clearly dual . now suppose in addition we have a hessian metric @xmath96 on @xmath29 , with local potential function @xmath98 . then in fact both @xmath176 and @xmath189 become khler manifolds : @xmath192 is a ( local ) khler potential on @xmath176 , defining a khler form @xmath193 . this metric is ricci - flat if and only if @xmath98 satisfies the real monge - ampre equation @xmath194 working locally with affine coordinates @xmath195 and complex coordinates @xmath196 , we compute @xmath197 which is clearly positive . furthermore , if @xmath198 , then @xmath199 is proportional to @xmath200 if and only if @xmath201 is constant . we write this khler manifold as @xmath202 . dually we have in local canonical coordinates @xmath203 on @xmath188 , the functions @xmath204 on @xmath188 induce a well - defined complex structure on @xmath189 , with respect to which the canonical symplectic form @xmath8 is a khler form of a metric . furthermore this metric is ricci - flat if and only if @xmath98 satisfies the real monge - ampre equation @xmath205 it is easy to see that an affine linear change in the coordinates @xmath206 ( and hence an appropriate change in the coordinates @xmath207 ) results in a linear change of the coordinates @xmath208 , so they induce a well - defined complex structure invariant under @xmath209 , and hence a complex structure on @xmath189 . then one computes that @xmath210 where @xmath211 . then the metric is ricci - flat if and only if @xmath212 , if and only if @xmath213 . as before , we call this khler manifold @xmath214 . this motivates the definition an affine manifold with metric of hessian form is a _ monge - ampre manifold _ if the local potential function @xmath98 satisfies the monge - ampre equation @xmath215 . monge - ampre manifolds were first studied by cheng and yau in @xcite . [ caniso ] show that the identification of @xmath216 and @xmath188 given by a hessian metric induces a canonical isomorphism @xmath217 of khler manifolds , where @xmath171 is the legendre transform of @xmath165 . finally , we note that a @xmath29-field can be introduced into this picture . to keep life relatively simple ( so as to avoid having to pass to generalized complex structures @xcite , @xcite , @xcite ) , we view the @xmath29-field as an element @xmath218 , where @xmath219 . noting that a section of @xmath220 over an open set @xmath103 can be viewed as a section of @xmath221 , such a section acts on @xmath221 via translation , and this action is in fact holomorphic with respect to the standard semi - flat complex structure . thus a ech 1-cocycle @xmath222 representing @xmath223 allows us to reglue @xmath176 via translations over the intersections @xmath224 . this gives a new complex manifold @xmath225 . if in addition there is a multi - valued potential function @xmath98 defining a metric , these translations preserve the metric and yield a khler manifold @xmath226 . thus the full toy version of mirror symmetry is as follows . the data consists of an affine manifold @xmath29 with potential @xmath98 and @xmath29-fields @xmath227 , @xmath228 . now it is not difficult to see , and you will have seen this already if you ve done exercise [ caniso ] , that the local system @xmath229 defined using the affine structure on @xmath29 is the same as the local system @xmath46 defined using the affine stucture on @xmath167 . so we say the pair @xmath230 is mirror to @xmath231 this provides a reasonably fulfilling picture of mirror symmetry in a simple context . many more aspects of mirror symmetry can be worked out in this semi - flat context , see @xcite . however , ultimately this only sheds limited insight into the general case . the only compact calabi - yau manifolds with semi - flat ricci - flat metric which arise in this way are complex tori ( shown by cheng and yau in @xcite ) . to deal with more interesting cases , we need to allow singular fibres , and hence , singularities in the affine structure of @xmath29 . to deal with singular fibres , we define a _ ( tropical , integral ) affine manifold with singularities _ is a @xmath232 manifold @xmath29 with an open subset @xmath30 which carries a ( tropical , integral ) affine structure , and such that @xmath233 is a locally finite union of locally closed submanifolds of codimension @xmath234 . by way of example , let s explain how the batyrev construction gives rise to a wide class of such manifolds . this construction is taken from @xcite , where a more combinatorially complicated version is given for complete intersections ; see @xcite and @xcite for an alternative construction . let @xmath235 be a reflexive polytope in @xmath236 , where @xmath237 ; let @xmath238 be the dual lattice , @xmath239 the dual polytope given by @xmath240 we assume @xmath241 is the unique interior lattice point of @xmath235 . let @xmath242 be the normal fan to @xmath17 , consisting of cones over the faces of @xmath235 . suppose we are given a star subdivision of @xmath235 , with all vertices being integral points , inducing a subdivision @xmath243 of the fan @xmath242 . in addition suppose that @xmath244 is an ( upper ) strictly convex piecewise linear function on the fan @xmath243 . also , let @xmath245 be the piecewise linear function representing the anti - canonical class of the toric variety @xmath246 ; i.e. @xmath247 takes the value @xmath81 on the primitive generator of each one - dimensional cone of @xmath242 . finally , assume that @xmath248 is chosen so that @xmath249 is a ( not necessarily strictly ) convex function . define , for any convex piecewise linear function @xmath250 on the fan @xmath243 , the newton polytope of @xmath250 , @xmath251 in particular , @xmath252 where @xmath253 denotes minkowski sum . our goal will be to put an affine structure with singularities on @xmath254 . our first method of doing this requires no choices . let @xmath255 be the set of proper faces of @xmath256 . furthermore , let @xmath257 denote the first barycentric subdivision of @xmath255 and let @xmath258 be the union of all simplices of @xmath257 not containing a vertex of @xmath255 ( a zero - dimensional cell ) or intersecting the interior of a maximal cell of @xmath255 . if we then set @xmath259 , we can define an affine structure on @xmath260 as follows . @xmath260 has an open cover @xmath261 where @xmath262 , the interior of @xmath263 , and @xmath264 is the ( open ) star of @xmath265 in @xmath257 . we define an affine chart @xmath266 given by the inclusion of @xmath267 in @xmath268 , the affine hyperplane containing @xmath263 . also , take @xmath269 to be the projection , where @xmath265 , being a vertex of @xmath256 , can be written uniquely as @xmath270 with @xmath271 a vertex of @xmath17 and @xmath272 a vertex of @xmath273 . one checks easily that for @xmath274 , @xmath275 is affine linear with integral linear part ( integrality follows from reflexivity of @xmath235 ! ) so @xmath29 is a tropical affine manifold with singularities . furthermore , if @xmath248 was chosen to have integral slopes , then @xmath29 is integral . we often would like to refine this construction , to get a finer polyhedral decomposition @xmath255 of @xmath29 and with it a somewhat more interesting discriminant locus @xmath276 . one reason for doing so is that this construction is clearly not mirror symmetric , as it depends only on a star subdivision of @xmath235 and not of @xmath17 . furthermore , a maximal star subdivision of @xmath17 corresponds to what batyrev terms a mpcp ( maximal projective crepant partial ) resolution of @xmath277 , and normally , we will wish to study hypersurfaces in a mpcp resolution of @xmath277 rather than in @xmath277 itself . to introduce this extra degree of flexibility , we need to make some choices , which is done as follows . first , choose a star subdivision of @xmath17 , with all vertices being integral points , inducing a refinement @xmath278 of the fan @xmath279 which is the normal fan to @xmath235 . this induces a polyhedral subdivision of @xmath280 , and we write the collection of cells of this subdivision as @xmath281 . note that because @xmath282 , we have @xmath283 a subdivision @xmath255 of @xmath284 is _ good _ with respect to @xmath281 if it is induced by a subdivision @xmath285 of @xmath256 satisfying the following three properties : 1 . @xmath273 is a union of cells in @xmath285 . all vertices of @xmath285 are contained either in @xmath284 or in @xmath273 . every cell @xmath286 with @xmath287 and @xmath288 can be written as @xmath289 with @xmath290 and @xmath291 the corresponding cone in @xmath278 . if @xmath248 has integral slopes and all vertices of @xmath285 are integral , then we say @xmath255 is integral . the following picture shows what such a good subdivision may look like , in the case that @xmath17 is the newton polytope of @xmath292 : given a good decomposition @xmath255 of @xmath293 we once again obtain an affine structure with singularities on @xmath29 , much as before , defining the discriminant locus @xmath258 in terms of the first barycentric subdivision @xmath257 of this new polyhedral decomposition @xmath255 . then as before @xmath259 has an open cover @xmath261 where @xmath262 , the interior of @xmath263 , and @xmath264 is the ( open ) star of @xmath265 in @xmath257 . we define an affine chart @xmath266 given by the inclusion of @xmath267 in @xmath268 , the affine hyperplane containing @xmath263 . also , take @xmath294 to be the projection , where @xmath265 can be written uniquely as @xmath270 with @xmath271 an integral point of @xmath17 and @xmath295 . as before , one checks easily that for @xmath274 , @xmath275 is affine linear with integral linear part so @xmath29 is a tropical affine manifold with singularities . furthermore , if @xmath248 has integral slopes , and @xmath255 is integral , then the affine structure on @xmath29 is in fact integral . [ quintic ] let @xmath296 be the convex hull of the points @xmath297 so @xmath17 is the convex hull of the points @xmath298 take @xmath299 and choose a star triangulation of @xmath17 . in this case it is easy to see the affine structure on @xmath260 in fact extends across the interior of all three - dimensional faces of @xmath17 . this gives a smaller discriminant locus @xmath276 which , given a nice regular triangulation of @xmath17 , looks like the following picture in a neighbourhood of a @xmath301-face of @xmath17 : the light lines giving the triangulation and the dark lines the discriminant locus @xmath276 . in this picture , @xmath276 is a trivalent graph , with two types of trivalent vertices . the ones along the edge are non - planar : the two additional legs of @xmath276 drawn in this picture are contained in other two - faces of @xmath17 . however @xmath276 is planar in the interior of this two - face . in general , if the subdivisions @xmath278 and @xmath243 of @xmath279 and @xmath242 respectively represent maximal projective crepant partial resolutions of @xmath277 and @xmath246 , and @xmath302 , then only these sorts of trivalent vertices occur . more specifically , one knows what the monodromy of the local systems @xmath46 and @xmath229 on @xmath260 look like at these vertices . if @xmath303 is a vertex contained in the interior of a two - face , then it is clear that the tangent space to that two - face is invariant under parallel transport , in a neighbourhood of @xmath265 , of @xmath46 . a more careful analysis yields that the monodromy matrices for @xmath46 take the form , in a suitable basis , @xmath304 here @xmath305 are given by parallel transport about loops around the three edges of @xmath276 coming out of @xmath265 . of course , the monodromy of @xmath229 is the transpose inverse of these matrices . similarly , if @xmath265 is a vertex of @xmath276 contained in an edge of @xmath256 , then the monodromy will take the form @xmath306 so we see that the monodromy of the two types of vertices are interchanged by looking at @xmath46 and @xmath229 . one main result of @xcite is if @xmath29 is a three - dimensional tropical affine manifold with singularities such that @xmath276 is trivalent and the monodromy of @xmath46 at each vertex is one of the above two types , then @xmath307 can be compactified to a topological fibration @xmath190 . dually , @xmath308 can be compactified to a topological fibration @xmath191 . we wo nt give any details here of how this is carried out , but it is not particularly difficult , as long as one restricts to the category of topological ( not @xmath309 ) manifolds . however , it is interesting to look at the singular fibres we need to add in this compactification . if @xmath310 is a point which is not a vertex of @xmath276 , then @xmath311 is homeomorphic to @xmath312 , where @xmath313 denotes a kodaira type @xmath313 elliptic curve , i.e. a pinched torus . if @xmath109 is a vertex of @xmath276 , with monodromy of the first type , then @xmath314 , with @xmath315 if @xmath316 or @xmath317 , where @xmath318 is identified with the unit circle in @xmath319 this is the three - dimensional analogue of a pinched torus , and @xmath320 . we call this a _ positive _ fibre . if @xmath109 is a vertex of @xmath276 , with monodromy of the second type , then @xmath311 can be described as @xmath321 , with @xmath315 if @xmath316 or @xmath317 , @xmath322 , or @xmath323 . the singular locus of this fibre is a figure eight , and @xmath324 . we call this a _ negative _ fibre . so we see a very concrete local consequence of syz duality : namely in the compactifications @xmath176 and @xmath189 , the positive and negative fibres are interchanged . of course , this results in the observation that euler characteristic changes sign under mirror symmetry for calabi - yau threefolds . continuing with example [ quintic ] , it was proved in @xcite that @xmath189 is homeomorphic to the quintic and @xmath176 is homeomorphic to the mirror quintic . haase and zharkov in @xcite gave a different description of what is the same affine structure . their construction has the advantage that it is manifestly dual . in other words , in our construction , we can interchange the role of @xmath235 and @xmath17 to get two different affine manifolds , with @xmath325 the affine manifold with singularities structure on @xmath326 and @xmath327 the affine manifold with singularities structure on @xmath284 . it is not obvious that these are `` dual '' affine manifolds , at least in the sense that @xmath328 is homeomorphic to @xmath329 and @xmath330 is homeomorphic to @xmath331 . in the construction given above , this follows from the discrete legendre transform we will discuss in 7 . on the other hand , the construction i give here will arise naturally from the degeneration construction discussed later in this paper . ruan in @xcite gave a description of _ lagrangian _ torus fibrations for hypersurfaces in toric varieties using a symplectic flow argument , and his construction should coincide with a _ symplectic _ compactification of the symplectic manifolds @xmath332 . in the three - dimensional case , such a symplectic compactification has now been constructed by ricardo castao - bernard and diego matessi @xcite . if this compactification is applied to the affine manifolds with singularities described here , the resulting symplectic manifolds should be symplectomorphic to the corresponding toric hypersurface , but this has not yet been shown . i should also point out that the explicit compactifications mentioned in three dimensions can be carried out in all dimensions , and will be done so in @xcite . we will show there in a much more general context that these compactifications are then homeomorphic to the expected calabi - yau manifolds . the previous section demonstrates that the syz conjecture gives a beautiful description of mirror symmetry at a purely topological level . this , by itself , can often be useful , but unfortunately is not strong enough to get at really interesting aspects of mirror symmetry , such as instanton corrections . for a while , though , many of us were hoping that the strong version of duality we have just seen would hold at the special lagrangian level . this would mean that a mirror pair @xmath333 would possess special lagrangian torus fibrations @xmath23 and @xmath24 with codimension two discriminant locus , and the discriminant loci of @xmath33 and @xmath34 would coincide . these fibrations would then be dual away from the discriminant locus . there are examples of special lagrangian fibrations on non - compact toric varieties @xmath5 with this behaviour . in particular , if @xmath334 with a @xmath335 action on @xmath5 preserving the holomorphic @xmath7-form , and if @xmath5 in addition carries a ricci - flat metric which is invariant under this action , then @xmath5 will have a very nice special lagrangian fibration with codimension two discriminant locus . ( see @xcite and @xcite ) . however , dominic joyce ( @xcite and other papers cited therein ) began studying some three - dimensional @xmath318-invariant examples , and discovered quite different behaviour . there is an argument that if a special lagrangian fibration is @xmath309 , then the discriminant locus will be ( hausdorff ) codimension two . however , joyce discovered examples which were not differentiable , but only piecewise differentiable , and furthermore , had a codimension one discriminant locus : define @xmath336 by @xmath337 with @xmath338 and @xmath339 it is easy to see that if @xmath340 , then @xmath341 is homeomorphic to @xmath342 , while if @xmath343 , then @xmath341 is a cone over @xmath344 : essentially , one copy of @xmath318 in @xmath342 collapses to a point . in addition , all fibres of this map are special lagrangian , and it is obviously only piecewise smooth . the discriminant locus is the entire plane given by @xmath343 . this example forces a reevaluation of the strong form of the syz conjecture . in further work joyce found evidence for a more likely picture for general special lagrangian fibrations in three dimensions . the discriminant locus , instead of being a codimension two graph , will be a codimension one blob . typically the union of the singular points of singular fibres will be a riemann surface , and it will map to an amoeba shaped set in @xmath29 , i.e.the discriminant locus looks like the picture on the right rather than the left , and will be a fattening of the old picture of a codimension two discriminant . joyce made some additional arguments to suggest that this fattened discriminant locus must look fundamentally different in a neighbourhood of the two basic types of vertices we saw in the previous section , with the two types of vertices expected to appear pretty much as depicted in the above picture . thus the strong form of duality mentioned above , where we expect the discriminant loci of the special lagrangian fibrations on a mirror pair to be the same , can not hold . if this is the case , one needs to replace this strong form of duality with a weaker form . it seems likely that the best way to rephrase the syz conjecture is in a limiting form . mirror symmetry as we currently understand it has to do with degenerations of calabi - yau manifolds . given a flat family @xmath345 over a disk @xmath2 , with the fibre @xmath346 over @xmath347 singular and all other fibres @xmath7-dimensional calabi - yau manifolds , we say the family is _ maximally unipotent _ if the monodromy transformation @xmath348 ( @xmath349 non - zero ) satisfies @xmath350 but @xmath351 . it is a standard fact of mirror symmetry that mirrors should be associated to maximally unipotent degenerations of calabi - yau manifolds . in particular , given two different maximally unipotent degenerations in a single complex moduli space for some calabi - yau manifold , one might obtain different mirror manifolds . sometimes these different mirror manifolds are birationally equivalent , as studied in @xcite , or are genuinely different , see @xcite . we recall the definition of gromov - hausdorff convergence , a notion of convergence of a sequence of metric spaces . let @xmath352 , @xmath353 be two compact metric spaces . suppose there exists maps @xmath354 and @xmath355 ( not necessarily continuous ) such that for all @xmath356 , @xmath357 and for all @xmath358 , @xmath359 and the two symmetric properties for @xmath360 hold . then we say the gromov hausdorff distance between @xmath5 and @xmath360 is at most @xmath361 . the gromov hausdorff distance @xmath362 is the infimum of all such @xmath361 . it follows from results of gromov ( see for example @xcite , pg . 281 , cor . 1.11 ) that the space of compact ricci - flat manifolds with diameter @xmath363 is precompact with respect to gromov - hausdorff distance , i.e. any sequence of such manifolds has a subsequence converging with respect to the gromov - hausdorff distance to a metric space . this metric space could be quite bad ; this is quite outside the realm of algebraic geometry ! nevertheless , this raises the following natural question . given a maximally unipotent degeneration of calabi - yau manifolds @xmath364 , take a sequence @xmath365 converging to @xmath347 , and consider a sequence @xmath366 , where @xmath367 is a choice of ricci - flat metric chosen so that @xmath368 remains bounded . what is the gromov - hausdorff limit of @xmath369 , or the limit of some convergent subsequence ? consider a degenerating family of elliptic curves , with periods @xmath81 and @xmath370 . if we take @xmath371 approaching @xmath347 along the positive real axis , then we can just view this as a family of elliptic curves @xmath372 with period @xmath81 and @xmath373 with @xmath374 . if we take the standard euclidean metric @xmath96 on @xmath372 , then the diameter of @xmath372 is unbounded . to obtain a bounded diameter , we replace @xmath96 by @xmath375 ; equivalently , we can keep @xmath96 fixed on @xmath319 but change the periods of the elliptic curve to @xmath376 . it then becomes clear that the gromov - hausdorff limit of such a sequence of elliptic curves is a circle @xmath377 . this simple example motivates the first conjecture about maximally unipotent degenerations , conjectured independently by myself and wilson on the one hand @xcite and kontsevich and soibelman @xcite on the other . [ gwksconj ] let @xmath364 be a maximally unipotent degeneration of simply - connected calabi - yau manifolds with full @xmath378 holonomy , @xmath365 with @xmath379 , and let @xmath380 be a ricci - flat metric on @xmath381 normalized to have fixed diameter @xmath382 . then a convergent subsequence of @xmath383 converges to a metric space @xmath384 , where @xmath385 is homeomorphic to @xmath386 . furthermore , @xmath387 is induced by a riemannian metric on @xmath388 , where @xmath389 is a set of codimension two . here the topology of the limit depends on the nature of the non - singular fibres @xmath390 ; for example , if instead @xmath390 was hyperkhler , then we would expect the limit to be a projective space . also , even in the case of full @xmath378 holonomy , if @xmath390 is not simply connected , we would expect limits such as @xmath391-homology spheres to arise . conjecture [ gwksconj ] is directly inspired by the syz conjecture . suppose we had special lagrangian fibrations @xmath392 . then as the maximally unipotent degeneration is approached , it is possible to see that the volume of the fibres of these fibrations go to zero . this would suggest these fibres collapse , hopefully leaving the base as the limit . this conjecture was proved by myself and wilson for k3 surfaces in @xcite . the proof relies on a detailed analysis of the behaviour of ricci - flat metrics in the limit , and also on the existence of explicit local models for ricci - flat metrics near singular fibres of special lagrangian fibrations . the motivation for this conjecture from syz also provides a limiting form of the conjecture . there are any number of problems with trying to prove the existence of special lagrangian fibrations on calabi - yau manifolds . even the existence of a single special lagrangian torus near a maximally unipotent degeneration is unknown , but we expect it should be easier to find them as we approach the maximally unipotent point . furthermore , even if we find a special lagrangian torus , we know that it moves in an @xmath7-dimensional family , but we do nt know its deformations fill out the entire manifold . in addition , there is no guarantee that even if it does , we obtain a foliation of the manifold : nearby special lagrangian submanifolds may intersect . ( for an example , see @xcite . ) so instead , we will just look at the moduli space of special lagrangian tori . suppose , given @xmath379 , that for @xmath393 sufficiently close to zero , there is a special lagrangian @xmath394 whose homology class is invariant under monodromy , or more specifically , generates the space @xmath395 of the monodromy weight filtration ( this is where we expect to find fibres of a special lagrangian fibration associated to a maximally unipotent degeneration ) . let @xmath396 be the moduli space of deformations of this torus ; every point of @xmath396 corresponds to a smooth special lagrangian torus in @xmath381 . manifold then comes equipped with the mclean metric and affine structures defined in 2 . one can then compactify @xmath397 , ( probably by taking the closure of @xmath396 in the space of special lagrangian currents ; the details are nt important here ) . this gives a series of metric spaces @xmath398 with the metric @xmath399 induced by the mclean metric . if the mclean metric is normalized to keep the diameter of @xmath400 constant independent of @xmath163 , then we can hope that @xmath398 converges to a compact metric space @xmath401 . here then is the limiting form of syz : if @xmath383 converges to @xmath402 and @xmath398 is non - empty for large @xmath163 and converges to @xmath401 , then @xmath403 and @xmath385 are isometric up to scaling . furthermore , there is a subspace @xmath404 with @xmath405 of hausdorff codimension 2 in @xmath403 such that @xmath406 is a monge - ampre manifold , with the monge - ampre metric inducing @xmath387 on @xmath406 . essentially what this is saying is that as we approach the maximally unipotent degeneration , we expect to have a special lagrangian fibration on larger and larger subsets of @xmath381 . furthermore , in the limit , the codimension one discriminant locus suggested by joyce converges to a codimension two discriminant locus , and ( the not necessarily monge - ampre , see @xcite ) hessian metrics on @xmath396 converge to a monge - ampre metric . the main point i want to get at here is that it is likely the syz conjecture is only `` approximately '' correct , and one needs to look at the limit to have a hope of proving anything . on the other hand , the above conjecture seems likely to be accessible by currently understood techniques , though with a lot of additional work , and i would nt be surprised to see it proved in the next few years . how do we do mirror symmetry using this modified version of the syz conjecture ? essentially , we would follow these steps : 1 . we begin with a maximally unipotent degeneration of calabi - yau manifolds @xmath364 , along with a choice of polarization . this gives us a khler class @xmath407\in h^2({{\mathcal{x}}}_t,{\mathbb{r}})$ ] for each @xmath408 , represented by @xmath409 the khler form of a ricci - flat metric @xmath410 . 2 . identify the gromov - hausdorff limit of a sequence @xmath411 with @xmath379 , and @xmath412 a scale factor which keeps the diameter of @xmath381 constant . the limit will be , if the above conjectures work , an affine manifold with singularities @xmath29 along with a monge - ampre metric . 3 . perform a legendre transform to obtain a new affine manifold with singularities @xmath167 , though with the same metric . 4 . try to construct a compactification of @xmath413 for small @xmath181 to obtain a complex manifold @xmath414 . this will be the mirror manifold . actually , we need to elaborate on this last step a bit more . the problem is that while we expect that it should be possible in general to construct symplectic compactifications of the symplectic manifold @xmath332 ( and hence get the mirror as a symplectic manifold ) , we do nt expect to be able to compactify @xmath413 as a complex manifold . instead , the expectation is that a small deformation of @xmath413 is necessary before it can be compactified . furthermore , this small deformation is critically important in mirror symmetry : _ it is this small deformation which provides the @xmath29-model instanton corrections_. because of the importance of this last issue , it has already been studied by several authors : fukaya in @xcite has studied the problem directly using heuristic ideas , while kontsevich and soibelmann @xcite have modified the problem of passing from an affine manifold to a complex manifold by instead producing a non - archimedean space . we will return to these issues later in this paper , when i discuss my own work with siebert which has been partly motivated by the same problem . because this last item is so important , let s give it a name : [ reconstruct1 ] given a tropical affine manifold with singularities @xmath29 , construct a complex manifold @xmath415 which is a compactification of a small deformation of @xmath416 . i do not wish to dwell further on this version of the syz conjecture here , because it lies mostly in the realm of analysis and differential geometry and the behaviour of ricci - flat metrics , and will give us little insight into what makes subtler aspects of traditional mirror symmetry work : for example , how exactly do instanton corrections arise ? so we move on to explain how the limiting form of syz inspired a more algebro - geometric form of syz , which in turn avoids all analytic problems and holds out great promise for understanding the fundamental mysteries of mirror symmetry . we now have two notions of limit : the familiar algebro - geometric notion of a flat degenerating family @xmath364 over a disk on the one hand , and the gromov - hausdorff limit on the other . kontsevich had an important insight ( see @xcite ) into the connection between these two . in this section i will give a rough idea of how and why this works . very roughly speaking , the gromov - hausdorff limit @xmath369 as @xmath379 should coincide , topologically , with the dual intersection complex of the singular fibre @xmath346 . more precisely , in a relatively simple situation , suppose @xmath345 is relatively minimal ( in the sense of mori ) and normal crossings , with @xmath346 having irreducible components @xmath417 . the dual intersection complex of @xmath346 is the simplicial complex with vertices @xmath418 , and which contains a simplex @xmath419 if @xmath420 . let us explain roughly why this should be , first by looking at a standard family of degenerating elliptic curves with periods @xmath81 and @xmath421 for @xmath7 a positive integer . such a family over the punctured disk is extended to a family over the disk by adding an @xmath422 ( a cycle of @xmath7 rational curves ) fibre over the origin . taking a sequence @xmath379 with @xmath393 real and positive gives a sequence of elliptic curves of the form @xmath423 where : @xmath424 and @xmath425 . in addition , the metric on @xmath423 , properly scaled , comes from the constant hessian metric on @xmath29 . so we wish to explain how @xmath29 is related to the geometry near the singular fibre . to this end , let @xmath426 be the irreducible components of @xmath346 ; these are all @xmath427 s . let @xmath428 be the singular points of @xmath346 . we ll consider two sorts of open sets in @xmath429 . for the first type , choose a coordinate @xmath430 on @xmath431 , with @xmath432 given by @xmath433 and @xmath434 given by @xmath435 . let @xmath436 be the open set @xmath437 for some small fixed @xmath438 . then one can find a neighbourhood @xmath439 of @xmath440 in @xmath429 such that @xmath439 is biholomorphic to @xmath441 for @xmath442 sufficiently small , @xmath443 a disk of radius @xmath444 in @xmath319 , and @xmath445 is the projection onto @xmath443 . on the other hand , each @xmath432 has a neighbourhood @xmath446 in @xmath429 biholomorphic to a polydisk @xmath447 on which @xmath33 takes the form @xmath448 . if @xmath438 and @xmath449 are chosen correctly , then for @xmath371 sufficiently close to zero , @xmath450 form an open cover of @xmath390 . now each of the sets in this open cover can be written as @xmath451 for some @xmath103 a one - dimensional ( non - compact ) affine manifold and @xmath452 . if @xmath103 is an open interval @xmath453 , then @xmath451 is biholomorphic to the annulus @xmath454 as @xmath455 is a holomorphic coordinate on @xmath456 . thus @xmath457 with @xmath452 . as @xmath458 , the interval @xmath459 shrinks to a point . so @xmath460 is a smaller and smaller open subset of @xmath390 as @xmath458 when we view things in this way . this argument suggests that every irreducible component should be associated to a point on @xmath29 . now look at @xmath461 . this is @xmath462 with @xmath452 . this interval approaches the unit interval @xmath463 as @xmath458 . so the open set @xmath464 ends up being a large portion of @xmath390 . we end up with @xmath390 , for small @xmath371 , being a union of open sets of the form @xmath465 ( i.e. @xmath466 ) and @xmath467 ( i.e. @xmath460 ) for @xmath468 , @xmath469 sufficiently small . these should glue , at least approximately , to give @xmath415 . so we see that irreducible components of @xmath346 seem to coincide with points on @xmath29 , but intersections of components coincide with lines . in this way we see the dual intersection complex emerge . let us make one more observation before beginning with rigorous results in the next section . suppose more generally we had a _ gorenstein toroidal crossings _ degeneration of calabi - yau manifolds @xmath345 . this means that every point @xmath470 has a neighbourhood isomorphic to an open set in an affine gorenstein ( i.e. the canonical class is a cartier divisor ) toric variety , with @xmath33 given locally by a monomial which vanishes exactly to order @xmath81 on each codimension one toric stratum . this is a generalization of the notion of normal crossings , see @xcite . very roughly , the above argument suggests that each irreducible component of the central fibre will correspond to a point of the gromov - hausdorff limit . the following exercise shows what kind of contribution to @xmath29 to expect from a point @xmath471 which is a zero - dimensional stratum in @xmath346 . [ gorensteinlimit ] suppose there is a point @xmath471 which has a neighbourhood isomorphic to a neighbourhood of a dimension zero torus orbit of an affine gorenstein toric variety @xmath472 . such an affine variety is specified as follows . set @xmath237 , @xmath236 , @xmath473 , @xmath474 as in 4 . then there is a lattice polytope @xmath475 , @xmath476 , @xmath477 , @xmath478 the monoid determined by the dual of the cone @xmath479 , and finally , @xmath480 $ ] , and @xmath33 coincides with the monomial @xmath481 . let us now take a small neighbourhood of @xmath482 of the form @xmath483\,|\,\hbox{$|z^p|<\delta$ for all $ p\in p$}\}.\ ] ] this is an open set as the condition @xmath484 can be tested on a finite generating set for @xmath485 , provided that @xmath486 . then show that for a given @xmath371 , @xmath487 and @xmath488 , if @xmath489 then @xmath490 note that @xmath491 so @xmath492 is an open subset of @xmath263 , and as @xmath458 , @xmath492 converges to the interior of @xmath263 . this observation will hopefully motivate the basic construction of the next section . we now return to rigorous statements . i would like to explain the basic ideas behind the program launched in @xcite . while i will use the previous sections as motivation , this work actually got its start when siebert began a program of studying mirror symmetry via degenerations of calabi - yau manifolds . work of schrer and siebert @xcite , @xcite led siebert to the idea that log structures on degenerations of calabi - yau manifolds would allow one to view mirror symmetry as an operation performed on degenerate calabi - yau varieties . siebert observed that at a combinatorial level , mirror symmetry exchanged data pertaining to the log structure and a polarization . this will be explained more clearly in the following section , where we introduce log structures . together , siebert and i realised that the combinatorial data he was considering could be encoded naturally in the dual intersection complex of the degeneration , and that mirror symmetry then corresponded to a discrete legendre transform on the dual intersection complex . it then became apparent that this approach provided an algebro - geometrization of the syz conjecture . here i will explain this program from the opposite direction , starting with the motivation of the previous section for introducing the dual intersection complex , and then work backwards until we arrive naturally at log structures . much of the material in this section comes from @xcite , 4 . [ toricdegen ] let @xmath345 be a proper flat family of relative dimension @xmath7 , where @xmath2 is a disk and @xmath429 is a complex analytic space ( not necessarily non - singular ) . we say @xmath33 is a _ toric degeneration _ of calabi - yau varieties if 1 . @xmath390 is an irreducible normal calabi - yau variety with only canonical singularities for @xmath493 . ( the reader may like to assume @xmath390 is smooth for @xmath493 ) . if @xmath494 is the normalization , then @xmath495 is a disjoint union of toric varieties , the conductor locus @xmath496 is reduced , and the map @xmath497 is unramified and generically two - to - one . ( the conductor locus is a naturally defined scheme structure on the set where @xmath85 is not an isomorphism . ) the square @xmath498 is cartesian and cocartesian . 3 . @xmath346 is a reduced gorenstein space and the conductor locus @xmath382 restricted to each irreducible component of @xmath495 is the union of all toric weil divisors of that component . 4 . there exists a closed subset @xmath499 of relative codimension @xmath234 such that @xmath500 satisfies the following properties : @xmath500 does not contain the image under @xmath85 of any toric stratum of @xmath495 , and for any point @xmath501 , there is a neighbourhood @xmath502 ( in the analytic topology ) of @xmath482 , an @xmath503-dimensional affine toric variety @xmath472 , a regular function @xmath504 on @xmath472 given by a monomial , and a commutative diagram @xmath505 where @xmath506 and @xmath507 are open embeddings and @xmath508 . furthermore , @xmath504 vanishes precisely once on each toric divisor of @xmath472 . take @xmath429 to be defined by the equation @xmath509 in @xmath510 , where @xmath2 is a disk with coordinate @xmath371 and @xmath511 is a general homogeneous quartic polynomial on @xmath512 . it is easy to see that @xmath429 is singular at the locus @xmath513 as @xmath346 is the coordinate tetrahedron , the singular locus of @xmath346 consists of the six coordinate lines of @xmath512 , and @xmath429 has four singular points along each such line , for a total of 24 singular points . take @xmath514 . then away from @xmath500 , the projection @xmath364 is normal crossings , which yields condition ( 4 ) of the definition of toric degeneration . it is easy to see all other conditions are satisfied . [ toricdegenexample2 ] let @xmath515 be a reflexive polytope with dual @xmath516 . choose an _ integral _ strictly convex piecewise linear function @xmath517 as in 4 . consider the family in @xmath518 defined by @xmath519 here @xmath520 denotes the section of @xmath521 determined by the lattice point @xmath522 and @xmath523 is a general choice of coefficient . so @xmath524 is the section which vanishes once on each toric boundary component of @xmath277 . now for most choices of @xmath248 , this does not define a toric degeneration : the singularities are too bad . however , there is a natural toric variety in which to describe this degeneration . set @xmath525 the assumption that @xmath526 is convex implies @xmath527 is convex , but of course is non - compact . let @xmath528 be the normal fan to @xmath527 in @xmath529 . let @xmath530 denote the toric variety defined by the fan @xmath528 . then @xmath527 is the newton polytope of a line bundle @xmath531 on @xmath530 , and terms of the form @xmath532 can be interpreted as sections of this line bundle , corresponding to @xmath533 . in addition , there is a natural map @xmath534 defined by projection onto @xmath55 , and this map defines the regular function @xmath371 . thus ( [ toricdegenexam ] ) defines a hypersurface @xmath535 and @xmath371 defines a map @xmath536 . this is a toric degeneration . without going into much detail , choosing a star decomposition of @xmath17 as in 4 and a good polyhedral decomposition @xmath255 of @xmath284 is essentially the same as choosing a refinement @xmath537 of @xmath278 with particularly nice properties . this yields a partial resolution @xmath538 . then the proper transform of @xmath539 in @xmath540 , @xmath541 , yields another toric degeneration @xmath542 . this is necessary to get a toric degeneration whose general fibre is a mpcp resolution of a hypersurface in a toric variety . for proofs , see @xcite , where the construction is generalized to complete intersections in toric varieties . given a toric degeneration @xmath345 , we can build the _ dual intersection complex _ @xmath543 of @xmath33 , as follows . here @xmath29 is an integral affine manifold with singularities , and @xmath255 is a _ polyhedral decomposition _ of @xmath29 , a decomposition of @xmath29 into lattice polytopes . in fact , we will construct @xmath29 as a union of lattice polytopes . specifically , let the normalisation of @xmath346 , @xmath544 , be written as a disjoint union @xmath545 of toric varieties @xmath431 , @xmath546 the normalisation . the _ strata _ of @xmath346 are the elements of the set @xmath547 here by toric stratum we mean the closure of a @xmath548 orbit . let @xmath549 be a zero - dimensional stratum . applying definition [ toricdegen ] ( 4 ) to a neighbourhood of @xmath482 , there is a toric variety @xmath472 such that in a neighbourhood of @xmath482 , @xmath345 is locally isomorphic to @xmath550 , where @xmath504 is given by a monomial . now the condition that @xmath504 vanishes precisely once along each toric divisor of @xmath472 is the statement that @xmath472 is gorenstein , and as such , it arises as in exercise [ gorensteinlimit ] . indeed , let @xmath551 be as usual , with @xmath552 . then there is a lattice polytope @xmath553 such that @xmath554 is the cone defining the toric variety @xmath472 . as we saw in exercise [ gorensteinlimit ] , a small neighbourhood of @xmath482 in @xmath429 should contribute a copy of @xmath555 to @xmath29 , which provides the motivation for our construction . we can now describe how to construct @xmath29 by gluing together the polytopes @xmath556 we will do this in the case that every irreducible component of @xmath346 is in fact itself normal so that @xmath557 is an isomorphism . the reader may be able to imagine the more general construction . with this normality assumption , there is a one - to - one inclusion reversing correspondence between faces of @xmath555 and elements of @xmath558 containing @xmath482 . we can then identify faces of @xmath555 and @xmath559 if they correspond to the same strata of @xmath346 . some argument is necessary to show that this identification can be done via an integral affine transformation , but again this is not difficult . making these identifications , one obtains @xmath29 . one can then prove if @xmath346 is complex @xmath7 dimensional , then @xmath29 is an real @xmath7 dimensional manifold . see @xcite , proposition 4.10 for a proof . now so far @xmath29 is just a topological manifold , constructed by gluing together lattice polytopes . let @xmath560 there is a one - to - one inclusion reversing correspondence between strata of @xmath346 and elements of @xmath255 . it only remains to give @xmath29 an affine structure with singularities . let @xmath257 be the first barycentric subdivision of @xmath255 , and let @xmath276 be the union of simplices in @xmath257 not containing a vertex of @xmath255 or intersecting the interior of a maximal cell of @xmath255 . if we then set @xmath259 , we can define an affine structure on @xmath260 as follows . @xmath260 has an open cover @xmath261 where @xmath262 , the interior of @xmath263 , and @xmath264 is the ( open ) star of @xmath265 in @xmath257 , just as in 4 . we now define charts . as a maximal cell of @xmath255 is of the form @xmath553 , this inclusion induces a natural affine chart @xmath561 . on the other hand , a vertex @xmath265 of @xmath255 corresponds to a codimension @xmath347 stratum of @xmath346 , i.e. to an irreducible component @xmath431 for some @xmath163 . because this is a compact toric variety , it is defined by a complete fan @xmath562 in @xmath563 . furthermore , there is a one - to - one correspondence between @xmath57-dimensional cones of @xmath562 and @xmath57-dimensional cells of @xmath255 containing @xmath265 as a vertex , as they both correspond to strata of @xmath346 contained in @xmath431 . there is then a continuous map @xmath564 which takes @xmath565 , for any @xmath566 containing @xmath265 as a vertex , into the corresponding cone of @xmath562 _ integral affine linearly_. such a map is uniquely determined by the combinatorial correspondence and the requirement that it be integral affine linear on each cell . it is then obvious these charts define an integral affine structure on @xmath260 . thus we have constructed @xmath543 . let @xmath345 be a degeneration of elliptic curves to an @xmath422 fibre . then @xmath29 is the circle @xmath567 , decomposed by @xmath255 into @xmath7 line segments of length one . continuing with example [ toricdegenexample2 ] , the dual intersection complex constructed from the toric degenerations @xmath542 is the affine manifold with singularities structure with polyhedral decomposition @xmath543 constructed on @xmath568 in 4 . this is not particularly difficult to show . again , for the proof and more general complete intersection case , see @xcite . is the dual intersection complex the right affine manifold with singularities ? the following theorem provides evidence for this , and gives the connection between this construction and the syz conjecture . [ complextheorem ] let @xmath364 be a toric degeneration , with dual intersection complex @xmath543 . then there is an open set @xmath101 such that @xmath569 retracts onto the discriminant locus @xmath276 of @xmath29 , such that @xmath390 contains an open subset @xmath570 which is isomorphic as complex manifolds to a small deformation of a twist of @xmath451 , where @xmath571 . we will not be precise here about what we mean by small deformation ; by twist , we mean a twist of the complex structure of @xmath451 by a @xmath29-field . see @xcite for a much more precise statement ; the above statement is meant to give a feel for what is true . the proof , along with much more precise statements , will eventually appear in @xcite . if @xmath364 is a _ polarized _ toric degeneration , i.e. if there is a relatively ample line bundle @xmath531 on @xmath429 , then we can construct another affine manifold with singularities and polyhedral decomposition @xmath572 , which we call the _ intersection complex _ , as follows . for each irreducible component @xmath431 of @xmath346 , @xmath573 is an ample line bundle on a toric variety . let @xmath574 denote the newton polytope of this line bundle . there is then a one - to - one inclusion preserving correspondence between strata of @xmath346 contained in @xmath431 and faces of @xmath575 . we can then glue together the @xmath575 s in the obvious way : if @xmath360 is a codimension one stratum of @xmath346 , it is contained in two irreducible components @xmath431 and @xmath576 , and defines faces of @xmath575 and @xmath577 . these faces are affine isomorphic because they are both the newton polytope of @xmath578 , and we can then identify them in the canonical way . thus we obtain a topological space @xmath167 with a polyhedral decomposition @xmath579 . we give it an affine structure with singularities in a similar manner as before . again , let @xmath276 be the union of simplices in @xmath580 not containing a vertex of @xmath579 or intersecting the interior of a maximal cell of @xmath579 . setting @xmath581 , this again has an open cover @xmath582 as usual , as @xmath267 is the interior of @xmath263 , it comes along with a canonical affine structure . on the other hand , a vertex @xmath265 of @xmath579 corresponds to a dimension zero stratum @xmath482 of @xmath346 , and associated to @xmath482 is the polytope @xmath553 . let @xmath583 be the normal fan to @xmath555 in @xmath584 . then there is a one - to - one inclusion preserving correspondence between cones in @xmath583 and strata of @xmath346 containing @xmath482 . this correspondence allows us to define a chart @xmath585 which takes @xmath586 , for any @xmath587 containing @xmath265 as a vertex , into the corresponding cone of @xmath583 in an integral affine linear way . this gives the manifestly integral affine structure on @xmath588 , and hence defines the intersection complex @xmath572 . analogously to theorem [ complextheorem ] , we expect let @xmath364 be a polarized toric degeneration , with intersection complex @xmath572 . let @xmath409 be a khler form on @xmath390 representing the first chern class of the polarization . then there is an open set @xmath589 such that @xmath590 retracts onto the discriminant locus @xmath276 of @xmath167 , such that @xmath390 is a symplectic compactification of @xmath591 for any @xmath371 . i do nt expect this to be particularly difficult : it should be amenable to the techniques of ruan @xcite , but has not been carried out . the relationship between the intersection complex and the dual intersection complex can be made more precise by introducing multi - valued piecewise linear functions , in analogy with the multi - valued convex functions of definition [ multivaluedconvex ] : let @xmath29 be an affine manifold with singularities with polyhedral decomposition @xmath255 . then a multi - valued piecewise linear function @xmath592 on @xmath29 is a collection of continuous functions on an open cover @xmath593 such that @xmath594 is affine linear on each cell of @xmath255 intersecting @xmath440 , and on @xmath160 , @xmath595 is affine linear . furthermore , for any @xmath566 , in a neighbourhood of each point @xmath596 , there is an affine linear function @xmath597 such that @xmath598 is zero on @xmath263 . to explain this last condition , and to clarify additional structure on @xmath29 , let us examine a property of the polyhedral decomposition @xmath255 of @xmath29 when @xmath543 is a dual intersection complex . consider any @xmath57-dimensional cell @xmath566 . this corresponds to an @xmath599-dimensional stratum @xmath600 , and as such , it is a toric variety defined by a fan @xmath601 in @xmath602 . now for any vertex @xmath265 of @xmath263 , @xmath603 is a toric stratum of the irreducible component @xmath604 of @xmath346 . thus @xmath601 can be obtained as a _ quotient fan _ of @xmath605 . in other words , there is a @xmath57-dimensional cone @xmath606 of @xmath605 corresponding to @xmath263 such that @xmath607 in particular , there is an open neighbourhood @xmath608 of @xmath609 and an integral linear map @xmath610 such that @xmath611 let @xmath612 be a small open neighbourhood of @xmath613 in @xmath29 ; if taken sufficiently small the maps @xmath614 can be viewed as being defined on open subsets of @xmath612 and patch to give an integral affine submersion @xmath615 , where @xmath616 is an open neighbourhood of @xmath613 . this map has the property that @xmath617 in general we call a polyhedral decomposition _ toric _ if for all @xmath566 there is always such an integral affine linear map @xmath618 from a neighbourhood @xmath612 of @xmath613 and a fan @xmath601 in @xmath619 with the above property . ( see @xcite , definition 1.22 for a perhaps too precise definition of toric polyhedral decompositions . the definition there is complicated by allowing cells to be self - intersecting , or equivalently , allowing irreducible components of @xmath346 to be non - normal . ) we can think of the fan @xmath601 as being the fan structure of @xmath255 transverse to @xmath263 at a point in the interior of @xmath263 . the main point for a dual intersection complex is that this fan structure is determined by @xmath603 , and this is independent of the choice of the point in the interior of @xmath263 . now let us return to piecewise linear functions . suppose we are given a polarized toric degeneration @xmath364 . we in fact obtain a piecewise linear function @xmath592 on the dual intersection complex @xmath543 as follows . restricting to any toric stratum @xmath603 , @xmath620 is determined completely by an integral piecewise linear function @xmath621 on @xmath601 , well - defined up to a choice of linear function . pulling back this piecewise linear function via @xmath622 to @xmath612 , we obtain a collection of piecewise linear functions @xmath623 . the fact that @xmath624 for @xmath625 implies that on overlaps @xmath626 and @xmath627 differ by at most a linear function . so @xmath628 defines a multi - valued piecewise linear function . the last condition in the definition of multi - valued piecewise linear function then reflects the need for the function to be locally a pull - back of a function via @xmath622 in a neighbourhood of @xmath263 . in fact , given any multi - valued piecewise linear function @xmath592 on @xmath543 with @xmath255 a toric polyhedral decomposition of @xmath29 , @xmath592 is determined by functions @xmath621 on @xmath601 for @xmath566 , via pull - back by @xmath622 . if @xmath531 is ample , then the piecewise linear function determined by @xmath620 is strictly convex . so we say a multi - valued piecewise linear function is _ strictly convex _ if @xmath621 is strictly convex for each @xmath566 . now suppose we are given abstractly a triple @xmath629 with @xmath29 an integral affine manifold with singularities with a toric polyhedral decomposition @xmath255 , and @xmath592 a strictly convex multi - valued piecewise linear function on @xmath29 . then we construct the _ discrete legendre transform _ @xmath630 of @xmath629 as follows . @xmath167 will be constructed by gluing together newton polytopes . if we view , for @xmath265 a vertex of @xmath255 , the fan @xmath605 as living in @xmath563 , then the newton polytope of @xmath631 is @xmath632 there is a one - to - one order reversing correspondence between faces of @xmath633 and cells of @xmath255 containing @xmath265 . furthermore , if @xmath263 is the smallest cell of @xmath255 containing two vertices @xmath265 and @xmath271 , then the corresponding faces of @xmath633 and @xmath634 are integral affine isomorphic , as they are both isomorphic to the newton polytope of @xmath621 . thus we can glue @xmath633 and @xmath634 along this common face . after making all these identifications , we obtain a cell complex @xmath572 , which is really just the dual cell complex of @xmath543 . of course , we have some additional information , namely an affine structure on the interior of each maximal cell of @xmath579 . to give @xmath167 an integral affine structure with singularities , one proceeds as usual , using this affine structure along with an identification of a neighbourhood of each vertex of @xmath579 with the normal fan of the corresponding maximal cell of @xmath255 . finally , the function @xmath592 has a discrete legendre transform @xmath247 on @xmath572 . we have no choice but to define @xmath247 in a neighbourhood of a vertex @xmath635 dual to a maximal cell @xmath566 to be a piecewise linear function whose newton polytope is @xmath263 , i.e.@xmath636 this gives @xmath630 , the discrete legendre transform of @xmath629 . if @xmath29 is @xmath637 , then this coincides with the classical notion of a discrete legendre transform . the discrete legendre transform has several relevant properties : * the discrete legendre transform of @xmath630 is @xmath629 . * if we view the underlying topological spaces @xmath29 and @xmath167 as being identified by being the underlying space of dual cell complexes , then @xmath638 and @xmath639 , where the subscript denotes which affine structure is being used to define @xmath46 or @xmath229 . this hopefully makes it clear that the discrete legendre transform is a suitable replacement for the duality provided to us by the legendre transform of 2 . finally , it leads to what we may think of as an _ algebro - geometric syz procedure_. in analogy with the procedure suggested in 5 , we follow these steps : 1 . we begin with a toric degeneration of calabi - yau manifolds @xmath364 with an ample polarization . 2 . construct @xmath629 from this data , as explained above . 3 . perform the discrete legendre transform to obtain @xmath640 . 4 . try to construct a polarized degeneration of calabi - yau manifolds @xmath641 whose dual intersection complex is @xmath630 . the discrete legendre transform enables us to reproduce batyrev duality . returning to the construction of 4 and example [ toricdegenexample2 ] , choosing a strictly convex piecewise linear function on @xmath537 corresponding to a line bundle @xmath531 induces a polarization of @xmath541 . this then gives us a strictly convex multi - valued piecewise linear function @xmath642 on @xmath543 , hence a discrete legendre transform @xmath643 . in @xcite i showed that this is the dual intersection complex associated to some choice of subdivision @xmath644 of @xmath645 obtained by interchanging the roles of @xmath17 and @xmath235 in the construction of 4 . as an exercise , you can check the following for yourself . if we take @xmath299 , and in addition define @xmath646 to take the value @xmath81 on the primitive generator of each one - dimensional cone on @xmath279 , the normal fan to @xmath235 , then from 4 we obtain an affine structure with singularities on @xmath300 , and completely symetrically using @xmath647 we also obtain such a structure on @xmath648 . these manifolds come with polyhedral decomposition @xmath255 and @xmath579 consisting of all proper faces of @xmath17 and @xmath235 respectively . the anti - canonical polarizations on @xmath277 and @xmath246 induce multi - valued piecewise linear functions @xmath649 on @xmath29 and @xmath167 respectively . then show @xmath650 and @xmath651 are discrete legendre transforms of each other . thus batyrev ( and batyrev - borisov ) duality is a special case of this construction . the only step missing in this mirror symmetry algorithm is the last : [ reconstruct2 ] given @xmath629 , is it possible to construct a polarized toric degeneration @xmath364 whose dual intersection complex is @xmath629 ? it is fairly obvious how to reconstruct the central fibre @xmath346 of a degeneration from the data @xmath630 , and we will see this explicitly in 8 . one could naively hope that this reducible variety has good deformation theory and it can be smoothed . however , in general its deformation theory is ill - behaved . as initially observed in the normal crossings case by kawamata and namikawa in @xcite , one needs to put some additional structure on @xmath346 before it has good deformation theory . this structure is a _ log structure _ , and introducing log structures allows us to study many aspects of mirror symmetry directly on the degenerate fibre itself . we shall do this in the next section , but first , let me address the question of how general this mirror symmetry construction might be : if @xmath345 is a large complex structure limit degeneration , then @xmath33 is birationally equivalent to a toric degeneration @xmath652 . the condition of being a large complex structure limit , as defined by morrison in @xcite , is a stronger one than maximally unipotent . why should i imagine something like this to be true ? well , fantasizing freely , we would expect that after choosing a polarization and ricci - flat metric on fibres @xmath390 , we have a sequence @xmath653 converging to @xmath29 an affine manifold with singularities . now in general an affine manifold with singularities need not arise as the dual intersection complex of a toric degeneration , first of all because it need not have a toric polyhedral decomposition . for example , even in two dimensions there are orbifold singularities ( corresponding to singular elliptic fibres which are not semi - stable ) which do not arise in dual intersection complexes of toric degenerations , yet can occur as the base of a special lagrangian fibration on a k3 surface . however , the _ general _ base does arise as the dual intersection complex of a toric degeneration in the k3 case . the hope is that the condition of large complex structure limit forces the singularities of @xmath29 to be `` sufficiently general '' so that one can construct a nice toric polyhedral decomposition @xmath255 on @xmath29 , and from this construct a toric degeneration . presumably , this toric degeneration will be , if picked correctly , birational to the original one . this argument of course is rather hand - wavy , but i believe it provides some moral expectation that there might be a large class of degenerations for which our method applies . i note that one can prove the conjecture in the case of k3 surfaces . we now come to the technical heart of the program laid out in @xcite . some aspects of this program are quite technical , so the goal here is to explain the highlights of @xcite as simply possible . we first introduce the log structures of fontaine - illusie and kato ( @xcite , @xcite ) . a log structure on a scheme ( or analytic space ) @xmath5 is a ( unital ) homomorphism @xmath654 of sheaves of ( multiplicative and commutative ) monoids inducing an isomorphism @xmath655 . the triple @xmath656 is then called a _ log space_. we often write the whole package as @xmath657 . a morphism of log spaces @xmath658 consists of a morphism @xmath659 of underlying spaces together with a homomorphism @xmath660 commuting with the structure homomorphisms : @xmath661 the key examples : [ logexamples ] ( 1 ) let @xmath5 be a scheme and @xmath662 a closed subset of codimension one . denote by @xmath663 the inclusion . then the inclusion @xmath664 of the sheaf of regular functions with zeroes contained in @xmath360 is a log structure on @xmath5 . this is called a _ divisorial log structure _ on @xmath5 . \(2 ) a _ prelog structure _ , i.e. an arbitrary homomorphism of sheaves of monoids @xmath665 , defines an associated log structure @xmath666 by @xmath667 and @xmath668 . \(3 ) if @xmath354 is a morphism of schemes and @xmath669 is a log structure on @xmath360 , then the prelog structure @xmath670 defines an associated log structure on @xmath5 , the _ pull - back log structure_. \(4 ) in ( 1 ) we can pull back the log structure on @xmath5 to @xmath360 using ( 3 ) . thus in particular , if @xmath364 is a toric degeneration , the inclusion @xmath671 gives a log structure on @xmath429 and an induced log structure on @xmath346 . similarly the inclusion @xmath672 gives a log structure on @xmath2 and an induced one on @xmath347 . here @xmath673 , where @xmath674 is the ( additive ) monoid of natural ( non - negative ) numbers , and @xmath675 @xmath676 is usually called the standard log point . we then have log morphisms @xmath677 and @xmath678 . \(5 ) if @xmath679 is a convex rational polyhedral cone , @xmath680 the dual cone , let @xmath681 : this is a monoid . the affine toric variety defined by @xmath263 can be written as @xmath682 $ ] . we then have a pre - log structure induced by the homomorphism of monoids @xmath683\ ] ] given by @xmath684 . there is then an associated log structure on @xmath5 . this is in fact the same as the log structure induced by @xmath685 , where @xmath686 is the toric boundary of @xmath5 , i.e. the union of toric divisors of @xmath5 . if @xmath687 , then the monomial @xmath688 defines a map @xmath689\quad ( = { \operatorname{spec}}{\mathbb{c}}[t])$ ] which is a log morphism with the log structure on @xmath690 $ ] induced similarly by @xmath691 $ ] . the fibre @xmath692/(z^p)$ ] is a subscheme of @xmath5 , and there is an induced log structure on @xmath693 , and a map @xmath694 as in ( 4 ) . @xmath33 is an example of a _ log smooth _ morphism . essentially all log smooth morphisms are tale locally of this form ( if @xmath674 is replaced by a more general monoid ) . see @xcite for details . condition ( 4 ) of definition [ toricdegen ] in fact implies that locally , away from @xmath500 , @xmath695 and @xmath696 are of the above form . so we should view @xmath677 as log smooth away from @xmath500 , and from the log point of view , @xmath696 can be treated much like a non - singular scheme away from @xmath5 . we will see this explicitly below when we talk about differentials . on a log scheme @xmath657 there is always an exact sequence @xmath697 where we write the quotient sheaf of monoids @xmath698 additively . we call @xmath698 the _ ghost sheaf _ of the log structure . i like to view @xmath698 as specifying the combinatorial information associated to the log structure . for example , if @xmath657 is induced by the cartier divisor @xmath699 with @xmath5 normal , then the stalk @xmath700 at @xmath358 is the monoid of effective cartier divisors on a neighbourhood of @xmath482 supported on @xmath360 . [ ghostexercise ] show that in example [ logexamples ] , ( 5 ) , @xmath701 if @xmath702 and @xmath482 is the unique zero - dimensional torus orbit of @xmath5 . more generally , @xmath703 when @xmath358 is in the torus orbit corresponding to a face @xmath704 of @xmath263 . in particular , @xmath704 can be recovered as @xmath705 , where @xmath706 is the additive monoid of non - negative real numbers . another important fact for us is that if @xmath707 is a morphism with @xmath5 carrying a log structure , and @xmath360 is given the pull - back log structure , then @xmath708 . in the case that @xmath666 is induced by an inclusion of @xmath662 , @xmath698 is supported on @xmath360 , so we can equate @xmath698 and @xmath709 , the ghost sheaves for the divisorial log structure on @xmath5 and its restriction to @xmath360 . putting this together with exercise [ ghostexercise ] and the definition of dual intersection complex , we see that given a toric degeneration @xmath364 the dual intersection complex _ completely determines _ the ghost sheaf @xmath710 off of @xmath500 . we in fact take the view that anyway the log structure on @xmath500 is not particularly well - behaved , and we always ignore it on @xmath500 . in fact , given a log structure @xmath711 on @xmath712 , this defines a push - forward log structure @xmath713 . there is an induced map @xmath714 , as @xmath715 because @xmath346 is cohen - macaulay . thus in what follows , if we have determined a log structure on @xmath712 , we just as well get a log structure on @xmath346 and will not concern ourselves with the behaviour of this log structure along @xmath500 . all this gives the necessary hint for working backwards , to go from @xmath543 to @xmath696 . suppose we are given an integral affine manifold with singularities @xmath29 with _ toric _ polyhedral decomposition @xmath255 . at each vertex @xmath265 of @xmath255 , @xmath255 locally looks like a fan @xmath605 , defining a toric variety @xmath604 . for every edge @xmath716 with endpoints @xmath265 and @xmath717 , @xmath8 defines a ray in both fans @xmath605 and @xmath718 , hence toric divisors @xmath719 , @xmath720 . the condition that @xmath255 is a toric polyhedral decomposition tells us that @xmath721 and @xmath722 are isomorphic toric varieties , and we can choose a torus equivariant isomorphism @xmath723 for each edge @xmath8 . if we choose these gluing maps to satisfy a certain compatibility condition on codimension two strata ( we leave it to the reader to write down this simple compatibility condition ) , then we can glue together the @xmath604 s to obtain , in general , an algebraic space we write as @xmath724 , where @xmath725 is the collection of gluing maps . ( in @xcite , we describe the gluing data in a slightly different , but equivalent , way ) . we call @xmath726 _ closed gluing data_. this is how we construct a potential central fibre of a toric degeneration . now @xmath724 can not be a central fibre of a toric degeneration unless it carries a log structure of the correct sort . there are many reasons this may not happen . first , if @xmath726 is poorly chosen , there may be zero - dimensional strata of @xmath724 which do not have neighbourhoods locally tale isomorphic to the toric boundary of an affine toric variety ; this is a minimum prerequisite . as a result , we have to restrict attention to closed gluing data induced by what we call _ open gluing data_. explicitly , each maximal cell @xmath566 defines an affine toric variety @xmath727 given by the cone @xmath728 , assuming we view @xmath475 as a lattice polytope . let @xmath729 be the toric boundary . it turns out , as we show in @xcite , that a necessary condition for @xmath724 to be the central fibre of a toric degeneration is that it is obtained by dividing out @xmath730 by an tale equivalence relation . in other words , we are gluing together the @xmath731 s to obtain an algebraic space , and those tale equivalence relations which produce algebraic spaces of the form @xmath724 are easily determined . this is carried out in detail in @xcite , 2 . the construction there appears technically difficult because of the necessity of dealing with algebraic spaces , but is basically straightforward . the basic point is that if @xmath732 are two maximal cells , with @xmath733 , then @xmath704 determines faces of the cones @xmath734 and @xmath735 , hence open subsets @xmath736 , with toric boundaries @xmath737 . now in general there is no _ natural _ isomorphism between @xmath738 and @xmath739 : this is a problem when @xmath740 , where @xmath276 is as usual the singular locus of @xmath29 . however , crucially @xmath741 and @xmath742 are naturally isomorphic , and we can choose compatible equivariant isomorphisms to obtain _ open gluing data_. choosing open gluing data allows us to define the tale equivalence relation : we are just gluing any two sets @xmath743 via the chosen isomorphism between @xmath741 and @xmath742 . any choice of open gluing data @xmath726 gives rise in this way to an algebraic space @xmath724 , and to any choice of open gluing data there is associated closed gluing data @xmath744 such that @xmath745 . the advantage of using open gluing data is that each @xmath731 for @xmath746 carries a log structure induced by the divisorial log structure @xmath729 . unfortunately , these log structures are not identified under the open gluing maps , precisely because of a lack of a natural isomorphism between the @xmath747 s cited above . however , the ghost sheaves of the log structures are isomorphic . so the ghost sheaves @xmath748 glue to give a ghost sheaf of monoids @xmath749 . summarizing what we have said so far : ( this is a combination of results of @xcite , 2,4 ) given @xmath543 , if @xmath726 is closed gluing data , and @xmath750 is the central fibre of a toric degeneration @xmath364 with dual intersection complex @xmath543 , then @xmath726 is induced by open gluing data and @xmath751 . this is as far as we can get with the combinatorics . the next point is to attempt to construct @xmath752 . the idea is that @xmath752 is an extension of @xmath749 by @xmath753 , so we are looking for some subsheaf of the sheaf @xmath754 here the superscript @xmath755 denotes the grothendieck group of the monoid . any extension of @xmath756 by @xmath753 gives rise to a sheaf of groups @xmath757 surjecting onto @xmath756 , and the inverse image of @xmath758 is a sheaf of monoids @xmath752 . of course , one also needs a map @xmath759 , and this complicates things a bit . to make a long story short , we can identify a subsheaf of extensions which yield genuine log structures . a section of this subsheaf determines a log structure on @xmath724 with the correct ghost sheaf . however , this is not precisely what we want . what we really want is a log structure on @xmath724 along with a log morphism @xmath760 which is log smooth . ( we will address the question of the bad set @xmath761 shortly . ) we call such a structure a _ log smooth structure _ on @xmath724 . it turns out these structures are given by certain sections of @xmath762 , where @xmath763 is the canonical section of @xmath749 whose germ at @xmath764 is @xmath81 for @xmath765 a generic point of an irreducible component of @xmath724 . so in fact , we can identify a subsheaf of @xmath766 , which we call @xmath767 , whose sections determine a log structure on @xmath724 _ and _ a log smooth morphism @xmath768 , i.e. a log smooth structure . the technical heart of @xcite is an explicit calculation of the sheaf @xmath767 . this is carried out locally in @xcite , theorem 3.22 , where the sheaf is calculated on the ( tale ) open subsets @xmath731 of @xmath724 , and globally in @xcite , theorem 3.24 . i will not state the precise results , but go into detail in a special case , which illustrates the most important features of the theory . suppose @xmath724 is normal crossings , i.e. every cell of @xmath255 is affine isomorphic to a standard simplex . then we have the local @xmath769 sheaf , @xmath770 this is a line bundle on @xmath771 . then one can show @xmath767 is the @xmath772-torsor associated to @xmath769 . this brings us back to friedman s condition of @xmath773-semistability @xcite . a variety with normal crossings is _ @xmath773-semistable _ if @xmath774 . thus we recover kawamata and namikawa s result @xcite showing that @xmath724 carries a normal crossings log structure over @xmath676 if and only if @xmath724 is @xmath773-semistable . this is because , of course , the @xmath772-torsor associated to @xmath769 has a section if and only if @xmath774 . now theorem 3.24 of @xcite tells us that in general @xmath767 is not a trivial @xmath772-torsor . the sheaf depends continuously on @xmath726 , but discretely on monodromy of the singularities of @xmath29 . let s explain the latter point explicitly if @xmath775 . the irreducible components of @xmath776 are in one - to - one correspondence with one - dimensional cells of @xmath255 . if @xmath777 is such an edge , suppose it contains one singularity of @xmath29 such that @xmath46 has monodromy @xmath778 in a suitable basis around a loop around the singularity . then @xmath769 restricted to the one - dimensional stratum @xmath779 of @xmath724 is @xmath780 . to make this statement completely accurate , one needs to define @xmath7 so that it is independent of the choice of basis and loop . to do this , one chooses a loop which is counterclockwise with respect to the orientation determined by the chosen basis of @xmath781 , where @xmath782 is the base - point of the loop . if all the @xmath7 s appearing are positive , then for some choices of gluing data @xmath726 , we may hope to have a section @xmath371 of @xmath769 which vanishes only at a finite set of points @xmath500 . if @xmath500 does not contain a toric stratum ( i.e. a triple point ) then we obtain a log structure on @xmath783 of the desired sort , hence a log structure on @xmath724 ( log smooth off of @xmath500 ) by push - forward . we then have in the situation of this example , with @xmath775 and @xmath784 a section vanishing on a finite set @xmath500 not containing a triple point , there exists a smoothing @xmath364 of @xmath724 such that the singular locus of @xmath429 is @xmath785 , and the induced log morphism @xmath786 coincides with @xmath768 determined by @xmath371 . the proof of this is a rather simple application of friedman s or kawamata and namikawa s results . to apply these results , however , we need to deal with the singular set @xmath500 . this is done by normalizing @xmath724 , choosing to blow up one point in the inverse image of each point of @xmath500 , and then regluing along the proper transform of the conductor locus . this produces a @xmath773-semistable variety , in the language of friedman , or a log smooth scheme , which can then be smoothed . ( such an approach seems difficult in higher dimensions . ) on the other hand , if @xmath787 for some singular point of @xmath29 , we run into problems , and there is in fact no smoothing of @xmath724 . this should not be surprising for the following reason . if @xmath788 , it turns out we would have to compactify the torus fibration @xmath789 by adding a strange sort of @xmath313 fibre over such a singular point . an @xmath313 fibre is an immersed sphere , and the intersection multiplicity of the two sheets at the singular point of the fibre is @xmath790 for an ordinary @xmath313 fibre . however , when the monodromy is given by @xmath788 , the intersection multiplicity is @xmath791 . this does not occur for a special lagrangian @xmath344-fibration , so it is not surprising we ca nt construct a smoothing in this case . if @xmath775 and @xmath792 for all singularities on @xmath29 , then we say @xmath29 is _ positive_. one can generalize this notion of positive to higher dimensional @xmath29 with polyhedral decompositions , see @xcite , definition 1.54 . positivity of @xmath29 is a necessary condition for @xmath724 to appear as the central fibre of a toric degeneration . all the examples of 4 are positive ; this in fact follows from the convexity of reflexive polytopes , and the positivity condition can be viewed as a type of convexity statement . passing back to the general case now , with no restriction on the dimension of @xmath29 or the shape of the cells of @xmath255 , it follows from @xcite , theorem 3.24 , that @xmath767 is a subsheaf of _ sets _ of a coherent sheaf we call @xmath793 . this sheaf is a direct sum @xmath794 , where @xmath795 is a line bundle on the toric stratum of @xmath724 corresponding to @xmath8 . furthermore , as in the two - dimensional normal crossings case , @xmath795 is a semi - ample line bundle if @xmath29 is positive . a section @xmath796 which is a section of @xmath767 outside of the zero set @xmath500 of @xmath371 determines a log smooth structure on @xmath783 . in particular , if @xmath500 does not contain any toric stratum , we are in good shape . we then obtain a log morphism @xmath768 which is log smooth away from @xmath500 . we call such a structure a _ log calabi - yau space_. let s review : given data * @xmath726 open gluing data ; * @xmath796 , with @xmath371 a section of @xmath767 over @xmath797 for some set @xmath500 which does not contain any toric stratum of @xmath724 ; we obtain @xmath768 . conversely , we show in @xcite that if @xmath364 is a toric degeneration , then @xmath678 is obtained in this way from the dual intersection complex @xmath543 from some choice of data @xmath726 and @xmath371 . to complete this picture , it remains to answer [ smoothingconjecture ] suppose @xmath543 is positive . 1 . what are the possible choices of @xmath726 and @xmath371 which yield log calabi - yau spaces ? 2 . given @xmath768 , when does it arise as the central fibre of a toric degeneration @xmath798 ? as we have sketched it , this question is now the refined version of our basic reconstruction problem question [ reconstruct2 ] . the choice of the data @xmath726 and @xmath371 determine the moduli of log calabi - yau spaces arising from a given dual intersection complex . so far we havent even made the claim that this moduli space is non - empty , and for general choice of @xmath543 , i do not know if this is the case or not , though it is non - empty if @xmath775 or @xmath799 . however , one would like a more explicit description of this moduli space in any event . in general the moduli space is a scheme , but it may be singular ( an example is given in @xcite , example 4.28 ) . some additional hypotheses are necessary to solve this problem . to motivate the necessary hypothesis , let s go back to 1 , where we introduced the notion of simplicity . we saw that the basic topology of mirror symmetry works only when the fibration is simple . so maybe we should expect the current construction to work better when we have simplicity . there is one technical problem with this : the definition of simplicity assumes the existence of a torus fibration @xmath23 . instead , we want to define simplicity entirely in terms of @xmath29 itself . unfortunately , the solution to this is rather technical , and produces a definition which is very difficult to absorb ( definition 1.60 of @xcite ) . let us just say here that if @xmath29 is simple in this new sense and @xmath800 was compactified in a sensible manner to a topological torus fibration @xmath190 , then @xmath33 would be simple in the sense of 1 , provided that @xmath801 . in higher dimensions , this new simplicity does not necessarily imply the old simplicity ; see the forthcoming ph.d . thesis of helge ruddat . this arises in situations where orbifold singularities arise in @xmath176 ; as is well - known , such singularities can not be avoided in higher dimension . once we accept this definition , life simplifies a great deal . extraordinarily , the a priori very complicated moduli space of log calabi - yau spaces with a given dual intersection complex has a very simple description when @xmath29 is simple ! one very difficult main result of @xcite , ( theorem 5.4 ) is given @xmath543 positive and simple , the set of log calabi - yau spaces with dual intersection complex @xmath543 , modulo isomorphism preserving @xmath29 , is @xmath802 . an isomorphism is said to preserve @xmath29 if it induces the identity on the dual intersection complex . so the moduli space is an algebraic torus ( or a disjoint union of algebraic tori ) of dimension equal to @xmath803 . note that this is the expected dimension predicted by the syz conjecture . indeed , if a smoothing of @xmath804 exists and it was a topological compactification @xmath176 of @xmath805 , with a simple torus fibration @xmath190 extending @xmath307 , then @xmath806 , so by simplicity , @xmath807 . the discussion of 1 suggests that @xmath808 is @xmath809 of the smoothing , which is of course the dimension of the complex moduli space of the smoothing . this argument can be made rigorous by introducing _ log differentials_. let @xmath810 be a morphism of logarithmic spaces . a _ log derivation _ on @xmath657 over @xmath811 with values in an @xmath812-module @xmath813 is a pair @xmath814 , where @xmath815 is an ordinary derivation of @xmath816 and @xmath817 is a homomorphism of abelian sheaves with @xmath818 ; these fulfill the following compatibility condition @xmath819 for all @xmath820 . we denote by @xmath821 the sheaf of log derivations of @xmath657 over @xmath811 with values in @xmath812 . we set @xmath822 . this generalizes the more familiar notion of differentials with logarithmic poles along a normal crossings divisor . if @xmath662 is a normal crossings divisor inducing a log structure on @xmath5 , then @xmath823 is the sheaf of differentials with logarithmic poles along @xmath360 , and @xmath824 is the restriction of this sheaf to @xmath360 . in general , @xmath825 is locally free if @xmath826 is log smooth . as a result , one can do deformation theory in the log category for log smooth morphisms ( see @xcite ) . this is one of the principal reasons for introducing log geometry into our picture . if @xmath768 is a log calabi - yau space , then the morphism is log smooth off of @xmath500 . define @xmath827 where @xmath828 is the inclusion . then one has [ hodgedecomp ] suppose @xmath543 is positive and simple , and suppose we are given a log calabi - yau space @xmath768 which occurs as the central fibre of a toric degeneration @xmath798 whose general fibre @xmath390 is non - singular . then for @xmath829 and @xmath7 with @xmath830 , we have isomorphisms @xmath831 where @xmath832 and @xmath833 are the ordinary sheaves of holomorphic poly - vector fields and holomorphic differentials on a smooth fibre @xmath390 . the proof of this result , along with a number of other results , appears in @xcite . the result holds for all @xmath834 when additional hypotheses are assumed , essentially saying the mirror to @xmath390 is non - singular . note in particular , since @xmath46 and @xmath229 are interchanged under discrete legendre transform , we get the interchange of ordinary hodge numbers from this result . in the more general situation where the calabi - yaus arising are singular , one might speculate about the relationship between these groups , the actual hodge numbers and stringy hodge numbers . these issues are addressed in the forthcoming ph.d . thesis of helge ruddat . this section is purely philosophical . in most of our discussion in 7 and 8 , we focused on the dual intersection complex , and in particular , focused on the question of constructing a degeneration from its dual intersection complex . since our primary goal was to solve the reconstruction problem question [ reconstruct1 ] , and as the dual intersection complex is related to the complex structure ( theorem [ complextheorem ] ) it seems natural to focus on the dual intersection complex . we will see in the next section that this intuition may not always be correct . so far , the intersection complex only seemed to arise when talking about mirror symmetry . however , mirror symmetry instructs us to view both sides of the picture on the same footing . when we construct a degenerate calabi - yau space from a dual intersection complex , we say we are in the _ fan picture _ , while if we construct a degenerate calabi - yau space from an intersection complex , we say we are in the _ cone picture_. more precisely , we have seen how given an integral affine manifold with singularities with toric polyhedral decomposition @xmath543 , then an additional choice of open gluing data @xmath726 specifies a space @xmath724 , along with a sheaf of monoids @xmath835 . some additional data may specify a log structure on @xmath724 with this ghost sheaf . the irreducible components of @xmath724 are defined using fans , given by the fan structure of @xmath255 at each vertex of @xmath255 . this is why we call this side the fan picture . on the other hand , given @xmath543 we can also construct a _ projective _ scheme @xmath836 given suitable gluing data @xmath837 . the irreducible components of this scheme are in one - to - one correspondence with the maximal cells of @xmath255 ; given such a maximal cell @xmath263 , viewing it as a lattice polytope in @xmath637 determines a projective toric variety , and @xmath838 is obtained by gluing together these projective toric varieties using the data @xmath837 . this is not quite the same data as occurred in the fan picture , because we also need to glue the line bundles , and this is additional data . the reason for calling this side the cone picture is that each irreducible component can be described as follows . given @xmath475 , let @xmath839 . then the corresponding projective toric variety is @xmath840 $ ] , where @xmath841 $ ] is graded using the projection of @xmath842 onto @xmath77 . hence the irreducible components and strata arise from cones over elements of @xmath255 . we summarize the duality between the cone and fan pictures : ( some restrictions may apply to gluing data on both sides in order for @xmath592 to yield the desired data . ) in particular , mirror symmetry interchanges discrete information about the log structure ( i.e. @xmath835 ) and discrete information about the polarization ( i.e the class of the line bundle on each irreducible component ) . so far we have seen only the most elementary aspects of mirror symmetry emerge from this algebro - geometric version of syz , e.g. the interchange of hodge numbers . however , the real interest in this approach lies in hints that it will provide a natural explanation for rational curve counting in mirror symmetry . if we follow the philosophy of the previous section , we need to identify structures on affine manifolds with singularities which in one of the two pictures has to do with rational curves and in the other picture has to do with periods . i believe the correct structure to study is that of tropical curves on affine manifolds with singularities @xmath29 . see @xcite,@xcite for an introduction to tropical curves in @xmath637 . here , we can take @xmath29 to be tropical , rather than integral ; hence the name . let @xmath29 be a tropical affine manifold with singularities with discriminant locus @xmath235 . let @xmath58 be a weighted , connected finite graph , with its set of vertices and edges denoted by @xmath843}$ ] and @xmath844}$ ] respectively , with weight function @xmath845 } \rightarrow{\mathbb{n}}\setminus\{0\}$ ] . a parametrized tropical curve in @xmath29 is a continuous map @xmath846 satisfying the following conditions : 1 . for every edge @xmath847 , @xmath848 is an embedding , @xmath849 is dense in @xmath850 , and there is a section @xmath851 which is tangent to @xmath852 at every point of @xmath853 . we choose this section to be primitive , i.e. not an integral multiple of another section of @xmath854 . 2 . for every vertex @xmath855}$ ] , let @xmath856}$ ] be the edges adjacent to @xmath265 . let @xmath857 be the section of @xmath858 promised by ( 1 ) , chosen to point away from @xmath265 . this defines germs @xmath859 . 1 . if @xmath860 , the following _ balancing condition _ holds in @xmath861 : @xmath862 2 . if @xmath863 , then the following balancing condition is satisfied in @xmath864 : @xmath865 the latter group is interpreted as follows . let @xmath866 be a point near @xmath867 , and identify , via parallel transport along a path between @xmath867 and @xmath109 , the groups @xmath864 and @xmath868 with local monodromy invariant subgroups of @xmath869 and @xmath870 respectively . then @xmath871 is a subgroup of @xmath869 , and the intersection makes sense . it is independent of the choice of @xmath109 and path . so tropical curves behave away from the discriminant locus of @xmath29 much as the tropical curves of @xcite,@xcite do , but they may have legs terminating on the discriminant locus . as we are interested in the case that @xmath29 is compact , we do not want legs which go off to @xmath872 . i warn the reader , however , that this definition is provisional , and the behaviour in ( 2 ) ( b ) may not be exactly what we want . now let us connect this to the question of counting curves . in the situation of a degeneration , @xmath873 , it is natural to consider families of maps of curves : @xmath874^f\ar[d]_{\pi}&{{\mathcal{x}}}\ar[d]^{\varphi}\\ d\ar[r]_g&d}\ ] ] here @xmath96 may be a ramified covering , and @xmath826 is a flat morphism with reduced one - dimensional fibres . in the case of interest , @xmath875 should be a stable map of curves for @xmath493 . let us assume that @xmath876 is a non - singular curve for @xmath493 . in the logarithmic context , it is then natural to put the log structure induced by the divisor @xmath877 on @xmath878 , and so get a diagram @xmath879^f\ar[d]_{\pi}&{{\mathcal{x}}}^{\dagger}\ar[d]^{\varphi}\\ d^{\dagger}\ar[r]_g&d^{\dagger}}\ ] ] of log morphisms . restricting to the central fibre , we obtain a diagram @xmath880^f\ar[d]_{\pi}&{{\mathcal{x}}}^{\dagger}_0\ar[d]^{\varphi}\\ 0^{\dagger}\ar[r]_g&0^{\dagger}}\ ] ] this suggests that we should build up a theory of stable log maps and log gromov - witten invariants . this theory should generalize the theories developed by li and ruan @xcite and jun li @xcite . i will say little about this here , as this rapidly gets quite technical . there is work in progress of siebert on this subject . this point of view has already been used in @xcite for counting curves in toric varieties , so some more hints of this approach can be found there . instead , i wish to sketch how such a diagram yields a tropical curve . to do so , consider a situation where @xmath826 is normal crossings , and the induced map @xmath881 has no infinitesimal log automorphisms over @xmath676 . ( this is the log equivalent of the notion of stable map ) . let @xmath543 be the dual intersection complex of the log calabi - yau space @xmath696 . we can define the _ dual intersection graph _ of @xmath882 , which will be a parameterized tropical curve on @xmath29 . i will only do the case here when the image of @xmath33 is disjoint from the set @xmath499 of definition [ toricdegen ] , ( 4 ) ; otherwise there are some technicalities to worry about . first we build @xmath58 . let @xmath883 be the irreducible components of @xmath884 . assume these components are normal for ease of describing this construction . set @xmath843}=\{v_1,\ldots , v_m\}$ ] . on the other hand , @xmath844}$ ] will contain an edge @xmath885 joining @xmath886 and @xmath887 whenever @xmath888 . to define @xmath846 , we first describe the image of each vertex . let @xmath889 be the minimal stratum of @xmath346 containing @xmath890 , where @xmath891 . let @xmath892 be the generic point of @xmath893 , @xmath894 . then we have an induced map @xmath895 , as @xmath33 is a log morphism . this induces a diagram on stalks of ghost sheaves @xmath896_{\bar f^{\ # } } \\ { \overline{{\mathcal{m}}}}_{0 } \ar[u]^{\bar\pi^{\ # } } & { \overline{{\mathcal{m}}}}_{0}\ar[l]^{\bar g^{\ # } } \ar[u]_{\bar\varphi^{\#}}}\ ] ] now @xmath897 ( see example [ logexamples ] , ( 4 ) ) and @xmath898 since @xmath826 is normal crossings . on the other hand , @xmath899 is the identity and if @xmath96 is a branched cover of degree @xmath773 , then @xmath900 is multiplication by @xmath773 . by exercise [ ghostexercise ] , @xmath901 but @xmath902 so @xmath903 is determined by an element @xmath904 of @xmath905 . now @xmath906 while @xmath907 thus @xmath908 , and @xmath909 . we define @xmath910 . this is a point of @xmath575 which is contained in @xmath29 . if @xmath888 , there is a minimal stratum @xmath911 containing @xmath912 . of course @xmath913 . in particular , @xmath914 contains @xmath575 and @xmath577 . we take @xmath915 to be the straight line joining @xmath916 and @xmath917 inside @xmath914 . furthermore , if @xmath918 is embedded as a lattice polytope , let @xmath919 be a primitive lattice element parallel to @xmath920 , and we take @xmath921 to be defined by the equation @xmath922 we do not give a proof here . the case where @xmath889 is always an irreducible component of @xmath346 is essentially covered in @xcite . instead , we ll do another extremal case , which exhibits some interesting features of log geometry . suppose a component @xmath923 of @xmath884 and all components @xmath924 intersecting @xmath923 are mapped by @xmath33 to a zero dimensional stratum @xmath603 of @xmath346 . without loss of generality we can assume @xmath925/(z^{(0,1)})\ ] ] as defined in 8 . thus @xmath647 maps @xmath926 into points @xmath927 . let us understand why the balancing condition holds at @xmath928 . let @xmath929 be an open neighbourhood of @xmath923 which only intersects @xmath930 , so @xmath647 is constant on @xmath103 as an ordinary morphism ( but not as a log morphism ) . restrict the log structure on @xmath884 to @xmath103 . we have an exact sequence @xmath931 taking global sections , we get @xmath932 a section @xmath933 defines an @xmath934-torsor @xmath935 , whose class in the picard group of @xmath103 is @xmath936 . it is an easy exercise in log geometry to show that if @xmath726 is the @xmath163th standard basis vector for @xmath937 , then @xmath938 . note @xmath939 for @xmath940 and @xmath941 as @xmath942 in @xmath878 . now observe that @xmath943 acting on the sheaves of monoids induces a diagram @xmath944^{\cong } & { \mathcal{m}}^{{{\operatorname{gp}}}}_{v(\sigma),x}\ar[r]^{f^{\#}}\ar[d]&\gamma(u,{\mathcal{m}}_u^{{{\operatorname{gp}}}})\ar[d]^p&\\ n\oplus{\mathbb{z}}\ar[r]_{\cong}&{\overline{{\mathcal{m}}}}^{{{\operatorname{gp}}}}_{v(\sigma),x}\ar[r]_{\bar f^{\#}}&\gamma(u , { \overline{{\mathcal{m}}}}_u^{{{\operatorname{gp}}}})\ar[r]_{\cong}&{\mathbb{z}}^t.}\ ] ] the map @xmath903 , by construction , is given by @xmath945 on the other hand , in order for @xmath903 to lift to @xmath943 , the @xmath934 torseur @xmath946 must have a section for every @xmath947 , i.e. must be trivial in the picard group . this implies @xmath948 or @xmath949 for all @xmath947 . but this is equivalent to @xmath950 which is the balancing condition . following the logic of mirror symmetry , this suggests that tropical curves on the cone side should have to do with periods . it is only recently that an understanding of this has begun to emerge , and unfortunately , i do not have space or time to elaborate on this . let me say that in @xcite , siebert and i have given a solution to question [ smoothingconjecture ] , ( 2 ) , given some hypotheses on @xmath804 , which are implied by simplicity of @xmath29 . in this solution , we construct explicit deformations of a log calabi - yau space , order by order . formally , our construction looks somewhat similar to that of kontsevich and soibelman @xcite for constructing non - archimedean k3 surfaces from affine manifolds , and we apply a key lemma of @xcite . however , kontsevich and soibelman work on what we would call the fan side , while we work on the cone side . this may be surprising given that all of our discussions involving the strategy of building log calabi - yau spaces was done on the fan side . however , if we take the mirror philosophy seriously , and we want to see tropical curves appear in a description of a smoothing , we need to work on the cone side . it turns out to be extremely natural . in fact , all tropical rational curves play a role in our construction . ultimately , all periods can be calculated in terms of the data involved in our construction , and in particular , there is a clear relationship between the period calculation and the existence of tropical rational curves on @xmath29 . once this is fully understood , this will finally give a firm understanding of a geometric explanation of mirror symmetry . cheng and s .- t . yau , _ the real monge - ampre equation and affine flat structures _ , in _ proceedings of the 1980 beijing symposium on differential geometry and differential equations _ , vol . 1 , 2 , 3 ( beijing , 1980 ) , 339370 , science press , beijing , 1982 . k. fukaya , _ multivalued morse theory , asymptotic analysis and mirror symmetry , _ in _ graphs and patterns in mathematics and theoretical physics _ , 205278 , proc . sympos . pure math . , * 73 * , amer . soc . , providence , ri , 2005 . j. richter - gebert , b. sturmfels , and t. theobald , _ first steps in tropical geometry , _ in _ idempotent mathematics and mathematical physics _ , 289317 , contemp . , * 377 * , amer . math . soc . , providence , ri , 2005 .

we trace progress and thinking about the strominger - yau - zaslow conjecture since its introduction in 1996 . in particular , we aim to explain how the conjecture led to the algebro - geometric program developed by myself and siebert , whose objective is to explain mirror symmetry by studying degenerations of calabi - yau manifolds . we end by outlining how tropical curves arise in the mirror symmetry story . # 1 # 1 # 1#2#30#1#2#30 # 1 # 1 # 1#1^

i would like to thank dieter zeppenfeld and david rainwater for their fruitful collaboration on this project .

we motivate the study of `` minijets '' ( that is jets soft with respect to the hard scattering , but hard with respect to @xmath0 ) as a means to study the underlying qcd colour flow in events . we discuss , with the aid of a simplistic model , currently available events in which minijets are observed , that is the events have high multiplicity , as a means of understanding the physics of minijets . we encourage both theoreticians and experiments to continue the study of minijets . = -2 cm = -1.5 cm = 23.0truecm = 16.5truecm cern th/97182 1.cm * minijets as a perturbative probe of colour charge * 1.3 cm david j. summers .2 cm th division , cern , ch-1211 genve 23 , suisse contribution to 32nd rencontres de moriond : qcd and hadronic interactions cern th/97182 in the past the concept of qcd colour flow has been an interesting prediction , and hence test , of qcd . one of the more surprising examples of qcd colour flow has been the rapidity gaps seen at hera @xcite . in the majority of jet events observed at hera the incoming electron / positron radiates a photon , which scatters off one of the coloured constituents of the proton , to produce the observed jet . this hard scattering leaves the proton remnant and the pre - jet in a non colour singlet state , during hadronization the colour of these objects is rearranged to form colourless hadrons , and these hadrons fill the gap between the produced jet and the proton remnant with an underlying event . however among the hera events are some with an observed jet produced at central rapidities ; but with no hadrons between the jet and proton remnant direction ; _ i.e. _ there is no underlying event , and this means that the mechanism for producing this jet can not be that just described . the lack of hadrons between the observed jet and the proton remnant means that there is no qcd colour is exchanged between the proton and the photon , and hence that photon must have scattered off a colour singlet object in the proton , the pomeron . so the flow of hadrons in an event has told us something about the structure of the proton , that the struck object inside the proton that took part in the hard scattering is a qcd colour singlet . this ability to probe the qcd colour of the object that takes part in the hard scattering is , in the future , likely to provide a strong experimental probe . a prime example is in the detection of a heavy ( @xmath1 gev ) higgs boson at the lhc , such a higgs is produced via two different mechanisms . approximately two thirds of heavy higgs are produced from @xmath2 fusion via a top quark loop , with the remaining third produced via vector boson fusion , show in fig.[fig : wwh]a . such a heavy higgs typically decays into two massive vector bosons , either @xmath3 or @xmath4 bosons , and the back ground from continuum diboson production , shown in fig.[fig : wwh]b , can be large . however the qcd colour flow in the diboson fusion signal process is very different from the background continuum diboson production , the leading colour flow is shown in fig.[fig : wwh ] as a dotted line . for the diboson fusion process qcd colour flows between the proton remnant and the parton that emits the vector boson , this parton travels in very much the same direction as the proton remnant . this means that the hadrons associated with the underlying event will travel in a similar direction to the proton remnant , i.e. at very large rapidities . for the diboson continuum background ( as well as the higgs signal from @xmath2 fusion ) qcd colour is exchanged between the two protons , and this means that the underlying event will produce hadrons at all rapidities . this difference in the flow of hadrons gives us a method of separating the vector boson fusion higgs signal from the continuum background , effectively the higgs signal has a double rapidity gap on both sides of the event , with the higgs decay taking place at central rapidities @xcite . however this method for separating signal from background by the flow of hadrons has many difficulties . on the theoretical side we make predictions and the rate of the signal and background based upon perturbative qcd ( pqcd ) . pqcd s fundamental fields are partons ( quarks and gluons ) and it is about these fields that pqcd makes predictions . by local parton hadron duality we expect that the produced partons will hadronize into jets ; however this tells us nothing about individual hadrons and it is in terms of individual hadrons that rapidity gaps are defined . this means that we can make no absolute predictions based upon pqcd of the signal rate of events where a rapidity gap is observed , or indeed the background rate where a fluctuation leads to an observed rapidity gap , despite qcd colour being exchanged between the protons . on the experimental side the cross section for higgs productions drops very rapidly as the higgs gets heavy , and for very heavy higgs the cross section is small enough that it is only with very high luminosities that we produce any higgs events at all . however at very high luminosities there are likely to be many interactions per bunch crossing , and each of these interactions will produce its own underlying event , which will wash out the rapidity gap at central rapidities . one possible solution to these problems is rather that look for events with a rapidity gap , _ i.e. _ no hadrons at central rapidities , instead to look for events with no soft `` minijets '' at central rapidities @xcite . naively we expect additional jets to be suppressed by a factor of @xmath5 , however there can often be a large volume of phase space in which additional jets can be radiated , and this can enhance the additional jet rate by a large logarithm such that multijet emission is not suppressed . in particular jet emission that is soft with respect to the hard scattering is enhanced by such a large logarithm . if we define soft jets in such a way that the background divector boson continuum ( which has more qcd colour acceleration , and hence more jet activity ) typically has additional jet emission , while the vector boson fusion signal ( with less qcd colour accelerated ) typically has no additional jet activity ; then the difficulties with using hadrons to define a rapidity gap are overcome . additional soft jet production we expect to be predictable by pqcd as long as the scale for soft jet emission is far above @xmath0 . on the experimental side we expect that the majority of interactions will be very soft , and although producing an underlying event the majority of interactions will not produce observed jets , and so not interfere with tagging a higgs signal event . this means that firm theoretical predictions can , in principle , be made about a minijet tag of a higgs signal , and that a minijet tag can be used at high luminosities where there are overlapping events . however before we can use minijet emission as a probe of qcd colour charge it is necessary to understand minijet emission both on a theoretical and experimental level . in order for multiple jet emission to be important we need a large phase space for emission of those jets , and this means that we require the hard scattering to be as energetic as possible . currently the most energetic events are to be found at the tevatron , and cdf have studied the multijet distributions in there most energetic events @xcite . they apply the cuts , @xmath6 where @xmath7 is the scattering angle of the highest @xmath8 jet in the multijet centre of mass frame . due to the high invariant mass of these events the events are forced to be fairly central , and this means that at least two jets are always observed . in figure [ fig : mult ] we show the multiplicity distribution observed in the cdf events , although 3rd and additional jets are naively produced at a higher order in perturbation theory because of the large volume of phase space such additional jet activity is not suppressed , with typical events with 3 or 4 jets observed . the high multiplicity of these events means that pqcd in its lowest order is not not reliable . the large logs that give rise to the high multiplicity need to be dealt with in the theoretical calculation . now most of the additional jet activity takes place at low @xmath8 , and these soft jets typically originate from gluons . now for soft gluons the matrix element factorises , the probability to radiate @xmath9 additional soft gluons is given by , @xmath10 the phase space to radiate @xmath9 additional gluons is given by , @xmath11 where the @xmath12 term arises from the symmetry factor for integrating @xmath9 gluons over the same phase space . now if @xmath13 , that is we make no experimental requirements about the observed properties of the gluons , then in moment space we find , @xmath14 this means that the probability to see @xmath9 gluon jets is given by , @xmath15 and so , @xmath16 however for the case of jets defined by the cdf algorithm @xmath17 , also not all jets are soft , and not all jets are gluon initiated . this means that this simple factorisation of the additional jet rate does not hold , however it is a reasonable ansatz to assume equation [ eq : atz ] . this is not the full story though , equation [ eq : atz ] does not conserve probability , this happens because the rate to radiate an addition gluon @xmath18 is an inclusive quantity , whereas if we are to make predictions about the specific multiplicity distributions then we require the exclusive rate to produce _ exactly _ @xmath9 additional jets . we can obtain the exclusive jet rate by adding a sudakov suppression factor @xmath19 @xcite that excludes all additional jet activity , and then equation [ eq : atz ] produces exactly @xmath9 exclusive jets . this gives the ansatz for additional jet activity over the hard scattering as , @xmath20 where @xmath21 is the average multiplicity @xmath22 , and at the leading logarithmic level is given by , @xmath23 this formula has the advantage over strictly tree level pqcd of being still valid even when @xmath24 . also we have , @xmath25 and so the 3 jet cross section has the same accuracy as at leading order . if we compare equation [ eq : satz ] with the measured cdf multiplicity distribution , fitting only @xmath26 to the observed average excess multiplicity over the 2 hard jets always observed we find the curve shown in figure [ fig : mult ] . although the point for multiplicity 4 is 3.5 standard deviations above the fitted curve , and thus the equation [ eq : satz ] has limitations , it is clear that the fit has got the gist of the multiplicity distribution correct and gives a place to start when studying event multiplicities , and hence the physics of minijets . if we now try to predict the value for @xmath26 from pqcd using equation [ eq : pfit ] with a tree level calculation for @xmath27 and a next - to - leading - order calculation for @xmath28 @xcite then we find the results shown in figure [ fig : nbar ] plotted against the minimum @xmath8 used to define a jet @xcite . for the leading order calculation of @xmath27 we have several different choices of scale . for the renormalisation scale of @xmath5 ; this can either be some multiple of the hard scattering scale , @xmath29 , or some multiple of the soft scale at which the jet is emitted @xmath30 . for the factorisation scale at which the parton densities are evaluated we similarly can choose a hard scale related to @xmath29 , if we feel we should sum up the initial state radiation , or a soft scale related to @xmath30 if we feel that explicitly calculating high order emission in @xmath27 means that we should only generate higher order corrections bellow that scale . a case can be made for each of these choices , and from figure [ fig : nbar ] we can see the prediction for @xmath22 changes by a factor of 3 for the experimental value of @xmath31 . this happens because we have two natural scales in the calculation of @xmath27 the hard scattering scale , and the @xmath8 of the emisson , and as these have very different values the predictions for @xmath22 vary accordingly . the theoretical predictions for @xmath22 span the experimental measurement , but we can not make an accurate theoretical predition at this time . without improvements to the theoretical predictions we do not currently have the ability to accurately predict the rate of minijet emission , and this means that we do not currently have the technology to use minijets as a absolute theoretical prediction to probe qcd colour charge . on the theoretical side the large scale dependence of @xmath27 can be cured by working at higher order in perturbation theory @xcite ; however this would necessitate an improvement in the simple minijet model , equation [ eq : satz ] , which is non consistent we such a higher order calculation . alternatively it may be possible to show which logarithmic terms need to be resummed in order to remove the large scale dependence , without a full calculation of the higher order finite corrections . in addition equation [ eq : satz ] does not take advantage of our knowledge of the tree level rate for high jet multiplicities @xcite . clearly there is much theoretical work that remains to be done . there is also much work to be done on the experimental side to help our understanding of minijets , for example the different theoretical scale choices not only affect the average multiplicity , but also affect the shape of distributions . this can be seen in figure [ fig : nbar ] where a renormalisation scale related to the @xmath8 of the emitted jet has the average multiplicity growing far more rapidly as the @xmath8 cut used to define a jet is decreased , than a harder renormalisation scale . if the experimenter were able to measure this distribution then this information can be used to motivate different choices for the scale choice .

high - power lasers and plasmas may be used to accelerate electrons by electric fields that are orders of magnitude greater than those achievable using conventional methods @xcite . an intense laser pulse is used to drive a wave in a plasma and , for sufficiently large fields , non - linearities lead to collapse of the wave structure ( ` wave - breaking ' ) due to sufficiently large numbers of electrons becoming trapped in the wave . hydrodynamic investigations of wave - breaking were first undertaken for cold plasmas @xcite and thermal effects were later included in non - relativistic @xcite and relativistic contexts @xcite ( see @xcite for a discussion of the numerous approaches ) . however , it is clear that the value of the electric field at which the wave breaks ( the electric field s ` wave - breaking limit ' ) is highly sensitive to the details of the hydrodynamic model . plasmas dominated by collisions are described by a pressure tensor that does not deviate far from isotropy , whereas an intense and ultrashort laser pulse propagating through an underdense plasma will drive the plasma anisotropically over typical acceleration timescales . thus , it is important to accommodate 3-dimensionality and allow for anisotropy when investigating wave - breaking limits . our aim is to uncover the relationship between wave - breaking limits and the shape of the @xmath1-particle distribution @xmath2 . in general , the detailed structure of @xmath2 can not be reconstructed from a few low - order moments so we adopt a different approach based on a particular class of piecewise constant @xmath1-particle distributions . our approach may be considered as a multi - dimensional generalization of the 1-dimensional relativistic ` waterbag ' model employed in @xcite ( for a discussion of the relationship between our approach and @xcite see @xcite ) . we employ the einstein summation convention throughout and units are used in which the speed of light @xmath3 and the permittivity of the vacuum @xmath4 . lowercase latin indices @xmath5 run over @xmath6 . our attention is focussed on plasmas evolving over timescales during which the ` discrete ' nature ( collisions ) of the plasma electrons can be neglected and the plasma ions can be prescribed as a background . such configurations are well described by the covariant vlasov - maxwell system @xcite which , for the purposes of this paper , is most usefully expressed in the language of exterior calculus ( see , for example , @xcite ) . we will now briefly summarize the particular formulation of the vlasov - maxwell system employed here . let @xmath7 be a spacetime with signature @xmath8 for the metric tensor @xmath9 . each point @xmath10 is associated with a space @xmath11 of future - directed unit normalized vectors on @xmath12 , @xmath13 where @xmath14 are the components of the metric @xmath9 in a coordinate system @xmath15 whose patch contains @xmath16 and @xmath17 are induced coordinates on @xmath18 . the total space @xmath19 of the bundle @xmath20 is the union of @xmath21 over @xmath10 and @xmath22 is the restriction to @xmath19 of the canonical projection on @xmath18 . naturally induced tensors on @xmath18 include the dilation vector field @xmath23 @xmath24 the vertical lift @xmath25 of the volume @xmath26-form @xmath27 from @xmath12 to @xmath18 and the horizontal @xmath26-form @xmath28 @xmath29 where @xmath30 is the horizontal lift of @xmath31 from @xmath12 to @xmath18 ( see appendix [ appendix : vertical_and_horizontal ] for further details ) and @xmath32 is the matrix of components of @xmath9 . the vlasov - maxwell system for @xmath2 ( a scalar field on @xmath18 whose restriction to @xmath19 is the plasma electron @xmath1-particle distribution ) and the electromagnetic field @xmath33 may be written @xmath34 where @xmath35 indicates equality on restriction by pullback from @xmath18 to @xmath19 and @xmath36 with @xmath37 the mass and @xmath38 the charge of the electron ( @xmath39 ) and @xmath40 the components of the electromagnetic 2-form @xmath41 on @xmath12 . the metric dual of a vector @xmath42 is defined by @xmath43 for all vectors @xmath44 on @xmath12 and @xmath45 is the hodge map induced from the volume @xmath26-form @xmath27 @xmath46 on @xmath12 . the components of the electron number @xmath26-current @xmath47 at @xmath10 are given as an integral over the fibre @xmath48 @xmath49 and the ion number @xmath26-current @xmath50 is prescribed as data . the measure on @xmath21 in ( [ electron_current ] ) is induced from the @xmath0-form @xmath51 , @xmath52 where @xmath53 is the alternating symbol with @xmath54 . the vlasov - maxwell equations constitute a non - linear integro - differential system . direct calculation of its solutions for general plasma configurations is difficult and , to proceed analytically , it is common to approximate the above as a finite number of moments of @xmath2 in @xmath55 satisfying a non - linear field system on @xmath12 ( a so - called ` fluid ' model ) . however , there are difficult issues associated with closing the resulting field system ( see , for example , @xcite ) so we opt for a different approach . our strategy is to reduce the system by employing a discontinuous @xmath2 , and to proceed we need to cast ( [ lvlas ] ) as an integral . one may rewrite ( [ lvlas ] ) as @xmath56 integrating ( [ dvlas ] ) over a 7-chain @xmath57 and applying stokes s theorem yields @xmath58 with @xmath59 the boundary of @xmath60 . for differentiable distributions , this equation is equivalent to ( [ dvlas ] ) ; however , since it makes no reference to the differentiability of @xmath2 , it may be regarded as a generalisation of ( [ dvlas ] ) applicable to discontinuous @xmath2 . equation ( [ intvlas ] ) may be used to develop an equation of motion for a discontinuity , which we choose as a local hypersurface @xmath61 . suppose that @xmath60 in ( [ intvlas ] ) is a @xmath62-dimensional ` pill - box ' straddling @xmath61 . we may write @xmath63 where @xmath64 and @xmath65 are the ` top ' and ` bottom ' of the pill - box and @xmath66 is the ` sides ' of the pill - box . thus , in the limit as the volume of @xmath60 tends to zero with @xmath64 tending to @xmath67 and @xmath65 tending to @xmath68 , we recover the condition @xmath69 \sigma^\ast \omega = 0,\ ] ] where the image of @xmath67 is in @xmath61 and @xmath70 = \sigma^*_+f + \sigma^*_-f$ ] . thus it follows that a finite discontinuity in @xmath2 can occur only across the image of a chain @xmath71 satisfying @xmath72 suppose that @xmath71 may be written locally @xmath73 for @xmath74 , where @xmath75 denotes the @xmath76 component of @xmath71 , and @xmath77 . it is then possible to translate ( [ sigma_discon ] ) into a field equation for a family of vector fields @xmath78 on @xmath79 given as @xmath80 where , since @xmath81 at @xmath82 , it follows @xmath83 using ( [ omega ] , [ sigma_discon ] ) it follows @xmath84 consider first the term @xmath85 in equation ( [ split ] ) : @xmath86 where ( [ liou ] ) and @xmath87 have been used ( see ( [ v_on_v ] ) in appendix [ appendix : vertical_and_horizontal ] ) . thus @xmath88 since @xmath89 ( see ( [ v_on_h_and_h_on_v ] ) in appendix [ appendix : vertical_and_horizontal ] ) . furthermore , using ( [ h_lift_of_dx ] ) in appendix [ appendix : vertical_and_horizontal ] , it follows @xmath90 where @xmath91 is the exterior covariant derivative on @xmath12 and @xmath92 is the exterior derivative on @xmath93 . using ( [ i_x_hash_1_on_tm ] , [ pullback_dx_h ] ) it follows @xmath94 and ( [ term_a ] ) is @xmath95 since @xmath96 ( @xmath97 ) and @xmath98 ( dim(@xmath93)=2 ) . since @xmath99 and @xmath100 , it follows ( [ simplified_term_a ] ) may be written @xmath101 where the family of @xmath102-forms @xmath103 on @xmath79 is @xmath104 the second term @xmath105 in ( [ split ] ) can be rewritten using a similar procedure : @xmath106 and from ( [ liou ] , [ force ] , [ i_x_hash_1_on_tm ] ) it follows @xmath107 where ( [ h_on_h ] ) and ( [ v_on_h_and_h_on_v ] ) have been used . then @xmath108 and @xmath109 combining ( [ term_a_on_m ] , [ term_b_on_m ] ) and ( [ split ] ) yields @xmath110 acting on ( [ discont_on_m ] ) successively with @xmath111 , @xmath112 and @xmath113 , and noting that @xmath114 and @xmath115 , yields @xmath116 thus , solutions to ( [ discon ] ) may be obtained by demanding that @xmath78 is driven by the lorentz force @xmath117 however , although ( [ lorentz_force_law ] ) is simpler than ( [ discon ] ) , there are simple solutions to ( [ discon ] ) that do not satisfy ( [ lorentz_force_law ] ) ; we will return to this point shortly . a laser pulse travelling through a plasma can excite plasma oscillations , which induce very high longitudinal electric fields . due to nonlinear effects , there is a maximum amplitude of electric field ( the ` wave - breaking limit ' ) that can be sustained in the plasma and , as mentioned in the introduction , our aim is to investigate the relationship between the shape of the distribution ( i.e. @xmath71 ) and the wave - breaking limit . for simplicity , we choose to describe the plasma using a distribution @xmath2 where @xmath118 is a positive constant inside a @xmath62-dimensional region @xmath119 and @xmath120 outside . in particular , we consider @xmath121 to be the union over each point @xmath10 of a domain @xmath122 whose boundary @xmath123 in @xmath124 is topologically equivalent to the @xmath102-sphere . such distributions are sometimes called ` waterbags ' in the literature and can be completely characterized by @xmath78 and the constant @xmath125 . our approach may be considered as a multi - dimensional generalization of the purely 1-dimensional relativistic waterbag model used in @xcite to examine wave - breaking . we work in minkowski spacetime @xmath7 and assume that the ions constitute a homogeneous immobile background . we employ an inertial coordinate system @xmath15 adapted to the ions : @xmath126 where @xmath127 and the ion proper number density @xmath128 is constant . to proceed further we seek a form for @xmath71 axisymmetric about @xmath129 whose pointwise dependence in @xmath12 is on the wave s phase @xmath130 only , where @xmath131 is constant and @xmath132 . we suppose that all electrons described by @xmath2 are travelling slower than the wave , and the wave ` breaks ' if the longitudinal velocity of any plasma electron equals @xmath131 ( i.e. an electron ` catches up ' with the wave ) . introduce @xmath133 and decompose @xmath134 as @xmath135\ , \bm{e}^1 + \psi(\xi^1,\zeta)\ , \bm{e}^2 \,\,+ r\sin(\xi^1)\cos(\xi^2)dx^1 + r\sin(\xi^1)\sin(\xi^2)dx^2 \label{v_ansatz}\ ] ] for @xmath136 , @xmath137 where @xmath138 is constant . here , @xmath139 , with @xmath140 , is an orthonormal coframe on @xmath12 adapted to @xmath141 . since @xmath78 is future - directed and timelike , and @xmath142 is timelike , it follows @xmath143 and @xmath144 . the component @xmath145 is determined using ( [ norm ] ) , @xmath146 ^ 2 - \gamma^2[1 + r^2 \sin^2(\xi^1)]},\ ] ] where the negative square root is chosen because no electron is moving faster along @xmath147 than the wave . substituting the _ ansatz _ ( [ v_ansatz ] ) together with a purely longitudinal electric field depending only on @xmath141 , @xmath148 into ( [ lorentz_force_law ] ) yields @xmath149 equation ( [ f_dmu ] ) is used to eliminate @xmath150 from maxwell equations ( [ max ] ) and obtain a differential equation for @xmath151 . the electron number current is calculated using ( [ electron_current ] ) : @xmath152 and ( [ max ] , [ v_ansatz ] , [ psi ] ) yield @xmath153 ^ 2 \,\,- \gamma^2[1 + r^2 \sin^2(\xi^1)]\bigg)^{1/2}\sin(\xi^1)\,\cos(\xi^1)\ , d\xi^1\end{aligned}\ ] ] and @xmath154 where @xmath125 is the value of @xmath2 inside @xmath122 . the form of the 2nd order autonomous non - linear differential equation ( [ ode_mu ] ) for @xmath151 is fixed by specifying the generator @xmath155 of @xmath156 subject to the normalization condition ( [ norm_a ] ) . the form of the integrand in ( [ ode_mu ] ) ensures that the magnitude of oscillatory solutions to ( [ ode_mu ] ) can not be arbitrarily large . for our model , the wave - breaking value @xmath157 is the largest @xmath151 for which the argument of the square root in ( [ ode_mu ] ) vanishes , @xmath158 because @xmath159 yields an imaginary integrand in ( [ ode_mu ] ) for some @xmath160 . the positive square root in ( [ mu_wave - breaking ] ) is chosen because , as discussed above , @xmath161 and in particular @xmath162 . the wave - breaking limit @xmath163 is obtained by evaluating the first integral of ( [ ode_mu ] ) between @xmath157 where @xmath150 vanishes and the equilibriumnote that the equilibrium of @xmath151 need not coincide with the plasma s thermodynamic equilibrium . ] value @xmath164 of @xmath151 where @xmath150 is at a maximum . using ( [ norm_a ] ) to eliminate @xmath125 it follows that @xmath164 satisfies @xmath165 ^ 2 - \gamma^2[1 + r^2 \sin^2(\xi^1)]\bigg)^{1/2 } \sin(\xi^1)\cos(\xi^1 ) d\xi^1\end{aligned}\ ] ] with @xmath166 since @xmath167 . equation ( [ ode_mu ] ) yields the maximum value @xmath168 of @xmath150 , @xmath169 ^ 2 - \gamma^2 [ 1 + r^2 \sin^2(\xi^1)]\bigg)^{1/2 } \sin(\xi^1)\cos(\xi^1 ) d\xi^1\,d\mu\bigg ] . \label{e_max}\end{aligned}\ ] ] the above is a general expression for @xmath168 given @xmath155 as data . in the following , we determine a simple expression for an upper bound on @xmath168 when @xmath170 where @xmath171 is a positive constant ( @xmath172 ensures ( [ a_negativity ] ) is satified ) . using ( [ e_max ] ) it follows @xmath173\,d\mu \bigg\},\end{aligned}\ ] ] where @xmath174 ^ 2 - \gamma^2[1 + r^2 ( 1-\chi^2)]\bigg)^{1/2}\chi\,d\chi\ ] ] furthermore , ( [ mu_equilibrium ] ) may be written @xmath175 = 1\ ] ] and since @xmath176 and @xmath177 for @xmath178 , using ( [ e_max_simple ] , [ mu_equilibrium_simple ] ) @xmath179\ ] ] furthermore @xmath180 and @xmath181 so @xmath182 where @xmath183 is the plasma angular frequency ( in units where @xmath4 and @xmath3 ) . in tackling ( [ discon ] ) , one may opt to seek only those @xmath78 satisfying ( [ lorentz_force_law ] ) ; this approach was followed in the preceeding sections . although , at first sight , this method appears to be a simpler than attempting to solve ( [ discon ] ) , it is not always the simplest option . there are potential advantages in considering ( [ discon ] ) in its generality , as we will now argue . the component of the magnetic field parallel to the velocity of a point charge does not contribute to the lorentz force on that point charge . a similar observation may also be applied to certain @xmath78 in ( [ discon ] ) even though the @xmath184 components of @xmath78 are non - zero . furthermore , the results of the previous section are unaffected by a constant magnetic field aligned along @xmath147 . the axially symmetric @xmath78 introduced above is of the general form @xmath185 where @xmath186 and @xmath187 . suppose @xmath33 is of the form @xmath188 we have @xmath189 and furthermore @xmath190 so @xmath191 hence , using ( [ surface_form ] ) it follows @xmath192 and from ( [ discon ] ) @xmath193 therefore , if @xmath78 satisfies ( [ discon ] ) with @xmath194 , the same velocity field also satisfies ( [ discon ] ) with @xmath195 . it follows that a longitudinal magnetic field does not influence an axially symmetric discontinuity in the electron distribution and the results of the previous section hold for non - zero constant @xmath196 . the previous discussion clearly shows that there is merit in considering ( [ discon ] ) in its generality . we argue that for future extension of this work to fields with more complicated spacetime dependence , it is prudent to eschew ( [ lorentz_force_law ] ) in favour of ( [ discon ] ) . we now illustrate this point further using a very simple example . let the chain @xmath71 be such that @xmath197 where @xmath198 is a function on @xmath12 and , using ( [ surface_form ] ) , @xmath199 the @xmath26-acceleration of @xmath78 is @xmath200 and , for simplicity , we assume that ( [ acceleration ] ) is in response to a longitudinal electric field , @xmath201 which contributes to ( [ discon ] ) as @xmath202 clearly the lorentz force equation ( [ lorentz_force_law ] ) can not be satisfied for general @xmath198 , since @xmath203 contains terms in @xmath204 , @xmath205 which can not cancel against terms in @xmath206 . however , the only nonzero contribution to the left - hand side of ( [ discon ] ) can be made to vanish by requiring @xmath207 inspection of the @xmath208 dependences in ( [ constant_energy_v ] ) and ( [ vr ] ) reveals that @xmath198 can depend only on @xmath147 and @xmath209 so @xmath150 also depends only on @xmath147 . to compare the above with solutions to ( [ lorentz_force_law ] ) , we seek a reparameterisation of @xmath71 whose corresponding family of @xmath26-velocities satisfies ( [ lorentz_force_law ] ) . in particular , we consider a map @xmath210 @xmath211 where @xmath74 and @xmath212 . then given @xmath71 satisfying @xmath213 , @xmath214 the chains @xmath71 and @xmath215 locally represent the same discontinuity , and are physically equivalent . however , the families @xmath216 and @xmath217 of vector fields , where @xmath218 are different . we demand @xmath219 and using ( [ reparam_discont ] ) it follows @xmath220 one may solve ( [ thetaeq ] ) to determine @xmath221 , but is clear that ( for general @xmath150 ) solving for the discontinuity in terms of @xmath71 is a simpler task . furthermore , we expect this state of affairs to hold for more complicated configurations . we have developed a covariant formalism for tackling discontinuities in @xmath1-particle distributions . we have used it to develop wave - breaking limits for models of thermal plasmas whose distributions have effectively @xmath1-dimensional spacetime dependence but are @xmath0-dimensional in velocity . given tensors on a manifold @xmath12 , there are a number of ways to lift them onto the tangent manifold @xmath18 . the simplest and best known of these are the vertical lift and ( given a connection on @xmath12 ) the horizontal lift . the following is a summary of some important properties of these lifts ; for more details see , for example , @xcite . the vertical lift is a tensor homomorphism . acting on forms , it is equivalent to the pull - back with respect to the canonical projection map @xmath222 : @xmath223 acting on a vector @xmath44 , the vertical lift is @xmath224 note from these definitions , contraction of the vertical lift of a vector with the vertical lift of a form vanishes : @xmath225 the horizontal lift makes use of a connection @xmath226 , with coefficients @xmath227 in a coordinate basis @xmath228 : @xmath229 the horizontal lift @xmath230 of the basis form @xmath31 is @xmath231 while that of the basis vector @xmath232 is @xmath233 the horizontal lifts of more general 1-forms and vectors may be determined from the relations @xmath234 for any function @xmath2 on @xmath12 . two other useful identities relate to the contractions of vertical and horizontal lifts of forms and vectors : @xmath236 we thank ra cairns , b ersfeld , j gratus , ajw reitsma and rmgm trines for useful discussions . this work is supported by epsrc grant ep / e022995/1 and the cockcroft institute .

we develop a new method for describing the dynamics of @xmath0-dimensional thermal plasmas . using a piecewise constant @xmath1-particle distribution , we reduce the vlasov equation to a generalized lorentz force equation for a family of vector fields encoding the discontinuity . by applying this equation to longitudinal electrostatic plasma oscillations , and coupling it to maxwell s equations , we obtain a limit on the magnitude of the electric field in relativistic thermal plasma oscillations . we derive an upper bound on the limit and discuss its applicability in a background magnetic field .

quantum convolutional codes provide an alternative to quantum block codes to protect quantum information for reliable quantum communication . ollivier and tillich launched the stabilizer framework for quantum convolutional codes @xcite . using this stabilizer framework forney _ et al . _ constructed rate @xmath0 quantum convolutional codes @xcite . recently , two of us constructed quantum convolutional codes from product codes @xcite and derived an algorithm to construct non - catastrophic encoders and encoder inverses @xcite . in @xcite , a generalized singleton bound for a class of quantum convolutional codes has been established , together with a family of codes based on generalized reed - solomon codes meeting this bound . unit memory convolutional codes are an important class of codes that appeared in a paper by lee @xcite . he also showed that these codes have large free distance @xmath1 among other codes ( multi - memory ) with the same rate . convolutional codes are often designed heuristically . however , classes of unit memory codes were constructed algebraically by piret based on reed - solomon codes @xcite and by hole based on bch codes @xcite . in a recent paper , doubly - cyclic convolutional codes are investigated which include codes derived from reed - solomon and bch codes @xcite . these codes are related , but not identical to the codes defined in this paper . the main results of this paper are : a method to construct convolutional codes from block codes a new class of convolutional stabilizer codes based on bch codes . these codes have non - catastrophic dual encoders making it possible to derive non - catastrophic encoders for the quantum convolutional codes . we briefly recall the basic facts about classical convolutional codes relevant for our discussion . let @xmath2 be a finite field with @xmath3 elements . a _ convolutional code _ @xmath4 of length @xmath5 and dimension @xmath6 over @xmath2 is a free module of rank @xmath6 that is a direct summand of @xmath7^n$ ] . a matrix @xmath8 in @xmath7^{k\times n}$ ] such that @xmath9^k\}$ ] is called a _ basic generator matrix _ of @xmath4 , and a matrix @xmath10^{(n - k)\times n}$ ] such that @xmath11^n,{\mathbf{v}}h^t=0\}$ ] is called a _ basic parity check matrix _ of @xmath4 . the existence of a convolutional code @xmath4 is equivalent to the existence of four matrices @xmath12^{k\times n}$ ] , @xmath13^{(n - k)\times n}$ ] , @xmath14^{n\times k}$ ] , and @xmath15^{n\times ( n - k)}$ ] such that @xmath16 , @xmath17^k}$ ] , and @xmath18^{n - k}}=hl$ ] . let @xmath19 denote the maximum of the degrees among the polynomials in the @xmath20th row of a basic generator matrix @xmath8 , and let the _ memory _ @xmath21 be the maximal value of @xmath19 . a basic generator matrix of a convolutional code @xmath4 is called _ reduced _ if the _ overall constraint length _ @xmath22 has the smallest value among all basic generator matrices of @xmath4 . it is often convenient to express the generator matrix as @xmath23 , where @xmath24 . let @xmath25 be the field of laurent series consisting of elements of the form @xmath26 for @xmath27 and @xmath28 for @xmath29 for some @xmath30 . we associate with a convolutional code @xmath4 another module @xmath31 , the entries of a generator matrix @xmath8 of @xmath32 might be rational functions . let @xmath33 where @xmath34 . then we can identify @xmath35 with an element in @xmath36 as @xmath37 , where @xmath38 . we define the weight of @xmath35 as @xmath39 . a generator matrix @xmath8 is called _ catastrophic _ if there exists a @xmath40 of infinite hamming weight such that @xmath41 has finite hamming weight . the free distance @xmath1 of @xmath4 is defined as @xmath42 a rate @xmath43 convolutional code with memory @xmath21 , overall constraint length @xmath44 , and free distance @xmath1 is denoted by @xmath45 . sometimes a shorter notation @xmath46 is also used . the _ euclidean inner product _ of two @xmath5-tuples @xmath47 and @xmath48 in @xmath7^n$ ] is defined as @xmath49 . the euclidean dual of a convolutional code @xmath4 is denoted by @xmath50^n\mid { \langle { \mathbf}{u}(d)|{\mathbf}{v}(d)\rangle}=0 \text { for all } { \mathbf}{v}(d)\in c\}$ ] . note that @xmath51 , the parity check matrix of @xmath4 , _ does not _ generate the euclidean dual in general . instead , one has to reverse the order of the coefficients of the polynomials in @xmath51 , _ i.e. _ consider the matrix @xmath52 , where @xmath53 is the memory of the code generated by @xmath51 . for codes over @xmath54 , we define the hermitian inner product as @xmath55 , where @xmath56 and @xmath57 . the hermitian dual of @xmath4 is then @xmath58^n\mid { \langle { \mathbf}{u}(d)|{\mathbf}{v}(d)\rangle_h}=0 \text { for all } { \mathbf}{v}(d)\in c\}$ ] . we briefly describe the stabilizer framework for quantum convolutional codes , see also @xcite . the stabilizer is given by a matrix @xmath59^{(n - k)\times 2n}.\ ] ] which satisfies the symplectic orthogonality condition @xmath60 . let @xmath61 be a quantum convolutional code defined by a stabilizer matrix as in eq . ( [ stab - mat ] ) . then @xmath5 is called the frame size , @xmath6 the number of logical qudits per frame , and @xmath43 the rate of @xmath61 . it can be used to encode a sequence of blocks with @xmath6 qudits in each block ( that is , each element in the sequence consists of @xmath6 quantum systems each of which is @xmath3-dimensional ) into a sequence of blocks with @xmath5 qudits . the memory of the quantum convolutional code is defined as @xmath62 . we use the notation @xmath63_q$ ] to denote a quantum convolutional code with the above parameters . we can identify @xmath64 with the generator matrix of a self - orthogonal classical convolutional code over @xmath2 or @xmath54 , which gives us a means to construct convolutional stabilizer codes . analogous to the classical codes we can define the free distance , @xmath1 and the degree @xmath44 , prompting an extended notation @xmath65_q$ ] . all the parameters of the quantum convolutional code can be related to the associated classical code as the following propositions will show . for proof and further details see @xcite . [ pr : css ] let @xmath66 be a convolutional code such that @xmath67 , where the dimension of @xmath68 is given by @xmath69 . then an @xmath65_q$ ] convolutional stabilizer code exists whose free distance is given by @xmath70 , which is said to be pure if @xmath71 . [ pr : c2qherm ] let @xmath4 be an @xmath72 convolutional code such that @xmath73 . then there exists an @xmath65_q$ ] convolutional stabilizer code , where @xmath74 . in this section , we give a method to construct convolutional codes from block codes . this generalizes an earlier construction by piret @xcite to construct convolutional codes from block codes . one benefit of this method is that we can easily bound the free distance using the techniques for block codes . another benefit is that we can derive non - catastrophic encoders . given an @xmath75_q$ ] block code with parity check matrix @xmath76 , it is possible to split the matrix @xmath76 into @xmath77 disjoint submatrices @xmath78 , each of which has @xmath5 columns , such that @xmath79 . \label{eq : splith}\end{aligned}\ ] ] then we can form the polynomial matrix @xmath80 where the number of rows of @xmath81 equals the maximal number @xmath82 of rows among the matrices @xmath78 . the matrices @xmath83 are obtained from the matrices @xmath78 by adding zero - rows at the bottom such that the matrix @xmath83 has @xmath82 rows in total . then @xmath81 generates a convolutional code . the fact that the @xmath78 come from a common block code allows us to characterize the parameters of the convolutional code and its dual using the techniques of block codes . our first result concerns a non - catastrophic encoder for the code generated by @xmath81 . [ th : noncatadualenc ] let @xmath84 be an @xmath75_q$ ] linear code with parity check matrix @xmath85 . assume that @xmath76 is partitioned into submatrices @xmath86 as in eq . ( [ eq : splith ] ) such that @xmath87 and @xmath88 for @xmath89 . define the polynomial matrix @xmath81 as in eq . ( [ eq : cch ] ) . then we have : [ lm : ccbasic ] the matrix @xmath81 is a reduced basic generator matrix . [ lm : ccdual ] if the code @xmath4 contains its euclidean dual @xmath90 , respectively its hermitian dual @xmath91 , then the convolutional code @xmath92\}$ ] is contained in its dual @xmath93 , respectively its hermitian dual @xmath94 . [ lm : ccdist ] let @xmath1 and @xmath95 respectively denote the free distances of @xmath96 and @xmath93 . let @xmath97 be the minimum distance of the code @xmath98 , and let @xmath99 denote the minimum distance of @xmath68 . then the free distances are bounded by @xmath100 and @xmath101 . to prove the claim ( a ) , it suffices to show that ( i ) @xmath102 has full rank @xmath82 , ( ii ) @xmath103 , has full rank @xmath82 , where for @xmath104 we define @xmath105 , and ( iii ) @xmath81 is non - catastrophic ; cf . * theorem 2.16 and theorem 2.24 ) . by definition , @xmath106 has rank @xmath82 , so ( i ) is satisfied . condition ( ii ) is satisfied , since the rows of @xmath76 are linearly independent ; thus , the rows of the highest degree coefficient matrix are independent as well . it remains to prove ( iii ) . seeking a contradiction , we assume that the generator matrix @xmath81 is catastrophic . then there exists an input sequence @xmath107 with infinite hamming weight that is mapped to an output sequence @xmath108 with finite hamming weight , _ @xmath109 for all @xmath110 . we have @xmath111 where @xmath112 and @xmath113 . by construction , the vector spaces generated by the rows of the matrices @xmath78 intersect trivially . hence @xmath109 for @xmath110 implies that @xmath114 for @xmath115 . the matrix @xmath116 has full rank . this implies that @xmath117 for @xmath110 , contradicting the fact that @xmath118 has infinite hamming weight ; thus , the claim ( a ) holds . to prove the claim ( b ) , let @xmath119 , @xmath120 be any two codewords in @xmath121 $ ] . then from eq . ( [ eq : encoding ] ) , we see that @xmath122 and @xmath123 are in the rowspan of @xmath76 _ i.e. _ they are elements of @xmath68 , for any @xmath124 . since @xmath125 , it follows that @xmath126 , for any @xmath127 which implies that @xmath128 . hence @xmath129 . similarly , we can show that if @xmath130 , then @xmath131 . for the claim ( c ) , without loss of generality assume that the codeword @xmath132 is in @xmath93 , with @xmath133 . it follows that @xmath134 for @xmath135 , where @xmath136 denotes the @xmath137th row of @xmath81 . in particular we have @xmath138 and @xmath139 . it follows that @xmath140 and @xmath141 . if @xmath142 , then @xmath143 and @xmath144 implying @xmath145 . if @xmath146 , then @xmath147 implies @xmath148 for @xmath149 , whence @xmath150 and @xmath151 , implying that @xmath152 . it follows that @xmath153 , giving the lower bound on @xmath95 . for the upper bound note that if @xmath154 is a codeword of @xmath4 , then @xmath155 . from @xmath154 we can construct a codeword @xmath156 by padding with zeros . now , @xmath157 and hence @xmath158 . since @xmath159 we obtain that @xmath160 . finally , let @xmath161 be a non - zero codeword in @xmath96 . we saw earlier in the proof of ( b ) that every @xmath162 is in @xmath68 . thus @xmath163 . a special case of our claim ( a ) has been established by a different method in ( * ? ? ? * proposition 1 ) . one of the attractive features of bch codes is that they allow us to design codes with desired distance . there have been prior approaches to construct convolutional bch codes , see @xcite , and most notably @xcite , where one can control the free distance of the convolutional code . here we focus on codes with unit memory . our codes have better distance parameters as compared to hole s construction @xcite and are easier to construct compared to @xcite . let @xmath2 be a finite field with @xmath3 elements , @xmath5 be a positive integer such that @xmath164 . let @xmath165 be a primitive @xmath5th root of unity . a bch code @xmath4 of designed distance @xmath166 and length @xmath5 is a cyclic code with generator polynomial @xmath167 in @xmath168/\langle x^n-1\rangle$ ] whose defining set is given by @xmath169 , where @xmath170 . let @xmath171.\ ] ] then @xmath172 . if @xmath173 , then a parity check matrix @xmath76 for @xmath4 is given by writing every entry in the matrix @xmath174 as a column vector over some @xmath2-basis of @xmath175 , and removing any dependent rows . let @xmath176 denote a basis of @xmath175 over @xmath2 . suppose that @xmath177 is a vector in @xmath178 , then we can write @xmath179 for @xmath180 . let @xmath181 be vectors in @xmath182 with @xmath183 , for a vector @xmath184 in @xmath182 , we have @xmath185 if and only if @xmath186 for all @xmath183 . for a matrix @xmath187 over @xmath175 , let @xmath188 denote the matrix that is obtained by expanding each row into @xmath189 rows over @xmath2 with respect to the basis @xmath190 , and deleting all but the first rows that generate the rowspan of the expanded matrix . then @xmath191 . it is well known that the minimum distance of a bch code is greater than or equal to its designed distance @xmath166 , which is very useful in constructing codes@xcite . before we can construct convolutional bch codes we need the following result on the distance of cyclic codes . [ lm : hartman ] let @xmath164 and @xmath192 . let @xmath193 be a cyclic code with defining set @xmath194 the minimum distance @xmath195 of @xmath4 is lower bounded as @xmath196 our goal is to bound the distance of @xmath4 using the hartmann - tzeng bound ( for instance , see @xcite ) . suppose that there exists @xmath197 such that @xmath198 . suppose further , that there exists @xmath199 , where @xmath200 and @xmath201 for all @xmath202 . then by ( * ? ? ? * theorem 4.5.6 ) , the minimum distance of @xmath4 is @xmath203 . we choose @xmath204 , so that @xmath205 is satisfied for any @xmath206 . next we choose @xmath207 such that @xmath208 and @xmath209 for @xmath202 , with @xmath210 as large as possible . now two cases can arise . if @xmath211 , then there _ may not _ always exist a set @xmath212 such that @xmath208 . in this case we relax the constraint that @xmath208 and choose @xmath212 as the set of maximum number of consecutive elements . then @xmath213 and @xmath214 giving the distance @xmath215 . if @xmath216 , then we can always choose a set @xmath217 such that @xmath208 . as we want to make @xmath210 as large as possible , the worst case arises when @xmath218 . since @xmath209 holds for @xmath202 , it follows @xmath219 . thus @xmath220 . thus the distance @xmath221 . [ th : bchcc ] let @xmath5 be a positive integer such that @xmath164 , @xmath173 and @xmath222 , where @xmath223)\right\rfloor.\ ] ] then there exists a unit memory rate @xmath43 convolutional bch code with free distance @xmath224 and @xmath225 , where @xmath226 . the free distance of the dual is @xmath227 . let @xmath84 be a narrow - sense bch code of designed distance @xmath228 and let @xmath190 a basis of @xmath175 over @xmath2 . recall that a parity check matrix for @xmath4 is given by @xmath229 . further , let @xmath230 , then from @xmath231,\label{eq : bchh}\end{aligned}\ ] ] it follows that @xmath232^t$ ] , where @xmath233 is obtained from @xmath234 by removing all rows common to @xmath235 . the code @xmath236 with parity check matrix @xmath230 coincides with the narrow - sense bch code of length @xmath5 and designed distance @xmath237 . by ( * theorem 10 ) , we have @xmath238 and @xmath239 which implies @xmath240 , @xmath241 , and @xmath242 . for @xmath243 , we have @xmath244 ; therefore , @xmath245 . by theorem [ th : noncatadualenc]([lm : ccbasic ] ) , the matrix @xmath76 defines a reduced basic generator matrix @xmath246 of a convolutional code of dimension @xmath82 , while its dual which we refer to as a convolutional bch code is of dimension @xmath247 . now @xmath233 is the parity check matrix of a cyclic code , @xmath248 of the form given in lemma [ lm : hartman ] , _ i.e. _ the defining set of @xmath248 is @xmath249 as defined in ( [ eq : defining_set ] ) with @xmath250 and @xmath251 . since @xmath233 is linearly independent of @xmath252 we have @xmath253 in the definition of @xmath249 . by theorem [ th : noncatadualenc]([lm : ccdist ] ) , the free distance of the convolutional bch code is bounded as @xmath254 . by lemma [ lm : hartman ] , @xmath255 and by the bch bound @xmath256 . thus @xmath257 . the dual free distance also follows from theorem [ th : noncatadualenc]([lm : ccdist ] ) as @xmath258 . but @xmath259 by ( * ? ? ? * lemma 12 ) . under some restrictions on the designed free distance , we can use convolutional codes derived in the previous section to construct quantum convolutional codes . [ th : bchqcceuclid ] assume the same notation as in theorem [ th : bchcc ] . then there exists a quantum convolutional code @xmath61 with parameters @xmath260_q$ ] , where @xmath261 . for the free distance of @xmath61 the bound @xmath262 holds and it is pure to @xmath263 . we construct a unit memory @xmath264 classical convolutional bch code as per theorem [ th : bchcc ] . its polynomial parity check matrix @xmath81 is as given in eq . ( [ eq : bchhd ] ) . using the notation as in the proof of theorem [ th : bchcc ] , we see that the code contains its dual if @xmath76 is self - orthogonal . but given the restrictions on the designed distance , we know from ( * ? ? ? * theorem 3 ) that the bch block code defined by @xmath76 contains its dual . it follows from theorem [ th : noncatadualenc]([lm : ccdual ] ) that the convolutional bch code contains its dual . from proposition [ pr : css ] we can conclude that there exists a convolutional code with the parameters @xmath260_q$ ] . by theorem [ th : bchcc ] the free distance of the dual is @xmath265 , also implying its purity . another useful method to construct quantum codes makes use of codes over @xmath54 . [ lem : qcchermitianbch ] let @xmath266 , where and @xmath267 . then there exist quantum convolutional codes with parameters @xmath268_q$ ] and free distance @xmath269 , where @xmath270 . by theorem [ th : bchcc ] there exists an @xmath271 convolutional bch code with the polynomial parity check matrix as in eq . ( [ eq : bchhd ] ) . the parent bch code has design distance @xmath228 and given the range of @xmath166 , we know by ( * ? ? ? * theorem 14 ) that it contains its hermitian dual . by theorem [ th : noncatadualenc]([lm : ccdual ] ) , the convolutional code also contains its hermitian dual . by proposition [ pr : c2qherm ] , we can conclude that there exists an @xmath260_q$ ] code with @xmath224 . we conclude by noting that the convolutional codes in theorems [ th : bchqcceuclid ] and [ lem : qcchermitianbch ] have non - catastrophic encoders and encoder inverses . this follows directly from the fact that @xmath81 in eq . ( [ eq : bchhd ] ) is a basic generator matrix ( cf . we would like to thank one of the referees for drawing our attention to @xcite . this research was supported by nsf career award ccf 0347310 , nsf grant ccf 0622201 , and a texas a&m titf initiative .

quantum convolutional codes can be used to protect a sequence of qubits of arbitrary length against decoherence . we introduce two new families of quantum convolutional codes . our construction is based on an algebraic method which allows to construct classical convolutional codes from block codes , in particular bch codes . these codes have the property that they contain their euclidean , respectively hermitian , dual codes . hence , they can be used to define quantum convolutional codes by the stabilizer code construction . we compute bch - like bounds on the free distances which can be controlled as in the case of block codes , and establish that the codes have non - catastrophic encoders .

dynamical effects caused by chern - simons terms have been known in various places.for examples it is well - known to explain the quantum hall effects , so called statistics [email protected] the other hand vortex has an important role to wash away long - range order in xy model or condensate of ( 2 + 1)-dimensional model for superfluid helium film @xcite,@xcite.these are known as kosterlitz - thouless transition . however its detailed dynamics have not been known.it is an interesting problem to study the dynamics of the above phase transition.in this sense qed@xmath0 with chern - simons term gives us an example to show an dynamical effects of vortex on chiral condensate . qed@xmath0 with chern - simons term has a vortex solution for vector potential by solving maxwell equation with charged particle @xcite . therefore the situation is similar to the isolated vortex inside superfluid at high temperature.we can examine the phase transition by solving dyson - schwinger equation for the fermion self - energy for chiral condensate and its destruction by vortices.in this model chern - simons term is absorbed to parity odd part of the gauge boson propagator and gauge boson acquires a mass.pure qed part of gauge boson contributes to the condensation of @xmath2,while the latter may wash away the condensate at zero temperature.these are the main goals of our analysises.in 4-dimensional representation of spinor we have chiral symmetry for massless fermion.if it breaks dynamically we have two kinds of mass as chiral symmetry breaking and parity violation.for infinitesimal value of the topological mass it has been pointed out by k.i.kondo and p.maris that chiral symmetry restores and parity violating phase remains within @xmath3 expansion and nonlocal gauge of dyson - schwinger equation @xcite , raya,@xcite.here we consider weak coupling and gauge covariant approximation which satisfy ward - takahashi relation at first.after that we examine the @xmath3 expansion in the landau gauge.in the final section we evaluate the spectral function of the fermion propagator @xcite,@xcite .this method is helpful to determine the infrared behaviour or one particle singularity of the propagator in the existence of massless boson as photon , which has been known in [email protected] qed@xmath5 is super renormalizable we get a short distance behaviour of the fermion propagator too.as a result we show the effect of vortex on the lowest order spectral function of parity even scalar part of the propagator.so that we find the critical value of topological mass above which chiral condensate is washed away . fortunately this value coincides with that obtained by numerical analysis of dyson - schwinger equations . the lagrangian density of topologically massive qed@xmath0 is writtendjt . @xmath6 this system is characterized by equations of motion @xmath7 under the gauge transformation @xmath8 lagrangian is invariant but the gauge fixing term changes by a total derivative@xmath9 however action @xmath10 is gauge invariant.it is easily seen by partial integration . there exists a vortex solution of vector potential in topologically massive qed@xmath0 @xcite.it has been known by solving equation of motion at large @xmath11 @xmath12 we separate the above equation into time and space components@xmath13 where @xmath14 . using dual field strength @xmath15 relativistic wave equation for electric and magnetic fields are derived and its solution is given.here we review that part shortly.equation of motion in dual field strength is written@xmath16 multiply @xmath17 and taking trace we obtain @xmath18 and@xmath19 taking divergence of the above equation , when the current is conserved @xmath20,we have@xmath21 using equation of motion @xmath22 twice , we get@xmath23 solution of dual field strength is given @xmath24 where @xmath25 in position space we have the solution of wave equation with source @xmath26@xmath27 here we neglect the homogeneous solution.we have shown here that both electric and magnetic fields are massive and short range which decreases exponentially @xmath28let us return to the equation of motion.surface integral of first equation at large radius yields@xmath29 radial component of electric field @xmath30 decrease as @xmath31 for large @xmath32.we can safely neglect the first term at spatial infinity.the @xmath33 , is time - independent which follows from conservation of @xmath34 and proportional to total charge.static solution of the vector potential is given for classical field@xmath35@xmath36 that is even though magnetic field is short range , magnetic potential is long range.therefore we find that if qed action contains chern - simons term and charged particle , we have a vortex solution for vector potential as an aharonov - bohm effect.the above vortex solution with quantization of @xmath37 is well known in quantum hall systems which explains quantization of hall conductance and the possibility of statistics transmutaion.hereafter we will examine possibility of wash away condensate of chiral order parameter in qed@xmath38 with massless four component fermion by the existence of vortex . here we consider the dynamical effects as mass generation in the presence of chern - simons term . if the fermion is massless @xmath39 , has @xmath40 symmetry generated by @xmath41 for @xmath42 @xmath43/2=\left ( \begin{array}{cc } i & 0 \\ 0 & -i% \end{array } \right ) .\gamma_{3}$ ] and @xmath44 act as chiral transformation@xmath45 scalar density is mixed by above transformation@xmath46 dynamical mass generation breaks @xmath40 symmetry down to @xmath47 which is generated by @xmath48 , @xcite.there exists discrete parity transformation . @xmath49 parity even and odd mass are transformed @xmath50 both of these mass terms are invariant under @xmath51 transformation.@xmath52 we find the eigenvalue of the free particle hamiltonian @xmath53 as @xmath54where @xmath55 is a @xmath56 unit matrix and @xmath57 is a operator defined above.two kinds of mass are written by@xmath58 so that we may split 4-component spinor into upper and lower components by projection operator@xmath59 from the lagrangian @xmath60 we have the free propagator @xmath61 since chern - simons term induces parity odd fermion mass , it is convenient to introduce parity odd mass from the beginnings and choose the basis of the eigenvalue of the hamiltonian.chiral representation is enough for this purposes.in our 4-component spinor representation parity violating mass @xmath62 is a pseudo scalar and current @xmath63 transform as an vector , which mix each other under parity and charge conjugation transformation with arbitrary phase.starting from dirac equation@xmath64 combination of above two equation , we obtain@xmath65 where @xmath66 a more general proof by ward - takahashi - relation of the vector current,@xmath67 by operator of symmetry transformation , @xmath68 & = \delta\psi(y)\delta ^{(3)}(x - y)=\gamma_{3,5}\psi(y)\delta^{(3)}(x - y ) , \\ \delta(x_{0}-y_{0})[j_{3,50}(x),\overline{\psi}(y ) ] & = \delta\overline{\psi } % ( y)\delta^{(3)}(x - y)=\overline{\psi}(y)\gamma_{3,5}\delta^{(3)}(x - y).\end{aligned}\ ] ] in terms of the vertex function it is written@xmath69@xmath70 where the subscript @xmath71 means truncation of the fermion propagator.for the proof of ward - takahashi relation in terms of integral equation for @xmath72 see @xcite.for vanishing bare mass @xmath73,current conservation is spontaneously broken by parity even mass generation@xmath74 in other form it is well known @xmath75 taking vacuum expectation value of both sides of equation , non - vanishing of the r.h.s indicates the existence of massless scalar ng boson , in the vertex @xmath76,of the form@xmath77 where @xmath78 is a bethe - salpeter amplitude for total momentum @xmath79.it is equal to scalar part of propagator by ward - takahashi relation . if the right hand side is finite and we have two order parameter @xmath80 of chiral symmetry breaking and parity violation . dyson - schwinger equation is familiar to obtain non - perturbative propagator and dynamical mass.therefore we set up the dyson - schwinger equation for the electron propagator in this representation . propagator of fermion and gauge boson are given @xmath81.\end{aligned}\ ] ] the quenched schwinger - dyson equation for the self - energy @xmath82 is written @xmath83 where @xmath84 in this normalization @xmath85 is real in the region @xmath86we separate the vector and scalar part of the propagator by taking the trace of @xmath87trace formulae for @xmath88 @xmath89 matrices are reduced to @xmath90 ones with projection operator @xmath91@xmath92 in chiral representation parity violating mass and vector part are different sign for up and down component.we have the coupled integral equation for @xmath93 and @xmath94 in the landau gauge by taking trace@xmath95 to keep gauge invariance of the action we adopt the gauge covariant approximation by introducing ball - chiu vertex ansatz for longitudinal part[9].@xmath96 in this case the dyson - schwinger equations are following equations in the landau gauge @xmath97@xmath98,\]]@xmath99\end{aligned}\ ] ] where @xmath100 , @xmath101 angular integral formulae are given in appendices a. for bare vertex we set @xmath102 . recently we obtained the numerical solutions with this approximation@xcite , hiroshima.it is said that the vortex destroys condensate in condensed matter physics.at bare vertex or inclusion the bc vertex we find @xmath103 which is the same order of magnitude with raya et.al@xcite in fig.1.simple vertex correction as @xmath104 yields @xmath105 [ ptb ] fig1.eps numerically some instability arise from @xmath106 term in @xmath107 equation.this term is mass changing and compete the reduction of mass by parity violating mass term.in pure qed this term is sensitive for low momentum and infrared cut - off of momentum integration . the effects on vacuum expectation value is relatively large.if we demand that the mass @xmath108 vanishes at the transition point and neglect this term , we find the clear transition as the bare vertex case with the same critical point.we can determine it by introducing renormalization of the topological mass as@xmath109 where @xmath110is determined self - consistently in fig.2.by this analysis critical point @xmath111 in the bare vertex case turned to be the true one . renormalization of topological mass is shown in appendix.at least for low energy we can safely neglect @xmath112 terms and find clear phase transition.in fig.3 we see the parity violating order parameter @xmath113 as a function of topological mass.near the critical point @xmath111,@xmath114 is lower than @xmath115 which is shown in fig.4 and first order phase transition occurs . [ ptb ] fig2.eps [ ptb ] fig3.eps [ ptb ] fig4.eps [ ptb ] fig5.eps [ ptb ] fig6.eps using free vector propagator , @xmath116 we shall calculate one - loop corrections to the vector propagator@xmath117 and set up the dyson - schwinger equation with bare vertex@xmath118 one - loop vacuum polarization is evaluated for fermion mass @xmath119 with pauli - villars or dimensional reguralization to remove gauge nonivariant cut - off dependent term and parity - odd chern - simons term is not induced here.see appendix and @xcite .@xmath120 , \\ t_{\mu\nu } & = ( g_{\mu\nu}-\frac{k_{\mu}k_{\nu}}{k^{2}}),\overline{d^{3}}p=% \frac{d^{3}p}{(2\pi)^{3}},\end{aligned}\ ] ] polarization function @xmath121 is @xmath122 , \notag \\ & = \frac{-e^{2}}{8}n\sqrt{-k^{2}}(k^{2}>0,m=0 ) , \notag \\ & = \frac{-e^{2}n}{6\pi m}k^{2}+o(k^{4})(k^{2}/m\ll1).\end{aligned}\ ] ] for @xmath39 case @xmath123 gives a non - perturbative correction to the vector propagator for low energy in the chiral symmetric phase.from the relation@xmath124 we have the inverse of the bare photon propagator@xmath125 adding the vacuum polarization of massless loop @xmath126+i\frac{p_{\mu}p_{\nu}}{\xi},\end{aligned}\ ] ] we obtain the full propagator @xmath127 in euclid space it has the form@xmath128 where @xmath129 including vacuum polarization for photon the dyson - schwinger equation has the following form @xmath130 in euclid space with @xmath131 gauge , we have@xmath132(k^{2}+% \pi(k^{2}))^{2}+\mu^{2}k^{2 } ) } \notag \\ & \mp\frac{\mu a_{\pm}(q)(q\cdot k)}{[a_{\pm}^{2}(q)q^{2}+b_{% \pm}^{2}(q)](k^{2}+\pi(k^{2}))^{2}+\mu^{2}k^{2 } ) } ] , \\ a_{\pm}(p ) & = 1+\frac{e^{2}}{4\pi^{3}p^{2}}\int d^{3}q[\frac{(k^{2}+\pi ( k^{2}))a_{\pm}(q)(p\cdot k)(q\cdot k)}{[a_{\pm}^{2}(q)q^{2}+b_{% \pm}^{2}(q)](k^{2}+\pi(k^{2}))^{2}+\mu^{2}k^{2})k^{2 } } \notag \\ & \mp\frac{\mu b_{\pm}(q)(p\cdot k)}{[a_{\pm}^{2}(q)q^{2}+b_{% \pm}^{2}(q)](k^{2}+\pi(k^{2}))^{2}+\mu^{2}k^{2})}].\end{aligned}\ ] ] here we apply the ball - chiu vertex as in the quenched case.kondo and maris applied @xmath3 approximation to solve the equation.however in the gauge covariant approximation which satisfy ward - takahashi relation has been shown that chiral order parameter in quenched and unquenched case are same at @xmath133 for weak coupling for @xmath134 case.so we choose covariant gauge and weak coupling.it is easy to take bc vertex and improve the dyson - schwinger equation as in the quenched case.but we have not enough memories to run pc.so that we only show the massless loop correction in the landau gauge case . in this case critical value @xmath111 in unquenched case is approximately the same value for quenched case.for example we find numerically @xmath135 for @xmath133 and @xmath136 for @xmath137at the critical point we find that @xmath138 , @xmath139 and @xmath140 vanishes above the critical point.this is the destruction mechanism of superfluidity by vortex in our model . for @xmath141 case we have very small values of order parameter for small topological mass.for large topological mass order parameter changes its sign at some value of topological mass.there may be a strong coupling phase for @xmath142 at least for small topological mass , where vacuum expectation value @xmath143 vanishes.for these cases @xmath144 expansion may be a good way to study the phase structure for strong coupling region in our model.a famous critical number of flavour which was derived in the linearized schwinger - dyson equation has been know as @xmath145 above which the chiral symmetry is restored.the results of @xmath3 expansion and the phase structure derived by ki.kondo and p.maris may be realized@xcite . in this section we would like to determine critical value of topological mass above which chiral condensate is washed away theoretically.first we notice that the spectral function for the propagator as one of the possibility citeops.in three dimension , absence of ultraviolet divergences is important.if we know only infrared behaviour or the leading logarithm of infrared divergence near the mass shell , it is possible to determine the whole region of the propagator in position space by the anomalous dimension , which is supplied by lowest ordered spectral function.by this method we find that only short distance behaviour of the propagator is modified and we have a finite chiral condensate for pure qed@xmath5 hoshino.if we choose soft - photon exponentiation to include all orders of soft - photon emission by electron , its spectral function may be written as @xmath146 , where @xmath147 is a model independent spectral function of the lowest order in the coupling constant for pure qed@xmath5 @xmath148 here we consider the fermion spectral function.the vacuum expectation value of the anticommutator has the form @xcite @xmath149.\end{aligned}\ ] ] we introduce the spectral amplitude by grouping together in the sum over @xmath150 all states of given three - momentum @xmath151@xmath152 and set out to construct its general form from invariance arguments.@xmath153 is a @xmath154 matrix and may be expanded in terms of @xmath155 linearly independent products of @xmath89 matrices.under the assumptions of lorentz invariance and parity transformation it reduces to the form@xmath156 second term in ( 82 ) can be related directly to ( 83 ) with the aid of pct invariance of the vacuum @xcite,@xcite.parity , charge conjugation and time reversal transformation are defined in our representation of @xmath89 matrices @xmath157@xmath158@xmath159@xmath160 effects of pct transformation on @xmath161 is@xmath162 inserting ( 98 ) along with ( 83 ) into ( 82 ) and using @xmath163,we obtain finally @xmath164_{\alpha\beta } e^{-iq\cdot(x - y ) } \notag \\ & + \{\gamma^{1}[\gamma\cdot q\rho_{1}(q^{2})+\rho_{2}(q^{2})]\gamma ^{1}\}_{\alpha\beta}e^{iq\cdot(x^{\prime}-y^{\prime } ) } ) \notag \\ & = \int\frac{d^{3}q}{(2\pi)^{2}}\theta(q_{0})[\rho_{1}(q^{2})i\gamma \cdot\partial_{x}+\rho_{2}(q^{2})]_{\alpha\beta}(e^{-iq\cdot(x - y)}-e^{iq% \cdot(x^{\prime}-y^{\prime } ) } ) \notag \\ & = \int\frac{d^{3}q}{(2\pi)^{2}}[\theta(q_{0})\gamma\cdot q\rho_{1}(q^{2})+\epsilon(q_{0})\rho_{2}(q^{2})]_{\alpha\beta}e^{-iq% \cdot(x - y)},\end{aligned}\ ] ] where @xmath165since @xmath166 vanishes for space - like @xmath167,we may also write this as an integral over mass spectrum by introducing@xmath168 we find @xmath169i\delta(x - y;\sqrt{s } ) \notag \\ & = \int ds\{\rho_{1}(s)is(x - y;\sqrt{s})+[\sqrt{s}\rho_{1}(s)-\rho _ { 2}(s)]i\delta(x - y;\sqrt{s})\}\end{aligned}\ ] ] where invariant @xmath170 function is given@xmath171|0\right\rangle \\ & = \sum_{n}\left\langle 0|\phi(0)|n\right\rangle \left\langle n|\phi ( 0)|0\right\rangle ( e^{-ip_{n}\cdot(x - y)}-e^{ip_{n}\cdot(x - y ) } ) \notag \\ & = \frac{1}{(2\pi)^{2}}\int d^{3}q\rho(q^{2})\theta(q_{0})(e^{-iq\cdot ( x - y)}-e^{iq\cdot(x - y ) } ) \\ & = \frac{1}{(2\pi)^{2}}\int_{0}^{\infty}ds\rho(s)\int d^{3}q\delta ( q^{2}-s)\epsilon(q_{0})e^{-iq\cdot(x - y ) } \notag \\ & = \int_{0}^{\infty}ds\rho(s)i\delta(x - y,\sqrt{s}).\end{aligned}\ ] ] the above spectral representation goes through unchanged for the vacuum expectation value of the time - ordered product of dirac field;it is necessary only to replace the @xmath172 and @xmath173 by the feynman propagator @xmath174 and @xmath175.if we know the matrix element @xmath176,we can determine the spectral function @xmath177and @xmath178.perturbative @xmath179 spectral function can be obtained by the usual definition @xmath180 in perturbation theory one photon emission matrix element is given@xmath181r;k\text { in}\right\rangle \notag \\ & = -i\left\langle 0^{in}|t[\psi^{in}(0),e\int d^{3}x\overline{\psi}% ^{in}(x)\gamma_{\mu}\psi^{in}(x)a_{\mu}^{in}(x)]|r;k\text { in}\right\rangle \\ & = -ie\int d^{3}xs_{f}^{0}(0-x)\gamma_{\mu}\left\langle 0|\psi^{in}(x)|r\right\rangle \left\langle 0|a_{\mu}^{in}(x)|k\right\rangle \notag \\ & = \frac{-ie}{\gamma\cdot(r+k)-m+i\epsilon}\gamma_{\mu}\epsilon_{\lambda } ^{\mu}(k)u_{s}(r)\sqrt{\frac{m}{e_{r}}}\frac{1}{\sqrt{2k_{0}}}.\end{aligned}\ ] ] for the evaluation of @xmath182,if we integrate @xmath183 first , we obtain @xmath184 for energy - momentum conservation.in our case first we integrate @xmath185.after that we exponentiate the function @xmath147 and integrate @xmath186 in the non perturbative case.at that stage the results are position dependent.finally we obtain the spectral function in position space @xmath187 for infinite number of photon emission@xmath188 where the notation @xmath189 @xmath190 is used and @xmath191 may be replaced to @xmath192 in our approximation.the polarization sum for gauge boson is given @xmath193.\ ] ] since we do not fix the numbers of photon @xmath150 we sum up from @xmath194equals zero to infinity.the function @xmath195 is a phase space integral of one photon intermediate state which may be exponentiated as @xmath196we use the trace formula to get @xmath197 to order @xmath198 we have model independent spectral function for 4-component fermion @xmath199.\ ] ] hereafter we keep in mind the projection operator of positive energy @xmath200 for the propagator.from this form @xmath201 in our approximation.to obtain the explicit form of the function @xmath147 , we use parameter trick @xmath202 combined with positive frequency part of the propagator with bare mass @xmath37@xmath203 we obtain the function @xmath147 symbolically as @xmath204 we get each terms @xmath205@xmath206 where @xmath207 . for short distance we have @xmath208 above spectral function contains linear infrared divergent term proportional to @xmath209this term depends on the gauge parameter.so we choose @xmath210 gauge to drop it.other terms are independent of @xmath211 in 4-dimension well - known infrared behaviour of charged particle is reproduced in this way.in qed@xmath0 we have used this technique to determine a short - distance behaviour of the propagator.in the long distance region propagator behaves as free one for finite @xmath37.if we set anomalous dimension which is a coefficient of @xmath212 to be unity @xmath213we obtain the propagator at short distance as@xmath214 which shows condensation @xmath215.in this case the trace means positive energy part only.for massless gauge boson vacuum expectation value is infrared cut - off dependent.if we set @xmath216we have @xmath217 which is very close to the numerical value by solving dyson - schwinger equation in euclid momentum space.in our topologically massive qed , there are two order parameters @xmath218 that have non vanishing value for small topological mass . just above the critical point there exists only parity violating mass with non - vanishing order parameter @xmath219.so that we may expect that the chiral symmetry breaking mass vanishes and parity violating condensate remains at the critical point.we show this is the case in evaluating the lowest order spectral function with chern - simons term.in the case of parity violation , we may include parity violating part of the spectral function @xmath220.since we adopt the chiral representation of the propagator in solving the dyson - schwinger equation , we will determined spectral functions in chiral representation too . contribution of chern - simons term for the spectral function in chiral representation is given in the appendix b@xmath221\delta(k^{2}-\mu^{2}).\ ] ] after the integration we obtain the correction of wave function renormalization by chern - simons term ( @xmath222 term ) as @xmath223 @xmath224 for @xmath225.at short distance we take into account only @xmath226 using @xmath227have @xmath228 final form is written as @xmath229 for short distance.in chiral representation if @xmath230we have non vanishing condensation of @xmath231.otherwise @xmath232.below the critical value of the topological mass @xmath233 there are two kind of mass with different anomalous dimension.in that case we can not separate them in the chiral representation.however at the @xmath234 there is only parity odd mass and and its condensate.so that we set wave function renormalization @xmath235,other contribution is only logarithmic.therefore we set @xmath236 and have @xmath237 for critical value of @xmath37.in this case we have only parity violating order parameter @xmath238 above the critical point.this is just the desirable form to give parity odd order parameter @xmath239so that at the critical point chiral order parameter vanishes and it undergoes into parity violating phase . so we conclude that the critical value of the topological mass is @xmath240 which is totally consistent with the value in our numerical analysis of dyson - schwinger equation for quenched case.this transition is the same as kosterlitz - thouless type at finite temperature where single vortex excitation destroys the superfluidity.so that the topological massive qed@xmath0 is the same with qed@xmath0 with single vortex at zero temperature except for small topological mass.in the statistical model , behaviour of the propagator near the critical point is studied with renormalization group equation for vortex number ( chemical potential ) and the inverse temperature[2,3 ] . in our model spectral function provide us anomalous dimension and we can determine the ultraviolet behaviour and critical point in the existence of topological mass.fortunately our dyson - schwinger equation successfully determines the structure of the propagator and dynamical mass near the critical region . in this work we studied the dynamics of kosterlitz - thouless type transition in topologically massive abelian gauge theory in three dimensional space-time.in this model equation of motion contain vortex solution for the vector potential.we showed that there exists a critical value of the topological mass above which chiral condensate is washed away for four component fermion.this phenomenon turned out to be gauge invariant by the choice of bc vertex in solving dyson - schwinger equation for fermion self-energy.in the analysis of spectral function we showed it modify the short distance behaviour of the propagator in position space.in that case anomalous dimension control the order parameter of chiral condensate in the absence of topological mass.however if we add chern - simons term to the lagrangian parity odd part of the gauge boson propagator is destructive to parity even part . in this way lowest order spectral function vanishes at the critical value of topological mass and the chiral condensate is washed away . our next step is to evaluate critical temperature by solving dyson - schwinger equation or spectral function , temperature dependence of specific heat and compare them with the experiment . 99 a.m.polyakov,fermi-bose transmutations induced by gauge fields,mod.phys.lett.a**3**,3(1988)325 - 328 . j. m. kosterlitz and d. j.thouless,ordering,metastability and phase transition in two - dimensional systems,j.phys.c**6 * * , 1181 - 1203 ( 1973 ) . j. m. kosterlitz,the critical properties of the two - dimensional xy model,j.phys.c**7**,1046 - 1060 ( 1974 ) . s. deser , r . jackiw and s. templeton,topologically massive gauge theories,*annals of physics * 281,409 - 449(2000 ) . k. i. kondo , p . spontaneous chiral - symmetry breaking in three - dimensional qed with chern - simons term,phys.rev.*d52 , 1212 - 1233 ( 1995 ) . * s. s. madrigal , c . p. hofmann and a. raya,dynamical mass generation and confinement in maxwell - chern - simons planar electrodynamics,*j.phys.conf.ser*.287:012028,2011 . yuichi hoshino , tomohiro inagaki and yuichi mizutani,gauge covariant solution for schwinger - dyson equation in 3d qed,*ptep2 * 2015,2,023b03 . r.jackiw and l.soloviev,low-energy-theorem approach to single particle singularity in the presence of massless bosons,phys.rev.*173,1458 ( 1968 ) . * yuichi.hoshino,mass singularity and confining property in qed@xmath0,jhep09(2004)048 . c.j.burden,bound states from the bethe - salpeter equation in qed3,nucl.phys.b**387**,419 - 446 ( 1992 ) . t.maskawa and h.nakajima,spontaneous breaking of chiral symmetry in a vector - gluon model,prog.theo.phys.*56*(1974)1326 . c.s.fishr,r.alkofer,t.dham,p.maris,dynamical breaking of chiral symmetry in unquenched qed@xmath0,phys.rev.d*70*:073007,2004 . james d.bjorken and sidney d.drell,*relativistic quantum fields*,mcgraw - hill book company . james d.bjorken and sidney d.drell,*relativistic quantum mechanics*,mcgraw - hill book company . after angular integral we have the following coupled integral equation @xmath242\mp\mu a(q)_{\pm}i_{2}(p , q)_{- } ) \notag \\ & + \{\delta a_{\pm}(b(q)_{\pm}\mp\frac{\mu}{2}a(q)_{\pm})-\delta b_{\pm } a(q)_{\pm}\}i_{1}(p , q)],\end{aligned}\]]@xmath243.\end{aligned}\ ] ] for unquenched case , to evaluate angular integral we may use complex number to represent parity even and odd piece of the photon propagator @xmath244 it has been known that the integral kernel is a logarithmic function as@xmath245 where @xmath246 so that the angular integral with finite topological mass @xmath37 may be a analytic continuation from the case @xmath134 to @xmath247@xmath248 for example@xmath249 the integration of the type @xmath250 is rewritten @xmath251 in the same way @xmath252 @xmath253 leads@xmath254 since@xmath255@xmath256.\end{aligned}\ ] ] the dyson - schwinger equations are rewritten as @xmath257}\pm\frac{a_{\pm}(q)j_{2}(p , q)_{-}}{% [ a_{\pm}^{2}(q)q^{2}+b_{\pm}^{2}(q ) ] } ] , \notag \\ a_{\pm}(p ) & = 1+\frac{e^{2}}{2\pi^{2}p^{2}}\int q^{2}dq[\frac{a_{\pm } ( q)j_{3}(p , q)}{[a_{\pm}^{2}(q)q^{2}+b_{\pm}^{2}(q)]}\pm\frac{% b_{\pm}(q)j_{2}(p , q)_{+}}{[a_{\pm}^{2}(q)q^{2}+b_{\pm}^{2}(q)]}].\end{aligned}\ ] ] in a two - dimensional representation , the trace of products of four @xmath258-matrices are : @xmath259@xmath260@xmath261 substituting the chiral representation of the propagator into @xmath262@xmath263@xmath264@xmath265@xmath266@xmath267 where@xmath268\delta_{\mu\nu}-\frac{1}{3}p^{\prime 2}\delta_{\mu\nu}-2k_{\mu}k_{\nu}x(1-x).\end{aligned}\ ] ] regulating the ultraviolet divergence by cut - off @xmath269 we obtain the vacuum polarization tensor for two - component fermion@xmath270+% \frac{e^{2}\lambda}{3\pi^{2}}\delta_{\mu\nu}.\end{aligned}\ ] ] in minkowski space we have@xmath271 from the above expression , vacuum polarization tensor for chiral representation of fermion is given @xmath272@xmath273 subsectionspectral function here we derive the contributions of chern - simons term to the spectral function.for scalar and vector part are given by trace@xmath274@xmath275 where @xmath89 matrices are @xmath276first we evaluate @xmath277@xmath278\chi_{+}\gamma_{\mu}(r\cdot\gamma+m)\gamma_{% \nu})i\epsilon_{\mu\nu\rho}\frac{k_{\rho}}{\mu } \notag \\ & = [ ( 2(m^{2}+r\cdot k)+k^{2})r_{\sigma}tr(\chi_{+}\gamma_{\mu}\gamma_{\sigma } \gamma_{\nu})+2m^{2}(r+k)_{\sigma}tr(\chi_{+}\gamma_{\sigma}\gamma_{\mu } \gamma_{\nu})]i\epsilon_{\mu\nu\rho}\frac{k_{\rho}}{\mu}.\end{aligned}\]]@xmath279 next we evaluate @xmath280@xmath281@xmath282

there is chiral like symmetry for 4-component massless fermion in ( 2 + 1)-dimensional gauge theory.since qed@xmath0 with chern - simons term contains vortex solution for vector potential , one may expect vortex driven phase transition as kosterlitz - thouless type where chiral condensate is washed away at zero temperature.to study this possibility , we evaluate the fermion propagator by dyson - schwinger equation numerically and spectral function analytically in the landau gauge.for quenched case we adopt ball - chiu vertex to keep gauge invariance of the results.the critical value of topological mass , above which chiral condensate washed away , turned out to be @xmath1 at least for weak coupling in both cases .

while conventional causal inference methods @xcite use conditional independences to infer a directed acyclic graph of causal relations among at least three random variables , there is a whole family of recent methods that employ more information from the joint distribution than just conditional independences @xcite . therefore , these methods can even be used for inferring the causal relation between just two observed variables ( i.e. , the task to infer whether @xmath0 causes @xmath1 or @xmath1 causes @xmath0 , given that there is no common cause and exactly one of the alternatives is true , becomes solvable ) . as theoretical basis for such inference rules , @xcite postulate the following asymmetry between cause and effect : if @xmath0 causes @xmath1 then @xmath2 and @xmath3 represent independent mechanisms of nature and therefore contain no information about each other . here , `` information '' is understood in the sense of description length , i.e. , knowing @xmath2 provides no shorter description of @xmath3 and vice versa , if description length is identified with kolmogorov complexity . this makes the criterion empirically undecidable because kolmogorov complexity is uncomputable @xcite . @xcite pointed out that `` information '' can also be understood in terms of predictability and used this to formulate the following hypothesis : semi - supervised learning ( ssl ) is only possible from the effect to the cause but not visa versa . this is because , if @xmath0 causes @xmath1 , knowing the distribution @xmath4 may help in better predicting @xmath0 from @xmath1 since it may contain information about @xmath5 , but @xmath2 can not help in better predicting @xmath1 from @xmath0 . information - geometric causal inference ( igci ) @xcite has been proposed for inferring the causal direction between just two variables @xmath0 and @xmath1 . in its original formulation it applies only to the case where @xmath0 and @xmath1 are related by an invertible functional relation , i.e. , @xmath6 and @xmath7 , but some positive empirical results have also been reported for noisy relations @xcite . we will also restrict our attention to the noiseless case . this is because attempts to generalize the theory to non - deterministic relations are only preliminary @xcite . moreover , the deterministic toy model nicely shows _ what kind _ of dependences between @xmath4 and @xmath5 occur while the dependences in the non - deterministic case are not yet well understood . we first rephrase how igci has been introduced in the literature and then explain our new interpretations . they also provide a better intuition about the relation to ssl . for didactic reasons we restrict the attention to the case where @xmath8 is a monotonously increasing diffeomorphism of @xmath9 $ ] . we assume that @xmath2 and @xmath4 have strictly positive densities @xmath10 and @xmath11 . we often write @xmath12 instead of @xmath13 whenever this causes no confusion . then @xcite assumes : [ psot : uncorr ] for @xmath0 causing @xmath1 , @xmath14 the interpretation that ( [ uncorr ] ) is an independence condition becomes more clear when the functions @xmath15 and @xmath16 are interpreted as random variables on @xmath9 $ ] . then , the difference between the left and the right hand side of ( [ uncorr ] ) is the covariance of @xmath10 and @xmath17 with respect to the uniform distribution @xcite . the intuition is that it is unlikely , if @xmath8 and @xmath2 are chosen independently , that regions where the slope of @xmath8 is large ( i.e. large @xmath17 ) , meet regions where @xmath10 is large and others where @xmath10 is small . simple calculations @xcite show that ( [ uncorr ] ) implies that @xmath11 is positively correlated with the slope of @xmath18 since @xmath19 with equality iff @xmath20 . this is illustrated in figure [ fig : corr ] a ) . a ) ( taken from @xcite ) if the fluctuations of @xmath10 do nt correlate with the slope of the functions , regions of high density @xmath11 tend to occur more often in regions where @xmath8 is flat . b ) @xmath21 maps the cube to itself . the regions of points @xmath22 with large @xmath23 ( here : the leftmost sphere ) can only be a small fraction of the cube.,title="fig:",scaledwidth=30.0% ] a ) ( taken from @xcite ) if the fluctuations of @xmath10 do nt correlate with the slope of the functions , regions of high density @xmath11 tend to occur more often in regions where @xmath8 is flat . b ) @xmath21 maps the cube to itself . the regions of points @xmath22 with large @xmath23 ( here : the leftmost sphere ) can only be a small fraction of the cube.,title="fig:",scaledwidth=33.0% ] moreover , using @xmath24 , eq . ( [ uncorr ] ) implies @xmath25 using @xmath26 we get @xmath27 with equality only for @xmath28 . for empirical data @xmath29 , with @xmath30 ( and hence @xmath31 ) , this suggests the following inference method : [ def : igci ] infer @xmath32 whenever @xmath33 some robustness of igci with respect to adding noise has been reported @xcite when the following modification is used : on the left hand side of ( [ empigci ] ) the @xmath34-tuples are ordered such that @xmath30 , while the right hand side assumes @xmath35 . note that in the noisy case , the two conditions are not equivalent . moreover , the left hand side of ( [ empigci ] ) is no longer minus the right hand side since eq . ( [ minus ] ) no longer makes sense . albeit hard to formulate explicitly , it is intuitive to consider the left hand side as measuring `` non - smoothness '' of @xmath3 and the right hand side the one of @xmath5 . then , the causal direction is the one with the smoother conditional . to describe the information theoretic content of ( [ uncorr ] ) and ( [ pc ] ) , we introduce the uniform distributions @xmath36 and @xmath37 for @xmath0 and @xmath1 , respectively . their images under @xmath8 and @xmath18 are given gy the probability densities @xmath38 and @xmath39 , respectively . we will drop the superscripts @xmath8 and @xmath18 whenever the functions they refer to are clear . then ( [ uncorr ] ) reads @xmath40 and is equivalent to the following additivity of relative entropies @xcite : @xmath41 likewise , ( [ pc ] ) reads @xmath42 in the terminology of information geometry @xcite , ( [ orth1 ] ) means that the vector connecting @xmath10 and @xmath36 is orthogonal to the one connecting @xmath36 and @xmath43 . thus the `` independence '' between @xmath10 and @xmath8 has been phrased in terms of orthogonality , where @xmath8 is represented by @xmath44 . likewise , the dependence between @xmath11 and @xmath18 corresponds to the fact that the vector connecting @xmath11 and @xmath37 is not orthogonal to the one connecting @xmath37 and @xmath45 . the information - theoretic formulation motivates why one postulates uncorrelatedness of @xmath10 and @xmath17 instead of one between @xmath10 and @xmath46 itself . a further advantage of this reformulation is that @xmath36 and @xmath37 can then be replaced with other `` reference measures '' , e.g. , gaussians with the same variance and mean as @xmath10 and @xmath11 , respectively ( which is more appropriate for variables with unbounded domains ) @xcite . however , both conditions ( [ uncorr ] ) and ( [ orth1 ] ) are quite abstract . therefore , we want to approach igci from completely different directions . in section [ sec : untyp ] we will argue that a large positive value for @xmath47 shows that the observed @xmath34-tuple @xmath48 is untypical in the space of all possible @xmath34-tuples . in section [ sec : numb ] we show that condition ( [ empigci ] ) implies that there are , in a sense , more functions from @xmath0 to @xmath1 than vice versa . in section [ sec : ssl ] we explain why the correlation between distribution and slope that occurs in the anticausal direction helps for unsupervised and semi - supervised learning . let us consider again a monotonously increasing diffeomorphism @xmath49\rightarrow [ 0,1]$ ] and explain in which sense a point @xmath50 $ ] can have a `` typical '' or an `` untypical '' position relative to @xmath8 . consider the function @xmath8 shown in figure [ fig : one_point ] a ) . the point @xmath51 is untypical because it meets @xmath8 in a region whose slope is larger than for the majority of points . of course , @xmath51 can also be untypical in the sense that the slope of @xmath51 is smaller than for the majority of points , see figure [ fig : one_point ] b ) . there is , however , an important asymmetry between large slope and small slope : if the slope at @xmath51 is significantly higher than the _ average _ slope over the entire domain , then @xmath51 is necessarily untypical because the slope can significantly exceed the average only for a small fraction of points . if the slope is significantly _ below _ the average , this does not mean that the point is untypical because this may even be the case for most of the points , as one can easily see on figure [ fig : one_point ] a ) . this asymmetry is known from statistics : a non - negative random variable may quite often attain values that are smaller than their expected value by orders of magnitude , but exceeding the expectation by a large factor is unlikely due to the markov inequality . ( a ) @xmath51 has an untypical position in both cases because it meets in a small region with large slope or in a region with small slope ( b ) . c ) function on a grid . the large dots denote the observed points , the small ones visualize one option of interpolating by a monotonic function that crosses @xmath52 and @xmath53.,title="fig:",scaledwidth=30.0% ] ( a ) @xmath51 has an untypical position in both cases because it meets in a small region with large slope or in a region with small slope ( b ) . c ) function on a grid . the large dots denote the observed points , the small ones visualize one option of interpolating by a monotonic function that crosses @xmath52 and @xmath53.,title="fig:",scaledwidth=30.0% ] ( a ) @xmath51 has an untypical position in both cases because it meets in a small region with large slope or in a region with small slope ( b ) . c ) function on a grid . the large dots denote the observed points , the small ones visualize one option of interpolating by a monotonic function that crosses @xmath52 and @xmath53.,title="fig:",scaledwidth=30.0% ] the above idea straightforwardly generalizes to mappings between multi - dimensional spaces : then a point @xmath54 can be untypical relative to a function @xmath21 in the sense that the jacobian of @xmath21 is significantly larger than the average . this is , for instance , the case for the points in the leftmost sphere of figure [ fig : corr ] b ) . we first introduce `` untypical '' within a general measure theoretic setting : [ thm : unt ] let @xmath55 and @xmath56 be probability distributions on measure spaces @xmath57 and @xmath58 , respectively . let @xmath59 be measurable and let the image of @xmath55 under @xmath21 have a strictly positive density @xmath60 with respect to @xmath56 . then , points @xmath22 for which @xmath61 are unlikely ( `` untypical '' ) in the sense that @xmath62 for all @xmath63 . [ cor : hyper ] let @xmath64^n \rightarrow [ 0,1]^n$ ] be a diffeomorphism . then the volume of all points @xmath22 for which @xmath65 is at most @xmath66 . the corollary is very intuitive : since the average of @xmath23 over the hypercube is @xmath67 , the fraction of @xmath22 for which the jacobian is significantly larger than @xmath67 is small . whenever we observe a point @xmath22 whose jacobian is significantly larger than @xmath67 , we are skeptical about whether it has been chosen independently of @xmath21 . we now describe in which sense igci rejects observations @xmath68 that are untypical . for @xmath69 we observe @xmath70 if the right hand side of ( [ fact ] ) is significantly larger than zero , then @xmath22 is untypical because this holds only for a small fraction of the hypercube . this suggests the following reinterpretation of igci : due to ( [ expec ] ) and ( [ minus ] ) , the right hand side of ( [ empigci ] ) will usually be positive when @xmath32 is the true causal direction . whenever it attains a `` large '' positive value , the expression @xmath71 is also large because the former is an approximation of the latter . then , @xmath72 is untypical for the function @xmath73 , which makes us rejecting the causal hypothesis @xmath74 . we now argue that igci , roughly speaking , amounts to choosing the direction for which there is a larger number of functions that fit the data . we will also discuss some connections to inductive principles of statistical learning theory @xcite . to get a clear sense of `` number of functions '' , we discretize the interval @xmath9 $ ] for @xmath0 and @xmath1 and assume that all @xmath75 are taken from the grid @xmath76 as in figure [ fig : one_point ] c ) . we assume furthermore that @xmath77 and , similarly , @xmath78 and denote these observations by @xmath68 and @xmath72 . let @xmath79 be the set of all monotonic functions for which @xmath80 with @xmath81 . our decision which causal direction is more plausible for the observation @xmath82 will now be based on the following generating models : a function @xmath8 is chosen uniformly at randomly from @xmath83 , i.e. , the set of functions from @xmath0 to @xmath1 that pass the points @xmath52 and @xmath53 . then , each @xmath84 with @xmath85 is chosen uniformly at random from @xmath86 . this yields the following distribution on the set of possible observations : @xmath87 likewise , we obtain a distribution for the causal direction @xmath74 given by @xmath88 where @xmath89 denotes the corresponding set of functions from @xmath1 to @xmath0 . + for a general grid @xmath90 , elementary combinatorics shows that the number of monotonic functions from @xmath0 to @xmath1 that pass the corners @xmath52 and @xmath91 is given by @xmath92 therefore , @xmath93 the pair @xmath82 defines @xmath94 grids @xmath95 and @xmath96 is the product of the numbers for each grid . thus , @xmath97 where we have applied rule ( [ comb ] ) to each grid @xmath98 . combining ( [ genxy ] ) , ( [ genyx ] ) , ( [ comb ] ) , and ( [ combcomb ] ) yields @xmath99 we now consider the limit of arbitrarily fine grid , i.e. , @xmath100 ( while keeping the ratios of all @xmath84 and those of all @xmath101 constant ) . then expression ( [ igcid ] ) becomes independent of the grid and can be replaced with @xmath102 thus , igci as given by definition [ def : igci ] simply compares the loglikelihoods of the data with respect to the two competing generating models above since ( [ discrigci ] ) coincides with the left hand side of ( [ empigci ] ) after normalizing @xmath103 such that @xmath104 . the above link is intriguing , but the function counting argument required that we discretized the space , leading to finite function classes , and it is not obvious how the analysis should be done in the continuous domain . in statistical learning theory @xcite , the core of the theoretical analysis is the following : for consistency of learning , we need uniform convergence of risks over function classes . for finite classes , uniform convergence follows from the standard law of large numbers , but for infinite classes , the theory builds on the idea that whenever these classes are evaluated on finite samples , they get reduced to finite collections of equivalence classes consisting of functions taking the same values on the given sample . in transductive inference as well as in a recently proposed inference principle referred to as inference with the `` universum , '' the size of such equivalence classes plays a central role @xcite . the proposed new view of the igci principle may be linked to this principle . universum inference builds on the availability of additional data that is not from the same distribution as the training data in principle , it might be observed in the future , but we havent seen it yet and it may not make sense for the current classification task . let us call two pattern recognition functions equivalent if they take the same values on the training data . we can measure the size of an equivalence class by how many possible labellings the functions from the class can produce on the universum data . a classifier should then try to correctly separate the training data using a function from a large equivalence class i.e. , a function from a class that allows many possible labellings on the universum data , i.e. , one that does not make a commitment on these points . loosely speaking , the universum is a way to adjust the geometry of the space such that it makes sense for the kind of data that might come up . this is consistent with a paper that linked the universum - svm @xcite to a rescaled version of fisher s discriminant @xcite . taken to its extreme , it would advocate the view that there may not be any natural scaling or embedding of the data , but data points are only meaningful in how they relate to other data points . in our current setting , if we are given a set of universum points in addition to the training set , we use them to provide the discretization of the space . we consider all functions equivalent that interpolate our training points , and then determine the size of the equivalence classes by counting , using the universum points , how many such functions there are . the size of these equivalence classes then determines the causal direction , as described above our analysis works exactly the same no matter whether we have a regular discretization or a discretization by a set of universum points . we now argue that the correlations between @xmath11 and @xmath105 are relevant for prediction in two respects : first , knowing @xmath18 tells us something about @xmath11 , and second , @xmath11 tells us something about @xmath18 . note that section [ sec : untyp ] already describes the first part : assume @xmath106 is large . then , knowing @xmath18 ( and , in addition , a lower bound for @xmath107 ) restricts the set of possible @xmath34-tuples to a region with small volume . we know explore what @xmath11 tells us about @xmath18 . this scenario is the one in unsupervised and semi - supervised learning ( ssl ) @xcite since the distribution of unlabeled points is used to get information about the labels . @xcite hypothesized that this is not possible in causal direction , i.e. , if the labels are the effect . in anticausal direction , the labels are the cause and unsupervised learning employs the information that @xmath108 contains about @xmath109 . as opposed to the standard notation in machine learning , where @xmath1 is the variable to be predicted from @xmath0 , regardless of which of the variables is the cause , we prefer to keep the convention that @xmath0 causes @xmath1 throughout the paper . thus , we consider the task of predicting @xmath0 from @xmath1 and discuss in which sense knowing the distribution @xmath11 helps . we study this question within the finite grid @xmath110 to avoid technical difficulties with defining priors on the set of differentiable functions . we use essentially the generating model @xmath32 from section [ sec : numb ] with monotonic functions on the grid with the following modification : we restrict the set of functions to the set of surjective functions @xmath111 to ensure that the image of the uniform distribution is a strictly positive distribution . to avoid that this is a strong restriction , we assume that @xmath112 . since we use the grid only to motivate ideas for the continuous case , this does not introduce any artificial asymmetry between @xmath0 and @xmath1 . then we assume that a function @xmath8 is drawn uniformly at random from @xmath111 . afterwards , @xmath34 @xmath113-values are drawn uniformly at random from @xmath114 . this generating model defines a joint distribution for @xmath34-tuples @xmath115 and functions @xmath8 via @xmath116 where @xmath117 denotes the application of @xmath8 in each component to @xmath115 yields the same distribution as in section [ sec : numb ] up to the technical modifications of having fixed endpoints and surjective functions . ] . in analogy to the continuous case , we introduce the image of the uniform distribution on @xmath114 under @xmath8 by @xmath118 and obtain @xmath119 hence , @xmath120 where we have used the fact that all functions are equally likely . we rephrase ( [ x > y ] ) as @xmath121 where @xmath11 denotes the distribution of empirical relative frequencies defined by the @xmath34-tuple @xmath122 and @xmath123 is a summand that does not depend on @xmath8 . ( [ predf ] ) provides a prediction of @xmath8 from @xmath122 . we now ask why this prediction should be _ useful _ although it is based on the wrong model because we assume that the true data generating model does not draw @xmath113-values from the uniform distribution ( instead , @xmath10 only `` behaves like the uniform one '' in the sense of ( [ uncorr ] ) ) . to this end , we show that the likelihood of @xmath8 is unusually high compared to other functions that are , in a sense , equivalent . to define a set of equivalent functions , we first represent @xmath8 by the following list of non - negative integers : @xmath124 and observe that this list describes @xmath8 uniquely because @xmath8 is monotonic . then every permutation @xmath125 on @xmath126 defines a new monotonic function @xmath127 by the list @xmath128 with @xmath129 . note that @xmath130 . therefore , one can easily argue that for large @xmath131 , most permutations induce functions @xmath127 for which @xmath132 this is because the difference between left and right hand side can be interpreted as covariance of the random variables @xmath11 and @xmath133 with respect to the uniform distribution on @xmath126 ( see also section [ sec : int ] ) and a random permutation yields approximately uncorrelated samples with high probability and upper bounds on @xmath11 , which goes beyond the scope of this paper . ] . therefore , if we observe that @xmath134 in the sense of significant violation of equality , the true function @xmath8 has a higher likelihood than the overwhelming majority of the functions @xmath127 . in other words , @xmath135 prefers the true function within a huge set of functions that are equivalent in the sense of having the same numbers of pre - images . translating this into the continuous setting , we infer @xmath8 from @xmath11 by defining a loglikelihood function over some appropriate set of sufficiently smooth functions via @xmath136 with a free parameter @xmath137 , since we have explained in which sense this provides a useful prediction in the discrete setting . rather than getting a distribution over the possible functions for @xmath8 we often want to get a single function @xmath138 that predicts @xmath0 from @xmath1 , i.e. , an estimator for @xmath139 . we define @xmath140 and observe that @xmath138 maps @xmath11 to the uniform distribution due to @xmath141 , i.e. , @xmath138 provides the correct prediction if @xmath10 is uniform . moreover , its inverse @xmath142 is the unique maximizer of ( [ flike ] ) since it maps @xmath36 to @xmath11 . to understand in what sense @xmath138 still provides a good prediction even if @xmath10 strongly deviates from @xmath36 , we observe that the error remains small if the cumulative distribution function @xmath143 does not deviate too much from the one for the uniform distribution . furthermore , @xmath138 shares some qualitative behavior with @xmath144 because it tends to have large slope where @xmath144 has large slope because @xmath145 correlates with @xmath146 due to ( [ pc ] ) . figure [ fig : exp ] visualizes unsupervised prediction based on @xmath138 for a simple function . a simple function @xmath8 from @xmath0 to @xmath1 ( left ) and the functions @xmath138 inferred from the empirical distributions of @xmath1 for two different input distributions @xmath10.,scaledwidth=110.0% ] we now argue that information theory provides theoretical results on how close @xmath138 is to @xmath144 . to this end , we define an ( admittedly uncommon ) distance of functions by the relative entropy distance of the densities that they map to the uniform distribution . thus , @xmath147 measures the distance between @xmath138 and @xmath144 . since relative entropy is conserved under bijections @xcite , we have @xmath148 i.e. , the deviation between @xmath138 and @xmath144 coincides with the deviation of @xmath10 from the uniform distribution . together with ( [ orth1 ] ) , ( [ transf ] ) implies @xmath149 with equality only for @xmath150 . note that @xmath10 represents the functions @xmath151 obtained from the analog of ( [ ghat ] ) when trying to infer @xmath8 from @xmath10 ( although we know that this is pointless when @xmath10 and @xmath8 are chosen independently ) . since @xmath44 represents the true function @xmath8 , we conclude : no matter how much @xmath138 deviates from @xmath144 , @xmath151 deviates even more from @xmath8 , i.e. , the error of unsupervised prediction in causal direction always exceeds the one in anticausal direction . for the semi - supervised version , we are given a few labeled points @xmath152 as well as a large number of unlabeled points @xmath153 . we consider again the limit where @xmath34 is infinite and the observations tell us exactly the distribution @xmath11 . then we use the information that @xmath11 provides on @xmath8 for interpolating between the labeled points via @xmath154 whenever @xmath155 $ ] . note that the above schemes for un- and semisupervised prediction are not supposed to compete with existing methods for real - world applications ( the assumption of a noiseless invertible relation does not occur too often anyway ) . the goal of the above ideas is only to present a toy model that shows that the independence between @xmath156 and @xmath157 typically yields a dependence between @xmath108 and @xmath158 that can be employed for prediction . generalizations of these insights to the noisy case could be helpful for practical applications . the authors are grateful to joris mooij for insightful discussions . p. hoyer , d. janzing , j. mooij , j. peters , and b schlkopf . nonlinear causal discovery with additive noise models . in d. koller , d. schuurmans , y. bengio , and l. bottou , editors , _ proceedings of the conference neural information processing systems ( nips ) 2008 _ , vancouver , canada , 2009 . mit press . http://books.nips.cc/papers/files/nips21/nips2008_0266.pdf . j. peters , d. janzing , and b. schlkopf . identifying cause and effect on discrete data using additive noise models . in _ proceedings of the thirteenth international conference on artificial intelligence and statistics ( aistats ) 2010 , jmlr : w&cp 9 , chia laguna , sardinia , italy , 2010_. http://jmlr.csail.mit.edu / proceedings / papers / v9/. j. peters , j. mooij , d. janzing , and b. schlkopf . identifiability of causal graphs using functional models . in _ proceedings of the 27th conference on uncertainty in artificial intelligence ( uai 2011)_. http://uai.sis.pitt.edu/papers/11/p589-peters.pdf . jason weston , ronan collobert , fabian sinz , lon bottou , and vladimir vapnik . inference with the universum . in _ in icml 06 : proceedings of the 23rd international conference on machine learning _ , pages 10091016 . acm , 2006 . sinz , o. chapelle , a. agarwal , and b. schlkopf . an analysis of inference with the universum . in jc platt , d koller , y singer , and s roweis , editors , _ advances in neural information processing systems 20 _ , pages 13691376 , 9 2008 .

information geometric causal inference ( igci ) is a new approach to distinguish between cause and effect for two variables . it is based on an independence assumption between input distribution and causal mechanism that can be phrased in terms of orthogonality in information space . we describe two intuitive reinterpretations of this approach that makes igci more accessible to a broader audience . moreover , we show that the described independence is related to the hypothesis that unsupervised learning and semi - supervised learning only works for predicting the cause from the effect and not vice versa .

intro a remarkable phenomenon quite ubiquitous in nature is that of collective synchronization , in which a large population of coupled oscillators spontaneously synchronizes to oscillate at a common frequency , despite each constituent having a different natural frequency of oscillation @xcite . one witnesses such a spectacular cooperative effect in many physical and biological systems over length and time scales that span several orders of magnitude . some common examples are metabolic synchrony in yeast cell suspensions @xcite , synchronized firings of cardiac pacemaker cells @xcite , flashing in unison by groups of fireflies @xcite , voltage oscillations at a common frequency in an array of current - biased josephson junctions @xcite , phase synchronization in electrical power distribution networks @xcite , rhythmic applause @xcite , animal flocking behavior @xcite ; see @xcite for a survey . the kuramoto model provides a simple theoretical framework to study how synchronization may emerge spontaneously in the dynamics of a many - body interacting system @xcite . the model comprises globally - coupled oscillators of distributed natural frequencies that are interacting via a mean - field coupling through the sine of their phase differences , with the phases following a first - order dynamics in time . over the years , many aspects of the model , including applications cutting across disciplines , from physical and biological to even social modelling , have been considered in the literature @xcite . an early motivation behind studying the kuramoto model was to explain the spectacular phenomenon of spontaneous synchronization among fireflies : in parts of south - east asia , thousands of male fireflies gather in trees at night and flash on and off in unison . in this respect , focussing on fireflies of a particular species ( the _ pteroptyx mallacae _ ) , a study due to ermentrout revealed that the approach to synchronization from an initially unsynchronized state is faster in the kuramoto setting than in reality @xcite . ermentrout proposed a route to reconciliation by elevating the first - order dynamics of the kuramoto model to the level of second - order dynamics . including also a gaussian noise term that accounts for the stochastic fluctuations of the natural frequencies in time @xcite , one arrives at a generalized kuramoto model including inertia and noise , in which oscillator phases have a second - order dynamics in time @xcite . one can prove that the resulting dynamics leads to a nonequilibrium stationary state ( ness ) at long times @xcite . study of nesss is an active area of research of modern day statistical mechanics @xcite . such states are characterized by a violation of detailed balance leading to a net non - zero probability current around a closed loop in the configuration space . one of the primary challenges in this field is to formulate a tractable framework to analyze nonequilibrium systems on a common footing , similar to the one due to gibbs and boltzmann that has been established for equilibrium systems @xcite . in a different context than that of coupled oscillators , the dynamics of the generalized kuramoto model also describes a long - range interacting system of particles moving on a unit circle under the influence of a set of external drive in the form of a quenched external torque acting on the individual particles , in the presence of noise . with the noise , but without the external torques , the resulting model is the so - called brownian mean - field ( bmf ) model @xcite , introduced as a generalization of the celebrated hamiltonian mean - field ( hmf ) model that serves as a prototype to study statics and dynamics of long - range interacting systems @xcite . in recent years , there has been a surge in interest in studies of systems with long - range interactions . in these systems , the inter - particle potential in @xmath0 dimensions decays at large separation @xmath1 as @xmath2 , with @xmath3 @xcite . examples are gravitational systems @xcite , plasmas @xcite , two - dimensional hydrodynamics @xcite , charged and dipolar systems @xcite , etc . unlike systems with short - range interactions , long - range interacting systems are generically non - additive , implying that dividing the system into macroscopic subsystems and summing over their thermodynamic variables such as energy do not yield the corresponding variables of the whole system . non - additivity leads to many significant thermodynamic and dynamical consequences , such as negative microcanonical specific heat , inequivalence of statistical ensembles , and others , which are unusual with short - range interactions @xcite . in this review , starting with the first - order mean - field dynamics of the original kuramoto model , we progressively modify the dynamics by including first the effects of a gaussian noise , and then the consequences of an inertial term that makes the dynamics second order in time . in each case , we discuss the possible transitions to synchrony that the resulting stationary state exhibits . here , we will explicitly consider a unimodal distribution of the natural frequencies . while the derivation of the phase diagram in the original model is based on an insightful self - consistent approach due to kuramoto , inclusion of gaussian noise allows to employ usual tools of statistical mechanics and explicitly study the evolution of the phase space distribution by using a fokker - planck approach . in both these cases , the transition between the unsynchronized and the synchronized phase turns out to be continuous or second order . we conveniently study the dynamics of the generalized kuramoto model that includes the effects of both inertia and noise by introducing a reduced parameter space involving dimensionless moment of inertia , temperature , and width of the frequency distribution . we point out the relation of the model to the bmf model , thereby making references to the literature on long - range interacting systems . we give a rigorous proof that the system at long times settles into a ness unless the width of the frequency distribution is zero when it has an equilibrium stationary state . we highlight that the generalized dynamics exhibits a nonequilibrium first - order transition from a synchronized phase at low parameter values to an unsynchronized phase at high values . as a result , the system as a function of the transition parameters switches over in a discontinuous way from one phase to another , thereby mimicking an abrupt off - on switch . this may be contrasted to the case of no inertia when the transition is continuous . in proper limits , we discuss how one may recover the known continuous phase transitions in the kuramoto model and in its noisy extension , and an equilibrium continuous transition in the bmf model . the present approach offers a complete and consistent picture of the phase diagram , unifying previous results with new ones in a common framework . in the last part of the review , we consider deviations from the mean - field setting of the kuramoto model . to this end , we analyze the generalized kuramoto dynamics on a one - dimensional periodic lattice on the sites of which the oscillators reside and interact with a coupling that decays as an inverse power - law of their separation along the lattice . we consider two specific cases of the dynamics , namely , in the absence of noise and inertia , and in the case when the natural frequencies are the same for all the oscillators ( giving rise to the so - called @xmath4-hmf model ) . for the latter case , we consider both overdamped and underdamped dynamics . in particular , we discuss how the long - time transition to synchrony is governed by the dynamics of the mean - field mode ( zero fourier mode ) of the spatial distribution of the oscillator phases . in this review , besides extensive numerical simulations , aspects of phase diagram are derived analytically by performing a linear stability analysis of the mean - field incoherent stationary state . moreover , for the case of the overdamped dynamics of the generalized kuramoto model on the lattice with the same natural frequency for all the oscillators , we present analytical results also on the linear stability analysis of the mean - field synchronized stationary state . we end the review with conclusions and perspectives . kuramoto - bare we start with a derivation of the dynamics of the kuramoto model by following ref . @xcite . consider first a single landau - stuart oscillator . its dynamics is given in terms of the complex variable @xmath5 as = iq + ( -|q|^2)q , ls with @xmath6 , and additionally , @xmath7 . in ref . @xcite , it is explained that the oscillator represented by eq . ( [ ls ] ) is a simple model for self - organized systems like , e.g. , reacting chemical species . writing @xmath5 in terms of its argument and modulus as @xmath8 with @xmath9 and @xmath10 $ ] , we see from equation ( [ ls ] ) that @xmath11 rotates uniformly in time with angular frequency equal to the parameter @xmath12 , while @xmath13 has a stable value @xmath14 , such that @xmath15 , and @xmath16 . then , setting @xmath17 leads to self - sustained limit - cycle oscillations at frequency @xmath12 with amplitude @xmath18 ; the phase @xmath11 varies as a function of time as = . next , consider a population of @xmath19 interacting landau - stuart oscillators with varying frequencies @xmath20 distributed according to a given probability distribution @xmath21 . the dynamics of the @xmath22th oscillator with angular frequency @xmath23 may be modelled as = i_i q_i + ( -|q_i|^2)q_i + _ j=1,ji^n k_ij q_j , ls - ii where the real parameter @xmath24 describes the coupling between the @xmath22th and the @xmath25th oscillator . in deriving his model while starting from the dynamics ( [ ls - ii ] ) , kuramoto considers three simplifying premises , namely , 1 . the limit of an infinite number of oscillators : @xmath26 , 2 . the coupling @xmath27 scaling as @xmath28 with @xmath29 finite , implying thereby that every oscillator is coupled _ weakly _ and with equal strength to every other oscillator , and 3 . the limit @xmath30 , while keeping @xmath31 fixed and finite , and , moreover , @xmath32 being finite . writing @xmath33 , we see that because of the above assumptions , @xmath34 while starting from an initial value will relax over a time of @xmath35 to its limit - cycle value equal to @xmath36 . as a result , the long - time dynamics corresponds to self - sustained limit - cycle oscillations for each oscillator , which is described by the evolution equation = _ i+_j=1^n ( _ j-_i ) . kuramoto - eom equation ( [ kuramoto - eom ] ) is the governing dynamical equation of the kuramoto model . kuramoto most investigations of the kuramoto model have been for a unimodal @xmath21 , i.e. , one which is symmetric about the mean @xmath37 , and which decreases monotonically and continuously to zero with increasing @xmath38 . we will denote by @xmath39 the width of the distribution @xmath21 ( e.g. , for a gaussian distribution , @xmath39 is the standard deviation ) . as mentioned in the introduction , we will in this review consider specifically such frequency distributions . by going to a comoving frame rotating with frequency @xmath37 with respect to the laboratory frame , one may from now on consider in the dynamics ( [ kuramoto - eom ] ) the @xmath23 s to have zero mean without loss of generality ; we will implement this in the rest of the review . in his early works , kuramoto adduced a remarkable self - consistent analysis to predict the long - time stationary state of the dynamics ( [ kuramoto - eom ] ) . this analysis and its further generalizations have established that the stationary state is characterized by one of two possible phases , depending on whether @xmath29 is below or above a critical value @xmath40 , given by @xcite k_c=. kura - barekc the system for @xmath41 is in an unsynchronized or incoherent phase in which the oscillators exhibit independent oscillations , while for @xmath42 in a synchronized phase in which a macroscopic fraction of oscillators are in synchrony . on tuning @xmath29 , a continuous phase transition occurs between the two phases . more precisely , the kuramoto model being a dynamical system , one does not quite have a phase transition in the sense of thermodynamics , but rather a bifurcation for the order parameter , see below . for noisy dynamics , cases of which will be studied later in the review , a phase transition of course has its usual meaning as in thermodynamics . here , by synchrony , we mean that in the limit @xmath26 , the oscillators have time - independent phases in the comoving frame , and have therefore phases that rotate uniformly in time with the same frequency @xmath37 in the laboratory frame . the magnitude @xmath43 and the phase @xmath44 of the complex order parameter , defined as ( t)=r(t)e^i(t)_j=1^n e^i_j(t ) , r - defn measure the amount of synchronization and the average phase , respectively . for @xmath41 , @xmath43 while starting from any initial value relaxes at long times to zero , corresponding to an incoherent stationary state . on the other hand , for @xmath42 , @xmath43 has a non - zero stationary state value @xmath45 that increases continuously with @xmath29 , and is such that @xmath46 . in the limit @xmath47 , all the oscillators are synchronized and at the same phase , so that @xmath48 . in terms of @xmath43 and @xmath44 , the dynamics ( [ kuramoto - eom ] ) reads = _ i+k r ( -_i ) . kuramoto - eom1 we now briefly recall a self - consistent analysis due to kuramoto @xcite that leads to equation ( [ kura - barekc ] ) . the starting point is to note that in the stationary state , the single - oscillator distribution @xmath49 , giving the fraction of oscillators with natural frequency @xmath12 that has phase @xmath50 at time @xmath51 , converges to the time - independent form @xmath52 . note that @xmath13 satisfies @xmath53 , and the normalization @xmath54 . when synchronized , the average phase in the stationary state , @xmath55 , will be a constant that may be set to zero by choosing properly the origin of the phase axes . the stationary value @xmath56 of @xmath43 , on the other hand , satisfies r_st= g()e^i _ st ( , ) . r - st since @xmath56 is real , the above equation implies that the imaginary part of the right hand side should vanish . as @xmath21 is symmetric : @xmath57 , the vanishing of the imaginary part is ensured if @xmath58 . this is indeed the case , as we will see below . on the basis of the above discussion , one may rewrite equation ( [ kuramoto - eom1 ] ) in the stationary state as = _ r_st _ i. kuramoto - eom2 the result ( [ kura - barekc ] ) follows from a self - consistent equation derived by adopting the following strategy : at a fixed @xmath29 , and for a given value of @xmath56 , ( i ) obtain the stationary state distribution @xmath52 implied by the dynamics ( [ kuramoto - eom2 ] ) , and ( ii ) require that the distribution when plugged into the right hand side of equation ( [ r - st ] ) reproduces the given value of @xmath56 on the left hand side , thereby yielding the self - consistent equation ; we now demonstrate this procedure . first , it follows from equation ( [ kuramoto - eom2 ] ) that the dynamics of oscillators with @xmath59 approaches in time a stable fixed point defined implicitly by _ i = kr_st_i , so that the @xmath22th oscillator in this group has the time - independent phase @xmath60 $ ] ; @xmath61 . this group of oscillators are thus `` locked '' or synchronized , and has the distribution _ st(,)=k r_st ( - k r_st ) ( ) ; || kr_st , kura - distr - locked where @xmath62 is the heaviside step function . on the other hand , oscillators with @xmath63 have ever drifting time - dependent phases . however , to be consistent with the fact that we have a time - independent average phase , it is required that @xmath64 for this group of `` drifting '' oscillators has the form _ st(,)= ; || > kr_st ; kura - distr - drifting this ensures that oscillators are more crowded at @xmath50-values with lower local velocity @xmath65 than at values with higher local velocity . the constant @xmath66 in equation ( [ kura - distr - drifting ] ) is fixed by the normalization condition @xmath67 , yielding c=. we now require that the given value of @xmath56 coincides with the one implied by the distributions in equations ( [ kura - distr - locked ] ) and ( [ kura - distr - drifting ] ) . plugging the latter forms into equation ( [ r - st ] ) , we get r_st&=&_-^_|| > kr_st g ( ) e^i + & + & _ -^ _ || kr_st g ( ) e^i k r_st ( - k r_st ) . the first integral on the right hand side vanishes due to the symmetry @xmath57 combined with the property that @xmath68 for the `` drifting '' oscillators , see equation ( [ kura - distr - drifting ] ) . the imaginary part of the second integral vanishes on using @xmath57 and @xmath58 for the `` locked '' oscillators , see equation ( [ kura - distr - locked ] ) ; the real part , after integration over @xmath12 , finally yields r_st = kr_st_-/2^/2 ^2 g(kr_st ) , which is the desired self - consistent equation . this equation has the trivial solution @xmath69 , valid for any value of @xmath29 , corresponding to the incoherent phase with @xmath70 . there can however be another solution corresponding to @xmath71 that satisfies 1=k_-/2^/2 ^2 g(kr_st ) . kura - bifurcation this solution bifurcates continuously from the incoherent solution at the value @xmath72 given by equation ( [ kura - barekc ] ) that follows from the above equation on taking the limit @xmath73 . since for a unimodal @xmath21 , one has a negative second derivative at @xmath74 , @xmath75 , one finds by expanding the integrand in equation ( [ kura - bifurcation ] ) as a powers series in @xmath56 that the bifurcation in this case is supercritical . as a matter of fact , it is not difficult to see that for a unimodal @xmath21 , a solution of equation ( [ kura - bifurcation ] ) exists only for @xmath76 . indeed , the right hand side of equation ( [ kura - bifurcation ] ) is equal to @xmath77 for @xmath78 , while its partial derivative with respect to @xmath56 , given by k^2_-/2^/2 ^2 g(kr_st ) , kura_r_deriv is negative definite ( here and henceforth , prime will denote derivative ) . on the other hand , for @xmath79 , the right hand side of equation ( [ kura - bifurcation ] ) after the change of variable @xmath80 can be written as _ -k^k u ( 1- ) ^g(u ) , kura_r_eq1 which is clearly smaller than @xmath81 , tending to @xmath81 as @xmath47 . finally , its derivative with respect to @xmath29 is _ -/2^/2 ^2 g(kr_st ) + kr_st _ -/2^/2 ^2 g(kr_st ) + = _ -/2^/2 ^2 g(kr_st ) , kura_k_deriv which is positive . these properties imply that a solution @xmath56 of equation ( [ kura - bifurcation ] ) exists for @xmath82 , which equals @xmath83 for @xmath84 , and which increases with @xmath29 and approaches unity as @xmath85 . the linear stability of the incoherent solution , @xmath86 @xcite , will be considered in sec . [ seckurlongrange ] , where it will appear as a special case of the kuramoto model with non - mean - field long - range interactions . the stability analysis will establish that the incoherent state is neutrally stable below @xmath40 and unstable above . noisykuramoto in order to account for stochastic fluctuations of the @xmath23 s in time , the dynamics ( [ kuramoto - eom ] ) with an additional gaussian noise term @xmath87 on the right hand side was studied by sakaguchi @xcite . the dynamical equations are = _ i+k r ( -_i)+_i(t ) , kura - noise - eom where _ i(t ) = 0 , _ i(t)_j(t ) = 2d_ij(t - t ) , with the parameter @xmath88 standing for the noise strength , while here and from now on , angular brackets will denote averaging with respect to noise realizations . in presence of @xmath87 , the continuous transition of the bare model is sustained , with @xmath40 shifted to @xcite k_c(d ) = 2 ^-1 . kura - noise - kc on taking the limit @xmath89 in the above equation , one recovers the transition point ( [ kura - barekc ] ) for the bare model . in the following , we briefly sketch the derivation of equation ( [ kura - noise - kc ] ) , following ref . @xcite . the starting point is to write down a fokker - planck equation for the time evolution of the distribution @xmath49 , which follows straightforwardly from the dynamics ( [ kura - noise - eom ] ) as = -+d . fokplaeqnoisy as before , in the stationary state , we set @xmath90 , and obtain from the above equation the result _ st(,)&=&()_st(0 , ) + & & , kura - noise - distr where @xmath91 is fixed by the normalization @xmath92 . for @xmath69 the above expression reduces to the incoherent state @xmath86 . substituting equation ( [ kura - noise - distr ] ) into equation ( [ r - st ] ) , one obtains a self - consistent equation for @xmath56 . as for the kuramoto model , it has the trivial solution @xmath93 , corresponding to the incoherent state . in finding the other solution , one observes that the imaginary part of the right hand side of equation ( [ r - st ] ) is zero due to the symmetry @xmath57 together with the property @xmath94 , see equation ( [ kura - noise - distr ] ) ; thus , only the real part contributes . expanding the resulting equation in powers of @xmath95 , and taking the limit @xmath73 @xcite yield the critical coupling strength @xmath96 given by equation ( [ kura - noise - kc ] ) . linear - stability - kura - noise the stability analysis of the incoherent state @xmath86 is performed by studying the linearized fokker - planck equation obtained from equation ( [ fokplaeqnoisy ] ) after expanding @xmath97 as ( , , t ) = + ( , , t ) ; || 1 . linearnoisy writing explicitly the expression for @xmath1 , the resulting linear equation is fokplalinearnoisy ( , , t ) = - ( , , t ) + d ( , , t ) + + _ -^ g( ) ( - ) ( ,,t ) . with the fourier expansion fouriernoisy ( , , t ) = _ k=-^+ _ k(,t)e^ik , equation ( [ fokplalinearnoisy ] ) gives fokplalinearnoisyfourier _ k(,t ) = -ik_k(,t ) - d k^2 _ k(,t ) + ( _ k,1 + _ k,-1 ) g()_k(,t ) . for @xmath98 the integral term vanishes , and we have fokplalinearnoisyfourierkne1 _ k(,t ) = -ik_k(,t ) - d k^2 _ k(,t ) , so that with @xmath12 varying in the support of @xmath21 , one has a continuous spectrum of stable modes that decay exponentially in time with rate @xmath99 . for @xmath100 , after posing expokeq1 _ 1(,t ) = _ 1(,)e^t , we have fokplalinearnoisyfourierk1 _ 1 ( , ) = g()_1( , ) . this equation also admits a continuous spectrum of stable modes , given by @xmath101 for each @xmath102 in the support of @xmath21 . the modes , normalized so that the right hand side of equation ( [ fokplalinearnoisyfourierk1 ] ) is equal to 1 , are given by _ 1(,i _ 0 -d ) = i p + c_1(_0 ) ( - _ 0 ) , eigen_cont_pm1_noisy with c_1(_0 ) g(_0 ) = i p , eigen_factor_noisy where @xmath103 denotes the principal value . however , unlike for @xmath104 , there is also a discrete spectrum for @xmath105 . from equation ( [ fokplalinearnoisyfourierk1 ] ) , we have _ 1 ( , ) = _ 1( , ) g( ) . fokplalinearnoisyfourierk1_discr in order to have a non - trivial solution of the above equation , the integral on the right hand side must not vanish . we can impose that this integral is equal to @xmath81 , since equation ( [ fokplalinearnoisyfourierk1 ] ) is linear . we then obtain the dispersion relation _ disper_relat_noisy decomposing @xmath106 into real and imaginary parts , @xmath107 , we obtain from equation ( [ disper_relat_noisy ] ) that _ -^+ g ( ) = 1 , disper_relatr_noisy + _ -^+ g ( ) = 0 . disper_relati_noisy with the change of variable @xmath108 , the integral in the second equation can be transformed to _ 0^+ x . disper_relati_noisy_b one may check that for unimodal @xmath21 , the above expression can be equal to @xmath83 only for @xmath109 . so equation ( [ disper_relatr_noisy ] ) becomes _ -^+ g ( ) = 1 , disper_relatr_noisy_b with @xmath106 real . this equation shows that only solutions @xmath110 are possible ; when such a solution is not present , there is no discrete spectrum , and the incoherent state is stable . however , stability holds also when there is a solution @xmath111 , since we have seen that all the eigenvalues of the continuous spectrum have a negative real part . the change of variable @xmath112 transforms equation ( [ disper_relatr_noisy_b ] ) to _ -^+ y g = 1 . disper_relatr_noisy_c the left hand side tends to @xmath83 as @xmath113 , while its derivative with respect to @xmath106 is _ -^+ y g , disper_relatr_noisy_der which is negative . therefore , a solution for @xmath106 exists only when the value of the left hand side of equation ( [ disper_relatr_noisy_c ] ) for @xmath114 is larger than @xmath81 . in particular , we have stability when this solution is negative ; the threshold for stability is thus given by _ y g(dy ) = 1 , disper_relatr_noisy_thres that gives the critical value ( [ kura - noise - kc ] ) . chap2 in the generalized dynamics , an additional dynamical variable , namely , angular velocity , is assigned to each oscillator , thereby elevating the first - order dynamics of the kuramoto model to the level of second - order dynamics ; the equations of motion are @xcite : = v_i , + eom + m =- v_i+ r(-_i)+_i+_i(t ) . here , @xmath115 is the angular velocity of the @xmath22th oscillator , @xmath116 is the moment of inertia of the oscillators , @xmath117 is the friction constant , @xmath118 is the strength of the coupling between the oscillators , while @xmath119 is a gaussian noise with _ i(t ) = 0 , _ i(t)_j(t ) = 2_ij(t - t ) . etatide - prop in the limit of overdamped motion ( @xmath120 at a fixed @xmath121 ) , the dynamics ( [ eom ] ) reduces to = r(-_i)+_i+_i(t ) . eom - overdamped1 then , defining @xmath122 and @xmath123 so that @xmath124 , the dynamics ( [ eom - overdamped1 ] ) for @xmath125 becomes that of the kuramoto model , equation ( [ kuramoto - eom ] ) , and for @xmath126 that of its noisy version , the dynamics ( [ kura - noise - eom ] ) . in appendix a , we illustrate how the dynamics ( [ eom ] ) without the noise term , studied in @xcite , arises in a completely different context , namely , in electrical power distribution networks comprising synchronous generators ( representing power plants ) and motors ( representing customers ) @xcite ; the dynamics arises in the approximation in which every node of the network is connected to every other . longrangemodel we now discuss that in a different context than that of coupled oscillators , the dynamics ( [ eom ] ) describes a long - range interacting system of particles moving on a unit circle , with each particle acted upon by a quenched external torque @xmath127 . much recent exploration of the static and dynamic properties of long - range interacting systems has been pursued within the framework of an analytically tractable prototypical model called the hamiltonian mean - field ( hmf ) model @xcite . the model comprises @xmath19 particles of mass @xmath116 moving on a unit circle and interacting through a long - range interparticle potential that is of the mean - field type : every particle is coupled to every other with equal strength . this system can also be seen as a set of @xmath128-rotators that reside on a lattice and interact through ferromagnetic coupling . the structure and dimensionality of the lattice need not be specified , since the coupling between each pair of rotators is the same ( mean - field system ) . since the configuration of an @xmath128-rotator is defined by a single angle variable , one might also view the rotators as spin vectors . however , one should be aware that this identification is not completely correct . this is because the dynamics of spins is defined differently , through poisson brackets , or , equivalently , through the derivative of the hamiltonian with respect to the spins , that yields the effective magnetic field acting on the individual spins . on the other hand , the @xmath128-rotators are more correctly identified with particles confined to a circle , with angle and angular momentum as canonically conjugate variables , and with the dynamics generated by the hamilton equations for these variables @xcite . apart from academic interest , the model provides a tractable reference to study physical systems like gravitational sheet models @xcite and the free - electron laser @xcite . the hamiltonian of the hmf model is @xcite h=_i=1^n+_i , j=1^n , hmf - h where @xmath129 $ ] gives the position of the @xmath22th particle on the circle , while @xmath130 is its conjugated angular momentum , with @xmath115 being the angular velocity . the time evolution of the system within a microcanonical ensemble follows the deterministic hamilton equations of motion given by & & = v_i , + hameq + & & m= r(-_i ) . here , the quantities @xmath1 and @xmath131 are as defined in equation ( [ r - defn ] ) . the dynamics conserves the total energy and momentum . under the evolution ( [ hameq ] ) , the system at long times settles into an equilibrium stationary state in which , depending on the energy density @xmath132 , the system could be in one of two possible phases . namely , for @xmath133 smaller than a critical value @xmath134 , the system is in a clustered phase in which the particles are close together on the circle , while for energies larger than @xmath135 , the particles are uniformly distributed on the circle , characterizing a homogeneous phase @xcite . a continuous phase transition between the two phases is effectively characterized by the quantity @xmath43 defined in equation ( [ r - defn ] ) , that in the present context may be interpreted as the specific magnetization of the system . the phase transition may be interpreted as one from a high - energy paramagnetic phase ( similar to the incoherent phase in the setting of coupled oscillators ) to a low - energy ferromagnetic phase ( similar to the synchronized phase ) . one may generalize the microcanonical dynamics ( [ hameq ] ) to include the effect of an interaction with an external heat bath at temperature @xmath136 . the resulting brownian mean - field ( bmf ) model has thus a canonical ensemble dynamics given by @xcite . & & = v_i , + bmfeq + & & m =- v_i+ r(-_i)+_i(t ) , where @xmath119 is defined in equation ( [ etatide - prop ] ) . a fluctuation - dissipation relation expresses the strength @xmath137 of the noise in terms of the temperature @xmath136 and the friction constant @xmath117 as = k_bt . d - relation we will set the boltzmann constant @xmath138 to unity in the rest of the paper . the canonical dynamics ( [ bmfeq ] ) also leads to a long - time equilibrium stationary state in which a generic configuration @xmath66 with energy @xmath139 occurs with the usual gibbs - boltzmann weight : @xmath140 . the phase transition in the hmf model within the microcanonical ensemble occurs within the canonical ensemble on tuning the temperature across the critical value @xmath141 . this latter critical value may be derived very simply by following standard procedure @xcite . due to full rotational @xmath142 symmetry , we may take the direction along which particles cluster to be along @xmath143 without loss of generality . then , the @xmath143-component of the magnetization @xmath1 , namely , @xmath144 , satisfies in equilibrium the equation r_x= . rx - hmf close to the critical point ( @xmath145 ) , expanding the above equation to leading order in @xmath146 , we get r_x(2-_-^ ^2 ) = 0 . with @xmath147 , we get the critical temperature as @xmath141 . let us now envisage the following situation : a set of quenched external torques @xmath148 acts on each of the particles , thereby pumping energy into the system . in this case , the second equation in the canonical dynamics ( [ bmfeq ] ) has an additional term @xmath149 on the right hand side . the resulting dynamics becomes exactly identical to the dynamics ( [ eom ] ) of the generalized kuramoto model . previous introducing inertia alone into the kuramoto dynamics ( equation ( [ eom ] ) without the noise term ) has significant consequences . et al . _ showed , mainly on the basis of numerical simulations , that finite large inertia leads to the synchronization transition becoming of first order , occurring in an abrupt way on tuning @xmath29 @xcite . analysis of the dynamics ( [ eom ] ) in the continuum limit , based on a suitable fokker - planck - like equation , was pursued in refs . it was shown that for a lorentzian @xmath21 , either larger inertia or larger frequency spread ( measured in terms of the width of the lorentzian @xmath21 ) makes the system harder to synchronize , leading to an incoherent stationary state . reducedspace we start our analysis of the dynamics ( [ eom ] ) by noting that the effect of @xmath39 may be made explicit by replacing @xmath23 in the second equation by @xmath150 . we thus consider from now on the dynamics ( [ eom ] ) with the substitution @xmath151 . in the resulting model , @xmath21 therefore has zero mean and unit width . also , we will consider in the dynamics ( [ eom ] ) the parameter @xmath137 to have the scaling ( [ d - relation ] ) . for @xmath152 , using dimensionless quantities dmsless1 & & t , + dmsless2 & & _ iv_i , + dmsless3 & & 1// , + dmsless4 & & / , + dmsless5 & & t/ , + dmsless6 & & _ i()_i(t)/ , the equations of motion become = _ i , + eom - scaled + = -_i+r(-_i)+_i+_i ( ) , where _ i()_j( ) = 2(/)_ij(- ) . for @xmath153 , using dimensionless time t(/ ) , and with @xmath154 and @xmath155 defined as above , the dynamics becomes the overdamped motion & & = r(-_i)+_i+_i ( ) , eom - overdamped where _ i()_j( ) = 2_ij(- ) . we have thus reduced the dynamics ( [ eom ] ) involving five parameters , @xmath156 , to the dynamics ( [ eom - scaled ] ) ( or ( [ eom - overdamped ] ) in the overdamped limit ) that involves only three dimensionless parameters , @xmath157 . from now on , we consider the dynamics in this reduced parameter space , dropping overbars for simplicity of notation . with @xmath158 ( i.e. @xmath159 @xcite,@xcite ) , the dynamics ( [ eom - scaled ] ) is that of the bmf model with an equilibrium stationary state . for other @xmath21 , the dynamics ( [ eom - scaled ] ) violates detailed balance , leading to a nonequilibrium stationary state ( ness ) @xcite . in section [ detailedbalance ] , we give a rigorous proof of this statement . phasetransition in this section , we report on a very interesting non - equilibrium phase transition that occurs in the stationary state of the dynamics ( [ eom - scaled ] ) . as discussed above , the three relevant parameters of the dynamics are @xmath160 . in this three - dimensional space of parameters , let us first locate the phase transitions in the kuramoto model and in its noisy extension , discussed in section [ kuramoto - bare ] . * the phase transition of the kuramoto dynamics ( @xmath161 , @xmath162 ) corresponds now to a continuous transition from a low-@xmath39 synchronized to a high-@xmath39 incoherent phase across the critical point @xmath163 . * extending the kuramoto dynamics to @xmath164 , the above mentioned critical point becomes a second - order critical line on the @xmath165-plane , given by solving @xmath166 $ ] . * the transition in the bmf dynamics ( @xmath167 ) corresponds now to a continuous transition occurring at the critical temperature @xmath168 . ) in terms of dimensionless moment of inertia @xmath116 , temperature @xmath136 , and width of the frequency distribution @xmath39 . here , the shaded blue surface is a first - order transition surface , while the thick red lines are second - order critical lines . the system is synchronized inside the region bounded by the surface , and is incoherent outside . the transitions of known models are also marked in the figure . the blue surface in ( a ) is bounded from above and below by the dynamical stability thresholds @xmath169 and @xmath170 of respectively the synchronized and the incoherent phase , which are estimated in @xmath19-body simulations from hysteresis plots ( see fig . [ fig : hys - mvary ] for an example ) ; the surfaces @xmath169 and @xmath170 for @xmath171 in the case of a gaussian @xmath21 with zero mean and unit width are shown in panel ( b).,title="fig:",scaledwidth=60.0% ] ) in terms of dimensionless moment of inertia @xmath116 , temperature @xmath136 , and width of the frequency distribution @xmath39 . here , the shaded blue surface is a first - order transition surface , while the thick red lines are second - order critical lines . the system is synchronized inside the region bounded by the surface , and is incoherent outside . the transitions of known models are also marked in the figure . the blue surface in ( a ) is bounded from above and below by the dynamical stability thresholds @xmath169 and @xmath170 of respectively the synchronized and the incoherent phase , which are estimated in @xmath19-body simulations from hysteresis plots ( see fig . [ fig : hys - mvary ] for an example ) ; the surfaces @xmath169 and @xmath170 for @xmath171 in the case of a gaussian @xmath21 with zero mean and unit width are shown in panel ( b).,title="fig:",width=377 ] ) , the figure shows @xmath1 vs. adiabatically tuned @xmath39 for different @xmath116 values at @xmath172 , showing also the stability thresholds , @xmath170 and @xmath169 , for @xmath173 . the data are obtained from simulations with @xmath171 . for a given @xmath116 , the branch of the plot to the right ( left ) corresponds to @xmath39 increasing ( decreasing ) ; for @xmath174 , the two branches almost overlap . the data are for the gaussian @xmath21 given by equation ( [ gomega - gaussian]).,width=377 ] fig : hys - mvary ) , the figure shows @xmath1 vs. adiabatically tuned @xmath39 at @xmath175 . the data are obtained from simulations with @xmath171 . the branch of the plot to the right ( left ) corresponds to @xmath39 increasing ( decreasing ) . the data are for a lorentzian @xmath21 with zero mean and unit width . , width=377 ] fig : hys - lor ) , the figure shows @xmath1 vs. adiabatically tuned @xmath39 for different temperatures @xmath176 at a fixed moment of inertia @xmath177 . the data are obtained from simulations with @xmath171 . for a given @xmath136 , the branch of the plot to the right ( left ) corresponds to @xmath39 increasing ( decreasing ) ; for @xmath178 , the two branches almost overlap . the data are for the gaussian @xmath21 given by equation ( [ gomega - gaussian]).,width=377 ] fig : hys - tvary ) at @xmath179 , and the gaussian @xmath21 given by equation ( [ gomega - gaussian ] ) , ( a ) shows at @xmath180 , the numerically estimated first - order phase transition point , @xmath1 vs. time in the stationary state , while ( b ) shows the distribution @xmath181 at several @xmath39 s around @xmath182 . the data are obtained from simulations with @xmath183.,title="fig:",width=377 ] + ) at @xmath179 , and the gaussian @xmath21 given by equation ( [ gomega - gaussian ] ) , ( a ) shows at @xmath180 , the numerically estimated first - order phase transition point , @xmath1 vs. time in the stationary state , while ( b ) shows the distribution @xmath181 at several @xmath39 s around @xmath182 . the data are obtained from simulations with @xmath183.,title="fig:",width=377 ] fig : r - vs - t - pr the complete phase diagram is shown schematically in fig . [ fig : phdiag](a ) , where the thick red second - order critical lines denote the continuous transitions mentioned above . on the other hand , for @xmath184 all non - zero , we demonstrate below that the synchronization transition becomes first order , occurring across the shaded blue transition surface . this surface is bounded by the second - order critical lines on the @xmath165 and @xmath185 planes , and by a first - order transition line on the @xmath186-plane . let us remark that all phase transitions for @xmath187 are in nesss , and are interpreted to be of dynamical origin , accounted for by stability considerations of stationary solutions of equations ( for example , the kramers equation discussed below ) for temporal evolution of phase space distributions . more rigorously , to qualify as thermodynamics phases , one needs to show that the different phases extremize a free energy - like quantity ( e.g. , a large deviation functional @xcite ) . such a demonstration in this nonequilibrium scenario is a daunting task , while for @xmath158 , the phases have actually been shown to minimize the equilibrium free energy @xcite . in order to demonstrate the first - order nature of the transition , we performed @xmath19-body simulations for a representative @xmath21 , i.e. , the gaussian given by equation ( [ gomega - gaussian ] ) . for given @xmath116 and @xmath136 , we prepared an initial state with all oscillators at @xmath188 and frequencies @xmath115 s sampled from a gaussian distribution with zero mean and standard deviation @xmath189 . we then let the system equilibrate at @xmath158 , and subsequently increase @xmath39 adiabatically to high values and back in a cycle . the simulations involved integrations of the @xmath190 coupled equations of motion ( [ eom - scaled ] ) , see appendix b for details . figure [ fig : hys - mvary ] shows the behavior of @xmath1 for several @xmath116 s at a fixed @xmath136 less than the bmf transition point @xmath168 , where one may observe sharp jumps and hysteresis behavior expected of a first - order transition . with decrease of @xmath116 , the jump in @xmath1 becomes less sharp , and the hysteresis loop area decreases , both features being consistent with the transition becoming second - order - like as @xmath120 , see fig . [ fig : phdiag](a ) . for @xmath173 , we mark in fig . [ fig : hys - mvary ] the approximate stability thresholds for the incoherent and the synchronized phase , denoted respectively by @xmath170 and @xmath169 . the actual phase transition point @xmath191 lies in between the two thresholds . let us note from the figure that both the thresholds decrease and approach zero with the increase of @xmath116 . a qualitatively similar behavior is observed for a lorentzian @xmath21 , see fig . [ fig : hys - lor ] . figure [ fig : hys - tvary ] shows hysteresis plots for a gaussian @xmath21 at a fixed @xmath116 and for several values of @xmath192 , where one observes that with @xmath136 approaching @xmath193 , the hysteresis loop area decreases , jumps in @xmath1 become less sharp and occur between smaller and smaller values that approach zero . moreover , the @xmath1 value at @xmath158 decreases as @xmath136 increases towards @xmath193 , reaching zero at @xmath193 . disappearance of the hysteresis loop with increase of @xmath136 similar to that in fig . [ fig : hys - tvary ] was reported in ref . our findings suggest that the thresholds @xmath170 and @xmath169 coincide on the second - order critical lines , as expected , and moreover , they asymptotically come close together and approach zero as @xmath194 at a fixed @xmath136 . for given @xmath116 and @xmath136 , and for @xmath39 in between @xmath195 and @xmath169 , fig . [ fig : r - vs - t - pr](a ) for @xmath1 as a function of time in the stationary state shows bistability , whereby the system switches back and forth between incoherent ( @xmath196 ) and synchronized ( @xmath197 ) states . the distribution @xmath181 depicted in figure [ fig : r - vs - t - pr](b ) is bimodal with a peak around either @xmath196 or @xmath198 as @xmath39 varies between @xmath195 and @xmath169 . figure [ fig : r - vs - t - pr ] lends further credence to the phase transition being first order . analysis we now turn to an analytical characterization of the dynamics ( [ eom - scaled ] ) in the continuum limit @xmath26 . to this end , we define the single - oscillator distribution @xmath199 that gives at time @xmath51 and for each @xmath12 the fraction of oscillators with phase @xmath50 and angular velocity @xmath200 . the distribution is @xmath201-periodic in @xmath11 , and obeys the normalization _ -^ _ -^ v f(,v,,t)=1 , while evolving following the kramers equation @xcite = -v+(--r(-))f+ , kramers where re^i = v g()e^if(,v,,t ) . let us briefly sketch the derivation of equation ( [ kramers ] ) , while the details may be found in ref . we will along the way also indicate how one may prove rigorously that the dynamics ( [ eom - scaled ] ) does not satisfy detailed balance unless @xmath158 . for simplicity of presentation , we first consider the case of a discrete bimodal @xmath21 , and then in the end extend our discussion to a general @xmath21 . then , consider a given realization of @xmath21 in which there are @xmath202 oscillators with frequencies @xmath203 and @xmath204 oscillators with frequencies @xmath205 , where @xmath206 . let us then define the @xmath19-oscillator distribution function @xmath207 as the probability density at time @xmath51 to observe the system around the values @xmath208 . in the following , we use the shorthand notations @xmath209 and @xmath210 . note that @xmath211 satisfies the normalization ( _ i=1^nz_i)f_n(,t)=1 . the distribution @xmath212 evolves in time according to the following fokker - planck equation that may be derived straightforwardly from the equations of motion ( [ eom - scaled ] ) : @xmath213-\sigma\sum_{j=1}^{n}\big(\omega^{t}\big)_{j}\frac{\partial f_{n}}{\partial v_{j}}+\frac{t}{\sqrt{m}}\sum_{i=1}^{n}\frac{\partial^{2}f_{n}}{\partial v_{i}^{2}}\nonumber \\ & -&\frac{1}{2n}\sum_{i , j=1}^{n}\sin(\theta_{j}-\theta_{i})\big[\frac{\partial f_{n}}{\partial v_{i}}-\frac{\partial f_{n}}{\partial v_{j}}\big ] , \label{eq : fp - eqn}\end{aligned}\ ] ] where the @xmath214 column vector @xmath215 has its first @xmath202 entries equal to @xmath203 and the following @xmath204 entries equal to @xmath205 , and where the superscript @xmath136 denotes matrix transpose operation : ^t . detailedbalance let us rewrite the fokker - planck equation ( [ eq : fp - eqn ] ) as @xmath216}{\partial x_{i}}+\frac{1}{2}\sum_{i , j=1}^{2n}\frac{\partial^{2}[b_{i , j}(\mathbf{x})f_{n}(\mathbf{x})]}{\partial x_{i}\partial x_{j}},\label{eq : fp - compact}\end{aligned}\ ] ] where @xmath217 and @xmath218 here , the drift vector @xmath219 is given by @xmath220 while the diffusion matrix @xmath221 is @xmath222 the dynamics described by the fokker - planck equation of the form ( [ eq : fp - compact ] ) satisfies detailed balance if and only if the following conditions are satisfied @xcite : @xmath223}{\partial x_{j}},\label{eq : detailed - balance - cond2}\end{aligned}\ ] ] where @xmath224 is the stationary solution of equation ( [ eq : fp - compact ] ) . here , @xmath225 denotes the parity with respect to time reversal of the variables @xmath226 s : under time reversal , we have @xmath227 , where @xmath228 ( respectively , @xmath229 ) depending on whether @xmath226 is odd ( respectively , even ) under time reversal . for example , @xmath230 s are even , while @xmath231 s are odd . using equation ( [ eq : bij - defn ] ) , the condition ( [ eq : detailed - balance - cond1 ] ) is trivially satisfied , while to check the condition given by ( [ eq : detailed - balance - cond2 ] ) , we formally solve this equation for @xmath224 and check if the solution solves equation ( [ eq : fp - compact ] ) in the stationary state . from equation ( [ eq : detailed - balance - cond2 ] ) , we see that for @xmath232 , the condition reduces to @xmath233 the above equation , using equation ( [ eq : ai - defn ] ) , is obviously satisfied . for @xmath234 , we have @xmath235 solving which we get @xmath236 , \label{eq : stationary - soln1}\end{aligned}\ ] ] where @xmath237 is a function to be determined . substituting the distribution ( [ eq : stationary - soln1 ] ) into equation ( [ eq : fp - compact ] ) and requiring that it is a stationary solution implies that @xmath39 has to be equal to zero , while d(_1,_2, ,_n)=(-_i , j=1^n ) . thus , for @xmath158 , when the dynamics reduces to that of the bmf model , we get the stationary solution as @xmath238.\label{eq : bmf - soln}\ ] ] where @xmath239 is the hamiltonian ( [ hmf - h ] ) ( expressed in terms of dimensionless variables introduced above ) . the lack of detailed balance for @xmath187 obviously extends to any distribution @xmath21 . kramers the starting point is to define the reduced distribution function @xmath240 , with @xmath241 and @xmath242 as @xcite @xmath243 note that the following normalizations hold for the single - oscillator distribution functions : z_1 f_1,0(z_1,t)=1@xmath244z_n_1 + 1 f_0,1(z_n_1 + 1,t)=1 . assuming that 1 . @xmath211 is symmetric with respect to permutations of dynamical variables within the same group of oscillators , and 2 . @xmath212 , together with the derivatives @xmath245 , vanish on the boundaries of the phase space , and then using equation ( [ eq : fp - eqn ] ) in equation ( [ eq : fs - defn ] ) , one obtains the bogoliubov - born - green - kirkwood - yvon ( bbgky ) hierarchy equations for the dynamics ( [ eom - scaled ] ) ( for details , see @xcite ) . in particular , the first equations of the hierarchy are @xmath246 and a similar equation for @xmath247 . in the limit @xmath248 , writing @xmath249,\ ] ] one can express equation ( [ eq : fp-1 ] ) in terms of @xmath21 . in order to generalize the above treatment to the case of a continuous @xmath21 , note for this case that the single - oscillator distribution function is @xmath250 . the first equation of the hierarchy is then @xmath251 in the continuum limit @xmath252 , one may neglect oscillator - oscillator correlations , and approximate @xmath253 as f(,v,,v,,,t ) & = & f(,v,,t)f(,v,,t ) + & + & corrections subdominant in n , so that equation ( [ eq : final - eqn ] ) reduces to the kramers equation ( [ kramers ] ) . solutionkramers the stationary solutions of equation ( [ kramers ] ) are obtained by setting the left hand side to zero . for @xmath158 , the stationary solution is f_st(,v ) , that corresponds to canonical equilibrium , with @xmath56 determined self - consistently @xcite , see equation ( [ rx - hmf ] ) . for @xmath187 , the incoherent stationary state is @xcite f^inc_st(,v,)=1/((2)^3/2 ) . inc - state ) , here we show the marginal distributions , @xmath254 and @xmath255 , corresponding to the incoherent phase for @xmath256 . the points denoting simulation data are for @xmath257 for one fixed realization of the @xmath23 s sampled from the gaussian distribution ( [ gomega - gaussian ] ) , while the continuous lines denote theoretical results ( [ marginalv ] ) and ( [ marginalth ] ) . , title="fig:",width=377 ] ) , here we show the marginal distributions , @xmath254 and @xmath255 , corresponding to the incoherent phase for @xmath256 . the points denoting simulation data are for @xmath257 for one fixed realization of the @xmath23 s sampled from the gaussian distribution ( [ gomega - gaussian ] ) , while the continuous lines denote theoretical results ( [ marginalv ] ) and ( [ marginalth ] ) . , title="fig:",width=377 ] in the class of unimodal frequency distributions , let us consider a representative @xmath21 , namely , a gaussian : g()= . gomega - gaussian we then have for the marginal angular velocity distribution p^inc_st(v)&=&_-^ g ( ) _ -^ f^inc_st(,v , ) + & = & , marginalv and the marginal angle distribution p^inc_st()&=&_-^ g ( ) _ -^ v f^inc_st(,v,)= , marginalth both correctly normalized to unity . in figs . [ marginal - thv - inc ] and [ marginal - thv - syn ] , we compare our theoretical predictions , ( [ marginalv ] ) and ( [ marginalth ] ) , with numerical simulation results . ) , the figure shows the marginal distributions , @xmath258 and @xmath259 , corresponding to the synchronized phase for @xmath260 . the points denoting simulation data are for @xmath257 for one fixed realization of the @xmath23 s sampled from the gaussian distribution ( [ gomega - gaussian]).,title="fig:",width=377 ] ) , the figure shows the marginal distributions , @xmath258 and @xmath259 , corresponding to the synchronized phase for @xmath260 . the points denoting simulation data are for @xmath257 for one fixed realization of the @xmath23 s sampled from the gaussian distribution ( [ gomega - gaussian]).,title="fig:",width=377 ] marginal - thv - syn the existence of the synchronized stationary state is borne out by our simulation results in fig . [ marginal - thv - syn ] , although its analytical form is not known . linearstability - incoherent - inertia we now discuss about the linear stability analysis of the incoherent state ( [ inc - state ] ) ; a similar analysis for the bmf model is discussed in ref . following ref . @xcite , we linearize equation ( [ kramers ] ) about the state by expanding @xmath261 as f(,v,,t)=f^inc_st(,v,)+e^tf(,v , ) , expansion_linear where @xmath262 satisfies the linearized kramers equation : @xmath263 since @xmath261 and @xmath264 are normalized , we have _ -^_-^ v f(,v,)=0 . norm substituting f(,v , ) = _ n=-^ b_n(v,,)e^i n try in equation ( [ linear - kramers ] ) , one gets & & + ( v-)+(1 - -inv)b_n + & & = ( i _ n,1 - i _ n,-1 ) 1,b_n , b - eqn where one has the scalar product @xmath265 with @xmath266 denoting complex conjugation . since @xmath267 is real , one has @xmath268 , while equation ( [ norm ] ) implies that @xmath269 . we can then restrict to consider only @xmath270 . next , equation ( [ b - eqn ] ) is transformed into a nonhomogeneous parabolic cylinder equation by the transformations & & b_n(v , , ) = _ n ( z , , ) , trans + & & z = ( v- + 2nti ) , which when substituted into equation ( [ b - eqn ] ) yield @xmath271\ , \beta_n\nonumber\\ = i\pi \sqrt{m}\fr{\partial f_{\rm st}^{\rm inc}}{\partial v } e^{{1\over 4}(z-2i\sqrt{mt } ) ^{2}}\ , \langle 1 , e^{-{1\over 4}(z-2i\sqrt{mt } ) ^{2 } } \beta_1 \rangle\ , \delta_{n,1}. \end{aligned}\ ] ] for @xmath272 , the right hand side of the above equation is zero , yielding the eigenvalues @xmath273 and the corresponding eigenfunctions @xmath274 that do not depend on @xmath275 and @xmath12 ; here , @xmath276 and @xmath277 are respectively the parabolic cylinder function and the hermite polynomial of degree @xmath278 @xcite . the eigenvalues @xmath279 form a continuous spectrum . all of them have negative real parts , thus leading to linear stability of the incoherent state ( [ inc - state ] ) , for @xmath280 and @xmath281 . for @xmath282 the eigenvalues have also negative real parts unless those with @xmath283 , that have a vanishing real part . they would correspond to neutrally stable modes ; however , the modes with @xmath282 have zero amplitude due to the normalization condition ( [ norm ] ) . for @xmath284 , solving ( [ b - eqn ] ) 1(z,,)&=&-i1,e^- ( - i)^2 _ 1 + & & _ p=0^ d_p(z ) , where @xmath285 ' & = & \left . \fr{\partial f_{\rm st}^{\rm inc}}{\partial v } \right|_{v= \sigma\omega\sqrt{m } - i2t\sqrt{m } + \sqrt{t } z}= - { ( z - 2i\sqrt{mt})\over ( 2\pi)^{{3\over 2}}t}e^{-{1\over 2}(z - 2i\sqrt{mt})^{2}},\end{aligned}\ ] ] using the above expression to compute @xmath286 , one obtains from the resulting self - consistent equation the following eigenvalue equation for @xmath106 @xcite : _ p=0^_-^ = 1 . stability - eqn -plane , ( b ) , corresponding to the loop in the complex @xmath106-plane , ( a ) , as determined by the function @xmath287 in equation ( [ eqfond]).,width=377 ] eigenvalueequation a detailed analysis of the eigenvalue equation ( [ stability - eqn ] ) , carried out in ref . @xcite , shows that the equation admits at most one solution for @xmath106 with a positive real part , and when the solution exists , it is necessarily real . we now briefly sketch the analysis . we rewrite equation ( [ stability - eqn ] ) as @xmath288 where @xmath21 is unimodal . the incoherent state ( [ inc - state ] ) is unstable if there is a @xmath106 with a positive real part that satisfies the above eigenvalue equation . we first look for possible pure imaginary solutions @xmath289 . separating equation ( [ eqfond ] ) into real and imaginary parts , we have @xmath290\nonumber \\ & & = \frac{e^{mt}}{2t}\sum_{p=0}^\infty \frac{\left(-mt\right)^p}{p ! } \int { \mbox{d}}\omega \ , g(\omega ) \frac{\left(p+mt\right)^2}{\left(p+mt\right)^2+\left(m\sigma \omega + \sqrt{m}\mu\right)^2 } - 1 = 0 , \\ \label{iml_imag } & & { \rm i m } \left[f(i\mu;m , t,\sigma)\right]\nonumber \\ & & = -\frac{e^{mt}}{2 t } \sum_{p=0}^\infty \frac{\left(-mt\right)^p}{p!}\int { \mbox{d}}\omega \ , g(\omega ) \frac{\left(p+mt\right)\left(m\sigma \omega + \sqrt{m}\mu \right ) } { \left(p+mt\right)^2+\left(m\sigma \omega + \sqrt{m}\mu\right)^2}= 0.\end{aligned}\ ] ] in the second equation above , making the change of variables @xmath291 , and exploiting the parity in @xmath143 of the sum , we get @xmath292&=&-\frac{e^{mt}}{2 t } m\sigma\int_0^\infty { \mbox{d}}x\big\ { \left[g\left(x-\frac{\mu}{\sqrt{m}\sigma}\right ) -g\left(-x-\frac{\mu}{\sqrt{m}\sigma}\right)\right]\nonumber \\ & \times & x \sum_{p=0}^\infty \frac{\left(-mt\right)^p}{p ! } \frac{p+mt}{\left(p+mt\right)^2+m^2\sigma^2 x^2 } \big\ } = 0.\end{aligned}\ ] ] it is possible to show that the sum on the right - hand side is positive definite for any finite @xmath39 , while for our class of unimodal @xmath21 s , the term within the square brackets is positive ( respectively , negative ) definite for @xmath293 ( respectively , for @xmath294 ) . therefore , the last equation is never satisfied for @xmath295 , implying thereby that the eigenvalue equation ( [ eqfond ] ) does not admit pure imaginary solutions ( the proof holds also for the particular case @xmath296 ) . this analysis also proves that there can be at most one solution of equation ( [ eqfond ] ) with positive real part . in fact , let us consider , in the complex @xmath106-plane , the loop @xmath297 depicted in fig . [ sm - fig1](a ) , with @xmath298 and @xmath66 representing @xmath299 , respectively , and the radius of the arc @xmath300 going to @xmath301 . due to the sign properties of @xmath302 $ ] just described , we obtain in the complex-@xmath287 plane the loop qualitatively represented in fig . [ sm - fig1](b ) . while the point @xmath303 is obtained when @xmath106 is at the points @xmath298 and @xmath66 , the point @xmath304 is determined by the value of @xmath305 given by @xmath306 from a well - known theorem of complex analysis @xcite , we therefore conclude that for @xmath307 , there is one and only one solution of the eigenvalue equation with positive real part and no solution for @xmath308 . when the single solution with positive real part exists , it is necessarily real , since a complex solution would imply the presence of its complex conjugate . for @xmath158 , one has @xmath309 . for @xmath310 , the value of @xmath311 depends on the distribution function @xmath21 . it is possible to prove that the value is always smaller than @xmath312 ; this is reasonable since if the incoherent state is stable for @xmath313 , which happens when @xmath314 , it is _ a fortiori _ stable for @xmath310 . the surface delimiting the region of instability of the incoherent state ( [ inc - state ] ) in the @xmath315 phase space is implicitly given by equation ( [ rel_real ] ) that may be solved to obtain the stability threshold @xmath316 . it is reasonable to expect on physical grounds that the threshold is a single - valued function , and that for any given value of @xmath116 , it is a decreasing function of @xmath136 for @xmath317 , reaching @xmath83 for @xmath318 . these facts may be proved analytically for the class of unimodal distributions functions @xmath21 considered in this work . also , one can prove for any @xmath21 that @xmath170 approaches @xmath83 as @xmath319 , by using the integral representation [ integr_repr_2 ] & & _ p=0^ + & & = e^-mt- ( m ) _ 0^t ( mt ) . for @xmath320 , as @xmath319 , the term with the integral in the right - hand side of the last equation tends to @xmath321 , so that equation ( [ rel_real ] ) gives @xmath322 . combined with the fact that @xmath309 , we get @xmath323 . turning to a representative gaussian case , equation ( [ gomega - gaussian ] ) , and using the subscript @xmath324 to distinguish results for this case , equation ( [ integr_repr_2 ] ) gives @xmath325.\end{aligned}\ ] ] the equation @xmath326 defines implicitly the function @xmath170 . we can show that this is a single - valued function with the properties @xmath327 and @xmath328 . we have @xmath329,\end{aligned}\ ] ] which is negative as @xmath330 is positive for @xmath331 . from the implicit function theorems , we then derive that @xmath327 . on the other hand , we have @xmath332,\end{aligned}\ ] ] which is clearly negative . since we are considering @xmath333 , multiplying equation ( [ rel_real_b_gauss_3 ] ) by @xmath334 gives @xmath335.\end{aligned}\ ] ] considering the integral on the right - hand side , since @xmath336 is negative for @xmath331 , the @xmath136-derivative of the integral is negative , while its second @xmath136 derivative is positive . then the right - hand side of equation ( [ rel_real_b_gauss_4 ] ) for @xmath333 can be zero for at most one value of @xmath136 . furthermore , since for fixed @xmath337 and @xmath116 the value of @xmath338 decreases if @xmath39 increases , the @xmath136 value for which @xmath326 decreases for increasing @xmath39 at fixed @xmath116 . this concludes the proof . furthermore , for what we have seen before , @xmath339 and @xmath340 for @xmath341 . it is evident from the above analysis that the proof is not restricted to the gaussian case , but works equally well for any @xmath21 such that @xmath342 is positive for any @xmath343 . however , on physical grounds , we are led to assume that the same conclusions hold for any even single - humped @xmath21 . we conclude on the basis of the above analysis that @xmath344 at the point of neutral stability , so that equation ( [ stability - eqn ] ) gives @xmath170 to be satisfying = _ p=0^_-^ . sigma_inc in the @xmath315 space , the above equation defines the stability surface @xmath195 . there will similarly be the stability surface @xmath345 . the two surfaces coincide on the critical lines on the @xmath165 and @xmath185 planes where the transition becomes continuous ; outside these planes , the surfaces enclose the first - order transition surface @xmath191 i.e. , @xmath346 . let us show by taking limits that the surface @xmath195 meets the critical lines on the @xmath165 and @xmath185 planes , and also obtain its intersection with the @xmath186-plane . on considering @xmath120 at a fixed @xmath136 , only the @xmath283 term in the sum in equation ( [ sigma_inc ] ) contributes , so that one has _ m 0 , t fixed^inc(m , t)=_c(m=0,t ) , with the implicit expression of @xmath347 given earlier . similarly , one has _ t t_c^- , m fixed^inc(m , t)=0 . when @xmath348 at a fixed @xmath116 , we get ^inc_noiseless(m)_t 0 , m fixed^inc(m , t ) , with & & 1=- _ -^d . for the representative case of the gaussian @xmath21 , equation ( [ gomega - gaussian ] ) , we get from equation ( [ sigma_inc ] ) that 1= _ p=0^ , gaussian - eqn where @xmath349 is the complementary error function : @xmath350 . ) , we show here schematic landau free energy @xmath351 vs. @xmath1 for first - order transitions at fixed @xmath116 and @xmath136 while varying @xmath39 . panels ( i ) and ( vii ) correspond to the synchronized and incoherent phase being at the global minimum . in panel ( iii ) ( respectively , ( v ) ) , the synchronized ( respectively , incoherent ) phase is at the global minimum , while the incoherent ( respectively , synchronized ) phase is at a local minimum , hence , metastable . panel ( iv ) corresponds to the first - order transition point , with the two phases coexisting at two minima of equal heights.,width=453 ] fig : fe numerics choosing @xmath352 , and @xmath353 , equation ( [ gaussian - eqn ] ) gives @xmath354 . then , starting with the incoherent state ( [ inc - state ] ) at a given @xmath39 and evolving under the dynamics ( [ eom - scaled ] ) , our theoretical continuum - limit analysis predicts that the order parameter @xmath1 for @xmath355 relaxes at long times from its value equal to @xmath83 to its stationary state value corresponding to the synchronized phase . for @xmath356 , on the other hand , @xmath1 remains zero for all times . [ cols= " < , > " , ] alphahmf - overdamped we now turn our attention to the case when all the oscillators have the same natural frequency , say , @xmath37 . as discussed in section [ kuramoto - bare ] , one can scale out @xmath37 from the equations of motion by going to a comoving frame rotating uniformly with frequency @xmath37 with respect to the laboratory frame . thus , we are effectively considering the dynamics without the presence of quenched torques . we will first study an overdamped model with noise @xcite . namely , the model obtained from the one studied in section [ seckurlongrange ] by adding gaussian noise terms to the equations of motion , but excluding the frequency terms . later , we will focus on the underdamped model with noise . let us then begin with the equation of motion for the @xmath22th oscillator : = _ j=1^n + _ i(t ) , kuramoto - eom_decaying_noise where @xmath87 is a gaussian white noise : _ i(t ) = 0 , + _ j(t ) = 2t_ij(t - t ) . noise_statistics the equation of motion ( [ kuramoto - eom_decaying_noise ] ) describes the overdamped dynamics of the so - called @xmath4-hmf model @xcite , within a canonical ensemble ( see equation ( [ long - range - inertia ] ) below ) . the definition of @xmath357 and the closest distance convention are the same as in section [ seckurlongrange ] , and so is the procedure to obtain the continuum limit ( @xmath358 , keeping the product @xmath359 constant at unity , where @xmath360 stands for the lattice constant ) . to discuss this limit , we introduce the variable @xmath361 $ ] , obtained as the continuum limit of @xmath362 . however , contrary to section [ seckurlongrange ] , the one - particle distribution function will now not depend on the frequency . here , we introduce the one - particle distribution function @xmath363 , defined such that the quantity @xmath364 represents the fraction of oscillators located between @xmath365 and @xmath366 that at time @xmath51 has their phase between @xmath11 and @xmath367 . the normalization is _ -^ ( , s , t ) = 1 s. norm_cond_rhoq in the continuum limit , the equation of motion takes the form = _ 0 ^ 1 s _ -^ ( ,s,t ) + ( s , t ) , kuramoto - eom_decaying_noise_cont where the normalizing factor @xmath368 and the closest distance convention @xmath369 are given in equations ( [ norm_fact ] ) and ( [ clos_dist_conv_cont ] ) , respectively . the statistical properties of the noise become ( s , t ) = 0 , + ( s , t ) ( s,t ) = 2t(s - s)(t - t ) . noise_statistics_cont the fokker - planck equation governing the evolution of @xmath363 is fokkerplanck_eq_decaying + = - \ { ( , s , t ) } + t . the generic stationary solution @xmath370 of the fokker - planck equation ( [ fokkerplanck_eq_decaying ] ) is obtained by setting the left hand side to zero , yielding _ 0(,s ) = a(s ) , noise_stationary where the constants @xmath371 for every @xmath365 are determined by the normalization condition ( [ norm_cond_rhoq ] ) . there are also consistency relations to be satisfied , as we now show . let us denote by @xmath372 and @xmath373 the two components of the local magnetization : m_x(s ) _ -^ _ 0(,s ) , mxsdef + m_y(s ) _ -^ _ 0(,s ) . mysdef from the definition of the modified bessel function of the first kind of order @xmath275 , i_n(x ) = _ -^ ( n ) e^x , defbessel and denoting _ x^()(s ) = _ 0 ^ 1 s , mxsaldef + _ y^()(s ) = _ 0 ^ 1 s , mysaldef one obtains for the normalization constant the equation a(s ) = ^-1 , normconst together with the self - consistency relations = ( ) . selfalcons we note for later use that if we choose an @xmath365-independent stationary distribution @xmath374 , then @xmath371 , @xmath372 and @xmath373 are also @xmath365-independent , with @xmath375 . linear - stability - alphahmf - overdamped let us now consider the @xmath365-independent ( that is , the mean - field ) incoherent state , obtained when @xmath376 , i.e. , _ 0 ( ) = . stationincoh as before , its linear stability can be analyzed by posing ( , s , t ) = + ( , s , t ) ; || 1 , alphaperturbation_b and studying the linearized fokker - planck equation for @xmath377 : fokkerplanck_eq_decaying_linear = _ 0 ^ 1 s _ -^ ( ,s,t ) + t . the procedure for stability analysis is the same as that adopted in the preceding subsection . we first perform a fourier expansion in @xmath11 : ( , s , t ) = _ k=-^+ _ k(s , t ) e^ik , fourier_theta_noise which when used in equation ( [ fokkerplanck_eq_decaying_linear ] ) gives linear_cont_eq_four_noise =( _ k,1 + _ k,-1 ) _ 0 ^ 1 s -k^2 t_k(s , t ) . for @xmath378 , the first term on the right hand side of equation ( [ linear_cont_eq_four_noise ] ) vanishes , and we have _ k(s , t ) = _ k(s,0 ) e^-k^2 t t ; k 1 ; damped_modes these are perturbations that decay exponentially in time , and thus correspond to stable modes . the equation for @xmath100 , = _ 0 ^ 1 s -t_1(s , t ) , linear_cont_eq_fourpm1_noise is studied by performing a further fourier expansion in @xmath365-space : _ 1(s , t ) = _ n=-^+ _ 1,n(t ) e^2i ns . fourier_sspace_noise substituting in equation ( [ linear_cont_eq_fourpm1_noise ] ) , we obtain = _ 1,n(t ) -t_1(t ) , linear_cont_eq_fourpm1s_noise where @xmath379 is given by equation ( [ eigenlambda ] ) . we therefore have _ 1,n(t ) = _ 1,n(0 ) . solution_n1_noise for a fixed @xmath29 , this expression determines the value of the temperature for which the mode @xmath380 is stable . precisely , the mode @xmath380 decays exponentially in time and is therefore stable for @xmath381 , while it is unstable , growing exponentially in time , for @xmath382 . therefore , the critical temperature for the neutral stability of @xmath383 is t_c , n = . crit_temper since , as previously explained , @xmath384 , and @xmath379 is a decreasing function of @xmath385 , we have @xmath386 , and = t_c,0 > t_c,1 > t_c,2 > t_c,3 > order_critical we note in particular that for @xmath387 , we have @xmath388 for @xmath389 , so that the modes @xmath380 for @xmath390 never destabilize . numerics - alphahmf - overdamped in the following , we take @xmath391 without loss of generality ( with a rescaling of the time unit , it is always possible to reduce to such a case ) . from the analysis presented above , we see that for @xmath314 , the incoherent state is stable . decreasing the temperature , the first perturbation mode to destabilize will be @xmath392 , which happens at @xmath318 . decreasing further the temperature , the modes @xmath380 with @xmath390 will progressively destabilize . we now discuss the results of simulations of the equation of motion ( [ kuramoto - eom_decaying_noise ] ) for a system with @xmath393 oscillators with @xmath394 . the effect of the stochastic noise has been taken into account with the same method as that described in appendix b , equation ( [ formalintegration3 ] ) for the case of systems with inertia . we have studied the observables @xmath395 defined in equation ( [ discrete_order_params ] ) . in fig . [ sas - fig1 ] , we show the time evolution of @xmath396 , @xmath397 , @xmath398 and @xmath399 for simulations performed at @xmath400 , with initial conditions reproducing the incoherent state , @xmath401 for all @xmath275 , obtained by taking the phases independently and uniformly distributed between @xmath83 and @xmath201 . from equation ( [ crit_temper ] ) , we find that @xmath402 lies between @xmath403 and @xmath404 . then , in particular , the observables plotted in fig . [ sas - fig1 ] should all increase exponentially in time . the plot shows that the agreement between the numerical and the theoretical growth rates is very good . ) , the figure shows the time evolution of the observables @xmath405 , and @xmath399 starting from an initial state @xmath406 that has been obtained by extracting the @xmath407 s uniformly in @xmath408 $ ] . therefore the initial state is the incoherent one . in these runs @xmath409 and @xmath400 . for these values of @xmath4 and @xmath136 , the fourier modes @xmath410 are all linearly unstable . in particular , @xmath411 , @xmath412 , @xmath413 , @xmath414 . consequently , @xmath405 , and @xmath399 all grow exponentially in time , initially , from their initial values at @xmath415 . the simulations have been performed with @xmath416 oscillators , and the plotted data involve and average over @xmath417 independent initial conditions and dynamical realizations . the plots show that after the initial exponential growth , @xmath396 attains a value very close to unity , while @xmath418 all decay to a value very close to zero ( and compatible to @xmath83 considering the finite size effects ) . the straight lines show the theoretical initial exponential growths , with rates given by @xmath419 . the agreement of the growth rates between theory and simulations is very good.,width=377 ] as is evident from the plot in fig . [ sas - fig1 ] , the final state reached by the system is the mean - field synchronized one , similar to what was observed for the model ( [ kuramoto - eom_decaying ] ) . one can argue that this is the gibbs - boltzmann equilibrium state of our system . for example , it has been proved in ref . @xcite that the equilibrium state of a system of oscillators interacting by long - range interactions on a lattice with periodic boundary conditions is the same as that of the corresponding mean - field system . in the following subsection , we show that indeed such a state is dynamically stable at temperatures @xmath420 . linear - stability - synchronized - alphahmf - overdamped let us then consider the stationary state ( [ noise_stationary ] ) in which there is only @xmath11 and no @xmath365 dependence : _ 0 ( ) = a , noise_stationary_nos where we have put @xmath391 . using the definitions ( [ norm_fact ] ) of @xmath368 and of the magnetization components ( [ mxsdef ] ) and ( [ mysdef ] ) , and exploiting the global rotational invariance in @xmath11 of the system ( [ kuramoto - eom_decaying_noise ] ) to put @xmath421 , we rewrite this state as _ 0 ( ) = a . noise_stationary_nos_b the expression of the normalization constant becomes a = ^-1 , normconst_nos while the self - consistency relation is m_x = ( ) . selfalcons_nos the last equation gives a non - vanishing @xmath422 for @xmath423 , as follows from the properties of @xmath424 and @xmath425 , see ref . @xcite ; for @xmath426 , the state ( [ noise_stationary_nos_b ] ) reduces to the uniform one , equation ( [ stationincoh ] ) . as before , the stability of the state ( [ noise_stationary_nos_b ] ) is studied by analyzing the linearized equation obtained by inserting in equation ( [ fokkerplanck_eq_decaying ] ) the expansion ( , s , t ) = _ 0 ( ) + ( , s , t ) ; || 1 . exp_nonunif in equation ( [ exp_nonunif ] ) , both @xmath427 and @xmath374 are normalized , implying that _ -^ ( , s , t ) = 0 . zeronorm from equation ( [ fokkerplanck_eq_decaying ] ) , we have at leading order in @xmath428 the linearized equation fokkerplanck_eq_decaying_nonunif = m_x ( ( , s , t ) ) + - ( _ 0 ( ) ) + t . since the stationary state @xmath374 is not uniform in @xmath11 , a fourier expansion in @xmath11 is not useful . performing a fourier expansion in @xmath365 , ( , s , t ) = _ n=-^+ _ n(,t ) e^2i ns , fourier_sspace_nonunif we have fokkerplanck_eq_decaying_nonunif_s = m_x ( _ n(,t ) ) + -_n ( ) ( _ 0 ( ) ) + t , where we have used _ n ( ) . def_lambda from the definitions of @xmath379 and @xmath368 , we have @xmath429 . let us now look for solutions of equation ( [ fokkerplanck_eq_decaying_nonunif_s ] ) of the form _ n(,t ) = _ n ( , ) e^t . posit_spectrum equation ( [ fokkerplanck_eq_decaying_nonunif_s ] ) then gives fokkerplanck_eq_decaying_nonunif_mu _ n ( , ) = m_x ( _ n ( , ) ) + -_n ( ) ( _ 0 ( ) ) + t . to solve this equation and to compute the eigenvalues @xmath430 , we adopt the following strategy . the function @xmath431 being @xmath201-periodic in @xmath11 , it can be expanded in the basis functions @xmath432 with @xmath433 . then , we multiply equation ( [ fokkerplanck_eq_decaying_nonunif_mu ] ) in turn by the basis functions , and then integrate over @xmath11 from @xmath83 to @xmath201 to obtain a system of algebraic equations . one gets an identity for @xmath283 , while for @xmath434 , we obtain the system _ x , n^(p ) = pm_x - tp^2 _ x , n^(p ) + _ n()p_x , n^(1 ) , system_x + _ y , n^(p ) = pm_x - tp^2 _ y , n^(p)+_n()p_y , n^(1 ) , system_y where we have introduced the notations ( _ x , n^(p),_y , n^(p ) ) _ -^ ( p , p ) _ n ( , ) , tildemag and m_x^(p ) _ -^ p _ 0 ( ) = ( ) . orderpmag now , clearly , @xmath435 , @xmath436 and @xmath437 . there is one system of equation given by equations ( [ system_x ] ) and ( [ system_y ] ) for each value of @xmath438 . these systems are associated with non - hermitian matrices ; therefore , the eigenvalues @xmath430 will in general be complex . the stationary state ( [ noise_stationary_nos_b ] ) is linearly stable if the eigenvalues of all these systems have negative real parts . we have evaluated numerically the spectrum , and the analysis has put in evidence that this is the case . actually , there is also a zero eigenvalue , and in general , the presence of purely imaginary eigenvalues ( zero being a particular case ) implies that the stationary state is only spectrally stable , while it might be linearly unstable . however , proving that the zero eigenvalue has multiplicity one ( see below ) ensures that linear stability holds @xcite . before describing the result of the numerical analysis of the systems ( [ system_x ] ) and ( [ system_y ] ) , we give an argument that points towards the stability of the stationary state ( [ noise_stationary_nos_b ] ) . let us define the entropy functional s= -_0 ^ 1 s _ -^ ( , s , t ) , entr_funct and the energy functional e= _ 0 ^ 1 s _ -^ ( , s , t ) u ( , s , t ) , ener_funct where @xmath439 is the mean - field potential u(,s , t ) = - _ 0 ^ 1 s _ -^ ( ,s,t ) . meanfieldpot with the dynamics of @xmath363 governed by the fokker - planck equation ( [ fokkerplanck_eq_decaying ] ) ( with @xmath391 in the present analysis ) , it is not difficult to obtain that htheorem ( e - ts ) f + = - _ 0 ^ 1 s _ -^ ( ( , s , t ) + t)^2 0 . we thus see that there is an @xmath239-theorem @xcite associated with the evolution of @xmath363 , with the @xmath239-function being the free energy @xmath440 $ ] ; this is in analogy with the mean - field case ( @xmath441 ) studied in ref . the right hand side of the last equation vanishes only for the stationary states given in ( [ noise_stationary ] ) , and in particular , for the state ( [ noise_stationary_nos_b ] ) . in addition , as proved in @xcite , the @xmath365-independent stationary state ( [ noise_stationary_nos_b ] ) realizes the minimum of the free energy . therefore , equation ( [ htheorem ] ) suggests that if this state is perturbed , the dynamics tends to restore it . the eigenvalues of the system ( [ system_x ] ) and ( [ system_y ] ) have been numerically evaluated by truncating the system at a finite value of @xmath278 , denoted by @xmath442 . as a matter of fact , we have found that the eigenvalues @xmath430 of the systems ( [ system_x ] ) and ( [ system_y ] ) always have a negative real part for any value of @xmath443 between @xmath83 and @xmath81 and for any temperature in the range @xmath444 ( except for the zero eigenvalue that we will consider in detail below ) . we recall that varying @xmath275 and @xmath4 , the factor @xmath443 can take any value in that range . obviously , by truncating the system , one can find only a finite number of eigenvalues , but by increasing the truncation value @xmath442 , we have checked that the new eigenvalues have negative real parts with larger absolute values , and the eigenvalues with negative real parts that have smaller absolute values converge extremely fast . we have also found that for @xmath136 not close to @xmath83 , the eigenvalues are in addition real . this can be understood by considering the systems ( [ system_x ] ) and ( [ system_y ] ) for @xmath445 . in that case , since @xmath446 for @xmath447 , they reduce to system_x_diag _ x , n^(p ) = - tp^2 _ x , n^(p ) + _ p,1_n ( ) _ x , n^(1 ) , + system_y_diag _ y , n^(p ) = - tp^2 _ y , n^(p ) + _ p,1_n()_y , n^(1 ) . the right hand sides give directly the eigenvalues . they are real and all negative , since @xmath445 and @xmath448 ( except for @xmath136 exactly equal to @xmath449 and for @xmath282 , where @xmath450 and then the right hand sides for @xmath451 are zero ) . by continuity , the eigenvalues will be real for at least a range of temperatures @xmath136 smaller than @xmath449 . we conclude the analysis by studying the zero eigenvalue for @xmath452 . for this , it is not convenient to analyze the systems ( [ system_x ] ) and ( [ system_y ] ) , but to start directly from equation ( [ fokkerplanck_eq_decaying_nonunif_mu ] ) with @xmath453 , i.e. , m_x ( _ n(,0 ) ) -_n ( ) ( _ 0 ( ) ) + + t = 0 . fokkerplanck_eq_decaying_nonunif_mu_0 the solution of this equation that satisfies the periodicity condition and equation ( [ zeronorm ] ) is _ n(,0 ) = _ n ( ) , solution_mu_0 where the normalization constant @xmath298 is given in equation ( [ normconst_nos ] ) , and where we have used the definition ( [ tildemag ] ) . this equation shows that in order to have a non - trivial solution , @xmath454 and @xmath455 can not both be equal to @xmath83 . we still have to satisfy equation ( [ tildemag ] ) as a self - consistent equation . multiplying equation ( [ solution_mu_0 ] ) by @xmath456 and by @xmath457 , we obtain self_mtildex _ x , n^(1 ) = _ x , n^(1 ) ( 1 - t - m_x^2 ) , + _ y , n^(1 ) = _ y , n^(1 ) _ n ( ) . self_mtildey the first of these equations is satisfied by @xmath458 , or by self_mtildex_b m_x = , that must be satisfied together with the self - consistent relation ( [ selfalcons_nos ] ) . in fig . [ sas - fig3 ] , we plot @xmath422 as a function of @xmath136 as determined by the self - consistent relation ( [ selfalcons_nos ] ) and by equation ( [ self_mtildex_b ] ) for @xmath459 . we see that there is no solution for @xmath452 . since the right hand side of equation ( [ self_mtildex_b ] ) decreases for decreasing @xmath443 , this also proves that there is no solution for any @xmath443 . therefore , the only solution of equation ( [ self_mtildex ] ) is @xmath458 . this requires that @xmath460 , and then equation ( [ self_mtildey ] ) becomes @xmath459 . this is verified only for @xmath282 . we have finally arrived at the conclusion that equation ( [ fokkerplanck_eq_decaying_nonunif_mu_0 ] ) admits a solution only for @xmath282 , and that this solution is unique and is given by _ 0(,0 ) = _ 0 ( ) , solution_mu_0_n0 with @xmath460 . this solution represents a global rotation of all oscillators , and is a neutral mode due to the global rotational invariance . the uniqueness of the mode associated with the zero eigenvalue assures that there are no secular terms with a linear growth , thus completing the proof of the linear stability of @xmath374 . as a function of @xmath136 as determined implicitly by the self - consistent relation ( [ selfalcons_nos ] ) and by equation ( [ self_mtildex_b ] ) with @xmath461 . the two curves do not intersect at any @xmath136 in the range @xmath452 , showing that there is no solution satisfying both relations.,width=377 ] alphahmf we will now be concerned with the model with inertia , that in the overdamped limit reduced to the model studied in the preceding subsection . the equations of motion are = v_i , + long - range - inertia + m= - v_i + _ j=1^n + _ i(t ) , with the same definitions as before of @xmath357 and of the closest distance convention . we recall the statistical properties of the gaussian white noise @xmath119 : _ i(t ) = 0 , _ i(t)_j(t ) = 2 t _ ij(t - t ) . etatide - prop_bis the equations of motion ( [ long - range - inertia ] ) describe the evolution of the @xmath4-hmf model @xcite , within a canonical ensemble . by performing the reduction to dimensionless quantities as in equations ( [ dmsless1])-([dmsless6 ] ) , the equations of motion become = v_i , + long - range - inertia_dms + = - v_i + _ j=1^n + _ i(t ) , where we have disregarded the overbars of the dimensionless quantities for notational convenience , and we have _ i(t ) = 0 , _ i(t)_j(t ) = 2 ( t/ ) _ ij(t - t ) . etatide - prop_dms the continuum limit of the dynamics is implemented in a manner analogous to that in preceding sections , by introducing the variable @xmath361 $ ] as the continuum limit of @xmath362 . the one - particle distribution function @xmath462 is such that @xmath463 is the fraction of oscillators located between @xmath365 and @xmath366 that at time @xmath51 has phase between @xmath11 and @xmath367 and angular velocity between @xmath200 and @xmath464 . the normalization is _ -^ _ -^v f(,v , s , t ) = 1 s. norm_cond_finert the equations of motion in the continuum limit are = v(s , t ) , + long - range - inertia_cont + = - v(s , t ) + _ 0 ^ 1 s _ -^ _ -^v f(,v,s,t ) + ( s , t ) , with @xmath368 defined previously . the kramers equation for @xmath462 is kramers_decaying = - v + + + . linear - stability - alphahmf below we will present the results of numerical simulations by plotting , as before , the first few of the observables defined in equation ( [ discrete_order_params ] ) . before that , we perform a stability analysis of the @xmath365-independent incoherent stationary state of the kramers equation ( [ kramers_decaying ] ) given by f_0(v ) = e^- . homog_kramers similarly to equation ( [ expansion_linear ] ) , we linearize the kramers equation by posing f(,v , s , t ) = f_0(v ) + e^t f(,v , s ) ; || 1 , expansion_decaying where normalization of both @xmath465 and @xmath466 implies that _ -^ _ -^v f(,v , s ) = 0 s. norm_cond_delta_0 at leading order , we obtain from equation ( [ kramers_decaying ] ) that kramers_decaying_linear f(,v , s ) = - v + ( v f(,v , s ) ) + , + - _ 0 ^ 1 s _ -^ _ -^v f(,v,s ) . the analysis of equation ( [ kramers_decaying_linear ] ) is very similar to that followed in section [ linearstability - incoherent - inertia ] , and therefore , we do not repeat all the details here . labelling @xmath267 with the eigenvalue @xmath467 , we pose expansion_nu f(,v , s , ) = _ k=-^b_k(v , s , ) e^ik , with @xmath468 and @xmath469 . substituting in equation ( [ kramers_decaying_linear ] ) , we have + + ( 1 - -ikv)b_k(v , s , ) + = i ( _ k,1 - _ k,-1 ) _ 0 ^ 1 s 1,b_k ( s , ) , eq_b_deltaomega where the scalar product is defined by , ( s,s ) _ -^ v ^*(v , s)(v , s ) . scalar - prod_sdep for @xmath470 , when the right hand side is equal to @xmath83 , equation ( [ eq_b_deltaomega ] ) is identical to equation ( [ b - eqn ] ) with @xmath471 . therefore , we can immediately write down the negative eigenvalues as _ p , k = - - k^2 t ; p = 0 , 1 , 2, eigenvalues_decaying for @xmath100 , we proceed as follows . let us consider only @xmath472 , since @xmath473 . we perform the expansion b_1(v , s , ) = _ n=-^+ b_1,n(v,)e^2i ns . expansion_thetas substituting in equation ( [ eq_b_deltaomega ] ) , we obtain + + ( 1 - -ikv)b_k(v , s , ) + = _ n ( ) i 1,b_1,n , eq_b_deltaomega_k1 where now the scalar product @xmath474 does not depend on @xmath365 , and where @xmath443 is defined in equation ( [ def_lambda ] ) . comparing with equation ( [ b - eqn ] ) , it is evident that on performing the same analysis as in section [ linearstability - incoherent - inertia ] , one arrives at the following implicit equation for the eigenvalue @xmath467 : _ n()_p=0^ - 1 = 0 . stability - eqn_deltaomega the stability threshold is again given by the value of the last expression for @xmath475 : - 1 = 0 . stability - eqn_deltaomega_crit we therefore obtain the same critical temperature for the @xmath275th mode as given in equation ( [ crit_temper ] ) ( where now @xmath391 ) . numerics - alphahmf in simulations , we monitor as in the previously discussed cases the observables @xmath395 defined in equation ( [ discrete_order_params ] ) . in fig . [ fignotesinertia ] , we show the evolution of @xmath396 , @xmath397 , @xmath398 and @xmath399 , in simulations of the equations of motion ( [ long - range - inertia ] ) with @xmath476 , at temperature @xmath477 and @xmath394 , starting from an initial state uniform in @xmath11 and gaussian in the velocity , equation ( [ homog_kramers ] ) . for these values of @xmath136 and @xmath4 , the modes @xmath478 for @xmath410 are all unstable ( see the critical temperatures in the caption of fig . [ sas - fig1 ] ) . the simulation has been performed with @xmath479 oscillators . ) , the figure shows the time evolution of the observables @xmath480 for a system of @xmath479 oscillators with @xmath481 , while starting from an initial state uniform in @xmath11 and gaussian in the velocity , equation ( [ homog_kramers]).,width=377 ] fignotesinertia we see that , as expected , the incoherent state is not stable , since the order parameter @xmath482 grows exponentially and reaches an asymptotic value that at this temperature is very close to @xmath81 . this is similar to what happens in the simulations of the models in the preceding subsections . also , the long - time decay of @xmath483 with @xmath389 is similar . however , contrary to the cases in the previous subsections , now the initial exponential growth of these parameters is not visible in fig . [ fignotesinertia ] . this is probably due to finite - size effects . we stress that these effects for a given number of oscillators @xmath19 are expected to be more marked for a system with inertia than in an overdamped system , since the former has two dynamical variables per oscillators , and consequently , the distribution @xmath261 depends on two dynamical variables . in fig . [ fignotesinertianbig ] , we plot the results of a simulation run with @xmath484 oscillators . for this larger system , the initial exponential growth of all the @xmath395 s is clearly visible . the theoretical rates are also shown with full lines . the agreement with the simulation is satisfactory . the attainment of the asymptotic value of @xmath482 and the decay to zero of the parameters @xmath483 with @xmath389 occur at later times with respect to the smaller system , and are not displayed in the figure . ) , the figure shows the time evolution of the observables @xmath480 for a system of @xmath484 oscillators with @xmath481 , while starting from an initial state uniform in @xmath11 and gaussian in the velocity , equation ( [ homog_kramers ] ) . the full straight lines show the theoretical growth rates.,width=377 ] fignotesinertianbig conclusions spontaneous synchronization appears naturally out of a competition between two qualitatively different dynamical regimes in systems that can be described as a set of interacting oscillators . the main purpose of this work was to show that in addition to this purely dynamical view , introducing noise into the dynamics allows one to study synchronization in the framework of statistical mechanics . this is not only possible , but also very useful , since it paves the way for use of efficient analytical tools commonly employed in the study of the statistical behaviour of many - body systems . one then derives that synchronization is a phase transition , characterized by the appearance of a non - vanishing value of an appropriate order parameter . we have shown that , interpreted as a system of interacting particles , the kuramoto model and its various extensions are long - range interacting lattice systems . furthermore , they are in their original setting mean - field systems , the extreme case of long - range interacting systems , where all pairs of particles interact with equal coupling strengths . we have also considered models where the coupling strengths decay slowly with the distance between the lattice sites . long - range systems often enjoy peculiar features , both in and out of equilibrium , due to the non - additivity of the interaction energy between subparts of the system . the form of the interaction in the kuramoto model and its extensions may be used as a prototypical interaction for studying these features . statistical dynamics of long - range systems , in the limit of a very large number of particles , can be very well described by equations that involve only the one - particle distribution function . the particular time evolution equation for the one - particle distribution depends on the system at hand : ( i ) for overdamped systems , it is the continuity equation in the case of noiseless dynamics , and the fokker - planck equation for the noisy case ; ( ii ) for underdamped systems , it is the vlasov equation for the noiseless dynamics , and the kramers equation for the noisy case . stable stationary solutions of these equations correspond to stationary states in which the system remains trapped for a time that diverges with the system size ( the limit in which the equations for the one - particle distribution function become exact ) . as exemplified in this review , the phase transitions mentioned above are then a change of the stability properties of the stationary states corresponding to the synchronized and the unsynchronized state . the presence of distributed natural frequencies in the dynamics of the kuramoto model leads to a violation of detailed balance in the stationary state , thereby resulting in long - time stationary states which are out of equilibrium . these are the so - called nonequilibrium stationary states ( nesss ) characterized by a net non - zero probability current around a closed loop in the configuration phase . as a result , one can not use the free energy as a thermodynamic potential to determine the nature of the stationary state of the system . however , to this end , we have shown that it is possible to employ successfully at least in numerical simulations probability distributions of the order parameter analogous to those employed in equilibrium statistical mechanics . in this review , we have restricted our analysis to unimodal distributions for the natural frequencies . let us comment on this point with respect to the mean - field models . for such distribution functions , the synchronization transition in the kuramoto model and in its noisy extension is continuous , i.e. , the order parameter of the stable stationary state grows continuously from zero as the coupling strength increases beyond a critical value at a given temperature , or , equivalently , as the temperature is decreased below a critical value for a given coupling strength . on the other hand , in the model with inertia , the transition becomes of first - order type : we have found that in certain ranges of the parameters , both the incoherent and the synchronized state are dynamically stable ( this situation is often referred to as bistability ) , with one of the two states being the globally stable state , depending on the variation of the parameters within the range . this gives rise to the existence of hysteresis loops . the overall picture is probably different with more general frequency distributions . there have been studies of the original kuramoto model with non - unimodal @xmath21 . it has been shown that in the case of a uniform distribution , that can be considered a limiting case ( although with a singular derivative ) of a unimodal distribution , the transition becomes of a first - order type , since at the threshold value @xmath40 of the coupling parameter given by eq . ( [ kura - barekc ] ) , there appears a solution of eq . ( [ kura - bifurcation ] ) with @xmath485 @xcite . also , continuous bimodal distributions have been studied , and it has been shown that , depending on the structure of the distribution , there can be bistability and states with clusters of oscillators locked at different frequencies ( related to the maxima of the bimodal distribution ) @xcite . here , we have discussed models in which the mean - field interaction is replaced by coupling strengths between the oscillators that decay as a power law with the distance between the oscillators residing on the sites of a lattice . in particular , we have focussed on one - dimensional lattices , with the parameter @xmath4 characterizing the decay being smaller than @xmath81 , to remain within the framework of long - range interactions . furthermore , we have imposed periodic boundary conditions . as we have stressed while introducing this class of systems , periodic boundary conditions cause the uniformity on the lattice of the equilibrium or stationary states , but they do not a priori rule out the influence of the lattice structure on the dynamical behavior . however , we have shown that the mean - field fourier mode of the spatial distribution of oscillator phases dominates the out - of - equilibrium dynamics at long times , since it is this mode that gets destabilized first on increasing the coupling constants or decreasing the temperature . on the other hand , the non - zero fourier modes destabilize at higher coupling constants or smaller temperatures . this is common to all the lattice models we have analyzed . in particular , the mean - field mode dominates in the underdamped noisy dynamics considered in section [ alphahmf ] ; a similar dominance has been found in the study of the microcanonical ensemble dynamics of this system , i.e. , without the noise @xcite . although we do not have analytical or numerical evidences , we feel that it is not unreasonable to adduce the hypothesis that also with more general boundary conditions , the mean - field mode is the one relevant for the dynamics ; however , we understand that without a detailed analysis , this statement remains at the level of speculation . let us remark that there have been earlier works on the kuramoto model with coupling constants decaying with a power law on one - dimensional periodic lattices , with the purpose to study the existence of the synchronized phase as a function of the power - law parameter @xmath4 in the limit @xmath26 . the critical value of @xmath4 for the existence of the synchronized phase has been numerical evaluated to be ( about ) @xmath486 @xcite ; in ref . @xcite , a spin wave approximation and simulations performed at larger @xmath19 values suggest on the other hand that the critical value is @xmath487 . we now point out some important issues that have not been discussed in this review , e.g. , details of the dynamical behavior of the system for a large but finite number of oscillators . among finite - size effects , of particular relevance are slow processes out of equilibrium , and the scaling of the associated timescales with the system size . when the stationary states of the single - particle equations are unstable , we expect that such relaxation does not depend on the size of the system , for large enough system size . on the other hand , stable states could be destabilized by finite - size effects . for example , in hamiltonian long - range systems , finite systems slowly evolve in time out of the stationary states of the vlasov equation that governs the dynamics for @xmath488 . the lifetime of these quasi - stationary " states generally diverges with the system size as a power law @xcite . we expect that something similar may happen in the noisy driven systems studied in this review . this could affect , e.g. , the rate of hopping between stationary states when there is bistability . another property that is affected by the finite size of the system is the stability property of the incoherent state of the original kuramoto model . as a matter of fact , it has been shown that the incoherent state , although neutrally stable in the limit @xmath26 , becomes fully stable for finite @xmath19 @xcite , due to a mechanism very similar to that of landau damping in plasma physics @xcite . in conclusion , we would like to stress that the kuramoto model and its extensions , besides being related to real systems as emphasized in the introduction , provide an interesting benchmark to study and analyze a variety of physical properties . in fact , they offer the possibility to consider synchronization both as a purely dynamical effect and as an emerging phenomenon typical of the statistical behavior of many - body systems . furthermore , the long - range character of the interaction gives rise to some peculiar properties that are typical for this class of systems . we hope that this review has succeeded in giving a flavor of the aforementioned issues , and will serve as an invitation to indulge in further studies of the kuramoto model . we acknowledge the hospitality of ens - lyon , and support of the cefipra grant 4604 - 3 ( s.g . ) and the grant anr-10-cexc-010 - 01 ( s.g . and s.r . ) . s. g. and a. c. acknowledge the hospitality of the universit di firenze . we warmly thank f. bouchet , t. dauxois , a. ghosh , m. komarov , d. mukamel , c. nardini , h. park , a. patelli , a. pikovsky , m. g. potters , and h. touchette for fruitful discussions over the years on topics reported in this review . we thank the galileo galilei institute for theoretical physics ( florence ) for the hospitality and infn for partial support during the completion of this work . app - power - distribution here , we briefly discuss , following refs . @xcite , how the dynamics ( [ eom ] ) arises in connection with electrical power distribution networks . the essential elements of an electrical power distribution network or grid are synchronous generators located at power plants and motors located with the consumers . while a generator converts mechanical ( or other forms of energy ) into electrical energy , the reverse is true for a motor . let @xmath489 denote the power , which being generated is a positive quantity for a generator and being consumed is negative for a motor . either unit basically consists of a rotating turbine whose state for the @xmath25th unit is represented by its phase _ j(t)=t+_j(t ) , where @xmath215 is the standard supply frequency , @xmath490 hz typically , while @xmath491 is the deviation from uniform rotation . from considerations of energy conservation , the generated or consumed power @xmath492 of the @xmath22th element equals the sum of the power @xmath493 exchanged with the grid , the power @xmath494 accumulated in the turbine , and the amount @xmath495 dissipated in overcoming friction , where @xmath496 is the moment of inertia of the turbine and @xmath497 is the friction constant . the power transmitted between two elements @xmath25 and @xmath22 connected by a transmission line depends on the phase difference across the ends of the transmission line , and is given by @xmath498 , where @xmath499 is the maximum capacity of the transmission line . with @xmath500 , we then have p^source_i=()^2+()^2+_j p_max;ji ( _ j-_i ) . with the assumption that @xmath501 , one arrives at the equation of motion @xcite = p_i--_jk_ji(_j-_i ) , eom - electric where & & p_i= , + & & = , + & & k_ji=. in the mean - field approximation , where every unit @xmath22 is connected to every other unit @xmath25 with equal strength and @xmath502 , where @xmath19 is the total number of nodes in the network , equation ( [ eom - electric ] ) reduces to = p_i--_j(_j-_i ) . eom - electric - mfa note that the @xmath503 s are intrinsic to the units and in general vary from one unit to another , so that they may be regarded as quenched random variables . the above dynamics is similar to the generalized kuramoto model dynamics ( [ eom ] ) in the absence of noise @xmath87 . app - simulation here we describe the method to simulate the dynamics ( [ eom - scaled ] ) for given values of @xmath160 ( note that we are dropping overbars for simplicity of notation ) , and for a given realization of @xmath23 s , by employing a numerical integration scheme @xcite . to simulate the dynamics over a time interval @xmath504 $ ] , we first choose a time step size @xmath505 . next , we set @xmath506 as the @xmath275-th time step of the dynamics , where @xmath507 , and @xmath508 . in the numerical scheme , we first discard at every time step the effect of the noise ( i.e. , consider @xmath509 ) , and employ a fourth - order symplectic algorithm to integrate the resulting symplectic part of the dynamics @xcite . following this , we add the effect of noise , and implement an euler - like first - order algorithm to update the dynamical variables . specifically , one step of the scheme from @xmath510 to @xmath511 involves the following updates of the dynamical variables for @xmath512 : for the symplectic part , we have , for @xmath513 , @xmath514 ; \nonumber \\ & & r\big(t_n+\frac{(k-1)\delta t}{4}\big)=\sqrt{r_x^2+r_y^2},\psi\big(t_n+\frac{(k-1)\delta t}{4}\big)=\tan^{-1}\frac{r_y}{r_x } , \nonumber \\ & & r_x=\frac{1}{n}\sum_{j=1}^n \sin\big[\th_j\big(t_n+\frac{(k-1)\delta t}{4}\big)\big],r_y=\frac{1}{n}\sum_{j=1}^n \cos\big[\th_j\big(t_n+\frac{(k-1)\delta t}{4}\big)\big ] , \nonumber \\ \label{formalintegration1 } \\ & & \th_i\big(t_{n}+\frac{k\delta t}{4}\big)=\th_i\big(t_n+\frac{(k-1)\delta t}{4}\big)+a(k)\delta t ~v_i\big(t_n+\frac{k\delta t}{4}\big ) , \label{formalintegration2 } \end{aligned}\ ] ] where the constants @xmath515 s and @xmath516 s are obtained from ref . @xcite . at the end of the updates ( [ formalintegration1 ] ) and ( [ formalintegration2 ] ) , we have the set @xmath517 . next , we include the effect of the stochastic noise by keeping @xmath518 s unchanged , but by updating @xmath519 s as @xmath520+\sqrt{2\delta t\frac{t}{\sqrt{m}}}\delta x(t_{n+1 } ) . \label{formalintegration3}\ ] ] here @xmath521 is a gaussian distributed random number with zero mean and unit variance . app - fast_algo for the models discussed in section [ chap3 ] , the interaction term in the equation of motion for each of the @xmath19 oscillators involves a sum over @xmath19 terms . this would imply at each time step of numerical simulation of the dynamics a computation time that scales as @xmath522 . here we discuss an alternative and efficient numerical algorithm @xcite that transforms the interaction term into a convenient form , allowing for its computation by a fast fourier transform ( fft ) scheme in a time scaling as @xmath523 . use of fft requires that we choose a power of @xmath486 for @xmath19 . let us denote with @xmath524 the sum appearing in the equations of motion ( [ kuramoto - eom_decaying ] ) , ( [ kuramoto - eom_decaying_noise ] ) and ( [ long - range - inertia_dms ] ) : j_i = _ j=1^n , eom1 where @xmath525 is the shortest distance between sites @xmath22 and @xmath25 on a one - dimensional periodic lattice of @xmath19 sites . our simulations results presented in section [ chap3 ] were obtained by considering the lattice constant @xmath360 to be unity . therefore , @xmath525 for @xmath526 is given by d_ij=\ { ll |j - i| ; & , + n-|j - i| ; & , . + [ dij ] while , as explained in the main text , we choose the value of @xmath527 , irrelevant for the equations of motion , equal to @xmath81 . equation ( [ eom1 ] ) may be rewritten as j_i = _ i _ j=1^nv_ij_j - _ i _ j=1^nv_ij _ j . eom2 the first summation may be interpreted as the @xmath22th element of the column vector formed by the product of an @xmath528 matrix @xmath529_{i , j=1,2,\ldots , n}$ ] with the column vector @xmath530 , where @xmath531 . similarly , the second summation may be interpreted as the @xmath22th element of the column vector formed by the product of @xmath532 with the column vector @xmath533 . the matrix @xmath532 has the form v = , with @xmath534 , and v_q=\ { ll 1/(q-1)^ & , + 1/(n - q+1)^ & . . + [ vr ] thus , @xmath532 is a circulant matrix fully specified by the elements in the first column . the remaining columns of @xmath532 are cyclic permutations of the elements in the first column , with offset equal to the column index . note that @xmath532 can be written as v = v_1 i+ v_2 p + v_3 p^2+ +v_np^n-1 , vp where @xmath489 is the cyclic permutation matrix , p = . since @xmath535 , the @xmath528 identity matrix , the eigenvalues of @xmath489 are given by @xmath536 ; the @xmath537s are the @xmath19-th root of unity . equation ( [ vp ] ) then implies that the eigenvalues of @xmath532 are given by @xmath538 for @xmath539 . it is straightforward to check that the eigenvectors of @xmath532 are the columns of the @xmath528 unitary discrete fourier transform matrix @xmath540_{j , k=1,2,\ldots , n}$ ] , where f_jk = e^-i2(j-1)(k-1)/n for 1 j , k n. then , one has @xmath541_{ij}=\lambda_j \delta_{ij}$ ] . in terms of the matrices @xmath542 and @xmath543 , one can rewrite equation ( [ eom2 ] ) as j_i = _ i _ j=1^n ( f^-1)_ij_j ( f ) _ j -_i _ j=1^n ( f^-1)_ij_j ( f ) _ j , eom3 where @xmath544 ( respectively , @xmath545 ) is the @xmath25th element of the column vector formed by multiplying the matrix @xmath542 with the column vector @xmath546 ( respectively , @xmath547 ) . @xmath544 and @xmath545 are just discrete fourier transforms , and may be computed very efficiently by standard fft codes ( see , e.g. , ref . the simulations reported in section [ chap3 ] were performed by using equation ( [ eom3 ] ) .

the phenomenon of spontaneous synchronization , particularly within the framework of the kuramoto model , has been a subject of intense research over the years . the model comprises oscillators with distributed natural frequencies interacting through a mean - field coupling , and serves as a paradigm to study synchronization . in this review , we put forward a general framework in which we discuss in a unified way known results with more recent developments obtained for a generalized kuramoto model that includes inertial effects and noise . we describe the model from a different perspective , highlighting the long - range nature of the interaction between the oscillators , and emphasizing the equilibrium and out - of - equilibrium aspects of its dynamics from a statistical physics point of view . in this review , we first introduce the model and discuss both for the noiseless and noisy dynamics and for unimodal frequency distributions the synchronization transition that occurs in the stationary state . we then introduce the generalized model , and analyze its dynamics using tools from statistical mechanics . in particular , we discuss its synchronization phase diagram for unimodal frequency distributions . next , we describe deviations from the mean - field setting of the kuramoto model . to this end , we consider the generalized kuramoto dynamics on a one - dimensional periodic lattice on the sites of which the oscillators reside and interact with one another with a coupling that decays as an inverse power - law of their separation along the lattice . for two specific cases , namely , in the absence of noise and inertia , and in the case when the natural frequencies are the same for all the oscillators , we discuss how the long - time transition to synchrony is governed by the dynamics of the mean - field mode ( zero fourier mode ) of the spatial distribution of the oscillator phases . keywords : stochastic particle dynamics ( theory ) , stationary states , phase diagrams ( theory )

we describe the details of our method to calculate the density matrix @xmath109 and the probability @xmath19 referred to in the main article . note that we use the notation of the doubled hilbert space , _ i.e. _ the liouville space , where a matrix @xmath110 in the normal hilbert space becomes a vector @xmath111 with the hilbert - schmidt inner product , @xmath112 $ ] @xcite . a linear matrix operation becomes a linear map on the liouville space , _ i.e. _ a superoperator . this description is useful for both the formulation and the numerical calculation in open quantum systems . let us first consider the closed exciton system [ eq . ( 1 ) in the main article ] with no interaction with the ambient environment . in this case , the dynamics of the initial density matrix @xmath113 is fully determined by the system hamiltonian @xmath114 . the time - evolution during the time interval @xmath115 $ ] is given by the unitary superoperator , @xmath116t/\hbar],\end{aligned}\ ] ] where @xmath117 ( @xmath118 ) is the map from any operator @xmath119 in a hilbert space to the superoperator @xmath120 ( @xmath121 ) in the liouville space that makes the operator act to a matrix from left ( right ) : @xmath122 and @xmath123 . the density matrix after the time - evolution @xmath124 is consistent with the trivial results @xmath125)\breve{\mathfrak{r}}(\exp[i\hat{h}t]){\ensuremath{{\boldmath \big|}\hat{\rho}_0\big\rangle\!\!\big\rangle}}\notag\\ & = { \ensuremath{{\boldmath \big|}\exp[-i\hat{h}t]\hat{\rho}_0\exp[i\hat{h}t]\big\rangle\!\!\big\rangle}}. \label{eq : unitaryevolution}\end{aligned}\ ] ] note that in the first line in eq . , we use the commutative property between the superoperators mapped by @xmath117 and @xmath118 , @xmath126 , and their homomorphic and antihomomorphic properties : @xmath127 and @xmath128 , and @xmath129 and @xmath130 . from now on , we shall consider the time - evolution in the open exciton system with photon dissipation to the environment based on the continuous measurement theory of photon counting @xcite . assuming that the environment does not absorb more than one photon during any infinitesimal time interval , we consider here only two fundamental processes : no - absorption and one - photon absorption . the one - photon absorption process during any infinitesimal time interval @xmath131 is represented by the superoperator @xmath132 with @xmath133 being the absorption rate in the presence of a free photon , which converts an @xmath134-photon state to an @xmath135-photon state @xcite . note that @xmath136 is an adequate quantum dynamical map because it is linear and completely positive ( cp ) @xcite . since the normalized pre - absorption state @xmath137 is converted to the non - normalized post - absorption state @xmath138 by the immediate absorption , @xmath139 , the associated absorption probability is given by @xmath140={\ensuremath{\big\langle\!\!\big\langle\hat{i}\big|}}\breve{j}\delta t{\ensuremath{{\boldmath \big|}\hat{\rho}_{-}\big\rangle\!\!\big\rangle}}=\kappa\delta t{\mbox{tr}}[\hat{a}^\dagger\hat{a}\hat{\rho}_{-}]$ ] . next , we consider the no - absorption process . when the superoperator @xmath141 represents the time - evolution under the condition that no photon is absorbed in the interval @xmath115 $ ] , the map @xmath142 should be continuous and homomorphic because the no - absorption process during @xmath115 $ ] must be divided into a sequential process of several no - absorption processes : @xmath143 for all @xmath144 , where @xmath145 is the identity superoperator . these properties are the same as those of the unitary time - evolution @xmath146 . the no - absorption process is , however , different from @xmath146 due to the presence of the environment . the canonical form is given by @xcite @xmath147t/\hbar],\\ \mathcal{h}&\equiv \hat{h}-i\hbar\kappa\hat{a}^\dagger\hat{a}/2.\end{aligned}\ ] ] the anti - hermitian part of @xmath148 ensures a decrease in the associated no - absorption probability in the infinitesimal time interval @xmath131 , @xmath149 $ ] , which compensates for an increase in the photon - absorption probability @xmath150 $ ] . it is noteworthy that @xmath141 is the cp superoperator that satisfies eq . and eq . , and approaches @xmath146 in the limit of @xmath151 the actual time - evolution superoperator @xmath152 is defined such that it represents a time - evolution of a density matrix during the time interval @xmath115 $ ] with no information about the absorption . accordingly , it is given by the sum of all possible combinations of the two fundamental processes : @xmath153 where @xmath154 represents the @xmath16-photon absorption process during the time interval @xmath115 $ ] . the actual time - evolution is also described by the recursive relation , @xmath155 which clearly indicates that the actual time - evolution consists of the immediate one - photon absorption process and the no - absorption process . in addition , @xmath152 satisfies the following equation - of - motion ( eom ) : @xmath156\breve{w}(t ) . \label{eq : timeevolution_w}\end{aligned}\ ] ] where the liouvillians are defined by @xmath157/\hbar+\kappa\breve{\mathfrak{l}}(\hat{a})\breve{\mathfrak{r}}(\hat{a}^\dagger)-\kappa[\breve{\mathfrak{l}}(\hat{a}^{\dagger}\hat{a})+\breve{\mathfrak{r}}(\hat{a}^{\dagger}\hat{a})]/2 $ ] , @xmath158/\hbar$ ] , and @xmath159/\hbar$ ] , respectively . here @xmath160 is defined in eq . ( 1 ) of the main article . accordingly , we can obtain the density matrix after the actual time - evolution @xmath161 by solving the eom . note that the eom has a lindblad form @xcite and @xmath162 is normalized as @xmath163={\ensuremath{\big\langle\!\!\big\langle\hat{i}\big|\hat{\rho}(t)\big\rangle\!\!\big\rangle}}={\ensuremath{\big\langle\!\!\big\langle\hat{i}\big|}}\breve{w}(t){\ensuremath{{\boldmath \big|}\hat{\rho}_0\big\rangle\!\!\big\rangle}}={\mbox{tr}}[\hat{\rho}_0]$ ] . it is noted that if there is a photon decoupled from the excitons at @xmath18 , the population decays exponentially with the lifetime of @xmath97 , which is calculated analytically @xcite . let us turn to the calculation of probability @xmath19 that @xmath16-photons are absorbed during the time interval @xmath115 $ ] . from the above discussion , we have already obtained the superoperator that represents the @xmath16-photon absorption process during @xmath115 $ ] @xmath154 : @xmath164 although it is difficult to calculate the probability analytically due to the phonon interaction , we are able to numerically calculate it with an approximation based on the moment generating function ( mgf ) of the probability , @xmath165 where we define the generalized density matrix by @xmath166 with @xmath167 being the time - evolution superoperator . the eom for @xmath168 is therefore given by the following equation , @xmath169{\ensuremath{{\boldmath \big|}\hat{\rho}(t;\lambda)\big\rangle\!\!\big\rangle } } , \label{eq : timeevolution_gdm}\end{aligned}\ ] ] with @xmath170 . the boundary condition is given by the actual density matrix at @xmath18 : @xmath171 . here , the initial density matrix @xmath113 is assumed to be the tensor product state of a bright exciton state and the phonon bath in the thermal equilibrium state , which reasonably describes the state just after the exciton generation in accordance with the franck - condon principle : @xmath172 where @xmath173 and @xmath174/{\mbox{tr}}_\mathrm{b}[\exp[-\hat{h}_\mathrm{b}/k_{b}t]]$ ] . here , @xmath33 is the partial trace with respect to the subspace of the phonon bath , @xmath175 is the boltzmann constant , and @xmath29 is the temperature of the bath . equation is equivalent to eq . ( 2 ) in the main article , which is written in the ordinary hilbert space . because the actual time - evolution of the density matrix is given by @xmath176 [ compare eq . and eq . ] , we hereafter focus on the calculation of the eom . the probability @xmath19 is given by the inverse fourier transformation of the mgf . in the main article , we assume that the coupling between the phonons and the excitons is weak . hence , we include the phonon effects on the exciton - photon system within the second order perturbation in terms of the interaction liouvillian @xmath177 . the phonon bath is also assumed to remain in the thermal equilibrium at all time . because the time - evolution equation has a form very similar to that for the standard density matrix , we make use of an established method developed in the quantum master equation ( qme ) . then , we derive a closed eom for the generalized reduced density matrix in the exciton - photon subspace @xmath178 defined by @xmath179 where @xmath180 means the partial trace of the phonon bath subspace @xmath181\big\rangle\!\!\big\rangle}}_{\rm s}={\ensuremath{{\boldmath \big|}{\mbox{tr}}_{\rm b}[\hat{\rho}]\big\rangle\!\!\big\rangle}}_{\rm s}$ ] . in accordance with the standard second - order time - convolutionless projection operator method @xcite , we start with the density matrix in the interaction picture , @xmath182{\ensuremath{{\boldmath \big|}\hat{\rho}(t;\lambda)\big\rangle\!\!\big\rangle}}\end{aligned}\ ] ] because @xmath183 satisfies the following eom , @xmath184 with @xmath185\breve{\mathcal{l}}_\textrm{int}(t)\exp[(\breve{\mathcal{l}}_\mathrm{s}(\lambda)+\breve{\mathcal{l}}_\mathrm{b})t]$ ] , the formal solution is given by @xmath186{\ensuremath{{\boldmath \big|}\hat{\rho}_0\big\rangle\!\!\big\rangle}}.\end{aligned}\ ] ] introducing the explicit coupling constant @xmath187 to the liouvillian , @xmath188 , we expand the above formal solution within the second order @xcite . then , we obtain the following eom for the generalized reduced density matrix in the interaction picture @xmath189 , @xmath190\breve{\upsilon}(t;\lambda)\exp[\breve{\mathcal{l}}_\mathrm{s}(\lambda)t]{\ensuremath{{\boldmath \big|}\hat{\rho}_{\textrm{i}}^{\rm s}(t;\lambda)\big\rangle\!\!\big\rangle}}_{\rm s},\end{aligned}\ ] ] where we define @xmath191.\end{aligned}\ ] ] here , @xmath192 and @xmath193\breve{b}_{r}^{\pm}\exp[-\breve{\mathcal{l}}_{\mathrm{s}}(\lambda)t]$ ] . the influence of phonons on the excitons is represented by the spectral density @xmath194 for @xmath35 , which is here assumed to have an ohmic form ; @xmath195 where we introduce a cutoff frequency @xmath39 and the effective coupling strength @xmath196 . converting the equation in the interaction picture to that in the schrdinger picture , we obtain the generalized qme for numerical calculation : @xmath197{\ensuremath{{\boldmath \big|}\hat{\rho}^{\rm s}(t;\lambda)\big\rangle\!\!\big\rangle}}_{\rm s}. \label{eq : timeevolution_rgdm}\end{aligned}\ ] ] the initial condition is given by @xmath198 . with using @xmath199 , the mgf and the actual reduced density matrix of the exciton - photon subsystem @xmath200 are calculated as @xmath201 we solve the generalized qme numerically . in the liouville space where a density matrix is vectorized and superoperators can be represented by matrices , the qme becomes just a system of linear differential equations . hence , we use ordinary methods : fourth - order runge - kutta ( rk4 ) method and eigen - decomposition ( ed ) . at the initial time , @xmath202 strongly depends on the time @xmath48 . therefore , we solve the equation by the rk4 method until @xmath202 becomes time - independent . we take the unit time step typically as @xmath203 ps . after sufficiently long time , @xmath202 hardly changes in time and becomes a nearly time - independent superoperator , which is known as the markovian limit . in the present calculations , the convergent time @xmath204 is typically tens of ps and is shorter than @xmath205 ps . after the convergence , we carry out the ed of the generator @xmath206 : @xmath207 where @xmath208 is the matrix composed of the eigenvectors , and @xmath209 is the diagonal matrix constructed from the corresponding eigenvectors . the generalized reduced density matrix @xmath210 at any time after @xmath204 is directly given by @xmath211\breve{p}^{-1}(\lambda){\ensuremath{{\boldmath \big|}\hat{\rho}^{\rm s}(t_\mathrm{c};\lambda)\big\rangle\!\!\big\rangle}}.\end{aligned}\ ] ] the method combining rk4 and ed works efficiently for the long - term dynamics calculation over several orders of time . we have confirmed that the results calculated from the combined method are consistent with those obtained only from the rk4 method in the interval @xmath212 $ ] .

energy dissipation and decoherence are at first glance harmful to acquiring long exciton lifetime desired for efficient photovoltaics . in the presence of both optically forbidden ( namely , dark ) and allowed ( bright ) excitons , however , they can be instrumental as suggested in photosynthesis . by simulating quantum dynamics of exciton relaxations , we show that the optimized decoherence that imposes a quantum - to - classical crossover with the dissipation realizes a dramatically longer lifetime . in an example of carbon nanotube , the exciton lifetime increases by nearly two orders of magnitude when the crossover triggers stable high population in the dark exciton . sunlight is a clean , abundant , and sustainable energy source . hence , the effective energy conversion from sunlight into electricity is a grand challenge in science and technology , which leads to an emerging interest on the post - silicon photovoltaic materials such as carbon nanotubes , quantum dots , transition metal dichalcogenides for their promising applications @xcite . the energy conversion consists of three processes of nonequilibrium quantum dynamics of excitons ( bound electron - hole pairs ) : exciton generation from a photon , exciton energy transfer , and charge separation of the exciton into electrodes . in the first two processes , the efficient conversion requires both a high photon absorption rate _ and long exciton lifetime_. however , the optimization is hampered by the reversibility between absorption and emission of a photon in the quantum dynamics : while a high absorption rate of photon is desired in the exciton generation process , it also leads to a high charge recombination rate of exciton . in this work , we show that a desired remarkable enhancement of the exciton lifetime by simultaneously keeping high absorption rate is achieved by utilizing nonequilibrium energy dissipation and decoherence by phonons . though the dissipation is in general an obstacle to high efficiency of energy conversion , there are cases where nonequilibrium dissipations and decoherence are actually instrumental in achieving long exciton lifetime because they can suppress the exciton recombination by making quantum dynamics irreversible via concomitant quantum - to - classical crossover @xcite . in fact , irreversible exciton dynamics is exploited in photosynthesis that also includes the above mentioned three processes , namely , the exciton generation , the exciton energy transfer , and the charge separation @xcite . photosynthesis achieves a remarkably high quantum efficiency reaching nearly 100% @xcite , which means that an absorbed photon is converted to an exciton with no recombination . in the exciton generation process , the absorbed energy is irreversibly transferred from the optically allowed bright exciton to an optically forbidden dark exciton @xcite , which can act as a ratchet between the exciton generation process and the next energy transfer process where the quantum coherence plays a role again @xcite . recently , photovoltaic models inspired by photosynthesis were studied from the viewpoint of steady - state heat engines modified with discrete exciton states @xcite . a mechanism for the enhancement of the photocurrent was proposed by designing the ultrafast classical transition from bright to dark excitons within a hundred femtoseconds , where quantum coherence between photons and excitons and the resultant photoluminescence causing the recombination were also ignored @xcite . however , these classical descriptions of the photocell dynamics make the validity of the enhancement questionable because the inherent dichotomy between quantum coherence and decoherence is a crucial issue in enhancing the quantum efficiency in the target time domain : on the one hand the desired ultrafast transition from the bright to the dark excitons necessarily requires quantum coherent dynamics , whereas on the other hand , the decoherence that imposes a quantum - to - classical crossover is assumed to immediately occur . for the microscopic understanding of optimal energy conversions , here we study the transient dynamics for the exciton generation process , taking account of the dichotomy in a unified quantum manner . to this end , we construct and investigate an open quantum model that consists of bright and dark excitons coupled with phonons and a dissipative photon , with using the combined method of the generalized quantum master equation ( gqme ) @xcite and the quantum continuous measurement theory of photon counting @xcite . we found that realistic coexistence of coherence and decoherence is the key for high quantum efficiency of photovoltaics . the crossover from quantum to classical dynamics due to dissipation and decoherence by phonons assists the long exciton lifetime : rapid quantum energy transfer from the bright to dark excitons suppresses the initial radiative loss , whereas the dark excitons become stable through the concomitant crossover . our results indeed reveal why high efficiency of exciton generation from photons can be compatible with the low photoluminescence by violating the reversibility as observed experimentally . . the photons are dissipated to environment with the rate of @xmath0 . the bright exciton is coupled with two dark excitons via a phononcontinuum that causes energy dissipation and decoherence.,width=325 ] by taking a typical semiconducting single - walled carbon nanotube ( swcnt ) with the chirality ( 6,5 ) , as a model material , we demonstrate that the exciton lifetime becomes nearly two - orders of magnitude longer than the case without dark excitons . moreover , our results are consistent with the experimental indications @xcite . in the swcnt , at least two dark exciton states exist with energy below the lowest bright one @xcite : even - parity dark exciton @xcite and spin - triplet dark exciton @xcite . the weak couplings between the bright and the dark excitons can strongly affect the exciton dynamics after the photon absorption . hence , we construct a hamiltonian consisting of one bright and two dark exciton states coupled to a phononcontinuum and a single photon , as a minimal model of the swcnt ( see also fig . [ fig1 ] ) . the model hamiltonian is given by @xmath1 the first term is the hamiltonian of the exciton - photon system , @xmath2 with @xmath3 , @xmath4 , and @xmath5 being the bosonic creation operators for the photon , the bright exciton , and the two dark excitons ( dark-1 and dark-2 ) , respectively , where the dark-1 ( dark-2 ) exciton has even - parity ( triplet ) symmetry . the bright exciton is coupled to the photon with the dipole coupling strength @xmath6 . the second term represents the phononic bath , @xmath7 where @xmath8 reads the bosonic creation operator of a phonon with a momentum @xmath9 . the energy of the photon , the excitons , and the phonons are represented by @xmath10 , @xmath11 , and @xmath12 , respectively . the last term indicates the phonon - mediated coupling between the bright and dark excitons , @xmath13 with the effective coupling strength @xmath14 . simulating radiative lifetime of excitons requires non - unitary dissipations of the photon to environment or by photo detections . we simulate the non - unitary events by using quantum continuous measurement theory of photon counting with the photon dissipation rate , @xmath15 @xcite . the photon counting gives the probability that @xmath16 photons are absorbed to environments ( or disappear at the instrument ) until a time @xmath17 after the absorption of a photon ( or the creation of the bright exciton ) at @xmath18 : @xmath19 . for feasible simulations of the excitons with the photon counting , we use the moment generating function ( mgf ) of the probability @xmath20\mathcal{p}(m;t)$ ] , with the conjugate variable @xmath21 to the absorbed photon number @xmath16 . the mgf dynamics can be explicitly determined by the following equation of motion for a generalized density matrix with @xmath21 : @xmath22/\hbar+\kappa\mathrm{e}^{i\lambda } \hat{a}^{}\hat{\rho}(t;\lambda)\hat{a}^\dagger\notag\\ & \quad -\kappa[\hat{a}^\dagger\hat{a}^{}\hat{\rho}(t;\lambda)+\hat{\rho}(t;\lambda)\hat{a}^\dagger\hat{a}^{}],\end{aligned}\ ] ] where the mgf is given by the trace of @xmath23 as @xmath24 $ ] for @xmath17 with the initial boundary condition , @xmath25 . here , @xmath26 is the initial density matrix just after the photon absorption . applying the franck - condon principle to the initial state at @xmath18 , the initial state is reasonably given by a separable state of the form @xmath27 $ ] where @xmath28 is the initial density matrix of the exciton - photon system with one bright exciton and @xmath29 is the temperature of the phononic bath . from now on , we assume that the coupling strength @xmath30 is weak and take the influence into account by the standard second - order perturbation approximation in terms of @xmath31 @xcite . in this approximation , the exciton dynamics is adequately described by the gqme for the generalized reduced density matrix ( rdm ) , @xmath32 $ ] where @xmath33 means the partial trace of the phononic bath , and the influence of the phonons on the exciton system is represented by the spectral density @xmath34 for @xmath35 . here , we assume that @xmath36 has the standard ohmic form @xcite , @xmath37 where @xmath38 and @xmath39 denote the effective coupling strength between the exciton and phonons and the energy cutoff of the spectral density , respectively . solving the gqme numerically , we calculate the generalized rdm that gives the mgf as @xmath40 $ ] . the probability is obtained from the inverse fourier transform of the mgf . note that the generalized rdm @xmath41 coincides with the actual density matrix of the reduced system where we ignore the counting history , @xmath42 @xcite . hereafter , we omit the script `` s '' representing the subspace of the exciton - photon system such as @xmath43 for simplicity . although the procedure is straightforward , the details of the formalism of the gqme are described in the supplemental material @xcite . actual time - resolved photoluminescence experiments have time - resolution limit @xmath44 , which is introduced into our theory phenomenologically by the gaussian averaging : @xmath45 d\tau$ ] . from the time - averaged probability with @xmath46 ps , the time - resolved photoluminescence , @xmath47 , and the quantum yield at a time @xmath48 , @xmath49 , are given by @xmath50 where @xmath51 is the initial number of the bright exciton at @xmath18 . note that @xmath49 represents the maximum quantum yield of photoluminescence experiments because the other environmental - photon - absorption is also included in @xmath49 . for a model calculation of ( 6,5 ) swcnts at room temperature , we choose the value of the parameters as follows : @xmath52 ev , @xmath53 ev , @xmath54 ev , @xmath55 k , @xmath56 ev , @xmath57 mev , @xmath58 mev , and @xmath59 mev . effects of the environment are described by the single parameter , namely , the decay rate of photon , @xmath0 . since @xmath0 strongly depends on the ambient solvents , matrices , and/or substrates , we vary @xmath0 over several orders of magnitude with keeping @xmath0 as a small parameter . the energy levels of bright and dark excitons are estimated from the photoluminescence experiments @xcite . the cutoff frequency of the spectral density @xmath60 is determined from the density of states of phonons in swcnts @xcite . the coupling strength @xmath61 is determined from the numerical fitting that reproduces the experiment @xcite ( see fig . [ fig3](c ) ) . the dipole coupling strength @xmath6 is taken to be @xmath62 . note that one confirms that one order of magnitude difference in @xmath6 does not affect our main results . and quantum yield @xmath49 . the parameter of radiative dissipation to environment is fixed at @xmath63 ns . the solid red line indicates the tri - exponential fitting of the numerical result , @xmath64 . the time - dependent energy @xmath65 is also plotted in the same figure . ( b ) population dynamics of the exciton - photon system with @xmath66 $ ] and @xmath67 $ ] . ( c ) time evolution of von neumann entropy @xmath68 and the quantum mutual information @xmath69 . the shadow region indicates the interval between the two peaks of the von neumann entropy.,width=325 ] first , we show the time - resolved photoluminescence @xmath47 for @xmath63 ns in fig . [ fig2](a ) , which may be a typical result for swcnts in aqueous solutions . the simulated @xmath47 is accurately fitted by a tri - exponential function : @xmath70 with @xmath71 . we note that similar multi - exponential curves are also observed in the experiments @xcite . the decay constants obtained in our calculation differ by several orders of magnitude : the fast decay is characterized by @xmath72 @xmath73 65 ps while the intermediate and slow decay constants , @xmath74 and @xmath75 , are found at 890 ps and 1 @xmath76s , respectively . while the luminescence rapidly decreases within 1 ns as shown in fig . [ fig2](a ) , the energy of the exciton - photon system @xmath77 $ ] remains over 80% of @xmath78 even at 100 ns with the plateau from 5ns to 50 ns where the quantum yield holds lower than 10% . the significant difference in the decays of energy and luminescence qualitatively explains both the seemingly contradictory experiments of the time - resolved photoluminescence @xcite and the pump - probe transient absorption spectroscopy @xcite : some time - resolved photoluminescence spectroscopies show the rapid luminescence decay with tens of ps and a low quantum yield @xcite whereas a dark exciton survives over 10 @xmath76s in the pump - probe transient absorption spectroscopy @xcite . exciton lifetime is a crucial factor characterizing the dynamics . here , we define the energy lifetime of excitons @xmath79 as the energy decay time to @xmath80 of the initial value for consistency with the single - exponential decay model : @xmath81 from the definition , we obtain @xmath82 ns , which is longer than forty times of that in the system with no dark exciton given by the inverse of the rate of the photon absorption to the environment @xmath83 ns as detailed later . here , the factor `` two '' in @xmath84 results from the halved residence time of the photon state by the rabi oscillation between the photon and the bright exciton @xcite . the enhanced energy lifetime of the exciton - photon system is attributed to the fast irreversible relaxation pathway from the bright exciton to the dark excitons accompanied by quantum - to - classical crossover . we see the relaxation in the population dynamics in fig . [ fig2](b ) . the initial dynamics has a quantum nature where the populations of the photon and the bright exciton are oscillating with the anti - phase with the frequency of @xmath85 , which clearly indicates the rabi oscillation . then , the population is gradually transferred to the dark excitons after 10 ps with a reduction of the oscillation due to decoherence effects by phonons . in particular , the dark-2 exciton is stabilized from a few ns to 100 ns with a high population via the quantum - to - classical crossover . the irreversible crossover is confirmed by using the time - dependence of the von neumann entropy and the quantum mutual information as shown in fig . [ fig2](c ) . the von neumann entropy , @xmath86 $ ] , initially remains small , which means that the initial dynamics is dominated by the time - evolution of a quantum pure state . then , the von neumann entropy increases due to the decoherence by phonons , and shows two peaks approximately at 300 ps and 600 ns with the strong mixing of states . these mixings , however , have different origins , _ i.e. _ the correlation between particles including quantum entanglement at @xmath87 ps and the classical stochastic mixing of the state at @xmath88 ns . to distinguish these two distinct origins of the mixing , we introduce quantum mutual information : the quantum mutual information provides a measure of correlation between subsystems of quantum states . here , we calculate the quantum mutual information between the subsystem x consisting of the photon and the bright exciton and the subsystem y of the dark excitons , @xmath89 . here @xmath90 $ ] is the quantum relative entropy that gives a measure of difference between two quantum states , and @xmath91 and @xmath92 are the rdms defined by the partial trace of the subsystem of y and x , respectively : @xmath93 $ ] and @xmath94 $ ] . while the correlation between the subsystems is enhanced approximately at the first peak of @xmath95 , @xmath96300 ps , the quantum mutual information gives no peak at the second peak of @xmath95 , which implies that the system becomes a nearly separable state with essentially no quantum entanglement after the first peak as shown in fig.[fig2](c ) . hence , the crossover occurs near the first peak , which leads to the superselection of dark exciton states imposed by the decoherence @xcite . ultimately , the fast relaxation from the bright exciton to the dark excitons and the rapid quantum - to - classical crossover do not contradict each other in the time domain . . the solid red lines indicate the tri - exponential fitting functions on the numerical results . ( b ) @xmath0-dependence of decay constants in tri - exponential fitting function . ( c ) correlation between @xmath72 and @xmath74 for several choices of @xmath97 , which ranges from 10 ps to 1 @xmath76s ( black filled circle ) . the filled square ( red ) symbols indicate the experimentally measured decay constants for ( 6,5 ) swcnt reproduced from ref . @xcite.,width=325 ] next , we examine the environmental effects on the time - resolved photoluminescence by monitoring @xmath0 as shown in fig . [ fig3](a ) . the photoluminescence decays are well fit by a tri - exponential function irrespective of @xmath0 though the decay constants strongly depend on @xmath0 . the two decay constants @xmath72 and @xmath74 for the fast and intermediate decays , respectively , are saturated with an increase in @xmath97 as shown in fig . [ fig3](b ) . besides , the largest decay constant @xmath75 shows no saturation and proportional to @xmath98 with increasing @xmath97 . these decay constants share the common origin : the positive correlation between @xmath72 and @xmath74 is clearly seen in fig . [ fig3](c ) . the convex curve is consistent with the results of the experiment @xcite . as a function of time and @xmath0 for the ( 6,5 ) swcnt model ( dissipative system ) , and for the hypothetical model with the same parameters as those in the dissipative system except for the energy levels , @xmath99 ( non - dissipative system ) . solid lines indicate the contours of the normalized energy . ( c ) energy lifetime @xmath79 as a function of @xmath0 in the dissipative and non - dissipative systems . the lifetime is normalized by @xmath100 that is the lifetime in the system with no dark exciton . ( d ) normalized lifetime in the dissipative system as a function of the imaginary part of the dielectric function at @xmath101 of the ambient dielectric medium , @xmath102=\kappa/\omega_\mathrm{ph}$].,width=325 ] finally , we emphasize the importance of the energy dissipation for the enhancement of the exciton lifetime . figures [ fig4](a ) and ( b ) show the system energy @xmath65 as a function of time and @xmath0 for the systems where ( a ) the dark excitons have the lower energy compared to the bright one ( dissipative system ) or ( b ) all the excitons have the same energy level ( non - dissipative system ) . the dissipative system obviously survives longer than the non - dissipative system for small @xmath0 where the irreversible relaxation becomes dominant at the initial dynamics for @xmath103100 ps . the difference is quantified in the energy lifetime @xmath79 as shown in fig . [ fig4](c ) . although the lifetime of the non - dissipative system is enhanced compared to that in the system with no dark exciton ( no - dark system ) @xmath104 , the enhancement factor is not large even for large @xmath97 : the factor is about two . while , the lifetime in the dissipative system is about 45 times as long as @xmath105 for large @xmath97 . the large enhancement of the lifetime suddenly occurs within a small range of @xmath0 . when all the energy loss to the environment can be represented by the dielectric loss of the ambient homogeneous linear dielectric medium , @xmath0 is related with the background dielectric function @xmath106 with @xmath107 being a frequency by using the poyinting s theorem in the classical electrodynamics @xcite : @xmath108 $ ] . in the case , the sudden enhancement of lifetime is estimated to occur at @xmath102\simeq 2 \times 10^{-6}$ ] as shown in the inset of fig . [ fig4](c ) , which is comparable to that of water in room temperature , @xmath102\simeq1\times10^{-5}$ ] @xcite . therefore , the exciton lifetime of the swcnt placed in the medium with smaller @xmath102 $ ] , _ e.g. _ air , can be strongly enhanced . in summary , we have investigated the effect of dissipation and decoherence on the long - term exciton dynamics in a model of the semiconducting ( 6,5 ) carbon nanotube . counterintuitively , the dissipation and decoherence enhance the exciton lifetime in the relaxation pathway through the dark excitons , which paradoxically appears in experiments as a rapid decay of photoluminescence and a slow growth of quantum yield . in the search for efficient solar cells , such bad radiators are at first glance unpromising . however , we have shown this is not necessarily true . our study confirms that semiconducting carbon nanotubes become potentially efficient photovoltaic materials as suggested in the experiments @xcite , and provides a further guideline and insight into high potentials of other unexplored materials . we acknowledge the financial supports by a grant - in - aid for scientific research ( no . 22104010 and no . 22340090 ) from mext , japan , and a grant - in - aid for scientific research on innovative areas ` materials design through computics : complex correlation and non - equilibrium dynamics ' . this work was also supported by mext hpci strategic programs for innovative research ( spire ) ( under the grant number hp130007 and hp140215 ) and computational materials science initiative ( cmsi ) . 44 natexlab#1#1bibnamefont#1#1bibfnamefont#1#1citenamefont#1#1url#1`#1`urlprefix[2]#2 [ 2][]#2 , , , , * * , ( ) . , , , , , * * , ( ) . , , , , , * * , ( ) . , * * , ( ) . , , , , * * , ( ) . , , , , , , , , , , , * * , ( ) . , * * , ( ) . , , , , , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , , , , , , * * , ( ) . , , , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , , , , * * , ( ) . , , , , * * , ( ) . , , , , * * , ( ) . , _ _ ( , , ) . , * * , ( ) . , * * , ( ) . , , , * * , ( ) . , , , , * * , ( ) . , , , * * , ( ) . , , , , , , * * , ( ) . , * * , ( ) . , , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , , , * * , ( ) . , , , , , , , , * * , ( ) . , , , * * , ( ) . , , , , , , , , , * * , ( ) . , , , * * , ( ) . , , , , * * , ( ) . , * * , ( ) . , , , , , , , * * , ( ) . , * * , ( ) . , , , , , , , , * * , ( ) . , , , , * * , ( ) . , , , , , , * * , ( ) . , _ _ ( , , ) . , * * , ( ) .

the effects of quantum electrodynamics ( qed ) may occur in a strong laser pulse interaction with a counterpropagating electron beam . in the well - known experiment @xcite these effects were weak and barely observable . if the laser pulse intensity is increased up to @xmath0 the qed effects control the laser - beam interaction and result in multiple pair production from a single beam electron . * qed strong fields . * in qed an electric field , @xmath2 , should be treated as strong if it exceeds the schwinger limit : @xmath3 ( see @xcite ) . such field is potentially capable of separating a virtual electron - positron pair providing an energy , which exceeds the electron rest mass energy , @xmath4 , to a charge , @xmath5 , over an acceleration length as small as the compton wavelength , @xmath6 typical effects in qed strong fields are : electron - positron pair creation from high - energy photons , high - energy photon emission from electrons or positrons and the cascade development ( see @xcite- @xcite ) resulting from the first two processes . less typical is direct pair separation from vacuum . this effect may only be significant if the field invariants as defined in @xcite , @xmath7 , @xmath8 , are large enough . here the case of _ weak _ field invariants is considered : @xmath9 , and any corrections of the order of @xmath10 are neglected ( see @xcite about such corrections ) . below , the term strong field is only applied to the field _ experienced by a charged particle_. * qed - strong laser fields . * qed - strong fields may be created in the focus of an ultra - bright laser . consider qed - effects in a relativistically strong pulsed field @xcite : @xmath11 being the vector potential of the wave . in the laboratory frame of reference the electric field is not qed - strong for achieved laser intensities , @xmath12 @xcite , and even for the @xmath13 intensity projected @xcite . moreover , both field invariants vanish for 1d waves , reducing the probability of direct pair creation from vacuum by virtue of the laser field s proximity to 1d wave . nonetheless , a counterpropagating particle in a 1d wave , @xmath14 , may experience a qed - strong field , @xmath15 , because the laser frequency , @xmath16 , is doppler upshifted in the frame of reference co - moving with the electron . herewith the electron dimensionless energy , @xmath17 , and its momentum are related to @xmath4 , and @xmath18 correspondingly , and subscript @xmath19 herewith denotes the vector projection on the direction of the wave propagation . the lorentz - transformed field exceeds the schwinger limit , if @xmath20 . numerical values of the parameter , @xmath21 , may be expressed in terms of the local instantaneous intensity of the laser wave , @xmath22 : @xmath23}}.\ ] ] in the slac experiment @xcite an electron beam of energy @xmath24 gev interacted with a counterpropagating terawatt laser pulse of intensity @xmath25 ( @xmath26 ) . a value of @xmath27 had been achieved . an increase in the laser field intensity up to @xmath28 ( @xmath29 ) with the use of the same electron beam , would allow us to reach a regime of multiple pair creation at @xmath30 . * radiation back - reaction . * the creation of pairs in qed - strong fields is a particular form of the radiation losses from charged particles . at high @xmath21 an intermediate stage in the pair creation process is the emanation of a high - energy photon by a charged particle : @xmath31 ( in contrast with @xmath32 case , in which the `` equivalent '' photons from the electron coulomb potential mostly contribute to the pair creation - see @xcite and the papers cited therein ) . this photon is then absorbed in the strong field , generating an electron - positron pair : @xmath33 . although the energy - momentum _ gain _ from the strong laser field is crucial in the course of both emission and pair creation , still a way to quantify the irreversible radiation _ losses _ may be found . specifically , in the 1d wave field the transfer of energy , @xmath34 , from the wave to a particle may be interpreted as the absorption of some number of photons , @xmath35 : @xmath36 . accordingly , the momentum from the absorbed photons is added to the parallel momentum of the particle : @xmath37 . so , both energy and parallel momentum are not conserved , however , their difference is : @xmath38 . to get the lorentz - invariant formulation , introduce the four - vector of the particle momentum , @xmath39 , and the wave four - vector , @xmath40 for the 1d wave field . their four - dot - product , @xmath41 , is conserved in any particle interaction with the 1d wave field , including its motion , photon emission , pair creation etc . the sum of this quantity , @xmath42 , over all particles in the final ( _ f _ ) state is equal to that for the particles in the initial ( _ i _ ) state : @xmath43 . each term in this conserving sum is positive ( we use the metric signature @xmath44 ) . therefore , any contribution to this sum from a newborn particle exacts a contribution from its parent . regarding the high - energy electron beam interaction with the ultra - strong laser pulse , the initially high value of @xmath30 ensures multiple pair creation . the radiation back - reaction , however , splits the initially high value of @xmath45 between all newborn particles . the reduced values of @xmath45 result in smaller values of @xmath46 . the cascade terminates , when all particles have @xmath32 and become incapable of creating new pairs . the radiation losses , thereby limit the cascading pair creation . particularly , emission of softer @xmath1 photons even may be described within the radiation force approximation , which is traditionally used to account for the radiation back - reaction ( see @xcite,@xcite,@xcite,@xcite,@xcite ) . the discussed processes are described by the kinetic equations for the involved particles ( electrons , positrons , @xmath1-photons ) . for circularly polarized 1d wave of constant amplitude , the system of three 1d integro - differential kinetic equations is reducible to a large system of odes , which is solved here numerically . the emission probability in the qed - strong 1d wave field may be found in sections 40,90,101 in @xcite , as well as in @xcite . however , to simulate highly dynamical effects in pulsed fields , one needs a reformulated emission probability , related to short time intervals ( not @xmath47 ) , which is rederived in appendix a with careful attention to consistent problem formulation . again , the energy , @xmath48 , and momentum , @xmath49 , of the emitted photon are normalized to @xmath4 and @xmath18 . the four - dot - product , @xmath45 , is the motional invariant for an electron and it is also conserved in the process of emission : @xmath50 . a subscript @xmath51 denotes the electron in the initial ( _ i _ ) or final ( _ f _ ) state . in the 1d wave field the emission probability may be conveniently related to the interval of the wave phase , @xmath52 , which should be taken along the electron trajectory . the interval of time , @xmath53 , and that of the electron proper time , @xmath54 , are related to @xmath52 as follows : @xmath55 $ ] . the phase volume element for the emitted photon is chosen in the form @xmath56 . the emission probability , @xmath57 , is integrated over @xmath58 , therefore , it is related to the element of the phase volume , @xmath59 ( see detail in appendix a ) : @xmath60 @xmath61 here @xmath62 is the macdonald function and @xmath63 . * collision integral . * in qed - strong fields we introduce @xmath21-parameter not only for electrons but also for @xmath1-photons and relate the emission probability to @xmath64 : @xmath65 @xmath66,\ ] ] here @xmath67 $ ] , @xmath68 . the electron parameter , @xmath69 , is taken for the initial state and its value in the final state is @xmath70 . the distribution functions for electrons and photons may be also integrated over @xmath71 and @xmath72 correspondingly . thus integrated functions are distributed over @xmath45 , @xmath73 . we can parameterize _ locally _ these distributions via @xmath74 , @xmath75 and introduce the 1d distribution functions , @xmath76 and @xmath77 . the collision integral ( see @xcite ) describes the change in the particle distributions due to emission and accounts for the electrons , leaving the given phase volume , @xmath78 , and those arriving into it within the interval , @xmath79 : @xmath80 @xmath81 * radiation force approximation*. one may exclude the emission of softer @xmath1-photons with @xmath82 from the collision integral by changing the spans as follows : @xmath83 @xmath84 the excluded process should be treated separately : @xmath85,\ ] ] @xmath86 where the expression for the _ radiation force _ , @xmath87 is obtained using the standard fokker - planck development ( see @xcite ) of the collision integral at small @xmath88 : @xmath89 the advective term like that in eq.([eq : radforce ] ) , once introduced to the kinetic equation , describes the electron transport along the characteristic lines , @xmath90 . this effect is equivalent to that from an extra four - force term , @xmath91 , in the dynamical equation for the electron four - momentum , @xmath92 , the force being such that : @xmath93 the radiation force is directed along @xmath94 . the two terms describe the energy - momentum lost for radiation and those absorbed from the 1d wave field in the course of emission , their total being orthogonal to @xmath92 ( see @xcite,@xcite ) . the force magnitude may be found from ( [ eq : rfproject ] ) : @xmath95 in the first component of this equation the term @xmath96 controls the radiation energy loss rate , @xmath97 . in dimensional form and related per time interval , @xmath98 . at @xmath99 , @xmath97 tends to the expression for the radiation energy loss rate found within the framework of classical electrodynamics . when the radiation force approach is generalized for large @xmath100 , the emission spectrum is modified by the qed effects and only the part of this spectrum ( which is minor at @xmath100 ) is embraced by the radiation force approximation . the absorption probability for photons in the 1d field is derived in appendix b. an electron - positron pair ( _ e , p _ ) is generated in the photon absorption with the conservation law : @xmath101 . the phase volume element for the created electron , again is chosen in the form @xmath102 . the absorption probability , @xmath103 , is integrated over the transversal momenta components and related to the element of the phase volume of electron , @xmath104 , resulting in the following collision integral : @xmath105 @xmath106 here @xmath107 $ ] , @xmath108 and @xmath109.\ ] ] as long as the distribution functions are integrated over the transversal components of momentum and expressed in terms of the motional integrals , @xmath110 , their evolution is controlled by the collision integrals : @xmath111 the derivatives , @xmath112 , are taken at constant @xmath45 . eqs.([eq : tosolve ] ) are easy - to - solve for the 1d wave field of any shape , however , for circularly polarized wave of constant amplitude the solution is especially simple . in this case @xmath45 are different from @xmath21 by a constant factor , and eqs.([eq : tosolve ] ) may be solved with derivatives , @xmath113 , at constant @xmath21 for the functions , @xmath114 . we solve eqs.([eq : tosolve ] ) numerically , by discretizing them at a uniform grid , @xmath115 , @xmath116 , with the choice of @xmath117 , @xmath118 . the @xmath119-dependent distribution functions at this grid obey the system of @xmath120 odes , which is integrated numerically . at initialization , electrons with @xmath121 , @xmath122 , counterpropagate in the circularly polarized wave field with @xmath123 . this choice corresponds to the slac electron beam and the laser intensity of @xmath124 for @xmath125 m , to be achieved soon . , and a spectrum of emission , @xmath126 , after the interaction of 46.6 gev electrons with one cycle ( left panel ) and five cycles ( right panel ) of a laser pulse of intensity @xmath127 ( so that @xmath128 $ ] see eq.([eq:2 ] ) ) . here @xmath129 is the distribution of the beam electrons and @xmath130 . , title="fig : " ] , and a spectrum of emission , @xmath126 , after the interaction of 46.6 gev electrons with one cycle ( left panel ) and five cycles ( right panel ) of a laser pulse of intensity @xmath127 ( so that @xmath128 $ ] see eq.([eq:2 ] ) ) . here @xmath129 is the distribution of the beam electrons and @xmath130 . , title="fig : " ] in fig.[fig1 ] the beam - wave interaction is traced during @xmath131 cycles of the incident laser pulse ( @xmath132 fs ) . the initial beam electron energy is rapidly converted into @xmath1-photons with high @xmath133 , which then rapidly produce pairs , the typical rates of the processes being of the order of the inverse light period . however , the larger fraction of the new particles is born at @xmath134 , with strongly reduced pair production rate . slow absorption of photons with @xmath135 maintains pair production even after tens of wave periods , as shown in fig.[fig2 ] . . we see that the laser - beam interaction may be accompanied by multiple pair production . the initial energy of a beam electron is efficiently spent for creating pairs with significantly lower energies as well as softer @xmath1-photons . this effect may be used for producing a pair plasma . it could also be employed to deactivation after - use electron beams , reducing radiation hazard . the way to solve the kinetic equations is accurate and it does not employ the monte - carlo method . the solution can be used to benchmark numerical methods designed to simulate processes in qed - strong laser fields . * we acknowledge * help and advice we received from s. s. bulanov , m. hegelich , j. g. kirk , h. ruhl , t. schlegel and t. tajima . one of us ( i.s . ) is supported by the doe nnsa under the predictive science academic alliances program by grant de - fc52 - 08na28616 . in weaker fields , especially for the particular case of a harmonic wave , the emitted power is given by an integral over many periods of the wave . in this case , the solution of the emission problem in the weak harmonic wave field ( under the requirement on the wave amplitude opposite to that formulated in ineq.([eq : strong ] ) ) is given in section 101 of @xcite as a sum over multi - photon - orders , resulting from the fourier - series expansion for the ( infinitely long ) periodic wave . this standard approach , however , may become meaningless as applied to ultra - strong laser pulses , for many reasons . these pulses may be so short that they can not be thought of as harmonic waves . their fields may be strong enough to force an electron to expend its energy on radiation in less than a single wave period . however , an even more important point is that the radiation loss rate and even the spectrum of radiation is no longer an integral characteristic of the particle motion through a number of wave periods : a local dependence of emission on both particle and field characteristics is typical for the strong fields . the latter statement may be found in @xcite , section 101 . for the particular case of the 1d wave field the evaluation of the _ formation time _ for emission is provided in @xcite within the framework of classical electrodynamics . it is shown that the formation time is much shorter than the wave period as long as ineq.([eq : strong ] ) is fulfilled . here the emission problem is discused for qed - strong fields . we consider a 1d wave field taken in the lorenz gauge @xcite : @xmath136 @xmath137 , @xmath138 and @xmath139 being the 4-vectors of the potential , the wave and the coordinates . herewith the 4-dot - product is introduced in a usual manner : @xmath140 etc . space - like 3-vectors ( i.e. , the first to the third components of a 4-vector ) in contrast with 4-vectors are denoted in bold , 4-indices are denoted with greek letters . recall , that a metric signature @xmath44 is used , therefore , for space - like vectors the 3d scalar product and 4-dot - product have opposite signs , particularly : @xmath141 a method facilitating many derivations involves the introduction of a specific time - space coordinate frame . introduce a transformed space - time ( tst ) : @xmath142 subscript @xmath143 denoting the vector components orthogonal to @xmath144 . the properties of the tst provide a convenient description for the classical motion of an electron in the 1d wave field . note first that @xmath145 second , the generalized momentum components , @xmath146 and @xmath147 , are conserved . third , the metric tensor in the tst is : @xmath148 note the unusual off - diagonal structure of the metric tensor , resulting in a strange relationship between contravariant and covariant coordinates : @xmath149 , @xmath150 . specifically , the ( contravariant ) component of the electron momentum , @xmath146 , is a motional invariant , as long as the vector - potential ( and the hamiltonian , @xmath151 ) does not depend on the ( covariant ) coordinate , @xmath152 ( despite its dependence on @xmath153 ) : @xmath154 finally , the identity , @xmath155 , being expanded in the tst metric , gives : @xmath156 the derivative over @xmath153 or , the same , over @xmath157 is conveniently related to the derivative over the proper time for electron : @xmath158=c(k\cdot p)\frac{d}{d\xi}.\ ] ] many characteristics of emission may be expressed in terms of the relationship between the 4-momenta of the electron at different instants : @xmath159 where @xmath160 as a consequence from eq.([eq : classic ] ) , one can obtain the expression for the _ momenta retarded product _ ( mrp ) : @xmath161 note , that the mrp is given by eq.([eq : pp ] ) for an arbitrary difference between @xmath157 and @xmath162 , but only for the particular case of the 1d wave field . however the limit of this formula as @xmath163 , which is as follows : @xmath164 or , in terms of the mrp in the proper time , @xmath165 : @xmath166 has a much wider range of applicability . eq.([eq : gencovariance ] ) is derived from the equation of motion : @xmath167 using the identities : @xmath168 @xmath169 here @xmath170 is the lorentz four - force : @xmath171 the dirac equation which determines the evolution of the wave function , @xmath172 , for a _ non - emitting _ electron in the external field , reads : @xmath173\psi=\psi,\ ] ] @xmath174 being the dirac @xmath175 matrices , @xmath176 . the relativistic dot - product of the dirac matrices by 4-vector , such as @xmath177 , is the linear combination of the dirac matrices : @xmath178 . such a linear combination , which is also a @xmath175 matrix , may be multiplied by another matrix of this kind or by 4-component bi - spinor , such as @xmath172 , following matrix multiplication rules . for example , @xmath179 is a bi - spinor , as is the matrix , @xmath177 multiplied from the right hand side by the bi - spinor , @xmath172 . the solution of eq.([eq : dirac ] ) in a form of a plane electron wave can be conveniently expressed in terms of the classical solution : @xmath180 the normalization coefficient @xmath181 . by expanding the phase multiplier in the tst , a more convenient form can be provided : @xmath182 } { \lambdabar_c } \right].\ ] ] here @xmath183 is plane wave bi - spinor amplitude , which satisfies the system of four linear algebraic equations : @xmath184 as well as the normalization condition : @xmath185 the @xmath157-dependent phase multiplier , @xmath186 , is as follows : @xmath187 or , @xmath188 using eq.([eq : classic ] ) , one can find : @xmath189 \right)}{2(k\cdot p)}\right]u(p(\xi_1))\ ] ] and verify that eq.([eq : volkov ] ) satisfies the dirac equation . to prove all these assertions , note that the plane wave bi - spinor amplitude once expressed in terms of the classical solution eq.([eq : classic ] ) , satisfies the following commutation rule : @xmath190\left[1+\frac{(\gamma\cdot k ) } { 2(k\cdot p)}(\gamma\cdot \delta a)\right]= \left[1+\frac{(\gamma\cdot k ) } { 2(k\cdot p)}(\gamma\cdot \delta a)\right]\left[(\gamma\cdot p(\xi_1))\pm 1\right].\ ] ] the latter can be proved using the commutation rules as in eqs.(22.5 ) from @xcite for the dirac matrices as well as the identities , @xmath191 : @xmath192 \left[1+\frac{(\gamma\cdot k)(\gamma\cdot \delta a)}{2(k\cdot p ) } \right]=\ ] ] @xmath193=\ ] ] @xmath194 \left[(\gamma\cdot p(\xi_1 ) ) \pm 1\right].\ ] ] thus , eq.([eq : commutation ] ) is proved . this commutation rule allows us to verify that @xmath183 is indeed a bi - spinor amplitude of a planar wave with the four - vector @xmath195 , satisfying eq.([bispinor ] ) . finally , to verify that eq.([eq : volkov ] ) gives the solution of the dirac equation , one should account for eq.([bispinor ] ) as well as the identity , @xmath196}{d\xi}=(\gamma\cdot k)(\gamma\cdot k)\left(\gamma\cdot \frac{da}{d\xi}\right ) u(p_0)=0,\ ] ] which is valid , because @xmath197 , the advantage of the approach used here as compared to the known volkov solution ( see section 40 in @xcite ) is that the wave function in eqs.([eq : volkov]-[eq : qedpropagator ] ) is described in a self - contained manner within some finite time interval , @xmath198 ( in fact , this interval is assumed to be very short below ) in terms of the local parameters of the classical trajectory of electrons . this approach is better applicable to strong fields , in which the time interval between subsequent emission occurrences , which destroys the unperturbed wave function , becomes very short . the emission problem is formulated in the following way . the electron motion in the strong field may be thought of as a sequence of short intervals . within each of these intervals the electron follows a piece of a classical trajectory , as in eq.([eq : classic ] ) , and its wave function ( an electron state ) is given by eq.([eq : volkov ] ) . the transition from one piece of the classical trajectory to another , or , the same , from one electron state to another occurs in a probabilistic manner . the probability of this transition , which is accompanied by a photon emission is calculated below using the qed perturbation theory . we consider the laser wave field as a classical field ( @xmath199 ) and apply first - order perturbation theory with respect to the hamiltonian of the electron - photon interaction , @xmath200 . the only difficulty specific to strong pulsed fields is that the short piece of the electron trajectory is strictly bounded in space and in time , while the qed invariant perturbation theory is based on the matrix element , which is the integral over infinite 4-volume . to avoid this difficulty the following method is suggested , which is analogous to the dipole emission theory as applied in tst . introduce domain , @xmath201 , bounded by two hypersurfaces , @xmath202 and @xmath203 ( see fig.[fig_1 ] ) . the difference @xmath204 is bounded as described below , so that @xmath205 covers only a minor part of the pulse . a volume , @xmath206 is a section of @xmath205 subtended by a line @xmath207 . with the following choice for the normalization coefficient in eq.([eq : volkov ] ) : @xmath208 the integral of the electron density in the volume @xmath209 @xmath210 is set to unity , i.e. there is a single electron in the volume @xmath209 . this statement follows from eq.([eq : volkov ] ) and the known property of normalized bi - spinor amplitudes : @xmath211 . here the hat means the dirac conjugation . for a photon of wave vector , @xmath212 , and polarization vector , @xmath213 , introduce the wave function : @xmath214 } { \sqrt{n_p } } l^\mu,\ ] ] or , by expanding this in the tst : @xmath215 l^\mu,\ ] ] where : @xmath216.\ ] ] here the photon momentum and photon energy are related to @xmath18 and @xmath4 correspondingly , or , equivalently , dimensionless @xmath212 equals dimensional @xmath212 multiplied by @xmath217 . the choice of the normalization coefficient , @xmath218 corresponds to a single photon in the volume , @xmath209 . the emission probability , @xmath219 , is given by an integral over @xmath205 : @xmath220 here @xmath221 is the number of states for the emitted photon . the transformation of the phase volume as in eq.([eq : firstphase ] ) is based on the following jacobian : @xmath222 which is also used below in many places . a subscript @xmath51 denotes the electron in the initial ( _ i _ ) or final ( _ f _ ) state . the number of electron states in the presence of the wave field , @xmath223 , should be integrated over the volume @xmath209 @xmath224 the integration by @xmath225 results in three @xmath226 functions , expressing the conservation of totals of @xmath227 and @xmath228 , for particles in initial and final states : @xmath229 twice integrated with respect to @xmath230 , the probability @xmath219 is proportional to a long time interval , @xmath231 , if the boundary condition for the electron wave at @xmath232 is maintained within that long time . on transforming the integral over @xmath233 to that over @xmath52 , one can find : @xmath234 @xmath235 to take the large value of @xmath236 seems to be the only way to calculate the integral , however , the emission probability calculated in this way relates to multiple electrons in the initial state , each of them locating between the wave fronts @xmath232 and @xmath203 during much shorter time , @xmath237 for a single electron the emission probability becomes : @xmath238 using @xmath226 functions it is easy to integrate eq.([eq : probab ] ) over @xmath239 : @xmath240 where @xmath241,\ ] ] @xmath242 and @xmath243 are the electron phase multipliers , @xmath186 , for the electron in initial and final states and @xmath244 if in the expression for @xmath245 the numerator of the fraction is taken for the initial state of electron , then the denominator should be taken for the final state and vise versa . prior to discussing eq.([eq : cons ] ) , return to eq.([eq : classic ] ) and analyze it component - by - component in the tst . it appears that three of the four components of that equation describe the conservation of @xmath45 and @xmath246 for electron _ in the course of its emission - free motion_. at the same time , yet another component of eq.([eq : classic ] ) , specifically , @xmath247 , directed along @xmath138 , describes the energy - momentum exchange between the electron and the 1d wave field , maintaining the identity , @xmath248 . now turn to eq.([eq : cons ] ) . again , three of the four components express the conservation of the same variables _ in the course of the photon emission _ , while the @xmath247 component , directed along @xmath138 describes the absorption of energy and momentum from the wave field in the course of the photon emission . note , that in the case of a strong field , the energy absorbed from field is not an integer number of quanta , and that for short non - harmonic field it is not even a constant , but a function of the local field . to calculate the matrix element , one can re - write it as the double integral over @xmath249 . the integrand in this double integral includes the phase multiplier and the following matrix product : @xmath250^*= ... \ ] ] using eq.([eq : qedpropagator ] ) one can reduce the multipliers of the kind of @xmath251 to @xmath252 , which can be expressed in terms of the polarization density matrix at the position @xmath162 . although in a strong wave electrons may be polarized ( see @xcite ) , in the present work the emission probability is assumed to be averaged over the electron initial and final polarizations . therefore , the polarization matrix is used in the form of @xmath253 $ ] and one can re - write the integrand as follows : @xmath254)\right]\times\ ] ] @xmath255(\gamma\cdot l^*)\times\nonumber\\ \left[1+\frac{(\gamma\cdot k)}{2(k\cdot p_i ) } ( \gamma\cdot [ a(\xi)-a(\xi_1)])\right ] \left[(\gamma\cdot p_i(\xi_1))+1\right](\gamma\cdot l ) \}= ... \ ] ] in transforming the trace of matrix in addition to the mentioned above relationships and the dirac matrix algebra we use the commutation rule eq.([eq : commutation ] ) , the conservation law eq.([eq : cons ] ) , and the identity @xmath256 for a spatial unity vector of the emitted photon polarization , which can be also written as @xmath257 . move @xmath258 to the first position using the commutation rule eq.([eq : commutation ] ) and then move it to the last position . in the product @xmath259(\gamma\cdot l)[(\gamma\cdot p_f(\xi_1))+1]$ ] we should keep only odd powers in @xmath1 , as long as in the preceding multiplier only odd powers of @xmath1 are present : @xmath260)\right ] ( \gamma\cdot l^*)\times\ ] ] @xmath261)\right ] \left[(\gamma\cdot p_i(\xi_1))(\gamma\cdot l)(\gamma\cdot p_f(\xi_1))+(\gamma\cdot l)\right ] \}=\ ] ] in the product of the first three multipliers separate the even powers in @xmath262 , which are @xmath263 $ ] . the traces of the products of two or four dirac matrices are calculated using eqs.(22.9 - 10 ) from @xcite . @xmath264 \left[1+\frac{(l^*\cdot k)(l\cdot k)}{2(k\cdot p_i)(k\cdot p_f ) } ( a(\xi)-a(\xi_1))^2\right]+(l^*\cdot p_i(\xi_1 ) ) ( l\cdot p_f(\xi_1))+\ ] ] @xmath265(a(\xi)-a(\xi_1))^2+\ ] ] in one of the two terms linear in @xmath262 we move @xmath266 from the first to the last position : @xmath267)(\gamma\cdot p_i(\xi_1))[2(l^*\cdot p_f(\xi_1))(\gamma \cdot l)+ ( \gamma\cdot p_f(\xi_1))]\right\}-\ ] ] @xmath268 ) [ 2(l^*\cdot p_i(\xi_1))(\gamma \cdot l)+(\gamma\cdot p_i(\xi_1 ) ) ] ( \gamma\cdot p_f(\xi_1 ) ) \right\}=\ ] ] calculate traces for the products of four dirac matrices in the terms linear in @xmath262 : @xmath264 \left[1+\frac{(l^*\cdot k)(l\cdot k)}{2(k\cdot p_i)(k\cdot p_f ) } ( a(\xi)-a(\xi_1))^2\right]+(l^*\cdot p_i(\xi_1 ) ) ( l\cdot p_f(\xi_1))+\ ] ] @xmath269 ( a(\xi)-a(\xi_1))^2+\ ] ] @xmath270\cdot p_i(\xi_1))+\ ] ] @xmath271\cdot p_f(\xi_1 ) ) -\frac 12(\left[a(\xi)-a(\xi_1)\right]\cdot p_f(\xi_1))+ \frac{(k\cdot p_f)}{2(k\cdot p_i ) } ( \left[a(\xi)-a(\xi_1)\right]\cdot p_i(\xi_1))-\ ] ] @xmath272\cdot l^*\right ) \left(l\cdot\left[a(\xi)-a(\xi_1)\right]\right)+ \frac{([a(\xi)-a(\xi_1)]\cdot p_i(\xi_1))(k\cdot l)(l^*\cdot p_f(\xi_1))}{(k\cdot p_i)}+\ ] ] @xmath273\cdot p_f(\xi_1))(k\cdot l)(l^*\cdot p_i(\xi_1))}{(k\cdot p_f)}\}=\ ] ] group the terms proportional to @xmath274 @xmath275+k\frac{([a(\xi)-a(\xi_1)]\cdot p_f(\xi_1))}{(k\cdot p_f)}\right\}\right)+\ ] ] @xmath276+k\frac{([a(\xi)-a(\xi_1)]\cdot p_i(\xi_1))}{(k\cdot p_i)}\right\}\right)+\ ] ] group in an analogous manner the terms , proportional to @xmath277 : @xmath278+k\frac{([a(\xi)-a(\xi_1)]\cdot p_f(\xi_1))}{(k\cdot p_f)}\right\}\right)+\ ] ] @xmath279+k\frac{([a(\xi)-a(\xi_1)]\cdot p_i(\xi_1))}{(k\cdot p_i)}\right\}\right)-1+\ ] ] the residual terms are @xmath280 : @xmath281 ( l^*\cdot k)(l\cdot k ) -(l^*\cdot k)(l\cdot p_i(\xi_1))(k\cdot p_f)- ( l^*\cdot k)(l\cdot p_f(\xi_1))(k\cdot p_i)\right\}\times\ ] ] @xmath282 use eq.([eq : cons ] ) and introduce wherever reasonable the functions of @xmath157 : @xmath283 @xmath284 ^ 2}{2(k\cdot p_f)}+ ( l^*\cdot p_{f}(\xi_1))\left(l\cdot k\right)\frac{\left[a(\xi)-a(\xi_1)\right]^2}{2(k\cdot p_i)}+\ ] ] @xmath281 ( l^*\cdot k)(l\cdot k ) -(l^*\cdot k)(l\cdot p_i(\xi_1))(k\cdot p_f)- ( l^*\cdot k)(l\cdot p_f(\xi_1))(k\cdot p_i)\right\}\times\ ] ] @xmath285 the total of all terms proportional to @xmath286 given above vanishes . this can be shown using the following properties : the tensor @xmath287 is hermitian ; the difference @xmath288 is the total of two terms , proportional to @xmath289 and @xmath290 , the latter being orthogonal to @xmath291 . the residual terms are : @xmath292 ^ 2}{2(k\cdot p_f)}\right\}\right)+\ ] ] @xmath293 ^ 2}{2(k\cdot p_i)}\right\}\right)-\ ] ] the last term , @xmath294 , we transform using eq.([eq : pp ] ) @xmath295 ^ 2-\frac12(p_i(\xi)\cdot p_i(\xi_1))- \frac12(p_f(\xi)\cdot p_f(\xi_1))=\ ] ] @xmath296\cdot\left[p_f(\xi_1)-p_i(\xi_1)\right]\right)+\ ] ] @xmath297 ^ 2\left((k\cdot p_i)-(k\cdot p_f)\right)^2}{4 ( k\cdot p_i)(k\cdot p_f)}.\ ] ] on substituting this into the integral expression for the matrix element , some terms give zero contributions to the integral . particularly , the following integrals vanish , as long as the expressions in the square brackets are the perfect time derivatives : @xmath298d\xi}=0.\ ] ] from here it is also easy to derive that : @xmath299d\xi}=0.\ ] ] therefore , with the transformed product of the dirac matrices , which should be multiplied by a factor of two ( as long as we should sum over the final electron polarization instead of averaging ) , we obtain : @xmath300d d\xi d\xi_1}}}{(2\pi\lambdabar_c)^2 ( k\cdot k^\prime)(k\cdot p_i)(k\cdot p_f)},\ ] ] where @xmath301 ^ 2\left((k\cdot p_i)-(k\cdot p_f)\right)^2 } { 8 ( k\cdot p_i)(k\cdot p_f)}.\ ] ] the matrix element may also be summed , if desired , for two possible directions of the polarization vector . the second term in the integrand is simply multiplied by two , while in the first one the negative of the metric tensor should be substituted for the product of the polarization vectors ( see section 8 in @xcite ) , so that @xmath302 substitutes for @xmath303 . the latter may be transformed using eq.([eq : pp ] ) , thus , giving : @xmath304 ^ 2\left((k\cdot p_i)^2+(k\cdot p_f)^2\right ) } { 4 ( k\cdot p_i)(k\cdot p_f)}.\ ] ] now perform integration over @xmath305 . on developing the dot - product , @xmath306 , in @xmath245 in the tst metric , @xmath307 , one can find : @xmath308,\ ] ] where @xmath309 @xmath310 ^ 2d\xi_2}\right\}}{2\lambdabar_c(k\cdot p_i)(k\cdot p_f)},\ ] ] @xmath311 integration over @xmath312 is performed using the following formula : @xmath313 the right hand side is proportional to @xmath314 , where the rapidly oscillating at large @xmath315 term results in a vanishing contribution to the integral over @xmath249 and should be neglected . the integrated over @xmath305 probability is : @xmath316 in strong fields the following estimates may be applied : @xmath317 now the bounds for @xmath204 can be _ consistently _ introduced : @xmath318 under these bounds , first , the time interval ( [ eq : time ] ) is much greater than the _ formation time_. therefore , once the double integral over @xmath249 is transformed into the integral over @xmath319 , the span to integrate over @xmath320 can be extended towards @xmath321 . hence , the emission probability becomes linear in @xmath204 , because only the integration over @xmath322 is performed from @xmath323 till @xmath324 : @xmath325 on the other hand the difference , @xmath204 , is to be small enough , so that the probability of emission within the time interval of eq.([eq : time ] ) is much less ( or at least less ) than unity : @xmath326 therefore , perturbation theory is applicable . in addition , the emission probability can be expressed in terms of the local electric field . note , that consistency in ( [ consistency ] ) is ensured in relativistically strong electromagnetic fields as long as @xmath327 , with no restriction on the magnitude of the electromagnetic field experienced by an electron . under the condition ( [ consistency ] ) in the integral over @xmath328 the small differences in the vector potential may be linearized : @xmath329 ^ 2= [ { \bf a}(\xi)-{\bf a}(\xi_1 ) ] ^2= \left ( \frac{d{\bf a}}{d\xi}\right)^2\theta^2,\ ] ] @xmath330 after this the integral , @xmath331 by means of a substitution , @xmath332 , may be expressed in terms of the macdonald functions , using the following equations : @xmath333dz}=\frac2{\sqrt{3}}k_{1/3}(r),\ ] ] ( see eq.(8.433 ) in @xcite ) @xmath334dz}=\frac2{\sqrt{3}}k_{2/3}(r),\ ] ] @xmath335dz}=-\frac2{\sqrt{3 } } \int\limits_r^{+\infty}{k_{1/3}(r^\prime)dr^\prime},\ ] ] as well as the recurrent relationships between the macdonald functions : @xmath336 and the identity , @xmath337 . in this way we arrive at eq.([eq : probabf ] ) . the advantage of the macdonald functions is the simplicity and fast convergence of the integral representation for them : @xmath338\cosh(\nu z ) dz } , \ ] ] @xmath339\frac{\cosh(\nu z)}{\cosh ( z ) } dz } , \ ] ] ( see eq.(8.432 ) in @xcite ) . in numerical simulations , therefore , the macdonald functions are very easy to use . here we consider a photon with the wave four - vector @xmath290 and find a probability of its absorption in the strong 1d wave field with producing electron and positron , their four - momenta being @xmath340 and @xmath341 correspondingly . the matrix element for the absorption probability is now related to the element of the electron phase volume . the wave function for the absorbed photon is the complex conjugated wave function of the emitted photon . the bi - spinor amplitude for positron is the dirac conjugated amplitude for electron . with these minor changes , the matrix element for the absorption probability is transformed as follows : @xmath342 where @xmath343,\ ] ] and @xmath344 eq.([eq : cons1 ] ) may be obtained from eq.([eq : cons ] ) as well as the new phase multiplier may be obtained from the earlier derived phase multiplier by virtue of the following transformation : @xmath345 because the photon emission and the photon absorption are two cross - invariant channels of the same reaction . then , to calculate the matrix element , one can re - write it as the double integral over @xmath249 . the integrand in this double integral includes the phase multiplier and the following matrix product : @xmath346^*= ... \ ] ] now we should follow the way we used to expand eq.([eq : matproduct ] ) with the following modifications . the polarization matrix for the positron is @xmath347 $ ] . while applying eq.([eq : qedpropagator ] ) to a positron , the dimensionless vector potential , which had been defined above in terms of a negative charge of electron , should now be taken with the opposite sign , therefore : @xmath348)\right]\times\ ] ] @xmath349(\gamma\cdot l)\times\nonumber\\ \left[1-\frac{(\gamma\cdot k)}{2(k\cdot p_p ) } ( \gamma\cdot [ a(\xi)-a(\xi_1)])\right ] \left[(\gamma\cdot p_p(\xi_1))-1\right](\gamma\cdot l^ * ) \}.\ ] ] we see that again the transformation eq.([crossinv ] ) allows us to obtain the above matrix product from that one we derived in developing eq.([eq : matproduct ] ) in appendix a. however , in addition to the modifications listed in eq.([crossinv ] ) we need to transform @xmath350 and change the whole sign of the matrix product . now we can apply this transformation procedure to the resulting expression for the emission probability , eq.([eq : dwfi ] ) , and obtain the result for the absorption probability : @xmath351d d\xi d\xi_1}}}{(2\pi\lambdabar_c)^2 ( k\cdot k^\prime)(k\cdot p_e)(k\cdot p_p)},\ ] ] where @xmath352 ^ 2(k\cdot k^\prime)^2 } { 8 ( k\cdot p_p)(k\cdot p_e)}.\ ] ] note an interesting polarization property : in the field of linearly polarized 1d wave the emission probability is _ maximal _ for photons with the same polarization as that for 1d wave , but the absorption probability for such photons is _ minimal _ ( the polarization - dependent term in @xmath353 is negative ) . we still employ here the probabilities averaged over the photon polarizations , but we think this approach may be accurate only when applied to the processes in circularly polarized strong wave fields . 99 c. bula _ et al _ , phys . lett . * 76 * , 3116 ( 1996 ) ; d.l . al _ , phys . lett . * 79 * , 1626 ( 1997 ) ; c. bamber _ et al _ , phys . rev . d * 60 * , 092004 ( 1999 ) . j. schwinger , phys . rev . * 82 * , 664 ( 1951 ) ; e. brezin and c. itzykson , phys . d * 2 * , 1191 ( 1970 ) . m. marklund and p. k. shukla , rev . mod . phys . * 78 * , 591 ( 2006 ) ; y. i. salamin _ et al _ , phys . reports * 427 * , 41 ( 2006 ) ; a. m. fedotov et al , phys . lett . * 105 * , 080402 ( 2010 ) . a. r. bell and j. g. kirk , phys . rev . * 101 * , 200403 ( 2008 ) ; h. hu , c. mueller and c.h . keitel , phys . * 105 * , 080401 ( 2010 ) . l. d. landau and e. m. lifshits , _ the classical theory of fields _ ( pergamon , new york , 1994 ) ; 1st edition : ( moscow , gostekhizdat , 1941 ) . a. di piazza , k. z. hatsagortsyan , and c. h. keitel , phys . * 97 * , 083603 ( 2006 ) ; s. s. bulanov , phys . e * 69 * , 036408 ( 2004 ) . et al _ , opt . * 29 * , 2837 ( 2004 ) ; v. yanovsky _ et al _ , optics express * 16 * , 2109 ( 2008 ) . http://eli-laser.eu/ ; e. gerstner , nature * 446 * , 16 ( 2007 ) ; t. feder , phys . today * 63 * ( 6 ) , 20 ( 2010 ) . j. d. jackson , _ classical electrodynamics _ ( wiley , new york , 1999 ) . a. zhidkov _ et al _ , phys . lett . * 88 * , 185002 ( 2002 ) ; j. koga , t. zh . esirkepov and s. v. bulanov , phys . plasmas * 12 * , 093106 ( 2005 ) . i. v. sokolov , jetp * 109 * , 207 ( 2009 ) ; i. v. sokolov _ et al _ , phys . plasmas * 16 * , 093115 ( 2009 ) . i. v. sokolov _ et al _ , phys . e. * 81 * , 036412 ( 2010 ) . v. b. berestetskii , e. m. lifshitz , and l. p. pitaevskii , _ quantum electrodynamics _ ( pergamon , oxford , 1982 ) . a. i. nikishov and v. i. ritus , sov . usp . * 13 * , 303 ( 1970 ) ; v. i. ritus , j. rus . 6 , 497 ( 1985 ) ; v. i. ritus , in _ issues in intense - field quantum electrodynamics _ , ed . by v. l. ginzburg ( nova science , commack , 1987 ) ; see also the papers cited there : a. i. nikishov and v. i. ritus , sov . jetp * 19 * , 529 ( 1964 ) ; * 19 * , 1191 ( 1964 ) ; * 20 * , 757 ( 1965 ) ; n. b. narozhny et al . jetp * 20 * , 622 ( 1965 ) ; m. v. galynsky and s. m. sikach , physics of particles and nuclei * 29 * , 469 ( 1998 ) ; l. dongguo _ et al _ , jpn * 42 * , 5376 ( 2003 ) ; e. nerush and i. kostyukov , phys . e * 75 * , 057401 ( 2007 ) . e. m. lifshitz and l. p. pitaevskii , _ physical kinetics _ ( pergamon , oxford , 1981 ) . about the hystorical roots of guage invariance , see : j. d. jackson and l. b. okun , rev . phys . * 73 * , 663 . ( 2001 ) t. omori _ et al . _ , lett . * 96 * , 114801 ( 2006 ) . i. s. gradshtein and i. m. ryzhik , _ table of integrals , series , and products _ ( academic press , new york , 1965 ) .

qed - effects are known to occur in a strong laser pulse interaction with a counterpropagating electron beam , among these effects being electron - positron pair creation . we discuss the range of laser pulse intensities of @xmath0 combined with electron beam energies of tens of gev . in this regime multiple pairs may be generated from a single beam electron , some of the newborn particles being capable of further pair production . radiation back - reaction prevents avalanche development and limits pair creation . the system of integro - differential kinetic equations for electrons , positrons and @xmath1-photons is derived and solved numerically .

the development of strategies to control the dynamic of a viral spread in a population is a central problem in public health and network security@xcite . in particular , how to control the traffic between subpopulations in the case of an epidemic outbreak is of critical importance . in this paper , we analyze the problem of controlling the spread of a disease in a population by regulating the traffic between subpopulations . the dynamic of the spread depends on both the characteristics of the subpopulation , as well as the structure and parameters of the transportation infrastructure . our work is based on a recently proposed variant of the popular sis epidemic model to the case of populations interacting through a network @xcite . we extend this model to a metapopulation framework in which large subpopulations ( i.e. , cities ) are represented as nodes in a metagraph whose links represent the transportation infrastructure connecting them ( i.e. , roads ) @xcite . we propose an extension of the susceptible - infected - susceptible ( sis ) viral propagation model to metapopulations using stochastic blockmodels @xcite . the stochastic blockmodel is a complex network model with well - defined random communities ( or blocks ) . we model each subpopulation as a random regular graph and the interaction between subpopulations using random bipartite graphs connecting adjacent subpopulations . the main advantage of our approach is that we can find the optimal traffic among subpopulations to control a viral outbreak solving a standard form convex semidefinite program . in this section we introduce some graph - theoretical nomenclature and the dynamic spreading model under consideration . let @xmath0 denote an undirected graph with @xmath1 nodes , @xmath2 edges , and no self - loops . we denote by @xmath3 the set of nodes and by @xmath4 the set of undirected edges of @xmath5 . the number of neighbors of @xmath6 is called the _ degree _ of node i , denoted by @xmath7 . a graph with all the nodes having the same degree is called regular . the adjacency matrix of an undirected graph @xmath5 , denoted by @xmath8 $ ] , is an @xmath9 symmetric matrix defined entry - wise as @xmath10 if nodes @xmath6 and @xmath11 are adjacent , and @xmath12 otherwise for all @xmath6 . ] . since @xmath13 is symmetric , all its eigenvalues , denoted by @xmath14 , are real . in a regular graph , the largest eigenvalue @xmath15 is equal to the degree of its nodes @xcite , and the associated eigenvector is @xmath16 ( where @xmath17 is the vector of all ones of size @xmath1 ) . our modeling approach is based on the n - intertwined sis model proposed by van mieghem et at . in @xcite . consider a network of @xmath1 individuals described by the adjacency matrix @xmath18 $ ] . the infection probability of an individual at node @xmath19 at time @xmath20 is denoted by @xmath21 . let us assume , for now , that the viral spreading is characterized by the infection and curing rates , @xmath22 and @xmath23 , . hence , the linearized n - intertwined sis model in @xcite is described by the following differential equation : @xmath24 where @xmath25 , @xmath26 , and @xmath27 . concerning the non - homogeneous epidemic model , we have the following result : [ prop : heterogeneous sis stability condition]consider the heterogeneous n - intertwined sis epidemic model in ( [ eq : heterosis ] ) . then , if @xmath28 an initial infection @xmath29^{n}$ ] will die out exponentially fast , i.e. , there exists an @xmath30 such that metapopulation models are useful to characterize the dynamics of systems composed by connected subpopulations @xcite . for example , consider a population of @xmath1 individuals distributed over @xmath31 cities connected via a collection of roads . at a lower level , we can described the pattern of interactions in the entire population as a massive graph with @xmath1 nodes ( individuals ) , where an edge @xmath32 represents the interaction between two individuals @xmath6 and @xmath11 . alternatively , we can describe this population at a higher level using a much smaller graph , called the metagraph , in which nodes represent cities and edges represent roads connecting cities . in the metapopulation model , there are two elements to take into consideration . on the one hand , we have the intrapopulation evolution , which is related to the evolution of an infection within each subpopulation , as in isolation . on the other hand , we have the subpopulation interaction , which is related to encounters between individuals from different subpopulations . we describe both elements in the following subsections . assume we partition the whole population of @xmath1 individuals into @xmath31 subpopulations of sizes @xmath33 . we denote the sets of nodes in each subpopulation by @xmath34 . we also assume that we are not given any information about the connectivity of individuals inside each subpopulation , apart from the number of nodes , @xmath35 , and the average degree @xmath7 of the individuals inside the @xmath6-th subpopulation . hence , it is reasonable to model the connectivity structure of each subpopulation as a random regular graph of size @xmath35 and degree @xmath7 . we denote the @xmath36 adjacency matrix of this random regular graph as @xmath37 . as we mentioned above , the largest eigenvalue of this random regular graphs is @xmath38 and the associated eigenvector is @xmath39 . the interaction between subpopulations is a crucial component that strongly influences the entire dynamics of the system . to model the interaction between subpopulations @xmath6 and @xmath11 , we assume that a random collection of @xmath40 individuals in @xmath41 connect to a random collection of @xmath40 individuals in @xmath42 . we can algebraically represent this connectivity pattern by defining a @xmath43 matrix @xmath44 representing to the structure of a random bipartite graph connecting two sets of nodes of sizes @xmath35 and @xmath45 via @xmath40 edges . the @xmath9 adjacency matrix of the whole population of individuals described above is a random matrix , denoted by @xmath46 , that can be defined according to block matrices , as follows . first , the @xmath47-the diagonal block of the population adjacency matrix is the @xmath36 random matrix @xmath37 , defined above . second , the @xmath48-th off diagonal block is the @xmath43 matrix @xmath44 , defined above . hence , the adjacency matrix of the complete population is @xmath49.\ ] ] in the following subsection , we apply the n - intertwined sis epidemic model to the above adjacency matrix to derive a spectral condition for stability of a small initial infection . we use ( [ eq : heterosis ] ) to model the dynamics of an sis spreading process in a metapopulations . assume that the the sis model spreads through the individuals in population @xmath41 with an spreading rate @xmath50 . the recovery rate of individuals in population @xmath41 is @xmath23 . we assume that the spreading rate of a virus from an infected individual in population @xmath11 towards a susceptible individual in population @xmath41 is equal to @xmath51 . let us define the vector @xmath52 as the @xmath35-dimensional vector containing the probabilities of infection of all the individuals in the subpopulation @xmath41 at time @xmath20 . hence , according to ( [ eq : heterosis ] ) , this vector of infection probabilities evolves as @xmath53 where the first and last terms represent the spreading and recovery dynamics within subpopulation @xmath6 . the second term accounts for the spreading of the disease from subpopulations @xmath11 to @xmath6 . we can stack the vectors @xmath54 into an @xmath1-dimensional vector to write the evolution of the whole population as @xmath55 , where @xmath56^{t}$ ] , and @xmath57,\end{aligned}\ ] ] according to proposition [ prop : heterogeneous sis stability condition ] , a small initial infection dies out exponentially fast if the largest eigenvalue of @xmath58 is strictly negative . in what follows , we study the largest eigenvalue of @xmath58 in terms of metapopulation parameters . notice that @xmath58 is a random matrix , since its blocks represent random graphs . to analyze the largest eigenvalue of this random matrix , we make use of the following spectral concentration result from @xcite : consider the random matrix @xmath58 , then , almost surely , @xmath59 where @xmath60 is the expectation of @xmath58 . @xmath61\ ] ] and @xmath62 is the largest eigenvalue under study . or equivalently , @xmath63 where @xmath64 $ ] , @xmath65 , @xmath66 , @xmath67 , @xmath68 , @xmath69 $ ] , and @xmath70 . therefore , we can approximate the largest eigenvalue of the @xmath9 matrix @xmath58 ( where @xmath1 is the number of individuals in the population ) , using the largest eigenvalue of the @xmath71 matrix @xmath72 ( where @xmath31 is the number of subpopulations in the model ) . hence , the condition under which the epidemic is guaranteed die out at rate @xmath73 if @xmath74 is satisfied . in the following section we use this result to find an optimal distribution of traffic between subpopulations in order to contain the epidemic spread . we assume that ( [ die ] ) is not satisfied without implementing a travel restriction policy . we define the travel restriction policy as @xmath75 $ ] where @xmath76 with cost @xmath77 convex . the problem travel restriction problem is formally stated as @xmath78 where , since we do not consider permanent relocation between cities , we assume that @xmath79 . the following lemma states the condition on the model parameters under which the travel restriction problem is feasible . [ feas ] there exists a set @xmath80 satisfying the constraints in ( [ problem ] ) if @xmath81 for all cities @xmath82 . included in proof of theorem [ sdp ] the constraint in lemma [ feas ] is equivalent to the condition that in each individual city the virus would die out at a rate @xmath73 with no intercity connections . the virus can not be forced to die out in the whole system by controlling intercity connections if it can persist in any city in isolation . [ sdp ] the traffic restriction problem given in ( [ problem ] ) is equivalent to the standard form semidefinite program @xmath83 the eigenvalue constraint in ( [ problem ] ) is equivalent to @xmath84 because @xmath85 can be expressed with any basis . multiplying by the positive definite matrix @xmath86 preserves the sign of the largest eigenvalue so we can express the relation as @xmath87 since @xmath88 is a symmetric matrix , we can express ( [ sym ] ) as the semidefinite constraint given in ( [ sdp ] ) , completing the proof of theorem [ sdp ] . to prove lemma [ feas ] , we construct a feasible point @xmath89 satisfying the equivalent constraint ( [ sym ] ) and the box constraint @xmath90 . let @xmath91 from ( [ feascond ] ) , @xmath92 . applying the triangle inequality @xmath93 choosing @xmath94 where @xmath95 guarantees that @xmath96 and that @xmath90 , completing the proof . standard form semidefinite programs are solvable in polynomial time via convex optimization methods therefore , a central authority can set traffic limits on all cities in order to guarantee the epidemic dies out at rate @xmath73 while minimizing the cost . in many cases it may not be possible to compute or implement a centralized policy . if we suppose the costs are incurred locally by each city directly effected by the restriction , @xmath97 then we can compute a heuristic local solution by first having each city manager solve @xmath98 then coordinating with neighboring cities allowing traffic @xmath99 [ heur ] the local heuristic solution defined in ( [ local1 ] ) and ( [ local2 ] ) yields a feasible solution to the traffic restriction problem , ( [ problem ] ) . equation ( [ local2 ] ) guarantees that @xmath100 is symmetric and @xmath101 . combining with the first constraint in ( [ local1 ] ) , @xmath102 the box constraint in ( [ local1 ] ) guarantees that @xmath76 . consider the matrix @xmath103 whose diagonal entries are strictly negative according to lemma [ feas ] . equation ( [ ddom ] ) guarantees that the matrix ( [ mat ] ) is diagonally dominant . theorem 6.1.10 from @xcite guarantees that the matrix in ( [ mat ] ) is negative semidefinite , satisfying the eigenvalue constraint and completing the proof . the heuristic solution proposed in ( [ local1 ] ) and ( [ local2 ] ) is local in the sense that traffic restrictions on the edges in the intercity network can be computed by each city solving the optimal restrictions for the edges connecting them to other cities . two cities will not necessarily compute the same optimal restriction so the minimum of the two values is used . this is consistent with our model because all travel assumed to be is round trip , thus the realized traffic can be at most the minimum of the traffic allowed by the two cities involved . theorem [ heur ] formally guarantees that this local method yields a solution that causes the virus to die out at at least rate @xmath73 . the fraction of people who are infected is driven to 0 in every city via traffic control . the cost incurred by the local heuristic is about 33% greater than the optimal cost . ] we demonstrate the relative performance of our centralized and local solutions using a sample problem with randomly generated parameters . we choose the cost function @xmath104 because it is convex and satisfies ( [ sep ] ) . furthermore , ( [ cost ] ) intuitively captures the cost of restricting traffic on each edge because there is no cost when traffic is unrestricted ( i.e. @xmath105 ) but the cost tends to @xmath106 as @xmath107 . it is not possible to completely delete an intercity connection . figure [ sim ] shows that with no traffic restrictions all cities go to 100% infection rate while both the optimal solution and local heuristic force the virus to die out . the heuristic solution incurs a higher total cost but also forces the epidemic to die out faster . figure [ net ] shows the network of cities and the traffic restrictions on the edges . points representing cities are scaled proportional to their populations @xmath35 . edges are scaled proportional to the the unrestricted traffic @xmath108 . the color of each edge is linearly scaled from green ( @xmath109 ) to red ( @xmath110 ) . it significant to note that despite the local approach , the restrictions imposed are very similar to the optimal case . more traffic restrictions are required by the heuristic solution , however there is a strong correlation between which edges are restricted by the heuristic solution and by the optimal solution . ] we have proposed a convex framework to contain the propagation of an epidemic outbreak in a metapopulation model by controlling the traffic between subpopulations . in this context , controlling the spread of an epidemic outbreak can be written as a spectral condition involving the eigenvalues of a matrix that depends on the network structure and the parameters of the model . based on our spectral condition , we can find cost - optimal approaches to traffic control in epidemic outbreaks by solving an efficient semidefinite program . 1 r.m . anderson and r.m . may , _ infectious diseases of humans : dynamics and control _ , oxford university press , 1991.[27 ] i. hanski and m. gilpin , metepopulation biology : ecology , genetics and evolution ( academic press , san diego , 1997 ) .

we propose a novel framework to study viral spreading processes in metapopulation models . large subpopulations ( i.e. , cities ) are connected via metalinks ( i.e. , roads ) according to a metagraph structure ( i.e. , the traffic infrastructure ) . the problem of containing the propagation of an epidemic outbreak in a metapopulation model by controlling the traffic between subpopulations is considered . controlling the spread of an epidemic outbreak can be written as a spectral condition involving the eigenvalues of a matrix that depends on the network structure and the parameters of the model . based on this spectral condition , we propose a convex optimization framework to find cost - optimal approaches to traffic control in epidemic outbreaks .

the essence of radiotherapy with protons and heavier ions lies in precise control of incident particles that are designed to stop in tumor volume . the targeting precision will be inevitably deteriorated by multiple scattering in beam modifiers and patient body and such effects must be accurately handled for dose calculations in treatment planning . on the other hand , simplicity and efficiency are also essential in clinical practice and there has always been need for a computational method that balances all these demanding and conflicting requirements . fermi and then eyges ( 1948 ) developed a general theory for charged particles that undergo energy loss and multiple scattering in matter . a group of particles is approximated as a gaussian beam growing in space with statistical variances @xmath0 where @xmath1 is the longitudinal position and @xmath2 and @xmath3 are the projected particle position and angle . the original fermi - eyges theory adopted purely gaussian approximation ( rossi and greisen 1948 ) with ( projected ) scattering power @xmath4 where @xmath5 mev is a constant , @xmath6 is the radiation length of the material , and @xmath7 , @xmath8 , and @xmath9 are the charge , the momentum , and the velocity of the particles . the fermi rossi formula totally ignores effects of large - angle single scattering ( hanson 1951 ) and was found to be inaccurate ( wong 1990 ) . based on formulations by highland ( 1975 , 1979 ) and gottschalk ( 1993 ) , kanematsu ( 2008b ) proposed a scattering power with correction for the single - scattering effect , although within the gaussian approximation , @xmath10 where @xmath11 is the radiative path length . although it would be difficult to calculate integrals because of the embedded integral in the @xmath12 terms , kanematsu ( 2008b ) further derived an approximate formula for the rms displacement of incident ions at the end point in homogeneous matter as @xmath13 where @xmath14 is the ion mass in units of the proton mass , @xmath15 is the expected in - water range on the incidence , @xmath16 is the stopping - power ratio of the matter relative to water , and @xmath17 and @xmath18 are constants . despite the complex involvement of variable @xmath15 in , kanematsu ( 2008b ) found the @xmath19@xmath15 relation for ions in water to be very linear . in fact , preston and kohler of harvard cyclotron laboratory knew the linear relation and derived universal curve @xmath20 with @xmath21 for relative growth of rms displacement @xmath22 in homogeneous matter in an unpublished work in 1968 . starting with the empirical linear relation , this work is aimed to develop a simple and general multiple - scattering model to improve efficiency of numerical heterogeneity handling and to enable further analytical beam modeling . linear approximation @xmath23 for homogeneous systems greatly simplifies to @xmath24 where @xmath25 cm is the radiation length of water , @xmath26 is the scattering / stopping ratio of the material relative to water , and @xmath27 is the geometrical range . was calibrated to for water at @xmath28 . equation @xmath29 at the end point @xmath30 associates and as @xmath31 to lead to another scattering power @xmath32 where @xmath33 is the particle - type - dependent factor . the scattering power is inherently applicable to any heterogeneous system by numerical integral of . we examined these fermi - rossi , extended highland , and linear - displacement models with unpublished measurements by phillips ( hollmark 2004 ) , those by preston and kohler ( kanematsu 2008b ) , and molire - hanson calculations by deasy ( 1998 ) . we took growths of rms displacement @xmath34 with depth @xmath1 for @xmath35 cm protons , @xmath35 cm helium ions , and @xmath36 cm carbon ions in water and rms end - point displacements @xmath37 for them with varied incident range @xmath15 . for a point mono - directional ion beam with in - water range @xmath15 incident into homogeneous matter with constant @xmath16 and @xmath6 , equations are analytically integrated to @xmath38 as a function of residual range @xmath39 at distance @xmath1 . in fact reduces to the universal curve by preston and kohler . a radiation field at a given @xmath1 position can be effectively or virtually modeled with a source , ignoring the matter ( icru-35 1984 ) . the effective extended source is at @xmath40 , where @xmath41 would be minimum in vacuum . the virtual point source is at @xmath42 , from which radiating particles would form a field of equivalent divergence . similarly , the effective scattering point is at @xmath43^{1/2}$ ] , at which a point - like scattering would cause equivalent rms angle and displacement ( gottschalk 1993 ) . with depth @xmath1 and ( b ) rms end - point displacement @xmath19 with incident range @xmath15 for protons , helium ions , and carbon ions in water , calculated with the fermi - rossi ( dashed ) , extended highland ( solid ) , linear - displacement ( dotted overlapping with the solid ) models along with a molire - hanson calculation by deasy ( @xmath44 ) and measurements by phillips ( @xmath45 , @xmath46 , @xmath47 ) and by preston and kohler ( @xmath48 ) . ( c ) end - point lateral shape of a proton pencil beam in water ( @xmath49 cm ) measured by preston and kohler ( @xmath48 ) and estimated by fermi - rossi ( dashed ) and the present ( solid ) formulations . ] \(a ) shows rms - displacement growths @xmath34 and ( b ) shows rms end - point displacements @xmath37 . the present model was virtually identical to the extended highland model with deviations from measurements or molire - hanson calculations within either 2% or 0.02 cm , while the fermi - rossi model overestimated the rms displacements by nearly 10% . ( c ) shows an end - point shape of a pencil beam measured by preston and kohler for @xmath49 cm protons in water and curves estimated by fermi - rossi and the present formulations with additional @xmath50 to @xmath51 for incident beam emittance in their experiment . the gaussian approximation was in fact adequate in this case . \(a ) shows analytical variances @xmath52 , @xmath53 , and @xmath41 growing with depth @xmath1 for protons in water . ( b ) shows the relative distances from current position @xmath1 to the effective extended source @xmath54 , the virtual point source @xmath55 , and the effective scattering point @xmath56 . they approach @xmath57 , @xmath58 , and @xmath59 at the @xmath60 limit . increase of the scattering power with depth moves these points relatively closer to the current position . in application of bragg peaks , the end - point displacement is the most important , for which the gaussian approximation was valid in the proton experiment by preston and kohler . in fact , the single - scattering effect is theoretically small for a thick target ( hanson 1951 ) . although ions suffer nuclear interactions with resultant fragments that generally scatter at large angles ( matsufuji 2005 ) , their contributions may be relatively less significant at the bragg peaks . the present formulation would be thus adequate for radiotherapy . the linear - displacement model with the fermi - eyges theory has brought general formulas for ions , including the universal curve intuitively derived by preston and kohler without explicit formulation of the scattering power . in the present model , the kinematic properties are encapsulated in residual range @xmath61 that is always tracked in beam transport . the ion - type dependence in scattering angle and displacement is simply as proportional to @xmath62 , which leads to 50% for @xmath63he , 28% for @xmath64c , and 24% for @xmath65o with respect to that of protons for a given incident range . these numbers coincide with detailed numerical calculations by hollmark ( 2004 ) . the linearity between end - point displacement and range observed for water is the basis of the present model . its validity for general body - tissue materials is not obvious . we here examine water and two extreme elements hydrogen ( @xmath66 cm ) and calcium ( @xmath67 cm ) among major elements of body tissues ( icru-46 1992 ) , using in the extended highland model . shows their @xmath15@xmath19 relations with geometrical scale correction and indicates that the linearity will generally hold for body - tissue elements . however , the elemental linearity may not truly warrant the validity for systems heterogeneous in atomic compositions . the scattering power only depends on the residual range that is irrelevant to the multiple scattering accumulated in the other upstream materials , whereas the accumulation should influence the single - scattering effect ( kanematsu 2008b ) . in other words , the present model implicitly assumes heterogeneity in density only . in the current practice of treatment planning , the patient heterogeneity is normally modeled with variable - density water ( kanematsu 2003 ) , for which the present model is rigorous with further simplified scattering power @xmath68 the simplicity will minimize computation of the integrands in path integrals for demanding dose calculations ( kanematsu 1998 , 2006 , 2008a ) . a novel multiple - scattering model has been formulated based on the fact such that the rms end - point displacement is proportional to the incident range in water . the model was designed to be equivalent with the extended highland model for stopping ions in water and agreed with measurements within 2% or 0.02 cm in rms displacement . the resultant scattering - power formula that is only inversely proportional to residual range is much simpler than former formulations and can be used in the framework of the fermi - eyges theory for gaussian - beam transport in tissue - like matter . the simplicity enables analytical beam modeling for homogeneous systems and improves efficiency of numerical path integrals for heterogeneous systems . the present model is ideal for demanding dose calculations in treatment planning of heavy - charged - particle radiotherapy . hollmark m , uhrdin j , d b , gudowska i and brahme a 2004 influence of multiple scattering and energy loss straggling on the absorbed dose distributions of therapeutic light ion beams : i. analytical pencil beam model 324765 kanematsu n , akagi t , futami y , higashi a , kanai t , matsufuji n , tomura h and yamashita h 1998 a proton dose calculation code for treatment planning based on the pencil beam algorithm _ jpn phys . _ * 18 * 88103 matsufuji n , komori m , sasaki h , akiu k , ogawa m , fukumura a , urakabe e , inaniwa t , nishio t , kohno t and kanai t 2005 spatial fragment distribution from a therapeutic pencil - like carbon beam in water 3393403

dose calculation for radiotherapy with protons and heavier ions deals with a large volume of path integrals involving a scattering power of body tissue . this work provides a simple model for such demanding applications . there is an approximate linearity between rms end - point displacement and range of incident particles in water , empirically found in measurements and detailed calculations . this fact was translated into a simple linear formula , from which the scattering power that is only inversely proportional to residual range was derived . the simplicity enabled analytical formulation for ions stopping in water , which was designed to be equivalent with the extended highland model and agreed with measurements within 2% or 0.02 cm in rms displacement . the simplicity will also improve the efficiency of numerical path integrals in the presence of heterogeneity .

the milky way ( mw ) and andromeda galaxy ( m31 ) are the dominant galaxies in the local group ( lg ) . astronomical observations of their mass distribution impose constraints on the standard cosmological model . the satellite overabundance problem @xcite , tidal disruption features @xcite and the disk dominated morphology @xcite are examples of how lg studies are linked to the cosmological context . detailed studies on the magellanic clouds dynamics and their possible link to m31 add to the interest of understanding the details of the lg kinematics and dynamics @xcite . however , a general concern in the use of the lg as a tool for near - field cosmology @xcite is how typical is the lg regarding the properties of interest @xcite . a new valuable piece of information in this issue is the recent observational determination of the proper - motion measurements of m31 , which until recently had been out of reach @xcite . the reported measurements set an upper bound for the tangential velocity of m31 with respect to the mw of @xmath17 km s@xmath1 . together with the values of the relative radial velocity of @xmath18 km s@xmath1observations show that the relative motion of the mw and andromeda is consistent with a head - on collision . with this information it is possible to quantify how common such a kinematic configuration is in a @xmath3cdm universe . this letter presents such a study . we use a large volume , high resolution dark matter only n - body simulation in the concordance @xmath3cdm cosmology to find a set of halo pairs with similar characteristics as inferred in the lg . we quantify these results in terms of the number of pairs with given radial and tangential velocities in the galactocentric rest frame . we also find the pairs that are consistent with a head on collision in terms of the ratio of the radial to tangential velocity @xmath19 and present these results in terms of the reduced angular momentum , mechanical energy and dimensionless spin parameter . in addition we make use of three constrained n - body simulations which are constructed to reproduce the observed large scale structure of the local universe on scales of a few tens of mpc . the special feature of these simulations is that each volume features a pair of halos with the right characteristics to be considered lg - like objects . this letter is structured as follows . in the next section we present the n - body simulations and the criteria we use to select lg - like halo pairs . in section [ sec : results ] we present the results for the dynamics in these pairs in terms of the tangential / radial velocities and the orbital angular momentum / mechanical energy . in the same section we summarize these dynamical results in terms of the dimensionless spin parameter of the pairs . finally , in the last section we comment and conclude about the implications of these results in the context of the @xmath3cdm model . the bolshoi simulation follows the non - linear evolution of the dark matter density field using n - body techniques . the simulation has a cubic comoving volume of @xmath20 , sampled with @xmath21 particles . the cosmological parameters used in the simulation are @xmath22 and @xmath23 , corresponding to the matter density , vacuum energy density , the normalization of the power spectrum , the dimensionless hubble constant and the index of the slope in the initial power spectrum . this set of parameters is compatible with the analysis of the seventh year of data from the wilkinson microwave anisotropy probe ( wmap ) @xcite . a detailed description of this simulation can be found in @xcite . with these parameters the mass per particle is @xmath24 . in this paper we use halos obtained through the bound density maxima ( bdm ) algorithm @xcite . the halos are selected to have an overdensity of 200 times the critical density . furthermore , we only include in the analysis halos whose center is located outside the virial radius of any other halo . we have obtained the data through the public available multidark database @xcite . the database allows us to obtain the comoving positions , peculiar velocities and masses for all the halos in the simulation volume at @xmath25 . the positions and velocities of these haloes correspond to the average values of the @xmath26 most bound particles . the hubble flow is taken into account to convert the peculiar velocities into physical velocities and allow for a comparison with observations . we have verified that the main conclusions of this paper hold in the case of halos defined by a fof algorithm with a linking length 0.17 times the mean interparticle distance . the constrained simulations we use in this letter are part of the constrained local universe simulations ( clues ) project whose main objective is to reproduce the large scale structure in the local universe as accurately as possible . the algorithm and observational constraints to construct the initial conditions are described in @xcite . we use three dark matter only simulations , each has a cubic volume of @xmath27on a side , with the density field sampled with @xmath28 particles . the cosmological density parameter is @xmath29 , the cosmological constant @xmath30 , the dimensionless hubble parameter @xmath31 , the spectral index of the primordial density perturbations @xmath32 and the power spectrum normalization @xmath33 , also consistent with wmap 7th year data . based on the bdm catalogs in the bolshoi simulation we construct a halo pair sample with the dynamical properties consistent with those of the mw and m31 . the criteria we impose to define a lg - like halo pair are the following : 1 . each halo has a mass in the range @xmath34 . 2 . with respect to each halo , there can not be any other halo within the mass range @xmath34 closer than its partner . it means that there can not be ambiguity on the identity of the pair members . the relative radial velocity between the two halos is negative @xcite . 4 . the distance between the center of mass of the halos must be less than @xmath35mpc @xcite . there can not be halos more massive than @xmath36 within a radius of @xmath10mpc with respect to every object centre @xcite . there can not be halos more massive than @xmath37 within a radius of @xmath38mpc with respect to every object centre @xcite . throughout this letter we refer to this sample as the full sample . this sample in the bolshoi simulation has @xmath39 pairs . additionally , there is a sample of three ( 3 ) pairs constructed from the three constrained realizations . these pairs fulfill all the above mentioned conditions and additionally are located in a place with the right distances with respect to the virgo cluster in the simulation . the full observational characteristics that we take in this letter for the mw - m31 pair are listed in table [ table:1 ] . a more reduced sample from the full sample has been constructed so as to obey the observational bounds on the masses and separation of the two main halos . these amount to : 1 . the separation between the center of mass of the halos is in the range @xmath40kpc @xcite . the total mass of the two halos is in the range @xmath41@xcite . including these conditions the full sample is reduced from @xmath39 to @xmath42 pairs . we refer to this sample as the reduced sample . note that only one constrained lg - like object is included in this subset . .summary of kinematic observational constraints . [ cols="^,^,^ " , ] notes : + 1 . in columns 2 and 3 consistency is defined from 1-@xmath4 uncertainties for @xmath43-@xmath44 , @xmath5-@xmath45 and @xmath46 . in the case of @xmath47 , is defined from the constrain @xmath48 derived from the 1@xmath49 observational uncertainties in the radial and tangential velocities . 2 . in columns 4 and 5 the number of pairs around the preferred region in @xmath50 in @xmath3cdm ( as shown in fig . [ fig : rt ] and summarized in table [ table:2 ] ) are calculated using the same observational 1-@xmath4 absolute uncertainty on @xmath43 and @xmath44 . [ table:3 ] figure [ fig : rt ] summarizes the central finding of this letter . most of the pairs in the two samples constructed from the bolshoi simulation have radial and tangential velocities notably different from the observational constraints at the 1-@xmath4 level . the most probable radial and tangential velocities in @xmath3cdm are summarized in table [ table:2 ] . for the full sample we have @xmath51 km s@xmath1and @xmath52 km s@xmath1 . for the reduced sample @xmath53 km s@xmath1and @xmath52 km s@xmath1 , where the uncertainties in these values reflect the minimum grid size needed to obtain robust statistics for the 2d histogram . the number of pairs compatible with lg observations at the 1-@xmath4 level are listed in table [ table:3 ] ( columns 2 and 3 ) . in the same table ( columns 4 and 5 ) we summarize the number of pairs around the @xmath3cdm values within the same range of absolute observational uncertainty ( i.e. @xmath54 km s@xmath1and @xmath55 km s@xmath1 . ) from these results we infer that the pairs around the preferred phase space for @xmath3cdm are at least @xmath11 times more common than pairs with the observed velocities for the lg . we highlight that this is a lower bound given that in the reduced sample none of the pairs is found in the interval allowed by observations . the high eccentric orbit of the observed lg constitutes an unlikely configuration for the @xmath3cdm lg - like objects . this holds for the full and the reduced sample of objects . these conclusions are valid at 1-@xmath4 level in the observational uncertainties . an equivalence in the abundance between the pairs around the observational values and those around the preferred @xmath3cdm velocities is reached only at the 4-@xmath4 level on the observational uncertainties . from figure [ fig : rt ] it is also clear that there is a significant number of pairs with a high tangential - to - radial velocity ratio . the peak in the pair number density is located around a region of @xmath56 , while the observations suggest @xmath57 . as summarized in table [ table:3 ] we find that only between @xmath58 to @xmath59 of the pairs are consistent with the observational constraint . the three pairs from the constrained realizations are also show higher @xmath60 ratios than observations : @xmath61 . the two - body problem of point - like masses can serve as a proxy for the dynamics of the lg and can be used as a tool for studying the lg within the framework of the standard cosmological model . within the model the dynamics of the lg is governed by the ( center of mass ) reduced angular momentum @xmath62 and reduced energy @xmath63 , where the total mass is @xmath64 , the reduced mass is @xmath65 and @xmath66 is the gravitational constant . this formulation has a clear theoretical advantage if one considers the angular momentum and the mechanical energy as quasi - conserved dynamical quantities . this means that , after some formation time , these quantities do not significantly evolve . however , this formulation has an observational disadvantage . the reduced energy and angular momentum are derived from very different kinds of observations , increasing the uncertainties in their final determination . we use monte carlo sampling to estimate the reduced energy and angular momentum from the observed properties in the lg listed in table[table:1 ] . lower panels in figure [ fig : rt ] also presents the results in the plane @xmath67 contour in the lower panels of figure [ fig : rt ] does not cross the zero level , as can be naively expected from figure [ fig : rt ] . this happens for the @xmath68 contour level . ] . as expected , the @xmath5 and @xmath6 constraints are less restrictive compared with the radial and tangential velocity constraints ; 15% ( 24% ) of pairs in the full ( reduced ) sample obey the 1-@xmath4 observational constraints . we note that for the reduced sample the preferred region is outside the 1-@xmath4 observational contours . all the constrained lgs of the full and reduced sample are consistent with the 1-@xmath4 observational uncertainty . however a an hypothetical increase by a factor of @xmath69 in accuracy of the tangential velocity , showing that it is below @xmath70 km s@xmath1 , would bring these pairs and the @xmath3cdm expectation outside this uncertainty region . the dimensionless spin parameter @xmath7 @xcite is used here to characterize the dynamical state of the observed and simulated lgs . the parameter measures the dynamical role of the angular momentum in terms of the gravitational attraction and is defined by @xmath71 we compare the distribution for @xmath7 obtained from the pair population in the bolshoi simulation and the distribution from the monte carlo simulation used to estimate the uncertainties on @xmath5 and @xmath6 . figure [ fig : lambda ] shows the likelihood distribution of the observational errors and the spin parameter of the lg - like pairs of the full ( left panel ) and the reduced ( right panel ) samples . the value of @xmath7 of the constrained lgs is shown as well . the expected observational value for @xmath7 is slightly inconsistent with the statistics derived from halo pairs in @xmath3cdm . the spin parameter of the simulated lg - like objects is skewed towards higher values compared with the observational estimation . inspection of figure [ fig : lambda ] also shows that the @xmath7 distribution is close to a log - normal one and thus resembles the distribution of the spin parameter of dm halos in simulation @xcite . the estimated value of @xmath46 of the observed lg is @xmath72 , lower than the mean value of @xmath46 of @xmath73 ( @xmath74 ) for the full ( reduced ) sample . the values of the spin parameter of the 3 constrained lgs are @xmath75 , @xmath76 and @xmath77 , with the latter value belonging to the sole constrained lg of the reduced sample . we have presented a comparison between the observed kinematics for the m31 in the galactocentric rest - frame and the expectations for a large n - body cosmological simulation in the @xmath3cdm cosmology . in the simulation we select a large sample of pairs that fulfill isolation criteria for the lg . we select a sub - sample which obeys more stringent observational constraints on the mass and separation of the mw and m31 . while the observations show that m31 moves towards us on a highly eccentric orbit , the simulation shows that the most common configuration at @xmath25 has values @xmath78 km s@xmath1and @xmath79 km s@xmath1 , where the error bars are estimated to include the results from the two pair samples . using the same absolute values for the uncertainty in the observed velocity components , we find that none of the halos in the reduced sample are compatible with observations . this makes pairs with the preferred @xmath3cdm values at least @xmath11 times more common than pairs compatible with the observational constraint . additionally , pairs with a fraction of tangential to radial velocity @xmath57 ( similar to observations ) represent @xmath58 to @xmath59 of the pairs . approximating the lg as two point masses we express the above mentioned results in terms of the orbital angular momentum @xmath80 and the mechanical energy @xmath5 per unit of reduced mass . we find that the uncertainties in the tangential velocity , the square of the norm of the velocity and the total mass in the lg are less constraining on the number of simulated pairs that are consistent with the observations . nevertheless , in the case of the reduced sample there is a slight tension between simulation and observation . a reduction by a factor of @xmath69 in the observational uncertainty on the radial velocity would clarify this issue . we also use the @xmath7 spin parameter to gauge the dynamical state of pairs . the values for the orbital angular momentum and energy , merged into the @xmath7 spin parameter , are in mild disagreement with the observational constraints . in the three pairs from constrained simulations we find kinematics dominated by radial velocities . however their velocity components differ from the observational constraints and their mechanical energy and orbital angular momentum are in broad concordance with observations . there is only one pair that fulfills all the separation , total mass constraints and matches the most probable value for the dimensionless spin parameter @xmath7 inferred from observations . summarizing , we see a broad agreement in the total angular momentum and energy and a marked difference with the precise balance between today s radial and tangential velocities , i.e. head - on collisions of mw - like halos , as described by 1-@xmath4 uncertainties in observations , are not common in @xmath3cdm . under the approximation of conservation of the orbital angular momentum and mechanical energy , this could only be explained if the initial conditions for the formation of the local group are special in comparison to the initial conditions of any other pair of dark matter halos in the @xmath3cdm cosmology . investigating this perspective , by a thorough characterization of the small pair sample consistent with observations in terms of halo properties such as concentration , spin and place in the cosmic web @xcite is underway ( bustamante et al . in prep . ) . the results presented in this letter open a new window into the question of how unique , if at all , is the lg in a cosmological context . this will continue to be studied within the framework of the clues project . [ sec : conclusions ] jef - r acknowledges financial support from universidad de los andes through its _ fondo de apoyo a profesores asistentes _ and the peter and patricia gruber foundation through its fellowship administered by the iau . jef - r also acknowledges early discussions with alejandro garca that motivated this work . yh has been supported by the isf ( 1013/12 ) gy acknowledges support from mineco through research grants aya 2009 13875 -c03 - 02 , aya 2012 31101 and consolider syec csd 2007 0050 . support from comunidad . de madrid through astromadrid research grant is also acknowledge the data and source code and instructions to replicate the results of this paper can be found here ` https://github.com/forero/lg_kinematics/ ` . thanks to the ipython community @xcite . thanks to jessica kirkpatrick for releasing her python code to make nice plots of 2d histograms . the multidark database used in this paper was constructed as part of the activities of the german astrophysical virtual observatory as result of a collaboration between the leibniz - institute for astrophysics potsdam ( aip ) and the spanish multidark consolider project csd2009 - 00064 . the bolshoi simulation was run on the nasa s pleiades supercomputer at the nasa ames research center . the clues simulation was run at lrz munich . , n. , bennett , c. l. , dunkley , j. , gold , b. , greason , m. r. , halpern , m. , hill , r. s. , hinshaw , g. , kogut , a. , komatsu , e. , larson , d. , limon , m. , meyer , s. s. , nolta , m. r. , odegard , n. , page , l. , smith , k. m. , spergel , d. n. , tucker , g. s. , weiland , j. l. , wollack , e. , & wright , e. l. 2011 , , 192 , 14 , a. w. , irwin , m. j. , ibata , r. a. , dubinski , j. , widrow , l. m. , martin , n. f. , ct , p. , dotter , a. l. , navarro , j. f. , ferguson , a. m. n. , puzia , t. h. , lewis , g. f. , babul , a. , barmby , p. , bienaym , o. , chapman , s. c. , cockcroft , r. , collins , m. l. m. , fardal , m. a. , harris , w. e. , huxor , a. , mackey , a. d. , pearrubia , j. , rich , r. m. , richer , h. b. , siebert , a. , tanvir , n. , valls - gabaud , d. , & venn , k. a. 2009 , , 461 , 66

recent observations constrained the tangential velocity of m31 with respect to the milky way ( mw ) to be @xmath0 km s@xmath1and the radial velocity to be in the range @xmath2 km s@xmath1@xcite . in this study we use a large volume high resolution n - body cosmological simulation ( bolshoi ) together with three constrained simulations to statistically study this kinematics in the context of the @xmath3cdm . the comparison of the ensembles of simulated pairs with the observed lg at the 1-@xmath4 level in the uncertainties has been done with respect to the radial and tangential velocities , the reduced orbital energy ( @xmath5 ) , angular momentum ( @xmath6 ) and the dimensionless spin parameter , @xmath7 . our main results are : ( i ) the preferred radial and tangential velocities for pairs in @xmath3cdm are @xmath8 km s@xmath1 , @xmath9 km s@xmath1 , ( ii ) pairs around that region are @xmath10 to @xmath11 times more common than pairs within the observational values , ( iii ) @xmath12 to @xmath13 of lg - like pairs in @xmath3cdm have energy and angular momentum consistent with observations while ( iv ) @xmath14 to @xmath15 of pairs in the same sample show similar values in the inferred dimensionless spin parameter . it follows that within current observational uncertainties the quasi - conserved quantities that characterize the orbit of the lg , i.e. @xmath5 , @xmath16 and @xmath7 , do not challenge the standard @xmath3cdm model , but the model is in tension with regard to the actual values of the radial and tangential velocities . this might hint to a problem of the @xmath3cdm model to reproduce the observed lg . galaxies : kinematics and dynamics , local group , methods : numerical

an important topic of interest , in modeling of saturated flow through porous media , is the analysis of the phenomenon when there is a microstructure present in the rock matrix . some of the achievements in the preexisting literature address the case when the microstructure is periodic ; see @xcite for the analytical approach and @xcite for the numerical point of view . from a different perspective , in @xcite is presented the homogenization analysis for a non - periodic fissured system where the geometry of the cracks surface satisfies @xmath0-smoothness hypotheses . however , none of these mathematical analysis accomplishments , takes in consideration a fractal geometric structure of porous media , which is an important case due to the remarkable evidence of this fact ; in particular in @xcite the authors found that pore space and pore interface have fractal features . also , see @xcite for pore structure characterization , including random growth models . see @xcite for fractal geometry results in tracing experiments within a porous medium including dispersion , fingering and percolation . finally , @xcite discusses the use of fractal surfaces in modeling the storage phenomenon in gas reservoirs . this paper concentrates on finding adequate fluid transmission conditions for flow in porous media with fractal interface , as well as the well - posedness of the corresponding weak variational formulations . our goal is to blend " the modeling of porous media flow with the fractal roughness of the microstructure . we differ from the previously mentioned achievements since we replace the geometric feature of periodicity by that of self - similarity and explore , numerically , the effect of some randomness in the fractal geometry . on the other hand , the present study has a very different approach to the analysis on fractals from the preexisting literature , given that the mainstream pde analysis on fractals concentrates its efforts in solving strong forms on the fractal domain @xcite , the analysis of the associated eigenvalues and eigenfunctions @xcite , or the study of the adequate function spaces @xcite . in order to start understanding the key features of the phenomenon , we limit the study to the 1-d setting , defining the domain of analysis as @xmath1 . additionally , we will use and adjustment of the classic stationary diffusion problem below , in order to introduce the fluid exchange transmission conditions across the fractal interface @xmath2 here @xmath3 stands for the pressure , @xmath4 indicates the flux according to darcy s law and @xmath5 denotes de permeability , which will be set as @xmath6 throughout this work . in addition , we set dirichlet and neumann boundary conditions on the extremes of the interval . we use classical notation and results on function spaces @xmath7 and indicate with standard letters @xmath8 the functions on these spaces . we adopt the letters @xmath9 , @xmath10 to denote the microstructure and its elements respectively , in particular , the following classical hilbert space will be frequently used [ eqn hilbert space on fractal ] @xmath11 endowed with its natural inner product @xmath12 in the next section , we introduce the geometry of the fractal interface , together with the adequate mathematical setting in order to include it successfully in the pde model . throughout this work we limit to a particular type of fractal microstructure , first we need to introduce a previous definition and a related result ( see @xcite ) let @xmath13 be a closed set a. a function @xmath14 is said to be a * contraction * if there exists a constant @xmath15 such that @xmath16 for all @xmath17 . additionally , we say that @xmath18 is a * similarity * if @xmath19 for all @xmath17 and @xmath20 is said to be its * ratio*. b. a finite family of contractions , @xmath21 on @xmath22 with @xmath23 is said to be an * iterated function system * or * ifs*. c. a non - empty compact subset @xmath24 is said to be an * attractor * of the ifs @xmath21 if @xmath25 in particular , if every contraction of the ifs is a similarity then , the attractor @xmath26 is said to be a * strictly self - similar * set . consider the ifs given by the contractions @xmath21 on @xmath13 then , there is a unique attractor @xmath24 satisfying the identity . see _ theorem _ 9.1 @xcite . [ rem the attractor ] in this work , the attractor @xmath26 of an ifs will prove to be important in an indirect way : not for the definition of a microstructure , but for analyzing the nature of the attained conclusions . it is a well - known fact that under certain conditions ( see @xcite , _ lemma _ 9.2 ) a strictly self - similar set @xmath26 has both , hausdorff and box dimensions which are equal ( see @xcite , _ theorem _ 9.3 ) , namely @xmath27 . moreover , if @xmath28 is the ratio of the similarity @xmath29 , then @xmath30 from now on , we limit our attention to fractal structures satisfying strict self - similarity . finally , we introduce the type of microstructure to be studied in throughout this work . [ def fractal microstructure ] we say that a set @xmath31 $ ] is a * fractal microstructure * if it is countable and there exists a sequence of finite subsets @xmath32 , together with an iterated function system of similarities @xmath33 on @xmath34 $ ] , satisfying the following conditions a. the set @xmath35 is finite and @xmath36 is recursively defined by @xmath37 b. the sequence of sets is monotonically increasing and @xmath38 in the following , we refer to @xmath39 as a @xmath40**-finite development * * of @xmath9 . [ rem properties of the sigma finite development ] let @xmath31 $ ] be a fractal microstructure and let @xmath33 be its corresponding system of similarities , notice the following a. there may exist more than one @xmath41-finite development of @xmath9 . b. given a @xmath41-finite development @xmath32 then , for each @xmath42 , the following relationships of cardinality must hold @xmath43 c. the fractal microstructure @xmath9 is necessarily contained in the unique fractal attractor @xmath26 of the ifs of similarities @xmath33 . let @xmath9 be a microstructure set , let @xmath44 be a @xmath41-finite development and consider the following sequence of strong interface problems [ pblm strong problem n - stage ] @xmath45 .\ ] ] with the interface conditions @xmath46 and the boundary conditions @xmath47 here , @xmath48 is a monotone ordering of @xmath49 i.e. , @xmath50 . the forcing term @xmath26 belongs to @xmath51 , @xmath52 is a storage fluid exchange coefficient , and @xmath53 is a source on the interface . in order to attain the variational formulation of the problems above we define the following function space [ def function space setting ] define the space @xmath54 endowed with the inner product @xmath55 @xmath56 and the norm @xmath57 . the variational formulation of the _ problem _ above is given by @xmath58 [ th boundedness of the sequence of solutions ] the problem is well - posed . moreover if @xmath59 the sequence of solutions @xmath60 is bounded in @xmath61 and for each @xmath62 it holds that @xmath63 clearly , the bilinear form @xmath64 is continuous and @xmath61-elliptic . in addition , it is direct to see that @xmath65 is linear and continuous . consequently , the well - posedness of _ problem _ follows from the lax - milgram theorem ( see @xcite ) . for the boundedness of @xmath66 , we test with @xmath67 and get @xmath68 the second and third lines were obtained applying the cauchy - schwartz inequality in @xmath51 , @xmath69 and @xmath70 respectively . also , @xmath71 is the poincar constant associated to the domain @xmath72 ; in this particular case @xmath73 ( see @xcite ) . hence , @xmath74 hence , the _ estimate _ follows . in the theorem above , particularly due to the a - priori _ estimate _ , we observe that the sequence of solutions @xmath75 contains extra information , which is given by the boundedness of the term @xmath76 . implicitly , this fact gives the subspace of convergence . [ def space of convergence ] define the * fractal interface space * @xmath77 endowed with the inner product @xmath78 @xmath79 and the norm @xmath80 . [ rem the little ell 2 equivalence ] a. in the following we refer to the trace operator @xmath81 , @xmath82 as the * fractal trace operator * on @xmath9 , which is well - defined for functions @xmath83 . b. the space @xmath84 is the subspace of functions @xmath85 whose trace on the microstructure set , @xmath86 , belongs to @xmath87 . c. the space @xmath84 is a hilbert space due to the completeness of @xmath61 with the norm @xmath88 and the completeness of @xmath87 with the norm @xmath89 . [ th weak convergence of the whole sequence ] let @xmath90 be the sequence of solutions to the family of _ problems _ . then , if it converges weakly in @xmath61 to an element @xmath91 , the sequence of functionals @xmath92 defined by @xmath93 converges weakly to @xmath94 in @xmath95 . in particular , @xmath96 and @xmath97 { } \sum_{\b \ , \in \ , \b } \xi(\b ) \ , q(\b ) \ , , \quad \forall\ , q\in \vb.\ ] ] clearly @xmath98 for all @xmath62 and due to the riesz representation theorem , we have that @xmath99 where the last inequality holds due to the _ estimate _ . consequently , there must exist a subsequence @xmath100 and an element @xmath101 such that @xmath102 { } \sum_{\b \ , \in \ , \b } \eta ( \b)\ , u(\b ) \,,\quad \forall \ , u\in \ell^ { \ , 2}(\b).\ ] ] in particular , for any @xmath103 , let @xmath104 be such that @xmath105 for all @xmath106 then , recalling that @xmath107 , the expression above yields @xmath108 { } \sum_{\b \ , \in \ , \b } \eta ( \b)\ , \ , \ind_{\{\b_{0}\}}(\b ) = \eta ( \b_{0 } ) \ , , & & \forall \ ; \b_{0}\in\b.\end{aligned}\ ] ] on the other hand , since @xmath109 this implies that @xmath110 for all @xmath111 $ ] . in particular @xmath112 for all @xmath103 , consequently @xmath113 and @xmath96 . moreover , since the above holds for any convergent subsequence of @xmath114 it follows that the whole sequence is weakly convergent i.e. , @xmath115 from here , the convergence statement follows trivially . [ rem partial interface development functionals ] it is important to observe that the weak convergence of the functionals @xmath116 in @xmath117 hypothesis is a stronger condition than the _ statement _ , since it can not be claimed that the fractal trace operator @xmath118 , @xmath119 is surjective . moreover , it will be shown that in most of the interesting cases this operator is not surjective . consider the variational problem with microstructure interface @xmath120 where , @xmath121 and @xmath122 . we claim that the solution @xmath3 of the problem above is the weak limit @xmath123 i.e. , _ problem _ is the * limit * " of _ problems _ . [ th well - posedness fractal interface problem ] the _ problem _ is well - posed . consider the bilinear form @xmath124 using the cauchy - schwartz inequality in each summand of the expression above , we conclude the continuity of the bilinear form . on the other hand , it is direct to see that @xmath125 , which implies that the bilinear form is @xmath84-elliptic . applying the lax - milgram theorem , the result follows , see @xcite . now we are ready to prove the weak convergence of the whole sequence of solutions of _ problems _ to the solution @xmath3 of problem . [ th weak convergence of the solutions in v ] let @xmath90 be the sequence of solutions of _ problems _ , then , it converges weakly in @xmath61 to the unique solution @xmath3 of _ problem _ . since @xmath60 is bounded in @xmath61 there must exist a weakly convergent subsequence @xmath126 and a limit @xmath91 . test with @xmath127 arbitrary , this gives @xmath128 letting @xmath129 in the expression above and , in view of _ theorem _ [ th weak convergence of the whole sequence ] , we get @xmath130 additionally , _ [ th weak convergence of the whole sequence ] implies that @xmath96 . therefore @xmath131 is a solution to _ problem _ , which is unique due to _ theorem _ [ th well - posedness fractal interface problem ] ; therefore we conclude that @xmath132 . since the above holds for any weakly convergent subsequence of @xmath133 it follows that the whole sequence must converge weakly to @xmath3 . [ th norms convergence ] let @xmath123 be the sequence of solutions of _ problems _ then [ stmt strong convergence ] @xmath134 { } 0 .\ ] ] @xmath135 { } 0 .\ ] ] where @xmath136 is the fractal trace operator on @xmath61 . we know that @xmath137 weakly in @xmath61 and @xmath138 weakly in @xmath139 from _ theorems _ [ th weak convergence of the solutions in v ] and [ th weak convergence of the whole sequence ] respectively , therefore @xmath140 on the other hand , testing on the diagonal @xmath67 , we get @xmath141 letting @xmath142 in the expression above we get @xmath143 hence @xmath144 from here , a simple exercise of real sequences shows that both sequences @xmath145 and @xmath146 converge . combining these facts with _ inequality _ we have @xmath147 therefore , if @xmath148 or @xmath149 , the equality above could not be possible . then , it holds that @xmath150 and @xmath151 finally , the convergence of norms together with the weak convergence on the underlying spaces , imply the strong convergence of both sequences . we start this section proving a lemma which is central in the understanding of the space @xmath84 . [ th microstructure accumulation points ] let @xmath9 be a microstructure in @xmath34 $ ] and let @xmath152 be such that @xmath153 with @xmath154 . then , if @xmath155 is an accumulation point of @xmath9 , it must hold that @xmath156 . let @xmath157 satisfy the hypotheses and let @xmath155 be an accumulation point of @xmath9 such that @xmath158 . since @xmath152 , it is absolutely continuous and there exists @xmath159 such that @xmath160 for all @xmath161 ; hence @xmath162 \vert \b - \b_{0}\vert < \epsilon } } \vert q(\b ) \vert^{p } \geq \operatorname*{card}\big(\ { \b\ , \in \,\b : \ \vert \b - \b_{0}\vert < \epsilon\}\big ) \ , \frac{1}{2^{p}}\,\vert q(\b_{0 } ) \vert^{p}.\ ] ] however , the set @xmath163 contains infinitely many points because @xmath164 is an accumulation point of @xmath9 , therefore @xmath165 does not belong to @xmath166 which is absurd . [ th dense microstructure ] suppose that @xmath9 is dense in @xmath34 $ ] then @xmath167 . due to _ lemma _ [ th microstructure accumulation points ] if @xmath168 it must hold that @xmath169 for every @xmath10 accumulation point of @xmath9 . since @xmath9 is dense in @xmath72 every point of @xmath9 is an accumulation point of @xmath9 , therefore @xmath169 for all @xmath170 . on the other hand , @xmath157 is absolutely continuous because it belongs to @xmath171 and due to the density of @xmath9 in @xmath34 $ ] it follows that @xmath172 . the _ lemma _ [ th microstructure accumulation points ] states that the accumulation points of the microstructure contained in @xmath9 define an important property of @xmath84 . in addition , the next property follows trivially [ th dense microstructure trivializes ] let @xmath31 $ ] be a dense microstructure then , the _ problem _ becomes trivial and the sequence @xmath66 satisfies @xmath173 the facts presented in _ lemma _ [ th microstructure accumulation points ] and _ corollary _ [ th dense microstructure trivializes ] are unfortunate , since several important fractal microstructures are dense in @xmath34 $ ] . on the other hand , if a microstructure @xmath9 is not dense in @xmath34 $ ] but it is a perfect set ( which is also an important case ) the _ problem _ , without becoming trivial , becomes fully decoupled and consequently uninteresting . an important example of the first case are the dyadic numbers in @xmath34 $ ] and an example of the second case is the collection of extremes of the removed intervals in the construction of the cantor set . as an alternative , it is possible to strengthen the conditions on the interface forcing term @xmath53 , seeking to weaken the summability properties of the limit function fractal trace @xmath174 . however , this approach yields estimates equivalent to _ inequality _ in _ theorem _ [ th boundedness of the sequence of solutions ] and consequently @xmath175 . therefore , this is not a suitable choice either . the unsatisfactory conclusions shown in the previous section are , essentially , due to a physical fact assumed in the model : that the storage fluid exchange coefficient @xmath176 is constant all over the microstructure @xmath9 . consequently , the storage effect across the microstructure adds up to infinity . hence , the modeling of @xmath176 has to avoid this hypothesis . on one hand , we need to assure that the form @xmath177 , defined by @xmath178 is bilinear and continuous . on the other hand , _ lemma _ [ th microstructure accumulation points ] states the need to avoid global estimates for @xmath179 for any @xmath180 . therefore , if @xmath181 and recalling that @xmath182 for all @xmath183 , the bilinear form satisfies @xmath184 in order to attain this condition , it is natural to assume that the storage coefficient @xmath176 , scales consistently with the properties of the fractal microstructure . this motivates the following definition [ def scaled storage coefficient ] let @xmath31 $ ] be a fractal microstructure with @xmath185 as given in _ definition _ [ def fractal microstructure ] . then , a storage coefficient @xmath186 is said to * scale consistently * with a given @xmath41-finite development @xmath187 , if it satisfies @xmath188 [ th scaled storage coefficient convergence ] let @xmath186 be a storage coefficient consistently scaled with @xmath9 then @xmath189 . since @xmath190 is the @xmath41-finite development of @xmath9 with @xmath191 , the cardinality _ identity _ implies @xmath192 since @xmath193 , as stated in _ definition _ [ def scaled storage coefficient ] , the expression above is convergent and the result follows . it is immediate to see some variations of _ definition _ [ def scaled storage coefficient ] based on the geometric series properties , for instance if @xmath194 is a sequence such that @xmath195{\alpha_{n } } < \frac{1}{l}$ ] then , consider @xmath196 the storage coefficient defined above will also satisfy that @xmath189 and permit the desired _ estimate _ . a more sophisticated variation includes probabilistic uncertainty for the values @xmath197 , whether on its decay rate or the distributed values of tolerance centered at the self - similarity parameter @xmath198 . a very basic probabilistic version of the latter will be numerically illustrated in _ section _ [ sec numerical experiments ] . in the following it will be assumed that @xmath121 . for simplicity of notation we understand that both , the storage coefficient and the interface forcing term are defined on the whole domain @xmath34 $ ] , with @xmath199 . it is also assumed that the storage coefficient @xmath200 scales consistently with the fractal microstructure @xmath9 and that the forcing term is summable @xmath201 . [ th well posedness good coefficient problems ] the following problems are well - posed . @xmath202 @xmath203 consider the bilinear forms [ def bilinear forms scaled coefficient ] @xmath204 @xmath205 since @xmath189 , the _ inequality _ is satisfied and the bilinear forms @xmath206 , @xmath207 for each @xmath62 are continuous ; consequently the bilinear forms are also continuous . on the other hand , since the coefficient @xmath176 is non - negative the bilinear forms are both coercive on @xmath61 . applying the lax - milgram theorem the result follows . let @xmath66 be the sequence of solutions of problems and let @xmath3 be the solution to problem . then a. @xmath66 converges weakly to @xmath3 in @xmath61 . b. the following convergence statements hold + [ stmt convergence interface functionals ] @xmath208 { } \sum_{\b \ , \in \ , \b } \beta(\b)\ , p^{2 } ( \b).\ ] ] @xmath209 { } \sum_{\b \ , \in \ , \b } f(\b)\ , p ( \b).\ ] ] c. @xmath66 converges strongly to @xmath3 in @xmath61 . a. the result follows using the techniques presented in _ theorem _ [ th boundedness of the sequence of solutions ] yield the boundedness of the sequence @xmath66 and using the reasoning of _ theorem _ [ th weak convergence of the solutions in v ] the weak convergence follows . b. due to the weak convergence of the solutions and the fact that the evaluation is a continuous linear functional , it follows that @xmath210 and @xmath211 for all @xmath170 . let @xmath212 be such that @xmath213 and @xmath214 for all @xmath62 . fix @xmath215 such that @xmath216 implies @xmath217 , which we know to exist since @xmath189 . then , for any @xmath216 it holds that @xmath218 since @xmath215 is fixed choose @xmath219 such that @xmath220 implies @xmath221 hence , combining with the previous estimate , the convergence _ statement _ follows . for the _ statement _ it is enough to combine the strong convergence of the forcing terms @xmath222{}0 $ ] with the weak convergence of the solutions @xmath137 . this completes the second part . c. again , the result follows combining the convergence _ statements _ and , with the techniques presented in _ lemma _ [ th norms convergence ] used to attain the strong convergence of the solutions . next , we present the closest version of a strong form of _ problem _ . [ th pseudo strong form ] let @xmath31 $ ] be a microstructure and let @xmath223 be its set of limit points . then , _ problem _ is a weak formulation of the following strong problem . [ pblm strong microstructure good coefficients ] @xmath224 @xmath225 together with the interface fluid transmission conditions for isolated points of @xmath9 @xmath226 and the non - localizable fluid transmission conditions for limit points of @xmath9 @xmath227 the boundary condition holds because @xmath228 . let @xmath229 , then , there exists @xmath230 such that @xmath231 for all @xmath232 . problem _ with @xmath233 to get @xmath234 since the above holds for each smooth function on @xmath235 , we conclude that @xmath236 in @xmath237 . similarly , it follows that @xmath236 in @xmath238 , therefore @xmath239 and the statement follows . now , choose @xmath240 and test _ problem _ to get @xmath241 next , in the expression above , we break the interval conveniently in order to use integration by parts @xmath242 from the initial part of the proof , the first summand of the left hand side cancels with the first two summands of the right hand side . evaluating the boundary terms and recalling that @xmath243 this gives @xmath244 consequently , the normal flux balance condition in follows for points @xmath245 . the normal stress balance condition in follows from the continuity of @xmath3 across the interface which holds for any function in @xmath246 . the normal stress balance in _ statement _ holds due to the continuity of @xmath3 at any point of @xmath183 . now , fix @xmath247 and test the _ problem _ with @xmath248 , where @xmath249 ; this gives @xmath250 taking limits in the expression above and recalling the lebesgue dominated convergence theorem the result follows . notice that , by definition of distributions , the following equality holds in the sense of distribution for the solution of _ problem _ @xmath251 where @xmath252 is the dirac evaluation functional . it is easy to observe that if @xmath253 is a limit point of @xmath9 , both series in the expression above will pick up infinitely many non - null terms for test functions such that @xmath254 . therefore , the techniques used in _ theorem _ [ th pseudo strong form ] to derive point - wise statements from weak variational ones , do not apply . again , this fact is unfortunate because , as pointed out at the end of _ section _ [ sec a sigma finite interface microstructure ] , several important fractal microstructures are dense in @xmath34 $ ] or perfect sets . if @xmath9 is dense very little can be said about the point - wise behavior at any point of the domain and if @xmath9 is a perfect set , the transmission conditions can not be described locally at any point of the interface . a positive aspect of this model is that the aforementioned cases do not become trivial or uninteresting . additionally , if the set of accumulation points @xmath255 has null lebesgue measure the non - localizable interface conditions are not relevant for the global phenomenon . fortunately , this is the case for very important fractal microstructures e.g. , the collection of extremes of the removed intervals in the construction of the cantor set , the sierpinski triangle in 2-d , etc . in this section we present two types of numerical experiments . the first type are verification examples , supporting our homogenization conclusions for a problem whose asymptotic behavior is known exactly . the second type are of exploratory nature , in order to gain heuristic understanding of the probabilistic variations of the model . the experiments are executed in a matlab code using the finite element method ( fem ) ; it is an adaptation of the code * fem1d.m * @xcite . in all the examples to be presented below the following holds . for the sake of simplicity , the microstructure @xmath31 $ ] is the collection of extremes of the removed intervals in the construction of the cantor set . in addition , its @xmath41-finite development @xmath187 is the natural one i.e. , @xmath256 for the experiments it will be shown that the sequences @xmath66 are cauchy ; where @xmath62 indicates de stage of the cantor set construction . in all the cases , the computations are made for the stages @xmath257 . for each example we present graphics for values of an @xmath258-stage chosen from @xmath259 , based on optical neatness ; two or three graphs for the solution and the graphs of the corresponding derivative . also , for visual purposes we introduce vertical lines in the graph of the derivatives , to highlight the fractal structure of the function . for all the examples we set the forcing term @xmath121 equal to zero and he interface forcing term @xmath260 , is given by the following expression @xmath261 [ ex basic example ] in the first example we implement the model studied in _ [ sec a sigma finite interface microstructure ] , in order to verify its conclusions . in particular , it must hold that @xmath262 . we set the storage coefficient @xmath263 and the forcing terms @xmath264 , @xmath53 defined in above . the convergence table is displayed below , together with the corresponding graphics . _ example _ [ ex basic example ] : convergence table , @xmath263 . [ cols="^,^,^,^,^,^",options="header " , ] in the table below the @xmath265 and @xmath266-norms for the difference of the partial averages @xmath267 and the cesro average are displayed . clearly convergence is observed in both cases and , due to the law of large numbers , better behavior is observed at the stage @xmath268 over the stage @xmath269 , as expected . the average and variance are also presented at the bottom of the table , which shows better behavior for the more developed stage @xmath270 over the stage @xmath269 not only in terms of centrality , but also in terms of deviation . + a. the authors tried to find , experimentally , a rate of convergence using the well - know estimate @xmath271 the sampling was made on the sequence of stages @xmath272 . experiments were run for _ examples _ [ ex basic example ] and [ ex scaled example ] however , in none of the cases , solid numerical evidence was detected that could suggest an order of convergence for the phenomenon . b. additional experiments for the random behavior _ example _ [ ex random behavior example ] were executed . the probabilistic variations involved were a. experiments for the stages @xmath273 and @xmath274 . b. experiments with different number of realizations , namely @xmath275 , @xmath276 and @xmath277 , depending on the computational time demanded . c. using a storage coefficient @xmath278 , with @xmath279 different intervals and @xmath176 with uniform distribution normal distribution . d. probabilistic variations of the forcing terms . e. combinations of one or more of the previous factors . f. execution of the aforementioned probabilistic variations adjusted to the scaled _ [ ex scaled example ] . + in all the cases , convergence behavior was observed as expected . naturally , the quality of convergence deteriorates depending on the deviation of the distributions and the combination of uncertainty factors introduced in the experiment . c. all the previously mentioned scenarios were also executed in the same code for the case of homogeneous dirichlet boundary conditions on both ends i.e. , @xmath280 . as expected convergence behavior is observed which is comparable with the corresponding analogous version setting the boundary conditions @xmath281 , @xmath282 , used along the analysis of this paper . the present work yields several accomplishments as well as limitations listed below . a. the unscaled storage model presented in _ section _ [ sec a sigma finite interface microstructure ] is in general not adequate , as it excludes most of the important cases of fractals . b. the scaled storage model presented in _ section _ [ dec fractal scaling modeling ] is suitable for an important number of fractal microstructures . however , the asymptotic variational model is equivalent to a pointwise strong model only if the closure of the microstructure @xmath283 is negligible i.e. , if it has null lebesgue measure . such hypothesis excludes important cases e.g. the family of fat " cantor sets , see @xcite in 1-d or the fat " versions of the classic fractal structures in 2-d or 3-d , where only the averaged normal flux " _ statement _ can be concluded for the accumulation points of the microstructure @xmath223 . c. in order to overcome the deficiency previously mentioned a first approach would be to take the traditional treatment of solving strong forms in fractal domains ( e.g. @xcite ) and then try to blend " it with the point of view presented here . such analysis is to be pursued in future research . d. the fractal microstructures addressed in this work are self - similar . this requirement is important only for the second model ( _ section _ [ dec fractal scaling modeling ] ) because it scales the storage according to the geometric detail of the structure at every level . in particular , we need an accurate idea of the growth rate of the microstructure @xmath9 from one level to the next , in order to scale the storage adequately . e. the self - similarity requirement for the microstructure @xmath9 in the introduction of the scaled model excludes the important family of the self - affine fractals ; this type of microstructures is a topic for future work . on the other hand , the unscaled model of _ section _ [ sec a sigma finite interface microstructure ] does not require such detailed knowledge of the microstructure precisely because it avoids scaling , although it is likely to be unsuited for the analysis of self - affine microstructures it may be a good starting point in order to detect the needs of the modeling for this case . f. it is important to observe the relevance of self - similarity versus the fractal dimension in the scaled model . while the self - similarity of the microstructure is the corner stone of the scaling ( hence it can not be given up ) the model is more flexible with respect to the fractal dimension of @xmath9 as long as it is not the same of the host " domain @xmath284 . g. we also stress that the input needed by the present result , in terms of geometric information on the fractal structure , does not have to be as detailed as in the strong forms pde analysis on fractals . h. the random experiments presented in _ example _ [ ex random behavior example ] , as well as those only mentioned in _ subsection _ [ sec closing observations numerical ] , furnish solid numerical evidence of good behavior for probabilistic versions of the unscaled and the scaled models respectively . additionally , it is important to handle certain level of uncertainty because , assuming to have a deterministic description of the fractal microstructure is a too strong hypothesis to be applicable in realistic scenarios . in the authors opinion , this is justification enough to pursue rigorous analysis of these problems to be addressed in future work . i. in _ example _ [ ex random behavior example ] uncertainty was introduced in the storage coefficient @xmath176 , or the forcing term @xmath53 , however the geometry of the microstructure was never randomized . in several works ( e.g. @xcite ) the self - similarity is replaced by the concept of statistical self - similarity in the sense that scaling of small parts have the same statistical distribution as the whole set . clearly , this random property is consistent with the scaling of @xmath176 in _ definition _ [ def scaled storage coefficient ] ; consequently , the statistical self - similarity is a future line of research for handling geometric uncertainty of the fractal microstructure @xmath9 . in particular , the fractal percolation microstructures , are of special interest for real world applications , see @xcite . j. the execution of all the numerical experiments shows that the code becomes unstable beyond the 9th stage of the cantor set construction . this suggests that in order to overcome these issues , an adaptation of the fem method has to be developed , targeted to the microstructure of interest . this aspect is to be analyzed in future research . k. the study of fractal microstructures in 2-d and 3-d are necessary for practical applications and in higher dimensions for theoretical purposes . given that passing from one dimension to two or more dimensions increases significantly the level of complexity in the microstructure and in the equation , considerable challenges are to be expected in this future line of research . l. another important analysis to be developed is the study of the models both , scaled and unscaled , in the mixed - mixed variational formulation introduced in @xcite . on one hand , this approach allows great flexibility for the underlying spaces of velocity and pressure which can constitute an advantage with respect to the treatment presented here . on the other hand , the mixed - mixed variational formulation allows modeling fluid exchange conditions across the interface of greater generality than , used in the present work ; this fact can contribute significantly to the development of the field . the authors wish to acknowledge universidad nacional de colombia , sede medelln for its support in this work through the project hermes 27798 . the authors also wish to thank professor magorzata peszyska from oregon state university , for authorizing the use of code * fem1d.m * @xcite in the implementation of the numerical experiments presented in _ section _ [ sec numerical experiments ] . it is a tool of remarkable quality , efficiency and versatility which has contributed significantly to this work .

we seek suitable exchange conditions for saturated fluid flow in a porous medium , where the interface of interest is a fractal microstructure embedded in the porous matrix . two different deterministic models are introduced and rigorously analyzed . also , numerical experiments for each of them are presented to verify the theoretically predicted behavior of the phenomenon and some probabilistic versions are explored numerically to gain further insight . coupled pde systems , fractal interface , porous media . 35q35 , 37f99 , 76s05

the authors thank jian - wei qiu and enke wang for helpful discussion . this work was supported by alexander von humboldt foundation , by bmbf , by nsfc under project no . 10405011 , by moe of china with project no . cfkstip-704035 , the u.s . department of energy under contract no . de - ac02 - 05ch11231 , and by the us nsf under grant no . phy-0457265 , the welch foundation under grant no . a-1358 . m. gyulassy , i. vitev , x. n. wang and b. w. zhang , _ quark - gluon plasma 3 _ , r. c. hwa and x .- n wang , eds . ( world scientific , singapore , 2003 ) . j. w. qiu and g. sterman , int . j. mod . e * 12 * ( 2003 ) 149 . x. f. guo and x .- wang , phys . lett . * 85 * , 3591 ( 2000 ) ; x .- n . wang and x. f. guo , nucl . a * 696 * , 788 ( 2001 ) . b. w. zhang and x .- wang , nucl . a * 720 * , 429 ( 2003 ) ; b. w. zhang , e. wang and x. n. wang , phys . lett . * 93 * , 072301 ( 2004 ) ; nucl . phys . a * 757 * , 493 ( 2005 ) ; b. w. zhang , [ arxiv : nucl - th/0411076 ] .

quark - quark double scattering in ea dis and its contribution to quark and anti - quark fragmentation functions are investigated with the generalized factorization of the relevant twist - four processes in pqcd . it is shown that the resulting modifications to quark and anti - quark fragmentation functions are different . while the numerical size of these effects can not be determined from pqcd , the structure of our result leads to a number of qualitative predictions for the relative size of the effect for different hadrons . these qualitative predictions agree with the multiplicity ratios for positive and negative hadrons as observed by hermes . , and multiple scattering in a nuclear medium is a key process to understand how hard processes are modified in nuclei . detailed experimental information , which is available for such diverse reactions as electron - nucleus collisions ( @xmath0 ) in fixed target kinematics and hadron - nucleus collisions ( @xmath1 ) and nucleus - nucleus collisions ( @xmath2 ) at rhic or lhc is confronted by highly advanced theoretical descriptions @xcite . still , however , a truly comprehensive and reliable description of all phenomena has not been reached . in this contribution we discuss one element which should be contained in such an ultimate theoretical description . we analyse @xcite quark - quark double scattering in nuclei and consider the nuclear effects on the quark and anti - quark fragmentation functions ( ff ) , using the generalized factorization theory by lqs @xcite in perturbative qcd . we find that quark - quark double scattering will give different corrections to quark and anti - quark ff , and obtain qualitative predictions which are in agreement with quite remarkable observation by the hermes collaboration @xcite . in general , there are two different kinds of double scattering with nuclear enhancement in ea dis : quark - gluon double scattering and quark - quark double scattering . quark - gluon double scattering gives the dominant contribution to the leading quark energy loss and has been studied in detailed @xcite , while quark - quark double scattering may mixes quark and gluon fragmentation functions and therefore gives rise to new nuclear effects , as pointed out by wang and guo @xcite . this is illustrated in fig . [ fig:4q - ex](a ) . for the central - cut diagram , the final parton is a gluon fragmenting into the hadron considered . for the left - cut or right - cut diagrams , however , the final fragmenting parton is a quark . therefore , adding up the contributions of these three cut diagrams we will get a mixture of quark and gluon fragmentation functions . moreover , we can see from the central cut of diagram ( a ) that the quark jet is converted into a gluon jet by annihilation with an anti - quark in the nucleus , which provides a mechanism to change the jet flavor @xcite . for quark - quark double scattering , there are three cut diagrams in leading - order as shown in fig . [ fig:4q - ex](a ) , and 46 cut diagrams in next - to - leading order , see fig . [ fig:4q - ex](b ) @xcite . by summing all these contributions and including virtual corrections we can rewrite the semi - inclusive tensor in terms of a modified fragmentation function taking into account quark - quark double scattering : @xmath3 where @xmath4 is the partonic part for single scattering , @xmath5 are the leading - twist quark fragmentation functions in the vacuum , and the @xmath6 are related to the twist - four parton correlation operators @xcite and could be approximated as : @xmath7 c\ , f_{\bar q}^n(x_t ) \nonumber \\ & & + \int_0^{\mu^2 } \frac{d\ell_t^2}{\ell_t^2 } \frac{\alpha_s^2 x_b}{x_a q^2 } \int_{z_h}^1 \frac{dz}{z } d_{q \to h}(\frac{z_h}{z } ) \frac{1+z^2}{(1-z)^2_+ } \nonumber \frac{c_f}{n_c } [ 1-e^{-x_l^2/x_a^2 } ] c\ , f_{\bar q}^n(x_t ) \nonumber \\ & & -\int_0^{\mu^2 } \frac{d\ell_t^2}{\ell_t^2 } \frac{\alpha_s^2 x_b}{x_a q^2 } \int_{0}^1 dz d_{q \to h}(z_h ) \frac{1+z^2}{(1-z)^2 } \nonumber \frac{c_f}{n_c } [ 1-e^{-x_l^2/x_a^2 } ] c\ , f_{\bar q}^n(x_t)\ , . \nonumber\end{aligned}\ ] ] here @xmath8 , @xmath9 is the quark distribution inside a nucleus , @xmath10 is the anti - quark distribution inside a nucleon and @xmath11 is assumed to be a constant , @xmath12 and @xmath13 is the typical transverse momentum carried by the antiquark . if we consider the modification to an anti - quark ff all the calculations are the same with @xmath14 . therefore , we have @xmath15 . hermes @xcite has measured the multiplicity ratio @xmath16 , which represents the ratio of the number of hadrons of type @xmath17 produced per dis event for a nuclear target of mass a to that from a deuterium target ( d ) : @xmath18 with @xmath19 the number of semi - inclusive hadrons in a given ( @xmath20)-bin , and @xmath21 the number of inclusive dis positrons in the same @xmath22-bin . in the constituent quark model , we have @xmath23 , @xmath24 . because the gluon ff for a positive hadron and its antiparticle are the same , and as we are interested only in the difference between the multiplicities of positive hadrons and negative hadrons , we neglect the contributions from gluon ff and simply assume : @xmath25 where @xmath26 is a constituent quark of hadron @xmath17 , and @xmath27 is the number of these constituent quarks . therefore we obtain @xmath28 \ , \ , , \label{eq : drm - pion } \\ & & \delta r_m^{p } \approx \delta\widetilde{d}_{q\to h}(z,\nu ) \propto f_{\bar q}^n(x_t ) \,\ , , \delta r_m^{\bar p } \approx \delta\widetilde{d}_{\bar q\to h}(z,\nu ) \propto f _ { q}^n(x_t ) \,\ , , \label{drm - proton } \\ & & \delta r_m^{\pi^+}(z,\nu ) \simeq \delta r_m^{\pi^- } \simeq \delta r_m^{\pi^0 } \,\ , , \label{eq : pion - order } \\ & & \delta r_m^{\bar p}(z,\nu ) > \delta r_m^{p}(z,\nu ) \,\ , . \label{eq : proton - order}\end{aligned}\ ] ] similarly we get @xmath29 the qualitative statements of eq . ( [ eq : pion - order ] ) , ( [ eq : proton - order ] ) , and ( [ eq : hadron - order ] ) agree with the observations made by the hermes collaboration @xcite .

the discovery of the top quark in @xmath5 collisions @xcite was a milestone for the standard model . the next steps beyond that milestone will be to improve the measurement of the top mass , and to test our understanding of qcd further by studying the details of its production and decay . we focus here on the `` single - lepton '' channel defined by @xmath6 with one top decaying hadronically ( @xmath7 , or its charge conjugate ) and the other decaying leptonically ( @xmath8 , or its charge conjugate , with @xmath9 or @xmath10 ) . this channel is experimentally favorable because the high @xmath11 lepton and large missing @xmath11 due to the neutrino provide a clean event signature that naturally discriminates against backgrounds . we will concentrate on the hadronically decaying top , since the treatment of the leptonically decaying one is complicated by errors in the measurement of the transverse momentum of the neutrino , via missing @xmath11 , and by ambiguity in its longitudinal momentum . we will not deal here with the question of how @xmath4 events are to be identified @xcite . rather we will concentrate on the problem of measuring @xmath12 once the events have been isolated . the mass determination relies on measuring the energies and directions of jets , and inferring from them the energies and directions of the original quarks . in addition to instrumental calibrations , this requires corrections for the hard - scale branching of partons , and for non - perturbative hadronization effects . these hard and soft qcd effects can be modeled by event generators such as herwig @xcite , which have been shown to describe the appropriate physics in @xmath13 decay at the @xmath14 collider lep @xcite , where the original @xmath15 energy is precisely known . part of the top quark analysis involves measuring the mass of the hadronically decaying @xmath2 boson . an eventual goal of the analysis should be to confirm the qcd effects in @xmath2 decay that are similar to those seen in great detail for @xmath13 decay . in the meantime , we can assume that the qcd effects are understood , and use the @xmath2 mass for help in determining the correct jet assignments in @xmath4 events . the decay @xmath16 is also a potentially useful tool to calibrate the detector , since the decay jets probe the detector at a wide range of angles and jet transverse momenta , and the @xmath2 mass is accurately known @xcite . this calibration is unavailable outside of top quark events , because hadronic @xmath2 decays are generally obscured by large qcd backgrounds @xcite . we will investigate two ideas that might improve the mass resolution . the first idea , examined in sect . [ sec : cones ] , is to use fixed cones in the rest frames of the decaying @xmath1 and @xmath2 to define the jets , in place of the usual cones defined in the rest frame of the detector . the second idea , examined in sect . [ sec : angles ] , is to use the angles between jets in the @xmath1 rest frame to measure @xmath3 directly . the first idea that we will investigate can be understood most clearly by considering the @xmath2 mass measurement . since the @xmath2 is color neutral , its decay products have nothing to do with other jets present in the final state . in the rest frame of the @xmath2 , the decays will look very similar to the @xmath13 decays studied at lep . they will almost always appear as back - to - back jets that can be defined by two fixed cones of half - angle @xmath17 oriented in opposite directions . reasonable cone sizes are on the order of @xmath18 . cones of this size are large enough to contain a good fraction of the energies of the jets , which are broadened by the collinear radiation singularity of qcd ; while they are small enough to leave a large fraction @xmath19 of the full @xmath20 of solid angle outside the two cones , so not too much `` background '' is included in them . although cone algorithms are not the traditional jet definitions in @xmath14 physics , they have been shown to work there , and even to provide superior resolution in jet angle and energy @xcite . this proposal to try fixed cones in the @xmath2 rest frame is quite different from the current experimental procedure for top analysis , where jets are defined by cones @xmath21 in the lego variables of the detector . those variables , @xmath22 pseudorapidity and @xmath23 azimuthal angle , are lorentz invariant only under boosts along the beam direction . meanwhile , the @xmath2 can have a rather large component of boost transverse to the beam . for example , in @xmath4 events with typical acceptance cuts , the middle @xmath24 of the @xmath2 transverse momentum distribution extends from @xmath25 to @xmath26 , so @xmath27 is not small compared to @xmath28 . lego - defined jet cones therefore correspond to a wide variety of sizes and shapes of `` cone '' in the @xmath2 rest frame , depending on the boost and hence depending on details that have nothing to do with the @xmath2 decay . similarly , the jet cone for the @xmath29-quark jet from @xmath30 is most reasonably defined by a fixed angle in the @xmath1 rest frame . since the @xmath29 carries color , the argument for a cylindrically symmetric cone in this frame is weaker than in the case of the @xmath2 decay . but the @xmath1 rest frame is nevertheless at least more logical than using the rest frame of the detector . it would similarly be reasonable to define the jet cone for the other @xmath29-jet , from the semileptonically decaying top , in the rest frame of that top . that should be done when this method is applied to real data . but we ignore it for the moment , in order to avoid bringing in the issue of how to estimate the longitudinal momentum of the neutrino , which is needed to find that frame . the idea of using fixed cones in the decay rest frames to define the jets is easy to implement as a follow - on to the traditional top quark analysis with `` kinematic reconstruction '' @xcite , _ i.e. _ , with explicit matching of the observed jets to decay partons . the procedure for each event is : 1 . identify the four - momenta of the jets corresponding to @xmath31 and @xmath32 , where @xmath33 , @xmath34 come from @xmath2 decay and @xmath35 , @xmath36 are @xmath29-quark jets , by the usual procedures of cdf or d . @xmath37-jet tagging is helpful for this , but not essential . there may of course be additional jets observed in the event . 2 . define @xmath38 and @xmath39 to be four - momenta directed in the forward and backward beam directions : @xmath40 and @xmath41 for beams in the @xmath42-direction . 3 . redefine the 4-momentum of @xmath33 as the sum of the 4-momenta observed in the calorimeter detector in all cells ( `` towers '' ) that lie within @xmath17 of the old @xmath33 in the rest frame of the old @xmath43 . however , do not include cells that lie closer in angle to @xmath34 , @xmath35 , @xmath36 , @xmath38 , or @xmath39 than to @xmath33 . + similarly redefine the 4-momentum of @xmath34 as the sum of the 4-momenta of all cells that lie within @xmath17 of the old @xmath34 in the rest frame of the old @xmath43 , and are not closer in angle to @xmath33 , @xmath35 , @xmath36 , @xmath38 , or @xmath39 . + similarly redefine the 4-momentum of @xmath35 as the sum of the 4-momenta in cells that lie within @xmath17 of the old @xmath35 in the rest frame of the old @xmath44 , and are not closer in angle to @xmath33 , @xmath34 , @xmath36 , @xmath38 , or @xmath39 . 4 . replace the old jet 4-momentum estimates by the new ones and repeat step 3 a few times . this iteration converges completely for @xmath45 of events . a maximum of 10 iterations is sufficient , since the momenta stop changing before that for @xmath46 of events . the major new work is in step 3 , which is very easy to implement in the following way . treat the energy in every cell of the detector that receives energy above its noise level as if it came from a zero - mass particle . there will typically be a few hundred such cells in each event . make a list of their four - momenta . the appropriate elements of this list to be added in the various parts of step 3 can be found without making explicit lorentz transformations , using the exact formula for the angle @xmath47 between @xmath48 and @xmath49 in the rest frame of @xmath50 @xmath51 \ , [ ( q \ ! \cdot \ ! r)^2 - q^2 \ , r^2 ] } } \ ; , \label{eq : costhe}\end{aligned}\ ] ] which holds for any four - vectors @xmath52 , @xmath53 , @xmath50 . to test the idea , herwig 5.7 @xcite was used to simulate the single - leptonic top channel at the tevatron energy @xmath54 . typical experimental cuts were approximated by @xmath55 , @xmath56 , @xmath57 . the four hadron jets were required , at the partonic level , to have @xmath58 and to be isolated in lego from the lepton by @xmath59 and from each other by @xmath60 . the detector was simulated by an array of @xmath61 cells in @xmath62 , covering the region @xmath63 . gaussian energy resolutions with @xmath64 for charged hadrons and @xmath65 for leptons and photons were assumed ( in @xmath66 units ) . a small amount of shower spreading was also included : the energy of each final hadron was spread over a disk of radius @xmath67 in the @xmath62 plane , before being deposited in the appropriate calorimeter cells . a cone - type jet finding algorithm described previously @xcite was used to search for jets in the simulated calorimeter . a small cone size @xmath68 was used , as is standard practice in top quark analysis ( cdf uses @xmath69 ) to reduce the loss of events from overlapping jets , since one is asking for 4 jets within a rather small area in lego _ e.g. _ , all four decay jets lie within some strip of width @xmath70 in @xmath71 in @xmath72 of the events . a minimum observed jet @xmath11 of @xmath73 was required . a feature of my jet - finder @xcite is that it finishes with an iteration in which each cell of the detector that lies within @xmath74 of at least one jet axis is assigned to the nearest jet axis . the jet momenta are then recomputed and this procedure is repeated . this feature tends to improve the mass resolution for objects decaying into jets . it is similar to the procedure being tried here , except that the nearest jets are now to be defined by angles in the appropriate rest frames . to obtain a sample that could be cleanly interpreted , the `` observed '' jets in the simulation were matched to their original quark partons , using a criterion based on the agreement in lego variables @xmath71 and @xmath75 , along with a contribution from the agreement in @xmath11 . a good match was obtained for @xmath76 of the events , upon which fig . 1 and fig . 2 are based . 1 shows the histogram of the mass observed for @xmath77 . the _ dotted curve _ is for jets defined by @xmath78 in the lego variables of the detector . note that the center of the peak is well below the actual @xmath79 assumed in the simulation , and that the peak is very asymmetrical , with a tail toward lower @xmath80 masses . this skewing of the mass distribution toward low mass is a direct result of gluon radiation falling outside the cone , which is characteristic of qcd , and which has been verified at lep . it demonstrates that making qcd corrections is a very important part of the top quark mass measurement , especially if a cone size as small as @xmath81 is used . the _ solid curve _ in fig . 1 shows the result of redefining the jets using a larger cone size @xmath82 . this included an iteration in which each calorimeter cell that lies within @xmath74 of at least one of the 3 jet axes from hadronic @xmath1 decay is reassigned to the jet with the nearest of those axes . this iteration is crucial to obtaining the improvement in mass resolution shown . note that the qualitative qcd feature of skewing toward lower mass is still visible , but much less pronounced . also , the center of the peak is at a higher mass , closer to the true partonic mass . the cone size @xmath83 is about optimal . if the width of the mass peak is defined by the range @xmath84 that contains the middle @xmath85 of the probability distribution , then @xmath86 goes from @xmath87 to @xmath88 to @xmath89 as @xmath74 is increased from @xmath81 to @xmath90 to @xmath91 . the _ dashed curve _ in fig . 1 shows the result of redefining the jets in the appropriate decay rest frames by the steps enumerated above . the curve shown is for @xmath18 , which is about optimal . one sees that the mass resolution obtained this way is quite good ( @xmath92 ) , but not as good as the @xmath83 curve . provide as good or slightly better resolution , as shown here . ] 2 shows a similar histogram of the mass observed for @xmath93 . one again sees qcd skewing of the peak toward lower masses , relative to the value @xmath94 assumed in the simulation . this mass shift includes a contribution of @xmath95 from losses in observed jet energy due to neutrinos . one again sees a dramatic reduction in skewing and improvement in the resolution , when the cone size is increased from @xmath96 ( _ dotted curve _ ) to @xmath97 ( _ solid curve _ ) . we emphasize that to obtain this improvement , it is essential to use the iterative procedure whereby each calorimeter cell is assigned to the nearest decay jet axis . the cone size @xmath83 is again about optimal . defining as before the width of the peak by the range @xmath84 that contains the middle @xmath85 of the probability distribution , @xmath86 goes from @xmath98 to @xmath99 to @xmath100 as @xmath74 is increased from @xmath81 to @xmath90 to @xmath91 . the suggestion to define the jets using a fixed cone angle in the rest frame of the @xmath1 for the @xmath29-jet , and in the rest frame of the @xmath2 for its decay jets , which was proposed in this section , leads to the _ dashed curve _ in fig . 2 . one sees that , as in fig . 1 , this plausible idea does not in fact improve the mass resolution ( @xmath101 ) , compared to @xmath97 cones although it is not much worse , either . so far , we have computed the observed masses of @xmath1 and @xmath2 directly from the 4-momenta of the jets into which they decay , as simply the invariant mass of the sum of those jet momenta . the estimated 4-momentum of each jet is obtained by adding the 4-momenta detected by the calorimeter cells inside its jet cone . the @xmath1 and @xmath2 are thus formally treated as decaying into a large number of zero - mass particles that represent the energies deposited in the detector . defined in this way , the individual jets have sizeable invariant masses often on the order of @xmath102 or higher . a possible alternative method to determine the top mass in each event would be to use only the _ directions _ of the jets in the top rest frame , treating the jets as zero - mass objects . to do this , let @xmath103 and @xmath104 be the angles between the @xmath29-jet and the two jets from @xmath2 decay in that frame . then @xmath105 is the angle between the @xmath2 decay jets . according to zero - mass kinematics , @xmath106 the angles needed in this formula can be obtained from the measured jet 4-momenta in any frame using eq . ( [ eq : costhe ] ) . at the partonic level , the use of zero - mass kinematics is promising since , _ e.g. _ , the energy of the @xmath29-quark is @xmath107 , so the influence of its mass @xmath108 is very small . an advantage of this technique is that it measures the top quark mass in units of @xmath28 , which is independently known @xcite . because a dimensionless ratio is measured , the result for @xmath12 is insensitive to any constant overall scale factor in the energy measurements . for comparison with the results of sect . [ sec : cones ] , the mass distribution obtained using this jet angle method are plotted as the _ dot - dash curve _ in fig . 2 , by scaling with the value of @xmath28 assumed in the simulation . _ the mass distribution is narrower than the best previous result by about @xmath109 . _ in addition to improved mass resolution , the curve is seen to be nearly symmetrical , and the center of the peak is very close to the mass value @xmath110 assumed in the simulation . the shift in the peak value due to the non - detection of neutrinos in the jets is also somewhat smaller when the jet angle method is used . when the simulation is repeated using @xmath111 , the center of the peak in @xmath112 determined by the jet angle method again comes out within @xmath113 of the input @xmath114 , so the good features found above are not specific to the assumed mass value . the nearly exact agreement with the input mass is of course partly accidental _ e.g. _ , the output mass would shift upward by @xmath115 if the calorimeter were able to detect neutrinos . the important result is that the output mass is close to the input one , and tracks it proportionally . the curve shown is for jets defined by @xmath97 , which is approximately optimal ; but a further advantage of the jet angle method is that it is found to be less sensitive to the choice of @xmath74 than is the traditional invariant mass method . defining the jets using fixed cones in the rest frames as described in sect . [ sec : cones ] , and then using the angle method to compute @xmath116 also works , but not quite as well as simply using the lego definition with @xmath117 . so far , we have explicitly used the original parton momenta , which are known in the simulation program , to determine the correct assignments of the observed jets . in real life , of course , that luxury is unavailable , so one faces a combinatoric problem in analyzing semileptonic @xmath4 events . there are 12 ways to assign 4 observed jets to the @xmath29-jet from semileptonic @xmath1 decay , the @xmath29-jet from hadronic @xmath1 decay , and the two jets from hadronic @xmath2 decay . there may be fewer combinations to try if one or both of the @xmath29-jets has been tagged ; but there may be many more , since frequently there are additional jets , _ e.g. _ , from initial state radiation . hence mass histograms like figs . 1 and 2 will be contaminated by a background from events in which the jets are misassigned . in order to test our methods directly , the herwig events and calorimeter simulation described above were used without making parton level cuts , and without using any of the partonic information to determine the correct jet assignments , in the following simplified event analysis : 1 . jets with observed @xmath118 are found using @xmath119 cones in @xmath62 space . if fewer than 4 jets are found , the event is rejected . 2 . if more than 4 jets are found , only those with the 4 largest values of @xmath11 are kept . we ignore @xmath29-tagging here , so all 12 ways of assigning the jets to the 4 expected jets are examined . ( by trying all possible subsets of 4 jets , it would be possible to reclaim an additional @xmath120 of good events ; but the combinatoric and non - top backgrounds to those events would be considerably larger , and the mass resolution not quite as good , so this will be of doubtful value even in the final data analysis . ) an iteration similar to that described in sect . [ sec : cones ] is carried out , so that each calorimeter cell within @xmath97 of any one of the 4 significant jet axes is associated with the nearest of those axes . the transverse momentum of the neutrino is estimated as usual as the negative of the total observed @xmath121 . its longitudinal momentum is estimated by requiring @xmath122 to equal the @xmath2 mass used in the simulation . the magnitude of the rapidity difference @xmath123 is then given by @xmath124 \ ; . \label{eq : massellnu}\end{aligned}\ ] ] the sign of @xmath125 is assumed to be the same as the sign of @xmath126 , which corresponds to choosing between the two possible solutions for @xmath127 the one giving the smaller energy for the leptonically decaying @xmath2 . if eq . ( [ eq : massellnu ] ) has no solution ( because @xmath128 ) , we take @xmath129 @xcite ; or reject the event if there is no solution with @xmath130 we reject some candidate jet assignments immediately on the basis of very loose cuts : we require @xmath131 for the leptonically decaying top and @xmath132 for the hadronically decaying @xmath2 . ( broad cuts on the hadronic top mass , or on the minimum non-@xmath2 dijet mass in the hadronic top might also be useful at this point @xcite . ) 6 . to select the correct jet assignment according to the best fit to the leptonic top and hadronic @xmath2 masses , we define a measure of fit by @xmath133 the parameters @xmath134 , @xmath135 , @xmath136 , @xmath137 in @xmath66 units were found using the previous simulation , by making the range @xmath138 correspond to the middle @xmath24 of the probability distribution for @xmath139 for the correct jet assignment , while @xmath140 corresponds similarly to the middle @xmath24 of the @xmath141 distribution for @xmath2 decay . the cubic ( @xmath142 ) and quadratic ( @xmath143 ) forms in eq . ( [ eq : d1d2 ] ) were chosen because these quantities have more symmetrical probability distributions than the qcd - skewed masses themselves . the factors in the denominators were chosen to make the constants @xmath144 and @xmath145 correspond directly to the widths of the distributions . + when it comes time to analyze real data , the parameters in eq . ( [ eq : d1d2 ] ) can be tuned using monte carlo events . also , the parameters in @xmath146 should be adjusted to be self - consistent with @xmath12 as measured by the hadronic @xmath1 decays . actually , the analysis could no doubt be improved by allowing for differences in the expected range for the observed @xmath139 , based on the configuration of @xmath147 and @xmath148 momenta . additional contributions to the measure of fit for each possible jet assignment could also be included , _ e.g. _ , based upon the transverse and longitudinal momenta of the @xmath1 and @xmath149 . the jet assignment that gives the lowest value of @xmath150 is assumed to be the correct one . a small number of events ( @xmath151 ) with a poor fit ( @xmath152 ) are rejected . a further @xmath153 of events are rejected when any two of the four final jet axes are separated by less than @xmath154 in lego . otherwise , the corresponding mass for @xmath155 is entered into a histogram to create fig . note that we adopt the idea advocated in ref . @xcite that the @xmath1 mass measurement should be based on the hadronically decaying top . the mass of the leptonically decaying one contributes only to choosing the correct set of jet assignments . this avoids a number of possibilities for systematic errors due to uncertainties in the neutrino momentum measurement . only the central value assumed for the leptonic decay mass needs to be tuned to match the final average @xmath12 found in the experiment . if the assumed value is incorrect , it will reduce the signal rate , but should not change the measured hadronic top mass . 3 shows the @xmath112 distributions for @xmath156 . the _ dotted curve _ shows the invariant mass given by the original @xmath119 cones . the _ solid curve _ shows the invariant mass given by @xmath97 cones , including the iteration that assigns calorimeter cells to the nearest decay jet axis . note that going from @xmath119 to @xmath97 yields a dramatic improvement in the sharpness of the peak , and considerably reduces the downward shift in mass relative to the @xmath157 input to the simulation . the benefits of going to @xmath97 cones is actually even greater than what is shown here , because the larger cones were used to infer the correct jet assignments for both curves . the _ dot - dashed curve _ shows the result of calculating @xmath112 by the jet angle method of sect . [ sec : angles ] . it shows a clear peak that is more symmetrical and is centered closer to the input value of @xmath12 , confirming the results of sect . [ sec : angles ] . the peak is not quite as high as the @xmath97 result , which may be due to the fact that the ordinary cone method is somewhat less affected by misassigned jets , since the value it gives for @xmath112 is independent of which of the three dijet pairs is identified as the hadronically decaying @xmath2 . the jet angle method and the ordinary invariant mass offer two largely independent measurements of @xmath112 in each event . we can combine the two measurements to obtain an average that has smaller errors than either measurement by itself . to allow for the fact that the two measurements contain different systematic mass shifts , it is seems appropriate to combine them by taking their geometric mean @xmath158 , although the distribution of their simple average is almost identical . the resulting mass histogram , shown as the _ dashed curve _ in fig . 3 , displays the best peak of all . this mass variable is therefore most promising for future analyses of @xmath4 data . in summary , we have seen that 1 . to obtain the best mass resolution for hadronic top decays , it is essential to use a sufficiently large cone size @xmath159 in the lego variables @xmath160 , @xmath161 of the detector , and to include an iteration whereby final particles within each of the four jet cones @xmath21 of the @xmath4 final state are self - consistently assigned to the nearest axis . quantitatively , going from @xmath68 to @xmath162 reduces the width of the top mass peak by a factor @xmath163 . going on from @xmath162 to @xmath83 reduces it by a further factor @xmath164 . 2 . a suggestion to define the jets by cones of fixed angle in the rest frames of the decaying @xmath1 and @xmath2 was argued to be plausible and shown to be simple to carry out , but did not in fact improve the mass resolution . instead , it gave results similar to @xmath117 . 3 . a suggestion to compute the top mass in each event on the basis of the jet angles in the measured @xmath1 rest frame , using the zero - mass kinematical formula eq . ( [ eq : angles ] ) , was shown to make a significant improvement in the measurement of @xmath12 . it reduces the width of the top mass peak slightly , and makes the peak more symmetrical and less sensitive to the choice of @xmath74 . this jet angle method is very easy to apply , since the necessary angles are easily computed from the jet four - momentum estimates using eq . ( [ eq : costhe ] ) . a further advantage of this method is that it measures @xmath12 as a ratio to the known @xmath28 , and is thus insensitive to the absolute calibration for jet energy measurements . an even better measure of @xmath12 in each event can be obtained using the geometric mean @xmath165 of the two measures of @xmath112 given by the ordinary lorentz - invariant mass of the three - jet system , and the jet angle method . these conclusions have been confirmed by the simulation of sect . [ sec : direct ] , which is a simplified version of the essence of actual @xmath4 data analysis . to clarify the point regarding cone size , let us examine the standard experimental practice of `` correcting '' the jet @xmath11 measured in a small cone , by a factor that is determined from monte carlo simulation , or experimentally , _ e.g. _ , by studying the @xmath11 balance of hard scattering in events with a direct photon and a jet . this correction factor could in general be a function of @xmath11 and @xmath71 . however , no such `` out - of - cone correction '' can improve the trijet mass resolution on a par with actually _ using _ larger cones , since when the latter is done , the revisions in @xmath11 are found to vary widely from case to case as a direct result of the qcd collinear radiation singularity . this is shown in fig . 4 , which displays the distribution of jet transverse energy increases in going from @xmath119 to @xmath97 in the simulation of sect . [ sec : direct ] , for jets from events in the signal peak with a typical @xmath166 in the smaller cones . quantitatively , the middle @xmath167 of the probability distribution extends from @xmath168 to @xmath169 for the @xmath29-jets , and from @xmath170 to @xmath171 for the jets from @xmath2 decay . the importance of ample cone size , with final state energies assigned to the nearest cone , has been clearly demonstrated . the jet angle method has been shown to have significant advantages . although the rest - frame cone method was found to be no better than the standard procedure , it should still be pursued on the grounds that its systematic errors might be different from the conventional analysis . most importantly , the mass variable @xmath172 , defined as the geometric mean of the two estimates of @xmath12 using @xmath97 cones , should be tried since it produced the best mass resolution of all methods examined here . the improved resolution should help to reduce the statistical uncertainty in the location of the mass peak , which will arise from the rather small number of events ( @xmath173 ) expected in each experiment by the end of the current tevatron running period . this technique should therefore improve the accuracy of the top quark mass measurement . the next steps , which can only be carried out by the experimenters of cdf and d , must be to try these procedures using full detector simulations , and then real data . the improvements in mass resolution could also help with the the difficult problem of observing doubly hadronic top events @xmath174 @xcite . i thank s. kuhlmann , j. huston , j. linnemann , t. ferbel , and s. snyder for information about the experiments . this work was supported in part by u.s . national science foundation grant number phy-9507683 .

two possible ways to improve the mass resolution for observing hadronic top quark decay @xmath0 are studied : ( 1 ) using fixed cones in the rest frames of the @xmath1 and @xmath2 to define the decay jets , instead of the traditional cones in the rest frame of the detector ; and ( 2 ) using the jet angles in the top rest frame to measure @xmath3 . by monte carlo simulation , the second method is found to give a useful improvement in the mass resolution . it can be combined with the usual invariant mass method to get an even better mass measurement . the improved resolution can be used to make a more accurate determination of the top quark mass , and to improve the discrimination between @xmath4 events and background for studies of the production mechanism .

pulsating da ( h - rich atmospheres ) white dwarfs , also known as zz ceti or dav variable stars , comprise the most numerous and studied class of degenerate pulsators . their instability strip , probably a pure one , is centered at an effective temperature of around @xmath3 k ( winget & kepler 2008 ; fontaine & brassard 2008 ; althaus et al . 2010a ) . the luminosity variations ( of up to @xmath4 mag ) are caused by nonradial @xmath1-mode pulsations of low degree ( @xmath5 ) and periods between @xmath6 and @xmath7 s , excited through a combination of the @xmath8 mechanism acting in the hydrogen partial ionization zone ( dolez & vauclair 1981 ; winget et al . 1982 ) and the `` convective driving '' mechanism once a sufficiently thick outer convective zone has been developed ( brickhill 1991 ; goldreich & wu 1999 ) . zz ceti asteroseismology the comparison between the observed periods and the periods computed for a suite of representative stellar models has the potential of disentangle the internal structure of da white dwarfs , allowing to place constraints on the stellar mass , the thickness of the outer envelopes , the core chemical composition , weak magnetic fields and slow rotation rates from the observed period patterns ( winget & kepler 2008 ; fontaine & brassard 2008 ; althaus et al . 2010a ) . a total of 148 zz ceti stars are currently known ( castanheira et al . 2010 ) . to this list we must add the recently discovered dav star wd [email protected] ( _ kepler _ i d 4552982 hereinafter wd j1916@xmath03938 ) , a zz ceti located in the _ kepler _ mission field and identified through ground - based times series photometry ( hermes et al . 2011 ; hereinafter hea11 ) . this star ( @xmath9 k ; @xmath10 ) exhibits low - amplitude luminosity variations with periods between @xmath11 and @xmath12 s ( see column 2 of table [ table1 ] ) . it is expected that an extended monitoring of this object with _ kepler _ could bring the best lightcurve of a pulsating white dwarf ever recorded ( hea11 ) , even with a better quality than the lightcurves of the brighter stars observed through uninterrupted ground - based campaigns of the whole earth telescope ( wet ; nather et al . 1990 ) . in order for white dwarf asteroseismology to provide realistic constraints on the internal structure of white dwarfs , it is crucial the use of stellar models characterised by consistent and detailed chemical profiles to accurately assess the adiabatic pulsation periods . this requirement has been tackled successfully in the case of the hot dovs or gw vir stars ( see crsico et al . 2007a , 2007b , 2008 , 2009 ) . very recently , romero et al . ( 2011 ) have performed the first asteroseismological analysis of a set of 44 zz ceti stars based on the new generation of fully evolutionary da white dwarf models presented in althaus et al . ( 2010b ) . these models are characterised by realistic chemical profiles throughout the star and cover a wide range of stellar masses , thicknesses of the h envelope and effective temperatures . the analysis of a large number of zz ceti stars has the potential to characterise the global properties of the class , in particular the thicknesses of the h envelope and the stellar masses . the study of romero et al . ( 2011 ) revealed that da white dwarfs in the solar neighbourhood could harbor a broad range of h - layer thickness . on the other hand , this analysis was successful in finding , for the first time , an univocal asteroseismological model for the archetypal zz ceti star g117@xmath13b15a , one of the targets of that analysis . motivated by the exciting discovery of the first pulsating da white dwarf in the field of the _ kepler _ mission , and encouraged by the availability of the most complete set of detailed evolutionary and pulsation models for dav stars up to date ( romero et al 2011 ) , we present in this paper a first asteroseismological analysis of wd j1916@xmath03938 . we are well aware that the asteroseismological potential of this pulsating star will probably be fully exploited with further observations by _ kepler _ , which will allow to have available a period spectrum substantially richer than that we have at hand at present just from ground - based photometry . however , we think that a first asteroseismological analysis of this star based on the seven periods a representative number of periods for zz ceti standards currently available is worth being attempted . by considering the mean period spacing exhibited by the star , we found that wd j1916@xmath03938 should be substantially massive , in excellent agreement with the spectroscopic inference . we also found that period - to - period fits favour an asteroseismological model characterised by a high stellar mass , a low effective temperature , and a rather thin h envelope . the paper is organized as follows . in the next section we briefly describe the computer codes and stellar models employed . the short section [ smd ] is devoted to the derivation of the spectroscopic mass of wd j1916@xmath03938 . we make inferences about the stellar mass from the period spacing data in section [ period - spacing ] , taking into account its dependence not only with the effective temperature and mass , but instead also with the thickness of the h envelope . in section [ period - fit ] we extract information of the star through asteroseismological period fits . section [ rotation ] is devoted to infer a rotation period of the star . finally , we enumerate our results in the conclusions ( section [ conclu ] ) . .the observed frequencies , periods , and amplitudes of wd j1916@xmath03938 ( columns 1 to 3 ) , according to hea11 , and the periods of a phenomenological model ( see section [ period - spacing ] ) with a fixed ( @xmath14 ) period spacing of @xmath15 s ( columns 4 to 6 ) . [ cols="^,^,^,^,^,^ " , ] [ table2 ] next , we tried to find a best - fit model by discarding the period at 834.1 s ( so , @xmath16 ) , to be consistent with the assumption made to derive a mean period spacing of the star in sect . [ period - spacing ] . we found the same best - fit model than before ( see table [ table2 ] , model 2 ) , but the quality of the fit is , as expected , substantially improved ( @xmath17 s ) . in a further step , we relaxed the constraint of @xmath14 for all the observed periods of wd j1916@xmath03938 , and we redone our period fit by allowing the values @xmath14 or @xmath18 . we obtain exactly the same best - fit model as before . in table [ table2 ] we show the theoretical periods and the @xmath19- and @xmath20-identifications ( model 3 ) . for this period fit , we obtain @xmath21 s. this is the best agreement with the observed periods we obtain , although in this case the @xmath20-identification is not the same as that assumed in sect . [ period - spacing ] to derive the period spacing value of the star . finally , we have carried out an additional period - fit , this time supposing that the periods at 823.9 s and 834.1 are the @xmath22 and @xmath23 components of a rotational triplet , in which the central component ( @xmath24 ) is missing ( see next section ) . under this assumption , we estimate the period of the @xmath25 component as the average of the @xmath22 and @xmath26 components . thus , we consider a period of 829 s in our period - fit procedure , instead the two periods at 823.9 s and 834.1 . here , we adopt @xmath14 for the set of @xmath16 periods at the outset . in this case , we obtain virtually the same asteroseismological model than before , with @xmath27 , @xmath28 k , @xmath29 , and @xmath30 s. the theoretical periods and @xmath31 values are shown in table [ table2 ] ( model 4 ) . we conclude that , even by adopting several different assumptions for the assignation of the @xmath20 and @xmath19 values to the observed periods , we found the same asteroseismological model which is characterised by an effective temperature of @xmath32 k , a surface gravity of @xmath33 , a stellar mass of @xmath34 , a mass of h of @xmath35 , and a he content of @xmath36 . the location of the best - fit model in the @xmath37 plane is depicted in fig . [ gteff ] . here , we suppose that the periods at 823.9 s and 834.1 are the @xmath22 and @xmath26 components , respectively , of a rotational triplet in which the central component ( @xmath24 ) is absent from the pulsation spectrum of wd j1916@xmath03938 . if this assumption is valid , we can infer a frequency spacing of @xmath38 @xmath39hz between the extreme components of the triplet . by assuming rigid and slow rotation , we have @xmath40 , being @xmath41 @xmath39hz and @xmath42 in the asymptotic limit for @xmath14 . then , we can infer a mean rotation period of @xmath43 hs for wd j1916@xmath03938 . this is in line with the values derived for other zz ceti stars from asteroseismology ( see , for instance , table 4 of fontaine & brassard 2008 ) . the fact that a da white dwarf like this , characterised by an effective temperature of @xmath44 k , and pulsating in high order @xmath1-modes with long periods , which usually have associated rotational kernels that are mostly confined to the outer layers of the star , prevent us to attempt a more profound analysis of the internal rotation of wd j1916@xmath03938 , like the one performed very recently by crsico et al . ( 2011 ) for the gw vir star pg 0122 + 200 . tempted by the exciting discovery of the first pulsating da white dwarf in the _ kepler _ mission field by hea11 , and encouraged by the availability of the most complete set of detailed evolutionary and pulsation models for dav stars up to date ( romero et al . 2011 ) , in this paper we have presented a first asteroseismological analysis of wd j1916@xmath03938 . this , even with the certainty that the full asteroseismological potential of this dav star will be exploited with further observations by _ kepler _ in the next years . we summarize our main findings below : * we derive a spectroscopic mass of @xmath45 for wd j1916@xmath03938 on the basis of the new da white dwarf evolutionary tracks of romero et al . this value is smaller than the estimate of hea11 , of @xmath46 , based on the older models of wood ( 1990 ) . in our analysis , we have taken into account the dependence of the tracks in the @xmath47 diagram with the thickness of the h envelope ( see fig . [ gteff ] ) . * by assuming that all the periods exhibited by wd j1916@xmath03938 are normal modes of the star , and supposing that all the periods except one ( at 834.1 s ) are associated to @xmath14 modes , we derive an average period spacing of @xmath48 s. on the basis of the asymptotic theory of stellar pulsations , we estimate a stellar mass of @xmath49 , corresponding to a thick h envelope . if we take into account the strong dependence of the asymptotic period spacing with the thickness of the h envelope , we found an evident degeneracy of @xmath50 with @xmath51 and @xmath52 . in virtue of this , we can conclude that @xmath53 , but we can not say anything about the thickness of the h envelope . finally , if we ignore the constraint imposed by the effective temperature as derived by spectroscopy , we found that wd j1916@xmath03938 should have a stellar mass @xmath54 in order for the star to be within the zz ceti instability strip . * by using the individual pulsation periods exhibited by wd j1916@xmath03938 , and taking into account several assumptions for their @xmath20- and @xmath19-identification , we found an asteroseismological model with @xmath55 k , @xmath56 , @xmath57 , @xmath58 , and @xmath59 . the stellar mass of the asteroseismological model is in good agreement with the spectroscopic estimate ( @xmath60 ) and with the derivation from the period spacing ( @xmath61 ) . this is quite encouraging , because the three estimates have been derived from very different approaches and assumptions . on the other hand , the effective temperature of the asteroseismological model is somewhat higher than the spectroscopic one ( @xmath62 k ) . * in passing , we have estimated a rotation period of 18.77 hs for wd j1916@xmath03938 . the reliability of this result rests on the assumption that the modes with periods at 823.9 s and 834.1 are the @xmath63 and @xmath23 components , respectively , of a @xmath14 frequency triplet due to rotation . in closing , we think that the conclusion that wd j1916@xmath03938 should be a rather massive zz ceti star , with @xmath64 , appears to be robust . the possibility that this rather massive pulsating white dwarf can be further monitored with _ kepler _ with a high degree of detail turns this star into a promising and unique target to study the physics of crystallization and carbon / oxygen phase diagrams at high densities . in this regard , we note that , for a stellar mass of @xmath65 , crystallization of matter should occur at the centre of a white dwarf star at @xmath66 k according to the phase diagram of segretain & chabrier ( 1993 ) , and at @xmath67 k if the new phase diagram recently proposed by horowitz et al . ( 2010 ) is considered ( althaus et al . 2011 ) . indeed , based on direct molecular dynamics simulations for dense carbon - oxygen mixtures , horowitz et al . ( 2010 ) find substantially lower crystallization temperatures and that the shape of the carbon - oxygen phase diagram is of the azeotropic form and not of the spindle type as previously believed . our analysis from the period spacing of wd j1916@xmath03938 suggests that this star could well have a stellar mass much higher than @xmath68 and necessarily a rather thin h envelope . if this were the case , this star could have a substantial fraction of its interior crystallized depending on the adopted carbon - oxygen phase diagram , an issue that could be explored with future asteroseismological analysis that exploit the new high - quality photometric data coming from the ongoing monitoring by the _ kepler _ mission . part of this work was supported by agencia through the programa de modernizacin tecnolgica bid 1728/oc - ar , and by the pip 112 - 200801 - 00940 grant from conicet . this research has made use of nasa s astrophysics data system . althaus , l. g. , garca - berro , e. , isern , j. , crsico , a. h. , & miller bertolami , m. m. 2011 , a&a , in press ( arxiv:1110.5665 ) althaus , l. g. , crsico , a. h. , isern , j. , & garca - berro , e. 2010a , a&ar , 18 , 471 althaus , l. g. , crsico , a. h. , bischoff - kim , a. , romero , a. d. , renedo , i. , garca - berro , e. , & miller bertolami , m. m. 2010b , apj , 717 , 897 brickhill , a. j. 1991 , mnras , 251 , 673 castanheira , b. g. , kepler , s. o. , kleinman , s. j. , nitta , a. , & fraga , l. 2010 , mnras , 405 , 2561 crsico , a. h. , althaus , l. g. , kawaler , s. d. , miller bertolami , m. m. , garca - berro , e. , kepler , s. o. 2011 , mnras , 418 , 2519 crsico , a. h. , althaus , l. g. , miller bertolami , m. m. , & garca - berro , e. 2009 , a&a , 499 , 257 crsico , a. h. , althaus , l. g. , kepler , s. o. , costa , j. e. s. , & miller bertolami , m. m. 2008 , a&a , 478 , 869 crsico , a. h. , althaus , l. g. , miller bertolami , m. m. , & werner , k. 2007a , a&a , 461 , 1095 crsico , a. h. , miller bertolami , m. m. , althaus , l. g. , vauclair , g. , & werner , k. 2007b , a&a , 475 , 619 crsico , a. h. , & althaus , l. g. 2006 , a&a , 454 , 863 dolez , n. , & vauclair , g. 1981 , a&a , 102 , 375 fontaine , g. , & brassard , p. 2008 , pasp , 120 , 1043 goldreich , p. , wu , y. 1999 , apj , 511 , 904 hermes , j. j. , mullally , f. , stensen , r. h. , et al . 2011 , apj , 741 , l16 ( hea11 ) horowitz , c. j. , schneider , a. s. , & berry , d. k. 2010 , physical review letters , 104 , 231101 nather , r. e. , winget , d. e. , clemens , j. c. , hansen , c. j. , & hine , b. p. 1990 , apj , 361 , 309 romero , a. d. , crsico , a. h. , althaus , l. g. , kepler , s. o. , castanheira , b. g. , miller bertolami , m. m. 2011 , mnras , in press ( arxiv:1109.6682 ) salaris , m. , serenelli , a. , weiss , a. , & miller bertolami , m. 2009 , apj , 692 , 1013 segretain , l. , & chabrier , g. 1993 , a&a , 271 , l13 tassoul , m. , fontaine , g. , winget , d. e. 1990 , apjs , 72 , 335 winget , d. e. , & kepler , s. o. 2008 , araa , 46 , 157 winget , d. e. , van horn , h. m. , tassoul , m. , fontaine , g. , hansen , c. j. , carroll , b. w. 1982 , apj , 252 , l65 wood , m. a. 1990 , ph.d . thesis , univ . texas , austin

dav stars , also called zz ceti variables , are pulsating white dwarfs with atmospheres rich in h. asteroseismology of dav stars can provide valuable clues about the origin , structure and evolution of da white dwarfs . recently , a new dav star , wd [email protected] , has been discovered in the field of the _ kepler _ spacecraft . it is expected that further monitoring of this star in the next years will enable astronomers to obtain the best lightcurve of a pulsating da white dwarf ever recorded , and thus to know with unprecedented precision the hidden details of the internal structure of this star . in this paper , we perform a first asteroseismological analysis of wd [email protected] on the basis of fully evolutionary da white - dwarf models . specifically , we employ a complete set of evolutionary da white - dwarf structures covering a wide range of effective temperatures , stellar masses , and h envelope thicknesses . these models have been obtained on the basis of a complete treatment of the evolutionary history of progenitors stars . we compute @xmath1-mode adiabatic pulsation periods for this set of models and compare them with the pulsation periods exhibited by wd [email protected] . based on a tentative estimation of the mean period spacing of the star , we find that the stellar mass should be substantially large ( @xmath2 ) , in agreement with the spectroscopically derived stellar mass . also , from period - to - period fits we find an asteroseismological model characterised by a low effective temperature , rather high stellar mass and a thin h envelope . the possibility that this rather massive pulsating white dwarf can be further monitored with _ kepler _ with a high degree of detail turns the star wd [email protected] into a promising and unique object to study the physics of crystallization and carbon / oxygen phase diagrams at high densities . [ firstpage ] asteroseismology stars : evolution stars : interiors stars : oscillations stars : individual : zz ceti stars stars : white dwarfs

nestings and crossings are equidistributed in many combinatorial objects , such as matchings , set partitions , permutations , and embedded labelled graphs @xcite . more surprising is the symmetric joint distribution of the crossing and nesting numbers : a set of @xmath0 arcs forms a @xmath0-crossing ( nesting ) if each of the @xmath1 pairs of arcs cross ( nest ) . the crossing number of an object is the largest @xmath0 for which there is a @xmath0-crossing , and the nesting number is defined similarly . et al . _ @xcite proved the symmetric joint distribution of the nesting and crossing numbers for set partitions and matchings . although they describe explicit involutions , they do not use simple local operations on the partitions . de mier @xcite , interpreted the work of krattenthaler @xcite to show that @xmath0-crossings and @xmath0-nestings satisfy a similar distribution in embedded labelled graphs . a hole in this family of results is the extension of @xmath0-crossings and @xmath0-nestings in permutations . this note fills this gap . we also give exact enumerative formulas for permutations of size @xmath2 with crossing numbers @xmath3 ( non - crossing ) and @xmath4 . the _ arc annotated sequence _ associated to the permutation @xmath5 is the directed graph on the vertex set @xmath6 with arc set @xmath7 , drawn in a particular way . it is also known as the standard representation , or simply , the arc diagram . it is embedded in the plane by drawing an increasing linear sequence of the vertices , with edges @xmath8 satisfying @xmath9 drawn above the vertices ( the upper arcs ) , and the remaining lower arcs satisfying @xmath10 drawn below . we refer to this graph as @xmath11 ; the subgraph induced by the upper arcs and @xmath12 is @xmath13 ; and the subgraph induced by the lower arcs and @xmath12 is @xmath14 . additionally , we reverse the orientation of the arcs in @xmath14 , and view it as a classic arc diagram above the horizon . because of these rules , the direction of the arcs is determined , and hence we simplify our drawings by not showing arrows on the arcs . these two subgraphs are arc diagrams in their own right : for example @xmath14 represents a set partition , and @xmath15 is a set partition with some additional loops . for convenience purposes , we draw @xmath16 flipped above the axis , so that it resembles other arc diagrams . . in this example , @xmath17 , @xmath18 , and the degree sequence is given by @xmath19,title="fig:",width=226 ] + . in this example , @xmath17 , @xmath18 , and the degree sequence is given by @xmath19,title="fig:",width=226 ] + . in this example , @xmath17 , @xmath18 , and the degree sequence is given by @xmath19,title="fig:",width=226 ] + crossings and nestings are defined for permutations by considering the upper and lower arcs separately . a crossing is a pair of arcs @xmath20 satisfying either @xmath21 ( an upper crossing ) or @xmath22 ( a lower crossing ) . a nesting is a pair of arcs @xmath23 @xmath24 satisfying @xmath25 ( an upper nesting ) or @xmath26 ( a lower nesting ) . there is a slight dissymmetry to the treatment of upper and lower arcs in this definition which we shall see is inconsequential . however , the reader should recall that what is considered a crossing ( nesting ) in the upper diagram is elsewhere called an _ enhanced _ crossing ( resp . enhanced nesting ) . crossings and nestings were defined in this way by corteel @xcite because they represent better known permutation statistics . corteel s theorem 1 states that the number of top arcs in this representation of a permutation is equal to the number of weak exceedences , the number of arcs on the bottom is the number of descents , each crossing is equivalent to an occurrence of the pattern @xmath27 , and each nesting is an occurrence of the pattern @xmath28 . corteel s proposition 4 states nestings and crossings occur in equal number across all permutations of length @xmath2 . to generalize her work we define @xmath0-crossings and @xmath0-nestings in the same spirit as set partitions and matchings . a @xmath0-crossing in a permutation ard diagram @xmath11 is a set of @xmath0 arcs @xmath29 that satisfy either the relation @xmath30 ( upper @xmath0-crossing ) or @xmath31 ( lower @xmath0-crossing ) . similarly , a @xmath0-nesting is a set of @xmath0 arcs @xmath32 that satisfy either the relation @xmath33 ( upper @xmath0-nesting ) or @xmath34 ( lower @xmath0-nesting ) . the _ crossing number _ of a permutation @xmath35 , denoted by @xmath36 , is the size of the largest @xmath0 such that @xmath11 contains a @xmath0-crossing . in this case we also say @xmath35 is @xmath37-noncrossing . likewise , the nesting number of a permutation @xmath38 is the size of the largest nesting in @xmath11 , and define @xmath37-noncrossing similarly . occasionally we consider the top and lower diagrams in their own right as _ graphs _ , and then we use the definition of demier @xcite , and hence distinguish separately the enhanced crossing number of the _ graph _ @xmath15 denoted @xmath39 from the permutation crossing number , and likewise for the enhanced nesting number @xmath40 . the number of permutations of @xmath41 with crossing number equal to @xmath0 is @xmath42 , and we likewise define @xmath43 for nestings . the _ degree sequence _ @xmath44 of a graph @xmath45 is the sequence of indegree and outdegrees of the vertices , when considered as a directed graph : @xmath46 some sources call these left - right degree sequences since in other arc diagrams the incoming arcs always come in on the left , and the outgoing arcs go out to the right . as a graph , the degree sequence of a permutation is trivial : @xmath47 , since a permutation is a map in which every point has a unique image , and a unique pre - image . to define a more useful entity , we define the degree sequence of a permutation to be the degree sequence of only the upper arc diagram : @xmath48 . the degree sequence defined by the lower arc diagram can be computed coordinate - wise directly from the upper by simple transformations given in table [ tab : vtypes ] , and we denote this sequence @xmath49 . ( the sums of the vertex degrees is not ( 1,1 ) because the lower arcs have their orientation reversed , and hence the indegree , and the out degree have switched ) an example is in figure 1 . the vertices with degree @xmath50 are called `` openers '' and those with degree @xmath51 are `` closers '' . [ tab : vtypes ] & @xmath52 + opener & + + & ( 1,0 ) & ( 1,0 ) + closer & + + & ( 0,1 ) & ( 0,1 ) + loop & + + & ( 1,1 ) & ( 0,0 ) + upper transient & + + & ( 1,1 ) & ( 0,0 ) + lower transient & + + & ( 0,0 ) & ( 1,1 ) + the main theorem can now be stated . [ thm : main ] let @xmath53 be the number of permutations of @xmath2 with crossing number @xmath54 , nesting number @xmath55 , and left - right degree sequence specified by @xmath56 . then @xmath57 there is an explicit involution behind this enumerative result . the number of permutations of @xmath41 with crossing number equal to @xmath0 is directly computable for small values of @xmath2 and @xmath0 . ._@xmath42 : the number of permutations of @xmath41 with crossing number @xmath0 . a crossing number of 1 is equivalent to non - crossing . _ [ cols="^,>,>,>,>,>,>,>,>,>",options="header " , ] we immediately notice the first column of table [ tab : enum ] , the non - crossing permutations , are counted by catalan numbers : @xmath58 . this has a simple explanation : non - crossing _ partitions _ have long been known to be counted by catalan numbers and there is a simple bijection between non - crossing permutations and non - crossing partitions . essentially , to go from a non - crossing permutation to a non - crossing partition , flip the arc diagram upside down , convert the loops to fixed points , and then remove the lower arcs . this defines a unique set partition , and is easy to reverse . this bijection is easy to formalize , but it is not the main topic of this note . to get a sense of how theorem 1 is proved , and to obtain some new enumerative results , we consider the set of maximum nestings and crossings . a _ maximum nesting _ is the largest possible : a @xmath4-nesting is maximum in a permutation on @xmath2 elements . we can compute @xmath59 explicitly . the number of permutations with a maximum nesting satisfies the following formula : @xmath60 we divide the result into a few cases , but each one is resolved the same way : for each permutation @xmath5 with a maximum nesting , the @xmath61-nesting comes from either @xmath15 or @xmath14 , and in most cases defines that subgraph . once one side is fixed , and there is a given degree sequence , it is straightforward to compute the number of ways to place the remaining arcs . some cases are over counted , and tallying these gives the final result . to achieve an @xmath63-nesting , it must be an enhanced nesting in the upper arc diagram , and it uses all vertices , including a loop : @xmath64 . it remains to define @xmath65 for @xmath66 . the lower degree sequence is fixed , and so @xmath67 for each @xmath54 , but other than that there is no restriction . thus , there are @xmath68 possibilities . the even case is slightly more complicated , owing to the fact that three different ways to achieve an @xmath70-nesting : an @xmath70-nesting in @xmath15 : : these permutations satisfy @xmath71 . as before , there are @xmath68 ways to define @xmath72 . an @xmath70-nesting in @xmath14 : : these permutations satisfy @xmath73 . again , there are @xmath68 possibilities to define @xmath74 . only the involution @xmath75 $ ] is in the intersection of these sets . an enhanced @xmath70-nesting in @xmath13 : : if the @xmath70-nesting uses only @xmath76 vertices , there is one left over . it must either be a lower transient vertex , or a loop since there is nothing left to connect to it . we count these by considering the different ways to construct it from a smaller permutation diagram . suppose we have a permutation with an @xmath70-nesting on @xmath76 vertices . by the first part , we know there are @xmath77 of these . we place it on @xmath78 points , by first selecting our special vertex @xmath54 , and placing the permutation on the rest . there are @xmath78 ways to pick this special vertex . finally , we create the new permutation @xmath35 by connecting the new vertex to the rest of the structure . we choose a point @xmath55 to be the value @xmath65 . we can choose @xmath54 and thus @xmath54 is a loop . otherwise , @xmath55 must be before the loop in @xmath79 . we then set @xmath80 to be @xmath81 . there are @xmath70 choices for @xmath65 . over counting : : we have counted twice the family of diagrams with two loops in the center . there are @xmath77 of these . putting all of the pieces together , and simplifying the expression we get the formula : @xmath82 this proof suggests a direct involution on the permutation which switches a maximum nesting for a maximum crossing , since the degree sequence of a nesting and a crossing have the same shape . thus the formula for maximum crossings is the same . this involution might send a @xmath0-crossing to a @xmath37-nesting , and so it can not be used for general @xmath0 . from the formula , we see that a very smal proportion of permutations have maximum crossings , ( @xmath83 ) or are non - crossing @xmath84 . what can be said of the nature of the distribution , or the average crossing number ? what is the nature of the generating function @xmath85 where @xmath86 marks the crossing number , or even simply the generating function for @xmath0-noncrossing permutations ? bousquet - mlou and xin @xcite consider this question for permutations : 2-noncrossing partitions are counted by catalan numbers , ( as we mentioned before ) , and thus the generating function is algebraic ; the counting sequence for 3-noncrossing partitions is p - recursive , and so the generating function is d - finite , and they conjecture that the generating function for @xmath0-noncrossing partitions , @xmath87 are likely not d - finite . how can these results be adapted to permutations , given the similar structure ? the proof of the main theorem first decomposes a permutation into its upper and lower arc diagrams and then applies the results for set partitions separately to each part . we can apply a result of demier ( * ? ? ? * theorem 3.3 ) almost directly to prove our result by describing an involution @xmath88 which preserves the left - right degree sequence and swaps nesting and crossing number . that is , @xmath89 , and @xmath90 , @xmath91 . this involution happens on the ferrer s diagram , and is described in detail by krattenthaler @xcite . her theorem 3.3 states that any left - right degree sequence @xmath56 , the number of @xmath0- noncrossing graphs on @xmath56 equals the number of @xmath0-nonnesting graphs on @xmath56 . actually , she proves more : the number of graphs with crossing number @xmath54 and nesting number @xmath55 have a symmetric distribution across all graphs with a given fixed degree sequence . in order to apply her results , the first step is re - write @xmath15 so that we only consider proper crossings and nestings instead of enhanced crossing and nestings . this is a common trick , known as inflation . essentially , we create the graph @xmath45 from @xmath15 by adding some supplementary vertices to eliminate loops and transitory vertices : + now each nesting and crossing is proper , and by ( * ? ? ? * lemma 3.4 ) @xmath92 and @xmath93 . in the course of her proof , she relies on an involution of krattenthaler @xcite on fillings of ferrer s diagrams . let @xmath94 be the map on embedded labelled graphs described implicitly in her proof . because @xmath94 is a left - right degree preserving map , we can identify the supplementary vertices in @xmath95 to get a graph with the correct kind of vertices . call this new graph @xmath96 . we now extend the definition of @xmath94 to @xmath15 by @xmath97 consider the pair of graphs @xmath98 . proving our main theorem now reduces to showing that there is a unique @xmath99 such that @xmath100 , which we do next . for every vertex in @xmath101 the indegree and the outdegree are equal to one this is because the left - right degree sequence of both the top and the bottom are preserved in the map , and hence the vector sum of their degree sequence is unchanged , i.e. @xmath47 , and has all the correct partial sum properties . the map is a bijection and so @xmath102 is unique . this map swaps the upper nesting and the upper crossing number , and also the lower nesting and the lower crossing number . thus @xmath103 . thus , the crossing and the nesting number are switched under the map @xmath94 . set partitions can be viewed as an instance of fillings of ferrer s diagrams , and krattenthaler s involution reduces to the involutions that chen _ et al . _ describe to prove their ( * ? ? ? * theorem 1 ) . figure [ fig : psisigma ] illustrates our involution on an example . remark that the degree sequence is fixed . and its image in the involution @xmath104 . note that @xmath105 , @xmath106.__,title="fig:",width=226 ] + and its image in the involution @xmath104 . note that @xmath105 , @xmath106.__,title="fig:",width=226 ] + involutions are in bijection with partial matchings , and have thus been previously considered . what of other subclasses of permutations ? the map presented here does not fix involutions , because loops are mapped to upper transient vertices , but it does fix any class that is closed under degree sequence , for example , permutations with no lower transitory vertices , or permutations with no upper transitory vertices nor loops . these conditions have interpretations in terms of other permutation statistics , if we consider the initial motivations of corteel . the main open question , aside from the enumerative , and probabilistic questions we have already raised , is to find a direct permutation description of our involution , i.e. a description avoiding the passage through tableaux or fillings of ferrer s diagrams . is this involution already part of the vast canon of permutation automorphisms ? which subclasses of permutations preserve the symmetric distribution ? from our example , we remark that cycle type is not neccesarily conserved ( since loops are always mapped to upper transitory vertices ) , but non - intersecting intervals are preserved . involution permutations are in bijection with partial matchings , and so this subclass has this property . ultimately we have considered a type of graph with two edge colours and strict degree restrictions . can this be generalized to a larger class of graphs with fewer degree restrictions ? what of a generalization of graphs with multiple edge colours ? the authors thank cedric chauve , lily yen , sylvie corteel and eric fusy for useful discussions . sb and mm are both funded by nserc ( canada ) , and mm did part of this work as part of a cnrs poste rouge ( france ) .

we introduce @xmath0-crossings and @xmath0-nestings of permutations . we show that the crossing number and the nesting number of permutations have a symmetric joint distribution . as a corollary , the number of @xmath0-noncrossing permutations is equal to the number of @xmath0-nonnesting permutations . we also provide some enumerative results for @xmath0-noncrossing permutations for some values of @xmath0 .

the systematic study of gaseous nebulae started in the decade of 1930 s with a long series of papers by menzel and collaborators ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * among others ) . in these papers , the basis was laid for the quantitative analysis of the hydrogen spectrum and the determination of physical conditions in gaseous nebulae . the hydrogen spectrum was recognized to form in the radiative cascade following recombination of h@xmath1 atoms , and quantitative predictions for the line intensities based on this scenario were made . two limiting cases were considered , which depend on the optical thickness of the nebula in the lyman lines . in case a , the nebula is assumed optically thin in all recombination lines , including those of the lyman series ; in case b , the nebula is assumed thick to lyman radiation , which is consequently scattered until conversion into higher series photons plus ly@xmath2 or 2-@xmath3 continuum emission takes place . case b was soon shown to be the most appropriate description for most nebulae , both on observational and theoretical grounds ( e.g. , * ? ? ? * ; * ? ? ? a third scenario , named case c , was also initially introduced to represent the case of optically thin nebulae with stellar continuum pumping as an additional mechanism for the excitation of ly@xmath2 . contrarily to cases a and b , however , case c has barely had any follow - up in the literature , with very sporadic exceptions ( e.g. , * ? ? ? * ; * ? ? ? as pointed out by @xcite , a reason for this scarce popularity is the _ a priori _ assumption of the relative lack of photons at the relevant frequencies due to the absorption features of stellar atmospheres , which would make the mechanism inefficient and imply that case c might be a merely academical scenario with no actual physical applications . ( the initial calculation by baker & menzel did not suffer from such limitation , since the stellar source was approximated by a blackbody . ) a further reason for the oblivion of case c is that the scenario it modifies ( case a ) is inadequate to represent most detectable objects . in other words , the lack of interest toward case c might not have to do with the correction it applies to case a , but rather with case a itself ; after all , that nebulae _ are _ irradiated by stellar light is beyond doubt . if this is the case , one might wonder why not considering the effect of stellar irradiation as a correction to case b , rather than case a. since case c is the irradiated counterpart to case a , it seems natural to christen `` case d '' the irradiated counterpart to case b ( table [ tab : cases ] ) . note that cases a and b depend solely on the physical conditions in the gas , while cases c and d also depend on the illuminating source , so they are not cases in the same ` local ' sense of cases a and b. case d has never been taken , to the best of our knowledge , into consideration . although this might seem somewhat surprising at first glance , we believe that there are reasons for it ; the arguments developed in this paper will help to understand why , so we defer a discussion of this point to the last section . l|cc & thin & thick + non - irradiated & a & b + irradiated & c & d + in section [ sec : mechanism ] , we will argue that case d is a better representation of real nebulae than case b , the difference between the two scenarios being mostly small but substantial by modern standards of accuracy . curiously enough , case d is implicitly assumed in many ( but not all ) photoionization models , although in a biased way ; the bias can only be eliminated by feeding into the models high - resolution near - uv spectra ( section [ sec : synthesis ] ) . we will also show that the differential fluorescent contribution to the various balmer lines can also affect the intrinsic balmer decrement , pushing it toward either the blue or the red ( section [ sec : spatial ] ) . the resulting spatial distribution of the balmer decrement depends on the spatial distribution of stars and the gas around them . a brief discussion of the dependence of fluorescence enhancement on the nebular and the stellar population s parameters is given in section [ sec : dependencies ] , and on details of the assumed stellar atmospheres is given in section [ sec : atmospheres ] . the main focus of section [ sec : discussion ] is the comparison of fluorescence enhancement to the other major deviation to case b , namely collisional excitation of h@xmath4 . finally , a short summary of the paper is given in section [ sec : summary ] . the abundance analysis of gaseous nebulae is based on a very simple principle . the intensity of a line with respect to , e.g. , @xmath5 is given by the abundance of the emitting ion relative to h@xmath1 times a mathematical expression describing the physical mechanisms of line formation involved . if such mechanisms are known and all the parameters they depend on ( such as the electronic temperature @xmath6 ) can be determined , then the observed value of a line ratio can be used to measure the ionic abundance relative to hydrogen . the specific mechanisms involved depend on the ion and the regime considered . to mention but a few examples , it is known that he i lines in h ii regions mainly form in the recombination cascade of he@xmath1 , plus a density - dependent contribution by collisional excitation from the metastable level 2@xmath7s , which affects mainly the triplet lines @xcite . he ii lines form in the recombination cascade of he@xmath8 . heavy element lines mainly form through collisional excitation from the ground state of abundant ionic stages . finally , h i lines - a standard reference for the intensity of the other lines in the optical range - form mostly in the radiative cascade following recombination of h@xmath1 under case b conditions . it is also known that case b is only a limiting case and not the exact description of conditions in h ii regions ; strictly speaking , it is not even self - consistent , as it provides no mechanism to depopulate the 2_p _ level ( see , e.g. , * ? ? ? however , the predictions made by assuming case b have generally proved to be accurate , and only in recent times has it become possible ( and relevant ) to point out exceptions to it . one such exception is the occurrence of collisional excitation from the ground state , which has recently been rediscovered and closely studied as a consequence of the increasing need for accuracy in abundance determinations @xcite . a further way in which hydrogen may deviate from case b behavior , which is the subject of this paper , is continuum pumping in the lyman lines and the related balmer fluorescent enhancement . continuum pumping is the excitation of a transition by means of the absorption of a stellar continuum photon of the appropriate wavelength . such excitation is followed by a radiative cascade which , in optically thick environments such as those of h ii regions , ends up producing a balmer series photon plus either ly@xmath2 or 2-@xmath3 continuum . in such a way , a ly@xmath9 absorption at 1026 can lead to the emission of an h@xmath2 photon ; a ly@xmath3 absorption at 973 can lead to the emission of either h@xmath9 or pa@xmath2 plus h@xmath2 ; and so on . these balmer photons appear independently of , and in addition to , balmer photons formed in the recombination cascade of h@xmath1 , so these two contributions add up in the total spectrum . as for ly@xmath2 photons , they are scattered and gradually shifted in frequency until they escape the nebula , or are absorbed by dust ; in dustless nebulae , a significant fraction of ly@xmath2 photons is converted into 2-@xmath3 continuum . dust can also destroy higher lyman photons , thus inhibiting pumping ; this possibility will be briefly discussed in section [ sec : discussion ] . it must be stressed that this balmer enhancement mechanism is triggered by pumping at the lyman - line wavelengths , i.e. it depends on the shape of the _ non - ionizing _ stellar field . this runs counter the common wisdom that , in a photoionized object , only the ionizing part of the stellar field matters ; after all , the ionization structure is governed by the ionizing radiation , and the temperature structure is determined by the balance between heating due to photoionization and cooling due to collisional excitation of abundant metals and , in less metal - rich environments , ly@xmath2 emission . as a consequence , in photoionization modeling the effect of non - ionizing stellar radiation on the nebula is generally neglected . yet , the emission spectrum of an h ii region can be altered by continuum pumping even if this mechanism has no other effect on the structure of the nebula . in the next section , we will justify this statement by discussing the role of stellar continuum pumping in case c nebulae , referring in particular to @xcite s @xcite results . the following sections will be dedicated to the role of stellar pumping in the optically thick case , which we have labelled above `` case d '' . in particular , we will estimate the contribution of stellar pumping on balmer intensities and show that it can not be neglected . a good starting point to case d is case c , which applies to nebulae optically thin in the lyman lines irradiated by a stellar continuum . in an optically thin environment , stellar continuum pumping results in a significant enhancement of ly@xmath2 emission , sufficient to bring the ly@xmath2/@xmath5 ratio well above the case b value , rather than below it as predicted for case a. @xcite explains this by showing that , in the vicinity of the star ( i.e. , before stellar lyman - line photons have substantially been absorbed ) continuum pumping provides a source of balmer photons comparable to recombination . if the gas slab is optically thin , recombination and fluorescence act concurrently in the whole volume , so the total emissivity in ly@xmath2 is largely enhanced by the latter mechanism . this argument anticipates one reason why case d has never been taken into consideration : if the gas thickness is large ( i.e. the optical depth is large in the lyman continuum ) , lyman - line photons are used up much earlier than lyman continuum photons , so that the relative effect of continuum pumping on the total emitted spectrum is diluted ; the larger the optical depth , the more diluted the effect , until it possibly becomes negligible altogether . if this were always the case , case d would not have any actual applications . there are , however , three possible exceptions to this conclusion : a ) a high accuracy in abundance analysis is required ; b ) the nebular region affected by fluorescence has an unusually large weight in the observation , due to aperture effects ; c ) the balance between the lyman - line photons and the lyman - continuum photons is more in favour of the former than in the case considered by @xcite . the first circumstance is certainly met in primordial helium abundance determinations , a goal that requires knowing the line formation mechanism of h and he lines with an accuracy better than 1% . the second circumstance is met in spatially - resolved observations sampling small regions in the vicinity of the ionizing cluster ; observations of this kind are more and more common as technology improves . the third circumstance does not seem at first glance very probable when considering that the input spectrum of @xcite is a blackbody ; these results , one may be tempted to conclude , overestimate rather than underestimate the amount of continuum pumping , since a blackbody has no absorption features . although this is true , it must be considered that a blackbody has no lyman jump either , so the lyman line - to - continuum flux of @xcite might actually be an underestimate . to demonstrate this point , in figure [ fig : synthesis_vs_blackbody ] we compare the @xmath10 k blackbody of that paper to the synthetic spectrum of a @xmath11 myr population emitting the same rate of ionizing photons ( i.e. , with the same potential to produce @xmath12 photons by recombination ; more details on this spectrum will be given in section [ sec : cased ] ) . the figure shows that the blackbody does not lie _ above _ the synthesis spectrum : rather , it crosses the near - bottom of the relevant absorption lines , thus underestimating the number of photons capable of continuum pumping . concluding , we see that there is no _ a priori _ reason to discard the relevance of case d. in the following section , we will quantify in a specific case the correction represented by case d with respect to case b , and discuss how this result is expected to depend on the relevant h ii region s parameters . the upper panel of figure [ fig : nehdentehb_m1 ] shows the physical conditions in a model giant h ii region , named m1 , computed with cloudy ( version c08.00 ; * ? ? ? this version of the code includes a full model of the hydrogen atom with strong radiation field physics ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * and appendix [ sec : appendix ] of this paper ) . for the sake of simplicity , we assumed a dustless nebula with a simple spherical geometry , with starting radius @xmath13 pc , hydrogen number density @xmath14 @xmath15 and filling factor @xmath16 . although a blister model would be more appropriate to represent many , if not most , h ii regions , we prefer to stick to the simplest possible case . the volume - integrated predictions presented in this paper do not depend on the assumed geometry , while the spatially - resolved predictions do ; it is therefore impossible to give figures suitable for any geometry and any line of sight , so the reader is warned that our spatially - resolved predictions are given more as a rough indication of the magnitude of the effect than as a definitive estimate . the model s total metallicity is @xmath17 , and the detailed elementary abundances are given in table [ tab : chem ] ; dust is not important at such low metallicities . thermal line widths are assumed ; the influence of a non - thermal contribution is discussed in section [ sec : turbulence ] . the ionizing source is the spectral energy distribution ( sed ) of a @xmath11 myr stellar population , emitting ionizing photons at a rate log @xmath18 = 52.09 s@xmath19 . the sed has been computed with sed@ ( * ? ? ? * and references therein ) , and does not include stellar winds . the effect of stellar winds on the predictions is discussed in section [ sec : atmospheres ] . assuming that the model s emission is described by case b , a fairly constant @xmath12 emissivity profile is expected , given the shallow temperature gradient . this , however , is not the case : two exceptions to case b behavior can be seen in the @xmath12 profile ( figure [ fig : nehdentehb_m1 ] , bottom panel ) . [email protected] & & + element & + & & + he & 10&89 + c & 7&62 + n & 6&38 + o & 7&87 + ne & 7&15 + s & 6&47 + cl & 4&48 + ar & 5&50 + fe & 5&81 + in the external part ( @xmath20 70 pc ) there is a small bump in the emissivity profile : this bump is caused by collisional excitation of neutral hydrogen , a mechanism ignored in case b and which only becomes effective in the outer part , thanks to the drop in the hydrogen ionization fraction . a very low metallcity gas , which we consider here , will have a high kinetic temperature so these collisional processes are especially important ; more details on this can be found in @xcite . in the inner part there is a much larger bump , extending through a substantial fraction of the total radius and whose maximum value more than doubles the expected case b value , which is shown as a dotted line . when the upper and lower panels of figure [ fig : nehdentehb_m1 ] are simultaneously considered , it becomes clear that density and temperature can not be responsible for the excess emissivity , as one can find pairs of points with the same @xmath6 and @xmath21 values but very different emissivities . in other words , the cause of this behavior can not be sought in the gas : this leaves the radiation field as the only alternative . indeed , the excess emissivity is caused by fluorescent excitement of @xmath12 by continuum pumping of lyman photons ( mainly ly@xmath9 ) . this is consistent with the position of the bump , close to the ionizing source ( note that this bump has nothing to do with the breaks in the @xmath21 and @xmath22 gradients seen at 6 pc , which are caused by the recombination of he@xmath8 and o@xmath23 ) . in the following sections , we will discuss whether the behavior of this model is representative of real nebulae and which its consequences are for the analysis of h ii regions , considering several specific issues . we will start by considering a simple model of a spherical h ii region , ionized by a population synthesis model ( section [ sec : synthesis ] ) , and show the effect of continuum pumping on the balmer intensities and ratios . in the following sections we will complicate this basic scenario by discussing the effect of dust , turbulence , geometry , or a different metallicity ( section [ sec : dependencies ] ) . since the amount of continuum pumping depends on the detailed shape of the spectrum , we will also discuss the effect of assuming stellar atmospheres with shocked stellar winds . since no high - resolution synthesis models are available with shocked winds , we will consider the case of a model h ii region ionized by individual stars ( section [ sec : atmospheres ] ) . sections [ sec : discussion ] and [ sec : summary ] will be devoted to the discussion and conclusions , respectively . the black thick curve in figure [ fig : sed_uv ] shows the near - uv portion of the sed fed into the model described in section [ sec : cased ] ; we stress again that this sed , as all the synthesis models used in this work , has been computed with stellar spectra with no emission lines . in this wavelength range , the spectral sampling of this sed is @xmath24 ; this value is determined by the grid of stellar libraries used , i.e. it does not depend on the way in which population synthesis is carried out ; any synthesis model using the same stellar library would have the same resolution ( this is the case , for example , of starburst99 models : * ? ? ? * ) . at such resolution , stellar absorption lines are smoothed out , so that the number of lyman photons in the spectrum is overestimated . since the mechanism studied in this work depends directly on the number of lyman - line photons emitted , it is important to establish whether fluorescent enhancement of @xmath12 can actually occur in h ii regions or it is just an artifact of low - resolution synthesis models . the only way to answer this question is by feeding a high - resolution sed into the photoionization code . to address the issue mentioned in the previous section , we computed a new photoionization model , labelled m2 , identical to m1 in all respects except for the stellar source , which has been replaced with a high - resolution sed in the 900 - 2000 range . this sed represents a population of the same stellar parameters ( age and metallicity ) as those of the original sed , but has a much better spectral sampling @xmath25 = 0.3 in the near - uv range ; a portion of this spectrum is shown in figure [ fig : sed_uv ] ( thin grey curve ) . the sed is taken from a library of high - resolution synthesis models computed with the code sed@ @xcite to carry out this work . the library has been computed with the non - rotating tracks of the geneva group @xcite , including standard mass - loss rates computed at five metallicities : @xmath26=0.040 , 0.020 , 0.008 , 0.004 and 0.001 . the isochrones are calculated by means of a parabolic interpolation between the tracks ( log @xmath27 , log @xmath28 ) in the hr diagram on a variable mass grid as prescribed in @xcite . the isochrones were generated at specific ages with a time step @xmath29 myr for 1 - 10 myr . we have assumed a power - law imf , with slope @xmath30 @xcite and low- and high - mass cutoffs @xmath31 and @xmath32 , respectively . to model the sed , we have combined two different sets of atmosphere models , with low and high resolution respectively . low resolution atmosphere models are standard kurucz lte atmospheres . high resolution atmosphere models cover the wavelength range from 900 to 2000 , and have been computed using the same atmospheres as the ones described in @xcite but in the uv domain . hot stars ( @xmath33 k ) are modelled as line - blanketed , non - lte , plane - parallel , hydrostatic atmospheres @xcite , and stars with @xmath34 k are modelled with kurucz lte atmospheres @xcite . we have assumed no contribution in the uv from stars with lower temperatures . the high - resolution synthetic spectra are obtained with the program synthe by @xcite . the portions of the sed shortward of 900 and longward of 2000 are equal in the low- and high - resolution models . the main features of these models are given in table [ tab : models ] , together with those of the models presented in the remainder of the paper . a summary of the difference between m1 and m2 from the point of view of the fluorescent contribution to @xmath12 is given in table [ tab : hrvslr ] . the three rows in the table correspond to different ways of integrating the emissivity along the shell structure of the model : in the 1d integration the emissivity of each shell is weighted by the thickness of the shell @xmath35 ; in the 2d integration it is weighted by the ring surface @xmath36 ; and in the 3d integration it is weighted by the shell volume @xmath37 . observationally , the three cases correspond to a beam - like aperture , a narrow slit , and an aperture including the whole object respectively . since fluorescent excitation takes place in the inner part of the nebula , we expect that the effect on integrated luminosities of differences in the local emissivity is amplified in 1d integrations with respect to 3d integrations . as expected , table [ tab : hrvslr ] shows that @xmath12 is spuriously enhanced in m1 with respect to m2 as a consequence of the poor resolution of its input spectrum , which partially suppresses stellar absorption lines . the bias amounts to as much as 0.04 dex ( 9% ) in the case of a beam - like simulated aperture , relative to the _ total _ h@xmath2 intensity . cccccccc model & & nebular boundary & + & @xmath38 2000 & 2000 - 900 & @xmath39 900 & & f & r & ce + m1 & lr & lr & lr & radiation bounded & + & = & = + m2 & lr & hr@xmath40 & lr & radiation bounded & = & = & = + m3 & lr & suppressed@xmath41 & lr & radiation bounded & @xmath42 & = & = + m4 & lr & hr@xmath40 & lr & density bounded & = & = & @xmath42 + m5 & lr & suppressed@xmath41 & lr & density bounded & @xmath42 & = & @xmath42 + ccc & & + integration & @xmath43 -1 + & & + 1d & 0.09 + 2d & 0.03 + 3d & 0.02 + since predicting balmer line intensities is not a common application of photoionization models , which is the reason for pointing out this bias ? the main reason to do it is estimating the amount of fluorescent excitation in real objects so that it can be taken into account when chemical abundances are determined . the exercise realized in this section provides a necessary piece of information in order to know whether such estimation can be carried out with low - resolution models , or high - resolution models are needed instead . the results above indicate that the bias in @xmath12 introduced by low - resolution models is substantial if compared to our best possible guess of the total emitted @xmath12 intensity , which is given by m2 . therefore , the results of table [ tab : hrvslr ] indicate that low - resolution synthesis spectra are not suitable to study the effect of fluorescence , and that high - resolution spectra are needed instead ; they also indicate that the predicted spatially - resolved line ratios relative to @xmath5 predicted are grossly overestimated in the direction of the center ; this must be taken into account when photoionization models are used to fit spatially - resolved spectroscopic data , as done , e.g. , in @xcite . note that cloudy automatically rebins any input spectrum , so that to take advantage of the larger spectral resolution it has been necessary to increase cloudy s default continuum mesh , as explained in the code s documentation ( hazy , part 3 , page 410 ) . we carried out several different experiments in this direction exploring the resolution range @xmath44 , corresponding to spectral samplings @xmath45 at @xmath46 1000 . some of the spectra used internally by cloudy after such rebinning are shown in figure [ fig : comparing_rcloudy_uv ] in the ly@xmath9 wavelength region . we found no significant differences in the emission - line spectrum below @xmath47 . as a final comment , we note that , since real massive stars may develop p - cygni line profiles due to the presence of strong stellar winds , at very high spectral resolution one would also need to consider both the detailed line profile and the relative velocities of the star and the nebula , which determine which part of the spectrum fall at the relevant wavelengths . having established that low - resolution spectra are unsuitable to compute the fluorescent contribution to balmer lines , we go on to address this problem by means of high - resolution spectra , considering the specific case of model m2 . the fluorescent intensities in m2 can be estimated by computing a new model m3 , identical to m2 in all respects except for fluorescent excitation , which is suppressed by setting to zero the relevant portion of the continuum ( but it can also be switched off by the cloudy command `` no induced processes '' ) . the fluorescent enhancement of m2 is computed as the difference between the balmer line intensities of m2 and m3 ( table [ tab : hr2vsnopumping ] ) . for the reasons explained in section [ sec : photo_models_hr ] , the enhancement is particularly large in the beam - like integration : specifically , in our models _ as much as 10% of the total intensity in h@xmath2 and 14% in h@xmath9 seen in the direction of the nebular center is caused by fluorescence rather than recombination_. for completeness , we have also computed the fluorescent enhancement of balmer lines of model m1 , to quantify the fraction of emissivity that would - erroneously - be attributed to fluorescence enhancement , had we used model m1 to this aim : in that case , the estimated fluorescent enhancement is almost twice as large as in the unbiased case ( table [ tab : hr2vsnopumping ] ) . as anticipated in section [ sec : photo_models_hr ] , this demonstrates that inputting low - resolution spectra to photoionization models grossly biases the estimated fluorescence contribution . thus , the first general result is that fluorescence does indeed has a sizeable influence on the emitted balmer spectrum . in the remainder of this section we will discuss the implications of such influence . in the following sections we will briefly discuss how our results depend on the model s parameters , such as the geometrical assumptions on the nebular distribution and the parameters that define the stellar populations . cccccccc integration & & & + & h@xmath2 & h@xmath9 & h@xmath3 & & h@xmath2 & h@xmath9 & h@xmath3 + 1d & 0.10 & 0.14 & 0.14 & & 0.17 & 0.21 & 0.26 + 2d & 0.03 & 0.04 & 0.05 & & 0.06 & 0.06 & 0.09 + 3d & 0.02 & 0.02 & 0.02 & & 0.04 & 0.03 & 0.04 + the results listed in table [ tab : hr2vsnopumping ] have important implications for abundance determinations . when a beam aperture is applied to the center of model m3 , at least 10% of the total balmer intensity comes from fluorescent excitation . if model m3 were a real object analyzed under the assumption of case b emissivity , and assuming that no fluorescent excitation is taking place in other ions , the inferred abundances of helium and heavy elements would be underestimated by up to 0.06 dex . even in the most favourable case , the large aperture corresponding to a 3d integration , a 0.01 dex ( 2% ) bias would be introduced ; although this is negligible for most abundance measurements , it would be crucially large for high - precision applications such as the determination of primordial helium . although helium lines might also be affected by fluorescent excitation , line ratios will in general be different from what predicted when neglecting fluorescent excitation . a second implication concerns extinction measurements . the upper panel of figure [ fig : ha2hb_hahb_r ] shows the local @xmath12/@xmath5 radial profile of model m2 . in the innermost 20 pc and the outermost 20 pc , this ratio differs considerably from the value expected under case b. the deviation in the external region is caused by collisional excitation : @xmath48 rises sharply because collisional excitation is more effective for @xmath12 than for @xmath5 @xcite . the deviation in the internal region is caused by fluorescence , as can be shown by an analysis of the pattern it follows . fluorescent excitation of @xmath12 and @xmath5 is mainly triggered by ly@xmath9 and ly@xmath3 photons , respectively ; the optical depth in ly@xmath9 is larger than that in ly@xmath3 , so that @xmath12 peaks at a smaller radius than @xmath5 ( figure [ fig : ha2hb_hahb_r ] , bottom panel ) . as a consequence , the @xmath12/@xmath5 ratio initially increases , then it decreases as ly@xmath9 photons become exhausted and ly@xmath3 photons are still available , and eventually increases again to approach case b as both ly@xmath9 and ly@xmath3 photons are exhausted . this behavior is to be expected no matter which the initial proportion of ly@xmath9 and ly@xmath3 photons is , as it only depends on the optical depth at the two wavelengths . a spectrum with a larger flux in ly@xmath9 might produce a @xmath12/@xmath5 ratio larger than the case b one , but the succession between increasing and decreasing stretches would persist . this interpretation of the @xmath12/@xmath5 behaviour has independently been verified by running models identical in all respects to model m1 except for the stellar continuum , in which small wavelength ranges around the lyman lines wavelengths have been selectively suppressed . for example , in the model with no continuum between 1000 and 1050 the @xmath48 enhancement disappears completely , leaving only a @xmath48 depression around @xmath49 10 pc . for brevity , the detailed description of these models has been omitted . the difference in fluorescent enhancement of different lines is illustrated by table [ tab : hr2vsnopumping ] for @xmath12 , @xmath5 and @xmath50 . these differences are model - specific and depend on the number of photons available to excite the various lyman lines ( which in turn depends on the precise shape of the spectrum and , in a real nebula , on line broadening and doppler shifts ) , and on the electron distribution among different sublevels of a given level , which influences the downward radiative paths available and which depends on the local gas density ( @xmath51-mixing is quite ineffective at such low densities ) . we can not say whether current synthesis model are accurate enough for these differences to be real ; but , if we take them at face value , the straightforward consequence of this differential enhancement is a change in the intrinsic balmer decrement with respect to its case b value . table [ tab : ha2hb ] quantifies this effect for model m2 , by comparing its @xmath48 values to those of model m3 . the case b value for these models temperature is 2.81 ; the differences between such value and the numbers in the table are determined by the competition among three factors . on one hand , fluorescent excitation enhances @xmath5 more than @xmath12 ( table [ tab : hr2vsnopumping ] ) : this tends to _ decrease _ the balmer decrement . on the other hand , collisional excitation enhances @xmath12 more than @xmath5 @xcite , tending to _ increase _ the balmer decrement . finally , fluorescent excitation takes place in the center and is therefore favoured in 1d integrations , whereas collisional excitation takes place in the external part of the model and is favoured in 3d integrations . these dependencies explain the patterns seen in the table : in model m2 ( with a `` normal '' amount of continuum pumping ) , fluorescence bluening dominates over collisional reddening in the 1d aperture , while collisional reddening dominates in the 3d aperture ; in model m3 ( where continuum pumping has been suppressed ) , fluorescence is quenched , so all the ratios are larger than the case b decrement , and the difference increases proportionally to the weight given to external regions due to the increasing role of collisional excitations . the consistency of this interpretation has been verified by computing density - bounded versions of models m2 and m3 , labelled m4 and m5 . in m4 and m5 , the nebula is truncated at a radius of 70 pc , so that the region of collisional excitation is not included in the computation : as a result , the @xmath48 ratios are either lower than ( m4 , density - bounded variant of m2 ) or close to ( m5 , density - bounded variant of m3 ) the case b value . we can go a bit further by recalling again that the optical depth is different for different lyman - line photons , in such a way that they are selectively absorbed at different depths into the nebula . note that this is true no matter which the initial proportion of lyman - line photons is , so that no role is played here by the caveat of section [ sec : effect_balmer_dec ] regarding the accuracy of the spectrum . as by - products of the absorption of lyman photons , different balmer lines are also produced in different proportions at different depths into the nebula ( figure [ fig : ha2hb_hahb_r ] , bottom ) : as a consequence , the purely fluorescent @xmath48 ratio is expected to vary as a function of depth , being in general different from the case b value ( figure [ fig : ha2hb_hahb_r ] , top ) . therefore , the projected @xmath48 ratio also depends on the particular direction considered , as it depends on the proportion of fluorescence - excited atoms intercepted along the particular line - of - sight , and therefore on the distance from the exciting stars . of course , the amplitude and scale length of this variation depend on the scale of the fluorescence zone , which depends on nebular geometry and is probably very thin in most objects ; specifically , it will be shown in section [ sec : nebular_geometry ] that the more plane - parallel is the gas distribution , the thinner is the fluorescence zone . but in regions with thicker geometries , such dependence may lie above the spatial resolving power and be misinterpreted as a spatial variation of interstellar extinction . there are numerous observations reporting systematic changes in the balmer decrement along different lines of sight and attributing them to differences in dust extinction ( see , e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . to investigate the possible role of continuum pumping in such differences , we simulate spatially resolved observations by plotting the predicted @xmath48 ratio along a simulated narrow slit ( figure [ fig : ha2hb_projected ] ) note that this is not the same as the 2d integration , which describes the _ integrated _ emission of a narrow slit . figure [ fig : ha2hb_projected ] shows that , with fluorescent excitation properly taken into account , the @xmath48 ratio actually increases from a minimum value near the center to a maximum in the outskirts . with no other observations available , this pattern could be interpreted as a variation in true extinction and bias the dust distribution inferred from the @xmath48 ratio . we believe that , for the most part and in most cases , the observed @xmath48 gradient does indeed reflect the dust distribution ; furthermore , it makes physical sense that such distribution is correlated to the spatial position of massive stars . however , we wish to call attention to the fact that a part of the gradient might also be intrinsic , arising as a consequence of the spatial segregation of fluorescent enhancement and collisional excitation of h@xmath4 . even adopting a conservative stance , the evidence presented in this section should be interpreted as a measure of the minimum uncertainty affecting interstellar extinction measurements . a rough estimate based on model m2 situates such uncertainty around @xmath520.02 dex in @xmath48 , or @xmath52 0.05 dex in @xmath53@xmath5@xmath54 . ccccc integration & + & m2 & m3 & m4 & m5 + 1d & 2.70 & 2.84 & 2.68 & 2.83 + 2d & 2.82 & 2.85 & 2.80 & 2.83 + 3d & 2.86 & 2.86 & 2.82 & 2.83 + assuming that the synthesis model used in this work is a good approximation to an actual sed , the effect of continuum pumping is inescapable . on the other hand , a natural objection to this conclusion is that it would cause apparent abundance gradients of appreciable amplitude , which would have hardly gone unnoticed . existing evidence is not conclusive : for example , in the works by @xcite and @xcite , radial gradients in the inferred abundance can neither be confirmed nor dismissed . there may be several causes for this . first of all , the error bar in the derived abundances is of the same order of the effect we are concerned with . secondly , we are considering here a spherical geometry , whereas there are observational evidences of a blister geometry in some resolved h ii regions ; in such a case , the region affected by fluorescence enhancement would be compressed in a thin shell . finally , the spatial distribution of stars is not always centrally concentrated as in our naf , highly idealized model . this comment mainly concerns giant h ii regions , in which the scale of the system exceeds the strmgren radius of individual stars . in such cases , we can expect a complex geometry , which can not be represented by our simple model . in the extreme case of homogeneously distributed sources , local predictions would approach the volume - averaged result of a centrally concentrated source . in this section we will briefly discuss how fluorescent excitation , whose effectiveness is a function of the lyman - line photons available in the exciting spectrum , is affected by the stellar population s parameters . only a qualitative description of the input synthesis models will be used to this aim . the discussion below is based on models with @xmath55 and @xmath56 and ages between 1 and 15 myr . the outcome of a comparison between different seds depends on which variables are held fixed . in the present case the comparison must be carried out at fixed @xmath18 , because we wish to compare the balmer line intensity due to fluorescent excitation to the intensity due to recombination , which is determined by @xmath18 . the fraction of @xmath12 provided by fluorescent excitation monotonously increases as age increases , amounting to , e.g. , 3% , 4% and 6% at @xmath57 , @xmath58 and @xmath59 myr , respectively . note that this increase is triggered by the progressive decrease in the number of ionizing photons , which declines with age ; this is only partially offset by the increase in the absolute value of the equivalent width of the lyman @xmath2 absorption line taking place after 3 myr . this can be seen in figure [ fig : age ] , which shows a sequence of seds in the age range 1 - 10 myr normalized to the number of ionizing photons . comparison of models spanning a wide metallicity range ( @xmath60 ) suggests that metallicity has no effect in very young objects , and only a slight effect in older objects , in the sense of larger fluorescent contributions at larger metallicities . our predictions for the 3d integration of models with ages in the 1 - 5 myr range are listed in table [ tab : z ] . relative contributions for other apertures might be estimated on the basis of these . lccccccc & + & 1 & 3 & 5 + 0.001 & 0.02 & 0.02 & 0.03 + 0.004 & 0.02 & 0.02 & 0.04 + 0.008 & 0.02 & 0.03 & 0.04 + 0.02 & 0.02 & 0.03 & 0.04 + by `` nebular geometry '' we exclusively refer to the distinction between thick ( spherical ) and thin ( plane - parallel ) models . the volume - integrated fluorescent contribution does not depend on geometry : roughly speaking , the relative excitation rates depend on the number of photons available for each process ( ionizing photons in the case of recombination , lyman - line photons in the case of fluorescent excitation ) . on the other hand , the beam - integrated fluorescent contribution does depend on nebular geometry : the more plane - parallel the model , the closer its beam- and volume - integrated line intensities . asymptotically , the intensities seen through different integrations coincide . density - bounded models differ from ionization bounded models in that the outermost part is missing . the fraction of fluorescent excitation in these models increases proportionally to the extension of the missing part , until the slab becomes so thin that it falls under case c. our simulations so far have assumed a dustless nebula . this is not a realistic assumption in most nebulae , and the difference may be relevant in the context of this research since dust grains absorb uv photons . to understand this issue , we have computed a series of dusty models with varying metal abundances and dust content . our models , which span the metallicity range @xmath61 , show no dependence of the relative fluorescent contribution on dust content ( although , of course , absolute fluxes are altered ) . the reason for this behaviour is the approximate constancy of dust opacity in the relevant range , which roughly goes from the energy of ly@xmath9 to a few rydbergs . the grains used in these models are orion - type , that is relatively large ; using ism - type grains , with a distribution extending to smaller wavelengths , uv photons would be more effectively blocked , resulting in smaller and dimmer nebulae . however , the fluorescent relative contribution would stay constant . this is a consequence of the small relative variation of dust opacity in the relevant range , which goes from ly@xmath9 to @xmath62 ( this range encloses more than 90% of the total number of ionizing photons ) ; see , e.g. , figure 19 of @xcite , which shows that the total variation of the absorption coefficient in the relevant range is within a factor of two , which defines the maximum bias that might affect our dustless results . the line widths of actual h ii regions are larger than the expected thermal width . the origin of this non - thermal contribution is currently unknown , although it has been suggested that it might be turbulence @xcite . whatever its cause , the effect of an increased line width is to make continuum pumping more efficient by increasing the line flux available ( see , e.g. , * ? ? ? since our models so far have been computed under the assumption of thermal line width , we will discuss here the consequences of assuming increasing line widths , assuming they are caused by turbulence . figure [ fig : turbulence ] illustrates the effect of turbulence on our reference model m2 at different values of the microturbulent velocity , which is assumed gaussian . the fluorescent component increases linearly with the turbulent velocity in all the range explored ; specifically , from 0 to 100 km s@xmath19 it doubles in 1d integrations and triples in 2d integrations , whereas the increase in 3d integrations is even larger . this gives an indication of how this paper s estimates should be increased to take turbulence into account . at even larger velocities , the effect would continue to increase . in this section we will consider the case of h ii regions around individual stars to discuss how our results depend on the details of the stellar atmosphere . in section [ sec : casec ] we mentioned two stellar spectral features , not present in blackbodies , that may affect the balance between the lyman - line photons and the lyman - continuum photons , and , hence , the relative importance of fluorescent enhancement in balmer lines : the lyman jump and the stellar absorption lines . these two features affect continuum pumping in opposite directions . on one hand , the ratio of non - ionizing to ionizing photons depends on the size of the lyman jump : hydrogen continuum pumping has a greater relative importance in a stellar atmosphere with a lyman jump than in a blackbody case . on the other , absorption lines at the lyman frequencies decrease the number of continuum photons available to pump , thus quenching continuum pumping . both features depend on the characteristics of the star , i.e. on its spectral type and luminosity class , metallicity , and strength of the stellar wind ; for example , the size of the lyman jump is larger in an o9v star than in an o3v star . the stellar flux emitted by massive stars in the spectral range between the lyman jump and 1200 has never been studied in terms of its effect on the surrounding nebular material : usually , only the ionizing part of the stellar fluxes is considered to be important . however , the arguments of section [ sec : mechanism ] show the importance of the non - ionizing part . fortunately , our knowledge of the characteristics of the stellar flux in this spectral range has enormously improved in the last decades thanks to the far - uv spectroscopic observations provided by space telescopes ( mainly copernicus and fuse : see , e.g. , * ? ? ? * ; * ? ? ? * ) and the development of stellar atmosphere codes for massive stars ( e.g. , * ? ? ? we review these characteristics here and give examples of how they may affect the behavior of photoionization models including continuum pumping . to this aim we considered two sources of information : the far - uv spectral atlases compiled by @xcite and @xcite , and the predictions of wm - basic models @xcite . below , we describe the main stellar spectral features that may affect case d predictions . in particular , we are interested in the lyman jump and those spectral features affecting the stellar flux emitted at the lyman wavelengths . in figure [ fig : wmbasic_spt ] we show the spectra of wm - basic models corresponding to four different spectral type - luminosity class combinations , in the wavelength range 900 - 1070 , which includes the lyman jump , the relevant h lyman lines and other stellar features of interest . only this range is represented because we are only interested in those lines that can produce balmer lines through continuum pumping , i.e. ly@xmath9 and lyman lines of shorter wavelengths . two types of wm - basic models were computed and included in the figure ( see also table [ tab : wmbasic_pars ] ) : black solid lines represent models with a smooth wind velocity law , light dashed lines represent models including the effect of shocks in the wind . this second type of models are necessary to reproduce some of the lines observed in the far - uv spectra of massive stars , such as o vi lines ( see below ) . this feature , located at @xmath63 , is the limit between the ionizing and the non - ionizing part of the stellar flux . it is produced by the presence of a sharp limit in the ionization cross section for the ground level of h i. therefore , its size depends on the amount of neutral h i atoms in the atmosphere , which is determined by the temperature and density in the region of the stellar atmosphere where the stellar continuum at this wavelength is formed . generally speaking , the lyman jump will be larger at lower effective temperatures and larger gravities ( since in these cases hydrogen is expected to be more recombined ) . note , however , that the size of the lyman jump may also be affected by other features of massive stars winds ( see , e.g. , * ? ? ? * ; * ? ? ? since practically no observations of the lyman jump exist ( the non - ionizing part of the stellar flux is generally not accesible to direct observations ) , we have to rely on stellar atmosphere models . as already mentioned , the lyman jump is larger in stars with larger gravities ( dwarf stars ) and later spectral types ( figure [ fig : wmbasic_spt ] ) . both observations and models show that the stellar flux at the lyman wavelengths may be affected not only by the h lyman lines , but also by resonant lines of heavy elements at different wavelengths that may develop p - cygni profiles if the characteristics of the stellar wind are appropriate . if the p - cygni profile is completely developed and the wind terminal velocity is large enough , the saturated blueshifted absorption may reach bluewards the wavelength of the closest h lyman line ( see figure [ fig : wmbasic_spt ] ) . as illustrative examples , we discuss here the behavior of the o vi @xmath641032 , 1038 doublet and the c iii @xmath65979 lines as a function of spectral type , luminosity class and wind characteristics . figure [ fig : wmbasic_spt ] also shows the location of the s vi lines , which may affect other h lyman lines . however , this is only given for information and its behavior is not discussed here . the o vi @xmath641032 , 1038 lines may affect the spectral region of ly@xmath9 . the spectroscopic catalogues by @xcite and @xcite show that these lines exhibit a single p - cygni wind profile in early - type o stars , both in dwarfs and supergiants . in later o - type spectra , two distinct blueshifted absorption features are often visible and , in the case of supergiants , o vi lines are still visible in b1i stars . these lines depend only weakly on the temperature , luminosity , and metallicity of the star , and can still be seen even at metallicities as low as in the smc . the presence of lines of this `` superionization '' species in the spectrum of ob stars has been commonly attributed to the effect of shock - produced x - rays on the ionization distribution in hot stars winds ( see , e.g. , * ? ? ? * for a discussion concerning the origin of this exotic species ) . in fact , only stellar atmosphere models including shocks in the stellar wind can produce the o vi lines . figure [ fig : wmbasic_spt ] shows the synthetic uv spectra of wm - basic models with and without shocks . in the case of o3 stars , strong o vi p - cygni profiles are only developed if shocks are included in the models . something similar occurs in the o9.5 models where , although a p - cygni profile is not developed , we can still find strong extended blueshifted absorptions . in all the cases , the presence of the o vi lines totally quenches the stellar continuum flux in ly@xmath9 . this only happens if the terminal velocity of the wind is above @xmath661700 kms@xmath67 . our second example is the c iii @xmath65979 line . @xcite show that , in supergiants , this line appears as a broad p - cygni profile with a broad absorption component in spectral types from o6 to o9.7 , while a broad absorption feature is detectable in subtypes o9.5 to b3 v. on the other hand , @xcite showed that in smc supergiants the line saturates despite the lower abundances , and does not appear weaker than in their lmc counterparts . figure [ fig : wmbasic_spt ] shows the behavior of this line as predicted by wm - basic models for the stellar cases considered . in both o9.5 models the c iii line develops a p - cygni profile whose blue absorption affects the flux in ly@xmath3 . interestingly , in the o9.5v star this line is inhibited by the inclusion of shocks . in this section we discuss how the spectral features considered above make ionized nebulae deviate from case b to case d. to this aim we constructed a set of spherical , dustless , constant density photoionization models with cloudy . the internal radius and density of the models were @xmath68=1 pc and @xmath69=200 @xmath15 , respectively , and no turbulent line widening was assumed . the seds predicted by wm - basic for the stellar models summarized in table [ tab : wmbasic_pars ] were included as input in these cloudy models , normalized to log@xmath70(h@xmath4)=52.09 s@xmath19 . figures [ fig : o3i ] through [ fig : o95v ] show the predicted h@xmath2 and h@xmath9 emissivities and the @xmath48 ratio for different stellar cases : similarly to figure [ fig : wmbasic_spt ] , dashed and solid lines indicate results obtained with or without shocks , respectively , with full inclusion of fluorescent excitation , while dotted lines are results obtained by switching fluorescent excitation off . the behavior of ratios can be easily understood by considering the spectral features of each model as a function of its properties . in o3 stars , we see that while the behavior of the @xmath48 line ratio is very close to case b predictions when a smooth stellar wind velocity law is assumed , it is decreased with respect to case b predictions when a shocked wind stellar model is considered . in the case of a smooth wind law , even though the lyman jump is small and the stellar continuum flux at the ly@xmath9 and ly@xmath3 wavelengths is dimished , the local h@xmath2 and h@xmath9 emissivities are noticeably affected by stellar continuum pumping in the region close to the stars , resulting in a strong deviation from case b predictions . however , the h@xmath2 and h@xmath9 variation are similar , in such a way that their ratio is close to case b predictions . on the other hand , in the shocked - wind stellar atmosphere models , the presence of the o vi p - cygni line ( see figure [ fig : wmbasic_spt ] ) leaves no ly@xmath9 photons , and hence the fluorescent contribution to h@xmath2 is negligible . since the fluorescent excitation of h@xmath9 is similar to the previous case , the combined effect is a lower @xmath48 ratio in the inner part of the nebula as compared to case b predictions . a similar line of reasoning can be followed to understand the other stellar cases represented in figures [ fig : o3i ] through [ fig : o95v ] . for example , the large h@xmath2/h@xmath9 line ratio predicted in the o9.5i case when a smooth wind law is used in the stellar atmosphere model calculation is a consequence of the large flux in ly@xmath9 ( which triggers a large fluorescent contribution to @xmath12 ) and the small flux in ly@xmath3 , due to the presence of the ciii line ( which quenches the fluorescent contribution to @xmath5 ) . ( note that in this case the lyman jump is larger . ) this large deviation from case b vanishes when the shocked - wind stellar atmosphere model is considered , due to the development of a broad absorption feature in the o vi lines that affects ly@xmath9 . in this case , the continuum pumping at ly@xmath9 and ly@xmath3 is zero ( note , however , that the h@xmath2 and h@xmath9 emissivities slightly deviate from case b. this is produced by pumping in ly@xmath71 ) . this example illustrates how deviation from case b to case d depends on the spectral features appearing in the spectral region considered . generally speaking , we can conclude that the characteristics of the wind has a sizeable effect on the predicted balmer line ratios . models with shocked winds have lower @xmath48 wih respect to models with smooth wind - velocity laws . lrrrrr & & + & & & & & + & & & & & + & & + o3@xmath72i & 42233 & 3.73 & 18.6 & -5.15 & 2800 + o9.5@xmath72i & 30463 & 3.19 & 22.1 & -5.67 & 2200 + o3@xmath72v & 44852 & 3.92 & 13.8 & -5.77 & 2800 + o9.5@xmath72v & 31884 & 3.92 & 7.2 & -6.89 & 1300 + the discussion above has made clear that the amount of lyman continuum pumping in a nebula is crucially dependent on the line profiles . these depend on both the wind structure and the relative motion between the stellar source and the gas , so the case should be considered in which the lyman lines in the stellar continuum are in emission rather than absorption ; in such a case , the nebula could produce more excitation than expected . we have taken such possibility into account by computing a sequence of models with varying blocking factors , whose effect is to amplify the amount of continuum pumping by a given factor , which can be either smaller or larger than one ( see appendix a in * ? ? ? * ) . the result is illustrated by figure [ fig : bf ] , which shows the ratio between the total h@xmath9 emission and the case b h@xmath9 emission as a function of the blocking factor in a radiation - bounded spherical model ionized by a stellar continuum ( solid line , left - hand scale ) . on the same figure , we have plotted the predictions for the first zone of the same models , a case corresponding to classical case c. in either case , a blocking factor larger than 1 implies that pumping is amplified with respect to the predicted value . it can be seen that increasing values of pumping trigger @xmath5 intensities well above predictions , particularly under case c conditions . in this section we highlight the differences between fluorescent and collisional excitation , in order to provide some useful guidelines to minimize their impact on observations aimed at abundance determinations . both mechanisms provide channels to increase the intensity of balmer lines , but they act with different effectiveness depending on the location within the nebula and the physical conditions regime . the following is a quick summary of such analogies and differences . along any given line of sight , fluorescent excitation takes place in the nebular region which lies the closest to the illuminating stars . collisional excitation , on the contrary , takes place preferently in the recombination zone that marks the outer limit of the h ii region . such region may be lacking totally or in part if the nebula is density - bounded , whereas a fluorescent region always exists . a further consequence of the different position in the nebula is that the relative contribution of fluorescence is enhanced in slit observations . finally , fluorescence appears to be happening in all kind of nebulae , with no strong preferences for metallicities and with a tendency to increase at increasing ages , whereas collisional excitation is effective only in hot , young nebulae . all in all , then , fluorescence appears to be a more ubiquitous process . the _ integrated _ value of fluorescent excitation does not depend on the geometrical distribution of the gas ; neither does the _ integrated _ value of collisional excitation , provided the region is radiation bounded . therefore , the two processes are pretty much geometry - free as long as the whole nebula is integrated , and a detailed knowledge of geometry is not necessary . fluorescent excitation might depend much more than collisional excitation on the distribution of stars in the case of spatially resolved observations . this circumstance is further complicated by the fact that it is not necessarily dominated by the hottest stars , as are recombination and , probably , collisional excitation . both processes also depend on the illuminating sed , but collisional excitation depends on it in a much weaker way than fluorescent excitation . the dependence of collisional excitation on the sed arises on the heating power of the ionizing photons : the harder the average photon , the hotter the gas , so younger clusters are more collisionally excited than older clusters . however , a rough knowledge of the cluster s age and its sed is usually sufficient to this aim . on the other hand , fluorescent excitation crucially depends on the sed , and specifically on the detailed profile of the relevant spectral features , which determine the proportion of photons available to excite the various lyman lines . theoretically , such shape depends on the stellar mix , the metallicity and details of the atmosphere . examples of this are given in section [ sec : atmospheres ] and in @xcite , where it is shown how the stellar ly@xmath9 flux may be affected by a neighbouring o vi line . this extreme dependence implies that moderate nebular motions might drastically change the result by doppler - shifting the line photons seen by the gas . in sum , uncertainties related to the sed affect the estimation of fluorescent excitation , while they have a much smaller impact on the evaluation of collisional excitation . as for the @xmath48 ratio , it can only be increased by collisions , while it can either be increased or decreased by fluorescence , depending on the details of the sed and the spatial position considered . as noticed in the introduction , not much has been written on the subject of optically thick nebulae illuminated by stellar radiation , which we have called `` case d '' . there are several specific reasons for this , at first surprising , fact . one of them has already been mentioned earlier in this section : absorption lines in stellar continua are expected to quench continuum pumping . even if the hypothesis is made that enough photons survive , a quantitative analysis must by necessity rely on a detailed model spectrum ; if this ingredient is lacking , even qualitative assessments become difficult to make ( see , e.g. , section [ sec : casec ] ) . only recently have high - resolution stellar libraries and synthesis spectra become available , making possible this kind of study . from this point of view , the forthcoming publication of high - resolution spectral synthesis models will be a giant leap forward . for lines of other ions , fluorescence is a well - known excitation mechanism ( see , e.g. , * ? ? ? * ; * ? ? ? its effect on the helium spectrum has been recently discussed by @xcite . in the case of hydrogen , though , fluorescence only provides a small fraction of the total intensity , and until recently neither theory nor observations were accurate enough to justify an analysis at that level of precision . as an extreme example , consider that at the time of publication of the baker and collaborators series of papers , gaunt factors equal to one were still considered a reasonable approximation ; and in 1959 , errors on gaunt factors as great as 20% were still acceptable ( seaton 1959 ) . a final point is brought about by improvements in technology , and in particular by the increasingly better spatial resolution . since fluorescent contribution is the largest in beam - like observations close to the stars , the availability of spatially - resolved information might actually signify a worsening in accuracy . this is analogous to what happens in population synthesis , where the straightforward modelization of stellar populations is made difficult by the poor sampling associated with small observing elements , in such a way that a probabilistic treatment becomes imperative ( i.e. , * ? ? ? case d is the theoretical scenario describing optically thick nebulae irradiated by stellar continua . in these objects , the balmer spectrum has a fluorescent component superposed to the recombination component : while the ionizing part of the stellar continuum is responsible for the recombination component , the non - ionizing part produces a generally small but non negligible fluorescent component . in the same way as collisional excitation of @xmath73 , which is discussed in various papers cited in this work , this mechanism represents an additive correction to recombination , and as such it must be taken into account in abundance analysis to avoid biased results . case d represents a correction to the more famous case b , but , as case b , it is a limiting scenario which only asymptotically describes the behavior of real nebulae . the label `` case d '' has been given for symmetry with case c , but cases c and d are inherently different from cases a and b in that they are not descriptions of intrinsic gas properties , but rather descriptions of complex systems which depend as much on the intrinsic properties of hydrogen as on the `` boundary conditions '' represented by the stellar pumping . to estimate the fluorescent contribution predicted under case d it is necessary to understand how it depends on the object s physical parameters and how it is affected by the observational conditions . with respect to the first point , our bottom line in this work is that a proper modelization requires inputting high - resolution synthesis models into the photoionization code . even this might be only a rough approximation as doppler shifts and various line - broadening mechanisms might be acting in real nebulae , thus altering the pattern of photons absorption ; but at least it is a physically - based approximation , and a good starting point for further explorations . regarding the influence of observational conditions , we have shown that the relative fluorescent intensity observed depends on the aperture . such dependence is straightforward in the case of a centrally concentrated cluster , but it becomes non trivial as more complicated spatial configurations are considered . 3d models including continuum transfer and photoionized by high - resolution seds might help in these cases , but even these could be insufficient to account for actual complexities , as a substantial fraction of the fluorescent component might come from stars of stellar type later than o. when large apertures are considered , geometrical features bear no influence . a comparison with the other major deviation from pure recombination , collisional excitation , is interesting both to clarify both mechanisms and to devise observational strategies to deal with them , which usually means trying to minimize them . our study suggests that fluorescence might be much more important than collisional excitation in two importanty ways : on one hand , it may contribute a larger fraction of the intensity ; on the other hand , it is predicted to affect virtually any h ii region , whereas collisional excitation is effective only in hot , young , metal poor , radiation - bounded h ii regions . one important difference between the two is that while collisional excitation takes place in the outer parts of h ii regions , fluorescent excitation takes place in the vicinity of stars : the observer willing to minimize its contribution on the total intensity is thus placed before a dilemma , as the former mechanism is minimized by spatially - resolved observations , while the latter is minimized by observations of the complete nebula . finally , collisional excitations always produce a reddening in the spectrum , whereas fluorescence might produce both a bluening and a reddening at different points in the nebula . both mechanisms must be taken into account when optical spectra are used to study the dust distribution of observed objects , as they can introduce small biases in spatially - resolved extinction measurements . since this is an exploratory paper , we do not attempt to establish here an exhaustive calibration of fluorescent excitation of balmer lines across the whole h ii regions taxonomy ; rather , we aim at pointing out a problem and the tools necessary to solve it . fluorescent excitation adds a small but non - negligible contribution to the observed balmer spectra of h ii regions and , as such , it should be studied in detail and thoroughly understood since neglecting it might introduce biases in abundance determinations . such biases might be particularly significant when determining the abundance of primordial helium , where a high accuracy is required . for such application , it is also necessary to take into account the effect of continuum fluorescence on the helium emission lines . this work has been funded with support from the spanish ministerio de educacin y ciencia through grants aya2004 - 63030 and aya2007 - 64712 , co - financed with feder funds . gjf acknowledges support from nsf ( ast 0607028 ) and nasa ( 07-atfp07 - 0124 ) . ssd acknowledges support by the spanish mec / fullbright postdoctoral fellowship program . the title was inspired by l. lucy s paper . the plots have been made with the idl package _ jmaplot _ by jess maz . fruitful discussions have been conducted with jorge garca rojas , antonio peimbert , manuel peimbert , enrique prez , grazyna stasinska , and ryszard szczerba . vl acknowledges the instituto de astronoma - unam for hospitality . version c08 of cloudy has the ability to resolve atoms and ions of the h - like iso - electronic sequence . for most of its history the code has used the compact model of the h - atom originally described by @xcite , @xcite and updated by @xcite . that model assumed a well @xmath51-mixed @xmath74-configuration as pioneered by @xcite . under this assumption the populations of @xmath51-terms within an @xmath74 configuration are proportional to the statistical weight 2(@xmath51 + 1 ) . term - averaged transition probabilities , such as those given by nist , are used . with the advent of faster processors it is now possible to resolve the @xmath75 terms within an @xmath74 configuration . the number of states that must be considered is substantially larger . for instance there are 210 terms within the first 20 levels of h@xmath4 , so on most processors the linear algebra will be roughly 2 dex slower . in the current implementation it is possible to adjust the number of @xmath74-configurations which are resulted into @xmath51-terms . these are then supplemented with several additional @xmath51-mixed @xmath74 configurations to `` top off '' the model atom . only lines produced within the @xmath75 resolved levels are reported . the predictions of the resolved and @xmath51-mixed models can be different at low densities where the real @xmath51-terms are not mixed by collisions . continuum fluorescence , the topic of this paper , has different effects in these two approximations , as stressed by ferland ( 1999 ) . the reason is that @xmath51-averaged transition probabilities between excited @xmath74-configurations and the ground @xmath74=1 configuration are dramatically different because only @xmath76 transitions occur . at low densities the rate of decay from an excited @xmath74 to ground will be overestimated if the @xmath51-mixed transition probabilities are used . another complication is that the relationships between the einstein @xmath77 and @xmath78 coefficients are not correct if the total statistical weight of the upper configuration is used . all of these problems are solved with the @xmath51-resolved model now used in cloudy . the current version of cloudy predicts an h i spectrum in excellent agreement with @xcite when the number of resolved levels is made large enough . predictions of the fluorescent excitation spectrum can have even greater differences . table [ tab : appendix ] gives a comparison of the fluorescent - excited spectrum for an optically thin cell of pure hydrogen gas with unit density . it is exposed to a 50,000@xmath79 blackbody with an ionization parameter of 0.01 . the emission would be close to case a if there were no external continuum fluorescing the lyman lines . it is in the case c limit when continuum excitation is included . the first column gives the wavelength of a few lyman , balmer , and paschen lines . the second column gives the @xcite case a intensities relative to their predicted h@xmath9 . the last two columns give the case c predictions for the @xmath51-resolved ( current cloudy ) and @xmath51-mixed ( versions before c08 ) cases . l|[email protected]@[email protected] line & & & + 1216 & 33&1 & 146 & & 165 & + 6563 & 2&82 & 4&02 & 8&02 + 4861 & 1 & & 1&45 & 2&35 + 4340 & 0&473 & 0&687 & 1&00 + 4102 & 0&264 & 0&384 & 0&523 + 1.875 m & 0&455 & 0&496 & 0&807 + 1.282 m & 0&215 & 0&235 & 0&347 + baker , j. g. & menzel , d. h. 1938 , , 86 , 70 baldwin , j. a. , ferland , g. j. , martin , p. g. , corbin , m. r. , cota , s. a. , peterson , b. m. , & slettebak , a. 1991 , , 374 , 580 benjamin , r. a. , skillman , e. d. , & smits , d. p. 2002 , , 569 , 288 burgess , a. 1958 , , 118 , 477 cannon , j. m. , skillman , e. d. , garnett , d. r. , & dufour , r. j. 2002 , , 565 , 931 capriotti , e. r. 1964 , , 225 cervio , m. , & luridiana , v. 2006 , , 451 , 475 cervi~ no , m. , g ' omez - flechoso , m.a . , castander , f.j . , schaerer , d. , moll ' a , m. , kn " odlseder , j. , & luridiana , v. 2001 , a&a , 376 , 422 chamberlain , j. w. 1953 , , 117 , 399 charbonnel , c. , meynet , g. , maeder , a. , schaeller , g. , & schaerer , d. 1993 , a&as , 101 , 415 cota , s. a. , & ferland , g. j. 1988 , , 326 , 889 ferguson , j. w. , & ferland , g. j. 1997 , , 479 , 363 ferland , g. j. 1980 , , 92 , 596 ferland , g. j. 1992 , , 389 , l63 ferland , g. j. 1999 , , 111 , 1524 ferland , g. j. , & rees , m. j. 1988 , , 332 , 141 ferland , g. j. , korista , k. t. , verner , d. a. , ferguson , j. w. , kingdon , j. b. , & verner , e. m. 1998 , pasp , 110 , 761 gabler , r. , gabler , a. , kudritzki , r. p. , puls , j. , & pauldrach , a. 1989 , , 226 , 162 gonzlez delgado , r. m. , cervio , m. , martins , l. p. , leitherer , c. , & hauschildt , p. h. 2005 , , 357 , 945 gonzlez delgado , r. m. , leitherer , c. , & heckman , t. 1997 , , 489 , 601 hillier , d. j. & miller , d. l. 1998 , , 496 , 407 hubeny , i. , lanz , t. , & jeffery , c.s . 1995 , synspec - a users guide izotov , y. i. , schaerer , d. , blecha , a. , royer , f. , guseva , n. g. , & north , p. 2006 , , 459 , 71 izotov , y. i. , thuan , t. x. , & stasiska , g. 2007 , , 662 , 15 kehrig , c. , vlchez , j. m. , snchez , s. f. , telles , e. , prez - montero , e. , & martn - gordn , d. 2008 , , 477 , 813 kurucz , r.l . 1993 , kurucz cd - rom 13 , atlas9 stellar atmosphere programs and 2 km / s grid ( cambridge : sao ) lanz , t. , & hubeny , i. 2003 , apj , 146 , 417 leitherer , c. , et al . 1999 , , 123 , 3 lucy , l. b. 1995 , , 294 , 555 luridiana , v. , & peimbert , m. 2001 , , 553 , 633 luridiana , v. , peimbert , a. , peimbert , m. , & cervio , m. 2003 , , 592 , 846 luridiana , v. 2008 , ( in preparation ) macfarlane , j. j. , waldron , w. l. , corcoran , m. f. , wolff , m. j. , wang , p. , & cassinelli , j. p. 1993 , , 419 , 813 maz - apellniz , j. , mas - hesse , j. m. , muoz - tun , c. , vlchez , j. m. , & castaeda , h. o. 1998 , , 329 , 409 martins , l. p. , gonzlez delgado , r. m. , leitherer , c. , cervio , m. , & hauschildt , p. 2005 , , 358 , 49 martins , f. , schaerer , d. , & hillier , d. j. 2005 , , 436 , 1049 menzel , d. h. 1937 , , 85 , 330 menzel , d. h. & baker , j. g. 1937 , , 86 , 70 najarro , f. , kudritzki , r. p. , cassinelli , j. p. , stahl , o. , & hillier , d. j. 1996 , , 306 , 892 osterbrock , d. e. , & ferland , g. j. 2006 , astrophysics of gaseous nebulae and active galactic nuclei , 2nd . ed . by d.e . osterbrock and g.j . ferland . sausalito , ca : university science books , 2006 pauldrach , a. w. a. , hoffmann , t. l. , & lennon , m. 2001 , , 375 , 161 peimbert , m. , luridiana , v. , & peimbert , a. , 2007 , , 666 , 636 pellegrini , e. w. , et al . 2007 , , 658 , 1119 pellerin , a. , et al . 2002 , , 143 , 159 pquignot , d. , petitjean , p. , & boisson , c. 1991 , , 251 , 680 porter , r. l. , ferland , g. j. , van hoof , p. a. m. , and williams , r. j. r. , in preparation pottasch , s. r. 1960 , , 131 , 202 relao , m. , beckman , j. e. , zurita , a. , rozas , m. , & giammanco , c. 2005 , , 431 , 235 rozas , m. , sabalisck , n. , beckman , j. e. , & knapen , j. h. 1998 , , 338 , 15 salpeter , e.e . 1955 , apj , 121 , 161 schaller , g. , schaerer , d. , meynet , g. , & maeder , a. 1992 , a&as , 96 , 269 schaerer , d. , charbonnel , c. , meynet , g. , maeder , a. , & schaller , g. 1993a , a&as , 102 , 339 schaerer , d. , meynet , g. , maeder , a. , & schaller , g. 1993b , a&as , 98 , 523 seaton , m. j. 1959 , , 119 , 81 seaton , m. j. 1968 , , 139 , 129 stasiska , g. , & izotov , y. i. 2001 , , 378 , 817 storey , p. j. , & hummer , d. g. 1995 , , 272 , 41 ( on the web at http://adc.gsfc.nasa.gov/adc-cgi/cat.pl?/catalogs/6/6064/ ) walborn , n. r. , fullerton , a. w. , crowther , p. a. , bianchi , l. , hutchings , j. b. , pellerin , a. , sonneborn , g. , & willis , a. j. 2002 , , 141 , 443

non - ionizing stellar continua are a potential source of photons for continuum pumping in the hydrogen lyman transitions . in the environments where these transitions are optically thick , deexcitation occurs through higher series lines . as a result , the emitted flux in the affected lines has a fluorescent contribution in addition to the usual recombination one ; in particular , balmer emissivities are systematically enhanced above case b predictions . the effectiveness of such mechanism in h ii regions and the adequacy of photoionization models as a tool to study it are the two main focuses of this work . we find that photoionization models of h ii regions illuminated by low - resolution ( @xmath0 ) population synthesis models significantly overpredict the fluorescent contribution to the balmer lines ; the bias has typical values of the order of a few hundredths of a dex , with the exact figure depending on the parameters of the specific model and the simulated aperture . conversely , photoionization models in which the non - ionizing part of the continuum is omitted or is not transferred significantly underpredict the fluorescent contribution to the balmer lines , producing a bias of similar amplitude in the opposite direction . realistic estimations of the actual fluorescent fraction of the balmer intensity require photoionization models in which the relevant portion of the stellar continuum is adequately represented , i.e. its resolution is high in the region of the lyman lines . in this paper , we carry out such estimation and discuss the variations to be expected as the simulated observational setup and the stellar population s parameters are varied . in all the cases explored , we find that fluorescent excitation provides a significant contribution to the total balmer emissivity . we also show that differential fluorescent enhancement may produce line - of - sight differences in the balmer decrement , mimicking interstellar extinction . fluorescent excitation emerges from our study as a small but important mechanism for the enhancement of balmer lines . as such , we recommend to take it into account in the abundance analysis of photoionized regions , particularly in the case of high - precision applications such as the determination of primordial helium .

we report on a search for the decay @xmath3 ( @xmath4 ) . this is a lepton - family - number - violating ( lfnv ) decay and , thus , is strictly forbidden in the standard model ( sm ) with massless neutrinos . incorporating massive neutrinos into the sm , which is required by the growing evidence for the lepton family transformations in the neutrino sector@xcite , results in a prediction of lfnv kaon decays at an unobservably low level@xcite . thus an observation of a lfnv process like the decay @xmath4 would serve as a clear indication of physics beyond the sm . moreover , extensions of the sm such as extended gauge theories@xcite , technicolor@xcite and supersymmetry@xcite do allow lfnv processes . a first search for @xmath4 , performed during the 1970s , resulted in an upper limit on the decay s branching ratio ( b ) of @xmath5 ( 90% c.l.)@xcite . a search for this decay in bnl was initiated by experiment 777 , in which the upper limit @xmath6 was established@xcite . the data collected in 1995@xcite and 1996@xcite by bnl experiment 865 , a successor to e777 , allows us to lower the combined upper limit to @xmath7 ( 90% c.l.)@xcite . nevertheless a null search result is usefull since it can either put constraints on the parameters of the existing extension models that allow lfnv or rule them out . for example , in the extended technicolor model ( etc)@xcite the transition between the leptons of different generations can be mediated by a horizontal etc boson ( @xmath8 ) . the corresponding diagram , shown in fig . 1 is similar to the diagram of the familiar decay @xmath9 . using the assumption that @xmath8 and @xmath10 boson couplings are approximately equal to each other , one can relate the @xmath8 boson mass , @xmath11 , to the @xmath4 branching ratio@xcite : @xmath12^{1/4}\ ] ] the quoted limit @xmath13 corresponds to a 65 tev lower limit on the @xmath11 mass scale . a competitive constraint on the new physics mass scale may be obtained from another lfnv kaon decay , @xmath14 ( @xmath15 ) , for which an upper limit on its branching ratio is set at @xmath16@xcite . however , the fact that the decay @xmath15 is sensitive to the axial - vector and pseudoscalar ( @xmath17 ) quark transitions , while @xmath4 is sensitive to vector , scalar , and tensor transitions , makes these two processes complementary in the search for new phenomena . this illustrates the significance of the rare decay searches which can be performed at an easily attainable low energy but at the same time can effectively probe the high mass region . the analysis outlined here is based on data recorded at the brookhaven alternating gradient synchrotron ( ags ) during a five month run in 1998 , employing the e865 detector . to perform a successful search for a rare decay , like the @xmath4 , two basic conditions must be satisfied : first , the parent particle ( @xmath18 ) must be produced in copious amounts and second , the detector must be able both to detect signal and to suppress backgrounds to a level low enough for a signal to be seen . the first condition was met by employing the high intensity proton source ( ags ) to create the @xmath18 beam . with the projected @xmath4 single event sensitivity of @xmath19 , in order to meet the second condition we designed the e865 detector to have excellent event reconstruction and particle identification ( pid ) capabilities . the particular requirements for the detector s event reconstruction and pid performance were derived from the careful consideration of the possible backgrounds . the dominant way to get a @xmath20 , @xmath21 and @xmath22 from a kaon decay is through the decay chain @xmath23 ( @xmath24 ) , @xmath25 , @xmath26 , which has a combined branching ratio of @xmath27 . the fact that the in - flight pion decays have to happen within the detector and the neutrinos carry off momentum , distorting an event s kinematics , led to the design of a high - resolution momentum analyzing spectrometer which allowed charged particle reconstruction and momentum measurement . the study of the monte carlo simulated events and off - line event reconstruction ( to be discussed later ) showed that such backgrounds could be suppressed to the level of @xmath28 by applying simple kinematic cuts . however , the @xmath24 decay would also mimic the @xmath4 if a @xmath29 was misidentified as @xmath22 and @xmath20 was either misidentified as @xmath21 or underwent an in - flight decay ( @xmath25 ) . another background was caused by rather common kaon decays : @xmath30 ( @xmath31 ) , with a @xmath20 misidentified as @xmath21 , and @xmath32 ( @xmath33 ) . both decays would mimic @xmath4 if the @xmath34 underwent a dalitz decay ( @xmath35 ) and the @xmath36 was misidentified as a @xmath20 . in the following we will refer to both decays ( @xmath31 and @xmath33 followed by a dalitz decay of @xmath34 ) as @xmath37 . the suppression of the @xmath24 and @xmath37 backgrounds required a pid system with abilities to veto @xmath29 ( while still being efficient in identifying @xmath22 ) , veto @xmath36 and to discriminate @xmath21 from @xmath20 . a detailed description of the apparatus ( see fig . [ detector ] ) and the beam arrangement is presented elsewhere@xcite . here we give only a brief overview of the kaon beam and detector elements . the ags at brookhaven national laboratory served as a source of the primary 25.5 gev / c proton beam , and delivered about 1@xmath38 @xmath39 protons to the kaon production target in a spill of 2.8 s with a cycle time of 5.1 s. before the extraction from the ags , the proton beam was debunched , _ i.e. _ particles were distributed over the circumference of the ring to obtain a uniform beam intensity during the spill . the extracted beam of protons , hitting a 15 cm long water - cooled copper target with transverse dimensions 5 @xmath38 5 @xmath40 , produced a large variety of particles , but mainly pions , protons and kaons survive . after the production target , the unseparated secondary beam was momentum selected at 6 gev / c ( @xmath41 ) and was transported to the beginning of the e865 detector through the specially designed 28 m long a2 beam line@xcite . a major concern during the design of the a2 beam line was the muon halo , produced by kaon and pion decays . it was greatly suppressed by tight collimation and two foci followed by dipole magnets , which effectively swept momentum - degraded muons out of the acceptance@xcite . during data taking , the beam flux in the a2 beam line was estimated to be 2 @xmath38 @xmath42 @xmath18 , 4 @xmath38 @xmath43 @xmath20 , and 2 @xmath38 @xmath43 protons per spill . the e865 detector was located at the end of the a2 beam line , and its first element was the 5-m long evacuated decay region within which about 6@xmath44 of the entering kaons downstream of the decay region , a dipole magnet ( d5 ) swept the charged decay products away from the beam , with negative particles going mostly to the left , and positive to the right side of the apparatus . the momentum - analyzing spectrometer consisted of proportional wire chambers ( pwc ) , p1-p4 and a second dipole magnet ( d6 ) , which steered the particles back into the acceptance region of the detector elements located further downstream the pwcs , each consisting of four wire planes , were desensitized in the region where the beam passed . this arrangement yielded a momentum resolution of @xmath45 @xmath46 gev / c , where @xmath47 , the momentum of the decay products in gev / c , was typically in the range 0.6 to 3.5 gev / c . the first part of the pid system consisted of two large atmospheric - pressure erenkov counters ( c1 and c2 ) , located upstream and downstream of the p3 wire chamber . each counter was separated by a thin membrane into two ( left and right ) parts providing independent particle identification for positive and negative tracks and allowing use of different gases . for the purpose of reducing the possibility of misidentifying a @xmath29 as @xmath22 the left sides ( c1l , c2l ) were filled with a high erenkov threshold gas , hydrogen ( @xmath48 ) , and had a light yield of about 2.3 photoelectrons ( p.e . ) for @xmath22 . to effectively register and veto @xmath36 , the right sides ( c1r , c2r ) were filled with a low - threshold gas , methane ( @xmath49 ) , and had a light yield of about 5.8 p.e . for @xmath36 . in order to reduce beam erenkov radiation in c1r and c2r , closed tubes filled with hydrogen gas were placed in the beam region . the next pid detector element was an electromagnetic calorimeter of the shashlyk design@xcite . it incorporated 582 modules , 11.4 cm by 11.4 cm by 15 radiation lengths each , assembled in a 30@xmath3820 array with 18 modules in the middle removed for beam passage . the approximate energy resolution for electrons was 8@xmath44 /@xmath50 . typical energy deposition of a minimum ionizing particle was 250 mev . the last pid detector element was a muon detection system , that was located downstream of the calorimeter and consisted of 24 planes of proportional tubes inter - spaced by iron plates . the plate thickness was 5 cm between the first eight pairs of planes and 10 cm between the last four . finally , four arrays of scintillator counter hodoscopes , a , b , c and d , were used for triggering purposes . the hardware trigger , described in detail in@xcite , was designed as a four - level structure with increasing sophistication and response time at each level . the lowest trigger level ( t0 ) selected events with three charged - particle tracks , one on the left and two on the right side , by requiring at least three coincidences between an a - counter slat and the calorimeter module behind it ( a@xmath51sh ) . by using a programmable matrix lookup unit ( mlu ) , for each combination of coincidences between individual counters on the right , only a limited , kinematically acceptable , region on the left was allowed . the mlu was programmed to maximize the acceptance of the three charged body kaon decays by using simulated @xmath4 events . in order to reduce contamination from decays occurring downstream of the decay region , at least one coincidence on both left and right sides between the d - counter and a@xmath51sh was required as well . the t0 signal was ready in about 175 ns after the particle hit the a - counter and had a rate of about 2.5 khz . the next trigger level ( t1 ) used information from the erenkov counters and the muon range telescope for purpose of pid . in the case of @xmath4 the dedicated trigger ( mue ) demanded the presence of an electron and a muon . consequently , the mue trigger required signals corresponding to at least 0.25 p.e . in both erenkov counters on the left side ( c1l , c2l ) to select @xmath22 and a hit in the muon range telescope ( b - hodoscope ) to be spatially consistent with the a@xmath51sh hit in order to select @xmath21 . the t1 signal was available about 130 ns after the t0 , and had a rate of about 1.3 khz . the next trigger level ( t2 ) was designed to discriminate between events with high and low @xmath52 invariant mass . information from coincident a@xmath51sh hits , and erenkov counters was used to select events consistent with low vertical separation of @xmath53 tracks ( low invariant mass ) . that trigger was not utilized in the @xmath4 search . the final trigger level ( t3 ) compared the number of hits in each of the pwc plane with a predetermined likelihood distribution obtained offline . this trigger rejected a small fraction of events that had either too few or too many pwc hits . during data taking , the average rate of t3 was about 1800 per 2.8 s spill or about 0.6 khz . in addition to the mue trigger , several prescaled monitor triggers were recorded , e.g. a minimum bias trigger ( tau ) , which required only t0 signal and was dominated by the @xmath24 events , and two triggers sensitive to the @xmath37 events . the first one ( eeps ) demanded the presence of an @xmath22 and @xmath36 by requiring signals ( at least 0.25 p.e . ) in both erenkov counters on the left and right sides ( c1l , c2l , c1r , c2r ) and was used for pid studies and background estimates . the second @xmath37 sensitive trigger ( cerenk ) was designed to obtain an unbiased response of the erenkov counters and required only three out of four sides of the erenkov counters ( c1l , c2l , c1r , c2r ) to have signals above 0.25 p.e . during a five month data taking period in 1998 we recorded a total of 1.3@xmath54 triggers on tape , of which 0.3 @xmath54 were mue triggers , successfully reconstructed in the offline analysis . the process of kinematic reconstruction of an event followed the t0 hardware trigger requirements ( a@xmath51sh ) and started with the `` clump '' finding , _ i.e. _ determining positions of tracks in the calorimeter based on their energy depositions and correlated a - counter hits . for each `` clump '' found in the calorimeter , a window in the pwc p4 was defined and space points , requiring at least three of the four wire planes in the chamber , were searched for . likewise , the space points found in p4 determined windows in p3 and so forth . a track was formed if at least three pwcs contributed with a space point . next , using the measured magnetic field map inside the dipole magnets , the momentum of each track was fit . for events containing at least three reconstructed tracks , a fitting algorithm was used to determine the decay vertex position by minimizing the quantity @xmath55 , the r.m.s . of the closest approach of the three charged tracks to the common vertex point . for events containing more than three tracks the combination with the smallest @xmath55 was tagged as the most probable set of track candidates from a kaon decay . studies of fully reconstructed @xmath24 events showed that the mean value of @xmath55 increased for vertices found further upstream ( lower @xmath56 ) consistent with monte carlo simulations . the dependence could be fitted with a second order polynomial ( denoted @xmath57 ) . the normalized quantity @xmath58 , which is independent of @xmath56 , was then used to judge the vertex quality in the @xmath4 analysis . as the last step , the information from the pid detectors was assigned to each found track . basic requirements for @xmath4 included the presence of a three - charged - track vertex within the decay region of an acceptable quality , @xmath59 . in addition , in order to ensure that reconstructed tracks came from a real kaon decay with all the particles being detected , the reconstructed total momentum vector was tracked back to the production target using the first - order beam - line transfer matrix , and its phase space ( the kaon momentum @xmath60 , the position @xmath61 , @xmath62 , and the direction , @xmath63 , @xmath64 ) was compared to that of the kaon beam . the five - dimensional phase space of the beam was broken up into three two dimensional distributions : @xmath61 versus @xmath63 , @xmath62 versus @xmath64 and @xmath60 vs @xmath61 . the properties of the kaon beam were determined from fully reconstructed @xmath24 events and were used as input for a corresponding logarithmic likelihood function @xmath65 ( see figures [ fig : tgtlik ] and [ final_dist ] , and also sec . [ sec : likelihood ] ) . the invariant mass @xmath66 , calculated using the daughter particles masses ( @xmath67 @xmath68 , @xmath69 ) and their measured momenta , was required to be consistent with the kaon mass , 493.67 mev , within the resolution , @xmath70 mev for simulated @xmath4 events . finally we required an unambiguous identification of @xmath22 , @xmath20 and @xmath21 . electron pid required a signal ( at least 0.3 p.e . ) in the left sides of both erenkov counters ( c1l , c2l ) with corrected timing within @xmath715 ns . in addition , the energy loss in the calorimeter ( @xmath72 ) was required to be consistent with the momentum of the track ( @xmath47 ) , @xmath73 the ratio @xmath74 to be between 0.65 and 1.5 . pions were identified requiring the absence of a signal above 1.2 p.e . on the right sides of the erenkov counters ( c2r , c2r ) and corresponding corrected timing of more than @xmath715 ns . in addition , energy loss in the calorimeter was required to be consistent with that of a minimum ionizing particle or a hadron shower ( @xmath75 ) . muons were identified by requiring the absence of a signal above 1.2 p.e . in both erenkov counters ( c1,c2 ) , energy deposited in the calorimeter consistent with minimum ionization ( @xmath76450 mev ) and penetration depth in the muon range consistent with its momentum . table [ tab : pid_pass3_eff ] summarizes the pid efficiencies and probabilities of misidentification achieved by the above mentioned selection cuts@xcite . although the selection criteria described above provided a high sensitivity for the signal ( respective efficiency of each cut was approximately 95% ) they let a considerable number of background events pass . .identification efficiencies and probability of misidentification for the primary selection pid cuts . the symbol `` @xmath77 '' stands for identified as . [ cols="<,^,^,^",options="header " , ] [ tab : corr_fact ] the correction factor @xmath78 in eq . ( [ eq : upper_limit_mc ] ) contains efficiencies to account for the difference in reconstructing @xmath24 and @xmath4 events . as described in section [ sec : tau_total ] , most of the kinematic cuts for the @xmath24 sample selection were also applied to select @xmath4 events and the corresponding efficiencies were the same for both decay modes . the correction factor was calculated , as a reciprocal of the the product of the selection cut efficiencies specific to the @xmath4 ( see table [ tab : corr_fact ] ) , to be @xmath79 . the bulk of @xmath78 , 2.17 , is the reciprocal of the product of the @xmath20 , @xmath21 and @xmath22 pid efficiencies , while the remaining factor of 1.53 results from the specific @xmath4 kinematic cuts . with the @xmath80@xcite , and all the other factors in eq . ( [ eq : upper_limit_mc ] ) determined , the upper limit on the @xmath4 branching ratio is set at : @xmath81 the details of the analysis of the final data set obtained by e865 are described in this paper . here , we summarize the search for the decay @xmath82 at bnl , which included a series of four measurements in which the following 90% c.l . upper limits were obtained : @xmath83@xcite , @xmath84@xcite , @xmath85@xcite , and @xmath86 ( this paper ) . since the last two measurements were not background free , we analyzed the combined likelihood function constructed as the product of the likelihood functions for the separate measurements to establish a combined upper limit . for the final data set , this function is described above in eq . [ eq : poisson_lik1 ] . the analogous function , based on the distributions shown in fig . 4 of ref . @xcite was used for the 1996 data . the only common parameter was a @xmath87 branching ratio @xmath88 , introduced into the likelihood function via @xmath89 where @xmath90 was the effective ( corrected for @xmath4 acceptance ) number of kaon decays in a given experimental run . for the background and signal free measurements of refs . @xcite ( e777 ) and @xcite , the likelihood function ( eq . [ eq : poisson_lik1 ] ) is reduced to the simple form @xmath91 . although no evidence of new physics was found , the parameter space of the existing extension theories that allow lfnv was reduced . particularly , as discussed in sec . [ sec : introduction ] , the mass , @xmath11 , of the corresponding extended technicolor boson , with strength equal to that of the weak interaction , is bound by the limit obtained on the @xmath4 branching ratio to be greater then @xmath93 tev . we gratefully acknowledge the contributions to the success of this experiment by dave phillips , the staff and management of the ags at the brookhaven national laboratory , and the technical staffs of the participating institutions . a. b. mcdonald _ et al . _ [ sno collaboration ] , aip conf . * 646 * , 43 ( 2003 ) . p. langacker , s. u. sankar and k. schilcher , phys . d * 38 * , 2841 ( 1988 ) . r. n. cahn and h. harari , nucl . b * 176 * , 135 ( 1980 ) .

based on results of a search for the lepton - family - number - violating decay @xmath0 with data collected by experiment e865 at the alternating gradient synchrotron of brookhaven national laboratory ( bnl ) , we place an upper limit on the branching ratio at @xmath1 ( 90% c.l . ) . combining the results with earlier e865 data and those of a previous experiment , e777 , an upper limit on the branching ratio of @xmath2 ( 90% c.l . ) is obtained . 3k^+ ^0 e^+ 3gk^+^0 e^+ 3k^+^0^+ 2k^+^0^+ 4k^+^+^-e^+ 3pik^+^+ ^0 ^0 2m_^2

a _ delaunay set _ @xmath0 has positive numbers @xmath1 such that every open ball of radius @xmath2 contains at most one point , and every closed ball of radius @xmath3 contains at least one point of @xmath4 . such sets were introduced as @xmath5-systems by boris n. delaunay in 1924 . by a _ triangulation _ of @xmath4 , we mean a simplicial complex , @xmath6 , whose vertex set is @xmath4 and whose underlying space is @xmath7 . this triangulation is _ uniformly bounded _ if there is a real number @xmath8 such that the circumsphere of every @xmath9-simplex in @xmath6 has radius smaller than or equal to @xmath10 . a particular triangulation is the _ delaunay triangulation _ , denoted as @xmath11 , whose @xmath9-simplices satisfy the additional condition that all other vertices lie outside their circumspheres . it exists if @xmath4 is generic , as will be explained shortly . the delaunay triangulation of a delaunay set is necessarily uniformly bounded . we also consider delaunay triangulations of _ finite _ sets of points , for which the underlying space of the simplicial complex is the convex hull of the points . writing @xmath12 for the set of all @xmath9-simplices in @xmath7 , we consider functionals @xmath13 for which there are constants @xmath14 and @xmath15 such that @xmath16 for every @xmath9-simplex @xmath17 whose edges are longer than or equal to @xmath18 and whose circumsphere has a radius smaller than or equal to @xmath10 . writing @xmath19 for this class of functionals , we define subclasses @xmath20 by requiring additional conditions . briefly , @xmath21 belongs to @xmath22 if the sum of values over the @xmath9-simplices of the delaunay triangulation of @xmath23 points in @xmath7 is smaller than or equal to the sum over the @xmath9-simplices in the other triangulation , and @xmath21 belongs to @xmath24 is it satisfies a similar condition for all finite sets of points . for a triangulation , we define the _ density _ of the functional by taking sums and lower limits over a growing sequence of balls : @xmath25 where @xmath26 is the closed ball with radius @xmath27 and center at the origin of @xmath7 , and @xmath28 is its volume . with these definitions , we can give our main result : * in @xmath7 , @xmath29 implies that the delaunay triangulation minimizes the density of @xmath21 among all uniformly bounded triangulations of a delaunay set , and in @xmath30 , @xmath31 suffices to reach the same conclusion . there are many concrete functionals studied in the literature to which our result applies . here , we just mention two : * the functional that maps every triangle in @xmath30 to the radius of its circumcircle ; see @xcite , * the functional that maps every @xmath9-simplex to the sum of squares of its edge lengths times the volume ; see @xcite . the remainder of this paper presents the detailed results in two sections . in this section , we introduce the background on delaunay sets , their uniformly bounded triangulations , and functionals on such triangulations . we recall from section [ sec1 ] that @xmath0 is a _ delaunay set _ if there are positive constants @xmath1 such that ( i ) every open ball of radius @xmath2 contains at most one point of @xmath4 , and ( ii ) every closed ball of radius @xmath3 contains at least one point of @xmath4 . hence , @xmath4 has no tight cluster and leaves no large hole . [ [ counting - points . ] ] counting points . + + + + + + + + + + + + + + + + condition ( i ) implies that every bounded subset of @xmath7 contains only finitely many points of @xmath4 . indeed , the subset can be covered by finitely many open balls of radius @xmath2 , and each such ball contains at most one point . condition ( ii ) implies that every cone with non - zero volume contains infinitely many points of @xmath4 . indeed , the cone contains an infinite string of disjoint closed balls of radius @xmath3 , and each such ball contains at least one point of @xmath4 . we quantify the first observation by giving concrete estimates . let @xmath32 be the closed ball with radius @xmath27 and center @xmath33 , and call the difference between two concentric balls an _ annulus_. let @xmath4 be a delaunay sets with parameters @xmath1 in @xmath7 . 1 . there are constants @xmath34 and @xmath35 such that the number of points of @xmath4 in @xmath32 is between @xmath36 and @xmath37 . 2 . there is a constant @xmath38 such that the number of points of @xmath4 in @xmath39 is at most @xmath40 . proof . to prove ( i ) , we note that @xmath32 can be covered by some constant times @xmath41 balls of radius @xmath2 , and that we can pack some other constant times @xmath42 balls of radius @xmath3 in it . the lower and upper bounds follow . to prove ( ii ) , we cover the annulus with a constant times @xmath43 balls of radius @xmath2 . the upper bound follows . it should be clear that the bound in ( ii ) also holds for annuli of constant width , but not for annuli whose width is a positive fraction of the radius . [ [ delaunay - triangulations . ] ] delaunay triangulations . + + + + + + + + + + + + + + + + + + + + + + + + following the original idea of boris n. delaunay , we consider @xmath9-simplices with vertices from @xmath4 such that the open ball bounded by the @xmath44-dimensional circumsphere contains no points of @xmath4 . we call such @xmath9-simplices _ empty_. here , it is convenient to assume that @xmath4 is _ generic _ in the sense that no @xmath23 points in @xmath4 lie on a common @xmath44-sphere . under this assumption , the empty @xmath9-simplices fit together without gap and overlap . now consider two not necessarily empty but non - overlapping @xmath9-simplices that share a @xmath44-simplex , which is a face of both . assuming the two @xmath9-simplices belong to a triangulation , we call this face _ locally delaunay _ if the @xmath45-st vertex of the second @xmath9-simplex lies outside the circumsphere of the first @xmath9-simplex . note that the condition is symmetric because the two circumspheres intersect in the @xmath46-sphere that passes through the vertices of the face , and either both @xmath45-st vertices lie outside or both lie inside the respective other circumsphere . delaunay considered both conditions and proved that they are equivalent @xcite . let @xmath4 be a generic delaunay set in @xmath7 . 1 . the collection of empty @xmath9-simplices together with their faces form a triangulation of @xmath4 , commonly known as the _ delaunay triangulation _ , @xmath11 . if all @xmath44-simplices of a triangulation @xmath6 of @xmath4 are locally delaunay , then @xmath47 . the equivalence between the local and the global conditions expressed in ( ii ) also holds for finite sets @xmath4 . in the plane , it means that a triangulation of a generic set @xmath4 is delaunay iff for each edge the sum of opposite angles in the two incident triangles is less than @xmath48 . let @xmath4 be a generic delaunay set in @xmath7 , and let @xmath6 be a triangulation of @xmath4 . we recall that this means that @xmath6 is a simplicial complex with vertex set @xmath4 whose underlying space is @xmath7 . recall also that @xmath6 is _ uniformly bounded _ if there is a real number @xmath8 such that the radius of the circumsphere of every @xmath9-simplex in @xmath6 is smaller than or equal to @xmath10 . it follows that no edge of @xmath6 is longer than @xmath49 . note that the delaunay triangulation of @xmath4 is uniformly bounded with @xmath50 . [ [ not - every - triangulation - is - uniformly - bounded . ] ] not every triangulation is uniformly bounded . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we begin by showing that every delaunay set has triangulations that are not uniformly bounded . given @xmath4 , we construct such a triangulation in three steps . 1 . for every point @xmath51 and every @xmath52 , we can find many points @xmath53 such that the edge @xmath54 is longer than @xmath55 and does not pass through any other points of @xmath4 . indeed , there is such an edge near every direction out of @xmath56 . to see this , we consider the set of points in @xmath4 that lie within the closed ball of radius @xmath55 around @xmath56 . there are only finitely many such points , which implies that within each cone with non - zero volume and apex @xmath56 , we can find a subcone , again with non - zero volume and apex @xmath56 , that does not contain any of the points of @xmath4 inside the ball . however , as argued in section [ sec21 ] , the cone contains infinitely many points of @xmath4 , so they must all be at distance larger than @xmath55 from @xmath56 . among these , let @xmath57 be the point closest to @xmath56 . using the knowledge about long edges , we construct a triangulation inductively , one phase at a time . after the @xmath58-th phase , we will have a triangulation @xmath59 of a finite subset of @xmath4 that includes all points at distance @xmath58 or less from the origin . in addition , we will make sure that @xmath59 contains at least one edge longer than @xmath58 , and that every point of @xmath4 that belongs to the underlying space of @xmath59 is a vertex of @xmath59 . we start with @xmath60 consisting of a single edge connecting the two points of @xmath4 that are closest to the origin of @xmath7 . 1 . in the @xmath61-st phase , we let @xmath56 be a vertex in the boundary of @xmath59 . let @xmath57 be another point of @xmath4 such that the edge @xmath54 is longer than @xmath62 and does not intersect the simplices in @xmath59 other than at @xmath56 . since @xmath59 is a triangulation , its underlying space is convex , and its boundary is triangulated . let @xmath17 be an @xmath63-simplex in the boundary of @xmath59 that is visible from @xmath57 . we extend @xmath59 by adding the @xmath64-simplex formed by @xmath57 and the vertices of the @xmath63-simplex . doing this for @xmath57 and all visible simplices in the boundary of @xmath59 , we obtain a simplicial complex @xmath65 by _ starring _ from @xmath57 . similarly , we add a point @xmath33 at distance @xmath62 or less from the origin that lies outside the underlying space by starring to the triangulation . repeating this operation for all such points @xmath33 , we eventually get a simplicial complex @xmath66 . 2 . while @xmath66 is a valid simplicial complex in @xmath7 , some of the new simplices may contain points of @xmath4 in their interiors . by construction , @xmath54 is not among these simplices . let @xmath67 be such a point , and @xmath68 the simplex of lowest dimension , @xmath69 , that contains @xmath70 in its interior . note that @xmath71 . we fix the situation by decomposing @xmath17 into @xmath72 @xmath69-simplices , each the convex hull of @xmath70 and @xmath69 vertices of @xmath17 . similarly , we decompose each simplex that contains @xmath17 as a face into @xmath72 simplices of the same dimension by starring from @xmath70 . repeating this procedure for all such points @xmath70 , we eventually get a triangulation @xmath73 such that all points in @xmath4 that belong to the underlying space of @xmath73 are in fact vertices of @xmath73 . observe that the edge @xmath54 added to @xmath59 in step 2 remains undivided until the end . this implies that @xmath73 indeed contains an edge longer than @xmath62 , as required . it follows that the triangulation thus constructed by transfinite induction is not uniformly bounded . [ [ measuring - volume . ] ] measuring volume . + + + + + + + + + + + + + + + + + the remainder of this section states and proves properties of uniformly bounded triangulations . we begin with the volume of their simplices . let @xmath4 be a delaunay set with parameters @xmath1 in @xmath7 , and let @xmath6 be a uniformly bounded triangulation with parameter @xmath10 of @xmath4 . 1 . in @xmath30 , there is a positive constant @xmath74 such that @xmath75 for every triangle @xmath17 in @xmath6 . 2 . in @xmath7 , there is a constant @xmath76 such that @xmath77 for every @xmath9-simplex @xmath17 in @xmath6 . proof . we prove ( i ) by expressing the area of a triangle in terms of the three edge lengths and the radius of the circumcircle : @xmath78 . the edges can not be shorter than @xmath18 , and the radius can not be larger than @xmath10 , which implies @xmath79 . to prove ( ii ) , we note that every @xmath9-simplex is contained in the ball bounded by its circumsphere . since the radius is at most @xmath10 , this ball is smaller than @xmath80 . if we remove the requirement of uniform boundedness , then the proof of the volume lemma breaks down . it is not clear whether the upper bound fails . in this context , we mention a related question asked by l. danzer and independently by m. boshernitzen : `` _ is it true that for every planar delaunay set there exists a triangle with arbitrarily large area that contains no points in its interior ? _ '' this question is still open . next , we describe a delaunay set in @xmath81 that has tetrahedra of arbitrarily small volume in the delaunay triangulation . it shows that the limitation of the lower bound in ( i ) to two dimensions is necessary . consider the standard cubic lattice , @xmath82 . let @xmath83 , for every @xmath84 , and move every point @xmath85 to @xmath86 , denoting the new point set by @xmath4 . to study the volume of the tetrahedra in the delaunay triangulation , we consider a single integer cube , for which we get a tetrahedron of volume about @xmath87 in the middle , four tetrahedra of volume about @xmath88 across each face , and two flat tetrahedra at the top and the bottom ; see figure [ fig : cube ] . these volume estimates assume arbitrarily small values of @xmath89 . if the third coordinates of the original vertices are @xmath58 and @xmath62 , then the volume of the top tetrahedron is @xmath90 , and that of the bottom tetrahedron is @xmath91 . since among the @xmath89 there are arbitrarily small numbers , there are tetrahedra in @xmath11 whose volume is arbitrarily close to @xmath92 . it is not difficult to extend this example to four and higher dimensions . [ [ counting - simplices . ] ] counting simplices . + + + + + + + + + + + + + + + + + + + recall the volume lemma , which states that every triangle in a uniformly bounded triangulation of a delaunay set in @xmath30 has an area that exceeds a positive constant . since a disk of radius @xmath27 has area @xmath93 , this implies that the number of triangles contained in this disk is at most some constant times @xmath94 . a similar result holds in three and higher dimensions , but the lack of a lower bound on the volume of a @xmath9-simplex requires a different argument , which we present as the proof of the following bounds . let @xmath4 be a delaunay set with parameters @xmath1 in @xmath7 , and let @xmath6 be a uniformly bounded triangulation with parameter @xmath10 of @xmath4 . 1 . the number of @xmath9-simplices sharing a common vertex is bounded from above by a constant @xmath95 . 2 . there are positive constants @xmath96 and @xmath97 such that the number of simplices contained in a ball of radius @xmath98 is between @xmath99 and @xmath100 . 3 . there is a constant @xmath101 such that the number of @xmath9-simplices contained in a ball of radius @xmath102 but not in the concentric ball of radius @xmath27 is at most @xmath103 . proof . to prove ( i ) , we let @xmath56 be the shared vertex , and we note that all incident @xmath9-simplices are contained in the ball of radius @xmath104 centered at @xmath56 . by the point count lemma , the number of points in this ball is bounded from above by @xmath105 . we have at most one @xmath9-simplex for every combination of @xmath9 of these points , which gives @xmath106 . the upper bound in ( ii ) is now easy : by the point count lemma , the number of points inside the ball of radius @xmath27 is at most @xmath107 . multiplying @xmath108 with @xmath106 gives @xmath109 . to get the lower bound , we restrict ourselves to the ball of radius @xmath110 . by the point count lemma , the number of points in this smaller ball is at least @xmath111 . every @xmath9-simplex incident to one of these points is contained in the ball of radius @xmath27 . each point belongs to at least @xmath112 @xmath9-simplices , which implies that the lower bound on the number of points also applies to the @xmath9-simplices . finally , @xmath113 is at least @xmath114 . to prove ( iii ) , we use the upper bound of @xmath115 on the number of points in the annulus . each @xmath9-simplex we count is incident to at least one of these points . multiplying @xmath116 with @xmath117 gives @xmath118 . recall that @xmath12 denotes the set of simplices , including degenerate ones . we are interested in functionals that have constant upper and lower bounds for the simplices that arise in uniformly bounded triangulations of delaunay sets . for other degenerate simplices we also allow infinity as a value . let @xmath19 be the class of functionals @xmath13 for which there are constants @xmath14 and @xmath15 such that @xmath16 for all @xmath9-simplices @xmath17 with edges of length at least @xmath18 and radius of the circumsphere at most @xmath10 . in this section , we extend the functionals from simplices to triangulations , and we introduce subclasses that favor delaunay triangulations for finite sets of points . [ [ densities . ] ] densities . + + + + + + + + + + as already mentioned in section [ sec1 ] , we define the density of a functional on a triangulation by taking the lower limit over a growing ball , of the sum of values over all @xmath9-simplices in the ball divided by the volume of the ball : @xmath119 there are other possibilities , such as taking the upper limit , or taking the average over the simplices . our results extend to both modifications of the definition . one of delaunay s motivations for defining @xmath120-systems was to generalize lattices in @xmath7 to a larger class of sets . for the delaunay triangulation of any lattice @xmath121 , the limit of the expression in , in which we substitute @xmath122 for @xmath6 , is well defined . unfortunately , this is not generally the case for delaunay triangulations of delaunay sets , which is the reason for taking the lower limit . since this might not be entirely obvious , we will prove shortly that for a broad class of functionals in @xmath19 , the limit does not generally exist . before that , we prove some positive results , namely that the density of every functional is bounded and independent of the choice of origin . specifically , we define @xmath123 for every point @xmath124 , and we prove that all choices of @xmath33 give the same result . let @xmath21 be a functional in @xmath19 . 1 . there is a constant @xmath125 such that @xmath126 for every uniformly bounded triangulation of a delaunay set in @xmath7 . 2 . @xmath127 for every @xmath124 . proof . to prove ( i ) , we recall the simplex count lemma , which implies that the number of @xmath9-simplices contained in @xmath26 is bounded from above by @xmath109 . multiplying with @xmath128 gives @xmath129 . to prove ( ii ) , we let @xmath130 and @xmath131 be the expressions in and without taking the lower limit , so that @xmath132 , and similarly for @xmath133 and @xmath131 . it suffices to prove @xmath134 & = & 0 , \label{eqn : limit1 } \end{aligned}\ ] ] which we do in two steps , namely by proving @xmath135 & = & 0 , \label{eqn : limit2 } \\ \lim_{\alpha \to \infty } [ f(t,\alpha+l)-f_z(t,\alpha ) ] & = & 0 , \label{eqn : limit3 } \end{aligned}\ ] ] where @xmath136 is the distance between @xmath33 and the origin . to prove , we write @xmath137 for the volume of the unit ball in @xmath7 , and we note that @xmath138 . furthermore , we write @xmath139 and @xmath140 for the sums over the @xmath9-simplices contained in the smaller ball and the extra @xmath9-simplices contained in the larger ball : @xmath141 by the simplex count lemma , @xmath139 is at most some positive constant times @xmath142 , while @xmath140 is at most some constant times @xmath143 . hence , @xmath144 the first term in goes to zero because @xmath140 grows slower than @xmath142 , and the second term goes to zero because @xmath145 grows slower than @xmath142 . the argument for is similar . indeed , all we need is to notice that the set of @xmath9-simplices contained in @xmath146 but not contained in @xmath32 is a subset of those contained in @xmath146 but not contained in @xmath147 . by the simplex count lemma , the number of simplices thus defined is bounded from above by a constant times @xmath143 , so that the argument goes through as before . [ [ non - existence - of - limits . ] ] non - existence of limits . + + + + + + + + + + + + + + + + + + + + + + + + we now show that taking the lower limit in the definition of density is necessary because the limit does not generally exist . indeed , functionals for which the limit exists , even just for all delaunay triangulations of delaunay sets , are the exception . this is true in particular for the functionals that are invariant under isometries , which include all examples we discuss in this paper . we begin by exhibiting a construction in @xmath30 that acts as a stepping stone in our argument . let @xmath17 and @xmath148 be two _ compatible _ triangles , by which we mean that they share an edge , the two angles opposite that edge add up to less than @xmath48 , and the remaining four angles are all acute . the condition implies that at least one of the triangles is acute , and we assume @xmath17 is . using a linear sequence of congruent copies of @xmath17 , we form a strip @xmath149 , which we call _ wide _ , and using copies of @xmath148 , we form a strip @xmath150 , which we call _ narrow _ ; see figure [ fig : strips ] . gluing strips together so that they match up at boundary edges , we get a delaunay triangulation , provided no two narrow strips are glued to each other . let @xmath151 be an infinite sequence of odd integers . we construct a delaunay triangulation @xmath152 inductively , starting with a block of @xmath153 wide strips . on each side , we add a block of @xmath154 strips alternating between narrow and wide , then a block of @xmath155 wide strips , then a block of @xmath156 strips again alternating between narrow and wide , and so on . for each @xmath157 , let @xmath158 be the total width of the first @xmath159 blocks . making sure that the origin lies on the center line of the first block , @xmath160 is the largest disk centered at the origin that is still contained in the union of the first @xmath159 blocks . we consider the sequence @xmath161 assuming that the limit in exists , the sequence of @xmath162 must converge for every sequence of @xmath163 . this is indeed the case if we measure area , because the number of triangles that intersect @xmath26 but are not contained in it is bounded from above by a constant times @xmath27 . hence , @xmath164 , where we abbreviate @xmath165 and take the sum over all triangles contained in @xmath26 , as usual . this motivates us to consider the ratios of the terms in the two sequences . assuming @xmath166 contains @xmath167 congruent copies of @xmath17 and @xmath168 congruent copies of @xmath148 , this gives @xmath169 define @xmath170 , @xmath171 , and @xmath172 . assuming @xmath173 , we have @xmath174 and define @xmath175 . by definition , we choose @xmath154 large enough so that @xmath177 , which is possible because @xmath178 goes to @xmath179 as @xmath154 goes to infinity . then we choose @xmath155 large enough so that @xmath180 , and so on , alternating between being close to @xmath181 and @xmath182 . we thus arrive at a contradiction because there is a gap of size @xmath183 between the terms with odd and even indices . in other words , we need @xmath184 for the limit to exist . we finally show that the non - existence of the limit is not an artifact of the particular delaunay set we used in the construction of @xmath152 . let @xmath185 and @xmath186 be arbitrary triangles with longest edges of lengths @xmath187 and @xmath188 and opposite angles @xmath189 and @xmath190 . setting @xmath191 , we construct triangles @xmath17 with edges of length @xmath192 and @xmath148 with edges of length @xmath193 . it is easy to verify that @xmath185 and @xmath17 are compatible , and so are @xmath17 and @xmath148 , and @xmath148 and @xmath186 . repeating the construction with the strips and blocks three times , we see that if @xmath21 is invariant under isometries of @xmath30 and the limit exists for the delaunay triangulations of all delaunay sets , then @xmath194 . conversely , among the functionals invariant under isometries , only the ones proportional to the area have the limit defined for the delaunay triangulations of all delaunay sets . [ [ subclasses . ] ] subclasses . + + + + + + + + + + + we are interested in two subclasses of functionals , @xmath20 , which we now introduce . to define @xmath22 , let @xmath195 be a generic set of @xmath23 points in @xmath7 such that no point lies inside the convex hull of the others . the non - degenerate @xmath9-simplices spanned by the points cover the convex hull twice ; see radon @xcite . indeed , we can split them into two collections such that each forms a triangulation of @xmath195 : the delaunay triangulation , @xmath196 , and the other triangulation , @xmath6 . changing one triangulation into the other is a _ flip _ , a name motivated by the planar case in which it replaces one diagonal of a convex quadrilateral with the other . we give the flip a direction , leading from @xmath6 to @xmath152 . let now @xmath21 be a functional , let @xmath197 be the sums of @xmath198 over all @xmath9-simplices in @xmath6 , and define @xmath199 similarly . the class @xmath22 consists of all functionals @xmath200 for which @xmath201 . in @xmath30 , the extra property of functionals in @xmath22 suffices to prove our main result . in @xmath7 , for @xmath202 , we need more structure . the reason is the existence of triangulations that can not be turned into the delaunay triangulation by a sequence of directed flips ; see @xcite for finite examples in @xmath81 . such examples do not exist in @xmath30 ; see @xcite . let now @xmath195 be a finite set of points in @xmath7 . as before , we assume that @xmath195 is generic . let @xmath203 be a simplicial complex with vertex set @xmath195 , but note that we do not require that @xmath203 be a triangulation of @xmath195 . for example , we could start with a triangulation of @xmath195 and construct @xmath203 as the subset of @xmath9-simplices that do not belong to the delaunay triangulation together with their faces . let @xmath204 be the subset of simplices in @xmath205 contained in the underlying space of @xmath203 . finally , let @xmath206 be the sum of @xmath198 over all @xmath9-simplices in @xmath203 , and define @xmath207 similarly . the class @xmath24 consists of all functionals @xmath200 for which @xmath208 . the condition for @xmath21 to belong to @xmath24 is at least as strong as that for @xmath21 to belong to @xmath22 , which implies @xmath209 . in this section , we state and prove our main theorem and some of its implications . as already mentioned in section [ sec1 ] , the main result of this paper is an extension of optimality results for delaunay triangulations from finite sets to delaunay sets , which are necessarily infinite . section [ sec2 ] provides all the technical concepts needed to give a formal proof of the theorem that facilitates this result . let @xmath4 be a delaunay set in @xmath7 . 1 . in @xmath30 , @xmath31 implies @xmath210 for all uniformly bounded triangulations @xmath6 of @xmath4 . 2 . in @xmath7 , @xmath29 implies @xmath210 for all uniformly bounded triangulations @xmath6 of @xmath4 . proof . fix a uniformly bounded triangulation @xmath6 with parameter @xmath10 of @xmath4 . we prove the inequalities by comparing subsets of @xmath9-simplices of @xmath6 and @xmath211 . we begin with ( ii ) . for every radius @xmath27 , we write @xmath212 and @xmath213 for the sets of simplices contained in @xmath26 . furthermore , we write @xmath214 for the set of simplices contained in the underlying space of @xmath215 . summing @xmath21 over the @xmath9-simplices in these sets , we have @xmath216 + [ \sigma_{d'(\alpha ) } - \sigma_{d(\alpha ) } ] + \sigma_{d(\alpha ) } . \label{eqn : sigma } \end{aligned}\ ] ] the first difference on the right - hand side is non - negative by assumption of @xmath29 . we prove shortly that the second difference is bounded from above by a constant times @xmath143 . this implies that dividing by @xmath217 and taking the lower limit gives @xmath218 , as required . to prove the bound for the second term , we assume @xmath219 , so that the simplex count lemma implies that the number of @xmath9-simplices in @xmath220 is bounded from above by a constant times @xmath143 . we get the same bound for @xmath221 because @xmath215 is uniformly bounded , with parameter @xmath10 , so it covers all of @xmath222 , which implies @xmath223 . the proof of ( i ) is similar , except that we have to do more work to construct the sets of @xmath9-simplices , now triangles needed for the comparison . we assume @xmath224 and let @xmath215 and @xmath225 be as before . we construct @xmath226 by modifying @xmath215 through a sequence of directed flips applied to non - locally delaunay edges . iterating the directed flip , we can guarantee that all interior edges of @xmath227 are locally delaunay . a directed flip does not increase the size of the larger circumcircle ( see e.g. @xcite ) , so flipping does not take us outside the class of uniformly bounded triangulations . importantly , the flips turn a large portion of @xmath215 into delaunay triangles , namely @xmath228 . to see this , we note that every triangle @xmath229 contained in @xmath230 is also in @xmath152 . indeed , its circumsphere is contained in @xmath231 , which is contained in the underlying space of @xmath226 . if @xmath17 were not empty , we would have a vertex inside the circumcircle , which would imply an edge that is not locally delaunay between this vertex and @xmath17 , which is a contradiction ; see also @xcite where this argument is used to prove part ( ii ) of the delaunay triangulation theorem . finally , the triangles of @xmath226 contained in @xmath232 cover @xmath233 , which implies @xmath228 , as required . for the comparison , we consider @xmath234 + [ \sigma_{t'(\alpha ) } - \sigma_{d(\alpha-6q ) } ] \\ & & + [ \sigma_{d(\alpha-6q ) } - \sigma_{d(\alpha ) } ] + \sigma_{d(\alpha ) } . \end{aligned}\ ] ] the first difference is non - negative , and the second and third differences are bounded from above by a constant times @xmath143 . dividing by @xmath235 and taking the lower limit , as @xmath27 goes to infinity , we get @xmath218 , as required . there are many functionals on triangles that are known to be in @xmath22 . applying the main theorem thus gives many optimality results for delaunay triangulations of delaunay sets . let @xmath17 be a triangle in @xmath30 , with edges of length @xmath236 , let @xmath237 and @xmath238 be constants , and consider the following list of functionals : * @xmath239 . * @xmath240 . * @xmath241 . * @xmath242 . * @xmath243 . * @xmath244 . then @xmath245 for every delaunay set @xmath246 , for every uniformly bounded triangulation @xmath6 of @xmath4 , and for @xmath247 . it is easy to see that all listed functionals belong to @xmath19 . for finite sets , the optimality of the delaunay triangulation for @xmath248 and @xmath249 was proved in @xcite , for @xmath250 and @xmath251 it was proved in @xcite , for @xmath252 it was proved in @xcite , and for @xmath253 it was proved in @xcite . it follows that @xmath254 for @xmath247 , so the claim follows from ( i ) in the main theorem . we have one example of a functional on @xmath9-simplices that is in @xmath24 , namely the extension of @xmath255 to three and higher dimensions . writing @xmath256 to @xmath257 for the lengths of the @xmath258 edges of a @xmath9-simplex @xmath17 , we define @xmath259 ; see also @xcite . rajan proved that for finite sets in @xmath7 , the density of @xmath260 attains its minimum for the delaunay triangulation . we will extend his proof to show that @xmath260 belongs to @xmath24 . with this , we get another consequence of the main theorem . we have @xmath261 for every delaunay set @xmath0 and for every uniformly bounded triangulation @xmath6 of @xmath4 . proof . the main tool in this proof is the lifting of a point @xmath262 to the point @xmath263 , an idea that goes back to voronoi @xcite . note that @xmath264 lies on the graph of the function @xmath265 defined by @xmath266 . for @xmath267 , write @xmath268 for the corresponding set of lifted points , and let @xmath269 be its convex hull . assuming @xmath195 is finite and generic , @xmath269 is a convex polytope whose faces are simplices . we distinguish between _ lower _ faces whose outward normals point down against the direction of the @xmath45-st coordinate axis and _ upper _ faces whose outward normals point up . importantly , if we project all lower faces vertically to @xmath7 , then we obtain the delaunay triangulation of @xmath195 . consider now the functional @xmath270 that maps a @xmath9-simplex @xmath17 in @xmath7 to the @xmath45-dimensional volume between the convex hull of the @xmath112 lifted vertices and the graph of @xmath271 . more precisely , it is the volume of the portion of the vertical @xmath45-dimensional prism over @xmath17 that is bounded above by the convex hull of the lifted vertices and below by the graph of @xmath271 . it is not difficult to prove that @xmath270 is invariant under isometries of @xmath7 , and to use this fact to show that @xmath272 . the reason for our interest in @xmath273 is the relation @xmath274 proved in @xcite . since the two functionals differ only by a multiplicative constant , it follows that @xmath260 also belongs to @xmath19 . it remains to prove that @xmath260 belongs to @xmath24 , which we do by showing that @xmath273 belongs to @xmath24 . indeed , this should be clear from the lifting result : the lifted images of the @xmath9-simplices in the delaunay triangulation are closer to the graph of @xmath271 than those of other @xmath9-simplices . more specifically , if @xmath203 is a simplicial complex with finite vertex set @xmath195 in @xmath7 , and all simplices of @xmath275 are contained in the underlying space of @xmath203 , then the total @xmath45-dimensional volume we get for @xmath203 is larger than or equal to that we get for @xmath204 . but this implies @xmath276 , and therefore @xmath277 , as required . the main contribution of this paper is an extension of optimality results that hold for delaunay triangulations of finite sets to delaunay sets , which are necessarily infinite . in the plane , this extension holds for all functionals that improve upon flipping an edge that is not locally delaunay . in three and higher dimensions , we need stronger properties to prove the extension . it would be interesting to know whether these stronger properties are necessary . specifically , is it true that @xmath31 implies that the density of @xmath21 attains its minimum at the delaunay triangulation of a delaunay set in @xmath7 , also for @xmath202 ? similarly , are there functionals in @xmath24 that are not in @xmath22 , or is @xmath278 ?

we study densities of functionals over uniformly bounded triangulations of a delaunay set of vertices , and prove that the minimum is attained for the delaunay triangulation if this is the case for finite sets .

let @xmath1)$ ] be a conformal manifold of dimension @xmath2 . the @xmath3 gjms operator @xcite is a conformally invariant operator acting on densities @xmath4 , which is defined for all @xmath5 if @xmath6 is odd and for integers within the range @xmath7 if @xmath6 is even . this operator has a universal expression in terms of any representative metric @xmath8 $ ] with leading term the @xmath3 power of the laplacian . the idea for the construction is realizing densities as functions on the metric cone @xmath9 and computing the obstruction of its formal harmonic extension to the ambient space @xmath10 , where @xmath11 is an ambient metric of fefferman graham @xcite . after the appearance of @xcite , other gjms - like conformally invariant differential operators have been constructed in , e.g. , @xcite . in this article , we establish another variant of the gjms construction . our operators @xmath12 act on weighted trace - free symmetric ( covariant ) 2-tensors : @xmath13 here , the values that @xmath14 takes are the same as in the density case , @xmath15 is the space of trace - free symmetric 2-tensors on @xmath16 , and @xmath17 . the main tool of our construction is the lichnerowicz laplacian of the ambient metric @xmath11 , which is defined by @xmath18 where @xmath19 is the connection laplacian and @xmath20 , @xmath21 are the following tensorial actions of the ricci and riemann curvature tensors of @xmath11 : @xmath22 our intention to study the gjms construction for 2-tensors is because of its relation to the second variation of the total @xmath0-curvature , i.e. , the integral of branson s @xmath0-curvature @xcite . recall that , for 4-dimensional compact conformal manifold @xmath1)$ ] of positive - definite signature , the chern gauss bonnet formula for the total @xmath0-curvature @xmath23 is @xmath24 where @xmath25 is the euler characteristic , @xmath26 is the weyl tensor , and @xmath8 $ ] is any representative . one can deduce from this that @xmath27 and the equality holds if and only if @xmath1)$ ] is conformally flat . it turns out that there is a partial generalization of this fact to the higher dimensions . recall that a conformal metric @xmath28 $ ] is identified with a weighted 2-tensor @xmath29 . let @xmath30}$ ] be the conformal killing operator . then we have the following theorem , which is due to mller rsted @xcite . [ thm : sphere ] let @xmath31 be the sphere of even dimension @xmath32 . then , for any smooth 1-parameter family @xmath33 of conformal metrics on @xmath31 such that @xmath34 and @xmath35}$ ] , the total @xmath0-curvature @xmath36 attains a local maximum at @xmath37 . our main theorem contains theorem [ thm : sphere ] as a special case . consider the following decomposition of @xmath38 , which is valid for any compact positive - definite conformal manifold @xmath1)$ ] and a representative @xmath8 $ ] ( see @xcite ) : @xmath39}\oplus\cas^g_\mathrm{tt}(2).\ ] ] here @xmath40 is the space of tt - tensors ( trace - free and divergence - free tensors ) with respect to @xmath41 . this is an orthogonal decomposition with respect to the @xmath42-inner product , and if @xmath41 is einstein , the lichnerowicz laplacian @xmath43 of @xmath41 respects this decomposition . [ thm : qcurvlocalmaximum ] let @xmath1)$ ] be a compact conformally einstein manifold of positive - definite signature with even dimension @xmath32 , and @xmath41 an einstein representative with schouten tensor @xmath44 . then , for any smooth 1-parameter family @xmath33 of conformal metrics such that @xmath45 , the second derivative of the total @xmath0-curvature at @xmath37 is @xmath46 where @xmath47 is the @xmath48-component of @xmath49 with respect to . in particular , suppose there is an einstein representative @xmath41 with @xmath50 such that the smallest eigenvalue @xmath51 of @xmath52 satisfies @xmath53 then , for any @xmath33 for which @xmath54}$ ] , the total @xmath0-curvature attains a local maximum at @xmath37 . for @xmath55 , @xmath56 and @xmath57 . therefore the assumption for the latter half of theorem [ thm : qcurvlocalmaximum ] is satisfied , and hence theorem [ thm : sphere ] follows . some ideas for the proof of theorem [ thm : qcurvlocalmaximum ] are in order . let @xmath1)$ ] be a compact conformal manifold of even dimension @xmath32 ( here we may allow arbitrary signature ) . if we are given a smooth family @xmath33 of conformal metrics on @xmath16 such that @xmath45 , then the derivative @xmath58 is trace - free with respect to @xmath33 . as shown by graham hirachi @xcite , the derivative of @xmath36 is given by @xmath59 where @xmath60 is the fefferman graham ambient obstruction tensor of @xmath33 @xcite . in particular , if @xmath1)$ ] has vanishing obstruction tensor , which is the case if @xmath1)$ ] is conformally einstein for instance , then @xmath36 stabilizes at @xmath37 . in this case the second derivative of @xmath36 at @xmath37 is of interest . it is given by @xmath61 where @xmath62 is the linearization at @xmath63 of the obstruction tensor operator ( @xmath64 is trace - free because @xmath63 is obstruction - flat ) . this shows that it suffices to compute @xmath65 to derive the second variational formula of the total @xmath0-curvature . the construction of our operators @xmath12 leads to the fact that @xmath66 is equal to @xmath65 up to a constant factor for obstruction - flat manifolds . ( for @xmath67 and @xmath68 , since an explicit formula of the obstruction tensor is known , one can directly compute its linearization . in higher dimensions our result is really new , because there is no such concrete formula for @xmath69 . ) thus our gjms construction adds new knowledge of @xmath65 , which is previously studied in @xcite . if we specialize to the case of conformally einstein manifolds , explicit computation is possible thanks to a well - known associated ambient metric . we will derive a formula of @xmath12 restricted to @xmath70 with respect to an einstein representative @xmath41 with schouten tensor @xmath44 : @xmath71 then theorem [ thm : qcurvlocalmaximum ] is an immediate consequence . this article is organized as follows . preliminaries about ambient metrics and some preparatory lemmas are included in section [ sec : preliminaries ] . in section [ sec : gjms ] , our operators @xmath12 are constructed . one of the characterizations of @xmath12 is that it gives the obstruction to dilation - annihilating tt - harmonic extension of @xmath72 with respect to the ambient lichnerowicz laplacian @xmath73 . in section [ sec : variations ] , we first show that the variation of the normal - form ambient metric modified by adding a certain tensor in the image of the killing operator of @xmath11 is a best possible approximate solution to the harmonic extension problem mentioned above . using this fact , we prove that the trace - free part of @xmath65 equals to @xmath74 in general . in section [ sec : confeinstein ] , we work on conformally einstein manifolds and prove theorem [ thm : qcurvlocalmaximum ] . in this article , `` conformal manifolds '' are of arbitrary signature unless otherwise stated . index notation is used throughout . on ambient spaces we use @xmath75 , @xmath76 , @xmath77 , @xmath78 as indices , while on the original manifolds @xmath79 , @xmath80 , @xmath14 , @xmath81 are used . i wish to thank kengo hirachi for a suggestion to take a variational approach to the @xmath0-curvature and for insightful advice , and robin graham for discussion on our formula and letting me know the work @xcite . i also thank bent rsted for related discussions . moreover , colin guillarmou informed me that he has given another proof of with sergiu moroianu and jean - marc schlenker in a recent work @xcite . let @xmath1)$ ] be a conformal manifold of dimension @xmath6 of signature @xmath82 with metric cone @xmath9 . with a fixed representative metric @xmath8 $ ] , @xmath9 is trivialized as @xmath83 let @xmath84 be the associated ambient space : @xmath85 in our index notation , if @xmath84 is trivialized as above , we use the indices @xmath86 and @xmath87 for the @xmath88- and @xmath89-components , respectively . the space @xmath9 carries a natural @xmath90-bundle structure . the dilation @xmath91 , @xmath92 , is by definition the action of @xmath93 , and the infinitesimal dilation field is denoted by @xmath94 . the spaces of the densities , weighted 1-forms , and weighted covariant symmetric 2-tensors ( all of weight @xmath95 ) are denoted by @xmath96 , @xmath97 , and @xmath98 . by the metric cone @xmath9 , these spaces are realized as follows : @xmath99 the @xmath90-action extends to @xmath100 and so does @xmath94 . in terms of the extended @xmath94 , we define as follows : @xmath101 when @xmath102 satisfies @xmath103 , then @xmath104 makes sense as a section in @xmath98 via the identification . we use the notation @xmath105 to express this weighted tensor . let @xmath11 be a pre - ambient metric . this means that @xmath106 is a homogeneous pseudo - riemannian metric of signature @xmath107 defined on a dilation - invariant open neighborhood of @xmath9 in @xmath84 such that its pullback to @xmath9 is equal to @xmath29 . in the sequel we only work asymptotically near @xmath9 , so we may assume that all pre - ambient metrics are defined on the whole @xmath84 . we next introduce the _ straightness _ condition : @xmath108 if this is the case , the differential of the canonical defining function @xmath109 of @xmath9 is @xmath110 recall that it follows immediately from that @xmath111 the fefferman graham theorem states that there is a straight pre - ambient metric @xmath11 with @xmath112 in this article , such a metric @xmath11 is called an _ ambient metric_. when @xmath6 is odd , ambient metrics are unique modulo @xmath113 and the action of dilation - invariant diffeomorphisms on @xmath84 leaving points on @xmath9 fixed ( such diffeomorphisms are called _ ambient - equivalence maps _ in the sequel ) . if is @xmath6 even , the situation is subtle . for a 1-form @xmath114 , we define @xmath115 we say that @xmath102 is @xmath116 if 1 . @xmath117 ; 2 . @xmath118 and hence @xmath119 makes sense ; and 3 . @xmath120 is trace - free with respect to @xmath63 . then , ambient metrics are unique modulo @xmath121 and the action of ambient - equivalence maps . by ( * ? ? ? * equation ( 3.13 ) ) , the condition @xmath122 for ambient metrics actually forces @xmath123 let @xmath8 $ ] and consider the induced trivialization @xmath124 . if a straight pre - ambient metric @xmath11 is near @xmath9 of the form @xmath125 where @xmath126 is a 1-parameter family of metrics on @xmath16 with @xmath127 , then @xmath11 is said to be in _ normal form relative to @xmath41_. for any straight pre - ambient metric @xmath11 and a choice of @xmath8 $ ] , it is known ( * ? ? ? * proposition 2.8 ) that there exists an ambient - equivalence map @xmath128 such that @xmath129 is in normal form relative to @xmath41 . [ lem : actionofnablat ] let @xmath11 be a straight pre - ambient metric . for @xmath114 and @xmath102 , @xmath130 let @xmath131 . then , since the levi - civita connection is torsion - free , @xmath132+\tilde{\nabla}_{\tilde{\xi}}t)\\ & = t(\tilde{\tau}(\tilde{\xi } ) ) -\tilde{\tau}(\cal_t\tilde{\xi})-\tilde{\tau}(\tilde{\nabla}_{\tilde{\xi}}t ) = ( \cal_t\tilde{\tau})(\tilde{\xi})-\tilde{\tau}(\tilde{\xi } ) = ( w-1)\tilde{\tau}(\tilde{\xi } ) . \end{split}\ ] ] the second equality is proved similarly . now let @xmath11 be a fixed ambient metric . let @xmath133 be the subspace of formally trace - free tensors of @xmath134 , and @xmath135 the subspace of formally tt - tensors . moreover , we define as follows : @xmath136 if @xmath6 is odd , these spaces are invariant under @xmath113-modifications of @xmath11 . if @xmath6 is even , we need some technically - defined tensor spaces . for @xmath137 , we set @xmath138 ( `` att '' is for `` approximately tt '' ) and @xmath139 where @xmath140 is the divergence operator @xmath141 , and @xmath142 is the smallest integer not less than @xmath143 . then @xmath144 does not depend on the @xmath121-ambiguity of @xmath11 . to check this , let @xmath145 be another ambient metric with @xmath146 . then @xmath147 . since @xmath148 for any @xmath149 , the trace condition is not affected . the christoffel symbol of @xmath150 is given by @xmath151 where @xmath152 hence @xmath153 let @xmath154 . then @xmath155 and , because @xmath156 , @xmath157 therefore , if @xmath158 , @xmath159 . [ lem : ambientlift ] let @xmath11 be an ambient metric and @xmath72 , where @xmath5 . if @xmath6 is odd , then there exists @xmath160 such that @xmath161 . if @xmath6 is even , there exists @xmath162 such that @xmath161 as long as @xmath163 . in both cases , the restriction @xmath164 is uniquely determined . to prove the existence part , take any @xmath165 for which @xmath166 . we shall inductively construct @xmath167 for nonnegative integers @xmath168 such that @xmath169 suppose we have @xmath170 with @xmath171 . if @xmath167 , then @xmath172 is automatically guaranteed : @xmath173 we seek for @xmath174 assuming that it is of the form @xmath175 where @xmath176 satisfies @xmath177 , @xmath178 , and @xmath179 is such that the whole expression is trace - free and vanishes if contracted with @xmath94 ( hence @xmath180 , @xmath181 ) . minus of the divergences of the additional three terms on the right - hand side of are @xmath182 therefore , we first put @xmath183 , and set @xmath184 so that the @xmath185-term of the divergence of vanishes . this is possible for all @xmath168 if @xmath6 is odd , and until @xmath186 if @xmath6 is even . applying borel s lemma , the proof of the existence for @xmath6 odd is complete . when @xmath6 is even , we get @xmath187 . furthermore , if @xmath188 is an even number , then @xmath189 can be made @xmath190 . anyway , @xmath189 finally becomes @xmath191 , and the existence for @xmath6 even is proved . let us once again take @xmath192 as we did in the beginning of this proof . if @xmath149 is as in the statement , then since @xmath193 and @xmath161 , @xmath149 must be written as @xmath194 where @xmath195 . moreover , in order @xmath196 to be satisfied , @xmath197 should be @xmath198 . then @xmath199 therefore , @xmath200 mod @xmath201 is determined by the condition @xmath202 . if we put @xmath203 into @xmath200 , then the right - hand side will be @xmath204 . thus @xmath200 is uniquely determined in order to satisfy @xmath205 . we call @xmath206 in lemma [ lem : ambientlift ] the _ ambient lift _ of @xmath207 . let @xmath1)$ ] be a conformal manifold of dimension @xmath2 and @xmath11 an ambient metric . we shall play with the following three operators : @xmath208 just as in the case of the classical gjms construction , one can verify the following . the operators @xmath143 , @xmath209 , @xmath210 enjoy the @xmath211 commutation relations : @xmath212=2x,\qquad [ h , y]=-2y,\qquad [ x , y]=h.\ ] ] the proof is left to the reader . consequently we have the following identities : @xmath213&=-my^{m-1}(h - m+1),\\ \label{eq : sl2commutations2 } [ x^m , y]&=mx^{m-1}(h+m-1),\\ \label{eq : sl2commutations3 } y^{m-1}x^{m-1}&=(-1)^{m-1}(m-1)!h(h+1)\cdots(h+m-2)+xz_m,\end{aligned}\ ] ] where @xmath214 is some polynomial in @xmath143 , @xmath209 , @xmath210 . we are going to verify that @xmath143 , @xmath209 , and @xmath210 preserve the subspaces @xmath215 when @xmath6 is odd and @xmath216 when @xmath6 even . for this we need two lemmas . for @xmath217 , @xmath114 , @xmath218 in , we may also replace @xmath219 with the hodge laplacian @xmath220 . first we compute @xmath221 : @xmath222 hence it is clear that @xmath223 implies @xmath224 and that @xmath225 implies @xmath226 . so , to prove , it remains to show that @xmath227 is @xmath228 . this is checked directly : @xmath229 and therefore @xmath230 by bochner s formula @xmath231 , @xmath232 is clear . let @xmath233 be defined by @xmath234 then it is known that , on any symmetric 2-tensor , @xmath235 [ lem : dric ] when @xmath6 is even and @xmath137 , @xmath236 let @xmath237 . then @xmath238 therefore @xmath239 on the other hand , since @xmath240 is at least @xmath201 , we can write @xmath241 . hence , by and , @xmath242 consequently , @xmath243 . [ prop : mappingpropertyofsl2 ] if @xmath6 is odd , then for any @xmath95 , @xmath244 if @xmath6 is even , @xmath245 since the case @xmath6 odd is easier to prove , we discuss the case @xmath6 even . it is clear that @xmath246 . for @xmath247 , @xmath248 , @xmath249 , and @xmath250 hence @xmath251 . it remains to show that @xmath252 . the trace of @xmath253 is @xmath254 by . furthermore , @xmath255 and hence @xmath256 the last equality is because of . this implies @xmath257 . finally , and lemma [ lem : dric ] show @xmath258 . [ thm : gjmsaslaplacianpower ] let @xmath5 if @xmath6 odd , and @xmath259 if @xmath6 even . for any @xmath72 , let @xmath260 be any extension of the ambient lift @xmath206 . then @xmath261 depends only on @xmath262 and not on the extension . furthermore , @xmath263 makes sense as a section in @xmath264 . we work on the case @xmath6 even only . any two extensions of @xmath206 differs by a tensor of the form @xmath265 , where @xmath266 . equation shows that the commutator @xmath267 $ ] vanishes on @xmath268 and hence @xmath269 . in particular , using lemma [ lem : ambientlift ] one can take @xmath270 as an extension of @xmath206 . then by proposition [ prop : mappingpropertyofsl2 ] , @xmath271 and @xmath263 is defined . [ thm : gjmsasobstruction ] let @xmath5 if @xmath6 odd , and @xmath259 if @xmath6 even . let @xmath72 and @xmath206 its ambient lift . then there exists a solution @xmath272 if @xmath6 odd , and @xmath270 if @xmath6 even , to the problem @xmath273 which is unique modulo @xmath274 . for any such @xmath149 , @xmath275 is independent of the ambiguity that lives in @xmath149 , and agrees with @xmath276 up to a constant factor : @xmath277 we work on the case @xmath6 even only . let us begin with an arbitrary extension @xmath278 of @xmath206 . if an extension @xmath279 satisfies @xmath280 , then it has a modification @xmath281 , @xmath282 , which is unique modulo @xmath283 , satisfying @xmath284 . in fact , by , we have @xmath285 therefore @xmath286 can be taken so that @xmath284 unless @xmath287 . hence there is @xmath149 with the property stated in the theorem . let @xmath288 , @xmath289 . then , by , @xmath290 . hence . except in the case where @xmath6 is even and @xmath291 , @xmath292 is trace - free since @xmath293 and @xmath294 are both @xmath295 . let @xmath1)$ ] be a conformal manifold of dimension @xmath2 and @xmath11 an ambient metric . we call @xmath296 the _ gjms operator on trace - free symmetric 2-tensors _ , where @xmath260 is any extension of the ambient lift of @xmath262 . ( one can remove @xmath297 unless @xmath6 is even and @xmath291 . ) in particular , when @xmath298 even , @xmath299 is called the _ critical gjms operator on trace - free symmetric 2-tensors_. [ thm : invarianceofgjms ] the gjms operators on trace - free symmetric 2-tensors do not depend on the choice of @xmath11 , and hence are conformally invariant differential operators . for the case where @xmath6 is even and @xmath291 , the direct verification of the conformal invariance is not easy . we will see in theorem [ thm : differentialofobstruction ] that , up to a constant factor , @xmath300 is equal to @xmath301 , which is clearly conformally invariant . here , we prove the theorem in the case @xmath6 odd and the case @xmath6 even , @xmath302 . by theorem [ thm : gjmsasobstruction ] , we may work with @xmath303 instead of @xmath304 . let @xmath11 be an ambient metric , @xmath72 and @xmath149 a solution to the problem stated in theorem [ thm : gjmsasobstruction ] . then , if @xmath128 is an ambient - equivalence map , @xmath305 solves the same problem with respect to @xmath129 . since @xmath306 , the restrictions of @xmath307 and @xmath308 to @xmath309 therefore we may assume that @xmath11 is in normal form . when @xmath6 is odd , the assertion is now clear because @xmath11 is formally unique if it is in normal form . so we assume that @xmath6 is even in what follows . it suffices to show that , if @xmath11 , @xmath310 are ambient metrics in normal form and @xmath270 , @xmath311 let @xmath312 . from * equation ( 3.16 ) ) , one concludes that @xmath313 and @xmath314 . therefore @xmath315 in addition , @xmath316 and @xmath317 by ( * ? ? ? * equation ( 6.1 ) ) . hence @xmath318 . moreover , if @xmath319 , then @xmath320 again by ( * ? ? ? * equation ( 6.1 ) ) , we have @xmath321 , @xmath322 , @xmath323 and hence @xmath324 . we close this section with a lemma that is proved just like the construction of @xmath149 in theorem [ thm : gjmsasobstruction ] . [ lem : solvinglaplacian ] let @xmath5 . for any @xmath325 , there exists @xmath326 such that @xmath327 likewise , for any @xmath328 , there exists @xmath329 such that @xmath330 in both problems , we may arbitrarily prescribe the values along @xmath9 ; if we prescribe @xmath331 , @xmath332 , then @xmath333 , @xmath334 are unique modulo @xmath274 . let @xmath11 be an ambient metric for a conformal manifold @xmath1)$ ] of dimension @xmath2 . recall that , from general calculations on ( pseudo-)riemannian curvature tensors , the differential of the ricci tensor operator ( which we write @xmath335 here ) is @xmath336 where @xmath337 is the dual of the divergence @xmath338 and @xmath339 is defined by @xmath340 . therefore , for @xmath6 even , a solution @xmath341 to the problem in theorem [ thm : gjmsasobstruction ] approximately solves @xmath342 , and hence it is expected that we can read off @xmath64 from the asymptotics of @xmath149 . this will finally turn out to be true , but since the definition of @xmath343 depends on the existence theorem of normal - form ambient metrics , in order to capture @xmath64 our starting point has to be infinitesimal modifications of ambient metrics in normal form . the differential equation that they ( approximately ) satisfy is different from @xmath344 . so we shall establish a method for translating solutions of the two equations . let @xmath1)$ ] be an @xmath6-dimensional conformal manifold with @xmath32 even and @xmath345 . suppose that @xmath346 is a family of conformal metrics ( here we use @xmath347 for the parameter , because @xmath88 will denote a coordinate on @xmath84 ) with @xmath45 such that @xmath348 . let @xmath8 $ ] be any representative metric , and @xmath349 the corresponding representatives of @xmath346 . by the method of @xcite , we can construct a family of ambient metrics @xmath350 such that @xmath351 and @xmath352 smoothly depends on the two variables @xmath89 , @xmath347 . all these metrics satisfy @xmath353 and @xmath354 . differentiating these equations , we conclude that @xmath355 solves @xmath356 note that it satisfies @xmath357 , @xmath358 , and hence @xmath359 therefore it holds that @xmath360 since the obstruction tensor @xmath361 is defined by @xmath362 we have @xmath363 [ lem : harminizationofnormalsolution ] let @xmath364 be as above . then , there exists a dilation - invariant vector field @xmath365 on @xmath84 such that @xmath366 and @xmath367 where @xmath368 is the killing operator : @xmath369 . such @xmath365 is unique modulo @xmath370 and satisfies @xmath371 , @xmath372 . the equation to be solved is @xmath373 . by a straightforward calculation , @xmath374 since @xmath375 for any @xmath365 satisfying @xmath366 , the equation simplifies to @xmath376 . by lemma [ lem : solvinglaplacian ] , @xmath365 is uniquely determined up to an @xmath370 ambiguity . if we write @xmath377 , then @xmath378 . on the other hand , @xmath379 should be @xmath295 by . consequently @xmath380 , i.e. , @xmath381 . moreover , @xmath382 . let @xmath383 . it is a consequence of the fact that the ricci operator commutes with diffeomorphisms that @xmath384 . since @xmath366 , @xmath122 , and @xmath385 , @xmath386 itself is @xmath387 and @xmath388 . therefore @xmath389 , @xmath390 . moreover , @xmath391 and hence @xmath392 , @xmath393 . thus we conclude @xmath394 [ lem : normalsolutionmodifiedtoharmonicsolution ] let @xmath364 and @xmath365 be as in lemma [ lem : harminizationofnormalsolution ] . then @xmath395 and it is a solution to . it remains to show that @xmath396 . by taking the trace of , we obtain @xmath397 . in addition , since @xmath372 , we have @xmath398 . hence , by lemma [ lem : solvinglaplacian ] , @xmath399 . then @xmath391 implies @xmath400 . furthermore , @xmath401 and @xmath402 since @xmath403 for @xmath404 , one can determine @xmath333 so that @xmath405 . then @xmath406 is still @xmath295 , and hence @xmath407 by lemma [ lem : solvinglaplacian ] . [ lem : lichnerowiczencodesobstruction ] let @xmath364 and @xmath365 be as in lemma [ lem : harminizationofnormalsolution ] and set @xmath408 . then @xmath409 , and @xmath410 vanishes . recall that @xmath411 let @xmath412 and @xmath377 . we proved in lemma [ lem : harminizationofnormalsolution ] that @xmath195 . as in the proof of lemma [ lem : actionofnablat ] , we compute @xmath413 thus @xmath414 vanishes . on the other hand , if we write @xmath415 , then @xmath416 and hence @xmath417 . this completes the proof . [ thm : differentialofobstruction ] let @xmath1)$ ] be a conformal manifold of even dimension @xmath6 . then the differential of the obstruction tensor @xmath65 is given by @xmath418 let @xmath364 , @xmath365 as in lemma [ lem : harminizationofnormalsolution ] and @xmath408 . by lemma [ lem : normalsolutionmodifiedtoharmonicsolution ] , @xmath300 is equal to the trace - free part of @xmath419 . by the previous lemma , @xmath420 . therefore , @xmath421 on the other hand , @xmath422 , for @xmath423 for any @xmath63 . hence . combining the theorem above and equation , we obtain the following . let @xmath1)$ ] be a compact conformal manifold of even dimension @xmath32 with vanishing obstruction tensor . let @xmath33 be a family of conformal structures such that @xmath45 . then the second derivative of the total @xmath0-curvature at @xmath37 is @xmath424 where @xmath49 and @xmath425 is the critical gjms operator on trace - free symmetric 2-tensors . recall that , for @xmath8 $ ] einstein with @xmath426 so that @xmath44 , the following formula gives an ambient metric that is ricci - flat : @xmath427 the inverse of @xmath11 is @xmath428 and the christoffel symbol of @xmath11 is given by @xmath429 a direct computation shows that @xmath430 , where @xmath431 and @xmath26 are the weyl tensors of @xmath11 and @xmath41 , respectively ( the latter is trivially extended to @xmath432 ) . the other components of @xmath431 are zero . [ lem : ambientlichnerowiczofconfeinstein ] let @xmath11 as above , and suppose that @xmath433 is of the form @xmath434 where @xmath435 is a symmetric 2-tensor on @xmath436 . then @xmath437 where @xmath438 is the lichnerowicz laplacian of @xmath41 . [ thm : criticalgjmsforconfeinstein ] let @xmath1)$ ] be a conformally einstein manifold with @xmath443 , and @xmath8 $ ] an einstein representative with schouten tensor @xmath44 . then , the action of @xmath12 restricted to @xmath70 is given by . let @xmath444 and @xmath445 . then @xmath446 since @xmath447 is a tt - tensor on @xmath436 , @xmath149 itself is a tt - tensor with respect to @xmath11 , and hence is an extension of the ambient lift of @xmath262 . we may compute @xmath304 by lemma [ lem : ambientlichnerowiczofconfeinstein ] . by taking the value along @xmath9 and trivializing with respect to @xmath41 , we obtain . let @xmath448}\xi+\varphi^g_\mathrm{tt}$ ] be the decomposition of @xmath49 with respect to and @xmath449 the flow generated by @xmath450 . then @xmath451 satisfies @xmath452 and the total @xmath0-curvature of @xmath453 is equal to @xmath36 . therefore @xmath454 and thus follows from theorem [ thm : criticalgjmsforconfeinstein ] . under the assumption of the latter half of the theorem , any eigenvalue of @xmath455 is strictly positive for @xmath456 . therefore , if @xmath457 , the second derivative of @xmath36 at @xmath37 is negative . gjms a. l. besse , _ einstein manifolds _ , ergebnisse der mathematik und ihrer grenzgebiete ( 3 ) , 10 , springer - verlag , berlin , 1987 , xii+510 pp . t. p. branson , sharp inequalities , the functional determinant , and the complementary series , _ trans . soc . _ * 347 * ( 1995 ) , 36713742 . t. branson , @xmath0-curvature and spectral invariants , _ rend . circ . mat . palermo ( 2 ) suppl . _ * 75 * ( 2005 ) , 1155 . t. branson and a. r. gover , conformally invariant operators , differential forms , cohomology and a generalisation of q - curvature , _ comm . partial differential equations _ * 30 * ( 2005 ) , 16111669 . t. p. branson and a. r. gover , the conformal deformation detour complex for the obstruction tensor , _ proc . * 135 * ( 2007 ) , 29612965 . t. p. branson and a. r. gover , origins , applications and generalisations of the @xmath0-curvature , _ acta appl . _ * 102 * ( 2008 ) , 131146 . c. fefferman and c. r. graham , conformal invariants , in _ the mathematical heritage of lie cartan ( lyon , 1984 ) _ , astrisque , numero hors serie , soc . math . france , montrouge , 1985 , 95116 . c. fefferman and c. r. graham , _ the ambient metric _ , annals of mathematics studies , 178 , princeton university press , princeton , nj , 2012 , x+113 pp . a. r. gover and l. j. peterson , the ambient obstruction tensor and the conformal deformation complex , _ pacific j. math . _ * 226 * ( 2006 ) , 309351 . c. r. graham and k. hirachi , the ambient obstruction tensor and @xmath0-curvature , in _ ads / cft correspondence : einstein metrics and their conformal boundaries _ , irma lect phys . 8 , eur . soc . , zrich , 2005 , 5971 . c. r. graham , r. jenne , l. j. mason , and g. a. j. sparling , conformally invariant powers of the laplacian . i. existence , _ j. london math . ( 2 ) _ * 46 * ( 1992 ) , 557565 . c. guillarmou , s. moroianu , and j schlenker , the renormalized volume and uniformisation of conformal structures , preprint , ` arxiv:1211.6705 ` . n. m. mller and b. rsted , rigidity of conformal functionals on spheres , preprint , ` arxiv:0902.4067 ` .

we construct a series of conformally invariant differential operators acting on weighted trace - free symmetric 2-tensors by a method similar to graham jenne mason sparling s . for compact conformal manifolds of dimension even and greater than or equal to four with vanishing ambient obstruction tensor , one of these operators is used to describe the second variation of the total @xmath0-curvature . an explicit formula for conformally einstein manifolds is given in terms of the lichnerowicz laplacian of an einstein representative metric .

the astra ( astrometric and phase - referencing astronomy ) upgrade program , funded by the nsf major research instrumentation program , aims at extending the sensitivity and spectral resolution of the keck interferometer ( ki ) through phase referencing and to implement a narrow - angle astrometry mode to enable a much broader scientific use of the ki . in this paper we report on results of the successful implementation of the first astra mode , the self phase referencing ( spr ) mode . in the spr mode , 55 % of the light from the on - axis science target is used to provide fringe tracking ( i.e. , phase referencing ) information while 20 % is used to make the science measurement . due to the stabilized fringes , @xmath8100x longer integrations can be taken on the second camera . therefore , it is possible in spr to increase the spectral resolution of the interferometric measurement by at least an order of magnitude . spr as currently implemented and provided to the general user , offers a spectral resolution @xmath9 of 2000 , which is nyquist sampled at r = 1000 ( 330 pixels across the k-band ) at k @xmath10 . further technical aspects and results from the commissioning are discussed by @xcite . besides simple amplitude spectro - interferometry at this resolution , the here presented data demonstrate that the differential visibility phase can be retrieved at a precision of at least 3 mrad for a bright star . non - linear phase changes in the source spectrum can be measured , and typically indicate translations of the photo - center over the respective spectral channels on the sky . the operational readiness of ki - spr was successfully demonstrated with the here discussed observations of on the night of april 25 , 2008 ( ut ) . in addition , several young stellar objects were observed successfully @xcite . we show at the example of 48 lib the wealth of spatial and spectral information that can be provided by spr data . 48 lib is a well studied classical be - star with circumstellar shell - emission lines . previous optical - nearinfrared long - baseline interferometric ( olbi ) measurements of be - star - shells demonstrated that 3d - models for h - line and continuum emission often overpredict the true size , which demonstrates the need for direct olbi measurements . narrow - band observations with the mark iii interferometer , resulted in be - star disk diameter estimates ranging in 2.6 to 4.5 mas @xcite , well suited for the angular resolution of the 85 m baseline of the ki ( @xmath11 ) . we chose 48 lib as science demonstration target because recent detection of very narrow satellite absorption features in fe ii shell lines reliably indicate very high inclination angles . this ensures strong features in the differential visibility and phase signals across the emission lines . a large fraction of be stars show cyclic variations in the ratio between the violet and red flux of the hi emission lines ( v / r ) , an enigmatic phenomenon studied since almost a century @xcite . early measurements with the gi2 t interferometer spatially and timely resolved the @xmath12 emission of @xmath13-cas , showing spatial variation of the line emission region with time @xcite . the fast rotation of be - stars ( near - critical rotational velocities for cooler stars like 48 lib , * ? ? ? * ) results in a flattening of the stellar shape , which can be observationally confirmed by olbi . while the rapid rotation might play a role in the ejection of the circumstellar envelope , the quadrupole moment of the gravitational potential of a flattened be star is expected to create a prograde one - armed spiral density pattern , slowly precessing in the disk and creating the v / r cycles . the latter authors calculate that the radial extension of such a disk oscillation mode should be confined to a few stellar radii , which can be probed by interferometry . central properties of this theory have been observed indirectly by the continuous observation of individual line profiles , and the detection of phase - shifts between different emission line cycles pointing towards a spiral density perturbation @xcite . spectro - interferometric observations of the photocenter shift in hydrogen lines of @xmath14 tau appear to directly confirm a one - armed density mode in the disk . after summarizing the observation and data reduction ( sect . [ sec:2 ] ) , we will discuss in sect . [ sec:3 ] that our data show that the radial dependence of the hydrogen emission regions can be observed _ directly _ with ki - spr , thanks to resolving several emission lines of the pfund series together with @xmath6 at unprecedented differential precision in the @xmath15-band . concluding remarks are given in sect . [ sec:5 ] . ccccc name & @xmath16@xmath17 & ut & bl@xmath18 & stellar diam + & & & & [ mas ] @xmath19 + ( tar , b4iii@xmath20 ) & 4.9/4.8/4.6 & 13:20:49 & 66.6 m , 39@xmath21 eon & @xmath22 + ( cal , a4v ) & 5.4/5.3/5.1 & 13:28:17 & & @xmath23 + details of the observations are given in table [ tab:11 ] . each dataset consists of 155 frames of 0.5 sec integration time , taken on the second fringe camera . the frames were selected to have an absolute group delay of less then 4.5 @xmath4 m to ensure high snr and limited impact of the dispersion introduced by the air - filled delay lines ( see appendix c in * ? ? ? the remaining 137 ( 151 ) frames of target ( calibrator ) were corrected for group delay ( linear ) and dispersion ( quadratic ) wavenumber slopes . the resulting pre - processed frames were stacked by calculating the mean in each wavelength bin , and the standard deviation of this mean is used throughout the paper as error . this statistical error is adequately describing the differential uncertainty of the data . the total flux data were flux normalized before the averaging . this is necessary to avoid that the larger absolute flux variation due to the spatial filtering of the single - mode fibre fed fringe camera dominates the here interesting differential flux variation between adjacent lambda bins . the resulting measurables of the spr observation ( flux , @xmath24 , and @xmath25 ) are shown in fig . [ fig:1 ] . to achieve the rest wavelength calibration shown in the figure , two steps were performed . first , the observed wavelength calibration needs to be rectified . the slope is measured internally by a fourier transform based technique . in standard ki observing setup , this slope is applied automatically to the data provided through the nexsci archive . furthermore , we used telluric absorption lines in the raw spectra to estimate an offset in the wavelength calibration ( typically of order of a few tenths of one spectral pixel ) , and to verify the archived rectification . the fact that our data do not show the full @xmath0 band is due to an alignment problem of the early instrument setup used . for current performance , we refer to the ki support webpage . in a second step , the corrections for barycentric motion of the earth ( -13 @xmath26 ) and peculiar motion of the star ( -6 @xmath26 , * ? ? ? * ) have been applied . note for @xmath6 , that the offset between zero @xmath25 and rest wavelength is significantly larger than the accuracy of our wavelength calibration . it is attributed to observing a profile which is significantly asymmetric at the level of the used spectral resolution . the quality of the resulting wavelength calibration is shown by overplotting the hi - recombination line centers in fig . [ fig:1 ] ( dotted lines ) . it is reassuring , that ( 1 ) the flux line profiles are centered , that ( 2 ) the various pfund lines show the same line profile , and most importantly , that ( 3 ) the @xmath25 of both @xmath6 and pfund lines crosses zero close to the transition wavelengths , and with the _ same _ sign of the slope . ( 3 ) is expected for an inclined keplerian disk . the target data were calibrated with a subsequently measured unresolved star , observed only 8 min later under equal atmospheric conditions . we applied simple calibration strategies : the raw target flux was divided by the calibrator , and multiplied by a measured spectral template from the irtf spectral library ( a3v , vacca , personal comm . ) containing the hi absorption features of the calibrator of similar spectral type . the spectral template provides a relative flux precision of 1 % , and was folded to @xmath27 , and divided by a black body continuum fit to the line free regions before application to our data . the resultant spectrum ( upper panel in fig . [ fig:1 ] ) thus shows in each spectral bin the flux with respect to the calibrator . the @xmath24 was corrected for the system visibility , estimated by observing an unresolved calibrator . we did not apply any flux bias correction in our data calibration . although the primary fringe tracker at ki has a flux bias of a few percent per magnitude internal tests have demonstrated that the detector used for these science observations does not display such a bias . the differential phase is calibrated by subtraction of the calibrator @xmath25 . to show the impressive differential stability of the data over the used band , we overplotted in red linear continua , fitted to the line - free regions ( @xmath28 m and @xmath29 m ) , over flux ratio and @xmath24 . the slightly red slope of the linear flux ratio continuum probably originates in the shell - induced nir excess of 48 lib , which photosphere is hotter than the calibrator of spectral type a4v . the continuum trend in the visibility is due to resolving the innermost circumstellar gas , and is discussed in sect . [ sec:31 ] . the red line in the @xmath25 panel marks zero phase , and is not a fit . the resulting pixel - to - pixel errors are @xmath30 , @xmath31 , and @xmath32 . the higher relative precision of the @xmath25 signal with respect to the precision of the visibility is expected for such spectro - interferometric measurements @xcite . this differential stability at the level of a few @xmath5 is unprecedented for spectrally dispersed olbi data at the spectral resolution and sensitivity offered by ki - spr . it is the key to derive the scientific results presented in the next section . a more comprehensive description of the spr mode and its sensitivity , performance , and technology will be given elsewhere @xcite . and @xmath7-emission lines are clearly visible _ center : _ mean calibrated @xmath24 of 48 lib showing that both the nir continuum and the recombination line emission are spatially resolved by the interferometer . a linear continuum ( red line ) was fitted to the line free regions . _ bottom : _ the calibrated differential phase data . the red line marks zero phase . all plots show the rest wavelength of the target . the vertical dotted lines indicate the rest wavelength of the recombination lines . it is apparent that all lines show the same slope at the line center , as expected for disk emission . [ fig:1 ] ] lccc @xmath33 ( gauss - disk only ) : & @xmath34 & [email protected] mas & + @xmath33 ( gauss - disk + star ) : & @xmath36 & [email protected] mas & + @xmath6 & @xmath37 & @xmath38 & @xmath39 + @xmath40 ( norm . ) & @xmath41 & @xmath42 & @xmath43 + @xmath44 ( norm . ) & @xmath45 & @xmath46 & @xmath47 + @xmath48 ( mas ) & @xmath49 & @xmath50 & @xmath51 + _ pc _ @xmath52 ( mas ) & @xmath53 & @xmath54 & @xmath51 + @xmath55 @xmath56 & @xmath57 & @xmath58 & @xmath39 + @xmath40 ( norm . ) & @xmath59 & @xmath60 & @xmath61 + @xmath44 ( norm . ) & @xmath62 & @xmath62 & @xmath61 + @xmath48 ( mas ) & @xmath63 & @xmath64 & @xmath65 + _ pc _ @xmath52 ( mas ) & @xmath66 & @xmath66 & @xmath67 + fig . [ fig:1 ] shows that different components have been detected in the spectra of : a clearly resolved continuum emission , and hi recombination line emission of @xmath6 and various pfund lines . the visibility data suggest that the continuum and the line emission stem from spatially separated , distinct regions in the circumstellar disk . also , they derive from different emission processes . therefore , we discuss in the following the two components separately . the disk continuum emission , emitted by bound - free and free - free thermal emission in the disk , is clearly resolved . to facilitate the comparison of our results with other published disk diameters , we give the size of a disk - only fit to the continuum visibilities ( @xmath68 ) in table [ tab:2 ] . we model the disk continuum by a maximally inclined gaussian profile , with a position angle of 50@xmath21 which is orthogonal to the axis of polarization @xcite . however the measured continuum emission consists of both stellar and disk contributions . taking into account , that the disk continuum of 48 lib contributes one third of the @xmath15-band emission , the resulting disk size scale ( @xmath69 ) is seven times larger than the stellar photospheric diameter . source variability may compromise this estimation of @xmath70 . but a flux calibration of the ki photometry of 48 lib , based on hd 145607 is consistent with the 2mass point source flux of our target @xcite to better than 10% , which is the precision level of our absolute flux calibration . report an irregular variability in the hipparcos band at the 5 % level . therefore , we can expect that variability of the star - disk flux ratio has no strong effect on the @xmath70 value reported above . note that although important for the estimation of the disk size , we can not derive the nir disk excess from the visibility data due to our small @xmath71-coverage . similarly , two - dimensional physical or phenomenological disk models of the continuum emission are not well constraint by a single baseline measurement . @xcite showed for several be stars , that , even for significantly two - dimensional @xmath71-coverages , physical models , using radiative transfer calculations , and phenomenological models , using gaussian intensity profiles are difficult to distinguish by the visibility data alone . physical and phenomenological fwhm estimates of the average disk compactness may differ by 50 - 100 % @xcite . this may be used as an accuracy estimate for a physical interpretation of the here derived @xmath70 . since near - infrared disk excess and @xmath70 of 48 lib are within the typical range of be star disks , we can expect that a physical modeling of the disk continuum , as done by @xcite , would lead to similar densities as described therein . the estimated disk continuum - to - stellar radii ratio matches other interferometrically observed edge - on systems with comparable properties ( e.g. @xmath14 tau in * ? ? ? the typically found moderate extent of the @xmath15-band emission size ( 5 - 10 times the stellar radius ) is significantly smaller than recently modeled , confirming the need for interferometric constraints to understand the physical conditions in the stellar outflows of be - stars . there is several indications in our data , that the disk continuum emission zone is inside the @xmath6 and smaller or equal to the @xmath72 recombination line emission zones . all fitted linear characteristics of the violet and red part of the @xmath6 ( @xmath55 ) line emission zones ( that is @xmath48 and the photocenter shifts _ pc _ @xmath52 , for details see the next section ; violet and red labels negative and positve doppler shifts , following standard be - star terminology ) indicate that the majority of the line emission is created at stellocentric radii larger than ( comparable to ) the radii dominating the continuum emission . furthermore , the @xmath55 line center in total and correlated flux drops below the continuum level ( fig . [ fig:3 ] ) . this points to a significant absorption of the continuum flux by gas _ in front _ of the continuum source , that is the continuum emission is embedded and surrounded by the zone dominating the line emission . therefore , we confirm the findings of ( * ? ? ? * studied @xmath12 versus continuum size ) and that the observed @xmath6 ( and @xmath55 ) hydrogen line emission in be stellar shells appears to occur at larger ( or equal ) stellocentric radii than the disk continuum emission . the flux ratio ( _ top _ ) is normalized to the linear continuum shown in fig . [ fig:1 ] . the _ central _ panel shows the correlated flux , normalized to the linear continuum fit . note that the linear flux ratio continuum fit may overestimate the actually observed continuum emission due to absorption by gas _ in front _ of the continuum emission , which would result in underestimated fluxes at the line center . the measured differential phase is plotted in the _ bottom _ panel . at the top right of each figure , the integrated violet and red line profile is given . the red area marks the region of integration ( 100 - 250 km / s ) , which excludes the uncertain central continuum absorption . the wavelength - differential precision is indicated by the red errorbar in the left of each panel . [ fig:2 ] ] profile at the flux levels of pf-(5 - 24 ) in the same way as in fig . [ fig:2 ] . the smoother line profile with respect to @xmath6 derives from the line averaging over the linearly interpolated individual pfund lines . the hatched area indicates the beginning of the velocity region where the shape of the mean line profile is significantly compromised by the neighboring lines due to the averaging . the profile at large velocities ( @xmath73 ) are a result of our data processing , and unreal . we show this part here to demonstrate that the central , real line profile is more significant than these relatively large profile changes in the outer velocity - space compromised from the averaging . the statistical error of the mean shapes shown is indicate by the red errorbar in the left of each figure . [ fig:3 ] ] the emission line widths of be - stars are dominated by large - scale kinematic doppler broadening ( due to the fact that the ionized gas is located in a rotating disk around the star ) , without significant contributions of non - kinematic scattering processes . the disk of 48 lib is known to be edge on , which simplifies the geometric and kinematic interpretation of the line profiles . the spectrometer does not fully nyquist - sample the double peak of the @xmath6 and pfund lines , but the central dip in the flux and the shape of the @xmath25 signal show that the violet ( v ) and red ( r ) part of the disk are clearly separated by the spectrometer , and that meaningful v / r ratios can be derived . @xmath7 - 24 .. 34 have been detected for the first time in an interferometric spectrum . the five strongest individual @xmath7-lines are clear detections ( @xmath74 ) above the differential precision level of flux ratio , visibility , and @xmath25 . to improve on the snr for the @xmath72lines , and to minimize the effects of the limited sampling of the line - profile , we average over the five strongest , detected @xmath72lines ( 24 .. 28 ) . to do so , the @xmath7225 .. 28 lines were scaled to match flux levels of @xmath7224 . this avoids that the final average profile ( @xmath55 ) is affected by the individual line fluxes . we checked that none of the features of @xmath55 discussed below is significantly altered or vanishes when averaging over all 11 detected pfund lines . however , our choice of averaging over the five strongest lines to create @xmath55 leads to the highest snr . in addition , the omitted higher - order pfund lines are closer in rest wavelength , and therefore , overlapping of adjacent line profiles would become apparent and compromise the profile of @xmath55 . similarly , at velocities @xmath75@xmath26 ( indicated by the dashed region in fig . [ fig:3 ] ) , the average profile is affected by adjacent lines . we purposefully show in the figure a large velocity space including these artifacts to demonstrate that @xmath55 emerges even out of these averaging artifacts . unless indicated otherwise , we use a linear flux ratio continuum fit . using a black body instead of a linear shape for the flux ratio continuum leads to a comparably good fit , and slightly increases the off - center line emission of @xmath55 . we include this uncertainty in the shape of the continuum at the red end of the k-band in the errors given in table [ tab:2 ] . the statistical precision of the averaged profiles are shown as red errorbars . we compared the individual properties ( flux , visibility , and differential phase ) of the @xmath72(24 .. 28 ) included in @xmath55 to check if meaningful geometric constraints on the pfund line emission zone can be derived from @xmath55 . the decreasing signature in visibility and phase with increasing excitation level is consistent with the @xmath7-n/@xmath7224 line flux ratios . this means that a similar geometry is encoded in the measured visibility and phase of each line , suggesting a common origin of the lines , and justifying the averaging . the individual line ratios used match the respective pfund - line ratio trends measured for other be - star envelopes , such as and @xcite . furthermore , these flux ratios vary by no more than up to about 30 % ( @xmath7 - 28/@xmath7224 ) , which indicates together with a similarly moderate variation of the respective einstein coefficients that the lines included in @xmath55 are emitted under similar conditions . therefore , they most likely originate from similar regions in the disk . the geometric constraints , derived from @xmath55 , can not be more accurate than this level , which matches the numbers , given in table [ tab:2 ] . the precision of the @xmath76-estimate of the individual pfund lines is clearly limited by our spectral resolution and the sampling of the line profile . apparent ratios of up to 20 % above and below unity for the pfund lines included in @xmath55 are regarded as not significantly different from unity . therefore , the lines show comparable @xmath76 . given the previous arguments for a common origin of the individual lines in the disk , we can expected that averaging does improve the @xmath76-signal for the pfund emission zone . we concentrate on answering the key question qualitatively : are ki - spr data suitable to measure directly , and efficiently in a single dataset , core parameters like radial extent and radius - dependence of the often discussed slowly rotating density inhomogeneity in the gas around be stars ? to do so , violet and red velocity bins outside of the line center were defined to derive average v - r properties of the lines , while avoiding the zone of highest absorption and uncertain continuum ( see below ) . the velocity bins enclose the maximum line emission with a width of 150 @xmath26 , which is chosen to be comparable to the spectral resolution ( bins are marked as colored background in fig . [ fig:2]&[fig:3 ] ) . the profile averages for each bin are given in the figures . we derived intrinsic line emission sizes and photo - center shifts from the data ( table [ tab:2 ] ) by assuming the superposition of a linear continuum and additional @xmath6 disk emission . note that the intrinsic photocenter shifts are on the order of the fringe spacing ( @xmath77 ) , which requires the use of the exact complex relation between measured and intrinsic phase to derive the correct _ pc _ @xmath52 . table [ tab:2 ] reports the mean linear properties fitted to the linearly interpolated data for each velocity bin . for our analysis , we used a linear continuum extrapolation of the continuum flux ratio into the line center to derive the quantitative constraints in table [ tab:2 ] . however , the actual continuum at the @xmath6 line center may deviate from this linear continuum . the stellar continuum has some photospheric @xmath6 absorption , and since we observe an edge - on disk in emission with , at least partially , optically thick @xmath6 emission , additional absorption and emission makes it difficult to estimate the exact continuum shape at the line center . this presents a difficulty for isolating the correlated flux , visibility and intrinsic @xmath25 of the line emission at the line center from the measured data , which contain a flux - weighted average of both continuum and line emission . an absorbed continuum profile ( in contrast to the here assumed linear extrapolation of the off - line continuum ) at the center of @xmath6 would increase the ( correlated ) line flux , and therefore decrease the derived intrinsic photocenter shift . to quantify this , we also applied a b4iii @xmath6 template to the data ( see also footnote @xmath78 in table [ tab:11 ] ) . this would not change the measured v / r asymmetries significantly , but it would increase the given @xmath48by about 40 % and decrease _ pc _ @xmath52 by 20% , which gives the order of magnitude of this systematic uncertainty in our quantitative analysis . and @xmath55 , as predicted by , but the here shown pattern is not based on a model calculation . the exact shape of the pattern is not constrained by a single ki - spr dataset . the relative location of the optical @xmath79 line is added for completeness , matching previous , single telescope velocity and _ v / r _ measurements @xcite . [ fig:4 ] ] @xmath6 and @xmath55 profiles differ . above all , they show different v / r ratios and central absorption levels . the total and correlated @xmath6 line profiles are with @xmath80 , and 1.6 respectively , significantly asymmetric . in contrast , @xmath48and _ pc _ @xmath52 of the violet and red @xmath6 emitting regions coincide within the uncertainties . resolved the @xmath6 disk emission of @xmath14 tau with the vlti with a spectral and spatial resolution similar to our data . they found a larger intrinsic line photocenter shift of the brighter wing , emitted , at the time of their observation , from the south - eastern , red part of the disk of @xmath14 tau . a non - lte disk model can explain such a trend in the @xmath25 as well as an extensive multi - wavelength and multi - technique dataset with a global oscillation of a spiral density pattern . our data might show a similar trend of a slightly larger _ pc _ @xmath52 of the brighter side of @xmath6 . however , this trend is not significant with respect to the estimated uncertainties . we would encourage a similar , complex modeling effort of the disk of 48 lib , including the new differential constraints from the pfund emission and based on repeated observations to improve on the significance of the @xmath25 trend of @xmath6 . here we concentrate on the overall line emission zone properties . the bulk of the @xmath6 emission appears to come from similar stello - centric radii , and the enhanced emission on the violet side indicates a locally enhanced density . the observed asymmetry is inline with the @xmath819 yr periodicity of cyclic changes of v / r(@xmath82 ) . the @xmath12 spectra of 48 lib , provided by the bess database also report a v / r @xmath83 in 2008 . in contrast , the @xmath55 profile appears to be symmetric in all properties within the errors . again , @xmath48and _ pc _ @xmath52 of @xmath55 are comparable for both sides , but the smaller _ pc _ points to significantly smaller stello - centric radii of the bulk of the @xmath55 emission , if compared to @xmath6 . both measured v / r(@xmath6 ) and v / r(@xmath55 ) can be reconciled with a one - armed density perturbation , if a radial dependence of the perturbation is allowed , as in the case of a spiral density wave , precessing through the disk . there is several indications for that the pfund emission emerges from inside the bulk of the @xmath6 emission . the mean _ pc _ @xmath52 ( @xmath6 ) is with 2.1 mas about twice as large as _ pc _ @xmath52 ( @xmath55 ) . furthermore , the bins are centered at different stello - centric radii and velocities . the peak velocities appear offset ( @xmath6 : @xmath84@xmath26 versus @xmath55 : @xmath85@xmath26 ) . these velocity and _ pc _ @xmath52 numbers are consistent with keplerian gas motion , a typical assumption for be star disk . the keplerian rotation assumption is further supported by the measured double - peaked line profiles ( in contrast to highly disordered or asymmetric profiles ) . since be - star disks are ( at least partially ) optically thick in the hydrogen emission lines , it is expected that the stellocentric radius of the respective line anti - correlates with the specific line absorption properties . comparing the respective einstein absorption coefficients , we find @xmath86 , which suggests that the @xmath55 is emitted inside of the @xmath6 as described by our data . @xmath6 should be emitted at smaller stellocentric radii than @xmath87 , as indirectly confirmed by the data of @xcite . also the smaller @xmath87 peak velocities ( the spectrum in fig . 1 of * ? ? ? * suggests about 110 km /s ) are consistent with larger stellocentric radii compared to @xmath6 and @xmath55 . another indication for the different location of the pfund line emission are the total and correlated flux profiles . both show a clear absorption below the continuum emission level . this , together with the symmetric line profile might indicate that the gas dominating the pfund emission is located _ in front _ of the inner disk and the star ( responsible for the continuum ) . as discussed , we can not estimate what fraction of this continuum absorption is due to intrinsic photospheric absorption . but a domination of the measured continuum absorption by photospheric absorption is unlikely for a be star . having the majority of the @xmath6 emission at larger stellocentric radii on the violet side , but the pfund emission at smaller radii in front of the star suggests that the over - density pattern follows a radial , one - armed spiral pattern . we depict the radial structure of the disk emission as measured by the ki in a schematic way in fig . [ fig:4 ] . however , our single dataset obviously can not distingiush a spiral wave from other radius - dependent perturbations patterns . our data provide a further quantitative test of papaloizou s density wave model . discuss that the temporal prediction of v / r cycles of about 10 yr periods fits the available spectroscopic observations . interferometric observations directly resolve the expected linear scale of the predicted density waves , as opposed to indirect methods such as polarimetry ( as discussed for instance in * ? ? ? the mean photocenter shifts result in stellocentric radii of about 18 ( 8) @xmath88 for @xmath6 ( @xmath55 ) , which is consistent with both hydrogen emission line modeling of be stars and with density wave models predicting that the local density perturbation , induced by the rotational flattening of the be stars , disappears outside a few stellar radii . we presented the first interferometric detection of several pfund lines at the red end of the @xmath0-band . thanks to the simultaneous detection of @xmath6 in the spectrum of a single ki - spr observation , we can derive stello - centric radius dependent properties the gas disk of the classical be - star 48 lib at high precision . the geometry and kinematics of the emission zones of @xmath6 and of eleven higher pfund lines are clearly detected in the differential visibility and phase signals , and the pfund emission originates from smaller stello - centric radii than @xmath6 . the radial separation of the various emission lines is convincingly explained by the physics of optically thick line emission in the be circumstellar disks . in addition , the @xmath0-band disk continuum emission was resolved as well , and shown to be emitted from inside the @xmath6 ( v / r @xmath89 ) hydrogen line zone , at radii comparable to the bulk of the @xmath7-emission ( with v / r @xmath90 ) . our findings match qualitatively and quantitatively theoretical be - star disk model calculations , predicting a precessing one - armed over - density pattern . the spatial constraints from our interferometric data coincide with the common notion that be - star disks have a keplerian - like radial velocity profile . observations similar to the here presented will shed more light on the enigmas of the creation and properties of circumstellar be shells at all stages of evolution . in particular time - resolved monitoring of be circumstellar environments with ki - spr together with radiative transfer modeling of the disk emission will deliver strong constraints on the actual radial and azimuthal geometry of the disk inhomogeneities likely to be responsible for the hi line emission asymmetries . in addition to the presented scientific results on the disk of 48 lib , we demonstrate with the detection of the pfund lines , that even weak line emission can produce a significant signal in the differential phase . ki - spr data of bright stars can be calibrated to a differential precision of @xmath30 , @xmath31 , and @xmath32 . this extreme precision in the differential phase equals an on - sky centroid shift of 3 @xmath4as ( a relative precision of 10@xmath91 ) , which is an improvement by two orders of magnitude over state - of - the - art single - telescope spectro - astrometry . this improvement is however limited to the relatively small interferometric field of view ( @xmath8130 mas ) and sensitivity of the current ki - spr mode . dual field phase referencing , the second phase of the ki - astra upgrade , is currently under construction , and will allow fainter science targets to be observed . using a different dispersive element , even higher spectral resolution could be achieved in spr mode . we are grateful to w. vacca and a. seifahrt for many useful discussions and contributions to this work . the excellent support of the ki team at wmko and nexsci helped making these observations a success . the realization of the ki - astra upgrade is supported by the nsf mri grant , ast-0619965 . the data presented herein were obtained at the w.m . keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california and the national aeronautics and space administration . the observatory was made possible by the generous financial support of the w.m . keck foundation . the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . the keck interferometer is funded by the national aeronautics and space administration as part of its exoplanet exploration program . this research has made use of the simbad database , operated at cds , strasbourg , france . this research has made use of nasa s astrophysics data system bibliographic services . this work has made use of the bess database , operated at gepi , observatoire de meudon , france . carciofi , a. c. , okazaki , a. t. , le bouquin , j .- b . , tefl , s. , rivinius , t. , baade , d. , bjorkman , j. e. , & hummel , c. a. 2009 , , 504 , 915 chesneau , o. , et al . 2005 , , 435 , 275 colavita , m. m. , et al . 1999 , , 510 , 505 cranmer , s. r. 2005 , , 634 , 585 curtiss , r. h. 1925 , popular astronomy , 33 , 537 domiciano de souza , a. , kervella , p. , jankov , s. , abe , l. , vakili , f. , di folco , e. , & paresce , f. 2003 , , 407 , l47 dougherty , s. m. , waters , l. b. f. m. , burki , g. , cote , j. , cramer , n. , van kerkwijk , m. h. , & taylor , a. r. 1994 , , 290 , 609 hanuschik , r. w. 1989 , , 161 , 61 hanuschik , r. w. , hummel , w. , dietle , o. , & sutorius , e. 1995 , , 300 , 163 hanuschik , r. w. , & vrancken , m. 1996 , , 312 , l17 hony , s. , et al . 2000 , , 355 , 187 hummel , w. , & vrancken , m. 2000 , , 359 , 1075 hummel , w. , & hanuschik , r. w. 1994 , pulsation ; rotation ; and mass loss in early - type stars , 162 , 382 mcdavid , d. 1999 , , 111 , 494 mourard , d. , bosc , i. , labeyrie , a. , koechlin , l. , & saha , s. 1989 , , 342 , 520 millour , f. , vannier , m. , petrov , r. g. , lopez , b. , & rantakyr , f. 2006 , iau colloq . 200 : direct imaging of exoplanets : science & techniques , 291 papaloizou , j. c. , savonije , g. j. , & henrichs , h. f. 1992 , , 265 , l45 pott , j .- u . , woillez , j. , akeson , r. l. , berkey , b. , colavita , m. m. , cooper , a. , eisner , j. a. , ghez , a. m. , graham , j. r. , hillenbrand , l. , hrynewych , m. , medeiros , d. , millan - gabet , r. , monnier , j. , morrison , d. , panteleeva , t. , quataert , e. , randolph , b. , smith , b. , summers , k. , tsubota , k. , tyau , c. , weinberg , n. , wetherell , e. , wizinowich , p. l. , 2008 , arxiv:0811.2264 , _ astrometry with the keck - interferometer : the astra project and its science _ , proceedings of the summerschool `` astrometry and imaging with the very large telescope interferometer '' , 2 - 13 june , 2008 , keszthely , hungary , in press shao , m. , colavita , m. m. , hines , b. e. , staelin , d. h. , & hutter , d. j. 1988 , , 193 , 357 shao , m. , & colavita , m. m. 1992 , , 262 , 353 skrutskie , m. f. , et al . 2006 , , 131 , 1163 slettebak , a. 1982 , , 50 , 55 stee , p. , & bittar , j. 2001 , , 367 , 532 tefl , s. , et al . 2009 , , 504 , 929 weigelt , g. , et al . 2007 , , 464 , 87 wisniewski , j. p. , kowalski , a. f. , bjorkman , k. s. , bjorkman , j. e. , & carciofi , a. c. 2007 , , 656 , l21 wisniewski , j. p. , draper , z. h. , bjorkman , k. s. , meade , m. r. , bjorkman , j. e. , & kowalski , a. f. 2010 , , 709 , 1306 wizinowich , p. , graham , j. , woillez , j. , et al . nsf - mri award abstract # 0619965 woillez , j. , et al . 2010 , pasp , submitted

we report on the successful science verification phase of a new observing mode at the keck interferometer , which provides a line - spread function width and sampling of 150 km / s at @xmath0-band , at a current limiting magnitude of @xmath1 mag with spatial resolution of @xmath2 and a measured differential phase stability of unprecedented precision ( 3 mrad at @xmath3 , which represents 3 @xmath4as on sky or a centroiding precision of @xmath5 ) . the scientific potential of this mode is demonstrated by the presented observations of the circumstellar disk of the evolved be - star 48 lib . in addition to indirect methods such as multi - wavelength spectroscopy and polaritmetry , the here described spectro - interferometric astrometry provides a new tool to directly constrain the radial density structure in the disk . we resolve for the first time several pfund emission lines , in addition to @xmath6 , in a single interferometric spectrum , and with adequate spatial and spectral resolution and precision to analyze the radial disk structure in 48 lib . the data suggest that the continuum and @xmath7-emission originates in significantly more compact regions , inside of the @xmath6 emission zone . thus , spectro - interferometric astrometry opens the opportunity to directly connect the different observed line profiles of @xmath6 and pfund in the total and correlated flux to different disk radii . the gravitational potential of a rotationally flattened be star is expected to induce a one - armed density perturbation in the circumstellar disk . such a slowly rotating disk oscillation has been used to explain the well known periodic v / r spectral profile variability in these stars , as well as the observed v / r cycle phase shifts between different disk emission lines . the differential line properties and linear constraints set by our data are consistent with theoretical models and lend direct support to the existence of a radius - dependent disk density perturbation . the data also shows decreasing gas rotation velocities at increasing stello - centric radii as expected for keplerian disk rotation , assumed by those models .

rare @xmath5 decays mediated via flavour changing neutral currents have received much attention because of their sensitivity to physics beyond the standard model . in the standard model these decays occur via penguin and box diagrams with virtual electroweak bosons and up - type quarks in the loops . because of the large top quark mass , the contribution with a top quark in the loop dominates . at energy scales below the mass of the top quark and @xmath6 boson , it is convenient to switch to an effective theory where the top quark and the weak bosons have been integrated out of the theory . the @xmath7 transition is then mediated by the effective hamiltonian @xcite @xmath8 where the operators are commonly defined by @xmath9 here @xmath10 are the usual left and right handed chiral projection operators . the values of the wilson coefficients @xmath11 have been calculated in the next to leading log approximation @xcite in the standard model and are given in table [ wilson_coefficients ] . [ wilson_coefficients ] .the wilson coefficients @xmath11 in the next - to - leading log approximation . [ cols="^,^,^,^,^,^,^,^,^ " , ] physics beyond the standard model will generally introduce new contributions to the loop and will therefore modify the values of these coefficients @xcite . measurements of the wilson coefficients may therefore indirectly constrain new physics scenarios . for example , the decay @xmath12 is proportional to @xmath13 , and the recent measurements of the branching ratio @xmath14 @xcite and inclusive rate @xmath15 @xcite have placed constraints on models of physics beyond the standard model which modify the magnitude of @xmath16 @xcite . the decay @xmath17 is suppressed , relative to @xmath18 , by an additional factor of the electromagnetic coupling constant and has not yet been observed @xcite . it has , however , the appeal of being sensitive to the signs and magnitudes of @xmath16 , @xmath19 , and @xmath20 , making it a potentially more powerful probe than @xmath18 of beyond the standard model physics . experimental studies of this process impose cuts on the available phase space . this is primarily due to the necessity of removing the resonance from @xmath21 with the @xmath22 decaying into two leptons . we incorporate representative cuts into the theoretical analysis . using an operator product expansion ( ope ) several observables of the inclusive decay @xmath0 have been calculated including the leading non - perturbative corrections @xcite . in this framework these leading corrections arise as matrix elements of dimension five operators , suppressed by two powers of the @xmath23 quark mass , and are conventionally parameterized by two quantities , @xmath3 and @xmath24 . a third parameter @xmath4 enters through the difference of the @xmath23 quark mass and @xmath5 meson mass @xmath25 whereas @xmath24 can be determined from the @xmath26 mass splitting , @xmath27 , no such simple relation exists for @xmath3 . it has been estimated using various methods to lie in the range @xmath28 @xcite . in a previous paper we extended the analysis of the total rate for @xmath29 to one order higher in the ope @xcite . the dimension six operators arising at this order can be parametrized by six quantities , commonly labelled @xmath30 and @xmath31 , all of which are unknown . we found that the uncertainties introduced by these six parameters can be significant , depending primarily on the actual values of the matrix elements and the amount of accessible phase space . in this paper we give the details of that analysis and also present calculations for the forward - backward asymmetry and moments of the hadron invariant mass spectrum at @xmath1 . as in our previous analysis , we neglect perturbative effects and effects due to the finite mass of the @xmath32 quark , which have been considered elsewhere @xcite . it has also been proposed that , rather than use this decay to search for new physics , it might instead be used to extract the parameters @xmath4 and @xmath3 through a measurement of its hadronic invariant mass moments @xcite . we estimate the uncertainties in this extraction due to the unknown matrix elements of the dimension six operators . this paper is organized as follows . in section [ section_formalism ] we briefly introduce the formalism used to calculate the nonperturbative corrections , and we present the results for the decay rate in section [ section_partial ] . in section [ section_fbasym ] we calculate the forward - backward asymmetry of the lepton pair . we then proceed in section [ section_moments ] to calculate moments of the hadronic invariant mass spectrum and estimate uncertainties in extracting @xmath4 and @xmath3 from these moments . finally we discuss the results and state our conclusions . the procedure for calculating nonperturbative contributions to heavy hadron decays has been thoroughly discussed in the literature @xcite , and we present here only a brief outline of the technique . the differential rate is proportional to the product of a lepton tensor @xmath33 and a hadron tensor @xmath34 and for the process in question it may be written as @xmath35 where @xmath36 denotes the three body phase space . the spin - summed tensor @xmath33 for massless leptons is @xmath37.\ ] ] the hadron tensor @xmath34 is related via the optical theorem to the imaginary part of the forward scattering matrix element @xmath38 where @xmath39 in this equation @xmath40 denotes the current mediating this transition , and is given by @xmath41 b\ ] ] where @xmath42 is the dilepton momentum reduces to the @xmath43 current when @xmath44 , @xmath45 , and @xmath46 . this provides some useful cross - checks with known results for semileptonic @xmath5 decays @xcite . ] . in accordance with convention , we have defined two effective wilson coefficients : @xmath47 and @xmath48 . the latter contains the operator mixing of @xmath49 into @xmath50 as well as the one loop matrix elements of @xmath51 @xcite . the full analytic expression for @xmath48 is quite lengthy and may be found in @xcite . since in the decay of a @xmath23 quark the momentum transfer to the final state parton is large , the time ordered product ( [ discontinuity ] ) can be expanded in terms of local operators @xcite @xmath52 , \label{ope}\ ] ] where @xmath53 represents a set of local operators of dimension @xmath54 , each operator containing @xmath55 derivatives . for a generic current @xmath56 the expressions for these operators are quite lengthy . the complete set of operators for @xmath57 @xcite and @xmath58 @xcite appear in the literature . in this study we include operators up to and including dimension @xmath58 . the standard lorentz decomposition for the forward scattering amplitude is @xmath59 where @xmath60 is the four velocity of the initial @xmath23 quark @xmath61 . since in this paper we treat the final state leptons as massless @xmath62 , the form factors @xmath63 do not contribute to observables . it is clear from ( [ discontinuity ] ) and ( [ schematicope ] ) that to calculate these form factors we must take matrix elements of the operators @xmath53 . matrix elements of dimension four operators vanish at leading order in the @xmath64 expansion @xcite and matrix elements of dimension five operators may be parameterized by @xmath3 and @xmath24 @xcite @xmath65 where @xmath66 , and @xmath67 is an arbitrary dirac structure . finally , the dimension six operators may be parameterized by the matrix elements of two local operators @xcite @xmath68 and by matrix elements of two time ordered products @xmath69 arising from a mismatch between the states @xmath70 of the effective theory and @xmath71 of the full theory . the contributions from @xmath31 can most easily be incorporated by making the replacements @xcite @xmath72 in the parton level results . in addition , as we will show later there is a contribution to the total rate from the dimension six four fermion operator @xmath73,\ ] ] the matrix element of which we define as @xmath74 the form factors up to @xmath75 have appeared in the literature @xcite . the @xmath1 contributions to the form factors proportional to @xmath30 are presented in appendix [ formfactorappendix ] . the dependence on @xmath31 is obtained by making the replacements ( [ tsubs ] ) in the @xmath75 form factors . the triple differential branching ratio is given by @xmath76 where we have defined kinematic variables @xmath77 , @xmath78 , and @xmath79 $ ] . in terms of these leptonic variables the limits of phase space are given by @xmath80 where @xmath81 \left[\hat{s}-(1-\hat{m}_s)^2 \right]}$ ] . for the calculation of the hadron invariant mass moments it will be convenient to express the phase space in terms of the parton energy fraction @xmath82 and the parton invariant mass fraction @xmath83 . they are related to the leptonic variables introduced above via @xmath84 the phase space can then be expressed as @xmath85 since the form factors @xmath86 are independent of @xmath87 , this first integration is trivial and we arrive at @xmath88 \right\}\end{aligned}\ ] ] in the above expressions we use the same conventions as in @xcite and normalize the @xmath89 branching ratio to the semileptonic branching ratio @xmath90 introducing the normalization constant @xmath91 in this expression @xmath92 is the well - known phase space factor for the @xmath93 parton decay rate @xmath94 and @xmath95 includes the @xmath96 qcd corrections as well as the nonperturbative corrections up to @xmath97 @xmath98 where @xmath99 the analytic expression for the perturbative function @xmath100 can be found in @xcite . an interesting experimentally accessible quantity is the dilepton invariant mass spectrum . evaluating the @xmath87 integral in ( [ triplelepdiff ] ) , and doing the integral over @xmath101 by picking out the residues , we find for the dilepton invariant mass spectrum @xmath102 \left ( |c_9^{\mbox{eff}}(\hat{s } ) |^2 + c_{10}^2 \right ) \right.\nonumber \\ & & \left . + \left [ \frac{4}{3 } ( 1-\hat{s})^2 ( 2 + \hat{s } ) \ ; ( 2 + \frac{\lambda_1}{m_b^2 } ) + 4 \left ( -6 -3 \hat{s } + 5 \hat{s}^3 \right ) ( \frac{\lambda_2}{m_b^2 } - \frac{\rho_2}{m_b^3 } ) \right . \right.\nonumber\\ & & \left.\left.\;\;\;\;\;\;\;\;\;\;\;\;\;\ ; - \frac{4(5 \hat{s}^4 + 19 \hat{s}^3 + 9 \hat{s}^2 - 7 \hat{s } + 22)}{9(1-\hat{s } ) } \frac{\rho_1}{m_b^3}\right ] \frac{|c_7^{\mbox{eff}}|^2}{\hat{s } } \right.\nonumber \\ & & \left . + \left [ 4 ( 1-\hat{s})^2 ( 2 + \frac{\lambda_1}{m_b^2 } ) + 4 \left ( -5 -6 \hat{s } + 7 \hat{s}^2 \right ) ( \frac{\lambda_2}{m_b^2 } - \frac{\rho_2}{m_b^3 } ) \right . \nonumber\\ & & \left.\left.\;\;\;\;\;\;\;\;\;\;\;\;\;\ ; + \frac{4 ( 3 \hat{s}^3 - 17 \hat{s}^2 + \hat{s } - 3)}{3(1-\hat{s } ) } \frac{\rho_1}{m_b^3 } \right ] re(c_9^{\mbox{eff}}(\hat{s } ) ) \ , c_7^{\mbox{eff } } \right . \nonumber\\ & & \left . - \frac{16}{3 } \frac{\rho_1}{m_b^3 } \delta(1-\hat{s } ) \left(c_{10}^2 + \left(c_9^{\mbox{eff } } ( \hat{s})+2 c_7^{\mbox{eff}}\right)^2 \right ) \right\}\ , . \label{eqn : dbds0}\end{aligned}\ ] ] the dependence on @xmath31 can be obtained by making the replacements ( [ tsubs ] ) in ( [ eqn : dbds0 ] ) . in this expression we have taken the limit @xmath103 . the corresponding expression with full @xmath104 dependence is given in appendix [ specwithmass ] . = 10 cm a plot of this distribution is shown in fig . [ diff_spec ] , where we have used the values for the nonperturbative matrix elements @xmath105 for the matrix elements of the dimension six operators we use the generic size @xmath106 as suggested by dimensional analysis . the vacuum saturation approximation @xcite predicts @xmath107 , as shown , and we find the displayed spectrum is fairly insensitive to the sign of the other dimension six matrix elements . one immediately notices divergences at both endpoints of this spectrum . the divergence at the @xmath108 endpoint is due to the intermediate photon going on shell and is a well known feature of the decay @xmath0 @xcite . in this limit one expects this spectrum to reduce to the @xmath15 rate with an on shell photon in the final state , convoluted with the fragmentation function giving the probability for a photon to fragment into a lepton pair . this correspondence is explicitly verified by the analytic form of the divergent term @xmath109 where the term multiplying @xmath110 is proportional to the total rate for @xmath15 @xcite . as mentioned above , experimental cuts require us to stay away from this endpoint , and therefore automatically regulate this divergence . the divergence at the @xmath111 endpoint is entirely due to the @xmath112 operators as can be seen from fig . [ diff_spec ] . in this case the analytic form of the divergent term is @xmath113 this leads , upon integration , to an unphysical logarithmic divergence in the expression for the total rate that is regulated by the mass of the @xmath32 quark . ( of course , it is only consistent to include the mass of the @xmath32 quark in the upper limit of integration if one uses the spectrum with the full @xmath114 dependence as given in appendix [ specwithmass ] . ) this divergence can be understood by considering a similar effect in the semileptonic decay @xmath115 @xcite . in that context , the origin of this divergence can be clarified by performing an ope for the total , rather than the differential , rate @xcite . including dimension six operators in this ope one finds a four fermion operator of the form @xmath116 contributing to the rate . in @xcite the matrix element of this operator was calculated at leading order in perturbation theory by integrating out the @xmath117 quark and its contribution to the total rate was found to be @xmath118 . to calculate this matrix element it was essential that the mass of the @xmath117 quark be large compared to the qcd scale @xmath119 . consequently , for the decay @xmath0 where the same operator with @xmath32 quarks rather than @xmath117 quarks appears , these methods are not applicable because the @xmath32 quark is too light . including higher orders in perturbation theory the matrix element of the four fermion operator contains logarithms of the form @xmath120 which are of order unity , making a perturbative calculation of this matrix element impossible . thus , a seventh non - perturbative matrix element @xmath121 defined in ( [ f1definition ] ) is required . it contributes only at the @xmath111 endpoint of the spectrum and cancels the logarithmic divergence proportional to @xmath122 in the total rate @xmath123 another noticeable feature in the dilepton invariant mass spectrum is the cusp due to the @xmath124 threshold . near this value of @xmath125 the methods we used to calculate the physical spectrum fail because of long distance contributions from the resonant decay @xmath126 , where the @xmath127 subsequently decays into two leptons . experimentally one deals with this resonance region by simply cutting it out . thus , to compare reliably to experiment we should include such a cut in our calculation . defining the partially integrated branching ratio by @xmath128 = 10 cm we plot the contribution of the individual matrix elements relative to the leading order parton result in fig . [ cut_plot ] . for the generic size @xmath129 used in this plot , the contribution from @xmath130 is of the same size as the contribution from dimension five operators . this implies that including the @xmath75 corrections for this decay does not significantly decrease the nonperturbative uncertainties . we see that the nonperturbative contributions become more dominant as the accessible phase space is decreased contributions shown here should not be taken as accurate indications of the actual size of the corrections , but rather as estimates of the uncertainty in the prediction . ] . for @xmath131 the uncertainty from the @xmath130 matrix element is of the same size as the parton model prediction . this is a clear signal that the ope is no longer valid if the phase space is restricted to be too close to the endpoint @xmath132 . this breakdown of the ope close to the endpoint is a well known feature encountered in this approach to the study of inclusive decays @xcite . unfortunately , in this endpoint region a shape function does not exist and an alternate approach , such as heavy hadron chiral perturbation theory for exclusive final states , must be used @xcite . a cut of @xmath133 has been suggested by the cleo collaboration in order to eliminate the resonance region @xcite . for this value the partially integrated rate is @xmath134 at this value of the cut @xmath135 the coefficients of the nonperturbative matrix elements clearly indicate a poorly converging ope . one can estimate the uncertainty induced by the @xmath1 parameters by fixing @xmath136 to the values given in ( [ parameter_set ] ) , then randomly varying the magnitudes of the parameters @xmath137 and @xmath121 between @xmath138 and @xmath139 as suggested by dimensional analysis . we also impose positivity of @xmath130 as indicated by the vacuum saturation approximation @xcite , and we enforce the constraint @xmath140 derived from the ground state meson mass splittings @xmath141 @xcite . here @xmath142 is the usual coefficient of the beta function @xmath143 for @xmath144 light flavours . taking the 1 @xmath145 deviation as a reasonable estimate of the uncertainties from @xmath1 contributions , we find the uncertainty in @xmath146 to be at the 10% level . it is clear from ( [ cleobranchingratio ] ) that the @xmath130 contribution is large , and relaxing the positivity constraint on @xmath130 enlarges the uncertainty to about 20% . a similar statement can be made regarding the @xmath147 analysis @xcite where the phase space cut is slightly higher , and the nonperturbative corrections are correspondingly somewhat larger . since the cut on @xmath125 can not be lowered because of the @xmath148 resonance , these uncertainties are intrinsic to our approach in the large dilepton invariant mass region . it is important to notice that in the invariant mass region below the @xmath127 resonance , the uncertainties from these matrix elements are much smaller . for example , integrating the differential spectrum up to the cut specified in the cleo analysis @xcite @xmath149 we find @xmath150 it is still true that the coefficient of the @xmath130 term is @xmath151 times larger than that of the @xmath3 term , but both sets of nonperturbative corrections are small relative to the parton level result in this region . although this does not allow us to draw a strong conclusion about the convergence of the ope , we can conclude that in this region the @xmath1 nonperturbative corrections are not a significant source of theoretical uncertainty . the differential forward - backward asymmetry is defined by @xmath152 where @xmath153 parameterizes the angle between the @xmath23 quark and the @xmath154 in the dilepton cm frame . it has been shown @xcite that new physics can modify this spectrum , so it is interesting to see how @xmath1 terms contribute to the sm prediction . integrating the triple differential decay rate ( [ triplelepdiff ] ) we find @xmath155 here we have again omitted the trivial dependence on @xmath31 . it is clear from this expression that the third order terms do not have abnormally large coefficients , and therefore introduce only small variations relative to the second order expression . an experimentally more useful quantity is the normalized fb asymmetry defined by @xmath156 unfortunately , this normalized asymmetry has inherited the poor behavior of the differential branching ratio in the endpoint region . in fig . [ asymm ] = 10 cm we illustrate the uncertainties of the normalized fb asymmetry originating from the matrix elements of the dimension six operators . the three curves show the mean value and the @xmath157 uncertainty of the forward backward asymmetry , obtained in a way similar to that explained in section [ section_partial ] . we can see that up to a value of @xmath158 the uncertainties are small , but for larger values of the dilepton invariant mass the uncertainties increase rapidly . because of the necessity of the cut to eliminate the @xmath124 resonances , the accessible high dilepton invariant mass region is therefore restricted to a few hundred mev . for the dilepton invariant mass region below the @xmath127 resonance @xmath159 the uncertainties are small . throughout this paper we have fixed the values of the leading non - perturbative parameters @xmath4 , @xmath3 , and @xmath24 . however , these values must be determined from experiment . the most sensitive observables for this purpose are those which vanish in the parton model . it is interesting to ask how severely our ignorance of the values of the @xmath1 parameters compromises our ability to extract the values of @xmath4 and @xmath3 from such a measurement . it has been suggested by ali and hiller @xcite that one use the first two moments of the hadron invariant mass spectrum defined by @xmath160 this idea is similar to the approaches used in the semileptonic @xmath161 @xcite and the rare radiative @xmath15 decays @xcite , though the experimental task is more difficult in this case due to the small size of the branching ratio . to calculate these hadronic moments we relate them to calculable partonic moments via @xmath162 where we have used the mass relation @xmath163 appropriate at this order in the ope . we therefore have to calculate the first two moments of the parton energy @xmath164 and parton invariant mass @xmath165 , as well as the mixed moment @xmath166 . defining @xmath167 we give the results for the required partonic moments in appendix [ parton_moments ] . as before , we have included the dependence on the cut on the lepton invariant mass in these results . it is important to note that the results for the partonic moments given in appendix [ parton_moments ] are expressed in terms of the @xmath23 quark mass @xmath168 and must be re - expressed in terms of the @xmath5 meson mass @xmath169 using the mass relation ( [ mass_relation ] ) . using again for the cut on the invariant mass the value proposed by cleo @xmath170 , we find for the two moments @xmath171\\ \langle s_h^2 \rangle & = & m_b^4 \left [ - 0.05 \frac{\lambda_1}{m_b^2 } + 0.14 \frac{\bar{\lambda}^2}{m_b^2 } - 0.53 \frac{\rho_1}{m_b^3 } - 0.21 \frac{\rho_2}{m_b^3 } + 0.63 \frac { \bar{\lambda } \lambda_1}{m_b^3 } + 0.46 \frac { \bar{\lambda } \lambda_2 } { m_b^3 } \right . \nonumber \\ & & \qquad\qquad \left . - 0.05 \frac{{\cal t}_1 + 3 { \cal t}_2}{m_b^3 } \right]\end{aligned}\ ] ] consider first the expression for @xmath172 . the @xmath3 term has a small coefficient and tends to cancel against higher order corrections , making this moment particularly insensitive to @xmath3 . we can see the problem another way by solving this equation for @xmath3 : the solution exhibits a pole near @xmath173 , close to the expected value of @xmath4 @xcite . as a result , the extracted value of @xmath3 is extremely sensitive to the values of the higher order parameters . since the presence of this pole persists as the value of the cut is changed , we conclude that this observable is unsuitable for extracting @xmath3 . for the first moment @xmath174 the convergence of the ope is much better . estimating the uncertainties from the unknown values of the dimension six operators by the method explained in section [ section_partial ] , we present the resulting constraint in the @xmath175 plane in fig . [ band ] . = 10 cm superimposed on this figure is the ellipse obtained in @xcite from an analogous study of moments of @xmath176 . unfortunately , the bound from our analysis is nearly parallel to the major axis of this ellipse and , since it is only the relative orientation of the constraints which has meaning in this figure , this moment does not provide much additional information about the values of @xmath4 or @xmath3 . our purpose in this paper has been to study nonperturbative uncertainties in the rare inclusive decay @xmath0 . building on previous studies which evaluated the leading nonperturbative corrections , we have parameterized the corrections arising at @xmath1 in terms of two matrix elements of local operators @xmath177 , four matrix elements of non local operators @xmath31 , and one matrix element of a four fermion operator @xmath121 . the numerical values of these parameters are unknown , yet even so a knowledge of the analytic form of the corrections allows us to study the convergence properties of the operator product expansion in various regions of phase space . furthermore , the assumption that these parameters , being of nonperturbative origin , should be @xmath178 permits us to make numerical estimates of theoretical uncertainties in observable quantities . we first considered the corrections to the differential spectrum @xmath179 . the experimental spectrum contains two prominent resonances due to intermediate @xmath127 and @xmath148 production , and the necessity of cutting these resonances out divides the accessible spectrum into two parts : the region of low dilepton invariant mass below the @xmath127 resonance , and the region of high dilepton invariant mass above the @xmath148 resonance . in the first region , we find that the parton level calculation dominates , and that nonperturbative corrections are small . the operator product expansion appears to be converging according to expectation and , should it be possible to take experimental data in this region , the results will not suffer from significant nonperturbative uncertainties . on the other hand we do find that , as expected , the nonperturbative uncertainties increase as one moves into the high dilepton invariant mass region . it is well known that as one approaches this endpoint , the operator product expansion breaks down . the interesting result of our study , however , is that the expansion breaks down somewhat earlier than was anticipated , with the @xmath1 uncertainties coming to dominate the integrated rate once the available range of @xmath125 is reduced to about one quarter of the full range . we also showed that the rate obtained by integrating over the entire region above the @xmath148 resonance contains uncertainties from dimension six matrix elements at the @xmath180 level , the uncertainties being dominated by those from the @xmath130 matrix element . this result may impact the potential for doing precise searches for new physics using data from this region of phase space . we also studied the contributions from dimension six operators to the forward - backward asymmetry . this quantity probes different combinations of wilson coefficients than the rate , and has been proposed as a complimentary source of information about possible new physics effects . as with the rate , the spectrum @xmath181 is divided into two regions by the @xmath127 and @xmath148 resonances . we find that the dimension six contributions are not unduly large anywhere in the phase space , suggesting that this observable has a well behaved ope . however if the differential forward - backward asymmetry is normalized to the differential rate , the resulting spectrum contains large uncertainties in the high dilepton invariant mass region . in the region of phase space below the @xmath127 resonance , however , we find the nonperturbative corrections to be small . finally we addressed a recent proposal suggesting that hadron invariant mass moments of the differential spectrum for @xmath0 could , due to the sensitivity of these moments to nonperturbative effects , be used to constrain the values of the hqet parameters @xmath3 and @xmath4 . in the low @xmath182 region the moments are suppressed relative to the rate and , considering the already tiny branching ratio , it is unlikely that experimental measurements in this region will be forthcoming . therefore we focused our attention on the high invariant mass region . in this region , we found that the first of these moments @xmath174 provided a constraint in the @xmath175 plane , but that this constraint was nearly the same as those derived from other , more experimentally promising , processes , and therefore seems to be of limited interest for this purpose . as for the second invariant mass moment @xmath172 , we found that the nonperturbative uncertainties were such that it was not possible to extract a stable constraint on the values of @xmath4 or @xmath3 . from these results , we conclude that these moments are not well suited to the extraction of these parameters . we would like to thank michael luke for helpful discussions . this work was supported in part by the national science and engineering research council of canada . in this appendix we present the form factors @xmath86 . these form factors have been calculated previously up to @xmath75 @xcite , and we do not reproduce those results here . we decompose the new contributions arising at @xmath97 as @xmath183 for completeness we have included the full @xmath104 dependence in these expressions , though in our analysis we set @xmath184 . defining @xmath185 with @xmath186 , @xmath187 and @xmath188 , , we find that the third order contributions are @xmath189 in this appendix we present the dilepton invariant mass spectrum for a finite @xmath32-quark mass @xmath114 . the spectrum originating from operators of dimension @xmath57 has been presented in eq . ( 47 ) of @xcite . the contributions from the time - ordered operators @xmath31 can be obtained by making the replacement ( [ tsubs ] ) in this equation . since the dilepton invariant mass distribution is independent of the definition of the four velocity of the heavy quark , the contribution proportional to @xmath190 is related by reparameterization invariance @xcite to the @xmath24 contribution . it can be obtained by the replacement @xmath191 in the results of @xcite . thus , the only term we have to add to the existing literature to obtain the complete expression including all @xmath112 contributions is the term originating from the darwin operator whose matrix element is @xmath130 . this contribution is given by @xmath192_*\ , \nonumber \\ & + & i_\delta \ , \delta(\hat{s}_u - \hat{s } ) \bigg]\,,\end{aligned}\ ] ] with limits of integration defined in ( [ lepton_phase ] ) @xmath193 with @xmath194 the function @xmath195 is singular at the upper limit of integration . to regulate this divergence we defined a `` star function '' @xmath196 _ * = \lim_{\beta\to 0 } \left\{f(\hat{s } , \hat{m}_s ) \theta(\hat{s}_u - \hat{s } - \beta ) - \delta ( \hat{s}_u- \hat{s } - \beta ) \int_{\hat{s}_l}^{\hat{s}_u - \beta } \!\!\!\!\!\!\!d \hat{s } \,f(\hat{s } , \hat{m}_s ) \right\}.\ ] ] this `` star function '' is analogous to the common plus distribution . the functions appearing above are @xmath197\\ % i^{c_7 ^ 2 } & = & - \frac{8}{9 \hat{s } } \bigg [ { \left ( 1 - \hat{s } \right ) } ^2\ , \left ( 22 - 7\,\hat{s } + 9\,{\hat{s}}^2 + 19\,{\hat{s}}^3 + 5\,{\hat{s}}^4 \right ) \nonumber\\ & & - \left ( 78 - 34\,\hat{s } - 105\,{\hat{s}}^2 + 376\,{\hat{s}}^3 - 40\,{\hat{s}}^4 + 18\,{\hat{s}}^5 - 5\,{\hat{s}}^6 \right ) \ , { \hat{m}_s}^2 \nonumber\\ & & + \left ( 70 + 67\,\hat{s } - 222\,{\hat{s}}^2 - 188\,{\hat{s}}^3 - 32\,{\hat{s}}^4 - 15\,{\hat{s}}^5 \right ) \ , { \hat{m}_s}^4 \nonumber\\ & & + \left ( 50 + 60\,\hat{s } + 258\,{\hat{s}}^2 + 136\,{\hat{s}}^3 \right ) \ , { \hat{m}_s}^6 - \left ( 110 + 157\,\hat{s } + 111\,{\hat{s}}^2 - 50\,{\hat{s}}^3 \right ) \ , { \hat{m}_s}^8 \nonumber\\ & & + \left ( 38 + 2\,\hat{s } - 75\,{\hat{s}}^2 \right ) \,{\hat{m}_s}^{10 } + 9\,\left ( 2 + 5\,\hat{s } \right ) \,{\hat{m}_s}^{12 } - 10\,{\hat{m}_s}^{14}\bigg]\\ % i^{c_7 c_9 } & = & - \frac{8}{3 } \bigg [ { \left ( 1 - \hat{s } \right ) } ^2\ , ( 3 - \hat{s } + 17\,{\hat{s}}^2 - 3\,{\hat{s}}^3 ) \nonumber\\ & & - \left ( 10 - 13\,\hat{s } + 56\,{\hat{s}}^2 + 58\,{\hat{s}}^3 - 10\,{\hat{s}}^4 - 5\,{\hat{s}}^5 \right ) \,{\hat{m}_s}^2 \nonumber\\ & & + \left ( 5 - 22\,\hat{s } + 12\,{\hat{s}}^2 - 34\,{\hat{s}}^3 - 25\,{\hat{s}}^4 \right ) \ , { \hat{m}_s}^4 + 2\,\left ( 10 + 29\,\hat{s } + 36\,{\hat{s}}^2 + 25\,{\hat{s}}^3 \right ) \ , { \hat{m}_s}^6 \nonumber\\ & & - \left ( 35 + 67\,\hat{s } + 50\,{\hat{s}}^2 \right ) \,{\hat{m}_s}^8 + \left ( 22 + 25\,\hat{s } \right ) \,{\hat{m}_s}^{10 } - 5\,{\hat{m}_s}^{12}\bigg]\\ i_\delta & = & - \frac{16(1-\hat{m}_s^2)^5}{3 \sqrt{1 - 4\hat{m}_s^2}}\left(c_{10}^2 + \left(c_9^{\mbox{eff } } ( \hat{s})+2 c_7^{\mbox{eff}}\right)^2 \right ) \end{aligned}\ ] ] notice that these terms correctly reproduce the expression ( [ eqn : dbds0 ] ) in the limit @xmath103 . we write the moments in the form @xmath198 where @xmath199 is given in eq . ( [ bchi ] ) . the coefficient @xmath200 depends on the parameters @xmath201 and @xmath202 as explained in section [ section_formalism ] , so we express the moments @xmath203 and @xmath204 as integrals which we evaluate numerically , @xmath205 \bigg\ } |c_9^{\mbox{eff}}(x_0,\hat{s}_0)|^2 \bigg ] \nonumber\\ m^{(m , n)}_{7,9 } & = & \frac{16}{2m_b } \left(-\frac{1}{\pi } \right ) \int_{\hat{m}_s}^{\frac{1}{2}(1-\chi ) } \ , dx_0 \int_{\hat{m}_s^2}^{x_0 ^ 2 } \ , d\hat{s}_0 \ , x_0^m\ , \hat{s}_0^n \bigg [ \sqrt{x_0 ^ 2 - \hat{s}_0 } \;\mbox{im } \bigg\ { \bigg [ 2 ( 1 - 2 x_0 + \hat{s}_0 ) t_1^{c_7 ( c_9\pm c_{10 } ) } \nonumber \\ & & \qquad + \frac{x_0 ^ 2 - \hat{s}_0}{3 } t_2^{c_7 ( c_9\pm c_{10 } ) } \bigg ] \bigg\}re\left [ c_9^{\mbox{eff}}(x_0,\hat{s}_0)\right ] \bigg ] \end{aligned}\ ] ] for the other contributions we find @xmath206 99 b.grinstein , m.savage , and m.b.wise , . m. misiak , ; * erratum * , . a.j . buras and m.mnz , . a.ali , g.f.giudice , and t.mannel , z.phys.*c67 * ( 1995 ) 417 ; c.huang , w.liao , and q.yan , ; p.cho , m.misiak , and d.wyler , ; j.hewett and j.d.wells , ; t.goto , y.okada , y.shimizu , and m.tanaka , ; e.lunghi , a.masiero , i.scimemi , and l.silvestrini , hep - ph/9906286 . r. ammar _ et al . _ ( cleo collaboration ) , . m.s . alam _ ( cleo collaboration ) , phys . * 74 * ( 1995 ) 2885 ; hep - ph/9908022 . j.hewett , hep - ph/9406302 ; t.g.rizzo , ; f.borzumati and c.greub , ; w.de boer _ et al . _ , ; h.baer _ et al . _ , ; m.ciuchini _ et al . _ , ; j.agrawal _ et al . _ , ( 1996 ) 2263 ; m.a.diaz _ et al . _ , ; g.v.kranrotis , z.phys.*c71 * ( 1996 ) 163 ; a.l.kagan and m.neubert , _ eur.phys.j._*c7 * ( 1999 ) 5 . ( cleo collaboration ) , ; b.abbott _ et al . _ ( d collaboration ) , ; c.albajar _ et al . _ ( ua1 collaboration ) , . also , a recent search in exclusive channels : t.affolder _ et al . _ ( cdf collaboration ) , hep - ex/9905004 . a. ali , l.t.handoko , g.hiller , t.morozumi , . a. ali and g. hiller , . b.grinstein , m.savage , and m.b.wise , . z.ligeti , y.nir , ; m.gremm , a.kapustin , z.ligeti , m.wise , ; m.neubert , . c. bauer and c.burrell , . j. chay , h. georgi and b. grinstein , phys . lett . * b247 * ( 1990 ) 399 . bigi , m. shifman , n.g . uraltsev , a. vainshtein , phys . * 71 * ( 1993 ) 496 , b. blok , l. koyrakh , m. shifman , a. vainshtein , phys * d49 * ( 1994 ) 3356 , erratum - ibid . * d50 * ( 1994 ) 3572 , a. v. manohar , m. b. wise , phys . rev . * d49 * ( 1994 ) 1310 ; i.i.bigi , n.g.uraltsev , a.i.vainshtein , ; * erratum * , . a.falk , m.luke , and m.savage , . c. bauer , phys * d57 * ( 1998 ) 5611 ; * erratum * . m. gremm and a. kapustin , phys . rev . * d55 * ( 1997 ) 6924 . a.f . falk and m.neubert , phys * d47 * ( 1993 ) 2965 ; . t. mannel , phys . rev . * d50 * ( 1994 ) 428 . a. f. falk , m. luke , m. savage , phys . * d53 * ( 1996 ) 2491 . shifman and m.b . voloshin , sov . ( 1987 ) 292 ; m.b . voloshin , n.g . uraltsev , v.a . khoze and m.a . shifman , sov . j. nucl . 46 ( 1987 ) 112 . b. blok , r.d . dikeman and m. shifman , phys . rev . * d51 * , 6167 ( 1995 ) . c. bauer , a. falk and m. luke , . g.buchalla and g.isidori , . m. neubert , phys . rev . * d49 * ( 1994 ) 3367 ; phys . rev . * d49 * ( 1994 ) 4623 . a. falk and m. luke , . m.luke and a.manohar , .

we study nonperturbative corrections to the inclusive rare decay @xmath0 by performing an operator product expansion ( ope ) to @xmath1 . the values of the matrix elements entering at this order are unknown and introduce uncertainties into physical quantities . we study uncertainties introduced into the partially integrated rate , moments of the hadronic spectrum , as well as the forward - backward asymmetry . we find that for large dilepton invariant mass @xmath2 these uncertainties are large . we also assess the possibility of extracting the hqet parameters @xmath3 and @xmath4 using data from this process . # 1/ # 1#2#3 ( # 2 ) # 3 # 1#2#3 ( # 2 ) # 3 # 1#2#3 ( # 2 ) # 3 # 1#2#3 ( # 2 ) # 3 # 1#2#3b * # 1 * ( # 2 ) # 3 # 1#2#3 ( # 2 ) # 3 # 1#2#3d * # 1 * ( # 2 ) # 3 # 1#2#3 ( # 2 ) # 3 # 1#2#3 ( # 2 ) # 3 # 1#2#3 ( # 2 ) # 3 # 1#2#3 ( # 2 ) # 3 # 1#2#3 * # 1 * ( # 2 ) # 3 # 1#2#3_ibid , _ * # 1 * ( # 2 ) # 3

in 1972 j. tits @xcite proved that a finitely generated linear group either is virtually solvable , or contains a copy of the free group @xmath1 of rank @xmath0 . nowadays such a dichotomy is called the _ tits alternative_. this alternative is a powerful result with important consequences . it was used by m. gromov in his proof of the famous `` _ _ polynomial growth _ _ '' theorem @xcite ; it also shows that linear groups satisfy _ von neumann conjecture _ : every linear group is either amenable or contains a non - abelian free subgroup . the tits alternative can be naturally modified by substituting `` finitely generated '' , `` virtually solvable '' and `` contains a non - abelian free subgroup '' with other conditions of a similar form . an example of a result of this type is the theorem of g. noskov and e. vinberg @xcite claiming that every subgroup of a finitely generated coxeter group is either virtually abelian or large . recall that a group is said to be _ large _ if it has a finite index subgroup which maps onto a non - abelian free group . a large group always contains a non - abelian free subgroup , but not vice - versa . let us now formally describe the possible versions of the tits alternative that we are going to consider . let @xmath2 be a collection of cardinals ; a group @xmath3 is said to be _ @xmath2-generated _ , if there is a generating set @xmath4 of @xmath3 and @xmath5 such that @xmath6 suppose that @xmath7 is a collection of cardinals , @xmath8 is a class of groups and @xmath3 is a group . we will say that @xmath3 _ satisfies the tits alternative relative to _ @xmath9 if for any @xmath2-generated subgroup @xmath10 either @xmath11 or @xmath12 contains a non - abelian free subgroup . the group @xmath3 _ satisfies the strong tits alternative relative to _ @xmath9 if for any @xmath2-generated subgroup @xmath10 either @xmath11 or @xmath12 is large . in this terminology the theorem of tits @xcite , mentioned above , tells us that linear groups satisfy the tits alternative relative to @xmath13 where @xmath14 is the collection of all finite cardinals and @xmath15 is the class of virtually solvable groups . for the class @xmath16 , of virtually abelian groups , the tits alternative relative to @xmath17 is known to hold in any word hyperbolic group @xcite , @xmath18 @xcite , where @xmath19 is the free group of rank @xmath20 , and in any group acting freely and properly on a finite dimensional cat(@xmath21 ) cubical complex @xcite . let @xmath22 be the collection of all countable cardinals . the result of g. noskov and e. vinberg @xcite in this language becomes : finitely generated coxeter groups satisfy the strong tits alternative relative to @xmath23 . the goal of this paper is to prove that many forms of tits alternative are stable under graph products . let @xmath24 be a simplicial graph and suppose that @xmath25 is a collection of groups ( called _ vertex groups _ ) . the _ graph product _ @xmath26 , of this collection of groups with respect to @xmath24 , is the group obtained from the free product of the @xmath27 , @xmath28 , by adding the relations @xmath29=1 \text { for all } g_v\in g_v,\ , g_u\in g_u \text { such that $ \{v , u\}$ is an edge of } { \gamma}.\ ] ] the graph product of groups is a natural group - theoretic construction generalizing free products ( when @xmath24 has no edges ) and direct products ( when @xmath24 is a complete graph ) of groups @xmath27 , @xmath28 . graph products were first introduced and studied by e. green in her ph.d . thesis @xcite . further properties of graph products have been investigated by s. hermiller and j. meier in @xcite and by t. hsu and d. wise in @xcite . basic examples of graph products are _ right angled artin groups _ , also called _ graph groups _ ( when all vertex groups are infinite cyclic ) , and _ right angled coxeter groups _ ( when all vertex groups are cyclic of order @xmath0 ) . throughout this paper @xmath30 will denote the group of integers under addition and @xmath31 will denote the infinite dihedral group . we will be interested in collections of cardinals @xmath2 and classes of groups @xmath8 satisfying the following properties : * ( _ closed under isomorphisms _ ) if @xmath32 are groups , @xmath33 and @xmath34 then @xmath35 ; * ( _ closed under @xmath2-generated subgroups _ ) if @xmath33 and @xmath36 is an @xmath2-generated subgroup , then @xmath35 ; * ( _ closed under direct products of @xmath2-generated groups _ ) if @xmath37 are @xmath2-generated then @xmath38 ; * ( _ contains the infinite cyclic group _ ) @xmath39 ; * ( _ contains the infinite dihedral group _ ) if @xmath40 then @xmath41 . suppose that @xmath2 is a collection of cardinals and @xmath42 is a class of groups enjoying ( p@xmath21)(p@xmath43 ) . let @xmath24 be a finite graph and let @xmath44 be a family of groups . then the graph product @xmath45 satisfies the tits alternative relative to @xmath46 provided each vertex group @xmath27 , @xmath47 , satisfies this alternative . to establish the strong version of tits alternative we will need one more condition on @xmath2 and @xmath8 : * ( _ @xmath2-locally profi _ ) if @xmath33 is non - trivial and @xmath2-generated then @xmath48 possesses a proper finite index subgroup . let @xmath2 be a collection of cardinals and let @xmath8 be a class of groups enjoying the properties ( p@xmath21)-(p@xmath49 ) , such that @xmath7 contains all finite cardinals or at least one infinite cardinal . suppose that @xmath24 is a finite graph and @xmath25 is a family of groups . then the graph product @xmath45 satisfies the strong tits alternative relative to @xmath46 provided each vertex group @xmath27 , @xmath28 , satisfies this alternative . examples of classes of groups with properties ( p@xmath21)-(p@xmath49 ) , for @xmath50 , are the classes consisting of virtually abelian groups , virtually nilpotent groups , ( virtually ) polycyclic groups , ( virtually ) solvable groups and , more generally , elementary amenable groups . it is easy to see that all of these properties are necessary . for example , if groups from @xmath8 contain no free subgroups and @xmath51 is a non - trivial group without proper finite index subgroups then @xmath52 will possess no non - trivial finite quotients . it follows that @xmath52 can not be large ; on the other hand , @xmath53 as it contains a copy of @xmath1 . thus we see that property ( p@xmath49 ) is necessary for the claim of theorem b. for any @xmath54 , let @xmath55 be the class of all solvable groups of derived length at most @xmath20 . denote by @xmath56 the class of all groups that are virtually in @xmath55 . it is easy to see that the pairs @xmath57 and @xmath58 enjoy the properties ( p@xmath21)-(p@xmath49 ) , for all @xmath59 with @xmath60 ( when @xmath61 the class @xmath62 , of abelian groups , does not satisfy ( p@xmath43 ) ) . hence applying theorem b we achieve [ cor : vsol - n ] suppose that @xmath63 for some @xmath60 or @xmath64 for some @xmath65 . let @xmath3 be a graph product of groups from @xmath8 . then any finitely generated subgroup @xmath12 of @xmath3 either is large or belongs to @xmath42 . observe that the graph product @xmath3 could be taken over an infinite graph in the above corollary because a finitely generated subgroup is always contained in a `` sub - graph product '' with only finitely many vertices ( see remark [ rem : fg_sbgp - fin_supp ] ) . it is also worth noting that the group @xmath66 , of rational numbers , can be embedded into a finitely generated group from @xmath67 ( this follows from the fact that there are finitely generated center - by - metabelian groups whose center is the free abelian group of countably infinite rank see @xcite ; taking a quotient of such a group by an appropriate central subgroup produces an embedding of an arbitrary countable abelian group into the center of a finitely generated group from @xmath67 ) . consequently , the free square of a finitely generated group from @xmath67 can contain @xmath68 , which is neither solvable nor large . thus the assumption that @xmath12 is finitely generated in corollary [ cor : vsol - n ] is essential . applying theorem b to finitely generated right angled coxeter groups , we see that these groups satisfy the strong tits alternative relative to @xmath69 , where @xmath70 is the smallest class of groups containing @xmath71 , closed under isomorphisms , taking subgroups and direct products . it is easy to see that @xmath72 , where @xmath73 denotes the class of polycyclic groups ; in particular all groups in @xmath70 will be finitely generated . a theorem of f. haglund and d. wise @xcite states that every finitely generated coxeter group is virtually a subgroup of some finitely generated right angled coxeter group . therefore we recover the result of noskov - vinberg @xcite , mentioned above : [ cor : coxeter ] let @xmath3 be a finitely generated coxeter group . if @xmath74 is an arbitrary subgroup then either @xmath12 is large or @xmath12 is finitely generated and virtually abelian . finally we would like to suggest the strongest possible ( in our opinion ) form of tits alternative as follows . suppose that @xmath7 is a collection of cardinals , @xmath8 is a class of groups and @xmath3 is a group . we will say that @xmath3 _ satisfies the strongest tits alternative relative to _ @xmath9 if for any @xmath2-generated subgroup @xmath10 either @xmath11 or @xmath12 has an epimorphism onto the free group @xmath1 of rank @xmath0 . the group @xmath75 is an example of a torsion - free large group ( @xcite ) which does not map onto @xmath1 ( @xcite ) . thus the strongest tits alternative is indeed more restrictive than the strong tits alternative . the evident groups , satisfying the strongest tits alternative relative to the class of torsion - free abelian groups ( and arbitrary @xmath7 ) , are residually free groups . for a collection of cardinals @xmath7 , we define a new collection of cardinals @xmath76 , by saying that a cardinal @xmath77 belongs to @xmath76 if and only if @xmath78 ( see sub - section [ s : cardinals ] for the definition of the addition of cardinals ) . for instance , if @xmath79 then @xmath80 ; note also that @xmath81 and @xmath82 . in order to prove the strongest alternative , we need an additional property of the pair @xmath46 : * ( _ @xmath76-locally indicable _ ) if @xmath33 is non - trivial and @xmath76-generated then @xmath48 has an infinite cyclic quotient . evidently ( p@xmath83 ) implies that every group in @xmath8 is torsion - free ( provided @xmath76 contains at least one non - zero cardinal ) . basic examples of groups satisfying ( p@xmath83 ) with @xmath50 are torsion - free nilpotent groups ( see @xcite ) . let @xmath2 be a collection of cardinals and let @xmath8 be a class of groups enjoying the properties ( p@xmath21)-(p@xmath84 ) and ( p@xmath83 ) . suppose that @xmath24 is a finite graph and @xmath25 is a collection of groups . then the graph product @xmath45 satisfies the strongest tits alternative relative to @xmath46 provided each vertex group @xmath27 , @xmath28 , satisfies this alternative . consider any pair @xmath9 satisfying the conditions of theorem c , such that @xmath85 . then for an arbitrary non - trivial @xmath76-generated group @xmath33 , @xmath86 is @xmath2-generated and contains a copy of @xmath1 , hence @xmath87 . so , the strongest tits alternative ( relative to @xmath46 ) for @xmath3 would imply that @xmath3 maps onto @xmath1 , hence the image of @xmath48 under this homomorphism is non - trivial , and , so it possesses an epimorphism onto @xmath88 . thus the assumption of ( p@xmath83 ) in theorem c is indeed necessary . for the first application of theorem c , let us take @xmath79 and let @xmath42 be any class of torsion - free groups enjoying the properties ( p@xmath21)-(p@xmath84 ) ( e.g. , @xmath42 could be the class of torsion - free amenable groups ) . in this case @xmath76-local indicability is equivalent to torsion - freeness , and so ( p@xmath83 ) holds automatically . recalling that a @xmath0-generated group maps onto @xmath1 if and only if it is isomorphic to @xmath1 , we achieve [ cor:2-gen ] let @xmath42 be a non - empty class of torsion - free groups , closed under isomorphisms , direct products and taking subgroups . suppose that @xmath3 is a graph product with all vertex groups from @xmath8 . then for any @xmath0-generated subgroup @xmath10 , either @xmath89 or @xmath90 . corollary [ cor:2-gen ] generalizes a classical result of a. baudisch @xcite , who proved that any two non - commuting elements of a right angled artin group generate a copy of @xmath1 . in fact , we are now able to say much more about subgroups of right angled artin groups . indeed , taking @xmath91 and @xmath8 to be the class of finitely generated torsion - free abelian groups , and applying theorem c we obtain the following result : [ cor : raag ] any subgroup of a finitely generated right angled artin group is either free abelian of finite rank or maps onto @xmath1 . the finite generation assumption in the above statement is not very important . indeed , using corollary [ cor : raag ] together with remark [ rem : fg_sbgp - fin_supp ] , it is easy to show that any non - abelian subgroup of an arbitrary right angled artin group maps onto @xmath1 . an easy consequence of corollary [ cor : raag ] is that for every non - cyclic subgroup @xmath12 , of a right angled artin group , the first betti number ( i.e , the @xmath66-rank of the abelianization ) is at least two . as far as we know , corollary [ cor : raag ] gives the first non - trivial family of groups satisfying the strongest tits alternative ( it is not hard to construct right angled artin groups that are not residually free ) . the family of subgroups of right angled artin groups is very rich and includes a lot of examples examples ( see @xcite ) . however , corollary [ cor : raag ] can be used to show that many groups are not embeddable into right angled artin groups . more precisely , one can obtain information regarding solutions of an equation in a right angled artin group from the solutions of the same equation in free and free abelian groups . for example , we have [ cor : abc ] if three elements @xmath92 of a right angled artin group @xmath3 satisfy @xmath93 for @xmath94 , then these elements pairwise commute . the above fact is an immediate consequence of corollary [ cor : raag ] and the result of r. lyndon and m. schtzenberger @xcite , who proved the same statement when @xmath3 is free . corollary [ cor : abc ] generalizes a theorem of j. crisp and b. wiest ( * ? ? ? 7 ) , who established its claim in the case when @xmath95 . the paper is organized as follows . in section [ sec : basic_prop ] we recall some basic properties of graph products , and in section [ s : parabolic ] we develop the theory of parabolic subgroups of graph products . the results of this section , together with bass - serre theory , allow us to prove , in section [ s : structuretheorem ] , the structure theorem for subgroups of graph products which do not contain a copy of @xmath1 ( theorem [ thm : structure ] ) . the structure theorem is one of the main results of this work , and it is intensively used in the proofs of theorems a , b and c. we apply the structure theorem to give a more detailed description of `` small '' subgroups of graph products ( see theorem [ thm : sharpa ] ) , and theorem a is a consequence of this description . in order to prove theorems b and c , we study the kernel of the canonical retraction of a graph product onto one of the vertex groups and prove that this kernel is again a graph product ( theorem [ thm : kernel ] ) . this fact has a number of interesting applications , which are established in section [ s : kernel ] . in section [ s : proof ] we establish a sufficient criterion for a subgroup of a graph product to be large or to map onto @xmath1 ( theorem [ thm : main_crit ] ) , and then we prove theorems b and c. finally , in section [ s : example ] we construct two examples showing that the claims of corollaries [ cor:2-gen ] and [ cor : raag ] are in a certain sense optimal . example [ ex:2-gen_non_virt_cyc ] shows that without the assumption of torsion - freeness , there is no control over @xmath0-generated subgroups of graph products ; and proposition [ prop : ex_fr_rank ] demonstrates that there is no connection between the rank of a non - abelian subgroup of a right angled artin group and the maximal rank of the free group onto which it can be mapped . the latter is based on a rips - type construction , which is presented in proposition [ prop : rips_constr ] . our reference for cardinal numbers is @xcite . cardinals are traditionally well - ordered by injection ( @xmath96 if there is an injective map from @xmath4 to @xmath97 ) , but , assuming the axiom of choice , this ordering is equivalent to the ordering defined via surjections , which we will use ( see ( * ? ? ? * proposition 2.1.8 ) ) . formally , if @xmath77 , @xmath98 are two cardinals , and @xmath4 , @xmath97 are sets with @xmath99 , @xmath100 , then @xmath101 if and only if there exists a surjective map @xmath102 the addition of cardinals @xmath77 and @xmath98 is defined in a natural way : @xmath103 is the cardinal of the union @xmath104 of two disjoint sets @xmath4 and @xmath97 with @xmath99 and @xmath100 . a well - known fact tells us that @xmath105 if at least one of the cardinals @xmath77 , @xmath98 is infinite ( see ( * ? ? ? * corollary 7.6.2 ) ) . if @xmath106 and @xmath77 is a cardinality of a set @xmath4 then @xmath107 denotes the cardinality of the disjoint union of @xmath108 copies of @xmath4 ; thus if @xmath77 is an infinite cardinal then @xmath109 for any @xmath110 . let @xmath2 be a collection of cardinals . we say that a group @xmath3 is @xmath2-generated , if there is a generating set @xmath4 of @xmath3 of cardinality @xmath77 , and a cardinal @xmath111 such that @xmath101 . remark that the image of an @xmath2-generated group under a homomorphism is again @xmath2-generated . it also follows from the definition that if @xmath112 is another collection of cardinals , such that for every @xmath113 there is @xmath114 with @xmath101 , then any @xmath112-generated group is also @xmath7-generated . we will say that a collection of cardinals @xmath7 is _ ample _ if it contains all finite cardinals or if it contains at least one infinite cardinal . [ lem : card_of_fi ] let @xmath7 be an ample collection of cardinals . if @xmath3 is an @xmath7-generated group and @xmath115 is a subgroup of finite index then @xmath116 is also @xmath7-generated . by the assumptions , @xmath117 for some @xmath118 , where @xmath119 . choose a generating set @xmath4 of @xmath3 with @xmath120 for some @xmath121 . by a theorem of reidemeister and schreier ( see @xcite ) , @xmath116 can be generated by a subset @xmath97 of the set @xmath122 , where @xmath123 . consequently @xmath124 . if @xmath4 is finite then @xmath97 is also finite , and hence @xmath116 is @xmath7-generated by the assumptions on @xmath7 . on the other hand , if @xmath4 is infinite , then @xmath77 is an infinite cardinal , and so @xmath125 , as required . let @xmath126 be a graph without loops or multiple edges . we will use @xmath127 and @xmath128 to denote the set of vertices and the set of edges of @xmath126 respectively . an edge can be considered as a @xmath0-element subset of @xmath127 . a path in @xmath126 of length @xmath20 from @xmath129 to @xmath130 is a sequence of vertices @xmath131 where @xmath132 , for @xmath133 . for any subset @xmath134 , by @xmath135 we will denote the _ full subgraph _ of @xmath126 with vertex set @xmath136 that is , @xmath137 and @xmath138 if and only if @xmath139 . the _ link _ @xmath140 , of a vertex @xmath141 , is the subset of vertices adjacent to @xmath130 ( excluding the @xmath130 itself ) ; in other words , @xmath142 for a subset @xmath143 , we define @xmath144 . let @xmath145 be a family of groups and let @xmath146 be the corresponding graph product . any element @xmath147 may be represented as a _ word _ @xmath148 where each @xmath149 , called a _ syllable _ of @xmath150 , is an element of some @xmath27 and @xmath151 . the number of syllables is the length of the word . for @xmath152 , we will say that the syllables @xmath149 and @xmath153 _ can be joined together _ if @xmath154 for some @xmath141 and @xmath155 , @xmath156 for @xmath157 . in this case , in @xmath158 the word @xmath150 represents the same element as the words @xmath159 and @xmath160 , whose lengths are strictly smaller than the length of @xmath150 . a word @xmath148 is _ reduced _ either if it is empty or if @xmath161 for all @xmath162 , and no two distinct syllables of @xmath150 can be joined together . we define the following transformations for the word @xmath150 . 1 . remove a syllable @xmath149 if @xmath163 in @xmath3 . 2 . replace two consecutive syllables @xmath149 and @xmath164 in the same vertex group @xmath27 with the single syllable @xmath165 . ( _ syllable shuffling _ ) for consecutive syllables @xmath166 @xmath167 with @xmath168 an edge of @xmath169 interchange @xmath149 and @xmath170 note that transformation ( t3 ) preserves the length of the word , and transformations ( t1),(t2 ) decrease it by @xmath171 . evidently starting with some word @xmath150 and applying finitely many transformations ( t1)-(t3 ) we can obtain a reduced word @xmath172 , representing the same element of the group @xmath173 . the following theorem was first proved by e. green ( * ? ? ? 3.9 ) in her thesis . [ thm : norm_form ] every element @xmath174 can be represented by a reduced word . moreover , if two reduced words represent the same element of the group , then one can be obtained from the other after applying a finite sequence of syllable shuffling . in particular , the length of a reduced word is minimal among all words representing @xmath175 , and a reduced word represents the identity if and only if it is the empty word . let @xmath147 and @xmath176 be a reduced word representing @xmath177 we define the _ length _ of @xmath175 in @xmath3 to be @xmath178 and the @xmath179 of @xmath175 in @xmath3 to be @xmath180 by the normal form theorem , the length and support of an element are well defined . for a subset @xmath181 , the _ support of _ @xmath4 , @xmath182 will be defined by @xmath183 . observe that for a subgroup @xmath184 one has @xmath185 . in particular , if @xmath12 is finitely generated then @xmath186 is a finite subset of @xmath127 . for every @xmath187 , theorem [ thm : norm_form ] also allows to define the set @xmath188 , consisting of all @xmath189 such that some reduced word @xmath150 , representing @xmath175 , starts with a syllable from @xmath27 . similarly , @xmath190 will denote the subset of all @xmath191 such that some reduced word for @xmath175 ends with a syllable from @xmath192 . evidently @xmath193 . throughout this section @xmath24 will denote a simplicial graph , @xmath194 , @xmath195 will be a family of groups and @xmath45 will be the corresponding graph product . for any subset @xmath196 the subgroup @xmath197 , generated by @xmath198 , is called a _ full subgroup _ ; according to a standard convention , @xmath199 . by the normal form theorem , @xmath200 is the graph product of the groups @xmath201 with respect to the graph @xmath135 . it is also easy to see that there is a _ canonical retraction _ @xmath202 defined ( on the generators of @xmath3 ) by @xmath203 for each @xmath204 with @xmath205 , and @xmath206 for each @xmath207 with @xmath208 . [ rem : fg_sbgp - fin_supp ] if @xmath74 is a finitely generated subgroup then @xmath12 is contained in the full subgroup @xmath200 , of @xmath3 , where @xmath209 is a finite subset of @xmath210 , and thus @xmath200 is a graph products of the family @xmath201 over a finite graph @xmath211 . for any @xmath47 the group @xmath212 naturally splits as a free amalgamated product : @xmath213 , where @xmath214 , @xmath215 and @xmath216 ( cf . * lemma 3.20 ) ) . the goal of this section is to develop the theory of _ parabolic subgroups _ ( i.e. , conjugates of full subgroups ) of graph products . some of the statements in this section are similar to those already known about parabolic subgroups of coxeter groups ( see ( * ? ? ? * sections 2,3 ) ) and about parabolic subgroups of right angled artin groups ( see ( * ? ? ? * section 2 ) ) . [ lem : sm_aux ] suppose that @xmath217 and @xmath218 are some elements of @xmath45 such that @xmath219 , @xmath220 and @xmath221 . then @xmath175 can be represented by a reduced word @xmath222 , where @xmath223 and @xmath224 , with @xmath225 . choose some reduced words @xmath226 , @xmath227 and @xmath228 representing the elements @xmath175 , @xmath229 , and @xmath230 in @xmath3 respectively . the statement will be proved by induction on @xmath231 . if @xmath232 then the claim clearly holds , so assume that @xmath65 and the claim has already been established for all @xmath187 , satisfying the assumptions of the lemma , with @xmath233 . if @xmath234 , there is nothing to prove . otherwise , the equality @xmath219 shows that @xmath235 , representing the left - hand side , can not be a reduced word . since the word @xmath226 is reduced and @xmath236 , there must exist @xmath237 such that @xmath238 can be joined with @xmath239 in @xmath150 . it follows that @xmath240 because otherwise the word @xmath226 would not be reduced . it also follows that @xmath175 is represented by the word @xmath241 and @xmath242 , where @xmath243 . hence we have @xmath244 , where @xmath245 is the element represented by the word @xmath246 . note that the word @xmath247 is reduced as it was obtained from a reduced word for @xmath175 by removing its last syllable ; in particular , @xmath248 . suppose that there is some @xmath249 . then @xmath250 for some @xmath251 , @xmath252 , with @xmath253 and @xmath254 . observe that @xmath238 must commute with @xmath255 , as @xmath256 , therefore @xmath257 , implying that @xmath258 , which contradicts to our assumptions . thus @xmath259 and we can apply the induction hypothesis to conclude that @xmath260 is represented by a reduced word @xmath261 , where @xmath262 , @xmath223 and @xmath263 . after setting @xmath264 , we can conclude that @xmath265 is represented by the word @xmath266 ( which must be reduced as its length is @xmath231 ) , satisfying the required properties . [ lem : inter_spec ] let @xmath267 be an arbitrary collection of subsets of @xmath268 . then @xmath269 , where @xmath270 . evidently , @xmath271 . to show the reverse inclusion , consider any @xmath272 . by theorem [ thm : norm_form ] , @xmath273 for every @xmath274 , hence @xmath275 , thus @xmath276 . [ prop : inter_parab ] consider arbitrary @xmath277 and @xmath187 . then there exist @xmath278 and @xmath279 such that @xmath280 . obviously one can write @xmath281 , where @xmath282 , @xmath283 and @xmath284 . then @xmath279 and @xmath285 . if @xmath286 for some @xmath287 with @xmath288 , then evidently @xmath289 , that is @xmath260 can be shortened . thus without loss of generality we can assume that @xmath290 . for every element @xmath291 , take some @xmath292 such that @xmath293 . applying lemma [ lem : sm_aux ] we see that @xmath294 . the latter implies that @xmath295 , i.e. , every syllable of @xmath229 commutes with every syllable of @xmath260 . setting @xmath296 we see that @xmath297 by lemma [ lem : inter_spec ] , @xmath298 , where @xmath299 , and so @xmath300 as claimed . any subgroup @xmath301 , conjugate to a full subgroup , is called _ parabolic_. moreover , if there are @xmath302 and @xmath303 such that @xmath304 , then @xmath116 is said to be a _ proper parabolic subgroup_. in the latter case @xmath305 unless @xmath306 for all @xmath307 . clearly any parabolic subgroup @xmath304 , where @xmath308 and @xmath303 , is a retract of @xmath3 , with the retraction @xmath309 , defined by @xmath310 for all @xmath187 . proposition [ prop : inter_parab ] immediately yields [ cor : inter_parab ] the intersection of two parabolic subgroups of a graph product is again a parabolic subgroup . [ lem : conj_of_parab_in_another_parab ] if @xmath277 and @xmath311 satisfy @xmath312 then @xmath313 . if , in addition , @xmath314 then there is @xmath315 such that @xmath316 . denote @xmath317 and let @xmath309 be the retraction defined above . by definition , @xmath318 , and as @xmath319 , we have @xmath320 . therefore @xmath321 , and @xmath322 . it follows that @xmath323 and @xmath324 . now , if @xmath314 , we have @xmath325 . hence @xmath326 , where @xmath327 . [ cor : non - tr_parab_in_other ] if @xmath277 and @xmath311 are such that @xmath312 and @xmath328 for all @xmath329 then @xmath330 . [ lem : conj_of_parab ] if @xmath116 is a parabolic subgroup of a graph product @xmath3 and @xmath331 for some @xmath187 , then @xmath332 . let @xmath309 be a retraction of @xmath3 onto @xmath116 . for any @xmath333 , take @xmath334 . then @xmath335 , hence @xmath336 for all @xmath333 , thus @xmath337 . [ prop : min_parab ] let @xmath4 be a subset of the graph product @xmath45 such that at least one of the following conditions holds : * the graph @xmath126 is finite ; * the subgroup @xmath338 is finitely generated . then there exists a unique minimal parabolic subgroup of @xmath3 containing @xmath4 . the uniqueness is clear from corollary [ cor : inter_parab ] . to establish the existence , suppose , at first , that @xmath126 is finite . then there is a subset @xmath308 , of minimal cardinality , such that @xmath339 for some @xmath340 . let us show that @xmath341 is a minimal parabolic subgroup containing @xmath4 . suppose that there is some parabolic subgroup @xmath342 of @xmath3 such that @xmath343 . according to lemma [ lem : conj_of_parab_in_another_parab ] , @xmath313 , and the minimality of @xmath344 implies that @xmath345 . consequently , @xmath346 and @xmath347 . therefore , applying lemma [ lem : conj_of_parab ] , we can conclude that @xmath348 , as required . in the case ( ii ) , when @xmath349 for some finite subset @xmath350 , it is clear that @xmath351 , where @xmath352 is a finite subset of @xmath353 . recall that the full subgroup @xmath200 , of @xmath3 , is itself a graph product ( of the groups @xmath201 with respect to the graph @xmath211 ) and any parabolic subgroup of @xmath200 is a parabolic subgroup of @xmath3 . since @xmath354 , by the first part of the proof @xmath4 is contained in a minimal parabolic subgroup @xmath341 of @xmath200 for some @xmath355 and @xmath356 . if @xmath357 for some @xmath358 and @xmath359 , then @xmath360 and so , by lemma [ lem : conj_of_parab_in_another_parab ] , @xmath342 is a parabolic subgroup of @xmath200 . therefore , using minimality of @xmath341 , we achieve @xmath361 ; thus the proposition is proved . suppose that a subset @xmath362 is contained in a minimal parabolic subgroup of @xmath3 . then this parabolic subgroup will be called the _ parabolic closure _ of @xmath4 and will be denoted by @xmath363 . recall that the normalizer @xmath364 , of a subset @xmath365 , is the subgroup of @xmath3 defined by @xmath366 . [ lem : parab_normal ] let @xmath45 and @xmath365 . suppose that the parabolic closure of @xmath4 in @xmath3 exists . then for any @xmath187 with @xmath367 one has @xmath368 , where @xmath369 . in particular , @xmath370 , i.e. , any element normalizing @xmath4 also normalizes @xmath371 . clearly for any @xmath187 , with @xmath367 , @xmath4 is contained in @xmath372 , which is also a parabolic subgroup . hence @xmath373 , and lemma [ lem : conj_of_parab ] implies that @xmath374 , thus @xmath375 . [ prop : normalizer_of_parab ] let @xmath116 be a non - trivial parabolic subgroup of the graph product @xmath45 . choose @xmath303 and @xmath376 so that @xmath377 and @xmath378 for all @xmath379 . then @xmath380 ; in particular the normalizer @xmath381 is a parabolic subgroup of @xmath3 . after conjugating everything by @xmath382 , we can assume that @xmath383 . observe that @xmath384 since @xmath385 . consider any element @xmath386 , any @xmath379 and any @xmath387 . then @xmath219 for some @xmath388 and lemma [ lem : sm_aux ] implies that @xmath389 . since the latter holds for every @xmath379 , we see that @xmath390 , where @xmath391 by lemma [ lem : inter_spec ] . thus @xmath392 , and evidently @xmath393 , hence @xmath394 . for any graph @xmath24 one can define its _ complement graph _ @xmath395 to be the graph with the same vertex set @xmath396 such that the edge set @xmath397 is the complement of @xmath398 in the set of two - element subsets of @xmath210 ; in other words , for any @xmath399 , @xmath400 if and only if @xmath401 . a graph @xmath24 will be called _ irreducible _ if @xmath210 can not be represented as a union of two disjoint non - empty subsets @xmath402 such that @xmath403 . it is easy to see that @xmath24 is irreducible if and only if @xmath395 is connected . in the case when @xmath24 is not irreducible , any graph product @xmath45 , with respect to @xmath24 , naturally splits as a direct product @xmath404 , where @xmath405 and @xmath406 . [ cor : norm_in_irred ] assume that @xmath24 is a finite irreducible graph and @xmath45 is the graph product of a family of groups @xmath407 with respect to @xmath24 . suppose that @xmath408 is a non - trivial subgroup such that @xmath409 is a proper parabolic subgroup of @xmath3 . then @xmath410 is contained in a proper parabolic subgroup of @xmath3 . combining proposition [ prop : min_parab ] with lemma [ lem : parab_normal ] we see that @xmath411 , where @xmath412 . by the assumptions , @xmath377 for some @xmath302 and @xmath303 ; after discarding all @xmath379 with trivial vertex groups , we can suppose that @xmath378 for each @xmath379 . then @xmath413 , according to proposition [ prop : normalizer_of_parab ] . since @xmath414 ( as @xmath415 ) , we can conclude that @xmath381 is a proper parabolic subgroup of @xmath3 because @xmath416 as @xmath24 is irreducible . [ lem : virt_cyc - parab ] if @xmath24 is a finite irreducible graph and @xmath10 is a virtually cyclic subgroup which is not contained in any proper parabolic subgroup of @xmath3 , then @xmath417 is finite for each proper parabolic subgroup @xmath116 of @xmath3 . suppose that @xmath418 is infinite for some proper parabolic subgroup @xmath419 . then there is an infinite cyclic subgroup @xmath420 such that @xmath421 . observe that @xmath422 is a proper parabolic subgroup and @xmath423 , therefore , @xmath12 must also be contained in a proper parabolic subgroup by corollary [ cor : norm_in_irred ] , contradicting to our assumptions . hence @xmath424 . [ prop : fin_parab_cl ] suppose that the graph @xmath24 is finite and @xmath362 . then there is a finite subset @xmath425 such that @xmath426 . by proposition [ prop : min_parab ] , for every finite subset @xmath427 , there exist @xmath428 and @xmath429 such that @xmath430 and @xmath431 for all @xmath432 . since @xmath268 is finite , the function , that corresponds to each finite subset @xmath427 the integer value @xmath433 , attains its maximum on some finite subset @xmath425 . evidently , to prove the claim it is enough to show that @xmath434 . for any @xmath435 we can find @xmath358 and @xmath187 such that @xmath436 and @xmath328 for all @xmath437 . by the choice of @xmath438 , we have @xmath439 and @xmath440 , where @xmath441 and @xmath442 . it follows that @xmath443 ( corollary [ cor : non - tr_parab_in_other ] ) , which , in view of @xmath439 , implies that @xmath444 . finally , the inclusion @xmath445 together with lemma [ lem : conj_of_parab ] yield @xmath446 , allowing to conclude that @xmath447 for all @xmath435 , thus finishing the proof . if @xmath24 is finite and @xmath45 is the graph product of groups with respect @xmath24 , then any descending / ascending chain of parabolic subgroups stabilizes . this is an easy consequence of propositions [ prop : min_parab ] and [ prop : fin_parab_cl ] , and is left as an exercise for the reader . throughout this section @xmath126 will be a finite simplicial graph , @xmath268 , @xmath45 will be the graph product of a family of groups @xmath195 with respect to @xmath24 . suppose that @xmath3 acts on a simplicial tree @xmath448 without edge inversions . it is well known that any element @xmath187 either fixes a vertex of @xmath448 or there exists a unique minimal @xmath449-subtree of @xmath448 , which is a bi - infinite geodesic path ( called _ the axis _ of @xmath175 ) , where @xmath449 acts by translation ( see ( * ? ? ? * i.4.11,i.4.13 ) or @xcite ) . in the former case @xmath175 is said to be _ elliptic _ , and in the latter case it is said to be _ hyperbolic_. the _ translation length _ @xmath450 , of @xmath175 , is the integer defined by @xmath451 , where @xmath452 is the standard edge - path metric on @xmath448 . if the element @xmath187 is hyperbolic then @xmath453 for every vertex @xmath454 ( see @xcite ) . the following theorem allows us to describe the structure of any subgroup of a graph product ( with respect to a finite graph ) , which has no non - abelian free subgroups . [ thm : structure ] suppose that the graph @xmath24 is finite , irreducible and has at least two vertices . if @xmath10 is a subgroup then at least one of the following is true : 1 . @xmath12 is contained in a proper parabolic subgroup of @xmath3 ; 2 . @xmath455 @xmath456 ; 4 . [ it:4 ] for every proper subset @xmath302 there is a subgroup @xmath457 such that @xmath458 and @xmath459 in @xmath3 for all @xmath187 . suppose that does not hold , i.e. , there is a subset @xmath460 such that every non - abelian free subgroup of @xmath12 intersects non - trivially some conjugate of @xmath461 in @xmath3 ( note that this is also the case if @xmath12 contains no non - abelian free subgroups at all ) . choose any vertex @xmath462 and set @xmath463 , @xmath464 and @xmath465 ; observe that @xmath355 . as @xmath24 is irreducible and @xmath466 , @xmath48 and @xmath467 are non - empty proper subsets of @xmath353 . then @xmath3 naturally splits as the amalgamated free product : @xmath213 . let @xmath448 be the bass - serre tree , corresponding to this splitting , on which @xmath3 acts simplicially and without edge inversions . the vertex stabilizers of this splitting are conjugates of @xmath200 and @xmath468 , and the edge stabilizers are conjugates of @xmath469 ( see ( * ? ? ? * i.4.1 , thm . 7 ) ) , all of which are proper parabolic subgroups in @xmath3 . since every non - abelian free subgroup of @xmath12 has non - trivial intersection with some conjugate of @xmath200 ( and thus it can not act freely on @xmath448 ) , by theorem 2.7 from @xcite the induced action of @xmath12 on @xmath448 must be _ reducible _ , which means that one of the following occurs : * every element of @xmath12 fixes a vertex of @xmath448 ; * there is a bi - infinite geodesic line @xmath470 which is invariant under the action of @xmath12 ; * there is an end of @xmath448 which is fixed by @xmath12 . in case ( i ) , by proposition [ prop : fin_parab_cl ] , there is a finite subset @xmath471 such that @xmath472 . the subgroup @xmath473 is finitely generated and every element of @xmath474 fixes a vertex of @xmath448 , therefore @xmath474 fixes some vertex of @xmath448 ( see ( * ? ? ? * i.6.5 , cor . 3 ) ) . but full @xmath3-stabilizers of vertices of @xmath448 are proper parabolic subgroups of @xmath3 , therefore @xmath475 is a proper parabolic subgroup of @xmath3 . thus @xmath12 will be contained in a proper parabolic subgroup of @xmath3 . assume , now , that we are in case ( ii ) . then @xmath12 acts on the bi - infinite geodesic line @xmath470 , giving rise to a homomorphism @xmath476 , where @xmath477 is the infinite dihedral group of all simplicial isometries of @xmath470 . moreover , we can suppose that @xmath478 is infinite , because otherwise @xmath12 would fix a vertex of @xmath470 and we would be in case ( i ) . denote @xmath479 ; then @xmath480 fixes every vertex of @xmath470 and so it is contained in a proper parabolic subgroup of @xmath3 by case ( i ) . if @xmath481 then @xmath12 is isomorphic to an infinite subgroup of @xmath71 , which is either infinite cyclic or infinite dihedral . if @xmath482 then @xmath423 is contained in a proper parabolic subgroup of @xmath3 by corollary [ cor : norm_in_irred ] . finally , assume that we are in case ( iii ) , i.e. , @xmath12 fixes some end @xmath483 of @xmath448 . if an element @xmath484 fixes some vertex @xmath485 of @xmath448 , then it will have to fix every vertex of the unique infinite geodesic ray between @xmath485 and @xmath483 . if @xmath486 is another elliptic element , then it will fix ( pointwise ) another geodesic ray converging to @xmath483 . but any two rays converging to @xmath483 are eventually the same , in particular they will have a common vertex , which will then be fixed by both @xmath175 and @xmath487 . thus @xmath488 will also be elliptic . it follows that the subset @xmath489 , of all elliptic elements of @xmath12 , is a normal subgroup of @xmath12 . the argument for case ( i ) shows that @xmath480 is contained in a proper parabolic subgroup of @xmath3 . and if @xmath482 then corollary [ cor : norm_in_irred ] yields that @xmath423 is contained in a proper parabolic subgroup of @xmath3 . thus , we can assume that @xmath481 , in other words all non - trivial elements of @xmath12 are hyperbolic . let @xmath490 be an element of minimal translation length ( for the action of @xmath12 on @xmath448 ) . for any other element @xmath491 , the intersection of @xmath492 and @xmath493 is an infinite geodesic ray @xmath494 , starting at some vertex @xmath495 of @xmath448 and converging to @xmath483 . after replacing @xmath229 and @xmath230 with their inverses , where necessary , we assume that @xmath496 and @xmath497 . write @xmath498 , where @xmath106 , @xmath499 and @xmath500 . then @xmath501 and so @xmath502 , thus @xmath503 , hence @xmath504 by the minimality of @xmath505 . thus @xmath506 for all @xmath507 , i.e. , @xmath508 is cyclic . since @xmath229 is hyperbolic , it must have infinite order ( cf . * i.4.11 ) ) , implying that @xmath455 . [ rem:4-core ] in the assumptions of theorem [ thm : structure ] , assume that the condition is satisfied . then @xmath12 has a finitely generated subgroup @xmath509 such that @xmath509 is not contained in any proper parabolic subgroup of @xmath3 and @xmath509 contains a free subgroup of rank @xmath0 . indeed , simply let @xmath509 to be the subgroup of @xmath12 generated by @xmath510 . the above structure theorem can be used to give a more precise description of subgroups of graphs products that do not contain non - abelian free subgroups . let @xmath2 be a collection of cardinals . for @xmath511 we define classes of groups @xmath512 as follows : @xmath513 is the class consisting of all @xmath2-generated groups , which are isomorphic to subgroups of @xmath514 @xmath515 and which do not contain a copy of @xmath1 . @xmath516 is the smallest class , closed under isomorphisms , containing all @xmath2-generated groups @xmath12 , such that @xmath12 is a subgroup of @xmath517 for some @xmath518 , and satisfying the next conditions : @xmath519 and , if for some @xmath520 the vertex group @xmath27 has @xmath0-torsion , then @xmath521 . @xmath512 , @xmath522 , is the class consisting of all @xmath2-generated groups @xmath12 , such that @xmath12 is isomorphic to a subgroup of @xmath517 for some @xmath523 and @xmath524 with @xmath525 , @xmath526 . [ thm : sharpa ] suppose that @xmath2 is a collection of cardinals , @xmath24 is a non - empty finite graph and @xmath45 is the graph product of a family of groups @xmath407 with respect to @xmath24 . if @xmath10 is an @xmath2-generated subgroup , which does not contain a copy of @xmath1 , then @xmath12 is in the class @xmath527 observe that @xmath528 whenever @xmath529 . the proof argues by induction on the number of vertices in @xmath530 the claim is evident if @xmath531 . so we will assume that @xmath532 and that the result holds for every graph product over a graph with fewer vertices than @xmath530 let @xmath12 be an @xmath2-generated subgroup of @xmath3 that does not contain a non - abelian free subgroup ; we have to show that @xmath533 there are two cases to consider . _ case 1 : @xmath24 is irreducible . _ since @xmath534 we are under the assumptions of the structure theorem ( theorem [ thm : structure ] ) . therefore @xmath12 is either contained in a proper parabolic subgroup ( and , hence , the result follows by induction ) , or @xmath535 or @xmath456 . by the assumptions , @xmath536 for all @xmath537 . on the other hand , the case @xmath538 can only happen if @xmath3 has @xmath0-torsion , which , in its own turn , only happens when @xmath27 has @xmath0-torsion for some @xmath28 ( see , for example , corollary [ cor : torfree ] below ) , and then @xmath71 is in @xmath539 for any @xmath537 . _ case 2 : @xmath24 is not irreducible . _ then there exist non - empty subsets @xmath48 and @xmath467 of @xmath210 such that @xmath540 and @xmath541 . as @xmath12 is @xmath2-generated , so are @xmath542 and @xmath543 since @xmath12 does not contain a non - abelian free subgroup , neither @xmath542 nor @xmath544 contain a copy of @xmath1 . using the induction hypothesis , @xmath545 and @xmath546 . it remains to observe that @xmath12 is contained in the subgroup of @xmath3 , generated by @xmath542 and @xmath544 , which is naturally isomorphic to @xmath547 , as required . theorem a is almost an immediate consequence of theorem [ thm : sharpa ] . if the collection @xmath7 contains no non - zero cardinals , then the statement trivially holds ( since only the trivial group could be @xmath2-generated in this case ) , therefore we will further assume that @xmath7 contains at least one non - zero cardinal , implying that any cyclic group is @xmath7-generated . by the assumptions , all vertex groups @xmath27 , @xmath28 , satisfy the tits alternative relative to ( @xmath548 ) . this means that @xmath549 . by ( p@xmath21),(p@xmath171 ) and ( p@xmath0 ) , @xmath2-generated subgroups of direct products of groups from @xmath513 are in @xmath42 . by ( p@xmath84 ) , @xmath550 ; and if @xmath27 has @xmath0-torsion for some @xmath515 then @xmath40 ( according to the tits alternative for @xmath27 ) and so @xmath551 by ( p@xmath43 ) . therefore @xmath552 again , by ( p@xmath21),(p@xmath171 ) and ( p@xmath0 ) , for @xmath553 all groups from @xmath512 are in @xmath554 now , consider an arbitrary @xmath2-generated subgroup @xmath555 , which does not contain a non - abelian free subgroup . without loss of generality we can suppose that the graph @xmath24 is non - empty , and so , by theorem [ thm : sharpa ] , @xmath556 , finishing the proof . taking @xmath7 to be the collection of all countable cardinals , one can use theorem [ thm : sharpa ] to obtain another interesting corollary , the proof of which is left for the reader : [ cor : rank_ab_in_raag ] if @xmath24 is a finite graph and @xmath3 is a right angled artin group corresponding to @xmath24 ( that is , @xmath3 is the graph product of infinite cyclic groups with respect to @xmath24 ) then the rank of any abelian subgroup of @xmath3 does not exceed @xmath557 . let @xmath24 be a simplicial graph and let @xmath558 be a graph products of a family of groups @xmath559 , with respect to @xmath24 . in this section we will use the normal form theorem ( theorem [ thm : norm_form ] ) to prove that for any @xmath560 , the kernel @xmath561 is itself a graph product . let @xmath268 , @xmath562 , @xmath563 and @xmath564 then @xmath353 is the disjoint union of @xmath565 and @xmath566 for every @xmath567 take a copy @xmath568 of @xmath569 , then construct the graph @xmath570 by gluing together the graphs @xmath571 along the corresponding copies of @xmath572 . more precisely , the vertex set of @xmath570 is the set of equivalence classes of pairs @xmath573 \mid g \in g_a \mbox { and } u \in { b\cup c}\}$ ] , where two such pairs @xmath574 and @xmath575 are equivalent @xmath576 , if and only if @xmath577 . two vertices @xmath578 are connected by an edge if and only if there are @xmath567 and @xmath579 such that @xmath580 $ ] , @xmath581 $ ] and @xmath582 . note that if @xmath200 is infinite and @xmath583 , then the new graph @xmath570 will have infinitely many vertices . [ ex:1 ] let @xmath24 be the simple path of length @xmath0 . that is , @xmath24 has three vertices @xmath584 and two edges @xmath585 suppose that @xmath586 . for @xmath587 we have @xmath588 , @xmath589 and @xmath590 figure [ fig:1 ] below illustrates the construction of the graph @xmath570 in this case . for each @xmath591\in v{\delta},$ ] let @xmath592}$ ] be an isomorphic copy of @xmath27 ( with a fixed isomorphism ) , and for any @xmath593},$ ] let @xmath594 denote the corresponding element in @xmath27 . set @xmath595 and let @xmath596 be the corresponding graph product . define a map @xmath597 by @xmath598 whenever @xmath593}.$ ] note that this is well defined , since if @xmath599 then @xmath600 and @xmath601 clearly @xmath602 can be uniquely extended to a homomorphism @xmath603 , because any defining relation of @xmath596 is mapped to a relation of @xmath26 . [ thm : kernel ] suppose that @xmath604 is a graph product of groups as above and @xmath605 is a vertex of @xmath24 . then there is a group isomorphism @xmath606 , where the graph @xmath570 and the family of groups @xmath607 are defined above . in particular , every vertex group @xmath608 , @xmath609 , is isomorphic to some @xmath27 , @xmath28 . moreover , for every @xmath610 we have @xmath611 let @xmath603 be the homomorphism defined above . we will show that @xmath612 is injective and its image is @xmath613 . we will then take @xmath614 to be the inverse of @xmath612 . * injectivity of @xmath612 * let @xmath615 be a reduced word representing a non trivial element of @xmath596 . for any given @xmath616 , we know that @xmath617 for some @xmath618 . suppose that @xmath619}$ ] for some @xmath620 . if @xmath621 then @xmath622 is uniquely determined by the equivalence class @xmath623 $ ] . otherwise , if @xmath624 ( this is equivalent to @xmath625 ) , we have freedom to ( recursively ) choose @xmath618 as follows : if @xmath626 we take @xmath314 , and if @xmath627 we take @xmath628 . then @xmath612 maps the element of @xmath596 , represented by @xmath150 , to the element of @xmath26 , represented by the word @xmath629 . arguing by contradiction , suppose that @xmath630 in @xmath26 . let @xmath631 be the word obtained from @xmath632 by removing all the trivial @xmath200-syllables . since @xmath633 and @xmath65 , it follows that two syllables of @xmath631 can be joined together . however , from the choice of @xmath149 s it is clear that no two ( non - trivial ) @xmath200-syllables of @xmath631 can be joined together ( because if @xmath634 and @xmath627 then @xmath635 ) . therefore there must exist integers @xmath636 , @xmath637 , such that @xmath638 can be joined with @xmath639 in @xmath631 . there are two cases to consider . _ case 1 . _ @xmath640 for some @xmath641 . since these two syllables can be joined together in @xmath631 , we have @xmath642 for each @xmath643 . hence @xmath644 , where @xmath645 \in v{\delta}$ ] , and for every @xmath643 one has @xmath646 . in other words , the syllables @xmath647 and @xmath648 can be joined together in @xmath150 , contradicting the assumption that @xmath150 is reduced . _ case 2 . _ @xmath640 for some @xmath649 . since @xmath650 we can conclude that for all @xmath651 , @xmath652 , @xmath653 in @xmath200 , and @xmath642 for every @xmath643 . therefore @xmath644 , where @xmath654 \in v{\delta}$ ] , and for every @xmath643 one has @xmath646 . this again yields a contradiction with the assumption that @xmath150 was reduced . thus @xmath655 in @xmath26 , implying that @xmath612 is injective . * the image of @xmath612 is @xmath656 * first notice that each generator of @xmath657 is mapped under @xmath612 to an element of @xmath656 . therefore @xmath658 on the other hand , any element @xmath659 can be represented by a word @xmath660 , where @xmath661 , every @xmath662 is a subword without @xmath200-syllables and each @xmath663 . as @xmath659 we see that @xmath664 in @xmath200 . for an element @xmath665 and a word @xmath666 , with @xmath667 for some @xmath668 , @xmath669 , we define @xmath670 to be the word @xmath671 . observe that @xmath670 represents the @xmath612-image of the word @xmath672 , where @xmath673}$ ] and @xmath674 for all @xmath162 , and the length of this word is the same as the length of @xmath675 . note that in @xmath676 the element @xmath175 is also represented by the word @xmath677 evidently @xmath172 is the @xmath612-image of some word , representing an element of @xmath657 . hence @xmath612 is surjective . * rewriting in @xmath657 * let @xmath678 be the inverse of @xmath612 . from the previous construction of @xmath172 it is clear that @xmath679 . and if @xmath680 , then @xmath529 ( as @xmath681 but @xmath682 for @xmath683 ) , proving the last claim of the theorem . let @xmath24 be a simplicial graph and let @xmath45 be the graph product of a family of groups @xmath407 with respect to @xmath24 . given any subset @xmath365 , such that the parabolic closure of @xmath4 exists , the _ essential support _ @xmath684 , of @xmath4 , is the minimal subset @xmath685 such that @xmath686 for some @xmath187 . [ rem : es_supp ] if @xmath687 for some @xmath688 then @xmath689 , where @xmath690 , and , according to corollary [ cor : non - tr_parab_in_other ] , @xmath691 . in particular for every @xmath692 , @xmath431 , hence @xmath693 by corollary [ cor : non - tr_parab_in_other ] . [ rem : es_supp - prop_parab ] for any finite subset @xmath365 , the essential supports @xmath684 and @xmath694 are well - defined by proposition [ prop : min_parab ] , and @xmath695 is a finite subset of @xmath210 . moreover , @xmath696 if and only if @xmath4 is not contained in a proper parabolic subgroup of @xmath3 . [ lem : proj_on_esupp ] let @xmath12 be a subgroup of @xmath45 such that @xmath697 is well - defined ( e.g. , if @xmath24 is finite or if @xmath12 is finitely generated ) . then the canonical projection @xmath698 is injective on @xmath12 and @xmath699 , i.e. , @xmath700 is not contained in a proper parabolic subgroup of @xmath701 , where @xmath701 is equipped with the natural graph product decomposition with respect to @xmath702 . by the assumptions , @xmath703 for some @xmath187 and @xmath704 is the smallest subset of @xmath210 with this property . evidently the restriction of @xmath705 to @xmath706 acts as conjugation by @xmath707 . it follows that @xmath705 is injective on @xmath12 and @xmath12 is not contained in any proper parabolic subgroup of @xmath701 . given a graph @xmath570 and a subset @xmath708 , we will say that @xmath675 is _ irreducible _ if the full subgraph @xmath709 is irreducible . in other words , @xmath675 is not irreducible if and only if @xmath710 such that @xmath711 and @xmath712 . [ prop : irred_subset ] let @xmath24 be a finite irreducible graph with at least two vertices and let @xmath45 be the graph product of a family of groups @xmath407 with respect to @xmath24 . consider a finite subset @xmath365 such that @xmath4 is not contained in any proper parabolic subgroup of @xmath3 and @xmath713 for some @xmath560 . using theorem [ thm : kernel ] , the kernel @xmath714 can be identified with the graph product @xmath715 as above . then the essential support @xmath716 is irreducible and @xmath717 . we will use the same notation ( @xmath48 , @xmath467 , @xmath718 , etc . ) as in the beginning of the section . note that @xmath719 is well - defined by proposition [ prop : min_parab ] . let us start with observing two facts . fact 1 : for any given element @xmath567 , @xmath716 can not be contained in @xmath573 \mid u \in b\cup c\}$ ] ( otherwise @xmath720 , which is a proper parabolic subgroup of @xmath3 ) . hence there are @xmath721 and @xmath722 such that @xmath723 and @xmath724,[(g_2,u_2 ) ] \in \operatorname{esupp}_{\delta}(x)$ ] . it follows that @xmath724 \neq [ ( g_2,u_2)]$ ] , and , in particular , @xmath717 . fact 2 : for every @xmath725 there is @xmath567 such that @xmath726 \in \operatorname{esupp}_{\delta}(x)$ ] . indeed , suppose that there is @xmath725 such that @xmath727 for some @xmath728 and @xmath729 with @xmath726 \notin s$ ] for all @xmath567 . it follows , by definition , that @xmath730 where @xmath731 . then @xmath732 , which is a proper parabolic subgroup of @xmath3 , contradicting to our assumptions . arguing by contradiction , let us assume that @xmath716 is not irreducible , that is @xmath733 such that @xmath734 are non - empty and @xmath735 in @xmath570 . if @xmath736,[(f_2,v_2 ) ] \in \operatorname{esupp}_{\delta}(x)$ ] are such that @xmath737 are distinct elements of @xmath200 and @xmath738 , then these vertices are not adjacent in @xmath570 , and thus they must both lie either in @xmath739 or in @xmath740 . without loss of generality we can suppose that the vertex @xmath724 $ ] from fact 1 lies in @xmath739 . then @xmath741\in p$ ] and , in view of fact 2 , for every @xmath649 and each @xmath742 \in \operatorname{esupp}_{\delta}(x)$ ] , @xmath742 \in p$ ] ( because such a vertex can not be adjacent to both @xmath724 $ ] and @xmath741 $ ] in @xmath570 ) . consequently , @xmath743 \mid g \in g_a , \ , u \in b\}= \{[(1,u ) ] \mid u \in b\}$ ] . let @xmath744 and @xmath745 be the projections of @xmath739 and @xmath740 to @xmath746 , i.e. , @xmath747\in p \mbox { for some } f \in g_a\}\subseteq b \cup c \mbox { and } \\ \overline{q}&:=\{u \in b \mid [ ( 1,u)]\in q\}\subseteq b.\end{aligned}\ ] ] it is easy to see that @xmath744 and @xmath745 are non - empty and disjoint because @xmath739 and @xmath740 were disjoint , and @xmath748 by fact 2 . since @xmath735 , we see that for all @xmath749 and @xmath750 there is @xmath751 and such that @xmath742\in p$ ] is adjacent to @xmath752=[(f , u)]\in q$ ] in @xmath570 , which implies that @xmath130 is adjacent to @xmath129 in @xmath24 . therefore @xmath753 for all @xmath749 and all @xmath750 , i.e. , @xmath754 . recalling that @xmath755 and @xmath756 , we arrive to a contradiction with irreducibility of @xmath24 . we have shown that @xmath716 is irreducible and has at least two elements ( fact 1 ) , thus the proposition is proved . for a finite subset @xmath757 , the _ length of @xmath4 with respect to _ @xmath26 will be defined by @xmath758 [ p : comp ] suppose that @xmath24 is a simplicial graph ( not necessarily finite ) and @xmath45 is the graph product of a family of groups @xmath559 with respect to @xmath24 . let @xmath12 be a subgroup of @xmath26 and let @xmath4 be a finite subset of @xmath759 then there exist a finite graph @xmath570 , a graph product @xmath596 over @xmath570 , and a homomorphism @xmath760 with the following properties : * each vertex group @xmath608 , @xmath609 , is isomorphic to some @xmath27 , @xmath28 ; * @xmath761 ; * @xmath762 is injective on @xmath763 ; * @xmath764 ; * @xmath765 in @xmath596 for all @xmath766 . moreover , if @xmath684 contains at least two elements and is irreducible in @xmath24 then @xmath570 will be irreducible and @xmath767 . if , additionally , @xmath768 in @xmath26 for some @xmath769 , then @xmath770 the required graph product can be constructed by applying the following two - step procedure several times : step 1 . let @xmath771 be the full subgraph @xmath702 of @xmath24 , where @xmath772 . since @xmath773 one sees that @xmath704 and the graph @xmath771 are both finite ( cf . remarks [ rem : es_supp ] and [ rem : es_supp - prop_parab ] ) . note that the canonical retraction @xmath774 is length - reducing : @xmath775 , where the latter length is measured in the full subgroup @xmath701 of @xmath3 , with @xmath701 being considered as a graph product of the family @xmath776 with respect to @xmath771 . moreover , @xmath705 is injective on @xmath763 and @xmath777 by lemma [ lem : proj_on_esupp ] . set @xmath778 and @xmath779 . now , if for all @xmath780 , @xmath781 in @xmath701 then @xmath701 is the required graph product @xmath596 where @xmath782 is the restriction of @xmath705 to @xmath12 . otherwise , proceed to step 2 . suppose that there is @xmath783 such that @xmath784 in @xmath701 , i.e. , @xmath785 . then , by theorem [ thm : kernel ] , @xmath561 is a graph product @xmath596 of groups , each of which is isomorphic to some @xmath27 , @xmath28 . by our assumptions and remark [ rem : es_supp ] , there is some @xmath786 such that @xmath787 . hence , by theorem [ thm : kernel ] . , @xmath788 and @xmath789 for all @xmath790 . consequently @xmath791 we also have that @xmath792 and if @xmath793 was irreducible in @xmath24 and contained at least two elements , then @xmath794 is an irreducible graph with at least two vertices , hence @xmath795 will be irreducible in @xmath570 and @xmath796 by proposition [ prop : irred_subset ] . since the non - negative integer @xmath797 is being decreased with each application of step 2 , after repeating this two - step procedure finitely many times we will obtain a graph product with all the desired properties . the following fact ( cf . * cor . 3.28 ) ) can be deduced immediately from the above proposition : [ cor : torfree ] let @xmath495 be a prime number . if each vertex group @xmath27 , @xmath28 , has no @xmath495-torsion then the graph product @xmath45 has no @xmath495-torsion . thus if each vertex group is torsion - free then @xmath3 is torsion - free . if @xmath3 contains an element @xmath175 , of order @xmath495 , then take @xmath798 and @xmath799 . applying proposition [ p : comp ] , we will find @xmath520 a homomorphism @xmath800 such that @xmath801 . since @xmath12 is a cyclic group of prime order , the latter implies that @xmath802 is injective , hence @xmath803 has order @xmath495 . let ( p ) be a property of groups ( preserved by group isomorphisms ) and let @xmath2 be a collection of cardinals . we will say that a group is @xmath2-_locally _ ( p ) , if every non - trivial @xmath2-generated subgroup has ( p ) . in this work we will mainly be interested in the properties of being indicable and profi . a group @xmath3 is said to be _ indicable _ if it has an infinite cyclic quotient ; @xmath3 will be called _ profi _ if it possesses a proper finite index subgroup . clearly every indicable group is profi , but not vice - versa . note that if a group @xmath3 maps onto an indicable ( resp . profi ) group , then @xmath3 itself is indicable ( resp . more generally , we shall say that ( p ) is a _ lifting _ property provided the following holds : if @xmath3 is a group possessing a non - trivial quotient @xmath740 with property ( p ) then @xmath3 itself has ( p ) . thus the properties of indicability and profi are lifting properties . [ prop : lift ] suppose that @xmath2 is a collection of cardinals , @xmath24 is a simplicial graph , @xmath45 is the graph product of a family of groups @xmath559 with respect to @xmath24 . let ( p ) be a lifting property of groups . if every vertex group @xmath27 , @xmath28 , is @xmath2-locally ( p ) then the graph product @xmath3 is also @xmath2-locally ( p ) . let @xmath12 be a non - trivial @xmath2-generated subgroup of @xmath3 . pick any @xmath804 and set @xmath805 . by proposition [ p : comp ] we can find the graph product @xmath596 and a homomorphism @xmath806 enjoying all of its claims . in particular @xmath807 , hence @xmath570 has at least one vertex @xmath609 . since @xmath808 in @xmath608 and @xmath608 is @xmath7-locally ( p ) ( as it is isomorphic to a vertex group of the original graph product ) , we can conclude that @xmath809 has ( p ) , and so @xmath12 has ( p ) because this property is lifting . proposition [ prop : lift ] has the following immediate corollary , which will later be used in the proofs of theorems b and c : [ cor : loc_ind - profi ] let @xmath2 be a collection of cardinals . then the graph product of groups is @xmath2-locally indicable ( resp . @xmath2-locally profi ) if and only if each vertex group is @xmath2-locally indicable ( resp . @xmath2-locally profi ) . a group @xmath3 is _ right orderable _ if there a total order invariant under the right action of @xmath3 on itself by multiplication . clearly , subgroups of right orderable groups are right orderable . by ( 2 ) , a group is right orderable if and only if every non - trivial finitely generated subgroup maps onto a non - trivial right orderable group . using this fact together with an argument , similar the one applied in proposition [ prop : lift ] , one can achieve the following statement : [ cor : orderable ] a graph product of right orderable groups is right orderable . a different proof of corollary [ cor : orderable ] has been independently obtained by i. chiswell in @xcite , where it is also proved that a graph product of bi - orderable groups is bi - orderable . [ l : subgraph_irred ] let @xmath24 be a non - empty finite irreducible graph . then there is a vertex @xmath810 such that the full subgraph @xmath211 , of @xmath24 , is again irreducible , where @xmath465 . recall that the graph @xmath24 is irreducible if and only if the complement graph @xmath395 is connected . it is not difficult to see that @xmath395 , as any non - empty finite connected graph , has a vertex @xmath130 that is not a cut vertex . then the full subgraph of @xmath395 spanned by @xmath465 ( obtained from @xmath395 by removing @xmath130 and all the edges adjacent to it ) , is connected . thus @xmath211 is irreducible . our methods for proving of theorems b and c are similar , and so we are going to prove both of them in parallel . [ lem : h_large_mapsonto ] suppose that @xmath2 is a collection of cardinals and @xmath3 is a group splitting as a free amalgamated product @xmath811 , where @xmath812 , @xmath813 . let @xmath10 be an @xmath2-generated subgroup such that @xmath12 is not virtually cyclic , @xmath12 is not contained in a conjugate of @xmath814 , @xmath815 , in @xmath3 , and @xmath816 for all @xmath187 . then the following are true . 1 . if @xmath7 is ample and @xmath3 is @xmath7-locally profi , then @xmath12 is large . 2 . if @xmath3 is @xmath76-locally indicable , then @xmath12 maps onto @xmath1 . by a generalization of kurosh s subgroup theorem ( see ( * ? ? ? * i.5.5 , thm . 14 ) ) , @xmath817 where @xmath818 is a free group , @xmath819 is a set of indices ( possibly empty ) , and for each @xmath820 there are @xmath821 and @xmath822 such that @xmath823 . since @xmath12 is @xmath2-generated , @xmath818 and every @xmath824 are also @xmath7-generated , as retracts of @xmath12 . if the rank of @xmath818 is at least @xmath0 , then @xmath12 maps onto @xmath1 and both ( i ) and ( ii ) hold . thus we can further assume that @xmath818 is either trivial or is isomorphic to @xmath825 \(i ) if @xmath826 , since @xmath12 is not cyclic we have @xmath827 , thus there exists some @xmath828 . since @xmath3 is @xmath7-locally profi , @xmath829 possesses a normal subgroup @xmath830 of finite index @xmath537 . then @xmath12 has an epimorphism onto the virtually free group @xmath831 . the above conditions on @xmath20 and @xmath818 imply that @xmath832 contains a non - abelian free subgroup , and so it is not virtually cyclic . thus @xmath832 is large and , hence , so is @xmath12 . in the case when @xmath833 , @xmath819 must have at least two elements because @xmath12 is not contained in a conjugate of @xmath814 , @xmath815 , by assumptions . if @xmath834 , choose three distinct indices @xmath835 . since @xmath3 is @xmath7-locally profi , each @xmath836 possesses a normal subgroup @xmath837 of finite index @xmath838 , @xmath839 . then @xmath12 maps onto the free product @xmath840 , of three non - trivial finite subgroups , and hence it is large . if @xmath841 and @xmath842 for @xmath843 , then @xmath844 contradicting to @xmath12 not being virtually cyclic . thus without loss of generality we can assume that @xmath845 . again , since @xmath3 is @xmath7-locally profi , @xmath836 possesses a normal subgroup @xmath837 of finite index @xmath838 for @xmath843 . moreover , if @xmath846 then @xmath847 and it is @xmath7-generated by lemma [ lem : card_of_fi ] , hence @xmath848 is profi and it must contain a proper finite index subgroup . therefore we can assume that @xmath849 . then @xmath12 maps onto the free product @xmath850 of two finite subgroups , of orders @xmath851 and @xmath852 , and so @xmath12 is large . this completes the proof for ( i ) . \(ii ) if @xmath826 , since @xmath12 is not cyclic there exists @xmath828 . by the grushko - neumann theorem ( * ? ? ? * i.10.6 ) , @xmath829 is @xmath76-generated because @xmath12 is @xmath2-generated . hence @xmath829 maps onto @xmath30 as @xmath3 is @xmath76-locally indicable . it follows that @xmath12 maps onto @xmath853 which is isomorphic to @xmath1 . if @xmath833 , the assumptions imply that @xmath854 , so we can find distinct @xmath855 . since both @xmath829 and @xmath856 are non - trivial and @xmath12 is @xmath2-generated , the grushko - neumann theorem implies that these free factors are @xmath76-generated . the @xmath76-local indicability of @xmath3 yields that @xmath829 and @xmath856 each map onto @xmath30 , hence @xmath12 maps onto @xmath857 , as required . for the rest of this section @xmath24 will be a finite simplicial graph , @xmath858 will be a family of groups and @xmath558 will be the corresponding graph product . before proving the main results , we are going to establish two sufficient criteria for a subgroup of a graph product to be large or to map onto @xmath1 . the following elementary observation will be useful : [ rem : non - tr_proj ] if @xmath10 non - trivially projects to every vertex group then @xmath12 is not contained in any proper parabolic subgroup of @xmath3 . [ thm : prebc ] let @xmath2 be a collection of cardinals and let @xmath24 be a finite irreducible graph with @xmath859 . suppose that @xmath555 is an @xmath2-generated subgroup such that @xmath12 is not virtually cyclic and @xmath860 for every @xmath861 1 . if @xmath7 is ample and @xmath27 is @xmath7-locally profi for every @xmath28 , then @xmath12 is large . 2 . if @xmath27 is @xmath76-locally indicable for each @xmath28 , then @xmath12 maps onto @xmath1 . if the collection @xmath7 contained no cardinals greater than @xmath171 , we would get a contradiction with our assumptions because only cyclic subgroups could be @xmath2-generated in this case , but @xmath12 is not virtually cyclic by the assumptions . therefore @xmath76 contains at least one non - zero cardinal , implying that any cyclic subgroup of @xmath3 is @xmath862-generated . observe , also , that , by corollary [ cor : loc_ind - profi ] , in case ( a ) @xmath3 is @xmath7-locally profi and in case ( b ) @xmath3 is @xmath862-locally indicable . the proof will proceed by induction on @xmath557 . base of induction : @xmath863 . suppose that @xmath864 and set @xmath562 , @xmath865 . in this case @xmath866 ( as @xmath24 is irreducible ) and @xmath12 is not contained in a conjugate of @xmath200 or @xmath468 by remark [ rem : non - tr_proj ] . therefore all the assumptions of lemma [ lem : h_large_mapsonto ] are satisfied , hence @xmath12 has the required property . step of induction : @xmath867 and we can suppose that the required statement has been proved for all @xmath2-generated subgroups , of ( possibly different ) graph products over finite irreducible graphs with fewer vertices , mapping non - trivially to each vertex group . by lemma [ l : subgraph_irred ] there is a vertex @xmath868 such that the full subgraph @xmath211 , of @xmath24 , is again irreducible , where @xmath465 . let @xmath869 and @xmath870 , then @xmath3 splits as the amalgamated free product @xmath871 . note that @xmath872 and @xmath873 because @xmath24 is irreducible , hence @xmath48 and @xmath467 are proper subsets of @xmath353 . recall that for the non - empty proper subset @xmath874 @xmath875 is a graph product with fewer vertices than @xmath3 , where @xmath876 , and @xmath542 is @xmath2-generated . moreover , @xmath542 projects non - trivially to each vertex group of @xmath200 , because for every @xmath205 the canonical retraction @xmath877 factors through the retraction @xmath878 ( that is , @xmath879 ) . thus if @xmath542 is not virtually cyclic in @xmath200 , by induction we know that @xmath542 satisfies the required properties ( @xmath542 is large in case ( a ) or maps onto @xmath1 in case ( b ) ) , and these properties are inherited by @xmath12 . hence we can suppose that @xmath542 is virtually cyclic . applying theorem [ thm : structure ] and remark [ rem : non - tr_proj ] , we see that either @xmath880 or @xmath881 _ case 1 . _ if @xmath882 , then @xmath542 has trivial intersection with any proper parabolic subgroup of @xmath200 by remark [ rem : non - tr_proj ] and lemma [ lem : virt_cyc - parab ] . take any @xmath187 , set @xmath883 and , recalling the inclusion @xmath873 , observe that @xmath884 since @xmath885 is a proper parabolic subgroup of @xmath200 . by remark [ rem : non - tr_proj ] and our hypothesis , @xmath12 is not contained in a conjugate of @xmath200 or @xmath468 , therefore we can apply lemma [ lem : h_large_mapsonto ] to reach the necessary conclusion . _ if @xmath886 then @xmath3 contains @xmath0-torsion , hence @xmath27 contains @xmath0-torsion for some @xmath47 , by corollary [ cor : torfree ] . the latter is impossible in case ( b ) since the cyclic group of order @xmath0 is @xmath76-generated but is not indicable . hence we are in case ( a ) , and we need to show that @xmath12 is large . by remark [ rem : non - tr_proj ] and lemma [ lem : virt_cyc - parab ] , @xmath542 has finite intersection with any proper parabolic subgroup of @xmath200 , hence the infinite cyclic subgroup @xmath887 , of index @xmath0 , will have trivial intersection with each proper parabolic subgroup of @xmath200 . set @xmath888 and note that @xmath889 , @xmath890 and @xmath891 . by the same argument as in case 1 , @xmath892 for all @xmath187 . if @xmath893 were contained in a conjugate of @xmath200 or @xmath468 , then @xmath894 would be contained in a proper parabolic subgroup of @xmath3 by corollary [ cor : norm_in_irred ] , contradicting our assumptions together with remark [ rem : non - tr_proj ] . we also need to recall lemma [ lem : card_of_fi ] , which claims that @xmath893 is @xmath7-generated . thus we see again that all of the conditions of lemma [ lem : h_large_mapsonto].(i ) are satisfied , and so @xmath893 is large . and since @xmath893 has index @xmath0 in @xmath12 , we are able to conclude that @xmath12 is large as well . [ thm : main_crit ] let @xmath2 be a collection of cardinals and let @xmath24 be a finite graph . suppose that @xmath555 is an @xmath2-generated subgroup such that @xmath12 is not virtually cyclic and @xmath895 is an irreducible subset of @xmath210 with @xmath896 . 1 . if @xmath7 is ample and @xmath27 is @xmath7-locally profi for every @xmath28 , then @xmath12 is large . 2 . if @xmath27 is @xmath76-locally indicable for each @xmath28 , then @xmath12 maps onto @xmath1 . according to lemma [ lem : proj_on_esupp ] , we can project everything onto @xmath701 , where @xmath897 , and further assume that @xmath898 , that is @xmath12 is not contained in any proper parabolic subgroup of @xmath3 . by theorem [ thm : structure ] and remark [ rem:4-core ] , @xmath12 contains a finitely generated subgroup @xmath509 such that @xmath509 is not virtually cyclic and @xmath899 . choose any finite generating set @xmath4 of @xmath509 and apply proposition [ p : comp ] to find a finite irreducible graph @xmath570 with at least @xmath0 vertices , a graph product @xmath900 , and a homomorphism @xmath901 such that every vertex group @xmath608 , @xmath609 , is isomorphic to some @xmath27 , @xmath28 , @xmath762 is injective on @xmath509 and @xmath902 projects non - trivially to each vertex group of @xmath116 . since @xmath903 is not virtually cyclic , we can conclude that @xmath902 is not virtually cyclic . therefore both claims of the theorem now follow from theorem [ thm : prebc ] . we are now in a position to prove theorems b and c. the argument will proceed by induction on @xmath557 . note that the base of our induction , @xmath904 , holds because the vertex groups satisfy the corresponding form of tits alternative . therefore we can suppose that @xmath905 and the required statement has been proved for all ( possibly different ) graph products over graphs with fewer vertices . observe that , in theorem b , the strong tits alternative for @xmath27 together with ( p@xmath49 ) imply that @xmath27 is @xmath7-locally profi for all @xmath28 . similarly , in theorem c , every vertex group @xmath27 will be @xmath76-locally indicable by the strongest tits alternative combined with ( p@xmath83 ) . let @xmath12 be an @xmath2-generated subgroup of @xmath906 by theorem a we know that if @xmath12 does not contain a copy of @xmath1 , then @xmath12 belongs to @xmath554 thus we can further suppose that @xmath12 is not virtually cyclic . for every non - empty proper subset @xmath874 @xmath875 is a graph product with fewer vertices than @xmath3 , where @xmath876 , and @xmath542 is @xmath2-generated . then , by the induction hypothesis we know that either @xmath542 is in @xmath42 or is large ( in theorem b ) , or maps onto @xmath1 ( in theorem c ) . in the latter two cases @xmath12 inherits the needed properties from @xmath542 . summarizing , we can restrict to the situation when @xmath12 is not virtually cyclic and for all @xmath874 @xmath542 belongs to @xmath42 . if @xmath24 is not irreducible then there exist @xmath907 such that @xmath908 in this case @xmath909 belongs to @xmath42 by ( p@xmath171 ) and ( p@xmath0 ) . suppose that @xmath24 is irreducible . if @xmath910 , then @xmath911 ( see lemma [ lem : proj_on_esupp ] ) belongs to @xmath8 . thus we can assume that @xmath912 , which enables us to apply theorem [ thm : main_crit ] and achieve that either @xmath12 is large ( in theorem b ) or @xmath12 maps onto @xmath1 ( in theorem c ) . the goal of this section is to show that the claims of corollaries [ cor:2-gen ] and [ cor : raag ] from the introduction can not be improved in an obvious way . for instance , it is natural to ask , whether , after dropping the torsion - freeness assumption from corollary [ cor:2-gen ] , one can still describe the structure of an arbitrary @xmath0-generated subgroup . the example below shows that this may not be easy , as such a subgroup can be large with infinite center ; in particular , it will not be virtually free . [ ex:2-gen_non_virt_cyc ] let @xmath24 be the simple path of length @xmath0 ( with @xmath84 vertices and @xmath0 edges ) , and let @xmath3 be the graph product with respect to @xmath24 , such that the vertex group in the middle is @xmath30 and the vertex groups at the end - points are @xmath913 . thus @xmath3 has the presentation @xmath914 . consider the @xmath0-generated subgroup @xmath915 . it is not difficult to see that @xmath12 is isomorphic to the group given by the presentation @xmath916 ( under the map sending @xmath229 to @xmath917 and @xmath230 to @xmath918 ) . it follows that @xmath12 has infinite cyclic center generated by @xmath919 , and the quotient of @xmath12 by its center is isomorphic to the free product @xmath920 , which is virtually a non - abelian free group . [ prop : rips_constr ] for every finitely generated recursively presented group @xmath739 there exists a finitely generated group @xmath116 and a finitely generated normal subgroup @xmath921 such that all of the following hold : first , using the celebrated higman embedding theorem @xcite , we embed @xmath739 into a finitely presented group @xmath923 . next we apply ( * ? ? ? 3.1 ) to embed @xmath923 into an infinite finitely presented group @xmath924 such that @xmath924 has no proper finite index subgroups . applying the haglund - wise modification of rips s construction ( see ( * ? ? ? 10.1 ) ) , we find a finitely generated group @xmath893 and a finitely generated normal subgroup @xmath925 such that @xmath926 . moreover , by theorem 5.9 from @xcite , we know that some finite index subgroup @xmath927 is a subgroup of some finitely generated right angled artin group @xmath3 . let @xmath928 be an epimorphism with @xmath929 . denote by @xmath930 the restriction of @xmath931 to @xmath12 ; then @xmath902 will have finite index in @xmath924 , hence @xmath932 by the construction of @xmath924 . since @xmath933 , we can define @xmath934 by @xmath935 . note that @xmath936 is finitely generated because it is a finite index subgroup of @xmath937 , which is finitely generated , and @xmath922 . since @xmath739 is finitely generated , it follows that @xmath116 is also finitely generated . suppose that @xmath480 is abelian . it is well - known that finitely generated right angled artin groups are linear over @xmath30 ( for instance , because they can be embedded into finitely generated right angled coxeter groups ( * ? ? ? 3.6 ) , and the standard geometric representation of the latter is a faithful representation by matrices with integer coefficients ( * ? ? ? * cor . 5.4 ) ) . hence @xmath10 can be considered as a subgroup of @xmath938 for some @xmath54 . d. segal proved ( see ( * ? ? ? * thm . 5 , p. 61 ) ) that every solvable subgroup of @xmath938 coincides with the intersection of finite index subgroups containing it ( in other words , it is closed in the profinite topology of that group ) . therefore the abelian subgroup @xmath480 will equal to an intersection of finite index subgroups of @xmath12 , which is equivalent to saying that @xmath939 is residually finite . but this is impossible because @xmath924 is an infinite group without proper subgroups of finite index . this contradiction shows that @xmath480 must be non - abelian , and so , by corollary [ cor : raag ] , both @xmath480 and @xmath116 possess epimorphisms onto @xmath1 . recall that the _ rank _ @xmath940 , of a group @xmath12 , is the least cardinality @xmath77 such that @xmath12 can be generated by a subset @xmath941 with @xmath99 . in the context of this work it is natural to define the _ free rank _ @xmath942 , of @xmath12 , as the supremum of all cardinals @xmath77 such that @xmath12 has an epimorphism onto a free group of rank @xmath77 . evidently , @xmath943 , and in view of corollary [ cor : raag ] , it makes sense to ask whether one can find a lower bound for @xmath942 in terms of @xmath940 for subgroups @xmath12 of a given finitely generated right angled artin group @xmath3 . note that for any non - trivial abelian subgroup @xmath10 , @xmath944 ( see corollary [ cor : rank_ab_in_raag ] ) , and so @xmath945 . however , the next statement shows that no such lower bound is possible in general . [ prop : ex_fr_rank ] there exists a finitely generated right angled artin group @xmath3 , a natural number @xmath948 and a collection of subgroups @xmath949 of @xmath3 such that @xmath950 and @xmath951 for all @xmath952 . in particular , @xmath953 , @xmath954 for all @xmath955 , and @xmath956 as @xmath957 . let @xmath958 be the ( restricted ) wreath product of two infinite cyclic groups . then @xmath739 is @xmath0-generated , recursively presented and contains a subgroup @xmath740 which is isomorphic to the free abelian group @xmath959 , of rank @xmath960 . according to proposition [ prop : rips_constr ] , there is a finitely generated right angled artin group @xmath3 and finitely generated subgroups @xmath961 such that @xmath962 and @xmath922 . let @xmath963 denote an epimorphism with @xmath964 . for each @xmath955 choose a subgroup @xmath965 of rank @xmath162 and let @xmath966 . now , let @xmath824 , @xmath967 , denote the full @xmath762-preimage of @xmath968 in @xmath116 and set @xmath969 . clearly @xmath970 for all @xmath967 . on the other hand , if @xmath967 and @xmath614 is an epimorphism from @xmath824 onto some free group @xmath971 , then @xmath972 is a finitely generated normal subgroup of @xmath971 . if @xmath973 then @xmath614 factors through the restriction of @xmath762 to @xmath824 , i.e. , @xmath971 is a homomorphic image of @xmath974 , which is a free abelian group . hence @xmath975 . if @xmath976 then @xmath972 must have finite index in @xmath971 ( see ( * ? ? ? * i.3.12 ) ) , hence , recalling schreier index formula ( ( * ? ? ? * i.3.9 ) ) , we obtain @xmath977 . therefore @xmath950 for all @xmath967 , as claimed . abraham karrass , alfred pietrowski and donald solitar , _ finite and infinite cyclic extensions of free groups_. collection of articles dedicated to the memory of hanna neumann , iv . j. austral . * 16 * ( 1973 ) , 458466 . jean - pierre serre , _ trees . _ translated from the french original by john stillwell . corrected 2nd printing of the 1980 english translation . springer monographs in mathematics . springer - verlag , berlin , 2003 .

we discuss various types of tits alternative for subgroups of graph products of groups , and prove that , under some natural conditions , a graph product of groups satisfies a given form of tits alternative if and only if each vertex group satisfies this alternative . as a corollary , we show that every finitely generated subgroup of a graph product of virtually solvable groups is either virtually solvable or large . as another corollary , we prove that every non - abelian subgroup of a right angled artin group has an epimorphism onto the free group of rank @xmath0 . in the course of the paper we develop the theory of parabolic subgroups , which allows to describe the structure of subgroups of graph products that contain no non - abelian free subgroups . we also obtain a number of results regarding the stability of some group properties under taking graph products .

the baryon ( @xmath10 ) and lepton ( @xmath11 ) numbers are considered to be conserved in the standard model ( sm ) and @xmath11 violation even in the sm @xcite . ] . however , no symmetry principle underlies these laws , such as , e.g. gauge invariance , which guarantees conservation of the electric charge . many extensions of the sm include @xmath10 and @xmath11 violating interactions , predicting the decay of protons and neutrons bounded in nuclei . various decay mechanisms with @xmath12=1 , 2 and @xmath13@xmath14=0 , 2 have been discussed in the literature intensively @xcite . a novel baryon number violating process , in which two neutrons in a nucleus disappear , emitting a bulk majoron @xmath15 , was proposed recently @xcite ; the expected mean lifetime was estimated to be @xmath1610@xmath17 y. additional possibilities for the nucleon ( @xmath1 ) decays are related to theories which describe our world as a brane inside higher - dimensional space @xcite . particles , initially confined to the brane , may escape to extra dimensions , thus disappearing for the normal observer ; the characteristic proton mean lifetime was calculated to be @xmath18=9.2@xmath19 y @xcite . observation of the disappearance of @xmath20 , @xmath1 , @xmath2 would be a manifestation of the existence of such extra dimensions @xcite . no evidence for nucleon instability has been found to date . experimental searches @xcite with the imb , frjus , ( super)kamiokande and other detectors have been devoted mainly to nucleon decays into strongly or electromagnetically interacting particles , where lower limits on the nucleon mean lifetime of @xmath21 y were obtained @xcite . at the same time , for modes where @xmath1 or @xmath2 pairs disappear or they decay to some weakly interacting particles ( neutrinos , majorons , etc . ) , the experimental bounds are a few orders of magnitude lower . different methods were applied to set limits for such decays ) means disappearance or decay to weakly interacting particles ( one or few neutrinos of any flavors , majorons , etc . ) . channel to @xmath22 , mentioned below , includes also decays to @xmath0 . ] ( see table 1 for summary ) : explanation + & & and c.l . & + @xmath23 & @xmath24 @xmath22 & 1.2@xmath2510@xmath26 & 1958 @xcite limit on @xmath27th spontaneous fission + & & 3.0@xmath2510@xmath26 & 1970 @xcite search for free @xmath28 in liquid scintillator + & & & enriched in deuterium ( @xmath29 ) + & & 4.0@xmath2510@xmath26 95% & 2001 @xcite free @xmath28 in d@xmath30o volume + & @xmath24 @xmath0 & 7.4@xmath2510@xmath31 & 1977 @xcite geochem . search for @xmath32te@xmath24 ... @xmath24@xmath33xe + & & 1.1@xmath2510@xmath34 & 1978 @xcite radiochem . search for @xmath35k@xmath24 ... @xmath24@xmath36ar + & & 1.9@xmath2510@xmath31 90% & 2000 @xcite search for @xmath37i decay in @xmath33xe detector + & & @xmath1610@xmath38 & 2002 @xcite free @xmath28 in d@xmath30o volume + & & 3.5@xmath2510@xmath38 90% & 2003 @xcite free @xmath28 in d@xmath30o volume + & & 1.1@xmath2510@xmath34 90% & 2003 [ @xmath39 ] search for @xmath3b decay in the ctf detector + @xmath28 & @xmath24 @xmath22 & 1.8@xmath2510@xmath26 & 1958 @xcite limit on @xmath27th spontaneous fission + & @xmath24 @xmath0 & 8.6@xmath2510@xmath31 & 1977 @xcite geochem . search for @xmath32te@xmath24 ... @xmath24@xmath33xe + & & 1.1@xmath2510@xmath34 & 1978 @xcite radiochem . search for @xmath35k@xmath24 ... @xmath24@xmath36ar + & & 4.9@xmath2510@xmath34 90% & 1993 @xcite search for @xmath40 with @xmath41=1950 mev emitted in + & & & @xmath42o deexcitation in kamiokande detector + & & 1.8@xmath2510@xmath43 90% & 2003 [ @xmath39 ] search for @xmath44c decay in the ctf + & @xmath24 @xmath45 & 5.0@xmath2510@xmath34 90% & 1979 @xcite massive liquid scint . detector fired by @xmath46 + & & & in result of @xmath28 decays in the whole earth @xmath47 + & & 1.2@xmath2510@xmath34 90% & 1991 @xcite frjus iron detector fired by @xmath46 @xmath48 + & @xmath24 @xmath49 & 3.0@xmath2510@xmath43 90% & 1991 @xcite frjus iron detector fired by @xmath50 @xmath51 + & @xmath24 @xmath52 & 2.3@xmath2510@xmath53 90% & 1997 @xcite search for bremsstr . @xmath40 with @xmath41@xmath54100 mev + & & & emitted due to sudden disapp . of @xmath28 magn . + & & & moment ( from kamiokande data ) @xmath55 + & @xmath24 @xmath56 & 1.7@xmath2510@xmath53 90% & 1997 @xcite the same approach @xmath55 + @xmath57 & @xmath24 @xmath58 & 6.0@xmath2510@xmath31 90% & 1991 @xcite frjus iron detector fired by @xmath46 @xmath59 + & @xmath24 @xmath60 & 1.2@xmath2510@xmath43 90% & 1991 @xcite frjus iron detector fired by @xmath50 @xmath61 + & @xmath24 @xmath0 & 1.2@xmath2510@xmath43 90% & 2000 @xcite search for @xmath62xe decay in @xmath33xe detector + & & 4.2@xmath2510@xmath43 90% & 2002 @xcite radiochem . search for @xmath35k@xmath24 ... @xmath24@xmath36ar @xmath63 + & & 4.9@xmath2510@xmath43 90% & 2003 [ @xmath39 ] search for @xmath64c and @xmath65o decay in the ctf + @xmath66 & @xmath24 @xmath0 & 5.5@xmath2510@xmath26 90% & 2000 @xcite search for @xmath62te decay in @xmath33xe detector + & & 5.0@xmath2510@xmath43 90% & 2003 [ @xmath39 ] search for @xmath44be decay in the ctf detector + @xmath67 & @xmath24 @xmath0 & 2.1@xmath2510@xmath43 90% & 2002 @xcite radiochem . search for @xmath35k@xmath24 ... @xmath24@xmath36ar @xmath63 + + + + + + + + + \(1 ) using the limit on the branching ratio of spontaneous fission of @xmath27th under the assumption that @xmath23 or @xmath28 decay in @xmath27th will destroy the nucleus @xcite . the bound on the mean lifetime obtained in this way can be considered independent of the a @xmath23 or @xmath28 decay mode , since the @xmath27th nucleus can be destroyed either by strong or electromagnetic interactions of daughter particles with the nucleus or , in the case of @xmath1 disappearance , by subsequent nuclear deexcitation process ; \(2 ) search for a free @xmath28 created after @xmath23 decay or disappearance in the deuterium nucleus ( @xmath68=@xmath67 ) in a liquid scintillator enriched in deuterium @xcite or in a volume of d@xmath30o @xcite ; \(3 ) geochemical @xcite or radiochemical @xcite search for daughter nuclides which have appeared after @xmath1 decays in the mother nuclei ( valid for decays into invisible channels ) ; \(4 ) search for prompt @xmath69 quanta emitted by a nucleus in a de - excitation process after @xmath1 decays within the inner nuclear shell @xcite ( valid for invisible channels ) ; \(5 ) considering the earth as a target with nucleons which decay by emitting electron or muon neutrinos ; the @xmath50 , @xmath46 can be detected by a large underground detector @xcite ( valid for decay into neutrinos with specific flavors ) ; \(6 ) search for bremsstrahlung @xmath40 quanta emitted due to a sudden disappearance of the neutron magnetic moment @xcite ( limits depend on the number of emitted neutrinos ) ; \(7 ) study of radioactive decay of daughters ( time - resolved from prompt products ) , created as a result of @xmath1 or @xmath2 decays of the mother nuclei , incorporated into a low - background detector ( valid for decay into invisible channels ) . this method was first exploited by the dama group with a liquid xe detector @xcite . in the present paper we use the same approach to search for @xmath1 and @xmath2 instability with the counting test facility , a 4.2 ton prototype of the multiton borexino detector for low energy solar neutrino spectroscopy @xcite . the preliminary results were presented in @xcite . borexino , a real - time 300 ton detector for low - energy neutrino spectroscopy , is nearing completion in the gran sasso underground laboratory ( see @xcite and refs . therein ) . the main goal of the detector is the measurement of the @xmath70be solar neutrino flux via @xmath71 scattering in an ultra - pure liquid scintillator , while several other basic questions in astro- and particle physics will also be addressed . the counting test facility ( ctf ) , installed in the gran sasso underground laboratory , is a prototype of the borexino detector . detailed reports on the ctf results have been published @xcite , and only the main characteristics of the set - up are outlined here . the ctf consists of an external cylindrical water tank ( @xmath7211@xmath7310 m ; @xmath741000 t of water ) serving as passive shielding for 4.2 m@xmath75 of liquid scintillator ( ls ) contained in an inner spherical vessel of @xmath762.0 m. high purity water with a radio - purity of @xmath77@xmath78 g / g ( u , th ) , @xmath79@xmath80 g / g ( k ) and @xmath812 @xmath82bq/_l _ for @xmath83rn is used for the shielding . the ls was purified to the level of @xmath8410@xmath85 g / g in u / th contamination . we analyze here the data following the upgrade of the ctf ( ctf - ii ) . the liquid scintillator used at this stage was a phenylxylylethane ( pxe , c@xmath86h@xmath87 ) with p - diphenylbenzene ( para - terphenyl ) as a primary wavelength shifter at a concentration of 2 g/_l _ along with a secondary wavelength shifter 1,4-bis-(2-methylstyrol)-benzene ( bis - msb ) at 50 mg/_l_. the density of the scintillator is 0.996 kg/_l_. the scintillator principal de - excitation time is less than 5 ns which provides a good position reconstruction . in the ctf - ii an additional nylon screen between the scintillator vessel and pmts ( against radon penetration ) and a muon veto system were installed . the scintillation light is collected with 100 phototubes ( pmt ) fixed to a 7 m diameter support structure inside the water tank . the pmts are fitted with light concentrators which provide a total of 21% optical coverage . the number of photoelectrons measured experimentally is 3.54 per pmt for 1 mev electrons at the detector s center . for each event the charge and timing of hit pmts are recorded . each channel is supported by an auxiliary channel , the so - called second group of electronics , used to record all events coming within a time window of 8.2 ms after the trigger , which allows tagging of fast time - correlated events with a decrease of the overall dead time of the detector . for longer delays , the computer clock is used providing the accuracy of @xmath740.1 s. event parameters measured in the ctf include the total charge collected by the pmts during 0500 ns , used to determine an event s energy ; the charge in the `` tail '' of the pulse ( 48548 ns ) which is used for pulse shape discrimination ; pmt timing , used to reconstruct the event s position ; and the time elapsed between sequential events , used to tag time - correlated events . the energy of an event in the ctf detector is defined using the total collected charge from all pmt s . in a simple approach the energy is supposed to be linear with respect to the total collected charge . the coefficient linking the event energy and the total collected charge is called light yield ( or photoelectron yield ) . the light yield for electrons can be considered linear with respect to its energy only for energies above 1 mev . at low energies the phenomenon of `` ionization quenching '' violates the linear dependence of the light yield versus energy @xcite . the deviations from the linear law can be taken into account by the ionization deficit function @xmath88 , where @xmath89 is an empirical birks constant . for the calculations of the ionitazion quenching effect for pxe scintillator we used the kb program from the cpc library @xcite . the `` ionization quenching '' effect leads to a shift in the position of the full energy peak for gammas on the energy scale calibrated using electrons . in fact , the position of the 1461 kev @xmath90k gamma in the ctf - ii data corresponds to 1360 kev of energy deposited for an electron . the detector energy and spatial resolution were studied with radioactive sources placed at different positions inside the active volume of the ctf . a typical spatial 1@xmath91 resolution is 10 cm at 1 mev . the studies showed also that the total charge response of the ctf detector can be approximated by a gaussian . for energies @xmath92@xmath931 mev ( which are of interest here ) , the relative resolution can be expressed as @xmath94 ( @xmath92 is in mev ) @xcite for events uniformly distributed over the detector s volume . the energy dependence on the collected charge becomes non - linear for the energies @xmath95 mev in the first group of electronics because of the saturation of the adcs used . in this region we are using only the fact of observing or not - observing candidate events , hence the mentioned nonlinearity does nt influence the result of the analysis . further details on the energy and spatial resolutions of the detector and ionization quenching for electrons , @xmath40 quanta and @xmath96 particles can be found in @xcite . the upgrade of the ctf was equipped with a carefully designed muon veto system . it consists of 2 rings of 8 pmts each , installed at the bottom of the tank . the radii of the rings are 2.4 and 4.8 m. muon veto pmts are looking upward and have no light concentrators . the muon veto system was optimized in order to have a negligible probability of registering the scintillation events in the so - called `` neutrino energy window '' ( 250800 kev ) . the behaviour of the muon veto at the higher energies has been specially studied for the present work . experimental measurements with a radioactive source ( chain of @xmath97ra ) @xcite gave , for the probability @xmath98 of identification of an event with energy @xmath92 in the ls by the muon veto , the value of ( [email protected])% in the 1.82.0 mev region ( see fig . 3 later ) . the energy dependence of @xmath98 was also calculated by a ray - tracing monte carlo method accounting for specific features of the light propagation in the ctf which are detailed in @xcite . the calculated function was adjusted to reproduce correctly the experimental measurements with the @xmath97ra source . as it will be shown below , the candidate events , relevant for our studies , have to satisfy the following criteria : ( 1 ) the event should occur in the active volume of the detector and must not be accompanied by the muon veto tag ; ( 2 ) it should be single ( not followed by a time - correlated event ) ; ( 3 ) its pulse shape must correspond to that of events caused by @xmath40 or @xmath100 particles . the selection and treatment of data ( spatial cuts , analysis of an event s pulse shape to distinguish between electrons and @xmath96 particles , suppression of external background by the muon veto system , etc . ) is similar to that in ref . @xcite . the experimental energy spectra in ctf - ii , accumulated during 29.1 days of measurements , are shown in fig . the spectrum without any cuts ( spectrum 1 ) is presented on the top . the second spectrum is obtained by applying the muon cut , which suppressed the background rate by up to two orders of magnitude , depending on the energy region . on the next stage of the data selection we applied a cut on the reconstructed radius . in the energy region 12 mev we used @xmath101100 cm cut aiming to remove the surface background events ( mainly due to the @xmath90k decays outside the inner vessel ) and leave the events uniformly distributed over the detector volume . the efficiency of the cut has been studied with mc simulation and lays in the range of @xmath102 in the energy region 12 mev . additional @xmath103 discrimination @xcite was applied to eliminate contribution from @xmath96 particles ( spectrum 3 in fig . the time - correlated events ( that occurred in the time window @[email protected] ms ) were also removed ( spectrum 4 ) . the peak at 1.46 mev , present in all spectra , is due to @xmath90k decays outside the scintillator , mainly in the ropes supporting the nylon sphere . the peak - like structure at @xmath166.2 mev is caused by saturation of the electronics by high - energy events . the lower spectrum of fig . 1 presents all the candidate scintillation events in the search for the decay of radioactive nuclides created in the active volume of the ctf after the nucleon disappearance in parent nuclei . the decay characteristics of the daughter nuclides , resulting from @xmath1 and @xmath2 decays in parent nuclei @xmath3c , @xmath4c and @xmath5o contained in the sensitive volume of the ctf liquid scintillator or in the water shield , are listed in table 2 . .initial nuclei in the counting test facility , their abundance @xmath105 @xcite , processes with @xmath106 and characteristics of daughter nuclides @xcite . [ cols= " < , < , < , < , < " , ] after the disappearance of one or two nucleons in the parent nuclide , one or two holes appear in the nuclear shells ; these holes will be filled in a subsequent nuclear de - excitation process , unless the nucleons reside on the outermost shells . if the initial excitation energy , @xmath107 , is higher than the binding energy @xmath108 of the least bound nucleon ( @xmath1 is @xmath23 or @xmath28 ) , the nucleus will be de - excited by particle emission ( @xmath23 , @xmath28 , @xmath96 , @xmath68 , etc . ) ; otherwise ( @xmath109 ) a @xmath40 quantum will be emitted . in the following , we will take into consideration the @xmath1 and @xmath2 decays from the last filled single - particle levels , when only @xmath40 quanta could be emitted . thus the daughter nucleus is exactly known . we also neglect by the prompt signal from gamma quanta emitted in initial deexcitation process ( in particular , because of generally big half lives of daughter nuclei , and uncertainty in the number of emitted @xmath40 s , their energies , and rejection of high energy events by the muon veto ) , and search only for subsequent radioactive decay of created nuclei . the number of nucleons or nucleon pairs participating in the process can be calculated following refs . @xcite . for example , after the decay of a neutron with binding energy @xmath110 , the excitation energy of the @xmath111 daughter nucleus will be @xmath112 . the condition to emit only @xmath40 quanta in the deexcitation process , @xmath113 , gives this restriction on the neutron binding energy : @xmath114 similar equations can be written for @xmath23 , @xmath66 , @xmath67 and @xmath57 decays ( see ref . the values of the separation energies @xmath108 and @xmath115 were taken from ref . @xcite . for single - particle energies of nucleons @xmath116 on nuclear shells we used the continuum shell model calculations @xcite for @xmath3c and @xmath4c , and the hartree - fock calculations with the skyrme s interaction @xcite for @xmath5o the expected response functions of the ctf detector and related efficiencies for the decay of unstable daughter nuclei were simulated with the egs4 package @xcite . the number of initial electrons and @xmath40 quanta emitted in the decay of the nucleus and their energies were generated according to the decay schemes @xcite . the events were supposed to be uniformly distributed in the whole volume of the liquid scintillator ( and in a water layer close to the ls ) . the energy and spatial resolution of the detector @xcite , light quenching factors for electrons and gammas @xcite , @xmath117 veto and triggering efficiency were taken into account in the simulations . the calculated responses for the decay of @xmath44c and @xmath64c in the liquid scintillator ( created after @xmath28 and @xmath57 disappearance in @xmath3c , respectively ) and @xmath65o in the water shield ( @xmath57 decay in @xmath5o ) are shown in fig . 2 . in the last case only the water layer of 1 m thickness closest to the liquid scintillator was taken into consideration . contributions from layers further out are negligible . the ctf background was simulated as well in order to check the understanding of the detector . the simulated @xmath90k 1.46 mev gamma peak together with experimental data is shown in the inset of fig . the shift in the position of the full absorption peak is due to the ionization quenching effect ( see section 2.2 ) . one can see the good agreement between the experimental and simulated @xmath90k peak position and resolution . the experimental data ( see fig . 2 , where the spectrum with all cuts is shown in detail ) gives no strong evidence of the expected @xmath1 and @xmath2 decay response functions , thus allowing only bounds to be set on the processes being searched for . in the present study , in order to extract the limits on the relevant mean lifetimes , we assumed conservatively that _ all _ events in the ctf experimental spectrum in the corresponding energy range @xmath118 are due to nucleon decays . the mean lifetime limit was estimated using the formula @xmath119 where @xmath120 is the detection efficiency in the @xmath118 energy window calculated in the full simulation of the relevant process , taking into account the radial cut efficiency @xmath121 , and probability of identification by muon veto @xmath122 ; @xmath123 the number of parent nuclei ; @xmath124 the number of objects ( @xmath28 , @xmath23 or @xmath2 pairs ) inside the parent nucleus whose decay will give the specific daughter nucleus ; @xmath125 the time of measurements ; @xmath126 the number of events ( in the @xmath118 window ) due to a particular effect which can be excluded with a given confidence level on the basis of experimental data ; and @xmath127 the corresponding number of decays in the liquid scintillator / water . the parameters in ( 1 ) were defined in the following way . the mass of scintillator was defined measuring the buoyancy with a precision of 5@xmath128 . the total time of the data taking is 29.1 days and takes into account the dead time of the electronics of @xmath129 . the measurements were performed in runs of about 24 hours . the internal source of @xmath90k was used to check the energy calibration stability . the position of the @xmath90k peak was defined for every run with statistical accuracy of @xmath130 kev . no systematical shift in energy scale with a time has been found . the same is valid for the energy resolution , which is directly connected to the energy scale . in addition , all the response functions for the searched decays are wider than detector s resolution , hence the result is insensitive to the observed energy scale uncertainties . the probability of detecting scintillation events by the muon veto @xmath131 has been defined from the experimental data for the energy @xmath132 mev . this value is in agreement with the mc simulation . for @xmath133 mev@xmath134 the changes in the integral efficiency @xmath135 is negligible in the case of @xmath28 and @xmath57 decays , for the @xmath23 and @xmath66 decays the @xmath136 will decrease by 7@xmath128 . for the @xmath57 decay , there are four neutrons on the outermost @xmath137 level of @xmath3c which gives the number of @xmath57 pairs @xmath138 . the disappearance of the @xmath57 pair from this level will result in a @xmath64c nucleus in the ground state . taking into account the statistical uncertainty in the number of experimental events in the 2.03.0 mev energy window ( 276@xmath9917 ) , we calculate @xmath139 at 90% c.l . accounting for the associated efficiency @xmath140 , the limiting value for @xmath64c decays is @xmath141 . with the values of @xmath142 for @xmath3c nuclei and @xmath143 d , we obtain @xmath144c@xmath145 y with 90% c.l . for the @xmath5o nucleus we will account for only one @xmath57 pair on the outermost @xmath146 orbit ( @xmath57 decay in deeper levels will result the @xmath5o nucleus being too excited ) . with the number of @xmath5o nuclei ( in a 1 m thick water layer ) @xmath147 , @xmath148 and @xmath149 for the energy region 2.22.6 mev , the result is @xmath150o@xmath151 y with 90% c.l . with the assumption that the mean lifetime of the @xmath57 pair is the same in @xmath3c and @xmath5o nuclei , and that @xmath57 decays in both of them contribute to the experimental spectrum simultaneously , one can obtain a slightly more stringent limit for @xmath57 decay : @xmath152 y with 90% c.l . for @xmath28 decay in @xmath3c we used a similar approach , just demanding that the simulated response function for @xmath44c decay should be equal to the experimental spectrum in the energy region of 1.01.1 mev ( fig . 2 ) . in this way the value @xmath153 was determined for the full number of @xmath44c decay events . together with @xmath154 ( four neutrons on the @xmath137 level ) , it gives the following limit : @xmath155c@xmath156 y with 90% c.l . as for the @xmath23 and @xmath66 decays into invisible channels , the @xmath23 disappearance in @xmath4c will result in @xmath3b nuclei , the @xmath157 decaying with high energy release @xmath158=13.370 mev . the @xmath66 decays in @xmath4c will produce @xmath44be nuclei , also the @xmath157 decaying with @xmath158=11.508 mev ( with probability of decay to the ground state of 57% ) . to estimate the @xmath159 for @xmath23 and @xmath66 instabilities , we use the fact that no candidate scintillation events were observed in the ctf spectrum with energies higher than 4.5 mev ( fig . high energy release in the liquid scintillator can activate the muon veto of the ctf , resulting in rejection of the event . we take into account such a suppression of high energy tails in @xmath100 decays of @xmath3b and @xmath44be using the probability @xmath98 for identification of an event with energy @xmath92 in the ls by the muon veto ( fig . the beta spectra of @xmath3b and @xmath44be without and with suppression by the muon veto are also shown in fig . the part of the @xmath3b beta spectrum with @xmath160 mev reduced by a factor of @xmath135 gives an integrated efficiency @xmath161 . with zero observed events ( and with the assumption of zero expected background ) , the limiting value for the number of events is @xmath162 with 90% c.l . in accordance with the feldman - cousins procedure @xcite recommended by the particle data group @xcite . thus we arrive at the value @xmath163 for @xmath3b decay . together with @xmath164 for parent @xmath4c nuclei and the number of @xmath154 for protons in @xmath137 orbit , we obtain @xmath165c@xmath166 y with 90% c.l . in a similar way , for the @xmath66 decay in @xmath4c with the value @xmath138 for the number of @xmath66 pairs and @xmath167 , the mean lifetime limit is @xmath168c@xmath169 y with 90% c.l . all mean lifetime limits obtained here together with the numbers of parent nuclei , @xmath124 and the numbers of decay events are summarized in table 3 . & @xmath123 & @xmath124 & @xmath170 & @xmath171 , y + & & & & & + @xmath23 & @xmath172c@xmath24@xmath173b & 2.1@xmath2510@xmath53 & 4 & 6.2 & @xmath174 + @xmath28 & @xmath175c@xmath24@xmath176c & 1.9@xmath2510@xmath177 & 4 & @xmath178 & @xmath179 + @xmath57 & @xmath175c@xmath24@xmath180c & 1.9@xmath2510@xmath177 & 2 & @xmath181 & @xmath182 + & @xmath183o@xmath24@xmath184o & 9.8@xmath2510@xmath177@xmath185 & 1 & @xmath186 & @xmath187 + @xmath66 & @xmath172c@xmath24@xmath188be & 2.1@xmath2510@xmath53 & 2 & 6.7 & @xmath189 + & & & & & + + using the unique features of the borexino counting test facility the extremely low background , the large scintillator mass of 4.2 ton and the low energy threshold new limits on @xmath1 and @xmath2 decays into invisible channels ( disappearance , decays to neutrinos , majorons , etc . ) have been set : comparing these values with the data of table 1 , one can see that the established limits for @xmath57 and @xmath66 decays are the best up - to - date limits set by any method , including radiochemical and geochemical experiments . these limits are obtained in a very conservative assumption that all events in the corrresponding energy region are due to the nucleon decays . the data from the full scale borexino detector will improve the presented limits by at least two orders of magnitude . g. t hooft , phys . ( 1976 ) 8 . p. langacker , phys . 71 ( 1981 ) 185 . carlson , c.d . carone , phys . b 512 ( 2001 ) 121 . r.n . mohapatra , a. perez - lorenzana , c.a . de s. pires , phys . b 491 ( 2000 ) 143 . f.j . yndurain , phys . lett . b 256 ( 1991 ) 15 ; + g. dvali , g. gabadadze , g. senjanovic , hep - ph/9910207 ; + n. arkani - hamed , s. dimopoulos , g. dvali , phys . today 55 , february ( 2002 ) 35 . s.l . dubovsky , v.a . rubakov , p.g . tinyakov , phys . d 62 ( 2000 ) 105011 ; jhep 08 ( 2000 ) 041 ; + v.a . rubakov , phys . uspekhi 44 ( 2001 ) 871 . dubovsky , jhep 01 ( 2002 ) 012 . perkins , ann . nucl . part . 34 ( 1984 ) 1 ; + r. barloutaud , nucl . phys . b ( proc . suppl . ) 28a ( 1992 ) 437 . d.e . groom et al . ( particle data group ) , europ . j. c 15 ( 2000 ) 1 . g.n . flerov et al . , sov . dokl . 3 ( 1958 ) 79 . d. thesis , case western reserve university , 1970 . tretyak , yu.g . zdesenko , phys . lett . b 505 ( 2001 ) 59 . evans jr . , r.i . steinberg , science 197 ( 1977 ) 989 . fireman , proc . int . conf . on neutrino phys . and neutrino astrophys . neutrino77 , baksan valley , ussr , 18 - 24 june 1977 ( m. , nauka , 1978 ) , v.1 , p.53 ; + r.i . steinberg , j.c . evans , proc . int . conf . on neutrino phys . and neutrino astrophys . neutrino77 , baksan valley , ussr , 18 - 24 june 1977 ( m. , nauka , 1978 ) , v.2 , p.321 . r. bernabei et al . b 493 ( 2000 ) 12 . q.r . ahmad et al . ( sno collaboration ) , phys . 89 ( 2002 ) 011301 . yu.g . zdesenko , v.i . tretyak , phys . b 553 ( 2003 ) 135 . y. suzuki et al . , phys . b 311 ( 1993 ) 357 . j. learned , f. reines , a. soni , phys . 43 ( 1979 ) 907 . c. berger et al . b 269 ( 1991 ) 227 . j.f . glicenstein , phys . b 411 ( 1997 ) 326 . p. belli et al . , to be published . borexino collaboration , g. alimonti et al . , astropart . ( 2002 ) 205 . derbin , o.ju . smirnov , v.i . tretyak for the borexino collaboration , proc . xx - th int . conf . on neutrino physics and astrophysics ( may 25 - 30 , 2002 , munich , germany ) to be published . borexino collaboration , g. alimonti et al . meth . a 406 ( 1998 ) 411 ; astropart . ( 1998 ) 141 ; phys . b 422 ( 1998 ) 349 ; borexino collaboration , g. alimonti et al . , nucl . instrum . meth . a 440 ( 2000 ) 360 . birks , proc . a 64 ( 1951 ) 874 los arcos , f. ortiz , comp . 103 ( 1997 ) 83 . smirnov , instr . and technique 46 ( 2003 ) in press . borexino collaboration , h.o . back et al . , phys . b 525 ( 2002 ) 29 . m. johnson et al . , nucl . instrum . meth . a 414 ( 1998 ) 459 . rosman , p.d.p . taylor , pure appl . ( 1998 ) 217 . , ed . by c.m . lederer and v.s . shirley , 7th ed . , john wiley , n.y . , g. audi , a.h . wapstra , nucl . phys . a 595 ( 1995 ) 409 . w. fritsch , r. lipperheide , u. wille , nucl . phys . a 241 ( 1975 ) 79 . d. vautherin , d.m . brink , phys . c 5 ( 1972 ) 626 . nelson , h. hirayama , d.w.o . rogers , slac - report-265 , stanford , 1985 . feldman , r.d . cousins , phys . d 57 ( 1998 ) 3873 .

the results of background measurements with the second version of the borexino counting test facility ( ctf - ii ) , installed in the gran sasso underground laboratory , were used to obtain limits on the instability of nucleons , bounded in nuclei , for decays into invisible channels ( @xmath0 ) : disappearance , decays to neutrinos , etc . the approach consisted of a search for decays of unstable nuclides resulting from @xmath1 and @xmath2 decays of parent @xmath3c , @xmath4c and @xmath5o nuclei in the liquid scintillator and the water shield of the ctf . due to the extremely low background and the large mass ( 4.2 ton ) of the ctf detector , the most stringent ( or competitive ) up - to - date experimental bounds have been established : @xmath6 y , @xmath7 y , @xmath8 y and @xmath9 y , all at 90% c.l . , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , proton decay , baryon number conservation , scintillation detector 11.30 , 11.30.f , 12.60 , 29.40.m

one of the main problems in neuroscience is determining how information is represented in the brain . there have been many experimental and theoretical studies in this direction in the last few years . for example , aertsen and gerstein conducted a series of experiments , and proposed the concept of `` effective connectivity '' @xcite . tsukada et al . showed experimentally that the structure of the inter - spike - interval has a significant effect on the amplitude of long - term potentiation ( ltp ) in the hippocampus of guinea - pigs , @xcite and fujii et al . , based on a large number of experimental data , theoretically proposed the hypothesis of dynamical cell assembly in the central nervous system @xcite . however , how information is represented in the brain has not yet been elucidated because the brain is a very complicated and hierarchically organized system . it is necessary to acquire more data about information representation at the neuron level as well as at the network level . it should also be emphasized that determination of how memory information is represented in the brain is essential for understanding of higher brain functions , because all higher brain functions are based on coding of memory information . furthermore , not only experimental studies but also theoretical studies , involving use of simple biological neuron models or neural networks , are necessary for elucidation of the mechanism of memory information representation and its algorithm . the purpose of this work was to investigate the effect of the time structure of spike trains on the ltp amplitude in the hippocampus using a simple biological neuron model , and explain the basic synaptic mechanisms underlying memory information representation . for this purpose , we constructed a single neuron model considering both n - methyl - d - aspartate ( nmda ) and non - nmda receptors . then we applied three different time - structured stimuli ( regular , markov , and chaotic ) to our neuron model , and compared the ltp amplitude among the three cases . a striking finding was that the chaotic stimuli in the nonstationary regime produce the largest ltp . this suggests the significance of nonstationary chaotic information representation in memory information processing . neurons in area ca1 of the hippocampus are typical examples of neurons which exhibit ltp @xcite . since their physiological properties have been extensively studied experimentally , we can construct a single neuron model of this area . here , we adopt a discrete time model with a continuous membrane potential , a setup , however , which compromises the biological plausibility and the simulation time cost . we take into account the channel properties in the synaptic region . the basic time evolution of the activity of a single neuron is given by @xmath0 where @xmath1 is the output of the neuron at time @xmath2 , @xmath3 is the membrane potential at time @xmath4 , and @xmath5 is a probability function which describes the relation between the membrane potential and the output . here , we adopt the following form for @xmath5 : @xmath6 where @xmath7 is the threshold membrane potential ( @xmath8 ) and @xmath9 at @xmath10 . the membrane potential @xmath3 can be decomposed into three parts , i.e. , @xmath11 where @xmath12 is the post - synaptic potential induced by the input current at time @xmath4 , @xmath13 is the constant resting membrane potential , and @xmath14 is a small fluctuation of the membrane potential at time @xmath4 where the average value of @xmath14 is zero . the synaptic input through non - nmda receptors @xmath12 can be represented as follows : @xmath12 depends on both the synaptic efficacy @xmath15 and the time course of the input stimulus @xmath16 , which can be written as @xmath17 where @xmath18 is the synaptic efficacy at time @xmath19 , @xmath20 is the spike train of the input stimuli which represents the existence of the input spike at time @xmath19 ( @xmath21 with input , @xmath22 without input ) , and @xmath23 is the decay constant for non - nmda receptors . the time course of @xmath20 determines the statistical features of the input signals , such as the average frequency or the average interval of the input stimuli . now , let us consider ltp of this model in terms of the change of @xmath15 . physiologically , both depolarization of the membrane and calcium entry into the post - synaptic neuron are necessary for ltp to occur @xcite . therefore , we introduce the variable @xmath24 , which represents the concentration of calcium ions in the post - synaptic neuron and is expressed as : @xmath25 where @xmath26 is a constant , @xmath27 is a step function , @xmath28 is the membrane potential required for removal of magnesium ions from nmda receptors ( @xmath29 ) , and @xmath30 is the decay constant for nmda receptors . although it is not well known how the calcium - dependent second messenger system produces ltp , several experimental observations have revealed strong correlation between the ltp amplitude and the frequency of the input stimuli , that is , it is physiologically plausible that ( 1 ) regular stimuli at 3 - 10 hz can induce ltp , @xcite but ( 2 ) low - frequency stimuli ( @xmath31 1hz ) induce long - term depression ( ltd ) @xcite . we also need to consider that the value of @xmath15 saturates in the regime of large @xmath24 and @xmath15 due to the limited number of neurotransmitters and receptors . therefore , we propose the following model for the change of the synaptic efficacy @xmath32 : @xmath33 where @xmath34 , and @xmath35 are all parameters . a schematic of @xmath32 is given in fig . 1 , where the values of @xmath36 are fixed in our simulations . = 15.9 cm = 11.0 cm for @xmath37 and @xmath38 , @xmath32 is positive corresponding to ltp described by the experimental data @xcite . for @xmath39 and @xmath38 , @xmath32 is zero due to the saturation effect . for @xmath40 and @xmath41 , @xmath32 is zero because of less @xmath42 entry . for @xmath43 and @xmath44 , @xmath32 is negative corresponding to ltd at low frequency @xcite . when we consider @xmath45 neurons which are stimulated independently , the averaged output @xmath46 at time @xmath4 is defined by @xmath47 which can be written as @xmath48 in a large @xmath45 limit according to the central limit theorem . therefore , in this letter we consider @xmath49 to represent the characteristic measure of the ltp amplitude . in figs . 2 - 4 , the amplitude variable of the vertical axis shows @xmath50 for comparison with the physiological data in ref . 2 ; the notation @xmath51 represents the time average for a sample process and @xmath52 represents the ensemble average for a large number of sample processes . in the present letter , the number of ensembles is fixed at @xmath53 to @xmath54 . the values of the parameters used in this letter are chosen to be biologically plausible , i.e. , @xmath55 mv , @xmath56 mv , @xmath14 is generated from the uniform distribution of @xmath57 , @xmath58 mv , @xmath59 , @xmath60 mv , @xmath61 ms , @xmath62 ms , @xmath63 , and @xmath64 . to determine the effect of the spike structure on ltp , we consider three different spike stimuli : ( i ) regular , ( ii ) markov , and ( iii ) chaotic . for the markov stimuli , we use the same transition matrix as was used by tsukada et al . @xcite to compare our results with theirs ; it is given as @xmath65 = \left [ \begin{array}{ccc } ss & sm & sl \\ ms & mm & ml \\ ls & lm & ll \end{array } \right],\ ] ] where @xmath66 , @xmath67 , and @xmath68 stand for the markovian states of the signals with a 100 ms spike interval , 500 ms spike interval , and 900 ms spike interval , respectively . for example , the transition probability @xmath69 indicates that the @xmath66 ( 100 ms ) interval directly follows the @xmath67 ( 500 ms ) interval . the transition matrices used in the simulations are given as @xmath70 = \left [ \begin{array}{ccc } 0.70 & 0.20 & 0.10 \\ 0.45 & 0.10 & 0.45 \\ 0.10 & 0.20 & 0.70 \end{array } \right]\ ] ] @xmath71 = \left [ \begin{array}{ccc } 0.41 & 0.18 & 0.41 \\ 0.41 & 0.18 & 0.41 \\ 0.41 & 0.18 & 0.41 \end{array } \right]\ ] ] @xmath72 = \left [ \begin{array}{ccc } 0.10 & 0.20 & 0.70 \\ 0.45 & 0.10 & 0.45 \\ 0.70 & 0.20 & 0.10 \end{array } \right],\ ] ] where @xmath73 , @xmath74 , and @xmath75 correspond to the positive correlation , the independent correlation , and the negative correlation , respectively @xcite . the mean frequency of the input stimuli @xmath16 for all three cases is fixed at 2 hz . for the case of chaotic stimuli , we use the modified bernoulli map @xmath76 ( mod.1 ) to produce the input stimuli @xmath16 , @xmath77 where @xmath78 and @xmath79 are parameters . then , in order to fix the mean spike interval at 500 ms ( i.e. , the mean frequency is 2 hz ) , we transform the value of @xmath80 to the spike interval @xmath81 of the time course @xmath16 in the following way : @xmath82 the properties of this map were thoroughly studied by aizawa , @xcite and it should be noted that the map is stationary chaotic for @xmath83 , and nonstationary chaotic for @xmath84 . the time course of the input stimuli @xmath16 generated by eq . ( 3.6 ) reveals a typical intermittency for all regimes , and the @xmath85 spectrum is emphasized in the case for @xmath84 . let us now numerically investigate ltp induced by the time - structured stimuli mentioned above . the setup of the numerical ltp experiments is as follows . the unit of the discrete time is taken as 1 ms , and 20 impulses with 0.05 hz are applied as an initial control stimulus . then , we ensure that the amplitude @xmath15 reveals a controlled reference state corresponding to the threshold @xmath7 . after these preparations , 200 tetanic impulses are applied , and then 20 control impulses with 0.05 hz intervals ( called the latter control ) are applied . the ltp amplitude is measured in terms of the average value , @xmath86 , where @xmath4 stands for the duration of the latter control . these procedures for our simulations are the same as in the physiological experiments @xcite . the first case is that of the regular periodic stimuli ranging from 10 hz to 0.5 hz , and the results are shown in fig . we use 1,000 different sample processes of stimuli for each frequency , and the ensemble averages of ltp , @xmath50 , are shown by the filledcircles and the standard deviations by the errorbars . figure 2 clearly shows that a higher - frequency stimulus produces a larger ltp than lower - frequency ones , and that no ltp occurs at frequencies of less than 2 hz . these results are consistent with the results of experiments performed by tsukada et al . @xcite = 15.9 cm = 10.0 cm next , we apply the markov stimuli produced by the transition matrices in eqs . ( 3.1 ) , ( 3.2 ) , and ( 3.3 ) . for each type of stimulus , we consider 100,000 sample processes , and calculate the average and standard deviation of the ltp amplitude . the results are shown in fig . it is clear that the positive correlated stimuli produce the largest ltp , the independent ones the next largest ones , and the negative correlated ones the least ones . comparing fig . 3 and fig . 2 , we can conclude that the time - structured markov stimuli with 2 hz frequency produce a larger ltp than do the regular stimuli with the same frequency . this implies that the time structure of the input signals is important for ltp to occur . although our result for the independent correlation case does not coincide with the experimental result obtained by tsukada et al . , @xcite the positive and the negative cases fit well with their experimental results . furthermore , our results shown in figs . 2 and 3 are consistent with the results of previous work performed by sugiura et al . @xcite with a different neuron model . = 15.9 cm = 10.0 cm finally , we apply the chaotic intermittent stimuli expressed by eq . ( 3.6 ) to our neuron model . for each value of @xmath87 , we produce 10,000 sample processes . the results are shown in fig . 4 . as the value of @xmath87 increases , the average ltp amplitude increases almost monotonically . thus we can conclude that on average , the chaotic stimuli in the nonstationary regime produce larger ltp than do the stationary ones . comparing fig . 4 with figs . 2 and 3 , we can clearly see that intermittent chaotic stimuli have the strongest ltp effect among all the stimuli , though the fluctuations represented by the standard deviations are also much larger in the nonstationary regime than in the stationary regime . = 15.9 cm = 10.0 cm in this letter , we proposed a simple neuron model including the effects of nmda and non - nmda receptors , and discuss the relation between the time structure of the input stimuli and the amplitude of ltp . it is shown by numerical simulation that the results obtained using our neuron model fit the experimental data reasonably well , and indicate that the chaotic stimuli in the nonstationary regime will produce the largest ltp . the reason that the time structure of the input stimuli has a significant effect on the amplitude of ltp can be understood by the fact that the calcium concentration through nmda receptors has a long - lasting memory effect ( @xmath88 ms ) , as is shown in eq . thus , if the interval between two spikes is short , the effect of the first spike remains when the second one arrives , resulting in a higher calcium concentration in a post - synaptic neuron . on the other hand , when the interval between the two spikes is long , the effect of the first spike ceases before the second one arrives . therefore , the inter - spike - interval must be short for ltp to occur in our model . for the markov stimuli , the positive correlation case has the longest short spike interval , the independent one the next longest , and the negative one the shortest , and this explains the differences in the average ltp amplitude in fig . 3 . in the case for the chaotic stimuli , @xmath80 of the modified bernoulli map tend to localize around @xmath89 or @xmath90 in the large @xmath87 regime , and as the value of @xmath87 increases , the spikes tend to cluster in a short interval or a long interval for a longer duration . therefore , if the spikes occur within a short interval for a long duration , they will produce very large ltp . on the other hand , almost no ltp will occur when spikes occur within a long interval for a long duration . this results in the large deviation in the nonstationary regime in fig . 4 . since the short spike interval becomes longer as the value of @xmath87 increases , the average ltp amplitude also increases . moreover , the average ltp amplitude in the nonstationary chaotic regime is the largest among those for all the stimuli , because the short spike interval in the nonstationary chaotic regime is longer than that in any others . to determine how memory information is represented in the brain , it is necessary to elucidate the relationship between the time structure of the impulse train and its validity as a form of information representation . for this purpose , we obtained the basic data of the ltp formation at the single neuron level . our results strongly suggest that the time structure of the impulse train is important for memory information representation at the single neuron level . in particular , the result that the nonstationary chaotic spike trains produce the largest ltp implies that the brain may use chaotic intermittency for memory information representation even at the single neuron level . furthermore , the wide range of the ltp amplitude in the nonstationary regime , which is indicated by the large standard deviation , seems to play a very important role in the selection of the detailed information embedded in the diversity of the time - structured spike trains . however , in order to acquire the entire pictures of memory information representation and its algorithm in the brain , it is also necessary to determine the relationship between the time structure of the impulse train and its validity as a form of information representation at the network level . also , in particular , detailed study about the relationship between the time structure of the impulse train and the properties of the network connections , such as symmetric , asymmetric , excitatory , and inhibitory ones , is necessary . to this end , the neuron model proposed in this letter is suitable because it is simple enough for network - level simulation and has biological properties , though some details such as the form of eq . ( 2.6 ) can be further simplified . further research for analysis of the network structure and increasing the level of the sophistication of our neuron model is now in progress . 99 a. m. h. j. aertsen and g. l. gerstein : _ neuronal cooperativity _ , j. krger ( springer - verlag , berlin , 1991 ) p. 52 . m. tsukada , t. aihara , m. mizuno , h. kato and k. ito : biol . cybern . * 70 * ( 1994 ) 495 . h. fujii , h. ito , k. aihara , n. ichinose and m. tsukada : neural networks * 9 * ( 1996 ) 1303 . p. s. churchland and t. j. sejnowski : _ the computational brain _ ( mit press , cambridge , 1992 ) . e. r. kandel , j. h. schwartz and t. m. jessell : _ principle of neuroscience _ ( appleton & lange , norwalk , 1991 ) . s. m. dudek and m. f. bear : proc . usa * 89 * ( 1992 ) 4363 . y. aizawa : prog . * 99 * ( 1989 ) 2866 . y. sugiura , t. aihara and m. tsukada : technical report of ieice * nc93 - 104 * ( 1994 ) 181 [ in japanese ] .

we theoretically investigate long - term potentiation ( ltp ) in the hippocampus using a simple model of a neuron stimulated by three different time - structured input signals ( regular , markov , and chaotic ) . the synaptic efficacy change is described taking into account both n - methyl - d - aspartate ( nmda ) and non - nmda receptors . the experimental results are successfully explained by our neuron model , and the remarkable fact that the chaotic stimuli in the nonstationary regime produce the largest ltp is discussed .

there have been many studies of dynamical predator - prey systems that simulate biological processes . particularly interesting early work was done by bell @xcite , who showed that the immune response can be modeled quite effectively by such systems . in bell s work the time evolution of competing concentrations of one antigen and one antibody is studied . the current paper shows what happens if we combine two antibody - antigen subsystems _ in a @xmath0-symmetric fashion _ to make an immune system in which there are _ two _ antibodies and _ two antigens_. an unexpected conclusion is that even if one antigen is lethal ( because the antigen concentration grows out of bounds ) , the introduction of a _ second _ antigen can stabilize the concentrations of _ both _ antigens , and thus save the life of the host . introducing a second antigen may actually drive the concentration of the lethal antigen to zero . we say that a classical dynamical system is @xmath0 _ symmetric _ if the equations describing the system remain invariant under combined space reflection @xmath1 and time reversal @xmath2 @xcite . classical @xmath0-symmetric systems have a typical generic structure ; they consist of two coupled identical subsystems , one having gain and the other having loss . such systems are @xmath0 symmetric because under space reflection the systems with loss and with gain are interchanged while under time reversal loss and gain are again interchanged . systems having @xmath0 symmetry typically exhibit two different characteristic behaviors . if the two subsystems are coupled sufficiently strongly , then the gain in one subsystem can be balanced by the loss in the other and thus the total system can be in equilibrium . in this case the system is said to be in an _ unbroken _ @xmath0-symmetric phase . ( one visible indication that a linear system is in an unbroken phase is that it exhibits rabi oscillations in which energy oscillates between the two subsystems . ) however , if the subsystems are weakly coupled , the amplitude in the subsystem with gain grows while the amplitude in the subsystem with loss decays . such a system is not in equilibrium and is said to be in a _ broken _ @xmath0-symmetric phase . interestingly , if the subsystems are very strongly coupled , it may also be in a broken @xmath0-symmetric phase because one subsystem tends to drag the other subsystem . a simple linear @xmath0-symmetric system that exhibits a @xmath0 phase transition from weak to moderate coupling and a second transition from moderate to strong coupling consists of a pair of coupled oscillators , one with damping and the other with antidamping . such a system is described by the pair of linear differential equations @xmath3 this system is invariant under combined parity reflection @xmath1 , which interchanges @xmath4 and @xmath5 , and time reversal @xmath2 , which replaces @xmath6 with @xmath7 . theoretical and experimental studies of such a system may be found in refs . @xcite . for an investigation of a @xmath0-symmetric system of many coupled oscillators see ref . experimental studies of @xmath0-symmetric systems may be found in refs . @xcite . it is equally easy to find physical nonlinear @xmath0-symmetric physical systems . for example , consider a solution containing the oxidizing reagent potassium permanganate @xmath8 and a reducing agent such as oxalic acid @xmath9 . the reaction of these reagents is self - catalyzing because the presence of manganous @xmath10 ions increases the speed of the reaction . the chemical reaction in the presence of oxalic acid is @xmath11 thus , if @xmath12 is the concentration of permanganate ions and @xmath13 is the concentration of manganous ions , then the rate equation is @xmath14 where @xmath15 is the rate constant . this system is @xmath0 invariant , where @xmath1 exchanges @xmath4 and @xmath5 and @xmath2 replaces @xmath6 with @xmath7 . for this system , the @xmath0 symmetry is always broken ; the system is not in equilibrium . the volterra ( predator - prey ) equations are a slightly more complicated @xmath0-symmetric nonlinear system : @xmath16 this system is oscillatory and thus we say that the @xmath0 symmetry is unbroken . these equations are discussed in ref . . a nonlinear @xmath0-symmetric system of equations that exhibits a phase transition between broken and unbroken regions may be found in ref . @xcite . in analyzing elementary systems like that in ( [ e1 ] ) , which are described by constant - coefficient differential equations , the usual procedure is to make the _ ansatz _ @xmath17 and @xmath18 . this reduces the system of differential equations to a polynomial equation for the frequency @xmath19 . we then associate unbroken ( or broken ) phases with real ( or complex ) frequencies @xmath19 . if @xmath19 is real , the solutions to both equations are oscillatory and remain bounded , and this indicates that the physical system is in dynamic equilibrium . however , if @xmath19 is complex , the solutions grow or decay exponentially with @xmath6 , which indicates that the system is not in equilibrium . for more complicated nonlinear @xmath0-symmetric dynamical systems , we still say that the system is in a phase of broken @xmath0 symmetry if the solutions grow or decay with time or approach a limit as @xmath20 because the system is not in dynamic equilibrium . in contrast , if the variables oscillate and remain bounded as @xmath6 increases we say that the system is in a phase of unbroken @xmath0 symmetry . however , in this case the time dependence of the variables is unlikely to be periodic ; such systems usually exhibit _ almost periodic _ or _ chaotic _ behavior . to illustrate these possibilities we construct a more elaborate @xmath0-symmetric system of nonlinear equations by combining a two - dimensional dynamical subsystem whose trajectories are _ inspirals _ with another two - dimensional dynamical subsystem whose trajectories are _ outspirals_. for example , consider the subsystem @xmath21 this system has two saddle points and one stable spiral point , as shown in fig . [ f1 ] ( left panel ) . plane for the dynamical subsystem ( [ e4 ] ) with @xmath22 . the initial conditions are @xmath23 . right panel : an outspiral trajectory for ( [ e5 ] ) in the @xmath24 plane with @xmath22 . the initial conditions are @xmath25 . in the left panel @xmath6 ranges from @xmath26 to @xmath27 and in the right panel @xmath6 ranges from @xmath26 to @xmath28 . ] next , we consider the @xmath0 reflection ( @xmath29 , @xmath30 , @xmath31 ) of the subsystem in ( [ e4 ] ) : @xmath32 the trajectories of this system are outspirals , as shown in fig . [ f1 ] ( right panel ) . the time evolution of the four dynamical variables in fig . [ f1 ] , @xmath33 and @xmath34 , @xmath35 and @xmath36 , is shown in fig . [ f2 ] . . the four variables @xmath33 , @xmath34 , @xmath35 , and @xmath36 are plotted as functions of @xmath6 . ] let us now couple the two subsystems in ( [ e4 ] ) and ( [ e5 ] ) in such a way that the @xmath0 symmetry is preserved . the resulting dynamical system obeys the nonlinear equations @xmath37 in which @xmath38 and @xmath39 are the coupling parameters . this system has a wide range of possible behaviors . for example , for the parametric values @xmath40 , @xmath41 , and @xmath42 and the initial conditions @xmath43 we can see from figs . [ f3 ] and [ f4 ] that the system is in a broken-@xmath0-symmetric phase . -symmetric system ( [ e6 ] ) in a broken-@xmath0-symmetric phase , as indicated by the outspiral behavior in the @xmath44 $ ] and @xmath45 $ ] planes . the parametric values are @xmath40 , @xmath41 , and @xmath42 and the initial conditions are @xmath43 . in these plots @xmath6 ranges from @xmath26 to @xmath46 . ] , @xmath34 , @xmath35 , @xmath36 for the parametric values and initial conditions shown in fig . [ f3 ] . ] when the coupling parameters are chosen so that the system ( [ e6 ] ) is in a phase of unbroken @xmath0 symmetry , the initial conditions determine whether the behavior is chaotic or almost periodic . for example , for the same parametric values @xmath40 , @xmath41 , and @xmath47 the system in ( [ e6 ] ) is in an unbroken-@xmath0-symmetric phase . two qualitatively different behaviors of unbroken @xmath0 symmetry are illustrated in figs . [ f5 ] , [ f6 ] , [ f78 ] and [ f9 ] , [ f10 ] , and [ f1112 ] . the first three figures display the system in two states of chaotic equilibrium and the next three show the system in two states of almost - periodic equilibrium . the poincar plots in figs . [ f5 ] and [ f6 ] ( left panels ) and figs . [ f9 ] and [ f10 ] ( left panels ) distinguish between chaotic and almost periodic behavior . ) in a phase of chaotic unbroken @xmath0 symmetry . the parametric values are @xmath40 , @xmath48 , @xmath47 and the initial conditions are @xmath49 . left panel : poincar plot of @xmath50 versus @xmath51 when @xmath52 . the two - dimensional scatter of dots indicates that the system is chaotic . in this plot @xmath6 ranges from @xmath26 to @xmath53 . right panel : a plot of @xmath33 versus @xmath35 for @xmath6 ranging from @xmath26 to @xmath54 . ] symmetry . the parametric values and the ranges of @xmath6 are the same as in fig . [ f5 ] , but the initial conditions are now @xmath55 . ] plotted as a function of time . the chaotic behavior can be seen as the uneven oscillations . these oscillations are reminiscent of a trajectory under the influence of a pair of strange attractors.,title="fig : " ] + plotted as a function of time . the chaotic behavior can be seen as the uneven oscillations . these oscillations are reminiscent of a trajectory under the influence of a pair of strange attractors.,title="fig : " ] ) in a state of unbroken @xmath0 symmetry . the parametric values and the time ranges are the same as in fig . [ f5 ] , but the initial conditions are @xmath56 . the presence of one - dimensional _ islands _ in the poincar plot ( left panel ) shows that the time evolution of the system is almost periodic . ] ) in a different state of unbroken @xmath0 symmetry . the parametric values and the ranges of @xmath6 are the same as in fig . [ f9 ] , but the initial conditions are @xmath57 . the poincar plot ( left panel ) again shows that the time evolution of the system is almost periodic . ] plotted as a function of time . the almost periodic behavior is particularly evident in the graphs on the left , where the oscillations are quite regular.,title="fig : " ] + plotted as a function of time . the almost periodic behavior is particularly evident in the graphs on the left , where the oscillations are quite regular.,title="fig : " ] the choice of coupling parameters usually ( but not always ) determines whether the system is in an unbroken or a broken @xmath0-symmetric phase . to demonstrate this , we take @xmath40 and examine the time evolution for roughly 11,000 values of the parameters @xmath38 and @xmath39 . figure [ f13 ] indicates the values of @xmath38 and @xmath39 for which the system is in a broken or an unbroken ( chaotic or almost periodic ) phase . plane for the system ( [ e6 ] ) with the parametric value @xmath40 . the initial conditions are @xmath58 . the dots correspond to parametric values @xmath59 in the region of broken @xmath0 symmetry , and the white space corresponds to the region of unbroken @xmath0 symmetry . the edges of the regions are not completely sharp ; it can be difficult to determine the precise location of the boundary curves separating broken and unbroken regions because this requires integrating for extremely long times . ] having summarized the possible behaviors of coupled @xmath0-symmetric dynamical subsystems , in sec . [ s2 ] we construct and examine in detail a @xmath0-symmetric dynamical model of an antigen - antibody system containing _ two _ antigens and _ two _ antibodies . this system is similar in structure to that in ( [ e6 ] ) . we show that in the unbroken region the concentrations of antigens and antibodies generally become chaotic and we interpret this as a chronic infection . however , in the unbroken regions there are two possibilities ; either the antigen concentration grows out of bounds ( the host dies ) or else the antigen concentration falls to zero ( the disease is completely cured ) . some concluding remarks are given in sec . infecting an animal with bacteria , foreign cells , or virus may produce an immune response . the foreign material provoking the response is called an _ antigen _ and the immune response is characterized by the production of _ antibodies _ , which are molecules that bind specifically to the antigen and cause its destruction . the time - dependent immune response to a replicating antigen may be treated as a dynamical system with interacting populations of the antigen , the antibodies , and the cells that are involved in the production of antibodies . a detailed description of such an immune response would be extremely complicated so in this paper we consider a simplified mathematical model of the immune response proposed by bell @xcite . bell s paper introduces a simple model in which the multiplication of antigen and antibodies is assumed to be governed by lokta - volterra - type equations , where the antigen plays the role of prey and the antibody plays the role of predator . while such a model may be an unrealistic simulation of an actual immune response , bell argues that this mathematical approach gives a useful qualitative and quantitative description . following bell s paper we take the variable @xmath33 to represent the concentration of antibody and the variable @xmath34 to represent the concentration of antigen at time @xmath6 . assuming that the system has an unlimited capability of antibody production , bell s dynamical model describes the time dependence of antigen and antibody concentrations by the differential equations @xmath60 according to ( [ e7 ] ) , the antigen concentration @xmath61 increases at a constant rate @xmath62 if the antibody @xmath50 is are not present . as soon as antigens are bound to antibodies , the antibodies start being eliminated at the constant rate @xmath63 . analogously , the concentration of antibody @xmath50 decays with constant rate @xmath64 in the absence of antigens , while binding of antigens to antibodies stimulates the production of antibody @xmath50 with constant rate @xmath65 . the functions @xmath66 and @xmath67 denote the concentrations of bound antibodies and bound antigens . assuming that @xmath68 , an approximate expression for the concentration of bound antigens and antibodies is @xmath69 where @xmath15 is called an _ association constant_. with the scalings @xmath70 and @xmath71 and the change of variable @xmath72^{-1},\ ] ] system ( [ e7 ] ) becomes @xmath73 the system ( [ e9 ] ) exhibits four different behaviors : * if @xmath74 , there is unbounded monotonic growth of antigen . * if @xmath75 and @xmath76 , there is an outspiral ( oscillating growth of antigen ) . * if @xmath75 and @xmath77 , there is an inspiral ( the antigen approaches a limiting value in an oscillatory fashion ) . * if @xmath75 and @xmath78 , the system exhibits exactly periodic oscillations . this behavior is unusual in a nonlinear system and indeed ( [ e6 ] ) does not exhibit exact periodic behavior . subsequent to bell s paper @xcite there have been many studies that use two - dimensional dynamical models to examine the antigen - antibody interaction @xcite . however , in this paper we construct a _ four_-dimensional model consisting of two antigens and two antibodies . let us assume that an antigen @xmath61 attacks an organism and that the immune response consists of creating antibodies @xmath50 as described by ( [ e7 ] ) . however , we suppose that the organism has a second system of antibodies and antigens @xmath24 . this second subsystem plays the role of a @xmath0-symmetric partner of the system @xmath79 , where parity @xmath1 interchanges the antibody @xmath50 with the antigen @xmath80 and the antigen @xmath61 with the antibody @xmath50 , @xmath81 and time reversal @xmath2 makes the replacement @xmath82 . the time evolution of this new antibody - antigen system is regulated by the equations @xmath83 we assume that the interaction between antibody @xmath51 and antigen @xmath80 is controlled by the same constant @xmath15 as in ( [ e8 ] ) . we assume that because antibodies may have many possible binding sites , @xmath50 can also bind to antigen @xmath80 and that antibody @xmath51 can also bind to antigen @xmath61 . moreover , for this model we assume that we can scale the dynamical variables so that this interaction is the same as the interaction @xmath84 and @xmath85 . this means that after the scaling @xmath86 , @xmath71 , @xmath87 , and @xmath88 , the dynamical behavior of the total system @xmath89 is described by @xmath90 the production of the antibody @xmath51 is stimulated by the presence of the antigen @xmath80 . the terms involving the parameter @xmath38 describe the production of additional antibodies @xmath51 and additional elimination of antigens @xmath61 . similarly , @xmath39 terms describe the production of new antibodies @xmath50 and additional elimination of antigens @xmath80 . we remark that the system ( [ e11 ] ) with @xmath91 can be derived from the hamiltonian @xcite @xmath92 that is , it can be recovered from the hamilton equations @xmath93 where @xmath94 figure [ f14 ] displays a phase diagram of the @xmath0-symmetric model in ( [ e11 ] ) , where we have taken @xmath95 , @xmath96 , and @xmath97 . in this figure a portion of the @xmath59 plane is shown and the regions of broken and unbroken @xmath0 symmetry are indicated . unbroken-@xmath0-symmetric regions are indicated as hyphens ( blue online ) . there are two kinds of broken-@xmath0-symmetric regions ; x s ( red online ) indicate solutions that grow out of bounds and o s ( green online ) indicate solutions for which the concentration of antigen @xmath61 approaches 0 . portion of the @xmath59 coupling - parameter plane for the @xmath0-symmetric immune - response system ( [ e11 ] ) showing regions of broken and unbroken @xmath0 symmetry . we take as initial conditions @xmath98 and @xmath99 ; that is , we assume that the disease associated with antigen - antibody 1 is well established and that at @xmath100 a very small amount of antigen - antibody 2 is injected . points in the unbroken region are indicated as hyphens ( blue ) . in this region the concentrations @xmath101 are all oscillatory in time . in general , depending on the initial conditions , the solutions can be either almost periodic or chaotic . however , as shown in fig . [ f18 ] , the solutions to ( [ e11 ] ) are chaotic . thus , in this region the introduction of antigen - antibody 2 makes the potentially lethal infection chronic . the regions whose points are indicated as o s ( green ) and x s ( red ) have broken @xmath0 symmetry . in the x regions the solutions oscillate and grow out of bounds . in the o regions @xmath33 and @xmath34 vanish and @xmath35 and @xmath36 approach small finite values as @xmath20 . thus , in the x regions the host dies , but in the o regions the disease due to antigen @xmath61 is completely cured . ] figure [ f15 ] shows that the organism does not survive if the second antibody - antigen pair @xmath102 is not initially present . in this figure we take @xmath98 but we we take @xmath103 . ) is described by ( [ e7 ] ) because we take @xmath103 and thus @xmath35 and @xmath36 remain 0 for all @xmath6 . we have taken @xmath95 , @xmath96 , and @xmath97 . the initial conditions are @xmath98 . ] figure [ f16 ] shows what happens in a broken-@xmath0-symmetric phase when the organism does not survive . we take @xmath104 and @xmath105 , which puts us in the lower - left corner of fig . the initial conditions are @xmath106 and @xmath107 . note that the level of the @xmath34 antigen grows out of bounds . -symmetric phase in the lower - left corner of fig . [ f14 ] ; specifically @xmath104 and @xmath105 . the organism does not survive the antigen attack . the antigen - antibody dynamics is described by ( [ e11 ] ) , where @xmath62 , @xmath64 , @xmath63 , @xmath65 , and the initial conditions are the same as in fig . [ f14 ] . ] figure [ f17 ] shows what happens in the unbroken region in fig . the organism survives but the disease becomes chaotically chronic . and @xmath108 , which is in the unbroken-@xmath0 phase in the lower - left portion of fig . the antigen - antibody dynamics is described by ( [ e11 ] ) , where @xmath62 , @xmath64 , @xmath63 , @xmath65 , and the initial conditions are the same as in fig . [ f14 ] . the concentrations of antigens and antibodies behave chaotically in time . ] figure [ f18 ] demonstrates the chaotic behavior at a point in the upper - right unbroken-@xmath0 portion of fig . [ f14 ] , specifically at @xmath109 and @xmath110 . the figure shows a poincar map in the @xmath79 plane for @xmath111 . ) , where @xmath62 , @xmath64 , @xmath63 , @xmath65 , and the initial conditions are the same as in fig . [ f14 ] . in this plot @xmath109 and @xmath112 , which places the system in the unbroken phase in the upper - right corner of fig . the dynamical behavior is chaotic and the disease becomes chronic , as implied by the poincar map in which trajectory points are plotted in the @xmath79 plane for @xmath111 . the scatter of points indicates chaotic behavior . the time interval for the plot is from @xmath100 to @xmath113 . ] figure [ f19 ] shows what happens in the broken-@xmath0 region in the lower - right corner of fig . [ f14 ] at @xmath48 and @xmath114 . in this region the antigen @xmath34 completely disappears and the disease is cured . and @xmath114 in the broken-@xmath0 region in the lower - right corner of fig . [ f14 ] . in this plot @xmath62 , @xmath64 , @xmath63 , @xmath65 , and the initial conditions are the same as in fig . note that the antigen concentration @xmath34 decays to zero and the disease is cured . the concentration of antigen @xmath80 approaches a small nonzero value as @xmath20 and , as was noted in ref . @xcite , this value is so small that we regard it as negligible . ] in this paper we have extended bell s two - dimensional predator - prey model of an immune response to a four - dimensional @xmath0-symmetric model and have examined the outcomes in the broken- and the unbroken-@xmath0-symmetric phases . we have found that in the unbroken phase the disease becomes chronic ( oscillating ) while in the broken phase the host may die or be completely cured . in bell s model ( ref . @xcite ) an oscillating regime is assumed to be a transitory state and that either the antigen is completely eliminated at an antigen minimum or the host dies at an antigen maximum . however , there are many examples in which the immune system undergoes temporal oscillations ( occurring in pathogen load in populations of specific cell types , or in concentrations of signaling molecules such as cytokines ) . some well known examples are the periodic recurrence of a malaria infection @xcite , familial mediterranean fever @xcite , or cyclic neutropenia @xcite . it is not understood whether these oscillations represent some kind of pathology or if they are part of the normal functioning of the immune system , so they are generally regarded as aberrations and are largely ignored . a discussion of immune system oscillation can be found in ref . additional chaotic oscillatory diseases such as chronic salmonella , hepatitis b , herpes simplex , and autoimmune diseases such as multiple sclerosis , crohn s disease , and fibrosarcoma are discussed in ref . @xcite . in ref . @xcite it is not possible to completely eliminate the antigen , that is , to make the antigen concentration go to zero . however , it is possible to reduce the antigen concentration to a very low level , perhaps corresponding to less than one antigen unit per host , which one can interpret as complete elimination . however , we will see that in the @xmath0-symmetric model ( [ e11 ] ) the antigen @xmath50 can actually approach 0 in the @xmath0 broken phase . in ref . @xcite it is stated that the predicted oscillations of increasing amplitude should be viewed with caution . such oscillations are predicted to involve successively lower antibody minima , which in reality may not occur . however , in ref . @xcite a modified two - dimensional predator - prey model for the dynamics of lymphocytes and tumor cells is considered . this model seems to reproduce all known states for a tumor . for certain parameters the system evolves towards a state of uncontrollable tumor growth and exhibits the same time evolution as that of @xmath50 and @xmath61 in figs . [ f15 ] and [ f16 ] . for other parameters the system evolves in an oscillatory fashion towards a controllable mass ( a time - independent limit ) of malignant cells . in this case the temporal evolution is the same as that of @xmath51 and @xmath80 in fig . [ f19 ] . in ref . @xcite this state is called a _ dormant _ state . it is also worth mentioning that in ref . two_-dimensional dynamical system describing the immune response to a virus is considered ; this model can exhibit periodic solutions , solutions that converge to a fixed point , and solutions that have chaotic oscillations . ordinarily , a two - dimensional dynamical system can not have chaotic trajectories but the novelty in this system is that there is a time delay . finally , we acknowledge that it is not easy to select reasonable parameters if one considers the application of bell s model to real biological systems . in the @xmath0-symmetric model it is also difficult to make realistic estimates of relevant parameters . nevertheless , we believe that some of the qualitative features described in this paper may also be seen in actual biological systems . we thank m. rucco and f. castiglione for helpful discussions on the functioning of the immune system . cmb thanks the doe for partial financial support and mg thanks the fondazione angelo della riccia for financial support . c. m. bender , m. gianfreda , b. peng , s. k. ozdemir , and l. yang , phys . a * 88 * , 062111 ( 2013 ) . b. peng , s. k. ozdemir , f. lei , f. monifi , m. gianfreda , g. l. long , s. fan , f. nori , c. m. bender , l. yang , nat . * 10 * , 394 ( 2014 ) . c. m. bender , m. gianfreda , and s. p. klevansky , phys . a * 90 * , 022114 ( 2014 ) . j. rubinstein , p. sternberg , and q. ma , phys . lett . * 99 * , 167003 ( 2007 ) . a. guo , g. j. salamo , d. duchesne , r. morandotti , m. volatier - ravat , v. aimez , g. a. siviloglou , and d. n. christodoulides , phys . rev . lett . * 103 * , 093902 ( 2009 ) . c. e. rter , k. g. makris , r. el - ganainy , d. n. christodoulides , m. segev , and d. kip , nat . phys . * 6 * , 192 - 195 ( 2010 ) . k. f. zhao , m. schaden , and z. wu , phys . a * 81 * , 042903 ( 2010 ) . z. lin , h. ramezani , t. eichelkraut , t. kottos , h. cao , and d. n. christodoulides , phys . lett . * 106 * , 213901 ( 2011 ) . l. feng , m. ayache , j. huang , y .- l . xu , m. h. lu , y. f. chen , y. fainman , and a. scherer , science * 333 * , 729 ( 2011 ) . s. bittner , b. dietz , u. gnther , h. l. harney , m. miski - oglu , a. richter , and f. schfer , phys . rev . lett . * 108 * , 024101 ( 2012 ) . n. chtchelkatchev , a. golubov , t. baturina , and v. vinokur , phys . * 109 * , 150405 ( 2012 ) . c. zheng , l. hao , and g. l. long , phil . . r. soc . a * 371 * , 20120053 ( 2013 ) . j. schindler , a. li , m. c. zheng , f. m. ellis , and t. kottos , phys . a * 84 * , 040101(r ) ( 2011 ) . c. m. bender , d. d. holm , and d. w. hook , j. physics a : math . * 40 * , f793 ( 2007 ) . i. v. barashenkov and m. gianfreda , j. phys . a : math . theor . * 47 * , 282001(ftc ) ( 2014 ) . a. s. perelson and g. weisbuch , rev . mod . phys . * 69*,1219 ( 1997 ) . m. plank , j. math . 36 * , 7 ( 1995 ) . r. s. desowitz , _ the malaria capers : more tales of parasites and people , research and reality _ ( norton , new york , 1991 ) ; c. m. poser and g. w. bruyn , _ an illustrated history of malaria _ ( parthenon , new york , 1999 ) .

the study of @xmath0-symmetric physical systems began in 1998 as a complex generalization of conventional quantum mechanics , but beginning in 2007 experiments began to be published in which the predicted @xmath0 phase transition was clearly observed in classical rather than in quantum - mechanical systems . this paper examines the @xmath0 phase transition in mathematical models of antigen - antibody systems . a surprising conclusion that can be drawn from these models is that a possible way to treat a serious disease in which the antigen concentration is growing out of bounds ( and the host will die ) is to inject a small dose of a _ second _ ( different ) antigen . in this case there are two possible favorable outcomes . in the unbroken-@xmath0-symmetric phase the disease becomes chronic and is no longer lethal while in the appropriate broken-@xmath0-symmetric phase the concentration of lethal antigen goes to zero and the disease is completely cured . = 1

un rsultat lmentaire bien connu de thorie des nombres , d dirichlet , est le suivant : si je prends , de manire indpendante , deux nombres choisis uniformment entre @xmath1 et @xmath2 , alors , la probabilit @xmath3 que ces deux nombres soient premiers entre eux vrifie @xmath4 . un rsultat un peu moins connu analogue est que la probabilit @xmath5 quun nombre choisi uniformment entre @xmath1 et @xmath2 soit sans facteur carr ( cest dire quon ne peut le diviser par le carr dun entier diffrent de 1 ) possde le mme comportement . voici deux rsultats dessence probabiliste , qui pourtant , ne sont pas tout fait des grands classiques de la littrature universitaire en probabilits . ils sont en revanche bien connus en thorie analytique des nombres , et dans ce contexte , les preuves proposes ont une prsentation peu probabiliste . dans le texte qui suit , nous allons ramener ces rsultats dans un cadre probabiliste , notamment laide des lois zta , qui sont des lois trs simples dotes de jolies proprits arithmtiques . un avantage de lapproche probabiliste est que lon na pas besoin destimes trs fines pour montrer la convergence . dans une partie finale , on pourra donner ainsi une preuve relativement lmentaire du calcul de la densit asymptotique des couples dentiers de gauss premiers entre eux . les preuves les plus couramment trouves dans la littrature se divisent souvent en deux tapes : * montrer la convergence de @xmath6 ( ou @xmath7 ) * identifier la limite . pour la premire tape , une preuve simple repose sur la formule du crible : prenons @xmath8 et @xmath9 deux variables alatoires indpendantes suivant la loi uniforme sur @xmath10 . on exprime alors la probabilit cherche comme le complmentaire de `` un nombre premier divise @xmath8 et @xmath9 '' et `` le carr dun nombre premier divise @xmath8 '' . notant @xmath11 la suite des nombres premiers , on optient avec la formule du crible @xmath12 o @xmath13 est la fonction de mbius : @xmath14 si @xmath2 est divisible par un carr et @xmath15 si @xmath2 scrit comme produit de @xmath16 nombres premiers distincts . en rindexant les termes et en rajoutant de nombreux termes nuls , on a finalement @xmath17 et , de manire similaire @xmath18 en utilisant un argument de convergence domine , on obtient alors la convergence de ces deux suites vers @xmath19 . pour rfrence , mentionnons quune autre approche existe , consistant exprimer @xmath20 comme une somme de la fonction indicatrice deuler , que lon exprime elle - mme laide de la fonction de mbius ] . cette approche , utilise par exemple par tissier @xcite , est un cas particulier de la mthode de lhyperbole de dirichlet . pour une description gnrale de cette technique , nous renvoyons le lecteur bordells @xcite ou tenenbaum @xcite . dans tous les cas , le problme didentifier la somme de cette srie demeure . la manire la plus efficace est dutiliser un produit de dirichlet : on note que @xmath21 o @xmath22 dsigne la convolution arithmtique des suites . @xmath23 comme @xmath24 ( cest la formule dinversion de mbius ) , on obtient lidentification recherche . , o @xmath25 est la fonction nombre de diviseurs . ] donnons une brve preuve combinatoire de la formule dinversion de mbius : soit @xmath26 . * si @xmath27 , lidentit @xmath28 est immdiate . * sinon , si les facteurs premiers sont @xmath29 , les seuls diviseurs @xmath30 de @xmath31 qui apportent une contribution non nulle @xmath32 scrivent comme produit des lments dune partie de @xmath33 , de sorte que + @xmath34 lapplication @xmath35 cre clairement une bijection entre les parties de @xmath33 ne contenant pas @xmath36 et celles la contenant . lvidence , les deux parties ainsi associes apportent des contributions opposes la somme totale , qui est donc nulle . on notera que ces preuves sont essentiellement combinatoires , et font toutes lusage dau moins deux formules de convolution ou dinversion : en effet , la formule de poincar , dune part , la formule de mbius , dautre part , sont toutes deux des formules dinversion . mais si le problme se rsout laide de deux inversions , peut - tre pourrait - on nen faire aucune ? on est donc tent de se demander si on ne peut pas trouver une preuve plus directe , ou plus probabiliste . une manire simple de deviner la limite est de passer par les lois zta . pour @xmath37 , on appelle loi zta ( ou loi de zipf ) de paramtre @xmath38 la loi sur @xmath39 qui assigne la masse @xmath40 au point @xmath2 . le coefficient de renormalisation @xmath41 est la clbre fonction @xmath42 de riemann , dfinie pour @xmath37 par @xmath43 [ limzeta ] la fonction @xmath42 a une limite infinie en @xmath1 . cela se voit aisment par un argument de monotonie : comme la srie harmonique diverge , pour tout @xmath44 on peut trouver @xmath2 tel que @xmath45 . par continuit de la fonction @xmath46 , il existe @xmath47 tel que @xmath48 pour @xmath491,s_0\mathclose[$ ] . il est alors clair que @xmath50 pour @xmath491,s_0\mathclose[$ ] . les lois zta apparaissent naturellement comme consolatrices de linexistence dune loi uniforme sur @xmath39 . prcisment , on a [ calculproba ] soit @xmath51 . si @xmath8 est une variable alatoire sur @xmath39 telle que tout tout @xmath52 , la probabilit de lvnement @xmath53 ( @xmath2 divise @xmath8 ) vrifie @xmath54 , alors pour des nombres @xmath55 deux deux premiers entre eux , les vnements @xmath56 , , @xmath57 sont globalement ( ou mutuellement ) indpendants . soient @xmath58 des entiers distincts entre @xmath1 et @xmath31 . comme les entiers @xmath59 sont premiers entre eux , on a @xmath60 donc @xmath61 pour @xmath37 , considrons @xmath8 suivant la loi zta de paramtre @xmath38 . pour tout entier naturel @xmath2 non nul , on a @xmath62 ainsi , la loi zta de paramtre @xmath38 satisfait la proprit requise par le lemme , que lon peut donc appliquer : @xmath63 lgalit @xmath64 mrite quelques mots : dabord , on utilise le fait que si une famille finie dvnements est indpendante , alors la famille forme par leurs complmentaires lest aussi . pour cela , il suffit de montrer que si on remplace un des lments de la famille par son complmentaire , la famille alors obtenue est encore indpendante , puis ditrer . pour conclure , il suffit dutiliser le thorme de continuit squentielle dcroissante : pour toute suite dcroissante dvnements , la probabilit de lintersection est la limite des probabilits . nous venons ainsi de donner une preuve probabiliste dune formule clbre de la thorie des nombres , initialement de euler : @xmath65 en passant au logarithme , on obtient pour tout @xmath37 , @xmath66 en faisant tendre @xmath38 vers @xmath1 , on obtient la divergence de la srie de terme gnral @xmath67 et , par quivalent , de la srie des @xmath68 ainsi , pour @xmath490,1\mathclose]$ ] , sil existait une variable alatoire sur @xmath39 telle @xmath54 pour tout @xmath52 , le deuxime lemme de borel cantelli vrifient @xmath69 , alors la probabilit que lon voit simultanment la ralisation dune infinit des vnements @xmath70 vaut 1 . voir par exemple garet et kurtzmann @xcite . ] assurerait @xmath71 ce qui est absurde , car lvnement en question est vide . en particulier , il nexiste pas de variable alatoire sur @xmath39 telle @xmath72 pour tout @xmath52 . on restera donc avec @xmath37 . soit donc @xmath8 suivant la loi zta de paramtre @xmath38 . en procdant comme prcdemment , on a @xmath73 o la dernire galit vient de . de la mme manire , si @xmath8 et @xmath9 sont indpendants et suivent la loi zta de paramtre @xmath38 , la probabilit que @xmath8 et @xmath9 soient premiers entre eux est la probabilit quaucun nombre premier ne divise les deux ; par indpendance et avec le lemme [ calculproba ] , cest encore @xmath74 . lorsque @xmath38 tend vers @xmath1 , ces probabilits tendent vers @xmath75 , qui est prcisment la valeur des limites de @xmath3 et @xmath5 . cela nest pas surprenant . en thorie des nombres , on dit quune partie @xmath76 de @xmath39 admet une densit naturelle , et que cette densit naturelle vaut @xmath77 si @xmath78 on dit galement quune partie @xmath76 de @xmath39 admet une densit analytique , ou encore une densit de dirichlet , et que cette densit vaut @xmath77 si @xmath79 o @xmath80 est la loi zta de paramtre @xmath38 . la notion de densit de dirichlet , ou densit analytique , est trs utilise en thorie analytique et probabiliste des nombres . on peut par exemple se rfrer louvrage de tenenbaum @xcite . en particulier , il est bien connu que lexistence dune densit naturelle implique celle dune densit de dirichlet , et que dans ce cas , les deux densits concident . ainsi , si @xmath81 converge , ce ne peut tre que vers @xmath82 . par la mthode combinatoire classique , puis la limite est identifie en passant par la densit de dirichlet . ] pour lheure , nous navons pas encore de preuve purement probabiliste , qui ne require pas une certaine familiarit avec la thorie des nombres . cependant , on peut noter que la probabilit limite @xmath83 est prcisment la probabilit quune variable suivant la loi zta de paramtre 2 soit gale un . do lide dobtenir ce thorme de dirichlet comme corollaire dune convergence en loi . la notion de convergence en loi est une notion trs importante en thorie des probabilits . le paragraphe qui suit en fait une prsentation minimale , dans le cadre des probabilits sur un ensemble dnombrable . cette prsentation doit permettre de rendre lessentiel du texte accessible un probabiliste dbutant typiquement un tudiant de deuxime anne de licence des universits ou des classes prparatoires scientifiques . soit @xmath84 un ensemble fini ou dnombrable . on note @xmath85 lensemble des probabilits sur @xmath84 , cest - - dire des familles @xmath86 avec @xmath87 pour tout @xmath88 et @xmath89 . si @xmath90 et @xmath91 , on peut dfinir une fonction @xmath92 de @xmath84 dans @xmath93 par @xmath94 pour tout @xmath95 . on peut alors crire @xmath96={\ensuremath{\mathbb{e}}}[{1\hspace{-2.7mm}1}_a-{1\hspace{-2.7mm}1}_{a^c}]={\ensuremath{\mathbb{e}}}[\phi_a(x)]\\&=\sum_{x\in \omega}\phi_a(x){\ensuremath{\mathbb{p}}}(x).\end{aligned}\ ] ] lidentification des deux extrmits de lgalit ne ncessite pas la notion desprance , mais elle est alors moins naturelle . + ainsi , pour @xmath97 et @xmath91 , on a @xmath98 do @xmath99 . notons que lgalit est atteinte pour @xmath100 . ainsi , si lon pose @xmath101 , on a @xmath102 il est facile de voir quon a ainsi dfini une distance entre les probabilits sur @xmath84 . cette distance est appele distance de la variation totale . en prenant pour @xmath44 un singleton , on a facilement @xmath103 cependant @xmath104 do en faisant la somme @xmath105 on peut maintenant dmontrer le thorme suivant : soient @xmath106 des probabilits sur @xmath84 . on a quivalence entre 1 . @xmath107 2 . pour tout @xmath91 , @xmath108 3 . @xmath109 4 . pour tout @xmath88 , @xmath110 par dfinition de @xmath30 , @xmath111 entrane @xmath112 et @xmath113 et il est facile de voir que @xmath112 ou @xmath113 entrane @xmath114 . le plus difficile est de montrer que @xmath114 entrane @xmath111 , mais nous avons fait le travail prparatoire . supposons donc que @xmath115 tend vers @xmath116 pour tout @xmath117 . . de la convergence de la somme @xmath119 , on dduit quil existe @xmath120 fini tel que @xmath121 . on a @xmath122 comme la somme est finie , @xmath123 , donc pour @xmath2 suffisamment grand @xmath124 et donc @xmath125 . lorsque lune des conditions quivalentes est vrifie , on dit que la suite de probabilits @xmath126 converge en loi vers @xmath127 . si @xmath128 sont des variables alatoires sur un espace probabilis rgi par la probabilit @xmath127 , on dit que @xmath129 converge , @xmath129 converge en loi vers @xmath8 . la prcision nest apporte que dans les rares cas o il y a ambigut sur la probabilit sous - jacente . ] en loi vers @xmath8 si la loi de @xmath129 ( note usuellement @xmath130 ) converge vers la loi de @xmath8 ( @xmath131 ) . on peut galement noter que @xmath132,\end{aligned}\ ] ] donc @xmath133 . cette ingalit simple a des consquences importantes . la plus immdiate est que la convergence des variables alatoires entrane la convergence des lois associes . [ cvpsentraineloi ] si la suite de variables alatoires @xmath134 est valeurs dans un ensemble discret et que @xmath135 tend vers @xmath136 pour tout @xmath95 , alors @xmath129 converge en loi vers @xmath8 . posons @xmath137 . la variable @xmath138 est une variable alatoire valeurs dans @xmath0 , donc @xmath139 daprs les proprits classiques des fonctions de rpartition . comme @xmath140 , @xmath129 converge en loi vers @xmath8 . si @xmath141 est une suite valeurs dans @xmath85 , le procd diagonal dextraction permet den extraire une sous - suite @xmath142 telle que pour tout @xmath88 , @xmath143 converge . si lon pose alors , pour tout @xmath88 , @xmath144 , on a alors , pour toute partie finie @xmath145 de @xmath84 . @xmath146 en passant la borne suprieure , on obtient @xmath147 . bien sr , @xmath148 pour tout @xmath88 , mais cest insuffisant pour affirmer que @xmath149 , puisquil faudrait encore que @xmath150 . de fait , il peut y avoir une perte de mesure , comme on peut le voir en prenant @xmath151 et @xmath152 . cela amne dfinir la notion de tension : on dit quune suite valeurs dans @xmath85 est tendue si pour tout @xmath118 , il existe @xmath145 fini avec @xmath153 pour tout @xmath52 . on peut alors noncer le thorme suivant : de toute suite tendue de @xmath85 , on peut extraire une sous - suite qui converge en loi . reprenons la suite @xmath154 considre plus haut . cette fois - ci , pour @xmath118 , on peut trouver @xmath145 fini tel que @xmath153 pour tout @xmath52 . on a alors avec les notations prcdentes @xmath155 donc @xmath156 . comme @xmath157 peut tre pris arbitrairement petit , @xmath158 est , cette fois , une probabilit . une probabilit limite dune sous - suite extraite est appele loi limite de la suite . comme dans le cas des suites relles valeurs dans un compact , on peut noncer : si une suite tendue a une unique loi limite , alors elle est convergente . soit @xmath158 lunique valeur dadhrence . soit @xmath88 . la suite @xmath159 est valeurs dans @xmath160 $ ] qui est compact . soit @xmath154 une suite telle que @xmath161 . comme @xmath162 est tendue , @xmath142 lest aussi . on peut donc en extraire une sous - suite @xmath163 qui converge en loi . mais cette suite est extraite de @xmath164 qui na quune seule valeur dadhrence , donc @xmath165 , ce qui entrane que @xmath166 . ainsi , @xmath167 est lunique valeur dadhrence de la suite @xmath159 qui est valeurs dans un compact , donc @xmath168 . comme @xmath149 , @xmath162 converge en loi vers @xmath158 . les notions de convergence en loi , de tension , se gnralisent des lois ou des variables alatoires sur @xmath169 ou @xmath170 . pour plus dinformations , on pourra par exemple se rfrer @xcite . si @xmath171 est une famille infinie de variables alatoires , on appelle lois jointes des @xmath171 les lois des vecteurs alatoires @xmath172 , o @xmath173 dcrit lensemble des parties finies de @xmath174 . on admettra que la loi dun vecteur alatoire @xmath172 est pleinement dtermine par la fonction de rpartition inverse : @xmath175 cest un fait gnral do nt la dmonstration est immdiate dans le cas de la thorie de la mesure . on peut toutefois noter que dans le cas des variables discrtes , cela peut se dmontrer par rcurrence sur @xmath176 sans difficult particulire . on a besoin dintroduire une notation et quelques lemmes : dabord , on note @xmath177 lexposant de @xmath178 dans la dcomposition de @xmath2 en produit de facteurs premiers ( cest la valuation @xmath178-adique de @xmath2 ) . [ reconnais_lois_un ] la loi dune variable alatoire @xmath8 valeurs dans @xmath39 est caractrise par les lois jointes des variables @xmath179 . on a @xmath180 la suite @xmath181 converge ponctuellement vers @xmath8 , donc converge en loi vers @xmath8 , daprs le thorme [ cvpsentraineloi ] . si les lois jointes des variables @xmath179 sont connues , les lois des @xmath181 sont connues , et donc la loi de @xmath8 . la loi dune variable alatoire valeurs dans @xmath39 est caractrise par les valeurs de @xmath182 , o @xmath2 dcrit @xmath39 . daprs le lemme prcdent , il suffit de connatre les loi jointes des variables @xmath179 . mais pour cela , il suffit davoir les fonctions de rpartition inverse . or , @xmath183 do le rsultat . ainsi , si @xmath8 et @xmath9 sont des variables indpendantes suivant respectivement les lois zta de paramtre @xmath38 et @xmath184 , alors leur plus grand commun diviseur , not @xmath185 , vrifie @xmath186 , donc @xmath185 suit la loi zta de paramtre @xmath187 . dans le cas o @xmath8 suit la loi zta de paramtre @xmath38 , le calcul de fonction de rpartition effectu dans la preuve du lemme nous donne @xmath188 la fonction de rpartition inverse caractrisant la loi des vecteurs , on en dduit que les variables @xmath189 sont indpendantes et que @xmath190 suit la loi gomtrique de paramtre @xmath191 ( que lon note @xmath192 ) . soit @xmath128 des variables alatoires valeurs dans @xmath0 . on suppose que * @xmath193 est tendue . * pour tout @xmath194 , @xmath195 alors @xmath134 converge en loi vers @xmath8 . comme la famille est tendue , il suffit didentifier les lois limites : si @xmath196 tend en loi vers @xmath9 , @xmath197 tend vers @xmath198 ; autrement dit @xmath199 tend vers @xmath200 . donc @xmath201 et @xmath8 et @xmath9 ont mme loi . on peut maintenant noncer et dmontrer un rsultat probabiliste . [ dansz ] soient @xmath129 , @xmath202 des variables alatoires suivant la loi uniforme sur @xmath10 . on note @xmath203 et @xmath204 , o @xmath205 est le plus grand entier @xmath206 tel que @xmath207 divise @xmath2 . alors @xmath208 et @xmath209 convergent en loi vers la loi zta de paramtre @xmath210 . on a @xmath211 et @xmath212 les deux quantits convergent videmment vers @xmath213 , qui est la probabilit quune variable suivant la loi zta de paramtre 2 soit divisible par @xmath214 . reste montrer la tension . quels que soient les entiers naturels @xmath2 et @xmath206 , on a @xmath215 et de mme @xmath216 . soit alors @xmath118 . si je prends @xmath206 tel que @xmath217 , lensemble @xmath218 vrifie @xmath219 et @xmath220 , ce qui montre que lhypothse de tension est bien satisfaite . ce rsultat se gnralise aisment : soient @xmath221 des variables alatoires indpendantes suivant la loi uniforme sur @xmath10 . on note @xmath222 et @xmath223 , o @xmath224 est le plus grand entier @xmath206 tel que @xmath225 divise @xmath2 . alors @xmath208 et @xmath209 convergent en loi vers la mme loi zta de paramtre @xmath226 . la preuve est laisse au lecteur . on a ainsi retrouv un ancien rsultat , que son dcouvreur , ernest cesro @xcite , dcrivait en ces termes ( voir @xcite ) : la probabilit que la @xmath226-ime racine de la plus haute puissance @xmath226-ime , qui divise un nombre entier pris au hasard , appartienne un certain systme de nombres , ne diffre pas de la probabilit que le plus grand commun diviseur de @xmath226 entiers , pris au hasard , appartienne au mme systme . dans cette section , on propose quelques gnralisations / extensions des rsultats prcdents , toujours en privilgiant , lorsque cela est possible , lapproche probabiliste . on supposera maintenant que les thormes de convergence domine et de convergence monotone sont bien connus , particulirement lorsquon les applique des variables alatoires . une fonction @xmath227 est dit multiplicative si @xmath228 est vrifie ds que @xmath178 et @xmath229 sont premiers entre eux . si cest vrai pour tous les couples @xmath230 , alors la fonction est dite compltement multiplicative . soit @xmath231 une fonction multiplicative positive ou borne . on suppose que @xmath8 suit la loi zta de paramtre @xmath38 et que @xmath232 est une suite de variables alatoires indpendantes telles que @xmath233 . alors @xmath234&=\prod_{i=1}^{+\infty}{\ensuremath{\mathbb{e}}}(p_i^{n_i})\end{aligned}\ ] ] et @xmath235 en particulier , si @xmath231 est compltement multiplicative @xmath236 soient @xmath237 une suite de variables alatoires indpendantes telles que @xmath233 . supposons @xmath231 multiplicative et borne . daprs le lemme [ reconnais_lois_un ] , @xmath238 a mme loi que @xmath8 , donc @xmath239={\ensuremath{\mathbb{e}}}[\phi(\prod_{i=1}^{+\infty } p_i^{n_i})]$ ] ; mais @xmath231 est une fonction multiplicative , donc @xmath240={\ensuremath{\mathbb{e}}}[\prod_{i=1}^{+\infty } \phi(p_i^{n_i})]=\prod_{i=1}^{+\infty}{\ensuremath{\mathbb{e}}}(\phi(p_i^{n_i})).\ ] ] comme @xmath231 est borne , lgalit entre le deuxime et le troisime membre est une consquence du thorme de convergence domine et de lindpendance des variables @xmath241 . ainsi @xmath242 en simplifiant par @xmath243 , on obtient le rsultat voulu . on note @xmath244 $ ] lensemble des nombres de la forme @xmath245 , avec @xmath206 et @xmath246 dans @xmath247 . cest lanneau des entiers de gauss . pour @xmath248 $ ] , on pose @xmath249 . on a bien sr @xmath250 . ainsi , un inversible de @xmath244 $ ] est tel que @xmath251 est un inversible de @xmath247 : ce ne peut tre que @xmath252 . il est alors ais de voir que tout lment de @xmath253 $ ] non nul scrit dune manire unique sous la forme @xmath254 , avec @xmath255 , @xmath256 et @xmath257 : @xmath258 est le reprsentant privilgi de la classe de @xmath259 lorsque lon quotiente le semi - groupe @xmath260,\times)$ ] par ses lments inversibles . on a bien sr @xmath261 , de sorte que lapplication @xmath214 passe au quotient . ainsi , si on note @xmath262 lensemble des classes non nulles , on peut dfinir une application @xmath263 de @xmath262 dans @xmath0 par @xmath264 et lon a galement @xmath265 . la fonction @xmath263 peut se calculer explicitement , en utilisant un certain nombre de rsultats bien connus de lanneau @xmath253 $ ] , qui sont par exemple dcrits dans perrin @xcite . on sait en particulier que * @xmath253 $ ] est un anneau factoriel . * les irrductibles de @xmath253 $ ] sont * * les nombres premiers congrus @xmath266 modulo 4 ; ces nombres ne peuvent scrire sous forme de somme de deux carrs . * * les nombres de la forme @xmath245 tels que @xmath267 est un nombre premier ; tous les entiers naturels premiers qui ne sont pas congrus @xmath266 modulo 4 peuvent scrire sous la forme dune somme de deux carrs . si @xmath206 et @xmath246 sont des entiers naturels premiers entre eux , tout lment de @xmath262 de norme @xmath268 se factorise de manire unique sous la forme du produit de deux lments de @xmath262 de normes respectives @xmath206 et @xmath246 . soit @xmath269 avec @xmath270 . @xmath258 se factorise dans @xmath262 comme produits de classes dlments irrductibles de @xmath253 $ ] . la norme dun facteur divise @xmath270 , donc soit @xmath206 , soit @xmath246 puisque @xmath206 et @xmath246 sont premiers entre eux . soit @xmath117 le produit des facteurs ( pris avec leur multiplicit ) do nt la norme divise @xmath206 , @xmath271 le produit des facteurs do nt la norme divise @xmath246 . on a @xmath272 . @xmath273 divise @xmath270 et est premier avec @xmath246 donc @xmath273 divise @xmath206 . de mme @xmath274 divise @xmath246 . comme leur produit fait @xmath268 , on a @xmath275 et @xmath276 . maintenant si @xmath277 avec @xmath275 et @xmath278 , un facteur irrductible de @xmath279 est un facteur irrductible de @xmath259 do nt la norme divise @xmath206 : cest un facteur irrductible de @xmath117 . la valuation de @xmath229 dans @xmath279 ne peut tre plus grande que dans @xmath259 , puisque @xmath279 divise @xmath259 . or , par dfinition de @xmath117 la valuation de @xmath229 dans @xmath117 est gale la valuation de @xmath229 dans @xmath259 , donc @xmath229 a une valuation plus petite dans @xmath279 que dans @xmath117 , ce pour tout @xmath229 , donc @xmath279 divise @xmath117 . comme @xmath279 et @xmath117 ont la mme norme , ils sont gaux . de mme , @xmath280 . si @xmath269 est tel que @xmath281 , un facteur irrductible de @xmath258 a une norme qui est une puissance de deux . vu la caractrisation des irrductibles rappele plus haut , ce facteur ne peut tre que la classe de @xmath282 . finalement , la classe de @xmath283 est la seule classe de norme @xmath284 . si @xmath178 est un nombre premier congru @xmath1 modulo 4 , @xmath178 nest pas premier dans @xmath253 $ ] ; il se ramifie sous la forme @xmath285 , @xmath245 et @xmath286 sont premiers dans @xmath253 $ ] , non quivalents , de norme @xmath178 . si @xmath287 , un facteur premier de @xmath259 a une norme qui divise @xmath288 , donc @xmath178 : les facteurs de @xmath259 ne peuvent tre que les classes de @xmath245 et @xmath286 . ainsi @xmath259 scrit comme une puissance @xmath31-ime de la classe de @xmath245 et une puissance @xmath77-ime de la classe de @xmath286 . comme @xmath289 , cela nous donne exactement @xmath290 solutions . si @xmath178 est un nombre premier congru @xmath266 modulo 4 , @xmath178 est premier dans @xmath253 $ ] . si @xmath287 , un facteur premier de @xmath259 a une norme qui divise @xmath288 , donc @xmath178 : ce ne peut tre que @xmath178 , donc @xmath259 scrit @xmath291 , et on a @xmath292 . il ny a donc de solution que si @xmath293 est pair , et dans ce cas , elle est unique . soient @xmath232 une suite de variables alatoires indpendantes telles que @xmath233 . on a @xmath295=\prod_{i=1}^{+\infty } { \ensuremath{\mathbb{e}}}(s'(p_i^{n_i}))]$ ] ainsi , si @xmath298 est congru 1 modulo 4 , on a @xmath299=\frac1{1-p_i^{-s}}$ ] et si @xmath298 est congru 3 modulo 4 , @xmath300 . ainsi , dans tous les cas @xmath301 o @xmath302 est dfini par @xmath303 et @xmath304 . finalement , comme @xmath302 est compltement multiplicative , on a @xmath305&=\prod_{i=1}^{+\infty}{\ensuremath{\mathbb{e}}}(s'(p_i^{n_i}))\quad\text { avec~\eqref{decomp_proba}}\\ & = \prod_{i=1}^{+\infty}(1-p_i^{-s}\chi_4(p_i))^{-1}\\ & = \sum_{n=1}^{+\infty}\frac{\chi_4(n)}{n^s}\quad\text { avec~\eqref{decomp_complete}}\\ & = \sum_{n=0}^{+\infty}\frac{(-1)^n}{(2n+1)^s}=\beta(s)\end{aligned}\ ] ] comme on la fait pour les nombres entiers , on veut maintenant mesurer la propension quont les entiers de gauss tre premiers entre eux . + soit @xmath37 . en identifiant @xmath262 avec @xmath306 , on voit que @xmath262 sidentifie une partie dnombrable discrte de @xmath307 . en regroupant les termes suivant la valeur de @xmath251 , on voit que la quantit @xmath308 peut se rcrire @xmath309 o @xmath263 est la fonction tudie la sous - section prcdente . on en dduit que @xmath310 . en particulier , la srie de terme gnral @xmath311 converge , et on peut alors dfinir une loi @xmath312 sur @xmath262 par @xmath313 il est alors ais de voir que pour tout @xmath314 @xmath315 . pour @xmath321 et @xmath322 , on note @xmath323 . en identifiant un point de @xmath262 avec les @xmath324 points de @xmath253 $ ] quil contient , on a @xmath325 ainsi la loi @xmath319 est la seule loi limite possible . montrons donc la tension . on peut trouver @xmath326 tels que @xmath327 pour @xmath328 et @xmath329 pour tout @xmath330 . on en dduit @xmath331 comme la srie des @xmath332 converge , cela donne comme prcdemment la tension de @xmath333 . george e. collins and jeremy r. johnson . the probability of relative primality of gaussian integers . in _ symbolic and algebraic computation ( rome , 1988 ) _ , volume 358 of _ lecture notes in comput . _ , pages 252258 . springer , berlin , 1989 .

on ( re)visite ici avec un regard probabiliste un certain nombre de rsultats connus de la thorie analytique des nombres . au centre de larticle se trouvent les lois zta , qui nous sont une consolation de linexistence dune loi uniforme sur @xmath0 . elles nous permettront par exemple dtudier la densit naturelle des couples dentiers ou dentiers de gauss premiers entre eux , ainsi que dautres problmes analogues . au passage , on retrouvera la dcomposition de la fonction zta de riemann sous forme dun produit eulrien et une gnralisation aux sommes de fonctions multiplicatives .

the night sky brightness in the optical and near - infrared , the number of clear nights , the seeing , transparency and photometric stability are some of the most important parameters that qualify a site for front - line ground - based astronomy . there is limited control over all these parameters , and only in the case of the sky brightness is it possible to keep it at its natural level by preventing light pollution from the immediate vicinity of the observatory . previous to the installation of any observatory , extensive tests of these parameters are carried out in order to find the best locations , maximizing then the efficiency of these expensive infrastructures . however , most of these parameters are not constant along the time . we have started a program to determine the actual values of the main observing conditions for the calar alto observatory . the calar alto observatory is located at 2168 m height above the sea - level , in the sierra de los filabres ( almeria - spain ) at @xmath345 km from the mediterranean sea . it is the second largest european astronomical site in the north hemisphere , behind the observatorio del roque de los muchachos ( la palma ) , and the most important in the continental europe . currently there are six telescopes located in the complex , three of them operated by the centro astronomico hispano aleman a.i.e . ( csic - mpg ) , including the 3.5 m , the largest telescope in the continental europe . along its 26 years of operations there has been different attempts to characterize some of the main properties described before : ( i ) birkle et al . ( 1976 ) presented the first measurements of the seeing at calar alto , being the basis of the site testing ; ( ii ) leinert et al . ( 1995 ) determined the optical sky brightness corresponding to the year 1990 ; ( iii ) hopp & fernandez ( 2002 ) studied the extinction curve corresponding to the years 1986 - 2000 ; ( iv ) ziad et al . ( 2005 ) estimated the median site seeing in the observatory from a single campaign in may 2002 . a consistent study of all these properties was presented in snchez et al . ( 2007 ) , including the first results of our program . in that study , hereafter paper i , we focused on the optical properties of the night sky , presenting ( i ) the first optical night - sky spectrum , identifying the natural and light pollution emission lines and their strength , ( ii ) the moon - less night - sky brightness in different optical bands , ( iv ) the extinction and its yearly evolution and ( v ) the atmospheric seeing and its yearly evolution . it was found that most of these properties were similar to those of major astronomical sites . in particular calar alto is among the darkest places in the world in the optical range . in the current study we focus on the near infrared ( nir ) properties of the sky at calar alto . in particular we characterize here the typical and darkest sky brightness in the @xmath0 , @xmath1 and @xmath7-bands , comparing them with similar properties in other observatories world - wide . their dependencies with positioning in the sky and seasons are also analyzed . the typical seeing measured in real observations ( telescope seeing ) is derived and compared with the atmospheric seeing estimated by the seeing monitor ( out of the telescope ) . finally we estimate the fraction of useful time in the observatory based on the data obtained by the meteorological station along the last ten years ( relative humidity and gust wind ) and the presence of clouds based on the cloud sensor . the structure of this article is as follows : in section [ data ] we describe the dataset collected for the current study , including a description of data and the data reduction ; in section [ ana ] we show the analysis and results performed over the different types of data and the results derived for each one ; in section [ conc ] we summarize the main results and present the conclusions . all the magnitudes listed in this article are in the vega system . the main purpose of this study is the characterization of the sky brightness in the near infrared at the calar alto observatory . in order to address this goal we collected different observational data , taken with different telescopes and instruments and using different methods to calibrate , reduce and analyze the data . with this selection we try to minimize the effects on the results of a particular instrumental setup , maximizing the reliability of the dataset and its time coverage . as we indicated in detail in paper i the most obvious method to derive the sky brightness is to obtain direct imaging on calibration stars ( or well calibrated fields ) , using the well known photometry of the observed stars to self - calibrate the images and then derive the sky surface brightness . with this purpose , observations were taken using the magic nir camera ( herbst et al . 1993 ) at the 1.23 m telescope during the nights of the 30th and 31st of january 2008 . although a strong dependency of the nir sky brightness with the moon ( at least for the @xmath1 and @xmath7 bands ) is not expected , both nights were selected to be dark . in both nights we used the j , h , ks , k and km bands . the magic detector is an rockwell nicmos3 array of 256@xmath8256 pixels , with a scale of 1.2@xmath6/pixel and a field - of - view of 5.1@[email protected]@xmath9 , enough for the purpose of this study . magic is under operation at calar alto since 1995 , and it is not in regular use anymore . on the 30th of january we observed the core of m67 , taking 24 individual frames of 60s per band ( splitted in individual integrations of 2s each one ) . this cluster was selected since there are publically available nir photometry for most of its members and on the indicated date it remains near the zenith ( [email protected] ) during most of the night . on the 31st we took 24 individual frames of 60s per band on a single calibration star from the ukirt faint photometric standards ( casali & hawarden 1992 ) , observing it at different airmasses . in both nights the observations were started two hours after the twilight , in order to minize as maximum its possible effects ( if any ) . the data were reduced following standard procedures for the near infrared , using iraf @xcite and self programed routines . each frame was corrected by its corresponding dark frame and flatfielded using flux normalized domeflats . bad pixels were identified using both the dark and the domeflats and masked from the individual frames . since we are interested in the sky brightness no sky subtraction was applied . once the images were reduced we measured the number of counts within an aperture of 8@xmath6 for the brightest stars detected in the field ( in the case of the 30th of january data ) or the photometric standard ( in the case of the 31st of january data ) . the number of counts from the sky was determined by measuring it directly in an annular ring of 14@xmath6 and 27@xmath6 ( inner and outer radii ) . after subtracting the sky to the counts of the detected stars the photometric zeropoint for the image was determined using the formula : @xmath11 where mag@xmath12 is the magnitude zeropoint , mag@xmath13 is the apparent magnitude of the calibration star in the corresponding band , `` ext '' is the extinction , `` counts '' are the measured counts within the indicated aperture , and t@xmath14 is the exposure time . when more than one star was analyzed we adopted the mean value of the derived zeropoints as the image photometric zeropoint . after that the surface sky - brightness was derived for each image by using the formula : @xmath15 where sb@xmath16 is the surface brightness in magnitudes per square arcsec , mag@xmath12 is the zeropoint described before , counts@xmath16 is the sky counts level estimated and `` scale '' is the pixel scale in arcseconds . it is noticed that the magnitude zeropoint was corrected by the extinction , but the sky brightness was not , following the convention adopted in most of the recent studies of sky brightness ( paper i and references therein ) . correcting the sky brightness for extinction would be appropiate only if the extinguishing layer were known to be below all sources of sky brightness , which is not the case ( benn & ellison 1998a , b ) . for the extinction we used the formulae presented in paper i , that allows to estimate the extinction in the nir bands based on the @xmath17-band extinction from the calar alto extinction monitor ( cavex , pi : u.thiele , see paperi ) . in both nights the extinction in the nir was almost neglectible . we finally got 24 individual estimations of the surface sky brightness for each band and night . the alhambra survey ( moles et al . 2008 ) is the largest cosmological survey currently on going at the calar alto observatory . it comprises the observation of 4 discontinuous square degrees on the sky in 8 different locations using 20 middle - width filters in the optical range and three additional ones ( j , h and ks ) in the near infrared . the locations were selected to cover areas of low galactic extinction , without very bright stars in the field , and/or corresponding areas already covered by other extragalactic surveys ( like cosmos , hdfn or the groth - strip ) . the main goal of this survey is to derive high quality photometric redshift of about @xmath31.6 millon of galaxies , in order to study the evolution of galaxies up to @xmath181 . the nir observations of the survey are taken with the omega2000 camera mounted in the 3.5 m telescope at calar alto . this camera has installed a hawaii-2 detector with an array of 2048@xmath82048 pixels , each one with projected size of 0.447@xmath6 . thus , the detector covers a field - of - view of @xmath315@xmath9@xmath815@xmath9 . omega2000 is under operations at calar alto since 2001 , and it has been for a long period of time one of the nir cameras with the largest field - of - view in the world . to observe the proposed field - of - view and to achieve the proposed depth ( @xmath1920 mag , 5@xmath20 detection limit ) the alhambra survey obtains several individual frames of 80s ( j ) , 60s ( h ) and 46s ( ks ) each one , at many different locations on the sky , being an unique dataset to study the properties of the nir sky brightness ( as an unproposed legacy ) . the alhambra survey has accumulated observations with omega2000 along 75 nights , from august 2004 to march 2008 , with about @xmath3150 individual frames per night . that means about @xmath33500 individual frames on target per band . these nights corresponds to a @xmath367% of the allocated nights for the project in that time interval , which reinforces our claim in paper i that a @xmath370% of the nights are useful for astronomical observations at calar alto ( see section 3.3 for a full discussion ) . to automate and homogenize the data reduction a semiautomatic pipeline was created ( cristobal et al . 2008 , in prep . ) . this pipeline performs the typical reduction steps for nir imaging data , including the dark - current subtraction , flat - fielding , image combination , sky determination and subtraction and flux calibration . the flux calibration was performed self - calibrating the images using the nir magnitudes of the field stars extracted from the 2mass survey @xcite . the typical error of the magnitudes of each individual star is about @xmath30.03 mag , with a range of errors between @xmath30.02 and @xmath30.10 mag . the large field - of - view of omega2000 grants the detection of a sufficiently large number of them ( @xmath2120 ) to obtain an accurate flux calibration , better than @xmath30.025 mags ( cristobal et al . , in prep . ) . as a by product of the pipeline the sky background is determined for each combined image . due to its nature the different pointings of the survey are observed in different nights , at different azimuths and elevations ( and airmasses ) . therefore , the derived sky backgrounds from the combined frames can not be used to characterize particular nights and/or atmospheric conditions . however , they are still useful indicators of the average sky background in the observatory . for the present study we collected the night sky backgrounds available from the alhambra reduced dataset , all of them corresponding to 12 different nights of the same month , august 2004 ( the 1st run of the survey ) . these data have the advantage of their high signal - to - noise , due to the depth of the combined images , and the fact that they have been reduced using a refined pipeline . the final sample comprising a set of 17 individual estimations of the sky background per filter . a simpler procedure was developed to obtain the sky brightness for each individual frame obtained by the survey , which comprises a much larger dataset . for each individual frame taken on the alhambra fields the approximate location of the field stars with nir magnitudes derived by the 2mass survey were obtained . poststamp images with a field - of - view of 40@xmath6@xmath840@xmath6 around these locations were taken . in each individual poststamp the location of the field star was determined by a simple centroid determination , masking the pixels up to 5@xmath20 the standard deviation over the median background . then , the counts within an aperture of 6@xmath6 radius around the determined centroid were measured , subtracting the background counts estimated within a ring between 14@xmath6 and 20@xmath6 radius and the nominal dark - current of the instrument ( kovcs et al . simultaneously the median counts within four different boxes of 18@xmath818 pixels located at 10@xmath6 away from each centroid was derived as an estimation of the counts of the sky brightness . the estimated counts per star were used to derive a magnitude zero - point for each star in the field with known photometry using the formulae ( 1 ) , and the corresponding sky surface brightness using formulae ( 2 ) and the estimation of the sky counts . the number of stars detected per field with accurate 2mass photometry grants the quality of the derived surface sky - brightness , despite the fact that the accuracy of the individual photometry is not that good in a few cases , as indicated before . the fact that no flat - field correction was applied should not affect strongly the results due to the small variations in the efficiency pixel - to - pixel in the omega2000 array . we finally got a sample of 1954 , 2394 and 2963 individual estimations of the night sky surface brightness for the @xmath0 , @xmath1 and @xmath2 bands . due to the number of measurements and the time coverage it comprises one of the largest samples of this nature ever obtained to our knowledge . as a by - product of the analysis it was derived the average fwhm on each analyzed image . for doing so it was extracted an horizontal and vertical intensity cut along and across the estimated centroid in each poststamp image . the intensity cuts were then fitted with a single gaussian function plus a continuum intensity to remove the background . the average of the horizontal and vertial fwhms is taken as the fwhm of the image of each star , and then , the median value of the fwhms derived on the same image is stored . once removed the effect of the pixel scale , the final fwhm can be considered as the seeing in the detector plane in the considered band . a summary of all the data set collected for this study is listed in table [ tab_data ] , including the date of the observations or the projects from where they were extracted , the telescope and instrument used , the bands covered by the data and the number of individual frames analized ( the number of individual sky background estimations ) . the atmospheric conditions in the observatory has been continously monitorized by a computerized meteorological station during the last decade . this station measures the most important meteorological parameters , such as the surface temperature , the relative humidity , the wind speed and direction , the wind gust speed and the air preasure , and store it every 30 seconds . these data comprises a huge dataset with 3541161 individual measurements of each parameter for the time period between january 1998 and december 2008 , that allows to characterize the atmospheric conticions at calar alto . the near infrared sky brightness is dominated by many instrically narrow hydroxyl ( oh ) emission lines ( eg . , oliva & origlia 1992 ; rousselot et al . 2000 ; cox 2000 ) . a few other species ( eg . , molecular oxygen at 1.27 @xmath22 m ) also contribute , as do h@xmath23o lines at the long wavelength end of the @xmath7-bands . another contributions to the sky background is zodiacal emission ( eg . , harwit et al . 1969 ; content 1996 ; cox 2000 ) , thermal emission from the atmosphere and the telescope ( which follows a graybody curve , content 1996 ) , and the moonlight ( eg . , cox 2000 ) . all these effects contributes differently at different wavelength ranges . in the @xmath0-band the moon - light may have a strong effect , if it is observed near the moon . on the other hand , its effect is neglectible in the @xmath1 and @xmath7 . in general , the dominant effect in the @xmath0 and @xmath1 bands is the oh emission lines , although the effect of the zodical light is not neglectible , being similar in both bands ( eg . , content 1996 ) . however , in the @xmath7-band the thermal emission becomes the dominant effect , in particular the telescope one ( eg . , cox 2000 ) , being less important the oh emission lines , and the optical airglow is almost neglectible . the night sky brightness suffers strong variations along each night which imposse limitations to the concept of _ typical _ sky brightness . the oh emission varies diurnally , being stronger in the daytime and at twilight . it may decrease by a factor 2 or 3 in one or two hours after the twilight , and it does not increase again just before the sunrise . its emission roughly scales with the airmass , which produces an increase of the sky brightness at higher airmasses , expecially in the @xmath0 and @xmath1 bands . the variation in the night sky brightness due to all these effects can be appreciated in figure [ mag_ut ] . this figure shows all the individual night sky surface brightnesses that comprise the dataset derived from the alhambra survey ( described before ) , along the universal time corresponding to each of the 75 observed nights . the wider range of surface brightness it is found in the @xmath2-band , while the narrower one corresponds to the @xmath0-band . as a rough estimation of the typical sky brightness it was derived the median sky brightness of each data - set . for a better characterization it was also derived the darkest sky brightness and the brightest one , which will illustrate the strength of the variation of the sky brightness . for the full sample of the alhambra dataset , which includes data taken at many different epochs this latter parameter were not derived , since it turns to be meaningless . all these estimations of the sky brightness are listed in table [ tab_med ] , for each data - set and each band . the values listed in the table illustrate how strong is the variation of the sky brightness , not only within a certain night , but also night - to - night . in this context the median sky brightness is the less useful parameter to characterize the sky brightness , since it was included in its derivation data taken at different times from the twilight and at different airmasses . a good example of the biases that can be found by adopting the median sky brightness are the data of the two nights observed at the 1.23 m telescope . they correspond to observations on two consecutive nights , using the same instrument at the same telescope . however , there is a strong difference in the median sky brightness derived for each of these nights , at least for the @xmath0 and @xmath1 bands . as we will show later the main reason for these differences is due to the air - mass range of the observations taken each night . while the 1st night the observations were restricted almost to the zenith ( @xmath24 ) , in the second one it was explored a wide range of airmasses . this may explain also the strong differences found among the brightest sky brightness for the same bands for the different datasets . a more reliable parameter to characterize the sky brightness is the darkest sky brightness derived for each band in the different datasets . although this parameter also presents a variation , this is much reduced than the one found in the median sky brightness . however it does not illustrate the expected sky brightness at the observatory in any conditions , it just represents the darkest ever possible sky that can be found on the data set . as expected the darkest sky - brightness is found in the largest dataset , since it covers the wider range of atmospheric conditions , and therefore it has the largest probability of catching the optimal ones . as indicated before the sky brightness presents a dependency with the zenith distance . in the optical range this dependency is mostly due to the airglow , an effect that is stronger at largest airmasses and smaller at the zenith ( eg . , paper i and references therein ) . in the nir there are additional components that depend mostly on the oh emission ( and the thermal emission ) , and which effect is minimized also at the zenith . in general the dependecy of the sky brightness with the airmass ( ie . , the zenith distance ) can be modeled by the following formula : @xmath25+\kappa ( x-1)\ ] ] the first two parameters within the logarithm of this formula are described in patat ( 2003 ) , being @xmath26 the increase in sky brightness at a certain band and airmass ( @xmath27 ) , @xmath28 the fraction of the total sky brightness generated by airglow , and ( 1-@xmath28 ) the fraction produced outside the atmosphere ( hence including zodiacal light , faint stars and galaxies ) . finally , @xmath29 is the extinction coefficient at the corresponding wavelength . as already indicated in paper i these two parameters describe correctly the dependency of the sky brightness with the airmass for the optical range . indeed the fraction @xmath28 is almost constant for the optical range , being of the order of @xmath30.6 ( eg . , benn & ellinson 1998a , b ) . however in the near - infrared the fraction of night sky light generated in the atmosphere can be larger than the one generated outside the atmosphere , and therefore `` f '' can be larger than 1 . this formula is an approximation for low airmasses of the so - called van rhijn formula ( e.g. , chamberlain 1961 ) , which describe the increase of the airglow emission when the zenith angle increases . the extinction correction has to be taken into account since the light , emitted in a layer on top of the atmosphere , is attenuated by the atmospheric extinction . figure [ mag_air ] shows the distribution of the nir surface sky brightness along the airmass for the different bands observed in the night of the 31st of january 2008 . as expected , there is a clear dependency of the surface brightness with the airmass , particularly strong in the j and h bands . a similar result is found in any night within our different collected dataset . this particular night was shown just as an illustrative example , since its dataset covers the larger number of filters and a wide range of airmasses . the sky - brightness / air - mass distribution derived for each night and filter were fitted with the parametrization described before . for the @xmath0 and @xmath1 band it was obtained that a value of @xmath302.6 reproduces well the air - mass dependency of the night sky brightness . the similarities between both bands are mostly due to the fact that the sky brightness is dominated in both of them by very intense oh bands . the @xmath0-band covers the 7 - 4 and 8 - 5 oh bands , while the @xmath1-band covers the 4 - 2 , 5 - 3 and 6 - 4 ones ( eg . , chamberlain 1961 ; maihara et al . 1993 ; leinert et al . 1998 ) . on the other hand , the air - mass dependecy is weaker in the @xmath7-band . a `` f '' value of @xmath30.4 reproduces well its shape . the @xmath7-band is less affected by the oh bands . only a part of the 9 - 7 band enters at wavelengths lower than 2.2@xmath22 m , and there is wide gap without any oh band between 2.2@xmath22 m and 2.7@xmath22 m ( eg . , leinert et al . 1998 ) . for this band the thermal contribution starts to be important ( eg . , content 1996 ; leinert et al . 1998 ; glass 1999 ) . the results from the parametrization shown in the previous section were used to correct the sky - brightness from its dependency with the airmass and to transform all the measured values to the corresponding value at the zenith . as already indicated in paper i , and references therein , these values are a good characterization of the typical sky - brightness for an astronomical site . table [ tab_zenit ] lists the median sky - brightness for the different considered bands after applying the considered correction for the dataset corresponding to the full sample of alhambra data . in addition it is listed the median values of the uncorrected sky - brightnesses restricted to the airmasses near to the zenith ( @xmath31 ) for the considered dataset . although this later is a much limited sample ( @xmath3150 measurements ) , the good agreement between both estimations of the zenithal sky - brightness , within @xmath30.1 mag , reinforces the validity of the adopted correction . in order to test the effects of the thermal emission on the sky - brightness the data were splitted in two subsamples , one covering the colder epoch of the year ( from november to february , so called _ winter _ ) , and another comprising the rest of the year ( so called _ summer _ ) . the median zenithal sky - brightness were derived for both datasets and listed in table [ tab_zenit ] . only in the @xmath2-band is appreciated a clear seasonal difference , with the winter sky - brightness being @xmath30.8 mags darker than the summer one , on average . these differences in the sky brightness along the year are clearly appreciated in figure [ mag_date ] , where the median zenithal sky - brightness for each night and filter are represented along the year . again , only for the @xmath2-band the sky - brightness shows a clear pattern , with a change associated with the seasons , being brighter in the summer and darker in the winter . this result is consistent with the expected origin of the sky - brightness at the different bands , since the thermal contribution should affect strongly the @xmath2-band . to investigate the actual dependency of the @xmath2-band sky - brightness with the temperature we derived the average air temperature measured by the meteorological station for the nights with measured sky - brightness , comparing both parameters . in figure [ mag_temp ] it is shown the distribution of the average , air - mass corrected , sky brightness for the @xmath0 , @xmath1 and @xmath2-band along the average temperature for the considered nights . as expected there is a clear dependecy of the sky brightness with the temperature ( with a correlation coefficient of @xmath320.88 ) , only for the @xmath2-band data . a linear regression derives an increase of @xmath30.08 mag in the sky brightness per 1@xmath5 increase in temperature . the temperature at calar alto presents a seasonal sinusoudal pattern , with a minimun in winter ( december - january ) at @xmath48@xmath5 c and a maximum in summer ( july - august ) at @xmath312@xmath5 c. due to this thermal variance it should be possible to find differences in the sky brightness of @xmath32 mag between the coldest nights in winter and the hottest in summer . it has been long suspected that the thermal background in the 3.5 m telescope is larger than in rest of the telescopes at calar alto , due to its larger structure . however this claim has never been tested . although we do not have a similar sample of sky - backgrounds measured in the rest of the telescopes , the fact that the @xmath2-band background correlates so tightly with the atmospheric temperature make us to suspect that this claim is not correct . indeed , the two estimations of the sky - brightness taken at the 1.23 m , marked with two big circles in figure [ mag_temp ] lie exactly in between the remaining data , taken at the 3.5 m telescope . as indicated in the introduction the night sky - brightness is a fundamental parameter to estimate the quality of a certain astronomical site . a fair comparison of this parameter among different observatories is the best method to establish their real status . in order to perform this exercise we collected all the published estimations of the night sky - brightness available in the literature and in the observatories webpages at different sites . the result from this collection is listed in table [ tab_other ] , including the name of the observatory where the data were obtained , the kind of measurement listed ( average or darkest sky brightness ) , the measured sky - brigthness for the @xmath0 , @xmath1 and @xmath2-bands and the reference from where the data were extracted . to perform a fair comparison between the sky brightness at these different sites and calar alto it is required to use the same kind of measurement . for the darkest sky brightness we will use the corresponding values listed in table [ tab_zenit ] , and for the average ones we will use those listed in table [ tab_med ] . in some cases , like in cerron pachon and mt.graham , it is not clearly stated in the references if they correspond to the average or the darkest sky brightness . we adopted the less advantage situation for calar alto for the purposed comparison and consider their measurements as the average ( that is always brighter ) instead of the darkest . regarding the darkest sky - brightness it is clear that for the @xmath0 ( 16.96 mag ) and @xmath1-bands ( 14.98 mag ) calar alto can be quoted among the darkest astronomical sites in the world ( or at least the ones listed here ) . for the @xmath2 band ( 13.55 mag ) its sky - brightness is clearly comparible with most of the astronomical sites , apart from mauna kea , which is , by far , the darkest place for this considered wavelength range . this is expected since this site is located by far at the highest altitude , as it is seeing in table [ tab_other ] , with the most tenuous atmosphere , and therefore the thermal background there is much lower than in any of the other sites . for the average sky - brightness we can only compare with the published sky - brightnesses from el roque de los muchachos observatory at la palma island and mauna kea . in the @xmath0-band calar alto seems to be darker than la palma or mauna kea by @xmath30.3 mag and in the @xmath1-band the three places seem to be equally dark . on the other hand , in the @xmath2-band calar also is slighly brighter than la palma , by @xmath30.2 mag , and both places are much brighter than mauna kea , by @xmath31 mag , as expected . however , considering the inhomogeneity of the collected data and the lack of details of how they were derived we consider that only the differences found with mauna kea in the @xmath2-band are really significative . in paper i we published the median site seeing at the calar alto observatory estimated from a continuous night monitoring of the atmospheric seeing performed with a dimm monitor since august 2004 ( robodimm , pi : j.aceituno ) . however it is well known that the site seeing is not in most cases the actual seeing measured in the detector when an astronomical image is taken . there are several reasons for this effect , being the stronger ones : * the dome seeing . the turbulence created by the temperature differences between the air inside the dome and outside the dome creates a local seeing . for that reason modern telescopes are build with smaller domes or domes that open wider than tradditional telescopes . in this regards the large domes of the equatorial mounted telescopes at calar alto are clearly a handicap which in principle could create bad dome seeings . in the last years there has been a large upgrade in the domes at calar alto . new windows were opened in the dome and large fan systems were installed to create laminar air - flows that cool down the dome and reduce the turbulence , in order to limit the effects on the seeing . * the mirror seeing . the differences between the temperature on the dome and the surface of the primary mirror creates turbulence in its vicinity that alters the measured seeing . due to its larger thermal capacity , mirrors are cooled down slower than the dome , and therefore this effect may affect a sustantial fraction of the astronomical night . to reduce this effect it was installed a cooling system that pumps evaporated nitrogen to the primary mirror while the telescope is parked during the day , reducing the thermal gradient and its effect over the seeing . * optical problems with the telescope , like bad focusing , misaligments of the primary and secondary mirror ( or prime - focus instrument ) , astigmatig effects . * mechanical or electronical problems like bad balancing ( that creates vibrations and fleasures ) , tracking and/or guiding problems . for all these reasons it is interesting to estimate the seeing measured in the astronomical images and compare it with the site seeing to estimate the quality not only of the observatory but also of the telescopes installed on it . for doing so we used the seeing estimated for each frame of the full alhambra survey dataset , described in section [ alh ] . it is known that the seeing presents a strong wavelength dependency , being smaller in the redder bands than in the bluer ones . to perform a fair comparison with the site seeing measured by the dimm at the @xmath17-band it is required to transform the estimated seeing at each nir band to the former one . for doing so we adopted the transformation derived by sarazin & roddier ( 1990 ) : @xmath33 after applying this transformation its was obtained a final sample of 7311 individual seeing estimations on the detector at the @xmath17-band that can be directly compared with the site seeing measured by the seeing monitor . figure [ seeing ] shows the distribution of the seeing measured directly at the detector together with the distribution of the site seeing measured by the dimm monitor in the same time period . as expected the former distribution has an offset towards larger seeings than the previous one , but there is no evidence of a broader tail . this indicates that although the telescope seeing is worst than the site one , there is little evidence of disastrous seeings . the stronger contribution to the seeing at the detector is site one , and not the remaining effects descrived before . the median seeing measured at the detector is @xmath31.0@xmath6 ( converted to the v - band ) , while the site seeing is @xmath30.9@xmath6 for the same time period , with a standard deviation of @xmath30.2@xmath6 in both cases . therefore , the added degradation of the seeing due to the technical characteristics of the telescope / instrument is statistically restricted to @xmath310% of the site seeing value . for comparison purposes the measured seeing distribution in the @xmath2-band is also shown in fig . [ seeing ] ( dash - dotted line ) . the median seeing in this band is @xmath30.77@xmath6 , with a standard deviation of @xmath30.13@xmath6 . a @xmath350% of the nights the seeing is lower than this typical value . a fundamental parameter to estimate the quality of a certain astronomical site is the fraction of time that it is possible to observe . this concept is always relative to the instrumentation / telescopes installed in the considered observatory . for example , radio telescopes can operate both during night and day time , while optical ones can only operate during the so - called astronomical night . in this regards the fraction of useful time has to be computed differently for different telescopes . the telescopes at calar alto were designed to operate in a wide range of atmospherical conditions . in general they can operate with relative humidities down to a 95% and wind gust velocities down to 20 m / s . in certain conditions of temperatures , cloud coverage and atmospheric preasure they can still operate with relative humidities as high as a 98% and wind gust velocities as high as 26 m / s ( at least the 3.5 m telescope ) . therefore the former limit is a conservative one to define when it is possible to observe . the fraction of useful time was estimated for each night using the database of atmospheric parameters described in section [ meteo ] , and adopting the definition of useful time as that time when the telescopes can operate , on the basis of the conservative limit described before . for the decade comprised in the dataset it was found that it was possible to observe in a @xmath372% of the night time , and in a @xmath341% of the nights the atmospheric conditions allows to observe for the full night . this estimation agrees with the fraction of useful time derived on the basis of the data from the extinction monitor described in paper i. of the two parameters that define the adopted limit , the humidity is the more critical one since only in very rare cases a wind gust speed of 20 m / s is reached . figure [ humi ] shows the distribution of nightly average relative humidity along the time period sampled by the meteorological station . there is a clear seasonal pattern , with dry and more stable summers and wet and more instable winters . the peaks of high relative humidity in winter are due to low altitude clouds and rain . however , the driest periods are also present in winter , although less frequently and stable than in summer . these periods are due to freezing temperatures that drys the atmosphere . this pattern is reflected in the fraction of useful time , that in summer is @xmath385% in average while in winter drops to a @xmath355% . this estimation has the caveat that in certain conditions it is still possible to have a covered sky by clouds while the humidity is lower than a 95% . with these conditions it will be possible to observe or not depending on the thickness of the clouds . in these cases the extinction measured by the extinction monitor is a good estimation of the it is possible to observe or not . however , as we indicated before there is no big differences between the fraction of useful times derived by the analysis of the atmospheric conditions and by the measurements of the extinction . in order to quantify the presence of clouds it was installed in the observatory a cloud sensor in october 2006 . this monitor determines the presence of clouds by estimating the difference in temperature between the sky and the ground , on the basis of the infrared emission of the former one . this instrument is monitoring this temperature every minute since it become operational , which provides us with a sample of 466142 individual estimations of this parameter . although it does not distinguish between thick and thin clouds , it is possible to stablish when the sky completely covered , that it is when the sky is less than 15@xmath5 c colder than the ground . by impossing this additional limitation to our adopted definition of useful time its actual fraction drops to a @xmath368% . it is interesting to note here that this fraction is very similar to the fraction of useful time derived directly from the observations of the alhambra survey ( section [ alh ] ) . we conclude that the fraction of nights with low humidity and calm wind and covered by clouds is rather small . in this article we have continued the characterization of the main properties of the night - sky at the calar alto observatory that we started in paper i. these properties were compared with similar ones of other different astronomical sites . the main results of this article can be summarized in the following points : * the night sky brightness at the near - infrared presents a strong airmass dependency in the @xmath0 and @xmath1 band , and a mild one in the @xmath2-band . these dependencies can be modelled and corrected ( or predicted ) for any night and airmass by assuming that they are due to airmass dependency of the oh lines emission . obviously , @xmath0 and @xmath1-band observations are recomended to be performed near the zenith , while at the @xmath2-band this is less critical . this result should be taken into account when defining observing strategies in the future . * there is a seasonal pattern in the @xmath2-band sky brightness , which shows an average difference of @xmath30.8 mag from summer , when is brighter , to winter , when is darker . this pattern is due to the strong dependecy of the sky - brightness in this band with the atmospheric temperature , which may indicate that the contribution from the thermal background emission is the dominant component in this band . * the brightness of the night sky at calar alto is similar or darker than any in other major astronomical site in the @xmath0 and @xmath1 bands , including both paranal and mauna kea . in the @xmath2-band all the astronomical sites are very similar apart from mauna kea , which is clearly darker ( by @xmath31 mag ) . indeed , calar alto has a similar sky brightness than paranal or la palma at this wavelength range . * the seeing on the detector is only slightly larger than the site seeing measured by the dimm , outside the dome ( at least for the 3.5 m telescope ) . the effects different than the purely atmospheric ones produce a statistical increase of a @xmath310% in the measured seeing on the detector compared with the site one . * the fraction of useful time at the observatory is of the order of @xmath370% , with @xmath340% of the nights 100% completely useful . these fractions are better during summer and worst during winter . we finally conclude that our additional analysis of the astronomical conditions at calar alto agree with our previous results presented in paper i. the new data strength our previous conclusion that it is a good astronomical site , similar in many aspects to places where there are 10m - like telescopes under operation or construction . for both reasons we consider that this observatory is a good candidate for the location of future large aperture optical / nir telescopes . sfs thanks the spanish plan nacional de astronoma program aya2005 - 09413-c02 - 02 , of the spanish ministery of education and science and the plan andaluz de investigacin of junta de andaluca as research group fqm322 . llrrldate(s ) & telescope & instrument & bands & number of frames + 30/01/08 & 1.23 m & magic & j , h , ks , k , km & 24 per band + 31/01/08 & 1.23 m & magic & j , h , ks , k , km & 24 per band + alhambra(1 ) & 3.5 m & omega2000 & j , h , ks & 17 + alhambra(2 ) & 3.5 m & omega2000 & j , h , ks & 1954,2394,2963 + \(1 ) 7 nights of observations in august 2004 from the alhambra survey . analysis based on the reduced and combined frames . + ( 2 ) 75 nights of observations between august 2004 and march 2008 from the alhambra survey . analysis based on the individual frames . + lllllldescription & j & h & ks & k & km + + median & 16.45 & 14.36 & 12.57 & 11.97 & 12.78 + brightest & 16.18 & 14.18 & 11.49 & 11.88 & 12.69 + darkest & 16.69 & 14.54 & 12.73 & 12.15 & 12.88 + + median & 15.53 & 14.03 & 12.54 & 12.02 & 12.76 + brightest & 14.91 & 13.39 & 12.38 & 11.81 & 12.54 + darkest & 16.52 & 14.98 & 12.82 & 12.34 & 13.04 + + median & 16.04 & 14.04 & 12.20 & @xmath34 & @xmath34 + brightest & 15.45 & 13.45 & 11.93 & @xmath34 & @xmath34 + darkest & 16.57 & 14.80 & 12.70 & @xmath34 & @xmath34 + + median & 15.64 & 13.76 & 12.28 & @xmath34 & @xmath34 + darkest & 16.96 & 14.98 & 13.55 & @xmath34 & @xmath34 + distribution of the nir surface sky brightness in different bands , derived using the data from the alhambra survey , along the universal time , for the 75 nights covered by our dataset , expanding from august 2004 to march 2008 . the plot shows 1954 , 2394 and 2963 individual points for the @xmath0 , @xmath1 and @xmath2 bands , respectively . blue crosses represent the values for the @xmath0-band , green stars represent the values for the @xmath1-band , and gray triangles represent the values for the @xmath2-band . the range of night sky surface brightnesses for each band are clearly appreciated . , title="fig : " ] distribution of the nir surface sky brightness in different bands along the airmass for the night of the 31st of january 2008 . the symbols for the different bands are described in the plot . there is clear dependency of the surface brightness with the airmass , particularly strong in the j and h bands . this dependency can be reproduced assuming an increase of the background flux with the airmass as it can be seen with the corresponding overplotted lines . a similar distribution is appreciated in all nights from our dataset . , title="fig : " ] distribution of the airmass corrected nir surface sky brightness for the j , h and ks bands , averaged for each night , along the time . the symbols for each band are similar to those of the figure [ mag_air ] . there is a clear seasonal trend in the ks band , being the sky brighter in summer than in winter . , title="fig : " ] distribution of the airmass corrected surface sky brightness for the @xmath0 , @xmath1 and @xmath2 bands , averaged for each night , along the average night temperature . only for the @xmath2-band a clear correlation between both parameters the dashed line represent the best linear regression fit over the @xmath2-band data . the encircle data indicate the data taken with the 1.23 m telescope.,title="fig : " ] distribution of the telescope seeing at the v - band derived from the analysis of the 7311 nir individual images from the alhambra survey ( red solid line ) corresponding to the time period between august 2004 and march 2008 , compared with a similar distribution of the atmospheric seeing , derived from the dimm monitor in the same time period ( black dashed line ) . the orange dash - dotted line shows the distribution of the original seeing measured in the @xmath2-band images . , title="fig : " ] distribution of the average relative humidity per night along the time period between january 1998 and december 2007 . there is a clear seasonal pattern , with the relative humidity being more frequently higher and more instable in winter than in summer . the driest periods are also shown in winter , associated with freezing temperatures . , title="fig : " ]

we present the characterization of additional properties of the night - sky at the calar alto observatory , following the study started in snchez et al . ( 2007 ) , hereafter paper i. we focus here on the night sky - brightness at the near - infrared , the telescope seeing , and the fraction of useful time at the observatory . for this study we have collected a large dataset comprising 7311 near - infrared images taken regularly along the last four years for the alhambra survey ( @xmath0 , @xmath1 and @xmath2-bands ) , together with a more reduced dataset of additional near - infrared images taken for the current study . in addition we collected the information derived by the meteorological station at the observatory during the last 10 years , together with the results from the cloud sensor for the last @xmath32 years . we analyze the dependency of the near - infrared night sky - brightness with the airmass and the seasons , studying its origins and proposing a zenithal correction . a strong correlation is found between the night sky - brightness in the @xmath2-band and the air temperature , with a gradient of @xmath40.08 mag per 1 @xmath5 c. the typical ( darkest ) night sky - brightness in the @xmath0 , @xmath1 and @xmath2-band are 15.95 mag ( 16.95 mag ) , 13.99 mag ( 14.98 mag ) and 12.39 mag ( 13.55 mag ) , respectively . these values have been derived for the first time for this observatory , showing that calar alto is as dark in the near - infrared as most of the other astronomical astronomical sites in the world that we could compare with . only mauna kea is clearly darker in the @xmath2-band , but not only compared to calar alto but to any other observatory in the world . the typical telescope seeing and its distribution was derived on the basis of the fwhm of the stars detected in the considered near - infrared images . this value , @xmath31.0@xmath6 when converted to the v - band , is only slightly larger than the atmospheric seeing measured at the same time by the seeing monitor , @xmath30.9@xmath6 . therefore , the effects different than the atmosphere produce a reduced degradation on the telescope seeing , of the order of @xmath310% . finally we estimate the fraction of useful time based on the relative humidity , gust wind speed and presence of clouds . this fraction , @xmath372% , is very similar to the one derived in paper i , based on the fraction of time when the extinction monitor is working .

the superconductivity of 60-derived solids owes its origin in part to features of the vibrational spectrum of 60 . consequently , a microscopic understanding of superconductivity requires an understanding of the vibrational modes . while there have previously been detailed computational studies @xcite yielding vibrational frequencies in excellent agreement with experiment , additional insight into the nature of modes can be obtained by comparing the vibrational spectrum of quasi - spherical 60 with that of its continuum analog : a thin , elastic , spherical shell . the electronic structure of 60 has been previously described in a similiar fashion , where 60 is thought of roughly as a spherical system with continuous rotational symmetry , invariant under the group @xmath7 . this leads to electronic multiplets , labeled by @xmath8 with degeneracy @xmath9 . the true molecule has only discrete icosahedral ( @xmath10 ) symmetry ; and , as is commonly done in crystal field theory , a correspondence can be made between the multiplets under the approximate high symmetry group and their descendants under the lower symmetry . this correspondence between orbitals under @xmath7 and orbitals under @xmath10 was made by haddon @xcite the @xmath7 multiplets order with energy increasing with increasing @xmath8 , implying that the energy scale associated with changes in radial quantum number , @xmath11 , is much larger than the rotational energy scale , @xmath12 , where @xmath13 is the length scale of the spherical potential and @xmath14 is the electron mass . the most energetic of the 60 @xmath15-electrons therefore occupy the @xmath16 multiplet . under the lower symmetry @xmath10 , multiplets of the spherical model for @xmath17 are split under @xmath10 , as icosahedral symmetry can not support an orbital degeneracy larger than 5 . for the occupied orbitals , these gaps are smaller than the spacing between adjacent @xmath8 multiplets ; consequently , the correlation between @xmath8 multiplets and nearly degenerate manifolds in 60 is clear . we expect the comparison of vibrations of an elastic shell to that of the 60 molecule to give qualitative insight , but we certainly do not expect close quantitative agreement . the continuum elastic approximation is only justified for the lowest frequency modes such that @xmath18 , where @xmath19 is a typical nearest - neighbor distance , and @xmath20 is the transverse speed of sound on the shell . furthermore , the linear elastic approximation assumes only local interactions . it is known that for force constant models to get good agreement with the experimentally measured vibrations of 60 , it is necessary to include up to third - nearest neighbor interactions.@xcite hence , a linear elastic model would not expect to yield good quantitative agreement with experiment over the entire spectrum . nevertheless , to deepen our understanding of the character of the complex modes of 60 , we find it useful to compare the molecular modes with those of an elastic spherical shell , where the eigenmodes and eigenfrequencies can be described by compact analytic expressions . we summarize the results obtained by lamb @xcite concerning the vibrations of a thin spherical shell . the modes can be divided into two classes . modes of the first class have displacements which are purely tangential to the surface of the shell . we denote the @xmath8th of these shear modes by @xmath21 . the frequency spectrum of shear modes is given by @xmath22 where @xmath23 is the dimensionless frequency , @xmath24 @xmath13 is the radius of the shell , and @xmath20 is the transverse speed of sound on the shell . we note that @xmath23 monotonically increases with increasing @xmath8 , and the lowest frequency shear mode has @xmath25 . the modes of the second class have displacements which have radial character . of these mixed modes , the @xmath26 mode is purely radial . its dimensionless frequency is given by @xmath27 where @xmath28 is related to the poisson ratio of the shell @xmath29 , @xmath30 the poisson ratio being the dimensionless ratio of transverse compression to longitudinal extension . the mixed modes with @xmath31 have displacements with both radial and tangential character . the dimensionless frequency of the @xmath32 mode is @xmath33 in the case of mixed modes with @xmath34 , each frequency has two branches whose frequencies are @xmath35^{1\over 2 } \label{m-}\ ] ] where @xmath36 . we denote the mixed modes by @xmath37 . we plot selected shell frequencies as a function of the poisson ratio in figs.[hy2 ] and [ hy01 ] . we establish correlations between the modes of the shell and the modes of 60 . the breathing mode of 60 ( @xmath38 ) can be unambiguously matched with its counterpart mode of the shell , the mixed mode @xmath39 . both modes have exclusively radial character and transform as the trivial representation of their respective symmetry groups . 60 has another mode preserving the icosahedral symmetry , but with pure tangential character @xcite . we note that the decomposition of @xmath40 representation under @xmath10 contains @xmath41 @xcite . we conclude that the @xmath42 mode is predominantly a descendant of the @xmath43 shell mode . the lowest frequency mode of the shell , @xmath44 , has a degeneracy of 5 . under @xmath10 , we expect this degeneracy to still be supported . experimentally , the lowest frequency mode of 60 , the so - called squashing mode @xmath45 , has mixed displacement character @xcite and the same degeneracy as the @xmath44 mode on the shell ; thus , we infer that they are correlated . the next lowest frequency mode of the shell , @xmath46 , has a degeneracy of 7 , which decomposes under @xmath10 into @xmath47 and @xmath48 modes . we correlate @xmath46 with the nearly degenerate multiplet @xmath49 and @xmath50 . similarly , the @xmath51 and @xmath52 levels must have evolved primarily from the splitting of the @xmath53 mode . as the @xmath54 shell mode must remain unsplit , we correlate it with a @xmath55 type mode under @xmath10 . there are two @xmath55 modes which are close in frequency . from stereographic projections @xcite of the modes , we conclude that @xmath56 is primarily a descendant of @xmath54 , while @xmath57 results from the splitting of the @xmath58 mode . for @xmath59 , mixed shell modes have small spacings compared to the splittings under @xmath10 . there are no real predominant `` progenitor '' levels , as mixing among many shell modes must be involved in determining the final character of the displacements . we summarize our low frequency correlations in fig . [ corr ] . the spectrum of the shell results from one structural parameter and two elastic parameters . we adjust the elastic parameters to fit to the low frequency portion of the spectrum . we use a value of 3.56 for @xmath13 , as obtained from experiment . we estimate the following parameter values from a fit to the @xmath38 and @xmath60 modes using eqs . [ m0 ] and [ shearf ] : @xmath61 and @xmath62 m / s . we choose to fit to these modes as their correlation with shell modes is clear , and they stem from multiplets whose degeneracy is preserved under @xmath10 . combining eqs . [ m0 ] and [ m1 ] , we observe that @xmath63 _ independent _ of elastic parameters . based on our correlation diagram , we can compare this ratio to experiment . we use the value of 493 @xmath64 for @xmath65 @xcite , and a value of 577 @xmath64 for @xmath66@xcite to obtain a ratio of @xmath67 . the reasonable agreement with @xmath68 gives support to the interpretation of these modes as progenitors of the @xmath69 and @xmath70 modes . a comparison of selected spherical model frequencies to experimental frequencies is given in table i. perhaps the real utility of the correlations made is that the known analytic forms of the eigenmodes might be used to approximate displacement fields in calculations . we thank f. anderson , w. leenstra , and j. wu for valuable discussions . acknowledgment is made to the donors of the petroleum research fund , administered by the american chemical society , for support of this research .

the vibrational spectrum of 60 is compared to the spectrum of a classical isotropic elastic spherical shell . we show correlations between the low frequency modes of 60 and those of the spherical shell . we find the spherical model gives the approximate frequency ordering for the low frequency modes . we estimate a poisson ratio of @xmath0 and a transverse speed of sound of @xmath1 m / s for the equivalent elastic shell . we also find that @xmath2 for the shell modes @xmath3 and @xmath4 , independent of elastic constants . we find that this ratio compares favorably with an experimental value of @xmath5 .

in the early days of equilibrium many - body simulation , intercomparisons of results from constant - temperature monte carlo with those from constant - energy molecular dynamics were indirect , requiring the conversion of isothermal @xmath0 data to isoenergetic @xmath1 data . only for hard disks and hard spheres , where temperature and energy are proportional , are the two ensembles identical@xcite . shuichi nos provided a conceptual breakthrough linking temperature @xmath2 and energy @xmath3 by discovering an isothermal canonical - ensemble molecular dynamics@xcite . he started out by including an additional time - reversible `` time - scaling variable '' @xmath4 in his novel hamiltonian : @xmath5 + \phi(q ) + ( p_s^2/2 m ) + ( \#kt/2)\ln(s^2 ) \ .\ ] ] here @xmath6 is the number of degrees of freedom , _ including _ @xmath4 , and @xmath7 is the momentum conjugate to @xmath4 . @xmath8 is a free parameter . getting to the canonical ensemble from nos s hamiltonian equations of motion requires just two more steps ; [ i ] `` scaling the time '' , multiplying all of nos s time derivatives by @xmath4 : @xmath9 then [ ii ] replacing all the scaled momenta @xmath10 by @xmath11 . nos showed that the resulting distribution for the @xmath12 is gibbs canonical distribution . starting instead with the `` nos - hoover '' equations of motion@xcite , @xmath13/\tau^2 \ , \ ] ] hoover showed that gibbs canonical distribution , with @xmath4 absent and with @xmath14 _ not _ including that extraneous @xmath4 variable , is a stationary solution of the extended phase - space continuity equation , @xmath15 here @xmath16 is a `` friction coefficient '' proportional to nos s @xmath7 and the free parameter @xmath17 takes the place of nos s parameter @xmath8 . the stationary distribution function for the nos - hoover motion equations is canonical in the @xmath12 and gaussian in the friction coefficient @xmath16 : @xmath18 the relaxation time @xmath17 controls the decay rate of velocity fluctuations in a @xmath19-dimensional @xmath20-body hamiltonian system . in order for averages using the nos - hoover equations to agree with canonical - ensemble averages , it is necessary in principle that the dynamics be `` ergodic '' , meaning that it must reach all of the @xmath12 phase - space states . of course in practice only a representative sampling of such states can be achieved . whether or not the motion equations are `` ergodic '' in this sense , _ capable _ of reaching all of the states described by gibbs canonical ensemble , depends upon the details of the underlying hamiltonian @xmath21 . for a simple harmonic oscillator hoover pointed out that neither the four original unscaled nos equations nor the three nos - hoover motion equations are ergodic : @xmath22 \ ; \ [ \ { \rm nos\acute{e } , \ not \ ergodic } \ ] \ .\ ] ] @xmath23/\tau^2 \ ; \ [ \ { \rm nos\acute{e}-hoover , \ not \ ergodic } \ ] \ .\ ] ] these nos equations are singular while the nos - hoover equations are not . with the relaxation time @xmath17 equal to unity , numerical nos - hoover calculations show that about five percent of the initial conditions drawn from the gaussian distribution @xmath24 are chaotic , making up a `` chaotic sea '' in the @xmath25 space which is penetrated by an infinite number of holes . the remaining 95% of initial conditions generate regular periodic toroidal solutions which fill in these holes . evidently this nos - hoover thermostated oscillator is far from `` ergodic '' [ where in this case an ergodic solution would have a smooth analytic gaussian density throughout @xmath25 space ] : @xmath26 \ .\ ] ] in 1990 bulgac and kusnezov reiterated the usefulness of the phase - space continuity equation in formulating more complicated motion equations ( with two or three thermostating control variables ) . shortly thereafter they were able to use this approach to fill out the complete gibbs distribution for prototypical hamiltonian systems like the harmonic oscillator and the two - well `` mexican hat '' problems@xcite . there is no _ foolproof _ test for ergodicity . the only reliable diagnostic for space - filling ergodicity is the lyapunov spectrum@xcite . if this spectrum , which measures the long - time - averaged sensitivity of the dynamics to initial conditions , is not only chaotic , but also the _ same _ for _ all _ initial conditions , the system is likely ergodic . for the harmonic oscillator this means that _ all _ @xmath27 oscillator states , all the way to infinity , will _ eventually _ occur . in nonergodic systems the long - time - averaged spectrum depends upon the initial conditions . in ergodic systems the long - time spectrum is always the same . by 1996 there were three known simple types of differential equations [ hh = hoover - holian , jb = ju - bulgac , mkt = martyna - klein - tuckermann ] which were thought to produce ergodicity in a one - dimensional harmonic oscillator@xcite . in addition to propagating the phase variables ( @xmath28 ) all three included _ two _ additional control variables ( @xmath29 ) capable of generating gibbs canonical distribution . for the simple harmonic oscillator these three models take the form : @xmath30 \ ; \ ] ] @xmath31 \ ; \ ] ] @xmath32 \ .\ ] ] so far no one - thermostat oscillator system ( with just a single control variable ) has been shown to be ergodic though it certainly is possible that such a one will be discovered . we explicitly include the temperature @xmath2 in all of these models in order to set the stage for _ nonequilibrium _ systems incorporating both `` cold '' and `` hot '' degrees of freedom . it is worth pointing out that not all two - thermostat oscillator systems are ergodic . patra and bhattacharya showed that their very reasonable set of doubly - thermostated oscillator equations is not ergodic@xcite : @xmath33 \ .\ ] ] our work here has three different aspects . first we explore the equilibrium lyapunov instability which facilitates the ergodicity of all three thermostats . in the equilibrium case , where the phase - space distribution is a smooth gaussian , the lyapunov instability embedded in that gaussian has a totally different multifractal character . we show that the equilibrium hh , jb , and mkt thermostats obey the zeroth law when coupled with one another or with other hamiltonian systems . these thermostats can provide gibbs complete canonical distribution provided that any internal energy barriers are not too large relative to @xmath34 . second we consider _ nonequilibrium _ cases , where heat flows between thermostats with the thermostat temperatures set at different values . the nonequilibrium dynamics is still ergodic but the phase - space distribution is no longer a smooth gaussian . it is instead an intimate multifractal combination of the attractor - repellor pairs common to macroscopic time - reversible , but dissipative , systems . we also illustrate the application of ergodic thermostats to kinetic barrier - crossing problems . last of all , we summarize the lessons learned in this study of three different sorts of applications of our ergodic time - reversible thermostats . the most convincing evidence for the lack of ergodicity with one thermostat variable , and its lack or presence in two - thermostat systems , is the lyapunov spectrum . here we concentrate on thermostated oscillators , the prototypical `` difficult '' case . a harmonic - oscillator system of four ordinary differential equations , such as the hh , jb , mkt , and pb sets for @xmath35 has four such exponents @xmath36 . these exponents describe the deformation of an infinitesimal hypersphere or hypercube in the four - dimensional phase space which contains the motion . shimada and nakashima@xcite , as well as benettin s group@xcite described general approaches to determining the spectrum of exponents . plane . because a mirror reflection of the dynamics ( perpendicular to the @xmath37 axis ) changes the signs of both @xmath37 and @xmath38 the figure reveals an inversion symmetry . in the upper half plane red indicates a positive exponent and green a negative . the colors are reversed in the lower half plane . the probability density in the cross - sectional plane is a simple gaussian in @xmath37 and @xmath38 . nevertheless the local lyapunov exponents vary in a multifractal manner throughout the four - dimensional space , reflecting their sensitivity to bifurcations in their past histories . in figures 1 - 4 both @xmath37 and @xmath38 range from -3 to + 3 . ] in an @xmath39-dimensional space their algorithms require the solution of @xmath40 sets of @xmath39 equations . in addition to a `` reference trajectory '' the equations describe the motion of an associated orthonormal set of @xmath39 comoving basis vectors , centered on the reference trajectory and locating @xmath39 nearby `` satellite trajectories '' in the @xmath39-dimensional space . the first lyapunov exponent gives the time - averaged rate at which two neighboring solutions of the equations diverge from one another , @xmath41 . the rate at which the _ area _ defined by three nearby solutions ( the reference and two others ) diverges ( or converges ) @xmath42 defines @xmath43 , while the rate at which the volume defined by four solutions and the hypervolume defined by all five of them define @xmath44 and @xmath45 . although these exponents are typically `` paired '' for hamiltonian systems , with @xmath46 \rightarrow \sum_1 ^ 4\lambda_i \equiv 0 \ , \ ] ] with the sum total equal to zero , as implied by liouville s theorem , none of the two - thermostat systems is hamiltonian so that this instantaneous symmetry is missing in the time - dependent exponents @xmath47 . but because the equations of motion are time - reversible the time - averaged spectra @xmath48 _ are _ symmetric . for an _ ergodic _ system the long - time - averaged spectrum of exponents should be _ independent _ of the initial conditions . in practice , for the harmonic - oscillator system , a few hundred oscillations is time enough to indicate whether or not a chosen initial condition ( either from a grid or chosen randomly ) converges to a spectrum close to the unique long - time - averaged lyapunov spectrum . estimates of these long - time - averaged exponents , after a _ time _ @xmath49 using a fourth - order runge - kutta integrator with an adaptive timestep , are as follows : @xmath50 in the hoover - holian case , and @xmath51 in the ju - bulgac case , and @xmath52 in the martyna - klein - tuckerman case . , @xmath38 , @xmath16 , and @xmath53 as it is also in the @xmath54 plane displayed here . ] the increase in the largest lyapunov exponent with the number of quartic forces ( none for mkt , one for hh , and two for jb ) suggests , as emphasized by bulgac and kusnezov@xcite , that these terms are particularly well - suited to promoting chaos and/or ergodicity . it seems likely that sextic forces , controlling @xmath55 , would have no particular advantages and would slow numerical work . the probability densities for the three cases follow easily from the phase - space continuity equation . this makes it is easy to check for ergodicity by comparing relatively short - time estimates for the lyapunov spectrum starting out with initial conditions chosen according to the stationary multivariable gaussian distributions . the ergodicity of the martyna - klein - tuckerman oscillator was called into question by patra and bhattacharya in 2014@xcite , based on apparent `` holes '' in a @xmath56 = @xmath57 double cross - section plane , the analog of a poincar plane for a three - dimensional flow problem . to resolve this question , the subject of the 2014 ian snook prize@xcite , we chose one million different initial conditions from the appropriate four - dimensional gaussian distributions for the hh , jb , and mkt cases and determined that _ all of them _ were fully consistent with a unique chaotic spectrum . each of these initial conditions was followed for one million fourth - order runge - kutta timesteps of 0.005 each . the alternative possibility , a regular ( nonchaotic ) solution with all four of the long - time - averaged lyapunov exponents equal to zero , is easy to distinguish from the chaotic case . as an additional check the velocity moments generated by these three ergodic thermostat types likewise reproduce gibbs values for the second , fourth , and sixth velocity moments with five - figure accuracy : @xmath58 by contrast , the motion equations for a nos - hoover harmonic oscillator with a single friction coefficient @xmath59 and unit relaxation time , @xmath60 are known _ not _ to be ergodic despite their simple three - dimensional gaussian solution of the phase - space continuity equation . choosing initial conditions from this stationary distribution @xmath61 and measuring the tendency of a nearby initial condition to separate gives the largest _ lyapunov exponent _ for the chosen initial condition , @xmath62 . because this system is conservative and time - reversible the long - time - averaged lyapunov spectrum is symmetric , @xmath63 , and sums to zero . @xmath64 initial conditions followed for a time @xmath65 divide neatly into two groups : @xmath66 557 initial conditions correspond to a lyapunov exponent significantly different to zero while 9443 correspond to nonchaotic toroidal solutions for which all three lyapunov exponents vanish . evidently roughly 6% of the stationary measure is chaotic ( with a positive time - averaged largest lyapunov exponent ) . the remaining 94% consists of regular tori with a vanishing lyapunov spectrum : @xmath67 . like the other models this martyna - klein - tuckerman oscillator has a four - dimensional gaussian distribution in its @xmath56 phase space which the dynamics samples ergodically . ] in the language of control theory the nos - hoover equations control the time - averaged value of the kinetic temperature , @xmath68 . subsequent work strongly suggests that an ergodic oscillator requires control of at least two moments , not just one . such an approach requires four or more ordinary differential equations . bulgac and kusnezov@xcite as well as ju and bulgac@xcite added a fifth equation so as to be able to thermostat a free particle ( or a rotating and translating cluster of particles ) undergoing brownian motion . . like the other models this patra - bhattacharya oscillator has a four - dimensional gaussian distribution in its @xmath56 phase space which the dynamics samples ergodically . unlike the other models about half the pb measure is regular rather than chaotic . ] four_-dimensional phase - space continuity equation shows that both the martyna - klein - tuckerman and the hoover - holian thermostats are consistent with exactly the same four - dimensional gaussian , @xmath69 . unlike the nos - hoover single - thermostat model we believe that the three two - thermostat models are _ all _ ergodic . we believe that this proposition has been thoroughly confirmed by the present work , confirming the chaos of one million independent initial conditions in each case . in view of the lack of other suitable entries in the 2014 competition , we have awarded ourselves the snook prize for this finding . it should be noted that _ all _ of these thermostat models , ergodic or not , _ are _ time - reversible , with the friction coefficients @xmath70 and the momentum @xmath38 simply changing sign if the time - ordered sequence of coordinate values @xmath71 is reversed . for this reason the time - averaged lyapunov spectrum is symmetric , with @xmath72 . despite the simple analytic nature of the phase - space distribution , the chaos implied by a positive @xmath73 engenders an amazing complexity , singular in its spatial behavior , to which we turn next . in exploring the hh , jb , and mkt ergodic oscillator solutions we uncovered an amazingly intricate multifractal structure most simply described in terms of the local largest lyapunov exponent , @xmath74 . the three phase - space distributions @xmath75 are all multidimensional gaussians , of the form , @xmath76 where @xmath39 is 2 for hh and mkt and 4 for jb . a closer look , using lyapunov instability as a tool , reveals the interesting structures shown for the three thermostated oscillators in * figures 1 - 3 * . for completeness , the corresponding patra - bhattacharya cross section is shown as * figure 4 * . for each of the three ergodic models we show the sign of the local lyapunov exponent , @xmath74 at the location @xmath54 . the plots are analogous to a poincar section for the more usual three - dimensional flows . at the equilibrium , @xmath77 , these instability plots can be generated in either of two ways : [ i ] whenever a @xmath54 trajectory comes close to @xmath78 plot the sign of the corresponding lyapunov exponent at the location @xmath27 ; [ ii ] starting at @xmath54 integrate _ backwards _ in time for several hundred thousand timesteps , storing the backward trajectory . then process the reversed trajectory data in the forward direction , with lyapunov instability controlling a nearby constrained satellite trajectory , finding the local lyapunov exponent at the chosen endpoint location @xmath54 . it turns out that these two methods are roughly comparable in cost . the second method has the advantage of providing more detail in regions where the measure is small . the precision shown in figures 1 - 4 can be generated in simulations taking a day or two on a single processor . method [ ii ] above , reversing a stored trajectory , is particularly well - suited to parallel simulations . the figures use color to indicate the sign of the local exponent . these cross sections have a fractal look . in fact they are . the reason for this was pointed out in our paper with dennis isbister@xcite . as the `` past '' history of a point is being generated , backward from the `` present '' , `` bifurcations '' , where the direction of the trajectory is uncertain at the level of the numerical work limit the trustworthy scale of the lyapunov exponent when the stored integration is reversed , going forward in time . for the megapixel details shown in figures 1 - 4 a few hundred thousand fourth - order ( or fifth - order or adaptive ) runge - kutta timesteps suffice for reliable results at the resolution of the plotted sections . more detail can always be generated , at a cost exponential in the number of significant figures , by reducing the numerical error of the trajectory integration . the cross sections shown in the figures have inversion symmetry . as the source of this symmetry is not so obvious let us describe it . consider a solution of the equations of motion going forward in time , @xmath79 . viewed simultaneously in a mirror perpendicular to the @xmath37 axis an observor sees @xmath80 . evidently there is an inversion symmetry , with exactly the _ same _ lyapunov exponent at two different values of @xmath27 : @xmath81 this observation saves a factor of two in computer time . the two - dimensional cross sections where both of the friction coefficients vanish , @xmath82 are the starting points for the backward integrations . the figures indicate the magnitude of the largest lyapunov exponent obtained at the original starting point from a reversed trajectory , with red positive and green negative as the local exponents varies with position . this is the case for positive @xmath37 . the colors are reversed for negative coordinate values in order to eliminate the discontinuity that would otherwise occur at vanishing @xmath37 . as is usual the fluctuations in @xmath74 are an order of magnitude larger than its long - time - averaged value . the variation of the local lyapunov exponent is not quite continuous . if the local exponent is plotted along a line in the @xmath27 plane there are occasional jumps , present on all scales , in the exponent value . the fractal dimension accounts for the variation of the jump magnitudes as a function of resolution . in the mkt case@xcite the data along such a line gave a fractal dimension of 1.69 rather than 1.00 . studies of the jumps along such lines reveal their fractal structure , seen here for the first time in two dimensions . all three of the two - thermostat hh , jb , and mkt models for heat - bath control of a harmonic oscillator are ergodic , because of this it was a complete surprise that patra and bhattacharya s control of both the kinetic and the configurational@xcite oscillator temperatures , using two thermostat variables , was unsuccessful , leading to a mix of regular and chaotic solutions . their model was : @xmath83 \ .\ ] ] the stationary phase - space distribution function for the pb equations is the same as that characterizing the hh and mkt models : @xmath84 but unlike the hh and mkt equations the patra - bhattacharya model _ fails _ the lyapunov - exponent test for ergodicity . about half the initial conditions chosen from the four - dimensional gaussian distribution give regular rather than chaotic solutions . although controlling _ coordinate _ moments like @xmath85 may seem unphysical landau and lifshitz showed ( in the 1951 russian edition of their _ statistical physics _ ) that what later came to be termed a `` configurational temperature''@xcite , based on a mean - squared force , can be derived from gibbs canonical distribution . a single integration by parts , in the canonical ensemble , relates the mean - squared force to the curvature of the potential through the temperature @xmath2 : @xmath86e^{-{\cal h}/kt}dq \equiv \int [ \ ( \nabla{\cal h})^2 \ ] e^{-{\cal h}/kt}dq \ .\ ] ] so control of @xmath87 can be viewed as regulating the fluctuations in the _ forces_. this concept is more appealing than coordinate control . but the lack of a physical `` thermometer '' capable of measuring force - based temperature is a disabling disadvantage of configurational temperature . the possibility of divergent or negative values of the instantaneous configurational temperature are further caution flags . the campisi `` thermostat '' is an example of both defects : @xmath88 formally , campisi dynamics is consistent with a gaussian momentum distribution . although at first glance appealing , this simple model exhibits a _ negative _ configurational temperature , @xmath89 , in one dimension and a divergent one in two@xcite . disagreement between the kinetic and configurational temperatures is a clear violation of the zeroth law of thermodynamics , to which we turn next . to what extent do these minimalistic thermostats , described by one or two additional control variables , exhibit the characteristic properties of macroscopic thermal baths ? an essential characteristic of such baths is their consistency with the zeroth law of thermodynamics . two `` good '' thermal baths , both at a temperature @xmath2 , should , when coupled together or to another canonical equilibrium system at the same temperature @xmath2 , remain at that common temperature without modifying gibbs equilibrium distribution in either bath or in another equilibrium system . for our thermostated oscillators this desirable property can be tested by coupling _ pairs _ of heat baths together with a simple hookes - law spring . the modification of the equations of motion for oscillator number 1 and oscillator number 2 , both at unit temperature , are as follows : @xmath90 it is easy to verify that the two nos - hoover oscillators , coupled together by a spring , have nonergodic solutions , corresponding to a six - dimensional analog of the tori making up most of the single - oscillator phase space . one such solution occurs with initial values of the two oscillator momenta @xmath91 . accordingly we restrict our detailed investigation to the three ergodic two - thermostat heat baths to see whether or not they can obey the zeroth law . all the six combinations of like and unlike _ pairs _ of ergodic two - thermostat oscillators : [ hh+hh , hh+jb , hh+mkt , jb+jb , jb+mkt , mkt+mkt ] , + provide chaotic eight - dimensional regions in phase space with symmetric lyapunov spectrum . this symmetry implies ergodicity . * figure 5 * shows both the spectra and the spectral sums ( which give the growth rates of phase - space hyperspheres as functions of their dimensionality ) . even though there are relatively small but significant differences between the hh , jb , and mkt lyapunov exponents pairing _ any two _ of these systems together ( with a hookes - law spring ) gives no indication of the dissipation which would result if one phase - space bath were to grow at the expense of the other . we conclude from these six examples that the hh , jb , and mkt heat baths _ all _ obey the zeroth law of thermodynamics , quite a good outcome for dynamical systems which represent a thermodynamic heat bath with the low cost of only four phase - space dimensions . the largest lyapunov exponent in these eight - dimensional problems varies from 0.20 for two coupled mkt oscillators to 0.23 for two coupled jb oscillators . the exponent for two hh oscillators lies in between . evidently the quartic forces in the ju - bulgac and hoover - holian thermostats are slightly more effective in promoting chaos than are the cubic martyna - klein - tuckerman forces . this same conclusion follows from the time required to generate the structures seen in figures 1 - 4 . the additional chaos comes at the ( rather small ) price of an increased stiffness in the differential equations . if any of the three ergodic thermostats were not ergodic we would expect to see a dissipative strange attractor result in coupling it to an ergodic heat bath . in all six pairings of the equilibrium ergodic thermostats there is no evidence of dissipation or reduced phase - space dimensionality . evidently all three @xmath70 thermostats are good heat baths from the standpoint of the zeroth law of thermodynamics . coupling pairs of thermostated oscillators with a harmonic coupling reveals that all three of the ergodic models @xmath92 hh , jb , mkt @xmath93 remain ergodic when coupled and also provide the complete canonical distribution . the similar equilibrium behavior of the hh , jb , and mkt thermostats suggests a further comparison for simple heat - flow problems in which two coupled thermostats have _ different _ temperatures , leading to hot - to - cold heat flow and dissipation . dissipation is reflected in phase space by the formation of strange attractors with a fractional dimensionality reduced below the equilibrium value . in the linear - response regime the six two - bath possibilities must necessarily agree with green and kubo s theory as all the two - constant heat baths reproduce gibbs canonical @xmath27 distribution at equilibrium . let us be adventurous and treat instead the relatively far - from - equilibrium coupling of two thermostats with thermostat 1 at half the temperature of thermostat 2 : @xmath94 just as before all the masses , force constants , and relaxation times are chosen equal to unity . the far - from - equilibrium lyapunov data corresponding to the nine combinations of thermostats , with temperatures of 1 and 2 , are shown in * figure 6 * , both individually and as sums . the broad maximum corresponding to sums of three or four exponents indicates that the maximum growth rate in phase space applies to four - dimensional volumes . the summed - up spectrum crosses zero near an abscissa value of 7 indicating that the dimensionality loss for this far - from - equilibrium problem is on the order of 1 . considerably larger losses , comparable to the number of particles , occur in `` @xmath95-model '' chain simulations where each of a few dozen particles is tethered to its lattice site with a cubic restoring force@xcite . closer to equilibrium , with cold and hot temperature of 1.0 and 1.1 , all nine thermostat combinations have a lyapunov sum of order -0.01 while temperatures of 1.0 and 2.0 provide dissipative sums in the range from -0.20 to -0.30 . compare * figures 6 and 7*. just as in heat conduction according to fourier s law , we would expect a linear - response dissipation _ quadratic _ in the temperature difference driving the flow , varying as @xmath96 . loss of phase - space volume corresponds to thermodynamic dissipation ( through gibbs relationship of phase volume to entropy , @xmath97 , where @xmath98 is phase volume or number of states ) . coupling two baths , with the cooler bath first , gives the following `` entropy production rates '' : @xmath99 @xmath100 @xmath101 we used quote marks above to remind the reader that simply summing the spectrum is not the same as averaging the dissipation for the cold and hot thermostats . dissipation can also be reckoned in terms of the phase - space dimensionality of the dissipative strange attractor , using the kaplan - yorke linear interpolation between the last positive dimension and the first negative one . for the nine heat - bath pairings the dimensionalities are : @xmath102 @xmath103 @xmath104 these results should provide good benchmarks for nonlinear theories of transport . the double - well `` mexican hat '' potential@xcite , @xmath105 + \left [ \ \frac{q^4}{4 } \ \right ] \longrightarrow \phi(\pm 1 ) = - \left [ \ \frac{1}{4 } \ \right ] \ ; \ \phi(0 ) = 0 \ , \ ] ] has a barrier between its two minima of height @xmath106 . at a temperature of unity , where the barrier is negligible , simulations of the hat potential coupled to the hh , mkt , and jb thermostats result in excellent agreement with gibbs canonical distribution . at sufficiently low temperatures we expect that the frequency of successful jumps over the barrier will decline , as would also be the case using the metropolis , rosenbluths , and tellers monte carlo scheme with a reasonable jump length , perhaps of order 0.01 . a straightforward `` weak coupling '' of the two - minimum hat potential to any one of the ergodic oscillator heat baths , @xmath107 leads to `` jump '' frequencies that vary roughly as gibbs canonical barrier weight , @xmath108 . eyring s model , as mentioned in the wikipedia article on `` henry eyring ( chemist ) '' , and with a jump frequency varying as @xmath109 , is a relatively poor description and corresponds to the upper curves in the * figure 8 * . all three ergodic thermostat models behave similarly , as is shown in triplets of narrower lines in * figure 8 * . another possibility for modelling jumps is to use the metropolis , rosenbluths , and tellers monte carlo method . by adusting the monte carlo trial steplength the two approaches , dynamical and monte carlo , can be made to correspond . we have compared the number of jumps over the barrier in billion - jump simulations with maximum jump length @xmath110 to dynamical simulations with timestep 0.005 , where in all these latter dynamical cases the hat potential is coupled to an hh , mkt , or jb thermostat with temperatures varying from 0.010 to 1.00 , as shown in figure 8 . the lyapunov spectra for three varieties of time - reversible deterministic oscillator thermostats show that all three of them reproduce an extended ( four - variable ) form of gibbs canonical ( two - variable ) distribution , @xmath111 the hh and mkt equations have gaussian distributions for @xmath16 and @xmath112 while the jb distribution is gaussian in @xmath113 and @xmath112 . all three approaches are sufficiently robust to thermostat a harmonic oscillator . the three are also consistent with the zeroth law of thermodynamics in the sense that any pair of them at a common temperature @xmath2 generates gibbs canonical distribution for that temperature for both thermostats . this set of ergodic thermostats represents a good toolkit for simulations with a few degrees of freedom as it gives an idea of the dependence of dissipation on the chosen form of the thermostat . with many degrees of freedom , where ergodicity is not an issue [ poincar recurrence takes forever so that the size of fluctuations determines the usefulness of the results ] any of these thermostats , as well as the single thermostat nos - hoover model , are likely equally useful . we have considered two _ nonergodic _ time - reversible thermostats , the patra - bhattacharya and nos - hoover , which reproduce the specified kinetic temperature but do not reproduce all of gibbs canonical distribution . when the pb or nh thermostats are coupled to the hh , jb , or mkt thermostats two qualitatively different results occur : [ 1 ] a dissipative strange attractor in the case of the pb model and [ 2 ] a conservative integer - dimensional chaotic sea when coupled to the hamiltonian - based nos - hoover oscillator . these results are similar for each of the three ergodic thermostats . in the pb case the dissipation is an indicator of the nonhamiltonian nature of the dynamics . any of the ergodic thermostats , hoover - holian , ju - bulgac , martyna - klein - tuckerman , can be used successfully in both equilibrium and nonequilibrium simulations . the first two of these can also be applied in nonequilibrium simulations where it is desired to specify the kinetic temperature of selected degrees of freedom . though there is no problem in solving the mkt equations for nonequilibrium problems brad holian pointed out long ago@xcite that this thermostat fails to return its target temperature due to the nonzero correlation linking the two mkt friction coefficients , @xmath114 . this relatively thorough investigation of ergodicity , based on one million independent initial conditions , should settle ( at least from the numerical , as opposed to analytical , viewpoint ) the question of the ergodicity of the martyna - klein - tuckerman oscillator . the apparent `` holes '' in that oscillator s cross section are still a small `` puzzle '' . a second such `` puzzle '' is the nature of the local lyapunov exponents . their intricate multifractal lyapunov structure is well concealed within an innocent gaussian distribution . from the standpoint of dynamical - systems theory ( as opposed to thermodynamics ) we wish to emphasize the amazing nature of the chaos hidden within the simple ergodic gaussian distributions . the smoothness of the distributions and the good convergence of their moments conceals the fractal nature of their chaos , as shown in figure 1 through 3 . we recommend all three systems for further studies . the @xmath115 plane is only a single choice among the many possible . despite their fractal nature the cross - sectional values are well - behaved , accessible , and reproducible from the numerical standpoint . it is encouraging for the future that work originally chosen to settle some ergodicity questions turned out to generate a swarm of other problems still needing more detailed understanding . we thank carol hoover , aurel bulgac , and baidurya bhattacharya , for their able help . g. benettin , l. galgani , a. giorgilli , and j. m. strelcyn , `` lyapunov characteristic exponents for smooth dynamical systems and for hamiltonian systems ; a method for computing all of them . part 1 : theory '' , meccanica * 15 * , 9 - 20 ( 1980 ) .

the relative stability and ergodicity of deterministic time - reversible thermostats , both singly and in coupled _ pairs _ , are assessed through their lyapunov spectra . five types of thermostat are coupled to one another through a single hookes - law harmonic spring . the resulting dynamics shows that three specific thermostat types , hoover - holian , ju - bulgac , and martyna - klein - tuckerman , have very similar lyapunov spectra in their equilibrium four - dimensional phase spaces and when coupled in equilibrium or nonequilibrium pairs . all three of these oscillator - based thermostats are shown to be _ ergodic _ , with smooth analytic gaussian distributions in their extended phase spaces ( coordinate , momentum , and two control variables ) . evidently these three ergodic and time - reversible thermostat types are particularly useful as statistical - mechanical thermometers and thermostats . each of them generates gibbs universal canonical distribution internally as well as for systems to which they are coupled . thus they obey the zeroth law of thermodynamics , as a good heat bath should . they also provide dissipative heat flow with relatively small nonlinearity when two or more such bath temperatures interact and provide useful deterministic replacements for the stochastic langevin equation .

the hypothesis of limiting fragmentation in high energy hadron - hadron collisions was suggested nearly four decades ago @xcite . this hypothesis states that the produced particles , in the rest frame of one of the colliding hadrons , will approach a limiting distribution . these universal distributions describe the momentum distributions of the fragments of the other hadron . central to the original hypothesis of the limiting fragmentation in ref . @xcite was the assumption that the total hadronic cross sections would become constant at large center - of - mass energy . if this occurred , the excitation and break - up of a hadron would be independent of the center - of - mass energy and distributions in the fragmentation region would approach a limiting curve . we now know that the total hadronic cross - sections are not constant at high energies . instead , to the highest energies achieved , they grow slowly with the center - of - mass energy @xmath0 , with favored functional forms being either a power law behavior @xmath1 with @xmath2 @xcite or a @xmath3 @xcite . a measurement of the total cross - section at the lhc @xcite will further help constrain the possible functional forms . even though the cross - sections are not constant , limiting fragmentation appears to have a wide regime of validity . it was confirmed experimentally in @xmath4 and nucleus - nucleus collisions at high energies @xcite . more recently , the brahms and phobos experiments at the relativistic heavy ion collider ( rhic ) at brookhaven national laboratory ( bnl ) have performed detailed studies of the pseudo - rapidity distribution of the produced charged particles @xmath5 for a wide range ( @xmath6 ) of pseudo - rapidities , and for several center - of - mass energies ( @xmath7 ) in nucleus - nucleus ( au - au and cu - cu ) and deuteron - nucleus ( d - au ) collisions . results for pseudo - rapidity distributions have also been obtained over a limited kinematic range in pseudo - rapidity by the star experiment at rhic @xcite . these measurements in a - a and d - a collisions were performed for several centralities . in addition to the d - au and a - a data , there are @xmath8 data at @xmath9 and also at @xmath10 gev . ( in the near future , data at @xmath11 gev may become available . ) these measurements have opened a new and precise window on many scaling phenomena glimpsed at lower energies . in particular , they have performed detailed studies of the limiting fragmentation phenomenon . the pseudo - rapidity distribution @xmath12 ( where @xmath13 is the pseudo - rapidity shifted by the beam rapidity @xmath14 ) is observed to become independent of the center - of - mass energy @xmath0 in the region around @xmath15 @xmath16 where @xmath17 is the impact parameter . it is worth noting that this scaling is in a strong disagreement with boost invariant scenarios which predict a fixed fragmentation region and a broad central plateau growing with energy . it would therefore be desirable to understand the nature of hadronic interactions that lead to limiting fragmentation and the deviations away from it . in a recent article , biaas and jeabek @xcite , argued that some qualitative features of limiting fragmentation can be explained in a two - step model involving multiple gluon exchange between partons of the colliding hadrons and the subsequent radiation of hadronic clusters by the interacting hadrons . in this paper , we compute limiting fragmentation within the color glass condensate ( cgc ) approach @xcite to high energy hadronic collisions its relation to the biaas - jeabek model will be addressed briefly later . in the cgc formalism , gluon production in the limiting fragmentation region can be described , at leading order , in the framework of @xmath18-factorization . under this assumption , the inclusive cross - section for produced gluons is expressed as the convolution of the unintegrated parton distributions " in the projectile and target respectively which is probed at large @xmath19 , and `` target '' the nucleus @xmath20 which is probed at small values of @xmath21 . of course , our choice of semantics , reminiscent of what is used in fixed target experiments , is somewhat arbitrary in collider experiments where the lab frame and the center of mass frame are identical . ] , @xmath22 times a ( transverse momentum dependent ) vertex squared . as we shall emphasize later , the name unintegrated parton distributions " though conventional is somewhat imprecise because the objects @xmath23 that enter in this formula differ from the expectation value of the gluon number operator . in eq . ( [ eq:2phis ] ) , @xmath19 and @xmath21 are the longitudinal momentum fractions of the gluons probed in the projectile and target , namely , @xmath24 where @xmath25 is the beam rapidity , @xmath26 is the nucleon mass , and @xmath27 is the transverse momentum of the produced gluon . we should emphasize that the kinematics here is the @xmath28 eikonal kinematics , which provides the leading contribution to gluon production in the cgc picture . as we will discuss further later , @xmath18-factorization actually works best in the limiting fragmentation kinematics where @xmath29 and @xmath30 ( or vice versa ) . clearly , @xmath18-factorization tells us that the cross - sections depend in general on both @xmath31 and @xmath32 . however , target limiting fragmentation implies a dependence on solely @xmath33 of the spectrum of produced particles integrated over the transverse momentum . as we shall see in section [ sec:2 ] , such a scaling emerges in a straightforward manner in the @xmath18-factorization framework if a ) the typical transverse momentum in the projectile is much smaller than the typical transverse momentum in the target , and b ) if unitarity is preserved in the evolution of the target with @xmath21 . these two conditions are naturally fulfilled in the cgc picture when @xmath21 becomes so small that the target reaches the `` black disk '' limit . more importantly , by studying how this limit is reached , one gets some insight into systematic deviations away from the limiting fragmentation . the dynamical evolution of the unintegrated distributions with @xmath34 are described in the cgc by the renormalization group equations called the jimwlk equations @xcite . these equations , more generally , describe the @xmath34 evolution of @xmath35-point parton correlation functions . the dynamics in the evolution at high parton densities is characterized by a saturation momentum " @xmath36 . this scale is the typical transverse momentum of partons in the hadron or nuclear wave - function is also related to the density of color charges per unit of transverse area in the hadron under consideration . ] . partons with @xmath37 saturate in the wave - function with occupation numbers of order @xmath38 . the implications of saturated or black disc " distributions for limiting fragmentation were studied previously in the cgc approach by jalilian - marian @xcite and in the work of kharzeev and levin @xcite and later by kharzeev , levin and nardi @xcite . in @xcite , it was _ a priori _ assumed that the target is very dense and appears black to the dilute partons in the projectile . this leads to the simple formula @xmath39 \ ; , \ ] ] for the rapidity distribution of the produced hadrons in aa collisions , where @xmath40 and @xmath41 are the quark and gluon distributions in the projectile nucleus . with a suitable choice of a scale @xmath42 in these distributions , this prescription gave a reasonable description of the fragmentation region . in @xcite , the @xmath18-factorization formalism was employed , together with a saturation " ansatz for the unintegrated gluon distribution . our approach to computing the inclusive gluon production in the saturation scenario is very similar in spirit to the one pioneered in @xcite . here however , the unintegrated gluon distributions at small @xmath34 ( @xmath43 ) are computed using a mean field version of the jimwlk equations called the balitsky - kovchegov ( bk ) equation @xcite . this approximation is strictly valid in the large @xmath44 , large mass number @xmath45 and high energy limit . however , the bk equation may have a wider range of applicability beyond these asymptotic limits . remarkably , it has been shown recently that the bk equation lies in the same universality class as the fisher - kolmogorov - petrovsky - piscounov equation in statistical mechanics , which describe a wide range of reaction - diffusion phenomena in nature @xcite . we will discuss results for limiting fragmentation from solutions of the bk equation for unintegrated distributions in both the fixed and running coupling cases . for @xmath46 , the unintegrated distribution , computed using the bk equation , is smoothly matched on to a functional form that contains key features expected of large @xmath34 parton distributions . we will also discuss what more detailed comparisons to the data can teach us about these distributions . this article is organized as follows . in section 2 , we discuss inclusive gluon production in high energy hadron and nuclear collisions . in the kinematical region of interest , @xmath18-factorization is applicable ; this allows us to relate the distribution of produced gluons to the unintegrated gluon distributions in the hadron wave - functions . parton - hadron duality @xcite is assumed to compare the distribution of produced gluons to those of hadrons . ( the effect of fragmentation functions can be significant at higher energies a qualitative discussion is presented in section 4 . ) the evolution of the unintegrated gluon distributions ( with @xmath19 and @xmath21 respectively ) in the projectile and target wave - functions is described by solutions of the bk equation . the fixed and running coupling forms of the bk equation and its solution are discussed in section 3 . results for gluon distributions obtained from numerical solutions of the bk equation are compared to data on limiting fragmentation in section 4 . we compare our results to data from pp , deuteron - gold and aa collisions and discuss their dependence on the initial conditions , running coupling effects and on infrared regulators . our results are summarized and open problems emphasized in the concluding section . at very high energies , for @xmath47 , gluon production and fragmentation is the dominant mechanism for particle production . when the occupation number of partons is small in one of the colliding hadrons and large in the other ( as is the case in proton - nucleus or nucleus - nucleus collisions in the fragmentation region of one of the nuclei ) , the inclusive multiplicity distribution of produced gluons can be expressed in the @xmath18-factorized form @xcite , @xmath48 strictly speaking , the formula , as written here , is only valid at zero impact parameter and assumes that the nuclei have a uniform density in the transverse plane . indeed , the functions @xmath23 are defined for the entire nucleus . here , @xmath49 denotes the transverse area of the overlap region between the two nuclei , while @xmath50 are the total transverse area of the nuclei , and @xmath51 is the casimir in the fundamental representation . the longitudinal momentum fractions @xmath19 and @xmath21 were defined previously in eq . ( [ eq : kinematics ] ) . the functions @xmath52 and @xmath53 are obtained from the dipole - nucleus cross - sections for nuclei @xmath45 and @xmath20 respectively , @xmath54 where @xmath55 and where the matrices @xmath56 are adjoint wilson lines evaluated in the classical color field created by a given partonic configuration of the nuclei @xmath45 or @xmath20 in the infinite momentum frame . for a nucleus moving in the @xmath57 direction , they are defined to be @xmath58 \ ; . \label{eq : pathglue}\end{aligned}\ ] ] here the @xmath59 are the generators of the adjoint representation of @xmath60 and @xmath61 denotes the `` time ordering '' along the @xmath62 axis . @xmath63 is a certain configuration of the density of color charges in the nucleus under consideration , and the expectation value @xmath64 corresponds to the average over these color sources @xmath65 . in the mclerran - venugopalan ( mv ) model @xcite , where no quantum evolution effects are included , the @xmath66 s have a gaussian distribution , with a 2-point correlator given by @xmath67 , where @xmath68 . this determines @xmath23 completely @xcite , since the 2-point correlator is all we need to know for a gaussian distribution . as we will discuss later , the small @xmath34 quantum evolution of the correlator in eq . ( [ eq : unintbgv ] ) is given by the bk equation . these distributions @xmath23 , albeit very similar to the canonical unintegrated gluon distributions in the hadrons , should not be confused with the latter @xcite . however , at large @xmath18 ( @xmath69 ) , they coincide with the usual unintegrated gluon distribution and this determines their normalization ] this normalization is different from ref . @xcite - we have checked however that , when appropriately normalized , our expression for the inclusive gluon distributions is identical to that of ref . @xcite . ] . the @xmath18-factorized expression in eq . ( [ eq : ktfact ] ) is only valid for inclusive gluon production when one of the hadrons is dilute and the other is dense . in the cgc framework , this means that we keep only terms of orders @xmath70 and @xmath71 in the amplitudes , if @xmath45 and @xmath20 are the dilute and dense hadrons respectively . this factorization is therefore applicable for proton - proton collisions , or nucleus - nucleus collisions in the fragmentation region of one of the projectiles . clearly , it applies as well to proton / deuteron gold collisions . @xmath18-factorization breaks down in kinematic regions that do not satisfy this dilute dense criterion . it particular , it is not a good approximation in the central rapidity region . although there are some analytical attempts to address these violations of @xmath18-factorization @xcite , these have been computed only numerically @xcite thus far . for quark production , @xmath18-factorization is broken already at leading order @xcite . the magnitude of their breaking has been quantified recently @xcite . here we will consider only the @xmath18-factorized expression of eq . ( [ eq : ktfact ] ) with the understanding that this expression likely has significant corrections at central rapidities . these will be discussed further later in the paper . from eq . ( [ eq : ktfact ] ) , it is easy to see how limiting fragmentation emerges in the limit where @xmath72 . in this situation , the typical transverse momentum @xmath18 in the projectile is much smaller than the typical transverse momentum @xmath73 in the target , because these are controlled by saturation scales evaluated respectively at @xmath19 and at @xmath21 . therefore , at sufficiently high energies , it is legitimate to approximate @xmath74 by @xmath75 . when we integrate the gluon distribution over @xmath76 , we thus obtain @xmath77 this expression , in the stated approximation , is nearly independent of @xmath21 and therefore depends only weakly on on @xmath31 . to see this more clearly , note that to obtain the second line in the above expression , we used eq . ( [ eq : unintbgv ] ) and the fact that the wilson line @xmath56 is a unitary matrix . the details of the evolution are therefore not important to achieve limiting fragmentation , only that the evolution equation preserve unitarity . the residual dependence on @xmath21 comes from the the upper limit @xmath78 of the integral in the second line . this ensures the applicability of the approximation that led to the expression in the second line above . the integral over @xmath79 gives the integrated gluon distribution in the projectile , evaluated at a resolution scale of the order of the saturation scale of the target . therefore , the residual dependence on @xmath80 arises only via the scale dependence of the gluon distribution of the projectile . this residual dependence on @xmath80 is very weak at large @xmath19 because it is the regime where bjorken scaling is observed . the formula in eq . [ eq : dngdyaprox ] was used previously in ref . the nuclear gluon distribution here is determined by global fits to deeply inelastic scattering and drell - yan data . we note that the glue at large @xmath34 is very poorly constrained at present @xcite . the approach of bialas and jezabek @xcite also amounts to using a similar formula , although convoluted with a fragmentation function ( see eqs . ( 1 ) , ( 4 ) and ( 5 ) of @xcite in addition , both the parton distribution and the fragmentation function are assumed to be scale independent in this approach ) . we will discuss the effect of fragmentation functions later in our discussion . though limiting fragmentation can be simply understood as a consequence of unitarity in the high energy limit , what may be more compelling are observed deviations from limiting fragmentation and how these vary with energy . these deviations may probe more deeply our understanding of the dynamics of both large @xmath34 and small @xmath34 modes in hadronic wavefunctions . in particular , in the small @xmath34 case , it may provide further insight into the renormalization group equations that , while trivially preserving unitarity , demonstrate interesting pre - asymptotic behavior . these concerns provide the motivation for this detailed study with the balitsky - kovchegov renormalization group equation to be discussed in the following section . we begin by briefly recapitulating key features of the balitsky - kovchegov ( bk ) equation and its solution @xcite . readers are referred to recent review literature on the subject for a more detailed discussion @xcite . the bk equation is a non - linear evolution equation in rapidity @xmath81 for the forward scattering amplitude of a quark - antiquark dipole scattering off a target in the limit of very high center - of - mass energy @xmath0 . it was originally derived , within the dipole picture ( which assumes the large @xmath44 limit ) at small values of bjorken @xmath34 , by taking into account multiple rescatterings of the @xmath82 dipoles off a dense nuclear target . the bk equation for the amplitude is equivalent to the corresponding jimwlk equation @xcite of the color glass condensate , in a mean field ( large @xmath44 and large @xmath45 ) approximation where higher order dipole correlators are neglected . the parametrically suppressed @xmath44 and @xmath45 contributions can , in principle , be computed by solving the jimwlk equation . in momentum space , the bk equation takes the form @xmath83 where we denote @xmath84 . the operator @xmath85 is the well known bfkl kernel in momentum space @xcite , whose action on the function @xmath86 is given by @xmath87 the function @xmath88 is the bessel - fourier transform of the dipole - target scattering amplitude @xmath89 : @xmath90 where @xmath91 is the size of the @xmath82 dipole and @xmath18 is its conjugate transverse momentum . the dipole amplitude @xmath92 is defined in terms of the correlator of two wilson lines of gauge fields in the target as @xmath93 where we have assumed translation invariance in the transverse plane in order to set the quark transverse coordinate to @xmath94 . here @xmath95 is a wilson line in the fundamental representation , obtained by replacing in eq . ( [ eq : pathglue ] ) the adjoint generators @xmath59 by the generators @xmath96 in the fundamental representation . however , to evaluate the `` unintegrated gluon distribution '' defined in eq . ( [ eq : unintbgv ] ) , one needs to know the rapidity dependence of the correlator of two wilson lines in the _ adjoint representation _ , @xmath97 . this can be obtained as follows . in the cgc framework , the weight functional for the color sources in generating functional can be expressed in the large @xmath44 and large @xmath45 limit as a non - local gaussian distribution of color sources . in this limit , one can obtain closed form expressions for expectation values of both the fundamental and adjoint correlators of wilson lines . one therefore has @xmath98 where @xmath99 is the casimir in the adjoint representation . the function @xmath100 is proportional to the variance of the non - local gaussian weight functional in the generating functional and is therefore the same in both the fundamental and adjoint cases . as @xmath101 in the large @xmath44 limit , eqs . ( [ eq : largen - corr ] ) and ( [ eq : dipole ] ) give @xmath102 ^ 2\ ; .\ ] ] substituting the lhs side here by the rhs into eq . ( [ eq : unintbgv ] ) , we obtain @xmath103 ^ 2 \ ; . \label{eq : phiadjoint}\ ] ] we digress here to note that if instead we had used the correlator of two wilson lines in the fundamental representation in eq . ( [ eq : unintbgv ] ) , we would have obtained the following expression for the unintegrated gluon density @xmath104 \ ; . \label{eq : phifundamental}\ ] ] the change from the correlator of wilson lines in the fundamental representation to the adjoint one corresponds to the emission of the gluon from the triple pomeron vertex itself . this emission is not prohibited by the abramovsky gribov kancheli ( agk ) cutting rules @xcite for inclusive gluon production in high energy qcd . it was argued previously @xcite that the numerical difference in the resulting rapidity distributions is rather small . in figure [ fig:1a ] , we compare numerical solutions ( to be discussed further shortly ) for the unintegrated gluon distribution using the correlator of wilson lines in the adjoint representation ( eq . [ eq : phiadjoint ] ) ) with the one using the correlator of wilson lines in the fundamental representation ( eq.[eq : phifundamental ] ) . the shape of the distribution is very similar , whereas the position of the peak is different . at @xmath105 the position of the peaks differ by the ratio of the @xmath106 . at higher values of @xmath107 this difference increases due to the faster evolution of the correlator in the adjoint representation . the bk equation is the simplest evolution equation capturing both the leading @xmath108 bfkl dynamics at moderately small values of @xmath34 , as well as the recombination physics of high parton densities at very small values of @xmath34 . in the limit where the dipole scattering amplitude is small , @xmath109 , the non - linear term in eq . ( [ eq : kov ] ) can be ignored and the bk equation reduces to the bfkl equation . in this limit , the amplitude has the solution , @xmath110 where @xmath111 , @xmath112 and @xmath113 . if we define the saturation condition as @xmath114 one gets @xcite @xmath115 a more careful analysis of the bk - equation close to the saturation boundary gives solution @xmath116 . in the strong field ( saturation ) limit @xcite where @xmath117 , one finds @xmath118 with @xmath119 and where @xmath120 is an undetermined constant . in the large @xmath107 asymptotic regime , leading and next - to - leading corrections to the form of the bk amplitude and the saturation scale have been computed @xcite . the coupling constant @xmath121 in eq . ( [ eq : kov ] ) is fixed in the leading logarithmic approximation in @xmath122 . this leads to a rather strong dependence of the saturation scale in the leading logarithmic approximation : @xmath123 with @xmath124 . for typical values of @xmath125 , of order @xmath126-@xmath127 , @xmath128 is at least a factor of three higher than fits to the hera and rhic data would suggest @xcite . it is therefore desirable to include at least part of the next - to - leading corrections in @xmath108 to the bk equation . such corrections are known to reduce significantly the pomeron intercept in the case of the bfkl equation . the bk equation has been solved numerically for both fixed and running coupling @xcite . these solutions have the following features : * the dipole amplitude is shown to unitarize , and the solution exhibits geometrical scaling . * the saturation scale has the behavior in eq . ( [ eq : fixed - sat - scale ] ) or @xmath129 in the fixed and running coupling cases respectively . * the infrared diffusion problem of the bfkl solution is cured by the non - linear term in the bk equation . * ultraviolet diffusion is still present in the bk equation . however , it is attenuated by including running coupling effects @xcite . in this work , we solved the bk equation numerically , in both fixed and running coupling cases , to investigate limiting fragmentation in hadronic collisions . in the fixed coupling case , since realistic values of @xmath125 give too high a value of @xmath128 , we solved the bk equation with very small values of @xmath125 chosen to give @xmath128 values that are compatible with data from hera , rhic and hadron colliders . ( we will see that these lower values are indeed favored by the data . ) this is not unreasonable because it has been shown @xcite that resummed next - to - leading order corrections to the bfkl equation in the presence of a unitary boundary ( _ la _ bk ) give values of the saturation scale @xmath128 that are nearly independent of the energy , as in the lo case , but with a much smaller value of @xmath128 . the limited running coupling studies we performed did not give nearly as good agreement with the data than the fixed coupling studies with the lower values of @xmath125 . this is likely due to the fact that the energy dependence in this case is much too fast compared to the data . further , since our results are sensitive to small transverse momenta , they will also depend strongly on how one regulates the running of @xmath130 in the infrared . this requires a more detailed study than reported here . we shall therefore restrict ourselves to the fixed coupling case in discussing our results in the following section . as explained before , the `` unintegrated gluon distributions '' @xmath23 are obtained from solutions to eq . ( [ eq : kov ] ) . at small values of @xmath34 , where the gluon density is high , eq . ( [ eq : kov ] ) captures the essential physics of saturation ; we therefore use it to determine @xmath131 for values of @xmath132 with @xmath133 . for larger values of @xmath34 ( @xmath134 ) , we used the phenomenological extrapolation , @xmath135 the parameter @xmath136 is fixed by qcd counting rules @xcite . we checked that the results are not sensitive to the variation of this parameter in the range @xmath137 . the parameter @xmath138 was varied between @xmath94 and @xmath139 in the fits to data discussed in the following section . the initial condition for the bk evolution , that gives @xmath140 , is given by the mclerran - venugopalan ( mv ) model @xcite with a fixed initial value of the saturation scale @xmath141 . for a gold nucleus , extrapolations from hera and estimates from fits to rhic data suggest that @xmath142 . the saturation scale in the proton is taken to be @xmath143 . for comparison , we also considered initial conditions from the golec - biernat and wusthoff ( gbw ) model @xcite . the values of @xmath144 were varied in this study to obtain best fits to the data . experimental data are presented in terms of the measured distributions of produced charged hadrons . our expression in eq . ( [ eq : ktfact ] ) is for produced gluons . rapidity distributions are dominated by low @xmath145 ( @xmath146 ) particles ; the detailed mechanism of how such soft gluons fragment to form hadrons is unknown . however , in @xmath147 collisions , several studies have been performed of hadronization ; it is observed that at scales comparable to @xmath148 , the distribution of produced hadrons mirrors that of the produced gluons . this hypothesis is known as parton - hadron " duality @xcite and we shall adopt it here . spectral shapes of softer partons , albeit one would imagine them to interact more strongly , are closer to that of observed hadrons - than those with @xmath149 . this is fortuitous for the discussion of limiting fragmentation here since @xmath150 is dominated by soft momenta . ] we will discuss the effect of fragmentation functions shortly . limiting fragmentation is often experimentally studied in terms of the measured pseudo - rapidity @xmath151 of produced particles . for massless particles , @xmath151 and @xmath152 are the same , but they differ for massive particles . one obtains @xmath153 \ ; . \label{eq : yeta}\ ] ] we choose @xmath154 to be of order of @xmath155 . our expression for @xmath150 of gluons has a logarithmic divergence at low @xmath145 ; in addition to parton - hadron duality , for self consistency , we should regulate the corresponding expression for charged hadrons with the same mass as the one used in the conversion from @xmath152 to @xmath151 . we will later discuss the sensitivity of our results to @xmath154 . we now have all the ingredients necessary to calculate the multiplicity distributions in the color glass condensate framework . as discussed previously , experimental data on pseudo - rapidity distributions in the projectile fragmentation region scale with @xmath156 for different energies . in eq . ( [ eq : ktfact ] ) , @xmath157 depends on the difference @xmath158 and @xmath159 on the sum @xmath160 . to provide a sense of the @xmath34 s involved , consider gold - gold collisions at rhic . for @xmath161 , which lies in the fragmentation region , and @xmath162 gev , @xmath163 and @xmath164 . one is therefore probing very small values of @xmath21 in nuclei in this region can be larger than unity in a nucleus and could in principle take values up to @xmath45 , the number of nucleons . ] , where the saturation scale @xmath165 is rather large . as we discussed previously , in this kinematic region , because of unitarity , the gluon distribution depends only weakly on @xmath166 . this qualitatively explains the scaling in the limiting fragmentation region . we now turn to our results . to reiterate , they are obtained by a ) solving the bk equation to obtain eq . ( [ eq : unint - gluon ] ) , and thereby eq . ( [ eq : phiadjoint ] ) for the unintegrated distribution for @xmath167 , b ) using eq . ( [ eq : extrapolx ] ) to determine the large @xmath34 ( @xmath168 ) behavior , and c ) substituting these in eq . ( [ eq : ktfact ] ) to determine rapidity distribution of gluons . the pseudo - rapidity distributions are determined by the transformation in eq . ( [ eq : yeta ] ) . in figure [ fig:2 ] we show pseudo - rapidity distributions of the charged particles produced in nucleon - nucleon collisions for center of mass energies ranging from @xmath169 to @xmath170 . we have performed calculations for two different input distributions at starting value of @xmath133 , namely , the gbw and mv models . the normalization has been treated as a free parameter . extrapolations to large values of @xmath34 are performed using the formula in eq . ( [ eq : extrapolx ] ) . plots on the left hand side of figure [ fig:2 ] are obtained for @xmath171 whereas the right hand side plots are done for @xmath172 . the different values of @xmath128 are obtained by varying @xmath125 when solving the bk equation . note that while @xmath128 controls the growth of @xmath173 with energy , the amplitude in eq . ( [ eq : omega ] ) has the growth rate @xmath174 which is significantly lower . so @xmath175 , which gives reasonable fits ( more on this in the next paragraph ) to the pp data for the mv initial conditions , corresponds to @xmath176 , which is close to the value for the energy dependence of the amplitude in next - to - leading order resummed bk computations @xcite and in empirical dipole model comparisons to the hera data @xcite . we find that our computations are extremely sensitive to the extrapolation prescription to large @xmath34 . this is not a surprise since we are probing the wave - function of the projectile at fairly large values of @xmath19 . from our analysis , we see that the data naively favor a non - zero value for @xmath138 in eq . ( [ eq : extrapolx ] ) . the zero value of @xmath138 results in the distributions which , in both the mv and gbw cases give reasonable fits ( albeit with different normalizations ) at lower energies but systematically become harder relative to the data as the energy is increased . to fit the data in the mv model up to the highest ua5 energies , a lower value of @xmath128 than that in the gbw model is required . this is connected with the fact that mv model has tails which extend to larger values in @xmath18 than in the gbw model . as the energy is increased , the typical @xmath177 does as well . we will return to this point shortly . in figure [ fig:3 ] the same distributions are shown as a function of the @xmath178 . the calculations for @xmath171 , are consistent with scaling in the limiting fragmentation region . there is a slight discrepancy between the calculations and the data in the mid - rapidity region . this discrepancy may be a hint that one is seeing violations of @xmath18 formalism in that regime because @xmath18 formalism becomes less reliable the further one is from the dilute - dense kinematics of the fragmentation regions @xcite . this discrepancy should grow with increasing energy . however , our parameters are not sufficiently constrained that a conclusive statement can be made . in changing from rapidity to pseudo - rapidity distributions , one has to use a jacobian with a mass parameter . we have chosen this parameter , for consistency , to be the same as our @xmath145 cut - off for the integration over the transverse momentum in the formula for the gluon production . we have checked the sensitivity of our calculations to variations in the @xmath145 cut - off @xmath154 in the range @xmath179 . in figure [ fig:3a ] we show the results for the pp collisions with two different values of @xmath180 . the ` dip ' in the pseudo - rapidity distribution becomes more or less pronounced when the parameter @xmath154 is decreased or increased respectively . we note that in order to obtain a reasonable description of the data we also had to adjust the other parameters ( normalization and @xmath128 ) . in figure [ fig:4 ] we show the extrapolation to higher energies , in particular the lhc range of energies for the calculation with the gbw input . we observed previously that the mv initial distribution , when evolved to these higher energies , gives a rapidity distribution which is very flat in the @xmath181 regime . we noted that this is because the average transverse momentum grows with the energy giving a significant contribution from the high @xmath18 tail of the distribution in the mv input in eq . ( [ eq : extrapolx ] ) . the effect of fragmentation functions on softening the spectra in the limiting fragmentation region can be simply understood by the following qualitative argument . the inclusive hadron distribution can be expressed as @xmath182 where @xmath183 is the fragmentation function denoting the probability , at the scale @xmath184 , that a gluon fragment into a hadron carrying a fraction @xmath185 of its transverse momentum . for simplicity , we only consider here the probability for gluons fragmenting into the hadron . the lower limit of the integral can be determined from the kinematic requirement that @xmath186we obtain , @xmath187 if @xmath188 were zero , the effect of including fragmentation effects would simply be to multiply eq . ( [ eq : fragment ] ) by an overall constant factor . at lower energies , the typical value of @xmath189 is small for a fixed @xmath31 ; the value of @xmath188 is quite low . however , as the center of mass energy is increased , the typical @xmath189 value grows slowly with the energy . this has the effect of raising @xmath188 for a fixed @xmath31 , thereby lowering the value of the multiplicity in eq.([eq : fragment ] ) for that @xmath31 . note further that eq . ( [ eq : min - frac ] ) suggests that there is a kinematic bound on @xmath189 as a function of @xmath31only very soft gluons can contribute to the inclusive multiplicity . in figure [ fig : pt_pp ] we illustrate @xmath145 distributions obtained from the mv input compare to the ua1 data @xcite . we compare the calculation with and without the fragmentation function . the fragmentation function has been taken from @xcite . clearly the bare " mv model does not describe the data at large @xmath18 because it does not include fragmentation function effects which , as discussed , make the spectrum less flat . in contrast , because the @xmath18 spectrum of the gbw model dies exponentially at large @xmath18 , this unphysical " @xmath18 behavior mimics the effect of fragmentation functions see figure [ fig : pt_pp ] . hence extrapolations of this model , as shown in figure [ fig:4 ] give a more reasonable looking result . similar conclusions were reached previously in ref . @xcite . we next compute the pseudo - rapidity distribution in deuteron - gold collisions . in figure [ fig:6 ] we show the result for the calculation compared with the da data @xcite . the unintegrated gluons were extracted from the pp and aa data . the overall shape of the distribution matches well on the deuteron side with the minimum - bias data . the disagreement on the nuclear fragmentation side is easy to understand since , as mentioned earlier , it requires a better implementation of nuclear geometry effects . similar conclusions were reached in ref . @xcite in their comparisons to the rhic deuteron - gold data . we now apply our considerations to central nucleus nucleus collisions . in figure [ fig:5 ] we present fits to data on the pseudo - rapidity distributions in gold gold collisions from the phobos , brahms and star collaborations . the data @xcite are for @xmath190 and the brahms data @xcite are for @xmath191 . a reasonable description of limiting fragmentation is achieved in this case as well . one again has discrepancies in the central rapidity region as in the pp case . as mentioned previously , a natural explanation for this discrepancy is the violation of @xmath18-factorization . it is expected that these violations decrease the multiplicity in this regime @xcite relative to extrapolations using @xmath18-factorization . we find that values of @xmath192 gev for the saturation scale give the best fits . this value is consistent with other estimates @xcite . apparently the gold - gold data are better described by the calculations which have @xmath172 . this might be related to the difference in the large @xmath34 distributions in the proton and nucleus . also , slightly higher values of @xmath128 are preferred . this variation of parameters from aa to pp case might be also connected with the fact that in our approach the impact parameter is integrated out , so that there is no detailed information on the nuclear geometry . a more detailed calculation with the impact parameter dependence taken into account is left for future investigations . in fig . [ fig : auauext ] we show the extrapolation of two calculations to higher energy @xmath193 . we note that the calculation within the mv model gives results which would violate the scaling in the limiting fragmentation region by approximately @xmath194 at larger @xmath31 . this violation is partly because of the effect of fragmentation functions discussed previously and in part due to the fact that the integrated parton distributions from the mv model do not obey bjorken scaling at large values of @xmath34 . in the latter case , the violations are proportional to @xmath195 as discussed previously . the effects of the former are simulated by the gbw model - the extrapolation of which , to higher energies , is shown by the dashed line . the band separating the two therefore suggests the systematic uncertainity in such an extrapolation coming from a ) the choice of initial conditions and b ) the effects of fragmentation functions which are also uncertain at lower transverse momenta . finally , the transverse momentum distribution in the gold - gold collisions is presented in fig . [ fig : pt_auau ] . the data at @xmath196 are from the phobos collaboration @xcite . again , in this case , the calculation without fragmentation function tends to overshoot the data significantly at high values of @xmath197 . including fragmentation function effects @xcite results in a better agreement with the shape of the @xmath197 distribution as expected . a more detailed study to extract parameters that give the best chi - squared values will be performed at a later date . we studied here the phenomenon of limiting fragmentation in the color glass condensate framework . in the dilute - dense ( projectile - target ) kinematics of the fragmentation regions , one can derive ( in this framework ) an expression for inclusive gluon distributions which is @xmath18 factorizable into the product of unintegrated " gluon distributions in the projectile and target . from the general formula for gluon production(eq . ( [ eq : ktfact ] ) ) , limiting fragmentation is a consequence of two factors : * unitarity of the @xmath56 matrices which appear in the definition of the unintegrated gluon distribution in eq . ( [ eq : unintbgv ] ) . * bjorken scaling at large @xmath19 , namely , the fact that the integrated gluon distribution at large @xmath34 , depends only on @xmath19 and not on the scale @xmath165 . ( the residual scale dependence consequently leads to the dependence on the total center - of - mass energy . ) deviations from the limiting curve at experimentally accessible energies are very interesting because they can potentially teach us about how parton distributions evolve at high energies . in the cgc framework , the balitsky kovchegov equation determines the evolution of the unintegrated parton distributions with energy from an initial scale in @xmath34 chosen to be @xmath133 . this choice of scale is inspired by model comparisons to the hera data . we compared our results to data on limiting fragmentation from pp collisions at various experimental facilities over a wide range of collider energies , and to collider data from rhic for deuteron - gold and gold - gold collisions . we obtained results for two different models of initial conditions at @xmath198 ; the mclerran - venugopalan model ( mv ) and the golec - biernat wusthoff ( gbw ) model . in addition to the two parameters in the initial conditions ( @xmath128 and @xmath138 ) , we also studied the sensitivity of our results to an infrared momentum cut - off @xmath154 ( chosen to be the same value in the @xmath199 conversion for hadrons ) . we found reasonable agreement for this wide range of collider data for the limited set of parameters . clearly these can be fine tuned by introducing further details about nuclear geometry . that would introduce further parameters but on the other hand there is more data for different centrality cuts as well - we leave these detailed comparisons for future studies . in addition , an important effect , which improves agreement with data , is to account for the fragmentation of gluons in hadrons . in particular , the mv model , which has the right leading order large @xmath18 behavior , but no fragmentation effects , is much harder than the data . the latter falls as a much higher power of @xmath18 . since even rapidity distributions at higher energies data are more sensitive to larger @xmath18 , we expect this discrepancy to show up in our studies of limiting fragmentation and indeed it does . we noticed that taking this into account lead to much more plausible extrapolations of fits of existing data to lhc energies . gluon fragmentation " contribution also suggests that the flat " deviations that we found ( for fits with mv initial conditions ) for @xmath200 ( @xmath201 ) are diminished . clearly , the procedure employed in this paper needs further improvements . one of them is the impact parameter dependence of the unintegrated gluon distribution functions . we discussed briefly gluon fragmentation " effects ( with different fragmentation function sets ) which need to be taken into account . furthermore , large @xmath34 distributions also need to be better constrained and consistency with computations of other final states in the cgc framework established . finally , the factorization formula used in this paper involves only gluons ; for more realistic estimates , one should also include quark distributions in this framework . this work was inspired by discussions with a. biaas and l. mclerran and by a seminar at saclay by a. biaas that was attended by fg and rv . we would also like to thank m. baker and r. noucier of the phobos collaboration and m. murray and f. videbaek of the brahms collaboration for their help with their respective data sets and for useful comments . we thank b. mohanty for bringing the star data on limiting fragmentation to our attention . finally , fg wishes to thank the hospitality of the nuclear theory group at bnl . ams and rv were supported by doe contract no . de - ac02 - 98ch10886 . ams was also supported by the polish committee for scientific research , kbn grant no . 1 p03b 028 28 . j. jalilian - marian , a. kovner , l.d . mclerran and h. weigert , _ phys . _ * d55 * : 5414 ( 1997 ) ; j. jalilian - marian , a. kovner , a. leonidov and h. weigert , _ nucl . * b 504 * ( 1997 ) 415 ; _ phys . _ * d 59 * ( 1999 ) 014014 ; j. jalilian - marian , a. kovner and h. weigert , _ phys . * d 59 * ( 1999 ) 014015 ; e. iancu , a. leonidov and l.d . mclerran , _ nucl.phys . _ * a692 * ( 2001 ) 583 ; e. ferreiro , e. iancu , a. leonidov and l.d . mclerran , _ nucl . _ * a 703 * , 489 ( 2002 ) ; e. iancu , a. leonidov and l. d. mclerran , _ phys . lett . _ * b510 * , 145 ( 2001 ) . l. n. lipatov , _ sov . * 23 * ( 1976 ) 338 ; + e. a. kuraev , l. n. lipatov and v. s. fadin , _ sov . * 45 * ( 1977 ) 199 ; + i. i. balitsky and l. n. lipatov , _ sov . j. nucl . _ * 28 * ( 1978 ) 338 . e. gotsman , e.m . levin , m. lublinsky and u. maor , _ nucl . phys . _ * a 696 * ( 2001 ) 851 ; e. gotsman , e.m . levin , m. lublinsky and u. maor , _ eur . * c27 * ( 2003 ) 411 ; e. levin and m. lublinsky , _ nucl . phys . _ * a696 * ( 2001 ) 833 ; m. lublinsky , _ eur . phys . j. _ * c21 * ( 2001 ) 513 . m. hirai , s. kumano and m. miyama , _ phys . rev . _ * d64 * ( 2001 ) 034003 ; + m. hirai , s. kumano and t .- h . nagai , _ phys . _ * c70 * ( 2004 ) 044905 . k. j. eskola , v. kolhinen and c. salgado , _ eur . phys . j. _ * c9 * ( 1999 ) 61 , d. de florian and sassot , _ phys . rev . _ * d69 * ( 2004 ) 074028 .

limiting fragmentation in proton - proton , deuteron - nucleus and nucleus - nucleus collisions is analyzed in the framework of the balitsky - kovchegov equation in high energy qcd . good agreement with experimental data is obtained for a wide range of energies . further detailed tests of limiting fragmentation at rhic and the lhc will provide insight into the evolution equations for high energy qcd . 1 . service de physique thorique ( ura 2306 du cnrs ) + cea / dsm / saclay , bt . 774 + 91191 , gif - sur - yvette cedex , france 2 . brookhaven national laboratory , + physics department , nuclear theory , + upton , ny-11973 , usa 3 . h. niewodniczaski institute of nuclear physics , + polish academy of sciences , + ul . radzikowskiego 152 , 31 - 342 krakw , poland

the fermilab experiment e781 ( selex ) : a segmented large @xmath0 baryon spectrometer @xcite , which took data in the 1996/97 fixed target run at fermilab , is designed to perform high statistics studies of production mechanisms and decay physics of charmed and charmed - strange baryons such as @xmath1 , @xmath2 , @xmath3 and @xmath4 . the physics goals of the experiment require good charged particle identification to look for the different baryon decay modes . one must be able to separate @xmath5 , @xmath6 and @xmath7 over a wide momentum range when looking for charmed baryon decays like @xmath8 . a rich detector with a 2848 phototube photocathode array has been constructed @xcite to do this . the detector begins about @xmath9 downstream of the charm production target , with two analysis magnets with @xmath10 @xmath11-kick each in - between , and is surrounded by multi - wire proportional and drift chambers which provide particle tracking . the average number of tracks reaching the rich is about 5 per event . first results from this detector can be found in @xcite . a detailed description of the detector itself is given in @xcite . in this article we first describe shortly the main parts of the detector ( vessel , mirrors , photon detector ) , after this we will review its stability during the run and performance for physics analyses . finally we report about physics results obtained with the help of the rich up to now . a detailed description can be found in @xcite . here we will repeat the main features only . the selex rich detector consists of a @xmath12 long vessel with @xmath13diameter , filled with pure neon at atmospheric pressure . after initial filling to a slight ( @xmath14 ) overpressure , the vessel was perfectly tight for a period of more than 15 months . at the end of the vessel an array of 16 hexagonal spherical mirrors is mounted on a low - mass hex - panel with 3-point mounts to form a sphere of @xmath15 radius . each mirror is @xmath16 thick , made out of low - expansion class , and has a reflectivity @xmath17 at @xmath18 . before evaporating , the quality of every mirror was measured @xcite . the phototube matrix at the focal plane consists of 2848 @xmath19photomultipliers in a hexagonal closed packed matrix of 89 columns in 32 rows . one column of 32 phototubes is connected to one common high voltage . in the center part , alternating columns of hamamatsu r760 and feu60 tubes are employed , and in the outer part only feu60 tubes are used . every phototube is connected to a preamplifier - discriminator - ecl driver hybrid chip , and the readout of the signals is performed with a standard wire - chamber electronics @xcite with @xmath20 integration time . the whole vessel is tilted by @xmath21 so particles are not passing through the phototubes . the detector performance over the 15 month running period was very stable . in the central region ( mixed r760 and feu60 ) we observed on average 13.6 photons for a @xmath22 particle , corresponding to a figure of merit @xcite @xmath23 , and in the outside ( only feu60 ) we obtained @xmath24 . the refractive index , measured in several different ways , was stable to better than @xmath25 , which would change the ring radius at @xmath22 by less than @xmath26 of its resolution . the single photon resolution was estimated to be @xmath27 , with the biggest contribution coming from the size of the phototubes . the ring radius resolution for the typical multitrack environment was measured at @xmath28 . = = in fig . [ radii ] we show the distribution of ring radii for single track events , where one can identify eight particles and eight anti - particles , demonstrating the huge dynamic range of this detector . with the same dataset we measured the noise of the detector , by counting the number of hits not assigned to a ring , taking into account the integration time of the readout electronics . = the result is shown in fig . [ noise ] , for the two different types of phototubes used . for beam - off events we observed on average 6 hits . for particle identification , we calculate likelihoods for different particle hypotheses @xcite , using information from the tracking system . with the help of kinematically reconstructed signals we measure the identification efficiency for different cuts . for protons ( using @xmath29 decays ) in the momentum range @xmath30 , the identification efficiency is around @xmath31 , and even below proton threshold ( @xmath32 ) , the efficiency is above @xmath33 @xcite , with a mis - identification rate of a few percent , which can be attributed to tracking errors . the same is true for kaons @xcite . = pions from charm decays have usually lower momenta than other decay products ; for that reason in selex we accept as pions in charm analysis also tracks which do not reach the rich . should the track be within the rich acceptance , we require the ratio of the likelihood of that track to be a pion to all other particles to be @xmath34 . in fig . [ pions ] we show that the identification efficiency is above @xmath33 without a sharp turn - on . selex submitted first publications in 1999 , after first reports at conferences in 1998 . to date selex has published seven papers and two more are submitted . some papers on mesons and hyperons do not use the rich in their analyses @xcite , but for charm physics the rich proved to be essential in the analysis , reducing the background significantly but still having high efficiency . selex published measurements of the @xmath35 @xcite and @xmath36 @xcite lifetimes , where the rich was essential in suppressing reflection below the mass peak stemming from mis - identified daughter particles . the measurement of production properties of @xmath37 @xcite with 4 different beam particles ( @xmath38 , @xmath39 , @xmath7 , @xmath40 ) also benefitted largely from the rich . more hadroproduction results for @xmath41 , @xmath42 , and @xmath36 are in preparation @xcite . selex also observed for the first time a cabbibo suppressed decay mode of a charmed strange baryon , the @xmath43 @xcite . just recently , selex reported on the first observation of doubly - charmed baryons @xcite . the rich detector is an essential part of the physics analysis in selex , due to its high efficiency and large dynamic range . it fulfilled all the design criteria . with seven papers already published plus two submitted , the collaboration is preparing at this moment several more publications to be submitted in the near future . after a first pass over all data taken ( @xmath44 interactions , @xmath45 events on tape ) in 1998/9 , from which all of the publications to date were obtained , we finished a second pass with improved reconstruction software in 2001 . the analysis of these data is just starting and many more results are expected . the authors are greatful for stimulating discussions with the selex collaboration members . they wish to thank the technical staff of their institutes for their support . this work was supported by the us department of energy under contract no . de - ac02 - 76cho3000 , the russian ministry of science and technology , and consejo nacional de ciencia y tecnologa ( mexico ) . selex collaboration : carnegie - mellon university , fermi national accelerator laboratory , university of iowa , university of rochester , university of hawaii , university of michigan flint , ball state , petersburg nuclear physics institute , itep ( moscow ) , ihep ( protvino ) , moscow state university , university of so paulo , centro brasileiro de pesquisas fisicas ( rio de janeiro ) , universidade federal da paraiba , ihep ( beijing ) , university of bristol , tel aviv university , max - planck - institut fr kernphysik ( heidelberg ) , university of trieste , university of rome `` la sapienza '' , infn , universidad autnoma de san luis potos , bogazici university . + http://fn781a.fnal.gov selex collaboration , j.s . et al . _ , in : _ proceedings of the 29th international conference on high energy physics _ , 1998 , edited by a. astbury _ ( world scientific , singapore , 1998 ) , vol . ii , p. 1259 . preprint hep - ex/9812031 .

selex took data in the 1996/7 fixed target run at fermilab . the excellent performance parameters of the selex rich detector had direct influence on the quality of the obtained physics results .

coupled oscillators play important role in many scientific fields , e.g. biology , electronics , and mechanics , see awrejcewicz1991,kozlowski1995,kuznetsov2009 and references therein . in this paper we analyse two coupled oscillators , one of which is driven by an external periodic force . important example of such system is a dynamic vibration absorber which consists of a mass @xmath3 , attached to the primary vibrating system of mass @xmath4 @xcite . equations describing dynamics of such system are of form:@xmath5where @xmath6 , @xmath7 and @xmath8 , @xmath9 represent ( nonlinear ) force of internal friction and ( nonlinear ) elastic restoring force for mass @xmath4 and mass @xmath3 , respectively . in the present paper we do not assume that the ratio @xmath10 is small . in the present paper we shall consider a simplified model:@xmath11 dynamics of coupled periodically driven oscillators is very complicated awrejcewicz1991,kozlowski1995,kuznetsov2009 . we simplified the set equations ( [ model1 ] ) , ( [ simplified ] ) by reducing it to the problem of motion of two independent oscillators . more exactly , we derived the exact fourth - order nonlinear equation for internal motion as well as approximate second - order effective equation in @xcite . moreover , applying the krylov - bogoliubov - mitropolsky method to these equations we have computed the corresponding nonlinear resonances ( cf . @xcite for the case of the effective equation ) . dependence of the amplitude @xmath1 of nonlinear resonances on the frequency @xmath12 is much more complicated than in the case of duffing oscillator and hence new nonlinear phenomena are possible . in the present paper we study metamorphoses of the function @xmath13 induced by changes of the control parameters . the paper is organized as follows . in the next section derivation of the exact 4th - order equation for the internal motion and approximate 2nd - order effective equations in non - dimensional form are presented . in section 3 metamorphoses of amplitude profiles determined within the krylov - bogoliubov - mitropolsky approach for the approximate 2nd - order effective equation are studied and the case of the standard duffing equation is presented as well . more exactly , theory of algebraic curves is used to compute singular points on effective equation amplitude profiles - metamorphoses of amplitude profiles occur in neighbourhoods of such points . in section 4 examples of analytical and numerical computations are presented for the effective equation . our results are summarized and perspectives of further studies are described in the last section . in new variables , @xmath14 , @xmath15 , equations ( model1 ) , ( [ simplified ] ) can be written as : @xmath16 where @xmath17 , @xmath18 , @xmath19 , @xmath20 , @xmath21 , @xmath22 . adding equations ( [ model2 ] ) we obtain important relation between variables @xmath23 and @xmath24 : @xmath25 where @xmath26 . we can eliminate variable @xmath23 in ( [ model2 ] ) to obtain the following exact equation for relative motion : @xmath27where @xmath28 , @xmath29 and @xmath30 is a nondimensional parameter . equations ( [ 4th - a ] ) , ( relation ) are equivalent to the initial equations ( [ model1 ] ) , ( simplified ) @xcite . in the present work we assume : @xmath31 we thus get:@xmath32 we shall write eq . ( [ 4th - b ] ) in nondimensional form . introducing nondimensional time @xmath33 and rescaling variable @xmath24 : @xmath34 where:@xmath35we get the exact equation for motion of mass @xmath36 : @xmath37where nondimensional constants are given by:@xmath38 we shall consider hierarchy of approximate equations arising from ( 4th-4c ) . for small @xmath39 we can reject the second term on the left in ( [ 4th-4c ] ) to obtain the approximate equation:@xmath40which can be integrated partly to yield the effective equation:@xmath41where transient states has been omitted @xcite . and finally , for @xmath42 we get the duffing equation:@xmath43 we applied the krylov - bogoliubov - mitropolsky ( kbm ) perturbation approach @xcite to the effective equation ( [ effective ] ) obtaining for the @xmath44 resonance the following amplitude profile okninski2006:@xmath45now , for @xmath42 , we obtain the amplitude profile for the duffing equation ( [ duffing]):@xmath46 it is well known that dependence of the function @xmath47 , cf . ( [ aduff ] ) , on control parameters @xmath48 , @xmath49 is rather simple . on the other hand , dependence of the amplitude profile @xmath50 on control parameters @xmath48 , @xmath49 , @xmath51 , @xmath52 is more complicated and thus @xmath53 can describe new nonlinear phenomena . in the next section we shall study possible metamorphoses of @xmath47 , @xmath54 induced by changes of control parameters , the more complicated case of the 4th - order exact equation ( [ 4th-4c ] ) will be treated elsewhere . equations ( [ aduff ] ) , ( [ aeff ] ) define the corresponding amplitude profiles implicitly . such amplitude profiles can be classified as planar algebraic curves . firstly , we shall collect useful theorems on implicit functions which will be used below . let us write equations ( [ aeff ] ) , ( [ aduff ] ) as @xmath55 and @xmath56 , respectively , where @xmath57 , @xmath58:@xmath59 it follows from general theory of implicit functions @xcite that conditions for critical points of @xmath60 read : @xmath61moreover , critical points of the inverse function @xmath62 are given by:@xmath63 it may happen that in some points @xmath64 we have:@xmath65 such points are referred to as singular points of algebraic curve @xmath66 because they are in some sense exceptional . singular points of the algebraic curve defined by ( [ fd ] ) are given by:@xmath67 the set of equations ( [ fd ] ) , ( [ sd1 ] ) , ( [ sd2 ] ) can be written as:@xmath68where @xmath69 are positive . general solution reads:@xmath70 it follows that @xmath71 , @xmath72 and @xmath73 and thus the system of equations ( [ d1 ] ) , ( [ d2 ] ) , ( [ d3 ] ) has no acceptable solutions since we assume that @xmath74 are real and @xmath75 are non - negative . singular points of the algebraic curve defined by ( [ feff ] ) are given by equations:@xmath76 it follows from ( [ feff ] ) and ( [ se2 ] ) that either of equations must hold : [ cases ] @xmath77where @xmath69 are positive . let us start with eq.([case2 ] ) . in this case we obtain from ( [ feff ] ) , ( [ se1 ] ) and ( [ case2 ] ) the following rather special solution : @xmath78 let us now consider more general eq.([case1 ] ) . we can treat @xmath79 as arbitrary . then we obtain two solutions for @xmath80:@xmath81where the inequality@xmath82must hold . this means that for a chosen value of @xmath79 the parameter @xmath49 must obey@xmath83 solving equations ( [ feff ] ) , ( [ se1 ] ) , and ( [ y1 ] ) or ( [ y2 ] ) we get:@xmath84where @xmath85 and @xmath86 , @xmath87 are given by:@xmath88with@xmath89@xmath90@xmath91@xmath92 it follows from solutions obtained in the preceding section that we can control position of a singular point . more exactly , we choose a value of @xmath79 and then @xmath49 fulfilling inequality ( [ ineqh ] ) can be chosen as well . next we specify @xmath48 and then @xmath80 , @xmath51 , @xmath52 are computed from eqs . ( [ y ] ) , ( [ a ] ) , ( [ h ] ) . in this process the position of the singular point @xmath96 and values of control parameters @xmath52 , @xmath49 , @xmath48 , @xmath97 are determined ( provided that the solutions are real ) . bifurcation diagram for the effective equation ( [ effective ] ) for the following values of control parameters @xmath98 , @xmath99 , @xmath100 , @xmath101 is shown in fig . [ f1 ] ( cf . fig . 1 _ _ _ _ in @xcite ) where colours mark different initial conditions . @xmath102{fig1.eps}\ ] ] position of the @xmath44 resonance agrees well with the amplitude profile , computed for the same parameters , cf . [ f2 ] and discussion in @xcite . we shall now compute coordinates of a singular point using equations ( [ y ] ) , ( [ a ] ) , ( [ h ] ) . at first we choose the value of @xmath79 as @xmath103 @xmath104 . then we can select any value of @xmath49 obeying inequality ( [ ineqh ] ) . we thus put @xmath105 to get from eq . ( [ y1 ] ) @xmath106 @xmath107 . next we choose @xmath108 to compute from ( [ a ] ) , ( [ h ] ) @xmath109 , @xmath110 . in fig . [ f3 ] below we show amplitude profiles computed from eq . ( [ aeff ] ) for critical parameter values @xmath111 , and for two more values of @xmath52 , @xmath112 and @xmath113 . @xmath102{fig2.eps}\ ] ] @xmath102{fig3.eps}\ ] ] the critical ( red ) curve intersects itself in singular point @xmath114 or @xmath115 . green curve corresponds to @xmath116 while blue curve has been computed for @xmath117 ( other parameter values unchanged ) . the initial amplitude from fig . [ f1 ] was also shown ( black curve ) . the first bifurcation diagram , cf . [ f4 ] , was computed for @xmath118 and corresponds to the green curve in fig . we note that the small branch of the @xmath44 resonance is discontinuous in agreement with the amplitude profile shown in fig . [ f3 ] ( green curve ) . the next bifurcation digram , fig . [ f5 ] , has been computed for critical value @xmath119 determined numerically from eq . ( [ effective ] ) ( this differs slightly from the critical value @xmath110 determined from the kbm solution as described above ) . @xmath120{fig4.eps}\ ] ] @xmath120{fig5.eps}\ ] ] and finally , the last bifurcation diagram was computed for @xmath117 - and again the small branch of the @xmath44 resonance is continuous . @xmath120{fig6.eps}\ ] ] it follows from results presented in section that for @xmath103 , @xmath105 , @xmath121 there is another singular point . indeed , we can compute @xmath80 from another of equations ( [ y ] ) to get from eqs . ( [ a ] ) , ( [ h ] ) @xmath122 @xmath123 , @xmath124 , @xmath125 . @xmath102{fig7.eps}\ ] ] in fig . [ f7 ] amplitude profiles computed from eq . ( [ aeff ] ) for critical parameter values @xmath126 and for two other values of @xmath52 , @xmath127 and @xmath112 have been shown . bifurcation diagrams for @xmath128 and @xmath129 are shown below . @xmath120{fig8.eps}\ ] ] @xmath120{fig9.eps}\ ] ] in this work we have studied metamorphoses of amplitude profiles for the effective equation , describing approximately dynamics of two coupled periodically driven oscillators . our analysis has been analytical although based on the approximate kbm method . theory of algebraic curves has been used to compute singular points on effective equation amplitude profiles . it follows from general theory that metamorphoses of amplitude profiles occur in neighbourhoods of such points . in section 3 we have computed analytically positions of singular points for the amplitude profiles @xmath0 determined within the krylov - bogoliubov - mitropolsky approach for the approximate 2nd - order effective equation ( [ effective ] ) . in the first case the singular point corresponds to self - intersection of @xmath0 , see fig . f3 , while in the second case it is a isolated point , cf . [ f7 ] . it is interesting that the solution described in section 3 permits control of position of singular point : we choose arbitrary value of variable @xmath79 @xmath130 , then value of the parameter @xmath49 obeying inequality ( [ inequality ] ) is selected . finally the value of the control parameter @xmath48 is chosen and @xmath80 , @xmath51 , @xmath52 are computed from eqs . ( y ) , ( [ a ] ) , ( [ h ] ) ; it should be stressed that we have not come across any difficulties to obtain real solutions . we hope to carry full analysis of conditions guaranteeing existence of real solutions in our future papers . as a by - product we have demonstrated that there are no singular points for @xmath0 computed for the duffing equation in agreement with well established numerical experience . we have also computed numerically bifurcation diagrams in the neighbourhoods of singular points and indeed dynamics of the effective equation ( [ effective ] ) changes according to metamorphoses of the corresponding amplitude profiles . a. okniski , j. kyzio , perturbation analysis of the effective equation for two coupled periodically driven oscillators , differential equations and nonlinear mechanics 2006 ( 2006 ) , article i d 56146 , 10 pages .

we study dynamics of two coupled periodically driven oscillators . important example of such a system is a dynamic vibration absorber which consists of a small mass attached to the primary vibrating system of a large mass . periodic solutions of the approximate effective equation are determined within the krylov - bogoliubov - mitropolsky approach to get the amplitude profiles @xmath0 . dependence of the amplitude @xmath1 of nonlinear resonances on the frequency @xmath2 is much more complicated than in the case of one duffing oscillator and hence new nonlinear phenomena are possible . in the present paper we study metamorphoses of the function @xmath0 induced by changes of the control parameters .

the de broglie - bohm pilot wave interpretation,@xcite also known as bohmian mechanics , offers an alternate interpretation of quantum mechanics from the copenhagen standpoint . it is deterministic , involving the trajectories of particles in addition to the wavefunction . while the interpretation contains hidden variables , it is in agreement with bell s theorem due to explicit non - local effects.@xcite for a relativistic particle , the postulates are : ( 1 ) the @xmath0-field satisfies the dirac equation , ( 2 ) the particle s trajectory is determined by a first order equation @xmath1 for conserved current @xmath2 depending on @xmath0,@xcite and ( 3 ) the statistical ensemble of positions obeys the born rule : @xmath3 for probability density @xmath4 at some initial time slice . the universe is not inherently random in this interpretation , but our ignorance of the initial conditions yields a distribution obeying this limitation.@xcite the dirac equation is given by : @xmath5 \psi = 0\ ] ] where the @xmath6 @xmath7-matrices satisfy @xmath8 . we use the metric signature @xmath9 , and the representation : @xmath10 for pauli matrices @xmath11 . the conserved current is given by @xmath12 where @xmath13 . by the second postulate , the guiding equation is : @xmath14 where @xmath15 for affine parameter @xmath16 , and it follows that @xmath17 . the particle trajectory is then determined given an initial spacetime point @xmath18 and proper boundary conditions on @xmath0 . using @xmath19 $ ] with the dirac ( and adjoint ) equation , @xmath2 can be decomposed as : @xmath20 + \frac{i \hbar}{4 m } \partial_\nu \big ( \bar{\psi } [ \gamma^\mu , \gamma^\nu ] \psi \big ) \end{split}\ ] ] which has a similar form to the non - relativistic current . in analogy with the non - relativistic case , we can write the equations of motion in second order form ; from the guiding equation ( [ eq : dirac_guiding ] ) we have : @xmath21 for @xmath22 . defining the matrices @xmath23 and @xmath24 , the dirac & adjoint equations become : @xmath25 differentiating with respect to time yields @xmath26 , where @xmath27 & @xmath28 are defined as : @xmath29 the feynman - stueckelberg interpretation of anti - matter@xcite has been applied to negative energy states to resolve the klein paradox.@xcite our purpose is to show how it can be incorporated into the pilot wave interpretation for a consistent explanation of a single relativistic particle . we review the klein paradox and discuss its resolution through this interpretation in section [ section : paradoxa ] . numerical simulations of a gaussian packet incident on potential steps of various heights are presented in section [ section : paradoxb ] . the potential barrier is analyzed in a similar context in section [ section : barrier ] , and numerical simulations are presented in section [ section : barrier_sim ] . we start by reviewing the klein paradox@xcite in one dimension . the potential is : @xmath30 with @xmath31 and @xmath32 . setting @xmath33 , defining @xmath34 and @xmath35 , the time independent free dirac equation for a spin @xmath36 particle can be written as a two - component spinor : @xmath37 @xmath38 @xmath39 for @xmath40 , we plug in @xmath41 for the incoming and reflected solutions : @xmath42 such that @xmath43 and @xmath44 . defining @xmath45 , the solution in region i is given by : @xmath46 for @xmath47 , we plug in @xmath48 for transmitted wave : @xmath49 such that @xmath50 . similarly by defining @xmath51 , the solution is : @xmath52 continuity at the boundary @xmath53 yields : @xmath54 @xmath55 we can compute @xmath56 & @xmath57 as : @xmath58 and @xmath59 . the interpretation of probability will be given later . as @xmath60 , @xmath61 approaches a finite non - zero value ( since @xmath62 ) suggesting that an infinite potential has some transparency for the electron . for the current @xmath63 in region i : @xmath64 @xmath65 for incident ( @xmath66 ) & reflected ( @xmath67 ) waves . using eq . ( [ eq : dirac_guiding ] ) it follows : @xmath68 in region ii we have : @xmath69 @xmath70 for the transmitted wave . using eq . ( [ eq : dirac_guiding ] ) we find : @xmath71 the feynman - stueckelberg interpretation ( of anti - matter as matter going backwards in time ) enters when the particle is in region ii since it is a negative energy solution . to keep @xmath72 increasing in the phase of @xmath73 when @xmath74 , we change @xmath75 ( or @xmath76 ) in such regions . applying this change to the wavefunction yields @xmath77 satisfying the time - reversed dirac equation@xcite : @xmath78 which corresponds to the complex conjugate in one dimension ; this does nt affect @xmath79 , @xmath80 , or the correct boundary condition @xmath81 since the phase cancels out . applying the change @xmath82 to the guiding equation ( [ eq : dirac_guiding ] ) in region ii yields : @xmath83 this changes the sign of @xmath84 , and therefore @xmath85 . the direction of the velocity @xmath85 in the lab frame is to the _ right _ even though the particle s velocity ( with monotonically increasing affine parameter @xmath16 ) is to the _ left _ and backwards in time . the process can be described as packet @xmath86 at @xmath87 & @xmath88 and packet @xmath57 at @xmath89 & @xmath90 . packet @xmath86 moves forward in time , packet @xmath57 moves backwards in time , and both meet at the step boundary . they constructively interfere forming the reflected packet @xmath56 which moves forward in time from the step boundary to @xmath88 . applying the pilot wave interpretation , there is a probability that the particle is in either packet @xmath86 or packet @xmath57 ( even though the two packets are nt on the same time slice ) with a definite , continuous trajectory _ over all _ of spacetime . the particle s initial spacetime position used to uniquely solve the guiding equation is defined at @xmath91 ; since @xmath92 in region i , @xmath93 as @xmath94 implying an ensemble of initial positions in packet @xmath86 . in region ii , @xmath95 such that @xmath96 as @xmath94 ; this implies the ensemble also includes a distribution of initial positions in packet @xmath57 which will travel backwards in time by the guidance condition . if the particle starts in packet @xmath86 , it is reflected at the step boundary ; if the particle starts in packet @xmath57 , it goes backwards in time to the step boundary , and then moves forward in time with packet @xmath56 towards @xmath88 . in the lab frame , this would appear to describe the creation of a particle / anti - particle pair at the step boundary . as there are two particles in the lab frame , it would be more accurate to describe the ensemble as a single particle _ process _ since there is only a single , continuous trajectory over all of spacetime . if the particle s initial spacetime point corresponds to reflection , the overall charge is @xmath97 ( for electrons ) ; if the initial spacetime point corresponds to a trajectory backwards in time , the overall charge is @xmath98 since in the lab we observe pair creation at the step boundary . it follows that charge is conserved on each time slice . returning to the interpretation of probability , we see at @xmath91 , there are two spacetime packets containing the ensemble with coefficients @xmath86 & @xmath57 . solving for @xmath86 & @xmath57 : @xmath99 as @xmath100 , all the particles move to the packet with coefficient @xmath56 . since @xmath101 , the relation : @xmath102 implies the conservation of _ spacetime _ probability with respect to @xmath16 ; @xmath103 is probability of the particle starting in packet @xmath86 at @xmath91 , and @xmath104 is the probability of the particle starting in packet @xmath57 at @xmath91 . it follows that the average number of particle / anti - particle pairs produced at the step boundary is given by@xcite : @xmath105 we can now reinterpret klein s result : as @xmath60 , @xmath106 approaches a finite non - zero value . this implies that for a potential @xmath107 , there will be pair production at the step boundary even if @xmath60 . the new feature of the interpretation is the probability of an _ initial _ particle position at a _ future _ time slice . as we shall see in section [ section : barrier ] , this is nt encountered in the potential barrier case . the feature is due to the infinite extent of the potential @xmath108 to @xmath90 yielding an asymptotic negative energy solution . we present numerical simulations for an incident gaussian packet starting with the free particle ( case 0 ) . the three cases with @xmath109 are : @xmath110 ( case 1 ) , @xmath111 ( case 2 ) , and the klein paradox @xmath112 ( case 3 ) . setting @xmath113 , the packet is constructed by integrating plane wave solutions over energy . following the notation of nitta , kudo , and minowa,@xcite we start by defining : @xmath114 where @xmath115 is the heaviside step function . the forward evolving solution @xmath116 ( and corresponding boundary condition ) is replaced by the time - reversed wavefunction @xmath77 for case 3 . the time - dependent wavefunction @xmath117 is : @xmath118 where @xmath119 for case 1 , @xmath120 for case 2 , and @xmath121 for case 3 . the gaussian function @xmath122 is defined as : @xmath123 for mean momentum @xmath124 , mean position @xmath125 , and position spread @xmath126 . the mean momentum is taken to be @xmath127 ( except for case 0 ) such that the mean incident velocity is @xmath128 , and the mean energy is @xmath129 . we first consider the case of @xmath130 , @xmath131 , @xmath132 , @xmath133 , and @xmath134 . the probability density is plotted ( figure [ fig : free1 ] ) , and the trajectories for a random gaussian distribution are computed ( figure [ fig : free2 ] ) . since @xmath135 by our choice of units , the initial wavefunction is confined to a width smaller than the compton wavelength causing two packets to move out near the speed of light . the _ zitterbewegung _ motion visible in the trajectories is due to the interference between positive and negative energy states . in general the trajectories may converge , but do not cross ; this is expected by uniqueness from the first order guiding equation in one spatial dimension . next we consider case 1 : @xmath136 ( figure [ fig : step1 ] & [ fig : step2 ] ) taking @xmath137 and @xmath138 . similar to non - relativistic tunneling,@xcite particles starting closer to the step continue through it , and particles starting further from the step may never reach it ( and are reflected even though @xmath139 ) . this must be the case , again by uniqueness . there is a clear bifurcation point between reflected & transmitted trajectories based on initial position . we now consider case 2 : @xmath140 ( figure [ fig : decay1 ] & [ fig : decay2 ] ) with parameters @xmath141 and @xmath138 . the potential is chosen such that the negative energy continuum is not opened up . the solution corresponds to an exponential decay for @xmath47 , and therefore total reflection which can be seen in the trajectories . lastly we consider case 3 : the klein paradox with @xmath142 ( figure [ fig : klein1 ] & [ fig : klein2 ] ) for parameters @xmath143 and @xmath138 . it can be seen from the probability distribution ( figure [ fig : klein1 ] ) that the reflected packet is larger than the incident packet . the probability distributions can be integrated at initial time @xmath144 , and final time @xmath145 : @xmath146 @xmath147 and the probability conservation @xmath148 with respect to affine parameter @xmath16 can be verified numerically . the initial conditions for the particle distribution are on the @xmath144 and @xmath145 time slices , and correspond to a random gaussian centered on the packets . it can be seen from the trajectories ( figure [ fig : klein2 ] ) that all incident particles are reflected , and all particles starting at @xmath145 form a `` _ _ v _ _ '' shape corresponding to pair creation at the boundary . by the choice of mean energy and barrier height , the anti - particles in region ii move at a greater velocity than the particles in region i. a similar numerical analysis is performed on the potential barrier ( cases 1 - 3 ) taking @xmath138 . we choose @xmath174 such that the mean incident velocity is @xmath175 , and the mean energy is @xmath176 . the probability distributions are integrated , and the conservation equation @xmath177 is verified numerically . we start with case 1 : @xmath139 ( figure [ fig : barrier_step1 ] & [ fig : barrier_step2 ] ) for @xmath178 and barrier width @xmath179 . two reflected packets are visible from scattering off both sides of the barrier . the initial conditions on the @xmath144 time slice correspond to a random gaussian distribution centered around the packet . the particles slow down inside of the barrier , and reflection still occurs with small probability even though @xmath139 . next we consider case 2 : @xmath180 ( figure [ fig : barrier_decay1 ] & [ fig : barrier_decay2 ] ) with @xmath141 and barrier width @xmath181 . the small width is chose such that a non - negligible packet tunnels through the barrier . this most closely resembles non - relativistic tunneling through a potential barrier@xcite since the solution is exponentially decaying inside region ii . lastly we perform a simulation for case 3 : @xmath182 ( figure [ fig : barrier_klein1 ] & [ fig : barrier_klein2 ] ) taking @xmath143 and barrier width @xmath183 . the fabry - perot pattern described in section [ section : barrier ] is visible as ripples in the probability density . while the packet inside the barrier appears to contradict the initial wavefunction on the @xmath144 time slice , it should be kept in mind that this is due to the finite size of the box used for the simulation . the packet that travels backwards in time in region ii diminishes in size as @xmath93 such that the initial wavefunction as @xmath184 is a gaussian packet in region i. for the purposes of numerical verification of probability conservation , we must add : @xmath185 corresponding to conservation with respect to affine parameter @xmath16 . by increasing the box size and moving the initial packet further from the barrier , @xmath186 and equality is approached . the particles which tunnel through the barrier appear to emerge in region iii _ before _ the particle in region i reached the potential boundary . according to the interpretation , this is due to the particle going backwards in time inside region ii . in the lab frame , this corresponds to a particle pair being produced at the right side of the barrier ; the particle goes off to @xmath90 , and the anti - particle annihilates the incident particle on the left side of the boundary at a later time . as previously stated , particle trajectories can not cross by uniqueness . it follows that a particle going backwards in time inside region ii can not tunnel back into region i and continue to @xmath88 . instead it can be seen that incident particles are reflected well before reaching the barrier , and form distinct bands . by tracing the reflected trajectories of figure [ fig : barrier_klein2 ] backwards towards the barrier , one sees that they agree with the spacetime points where particles would have tunneled back into region i after multiple internal reflections in region ii ( analogous to the interferometer ) . this is due to the fact that the wavefunction _ can _ cross itself resulting in the fabry - perot pattern . since the particles are _ guided _ by the wave , this pattern is therefore visible in the trajectories.@xcite similar behavior occurs for the transmitted particles when tunneling out into region iii . the interpretation therefore produces the asymptotic behavior of the fabry - perot interferometer analogy . now we apply the interpretation to the potential barrier.@xcite the potential is given by : @xmath149 with @xmath150 and @xmath151 . defining @xmath45 , the solution in region i is : @xmath46 defining @xmath51 , in region ii : @xmath152 @xmath153 where @xmath154 corresponds to the time - reversed solution in region ii . continuity at the boundary @xmath81 yields : @xmath54 @xmath155 the solution in region iii is given by : @xmath156 with boundary condition @xmath157 such that : @xmath158 we can then compute @xmath56 & @xmath57 as : @xmath159 and the reflection & transmission probabilities : @xmath160 ( 1-\kappa^2)^2 } { \left ( 1+\kappa \right)^4+\left ( 1-\kappa \right)^4 - 2 \ \text{cos}[2 k l ] ( 1-\kappa^2)^2 } \\ \frac{|t|^2}{|a|^2 } = \frac{16 \kappa^2}{(1+\kappa)^4+(1-\kappa)^4- 2 \ \text{cos}[2 k l ] ( 1-\kappa^2)^2 } \end{split}\ ] ] from which it follows @xmath161 . using eq . ( [ eq : dirac_guiding ] ) , the velocity in region i is : @xmath162 similarly in region iii we have : @xmath163 where the solutions correspond to positive energy states in regions i & iii . in region ii : @xmath164 after changing @xmath165 for the negative energy state . the top signs corresponds to @xmath166 , and bottom to @xmath167 . in regions i & iii , @xmath168 as @xmath94 such that the distribution of initial position is at @xmath87 & @xmath88 . in region ii , @xmath169 as @xmath94 ; since there is no wave at @xmath89 in region ii,@xcite there is nt a distribution of initial positions at @xmath89 as we saw in the klein step . this implies the entire distribution is in region i at @xmath87 & @xmath88 . we do nt encounter the feature of the klein paradox ( with an `` initial '' distribution of particles at @xmath89 ) regarding probability since @xmath161 , but there is still finite transmission as @xmath60 . this again suggests that there will be pair creation at the right side of the barrier even if @xmath60 . as pointed out by thomson and mckellar,@xcite there is an analogy between the fabry - perot interferometer , and the incident wavefunction that enters region ii and moves backwards in time . the packet in region ii will diminish as smaller packets tunnel out of the barrier and move forward in time ( the limit of multiple internal reflections is studied in the appendix ) . this results in a fabry - perot pattern in regions i & iii at @xmath170 . as we will see in section [ section : barrier_sim ] , this pattern is visible in the particle trajectories . a comment should be made on the choice of words used to describe this process . the wave packet `` traveling backwards in time '' is used as an interpretational description similar to a particle going `` forwards / backwards '' in time . like the particle worldline , the wavefunction exists over all of spacetime , and therefore should nt be thought of as _ dynamically _ evolving in spacetime . it is equivalent to think of the wavefunction evolving forward in time from the initial conditions set at @xmath171 , but via the time reversed solution in region ii . the _ initial _ condition in region ii at @xmath172 can therefore be viewed as a _ final _ condition being evolved backwards to @xmath173 . we have shown that a consistent explanation can be given for a single relativistic particle process by combining the pilot wave and feynman - stueckelberg interpretations . the interpretation can be applied to scattering off a potential barrier as well as the klein paradox . in this context , the original statement of the paradox then becomes a statement about pair production at the boundaries of high potentials . it would seem that the klein step still has an interesting feature , even with this interpretation : the particle s initial condition is distributed with non - zero probability at a future time slice . this is due to the infinite extent of the potential @xmath31 yielding a negative energy solution at @xmath90 - something that does not occur in the potential barrier . for both cases the wave traveling backwards in time interferes with itself ; this causes constructive interference for the reflected packet in the klein step , and a fabry - perot pattern for the klein barrier . future work can include the application of this interpretation to bosons where the dirac sea picture breaks down , as well as a cohesive interpretation of a many particle process . ascribing a continuous trajectory to the particle in this deterministic interpretation provides an intuitive perspective on particle creation / annihilation , and a resolution to the historic paradox . we show for the case of the klein barrier that a packet going backwards in time inside region ii diminishes in size as @xmath168 . this is done by showing the probability of an arbitrarily large number of internal reflections approaches zero . the probability of an incident packet being reflected without entering region ii is @xmath187 . the probability of a transmitted packet not undergoing any internal reflection is @xmath188 . for each higher order @xmath189 of asymptotic reflection @xmath190 or transmission @xmath191 probability ( corresponding to @xmath192 internal reflections inside region ii ) , we multiply by @xmath193 : @xmath194 @xmath195 where the number of internal reflections is @xmath196 for asymptotically reflected packets , and @xmath197 for asymptotically transmitted packets . we can solve for @xmath193 as : @xmath198 \right)^2\ ] ] for the probability of an arbitrary number of internal reflections to approach zero , we require : @xmath199 since @xmath200 . using @xmath150 , @xmath182 , and being careful with signs : @xmath201 we conclude the packet traveling backwards in time inside region ii decays as @xmath168 .

the de broglie - bohm pilot wave interpretation of quantum mechanics is shown to provide a consistent explanation for a single relativistic particle ( more accurately , a single particle _ process _ since pair production is addressed ) . this is accomplished by incorporating the feynman - stueckelberg interpretation of anti - matter . the lack of conserved probability and existence of negative energy solutions is studied through this interpretation . we discuss the resolution of klein s original paradox through the pilot wave interpretation , and similarly analyze the potential barrier problem . simulations are presented for a wave packet incident on potential steps & barriers of various heights .

a number of studies over the past ten years have estimated the inter - galactic void probability function and investigated its departure from randomness . the basic random model is that arising from a poisson process of mean density @xmath0 galaxies per unit volume in a large box . then , in a _ given region of volume _ @xmath1 the probability of finding exactly @xmath2 galaxies is @xmath3 so the probability that the given region is devoid of galaxies is @xmath4 then it follows that the probability density function for the continuous random variable @xmath5 in the poisson case is @xmath6 for comparison with observations , the approximation fails for very large @xmath5 since a finite volume box is involved in any catalogue . a hierachy of @xmath7-point correlation functions needed to represent clustering of galaxies in a complete sense was devised by white @xcite and he provided explicit formulae , including their continuous limit . in particular , he made a detailed study of the probability that a sphere of radius @xmath8 is empty and showed that formally it is symmetrically dependent on the whole hierarchy of correlation functions . however , white concentrated his applications on the case when the underlying galaxy distribution was a poisson process , the starting point for the present approach which is concerned with geometrizing the parameter space of departures from a poisson process . we choose a family of parametric statistical models that includes ( [ negexp ] ) as a special case . there are of course many such families , but we take one that has been successful in modelling void size distributions in terrestrial stochastic porous media @xcite and has been used in the representation of clustering of galaxies @xcite . the family of gamma distributions has event space @xmath9 parameters @xmath10 and probability density functions given by @xmath11 then @xmath12 and @xmath13 and we see that @xmath14 controls the mean of the distribution while the spread and shape is controlled by , @xmath15 the square of the coefficient of variation . the special case @xmath16 corresponds to the situation when @xmath5 represents the random or poisson process in ( [ negexp ] ) with @xmath17 thus , the family of gamma distributions can model a range of stochastic processes corresponding to non - independent ` clumped ' events , for @xmath18 and dispersed events , for @xmath19 as well as the random case ( cf . thus , if we think of this range of processes as corresponding to the possible distributions of centroids of extended objects such as galaxies that are initially distributed according to a poisson process with @xmath20 then the three possibilities are : chaotic or random structure : : with no interactions among constituents , @xmath21 clustered structure : : arising from mutually attractive interactions , @xmath22 dispersed structure : : arising from mutually repulsive interactions , @xmath23 figure [ fgamma ] shows a family of gamma distributions , all of unit mean , with @xmath24 ( 300,220)(0,0 ) ( 80,0 ) ( 45,177)@xmath25 ( 125,160)@xmath26 ( 180,70)@xmath16 ( 310,50)@xmath27 ( 300,-7)void volume @xmath5 shannon s information theoretic ` entropy ' or ` uncertainty ' for such stochastic processes ( cf . jaynes @xcite ) is given , up to a factor , by the negative of the expectation of the logarithm of the probability density function ( [ feq ] ) , that is @xmath28 in particular , at unit mean , the maximum entropy ( or maximum uncertainty ) occurs at @xmath20 which is the random case , and then @xmath29 the ` maximum likelihood ' estimates @xmath30 of @xmath31 can be expressed in terms of the mean and mean logarithm of a set of independent observations @xmath32 these estimates are obtained in terms of the properties of @xmath33 by maximizing the ` log - likelihood ' function @xmath34 with the following result @xmath35 where @xmath36 and @xmath37 is the digamma function , the logarithmic derivative of the gamma function the usual riemannian information metric on the parameter space @xmath38 is given by @xmath39 for more details about the geometry see @xcite . the 1-dimensional subspace parametrized by @xmath16 corresponds to the available ` random ' processes . a path through the parameter space @xmath40 of gamma models determines a curve @xmath41\rightarrow { \cal{s } } : t\mapsto ( c_1(t),c_2(t))\ ] ] with tangent vector @xmath42 and norm @xmath43 given via ( [ gammametric ] ) by @xmath44 the information length of the curve is @xmath45 and the curve corresponding to an underlying poisson process has @xmath46 so @xmath47 and @xmath48 and the information length is @xmath49 as we know from elementary geometry , arc length is often difficult to evaluate analytically because it contains the square root of the sum of squares of derivatives . accordingly , we sometimes use the ` energy ' of the curve instead of length for comparison between nearby curves . energy is given by integrating the _ square _ of the norm of @xmath50 @xmath51 so in the case of the curve @xmath46 the energy is @xmath52 it is easily shown that a curve of constant @xmath14 has @xmath53 where @xmath54 and @xmath55 this has energy @xmath56 locally , minimal paths joining nearby pairs of points in @xmath40 are given by the autoparallel curves or geodesics @xcite defined by ( [ gammametric ] ) . some typical sprays of geodesics emanating from various points are provided in @xcite . the gaussian curvature of the surface @xmath40 @xcite actually controls all of the geometry of geodesics and it is given by @xmath57 @xmath58 for a general account of large - scale structures in the universe , see fairall @xcite . kauffmann and fairall @xcite developed a catalogue search algorithm for nearly spherical regions devoid of bright galaxies and obtained a spectrum for diameters of significant voids . this had a peak in the range 8 - 11 @xmath59 a long tail stretching at least to 64 @xmath59 and is compatible with the recent extrapolation models of baccigalupi et al @xcite which yield an upper bound on void diameters of about 100 @xmath60 we shall return to the data of kauffmann and fairall later in this section . simulations of sahni et al . @xcite found strong correlation between void sizes and primordial gravitational potential at void centres ; void topologies tended to simplify with time . ghigna et al . @xcite found in their simulations that void statistics are sensitive to the passage from cdm to chdm models . this suggested that the void distribution is sensitive to the type of dark matter but not to the transfer function between types . chdm simulations gave a void probability in excess of observations , cdm simulations being somewhat better . vogeley et al . @xcite compared void statistics with cdm simulations of a range of cosmological models ; good agreement was achieved for samples of very bright galaxies ( @xmath61 ) but for samples containing fainter galaxies the predicted voids were reported to be ` too empty ' . ghigna et al . @xcite compared observational data with gaussian - initiated @xmath7-body simulations in a @xmath62 box and found that at the @xmath63 scales the void probability for chdm was significantly larger than observed . ghigna et al . @xcite compared simulated galaxy samples with the perseus - pisces redshift survey . the void probability function did discriminate between dm and cdm models , the former giving particularly good agreement with the survey . little and weinberg @xcite used similar @xmath7-body simulations , and found that the void probability was insensitive to the shape of the initial power spectrum . watson and rowan - robinson @xcite found that standard cdm predictors do yield reasonably good void probability function estimates whereas voronoi foam models performed less well . lachize - rey and dacosta were of the opinion that the available samples of galaxies were insufficiently representative , because of greater apparent frequency of larger voids in the southern hemisphere . bernardeau @xcite started from a gaussian field and derived an expression for the void probability function , obtaining @xmath64 with @xmath65 the variance of the number of galaxies in volume @xmath66 this distribution has an extended large tail because it is asymptotically more like the exponential of @xmath67 than the poisson case which decays like the exponential of @xmath68 cappi et al . @xcite examined the dependence of the void probability function on scale for a range of galaxy cluster samples , finding a general scaling to occur up to void diameters of about @xmath69 kerscher et al . @xcite used the void probability function to obtain spatial statistics of clusters on scales @xmath70 obtaining satisfactory agreement in a model with a cosmological constant and in a model with breaking of scale invariance of perturbations . ( 300,220)(0,0 ) ( 80,0 ) ( 43,160)@xmath71 ( 300,-7)void diameter @xmath72 for our model we consider the diameter @xmath72 of a spherical void with volume @xmath73 having distribution ( [ feq ] ) . something close to the random variable @xmath72 has direct representation in some theoretical models , for example as polyhedral diameters in voronoi tesselations @xcite . ( 300,220)(0,0 ) ( 50,0 ) the probability density function for @xmath72 is given by @xmath74 then the mean @xmath75 variance @xmath76 and coefficient of variation @xmath77 of @xmath72 are given , respectively , by @xmath78 the fact that the coefficient of variation ( [ cvd ] ) depends only on @xmath79 gives a rapid fitting of data to ( [ dpdf ] ) . numerical fitting to ( [ cvd ] ) gives @xmath80 this substituted in ( [ meand ] ) yields an estimate of @xmath14 to fit a given observational mean . by way of illustration , this has been done in figure [ kfzcat ] for the zcat / src data from kauffmann and fairall @xcite figure 8a ( cf also @xcite figure 6.5 ) , both set to unit mean diameter the true mean for that catalogue was about @xmath81 the fitted values were @xmath82 and @xmath83 this fit is not particularly good if the reported peaks are not an artifact but , qualitatively , we observe that the fitted value @xmath84 is apparently significantly less than @xmath85 which would correspond to the random model . we conclude tentatively that , for the zcat / src data subjected to the kauffmann and fairall search algorithm , the new model suggests clumping rather than dispersion in the underlying stochastic process . a _ mathematica _ notebook for performing the fitting procedure is available from the author , who would welcome more sets of data . suppose that the parameter space @xmath40 of gamma - based models is a meaningful representation of the evolutionary process and that some coordinates such as @xmath86 @xmath87 in @xmath40 represent current data . then geodesics in @xmath40 through this point represent some kind of extremal path . moreover , we may consider a vector field @xmath88 on @xmath40 such that @xmath86 @xmath87 is the present endpoint of an integral curve of @xmath89 the initial point of this curve being presumably in an epoch when less clustering ( @xmath90 ) , a random state ( @xmath16 ) or even dispersion ( @xmath91 ) was present at higher density . it would be interesting to investigate the various candidate cosmologies for their appropriate vector fields , via the statistics of matter and voids they predict . the present gamma - related parametric statistical models provide the means to convert the catalogue statistics into the coordinate parameters and a background geometrization of the statistics on which dynamical processes may be formulated . the author wishes to thank a.p . fairall , f. labini and m.b . ribeiro for helpful comments during the preparation of this article . c.t.j . dodson . gamma manifolds and stochastic geometry . in : * proceedings of the workshop on recent topics in differential geometry * , santiago de compostela 16 - 19 july 1997 . _ public . geometra y topologa _ 89 ( 1998 ) 85 - 92 . m. kerscher , j. schmalzing , j. retzlaff , s. borgani , t. buchert , s. gottlober , v. muller , m. plionis and h. wagner . minkowski functionals of abell / aco clusters . _ mon . not . _ 284 , 1 ( 1997 ) 73 - 84 . t. piran , m.lecar , d.s . goldwirth , l. nicolaci da costa and g.r . limits on the primordial fluctuation spectrum : void sizes and anisotropy of the cosmic microwave background radiation . _ mon . not . _ 265 , 3 ( 1993 ) 681 - 8 .

a number of recent studies have estimated the inter - galactic void probability function and investigated its departure from various random models . we study a family of parametric statistical models based on gamma distributions , which do give realistic descriptions for other stochastic porous media . gamma distributions contain as a special case the exponential distributions , which correspond to the ` random ' void size probability arising from poisson processes . the random case corresponds to the information - theoretic maximum entropy or maximum uncertainty model . lower entropy models correspond on the one hand to more ` clustered ' structures or ` more dispersed ' structures than expected at random . the space of parameters is a surface with a natural riemannian structure , the fisher information metric . this surface contains the poisson processes as an isometric embedding and provides the geometric setting for quantifying departures from randomness and perhaps on which may be written evolutionary dynamics for the void size distribution . estimates are obtained for the two parameters of the void diameter distribution for an illustrative example of data published by fairall . + * keywords * gamma distribution , void distribution , randomness , information geometry + msc classes : 85a40 ; 60d05

in qcd with two flavors at high baryon density the superconducting phase is the lowest energy state @xcite . in this phase @xmath0 quarks with two different colors ( say `` red '' , `` green '' ) create cooper pairs whereas the `` blue '' quarks remain free . the chiral symmetry @xmath1 is restored and only gauge color group is broken . on the other hand at lower density one expects the state with broken chiral symmetry . this can be a hadronic phase but the free quark plasma is not ruled out . in such a situation the sequence of the phase transitions along the density line ( at zero or very low temperature ) follows the chain : hadronic phase / quark phase / superconducting phase . all the phase transitions are expected to be first order . in this letter we present an explicit model describing the transition between the normal quark and the superconducting phases . it occurs that the transition actually goes through the mixed state which consists of the non - uniform chiral phase and superconducting phase . the non - uniform chiral phase had been introduced in the context of the non - relativistic nuclear physics @xcite and then was extended for the systems of high density plasma of quarks @xcite . in this new phase the chiral fields condense in both scalar and pseudoscalar channels creating a static chiral wave along one direction in space with the wavelength @xmath2 where @xmath3 is a wave vector . it was shown that in some temperature - density range this configuration has lower energy then the uniform quark phase . on the other hand we have learned from the studies @xcite that at higher densities one should compare the non - uniform quark phase with the superconducting phase rather then with the uniform quark phase . as a result one can show that the phase diagram of strongly interacting matter is more complicated and contains a region of the mixed state . the high density matter exists in the neutron stars . if the density is high enough then the superconducting phase resides in their cores . in this situation one can also expect the mixed phase to be present there - the mixed state which separates uniform chiraly broken phase from the superconducting phase . we can not prove that this mixed state is the same as the one considered in this paper but we provide an explicit realization of such a scenario . in the next section we introduce a thermodynamic potential describing all the interesting phases . in the third section we present the numerical analysis and finaly we discuss the results in the conclusions . let us consider the njl model @xcite : @xmath4 -g^\prime ( \psi\tau_2t^ac\gamma_5\psi ) ^\dagger ( \psi\tau_2t^ac\gamma_5\psi ) \right\}\ ] ] where @xmath5 is the quark field , @xmath6 the quark chemical potential and the color , flavor and spinor indices are suppressed . the vector @xmath7 is the isospin vector of pauli matrices and @xmath8 are three color antisymmetric group generators . the coupling constant @xmath9 describes the interaction in the isospin singlet , lorentz scalar and the isospin triplet , lorentz pseudoscalar , quark - antiquark channels whereas @xmath10 describes the interaction in the color @xmath11 , flavor singlet , lorentz scalar diquark channel . both couplings are related through the fierz transformation but we treat them as the independent parameters . there is an additional parameter ( or function ) which regularize infinities in the model . in our case we use the simplest momentum cut - off @xmath12 . our main results are independent of this choice . we analize the four - fermion point interactions in the mean field approximation using the general anstaz : @xmath13 where the directions of symmetry breaking in the isospin and color spaces are chosen arbitrarily . the wave vector @xmath3 describes the chiral wave along some arbitrary direction in space . in the limit of vanishing @xmath3 we recover usual uniform chiral phase . the gap parameter @xmath14 describes the superconducting phase . let us notice that ( [ ansatz ] ) is not the analog of the loff phase in the superconductivity @xcite which arises due to the difference in the chemical potentials of the pairing species . in our case it is rather the dynamical interaction between the spin and isospin degrees of freedom which effectively lower the energy @xcite . using formulae ( [ h ] ) , ( [ ansatz ] ) in the mean - field approximation we get : @xmath15 where the dirac operator is now space dependent @xmath16 . one can eliminate this dependence by the unitary transformation to the new fermi fields @xmath17 which gives the free energy in the form : @xmath18 where @xmath19 is the vector of dirac spin matrices . the hamiltonian ( [ th_mf ] ) describes two kinds of interactions : the `` chiral '' part and the `` superconducting part '' . the chiral part mixes particles and antiparticles leaving color and isospin structure untouched . on the other hand the `` superconducting '' part mixes colors and isospins of particles . to be more precise , one can introduce the plane wave basis @xmath20 where @xmath21 , @xmath22 are dirac bispinors , @xmath23 @xmath24 is an annihilation operator of quark ( antiquark ) of color @xmath25 , flavor @xmath26 , spin @xmath27 , and momentum @xmath28 , satisfying usual anticommutation relations . in this basis hamiltonian ( [ th_mf ] ) creates the @xmath29 matrix which decays into three independent diagonal @xmath30 blocks . two of them are related to pairing between `` red '' , `` green '' @xmath0 quarks and antiquarks whereas the third one describes the pairing between `` blue '' quarks and antiquarks ( which do not create cooper pairs ) . the hamiltonian ( [ th_mf ] ) can be diagonalized with the positive energy eigenvalues : @xmath31 for @xmath0 , `` red '' , `` green '' quarks and @xmath32 for `` blue '' @xmath0 quarks . the negative eigenvalues has opposite signs . after diagonalization the hamiltonian ( [ th_mf ] ) takes the general form @xmath33 . the first term is a diagonal operator and the second term defines the energy of the ground state as a function of chemical potential . the vacuum expectation value @xmath34 and @xmath35 the first integral describes the contribution from the quarks which create the chiral condensate as well as cooper pairs . the factor 2 is the number of colors minus one ( `` red '' , `` green '' ) . the parameter @xmath12 reminds of the momentum cut - off of the divergent integral . the last two terms are connected to `` blue '' quarks which build only chiral condensate . the factor of 2 , this time , describes the number of flavors . let us rewrite ( [ h0 ] ) in the form : @xmath36 which distinguishes explicitly between the vacuum contribution - the first integral , and the finite density contributions - the last two integrals . the vacuum contribution can be expanded in the power of the wave vector @xmath3 ( nothing more but the gradient expansion in mesonic fields ) and we arrive at the final result : @xmath37 the coefficient at @xmath38 is model independent and it is related to the pion decay constant with its vacuum value @xmath39 mev . @xmath40 is the constituent quark mass at zero density . let us notice that in the limit of vanishing @xmath41 we recover the chiral hamiltonian @xcite whereas in the limit of vanishing @xmath3 we can find the hamiltonian describing cooper pairing @xcite . let us also mention that from the formula ( [ h03 ] ) one can easily guess the linear sigma model version of the hamiltonian . indeed the njl and linear sigma models at the mean - field level are different only with respect to the description of the vacuum contribution to the energy . thus instead of the formula ( [ h03 ] ) one can write in the linear sigma model : @xmath42 where @xmath43 is dimensionless coupling constant and @xmath44 is a mass of sigma meson . however in the numerical calculation we only use the formula ( [ h03 ] ) . the model parameters can be fitted to the values of the chiral condensate and the pion decay constant which give @xmath45 gev@xmath46 and the cut - off @xmath47 gev @xcite . for this set of parameters the chiral phase transition at @xmath48 and @xmath49 takes place at the chemical potential @xmath50 gev and the value of the constituent quark mass at zero density is @xmath51 mev . the coupling constant @xmath10 we treat as the additional model parameter . the main results which are robust against the change in @xmath10 are : * there is the first order phase transition from the homogeneous chiraly broken phase @xmath52 , @xmath48 , @xmath53 to the mixed phase @xmath54 , @xmath55 , @xmath56 . * by increasing chemical potential one decreases constituent mass @xmath57 and increases @xmath3 and @xmath41 . * there is additional first order phase transition from the mixed phase to the 2sc phase ( @xmath58 , @xmath48 , @xmath56 ) . from the above points one can see that the 2sc phase is separated from the chiraly broken phase by the the non - uniform mixed state . this conclusion is still valid for non - zero temperatures . we discuss this feature in the last section . on the other hand the numerical details depend on the choice of @xmath10 . in particular for the range @xmath59 we can find : * the transition to the mixed phase appears in the band of 10 mev around @xmath60 mev ( smaller @xmath61 larger critical chemical potential ) . * the values of the gap parameter @xmath41 can vary by a factor of two ( smaller @xmath61 larger @xmath41 ) . * the values of the wave vector @xmath62 can change by 40 - 50 per cent ( smaller @xmath61 larger @xmath63 ) . * the range of the mixed phase along @xmath6-axis can change from 10 to 100 mev ( smaller @xmath10 longer the range ) . in the fig . 1 we present the values of @xmath57 , @xmath62 , @xmath41 for @xmath64 as a function of chemical potential @xmath6 . the first order phase transition takes place at @xmath65 gev . the constituent mass drops to around half of its vacuum value ( fig . 1a ) , the wave vector @xmath66 gev ( fig . 1b ) and @xmath67 gev ( fig . 1c ) . the mixed phase exists up to the second critical point @xmath68 gev where another first order phase transition appears . the values of @xmath57 , @xmath3 drop to zero and the value of the gap @xmath69 gev . for larger @xmath70 we enter the homogeneous 2sc phase . = 5.5 cm = 5.5 cm = 5.5 cm in this letter we showed that the high density superconducting phase 2sc is separated from the low density chiraly broken phase by the region of the mixed non - homogeneous phase . this mixed state consists of the chiral waves with non - zero expectation values in the scalar and pseudoscalar channels @xmath71 described by the wave vector @xmath3 which points at arbitrary direction is space . in addition there is a non - zero value of the superconducting gap parameter @xmath41 . we can not prove that this mixed state is the lowest energy state however we showed that its energy is lower then that of the homogeneous phases . similar questions was already addressed in the context of the overhauser effect @xcite versus bcs pairing . it was shown that both types of pairing can be competitive @xcite with each other . however it was also shown that within the njl - type of models bcs state is favoured over the overhauser state @xcite which suggests that the non - homogeneous phase ( [ ansatz ] ) is lower then bcs and overhauser phases in the considered density range . nevertheless one has to keep in mind that other approaches ( instanton models ) favours overhauser effect over bcs pairing @xcite thus one can not drew the final conclusion and more work is required . we can then conclude that the superconducting phase and the normal phase are separated by some kind of crystal state . let us stress that the crystal structure ( [ ansatz ] ) is dynamical in origin , not `` kinematical '' in a sense of the loff state @xcite . our results are robust against the reasonable changes in the model parameters . what is more the same results one can derive not only in the njl model but also in the linear sigma model . thus also in that sense the results are model independent . the numerical details , like the location of the critical points , the values of the constituent mass , wave vector and superconducting gap , depends on the choice of the parameters . although we perform our calculations for massless quarks the introduction of the small current quark mass does not change the main conclusions . it merely distorts the shape of the chiral wave as was shown in @xcite for one dimensional system . one can also expect that the mixed phase exists for some range of temperatures ( below @xmath72 mev where superconductor melts ) and densities . however the detailed picture of the phase diagram remain to be done . the real qcd consists of two light flavors and one heavier , strange quark . in this situation one has to consider the three flavors njl model . however we expect that our main points remain the same . if one considers the superconducting phase inside the neutron stars then the question of the color and charge neutrality rises @xcite . it was shown by the explicit calculation @xcite that at non - zero lepton chemical potential the 2sc phase borders with the normal quark phase at lower densities and only at higher densities it changes into the cfl phase . thus even in the case of the three flavors qcd we can still have quark matter/2sc boundary in the neutron stars . this boundary is not just a line in the phase diagram but the mixed state region as we have shown . what kind of the crystal or non - uniformity is realized in the three flavors qcd has not been discoverd so far . * acknowledgement * this work was supported by a fellowship from the foundation for polish science . it was also supported in part by polish state committee for scientific research , grant no . 2p 03b 094 19 . 99 m. alford , k. rajagopal and f. wilczek , phys . b422 * ( 1998 ) 247 . r. rapp , t. schafer and e. v. shuryak , phys . * 81 * ( 1998 ) 53 . f. dautry , e. m. nyman , nucl . phys . * a319 * ( 1979 ) 323 . m. kutschera , w. broniowski and a. kotlorz , nucl . phys . * a516 * ( 1990 ) 566 ; phys . lett . * b237 * ( 1990 ) 159 . m. sadzikowski and w. broniowski , phys . lett . * b488 * ( 2000 ) 63 . m. sadzikowski , mod . * a16 * ( 2001 ) 1129 . a. i. larkin and yu . n. ovchinnikov zh . * 47 * ( 1964 ) 1136 ; p. fulde and r. a. ferrell , phys . bf 135 ( 1964 ) a550 . m. alford , j. berges and k. rajagopal , nucl . * b558 * ( 1999 ) 219 ; t. schfer and f. wilczek , phys . rev . * d60 * ( 1999 ) 074014 . s. p. klevansky , rev . * 64 * ( 1992 ) 64 . d. v. deryagin , d. yu . grigoriev and v. a. rubakov , int . * a7 * ( 1992 ) 659 . y . park , m. rho , a. wirzba and i. zahed , phys . rev . * d62 * ( 2000 ) 034015 . r. rapp , e. shuryak and i. zahed , phys . * d63 * ( 2001 ) 034008 . v. schon , m. thies , in m. shifman ( ed . ) _ at the frontier of particle physics _ , vol . 3 1945 - 2032 ( hep - th/0008175 ) . m. alford and k. rajagopal , jhep 0206:031 , 2002 ( hep - ph/0204001 ) . a. w. steiner , s. reddy and m. prakash , suny - ntg-02 - 05 - 25 ( 2002 ) ( hep - ph/0205201 ) .

we show that the superconducting 2sc phase at high density and normal chiraly broken quark phase at low density is separated by the mixed non - uniform phase along the baryon density line .

our motivation comes from three seemingly unrelated problems . the first is the problem of _ floating bodies of equilibrium in two dimensions_. from 1935 to 1941 , mathematicians at the university of lviv , among them stefan banach and mark kac , collected mathematical problems in a book , which became known as the scottish book , since they often met in the scottish coffee house . stanislaw ulam posed problem 19 of this book : `` is a sphere the only solid of uniform density which will float in water in any position ? '' the answer in the two - dimensional case , as it turns out , depends on the density of the solid . the second problem , known as the _ tire track problem _ , originated in the story , `` the adventure of the priory school '' by arthur conan doyle , where sherlock holmes and dr . watson discuss in view of the two tire tracks of a bicycle which way the bicycle went . the problem is : `` is it possible that tire tracks other than circles or straight lines are created by bicyclists going in both directions ? '' as shown in figure [ fig : ambitrack ] , the answer to this subtle question is affirmative . the third problem is the problem of describing the trajectories of _ electrons in a parabolic magnetic field_. all three problems turn out to be equivalent @xcite . ambiguous bicycle tracks . the rear - wheel track is the inner curve and the front - wheel track is the outer curve . one can not tell which way the bicycle went because a bicycle could have followed either one of two trajectories @xcite . , width=384 ] often in mathematics it is fruitful to discretize a problem . as such , s. tabachnikov proposed a `` discrete bicycle curve '' ( also known as a `` bicycle polygon '' ) @xcite , which is a polygon satisfying discrete analogs of the properties of a bicycle track . the main requirement turns out to be that , in the language of discrete differential geometry , the polygon is `` self - darboux '' . that is , the discrete differential geometric notion of a discrete darboux transformation @xcite , which relates one polygon to another , relates a discrete bicycle curve to itself . the topic of bicycle curves and polygons belongs to a number of active areas of research . on the one hand , it is part of rigidity theory . as an illustration , r. connelly and b. csiks consider the problem of classifying first - order flexible regular bicycle polygons @xcite . other work on the rigidity theory of bicycle curves and polygons can be found in @xcite . the topic is also part of the subject of discrete integrable systems . this point of view is taken in @xcite , where the authors find integrals of motion ( i.e. , quantities which are conserved ) of bicycle curves and polygons under the darboux transformation and recutting of polygons @xcite . in this paper , in analogy with bicycle polygons , we introduce a new concept called a periodic discrete bicycle path and study both its rigidity and integrals . a bicycle @xmath0-gon is an equilateral @xmath1-gon whose @xmath2 diagonals are of equal length @xcite . we consider the following analog . [ defbikepath ] define @xmath3 ( for brevity , @xmath4 ) to be a _ discrete periodic bicycle @xmath0-path _ ( or discrete @xmath0-path ) if the following conditions hold : 1 . @xmath5 , @xmath6 . ( periodicity condition ) 2 . ( equilateralness ) 3 . ( equality of @xmath2-diagonals ) an example of a discrete @xmath9-path.,width=384 ] definition [ defbikepath ] is meant to model the motion of a bicycle whose trajectory is spatially periodic . if the back wheel of the bike is at @xmath10 then the front wheel is at @xmath11 , and the condition that @xmath12 is independent of @xmath13 represents the rigidity of the frame of the bike . the condition that @xmath14 is independent of @xmath13 prescribes a constant speed for the motion of the bike . some natural questions regarding periodic @xmath0-paths are for which pairs @xmath0 they exist , how many there are and whether they are rigid or flexible . we consider these questions in section [ sec : rig ] . a simple example of a bicycle @xmath0-path , analogous to the regular @xmath0-polygon , is @xmath15 , i.e. when all vertices lie at equal intervals on the line . we call this the _ regular path_. since bicycle @xmath0-paths are discretized bicycle paths , it is also interesting to see if there are any integrals of motion . we show that this is indeed the case in section [ sec : integrals ] , by showing that area is an integral of motion . the following two lemmas will be helpful when analyzing the rigidity of discrete bicycle paths . [ lem : cyclic absolute values]let @xmath16 , @xmath17 for every @xmath18 and let @xmath19 then @xmath20 in particular , if @xmath1 is odd , then @xmath21 . first note that the candidate set is well - defined since @xmath22 . let @xmath23 . recall that @xmath24 and that @xmath25 . set @xmath26 and @xmath27 . then @xmath28 and @xmath29 it follows that any @xmath1-tuple in @xmath30 satisfies the conditions @xmath31 , @xmath32 and @xmath33 . the opposite inclusion is clear . [ lem : difference equation]let @xmath35 for every @xmath36 with @xmath37 and let @xmath2 and @xmath1 be coprime integers . assume that @xmath38 for each @xmath39 and that @xmath40 . then @xmath41 for each @xmath39 . define @xmath42 via @xmath43 . the hypothesis @xmath40 implies that @xmath44 since @xmath2 and @xmath1 are coprime , the difference @xmath45 is independent of @xmath39 . if @xmath46 is the order of @xmath2 in @xmath47 , then @xmath48 so that @xmath49 for every @xmath39 . @xmath34 the following theorem gives a classification of a family of periodic @xmath0-paths . [ thm : discretepath1 ] the discrete @xmath50-paths @xmath51 with @xmath52 are exactly those paths which satisfy @xmath53 and for each @xmath13 . in particular , if @xmath1 is odd then a discrete @xmath50-path must be regular . for every @xmath39 , @xmath54 and @xmath55 therefore @xmath56 and @xmath57 it follows that @xmath58 since @xmath52 , @xmath59 by lemma [ lem : difference equation ] , @xmath60 for each @xmath13 . now equation @xmath61 for all @xmath62 implies that @xmath63 so that @xmath64 by lemma [ lem : cyclic absolute values ] , we are done . @xmath34 [ thm : path2 ] the discrete @xmath65-paths @xmath51 with @xmath52 are exactly those paths which satisfy @xmath53 and for each @xmath13 . in particular , if @xmath1 is odd then a discrete @xmath65-path must be regular . set @xmath66 and @xmath67 . then @xmath68 and @xmath69 substituting , we get @xmath70 so that @xmath71 is constant . by lemma [ lem : difference equation ] , @xmath41 . it follows that @xmath72 is constant , so by lemma [ lem : difference equation ] we are done . @xmath34 any @xmath65-path is an @xmath50-path and vice verse . for an example , see figure [ sixfive ] . it is important to make a distinction between infinitesimal `` trapezoidal '' movement and infinitesimal `` parallelogram '' movement of the bicycle . informally , infinitesimal `` parallelogram '' movement can be seen as a trivial gliding of the bike , and thus we insist on infinitesimal `` trapezoidal '' movement . ( trapezoidal condition ) we will say that a discrete @xmath0-path satisfies the trapezoidal condition if @xmath73 for each @xmath74 . as an illustration of these concepts , consider figure [ sixfive ] : @xmath75 is a trapezoidal motion , while @xmath76 is a parallelogram motion . consequently , the bicycle path in the figure does not satisfy the trapezoidal condition . assuming the trapezoidal condition , we may view bicycle paths in terms of an important construction in discrete differential geometry called the darboux transformation @xcite . ( darboux transform ) we say that two polygons @xmath77 and @xmath78 are in darboux correspondence ( with parameter @xmath79 ) if each quadrilateral @xmath80 is an isosceles trapezoid with side length @xmath79 : @xmath81 two polygons in darboux correspondence . , width=288 ] we denote the map taking vertex @xmath82 to @xmath83 by @xmath84 . we will also refer to the map of polygons @xmath85 by the same letter , since no confusion ought to occur . consider a polygonal line @xmath77 with vertices @xmath86 . let @xmath87 be a vector with its origin at @xmath88 . having a vector @xmath89 at vertex @xmath90 , we obtain a vertex @xmath91 of the same length at @xmath92 via the trapezoidal condition . for example , in figure [ dcorrespondence ] , @xmath93 and @xmath94 . for a fixed length of @xmath87 , we may view the map taking @xmath87 to @xmath95 as a self - map of the circle of radius @xmath96 by identifying the circle at @xmath88 with circle at @xmath97 via parallel translation . ( monodromy map of the darboux transformation ) the monodromy map is the map acting on the identified circles at @xmath88 and @xmath98 which takes @xmath87 to @xmath99 . it is known that the monodromy map is a fractional linear transformation on a circle of fixed radius after we identify the circle with the real projective line via stereographic projection @xcite . we will assume throughout , unless otherwise stated , that the monodromy map is acting on a fixed point . in other words , we will assume that the darboux transform has been chosen so that the darboux vector at @xmath88 is the same as that at @xmath98 , where @xmath1 is the period . this is analogous to applying the darboux transform to a closed polygon and requiring that its image is closed also . + we mention in passing that in the case of closed polygons , darboux correspondence implies that the monodromies of the two polygons are conjugated to each other . the invariants of the conjugacy class of the monodromy , viewed as functions of the length parameter are consequently integrals of the darboux correspondence @xcite . * connection between darboux transformation and discrete @xmath0-paths*. a discrete @xmath0-path satisfying the trapezoidal condition may be interpreted in terms of the darboux transform . indeed , given such a path , we consider the periodic equilateral linkages @xmath100 for @xmath101 . the trapezoidal condition implies that there is a darboux correspondence @xmath102 of the same parameter ( since the @xmath0-path is equilateral ) for consecutive linkages ( see figure [ connection ] ) . conversely , given a series of such equilateral linkages in consecutive darboux correspondence , it is easy to see that we may form a bicycle @xmath0-path satisfying the trapezoidal condition . + viewing a discrete @xmath0-path satisfying the trapezoidal condition ( top left ) in terms of the darboux transformation . the path is decomposed into equilateral linkages ( top right ) . any two consecutive linkages are in darboux correspondence ( bottom).,title="fig : " ] viewing a discrete @xmath0-path satisfying the trapezoidal condition ( top left ) in terms of the darboux transformation . the path is decomposed into equilateral linkages ( top right ) . any two consecutive linkages are in darboux correspondence ( bottom).,title="fig : " ] viewing a discrete @xmath0-path satisfying the trapezoidal condition ( top left ) in terms of the darboux transformation . the path is decomposed into equilateral linkages ( top right ) . any two consecutive linkages are in darboux correspondence ( bottom ) . ] the darboux transformation also preserves the area of periodic linkages . more precisely , let @xmath103 for @xmath104 sufficiently large so that the linkage @xmath105 and its darboux transformation @xmath106 lie completely above @xmath103 . we define an area function as follows . let @xmath107 . we define the area of @xmath77 to be the signed area of the polygon @xmath108 and denote it by @xmath109 . we show that this area is preserved under darboux transformation ( see figure [ areafig ] ) . in particular , it will follow that the area of @xmath110 is equal to the area of @xmath111 for every @xmath112 . [ thm : area ] the darboux transformation is area preserving on periodic polygonal paths . two periodic paths @xmath77 and @xmath113 in darboux correspondence . by theorem [ thm : area ] , the two paths have equal areas under the curve.,title="fig:",width=384 ] two periodic paths @xmath77 and @xmath113 in darboux correspondence . by theorem [ thm : area ] , the two paths have equal areas under the curve.,title="fig:",width=384 ] let @xmath77 and @xmath106 be two periodic polygonal paths in darboux correspondence . we show that the difference of the areas of @xmath77 and @xmath106 is zero . we have @xmath114 and similarly for @xmath106 . therefore @xmath115 from the isosceles trapezoids , @xmath116 also , @xmath117 similarly , @xmath118 using equation [ eq : trapezoid triangle area equality ] , @xmath119 it follows that @xmath120 which telescopes to @xmath121 since @xmath122 and @xmath123 , it follows that @xmath124 , so that @xmath125 . @xmath34 we end our discussion with some research topics and questions of interest concerning bicycle @xmath0-paths . 1 . construct interesting families of bicycle @xmath0-paths . for example , ones for which the condition @xmath126 does not hold . 2 . what is the @xmath127th order infinitesimal rigidity theory of bicycle @xmath0-paths like ? 3 . for closed bicycle polygons , there are many integrals of motion @xcite . for example , a geometric center called the circumcenter of mass @xcite is invariant under darboux transformation for closed polygons . are there other integrals of motion for bicycle @xmath0-paths ?

a bicycle @xmath0-gon is an equilateral @xmath1-gon whose @xmath2 diagonals are of equal length . in this paper we consider periodic bicycle @xmath0-paths , which are a natural variation in which the polygon is replaced with a periodic polygonal path .

laser wakefield acceleration ( lwfa ) offers the potential to construct compact accelerators that has a numerous potential applications including the building blocks for a next generation linear collider and being the driver for compact light sources . as a result , lwfa has attracted extensive interest since it was originally proposed @xcite , and the last ten years has seen an explosion of experimental results . due to the strong nonlinear effects that are present in lwfa , developing predictive theoretical models is challenging @xcite ; therefore numerical simulations are critical . in particular , particle - in - cell ( pic ) simulations play a very important role in lwfa research because the pic algorithm follows the self - consistent interactions of particles through the electromagnetic fields directly calculated from the full set of maxwell equations . using a standard pic code to study a 10 gev stage in a nonlinear regime takes approximately 1 million core hours on today s computers and a 100 gev stage would take 100 million core hours . while computing resources now exist to do a few of such simulations , it is not possible to do parameter scans in full three - dimensions . therefore , reduced models such as combining the ponderomotive guiding center with full pic @xcite for the wake or with quasi - static pic @xcite are used for parameter scans . however , while these models are very useful , they can not model full pump depletion distances and the quasi - static approach can not model self - injection . another reduced model that has been recently proposed is to expand the fields in azimuthal mode numbers and truncate the expansion @xcite . this can reduce modeling a 3d problem with low azimuthal asymmetry into the similar computational cost as using a 2d @xmath3 code . recently , it was shown that by performing the simulation in an optimal lorentz boosted frame with velocity @xmath0 , the time and space scales to be resolved in a numerical simulation may be minimized @xcite . the basic idea is that in the boosted frame the plasma length ( the laser propagation distance ) is lorentz contracted while the plasma wake wavelength and laser pulse length are lorentz expanded . the number of laser cycles is an invariant ( assuming there is no reflected wave ) so the necessary number of cells needed to resolve the laser is also an invariant while the cell size and hence time step are lorentz expanded . the increase in time step and decrease in the plasma length lead to savings of factors of @xmath1 as compared to a lab frame simulation using the so called moving window @xcite . using such simulations , it has been shown that using a 13 @xmath4 laser one could generate 10 gev electron beam in a self - guided stage and 50 gev in a channel guided stage @xcite . for these cases the savings can be larger than factors of @xmath5 . however , in the boosted frame lwfa simulations noise from a numerical instability can be an issue . as discussed in @xcite , the noise results from a numerical cerenkov instability induced by the plasma drifting with relativistic speeds on the grid . according to the dispersion relation this numerical instability is attributed to the coupling between the wave - particle resonances with em modes ( including aliased modes ) in the numerical system . the pattern of the instability in fourier space can be found at the intersections of the em dispersion relation of the solver used in the simulation algorithm , and the wave - particle resonances @xcite . in order to mitigate this instability , it is preferable to use an em solver that eliminates the numerical instability at the main beam resonance . in this case , the instability occurs only at high @xmath6 modes which are far away from the physics of interest . as the em dispersion curves for most fdtd solvers inevitably bends down ( i.e. , supports waves with phase velocities less than the speed of light ) at high @xmath6 , numerical instabilities at the main beam resonance are found in these solvers . however , when using a spectral solver that spatially advances the em fields in fourier space , the dispersion curve assures no instability pattern at the main beam resonance . in addition , the pattern at the first space aliasing beam mode is found to indeed be located at high @xmath7 values that are far away from the interested physics . for the spectral solver the numerical cerenkov instability is located at a predicted pattern in @xmath8 space so it can be conveniently eliminated by applying simple filters directly in @xmath9 space . in this paper we describe the development of a fully parallelized three - dimensional electromagnetic spectral pic code called upic - emma that was rapidly built using components of the ucla pic framework ( upic ) @xcite . we demonstrate that through the use of appropriate filters , lorentz boosted frame simulations of lwfa at the optimum frame velocities can be carried out without limitations from the numerical cerenkov instability . we show that a simple low pass filter with a hard cutoff at @xmath6 works very well . this completely eliminates modes with @xmath6 above a selected value . meanwhile , it is not as easy to use such a filter in @xmath6 space using a fdtd solver ( and such solvers have instabilities at lower @xmath10 ) . as discussed in ref . @xcite , when using the fdtd code to simulate relativistic plasma drift , an optimized time step has to be chosen to minimize the instability growth rate . while the instability growth rate is minimized , this time step does not lead to complete elimination of the instability and it can lead to further errors in numerical dispersion . additional smoothing and filtering can help as well , but unlike when using a spectral code the instability can not be essentially eliminated . for the spectral code , the only errors in numerical dispersion arise from the use of finite time step . because , it is not necessary to use an optimum time step ( nor does one exist ) , one can minimize the errors in numerical dispersion for the em waves by choosing smaller time steps if needed . one disadvantage with the spectral code is that it is not easy to use a moving window , however for the optimum @xmath2 no moving window is needed . we note that the use of a pseudo - spectral algorithm has recently been discussed and analyzed @xcite . this can easily be included into upic - emma if the algorithm is shown to have advantages . we have benchmarked upic - emma by comparing the 2d and 3d simulation results of lwfa in lorentz boosted frames with the corresponding osiris @xcite lab frame simulations . good agreement is found between the osiris lab frame simulations , and upic - emma boosted frame simulations , in both linear , and nonlinear regimes . we also compare upic - emma simulations for different values of @xmath2 and excellent agreement is found . the remainder of this paper is organized as follows . in section [ sect : numinstab ] we discuss the numerical instability induced by relativistic drift . in section [ sect : spectralpic ] , we describe the development of upic - emma , and how using the algorithms in upic - emma can eliminate the instability induced by relativistic plasma drift . in section [ sect : lwfaboostedframesim ] , we discuss details of lwfa lorentz boosted frame simulations using upic - emma . in section [ sect : lwfasim ] , we benchmark upic - emma results with different @xmath2 and with osiris lab frame simulation . summary is given in section [ sect : conclusion ] . the numerical cerenkov instability induced by relativistic plasma drift has been extensively studied in @xcite . in a pic system , when the plasma is drifting relativistically , the velocity of the drifting particles can be equal ( be in resonance ) to the component of the phase velocity of the main em mode along the drift direction . in addition , its aliased modes can always be in resonance with the em modes . the resulting wave - particle resonance leads to a violent numerical instability known as the numerical cerenkov instability . due to the nature of wave - particle resonances , the numerical instability occurs at the intersections of the beam resonances and em modes determined by the maxwell solver used in the simulation . by carefully choosing the maxwell solver , the instability pattern can be manipulated so that mitigation can be achieved . as discussed in @xcite , when a spectral solver is used , there are no intersections of the em modes with the main beam resonance . as a result , the instability can be found only at the aliased resonances and the fatest growing modes are the first spatial aliases . these resonances reside at high @xmath7 in fourier space far away from the important physics . the instability pattern for the spectral solver can be found by investigating the corresponding dispersion relation , @xmath11 ^ 2-[\vec{k}]_e\cdot[\vec{k}]_b+[\vec{k}]_e[\vec{k}]_b)\vec{e}\nonumber\\ = ~&-\omega^2_p \sum_{\mu,\vec{\nu}}(-1)^\mu\biggl\{\int \frac{\overleftrightarrow{s_j}(-\vec{k}')\vec{p}d\vec{p}}{\gamma\omega'-\vec{k}'\cdot\vec{p } } \biggl\{[\omega]\overleftrightarrow{s_e}(\omega',\vec{k}')\vec{e}+ \frac{\vec{p}}{\gamma}\times\{\overleftrightarrow{s_b}(\omega',\vec{k}')([\vec{k}]_e\times\vec{e})\ } \biggr\}\cdot\frac{\partial f^n_0}{\partial \vec{p}}\biggr\}\nonumber\end{aligned}\ ] ] where @xmath12 and @xmath8 are the frequency and wavenumber of the modes in the simulation system ; @xmath13 and @xmath14 are the momentum and lorentz factor of the drifting plasma ; @xmath15 is the electric field ; @xmath16 is the normalized distribution function of the plasma ; @xmath17 is the plasma frequency ; @xmath18 , @xmath19 , and @xmath20 are the corresponding interpolation tensors for em fields and current ; @xmath21 $ ] and @xmath22_{e , b}$ ] are the finite difference time and space operators for the maxwell solver used in the algorithm . and @xmath23 where @xmath24 and @xmath25 are the time step , and grid sizes in the simulation . the sum over @xmath26 and @xmath27 is attributed to the finite grid size and time step used in the simulation . specific expressions of @xmath21 $ ] and @xmath22 $ ] can be found in appendix a of ref . @xcite . since the instability pattern is found near the intersections of the em modes and beam resonance @xcite , we can obtain a simple analytical expression for the instability pattern in the limit @xmath28 ( which leads to @xmath21=\omega$ ] ) . under this assumption , the equation for the em dispersion curves in the spectral solver is @xmath29 and the equations for the beam resonances are @xmath30 where @xmath31 . defining @xmath32 , we can obtain the expressions for the intersections as @xmath33 note that there are no solutions for @xmath34 for the spectral solver . the lowest order terms for the instability pattern are at the @xmath35 and @xmath36 resonances . in the limit of interest @xmath37 , we obtain @xmath38 in figure [ fig : bandfilter ] ( a ) we plot eq . ( [ eq : inst11 ] ) for @xmath39 and @xmath40 . note the `` ring '' pattern of the instability , which crosses the @xmath41 axis near the point @xmath42 ; therefore this mode is located at high @xmath7 values which are far away from the region of interesting physics . therefore , the numerical cerenkov instability can be effectively eliminated if the fatest growing modes ( @xmath43 ) are suppressed in the maxwell solver . in the simulations , we identify the unstable modes in fourier space using the approximate expression eq . ( [ eq : inst11 ] ) . we then apply filters with specific masks which multiply the undesired modes by zero . in figure [ fig : bandfilter ] ( b ) we plot the `` ring - shaped '' band - pass filter used in some of the two - dimensional simulations for testing the instability mitigation . we put `` ring '' in quotes because it is not true ring but rather a range between two parabolas . we also used a low pass filter with a hard cut - off . we filled the simulation box with neutral plasma drifting relativistically at @xmath44 in the @xmath45 direction , and ran cases without a filter , with the `` ring '' filter , and the low pass filter with a hard cutoff . as seen from figure [ fig : bandfilter ] ( c ) , these filters efficiently suppresses the instability modes at @xmath43 in @xmath46 @xcite . therefore , the mitigation of the instability using band - pass filters shows the flexibility and efficiency of a spectral solver in being able to pinpoint the suppression of the unphysical modes in pic simulations while leaving the modes near the interesting physics completely unaffected . , @xmath40 mode of numerical cerenkov instability for the 2d spectral solver in @xmath47 plot ; ( b ) shows the `` ring - shaped '' band - pass filter applied in the 2d spectral solver ; and ( c ) shows the @xmath46 energy evolutions for various simulation setups.,scaledwidth=104.0% ] as mentioned in the section [ sect : numinstab ] , an em - pic code with a spectral solver has superior properties in suppressing the numerical cerenkov instability induced by a relativistic plasma drift . they also have superior properties with respect to numerical dispersion errors and noise . in the following , we will briefly explain the algorithm of a spectral em - pic code , as well as discuss the challenges in optimizing the performances of such a code . spectral pic codes have a long history @xcite . however , despite their advantages in better accuracy and less noise , they are not currently as widely used because they use global field solvers which do not scale as well on parallel computers , and implementing boundary conditions is not as straight forward . a spectral em - pic code has the same basic flow chart as a finite - difference - time - domain ( fdtd ) pic code . in a spectral em - pic code both the charge and current are deposited on the mesh from the particles ; the forces exerted on the particles are interpolated from the mesh points , and particles are advanced using the lorentz forces . the main difference between the spectral pic code and fdtd pic code is the solver used to advance the electromagnetic field and that all field quantities , including the charge and current densities , are defined at the same locations on a cell ( no yee mesh @xcite is needed ) . in a spectral pic code the charge and current are directly deposited , and a strict charge conserving current deposit is not needed because gauss s law is solved at each time step using the charge density . this gives the longitudinal part of the electric field . the longitudinal component of the magnetic field is set to zero at each time step . faraday s law and ampere s law are used to advance the transverse electric and magnetic fields forward in time . note that because gauss s law is solved for directly at each time step , a charge conserving current deposit or boris correction to the longitudinal component of the electric field is not required to maintain that gauss s law is satisfied . the equation for the longitudinal component of electric field @xmath48 becomes : @xmath49 and the transverse electric field and magnetic field are leap - frogged forward in time using faraday s and ampere s law : @xmath50 where the transverse component of the current is : @xmath51 we also multiply @xmath52 and @xmath53 by a shape function @xmath54 where @xmath55 is the particle size . the fields are also multiplied by this shape function then interpolated to the particles @xcite . in addition , just as in a fdtd code , the particle positions and velocities ( and correspondingly the charge and current densities ) are defined at half integer values in time with respect to each other . if positions are defined at whole time steps and velocities ( momentum ) at half integer values , then the longitudinal and transverse components of the electric field are defined at whole time steps ( when particle positions are defined ) and the magnetic field is defined at half - integer values . once the fields are transformed back from @xmath9 space to real space then the particles can be pushed . the particle push is identical to that of a fdtd except for the interpolation of the forces because all field quantities are defined at the same locations in a cell . there are no dispersion errors for light waves due to the grid , however there are errors from the time step . this is a significant advantage of the spectral solver , whereas a fdtd code describes the @xmath56_i$ ] operator to @xmath57 , the spectral code has no errors in the @xmath22 $ ] operator . both a spectral and a fdtd code effectively truncate the highest @xmath58 to @xmath59 . in addition , when including time step errors , the numerical dispersion of a spectral pic code is superluminal , while that of the fdtd code is subluminal . a pseudo - spectral algorithm which also removes the time step errors has recently been described @xcite . as we discuss elsewhere in this paper , the more accurate and superluminal aspect of the em dispersion relation provided by the spectral solver ( together with the simple filters ) is crucial for eliminating the fastest growing modes of numerical cerenkov instability . this ensures no non - physical interaction between waves and particles in the first brillouin zone for the spectral pic code . the corresponding courant condition in 2d and 3d are ( for the square and cubic cells ) @xcite : @xmath60 a spectral pic code is also distinguished from a fdtd code in the way it is parallelized . for the field solver , the simulation box is usually partitioned in one dimension in 2d , and two dimensions in 3d , so that each processor holds global information in the dimension to be transformed . as a result , a parallel spectral pic code requires a fast parallel transpose routine to accomplish efficient fft in multi - dimensions . the nature of all - to - all communications in the fft routines makes it challenging for spectral pic code to scale to large number of processors @xcite . in many cases the decomposition for the particles is the same as that for the fields although this does not have to be the case . we have developed a multi - dimensional em - pic code using a spectral field solver called upic - emma . this code was rapidly put together using components provided by the upic framework , a pic framework with spectral solvers developed at ucla @xcite . upic - emma is fully relativistic and fully parallelized . inherited from the upic framework , upic - emma is coded in layers for convenient extension with different programming styles . the lowest layers are written in fortran77 for high performance . they can be easily extended to many other languages . on top of this layer exists a library of fortran90 wrapper functions which hide the complexity of the fortran77 layer and that provide simpler arguments which enables strict type checking . the code separates the physics procedures from the communication , and utilizes the message - passing interface ( mpi ) for parallel processing . in addition , a multi - tasking library was implemented to enable mixed multi - tasking and mpi messaging , where multi - tasking is used on a multiple cpu shared memory node , and message - passing is used between such nodes . upic - emma also features 3d load balancing where the fields and particles use different partitions . in section [ sect : numinstab ] we described general issues regarding numerical instability that arises when a plasma drifts near the speed of light . in this section we describe some details regarding issues specific to modeling lwfa in a lorentz boosted frame . we describe issues related to numerical dispersion in the lab frame , in the boosted frame , and in transforming from the boosted frame back to the lab frame for comparison . we also discuss the moving antenna and interactions between the laser and the drifting plasma boundary . as mentioned in the introduction , one of the first obstacles in modeling lwfa in a boosted frame is to mitigate the numerical cerenkov instability . for the fdtd pic code @xcite which uses a combination of a yee solver together with the momentum conserving field interpolation scheme , it is useful to choose the optimal time step @xmath61 , where @xmath62-direction is the plasma drifting direction , to minimize the numerical cerenkov instability growth rate . the need to use this time step eliminates the flexibility in tuning the time step to minimize numerical dispersion errors for the laser . in figure [ fig : groupvelocityerror ] we present the error in group velocity of an em wave on a grid in 2d ( we let @xmath63 ) . note that for the yee , and karkkainen solvers @xcite which were discussed in ref . @xcite , the most accurate dispersion relation occurs at their courant limit , but not the corresponding optimal time step at @xmath61 ( for momentum conserving field interpolation ) . on the other hand , for a spectral pic code the instability mitigation does not rely on the relation of grid sizes and time step . in particular , the em dispersion relation can be made arbitrarily accurate by reducing the time step [ see figure [ fig : reflection ] ( a ) ] . therefore , in general when simulating relativistically drifting plasma , a spectral pic code can provide more accuracy and flexibility over the fdtd pic code with respect to numerical dispersion in the simulated frame . note that recently in ref . @xcite a pseudo - spectral algorithm is described which can further improve the accuracy . of the 2d em dispersion relation for various cases . defining @xmath64 to be the laser propagating direction , this plot shows the propagation angle in @xmath65 . if the error @xmath66 is larger than zero , its corresponding point is in the right side of the vertical axis , and vice versa . the group velocity is calculated for the @xmath67 mode while we are using @xmath68 for the calculation.,scaledwidth=104.0% ] while numerical dispersion errors exist when using a finite size grid in vacuum , here we show that when modeling the lwfa in the lorentz boosted frame , these errors in the boosted frame are not necessarily an issue when the results are transformed back to the lab frame . while the value for @xmath2 in the boosted frame can be arbitrary , the speed up is generally larger as @xmath2 is increased . however , choosing @xmath69 , where @xmath70 is the phase velocity of the wake , is generally optimum because in this frame the plasma length and the laser pulse length are nearly matched . when the laser and plasma frequency are comparable each is resolved similarly , i.e. , there is no over - resolution of either the laser wavelength or plasma wavelength . in the boosted frame , the length of the plasma contracts by @xmath2 , the electron and ion mass are both @xmath2 times heavier , the plasma density is @xmath2 times larger , and the corresponding plasma frequency is a lorentz invariant . as for the laser , there is a @xmath71 stretch to the pulse length , while the rayleigh length contracts by @xmath2 . therefore , while the pulse waist does not change , the effective spot size at the rear of the pulse increases by a factor of @xmath72 . hence for sufficiently large @xmath2 an antenna is needed to launch the laser from the laser pulse waist that is moving backwards . the antenna is usually placed at the plasma boundary ( see section [ sect : antenna ] for details ) . in the lab frame simulation , a moving window which only models the region of interest around the laser is often used to reduce the simulation box size . implementation of a moving window is challenging in a spectral pic code due to the non - local nature of the field solver which necessitates knowledge of boundary condition at both of the moving boundaries . however , the relative range of @xmath45 and @xmath73 contracts when lorentz transforming the data of interest from lab frame to boosted frame . if @xmath2 is appropriately chosen , in this frame the length of the plasma column is of the same order as the laser pulse length . as a result , for @xmath74 it is feasible to conduct the boosted frame simulation without the moving window . in lwfa lab frame simulations , an em wave with frequency @xmath75 is incident on a stationary plasma slab . this leads to a reflected and transmitted waves , each having the incident frequency . their wave numbers are determined from the dispersion relation in vacuum ( reflected wave ) , and in plasma ( transmitted wave ) . in a simulation the same physics occurs except the em wave now satisfies the numerical dispersion relation in vacuum and plasma . in the boosted frame there is still a reflected and transmitted waves , except in this case the incident wave , reflected wave , and transmitted wave each have different frequencies . furthermore , numerical issues can lead to some subtle effects . an effective method to identify the frequencies of the reflected and transmitted waves is to use an @xmath76 diagram , which was previously used in studying the radiation generated from ionization fronts @xcite . at the plasma boundary @xmath77 , the phase of each wave @xmath78 must be the same , otherwise the continuity of fields can not be satisfied at every instant in time . this leads to @xmath79 where @xmath80 , @xmath81 , @xmath73 correspond to incident , reflected , and transmitted waves respectively . for example , if @xmath82 then @xmath83 . if the incident and reflected waves obey the vacuum dispersion relation @xmath84 then @xmath85 which can also be obtained from a double lorentz transformation . in a lorentz boosted frame the plasma is drifting but @xmath86 is lorentz transformed to @xmath87 and we want @xmath88 and @xmath89 [ where the @xmath90 sign refers to the boosted frame variables ] . in this frame @xmath91 where @xmath92 can be either the reflected or transmitted waves . the constant @xmath93 is obtained by lorentz transforming @xmath75 and @xmath94 into the boosted frame : @xmath95 and @xmath96 . therefore , @xmath97 regardless of the relationship between @xmath75 and @xmath94 . in a real system @xmath98 although numerical errors in the dispersion relation do not alter the constant @xmath99 . therefore , the reflected and transmitted waves must fall along the line @xmath100 in @xmath101 space ( here we are ignoring the aliasing modes ) . in addition , they must also fall on the dispersion curves for light in a plasma @xcite @xmath102 ^ 2=[k]^2+\frac{\omega'^2_p}{\gamma_b}s^2\frac{[\omega]-[k]v_b}{\omega - kv_b}\end{aligned}\ ] ] or in vacuum @xmath103 ^ 2=[k]^2\end{aligned}\ ] ] on the grid where we assume @xmath104 in eq . ( 19 ) of ref . @xcite , and @xmath105 is lorentz invariant where @xmath106 is the lab frame density , @xmath107 is the electron charge , and @xmath108 is the electron rest mass . the reflected and transmitted wave lie at the intersection between eq . ( [ eq : straightline ] ) , and eq . ( [ eq : plcurve ] ) in plasma or eq . ( [ eq : vacurve ] ) in vacuum . this is shown in figure [ fig : reflection ] ( a ) for a case where @xmath109 , @xmath110 , and @xmath111 . the line @xmath112 and the dispersion curve for a real plasma ( black dashed lines ) , a fdtd yee solver ( red lines ) , and a spectral solver ( green lines for inside the plasma ; magenta lines for in vacuum ) are shown . the vacuum dispersion relation is plotted for the spectral solver in the upper left quadrant for the reasons given in the next paragraph . in figure [ fig : reflection ] ( b ) , we have expanded the region in @xmath76 space near the origin to illustrate the frequency and direction of the transmitted wave which does not depend strongly on the solvers used . when @xmath113 then @xmath114 and @xmath115 . if @xmath116 then @xmath117 would be negative and the phase velocity and group velocity of the transmitted wave would be negative ; however , since @xmath118 the transmitted wave would still be in the plasma . figure [ fig : reflection ] ( a ) also illustrates that numerical errors to the dispersion relation effect the location of the reflected wave . in a real system where @xmath119 in vacuum and @xmath120 in the plasma , then the reflected wave would occur where @xmath121 intersects the vacuum curve , i.e. at @xmath122 , which is larger than the largest @xmath92 in the fundamental brillouin zone . however , for the numerical dispersion curves shown in figure [ fig : reflection ] ( a ) , the reflected wave resides at the intersection with the plasma dispersion relation in the lower right quadrant for the fdtd solver or with the vacuum dispersion relation in the upper left quadrant for the spectral solver . for the fdtd case , the reflected wave has a negative phase and group velocity . however , the group velocity is less than @xmath0 so the reflected wave propagates backwards while staying inside the plasma . for the spectral solver the group velocity is slightly larger than the speed of light so it resides outside the plasma . the predicted locations of the transmitted and reflected waves are confirmed in an osiris ( fdtd ) simulation . this is seen in figure [ fig : reflection ] ( c ) , in which the @xmath92 and @xmath123 spectrum is plotted from a simulation for parameters identical to those used to generate the theoretical plot in figure [ fig : reflection ] ( a ) . strong signals are seen at the predicted locations . for cases of interest the reflection coefficient is small [ the reflected signal is significantly smaller than the transmitted signal in figure [ fig : reflection ] ( c ) ] so the unphysical mode is not energetically important , and it does not complicate the physics . we have also investigated the invariance of transforming results back to the lab frame based on the numerical dispersion relations . we note that solving maxwell s equations on a grid using discrete time steps is not strictly lorentz invariant . for example , the group velocity of light in vacuum for a spectral solver is greater than the speed of light , and it depends on @xmath124 . nevertheless , when carrying out lwfa ( or other ) simulations in a boosted frame , the results are transformed back to the lab frame using the lorentz transformations . this is assumed to be reasonable if one is looking at modes which are properly resolved . as noted earlier , when @xmath2 is chosen near @xmath70 there is a balance between the laser pulse length and the plasma length . in addition , for @xmath125 the transmitted wave @xmath126 , and errors in the boosted frame due to the finite cell size are minimized . in figure [ fig : reflection ] ( d ) we show that when @xmath127 and @xmath128 for the transmitted wave are lorentz transformed back to the lab frame using the velocity addition formulas , @xmath129 where @xmath130 and @xmath131 are the phase and group velocity normalized to @xmath132 , that the numerical errors are nearly absent for sufficiently large @xmath2 . the values for @xmath127 and @xmath128 are calculated from the linear dispersion relation , where @xmath133 is given and @xmath134 is from the dispersion relation in the lab frame . in the boosted frame @xmath135 and @xmath136 . according to the plot , for @xmath137 there are clear numerical errors ; however , for @xmath138 , the numerical errors are minimized . these results can be understood as follows . in a lorentz boosted frame where @xmath139 , the group velocity @xmath140 , while the phase velocity @xmath141 in the numerical system , which when substituted back to eq . ( [ eq : phasegroupv ] ) leads to @xmath142 which are the accurate values for a continuous system . in addition , writing @xmath143 , and @xmath144 , where @xmath145 and @xmath146 corresponds to the correct values in the lab and boosted frame , and defining @xmath147 we can obtain the expressions of the error @xmath148 as @xmath149 note the @xmath150 in the denominator indicates that when @xmath2 is sufficiently large , the errors in velocity when transformed back to the lab frame will be small for any waves . we also note that the arguments going from eq . ( [ eq : phasegroupv ] ) to eq . ( [ eq : verrorgammab ] ) hold for any velocity including those of the particles . this indicates that if we choose the @xmath2 large enough that we would obtain more accurate results compared to a simulation done in the lab frame ( with typical cell sizes and time steps ) . an area of future work is to quantify the differences more accurately . and various em dispersion curves , while in ( b ) we magnified the region near the origin ; ( c ) shows an example of the @xmath151 spectrum of a 1d lwfa boosted frame simulation with the yee solver . the hot spots in ( c ) show where the transmitted and reflected waves are , and agrees with the prediction in ( a ) . ( d ) shows the dependence of the transformed phase and group velocity of the em waves in the plasma with @xmath2 . the phase and group velocity converges quickly as @xmath2 increases from 1.,scaledwidth=104.0% ] as discussed in @xcite and @xcite , the effective spot size of the laser increases by a factor of @xmath152 because the rayleigh length of the laser contracts by @xmath2 and the pulse length expands by @xmath153 . to prevent the need for using a simulation box with transverse size @xmath154 times that in needed in the lab frame , we utilize a thin slice of grids at the plasma boundary ( where the laser beam waist resides ) as an antenna to drive the laser pulse into the plasma . the antenna is moving together with the plasma boundary [ see figure [ fig : simsetup ] ] . . the laser is launched via the moving antenna ( moving together with the plasma column boundary at @xmath155 ) by initializing the appropriate curent in the green slice which has a typical width of @xmath156 . the laser is likewise plotted for @xmath157 . note when the laser is launched via the antenna , only the area within the antenna is initialized . , scaledwidth=104.0% ] the em field in the moving antenna as a function of @xmath158 and time @xmath73 can be derived as follows ( see also @xcite and @xcite ) . for instance , for a laser linearly polarized in the @xmath159 direction , the expression for the electric field @xmath160 of a gaussian pulse in the lab frame can be expressed as : @xmath161}{\exp\biggl[-\frac{2(x_1-ct)^2}{\sigma^2_s}\biggr]}\nonumber\\ & { \exp\biggl[ikx_1+ik\frac{x^2_2+x^2_3}{2r(x_1)}-i\textrm{arctan}\frac{x_1}{x_r}\biggr]}{\exp(-i\omega t)}\nonumber\end{aligned}\ ] ] with @xmath162 where @xmath163 is the amplitude , @xmath164 is laser pulse waist , @xmath165 is the laser pulse length , @xmath12 and @xmath166 are the laser frequency and wavenumber , and @xmath167 is the laser rayleigh length . a similar expression holds for the magnetic field @xmath168 . after lorentz transforming , we can readily obtain the new expression of the laser pulse in the boosted frame @xmath169}{\exp\biggl[-\frac{2[\eta-(1+\beta)ct']^2}{\sigma'^2_s}\biggr]}\nonumber\\ & { \exp\biggl[ik'\eta+ik'\frac{x'^2_2+x'^2_3}{2r'(\eta)}-i\textrm{arctan}\frac{\eta}{x'_r}\biggr]}{\exp(-i\omega ' t')}\end{aligned}\ ] ] where @xmath170 we have neglected the fact that @xmath46 may not equal @xmath171 when solving for the fields on a grid , i.e. , @xmath172 . in the spectral code , the transverse and longitudinal components of the fields are solved for separately . therefore , on the antenna we set @xmath173 so there are no longitudinal fields on it . when launching a laser from the antenna , we assign current ( in the direction of the laser polarization direction ) at every point inside the antenna such that @xmath15 has the desired form and polarization . the other components and the magnetic field follow naturally from the maxwell field solver . the antenna has a finite width of around @xmath156 where @xmath174 is the wavelength of the laser in vacuum to eliminate any backward propagating signal . the current for generating the laser is added after the current is deposited for all the particles in the system is finished . the moving antenna implemented in upic - emma is benchmarked by transforming the data back to the lab frame and then comparing it to data from an osiris lab frame simulation . in the osiris lab frame run , the laser propagates in the @xmath45 direction together with the moving window ; as in the upic - emma run , the laser is launched from a moving antenna . in the upic - emma simulation @xmath175 is used . periodic boundary condition are used for transverse directions in both cases . the transformed upic - emma boosted frame data ( to the lab frame ) are plotted together with the lab frame osiris data in figure [ fig : movant ] . good agreement is found between the two cases . note the shift in the laser wave packet between the osiris data and upic - emma data . we verified that the shift was attributed to the difference in group velocity between the yee solver and spectral solver ( transformed back to lab frame ) . earlier the mode numbers of the fastest growing modes of the numerical cerenkov instability in the spectral solver were identified as eq . ( [ eq : inst11 ] ) . based on this equation , we use filters that eliminate a range of @xmath9 s centered around this parabola . specifically , we muliply all modes by either 1 or 0 . those modes multiplied by 0 are those in the range : @xmath176 in 2d , and @xmath177 in 3d . @xmath178 is usually chosen to be @xmath179 direction ) @xmath151 field at @xmath180 , and ( b ) shows the laser @xmath151 field transformed back from the boosted frame data . ( c)(f ) shows the comparison of on - axis @xmath151 field between osiris data and upic - emma data at various time points . @xmath181 is the coordinates moving together with the moving window.,scaledwidth=104.0% ] we next present simulation results using upic - emma to model lwfa in a boosted frame . two - dimensional simulations in the linear and nonlinear regimes are presented for two different choices of @xmath2 and the results are compared to osiris simulation results in the lab frame ( the upic - emma results are transformed back to the lab frame ) . we also present 3d results from upic - emma including comparison with osiris lab frame simulations . for the linear cases @xmath182 , while for the nonlinear cases @xmath183=3.0 or 4.0 . in both the osiris and upic - emma simulations the time step is chosen near the courant limit . precise values for the simulations parameters are shown in tables [ tab : simpara1 ] and [ tab : simpara2 ] . in figure [ fig : ad1 ] , results from the @xmath183=0.1 case are shown . in the top row , the wakefield @xmath184 is shown at various lab frame times for an osiris lab frame simulations ( blue ) and for two upic - emma simulations where @xmath2=14 ( red ) and @xmath2=28 ( green ) respectively . this figure shows that for early times the three curves are inseparable while for later times the results from osiris lag behind . at later times the upic - emma results for the different @xmath2 remain inseparable . there is no evidence in these plots of any noise in the wake and laser fields due to the numerical cerenkov instability . the fact that the osiris lab frame result slips backwards is due to the numerical dispersion error in @xmath185 that was discussed earlier . in the bottom row of figure [ fig : ad1 ] , the laser field ( @xmath151 ) is plotted at the same times . the same colors are used to show the results from the three simulations . the slippage of the osiris lab frame curve is also seen in the laser field . we next show results for a more nonlinear case where @xmath183=3.0 . as before , there is a lab frame osiris simulation and two upic - emma boosted frame simulations with @xmath2=14 and @xmath2=28 . the same colors as in figure [ fig : ad1 ] are used to distinguish the data from these three simulations . we plot the accelerating field in the upper row and the laser field in the lower row at four various lab frame times ( different times than used in figure [ fig : ad1 ] ) . similarly to the linear @xmath182 case the wakefields from the three simulations are inseparable at early times while for later times the osiris results slip behind . while the agreement between the two boosted frame simulations is still good , it is not as good as for the previous case . the differences in the laser field are small for larger values of @xmath181 ( at the head of the laser ) and there are differences at later times . next in figures [ fig : a4e1 ] and [ fig : a4e3 ] , we present results from a case where the laser amplitude is increased to @xmath186 . in the top row of figure [ fig : a4e1 ] , the plasma density and wakefield in the lorentz boosted frame with @xmath175 are plotted at x@xmath187 . there is no evidence of the numerical cerenkov instability . only a small region of the simulation box , including where the instability is most robust , is plotted . in the lower row of figure [ fig : a4e1 ] we also plot in the lab frame the wakefields obtained in these three simulations . as in the two previous cases , good agreement is found in the wakefield amplitude . there is slippage of the wakefield in the osiris simulation and small differences between the two boosted frame simulations . it is worth noting that , in the case where @xmath188 ( green ) , the wakefield around the spike looks different compared with that of @xmath189 ( red ) and osiris lab frame simulation ( blue ) . the flattened part in the @xmath190 curve indicates more self - trapped charge is loaded in the wakefield . we believe these differences may be due to different statistics since each macro - particle represents much higher charge at higher @xmath2 . in the lwfa lab frame simulation we usually choose the longitudinal cell sizes ( along the laser propagation direction ) to be a fraction of @xmath191 , and the perpendicular cell size to be a fraction of @xmath192 , where @xmath75 and @xmath17 are the frequency of the laser and plasma . meanwhile , in the lorentz boosted frame the perpendicular cell size remains the same , while the longitudinal cell size increases by @xmath71 due to the stretch of the laser wavelength . at the same time , the plasma column contracts by @xmath2 thus the plasma density increases by @xmath2 . as a result , if we keep the number of particle per cell to be the same as in the lab frame , factors of @xmath193 savings can be achieved in the boosted frame simulation . howver , in this case the increase in longitudinal cell sizes and plasma density causes one macro - particle to represent @xmath193 more charge then in the lab frame . for example , if we use a cell with longitudinal size of @xmath194 , perpendicular sizes of @xmath195 , and 8 particles per cell , for a plasma density @xmath196 , each macro - particle represent @xmath197 real electrons , which corresponds to @xmath198 fc of charge . if @xmath199 , then one macro - particle in the boosted frame corresponds to @xmath200 pc . this shows that modeling self - trapping in lorentz boosted frames at large @xmath2 will require future work including identifying where the self - trapped particles come from and loading more particles per cell in these regions . figure [ fig : a4e3 ] shows the comparison of the laser @xmath151 fields for the three @xmath183=4.0 cases . in the top row we show line outs of the laser at four different propagation distances ( times ) . the osiris lab frame curve slips backwards . as in the other cases , the boosted frame curves line up at the front of the laser and as in the other nonlinear case differences in the curves are seen in the back of the laser . in addition , for the @xmath175 case we transform not only the on - axis data , but also the off - axis data in order to compare the 2d laser profile between osiris lab frame run and upic - emma boosted frame run . the osiris lab frame data is shown in the middle row and the upic - emma data in the bottom row . only a part of the the simulation box is shown . good agreement is found in how the laser pump depletes between the two runs and in how the shape evolves . the slippage of the osiris simulation results is seen . last , to illustrate that upic - emma is fully working in three - dimensions , we present the 3d results of upc - emma using the simulation parameters in table [ tab : simpara2 ] and @xmath183=4.0 . these parameters are similar to those in ref . @xcite . in figure [ fig : lu3d ] ( a ) and ( b ) we present 2d slices in the boosted frame of the plasma density and wakefield at the center of the box in the @xmath201-direction at @xmath202 . as in the 2d cases , no noise from the numerical cerenkov instability is evident . in figure [ fig : lu3d ] ( c ) , the wakefield at @xmath203 in the lab frame from the osiris lab frame simulation ( blue ) and upic - emma boosted frame simulation ( red ) are shown . the curves agree well but not perfectly . note that in ( c ) , there is no slippage because we are showing the result at a time where little slippage has occurred . future work will involve understanding these differences for these nonlinear cases . .simulation parameters for the 2d simulations , with @xmath204 ( related to figure [ fig : ad1 ] , [ fig : a3 ] , [ fig : a4e1 ] , and [ fig : a4e3 ] ) . the laser frequency @xmath75 and laser wave number @xmath94 are used to normalize simulation parameters , and @xmath205 . [ cols= " < , > " , ] and @xmath151 between osiris lab frame simulation , and upic - emma boosted frame simulation ( @xmath206 ) at various time steps , for @xmath182 . @xmath181 is the coordinates moving together with the moving window . , scaledwidth=100.0% ] and @xmath151 between osiris lab frame simulation , and upic - emma boosted frame simulation ( @xmath206 ) at various time steps , for @xmath207 . @xmath181 is the coordinates moving together with the moving window . , scaledwidth=105.0% ] ) for @xmath186 . first row shows the 2d plots of plasma electron density ( left ) , and the corresponding @xmath184 for @xmath208 in the boosted frame ( @xmath189 ) . the second row shows the on - axis @xmath184 comparison between osiris lab frame , and upic - emma boosted frame simulation ( @xmath206 ) . @xmath181 is the coordinates moving together with the moving window . , scaledwidth=104.0% ] field between osiris lab frame simulation , and upic - emma boosted frame simulation ( @xmath206 ) at various time steps , for @xmath186 . the first row shows on - axis @xmath151 comparison between osiris lab frame , and upic - emma boosted frame ( @xmath206 ) . the second and third rows show the 2d comparison between the osiris lab frame results and the transformed data from upic - emma boosted frame ( @xmath189 ) . @xmath181 is the coordinates moving together with the moving window . , scaledwidth=104.0% ] ) . ( a ) and ( b ) present 2d cross section plot of the plasma electron density , and @xmath184 at @xmath209 , while ( c ) shows the on - axis @xmath184 comparison at @xmath203 in the lab frame . @xmath181 is the coordinates moving together with the moving window.,scaledwidth=104.0% ] in this paper , we described the rapid development of a new three dimensional pic code that can be used to model laser wakefield acceleration in the lorentz boosted frames . in such simulations a plasma is drifting at relativistic speeds towards the laser , which leads to the numerical cerenkov instability . the growth rates and unstable mode numbers of the numerical cerenkov instability depends on the type of maxwell field solver used . the new code , called upic - emma , uses a spectral field solver , and is fully parallelized . it is built using the components of the upic framework , which is a set of modules for building parallelized pic codes with fft based ( spectral ) solvers . the use of a spectral solver in which the fields are solved for in fourier space allows for more convenient mitigation of the numerical cerenkov instability . the phase velocity of light in vacuum and in a plasma is always greater than the speed of light for a spectral solver . in such cases , the fastest growing modes of the numerical cerenkov instability are due to the lowest order aliased beam mode and they reside at large values of @xmath10 . these modes can be easily filtered out using a `` hard '' low pass or `` shell '' filters , thereby eliminating the fast growing modes of the instability . we presented examples of lwfa boosted frame simulations using upic - emma . several different values of the laser amplitude were simulated ranging from a very linear regime to a nonlinear regime . for the cases shown there was no evidence of the numerical instability and good agreement was found between osiris lab frame and upic - emma boosted frame simulations . the comparison showed that the wake and laser from osiris lab frame simulation slipped behind the results from the boosted frame simulations as expected from numerical dispersion errors . we showed that the dispersion errors become smaller when results are transformed back the lab frame . the results indicate that the use of a spectral code may be attractive for carrying out high fidelity lwfa simulations in boosted frames at high @xmath2 . future work will involve using pseudo - spectral methods , studying how self - trapping occurs in boosted frames where each macro - particle contains significant charge , studying and understanding the differences between boosted frame and lab frame simulations for these and more nonlinear regimes , and using a moving cathode to simultaneously launch trailing beams into laser driven wakes or to study beam driven wakes . this work was supported by doe awards de - fc02 - 07er41500 , de - sc0008491 , de - fg02 - 92er40727 , and de - sc0008316 , and by nsf grants nsf phy-0904039 and nsf phy-0936266 , and by nsfc grant 11175102 , thousand young talents program , and by fct ( portugal ) , grant expl / fis - pla/0834/1012 , and by the european research council ( erc-2010-adg grant 267841 ) . simulations were carried out on the ucla hoffman 2 cluster , and dawson 2 cluster . 00 t. tajima , j. m. dawson , phys . 43 , 267 ( 1979 ) . w. lu , et . al . , phys . st - accel . beams 10 , 061301 ( 2007 ) . s. f. martins , r. a. fonseca , w. lu , w. b. mori and l. o. silva , nat . vol . 6 , 311 ( 2010 ) d. f. gordon , w. b. mori , and t. m. antonsen , ieee trans . plasma sci . 28 , 1135 ( 2000 ) . p. mora and t. m. antonsen , phys . plasma . 4 , 217 ( 1997 ) c. huang , et . al . , j. comp . 217 , 658 ( 2006 ) a. f. lifschitz , et . al . , j. comp . phys . 228 , 1803 ( 2009 ) j. -l . vay , phys . 98 , 130405 ( 2007 ) s. f. martins , r. a. fonseca , l. o. silva , w. lu , w. b. mori , comp . 181 , 869 ( 2010 ) . vay , c. g. r. geddes , e. cormier - michel , d. p. grote , j. comp . 230 , 5908 ( 2011 ) . c. d. decker , w. b. mori , phys . 72 , 490 ( 1994 ) b. b. godfrey , j. comp . 15 , 504 ( 1974 ) b. b. godfrey , j. comp . 19 , 58 ( 1975 ) k. nagata , phd thesis , osaka university , 2008 . p. yu et.al , in proc . 15th advanced accelerator concepts workshop , austin , tx , 2012 . b. godfrey and j .- l . vay , j. comp . 248 , 33 - 46 ( 2013 ) x. xu , et . al . , comp . , vol . 184 , 25032514 ( 2013 ) j. m. dawson , rev . modern phys . 55 , no . 2 , 403 ( 1983 ) k. yee , ieee transactions on antennas and propagation , vol . 14 , 302 ( 1966 ) v. k. decyk , comp . 97 , 8794 ( 1995 ) v. k. decyk , comp . 177 , 95 ( 2007 ) . m. karkkainen , e. gjonaj , t. lau , t. weiland , in proceedings of the international computational accelerator physics conference , chamonix , france , 2006 , pp . i. haber , r. lee , h. klein , j. boris , in : proc . sixth conf . on num . plasmas , berkeley , ca , 1973 , pp . ; see also appendix a of [ j .- l . vay , i. haber , b. b. godfrey , j. comp . 243 , 260268 ( 2013 ) ] . c. h. lai , t. katsouleas , w. b. mori , d. w. whittum , ieee trans . plasma phys . , vol 21 , 18 ( 1993 ) j .- l . vay , c. g. r. geddes , e. cormier - michel , d. p. grote , arxiv:1009.2727v2 r. a. fonseca , et al . , in : p.m.a . sloot , et al . ( eds . ) , iccs , in : lncs , vol . 2331 , 2002 , pp . 342351 . f. s. tsung , r. narang , w. b. mori , c. joshi , r. a. fonseca , l. o. silva , phys . rev . 93 , 185002 ( 2004 ) . a. t. lin , j. m. dawson , h. okuda , phys . fluids , vol . 17 , 1995 ( 1974 ) .

simulating laser wakefield acceleration ( lwfa ) in a lorentz boosted frame in which the plasma drifts towards the laser with @xmath0 can speedup the simulation by factors of @xmath1 . in these simulations the relativistic drifting plasma inevitably induces a high frequency numerical instability that contaminates the interested physics . various approaches have been proposed to mitigate this instability . one approach is to solve maxwell equations in fourier space ( a spectral solver ) as this has been shown to suppress the fastest growing modes of this instability in simple test problems using a simple low pass or `` ring '' or `` shell '' like filters in fourier space . we describe the development of a fully parallelized , multi - dimensional , particle - in - cell code that uses a spectral solver to solve maxwell s equations and that includes the ability to launch a laser using a moving antenna . this new em - pic code is called upic - emma and it is based on the components of the ucla pic framework ( upic ) . we show that by using upic - emma , lwfa simulations in the boosted frames with arbitrary @xmath2 can be conducted without the presence of the numerical instability . we also compare the results of a few lwfa cases for several values of @xmath2 , including lab frame simulations using osiris , a em - pic code with a finite difference time domain ( fdtd ) maxwell solver . these comparisons include cases in both linear and nonlinear regimes . we also investigate some issues associated with numerical dispersion in lab and boosted frame simulations and between fdtd and spectral solvers . particle - in - cell , plasma simulation , laser wakefield accelerator , boosted frame simulation , spectral solver , numerical cerenkov instability

let @xmath1 be the set of positive integers and @xmath2 . let @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 , @xmath8 . the kostka number @xmath9 and the littlewood - richardson coefficient @xmath10 play an essential role in the representation theory of the symmetric groups and the special linear groups . their combinatorial definitions can be found in section [ definitions ] . the issue of their computational complexity has been a question of folklore . recently , in @xcite , e. rassart asked whether there exist fast ( polynomial time ) algorithms to compute kostka numbers and littlewood richardson coefficients ( question 1 , page 99 ) . we prove that these two quantities are # p - complete ( see theorems [ th : k ] , [ th : c ] ) , and thus answer his question in the negative under the hypothesis that a # p - complete quantity can not be computed in polynomial time . in @xcite , barvinok and fomin show how the set of all non - zero @xmath9 for a given @xmath11 can be produced in time that is polynomial in the total size of the input and output . they also give a probabilistic algorithm running in time , polynomial in the total size of input and output , that computes the set of all non - zero littlewood - richardson coefficients @xmath12 given @xmath13 and @xmath11 . in @xcite , methods for the explicit computation of the kostka numbers and littlewood - richardson coefficients using vector partition functions are discussed . combinatorially , the kostka number @xmath9 is the number of young tableaux that have shape @xmath13 and content @xmath11 ( @xcite , page 25 ) . the littlewood - richardson coefficient @xmath10 is the number of lr skew tableaux on the shape @xmath14 with content @xmath15 ( this follows from corollary 2 , ( v ) , page 62 and lemma 1 , page 65 of @xcite ) . representation theoretically , @xmath9 is the multiplicity of the weight @xmath11 in the representation @xmath16 of the lie algebra @xmath17 of the special linear group having highest weight @xmath13 and @xmath10 is the multiplicity of @xmath18 in the tensor product @xmath19 . they also appear in the representation theory of the symmetric groups ( see chapter 7 , @xcite ) . while there are formulae for @xmath9 and @xmath10 due to kostant and steinberg respectively ( @xcite , @xcite ) , the number of terms is , in general , exponential in the bit - length of the input . these numbers have interesting properties such as , for fixed @xmath13 , @xmath11 , @xmath20 , @xmath15 , @xmath21 and @xmath22 ( @xcite ) are polynomials in @xmath23 . whether @xmath24 can be answered in polynomial time ( see proposition [ prop : k_in_p ] ) , and so can the question of whether @xmath25 , though the latter is a highly non - trivial fact estabilished by ketan mulmuley and milind sohoni @xcite , and uses the proof of the saturation conjecture by knutson and tao @xcite . this fact plays an important role in the approach to the @xmath26 vs @xmath27 question @xcite due to ketan mulmuley and milind sohoni . let @xmath28 and @xmath29 . we reduce the # p - complete problem ( @xcite ) of finding the number @xmath30 of @xmath0 contingency tables that have row sums @xmath31 and column sums @xmath32 , to that of finding some @xmath33 . we reduce this in turn to finding some @xmath34 , where @xmath35 and @xmath36 can be computed in time polynomial in the size of @xmath37 . the main tool used in the reduction to finding kostka numbers is a correspondence found by donald e. knuth ( @xcite ) between the set @xmath38 of contingency tables and pairs of tableaux having contents @xmath39 and @xmath40 respectively . a * young diagram * ( @xcite , page 1 ) is a collection of boxes , arranged in left justified rows , such that from top to bottom , the number of boxes in a row is monotonically ( weakly ) decreasing . the first two shapes in figure @xmath41 are young diagrams . a * filling * is a numbering of the boxes of a young diagram with positive integers , that are not necessarily distinct . a * young tableau * or simply * tableau * is a filling such that the entries are 1 . weakly increasing from left to right across each row , and 2 . strictly increasing from top to bottom , down each column . @xmath26 and @xmath42 , in figure @xmath43 , are young tableau . a * skew diagram * is the diagram obtained from removing a smaller young diagram out of a larger one . the third shape in figure @xmath41 is a skew shape . a * skew tableau * is a filling of the boxes of a skew diagram with positive integers , non - decreasing in rows , and strictly increasing in columns ( see figure @xmath44 ) . if the number of boxes in the @xmath45 row of a tableau , for @xmath46 is @xmath47 and @xmath48 , it is said to have * shape * @xmath13 . if the tableau houses @xmath49 copies of @xmath50 for @xmath51 and @xmath52 , it is said to have * content * @xmath11 . thus , in figure @xmath43 , @xmath26 and @xmath42 have the same shape @xmath53 , but contents @xmath54 and @xmath55 respectively . [ fig : gamma ] and the skew shape @xmath14.,title="fig:",height=153 ] we represent a young diagram having @xmath56 rows and @xmath47 boxes in the @xmath45 row by @xmath57 . given two shapes @xmath58 and @xmath59 , @xmath60 is defined to be the skew - shape obtained by attaching the lower left corner of @xmath20 to the upper right corner of @xmath13 as in figure @xmath41 ( see @xcite , page 60 ) . @xmath61 denotes the number of bits used in the description of this tuple of vectors . for @xmath62 , let @xmath63 . for vectors @xmath13 , @xmath11 , we say that @xmath64 if @xmath65 and @xmath66 . in addition , if @xmath67 , we say @xmath68 . this ordering is called the dominance ordering . we call a tableau * littlewood - richardson * or * lr * , if , when its entries are read right to left , top to bottom , at any moment , the number of copies of @xmath69 encountered is greater than or equal to the number of copies of @xmath70 encountered ( @xcite , page 63 ) . we denote the set of all ( possibly skew ) tableaux of shape @xmath13 and content @xmath11 by @xmath71 , and its subset consisting of all lr ( possibly skew ) tableaux by lr@xmath72 . the * kostka number * @xmath73 is the number of tableaux of shape @xmath13 and content @xmath11 , i.e @xmath74 ( @xcite , page 25 ) . the * littlewood - richardson coefficient * @xmath75 is the number of lr skew tableaux of shape @xmath14 of content @xmath15 , i.e @xmath76 ( this follows from corollary 2 , ( v ) , page 62 and lemma 1 , page 65 of @xcite ) . valiant defined the counting class # p in his seminal paper @xcite , and proved that the computation of the permanent is # p - complete . the class # p is the class of functions @xmath77 , for which there exists a polynomial time turing machine @xmath78 and a polynomial @xmath79 such that @xmath80 . the tableau shapes @xmath81 and contents @xmath82 are described by vectors with integer coefficients . therefore the number of boxes in a tableau might be exponential in the size of the input . thus it needs to be estabilished that even though an object in @xmath72 or @xmath83 may be exponentially large , it has a polynomial length description , and its membership ( given this short description ) can be verified in polynomial time . [ prop : kp ] the problem of determining @xmath9 is in # p. * proof : * + let @xmath84 be defined as in the introduction . a tableau of shape @xmath13 is fully specified given the number @xmath85 of copies of @xmath69 , that are present in the @xmath86 row , for all @xmath87 and @xmath51 . thus its description is of length polynomial in @xmath88 , in fact it has length @xmath89 . given the set @xmath90 , this corresponds to the tableau obtained , by writing , in the @xmath45 row , from left to right , @xmath91 copies of @xmath41 , @xmath92 copies of @xmath43 and so on for each @xmath87 . to verify whether this filling has the content @xmath11 , we need to verify , that @xmath93 for each @xmath50 , which takes @xmath89 time . to verify whether this filling has the shape @xmath13 , we need to verify , that @xmath94 for each @xmath69 , which again takes @xmath89 time . to verify whether this filling is a tableau , we need to check that the entries in a column of this filling are strictly increasing from top to bottom . in other words , that , for each row @xmath95 and each @xmath96 @xmath97 this can be done in time @xmath89 if we maintained cumulative sums and calculated the next sum incrementally . therefore the task of computing @xmath9 is in # p. + @xmath98 [ prop : k_in_p ] given @xmath13 and @xmath11 , whether or not @xmath99 can be answered in polynomial time . * proof : * + let @xmath13 and @xmath11 be as in proposition [ prop : kp ] . for any permutation @xmath100 of the set @xmath101 , let @xmath102 be the vector @xmath103 . it is a known fact that @xmath104 ( see @xcite , page 26 ) . let @xmath100 be a permutation such that @xmath105 , @xmath106 . for any @xmath107 , whose components are arranged in non - increasing order , it is known that @xmath108 if and only if @xmath109 ( see @xcite , page 26 ) . whether @xmath110 can be checked in time that is @xmath111 . thus , whether or not @xmath99 can be answered in time @xmath112 , which is the time it takes to find a permutation @xmath100 that arranges the components of @xmath113 in non - increasing order . + @xmath98 [ prop : cp ] the problem of computing @xmath10 is in # p. * proof : * + let @xmath81 and @xmath15 be defined as in the introduction . given @xmath114 , we shall describe it as follows . let the number of occurrences of @xmath50 in the @xmath45 row of @xmath115 be @xmath116 . @xmath115 is described by @xmath117 and @xmath118 . conversely , given @xmath117 and @xmath119 , we verify that it has shape @xmath14 and content @xmath11 , by checking that 1 . @xmath120 , 2 . @xmath121 , and 3 . @xmath122 . this takes time @xmath123 . the description corresponds to the skew filling obtained , by writing in the shape @xmath14 , left to right , in the @xmath45 row , @xmath124 copies of @xmath41 , @xmath125 copies of @xmath43 and so on . such a skew filling would be a skew tableau if and only if the entries of each column strictly increased from the top to the bottom . in other words , @xmath126 for each row @xmath127 other than @xmath56 and each @xmath128 . thus , to verify from its description , that a given filling is an element of @xmath129 takes polynomial time . to verify that a specified filling is lr , we need to check , that , while scanning the entries of the tableau from left to right , top to bottom , at any instant , the number of copies of @xmath50 encountered is @xmath130 the number of copies of @xmath131 encountered . in other words , that the following inequality holds for every row @xmath132 and every @xmath133 , @xmath134 thus , to verify that a skew tableau is lr only takes time polynomial in the bit - length of its description thus , the problem of determining @xmath10 is in # p. + @xmath98 [ lemma : kp ] given @xmath135 , @xmath136 , and @xmath137 , let @xmath138 and @xmath139 . then , @xmath140 . * proof : * + [ fig : rsk ] and @xmath141 for @xmath142 , @xmath143.,title="fig:",height=76 ] the r - s - k ( robinson - schensted - knuth ) correspondence ( @xcite , or @xcite pages 40 - 41 ) gives a bijection between @xmath38 , the set of @xmath0 contingency tables with row sums @xmath39 and column sums @xmath40 , and pairs of tableaux @xmath144 having a common shape but contents @xmath39 and @xmath40 respectively . in other words , we have a bijection between @xmath38 and @xmath145 . a sample correspondence is shown in figure @xmath43 . [ fig : leavep ] and @xmath146 for @xmath142 and @xmath143.,title="fig:",height=76 ] for every shape @xmath147 , such that that @xmath148 , there is exactly one tableau having shape @xmath149 and content @xmath39 . for any other shape @xmath149 there is no tableau having shape @xmath149 and content @xmath39 . any tableau with content @xmath150 can have at most two rows , since the entries in a single column are all distinct . further , @xmath41 may be present only in the @xmath151 row , because top to bottom , each column has strictly increasing entries . therefore for a tableau to have content @xmath39 , it is necessary that @xmath152 and there be atleast @xmath153 boxes in the @xmath151 row . in other words , @xmath154 conversely , this condition together with @xmath136 , implies that @xmath155 . therefore the filling in which the first @xmath153 boxes of the top row contain @xmath41 and all others contain @xmath43 is a tableau ( see @xmath42 in figure @xmath156 ) . since all the copies of @xmath41 must be in the first row and must be in a contiguous stretch including the leftmost box , this is the only tableau in @xmath157 . hence the claim is proved . thus there is a bijection between @xmath141 and the set of tableaux of content @xmath40 having some shape @xmath148 . i.e , there is a bijection between @xmath141 and @xmath146 . an example of this is provided in figure @xmath156 . let us now consider the set @xmath146 . any tableau in @xmath146 can be extended to a tableau of the shape @xmath158 by filling the boxes that are in @xmath13 but not @xmath149 , with @xmath159 . this extension is a bijection between @xmath146 and @xmath72 . if there is a tableau of shape @xmath149 and content @xmath39 , @xmath160 , and @xmath161 . @xmath162 . therefore no two of the boxes in @xmath13 which are not in @xmath149 belong to the same column . those of these boxes , that are present in a given row , occupy a contiguous stretch that includes the rightmost box . therefore by filling them with @xmath159 we get a tableau in @xmath72 . conversely , given a tableau @xmath163 in @xmath72 , deleting all boxes of @xmath163 filled with @xmath159 gives a tableau in @xmath164 . these two maps are inverses of each other and hence provide a bijection between @xmath164 and @xmath72 . hence the claim is proved . an example of this correspondence has been illustrated in figure @xmath165 . [ fig : embed ] and @xmath72 , where @xmath166 and @xmath167.,title="fig:",height=76 ] therefore , @xmath168 . + @xmath98 [ th : k ] the problem of computing @xmath9 , even when @xmath13 has only @xmath43 rows , is # p - complete . * proof : * + computing @xmath9 is in # p by proposition [ prop : kp ] . now the result follows from lemma [ lemma : kp ] because the computation of @xmath30 is known to be # p - complete ( @xcite ) . + @xmath98 [ lemma : kc ] given @xmath169 , @xmath170 , and @xmath171 , let @xmath172 where @xmath173 , and @xmath174 , where @xmath175 , and @xmath176 . then @xmath177 . * proof : * + [ fig : kc ] and lr@xmath129 for @xmath178 and @xmath167 , @xmath179 and @xmath180.,title="fig:",height=124 ] @xmath34 is , by definition , @xmath76 , which is the number of lr tableaux on the skew shape @xmath14 that have content @xmath15 . the skew shape @xmath14 consists of a copy of @xmath13 and a copy of @xmath20 , as in figures @xmath41 and @xmath44 . for any skew tableau @xmath115 of shape @xmath14 , we shall denote by @xmath181 , the restriction of @xmath115 to the copy of @xmath20 and by @xmath182 , the restriction of @xmath115 to the copy of @xmath13 . thus , @xmath181 is a tableau of shape @xmath20 and @xmath182 is a tableau of shape @xmath13 . let @xmath114 . for @xmath183 , the @xmath45 row of @xmath181 must consist entirely of copies of @xmath69 . we proceed to show this by induction . let @xmath184 . the rightmost entry of the @xmath151 row of @xmath181 , is @xmath41 , by the lr condition . the entries of each row are non - decreasing , left to right . therefore each entry of the @xmath151 row must be @xmath41 . now assume for some @xmath185 , @xmath186 , that the claim is true for all @xmath187 . if @xmath188 , we are done , so let @xmath189 . the entries of columns of @xmath181 are strictly increasing , top to bottom . by the induction hypothesis , the entry directly above a square in the @xmath190 row is @xmath185 . therefore , any element in the @xmath190 row must be @xmath191 . but the rightmost entry of the @xmath190 row can not be anything @xmath192 , because this would violate the lr condition . the entries of each row are non - decreasing , left to right . therefore each entry of the @xmath190 row must be @xmath193 . hence , by induction , the claim is proved . consequently , @xmath182 must have content @xmath194 . in other words , @xmath195 . conversely , given any tableau @xmath196 , let @xmath197 be the skew tableau of shape @xmath198 in which @xmath199 and the @xmath45 row of @xmath200 consists entirely of copies of @xmath69 . while scanning @xmath197 right to left , top to bottom , the lr condition could not possibly be violated while on a square of @xmath200 . by the time we begin scanning the rightmost box of the first row of @xmath201 , for any @xmath202 , the number of copies of @xmath69 encountered is already @xmath203 more than the number of copies of @xmath70 encountered . therefore the lr condition could not possibly be violated on any square of @xmath201 either . therefore @xmath204 . @xmath199 , thus we have a bijection between @xmath83 , the set of lr skew tableaux of shape @xmath14 having content @xmath15 and @xmath72 , the set of tableaux of shape @xmath13 having content @xmath11 . hence @xmath205 as claimed . + @xmath98 [ th : c ] the problem of computing @xmath206 , even when @xmath13 has only @xmath43 rows is # p - complete . * proof : * + by proposition [ prop : cp ] , computing @xmath10 is in # p. we have already proved in theorem [ th : k ] , that the computation of @xmath9 is # p complete . the result now follows from lemma [ lemma : kc ] . we proved that the computation of kostka numbers and littlewood - richardson coefficients is # p complete . the reduction to computing kostka numbers was from the # p complete problem @xcite of computing the number of contingency tables having given row and column sums . the problem of computing kostka numbers was then reduced to that of computing littlewood - richardson coefficients . it is not known whether there exist fully polynomial randomized approximation schemes ( fpras ) to compute any of these quantities . it would be of interest to find fpras to compute kostka numbers and littlewood - richardson coefficients . these are conjectured to exist by k. mulmuley and m. sohoni , @xcite in their approach to the p vs np question . i wish express my gratitude to ketan mulmuley for suggesting the topic of this paper and for many valuable discussions . many thanks are due to ravi kannan for informing me about @xcite . i also sincerely thank etienne rassart , rahul santhanam , and lszl babai for helpful comments .

kostka numbers and littlewood - richardson coefficients play an essential role in the representation theory of the symmetric groups and the special linear groups . there has been a significant amount of interest in their computation ( @xcite , @xcite , @xcite , @xcite , @xcite ) . the issue of their computational complexity has been a question of folklore , but was asked explicitly by e. rassart @xcite . we prove that the computation of either quantity is # p - complete . the reduction to computing kostka numbers , is from the # p - complete problem @xcite of counting the number of @xmath0 contingency tables having given row and column sums . the main ingredient in this reduction is a correspondence discovered by d. e. knuth @xcite . the reduction to the problem of computing littlewood - richardson coefficients is from that of computing kostka numbers .

quantum spin ice ( qsi ) is the name given to a class of magnetic system on the pyrochlore lattice which exhibit a competition between classical frustration and quantum fluctuations.@xcite the particular kind of quantum fluctuations acting within the set of classical spin ice states ( characterized by their induction of a pauling residual entropy ) leads to a promising avenue to search for the elusive quantum spin liquid@xcite ( qsl ) state among an actively studied set of real compounds.@xcite recently , huang , chen , and hermele@xcite have proposed a specialized model of qsi , relevant to strong spin - orbit coupled materials exhibiting kramers doublets with dipolar - octupolar character in @xmath4-and @xmath5-electron systems . the model has several competing diagonal and off - diagonal terms in the hamiltonian , and is believed to harbor both gapless and gapped spin liquids in different parameter regimes . from this three - dimensional ( 3d ) hamiltonian , the model can be reduced to a two - dimensional ( 2d ) hamiltonian of interacting spins on the kagome lattice . @xcite this model , which we call _ quantum kagome ice _ ( qki ) , opens a theoretical and experimental avenue to search for two - dimensional qsl phases within the broader context of qsi physics . most importantly from a theoretical perspective , the quantum kagome ice hamiltonian ( like the original huang - chan - hermele hamiltonian ) is devoid of the quantum monte carlo ( qmc ) `` sign problem '' , allowing for systematic study of the phase diagram via large scale computer simulations . as mentioned above , the model hamiltonian for quantum kagome ice studied by carrasquillia _ et al . , _ @xcite originates from a certain class of @xmath4-and @xmath5-electron systems in which dipolar - octupolar kramers doublets arise on the sites of the pyrochlore lattice.@xcite in the localized limit , this model maps to a well - known xyz hamiltonian on the pyrochlore lattice . it is known to exhibit two distinct phases on the pyrochlore lattice dipolar qsi and octupolar qsi phases . @xcite in the presence of a magnetic field along the [ 111 ] crystallographic direction there is a partial lifting of degeneracy . the magnetic field pins one spin per tetrahedron and the model effectively decouples into alternating kagome and triangular layers of the original pyrochlore structure . @xcite the resulting quantum kagome ice hamiltonian is given by @xcite @xmath6-h\sum_l s_l^z , \label{k1}\end{aligned}\ ] ] where @xmath7 can be used to transform eq . into an xyz hamiltonian . the above model [ eq . ] is regarded as a projection of the qsi hamiltonian on a @xmath8d pyrochlore lattice onto the @xmath9d kagome lattice hamiltonian .@xcite the previously - studied u(1)-invariant xxz model @xcite ( equivalent to hard - core bosons ) is recovered when @xmath10 . as @xmath11 is turned on , the u(1 ) symmetry is explicitly broken , and eq . retains only @xmath12 symmetry : @xmath13 ; @xmath14 . note that the sign of the @xmath11 term can be changed by a canonical unitary transformation , @xmath15 , that leaves all other terms invariant . this unitary transformation corresponds to a @xmath16 rotation about the @xmath17-axis . recently , it has been shown using non - perturbative , unbiased qmc simulations,@xcite that eq . with @xmath18 promotes a qsl phase . the hamiltonian exhibits several distinct phases as a function @xmath19 and @xmath20 . at @xmath18 and small @xmath1 , there is an in - plane canted ferromagnet ( cfm ) , with spins partially ordered along the @xmath21- or @xmath22-direction ( depending on the sign of @xmath11 ) , that exists on the entire @xmath20 axis , arising from a spontaneous breaking of the hamiltonian @xmath12 symmetry . for very large @xmath1 , a fully polarized ( fp ) phase occurs with spins aligned along the @xmath17-direction . the nontrivial phases appear in the regime @xmath23 and @xmath24 and are called `` lobes '' on the phase diagram . the lobes have magnetization per site @xmath25 depending on the sign of @xmath1 , arising from the constraint that two spins point up and two down ( or vice - versa ) , per unit cell of 3 sites . these lobes were described in ref . [ ] as a candidate gapped featureless qsl state , with no evidence of conventional ordering of any kind . in this paper , we aim to complement the recent quantum monte carlo simulation work by carrasquillia _ et al . , _ @xcite through an analytical large-@xmath0 approach.@xcite we re - examine the phases observed in the qmc study from the perspective of the spin - wave expansion . we begin with @xmath26 , where the classical ground state is highly degenerate . to first - order in @xmath27 , we show that quantum fluctuations lift the degeneracy by selecting a unique ground state mirroring what is seen in the qmc . next , in the @xmath25 lobes ( @xmath3 and @xmath28 with @xmath18 ) the linear spin - wave analysis produces a ground - state energy , magnetization , and `` condensate fraction '' with similar features as the qmc , including the phase transition out of the cfm phase . the excitation spectrum at @xmath18 displays no soft modes for a wide range of parameter space . thus , at the level of our large-@xmath0 theory , there is no indication of any tendency toward any magnetic order inside the putative spin liquid lobes . kagome lattice with three - sublattice unit cell denoted by @xmath29 . the blue hexagon is the brillouin zone with @xmath30 ; @xmath31 ; @xmath32 . the lattice vectors @xmath33 , are given by @xmath34 . we set @xmath35 in the computations.,width=288 ] we begin with a study of the classical ground states of the hamiltonian in eq . . a crucial observation ( made above ) is that the sign of the @xmath11 term is not unique , i.e. , it can be changed by a canonical unitary transformation . at zero magnetic field , and in the limit @xmath18 ; @xmath36 , the ground state of eq . is an in - plane cfm state ordered along the @xmath21- or @xmath22-direction depending on the sign of @xmath11 . in contrast , the limit @xmath37 exhibits extensive accidental degeneracy ; it corresponds to the classical spin ice on kagome lattice.@xcite in the large-@xmath0 limit , the spin operators in eq . can be represented as classical vectors@xcite which can be written as @xmath38 , where @xmath39 is a unit vector . the classical hamiltonian under the unitary transformation , @xmath15 is given by , @xmath40 , \label{mf } \end{aligned}\ ] ] where @xmath41 ; @xmath42 when @xmath43 and @xmath44 are nearest neighbours , and 0 otherwise ( fig . ) . we have taken the magnetic field , @xmath1 , to be of order @xmath0 . @xmath45 is the number sites and @xmath46 is the number of unit cells . for @xmath10 , the classical energy is clearly minimized when @xmath47 is a constant , independent of the sublattice index . once @xmath48 , the energy is minimized for @xmath49 or @xmath16 , depending on the sign of @xmath11 . however , since the sign of @xmath11 can be changed by the aforementioned canonical transformation , we will fix it to be positive henceforth . in this case , the groundstate corresponds to @xmath50 . in the case when @xmath51 , the classical model exhibits an accidental degeneracy.@xcite indeed , introducing parameters @xmath52 and @xmath53 such that @xmath54 , and @xmath55 , we have , @xmath56 where @xmath57 , and @xmath58 is arbitrary . these equations are identical to those that arise in the classical limit of the xxz model , studied previously in ref . it follows that including a nonzero @xmath1 field , the topology of the classical phase diagram is the same as the xxz model , but the interpretation of the phases is different . for example , the superfluid phase in the xxz model is replaced by the cfm phase in our case ( @xmath18 ) , which only has a broken @xmath12 symmetry , in contrast to the broken u(1 ) of the superfluid . at @xmath59 , a phase transition occurs between the cfm phase and a phase with no cfm order , but with a finite magnetization @xmath25 in the @xmath17-direction . in contrast to the xxz model,@xcite the total @xmath60 is not conserved here , hence these lobes retain a small finite magnetic susceptibility.@xcite also , note that there exists another classical phase boundary between the cfm and the fp state at @xmath61 . in the previous section , we uncovered the classical phase diagram of the @xmath12-invariant quantum kagome ice model . in this section we perform a linear spin wave analysis of this model to study the role of quantum fluctuations.@xcite our main goal is to investigate the role of the quantum fluctuations introduced by the @xmath11 term in lifting the extensive degeneracy of the classical ground state . to facilitate the spin wave expansion , we rotate the coordinate axes so that the @xmath17 axis coincides with the local direction of the classical polarization . we rotate by the angle @xmath62 about the @xmath22 axis , which corresponds to the rotation matrix , @xmath63 following the standard approach,@xcite we employ a three - sublattice holstein primakoff transform with the bosonic operators , @xmath64 . after fourier transform we obtain @xmath65 where @xmath66 label the sublattices , and the coefficients are given by ^0=_aa,_bb,_cc , @xmath67 for the kagome lattice , we have @xmath68 with @xmath69 @xmath70 , where @xmath71 and @xmath72 where @xmath73 . it is convenient to write eq . in terms of the nambu operators @xmath74 and @xmath75 : @xmath76 with @xmath77 ; @xmath78 and @xmath79 where @xmath80 is a @xmath81 identity matrix ( @xmath82 ) and @xmath83 is a pauli matrix . the hamiltonian ( eq . ) can be diagonalized by the generalized bogoliubov transformation @xcite @xmath84 where @xmath85 is a @xmath86 matrix , and @xmath87 with @xmath88 being the quasiparticle operators . the matrix @xmath85 satisfies the relations , @xmath89 with @xmath90 and @xmath91 being the diagonal matrix of the quasiparticle energy eigenvalues . this is equivalent to saying that we need to diagonalize a matrix @xmath92 , whose eigenvalues are given by @xmath93 $ ] and the columns of @xmath94 are the corresponding eigenvectors . from eq . we have @xmath95 and @xmath96 yielding @xmath97 with the ground state energy given by @xmath98 with @xmath99 ; @xmath100 ; and @xmath101.,width=336 ] at @xmath26 , the classical ground state is highly degenerate . to see how the @xmath11 term lifts this degeneracy , we analyze an expression for the quantum fluctuation correction to the classical energy , which is given by , @xmath102 on kagome lattice for @xmath103 at @xmath100 . the qmc result @xcite is with system size @xmath104 and @xmath105 at inverse temperature @xmath106.,width=336 ] figure shows the plot of @xmath107 as a function of @xmath58 , which parameterizes the different classical ground states . quantum fluctuations select a particular @xmath58 as the lowest energy state . this value of @xmath58 is a solution of the equation @xmath108 . and @xmath109 for @xmath18 . the colormap represents @xmath110 per site . , width=336 ] vs. @xmath109 entering the lobes at @xmath111 . the trend captured in qmc simulation @xcite is consistent with this spin wave calculation . , width=336 ] vs. @xmath112 entering the lobes . the dash line corresponds to the u(1 ) model , which is well - known to be the same in qmc simulation . the symbol shows the trends of the @xmath12 model . there is no corresponding qmc result reported in ref . [ ] . , width=336 ] in fig . we show the ground state energy as a function of @xmath113 , together with the exact energy obtained by unbiased qmc simulations.@xcite clearly , the linear swt captures the trend of the dependence of the ground state energy on @xmath114 . at the rotationally symmetric point @xmath115 and @xmath116 , eq . becomes @xmath117,\nonumber\\&=\sum_{\la lm\ra}[-s_{l}^xs_{m}^x+s_{l}^ys_{m}^y+s_{l}^zs_{m}^z].\end{aligned}\ ] ] in the @xmath118-quantization axis , @xmath119 , hence the classical ground state is exact , which gives @xmath120 and corresponds to the maximum in fig . . next , we examine the @xmath121 region of the phase diagram . in the purely classical limit , @xmath122 the addition of a symmetry - breaking field provides an additional constraint . namely , for @xmath26 , each triangular plaquette of the lattice realizes a configuration with either two @xmath123 , one @xmath124 , or one @xmath123 , two @xmath124 . in contrast , for @xmath125 say , only the two @xmath123 , one @xmath124 configuration is energetically allowed for each triangle ( resulting in a magnetization per spin of @xmath126 ) . thus , the degeneracy is reduced compared to the @xmath26 case , but still remains extensive ( up to the critical field @xmath127 ) . including linear spin - wave corrections , we first compute the expectation value of the sublattice magnetization , given by@xcite @xmath128 where @xmath129 for the sublattice @xmath130 respectively . note that the expectation value of @xmath118 plays a role analogous to the condensate fraction in ref . . this will be useful in comparing our results to the qmc results of ref . . in fig . [ phase_lobe ] , we show a plot of the expectation value of @xmath118 as a function of @xmath112 and @xmath109 . the phase diagram consists of three distinct phases . there is a cfm phase with @xmath131 , and an fp phase which appears for @xmath132 ( not shown ) . in addition , our theory captures the appearance of two @xmath133 lobes for @xmath134 and @xmath135 , with vanishing @xmath110 and @xmath136 . these lobes are somewhat analogous to the well - known fixed - magnetization lobes of the xxz model,@xcite however , an important difference is that a finite susceptibility is retained throughout the lobes as mentioned above . this is a consequence of the fact that the hamiltonian only has @xmath12 symmetry , instead of u(1 ) in the xxz case . another feature of the phase diagram is that linear spin wave theory captures the same transition point ( @xmath137 ) as qmc simulations @xcite from the lobes to cfm . this is in contrast to the u(1 ) model,@xcite which suggests that quantum fluctuations are small in the model with @xmath138 . in the qmc simulations of the @xmath18 hamiltonian,@xcite the nature of the ground state within the lobes is clearly distinct from the xxz model,@xcitewhere a valence bond solid ( vbs ) is observed . in contrast , ref . observes a featureless spin liquid state . this is despite the fact that a perturbative argument taking @xmath139 suggests a similar vbs phase as in the xxz case.@xcite qmc simulation also captures a finite susceptibility in the lobes as corroborated here with spin wave theory calculation . for the in - plane fm and cfm ordered states along the brillouin zone paths with @xmath18 . , width=336 ] inside the @xmath25 lobe for @xmath140 , @xmath141.,width=336 ] in fig . we plot @xmath142 entering the lobe at @xmath111 and fig . shows the plot of @xmath142 as a function of the magnetic field inside the lobe . in the xxz model , @xmath143 inside the lobe . in our case , it is clear that we retain a small but finite susceptibility inside the lobe which is a consequence of the @xmath12 symmetry of our model as discussed above . these results are consistent with qmc simulations.@xcite finally , we examine the spin wave dispersion in the @xmath144 lobes . as shown in fig . , a special feature of the @xmath145 model in zero field is that the spin wave spectrum does not have a soft mode , which is a consequence of the discrete @xmath12 symmetry . there are two dispersive modes and one flat mode ; the dispersive modes touch at the @xmath146 point ( see fig . [ kagome ] ) in the brillouin zone and the flat band always has the lowest energy . the existence of the flat band is a well - known feature of the kagome lattice that follows from its geometry.@xcite when the field reaches the critical value @xmath147 , one of the modes softens signifying the transition to the lobes . in the lobe itself , the gap reopens and there are no soft modes again ( see fig . [ z2_lobe ] ) . motivated by recent quantum monte carlo ( qmc ) simulations of the quantum kagome ice model,@xcite we have presented another perspective of the ground state phase diagram of the model using the large-@xmath0 expansion . we focused mainly on the @xmath140 model , which has not been studied before using this approach . exploring the parameter regimes of this model , we uncovered three distinct phases , including a canted ferromagnet ( cfm ) , a fully - polarized ( fp ) state , and most interestingly finite - magnetization @xmath148 `` lobes '' . this picture is consistent with the phases observed in qmc simulations . at finite magnetic field and dominant ising coupling , the nature of the @xmath148 magnetized lobes is explicitly investigated and we find that all of the observables qualitatively agree with the qmc simulations . in particular , no evidence of any out - of - plane ( @xmath60 ) spin order , which would be expected in the most likely candidate for valence - bond solid ( vbs ) order is found within the lobes . however , in order to more thoroughly address the possibility of vbs order within the lobes , it would be prudent to undertake a study using schwinger bosons in the future . the authors would like to thank juan carrasquilla for invaluable discussions , and also for providing us with details of his qmc simulations . support was provided by nserc , the canada research chair program , and the perimeter institute ( pi ) for theoretical physics . research at perimeter institute is supported by the government of canada through industry canada and by the province of ontario through the ministry of research and innovation . 99 m. j. p. gingras and p. a. mcclarty , reports on progress in physics * 77 * , 056501 ( 2014 ) . zhihao hao , alexandre g. r. day , and michel j. p. gingras , , 214430 , ( 2014 ) . s. ryu , o. i. motrunich , j. alicea , and m. p. a. fisher , , 184406 , 2007 . s. b. lee , s. onoda , and l. balents , , 104412 , ( 2012 ) . r. applegate , n. r. hayre , r. r. p. singh , t. lin , a. g. r. day , and m. j. p. gingras , , 097205 , ( 2012 ) . y. wan and o. tchernyshyov , , 247210 , ( 2012 ) . k. a. ross , l. savary , b. d. gaulin , and l. balents , phys . x * 1 * , 021002 , ( 2011 ) . l. savary and l. balents , , 205130 , ( 2013 ) . h. r. molavian , p. a. mcclarty , m. j. p. gingras , arxiv:0912.2957 [ cond-mat.stat-mech ] . h. r. molavian , m. j. p. gingras , j. phys . : condens . matter * 21 * , 172201 ( 2009 ) . k. matsuhira , z. hiroi , t. tayama , s. takagi , and t. sakakbara , j. phys . : condens . matter * 14 * , l559 ( 2002 ) . m. udagawa _ et al_. , j. phys . * 71 * , 2365 ( 2002 ) . h. r. molavian , m. j.p . gingras , b. canals , , 157204 , ( 2007 ) . l. balents , nature * 464 * , 199 ( 2010 ) . y .- huang , g. chen , and m. hermele , , 167203 ( 2014 ) . k. kimura , s. nakatsuji , j - j . wen , c. broholm , m. b. stone , e. nishibori and h. sawa , nature communications * 4 * , 1934 , ( 2013 ) . t. fennell , m. kenzelmann , b. roessli , h. mutka , j. ollivier , m. ruminy , u. stuhr , o. zaharko , l. bovo , a. cervellino , m. k. haas , and r. j. cava , , * 112 * , 017203 , ( 2014 ) . anderson , phys . rev . * 86 * , 694 ( 1952 ) ; t. oguchi , phys . rev . * 117 * , 117 ( 1960 ) . a. b. harris , c. kallin , and a. j. berlinsky , , 2899 , ( 1992 ) . jolicoeur and j. c. le guillou , , 2727 , ( 1989 ) . r. moessner and s. l. sondhi , , 064411 , ( 2003 ) . s. v. isakov , k. s. raman , r. moessner , and s. l. sondhi , , 104418 , ( 2004 ) . a. j. macdonald , p. c. w. holdsworth and r. g. melko , j. phys . : condens . matter * 23 * , 164208 ( 2011 ) . s. v. isakov , s. wessel , r. g. melko , k. sengupta , and y. b. kim , , 147202 , 2006 . r. g. melko , j. phys . : matter * 19 * , 145203 ( 2007 ) . k. damle and t. senthil , , 067202 , ( 2006 ) . zhang , r. dillenschneider , y. yu , s. eggert , , 174515 , ( 2011 ) . k. sengupta , s. v. isakov , and y. b. kim , , 245103 , ( 2006 ) . s. a. owerre and m. b. paranjape , physics reports * 546 * , 1 , ( 2015 ) . s. miyashita , h. kawamura , j. phys . . jpn . * 54 * , 3385 , ( 1985 ) . b. kleine , e. mller - hartmann , k. frahm , and p. fazekas , z. phys . * b * 87 , 103 , ( 1992 ) . g. murthy , d. arovas , and a. auerbach , , 3104 , ( 1997 ) . j. -p . blaizot and g. ripka , quantum theory of finite systems ( mit press , cambridge , ma , 1986 ) . c. zeng and v. elser , , 8436 , ( 1990 ) . j. carrasquilla , private communication . a. lauchli and m. hermele , private communication .

we present a large-@xmath0 study of a quantum spin ice hamiltonian , introduced by huang _ et al . _ [ , 167203 ( 2014 ) ] , on the kagome lattice . this model involves a competition between the frustrating ising term of classical kagome ice , a zeeman magnetic field @xmath1 , and a nearest - neighbor transverse spin - flip term @xmath2 . recent quantum monte carlo ( qmc ) simulations by carrasquilla , _ et al . , _ [ nature communications * 6 * , 7421 ( 2015 ) ] , uncovered lobes of a disordered phase for large ising interaction and @xmath3 a putative quantum spin liquid phase . here , we examine the nature of this model using large-@xmath0 expansion . we show that the ground state properties generally have the same trends with those observed in qmc simulations . in particular , the large-@xmath0 ground state phase diagram captures the existence of the disordered lobes .

figure [ obs ] shows the observed velocity distribution in the @xmath1-@xmath2 plane , based on the hipparcos input ( radial velocities ) and output ( tangential velocities ) catalogues . the distribution is characterised by several distinct streams , and in particular a stream with an asymmetric drift of @xmath3 kms@xmath4 and @xmath5 , which hereafter we will refer to as the `` herculis '' stream according to a comoving eggen group ( skuljan et al . this stream most probably has a dynamical origin because its stars are mainly older than a few gyr and present a large range of metallicities ( raboud et al . 1998 ) , and is under - represented in the hipparcos sample , which is biased towards young stars . the average luminosity of stars in this sample indeed increases with distance and the sample only covers the vertical region of the galactic plane where the fraction of young stars is largest . we have realised a large 3d @xmath0-body simulation designed to evolve into realistic barred models of the milky way . the initial conditions are adapted from fux ( 1997 ) and @xmath6 particles , allowing to compute very detailed velocity distributions in small space regions . after the formation of the bar ( at @xmath7 gyr ) , the simulation reveals a complex velocity structure with multiple streams occuring almost everywhere in the disc . the streams , as well as the velocity dispersions , remain very time - dependent until the end of the simulation ( @xmath8 gyr ) , even within regions corotating with the bar . outside corotation and for an inclination angle of the bar @xmath9 , the simulation often displays a herculis - like stream ( figure [ sim ] ) . the model selected here is a snapshot at @xmath10 gyr , scaled to a corotation radius of @xmath11 kpc , where @xmath12 and @xmath13 are the galactocentric distances of the lagrangian points @xmath14 and @xmath15 , and to a circular velocity in the azimuthally averaged potential of @xmath16 kms@xmath4 at @xmath17 kpc . in these units , @xmath18 kpc and @xmath19 kms@xmath4 at @xmath20 and @xmath21 . our interpretation for the herculis stream follows the same lines as in raboud et al . ( 1998 ) . in a rotating barred potential , the value of the hamiltonian : @xmath22 where @xmath23 , @xmath24 is the gravitational potential and @xmath25 the pattern speed of the bar , is an integral of motion ( jacobi s integral ) . three critical values of this integral are associated with stars corotating at the lagrangian points ( @xmath26 and @xmath27 ) and stars on circular orbits at the outer lindblad resonance in the axisymmetriesed potential ( @xmath28 ) . stars with @xmath29 follow `` hot '' orbits which are susceptible to cross the corotation radius . in the simulation , even far beyond the solar circle , a substantial fraction of particles are on such orbits and the hamiltonian distribution frequently shows a secondary peak between @xmath30 and @xmath31 ( figure [ hist ] ) . if @xmath1 and @xmath2 are measured relative to the galactic centre , equation ( [ ham ] ) also writes : @xmath32 thus , neglecting the vertical velocity component @xmath33 , the contours of constant hamiltonian in the @xmath34 plane are circles centred on @xmath35 and of radius @xmath36 . figure [ sim ] shows that the herculis - like stream in the model falls just outside the @xmath30-contour and thus concerns particles on hot orbits . this interpretation is different from the one proposed by dehnen ( 1999 ) , who attributes the valley between the herculis stream and the hyades - pleiades branch to stars scattered off the olr . in our case , the valley is produced by the decline of the hot orbit population as @xmath38 . but in both cases the position of the olr is found near @xmath20 , in agreement with kalnajs ( 1991 ) conclusion based on the hyades and sirius streams . figure [ orb ] traces back in time the orbits of the particles in the herculis - like stream which were recently captured inside corotation ( using the instantaneous frozen rotating potential ) . these particles come from the low hamiltonian edge of the stream and represent about 10% of the particles in the stream . this percentage strongly depends on @xmath39 and @xmath40 . dehnen w. 1999 , letters , in press ( astro - ph/9908105 ) fux r. 1997 , a&a 327 , 983 kalnajs a.j . 1991 . in : sundelius b. ( ed . ) varberg conf . , dynamics of disc galaxies . gteborg , p. 323 raboud d. , grenon m. , martinet l. , fux r. , udry s. 1998 , a&a 335 , l61 skuljan j. , hearnshaw j.b . , cottrell p.l . 1999 , 308 , 731

a high - resolution @xmath0-body simulation suggests that stellar streams in the discs of barred galaxies are common and strongly time - dependent . the velocity distribution of stars in the solar neighbourhood betray many such streams , including a stream of outward moving stars with low angular momentum . this stream is interpreted as a signature of the galactic bar , in the sense that its stars have just enough energy ( jacobi s integral ) to cross the corotation resonance .

let @xmath6 be a stationary ergodic sequence of random variables with mean @xmath7 and finite variance , satisfying the central limit theorem , i.e. @xmath8 where the limit variance @xmath9 is given by the formula @xmath10 for practical applications of the central limit theorem , e.g. , for calculation of confidence intervals or testing hypothesis concerning the mean , we need to estimate the variance @xmath9 . we investigate the properties of a subsampling estimator of @xmath11 . given observations @xmath12 and a block length @xmath13 , we define the overall mean and the block means as @xmath14 respectively . by the central limit theorem , @xmath15 converges in distribution to @xmath16 as @xmath17 goes to infinity , and thus @xmath18 . thus , it is natural to estimate @xmath11 by the arithmetic mean of @xmath19 , @xmath20 $ ] . replacing the unknown mean @xmath7 by the sample mean @xmath21 , we obtain @xmath22 } \sqrt{\frac{\pi}{2 } } \sum_{i=1}^{[n / l ] } \frac{|s_i(l)-l\bar{x}_n|}{\sqrt{l } } \label{eq : b_n}\ ] ] as an estimator for @xmath11 . in what follows we will always write @xmath23 $ ] for the number of full blocks of length @xmath24 in the set @xmath25 . subsampling estimators of the limit variance @xmath9 of the arithmetic mean of a stationary stochastic process have been studied by various authors under different assumptions concerning the dependence structure . @xcite investigated @xmath26-mixing processes and the estimator @xmath27 } \sum_{i=1}^{[n / l ] } \left(\frac{|s_i(l)-l\bar{x}_n|}{\sqrt{l}}\right)^2\ ] ] for @xmath9 . @xcite studied properly normalized means of @xmath28-th powers of @xmath29 as an estimator for @xmath30 in the case of @xmath31-mixing processes . they consider overlapping blocks , i.e. @xmath32 where @xmath33 . @xcite studied the same estimator , in the case @xmath34 , for associated processes . @xcite investigated the case of weakly dependent ( in the sense of * ? ? ? * ) processes . in this paper , we extend the above mentioned results to processes that can be represented as functionals of strongly mixing processes . we assume that @xmath6 can be written as @xmath35 for all @xmath36 , where @xmath37 is a measurable function and where @xmath38 is an @xmath26-mixing process . recall that a process @xmath38 is called @xmath26-mixing ( or strongly mixing ) if @xmath39 as @xmath40 . here @xmath41 denotes the @xmath11-field generated by the random variables @xmath42 . the coefficients @xmath43 are called mixing coefficients . in addition , we have to require that the function @xmath37 defined above is continuous in a suitable sense . define @xmath44 in what follows , we will have to make assumptions concerning the behavior of the coefficients @xmath45 , @xmath46 and @xmath47 as @xmath48 . functionals of @xmath26-mixing processes cover most of the standard examples of weakly dependent processes known in the literature , for example arma- and garch - processes from time series analysis , many markov processes and hidden markov models , the sequence of digits and remainders in continued fraction expansion , and many chaotic dynamical systems , such as expanding , piecewise monotone maps of the unit interval . details can be found , e.g. , in @xcite and in @xcite . suppose @xmath1 is a stationary process that can be expressed as a functional of an @xmath26-mixing process . suppose that @xmath49 for some @xmath50 and that the approximation coefficients @xmath51 satisfy @xmath52 and the mixing coefficients @xmath53 satisfy @xmath54 for some @xmath55 . then , for any sequence @xmath56 satisfying @xmath57 and @xmath58 , we have @xmath59 in @xmath2 . [ th : fctl - ltwo ] @xcite study robust non - parametric tests for structural breaks in time series . they consider the model @xmath60 , where @xmath61 is a stationary stochastic process and @xmath62 is a sequence of unknown constants . the hypothesis of interest is @xmath63 to be tested against the alternative @xmath64 . investigating the asymptotic distribution of the two - sample hodges - lehmann test statistic @xmath65 dehling and fried show that under the above null hypothesis , we have @xmath66 as @xmath67 . here @xmath68 , where @xmath69 is an independent copy of @xmath5 , and @xmath70 where @xmath4 is the distribution function of @xmath5 . for statistical applications , the limit variance @xmath71 has to be estimated . we can directly apply theorem [ th : fctl - ltwo ] to the sequence @xmath72 , provided that it satisfies the conditions and that @xmath4 is known . the latter is generally not the case and thus @xmath4 has to be replaced by the empirical distribution function @xmath73 . in what follows , we will show that this leads to an @xmath2-consistent estimator . denoting the empirical distribution function by @xmath74 , we define the estimators @xmath75 } \sqrt{\frac{\pi}{2 } } \sum_{i=1}^{[n / l_n ] } \frac{|t_i(l_n)-l_n\ , \bar{u}_n|}{\sqrt{l_n } } , \\ \hat{d}_n&= & \frac{1}{[n / l_n ] } \sqrt{\frac{\pi}{2 } } \sum_{i=1}^{[n / l_n ] } \frac{|\hat{t}_i(l_n)-l_n\ , \tilde{u}_n|}{\sqrt{l_n } } , \label{eq : def - dhat}\end{aligned}\ ] ] where @xmath76 , @xmath77 , and @xmath78 let @xmath1 be a stationary process satisfying the assumptions of theorem [ th : fctl - ltwo ] . in addition , assume that @xmath79 , @xmath80 and that @xmath81 is lipschitz - continuous . then , as @xmath82 , @xmath83 and @xmath84 , we have @xmath85 in @xmath2 . [ th : edf - ltwo ] we investigate the performance of the proposed estimators and compare them to alternatives . the estimators are applied to arma(1,1 ) processes with gaussian innovations and different parameter settings , which are given in table [ ts_coefs_tab ] . for each parameter combination , 1000 samples are drawn , and variance , bias and mean squared error of the estimates are computed . the sample size is @xmath86 throughout . .types of arma(1,1 ) time series used in the simulation . ar and ma denote the corresponding coefficents of the arma(1,1 ) model in the usual notation . [ cols=">,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] besides comparing several estimators , another question addressed by the simulations is the applicability of carlstein s rule for choosing the block length . @xcite proved it to be optimal in terms of mean squared error for his estimator @xmath87 , cf . ( [ eq : carlstein ] ) , in case of an ar(1 ) process with known autocorrelation parameter . the simulation study consists of two stages : the first part is concerned with the estimation of @xmath9 , cf . ( [ eq : clt - var ] ) , the second part with the estimation of @xmath71 , cf . ( [ eq : clt - var2 ] ) . [ [ part-1 ] ] part 1 + + + + + + we compare @xmath88 , cf . ( [ eq : b_n ] ) , the carlstein estimator @xmath87 , referred to as ce in the following , and the kernel - based estimator of @xcite , abbreviated by jde in the following . while the former two depend on the block length , the latter is affected by the choice of the kernel and the bandwidth @xmath89 . we use the bartlett kernel , which is a common choice due to guaranteed positive semi - definiteness . to compute mean squared error and bias we compare the estimates to the true value of @xmath90 , not the infinite series ( [ eq : clt - var ] ) . we use the fixed block lengths 1 , 2 , 5 , 10 , 30 , 50 and 100 , and determine the data adaptive block length following carlstein as follows : an ar(1 ) process is fit to the sample ( regardless of whether it was generated as such ) using the r function arima ( ) @xcite , and the block length is computed from the estimated autocorrelation parameter . for the jde , the bandwidths 2.8 , 3.5 , 4.7 , 7.9 and 22.4 are used , corresponding to sixth to second root of @xmath91 . [ [ part-2 ] ] part 2 + + + + + + the second part of the simulation , which deals with the estimation of @xmath71 , compares @xmath92 , cf . ( [ eq : def - dhat ] ) , to the @xcite estimator , denoted by pse in the following . both depend on a block length , but contrary to @xmath92 , the pse is based on overlapping blocks . the data generation is exactly the same as in the first part ( sample size , number of repetitions and arma parameters ) , but the data evaluation is different : each sample is split into halves and the estimate of interest is computed as the mean of the two subsample estimates . this choice is motivated by the intended application to the change - point test by @xcite . the block lengths considered are 2 , 5 , 10 , 25 , 40 , 50 for @xmath92 and 5 , 10 , 25 , 40 , 45 for the pse . note that , while a block length 1 is of interest in the first simulation , since it yields a variance estimator for independent observation , it leads to data independent estimators in this situation and is hence neglected . for computing the true value of @xmath93 we make use of the fact that the population version of spearman s rho @xmath94 equals @xmath95 if @xmath96 and @xmath97 are jointly gaussian with correlation @xmath31 . this result can be traced back to @xcite ; for a newer reference see , e.g. , @xcite . hence , letting @xmath98 denote the correlation of @xmath5 and @xmath99 , we have @xmath100 , ce and jde as a function of the smoothing parameter ; computed from 1000 samples of size 500 generated from an arma(0.5 , 0.5 ) model . gray symbols correspond to the data adaptive block length , plotted at the mean block length of 16.04 . the x - axis is on log scale . ] , ce and jde as a function of the smoothing parameter ; computed from 1000 samples of size 500 generated from an arma(0.5 , 0.5 ) model . gray symbols correspond to the data adaptive block length , plotted at the mean block length of 16.04 . the x - axis is on log scale . ] [ [ results ] ] results + + + + + + + as far as the comparison of the estimators is concerned , we get similar results for all arma parameter settings , which are exemplified for the @xmath9-estimators at an arma(0.5,0.5 ) process in figure [ fig : bias.mse.1 ] and for the @xmath93-estimators at an ma(0.1 ) process in figure [ fig : bias.mse.2 ] . the performance of the estimators is generally quite similar . this is also true for @xmath92 and the pse , which may seem surprising , since the pse , in contrast to @xmath92 , uses overlapping blocks . in terms of bias , @xmath88 and @xmath92 slightly outperform their competitors ( see left panels of figures [ fig : bias.mse.1 ] and [ fig : bias.mse.2 ] ) , but they show slightly bigger variances , leading to a worse mean squared error ( see right panels ) . the choice of the smoothing parameter ( block length or bandwidth ) has altogether a much larger influence than the choice of the method . we furthermore observe the following : the data adaptive block lengths , which correspond to the gray symbols in the figures , perform quite well in terms of mean squared error ( mse ) unless the ma - coefficient is strongly negative . except for a few cases , @xmath9 is always underestimated if all autocorrelations of the underlying process are non - negative and overestimated if all are non - positive . in the white noise case , a block length of 1 is mse optimal , but other block lengths give very similar results . on the other hand , even for small non - zero ar- or ma - coefficients , not accounting for the dependence substantially impairs the accuracy of the estimation . in summary , all estimators considered show a very similar performance for good choices of the smoothing parameter despite some considerable differences in their construction . an unfavorable choice of the block length leads to an increase in both , bias and variance , but with varying extents for the different estimators . based on our results , we suggest to use the estimators @xmath88 , @xmath92 if one is willing to trade a smaller bias for a bigger variance . the data adaptive block length proposed by @xcite seems to yield acceptable results for all block length dependent estimators even if the assumption of an ar(1 ) process is not met . we also carried out some simulations for other values of @xmath101 . while the estimators performance , for fixed block length respectively bandwidth , clearly improves with the sample size , the ranking of the estimators is little affected . therefore we focus on a fixed sample size here . and pse as a function of the block length ; computed from 1000 samples of size 500 generated from an arma(0.5 , 0.5 ) model . gray symbols correspond to the data adaptive block length , plotted at the mean block length of 2.7 . the x - axis is on log scale . ] and pse as a function of the block length ; computed from 1000 samples of size 500 generated from an arma(0.5 , 0.5 ) model . gray symbols correspond to the data adaptive block length , plotted at the mean block length of 2.7 . the x - axis is on log scale . ] we will first state and prove a special case of theorem [ th : fctl - ltwo ] where the process @xmath6 is @xmath26-mixing . we need this result later on in the proof of theorem [ th : fctl - ltwo ] . suppose @xmath1 is a stationary , @xmath26-mixing process with + @xmath102 for some @xmath103 and mixing coefficients @xmath53 satisfying @xmath54 for some @xmath55 . then , for any sequence @xmath56 satisfying @xmath57 and @xmath58 , we have @xmath104 in @xmath2 . [ prop : sm - ltwo ] we will show that @xmath105 as @xmath40 . by definition of @xmath88 , we get @xmath106 as @xmath58 , we get by the central limit theorem @xmath107 in @xmath108 . thus , it remains to prove that @xmath109 by lemma [ lem3 ] we have @xmath110 and thus it suffices to show @xmath111 applying lemma [ lem1 ] with @xmath112 to the stationary @xmath26-mixing process @xmath113 , we obtain @xmath114 where we have used lemma [ lem1 ] with @xmath115 in the final step . as @xmath116 , we have thus proved proposition [ prop : sm - ltwo ] . defining @xmath117 we decompose @xmath6 into a part that is @xmath26-mixing and a remaining part , which is small . in addition , we define the corresponding arithmetic means @xmath118 and @xmath119 and the block sums @xmath120 furthermore , let @xmath121 for fixed @xmath122 , the process @xmath123 is @xmath26-mixing with coefficients @xmath124 . moreover , we have @xmath125 and @xmath126 . finally , one can show that @xmath127}^{\delta/(2+\delta ) } + \psi_{[j/3]}^{(1+\delta)/(2+\delta)}\right ) , \ ] ] see @xcite . observe that @xmath128 we will now bound each of the terms on the right hand side separately . to the third term , we apply the inequality @xmath129@xmath130 valid for any @xmath131 . we fix @xmath132 and choose @xmath133 sufficiently large such that for all @xmath134 @xmath135and hence by minkowski s inequality @xmath136since @xmath137 , we also have @xmath138 for all @xmath134 with @xmath133 sufficiently large . proposition [ prop : sm - ltwo ] implies that , for any fixed @xmath122 , @xmath139 in @xmath2 as @xmath140 . hence the first term on the right hand side of ( [ eq : main - th ] ) converges to @xmath141 in @xmath2 . finally , stationarity and lemma [ lem2 ] imply @xmath142 in @xmath2 as @xmath140 . this finishes the proof of theorem [ th : fctl - ltwo ] . as @xmath4 is lipschitz continuous , the process @xmath72 satisfies the conditions of theorem [ th : fctl - ltwo ] , hence @xmath143 in @xmath2 by theorem [ th : fctl - ltwo ] . thus it suffices to show that @xmath144 almost surely . we get @xmath145\sqrt{l_n } } \sqrt{\frac{\pi}{2 } } \sum_{i=1}^{[n / l_n ] } \left| t_i(l_n)-\hat{t}_i(l_n)-l_n(\bar{u}_n-\tilde{u}_n ) \right| \\ \leq \frac{\sqrt{\pi/2}}{[n / l_n]\sqrt{l_n } } \sum_{i=1}^{[n / l_n ] } & & \hspace{-2em } \left(\sum_{j=(i-1)l+1}^{il}|\hat{f}_n(x_j)-f(x_j)| + \frac{l_n}{n}\sum_{i=1}^n|\hat{f}_n(x_i)-f(x_i)| \right ) \\[0.5ex ] \leq \ , & & \hspace{-1em } 2 \sqrt{l_n}\sup_{t\in { { \mathds r } } } |\hat{f}_n(t)-f(t)| . \end{aligned}\ ] ] lemma [ lem : ep - lil ] implies @xmath146 almost surely , provided @xmath147 . this proves that @xmath92 is an @xmath2 consistent estimator of the limit variance @xmath148 . for ease of reference we state some results from the literature on weakly dependent processes that we need in the course of the proofs . @xcite [ lem : ep - lil ] let @xmath167 be a stationary sequence of strongly mixing random variables with @xmath79 , and let @xmath168 , @xmath159 , be a sequence of functionals with common distribution function @xmath169 , satisfying @xmath170 where @xmath171 . then @xmath172 almost surely , where @xmath173 . s. borovkova , r. m. burton , and h. dehling . limit theorems for functionals of mixing processes with applications to @xmath174-statistics and dimension estimation . _ transactions of the american mathematical society _ , 3530 ( 11):0 42614318 , 2001 . h. dehling and r. fried . asymptotic distribution of two - sample empirical @xmath174-quantiles with applications to robust tests for shifts in location . _ journal of multivariate analysis _ , 1050 ( 1):0 124140 , 2012 . doi : 10.1016/j.jmva.2011.08.014 . p. doukhan , j. jakubowicz , and j. r. len . variance estimation with applications . in i. berkes , r. bradley , h. dehling , m. peligrad , and r. tichy , editors , _ dependence in probability , analysis and number theory . _ , pages 203231 . heber city , ut : kendrick press , 2010 . k. pearson . _ mathematical contributions to the theory of evolution . xvi . on further methods of determining correlation_. drapers company research memoirs : biometric series iv . london : dulau & co. , 1907 .

we study subsampling estimators for the limit variance @xmath0 of partial sums of a stationary stochastic process @xmath1 . we establish @xmath2-consistency of a non - overlapping block resampling method . our results apply to processes that can be represented as functionals of strongly mixing processes . motivated by recent applications to rank tests , we also study estimators for the series @xmath3 , where @xmath4 is the distribution function of @xmath5 . simulations illustrate the usefulness of the proposed estimators and of a mean squared error optimal rule for the choice of the block length .

there have been a number of reports on the coexistence of magnetic order and superconductivity in the ruthenocuprate rusr@xmath0gdcu@xmath1o@xmath2 , synthesized for the first time in 1995 [ 1 ] . its peculiarity lies in the fact that , unlike previous compounds , magnetic order occurs at a temperature much higher than the superconducting transition temperature . this compound is characterised by a triple perovskitic cell similar to the high temperature superconducting cuprate ( htsc ) yba@xmath1cu@xmath3o@xmath4 , in that it contains two cuo@xmath0 layers while the cuo chains are replaced by a ruo@xmath0 layer . however , various experimental reports came to different conclusions . it has been suggested on the basis of transport measurements that its electronic behaviour is similar to an underdoped htsc [ 2 ] while , on the contrary , nmr measurements resulted comparable to those of an optimally doped htsc [ 3 ] . some other reports concluded that the magnetic order is ferromagnetic in the ruo@xmath0 layers [ 2,4,5,6 ] in which case there should be competition between the superconducting and magnetic order parameters resulting eventually in a spontaneous vortex phase formation or spatial modulation of the respective order parameters . however , powder neutron diffraction showed that the low - field magnetic order is predominantely antiferromagnetic [ 7 ] , with a small ferromagnetic component presumibly produced by spin canting . the spectrum of published data includes also non superconducting samples showing similar macroscopic magnetic behaviour [ 8 ] , samples showing zero resistance but no diamagnetic signal and finally samples with evidence of a resistive and magnetic transition . since the physical properties of this rutheno - cuprate material are strongly dependent on the details of the preparation procedure , and can be very different even in samples that turn out to be formally identical to a standard structural and chemical - physical characterization , we have conducted a systematic on the effects of sample preparation conditions on the properties of such hybrid compound . the crystal structure was determined by powder x - ray diffraction ( xrd ) using cu k@xmath5 radiation . dc resistivity and magnetic measurements were performed by the standard four - probe technique with 1 ma current in a closed - cycle helium cryostat in the temperature range 15 - 300 k and by a quantum design squid magnetometer respectively . measurements were performed on similar size bar - shaped sintered polycrystalline specimens allowing comparison of the results . polycrystalline samples with nominal composition rusr@xmath0gdcu@xmath1o@xmath2 ( hereafter referred as ru-1212 ) are commonly prepared by solid - state reaction technique from a mixture of high purity ruo@xmath6(99.95% ) , gd@xmath0o@xmath7(99.99% ) , cuo ( 99.9% ) and srco@xmath3 ( 99.99% ) [ 1,4,5,9,10,11 ] . the raw materials are : i first reacted in air at about 960c to decompose srco@xmath3 , ii heated in flowing n@xmath0 at 1010c , iii annealed in flowing o@xmath0 at temperatures ranging from 1050 to 1060c and iv - finally , a prolonged anneal in flowing o@xmath0 at 1060c is performed , during which the material densifies , granularity is substantially reduced [ 12 ] and ordering within the crystal structure develops [ 13 ] . because the superconducting and magnetic properties are affected by the details of the preparation process , which in turn affect the microscopic structure , a systematic work on the synthesis of ru-1212 and the effects of sample preparation on the magnetic and superconducting properties was developed [ 14 ] . basically a procedure as described commonly in literature and sketched in fig.1 has been adopted with the aim to give insight on the formation and stability of the various phases involved in the synthesis of this complex system . each reaction step was carried out on a mgo single crystal substrate to prevent reaction with the alumina crucible . between each step the products were throughly ground and pressed into pellets . gdcu@xmath1o@xmath2 ] i - the stoichiometric oxides were first calcined in air at different temperatures , t@xmath9 , for 12 h. xrd spectra performed on these calcined samples are shown in fig.2 ( a ) , ( b ) , ( c ) . the spectra show the peaks of the ru-1212 phase whose amount increases with the temperature of the thermal treatment . there are however reflections of second phases identified as srruo@xmath3 and gd@xmath0cuo@xmath10 , with higher amounts in samples calcined at the lower temperature , diminuishing with increasing the calcination temperature . = 900 , ( * b * ) 940 and ( * c * ) 960c in air for 12h . @xmath11 ru1212 , @xmath12 srruo@xmath3 and + gd@xmath0cuo@xmath10 ; ( * d * ) after annealing in n@xmath0 for 15h @xmath13 sr@xmath0gdruo@xmath14 , @xmath15 cu@xmath0o ] ii - the pellets were then annealed in flowing nitrogen at 1010c for 12 h. the sintering in n@xmath0 gas is required to suppress the srruo@xmath7 phase [ 10 ] . this step resulted in fact in the formation of a mixture of sr@xmath0gdruo@xmath16 and cu@xmath1o independently of the starting calcined mixture from step ( i ) ( hereafter named l - serie ) . typical xrd pattern is shown in fig.2 ( d ) obtained from the sample calcined at 900c which contained the highest amount of srruo@xmath3 . no detectable traces , within the resolution of the technique , of such very stable in oxidising environment [ 10 ] impurity phase , were observed . on behalf of these results , the synthesis of ru-1212 by using sro@xmath1 as starting reagent in place of srco@xmath3 was investigated ( sample i ) . raw materials were then heated directly in n@xmath0 flow at 1010c avoiding thus the first calcination step ( i ) in air . no significant differences were obtained in the composition of the products as inferred from xrd analysis with respect to previous results shown in fig.2(d ) . the l- serie mixture was then subjected to eight successive sintering steps in flowing o@xmath0 , each one lasting 15 h , at successively increasing temperatures in the range 1030c 1085c . each successive thermal treatment was performed at a temperature about 7c higher than the previous one . in order to investigate the effects of the thermal treatments the product was quenched to room temperature at the end of every step , fully characterized , reground , pressed into pellets and subjected to the successive thermal treatment . powder xrd patterns of all our samples show ru-1212 as the major phase , with zero to some amount of srruo@xmath7 as minor impurity depending on the sample preparation condition . traces of second phase srruo@xmath3 ( 2% vol . for sample l1 ) with decreasing amount up to sample l3 were detected . single phase materials were obtained afterwards . all peaks can be indexed assuming a tetragonal lattice and table 1 lists the lattice parameters calculated for these ru-1212 samples . in fig.3 the x - ray powder diffraction pattern of sample l5 , synthesized after five sintering steps up to 1067c for a total time t= ( 15 x 5 ) = 75 h , is reported . the same xrd spectra have been obtained for sample i , subjected to a subsequent thermal treatment at 1050c for 24 h and successively to a prolonged anneal at 1060c for a week in o@xmath0 flow . lllllll sample & @xmath17c ) & a ( @xmath18 ) & c ( @xmath18 ) & @xmath19 cm ) & @xmath20(k ) & @xmath21(k ) + l1 & 1031 & 3.826(2 ) & 11.516(7 ) & 154.8 & @xmath22 & 37 + l2 & 1037 & 3.821(3 ) & 11.528(8 ) & 141.7 & @xmath22 & 44 + l3 & 1044 & 3.831(1 ) & 11.545(7 ) & 140.9 & @xmath22 & 46 + l4 & 1053 & 3.831(1 ) & 11.547(6 ) & 70.6 & 17 & 48 + l5 & 1061 & 3.828(1 ) & 11.552(6 ) & 30.6 & 25 & 49 + l6 & 1067 & 3.844(1 ) & 11.585(3 ) & 14.0 & 21 & 44 + l7 & 1073 & 3.845(1 ) & 11.610(5 ) & 23.7 & 21 & 47 + l8 & 1084 & 3.835(1 ) & 11.580(5 ) & 22.0 & @xmath22 & 45 + @xmath22 no information available below 15 k. see text for a complete discussion . + [ tab1b ] parallel checks have been performed allowing us to conclude that reaching the `` optimal '' temperature directly in one step for a time which is the sum of the corresponding partial times of each single step covered up to the same temperature does not produce the same results of the longer procedure described above . single - phase formation seems to be kinetically hindered by the slow decomposition rate of the impurities which already formed upon calcination . repeated homogenisations , related to the sequence of grinding and annealing , improve the phase purity of the material and control the superconducting behaviour . morphologically , all the l - series samples show a high grain homogeneity with clean grain boundaries as probed by sem and microprobe analyses . fig.4 shows the typical granular morphology detectable at the beginning of the thermal treatment cycles ( sample l2 ) with an average grain size of about 2 @xmath23 m . a progressive grain growth and a corresponding increase in grain connectivity due to the different thermal treatments can be observed ( fig.5 , sample l6 , it can be noticed how some grains begin to coalesce into big aggregates dispersed in an almost unchanged granular matrix ) , without reaching , by the way , complete sintering at the highest temperature . such behaviour is related to the difference between the decomposition temperature of the 1212 phase and the maximum temperature of the thermal processes considered in this work . a geometric density variation of about 5% between the first and the final bulk sample has been measured ( d @xmath24 4.2 g/@xmath25 corresponding to about 63% of the theoretical crystallographic density ) . in general resistivity measurements of ru-1212 show a superconducting transition at 45 k with a very slight upturn in the vicinity of t@xmath26 reaching zero resistivity at a lower temperature between 20 30 k. the resistivity transitions at h = 0 are much broader than those observed in many of the other htsc . a metallic behaviour , with a t linear dependence at high temperatures above 100 k , is usually observed . a magnetic transition at t@xmath27 = 132 k manifest itself as a small yet noticeable kink / minimum in the resistivity related to the onset of the magnetic ordering of the ru lattice . since the oxygen stoichiometry is pratically unchanged in ru-1212 , the annealing turns out to influence mainly the granularity and ordering within the crystal structure . previous studies have shown that the semiconductor - like upturn and the zero resistivity temperature are critically dependent on the sample processing . [ 10,15 ] . in particular , according to [ 6 ] the slight upturn in the vicinity of t@xmath26 is related to grain boundary effects . high resolution tem study on ru-1212 has shown that prolonged thermal treatment at 1060c in o@xmath0 removes most of a multidomain structure , consisting predominantly of 90 rotations , as well as significantly reduces the semiconductor - like upturn [ 15 ] . part of the superconducting transition width may be due to structural disorder . however it must be underlined that a broad superconducting transition is also expected within the spontaneous vortex phase model [ 16 ] . /dt temperature dependence for sample l4 around @xmath28 ] curves of @xmath29 t ) of selected samples ( l - serie ) considered significative , for sake of clarity , of the overall process of synthesis are shown in fig.6 . all samples exhibit weakly pronounced or local minima in the dc resistivity near the magnetic transition temperature , of the order of about 132 k. this feature is more clearly visible in the inset of the figure where the derivative of the resistivity ( sample l4 ) is plotted . at low temperatures the dc resistance shows a semiconductor - like upturn followed by a sudden decrease in resistivity starting at t@xmath30 and achieving zero resistivity state for temperatures below 30 k as reported in table 1 . there is a small increase in the zero resistivity temperature for our best sample ( l6 ) and only a small reduction in the semiconductor - like upturn . summarizing the general trend , it can be stated that the resistivity is progressively decreased and a crossover from a semiconducting to metallic normal state resistivity behaviour is observed on going from l1 to l8 sample . we underline that zero resistivity has not been reached for samples from l1 , l2 , l3 and l8 even if , considering their strong resistivity drop detected below 45 k , a r = 0 value is expected at a temperature lower than 13 k for samples l2 , l3 and l8 . a comparison between the resistivity behaviours , independently of their granular nature , has been possible because different values are not related to the sample density variations , which as already noted is almost unchanged for all the samples . a semiconducting - like transport with no indications of transition to superconductivity at low temperatures is observed for sample e in all the temperature range considered , as shown in fig . it is noteworthy that a kink in resistivity is observed in the vicinity of t@xmath27 ( arrow in fig . this anomaly is due to a reduction of spin scattering . such a behaviour is also observed in srruo@xmath3 single crystals [ 17 ] at around its ferromagnetic transition temperature . the derivative of resistivity , shown in fig . 8 for l3 l8 samples , clearly shows two overlapping maxima , indicating that the resistive transition proceeds in two steps : a high temperature contribution , associated with the thermodynamic superconducting transition temperature and another one , at a lower temperature , which critically depends on sample processing conditions [ 10,14,15 ] as well as the zero resistance temperature value . /dt temperature dependence ] since from the analysis of x - ray patterns our superconducting and non superconducting samples are indistinguishable , more insight about the physical nature of the superconducting and magnetic states is expected from magnetic measurements . the magnetic characterization of the ruthenocuprate materials is a crucial and not trivial point . magnetic measurements are obviously a key tool to observe both the superconducting and magnetic behaviour of these samples , but many years after their successful synthesis [ 1 ] and in spite of a great experimental effort devoted to this problem , many doubts still survive about the magnetic ordering present in these type of samples [ 7 , 11 , 18 , 19 , 20 , 21 , 22 ] . first of all , we recall that generally granular samples are measured : therefore , all the problems related to the granular behaviour of htsc and , in general , to the distinction between intrinsic and extrinsic properties ( intra - granular and inter - granular ) must be born in mind . a first problem encountered in the observation of the superconducting behaviour is the fact that the standard diamagnetic signals , both in the field cooled ( fc ) and zero field cooled ( zfc ) mode , are not always seen in all the samples of such compounds [ 13 , 14 , 18 , 23 ] : what is more often observed is the shielding signal , rarely the diamagnetism related to the fc procedure . both signals are quickly removed by the application of even a small external magnetic field ( few tens of gauss ) . in contrast , even when in the magnetic measurement there is no trace of superconducting behaviour , it may be observed resistively and the application of even a high external magnetic field ( up to tesla ) does not destroy it [ 24 ] . the reason for such a contradictory phenomenology may be understood bearing in mind the simultaneous presence of magnetic and superconducting ordering in these samples . this fact implies consequences both on the sample physical behaviour and on the measurement technique used to monitor it . we recall that @xmath23sr measurements [ 5 ] indicate the homogeneous presence of an internal field that , at low temperature , may reach hundreds of gauss and may give rise to a spontaneous vortex phase ( svp ) in the temperature range where it exceeds the first critical field h@xmath31(t ) [ 25 ] . in a type ii superconductor at h@xmath32h@xmath31 the meissner effect is practically never observed for the presence inside the materials of `` pinning centres '' that are able to block the flux lines and prevent their expulsion . this is the reason why the fc diamagnetic signal may be very small and its difference from the zfc signal is an indication of the critical current density that a sample can carry . a vast literature related to high t@xmath33 superconductors illustrates unambiguously this item [ 26 , 27 ] . moreover , as noted in [ 28 ] , the magnetization of ruthenocuprate materials contains magnetic signals arising from different contributions : the gd paramagnetic spin lattice , the ru spin lattice and , finally , the diamagnetic signal related to the superconducting behaviour . both for gd and ru spin lattice the antiferromagnetic ordering is coupled with a ferromagnetic component that , in the case of ru , is attributed to a canting of the lattice and in the case of gd is simply related to the presence of the net ferromagnetic moment of the ru lattice [ 20 ] . the simultaneous presence of such opposite magnetic signals makes the magnetic measurement unsuitable for the observation of the superconductivity : in fact , such measurement can not separate the magnetic signal related to superconductivity from that related to the magnetic ordering . moreover , it is clear that the application of an external magnetic field exalts the magnetic signals and depresses the superconducting one , destroying very quickly the visibility of the superconductivity . in the light of these considerations we can understand the fact that the superconducting behaviour is often observed resistively but not magnetically : it depends on the competition between two opposite magnetic signals , one related to the magnetic ordering , the other to the superconducting one . `` more superconductivity '' is obviously related to many factors : the amount of superconducting phase inside the sample , the quality of the connection between the grain that makes the shielded volume and therefore the related magnetic signal smaller or larger , and the intrinsic properties of the ru-1212 phase that , as we will see , may change in connection with the grade of order of the material . now , dealing with the experimental problems , we point out the following . in order to enhance the superconducting behaviour it is suitable to apply magnetic fields as small as possible . * * this fact , due the peculiar modalities of elaboration of the instrumentation commonly used , must be considered in detail . the first problem is the exact knowledge of the field that is effectively seen by the sample , and the second is strictly related to the complexity of the magnetic signal present in these samples . a small remanent magnetic field in the superconducting coil of the experimental set - up is often present . it may be zeroed by a procedure that , starting from a value of some tesla , applies coercive fields of decreasing intensity . in such a way the field is zeroed but for a few gauss that may be zeroed in the central point of the magnet by applying a small counterfield . anyway , a very small residual field survives and turns out to be of the order of fractions of gauss . in the light of what has been said , a real zfc measurement can not be made and , since the fc magnetic moment is about one order of magnitude greater than the zfc , also a residual field of fractions of gauss may give a considerable magnetic signal whose polarity depends on the field polarity . in addition , the basic condition of a homogeneous magnetic moment required by the squid magnetometer is not fulfilled , in particular at low temperatures , where , as a consequence of the applied field , magnetic moments of opposite polarity will be present in the sample . finally , we recall that during the measurements the sample is moved for a length that is usually of few centimetres , so that it travels in a non uniform magnetic field that makes it follow a minor hysteresis loop . if the value of the moment is not constant during the scan , an asymmetric scan wave form will be observed and the quality of the measurement will drastically degrade [ 29 ] . all we have said is illustrated in fig . 9 where magnetization measurements are reported for both zfc and fc conditions . for sake of clarity we report data for some representative samples only . the cuspid at t @xmath34 30 k marks the magnetic ordering : there is a small variation in this temperature , which is smaller in the sample with higher superconducting temperature in agreement with the literature data [ 8 ] . it is remarkable to observe the different behaviour exhibited by the various samples : l3 gives no hint of superconductivity , l5 exhibits a very clear shielding corresponding to about 75% of the maximum diamagnetic signal at @xmath35h@xmath36 = 0.5 g while at @xmath23h@xmath37 = 5 g its transition is strongly worsened , l6 shows a diamagnetic shift after an ascent of the magnetization ( probably due to the instrumental effects we outlined before , for the presence of two opposite magnetic signals of similar magnitude ) , and l8 shows a behaviour very similar to l3 . at the lowest temperatures a large contribution from gd sublattice , which orders antiferromagnetically at 2.5 k , is clearly visible in fc magnetization curves for the magnetically non superconducting samples l3 and l8 . if the superconductivity is marked by the visibility of a diamagnetic shift of the zfc or fc signals , such behaviour is surely absent in l1 , l2 , l3 , l4 and l8 while , to different extent , it is observed in l5 , l6 and l7 . in the resistivity measurements , on the contrary , all the samples show a large drop of resistivity , but at the temperature of t=13 k ( the minimum value at which we measured resistively , while magnetically we reached t=5 k ) zero is reached for l4 , l5 , l6 and l7 samples . in lower fig . 9 the fc data for samples l3 , l5 , l6 and l8 are shown . a sudden onset of a spontaneous magnetic moment appears , related to a ferromagnetic component arising from ru spin ordering in ruo@xmath0 planes . such a spontaneous magnetization develops at a temperature in the 130 - 135 k range and below 110 k it rises almost linearly as temperature decreases down to about 50 k. we remark the very similar behaviour of l3 and l8 , already observed in upper part of fig . a clear diamagnetic behaviour is seen only in l5 : at the minimum applied field of 0.2 g , and to a minimum extent * _ _ * even at 1.2 g , a diamagnetic behaviour that quickly reenters is seen in the fc curve . at 5.5 g the diamagnetic effect is only seen as a constant value hindering the gd magnetic ordering . such behaviour has been already observed [ 30 ] . vs temperature for all the samples . fitting results are shown as solid lines ] in fig . 10 we present the inverse of ruthenium susceptibility as a function of temperature for all the samples in the series l1-l8 . in the calculation of the ruthenium susceptibility we have followed the procedure suggested by butera et al . [ 22 ] . such a procedure calculates the ru susceptibility by subtracting three magnetic contributions to the experimental value : 1 ) the paramagnetic contribution from gd ions , 2 ) the core diamagnetism for the 1212 compounds as deduced from the landolt - brnstein tables , and 3 ) a temperature - independent pauli - like contribution coming from the conduction electrons . the so obtained ruthenium susceptibility is fitted by the curie - weiss relationship @xmath38 and allows to calculate both the curie temperature @xmath39 and the effective magnetic moment @xmath40 for ru atom . although a maximum content of about 2 vol.% of srruo@xmath3 impurity phase was detected from x ray analyses in sample l1 , with decreasing amount to zero for sample l4 , a similar negligible error on the absolute values of @xmath41 and @xmath42 has been calculated with no significant effect on their general behaviour . and @xmath42 calculated from a best fit of @xmath43 vs temperature data ] the obtained results are reported in fig . 11 and in table 2 as a function of the number of annealing steps that give rise to the sequence l1-l8 . .@xmath41 and @xmath44 values as a function of annnealing steps and `` superconductivity '' status of all l serie samples [ cols= " < , < , < , < , < , < " , ] @xmath22 measurement technique utilized to detect superconductivity : resistivity , zfc and fc magnetization . + @xmath45 no information available below 15 k. see text for a complete discussion . [ tab1a ] starting from l1 the @xmath39 values increase , reach a maximum ( around l5-l6 of about 160 k ) , and then decrease going up to l8 . the @xmath40 values have a specular trend , decreasing from the value 3 for l1 down to a minimum value of about 2 at l6 , and then slightly increasing once again . since the superconductivity is better observed in the samples l5 , l6 , l7 both by resistivity and magnetic measurements , these data suggest that an improved superconducting behaviour may be related to small intrinsic variations in the structure of the sample that produces smaller effective magnetic moments for ru atom and higher curie temperatures . we give here only some suggestions to be explored . the @xmath40 values derived by the best fit imply that ru is in a mixed valence state between ru@xmath46 and ru@xmath47 . such a result has been firstly proposed by liu et al . [ 21 ] through xanes spectroscopy and successively confirmed by butera et al . [ 22 ] through magnetic measurements by means of the procedure we have outlined . these results definitively contradict the hypothesis that ru exhibits an effective moment @xmath48 1 @xmath49/ru atom , as proposed in [ 4 ] . on passing from l1 up to l8 the proportion of ru@xmath50 and ru@xmath47 changes . possible consequences of this fact are : slight variations in the carriers number and , as a consequence , in the critical temperature ( see the resistivity data in fig.6 and table 1 ) , different coupling between the superconducting and the magnetic planes both in term of total magnetic moment seen by the conduction electron ( with increased or decreased pair - breaking effect ) and in term of coupling between orbitals of superconducting and magnetic electrons [ 31 ] . the origin of these variations may be found in a different degree of lattice disorder following the various annealing steps performed at different temperatures that , as we have widely observed , produce significant variations in all the physical properties . moreover , the lattice disorder can imply a certain amount of cu@xmath51ru substitutions that are a possible candidate for the observed variations of the effective magnetic moment . also the variation of @xmath52 may be the consequence of the different coupling between ru atoms following the different valence state . the same trend we have seen in @xmath40 is observed in the saturation moment as seen by measuring magnetization up to the maximum field of 5.5 tesla at t=5 k. results are shown in fig 12 . the values change from the minimum value of 6.5 @xmath53 for l5 and l6 samples up to a maximum value of 7 @xmath53 in the l2 and l8 samples . we remark that in the experimental conditions we have used the saturation is not completely reached , but the hierarchy of the saturated magnetic moments is surely correct . the magnetic and superconducting properties of ru1212 have been studied and compared for a series of samples synthesized under different conditions with the aim to find out the fundamental parameters ruling out the phase formation and its related structural and physical properties . from our experimental work it results that the optimal annealing temperature lies in a narrow temperature range around 1060 - 1065c ; further temperature increase worsens the phase formation . subsequent grinding and annealing steps up to this temperature improve the phase homogeneity . a wide range of physical properties has been obtained on quenched samples from the same batch , which differ only in the synthesis procedure parameters . no other substantial differences were detected for these samples , all showing similar compositional and structural characteristics . it emerges that the preparation method plays an important role when dealing with the magnetic and superconducting properties of this hybrid compound . so far , published data on the ru1212 phase show the same general trend for what regards the measured physical properties . because most of the samples are chemically and structurally comparable , great care must be taken in the preparation process details such as the final sintering temperature and the number of homogeneization steps ( if any ) performed up to that temperature . only samples with the same thermal history / parameters can be compared . m. murakami : ` flux pinning of melt processed ybco superconductors and their applications ' . in : _ studies of high temparature superconductors vol.9_. ed a. narlikar ( nova science publisher , new york 1992 ) pp

a systematic study on the synthesis of the ru-1212 compound by preparing a series of samples that were annealed at increasing temperatures and then quenched has been performed . it results that the optimal temperature for the annealing lies around 1060 - 1065c ; a further temperature increase worsens the phase formation . structural order is very important and the subsequent grinding and annealing improves it . even if from the structural point of view the samples appear substantially similar , the physical characterizations highlight great differences both in electrical and magnetic properties related to intrinsic properties of the phase as well as to the connection between the grains as inferred from the resistive and the curie weiss behaviour at high temperature as well as in the visibility of zfc and fc magnetic signals .

in this paper , we outline a method for computing the homology of the two double - branched covers of any properly embedded arc in the solid torus . in particular , we use this method to partially answer a question posed by krebes in @xcite . let @xmath2 be a standardly embedded solid torus @xmath6 . then a genus-@xmath0 tangle is a properly embedded arc in @xmath2 . just as we may discuss embedding ordinary tangles in @xmath7 into knots and links ( see @xcite , @xcite , and @xcite ) , we may consider embedding genus-1 tangles in knots . we say that a genus-1 tangle @xmath1 embeds in a knot @xmath4 if @xmath1 can be completed by an arc exterior to @xmath2 to form the knot @xmath4 ; that is , there exists some arc in @xmath8 such that upon gluing this arc to @xmath1 along their boundary points , we have a knot in @xmath9 which is isotopic to @xmath4 . we say that @xmath4 is a closure of @xmath1 . let @xmath10 denote a longitude for @xmath2 which is contained in @xmath11 and avoids the genus-1 tangle . a closure @xmath4 of @xmath1 is called odd ( respectively , even ) with respect to @xmath10 if @xmath12 is odd ( respectively , even ) . if @xmath10 is chosen to be the longitude which circles the central hole of @xmath2 as in fig . [ krebestangle ] , and we span the longitude @xmath10 by a disk @xmath13 filling the hole , then @xmath12 is the number of transverse intersections counted with sign of the arc which completes @xmath1 to @xmath4 with @xmath13 . thus , in this case we can say more colloquially that @xmath4 is an odd ( respectively , even ) closure with respect to @xmath10 if the arc which completes @xmath1 to @xmath4 passes through the hole of @xmath2 an odd ( respectively , even ) number of times . @xmath10 at 377 185 in @xmath2 together with a specified longitude @xmath10.,height=96 ] in @xcite , krebes asks whether the genus-1 tangle @xmath5 given in fig . [ krebestangle ] embeds in the unknot . using the following results from @xcite , we are able to partially answer this question . note that when discussing this example , we always use the longitude @xmath10 drawn in fig . [ krebestangle ] . [ ruberman ] suppose @xmath14 is an orientable 3-manifold with connected boundary , and @xmath15 where @xmath16 is an orientable 3-manifold with @xmath17 torsion . then the inclusion map @xmath18 induces an injection of the torsion subgroup @xmath19 of @xmath20 into @xmath17 . this theorem has a useful corollary which can easily be proved directly using a meyer - vietoris sequence . [ rubermancor ] let @xmath14 and @xmath16 be as in theorem [ ruberman ] but suppose @xmath21 . then @xmath22 is torsion - free . one obtains an obstruction to embedding genus-1 tangles in knots from theorem [ ruberman ] by applying the result to branched covers of @xmath2 branched over genus-1 tangles . for any genus-1 tangle @xmath1 , the homology @xmath23 is free abelian on the two generators given by the meridian @xmath24 of @xmath1 and the longitude @xmath10 of @xmath2 . for a given @xmath25 , each @xmath25-fold cover of @xmath2 branched over @xmath1 is associated to a homomorphism @xmath26 which maps @xmath24 to one . the remaining generator @xmath10 of @xmath23 may be sent to any element of @xmath27 ; we use @xmath28 to index the @xmath25-fold branched covers . so , @xmath29 denotes the @xmath25-fold cover of @xmath2 branched over @xmath1 associated to the homomorphism @xmath30 which maps @xmath10 to @xmath31 . if a genus-1 tangle @xmath1 embeds in a knot @xmath4 , then the @xmath25-fold cover @xmath32 of @xmath9 branched over @xmath4 restricts to some @xmath25-fold cover @xmath29 of @xmath2 branched over @xmath1 . in this case , we say that the closure @xmath4 induces the cover @xmath29 . then according to theorem [ ruberman ] , the torsion subgroup @xmath33 of @xmath34 injects into @xmath35 . note that if @xmath4 is the unknot , then @xmath32 is @xmath9 and according to corollary [ rubermancor ] , the torsion subgroup @xmath33 is trivial . thus , if there is any torsion in the homology of @xmath29 , then any closure of @xmath1 which induces the cover @xmath29 is not the unknot . ruberman considered using theorem [ ruberman ] to study krebes question about the genus-1 tangle @xmath5 . he found that it seems that the homology of all the cyclic covers of the solid torus , branched along this arc , is torsion - free " ( see section 5 of @xcite ) . however , our detailed computation in sections 2 and 3 shows that the two - fold branched cover @xmath36 does have torsion in its homology , and we prove the following results . [ maintheorem ] if a knot @xmath4 in @xmath9 is an odd closure of @xmath5 , then @xmath37 is divisible by @xmath38 . [ maincor ] if @xmath5 embeds in the unknot , then the unknot is an even closure of @xmath5 . where we perform a meridional twist , and the genus-1 tangle which results from the twist.,height=124 ] before further discussion , we need to make a remark about the definition of genus-1 tangles . [ embeddingremark ] note that when defining genus-1 tangles , we fix a standardly embedded solid torus @xmath2 in the 3-sphere . the reason that we restrict to a fixed embedding is that there are many ways to re - embed a solid torus inside @xmath9 . for instance , if we perform a meridional twist on @xmath2 along the disk indicated in fig . [ krebestwist ] , the image of @xmath5 under this twist can be easily seen to embed in an unknot via the exterior arc pictured in fig . [ krebestwist ] . thus it is necessary to specify the embedding of @xmath39 in the case of genus-1 tangles , and we restrict to a fixed standardly embedded solid torus in our definition . for the purposes of this paper , we restrict our attention to double - branched covers of @xmath2 branched over @xmath5 . since a homomorphism @xmath40 must map the specified longitude @xmath10 to either zero or one , there are two double - branched covers , @xmath41 and @xmath42 . we call @xmath41 the even double - branched cover because it is induced by all even closures of @xmath5 ( with respect to @xmath10 ) . similarly , since @xmath42 is induced by all odd closures of @xmath5 , we call it the odd double - branched cover . in this section , we adapt rolfsen s technique to find surgery descriptions for these double - branched covers . following @xcite , we perform surgery near a carefully selected crossing ( see fig . [ surgery ] ) in such a way that after surgery we may essentially unwind @xmath5 ( via sliding its endpoints around the boundary in the complement of @xmath10 ) so that it looks trivial . this process , illustrated in fig . [ unwind ] , results in a nice surgery description of @xmath5 inside @xmath2 . note that in the last drawing of fig . [ unwind ] , we choose to draw this surgery description in a particular way because it makes constructing branched covers easier . to make it look trivial . the surgery curve is always given the blackboard framing.,width=465 ] now we construct the odd cover , @xmath42 . construction is dictated by the homomorphism @xmath40 corresponding to the cover . if @xmath30 maps a generator of @xmath43 to a non - zero element , then we cut the solid torus along a disk transverse to that generator . thus , we have two cuts to make in the case of the odd cover . first , we cut @xmath2 along a disk which is transverse to the meridian @xmath24 of @xmath5 and whose boundary is made up of the unwound genus-1 tangle @xmath5 together with an arc in @xmath11 . then , because @xmath30 sends @xmath10 to one , we cut @xmath2 along a disk which is transverse to @xmath10 and whose boundary is contained in @xmath11 . we then take two copies of the resulting manifold and glue them together carefully to obtain a surgery description for @xmath42 . this process is illustrated in fig . [ constructodd ] . although it is not needed in the proof of theorem [ maintheorem ] , we also give a surgery description of the even double - branched cover @xmath41 in fig . [ constructeven ] . of @xmath2 branched over @xmath5.,width=374 ] .,width=393 ] now we compute the homology of the odd double - branched cover . from fig . @xmath44 we see that the surgery description for @xmath42 is given by a 2-component surgery link inside a genus-2 handlebody . we denote the components of the surgery link by @xmath45 and @xmath46 , and let @xmath47 denote the genus-2 handlebody . the complement of @xmath47 in @xmath9 is a neighborhood of the handcuff graph @xmath48 , pictured in fig . [ oddhomology ] , which is composed of loops @xmath49 and @xmath50 joined together by an arc . then the complement of @xmath51 in @xmath47 can be viewed as the complement of @xmath52 in @xmath9 . one can see that @xmath53 is isomorphic to @xmath54 which is free on four generators : the meridians of @xmath45 , @xmath46 , @xmath49 , and @xmath50 . completing the surgery by gluing in two solid tori according to @xmath45 and @xmath46 introduces two relations on these four generators , which are given by the linking numbers of @xmath45 and @xmath46 with each of @xmath45 , @xmath46 , @xmath49 , and @xmath50 . then @xmath55 is isomorphic to @xmath54 modulo these two relations , and we can get a presentation for @xmath55 using linking numbers . thus , we have the following presentation matrix for @xmath55 : @xmath56 let @xmath4 be an odd closure of @xmath5 , and let @xmath32 denote the double cover of @xmath9 branched over @xmath4 . since @xmath4 is an odd closure of @xmath5 , it induces a restriction from @xmath32 to @xmath42 . then according to theorem [ ruberman ] , we have that @xmath59 . thus @xmath60 divides @xmath61 . we are unable to use this method to restrict all closures of @xmath5 because @xmath41 has a torsion - free first homology group . indeed , the statement in remark [ embeddingremark ] allows us to see that the even cover does embed in @xmath9 and so must have torsion - free first homology . of course , this can be verified by deriving a presentation for the homology of @xmath41 using the procedure above .

a genus-@xmath0 tangle @xmath1 is an arc properly embedded in a standardly embedded solid torus @xmath2 in the 3-sphere . we say that a genus-@xmath0 tangle embeds in a knot @xmath3 if the tangle can be completed by adding an arc exterior to the solid torus to form the knot @xmath4 . we call @xmath4 a closure of @xmath1 . an obstruction to embedding a genus-1 tangle @xmath1 in a knot is given by torsion in the homology of branched covers of @xmath2 branched over @xmath1 . we examine a particular example @xmath5 of a genus-1 tangle , given by krebes , and consider its two double - branched covers . using this homological obstruction , we show that any closure of @xmath5 obtained via an arc which passes through the hole of @xmath2 an odd number of times must have determinant divisible by three . a resulting corollary is that if @xmath5 embeds in the unknot , then the arc which completes @xmath5 to the unknot must pass through the hole of @xmath2 an even number of times .

the recent discovery of na@xmath2coo@xmath3h@xmath4 o superconductor @xcite has renewed interest in the study of exotic superconductivity ( sc ) . motivated by the triangular lattice structure of spin-@xmath6 co atoms of the reference coo@xmath4 layer , several authors @xcite have quickly proposed that it might be a realization of the long sought spin - liquid based superconductivity , originally proposed for high - t@xmath7 superconductors by anderson et al.@xcite . indeed several theoretical calculations @xcite indicate that the frustration of the antiferromagnetism ( afm ) on the triangular lattice of co favors the d+id superconducting symmetry . for a new superconducting material , the first step toward laying down the theory of the superconducting state is the experimental identification of the sc gap symmetry . nuclear spin - lattice relation rate 1/t@xmath1 and knight shift measurements have been the successful tools for this purpose and several measurements of 1/t@xmath1 and knight shift of this compound have already been reported . at present , however , these reports appear inconsistent with each other . regarding the coherence peak of co@xmath5 1/t@xmath1 , for example , kobayashi et al . @xcite and waki et al . @xcite reported the existence of a coherence peak ; the maximum value of 1/t@xmath1 t is about 1.1 to 1.5 times of the normal state value while an ideal s - wave case has almost twice of the normal state value@xcite . however , fujimoto et al . @xcite recently reported no coherence peak . furthermore , they reported more precise @xmath5 co 1/t@xmath1 below t@xmath7 and found a substantial region of linear temperature dependence in 1/t@xmath1 at low temperatures and t@xmath8 behavior near and below t@xmath7 . to fit their data a residual density of states ( dos ) of 0.65n(0 ) was needed assuming an impure d - wave sc state . from this calculation , these authors concluded that the non s - wave gap with the lines of nodes is most consistent with their 1/t@xmath1 data and the compatibility of the d+id symmetry gap having a full gap remains as a question . in this note , we calculated 1/t@xmath1 for d+id gap with impurities using self - consistent t - matrix approximation and quantitatively compared our results with experiments reported up to date . since the impurity effects on various symmetry gaps has been studied by many authors @xcite and the generic behaviors are well known , we summarize them . within t - matrix approximation , the impurity scattering renormalizes the self - energies both of normal and anomalous components . in nambu notation , they correspond to @xmath9 , and @xmath10 , respectively ; @xmath11 and @xmath10 are the two separate components of the normal self - energy according to particle - hole transformation and @xmath12 is the anomalous self - energy component . among them , the renormalization of @xmath10 usually vanishes for particle - hole symmetric band , which is a good approximation for most cases . for an isotropic s - wave gap , both @xmath11 and @xmath12 are renormalized and their effects largely cancel each other in the renormalized gap equation ; as a result the impurity effect is null for low temperature sc properties such as the changes of the order parameter ( op ) magnitude ( @xmath13 ) and t@xmath7 , and the residual dos etc . for a sign changing op such as the d - wave gap , the renormalization of @xmath12 vanishes , too , due to symmetry , and the impurity effects appear only in the renormalization of @xmath11 the normal self - energy . the case of d+id gap is interesting in that it has a full gap all over the fs and should display similar thermodynamic behaviors as in an isotropic s - wave gap . however , regarding the impurity effects , it is expected to show a mixed behavior of a d - wave gap and a s - wave gap because the sign changing op leads to a vanishing @xmath12 renormalization despite a full gap . we find that for unitary impurities the residual dos is created in the d+id gap as easily as in the d - wave gap . this happens because the resonant scattering process of unitary impurity is more efficient at lower dos . as a result , d+id gap displays a similar t - linear behavior in 1/t@xmath1 at low temperatures with the same amount of unitary impurities as in the simple d - wave case . therefore , the observation of the t - linear behavior in 1/t@xmath1 at low temperatures@xcite itself can not rule out the d+id symmetry gap on coo@xmath4 superconductor . moreover , in contrast to the d - wave gap the d+id gap with a smaller amount of impurities displays a reduced coherence peak reflecting full opening of the gap around the fs . putting together our theoretical results , we conclude that the d+id symmetry gap with unitary impurities provides the most consistent agreement with the currently available nuclear spin - lattice relaxation data . we also studied the effects of born limit impurity on both symmetry gaps . we find that the response to born limit impurities is qualitatively different for d+id and d - wave symmetry gaps , and both cases can not fit the experimental data . the effect of the impurity scattering is included with t - matrix approximation@xcite . as we explained above , for sign - changing op and assuming particle - hole symmetric band , we just need to calculate @xmath14 , where @xmath15 , @xmath16 the normal dos at the fermi energy , @xmath17 the impurity concentration . @xmath18 is defined in matsubara frequency as @xmath19},\ ] ] and @xmath20 where @xmath21 and the scattering strength parameter @xmath22 is related to the s - wave phase shift @xmath23 as @xmath24 . with this @xmath18 the following gap equation is solved self - consistently . @xmath25 for d+id gap equation , we assume the pairing potential @xmath26 to be a constant because thermodynamic properties of d+id sc are determined by the absolute magnitude of the gap @xmath27 which is the same as the isotropic s - wave gap . the property of the d+id gap in the above gap equation manifests itself only as the absence of @xmath12 renormalization . for comparison , we also calculate the d - wave case with a proper pairing potential @xcite , which also has no @xmath12 renormalization due to the sign - changing op . the only technical difference of the impurity effects on d+id- and d - wave symmetry gaps is the normal self - energy correction due to impurities in the sc states with a full gap and lines of nodes in the gap , respectively . with the gap function @xmath28 and @xmath29 obtained from eq.(1 ) ( @xmath29 is analytically continued from @xmath30 by pad approximant method ) , we calculate the @xmath0 nuclear spin - lattice relaxation rate @xcite @xmath31,\end{aligned}\ ] ] where @xmath32 and @xmath33 means the average over the fs . the first term in the bracket of eq . ( 4 ) is @xmath34 . the second term vanishes in our calculations because of the symmetry of the op . to calculate @xmath35 using eq.(4 ) , we need the full temperature dependence of the gap function @xmath28 . the gap equation eq . ( 3 ) is basically the bcs gap equation , therefore it gives the bcs temperature dependence for @xmath28 and @xmath36 and @xmath37 for the d+id- and d - wave gap solutions , respectively . we use the phenomenological formula , @xmath38 ; @xmath39 , with parameters @xmath40 and @xmath41 . the temperature dependence of @xmath42 ( @xmath43 ) is similarly extrapolated : @xmath44 , where @xmath45 is the normal state @xmath18 . then we only need to calculate @xmath28 and @xmath18 at zero temperature . in our numerical calculation , the bcs value @xmath46 and the ratio @xmath47 for d+id gap and @xmath48 for d - wave gap are used , respectively . note that the final results are insensitive to the precise value of @xmath40 . fig.1(a ) shows the normalized nuclear spin - lattice relaxation rate 1/t@xmath1 for d+id sc gap with varying impurity concentration of unitary impurities ( c=0 ) . the normalization is chosen so that the value of 1/t@xmath1 at t@xmath7 is 100 in arbitrary units . the experimental data ( black circles ) are also normalized in the same fashion . the inset shows the corresponding dos . the key result we find is that although d+id gap develops a full gap as in an isotropic s - wave gap , the unitary impurity scattering produces resonance states inside the gap ( at @xmath49 in our case of c=0 and a particle - hole symmetric band ) . a finite amount of unitary impurities , therefore , quickly produces a residual dos around @xmath49 inside the gap . as a result , the nuclear spin - lattice relaxation rate 1/t@xmath1 develops a growing t - linear region at low temperatures with increasing impurity concentration . the high - temperature 1/t@xmath1 below t@xmath7 initially follows the pure d+id curve ( black open squares ) with a small amount of impurities , but with increasing impurity concentration it follows a nongeneric power law dependence @xcite . with the impurity concentration of @xmath50 ( @xmath13 is the absolute magnitude of the pure d+id gap at zero temperature ) , the residual dos at @xmath49 n(0 ) is about 0.63 n@xmath51 ( n@xmath51 is the normal state dos at fermi level ) . compared with the experimental data of @xmath5co 1/t@xmath1 from t. fujimoto et al . ( ref.[8 ] ) it provides a reasonably good fit : the low temperature t - linear behavior and its magnitude , the high temperature power law ( experimental fit was t@xmath52 ) , and the absence of the coherence peak . therefore , the experimental observation of the substantial region of t - linear behavior @xcite in @xmath5co 1/t@xmath1 of na@xmath2coo@xmath3h@xmath4 o does nt rule out the possibility of d+id gap in this compound . furthermore , our calculations for d+id gap provide a resolution for the conflicting observation of the coherence peak by other groups @xcite . first , without impurities ( black open squares ) the height of the coherence peak is about 1.46 times of the value of 1/t@xmath1 at t@xmath7 , which is a little smaller than the value for isotropic s - wave case . this is because the second term in eq.(4 ) vanishes due to the symmetry of the d+id op . with impurities the coherence peak is quickly suppressed and there is a small peak visible with a smaller amount of impurities ( for example , see @xmath53 ( open purple squares ) ; note the logarithmic scale ) . we notice that there is a systematic trend of the height of the coherence peak with t@xmath7 of the samples used for @xmath5co 1/t@xmath1 measurement : no coherence peak with t@xmath7=3.9 k @xcite ; coherence peak with height 1.1 times of the value of 1/t@xmath1 at t@xmath7 with t@xmath7=4.5 k @xcite ; coherence peak height 1.4 times of the value of 1/t@xmath1 at t@xmath7 with t@xmath7=5.0 k @xcite . therefore , we can understand the seemingly inconsistent experimental reports on @xmath5co 1/t@xmath1 by different groups with d+id gap and as a result of the different samples with different scattering rates@xcite . we also show the d - wave results with unitary impurities ( c=0 ) in fig.1(b ) . the overall shape of dos for d - wave gap with and without impurities is qualitatively different from the case of d+id gap . nevertheless , the accumulation of the residual dos n(0 ) with impurities are very similar in magnitude , which is the main property determining the low temperature 1/t@xmath1 behavior . as a result , we can also fit the data of fujimoto et al . with a d - wave gap with a similar amount of impurities ( the necessary @xmath54 value should be between 0.32 and 0.16 , see fig.1(b ) ) ; the residual dos n(0 ) with @xmath55=0.32 is about 0.73 n@xmath51 . however , the d - wave results can not explain the observation of the coherence peak seen by some experimental groups @xcite . for completeness , we show the same calculations in the born limit ( c=1 ) in fig.2(a ) and fig.2(b ) . both d+id and d - wave gaps with born limit impurities does nt produce a close fit to the experimental data unless @xmath56 . even with @xmath57 , the detailed line shape of the calculated 1/t@xmath1 does nt quite fit the experimental data for both gaps . moreover , the reduction of t@xmath7 and the gap magnitude @xmath13 with @xmath58 is about 50 @xmath59 or more for both gap cases , which is not consistent with experiments . in summary , we have calculated the nuclear spin - lattice relaxation rate 1/t@xmath1 for both d+id- and d - wave superconducting states with impurities . we found that unitary impurities produce the residual density of states inside the gap due to the resonant scattering for both cases . interestingly , we found that the rate of accumulation of the residual density of states with unitary impurities is similar for both sc gaps . as a result , both d+id and d - wave gaps produce the t - linear behavior in 1/t@xmath1 at low temperatures and can fit the experimental data of fujimoto et al . @xcite with the unitary impurity concentration of @xmath60 . however , considering observations of the coherence peak in @xmath5co 1/t@xmath1 by other groups @xcite and a systematic trend of the height of the coherence peak with t@xmath7 of the samples , we conclude that the d+id symmetry gap is the most consistent with the currently available nuclear spin - lattice relaxation data of na@xmath2coo@xmath3h@xmath4 o. we thank prof . g .- zheng for providing the experimental data and discussions , in particular , about the details of experimental data and interpretation and dr . zhu and prof . b. s. shastry for discussions . this work was supported by us doe . was partially supported by the korean science and engineering foundation ( kosef ) through the center for strongly correlated materials research ( cscmr ) ( 2002 ) and through the grant no . 1999 - 2 - 114 - 005 - 5 . permanent address : department of physics , chonnam national university , kwangju 500 - 757 , korea . k. takada , h. sakurai , e. takayama - muromachi , f. izumi , r.a . dilanian and t. sasaki , nature * 422 * , 53 ( 2003 ) . g. baskaran , cond - mat/0303649 ; brijesh kumar and b. sriram shastry , cond - mat/0304210 ; qiang - hua wang , dung - hai lee and patrick a. lee , cond - mat/0304377 . masao ogata , cond - mat/0304405 ; jian - xin zhu and a. v. balatsky , cond - mat/0306253 . y. kobayashi , m. yokoi and m. sato , cond - mat/0305649 . t. waki , c. michioka , m. kato , k. yoshimura , k. takada , h. sakurai , e. takayama - muromachi and t. sasalki , cond - mat/0306036 . m. tinkham , p 83 , _ introduction to superconductivity _ , mcgraw - hill , inc . t. fujimoto , g .- q . zheng , y. kitaoka , r.l . meng , j. cmaidalka and c.w . chu , cond - mat/0307127 . l. j. buchholtz and g. zwicknagl , phys . b * 23 * , 5788 ( 1981 ) ; l. s. borkowski and p. j. hirschfeld phys . b * 49 * , 15404 ( 1994 ) ; r. fehrenbacher and m. r. norman , phys . rev . b * 50 * , 3495 ( 1994 ) . in fact , the temperature dependence of @xmath0 at high temperatures near t@xmath7 is not a generic property for any symmetry gaps ; it is determined by the temperature slope of @xmath61 near t@xmath7 , elastic and inelastic scattering rates , and the anisotropy of the gap , etc . often quoted t@xmath8 behavior of @xmath0 for a pure d - wave gap is a generic property at low temperatures reflecting the lines of nodes , but the same t@xmath8 behavior observed at high temperatures is a result from combined effects mentioned above . nevertheless , empirically once the coherence peak is suppressed the high temperature power law of @xmath62 can be fit with @xmath63 between 2 and 3 for the dirty d+id and d cases .

we calculated the nuclear spin - lattice relaxation rate @xmath0 for the d+id superconducting state with impurities . we found that small amount of unitary impurities quickly produces the residual density of states inside the gap . as a result , the t - linear behavior in 1/t@xmath1 is observed at low temperatures . our results show that the d+id pairing symmetry of the superconducting state of na@xmath2coo@xmath3h@xmath4 o is compatible with recent @xmath5co 1/t@xmath1 experiments of several groups . 2

quantum coherence is one of important topics of the quantum theory and has been developed as an important quantum resource in recent years ( @xcite ) . furthermore , it had been linked to quantum entanglement in many quantum phenomenon and plays an important role in fields of quantum biology and quantum thermodynamics ( @xcite-@xcite ) . in finite dimensional systems , the theory of ( quantifying ) quantum coherence has been undertaken by many authors ( @xcite-@xcite ) . quantum operations relative to coherence have become important approaches and resources , such as incoherent operations . recently , coherence breaking channels for the finite dimensional case , as a subclass of inherent operations , are defined and characterized in @xcite . as we know , continuous variable ( cv ) quantum systems are fundamental important from theoretical and experimental views . in particular , gaussian states can be produced and managed experimentally easy . several researchers focused on the theory of ( quantifying ) quantum coherence of gaussian states ( @xcite-@xcite ) . furthermore , gaussian maximally incoherent channels also be characterized in @xcite . in the paper , we characterize gaussian coherence - breaking channels and focus on some relative problems . the article is organized as follows . in section 2 , we give a complete characterization of gaussian coherence - breaking channels . furthermore , we show the inclusion relation among gpptcs , gcbcs , gebcs , gcqcs and gqccs . in section 3 , we discuss the property of the tensor product of a gcbc and arbitrary a gaussian channel with entanglement inputs , and show an observation that the tensor product of gcbc and arbitrary a gaussian channel maps all input states into product states , and its connection to channel capacity . we first give some necessary notations on gcbcs . * definition 2.1 * let @xmath1 be the complex separable infinite dimensional hilbert space , @xmath2 the trace class operators on @xmath1 . for the fixed reference basis @xmath3 , a state @xmath4 is incoherent if @xmath5 , otherwise , it is coherent . denote by @xmath6 the set of all incoherent states and @xmath7 the set of incoherent gaussian states . a gaussian channel @xmath8 is coherence - breaking ( cb ) if @xmath9 is always incoherent for arbitrary a gaussian input state @xmath10 . what is the characterization of gcbcs ? how can we arrive at it ? an @xmath0-mode gaussian state @xmath4 is represented by @xmath11 where @xmath12 is the weyl operator , its covariance matrix ( cm ) @xmath13 is a @xmath14 real symmetric matrix and the displacement vector @xmath15 is a @xmath16-dimensional real vector . so a gaussian state is described by its covariance matrix and displacement , we write @xmath17 $ ] . if @xmath18 is an @xmath0-mode ( bosonic ) gaussian channel , @xmath9 have the following covariance matrix and displacement : @xmath19 where @xmath20 is a @xmath16 dimensional real vector , @xmath21 are @xmath22 real matrices satisfying @xmath23 , @xmath24 is the transpose of @xmath25 and @xmath26 so we write a gaussian channel as @xmath27 . in the following theorem , we give a characterization of @xmath0-mode gcbcs . let us recall the boolean matrix is the matrix with all elements in @xmath28 and the column stochastic matrix is the matrix satisfying the sum of elements in each a column is 1 . * theorem 2.1 * _ for an @xmath0-mode gaussian channel @xmath29 , @xmath18 is coherence - breaking if and only if @xmath30 , @xmath31 and there exist scalars @xmath32 such that @xmath33 in particular , every one - mode gaussian coherence - breaking channel has the form @xmath34 . _ * noting that a single - mode gaussian state is incoherent ( thermal ) iff it has the diagonal covariance matrix , zero displacement and squeezing , and the @xmath0-mode gaussian incoherent state is the tensor product of single - mode gaussian incoherent states ( @xcite , @xcite ) . therefore , the if " part is obvious , let us check that the only if " part . for arbitrary an @xmath0-mode coherence - breaking gaussian channel @xmath27 , @xmath35 . since every gaussian incoherent state has zero displacement , it follows that @xmath36 . so without loss of generality , we assume that the gaussian channel in the remained proof has zero displacement . next let us begin with the proof of the claim in the part b of the fourth section in @xcite for the sake of self - containment . let @xmath37 , @xmath38 and each @xmath39 are @xmath40 real matrices . noting that an @xmath0-mode incoherent gaussian state @xmath4 has the covariance matrix @xmath41 since @xmath18 maps arbitrary incoherent states to themselves , we have that for each a set @xmath42 , there exist a set @xmath43 such that @xmath44 where as the covariance matrix of the thermal state , @xmath45s and @xmath46s are larger than 1 . it follows that for each @xmath47 , @xmath48@xmath49 furthermore , by arbitrariness of each @xmath50 , we have that @xmath51 and @xmath52 and thus , there exist unique @xmath53 in the @xmath47th column of the block matrix @xmath37 such that @xmath54 it follows from arbitrariness of @xmath50 that @xmath55 and @xmath56 , where @xmath57 is a @xmath58 orthogonal matrix . now we can describe the structure of @xmath59 as follows , for each @xmath47 , there exist the only one number @xmath60 such that @xmath61 in the @xmath47th column , while @xmath62 for @xmath63 . it means that for @xmath64 , there exist n single - mode gaussian channels @xmath65 and an @xmath66 column stochastic boolean matrix @xmath67 such that @xmath68 and @xmath69 where @xmath70 for the @xmath58 orthogonal matrix @xmath71 and @xmath72 i.e. , @xmath73 to the end , it only need to show @xmath30 , i.e. , each @xmath74 is zero . now let us focus on the operation on each a mode in ( 2.6 ) . according to the column stochastic boolean block matrix @xmath75 and ( 2.6 ) , it is reasonable to assume that for the transform @xmath76 on the @xmath77th mode in ( 2.6 ) , there is a subset @xmath78 such that for arbitrary input gaussian product state @xmath79 with its covariance matrix @xmath80 , @xmath81 where @xmath82 where @xmath83 is 2@xmath842 real orthogonal matrix . in order to show @xmath30 , it is enough to show each @xmath85 . next we will show @xmath85 in ( 2.8 ) . as we know , a 2@xmath842 real orthogonal matrix @xmath83 has to be one of the following forms : @xmath86 for each one - mode gaussian state @xmath87 , suppose that its cm @xmath88 where @xmath89 , and @xmath90 . next we deal with the two cases respectively . if @xmath83 has the first form in ( 2.9 ) , it follows from ( 2.7 ) that @xmath91 has to be diagonal for arbitrary one - mode input covariance matrix @xmath92 . it follows from a short computation that @xmath93=0 . \eqno ( 2.10)\ ] ] assume on the contrary that there is at least one scalar @xmath94 such that @xmath95 . taking @xmath96 and @xmath97 for all @xmath98 , we have that by ( 2.10 ) , @xmath99=0.\eqno(2.11)\ ] ] taking @xmath100 and @xmath101 in ( 2.11 ) , then @xmath102 , that is , @xmath103 . it follows that @xmath104 is always zero in eq ( 2.11 ) . taking @xmath105 again , we have either @xmath106 or @xmath107 . it follows from @xmath103 that @xmath108 . thus @xmath109 is not orthogonal . it is a contradiction . so all @xmath85 . if @xmath83 has the second form , it still follows from ( 2.7 ) that @xmath91 has to be diagonal for arbitrary one - mode input covariance matrix @xmath92 . it follows that @xmath110=0.$ ] it follows from a similar discussion to ( 2.11 ) that all @xmath85 . in summary , @xmath85 for all @xmath111 . thus @xmath30 . we complete the proof . @xmath112 next a picture about inclusion relation between gcbcs , gpptcs , gebcs , gcqcs and gqccs is showed . without loss of generality , next we assume that gaussian channels have zero displacement vectors in the remained part of the section . let us begin from the definition of several kinds of gaussian channel . * definition 2.2 * ( @xcite ) let @xmath113 be separable infinite dimensional hilbert spaces , a gaussian channel @xmath114 is entanglement - breaking ( gebc ) if @xmath115 is always separable for arbitrary a gaussian input state @xmath116 . @xmath18 is called the gaussian positive partial transpose channel ( gpptc ) if @xmath115 has the positive partial transpose for arbitrary a gaussian input state @xmath117 . * definition 2.3 * ( @xcite , @xcite ) a gaussian channel @xmath118 is classical - quantum ( gcqc ) if @xmath119 , where @xmath120 is the lebesgue measure and @xmath121 is the dirac s system satisfying @xmath122 . it is a direct analogy of cq maps in the finite dimensional systems . a gaussian channel @xmath114 is quantum - classical ( gqcc ) if @xmath123 , that is , such a channel maps quantum states to probability distributions on @xmath124 . where @xmath125 , @xmath126\phi_t(\omega){\rm d}^{2n}\omega$ ] and the operator characteristic function @xmath127.\ ] ] we denote by @xmath128 the sets of gpptcs , gebcs , gcqcs , gqccs and gcbcs respetively . next we uncover the inclusion relation among kinds of gaussian channels . let us start from the general gaussian channels . an @xmath0-mode gaussian channel @xmath129 satisfies @xmath130 where @xmath131 and @xmath132 . noting also that a gaussian channel @xmath133 is entanglement - breaking ( gebc ) if and only if there exist @xmath134 such that @xmath135 furthermore , @xmath18 is ppt iff @xmath136 ( @xcite ) . it follows that @xmath137 holevo ( @xcite ) showed that @xmath29 is a gcqc if @xmath138 and a gqcc if it satisfies @xmath139 a direct observation and theorem 2.2 say that @xmath140 next let us check that @xmath141 . if @xmath142 satisfies that @xmath143 , taking @xmath144 and @xmath145 . it follows that @xmath146 . so @xmath141 . finally , we have that @xmath147 the inclusion relation is described in figure 1 . next we are interested in gaussian coherence breaking channels with assistant entanglement inputs . such property is helpful to understand relative problems on channel capacities ( @xcite- @xcite ) . without loss of generality , we assume that gaussian channels in the section are with zero displacement . an observation first is obtained as follows . * theorem 3.1 * _ for arbitrary a gaussian channel @xmath148 on the system @xmath149 and an @xmath0-mode gaussian coherence breaking channel @xmath18 on quantum system @xmath150 , @xmath151 is always a product state for arbitrary input gaussian state @xmath117 on the composite system @xmath152 . _ * suppose @xmath153 and @xmath154 has the cm @xmath155 , then @xmath156 has the following cm @xmath157 where @xmath158 is the eigenvalue of thermal state in the @xmath77th mode of the system @xmath159 . it follows that @xmath160 thus , @xmath161 is a product state . we complete the proof . @xmath112 the observation says that the set of gcbcs has to be a proper subset of the set of gebcs . in order to link to applications into the theory of channel capacities , it is need to discuss the property of the @xmath162tensor - product for gcbcs and arbitrary channels with entangled inputs . suppose that the gaussian channel @xmath148 has the form @xmath163 with the zero displacement vector and @xmath164 acts on bipartite system @xmath165 . it follows that the gaussian channel @xmath166 can be described based on covariance matrices of input states as follows @xmath167 where diag@xmath168 , @xmath169 is the covariance matrix for input states of @xmath164 , @xmath170 is the eigenvalue of thermal state in the @xmath77th mode of the system @xmath159 . it follows that the cm @xmath171 of the output state @xmath172 without loss of generality , write the modulation covariance matrix for the gaussian channel @xmath173 @xmath174 ( for example , see @xcite , @xcite ) . similar to the above discussion , we have that the modulated output state has the covariance matrix @xmath175 @xmath176)}= \begin{pmatrix } \oplus_{i=1}^n \omega_i i_2 & 0 \cr 0 & x_{\psi } ( b+\nu'_{\rm mod } ) x_{\psi}^t + y_\psi \cr \end{pmatrix}^{\oplus k } , \eqno(3.1)\ ] ] where the @xmath177 is the submatrix of @xmath178 corresponding to @xmath67 . recall that capacity of quantum channels is a core topic in quantum information , and additivity question for channel capacity arises from the following one : whether or not the entangled inputs can improve the classical capacity of quantum channels ( @xcite-@xcite ) . the classical @xmath179-capacity of a channel @xmath18 is defined as @xmath180 , \eqno ( 3.2)\ ] ] where @xmath181 denote quantum entropy . the classical capacity @xmath182 is @xmath183 if @xmath18 is gaussian channel , its gaussian classical @xmath179-capacity is defined as @xmath184 where the input @xmath185 is @xmath0-mode gaussian state , @xmath186 is the so - called averaged signal state , @xmath187 is probability measure . note that the right side of eq ( 3.4 ) may be infinite . it is obvious that the gaussian classical capacity is a lower bound of classical capacity . the additivity of channel capacity can be described as @xmath188 where @xmath18 and @xmath148 are channels and @xmath189 denotes channel capacity ( @xcite ) . it is mentioned that the additivity of classical capacity for gebcs had been discussed in ( @xcite , @xcite ) . here , we are interested in giving a shorter proof of additivity of classical capacity of the tensor product of a gcbc and arbitrary a gaussian channel . for the gaussian channel @xmath148 has the form @xmath190 and the gcbc @xmath18 , it follows from ( 3.1 ) that @xmath191)-s(\rho[\nu_{\rm out}])\\ & = ks(\rho[\oplus_{i=1}^n \omega_i i_2]\otimes \rho[x_{\psi } ( b+\nu'_{\rm mod } ) x_{\psi}^t + y_\psi])-\\ & \ \ \ \ ks(\rho[\oplus_{i=1}^n \omega_i i_2]\otimes \rho[x_{\psi } b x_{\psi}^t + y_\psi ] ) \\ & = ks(\phi(\bar{\rho}_a))+ks(\psi(\bar{\rho}_e))-ks(\phi(\rho_a))-ks(\psi(\rho_e ) ) \\ & = k\chi(\phi)+k\chi(\psi ) . \end{array}\ ] ] it follows from that the gaussian classical capacity and gaussian classical @xmath179-capacity are both additive for the tensor product of a gcbc and arbitrary a gaussian channel . in finite dimensional systems , the additivity properties have been proven for many examples , such as entanglement breaking channels , unital qubit channels , depolarizing channels , and contravariant channels ( @xcite ) . for classical ( @xmath179- ) capacity in the infinite dimensional case , the additivity question is affirmatively answered for the tensor product of entanglement breaking channels and arbitrary quantum channels ( @xcite , @xcite ) . for gaussian channels , the classical capacity with gaussian inputs is considered as a lower bound of the classical capacity . the gaussian classical capacity of some special kinds of gaussian channels have been proved to be additive , such as the tensor product of lossy channels @xcite , classical noise channels and thermal noise channels @xcite , the fiducial channels ( @xcite ) . here , we will pay attention to additivity of quantum capacity and gaussian classical ( @xmath179- ) capacity for gcbcs . the quantum capacity @xmath192 of a channel @xmath18 is defined as @xmath193 where @xmath194 , the coherent information @xmath195 and entropy exchange @xmath196 ( @xmath117 is the joint entangled input state shared by system @xmath159 and a purifying reference system @xmath197 ) . here , we have given a complete characterization of arbitrary @xmath0-mode gaussian coherence breaking channels ( gcbcs ) , which constitute a proper subclass of gaussian entanglement breaking channels . indeed , gaussian coherence breaking channels have more rigorous properties , comparing with gaussian entanglement breaking ones . an obvious witness is our observation that the tensor product of a gcbc and arbitrary a gaussian channel maps all input states into product states . we also establish the inclusion relations among gcbcs and other common kinds of gaussian channels . furthermore , we discuss the property of the @xmath162tensor - product of gcbcs and arbitrary channels with entangled inputs , and show a simple application in the research on the theory of the channel capacity . * acknowledgements * thanks for comments . the work is completed when kan he studies at university of technology sydney as a visiting scholar supported by china scholarship council ( 2016 - 2017 ) . this work supported partially by program for the outstanding innovative teams of higher learning institutions of shanxi ( 2014 ) . p. wikliski , m. studzi@xmath198ski , m. horodecki , and j. oppenheim , limitations on the evolution of quantum coherences : towards fully quantum second laws of thermodynamics , phys . 115(2015 ) , 210403 j. schfer , e. karpov , r. garca patrn , o.v . pilyavets , n. j. cerf , equivalence relations for the classical capacity of singlemode gaussian quantum channels . 111(2013 ) , p. 0305031 - 0305035 j. eisert , m. m. wolf , gaussian quantum channels . in : quantum information with continuous variables of atoms and light , edited by n. j. cerf , g. leuchs , e. s. polzik , imperial college press , london . 1(2007 ) , p. 23 - 42 c. h. bennett , c. a. fuchs , j. a. smolin , entanglement - enhanced classical communication on a noisy quantum channel , in : quantum communication , computing and measurement , proc . qcm96 , ed . by o. hirota , a. s. holevo and c. m. caves , new york : plenum 1997 , pp .

in the paper , we give a characterization of arbitrary @xmath0-mode gaussian coherence - breaking channels ( gcbcs ) and show the tensor product of a gcbc and arbitrary a gaussian channel maps all input states into product states . the inclusion relation among gcbcs , gaussian positive partial transpose channels ( gpptcs ) , entanglement breaking channels ( gebcs ) , gaussian classical - quantum channels ( gcqcs ) and gaussian quantum - classical channels ( gqccs ) , is displayed .

boolean functions play an important role in various formal methods for specification , verification and analysis of software systems . in program analysis , boolean functions are often used to approximate properties of the set of states encountered at a given program point . for example , a conjunction @xmath2 could specify that variables @xmath3 and @xmath4 satisfy some property whenever control reaches a given program point . a boolean function @xmath5 could specify that if @xmath6 is satisfied at a program point ( perhaps depending on the unknown inputs to the program ) then also @xmath7 is satisfied . a disjunction @xmath8 could arise as a consequence of a branch in the control where @xmath6 and @xmath7 approximate properties of the ` then ` and ` else ` branches respectively . for program analysis using boolean functions , we often consider the _ positive _ boolean functions , @xmath1 . namely , those for which @xmath9 ( denoting @xmath10 and @xmath11 by @xmath12 and @xmath13 respectively ) . this restriction is natural as , due to the element of approximation , the result of an analysis is not a `` _ _ yes / no _ _ '' answer , but rather a `` _ _ yes / maybe not _ _ '' answer . in this case there is no `` negative '' information . sophisticated @xmath1-based analyzers implemented using binary decision diagrams @xcite have been shown @xcite to give good experimental results with regards to precision as well as the efficiency of the analyzers . however , scalability is a problem and inputs ( programs ) for which the analysis requires an exponential number of iterations or exponentially large data structures are encountered @xcite . the domain , @xmath0 , of definite boolean functions is a subdomain of @xmath1 . these are the positive functions whose sets of models are closed under intersection . the domain @xmath0 is less expressive than @xmath1 . for example , the formula @xmath14 is not in @xmath0 . however , @xmath0-based analyzers can be implemented using less complex data structures and can be faster than @xmath1-based analyzers . for goal dependent groundness analyses ( where a description of the inputs to the program being analyzed is given ) @xmath0 has been shown to provide a reasonable tradeoff between efficiency and precision @xcite . the work described in @xcite illustrates a series of pathological inputs for @xmath1-based groundness analysis . that paper defines a predicate @xmath15 using @xmath16 clauses and illustrates that its @xmath1-based groundness analysis requires @xmath17 iterations however , given that the size of the program ( the total number of arguments ) , is quadratic in @xmath16 ( @xmath18 ) , the number of iterations is sub - exponential in the size of the input ( @xmath17 or @xmath19 ) . it has been suggested that @xmath0 analyses might provide better scalability properties than @xmath1 due to the restriction to functions whose models are closed under intersection . this note makes two contributions : 1 . it demonstrates that the worst - case behavior of a @xmath0-based analysis is ( at least ) as bad as that described in @xcite for @xmath1-based analyses ; and 2 . it demonstrates that the worst - case behavior of a @xmath1-based analysis is exponential in the size of the input . we have not succeeded to demonstrate a worst - case analysis for @xmath0 for which the number of iterations is exponential in the size of the input , nor to prove that @xmath0-based groundness analysis has sub - exponential worst - case behaviour . this remains an open problem . consider an @xmath16-ary boolean function @xmath20 . a model @xmath21 of @xmath20 can be viewed as a sequence @xmath22 of zero s and one s such that @xmath23 . for the sake of our construction , we order @xmath16-ary models according to their value as @xmath16-digit binary numbers . so a model @xmath24 is smaller or equal to a model @xmath25 if and only if the binary number corresponding to @xmath24 is less or equal to the binary number corresponding to @xmath25 . the intersection of models is defined as usual so that @xmath26 where @xmath27 if and only if @xmath28 . let us first comment on the series of programs which demonstrates the potential worst - case behavior of a @xmath1-based groundness analysis @xcite . the analysis of the predicate @xmath29 enumerates the models of the ( constant ) @xmath16-ary boolean function @xmath13 ( @xmath11 ) in reverse order . starting from the initial approximation ( which has no models ) , each consecutive approximation is a function which has one new model that was not in the previous iteration . for example , when @xmath30 , the models accumulate in the following order : @xmath31 and the @xmath1-based analysis totals 8 iterations . in contrast the corresponding @xmath0-based analysis totals 4 iterations because at each iteration the current set of models is closed under intersection . so for example , in the third iteration , the set @xmath32 is closed to give @xmath33 . we now construct a series of programs which demonstrates the potential worst - case behavior of a @xmath0-based groundness analysis . this construction is based on the following observation : [ prop1 ] let @xmath21 be an @xmath16-ary model . then the set of @xmath16-ary models smaller or equal to @xmath21 is closed under intersection . the result follows from the following observation : if @xmath24 and @xmath25 are @xmath16-ary models , then @xmath34 is no larger than @xmath24 ( and no larger than @xmath25 ) . this is because @xmath34 is obtained from @xmath24 ( or from @xmath25 ) by changing some one s to zero s . a consequence of proposition [ prop1 ] is that the domain of definite boolean functions over @xmath16 variables contains a chain of length @xmath35 . to demonstrate such a chain consider an enumeration @xmath36 of the @xmath16-ary models according to their binary ordering ( so @xmath37 and @xmath38 ) . observe that @xmath39 is the @xmath16-ary binary representation of @xmath40 . define a sequence @xmath41 as follows : let @xmath42 be the boolean function with the empty set of models and for @xmath43 define @xmath44 to be the boolean function whose models are @xmath45 . > from the construction it is clear that @xmath46 forms a chain . moreover , the elements of @xmath46 are in @xmath0 : they are positive because they have @xmath47 as a model ; and from proposition [ prop1 ] , it follows that they are closed under intersection . the chain @xmath46 is of length @xmath48 because , for @xmath49 @xmath44 has exactly one model more than @xmath50 . this is the setting for our construction . the ( @xmath0-based ) groundness analysis of the following predicate `` @xmath51 '' iterates through the chain @xmath46 . the arguments typeset in boldface highlight the case for @xmath52 . the program size is quadratic in @xmath16 and consists of a single predicate of arity @xmath16 with @xmath53 binary clauses . the analysis of the program can be viewed as counting from zero to @xmath54 in its arguments . # 1 @xmath55 the @xmath0- and @xmath1-based groundness analyses of the predicate @xmath51 program in the series given in this note involve an exponential number of iterations and compute an @xmath16-ary boolean function . the same is true for the @xmath1-based analysis of the series given in @xcite . however , it is important to note that complexity is typically expressed in terms of the size of the input to a problem and that the size of the program defining @xmath51 in both series is quadratic in @xmath16 ( @xmath18 ) . hence , formally speaking , we have shown that both @xmath0 and @xmath1-based groundness analyses may potentially involve a number of iterations which is @xmath19 . this is bad enough , but sub - exponential . for @xmath1 , we can strengthen the result . the following program is of size linear in @xmath16 ( @xmath56 ) and its @xmath1-based groundness analysis requires @xmath54 iterations . @xmath57 \begin{array}{llcll } s(c~ , x_1 , \ldots , x_1 , & x_1 , c~ , \ldots , c~ ) .\\ s(w , a_1 , \ldots , a_{n-1 } , & w , b_1 , \ldots , b_{n-1 } ) & \leftarrow & s(a_1,\ldots , a_n , & b_1,\ldots , b_n).\\ \end{array } \end{array}\ ] ] intuitively , the @xmath58 arguments of the predicate @xmath59 represent two @xmath16-digit binary numbers ( the first is the successor of the second ) so that the @xmath16 recursive clauses from the program in section [ s2 ] can be simulated by two clauses for @xmath59 . the base case of @xmath59 corresponds to the last recursive clause . however , the analysis of @xmath59 does not follow an exponential chain so we still need the predicate @xmath51 to get the worst - case behaviour . this approach does not work for @xmath0 because the result in @xmath1 for @xmath59 is not closed under intersection . we have demonstrated a @xmath60 worst case complexity for @xmath1 and at least @xmath19 for @xmath0 ( where @xmath61 is the size of the program ) . it remains to be determined if the worst case for @xmath0 is really as bad as that for @xmath1 or perhaps @xmath0 has better worst - case behaviour . howe , j. m. , & king , a. ( 2000 ) . implementing groundness analysis with definite boolean functions . smolka , g. ( ed ) , _ european symposium on programming_. lecture notes in computer science , vol . springer - verlag . king , andy , smaus , jan - georg , & hill , pat . ( 1999 ) . quotienting share for dependency analysis . swierstra , doaitse ( ed ) , _ european symposium on programming_. lecture notes in computer science , vol . springer - verlag .

this note illustrates theoretical worst - case scenarios for groundness analyses obtained through abstract interpretation over the abstract domains of _ definite _ ( @xmath0 ) and _ positive _ ( @xmath1 ) boolean functions . for @xmath0 , an example is given for which any @xmath0-based abstract interpretation for groundness analysis follows a chain which is exponential in the number of argument positions as well as in the number of clauses but sub - exponential in the size of the program . for @xmath1 , we strengthen a previous result by illustrating an example for which any @xmath1-based abstract interpretation for groundness analysis follows a chain which is exponential in the size of the program . it remains an open problem to determine if the worst case for @xmath0 is really as bad as that for @xmath1 .

starting with a commercially available 50 nm thick low stress silicon nitride membrane , a 50 nm thick gold layer for the plasmonic metamaterial was thermally evaporated . the gold - coated membrane was structured with a focussed ion beam system ( fei helios 600 nanolab ) . the pattern of plasmonic resonators was milled and then the supporting silicon nitride membrane was cut into suspended strips with tapered ends . detailed dimensions of the nanostructure are given by fig . [ fig - sample]b . the cubic nano - opto - mechanical nonlinearity was measured by detecting the pump - induced modulation of the metamaterial s transmission ( fig . [ fig - modulation]a ) . the probe beam at wavelength 1310 nm was generated by a fiber - coupled telecom laser diode ( thorlabs fpl1053s ) . the pump beam from a fiber - coupled laser diode at wavelength 1550 nm ( thorlabs fpl1009s ) was modulated with a fiberized electro - optical modulator ( eospace ax-0k5 - 10-pfa - ul ) and then combined with the probe beam in a fiber coupler ( jdsu l2swm1310/1550bb ) . the transmitted beam was filtered to remove the pump and the probe signal was detected by an ingaas photodetector ( new focus 1811 ) and a lock - in amplifier ( stanford research sr844 ) . in all optical experiments and simulations the electric field was @xmath6-polarized ( parallel to the strips ) and incident on the silicon nitride side of the sample . the authors are grateful to artemios karvounis for fruitful discussions and to pablo cencillo - abad for assistance in preparing the manuscript . this work is supported by the leverhulme trust , the royal society , the u.s . office of naval research ( grant n000141110474 ) , the moe singapore ( grant moe2011-t3 - 1 - 005 ) and the uk s engineering and physical sciences research council ( grant ep / g060363/1 ) . the data from this paper can be obtained from the university of southampton eprints research repository : http://dx.doi.org/10.5258/soton/377861

from the demonstration of saturable absorption by vavilow and levshin in 1926 @xcite , and with invention of the laser , unavailability of strongly nonlinear materials was a key obstacle for developing optical signal processing , in particular in transparent telecommunication networks . today , most advanced photonic switching materials exploit gain dynamics @xcite and near - band and excitonic effects in semiconductors @xcite , nonlinearities in organic media with weakly - localized electrons @xcite and nonlinearities enhanced by hybridization with metamaterials @xcite . here we report on a new type of artificial nonlinearity that is nano - opto - mechanical in nature . it was observed in an artificial metamaterial array of plasmonic meta - molecules supported by a flexible nano - membrane . here nonlinearity is underpinned by the reversible reconfiguration of its structure induced by light . in a film of only 100 nanometres thickness we demonstrated modulation of light with light using milliwatt power telecom diode lasers . [ tbh ] * nonlinearity in nano - opto - mechanical metamaterial . * light - induced ( red ) forces in the plasmonic metamaterial array cause nanoscale reversible displacements of its small and light building blocks supported by an elastic nano - membrane . they move fast , with a response time reaching microseconds . these displacements change the plasmonic spectra of the metamaterial array and its transmission . this is used to modulate another , weaker beam of light ( green ) at a different wavelength . , title="fig:",width=302 ] some exceptional opportunities for developing engineered nonlinear media are provided by nano - optomechanics that take advantage of the changing balance of forces at the nanoscale @xcite . with the decrease in the physical dimensions of a system the electromagnetic forces between constituent elements grow , as may be illustrated by the repulsion of electrons as their separation diminishes . in contrast , elastic forces , such as the force restoring a compressed spring , decrease with size . moreover , the nanoscale metamaterial building blocks have high natural frequencies and thus can be moved very fast , potentially offering ghz switching bandwidth for elements of sub - micron size . recently this was exploited to develop reconfigurable metamaterials fabricated on nanoscale elastic membranes . these photonic metamaterials can be driven thermo - elastically @xcite and with electromagnetic forces , such as the coulomb force between charged elements of the nanostructure @xcite , or the lorentz force acting on currents running through conductive elements in magnetic field @xcite . their electro - optical and magneto - optical switching characteristics surpass those of natural media by orders of magnitude . metamaterials are in essence arrays of optical resonators where light - induced nano - opto - mechanical phenomena can play crucial roles . for instance it has been suggested that nearfield @xcite and casimir forces @xcite in metamaterials are modified in the presence of light . forces between oscillating plasmonic or displacement currents induced by light in metamolecules were also theoretically shown to be sufficient to drive reconfiguration of a metamaterial structure in the optical part of the spectrum @xcite , and proof - of - principle experiments on such interactions between individual metamolecules have been reported at microwave frequencies @xcite . here we show that reversible reconfiguration of a plasmonic metamaterial nanostructure driven by optical forces between its illuminated elements can be the source of a strong optical nonlinearity . we show that this cubic nonlinearity may be used to modulate light with light at milliwatt power levels . moreover , although for the majority of media the magnitude of nonlinearity tends to be proportional to its response time , the nano - opto - mechnical nonlinearity is three orders of magnitude faster than could be expected from this otherwise universal trend . the plasmonic nanomechanical metamaterial was fabricated on strips of suspended dielectric membrane of nanoscale thickness in such a way that plasmonic elements of the metamolecules were located on different strips . here light illumination leads to electromagnetic and thermal forces that induce nanoscale strip displacements and thus reconfiguration of individual metamolecules in a way that affects their plasmonic resonances , see fig . [ fig - concept ] . this leads to modulation of the metamaterial s transmissivity at different wavelengths . the nonlinear response of metamaterial , that is only 100 nanometres thick , can be observed at only a few milliwatts of continuous laser power , using lasers operating at telecommunication wavelengths . we developed nano - opto - mechanical metamaterial on the basis of a @xmath0-shaped resonator design known for exhibiting plasmon - induced transparency @xcite , see fig . [ fig - concept ] and fig . [ fig - sample]a . to allow mechanical deformation of the 700 nm @xmath1 700 nm plasmonic @xmath0 metamolecules , the horizontal and vertical gold bars were supported by different flexible silicon nitride strips of 28 @xmath2 m length spaced by alternating gaps of 95 nm and 145 nm . the nanostructure was fabricated by focused ion beam milling from a 50 nm thick silicon nitride membrane covered by 50 nm of gold . such structures can be reconfigured by ( i ) differential thermal expansion through light - induce heating of the bimorph gold and silicon nitride layers and ( ii ) by near - field optical forces acting between elements of illuminated plasmonic resonators . [ tbh ] * optically reconfigurable plasmonic metamaterial . * ( a ) scanning electron microscope image of the metamaterial nanostructure consisting of asymmetrically spaced gold ( yellow ) resonators on silicon nitride strips ( red ) . ( b ) transmission @xmath3 , reflection @xmath4 and absorption @xmath5 spectra of the metamaterial for @xmath6-polarized light incident on the silicon nitride side . dashed lines indicate the optical 1310 nm probe and 1550 nm pump wavelengths used in the modulation measurements of the main manuscript . the inset shows detailed unit cell dimensions in nanometers . the scanning electron micrograph was recorded by fei helios 600 nanolab and the optical spectra were measured using a microspectrophotometer ( craic technologies ) . , title="fig:",width=302 ] * optical spectra and optical forces . * ( a ) simulated metamaterial transmission @xmath3 , reflection @xmath4 and absorption @xmath5 spectra . insets show maps of the instantaneous values of the optically induced charges at the probe and pump wavelengths . ( b ) in - plane - of - metamaterial component of optical forces @xmath7 acting between the strip segments of an individual unit cell according to maxwell stress tensor calculations . ( c ) normal to the metamaterial component of optical forces @xmath8 acting on the strip segments along the light propagation direction . the total optical force on the unit cell ( green line ) is presented alongside the value expected from reflection and absorption ( dotted black line ) . forces are shown per unit cell in units of @xmath9 , where @xmath10 is the incident power per unit cell and @xmath11 is the speed of light in vacuum . all quantities are shown for @xmath6-polarized illumination of the silicon nitride side . , width=302 ] [ t ] experimental measurements and full 3d maxwell simulations show a pronounced near - infrared absorption resonance around 1240 nm , fig . [ fig - simulations]a and fig . [ fig - sample]b . maxwell stress tensor calculations @xcite assuming normal incidence illumination of the non - diffracting , periodic metamaterial reveal optical forces acting on the @xmath0-resonators around these resonances , see fig . [ fig - simulations]b - c . as the normally incident photons only carry momentum along the @xmath12-directionthe net force along @xmath12 must comply with the momentum transfer associated with absorption @xmath5 and reflection @xmath4 , @xmath13 , where @xmath10 is the incident power per unit cell area and @xmath11 is the speed of light in vacuum . in close agreement with these relationships , our simulations show substantial differential optical forces @xmath14 between the unit cell s strip segments reaching about 0.4 @xmath9 along @xmath15 and 2.8 @xmath9 along @xmath12 . optical forces in such metamaterial nanostructures can be understood as time - averaged coulomb and lorentz forces acting between the currents and oscillating dipole charges of the plasmonic resonators in the presence of the illuminating electromagnetic wave @xcite . for example , repulsive and attractive optical forces within the metamaterial plane for illumination at wavelengths of 1310 nm and 1550 nm correspond to repulsive and attractive interaction of the dipole charges on neighboring strips , see fig . [ fig - simulations]a , b . forces normal to the metamaterial plane result from a combination of optical pressure and the lorentz force acting on the moving charges of the plasmonic mode in the presence of the incident wave s magnetic field . we investigated the nonlinear optical properties of the metamaterial in a pump - probe experiment as presented in fig . [ fig - modulation]a . as the pump and probe optical sources we used fiber - coupled telecommunication laser diodes operating at the wavelengths of 1550 nm and 1310 nm , respectively . the intensity of the pump beam was modulated by a fibre - coupled electro - optical modulator . the pump and probe were then combined into a single beam using a fiber coupler , de - coupled into free space and focused on the sample placed in a microscope using focusing and collection objectives . the sample was placed in a cell evacuated to the pressure of 30 pa to prevent damping of the nano - mechanical motion . light transmitted through the sample was detected by an ingaas photodetector and a lock - in amplifier locked to the pump modulation frequency . in order to optimize the light - induced nano - opto - mechanical modulation , we pumped the nanostructure at 1550 nm just above its absorption resonance , where simulations predict significant in - plane forces . at low modulation frequencies , below 100 khz , the optical pump leads to pronounced modulation of the metamaterial s transmission at the probe wavelength , see fig . [ fig - modulation]c . for 0.9 mw pump power ( peak intensity 920 w/@xmath16 ) a modulation amplitude on the order of 1@xmath17 is observed at 25 khz modulation . the low - frequency component of the nonlinear response has a thermal nature and drops rapidly with increasing modulation frequency , fading at a few 100s of khz . in this frequency range , the modulation amplitude halves as the modulation frequency doubles which is consistent with a thermal deformation of the bimorph au / sinx membrane , see inset to fig . [ fig - modulation]b . this is consistent with thermal calculations for 50 nm gold @xcite and silicon nitride @xcite layers , which predict conductive cooling timescales on the order of 20 - 30 @xmath2s . the light - induced heating of the nanostructure is balanced by heat conduction along the metamaterial strips and according to our calculations can reach hundreds of degrees with pump excitation at 1 kw/@xmath16 . as the thermal expansion coefficient of gold ( 14.4 @xmath18/k ) exceeds that of silicon nitride ( 2.8 @xmath18/k ) 5-fold such large temperature changes will bend the strips . when heated by the modulated pump laser narrow strips will rise higher than wider strips due to more complete gold coverage ; the metamolecule changes its shape and therefore its plamonic absorption and transmission for the probe wavelength change . while the thermal nonlinearity easily dominates at low modulation frequencies , other modulation mechanisms are at play at higher frequencies . indeed , as differential thermal expansion in the layered structure can only drive motion out of the metamaterial plane , it can not drive in - plane resonant oscillations of the metamaterial strips . several resonant peaks of nonlinear response are seen at much higher frequencies around 600 khz , 1 mhz and 2 mhz , see fig . [ fig - modulation]c . mechanical eigenmode calculations identify the resonances as the fundamental modes of oscillation of the narrow and wide strips , corresponding to vibration normal to the metamaterial plane ( around 600 khz ) and motion within the metamaterial plane ( at mhz frequencies ) , see insets . comparing the sub-@xmath2s heating and cooling cycles to the characteristic strip cooling timescale , we argue for their non - thermal origin . to assess the origin of the high frequency modulation we estimate the optomechanical deformation that can be expected from the optical forces acting between different parts of the metamolacular resonator . based on the results of simulations presented in fig . [ fig - simulations]c , pump - induced optical forces reach @xmath19 ( @xmath20 ) for the narrow ( wide ) strips . for a strip pair with 29 unit cells and 920 w/@xmath16 pump intensity , this corresponds to about 220 fn ( 90 fn ) force per strip . estimating the spring constant @xmath21 from strip mass @xmath22 and measured resonance frequency @xmath23 as @xmath2428 fn / pm ( 56 fn / pm ) for out - of - plane deformation , we may expect the gap between strips and thus between elements of the metamolecule to change by 6 pm . corresponding estimates for in - plane deformation predict a gap change on the order of 1 pm . such modulation is not sufficient to create a detectable level of non - resonant high - frequency modulation . however , much larger displacements are expected at the structure s mechanical resonances , where the displacement will be enhanced by the quality factor of the mechanical resonator . our optical measurements show mechanical resonance quality factors on the order of 100 , which may serve as a lower limit for the quality factor of mechanical resonances of individual strips as the overall resonance profile is affected by inhomogeneous broadening due to slight structural variations from strip to strip . taking this enhancement into account , optical forces are definitely sufficient to induce resonant strip displacements on the order of 1 nm , which is consistent with the observed transmission modulation on the order of 1% . therefore we argue that the resonant response at 600 khz , 1 mhz and 2 mhz is predominantly driven by non - thermal , optical forces . * optical nonlinearities in metamaterials and conventional media . * cubic optical nonlinearities due to thermal effects , saturated absorption , molecular orientation and electronic anharmonicity follow a general trend of decreasing achievable modulation frequency @xmath25 with increasing magnitude of the nonlinearity : @xmath26 @xcite . metamaterial patterning of plasmonic metal ( points 1 and 2 , @xcite ) , hybridization of carbon nanotubes ( point 3 , @xcite ) , graphene ( point 4 , @xcite ) , and semiconductors ( points 5 and 6 , @xcite ) with metamaterials resonantly enhances the magnitude of the nonlinear response without affecting its response time . nano - opto - mechanical metamaterial demonstrated in the present work also departs from the common trend exhibiting @xmath27 that is three orders of magnitude bigger than in conventional nonlinear media ( point 7 ) . , width=302 ] absorption in a nonlinear medium is conventionally described by the expression @xmath28 , where @xmath29 is light intensity , @xmath12 is the propagation distance and @xmath30 and @xmath31 are the linear and nonlinear absorption coefficients . as the observed nonlinear transmission change is proportional to the pump power , we can quantify the nonlinearity of the reconfigurable photonic metamaterial by estimating its first nonlinear absorption coefficient @xmath31 . assuming that the nonlinear transmission change @xmath32 results from nonlinear absorption , @xmath33 , where @xmath34 is the metamaterial s thickness . at 1 mhz modulation frequency , @xmath35 m / w corresponding to a nonlinear susceptibility on the order of @xmath36 and a nonlinear imaginary refractive index change @xmath37 at 920 w/@xmath16 pump intensity . we argue here that the observed nano - opto - mechanical nonlinearity is special as it exhibits a departure from the prevailing relation between speed and magnitude of nonlinear responses ( fig . [ fig - nonlinearity ] ) . this happens in a way that allows larger nonlinearity or higher modulation frequency to be achieved than could be expected from traditional media . indeed , it appears that for many mechanisms of optical nonlinearity the achievable modulation frequency @xmath25 is related to the magnitude of the nonlinerity @xmath38 in such a way that @xmath26 @xcite . at the resonance of in - plane motion , mhz rate modulation can be achieved at about 1 kw/@xmath16 of light intensity corresponding to about @xmath39 . in summary , we provide the first experimental demonstration of a giant nano - opto - mechanical nonlinearity in plasmonic metamaterial that allows modulation of light with light at mhz frequencies and miliwatt power levels . bimorph deformation resulting from optical heating at low modulation frequencies and electromagnetic forces between elements of plasmonic resonators drive the nonlinear response . the magnitude and spectral position of the underlying optical resonance can be controlled across the optical spectral range by design of the plasmonic resonators and the supporting nano - membrane strips .

ocean - tide information has considerably many applications . the data obtained is used to solve vital problems in oceanography and geophysics , and to study earth tides , elastic properties of the earth s crust and tidal gravity variations . it is also used in space studies to calculate the trajectories of man - made satellites of the earth and to interpret the results of satellite measurements . the interaction of tides with deep sea ridges and chains of seamounts give rise to deep eddies which transport nutrients from the deep to the surface . the alternate flooding and exposure of the shoreline due to tides is an important factor in the determination of the ecology of the region . [ [ section ] ] one of the first mathematical explanation for tides was given by newton by describing tide generating forces . the first dynamic theory of tides was proposed by laplace . here we consider the tidal dynamics model proposed by marchuk and kagan @xcite . the existence and uniqueness of weak solutions of the deterministic tide equation and that of strong solutions of the stochastic tide equation with additive trace class gaussian noise have been proved in manna , menaldi and sritharan@xcite . in this work , we consider the stochastic tide equation with lvy noise and prove the existence and uniqueness and regularity of solution in bounded domains . control of fluid flow has numerous applications in control of pollutant transport , oil recovery / transport problems , weather predictions , control of underwater vehicles , etc . unification of many control problems in the engineering sciences have been done by studying the optimal control problem of navier - stokes equations ( see @xcite , @xcite ) . here we consider the initial data optimal control of the stochastic tidal dynamics model . we consider the stroock - varadhan martingale formulation @xcite of the stochastic model to prove the existence of optimal initial value control . [ [ section-1 ] ] the organization of the paper is as following . a brief description of the model has been given in section [ model ] . section [ setting ] describes the functional setting of the problem and states the monotonicity property of the non - linear operator . in section [ estimate ] we consider the a - priori estimates and prove the existence , uniqueness and regularity of strong solution . in section [ stochastic ] we consider the stochastic optimal control problem with initial value control . [ [ section-2 ] ] in the framework of gelfand triple @xmath0 we consider the following tidal dynamics model with lvy noise lcl du + [ au+b(u)+g]dt = f(t)dt+(t , u(t))dw(t)+_z h(u , z ) ( dt , dz ) , + , + + div(hu)=0 , + u(0)=u_0,(0)=_0 . the operators @xmath1 and @xmath2 are defined in sections [ model ] and [ setting ] . @xmath3 is an @xmath4-valued wiener process with trace class covariance . @xmath5 is a compensated poisson random measure , where @xmath6 denotes the poisson counting measure associated to the point process @xmath7 on @xmath8 , a measurable space , and @xmath9is a @xmath10-finite measure on @xmath11 . [ [ section-3 ] ] the following theorem states the main result of section [ estimate ] . the functional spaces appearing in the statement of the theorem have been defined in section [ setting ] . [ thmint1 ] let us consider the above stochastic tide model with @xmath12 and @xmath13 such that @xmath14 assume that @xmath15 and @xmath16 satisfy the following hypotheses : 1 . @xmath17\times{\mathbb{h}_0 ^ 1}(\mathcal{o});l_q(h_0,{\mathbb{l}^2 } ) ) , h\in\mathbb{h}^2_\lambda([0,t]\times z;{\mathbb{l}^2}(\mathcal{o}))$ ] , 2 . for all @xmath18 , there exists a positive constant @xmath19 such that for all @xmath20 @xmath21 3 . for all @xmath18 , there exists a positive constant @xmath22 such that for all @xmath23 @xmath24 then there exist path - wise unique adapted processes @xmath25 and @xmath26 with the regularity @xmath27 satisfying the above stochastic tide model in the weak sense . in section [ stochastic ] we consider the following stochastic optimal control problem with initial value control , lcl [ scint1 ] du + [ au+b(u)+g]dt = f(t)dt+(t , u(t))dw(t)+_z h(u , z ) ( dt , dz ) + , + [ scint2 ] + div(hu)=0 , + [ scint3 ] u(0)=u_0+u,(0)=_0 , where @xmath28 , @xmath29 and @xmath30 . the regularities on the initial values and the assumptions on @xmath15 and @xmath16 are the same as considered in [ thmint1 ] . the cost functional is given by @xmath31,\ ] ] where the function @xmath22 is defined in section [ stochastic ] . [ [ section-4 ] ] the main result of section [ stochastic ] is given @xmath28 , there exists a pair @xmath32 which gives a martingale solution to equations - , and rcl j(,)=\{&&j(u , u);(u , u)(l^2(0,t;_0 ^ 1())l^(0,t;^2 ( ) ) + & & ( 0,t;^-1()))l^2(0,t;^2 ( ) ) , + & & ( u , u ) + & & -}. under the assumptions that : ( 1 ) earth is perfectly solid , ( 2 ) ocean tides do not change earth s gravitational field , and ( 3 ) no energy exchange takes place between the mid - ocean and shelf zone , marchuk and kagan @xcite obtained the following mathematical model @xmath33 @xmath34 in @xmath35 $ ] , where @xmath36 is a bounded 2-d domain ( horizontal ocean basin ) with coordinates @xmath37 and @xmath38 represents the time . here @xmath39 denotes the time derivative , @xmath40 and @xmath41 are the laplacian , gradient and the divergence operators respectively . [ [ section-5 ] ] the unknown variables @xmath42 represent the total transport 2-d vector ( i.e. , the vertical integral of the velocity from the ocean surface to the ocean floor ) and the displacement of the free surface with respect to the ocean floor . for details on the coefficients and the domain description see manna , menaldi and sritharan @xcite . [ [ section-6 ] ] denote by @xmath1 the following matrix operator @xmath43 and the nonlinear vector operator @xmath44 where @xmath45 and @xmath46 are positive constants , @xmath47 is a strictly positive smooth function . in this model we assume the depth @xmath48 to be a continuously differentiable function of @xmath49 , nowhere becoming zero , so that @xmath50 where m is some positive constant which equals zero at a constant ocean depth . [ [ section-7 ] ] to reduce to homogeneous dirichlet boundary conditions consider the natural change of unknown functions @xmath51 and @xmath52 which are referred to the tidal flow and the elevation . the full flow @xmath53 which is given a priori on the boundary @xmath54 , has been extended to the whole domain @xmath35 $ ] as a smooth function and still denoted by @xmath53 . + + then the tidal dynamic equation can be written as @xmath55,\\ \partial_t { \hat{z}}+ div(hu)=0\quad\text{in}\quad\mathcal{o}\times[0,t],\\ u=0\quad\text{on}\quad\partial\mathcal{o}\times[0,t],\\ u = u_0,\;\;{\hat{z}}={\hat{z}}_0\quad\text{in}\quad\mathcal{o}\times\{0\ } , \end{array}\right.\ ] ] where lcl f^=f-+g_0^t div(hw^0)dt - aw^0 , + u_0(x)=w_0(x)-w^0(x,0 ) , + _ 0(x)=_0(x ) . we use the ( vector - valued ) sobolev spaces @xmath56 and @xmath57 , with : + the norm on @xmath58 as @xmath59 and the norm on @xmath4 as @xmath60 using the gelfand triple @xmath0 , we may consider @xmath61 or @xmath62 as a linear map from @xmath58 or @xmath4 into the dual of @xmath58 respectively . the inner product in @xmath4 or @xmath63 is denoted by @xmath64 . so @xmath65 for any @xmath66 and @xmath67 in @xmath4 . the duality pairing between @xmath68 and @xmath69 is denoted by @xmath70 . [ [ section-8 ] ] now we give the definitions and some properties of hilbert space valued wiener processes , lvy processes and skorokhod spaces , most of which have been borrowed from the books by da prato and zabczyk @xcite , applebaum @xcite and metivier @xcite . [ [ section-9 ] ] for a hilbert space @xmath71 , let us denote the space of all bounded linear operators from @xmath71 to @xmath71 by @xmath72 . we consider two hilbert spaces @xmath16 and @xmath71 , and a symmetric non - negative operator @xmath73 such that @xmath74 . a @xmath71-valued stochastic process @xmath75 , is called a @xmath76-wiener process if 1 . @xmath77 , 2 . @xmath78 has continuous trajectories , 3 . @xmath78 has independent increments , 4 . @xmath79 . if a process @xmath80\}$ ] satisfies @xmath81 and @xmath82 for @xmath83 $ ] the we say that @xmath78 is a @xmath76-weiner process on @xmath84 $ ] . [ [ section-10 ] ] let @xmath76 be a symmetric non - negative operator @xmath85 . define @xmath86 . then @xmath87 is a hilbert space equipped with inner product @xmath88 , @xmath89 where @xmath90 is the pseudo - inverse of @xmath91 , and @xmath76 is the covariance operator of the @xmath92-valued wiener process @xmath3 . [ [ section-11 ] ] let @xmath93 denote the space of linear operators @xmath94 such that @xmath95 is a hilbert - schmidt operator from @xmath92 to @xmath92 . define the norm on the space @xmath93 by @xmath96 . [ [ section-12 ] ] let @xmath97 be a separable and complete metric space . let @xmath98 denote the set of @xmath99-valued functions on @xmath84 $ ] which are cdlg ( if @xmath100 then for all @xmath101\;\;g$ ] is right continuous at @xmath38 and has a left limit at @xmath38 ) , endowed with the skorokhod topology . this topology is metrizable by the following metric @xmath102 rcl _ t(u , v):=__t & & , where @xmath103 is the set of increasing homeomorphisms of @xmath84 $ ] . more over , @xmath104 is a complete metric space . a sequence @xmath105 converges to @xmath106 iff there exists a sequence @xmath107 of homeomorphisms of @xmath84 $ ] such that @xmath108 tends to the identity uniformly on @xmath84 $ ] and @xmath109 tends to @xmath66 uniformly on @xmath84 $ ] . let @xmath110 be a filtered probability space , and @xmath111 be a banach space . a process @xmath112 with state space @xmath113 is called a lvy process if 1 . @xmath112 is adapted to @xmath114 , 2 . @xmath115 a.s . , 3 . @xmath116 is independent of @xmath117 if @xmath118 , 4 . @xmath112 is stochastically continuous , i.e. , @xmath119 , 5 . @xmath112 is cdlg , 6 . @xmath112 has stationary increments , i.e. , @xmath120= { \mathbb{e}}[x_{t - s}],0\leq s < t$ ] . the jump of @xmath121 at @xmath122 is given by @xmath123 . let @xmath124 . we define @xmath125 so , @xmath126 is the number of jumps of size @xmath127 which occur before or at time @xmath38 . @xmath126 is called the poisson random measure of @xmath112 . the differential form of this measure is written as @xmath128 . [ [ section-13 ] ] we call @xmath5 a compensated poisson random measure , where @xmath129 is known as the compensator of the lvy process @xmath112 . here @xmath130 denotes the lebesgue measure on @xmath131 , and @xmath9 is a @xmath15-finite measure on @xmath11 . [ lemw]for any real - valued smooth function @xmath132 and @xmath133 with compact support in @xmath134 , the following hold : lcl _ l^2 ^ 24_1_l^2_2_l^2 , + [ eq13 ] _ l^4 ^ 42_l^2 ^ 2_l^2 ^ 2 . for proof see @xcite . + + notice that by means of the gelfand triple we may consider @xmath1 , given by , as a mapping of @xmath58 into its dual @xmath135 . [ [ section-14 ] ] define the non - symmetric bilinear form @xmath136+\beta[(u_1,v_2)_{{\mathbb{l}^2}}-(u_2,v_1)_{{\mathbb{l}^2}}],\ ] ] on @xmath68 . thus if @xmath66 has a smooth second derivative then @xmath137 for every @xmath67 in @xmath68 . more over , the bilinear form @xmath138 is continuous and coercive in @xmath58,i.e . , lcl |a(u , v)|c_1u|__0 ^ 1v|__0 ^ 1u , v_0 ^ 1 , + ( au , u)_^2=a(u , u)=u__0 ^ 1 ^ 2 , for some positive constant @xmath139 . [ [ section-15 ] ] let us denote the nonlinear operator @xmath140 by @xmath141.\ ] ] then we have the following lemma : [ mon ] let @xmath66 and @xmath67 be in @xmath142 . then the following estimate holds : @xmath143 for proof see lemma 3.3 in @xcite . [ [ section-16 ] ] the nonlinear operator @xmath140 is a continuous operator from @xmath144 to @xmath4 , where lcr [ e2 ] b(v)_^2c_2v+w^0_^4 ^ 2 , + [ e1 ] b(u)-b(v)_^2c_2[u+w^0_^4+v+w^0_^4]u - v_^4 , where the constant @xmath145 is the sup - norm of the function @xmath146 . [ [ section-17 ] ] with the above functional setting , we write the stochastic tidal dynamics equation with the lvy forcing as lcl [ eq21 ] du + [ au+b(u)+g]dt = f(t)dt+(t , u(t))dw(t)+_z h(u , z ) ( dt , dz ) , + + [ eq22 ] + div(hu)=0 , + [ eq23 ] u(0)=u_0,(0)=_0 , where @xmath28 and @xmath29 . the operators @xmath1 and @xmath2 are defined through and respectively . @xmath3 is a @xmath4-valued wiener process with trace class covariance . @xmath147 is a measurable mapping from @xmath148 into @xmath92 . [ [ section-18 ] ] we define the space @xmath149\times z;{\mathbb{l}^2}(\mathcal{o}))$ ] as @xmath150\times z;{\mathbb{l}^2}(\mathcal{o}))=\{x:{\mathbb{e}}\left[\int_0^t\int_z \|x\|_{{\mathbb{l}^2}}^2 { \lambda(dz)}dt\right]<\infty\}.\ ] ] we assume that @xmath15 and @xmath16 satisfy the following hypotheses : 1 . @xmath17\times{\mathbb{h}_0 ^ 1}(\mathcal{o});l_q(h_0,{\mathbb{l}^2 } ) ) , h\in\mathbb{h}^2_\lambda([0,t]\times z;{\mathbb{l}^2}(\mathcal{o}))$ ] , 2 . for all @xmath18 , there exists a positive constant @xmath19 such that for all @xmath20 @xmath151 3 . for all @xmath18 , there exists a positive constant @xmath22 such that for all @xmath23 @xmath24 denote by @xmath152 where @xmath153 is any fixed orthonormal basis in @xmath4 with each @xmath154 . similarly denote by @xmath155 the n - dimensional subspace of @xmath63 . let @xmath156 denote the orthogonal projection of @xmath4 into @xmath157 . define @xmath158 and @xmath159 . [ [ section-19 ] ] we define @xmath160 as the solution of the following stochastic equation in variational form : lrl [ eqn5 ] & & ( du^n(t),v^n(t))_^2+(a(u^n(t),v^n(t))+(b(u^n(t)),v^n(t))_^2+(g^n(t ) , v^n(t))_^2)dt + & & = ( f(t),v^n(t))_^2dt+(^n(t , u^n(t)),v^n(t))_^2dw^n(t ) + & & + _ z ( h^n(u^n(t-),z),v^n(t))_^2 ( dt , dz)v^n_n^2 ( ) , + [ eqn6 ] & & ( ^n(t)+div(hu^n(t)),(t))_l^2=0l_n^2 ( ) , + & & u^n(0)=u^n_0^n(0)=^n_0 . [ prop ] under the above mathematical setting let @xmath161\times{\mathbb{h}_0 ^ 1}(\mathcal{o});l_q(h_0,{\mathbb{l}^2})),\;h\in\mathbb{h}^2_\lambda([0,t]\times z;{\mathbb{l}^2}(\mathcal{o})),\\ u_0\in{\mathbb{l}^2}(\mathcal{o}),{\hat{z}}_0\in l^2(\mathcal{o } ) . \end{array}\right.\ ] ] let @xmath162 be an adapted process in @xmath163 which solves the differential equations and . then we have the following a priori estimates : l [ eq14 ] [ u^n(t)_^2 ^ 2+^n(t)_l^2 ^ 2]+2[_0^tu^n(t)__0 ^ 1 ^ 2dt]c_1(2 ) t , + [ eq15 ] [ _ 0tt(u^n(t)_^2 ^ 2 + ^n(t)_l^2 ^ 2)]+2[_0^tu^n(t)__0 ^ 1 ^ 2dt]c_2(2 ) , where the constants @xmath164 and @xmath165 depend on the coefficients @xmath166 and the norms @xmath167 , @xmath168 and @xmath169 . applying it s formula to @xmath170 and the process @xmath162 lr [ eqn 7 ] d(u^n(t)_^2 ^ 2)+ 2u^n(t)^2__0 ^ 1dt+2(b(u^n(t)),u^n(t))_^2dt+2(g^n(t ) , u^n(t))_^2dt + = 2(f(t),u^n(t))_^2dt+2(^n(t , u^n(t)),u^n(t))_^2dw^n(t)+^n(t , u^n(t))_l_q^2dt + + _ z h^n(u^n(t-),z)_^2 ^ 2 n(dt , dz ) + 2_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) . + using the definition of the operator @xmath140 and lemma [ mon ] rcl ( b(u^n(t)),u^n(t))_^2&&_(x)|w^0(t)|^2u^n(t)dx + & & -w^0(t)_^4 ^ 2u(t)_^2 + & & -[w^0(t)_^4 ^ 4+u(t)_^2 ^ 2 ] . using the divergence theorem and the inequality @xmath171 we obtain , rcl |g(^n(t ) , u^n(t))_^2| & = & |-g(^n(t),div ( u^n(t)))_^2| + & & [ ^n(t)_l^2 ^ 2+div ( u^n(t))_l^2 ^ 2 ] . since the divergence is bounded by gradient in @xmath172-norm @xmath173.\ ] ] hence the energy equality yields rcl [ eq8 ] d&(&u^n(t)_^2 ^ 2)+ 2u^n(t)^2__0 ^ 1 dt + & & ( f(t)_^2 ^ 2+u^n(t)_^2 ^ 2 + [ w^0(t)_^4 ^ 4+u(t)_^2 ^ 2]+^n(t)_l^2 ^ 2)dt + & & + u^n(t)__0 ^ 1 ^ 2dt + 2(^n(t , u^n(t)),u^n(t))_^2dw^n(t)+^n(t , u^n(t))_l_q^2dt + & & + _ z h^n(u^n(t-),z)^2_^2 n(dt , dz ) + 2_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) . + using equation @xmath174 now rcl & & |(hdiv(u^n(t)),^n(t))_l^2|+|(u^n(t)h,^n(t))_l^2| + & & h_l^div(u^n(t))_l^2^n(t)_l^2 + & & + u^n(t)_^2h_l^^n(t)_l^2 . using the assumptions on @xmath175 rcl |(div(hu^n(t)),^n(t))_l^2| & & u^n(t)__0 ^ 1^n(t)_l^2 + mu^n(t)_^2^n(t)_l^2 + & & [ u^n(t)__0 ^ 1 ^ 2 + ^n(t)_l^2 ^ 2 ] + & & + [ u^n(t)_^2 ^ 2+^n(t)_l^2 ^ 2 ] . thus @xmath176 adding equations and lcl [ eq10 ] d(&&u^n(t)_^2 ^ 2+^n(t)_l^2 ^ 2)+u^n(t)^2__0 ^ 1dt + & & ( 1+m+)u^n(t)_^2 ^ 2 dt + ( + + m)^n(t)_l^2 ^ 2 dt + & & + w^0(t)_^4 ^ 4 dt + f(t)_^2 ^ 2 dt + ^n(t , u^n(t))_l_q^2 dt + & & + 2(^n(t , u^n(t)),u^n(t))_^2dw^n(t)+_z h^n(u^n(t-),z)^2_^2 n(dt , dz ) + & & + 2_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) . define @xmath177 integrating equation rcl [ eq11 ] & u^n & ( t_n)_^2 ^ 2+^n(t_n)_l^2 ^ 2+_0^t_nu^n(s)^2__0 ^ 1ds + & & ( 1+m+)_0^t_nu^n(s)_^2 ^ 2 ds + ( + + m)_0^t_n^n(s)_l^2 ^ 2 ds + & & + _ 0^t_nw^0(s)_^4 ^ 4 ds + _ 0^t_nf(s)_^2 ^ 2 ds + _ 0^t_n^n(s , u^n(s))_l_q^2 ds + & & + 2_0^t_n(^n(s , u^n(s)),u^n(s))_^2dw^n(s ) + & & + _ 0^t_n_z h^n(u^n(s-),z)^2_^2 ( ds , dz ) + & & + _ 0^t_n2_z ( h^n(u^n(s-),z),u^n(s-))_^2(ds , dz ) + & & + _ 0^t_n_z h^n(u^n(s-),z)^2_^2 ( dz)ds + u^n_0_^2 ^ 2 + ^n_0_l^2 ^ 2 . let @xmath178 since @xmath179 is strong 2-integrable w.r.t . @xmath180 , hence using hlder s inequality @xmath181 is strong 1-integrable w.r.t . @xmath180 . then by theorem 4.12 of @xcite rcl [ eq12 ] & & + & & 2_0^t_z . also the stochastic integrals @xmath182 are martingales with zero averages . + taking expectation of equation and using and the above property , we get rcl & [ & u^n(t_n)_^2 ^ 2+^n(t_n)_l^2 ^ 2]+ + & & c+ + & & + + + & & + 3 + & & + [ u^n_0_^2 ^ 2+^n_0_l^2 ^ 2 ] . using the assumption h.2 rcl & [ & u^n(t_n)_^2 ^ 2+^n(t_n)_l^2 ^ 2]+ + & & c+ + & & + + 3k + & & + [ u^n_0_^2 ^ 2+^n_0_l^2 ^ 2 ] . hence using gronwall s inequality and taking limit as @xmath183 we have the desired a priori estimate . proceeding similarly but taking supremum over @xmath184 $ ] over equation before taking expectation rcl [ eq17 ] & [ & _ 0tt_n ( u^n(t)_^2 ^ 2+^n(t)_l^2 ^ 2)]+ + & & c+ + & & + + 3k(t_n)+3k + & & + 2 + + & & + + & & + [ u^n_0_^2 ^ 2+^n_0_l^2 ^ 2 ] . using burkholder - davis - gundy inequality , young s inequality and assumption h.2 lcl [ eq18 ] 2 + c_3 ^ 1/2 + c_3 ^ 1/2 + c_3k^1/2 + c_3k^1/2 lcl + ( c_3k)^2 + + ( c_3k)^2(t_n ) . again using burkholder - davis - gundy inequality , young s inequality and assumption h.2 lcl [ eq19 ] 2 + c_4 ^1/2 + c_4 ^1/2 + c_4k ^1/2 + + ( c_4k)^2 + + ( c_4k)^2(t_n ) . substituting equations and in equation and rearranging the terms we have rcl & [ & _ 0tt_n ( u^n(t)_^2 ^ 2+^n(t)_l^2 ^ 2)]+2 + & & c^+ + & & + 2+c ( t_n)+2[u^n_0_^2 ^ 2+^n_0_l^2 ^ 2 ] , where @xmath185 $ ] and @xmath186 $ ] . now taking limit as @xmath183 and gronwall s inequality we get the desired a priori estimate . let @xmath188 . we assume that @xmath15 and @xmath16 satisfy the following hypotheses : 1 . @xmath17\times{\mathbb{h}_0 ^ 1}(\mathcal{o});l_q(h_0,{\mathbb{l}^2 } ) ) , h\in\mathbb{h}^p_\lambda([0,t]\times z;{\mathbb{l}^2}(\mathcal{o}))$ ] , 2 . for all @xmath18 , there exists a positive constant @xmath19 such that for all @xmath20 @xmath151 3 . for all @xmath18 , there exists a positive constant @xmath22 such that for all @xmath23 @xmath189 4 . for all @xmath18 , there exists a positive constant @xmath190 such that for all @xmath188 and @xmath20 @xmath191 let @xmath188 and @xmath192\times{\mathbb{h}_0 ^ 1}(\mathcal{o});l_q(h_0,{\mathbb{l}^2})),\;h\in\mathbb{h}^p_\lambda([0,t]\times z;{\mathbb{l}^2}(\mathcal{o})),\\ u_0\in{\mathbb{l}^2}(\mathcal{o}),{\hat{z}}_0\in l^2(\mathcal{o } ) . \end{array}\right.\ ] ] let @xmath162 be an adapted process in @xmath193;{\mathbb{l}^2}(\mathcal{o}))\cap l^2([0,t];{\mathbb{h}_0 ^ 1}(\mathcal{o}))\cap\mathcal{d}(0,t,{\mathbb{l}^2}(\mathcal{o}))$ ] which solves the differential equations and . then we have the following a priori estimate : l [ eq16 ] [ _ 0ttu^n(t)_^2^p+^n(t)_l^2^p]+p[_0^t u^n(t)_^2^p-2u^n(t)__0 ^ 1 ^ 2 dt]c_1(p ) , + where the constant @xmath194 depends on the coefficients @xmath166 and the norms @xmath195 @xmath196 and @xmath169 . the proposition can be proved using the same ideas used in proposition [ prop ] . a path - wise strong solution @xmath66 is defined on a given filtered probability space @xmath110 as a @xmath197 valued function which satisfies the stochastic tide equations and in the weak sense and also the energy inequalities in proposition [ prop ] [ thmuniq ] let @xmath12 and @xmath13 be such that @xmath198 suppose @xmath15 and @xmath16 satisfy the conditions in h.1-h.3 , then there exist a unique path - wise strong solution @xmath25 and @xmath26 with the regularity @xmath199 satisfying the stochastic equations - and the apriori bounds - * existence * : + define @xmath200 then @xmath201 using the a priori estimates - , it follows from the banach - alaoglu theorem that along a subsequence , the galerkin approximations @xmath202 have the following limits : rcl & & u^nu l^2(;l^(0,t,^2())l^2(0,t;_0 ^ 1 ( ) ) ) , + & & ^nl^2(;l^2(0,t;l^2 ( ) ) ) , + & & f(u^n)f_0l^2(;l^2(0,t;^-1 ( ) ) ) , + & & ( , u^n)_0l^2(;l^2(0,t;l_q ) ) , + & & h^n(u^n,)h_0_^2 ( [ 0,t]z;^2 ) , where @xmath66 has the differential form @xmath203 weakly in @xmath204 . + + applying it s formula to @xmath205 and the process @xmath162 rcl d[e^-ltu^n(t)_^2 ^ 2 & + & e^-ltg^n(t)_l^2 ^ 2 ] + & & = -le^-ltu^n(t)_^2 ^ 2 dt -le^-ltg^n(t)_l^2 ^ 2 dt + & & -e^-lt(2f(u^n(t)),hu^n(t))_^2 dt + & & + 2e^-lt(^n(t , u^n(t)),hu^n(t))_^2dw^n(t ) + & & + e^-lt^n(t , u^n(t))_l_q^2dt + & & + 2e^-lt_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) + & & + e^-lt_z h^n(u^n(t-),z)^2_^2 n(dt , dz ) . integrating from @xmath206 to @xmath169 and taking expectation lcl & [ & e^-ltu^n(t)_^2 ^ 2 + e^-ltg^n(t)_l^2 ^ 2 - u^n(0)_^2 ^ 2 - g^n(0)_l^2 ^ 2 ] + & = & -[_0^t le^-ltu^n(t)_^2 ^ 2 dt ] -[_0^t le^-ltg^n(t)_l^2 ^ 2 dt ] + & & -2[_0^t e^-lt(f(u^n(t)),hu^n(t))_^2 dt ] + & & + [ _ 0^t 2e^-lt(^n(t , u^n(t)),hu^n(t))_^2dw^n(t ) ] + & & + [ _ 0^t e^-lt^n(t , u^n(t))_l_q^2dt ] + & & + [ _ 0^t 2e^-lt_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) ] lcl & & + [ _ 0^t e^-lt_z h^n(u^n(t-),z)^2_^2 n(dt , dz ) ] . since @xmath207,\ ] ] and @xmath208,\ ] ] are martingales with zero averages , and since @xmath209 is strong 1-integrable w.r.t @xmath180 , lrl [ _ 0^t e^-lt_z h^n(u^n(t-),z)^2_^2 n(dt , dz ) ] + = [ _ 0^t e^-lt_z h^n(u^n(t-),z)^2_^2 ( dz)dt ] . so lcl [ e^-ltu^n(t)_^2 ^ 2 + e^-ltg^n(t)_l^2 ^ 2 - u^n(0)_^2 ^ 2 - g^n(0)_l^2 ^ 2 ] + = -[_0^t le^-ltu^n(t)_^2 ^ 2 dt ] -[_0^t le^-ltg^n(t)_l^2 ^ 2 dt ] + -2[_0^t e^-lt(f(u^n(t)),hu^n(t))_^2 dt ] + [ _ 0^t e^-lt^n(t , u^n(t))_l_q^2dt ] + + [ _ 0^t e^-lt_z h^n(u^n(t-),z)^2_^2 ( dz)dt ] . using the lower semi - continuity of @xmath92-norm lcl [ eq25 ] _ n \{-[_0^t le^-ltu^n(t)_^2 ^ 2 dt ] -[_0^t le^-ltg^n(t)_l^2 ^ 2 dt ] + -2[_0^t e^-lt(f(u^n(t)),hu^n(t))_^2 dt ] + [ _ 0^t e^-lt^n(t , u^n(t))_l_q^2dt ] + + [ _ 0^t e^-lt_z h^n(u^n(t-),z)^2_^2 ( dz)dt ] } + = _ n \{[e^-ltu^n(t)_^2 ^ 2 + e^-ltg^n(t)_l^2 ^ 2 + - u^n(0)_^2 ^ 2 - g^n(0)_l^2 ^ 2 ] } + [ e^-ltu(t)_^2 ^ 2 + e^-ltg(t)_l^2 ^ 2 - u(0)_^2 ^ 2 - g(0)_l^2 ^ 2 ] + = -[_0^t le^-ltu(t)_^2 ^ 2 dt ] -[_0^t le^-ltg(t)_l^2 ^ 2 dt ] + -2[_0^t e^-lt(f_0(t),hu(t))_^2 dt ] + [ _ 0^t e^-lt_0(t)_l_q^2dt ] + + [ _ 0^t e^-lt_z h_0(t , z)^2_^2 ( dz)dt ] . using the monotonicity property of @xmath210 and assumption h.3 , we have for all @xmath211 lcl -2[_0^t e^-lt(f(u^n(t))-f(v(t)),hu^n(t)-hv(t))_^2 dt ] + + [ _ 0^t e^-lt^n(t , u^n(t))-^n(t , v(t))_l_q^2dt ] + + [ _ 0^t e^-lt_zh^n(u^n(t-),z)-h^n(v(t-),z)^2_^2 ( dz)dt ] + -[_0^t le^-ltu^n(t)-v(t)_^2 ^ 2 dt ] + -[_0^t le^-ltg-^n_l^2 ^ 2 dt ] + 0 . rearranging the terms lcl & -2&[_0^t e^-lt(f(u^n(t)),hu^n(t))_^2 dt ] + [ _ 0^t e^-lt^n(t , u^n(t))_l_q^2dt ] + & + & [ _ 0^t e^-lt_zh^n(u^n(t-),z)^2_^2 ( dz)dt ] -[_0^t le^-ltu^n(t)_^2 ^ 2 dt ] + & -&[_0^t le^-ltg^n_l^2 ^ 2 dt ] + & & -2[_0^t e^-lt(f(v(t)),hu^n(t)-hv(t))_^2 dt ] -[_0^t e^-lt^n(t , v(t))_l_q^2dt ] + & & + 2[_0^t e^-lt(^n(u^n(t)),^n(v(t)))_l_qdt ] + & & -[_0^t e^-lt_zh^n(v(t-),z)^2_^2 ( dz)dt ] + & & + 2[_0^t e^-lt_z ( h^n(u^n(t-),z),h^n(v(t-),z))_^2 ( dz)dt ] + & & + [ _ 0^t le^-ltv(t)_^2 ^ 2 dt ] -2[_0^t le^-lt(u^n(t),v(t))_^2 dt ] + & & + [ _ 0^t le^-ltg_l^2 ^ 2 dt ] -2[_0^t le^-ltg(^n(t),(t))_l^2 dt ] + & & -2[_0^t e^-lt(f(u^n(t)),hv(t))_^2 dt ] . taking limit in @xmath212 and using lcl & -2&[_0^t e^-lt(f_0(t),hu(t))_^2 dt ] + [ _ 0^t e^-lt_0(t)_l_q^2dt ] + & + & [ _ 0^t e^-lt_zh_0(t , z)^2_^2 ( dz)dt ] -[_0^t le^-ltu(t)_^2 ^ 2 dt ] lcl & -&[_0^t le^-ltg_l^2 ^ 2 dt ] + & & -2[_0^t e^-lt(f(v(t)),hu(t)-hv(t))_^2 dt ] + & & -[_0^t e^-lt(t , v(t))_l_q^2dt ] + & & + 2[_0^t e^-lt(_0(t),(v(t)))_l_qdt ] + & & -[_0^t e^-lt_zh(v(t-),z)^2_^2 ( dz)dt ] + & & + 2[_0^t e^-lt_z ( h_0(t , z),h(v(t-),z))_^2 ( dz)dt ] + & & + [ _ 0^t le^-ltv(t)_^2 ^ 2 dt ] -2[_0^t le^-lt(u(t),v(t))_^2 dt ] + & & + [ _ 0^t le^-ltg_l^2 ^ 2 dt ] -2[_0^t le^-ltg((t),(t))_l^2 dt ] + & & -2[_0^t e^-lt(f_0(t),hv(t))_^2 dt ] . rearranging the terms lcl -2[_0^t e^-lt(f_0(t)-f(v(t)),hu(t)-hv(t))_^2 dt ] + + [ _ 0^t e^-lt_0(t)-(t , v(t))_l_q^2dt ] + + [ _ 0^t e^-lt_zh_0(t , z)-h(v(t-),z)^2_^2 ( dz)dt ] + -[_0^t le^-ltu(t)-v(t)_^2 ^ 2 dt ] + 0 . this estimate holds for any @xmath211 , for any @xmath213 . it is obvious from the density argument that the above inequality remains the same for any @xmath214 . in fact , for any @xmath214 there exists a strongly convergent subsequence @xmath215 satisfying the above inequality . + + taking @xmath216 , we get @xmath217 and @xmath218 . + now we take @xmath219 where @xmath220 and @xmath221 is an adapted process in @xmath222 . + then we have @xmath223\geq \lambda\mathbb{e}[\int_0^t e^{-lt}(f_0(t),hw(t))_{{\mathbb{l}^2 } } dt].\ ] ] dividing by @xmath224 on both sides on the inequality above and letting @xmath224 go to 0 , we have by the hemicontinuity of @xmath210 @xmath225\geq 0.\ ] ] since @xmath221 is arbitrary and @xmath175 is a positive , bounded , continuously differentiable function , @xmath226 . this proves the existence of a strong solution . + + * uniqueness * : + if @xmath66 and @xmath67 are two solutions then @xmath227 solves the stochastic differential equation rcl dw(t)+aw(t)dt&+&g((t)-(t))dt + & = & b(v(t))-b(u(t))dt+((t , u(t))-(t , v(t)))dw(t ) + & & + _ z ( h(u(t-),z)-h(v(t-),z))(dt , dz ) . applying it s formula to @xmath170 and to the process @xmath228 rcl d(w(t)_^2 ^ 2)&+&2w(t)__0 ^ 1 ^ 2 dt+2g(((t)-(t)),w(t))_^2 dt + & = & 2(b(v(t))-b(u(t)),u(t)-v(t))_^2dt + & & + 2((t , u(t))-(t , v(t)),w(t))_^2 dw(t ) + & & + ( t , u(t))-(t , v(t))_l_q^2 dt + & & + _ z h(u(t-),z)-h(v(t-),z)^2_^2 n(dt , dz ) + & & + _ z ( ( h(u(t-),z)-h(v(t-),z)),w(t-))_^2(dt , dz ) . using the result from lemma [ mon ] and equation rcl [ eq27 ] d&(&w(t)_^2 ^ 2)+2w(t)__0 ^ 1 ^ 2 dt + & & ( t)-(t)_l^2 ^ 2 dt+w(t)__0 ^ 1 ^ 2 dt + & & + 2((t , u(t))-(t , v(t)),w(t))_^2 dw(t)+(t , u(t))-(t , v(t))_l_q^2 dt + & & + _ z h(u(t-),z)-h(v(t-),z)^2_^2 n(dt , dz ) + & & + _ z ( ( h(u(t-),z)-h(v(t-),z)),w(t-))_^2(dt , dz ) . notice that @xmath229 taking inner product with @xmath230 , we have as in equation lcl [ eq28 ] d((t)-(t)_l^2 ^ 2 ) + mw(t)_^2 ^ 2 dt + ( + m)(t)-(t)_l^2 ^ 2 dt + w(t)__0 ^ 1 ^ 2 dt . let @xmath231 adding equations and rcl d&(&w(t)_^2 ^ 2+(t)-(t)_l^2 ^ 2)+w(t)__0 ^ 1 ^ 2 dt + & & c(w(t)_^2 ^ 2+(t)-(t)_l^2 ^ 2 ) dt + & & + 2((t , u(t))-(t , v(t)),w(t))_^2 dw(t)+(t , u(t))-(t , v(t))_l_q^2 dt + & & + _ z h(u(t-),z)-h(v(t-),z)^2_^2 n(dt , dz ) + & & + _ z ( ( h(u(t-),z)-h(v(t-),z)),w(t-))_^2(dt , dz ) . integrating from 0 to @xmath169 and taking expectation rcl & [ & w(t)_^2 ^ 2+(t)-(t)_l^2 ^ 2]+[_0^t w(t)__0 ^ 1 ^ 2 dt ] + & & c[_0^t(w(t)_^2 ^ 2+(t)-(t)_l^2 ^ 2 ) dt ] + & & + [ _ 0^t ( t , u(t))-(t , v(t))_l_q^2 dt ] + & & + [ _ 0^t _ z h(u(t-),z)-h(v(t-),z)^2_^2 ( dz)dt ] + & & + [ w(0)_^2 ^ 2+(0)-(0)_l^2 ^ 2 ] . using assumption h.3 rcl & [ & w(t)_^2 ^ 2+(t)-(t)_l^2 ^ 2]+[_0^t w(t)__0 ^ 1 ^ 2 dt ] + & & c[_0^t(w(t)_^2 ^ 2+(t)-(t)_l^2 ^ 2 ) dt ] + l[_0^t w(t)_^2 ^ 2 dt ] + & & + [ w(0)_^2 ^ 2+(0)-(0)_l^2 ^ 2 ] . in particular rcl [ w(t)_^2 ^ 2&+&(t)-(t)_l^2 ^ 2 ] + & & [ w(0)_^2 ^ 2+(0)-(0)_l^2 ^ 2 ] + & & + ( c+l)[_0^t(w(t)_^2 ^ 2+(t)-(t)_l^2 ^ 2 ) dt ] . + hence the uniqueness of pathwise strong solution follows using gronwall s inequality . we assume that @xmath15 and @xmath16 satisfy the following hypotheses : 1 . @xmath17\times { \mathbb{h}_0 ^ 1}(\mathcal{o});l_q(h_0,{\mathbb{h}_0 ^ 1 } ) ) , h\in\mathbb{h}^2_\lambda([0,t]\times z;{\mathbb{h}_0 ^ 1}(\mathcal{o}))$ ] , 2 . for all @xmath18 , there exists a positive constant @xmath19 such that for all @xmath232 @xmath233 3 . for all @xmath18 , there exists a positive constant @xmath22 such that for all @xmath234 @xmath235 let @xmath236\times { \mathbb{h}_0 ^ 1}(\mathcal{o});l_q(h_0,{\mathbb{h}_0 ^ 1})),\;h\in\mathbb{h}^2\lambda([0,t]\times z;{\mathbb{h}_0 ^ 1}(\mathcal{o})),\\ u_0\in{\mathbb{h}_0 ^ 1}(\mathcal{o}),{\hat{z}}_0\in h^1_0(\mathcal{o } ) . \end{array}\right.\ ] ] let @xmath162 be an adapted process in @xmath237;w^{2,2}_0(\mathcal{o}))\cap l^\infty([0,t];{\mathbb{h}_0 ^ 1}(\mathcal{o}))\cap\mathcal{d}(0,t,{\mathbb{h}_0 ^ 1}(\mathcal{o}))$ ] which solves the differential equations and . then we have lr _ t0_n ( _ 0tt_nu^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2 + _ 0^t_nu^n(t)^2_^2 dt + > ( n-1)+u^n_0__0 ^ 1 ^ 2+^n_0_h^1_0 ^ 2)=0 . applying it s formula to @xmath170 and to the process @xmath238 lr d(u^n(t)__0 ^ 1 ^ 2)+ 2u^n(t)^2_^2dt+2(b(u^n),u^n)_^2dt + + 2((g^n ) , u^n)_^2dt + = 2(f , u^n(t))_^2dt+2(^n(t , u^n(t)),u^n(t))_^2dw^n(t ) + + ^n(t , u^n(t))_l_q^2dt + _ zh^n(u^n(t-),z)_^2 ^ 2 n(dt , dz ) + + 2_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) . using divergence theorem lr d(u^n(t)__0 ^ 1 ^ 2)+ 2u^n(t)^2_^2 dt + = 2 ( b(u^n),u^n)_^2dt+2(g^n , u^n)_^2dt+2(f , u^n(t))_^2dt + + 2(^n(t , u^n(t)),u^n(t))_^2dw^n(t)+^n(t , u^n(t))_l_q^2dt + + _ zh^n(u^n(t-),z)_^2 ^ 2 n(dt , dz ) + + 2_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) . using young s inequality @xmath239 and @xmath240.\ ] ] using lemma [ lemw ] , equation and young s inequality lrl 2(b(u^n),u^n)_^2&&b(u^n)_^2 ^ 2+u(t)_^2 ^ 2 + & & c_2(u^n(t)^4_^4+w^0(t)_^4 ^ 4)+u(t)_^2 ^ 2 + & & c_2(2u^n(t)^2_^2u^n(t)__0 ^ 1 ^ 2+w^0(t)_^4 ^ 4)+u(t)_^2 ^ 2 . now using the a - priori estimate given by equation we get @xmath241 hence lr d(u^n(t)__0 ^ 1 ^ 2)+ u^n(t)^2_^2 dt + ( 2c_2c_1(2)+1)u^n(t)__0 ^ 1 ^ 2dt+^n(t)_h_0 ^ 1 ^ 2+f(t)__0 ^ 1 ^ 2dt+c_2w^0(t)_^4 ^ 4 + + 2(^n(t , u^n(t)),u^n(t))_^2dw^n(t)+^n(t , u^n(t))_l_q^2dt + + _ zh^n(u^n(t-),z)_^2 ^ 2 n(dt , dz ) + + 2_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) . using assumption a.2 lr [ eeeeq ] d(u^n(t)__0 ^ 1 ^ 2)+ u^n(t)^2_^2 dt + ( 2c_2c_1(2)+1)u^n(t)__0 ^ 1 ^ 2dt+^n(t)_h_0 ^ 1 ^ 2+f(t)__0 ^ 1 ^ 2dt+c_2w^0(t)_^4 ^ 4 + + 2(^n(t , u^n(t)),u^n(t))_^2dw^n(t)+k(1+u^n(t)__0 ^ 1 ^ 2)dt + + _ zh^n(u^n(t-),z)_^2 ^ 2 n(dt , dz ) + + 2_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) . using @xmath242 now rcl |(div(hu^n(t)),^n(t))_l^2| & & u^n(t)__0 ^ 1^n(t)_l^2 + mu^n(t)_^2^n(t)_l^2 + & & [ u^n(t)__0 ^ 1 ^ 2 + ^n(t)_l^2 ^ 2 ] + & & + [ u^n(t)_^2 ^ 2+^n(t)_l^2 ^ 2 ] . using poincar s inequality @xmath243 and rearranging @xmath244 adding equations and lr d(u^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2)+ u^n(t)^2_^2 dt + ( 2c_2c_1(2)+1+mc_p^2+)u^n(t)__0 ^ 1 ^ 2dt+^n(t)_h_0 ^ 1 ^ 2 dt + + ( + m)^n(t)_l^2 ^ 2 dt+f(t)__0 ^ 1 ^ 2dt+c_2w^0(t)_^4 ^ 4 + + 2(^n(t , u^n(t)),u^n(t))_^2dw^n(t)+k(1+u^n(t)__0 ^ 1 ^ 2)dt + + _ zh^n(u^n(t-),z)_^2 ^ 2 n(dt , dz ) + + 2_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) . define lrl _ n : = \{t : u^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2&+&_0^tu^n(s)^2_^2ds+^n(t)_^2 ^ 2 + & & > n+u^n_0__0 ^ 1 ^ 2+^n_0__0 ^ 1 ^ 2}. integrating and taking supremum over @xmath245 lr _ 0tt_n(u^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2)+ _ 0^t_nu^n(t)^2_^2 dt + s_0^t_n(u^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2 + ^n(t)_l^2 ^ 2+f(t)__0 ^ 1 ^ 2+w^0(t)_^4 ^ 4 + 1 ) dt + + 2_0tt_n_0^t((^n(t , u^n(t)),u^n(t))_^2dw^n(t ) + + _ 0^t_n_z ( h^n(u^n(t-),z),u^n(t-))_^2(dt , dz ) + + _ 0^t_n_zh^n(u^n(t-),z)_^2 ^ 2 n(dt , dz ) ) + + u^n_0__0 ^ 1 ^ 2+^n_0_h^1_0 ^ 2 , where @xmath246 . + hence lr ( _ 0tt_nu^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2 + _ 0^t_nu^n(t)^2_^2 dt . + . > ( n-1)+u^n_0__0 ^ 1 ^ 2+^n_0_h^1_0 ^ 2 ) + ( s_0^t_n(u^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2+^n(t)_l^2 ^ 2 . + f(t)__0 ^ 1 ^ 2+w^0(t)_^4 ^ 4 + 1 ) dt>(n-1)/2 ) + + 2(_0tt_n_0^t((^n(t , u^n(t)),u^n(t))_^2dw^n(t ) . + + _ 0^t_n_z ( h^n(u(t-),z),u^n(t-))_^2(dt , dz ) + .+_0^t_n_zh^n(u(t-),z)_^2 ^ 2 n(dt , dz ) ) > ( n-1)/2 ) . now lr ( _ 0tt_n_0^t((^n(t , u^n(t)),u^n(t))_^2dw^n(t ) . + + _ 0^t_n_z ( h^n(u(t-),z),u^n(t-))_^2(dt , dz ) + .+_0^t_n_zh^n(u(t-),z)_^2 ^ 2 n(dt , dz ) ) > ( n-1)/2 ) lr ( _ 0tt_n_0^t(^n(t , u^n(t)),u^n(t))_^2dw^n(t)>(n-1)/6 ) + + ( _ 0tt_n_0^t _ z ( h^n(u(t-),z),u^n(t-))_^2(dt , dz)>(n-1)/6 ) + + ( _ 0tt_n_0^t_zh^n(u(t-),z)_^2 ^ 2 n(dt , dz ) > ( n-1)/6 ) . using doob s inequality lr ( _ 0tt_n_0^t(^n(t , u^n(t)),u^n(t))_^2dw^n(t)>(n-1)/6 ) + [ _ 0^t_n u^n(t)__0 ^ 1 ^ 2^n(t , u^n(t))_l_q^2 dt ] + k[_0^t_n u^n(t)__0 ^ 1 ^ 2(1 + u^n(t)__0 ^ 1 ^ 2 ) dt ] + kn(n+1)(t_n ) . using doob s inequality again lr ( _ 0tt_n_0^t_z ( h^n(u(t-),z),u^n(t-))_^2(dt , dz)>(n-1)/6 ) + [ _ 0^t_n u^n(t)__0 ^ 1 ^ 2_zh^n(u(t-),z)_^2 ^ 2 ( dz)dt ] + k[_0^t_n u^n(t)__0 ^ 1 ^ 2(1 + u^n(t)__0 ^ 1 ^ 2 ) dt ] + kn(n+1)(t_n ) . using doob s inequality and strong 2-integrability of @xmath247 lr ( _ 0tt_n_0^t_zh^n(u(t-),z)_^2 ^ 2 n(dt , dz ) > ( n-1)/6 ) + [ _ 0^t_n_zh^n(u(t-),z)_^2 ^ 2 ( dz)dt ] + k[_0^t_n(1+u(t)__0 ^ 1 ^ 2 ) dt ] + k(n+1)(t_n ) . using chebychev s inequality lr ( s_0^t_n(u^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2+^n(t)_l^2 ^ 2 . + f(t)__0 ^ 1 ^ 2+w^0(t)_^4 ^ 4 + 1 ) dt>(n-1)/2 ) + [ s_0^t_n(u^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2+^n(t)_l^2 ^ 2 + + f(t)__0 ^ 1 ^ 2+w^0(t)_^4 ^ 4 + 1 ) dt ] lr 3ns ( t_n)+s[_0^t_n(f(t)__0 ^ 1 ^ 2+w^0(t)_^4 ^ 4 + 1)dt ] . hence lr _ t0_n ( _ 0tt_nu^n(t)__0 ^ 1 ^ 2+^n(t)_h_0 ^ 1 ^ 2 + _ 0^t_nu^n(t)^2_^2 dt . + . > ( n-1)+u^n_0__0 ^ 1 ^ 2+^n_0_h^1_0 ^ 2)=0 . let @xmath97 be a separable and complete metric space . let @xmath106 and let @xmath248 be given . a modulus of @xmath66 is defined by @xmath249,\mathbb{s}}(u,\delta):=\inf_{\pi_\delta}\max_{t_i\in\overline{\omega}}\sup_{t_i\leq s < t < t_{i+1}\leq t}\rho(u(t)u(s)),\ ] ] where @xmath250 is the set of all increasing sequences @xmath251 with the following property @xmath252 taking the path space @xmath258 into account , we call @xmath259 , where @xmath260 denotes the extended skorokhod topology , @xmath261 , where @xmath262 denotes the weak - star topology and @xmath263 , where @xmath221 denotes the weak topology and @xmath264 as the strong topology of @xmath265 . note that the spaces @xmath266 are completely regular and continuously embedded in @xmath267 . let @xmath268 be the supremum of four topologies , that is , @xmath269 . we can assume that @xmath19 is a closed subset of @xmath276 . by condition ( a ) the weak - star topology in @xmath277 induced on @xmath276 is metrizable . similarly by condition ( b ) the weak topology in @xmath278 induced on @xmath276 is metrizable . thus the compactness of a subset of @xmath276 is equivalent to its sequential compactness . let @xmath279 be a sequence in @xmath19 . by the banach - alaoglu theorem @xmath19 is compact in @xmath277 as well as @xmath278 . + we need to prove that @xmath279 is compact in @xmath280 . by condition ( a ) for every @xmath272 $ ] the set @xmath281 is bounded in @xmath4 . since the embedding @xmath282 is compact , @xmath281 is compact in @xmath135 . using this and condition ( c ) , theorem [ cmpct ] implies compactness of the sequence @xmath279 in the space @xmath280 . therefore , there exists a subsequence denoted again by @xmath279 such that @xmath283 since @xmath284 in @xmath285 in @xmath135 for all continuity points of @xmath66 . hence , by condition ( a ) and lebesgue dominated convergence theorem @xmath286 to complete the proof , we need to show that @xmath287 since embeddings in the gelfand triple are compact , by lions @xcite , for every @xmath288 , there exists a constant @xmath289 such that @xmath290 thus @xmath291 by condition ( b ) @xmath292 where @xmath293 . hence @xmath294 as @xmath288 is arbitrary , @xmath295 let @xmath97 be a separable and complete metric space . let @xmath296 be a probability space with the filtration @xmath297}$ ] satisfying the usual hypotheses , and let @xmath298 be a sequence of cdlg , @xmath210-adapted and @xmath99-valued processes . @xmath298 is said to satisfy the aldous condition iff @xmath299 such that for every sequence @xmath300 of stopping times with @xmath301 @xmath302 [ aldt ] let @xmath303 satisfy the aldous condition . let @xmath156 be the law of @xmath304 on @xmath305 . then for every @xmath288 there exists a subset @xmath306 such that @xmath307 and @xmath308}(u,\delta)=0.\ ] ] [ ald ] let @xmath309 be a separable banach space and let @xmath310 be a sequence of @xmath111-valued random variables . assume that for every sequence @xmath300 of @xmath210-stopping times with @xmath301 and for every @xmath213 and @xmath311 the following condition holds @xmath312\leq c\theta^{\beta},\ ] ] for some @xmath313 and some constant @xmath314 . then the sequence @xmath298 satisfies the aldous condition in the space @xmath111 . we use the tightness condition for the prokhorov - varadarajan theorem which states that a sequence of measures @xmath315 is tight on a topological space @xmath111 if for every @xmath288 there exists a compact set @xmath316 such that @xmath317 . hence the tightness of measure in @xmath276 is given by the following theorem . * there exists a positive constant @xmath318 such that @xmath319}\|x_n(s)\|_{{\mathbb{l}^2}}]\leq c_1,\ ] ] * there exists a positive constant @xmath145 such that @xmath320\leq c_2,\ ] ] * @xmath298 satisfies the aldous condition in @xmath135 . let @xmath156 be the law of @xmath304 on @xmath276 . then for every @xmath288 there exists a compact subset @xmath321 of @xmath276 such that @xmath322 and the sequence of measures @xmath323 is said to be tight on @xmath324 . let @xmath288 . by the chebyshev inequality and by ( a ) , we see that for any @xmath325 @xmath326}\|x_n(s)\|_{{\mathbb{l}^2}}>r)\leq\dfrac{\mathbb{e}[\sup_{s\in[0,t]}\|x_n(s)\|_{{\mathbb{l}^2}}]}{r}\leq \dfrac{c_1}{r}.\ ] ] let @xmath327 be such that @xmath328 . then @xmath326}\|x_n(s)\|_{{\mathbb{l}^2}}>r_1)\leq\dfrac{\epsilon}{3}.\ ] ] let @xmath329}\|u(s)\|_{{\mathbb{l}^2}}\leq r_1\}$ ] . + + by the chebyshev inequality and by ( b ) , we see that for any @xmath325 @xmath330}{r^2}\leq \dfrac{c_2}{r^2}.\ ] ] let @xmath331 be such that @xmath332 . then @xmath333 let @xmath334 . + + by lemma [ aldt ] , there exists a subset @xmath335 such that @xmath336 and @xmath337}(u,\delta)=0.\ ] ] it is sufficient to define @xmath321 as the closure of the set @xmath338 in @xmath276 . by theorem [ cpt ] , @xmath321 is compact in @xmath276 . 1 . @xmath17\times{\mathbb{h}_0 ^ 1}(\mathcal{o});l_q(h_0,{\mathbb{l}^2 } ) ) , h\in\mathbb{h}^2_\lambda([0,t]\times z;{\mathbb{l}^2}(\mathcal{o}))$ ] , 2 . for all @xmath18 , there exists a positive constant @xmath19 such that for all @xmath20 @xmath151 3 . for all @xmath18 , there exists a positive constant @xmath22 such that for all @xmath23 @xmath24 * @xmath343 is a filtered probability space with a filtration @xmath344 , * @xmath345 is a time homogeneous poisson random measure over @xmath343 with the intensity measure @xmath224 , * @xmath346 is a cylindrical wiener process over @xmath343 , * for all @xmath272 $ ] @xmath347 @xmath348 @xmath349 such that for all @xmath350 and for all @xmath351 , the following identities hold @xmath352-a.s . + llr & ( ( t),v)_^2+_0^t ( a((s)),v)_^2ds+_0^t(b((s)),v)_^2ds+_0^t(g(s ) , v)_^2ds & + & = ( u_0,v_^2)+(,v)_^2+_0^t(f(s),v)_^2ds + & + _ 0^t((s,(s)),v)dw(s)_^2 + _ 0^t_z ( h((s-),z),v)_^2 ( ds , dz ) . + & ( ( t),w)_l^2+_0^t ( div(h(s)),w)_l^2 ds=(z_0,w)_l^2 . we prove the tightness of the measures using the tightness criterion given in theorem [ sdcthm1 ] . by proposition [ prop ] , conditions ( a ) and ( b ) are satisfied . we use lemma [ ald ] to prove the aldous condition for the sequence @xmath355 in the space @xmath135 . let @xmath300 be a sequence of stopping times where @xmath356 . then using lrl u^n(t)=u^n_0+u^n-_0^t a(u^n(s))ds-_0^t b(u^n(s ) ) ds-_0^t g^n(s ) ds + + _ 0^t f(s)ds+_0^t ^n(s , u^n(s))dw^n(s)+_0^t _ z h^n(u^n(s-),z)(ds , dz ) + = j^n_1+j^n_2+j^n_3(t)+j^n_4(t)+j^n_5(t)+j^n_6(t)+j^n_7(t)+j^n_8(t),t . let @xmath357 . we need to show that holds for each @xmath358 . since @xmath359 and @xmath360 are independent of time , they clearly satisfy . + + since @xmath361 , for all @xmath350 @xmath362 hence using the above inequality @xmath363 therefore lrl & [ & j^n_4(_n+)-j^n_4(_n)_^-1 ] + & = & [ _ _ n^_n+b(u^n(s))ds_^-1 ] + & & c[__n^_n+b(u^n(s))_^2ds ] + & & c[__n^_n+ b(u^n(s))_^2ds ] + & & cc_2[__n^_n+u^n+w^0(s)_^4 ^ 2 ds ] + & & cc_2[__n^_n+ u^n+w^0(s)_^2u^n+w^0(s)__0 ^ 1 ds ] + & & c_3[(__n^_n+ u^n+w^0(s)_^2 ^ 2 ds)^1/2 ( _ _ n^_n+ u^n + w^0(s)__0 ^ 1 ^ 2 ds)^1/2 ] + & & c_3[(_0stu^n+w^0(s)_^2 ^ 2 ) ^1/2 ( _ 0^t u^n + w^0(s)__0 ^ 1 ^ 2 ds)^1/2 ] + & & c_4 ^ 1/2 . lrl [ j^n_7(_n+)-j^n_7(_n)^2_^-1]&=&[__n^_n+ ^n(s , u^n(s ) ) dw^n(s)^2_^-1 ] + & & c[__n^_n+ ^n(s , u^n(s ) ) dw^n(s)_^2 ^ 2 ] + & = & c [ _ _ n^_n+ ^n(s , u^n(s))_l_q^2 ds ] + & & ck [ _ _ n^_n+ ( 1 + u^n(s))_^2 ^ 2 ) ds ] + & & ck(1+[_0stu^n(s)_^2 ^ 2 ] ) + & & c_6 . thus @xmath376 satisfies @xmath365 with @xmath377 and @xmath378 . + since @xmath368 , using the hypothesis s.2 , and the isometry @xcite @xmath379=\mathbb{e}[\int_{0}^{t}\int_z \| h^n(u^n(s-),z)\|_{{\mathbb{l}^2}}^2\lambda(dz)ds],\ ] ] we get lrl [ j^n_8(_n+)-j^n_8(_n)^2_^-1]&=&[__n^_n+ _ z h^n(u^n(s-),z)(ds , dz)_^-1 ^ 2 ] + & & c[__n^_n+_z h^n(u^n(s-),z)(ds , dz)_^2 ^ 2 ] + & = & c[__n^_n+_z h^n(u^n(s-),z)_^2 ^ 2(dz)ds ] + & & ck[__n^_n+ ( 1 + u^n(s))_^2 ^ 2 ) ds ] + & & ck(1+[_0stu^n(s)_^2 ^ 2 ] ) + & & c_6 . using proposition [ prop ] , @xmath384\leq c_1 $ ] , hence using fubini s theorem @xmath385\leq c_1t.\ ] ] using the chebychev inequality , we see that for any @xmath325 @xmath386}{r^2}\leq \dfrac{c_1t}{r^2}.\ ] ] let @xmath327 be such that @xmath387 . then @xmath388 define @xmath389 hence @xmath390 . using the assumption @xmath392\leq c_c$ ] , hence using fubini s theorem @xmath393\leq c_ct.\ ] ] using the chebychev inequality , we see that for any @xmath325 @xmath394}{r^2}\leq \dfrac{c_ct}{r^2}.\ ] ] let @xmath327 be such that @xmath395 . then @xmath396 define @xmath397 hence @xmath390 . by lemma [ tight ] , the set of measures @xmath354 is tight on the space @xmath324 , by lemma [ tight2 ] , the set of measures @xmath382 is tight on the space @xmath398 , and by lemma [ tight3 ] , the set of measures @xmath399 is tight on the space @xmath398 . define @xmath400 . the set of measures @xmath401 is tight . define @xmath402 . the set of measures @xmath403 is tight . thus the set @xmath404 is tight . by the skorokhod embedding theorem @xcite , there exists a subsequence @xmath405 , a probability space @xmath406 , and on this space random variables @xmath407 , @xmath408 such that we denote these sequences again by @xmath415 and + @xmath416 . using the definition of the space @xmath276 , we have @xmath417 @xmath418 @xmath419 and @xmath420 since the random variables @xmath421 and @xmath160 are identically distributed , we have @xmath422\leq c_{1(2)},\ ] ] and @xmath423\leq c_{2(2)}.\ ] ] since the random variables @xmath424 and @xmath425 are also identically distributed , we have @xmath426\leq c_{1(2)}.\ ] ] using the assumption @xmath427<c_c$ ] , we have @xmath428\leq c_c.\ ] ] define for all @xmath350 lrl k^1(^n,^n,^n,^n,^n , v)(t)&=&(^n_0,v)_^2+(^n , v)_^2+_0^t ( a(^n(s)),v)_^2ds + & & + _ 0^t ( b(^n(s)),v)_^2ds-_0^t(^n(s),v)_^2 ds + & & + _ 0^t ( ^n(s,^n(s)),v)_^2d^n(s ) + & & + _ 0^t_z ( h^n(^n(s-),z),v)_^2^n(ds , dz ) , lrl k^1(u^*,z^*,u^*,n^*,w^*,v)(t)&=&(u^*_0,v)_^2+(u^*,v)_^2+_0^t ( a(u^*(s)),v)_^2ds + & & + _ 0^t ( b(u^*(s)),v)_^2ds-_0^t(z^*(s),v)_^2 ds + & & + _ 0^t ( ( s , u^*(s)),v)_^2dw^*(s ) + & & + _ 0^t_z ( h(u^*(s-),z),v)_^2^*(ds , dz ) , we show the term by term convergence of the above equation in @xmath437 . + since @xmath434 in @xmath277 , we have @xmath352-a.s . @xmath438 hence by and the vitali theorem @xmath439=0.\ ] ] hence @xmath440 since @xmath441 in @xmath442 , we have @xmath352-a.s . @xmath443 since @xmath444 , by the vitali theorem @xmath445=0.\ ] ] hence @xmath446 since @xmath434 in @xmath278 , therefore by vitali theorem , for all @xmath272 $ ] @xmath447=0.\ ] ] hence by the dominated convergence theorem @xmath448=0.\ ] ] since @xmath434 in @xmath265 , we have by the continuity of the non - linear term @xmath140 given by lrl [ |_0^t(b(^n(s)),v)_^2 ds| ] + v_^2 [ _ 0^t b(^n(s))_^2ds ] + c_2v_^2 [ _ 0^t ^n(s ) + w^0(s)_^4 ^ 2 ds ] + c_2v_^2 [ _ 0^t ^n(s ) + w^0(s)_^2^n(s ) + w^0(s)__0 ^ 1ds ] + c_2v_^2 [ ( _ 0^t ^n(s ) + w^0(s)_^2 ^ 2 ds)^1/2(_0^t ^n(s ) + w^0(s)__0 ^ 1 ^ 2ds)^1/2 ] + c_2v_^2 [ ( t_0tt^n(s ) + w^0(s)_^2 ^ 2)^1/2(_0^t ^n(s ) + w^0(s)__0 ^ 1 ^ 2ds)^1/2 ] + < . lrl [ |_0^t |(^n(s,^n(s))-(s , u^*(s)),v)_^2|^2 ds| ] + v_^2 ^ 2 [ _ 0^t ^n(s,^n(s))_l_q^2+(s , u^*(s))_l_q^2 ds ] + kv_^2 ^ 2 [ _ 0^t ( 2+^n(s)_^2 ^ 2 + u^*(s)_^2 ^ 2 ) ds ] + ktv_^2 ^ 2 [ ( 2+_0st^n(s)_^2 ^ 2+_0st u^*(s)_^2 ^ 2 ) ] + < . lrl [ |_0^t _ z |(h^n(^n(s-),z)-h(u^*(s-),z),v)_^2|^2 ( dz)ds| ] + v_^2 ^ 2 [ _ 0^t _ z h^n(^n(s-),z)_^2 ^ 2 + h(u^*(s-),z)_^2 ^ 2 ( dz)ds ] + kv_^2 ^ 2 [ _ 0^t ( 2+^n(s)_^2 ^ 2 + u^*(s)_^2 ^ 2 ) ds ] + ktv_^2 ^ 2 [ ( 2+_0st^n(s)_^2 ^ 2+_0st u^*(s)_^2 ^ 2 ) ] + < . hence by the dominated convergence theorem @xmath460 dt=0.\ ] ] to prove we see that since @xmath451 in @xmath398 , @xmath461 since both @xmath424 and @xmath462 satisfy the inequality , by the vitali theorem we show the term by term convergence of the above equation in @xmath437 . + since @xmath451 in @xmath463 , we have @xmath352-a.s . @xmath464 hence by and the vitali theorem @xmath465=0.\ ] ] hence @xmath466 since @xmath434 in @xmath467 , lrl _ 0^t [ |(u^*(t),v)_^2-k^1(u^*,z^*,u^*,n^*,w^*,v)(t)|^2]dt + _ 0^t [ |(u^*(t),v)_^2-(^n(t),v)_^2|^2]dt + + _ 0^t [ |(^n(t),v)_^2-k^1(^n,^n,^n,^n,^n , v)(t)|^2]dt + + _ 0^t [ |k^1(^n,^n,^n,^n,^n , v)(t)-k^1(u^*,z^*,u^*,n^*,w^*,v)(t)|^2]dt + = 0 . 1 . @xmath479\times{\mathbb{h}_0 ^ 1}(\mathcal{o})\times{\mathbb{l}^2}(\mathcal{o})$ ] is measurable , 2 . @xmath480 is lower semicontinuous @xmath481 $ ] where @xmath482 is the space @xmath58 endowed with the weak topology , and 3 . @xmath483\times{\mathbb{h}_0 ^ 1}(\mathcal{o})\times{\mathbb{l}^2}(\mathcal{o})$ ] where @xmath484 is non - negative and @xmath485 where @xmath491 . the first integral on the right - hand side is zero due to the lemma given below . due to the lower semicontinuity of @xmath492 , we have @xmath493 so @xmath494 now using the beppo - levi theorem on the bounded measurable functions @xmath492 we get the required semicontinuity . let @xmath487 in the @xmath268-topology and @xmath488 in @xmath265-weak . let @xmath495\times{\mathbb{h}_0 ^ 1}(\mathcal{o})\times{\mathbb{l}^2}(\mathcal{o})\rightarrow \mathbb{r}_+$ ] be a bounded measurable function such that , @xmath496 is lower semicontinuous . then @xmath497 define @xmath498 . for @xmath499 and @xmath500 , define @xmath501\times{\mathbb{l}^2}(\mathcal{o});\inf\limits_{|\langle y , u\rangle-\langle y , z\rangle|\leq 1/m}\theta(t , z , u)\leq -\gamma\right\},\ ] ] and @xmath502\times{\mathbb{l}^2}(\mathcal{o});\inf\limits_{|\langle y , u\rangle-\langle y , z\rangle|\leq 1/m}\theta(t , z , u)\leq -\gamma\right\}.\ ] ] clearly , @xmath503 . + + as @xmath487 , we have @xmath504 hence @xmath505 the lower semicontinuity of @xmath132 implies that each @xmath38-section of @xmath506 is closed . hence we have we restrict ourselves to admissible pairs @xmath512 which satisfy - in the martingale sense such that @xmath513 + + we have @xmath514 and @xmath515 now , @xmath516 implies that @xmath517\leq c(r),\ ] ] for some constant @xmath518 . hence by lemma [ tight3 ] there exists a sequence of tight measures @xmath519 in @xmath383 . by theorem [ martexist ] , there exists a corresponding sequence @xmath520 such that the pair @xmath521 gives a martingale solution to - , where @xmath522 in the @xmath268-topology and @xmath523 weakly in @xmath4 . by the same theorem we also have that @xmath524 solves - . + by lemma [ lowersemcont ] , @xmath525 is lower semi - continuous and hence by theorem 55 in chapter iii of @xcite , so is @xmath210 . so , @xmath526 * acknowledgements : * pooja agarwal would like to thank department of science and technology , govt . of india , for the inspire fellowship . utpal manna s work has been supported by the national board of higher mathematics of department of atomic energy , govt . of india , under grant no . nbhm / rp46/2013/fresh/421 . the authors would like to thank indian institute of science education and research(iiser ) - thiruvananthapuram for providing stimulating scientific environment and resources . martingale solutions to the 2d and 3d stochastic navier - stokes equations driven by the compensated poisson random measure ; preprint 13 . department of mathematics and computer sciences , lodz university . stochastic analysis of tidal dynamics equation ; _ infinite dimensional stochastic analysis , special volume in honor of professor hh . , edited by a. sengupta and p. sundar , world scientific publishers .

in this work we first present the existence , uniqueness and regularity of the strong solution of the tidal dynamics model perturbed by lvy noise . monotonicity arguments have been exploited in the proofs . we then formulate a martingale problem of stroock and varadhan associated to an initial value control problem and establish existence of optimal controls .

hassler whitney introduced matroids in 1935 and gave two major examples of matroids , the cycle matroid of a graph and a matroid defined by a finite collection of vectors in a vector space . matroids have since found many applications in other parts of mathematics , including , in particular , topology where the matroidal properties of a hyperplane arrangement are very closely related to the topological properties of the arrangement ( see @xcite and the references therein ) . in this paper we show that the matroids associated to graphs are intimately related to its topology we consider graphs on surfaces . suppose that a graph @xmath0 is embedded into a surface @xmath1 in a cellular manner ; that is each connected component ( face ) of the complement to the graph is homeomorphic to a disc . then we can define a dual graph @xmath2 embedded into the same surface @xmath1 in a natural way . we associate _ cycle matroid _ @xmath3 with the graph @xmath0 , and with the dual graph @xmath2 we associate its _ bond matroid _ @xmath4 _ dual _ to @xmath5 . in the planar case , when @xmath1 is a sphere , @xmath4 is isomorphic to @xmath3 . thus the difference between isomorphism classes of @xmath3 and @xmath4 can be considered as a measure of non - planarity of the embedding and should reflect the topology of the pair @xmath6 . various combinatorial parameters of the matroids @xmath3 and @xmath4 can be assembled into the las vergnas polynomial of a matroid perspective @xmath7 introduced in @xcite . on the other hand , the topological parameters of the embedding of @xmath0 into @xmath1 can be assembled into the krushkal polynomial introduced in @xcite . in this paper we show that the las vergnas polynomial is a specialization of the krushkal polynomial . in the process , we relate many combinatorial parameters of the matroids @xmath3 and @xmath4 with the topological parameters of the embedding . in section [ s : lv ] we briefly review matroids and their combinatorial parameters , and introduce the las vergnas polynomial . we introduce the krushkal polynomial in section [ s : kp ] . the main theorem is formulated in section [ s : th ] , wherein we also obtain the duality property of the las vergnas polynomial as a consequence of our main theorem . we begin to prove the main theorem in section [ s : th ] and finish in section [ s : mct ] which consists of three lemmas relating the topological parameters of the embedding @xmath8 with the combinatorial parameters of its matroids . we conclude in section [ s : kp - br ] with a discussion on the relation of the krushkal polynomial with the bollobs - riordan polynomial of ribbon graphs . this work has been done as a part of the summer 2010 undergraduate research working group ` http://www.math.ohio-state.edu/~chmutov/wor-gr-su10/wor-gr.htm ` knots and graphs " at the ohio state university . we are grateful to all participants of the group for valuable discussions and to the osu honors program research fund for the student financial support . we thank the anonymous referee and editors for various suggestions improving the exposition of the paper . for additional background on matroids we refer to @xcite , in addition to whitney s classical paper @xcite . _ a matroid _ is a finite set @xmath9 with a _ rank _ function @xmath10 that assigns a number to a subset of @xmath9 and satisfies the following axioms : * _ the rank of an empty subset is zero . _ * _ for any subset @xmath11 and any element @xmath12 , @xmath13 _ * _ for any subset @xmath14 and two elements @xmath15,@xmath16 not in @xmath14 , if @xmath17 , then @xmath18 . _ there are two major examples of matroids we will be focusing on , although these examples do not exhaust the collection of matroids . that is to say , there are matroids that do not represent either of the following situations . the first example is the cycle matroid @xmath3 of a graph @xmath0 . the underlying set @xmath9 is the set of edges @xmath19 and the rank function is given by @xmath20 , where @xmath21 is the number of vertices of @xmath0 and @xmath22 is the number of connected components of the spanning subgraph of @xmath0 consisting of all the vertices of @xmath0 and edges of @xmath14 . the second example is a finite set of vectors in a vector space . we may think about them as column vectors of a matrix . the rank function is the dimension of the subspace spanned by the subset of vectors , or the rank of the corresponding submatrix . this example generalizes the first one when the matrix ( over @xmath23 ) is the incidence matrix of @xmath0 , i.e. the matrix of the simplicial boundary map from the set of edges of @xmath0 to the set of vertices of @xmath0 . a subset @xmath11 is called _ independent _ if @xmath24 . in the first example , the independent subsets are those subsets of edges which do not contain cycles . in the second , independent subsets correspond to linearly independent subsets of vectors . a _ base _ of a matroid is a maximal independent set . a subset @xmath11 is called a _ circuit _ if @xmath25 . in the first example , this new notion of a circuit and the traditional notion of a circuit in a graph match . in the second example , a circuit is a subset of vectors with precisely one linear relation between them . given any matroid @xmath9 , there is a dual matroid @xmath26 with the same underlying set and with the rank function given by @xmath27 . in particular @xmath28 . any base of @xmath26 is a complement to a base of @xmath9 . the dual matroid to the cycle matroid of a graph @xmath0 is called the _ bond matroid _ of @xmath0 : @xmath29 . the circuits of @xmath30 are the minimal edge cuts , also known as the _ bonds _ of @xmath0 . these are minimal collections of the edges of @xmath0 which , when removed from @xmath0 , increase the number of connected components . the whitney planarity criteria @xcite says that a graph @xmath0 is planar if and only if its bond matroid @xmath30 is the cycle matroid of some graph . in this case , it will be the cycle matroid of the dual graph , @xmath31 . for two matroids @xmath9 and @xmath32 , a bijection @xmath33 is called a _ matroid perspective _ if any circuit of @xmath9 is mapped to a union of circuits of @xmath32 . equivalently , @xmath34 where @xmath35 and @xmath36 are the rank functions of matroids @xmath9 an @xmath32 . the _ tutte polynomial _ of a matroid perspective @xmath33 is the polynomial in variables @xmath37 defined by @xmath38 where @xmath39 is the nullity in @xmath9 . * properties . * the usual tutte polynomial of matroids @xmath9 and @xmath32 can be recovered from the tutte polynomial of matroid perspective in the following ways : + @xmath40 for graphs @xmath0 and @xmath2 dually cellularly embedded in a surface @xmath1 , the natural map of the bond matroid of @xmath2 onto the cycle matroid of @xmath0 , @xmath41 , is a matroid perspective . formally this follows from a theorem of j. edmonds @xcite . informally , a circuit @xmath42 of @xmath43 is a minimal cut of the dual graph @xmath2 which separates the vertices of @xmath2 into two sets . the vertices of @xmath2 correspond to the faces of the original graph @xmath0 . thus , the minimal cut separates the faces of @xmath0 . we may think about cutting an edge of @xmath42 as cutting the surface @xmath1 along the corresponding edge of @xmath0 . therefore , in terms of the original graph @xmath0 , the circuit @xmath42 corresponds to a subset of edges of @xmath0 which separate the surface @xmath1 . topologically it represents a zero - homologous cycle which ( in general ) consists of several circuits . hence , a circuit of @xmath43 is mapped to a union of circuits of @xmath3 . we call the tutte polynomial @xmath44 of the matroid perspective @xmath41 the _ las vergnas polynomial _ of the graph @xmath0 on the surface @xmath1 , and denote it @xmath45 . one goal of this paper is to give a topological interpretation of various combinatorial ingredients of this polynomial . in this paper we assume that @xmath1 is orientable . [ ex1 ] let @xmath0 be a graph with one vertex and two loops embedded into a torus as shown . then @xmath2 is the similar graph . @xmath46[25pt]{\begin{picture}(100,15)(0,0 ) \put(0,0){\includegraphics[width=100pt]{rg - torus - d.eps } } \put(80,50){$g^*$ } \end{picture}}}\hspace{2 cm } { \raisebox{-20pt}[0pt][0pt]{\begin{picture}(100,15)(0,0 ) \put(0,0){\includegraphics[width=100pt]{rg - torus.eps } } \put(0,45){$g$ } \end{picture}}}\ ] ] in this case the bond matroid @xmath47 has rank 2 , and the cycle matroid @xmath48 has rank 0 . for any subset @xmath14 : @xmath49 , @xmath50 , and @xmath51 . we have @xmath52 v. krushkal discovered his polynomial while researching topological quantum field theories and the algebraic and combinatorial properties of models of statistical mechanics . this polynomial can be seen as a generalization of the tutte and bollobas - riordan polynomials @xcite . for a graph @xmath0 embedded into a ( not necessarily connected , but orientable ) surface @xmath1 , @xmath53 where * @xmath22 is the number of connected components of the spanning subgraph @xmath14 ; * the restriction of the embedding @xmath54 to @xmath14 induces a map on the first homology groups and we define @xmath55 * @xmath56 is equal to twice the genus of a regular neighborhood of the spanning subgraph @xmath14 in @xmath1 ( the neighborhood is a surface with boundary , and its genus is defined as the genus of the closed surface obtained by attaching a disk to each boundary circle ) ; * @xmath57 is equal to twice the genus of the surface obtained by removing a regular neighborhood of @xmath14 from @xmath1 . v. krushkal indicates in his paper @xcite that the parameters @xmath56 and @xmath57 have the following interpretation in terms of the symplectic bilinear form on the vector space @xmath58 given by the intersection number . for a given spanning subgraph @xmath14 , let @xmath59 be its image in the homology group : @xmath60 for the subspace @xmath59 we can define its orthogonal complement @xmath61 in @xmath58 with respect to the symplectic intersection form . then @xmath62 [ ex2 ] continuing with example [ ex1 ] , for the graph @xmath0 on the torus @xmath1 , the map @xmath63 is not degenerate . therefore , @xmath64 for any subset @xmath14 . also , @xmath65 for any @xmath14 . if @xmath66 , then its regular neighborhood has genus 0 . hence , for such @xmath14 , @xmath67 , while @xmath68 . similarly , if @xmath69 , then the regular neighborhood of its complement also has genus 0 . so , for such @xmath14 , @xmath70 , while @xmath71 . combining all this we get @xmath72 suppose @xmath0 is cellularly embedded in an orientable surface @xmath1 of genus @xmath73 . then @xmath74 the summands corresponding to each subgraph @xmath14 of @xmath0 will be shown to be equal . applying the substitution in we arrive at summands of the form @xmath75 the cycle matroid rank of a graph is given by @xmath76 and that of a subgraph @xmath77 by @xmath78 . so @xmath79 and hence the powers of the @xmath80 factor coincide . it remains to prove the equality between the exponents of @xmath81 and @xmath16 : * @xmath82 * @xmath83 these will be proved separately in the next section . @xmath84 this follows from krushkal s formula @xcite : @xmath85 for cellular embeddings @xmath8 . in this section we relate the topological parameters of the embedding @xmath8 with the combinatorial parameters of the matroid perspective @xmath41 . [ l:4 - 1 ] @xmath82 . let @xmath86 . the rank of a dual matroid can be defined in terms of the rank of a matroid by @xmath87 , where @xmath14 is a subset of edges of @xmath2 that can be naturally identified with the corresponding subset of edges of @xmath0 . since the nullity is defined as the number of edges minus the rank we have that @xmath88 , where @xmath89 is the spanning subgraph of @xmath2 consisting of the edges not in @xmath14 and all vertices of @xmath2 . thus , because @xmath90 is the cycle matroid of the graph @xmath2 , we have @xmath91 obviously , @xmath92 and is equal to the number of connected component @xmath93 of @xmath1 . now we consider @xmath14 as a spanning subgraph of @xmath0 . we want to remove its regular neighborhood from the surface @xmath1 and count the number of connected components @xmath94 . first we remove small discs around all vertices of @xmath14 , i.e. all vertices of @xmath0 . these are exactly the faces of @xmath2 . so we are left with a regular neighborhood of @xmath2 in @xmath1 . secondly , removing neighborhoods of the edges of @xmath95 from @xmath1 will give us the same surface , topologically , as deleting the corresponding neighborhoods of the edges of @xmath96 , because these edges are transverse to each other . in other words @xmath97 is the number of components of @xmath98 . hence @xmath99 where @xmath14 is regarded as a spanning subgraph of @xmath0 . let us turn our attention to the number that we wish to show @xmath100 to be equal to . denote by @xmath101 the linear map induced by the composition of embeddings @xmath102 . we have @xmath103 . the topological pair @xmath104 gives us a long exact sequence of homology groups @xmath105 @xmath106 is trivial as @xmath14 is one dimensional , @xmath107 has dimension equal to the number of components of @xmath1 . and @xmath108 has dimension equal to the number of components of @xmath109 . so if we turn our attention to the short exact sequence @xmath110 wherein @xmath111 , we see that @xmath112 . thus the @xmath81 powers in the main theorem coincide . in fact , the argument with the long exact sequence was used by v. krushkal ( * ? ? ? * end of proof of theorem 3.1 ) where he essentially proved that @xmath113 using a slightly different terminology . however , the relation of this parameter to the matroidal @xmath100 was not addressed there . [ l:4 - 2 ] @xmath114 . @xmath115 , where @xmath116 is a spanning forest of @xmath0 . the bond rank of a graph @xmath2 is the maximal number of edges that one can delete from it without increasing the number of connected components , i.e. all edges but a spanning forest , @xmath117 , of the graph @xmath2 . so @xmath118 and @xmath119 . but the number of edges in a spanning forest of @xmath0 is equal to the number of vertices minus the number of components . similarly , @xmath120 , the number of edges in a spanning forest of @xmath2 , is the number of faces , @xmath121 , of @xmath0 minus the number of components . so @xmath122 . [ l:4 - 3 ] @xmath123 . consider @xmath124 . for either matroid , @xmath9 or @xmath32 , rank is equal to the number of edges minus nullity . so @xmath125 . we have seen in lemma [ l:4 - 1 ] that @xmath126 . formula ( 2.5 ) in krushkal s paper @xcite tells us that this nullity @xmath127 so we have @xmath124 equaling @xmath128 lemma [ l:4 - 2 ] is a particular case of lemma [ l:4 - 3 ] where we take @xmath129 . indeed , in this case we have @xmath130 and @xmath131 . lemmas [ l:4 - 2 ] and [ l:4 - 3 ] together imply @xmath132 thus the @xmath16 powers in the main theorem coincide . this completes the proof of the main theorem . a cellular embedding @xmath8 may be studied in terms of a _ ribbon graph _ which represents a regular neighborhood @xmath133 of @xmath0 in @xmath1 . working backwards , starting with a ribbon graph @xmath133 , we can construct a surface @xmath1 by capping all boundary components of @xmath133 by discs . then the core graph @xmath0 of @xmath133 , obtained by contracting all edge - ribbons to their central lines and all vertex - discs to their central points , can be cellularly embedded in @xmath1 . for ribbon graphs , we have the bollobs - riordan polynomial @xcite defined as @xmath134 where @xmath135 is the number of boundary components of the spanning ribbon subgraph @xmath136 . note that the exponent @xmath137 is equal to @xmath138 for oriented ribbon graphs . it was proved in ( * ? ? ? * theorem 2 ) that @xmath140 is universal in the class of polynomials satisfying the contraction / deletion property . the krushkal polynomial also satisfies _ a _ contraction / deletion property ( * ? ? ? * lemma 2.1 ) . based on that v. krushkal wrote that @xmath140 and @xmath141 carry equivalent information . however this is not the case as the contraction / deletion properties for @xmath140 and for @xmath141 are not quite the same . the problem arises when deletion of an edge of a ribbon graph changes its genus . the genus might decrease by 1 with removal of an edge . for example , if we delete a loop @xmath142 from the ribbon graph @xmath133 corresponding to @xmath0 from example [ ex1 ] , then the resulting graph with a single loop will have genus zero . so , while in the bollobs - riordan approach it is considered as a graph embedded into a sphere , in the krushkal approach it is still embedded into the torus . we can not apply the substitution to that graph since its embedding on the torus is no longer cellular . thus the krushkal polynomial does not satisfy the contraction / deletion property in the sense of bollobs and riordan . [ ex3 ] this is an example of a calculation of the three polynomials . here @xmath0 is a graph on torus with two vertices and three edges @xmath143 , @xmath144 , and @xmath42 . its dual @xmath2 has one vertex and three loops . the ribbon graph corresponding to @xmath0 is denoted @xmath133 . we use the same symbols @xmath143 , @xmath144 , @xmath42 to denote the corresponding edges in all three graphs . @xmath145[35pt]{\begin{picture}(100,15)(0,0 ) \put(0,0){\includegraphics[width=100pt]{gd - torus.eps } } \put(65,55){$a$}\put(35,34){$b$ } \put(58,6){$c$ } \end{picture}}}\ \hspace{1 cm } g={\raisebox{-25pt}[35pt][35pt]{\begin{picture}(100,15)(0,0 ) \put(0,0){\includegraphics[width=100pt]{g - torus.eps } } \put(70,36){$a$}\put(35,-6){$b$ } \put(55,4){$c$ } \end{picture}}}\ \ ] ] @xmath146[0pt]{\begin{picture}(80,15)(0,0 ) \put(0,0){\includegraphics[width=80pt]{cg - rg.eps } } \put(67,45){$a$}\put(15,-3){$b$ } \put(55,-4){$c$ } \end{picture}}}\ ] ] the matroid @xmath48 is of rank 1 , and for any nonempty subset @xmath14 , @xmath147 . the cycle matroid @xmath5 of the dual graph is of rank zero because @xmath2 has only loops . so its dual @xmath47 has rank 3 , all subsets @xmath14 are independent and @xmath49 . the next table shows the value of various parameters and contributions of all eight subsets @xmath148 to the three polynomials . @xmath149 \hline\hline & c(h ) & 2&1&1&1 & 1&1&1&1 \\ \hhline{~--------- } & k(h ) & 0&0&0&0 & 0&0&0&0 \\ \hhline{~--------- } & s(h ) & 0&0&0&0 & 0&0&0&2 \\ \hhline{~--------- } & s^\perp(h ) & 2&2&2&0 & 2&0&0&0 \\ \hhline{~--------- } \makebox(20,0){\raisebox{60pt}{\rotatebox{90}{krushkal } } } & p_{g,\s } & xb&b&b&1 & b&1&1&a \\ \hline\hline & r_m(h ) & 0&1&1&2 & 1&2&2&3 \\ \hhline{~--------- } & r_{m'}(h ) & 0&1&1&1 & 1&1&1&1 \\ \hhline{~--------- } & n_m(h ) & 0&0&0&0 & 0&0&0&0 \\ \hhline{~--------- } \makebox(20,0){\raisebox{50pt}{\rotatebox{90}{las vergnas } } } & lv_{g,\s } & \!\!\!(x-1)z^2&z^2&z^2&z & z^2&z&z&1 \\ \hline\hline & c(\ch ) & 2&1&1&1 & 1&1&1&1 \\ \hhline{~--------- } & n(\ch ) & 0&0&0&1 & 0&1&1&2 \\ \hhline{~--------- } & bc(\ch ) & 2&1&1&2 & 1&2&2&1 \\ \hhline{~--------- } \makebox(20,0){\raisebox{50pt}{\rotatebox{90 } { $ \genfrac{}{}{0pt}{0}{\mbox{bollob\'as}}{\mbox{riordan}}$ } } } & br_\cg & ( x-1)&1&1&y & 1&y&y&y^2z^2 \\ \end{array}\ ] ] thus @xmath150 @xmath151 one can readily confirm the relations and from here . now if we contract the edge @xmath42 , the graph @xmath152 still will be cellularly embedded into the same torus @xmath1 , and its regular neighborhood coincides with the ribbon graph @xmath153 . examples [ ex1 ] and [ ex2 ] and the right part of the table above give the following polynomials : @xmath154 meanwhile if we delete the edge @xmath42 , then @xmath155 but the graph @xmath156 is not cellularly embedded into the torus @xmath1 any more . thus the las vergnas and the bollobs - riordan polynomials are not defined for it . its regular neighborhood gives the ribbon graph @xmath157 which , after capping the discs to its two boundary components , results in the sphere @xmath158 . thus the graph @xmath156 embeds cellularly into the the sphere @xmath158 . for this embedding we have @xmath159 therefore @xmath160 but @xmath161 currently , according to relations and , the krushkal polynomial is the most general polynomial of graphs on surfaces and so it clearly deserves further research . also , because the two relations look quite different , the las vergnas and the bollobs - riordan polynomials seem to be independent . very recently the krushkal polynomial was generalized to higher dimensional simplicial complexes @xcite . it is related to a matroid on the sets of simplices of the middle dimension for a triangulation of an even dimensional sphere , where a subset of simplices is indpendent if and only if they are mapped into the the linear independent chains by the simplicial boundary map . it turns out that the dual matroid corresponds to the dual triangulation and the tutte polynomial of this matroid corresponds to the higher dimensional " tutte polynomial of the simplicial complexes ( see @xcite for details ) . * note added in proof . * after the paper has been submitted to the journal jonathan michel found the following example of two ribbon graphs with the same las vergnas polynomials but different krushkal polynomials . @xmath162[35pt]{\begin{picture}(60,15)(0,0 ) \put(0,0){\includegraphics[width=60pt]{g1-jm.eps } } \end{picture}}}\ \hspace{1 cm } \cg_2={\raisebox{-25pt}[35pt][35pt]{\begin{picture}(60,15)(0,0 ) \put(0,0){\includegraphics[width=60pt]{g2-jm.eps } } \end{picture}}}\ \ ] ] let @xmath163 ( resp . @xmath164 ) be a surface obtained from @xmath165 ( resp . @xmath166 ) by gluing a disc to its boundary component , and @xmath167 ( resp . @xmath168 ) be the corresponding core graph . then @xmath169 . but @xmath170 and @xmath171 jonathan michel also found two different ribbon graphs with the same krushkal polynomials . @xmath172[35pt]{\begin{picture}(100,15)(0,0 ) \put(0,0){\includegraphics[width=100pt]{g3-jm.eps } } \end{picture}}}\ \hspace{1 cm } \cg_4={\raisebox{-25pt}[0pt][0pt]{\begin{picture}(100,15)(0,0 ) \put(0,0){\includegraphics[width=100pt]{g4-jm.eps } } \end{picture}}}\ \ ] ] for the corresponding core graphs @xmath173 and @xmath174 we have @xmath175 v. krushkal , _ graphs , links , and duality on surfaces _ , combinatorics , probability and computing * 20 * ( 2011 ) 267287 . v. krushkal , d. renardy , _ a polynomial invariant and duality for triangulations_. preprint ` arxiv:1012.1310v1 [ math.co ] ` . m. las vergnas , _ the tutte polynomial of a morphism of matroids . i. set - pointed matroids and matroid perspectives _ , symposium la mmoire de franois jaeger ( grenoble , 1998 ) . fourier ( grenoble ) * 49*(3 ) ( 1999 ) 973-1015 .

for a graph embedded into a surface , we relate many combinatorial parameters of the cycle matroid of the graph and the bond matroid of the dual graph with the topological parameters of the embedding . this gives an expression of the polynomial , defined by m. las vergnas in a combinatorial way using matroids as a specialization of the krushkal polynomial , defined using the symplectic structure in the first homology group of the surface . 05c10 , 05c31 , 57m15 , 57m25 , 57m27 graphs on surfaces , ribbon graphs , matroids , krushkal polynomial , las vergnas polynomial , bollobs - riordan polynomial

conventionally , in order to construct the equilibrium configurations of static neutron stars the equations of hydrostatic equilibrium derived by tolman - oppenheimer - volkoff ( tov ) @xcite are widely used . in connection with this , it has been recently revealed in refs . @xcite that the tov equations are modified once all fundamental interactions are taken into due account . it has been proposed that the einstein - maxwell system of equations coupled with the general relativistic thomas - fermi equations of equilibrium have to be used instead . this set of equations is termed as the einstein - maxwell - thomas - fermi ( emtf ) system of equations . although in the tov method the condition of local charge neutrality ( lcn ) , @xmath10 is imposed ( see e.g. @xcite ) , the emtf method requires the less rigorous condition of global charge neutrality ( gcn ) as follows @xmath11 d^3r = 0,\ ] ] where @xmath12 is the electric charge density , @xmath13 is the fundamental electric charge , @xmath14 and @xmath15 are the proton and electron number densities , respectively . the integration is performed on the entire volume of the system . the lagrangian density accounting for the strong , weak , electromagnetic and gravitational interactions consists of the free - fields terms such as the gravitational @xmath16 , the electromagnetic @xmath17 , and the three mesonic fields @xmath18 , @xmath19 , @xmath20 , the three fermion species ( electrons , protons and neutrons ) term @xmath21 and the interacting part in the minimal coupling assumption , @xmath22 given as in refs . @xcite : @xmath23 where unless otherwise specified . ] @xmath24 the inclusion of the strong interactions between the nucleons is made through the @xmath25-@xmath26-@xmath27 nuclear model following ref . consequently , @xmath28 , @xmath29 , @xmath30 are the field strength tensors for the @xmath31 , @xmath27 and @xmath32 fields respectively , @xmath33 stands for covariant derivative and @xmath34 is the ricci scalar . the lorentz gauge is adopted for the fields @xmath35 , @xmath36 , and @xmath37 . the self - interaction scalar field potential is @xmath38 , @xmath39 is the nucleon isospin doublet , @xmath40 is the electronic singlet , @xmath41 stands for the mass of each particle - species and @xmath42 , where @xmath43 are the dirac spin connections . the conserved currents are given as @xmath44 , @xmath45 , @xmath46 , and @xmath47 , where @xmath48 is the particle isospin . in this chapter we adopt the nl3 parameter set @xcite used in ref . @xcite with @xmath49 mev , @xmath50 mev , @xmath51 mev , @xmath52 , @xmath53 , @xmath54 , plus two constants that give the strength of the self - scalar interactions , @xmath55 @xmath56 and @xmath57 . in the core - crust transition layer , whereas the bottom panel shows a specific example of a density profile inside a neutron star . in this plot we have used for the globally neutral case a density at the edge of the crust equal to the neutron drip density , @xmath58 g @xmath59.,title="fig : " ] [ fig : model ] g @xmath59 has been used . gcn and lcn stand for global and local charge neutrality cases , respectively ] thus , the system of the emtf equations @xcite is derived from the equations of motion of the above lagrangian . the solution of the emtf coupled differential equations gives a novel structure of the neutron star , as shown in fig [ fig : model ] : a positively charged core at supranuclear densities , @xmath60 g @xmath59 , surrounded by an electron distribution of thickness @xmath61 , which is negatively charged and a neutral ordinary crust at lower densities @xmath62 . the condition of the thermodynamic equilibrium is given by the constancy of the particle klein potentials @xcite extended to account for electrostatic and strong fields @xcite @xmath63={\rm constant},\ ] ] where the subscript @xmath64 stands for each kind of particle , @xmath65 is the particle chemical potential , @xmath66 is the particle electric charge , @xmath67 is the time component of the fluid four - velocity which satisfies @xmath68 and @xmath69 is the t t component of the spherically symmetric metric . for the static case we have only the time components of the vector fields , @xmath70 , @xmath71 , @xmath72 . @xmath73 the constancy of the klein potentials ( [ eq : klein ] ) leads to a discontinuity in the density at the core - crust transition and , respectively , this generates an overcritical electric field @xmath74 , where @xmath75 volt @xmath76 , in the core - crust boundary interface . the klein condition ( [ eq : klein ] ) is necessary to satisfy the requirement of thermodynamical equilibrium , together with the tolman condition ( constancy of the gravitationally red - shifted temperature ) @xcite , if finite temperatures are included @xcite . particularly , the continuity of the electron klein potential leads to a decrease of the electron chemical potential @xmath77 and density at the core - crust boundary interface . they attain values @xmath78 and @xmath79 at the basis of the crust , where global charge neutrality is achieved . as it has been shown in refs . @xcite , that the solution of the emtf equations along with the constancy of the klein potentials yield a more compact neutron star with a less massive and thiner crust . correspondingly , this results in a new mass - radius relation which prominently differs from the one given by the solution of the tov equations with local charge neutrality ; see fig . [ fig : mrstat ] . in this chapter the extension of the previous results obtained in refs . @xcite are considered . to this end the hartle formalism @xcite is utilized to solve the einstein equations accurately up to second order terms in the angular velocity of the star , @xmath1 ( see section [ sec:2 ] ) . for the rotating case , the klein thermodynamic equilibrium condition has the same form as eq . ( [ eq : klein ] ) , but the fluid inside the star now moves with a four - velocity of a uniformly rotating body , @xmath80 , with ( see @xcite , for details ) @xmath81 where @xmath82 is the azimuthal angular coordinate and the metric is axially symmetric independent of @xmath82 . the components of the metric tensor @xmath83 are now given by eq . ( [ eq : rotmetric ] ) below . it is then evident that in a frame comoving with the rotating star , @xmath67 , and the klein thermodynamic equilibrium condition remains the same as eq . ( [ eq : klein ] ) , as expected . this chapter is organized as follows : in section [ sec:2 ] we review the hartle formalism and consider both interior and exterior solutions . in section [ sec:3 ] the stability of uniformly rotating neutron stars are explored taking into account the keplerian mass - shedding limit and the secular axisymmetric instability . in section [ sec:4 ] the structure of uniformly rotating neutron stars is investigated . we compute there the mass @xmath84 , polar @xmath4 and equatorial @xmath5 radii , and angular momentum @xmath8 , as a function of the central density and the angular velocity @xmath1 of stable neutron stars both in the globally and locally neutral cases . based on the criteria of equilibrium we calculate the maximum stable neutron star mass . in section [ sec:5 ] we construct the new neutron star mass - radius relation . in section [ sec:6 ] we calculate the moment of inertia as a function of the central density and total mass of the neutron star . the eccentricity @xmath6 , the rotational to gravitational energy ratio @xmath85 , and quadrupole moment @xmath9 are shown in section [ sec:7 ] . the observational constraints on the mass - radius relation are discussed in section [ sec:8 ] . we finally summarize our results in section [ sec:9 ] . in his original article , hartle ( 1967 ) @xcite derived the equilibrium equations of slowly rotating relativistic stars . the solutions of the einstein equations have been obtained through a perturbation method , expanding the metric functions up to the second order terms in the angular velocity @xmath1 . under this assumption the structure of compact objects can be approximately described by the total mass @xmath84 , angular momentum @xmath8 and quadrupole moment @xmath9 . the slow rotation regime implies that the perturbations owing to the rotation are relatively small with respect to the known non - rotating geometry . the interior solution is derived by solving numerically a system of ordinary differential equations for the perturbation functions . the exterior solution for the vacuum surrounding the star , can be written analytically in terms of @xmath84 , @xmath8 , and @xmath9 @xcite . the numerical values for all the physical quantities are derived by matching the interior and the exterior solution on the surface of the star . the spacetime metric for the rotating configuration up to the second order of @xmath1 is given by @xcite @xmath86dr^2 \nonumber\\ & & - r^2\left(1 + 2k\right)\left[d\theta^2+\sin^2\theta\left(d\phi-\omega dt\right)^2\right]+o(\omega^3 ) \ , , \ ] ] where @xmath87 , @xmath88 , and @xmath89 are the metric functions and mass profiles of the corresponding seed static star with the same central density as the rotating one ; see eq . ( [ eq : metric1 ] ) . the functions @xmath90 , @xmath91 , @xmath92 and the fluid angular velocity in the local inertial frame , @xmath93 , have to be calculated from the einstein equations . expanding up to the second order the metric in spherical harmonics we have @xmath94 where @xmath95 is the legendre polynomial of second order . because the metric does not change under transformations of the type @xmath96 , we can assume @xmath97 . the functions @xmath90 , @xmath91 , @xmath92 have analytic form in the exterior ( vacuum ) spacetime and they are shown in the following section . the mass , angular momentum , and quadrupole moment are computed from the matching condition between the interior and exterior metrics . for rotating configurations the angular momentum is the easiest quantity to compute . to this end we consider only @xmath98 component of the einstein equations . by introducing the angular velocity of the fluid relative to the local inertial frame , @xmath99 one can show from the einstein equations at first order in @xmath1 that @xmath100 satisfies the differential equation @xmath101 where @xmath102 with @xmath103 and @xmath104 the metric functions of the seed non - rotating solution ( [ eq : metric1 ] ) . from the matching conditions , the angular momentum of the star is given by @xmath105 so the angular velocity @xmath1 is related to the angular momentum as @xmath106 the total mass of the rotating star , @xmath84 , is given by @xmath107 where @xmath108 is the contribution to the mass owing to rotation . the second order functions @xmath109 ( the mass perturbation function ) and @xmath110 ( the pressure perturbation function ) are computed from the solution of the differential equation @xmath111 where @xmath112 and @xmath113 are the total energy - density and pressure . turning to the quadrupole moment of the neutron star , it is given by @xmath114 where @xmath115 is a constant of integration . this constant is fixed from the matching of the second order function @xmath116 obtained in the interior from @xmath117\;,\\ \frac{d h_2}{dr}&=h_2\bigg\{-\frac{d\nu}{dr}+\frac{r}{r-2 m_0}\left(\frac{d\nu}{dr}\right)^{-1}\bigg[8\pi({\cal e}+p)-\frac{4 m_0}{r^3}\bigg ] \bigg\}-\frac{4 ( k_2+h_2)}{r ( r-2 m_0)}\left(\frac{d\nu}{dr}\right)^{-1}\nonumber\\ & + \frac{1}{6}\bigg[\frac{r}{2}\frac{d\nu}{dr}-\frac{1}{r-2 m_0}\left(\frac{d\nu}{dr}\right)^{-1}\bigg]r^3j^2\left(\frac{d\bar{\omega}}{dr}\right)^2\nonumber\\&-\frac{1}{3}\bigg[\frac{r}{2}\frac{d\nu}{dr}+\frac{1}{r-2 m_0}\left(\frac{d\nu}{dr}\right)^{-1}\bigg]r^2 \bar{\omega}^2\frac{dj^2}{dr}\;,\end{aligned}\ ] ] with its exterior counterpart ( see @xcite ) . it is worth emphasizing that the influence of the induced magnetic field owing to the rotation of the charged core of the neutron star in the globally neutral case is negligible @xcite . in fact , for a rotating neutron star of period @xmath118 ms and radius @xmath119 km , the radial component of the magnetic field @xmath120 in the core interior reaches its maximum at the poles with a value @xmath121 , where @xmath122 g is the critical magnetic field for vacuum polarization . the angular component of the magnetic field @xmath123 , instead , has its maximum value at the equator and , as for the radial component , it is very low in the interior of the neutron star core , i.e. @xmath124 . in the case of a sharp core - crust transition as the one studied by @xcite and shown in fig . [ fig : model ] , this component will grow in the transition layer to values of the order of @xmath125 @xcite . however , since we are here interested in the macroscopic properties of the neutron star , we can ignore at first approximation the presence of electromagnetic fields in the macroscopic regions where they are indeed very small , and safely apply the original hartle formulation without any generalization . in this subsection we consider the exterior hartle solution though in the literature it is widely known as the hartle - thorne solution . one can write the line element given by eq . ( [ eq : rotmetric ] ) in an analytic closed - form outside the source as function of the total mass @xmath84 , angular momentum @xmath8 , and quadrupole moment @xmath9 of the rotating star . the angular momentum @xmath8 along with the angular velocity of local inertial frames @xmath126 , proportional to @xmath1 , and the functions @xmath127 , @xmath116 , @xmath109 , @xmath128 , @xmath129 , proportional to @xmath130 , are derived from the einstein equations ( for more details see @xcite ) . following this prescriptions the eq . [ eq : rotmetric ] becomes : @xmath131dt^2\nonumber\\ & + \frac{4j}{r}\sin^2\theta dt d\phi-\left(1-\frac{2 { m}}{r}\right)^{-1}\times\bigg[1 - 2\left(k_1-\frac{6 j^{2}}{r^4}\right)p_2(\cos\theta ) \nonumber \\ & -2\left(1-\frac{2 { m}}{r}\right)^{-1}\frac{j^{2}}{r^4}\bigg]dr^2-r^2[1 - 2k_2p_2(\cos\theta)](d\theta^2+\sin^2\theta d\phi^2),\end{aligned}\ ] ] where @xmath132 and @xmath133 \ ; , \\ q_2 ^ 2(x)&=(x^2 - 1)\left[\frac{3}{2}\ln\left(\frac{x+1}{x-1}\right)-\frac{3x^3 - 5x}{(x^2 - 1)^2}\right ] \;,\end{aligned}\ ] ] are the associated legendre functions of the second kind , being @xmath134 the legendre polynomial , and @xmath135 . this form of the metric is known in the literature as the hartle - thorne metric . to obtain the exact numerical values of @xmath84 , @xmath8 and @xmath9 , the exterior and interior line elements have to be matched at the surface of the star . it is worth noticing that in the terms involving @xmath136 and @xmath9 , the total mass @xmath84 can be directly substituted by @xmath137 since @xmath108 is already a second order term in the angular velocity . in a sequence of increasing central density in the @xmath84-@xmath0 curve , @xmath138 , the maximum mass of a static neutron star is determined as the first maximum of such a curve , namely the point where @xmath139/@xmath140 . this derivative establishes the axisymmetric secular instability point , and if the perturbation obeys the same equation of state ( eos ) as the equilibrium configuration , it coincides also with the dynamical instability point ( see e.g. ref . @xcite ) . in the rotating case , the situation becomes more complicated and in order to find the axisymmetric dynamical instability points , the perturbed solutions with zero frequency modes ( the so - called neutral frequency line ) have to be calculated . friedman et al . ( 1988 ) @xcite however , following the works of sorkin ( 1981 , 1982 ) @xcite , described a turning - point method to obtain the points at which secular instability is reached by uniformly rotating stars . in a constant angular momentum sequence , the turning point is located in the maximum of the mass - central density relation , namely the onset of secular axisymmetric instability is given by @xmath141_{j=\rm constant}=0 \ ; , \ ] ] and once the secular instability sets in , the star evolves quasi - stationarily until it reaches a point of dynamical instability where gravitational collapse sets in ( see e.g. @xcite ) . the above equation determines an upper limit for the mass at a given angular momentum @xmath8 for a uniformly rotating star , however this criterion is a sufficient but not necessary condition for the instability . this means that all the configurations with the given angular momentum @xmath8 on the right side of the turning point defined by eq . ( [ eq : turningpoint ] ) are secularly unstable , but it does not imply that the configurations on the left side of it are stable . an example of dynamically unstable configurations on the left side of the turning - point limiting boundary in neutron stars was recently shown in ref . @xcite , for a specific eos . in order to investigate the secular instability of uniformly rotating stars one should select fixed values for the angular momentum . then construct mass - central density or mass - radius relations . from here one has to calculate the maximum mass and that will be the turning point for a given angular momentum . for different angular momentum their will be different maximum masses . by joining all the turning points together one obtains axisymmetric secular instability line ( boundary ) . this boundary is essential for the construction of the stability region for uniformly rotating neutron stars ( see next sections and figures ) . the maximum velocity for a test particle to remain in equilibrium on the equator of a star , kept bound by the balance between gravitational and centrifugal force , is the keplerian velocity of a free particle computed at the same location . as shown , for instance in @xcite , a star rotating at keplerian rate becomes unstable due to the loss of mass from its surface . the mass shedding limiting angular velocity of a rotating star is the keplerian angular velocity evaluated at the equator , @xmath142 , i.e. @xmath143 . friedman ( 1986 ) @xcite introduced a method to obtain the maximum possible angular velocity of the star before reaching the mass - shedding limit ; however @xcite and @xcite , showed a simpler way to compute the keplerian angular velocity of a rotating star . they showed that the mass - shedding angular velocity , @xmath144 , can be computed as the orbital angular velocity of a test particle in the external field of the star and corotating with it on its equatorial plane at the distance @xmath142 . it is possible to obtain the analytical expression for the angular velocity @xmath1 given by eq . ( [ eq : omegakep ] ) with respect to an observer at infinity , taking into account the parameterization of the four - velocity @xmath145 of a test particle on a circular orbit in equatorial plane of axisymmetric stationary spacetime , regarding as parameter the angular velocity @xmath1 itself : @xmath146 \ ; , \ ] ] where @xmath147 is a normalization factor such that @xmath148 . normalizing and applying the geodesics conditions we get the following expressions for @xmath147 and @xmath149 @xmath150 thus , the solution of the system of eq . ( [ eight ] ) can be written as @xmath151 where @xmath152 stands for co - rotating / counter - rotating orbits , @xmath153 and @xmath154 are the angular and time components of the four - velocity respectively , and a colon stands for partial derivative with respect to the corresponding coordinate . to determine the mass shedding angular velocity ( the keplerian angular velocity ) of the neutron stars , we need to consider only the co - rotating orbit , so from here and thereafter we take into account only the plus sign in eq . ( [ eight ] ) and we write @xmath155 . for the hartle external solution given by eq . ( [ ht1 ] ) we obtain @xmath156 \ ; , \ ] ] where @xmath157 and @xmath158 are the dimensionless angular momentum and quadrupole moment . the analytical expressions of the functions @xmath159 are given by @xmath160 the maximum angular velocity for a rotating star at the mass - shedding limit is the keplerian angular velocity evaluated at the equator ( @xmath142 ) , i.e. @xmath161 in the static case i.e. when @xmath162 hence @xmath163 and @xmath164 we have the well - known schwarzschild solution and the orbital angular velocity for a test particle @xmath165 on the surface ( @xmath166 ) of the neutron star is given by @xmath167 besides the above stability requirements , one should check if the neutron star is gravitationally bound . in the non - rotating case , the binding energy of the star can be computed as @xmath168 where @xmath169 is the rest - mass of the star , @xmath170 is the rest - mass per baryon , and @xmath171 is the total number of baryons inside the star . so the non - rotating star is considered bound if @xmath172 . in the slow rotation approximation the total binding energy is given by ref . @xcite @xmath173 where @xmath174- \frac{d u}{dp}\left(1-\frac{2 m}{r}\right)^{-1/2}\bigg\}\nonumber \\ & + ( { \cal e}-u)\left(1-\frac{2m}{r}\right)^{-3/2}\bigg[\frac{m_0}{r}+\frac{1}{3}j^2r^2\bar{\omega}^2\bigg]\nonumber \\&-\frac{1}{4\pi r^2}\left [ \frac{1}{12}j^2 r^4 \left(\frac{d\bar{\omega}}{dr}\right)^2-\frac{1}{3}\frac{dj^2}{dr}r^3 \bar{\omega}^2 \right]\;,\end{aligned}\ ] ] where @xmath175 is the internal energy of the star , with @xmath176 the baryon number density . we show now the results of the integration of the hartle equations for the globally and locally charge neutrality neutron stars ; see e.g. fig . [ fig : model ] . following refs . @xcite , we adopt , as an example , globally neutral neutron stars with a density at the edge of the crust equal to the neutron drip density , @xmath177 g @xmath59 . in fig . [ fig : mtotvsrhog ] we show the mass - central density curve for globally neutral neutron stars in the region close to the axisymmetric stability boundaries . particularly , we construct some @xmath8=constant sequences to show that indeed along each of these curves there exist a maximum mass ( turning point ) . the line joining all the turning points determines the secular instability limit . in fig . [ fig : mtotvsrhog ] the axisymmetric stable zone is on the left side of the instability line . g and the density is normalized to nuclear density @xmath178 g @xmath59 . the solid curve represents the configuration at the keplerian mass - shedding sequence , the dashed curve represents the static sequence , the dotted curves represent the @xmath8=constant sequences . the doted - dashed line joins all the turning points of the @xmath8=constant sequences , so it determines the axisymmetric secular instability boundary . ] -constant sequences . the doted - dashed line is the secular instability boundary . ] clearly we can transform the mass - central density relation in a mass - radius relation . in fig . [ fig : mtotvsreqg ] we show the mass versus the equatorial radius of the neutron star that correspond to the range of densities of fig . [ fig : mtotvsrhog ] . in this plot the stable zone is on the right side of the instability line . we can construct a fitting curve joining the turning points of the @xmath8=constant sequences line which determines the axisymmetric secular instability boundary . defining @xmath179 as the maximum stable mass of the non - rotating neutron star constructed with the same eos , we find that for globally neutral configurations the instability line is well fitted by the function @xmath180 where @xmath181 , and @xmath182 . the turning points of locally neutral configurations in the mass - central density plane are shown in fig . [ fig : mtotvsrhol ] . the corresponding mass - equatorial radius plane is plotted in fig . [ fig : mtotvsreql ] . g and the density is normalized to nuclear density @xmath178 g @xmath59 . the solid curve represents the configuration at keplerian mass - shedding sequence , the dashed curve represents the static sequence , the dotted curves represent the @xmath8=constant sequences . the doted - dashed line determines the axisymmetric secular instability boundary . ] = constant sequences . the doted - dashed line is the axisymmetric secular instability boundary . ] for locally neutral neutron stars , the secular instability line is fitted by @xmath183 where @xmath184 , and @xmath185 . we turn now to analyze in detail the behavior of the different properties of the neutron star along the keplerian mass - shedding sequence . the total mass of the rotating star is computed from eq . ( [ eq : mrot ] ) . in fig . [ fig : mtotvsf ] the total mass of the neutron star is shown as a function of the rotation frequency for the keplerian sequence . it is clear that for a given mass , the rotational frequency is higher for a globally neutral neutron star with respect to the locally neutral one . the configuration of maximum mass , @xmath186 , is obtained along the keplerian sequence , and it is found before the secular instability line crosses the keplerian curve . thus , the maximum mass configuration is secularly stable . this implies that the configuration with maximum rotation frequency , @xmath187 , is located beyond the maximum mass point , specifically at the crossing point between the secular instability and the keplerian mass - shedding sequence . the results are summarized in table [ tab : mrfmax ] and shown in fig . [ fig : mmaxrho ] . = constant sequences , the dotted - dashed line is axisymmetric secular instability boundary . the thin line joins @xmath188 with @xmath189 . ] in fig . [ fig : mmaxrho ] details near the maximum masses are illustrated . here we focus on the definition of maximum rotating mass @xmath188 , maximum static mass , maximum angular momentum @xmath190 and maximum angular velocity ( minimum rotation period ) @xmath191 . note , that @xmath192 is determined along the turning points of @xmath8=constant sequences ( axisymmetric secular instability line ) what is consistent with the results of stergioulas and friedman ( 1995 ) @xcite . at large scales the difference between axisymmetric secular instability line and the line joining @xmath188 with @xmath189 can not be seen ( for more details see @xcite ) . it is important to discuss briefly the validity of the present perturbative solution for the computation of the properties of maximally rotating neutron stars . the expansion of the radial coordinate of a rotating configuration @xmath193 in powers of angular velocity is written as @xcite @xmath194 where @xmath195 is the difference in the radial coordinate , @xmath196 , between a point located at the polar angle @xmath197 on the surface of constant density @xmath198 in the rotating configuration , and the point located at the same polar angle on the same constant density surface in the static configuration . in the slow rotation regime , the fractional displacement of the surfaces of constant density due to the rotation have to be small , namely @xmath199 , where @xmath200 and @xmath201 and @xmath202 are function of @xmath34 , proportional to @xmath130 . from table [ tab : mrfmax ] , we can see that the configuration with the maximum possible rotation frequency has a maximum fractional displacement @xmath203 as low as @xmath204 and @xmath205 , for the globally and locally neutral neutron stars respectively . in this line , it is worth quoting the results of ref . @xcite , where it has been shown that the inclusion of a third - order expansion @xmath206 in the hartle s method improves the value of the maximum rotation frequency by less than 1% for different eos . the reason for this is that as mentioned above , along the keplerian sequence the deviations from sphericity decrease with density ( see fig . [ fig : eccvsrho ] ) , which ensures the accuracy of the perturbative solution . turning to the increase of the maximum mass , it has been shown in ref . @xcite that the mass of maximally rotating neutron stars , computed with the hartle s second order approximation , is accurate within an error as low as @xmath207 . .maximum static mass @xmath208 and corresponding static radius @xmath209 of neutron stars as computed in ref . @xcite ; maximum rotating mass @xmath210 and corresponding equatorial radius @xmath211 of neutron stars as given in refs . @xcite ; increase in mass @xmath212 and radius @xmath213 of the maximum mass configuration with respect to its non - rotating counterpart ; maximum rotation frequency @xmath187 and corresponding minimum period @xmath214 . [ cols="^,^,^",options="header " , ] [ tab : mrprediction ] along with the observational constraints one should also take into account the theoretical constraints on the mass - radius relations of neutron stars @xcite . indeed only theoretical estimations give the upper and lower bounds for all quantities , describing the properties of neutron stars , to be measured from observations . the equilibrium configurations of static and uniformly rotating neutron stars in both global and local charge neutrality cases have been constructed . to achieve this goal the hartle approach has been applied to the seed static solution , derived from the integration of the einstein - maxwell - thomas - fermi equations @xcite . all physical quantities such as the static and rotating masses , polar and equatorial radii , eccentricity , angular momentum , moment of inertia , quadrupole moment , rotational kinetic energy and gravitational binding energy have been calculated as functions of the central density and the angular velocity of the neutron star . in order to investigate only stable configurations of rotating neutron stars the keplerian mass - shedding limit and the secular axisymmetric instability have been analyzed . this allowed one to construct the stability region , the boundary inside which all stable uniformly rotating neutron stars can be found . in addition the fitting formulas have been obtained for the secular instability boundary in eqs . ( [ eq : secularg ] ) and ( [ eq : secularl ] ) for global and local charge neutrality , respectively . with this analysis the maximum mass and maximum rotation frequency of the neutron star have been established . in order to favor or disfavor some models of neutron stars the current observational constraints on the mass - radius relations related to the maximum observed mass , maximum surface gravity , largest mass , maximum rotation frequency have been analyzed . all these constraints are of paramount importance not only in the physics of neutron stars , but also in nuclear physics to test theoretical hypothesis and assumptions made in the construction of the equations of state . as a result all observations favor stiff equations of state as indicated in ref . @xcite . finally , the results of this chapter have their immediate astrophysical implications in the physics of compact objects , gravitational waves , short and long gamma ray burst , x - ray phenomena occurring in the accretion disks around neutron stars such as quasi periodic oscillations . combining both the theory of compact objects and observational data from different phenomena one can infer information not only on the properties and parameters of neutron stars , but also constrain the equations of state , thus probe the nuclear physics theories @xcite . , j. , freire , p. c. c. , wex , n. , tauris , t. m. , lynch , r. s. , van kerkwijk , m. h. , kramer , m. , bassa , c. , dhillon , v. s. , driebe , t. , hessels , j. w. t. , kaspi , v. m. , kondratiev , v. i. , langer , n. , marsh , t. r. , mclaughlin , m. a. , pennucci , t. t. , ransom , s. m. , stairs , i. h. , van leeuwen , j. , verbiest , j. p. w. , whelan , d. g. , apr . 2013 . a massive pulsar in a compact relativistic binary . science 340 , 448 . 2012nupha.883 .... 1b belvedere , r. , pugliese , d. , rueda , j. a. , ruffini , r. , xue , s .- s . , jun . 2012 . neutron star equilibrium configurations within a fully relativistic theory with strong , weak , electromagnetic , and gravitational interactions . nuclear physics a 883 , 124 . 2012ijmps .. 12 ... 58b boshkayev , k. , rotondo , m. , ruffini , r. , mar . 2012 . on magnetic fields in rotating nuclear matter cores of stellar dimensions . international journal of modern physics conference series 12 , 5867 . , k. nov . rotating white dwarfs and neutron stars in general relativity . thesis http://padis.uniroma1.it/bitstream/10805/1934/1/thesis%20of%20kuantay%20boshkayev.pdf [ http://padis.uniroma1.it/bitstream/10805/1934/1/thesis%20of%20kuantay%20boshkayev.pdf ] heinke06 heinke , c. o. , rybicki , g. b. , narayan , r. , grindlay , j. e. , jun . a hydrogen atmosphere spectral model applied to the neutron star x7 in the globular cluster 47 tucanae . 644 , 10901103 . , m. , rueda , j. a. , ruffini , r. , xue , s .- s . the self - consistent general relativistic solution for a system of degenerate neutrons , protons and electrons in @xmath215-equilibrium . physics letters b 701 , 667671 .

in this chapter we review the recent results on the equilibrium configurations of static and uniformly rotating neutron stars within the hartle formalism . we start from the einstein - maxwell - thomas - fermi equations formulated and extended by belvedere et al . ( 2012 , 2014 ) . we demonstrate how to conduct numerical integration of these equations for different central densities @xmath0 and angular velocities @xmath1 and compute the static @xmath2 and rotating @xmath3 masses , polar @xmath4 and equatorial @xmath5 radii , eccentricity @xmath6 , moment of inertia @xmath7 , angular momentum @xmath8 , as well as the quadrupole moment @xmath9 of the rotating configurations . in order to fulfill the stability criteria of rotating neutron stars we take into considerations the keplerian mass - shedding limit and the axisymmetric secular instability . furthermore , we construct the novel mass - radius relations , calculate the maximum mass and minimum rotation periods ( maximum frequencies ) of neutron stars . eventually , we compare and contrast our results for the globally and locally neutron star models . * pacs * 97.60.jd , 97.10.nf , 97.10.pg , 97.10.kc , 26.60.dd , 26.60.gj , 26.60.kp , 04.40.dg . + * keywords : * neutron stars , equations of state , mass - radius relation .

although recent observations have revealed the outline of the cosmic reionization @xcite , its detailed history and the nature of ionizing sources are not yet fully understood . early forming galaxies can be strong ionizing sources , depending on the escape fraction of their ionizing photons ( @xmath6 ) . this @xmath6 is therefore a key quantity for understanding the cosmic reionization process but its typical ( or effective ) value is not yet clearly established ( e.g. , * ? ? ? the goal of this letter is to put together all existing information on the amount of ionizing photons released by galaxies into the intergalactic medium ( igm ) , both from direct observations and from indirect estimations based on the effects of the igm ionization itself . in order to make an homogeneous comparison , we introduce a quantity derived from @xmath6 , the _ escape _ flux density ( hz@xmath7 ) ratio of the lyman continuum ( lyc , @xmath8 rest - frame ) to the non - ionizing ultraviolet ( uv , @xmath9 rest - frame ) , @xmath10 . this @xmath10 is naturally measured by direct observations ( e.g. , * ? ? ? we compile all the observations of the lyc from galaxies and derive @xmath10 at @xmath11 . on the other hand , @xmath10 can be derived from the ionizing background intensity inferred from igm absorption measurements via a cosmological radiative transfer model . we derive @xmath10 at @xmath06 based on the recent reports of the background intensities . the cosmological parameters assumed in this letter are @xmath12 km s@xmath7 mpc@xmath7 , @xmath13 , and @xmath14 . the definition of @xmath6 in this letter is @xmath15 where @xmath16 and @xmath17 are the intrinsic and escaping lyc flux densities , respectively . to obtain @xmath17 , we have to correct the observed lyc flux density @xmath18 for the photoelectric absorption by the neutral hydrogen remained in the igm : @xmath19 we may estimate @xmath16 from the observed uv flux density @xmath20 as follows : @xmath21 where @xmath22 is the intrinsic lyc - to - uv flux density ratio and @xmath23 is the uv dust opacity in the interstellar medium of a galaxy . therefore , we have @xmath24 and @xmath25 this @xmath10 is called the lyc - to - uv escape flux density ratio in this letter . as found in equation ( 5 ) , we can obtain @xmath10 from the observed lyc - to - uv flux density ratio with a correction for the igm absorption against the lyc . this is similar to the approach introduced by @xcite . on the other hand , @xmath10 can be obtained from the observed ionizing background intensity through a cosmological radiative transfer model . we describe the different aspects of this estimation in the four following sub - sections . the mean specific intensity at the observed frequency @xmath26 as seen by an observer at redshift @xmath27 is given by ( e.g. , * ? ? ? * ) @xmath28 where the frequency @xmath29 , @xmath30 is the line element , @xmath31 is the emissivity per unit comoving volume , and @xmath32 is the effective igm opacity . although the upper limit of the integral could be infinity , we set it to @xmath33 . the line element is @xmath34 where @xmath35 is the speed of light and @xmath36 is the hubble parameter . the comoving emissivity can be written as @xmath37 where @xmath38 and @xmath39 are the qso and galactic emissivities , respectively . the igm emissivity can be omitted ( see 2.3 ) . the effective igm opacity is ( e.g. , * ? ? ? * ) @xmath40 where @xmath41 is the cloud number on a line of sight per unit redshift @xmath42 interval and per unit h i column density @xmath43 interval , and @xmath44 and @xmath45 are the lower and upper limits of the column density of the igm clouds . the cloud optical depth is @xmath46 with @xmath47 the hydrogen cross - section at the frequency @xmath48 . the frequency dependence of @xmath47 for the lyc is assumed to be @xmath49 . the effective igm opacity is determined by the cloud number function which can be expressed as ( e.g. , * ? ? ? * ) @xmath50 if we assume a single index @xmath51 @xmath52 for all @xmath43 and @xmath42 , and @xmath53 and @xmath54 , equation ( 9 ) can be approximated to @xcite @xmath55^{\beta-1 } ( 1+z')^\gamma dz'\,,\ ] ] where @xmath56 is the usual gamma function , @xmath57 is the number of the igm clouds with the column density between @xmath44 and @xmath45 on a line of sight per unit redshift interval , and @xmath48 . according to @xcite and @xcite , we assume @xmath58 , @xmath59 for @xmath60 and @xmath61 for @xmath62 , and @xmath63 @xmath64 . note that we have assumed the same number evolution along the redshift ( the index @xmath65 ) for the ly@xmath66 forest and for the lyman limit system ( and also for the damped ly@xmath66 clouds ) , whereas @xcite adopted different indices for these clouds . as shown in the appendix of @xcite , the same @xmath65 for all clouds is compatible with the recent observations of the lyman limit systems by @xcite . while the igm opacity model is used in this paper for the lyc , we display in fig . 1 the opacity predictions at ly@xmath66 line where comparisons with observations are possible . the ly@xmath66 cross - section is based on @xcite . note that the same igm clouds contribute to the lyc and ly@xmath66 opacities . our model ( solid line ) shows a very good agreement with the data at @xmath67 but a smaller opacity than the data at @xmath68 . an explanation is that the parameters of the cloud distribution taken from @xcite who analysed the ly@xmath66 forest at @xmath69 . if we assume @xmath70 for @xmath71 , the model ( dashed line ) shows a very good agreement with the data . the solid and dashed lines are the igm ly@xmath66 opacities calculated by equation ( 11 ) ( see the text for the detail ) . the circles and squares with error - bars are the observed opacity taken from songaila ( 2004 ) and fan et al . ( 2006 ) , respectively.,width=302 ] the lyc emissivity from galaxies can be expressed as @xmath72 where @xmath73 is the observed galactic uv emissivity per unit comoving volume , and @xmath74 is the lyc frequency dependence of galaxies . we assume that @xmath75 ( e.g. , * ? ? ? * ) , independently of the redshift . the effect of the spectral slope on the estimated @xmath10 is 15% at most if the index is changed from 0 to @xmath76 . a redshift dependence of @xmath10 is introduced . a key feature of equation ( 12 ) is that the lyc emissivity is directly related to the observed uv emissivity with the only parameter @xmath10 that we wish to determine . in other words , our estimate for @xmath77 is free from uncertain dust correction and from the intrinsic lyc production rate ( or @xmath78 ) which depends on the initial mass function . the @xmath73 values are obtained from the integration to zero luminosity of uv luminosity functions compiled from the literature . they are shown as a function of redshift in fig . 2 . the error - bars indicate observational 1-@xmath79 errors . for practical purpose , we express @xmath73 as the following functional form , @xmath80 , with @xmath81 the normalization adopted is @xmath82 erg s@xmath7 hz@xmath7 mpc@xmath83 based on @xcite . @xmath73 of equation ( 13 ) is displayed as the solid line in fig . 2 and called the standard case . because of the scatter at @xmath84 , we consider another case , the high - emissivity case , with @xmath85 for @xmath86 , displayed in fig . 2 as the dotted line . the qso emissivity is given by the prescription in @xcite who obtained the lyc emissivity of qsos based on the optical luminosity functions of @xcite and @xcite and an average qso spectrum . we here adopt a qso spectrum parameterized by @xcite . the ionized igm can contribute to a fraction of the lyc emissivity by the recombination process @xcite . however , it is obviously secondary . thus , we omit it . @xcite suggested a significant contribution to the lyc by the thermal emission from hot gas shock - heated by the cosmological structure formation . in spite of its hard spectrum , however , the estimated hydrogen ionization rate is an order of magnitude smaller than that reported by @xcite at @xmath874 and is less than one - third of that reported by @xcite at @xmath886 . thus , we also omit the thermal emission by the structure formation . the omission of the igm emissivity leads to an overestimate of @xmath10 that should be less than a factor of 2 . if we enter the igm opacity model described in 2.2 and the lyc emissivities described in 2.3 into equation ( 6 ) , we can predict the background intensity at the lyman limit as a function of @xmath10 . this latter quantity can be evaluated by a comparison between the predicted and observed intensities . the observed intensities are obtained in the following ways . at @xmath69 , @xcite presented intensities at the lyman limit based on their observations of the qso proximity effect . @xcite and @xcite estimated hydrogen ionization rates ( @xmath89 ) at @xmath874 and at @xmath886 , respectively , based on the observed igm ly@xmath66 opacity and a cosmological hydrodynamical simulation . the ionization rate @xmath89 is @xmath90 where the subscript l means the quantity at the lyman limit , @xmath91 is the planck constant , and @xmath66 is the power - law index when we assume a power - law background radiation ( @xmath92 ) . assuming @xmath93 , we have @xmath94 with @xmath95 @xmath96 @xcite . the uncertainty on @xmath97 resulting from the adopted @xmath66 is a factor of @xmath98 . with this conversion we obtain @xmath97 from the @xmath89 values of @xcite and @xcite . we should note here that @xcite determined their absolute @xmath89 values so as to fit those of @xcite which are somewhat smaller than those of @xcite . @xmath10 is determined by a comparison of these observed intensities at the lyman limit ( @xmath99 ) with the calculated theoretical intensities as @xmath100 where @xmath101 and @xmath102 are the lyman limit intensities originating from qsos and galaxies , respectively . this comparison is performed by a non - parametric way from high to low redshift . calculating @xmath102 ( and then @xmath10 ) at a redshift , we use the @xmath10 values at redshifts larger than the redshift . for three different references of the observed intensities , we estimate @xmath10 independently . the assumed initial redshifts in equation ( 6 ) are 5.0 , 5.0 , and 6.0 for the data from @xcite , @xcite , and @xcite , respectively . this choice does not affect the results if we take an enough high redshift because the igm opacity is very large at high redshift . in addition , we note that we assumed the higher opacity case at @xmath71 ( shown in fig . 1 as dashed line ) only for the data from @xcite . the values of @xmath10 obtained at different redshifts either from direct observations or from indirect estimations ( 2.4 ) are displayed in fig . 3 . that will be used in fig . 4.,width=302 ] their comparison reveals an evolution of @xmath10 , with @xmath10 getting larger at higher redshifts . if we put confidence in the upper limit of @xcite at @xmath103 , which is an average value of @xmath10 for 11 star - forming galaxies and is less affected than a single measurement by the randomness of the lyc escape , we can obtain a possible evolution , for example , as @xmath104 from @xmath105 to @xmath106 , which is shown as the dashed line in fig . since the uncertainties are still very large , this rather strong evolution is just an example . however , we note that an evolution of @xmath10 is also found even in the high - emissivity case ( open symbols ) . such an increasing @xmath10 was suggested by @xcite in his fig . 1 although the redshift coverage is smaller than ours . the measurement by @xcite is somewhat larger than our estimations . this may be caused by their sample selection bias . indeed , their sample galaxies are taken from the bluest quartile in the uv colour among lyman break galaxies . at lower redshifts , we have only upper limits , except for the measurement by @xcite , and a large dispersion . this is probably due to the small number of the observed galaxies . indeed , only 7 galaxies have been observed to date @xcite . the galaxy observed by @xcite ( haro 11 ) may be an exceptional one in the local universe as discussed in their paper . 4 shows a comparison of the calculated ionizing intensity with the observed ones . assuming the evolution of @xmath10 shown in fig . 3 as the dashed line , we obtain the dotted line for the galactic contribution and the solid line for the total intensity . the dashed line is the qso contribution . the main contributor to the ionizing background radiation changes from galaxies to qsos at @xmath107 . are measurements taken from scott et al . the triangles and squares are converted from the ionization rates of bolton et al . ( 2005 ) and fan et al . ( 2006 ) , respectively . the dashed line is the intensity from the qsos ( bianchi et al . the solid and dotted lines are the total and galactic intensities , respectively . the standard galactic emissivity and an evolution of the escape flux density ratio shown in fig . 3 as the dashed line are assumed.,width=302 ] we have discussed the escape flux density ratio of the lyc to the non - ionizing uv ( @xmath10 ) in a new framework for comparing the direct observations of the lyc from galaxies with the cosmic ionizing background intensity . we have found a redshift evolution of @xmath10 , with larger @xmath10 at higher redshifts . here , we translate @xmath10 into the escape fraction ( @xmath6 ) of ionizing photons from galaxies . by equation ( 4 ) , we can convert @xmath10 into @xmath6 if we have @xmath78 and @xmath23 . for normal stellar populations with the standard salpeter initial mass function and mass range 0.1100 @xmath108 , @xmath1090.3 ( e.g. , * ? ? ? * ) . from the observed uv slope , @xmath1102 for the lyman break galaxies at @xmath107 ( e.g. , * ? * ) although the uncertainty is very large . in this case , we have @xmath111 . if this relation is valid in the range @xmath06 , our result means that @xmath6 increases from a value less than 0.01 at @xmath3 to about @xmath4 at @xmath5 . if we explain the evolving @xmath10 by a constant @xmath6 , we have two possibilities according to equation ( 4 ) ; ( 1 ) an evolving @xmath78 and ( 2 ) an evolving @xmath23 . the evolving @xmath78 case needs an order of magnitude larger production rates of the lyc relative to the uv at @xmath5 than at @xmath3 . this might be realized by a top - heavy stellar initial mass function as suggested for zero metallicity , population iii ( popiii ) stars . very massive ( @xmath112 @xmath108 ) popiii stars show almost black - body spectra with an effective temperature about @xmath113 k ( e.g. , * ? ? ? * ) and would give @xmath114 . however , such a transition of the mass function would occur at a much higher redshift than @xmath1154 because it is thought to happen at a very low metallicity like @xmath116 solar value ( e.g. , * ? ? ? although @xcite proposed that 1030% of stars in @xmath1174 star - forming galaxies are massive popiii stars , we would need more such stars to obtain an enough large @xmath78 . the evolving @xmath23 case needs an increment of 23 mag of @xmath23 from @xmath3 to @xmath5 . such a systematic increase of @xmath23 for galaxies at high redshifts is not reported , whereas @xcite reported an opposite trend for @xmath874 ( it may be caused by a selection bias in high redshift sample ) . we are therefore left with an evolution of @xmath6 , with a larger @xmath6 at a higher redshift . the escape of the lyc is probably a random phenomenon . thus , the evolving @xmath6 suggests an increase of the fraction of galaxies showing a large escape . according to theoretical studies , a smaller galaxy can show a larger @xmath6 because of galactic winds @xcite and/or champagne flows @xcite . thus , the evolving @xmath6 may suggest a decrease at higher redshift of the average scale of galaxies , which is consistent with the cold dark matter scenario . the evolving @xmath6 also suggests a morphological evolution of galaxies ; at higher redshifts , a large fraction of galaxies would show very disturbed structure caused by galactic winds and/or champagne flows , yielding a large escape of the lyc . we note two issues which should be assessed in future . if there are a large number of type ii qsos at high redshifts as suggested by @xcite , the required @xmath10 may become much smaller . if the fraction of galaxies having a low - luminosity active galactic nucleus increases towards high redshifts as suggested by @xcite , a scenario with an evolving @xmath78 without an evolving @xmath6 may become possible . finally , we strongly encourage new direct measurements of the lyc from galaxies to confirm / reject the evolving @xmath10 . for @xmath107 , the ground - based large telescopes are useful , and for @xmath103 , the _ galex _ satellite is appropriate . we thank the referee , simone bianchi , for comments helping us to improve this letter .

the escape fraction of ionizing photons from galaxies is a crucial quantity controlling the cosmic ionizing background radiation and the reionization . various estimates of this parameter can be obtained in the redshift range , @xmath06 , either from direct observations or from the observed ionizing background intensities . we compare them homogeneously in terms of the observed flux density ratio of ionizing ( @xmath1 rest - frame ) to non - ionizing ultraviolet ( @xmath2 rest - frame ) corrected for the intergalactic absorption . the escape fraction is found to increase by an order of magnitude , from a value less than 0.01 at @xmath3 to about @xmath4 at @xmath5 . [ firstpage ] cosmology : observation diffuse radiation galaxies : evolution intergalactic medium

even though the master equation ( 2 ) involves only the fock states and not the full many - body hilbert space containing also their coherent superpositions , the problem still grows exponentially with both particle number @xmath0 and system size @xmath35 . therefore , in order to compute exact steady - state expectation values from eq . ( 2 ) , we have to resort to quantum - jump - type monte - carlo ( mc ) simulations ( crosses in figs . 1 and 2 ) [ 40 ] : we generate a random process in the classical space of sharp occupation numbers @xmath19 . namely , according to the sum and the relative weight of the many - body rates @xmath131 leading away from the current state @xmath19 , we draw both the time after which a quantum jump happens and the new state @xmath132 , respectively . expectation values like @xmath25 are computed by averaging over a random path . the monte - carlo method gives quasi exact results , in the sense that the accuracy is controlled by the length of the random path . it is instructive to release the restriction @xmath11 to @xmath133 , still assuming that every state @xmath10 can be reached from every other state @xmath9 via a sequence of finite - rate quantum jumps . in this case the set of selected states can not consist of different subsets with zero mutual rates , as this would leave the relative occupation of the subsets undetermined . moreover , the leading contribution to the asymptotic occupation @xmath25 of a non - selected state @xmath10 with rates @xmath134 for all selected states @xmath9 appears in a higher order of our expansion ( but is still independent of @xmath0 such that the occupations of the non - selected states saturate for large @xmath0 ) . this suggests that if there is one class of rates that are non - zero , but very small compared to the other rates , then the occupation numbers @xmath25 can be dominated by higher - order corrections , before for very large @xmath28 they eventually reach their asymptotic values given by eqs . ( 4 ) and ( 5 ) the rate imbalances @xmath117 form a skew - symmetric matrix @xmath135 . let the vector @xmath136 describe the occupations of the single - particle states in the leading order of our expansion in powers of @xmath66 . in this order only the selected states @xmath137 have non - zero positive occupation numbers @xmath138 , whereas @xmath139 for @xmath140 . we can define the vector @xmath141 . according to eq . ( 4 ) one has @xmath142 for @xmath137 and , moreover , according to eq . ( 5 ) one has to require @xmath143 for @xmath140 in order to have finite , positive occupations for the non - selected states . thus , the problem of finding an asymptotic mean - field solution can be reduced to the problem of finding a set @xmath63 of selected states such that @xmath141 with @xmath144 and @xmath142 for @xmath137 and @xmath139 and @xmath143 for @xmath140 . ( note that we exclude the non - generic situation with @xmath145 for some @xmath10 that requires fine - tuned rate matrices . this situation , however , is relevant at transition points , where the set of selected states changes in response to a parameter variation . ) assume there exist two different sets @xmath146 and @xmath147 both leading to physical solutions @xmath148 and @xmath149 giving rise to non - negative occupation numbers . then , it follows that @xmath150 , using @xmath151 . however , this requires that both @xmath152 and @xmath153 , such that @xmath154 _ and _ @xmath155 , respectively , leading to the contradiction @xmath156 . restrict @xmath63 to sets comprising an odd number @xmath30 of states ( generically even @xmath30 do not allow for solutions , as explained in the main text ) . each choice of @xmath63 gives rise to a possibly non - physical solution @xmath157 with @xmath158 . the vector of signs @xmath159 with @xmath160 if @xmath137 and @xmath161 if @xmath140 , with the overall sign fixed by the convention @xmath162 , distinguishes physical solutions ( @xmath163 for all @xmath10 ) from non - physical solutions . now one can observe : ( i ) each vector @xmath164 occurs at most once . namely , if @xmath146 and @xmath147 gave rise to the same vector @xmath159 then the modified rate imbalance matrix @xmath165 had the two physical solutions @xmath146 and @xmath147 in contradiction to the uniqueness . ( ii ) the number @xmath166 of possible vectors @xmath159 equals the number @xmath167 of possible sets @xmath63 . therefore , each vector @xmath159 occurs once , in particular the one with @xmath163 for all @xmath10 leading to the physical solution with positive occupations . a quantum floquet system , characterized by a time - periodic hamiltonian @xmath116 , possesses solutions of the form @xmath168 , with quasienergies @xmath169 and time - periodic floquet modes @xmath170 forming a complete orthonormal basis at every instant in time [ 46 - 48 ] . when coupled weakly to a thermal bath , according to floquet - markov theory [ 49 - 52 ] , the steady state is given by a time - periodic density matrix @xmath171 that is diagonal in the floquet modes and characterized by the time - independent probabilities @xmath42 obeying a master equation of the form ( 1 ) . we assume a bath of harmonic oscillators @xmath172 , with angular frequencies @xmath173 and annihilation operators @xmath174 , that couples to the system operator @xmath175 like @xmath176 . then the rates read @xmath177 with @xmath178 and bath correlation function @xmath179^{-1}$ ] . the latter is determined by the inverse temperature @xmath78 and the spectral density @xmath180 $ ] that we assume to be ohmic , @xmath181 . the floquet gas of @xmath0 non - interacting bosons is described by eq . ( 2 ) with @xmath19 denoting the occupation numbers of the single - particle floquet modes . remarkably , we obtain the temperature - independent expression @xmath182 for the rate differences . this implies that for a periodically driven system the set of selected states does not depend on the bath temperature . in particular , according to eqs . ( 4 ) and ( 5 ) the relative occupations among the selected states do not depend on temperature , whereas the occupations of the non - selected states ( and thus also the crossover density @xmath34 ) are temperature dependent . in the limit of vanishing periodic forcing ( i.e. for a system described by a time - independent hamiltonian ) the floquet modes and quasienergies approach the eigenmodes and eigenenergies of the time - independent hamiltonian , such that @xmath183 is non - zero only for @xmath184 . as a consequence the rates @xmath8 are given by a single term only @xmath185 and , thanks to the general property @xmath186 of the bath correlation function , give rise to detailed balance , with @xmath187 $ ] . in contrast , for finite periodic driving detailed balance is not guaranteed , because terms with different @xmath188 contribute to the rates . in a non - driven system detailed balance can be broken by coupling the system weakly to different baths of different temperature ; in this case the rates resulting from the different baths simply sum up , @xmath189 . the assumption of a diagonal steady - state density matrix , @xmath190 or @xmath191 , for weak enough coupling to a heat bath has to be discussed carefully for floquet systems . in these systems all quasienergies can be placed in a finite energy interval of size @xmath192 , leading to smaller and smaller near degeneracies as the size @xmath35 of the single - particle state space is increased . the assumption of a diagonal density matrix is fulfilled only , if for any two states , their coupling provided by the heat bath is smaller than their quasienergy splitting . in the limit @xmath193 this assumption can not be guaranteed and the consequences for single particle systems and their density matrix have been discussed in ref . [ 52 ] . for many - particle systems , even when @xmath35 is fixed , the state space increases exponentially with the particle number @xmath0 and therefore the size of the smallest quasienergy splitting decreases exponentially . quickly it will be much smaller than even the tiniest experimentally realizable coupling to a heat bath . this could destroy the diagonality of the density matrix for large particle numbers @xmath0 . this is not the case , however , for the here studied non - interacting gas . there all transitions between many - particle floquet states can be reduced to single - particle transitions for which the single - particle quasienergy splittings are relevant . therefore we can make the assumption of a diagonal density matrix for a non - interacting gas . a switching mechanism due to an avoided quasienergy crossing in a single - particle system was reported in ref . [ it allowed to switch the asymptotic steady state between the lower and the upper well of an asymmetric double - well potential by a periodic driving , that is much weaker than the asymmetry . it works even if the floquet states involved in the avoided crossing have just a small occupation . it would be interesting to study how such a switching mechanism due to an avoided crossing is reflected in the corresponding ideal bose gas and how it affects the set of bose selected states . rates @xmath194 for a quantum jump from state @xmath10 to state @xmath9 for the tight - binding chain in contact with two heat baths ( figs . 2a - c in the main text ) . from left to right the three plots correspond to the system coupled to the positive temperature bath only , coupled to both baths with equal strength , coupled to the bath with negative temperature only , respectively . ] mean occupations @xmath25 versus total particle number @xmath0 for a fully chaotic kicked rotor ( a ) and a driven quartic oscillator with mixed phase space ( b ) coupled to a thermal bath . the corresponding rate matrices are given in subfigures ( c ) and ( d ) , respectively . both ( a ) and ( b ) show bose selection for large @xmath0 . red ( blue ) lines indicate regular ( chaotic ) states , solid ( dashed ) lines result from mean - field ( asymptotic ) theory . insets : stroboscopic poincar plots of the classical dynamics in phase space in comparison to the husimi distribution of a representative chaotic and regular quantum state ( red shaded ) . ] in fig . s[fig : rates ] we plot the rate matrix used for the tight - binding chain coupled to two heath baths of positive and negative temparature , as it is described in figs . 2a - c of the main text . from left to right the three plots correspond to the system coupled only to the positive temperature bath , coupled to both baths with equal strength , coupled only to the bath with negative temperature , respectively . for the leftmost ( rightmost ) case simply the state @xmath62 with the lowest ( largest ) energy is selected , since for this state @xmath73 . for the middle case with coupling to both baths we find three selected states , namely the states 4 , 5 and 10 ( labeling the states by @xmath195 from the lowest to the highest energy ) . the selected states 4 and 5 are characterized by low rates to states of higher energy and from states with lower energy . however , we are not aware of a simple general strategy for guessing the selected sates simply from looking at the rate matrix . [ [ two - further - examples - for - bose - selection - in - time - periodically - driven - model - systems ] ] two further examples for bose selection in time - periodically driven model systems ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in order to emphasize the fact that bose selection is a generic and robust effect in driven - dissipative ideal bose gases , let us study the phenomenon also in the two time - periodically driven model systems in weak contact with a thermal bath that , on the single - particle level , were discussed in detail in ref . the first one is the kicked rotor with parameters giving @xmath196 floquet modes that in the semiclassical sense are all chaotic . the resulting rates @xmath8 ( plotted in fig . s[fig : floquet]c ) appear to be exponentially distributed and give rise to the bose selection of @xmath197 states , as shown in fig . s[fig : floquet]a . the second system ( fig . s[fig : floquet]b ) is a periodically driven quartic oscillator having a mixed phase space with a central region of 26 chaotic states , surrounded by infinitely many regular states that can be labelled by @xmath198 in radial direction . bose selection occurs among the chaotic states that are connected by random - like highly coordinated rates ( the rate matrix is plotted in fig . s[fig : floquet]d ) ; 21 states are selected . in turn , the regular states are coupled essentially to their nearest neighbor only and their occupation decays rapidly with increasing @xmath10 , allowing us to truncate modes with @xmath199 .

bose - einstein condensation , the macroscopic occupation of a single quantum state , appears in equilibrium quantum statistical mechanics and persists also in the hydrodynamic regime close to equilibrium . here we show that even when a degenerate bose gas is driven into a steady state far from equilibrium , where the notion of a single - particle ground state becomes meaningless , bose - einstein condensation survives in a generalized form : the unambiguous selection of an odd number of states acquiring large occupations . within mean - field theory we derive a criterion for when a single and when multiple states are _ bose selected _ in a non - interacting gas . we study the effect in several driven - dissipative model systems , and propose a quantum switch for heat conductivity based on shifting between one and three selected states . in quantum many - body physics there is currently a huge interest in non - equilibrium phenomena beyond the hydrodynamic description of systems retaining approximate local equilibrium . recently , intriguing results have been obtained for paradigmatic scenarios : the dynamics away from equilibrium in response to a parameter variation @xcite , the possible relaxation towards equilibrium @xcite versus many - body localization @xcite , and the control of many - body physics by means of strong periodic forcing @xcite . another fundamental scenario of quantum many - body dynamics are non - equilibrium steady states of driven - dissipative systems , with transport of , e.g. , mass or energy through the system @xcite . in this context one might ask simple questions : what are the properties of an ideal bose gas driven to a steady state far from equilibrium ? in particular , what happens in the quantum degenerate regime , where in equilibrium bose - einstein condensation occurs ? in this letter we investigate the quantum degenerate regime of driven - dissipative ideal bose gases of @xmath0 particles in steady states far from equilibrium , assuming weak coupling to the environment . examples of such systems comprise bosons coupled to two heat baths of different temperature and time - periodically forced bosons in contact with a single heat bath . for large densities these systems are found to exhibit an intriguing generic behavior , namely the single - particle states unambiguously separate into two groups : one , that we call _ bose selected _ , whose occupations increase linearly when the total particle number is increased at fixed system size , and another one whose occupations saturate . remarkably , this generalized form of bose condensation is a very consequence of bosonic indistinguishability , not relying on thermodynamic equilibrium . we show examples both with the number of selected states being extensive and of order one ; the latter corresponding to a fragmented condensate @xcite with a macroscopic occupation of each selected state ( not relying on ground - state degeneracy ) . we propose to switch the heat conductivity of a system by shifting between one selected state ( corresponding to standard bose condensation ) and three selected states . our findings are relevant for artificial many - body quantum systems such as superconducting and optical circuits @xcite , exciton - polariton fluids @xcite , or photons in a dye - filled cavity @xcite , that are intrinsically driven - dissipative . tailored dissipation has also been used or proposed as a powerful tool for quantum engineering in ultracold atomic quantum gases @xcite and trapped ions @xcite . our results , moreover , provide a connection between bose condensation in quantum systems and the phenomenon of real - space condensation in classical non - equilibrium models @xcite , where also condensation into multiple states has been found @xcite . consider an open quantum system of a single particle weakly coupled to an environment , with reduced density operator @xmath1 . the dynamical map for the time evolution of @xmath1 shall be given by a markovian master equation @xmath2 with linear liouvillian @xmath3 , guiding the system into a steady state @xmath4 that is diagonal with respect to the ( quasi)energy eigenstates @xmath5 @xcite . the dynamics of the diagonal elements @xmath6 is given by @xmath7 with rates @xmath8 for a quantum jump from @xmath9 to @xmath10 that , for simplicity , we assume to be strictly positive , @xmath11 . now we generalize the single - particle problem ( [ eq : spr ] ) to a gas of @xmath0 non - interacting bosons . the many - body steady state @xmath12 will be diagonal in the fock basis @xmath13 labeled by the occupation numbers @xmath14 of the single - particle states @xmath10 , obeying @xmath15 : @xmath16 . the @xmath0-boson rate equation reads @xmath17 where @xmath18 denotes the occupation numbers obtained from @xmath19 by transferring one particle from @xmath10 to @xmath9 @xcite . the steady state with @xmath20 is unique , if every state can be connected to every other one via a sequence of finite - rate quantum jumps @xcite . this is true for every @xmath0 , if it is true for the single - particle problem ( [ eq : spr ] ) . equation ( [ eq : mpr ] ) is classical in the sense that it involves the diagonal elements of the density matrix only . however , the bosonic quantum statistics is reflected in the fact that the rate for a jump from @xmath10 to @xmath9 depends both on @xmath21 and @xmath22 . this rate reads @xmath23 with @xmath24 for fermions , distinguishable particles , and bosons , respectively . ( color online ) ( a ) mean occupations @xmath25 versus density @xmath26 for one realization of the random - rate matrix @xmath8 with @xmath27 . crosses , solid and dashed lines from monte carlo , mean - field , and asymptotic theory , respectively . for large @xmath28 the bosonic quantum statistics leads to the bose selection of @xmath29 states whose occupation grows linearly with @xmath28 . inset : distribution of @xmath30 for the ensemble of rate matrices . ( b ) like ( a ) , but with @xmath31 and @xmath32 . thick lines : average occupation of a selected and a non - selected state , exactly ( dashed ) and assuming equal occupation @xmath28 for @xmath33 followed by saturation of the non - selected occupations ( solid ) . inset : crossover density @xmath34 versus system size @xmath35 . ] as a transparent model system , we will first consider rate matrices @xmath8 given by exponentially distributed , independent random numbers @xcite . this choice is motivated by the distribution of rates obtained for fully chaotic systems @xcite . the rates determine the system completely ( and there is no need to define , e.g. , the energy of a state @xmath10 . ) in this model the number of states @xmath35 corresponds to the system size and , thus , the filling factor @xmath36 to the density . in fig . [ fig : bs ] , we plot the mean steady - state occupations @xmath25 versus @xmath28 , for two realizations of @xmath8 with @xmath27 and @xmath31 ; here @xmath37 with number operator @xmath38 and @xmath39 . crosses result from quasi exact monte - carlo ( mc ) simulations @xcite , solid and dashed lines from the mean - field ( mf ) and asymptotic theory derived below , respectively . the filling @xmath28 directly controls quantum degeneracy : in the non - degenerate regime @xmath40 the relative occupations @xmath41 approach the @xmath28-independent single - particle probabilities @xmath42 . quantum - statistical corrections make themselves felt when entering the degenerate regime at @xmath43 . for even larger densities @xmath28 , around a crossover value @xmath34 , we observe that for a group of @xmath30 single - particle states the occupation grows linearly with @xmath0 , while the occupation of the remaining states saturates . this is the aforementioned effect of bose selection . asymptotically , in the ultra - degenerate regime @xmath44 , the relative occupations of the selected states , @xmath41 , as well as the absolute occupations @xmath25 of the non - selected states become again independent of @xmath28 . within the ensemble of rate matrices the number of selected states @xmath30 is found to be always odd ( e.g. fig . [ fig : bs]a , inset ) and on average @xmath45 with fluctuations @xmath46 ( a system with non - extensive @xmath47 is presented below ) . the crossover density @xmath34 at which the crossover to bose selection occurs , is given by the average number of particles that asymptotically ( for large @xmath28 ) occupies a non - selected state . this can be seen by assuming that for @xmath33 both groups of states are equally populated with @xmath28 particles on average , until at @xmath48 the average occupation of the non - selected states reaches and keeps the saturation value @xmath34 ( fig . [ fig : bs]b , thick solid lines ) . in the random - rate model @xmath34 increases like @xmath46 with @xmath35 ( fig . [ fig : bs]b , inset ) @xcite . therefore , in this model bose selection does not occur in the thermodynamic limit , @xmath49 keeping @xmath28 constant , but in finite systems . this behaviour resembles finite - temperature equilibrium bose condensation in one dimension . in order to treat large systems and to understand the behaviour visible in fig . [ fig : bs ] , we derive a mf theory from the equation of motion @xmath50 for the @xmath25 by approximating two - state correlations @xmath51 by the trivial ones given by wick decomposition , @xmath52 ( for @xmath53 ) . this gives a closed set of non - linear equations for the @xmath25 , @xmath54,\ ] ] and is equivalent to a factorized gaussian ansatz @xmath55 with @xmath56 . the mf theory is confirmed by the mc data ( figs . [ fig : bs]a , [ fig : chain]a , and [ fig : chain]d ) @xcite . an asymptotic theory for the ultra - degenerate regime , particle number to infinity at fixed system size , ( not to be confused with the thermodynamic limit : system size to infinity at fixed density ) can be derived from the mf eq . ( [ eq : mf ] ) for @xmath57 . the naive approximation @xmath58 leads to the set of equations @xmath59 that generally does not possess a physical solution with non - negative occupations @xmath60 , unless several of the @xmath25 vanish . this gives already a hint why bose selection occurs , but it does not tell us which states become selected , since , e.g. , @xmath61 would be a solution for _ any _ state @xmath62 . a systematic theory is obtained by assuming that there is some ( yet to be determined ) set @xmath63 of bose selected single - particle states with occupation numbers @xmath64 that are large compared to one as well as to the occupations of the non - selected states @xmath65 . this allows us to expand the @xmath25 in powers of @xmath66 . in leading order we obtain the closed set of linear equations for the bose selected states , @xmath67 the fact that @xmath68 is a skew - symmetric matrix guarantees a zero determinant and a solution of eq . ( [ eq : selected ] ) provided the set @xmath63 contains an odd number @xmath30 of states ( for even @xmath30 the existence of a solution requires fine tuned rates @xmath8 ) . thus generically one expects an odd number of selected states , in accordance with the numerically obtained distribution ( fig . [ fig : bs]a , inset ) . the next order describes the occupations of the non - selected states , @xmath69 and gives also corrections to the occupations of the selected states that we omit here ( even higher orders become essential when allowing also for zero rates , @xmath70 @xcite ) . equation ( [ eq : selected ] ) determines the relative occupations among the selected states . these are independent of the total particle number @xmath0 and , in turn , dictate the absolute occupations of the non - selected states via eq . ( [ eq : nonselected ] ) . the latter , thus , do not depend on the total particle number @xmath0 , corresponding to the saturation behaviour visible in fig . [ fig : bs ] . the total number of particles occupying the selected states , including corrections to the leading order ( [ eq : selected ] ) , is given by @xmath71 and increases linearly with @xmath0 ( since the `` depletion '' @xmath72 is independent of @xmath0 ) . this behavior is generic for the ultra - degenerate regime and generalizes bose condensation , where the occupation of a single state @xmath62 increases with @xmath0 . remarkably , bose selection is a very consequence of the bosonic quantum statistics and does not require any assumptions based on equilibrium statistical mechanics . the set @xmath63 of selected states has to be determined by the physical requirement that the occupations @xmath25 of both the selected and the non - selected states are non - negative . it can be shown that a unique physical solution with positive occupations exists @xcite , as expected from the fact that a unique steady state of eq . ( [ eq : mpr ] ) exists ( see supplemental material @xcite ) . we are not aware of an easy strategy ( beyond trial and error ) that generally allows to determine which and how many states are selected . however , if there is a ground - state - like single - particle state @xmath62 , characterized by @xmath73 for all @xmath74 , then only this state @xmath62 will be selected and @xmath75 , corresponding to bose - einstein condensation . namely , since eq . ( [ eq : selected ] ) is fulfilled trivially and eq . ( [ eq : nonselected ] ) gives positive occupations @xmath76^{-1}>0 $ ] for the non - selected states , this is the ( unique ) physical solution . in contrast , as soon as there is no such ground - state - like state @xmath62 anymore , then more than a single state must be selected . an important special case are rate matrices for a system with single - particle energies @xmath77 in weak contact with a thermal bath of inverse temperature @xmath78 for which the rate matrices obey @xmath79 $ ] . such rates guarantee detailed balance , i.e. the existence of an equilibrium steady state for which each summand on the right - hand side of eqs . ( [ eq : spr ] ) and ( [ eq : mpr ] ) vanishes independently . in the ultra - degenerate regime , one then recovers from eq . ( [ eq : nonselected ] ) the familiar expression @xmath80 - 1\}^{-1}$ ] for @xmath81 while @xmath82 . therefore ( excluding ground - state degeneracy @xmath83 ) a non - equilibrium steady state breaking detailed balance , as it is found in driven - dissipative systems , is a necessary condition for observing bose selection of more than a single state . however , breaking detailed balance is not sufficient , as can be inferred from the example of a system driven between two baths of different positive temperature . in this case the rates sum up @xmath84 and , despite the fact that the combined rates do not lead to detailed balance anymore , they still obey @xmath85 for all @xmath86 such that only the ground state will be selected . below we will discuss concrete systems of two classes for which @xmath87 is found naturally : ( i ) systems in weak contact with two baths , one with positive temperature and another , energy - inverted one with negative temperature , and ( ii ) time - periodically driven systems in weak contact with a thermal bath . ( color online ) ( a ) occupations @xmath25 from mean - field ( lines ) and monte - carlo ( crosses ) calculations for @xmath88 bosons on a tight - binding lattice of @xmath27 sites , weakly coupled with strengths @xmath89 to two baths of temperature @xmath90 , as depicted in ( c ) . ( b ) heat flow @xmath91 from bath 2 to bath 1 ( arbitrary units , @xmath92 kept constant ) , the shaded ( unshaded ) area corresponds to @xmath93 ( @xmath75 ) . ( d ) occupations of single - particle floquet states for the tight - binding chain with the coupling to bath 2 replaced by a driving term of strength @xmath94 as depicted in ( e ) . ] let us now investigate the effect of bose selection in the concrete physical model system of a one - dimensional tight - binding lattice . such a model describes , e.g. , an array of josephson junctions , ultracold atoms in optical lattices or vibrons in an ion chain @xcite . on the single - particle level , the lattice sites @xmath95 are coupled by tunneling , @xmath96 with @xmath97 . the eigenstates @xmath10 are delocalized , thus a highly coordinated rate matrix results from coupling a bath to a local operator like @xmath98 . the resulting rate matrix can be derived microscopically within the born - markov approximation @xcite ( see supplemental material for details of the ohmic baths used here and a plot of the rate matrix @xcite ) . in order to achieve bose selection with @xmath87 we consider two baths , as sketched in fig . [ fig : chain]c : one with positive temperature @xmath99 couples with strength @xmath100 to @xmath101 and another one with negative temperature @xmath102 couples with strength @xmath103 to @xmath104 @xcite . here the negative temperature models a bath with occupations that increase with energy . in fig . [ fig : chain]a we plot the mean occupations of the eigenstates of the tight - binding chain versus the relative coupling strength @xmath105 for large filling . one can observe bose selection as a clear separation between highly occupied states on the one hand and states with occupations @xmath106 on the other . for @xmath107 the system is in equilibrium and bose condensation , the selection of a single state , is found . when the coupling to the inverted bath is switched on , at @xmath108 three states become selected ( fig . [ fig : chain]a , shaded area ) . while the data of fig . [ fig : chain]a correspond to @xmath27 , for @xmath109 we have studied also larger systems with up to @xmath110 sites and always found 3 states selected . this suggests , that the model of fig . [ fig : chain]c is an example where , in contrast to the random - rate model , the number of selected states remains of order one ( while still being larger than one ) . this corresponds to a fragmented condensate with a macroscopic occupation of each selected state . as a striking signature for the selection of more than a single state , at the transition from @xmath75 to @xmath93 a significant steady - state heat flow @xmath91 from bath 2 to bath 1 is established abruptly ( fig . [ fig : chain]b ) . the heat flow from bath @xmath111 into the system reads @xmath112 . this explains an increase by orders of magnitude from @xmath64 to @xmath113 when the transition from one to three selected states occurs [ since @xmath114 ( @xmath115 ) for selected ( non - selected ) states ] . thus , the mechanism of bose selection might be used to design quantum devices working far from equilibrium that allow to switch the heat conductivity via the number of selected states . let us now consider time - periodically driven quantum systems ( floquet systems ) with hamiltonian @xmath116 @xcite . when coupled weakly to a thermal bath , these systems can be described within floquet - born - markov theory @xcite . one obtains eqs . ( [ eq : spr ] ) and ( [ eq : mpr ] ) with @xmath10 labeling single - particle floquet states . in the supplemental material we show that the rate differences @xmath117 are independent of the bath - temperature @xcite . according to eqs . ( [ eq : selected ] ) and ( [ eq : nonselected ] ) this implies that the selected states and their relative occupations are temperature independent , whereas the occupations of the non - selected states ( and thus also the crossover density @xmath34 ) are temperature dependent . replacing the energy - inverted bath coupled to one end of the tight binding chain by a coherent periodic driving term @xmath118 with @xmath119 ( fig . [ fig : chain]e ) , we obtain the occupations of the single - particle floquet states versus the driving strength @xmath94 ( fig . [ fig : chain]d ) . in this driven - dissipative system we observe again both bose condensation into a single state which one is controlled by the parameters and bose selection of @xmath93 states . two more examples that emphasize that bose selection is a generic and robust effect in open time - periodically driven systems are given in the supplemental material @xcite : the @xmath0-boson generalizations of the open kicked rotor and the open driven quartic oscillator of ref . @xcite . in figs . [ fig : chain]a and [ fig : chain]d , we can study the evolution of the occupations with respect to a parameter controlling the rate matrix . within the asymptotic theory ( [ eq : selected ] ) and ( [ eq : nonselected ] ) transitions of states between the groups of selected and non - selected states are triggered either by the occupation of a selected state approaching zero or by the occupation of a non - selected state diverging . both requires the fine - tuning of a single parameter . while at the transition point an even number of states is selected , after the transition again the generic situation with an odd number of states has to be recovered . thus , a second state has to make a transition at the transition point , too . when approaching the transition point from the other side , this second state plays the role of the triggering one . one finds three types of two - state processes , examples of which are labeled by i , ii , and iii in fig . [ fig : chain]a and [ fig : chain]d : the transition is triggered from one side by a selected and from the other one by a non - selected state ( i , @xmath30 changes by 2 ) , or the transition is triggered on both sides either by selected ( ii ) or non - selected ( iii ) states ( @xmath30 does not change ) . in future work , it will be interesting to study the impact of , e.g. , dimensionality , particle reservoirs , disorder , and interactions on the effect of bose selection in non - equilibrium steady states . a concrete application of bose selection in a physical system is the quantum switch for heat conductivity proposed here . we warmly thank martin holthaus for valuable discussions initiating this work and henning schomerus for providing a proof for the existence of a unique solution of the asymptotic theory . we acknowledge support through dfg forschergruppe 760 `` scattering systems with complex dynamics '' . 35 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , , , * * , ( ) . , * * , ( ) . , , , , * * , ( ) . d. m. basko , i. l. aleiner , and b. l. altshuler , ann . phys . * 321 * , 1126 ( 2006 ) . d. a. huse , r. nandkishore , v. oganesyan , a. pal , and s. l. sondhi , phys . rev . b * 88 * , 014206 ( 2013 ) . a. eckardt , c. weiss , and m. holthaus , phys . rev . lett . * 95 * , 260404 ( 2005 ) . a. zenesini , h. lignier , d. ciampini , o. morsch , and e. arimondo , phys . rev . lett . * 102 * , 100403 ( 2009 ) j. struck , c. lschlger , r. le targat , p. soltan - panahi , a. eckardt , m. lewenstein , p. windpassinger , and k. sengstock , science * 333 * , 996 ( 2011 ) . , , , * * , ( ) . , ( ) . , , , * * , ( ) . e. j. mueller , t .- l . ho , m. ueda , and g. baym , phys . rev . a * 74 * , 033612 ( 2006 ) . , , , * * , ( ) . , * * , ( ) . , , , * * , ( ) . , * * , ( ) . j. klaers , j. schmitt , f. vewinger , and m. weitz , nature * 468 * , 545 ( 2010 ) . , , , * * , ( ) . , , , , , , , , , * * , ( ) . , , , * * , ( ) . , , , , in _ _ , edited by , , ( , , ) , vol . , p. . , , , , , , * * , ( ) . , , , * * , ( ) . , , , , , , , , , , * * , ( ) . , * * , ( ) . m. r. evans and t. hanney , j. phys . a : math . gen . * 38 * , r195 . o. hirschberg , d. mukamel , and g. m. schtz , phys . rev . lett . * 103 * , 090602 ( 2009 ) . s. grosskinsky , f. redig and k. vafayi , j. stat . phys . * 142 * , 952 ( 2011 ) . a. schadschneider , d. chowdhury , and k. nishinari , _ stochastic transport in complex systems - from molecules to vehicles _ ( elsevier , amsterdam , 2011 ) , chapter 3 . m. r. evans and t. hanney , and s. n. majumdar , phys . rev . lett . * 97 * , 010602 ( 2006 ) . y. schwarzkopf , m. r. evans , and d. mukamel , j. phys . a : math . theor . * 41 * , 205001 ( 2008 ) . s .- w . kim , j. lee , and j. d. noh , phys . rev . e * 81 * , 051120 ( 2010 ) . a. g. thompson , j. tailleur , m. e. cates , and r. a. blythe , j. stat . mech . , p02013 ( 2010 ) . , _ _ ( , , ) . the many - body liouvillian generating the dynamics of @xmath120 is obtained from the single - particle one by replacing operators @xmath121 by fock - space operators @xmath122 with @xmath123 creating a boson in state @xmath10 . while the diagonal terms of @xmath124 solve eq . ( [ eq : mpr ] ) , the off - diagonal terms decay . , * * , ( ) . , * * , ( ) . the mean rate , here the only free parameter , determines the time scale of the evolution , but not the steady state . see supplemental online material . , * * , ( ) . the asymptotic occupation of a non - selected state ( [ eq : nonselected ] ) scales like @xmath125 . here the @xmath126 were first approximated by their average . while in the numerator the sum over rates is then @xmath127 , the denominator is a sum over rate differences of both signs dominated by fluctuations , giving @xmath46 only . residual deviations that are visible for some data points of the non selected states are largely cured when mf theory is augmented by keeping also non - trivial two - state correlations like @xmath128 for @xmath53 , while neglecting non - trivial three - state correlations @xmath129 . the mechanism of bose selection is not altered by such corrections . h. schomerus , private communication . , , , * * , ( ) . the symmetric choice @xmath130 is special and always leads to @xmath75 with either the state of lowest or largest energy selected . , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , , , , * * , ( ) . , , , * * , ( ) . , , , * * , ( ) . , , , * * , ( ) . r. ketzmerick and w. wustmann , phys . rev . e * 80 * , 021117 ( 2009 ) . , * * , ( ) .

one of the successes of the supersymmetric version of the standard model is that it provides a natural candidate for the cold dark matter of the universe . many authors @xcite have lauded this feature of supersymmetry . studies often assume that the lightest supersymmetric particle ( lsp ) is the bino ( @xmath5 ) , the supersymmetric partner of the hypercharge gauge boson , for two reasons . bino annihilation can give about the right relic density . second , minimal supergravity models with gauge coupling unification and with universal scalar and gaugino masses at the unification scale generically produce a bino lsp after the full renormalization group flow of all sparticle masses , yukawas , and gauge couplings @xcite . in general , higgsino - like lsp s annihilate too efficiently to provide cosmologically interesting amounts of cold dark matter . in this paper we are instead motivated primarily by data rather than by general theory , mainly by the @xmath0 event reported by cdf @xcite . this event has two different possible supersymmetric interpretations . one @xcite is that two selectrons are created which ultimately decay into electrons , photons , and a very light gravitino ( less than about @xmath6 ) as would happen in low scale gauge - mediated supersymmetry breaking models . however , these models have not as yet produced a compelling cold dark matter candidate , and so we do not consider the light gravitino interpretation further for this paper . the second supersymmetric interpretation of this event @xcite is based on ordinary gravity communicated supersymmetry breaking where the gravitino is sufficiently heavy to not be the lsp . in this interpretation , the decay chain which produces @xmath0 is e^+ e^- e^+ n_2 e^- n_2 e^+e^- where the photino - like second - lightest neutralino @xmath7 decays radiatively @xcite into the lightest neutralino ( @xmath8 ) and a photon . thus , @xmath8 is the lsp . here , motivated by the desire for a simple notation ( and one suitable for e - mail ) we denote neutralinos by @xmath9 ( and charginos by @xmath10 ) . if the supersymmetric interpretation of the fnal @xmath11 event is correct , then @xmath8 has been observed at fnal . of course , that it escapes the detector only proves it lives longer than @xmath12 sec , so its direct detection would be necessary before it could be finally accepted as the cold dark matter . in this paper we will demonstrate that @xmath8 can be an interesting dark matter candidate , despite the fact that it is higgsino - like , and we also discuss the direct detection prospects for such a particle . our analysis assumes a general low scale supersymmetric lagrangian , with no presumed relations among parameters . no assumptions are made about the exact form of the high scale theory except that it exists perturbatively , and no assumptions are made about common gaugino or scalar masses . rather , we assume that the low energy theory can be described by a superpotential plus general soft - breaking terms . we use the results of ref . @xcite for the mass and coupling requirements of the light supersymmetric states . in order to proceed with a discussion about the dark matter qualities of the @xmath8 lsp , we must discuss its composition and mass . for convention purposes we write down the neutralino mass matrix @xmath13 in the @xmath14 basis . if @xmath15 and @xmath16 is near one , then the two lightest eigenstates of the neutralino mass matrix are @xmath17 ( photino - like ) , and @xmath18 ( higgsino - like ) , where @xmath19 . this arrangement of lightest neutralino mass eigenstates enhances the important radiative neutralino decay @xmath20 , and along with the @xmath21 event of ref . @xcite implies @xmath22 and @xmath23 @xcite . we shall see that with @xmath16 near one the @xmath24 invisible width constraint is satisfied and @xmath8 provides an interesting amount of dark matter as well . from the invisible width determinations at lep , the @xmath24 is not allowed to decay into @xmath25 with a partial width more than about 5 mev ( at @xmath26 ) @xcite . in our approximation , the formula for the partial width is @xmath27 figure [ inv : fig ] has contours of the invisible width in units of mev . = 3.25 in as can be seen from the above equation as long as @xmath16 is close enough to 1 then the invisible width constraint can be satisfied . we think this constraint should not be applied too tightly at this stage since it could be affected by new physics . however , since @xmath28 is the natural region for the radiative decay requirements in ref . @xcite , this constraint does not cause any problems for our analysis here , even if it is applied with the most stringent assumptions @xcite . @xmath8 pairs annihilate through the @xmath24 into fermion pairs . to look at the prediction for @xmath29 , we can expand the thermally averaged annihilation cross section @xcite into fermions @xmath30 in the following way : @xmath31 where @xmath32 and @xmath33 depend only on one unknown , the lsp mass . applying the usual approximation method @xcite to solve the boltzmann equation , the relic abundance can be found : @xmath34 where @xmath35 , @xmath36 and @xmath37 must be solved for self - consistently . calculations such as these could be valid to a factor of two or better . in fig . [ omega : fig ] contours of @xmath29 are plotted in the @xmath38 plane . = 3.25 in since the annihilation cross section is proportional to @xmath39 when @xmath16 gets closer to 1 , @xmath29 begins to exceed one . the @xmath40 channel sfermion exchange is greatly suppressed ( since @xmath8 is mainly higgsino - like ) , but if the @xmath41 fraction of @xmath8 is large enough , then a @xmath40 channel sfermion diagram which couples like the @xmath42 gauge coupling could start to become important . however , this potentially efficient annihilation channel is suppressed by a factor of @xmath43 and @xmath44 is less than about @xmath45 @xcite . comparing @xmath46 with @xmath47 , we find that this channel can compete with the @xmath48 channel @xmath24 exchange only when @xmath16 is less than about @xmath49 . furthermore , as @xmath16 gets closer and closer to 1 , @xmath44 necessarily becomes smaller and smaller , and the number @xmath50 is actually an overestimate . at @xmath51 we find that the annihilation cross section is too small for all values of @xmath52 in fig . [ omega : fig ] , and therefore we set this as our lower bound on allowed @xmath16 in this scenario . consequently , the sfermion annihilation channel is not numerically important . the @xmath48 channel higgs exchange diagrams might possibly play an important role in the annihilation cross section . however , if the pseudo - scalar ( @xmath53 ) is sufficiently heavy then the pseudo - scalar and heavy scalar ( @xmath54 ) higgses will decouple ; in the absence of information about the heavy higgs bosons we assume they effectively decouple . in this limit it can be shown that the @xmath55 also vertex decouples . furthermore , since the lsp is sufficiently light not to annihilate into top quarks , or vector bosons , the light final state fermion masses contribute a further suppression of the @xmath56 mediated annihilation cross section . the allowed mass range for @xmath8 from ref . @xcite overlaps @xmath57 if consistency with the supersymmetry interpretation of the lep @xmath58 excess is required , probably @xmath59 , but it is premature to assume that . if @xmath60 it is necessary to do the resonant calculation very carefully @xcite . there is always a value of @xmath16 for which the curves of fig . 2 continue across @xmath61 smoothly , so we will wait until @xmath52 and @xmath16 are better measured to do the more precise calculations needed . we show results in fig . 2 for @xmath62 which we expect is the most relevant region . given the results of reference @xcite , the only channel that could complicate the simple analysis is coannihilation of the @xmath8 with the @xmath63 @xcite , if @xmath64 ( @xmath63 is the lightest stop mass eigenstate ) . we expect @xmath65 and @xmath66 , so probably this complication can be ignored , but until the masses are better determined it should be kept in mind . the hubble constant @xmath67 is probably between about @xmath68 and @xmath69 . assuming the cold dark matter constitutes 0.4 to 0.8 of @xmath70 , we expect that @xmath71 should lie somewhere between @xmath72 and @xmath68 in fig . [ omega : fig ] ( e.g. @xmath73 ) . we emphasize that fig . 2 follows from the results of ref . @xcite , and that apart from the approximations mentioned above this is a prediction of the supersymmetric interpretation of the cdf event . further , we note that the supersymmetric interpretation of the reported excess of @xmath74 decays at lep leads to the same region of parameters as ref . @xcite , with higgsino - like @xmath8 and with @xmath16 near 1 @xcite , and therefore can conservatively be viewed as consistent with this prediction , or optimistically as additional evidence for its correctness . we have established that the same @xmath8 which is necessary to explain the @xmath0 event at fermilab , and independently the lep @xmath75 excess , is also a viable cold dark matter candidate . future experiments at fermilab and lep will be able to determine if the supersymmetric interpretation of the cdf event is a valid one . however , these colliders can not determine experimentally if @xmath8 particles in fact are stable and comprise a significant portion of the cold dark matter in the universe . in this section we discuss some of the direct detection prospects for this particle . from kinematic analyses of the galactic rotation curves it has been estimated @xcite that the local density of cold dark matter is approximately @xmath76 . ( we will use the lower number in our subsequent calculations . ) several experiments are under way to look for weakly interacting massive particles ( wimps ) floating around our part of the galaxy . neutrino telescopes at amanda , etc . , hope to see the effects of wimp annihilations in the sun . the light higgsino dark matter candidate , which we propose here , would be difficult to detect at the large area neutrino telescopes since the muon threshold energy is about @xmath77 , roughly equivalent to the lsp mass range we are considering . the neutrinos produced by the lsp annihilations in the sun will be at energies below this threshold , and so the converted muons will not be energetic enough to be detected . other experiments @xcite are designed to measure direct annihilations of lsps in the galactic halo by seeing an excess of photons , electrons , protons , etc . in the spectrum . the steeply rising photon background ( as energy decreases ) makes the photon signal difficult to extract , and the broad energy spread of the electron / positron signal for @xmath78 annihilations in the galactic halo also complicates this detection possibility . lastly , numerous table - top experiments @xcite are being set up with the hope of seeing wimps interact with different nuclei . it is this last set of experiments that we wish to focus on in this analysis . we also wish to urge one change in notation , namely that the acronym wisp ( weakly interacting supersymmetric particle ) be used when the particle in question is known to be a possible state following from a supersymmetric lagrangian and consistent with phenomenological constraints ; many wimp s discussed in the literature are not wisp s . since we have argued that the @xmath24 coupling is the most important one , we concentrate on @xmath79 and @xmath80 which are well - suited for spin - dependent scattering of lsp s with nucleons , and are currently being considered by experimentalists for larger scale designs . the spin - dependent cross section in this scenario can be written as @xmath81 ^ 2\ ] ] where @xmath82 is the spin content of the neutron @xcite carried by quark @xmath83 , and @xmath84 is the mass of a silicon atom . from this we can estimate @xcite the rate of interactions per day : @xmath85 where @xmath86 quantifies the nuclear form factor suppression . we are not considering the spin - independent cross section in this analysis since , as we argued above , the couplings of @xmath87 to all scalar particles are very small . it is possible with a lighter pseudo - scalar mass to have larger couplings of @xmath87 to the higgs particles , which would contribute to a spin - independent cross - section , but to be conservative we have assumed that the higgs effects are decoupled . figure [ si29:fig ] is a plot of the event rate per kilogram per day of @xmath8 interacting on @xmath79 and @xmath80 . = 3.25 in the expected sensitivity @xcite for @xmath80 is at about the @xmath88 events contour in the near future , and about @xmath89 in the next round of experiments . thus , the entire region of the plot above about the @xmath89 event contour will soon be probed in the table - top detector . this is a good demonstration of how the table top experiment can sometimes do better than collider limits ( see fig . [ inv : fig ] ) for an interesting part of parameter space . @xmath90ge has a reduced event rate compared to silicon mainly because the nuclear land @xmath91 factor is smaller , and the nucleus mass is heavier . we interpret fig . [ si29:fig ] as implying that si and ge and related detectors may be able to observe a cold dark matter signal in the next round of attempts . it is remarkable that these calculations ( which could have given much smaller or much larger @xmath92 and are essentially free of parameters ) imply a wisp cold dark matter candidate that is cosmologically interesting , possibly detectable using table - top experiments , possibly already observed at fnal , and possibly associated with loop effects seen at lep . we would like to thank s. ambrosanio , b. cabrera , d. caldwell , e. diehl , g. kribs , s. martin and m. worah for useful discussions . this work was supported in part by the u.s . department of energy . for overview and references see for example , m. drees , m. nojiri , ; e. diehl , g.l . kane , c. kolda , j. wells , ; g. jungman , m. kamionkowski , k. griest , hep - ph/9506380 ; r. arnowitt , p. nath , hep - ph/9509260 ; and most recently a. gabutti , m. olechowski , s. cooper , s. pokorski , l. stodolsky , hep - ph/9602432 . r. roberts , g.g . ross , ; r. arnowitt , p. nath , ; s. kelley , j. lopez , d. nanopoulos , h. pois , k. yuan , ; m. carena , s. pokorski , c. wagner , ; g.l . kane , c. kolda , l. roszkowski , j.d . wells , ; d. castao , e. piard , p. ramond , ; v. barger , m. berger , p. ohmann , . a. djouadi ; m. boulware , d. finnell , ; j.d . wells , c. kolda , g.l . kane , ; d. garcia , j. sola , ; g.l . kane , r. stuart , j.d . wells , ; a. dabelstein , w. hollik , w. msle , hep - ph/9506251 ; p. chankowski , s. pokorski , ; j. ellis , j. lopez , d. nanopoulos , hep - ph/9512288 ; e. simmons , y. su , hep - ph/9602267 ; p. chankowski , s. pokorski , hep - ph/9603310 .

motivated by the supersymmetric interpretation of the cdf @xmath0 event and the reported @xmath1 excess at lep , we analyze the higgsino as a cold dark matter candidate . we examine the constraints as implied by the collider experiments , and then calculate its relic density . we find that this higgsino - like lightest supersymmetric particle is a viable cold dark matter candidate ( @xmath2 ) , and we discuss its favorable prospects for laboratory detection . @=11 caption#1[#2]#3 @=12 # 10= 0=0 1= 1=1 0>1 # 1 / .7ex .7ex m_t m_b m_t m_b m_z m_w _ s(m_z^2 ) t_l t_r t_1 t_2 zb| ( zb|b ) ( z ) r_b^ r_b^ r_b ^ ^0 ^2_w _ w _ w _ w m_h^ m_1/2 _ i.e. _ _ et al . _ _ e.g. _ n_1 # 1o(#1 ) # 1m_#1 # 1m_#1 # 1#2#3nucl . phys . * b#1 * ( 19#2 ) # 3 # 1#2#3phys . lett . * b#1 * ( 19#2 ) # 3 # 1#2#3phys . lett . * # 1b * ( 19#2 ) # 3 # 1#2#3phys . rev . * d#1 * ( 19#2 ) # 3 # 1#2#3phys . rev . lett . * # 1 * ( 19#2 ) # 3 # 1#2#3phys . rep . * # 1 * ( 19#2 ) # 3 # 1#2#3ann . rev . . astrophys . * # 1 * ( 19#2 ) # 3 # 1#2#3ann . rev . nucl . part . sci . * # 1 * ( 19#2 ) # 3 # 1#2#3mod . . lett . * a#1 * ( 19#2 ) # 3 # 1#2#3zeit . fr physik * c#1 * ( 19#2 ) # 3 # 1#2#3ap . j. * # 1 * ( 19#2 ) # 3 # 1#2#3mod . . lett . * a#1 * ( 19#2 ) # 3 * higgsino cold dark matter motivated by collider data * g.l . kane@xmath3 and james d. wells@xmath4 + 0.5 cm .5 cm

the magellanic clouds ( large , lmc , and small , smc ) are irregular , gas - rich dwarf galaxies in our local group . the presence of a bar - like structure especially prominent in the lmc suggests that they could have originally been barred spiral galaxies transformed into present - day irregular galaxies due to their tidal interactions with one another and with the milky way . at a distance of 48.5 kpc ( macri et al . 2006 ; freedman et al . 2010 ) , the lmc is the third closest galaxy to the milky way . with an inclination of @xmath9 ( haschke et al . 2011 ) it offers us an excellent view of its structure and content . the smc is located at a distance of @xmath10 kpc ( hilditch et al.2005 ) with an inclination of only @xmath11 ( subramanian et al . 2011 ) . probing the star formation history ( sfh ) in galaxies informs us about galaxy formation and evolution . star clusters revealed differences in the sfh between the lmc and the milky way ( hodge 1960 ; sagar & pandey 1989 ) : the lmc contains a larger fraction of young clusters relative to the milky way . other methods of determining the sfh include a comparison of the observed colour magnitude diagrams ( cmds ) with the model - built cmds ( e.g. , bertelli et al . harris & zaritsky ( 2009 ) used multi - color photometry of millions of stars over @xmath12 deg@xmath13 of the lmc to derive the sfh ; they found multiple episodes of star formation between 5 to 0.1 gyr ago in addition to an initial burst of star formation @xmath14 gyr ago . in a recent study , cignoni et al.(2013 ) studied various fields within the bar and wing of the smc and confirmed a dominant intermediate - age star formation epoch . their derived metallicity for the fields support a well mixed metal content of the smc . the cmd approach has also revealed that the most recent star formation , as young as @xmath15 myr , is confined to the central regions in both the lmc and smc ( indu & subramanian 2011 ) . our approach to investigate the sfh is based on employing long period variable stars ( lpvs ) . these are mostly asymptotic giant branch ( agb ) stars at their very late stage of evolution , as well as more massive red supergiants ( rsgs ) . the agb stars alone have low- to intermediate - mass ranges which relate to diverse ages and hence time epochs in the past , and they can thus be used to derive star formation rates ( sfrs ) over much of a galaxy s cosmological evolution . the agb stars develop unstable mantles which cause radial pulsation at periods of one or a few years ; they are luminous ( @xmath16 l@xmath1 ) and cool ( @xmath17 k ) and hence stand out at near - infrared wavelengths . their colours can be further reddened by enhanced atmospheric opacity due to carbon - dominated chemistry ( in carbon stars ) or due to attenuation by circumstellar dust produced in their winds . thanks to recent large - scale , extended monitoring projects such as ogle - i , ii , iii , macho and others , we are now in a position to use lpvs to study these nearby galaxies . the data and the methodology are described in section 2 . in section 3 we present our analysis and the results . in section 4 we present the sfh derived from our analysis . section 5 presents a discussion of the findings and a comparison with other studies . section 6 contains a summary and conclusions . in this paper , we use the catalogue of lpvs in the lmc from spano et al.(2011 ) and in the smc from soszyski et al . ( 2011 ) . in the case of the lmc , they used the eros-2 survey which started functioning efficiently from 1996 ; it employed two cameras enabling observations at two wavelengths , blue ( be ) and red ( re ) , simultaneously . after seven years of monitoring , it provided a database of 856,864 variables in the lmc of which 43,551 are considered lpvs ( spano et al . the eros-2 survey covered an area of 88 deg@xmath18 on the lmc , which is a superior coverage in comparison with other projects such as macho and ogle - iii which covered areas on the lmc of 13.5 deg@xmath18 and 40 deg@xmath18 , respectively . spano et al . cross - identified the eros-2 lpv candidates with variable stars from the ogle - iii and macho surveys . they found nearly 60% of lpvs in common between eros-2 and macho ; the difference is attributed to differences in sky coverage between the surveys ( spano et al . 2011 ) . for the smc , soszyski et al . used the ogle - iii project which has monitored an area of 14 deg@xmath18 on the smc . after 13 years of monitoring in the optical v- and near - infrared i - band , the ogle survey has provided a database of about 6 million stars in the smc of which 19,384 are identified as lpvs are detected ( soszyski et al . 2011 ) . as mentioned , agb stars become surrounded by dust , which causes reddening and makes them particularly faint in optical bands . therefore , infrared ( ir ) studies are preferred . for a comparison and more detailed study of the core region we also utilise ita et al.s ( 2004a ) data of variable stars in the magellanic clouds . this catalogue had been obtained by cross - identifying the ogle - ii and sirius data ( simultaneously observated in j- , h- , and k - band ) . the total area covered by the sirius survey is 3 deg@xmath18 and 1 deg@xmath18 in the lmc and smc , respectively , covering the core region of these galaxies in particular , the bar of the lmc ( see fig . 1 ) . spano et al . ( 2011 ) cross - matched the eros-2 survey with 2mass near - ir ( and _ spitzer _ mid - ir ) data , within a search radius of @xmath19 . for the smc , we cross - matched the original catalogue of soszyski et al . ( 2011 ) with 2mass , also within a search radius of @xmath19 . ita et al . classified the variable stars into nine groups for the lmc and eight groups for the smc , based on their location in the period k - band magnitude plane . the categories include cepheids , miras , semi regulars , irregular variables and eclipsing binaries ( ita et al . 2004a ) . in total , the number of variable stars in ita s catalogues of the lmc and smc are 8852 and 2927 , respectively . for the purpose of this study , we select lpvs pulsating in the fundamental mode , equal to sequences c and d in ita et al . ( 2004a ) above the ( first ascent ) red giant branch ( rgb ) tip ( @xmath20 and @xmath21 mag for the lmc and smc , respectively ) . spano et al . ( 2011 ) classified their variable stars using period luminosity relations ( @xmath22@xmath23 diagram ) , into six groups ( figs . 22 and 23 in their paper ) , from which sequence c is considered as miras pulsating in the fundamental mode . sequence d comprises mostly of stars with secondary long periods that are probably not due to the same radial pulsation we are after , but it does contain some dusty lpvs that have lengthened pulsation periods as a result of diminished mass ( ita et al.2004b ; spano et al . 2011 ) . for the purpose of this study we select variables belonging to sequence c , and variables with `` only '' one period in sequence d. an appropriate stellar evolution model is provided by the padova group ( marigo et al . lpvs are expected to be at the end - points of the agb ( or rsg ) part of the isochrones in cmds . thus , the padova isochrones were fitted to the cmds of the magellanic clouds in order to correct the reddened lpvs ( see fig.2 ) . two different groups of reddening slopes are seen , those with steeper slopes are associated with those stars surrounded by oxygeneous dust whilst some with shallower slopes are associated with carbonaceous dust . not all stars are reddened ; the correction is applied only to those stars which have @xmath24 mag . the total mass of stars formed between @xmath25 and @xmath26 depends on star formation rate , @xmath27 , as @xmath28 thus , @xmath29 , the number of stars formed , is defined by the following equation : @xmath30\,{\rm d}m(t),\ ] ] in which @xmath31 is the initial mass function . the nominator represents the total number of stars within the mass range of @xmath32@xmath33 , whilst the denominator gives their total mass . therefore , the fraction refers to the number of stars per mass unit . for the initial mass function , the kroupa ( 2001 ) model was used where we assumed @xmath34 m@xmath1 and @xmath35 m@xmath1 . the number of stars formed between @xmath25 and @xmath26 is then obtained as follows : @xmath36\,{\rm d}n(t).\ ] ] because the large - amplitude , long - period variability only comprises a short phase in the evolution of a star lasting @xmath37 , the number of such variable stars that were formed is reduced to @xmath38\,{\rm d}n(t).\ ] ] thus , the final equation for the star formation rate derived from the variable star counts is : @xmath39\,\frac{{\rm d}n^\prime(t)}{\delta t}.\ ] ] for reasons that will become clear below , we convert k - band magnitudes into stellar birth mass , first , before converting that into the age of the star . for the latter , we use the following mass age relation : @xmath40}}\ = \ a\,\log{m\,{\rm [ m}_\odot{\rm]}}+b,\ ] ] where @xmath41 and @xmath42 are given in javadi et al . ( 2011 ) , based on the models from marigo et al . ( 2008 ) . javadi et al . ( 2011 ) give the mass luminosity ( k - band magnitude ) relation as @xmath43}}\ = \ ak+b,\ ] ] where @xmath23 is the ( extinction - corrected ) k - band magnitude and @xmath41 and @xmath42 are defined in table 1 for various ranges of @xmath23 . we use a distance modulus of @xmath44 mag and @xmath45 mag for the lmc and smc , respectively . ccc + @xmath41 & @xmath42 & validity range + @xmath46 & 2.840 & @xmath47 + @xmath48 & 9.391 & @xmath49 + @xmath50 & @xmath51 & @xmath52 + @xmath53 & @xmath54 & @xmath55 + @xmath56 & @xmath57 & @xmath58 + @xmath59 & @xmath60 & @xmath61 + @xmath62 & @xmath63 & @xmath64 + @xmath65 & @xmath66 & @xmath67 + + @xmath41 & @xmath42 & validity range + @xmath68 & 1.1268 & @xmath69 + @xmath70 & @xmath71 & @xmath72 + @xmath73 & @xmath74 & @xmath75 + @xmath76 & @xmath77 & @xmath78 + @xmath79 & @xmath80 & @xmath81 + @xmath82 & @xmath83 & @xmath84 + @xmath85 & @xmath86 & @xmath87 + @xmath88 & @xmath89 & @xmath90 + @xmath91 & @xmath92 & @xmath93 + the light from the variable stars can be attenuated by interstellar and/or circumstellar dust ; this is wavelength dependent resulting also in reddening of the near - ir colours . to determine the intrinsic k - band magnitude , the photometry needs to be de - reddened . the ( de-)reddening slope in the cmd depends on whether the dust is oxygenous or carbonaceous in composition . stars with a birth mass in the range of 1.54 m@xmath1 are expected to have become carbon stars due to the third dredge - up of nculear - processed material material with a carbon : oxygen ratio in excess of unity . in stars with a birth mass @xmath94 m@xmath1 , third dredge - up is not sufficiently efficient , whilst , for stars with a birth mass @xmath95 m@xmath1 nuclear burning of the carbon at the bottom of the convection zone prevents the surface to be enriched in carbon . the reddening correction equation is @xmath96 where @xmath41 is the average slope for each dust type . the procedure followed was to first apply a correction assuming carbonaceous dust . if the mass resulting from this corrected k - band magnitude ended up in the range 1.54 m@xmath1 then the star was treated as a carbon star . otherwise , the correction was replaced by one assuming oxygenous dust , and a mass was derived accordingly . figure 3 shows the mass histogram for different choices of reddening correction . to appreciate the impact of these choices on the derived sfh , we also applied the reverse approach ; assuming that stars are oxygen - rich unless their masses would fall within 1.54 m@xmath1 . ( 5 ) depends on a correction factor ( @xmath37 ) for the duration of the evolutionary phase in which the star exhibits large - amplitude , long - period pulsations . generally , since a massive star evolves faster , its pulsation duration is shorter . thus , low mass stars are more likely to be identified as variables , in comparison to more massive stars . however , the pulsation phase depends sensitively on the star s effective temperature which , in turn , depends on metallicity , mass loss , et cetera . ccccc + @xmath97 & @xmath98 & @xmath41 & @xmath42 & @xmath99 + @xmath100 & 1 & 2.34 & 1.281 & 0.378 + & 2 & 1.32 & 0.460 & 0.165 + & 3 & 0.38 & 0.145 & 0.067 + + @xmath97 & @xmath98 & @xmath41 & @xmath42 & @xmath99 + @xmath101 & 1 & 2.29 & 1.217 & 0.408 + & 2 & 0.84 & 0.524 & 0.093 + & 3 & 0.87 & 0.206 & 0.088 + the mass pulsation relation from javadi et al . ( 2011 ) is used to connect the birth mass of stars to their pulsation duration . relative to its age , it is parameterised as @xmath102}}-b_i\right)^2}{2c_i^2}\right],\ ] ] where @xmath41 , @xmath42 , @xmath99 and @xmath97 are defined in table 2 . given the above argument , the star formation rate estimated using lpvs is shown to be underestimated by a factor of 10 ( javadi et al . 2013 ) in comparison to the star formation rate obtained using @xmath103 emission , for instance . this inaccuracy was exposed by a mismatch between the integrated mass loss and birth mass . to reconcile both the mass loss budget and star formation rate with independent evidence , the pulsation duration adopted from the marigo et al . ( 2008 ) models needs to be decreased by a factor 10 . this is a fairly robust result and we thus apply the correction in our work . given the uncertainties in the models , in particular with regard to the duration of the lpv phase , the star formation rates and stellar masses formed are possibly uncertain by up to a factor of a few . we thus derive the sfrs as a function of time , where we group the stars into bins of equal numbers i.e. , equal poissonian significance . figure 4 represents the sfhs for both reddening correction approaches ( see section 3.1.1 ) ; black lines indicate the case in which a carbonaceous - dust correction is first applied , and red lines demonstrate the reverse however , there is no noticeable difference in the sfrs between these two correction approaches . horizontal lines show time bins while vertical lines demonstrate statistical error bars derived as following ; @xmath104 where `` n '' is the number of stars in each age bin . for the lmc ( fig . 4 , left panel ) we find an ancient star formation episode with rate of @xmath105 m@xmath1 yr@xmath2 @xmath106 gyr ago ( @xmath107 ) , corresponding to the formation of the lmc : the total stellar mass produced in the lmc is @xmath108 m@xmath1 , of which @xmath109 per cent was formed during the first epoch . for the smc ( fig . 4 , right panel ) , two formation epochs are observed ; one with a sfr of @xmath110 m@xmath1 yr@xmath2 rate @xmath5 gyr ago ( @xmath111 ) and a secondary star formation episode @xmath7 gyr ago ( @xmath112 ) with a similar sfr of @xmath113 m@xmath114 yr@xmath2 . these findings are in good agreement with recent studies , e.g. weisz et al . ( 2013 ) . of a total stellar mass in the smc of @xmath115 m@xmath1 , @xmath116 per cent was produced during the first peak in sfr , and @xmath117 per cent during the second peak . before investigating the sfh in the central regions of the lmc and smc , and comparing it to the sfh in their peripheries , we consider the use of the ita et al . ( 2004a ) catalogue for the central regions . figure 5 shows the normalised k - band luminosity distributions . the peak at the brightest k - band magnitudes corresponds to the agb ; the second peak starts at @xmath20 mag for the lmc and @xmath118 mag for the smc , which corresponds to the rgb tip . the deeper ita et al . catalogue also contains fainter stars including cepheids . for the purpose of our study , based mainly on agb stars ( and small numbers of rsgs ) there is no indication that the ir catalogue of ita et al . of the lmc variables offers any major advantage over the spano et al . ( 2011 ) catalogue ( we remind the reader that the latter is originally an optical catalogue cross matched with 2mass ) . on the other hand , in the central part of the smc , the ita et al . catalogue is preferred over that of soszyski et al . ( 2011 ) as the former is much more complete for the agb stars . the sfh of the lmc was derived from the ita et al . ( 2004a ) catalogue , and for the overlapping region with the spano catalogue ( left panel in figure 6 ) . this area corresponds to the bar structure of the lmc ( see fig . 1 ) . for a meaningful comparison with the literature we report the sfr per physical area . the sfr in the lmc bar varies from @xmath119 to @xmath120 m@xmath1 yr@xmath2 kpc@xmath121 . two formation epochs are seen ; one epoch @xmath106 gyr ago ( @xmath122 ) when almost 48 per cent of the total stellar mass within the bar was formed , and a more recent epoch starting @xmath123 gyr ago ( @xmath124 ) which lasted until as recent as 500 myr ago ( @xmath125 ) , within which almost 30 per cent of the total stellar mass was formed ( the total stellar mass in the covered area is @xmath126 m@xmath1 ) . these two estimates are consistent between the use of both catalogues , except for the ancient epoch in which a higher sfr is found for the spano data . however , it appears as though the sfr based on spano et al . ( 2011 ) is more successful in recovering the ancient star formation . cioni et al . ( 2014 ) also found the star formation history of the lmc using vista survey of the magellanic clouds ( vmc ) which reported an old star formation at around @xmath127 as well as a secondary peak of star formation at @xmath128 for stars closer to the lmc center . these two peaks in sfr coincide with the peaks that we have determined , though we argue for a more sustained star formation over the most recent few gyr . the sfh of the smc based on the ita et al . ( 2004a ) catalogue was also derived and compared with that derived for the overlapping region in the soszyski et al . ( 2011 ) catalogue ( right panel in figure 6 ) . this corresponds to the central , approximately square - degree region of the smc ( see fig . 1 ) . the sfr varies from @xmath129 m@xmath1 yr@xmath2 kpc@xmath121 to @xmath130 m@xmath1 yr@xmath2 kpc@xmath121 . as in other studies ( e.g. , cignoni et al . 2012 ) , two formation episodes can be identified ; one episode @xmath5 gyr ago ( @xmath131 ) when almost 49 per cent of the total stellar mass was formed , and another one @xmath132 myr ago ( @xmath133 ) within which around 12 per cent of the total stellar mass was formed . as for the lmc , the sfrs obtained for both catalogues agree rather well , however , the ita et al . data indicate a relatively more pronounced secondary ( recent ) episode of star formation . by excluding the stars in the bar / central regions of the galaxies ( amounting to 3 deg@xmath18 and 1 deg@xmath18 in the lmc and smc , respectively ) , we also obtained the sfh for the outer regions . conversely , we also isolated the sfh of the bar / central regions by subtracting that derived for the immediately surrounding regions ( which could also be seen in projection against and/or mixed with any distinct sfhs belonging to the bar / central regions ) . figure 7 shows a comparison between the sfhs derived for the different regions . in case of the lmc , the second peak of star formation is not seen in the outer region whilst for the smc , another remarkable peak of star formation is still observed in outer region . the latter is in good agreement with the work by ripepi et al . ( 2014 ) who , using the optical step survey , found regions of high levels of star formation in the north and in the wing of the smc . figure 8 shows the ratio of the sfr in the central regions compared to that in the outer regions . each bin in the histograms contain an equal number of stars . note also the logarithmic scale . the lmc shows a steady concentration of younger stars in the central regions . a similar result was reported by meschin et al . ( 2014 ) , who found an ancient star formation epoch dating back 10 gyr , followed by more recent star formation activity 41 gyr ago . they also found younger stellar populations to be more centrally concentrated ; in other words , the stellar population ages as one moves away from the center of the galaxy . any such trend in the smc is less obvious , though might also be present . dynamical effects will have resulted in the migration of old stars towards the outskirts , so it is impossible to tell on the basis of this work whether star formation activity has always been centrally concentrated , or whether it has progressed further inwards as the galaxies matured . as the dynamical timescale is not much more than 100 myr , stars that were formed more than a gyr ago will have equilibrated their kinematics with the gravitational potential well unless they have been exposed to recent tidal disturbances . we find a significant difference in the ancient sfh of the lmc and the smc . for the lmc , the bulk of the stars formed @xmath134 gyr ago , while the strongest episode of star formation in the smc occurred a few gyr later . this is consistent with the findings of weisz et al . ( 2013 ) obtained through cmd analysis of optical ( hubble space telescope ) photometry obtained in a few small regions in the centres and outskirts of the magellanic clouds . they argued that the relative enhancement in the sfh of the smc is a consequence of the interaction with the lmc , with a stronger impact on the global star formation of the smc due to its much smaller mass compared to the lmc . the smaller size of the smc may also have resulted in the slower assembly at early times , similar to what was found in m33 ( javadi et al . this would be easier to understand if the smc and lmc did not form as a pair , but at least the smc was formed in isolation . a secondary peak of the sfh for the central part of the lmc has been reported in other studies , e.g. , smecker - hane et al . ( 2001 ) who found a dominant star formation activity within the bar of the lmc at intermediate ages . having isolated the sfh of the bar of the lmc from that of the disc , it has become clear that the bar formed a few gyr ago and has maintained its star formation until very recently ( @xmath135 myr ) . however , the secondary peak in sfr in the smc is much sharper ( in duration ) and occurred @xmath136 myr ago . there is indeed a similar peak noticeable in the sfh of the lmc , and this is thus distinct from the star formation activity related to the bar itself . the co - eval 700-myr bursts in the smc and lmc are arguably due to the tidal interaction between the magellanic clouds and possibly their approach of the milky way ( cf . bekki & chiba 2005 ; indu & subramaniam 2011 ) . the lmc catalogues we have worked with are limited to @xmath137 myr due to the lack of more massive stars ( brighter stars ) because of saturation of the detectors used in the surveys . however , ulaczyk et al . ( 2013 ) presented a catalogue of brighter stars that had been saturated in the main ogle catalogue . on the basis of this bright sample we derive the recent sfh in the lmc using the methods described above ( fig . we find a sfr of @xmath138 m@xmath1 yr@xmath2 at @xmath139 myr , i.e. similar to that over the past gyr or so ( compare , for instance , with figure 4 ) . after applying a correction to the pulsation duration of lpvs as predicted by the stellar evolutionary models ( see javadi et al . 2013 ) , the resulting sfrs are now in very good agreement with those obtained in other studies ( e.g. , harris & zaritsky 2004 , 2009 ; cignoni et al . however , a recent investigation by the padova group of their own models revealed a severe over - estimation of the lifetimes of thermal - pulsing agb stars ( girardi et al . 2013 ) which could also have implications for the duration of the radial pulsation phase . indeed , the techniques we have pioneered in m33 ( javadi et al . 2011 , 2013 ) , and now applied in the magellanic clouds , offer a promising avenue towards exposing the assembly and development of galaxies as well as improving models for the late stages of stellar evolution and mass loss . we wish to thank spano et al . , soszyski et al . and ita et al . for making their catalogues of variable stars public this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation . antoniou v. , zezas a. , hatzidimitriou d. , kalogera v. , 2010 , aspc , 424 , 230 bekki k. , chiba m. , 2005,mnras , 356 , 680 bekki k. , 2012 , mnras , 422 , 1957 bertelli g. , mateo m. , chiosi c. , bressan a. , 1992 , apj , 388 , 400 besla g. , kallivayalil n. , hernquist l. , van der marel r.p . , cox t.j . , kere d. , 2012 , mnras , 421 , 2109 cioni m .- r.l . , girardi l. , marigo p. , habing h.j . , 2006 , a&a , 448 , 77 cioni m .- r.l . , et al . , 2014 , a&a , 562a , 32 cignoni m. , et al . , 2012 , apj , 754 , 130 cignoni m. , et al . , 2013 , apj , 775 , 83 dolphin a.e . , 1999 , arxiv : 9910524v1 freedman w.l . , kennicutt r.c . , mould j.r . , 2010 , hia , 15 , 1 gieren w. , et al . , 2005 , apj , 627 , 224 girardi l. , marigo p. , bressan a. , rosenfield p. , 2013 , apj , 777 , 142 glatt k. , grebel e.k . , koch a. , 2010 , a&a , 517 , 50 harris j. , zaritsky d. , 2004 , aj , 127 , 1531 harris j. , zaritsky d. , 2009 , aj , 138 , 1243 haschke r. , grebel e.k . , duffau s. , 2011 , aj , 141 , 1865 holtzman j.a . , et al . , 1999 , aj , 118 , 2262 indu g. , subramaniam a. , 2011 , a&a , 535 , a115 ita y. , et al . , 2004a , mnras , 347 , 720 ita y. , et al . , 2004b , mnras , 353 , 705 javadi a. , van loon j.th . , mirtorabi m.t . , 2011 , mnras , 414 , 3394 javadi a. , van loon j. th . , khosroshahi h. , mirtorabi m. , 2013 , mnras , 432 , 2824 kerber l.o . , girardi l. , rubele s. , cioni m .- r.l . , 2009 , a&a , 499 , 697 luri x. , gomez a.e . , torra j. , figueras f. , mennessier m.o . , 1998 , a&a , 335 , 81 macri l.m . , stanek k.z . , bersier d. , greenhill l.j . , reid m.j . , 2006 , apj , 652 , 1133 marigo p. , girardi l. , bressan a. , groenewegen m.a.t . , silva l. , granato g.l . , 2008 , a&a , 482 , 883 maschberger t. , kroupa p. , 2011 , mnras , 411 , 1495 meschin i. , et al . , 2014 , mnras , 438 , 1067 monelli m. , et al . , 2011 , eas , 48 , 43 nanni a. , bressan a. , marigo p. , girardi l. , 2013 , mnras , 434 , 2390 nidever d.l . , et al . , 2011 , apj , 733 , l10 ripepi v. , et al . , 2014 , mnras , 442 , 1897 rubele s. , girardi l. , kozhurina - platais v. , goudfrooij p. , kerber l. , 2011 , mnras , 414 , 2204 rubele s. , et al . , 2013 , mnras , 430 , 2774 smecker - hane t.a . , cole a. , gallagher j.s . , stetson p. , 2002 , apj , 566 , 239 soszyski i. , et al . , 2011 , aca , 61 , 217 spano m. , et al . , 2011 , a&a , 536 a , 60 subramanian s. , subramaniam a. , 2011 , asinc , 3 , 144 ulaczyk k. , et al . , 2013 , aca , 63 , 159 van loon j.th . , zijlstra a.a . , whitelock p.a . , waters l.b.f.m . , loup c. , trams n.r . , 1997 , a&a , 325 , 585 van loon j.th . , et al . , 2008 , a&a , 487 , 1055 van loon j.th . , oliveira j.m . , gordon k.d . , sloan g.c . , engelbracht c.w . , 2010 , aj , 139 , 1553 weisz d.r . , et al . , 2013 , mnras , 431 , 364 whitelock p.a . , feast m.w . , van loon j.th . , zijlstra a.a . , 2003 , mnras , 342 , 86 wood p.r . , whiteoak j.b . , hughes s.m.g . , bessell m.s . , gardner f.f . , hyland a.r . , 1992 , apj , 397 , 552

we present the first reconstruction of the star formation history ( sfh ) of the large and small magellanic clouds ( lmc and smc ) using long period variable stars . these cool evolved stars reach their peak luminosity in the near - infrared ; thus , their k - band magnitudes can be used to derive their birth mass and age , and hence the sfh can be obtained . in the lmc , we found a 10-gyr old single star formation epoch at a rate of @xmath0 m@xmath1 yr@xmath2 , followed by a relatively continuous sfr of @xmath3 m@xmath1 yr@xmath2 , globally . in the core of the lmc ( lmc bar ) , a secondary , distinct episode is seen , starting 3 gyr ago and lasting until @xmath4 gyr ago . in the smc , two formation epochs are seen , one @xmath5 gyr ago at a rate of @xmath6 m@xmath1 yr@xmath2 and another only @xmath7 gyr ago at a rate of @xmath8 m@xmath1 yr@xmath2 . the latter is also discernible in the lmc and may thus be linked to the interaction between the magellanic clouds and/or milky way , while the formation of the lmc bar may have been an unrelated event . star formation activity is concentrated in the central parts of the magellanic clouds now , and possibly has always been if stellar migration due to dynamical relaxation has been effective . the different initial formation epochs suggest that the lmc and smc did not form as a pair , but at least the smc formed in isolation . [ firstpage ] stars : evolution stars : luminosity function , mass function magellanic clouds galaxies : star formation galaxies : stellar content galaxies : structure

recently , solitons on a finite background , generated by the fundamental nonlinear schrdinger ( nls ) equation with self - focusing , which have been known for a long time , have drawn renewed interest as prototypes of rogue waves in the ocean kharif_book , kharif , slunyaev , zakharov , ruban , andonowati , optical fibers solli1,akhmediev3,natphys , kibler1,hammani , erkintalo , plasmas @xcite and other nonlinear media . nls solutions for solitons on a finite background can be identified as the kuznetsov - ma soliton ( kms ) @xcite , which is localized in the transverse direction , the akhmediev breather ( ab ) @xcite , localized in the longitudinal direction , and the peregrine solution ( ps ) @xcite , which is a limit case of the kms and ab , being localized in both the transverse and longitudinal directions . the spatiotemporal localization predicted by the ps solution , and dynamical manifestations of the kms have been demonstrated in optical fibers natphys , kibler1 . splitting of the ps produced by deformed initial conditions , and the related modulational instability have been observed too @xcite . the dynamics accounted for by the ab was first observed in ref . @xcite and further analyzed in ref . collisions between abs @xcite , their dynamics in optical fibers with a longitudinally tailored dispersion @xcite , and shaping of an optical frequency comb into a generator of rogue waves @xcite have been demonstrated too . in these experiments , the abs and kmss were excited by weakly and strongly modulated continuous waves ( cws ) , respectively . generally , nls solutions initiated by perturbations added to a finite background feature strong temporal compression of emerging peaks @xcite . in optical systems , such dynamical scenarios can be used for the generation of supercontinuum , and of high - power ultrashort pulses and pulse trains dudley , solli3,mussot , dudley1,taki . to create high - repetition - rate and high - quality pulse trains , the delay - line interferometer was used to eliminate the finite background around the ab and allow doubling of the repetition rate of the pulse trains @xcite . a possibility of spontaneous transformation of abs into solitons and ultrashort pulses generated by the modulation instability have been considered too @xcite . extraction of high - power optical pulses from the ps by means of the spectral - filtering method was proposed in refs . . however , these results are only suitable for picosecond pulses with the width @xmath0 ps kodama , agrawal . in the present paper , using soliton solutions on the finite background for the hirota equation , we construct high - power strongly compressed pulse trains induced by modulated cws in the subpicosecond regime , an appropriate model for which must include higher - order terms , added to the usual nls equation . the paper is organized as follows . in sec . ii , the hirota equation , which models the pulse propagation in the nonlinear fiber , is presented , and solutions for solitons on the finite background are recapitulated for this equation . in the same section , using ab and kms solutions for the hirota equation , high - power compressed pulse trains induced by modulated cws are demonstrated in the model with the higher - order terms , making use of juxtaposition of two solutions with a half - period temporal shift and opposite signs . the influence of the raman effect on the high - power pulse trains is considered in sec . conclusions are summarized in sec . as said above , the standard nls equation is valid as the model for picosecond pulses with the width @xmath0ps , while in the subpicosecond or femtosecond regime higher - order terms should be included , to account for the third - order dispersion ( tod ) , self - steepening , and the raman effect kodama , agrawal . this accordingly modified nls equation is@xmath1 . \label{model}\end{gathered}\]]here @xmath2 is the slowly varying envelope of the electric field , @xmath3 , where @xmath4 and @xmath5 are the temporal variable and propagation distance , and @xmath6 is the group velocity . coefficients @xmath7 , @xmath8 , @xmath9 , @xmath10 and @xmath11 account for the second - order group - velocity dispersion ( gvd ) , tod , strength of the kerr nonlinearity , nonlinear dispersion , and the raman time - delay constant , respectively . equation ( [ model ] ) does not include the fiber loss , assuming , as usual , that it may by compensated by gain kodama , agrawal . at the end of sec . iii , we produce results of the analysis of the model which explicitly includes the loss , the conclusion being that the difference is relatively small except for a decreasing of peak power . by means of rescaling , @xmath12 , @xmath13 , and @xmath14 , with the temporal scale @xmath15^{1/2}$ ] and dispersion length @xmath16 , ( [ model ] ) is transformed into @xmath17where @xmath18 is @xmath19 and @xmath20 for the anomalous and normal gvd , respectively , and the other coefficients are @xmath21which , as well as their unscaled counterparts , are related to the tod , self - steepening , nonlinear dispersion , and the raman effect , respectively . in its general form , eq . ( [ he ] ) is not integrable . if the raman effect is omitted , i.e. , @xmath22 , and additional conditions @xmath23are satisfied , eq . ( [ he ] ) becomes the integrable hirota equation hirota . here we only consider the case of the anomalous gvd , i.e. @xmath24 . in this case , the solution of the hirota equation for a soliton placed on top of a finite background can be compactly written as ygy , lishuqing , ankiewicz@xmath25-\cosh ( b\xi ) } \right ] e^{i\xi } , \label{ab}\]]@xmath26where @xmath27 is a real constant parametrizing this class of the solutions kibler1 . when @xmath28 , solution ( [ ab ] ) goes over into its nls counterpart @xcite . for @xmath29 , constants @xmath30 , @xmath31 and @xmath32 in eq . ( [ ab ] ) are real . in this case , the solution describes the ab , and the maximally compressed pulse train is attained at point @xmath33 , in the form of @xmath34with period @xmath35 and background power @xmath36which corresponds to @xmath37 . the analysis reported in ref . @xcite has demonstrated that , in the range of @xmath38 , the evolution of the maximally compressed pulse train can be frozen with the help of a specifically tailoring the gvd profile , which leads to formation of a quasi - stable train of fundamental solitons , while at @xmath39 a delay - line interferometer can be used to create a high - power pulse train with the double repetition rate and zero background , as shown in ref . a natural question is whether such a double - repetition - rate high - power train can stably propagate in the framework of the fiber model . to address this issue , it is necessary to choose suitable initial conditions , which can excite a maximally compressed pulse array similar to the ab profile ( peak ) . it has been shown that such a maximally compressed pulse train can be excited by a modulated cw @xcite , corresponding to the initial condition@xmath40where @xmath41 is a small modulation amplitude , and @xmath30 is the modulation frequency , taken as per eq . ( [ abc ] ) . km , juxtaposed with its counterpart with the opposite sign , delayed by a half a period . here the blue dotted and green dashed - dotted curves depict , respectively , the real and imaginary parts of the numerical result , while the black solid curve shows the exact solution , which actually overlaps with the blue dotted curve . ( b ) the intensity distribution of the resulting double pulse train , produced as the superposition of the two original trains . ( c ) the evolution of the double train with zero background . ( d ) the evolution of the peak power of the double train with zero background . these results correspond to @xmath42 in eq . ( [ abc ] ) and @xmath43 in eq . ( [ ab_initial]).,width=321 ] we numerically simulated the integrable hirota equation with the initial modulated cw state ( [ ab_initial ] ) , where the system parameters are adjusted to the smf-28 fiber with @xmath7 @xmath44 ps@xmath45km@xmath46 , @xmath47 ps@xmath48km@xmath46 , and @xmath49 w@xmath46km@xmath46 around @xmath50 nm @xcite , and the initial power @xmath51 w. the simulations are necessary , as generic solutions of equations can not be obtained in an explicit analytical form . in fact , high - power pulse trains growing from the modulated cw do not necessarily reduce to exact breather solutions of the hirota equation , such as kms @xcite . in the present case , the normalized parameters are @xmath52 @xmath53 ps , @xmath54 km , and @xmath55 , respectively . the simulation demonstrates that the maximally compressed pulse train is attained at @xmath56 km , as shown in fig . [ fig1](a ) . at this position , we superpose the train and its counterpart with the opposite sign , delayed by half a period , to produce a double pulse train , which is displayed in fig . the procedure can be experimentally realized by using a delay line and a piezo - electrical device to impose the necessary delay of the pulse train and the @xmath57 phase - shift , as shown in ref . @xcite . then , we let the resulting pulse train propagate in the fiber , see the corresponding stable regime in fig . [ fig1](c ) . unlike its exact counterpart , the numerical solution at the point of the strongest compression includes an imaginary part [ shown by the green dashed - dotted curves in fig . [ fig1](a ) ] , whose maximum value is @xmath58 of the amplitude in the real part , which is a small perturbation that was included to test the stability of the propagation of the pulse train with the zero background . because the soliton number , @xmath59^{1/2}$ ] , takes value @xmath60 for the present parameter set , which exceeds @xmath61 , individual pulses in the resulting train , with peak power @xmath62 and full - width at half - maximum ( fwhm ) @xmath63ps , exhibit oscillations , as shown in fig . [ fig1](d ) . we stress that the oscillatory regime is a dynamically stable one , as is clearly shown by the simulations . next , we turn to the case of @xmath64 , which corresponds to the high background power , as per eq . ( [ bg ] ) . in this case , eq . ( [ abc ] ) yields imaginary @xmath30 and @xmath31 , converting the hyperbolic and trigonometric functions in eq . ( [ ab ] ) into trigonometric and hyperbolic ones , respectively , thus swapping localization and periodicity features of the solution . it now represents a bright oscillating pulse propagating on top of a finite background , which is generally classified as kms . accordingly , the maximally compressed pulse is attained at @xmath33@xmath65with @xmath66 . this pulse has a single peak at @xmath67 [ @xmath68 , and background value @xmath69 as @xmath70 \label{ma - initial}\]]with @xmath71/\cosh ^{-1}\{1[(4a-2)/(\sqrt{4a-2\sqrt{2a}+1}-1)-1]/\sqrt{2a}% \}$ ] and @xmath72 , which has the peak power and fwhm identical to those of the intensity profile , @xmath73/[\sqrt{% 2a}\cosh ( \widetilde{\omega } ( \tau + c\pi /\widetilde{b}))+1]\}^{2}$ ] , at the point of the minimal intensity , @xmath74 , where @xmath75^{1/2}$ ] . km and its counterpart with the opposite sign and temporal delay @xmath76 . ( b ) the intensity distribution of the train produced by the juxtaposition of both original trains . ( c ) the evolution of the resulting pulse train with zero background . ( d ) the evolution of the peak power of the resulting pulse train with zero background . here @xmath77 . the notation adopted in the panels is the same as in fig . , width=321 ] starting from input ( [ ma - initial ] ) , we have simulated , similar to what is shown above in fig . [ fig1 ] , the evolution produced by the juxtaposition of the maximally compressed pulse trains and its counterpart with the opposite sign and half - a - period delay . the results , summarized in fig . [ fig2 ] , show that , for our choice of the parameters , the maximally compressed pulse train is attained at @xmath78 km [ fig . [ fig2](a ) ] , with individual pulses in the train being almost identical to the profile of the maximally compressed pulse given by eq . ( [ ma - pulse ] ) . thus , combining the pulse train with its counterpart with the opposite sign and time delay @xmath79 , one can construct a double pulse train with zero background [ see fig . [ fig2](b ) ] . it propagates stably , exhibiting oscillatory behavior of the pulses , as the respective soliton number is @xmath80 , with fwhm @xmath81 ps , which again exceeds @xmath61 , see figs . [ fig2](c ) and ( d ) . similarly , unlike the exact solution , the numerical result includes an imaginary part [ see the green dash dotted curves in fig . [ fig2](a ) ] , with maximum value @xmath82 of the amplitude of the real part , which does not affect the stable propagation of the pulse train with zero background . finally , it should be pointed out that , in the case of @xmath83 , solution ( [ ab ] ) of the hirota equation reduces to the ps in the form of @xmath84 e^{i\xi } , \]]which is a superposition of the cw solution and a rational fraction function , forming a maximally compressed pulse at @xmath33 peregrine , ygy1 . this implies that the peregrine rogue wave , i.e. , the maximally compressed pulse , can be excited by a small localized ( single - peak ) perturbation placed on top of the cw background @xcite . thus , the high - power pulse extracted from the peregrine rouge wave can be realized by means of the spectral - filtering method , displaying a breather - like behavior @xcite . in the previous section , we have considered the generation and propagation of the pulse trains growing from the cws as solutions of the hirota equation , i.e. , eq . ( [ he ] ) under the integrability relations ( [ int ] ) , while the raman effect was not be included ( @xmath22 ) . this model does not precisely apply to real optical fibers , but the exact solutions provided by it help one to understand the pulse - train dynamics in more general models . in this section , we report results of simulations of eq . ( [ he ] ) with realistic values of parameters ( [ alpha ] ) , including a nonzero raman coefficient , and without imposing integrability conditions ( [ int ] ) : @xmath85 , @xmath86 , @xmath87 is determined by the delayed - nonlinear - response parameter @xmath88 , and @xmath89 for the usual value of the raman time constant , @xmath90 fs . actually , @xmath10 is a small quantity which has little effect on simulation results , therefore we set @xmath91 . thus , we focus on the influence of the raman effect on the dynamics of the high - power pulse trains growing from the modulated cws introduced as per eqs . ( [ ab_initial ] ) and ( [ ma - initial ] ) . in the analysis presented above the modulation amplitude @xmath92 and frequencies @xmath30 and @xmath93 in eqs . ( [ ab_initial ] ) and ( [ ma - initial ] ) were taken as functions of parameter @xmath27 , making it possible to compare the results with the exact solutions produced by the hirota equation , as shown in fig . [ fig1](a ) and fig . [ fig2](a ) . however , in the realistic fiber - optic model , the modulation amplitudes and frequencies can take arbitrary values . here , we use inputs given by eqs . ( [ ab_initial ] ) and ( [ ma - initial ] ) with different values of the parameters , aiming to excite the maximally compressed pulse trains , and perform subsequent simulations of the corresponding double pulse trains in the framework of eq . ( [ he ] ) . ) . ( b ) the corresponding power spectra . ( c ) the evolution of the peak power . here the parameters are @xmath94 , @xmath95 , @xmath90 fs , and @xmath96.,width=321 ] , but for the pulse train generated by the input in the form of eq . ( ma - initial ) . the parameters are @xmath97 , @xmath98 , @xmath90 fs , and @xmath99,width=321 ] fig . [ fig3 ] shows the evolution of the train initiated by input ( ab_initial ) with @xmath100 and @xmath101 . in fig . [ fig3](a ) one can see that the modulated cw initially undergoes strong temporal compression and increase of the peak power , attaining the maximally compressed state of the train at @xmath102 km . at that stage , the double pulse train is constructed , using the delay - line interferometer ( the method is outlined in the previous section , see also ref . @xcite ) , which robustly propagates along the fiber . fig . [ fig3](b ) displays the concomitant evolution of the pulse - train s power spectrum , showing that the spectrum initially expands until the point of the maximum compression , at which the spectral distribution is reduced to a half of the original spectrum @xcite . subsequently , the spectral width remains essentially constant , while the frequencies feature the red shift induced , as usual , by the raman effect @xcite . for the presently chosen parameters , the frequency shift of the train is about @xmath103 thz at @xmath104 km . in addition , the evolution of the peak power , which exhibits an oscillatory behavior , is displayed in fig . [ fig3](c ) . [ fig4 ] displays the evolution initiated by the input given by eq . ( [ ma - initial ] ) with @xmath97 and @xmath98 . this scenario exhibits three differences in comparison with fig . one is that the maximally compressed pulse train is attained at a shorter distance , i.e. , @xmath105 km , as seen in fig . [ fig4](a ) . the other is that the initial spectrum is narrower , while the raman - induced red frequency shift is larger , as seen in fig . [ fig4](b ) , in which the frequency shift of the train is about @xmath106 thz at @xmath104 km . the third difference is a higher peak power , as seen in fig . [ fig4](c ) . km in the fiber , versus the modulation intensity @xmath107 or @xmath92 , with the black solid and blue dotted curves pertaining to inputs ( ab_initial ) and ( [ ma - initial ] ) , respectively . here the parameters are @xmath108 , @xmath90 fs , and @xmath109,width=321 ] to summarize the dynamical results outlined above , in fig . [ fig5 ] we present the largest - compression position @xmath110 for the pulse trains , their peak power , the fwhm of the central pulse in the double pulse trains constructed with the help of the delay - line interferometer method , and the corresponding frequency shift accumulated after the @xmath104 km long propagation , as functions of parameters @xmath41 and @xmath92 in inputs ( ab_initial ) and ( [ ma - initial ] ) , respectively . in fig . [ fig5](a ) , one can see that , for the same modulation intensity , the position of the strongest compression corresponding to initial condition ( [ ma - initial ] ) is always smaller than its counterpart corresponding to input ( [ ab_initial ] ) . interestingly , for the double pulse train induced by the latter input , the peak power and the fwhm of the central pulse are almost constant for the present parameters , which leads to the nearly constant frequency shift at @xmath104 km . on the other hand , for the double pulse train induced by input ( [ ma - initial ] ) , the peak power increases , and the fwhm decreases , with the increase of the modulation intensity , which implies that this input can be used to generate the pulse train with the higher peak power and narrower width . however , in the same case , the frequency shift at @xmath111 km increases with the increase of the modulation intensity in a staircase fashion , as shown in fig . [ fig5](d ) . furthermore , from figs . [ fig5](b ) and ( c ) one can find that the corresponding soliton numbers are in the range of @xmath112 , each pulse in the double pulse trains eventually reshaping into the fundamental soliton , as seen in figs . fig3(c ) and [ fig4](c ) . comparing with figs . [ fig1](d ) and fig2(d ) , one can see that raman effect is helpful for the formation of solitons over a shorter propagation distance , as was shown in a different context in ref . @xcite . except for @xmath113.,width=321 ] from fig . [ fig5](c ) it can be seen that the pulse width in the trains is less than @xmath114 ps ( here the pulse width is taken equal to fwhm/@xmath115 agrawal ) , and it may fall in the subpicosecond range for larger values of modulation intensity @xmath92 . these widths are small enough to make it necessary using the extended model based on eq . ( [ model ] ) , rather than the simple nls equation which does not include higher - order terms . to additionally illustrate this point , we have performed simulations of the propagation dynamics of the train initiated by input ( [ ma - initial ] ) with @xmath116 in the subpicosecond range . in this case , the maximally compressed pulse train is attained at a shorter distance @xmath117 km in comparison with fig . subsequently , the use of the delay - line interferometer produces the double pulse train with the fwhm of the central pulse @xmath118 ps ( the corresponding pulse width is @xmath119 fs ) and peak power @xmath120 w. the corresponding robust evolution of the double pulse train is shown in fig . [ fig6](a ) . comparing with the results shown in fig . [ fig4 ] , we conclude that the spectrum range is wider , the frequency shift is larger , and the peak power is higher because the pulse width becomes narrower , as seen in figs . [ fig6](b ) and ( c ) . , but with fiber loss @xmath121 db / km.,width=321 ] note that the above considerations disregarded fiber loss , assuming that it might be compensated by gain here , we display an example which explicitly takes the loss into account , and the results are shown in fig . [ fig7 ] , in which the maximally compressed pulse train is attained at a somewhat longer distance , i.e. , @xmath122 @xmath123 km ( it was @xmath124 km without fiber loss ) and the peak power of the corresponding double pulse train is decreased to @xmath125 w , from @xmath126 w in the lossless fiber . comparing with fig . [ fig4 ] , we see that the propagation scenario is not strongly affected by the loss . of course , the loss produces stronger effects over large propagation distances ( @xmath127 km ) , in which case periodic compensation provided by gain should be explicitly taken into account . using exact solutions of the hirota equation for the soliton on the finite background , including the ab ( akhmediev breather ) and kms ( kuznetsov - ma solution ) , the double high - power pulse trains with zero - background have been constructed . the results demonstrate that the double trains , built as superpositions of the maximally compressed pulse trains with opposite signs and a half - period time delay , propagate preserving their shape . the dynamics of the trains generated by the modulated cws in the realistic fiber model has been simulated too , demonstrating that they can propagate robustly , with the red frequency shift caused by the raman effect . comparing different initial conditions , we have concluded that input ( [ ma - initial ] ) is more suitable for generating the pulse trains with a higher peak power and narrower width , the corresponding raman - induced frequency shift being larger too , exhibiting a staircase dependence on the cw - modulation parameters . the settings analyzed in this work suggest an experimental realization in nonlinear optical fibers . this research is supported by the national natural science foundation of china grant 61078079 and 61475198 , the shanxi scholarship council of china grant 2011 - 010 . a. bendahmane , a. mussot , p. szriftgiser , o. zerkak , g. genty , j. m. dudley , and a. kudlinski , experimental dynamics of akhmediev breathers in a dispersion varying optical fiber , * 39 * , 4490 - 4493 ( 2014 ) . b. frisquet , a. chabchoub , j. fatome , c. finot , b. kibler , and g. millot , two - stage linear - nonlinear shaping of an optical frequency comb as rogue nonlinear - schrdinger - equation - solution generator , * 89 * , 023821 ( 2014 ) . p. v. mamyshev , s. v. chernikov , e. m. dianov , and a. m. prokhorov , generation of a high - repetition - rate train of practically noninteracting solitons by using the induced modulational instability and raman self - scattering effects , * 15 * , 1365 - 1367 ( 1990 ) .

pulse trains growing from modulated continuous waves ( cws ) are considered , using solutions of the hirota equation for solitons on a finite background . the results demonstrate that pulses extracted from the maximally compressed trains can propagate preserving their shape and forming robust arrays . the dynamics of double high - power pulse trains produced by modulated cws in a model of optical fibers , including the raman effect and other higher - order terms , is considered in detail too . it is demonstrated that the double trains propagate in a robust form , with frequencies shifted by the raman effect .

time - resolved experiments are used to monitor time - dependent phenomena . in typical pump - probe experiments a short probe pulse follows a short pump pulse at some specific delay . femtosecond capabilities have been available for some years at visible wavelengths @xcite . however , there is a strong interest in extending these techniques to x - ray wavelengths , where one could directly probe structural changes with atomic resolution . this goal will be achieved with the realization of x - ray free electron lasers ( xfels ) @xcite . in their initial configurations , xfels will produce radiation pulses with duration of about a hundred femtosecond , which will allow time - resolved studies of transient structures of matter on the time - scale of chemical reactions . one of the main technical problems for the realization of pump - probe experiments is the relative synchronization of radiation pulses from xfel and optical laser . both x - ray and optical pulses are subject to time - jitter . in an xfel , radiation is produced by an electron bunch travelling through an undulator . the budget for time jitter of the electron bunch starts to accumulate from the photo - injector laser system . extra source of jitter is constituted by fluctuations of the electron energy , which transform to time jitter in magnetic bunch - compressors . the pump - laser itself has an intrinsic time jitter caused e.g. by mechanical vibrations in the laser oscillator or electrical - noise in the stabilization electronics . there exists a tendency for straightforward solution of the problem by means of implementation of a precise synchronization system at xfels , aiming all clocks and triggers within the facility to be perfectly synchronized @xcite . once this is done , temporal resolution of pump - probe experiment will be defined by the synchronization system plus intrinsic time - jitter of the pump - laser system . however , precise synchronization on a kilometer - scale facility is a rather challenging task , and it is not clear , at the moment , what will be the overall practical accuracy limit for the synchronization of xfel pulses to pulses from external lasers . approaches based on generation of two coherent radiation pulses of different colors by the same electron bunch passing through two distinct insertion devices have been proposed in @xcite . intrinsic feature of these schemes is an ideal mutual synchronization of the radiation pulses . with careful design of the optical transport system from source to sample it is possible to reach femtosecond level of synchronization . one of the schemes , using far infrared and vuv radiation pulses is being experimentally realized at flash , the free - electron laser in hamburg @xcite . recently , an alternative concept has been proposed for realization of pump - probe experiments with xfel pulses and powerful optical pulses from external lasers @xcite . the basic idea is to shift the attention from the problem of absolute synchronization to the problem of measurement of relative jitter between x - ray and optical pulses . if a relative delay is known for each pump - probe event , a sorting of results according to measured delay times gives the required experimental outcome . in this paper we propose a concept of time - arrival monitor allowing measurement of a relative delay between xfel and optical pulses on a femtosecond time scale . our scheme makes use of the same electron bunch to produce both an xfel pulse and a powerful laser - like ( bandwidth - limited and diffraction - limited ) optical pulse . the latter is generated downstream of the x - ray undulator , and will be naturally synchronized to the xfel pulse . relative synchronization is preserved by using the same mechanical support for optical elements transporting x - ray pulse and optical pulse down to the experimental area . there , single - shot cross - correlation between the optical pulse and the pump - laser pulse can be performed yielding , effectively , the temporal delay between xfel pulse and pump - laser pulse . implementation of the proposed scheme for pump - probe experiments into the project of the european xfel @xcite is described in detail . we show that it naturally fits into the project . a modulation at optical frequency is imprinted onto the electron bunch using the optical replica synthesizer ( ors ) setup @xcite that will be installed after the final beam compression stage at the energy of 2 gev . subsequently , electrons travel through the main undulator and radiate the xfel pulse . finally , the presence of a long straight section and a separation magnet downstream of the main undulator serve as optical radiator , where the modulated electron bunch produces a powerful pulse of coherent edge - radiation . the proposed arrival - time monitor can be realized without the addition of specific hardware in the accelerator and undulator tunnels , and requires minimal efforts . the applicability of the present scheme is not restricted to the european xfel setup . other projects , e.g. the lcls or scss facility @xcite , may benefit from our proposal too . scheme for pump - probe experiments with xfel pulses and pulses produced by an external laser . operation of the scheme is based on production by the electron bunch of an optical pulse naturally synchronized with the xfel pulse . relative synchronization is preserved by using the same radiation transport system to the experimental area . the optical pulse is then used to monitor the delay time between pump and probe by correlating it with the pump laser.,width=529 ] a basic scheme of the timing system is shown in fig . [ setup ] . discussions in this paper focus on design and parameters ( see fig . [ realscheme ] ) of the sase 1 line of the european xfel , operating in the wavelength range around @xmath1 nm . main elements of the timing system are energy modulator ( constituted by seed laser and undulator modulator ) , dispersion section and radiator . the energy modulator ( see fig . [ ors ] ) is located after the second bunch compressor ( bc2 ) , where the electron energy is about @xmath2 gev . a laser pulse at wavelength @xmath3 nm is used to modulate the electron energy at the same wavelength . the duration of the laser pulse is chosen to be about @xmath4 ps , much longer than the time - jitter of the electron pulse ( a fraction of a picosecond ) to avoid synchronization problems . the energy of the laser pulse is about @xmath5 mj . laser beam is focused onto the electron beam in a short modulator undulator , with a number of periods @xmath6 , resonant at the seed laser wavelength @xmath3 nm . energy modulation with an amplitude of about @xmath5 mev is achieved due to interaction with the electron bunch in the modulator undulator . subsequently , the electron bunch passes through the dispersion section with momentum compaction @xmath7 m , where the energy modulation induces a density modulation at the seed - laser wavelength . with parameters discussed above , the density modulation reaches an amplitude of about @xmath8 . following the dispersion section the bunch is accelerated up to the energy of @xmath9 gev in the main accelerator , it passes trough the collimator system and is directed to the sase 1 x - ray undulator . a high - current ( 5 ka ) electron beam is transported through the xfel linac , and it is therefore mandatory to include self - interaction effects in our analysis . during the passage of the bunch through the accelerator , the initial density modulation produces an energy modulation due to longitudinal impedance caused by space - charge fields . if the collimation system is properly tuned , such energy modulation can induce further modulation in density . calculations presented in section [ sec : modu ] show that exploitation of self - interaction mechanisms allow one to deliver , at the entrance of the sase 1 undulator , an electron bunch with a @xmath0 density modulation , and negligible energy modulation . the sase fel process in the baseline undulator is not perturbed by such level of density modulation : fluctuations of the electron current density serve as input signal for the radiation amplification process , which develops nearly in the same way as with an unmodulated electron bunch . as a result , at the exit of the sase 1 undulator the electron bunch produces the nominal x - ray pulse . finally , after the sase 1 undulator , the modulated electron bunch passes through a long straight section followed by a separation magnet , which separates the electron bunch from the sase pulse . due to the presence of the combination of straight section and separation magnet , the electron bunch emits an optical pulse of coherent edge radiation at the wavelength of the bunch density modulation . such an optical pulse carries about @xmath10 photons ( order of a microjoule ) and is delivered in a bandwidth - limited and diffraction - limited pulse . summing up , combination of an optical modulator after magnetic bunch compressor bc2 and optical radiator after the baseline undulator will allow to produce a powerful laser - like optical pulse at the entrance of the photon beamline . this laser - like pulse is naturally synchronized with the electron bunch and thereby with the x - ray pulse . thus , the problem of measuring the time - shift between a pump - laser pulse and a ( x - ray ) probe pulse is reduced to the problem of measuring the time delay between two ultrashort optical pulses , which may be solved with standard techniques . contrarily to the x - ray pulse , which is peaked on - axis , the edge - radiation pulse is peaked in the forward direction at an angle of a few tens of microradians . the distance from the separation magnet to the first ( x - ray ) mirror station is about @xmath11 m. according to calculations presented in section [ sec : rad ] , the diameter of the spatial distribution of edge radiation at this distance will be about @xmath12 cm , which fits with the aperture of the photon beamline . relative synchronization of x - ray and optical pulses must be preserved during propagation to the experimental area . the best way to do so is to use the same mechanical support for both optical elements transporting the x - ray pulse and optical elements transporting the optical pulse . for example , optical elements for the transportation of edge - radiation may be directly assembled on the x - ray mirrors . after passing through the transport system , the edge - radiation pulse will reach the experimental area , where a single - shot cross - correlation measurement with the pump - laser pulse can be performed . in reference @xcite the possibility of extracting the temporal shift between pulses from two ultrashort laser pulses was experimentally demonstrated . the method was based on sum - frequency generation in a non - linear crystal . the intrinsic accuracy of this technique was shown to be within the femtosecond range . successful measurements of the time - offset signal were performed even when one of the two pulses was very weak ( down to @xmath13 photons per pulse ) . integration of the time - arrival monitor in the european xfel setup makes use of the optical replica synthesizer ( ors ) . the ors is used to imprint density modulation onto electron beam . optical radiator is the combination of straight - section and separation magnet after the sase 1 undulator which produsese powerful optical pulse of coherent edge radiation.,width=529 ] the optical replica synthesizer ( ors ) will be used to superimpose density modulation at optical wavelength on the electron bunch . first , energy modulation is created by letting a laser interact with the bunch inside a modulation undulator . second , a dispersion section ( magnetic chicane ) induces the desired density modulation . the radiator undulator , used in optical - replica ultrashort electron bunch diagnostics , is switched off during operation of the arrival - time monitor.,width=529 ] the modulator to be used in our scheme is the optical replica modulator @xcite , which consists of three elements ( see fig . [ ors ] ) : an optical seed laser , a modulator undulator and a dispersion section ( magnetic chicane ) . the radiator undulator shown in fig . [ ors ] , used in optical - replica ultrashort electron bunch diagnostics , is switched off during operation of the arrival - time monitor . the seed laser pulse interacts with the electron beam in the modulator undulator , which is resonant with the laser wavelength @xmath14 . as a result , the electron bunch is modulated in energy . in the following dispersion section the energy modulation induces a density modulation at the optical wavelength . the dispersion section is designed to produce an energy dependence of the particles path length @xmath15 , where @xmath16 is the energy deviation of a particle in units of the rest mass , @xmath17 is the deviation from the path length of an electron with nominal energy @xmath18 in units of the rest mass , and @xmath19 is the compaction factor of the dispersion section . @xmath16 and @xmath17 may assume positive or negative values , while @xmath20 , the dispersion section being a chicane . suppose that at the entrance of the chicane we have an initial energy modulation . then , in units of the rest mass , the energy deviation of each particle due to this energy modulation is given by @xmath21 , where @xmath22 $ ] is the modulation phase , with @xmath23 the longitudinal velocity of a nominal electron , and @xmath24 , @xmath25 being the speed of light in vacuum . here @xmath26 is the amplitude of our small ( @xmath27 ) energy modulation taken with its own sign . at the exit of the chicane we may express the current as @xmath28 $ ] . here @xmath29 is the unmodulated electron beam current , and @xmath30 is the amplitude of our small ( @xmath31 ) density modulation , also taken with its own sign . neglecting collective effects , the amplitude of density modulation at the exit of the chicane , @xmath30 , approaches @xcite @xmath32~ , \label{a1}\end{aligned}\ ] ] where @xmath33 is the rms uncorrelated energy spread of the electron bunch in units of the rest mass , and @xmath34 is the reduced modulation wavelength . in the case of the european xfel ( see @xcite , fig . [ realscheme ] and fig . [ ors ] ) @xmath35 , corresponding to an energy of @xmath2 gev , @xmath36 , corresponding to @xmath4 mev rms uncorrelated energy spread , @xmath37 , corresponding to @xmath5 mev modulation , and @xmath3 nm , corresponding to the second harmonic of a ti : si laser . a value @xmath38 m leads to a modulation amplitude @xmath39 . note that in this case the exponential suppression factor in eq . ( [ a1 ] ) is about @xmath40 , and can be practically neglected . if collective effects could be neglected during the transport of the bunch , one could propagate the @xmath0 initial density modulation up to the radiator , accounting for velocity bunching and , possibly , for the presence of a non - zero compaction factor at the collimation section @xmath41 . in this case , estimations would remain within the framework of single - particle dynamics . tuning the energy in the optical seed laser pulse , one may easily achieve a final modulation @xmath42 at the optical radiator entrance . however , collective effects strongly influence the modulation process in our jitter - monitoring scheme . in other words , the problem of propagation of the induced beam density modulation through the setup depicted in fig . [ realscheme ] is a problem involving self - interactions . as the bunch progresses through the linac , the modulation of the bunch density produces an energy modulation due to longitudinal impedance caused by space - charge field . this process is complicated by the fact that , due to the presence of energy and density modulation , plasma oscillations can develop . one should account for these facts in order to quantify energy and density modulation before the collimation section . then , in the collimation section , the energy modulation induces extra - density modulation due to non - zero compaction factor . finally , longitudinal space - charge impedance is responsible for further energy modulation and further plasma oscillations during the passage of the beam in the main undulator . as a result , the initial beam modulation , in energy and density , will be modified by the passage through the setup . in order to study the feasibility of our scheme one needs to estimate what modifications will take place . at the entrance of the accelerator , the density modulation amplitude is given by @xmath30 , and the energy modulation amplitude by @xmath26 . due to energy modulation , particles undergo a phase shift , with respect to the phase @xmath43 , which is responsible for a change in the density modulation along the acceleration section . also the energy modulation is a varying function of @xmath44 because of the presence of longitudinal space - charge forces . indicating with @xmath45 and with @xmath46 the density and energy modulation amplitudes along the accelerator we may write @xmath47 where , additionally , the relativistic lorentz factor of the bunch , @xmath48 , accounts for the acceleration process . ( [ daiacc ] ) can be directly derived from eq . ( [ a1 ] ) substituting @xmath30 and @xmath26 with @xmath45 , and @xmath49 , neglecting the exponential suppression factor due to uncorrelated energy spread , remembering that the compaction factor for a free - space section of length @xmath17 is given by @xmath50 , and taking the limit for @xmath51 . one can estimate the derivative of @xmath49 along the xfel linac using results from papers studying microbunching instabilities like @xcite . we can write : @xmath52 where @xmath53 is the longitudinal impedance induced by space charge , @xmath54 is the free - space impedance , expressed in the same units of @xmath55 , and @xmath56 ka the alfven current . the longitudinal impedance induced by space charge in free space was studied in the case of an electron bunch with finite transverse profile in @xcite . in that reference we gave analytical expressions for the impedance in the steady state limit , which can be easily generalized in the case of adiabatic acceleration , when @xmath57 . the adiabatic acceleration limit can always be used in our case . assuming a constant acceleration gradient @xmath58 ( see @xcite ) , we have @xmath59 for @xmath60 , corresponding to the lowest energy of @xmath2 gev , and @xmath61 for @xmath62 , corresponding to the highest energy of @xmath9 gev , whereas the largest effects due to longitudinal impedance are expected in the first part of the acceleration process . then , using results in @xcite , which are valid for a gaussian transverse distribution of the electron bunch , we find that eq . ( [ modul ] ) can be written as @xmath63 ~\gamma\left[0,\frac{\epsilon_n \beta}{(\gamma_0+g z)^3 \lambdabar^2}\right]}~. \label{modul2}\end{aligned}\ ] ] here @xmath64 is the incomplete gamma function , @xmath65 is the average betatron function in the accelerator , and @xmath66 is the normalized emittance . the system of coupled differential equations constituted by eq . ( [ daiacc ] ) and eq . ( [ modul2 ] ) should be solved with given initial conditions @xmath67 and @xmath68 at the entrance of the accelerator in order to obtain density and energy modulation at the entrance of the collimator . such system of equations accounts for plasma oscillations of the electron bunch in the limit for adiabatic acceleration , with the help of a longitudinal impedance averaged along the transverse direction . a more detailed analysis of space - charge waves performed as a function of the transverse coordinates ( i.e. without averaging the longitudinal impedance ) is given in @xcite , where we studied the problem of plasma oscillations for an electron bunch with arbitrary transverse profile going along a straight section with uniform motion . we assume an average betatron function of about @xmath69 m along the main accelerator and a normalized emittance @xmath70 mm@xmath71mrad ( see @xcite ) . setting the acceleration length @xmath72 m ( see fig . [ realscheme ] ) and @xmath73 ka , numerical analysis shows that our initial conditions @xmath74 and @xmath75 yield , at the entrance of the collimator @xmath76 , @xmath77 and @xmath78 . the nominal value of the compaction factor of the collimator @xmath41 is set to zero , with possibility of fine tuning around this value of about @xmath79 m . taking advantage of this possibility and setting @xmath80 m , at the exit of the collimator one obtains a density modulation @xmath81 , whereas the energy modulation remains unvaried @xmath82 . further on , energy and density modulation should be propagated through a straight section followed by the main xfel undulator . propagation can be performed using the same set of equations eq . ( [ daiacc ] ) and eq . ( [ modul2 ] ) , setting @xmath83 , using an energy of @xmath9 gev and an average value of the betatron function @xmath84 m . this gives only a correction to the energy modulation , so that at the entrance of the xfel undulator , at @xmath85 , one still has @xmath86 , while @xmath87 . decrease of the energy modulation is related to an advantageous initial phase of plasma oscillation at the entrance of the straight section . numerical analysis shows that such an energy modulation , being smaller than the foreseen xfel spectral bandwidth @xmath88 @xcite , will not alter the xfel process ( see figs . [ fig : psat - enmod ] and [ fig : spectrum-0 - 4 ] ) . normalized spectrum of the fel radiation pulse at zero energy modulation ( left plot ) , and at @xmath89 mev ( right plot ) . numerical example corresponds to undulator sase 1 at the european xfel @xcite . amplitude and period of electron bunch modulation are 10% and 400 nm , respectively . simulations have been performed with the code fast @xcite.,width=529 ] normalized spectrum of the fel radiation pulse at zero energy modulation ( left plot ) , and at @xmath89 mev ( right plot ) . numerical example corresponds to undulator sase 1 at the european xfel @xcite . amplitude and period of electron bunch modulation are 10% and 400 nm , respectively . simulations have been performed with the code fast @xcite.,title="fig:",scaledwidth=45.0% ] normalized spectrum of the fel radiation pulse at zero energy modulation ( left plot ) , and at @xmath89 mev ( right plot ) . numerical example corresponds to undulator sase 1 at the european xfel @xcite . amplitude and period of electron bunch modulation are 10% and 400 nm , respectively . simulations have been performed with the code fast @xcite.,title="fig:",scaledwidth=45.0% ] similarly as before , the passage in the main xfel undulator has the effect of decreasing the energy modulation level , too . moreover , although the undulator is shorter than the straight section preceding it , the effect on the energy modulation is stronger . in fact , the longitudinal lorentz factor @xmath90 should be used in the undulator instead of @xmath91 ( see reference @xcite ) . since @xmath92 , @xmath93 and @xmath94 differ of about an order of magnitude , hence the different influence of the undulator compared with the straight section . in the undulator , ( [ daiacc ] ) and eq . ( [ modul2 ] ) are modified to @xmath95 and @xmath96 ~\gamma\left[0,\frac{\epsilon_n \beta}{\lambdabar^2 \gamma_0 \gamma_z^2}\right]}~. \label{modul2x}\end{aligned}\ ] ] solving numerically with previously found initial conditions @xmath97 and @xmath98 , and using @xmath99 m , one finds the energy and density modulation levels at the radiator entrance , @xmath100 and @xmath101 . as a final remark it should be noted that , pending design finalization , the @xmath41 value may be set to any value from @xmath102 mm to @xmath103 mm , different from the nominal value @xmath104 . depending on this value and , possibly , using the fine tuning option to increase or decrease the @xmath41 value up to @xmath105 m , different initial conditions should be set to obtain acceptable values for @xmath106 and @xmath107 . for example , if @xmath108 mm , setting @xmath109 and @xmath110 ( corresponding to about @xmath111 mev ) would yield @xmath100 and @xmath112 , corresponding to about @xmath12 mev energy modulation , which is perfectly compatible with our scheme . it follows that the optical modulator can induce about @xmath113 density modulation at the entrance of the optical radiator , @xmath114 , and acceptable energy modulation , independently of the design of the collimation section , without perturbation of the fel process in the baseline undulator . after collimation we deal with an electron bunch modulated in density at optical wavelength . this wavelength is much larger than the geometrical emittance of the beam , and in our case the electron bunch can be considered as a filament with no transverse dimension nor divergence , as far as optical wavelengths are concerned . an analysis of the problem in the space - frequency domain @xcite shows that when a filament beam modulated in density passes through a given trajectory , it produces coherent radiation very much likely a single electron . in general , one needs to solve maxwell s equation for given macroscopic sources . a paraxial treatment is possible , based on the ultrarelativistic assumption @xmath115 . consider the transverse components of the fourier transform of the electric field . they form a vector @xmath116 , dependent on transverse and longitudinal coordinates @xmath117 and @xmath44 . from the paraxial approximation follows that the envelope @xmath118}$ ] , does not vary much along @xmath44 on the scale of the reduced wavelength @xmath119 . with some abuse of language we will indicate @xmath120 as `` the field '' . the field obeys the following paraxial wave equation in the space - frequency domain : @xmath121 = \vec{g}(z , \vec{r},\omega)$ ] . here @xmath122 , and the differential operator @xmath123 is defined by @xmath124 , @xmath125 being the laplacian operator over transverse cartesian coordinates . the source - term vector @xmath126 is specified by the trajectory of the source electrons , and can be written in terms of the fourier transform of the transverse current density , @xmath127 , and of the charge density , @xmath128 , as @xmath129 \left({i\omega}/{c^2 } \cdot \vec{\bar{j } } -\vec{\nabla}_\bot \bar{\rho}\right)$ ] . vector @xmath130 and @xmath131 are regarded as given data . in this paper we will treat @xmath132 and @xmath131 as macroscopic quantities , without investigating individual electron contributions . from the previous discussion it follows that , as concerns emission of coherent optical radiation , our setup reduces to an upstream bending magnet ( corresponding to the last bend of the collimator ) followed by a straight section , an undulator ( the main sase 1 undulator ) , a second straight section and a downstream separation bending magnet , which divides the electron beam from the xfel pulse ( see fig . [ realscheme ] ) . we picture the upstream bending magnet as a `` switch - on '' of both harmonics of the electromagnetic sources and of the field . similarly , the downstream bend enforces a `` switch - off '' process . when a modulated beam passes through a straight section limited by upstream and downstream bending magnets , it produces edge radiation ( see among others , @xcite and references therein ) . in our case , trajectory from the collimator section up to the beam dump is more complicated than a single straight section limited by bending magnets , but conceptually the mechanism of radiation production is the same . moreover , the influence of bending magnet radiation to the field contribution can be shown to be negligible . to see this , it is sufficient to compare the radiation formation length of the field associated with bending magnets with the radiation formation length of the shortest straight section . deflection introduced in the collimation section corresponds to a bending radius of @xmath133 m. deflection introduced between sase 1 and sase 3 corresponds , instead , to a bending radius of about @xmath134 m. for a wavelength @xmath3 nm and bending radius @xmath135 m we obtain the longest formation length @xmath136 m. for the shortest straight section of length @xmath137 m , the edge - radiation formation length would be @xmath138 = 2 \gamma^2\lambdabar \simeq 150 $ ] m , which is about @xmath139 times longer than the formation length for the bends . it follows that field contributions from bending magnets can be neglected with an accuracy @xmath140 , and a sharp - edge approximation applies . understanding the operation of the optical radiator is made simpler with the help of the theoretical study in reference @xcite . in that reference we showed that radiation from an ultra - relativistic filament beam along the trajectory specified above can be interpreted as radiation from three virtual sources , located at specific longitudinal positions . specification of these virtual sources amounts to specification of three field distributions @xmath141 at three distinct locations @xmath142 . in principle these locations are a matter of taste , but there are particular choices of @xmath142 where @xmath141 exhibit plane wavefronts and are similar to waists of laser beams . in the case under study these privileged longitudinal positions are the center of the straight sections and the center of the undulator . let us indicate with @xmath143 and @xmath144 the lengths of upstream and downstream straight sections , with @xmath145 the length of the main xfel undulator and set @xmath146 in its center . in this case , the three sources are located at @xmath147 , @xmath148 and @xmath149 . since the virtual sources exhibit plane wavefronts , they are completely specified by real - valued amplitude distributions of the field . in the case of a single electron , these were derived from the far zone field distribution and were found to be @xcite : @xmath150\cr & & \times \frac{e l_1}{2\pi c \lambdabar^2 } \int d\vec{\theta } ~\vec{\theta}~ \mathrm{sinc}\left[\frac{1}{4 } \left(\frac{l_1 \theta^2}{\lambdabar}+\frac{l_1}{\gamma^2 \lambdabar}\right ) \right]\exp\left[\frac{i}{\lambdabar}\vec{r}\cdot \vec{\theta } \right ] ~,\label{vir1eaf}\end{aligned}\ ] ] @xmath151\exp\left[\frac{i}{\lambdabar}\vec{r}\cdot \vec{\theta } \right ] ~\cr & & \label{vir1ebf}\end{aligned}\ ] ] and @xmath152\cr & & \times \frac{e l_2}{2\pi c \lambdabar^2 } \int d\vec{\theta } ~\vec{\theta}~ \mathrm{sinc}\left[\frac{1}{4 } \left(\frac{l_2 \theta^2}{\lambdabar}+\frac{l_2}{\gamma^2 \lambdabar}\right ) \right]\exp\left[\frac{i}{\lambdabar}\vec{r}\cdot \vec{\theta } \right ] ~.\label{vir1edf}\end{aligned}\ ] ] note that the field is radially polarized . aside for a different phase , eq . ( [ vir1eaf ] ) , eq . ( [ vir1ebf ] ) and eq . ( [ vir1edf ] ) have a similar mathematical structure . however , for sase 1 , @xmath153 , so that @xmath154 is about an order of magnitude smaller than @xmath93 . it follows that the second term in the @xmath155 function of eq . ( [ vir1eaf ] ) or eq . ( [ vir1edf ] ) is much smaller than the analogous term in eq . ( [ vir1ebf ] ) . in particular , for our study case we have @xmath156 and @xmath157 of order unity , but @xmath158 . as a result , the @xmath155 function in eq . ( [ vir1edf ] ) is strongly suppressed , one may neglect the virtual source located in the center of the undulator and let @xmath159 , at least in first approximation . it should be remarked that the virtual source @xmath160 is responsible for a field contribution known as transition undulator radiation . typical expressions for tur emission @xcite consist of relations for the energy distribution of radiation in the far zone that account for the presence of the virtual source at @xmath161 alone , without considering further contributions due to other elements in the beamline , e.g. the straight sections in our case . as it was shown in @xcite and @xcite these expressions have no physical meaning . in our study case , as we have just seen , the contribution from transition undulator radiation can be even neglected . we are thus left with the contributions from two virtual sources located at @xmath162 and @xmath163 , accounting for edge radiation emission from the straight lines before and after the undulator . however , the contribution from the upstream straight section will be strongly suppressed by a photon stop inside the main undulator , whose main function is that of absorbing spontaneous radiation background . moreover , an extra photon stop might be installed at the exit of the main undulator , absorbing all but the sase pulse . as a result , one has to deal with the simple situation where the optical radiator is composed by a single straight section limited by bending magnets downstream of the main undulator . since @xmath164 is known , free - space propagation from the virtual source through the near zone and up to the far - zone can be performed with the help of the fresnel propagation as it is done for usual laser beams @xcite . the only difference is in the peculiar shape of @xmath164 , which reflects the particular trajectory followed by the filament beam . it is important to realize that the first optical element of the optical beamline will be placed at about @xmath11 m from the end of the straight section . such distance is comparable with the straight section length . thus , one needs to know how edge radiation propagates in the near zone in order to characterize the pulse at the optical element position . in this regard , it should be stressed that fresnel propagation allows one to calculate the electric field not only in the far zone , but in the near zone too , solving the full problem of free - space propagation . in principle , one may directly propagate eq . ( [ vir1edf ] ) . however , the integral in @xmath165 can not be solved analytically . numerical evaluations are simplified with the help of reference @xcite , where we showed that the virtual source in eq . ( [ vir1edf ] ) is equivalent to two virtual sources located at the edges of the straight section . these two sources still present a plane wavefront , and they can be described analytically in a simple way in terms of the modified bessel function of the first order @xmath166 . it is convenient to adopt this picture for computational purposes . shifting , for simplicity , the origin of the longitudinal axis in the center of the downstream straight section and letting @xmath167 we can write the two virtual sources as @xcite : contour plot of the spatial distribution of radiation simulated with the help of the srw code at the position of the first optical element , @xmath168 m. , width=529 ] photon density distribution as a function of the radial position at the first optical element , @xmath168 m. the solid line has been calculated with the help of eq . ( [ fiprf ] ) . black and white circles are obtained by cutting the contour plot in fig . [ srw2d ] respectively along the horizontal and vertical axis.,width=529 ] @xmath169 \frac{\vec{r}}{r } k_1\left(\frac{r}{\gamma \lambdabar } \right)~.\label{virpm05}\end{aligned}\ ] ] fresnel propagation can now be performed , and the radiation energy density can be calculated as @xcite @xmath170 with @xmath171 \exp\left[\frac{i}{4 } \frac{l}{\gamma^2\lambdabar}\right ] \right.\right.\cr & & \left.\left . \times \int_0^{\infty } d{r } ' { r } ' k_1\left(\frac{r'}{\gamma\lambdabar}\right ) j_1 \left(\frac{r{r}'}{\lambdabar({z}-l/2)}\right ) \exp\left[\frac{i{r}'^2 } { 2\lambdabar\left({z}-l/2\right)}\right ] \right\ } \right.\cr & & \left . + \left\{\frac{\vec{r}}{r } \left(\frac{2 i } { { z}+l/2}\right){\frac{l}{\gamma\lambdabar \sqrt{\lambdabar l } } } \exp\left[\frac{i r^2 } { 2 \lambdabar\left({z}+l/2\right)}\right ] \exp\left[-\frac{i}{4 } \frac{l}{\gamma^2\lambdabar}\right ] \right.\right.\cr & & \left.\left . \times \int_0^{\infty } d{r } ' { r } ' k_1\left(\frac{r'}{\gamma\lambdabar}\right ) j_1 \left(\frac{r{r}'}{\lambdabar({z}+l/2)}\right ) \exp\left[\frac{i{r}'^2 } { 2\lambdabar\left({z}+l/2\right)}\right ] \right\ } \right|^2~.\cr & & \label{fiprf}\end{aligned}\ ] ] since we are interested in the radiation energy from a filament beam with a given longitudinal profile , we should multiply the single - electron result by the squared - modulus of the fourier transform of the temporal profile of the bunch . near the modulation frequency @xmath172 , a bunch with modulated gaussian temporal profile and rms duration @xmath173 gives @xmath174 + \exp\left[-\frac{\sigma_t^2}{2 } \left(\omega+\frac{c}{\lambdabar}\right)^2\right]\right\}~ , \label{fbarrrr}\end{aligned}\ ] ] where @xmath175 is the fourier transform of the temporal profile of the bunch and @xmath176 is the number of electrons in the bunch . in order to calculate the spatial density distribution of the number of photons per pulse we should integrate eq . ( [ fipr ] ) in @xmath177 . since we are interested in coherent emission around the modulation wavelength , we can consider the wavelength in eq . ( [ fipr ] ) fixed . this amounts to a multiplication of eq . ( [ fipr ] ) by @xmath178 leading to @xmath179 where @xmath180 is the fine structure constant . we considered ( see @xcite ) the case for @xmath181 , @xmath182 m , @xmath183 fs and @xmath184 ( i.e. about @xmath4 nc ) . since the first element of the optical beamline is foreseen to be placed at about @xmath11 m from the separating magnet ( see fig . [ realscheme ] and fig . [ tunnel ] ) , and since we measure @xmath44 from the center of the straight section , we set our observation plane at @xmath185 m. then we use eq . ( [ endenee ] ) to calculate the photon density distribution . we cross - checked our analytical results with the code srw @xcite . a contour plot for the photon density distribution as calculated from srw is given in fig . [ srw2d ] . horizontal and vertical cuts along the contour plot are compared with eq . ( [ endenee ] ) in fig . [ profile ] . the total number of photons between the first two minima of the distribution function ( at @xmath186 cm and @xmath187 cm respectively ) is obtained integrating in @xmath188 . for parameters selected above , @xmath189 photons . schematic representation of the optical transport system in the european xfel.,width=529 ] the optical pulse must be transported to the experimental area preserving relative synchronization with the xfel pulse . the most convenient way to accomplish this task is to use the same physical support for both x - ray and edge - radiation optics , e.g. assembling mirrors for the transport of edge radiation directly on x - ray optical elements . in this way , mechanical vibrations of the transport system will not affect relative synchronization of the two pulses . a scheme of the xfel optical system is sketched in fig . [ tunnel ] . as mentioned before , the first optical elements will be located about @xmath11 m downstream of the separation magnet . once edge - radiation is transported to the experimental area , single - shot cross - correlation with the pump - laser pulse can be performed , yielding the time delay between the pump - laser and the xfel pulse . as discussed before , in reference @xcite a measurement of the time - offset signal between two ultrashort laser pulses was experimentally performed based on sum - frequency generation in a non - linear crystal . the intrinsic accuracy of the method is in the femtosecond range . actual measurements were successfully performed even when one of the two laser pulses is very weak , up to @xmath13 photons per pulse . this sensitivity allows our method to be employed even if a large number of optical photons are lost , from the total of @xmath10 photons per pulse , during the transport process along the optical system . our analysis demonstrates the feasibility of pump - probe experiments at xfels with femtosecond temporal resolution , based on timing of xfel pulses to optical pulses from an external pump - laser . the proposed scheme does not require absolute synchronization of pump and probe pulses . synchronization in the sub - picosecond range , which has been experimentally demonstrated , is sufficient for its operation . the present study includes an analysis of physical principles of operation , which are clear and transparent , and of fundamental physical effects of importance for the operation of the proposed time - arrival monitor . technical realization will be rather simple and cost - effective since it is essentially based on technical components ( optical - replica synthesizer ) being already included in the design of the european xfel . baseline parameters of the european xfel were used in our analysis . thus , the proposed scheme can be implemented at the first stage of xfel facility operation . we thank massimo altarelli ( desy ) and jochen schneider ( desy ) for their interest in this work , vitali kocharian ( desy ) for his help with srw simulations . bosch and o.v . chubar , `` long wavelength edge radiation in an electron storage ring '' , proc . sri97 , tenth us national conference , aip conf . 417 , edited by e. fontes ( aip , woodbury , ny ) , pp . 35 - 41 ( 1997 )

we propose a technique , to be used for time - resolved pump - probe experiments , for timing an x - ray free electron laser ( xfel ) to a high - power conventional laser with femtosecond accuracy . our method takes advantage of the same electron bunch to produce both an xfel pulse and an ultrashort optical pulse with the help of an optical radiator downstream of the x - ray undulator . since both pulses are produced by the same electron bunch , they are perfectly synchronized . application of cross - correlation techniques will allow to determine relative jitter between the optical pulse ( and , thus , the xfel pulse ) and a pulse from an external pump - laser with femtosecond resolution . technical realization of the proposed timing scheme uses an optical replica synthesizer ( ors ) setup to be installed after the final bunch - compression stage of the xfel . the electron bunch is modulated in the ors setup by an external optical laser . subsequently , it travels through the main undulator , and produces the xfel pulse . finally , a powerful optical pulse of coherent edge radiation is generated as the bunch passes through a long straight section and a separation magnet downstream of the main undulator . our study shows that at a moderate ( about @xmath0 ) density modulation of the electron bunch at the location of the optical radiator allows production of high power x - ray and optical pulses . relative synchronization of these pulses is preserved by using the same mechanical support for both x - ray and optical elements transporting radiation down to the experimental area , where single - shot cross - correlation between optical pulse and pump - laser pulse is performed . we illustrate the potential of the proposed timing technique with numerical examples referring to the european xfel facility . x - ray free - electron laser ( xfel ) , synchronization , longitudinal impedance , space - charge 41.60.cr , 42.25.-p , 41.75.-ht

this is to certify that : 1 . the thesis comprises only my original work towards the mphil ; 2 . due acknowledgement has been made to all other material used ; and 3 . the thesis is less than 50,000 words in length . [ cols="^",options="header " , ] in this section , we will prove that under certain conditions , the information state converges in distribution . this fact is already known for classical hidden markov models , and is quite robust : legland and mevel @xcite prove geometric ergodicity of the information state even when calculated from incorrectly specified parameters , while capp , moulines and rydn @xcite prove harris recurrence of the information state for certain uncountable state underlying chains . we will present a mostly elementary proof of convergence in the case of multiple observation processes . to determine the limiting behaviour of the information state , we begin by finding an explicit form for its one - step time evolution . [ def : rfunction ] for each observation process @xmath0 and each observed state @xmath1 , the * @xmath2-function * is the function @xmath3 given by @xmath4 where @xmath5 is the dirac measure on @xmath6 and @xmath7 is the @xmath8th component of @xmath9 . [ lem : informationstaterecurrence ] in a hidden markov model with multiple observation processes and a fixed policy @xmath10 , the information state satisfies the recurrence relation @xmath11 let @xmath12 and @xmath13 . by the markov property as in definition [ def : observationprocess ] , and the simplification ( [ eqn : noidependence ] ) , @xmath14 & \hspace{3cm}\times{\mathbb p}\big(x_{t+1}=x{\,\big|\,}x_t = j\big){\mathbb p}\big(x_t = j , y^{(i_{(t)})}_{(t)}=y_{(t)}\big)\nonumber\\ & = \frac1{k_{t+1}}\sum_jm^{(i_{t+1})}_{x , y_{t+1}}t_{j , x}k_tz_t(y_{(t)})_j\nonumber\\ & = \frac{k_t}{k_{t+1}}\sum_jm^{(i_{t+1})}_{x , y_{t+1}}t_{j , x}z_t(y_{(t)})_j\nonumber\\ & = \frac{\sum_jm^{(i_{t+1})}_{x , y_{t+1}}t_{j , x}z_t(y_{(t)})_j}{\sum_x\sum_jm^{(i_{t+1})}_{x , y_{t+1}}t_{j , x}z_t(y_{(t)})_j},\end{aligned}\ ] ] since @xmath15 does not depend on @xmath16 and @xmath17 . note that for each information state @xmath18 and each observation process @xmath0 , there are at most @xmath19 possible information states at the next step , which are given explicitly by @xmath20 for each observation @xmath21 . [ lem : informationdistributionrecurrence ] the information distribution satisfies the recurrence relation @xmath22 where the sum is taken over all observation processes @xmath0 and all observation states @xmath1 , @xmath23 is the dirac measure on @xmath24 , and @xmath25 is the matrix product considering @xmath9 as a row vector . since @xmath26 is a deterministic function of @xmath27 , given that @xmath28 , @xmath29 this depends only on @xmath30 and @xmath31 , so given that @xmath32 and @xmath33 , @xmath34 integration over @xmath35 gives @xmath36 by definition [ def : informationstate ] , @xmath37 is the posterior distribution of @xmath38 given the observations up to time @xmath39 , so @xmath40 , the @xmath16th coordinate of the vector @xmath9 . since @xmath37 is a function of @xmath27 , which is a function of @xmath41 and the observation randomness @xmath42 , by the markov property as in definition [ def : observationprocess ] , @xmath43 substituting ( [ eqn : pygivenz ] ) into ( [ eqn : mutplusone ] ) completes the proof . note that lemma [ lem : informationdistributionrecurrence ] shows that the information distribution is given by a linear dynamical system on @xmath44 , and therefore the information state is a markov chain with state space @xmath24 . we will use tools in markov chain theory to analyse the convergence of the information state , for which it will be convenient to give a name to this recurrence . [ def : transitionfunction ] the * transition function * of the information distribution is the deterministic function @xmath45 given by @xmath46 , extended linearly to all of @xmath44 by the recurrence in lemma [ lem : informationdistributionrecurrence ] . the coefficients @xmath47 are called the * @xmath48-functions*. we now give a criterion under which the information state is always positive recurrent . a discrete state markov chain @xmath38 is called * ergodic * if it is irreducible , aperiodic and positive recurrent . such a chain has a unique invariant measure @xmath49 , which is a limiting distribution in the sense that @xmath38 converges to @xmath49 in total variation norm @xcite . a discrete state markov chain @xmath38 is called * positive * if every transition probability is strictly positive , that is , for all @xmath50 , @xmath51 . this is a stronger condition than ergodicity . [ def : anchored ] we shall call a hidden markov model * anchored * if the underlying markov chain @xmath38 is ergodic , and for each observation process @xmath0 , there is a state @xmath52 and an observation @xmath53 such that @xmath54 and @xmath55 for all @xmath56 . the pair @xmath57 is called an * anchor pair*. heuristically , the latter condition allows for perfect information @xmath58 whenever the observation @xmath53 is made using observation process @xmath0 . this anchors the information chain in the sense that this state can be reached with positive probability from any other state , thus resulting in a recurrent atom in the uncountable state chain @xmath37 . on the other hand , since each information state can make a transition to only finitely many other information states , starting the chain at @xmath58 results in a discrete state markov chain , for which it is much easier to prove positive recurrence . [ lem : anchor ] in an anchored hidden markov model , for any anchor pair @xmath57 , @xmath59 for all @xmath60 . when @xmath56 , @xmath55 by definition [ def : anchored ] , so every term in the numerator of definition [ def : rfunction ] is zero except the coefficient of @xmath58 . since we know the coefficients have sum 1 , it follows that @xmath61 . [ lem : alpha ] in a positive , anchored hidden markov model , the @xmath48-functions @xmath62 , for each @xmath63 , are uniformly bounded below by some @xmath64 , that is , @xmath65 for all @xmath0 and @xmath18 . we can write @xmath66 by definitions [ def : transitionfunction ] and [ def : anchored ] , which is bounded below by @xmath67 since @xmath68 . since each @xmath54 , if all the entries of @xmath69 are positive , then @xmath70 is bounded below uniformly in @xmath18 for fixed @xmath0 , which then implies a uniform bound in @xmath18 and @xmath0 since there are only finitely many @xmath0 . [ def : orbit ] for each state @xmath71 , the * orbit * @xmath72 of @xmath73 under the @xmath2-functions is @xmath74 by requiring the @xmath48-functions to be positive , we exclude points in the orbit which are reached with zero probability . let @xmath75 . [ prop : discreteness ] in a positive , anchored hidden markov model , there exists a constant @xmath76 such that for all measures @xmath77 , the mass of the measure @xmath78 outside @xmath79 is bounded by @xmath80 , that is , @xmath81 . we can rewrite definition [ def : transitionfunction ] as @xmath82 where @xmath83 in this notation , the integral is the lebesgue integral of the function @xmath84 with respect to the measure @xmath85 . since @xmath84 takes values in the @xmath44 and @xmath85 is a probability , the integral also takes values in @xmath86 , thus @xmath87 maps the information state space @xmath44 to itself . since @xmath88 is a measure supported on the set of points reachable from @xmath18 via an @xmath2-function , and @xmath79 is a union of orbits of @xmath2-functions and therefore closed under @xmath2-functions , it follows that all mass in @xmath79 is mapped back into @xmath79 under the evolution function , that is @xmath89 on the other hand , by lemma [ lem : anchor ] , @xmath90 for all @xmath18 , hence @xmath91 putting these together gives @xmath92 setting @xmath93 gives @xmath94 , hence @xmath81 by induction . by lemma [ lem : alpha ] , @xmath95 , while @xmath96 since we can always choose a larger value . up to this point , we have considered the evolution function as a deterministic function @xmath45 . however , we can also consider it as a probabilistic function @xmath97 . by definition [ def : transitionfunction ] , @xmath87 maps points in @xmath79 to @xmath79 , hence the restriction @xmath98 gives a probabilistic function , and therefore a markov chain , with countable state space @xmath79 . by proposition [ prop : discreteness ] , the limiting behaviour of the information chain takes place almost entirely in @xmath79 in some sense , so we would expect that convergence of the restricted information chain @xmath99 is sufficient for convergence of the full information chain @xmath87 . this is proved below . [ prop : positiverecurrent ] in a positive , anchored hidden markov model , under any policy , the chain @xmath99 has at least one state of the form @xmath58 which is positive recurrent , that is , whose expected return time is finite . construct a markov chain @xmath100 on the set @xmath101 , with transition probabilities @xmath102 for all @xmath0 , @xmath103 whenever @xmath104 is nonempty , and all other transition probabilities zero . we note that this is possible since we allow each state a positive probability transition to some other state . since @xmath100 is a finite state markov chain , it must have a recurrent state . each state @xmath72 can reach some state @xmath105 , so some state @xmath105 is recurrent ; call it @xmath106 . consider a state @xmath107 of the chain @xmath87 which is reachable from @xmath108 , where @xmath109 is a composition of @xmath110 @xmath2-functions with corresponding @xmath48-functions nonzero . since the @xmath105 partition @xmath24 , one of them must contain @xmath111 ; call it @xmath112 . we will assume @xmath113 ; the proof follows the same argument and is simpler in the case when @xmath114 . by definition of the @xmath48-functions , @xmath115 this means that @xmath116 is reachable from @xmath108 in the chain @xmath87 , hence in the chain @xmath100 , @xmath112 is reachable from @xmath106 , and by recurrence of @xmath106 , @xmath106 must also be reachable from @xmath112 via some sequence of positive probability transitions @xmath117 by definition of @xmath118 , @xmath119 is nonempty , and thus contains some point @xmath120 , where @xmath121 is a composition of @xmath122 @xmath2-functions with corresponding @xmath48 nonzero . by definition [ def : orbit ] , each transition @xmath123 to @xmath124 in the information chain occurs with positive probability , so @xmath125 since @xmath126 , by anchoredness and positivity , @xmath127 the markov property then gives @xmath128 continuing via the sequence ( [ eqn : path ] ) , we obtain @xmath129 thus , for every state @xmath130 reachable from @xmath108 , we have found constants @xmath131 and @xmath132 such that @xmath133 by lemma [ lem : alpha ] , @xmath134 is uniformly bounded below , while @xmath135 depends only on the directed path ( [ eqn : path ] ) and not on @xmath18 , and thus is also uniformly bounded below since there are only finitely many @xmath105 , and hence it suffices to choose finitely many such paths . similarly , @xmath136 also depends only on the directed path ( [ eqn : path ] ) , and thus is uniformly bounded above . in particular , it is possible to pick @xmath136 and @xmath137 such that @xmath138 and @xmath139 . let @xmath140 be the first entry time into the state @xmath108 . by the above bound , we have @xmath141 for any initial state @xmath18 reachable from @xmath108 . letting @xmath142 and @xmath143 be independent copies of @xmath140 and @xmath85 , the time - homogeneous markov property gives @xmath144 & \hspace{2cm}\times{\mathbb p}\big(z_{ks}=z'\big|\tau > ks , z_0=z\big)\nonumber\\[.2 cm ] & \le\sup_{z'}{\mathbb p}\big(\tau>(k+1)s\big|\tau > ks , z_{ks}=z'\big)\nonumber\\ & = \sup_{z'}{\mathbb p}\big(\tau'>s\big|z'_0=z'\big)\nonumber\\ & \le1-c.\end{aligned}\ ] ] by induction , @xmath145 for all @xmath146 . dropping the condition on the initial distribution @xmath147 for convenience , we have @xmath148 & = \sum_{k\in{\mathbb z}^+}{\mathbb p}(\tau > k)=\sum_{k\in{\mathbb z}^+}\sum_{0\le t < s}{\mathbb p}(\tau > ks+t)\nonumber\\ & \le\sum_{k\in{\mathbb z}^+}\sum_{0\le t < s}{\mathbb p}(\tau > ks)\le\sum_{k\in{\mathbb z}^+}s(1-c)^k=\frac sc<\infty.\end{aligned}\ ] ] in particular , @xmath149<\infty$ ] , so @xmath108 is a positive recurrent state . [ lem : linearity ] the transition function @xmath87 , considered as an operator on the real banach space @xmath86 of signed radon measures on @xmath24 with the total variation norm , is linear with operator norm @xmath150 . linearity follows immediately from the fact that @xmath87 is defined as a finite sum of integrals . for each @xmath151 , let @xmath152 be the hahn decomposition , so that @xmath153 by definition of the total variation norm . if @xmath154 , then @xmath155 by linearity of @xmath87 . otherwise , let @xmath156 , so that @xmath157 . since @xmath87 maps probability measures to probability measures , @xmath158 , and similarly for @xmath159 , hence by linearity of @xmath87 and the triangle inequality , @xmath160 this shows that @xmath161 . picking @xmath162 to be any probability measure gives @xmath163 , hence @xmath150 . [ thm : convergence2 ] in a positive , anchored hidden markov model , irreducibility and aperiodicity of the restricted information chain @xmath99 is a sufficient condition for convergence in distribution of the information state @xmath37 to some discrete invariant measure @xmath164 . in this case , the restricted chain @xmath99 is irreducible , aperiodic and positive recurrent that is , ergodic and hence has a unique invariant probability measure @xmath165 , which can be considered as an element of @xmath44 supported on @xmath166 . we will show that the information distribution @xmath167 converges in total variation norm to @xmath165 . fix @xmath168 and @xmath64 , and pick @xmath169 such that @xmath170 , with @xmath171 as defined in proposition [ prop : discreteness ] . let @xmath172 be the restriction of the probability measure @xmath173 to @xmath79 , which is a positive measure with total mass @xmath174 , hence we can divide to obtain the probability measure @xmath175 supported on @xmath79 . since @xmath99 is ergodic and @xmath176 is a probability measure , @xmath177 converges to @xmath165 in total variation norm , so pick @xmath178 such that for all @xmath179 , @xmath180 . for any @xmath181 , we have the triangle inequality bound @xmath182 by lemma [ lem : linearity ] , @xmath87 is linear with operator norm 1 , so by proposition [ prop : discreteness ] , @xmath183 since @xmath184 , again using linearity and the fact that @xmath185 , @xmath186 finally , again using proposition [ prop : discreteness ] @xmath187 we see that for all @xmath181 , @xmath188 , so the information state @xmath167 converges to @xmath165 in total variation norm and therefore in distribution . the conditions of theorem [ thm : convergence2 ] can be weakened to the case when the underlying chain is only ergodic . given an ergodic finite transition matrix @xmath69 , some power @xmath189 will be positive , and thus we should still have convergence by taking time in @xmath146-step increments , since the information state will not fluctuate too much within those @xmath146 steps . the difficulty lies in the fact that the information state taken in @xmath146-step increments is not the same as the information state of the @xmath146-step chain . it is our belief that the information state converges in all but a small number of pathological examples ; however , we are only able to prove it in the above cases . if the information state does not converge , then it does not make sense to consider a limiting expected entropy . however , it is possible that a cesro sum of the expected entropy converges , and the limsup and liminf will certainly exist . alternatively , we could simply work with a finite time horizon . continuing further within the general case has proven to be quite difficult , so we will restrict the remainder of our results to a special case , where there are two states and two observation processes with two possible observations each , and each observation process observes a different underlying state perfectly . formally , @xmath190 , and the transition and observation matrices are @xmath191 in order to exclude trivial cases , we will require that the parameters @xmath192 , @xmath193 , @xmath194 and @xmath195 are contained in the open interval @xmath196 , and @xmath197 . note that this special case exactly corresponds to the problem of searching for a moving target studied by macphee and jordan @xcite , although our cost function , limiting expected entropy , is very different from theirs , expected cumulative sum of prescribed costs until the first zero - entropy observation . we will begin by proving that given this restriction , the information state always converges in distribution , except for one case which is pathological in a sense that will be explained later . this proof is both a correction and an improvement of the proof given in @xcite . in the special case , the space @xmath24 is a 1-simplex embedded in @xmath198 , which we can identify with the interval @xmath199 $ ] , via equating each point @xmath200 $ ] with the point @xmath201 , so that @xmath18 represents the mass at 0 in the distribution . by substituting these parameters into into definition [ def : transitionfunction ] , the transition function in the special case is @xmath202 where : * @xmath203 ; * @xmath204 ; * @xmath205 ; and * @xmath206 . note that the second line of ( [ eqn : specialcase ] ) consists of two point masses at 0 and 1 , which is a feature of the anchoredness condition . in the special case , it allows us to represent the location of masses by only two @xmath2-functions . we will continue to use the symbols @xmath207 , @xmath208 , @xmath209 and @xmath210 in their meaning above for the remainder of our discourse . note that we have simplified the notation , in that @xmath209 represents the @xmath2-function @xmath211 , while @xmath212 does not appear since @xmath213 identically , and similarly for the other symbols . since the special case satisfies the conditions of positivity and anchoredness , proposition [ prop : positiverecurrent ] applies , so irreducibility and aperiodicity of the information chain are sufficient for ergodicity . we now show that this occurs in all but two exceptional cases : * case 1 : each orbit is contained entirely in the policy region which maps to that orbit , that is , @xmath214 and @xmath215 . note that by ( [ eqn : specialcase ] ) , the @xmath48-functions are always strictly positive , so we must have @xmath216 and @xmath217 . * case 2 : the orbits alternate periodically between policy regions , that is , @xmath218 and @xmath219 , where @xmath2 is the combined @xmath2-function @xmath220 . let case 0 denote the general case when neither case 1 nor case 2 occurs . [ lem : irreducible ] the chain @xmath99 has only one irreducible recurrent class , except in case 1 , where it splits into two irreducible recurrent classes , both of which are positive recurrent . by proposition [ prop : positiverecurrent ] , without loss of generality , assume that the state 0 is positive recurrent . if 1 is reachable from 0 , that is , there is some @xmath39 such that @xmath221 , then 0 is also reachable from 1 since 0 is recurrent , hence 0 and 1 are in the same irreducible recurrent class . by ( [ eqn : specialcase ] ) , either 0 or 1 is reachable from every state , so there can not be any other recurrent classes . otherwise , if 0 is reachable from 1 but 1 is not reachable from 0 , then 1 is transient , and furthermore , all of @xmath222 is transient since any @xmath223 is reachable only via the transient state 1 , while all of @xmath224 is reachable from the recurrent state 0 and hence forms the only irreducible recurrent class . finally , if 0 and 1 are both unreachable from each other , then it must be the case that @xmath214 and @xmath215 , in which case the chain splits into two irreducible classes @xmath224 and @xmath222 , both of which are positive recurrent by the argument in proposition [ prop : positiverecurrent ] . [ lem : aperiodic ] the recurrent classes are aperiodic , except in case 2 , where the chain consists of a single irreducible recurrent class with period 2 . if any recurrent class is periodic , then at least one of 0 or 1 is recurrent and periodic ; without loss of generality suppose it is 0 . since 0 is periodic , it can not reach itself in 1 step , so must be contained in @xmath106 , thus 1 is reachable from 0 and hence is in the same irreducible recurrent class . note that by the same argument , @xmath225 , thus 0 reaches itself in 2 steps and hence the period must be 2 . this means 0 can not reach itself in an odd number of steps , so @xmath226 when @xmath146 is even and @xmath227 when @xmath146 is odd , and similarly for the orbit of 1 , which is the only possibility of periodicity . thus , there is one exception in which the information chain is periodic , and another in which it is reducible . as will be evident , both exceptions can be remedied . we begin by giving a class of policies under which reducibility can not occur . this class of policies is simple enough to be easily analysed , and as will be conjectured later , always includes the optimal policy in any hidden markov model within the special case . [ def : thresholdpolicy ] a policy @xmath10 is called a * threshold policy * if its preimages @xmath228 and @xmath229 are both intervals . a threshold policy is indeed given by a threshold , since there must be some boundary point between @xmath230 and @xmath106 , such that one observation process is used on one side and the other is used on the other side . outside the threshold case , it is unclear what the equivalent definition would be , since the concept of an interval does not generalise easily to higher dimensions . [ lem : r0bigger ] the linear fractional transformations @xmath209 and @xmath210 satisfy the inequality @xmath231 for all @xmath200 $ ] . we can write @xmath232 . note that the coefficient @xmath233 of @xmath194 is strictly positive and in the denominator , while @xmath234 is exactly the same with @xmath235 instead of @xmath194 . since @xmath236 , @xmath237 , hence @xmath231 . [ lem : monotonic ] the linear fractional transformations @xmath209 and @xmath210 are both strictly increasing when @xmath238 and both strictly decreasing when @xmath239 . the derivative of @xmath209 is @xmath240 , which is positive everywhere if @xmath238 and negative everywhere if @xmath239 . the same holds for @xmath210 , since it is identical with @xmath235 instead of @xmath194 . [ lem : fixedpoint ] the linear fractional transformations @xmath209 and @xmath210 have unique fixed points @xmath241 and @xmath242 , which are global attractors of their respective dynamical systems . split the interval @xmath199 $ ] into open subintervals at its interior fixed points , noting that by inspection , the two boundary points are not fixed . since linear fractional transformations have at most two fixed points , there must be between one and three such subintervals , all non - degenerate . by continuity , in any such subinterval @xmath243 , either @xmath244 for all @xmath245 , or @xmath246 for all @xmath245 . since @xmath247 , @xmath244 for all points @xmath18 in the leftmost subinterval , and since @xmath248 , @xmath246 for all points @xmath18 in the rightmost subinterval . this means that there are at least two subintervals . if there were three subintervals , @xmath249 , @xmath250 and @xmath251 in that order , then either @xmath244 for all @xmath252 , or @xmath246 for all @xmath252 . in the first case , the fixed point between @xmath249 and @xmath250 is attracting on the left and repelling on the right , and in the second case , the fixed point between @xmath250 and @xmath251 is repelling on the left and attracting on the right . however , such fixed points can only occur for parabolic linear fractional transformations with only one fixed point . this is a contradiction , hence there can not be three subintervals . thus , there are exactly two subintervals , and exactly one fixed point , which must then be an attractor across all of @xmath199 $ ] . [ lem : eta0bigger ] the fixed points satisfy @xmath253 . by lemma [ lem : r0bigger ] , @xmath254 . first consider the case @xmath238 . applying lemma [ lem : monotonic ] @xmath146 times gives @xmath255 , so the orbit of @xmath241 under @xmath210 is monotonically decreasing , but it also converges to @xmath242 by lemma [ lem : fixedpoint ] , hence @xmath253 . in the remaining case @xmath239 , suppose @xmath256 . then by lemma [ lem : monotonic ] , @xmath257 , which is a contradiction , hence @xmath253 . [ prop : exception ] the first exception to theorem [ thm : convergence2 ] , case 1 , can not occur under a threshold policy . suppose @xmath214 , @xmath215 , and the policy is threshold . since @xmath258 and @xmath259 , it follows that every point of @xmath224 is less than every point of @xmath222 . since @xmath224 and @xmath222 are the orbits of 0 and 1 under @xmath209 and @xmath210 respectively , by lemma [ lem : fixedpoint ] , they have limit points @xmath241 and @xmath242 respectively , hence @xmath260 . this contradicts lemma [ lem : eta0bigger ] , hence this can not occur . the remaining exception is when the information chain is periodic with period 2 , in which case the expected entropy oscillates between two limit points . the limiting expected entropy can still be defined in a sensible way , by taking the average , minimum or maximum of the two limit points , depending on which is most appropriate for the situation . thus , for threshold policies , it is possible to define optimality without exception . we conclude this section by writing down a closed form general expression for the limiting expected entropy . [ prop : entropyformula ] under the conditions of theorem [ thm : convergence2 ] , that is , in case 0 , the limiting expected entropy of a policy is given by @xmath261 where , for @xmath262 : * @xmath263 is the entropy function and @xmath264 is the constant function with value 1 ; * @xmath220 and @xmath265 are the combined @xmath2-function and combined @xmath48-function respectively , with @xmath209 , @xmath210 , @xmath207 and @xmath208 defined as in ( [ eqn : specialcase ] ) ; * @xmath266 $ ] are defined by the recursion @xmath267 * @xmath268 are defined by the recursion @xmath269 * @xmath270 is the linear functional @xmath271 by theorem [ thm : convergence2 ] , a unique invariant probability measure @xmath272 exists , so it suffices to show that @xmath273 where @xmath274 is the dirac measure , and @xmath275 is extended in the obvious way to a linear functional @xmath276 . by proposition [ prop : discreteness ] , the invariant measure is supported in the combined orbit set @xmath79 , and @xmath277 is the only point which can make a one - step transition under the restricted information chain @xmath99 to @xmath278 , with probability @xmath279 . since the masses at @xmath277 and @xmath278 are invariant , we must have @xmath280 it then follows that for some constants @xmath281 , @xmath282 the mass at @xmath277 is @xmath283 , which makes a transition under @xmath99 to 0 in one step with probability @xmath284 . since @xmath272 is an invariant measure , mass is conserved at @xmath285 , hence @xmath286 since @xmath287 as @xmath288 , by telescoping series , @xmath289 multiplying the right hand side of ( [ eqn : conservationofmass ] ) by ( [ eqn : telescope ] ) with @xmath290 , then collecting coefficients of @xmath291 and @xmath292 , yields @xmath293 note that conservation of mass at @xmath294 is now automatic , since mass is conserved on all of @xmath79 and at every other point in @xmath79 . the second equation comes from requiring @xmath272 to have total mass 1 , that is , @xmath295 the solution to ( [ eqn : conservationofmass2 ] ) and ( [ eqn : totalmassone ] ) is exactly the required result . we note that the denominator is zero only when @xmath296 and @xmath297 for all @xmath298 , which is exactly the excluded case @xmath214 and @xmath215 . this proof can be generalised easily to the case of more than two observation processes , as long as each one is anchored with only two states . we now present computational results , again in the special case , which will illustrate the nature of the optimal policy in the case of minimising limiting expected entropy . since any such discussion must cover how best to calculate the limiting expected entropy given a hidden markov model and a policy , this is a natural place to start . the simpliest approach is to use the formula in proposition [ prop : entropyformula ] , noting that each of @xmath299 , @xmath300 , @xmath301 and @xmath302 is a product of two infinite series , each of which is bounded by a geometric series and hence well - approximated by truncation . specifically , we write the limiting expected entropy as @xmath303 where : * @xmath304 ; * @xmath305 ; * @xmath306 . we can simplify the calculation slightly by recursively updating @xmath307 and @xmath308 , storing them as real - valued variables @xmath309 and @xmath310 . [ alg : entropy ] estimation of limiting expected entropy . 1 . define as functions the entropy @xmath263 , the mixed @xmath2-function @xmath311 and the mixed @xmath48-function @xmath312 ; 2 . pick a large number @xmath313 ; 3 . initialise variables @xmath314 , @xmath315 , @xmath316 , @xmath317 , @xmath318 and @xmath319 to 0 , @xmath320 and @xmath321 to 1 , @xmath111 to 0 and @xmath322 to 1 ; 4 . repeat the following @xmath323 times : 1 . add @xmath320 to @xmath318 and @xmath321 to @xmath319 ; 2 . add @xmath324 to @xmath316 and @xmath325 to @xmath317 ; 3 . if @xmath114 , then add @xmath326 to @xmath314 , and if @xmath327 , then add @xmath328 to @xmath315 ; 4 . let @xmath329 and @xmath330 ; 5 . multiply @xmath320 by @xmath331 and @xmath321 by @xmath332 ; 5 . the limiting expected entropy @xmath333 is estimated by @xmath334 [ prop : errorbound ] the estimate @xmath335 satisfies the error bound @xmath336 where @xmath337 is the denominator of @xmath335 , and @xmath338 the formula for @xmath48 follows from definition [ def : transitionfunction ] , since the @xmath48-functions are linear and hence their maxima occur at the endpoints @xmath339 or @xmath340 . note that @xmath341 follows from the requirement that @xmath342 . since @xmath343 is monotonic increasing , this implies that the error bound is finite and vanishes as @xmath344 . since each series has non - negative summands , each truncated series is smaller than the untruncated series , hence @xmath345 the @xmath146th summand in each series is bounded by @xmath346 , hence @xmath347 the same bound holds for @xmath348 , hence @xmath349 similarly , since @xmath350 , @xmath351 combining via the triangle inequality gives the required bound . note that this bound depends only on the quantities @xmath48 and @xmath343 , which are easily calculated . in particular , this allows us to prescribe a precision and calculate the limiting expected entropy to within that precision by running the algorithm with unbounded @xmath313 and terminating when error bound reaches the desired precision . furthermore , since @xmath343 appears only in the denominator and grows monotonically with @xmath313 , we can replace it with @xmath352 for some small , fixed @xmath353 to calculate a prior a sufficient number of steps for any given precision . [ eg : errorbound ] in later simulations , we will pick @xmath354 $ ] , so that @xmath355 . this gives an error bound of @xmath356 while the constant appears daunting at first glance , solving for a prescribed error of @xmath357 gives @xmath358 hence , we require @xmath359 for any realistic value of @xmath343 , this is easily within computational bounds , as each iteration requires at most 36 arithmetic operations , 2 calls to the policy function , and 4 calls to the logarithm function . an alternative approach to estimating limiting expected entropy would be to simulate observation of the hidden markov model under the given policy . the major drawback of this method is that it requires working with the information state @xmath37 , which takes values in @xmath360)$ ] , the set of probability measures on the unit interval , which is an infinite dimensional space . one possibility is to divide @xmath199 $ ] into subintervals and treat each subinterval as a discrete atom , but this produces a very imprecise result . even using an unrealistically small subinterval width of @xmath361 , the entropy function has a variation of over @xmath362 across the first subinterval , restricting the error bound to this order of magnitude regardless of the number of iterations . in comparison , example [ eg : errorbound ] shows that the direct estimation method has greatly superior performance . an improvement is to use the fact that the limiting distribution @xmath272 is discrete , and store a list of pairs containing the locations and masses of discrete points . since any starting point moves to either 0 or 1 in one step , at the @xmath313th iteration , the list of points must contain at least the first @xmath313 points in the orbit of either 0 or 1 . each such point requires a separate calculation at each iteration , and thus the number of computations is @xmath363 rather than @xmath364 as for algorithm [ alg : entropy ] . since the number of iterations @xmath313 corresponds to the last point in the orbit of 0 or 1 which is represented , for any given @xmath313 , this method differs from the direct computation method only in the masses on these points , thus we would expect the relationship between precision and number of iterations to be similar . since the simulation method has quadratically growing number of computations , this would suggest that it is slower than the direct computation method , and indeed , this is also indicated by empirical trials . we will use the direct computation method of estimating limiting expected entropy for all of our results . the problem of finding the policy which minimises limiting expected entropy is made much easier by restricting the search space to the set of threshold policies , as these can be represented by a single number representing the threshold , and a sign representing which observation process is used on which side of the threshold . the simplest approach is to pick a collection of test thresholds uniformly in @xmath199 $ ] , either deterministically or randomly , and test the limiting expected entropy at these thresholds , picking the threshold with minimal entropy as the optimal threshold policy . however , this method is extremely inefficient . proposition [ prop : discreteness ] shows that the policy only matters on the countable set @xmath365 $ ] , so moving the threshold does not change the policy as long as it does not move past a point in @xmath79 . as shown in figures [ fig : regioni][fig : regionvi ] , points in @xmath79 tend to be quite far apart , and thus the naive approach will cause a large number of equivalent policies to be tested . on the other hand , points in @xmath79 close to the accumulation points are closely spaced , so even with a very fine uniform subset , some policies will be missed when the spacing between points in @xmath79 becomes less than the spacing between test points . a better way is to decide on a direction in which to move the threshold , and select the test point exactly at the next point in the previous realisation of @xmath79 in the chosen direction , so that every threshold in between the previous point and the current point gives a policy equivalent to the previous policy this ensures that each equivalence class of policies is tested exactly once , thus avoiding the problem with the naive method . however , a new problem is introduced in that the set of test points depends on the iteration number @xmath313 , which determines the number of points of @xmath79 that are considered . this creates a circular dependence , in that the choice of @xmath313 depends on the desired error bound , the error bound depends on the policy , and the set of policies to be tested depends on @xmath313 . we can avoid this problem by adapting proposition [ prop : errorbound ] to a uniform error bound across all threshold policies . [ prop : errorboundthreshold ] for a threshold policy and @xmath366 , the error is @xmath367 where @xmath368 is the smallest integer such that @xmath369 , and @xmath370 first note that @xmath368 exists , since by lemma [ lem : eta0bigger ] , iterations of @xmath209 and @xmath210 converge to @xmath371 respectively . using proposition [ prop : errorbound ] , it suffices to prove that for @xmath366 , @xmath372 it is not possible for @xmath373 and @xmath374 , since this would mean @xmath375 and @xmath376 , which gives the ordering @xmath377 , but @xmath230 and @xmath106 are intervals for a threshold policy . hence , either @xmath378 or @xmath379 for some @xmath380 . if @xmath381 , then @xmath382 . since @xmath383 , this gives @xmath384 , as required . a similar argument holds in the case @xmath385 . the existence of this uniform bound for @xmath343 in the threshold case is closely related to proposition [ prop : exception ] , which states that the exception case 1 , where @xmath386 and @xmath387 , can not occur in a threshold policy . in this exceptional case , proposition [ prop : entropyformula ] does not hold , as the denominator is zero , and hence @xmath388 for all @xmath313 . the fact that this can not occur in a threshold policy is the key ingredient of this uniform bound . now that we have an error bound which does not depend on the policy , we can determine a uniform number of iterations @xmath313 that will suffice for estimating the limiting expected entropy for all threshold policies . this reduces the search space to a finite one , as each point in the orbits of 0 and 1 must be in one of the two policy regions , hence , there are at most @xmath389 policies . most of these will not be threshold policies , but since orbit points need not be ordered , there is no obvious bound on the number of threshold policies that need to be checked . simulation results later in this section will show that in most cases , the number of such policies is small enough to be computationally feasible . [ def : orientation ] the * orientation * of a threshold policy is the pair of whether @xmath230 is to the left or right of @xmath106 , and whether the threshold is included in the left or right interval . let @xmath390 $ ] , @xmath391(a_1]$ ] , @xmath392 $ ] and @xmath393(a_0]$ ] denote the four possibilities , with the square bracket indicating inclusion of the threshold and round bracket indicating exclusion . our strategy for simplifying the space of threshold policies that need to be considered is to note that the policy only matters on @xmath79 , the support of the invariant measure . although @xmath79 depends on the policy , for a given orientation and threshold @xmath39 , any policy with the same orientation and some other threshold @xmath394 such that no points of @xmath79 lie between @xmath39 and @xmath394 is an equivalent policy , in that sense that the invariant measure is the same , since no mass exists in the region where the two policies differ . thus , for each orientation , we can begin with @xmath395 , and at each iteration , move the threshold left past the next point in @xmath79 , since every threshold in between is equivalent to the previous threshold . although @xmath79 changes at each step , this process must terminate in finite time since we already showed that there are only finitely many policies for any given @xmath313 , and by testing equivalence classes of policies only once , it is likely to that far fewer steps are required than the @xmath389 bound . furthermore , since @xmath79 is a discrete set , every threshold policy has an interval of equivalent threshold policies , so we can assume without loss of generality that the threshold is contained in the interval to the right , that is , only test the orientations @xmath390 $ ] and @xmath392 $ ] . [ alg : optimalthreshold ] finding the optimal threshold policy . 1 . find @xmath368 , the smallest integer such that @xmath369 , by repeated application of @xmath209 and @xmath210 to 0 and 1 respectively ; 2 . prescribe an error @xmath396 and determine the number of iterations @xmath397 3 . start with the policy @xmath398 and @xmath399 $ ] with @xmath395 , that is , @xmath230 is the whole interval and @xmath106 is empty but considered as being to the right of @xmath230 , and loop until @xmath400 : 1 . run algorithm [ alg : entropy ] , and let the next value of @xmath39 be the greatest @xmath308 which is strictly less than the previous value of @xmath39 ; 2 . if entropy is less than the minimum entropy for any policy encountered so far , record it as the new minimum ; 4 . repeat with @xmath230 considered to be on the right of @xmath106 . we now calculate the location of the optimal threshold for a range of values of the parameters @xmath192 , @xmath193 , @xmath194 and @xmath195 . the results of this calculation , which will be presented in figure [ fig : thresholdregions ] , is the primary content of this section , as it will give an empirical description of the optimal threshold policy . since we have not been able to prove optimality analytically , except in the symmetric case of proposition [ prop : symmetric ] , this empirical description will provide our most valuable insight into the problem . in order to facilitate understanding , we will define six classes of threshold policies , depending on the location and orientation of the threshold . diagrams of the invariant measure under each of these classes , presented in figures [ fig : regioni][fig : regionvi ] , will demonstrate that these classes of thresholds are qualitatively different , which will further manifest itself in the next section . we partition the interval into three subintervals with endpoints at the attractors @xmath241 and @xmath242 , noting that this produces three regions which consist entirely of a single equivalence class of policy . we also assign colours to these regions to accommodate the data presented in figure [ fig : thresholdregions ] . [ def : thresholdregions ] the six * threshold regions * are defined by partitioning the set of all threshold policies first by whether the orientation is @xmath390 $ ] or @xmath392 $ ] , then by the location of the threshold @xmath39 in relation to the accumulation points @xmath242 and @xmath241 , with inclusion of endpoints defined more precisely overleaf . we number them using roman numerals i through vi . note that if either @xmath401 or @xmath402 , then either @xmath230 or @xmath106 occupies the whole interval , depending on the orientation . in particular , @xmath390 $ ] with @xmath400 is equivalent to @xmath392 $ ] with @xmath395 , since in both cases @xmath106 is the whole interval , and similarly @xmath390 $ ] with @xmath395 is equivalent to @xmath392 $ ] with @xmath400 . thus , the space of all possible threshold policies consists of two disjoint intervals @xmath403 $ ] , each of whose endpoints is identified with an endpoint of the other interval , which is topologically equivalent to a circle . to be technically correct , we note that identifying 0 and @xmath404 does not present a problem , since we can simply extend the interval to @xmath405 $ ] for some small @xmath64 , and identify 0 and @xmath406 instead . while this results in a subinterval of the circle corresponding to the same policy , this does not add additional complexity since every threshold policy has an interval of equivalent policies . this topology is illustrated overleaf in figure [ fig : thresholddomain ] , followed by precise definitions and examples of the six threshold regions in figures [ fig : regioni][fig : regionvi ] , and finally our computational results in figure [ fig : thresholdregions ] . ( i ) ( ii ) ( iii ) + $ ] , while the bottom line represents the orientation @xmath392 $ ] . the right end of the top line and the left end of the bottom line are both the policy @xmath407 $ ] and @xmath408 , so we can paste them together , and similarly for the left end of the top line and the right end of the bottom line . hence , we see that the set of threshold policies is topologically a circle.,title="fig : " ] + 0 @xmath242 @xmath241 1 + $ ] , while the bottom line represents the orientation @xmath392 $ ] . the right end of the top line and the left end of the bottom line are both the policy @xmath407 $ ] and @xmath408 , so we can paste them together , and similarly for the left end of the top line and the right end of the bottom line . hence , we see that the set of threshold policies is topologically a circle.,title="fig : " ] + ( iii ) ( iv ) ( v ) ( vi ) ( i ) + * region i * ( represented by red in figure [ fig : thresholdregions ] ) : @xmath390 $ ] with @xmath409 or @xmath392 $ ] with @xmath402 . when @xmath238 , every policy here is equivalent to the all-@xmath106 policy , as the mass in the orbit of 0 approaches 0 . , @xmath410 , @xmath411 , @xmath412 . entropy is 0.3145 . under the evolution function , any mass eventually enters the grey region since it contains both accumulation points in its interior , after which it can not escape , hence in the limit , there is zero mass in the orbit of 0 , and the policy is equivalent the all-@xmath106 policy . ] * region ii * ( yellow ) : @xmath390 $ ] with @xmath413 . this is the most difficult to understand of the threshold policies , as the orbits do not converge to the accumulation points @xmath241 and @xmath242 , but rather , oscillate around the threshold @xmath39 . , @xmath410 , @xmath411 , @xmath412 . entropy is 0.3251 . note that the masses do not converge to the accumulation points . ] * region iii * ( green ) : @xmath390 $ ] with @xmath414 or @xmath392 $ ] with @xmath401 . when @xmath238 , every policy here is equivalent to the all-@xmath230 policy , since the mass in the orbit of 1 approaches 0 . , @xmath410 , @xmath411 , @xmath412 . entropy is 0.3337 . under the evolution function , any mass eventually enters the white region since it contains both accumulation points in its interior , after which it can not escape , hence in the limit , there is zero mass in the orbit of 1 , and the policy is equivalent to the all-@xmath230 policy . ] * region iv * ( cyan ) : @xmath392 $ ] with @xmath415 . for a policy in this region , the first finitely many points of the orbit of 0 belongs to @xmath106 , while every other point lies in @xmath230 . note that @xmath401 is included in region iii . , @xmath410 , @xmath411 , @xmath412 . entropy is 0.3275 . the orbit of 1 approaches @xmath241 , while the orbit of 0 follows an approach sequence to @xmath242 for a finite number of steps ( in this case 2 steps ) before also approaching @xmath241 . ] * region v * ( blue ) : @xmath392 $ ] with @xmath416 . when @xmath238 , every policy here is equivalent to the policy @xmath417 and @xmath418 . , @xmath410 , @xmath411 , @xmath412 . entropy is 0.3265 . note that the orbit of 0 converges to @xmath242 while the orbit of 1 converges to @xmath241 . ] * region vi * ( magenta ) : @xmath392 $ ] with @xmath419 . for a policy in this region , the first finitely many points of the orbit of 0 belongs to @xmath106 , while every other point lies in @xmath230 . note that @xmath402 is included in region i. the lack of symmetry with region vi in terms of strict and non - strict inequalities is due to the choice that the threshold itself be included in the right region . , @xmath410 , @xmath411 , @xmath412 . entropy is 0.3182 . the orbit of 0 approaches @xmath242 , while the orbit of 1 follows an approach sequence to @xmath241 for a finite number of steps ( in this case 1 step ) before also approaching @xmath242 . ] on the major horizontal axis increasing rightwards , @xmath193 on the major vertical axis increasing downwards , @xmath194 on the minor horizontal axis increasing rightwards and @xmath195 on the minor vertical axis increasing downwards . each parameter is sampled by @xmath420 , omitting the trivial cases when @xmath421 , resulting in 152000 sample points . colours representing regions are defined in figures [ fig : regioni][fig : regionvi ] . ] the bound in proposition [ prop : errorboundthreshold ] is consistent with the empirical running time for the calculations used to generate figure [ fig : thresholdregions ] . our program experienced significant slowdowns when either @xmath192 and @xmath194 , or @xmath193 and @xmath195 were large , with the slowest occuring at the extreme point @xmath422 . this suggests that a uniform bound when @xmath423 is impossible , and that algorithm [ alg : optimalthreshold ] will fail for sufficiently large values of @xmath48 . on the other hand , either @xmath424 or @xmath425 results in trivial limiting behaviour of the underlying chain , so this situation is unlikely to occur in any practical application . note that in figure [ fig : thresholdregions ] , the half of the main diagonal where @xmath238 is entirely blue . this can be proven analytically . we note that the condition @xmath238 corresponds to the underlying markov chain having positive one - step autocorrelation , which is a reasonable assumption when the frequency of observation is greater than the frequency of change in the observed system , since in this case , one would not expect the system to oscillate with each observation . [ prop : symmetric ] in the symmetric case @xmath426 and @xmath427 , with @xmath238 , the optimal general policy is the unique region v policy @xmath428 given by @xmath417 and @xmath418 . proposition [ prop : entropyformula ] gives the limiting expected entropy as a convex combination of two quantities , hence @xmath429 note that equality is realised when the two quantities above are equal , which occurs under @xmath428 , since in that case @xmath430 and @xmath431 . let @xmath432 , @xmath433 and @xmath434 . in order to prove that @xmath428 is optimal , by symmetry , it suffices to prove that it minimises @xmath435 . first , we show that for each @xmath146 and any other policy @xmath10 , @xmath436 , where the notation @xmath437 makes the dependence on the policy explicit . by lemma [ lem : r0bigger ] , @xmath438 . by lemma [ lem : monotonic ] and our assumption that @xmath238 , iterations of @xmath209 and @xmath210 approach their limits monotonically , hence @xmath439 and @xmath440 . these inequalities are illustrated in figure [ fig : symmetric ] . combining these inequalities gives @xmath441 . entropy is concave on the interval @xmath199 $ ] , and therefore on the subinterval @xmath442 $ ] , hence @xmath443 since @xmath444 and entropy is increasing on @xmath445 $ ] , @xmath446 , which is equal to @xmath447 by symmetry . hence , the inequality ( [ eqn : symmetricineq ] ) reduces to @xmath448 , that is , @xmath449 . and @xmath450 in relation to 0 , @xmath242 , @xmath241 and 1 . all positions are fixed except that @xmath450 may be to the left of @xmath242 . since @xmath451 lies to the left of @xmath242 and the diagram is symmetric , @xmath451 has lower entropy than @xmath241 . since @xmath450 lies between @xmath451 and @xmath241 and entropy is concave , @xmath450 has higher entropy than @xmath451 . hence the @xmath428 minimises the entropy at @xmath452.,title="fig : " ] + 0 @xmath451 @xmath242 @xmath450 @xmath241 1 next , we show that for each @xmath146 and any other policy @xmath10 , @xmath453 let @xmath454 . since @xmath238 , by ( [ eqn : specialcase ] ) , @xmath208 is decreasing , while @xmath455 is increasing . we have already established that @xmath441 and @xmath456 , which implies @xmath457 and @xmath458 respectively . this shows that @xmath459 . for @xmath460 , write @xmath461 since @xmath462 and @xmath463 , decreasing @xmath464 increases @xmath465 , and the same is true for @xmath466 , since in that case we can write the expression in the same way with @xmath467 . this proves ( [ eqn : coeffdominance ] ) . using ( [ eqn : coeffdominance ] ) and @xmath468 , for any @xmath469 , @xmath470 since @xmath471 identically , the second series vanishes as @xmath472 , while the first series is always non - negative by ( [ eqn : coeffdominance ] ) , hence @xmath473 . thus , @xmath474 this proves the required minimisation . note that the proof above relies heavily on the fact that equality is attained in ( [ eqn : symmetrich ] ) . this occurs only in the symmetric case , and thus this approach does not generalise readily . the complexity of the proof in the symmetric case is indicative of the difficulty of the problem in general , and thus highlights the importance of the empirical description provided by figure [ fig : thresholdregions ] . in the course of performing the computations to generate figure [ fig : thresholdregions ] , we noticed that entropy is unimodal with respect to threshold , with threshold considered as a circle as in figure [ fig : thresholddomain ] . while we can not prove this analytically , it is true for each of the 152000 points in the parameter space considered . this allows some simplification in finding the optimal threshold policy , since finding a local minimum is sufficient . thus , we can alter algorithm [ alg : optimalthreshold ] to begin by testing only two policies , then testing policies in the direction of entropy decrease until a local minimum is found . however , the running time improvement is only a constant factor ; if we model entropy as a standard sinusoid with respect to threshold , then the running time decreases by a factor of 3 on average . the problem of determining the optimal general policy is much more difficult , due to the complexity of the space of general policies . since a policy is uniquely determined by the value of the policy function at the orbit points , this space can be viewed as a hypercube of countably infinite dimension , which is much more difficult to study than the space of threshold policies , which is a circle . one strategy is to truncate the orbit and consider a finite dimensional hypercube , justified by the fact that orbit points have masses which decay geometrically , and thus the tail contributes very little . however , a truncation at @xmath313 ( that is , force the policy to be constant on @xmath475 , and similarly for the orbit of 1 ) gives @xmath389 possible policies , which is still far too large to determine optimality by checking the entropy of each policy . the next approximation is to only look for locally optimal policies , in the sense that changing the policy at each of the @xmath476 truncated orbit points increases entropy , and hope that by finding enough such locally optimal policies , the globally optimal policy will be among them . since a hypercube has very high connectivity , regions of attraction tend to be large , which heuristically suggests that this strategy will be effective . [ alg : localoptimum ] finding a locally optimal truncated policy . 1 . pick @xmath313 , and a starting policy , expressed as a pair of sequences of binary digits @xmath477 , with @xmath478 ; 2 . cycle through the digits @xmath479 , flipping the digit if it gives a policy with lower entropy , otherwise leaving it unchanged ; 3 . if the previous step required any changes , repeat it , otherwise a locally optimal truncated policy has been found . we picked @xmath480 since this allows a policy to be easily expressed as two unsigned 64-bit integers , and for each of the 152000 uniformly spaced parameters of figure [ fig : thresholdregions ] , we generated 10 policies uniformly on the hypercube and applied algorithm [ alg : localoptimum ] . none of the locally optimal policies for any of the parameter values had lower entropy than the optimal threshold policy from figure [ fig : thresholdregions ] , and on average 98.3% of them were equivalent to the optimal threshold policy , up to a precision of 0.1% , indicating that the optimal threshold policy is locally optimal with a very large basin of attraction , which strongly suggests that it is also the globally optimal policy . in the special case , the infimum of entropy attainable under threshold policies is the same as that under general policies . the fact that a large proportion of locally optimal policies have globally optimal entropy gives a new method for finding the optimal policy . by picking 10 random truncated policies and running algorithm [ alg : localoptimum ] , at least one of them will yield an optimal policy with very high probability . empirical observations suggest that this method is slower than algorithm [ alg : optimalthreshold ] on average , but since the success rate remains high while algorithm [ alg : optimalthreshold ] becomes significantly slower as @xmath48 approaches 1 , this method is a better alternative for some parameter values . . darkness increases with the proportion of simulated locally optimal policies which have the same entropy as the optimal threshold policy , up to a precision of 0.1% . the average is 9.83 out of 10 , but the distribution is far from uniform local optima are exceedingly likely to be the same as the threshold optimum for some parameter values and exceedingly unlikely for others . the boundaries are approximately those of the threshold regions ( see figure [ fig : thresholdregions ] ) , with some imprecision due to the non - deterministic nature of the simulation data . ] one last policy of interest is the greedy policy . in the previous sections , we considered a long term optimality criterion in the minisation of limiting expected entropy , but in some cases , it may be more appropriate to set a short term goal . in particular , one may instead desire to minimise expected entropy at the next step , in an attempt to realise maximal immediate gain while ignoring future implications . [ def_greedy ] the * greedy * policy is the policy such that the expected entropy after one observation is minimised . up to an exchage of strict and non - strict inequalities , this is given by @xmath481:\alpha_0(z)h(r_0(z))<\alpha_1(z)h(r_1(z))\},\\ & a_1=\{z\in[0,1]:\alpha_0(z)h(r_0(z))\ge\alpha_1(z)h(r_1(z))\}.\end{aligned}\ ] ] the greedy policy has the benefit of being extremely easy to use , as it only requires a comparison of two functions at the current information state . since these functions are smooth , efficient numerical methods such as newton - raphson can be used to determine the intersection points , thus allowing the policy to be described by a small number of thresholds . in fact , only one threshold is required , as computational results show the greedy policy always a threshold policy . using the 152000 uniformly distributed data points from before , in each case the two functions defining the greedy policy crossed over at most once . the greedy policy is always a threshold policy . note that @xmath482 . it may appear at first glance that the factor of @xmath195 violates symmetry , but recall that @xmath18 maps to @xmath483 under relabelling . using this identity , the intersection points that define the greedy policy satisfy @xmath484 , where @xmath485 . it is easy to see that @xmath486 is monotonic decreasing on @xmath199 $ ] with range @xmath487 $ ] , hence @xmath488 is a well - defined one - parameter family of functions mapping @xmath199 $ ] to itself with fixed points at 0 and 1 . on the other hand , @xmath489 is also a one - parameter family of functions mapping @xmath199 $ ] to itself with fixed points at 0 and 1 . since the range of @xmath209 is contained in @xmath196 , we can discount the endpoints @xmath490 and @xmath491 , hence it suffices to show that the equation @xmath492 has at most one solution for @xmath493 . convexity methods may help in this last step but we have not been able to complete the proof . even when the greedy policy is not optimal , it is very close to optimal . of the 152000 uniformly distributed data points in figure [ fig : greedyoptimal ] below , the greedy policy is non - optimal at only 6698 points , or 4.41% , up to an error tolerance of @xmath494 . on average , the greedy policy has entropy 0.0155% higher than the optimal threshold policy , with a maximum error of 5.15% occuring at the sample point @xmath495 , @xmath496 , @xmath497 and @xmath498 . thus the greedy polices provides an alternative suboptimal policy which is very easy to calculate and very close to optimal . . light grey indicates that the greedy policy is the optimal threshold policy ; darker points indicate suboptimality with darkness proportional to error . similarly to figure [ fig : localoptimum ] , the suboptimal points lie on the boundaries of the threshold regions . ] we make a final remark that the likelihood of a locally optimal policy being globally optimal as shown in figure [ fig : localoptimum ] , and the closeness of the greedy policy to the optimal threshold policy as shown in figure [ fig : greedyoptimal ] , both exhibit a change in behaviour at the boundaries of the threshold regions as shown in figure [ fig : thresholdregions ] . this suggests that these regions are indeed qualitatively different , and are likely to be interesting objects for further study . this thesis presents initial groundwork in the theory of hidden markov models with multiple observation processes . we prove a condition under which the information state converges in distribution , and give algorithms for finding the optimal policy in a special case , which provides strong evidence that the optimal policy is a threshold policy . * the information state converges for an anchored hidden markov model with only ergodicity rather than positivity ; * the greedy policy is always a threshold policy ; * among threshold policies , the limiting expected entropy is unimodal with respect to threshold ; and * the optimal threshold policy is also optimal among general policies . possible approaches to these problems are likely to be found in @xcite , @xcite and @xcite . the author was not aware of these papers under after the date of submission , and thus was unable to incorporate their methods into this thesis . better algorithms and error bounds for finding the optimal policy are also a worthwhile goal . although our algorithms are computationally feasible with reasonable prescribed errors , our focus was on finding workable rather than optimal algorithms , and thus there is plenty of room for improvement . baum and t. ptrie . statistical inference for probabilistic functions of finite state markov chains . _ annals of mathematical statistics_. volume 6 , number 37 , pages 15541563 , 1966 . o. capp , e. moulines and t. rydn . _ inference in hidden markov models_. springer , new york , 2005 . g. casella and r.l . _ statistical inference _ , second edition . thomson , pacific grove , 2002 . t.m . cover and j.a . _ elements of information theory_. wiley , new york , 1991 . elliott , l. aggoun and j.b . _ hidden markov models : estimation and control_. springer - verlag , new york , 1995 . evans and v. krishnamurthy . optimal sensor scheduling for hidden markov model state estimation . _ international journal of control_. volume 18 , number 74 , pages 17371742 , 2001 . l.a . johnston and v. krishnamurthy . opportunistic file transfer over a fading channel : a pomdp search theory formulation with optimal threshold policies . _ ieee transactions on wireless communications_. volume 5 , number 2 , pages 394205 , 2006 . v. krishnamurthy . algorithms for optimal scheduling and management of hidden markov model sensors . _ ieee transactions on signals processing_. volume 6 , number 50 , pages 13821397 , 2002 . v. krishnamurthy and d. djonin . structured threshold policies for dynamic sensor scheduling a pomdp approach . _ ieee transactions on signals processing_. volume 5 , number 10 , pages 49384957 , 2007 . f. le gland and l. mevel . exponential forgetting and geometric ergodicity in hidden markov models . _ mathematics of control , signals and systems_. volume 13 , pages 6393 , 2000 . macphee and b.p . optimal search for a moving target . _ probability in the engineering and informational sciences_. volume 9 , number 2 , pages 159182 , 1995 . meyn and r.l . tweedie . _ markov chains and stochastic stability_. springer - verlag , london , 1993 . rabiner . a tutorial on hidden markov models and selected applications in speech recognition . _ proceedings of the ieee_. volume 2 , number 77 , pages 257286 , 1989 . sensor scheduling for optimal observability using estimation entropy . _ proceedings of the fifth annual ieee international conference on pervasive computing and communications workshops_. march 2007 . d. sinno and d. cochran . dynamic estimation with selectable linear measurements . _ proceedings of the 1998 ieee international conference on accoustics , speech and signal processing_. may 1998 . _ supermodularity and complementarity_. princeton university press , 1998 . separation of estimation and control for discrete time systems . _ proceedings of the ieee_. volume 59 , number 11 , 1971 . _ hidden markov models with multiple observation processes_. honours thesis , university of melbourne , 2007 .

we consider a hidden markov model with multiple observation processes , one of which is chosen at each point in time by a policy a deterministic function of the information state and attempt to determine which policy minimises the limiting expected entropy of the information state . focusing on a special case , we prove analytically that the information state always converges in distribution , and derive a formula for the limiting entropy which can be used for calculations with high precision . using this fomula , we find computationally that the optimal policy is always a threshold policy , allowing it to be easily found . we also find that the greedy policy is almost optimal .

let @xmath3 be a smooth bounded domain with boundary @xmath4 and @xmath5 the laplacian acting on functions with dirichlet boundary conditions . we are interested in describing the relationship between certain @xmath6 estimates for the associated spectral projector , obtained recently by h. smith and c. sogge @xcite , and strichartz inequalities for solutions to the wave equation in @xmath7 . this relationship turns out to be very simple , natural and optimal ( at least in some range of indexes ) , and is closely related to an earlier remark of mockenhaupt , seeger and sogge regarding fourier integral operators ( @xcite , corollary 3.3 ) . as an application we consider the critical semi - linear wave equation ( with real initial data ) in @xmath8 , @xmath9 which enjoys the conservation of energy @xmath10 our main result reads : [ th.1 ] for any @xmath11 there exists a unique ( global in time ) solution @xmath12 to in the space @xmath13 to our knowledge , the fact that weakened dispersion estimates can still imply optimal ( and scale invariant ) strichartz estimates for the solution to the wave equation was first noticed by the second author @xcite . observe that the results obtained in @xcite , though restricted only to the interior of strictly convex domains , are far more precise than the results presented here , and apply to the critical non linear wave equation in higher dimension as well . the difficulty in proving theorem [ th.1 ] is that we can not afford any loss in the strichartz estimates we prove : strichartz estimates obtained by tataru @xcite for lipschitz metrics ( see also the results by anton @xcite in the schrdinger context ) would certainly improve the classical result ( well posedness for the cubic non linear wave equation ) but it would not be enough to deal with the quintic nonlinearity . in other dimensions , we can still apply smith and sogge s spectral projectors results , leading to other strichartz type estimates . the strategy also works for obtaining semi - classical strichartz - type estimates for the schrdinger equation ( see burq , grard and tzvetkov @xcite ) . these questions will be addressed elsewhere . using the material in this paper , it is rather standard to prove existence of global smooth solutions , for smooth initial data satisfying compatibility conditions ( see @xcite ) . furthermore , the arguments developed in this paper apply equally well to more general defocusing non linearities @xmath14 satisfying @xmath15 finally , let us remark that our results can be localized ( in space ) and consequently hold also in the exterior of any obstacle , and we extend in this framework previous results obtained by smith and sogge @xcite for convex obstacles . we shall denote in the remaining of this paper , for @xmath16 , by @xmath17 the domain of @xmath18 ( @xmath19 for @xmath20 ) . * acknowledgments : * we thank p. grard for various enlightenments about the critical wave equation . the local ( in time ) existence result for is in fact an easy consequence of some recent work by smith and sogge @xcite on the spectral projector defined by @xmath21 . [ th.smso ] let @xmath3 be a smooth bounded domain , then @xmath22 we now derive from this result some strichartz estimates , which are optimal w.r.t scaling . [ th.2 ] assume that for some @xmath23 , the spectral projector @xmath24 satisfies @xmath25 then the solution to the wave equation @xmath26 satisfies @xmath27 the proof is rather simple . in fact , one can recast the spectral projector estimate as a square function ( in time ) estimate for the wave equation , and use sobolev in time , as was already observed in @xcite in the context of variable coefficients wave equations ( note that such square function estimates are useful in their own right in a nonlinear context : see e.g. @xcite , appendix b , @xcite ) . let @xmath28 be the eigenbasis of @xmath29 consisting in eigenfunctions of @xmath30 associated to the eigenvalues @xmath31 . let us define on this basis an abstract self adjoint operator @xmath32 e_\lambda\ ] ] ( where @xmath33 $ ] is the integer part of @xmath34 ) . now we prove the estimate with @xmath35 replaced by @xmath36 . we decompose @xmath37 with @xmath38 using plancherel formula ( for @xmath39 fixed ) , @xmath40 and consequently , using sobolev injection in the time variable for the first inequality , in the last inequality , with @xmath41 , and the fact that @xmath42 from line 3 to line 4 , @xmath43 consequently , @xmath44 is continuous from @xmath45 to @xmath46 . coming back to @xmath47 , it satisfies @xmath48 therefore , using duhamel formula , minkovski inequality and that both @xmath49 and @xmath50 are bounded on @xmath45 , @xmath51 [ cor.1 ] consider @xmath12 solution to @xmath52 then @xmath53 as a consequence , for any initial data @xmath54 , the critical non linear wave equation is locally well posed in @xmath55 ; h^1_0 ( \omega ) ) \cap l^5((0,t ) ; l^{10 } ( \omega ) ) \times c^0([0,t ] ; l^2 ( \omega))\ ] ] ( globally for small norm initial data ) . this corollary proves the uniqueness part in theorem [ th.1 ] to prove corollary [ cor.1 ] , we observe that according to theorems a and [ th.2 ] , the operator @xmath56satisfies @xmath57 applying the previous inequality to @xmath58 , and using the @xmath6 elliptic regularity result @xmath59 we get @xmath60 and consequently by ( complex ) interpolation between and , @xmath61 finally , by sobolev embedding @xmath62 to conclude , we simply observe that @xmath63 and @xmath64 is an isometry from @xmath29 to @xmath65 , leading to . the remaining of corollary [ cor.1 ] follows by a standard fixed point argument with @xmath66 in the space @xmath67 with a sufficiently small @xmath68 ( depending on the initial data @xmath69 ) . note that this local in time result holds irrespective of the sign of the nonlinearity . finally , to obtain the global well posedness result for small initial data , it is enough to remark that if the norm of the initial data is small enough , then the fixed point can be performed in @xmath70 . then the control of the @xmath71 norm by the energy ( which is conserved along the evolution ) allows to iterate this argument indefinitely leading to global existence . note that this result holds also irrespective of the sign of the nonlinearity because for small @xmath71 norms , the energy always control the @xmath71 norm . it turns out that our strichartz estimates are strong enough to extend local to _ global _ existence for arbitrary ( finite energy ) data , when combined with a trace estimate and non concentration arguments . we have @xmath84 the contributions of @xmath85 are easily dealt with , as previously . let us focus on the contribution of @xmath86 denote by @xmath87 ; interpolating between and , @xmath88 let @xmath89 . then there exist @xmath90 such that @xmath91 as a consequence , from @xmath92 , @xmath93 by duality we deduce that the operator @xmath94 defined by @xmath95 is bounded from @xmath96 to @xmath29 ( observe that @xmath97 ) ; using and boundedness of @xmath98 from @xmath99 to @xmath100 , we obtain @xmath101 and @xmath102 but @xmath103 and an application of christ - kiselev lemma @xcite allows to transfer this property to the operator @xmath104 now we remark that if @xmath105 , we can estimate @xmath106 interpolating between these two inequalities yields @xmath107 following ideas of struwe @xcite , grillakis @xcite and shatah - struwe @xcite , we will localize these estimates on small light cones and use the fact that the @xmath108 norm is small in such small cones . in this section we shall always consider solutions in @xmath110 of having bounded energy and obtained as limits in this space of smooth solutions to the analog of where the non linearity and the initial data have been smoothed out . consequently all the integrations by parts we will perform will be licit by a limiting argument . take @xmath113 a vector field whose restriction to @xmath114 is equal to @xmath115 and compute for @xmath116 @xmath117 u(t , x)\cdot { u } ( t , x ) dx dt\\ = \int_0^{t}\int_{\omega } \left((\partial_t^2- \delta ) z u(t , x)- z(\partial_t^2- \delta ) u(t , x)\right ) { u } ( t , x ) dx dt\,.\end{gathered}\ ] ] integrating by parts , we obtain @xmath118 u(t , x ) \cdot u(t , x ) dx dt= \int_0^{t } \int_{\partial \omega } \bigl|\frac{\partial u}{\partial \nu}\bigr|^2 d\sigma dt\\ { } + \int_0^{t}\int_{\omega } -(z u)u^5(t , x)+ z ( u^5 ) u ( t , x ) dx dt+\left[\int_\omega \partial_t ( zu)\cdot u dx \right]_0^{t}-\left[\int_\omega ( zu)\cdot \partial_t u dx \right]_0^{t } .\end{gathered}\ ] ] remark now that if @xmath119 , then integration by parts yields ( using the dirichlet boundary condition ) @xmath120 while @xmath121_0^{t}-\left[\int_\omega ( zu ) \partial_t u dx \right]_0^{t}\bigr| \leq c e(u),\ ] ] and @xmath122= - [ \delta , z]$ ] as a second order differential operator in the @xmath39 variable is continuous from @xmath65 to @xmath123 and consequently @xmath124 u(t , x ) \overline{u } ( t , x ) dx dt\bigr| \leq c e(u).\ ] ] as the constants are uniform with respect to @xmath116 , collecting , and yields . by time translation , we shall assume later that @xmath125 . let us first define @xmath126 @xmath127 @xmath128\times \partial \omega)\cap k_s^t ) \cup d_t \cup d_s \cup m_s^t\\ e(u ) & = \bigl(\frac { |\partial_t u|^2 + |\nabla_x u|^2 } 2 + \frac { |u|^6 } 6 , - \partial_t u \nabla _ x u\bigr ) \end{aligned}\ ] ] and the flux across @xmath129 @xmath130 where @xmath131 is the outward normal to @xmath132 and @xmath133 the induced measure on @xmath129 . remark that @xmath134 an integration by parts gives ( see rauch @xcite or ( * ? ? ? * ( 3.3 ) ) ) @xmath135 this proves that @xmath136 is bounded in @xmath137 ( uniformly with respect to @xmath138 ) and , since @xmath139 beeing a non - negative non - increasing function has a limit when @xmath140 , @xmath141 exists and satisfies @xmath142 we follow @xcite and simply have to take care of the boundary terms . we can assume that @xmath146 as otherwise these boundary terms disappear in the calculations below ( which in this case are standard ) . contrarily to @xcite we can not use any convexity assumption to obtain that these terms have the right sign , but we shall use proposition [ prop.apriori ] to control them . performing a space - time translation , we can assume @xmath147 . integrating over @xmath148 the identity @xmath149 we get ( see ( * ? ? ? * ( 3.9) ( 3.12 ) ) ) , @xmath150 let @xmath151 . using hlder s inequality and the conservation of energy , we get that the first term in the left tends to @xmath152 , whereas the last term is non negative . this yields @xmath153 on the other hand , by direct calculation ( see ( * ? ? ? * ( 3.11 ) ) ) , @xmath154 and ( see ( * ? ? ? * ( 3.12 ) ) ) @xmath155 as a consequence , we obtain @xmath156 where @xmath157 is the exterior normal to @xmath7 at point @xmath39 and @xmath158 is the surface measure on @xmath4 . taking into account ( and the dirichlet boundary condition ) , we obtain on @xmath4 @xmath159 however , for @xmath160 , given that @xmath161 , we have @xmath162 where @xmath163 is a unit vector tangent to @xmath114 at @xmath164 . consequently , as @xmath165 , @xmath166 and the right hand side in is bounded ( using proposition [ prop.apriori ] ) by @xmath167 therefore , @xmath168 finally , by hlder s inequality and , we obtain @xmath169 hence , @xmath170 for which the right hand side goes to @xmath152 as @xmath140 by . remark that in the calculations above all integrals on @xmath171 and @xmath172 have to be understood as the limits as @xmath151 of the respective integrals on @xmath148 and @xmath129 ( which exist according to , ) . in this section we consider @xmath12 the unique forward maximal solution to the cauchy problem in the space @xmath144 . assume that @xmath173 and consider a point @xmath174 ; our aim is to prove that @xmath12 can be extended in a neighborhood of @xmath175 , which will imply a contradiction . we perform a space time translation and assume that @xmath176 . [ lem.extension ] for any @xmath186 there exists @xmath187 such that for any @xmath188 and any @xmath189 , there exist a function @xmath190 ( independent of the choice of @xmath191 ) , satisfying @xmath192 in other words , we can extend functions in @xmath193 on the ball @xmath194 to functions in @xmath195 with uniform bounds with respect to ( small ) @xmath196 , for the @xmath71 _ and _ the @xmath6 norms respectively . let us first prove the first part of the lemma . let us first assume @xmath200 . we use the usual reflexion extension ( suitably cut off ) . fix a function @xmath201 equal to @xmath79 near @xmath79 . let us define @xmath202 ( in polar coordinates @xmath203 centered at @xmath204 ) by @xmath205 an elementary calculation shows that ( uniformly with respect to @xmath206 ) @xmath207 and ( using hardy inequality to control @xmath208 ) , @xmath209 since @xmath200 , the function @xmath202 is in @xmath65 if @xmath196 is sufficiently small . if @xmath210 , this is no longer true and we have to take care of the boundary condition @xmath211 . to do so the idea is to take a foliation by hypersurfaces tangent to the boundary and to perform a reflexion extension _ tangential _ to this foliation . we consider a change of variables @xmath212 such that ( near @xmath204 ) @xmath213 writing @xmath214 , and working in polar coordinates @xmath215 , the implicit function theorem ( and the assumption @xmath216 ) implies that there exists a smooth function @xmath217 such that for @xmath218 $ ] , @xmath219 we can now define an extension in a polar coordinate for @xmath220 , ( for @xmath221 ) by @xmath222 let @xmath223 equal to @xmath79 near @xmath152 . then the extension we consider is @xmath224 where @xmath225 is the extension we built in the interior case and @xmath226 is the extension we just built ( with the foliation ) . [ fig : foliation ] @xmath227{foliation } } { \aat{45}{31}{$\{|\psi(y)- \psi(0)|=|x - x_0|=r\}$}\aat{45}{26}{$\{\rho = r\zeta(\theta , \frac{y_3 } r , r)\}$}\aat{50}{18}{$\{y_3= cste\}$}\aat{38}{9}{$y'$}\aat{23}{7}{$\partial \omega$}}\ ] ] let us come back to the proof of proposition [ prop.l5w ] . let @xmath229 be the function given by the second part of lemma [ lem.extension ] . then @xmath230 is equal to @xmath231 on @xmath232 and @xmath233 on the other hand , @xmath234 and @xmath235 by ( complex ) interpolation , as in , @xmath236 let @xmath237 be the solution ( which , by finite speed of propagation , coincides with @xmath12 on @xmath232 ) of @xmath238 applying , and the sobolev embedding @xmath239 , we get @xmath240 finally , from proposition [ noconc ] , and the continuity of the mapping @xmath241 ( which takes value @xmath152 for @xmath242 ) , there exists @xmath163 ( close to @xmath152 ) such that @xmath243 and passing to the limit @xmath244 , @xmath245 we are now ready to prove the global existence result . let @xmath247 be close to @xmath152 and let @xmath35 be the solution to the linear equation @xmath248 then the difference @xmath249 satisfies @xmath250 let @xmath229 be the function given by lemma [ lem.extension ] from @xmath251 . we have @xmath252 let @xmath253 be the solution to @xmath254 by finite speed of propagation , @xmath237 and @xmath253 coincide in @xmath232 . on the other hand , using yields @xmath255 finally , for any ball @xmath256 , denote by @xmath257 since @xmath35 is a solution to the linear equation , @xmath258 recalling that @xmath259 inside @xmath260 , we obtain from and ( and the sobolev injection @xmath261 ) that there exists a small @xmath262 such that @xmath263 but , since @xmath264 , we have , by dominated convergence , @xmath265 consequently , there exists @xmath266 such that @xmath267 now , according to , the @xmath109 norm of @xmath12 remains smaller than @xmath268 on @xmath269 . as a consequence , the same proof as for proposition [ prop.l5w ] shows that the @xmath270 norm of the solution on the truncated cone @xmath271 is bounded . since this is true for all @xmath186 , a compactness argument shows that @xmath272 which , by duhamel formula shows that @xmath273 exists in @xmath274 and consequently @xmath12 can be extended for @xmath275 small enough , using corollary [ cor.1 ] . jeffrey rauch . i. the @xmath276 klein - gordon equation . anomalous singularities for semilinear wave equations . in _ nonlinear partial differential equations and their applications . collge de france seminar , vol . i ( paris , 1978/1979 ) _ , volume 53 of _ res . notes in math . _ , pages 335364 . pitman , boston , mass .

we prove that the defocusing quintic wave equation , with dirichlet boundary conditions , is globally well posed on @xmath0 for any smooth ( compact ) domain @xmath1 . the main ingredient in the proof is an @xmath2 spectral projector estimate , obtained recently by smith and sogge @xcite , combined with a precise study of the boundary value problem .

[ a ] self - similarity is a possible feature of multi - partons inside a proton at small _ x _ suggested by lastovicka @xcite and pursued by us in recent years @xcite . more recently , we have examined its consequences in @xcite double parton distribution function(dpdf ) to be tested at lhc , in froissart saturation @xcite and longitudinal structure function @xcite . one of the theoretical limitations of the model of ref @xcite is that it has a singularity at @xmath2 which is within the physically allowed range @xmath3 . even though outside its phenomenological range of validity @xmath4 , such singularity is beyond common expectations from any physically viable model of proton structure function @xmath5 . in the present paper , we therefore make a re - analysis of the model of ref@xcite , demanding it to be singularity free in the entire @xmath6-range of @xmath3 . present work will also study the consequences of the suggestion of ref@xcite in the context of the reported reanalysis . in section[b ] , we outline the formalism and in section[c ] , we discuss the results . section[d ] contains the conclusions . the self - similarity based model of the proton structure function of ref@xcite is based on transverse momentum dependent parton distribution function(tmd ) @xmath7 . here @xmath8 is the parton transverse momentum squared . choosing the magnification factors @xmath9 and @xmath10 , it is written as @xcite @xmath11= d_1.log\frac{1}{x}.log\left(1+\frac{k_t^2}{q_0 ^ 2}\right)+d_2.log\frac{1}{x } + d_3.log\left(1+\frac{k_t^2}{q_0 ^ 2}\right)+d_0^i\ ] ] where i denotes a quark flavor . here @xmath12 are the three flavor independent model parameters while @xmath13 is the only flavor dependent normalization constant . m@xmath14(=1 gev@xmath14 ) is introduced to make ( pdf ) @xmath15 as defined below ( in eqn([e2 ] ) ) dimensionless . the integral quark densities then defined as : @xmath16 as a result , the following analytical parametrization of a quark density is obtained by using eqn([e2 ] ) @xcite @xmath17 where , @xmath18 is flavor independent . using eqn([e3 ] ) in the usual definition of the structure function @xmath5 , one can get @xmath19 or it can be written as @xmath20 where @xmath21 . from hera data @xcite , eqn([e6 ] ) was fitted with @xmath22 in the kinematical region , @xmath23 the model of ref@xcite has two inherent limitations . first , the parameter @xmath24 is negative contrary to the expectations of positivity of the fractal dimensions @xcite . secondly , due to its negative value , eqn([e6 ] ) develops a singularity at @xmath25 @xcite as it satisfies the condition @xmath26 , contrary to the expectation of a physically viable form of structure function . the model can be made singularity free under following specific conditions of the model parameters . + + * case 1 :* if @xmath27 in eqn([e1 ] ) then the pdf eqn([e3 ] ) and the structure function eqn([e6 ] ) will be of the form : @xmath28 @xmath29 + + * case 2 :* in this case @xmath30 in eqn([e1 ] ) then the corresponding expressions for the pdf and structure function in this limit are respectively : @xmath31 @xmath32 + + * case 3 :* in this case , @xmath33 in eqn([e1 ] ) then the corresponding pdf and the structure function are set in the form : @xmath34 @xmath35 respectively . + + * case 4 :* this is the most general case for the singularity free model , eqn([e6 ] ) under the condition that @xmath36 are positive . it is to be noted that the variable in which the supposed fractal scaling in the quark distribution occur is not known for the underlying theory . in ref@xcite , the choice of @xmath37 as one of the magnification factors was taken presumably because of the power law form of the quark distribution at small @xmath6 found in glck - reya - vogt ( grv ) distribution @xcite . however , this form is not derived for the theory but rather inspired by the power law distribution in @xmath38 assumed for grv distribution for qcd evaluation . the choice of @xmath37 as the proper scaling variable is not an established result of the underlying theory . a more plausible variable appears instead to be @xmath39 @xcite which confirms to the froissart saturation @xcite of high energy interaction . with this choice , transverse momentum dependent structure function ( tmd ) and parton distribution function ( pdf ) now take the following forms @xcite @xmath40 and @xmath41 where , @xmath42 the structure function as defined in eqn([e6 ] ) becomes @xmath43 where @xmath44 for very small @xmath6 and large @xmath45 , the second term of eqn([e18 ] ) can be neglected leading to @xmath46 which satisfies the froissart saturation condition , if @xmath47 within @xmath48 corrections @xcite . to determine the parameters of the model @xmath49 we use recently compiled hera data @xcite instead of earlier data @xcite used in ref@xcite . following this procedure of ref@xcite we make @xmath50-analysis of the data and find the following results . + + * case 1 :* we note that @xmath51=1 is ruled out since it will make the structure function eqn([e10 ] ) @xmath6-independent . in table [ table1 ] we show the results . from the @xmath50-analysis , the model in case 1 is confined well with data for 0.35@xmath5270gev@xmath14 and 6.62@xmath53 10@xmath54 x @xmath55 0.08 . note that @xmath24 and @xmath56 are taken to be zero in this limit . here the number of @xmath57 data points is 222 . + 12cm|x|x|x|x|x|x|x| @xmath58 & @xmath56 & @xmath51 & @xmath24 & @xmath59 ( gev@xmath14 ) & @xmath50 & @xmath60 + -4.129 @xmath61 & 0 & 1.226 @xmath62 & 0 & - & 180.11 & 0.81 + + + + * case 2 :* the parameters @xmath63 and @xmath59 are determined ( given in table [ table2 ] ) in the similar manner as in case 1 but the validity range is new 0.35 @xmath64 gev@xmath14 and 6.62@xmath53 10@xmath54 x @xmath55 0.032 . as in case 1 , here too @xmath65 is ruled out since it will make eqn([e12 ] ) @xmath6-independent . the number of @xmath57 data points is 174 . + 12cm|x|x|x|x|x|x|x| @xmath58 & @xmath56 & @xmath51 & @xmath24 & @xmath59 ( gev@xmath14 ) & @xmath50 & @xmath60 + -6.125 @xmath66 & 0 & 1.214 @xmath62 & 0.531 @xmath62 & 0.053 @xmath67 & 138.67 & 0.80 + + + * case 3 :* here parameters are best fitted in the range 0.35@xmath5215 gev@xmath14 and 6.62@xmath53 10@xmath54 x @xmath55 0.02 and given in table [ table3 ] . the number of @xmath57 data points is 146 . + 12cm|x|x|x|x|x|x|x| @xmath58 & @xmath56 & @xmath51 & @xmath24 & @xmath59 ( gev@xmath68 & @xmath50 & @xmath60 + -3.533 @xmath69 & 0.411 @xmath70 & 0.582 @xmath71 & 0 & 0.035 @xmath72 & 117.31 & 0.80 + + + * case 4 :* parameters @xmath73 and @xmath59 are determined and given in table [ table4 ] and obtained in a more restrictive range 0.85@xmath5210 gev@xmath14 and 2@xmath53 10@xmath74 x @xmath550.02 . the number of @xmath57 data points is 95 . + 12cm|x|x|x|x|x|x|x| @xmath58 & @xmath56 & @xmath51 & @xmath24 & @xmath59 ( gev@xmath68 & @xmath50 & @xmath60 + -2.971 @xmath75 & 0.065 @xmath76 & 1.021 @xmath77 & 0.0003 & 0.20 @xmath78 & 18.829 & 0.20 + + in figure [ fig1 ] , we plot @xmath57 as a function of @xmath6 for eight representative values of @xmath45 ( @xmath45= 1.5 , 2 , 2.7 , 3.5 , 4.5 , 6.5 , 8.5 , 10 gev@xmath14 ) in the phenomenological allowed range 0.85@xmath5210 gev@xmath14 . we also show the corresponding available data from ref @xcite . our analysis indicates that the phenomenological range of validity of the present version of the model is more restrictive @xmath79 gev@xmath14 and @xmath80 to be compared with eqn([e8 ] ) of the previous version of ref@xcite . we also observe the following features of the model compared to data : at @xmath81gev@xmath14 data overshoots the theory . but as @xmath45 increases , the theoretical curve comes closer to data . at @xmath45=10 gev@xmath14 , on the other hand , the theory exceeds data . main reason of this feature is that the @xmath6-slope of the model is less than that of the data . this limitation can in principle be overcome by modification of the magnification factor @xmath37 to @xmath82 in eqn([e1 ] ) to accommodate proper large @xmath6 behavior of structure function as noted in ref@xcite . as an illustration we show results of two representative values of @xmath45= 4.5 and 6.5 gev@xmath14 in figure [ fig2 ] with such magnification using the parameters in table [ table4 ] . as a function of @xmath6 in bins of @xmath45 by recently compiled hera data.,width=234 ] it is interesting to predict the @xmath83-dependance of unintegrated parton distribution ( tmd ) from the @xmath6 and @xmath45 dependence of the integrated parton distribution function ( pdf ) . clearly this can be done within a model framework , as has been noted in ref@xcite as well as in ref@xcite . though it should be of interest to explore this approach to study @xmath83-dependence of @xmath7 , such a study makes sense only in the @xmath6-@xmath45 range where the approach works and where the parameters had been fitted namely @xmath80 in the present model . using eqn([e1 ] ) , @xmath7 has the form @xmath84 + which has dimension of mass@xmath85 consistence with eqn([e2 ] ) where the pdf @xmath15 is dimensionless . to evaluate eqn([e22 ] ) we take the mean value @xmath58 , @xmath56 , @xmath51 , @xmath24 and @xmath59 from the table[table4 ] . assuming @xmath86 and setting @xmath87 , we obtain @xmath88 . in figure [ fig3 ] , @xmath89 vs @xmath83 is shown for representative values of + ( i ) @xmath6 = 10@xmath90 ( ii ) @xmath6 = 10@xmath91 ( iii ) @xmath6 = 10@xmath85 ( iv ) @xmath6=0.02 setting @xmath92 gev@xmath14 . the allowed limit of @xmath83 is considered to be less than the average value @xmath83=0.25 gev@xmath14 @xcite as determined from data . similarly @xmath89 vs @xmath6 as shown in figure [ fig4 ] for representative values of + ( i ) @xmath83 = 0.01 gev@xmath14 ( ii ) @xmath83 = 0.1 gev@xmath14 ( iii ) @xmath83= @xmath59 = 0.20 gev@xmath14 ( iv ) @xmath83=0.25 gev@xmath14 . the present graphical analysis of tmd [ fig [ fig3],[fig4 ] ] is an improvement over the earlier one @xcite , in the sense that the experimental limit on @xmath83 as determined from data @xcite was not taken into consideration there . further it was also beyond the range of validity of the original model . proper dimension of the tmd has also been taken into account in the present case . + let us now compare the structure of the model tmd ( eqn[e22 ] ) with the suggested forms available in current literature . the standard way to study tmds is through the factorization approach @xcite where @xmath6 and @xmath45-dependence are factorized into a pdf @xmath15 and a gaussian transverse momentum dependent function @xmath93 @xmath94 where , @xmath95 with normalization condition , @xmath96 + such factorization property of tmd is not present in the model , ( eqn[e22 ] ) nor the gaussian form eqn[e24 ] . only in the absence of correlation term @xmath56 ( eqn[e1 ] ) such factorization property emerges . in this limit the @xmath83 dependent functional form of the tmd is given by @xmath97 in contrast to a gaussian function ( eqn[e24 ] ) . introducing a @xmath83 cut off @xmath98 eqn([e26 ] ) will satisfy the normalization condition ( eqn[e25 ] ) with a normalization constant @xmath99}\ ] ] + in figure [ fig5 ] we compare eqn([e24 ] ) and eqn([e26 ] ) with @xmath100= 0.25 gev@xmath14 in @xmath101 @xmath102 vs @xmath83 plot . @xmath102 vs @xmath83,width=264 ] the condition of froissart saturation in the parton distribution function in the model @xcite leads to modification of the magnification factor @xmath37 to @xmath39 with the additional condition eqn([e21 ] ) . for fixed @xmath103 and @xmath104 it can be satisfied only for a specific value of @xmath45 . as an illustration if one takes the mean values of the set of table [ table4 ] , i.e. @xmath105 -2.971 gev@xmath14 , @xmath106gev@xmath14 , @xmath107 1.021 gev@xmath14 , @xmath108 0.0003 gev@xmath14 , @xmath109 0.20 gev@xmath14 it is obtained at @xmath110 = 6.95@xmath111 gev@xmath14 , far beyond the phenomenological range of validity 0.85@xmath112 10 gev@xmath14 . however its leading @xmath113 behavior is to be manifested to any @xmath45 , @xmath114 or in a specific regime of @xmath45 , 0.11@xmath1151200 gev@xmath14 as in ref@xcite @xmath56 and @xmath51 should have proper @xmath45-dependence , a feature beyond the scope of the present method of parametrization . a leading @xmath113 behavior of the structure function @xmath5 with @xmath45-independent exponent is possible in the present model only if the correlation term proportional to @xmath56 in the definition of tmd ( eqn[e1 ] ) is negligible compared to @xmath51 and @xmath24 and it is redefined as @xmath116 instead of eqn([e22 ] ) , in such a case the corresponding pdf and structure function have form as @xmath117 and @xmath118 note that eqn([e29 ] ) and eqn([e30 ] ) corresponds to case 2 ( eqn[e11 ] and eqn[e12 ] ) of the original model . setting @xmath51= 2 eqn([e28]-[e30 ] ) will then yield the desired froissart saturation saturation in the model . it is to be noted that condition ( [ e21 ] ) [ eqn(28 ) of ref@xcite ] corresponds to the froissart compatibility of the pdf and not the structure finction @xmath5 as defined in eqn([e5 ] ) due to the multiplicative @xmath6-factor relative to the pdf . redefining of tmd in eqn([e28 ] ) resolves this anomaly . in the present paper , we have made a reanalysis of a structure function @xmath119 based on self - similarity @xcite . the present study is based on the notion that a physically viable model proton should be finite in the @xmath6-range @xmath120 hence singularity free . it also conforms to the expectation that @xmath121fractal dimension " associated with self - similarity are invariably positive definite . we have then studied the possibility of incorporating froissart saturation bound . our analysis indicates that the range of validity of the present version of the model approaches non - perturbative regime of lower @xmath45 , @xmath122gev@xmath14 instead of @xmath123gev@xmath14 of the ref@xcite . it is also suggested that the @xmath6-slope of the predicted structure function can be increased by proper redefinition of the defining magnification factors @xmath37 as noted in ref@xcite . interestingly , the range of validity of the present version of the model is close to that of ref@xcite based on holographic qcd . pattern of momentum fractions carried by the quarks and gluons in the present model [ case 1 - 4 ] is currently under study . we thank dr . kushal kalita for helpful discussions , dr . rupjyoti gogoi of tezpur university for collaboration at the initial stage of the work and dr . akbari jahan for useful comments on self - similarity . one of the authors ( b.s . ) acknowledges the ugc - rfsms for financial support . t. lastovicka , euro . j. c * 24 * , 529 ( 2002 ) , hep - ph/0203260 d.k . choudhury and rupjyoti gogoi , hep - ph/0310260 ; hep - ph/0503047 d.k . choudhury and rupjyoti gogoi , indian . * 80 * , 823 ( 2006 ) d.k . choudhury and rupjyoti gogoi , indian . j. phys.*81 * , 607 ( 2007 ) a. jahan and d.k . choudhury , mod . a * 27 * , 1250193 ( 2012 ) , hep - ph/1304.6882 a. jahan and d.k . choudhury , mod . a * 28 * , 1350056 ( 2013 ) , hep - ph/1306.1891 d.k . choudhury and a. jahan , int . j. mod.phys . a * 28 * , 1350079 ( 2013 ) , hep - ph/1305.6180 a. jahan and d.k . choudhury , phys . d * 89 * , 014014 ( 2014 ) , hep - ph/1401.4327 a. jahan and d.k . choudhury , hep - ph/1404.0808 h1:c . j. c * 21 * , 33 - 61 ( 2001 ) , hep - ex/0012053 zeus : j. breitweg _ et al . _ , b * 487 * , 53 ( 2000 ) , hep - ex/0005018 m. glck , e reya and a. vogt , euro . j. c * 5 * , 461 ( 1998 ) m.m . block , l. durand , p. ha and d.w . mckay , phys . d * 84 * , 094010 ( 2011 ) f. froissart phys . rev . * 123 * , 1053 ( 1961 ) a. martin and s.m . roy , hep - ph/1306.5210 h1 and zeus collaborations , jhep * 01 * , 109 ( 2010 ) p. zavada , phys . d * 83 * , 014022 ( 2011 ) , hep - ph/0908.2316 a. jahan and d.k . choudhury , hep - ph/1106.1145 m. anselmino _ et al . _ , phys . rev . d * 71 * , 074006 ( 2005 ) , hep - ph/0501196 m. anselmino _ et al . _ , d * 75 * , 054032 ( 2007 ) , hep - ph/0707.1197 m. anselmino _ suppl . * 191 * , 98 - 107 ( 2009 ) , hep - ph/0812.4366 akira watanabe and katsuhiko suzuki , hep - ph/1312.7114

in this paper we make re - analysis of a self - similarity based model of the proton structure function at small _ x _ pursued in recent years . the additional assumption is that it should be singularity free in the entire kinematic range @xmath0 . our analysis indicates that the model is valid in a more restrictive range of @xmath1 . we also discuss the possibility of incorporation of froissart saturation condition in the model . + + keywords : self - similarity , quark , gluon . + pacs nos . : 05.45.df , 24.85.+p

in this paper we will study the spectral asymptotics of schdinger operators of the form @xmath1 , where @xmath2 is the ( positive ) laplacian on @xmath0 , and @xmath3 is a compactly supported , possibly complex - valued potential . the essential spectrum of @xmath4 is @xmath5 and is absolutely continuous . the eigenvalue spectrum is finite and contained in @xmath6 , and hence it is the resonance set that plays the role of discrete spectral data in this setting . to define resonances , consider first the resolvent of @xmath2 , written in the form @xmath7 for @xmath8 . the well - known formula expressing the kernel of @xmath9 in terms of hypergeometric functions ( see @xcite ) shows immediately that the cutoff resolvent @xmath10 admits a meromorphic extension to @xmath11 , with poles of finite rank , for any @xmath12 . it is easy to extend this meromorphic continuation result to @xmath13 ; see [ potsc.sec ] for the details . we define the resonance set @xmath14 as the set of poles of @xmath15 , counted according to multiplicity given by the rank of the residue . rpotres3.pdf the resonance counting function is @xmath16 figure [ rpotres3.fig ] shows a sample of the resonance set for @xmath17 in @xmath18 , the characteristic function of the unit ball . the corresponding counting function shown in figure [ nv3.fig ] . these plots are based on explicit calculation of the resonance set in terms of legendre functions ; see [ scmatrix.sec ] for the formulas . it is essentially already known that @xmath19 that is , for real @xmath20 this is a special case of ( * ? ? ? 2.2 ) , and the extension to compactly supported complex potentials is straightforward . nv3.pdf ( 72,40)@xmath21 in the case @xmath22 , the resonance set is of course well - known : for @xmath23 odd we have @xmath24 while for @xmath23 even there are no resonances , @xmath25 . in the case of @xmath23 odd we thus have a simple asymptotic , @xmath26 for later reference , we label the asymptotic constant for the model case as @xmath27 for more general families of compactly supported perturbations of the laplacian on @xmath0 , including metric and topological perturbations as well as smooth real potentials , sharp estimates of @xmath28 were obtained in borthwick @xcite . our first result is an extension of this bound to non - smooth , complex - valued potentials . these estimates involve the integrated version of the counting function , @xmath29 a common usage in the theory of entire functions . the constant in the upper bound is expressed in terms of a _ indicator function _ , defined for @xmath30 by @xmath31_+}{x^{n+2}}\>dx,\ ] ] where @xmath32_+$ ] denotes the positive part and @xmath33 \\ & \qquad + \log \left| \frac{\cosh r - \sqrt{1 + \alpha^2 \sinh^2 r}}{\cosh r + \sqrt{1 + \alpha^2 \sinh^2 r } } \right| . \end{split}\ ] ] the corresponding asymptotic constant is the integral , @xmath34 ( the angular dependence of the indicator function will play a role later when we consider the distribution of resonances in sectors . ) [ upper.thm ] suppose that @xmath35 has support contained in a closed ball of radius @xmath36 . then @xmath37 the error estimate in theorem [ upper.thm ] improves on the @xmath38 result of @xcite , but this improvement is specific to the potential case . it is based an sharper scattering phase estimate that we will give in proposition [ sigma.bound ] . our main goal in this paper is to demonstrate the sharpness of theorem [ upper.thm ] in the case of radial potentials . this is analogous to the euclidean result for radial potentials in odd dimensions due to zworski @xcite , with the exact constant later computed by stefanov @xcite . [ asymp.thm ] suppose that @xmath39 is a radial potential with support in a ball of radius @xmath36 . if we assume that @xmath20 is continuous near @xmath40 and has finite order of vanishing in the sense that @xmath41 for some @xmath42 and @xmath43 . then @xmath44 in the final section of the paper we include some applications of this result . in the euclidean case , christiansen @xcite has established generic properties of resonance distributions for potential scattering , and the exact asymptotic for the radial case plays a key role in this work . using theorem [ asymp.thm ] we can prove some analogous theorems for @xmath0 . these results ( see [ dist.sec ] for the precise statements ) include : 1 . for generic ( real or complex ) potentials @xmath20 with support in a compact set @xmath45 with non - empty interior , the order of growth of the resonance counting function is optimal in the sense that @xmath46 2 . for generic ( real or complex ) potentials supported in the closed ball @xmath47 , @xmath48 i.e. the estimate in theorem [ upper.thm ] is generically optimal for potentials with support equal to @xmath47 . 3 . for generic ( real or complex ) potentials with support in the closed ball @xmath47 , there is a generic lower bound on the number of resonances contained in a sector near the critical line , with the optimal order of growth @xmath49 and a constant independent of the size of the sector . 4 . for potentials with support in @xmath47 for which @xmath50 , the asymptotic distribution of resonances in sectors is governed by the indicator function @xmath51 defined in . the same distribution holds , in a weighted average sense , for families of perturbations of of such potentials . the paper is organized as follows . in [ potsc.sec ] we introduce the basic spectral operators associated to @xmath52 , the scattering matrix in particular . in [ rescount.sec ] we establish the formula for counting resonances in terms of the relative scattering determinant , which is the basis for the rest of the analysis . we also prove some general estimates on the scattering determinant that will be needed later , and which in particular give the proof of theorem [ upper.thm ] . explicit formulas for scattering matrix elements in the radial case are worked out in [ scmatrix.sec ] . in [ asym.sec ] we develop precise recursive estimates for these matrix elements . from these estimates we derive asymptotics of the scattering determinant in [ rscdet.sec ] , yielding the proof of theorem [ asymp.thm ] . finally [ dist.sec ] contains the resonance distribution results for generic potentials as outlined above . * acknowledgment . * we would like to thank tanya christiansen for advice on the proof of lemma [ sigma.bound ] as well as for helpful discussions related to the material in [ dist.sec ] . consider a schrdinger operator @xmath52 in @xmath0 , with potential @xmath35 . the resolvent @xmath13 is defined by the spectral theorem for @xmath53 sufficiently large , and is related to the model resolvent @xmath9 by the identity @xmath54 as mentioned in the introduction , the cutoff resolvent @xmath10 admits a meromorphic continuation to @xmath11 . we can describe this more precisely in terms of weighted @xmath55 spaces . in terms of geodesic polar coordinates @xmath56 for @xmath57 , define @xmath58 . in terms of the poincar ball compactification of @xmath0 , @xmath59 is a boundary defining function . the model resolvent @xmath9 extends meromorphically to @xmath60 , as an operator @xmath61 . since @xmath62 is compact as an operator on @xmath63 , for @xmath60 , with arbitrarily small norm for @xmath53 sufficiently large , the analytic fredholm theorem yields a meromorphic inverse @xmath64 . in conjunction with , this establishes the following : the resolvent @xmath15 extends meromorphically to @xmath65 as @xmath66 with poles of finite rank . for any @xmath67 , @xmath15 is bounded as an operator @xmath61 for @xmath60 . with meromorphic continuation of @xmath15 established , we define the resonance set @xmath14 as the set of poles of @xmath15 , counted with multiplicities @xmath68 for later use we need an estimate on the cutoff resolvent in the physical plane . [ res.est ] suppose @xmath35 and @xmath69 , with @xmath70 on @xmath71 . there exist @xmath72 , @xmath73 such that for @xmath74 $ ] and @xmath75 , @xmath76 here @xmath77 depends only on the support of @xmath78 , while @xmath79 depends on @xmath80 and @xmath81 . set @xmath82 , and let @xmath83 be the projector given by multiplication by the characteristic function of @xmath84 . since @xmath85 , we can write @xmath86 and then by inverting this expression we have the identity @xmath87 this allows us to write the cutoff resolvent as @xmath88 ( this trick works just as in the euclidean case ; see , e.g. , @xcite . ) to estimate the terms involving @xmath9 , we can cite guillarmou ( * ? ? ? 3.2 ) , which gives the estimate @xmath89 for @xmath90 , @xmath91 . applying this to the the cutoff resolvent gives @xmath92 we can also apply to obtain , for @xmath93 sufficiently large , the estimates @xmath94 the claim then follows from . the scattering matrix @xmath95 associated to the potential @xmath20 can be defined in the same way as for any asymptotically hyperbolic manifold . we will recall the details rather briefly ; see @xcite for details and references . given @xmath96 and @xmath97 , there is a unique solution of @xmath98u=0 $ ] with the asymptotic @xmath99 for some @xmath100 . the scattering matrix , defined as the map @xmath101 , is a meromorphic family of pseudodifferential operators on @xmath102 . note that by construction , the scattering matrix satisfies @xmath103 . the relative scattering matrix @xmath104 is of determinant class , and the relative scattering determinant is defined as @xmath105 the reflection formula for the scattering matrix implies @xmath106 . this meromorphic function admits a hadamard factorization over the resonance sets @xmath14 and @xmath107 : @xmath108 where @xmath109 is a polynomial of degree at most @xmath49 , and @xmath110 for @xmath20 real - valued , the factorization is a special case of ( * ? ? ? 3.1 ) , and the extension to complex @xmath20 is straightforward . one consequence of is that we can count resonances with a contour integral over the scattering determinant . integrating @xmath111 around a half - circle contour centered at @xmath112 yields @xmath113 where @xmath114 counts the number of resonances @xmath115 with @xmath116 and @xmath117 , which occur only when @xmath118 is a discrete eigenvalue . there are only finitely many discrete eigenvalues , so @xmath119 is bounded . if we divide this contour integral by @xmath120 and integrate , we obtain the relative counting formula , @xmath121 for a general self - adjoint perturbation of @xmath2 , of the type considered in @xcite , the first integral could be expressed in terms of the _ scattering phase _ @xmath122 , which would be real - valued in that case . in the case of a metric perturbation there is a weyl law ( * ? ? ? 3.6 ) giving the asymptotic @xmath123 , with @xmath93 proportional to the volume of the perturbation . for potential scattering we would expect the scattering phase term to be of lower order , and that is indeed the case . [ sigma.bound ] for @xmath124 , we have @xmath125 for @xmath126 sufficiently large , where @xmath127 depends only on @xmath81 and @xmath71 . for real @xmath20 this gives the scattering phase estimate @xmath128 . we will defer the proof for a moment to observe the consequences for the resonance counting formula . applying proposition [ sigma.bound ] to , and using the asymptotics for @xmath129 given in , yields the following : [ count.prop ] for @xmath35 , @xmath130 in this subsection we will develop the proof of proposition [ sigma.bound ] . if we let @xmath131 , the resolvent identity implies the relation @xmath132 proposition [ res.est ] implies that @xmath133 for @xmath134 sufficiently large , in which case we can write @xmath135 [ rels.lemma ] the scattering matrices satisfy a relative scattering formula @xmath136 valid for @xmath137 with @xmath138 sufficiently large . using equation and its transpose we have @xmath139 the formulas for the scattering matrix can then be derived by multiplying the kernels by @xmath140 and taking the limit as @xmath141 . this gives @xmath142 the result follows after applying @xmath143 on the right and using . in order to apply lemma [ r0rv ] we need some estimates on hilbert - schmidt norms of the poisson operator . for this estimate it is easiest to write the poisson kernel in the @xmath144 model . recall that we use the boundary defining function @xmath145 , where @xmath146 is hyperbolic distance from the origin . the normalizing factor is included so that the induced metric on @xmath147 is the standard sphere metric . for this boundary defining function , the poisson kernel is given by @xmath148 where @xmath149 , @xmath150 . [ e0.hsnorm ] let @xmath151 . for @xmath126 , the poisson operator @xmath152 satisfies @xmath153 and @xmath154 the hilbert - schmidt norm is calculated directly : @xmath155^{\frac12}\ ] ] because @xmath78 is compactly supported , there is no convergence issue and the term in brackets is just a constant . the result follows from @xmath156 which is easily deduced from stirling s formula . the derivative estimate is similar . by virtue of lemma [ rels.lemma ] we can write this as @xmath157 where @xmath158 following the argument from froese ( * ? ? ? * lemma 3.3 ) , we will estimate the derivative @xmath159.\ ] ] for @xmath20 real , @xmath160 is unitary for @xmath161 , so that in this case @xmath162 . for a complex potential we need a separate estimate . note that for @xmath161 , @xmath163^ * = e_0(n - s)^t { \mathbbm{1}}_\omega,\ ] ] and we have the general relation ( for any @xmath164 ) , @xmath165 thus we can estimate , for @xmath161 with @xmath166 sufficiently large , @xmath167 the model resolvent estimate from guillarmou ( * ? ? ? 3.2 ) thus implies that @xmath168 for @xmath169 large , and also that @xmath170 for @xmath169 sufficiently large . we conclude that @xmath171 by using with , we can bound the derivative of the scattering phase by a trace norm , @xmath172 to control the trace norm , we have the hilbert - schmidt estimates on @xmath173 and derivatives from lemma [ e0.hsnorm ] . since guillarmou ( * ? ? ? 3.2 ) proves that the estimate also holds with @xmath9 replaced by the derivative @xmath174 , we can estimate @xmath175 for @xmath169 sufficiently large . putting these together ( and noting the extra factor of @xmath176 ) we obtain @xmath177 and the result follows . sharp upper bounds for @xmath178 were provided in borthwick ( * ? ? ? 5.4 ) , for a more general class of compactly supported `` black box '' perturbations of @xmath2 . however , if we restrict to potential scattering we can improve the error estimate . ( in the euclidean case this improvement was established by dinh - vu @xcite . ) theorem [ upper.thm ] follows immediately from the counting formula of proposition [ count.prop ] and the following : [ tau.upper ] assume that the support of @xmath20 is contained within a ball of radius @xmath36 . for @xmath179 for some @xmath180 and @xmath181 we have @xmath182 uniformly for @xmath181 , with @xmath51 the indicator function defined in . set @xmath183 for @xmath184 . let @xmath185 be a cutoff function with @xmath186 for @xmath187 and @xmath188 for @xmath189 . then set @xmath190 , so that @xmath191 for @xmath192 and @xmath193 for @xmath194 . then from the proof of ( * ? ? ? * lemma 4.1 ) we have @xmath195 r_v(s ) [ \delta,\chi_1 ] e_0(n - s),\ ] ] where @xmath196 is the unperturbed poisson operator on @xmath0 . as in the proof of ( * lemma 5.2 ) , this formula leads to an estimate @xmath197 here @xmath198 is the multiplicity of spherical harmonics of weight @xmath199 in dimension @xmath23 , @xmath200 the factor @xmath201 is a cutoff resolvent norm , @xmath202 r_v(s ) [ \delta_0,\chi_1 ] \bigr\vert.\ ] ] finally , the @xmath203 s , calculated from explicit formulas for the fourier decomposition @xmath196 , @xmath204^{\frac12 } \\ & \qquad \times\left [ \int_{r_2}^{r_3 } \bigl|p_{\nu}^{-k}(\cosh r)\bigr|^2\>\sinh r\>dr\right]^{\frac12 } , \end{split}\ ] ] where @xmath205 and @xmath206 . the key to the improved error in the potential scattering case is our ability to estimate @xmath9 for @xmath207 , as in proposition [ res.est ] . using a rather general commutator argument ( see , e.g. , ( * ? ? * lemma 9.8 ) ) , we can derive from proposition [ res.est ] the following bound : if @xmath208 are cutoffs with disjoint supports , then for @xmath93 sufficiently large , @xmath209 uniformly for @xmath30 . then , noting that @xmath210 , we have @xmath211 from ( * ? ? ? * lemma 5.3 ) we also quote the estimate @xmath212 valid for @xmath213 , @xmath30 , and @xmath214 . the statement in @xcite assumed that @xmath215 , but that condition was just a simple way to avoid the poles of @xmath216 . we can easily extend the estimate by noting that @xmath217 . note that the case @xmath218 occurs only if @xmath219 . in this case , the complicated legendre function bounds used in @xcite can be replaced by a simpler estimate : @xmath220 valid for fixed @xmath221 and @xmath222 $ ] . the remainder of the proof is similar to that of ( * ? ? ? * lemma 5.3 ) , with some modifications to improve the error term . by conjugation , it suffices to consider @xmath223 $ ] . let @xmath224 be the implicit solution of @xmath225 , as illustrated in figure [ hplot.fig ] . hplot.pdf ( 25,17)@xmath226 ( 55,55)@xmath227 ( 10,40)@xmath228 we will use @xmath226 to subdivide the sum on the right side of into several pieces , the first of which is @xmath229 using the estimate , @xmath230 with and , we have @xmath231 estimating the sum by an integral , and then substituting @xmath232 , yields @xmath233 by observing that @xmath234 and recalling that @xmath235 , we see that @xmath236 for @xmath237 , uniformly in @xmath238 . hence @xmath239 the second piece of , containing terms where @xmath240 , is defined as @xmath241 note that for @xmath242 , @xmath243 thus , in the range of @xmath244 , gives @xmath245 since @xmath244 contains @xmath246 terms , and @xmath247 , we conclude that @xmath248 finally , the third part of is @xmath249 in this range we have @xmath250 for some @xmath251 , when @xmath93 is sufficiently large . it follows easily that @xmath252 [ lind.lemma ] let @xmath253 denote the joint set of zeros and poles of @xmath254 . assuming @xmath255 for some @xmath256 , we have @xmath257 since @xmath258 , it suffices to prove the bounds for @xmath259 . proposition [ tau.upper ] gives the upper bound , @xmath260 for @xmath261 with @xmath262 . thus we have only to prove the lower bound . consider the hadamard products appearing in the factorization of @xmath254 given in proposition [ tau.factor ] . these products are of order @xmath49 but not finite type . to work around this , we consider products of the form @xmath263 . by lindelf s theorem ( see e.g. ( * ? ? ? 2.10.1 ) ) , such functions are of finite type and so satisfy estimates , @xmath264 the minimum modulus theorem ( * ? ? ? 3.7.4 ) , gives corresponding lower bounds , @xmath265 provided we stay a distance at least @xmath266 away from the zeros , for some @xmath256 . using these estimates together with proposition [ tau.factor ] gives @xmath267 provided @xmath268 and @xmath269 stay at least a distance @xmath266 away from the sets @xmath270 and @xmath271 . assuming @xmath272 $ ] , we already know @xmath273 from , provided @xmath274 stays at least a distance @xmath266 away from @xmath275 . similarly , for @xmath276 $ ] , we already have an upper bound on @xmath277 in conjunction with ( [ tt.bnd ] ) , these estimates give the desired lower bound of @xmath278 in the first quadrant , except that we have been required to excise disks near the points not only of @xmath253 , but also its rotations by @xmath279 . however , once we have obtained estimates of @xmath280 itself , the missing disks can be filled in using the maximum modulus theorem . for a radial potential , @xmath281 , the scattering matrix @xmath95 acting on @xmath102 is diagonalized by spherical harmonics . in this section we will develop a formula for the matrix elements of @xmath95 which will then be used to produce estimates in [ asym.sec ] . in geodesic polar coordinates , @xmath282 , and the laplacian takes the form , @xmath283 the spherical harmonic @xmath284 is an eigenfunction of @xmath285 satisfying @xmath286 the indices range over @xmath287 and @xmath288 with @xmath289 as in [ potsc.sec ] , we use the boundary defining function @xmath290 this choice is made so that the metric induced on the conformal infinity by @xmath291 is the standard sphere metric on @xmath102 . the scattering matrix elements @xmath292_l$ ] are the eigenvalues , meromorphic in @xmath164 , of the spherical harmonics @xmath293_l y_l^m.\ ] ] to compute @xmath292_l$ ] , we consider a generalized eigenfunction @xmath294 on @xmath0 . from the eigenfunction equation , @xmath295 we derive the coefficient equation @xmath296 u = 0.\ ] ] if we rewrite this equation in the variable @xmath59 , then indicial roots at @xmath297 are @xmath164 and @xmath298 , implying that @xmath299 will in general have a two - part asymptotic expansion with leading terms of these orders as @xmath300 . the scattering matrix eigenvalue associated to @xmath284 appears as the ratio of the leading coefficients , i.e. @xmath301_l \rho^{s } \right),\ ] ] as @xmath300 . the homogeneous equation ( @xmath22 ) is solved by legendre functions , with the independent solutions , @xmath302 where @xmath303 here @xmath304 is the normalized q - function introduced by olver @xcite , which is analytic in its parameters . it is related to the standard definition by @xmath305 . from the well - known asymptotics of the legendre functions we obtain @xmath306 and also @xmath307 if @xmath20 is assumed to have support in @xmath308 , there exists a solution @xmath309 of the full equation that is equal to @xmath310 for @xmath311 . this solution will generally have a leading singularity proportional to @xmath312 at @xmath313 , so that @xmath314 is not a smooth solution at the origin . however , since @xmath315 is an independent solution , we can cancel the singularity at @xmath313 by taking the combination , @xmath316 where the coefficients are given by the limits @xmath317 by the indicial equation at @xmath313 , canceling the leading @xmath312 term at @xmath318 removes that whole part of the expansion , so that expansion of @xmath319 at @xmath313 has only the part with leading term proportional to @xmath320 . hence @xmath321 is regular at the origin and defines a generalized eigenfunction on @xmath0 . we can therefore deduce from and that @xmath322_l = 2^{n-2s } \frac{\gamma(\frac{n}2 - s)}{\gamma(s-\frac{n}2 ) } \frac{f^k(n - s)}{f^k(s)},\ ] ] with @xmath323 . for future reference , we also introduce the unperturbed version of these coefficients , @xmath324 in this notation , the ( well - known ) formula for the unperturbed scattering matrix elements becomes @xmath325_l = 2^{n-2s } \frac{\gamma(\frac{n}2 - s)}{\gamma(s-\frac{n}2 ) } \frac{f^k_0(n - s)}{f^k_0(s)}.\ ] ] for a radial step potential of the form @xmath326 , with @xmath327 , we can write the functions @xmath309 explicitly . the coefficient solutions for @xmath328 are legendre functions @xmath329 , with the shifted parameter @xmath330 a simple ode matching problem at @xmath331 then shows that for @xmath326 , the coefficients @xmath332 appearing in are given by @xmath333\big|_{z = \cosh r_0}\ ] ] where @xmath334 denotes the wronskian . the resonances in the @xmath284 mode can then be characterized explicitly a the zeros of @xmath335 ; this is the basis of the resonance plots in figures [ rpotres3.fig ] and [ plot1i.fig ] . the main goal of this section is a precise estimate of the eigenvalues of the relative scattering matrix @xmath160 . for convenience , let us set @xmath336_l,\ ] ] where @xmath337 as before . [ lamk.prop ] for all @xmath338 , @xmath339 , and @xmath93 sufficiently large , and assuming that @xmath340 , we have @xmath341 with constants that depend only on the potential @xmath20 and on @xmath342 . ( for @xmath218 , the exponent @xmath343 is replaced by its limiting value , @xmath344 . ) the strategy for the proof is analogous to that of zworski @xcite . using and we can write @xmath345 by a standard application of variation of parameters to the ode for @xmath346 , with @xmath347 for @xmath311 as the boundary condition , we obtain the integral equation @xmath348 where the integral kernel is @xmath349(t)},\ ] ] with @xmath350 denoting the wronskian . using the well - known formula for the wronskian of a pair of legendre functions , this kernel reduces to @xmath351 . \end{split}\ ] ] formally , we can solve the integral equation for @xmath352 using the series @xmath353 , with @xmath354 the unperturbed solution and @xmath355 defined recursively by @xmath356 we first need asymptotic estimates on these @xmath355 which justify the convergence of this series , for @xmath357 sufficiently large . then we can derive estimates for the scattering matrix elements using . to analyze the sequence @xmath358 we first recall some asymptotic estimates on the legendre functions from @xcite , obtained using techniques from olver @xcite . set @xmath359 , so that @xmath360 . the liouville transformation takes the legendre equation to an approximate airy equation with the variable @xmath361 defined by @xmath362 where @xmath363.\ ] ] the legendre functions can then be approximated in terms of airy functions of @xmath361 . the asymptotics of @xmath364 can be worked out fairly easily . as @xmath318 , we have @xmath365 where @xmath366 and as @xmath367 , we have @xmath368 where @xmath369 the liouville transformation yields the following asymptotic result , derived in borthwick ( * ? ? ? * prop a.1 ) . assuming that @xmath213 , @xmath370 $ ] and @xmath371 , we have @xmath372^{\frac14 } } \>e^{-k p(\alpha ) } \bigl [ { { \rm ai}}\bigl(k^{\frac23 } e^{\frac{2\pi i}3 } \zeta\bigr ) + h_1(k , \alpha , r)\bigr],\ ] ] and @xmath373^{\frac14 } } \ > e^{kq(\alpha ) } \bigl [ { { \rm ai}}\bigl(k^{\frac23 } \zeta\bigr ) + h_0(k,\alpha , r)\bigr],\ ] ] with the error estimates , @xmath374 for the most part , we will be content with the sharp upper bounds derived from these asymptotics . from ( * ? ? ? * cor . a.3 ) we cite : [ pq.bounds ] assuming that @xmath375 , @xmath376 $ ] , and @xmath377 $ ] , we have the following estimates : @xmath378}\ ] ] and @xmath379},\ ] ] where @xmath77 depends only on @xmath36 and @xmath380 . beyond the upper bounds of proposition [ pq.bounds ] , which serve to control the error terms in our expansion , we need also a lower bound to apply to the leading term . [ p.asym.prop ] assume that @xmath376 $ ] , @xmath377 $ ] , and that for some sufficiently large @xmath381 we have both @xmath382 and @xmath383 . then @xmath384^{\frac14 } } \>e^{k\phi(\alpha;r)}.\ ] ] the assumption that @xmath376 $ ] implies that @xmath385 $ ] also . hence @xmath386 is bounded away from @xmath387 and we may apply the airy function asymptotic ( * ? ? ? * eq . ( 4.4.03 ) ) , @xmath388,\ ] ] for @xmath389 . this result , along with and , gives @xmath390^{\frac14 } } \>e^{k\phi(\alpha;r ) } \\ & \qquad \times \left [ 1 + o\left(k^{-1}\right ) + o\left(k^{-\frac13 } { \left|k\alpha \right|}^{-\frac23}\right ) + o_{\varepsilon}({\left|k\phi \right|}^{-1})\right ] \end{split}\ ] ] to complete the proof , we can deduce from the analysis of @xmath391 in the proof of ( * ? ? ? * prop a.1 ) that , for @xmath376 $ ] , @xmath392 for the application of proposition [ pq.bounds ] to the estimate of the iterated solutions @xmath355 , we essentially need only two estimates for the inductive step , corresponding to the two exponentials appearing in the asymptotics . for the second of these estimates we will need to bring in the hypothesis of theorem [ asymp.thm ] : that @xmath393 is continuous near @xmath40 and satisfies @xmath394 for some constants @xmath43 and @xmath395 . [ jekp.lemma ] for @xmath396 , @xmath397 and @xmath398 , we have @xmath399 where @xmath400 . under the same hypotheses , and assuming also , there exists @xmath401 such that for @xmath402 , @xmath403 ( \sinh r)^{-\frac{n-1}2}. \end{split}\ ] ] the assumptions on @xmath164 correspond to the hypotheses of proposition [ pq.bounds ] . by the conjugation symmetry we can assume @xmath404 . applying the estimates and in the definition of @xmath405 from gives @xmath406}\\ & \hskip.5in\qquad\times \int_r^{r_0 } \left [ e^{k{\operatorname{re}}(\phi(\alpha;r ) - 2\phi(\alpha;t ) ) } + e^{-k{\operatorname{re}}\phi(\alpha;r ) } \right ] { \left|v(t ) \right| } \sinh t\>dt . \end{split}\ ] ] since @xmath407 is increasing , we can replace the expression in brackets by @xmath408 . the claim then follows easily from an estimate based on stirling s formula : @xmath409 } \le ck^{-1},\ ] ] valid for @xmath410 and @xmath213 . for the estimate , the analog of , together with , yields @xmath411 { \left|v(t ) \right| } \sinh t\>dt . \end{split}\ ] ] the bound on the first term in the bracket is clear , since @xmath407 is increasing . for the second term , we need to apply prop [ laplace.prop ] and the assumption . to check the hypotheses , we simply compute the derivatives , @xmath412 where the prime denotes an @xmath146 derivative . in particular , for @xmath404 we have @xmath413 and @xmath414 proposition [ laplace.prop ] then gives @xmath415 for @xmath402 . in the estimates that follow , we consider expressions involving both @xmath164 and @xmath298 , but always with the convention that @xmath416 ( equivalently , @xmath410 ) . our recursive estimates are most straightforward in the case of @xmath417 , which corresponds to the expansion of the denominator in formula for the scattering matrix element . [ vj.good ] for @xmath396 , @xmath418 and @xmath419 , there exists a constant @xmath420 , such that @xmath421 } ( \sinh r)^{-\frac{n-1}2}.\ ] ] the proof is by induction over @xmath422 . the estimate for @xmath423 follows directly from corollary [ pq.bounds ] . and the inductive step that extends the estimate from @xmath424 to @xmath425 the follows immediately from the iterative formula and lemma [ jekp.lemma ] . to handle the terms involving @xmath426 , it is useful to note the legendre function identities , @xmath427 these imply the symmetry @xmath428 we can also write the @xmath429 solution in terms of @xmath430 and @xmath431 using . then for @xmath432 , proposition [ pq.bounds ] gives the estimate @xmath433 ( \sinh r)^{-\frac{n-1}2},\ ] ] where @xmath434 [ vj.bad ] for @xmath396 , @xmath418 and @xmath419 , @xmath435 ( \sinh r)^{-\frac{n-1}2}. \end{split}\ ] ] the @xmath436 case was already dealt with in . and if we start from the inductive assumption that @xmath437 ( \sinh r)^{-\frac{n-1}2},\ ] ] then under these assumptions lemma [ jekp.lemma ] implies the bound @xmath438 ( \sinh r)^{-\frac{n-1}2}. \end{split}\ ] ] the result follows by induction . from lemmas [ vj.good ] and [ vj.bad ] we deduce that for @xmath419 and @xmath439 there exists @xmath440 such that the two series , @xmath441 converge absolutely and uniformly for @xmath442 $ ] and @xmath443 . under these assumptions , we can then express the scattering matrix element @xmath444 given in as a sum over of the limiting values @xmath445 to analyze @xmath444 , we will need to consider the ratios @xmath446 . the dominant term is given by the following : [ a1a0.lemma ] for @xmath447 , @xmath448 , and @xmath449 , and @xmath450 , @xmath451 with constants that depend only on @xmath452 . the exact formula for @xmath453 can easily be deduced from the formula for @xmath454 , @xmath455 using the asymptotics and the definition of @xmath456 we find that @xmath457 to apply our legendre estimates , which require @xmath458 , we use the reflection formulas to flip the arguments from @xmath459 . the resulting expression for the ratio @xmath460 has two parts , corresponding to the two terms in the @xmath461-reflection formula . we will write these as : @xmath462 where @xmath463 and @xmath464 as usual , by the conjugation symmetry we assume that @xmath376 $ ] . by proposition [ p.asym.prop ] , for @xmath465 in this sector we can find @xmath401 such that for @xmath466 and @xmath467 we have @xmath468^{-\frac14 } \frac{e^{k\phi(\alpha;t ) - kp(\alpha)}}{\gamma(k+1)},\ ] ] with constants that depend only on @xmath380 and @xmath36 . thus , @xmath469 we can estimate the integral using the version of laplace s method given in prop [ laplace.prop ] . from the expressions for the derivatives of @xmath391 , we see that for @xmath376 $ ] , @xmath470 } { \left|\phi''(\alpha;\cdot ) \right| } \le c_{\varepsilon}.\ ] ] under the assumption , proposition [ laplace.prop ] then gives , for @xmath448 , @xmath471 with constants that depend only on @xmath380 and @xmath20 . the formula for @xmath472 was given in . for @xmath376 $ ] the integral estimate reduces to @xmath473 an application of stirling s formula gives the estimate , for @xmath474 and @xmath450 , @xmath475 + o(\log ( 1+{\varepsilon}^{-1 } ) ) . \end{split}\ ] ] we note also that by definition , @xmath476.\ ] ] the combined estimate , for @xmath447 , @xmath443 , @xmath449 , and @xmath450 , is @xmath477 with constants that depend only on @xmath380 and @xmath20 . to control the second integral , we apply corollary [ pq.bounds ] to estimate @xmath478 } { \left\vert v \right\vert}_\infty \sinh r_0,\ ] ] under the same assumptions as for @xmath479 . the stirling estimate then shows that @xmath480 . the other terms in the series expansion for @xmath444 can now be estimated using lemmas [ vj.good ] and [ vj.bad ] . [ aj.rem ] for @xmath419 , @xmath418 , and @xmath448 we have @xmath481 and @xmath482.\ ] ] from lemma [ vj.good ] we obtain the estimates , for @xmath396 , @xmath418 and @xmath483 , @xmath484}.\ ] ] in conjunction with the stirling estimate this gives . for the estimate in the other half - plane , we start by taking the @xmath485 limit in lemma [ vj.bad ] to obtain , for @xmath396 , @xmath418 and @xmath419 , @xmath486 after dividing by @xmath487 and substituting the definitions of @xmath488 and @xmath489 , we find that @xmath490 . \end{split}\ ] ] the first expression is estimated using and and the second by . by and lemmas [ vj.good ] and [ vj.bad ] , for @xmath448 we can represent @xmath491 \left[1 + \sum_{j=1}^\infty \frac{f^k_j({\tfrac{n}{2}}+k\alpha)}{f^k_0({\tfrac{n}{2}}+k\alpha ) } \right]^{-1}.\ ] ] under the hypotheses , gives @xmath492 while from we have @xmath493.\ ] ] assuming @xmath448 , these estimates give @xmath494 the estimate then follows from lemma [ a1a0.lemma ] ( after noting that @xmath495 ) . this completes the proof in the case @xmath448 . the legendre function estimates given in [ leg.asym.sec ] are applicable only for @xmath357 sufficiently large . although this covers the main region of interest , where @xmath357 and @xmath496 are comparable in magnitude , we still need to estimate for @xmath444 when @xmath357 is small . we could be satisfied with fairly rough estimates , since the low - frequency terms make a contribution of order @xmath93 to asymptotic of leading order @xmath497 . however , we need lower bounds in particular , and there is no general estimate that will provide these . fortunately , the asymptotics of the legendre functions for large @xmath498 with @xmath357 fixed are well - covered in the literature . from olver ( * ? ? ? * thm 12.9.1 and 12.12 ) , we have the following : [ pq.ik.prop ] for @xmath499 and @xmath500 $ ] we have @xmath501 using standard estimates for the modified bessel functions , we obtain the bounds @xmath502 for @xmath503 where @xmath504 for bounded @xmath357 the inductive estimates are furnished by the following : for @xmath505 with @xmath506 and @xmath507 , @xmath508 and @xmath509 ( \sinh r)^{-\frac{n-1}2}. \end{split}\ ] ] by proposition [ pq.ik.prop ] and we can estimate @xmath510 ( \sinh r)^{-\frac{n-1}2 } ( \sinh t)^{\frac{n+1}2}.\ ] ] the estimate the follows easily from the definition of @xmath511 . using the same estimate for @xmath512 we break up the left - hand side of into two terms . estimation of the term with @xmath513 in front works just as in the estimate for . the term with @xmath514 out front involves the integral @xmath515 using the laplace estimate from proposition [ laplace.prop ] we have @xmath516 if @xmath517 then we must also consider @xmath518 for @xmath93 sufficiently large , we will have @xmath519 , so that this extra term may be combined with the laplace estimate term . to complete the estimate , we will show that for @xmath357 fixed , @xmath500 $ ] , @xmath520 , and @xmath521 , we have @xmath522 with constants that depend only on @xmath523 . from the definition of @xmath524 , we can easily check , for @xmath146 fixed and @xmath525 , that @xmath526 thus implies in the case where @xmath357 is bounded . for this argument we set @xmath527 , and we make the assumption according to proposition [ pq.ik.prop ] and [ ik.bounds ] , we have @xmath528 by induction using lemma [ j.fminus ] , we obtain the estimates @xmath529 using in addition to proposition [ pq.ik.prop ] and [ ik.bounds ] yields @xmath530 ( \sinh r)^{-\frac{n-1}2},\ ] ] where @xmath531 and @xmath532 the induction argument corresponding to that of lemma [ vj.bad ] then gives @xmath533 ( \sinh r)^{-\frac{n-1}2}. \end{split}\ ] ] these estimates show that the two series @xmath534 converge for @xmath93 sufficiently large . by arguing as in lemma [ aj.rem ] , we find @xmath535 for @xmath93 sufficiently large , with constants that depend only on @xmath536 , and @xmath20 . as in the proof of lemma [ a1a0.lemma ] , we split @xmath537 where @xmath538 ^ 2 v(t)\sinh t\>dt,\ ] ] and @xmath539 using proposition [ pq.ik.prop ] and the well - known asymptotic @xmath540 valid for @xmath541 , we can reduce @xmath479 to @xmath542 proposition [ laplace.prop ] then gives @xmath543 applying stirling s formula , and assuming @xmath544 in addition to the other hypotheses , we find @xmath545 with constants depending on @xmath523 . the second integral , @xmath546 , is easily seen via proposition [ pq.ik.prop ] to be @xmath547 , so the result follows . in this section we will complete the proof of theorem [ asymp.thm ] by establishing an asymptotic for @xmath254 in the radial potential case . in terms of the matrix elements @xmath444 defined in , the relative scattering determinant is @xmath548 where @xmath337 and @xmath198 is the multiplicity . our main goal is the following : [ tau.asymp ] suppose @xmath254 is the relative scattering determinant corresponding to a radial potential @xmath549 $ ] , such that @xmath20 is continuous near @xmath36 and satisfies @xmath550 for some @xmath395 . then assuming @xmath520 and @xmath551 @xmath552 where @xmath51 was defined in . the leading contribution to this estimate comes from terms for which @xmath553 exhibits exponential growth for large @xmath240 . this is the case covered by proposition [ lamk.prop ] . we must also account for the cases where the exponent @xmath554 is near zero or negative , for which proposition [ lamk.prop ] gives no information . for the exponential decay estimate we turn to the formula for the relative scattering matrix used in @xcite . [ lamk.upper.prop ] for @xmath419 , @xmath467 , and @xmath555 @xmath556 where @xmath524 was defined in . set @xmath557 for @xmath184 . let @xmath185 be a cutoff function with @xmath188 for @xmath187 and @xmath186 for @xmath189 . then set @xmath558 , so that @xmath191 for @xmath192 and @xmath193 for @xmath194 . then from the proof of ( * ? ? ? * lemma 4.1 ) we have @xmath559 r_v(s ) [ \delta,\chi_1 ] e_0(n - s),\ ] ] where @xmath196 is the unperturbed poisson operator on @xmath0 . since @xmath20 is radial , @xmath95 is diagonalized by spherical harmonics and @xmath444 may be computed as a matrix element , @xmath560 y_l^m \right\rangle } \\ & = ( 2s - n ) \bigl\langle { \mathbbm{1}}_{[r_2,r_3 ] } e_0(s)y_l^m , [ \delta , \chi_2 ] r_v(s ) [ \delta,\chi_1 ] { \mathbbm{1}}_{[r_1,r_2 ] } e_0(n - s)y_l^m \bigr\rangle , \end{split}\ ] ] where @xmath561}$ ] denotes the multiplication operator of the characteristic function @xmath562}(r)$ ] . as in the proof of proposition [ tau.upper ] , we can derive from proposition [ res.est ] the bound @xmath563 r_v(s ) [ \delta_0,\chi_1 ] \bigr\vert \le c_{{\varepsilon},r_0}\delta^{-4},\ ] ] under the assumptions @xmath564 and @xmath565 , which keep @xmath566 bounded away from the spectrum of @xmath52 . also , using the decomposition of the poisson operator in polar coordinate from ( * ? ? ? 4.2 ) , we can compute @xmath567 where @xmath568 with @xmath569 as before . thus from and the resolvent estimate we have the bound @xmath570^{\frac12 } \\ & \qquad \left [ \int_{r_2}^{r_3 } { \left|e_k(s ; r ) \right|}^2\>(\sinh r)^n\>dr\right]^{\frac12}. \end{split}\ ] ] substituting in with the definition if @xmath571 and using the reflection identity @xmath572 , we can rewrite this as @xmath573^{\frac12 } \left [ \int_{r_2}^{r_3 } { \left|p^{-k}_{\nu}(\cosh r ) \right|}^2 \sinh r\>dr \right]^{\frac12}. \end{split}\ ] ] applying corollary [ pq.bounds ] and using the fact that @xmath574 is increasing allows us to reduce this estimate to @xmath575}.\ ] ] the result then follows from and the definition of @xmath524 . the final issue that we need to resolve before proving theorem [ tau.asymp ] is the behavior on the zone where @xmath576 . this region contains non - trivial zeros of @xmath444 , so to produce a lower bound on @xmath577 is delicate . the only tool we have for this situation is the minimum modulus theorem for entire functions . [ inter.prop ] for any @xmath256 we have @xmath578 under the assumption that @xmath579 . the solutions @xmath309 are analytic as functions of @xmath164 , so the functions @xmath332 appearing in the formula for the scattering matrix element are analytic as well . using stirling , we can easily produce a crude bound , @xmath580 for all @xmath581 with with @xmath77 independent of @xmath357 . from the estimates on the series @xmath582 produced in [ hfreq.sec ] and [ lfreq.sec ] , and the straightforward bound @xmath583 , we claim the same result holds in the perturbed case : @xmath584 here @xmath77 depends on @xmath20 but not on @xmath357 . ( note our bounds on @xmath585 and @xmath586 omit the sectors @xmath587 $ ] , but this restriction was necessary only for the lower bounds . a simple application of phragmn - lindelf extends the upper bounds to the missing sectors . ) we can now apply the minimum modulus theorem ( see e.g. ( * ? ? ? 3.7.4 ) ) to obtain a corresponding lower bound : for @xmath256 @xmath588 under the restriction that @xmath589 . ( here we use also the fact that for any @xmath357 the zeros of @xmath332 are included in the set @xmath590 . ) applying these upper and lower bounds to the matrix element formula yields the result . we have already noted the more general upper bound in proposition [ tau.upper ] , so our goal here is to produce a corresponding lower bound : @xmath591 by conjugation we can assume that @xmath592 . we need to estimate the sum @xmath593 from the definition we can estimate the multiplicities by @xmath594 for @xmath223 $ ] , let @xmath226 be the implicit solution of @xmath595 , as shown in figure [ hplot.fig ] . since @xmath596 for @xmath597 , we can see that @xmath598 for @xmath93 sufficiently large . thus , under the assumption @xmath599 , @xmath600 is bounded below by @xmath601 for @xmath602 . on the other hand , for @xmath603 , we have @xmath604 , and if @xmath605 is large we can use the approximation @xmath606 . we conclude that , under these assumptions , @xmath607 for all @xmath93 sufficiently large . thus for @xmath599 with @xmath93 sufficiently large and @xmath340 , proposition [ lamk.prop ] implies @xmath608 with this estimate in mind , we divide the sum into three pieces , roughly according to the sign of @xmath609 : @xmath610 the dominant term is @xmath611 , and from and we obtain @xmath612 } \frac{2k^{n-1}}{\gamma(n ) } h\left(\frac{ae^{i\theta}}{k } , r_0\right ) - o(a^{n}).\ ] ] the summand is monotonic as a function of @xmath357 , so we can estimate with an integral : @xmath613 } k^{n+1 } h\left(\frac{ae^{i\theta}}{k } , r_0\right)\>dk - o(a^n).\ ] ] we can then make the substitution @xmath232 to reduce this to @xmath614 using the fact that @xmath615 for @xmath616 large , we can extend the lower limit of integration by adjusting the error term , @xmath617 next we consider the middle term @xmath244 . proposition [ inter.prop ] implies that @xmath618 is @xmath619 for @xmath357 in the range defined by @xmath620 $ ] , and there are @xmath621 values of @xmath357 in this range . hence @xmath622 finally , for @xmath623 , we will use proposition [ lamk.upper.prop ] which shows that @xmath624 is exponentially small in this region , together with the simple estimate , @xmath625 this yields , for small @xmath626 , @xmath627 for @xmath628 $ ] , we have @xmath629 so by choosing @xmath630 , with @xmath631 sufficiently small , we can bound the exponent by @xmath632 . we obtain , for @xmath93 sufficiently large and @xmath520 , @xmath633 this yields @xmath634 combining with these estimates on @xmath244 and @xmath623 completes the proof of . under the hypotheses of the theorem , theorem [ tau.asymp ] gives @xmath635 assuming @xmath636 . to fill the gap near @xmath637 , we use the general estimate from lemma [ lind.lemma ] . this gives @xmath638 for @xmath93 in some unbounded set @xmath639 that depends on the choice of @xmath640 in the lemma . we conclude that @xmath641 for @xmath642 . taking @xmath643 then gives @xmath644 in combination with proposition [ count.prop ] , this completes the proof . in this section we ll develop the theorems on resonance distribution outlined in [ intro.sec ] . these results draw on techniques from christiansen @xcite , christiansen - hislop @xcite , and borthwick - christiansen - hislop - perry @xcite . we will only sketch the proofs for cases that are very similar to these earlier results . before stating the results we recall a standard definition from several complex variables . a _ pluripolar _ subset @xmath645 of a connected domain @xmath646 is the polar set of a plurisubharmonic function @xmath647 on @xmath84 , i.e. the set @xmath648 . pluripolar sets have lebesgue measure zero in @xmath649 . moreover , for @xmath650 the real part @xmath651 of a pluripolar set @xmath645 will have lebesgue measure zero in @xmath652 . the first result shows that the order of growth in the bound @xmath653 of theorem [ upper.thm ] is saturated for generic potentials . [ kgen.thm ] let @xmath654 or @xmath655 . given a compact subset @xmath45 with non - empty interior , the set @xmath656 is _ baire typical _ in @xmath657 ( i.e. is a dense @xmath658 subset ) . this is the hyperbolic analog of the main result from christiansen - hislop @xcite . those techniques were already adapted to hyperbolic manifolds in @xcite , so we will not repeat all the details here . suppose @xmath659 is a holomorphic family of potentials for @xmath660 , an open connected set . with a relatively simple adaption of the proof of ( * ? ? ? 1.1 ) , similar to the version given in @xcite , we can show that if @xmath661 holds for some particular @xmath662 , then it holds for @xmath663 , where @xmath645 is a pluripolar set . given any @xmath664 we can choose @xmath665 a radial potential supported in @xmath666 to which theorem [ asymp.thm ] applies and form the family @xmath667 . in this way we conclude that @xmath668 has maximal rate of growth except for @xmath669 in some pluripolar set . then one can argue exactly as in christiansen - hislop @xcite or @xcite to characterize the class of @xmath20 with maximal growth rate as baire typical . following christiansen @xcite , we can prove a variant of theorem [ kgen.thm ] involving the sharp asymptotic constant rather than just the order of growth , at the cost of restricting the supporting set @xmath666 to a closed ball @xmath670 . [ order.thm ] suppose @xmath671 is a holomorphic family of potentials for @xmath672 , an open connected subset . assume that @xmath673 for all @xmath669 , and that the condition , @xmath674 holds for some @xmath662 , where @xmath675 is the asymptotic constant defined in . then there exists a pluripolar set @xmath676 such that holds for @xmath677 . the proof is closely related to the proof of ( * ? ? ? * thm . the only complication is that in our case @xmath678 has infinitely many poles on the positive real axis for @xmath23 odd , at the points @xmath679 . to handle this , we introduce a function @xmath680 defined as a hadamard product @xmath681 where @xmath682 denotes an elementary factor . the extra zeros at @xmath683 are inserted so that @xmath684 will have a regularly distributed zero set in the sense of levin @xcite . by ( * thm ii.2 ) , there is a smooth indicator function @xmath685 such that @xmath686 uniformly for @xmath687 . we can then cancel off the extra poles of @xmath688 by introducing @xmath689 whose poles for @xmath690 correspond to the ( finitely many ) points of @xmath691 on that side . by proposition [ count.prop ] , we then have @xmath692 where @xmath693 and @xmath694 by , @xmath695 exists and is given by some constant @xmath696 . theorem [ upper.thm ] thus gives the bound @xmath697 and by hypothesis , the maximum is achieved at @xmath698 . for @xmath93 sufficiently large , the function @xmath699 is plurisubharmonic in @xmath669 , and so by lelong - gruman ( * ? ? ? 1.39 ) , there exists a pluripolar set @xmath676 such that @xmath700 for @xmath677 . this proves the claim . note that as in theorem [ kgen.thm ] , we could use theorem [ order.thm ] in conjunction with perturbation by radial potentials to show that the condition holds on a baire typical subset of @xmath701 . many of the results of christiansen @xcite concern the resonance counting function restricted to sectors . in the hyperbolic case it is natural to center these sectors at @xmath112 and define : @xmath702\bigr\}.\ ] ] the `` averaged '' sectorial counting function is denoted with a tilde : @xmath703 the results below refer to the indicator function @xmath51 , defined in and illustrated in figure [ indicator.fig ] . indicator.pdf ( -4,0)@xmath704 ( 95,0)@xmath705 ( 80,26)@xmath706 for @xmath707 [ msector.thm ] suppose @xmath708 has counting function satisfying @xmath50 . then for @xmath709 , with @xmath710 , @xmath711 note that @xmath712 contributes only if @xmath23 is odd , and then only if @xmath713 . it suffices to prove only the case @xmath714 , @xmath715 t^{n+1 } + o(t^{n+1}).\ ] ] if @xmath20 is real , then self - adjointness implies a conjugation symmetry in @xmath14 , and the @xmath716 case would be equivalent to this by reflection . in the general case , the indicator function @xmath717 still posseses the conjugation symmetry even though @xmath14 does not , so the proof for @xmath716 is identical to the one we will give for @xmath714 . the intermediate case @xmath718 \subset ( \tfrac{\pi}2 , \tfrac{3\pi}2)$ ] follows by subtracting the two endpoint cases . to begin we apply the argument principle to the integral of @xmath111 over a sector given by @xmath719 and @xmath720 . the result is @xmath721 the term @xmath722 is bounded by a fixed constant , the total number of discrete eigenvalues . and for the first segment of the contour integral , on critical line @xmath161 , we have a bound @xmath723 by proposition [ sigma.bound ] . by applying the cauchy - riemann equations to the integrands of the two remaining integrals , we obtain @xmath724 where @xmath725 ( this is a slight adaptation of the definition from levin @xcite , moving the center to @xmath726 . ) now , to eliminate the derivatives , we divide by @xmath120 and then integrate over @xmath120 from @xmath727 to @xmath93 and over @xmath728 from @xmath727 to @xmath238 . note that proposition [ sigma.bound ] implies @xmath729 . the result of these integrations is therefore that @xmath730d\eta \\ & \qquad = \frac{n+1}{2\pi } \int_{0}^{-\frac{\pi}2+\theta } \int_{-\frac{\pi}2}^{-\tfrac{\pi}2+\eta } \log { \left|\tau({\tfrac{n}{2}}+ ae^{i\omega } ) \right| } \>d\omega \>d\eta + \frac{n+1}{2\pi } \int_0^a j_\tau(t , \theta-\pi)\>\frac{dt}{t}\\ & \qquad\qquad + o(t^{n } ) . \end{split}\ ] ] to continue , we use the background function @xmath684 introduced in the proof of theorem [ order.thm ] to define @xmath731 . if @xmath1 has discrete eigenvalues , then @xmath732 will still have poles at a finite set @xmath733 . in this case we can simply replace @xmath734 which will remove the poles without affecting the asymptotics . for notational convenience , we will simply assume that @xmath732 is analytic for the rest of the proof . by proposition [ tau.upper ] and we have a bound for @xmath735 , @xmath736 as @xmath737 , where the maximum modulus principle is used to remove the restriction on the values of @xmath93 . thus @xmath738 for @xmath735 . the left - hand side is by definition the indicator function of @xmath739 , so ( * ? ? ? * thm i.28 ) gives , for any @xmath740 , @xmath741 for @xmath742 and @xmath743 . on the other hand , by the assumption on the asymptotics of @xmath21 , together with proposition [ count.prop ] , we have @xmath744 in conjunction with this implies @xmath745 then , by the final argument from ( * ? ? ? * thm iv.3 ) ( used also in ( * ? ? ? 2.2 ) ) , @xmath746 is of _ completely regular growth _ in the angle @xmath747 , with indicator function equal to @xmath748 . this means that @xmath749 uniformly for @xmath750 , for @xmath93 outside of some subset of zero relative measure in @xmath751 . by ( * lemma iii.2 ) , we have @xmath752 for @xmath735 . the corresponding limit holds also for @xmath753 , since @xmath680 has completely regular growth by construction . thus @xmath754 for @xmath735 . returning to , we can apply to obtain @xmath755d\eta \\ & \qquad = \frac{n+1}{2\pi } \int_{0}^{-\frac{\pi}2+\theta } \int_{-\frac{\pi}2}^{-\tfrac{\pi}2+\eta } h_{r_0}(\omega ) \>d\omega \>d\eta + \frac{1}{2\pi ( n+1 ) } h_{r_0}(\theta-\pi ) . \end{split}\ ] ] the integral over @xmath728 can be removed using ( * ? ? ? * lemma 5.4 ) , provided we avoid the point @xmath756 where @xmath717 fails to be differentiable . we then complete the proof using the fact that the asymptotic @xmath757 is equivalent to @xmath758 ( see , e.g , . * lemma 1 ) ) . one interesting feature of theorem [ msector.thm ] is that the indicator function remains the same whether @xmath52 is self - adjoint or not . even though the conjugation symmetry is broken in the non - self - adjoint case , the resonance distribution still exhibits this symmetry in an asymptotic sense . this situation is illustrated in figure [ plot1i.fig ] , which shows the resonance plots for real and imaginary radial step potentials in @xmath759 . the breaking of the conjugation symmetry is clear only in the vicinity of the origin . plot1i.pdf ( 15,0)@xmath17 ( 70,0)@xmath760 finally , and once again following @xcite , we present some results on distribution of resonances in sectors that hold in a generic sense . the first result concerns the resonances in a narrow sector bordering on the critical line @xmath161 , as shown in fig [ sector.fig ] . the theorem gives a lower bound independent of @xmath380 on the number of resonances in this strip that is independent of @xmath380 . as in the remark following theorem [ order.thm ] , we could use perturbation by radial potentials to show that this condition holds for a baire typical subset of @xmath701 . sector.pdf [ ngeneric ] suppose @xmath671 is a holomorphic family of potentials for @xmath672 , an open connected subset . assume that @xmath673 for all @xmath669 , and that the condition , @xmath761 holds for some @xmath662 , where @xmath675 is the asymptotic constant defined in . then for any @xmath762 , there exists a pluripolar set @xmath763 such that @xmath764 the corresponding result holds in the conjugate sector also , i.e. for @xmath765 . let us denote by @xmath766 the relative scattering determinant associated to @xmath767 . as in the proof of theorem [ kgen.thm ] we introduce the background function @xmath684 to cancel the poles of @xmath688 coming from @xmath107 ( necessary only if @xmath23 is odd ) . then @xmath768 is analytic for @xmath207 , except for finitely many poles related to the discrete spectrum . as in the proof of theorem [ msector.thm ] , we note that these poles are easily cancelled off ( see ( * ? ? ? * lemma 5.2 ) for details ) , and for the sake of exposition we assume that @xmath769 is analytic for the rest of the proof . from , assuming @xmath770 we have @xmath771 where @xmath772 and @xmath773 now we use the fact that @xmath774 is plurisubharmonic as a function of @xmath669 , and argue as in the proof of theorem [ order.thm ] . using the assumption on @xmath775 and ( * ? ? ? 1.39 ) , we find that there exists a pluripolar set @xmath776 such that @xmath777 \\ & \qquad = \frac{1}{2\pi(n+1 ) } h_{r_0}(\theta-\pi ) + \frac{n+1}{2\pi } \int_{0}^{-\frac{\pi}2+\theta } \int_{-\frac{\pi}2}^{-\frac{\pi}2+\eta } h_{r_0}(\omega)\>d\omega\>d\eta , \end{split}\ ] ] for @xmath778 . from this point we can simply follow the end of the proof of ( * ? ? ? 1.2 ) , to take the limit @xmath779 . we can be slightly more explicit about the constant appearing in theorem [ ngeneric ] , although it does nt reduce to a simple formula . we start from , written as @xmath780 where @xmath226 is the implicit solution of @xmath595 . we can easily compute @xmath781 noting also that @xmath782 , we obtain @xmath783 our final result in this section concerns the `` expected value '' of the resonance counting function , computed as a weighted average over a complex family of potentials . the following result says that such weighted averages will exhibit asymptotic behavior with optimal growth , both globally and in sectors . [ naverage ] under the hypotheses of theorem [ ngeneric ] , assume that @xmath784 satisfies @xmath785 where @xmath786 denotes lebesgue measure on @xmath649 . then @xmath787 furthermore , for @xmath709 , with @xmath710 , @xmath788 this result is the analog of christiansen ( * ? ? ? the only major adjustment required in the proof is the replacement of @xmath688 by @xmath768 in the case where @xmath23 is odd . since this change was already discussed in the proofs of theorems [ order.thm ] , [ msector.thm ] , and [ ngeneric ] , and otherwise the details are thoroughly covered in @xcite , we will omit the proof . suppose @xmath789 \to { \mathbb{c}}$ ] is a smooth function with @xmath790 . if @xmath791 is smooth and non - vanishing at @xmath792 , then the classical laplace s method gives the asymptotic @xmath793 as @xmath794 . in this appendix we will extend this classical result to include rougher assumptions on @xmath299 and an explicit estimate of the error . this is fairly straightforward , but we include the details because the uniformity of the error estimate is crucial in our application . [ laplace.prop ] assume that @xmath789 \to { \mathbb{c}}$ ] is a smooth function with @xmath413 , and that @xmath795 $ ] is continuous near @xmath792 and satisfies @xmath796 for some @xmath797 . for @xmath798 we have @xmath799 as @xmath800 . more precisely , assume that for @xmath801 @xmath391 satisfies the bounds @xmath802}{\left|\phi '' \right| } \le \frac{1}{\beta}.\ ] ] then , given @xmath803 , there exists @xmath804 such that @xmath805 implies @xmath806 to estimate @xmath817 we use taylor s theorem to obtain @xmath818 for @xmath819 $ ] . in particular , for @xmath357 sufficiently large we have @xmath820 } { \left|e^{kh(x ) } - 1 \right| } \le c\beta^{-1 } k^{-\frac12}.\ ] ] applying this estimate in @xmath817 , and then replacing the upper limit in the integral by @xmath821 , we obtain the estimate @xmath822 for the @xmath823 term we can simply use the standard estimate on an incomplete gamma function , @xmath824 now consider the second integral , @xmath546 . given any @xmath626 , we will have @xmath825 for all @xmath120 sufficiently close to @xmath792 . hence , for @xmath826 , @xmath827 by the same analysis we used on @xmath479 we find @xmath828,\ ] ] for sufficiently large @xmath357 . the third integral is estimated for @xmath466 by @xmath829\\ & \le b { \left\vert u \right\vert}_\infty e^{2k { \operatorname{re}}\phi(b ) } e^{-2k^{1/4}{\operatorname{re}}\phi'(b ) } \exp\left [ k^{-\frac12}/\beta \right ] , \end{split}\ ] ] again using taylor and the assumed bound on @xmath830 . collecting these estimates , we find that for @xmath831 , @xmath832 using the lower bounds on @xmath833 and @xmath834 , this becomes @xmath835 we can adjust @xmath836 as necessary to make the right - hand side smaller than @xmath837 for @xmath838 . d. borthwick , t. christiansen , p. d. hislop , and p. a. perry , resonances for manifolds hyperbolic near infinity : optimal lower bounds on order of growth , _ int . res . not . imrn _ ( 2011 ) , no . 19 , 44314470 .

we study the asymptotic distribution of resonances for scattering by compactly supported potentials in @xmath0 . we first establish an upper bound for the resonance counting function that depends only on the dimension and the support of the potential . we then establish the sharpness of this estimate by proving the a weyl law for the resonance counting function holds in the case of radial potentials vanishing to some finite order at the edge of the support . as an application of the existence of potentials that saturate the upper bound , we derive additional resonance asymptotics that hold in a suitable generic sense . these generic results include asymptotics for the resonance count in sectors .

the numerical model that we use to study odd convection was described in detail in , so here we mention only its most relevant features . the computational domain is a cartesian box with uniform gravity @xmath0 . we solve the evolution equations for momentum , temperature , and composition in the boussinesq approximation , which are @xmath1 here @xmath2 , @xmath3 , @xmath4 , and @xmath5 are perturbations of pressure , density , temperature , and mean molecular weight respectively , and @xmath6 is the vertical component of the velocity @xmath7 ; all of these quantities are assumed to be periodic in all three cartesian directions . we use subscript 0 s to denote the unperturbed , background fields , which we assume are functions of @xmath8 alone . the background gradients of temperature , @xmath9 , and mean molecular weight , @xmath10 , are both negative in the situation considered here , as is the adiabatic temperature gradient , @xmath11 we consider a small volume within which the diffusivities of momentum , @xmath12 , temperature , @xmath13 , and composition , @xmath14 , can be treated as constant , and the density perturbations can be expressed as a linear function of the temperature and composition perturbations , & & & = - t , & [ eq : rho ] + & & = -()_p , & = ( ) _ p , t . & both @xmath15 and @xmath16 are positive for any realistic equation of state . we approximate @xmath15 , @xmath16 , and the other coefficients in equations ( [ eq : mom - dim])([eq : mu - dim ] ) as constant within our small domain , and we suppose that the background temperature and molecular weight profiles are linear in @xmath8 , so that the total temperature and composition fields are @xmath17 and @xmath18 , with @xmath9 , @xmath19 , and @xmath10 all constant . the schwarzschild and ledoux stability criteria are then & & ( t_0z - t_0z^ad ) & > 0 & [ eq : schwarz ] + & & ( t_0z - t_0z^ad ) & > _ 0z & [ eq : ledoux ] respectively . we are concerned here with regions which satisfy ( [ eq : ledoux ] ) but not ( [ eq : schwarz ] ) , i.e. , regions in which @xmath20 , where @xmath21 is the `` inverse density ratio , '' @xmath22 from a linear stability analysis of equations ( [ eq : mom - dim])([eq : mu - dim ] ) , it can be shown that such regions are unstable to double - diffusive instability if @xmath21 lies in the range @xmath23 @xcite where @xmath24 is the critical value @xmath25 here , @xmath26 and @xmath27 are the diffusivity ratios & & & = /_t & + & & & = _ /_t . & in astrophysical objects , both @xmath26 and @xmath27 are typically of order @xmath28 or smaller , and so the instability window given by equation ( [ eq : window ] ) is rather wide . the instability is oscillatory , and has a typical horizontal lengthscale @xmath29 ( e.g. , * ? ? ? we are interested primarily in the fluxes of heat and composition through odd convection . as in we measure these fluxes in terms of nusselt numbers , defined as & & _ t & = 1 - & [ eq : nut - dim ] + & & _ & = 1 - , & [ eq : nuc - dim ] where the notation @xmath30 represents a volume average over the entire domain . the total vertical fluxes of heat , @xmath31 , and composition , @xmath32 , can be reconstructed from these as follows : & & & = - _ tt_0z & + & & & = -_t_tt_0z + ( _ t-1)_tt_0z^ad & [ eq : flux_t ] + & & f_&= - _ _ 0z & + & & & = -___0z , & [ eq : flux_c ] where @xmath33 is the specific heat . we note that , in general , the thermal nusselt number @xmath34 defined by equation ( [ eq : nut - dim ] ) is _ not _ the ratio of the total heat flux to the diffusive ( i.e. , radiative plus conductive ) heat flux , unless the adiabatic temperature gradient @xmath19 is zero . however , as in , we can introduce the total `` potential temperature '' @xmath35 which has background gradient @xmath36 and perturbation @xmath37 , in order to write the thermal nusselt number as @xmath38 so we may regard @xmath34 as the ratio of the total and diffusive fluxes of potential temperature . since @xmath39 and @xmath4 are equal , in what follows we will refer to @xmath39 as simply `` the temperature perturbation '' . in double - diffusive convection , an important quantity is the `` buoyancy flux ratio '' @xmath40 , which we define as @xmath41 @xcite as the ratio of the _ total _ fluxes , rather than the turbulent fluxes . see for more detail . ] . this is closely related to the core erosion `` efficiency factor '' @xmath42 introduced by @xcite , which measures the fraction of a planet s luminosity that is used to erode the stable compositional stratification at the boundary with the core . in the limit where the transport of both heat and composition is dominated by convection ( i.e. , @xmath43 ) this fraction is precisely @xmath40 . the gravitational potential energy released from the thermal stratification must exceed the potential energy used to erode the compositional stratification , which implies that @xmath44 . as shown in , @xmath40 plays an important role in the formation of layers , and possibly also controls the evolution of layers after their formation @xcite . for computational convenience , we work mainly with nondimensional quantities . as in we nondimensionalize equations ( [ eq : mom - dim])([eq : rho ] ) using the lengthscale @xmath45 from equation ( [ eq : d ] ) as well as the scales & & [ t ] & = d^2/_t , & + & & [ ] & = d|t_0z - t_0z^ad| , & + & & [ ] & = d(/)|t_0z - t_0z^ad| , & + & & [ p ] & = _ 0 . & the governing equations are then @xmath46 where all variables are now dimensionless . the definitions ( [ eq : nut - dim])([eq : nuc - dim ] ) of the nusselt numbers then become & & _ t & = 1 + & [ eq : nut ] + & & _ & = 1 + . & [ eq : nuc ] we will also make use of alternative definitions of the nusselt numbers , & & _ t & = 1 + & [ eq : mut ] + & & _ & = 1 + , & [ eq : muc ] which are based on the rates of thermal and compositional dissipation , and are sometimes called `` cox numbers '' . we note that @xmath47 and @xmath48 are guaranteed to be positive at all times , whereas @xmath34 and @xmath49 are not . however , it can be readily verified , from equations ( [ eq : divu])([eq : mu ] ) , that the definitions ( [ eq : nut])([eq : nuc ] ) and ( [ eq : mut])([eq : muc ] ) yield the same result after taking a time average in a statistically steady state @xcite . the advantage of retaining both forms is explained in section [ sec : measure ] . in principle , any particular instance of odd convection is characterized by the three dimensionless parameters @xmath21 , @xmath26 , and @xmath27 . however , in it was shown that , within the instability window given by equation ( [ eq : window ] ) , there is another critical value of @xmath21 , say @xmath50 , below which odd convection spontaneously transitions into layered convection . presented empirical scaling laws for @xmath50 , and for the transport of heat and composition prior to the development of layers . it was shown that , in the parameter regime relevant to astrophysical objects , for which @xmath51 , the critical value @xmath52 . this result implies that there is a significant range of density ratios @xmath21 for which layers form spontaneously , and so the transport through layered convection must be considered in models of astrophysical objects . a preliminary study of layered convection by @xcite demonstrated that the turbulent transport is strongly dependent on the layer height , @xmath53 say , in addition to @xmath26 , @xmath27 , and @xmath21 . their results suggest that @xmath34 and @xmath49 both follow a power law in @xmath53 , with an exponent between @xmath54 and @xmath55 . in what follows we present a more comprehensive study of the dependence of the turbulent fluxes on each of the parameters @xmath53 , @xmath26 , @xmath27 , and @xmath21 . in order to provide a clear description of the transport properties of layered convection , and the subtleties involved in measuring the thermal and compositional fluxes in the layered phase , we first revisit one particular simulation that was previously presented in . this simulation has @xmath56 , @xmath57 , @xmath58 , and a domain size of @xmath59 . as described in , this simulation spontaneously transitions from homogeneous odd convection to layered convection , and the three layers that initially form subsequently merge into two layers , then into one layer . this evolution is illustrated in figure [ fig : evolution ] , which shows snapshots of the composition perturbation @xmath5 during each phase of the simulation , as well as profiles of the horizontally - averaged total `` potential density , '' @xmath60 the profiles show that @xmath61 is roughly constant within each layer , but increases slightly with height . the local ledoux stability criterion can be expressed as @xmath62 ( cf . equation ( [ eq : ledoux ] ) ) , so this indicates that the layers are subject to overturning convection . within each layer , the total composition @xmath63 and the total potential temperature @xmath64 are both roughly uniform , and there are strong jumps in @xmath63 and @xmath64 across the interfaces between adjacent layers . ( top row ) and potential density profiles @xmath65 ( bottom row ) at four different times in a simulation with @xmath56 , @xmath66 , and a domain size of @xmath59 . the first column shows the homogeneous phase of odd convection prior to layer formation . the remaining columns show layered convection with 3 , 2 , and 1 layers respectively . because the perturbation is plotted in the top row , rather than the total composition , the layers appear as regions with a relatively weak vertical gradient , and the interfaces appear as regions with a strong vertical gradient of the opposite sign . the color bar for each plot is scaled to the maximum of the perturbation @xmath5 , which increases as the layers merge . in the profiles , @xmath61 has been rescaled such that the total change in @xmath61 across the domain is unity.,width=604 ] spontaneous layer formation similar to that observed here has previously been reported in simulations of double - diffusive convection in the `` fingering '' regime , for which the thermal stratification is stabilizing and the compositional stratification is destabilizing ( e.g. , * ? ? ? * ) , but only for prandtl numbers @xmath67 @xcite . however , in fingering convection the initial height of the layers is typically much larger than the lengthscale of the linear double - diffusive instability , whereas here we find that layers initially form on a scale comparable to that of the linear instability . in the simulation shown in figure [ fig : evolution ] , for example , the fastest growing linear mode has a wavelength of roughly @xmath68 , i.e. , @xmath69 of the domain size , and the initial layer height is @xmath70 of the domain size . figure [ fig : interface ] shows more detail of the interface structure during the 2-layer phase . we find that the interfaces are highly dynamic and turbulent , exhibiting structure on scales that are much smaller than the wavelength of the fastest growing linear mode . this is in marked contrast to the flat , laminar interfaces found in simulations and laboratory experiments of odd convection at larger prandtl number ( e.g. , * ? ? ? * ; * ? ? ? * both of which have @xmath71 ) . the composition field exhibits structure on finer scales than the temperature field , as expected because of the smaller diffusivity of composition . on larger scales , the temperature and composition fields display similar features , such that the temperature field resembles a coarse - grained version of the composition field . by contrast , the density field has even finer structure , since it depends on the difference between the composition and temperature fields . , at @xmath72 . regions where the perturbations are small have been made transparent in order to show the interfaces in more detail.,width=604 ] the fluxes through this simulation , measured in terms of nusselt numbers @xmath73 and @xmath74 , are plotted in figure [ fig : nu ] as a function of time , along with the mean - squared velocity , @xmath75 . as mentioned earlier , the mean value of each flux increases significantly when layers first form , and increases still further when layers merge @xcite . the mean kinetic energy also increases , roughly in proportion with the fluxes , indicating that the simulation becomes more turbulent as the layers become thicker . ) ([eq : muc ] ) , and the mean kinetic energy in the simulation shown in figures [ fig : evolution ] and [ fig : interface ] . the initial linear instability grows exponentially until @xmath76 , then saturates , leading to homogenous odd convection . layers first form at @xmath77 ; the duration of the 3 layer , 2 layer , and 1 layer phases are indicated at the top of the plot . the vertical lines indicate the two times shown in figure [ fig : min_max ] . using @xmath74 rather than @xmath73 to measure the mean fluxes has the advantage of filtering out fluctuations in the fluxes on short timescales , while retaining the important long timescale information.,width=377 ] we note that the two definitions for the nusselt numbers both yield consistent results for the mean flux , although the definition based on the instantaneous fluxes exhibits larger fluctuations in the homogeneous , pre - layer phase of the simulation , during which the dynamics are dominated by wave - like , oscillatory motions @xcite . between layer merger events the mean fluxes and kinetic energy oscillate quasi - periodically between high and low values , with relative departures from the mean flux of order unity . the oscillations occur as the flow in the simulation cycles between states that are highly turbulent and states that are more quiescent . in the quiescent state , the interfaces are roughly horizontal , whereas in the turbulent state the interfaces have undulations on a horizontal scale comparable to the thickness of the layers . for illustration , figure [ fig : min_max ] shows the composition perturbation , and the horizontally - averaged total composition @xmath63 , at two times during the one - layer phase . at the first time , @xmath78 , the interface is in a quiescent state , and at the second , @xmath79 , the interface is in a more turbulent state . these two times correspond to minima and maxima in the nusselt numbers , as shown in figure [ fig : nu ] . the quasi - periodic nature of the oscillations in the fluxes suggests that they are caused by large - scale gravity - wave motions of the interfaces . for an interfacial gravity wave of horizontal scale @xmath53 , the oscillation frequency is @xmath80 @xcite , where @xmath81 represents the jump in potential density across the interface . for a staircase of layers of thickness @xmath53 we have @xmath82 , and so the characteristic gravity - wave frequency is @xmath83 where @xmath84 is the buoyancy frequency associated with the mean stratification , which in dimensionless units is @xmath85 . in the simulation shown in figure [ fig : nu ] this corresponds to a ( dimensionless ) oscillation period of @xmath86 , which is approximately half of the typical period of the observed flux oscillations . we emphasize , however , that the disturbances to the interfaces seen in figure [ fig : min_max ] are far more complex in structure than linear gravity waves . moreover , the large - scale motions of the interfaces are not strictly periodic in time they grow and decay cyclically , but they do not oscillate about zero . and @xmath87 during the one - layer phase . the profiles show the horizontal average of the total composition @xmath63 , and the volume rendering shows the perturbation @xmath5 using the same color bar at both times.,height=188 ] previous models of layered convection in astrophysical objects ( e.g. , * ? ? ? * ) have assumed that the interfaces are horizontal , laminar , and diffusive . if that were the case , the fluxes though the interfaces could be determined simply by measuring the vertical gradients of the horizontally - averaged @xmath88 and @xmath63 profiles in the center of each interface . figures [ fig : nu ] and [ fig : min_max ] show that this type of model can not be valid here , because the fluxes are maximal when these gradients are weakest . in fact , since the interfaces are not horizontal , the fluxes depend on the local gradients within the interfaces _ and _ on the effective surface area of the interfaces . figure [ fig : min_max ] suggests that variations in the surface area are the main cause of the oscillations in the fluxes . the interface behavior seen here somewhat resembles that seen in the compressible , two - dimensional simulations of @xcite at low prandtl number . @xcite also attributed the transport across the interfaces to the breaking of interfacial gravity waves , although the interpretation was complicated by interactions with the boundaries of the domain . the disturbances to the interface in figure [ fig : min_max ] have the same horizontal lengthscale as the width of the domain , @xmath89 , suggesting that , here also , the dynamics may be sensitive to the choice of domain size and the periodic boundary conditions . however , as described in appendix [ sec : robust ] , similar behavior is seen in all of our layered simulations , and the results do not depend significantly on the size of the domain . our aim is to determine how the mean fluxes of heat and composition , measured in terms of dimensionless nusselt numbers , depend on the dimensionless parameters @xmath90 , @xmath26 , @xmath27 , and the average ( dimensionless ) layer height , which we denote as @xmath53 . as described in the last section , during each layered phase the fluxes of heat and composition oscillate quasi - periodically about a mean value , and reliably calculating the mean fluxes requires time averaging over several of these oscillations . we measure the mean fluxes and estimate the errors in our measurements as follows : 1 . we first identify the number of layers during each phase of the simulation , by counting the number of `` steps '' in the vertical profiles of the horizontally averaged fields , and matching each merger event with the corresponding increase in the nusselt numbers . once we have identified the time interval corresponding to a particular layered phase , we then divide this interval into four equal subintervals , and calculate the average values of @xmath34 , @xmath49 , @xmath47 , and @xmath48 during each subinterval using equations ( [ eq : nut])([eq : muc ] ) . thus we obtain eight separate measurements of the thermal and compositional fluxes during each layered phase . we treat each of our eight measurements as independent , and calculate their mean and standard deviation , which yields estimates for the mean flux and the measurement error . ( a similar method was used to estimate the measurement error in . ) 4 . in the same fashion , we calculate the buoyancy flux ratio @xmath40 for each quarter interval using equation ( [ eq : gamma ] ) , and for both definitions of the nusselt numbers . as before , we use the resulting eight measurements to estimate the mean value of @xmath40 and our measurement error . we have applied this procedure to each of the simulations from in which layers formed spontaneously , as well as to many additional simulations that have since been performed . the parameters of each simulation , including those from that formed into layers , are listed in table [ tab : runs ] . cccccc _ 0.1 & _ 0.1 & _ 1.1 & _ 100@xmath91 & _ 128@xmath91 & _ 2 + 0.1 & 0.1 & 1.1 & 100@xmath91 & 64@xmath91 & 2 + 0.1 & 0.1 & 1.1 & 50@xmath91 & 64@xmath91 & 1 + _ 0.3 & _ 0.3 & _ 1.1 & _ 100@xmath91 & _ 128@xmath91 & _ 2 + 0.3 & 0.3 & 1.1 & 100@xmath91 & 64@xmath91 & 2 + _ 0.3 & _ 0.3 & _ 1.15 & _ 100@xmath91 & _ 64@xmath92@xmath93128 & _ 2 + 0.3 & 0.3 & 1.15 & 100@xmath91 & 64@xmath91 & 2 + 0.3 & 0.3 & 1.15 & 200@xmath92@xmath93100 & 128@xmath91 & 2 + 0.3 & 0.3 & 1.15 & 100@xmath92@xmath93200 & 64@xmath92@xmath93256 & 4 + 0.33 & 0.33 & 1.15 & 200@xmath92@xmath93400 & 128@xmath92@xmath93512 & 8 + 0.1 & 0.1 & 1.2 & 100@xmath91 & 64@xmath91 & 2 + 0.1 & 0.1 & 1.2 & 200@xmath92@xmath93100 & 128@xmath92@xmath9364 & 2 + 0.3 & 0.3 & 1.2 & 100@xmath91 & 32@xmath91 & 3 + _ 0.3 & _ 0.3 & _ 1.2 & _ 100@xmath91 & _ 64@xmath91 & _ 2 + 0.3 & 0.3 & 1.2 & 178@xmath91 & 128@xmath91 & 3 + 0.3 & 0.3 & 1.2 & 89@xmath92@xmath93178 & 64@xmath92@xmath93128 & 4 + 0.3 & 0.3 & 1.2 & 89@xmath92@xmath93178 & 32@xmath92@xmath9364 & 4 + 0.3 & 0.3 & 1.2 & 200@xmath92@xmath93100 & 128@xmath91 & 2 + _ 0.1 & _ 0.1 & _ 1.25 & _ 100@xmath91 & _ 128@xmath91 & _ 2 + 0.1 & 0.1 & 1.25 & 100@xmath91 & 64@xmath91 & 2 + 0.1 & 0.1 & 1.25 & 50@xmath92@xmath93100 & 64@xmath92@xmath93128 & 3 + _ 0.3 & _ 0.3 & _ 1.25 & _ 100@xmath91 & _ 64@xmath91 & _ 2 + _ 0.01 & _ 0.01 & _ 1.5 & _ 100@xmath91 & _ 192@xmath91 & _ 2 + _ 0.03 & _ 0.03 & _ 1.5 & _ 100@xmath91 & _ 192@xmath92@xmath93256 & _ 3 + 0.03 & 0.03 & 1.5 & 50@xmath91 & 96@xmath91 & 2 + 0.03 & 0.03 & 1.5 & 50@xmath91 & 64@xmath91 & 2 + 0.03 & 0.03 & 1.5 & 50@xmath91 & 48@xmath91 & 2 + 0.03 & 0.03 & 1.5 & 50@xmath91 & 128@xmath91 & 2 + _ 0.1 & _ 0.1 & _ 1.5 & _ 100@xmath91 & _ 64@xmath91 & _ 2 + _ 0.01 & _ 0.01 & _ 2.0 & _ 100@xmath91 & _ 192@xmath91 & _ 3 + _ 0.3 & _ 0.03 & _ 1.1 & _ 100@xmath91 & _ 192@xmath91 & _ 3 + _ 0.3 & _ 0.1 & _ 1.1 & _ 100@xmath91 & _ 128@xmath91 & _ 2 + 0.3 & _ 0.1 & _ 1.1 & _ 100@xmath91 & _ 64@xmath91 & _ 2 + _ 0.03 & _ 0.3 & _ 1.1 & _ 100@xmath91 & _ 192@xmath91 & _ 1 + _ 0.1 & _ 0.3 & _ 1.1 & _ 100@xmath91 & _ 64@xmath91 & _ 2 + _ 0.3 & _ 0.1 & _ 1.2 & _ 100@xmath91 & _ 80@xmath91 & _ 2 + _ 0.03 & _ 0.3 & _ 1.2 & _ 100@xmath91 & _ 128@xmath91 & _ 2 + _ 0.1 & _ 0.3 & _ 1.2 & _ 100@xmath91 & _ 80@xmath91 & _ 2 + _ 0.3 & _ 0.03 & _ 1.25 & _ 100@xmath91 & _ 128@xmath91 & _ 4 + _ 0.3 & _ 0.1 & _ 1.4 & _ 100@xmath91 & _ 64@xmath91 & _ 3 + _ 0.3 & _ 0.03 & _ 1.5 & _ 100@xmath91 & _ 128@xmath91 & _ 3 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ tab : runs ] we present below the results of these measurements and describe their dependence on the parameters @xmath26 , @xmath27 , @xmath21 , and @xmath53 . in principle these four parameters are sufficient to characterize any particular regime of layered convection . however , in our numerical simulations we also specify the size of the computational domain and the numerical resolution , and so we must check whether these additional parameters have any effect on the results . appendix [ sec : robust ] describes some of the tests that we have performed to ensure that our measurements are robust . for reasons explained in the appendix , only the flux measurements from simulations performed in domains of width @xmath94 are presented below . we also note that the spontaneous emergence of layers described in is not the only mechanism by which layers can develop in odd convection . laboratory experiments and observations of earth s oceans indicate that the formation of layers can be `` triggered '' by external influences @xcite . layered convection might even be metastable under conditions for which the linear odd instability is absent @xcite . in this paper , however , we only consider the regime in which layers form spontaneously , @xmath95 , in order to guarantee the reproducibility of our results . to facilitate comparison of our results with those of previous studies , we introduce the thermal rayleigh number for layered convection , which is @xmath96 where @xmath97 is the ( dimensional ) layer height . owing to computational limitations , our results cover only a limited range of @xmath26 and @xmath27 values . moreover , by choosing to analyze only those simulations in which layers form spontaneously , we are restricted to a rather narrow range for @xmath21 . on the other hand , the range of @xmath98 values that can be achieved in each simulation is limited only by the height of the computational domain . moreover , the fact that layers merge allows us to test the dependence of the nusselt numbers on @xmath98 using data from a single simulation . the tallest simulation presented here has a domain of height @xmath99 , and initially formed eight layers . figure [ fig : nu_tall_box ] shows the nusselt numbers in this simulation , plotted as functions of time and @xmath98 . , @xmath100 , and domain size @xmath101 . the number of layers in each layered phase is indicated on the plot . the horizontal lines indicate the approximate duration of each layered phase and the mean flux determined by the method described in section [ sec : measure ] . b ) dependence of @xmath34 on @xmath98 in the same simulation . the errors in the measurements were determined by the method described in section [ sec : measure ] . the solid and dashed lines both have a slope of @xmath70 . , width=642 ] we find that both @xmath34 and @xmath49 are well approximated by a power law in @xmath98 with an exponent of @xmath70 . this result can be explained physically if we suppose that the mean flux depends only on the jumps in potential temperature and composition across the interfaces , @xmath102 and @xmath103 , say , and not on any properties of the layers . more precisely , let us suppose that the ( dimensional ) buoyancy fluxes @xmath104 and @xmath105 depend only on the buoyancy jumps @xmath106 and @xmath107 and the diffusivities @xmath13 , @xmath14 , and @xmath12 . on dimensional grounds , the fluxes must then be proportional to @xmath108 @xcite . in our simulations , the _ total _ change in temperature between the top and bottom of the domain is fixed , and so the jump across each interface is inversely proportional to the number of layers , and therefore proportional to the height of the layers . the exponent @xmath70 then follows from equation ( [ eq : rat ] ) . we now ask whether _ all _ of our nusselt number measurements can be fitted to the same power law . in order to fit all of the data simultaneously , we must make some assumption regarding the dependence not only on @xmath98 , but also on @xmath26 , @xmath27 , and @xmath21 . we consider the thermal and compositional nusselt numbers in turn , and then the flux ratio @xmath40 . many analytical and heuristic arguments have been made that the turbulent transport of heat through layered convection follows a power law in both the rayleigh number , @xmath98 , and the prandtl number , @xmath26 ( e.g. , * ? ? ? * ; * ? ? ? but its dependence on the parameters @xmath21 and @xmath27 is less clear . as a first step toward fitting our data to an analytical formula for @xmath34 , we therefore assume a dependence of the form @xmath109 where the exponents @xmath110 and @xmath111 , and the function @xmath112 , are to be determined . we group all of our @xmath34 measurements according to their values of @xmath21 and @xmath27 , which leads to 15 distinct groups : @xmath113 , @xmath114 , @xmath115 , @xmath116 , @xmath117 , @xmath118 , @xmath119 , @xmath120 , @xmath121 , @xmath122 , @xmath123 , @xmath124 , @xmath125 , @xmath126 . we then look for a fit of the form @xmath127 where @xmath128 represents the index of each group . to determine the coefficients @xmath110 , @xmath111 , and @xmath129 we perform a weighted least - squares minimization with weights proportional to the inverse square of the estimated error in the measurements of @xmath34 . the best fit to the data has exponents @xmath130 and @xmath131 . the best fit values for the coefficients @xmath129 are shown in figure [ fig : nut_fit ] , along with a comparison between the data and the fit . defined in equation ( [ eq : nut_guess2 ] ) . right panel : comparison between the measured values of @xmath132 and the best fit for the formula in equation ( [ eq : nut_guess2 ] ) with exponents @xmath133 and @xmath134 . in the online version of this figure , data from different parameter pairs @xmath135 are plotted in different colors . ] most of the data points lie within one standard error of the best fit line , although there are significant outliers for both low and high values of @xmath34 . the lowest values of @xmath34 are found in simulations that are only weakly turbulent , for which we would not expect a simple scaling law to hold . the presence outliers at the highest values of @xmath34 , on the other hand , may result from insufficient time - averaging or numerical resolution in the simulations with the tallest layers . the fact that the best fit has exponents @xmath136 suggests that the heat flux depends only on the parameters @xmath21 , @xmath27 , and the product @xmath137 . in that case the heat flux is independent of viscosity , as predicted by @xcite for layered convection with @xmath138 and @xmath139 . however , the model of @xcite predicts exponents @xmath140 , rather than @xmath70 , which is incompatible with our results . using the result that @xmath141 we can directly compare simulations performed with different values of @xmath26 but the same values of @xmath27 and @xmath90 . an example is given in figure [ fig : depend_nu ] , which overlays time series of @xmath34 in three simulations that each have @xmath142 , @xmath143 , and domain size @xmath59 . for these parameters , the best fit prediction is @xmath144 . since each of these simulations has the same ( dimensionless ) domain height , we can directly compare their two - layer ( @xmath145 ) and one - layer ( @xmath146 ) phases ; we find that , as predicted , @xmath147 has the same dependence on @xmath53 in each of these simulations . , in three simulations with @xmath142 , @xmath143 , and domain height @xmath89 , but different values of @xmath26 . the dashed lines indicate the flux predicted by the best fit for the two - layer and one - layer phases , @xmath148.,height=302 ] the best - fit values for the coefficients @xmath129 , presented in figure [ fig : nut_fit ] , show a weak but well - defined dependence on @xmath21 and @xmath27 ( for fixed @xmath98 and @xmath26 ) . in particular , we find that decreasing either the composition gradient @xmath10 or the compositional diffusivity @xmath14 produces a slight increase in the mean heat flux . we interpret this result as evidence that changing @xmath10 and @xmath14 affects the heat flux only indirectly , by changing the overall level of turbulence . we apply the same analysis for the compositional flux , i.e. , we suppose that @xmath149 and fit the data to determine @xmath110 , @xmath111 , and @xmath129 . this time the best - fit values for the exponents are @xmath150 and @xmath151 . this is roughly consistent with the trend observed in figure [ fig : nu_tall_box]b , but indicates that the compositional flux , unlike the heat flux , depends on viscosity even for @xmath152 . this is perhaps not surprising given that the composition field is more sensitive than the temperature field to the small - scale motions of the fluid ( e.g. , see figure [ fig : interface ] ) . figure [ fig : nuc_fit ] shows the best fit values for the coefficients @xmath129 and a comparison between the measured values of @xmath49 and the best fit for equation ( [ eq : nuc_guess2 ] ) . as found previously for @xmath34 , there is a weak increase in @xmath49 as @xmath21 is made smaller , i.e. , as the composition gradient is reduced . unlike @xmath34 , however , @xmath49 depends strongly on @xmath27 , i.e. , on the compositional diffusivity . in fact , we find that @xmath49 is roughly inversely proportional to @xmath27 , indicating that the mean turbulent flux of composition @xmath153 is approximately independent of compositional diffusivity in the parameter regime considered here . defined in equation ( [ eq : nuc_guess2 ] ) . the color bar has been rescaled to accommodate the larger range of values here . right panel : comparison between the measured values of @xmath154 and the best fit formula in equation ( [ eq : nuc_guess2 ] ) with exponents @xmath155 and @xmath156 . ] taken at face value , the empirical power laws for @xmath34 and @xmath49 derived above imply that the flux ratio @xmath157 . however , as discussed in section [ sec : model ] , @xmath40 can not exceed unity in a statistically steady state , and therefore can not follow a simple power law in @xmath98 . as illustrated in figure [ fig : gamma ] , we find that @xmath40 is generally an increasing function of @xmath98 , and possibly asymptotes to a constant value at large values of @xmath98 . further simulations at high @xmath98 are required to confirm this trend , and to determine how this asymptotic value depends on the other parameters @xmath21 , @xmath26 , and @xmath27 . figure [ fig : gamma ] suggests that @xmath40 is a decreasing function of @xmath21 , but again , more data is required to confirm this suggestion . if @xmath40 remains a monotonically increasing function of @xmath98 for large @xmath98 then there are important implications for the merging of layers , as we discuss in the next section . on @xmath98 for all simulations with @xmath158 , and for the large - domain simulation shown in figure [ fig : nu_tall_box ] , which has @xmath159 . different colors and symbols indicate different values of @xmath21.,width=377 ] we have measured the mean fluxes of heat and composition in an extensive suite of numerical simulations of layered odd convection . in these simulations , layers develop spontaneously from an initially homogeneously turbulent state , and subsequently merge into a single layer spanning the entire height of the simulation domain . the formation and merging of layers is , to some extent , a stochastic process , but nevertheless occurs at a similar rate in simulations with different domain sizes and numerical resolutions . the development of layers leads to a significant increase in the mean fluxes , and the fluxes increase still further each time layers merge . we find that the mean fluxes are approximately proportional to the average thickness of the layers to the power @xmath55 , as predicted by @xcite . the heat flux is independent of viscosity , but the composition flux is not . within the range of parameter values considered here , the thermal and compositional nusselt numbers are well approximated by & & _ t - 1 & = a^1/3_t^1/3 & [ eq : empirical_nut ] + & & _ - 1 & = b^-1 ^ 1/4_t^1/3 & [ eq : empirical_nuc ] where the prefactors @xmath160 and @xmath161 are weakly dependent on the parameters @xmath21 , @xmath27 , @xmath26 , and @xmath98 , and have typical values @xmath162 and @xmath163 respectively . the mean heat and composition flux can be calculated from these expressions using equations ( [ eq : flux_t ] ) and ( [ eq : flux_c ] ) . we emphasize , however , that the parameter values used in our simulations are rather far from true astrophysical values , and so any application of the formulae ( [ eq : empirical_nut])([eq : empirical_nuc ] ) to astrophysical conditions requires an extrapolation beyond the range in which these formulae have been derived . in all our simulations the interfaces between the convective layers exhibit complex dynamical behavior that is very different from the laminar `` interface scouring '' and entrainment seen in studies of layered convection at higher prandtl number ( e.g. , * ? ? ? the structure of the interfaces , and their effective surface area , is highly time dependent , which produces order unity variations in the fluxes of heat and composition during each layered phase . previous models for the transport by odd convection have assumed that the interfaces remain flat and laminar , implying that the fluxes are limited by the vertical gradients of the horizontally averaged profiles within the interfaces . our numerical results demonstrate that transport at low prandtl number is more efficient than predicted by such quasi - one - dimensional models . we argue that the interface dynamics must be considered in future models for transport under astrophysical parameter conditions . since the transport by layered convection is strongly dependent on the layer thickness , we must consider what determines this thickness in stellar and planetary interiors . recently , @xcite have proposed a model for the transport by odd convection in jupiter s interior . they use heuristic arguments to obtain upper and lower bounds for the height of layers in jupiter , and thereby derive bounds for the thermal and compositional transport . in all of our simulations , however , the layers eventually merge into a single layer with the same height as the computational domain , so our results make no prediction for the height of layers in real objects . whether simulations performed in taller domains would yield a saturation thickness for layers is unknown . for _ fingering _ convection , it has been argued by @xcite that there exists such a saturation thickness . the essence of his argument is that the buoyancy flux ratio @xmath40 increases every time layers merge , but that there exists a maximum permitted value of @xmath40 , which is attained at a finite layer thickness @xmath164 , say . layers of thickness @xmath165 therefore do not undergo further merging . in our simulations of _ odd _ convection , by contrast , @xmath40 appears to asymptote toward a limiting value as @xmath166 , so this argument does not apply . ( this is in contrast to a statement made by * ? ? ? * ) . an alternative picture for layer merging has been proposed by @xcite , but their model also predicts that the merging of layers proceeds indefinitely . we note that the boussinesq approximation assumed in our numerical model breaks down once the layer thickness becomes comparable to the local pressure scale height of the planet or star . in practice , this may be what limits the size of layers in real astrophysical objects . at this point , convection within the layers would be almost indistinguishable from `` normal '' overturning convection , allowing the usual mixing - length approximation for the transport to be applied . an important issue that has not been explicitly addressed so far is the timescale for layer merging . our results suggest that this timescale is an increasing function of layer height , which is in accordance with the theory of layer mergers proposed by @xcite . if the timescale for merging in an astrophysical object eventually becomes comparable to the global evolution timescale , then the layers will never reach a saturation thickness , and the evolution of the layers must then be modeled as an explicitly time - dependent process in 1d structure and evolution models . future papers will study the dynamics of layer mergers in more detail , in order to determine what sets the merging timescale , and how it depends on layer height . we thank b. r. ruddick , e. a. spiegel , and m. g. wells for useful comments and suggestions . g. and t. w. were supported by funding from the nsf ( nsf-0933759 ) . part of the computations were performed on the ucsc pleiades supercomputer , purchased with an nsf - mri grant . others used computer resources at the national energy research scientific computing center ( nersc ) , which is supported by the office of science of the us department of energy under contract de - ac03 - 76sf00098 . in this appendix we investigate the influence of numerical resolution and domain size on the initial layer height , the timescale for layer formation and merging , the time - averaged fluxes , and the amplitude of oscillations in the fluxes . because of the disparity between the diffusivities of temperature and composition , double - diffusive convection always features a wide range of dynamical length scales , making this a tough problem to model numerically . furthermore , as mentioned in section [ sec : typical ] , the emergence of layers is associated with an increase in the overall level of turbulence , and increased spectral energy at all scales in a given simulation @xcite . properly resolving the layered phase therefore requires even greater numerical resolution than resolving the homogenous phase . in order to check that the thermal and compositional fluxes in our simulations are not sensitive to the numerical resolution , we have run several simulations that have the same physical parameters and domain size , but different numbers of grid points in the vertical and horizontal directions . as an example , figure [ fig : res_check ] compares time series of @xmath73 and @xmath74 from four simulations performed with different numerical resolutions . all four were performed in domains of size @xmath167 with the same physical parameters as the simulation presented in section [ sec : typical ] , but with @xmath168 , @xmath169 , @xmath170 , and @xmath171 fourier modes respectively . the simulation with the coarsest resolution ( @xmath168 ) only resolves structures on lengthscales @xmath172 , but even in this simulation the linear double - diffusive instability is well resolved ( recall that the fastest growing linear mode has wavelength @xmath173 at these parameters ) . each simulation undergoes a homogeneous phase , then forms three layers , which later merge into two layers , then one layer . the duration of each layered phase varies between the four simulations , suggesting that the merging of layers is , to some extent , a stochastic process . however , the four simulations produce similar fluxes during each phase , and the variations in each flux about its mean are comparable in all four cases . , @xmath169 , @xmath170 , and @xmath171 fourier modes in each spatial direction . all four cases have @xmath174 , @xmath175 , and a domain size of @xmath167 . the two rows show the nusselt numbers calculated from the instantaneous fluxes ( [ eq : nut ] ) and ( [ eq : nuc ] ) , and the instantaneous dissipation rates ( [ eq : mut ] ) and ( [ eq : muc]).,width=453 ] in each simulation , the most stringent test of the numerical resolution occurs during the final merging event , which marks the start of the one - layer phase . figure [ fig : high_low ] compares the density fields at this point in the lowest and highest resolution simulations shown in figure [ fig : res_check ] . the lowest resolution simulation is clearly under - resolved , with significant features at the nyquist scale , whereas the highest resolution simulation displays no numerical artifacts . from the @xmath168 resolution simulation at @xmath176 ( left ) and the @xmath171 resolution simulation at @xmath177 ( both cases show cross - sections through the interface at the start of the one - layer phase , which is typically the least well - resolved stage in any simulation.,width=604 ] nevertheless , the mean fluxes through these two simulations are approximately equal in the one - layer phase , which suggests that the transport of both heat and composition is dominated by scales @xmath172 in these simulations . figure [ fig:4cases ] compares time series of the nusselt numbers from four simulations performed in domains of different heights and widths . each simulation has @xmath158 , @xmath100 , and the same numerical resolution , i.e. , the same ( dimensionless ) distance between neighboring grid points . these four simulations exhibit the following features , which we find to be typical : * the cases with wider domains ( @xmath178 ) take longer to form layers than the cases with narrower domains ( @xmath89 ) . * the initial layer height is not very sensitive to the size of the domain , provided that the domain is sufficiently wide . in figure [ fig:4cases ] , for example , the initial layer height is @xmath179 in all four cases . however , in other simulations ( not shown here , but see @xcite for instance ) it has been found that in very narrow domains layers form much earlier , and the initial layer height is smaller . this result indicates that the horizontally - periodic boundary conditions , which encourage horizontal coherence in the simulation , can artificially accelerate the formation of layers when the domain is sufficiently narrow . for this reason , we only use data from simulations that were performed in domains of width @xmath94 when calculating the mean fluxes in section [ sec : measure ] . * the mean fluxes of heat and composition , for layers of a particular height , are consistent between runs with different domain sizes . * during each layered phase , the cases with wider domains exhibit smaller variations in the value of each flux about its mean value . this suggests that the variations in the fluxes are horizontally localized , and therefore have less effect on the mean flux in wider domains . similarly , the simulation in the tallest domain ( of height @xmath99 ) exhibits the smallest variations in the fluxes during each layered phase , because in this simulation the volumetric mean flux represents an average over a greater number of layers and interfaces . figure [ fig:4cases ] also illustrates a practical difficulty that arises when comparing results from simulations performed in domains of different heights , which is that simulations in taller domains usually undergo more layered phases than simulations in shorter domains , and the length of each layered phase is correspondingly shorter . for example , the simulation with a domain height of @xmath99 passes through phases in which the average layer height is @xmath180 and @xmath181 ; these states can not be achieved in the simulations with a domain height of @xmath89 . as a result , the time series of the nusselt numbers in the taller simulation are less obviously step - like , and the mean flux during each layered phase is more difficult to measure accurately . this also makes it difficult to determine whether the domain height affects the rate at which layers merge . the full time series of both nusselt numbers from this simulation are shown in figure [ fig : nu_tall_box]a . we find that the time between layer merger events increases as the number of layers decreases , which is in accordance with the model of layer merging proposed by @xcite . another difficulty associated with measuring fluxes in simulations with multiple layers is that the layers are not necessarily all of the same height . in that case , a complete description of the state of the system would require ( at least ) measuring the height of each layer , and the jumps in temperature and composition across each interface , potentially introducing many new parameters on which the mean flux might depend . furthermore , in such cases we would not expect the system to be in statistical equilibrium , and so the mean flux might not be well defined . to avoid dealing with this additional complexity , in this study we assume that , in each layered phase , all the layers have approximately the same height , while recognizing that this assumption introduces an additional source of error into our flux measurements .

regions of stellar and planetary interiors that are unstable according to the schwarzschild criterion , but stable according to the ledoux criterion , are subject to a form of oscillatory double - diffusive ( odd ) convection often called `` semi - convection '' . in this series of papers , we use an extensive suite of three - dimensional ( 3d ) numerical simulations to quantify the transport of heat and composition by odd convection , and ultimately propose a new 1d prescription that can be used in stellar and planetary structure and evolution models . the first paper in this series demonstrated that under certain conditions odd convection spontaneously transitions from an initially homogeneously turbulent state into a staircase of convective layers , which results in a substantial increase in the transport of heat and composition . here , we present simulations of odd convection in this layered regime , we describe the dynamical behavior of the layers , and we derive empirical scaling laws for the transport through layered convection . [ [ sec : intro ] ] a long - standing problem in modeling stellar and planetary interiors is how to quantify the transport of heat and composition in regions that have a destabilizing ( i.e. , superadiabatic ) thermal stratification and a stabilizing compositional stratification , such that the total density stratification is stable according to ledoux s criterion @xcite . such regions are present , for instance , in giant planets @xcite , and in massive stars both main - sequence @xcite and post - main - sequence @xcite . it has long been recognized that , because the molecular diffusivity of composition is typically only a tiny fraction of the total ( radiative plus conductive ) diffusivity of temperature , these regions are subject to a form of double - diffusive instability @xcite . the turbulent convection that arises from this instability , which is often called `` diffusive '' or `` oscillatory '' convection , enhances the transport of heat and composition , changing the internal structure of the host object . oscillatory double - diffusive convection is closely related to the phenomenon known as `` semi - convection , '' but the latter is often also associated with the effects of convective overshoot and the dependence of thermal diffusivity on composition , neither of which are included in this study . instead , we consider a localized region away from any radiative convective boundary , in which the thermal and compositional diffusivities can be approximated as constant , and we study the instability and turbulence arising from the competition between thermal and compositional stratification . following @xcite we refer to convection in this situation as `` oscillatory double - diffusive ( odd ) convection . '' the linear instability mechanism that leads to odd convection can be understood from physical arguments . in the absence of thermal and compositional diffusion , vertically displaced parcels of fluid oscillate adiabatically about their equilibrium position with a frequency determined by the total density stratification . but the presence of diffusive processes can amplify the oscillation , producing an instability that resembles an overstable internal gravity wave @xcite . indeed , if the thermal diffusivity is sufficiently large in comparison with the compositional diffusivity , then a displaced parcel will acquire heat ( and therefore entropy ) while below its equilibrium position , and thus will be slightly more buoyant than its surroundings when it returns to its equilibrium position . the amplitude of the oscillation grows exponentially until nonlinear effects are sufficient to saturate the instability , leading to a state of weakly turbulent odd convection . after saturation , odd convection initially takes the form of a homogenously turbulent state , in which the time - averaged stratification remains uniform . previous work by @xcite and @xcite has shown that the turbulent transport of heat and composition in this homogeneous state is fairly weak . ( * ? ? ? * hereafter ) derived empirical formulae for the respective fluxes from an extensive suite of numerical simulations performed in local cartesian domains . however , in certain cases the homogeneous state was found to spontaneously transition into a new state of layered convection , for which the transport of heat and composition is greatly enhanced . derived a simple semi - analytical criterion that predicts under which conditions spontaneous layering occurs . under the physical conditions relevant to planetary and stellar interiors , spontaneous layering is expected within a significant fraction of the double - diffusive instability range , specifically in regions that are closer to being ledoux - unstable ( see figure 1 in ) . in this paper , we present further results from numerical simulations and use them to derive empirical formulae for the fluxes of heat and composition in layered convection . we determine how the fluxes depend on the thickness of the layers and on other governing parameters . we also describe some of the physical properties of layered convection , paying particular attention to the interfaces between the layers , and we compare our results with previous studies of odd convection . finally , we discuss our results in the light of recent layered models of giant planets .

various galaxy surveys in the optical , near - infrared , and sub - mm have uncovered different galaxy populations at redshift @xmath1 ( see steidel , cimatti , and chapman in these proceedings for discussions ) , but whether these galaxies are representative of the galaxy population at high redshifts or how they are related to the present - day population is not clear . for example , rest - frame uv - selected samples are presumably sensitive to active star - forming galaxies and biased against evolved , quiescent systems that are faint at ultraviolet wavelengths , while sub - mm samples select mostly dusty star - forming galaxies . in contrast , near - infrared based surveys are sensitive to early - type galaxies at @xmath2 that have red optical and near - infrared colors and are known to dominate the total stellar mass at @xmath3 . a complete sample of massive systems at different epoch allows us to study the history of stellar mass assembly and offers important clues for discriminating between different galaxy formation scenarios ( e.g. @xcite ) . but past studies based on near - infrared surveys have yielded inconsistent space density measurements of red galaxies at @xmath4 ( see @xcite for a list of references ) . early - type galaxies are strongly clustered and have no prominent narrow - band spectral features at uv wavelengths . if these red galaxies are progenitors of early - type galaxies we see in the local universe ( rather than dusty star - forming galaxies ) , then a wide - field infrared survey is required to minimize the effect of surface density variation between fields due to strong clustering ( e.g.@xcite ) . furthermore , deep galaxy spectroscopy is required to obtain a complete sample of these massive systems for spectral diagnostics of their ism and stellar content . the las campanas infrared survey ( lcirs ) and gemini deep deep survey ( gdds ) are complementary studies designed to probe the nature of red galaxies and to determine mass assembly history using near - infrared selected galaxies . in the following section i summarize the goals and results of the surveys . galaxies with @xmath5 in different survey areas @xmath6 . the surface densities were determined empirically within separate sub - areas randomly selected from the lcirs fields . the left panel shows the cumulative probability of different surface density measurements for @xmath7 . the vertical dotted line indicates the mean surface density of 0.83 arcmin@xmath8 determined over the entire lcirs @xmath9-band selected red galaxies . based on these curves , we show in the right panel the probability of obtaining a measurement within a factor of two accuracy of the nominal value versus @xmath6.,scaledwidth=95.0% ] the lcirs is a deep , wide - field near - infrared and optical imaging survey , designed to identify a large number of red galaxies at @xmath0 , while securing a uniform sample of galaxies of all types to @xmath10 using photometric redshift techniques ( @xcite ) . the primary objectives are : ( 1 ) to examine the nature of the red galaxy population and identify evolved galaxies at redshifts @xmath11 ; ( 2 ) to study the space density and luminosity evolution of early - type galaxies at redshifts @xmath12 ; and ( 3 ) to measure spatial clustering of massive galaxies , thereby inferring merger rates of these galaxies for constraining theoretical models . ( left ) and @xmath9 ( right ) galaxy luminosity function of @xmath13 galaxies identified at @xmath14 in the lcirs using photometric redshift techniques . the points were evaluated using a modified stepwise maximum likelihood analysis , explicitly incorporating redshift uncertainties from photometric redshift analysis @xcite . error bars were estimated using a bootstrap technique that takes into account photometric uncertainties , redshift errors , and sampling biases . the thick solid curve in each panel shows our best - fit schechter luminosity function . for comparison , we present measurements for @xmath9-selected galaxies at @xmath15 from the k20 survey team @xcite ( dashed curves ) . we also plot local measurements from the 2mass survey @xcite ( thin curves).,scaledwidth=100.0% ] we have completed the phase - i @xmath16-band survey to @xmath17 over 1.1 @xmath18 and are completing the phase - ii @xmath9-band survey to @xmath19 over 0.75 @xmath20 . the size of our survey area was determined in order to obtain a representative measure of the surface density of red galaxies and a significant signal in clustering analysis @xcite . five random fields have been chosen and each field covers @xmath21 , corresponding to a projected co - moving distance 20 @xmath22 mpc . this is twice the correlation length observed for early - type galaxies at @xmath3 ( see @xcite for references ) . the large survey area of the lcirs allows us to estimate empirically the uncertainties in the surface density measurements of red galaxies identified in smaller - area surveys , assuming that the lcirs galaxy sample are representative of these systems at @xmath4 . figure 1 shows the results based on red galaxies with @xmath23 found in non - overlapping , random sub - areas within the lcirs fields , in comparison to 0.83 arcmin@xmath8 as the nominal surface density of @xmath23 galaxies determined based on the entire lcirs red sample . we find based on the curves that a survey area of at least 70 arcmin@xmath24 is necessary to identify a statistically representative sample of red galaxies . our initial clustering analysis based on an @xmath16-band selected sample over 0.62 deg@xmath24 of sky shows that red galaxies with @xmath25 have a co - moving correlation length of @xmath26 mpc at @xmath27 , comparable to that of present - day early - type galaxies @xcite . photometric redshifts have been determined for the entire lcirs galaxy sample based on a combination of available optical @xmath28 and near - infrared @xmath29 photometry @xcite , making the lcirs a _ complete _ redshift survey of faint galaxies at @xmath4 . but because of relatively large redshift uncertainties , accounting for these errors in all analyses that make use of photometric redshift measurements is necessary to minimize systematic errors . adopting an @xmath16-band selected sample and using a technique that explicitly accounts for the _ non - gaussian characteristics _ of photometric redshift uncertainties , we have shown that the evolution of the rest - frame @xmath30-band co - moving luminosity density @xmath31 is characterized by @xmath32 and that luminous early - type galaxies exhibit only moderate evolution over @xmath33 @xcite . rest - frame @xmath34- and @xmath9-band luminosity functions have also been calculated using @xmath13 @xmath9-band selected galaxies at @xmath35 . figure 2 shows the best - fit luminosity functions in @xmath34 ( left panel ) and @xmath9 ( right panel ) using modified stepwise maximum likelihood method and sty approach as described in chen et al . the best - fit schechter parameters are @xmath36 , and @xmath37 . figure 2 also shows that our results are consistent at the bright end with the estimates presented by the k20 team for 170 galaxies found at @xmath15 over a total sky area of 52 arcmin@xmath24 @xcite ( dashed curves ) , but exhibit a hint of a turn - over at the faint - end . comparisons with existing measurements for local galaxies using 2mass data @xcite indicate that there is approximately @xmath38 mag fading in luminous ( @xmath39 ) near - infrared selected galaxies from @xmath40 to @xmath3 , the amount of which may be explained by stellar evolution over the same period of time . these results together suggest little / no evolution in the space density of massive early - type galaxies since @xmath41 and support the hypothesis that these systems were formed at redshift much beyond @xmath41 . the gdds is an ultra - deep spectroscopic survey of near - infrared selected galaxies at @xmath42 @xcite . the primary objectives are : ( 1 ) to obtain spectral diagnostics of the ism and stellar content of massive galaxies at high redshift ; ( 2 ) to construct the stellar mass function over the target redshift range ; and ( 3 ) to determine ages and star - formation histories of evolved systems at @xmath11 . the survey makes use of nod & shuffle observing mode @xcite for accurate sky subtraction and fringe removal at wavelength beyond 7500 . it is aimed at identifying near - infrared selected faint galaxies at @xmath43 to a high completeness level . we have observed 301 spectra in four fields using gmos on gemini north . each of the four fields covers @xmath44 field of view and is located in a separate lcirs field . the pointing and target selection of the gdds sample is guided by known information in the wider - field lcirs sample to maximize the efficiency of identifying a representative sample of faint galaxies at @xmath43 . for example , each gdds pointing is selected from within an lcirs field to have the number of red galaxies comparable to the global mean , which is important for reliable volume density analyses of early - type galaxies . in addition , we exclude foreground contaiminating sources at @xmath45 using photometric redshifts available for all lcirs sources . figure 3(a ) shows a comparison of the observed @xmath46 color versus @xmath47 for cfrs @xcite , gdds @xcite , and uv - selected galaxies @xcite . it demonstrates that the gdds is probing a representative galaxy population concentrated at @xmath48 with a broad - range of star formation history , from quiescent through active star - forming systems . conversely , the gdds results offer necessary tests and verification of the accuracy of lcirs photometric redshifts . figure 3(b ) shows that the rms scatter between @xmath49 and @xmath50 is @xmath51 , consistent with previous results based on a smaller number of galaxies at @xmath52 @xcite . color versus redshift for the gdds sample ( squares ) , in comparison to the cfrs sample @xcite ( crosses ) , uv - selected galaxies @xcite ( stars ) , and model predictions for a non - evolving elliptical galaxy ( solid curve ) , to an exponentially declining star formation rate ( sfr ) with e - folding time of @xmath53 ( dashed curve ) and 2 gyr ( long dash - dotted curve ) , and a constant sfr ( short dash - dotted curve ) . ( b ) comparison of photometric and spectroscopic redshifts for @xmath54 galaxies identified in the lcirs . spectroscopic redshifts for galaxies in the hdfs and cdfs were collected in part from the literature and in part from our own observations . spectroscopic redshifts for galaxies in the noao , 1511 , and ntt fields were obtained from the gdds.,scaledwidth=100.0% ] the gdds data have revealed a wealth of information regarding the ism and stellar content of massive galaxies at @xmath55 . sample spectra are presented in figure 4 , where we show various gdds composite spectra as well as individual sample spectra to demonstrate the data quality . the full sample of gdds spectra is presented in abraham et al . @xcite . specifically , we show in panel ( a ) the composite spectrum of early - type ( absorption - line dominated ) galaxies , panel ( b ) intermediate - type galaxies , and in panel ( c ) late - type ( emission - line dominated ) galaxies . for comparison , we include in panel ( a ) the sdss composite for luminous red galaxies ( lrg ) at @xmath56 @xcite ( red curve ) . our composite spectrum for early - type galaxies at @xmath57 is strikingly similar to the lrg composite spectrum at @xmath58 , showing little evolution in the stellar content of massive and quiescent systems over the redshift interval spanned by the two samples . we also show the spectrum of a galaxy at @xmath59 in panel ( d ) together with the spectrum of the host galaxy of 53w091 at @xmath60 @xcite ( red curve ) and a starburst galaxy at @xmath61 in panel ( e ) . these spectra demonstrate that we have good signal - to - noise ratios to obtain accurate and precise redshifts of individual galaxies using both rest - frame ultraviolet metal absorption features and weak continuum bumps originating in evolved stellar populations . . ( d ) the spectrum of a galaxy at @xmath59 from the gdds , together with the spectrum of the host galaxy of 53w091 at @xmath62 @xcite in red , and the model spectrum of a single burst system formed at @xmath63 in blue . ( e ) the spectrum of a starburst galaxy at @xmath64 , showing strong metal absorption features . ( f ) a normalized composite spectrum of 13 massive star - forming galaxies identified at @xmath65 from the gdds , together with a normalized composite spectrum of local starburst galaxies @xcite.,scaledwidth=100.0% ] absorption systems compiled from the literature ; and stars are for the ism of gamma - ray burst host galaxies . ( b ) rest - frame uv magnitude versus inferred stellar mass for galaxies in three redshift intervals . the solid line indicates the minimum uv flux of the galaxies included for the composite spectrum presented in panel ( f ) of figure 4.,scaledwidth=100.0% ] in addition , we have constructed a composite spectrum for 13 @xmath9-band selected galaxies at @xmath66 from the gdds sample that are also luminous in the uv with @xmath67 @xcite . a normalized version is shown in panel ( f ) of figure 4 , together with a normalized composite spectrum of local starburst galaxies from tremonti et al . the composite spectrum shows prominent absorption features from mgii , feii together with mgi and mnii , all of which are stronger than those observed in local starburst systems . it indicates that the ism of these massive star - forming galaxies has been heavily enriched with metals by @xmath68 . based on a curve - of - growth analysis of all available transitions , savaglio et al . ( @xcite ) find a mean feii column density of @xmath69 that is among the highest observed in either known damped ly@xmath70 absorption systems or in the ism of host galaxies of gamma - ray bursts ( figure 5a ) . while the metallicity is uncertain due to the unknown neutral hydrogen content , it is clear that the ism of these galaxies contains abundant heavy elements . because of the unknown nature of the galaxies that give rise to damped ly@xmath70 systems , direct column density measurements of the ism of luminous galaxies at high redshift offer a promising alternative for understanding cosmic chemical enrichment history . we also determine the formation epoch of the stars seen in quiescent systems that show no signs of recent star formation ( lack of emission features in the rest - frame uv ) . using a likelihood analysis , we derive a minimum age for each system by comparing predicted spectral energy distributions ( seds ) from different stellar synthesis models with the observed ones established from the gdds spectra and available lcirs optical and near - infrared photometry . an example is shown in panel ( d ) of figure 4 for a red galaxy at @xmath71 . the model ( blue curve ) was obtained for a single burst occurring at @xmath63 and passively evolving to lower redshift . the agreement between the observed spectrum and models suggests that this galaxy is at least 3.5 gyr old , pushing the formation epoch of this galaxy to beyond @xmath63 ( mccarthy et al . 2004 ) . finally , we estimate the total stellar mass of every galaxy in the gdds sample using a @xmath9-band mass - to - light ratio determined from comparing the observed multi - color data with a grid of model seds with varying dust extinction , metallicity , and star formation history . figure 5(b ) shows rest - frame uv flux ( expressed in @xmath72 ) versus the inferred stellar masses of individual galaxies in three redshift bins spanned by the gdds sample . we see that a large fraction of the gdds galaxies have masses @xmath73 , and these systems exhibit on average increasing rest - frame uv fluxes ( and therefore increasing star formation rate ) with increasing redshift . in addition , while few systems with stellar masses @xmath74 are seen at @xmath75 in the gdds sample due to the @xmath9 limit of the survey , a tendency for the most massive galaxies to be uv quiescent is found in the low - redshift sample ( crosses)a trend which is similar to what is seen at @xmath3 . this result supports the view that most star formation takes place in lower mass systems even at @xmath11 ( glazebrook et al . 2004 ) . in summary , the wide - field multi - color imaging data from the lcirs and the high - quality nearly complete spectral sample from the gdds together allow us to conduct a comprehensive study of massive early - type galaxies at @xmath57 . our luminosity function analysis indicates that most and perhaps all of the present - day infrared luminous galaxies were in place by @xmath76 . the rest - frame uv spectra of massive star - forming systems show abundant metals that arise in the ism of these galaxies , indicating an early heavy element enrichment by @xmath77 . finally , spectral synthesis modelling showed that near - infrared selected galaxies at @xmath11 are massive and that the quiescent ones with no evidence of recent star formation are genuinely old and likely to form at @xmath78 . together these results suggest a rapid and early formation process of massive galaxies , presenting difficulties for conventiional hierarchical formation scenarios .

on behalf of the survey teams i summarize the designs and results of the las campanas infrared survey and gemini deep deep survey , both of which were initiated to understand the nature of red galaxies and to study the history of stellar mass assembly . our results from luminosity function analysis , ism absorption line measurements , and spectral synthesis modeling show that near - infrared selected galaxies at @xmath0 are not only massive and abundant but also old and metal enriched , indicating rapid formation of massive systems at higher redshifts .

braneworld scenarios driven by extra dimensions @xcite have been identified through several strategies involving analytical procedures . in these scenarios , the observable universe is supported by a hypersurface with @xmath0 dimensions , the brane , inserted in a bulk with @xmath1 dimensions , where @xmath2 is the number of extra dimensions . paradigmatically , the extra dimension can be very large or even infinitely extended , and thereat particles and fields are supposed to be constrained by the brane properties , whereas gravity can evade as to permeate the entire bulk . matter localization is possible due to a warp factor parameter driving the metric that folds the brane through an approach that can explain , for instance , the hierarchy properties @xcite beyond the standard model . despite its challenging proposal , the search for beyond the standard model solutions of descriptions of the universe has become an arduous task . the brane scenarios considered here , and their matter localization profiles exhibited in some cases , are related to generic solutions of an effective action driven by a real scalar field , @xmath3 , coupled to @xmath4-dimensional gravity , given by @xmath5 where @xmath6 is the scalar curvature , @xmath7 denotes the metric tensor , with @xmath8 and @xmath9 running from @xmath10 to @xmath11 , and which gives rise to several strategies for investigating analytical braneworld scenarios . real scalar field theories coupled to gravity produce thick domain walls often related to integrable models through which analytical solutions of gravitating defect structures admit an internal structure . in this case , potentials driven by a single real scalar field support bps - like solutions @xcite of first - order differential equations that allow for generating braneworld scenarios through the insertion of deformed topological defects @xcite . in one - dimensional scalar field theories resumed by a scalar field @xmath12 ( from eq . ( [ 00000 ] ) ) , deformed @xmath13 and sine - gordon models @xcite , with field potentials respectively given by @xmath14 and @xmath15 , can thus trigger off such novel braneworld scenarios driven by @xmath3 . in order to guarantee the existence of defect solutions , one imposes that the potential @xmath16 generates a set of critical points , @xmath17 , such that @xmath18 as well as @xmath19 , with @xmath20 . deformed potentials exhibiting either a single or multiple minima engender two types of defects with different topologies . when it has at least two degenerate minima , it generates two new vacua corresponding to the classical field configurations , @xmath21 . in one spatial dimension the space boundaries consist in two critical points of the potential , i. e. those located at @xmath22 and @xmath23 . this spatial boundary topology is mirrored by the vacua configuration space topology of the fields , which also consists in two points , @xmath24 and @xmath25 . on the other hand , when the potential has one single minimum , @xmath26 , it leads to @xmath27 . consequently , a trivial map carries both boundary spatial points in the same field value ( @xmath28 or @xmath29 ) , which defines lump - like structures . likewise , for degenerate minima , the identity map carries @xmath22 into @xmath24 and @xmath23 into @xmath25 , which defines kink - like structures . in order to guarantee the model integrability as well as matter localization , several lump - like defects can be obtained from deformation chains ( as those from refs . @xcite and @xcite ) . departing from a primitive defect , @xmath30 , which triggers of a deformation chain - which might be identified by @xmath31 or eventually a sine - gordon kink solutions - one is able to engender @xmath32 - cyclic deformation chains through the use of a recurrence chain rule constrained by hyperbolic and trigonometric elementary relations ( see the appendix ) @xcite . in particular , the deformation chain results can be used to build up novel warp factor structures of braneworlds . to pursue such a proposal , this paper is organized as follows . in section ii , one reports about the braneworld equations of motion in @xmath33 dimensions , as well as to the general procedure for computing metric perturbations . in section iii , one introduces the _ ansatz _ for obtaining warp factors from one - dimensional lump - like solutions that support a set of novel braneworld scenarios also described in this section . finally , our conclusions are shown in section iv . let one considers the action from eq . ( [ 00000 ] ) for a @xmath4 dimensional model driven by the rs2 metric @xcite , @xmath34 where @xmath35 , with @xmath36 , and where the extra dimension appears in the warp factor function , @xmath37 , that localizes matter around @xmath38 . assuming a dependency of @xmath3 exclusively on @xmath39 , the equation of motion for the scalar field arising from the action ( [ 00000 ] ) is given by @xmath40 by varying the action , @xmath41 , with respect to the metric @xmath42 , one obtains @xmath43 and after some mathematical manipulations , one has @xmath44 which are the einstein equations rewritten in a convenient form . the scalar field potential is used to be written in terms of a _ superpotential _ , @xmath45 , in a particular form given by @xmath46 which has been largely considered in the literature @xcite . in order to demote from second to first order field equations , a superpotential @xmath47 is introduced as to set @xcite , @xmath48 such that the above introduced field equations can be rewritten as first order equations , @xmath49 @xmath50 in order to examine the distribution of matter inside the brane , one recovers the energy density derived from ( [ 00000 ] ) @xcite , @xmath51e^{2a(y ) } , \label{em}\ ] ] and , in the same scope , a quantum mechanical analogue to examine the stability of metric fluctuations is motivated by the interest of finding a resonance spectrum like those exhibited by _ volcano _ potentials . gravitational wave quantum analogue equations can be straightforwardly derived through tensor metric perturbations in eq . ( [ intervaloinvariante ] ) , as to have @xmath52 where a factor @xmath53 is extracted from the fluctuation term as to simplify subsequent equations , with @xmath54 in the form of transverse and traceless tensor perturbations , for which one obtains @xmath55 for the linearized gravity decoupled from the scalar field . if the zero mode solution of the resonance spectrum is normalizable - which can be translated into some kind of localization - it is possible to localize the four - dimensional gravity inside the brane . considering the convenient change of variable , @xmath56 and denoting @xmath57 , with @xmath58 arbitrary , one arrives at the one - dimensional schrdinger equation , @xmath59 where @xmath60 is the quantum mechanical analogue potential given by @xmath61 through eq . ( [ sch ] ) , one recovers the hamiltonian form given by @xmath62 with @xmath58 real and @xmath63 , as to not exhibit unstable modes . by means of eq . ( [ hamiltoniana ] ) , one finds the zero mode solution ( @xmath64 ) , @xmath65 where @xmath66 is the normalization constant . finally , a short comment related to reducing second order nonlinear equations to a system of firs order nonlinear equations is pertinent , in the sense discussed in the end of section v of the ref . @xcite , where it has been stablished that such order reduction must be followed by the correct absorption of the related integration constants , implicitly considered along our calculations . the above equations summarize the simplified mathematical description of the brane scenarios with one extra dimension , such that the warp factor calculation procedure can be setup in the next section . assuming that localized energy structures on the brane demand for warp factors delineated by a lump - like behavior , which guarantees the model integrability and matter localization , one identification of a localized behavior described by an arbitrary _ bell - shaped _ function , @xmath67 , can be performed through @xmath68 with @xmath67 known _ a priori_. thereat , @xmath69 is given straightforwardly by @xmath70 , \label{a}\ ] ] as to have @xmath71 a plethora of kink an lump - like analytical defects were released in the ref . @xcite generated from cyclic deformation chains ( see the appendix ) triggered by the solutions derived from a @xmath13 model ( besides other three deformed models ) . several of those lump - like solutions can be inserted into eq . ( [ a ] ) as to generate braneworld scenarios @xcite . departing from some previously known analytical profile for a defect , @xmath30 , a deformation procedure @xcite sets that the driving potential can be rewritten as @xmath72\left(\frac{d\tilde{\chi}}{d\chi}\right)^{-2},\ ] ] where the deformation function , @xmath73 , relates minimal points from the primitive defect , @xmath30 , to minimal points from deformed solutions , @xmath74 , in such a way that , @xmath75 , \label{def}\ ] ] where @xmath76 and @xmath77 are respectively primitive and deformed potentials . as for triggering a cyclic deformation chain , it has been shown that the deformation procedure can exhibit a topological mass constraint @xcite given by @xmath78 thereat the summation of the @xmath79 deformed defect topological masses is equal to the primitive defect mass , in a kind of cyclic chain of one - dimensional defect deformation . cyclic chains of kink and lump - like deformed defects have been built through the above procedure by means of deformation functions , @xmath80 , exhibiting either hyperbolic or trigonometric behaviors ( see the appendix ) @xcite . since deforming chains with three or four connected defects ( @xmath81-cyclic , for @xmath82 , and @xmath11-cyclic , for @xmath83 , deformations ) have been considered , the above indexed label @xmath74 is given in certain generic sense : the primitive kink solution is always labeled by @xmath30 , while different greek letters ( @xmath84 and @xmath85 ) are used to index the deformed defects . the primitive kink solutions that shall have triggered off the deformation defect generation depicted from @xmath86 - cyclic chains ( a @xmath81-cyclic one , with @xmath87 , and a @xmath11-cyclic one with @xmath88 ) from ref . @xcite have been given by @xmath89 , \\ \chi_4(y)&= & 2(2+y^2)^{-1},\end{aligned}\ ] ] from which , deformed defects identified by @xmath74 into eq . ( [ def ] ) could be obtained and identified with @xmath90 from eq . ( [ a ] ) , @xmath91 as to support warp factors that generate consistent braneworld scenarios driven by a scalar field @xmath3 as set by the action from ( [ 00000 ] ) . that is , the deformed defects depicted from @xmath81 and @xmath11-cyclic deformation chains from refs . @xcite that shall be used to construct novel braneworld scenarios parameterized by @xmath92 are summarized in table [ tabela ] , which describes a plenty of deformed lump - like solutions ( and the respective correspondence with refs . @xcite ) . they shall be analytically described in the sections where they are required . @*3l@ & & + ( lr)2 - 3 & @xmath93 & @xmath94 + & & @xmath95 + & & @xmath96 + & & @xmath97 + ( lr)2 - 3 & @xmath93 & @xmath95 + ( lr)1 - 1 according to eq . ( [ a ] ) , the localization profiles described by such deformed solutions , @xmath92 , apart from a normalization constant , shall reproduce the localization profiles of the warp factors , @xmath37 , for each one of those novel braneworld scenarios . in the following , one shall identify the results obtained for all the scenarios summarized by table [ tabela ] , from which the warp factor is identified by eq . ( [ a ] ) , and used as an input function for computing the energy density , @xmath98 ( c. f. eq . ( [ em ] ) ) , and the corresponding qm potential , @xmath99 ( c. f. eq . ( [ qmpotential ] ) ) . the first scenario that hereon considered is driven by the _ ansatz : @xmath101 \operatorname{sech}(n)]}{n},\ ] ] is thus summarized by @xmath102 \operatorname{sech}(n)]}{n } \bigg],\ ] ] @xmath103 ^ 2 -2 \operatorname{sech}(y)^2\tanh(y)\tanh[n \tanh(y ) ] \bigg ] , \\\ ] ] @xmath104 \nonumber\\ & & \quad\quad + n \operatorname{sech}(y)^2 \bigg(-4 \operatorname{sech}[n \tanh(y)]^2-\frac{\tanh[n \tanh(y)]^2}{\log[\cosh[n \tanh(y ) ] \operatorname{sech}(n)]}\bigg ) \bigg],\end{aligned}\ ] ] from which the zero mode is straightforwardly calculated through eq . ( [ modozero ] ) . the results are depicted in fig . [ lambda1 ] , where lump solutions depend on a free parameter @xmath79 , for which one notices that @xmath105 , with @xmath106 . a similar braneworld scenario obtained now for a deformed solution from a trigonometric chain with the _ ansatz _ : @xmath107 - \cos(n)}{n},\ ] ] is thus summarized by @xmath108}{n } ] , \\ t_{00}(y ) & = & \frac{3}{4 } \bigg [ n \cos[n \tanh(y ) ] \operatorname{sech}(y)^4 - 2 \operatorname{sech}(y)^2 \sin[n \tanh(y ) ] \tanh(y ) \bigg],\end{aligned}\ ] ] @xmath109 + 5 \cos[2 n \tanh(y ) ] ) \operatorname{sech}(y)^4 } { 32 ( \cos(n ) - \cos[n \tanh(y ) ] ) } \nonumber\\ & & \quad\quad\quad\quad\quad\quad\quad\quad\quad + \frac{3}{2 } \operatorname{sech}(y)^2 \sin[n \tanh(y ) ] \tanh(y).\end{aligned}\ ] ] for which the results are depicted in fig . [ lambda2 ] . the braneworld scenarios derived from four different deformed lumps ( see table [ tabela ] ) : @xmath111 - \operatorname{sech}(n)}{n } , \\ \varphi(\chi_2(y))&=&\frac{1}{n } \tanh[n \operatorname{sech}(y ) ] , \\ \varphi(\chi_3(y))&= & \frac{1}{n } \tanh[n \ln[1 + \operatorname{sech}(y)^2 ] ] , \\ \varphi(\chi_4(y))&= & \frac{1}{n } \tanh \bigg[\frac{2n}{(2 + y^2 ) } \bigg],\end{aligned}\ ] ] are summarized by the warp factors given by , @xmath112-\operatorname{sech}(n)}{n}\right]},\\ a_{\varphi(\chi 2)}(y ) & = & \frac{1}{2 } \ln \bigg[\frac{\tanh[n \operatorname{sech}(y)]}{n } \bigg],\\ a_{\varphi(\chi 3)}(y ) & = & \frac{1}{2 } \ln \bigg[\frac{\tanh[n \log[1 + \operatorname{sech}(y)^2]]}{n } \bigg],\\ a_{\varphi(\chi 4)}(y ) & = & \frac{1}{2 } \ln \bigg[\frac{1}{n}\tanh\left[\frac{2 n}{2 + y^2}\right ] \bigg].\end{aligned}\ ] ] from which ( and from this point ) the expressions for the energy density , @xmath98 ( c. f. eq . ( [ em ] ) ) , and the corresponding qm potential , @xmath99 ( c. f. eq . ( [ qmpotential ] ) ) have been suppressed due to their extensive size . the plots for such complete scenarios derived from the lumps @xmath113 , @xmath114 , @xmath115 and @xmath116 are depicted in figs . [ lambda3 ] , [ lambda4.1 ] , [ lambda4.2 ] and [ lambda4.3 ] respectively . the brane scenarios for models depicted in figs . [ lambda4.2 ] and [ lambda4.3 ] give rise to thick branes , most of them with no internal structures . however , for increasing values of the arbitrary parameter @xmath79 ( blue lines ) , the potential that drives the scalar field allows the appearance of thick branes that host internal structures in the form of a layer of an extra phase circumvented by two separate interfaces . it is related to a more concentrated energy density of the matter field with a corresponding extension / localization of the warp factor which exhibits a profile that approaches to a _ form in the region inside the brane . the mentioned internal structure is noticed from the energy profile . the last braneworld scenario obtained from a deformed lump @xmath117}{4 n}\ ] ] supported by a @xmath13 model leads to the warp factor , @xmath118}{4 n } \bigg],\ ] ] now , it is convenient do add some relevant comments to the results obtained up to this point . by observing the qm potentials obtained for the above solutions supported by @xmath13 model , for all the above obtained solutions , one notices that they exhibit a generalized _ volcano _ behavior and may therefore contain normalized zero modes . zero modes were indeed analytically found for all the cases and they are normalizable for the above solutions and for any value of the parameter @xmath79 . it reveals that our brane solutions can support four - dimensional gravity @xcite . another feature that arises from the plot observation is that the variation of the parameter @xmath79 leads to small changes in the brane profile , changing the values of their critical points . the plots related to the models derived from @xmath96 and @xmath97 were constructed with a different value of @xmath79 , @xmath105 . thereat , its possible to see the emergence of a _ plateau _ for @xmath119 , opposing to the lower values of @xmath79 , where this structure is not presented . furthermore , through the energy - momentum tensor profile , one notices that an interface interpolating two bulk phases is broken into another two separated interfaces . that is , there arises a new phase between the two interfaces , a phenomenon known as complete wetting in condensed matter@xcite . therefore , one can infer that this brane scenario supports internal structures @xcite . the region where the new phase arises corresponds to the _ plateau _ region exhibited by the warp factor . although the brane defect can not be analytically calculated , a taylor series analysis has been implemented as to obtain a robust double - kink approximation for the field solution : this is exactly the kind of defect that one expects in thick brane scenarios , where there internal structures are present @xcite . in fig . [ lambda5.1 ] , corresponding to the @xmath120 in the @xmath110cyclic hyperbolic chain , one also notices that @xmath119 . however , a different behavior arises when one compares it with those results found for @xmath121 and @xmath122 relative to the hyperbolic chain . through the @xmath37 profile , one notices that there is a lump interpolated by a _ plateau_. in this case , through careful observation of the energy - momentum tensor profile , one notices the appearance of several novel phases , generating four matter picks instead of two . this behavior indicates that the resulting defect has further details in its structure than those presented in the double - kink one . despite its relevance , analytical braneworld scenarios are sparing in the literature . in this work several novel braneworld scenarios have been generated via lump deformed defects inserted in the warp factor calculation . likewise , the qm analogue problem has been studied by means of metric perturbations . all deformed models are exhibited through an analytical description for which at least one normalizable zero mode has been obtained . in all these cases one therefore can recover @xmath110dimensional gravity inside the brane , near @xmath123 . in particular , one has noticed the two driving models for thick brane solutions where a double - kink approximation was found for the topological defect . this indicates that the brane supports an internal structure . for the model with a more intricate structure , one sees in the warp factor profile the formation of a lump on the _ plateau _ structure . for the first time in the literature , a braneworld model supports more than two peaks in the tensor - energy momentum , so that there is the possibility that this brane supports topological defects even more complex than the double - kink , which deserves a more careful investigation . _ the work of aeb is supported by the brazilian agency cnpq ( grant no . 300809/2013 - 1 and grant no . 440446/2014 - 7 ) . rdr is grateful to cnpq ( grants no . 303027/2012 - 6 and no . 473326/2013 - 2 ) , and to fapesp ( grant 2015/10270 - 0 ) for partial financial support . a generalized @xmath124-deformation process , with @xmath125 is identified by the @xmath124-derivative operation given by @xmath126}(\vartheta^{[\kappa ] } ) = \frac{d\vartheta^{[\kappa]}}{d\vartheta^{[\kappa-1]}},\ ] ] with @xmath127}$ ] identified by a real scalar fields corresponding to a primitive defect structures constructed from an @xmath32 deformation chain triggered by @xmath128}$ ] , such that one effectively has @xmath127 } \equiv \vartheta^{[\kappa]}(\chi)$ ] which can be cyclically recovered by successive deformations . a hyperbolic chain @xcite is defined through the systematic relations , @xmath129}_{\chi } & = & \tanh{(\chi ) } , \nonumber\\ \vartheta^{[1]}_{\chi } & = & \tanh{(\chi)}\operatorname{sech}{(\chi ) } , \nonumber\\ \vartheta^{[2]}_{\chi } & = & \tanh{(\chi)}\operatorname{sech}{(\chi)}^{2 } , \nonumber\\ & \vdots & \nonumber\\ \vartheta^{[n-1]}_{\chi } & = & \tanh{(\chi)}\operatorname{sech}{(\chi)}^{n-1 } , \nonumber\\ \vartheta^{[n]}_{\chi } & = & \operatorname{sech}{(\chi)}^{n } , \label{aa}\end{aligned}\ ] ] where the upper index corresponds to the respective derivative , which upon straightforward integrations results into @xmath129}(\chi ) & = & \ln{[\cosh{(\chi ) } ] } , \nonumber\\ \vartheta^{[1]}(\chi ) & = & - \operatorname{sech}{(\chi ) } , \nonumber\\ \vartheta^{[2]}(\chi ) & = & - \frac{1}{2 } \operatorname{sech}{(\chi)}^{2 } , \nonumber\\ & \vdots & \nonumber\\ \vartheta^{[n-1]}(\chi ) & = & - \frac{1}{n-1}\operatorname{sech}{(\chi)}^{n-1 } , \nonumber\\ \vartheta^{[n]}(\chi ) & = & \sinh{(\chi)}\ , _ 2f_1\left[\frac{1}{2},\ , \frac{1+n}{2},\ , \frac{3}{2},\ , -\sinh{(\chi)}^2\right ] , \nonumber\\ \label{bb}\end{aligned}\ ] ] where @xmath130 is the gauss hypergeometric function and the integration constants have been suppressed . from eqs . ( [ aa]-[bb ] ) one identifies @xmath131}(\vartheta^{[\kappa ] } ) = \operatorname{sech}{(\chi ) } \equiv \exp{[-\vartheta^{[0]}]},&~~ & \kappa = 1,\,2,\ , \ldots,\,n-1,\\ g^{[n]}(\vartheta^{[n ] } ) = 1/\sinh{(\chi ) } , & ~~&\end{aligned}\ ] ] and @xmath132}(\vartheta^{[\kappa ] } ) } = \frac{d\vartheta^{[n-1]}}{d\vartheta^{[0 ] } } = \operatorname{sech}{(\chi)}^{n-1 } \equiv \exp{[- ( n-1 ) \vartheta^{[0 ] } ] } , \label{dd}\ ] ] which set the complete form of the chain rule of the @xmath32-cyclic deformation , as it can be expressed by @xmath133}}{d\chi}\,\prod_{\kappa=1}^{n}{g^{[\kappa]}(\vartheta^{[\kappa]})}\,\frac{d\chi}{d\vartheta^{[n ] } } = 1 , \label{ee}\ ] ] from which a closed cycle of @xmath32 deformation functions is summarized . analogously , a trigonometric deformation chain can be engendered by changing @xmath134 and @xmath135 into eqs . ( [ aa]-[bb ] ) , as to have @xmath129}(\chi ) & = & \cos{(\chi ) } , \nonumber\\ \vartheta^{[1]}(\chi ) & = & \frac{1}{2}\cos{(\chi)}^{2 } , \nonumber\\ \vartheta^{[2]}(\chi ) & = & \frac{1}{3}\cos{(\chi)}^{3 } , \nonumber\\ & \vdots & \nonumber\\ \vartheta^{[n-1]}(\chi ) & = & \frac{1}{n}\cos{(\chi)}^{n } , \nonumber\\ \vartheta^{[n]}(\chi ) & = & -\frac{\cos{(\chi)^{n+1}}}{n+1}\ , _ 2f_1\left[\frac{1+n}{2},\ , \frac{1}{2},\ , \frac{3 + n}{2},\ , \cos{(\chi)}^2\right ] . \label{bb2}\end{aligned}\ ] ] as it has been performed in refs . @xcite . 99 l. randall and r. sundrum , phys . . lett . * 83 * , 4690 ( 1999 ) . v. a. rubakov , and m. e. shaposhnikov , phys . b*125 * , 136 ( 1983 ) . v. a. rubakov , and m. e. shaposhnikov , phys . b*125 * , 139 ( 1983 ) . n. arkani - hamed , s. dimopoulos and g. dvali , phys . b*429 * , 4237 ( 1998 ) . l. randall and r. sundrum , phys . lett . * 83 * , 3370 ( 1999 ) . r. rajaraman , _ solitons and instantons _ , ( north - holland , amsterdam , 1982 ) . d. bazeia , l. losano and j. m. c. malbouisson , phys . rev . d*66 * , 101701 ( 2002 ) . d. bazeia and l. losano , phys . rev . d*73 * , 025016 ( 2006 ) . a. e. bernardini and m. chinaglia , mod . phys . lett . a * 30 * , 50118 ( 2015 ) . a. e. bernardini and r.da rocha , ahep * 2013 * , 304980 ( 2013 ) . a. e. bernardini , m. chinaglia and r. da rocha , eur . j. plus * 130 * , 97 ( 2015 ) . n. sasakura , jhep * 0202 * , 026 ( 2002 ) ; phys . rev . d*66 * , 065006 ( 2002 ) . v. i. afonso , d. bazeia and l. losano , phys . b*634 * , 526 ( 2006 ) . o. dewolfe , d. z. freedman , s. s. gubser and a. karch , phys . rev . d*62 * , 046008 ( 2000 ) . m. gremm , phys . b*478 * , 434 ( 2000 ) . c. cski , j. erlich , t.j . hollowood and y. shirman , nucl . b*581 * , 309 ( 2000 ) . d. bazeia , j. menezes and r. menezes , phys . * 91 * , 241601 ( 2003 ) . o. arias , r. cardenas and i. quiros , nucl . b*643 * , 187 ( 2002 ) . n. barbosa - cendejas and a. herrera - aguilar , phys . rev . d*73 * , 084022 ( 2006 ) . a. e. bernardini and o. bertolami , phys . b*726 * , 512 ( 2013 ) . a. campos , k. holland and u. j. wiese , phys . rev . lett . * 81 * , 2420 ( 1998 ) . d. bazeia , c. furtado and a. r. gomes , jcap * 0402 * , 002 ( 2004 ) . d. bazeia , j. menezes and r. menezes , phys . lett . * 91 * , 241601 ( 2003 ) . a. t. avelar , d. bazeia , l. losano and r. menezes , eur . j c*55 * , 133 ( 2008 ) . respective to the @xmath100cyclic hyperbolic chain . warp factor ( @xmath136 column ) and zero mode solution ( @xmath137 column ) are presented in the first line ; qm potential as a function of @xmath138 ( @xmath136 column ) and the energy density ( @xmath137 column ) are presented in the second line . blue line represents the solution for @xmath139 , noting that @xmath105 , with @xmath106 . black lines range from dotted to dashed as @xmath79 increases . ] respective to the @xmath100cyclic trigonometric chain . warp factor ( @xmath136 column ) and zero mode solution ( @xmath137 column ) are presented in the first line ; qm potential as a function of @xmath138 ( @xmath136 column ) and the energy density ( @xmath137 column ) are presented in the second line . blue line represents the solution for @xmath139 while black lines range from dotted to dashed as @xmath79 increases ( as for the same parameters of fig . [ lambda1 ] ) . ] respective to the @xmath110cyclic hyperbolic chain . warp factor ( @xmath136 column ) and zero mode solution ( @xmath137 column ) are presented in the first line ; qm potential as a function of @xmath138 ( @xmath136 column ) and the energy density ( @xmath137 column ) are presented in the second line . blue line represents the solution for @xmath139 while black lines range from dotted to dashed as @xmath79 increases ( as for the same parameters of fig . [ lambda1 ] ) . ] respective to the @xmath110cyclic hyperbolic chain . warp factor ( @xmath136 column ) and zero mode solution ( @xmath137 column ) are presented in the first line ; qm potential as a function of @xmath138 ( @xmath136 column ) and the energy density ( @xmath137 column ) are presented in the second line . blue line represents the solution for @xmath139 while black lines range from dotted to dashed as @xmath79 increases ( as for the same parameters of fig . [ lambda1 ] ) . ] respective to the @xmath110cyclic hyperbolic chain . warp factor ( @xmath136 column ) and zero mode solution ( @xmath137 column ) are presented in the first line ; qm potential as a function of @xmath138 ( @xmath136 column ) and the energy density ( @xmath137 column ) are presented in the second line . blue line represents the solution for @xmath140 while black lines range from dotted to dashed as @xmath79 increases ( as for the same parameters of fig . [ lambda1 ] ) . ] respective to the @xmath110cyclic hyperbolic chain . warp factor ( @xmath136 column ) and zero mode solution ( @xmath137 column ) are presented in the first line ; qm potential as a function of @xmath138 ( @xmath136 column ) and the energy density ( @xmath137 column ) are presented in the second line . blue line represents the solution for @xmath140 while black lines range from dotted to dashed as @xmath79 increases ( as for the same parameters of fig . [ lambda1 ] ) . ] respective to the @xmath110cyclic trigonometric chain . warp factor ( @xmath136 column ) and zero mode solution ( @xmath137 column ) are presented in the first line ; qm potential as a function of @xmath138 ( @xmath136 column ) and the energy density ( @xmath137 column ) are presented in the second line . blue line represents the solution for @xmath140 while black lines range from dotted to dashed as @xmath79 increases ( as for the same parameters of fig . [ lambda1 ] ) . ]

novel braneworld scenarios supported by warp factors driven by a single extra dimension are obtained from deformed one - dimensional lump - like solutions known _ a priori_. through a novel _ ansatz _ , the internal energy structure , the braneworld warp factor , and the quantum mechanical analogue problem , as well as the associated zero mode solutions , are straightforwardly derived by means of an analytical procedure . the results allow one to identify thick brane solutions that support internal structures and that can hold the ( @xmath0)-dimensional gravity .

quantum phase transitions ( qpts)@xcite which separate competing ground states represent generic phenomena in solid - state systems at zero temperature . the transition is frequently found to be continuous , often times giving rise to a quantum critical point . in the neighborhood of a quantum critical point of a metallic system the finite temperature properties as a rule show non - fermi liquid behavior@xcite . in recent years , quantum phase transitions at the nanoscale have attracted much attention@xcite . much of the effort has been focused on the breakdown of the kondo effect in transport of a quantum dot due to its coupling to a dissipative environment . however , relatively less is known about the corresponding out - of - equilibrium properties@xcite . a finite bias voltage applied across a nanosystem is expected to smear out the equilibrium transition , but the current - induced decoherence might act quite differently as compared to thermal decoherence at finite temperature @xmath4 , resulting in exotic behavior near the transition . meanwhile , understanding the interplay of electron interactions and non - equilibrium effects in quantum systems is one of the most challenging open questions in condensed matter physics . many of the theoretical approaches that have been proven so successful in treating strongly correlated systems in equilibrium are simply inadequate once the system is out of equilibrium . the real - time schwinger - keldysh formalism@xcite has been known as the most successful approach to non - equilibrium dynamics since it offers a controlled perturbative expansion of the density operator . however , care must be taken to avoid the appearance of infrared divergences , in the perturbative approaches . though much is known for quantum impurity systems in equilibrium , understanding their properties in non - equilibrium steady - state is still limited . nevertheless , significant progress has been made by different approaches , such as ( 1 ) analytical approximations : perturbative renormalization group method ( rg ) @xcite , hamiltonian flow equations@xcite , functional rg@xcite , strong - coupling expansions@xcite , master equations@xcite ; ( 2 ) exact analytical solutions : field theory techniques@xcite , the scattering bethe ansatz@xcite , mapping of a steady - state non - equilibrium problem onto an effective equilibrium system@xcite ; ( 3 ) numerical methods : time - dependent density matrix renormalization group ( rg)@xcite , time - dependent numerical rg@xcite , diagrammatic monte carlo@xcite , and imaginary - time nonequilibrium quantum monte carlo@xcite . in this paper , we provide a comprehensive analysis of the non - equilibrium transport near a quantum phase transition in a dissipative resonant level model by employing the recently developed frequency - dependent rg@xcite and functional rg approaches@xcite , and extending our earlier work in ref . . we aim to address several fundamental questions related to the non - equilibrium transport in quantum dot settings , such as : what is the distinct non - equilibrium conductance profile at zero temperature compared to that in equilibrium at finite temperatures near the transition ? is there any scaling behavior of the conductance at finite temperatures and finite bias voltage near the transition ? for this purpose , we investigate three classes of typical nano - models comprising a spinless resonant level coupled to : ( i ) two non - interacting fermi - liquid leads subject to an ohmic dissipative environment , where an ohmic environment can be realized in a nanoscale resistor and has many applications in physics ranging from mesoscopic physics ( refs . ) to biological systems@xcite , ( ii ) two interacting fermion baths , in particular two fractional quantum hall edge ( fqhe)@xcite leads , or the `` chiral luttinger liquids '' where electrons on the edge of a 2d fractional quantum hall system show one - dimensional chiral luttinger liquid behaviors with only one species of electrons ( left or right movers ) , ( iii ) two interacting luttinger - liquid leads subject to an ohmic dissipative environment . in the class ( i ) model , the qpt separating the conducting and insulating phase for the level is solely driven by dissipation , which can be modeled by a bosonic bath . dissipation - driven qpts have been addressed theoretically and experimentally in various systems , such as : quantum dot systems@xcite , josephson junction arrays@xcite , superconducting thin film@xcite , superconducting qubit@xcite , qubits or resonant level systems coupled to photonic cavities@xcite , and biological systems@xcite . here , we focus on the non - equilibrium properties of the system near quantum phase transition . meanwhile , for the class ( ii ) model , tunneling of electrons or quasi - particles between two fqhe states may in general suffer from the electron - electron interactions in fqhe . interesting experimentally relevant questions arise regarding how interaction effects modify the nonequilibrium charge transport in such systems . furthermore , one can extend the above two classes of models to a more general class ( iii ) model where both electron - electron interactions and the dissipation are present in the fqhe setpups , which have not been explored both theoretically and experimentally . our results have relevance for the recent experiment in ref . where the electronic transport through a resonant level in a nanotube exhibits the luttinger liquid behavior , namely the conductance demonstrates a non - trivial power - law suppression as a function of bias voltage . this paper is organized as follows : in section ii a , the model hamiltonian of class ( i ) is introduced . in section ii b , we establish rigorious mappings of our model system , the class ( i ) model at a finite bias voltage , onto the out - of - equilibrium anisotropic kondo model as well as onto class ( ii ) and ( iii ) model systems subject to a finite voltage bias . we compute the current operator in section ii c for these three classes of models . we employ the nonequilibrium rg approach in section iii . our results on nonequilibrium transport near the quantum phase transition both at zero and finite temperatures are presented in section iv , followed by the results on the nonequilibrium finite - frequency current noise in section v. we make a few remarks on the important issues of nonequilibrium quantum criticality in section vi . finally , we draw conclusions in section vii . our hamiltonian in all of the three classes of models mentioned above takes the following generic form : @xmath5 c^{\dagger}_{k i}c_{k i } + t_{i } c^{\dagger}_{ki } d + h.c . \\ & + & \sum_{r } \lambda_{r } ( d^{\dagger}d-1/2 ) ( b_{r } + b^{\dagger}_{r } ) + \sum_{r } \omega_{r } b^{\dagger}_{r } b_{r } , \nonumber \label{rlm}\end{aligned}\ ] ] where @xmath6 is the ( real - valued ) hopping amplitude between the lead @xmath7 and the quantum dot , @xmath8 and @xmath9 are electron operators for the ( fermi - liquid type ) leads and the quantum dot , respectively . @xmath10 is the chemical potential shift applied on the lead @xmath7 ( @xmath0 will denote the bias voltage throughout this paper ) , while the dot level is at zero chemical potential . here , @xmath11 are the boson operators of the dissipative bath with an ohmic type spectral density@xcite : @xmath12 . note that usually we introduce a cutoff via a @xmath13 function in @xmath14 ; here , we assume that @xmath15 is a large energy scale comparable to the energy bandwidth of the reservoir leads . to simplify the discussion , we assume that the electron spins have been polarized through the application of a strong magnetic field . note that our model can be realized experimentally in a quantum dot coupled to resistive environment as shown in ref . . in this section , we briefly summarize the behavior of our model system at equilibrium which means in the absence of a finite bias voltage ( @xmath16 ) . a dissipative resonant - level systems in equilibrium coupled to several leads maps onto the anisotropic one - channel kondo model@xcite where the dimensionless transverse kondo coupling @xmath17 is proportional to the hopping @xmath18 between the level and the leads and the longitudinal coupling @xmath19 ( the exact prefactors are given in refs . ; see also sec . ii b and appendix a ) . here , the superscript @xmath20 in @xmath21 refers to the equilibrium couplings . the model exhibits a kosterlitz - thouless ( kt ) qpt from a delocalized ( kondo screened ) phase for @xmath22 , with a large conductance , @xmath23 , to a localized ( local moment ) phase for @xmath24 , with a small conductance , as the dissipation strength is increased ( see fig . 1 ) . for @xmath25 , the kt transition occurs at @xmath26 . as @xmath27 , the kondo temperature @xmath28 obeys@xcite : @xmath29 . note that here we assume our resonant level system exhibits the particle - hole ( p - h ) symmetry ; namely , the resonant - level energy @xmath30 is set to be zero ( @xmath31 ) . however , in a more general resonant - level model where p - h symmetry is abscent , an additional term of the form @xmath32 is present in the hamiltonian eq . ( [ rlm ] ) . in terms of its equivalent kondo model , this p - h symmetry breaking term plays the role as an effective local magnetic field @xmath33 acting on the impurity spin in the kondo model@xcite , which needs more involved treatments and exceeds the scope of a simple and generic model system considered in the present work . in equilibrium , the dimensionless scaling functions @xmath34 and @xmath35 at the transition are obtained via the renormalization - group ( rg ) equations of the anisotropic kondo model : @xmath36 where @xmath37 , with @xmath38 being the ultraviolet cutoff . having in mind a quantum dot at resonance , @xmath39 , with @xmath40 being the level spacing on the dot and @xmath41 the cut - off of the bosonic bath ; @xmath38 is of the order of a few kelvins . at low temperatures @xmath42 , the conductance drops abruptly with decreasing temperatures@xcite : @xmath43 ^{2}\propto \frac{1}{\ln ^{2}(t/\mathcal{d})}. \label{gteq}\ ] ] below , we analyze the non - equilibrium @xmath44 ) transport of our model system at the kt transition and in the localized phase in the double - barrier resonant tunneling regime where the dissipative resonant level couples symmetrically to the two leads ( @xmath45 ) . note , however , that when the dissipative resonant level couples asymmetrically to the leads @xmath46 , as has been observed experimentally in ref . , the system reaches the single - barrier tunneling regime , leading to luttinger liquid behavior in conductance with power - law dependence in bias voltage . + for the sake of convenience , we set the following units throughout the rest of the paper : @xmath47 , and the temperature @xmath4 is in unit of @xmath48 . our generic model hamiltonian eq . ( [ rlm ] ) in fact can be mapped onto various related model systems as we shall discuss below , including the anisotropic kondo model ( class ( i ) ) , the class ( ii ) and ( iii ) systems mentioned above . here , we will address the non - equilibrium transport through a dissipative resonant level based on one of the equivalent models : the two - lead anisotropic kondo model . the mappings for the three classes of models discussed below will be derived in an analogous way . the general scheme of these mappings is via bosonization followed by re - fermionization ( or in the opposite order)@xcite . first , we envision a non - equilibrium mapping revealing that the leads are controlled by distinct chemical potentials . through similar bosonization and refermionization procedures as in equilibrium , our model is mapped onto an anisotropic kondo model@xcite with the effective ( fermi - liquid ) left ( @xmath49 ) and right lead ( @xmath50)@xcite ( see appendix a for details ) : @xmath51c_{k\gamma \sigma } ^{\dagger } c_{k\gamma \sigma } \nonumber \\ & + & ( j_{\perp } ^{(1)}s_{lr}^{+}s^{-}+j_{\perp } ^{(2)}s_{rl}^{+}s^{-}+h.c.)\nonumber \\ & + & \sum_{\gamma = l , r}j_{z}s_{\gamma \gamma } ^{z}s^{z } , \nonumber \\ \label{hkondo}\end{aligned}\ ] ] where @xmath52 is the electron operator of the effective lead @xmath53 , with @xmath54 the spin quantum number , @xmath55 is the index for the effective non - interacting fermionic leads , @xmath56 , @xmath57 , and @xmath58 where @xmath59 describes the charge occupancy of the level . additionally , @xmath60 are the spin - flip operators between the effective leads @xmath61 and @xmath62 , @xmath63 embody the transverse kondo couplings , @xmath64 , and @xmath65 . it should be noted that this mapping is exact near the phase transition where @xmath66 or @xmath67 , and thus @xmath68 . note that the above mapping takes a spinless dissipative resonant level model with spinless fermionic baths @xmath69 to the anisotropic kondo model with a `` spinful '' quantum dot ( with spin operator given by @xmath70 ) and `` spinful '' conduction electron leads @xmath71 . the appearance of the `` pseudo - spin '' degrees of freedom in the effective kondo model can be understood in terms of the `` charge kondo '' effect : the the tunneling between a resonant level ( which can be represented by a `` qubit '' or a `` spin '' ) and the spin - polarized leads plays an equivalent role as the `` pseudo - spin '' flips between spin of a quantum dot and that of the conduction electrons ; and the coupling of the charge of the resonant level to the bosonic environment acts as the ising coupling between z - components of the pseudo - spins on the dot and in the effective leads@xcite . meanwhile , as mentioned above , when the resonant - level model shows p - h asymmetry , an additional term @xmath32 appears in the hamiltonian , which is equivalent to a local magnetic field acting on the impurity spin @xmath72 via the identification : @xmath73 , @xmath74 , and @xmath75 . for simplicity , we do not intend to study further this p - h asymmetry term and focus mainly on the effective kondo model in the absence of magnetic field . note also that the mapping has been derived earlier in ref . and is well - known at equilibrium ( ref . ) . in appendix a , we will provide more details regarding the different theoretical steps , in particular with a finite bias voltage . our analysis for the non - equilibrium transport of a dissipative resonant level model is applicable for describing a resonant level quantum dot coupled to two chiral luttinger liquid leads , which is relevant for describing quasiparticle tunneling between two fractional quantum hall edge ( fqhe ) states@xcite ( the class ( ii ) model mentioned above ) . in the absence of bias voltage , this case has been studied in refs . . via the standard bosonization , @xmath76 the hamiltonian of such system can be written as@xcite ( see appendix a. ) : @xmath77 where the lead term @xmath78 describes two chiral luttinger liquid leads with lead index @xmath79 , @xmath80 denotes the tunneling term and the bias voltage term @xmath81 is given respectively by : @xmath82 where the boson field @xmath83 denotes the chiral luttinger liquid in lead @xmath84 , the tunneling between lead and the resonant level is given by @xmath85 , @xmath0 is the bias voltage , and @xmath86 refers to the luttinger parameter . here , we set @xmath87 throughout the paper with @xmath88 being the lattice constant . through the similar refermionization , we arrive at the effective kondo model as shown in eq . ( [ hkondo ] ) with the bare kondo couplings @xmath89 , @xmath90 . the non - equilibrium rg scaling equations for @xmath91 have the same form as in eq . ( [ rg2 ] ) . + so far , we consider here just a dissipative resonant single - level coupled to two non - interacting leads . nevertheless , the mapping can be straigthforwardly generalized to the same system but with a spinless quantum dot which contains many energy levels . in this case , the effective luttinger liquid parameter @xmath92 is modified as : @xmath93 ( see eq . ( [ kprime ] ) in appendix a ) . more generally , the mapping can be further generalized to the system of a many - level ( single - level ) spinless quantum dot with ohmic dissipation coupled to two chiral luttinger liquid leads with luttinger parameter @xmath86 , giving rise to the effective luttinger liquid parameter @xmath94 defined as ( see eq . ( [ ktilde ] ) in appendix a ) : @xmath95 for a many - level spinless quantum dot and @xmath96 for a spinless quantum dot with a single resonant level . details of the mapping is given in appendix a. we may compute the non - equilibrium current operator in the effective models through the mappings . we will first compute the current operator within the effective anisotropic kondo model as it is the main focus of this paper . from the mapping described in sec . _ , we can establish the invariance of the net charge on the resonant level upon the mapping : @xmath97 , where @xmath98 represents the charge in lead @xmath99 , whereas @xmath100 represents the charge in the effective lead @xmath101 . this allows us to check that the averaged currents within the keldysh formalism@xcite are the same in the original and in the effective kondo model ( see appendix b for details ) : @xmath102 \nonumber \\ & = & { \it i } j_{\perp}^{(1 ) } ( s_{lr}^{- } s^{+ } - s_{rl}^{+ } s^{- } ) - ( 1\rightarrow 2 , l\rightarrow r).\nonumber \\ \label{currenti}\end{aligned}\ ] ] thus , the current @xmath103 can be computed from the kondo model due to the invariance of the average current upon the mapping mentioned above . note that through the various mappings mentioned above , it is straightforward to see that the current operator for other related models resonant level coupled to fqhe leads and dissipative resonant level ( both small and large in size ) coupled to interacting luttinger liquid leads take exactly the same form as shown in eq . ( [ currenti ] ) . now , we employ the non - equilibrium rg approach to the effective kondo model@xcite in eq . ( [ hkondo ] ) . in this approach , the anderson s poor - man scaling equations are generalized to non - equilibrium rg equations by including the frequency dependence of the kondo couplings and the decoherence due to the steady - state current at finite bias voltage@xcite . for the sake of simplicity , we assume that the resonant level ( quantum dot ) is symmetrically coupled to the right and to the left lead , @xmath104 ( or @xmath105 ) . we will discuss in appendix c the more general case with @xmath46 . the dimensionless kondo couplings as a function of frequency @xmath106 exhibit an extra symmetry due to the particle - hole symmetry of the effective kondo model : @xmath107 where @xmath108 is the initial value , with @xmath109 being the density of states per spin of the conduction electrons . here , we suppress the upper script symbol @xmath110 in the kondo couplings since we will now focus on the non equilibrium case v not zero . we obtain@xcite : @xmath111 ^{2}\theta _ { \omega + \frac{\beta v}{2 } } , \nonumber \\ \frac{\partial g_{\perp } ( \omega ) } { \partial \ln d } & = & -\!\!\!\sum_{\beta = -1,1}\!\!\!g_{\perp } \left ( \frac{\beta v}{2}\right ) g_{z}\left ( \frac{\beta v}{2}\right ) \theta _ { \omega + \frac{\beta v}{2 } } , \label{gpergz}\end{aligned}\ ] ] where @xmath112 , @xmath113 is the running cutoff . here , @xmath114 is the decoherence ( dephasing ) rate at finite bias which cuts off the rg flow@xcite . in the kondo model , @xmath114 corresponds to the relaxation rate due to spin flip processes ( which are charge flips in the original model ) , defined as the broadening @xmath115 of the dynamical transverse spin susceptibility @xmath116 in the effective kondo model@xcite : @xmath117 with @xmath116 being the time fourier transform of the spin susceptibility @xmath118\rangle = { \it i}\theta(t)\langle [ f^{\dagger}_{\downarrow}(t ) f_{\uparrow}(t ) , f^{\dagger}_{\uparrow}(0 ) f_{\downarrow}(0)]\rangle$ ] , and @xmath119 being @xmath120 . here , we take the pesudo - fermion representation of the spin operators @xmath121 with @xmath122 being the pesudo - fermion operator and @xmath123 being the pauli matrices@xcite . + in the original model the decoherence rate @xmath114 corresponds to the charge flip rates , defined as the broadening @xmath124 of the resonant - level ( @xmath125electron ) green s function ( or equivalently the imaginary part of the resonant - level self - energy @xmath126 ) : @xmath127 where the self - energy @xmath128 of the @xmath125electron green s function is defined via : @xmath129 with @xmath130 being the fourier transform of the resonant - level green s function @xmath131\rangle$ ] . these two definitions for @xmath114 agree with each other with the proper identification : @xmath132 , @xmath133 . at the transition as a function of bandwidth cutoff @xmath134 ( in unit of @xmath135 ) ; the bare couplings are @xmath136 ( in unit of @xmath135 ) . we have set @xmath137 ( in unit of @xmath135 ) . the decoherence rate @xmath114 is around @xmath138.,width=302 ] note that these rg equations in the present context were already discussed in the short ref . , but now we will elaborate the methodology . the configurations of the system out of equilibrium are not true eigenstates , but acquire a finite lifetime . the spectral function of the fermion on the level is peaked at @xmath139 , and therefore we have @xmath140 on the right hand side of eq . ( [ gpergz ] ) . other kondo couplings are not generated . from ref . via the fermi s golden rule of the spin - flip rates @xmath114 in the kondo model , we identify : @xmath141 , & & \nonumber\end{aligned}\ ] ] where @xmath142 is the fermi function . here , @xmath143 for the @xmath144 terms while @xmath145 for the @xmath146 terms with @xmath147 , @xmath148 being @xmath49 or @xmath50 . we have introduced the occupation numbers @xmath149 for up and down spins satisfying @xmath150 and @xmath151 . in the delocalized phase , we get @xmath152 , in agreement with the quantum boltzmann equation @xcite . at the kt transition , we can use that @xmath153 from the symmetry of the kondo model and that @xmath154 . finally in the localized phase , we have @xmath155 , and @xmath156 satisfies @xmath157 ( see refs . ) , which remains true at a finite bias voltage . at @xmath158 versus @xmath106 across the kt transition . the bias voltage is fixed at @xmath159 . @xmath160 develops a peak ( dip ) at @xmath161 in the delocalized ( localized ) phase , respectively . at @xmath162 ( vertical dotted lines ) , @xmath160 shows peaks ( for localized phase ) or dips ( for delocalized phase ) . note that @xmath160 weakly depends on @xmath106 for @xmath163 , @xmath164.,width=302 ] following the scheme of ref . , we solve eqs . ( [ gpergz ] ) and ( [ gammav ] ) self - consistently . first , we compute @xmath165 for a given cutoff @xmath134 . we then substitute the solutions back into the rg equations to get the general solutions for @xmath166 at finite @xmath0 , and finally extract the solutions in the limit @xmath167 . when the cutoff @xmath134 is lowered , the rg flows are not cutoff by @xmath0 but they continue to flow for @xmath168 until they are stopped for @xmath169 . + in fig.[rgflow ] we show a typical rg flow of @xmath170 at the kt transition as a function of bandwidth @xmath134 with the analytical approximation : @xmath171 for @xmath172 , @xmath173 for @xmath174 , and @xmath175 for @xmath176 . here , @xmath37 , with @xmath38 being the ultraviolet cutoff , and @xmath134 is the running cutoff scale set by the rg scaling equations for @xmath177 . this clearly shows that the rg flow of @xmath170 is stopped at @xmath114 , a much lower energy scale than @xmath0 . + note that the charge ( or pseudospin ) decoherence rate @xmath114 is a function of frequency , @xmath160 in the more general and rigorous functional renormalization group ( frg ) framework@xcite . here , @xmath178 within frg . nevertheless , we find @xmath179 at @xmath158 depends weakly on @xmath106 and can be well approximated by its value at @xmath161 , @xmath180 ( see fig . [ gamma_omega ] ) . we have checked that the non - equilibrium current @xmath181 and conductance @xmath182 obtained from this approximation ( @xmath180 ) agrees very well with that from the more rigorous frg approach based on the frequency - dependent decoherence rate @xmath183 ( see eq . ( [ gamma ] ) below ) as a consequence of the fact that the current and conductance are integrated quantities over the frequencies , and they are insensitive to weak frequency - dependence of @xmath114 . in fig . [ gamma - d ] we show the rg flow of the decoherence rate @xmath184 as a function of @xmath134 , using the same parameter as in fig . 1 . one observes that @xmath114 tends to a finite value as @xmath185 . the inset shows @xmath114 as a function of @xmath0 ( see also section vi ) . at @xmath158 versus @xmath134 ( in unit of @xmath135 ) at a fixed bias voltage @xmath186 ( vertical dashed line ) at the kt transition with bare kondo couplings @xmath187 . under rg , @xmath114 approaches a constant value as @xmath185 : @xmath188 . inset : @xmath114 as a function of @xmath0 for the same bare kondo couplings.,width=302 ] note that , unlike the equilibrium rg at finite temperatures where rg flows are cutoff by temperature @xmath4 , here in non - equilibrium the rg flows will be cutoff by the decoherence rate @xmath114 , an energy scale typically much higher than @xmath4 , but much lower than @xmath0 , @xmath189 . moreover , @xmath190 is a non - linear function in @xmath0 . ( for example , at the kt transition , @xmath191 ^ 2 $ ] . ) . the unconventional properties of @xmath190 lead to a non - equilibrium conductance ( @xmath182 ) distinct from that in equilibrium ( @xmath192 ) near the kt transition@xcite . in contrast , the equilibrium rg will lead to approximately frequency independent couplings , ( or `` flat '' functions @xmath193 ) . + notice that the mapping mentioned above works near the kt transition , @xmath194 . however , for a general case deep in the localized phase , the effective kondo couplings aquire an additional phase @xmath195 where the more general form of @xmath196 and its phase @xmath197 are derived and defined in eq . ( [ tilde - h - t ] ) of appendix a .. this results in a nonzero bare scaling dimension@xcite for @xmath196 , @xmath198 = \frac{1}{2}(\sqrt{2 } - \frac{1}{\sqrt{k}})^2 = 1-\sqrt{\frac{2}{k } } + \frac{1}{2k}$ ] . this slightly modifies the non - equilibrium rg scaling equations to the following form : @xmath111 ^{2}\theta _ { \omega + \frac{\beta v}{2 } } \nonumber \\ \frac{\partial g_{\perp } ( \omega ) } { \partial \ln d } & = & -\!\!\!\sum_{\beta = -1,1}\!\!\ ! [ \frac{1}{2}(1-\sqrt{\frac{2}{k } } + \frac{1}{2k } ) g_{\perp } \left ( \frac{\beta v}{2}\right ) \nonumber \\ & + & g_{\perp } \left ( \frac{\beta v}{2}\right ) g_{z}\left ( \frac{\beta v}{2}\right ) ] \theta _ { \omega + \frac{\beta v}{2 } } \label{rg2}\end{aligned}\ ] ] where the linear term @xmath199 in eq . ( [ rg2 ] ) for @xmath200 comes from the bare scaling dimension of @xmath196 terms mentioned above , and it vanishes in the limit of @xmath201 , as expected . in fact , this term applies to the three models ( case ( i ) , ( ii ) and ( iii ) ) through the mappings . note that the above scaling equations may be cast in the same form as in eq . ( [ gpergz ] ) through redefinition of the coupling @xmath202 : @xmath203 . all the previous results remain valid upon the above shift of @xmath202 . + at the transition at various bias voltages @xmath0 ( in unit of @xmath135 ) ; the bare couplings are @xmath136 ( in unit of @xmath135 ) . the arrows give the values of @xmath204 at these bias voltages . , width=283 ] in the section , we present our results for non - equilibrium conductance . all explicit results will be obtained for the kt transition point and the localized phase , but not for the delocalized phase . at the kt transition , we both numerically and analytically solve eqs . ( [ gpergz ] ) and ( [ gammav ] ) ( in the limit of @xmath167 ) . in particular , the approximated analytical solution within the approximation @xmath205 due to @xmath206 is obtained : @xmath207 } \\ & + & \theta ( v-|\omega-\beta v/2| ) \times \nonumber \\ & & \left [ \frac{1}{\ln[\mathcal{d}^{2}/ v \max(|\omega-\beta v/2|,\gamma ) ] } - \frac{1}{4 \ln\frac{\mathcal{d}}{v}}\right ] . \nonumber\end{aligned}\ ] ] the solutions at the transition ( denoted @xmath208 and @xmath209 ) are shown in fig . [ gpergzfig ] . since @xmath210 decreases under the rg scheme , the effect of the decoherence leads to minima ; the couplings are severely suppressed at the points @xmath211 . we also check that @xmath212 . + from the keldysh calculation up to second order in the tunneling amplitudes , the current reads : @xmath213 - ( l \leftrightarrow r ) . \nonumber\end{aligned}\ ] ] at @xmath158 , it simplifies as @xmath214 . then , we numerically evaluate the non - equilibrium current . the differential conductance is obtained as @xmath215 . the @xmath158 results at the kt transition are shown in fig . [ i ] and fig . [ icr ] . ) fits well with the numerical result over several decades ( from @xmath216 to @xmath217 , see inset ) . however , it starts to deviate from the numerical result at higher bias voltages.,width=283 ] first , it is instructive to compare the non - equilibrium current at the transition to the ( nave ) approximation : @xmath218 ^ 2 \approx \frac{\pi}{8 } \frac{v}{(\ln^2({t_d}/{v}))}.\end{aligned}\ ] ] as shown in fig . [ i ] , our numerically obtained non - equilibrium current fits well with the above analytical approximation for @xmath219 . however , it starts to deviate from its numerically obtained values for higher bias voltages @xmath220 . this deviation is due to the fact that the equilibrium form of the conductance at the transition is obtained by treating @xmath221 a flat function within @xmath222 : @xmath223 . we have checked that the equilibrium coupling @xmath224 indeed corresponds to @xmath225 , therefore the transport recovers the expected equilibrium form for @xmath226 . however , since @xmath200 is not a flat function for @xmath227 ( it has two minima at @xmath228 ) , with increasing @xmath0 ( say for @xmath229 ) the non - equilibrium current exhibits a distinct behavior due to the frequency dependence of the coupling . in fact , the more accurate approximate expression for the non - equilibrium current at the transition is found to be : at the kt transition . @xmath230 is the equilibrium conductance at the transition for @xmath231 : @xmath232 with the bare couplings @xmath233 . , width=283 ] @xmath234 ^ 2 \right ) \\ & + & \frac{\pi v}{2 } \left ( ( 1-\frac{\pi}{4 } ) \left[g_{\perp , cr}(\omega = v/2)\right]^2 \right ) , \nonumber \label{icr - app}\end{aligned}\ ] ] where @xmath235 ) . ( a ) @xmath236 at low bias follows the equilibrium scaling ( dashed lines ) . ( b ) the conductance @xmath237 is a function of @xmath238 where we have defined @xmath239 and @xmath240.,width=321 ] here , we have treated @xmath241 within the interval @xmath242 as a semi - ellipse . + as demonstrated in fig . [ icr ] , the conductance @xmath236 obtained via the approximation in eq . ( [ ivapprox ] ) fits very well with that obtained numerically over the whole range of @xmath243 . in the low - bias @xmath226 ( equilibrium ) limit , since @xmath244 , we have @xmath245 ; therefore the scaling of @xmath246 is reminiscent of the equilibrium expression in eq . ( [ gteq ] ) , @xmath247 . this agreement between equilibrium and non - equilibrium conductance at low @xmath0 persists up to a crossover scale @xmath248 ( determined for the parameters used in fig . [ icr ] ) . at larger biases , the conductance shows a unique non - equilibrium profile ; see eq . ( [ ivapprox ] ) . we find an excellent agreement of the non - equilibrium conductance obtained by three different ways pure numerics , analytical solution eq . ( [ gsolution ] ) and the approximation in eq . ( [ ivapprox ] ) . for large bias voltages @xmath249 , since @xmath250 approaches its bare value @xmath251 , the non - equilibrium conductance increases rapidly and reaches @xmath252 . note that the non - equilibrium conductance is always smaller than the equilibrium one , @xmath253 , since @xmath254 . additionally , in the delocalized phase for @xmath255 , the rg flow of @xmath251 is suppressed by the decoherence rate , and @xmath256 ( ref . ) . + in the localized phase , we first solve the equilibrium rg equations of the effective kondo model analytically , resulting in @xmath257 where @xmath258 . we introduce the energy scale @xmath259 ( which vanishes at the kt transition ) such that @xmath260 for @xmath261 , leading to @xmath262 . + ) . at large bias voltages @xmath0 , the non - equilibrium conductance @xmath236 ( solid lines ) is distinct from the equilibrium form ( dashed lines ) . the dot - dashed lines stem from an analytical approximation via eq . ( [ ivapproxloc]).,width=311 ] at a finite bias , we first solve for the self - consistent non - equilibrium rg equations both analytically and numerically , resulting in : @xmath263 + \frac{b}{2c } [ \gamma^c \sqrt{c^2+b^2\gamma^{2c } } \\ \nonumber & - & v^c \sqrt{c^2+b^2v^{2c } } ] \\ \nonumber & + & \frac{c}{2 } \ln\left [ \frac{b\gamma^c + \sqrt{c^2 + b^2\gamma^{2c}}}{bv^{c } + \sqrt{c^2+b^2 v^{2c}}}\right ] \\ \nonumber & + & \frac{c}{2 } \ln\left [ \frac{a v^{2c } + \sqrt{c^2+a^2 v^{4c}}}{a + \sqrt{c^2+a^2}}\right ] , \end{aligned}\ ] ] @xmath264 + \frac{b}{c } [ ( \frac{v}{2})^c \sqrt{c^2+b^2(\frac{v}{2})^{2c}}\\ \nonumber & - & v^c \sqrt{c^2+b^2 v^{2c}}]\\ \nonumber & + & \frac{c}{2 } \ln\left [ \frac{b(\frac{v}{2})^c + \sqrt{c^2+b^2 ( \frac{v}{2})^{2c}}}{bv^{c } + \sqrt{c^2+b^2 v^{2c } } } \right ] \\ \nonumber & + & \frac{c}{2 } \ln\left [ \frac{a v^{2c } + \sqrt{c^2 + a^2 v^{4c}}]}{[a + \sqrt{c^2+a^2 } } \right],\end{aligned}\ ] ] and similarily we get @xmath265 \\ \nonumber & + & \frac{b^2}{2c } v^{2c } [ 1-(\frac{\gamma}{v})^{2c } ] , \end{aligned}\ ] ] at the kt transition ( same unit as in fig . [ icr ] ) . ( a ) . for @xmath266 , the conductance follows the non - equilibrium scaling @xmath267 . ( b ) . for @xmath268 , now the conductance follows the equilibrium scaling @xmath269.,width=340 ] @xmath270\\ \nonumber & + & \frac{b^2}{c}v^{2c } [ 1 - 2^{-2c}],\end{aligned}\ ] ] where we unambiguously identify @xmath271 , @xmath272 ; in this expression , @xmath0 and @xmath114 have been normalized to @xmath135 . + the non - equilibrium current in the localized phase @xmath273 is obtained via the same approximation leading to eq . ( [ icr - app ] ) at the kt transition : @xmath274 ^ 2 \right ) \\ & + & \frac{\pi v}{2 } \left ( ( 1-\frac{\pi}{4 } ) \left[g_{\perp , loc}(\omega = v/2)\right]^2 \right ) . \nonumber \end{aligned}\ ] ] as shown in fig . [ gva ] , we numerically obtain the non - equilibrium conductance in the localized phase . for very small bias voltages @xmath226 , we find that the conductance reduces to the equilibrium scaling : @xmath275 ( see fig . [ gva ] ( a ) and ( b ) ) . for @xmath276 and @xmath277 , we get that the exponent @xmath278 , in perfect agreement with that obtained in equilibrium at low temperatures : @xmath279 ( ref . ) . at higher bias voltages @xmath280 , the conductance now follows a unique non - equilibrium form ( consult fig . [ gvb ] ) whose qualitative behavior is similar to that at the kt transition . our non - equilibrium conductance obtained numerically in this phase is in very good agreement with that from the above approximated analytical solutions in eq . [ ivapproxloc ] ( see fig . [ gvb ] ) . + ( in unit of @xmath135 ) . ( a ) . for @xmath266 , the conductance follows the non - equilibrium scaling @xmath281 . ( b ) . for @xmath282 , now the conductance follows the equilibrium scaling @xmath269.,width=321 ] we have also analyzed the finite temperature profile of the non - equilibrium conductance at the transition and in the localized phase . we distinguish two different behaviors . at the kt transition , for @xmath283 , the conductance @xmath284 exhibits the same non - equilibrium form as @xmath158 , @xmath182 ( see fig . [ gvt](a ) ) ; while as for @xmath268 it saturates at the value for the equilibrium conductance ( @xmath285 ) at finite temperatures ( see fig . [ gvt](b ) . in the localized phase , while for @xmath268 the conductance saturates at @xmath286 ( fig . [ gvt_loc](a ) ) , for @xmath3 , however , @xmath284 exhibits universal power - law scaling : @xmath287 ( see fig . [ gvt_loc](b ) ) . this universal power - law scaling behavior in @xmath284 looks qualitativly similar to that from the recent experiment on the transport through a dissipative resonant level in ref . . however , these two power - law behaviors in conductance at a finite bias and temperature are different in their orgins : the authors in ref . studied the quantum critical behavior of a dissipative resonant level in the regime of the delocalized phase ( @xmath288 ) . as the resonant level is detuned from the fermi level , the system at low temperatures exhibits power - law scaling in conductance at a large bias voltage @xmath3 : @xmath289 with @xmath290 . they showed further that this behavior is equivalent to that for a single - barrier tunneling of electrons through a luttinger liquid . by contrast , the luttinger - liquid - like power - law scaling in @xmath284 ( see eq . ( [ gvt - loc ] ) ) we find here is the generic feature of a dissipative resonant level in the localized phase ( @xmath291 ) , which has not yet been explored experimentally . therefore , our theoretical predictions on the nonequilibrium transport at a dissipative quantum phase transition offer moltivations for further experimental investigations in the regime of our interest . the above two qualitatively different behaviors in conductance for @xmath268 and @xmath3 crossover at @xmath292 . note that similar behavior has been predicted in a different setup consisting of a magnetic single electron transistor ( set ) in ref . where a true quantum critical point separates the kondo screened and the local moment phases . the solid lines represent conduction electron propagators ; the dashed lines denote the pseudo - fermion propagators . the current vertex functions @xmath293 are denoted by the shaded squares.,width=264 ] in addition to non - equilibrium current and conductance near the localized - delocalized transition addressed above , further insight on the phase transition can be obtained from the current fluctuations ( or noise ) . the zero frequency shot noise has been used to probe the fractional charge of quasiparticle excitations in fqhe state tunnelings@xcite . however , even more useful information can be found in the finite - frequency ( ff ) current noise , which can be used to probe the crossover between different quantum statistics of the quasiparticles@xcite . recently , there has been theoretical studies on the ff current noise of a non - equilibrium kondo dot@xcite . so far , these studies have not been extended to the non - equilibrium ff current noise of a dissipative quantum dot . + to address this issue , we combine recently developed functional renormalization group ( frg ) approach in refs . and the real - time frg approach in ref . . within our frg approaches , as the system moves from the delocalized to the localized phase , we find the smearing of the dips in current noise spectrum for frequencies @xmath162 ; more interestingly , we find a peak - to - dip crossover in the ac conductance at @xmath162 . these features are detectable in experiments and can serve as alternative signatures ( besides conductance ) of the qpt in the dissipative resonant level quantum dot . + first , via the above mapping , the current through the dissipative resonant level quantum dot is given by the transverse component of the current @xmath294 in the effective anisotropic kondo model as shown in eq . ( [ currenti])@xcite . following the real - time rg approach in ref . the keldysh current operator through the left lead in the effective kondo model ( via a generalization of eq . ( [ currenti ] ) ) is given by : @xmath295 : ( the squares ) . the solid lines represent conduction electron propagators ; the dashed lines denote the pseudo - fermion propagators . here , the kondo couplings @xmath296 are denoted by the circles . , width=264 ] @xmath297 \end{aligned}\ ] ] with @xmath298 , @xmath299 , @xmath300 . here , @xmath301 is the left current vertex matrix with bare ( initial ) matrix elements : @xmath302 , @xmath303 , @xmath304 , @xmath305 , and @xmath306 being the upper and lower keldysh contour , respectively . the emission component of the non - equilibrium ff noise of a kondo quantum dot , @xmath307 , is given by the current - current correlator : @xmath308 similarly , the absorption part of the noise is defined as : @xmath309 . note that the current operator @xmath295 is non - local in time under rg ; the current vertex function @xmath301 therefore acquires the double - time structure : it keeps track of not only the times electrons enter ( @xmath310 ) and leave ( @xmath311 ) the dot , but also the time @xmath18 at which the current is measured@xcite . the double - time structure of the current operator automatically satisfies the current conservation : @xmath312 ( ref . ) . at zero temperature in the delocalized phase with bare kondo couplings being @xmath313 , @xmath314 . the bias voltage is fixed at @xmath159.,width=321 ] the frequency - dependent current noise @xmath315 is computed via the second - order renormalized perturbation theory ( see diagram in fig . [ noise_dia ] ) . note that due to the double - time structure of the current vertex function @xmath316 , in the fourier ( frequency ) space , @xmath317 has a two - frequency structure ; it depends on the incoming ( @xmath318 ) and outgoing ( @xmath319 ) frequencies of the electron ( see fig . [ noise_dia ] ) . the result reads : @xmath320 where the correlator @xmath321 is computed by the diagram in fig . [ noise_dia ] : @xmath322^<,\nonumber \\ \chi_{\alpha\beta}(\omega,\omega ) & = & \int \frac{d\epsilon}{2\pi } \hat{g}_{\alpha}(\epsilon)\hat{g}_{\beta}(\epsilon+\omega+\omega)\nonumber \\ & \times & l_{\alpha\beta}^{\perp}(\epsilon+\omega , \epsilon ) l_{\beta\alpha}^{\perp}(\epsilon , \epsilon+\omega),\nonumber \\ \chi_f(\omega ) & = & \int \frac{d\nu}{2\pi } \hat{g}_f(\nu ) \hat{g}_f(\nu+\omega ) , \end{aligned}\ ] ] where @xmath323 is the green s function in @xmath324 keldysh space , and its lesser and greater green s function are related to its retarded , advanced , and keldysh components by : @xmath325 the lesser ( @xmath326 ) and greater ( @xmath327 ) components of green s function of the conduction electron in the leads and of the quantum dot ( impurity ) are given by : @xmath328 where @xmath329 is the density of states of the leads , @xmath330 is the occupation number of the pseudofermion which obeys @xmath331 , @xmath332 in the delocalized phase and @xmath333 , @xmath334 in the localized phase@xcite . here , the pseudofermion occupation number @xmath335 and the occupation number on the dot @xmath336 are related via @xmath337 ( refs . ) . the renormalized current vertex function @xmath338 and the kondo couplings @xmath200 , @xmath339 are obtained from the non - equilibrium functional rg approaches in ref . and refs . , respectively . carrying out the calculations , the finite - frequency noise spectrum reads : at zero temperature in the localized phase with bare kondo couplings being @xmath313 , @xmath340 . the bias voltage is fixed at @xmath159.,width=321 ] @xmath341 where @xmath342 is the fermi function of the lead @xmath343 given by @xmath344 . the symmetrized noise spectrum reads : @xmath345\ ] ] with the relation between emission and absorption parts of the noise spectrum in frequency space @xmath346 being used . the frequency - dependent kondo couplings @xmath347 and current vertex functions @xmath348 are obtained self - consistently within the frg approaches , which can be divided into two parts . first , the kondo couplings @xmath347 are solved via eq . ( [ gpergz])@xcite together with the generalized frequency - dependent dynamical decoherence rate @xmath160 appearing in @xmath349 in eq . ( [ gpergz ] ) . here , @xmath160 is obtained from the imaginary part of the pseudofermion self energy@xcite : @xmath350\nonumber \\ & + & g_{z}(\epsilon+\omega ) g_{z}(\epsilon ) [ f^{l}_{\epsilon}-f^{l}_{\epsilon+\omega}]\nonumber \\ & + & ( l\rightarrow r ) . \label{gamma}\end{aligned}\ ] ] note that the zero - frequency decoherence rate @xmath184 corresponds to the decoherence rate @xmath114 obtained in eq . ( [ gammav])@xcite . we have solved the rg equations eq . ( [ gpergz ] ) subject to eq . ( [ gamma ] ) self - consistently@xcite . + the solutions for @xmath200 , @xmath351 and @xmath160 close to the kt transition are shown in refs . ( see also fig . [ gamma_omega ] and fig . [ gpergzfig ] ) . as the system goes from the delocalized to localized phase , the features in @xmath200 at @xmath352 undergoes a crossover from symmetric two peaks to symmetric two dips , while the symmetric two peaks in @xmath353 still remain peaks . the finite - frequency non - equilibrium decoherence rate @xmath160 monotonically increases with increasing @xmath106 , it shows logarithmic sigularities at @xmath354 in the delocalized phase@xcite . as the system moves to the localized phase , the overall magnitude of @xmath160 decreases rapidly and the singular behaviors at @xmath355 get smeared out@xcite . + at zero temperature versus @xmath106 across the kt transition . the bias voltage is fixed at @xmath159 . inset : @xmath356 at zero temperature versus @xmath106 normalized to @xmath357.,width=283 ] next , following ref . , we generalize the rg scaling equation for the general current vertex function @xmath358 for the anisotropic kondo model ( see diagrams in fig . [ current ] and also in fig . 1 of ref . ) . the rg scaling equations for the general vertex functions @xmath359 can be simplified as : @xmath360 where we make the following identifications : @xmath361 , @xmath362 . similarly , @xmath363 refers to only the transverse component of the current vertex function @xmath293 ; while @xmath364 refers only to the longitudinal part of @xmath365 . here , the frequency - dependent kondo couplings @xmath366 in eq . ( [ lij ] ) are obtained from eq . ( [ gpergz ] ) and eq . ( [ gamma ] ) . note that the scaling equations for @xmath293 via ref . can also be expressed within the rg approach in ref . via a straightforward generalization by allowing for the two - frequency dependent vertex functions @xmath293 where @xmath367 refers to the incoming ( outgoing ) frequency ( see fig . [ noise_dia ] and fig . [ current ] ) . + defined in eq . ( [ gaceq ] ) versus @xmath106 across the kt transition . ( b ) @xmath368 at zero temperature versus @xmath106 normalized to @xmath369 . the bias voltage is fixed at @xmath159.,width=321 ] we solved the self - consistent rg scaling equations eq . ( [ lij ] ) for the current vertex functions with the help of the solutions for the renormalized kondo couplings via eq . ( [ gpergz ] ) and eq . ( [ gamma ] ) . the typical results at zero temperature are shown in fig . [ l12_deloc ] and fig . [ l12_loc ] ; they exhibit the following symmetry : @xmath370 . note that since the initial conditions for the current vertex function have the following structures : @xmath371 , @xmath372 , we find @xmath373 . in the delocalized ( kondo ) phase , a sharp peak is developed in @xmath374 for @xmath375 ; while as a small dip is formed for @xmath376 . meanwhile , in general @xmath374 is maximized at @xmath377 for fixed @xmath378 . this agrees perfectly with the result in ref . . in the localized phase , however , we find the opposite : @xmath374 develops a sharp dip at @xmath375 ; and it is minimized @xmath377 for fixed @xmath378 . the peak - dip structure of the current vertex function @xmath379 plays a crucial role in determining the noise spectrum both in delocalized and in the localized phases . + substituting the numerical solutions for @xmath293 and @xmath380 into eq . ( [ snu ] ) , we get the zero - temperature ff noise @xmath315 . the results at zero temperature are shown in fig . [ s_scaled ] . first , the overall magnitude of @xmath315 decreases rapidly as the system crosses over from the delocalized to the localized phase . this can be understood easily as the current decreases rapidly in the crossover , leading to a rapid decrease in the magnitude of noise . for @xmath381 , @xmath315 in both phases increases monotonically with increasing @xmath106 due to the increase of the photon emission at higher energies@xcite . for @xmath382 , however , it changes from a peak to a dip centered at @xmath383 as the system crosses overs from delocalized to localized phase ( see fig . [ s_scaled ] ) . at @xmath354 , @xmath315 exhibits a dip ( minima ) in the delocalized phase , a signature of the non - equilibrium kondo effect ; while as the system crosses over to the localized phase the dips are gradually smeared out and they change into a `` kink''-like singular point at @xmath384 , connecting two curves between @xmath385 and @xmath386 . + and ( b ) @xmath114 as a function of @xmath387 near the kt transition . , width=340 ] we furthermore computed the non - equilibrium ac conductance at zero temperature@xcite : @xmath388 across the transition . note that @xmath389 corresponds to the non - equilibrium differential conductance . as shown in fig . [ gac ] ( a ) , in the delocalized phase the splitted peaks in @xmath390 at @xmath355 are signatures of the kondo resonant at finite bias , and are consistent with the dips at seen in the noise spectrum . as the system moves to the localized phase , the overall magnitudes of @xmath390 as well as the pronounced splitted kondo peaks at @xmath355 get suppressed ; they change into dips deep in the localized phase ( see fig . [ gac ] ( b ) ) . in response to this change in the splitted kondo peaks , the overall shape of @xmath391 shows a dip - to - hump crossover near @xmath161 . note that the suppression of the kondo peaks for @xmath390 at @xmath355 corresponds to the smearing of the dips at @xmath355 shown in the noise spectrum @xmath315 ( see fig . [ s_scaled ] ) . the above evolution in the noise spectrum matches well with the non - equilibrium transport properties studied in refs . , and can serve as alternative signatures of the localized - delocalized transition in future experiments . we would like to make a few remarks before we conclude . firstly , the distinct non - equilibrium scaling behavior seen here is in fact closely tied to the non - trivial ( non - linear ) @xmath0 dependence of the decoherence rate @xmath392 which cuts off the rg flow ( see fig . [ gammavfig ] ( a ) and ( b ) ) . the decoherence rate @xmath114 near the transition clearly plays a very different role as compared to the temperature near the transition . in particular , at @xmath158 we find that @xmath393 is a highly non - linear function in @xmath0 , resulting in the observed deviation of the non - equilibrium scaling from that in equilibrium . in fact , we can obtain the analytical form via the approximation in eq . ( [ ivapprox ] ) and eq . ( [ ivapproxloc ] ) . at the kt transition , @xmath394 shows a logarithimic decrease as @xmath0 decreases ( see eqs . ( [ icr - app ] ) , ( [ gcr - app ] ) ) ; while in the localized phase it exhibits a combined power - law and logarithmic dependence on @xmath0 ( see eqs . ( [ gp - loc - app - a ] ) , ( [ gp - loc - app - b ] ) , ( [ gz - loc - app - a ] ) , ( [ gz - loc - app - b ] ) , ( [ ivapproxloc ] ) ) . + by contrast , the equilibrium decoherence rate @xmath395 shows a clear power - law behavior in the localized phase at low temperatures , @xmath396 ( see eq . ( [ gt - loc ] ) ) : @xmath397 which is consistent with the prediction made in ref . for the electron lifetime in luttinger liquids . meanwhile , at the kt transition and in the localized phase , since @xmath398 , the rg flow for @xmath177 are cut off at an energy scale @xmath114 much lower than @xmath0 , leading to smaller renormalized couplings @xmath177 in magnitude compared to their corresponding equilibrium values @xmath399 , @xmath400 . this results in smaller conductance than that in equilibrium , @xmath401 . + secondly , it is of fundamental importance to study further the possible scaling behaviors in non - equilibrium dynamical quantities near the transition , such as the @xmath402 scaling in dynamical charge susceptibility at the kt transition and in the localized phase . in particular , the question has been raised on the existence of the concept of `` effective temperature '' that allows one to extend the fluctuation - dissipation theorem in equilibrium to the non - equilibrium ( non - linear ) regime@xcite . it is also interesting to address the crossover between delocalized phase with @xmath403 where @xmath29 to kt point with @xmath404 and further to the localized phase with power - law conductance @xmath405 . to date , the full crossover function of the conductance is not known yet . further study is therefore needed to investigate these issues . in summary , we have investigated the non - equilibrium transport at a qpt using a standard nano - model , the dissipative resonant level model . by employing an exact mapping onto the anisotropic kondo model and by applying a controlled energy - dependent rg and functional rg approaches to our model system we have calculated the renormalized coupling functions @xmath347 , the decoherence rate @xmath114 , the current @xmath103 , differential conductance @xmath284 , and the current noise spectrum @xmath315 . for @xmath1 , the conductance @xmath406 follows the equilibrium behavior ; by increasing @xmath0 , the frequency - dependence of the couplings begins to play an important role and therefore we systematically find scaling behavior of the non - equilibrium conductance very distinct from that of the equilibrium counterpart . we have also analyzed the finite temperature profile of @xmath284 at the transition as well as in the localized phase and found that the conductance shows different behaviors for @xmath3 and @xmath268 ; it exhibits @xmath407 scaling behavior for @xmath282 . + regarding transport properties of our system near the transition , the role played by the bias voltage is very different from that played by the temperature . the key to these very different behaviors lies in the fact that the non - equilibrium charge ( or effective spin ) decoherence rate , which serves as a cutoff for the rg flows of the kondo couplings , is a highly non - linear function of the bias voltage . further investigations are needed to address the full crossover function in conductance as well as the scaling behaviors of the dynamical quantities near the transition in a search for the existence of the `` effective temperature '' that allows one to generalize the equilibrium fluctuation - dissipation theorem to the non - equilibrium regime . furthermore , we provide signatures of the localized - delocalized transition in the finite - frequency current noise spectrum and the ac conductance . our results have a direct experimental relevance for dissipative two - level systems ; moreover , they are applicable for describing non - equilibrium transport of a resonant level coupled to interacting chiral luttinger liquid generated by fractional quantum hall edge states via the mappings discussed in appendix a. finally , our model system has direct relevance for the recent experiment in a quantum dot coupled to resistive environment as shown in ref . . our work motivates future experimental as well as theoretical investigations on dissipative quantum phase transitions in nanosystems . + we are grateful to d. goldhaber - gordon , p. moca , g. zarand , and s. kirchner for stimulating discussions , and to r.t . chang and k.v.p . latha for technical support . we acknowledge the generous support from the nsc grant no.98 - 2112-m-009 - 010-my3 , no.101 - 2628-m-009 - 001-my3 , the moe - atu program , the cts of nctu , the ncts of taiwan , r.o.c . ( c.h.c . ) , the department of energy in usa under the contract de - fg02 - 08er46541 ( k.l.h . ) , the german - israeli foundation via g 1035 - 36.14/2009 , the dfg via for 960 ( m.v.,p.w . ) , grk 1621 ( m.v . ) , and the center for functional nanostructures cfn ( p.w . ) , the u.s . department of energy , office of basic energy sciences , division of materials sciences and engineering under award de - sc0002765 ( g.f . ) . has benefitted from the visiting programs of kitp , ictp , and mpi - pks . k.l.h . acknowledges kitp for hospitality . g.f . thanks ncts , nctu of taiwan r.o.c . for hospitality during the visit . in this appendix , we provide detailed derivations on various mappings mentioned in sec . ii . via bosonization and refermionization techniques , the three mappings described below will follow one from the other , but there are a few technical details that will change . we describe in details on the mapping of dissipative resonant level model in eq . ( [ rlm ] ) onto anisotropic kondo model in eq . ( [ hkondo ] ) . our goal is to connect the parameters of these two equations in the main text . we first start from eq . ( [ rlm ] ) : @xmath408 where @xmath6 is the ( real - valued ) hopping amplitude between the lead @xmath7 and the quantum dot , @xmath8 and @xmath9 are electron operators for the ( fermi - liquid type ) leads and the quantum dot , respectively . @xmath10 is the chemical potential applied on the lead @xmath7 ( @xmath0 denotes the bias voltage ) , while the dot level is at zero chemical potential . here , @xmath11 are the boson operators of the dissipative bath with an ohmic type spectral density . it proves to be more convenient to re - express the dissipative boson fields @xmath409 and @xmath410 in terms of the canonical fields @xmath411 and @xmath412 as : @xcite : @xmath413 e^{-a |p|/2}\nonumber \\ \hat{\pi}_0(x , t ) & = & \partial_t \hat{\phi}_0(x , t)\nonumber \\\end{aligned}\ ] ] where @xmath414 with @xmath415 being the phonon velocity , and the boson fields @xmath411 and @xmath412 satisfies the commutation relation : @xmath416 = { \it i } \delta ( x - x^{\prime})$ ] . the dissipative boson bath can therefore be re - expressed as : @xmath417 . \label{rlm - diss}\end{aligned}\ ] ] here , the velocity of the boson field @xmath418 is set to be @xmath419 . we start the mapping by bosonizing the fermionic operators in the leads : @xmath420 where we have introduced the ( standard ) klein factors @xmath421 ensuring anti - commutation relations and @xmath88 is a short - distance cutoff ( lattice spacing ) . the fermionic baths of conduction electrons can be re - written as : @xmath422 \label{rlm - leads}\end{aligned}\ ] ] where the fermi velocity of the electrons is set to be @xmath419 . the level on the quantum dot can be mapped onto a pseudo - spin : @xmath423 and @xmath424 ; @xmath79 represent the two leads . the coupling between the dot and the dissipation bath ( @xmath425 term ) can be absorbed in the tunneling part of the hamiltonian through the unitary transformation @xmath426@xcite : @xmath427 with @xmath428 . here , @xmath84 refers to the strength of the coupling between the resonant level and the dissipative boson bath , and we set @xmath429 for simplicity . we can simplify our variables even further by combining the above fields describing the leads and the noise : @xmath430 , @xmath431 , where @xmath432 . note that here @xmath86 may be interpreted as the effective luttinger liquid parameter as the effect of ohmic dissipation on the quantum dot plays a similar role as interactions in the luttinger liquid leads coupled to the dot with the identification @xmath433 . the comined bosonic and fermionic bath @xmath434 can be re - expressed in terms of these new boson fields : @xmath435 \nonumber \\ & + & \frac{1}{2 } \int dx [ ( \partial_x \hat{\phi}_0)^2(x , t ) + \hat{\pi}_0 ^ 2(x , t)]\nonumber \\ & = & \frac{1}{2 } \int dx \sum_{\alpha=1,2}[(\partial_x \tilde{\phi}_{s,\alpha})^2(x , t ) + \hat{\pi}_{s,\alpha}^2(x , t)\nonumber \\ & + & ( \partial_x \tilde{\phi}_{a,\alpha})^2(x , t ) + \hat{\pi}_{a,\alpha}^2(x , t ) ] , \label{tilde - h - bath}\end{aligned}\ ] ] where @xmath436 fields are canonically conjuate to the fields @xmath437 . note that as we shall see below only the fields from symmetric combinations @xmath438 and @xmath439 couple to the tunneling and chemical potential terms , the antisymmetric combinations @xmath440 and @xmath441 are de - coupled from the rest of the hamiltonian . the tunneling and chemical potential parts of the hamiltonian now become : @xmath442 close to @xmath443 ( transition ) , we can map our model onto the 2-channel aniotropic kondo model . after applying the two unitary transformations @xmath444 and @xmath445 , we obtain : @xmath446 s^{- } + h.c . \\ \nonumber & - & ( \sqrt{2 } - \frac{1}{\sqrt{k } } ) ( \partial_x \tilde{\phi}_{s,1 } + \partial_x \tilde{\phi}_{s,2 } ) s_z \label{tilde - h - t}\end{aligned}\ ] ] note that there are additional phase factors @xmath447 in the hopping terms . since we are interested in the physics close to the localized - delocalized transition , i.e. , @xmath448 , we may drop these phase factors in the following analysis . the chemical potential term after the above two transformations now becomes @xmath449 \label{tilde - h - mu}\end{aligned}\ ] ] note that since the hoping @xmath450 and chemical potential @xmath451 terms involve only @xmath438 fields , @xmath440 fields are decoupled from the hamiltonian . now , we can refermionize the bosons and map our transformed hamiltonian @xmath452 onto the anisotropic kondo model in eq . ( [ hkondo ] ) via the following identifications : @xmath453 where @xmath454 is the klein factor for the effective lead @xmath49 and @xmath50 , respectively . + to see the equivalence between these two models , we bosonize eq . ( [ hkondo ] ) and compare it with eq . ( [ maped - rlm ] ) : @xmath455c_{k\gamma \sigma } ^{\dagger } c_{k\gamma \sigma } \nonumber \\ & = & \frac{1}{2 } \int dx \sum_{\alpha = l , r } [ ( \partial_x \phi_\alpha)^2(x , t ) + \pi_\alpha^2(x , t)]\nonumber \\ & - & \frac{v}{2 } \sqrt{\frac{1}{2k } } \sum_{\sigma=\uparrow,\downarrow } [ \partial_x \phi_{l}^{\sigma } - \partial_x \phi_{r}^{\sigma}],\nonumber \\ h_{j_{\perp } } & = & j_{\perp } ^{(1)}s_{lr}^{+}s^{-}+j_{\perp } ^{(2)}s_{rl}^{+}s^{-}+h.c.\nonumber \\ & = & j_{\perp } ^{(1 ) } f_l^{\dagger \uparrow } f_r^{\downarrow } e^{{\it i}\phi_{l}^{\uparrow } - { \it i}\phi_{r}^{\downarrow } } + j_{\perp } ^{(2 ) } f_r^{\dagger \uparrow } f_l^{\downarrow } e^{{\it i}\phi_{r}^{\uparrow } - { \it i}\phi_{l}^{\downarrow } } , \nonumber \\ h_{j_z } & = & \sum_{\gamma = l , r}j_{z}s_{\gamma \gamma } ^{z}s^{z}\nonumber \\ & = & -j_z \sum_{\alpha = l , r } [ \partial_x \phi_{\alpha}^{\uparrow } - \partial_x \phi_{\alpha}^{\downarrow}]\nonumber \\ \label{compare}\end{aligned}\ ] ] with the proper redefinitions of the klein factors : @xmath456 , @xmath457 , and the identifications : @xmath132 , @xmath133 , @xmath458 , @xmath459 , @xmath460 , we finally establish the equivalence between a kondo model with the effective left ( @xmath49 ) and right lead ( @xmath50 ) in eq . ( [ hkondo ] ) and a dissipative resonant level model in eq . ( [ maped - rlm ] ) . we provide details here on the mapping of a dissipative resonant level model eq . ( [ rlm ] ) onto a resonant level coupled to fractional quantum hall edge states ( fqhe ) as shown in eq . ( [ dot - luttinger ] ) . we start from the hamiltonian eq . ( [ dot - luttinger ] ) describing a resonant level coupled to two fqhe states : @xmath461 where the lead term @xmath78 describes two chiral luttinger liquid leads with lead index @xmath79 , @xmath80 denotes the tunneling term and the bias voltage term @xmath81 is given respectively by : @xmath462 where the boson field @xmath83 denotes the chiral luttinger liquid in lead @xmath84 , the tunneling between lead and the resonant level is given by @xmath85 , @xmath0 is the bias voltage , and @xmath86 refers to the luttinger parameter . via similar unitary transfermations mentioned above , @xmath463 and @xmath464 , eq . ( [ dot - luttinger ] ) now becomes : @xmath465 where the tunneling term @xmath80 in eq . ( [ h - ll - appendix ] ) becomes ( assumming @xmath45 ) : @xmath466 s^{- } + h.c.\\ \nonumber & - & ( 1 - \sqrt{\frac{1}{2 k } } ) ( \partial \sqrt{2}\varphi_{1 } + \partial \sqrt{2 } \varphi_2 ) s_z , \end{aligned}\ ] ] and the chemical potential term in eq . ( [ h - ll - appendix ] ) becomes @xmath467 . \end{aligned}\ ] ] the equivalence between a resonant level coupled to fqhe eq . ( [ dot - luttinger ] ) and a dissipative resonant level model eq . ( [ rlm ] ) is established by comparing the transformed hamiltonian @xmath468 in eq . ( [ barh - fqhe ] ) for the former model and @xmath469 ( see eq . ( [ tilde - h - bath ] ) , ( [ tilde - h - t ] ) , and ( [ tilde - h - mu ] ) ) for the latter one . below we provide details on the mapping of a large dissipative resonant level onto a large resonant level ( spinless quantum dot ) with ohmic dissipation coupled to two chiral luttinger liquid leads . the mapping is easily extended to the latter case with a small ( single - level ) resonant level . first , we take the same dissipative boson environment as shown in eq . ( [ rlm - diss ] ) . via standard bosonization ( see eq . ( [ bosonization - ll ] ) ) , the luttinger leads and the chemical potential term take the same bosonized form as eq . ( [ rlm - leads ] ) and eq . ( [ h - ll ] ) , respectively . the remaining parts of the hamiltonian are modified as follows : @xmath470 where @xmath471 referrs to the energy spectrum of the many - levl dot , the electron destruction operator on the dot @xmath472 is defined as : @xmath473 , and spin - flip operator @xmath474 represents for the hoping of charge between lead and the dot@xcite . we then bosonize the electron operators in the leads ( see eq . ( [ bosonization - ll ] ) ) and on the dot : @xmath475 . via the unitary transformation @xmath426 defined in eq . [ ub ] , we arrive at : @xmath476 to further simplify the hoping term , we define new boson fields @xmath477 via linear combinations of the fields @xmath478 and @xmath479 : @xmath480 with @xmath481 . the combined fermionic baths of the leads and the dot are given by : @xmath482 , \label{kprime}\end{aligned}\ ] ] where @xmath483 are canonically conjugate boson fields to @xmath477 fields . in terms of the new fields @xmath477 , the hoping and chemical potential terms now become : @xmath484 . \end{aligned}\ ] ] we may furthermore combine the boson fields from the leads @xmath485 and from the dissipative bath @xmath418 via the following definitions : @xmath486 where @xmath487 . upon applying the unitary transformation , the combined fermionic and bosnoic baths terms become : @xmath488,\end{aligned}\ ] ] where @xmath489 are canonically conjugate boson fields to @xmath437 fields . meanwhile , the corresponding hoping and chemical potential terms become : @xmath490 . \end{aligned}\ ] ] via the similar unitary transfermation mentioned above , @xmath491 and @xmath492 , the tunneling term becomes ( assumming @xmath45 ) : @xmath493 s^{- } + h.c.\\ \nonumber & - & ( 1 - \sqrt{\frac{1}{2 \tilde{k } } } ) ( \partial \sqrt{2}\tilde{\phi}_{s,1 } + \partial \sqrt{2 } \tilde{\phi}_{s,2 } ) s_z . \end{aligned}\ ] ] the chemical potential term therefore becomes @xmath494 . \end{aligned}\ ] ] we may now follow the same refermionization procedure as shown in eq . ( [ refermionization ] ) to map our hamiltonian onto the anisotropic kondo model in the same form as eq . ( [ hkondo ] ) with the following identifications : @xmath495 the above mapping can easily be generalized to a small quantum dot with single resonant level with @xmath94 given by eq . ( [ k - small ] ) where the contribution from the many - level big dot is absent here . in this appendix , we prove that the average currents in the original model @xmath496 is equivalent to that in the effective kondo model @xmath497 . the current operators in both models are given by : @xmath498 @xmath499 on the other hand , from the bosonized forms of the two models , ( at the transition ) we have : @xmath500 \rangle \\ \nonumber & = & \int \ dx\sqrt{\frac{1}{2k } } \langle \frac{d}{dt } [ \partial_x ( \sqrt{2}\tilde{\phi}_{s,1 } ) - \partial_x ( \sqrt{2}\tilde{\phi}_{s,2 } ) ] \rangle\end{aligned}\ ] ] @xmath501 \rangle \\ \nonumber & = & \int dx\ \langle d / dt [ \partial_x ( \sqrt{2}\tilde{\phi}_{s,1 } ) - \partial_x ( \sqrt{2}\tilde{\phi}_{s,2 } ) ] \rangle\end{aligned}\ ] ] therefore , we have @xmath502 ( or @xmath503 ) . the above relation obtained so far from the mapping is exact at finite bias voltages . in the limit of our interest @xmath504 , @xmath505 . + we can also prove this equivalence through keldysh perturbation theory . we now would like to prove that @xmath506 where @xmath507 \nonumber \\ z_{k } & = & tr[e^{-\beta h_{k } } t_c(s_c^{k}(-\infty,\infty))]\nonumber \\ s_c^{k}(-\infty,\infty ) & = & e^{-{\it i } \int_c dt ' h_{k}^{eq}(t')}\end{aligned}\ ] ] and @xmath508 \nonumber \\ z_{ori } & = & tr[e^{-\beta h } t_c(s_c^{ori}(-\infty,\infty))]\nonumber \\ s_c^{ori}(-\infty,\infty ) & = & e^{-{\it i } \int_c dt ' h^{eq}(t')}\end{aligned}\ ] ] here @xmath509 is the kondo ( original ) hamiltonian in equilibrium ( @xmath510 ) , @xmath511 orders the operators along the keldysh contour @xmath512 . we first show that @xmath513 ( the two partition functions from the original and the effective kondo models are equivalent ) where @xmath514\\ \nonumber s_c^{ori}(-\infty,\infty ) & = & e^{-{\it i } \int_c dt ' h^{eq}(t')}\end{aligned}\ ] ] to prove this , we first note that the original and the effective kondo models are related by the above - mentioned unitary transfromations : @xmath515 with @xmath516 . the similar relation holds for the current operators : @xmath517 . using the following identities : @xmath518^n\end{aligned}\ ] ] where @xmath519 & = & tr[\hat{c}(\tau ) \hat{a}(\tau ) \hat{b}(\tau)]\nonumber \\ & = & tr[\hat{b}(\tau ) \hat{c}(\tau ) \hat{a}(\tau ) ] = \cdots\end{aligned}\ ] ] with @xmath520 the imaginary time and @xmath521 being any quantum mechanical operators we can then show that @xmath513 . in other words , when the hamiltonian is under above unitary transformations , the trace in the partition function remains unchanged . \2 . in the similar way , we can prove that : @xmath522 \\ \nonumber & = & tr[e^{-\beta h } t_c(s_c^{ori}(-\infty,\infty ) i_{ori}(t))]\end{aligned}\ ] ] where we have used eq . ( [ identity ] ) and @xmath517 . from 1 . and 2 . mentioned above , we conclude that @xmath523 holds for all orders in keldysh perturbation theoy . in this appendix , we derive the general expression for the average current for @xmath524 . from eq . ( [ ikondo ] ) , the average current in the kondo model is given by : @xmath525 where @xmath526 . following ref . , the dyson s equation for @xmath527 is given by : @xmath528 -(1\rightarrow 2 , l\rightarrow r)\end{aligned}\ ] ] where @xmath529 , @xmath530 , @xmath531 , @xmath532 , and @xmath533 are time - order and anti - timeordered green s functions , respectively . the following relations hold among these correlation functions : @xmath534 where @xmath535 is the retarded ( advanced ) green s function , respectively . straightforward calculation gives : @xmath536 the average current reads @xmath537 ( \chi_d^r(\omega ) - \chi_d^a(\omega))\\ \nonumber & + & ( \tilde{\gamma}_1-\tilde{\gamma}_2 ) \chi_d^<(\omega)\end{aligned}\ ] ] where @xmath538 with @xmath539 being the constant density of states of the leads . j. schwinger , j. math . phys . * 2 * , 407 ( 1961 ) ; l. v. keldysh , soviet physics jetp * 20 * , 1018 ( 1965 ) ; j. rammer and h. smith , rev . mod . phys . * 58 * , 323 ( 1986 ) ; a. p. jauho , n. s. wingreen and y. meir , phys . b * 50 * , 5528 ( 1994 ) . c. mora , p. vitushinsky , x. leyronas , a. a. clerk , and k. le hur , phys . rev . b * 80 * , 155322 ( 2009 ) ; p. vitushinsky , a. a. clerk , and k. le hur , phys . rev . lett . * 100 * , 036603 ( 2008 ) ; c. mora , x. leyronas and n. regnault , phys . rev . lett . * 100 * , 036604 ( 2008 ) ; z. ratiani and a. mitra , phys . b * 79 * , 24511 ( 2009 ) . m. schiro and m. fabrizio , phys . rev . b * 79 * , 155302 ( 2009 ) ; p. werner , t. oka and a. j. millis , phys . rev . b * 79 * , 035320 ( 2009 ) ; t. l. schmidt , p. werner , l. muehibacher , a. komnik , phys . rev . b * 78 * , 235110 ( 2008 ) ; p. werner et al . , b * 81 * , 035108 ( 2010 ) ; l. muehlbacher , d. f. urban , and a. komnik , phys . rev . b * 83 * , 075107 ( 2011 ) ; e. gull , a. j. millis , a.i . lichtenstein , a. n. rubtsov , m. troyer , and p. werner , rev . phys . * 83 * , 349 ( 2011 ) . see , for example : a. feiguin , p. fendley , m. fisher , and c. nayak , phys . * 101 * , 236801 ( 2008 ) ; a. m. chang , rev . phys . * 75 * , 1449 ( 2003 ) ; m. grayson , d. c. tsui , l. n. pfeiffer , k. w. west , and a. m. chang , phys . lett . * 86 * , 2645 ( 2001 ) . y. bomze , h. mebrahtu , i. borzenets , a. makarovski , and g. finkelstein , phys . b * 79 * , 241402(r ) ( 2009 ) ; h. t. mebrahtu , i. v. borzenets , d. e. liu , h. zheng , y. v. bomze , a. i. smirnov , h. u. baranger , and g. finkelstein , nature * 488 * , 61 , ( 2012 ) . delbecq , v. schmitt , f.d . parmentier , n. roch , j.j . viennot , g. fve , b. huard , c. mora , a. cottet and t. kontos , phys . lett . * 107 * , 256804 ( 2011 ) ; k. d. petersson _ et al . _ , nature * 490 * , 380 ( 2012 ) ; h. toida _ et al . _ , arxiv:1206.074 ; s. j. chorley _ et al . _ , lett . * 108 * , 036802 ( 2012 ) ; t. frey _ et al . _ , b * 86 * , 115303 ( 2012 ) . o. astafiev , a.m. zagoskin , a.a . abdumalikov jr . , a. pashkin , t. yamamoto , k. inomata , y. nakamura , and j.s . tsai , science * 327 * , 840 ( 2010 ) ; k. le hur , phys . b * 85 * , 140506(r ) ( 2012 ) ; i .- c . hoi , c.m . wilson , g. johansson , t. palomaki , b. peropadre , p. delsing , phys . lett . * 107 * , 073601 ( 2011 ) ; m. goldstein , m. h. devoret , m. houzet , l. i. glazman , arxiv:1208.0319 ; m. marthaler , y. utsumi , d. s. golubev , a. shnirman , and gerd schn , phys . lett . * 107 * , 093901 ( 2011 ) ; p .- q . jin , m. marthaler , j. h. cole , a. shnirman , and gerd schn , phys . b * 84 * , 035322 ( 2011 ) ; h. zheng , d. j. gauthier , and h. u. baranger , phys . rev . a * 82 * , 063816 ( 2010 ) ; h. zheng , d. j. gauthier , and h. u. baranger , phys . a * 85 * , 043832 ( 2012 ) ; p. longo , p. schmitteckert and k. busch phys . lett . * 104 * , 023602 ( 2010 ) ; j .- shen and s. fan , phys . rev . lett . * 98 * , 153003 ( 2007 ) ; j .- shen and s. fan , phys . a * 76 * , 062709 ( 2007 ) ; a. leclair , f. lesage , s. lukyanov , and h. saleur , phys . a * 235 * , 203 ( 1997 ) ; r. konik and a. leclair , phys . b * 58 * , 1872 ( 1998 ) . the frg approach for the kondo couplings @xmath347 used here@xcite ( eq . ( [ gpergz ] ) and eq . ( [ gamma ] ) ) is somewhat different from that in ref . : in ref . the rg scaling equation for pseudofermion self - energy is formulated and is solved self - consistently along with the rg scaling equations for the kondo couplings . here , following ref . and ref . the pseudofermion self energy is included self - consistently through the decoherence rate @xmath160 where it is computed via the fermi - golden rule within the renormalized perturbation theory . nevertheless , we have checked that the results on @xmath347 via both approaches agree very well .

the resonant - level model represents a paradigmatic quantum system which serves as a basis for many other quantum impurity models . we provide a comprehensive analysis of the non - equilibrium transport near a quantum phase transition in a spinless dissipative resonant - level model , extending earlier work [ phys . . lett . * 102 * , 216803 ( 2009 ) ] . a detailed derivation of a rigorous mapping of our system onto an effective kondo model is presented . a controlled energy - dependent renormalization group approach is applied to compute the non - equilibrium current in the presence of a finite bias voltage @xmath0 . in the linear response regime @xmath1 , the system exhibits as a function of the dissipative strength a localized - delocalized quantum transition of the kosterlitz - thouless ( kt ) type . we address fundamental issues of the non - equilibrium transport near the quantum phase transition : does the bias voltage play the same role as temperature to smear out the transition ? what is the scaling of the non - equilibrium conductance near the transition ? at finite temperatures , we show that the conductance follows the equilibrium scaling for @xmath2 , while it obeys a distinct non - equilibrium profile for @xmath3 . we furthermore provide new signatures of the transition in the finite - frequency current noise and ac conductance via the recently developed functional renormalization group ( frg ) approach . the generalization of our analysis to non - equilibrium transport through a resonant level coupled to two chiral luttinger - liquid leads , generated by the fractional quantum hall edge states , is discussed . our work on dissipative resonant level has direct relevance to the experiments in a quantum dot coupled to resistive environment , such as h. mebrahtu _ et al . _ , nature * 488 * , 61 , ( 2012 ) .

the really novel feature of general relativity ( henceforth abbreviated gr ) , as compared to other field theories in physics , is that spacetime is not a fixed background arena that merely stages physical processes . rather , spacetime is itself a dynamical entity , meaning that its properties depend in parts on its specific matter content . hence , contrary to the newtonian picture , in which spacetime acts ( via its inertial structure ) but is not acted upon by matter , the interaction between matter and spacetime now goes both ways . saying that the spacetime is ` dynamic ' does not mean that it ` changes ' with respect to any given external time . time is clearly within , not external to spacetime . accordingly , solutions to einstein s equations , which are whole spacetimes , do not as such describe anything evolving . in order to take such an evolutionary form , which is , for example , necessary to formulate an initial value problem , we have to re - introduce a notion of ` time ' with reference to which we may speak of ` evolution ' . this is done by introducing a structure that somehow allows to split spacetime into space and time . ( -235,57)@xmath0 ( -145,58)@xmath1 ( -145,38)@xmath2 ( -145,78)@xmath3 ( -60,57)@xmath4 ( -60,23)@xmath5 ( -60,95)@xmath6 ( -105,98)@xmath7 let us explain this in more detail : suppose we are give a spacetime , that is , a four dimensional differentiable manifold @xmath7 with lorentzian metric @xmath8 . we assume that @xmath7 can be foliated by a family @xmath9 of spacelike leaves . that is , for each number @xmath10 there is an embedding of a fixed 3-dimensional manifold @xmath0 into @xmath7 , @xmath11 whose image @xmath12 is just @xmath4 , which is a spacelike submanifold of @xmath7 ; see fig.[fig : nico1 ] . it receives a riemannian metric by restricting the lorentzian metric @xmath8 of @xmath7 to the tangent vectors of @xmath4 . this can be expressed in terms of the 3-manifold @xmath0 . if we endow @xmath0 with the riemannian metric @xmath13 then @xmath14 is isometric to the submanifold @xmath4 with the induced metric . each three dimensional leaf @xmath4 now corresponds to an instant of time @xmath10 , where @xmath10 is so far only a topological time : it faithfully labels instants in a continuous fashion , but no implication is made as to its relation to actual clock readings . the statement of such relations can eventually only be made on the basis of dynamical models for clocks coupled to the gravitational field . by means of the foliation we now recover a notion of time : we view spacetime , @xmath15 , as the one - parameter family of spaces , @xmath16 . spacetime then becomes nothing but a ` trajectory of spaces ' . in this way we obtain a dynamical system whose configuration variable is the riemannian metric on a 3-manifold @xmath0 . it is to make this point precise that we carefully distinguish between the manifold @xmath0 and its images @xmath4 in @xmath7 . in the dynamical formulation given now , there simply is no spacetime to start with and hence no possibility to embed @xmath0 into something . only _ after _ solving the dynamical equations can we construct spacetime and interpret the time dependence of the metric of @xmath0 as being brought about by ` wafting ' @xmath0 through @xmath7 via a one - parameter family of embeddings @xmath1 . but initially there is only a 3-manifold @xmath0 of some topological type , that is , solutions exist for _ any _ topology . however , one often imposes additional requirements on the solution . for example , one may require that there exists a moment of time symmetry , which will make the corresponding instant @xmath4 a totally geodesic submanifold of @xmath7 , like e.g. in recollapsing cosmological models at the moment of maximal expansion . in this case the topology of @xmath0 will be severely restricted . in fact , most topologies @xmath0 will only support geometries that always expand or contract somewhere . ] and the equations of motion together with some suitable initial data . for a fuller discussion we refer to the comprehensive work by isham and kucha @xcite . whereas a specified motion of @xmath0 through @xmath7 , characterized by the family of embeddings ( [ eq : sigmaembedding ] ) , gives rise to a one - parameter family of metrics @xmath17 , the converse is not true . that is to say , it is not true that _ any _ one - parameter family of metrics @xmath17 of @xmath0 can be obtained by finding a spacetime @xmath15 and a one parameter family of embeddings @xmath1 , such that ( [ eq : spatialmetric ] ) holds . moreover , there is clearly a huge redundancy in creating @xmath15 from the family @xmath18 , since there are obviously many different motions of @xmath0 through the same @xmath7 , which give rise to apparently different solution curves @xmath17 . this redundancy can be locally parameterized by four functions , on @xmath0 : a scalar field @xmath19 and a vector field @xmath20 . in the embedding picture , they describe the components of the velocity vector field @xmath21 normal and tangential to the leaves @xmath4 respectively . we write @xmath22 where @xmath23 is the normal to @xmath4 . the tangential component , @xmath20 , just generates intrinsic diffeomorphisms on each @xmath4 , whereas the normal component , @xmath19 , really advances one leaf @xmath4 to the next one ; see fig.[fig : nico2 ] . ( -180,11)@xmath24 ( -180,78)@xmath25 ( -135,14)@xmath26 ( -49,88)@xmath27 ( -95,-5)@xmath20 ( -40,45)@xmath28 ( -105,45)@xmath29 for the initial - value problem it is the derivative along the normal @xmath23 of the 3-metric @xmath30 , denoted by @xmath31 , that gives the essential information . hence we write @xmath32 in the embedding picture @xmath33 is the extrinsic curvature of @xmath4 in @xmath7 . the first order evolution equations that result from einstein s field equations are then of the general form @xmath34 where @xmath35 in ( [ eq : evolution1 ] ) is given by the right - hand side of ( [ eq : hdot ] ) . @xmath36 is a more complicated function whose precise structure need not interest us now and which also depends on matter variables ; see e.g. @xcite . as we have seen , the initial data for the gravitational variables consist of a differentiable 3-manifold @xmath0 , a riemannian metric @xmath30 the configuration variable , and another symmetric second rank tensor field @xmath31 on @xmath0 the velocity variable . however , the pair @xmath37 can not be chosen arbitrarily . this is because there is a large redundancy in describing a fixed spacetime @xmath7 by a foliation ( [ eq : sigmaembedding ] ) . on the infinitesimal level this gauge freedom is just the freedom of choosing @xmath19 and @xmath20 . the gauge transformations generated by @xmath20 are just the spatial diffeomorphisms of @xmath0 . @xmath20 may be an arbitrary function of @xmath10 , which corresponds to the fact that we may arbitrarily permute the points in each leaf @xmath4 separately ( only restricted by some differentiability conditions ) . the gauge transformations generated by @xmath19 correspond to pointwise changes in the velocities with which the leaves @xmath4 push through @xmath7 . these too may vary arbitrarily within the leaves as well as with coordinate time @xmath10 . whenever there is gauge freedom in a dynamical theory , there are so - called _ constraints _ , that is , conditions which restrict the initial data ; see e.g. @xcite . for each gauge freedom parameterized by an arbitrary function , there is one functional combination of the initial data which has to vanish . in our case there are four gauge functions , @xmath19 , and the three components of @xmath20 . accordingly there are four constraints , which group into one _ scalar or hamiltonian constraint _ , @xmath38=0 $ ] , and three combined in the _ vector or diffeomorphism constraint _ , @xmath39=0 $ ] . their explicit expressions are : and use the abbreviation @xmath40 . hence @xmath41 has the physical dimension of @xmath42 . we shall set @xmath43 throughout . ] @xmath44 & & \,=\,(2\kappa)^{-1}\,g^{ab\,cd}k_{ab}k_{cd } -(2\kappa)^{-1}\sqrt{h}\bigl({}^{\scriptscriptstyle ( 3)}\!r-2\lambda\bigr ) + \,\sqrt{h}\rho\,,\\ \label{eq : diffconstraint1 } & d^a[h , k ] & & \,=\,-\,\kappa^{-1}\,g^{ab\,cd}\nabla_bk_{cd}+\,\sqrt{h}j^a\ , .\end{aligned}\ ] ] here @xmath45 and @xmath46 are the energy- and momentum densities of the matter , @xmath47 and @xmath48 are the levi - civita connection and its associated scalar curvature of @xmath49 . finally @xmath50 is the so called dewitt metric , which at each point of @xmath0 defines an @xmath30-dependent lorentzian metric on the @xmath51 dimensional space of symmetric second - rank tensors at that point . its explicit form is given by @xmath52 note that the linear space of symmetric second - rank tensors is viewed here as the tangent space ( ` velocity space ' ) of the space @xmath53 of riemannian metrics on @xmath0 . from ( [ eq : wdwmetric1 ] ) one sees that it is the trace part of the ` velocities ' , corresponding to changes of the scale ( conformal part ) of the riemannian metric , that span the negative - norm velocity directions . the presence of constraints and their relation to the evolution equations is the key structure in canonical gr . it is therefore instructive to point out how this structure arises from the conventional , four dimensional form of einstein s equations . before doing this , it is useful to first remind ourselves on the analogous situation in electrodynamics . so let us first consider electrodynamics in minkowski space . as usual , we write the field tensor @xmath54 as exterior differential of a vector potential @xmath55 , that is @xmath56 . in components this reads @xmath57 . here @xmath58 are the components of the electric , @xmath59 of the magnetic field , where @xmath60 is a cyclic permutation of 123 . the homogeneous maxwell equations now simply read @xmath61 , whereas the inhomogeneous maxwell equations are given by ( in components ) : @xmath62 where here @xmath63 is the electric four - current . due to its antisymmetry , the field tensor obeys the identity @xmath64 taking the divergence of ( [ eq : inhmaxwelleq ] ) and using ( [ eq : maxwellid1 ] ) leads to @xmath65 showing the well known fact that maxwell s equations imply charge conservation as integrability condition . let us now interpret the role of charge conservation in the initial - value problem . decomposing ( [ eq : maxwellid1 ] ) into space and time derivatives yields @xmath66 even though the right - hand side contains third derivatives in the field @xmath67 , time derivatives appear at most in second order ( since @xmath68 is spatial ) . hence , since it is an identity , @xmath69 contains time derivatives only up to first order . but the initial data for the second order equation ( [ eq : inhmaxwelleq ] ) consist of the field @xmath67 and its first time derivative . hence the time component @xmath70 of maxwell s equations gives a relation amongst initial data , in other words , it is a _ constraint_. clearly this is just the gau constraint @xmath71 ( here @xmath45 is the electric charge density ) . only the three spatial components of ( [ eq : inhmaxwelleq ] ) contain second time derivatives and hence propagate the fields . they provide the evolutionary part of maxwell s equations . now , assume we are given initial data satisfying the constraint @xmath72 , which we evolve according to @xmath73 . how can we be sure that the evolved data again satisfy the constraint ? to see when this is the case , we use the identity ( [ eq : maxwellid2 ] ) and solve it for the time derivative of @xmath70 : @xmath74 this shows : if initially @xmath73 ( and hence @xmath75 ) , then the constraint @xmath72 is preserved in time if and only if @xmath76 . charge conservation is thus recognized as the necessary and sufficient condition for the compatibility between the constraint part and the evolutionary part of maxwell s equations . finally we wish to make another remark concerning the interplay between constraints and evolution equations . it is clear that a solution @xmath77 to ( [ eq : inhmaxwelleq ] ) satisfies the constraint on _ any _ simultaneity hypersurface of an inertial observer ( i.e. spacelike plane ) . if the normal to the hypersurface is @xmath78 , this just states that @xmath79 implies @xmath80 . but the converse is obviously also true : if @xmath80 for all timelike @xmath78 , then @xmath79 . in words : given an electromagnetic field that satisfies the constraint ( for given external current @xmath63 ) on _ any _ spacelike plane in minkowski space , then this field must necessarily satisfy maxwell s equations . in this sense , maxwell s equations are the _ unique _ propagation law that is compatible with gau constraint . after this digression we return to gr , where we can perform an entirely analogous reasoning . we start with einstein s equations , in which the spacetime metric @xmath81 is the analog of @xmath67 and the einstein tensor @xmath82 is the analog of @xmath83 . they read @xmath84 due to four dimensional diffeomorphism invariance , we have the identity ( twice contracted second bianchi - identity ) : @xmath85 which is the analog of ( [ eq : maxwellid1 ] ) . taking the covariant divergence of ( [ eq : einsteinequation ] ) and using ( [ eq : bianchiid ] ) yields @xmath86 which is the analog of ( [ eq : maxwellid2 ] ) . hence the vanishing covariant divergence of @xmath87 is an integrability condition of einstein s equations , just as the divergencelessness of the electric four - current was an integrability condition of maxwell s equations . does not as such imply a conservation law . only in presence of a spacetime symmetry , i.e. a killing vector field @xmath88 , the current @xmath89 is conserved , @xmath90 , and hence gives rise to a conserved quantity . ] in order to talk about ` evolution ' , we consider the foliation ( [ eq : sigmaembedding ] ) of @xmath7 and locally use coordinates @xmath91 such that @xmath92 is the normal @xmath23 to the leaves and all @xmath93 are tangential . expanding ( [ eq : bianchiid ] ) in terms of partial derivatives gives : @xmath94 which is the analog of ( [ eq : maxwellid3 ] ) . now , since the @xmath95 contain at most second and the @xmath96 at most first derivatives of the metric @xmath81 , this identity immediately shows that the four components @xmath97 ( @xmath98 ) contain at most first time derivatives @xmath92 . but einstein s equations are of second order , hence the four equations @xmath99 are relations amongst the initial data , rather than being evolution equations . in fact , up to a factor of -2 they are just the constraints ( [ eq : hamconstraint1][eq : diffconstraint1 ] ) : @xmath100 moreover , the remaining purely spatial components of einstein s equations are equivalent to the twelve first - order evolution equations ( [ eq : evolution1]-[eq : evolution2 ] ) . the interplay between constraints and evolution equations can now be followed along the very same lines as for the electrodynamic analogy . expanding the left equality of ( [ eq : einsteinid ] ) in terms of partial derivatives gives @xmath101 which is the analog of ( [ eq : maxwellid4 ] ) . it shows that the constraints are preserved by the evolution if and only if the energy - momentum tensor of the matter has vanishing covariant divergence . let us now turn to the last analogy : the uniqueness of the evolution preserving constraints . clearly einstein s equations @xmath102 imply @xmath103 for any timelike vector field @xmath78 . hence the constraints are satisfied on any spacelike slice through spacetime . again the converse is also true : given a gravitational field such that @xmath103 for any timelike @xmath78 ( and given external @xmath87 ) , then this field must necessarily satisfy einstein s equations . in this sense einstein s equations follow uniquely from the condition of constraint preservation . this property will be crucial for the interpretation of the quantum theory discussed below . we know from quantum mechanics that the classical trajectories have completely disappeared at the fundamental level . as we have discussed above , the analogue to a trajectory is in gr provided by a spacetime given as a set of three dimensional geometries . in quantum gravity , the spacetime will therefore disappear like the classical trajectory in quantum mechanics . it is therefore not surprising that the evolution equations and will be absent in quantum gravity . all the information will be contained in the quantized form of the constraints and . we have seen above that einstein s equations can be written as a dynamical system ( [ eq : evolution1][eq : evolution2 ] ) with constraints ( [ eq : hamconstraint1][eq : diffconstraint1 ] ) . here we wish to give its canonical formulation . basically this means to introduce momenta for the velocities and write the first - order equations of motions as hamilton equations . this means to identify the poisson structure and the hamiltonian . the result is this : as before , the configuration variable is the riemannian metric @xmath104 on @xmath0 . its canonical momentum is now given by @xmath105 so that the poisson brackets are @xmath106 where @xmath107 is the dirac distribution on @xmath0 . elimination of @xmath108 in favour of @xmath109 in the constraints leads to their canonical form : @xmath110 & & \,=\,2\kappa\,g_{ab\,cd}\pi^{ab}\pi^{cd } \,-\,(2\kappa)^{-1}\sqrt{h}({}^{\scriptscriptstyle ( 3)}\!r-2\lambda)\ , + \,\sqrt{h}\rho\,,\\ \label{eq : diffconstraint3 } & d^a[h,\pi ] & & \,=-2\nabla_b\pi^{ab}\,+\,\sqrt{h}j^a\,,\end{aligned}\ ] ] where now ) . @xmath111 is the inverse to @xmath50 , i.e. @xmath112 , and _ not _ obtained by simply lowering the indices using @xmath104 . ] @xmath113 likewise , rewriting ( [ eq : evolution1][eq : evolution2 ] ) in terms of the canonical variables shows that they are just the flow equations for the following hamiltonian : @xmath114=\int_\sigma d^3x \bigl\{\alpha(x ) h[h,\pi_{ab}](x)+\beta^a(x)d_a[h,\pi](x)\bigr\}+\text{boundary terms}\,.\ ] ] the crucial observation to be made here is , that , up to boundary terms , the total hamiltonian is a combination of pure constraints . the boundary terms generally appear if @xmath0 is non - compact , as it will be the case for the description of isolated systems , like stars or black holes . in this case the boundary terms are taken over closed surfaces at spatial infinity and represent conserved poincar charges , like energy , linear- and angular momentum , and the quantity associated with asymptotic boost transformations . if , however , @xmath0 is closed ( i.e. compact without boundary ) all of the evolution will be generated by constraints , that is , pure gauge transformations ! in that case , evolution , as described here , is not an observable change . for that to be the case we would need an extrinsic clock , with respect to which ` change ' can be defined . but a closed universe already contains by definition everything physical , so that no external clock exists . accordingly , there is no external time parameter . rather , all physical time parameters are to be constructed from within our system , that is , as functional of the canonical variables . a priori there is no preferred choice of such an intrinsic time parameter . the absence of an extrinsic time and the non - preference of an intrinsic one is commonly known as the _ problem of time _ in hamiltonian ( quantum- ) cosmology . finally we turn to the commutation relation between the various constraints . for this it is convenient to integrate the local constraints ( [ eq : hamconstraint3][eq : diffconstraint3 ] ) over lapse and shift functions . hence we set ( suppressing the phase - space argument @xmath115 $ ] ) @xmath116 a straightforward but slightly tedious computation gives @xmath117)\,,\\ \label{eq : calgebra2 } & \{\mathcal{d}(\beta),\mathcal{h}(\alpha)\ } & & \,=\,\mathcal{h}(\beta(\alpha))\,,\\ \label{eq : calgebra3 } & \{\mathcal{h}(\alpha),\mathcal{h}(\alpha')\ } & & \,=\,\mathcal{d}(\alpha\nabla\alpha'-\alpha'\nabla\alpha)\,.\end{aligned}\ ] ] there are three remarks we wish to make concerning these relations . first , ( [ eq : calgebra1 ] ) shows that the diffeomorphism generators form a lie subalgebra . second , ( [ eq : calgebra2 ] ) shows that this lie subalgebra is not a lie ideal . this means that the flow of the hamiltonian constraint does not leave invariant the constraint hypersurface of the diffeomorphism constraint . finally , the term @xmath118 in ( [ eq : calgebra3 ] ) contains the canonical variable @xmath30 , which is used implicitly to raise the index in the differential in order to get the gradient @xmath47 . this means that the relations above do not make the set of all @xmath119 and all @xmath120 into a lie algebra . in this section we wish to point out that the relations ( [ eq : calgebra1][eq : calgebra3 ] ) follow a general pattern , namely to represent the ` algebra ' of hypersurface deformations , or in other words , infinitesimal changes of embeddings @xmath121 . to make this explicit , we introduce local coordinates @xmath122 on @xmath0 and @xmath123 on @xmath7 . an embedding is then locally given by four functions @xmath124 , such that the @xmath125 matrix @xmath126 has its maximum rank 3 ( we write @xmath127 ) . the components of the normal to the image @xmath128 are denoted by @xmath129 , which should be considered as functional of @xmath124 . the generators of normal and tangential deformations of @xmath130 with respect to the lapse function @xmath19 and shift vector field @xmath20 are then given by @xmath131\,\frac{\delta}{\delta y^\mu(x)}\,,\\ \label{eq : tangentdef } & t_\beta & & \,=\ , \int_\sigma d^3x\ \beta^a(x)\,y^\mu_{,a}(x)\ , \frac{\delta}{\delta y^\mu(x)}\,,\end{aligned}\ ] ] which may be understood as tangent vectors to the space of embeddings of @xmath0 into @xmath7 . a calculation ) is immediate . to verify ( [ eq : alghypdef2][eq : alghypdef3 ] ) one needs to compute @xmath132/\delta y^{\nu}(x')$ ] . this can be done in a straightforward way by varying @xmath133 n^\nu[y(x)]=-1 \quad\text{and}\quad g_{\mu\nu}(y(x))\,y^\mu_{,a}(x)\,n^\nu[y(x)]=0\ ] ] with respect to @xmath134 . ] then leads to the following commutation relations @xmath135 & & \,=\,-\ , t_{[\beta,\beta']}\,,\\ \label{eq : alghypdef2 } & [ t_\beta\,,\,n_\alpha ] & & \,=\,-\ , n_{\beta(\alpha ) } \,,\\ \label{eq : alghypdef3 } & [ n_\alpha\,,\,n_{\alpha ' } ] & & \,=\,-\ , t_{\alpha\nabla\alpha'-\alpha'\nabla\alpha}\,.\end{aligned}\ ] ] up to the minus signs this is just ( [ eq : calgebra1]-[eq : calgebra3 ] ) . the minus signs are just the usual ones that one always picks up in going from the action of vector fields to the poisson action of the corresponding phase - space functions . ( in technical terms , the mapping from vector fields to phase - space functions is a lie-_anti_-homomorphism . ) this shows that ( [ eq : calgebra1][eq : calgebra3 ] ) just mean that we have a hamiltonian realization of hypersurface deformations . in particular , ( [ eq : calgebra1][eq : calgebra3 ] ) is neither characteristic of the action nor the field content : _ any _ four dimensional diffeomorphism invariant theory will gives rise to this very same ` algebra ' . it can be shown that under certain general locality assumptions the expressions ( [ eq : hamconstraint3 ] ) and ( [ eq : diffconstraint3 ] ) give the unique 2-parameter ( here @xmath41 and @xmath136 ) family of realizations for @xmath137 and @xmath138 satisfying ( [ eq : alghypdef1][eq : alghypdef3 ] ) on the phase space parameterized by @xmath139 ; see @xcite and also @xcite . as we have just discussed , einstein s equations take the form of a constrained hamiltonian system if put into canonical form . the unconstrained configuration space is the space of all riemannian metrics on some chosen 3-manifold @xmath0 . this space is denoted by @xmath140 . any two riemannian metrics that differ by an action of the diffeomorphism constraint are gauge equivalent and hence to be considered as physically indistinguishable . let us briefly mention that the question of whether and when the diffeomorphism constraint actually generates all diffeomorphisms of @xmath0 is rather subtle . certainly , what is generated lies only in the identity component of the latter , but even on that it may not be onto . this occurs , for example , in the case where @xmath0 contains asymptotically flat ends with non - vanishing poincar charges associated . asymptotic poincar transformations are then not interpreted as gauge transformations ( otherwise the poincar charges were necessarily zero ) , but as proper physical symmetries ( i.e. changes of state that are observable in principle ) . leaving aside the possible difference between what is generated by the constraints and the full group @xmath141 of diffeomorphisms of @xmath0 , we may consider the quotient space @xmath142 of riemannian _ geometries_. this space is called _ superspace _ in the relativity community ( this has nothing to do with supersymmetry ) , which we denote by @xmath143 . now from a topological viewpoint @xmath140 is rather trivial . it is a cone of @xmath144 is again an element of @xmath140 . ] in the ( infinite dimensional ) vector space of all symmetric second - rank tensor fields . but upon factoring out @xmath141 the quotient space @xmath143 inherits some of the topological information concerning @xmath0 , basically because @xmath141 contains that information @xcite . this is schematically drawn in fig.[fig : nico3 ] . ( 5,130)@xmath140 ( 5,18)@xmath143 in a certain generalized sense , gr is a dynamical system on the phase space ( i.e. cotangent bundle ) built over superspace . the topology of superspace is characteristic for the topology of @xmath0 , though in a rather involved way . note that , by construction , the hamiltonian evolution is that of a varying embedding of @xmath0 into spacetime . hence the images @xmath4 are all of the same topological type . this is why canonical gravity in the formulation given here can not describe transitions of topology . note , however , that this is not at all an implication by einstein s equations . rather , it is a consequence of our restriction to spacetimes that admit a global spacelike foliation . there are many solutions to einstein s equations that do not admit such foliations globally . this means , that these spacetimes can not be constructed by integrating the equations of motions ( [ eq : evolution1][eq : evolution2 ] ) successively from some initial data . should we rule out all other solutions ? the general feeling seems to be , that at least in quantum gravity , topology changing classical solutions should not be ruled out as possible contributors in the sum over histories ( path integral ) . fig.[fig : nico4 ] shows two such histories . whereas in the left picture the universe simply ` grows a nose ' , it bifurcates in the right example to become disconnected . ( -262,14)@xmath145 ( -260,160)@xmath146 ( -290,85)@xmath7 ( -47,20)@xmath145 ( -64,190)@xmath147 ( -95,190)@xmath148 ( -82,100)@xmath7 one may ask whether there are topological restrictions to such transitions . first of all , it is true ( though not at all obvious ) that for any given two 3-manifolds @xmath145 , @xmath146 ( neither needs to be connected ) there is a 4-manifold @xmath7 whose boundary is just @xmath149 . in fact , there are infinitely many such @xmath7 . amongst them , one can always find some which can be endowed with a globally regular lorentz metric @xmath8 , such that @xmath145 and @xmath146 are spacelike . however , if topology changes , @xmath15 necessarily contains closed timelike curves @xcite . this fact has sometimes been taken as rationale for ruling out topology change in ( classical ) gr . but it should be stressed that closed timelike curves do not necessarily ruin conventional concepts of predictability . in any case , let us accept this slight pathology and ask what other structures we wish to define on @xmath7 . for example , in order to define fermionic matter fields on @xmath7 we certainly wish to endow @xmath7 with a @xmath150 spin structure . this is where now the first real obstructions for topological transitions appear @xcite . be the spacelike boundary of the lorentz manifold @xmath7 , then @xmath151 has to be even for @xmath7 to admit an @xmath150 spin structure . ] it is then possible to translate them into selection rules for transitions between all known 3-manifolds @xcite . so far the considerations were purely kinematical . what additional obstructions arise if the spacetime @xmath15 is required to satisfy the field equations ? here the situation becomes worse . it is , for example , known that any topology - changing spacetime that satisfies einstein s equations with matter that satisfies the weak - energy condition @xmath152 for all lightlike @xmath153 must necessarily be singular . for all lightlike @xmath153 . ] hence it seems that we need to consider degenerate metrics already on the classical level if topology change is to occur . can we relax the notion of ` solution to einstein s equations ' so as to contain these degenerate cases as well ? the answer is ` yes ' if instead of taking the metric as basic variable we rewrite the equations in terms of vierbeine and connections ( first oder formalism ) . it turns out that the kind of singularities one has to cope with are very mild indeed : the vierbeine become degenerate on sets of measure zero but , somewhat surprisingly , the curvature stays bounded everywhere . in fact , there is a very general method to generate an abundance of such solutions @xcite . it is a much debated question whether topology changing amplitudes are suppressed or , to the contrary , needed in quantum gravity . on one hand , it has been shown in the context of specific lower dimensional models that matter fields on topology - changing backgrounds may give rise to singularities corresponding to infinite densities of particle production @xcite . on the other hand , leaving out topology changing amplitudes in the sum - over - histories approach is heuristically argued to be in conflict with expected properties of localized pseudo - particle - like excitations in gravity ( so called geons ) , like , for example , the usual spin - statistic relation @xcite . here there still seems to be much room for speculations . just in the same way as any lagrangian theory endows the configuration space with the kinetic - energy metric , @xmath140 inherits a metric structure from the ` kinetic - energy ' part of ( [ eq : hamconstraint1 ] ) . tangent vectors at @xmath144 are symmetric second - rank tensor fields on @xmath0 and their inner product is given by the so - called _ wheeler dewitt metric _ : @xmath154 due to the pointwise lorentzian signature ( 1 + 5 ) of @xmath50 it is of a hyper - lorentzian structure with infinitely many negative , null , and positive directions each . however , not all directions in the tangent space @xmath155 correspond to physical changes . those generated by diffeomorphism , which are of the form @xmath156 for some vector field @xmath20 on @xmath0 are pure gauge . we call them _ vertical_. ( 2,50)@xmath140 ( 2,0)@xmath143 ( -72,124 ) 92vertical ( -63,113 ) 15horizontal ( -199,-9)@xmath157 $ ] ( -168,-9)@xmath158 $ ] ( -113,-9)@xmath159 $ ] ( -230,122)@xmath30 ( -171,64)@xmath160 ( -125,162)@xmath161 the diffeomorphism constraint ( [ eq : diffconstraint3 ] ) for @xmath162 a case to which we now restrict for simplicity now simply says that @xmath163 must be @xmath164orthogonal to such vertical directions . we call such orthogonal directions _ moreover , it is easily seen that the inner product ( [ eq : dewittmetric ] ) is invariant under @xmath141 . all this suggests how to endow superspace , @xmath143 , with a natural metric : take two tangent vectors at a point @xmath157 $ ] in @xmath143 , lift them to horizontal vectors at @xmath30 in @xmath140 and there take the inner product according to ( [ eq : dewittmetric ] ) . however , this procedure only works if the horizontal subspace of @xmath155 is truly complementary to the vertical space of gauge directions . however , this is not guaranteed due to @xmath165 not being positive definite : whenever there are vertical directions of zero @xmath164norm , there will be non - trivial intersections of horizontal and vertical spaces . sufficient conditions on @xmath30 for this _ not _ to happen can be derived , like , for example , a strictly negative ricci tensor @xcite . the emerging picture is that there are open sets in @xmath143 in which well defined hyper - lorentzian geometries exist , which are separated by closed transition regions in which the signature of these metrics change . the transition regions precisely consist of those geometries @xmath157 $ ] which possess vertical directions of zero @xmath164norm ; see fig.[fig : nico5 ] . einstein s theory of gr has now been brought into a form where it can be subject to the procedure of canonical quantization . as we have argued above , all the information that is needed is encoded in the constraints and . however , quantizing them is far from trivial @xcite . one might first attempt to solve the constraints on the classical level and then to quantize only the reduced , physical , degrees of freedom . this is already impossible in quantum electrodynamics ( except the case of freely propagating fields ) , and it is illusory to achieve in gr . one therefore usually follows the procedure proposed by dirac and tries to implement the constraints as conditions on physically allowed wave functionals . the constraints and the become the quantum conditions @xmath166 where the ` hat ' is a symbolic indication for the fact that the classical expressions have been turned into operators . this procedure also applies if other variables instead of the three - metric and its momentum are used ; for example , such quantum constraints also play the role in loop quantum gravity , cf . the contributions of nicolai and peeters as well as thiemann to this volume . in the present case the resulting formalism is called quantum geometrodynamics . quantum geometrodynamics is defined by the transformation of @xmath167 into a multiplication operator and @xmath168 into a functional derivative operator , @xmath169 . the constraints and then assume the form , restricting here to the vacuum case for simplicity , @xmath170 equation is called the _ wheeler dewitt equation _ in honour of the work by bryce dewitt and john wheeler ; see e.g. @xcite for details and references . in fact , these are again infinitely many equations ( one equation per space point ) . the constraints are called the _ quantum diffeomorphism ( or momentum ) constraints_. occasionally , both ( [ wdw ] ) and ( [ momentum ] ) are referred to as wheeler dewitt equations . in the presence of non - gravitational fields , these equations are augmented by the corresponding terms . the argument of the wave functional @xmath171 is the three - metric @xmath167 ( plus non - gravitational fields ) . however , because of , @xmath171 is invariant under coordinate transformations on three dimensional space ( it may acquire a phase with respect to ` large diffeomorphisms ' that are not connected with the identity ) . a most remarkable feature of the quantum constraint equations is their ` timeless ' nature the external parameter @xmath10 has completely disappeared . instead of an external time one may consider an ` intrinsic time ' that is distinguished by the kinetic term of . as can be recognized from the signature of the dewitt metric , the wheeler dewitt equation is locally hyperbolic , that is , it assumes the form of a local wave equation . the intrinsic timelike direction is related to the conformal part of the three - metric . with respect to the discussion in the last section one may ask whether there are regions in superspace where the wheeler dewitt metric exists and has precisely one negative direction . in that case the wheeler dewitt equation would be strictly hyperbolic ( rather than ultrahyperbolic ) in a neighbourhood of that point . it has been shown that such regions indeed exist and that they include neighbourhoods of the standard round three - sphere geometry @xcite . this implies that the full wheeler dewitt equation that describes fluctuations around the positive curvature friedmann universe is strictly hyperbolic . in this case the scale factor of the friedmann universe could serve as an intrinsic time . the indefinite nature of the kinetic term reflects the fact that gravity is attractive @xcite . there are many problems associated with the quantum constraints and . an obvious problem is the ` factor - ordering problem ' : the precise form of the kinetic term is open there could be additional terms proportional to @xmath172 containing at most first derivatives in the metric . since second functional derivatives at the same space point usually lead to undefined expressions such as @xmath173 , a regularization ( and perhaps renormalization ) scheme has to be employed . connected with this is the potential presence of anomalies , cf . the contribution by nicolai and peeters . another central problem is what choice of hilbert space one has to make , if any , for the interpretation of the wave functionals . no final answer to this problem is available in this approach @xcite . what about the semiclassical approximation and the recovery of an appropriate external time parameter in some limit ? for the full quantum constraints this can at least be achieved in a formal sense ( i.e. , treating functional derivatives as if they were ordinary derivatives and neglecting the problem of anomalies ) ; see @xcite . the discussion is also connected to the question : where does the imaginary unit i in the ( functional ) schrdinger equation come from ? the full wheeler dewitt equation is real , and one would thus also expect real solutions for @xmath171 . an approximate solution is found through a born oppenheimer - type of scheme , in analogy to molecular physics . the state then assumes the form @xmath174/\hbar ) \ , \psi[h,\phi]\ , \ ] ] where @xmath30 is an abbreviation for the three - metric , and @xmath175 stands for non - gravitational fields . in short , one finds that * @xmath176 obeys the hamilton jacobi equation for the gravitational field and thereby defines a classical spacetime which is a solution to einstein s equations ( this order is formally similar to the recovery of geometrical optics from wave optics via the eikonal equation ) . * @xmath177 obeys an approximate ( functional ) schrdinger equation , @xmath178 where @xmath179 denotes the hamiltonian for the non - gravitational fields @xmath175 . note that the expression on the left - hand side of is a shorthand notation for an integral over space , in which @xmath47 stands for functional derivatives with respect to the three - metric . semiclassical time @xmath10 is thus defined in this limit from the dynamical variables . * the next order of the born - oppenheimer scheme yields quantum gravitational correction terms proportional to the inverse planck mass squared , @xmath180 . the presence of such terms may in principle lead to observable effects , for example , in the anisotropy spectrum of the cosmic microwave background radiation . oppenheimer expansion scheme distinguishes a state of the form from its complex conjugate . in fact , in a generic situation both states will decohere from each other , that is , they will become dynamically independent @xcite . this is a type of symmetry breaking , in analogy to the occurrence of parity violating states in chiral molecules . it is through this mechanism that the i and the @xmath10 in the schrdinger equation emerge . the recovery of the schrdinger equation raises an interesting issue . it is well known that the notion of hilbert space is connected with the conservation of probability ( unitarity ) and thus with the presence of an external time ( with respect to which the probability is conserved ) . the question then arises whether the concept of a hilbert space is still required in the _ full _ theory where no external time is present . it could be that this concept makes sense only on the semiclassical level where holds . the major physical applications of quantum gravity concern cosmology and black holes . although the above presented formalism exists , as yet , only on a formal level , one can study models that present no mathematical obstacles . typically , such models are obtained by imposing symmetries on the equations @xcite . examples are spherical symmetry ( useful for black holes ) and homogeneity and isotropy ( useful for cosmology ) . quantum cosmology is the application of quantum theory to the universe as a whole . let us consider a simple example : a friedmann universe with scale factor @xmath181 containing a massive scalar field @xmath175 . in this case , the diffeomorphism constraints are identically fulfilled , and the wheeler dewitt equation reads @xmath182 this equation is simple enough to find solutions ( at least numerically ) and to study physical aspects such as the dynamics of wave packets and the semiclassical limit @xcite . there is one interesting aspect in quantum cosmology that possesses far - reaching physical consequences . because does not contain an external time parameter @xmath10 , the quantum theory exhibits a kind of determinism drastically different from the classical theory @xcite@xcite . consider a model with a two - dimensional configuration space spanned by the scale factor , @xmath183 , and a homogeneous scalar field , @xmath175 , see fig . ( such a model is described , for example , by with @xmath184 . ) the classical model be such that there are solutions where the universe expands from an initial singularity , reaches a maximum , and recollapses to a final singularity . classically , one would impose , in a lagrangian formulation , @xmath185 ( satisfying the constraint ) at some @xmath186 ( for example , at the left leg of the trajectory ) , and then the trajectory would be determined . this is indicated on the left - hand side of fig . 6 . in the quantum theory , on the other hand , there is no @xmath10 . the hyperbolic nature of a minisuperspace equation such as suggests to impose boundary conditions at @xmath187 constant . in order to represent the classical trajectory by narrow wave packets , the ` returning part ' of the packet must be present ` initially ' ( with respect to @xmath183 ) . the determinism of the quantum theory then proceeds from small @xmath183 to large @xmath183 , not along a classical trajectory ( which does not exist ) . this behaviour has consequences for the validity of the semiclassical approximation and the arrow of time . in fact , it may in principle be possible to understand the origin of irreversibility from quantum cosmology , by the very fact that the wheeler dewitt equation is asymmetric with respect to the intrinsic time given by @xmath183 . the framework of canonical quantum cosmology is also suitable to address the quantum - to - classical transition for cosmological variables such as the volume of the universe @xcite@xcite . using the approach of loop quantum gravity ( see thiemann s contribution ) one arrives at a wheeler dewitt equation in cosmology which is fundamentally a difference equation instead of a differential equation of the type . in the ensuing framework of loop quantum gravity it seems that the classical singularities of gr can be avoided . singularity avoidance for collapse situations can also be found from spherically symmetric models of quantum geometrodynamics . for example , in a model with a collapsing null dust cloud , an initially collapsing wave packet evolves into a superposition of collapsing and expanding packet @xcite . this leads to destructive interference at the place where the singularity in the classical theory occurs . other issues , such as the attempt to give a microscopic derivation of the bekenstein hawking entropy ( see the contribution by c. kiefer to this volume ) , have been mainly addressed in loop quantum gravity . a final , clear - cut , derivation remains , however , elusive . 99 anderson , a. and dewitt , b. ( 1986 ) . does the topology of space fluctuate . _ foundations of physics _ , * 16 * , 91 . geroch , r. ( 1967 ) . topology in general relativity . _ journal of mathematical physics _ , * 8 * , 782 - 786 . gibbons , g.w . and hawking , s.w . selection rules for topology change . _ communications in mathematical physics _ , * 148 * , 345 - 352 . giulini , d. ( 1992 ) . on the selection rules for spin - lorentz cobordisms . _ communications in mathematical physics _ , * 148 * , 353 - 357 . giulini , d. and kiefer , c. ( 1994 ) . dewitt metric and the attractivity of gravity . _ physics letters a _ , * 193 * , 2124 . giulini , d. ( 1995a ) . on the configuration - space topology in general relativity _ helvetica physica acta _ , * 68 * , 87 - 111 . giulini , d. ( 1995b ) . what is the geometry of superspace . _ physical review d _ , * 51 * , 5630 - 5635 . giulini , d. ( 1998 ) . on the construction of time - symmetric black - hole initial data . in _ black holes : theory and observation _ ( ed . f. hehl , c. kiefer , and r. metzler ) , pp.224243 . lecture notes in physics 514 . springer , berlin . hjek , p. ( 2003 ) . quantum theory of gravitational collapse ( lecture notes on quantum conchology ) . in _ quantum gravity : from theory to experimental search _ d. giulini , c. kiefer , and c. lmmerzahl ) , pp.255299 . lecture notes in physics 631 . springer , berlin . henneaux , m. and teitelboim , c. ( 1992 ) . _ quantization of gauge systems _ ( princeton university press , princeton ) . hojman , s.a . and kucha , k. and teitelboim , c. ( 1976 ) . geometrodynamics regained . _ annals of physics _ , * 96 * , 88135 . horowitz , g. ( 1991 ) . topology change in classical and quantum gravity . _ classical and quantum gravity _ , * 8 * , 587 - 602 isham , c. and kucha , k. ( 1985a ) . representations of spacetime diffeomorphisms i : canonical parametrised spacetime theories . _ annals of physics _ * 164 * , 288 - 315 . isham , c. and kucha , k. ( 1985b ) . representations of spacetime diffeomorphisms ii : canonical geometrodynamics . _ annals of physics _ * 164 * , 316 - 333 . joos , e. , zeh , h. d. , kiefer , c. , giulini , d. , kupsch , j. , and stamatescu , i .- o . ( _ decoherence and the appearance of a classical world in quantum theory _ , 2ndedn ( springer , berlin ) . kiefer , c. ( 2004 ) . _ quantum gravity _ ( oxford university press , oxford ) . kiefer , c. ( 2006 ) . quantum gravity : general introduction and recent developments . _ annalen der physik _ , * 15 * , 129148 . kucha , k. ( 1974 ) . geometrodynamics regained : a lagrangian approach . _ journal of mathematical physics _ , * 15 * , 708715 . sorkin , r. ( 1997 ) . forks in the road , on the way to quantum gravity . _ international journal of theoretical physics _ , * 36 * , 2759 - 2781 . zeh , h. d. ( 2001 ) . _ the physical basis of the direction of time _ , 4thedn ( springer , berlin ) . see also http://www.time-direction.de .

we give an introduction to the canonical formalism of einstein s theory of general relativity . this then serves as the starting point for one approach to quantum gravity called quantum geometrodynamics . the main features and applications of this approach are briefly summarized .

star formation occurs in the coldest phase of the interstellar medium ( ism ) , where the gas forms extended structures known as giant molecular clouds ( gmcs ) . however , the internal mechanisms within the clouds that produce a galaxy s star formation rate remain poorly understood . the clouds themselves can not be freely collapsing to form stars , since the rate of gas conversion over a cloud s free - fall time would lead to a star formation rate more than a hundred times higher than observed in the milky way @xcite . two main options exist for avoiding this conundrum . the first is that the clouds are destroyed by its first generation of stars , redistributing the majority of its gas back into the warm ism @xcite . however , previous simulations have failed to successfully disperse the clouds with stellar feedback @xcite . while it is possible that the feedback methods employed within the simulations are ineffective for numerical reasons , there is also strong observational evidence to suggest that clouds are not in free - fall . rather , internal turbulent motions and magnetic fields seem to control the dynamics of the clouds and prevent the rapid conversion of gas into stars @xcite . support for clouds being turbulent structures stretches back to the early 1980s with the seminal paper by @xcite . since then , it has become widely accepted that gmcs are supersonically turbulent and this must play a crucial role in star formation @xcite . this then leads to the question of the origin of this turbulence . without a fresh injection of energy , turbulent modes inside a cloud will decay over a crossing time , @xmath11 and the clouds would quickly collapse and form stars at a high rate @xcite . a force is therefore needed to drive the turbulence in the clouds , either from the inside through stellar feedback , or from the outside via shear or interactions with neighbouring clouds @xcite . evidence for the latter comes from previous simulations of cloud formation . even without the inclusion of star formation , global simulations forming clouds through self - gravitational collapse find good agreement with observations for the typical properties such as mass , radius and velocity dispersion @xcite . including turbulence in analytical theories of star formation also produces a good match with observed star formation rates and predictions of the initial mass function of stars @xcite and star formation rates and efficiencies @xcite . this suggests that gmc evolution may be largely governed by gravity and turbulence driven by interactions within their galactic environment . indeed , @xcite and @xcite find that cloud collisions are common , occurring multiple times per orbital period . observations also support this view , with suggestions that the turbulence itself is dominated by larger - scale modes and the existence of clouds such as the maddalena cloud and pipe nebula which have a low star formation rate @xcite . to understand how gmcs convert their gas into stars , we must first understand the properties of the turbulence . information about the turbulence can be found in the gas probability density function ( pdf ) . in an isothermal system , the gas density pdf has a log - normal form @xcite , with the width of the pdf increasing with mach number ; @xmath12 , where @xmath1 is the standard deviation of the gas density , @xmath13 , and @xmath14 is the mean density , @xmath2 is the rms mach number of the turbulence and @xmath0 is a proportionality factor related to whether the turbulence is driven by a solenoidal ( divergence - free ) driver or by a compressive ( curl - free ) driver @xcite . solenoidal and compressive driving of the turbulence gives rise to extremely different density distributions and the value of @xmath0 is determined by the relative mix of the two . in the case of fully solenoidal driving , the proportionality term becomes @xmath15 , while fully compressive driving yields @xmath16 @xcite . in a @xmath17 sized simulation of supernova - driven turbulence performed by @xcite , @xcite found a ratio of compressible to solenoidal velocity modes of about @xmath18 , which @xcite suggested may relate to an average @xmath19 . this suggests a significant fraction of compressible modes in the turbulence driving . observationally , @xcite estimated @xmath20 from @xmath21co line observations of the taurus molecular cloud , using the projected 2d column density . @xcite found a slightly lower ( more solenoidal ) value of @xmath22 , for the milky way s infrared dark ( non - star forming ) clouds , however , they did not take into account the magnetic pressure contribution to the standard deviation mach number relation . this can result in an underestimate of @xmath0 @xcite . @xcite also observed a non - star forming cloud and concluded the turbulence was compressively driven with a lower limit of @xmath23 . most recently , @xcite determined the @xmath24 parameter in the central molecular zone ( cmz ) cloud g0.253 + 0.016 ` brick ' , indicating primarily solenoidal driving due to strong shearing motions in the cmz . this is in contrast to clouds in the galactic disc for which @xmath19 , indicating strong contributions from compressive modes in the turbulence driving ( see review of the currently available disc clouds for which estimates of @xmath0 are available in * ? ? ? * ) . in this paper , we explore the turbulence properties of clouds formed in a high - resolution ( minimum computational cell size of @xmath25pc ) galactic - scale simulation . our main motivation is to explore how turbulence that is driven on the galactic scale through disc shear and cloud interactions travels down to the cloud scale . by measuring the value of @xmath0 in the resultant cloud population , we explore the relative strengths of the solenoidal and compressive modes and how these vary between the clouds and over their lifetime . no initial @xmath0 parameter is imposed on the simulation and the turbulence is allowed to develop self - consistently through hydrodynamical and gravitational interactions in the disc . section [ sec : methods ] explains the galaxy simulation and the methods for cloud identification and tracking , section [ sec : results ] presents our results and in section [ sec : conclusions ] , we summarise our conclusions . our simulation was performed using _ enzo _ ; a three - dimensional adaptive mesh refinement ( amr ) hydrodynamics code @xcite . the galaxy is simulated in a three - dimensional box of side @xmath26 with a root grid of @xmath27 cells and eight levels of refinement , giving a limiting resolution of @xmath28 . cells were refined whenever the jeans length dropped to less than four times the cell size @xcite or whenever the density contrast between neighbouring cells exceeded a factor of 10 . to ensure the galaxy is stable in the absence of cooling , the whole disc region is kept at a minimum of two levels of refinement in addition to the root grid ( @xmath29pc ) throughout the simulation , over a height of 1kpc above and below the disc . within the radii of our main region of analysis ( see below ) and up to 100pc from the mid - plane , all gas is kept refined to at least level three ( @xmath30pc ) . the adaptive mesh then acts to refine 23% of the gas to level four ( @xmath31pc ) , 10% of the gas to level five ( @xmath32pc ) , 4% to level six ( @xmath33pc ) , 1.3% of the gas to level seven ( @xmath34pc ) and 0.5% of the gas to level eight ( @xmath35pc ) . however , within the clouds themselves , 89% of the gas is covered by level eight . _ enzo _ evolves the gas using a three - dimensional version of the zeus hydro - scheme , which uses an artificial viscosity term to handle shocks @xcite . the variable associated with this is the quadratic artificial viscosity term and was set to its default value of @xmath36 . the gas cools radiatively to @xmath37 using a one - dimensional cooling curve created from the cloudy package s cooling table for metals and _ enzo _ s non - equilibrium cooling rates for atomic species of hydrogen and helium @xcite . this minimum value is close to that of the gmcs . at this temperature , the sound speed of the gas is @xmath38 . the weakness of this model is that the gas is assumed to be optically thin and non - molecular ; both untrue within the gmcs . for identifying the gmcs themselves , the lack of molecular gas is not necessarily an issue , since it is thought to be a good tracer the overall gas distribution @xcite . the problem with the cloud internal dynamics is fundamentally a problem with resolution which we discuss more thoroughly at the end of this paper . to prevent unresolved gas from collapsing , a pressure floor is imposed when the jeans length becomes less than four times the cell size @xcite . when this operates , an artificial pressure is calculated which results in a change to the velocity flow into that cell , suppressing collapse . the true pressure and internal energy is unchanged , so this adjustment is not advected through the simulation but calculated locally when required . the disc also includes a radially dependent photoelectric heating described in the detail in @xcite , but does not include any star formation , stellar feedback or magnetic fields . the simulation therefore concentrates on the early stages of star formation and the environment of the young gas cloud evolution . to minimise artefacts from the orientation of the cartesian mesh , the simulation sets a co - rotation point at a radius of @xmath39 @xcite . at this radius , gas does not move with respect to the grid , removing potential numerical artefacts during the circular gas flow through the cartesian system . to maximise our efforts on this region , the highest refinement level is confined to an annulus of thickness @xmath40 over the co - rotation point , from @xmath41 to @xmath42 . our clouds are therefore analysed from inside this region . this refinement strategy allows us to achieve the best possible resolution for the clouds and at the same allows us to maintain realistic boundary conditions by modelling an entire galactic disc . the galaxy itself is an isolated milky way - type disc , with a flat rotation curve and circular velocity of @xmath43kms@xmath44 . as the disc cools , it gravitationally fragments as the toomre @xmath45 parameter drops below the stability threshold . toomre @xmath45 is defined as @xmath46 , where @xmath47 is the epicyclic frequency , @xmath48 is defined as @xmath49 where @xmath50 is the 1d velocity dispersion of the gas motion in the plane of the disc and @xmath51 is the sound speed ( our discs are initialized with @xmath52 ) , and @xmath53 is the gas surface density . a value @xmath54 corresponds to a gravitationally stable disc , while @xmath55 suggests an unstable state . in our main disc region between @xmath56@xmath57 , we set the @xmath45 initially to @xmath58 . as the disc cools , the gas fragments into the clouds , which can then grow and evolve through mergers and interactions . the initial fragmentation and evolution of the gas disc is described in @xcite . by modelling the global disc , the effect of galactic shear and cloud interactions are naturally included . these are some of the driving forces for the cloud internal turbulence @xcite with many other driving mechanisms contributing @xcite . we note that magnetic fields were not included in the simulations . magnetic fields are important for the structure and dynamics of the ism @xcite , but instead of including all of the relevant detailed physics , we here focus on improving the aspect of initial and boundary conditions for cloud formation and dynamics . follow - up simulations are need that include magnetic fields , detailed molecular chemistry and cooling , as well as star formation feedback , such as supernova explosions @xcite . the strength of the present simulations is that they model a global galactic disc , thus providing better initial and boundary conditions for cloud formation than previous simulations . we identify clouds in our simulation using constant density contours at @xmath59 ( @xmath60 ) , where the clouds are then the coherent structures within the contour line ( as in * ? ? ? the threshold density for the cloud contour was selected based on the observed mean density of typical gmcs . because the internal properties of the smaller clouds will be poorly resolved , we restrict our cloud analysis to clouds containing more than 1000 cells , giving an approximate 10 cells in each dimension , which is a the absolute minimum to capture the main fraction of rotational modes @xcite . the exception to this is the resolution studies in section [ sec : resolution ] , where differences in cell size result in too stringent a sub - selection of the cloud population , so we do not impose a minimum number of cells for the presentation of the resolution study below . at our two main analysis times of 60myr and 240myr , the total number of clouds in our highest - resolution simulation ( @xmath61 ) analysed is 1529 and 1313 , respectively . to follow the evolution of the clouds , we output the simulation data every @xmath62 and track the clouds between these data sets , assigning a unique tag number to the same structure at different times . if a cloud disappears during the evolution , we search for possible merger events and if unsuccessful , mark it as dispersed . a detailed description of the tracking algorithm can be found in @xcite , although the cloud identification matches that in @xcite . as described in the introduction , a key value for quantifying the cloud internal turbulence is the @xmath0 parameter , defined as the proportionality constant of the standard deviation mach number relation for non - magnetized and non - self - gravitating gas . for each cloud , @xmath63 where @xmath1 is the standard deviation of the gas density , @xmath13 , within the cloud , normalised by the cloud mean density , @xmath14 . the cloud s turbulent mach number , @xmath64 , is given by the volume - weighted average of the local ( cell - based ) mach number , where @xmath65 and @xmath51 are the three - dimensional velocity dispersion and sound speed , respectively . finally , @xmath0 is the turbulence driving parameter @xcite , which we concentrate on measuring here . this can be related to the gas density pdf if the latter is assumed to follow a log - normal distribution , @xmath66 , \label{eq : log_normal}\end{aligned}\ ] ] where @xmath67 is the standard deviation of the logarithmic gas density with @xmath68 . in which case , @xmath0 can be related to @xmath67 with @xmath69 . however , note that this expression is dependent on the log - normal distribution of the gas , while equation [ eq : b_parameter ] applies for any density pdf @xcite . the turbulence driving parameter @xmath0 is determined by the mix of the solenoidal and compressive turbulent modes in the acceleration field that drives the turbulence . for fully solenoidal forcing ( divergence - free ) , @xmath70 , while fully compressive forcing ( curl - free ) finds @xmath71 in three - dimensional simulations @xcite . to estimate @xmath0 from the turbulent motions inside a cloud , we must first subtract the cloud s global motions @xcite . the velocity dispersion is therefore calculated after the subtraction of the linear and angular momentum from the cloud motion . this is done by first subtracting the bulk velocity from the cloud s centre - of - mass . the cloud can then be regarded as a rigid body whose circular velocity about the core can be subtracted : @xmath72 , where @xmath73 is the position vector relative to the centre - of - mass , @xmath74 is the angular velocity and @xmath75 is the total angular momentum . the three - dimensional velocity dispersion about the cloud s centre - of - mass then becomes , @xmath76 where @xmath77 is the velocity of the gas , @xmath78 is the cloud s centre - of - mass velocity and @xmath79 is the circular velocity from the angular momentum about the cloud s centre . the co - rotating , high - resolution region where we define the clouds is within the two white circles . ] the face - on view of the surface density in the whole galactic disc at @xmath80 ( one orbital period at the outer edge of our analysis region at @xmath81kpc . ) is shown in figure [ fig : figure1 ] . the two white circles mark the inner and outer boundary of the high - resolution co - rotating region where we analyse the clouds . using this technique we have a region with high resolution ( @xmath82 ) and at the same time include the global disc potential , with the appropriate boundary conditions for the high - resolution region . since we do not drive a global spiral potential , the disc gas has a predominately flocculent structure , with the dense knots of cloud gas being connected by a filamentary warm ism . the environment is clearly not quiescent , but a continuously interacting medley of collisions and tidal forces that shape the clouds . figure [ fig : figure2 ] describes the thermodynamical condition in the whole disc ism via a two - dimensional density - temperature diagram of the distribution in the gas mass . notably , there is no clear division in the gas between the hot , warm and cold ism phases , but a continuous distribution connecting the cold clouds with the surrounding material . at densities @xmath83 , the distribution becomes a steeply rising linear relation corresponding to the artificial pressure floor defined in section [ sec : code ] . in this region , gas follows a polytrope with an adiabatic index of @xmath84 , which is shown by the dashed black line . the pressure floor halts the collapse and prevents the formation of isolated cells with very high densities in which the jeans length would otherwise not be resolved . while this regime was necessary to minimise numerical artifacts , it is clear that even on parsec scales , it is hard to resolve the molecular gas component of the clouds . exactly where the simulations start to become resolution limited will be studied further in section [ sec : resolution ] . ) for the high - resolution region of the disc ( between the white lines in figure [ fig : figure1 ] ) with a height of @xmath85 . ] the evolution of the structure of the ism can be explored using the density . figure [ fig : figure3 ] shows the volume - weighted density pdf for the high - resolution region of the galaxy marked between the two white circles in figure [ fig : figure1 ] . four different simulation times are shown at 30myr , 60myr , 120myr and 240myr , with the additional vertical black dotted line showing the location of the cloud identification threshold , which sits at @xmath59 . at the earliest time of 30myr , the galaxy disc is only partially fragmented . the pdf has a narrow profile that broadens as clouds form and begin to interact in the disc . from 60myr to 240myr , the profile continues to broaden but at a slower rate . equation [ eq : b_parameter ] implies that the broadening of the pdf profile is due to an increase in the standard deviation of the logarithmic density . previous research has suggested that this can occur due to a shift from solenoidal turbulence driving to compressive driving for the same mach number @xcite . the evolution from a narrow to wide density pdf may therefore indicate a change from a largely solenoidal turbulence - dominated system to a more compressive regime as cloud and gas structures interact and collide . the cloud gas ( right of the vertical dashed line ) extends smoothly to higher densities , but with a profile that is not completely log - normal . while a log - normal distribution is expected from isothermal studies , deviations are frequently seen in other models due to shocks and other intermittent events @xcite . moreover , our gas is not isothermal and the highest density end may be gravitationally collapsing , with support coming from the pressure floor ( @xmath86 ) rather than internal turbulence . clouds are formed in the galaxy disc through hydrodynamical and gravitational fragmentation . figure [ fig : figure4 ] shows a close - up of this process , displaying the surface density of the disc in a 2kpc region at early ( 60myr ) and late ( 240myr ) times . both images are taken within the analysis region shown in figure [ fig : figure1 ] . the cloud environment changes substantially between these stages . at early times , the disc is in the process of fragmenting . its rotational motion supports the gas against radial collapse , causing the initial structures to be tangential filaments . as the disc becomes more unstable , the rotation can no longer support the gas which collapses to create gmcs . as the dynamical time is shortest in the inner disc , the left - hand image in figure [ fig : figure3 ] shows this process occurring from right to left , with the outer gas on the right - hand side in an earlier state of collapse compared to the inner left - hand edge . by @xmath80 , the disc has completely fragmented and interactions between the clouds now dominate the local dynamics . close - ups of two typical clouds in the simulation are shown in figure [ fig : figure5 ] . the properties for these clouds match the peak values in the distributions shown in figure [ fig : figure6 ] . the left - hand cloud shows an interacting structure fragmented into three parts , while the right - hand cloud is a more quiescent object with a single core . despite these visual differences , both clouds have similar mass , radii and velocity dispersions . figure [ fig : figure6 ] turns to the properties of the clouds themselves , showing the distribution in their values at the early ( 60myr in blue dashes ) and late ( 240myr in green solid ) evolution times . as mentioned in section [ sec : cloud_definition ] , only clouds with more than 1000 cells are included in this analysis . overlaid with the red dotted line are the observational results from the grs milky way survey from @xcite . the average and the standard deviation of cloud properties are summarised in table [ tab : table1 ] . .average and standard deviation of the cloud properties for clouds with more than 1000 cells ( mass @xmath87 , radius @xmath88 , 1d velocity dispersion @xmath89 , virial parameter @xmath90 , mach number @xmath2 and density standard deviation @xmath1 ) at @xmath91 and @xmath80 . [ cols="^,^,^,^,^,^ " , ] the top - left plot shows the distribution of the cloud mass . at 60myr , the peak value sits at @xmath92 . this has shifted slightly lower by 240myr . the change is due to the clouds fragmenting into smaller structures as the disc finishes its period of instability . both plots overlay the observed milky way profile , with approximately equivalent peak values . the simulation results have a high - mass tail due to the lack of star formation to remove the densest gas which increases over time . this creates a population of massive clouds whose bulk has built up via successive mergers and accretion . its existence underscores the importance of interactions in the evolution of the clouds . the high - mass tail causes the standard deviation in cloud mass at @xmath80 ( @xmath93 ) to be significantly larger than that at @xmath94 ( @xmath95 ) , as shown in table [ tab : table1 ] . note that the mass here is plotted on a logarithmic scale ( non - gaussian distribution ) , indicating that the clouds in this later stage have a wide range of masses . the smaller clouds in the milky way data are not seen in the simulation plot , as we focus on our best - resolved population containing more than 1000 cells per cloud . the distribution in cloud radius is shown in the top right plot of figure [ fig : figure6 ] . we define the average radius of the cloud as @xmath96 where @xmath97 is the projected area of the cloud in the x - y plane , @xmath98 in y - z plane and @xmath99 in z - x plane . the peak cloud radius remains the same between 60myr and 240myr , but the profile becomes broader at later times . the average radius remains the same , with @xmath100 . however , figure [ fig : figure4 ] suggests that the cloud shape will be changing more considerably as it goes from being dominated by a fragmenting filament to one controlled both by the warm ism filaments and cloud interactions . the range in values at both times are consistent with those measured in the milky way , but once again we omit the smaller clouds from our analysis . the middle - left panel shows the distribution of the one - dimensional volume - weighted velocity dispersion after the subtraction of the rotational velocity . the evolution shows a broadening in the range of values , with a late population of high velocity dispersion clouds . these likely correspond to the high - mass tail seen in the mass distribution , whose increased gravitational potential from successive mergers boosts the velocity dispersion . while this larger population is not seen in the milky way the peak value at both 60myr and 240myr is close to that seen observationally at about @xmath9@xmath101 . the virial parameter , @xmath90 , describes the ratio of kinetic to gravitational energy using observational parameters . the value is defined as @xmath102 where @xmath103 , @xmath104 is the volume - weighted average cloud temperature , @xmath105 and @xmath106 , the molecular weight in a gmc . the 1d velocity dispersion is @xmath107 , where the velocity components @xmath108 , @xmath109 , and @xmath110 were corrected for linear and angular momentum contributions according equation [ eq : veldisp_3d ] , @xmath88 is the cloud radius and @xmath87 is the cloud mass . a value of @xmath111 implies that kinetic energy is dominant , while @xmath112 indicates that gravity is dominant and therefore , the cloud is gravitationally bound @xcite . the virial parameter distribution is shown in the middle right panel of figure [ fig : figure6 ] . the averages given in table [ tab : table1 ] show @xmath113 at @xmath94 and @xmath114 at @xmath80 . at later times , the virial parameter average shifts to a slightly higher value . however , the distribution shows that the peak value remains unchanged at just greater than unity , suggesting most clouds are borderline bound throughout their lives . at later times , a population of less - bound clouds appears , suggesting that the products of successive mergers and interactions tend to not be gravitationally bound . the milky way data likewise finds that clouds sit on the brink of being bound , with the data showing a broad range of values . there are clearly differences between the simulated and observed distributions , but given the uncertainties in both the simulations and the observations , we do not expect to find significantly better agreement . left on the bottom row shows the distribution in the cloud mach number , @xmath64 , while the bottom right is the density standard deviation . these are the two properties that control the value of the turbulence parameter , @xmath0 , which we will focus on in the next subsection . both distributions have a minor dependence on time . the peak value of mach number at @xmath80 is @xmath115 , which is slightly lower than @xmath116 at @xmath94 . this comes from the broader range in velocity dispersion at later times that we saw in the plot above ; the broadening at 240myr causes the velocity dispersion to peak at a lower value . a similar but weaker trend can be seen in the distribution of the standard deviation , @xmath1 ; namely clouds in the later stage generate lower @xmath1 than early on in the simulations . the distribution is also slightly broader , with clouds at 240myr showing a more varied internal density range than at 60myr . this likely stems directly from each cloud s history , determined by interactions , birth location and the local environment . the average values for the standard deviation in table [ tab : table1 ] are high due to being dominated by 1% of outlying clouds that are undergoing gravitational collapse . and the standard deviation of density fluctuations @xmath1 . teal points show individual clouds at @xmath94 and purple points are clouds at @xmath80 . the lines show the gradient corresponding to different values of @xmath0 , based on equation [ eq : b_parameter ] . ] one of the most important results in our study is shown in figure [ fig : figure7 ] . it compares the standard deviation @xmath1mach number @xmath2 relation for the clouds at the early @xmath91myr simulation time and those at @xmath80 to investigate the change in turbulence modes over the course of the simulation . as with figure [ fig : figure6 ] , we only plot clouds with more than 1000 cells . the purple points correspond to clouds at @xmath80 and the teal points for @xmath94 , with the red line showing the gradient for @xmath117 ( pure compressive driving ) in equation [ eq : b_parameter ] and the green and blue lines marking @xmath118 and @xmath119 ( pure solenoidal driving ) , respectively . surprisingly , despite the significant difference in disc structure at the 60myr and 240myr times shown in figure [ fig : figure4 ] , the best fit for the cloud turbulence modes is @xmath120 for both times . the reason for this agreement can be seen from the last panels in figure [ fig : figure6 ] , where the peaks in both the mach number and standard deviation @xmath1 shift slightly towards lower values with time , compensating for one another to give the same average @xmath0 . physically speaking , this supports the picture that the clouds break into smaller clouds during the disc gravitational fragmentation and subsequent cloud interactions . while some clouds merge to form the high - end tail of the mass distribution , a substantial fraction of clouds have a lower mach number and density contrast post - fragmentation , which balance to give a similar mix of turbulence modes . this average value of @xmath0 indicates that the clouds in the galaxy simulation are typically dominated by solenoidal turbulence driving . however , in the next section we will note this is a strictly lower limit , because of the limited numerical resolution currently achievable in these simulations of disc galaxies . more important than the average value of @xmath0 ( which is not converged with resolution ) , we find a very wide distribution of @xmath0 , with values covering the full range from purely solenoidal ( @xmath4 ) to purely compressive driving ( @xmath5 ) and mixed values in between . thus , the turbulence driving parameter for individual clouds is far from constant . this shows that @xmath0 should be considered a varying parameter that changes between individual clouds . the range of @xmath0 values in the cloud population increases with simulation time , as the distribution at 240myr in figure [ fig : figure7 ] is somewhat broader than at 60myr . this suggests that a cloud s history or its environment might be controlling the evolution of @xmath0 . to test the idea that clouds are born with a constant @xmath0 and interactions drive more compressive or solenoidal driving , we plot the same relation as in figure [ fig : figure7 ] for clouds of different ages . this is shown in figure [ fig : figure8 ] . for cloud populations at different ages . the clouds are tracked every @xmath62 from @xmath121 to @xmath80 and then grouped into populations based on the time since the cloud formed . the age of clouds with blue data points is @xmath122@xmath62 , green is @xmath9@xmath123 , red is @xmath124@xmath125 and cyan cloud points are for clouds with age @xmath126@xmath127 . the lines show different values of @xmath0 , with the blue - dashed , green - solid , red - dotted and cyan - dot - dash showing the best fit for @xmath0 in each of the cloud age data sets with the same colour . ] as described in section [ sec : cloud_definition ] , we followed the evolution of @xmath0 for individual clouds by tracking them between simulation outputs . we did this for clouds between 220240myrs , during the fully fragmented phase of the disc where cloud interactions dominated their evolution . in figure [ fig : figure8 ] , clouds that are between @xmath122@xmath128myr old are marked in dark blue , clouds between @xmath9@xmath129myr old are in green , @xmath124@xmath130myr old clouds are in red and @xmath126@xmath131myr old clouds are cyan points . lines of the same colour show the best - fit @xmath0 parameter for each population . there is initially a small amount of evolution in @xmath0 as the cloud ages , moving from an average @xmath132 for the youngest clouds towards an average @xmath133 for clouds between @xmath124@xmath130myr . however , this number decreases again slightly for the oldest clouds . while not major changes , these are likely to be due to the changes in compressive modes as the cloud undergoes interactions during its life . however , this effect is clearly weak , suggesting no strong relation between the turbulence modes and cloud mergers or lifetime . we also attempted to find a stronger correlation by examining the values of @xmath0 before and after individual recorded mergers . no emerging trend was seen , with @xmath0 staying approximately constant during the interaction . it appears instead that clouds overall retain the @xmath0 they are formed with , although this value can take on a wide range of numbers . , ( thin line histogram ) and its turbulent component , @xmath134 ( thick line histogram ) . the mean and standard deviation of these distributions are @xmath135 and @xmath136 . the thin and thick curves are log - normal fits to the distributions as in @xcite . ] our main goal here is to quantify the relative strength of solenoidal to compressive modes in the turbulence driving of molecular clouds formed by galactic dynamics and cloud - cloud collisions . an important diagnostic of this is the _ compressive ratio _ @xcite , which quantifies the fraction of compressive modes in the turbulent velocity field . here we follow the recent definition of the compressive ratio proposed in @xcite , @xmath137 where @xmath138 and @xmath139 are the compressive and solenoidal components of the velocity field , respectively . the two velocity components are derived with the helmholtz decomposition of the velocity field in fourier space ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? in addition to the total compressive ratio @xmath140 given by equation ( [ eq : compratio ] ) , we consider a purely turbulent version , @xmath134 , where we subtracted the centre - of - mass velocity as well as the mean velocities @xmath141 , @xmath142 and @xmath143 in spherical shells centred on the centre of mass of each cloud in spherical coordinates ( @xmath144,@xmath145,@xmath146 ) @xcite . this removes systematic motions from the velocity field such that only turbulent fluctuations remain . this method is similar to the method used in @xcite to remove systematic motions and to isolate the turbulence in the clouds . figure [ fig : figure9 ] shows the frequency distribution of the total compressive ratio @xmath140 and the turbulent compressive ratio @xmath134 of all the clouds with at least 1000 cells in our sample . similar to the distributions of the turbulent driving parameter @xmath0 , we find broad distributions of the compressive ratio , with values of @xmath140 covering a wide range from purely solenoidal ( @xmath147 ) to strongly compressive ( @xmath148 ) . the mean and standard deviation of these distributions are @xmath135 and @xmath136 , suggesting that most clouds are dominated by solenoidal modes . this is similar to the conclusion drawn from the sample of clouds in @xcite , who find broad distributions and typical values of @xmath149 from simulations of supernova driving . our values of @xmath140 are slightly lower , which is expected for turbulence purely driven by galactic dynamics ( which drives solenoidal motions due to galactic shear ) compared to turbulence driven by supernova explosions ( which drive more compressive modes compared to galactic shear ) @xcite . in agreement with @xcite we do not find a significant difference between the total and turbulent compressive ratio . the most likely explanation for the similarity of @xmath140 and @xmath134 is that for @xmath134 one removes systematic compressive motions ( infall ) from the numerator @xmath150 and at the same time removes systematic solenoidal motions ( large - scale rotation ) from the denominator @xmath151 , such that the ratio of the two nearly stays the same . in summary , we find broad distributions of the compressive ratio of our cloud sample formed by galactic dynamics and cloud - cloud collisions , with typical values of @xmath152 , i.e. , dominated by solenoidal modes . following the suggestions by @xcite , this value of @xmath140 may be related to the turbulence driving parameter @xmath153^{1/2}\sim0.4 $ ] , which is broadly consistent with the average value of @xmath0 found directly in section [ sec : drivingmode ] . , with resolution . we find that the average @xmath0 value increases systematically with increasing resolution , which means that our measurements represent lower limits . however , every resolution shows that @xmath0 varies by factors of a few within the cloud population . ] one important consideration in our quantification of the turbulence driving parameter @xmath0 is the effect of numerical resolution of the simulations . while the simulation refines down to cells with sizes @xmath154 , the internal cloud turbulence is a challenging property to resolve and converge on . to test the robustness of our quantitative values , we compared simulations at four different refinement levels , corresponding to limiting resolutions ( smallest cell size ) of @xmath155 , @xmath156 , @xmath157 and our main simulation resolution at @xmath154 . the resulting values for the cloud parameter @xmath0 at each resolution are shown in figure [ fig : figure10 ] . the dark blue line shows the distribution for our main simulation run , while green , red and cyan mark the distributions for simulations of decreasing resolution . the legend shows the average @xmath0 value with its standard deviation . due to the large size of the cells in the lower - resolution run , figures [ fig : figure10 ] and accompanying figure [ fig : figure11 ] show all clouds in the simulation , not only those with cell number above 1000 . our resolution test shows a systematic dependence of @xmath0 on @xmath158 , with the turbulence value increasing at higher resolutions ( decreasing @xmath158 ) . in our lowest resolution simulation with @xmath155 ( cyan line ) , the average @xmath0 for all clouds is 0.09 . this increases to @xmath18 for a limiting resolution of @xmath154 , showing an increase in the resolved compressive modes . such a trend means that we can not determine an absolute value for @xmath0 in our global models and the average value of @xmath120 should be considered a lower limit even in the highest - resolution simulations . the standard deviation in the values for @xmath0 also increases with resolution , although this is likely to be due to the wider range in cloud sizes that the smaller cells allow , as we will investigate next . the origin of the resolution dependence of @xmath0 can be seen in figure [ fig : figure11 ] . the three panels show how the profiles for the 3d volume - weighted mach number , standard deviation @xmath1 and the cloud radius vary with resolution . the mach number profile shows a nearly constant distribution shape past our lowest resolution level , suggesting that this property has converged in the simulations . however , both the standard deviation of density fluctuations and the radius of the clouds shift to systematically higher and lower values , respectively , with increasing resolution . full convergence is likely to be achieved when the simulation resolves the sonic scale ( @xmath159pc ) , where the turbulence transitions from supersonic to subsonic @xcite . this is because below the sonic scale , the turbulent density fluctuations become small compared to the large - scale supersonic fluctuations . the sonic scale is also the scale on which filaments and dense cores are expected to form @xcite . unfortunately , this is still out of reach for our global galaxy simulations . future galaxy simulations , however , will be able to reach down to sub - pc scales . for the main result of this work , we emphasise that despite the resolution dependence of the exact value of @xmath0 , all resolutions show that the cloud population has a wide distribution of @xmath0 values , indicating that galaxy dynamics and cloud - cloud collisions alone ( without feedback from star formation ) can produce the full range of cloud driving parameters , from purely solenoidal driving ( @xmath4 ) to purely compressive driving ( @xmath5 ) . we thus expect that these clouds can have a wide range of star formation activity , as star formation depends critically on the driving and statistics of the turbulence @xcite . we performed a three - dimensional hydrodynamical simulation of a global isolated disc galaxy and explored the turbulence properties in the giant molecular cloud population formed through self - gravity and cloud interactions . the turbulent modes were quantitatively explored by using the turbulence driving parameter @xmath0 , which is around @xmath18 for solenoidal modes and @xmath128 for purely compressive driving of the turbulence . our main results are as follows . 1 . in the early and the late stages of galaxy disc evolution , the average @xmath0 for the cloud population is almost the same at @xmath160 . this is despite the morphology of the disc structure being quite different , due to only being partially fragmented at the earlier times . our subsequent resolution study indicated that this average value of @xmath0 is a strictly lower limit on the true value and resolution down to @xmath161pc is likely required to achieve convergence . 2 . the distribution of @xmath0 exhibits wide scatter through the cloud population with @xmath0 varying by factors of a few . this means that each cloud can have a significantly different turbulence driving mode , ranging from purely solenoidal , divergence - free driving ( @xmath162 ) to purely compressive , curl - free driving ( @xmath163 ) . the main result of a wide distribution of turbulence properties of the clouds is supported by calculations of the compressive ratio ; the ratio of compressive to solenoidal modes in the turbulent velocity field , @xmath140 . values of @xmath140 likewise took on a wide range , with the mean value of @xmath164 , suggesting primarily solenoidal modes on average , but with individual clouds having values reaching far into the highly compressible regime . we tracked the clouds through the simulation to compare changes during the cloud lifetime . clouds live in a highly dynamic environment with multiple interactions occurring over the cloud s life . despite this , we find that the @xmath0 distribution does not depend significantly on the cloud age . we conclude that the turbulent modes within gmcs do not significantly depend on the evolutionary stage of the galaxy disc , nor on the merger history of a cloud . however , there is always a wide distribution of modes found within the cloud population . this means that a constant @xmath0 can not be selected , but must instead be treated as a varying parameter when looking at individual clouds . we thank the anonymous referee for their detailed comments , which helped to improve this work . numerical computations were carried out on cray xc30 at the center for computational astrophysics ( cfca ) of the national astronomical observatory of japan . the authors would like to thank the yt development team for many helpful analysis support @xcite . c.f . acknowledges funding provided by the australian research council s discovery projects ( grants dp150104329 and dp17010603 ) . further thanks the leibniz rechenzentrum and the gauss centre for supercomputing ( grants pr32lo , pr48pi and gcs large - scale project 10391 ) , the partnership for advanced computing in europe ( prace grant pr89mu ) , and the australian national computational infrastructure ( grant ek9 ) , as well as the pawsey supercomputing centre with funding from the australian government and the government of western australia . was partially funded by the mext grant for the tenure track system . d.m.s . gratefully acknowledges funding from the australian government s new colombo plan . this work is partly supported by jsps grant - in - aid for scientific research number 15k0514 . , g. l. , & norman , m. l. 1997 , in astronomical society of the pacific conference series , vol . 123 , computational astrophysics ; 12th kingston meeting on theoretical astrophysics , ed . d. a. clarke & m. j. west , 363

we determine the physical properties and turbulence driving mode of molecular clouds formed in numerical simulations of a milky way - type disc galaxy with parsec - scale resolution . the clouds form through gravitational fragmentation of the gas , leading to average values for mass , radii and velocity dispersion in good agreement with observations of milky way clouds . the driving parameter ( @xmath0 ) for the turbulence within each cloud is characterised by the ratio of the density contrast ( @xmath1 ) to the average mach number ( @xmath2 ) within the cloud , @xmath3 . as shown in previous works , @xmath4 indicates solenoidal ( divergence - free ) driving and @xmath5 indicates compressive ( curl - free ) driving . we find that the average @xmath0 value of all the clouds formed in the simulations has a lower limit of @xmath6 . importantly , we find that @xmath0 has a broad distribution , covering values from purely solenoidal to purely compressive driving . tracking the evolution of individual clouds reveals that the @xmath0 value for each cloud does not vary significantly over their lifetime . finally , we perform a resolution study with minimum cell sizes of @xmath7 , @xmath8 , @xmath9 and @xmath10 and find that the average @xmath0 value increases with increasing resolution . therefore , we conclude that our measured @xmath0 values are strictly lower limits and that a resolution better than @xmath10 is required for convergence . however , regardless of the resolution , we find that @xmath0 varies by factors of a few in all cases , which means that the effective driving mode alters significantly from cloud to cloud . hydrodynamics - turbulence - methods : numerical - ism : clouds - ism

an isothermal fluid model has been presented in the literature @xcite . it successfully reproduces most of the major features of recent experimental work . this model was motivated by trends observed in more sophisticated treatments . the basic ideas of the model will be reviewed here , and an extension to the case of a gaussian distribution with elliptical symmetry will be presented . the initial ion density distribution is proportional to the gaussian distribution of the neutral atom cloud from which the plasma is created . for a spherically symmetric cloud , the initial distribution can be written as @xmath2 . because the electrons thermalize much faster than the ions , in this model we take the initial electron density distribution to be the thermal equilibrium distribution , given by the boltzmann factor : @xmath3\ ] ] the lowest temperature plasmas are nearly charge - neutral , and the electron density distribution is approximately equal to the ion density . in this limit , it is shown in @xcite that for a spherically symmetric plasma that to within an arbitrary additive constant the electrical potential energy can be written as @xmath4 = - { k_b t \beta r^2 } .\ ] ] the force is the negative gradient of this potential energy . it is manifestly radial , spherically symmetric , and linearly proportional to the radial coordinate @xmath5 , measured from the center of the plasma . the velocity , which is the time integral of the acceleration , is also linearly proportional to @xmath5 . the consequence is that if the distribution is gaussian initially , it will remain gaussian at all times in the expansion . for the case of non - spherical symmetry , the approach is more or less the same , although the isothermal nature of the plasma has a more restricted meaning . we will take the initial ion density distribution to be gaussian , symmetric in the @xmath6 plane , and initially elongated in the @xmath7 direction : @xmath8 \label{eqn : density } , \ ] ] the initial conditions are @xmath9 and @xmath10 , and the parameter @xmath11 defines the elipticity of the system . the plasma fluid equations for our system are written @xmath12 following the derivation used in the case of spherical symmetry , the velocity is taken to be @xmath13 . inserting this and the density profile of eq . [ eqn : density ] into eq . [ eqn : continuity ] gives @xmath14 because @xmath15 , @xmath16 , and @xmath7 are independent variables , the only non - trivial solution of this equation is @xmath17 where we have dropped the subscripts because all components have this same form of solution . solving eq . [ eqn : newton ] requires a little more care . it is straight- forward to write down the acceleration vector following the derivation of the spherically symmetric case . however , the solution requires that the temperature be isothermal in a given dimension , but anisotropic in space . this condition allows the density distribution to reduce to the proper form in the limiting case of a plasma infinitely long in the @xmath7 dimension . this decouling requires energy to be conserved separately in the @xmath6 plane and in the @xmath7 dimension . the plasma equations for these two spaces are now exactly identical and completely separable . equation [ eqn : newton ] can be re - written ( dropping the subscripts ) as @xmath18 and the conservation of energy is @xmath19 where we have neglected the energy due to electron - ion recombination . equations [ eqn : gamma1 ] , [ eqn : velocity2 ] , and [ eqn : energy ] are exactly identical to eq . 2 of ref . @xcite . using eqs . [ eqn : gamma1 ] and [ eqn : energy ] , we now have both @xmath20 and @xmath21 in terms of @xmath22 . inserting this into eq . [ eqn : velocity2 ] gives @xmath23 where we have made the substitution @xmath24 . the solution to this equation is @xmath25 , where @xmath26 is an integration constant . the constant @xmath26 must be equal to zero to meet the condition that the ion velocity is initially zero at @xmath27 @xcite . the time evolution of the density and velocity functions can now be written as @xmath28 } { \pi^{3/2 } \sigma_1 ^ 2(t ) \sigma_2(t ) } \label{eqn : dens } \\ \vec{v}(\vec{r},t ) & = & v_e^2 t \left ( \frac{x}{\sigma_1 ^ 2(t)}\hat{x } + \frac{y}{\sigma_1 ^ 2(t)}\hat{y } + \frac{z}{\sigma_2 ^ 2(t)}\hat{z } \right ) \label{eqn : bigvelocity}\\ \sigma_1 ^ 2(t ) & = & \sigma_0 ^ 2 + v_e^2t^2 \\ \sigma_2 ^ 2(t ) & = & \alpha^2\sigma_0 ^ 2 + v_e^2t^2 .\end{aligned}\ ] ] we note that for two - dimensional planar gaussian charge distributions , a closed - form expression for the electric field has been derived @xcite . if such a compact analytical solution could be written in the three - dimensional case , it would remove the decoupling constraint that we imposed in order to solve the plasma equations . however , such a solution is not readily apparent . the geometry of our measurements is shown in fig . [ fig : fluorescence ] . in the experiment , the probe laser beam is spatially filtered and focused into the plasma with a confocal beam parameter that is long compared to all plasma dimensions . the position offset of this probe laser relative to the plasma is denoted by the parameter @xmath29 . after the plasma is created , the number of atoms in the column defined by the probe laser beam changes dramatically . when the laser beam passes through the center of the plasma ( @xmath30 ) , the number of ions in the beam monotonically decreases . however , if the laser is outside of the initial plasma distribution , as shown in fig . [ fig : fluorescence ] , the number first increases as the plasma moves into the laser beam , and then decreases as the plasma disperses . -dimension ( labeled region 1 ) . the probe laser is represented by the solid dark arrow parallel to the @xmath7 axis , displaced by a variable distance @xmath29 . at later times the plasma expands slowly in @xmath7 , but quickly in the @xmath6 plane . regions 2 and 3 represent the plasma size at later and later times . as the plasma expands , the number of atoms in the column defined by the probe laser changes . _ right _ a partial level diagram of ca@xmath31 . the probe laser is locked to the 397 nm resonance transition , and we measure 397 nm light scattered by plasma ions . , title="fig:",width=326 ] + the fluorescence signal depends on both the number of ions in the column defined by the probe laser beam ( gaussian beam profile with @xmath32 radius @xmath33 ) and the velocity distribution of the ions in the plasma . because the laser has a narrow bandwidth , atoms moving at velocities greater than @xmath349 m / s are doppler - shifted out of resonance . in this section we will use the results of the previous section to derive an expression for how the plasma ion fluorescence signal should change with time for different values of the offset parameter @xmath29 . the fluorescence signal @xmath35 is proportional to the absorption of the probe laser beam . using beer s law and the standard approximation of small optical depth , @xmath35 can be written as @xmath36 where @xmath37 is the spatial profile of the probe laser beam , and @xmath38 is the absorption cross section as a function of @xmath39 , the difference between the laser frequency and the atomic resonance frequency . removing the limitation of small optical depth is trivial . we can simplify this expression by setting the laser frequency equal to @xmath40 , and recognizing @xmath41 as the doppler shift due to the velocity @xmath42 of the atoms , where @xmath43 is the optical wavelength of the transition . equation [ eqn : bigvelocity ] gives the position - dependent velocity of the ions . we define a length @xmath44 , where @xmath45 is natural width of the transition , @xmath46 is the lifetime of the transitions s upper state , and use it to write the absorption profiles , @xmath47^{-1 } & \mbox{lorentzian , } \\ \\ \exp\left [ -2 \left ( { v_{th}}/{2 \gamma \lambda}\right)^2 \left({z}/{\ell } \right)^2 \right ] & \mbox{gaussian , } \end{array } \right.\ ] ] where @xmath48 is the absorption cross - section on resonance and @xmath49 is the rms velocity of a thermal distribution . the true absorption lineshape is better represented by a voigt profile . however , as the voigt profile can be approximated by a linear combination of the lorentzian and gaussian line profiles @xcite , we will only write down these two limiting forms . power broadening of the line can be included in a straightforward manner @xcite . we take the plasma density profile from eq . [ eqn : dens ] , and write the spatial profile of the probe laser beam as @xmath50,\ ] ] which corresponds to the geometry represented in fig . [ fig : fluorescence ] . performing the integration in eq . [ eqn : sig ] gives , @xmath51 where @xmath52 and @xmath53 . this expression for the lorentzian lineshape is proportional to eq . we create ultracold neutral plasmas by photoionizing laser - cooled calcium atoms in a magneto - optical trap ( mot ) . the calcium mot is formed in the usual way by three pairs of counter - propagating laser beams that intersect at right angles in the center of a magnetic quadrupole field @xcite . the 423 nm laser light required for the calcium mot is generated by frequency - doubling an infrared laser in periodically - poled ktp ( ppktp ) , and has been described previously @xcite . a diode laser master - oscillator - power - amplifier ( mopa ) system delivers 300 mw single frequency at 846 nm , as shown in fig . [ fig : exptsketch ] . this laser is phase - locked to a build - up cavity using the pound - drever - hall technique @xcite , giving a power enhancement of 30 . a 20 mm long ppktp crystal in the small waist of the build - up cavity is used to generate typically 45 mw output power at 423 nm @xcite . a schematic drawing of the mot laser system and frequency stabilization electronics used in these experiments.,width=326 ] the laser is further stabilized by locking the 423 nm light to the calcium resonance transition using saturated absorption spectroscopy in a calcium vapor cell @xcite . our vapor cell differs from ref . @xcite in that it has a stainless steel body with conflat metal seals and windows and a valve . an acousto - optic modulator ( aom ) in one arm of the saturated absorption laser beams shifts the laser frequency so that the laser beam sent to the mot is 35 mhz ( one natural linewidth ) below the atomic resonance . we also use the aom to chop this beam and use a lock - in amplifier to eliminate the doppler background in the saturated absorption signal . because the 846 nm laser is already locked to the frequency - doubling cavity , the feedback from this second lock circuit servos the frequency - doubling cavity length . the trap is loaded from a thermal beam of calcium atoms that passes through the center of the mot . the thermal beam is formed by heating calcium in a stainless steel oven to @xmath54 c. the beam is weakly collimated by a 1 mm diameter , 10 mm long aperture in the oven wall . the distance between the oven and the mot is approximately 10 cm . as the beam passes through the mot , the slowest atoms in the velocity distribution are cooled and trapped . an additional red - detuned ( 140 mhz , or four times the natural linewidth ) laser beam counter - propagates the calcium atomic beam , significantly enhancing the mot s capture efficiency . to prevent optical pumping into metastable dark states we also employ a diode laser at 672 nm . the density profile of the mot has an asymmetric gaussian profile and is well - represented by eq . [ eqn : dens ] with the peak density equal to @xmath55 @xmath56 , @xmath57 mm , and @xmath58 . we photo - ionize the atoms in the mot using a two - color , two - photon ionization process . a portion of the 846 nm diode laser radiation from the mot laser is pulse - amplified in a pair of yag - pumped dye cells and frequency doubled . this produces a 3 ns - duration laser pulse at 423 nm with a pulse energy around 1 @xmath59j . this laser pulse passes through the mot , and its peak intensity is a few thousand times greater than the saturation intensity . a second yag - pumped dye laser at 390 nm counter - propagates the 423 nm pulse and excites the mot atoms to low - energy states in the region of the ionization potential . we photo - ionize 85 - 90% of the ground - state atoms in the mot . the minimum initial electron temperature is limited by the bandwidth of the 390 nm laser to about 1 k. ions in the plasma scatter light from a probe laser beam tuned to the ca ii @xmath60 transition at 397 nm . the probe laser is generated by a grating - stabilized violet diode laser . this laser , as well as the 672 nm laser used in the neutral atom trap , is locked to the calcium ion transition using the davll technique @xcite in a large - bore , low - pressure hollow cathode discharge of our own design ( see fig . [ fig : cathode ] ) . ballast resistor is connected in series with the discharge . , title="fig:",width=326 ] + the probe laser is spatially filtered . the typical probe laser intensity is a few hundred @xmath59w focused to a gaussian waist of 130 @xmath59 m in the mot . we average repeated measurements of the scattered 397 nm radiation with the probe laser in a given position , denoted by the parameter @xmath29 in fig . [ fig : fluorescence ] . this produces a time - resolved signal proportional to the number of atoms resonant with the probe beam in a particular column of the plasma . by translating a mirror just outside the mot chamber , we scan the probe laser across the ion cloud . in this manner we obtain temporal and spatial information about the plasma expansion . one comparison of the isothermal model with experimental data was presented in ref . @xcite . in that work , the initial electron energy of the plasma , and therefore the expansion velocity @xmath61 , was fixed , and the parameter @xmath29 was varied from 0 to 4@xmath62 . in the following , we present a complementary comparison . the solid line in fig . [ fig:550 ] shows ion fluorescence signal with the probe laser tuned to the ion resonance frequency , and with the probe laser beam propagating along the @xmath7 axis ( @xmath30 in fig . [ fig : fluorescence ] ) for a range of initial electron energies . as discussed in ref . @xcite , the early rise in the fluorescence signal shows the increasing number of plasma ions in the @xmath63 level . this feature is easily explained in terms of the classic rabi two - level atom with damping . the probe laser intensity is in the range of 5 to 10 times the resonance rabi frequency . our numerical integration of the optical bloch equations mimics the approximately 20 ns rise time observed in the experimental data , and shows that the gaussian spatial profile of the probe laser beam washes out subsequent oscillations in the excited state fraction . following the initial rise , the fluorescence decays . at approximately @xmath64 ns the decay slows down due to correlation - induced heating in the plasma @xcite . ) . the electron energy , @xmath65 , is measured relative to the ionization limit . the solid line is the fluorescence signal , the dotted line is a fit using the lorentzian lineshape in eq . [ eqn : final ] . the model is normalized to the fluorescence signal at @xmath66 s , and the single fit parameter is the expansion velocity , @xmath61 . in the top panel , the plasma is spontaneously generated after exciting the atoms to a rydberg state with @xmath67.,width=326 ] relative to the center of the initial distribution . the solid line is the data , and the dotted line is the fit using the lorentzian lineshape in eq . [ eqn : final ] . the model is normalized to the peak of the data.,width=314 ] these two processes are not included in the isothermal model . we therefore begin the comparison of the model with the data at time @xmath66 s. the model is normalized to the data and fit using @xmath61 as the only fit parameter in an un - weighted least - squares fitting routine . the model uses the lorentzian lineshape in eq . [ eqn : final ] . the justification for using this lineshape arises from the analysis presented in ref . @xcite . in fig 2 of that reference , the @xmath27 velocity is 6 m / s . this is the rms velocity of a boltzmann distribution , and it is due almost entirely to correlation - induced heating . this velocity width gives a doppler width smaller than the natural line width , and the voigt profile is close to a pure lorentzian . furthermore , as the plasma evolves , the ion temperature falls due to the adiabatic expansion . the time scale for this is @xmath68s . this is shorter than the time scale for heating the ions by collisions with the electrons , @xmath69s . it is therefore not surprising that the gaussian lineshape gives a poorer fit to the data . in the fitting procedure , the gaussian lineshape produces expansion velocities that do not grow as the initial electron energy increases . the figure shows that the lorentzian model describes the data well over a few orders of magnitude in time . the velocities extracted from these data are shown in fig . [ fig : vel ] . the differences between the model and the signal are not negligible . for all initial electron energies , the model is slightly too low at @xmath70 ns , and too high at @xmath71s . the data are not corrected for optical pumping into the @xmath72states , which we measure to have a time constant of about 5 @xmath59s . the differences between the model and the fluorescence signal could indicate internal heating processes that manifest themselves in the ion velocity on the few hundred ns time scale . they could also indicate ions that appear in the plasma from rydberg states at late times or collective plasma density variations . these differences can be studied in future work . we also compare the fluorescence signal and the model over a range of initial electron energies with the probe laser beam shifted to @xmath73 . in this arrangement , the ion fluorescence signal is initially small , and grows as ions move into the probe laser beam . typical data are shown in fig . [ fig:600 ] . for these data , the model is fit to the @xmath74s signal using a least - squares procedure , with @xmath61 as the fit parameter . the velocity extracted from this fit is shown in fig . [ fig : vel ] . it is possible to fit the data using @xmath61 so that the peak in the model coincides with the peak of the fluorescence signal . these velocities are also plotted in the figure . and [ fig:600 ] . the triangles ( @xmath75 ) represent the velocity extracted from fig . [ fig:550 ] , and the circles ( @xmath76 ) from fig . [ fig:600 ] . the stars ( @xmath77 ) represent the velocity extracted from centering the peak of the model on the peak of the data in fig . [ fig:600 ] . the line is a fit of the data with @xmath78 @xmath79 . the fitted result is @xmath80 , where @xmath81 is the mass of the calcium ion . [ fig : vel],width=314 ] . for these data , the signal in the first @xmath59s is small compared to the model . moreover , there are variations in the fluorescence signal that do not appear in the model . this suggests that at the edges of the plasma expansion , the density distribution is distinctly non - gaussian , even at early times before any ion motion is possible . it also appears , as pointed out in ref . @xcite , that the gaussian density profile is recovered at late times for all initial electron energies . we present an extension of the isothermal plasma expansion model of refs . @xcite for quasi - two - dimensional geometry . we include velocity effects and predict a fluorescence or absorption signal vs. time for given initial conditions . the model matches the correct order of magnitude and general features of the fluorescence signal . some discrepancies are pointed out , which can be studied in future work . making the plasmas more ideally two - dimensional should improve applicability of the model , and further suppress effects due to expansion in the long dimension . increasing the range of initial electron temperatures and plasma densities can probe interesting regions of phase space where differences between the data and the model are likely to be more pronounced . as an example , it should be possible to extend this model to include effects due to electron - ion recombination at early times . for strongly - coupled neutral plasmas , the three - body recombination rate should be on the order of the plasma frequency . using the high - sensitivity and fast time - response of fluorescence spectroscopy , it should be possible to directly measure spectroscopic recombination signatures in low - temperature , low - density plamsas , where the predicted three - body recombination rate is greater than the plasma frequency . for example , reducing the plasma density to 10@xmath82 @xmath56 will make the recombination time @xmath83 ns . increasing the probe laser intensity to 100 times the saturation intensity will shorten the early rise time of the fluorescence signal to around 10 ns . optical pumping time will be comparable to the correlation - induced heating time , around 200 ns . this research is supported in part by brigham young university , the research corporation , and the national science foundation ( grant no . phy-9985027 ) . one of us ( sdb ) also acknowledges the support of the alexander von humboldt foundation .

we extend an isothermal thermal model of ultracold neutral plasma expansion to systems without spherical symmetry , and use this model to interpret new fluorescence measurements on these plasmas . by assuming a self - similar expansion , it is possible to solve the fluid equations analytically and to include velocity effects to predict the fluorescence signals . in spite of the simplicity of this approach , the model reproduces the major features of the experimental data . ultracold plasmas are produced from photo - ionized laser - cooled gases @xcite . in these laboratory plasmas , it is possible to study the kinetics and thermodynamics of multi - component , strongly - interacting coulomb systems . these systems are characterized by the ratio of the nearest - neighbor coulomb energy to the average kinetic energy , denoted as @xmath0 , with @xmath1 being the interparticle spacing . a new class of ultracold plasma experiments has recently become available in which it is possible to spectroscopically study the plasma ions @xcite . these plasmas are made using alkaline - earth atoms , because the resonance transition wavelengths of the ions are readily generated using standard laser methods . the spatially - resolved time evolution of the plasma ion temperature and density can be measured using absorption and fluorescence techniques . these ultracold neutral plasmas are not trapped , although efforts are underway in a few laboratories to trap them . the untrapped plasmas freely expand , and as they expand the density and temperature change radically . processes of recombination , collisional and thermal ionization , radiative cascade , and adiabatic and evaporative cooling all play important roles in how the system evolves and equilibrates . a variety of models have been used to investigate the properties of these plasmas @xcite . one particularly simple isothermal fluid model @xcite has been surprisingly successful in predicting the general features of these plasmas @xcite . in this paper we extend this model from the spherically - symmetric gaussian plasma distributions to gaussian distributions with elliptical symmetry . the elliptical symmetry has important experimental advantages . in such systems the plasma expands primarily in two dimensions . the practical advantage is that the density falls more slowly than in the three dimensional case , making it possible to study the plasma for longer times . the doppler - shift due to the directed expansion of the plasma is also suppressed . it should therefore be possible to study plasma oscillations and heating effects for greater time periods before these oscillations are masked by the directed expansion of the plasma . finally , if the plasmas are generated from a density - limited neutral atom trap , the elongated symmetry allows a greater number of atoms to be trapped initially , corresponding to a greater column density of plasma ions . this directly increases the visibility of fluorescence and absorption signals .

a prototype problem in the field of nuclear condensed matter physics is to determine and understand the magnetic hyperfine field ( hff ) at any of the elements of the periodic table , incorporated as a substitutional impurity in a simple ferromagnetic host such as bcc fe . understanding hyperfine fields forms a critical test for our understanding of condensed matter . moreover , they provide a convenient tool for nuclear physicists to determine nuclear magnetic moments : two features that explain the decades of experimental@xcite and theoretical@xcite efforts that have been devoted to this problem . today , the hyperfine fields of all elements as substitutional impurities in bcc fe are well - understood up to about z=55 @xcite . for the heavier 5d impurities , sizeable deviations between theory and experiment remain@xcite . the hyperfine fields of very light impurities at interstitial sites in fe have been calculated as well@xcite . lanthanide impurities in fe are much less understood , both experimentally and theoretically . as far as experiment is concerned , it is hard to obtain _ reliable _ values for the lanthanide hyperfine fields ( and also for their electric - field gradients , see below ) . this problem is illustrated by the rather desperate conclusion of l. niesen in a still useful review@xcite back in 1976 : _ ( ... ) if we can not perform experiments that yield unambiguous results , we should better do no experiments at all . _ " on the theoretical side , no _ ab initio _ studies have been performed yet for lanthanide impurities in fe ( an approach using a model hamiltonian is developed in ref . ) . the reason for this lies in a known failure of the widely used local density approximation ( lda ) within density functional theory ( dft ) : lda is not suitable to describe strong electron correlations@xcite . as a result , the strongly correlated and mainly _ localized _ 4f states in lanthanides are rendered _ itinerant _ by lda . it can therefore be anticipated that for a lanthanide impurity in bcc fe , lda is incapable to describe correctly the interaction between these localized and strongly correlated 4f electrons and the itinerant 3d states of the host material . this could be overcome by treating the 4f states as core states , an approach that has been used in the past with some success for lanthanides@xcite . in this way , however , one forces the f - electrons to behave exactly as in free atoms , which is not entirely correct . an efficient and popular way to improve on the lda failure without resorting to fully atomic 4f behavior , is to use the lda+u method @xcite . in lda+u , the correlation absent in lda is reintroduced by an on - site coulomb repulsion parameter u , to which an _ a priori _ value has to be assigned . the lda+u method has been used in the recent past with considerable success ( recent examples are refs . and many others ) , but it is not yet clear where the boundaries of its range of applicability are . in this work , we will examine how well lda+u performs on a delicate quantity as the hff . an extra feature for lanthanide impurities in fe that is absent for lighter impurities , is the presence of a large electric - field gradient ( efg ) at the lanthanide nucleus . at a site with cubic point symmetry such as a substitutional site in bcc fe the efg tensor must be necessarily zero . due to the strong spin - orbit coupling for these heavy impurities , however , the crystalline cubic point symmetry at the lanthanide site is lowered to a tetragonal one . this allows the existence of a large efg at the nucleus of the impurity . the same happens for 5d impurities in fe@xcite , but there the efg is 2 orders of magnitude smaller . we have calculated and analyzed this efg for lanthanides , and compare it with the sparse experimental data . the goals of this work can be summarized as follows . on the physical side , we want to obtain better _ quantitative _ and _ qualitative _ insight in magnetic hyperfine fields and electric - field gradients of lanthanide impurities in fe . this should allow us to asses better the reliability of the existing experiments , and to derive the underlying physical mechanism . on the technical side , we want to examine whether the range of applicability of the lda+u method can be extended to problems as delicate and sensitive as magnetic and electric hyperfine interactions of heavy impurities in a transition metal host . it will be shown that in the course of this analysis unexpected new results and questions show up , such as the influence of the lanthanide 5p electrons ( sec . [ sec - ldau ] and [ sec - efgldau ] ) and the position of the delocalization - localization transition in this system ( sec . [ sec - ldau ] and [ sec - pressure ] ) . all our calculations were performed within density functional theory@xcite , using the augmented plane waves + local orbitals ( apw+lo ) method@xcite as implemented in the wien2k package@xcite to solve the scalar - relativistic kohn - sham equations . in the apw+lo method , the wave functions are expanded in spherical harmonics inside nonoverlapping atomic spheres of radius @xmath1 , and in plane waves in the remaining space of the unit cell ( = the interstitial region ) . for the fe atoms a @xmath1 value of 2.20 a.u . was chosen , while for the lanthanide impurity we used @xmath22.45 a.u . the maximum @xmath3 for the expansion of the wave function in spherical harmonics inside the spheres was taken to be @xmath4 . the plane wave expansion of the wave function in the interstitial region was made up to @xmath5 a.u.@xmath6 , and the charge density was fourier expanded up to @xmath7 . the lattice constant of fe was fixed at the experimental value of 2.87 . in order to reproduce the situation of an isolated impurity in bulk fe , we used the supercell approach with a @xmath8 supercell where one iron atom was replaced by a lanthanide atom . the neighboring fe atoms will be displaced by the presence of this impurity , as was documented before for lighter impurities in fe@xcite . we took this effect into account in an average way by relaxing the nearest neighbors for eu as an impurity ( which is in the middle of the lanthanide series ) , and kept the same relaxation fixed for all other lanthanides . the eu - fe distance was 2.60 which is an increase of 0.11 with respect to the fe - fe distance and which is almost identical to the distance between 5p impurities and their fe neighbors@xcite . it was tested for another lanthanide ( er ) that there was only a marginal difference of less than 1 t between the hff obtained with the eu - fe distance and the correct er - fe distance ( 2.58 ) . a test for an extended supercell of 32 atoms was also performed . we relaxed the first four nearest neighbors . the eu - fe distance hardly changed ( 2.63 ) and the fermi contribution to the hyperfine field changed with 5 t. for the sampling of the brillouin zone ( bz ) of the @xmath8 supercell we took 75 special k - points in the irreducible part of the bz , which corresponds to a @xmath9 mesh . as exchange - correlation functional , the local density approximation ( lda)@xcite was used . spin - orbit ( so ) coupling was taken into account in all the calculations by a second - variational step scheme@xcite , using a cut - off energy @xmath10 ry . relativistic local orbitals ( rlo ) for the lanthanide 5p states were added to the basis set , because it is known that for actinides this allows to reduce the basis set size needed for the second variational step@xcite ( = lower @xmath11 ) . limitations in the implementation prevent to obtain correct efg s and dipolar hff s when rlo s are used . therefore , whenever such information was needed , the calculations were repeated without rlo s . this never had a large influence on the obtained values , however . for the lda+u method , the ` around the mean field ' ( amf ) scheme of czyyk and sawatzky@xcite was used . the choice of the u and j parameters is discussed in detail in sec . [ sec - ldau ] . free atoms were simulated by a supercell containing only one lanthanide and vacuum otherwise , leading to a separation of 9.4 between two ` neighboring ' lanthanides . for free ion calculations , this cell was charged . all the other parameters were chosen exactly the same as in the calculations for lanthanides in fe . let us first have a look at the experimental data set for the hff ( fig [ fig1]-a ) . dotted - dashed line : lda results for the antiferromagnetic orientation ( campbell - brooks orientation ) . dotted line : lda results for the ferromagnetic orientation when this one has the lowest energy . full line : lda+u value for trivalent lanthanides . dashed line : lda+u value for divalent lanthanides . [ fig1],width=302 ] only in 4 cases the magnitude of the hff _ and _ its sign are known with high reliability ( the sign of the hff indicates whether the field is parallel ( + ) or antiparallel ( - ) with respect to the magnetization of the fe host lattice ) . these cases are la ( -47(1 ) t)@xcite , ce ( -41(2 ) t)@xcite and lu ( -73.12(36 ) t)@xcite for which nuclear magnetic resonance on oriented nuclei ( nmr / on ) has been performed , and yb ( -125(8 ) t)@xcite on which time - dependent perturbed angular correlation spectroscopy ( tdpac ) has been applied . the latter technique has also been used for gd@xcite , albeit on a recoil - implanted sample which is not necessarily clean . the value of -26(8 ) t obtained in this way agrees well with an in principle reliable mssbauer measurement of -37 t , which is unfortunately not very well documented@xcite . three time - integrated perturbed angular correlation ( ipac ) measurements are available for gd as well ipac is a method that is rather unreliable , and can merely be used to determine the sign and an order of magnitude . they yield -20(5 ) t@xcite , -18(9 ) t@xcite and -7 t@xcite . a hff of -30(10 ) t can therefore be assigned to gd in fe in a reliable way . in 4 other cases the magnitude of the hff but not its sign has been measured with an accurate method as mssbauer spectroscopy ( ms ) : eu ( 148.2(9 ) t)@xcite , dy ( 610(7 ) t)@xcite , er ( 768(13 ) t)@xcite and tm ( 671 t)@xcite . for pr@xcite and tb@xcite , the magnitude but not the sign has been measured with low temperature nuclear orientation ( ltno ) . this non - resonant technique provides data that are less accurate than the previous ones , although they still are reasonably reliable . finally , in the case of ce@xcite , nd@xcite , sm@xcite , eu@xcite , gd@xcite , dy@xcite , er@xcite and tm@xcite ipac experiments have been reported , from which only the sign information can be reasonably trusted ( see e.g. the agreement with other experiments in fig . [ fig1]-a and ref . ) . due to the the latter sign information , the hff of the light lanthanides is guessed to be positive , while for the heavy lanthanides it is negative . the line in fig . [ fig1]-a summarizes the most likely interpretation of this data set . fig.[fig1]-a can be understood in terms of hund s rules and the model of campbell and brooks . based on heuristic arguments ( ref . ) and first principles calculations ( ref . ) , campbell and brooks showed that the _ interatomic _ exchange interaction between a transition metal 3d spin moment and a lanthanide 5d spin moment is antiferromagnetic ( fig . [ fig2 ] ) . -a ) . due to hund s third rule , the lanthanide 4f orbital moment is antiparallel ( parallel ) to the 4f spin moment in the first ( second ) half of the series . because the orbital contribution to the hff ( which is parallel to the orbital moment ) is dominant for lanthanides , the total hff is parallel ( antiparallel ) to the fe moment and hence called positive ( negative ) in the first ( second ) half of the series . [ fig2],width=302 ] the lanthanide 4f moment is localized at the lanthanide site and can not directly interact with its transition metal neighbors , but it has a ferromagnetic _ intra - atomic _ exchange interaction with the lanthanide 5d moment . the result is a net antiferromagnetic coupling between the lanthanide 4f moment and the transition metal 3d moment ( fig . [ fig2 ] ) . according to hund s third rule , the lanthanide orbital moment is antiparallel to the lanthanide spin moment for the 7 lightest lanthanides , and parallel to it for the 7 heaviest lanthanides . the dominant contribution to the hff is the orbital hff ( see sec . [ sec - lda ] and fig . [ fig4 ] ) , , width=302 ] which is parallel to the orbital moment . therefore , one expects the total hff to be parallel to the fe magnetization ( and hence positive ) for the light lanthanides , and antiparallel ( negative ) for the heavy ones , as is seen indeed in fig . [ fig1]-a . as a first step , we calculate the magnetic hff with the common lda . this will provide us with a data set to which we can later compare the possible improvement by lda+u , and it allows to introduce some peculiarities that will play a role in all later calculations as well . as is usual with this type of methods , our calculations involve an iterative procedure ( ` self - consistent field ' procedure ) that yields in the end a possible state of the calculated system , which is not necessarily the desired ground state : in the space of possible solutions , this self - consistent field procedure finds a _ local _ minimum , but not necessarily the _ global _ minimum . the local minimum that is obtained , depends to some degree on the starting configuration that was initially chosen . this behavior is prominently present for lanthanides in fe . if the spin moment of the lanthanide initially is put parallel to the fe spin moment , then this parallelism is maintained throughout the iterative procedure ( except for la , ce and pr , where the moment always spontaneously turns to an antiparallel orientation ) . we call this from now on the ferromagnetic solution . with an initially antiparallel configuration , an antiparallel ( or antiferromagnetic ) solution is obtained . if the lanthanide was given initially no spin moment , then a solution with a spin moment that is much reduced compared to the two preceding solutions was found . in order to decide which of those is the ground state , one has to look at the total energy of each solution . the total energy of the case with reduced moment was much higher than the others , and we will not consider it further . the energy differences between the other two solutions are given in fig . [ fig3 ] . - e@xmath12 ) . if this energy difference is positive , the antiferromagnetic situation has the lowest energy . energy difference for lda are compared with energy differences for lda+u with progressively larger u. the lda result is equivalent to u=0.0 ry . there are some gaps in the picture , because not for every situation a converged solution could be found . for pm the calculated energy difference is much larger ( 0.42 ry for u=0.6 ry ) and is out of scale . [ fig3],width=302 ] for all lanthanides up to tb , the antiferromagnetic solution has the lower energy . starting with dy , the ferromagnetic solution becomes the ground state . in fig . [ fig4 ] , the different contributions to the magnetic hff are given for both types of solutions . a hff is a magnetic field at the position of the nucleus , and it is built mainly from 3 contributions : the spin dipolar field , the fermi contact field and the orbital field . the spin dipolar field is generated by the spin moments of the electrons surrounding the nucleus . for cubic point symmetry , this contribution vanishes . the fermi contact field@xcite is of dipolar nature as well , but is due to the penetration of s - electrons into the nucleus . it does not vanish for cubic symmetry , and it is the dominant ( and almost only ) contribution for impurities up to z=55 in fe . the orbital field stems from the electric charge of the electron that orbits the nucleus , and it vanishes for cubic symmetry . as fig . [ fig4 ] shows , the dipolar field does not exceed a few tesla , while the fermi contact field lies between -20 t and -40 t. the orbital field is the dominant contribution , and can reach almost @xmath13 t. at first sight , one would expect a zero orbital and dipolar field for a substitutional impurity in fe , as the point symmetry is cubic . the reason why this is not the case for lanthanides , is purely due to the presence of spin - orbit coupling , which breaks the crystalline cubic point group symmetry . the oscillatory behavior of the orbital moment reflects hund s third rule : the orbital moment is antiparallel ( parallel ) to the spin moment in the first ( second ) half of the lanthanide series . because the orbital field is parallel to the orbital moment , it will for the antiferromagnetic solution be positive in the first half of the series and negative in the second half ( and vice versa for the ferromagnetic solution ) . according to the lda total energies , we have to accept the antiferromagnetic solution as the ground state up to tb , and the ferromagnetic solution starting from dy . this leads to positive hff s for almost all lanthanides ( fig . [ fig1]-b ) , which is contradiction with the current interpretation of the experimental data set and with the model of campbell and brooks . we will demonstrate in sec . [ sec - ldau ] that this is a new , clear example of a failure of lda , to be added to the list of notorious shortcomings as the general overbinding behavior , the prediction of the wrong crystal structure for fe , and the prediction of metallicity for some strongly correlated insulators as nio and la@xmath14cuo@xmath15 . even if we would select the antiferromagnetic solution throughout ( as the campbell - brooks model suggests ) , then still the quantitative agreement with the experimental hff s is rather poor ( fig . [ fig1]-b ) . using and interpreting lda+u calculations brings some complications that are absent for lda . first , lda+u schemes are not fully _ ab initio _ : they involve an on - site coulomb repulsion parameter @xmath16 and an on - site exchange interaction constant @xmath17 that have to be chosen _ a priori _ for every atom with strong correlations . in our case we have to choose one @xmath16 and one @xmath17 for the f - states of the lanthanide impurity . in line with the strategy adopted for the relaxation ( sec . [ sec - compdetails ] ) , we strive for reasonable overall agreement and do not focus on agreement for individual cases too much . therefore we take the same @xmath16 and @xmath17 for all lanthanides . looking at other calculations@xcite and experiments@xcite , @xmath18ry is a reasonable choice . the value of @xmath17 is usually an order of magnitude smaller , and it does not affect the results as much as @xmath16 does . therefore we take @xmath17 as 10% of @xmath16 . second complication with lda+u is that due to the introduction of tuneable variables @xmath16 and @xmath17 , the total energy looses its exact mathematical meaning as a variational quantity . if lda+u is used , a lower energy for one solution does therefor not necessarily mean that this solution is to be preferred over another one with a higher energy . this does not mean that these total energies are meaningless : for instance , lda+u has been used successfully to determine the equilibrium volume of materials by energy minimization@xcite . the current general opinion is that if the electronic structure does not _ qualitatively _ change between two slightly different situations , then the total energies obtained by lda+u for these situations can probably be compared . the criterion to decide whether there is a qualitative difference between two electronic structures or not , is provided by the 4f density matrix : in qualitatively similar cases , the occupation of the 14 m - orbitals (= the diagonal elements of the two spin - polarized @xmath19 density matrices ) should be more or less identical . finally and third there are much more local minima in the space of solutions when lda+u is used , and a calculation gets easily trapped in one of them . because the density matrices for two different solutions are necessarily different from each other , the total energy can not be used to determine which solution is the ground state . this problem is illustrated in tab . [ tab1 ] , [ cols="<,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] tab . [ tab2 ] shows that the main contribution to the total @xmath20 is due to the single 4f - up electron : @xmath21 . there is a large contribution of 5.6 + 9.6=15.2 with the opposite sign due to p - electrons . what is surprising is that this p - contribution does not stem from the valence 6p electrons , but from the entirely filled and strongly bound 5p shell , which lies more than 20 ev below the fermi energy . intuitively , one would have assumed such a filled and well - bound shell to be entirely spherically symmetric , which would mean @xmath22 . and indeed , the 5p - anisotropy @xmath23 is very small : 0.0030 ( up ) and 0.0040 ( down ) ( it is shown in ref . that @xmath24 is proportional to @xmath25 ; @xmath26 is the number of electrons in the @xmath27 orbital , and @xmath24 measures the unequal occupation of the 3 p - orbitals ) . however , a considerable part of this anisotropy stems from a region very close to the nucleus and hence gets amplified by the @xmath0 factor . this is demonstrated in fig . [ fig11 ] , for tb in fe ( sum of up and down electrons , arbitrary units for y - scale ) . b ) left axis ( arbitrary units ) : @xmath28 for tb in fe . right axis ( arbitrary units ) : integral of @xmath28 , which is called @xmath29 ( apart from a constant factor with negative sign ) . c ) and d ) : idem , but for the f - f contribution . [ fig11],width=264 ] where for the same tb - configuration the bare anisotropic p - p charge density @xmath30 is shown , before ( fig . [ fig11]-a ) and after ( fig . [ fig11]-b ) weighing with a 1/r@xmath31 factor , and also after integration ( fig . [ fig11]-b , right axis ) . figs . [ fig11]-c and [ fig11]-d repeat this for the f - f contribution . the final integrals are clearly determined exclusively by anisotropies in a region closer than 0.05 to the nucleus . in the expression for a spin dipolar field , the same factor @xmath0 is present , explaining why a similar dependence of the 5p dipolar field on the 4f occupation was observed there ( fig . [ figc]-c , d , e ) . a generalization of this analysis is given in fig . [ figd ] , which shows the same individual 4f m - orbital contributions to @xmath32 for lanthanides split into 4f and 5p ( i.e. when a given 4f m - orbital is occupied , fig . [ figd]-a , b shows the direct contribution from this orbital , while fig . [ figd]-e shows the corresponding induced contribution of the 5p - shell ( up and down summed ) ) . this shows that the opposite signs for 5p and 4f as seen in the example of tb is a general effect : occupying the 4f m=0 orbital gives a negative direct 4f contribution but induces a positive 5p contribution , etc . this can be understood as follows . a negative @xmath32 corresponds to charge accumulation along the z - axis , a positive @xmath32 to charge accumulation in the xy - plane . the shape of the 4f orbitals is such that m=0 has its charge mainly along the z - axis ( reflected in a negative @xmath33 ) , while the xy - plane is more and more occupied for larger @xmath34 . apparently the 4f electrons dispel the 5p electrons : if m=0 is occupied , then the 5p electrons are forced away from the z - axis into the xy - plane , resulting in a positive @xmath25 and @xmath24 . as tab . [ tab2 ] shows , a 4f - up electron distorts the 5p orbitals with either spin : this is an interaction between the electron charges , not between the spins . recently , the efg of the actinide u has been analyzed in uo@xmath14 by r. laskowski et al . ( ref . ) , using the same apw+lo method as used in this work . these authors show in their fig . 2 the contributions of @xmath29 , @xmath35 and @xmath36 as a function of the deformation of the oxygen cage that surrounds the u atom . they do not further divide the p - p contribution in 6p and 7p ( here the 6p shell is entirely filled and well - bound ) . without deformation of the oxygen cage , the crystallographic surrounding is cubic and the slightly non - zero @xmath37 is due to spin - orbit coupling only , as is the case for lanthanides in fe ( compare with nd in fig . [ fig8 ] ) . this small @xmath32 is a sum of a f - f contribution of + 20 and a p - p contribution of -21 ( the d - d contribution is small : -1 ) . this can be compared to fig . [ figd ] for nd with the m=(-3 , -2 , -1 ) orbitals filled , which leads to a 4f - contribution + 14 and a 5p - contribution of -12 , quite similar values . because of this analogy , we suggest that also for u in uo@xmath14 th p - p contribution for an undistorted oxygen cage is due to the completely filled 6p shell . this interpretation would furthermore imply that as a function of oxygen cage deformation , the 6p - contribution in fig . 2 of ref . would remain almost constant ( just as the 5f - contribution does ) , and that the strong decrease of the total p - p contribution is due to 7p only . this makes sense , as this decrease is attributed @xcite to the tails of the o-2p wave functions , and hence should appear near the fermi energy (= the region of the 7p ) . it is well known experimentally@xcite that the pure lanthanides exhibit a large variety of structural phase transitions as a function of external pressure . at certain pressures , volume collapses are sometimes observed and attributed to the delocalization of the f electrons . these delocalization pressures have been determined for 6 elements of the lanthanide series : ce @xcite , pr @xcite , nd @xcite , sm @xcite , gd @xcite and dy @xcite ( fig . [ pressure ] ) . for ce the delocalization of the f electrons occurs around the pressure of 1 gpa and is accompanied by a volume collapse of 16% at the isostructural transition to another fcc phase ( @xmath38-ce ) . pr transforms to a @xmath38-u structure at 20 gpa with a volume collapse of 9 - 12% . in gd 4f delocalization occurs at 59 gpa when the structure changes to body - centered monoclinic ( bcm ) with a volume collapse of 10.6% and in dy this happens at 73 gpa with a volume collapse of 6% . for nd and sm no volume collapse has been observed . in these two cases the delocalization of the 4f electrons was associated with the appearance of low - symmetry structures ( similar with those that appear in pr , gd and dy cases with volume collapse ) . this is a somewhat ambiguous procedure . for nd two transition pressures have been proposed : 40 gpa ( corresponding to the transition to an hp3 structure @xcite ) and 113 gpa ( corresponding to the transition to the @xmath38-u phase @xcite ) . for sm the delocalization pressure is proposed to be 37 gpa , when the sm structure changes to hp3 . for pm we have only a lower limit for the transition pressure , 60 gpa ( until this pressure no low - symmetry structure has been observed@xcite ) . , width=264 ] in our calculations for lanthanide impurities we replace an fe atom from an iron lattice by a lanthanide atom . obviously the lanthanide atom ( which has a much larger volume ) will feel a chemical pressure or effective pressure . how large will this effective pressure be ? we concluded in sec . [ sec - ldau ] that at least ce and pr are delocalized . hence , the effective pressure which we assume to be independent on the lanthanide in a first approximation should be at least 20 gpa ( fig . [ pressure ] ) . the hyperfine fields for nd and sm are only very approximately measured ( fig . [ fig1]-a ) , such that one can not conclude whether they are localized or not . based on isomer shift and efg ( fig . [ fig8 ] ) , eu is definitely localized , as are all heavier lanthanides . therefore , two qualitatively different proposals for the effective pressure are possible : about 25 gpa ( everything starting with nd is localized ) or about 40 gpa ( everything below eu is delocalized , except for pm and maybe nd ) . assuming an effective pressure that is not constant ( motivated by the decreasing volume of heavier lanthanides ) does not change this picture ( dotted lines in fig . [ pressure ] these lines qualitatively take the lanthanide contraction into account ) . a more accurate experimental determination of hff and efg for nd , pm and sm in fe would allow to distinguish between both scenarios and would allow to determine the real position of the delocalization - localization transition in this system . experimental data for nd in fe are also for another reason interesting : if nd in fe would be found to be localized ( itinerant ) and the effective pressure of 40 gpa would known to be correct ( from a sm - measurement for instance ) than the delocalization pressure of 113 gpa ( 35 gpa ) for bulk nd is probably correct . if the effective pressure of 25 gpa would be correct , no such conclusion can be made . in conclusion for this section , accurate measurement of hff and efg for nd , pm or sm in fe would offer a lot of information . lda+u calculations for free lanthanide 3@xmath39 ions were already mentioned in figs . [ figa]- [ figd ] . a quite complete experimental data set exists for this situation as well , both for the hff ( fig . [ fig10]-a ) and the efg ( fig . [ fig10]-b ) ( data are copied from ref . , the original data are in ref . ) . , [ figc ] and [ fig6 ] , both for divalent and trivalent lanthanides . b ) experimental value for @xmath32 in free lanthanide ions , compared with lda+u predictions based on fig . [ figd ] , both for divalent and trivalent lanthanides . [ fig10],width=302 ] in the experiments , the ions were not really free but were incorporated in a paramagnetic salt , and the effect of crystal fields was removed later in order to find the free ion values . by the same procedure as explained for lanthanides in fe , we extracted lda+u predictions for the hff and @xmath32 for divalent and trivalent lanthanide ions , which are given by dashed and full lines in fig . [ fig10 ] , respectively . for lanthanides in fe , the positive z - direction was naturally defined by the moments of the ferromagnetic fe host atoms . for free lanthanides , the total ( = spin+orbital ) angular momentum @xmath17 determines the positive z - direction . due to this different choice of axes , there is an _ apparent _ sign change for the heavy lanthanides between fig . [ fig1 ] ( in fe ) and fig . [ fig10]-a ( free ) . the agreement with experiment is again quite nice . eu is divalent ( it was trivalent in fe ) , while yb is trivalent ( it was divalent in fe ) . for sm there is a large deviation , both for the hff and @xmath32 . but this is no surprise : it is well - known there are low - lying excited states in sm which will mix with the ground state , such that our procedure which is based on hund s rules ground states is expected to fail . the possibility to consider non - collinear magnetism at every infinitesimal region of space has recently been implemented@xcite in the wien2k code . in principal this can be an important feature even for collinear antiferromagnets as we are dealing with here : it allows the spin moment to turn _ gradually _ from the fe - orientation to the opposite lanthanide - orientation , and this is a better replication of what happens also in nature . we did not attempt a full study , but calculated hff and efg for tm in fe only . all technical parameters were chosen exactly the same as in the collinear calculations . for tm in fe as test example , the total hff changes from -822 t in the collinear lda+u calculation to -838 t in a non - collinear one , while the efg remains exactly the same : -38.1 @xmath40 v / m@xmath41 . such a change of 16 t is not small in absolute value , but is negligible compared to the large values of the hff s in this problem . therefore we conclude that non - collinear magnetism does nt play an important role for lanthanides in fe . we have demonstrated that lda leads to a qualitatively wrong behavior for the magnetism of lanthanides in fe . using lda+u , qualitative and quite reasonable quantitative agreement with experiment is obtained , both for hff and efg . this shows that the semi _ ab initio _ lda+u method is a useful tool even for such sensitive quantities as hyperfine parameters of strongly correlated impurities in an itinerant magnetic host . we could come to these conclusions only after applying a careful strategy in order to cope with the lack of a good criterion to determine the true ground state if lda+u is used . for all lanthanides the 4f spin moment couples antiferromagnetically to the fe 3d moment , in agreement with the model of campbell and brooks . the orbital hff is by far the dominant contribution to the total hff ( figs . [ fig1 ] and [ fig4 ] ) . the calculated value of the efg for lanthanides in fe agrees well with the few experimental data ( fig . [ fig8 ] ) . we discovered an unexpectedly strong contribution of the completely filled 5p shell to the dipolar hff and the efg , which can be explained by their common @xmath0 dependence : small deformations of the 5p shell in a region close to the nucleus are strongly emphasized . a reinterpretation of recent efg calculations@xcite for uranium in uo@xmath14 suggests that the same is true for the 6p shell in actinides . furthermore , we conclude that yb is divalent in an fe host , while all other lanthanides are trivalent ( including eu , with perhaps an exception for tm ) . the lightest lanthanides ( at least up to pr ) show delocalized 4f behavior , and we conclude that the delocalization - localization transition that typically happens already at ce is postponed for lanthanides in fe : it falls at least after pr and current experiments do not exclude that it could go up to sm ( although pm is certainly localized ) . this can be explained by the large effective pressure that is felt by these lanthanide impurities ( either 25 or 40 gpa - fig . [ pressure ] ) , leading to a larger overlap between the 4f wave functions and the neighboring fe-3d . the question of a postponed localization transition has never been touched before in the 40 years of experiments on this system . this illustrates what can be the added value of ab - initio calculations for hyperfine interactions studies . also in the case of free lanthanide ions , hff and efg can be quantitatively reproduced . remarkably , eu is divalent in this case and yb is trivalent just the opposite as for lanthanides in fe . the effect of fully non - collinear magnetism on this problem was tested to be negligible . obviously , the numerical agreement of the calculated hyperfine fields with experiment is for these lanthanides still much less satisfactorily then it is for lighter impurities . with the current methods , there are only limited possibilities to improve the accuracy of the calculations . the supercell can be extended to e.g. 32 atoms and relaxation of the fe - neighbours can be calculated for every individual element ( this requires the calculation of forces including spin - orbit coupling and lda+u , which is time - consuming and not yet fully implemented in wien2k ) . but as inevitably a rather arbitrary choice remains to be made for the value of @xmath16 , it is not clear whether these sophistications will really improve the agreement with experiment . and most likely they will not add anything new to the physical insight . in our opinion , new progress in this topic will have to come from experiment . many of the experimentally determined hff s and efg s carry still large error bars . accurate measurements for instance with the nmr / on method are desirable ( note that nmr / on has not yet been applied for any of the lanthanides with a large efg and/or hff : such large hyperfine interactions put severe requirements on the equipment ) . the predicted hff s and efg s from this work should allow to reduce considerably the frequency domain that has to be scanned in an nmr / on experiment , and warrants a more physical and reliable interpretation of the observed resonances . as most worthwhile experiments , we suggest a more accurate determination of hff and efg for pr to sm : this would allow to examine experimentally the position of the delocalization - localization transition . once a data set with improved accuracy will be available , it can serve in its turn as a testing ground for future generations of _ ab initio _ many body methods . illuminating discussions with th . mazet ( nancy ) , p. blaha ( vienna ) , p. novk ( prague ) , m. divi ( prague ) , b. barbara ( grenoble ) , n. severijns ( leuven ) and p. schuurmans ( sck / cen - mol , belgium ) are gratefully acknowledged . the calculations were performed on a pc - cluster in leuven , in the frame of projects g.0239.03 of the _ fonds voor wetenschappelijk onderzoek - vlaanderen _ ( fwo ) , the concerted action of the kuleuven ( goa/2004/02 ) and the inter - university attraction pole ( iuap p5/1 ) . the authors are indebted to l. verwilst and j. knuts for their invaluable technical assistance concerning this pc - cluster .

the magnetic hyperfine field and electric - field gradient at isolated lanthanide impurities in an fe host lattice are calculated from first principles , allowing for the first time a qualitative and quantitative understanding of an experimental data set collected over the past 40 years . it is demonstrated that the common local density approximation leads to quantitatively and qualitatively wrong results , while the lda+u method performs much better . in order to avoid pitfalls inherent to the lda+u method , a careful strategy had to be used , which will be described in detail . the lanthanide 4f spin moment is found to couple antiferromagnetically to the magnetization of the fe lattice , in agreement with the model of campbell and brooks . there is strong evidence for a delocalization - localization transition that is shifted from ce to at least pr and maybe further up to sm . this shift is interpreted in terms of the effective pressure felt by lanthanides in fe . implications for resolving ambiguities in the determination of delocalization in pure lanthanide metals under pressure are discussed . for the localized lanthanides , yb is shown to be divalent in this host lattice , while all others are trivalent ( including eu , the case of tm is undecided ) . the completely filled and well - bound 5p shell of the lanthanides is shown to have a major and unexpected influence on the dipolar hyperfine field and on the electric - field gradient , a feature that can be explained by their @xmath0 dependence . an extrapolation to actinides suggests that the same is true for the actinide 6p shell . the case of free lanthanide atoms is discussed as well .

binary exponential backoff ( beb ) is widely adopted as a key collision resolution mechanism in popular random - access networks , such as ieee @xmath4 ethernet and ieee @xmath5 wireless local area network ( wlan ) . with exponential backoff ( eb ) , a packet is transmitted after waiting a number of time slots randomly selected from a contention window , the size of which increases multiplicatively on collisions . mathematically , the contention window @xmath6 after @xmath7 consecutive collisions of a packet . here , @xmath8 with @xmath2 is the backoff function for eb must be an increasing function for the backoff process to be meaningful . therefore , @xmath9 must be larger than unity . ] and @xmath10 is the initial contention window size . beb is a special case with @xmath11 . most of the research attention has been focused on investigating the throughput provided by eb . thanks to the seminal work of bianchi@xcite , the throughput is now well understood through a fixed point equation that characterizes the backoff process . subsequently , @xcite shows that the throughput of eb is stable against the network size in the sense that the throughput converges to a nonzero constant when the network size goes to infinity ( assuming no retry limit is enforced ) . throughput stability has been the most intriguing aspect of eb , and has enabled eb - based mac protocols to support a wide range of throughput oriented applications regardless of the network congestion level . with the recent boom of delay - sensitive multimedia applications such as voip and video conferencing , research interests are being shifted to other aspects of system performance such as delay , delay jitter , and short - term fairness @xcite . indeed , it can be shown that delay jitter significantly affects the users perception of quality of real - time multimedia services . eb , despite its good throughput performance , has been shown to suffer poor performance in delay and short - term fairness . more specifically , eb could induce divergent ( i.e. , infinite ) second- and high - order moments of medium access delay , yielding extraordinarily large delay jitter and severe transmission starvation of users @xcite . essentially , the medium access delay follows a power law distribution , implying that a non - negligible number of packets may experience much larger delay than the average @xcite . as a motivating example , we monitor the packet transmission during a @xmath12 second period in a @xmath13-node ieee @xmath5 g wlan , where beb is adopted . alarmingly , @xmath14 out of the @xmath13 nodes experience severe transmission starvation , as illustrated in fig . @xmath15 . the figure shows that node @xmath16 and @xmath17 perceive starvation for a duration of @xmath18 and @xmath19 seconds , respectively . even worse , node @xmath14 barely receives any service throughout the entire simulation time . consecutive seconds for @xmath14 of @xmath13 nodes in a 802.11 g system . assume that all nodes are continuously backlogged and no retry limit is enforced.,width=336 ] in an attempt to address the above issues , this paper seeks to understand the following important questions . 1 . what is the root cause of the power - law delay distribution of eb . is it an intrinsic issue of eb , or can be avoided by adjusting the backoff exponent @xmath9 . 2 . if the problem is intrinsic with eb , can we find an alternative backoff function that does not suffer the same problem . in general , what is the necessary and sufficient condition for a backoff function to have convergent delay moments , i.e. , not to experience power law delay . 3 . is it possible to achieve throughput stability and convergent delay moments at the same time by certain backoff functions . if not , are there any backoff functions that exhibit convergent delay moments and good throughput performance at the same time when the network size is within a finite and practical range . in the literature , @xmath20 has been partly addressed . @xcite first finds that the medium access delay distribution of eb is heavy - tailed when retry limit @xmath21 is infinite , regardless of the backoff exponent @xmath9 . @xcite later proves that the medium access delay indeed follows a power law distribution , the slope of which is obtained as a function of the backoff exponent and the collision probability . noticeably , the effect of power law delay can not be eliminated even if a finite retry limit k is enforced in practical systems . @xcite and @xcite observe that the medium access delay follows a truncated power law distribution , implying that small retry limit does not eliminate the power law characteristics induced by eb . this directly translates to high packet loss rate , if packets have to be discarded upon reaching a retry limit @xmath21 . indeed , our simulation results show that beb suffers @xmath22 packet loss rate in a @xmath23-node network with @xmath24 , leading to an equal percentage reduction of throughput as that with @xmath25 . the analysis in these prior work can be treated as a special case of the analysis for general backoff functions in this paper . as to @xmath26 , there are some initial attempts to replace eb with other more moderate backoff algorithms , such as linear backoff ( @xmath27 ) and polynomial backoff ( @xmath28)@xcite . observations made by @xcite showed that linear and polynomial backoffs with appropriate parameter settings can improve upon beb in terms of throughput and delay performance . @xcite observed that pb can achieve a similar saturation throughput as eb but with much smaller delay jitter . however , to the authors best knowledge , no analysis was provided to explain the root cause behind the phenomenon . to fully address the important questions @xmath29 , this paper attempts to uncover the fundamental laws that govern the throughput stability and tail distribution of medium access delay . our main contributions are detailed below . * we find that the heaviness of the tail distribution of medium access delay is closely related to how rapidly the contention window is augmented with each collision . specifically , eb always induces power law delay distribution regardless of the choice of backoff exponent @xmath9 . meanwhile , power law delay is mitigated as long as the backoff function is slower than exponential functions , i.e. , @xmath1 for all @xmath2 , where @xmath30 will be defined more rigorously later . this explains the observations made by @xcite and @xcite . furthermore , we find that delay distribution becomes light - tailed if the backoff function increases linearly or sub - linearly . * we prove that throughput stability is achieved only when the backoff function is at least as fast as an exponential function , i.e. , @xmath3 for some @xmath2 , where @xmath31 will be defined rigorously later . in other words , pb fails to sustain non - zero asymptotic throughput , although they yield convergent delay moments . this presents a fundamental tradeoff between throughput stability and the heaviness of tail delay distribution . * we find that super - linear polynomial backoff achieves high throughput across a wide range of practical network size , despite its throughput instability asymptotically . this , together with our findings in @xmath32 , suggests that super - linear polynomial backoff is a better alternative than eb in supporting broadband network applications that call for both high throughput and low delay and delay jitter . our study on the delay tail distribution of backoff process is not only for theoretical interest but also closely related to engineering applications . in the past few years , a number of modified exponential backoff schemes , including quality of service enhancing protocols , have been proposed to improve the delay performance of conventional beb @xcite . for instance , @xcite proposed a lmild backoff algorithm , in which the contention window doubles upon collisions whereas decreases linearly upon successful transmissions . besides , the enhanced distributed channel access ( edca ) scheme , which is adopted in the @xmath33 standard , gives priority to delay - sensitive applications by setting a shorter contention window and shorter arbitration inter - frame space @xcite . despite their respective contributions , they do not eliminate the fundamental feature of power law delay distribution induced by exponential backoff , and thus may still perceive relatively large delay jitter or high packet loss rate . instead , we propose to fundamentally solve the power law delay problem by replacing eb with pb . meanwhile , we show that high throughput can be achieved in a wide range of practical network size through parameter tuning of pb . in this sense , we can mitigate the power law delay distribution of eb without hurting the advantageous throughput performance . our simulation results show that pb with reasonable backoff parameter outperforms beb regardless of the existence of the retry limit . with current hardware processing power , the implementation of pb in random access networks incurs minor extra cost . therefore , we believe it is a promising algorithm with broad applications in future random access networks . the rest of the paper is organized as follows . we briefly review the backoff protocols and introduce some background information in section ii . in section iii , the main results of this paper is summarized . in section iv , we analyze the power law tail distribution of medium access delay for general backoff protocols . in section v , we derive the condition to sustain stable throughput . simulation results are presented in section vi , where we show that pb is a better alternative than eb in random access networks . finally , the paper is concluded in section vii . in this section , we first briefly review the operation of general backoff protocols . we then introduce the notion of medium access delay and some important metrics that will be used in later sections to evaluate the performance of different backoff schemes . we consider a fully connected wlan consisting of @xmath34 continuously backlogged nodes . illustrated in fig . @xmath35 , the transmission of @xmath14 nodes is coordinated by a backoff mechanism . at each packet transmission , a node sets a backoff counter value @xmath36 by randomly choosing an integer from a contention window @xmath37 $ ] , where @xmath38 is the size of contention window . at the initial transmission attempt of a packet , @xmath38 is set to its minimum value @xmath10 . the contention window size is incremented on each collision . after @xmath39 collision , we say the node is in its @xmath39 backoff stage and the contention window size @xmath6 , where @xmath0 is an increasing backoff function characterizing the backoff process . for example , @xmath8 , @xmath40 for eb , @xmath41 , @xmath40 and @xmath42 for sub - exponential backoff ( seb ) , @xmath43 , @xmath44 for polynomial backoff ( pb ) . we denote the backoff counter value at the @xmath39 backoff stage by @xmath45 . the backoff counter value decreases by one following each time slot , which could either be an idle slot or a transmission time slot . the packet is transmitted once the backoff counter reaches zero . when there is a finite retry limit @xmath21 , a packet is dropped if it has not been successfully transmitted after @xmath21 retransmissions . the backoff process is illustrated in fig . @xmath35 . for example , a tagged packet at node @xmath16 experiences three backoff stages before a successful transmission . in backoff stage @xmath46 , one backoff countdown slot is occupied by the collision between node @xmath17 and @xmath14 , while the other countdown slots are idle time slots . after the countdown process in stage @xmath46 , node @xmath16 collides with node @xmath17 s transmission and enters backoff stage @xmath16 . the process repeats until it successfully transmits after backoff stage @xmath17 . and backoff schemes . the initial backoff window @xmath47 and the retry limit @xmath25.,width=336 ] the exact backoff process is very complex and often intractable , since the success and collision processes of various nodes are coupled and strongly correlated @xcite . a common technique adopted by most of the prior work on saturation analysis is the mean field decoupling approximation , where the backoff process at one node is decoupled and treated as if it is independent from the backoff processes at the other nodes . specifically , it assumes that a node encounters a collision probability @xmath48 when it transmits , regardless of its own backoff stage . moreover , the average attempt rate of an arbitrary node in a generic time slot , denoted by @xmath49 , is assumed to be constant and does not vary with the backoff stage @xcite . the validity of mean field approximation for eb has been recently verified in both theorem and experiments @xcite . to validate the assumption for general backoff functions , we reproduce the experiments in @xcite and plot @xmath48 against backoff stage for three representative backoff functions in fig . @xmath50 . the figure shows that , @xmath48 is largely independent of the backoff stage for all the backoff functions in consideration . in particular , the variances of @xmath48 across backoff stage are less than @xmath51 for all three backoff functions . the simulation results match the analytical results based on mean field approximation , which will be introduced in ( [ 1 ] ) and ( [ 3 ] ) . therefore , we can safely adopt the assumptions in this paper to analyze the throughput and delay performance of general backoff schemes . under mean field approximation , the probabilities of a time slot being an idle time slot , successful transmission or a collision can be calculated as @xmath52 with ( [ 19 ] ) , normalized throughput @xmath53 , defined as the portion of time occupied by successful packet transmissions , is given by @xmath54 where @xmath55 , @xmath56 and @xmath57 denote the lengths of idle , success and collision time slots , respectively . an important metric of system performance is throughput stability . here , we say a backoff scheme is throughput - stable if it can yield non - zero asymptotic throughput when the network size approaches infinity ( i.e. , becomes extraordinarily large ) . it is shown in @xcite that the transmission probability @xmath49 of a saturation network in steady - state is the root of a fixed point system @xmath58},\ ] ] where @xmath48 is the probability of a node encounters a collision when it transmits , which is given by @xmath59 the above fixed point system always has one unique solution as long as the backoff function @xmath0 is non - decreasing for @xmath60 @xcite . as we will show in later sections , some important properties of system performance , such as power law delay behavior and asymptotic throughput , are closely related to the value of @xmath48 . it is worth noting that , @xmath61 always holds in a system under steady - state , despite that the limit of @xmath48 could be @xmath16 . otherwise , if @xmath62 , all nodes will continuously encounter collisions and enter the next backoff stage . in this case , the limiting distribution of backoff stages does not exist and the system can never be in steady - state . for simplicity , we assume that the retry limit @xmath21 is infinite hereafter , so as to better understand the factors that fundamentally affect the properties of a backoff function without considering the implementation details . in this case , ( [ 2 ] ) becomes @xmath63}.\ ] ] the results obtained from the infinite-@xmath21 model can be easily translated to the standard systems with finite @xmath21 . for instance , an infinite - variance medium access delay distribution in the infinite-@xmath21 model indicates high packet loss rate in the standard networks . in fact , we will show this implication in the simulations section where results for both cases are presented . unless otherwise stated , we use delay " and medium access delay " interchangeably throughout the paper . illustrated in fig . @xmath35 , the medium access delay of a packet , denoted by @xmath64 , is the time period from the instant it becomes the head - of - line ( hol ) packet to the instant at which the packet is successfully transmitted . medium access delay of a packet consists of three parts , namely a series of backoff countdowns , collisions involving the tagged node and successful transmissions of the tagged node@xcite . in particular , the backoff countdown slots seen by a tagged node include idle slots as well as the busy slots , successful or collided , due to other nodes . for example , the medium access delay of the tagged packet at node @xmath16 in fig . @xmath35 includes @xmath14 backoff countdown stages , @xmath17 collision slots involving node @xmath16 and a successful transmission slot . suppose that a packet is successfully transmitted after @xmath65 collisions . then , the medium access delay of this packet denoted by @xmath66 , is @xmath67 here , @xmath68 is the time consumed on backoff countdown at the @xmath39 backoff stage . it is the summation of a number of backoff countdown slots , given by @xmath69 where @xmath70 is the length of its @xmath71 countdown slot , which could either be @xmath57 , @xmath56 or an idle time slot @xmath55 . we denote the probability density function of medium access delay @xmath64 by @xmath72 and its tail distribution function @xmath73 . in many engineering applications , we often encounter heavy - tailed distributions whose tail decaying rate is slower than exponential@xcite . for instance , it is observed that both the size of data files stored in web servers and the process execution time in a computing environment follow heavy - tailed distribution@xcite . power law distribution belongs to a subclass of heavy - tailed distribution , whose tail distribution follows a power law decaying rate . conversely , a probability distribution is light - tailed if it is not heavy - tailed . the definitions of the different distributions with respect to the decay rate of tail distribution are formally defined as follows and illustrated in fig . @xmath74 . * definition 1 : * a probability distribution @xmath72 is power law distribution with slope parameter @xmath75 , if its tail distribution function @xmath76 satisfies @xmath77 where @xmath78 is a slow varying function ( i.e. , slower than any power function , such that @xmath79 for all @xmath80 . for instance , @xmath81 ) and the notation @xmath82 means @xmath83 . * remark 1 : * power law distribution can also be characterized by the moments of @xmath64 . for a power law distribution with slope parameter @xmath75 , @xmath84 $ ] is finite for all @xmath85 and is infinite for all @xmath86 @xcite . in fact , the tail decaying rate of a probability distribution is closely related to the convergence of moments . specifically , a finite @xmath84 $ ] indicates that the tail distribution decays faster than a polynomial function with power @xmath87 ( cf . @xcite , p. @xmath88 ) . if @xmath84 $ ] are finite for all @xmath89 , the tail distribution of @xmath72 decays faster than all power law functions and @xmath72 belongs to region @xmath14 or @xmath90 in fig . @xmath74 . in this case , we say that the power law distribution is mitigated . * definition 2 : * a probability distribution @xmath72 is heavy - tailed distribution if its moment generating function diverges , i.e. , @xmath91 using taylor expansion to ( [ 21 ] ) , it holds that @xmath92.\ ] ] * remark 2 : * the tail decay rate of a heavy - tailed distribution is slower than any exponential functions . from ( [ 26 ] ) , any divergent moment @xmath93 $ ] would indicate that @xmath72 is heavy - tailed distribution , but not the reverse . the rhs of ( [ 26 ] ) could still be infinite even if all moments are finite . for example , weibull distribution with shape parameter smaller than @xmath16 is a heavy - tailed distribution but not a power law distribution . the relationship between power law distribution and heavy - tailed distribution is shown in fig . weibull distribution with shape parameter smaller than @xmath16 belongs to the set of distributions in region @xmath14 in fig . @xmath74 . generally speaking , if a delay distribution is identified as a power law or heavy - tailed distribution , the probability of extremely large delay occurs is non - negligible . for instance , it is shown in @xcite and @xcite that the delay distribution of eb follows power law distribution , which is considered as the root cause of poor delay performance and user unfairness in current wlan systems . in this paper , we develop a unified framework to study the heaviness of delay tail distribution of any general backoff functions , making the study in @xcite and @xcite a special case of ours . before leaving this session , we introduce the following two important notations to describe the limiting behavior of functions . * definition 3 : * a function @xmath94 , if @xmath95 and @xmath96 such that @xmath97 , @xmath98 . * definition 4 : * a function @xmath99 , if @xmath100 , @xmath96 such that @xmath101 , @xmath98 . * remark 3 : * loosely speaking , @xmath102 is asymptotically no slower " than @xmath72 if @xmath103 and slower " than @xmath72 if @xmath104 . the definitions of @xmath31 and @xmath105 can be straightforwardly extended to discrete functions @xmath0 and @xmath106 with the replacements of @xmath107 by @xmath7 and @xmath108 by @xmath109 , where @xmath110 . in the strict sense , @xmath111 does not imply @xmath112 , and vice versa . however , under a mild condition that the limit of @xmath113 exists ( @xmath114 ) , the two notations @xmath115 and @xmath112 are complementary ( cf . @xcite , ch . 3 ) . specifically , @xmath112 if @xmath116 and @xmath115 otherwise . a special interest of this paper is to compare the growth rate of a general backoff function @xmath0 with an exponential function , i.e. @xmath117 . in this case , the limit @xmath118 exists for most of the practical backoff functions , such as eb , seb and pb . without causing confusions , we discuss in the following in the weaker sense that @xmath119 if and only if @xmath120 fails . we summarize in this section the key results of this paper . the proofs of the results are deferred to section iv and v. * * power law delay * : a random - access network with an increasing backoff function @xmath0 does not suffer a power law delay if and only if @xmath1 , @xmath121 . in other words , the system observes power law delay if and only if @xmath122 such that @xmath123 ( proved in section iv ) . * * heavy - tailed delay * : the distribution of medium access delay is heavy - tailed with eb , seb and superlinear pb , while light - tailed with linear - sublinear pb ( proved in appendix b ) . * * throughput stability * : an increasing backoff function @xmath0 is throughput - stable if and only if @xmath122 such that @xmath3 . in other words , the network is throughput - unstable if and only if @xmath1 , @xmath121 ( proved in section v ) . our results show that the power law behavior of eb is essentially attributed to its exponential function growth rate , and can be mitigated if eb is replaced by a slower " backoff function , such as pb and seb . however , it is also the exponential growth rate that ensures non - zero asymptotic throughput of eb when the network size becomes extraordinarily large . this also implies that it is impossible to achieve stable throughput and non - power law medium access delay distribution at the same time . using pb , seb and eb as examples , fig . @xmath124 summarizes the key results of this paper . it shows that the heaviness of tail distribution improves from a power law tail with eb , to a heavy but non - power law tail with seb and superlinear pb , and eventually to a light tail with linear - sublinear pb . however , stable throughput is unattainable when eb is replaced by the other backoff schemes . in this section , we characterize the power law behavior of medium access delay distribution for general backoff functions . this is achieved by studying the convergence of moments of medium access delay . we prove that convergent delay moments are attainable if and only if the backoff functions are slower " than exponential function . accordingly , backoff functions such as eb and @xmath41 with @xmath125 , always induce power law delay distribution . in contrast , pb and seb can fully eliminate the power law tail of medium access delay . following the definition of @xmath48 , the probability that a packet is successfully transmitted after @xmath65 consecutive collisions is @xmath126 . therefore , the @xmath127 ( @xmath89 ) moment of the medium access delay @xmath64 is @xmath128\\ = & ( 1-p_c ) \sum_{j=0}^{\infty}p_c^je\left[x_j^n\right ] \\ = & ( 1-p_c ) \sum_{j=0}^{\infty}p_c^je\left[\left\{\sum_{k=0}^{j}c_k + jt_{coll } + t_{succ}\right\}^n\right]\\ = & ( 1-p_c)\sum_{j=0}^{\infty}p_c^je\left[\left(\sum_{k=0}^{j}c_k\right)^n + \right.\\ & \left.\text { other terms with power of } \sum_{k=0}^{j}c_k \text { lower than } n-1\right ] , \end{aligned}\ ] ] where @xmath68 is the time consumed on backoff countdown at the @xmath39 backoff stage , given in ( [ 32 ] ) . we can see that the convergence of @xmath84 $ ] is determined by the most significant term in the rhs of ( [ 9 ] ) , i.e. @xmath129 . \end{aligned}\ ] ] that is to say , the convergence of medium access delay is equivalent to that of the integrated backoff countdown process . similarly , let @xmath130 denote the total number of backoff countdowns before the packet successfully transmits . if a packet is successfully transmitted after @xmath65 collisions , the total number of backoff countdowns denoted by @xmath131 , is @xmath132 where @xmath45 is the backoff counter value at the @xmath39 backoff stage . therefore , the @xmath127 moment of @xmath130 is @xmath133=(1-p_c)\sum_{j=0}^{\infty}p_c^j e\left[\left(\sum_{k=0}^j b_k\right)^n\right ] . \end{aligned}\ ] ] recall that @xmath134 . besides , the countdown time slot @xmath70 is bounded as @xmath135 therefore , ( [ 10 ] ) is lower bounded by @xmath136\ ] ] meanwhile upper bounded by @xmath137.\ ] ] it can be seen that ( [ 10 ] ) converges if and only if @xmath138 $ ] converges . that is to say , the convergence of @xmath84 $ ] is equivalent to that of @xmath138 $ ] . in this sense , we mainly focus on the convergence properties of @xmath138 $ ] in the following discussions . theorem @xmath16 presents the relation between power law delay distribution and backoff function growth rate . * theorem 1 : * a random - access network with an increasing backoff function @xmath0 suffers a power law delay if @xmath122 such that @xmath123 , and does not suffer a power law delay if @xmath1 , @xmath121 . _ proof _ : we first prove that a @xmath0 suffers a power law delay if @xmath122 such that @xmath123 . to prove the argument , we only need to show that there exists an infinite @xmath138 $ ] . by the jensen s inequality , @xmath139 \geq \left[e\left(\sum_{k=0}^{j}b_k\right)\right]^n = \left(\sum_{k=0}^{j}e\left[b_k\right]\right)^n,\ ] ] for @xmath140 . substituting ( [ 33 ] ) into ( [ 11 ] ) , we have the lower bound of @xmath141 $ ] , where @xmath142&\geq\left(1-p_c\right)\sum_{j=0}^{\infty}p_c^j \left(\sum_{k=0}^{j}e\left[b_k\right]\right)^n\\ & \geq \left(1-p_c\right)\sum_{j=0}^{\infty}\left\{p_c^j \sum_{k=0}^{j}\left(e\left[b_k\right]\right)^n\right\}\\ & = \frac{1}{2^n}\left(1-p_c\right)\sum_{j=0}^{\infty}\left\{p_c^j \sum_{k=0}^{j}\left(w_k-1\right)^n\right\}. \end{aligned}\ ] ] the last equality holds because @xmath143=\frac{w_k-1}{2}$ ] . for a @xmath123 , there always @xmath95 and @xmath144 such that @xmath145 , @xmath146 . with @xmath6 , the rhs of ( [ 22 ] ) is be lower bounded by @xmath147 notice that the above lower bound of @xmath141 $ ] becomes infinite when @xmath148 , or equivalently @xmath149 . this leads to the proof that @xmath0 suffers a power law delay then , we prove that @xmath0 does not suffer a power law delay if @xmath1 , @xmath121 . this is equivalent to show that @xmath138 $ ] is finite for all @xmath89 . according to the holder s inequality , it holds that @xmath150 taking the expectations on the both sides , we have @xmath151\leq e\left[\left(j+1\right)^{n-1}\sum_{k=0}^{j}b_k^n \right].\ ] ] from ( [ 11 ] ) , @xmath141 $ ] is upper bounded by @xmath152&\leq\left(1-p_c\right)\sum_{j=0}^{\infty}\left\{p_c^j\left(j+1\right)^{n-1 } \sum_{k=0}^{j}e\left[b_k^n \right]\right\}. \end{aligned}\ ] ] by assumption , @xmath45 is uniformly generated from @xmath153 $ ] . thus , we have @xmath154&=\frac{1}{w_k}\cdot \sum_{l=0}^{w_k-1 } l^n \leq \frac{1}{w_k } \int_{0}^{w_k}x^n\ dx = \frac{1}{n+1}w_k^{n}. \end{aligned}\ ] ] substituting @xmath155\leq \frac{1}{n+1}w_k^{n}$ ] into ( [ 12 ] ) , we have @xmath156\leq\frac{1-p_c}{n+1}\sum_{j=0}^{\infty}\left\{p_c^j\left(j+1\right)^{n-1}\cdot \sum_{k=0}^{j}w_k^{n}\right\}.\ ] ] by definition , for any @xmath2 and @xmath157 there exists a @xmath158 , such that @xmath159 for all @xmath160 . then , the following inequality holds for for all @xmath2 and @xmath157 , @xmath161\leq\frac{\left(1-p_c\right)w_0^n}{n+1}\biggl\{\sum_{j=0}^{k_r}\left[p_c^j\left(j+1\right)^{n-1}\cdot \sum_{k=0}^{j}g^n(k)\right]+ \\ & \sum_{j = k_r+1}^{\infty}\left[p_c^j\left(j+1\right)^{n-1 } \left(\sum_{k=0}^{k_r}g^n(k ) + \sum_{k = k_r+1}^{j}c^nr^{nk}\right)\right]\biggr\}. \end{aligned}\ ] ] the second term in the rhs of ( [ 61 ] ) , which determines the convergence of the upper bound , can be expressed as @xmath162 where @xmath163 is a finite constant for a given @xmath9 . noticeably , ( [ 62 ] ) is the upper bound on @xmath138 $ ] for any @xmath2 . thus , we can safely say @xmath138 $ ] is finite as long as there exists an @xmath2 such that ( [ 62 ] ) is finite . notice that the first term in ( [ 62 ] ) @xmath164 is convergent for all @xmath89 . then , ( [ 62 ] ) is finite if and only if @xmath165 is finite . this can be achieved by selecting a @xmath166 in other words , @xmath167 we can always find a finite upper bound of @xmath141 $ ] by selecting a @xmath9 as in ( [ 65 ] ) . this implies that all the delay moments are finite with the backoff function @xmath0 , i.e. power law delay is mitigated . @xmath168 * remark 4 : * loosely speaking , theorem @xmath16 implies that a backoff function @xmath0 will induce power law delay distribution if and only if @xmath169 for some @xmath2 , i.e. , @xmath0 grows at least as fast as an exponential function . therefore , eb will always induce power law delay distribution , while pb and seb can mitigate the power law delay . this also explains the observations made by @xcite and @xcite that pb achieves better delay performance than eb . * corollary 1 : * for an increasing backoff function @xmath0 , if @xmath170 exists and is denoted by @xmath171 where @xmath172 . then , @xmath138 $ ] is finite if and only if @xmath173 . this implies that @xmath0 yields a power law delay if @xmath174 , and a non - power law delay if @xmath175 . _ proof : _ see appendix a. @xmath168 the backoff functions discussed in corollary @xmath16 are special cases of the general ones discussed in theorem @xmath16 , in that the limit of @xmath176 exists as @xmath177 . it is easy to check that @xmath178 for eb ( @xmath8 ) and @xmath179 for pb ( @xmath180 , @xmath44 ) and @xmath181 for seb ( @xmath182 , @xmath183 ) . thus , eb suffers a power law delay , while pb and seb do not . this is consistent with theorem @xmath16 . moreover , following corollary @xmath16 , we see that @xmath184 $ ] for eb is finite if @xmath185 and infinite if @xmath186 . this is consistent with the results in @xcite and @xcite . notice that theorem @xmath16 is more general in that it applies even if the limit in the lhs of ( [ 78 ] ) does not exist . with the even entries equal to @xmath187 and the odd entries equal to @xmath188 . in this case , @xmath189 but @xmath170 does not exist . ] before leaving the section , note that a non - power law distribution could still be heavy - tailed , such as weibull distribution with slope parameter smaller than @xmath16 . it is of mathematical interest to further investigate the heavy - tailed behavior of delay distribution when different backoff functions are used , although the study may be of little practical relevance . for the completeness of mathematics , we provide detailed analysis on heavy tailed delay distribution in the appendix b. interestingly , our results show that , although seb and super - linear pb ( @xmath190 ) can mitigate the power law tail distribution , the delay distributions are still heavy - tailed . meanwhile , as we further decrease the growth rate of backoff function , delay distribution eventually becomes light - tailed when a linear - sublinear pb is used ( @xmath191 ) . a point to mention is that , our analysis of heavy - tailed behavior of delay distribution is mainly for theoretical interest instead of engineering applications . on one hand , linear - sublinear pb is impractical in the sense that it may yield prohibitively low throughput . on the other hand , although the delay distributions are heavy - tailed with seb and super - linear pb , their tail decaying rates are essentially faster than all power law functions , which is sufficient for any practical engineering implementations . in fact , we will show the superior delay performance of seb and pb in simulations section . in this section , we show that stable throughput is attainable if and only if the backoff function grows at least as fast as an exponential function . here , we say a backoff scheme is throughput - stable if it yields non - zero asymptotic throughput when the network size approaches infinity ( i.e. , becomes extraordinarily large ) . in other words , a throughput - stable backoff scheme guarantees non - zero throughput regardless of the network size , and thus is suitable for practical deployment where the network size can vary randomly over time . the collision probability @xmath48 increases as the network size @xmath34 increases due to higher contention level . meanwhile , @xmath49 decreases with @xmath34 , since the increased contention window size results in a smaller probability of transmission in a given time slot @xcite . since @xmath61 and monotonically increases with @xmath34 , the limit @xmath192 exists . similar argument applies to @xmath49 , such that @xmath193 exists as well . besides , as @xmath34 approaches infinity , it holds that @xmath194 because almost all the nodes are in extremely high backoff stage when the network is stable , such that the probability of transmission approaches zero ( to be justified in ( [ 68 ] ) ) . taking limit on both sides of ( [ 1 ] ) , we have @xmath195 where the last equality holds because @xmath196 . for simple illustration , we assume @xmath197 . consequently , the normalized throughout in ( [ 49 ] ) becomes @xmath198 . notice that the conclusions in this section apply even without this assumption . taking logarithm on both sides of ( [ 46 ] ) , we have @xmath199 then , the asymptotic @xmath200 in ( [ 19 ] ) is @xmath201 where the last equality holds because @xmath202 . for eb , it has been proved in @xcite that @xmath203 . evidently , its asymptotic throughput is non - zero following ( [ 14 ] ) . the throughput stability of general backoff functions is studied in the following subsection . the following theorem @xmath17 provides a criterion to determine the throughput stability of a general backoff function . accordingly , we find that the throughput of seb and pb collapses to zero when the network size is extraordinarily large . * theorem 2 : * for a system in steady - state , an increasing backoff function @xmath0 is throughput - stable if @xmath122 such that @xmath3 , and throughput - unstable if @xmath1 for all @xmath2 . _ proof _ : as per ( [ 2 ] ) and ( [ 1 ] ) , the fixed point system for a backoff function @xmath0 is @xmath204 and @xmath205 the limit of @xmath49 is @xmath206 as @xmath61 . strictly holds in a stable network , although the limit of @xmath48 could be @xmath16 . ] from ( [ 72 ] ) , this indicates that @xmath207 . meanwhile , the fixed point system in ( [ 72 ] ) and ( [ 73 ] ) can be compactly written as @xmath208 . here , @xmath209 is an decreasing function in @xmath48 . this can be justified by calculating @xmath210>0 , \end{aligned}\ ] ] where the inequality holds because @xmath0 is an increasing function . as illustrated in fig . @xmath211 , the solution of the fixed point system is the intersecting point of the @xmath48 and @xmath209 curves . we first prove the first part of theorem @xmath17 . that is , the asymptotic throughput is strictly larger than zero if @xmath3 for some @xmath2 . by definition , there @xmath144 and @xmath95 such that @xmath212 , @xmath213 . then , we have @xmath214 p_c^k . \end{aligned}\ ] ] let @xmath215 denote a constant parameter , we have @xmath216 , @xmath209 and the fixed point solutions for backoff functions @xmath217 and a @xmath169.,width=268 ] in the following , we show that @xmath218 . this is proved by comparing @xmath48 resulting from @xmath0 with that resulting from an exponential backoff function @xmath219 . to distinguish from the notation @xmath220 for @xmath0 , we use @xmath221 to denote the fixed point system when @xmath0 is replaced by @xmath106 . the corresponding fixed point solution is denoted by @xmath222 . from ( [ 64 ] ) , we can infer that @xmath223 we illustrate @xmath221 and @xmath209 in fig . @xmath211 , where we observe that @xmath224 . in fact , this can also be rigorously proved by contradiction . assuming @xmath225 , we have @xmath226 where the first inequality holds because @xmath221 is an decreasing function of @xmath48 and the second inequality is from ( [ 81 ] ) . clearly , ( [ 66 ] ) leads to an contradiction to the assumption @xmath225 . for eb with backoff function @xmath227 , @xmath228 regardless of the value of @xmath229 . can be considered as a scaling factor to the initial contention window size @xmath10 , which is unrelated to the asymptotic behavior of eb . ] accordingly , we have @xmath230 . since @xmath48 is increasing with @xmath34 , it suffices to conclude that @xmath231 . in this case , the asymptotic throughput in ( [ 14 ] ) is non - zero , indicating that @xmath0 is a throughput - stable scheme . next , we prove the second part of theorem @xmath17 . that is , the asymptotic throughput is zero if @xmath1 for all @xmath2 . by definition , for any @xmath2 and @xmath157 there exists a @xmath158 , such that @xmath159 for all @xmath160 . for a given @xmath2 , we have @xmath232 since @xmath207 and the inequality in ( [ 80 ] ) holds for all @xmath2 , we have @xmath233 for all @xmath2 . in other words , @xmath234 must hold for all @xmath2 . this is achievable only if @xmath235 , otherwise we can always find a @xmath236 such that @xmath237 . with @xmath238 , we can derive @xmath239 indicating a zero asymptotic throughput . this completes the proof of the second part of theorem @xmath17 . @xmath168 theorem @xmath17 implies that a backoff scheme is throughput - stable if and only if the backoff function grows at least as fast as an exponential function . accordingly , stable throughput is attainable with eb , but unattainable with pb or seb . evidently , there exists a tradeoff between the tail decaying rate of medium access delay and throughput stability . faster " backoff function is used , the tail distribution of delay becomes heavier while throughput stability improves , and vice versa . in practice , we need to select backoff functions that achieve a balance of the tradeoff . interestingly , we show in the simulation section that pb can achieve good throughput and delay performance within practical range of user population , when the order of backoff function is set properly . we therefore advocate pb as a potential candidate to replace eb in current random access networks , now that there are increasingly more multimedia applications with stringent delay requirements in the network . in this section , we first verify our analysis in previous sections . we then illustrate through numerical simulations that superlinear pb is a good alternative of eb due to its good throughput performance and finite delay moments . .system parameters of @xmath240 [ cols="<,<",options="header " , ] unless otherwise stated , we use the dcf basic - access mode in @xmath240 , where the system parameters are listed in table i. the slot lengths are @xmath241 .,width=336 ] fig . @xmath242 compares saturation throughput and delay performance of eb , seb and pb . here , the total simulation time is @xmath243 time slots . the backoff functions in consideration are eb ( @xmath244 ) , seb ( @xmath245 ) and pb ( @xmath246 ) . the coefficients in the backoff functions are chosen to align the throughputs of different backoff schemes for fair comparison . no retry limit is imposed . when the network size increases from @xmath16 to @xmath23 , normalized saturation throughput and the variance of medium access delay are plotted in fig . @xmath247 and fig . @xmath248 , respectively . we can see that the three schemes yield similar saturation throughput . however , the delay variance of eb is much larger than those of seb and pb . specifically , the delay variance of eb is about @xmath249 times of pb . the root cause of this phenomenon is that the delay variance of eb is infinite following theorem @xmath16 in section iv ( here @xmath250 when @xmath251 ) . the delay variance of eb in fig . @xmath248 are bounded only due to finite simulation time . similarly , we compare the three schemes in fig . @xmath252 when the retry limit @xmath24 . the figure shows that eb perceives much higher packet loss rate than seb and pb due to its inherent characteristics of power - law delay distribution . the higher packet loss rate of eb also causes notably lower throughput compared with seb and pb . an intuitive explanation is that the delay distribution of eb has a heavier " tail than pb and seb when @xmath25 . by setting @xmath24 , it is analogously truncating the delay distribution at a certain point and calculating the tail distribution beyond as the packet loss rate . accordingly , eb with a heavier " tail also yields higher packet loss rate as well . results in fig . @xmath242 and @xmath252 suggest that seb and pb achieve better delay performance than eb under the same network throughput , which is consistent with our results in fig . @xmath124 . in particular , pb yields the smallest delay variance and lowest packet loss rate . .,width=336 ] in fig . @xmath253 , we compare the fairness among users when the three backoff schemes are used . the histogram of the number of successfully transmitted packets is plotted for @xmath12 nodes . the result is an average of @xmath18 independent simulations each with @xmath254 time slots . the average number of packets transmitted per node is @xmath255 $ ] for eb , seb and pb , respectively . this means that the throughput of the three backoff schemes are very similar . we can see in fig . @xmath256 that with eb , the difference in the amount of service received is significant across different nodes . the number of successfully transmitted packets by a node can vary all the way from @xmath46 to @xmath257 . worse still , severe transmission starvation is observed with eb . on average more than @xmath258 of nodes transmit very few or even zero packets during the entire simulation time . in fig . @xmath259 , seb performs much better than eb , where the disparity of successful transmission is smaller and the maximum number of successfully transmitted packets is reduced to around @xmath260 . however , we can still observe around @xmath261 of the nodes in transmission starvation . this is because large delay can still occur with non - negligible probability when seb is adopted , although the power law tail is mitigated . in vivid contrast , we can see in fig . @xmath262 that the range of the number of successful transmissions is significantly reduced and no transmission starvation occurs when pb is implemented . this indicates that pb achieves the fairest air time allocation among nodes . such observation is consistent with our results in fig . @xmath124 that pb has the lightest " tail of delay distribution among the three schemes . with different backoff parameters , we plot the normalized saturation throughput of pb in fig . @xmath263 when the number of contending nodes varies from @xmath16 to @xmath264 . besides , the saturation throughput of beb is also presented for comparison . we can see that the throughput of pb gradually decreases as @xmath34 increases . in fact , the throughput will decrease to zero when @xmath34 becomes significantly large . on the other hand , the throughput of exponential backoff converges to a constant as @xmath34 increases . these observations verify out analysis in theorem @xmath17 that the throughput is stable with eb while unstable with pb . however , we also see that pb with @xmath265 can sustain higher saturation throughput than beb for all @xmath266 . this implies that high efficiency can be obtained with pb in practical scenarios when the order of backoff exponent is set properly . therefore , we can safely enjoy the small delay jitter and better user fairness brought by pb without worrying about the instability of asymptotic throughput in practical systems . ) when @xmath267 varies.,width=336 ] simulation results in this section show that pb can achieve high throughput , smaller delay jitter and good user fairness at the same time within practical range of user population . it is therefore a better alternative than eb , especially for carrying real - time traffics with stringent delay requirements . in this paper , we have analyzed the tail delay distribution and throughput stability of general backoff functions . a tradeoff has been established between the tail decaying rate of medium access delay distribution and the stability of throughput . in particular , we found that power law delay distribution can be avoided if the backoff functions are slower " than an exponential function . examples of such slow " backoffs are pb and seb . in addition , the delay distribution becomes light tailed when linear - sublinear pb is used . on the other hand , non - zero asymptotic throughput is attainable only when backoff functions grow at least as fast as an exponential function , such as eb . for practical implementation , we show that pb obtains good throughput performance within a practical range of user population . meanwhile , all delay moments with pb are finite as opposed to the infinite delay moments with eb . as such , we advocate pb as a better alternative than eb , now that there are increasingly more multimedia applications with stringent delay requirements in the network . _ proof : _ we first show that the limit @xmath268^n}\ ] ] exists . since @xmath269 , it holds that @xmath270 , @xmath271 such that @xmath272 equivalently , we have @xmath273 then , it holds that @xmath274 the second term of the rhs of ( [ 70 ] ) can be rewritten as @xmath275^n } = \frac{1}{\sum_{k=0}^{k_0}\left[\frac{g(k)}{g(j+1)}\right]^n + \sum_{k = k_0 + 1}^{j}\left[\frac{g(k)}{g(j+1)}\right]^n},\ ] ] which is upper bounded by @xmath276^n + \sum_{k=1}^{j - k_0}\left(\frac{1}{\gamma+\epsilon}\right)^{kn } } \end{aligned}\ ] ] and lower bounded by @xmath277^n + \sum_{k=1}^{j - k_0}\left(\frac{1}{\gamma-\epsilon}\right)^{kn}}. \end{aligned}\ ] ] notice that @xmath278 ( @xmath279 ) can be made arbitrarily small ( large ) as @xmath177 . when @xmath174 , we see that the lower bound and upper bounds converge to @xmath280 as @xmath177 . otherwise , when @xmath175 , the lower bound and upper bounds converge to @xmath46 as @xmath177 . in both cases , @xmath281 next , we prove that @xmath138 $ ] is finite only if @xmath282 . from ( [ 22 ] ) , we have @xmath283\geq\frac{1}{2^n}\left(1-p_c\right)\sum_{j=0}^{\infty}\left\{p_c^j \sum_{k=0}^{j}\left(w_k-1\right)^n\right\}. \end{aligned}\ ] ] letting @xmath284 we see that the limit @xmath285 exists following the similar argument in ( [ 70])-([76 ] ) . using the ratio test , the lower bound in ( [ 75 ] ) is finite only if @xmath286 since @xmath287 , we have @xmath288 equivalently , it holds that @xmath289 with ( [ 41 ] ) and ( [ 42 ] ) , the lower bound converges only if @xmath290 from ( [ 63 ] ) , the above condition can be rewritten as @xmath291 . we therefore reach the proof of _ only if _ part , since a finite lower bound is a necessary condition for finite @xmath141 $ ] . then , we prove that @xmath138 $ ] is finite if @xmath292 . from ( [ 74 ] ) , we have @xmath293\leq\frac{1-p_c}{n+1}\sum_{j=0}^{\infty}\left\{p_c^j\left(j+1\right)^{n-1}\cdot \sum_{k=0}^{j}w_k^{n}\right\}.\ ] ] similarly to the argument in the proof of the _ only if _ argument , the following limit exists , where @xmath294 using the ratio test , the upper bound in ( [ 74 ] ) converges if @xmath291 . a convergent upper bound is a sufficient condition for finite @xmath141 $ ] , which leads to the proof that @xmath138 $ ] is finite . notice that @xmath61 strictly holds in a random access network with finite number of nodes . therefore , @xmath173 holds for all @xmath89 when @xmath175 . when @xmath174 , however , @xmath173 is violated for all @xmath295 . in other words , there always exists infinite delay moment when @xmath174 , while all the delay moments are finite when @xmath175 . following the definition of power law distribution , we reach the proof of corollary @xmath16 . @xmath168 as illustrated in fig . @xmath74 , power law distribution belongs to a subclass of heavy - tailed distribution ( region @xmath17 in fig . @xmath74 ) . however , the converse statement , that a non - power law distribution is not heavy - tailed , is not true ( region @xmath14 ) . this motivates us to further study the heavy - tailed behavior of medium access delay distribution of different backoff functions . interestingly , we show that superlinear pb and seb , which eliminate the power law tail , still yield heavy - tailed delay distribution ( region @xmath14 ) . meanwhile , linear - sublinear pb , i.e. , @xmath180 and @xmath296 , yields a light - tailed delay distribution ( region @xmath90 ) . since medium access delay distribution is a power law distribution with eb , it is also a heavy - tailed distribution . for pb , we show that the delay distribution is heavy - tailed if @xmath297 , whereas lighted - tailed if @xmath296 . from ( [ 16 ] ) , it holds that @xmath298 \geq \text{e}\left[\lambda^n\right]\cdot l_{\text{min}}^n,\ ] ] where @xmath299 . from ( [ 22 ] ) , we have @xmath300\geq \left(1-p_c\right)\sum_{j=0}^{\infty}\left\{p_c^j \sum_{k=0}^{j}\left(\text{e}\left[b_k\right]\right)^n\right\}. \end{aligned}\ ] ] for pb , we substitute @xmath143=\frac{1}{2}\left(1+k^b\right)w_0 $ ] into the rhs of ( [ 45 ] ) , @xmath301 & \geq \left(1-p_c\right)\left(\frac{w_0}{2}\right)^n\sum_{j=0}^{\infty}\left\{p_c^j \sum_{k=0}^{j}\left(1+k^b\right)^n\right\}\\ & \geq \left(1-p_c\right)\left(\frac{w_0}{2}\right)^n\sum_{j=0}^{\infty}\left\{p_c^j \sum_{k=0}^{j}k^{bn}\right\}\\ & \geq \left(1-p_c\right)\left(\frac{w_0}{2}\right)^n\sum_{j=0}^{\infty}\left\{p_c^j \int_{0}^{j } t^{bn } dt\right\}\\ & = \frac{1-p_c}{bn+1}\left(\frac{w_0}{2}\right)^n\sum_{j=0}^{\infty}p_c^j j^{bn+1}. \end{aligned}\ ] ] here , @xmath302 can be represented as @xmath303 where @xmath304 is a lerch s transcendent . when @xmath305 and @xmath306 , it holds that @xmath307 ( cf . @xcite , p. @xmath308 ) , where @xmath309 is a gamma function . correspondingly , we can write @xmath310&\geq \frac{1-p_c}{bn+1}\left(\frac{w_0}{2}\right)^n\phi \left(p_c , -(bn+1 ) , 0\right)\\ & \approx\frac{1-p_c}{bn+1}\left(\frac{w_0}{2}\right)^n\gamma(bn+2 ) \left(\ln\frac{1}{p_c}\right)^{-\left(bn+2\right)}. \end{aligned}\ ] ] recall in ( [ 26 ] ) that @xmath72 is heavy - tailed if and only if @xmath311=\infty,\ \ \forall \lambda>0.\ ] ] we substitute the inequality in ( [ 31 ] ) into the lhs of ( [ 38 ] ) , @xmath312 \\ \geq & \sum_{n=0}^{\infty } \frac{\lambda^n}{n ! } \text{e}\left[\lambda^n\right]l_{\text{min}}^n \\ \geq & \sum_{n=0}^{\infty } \frac{1-p_c}{bn+1}\left(\frac{w_0l_{\text{min}}\lambda}{2}\right)^n\frac{1}{n ! } \gamma(bn+2 ) \left(\ln\frac{1}{p_c}\right)^{-\left(bn+2\right)}. \end{aligned}\ ] ] the lhs of ( [ 38 ] ) is finite only if the lower bound is finite . using the ratio test to lower bound in ( [ 48 ] ) , we obtain the test parameter as @xmath313 the delay distribution of pb is heavy - tailed if @xmath314 is larger than @xmath16 for all @xmath315 . note that , when @xmath107 is a very large real positive number , gamma function can be well approximated by @xmath316 ( cf . @xcite , p. @xmath317 ) . we denote @xmath318 , then @xmath319 by lhospital s rule , @xmath320 . then , the rhs of ( [ 69 ] ) equals to @xmath321 therefore , we have @xmath322 accordingly , the test parameter @xmath314 in ( [ 30 ] ) is @xmath323 we can see that @xmath324 for all @xmath325 when @xmath297 . therefore , the delay distribution of a super - linear pb is heavy - tailed . for linear - sublinear pb with @xmath326 , however , we currently have not obtained conclusive analytical results to verify its heavy - tailed behavior . instead , we numerically calculate the probability mass function of @xmath130 and find it matches the features of light - tailed distribution when @xmath296 . the probability mass function can be obtained through calculating its probability generating function ( cf . @xcite , p. @xmath327 ) . we plot @xmath328)$ ] against @xmath87 with different @xmath296 in fig . @xmath329 . besides , the results of linear regressions and corresponding @xmath330 values are also provided . in all three cases , we can see that @xmath328)\thicksim -\lambda_0 n$ ] for some @xmath331 , indicating @xmath332\thicksim e^{-\lambda_0 n}$ ] . thus , the delay distribution is light - tailed distribution with an exponentially decaying tail . it can be easily shown that seb also suffers heavy - tail delay distribution . to see this , we substitute @xmath336=\frac{1}{2}r^{k^a}w_0 $ ] into ( [ 45 ] ) , then @xmath337\geq \left(1-p_c\right)\left(\frac{w_0}{2}\right)^n\sum_{j=0}^{\infty}\left\{p_c^j \sum_{k=0}^{j}r^{nk^a}\right\}. \end{aligned}\ ] ] for any @xmath183 , @xmath2 and @xmath297 , there always @xmath338 so that @xmath339 as we have proved in previous subsection , the following inequality holds for pb with @xmath340 and @xmath297 , @xmath341=\infty,\ ] ] @xmath342 . following ( [ 34 ] ) and ( [ 35 ] ) , it suffices to claim to that the moment generating function of seb diverges , i.e. , @xmath343=\infty,\ \ \forall \lambda>0.\ ] ] therefore , the distribution of medium access delay is heavy - tailed . y. j. zhang , s. c. liew , and d. r. chen , sustainable throughput of wireless lans with multipacket reception capability under bounded delay - moment requirements , " _ ieee transactions on mobile computing _ , vol 9 , pp.1226 - 1241 , sept 2010 . a. kumar , e. altman , d. miorandi , and m. goyal , new insights from a fixed point analysis of single cell ieee 802.11 wlans , " _ ieee / acm transactions on networking _ , vol.15 , no . 588 - 601 , aug . 2007 .

exponential backoff ( eb ) is a widely adopted collision resolution mechanism in many popular random - access networks including ethernet and wireless lan ( wlan ) . the prominence of eb is primarily attributed to its asymptotic throughput stability , which ensures a non - zero throughput even when the number of users in the network goes to infinity . recent studies , however , show that eb is fundamentally unsuitable for applications that are sensitive to large delay and delay jitters , as it induces divergent second- and higher - order moments of medium access delay . essentially , the medium access delay follows a power law distribution , a subclass of heavy - tailed distribution . to understand and alleviate the issue , this paper systematically analyzes the tail delay distribution of general backoff functions , with eb being a special case . in particular , we establish a tradeoff between the tail decaying rate of medium access delay distribution and the stability of throughput . to be more specific , convergent delay moments are attainable only when the backoff functions @xmath0 grows slower than exponential functions , i.e. , when @xmath1 for all @xmath2 . on the other hand , non - zero asymptotic throughput is attainable only when backoff functions grow at least as fast as an exponential function , i.e. , @xmath3 for some @xmath2 . this implies that bounded delay moments and stable throughput can not be achieved at the same time . for practical implementation , we show that polynomial backoff ( pb ) , where @xmath0 is a polynomial that grows slower than exponential functions , obtains finite delay moments and good throughput performance at the same time within a practical range of user population . this makes pb a better alternative than eb for multimedia applications with stringent delay requirements . medium access control , backoff algorithms , wireless lan ( wlan ) , power law delay .

the tau decay into 3 pseudoscalar particles has been studied since the discovery of the tau lepton . moreover , decays to final states containing kaons provide a direct determination of the strange quark mass and cabbibo kobayashi maskawa ( ckm ) matrix element @xmath8 . although the measurements of the @xmath0 decay from cleo used sufficiently large statistics , the overall uncertainty is as large as @xmath110% @xcite . the resonant states in @xmath0 decay have been studied only with poor statistics by cleo @xcite . the resonance study is itself of interest for strange spectral function measurements and leptonic violation studies . theoretically , two intermediate resonances , the @xmath9 and @xmath10 , are both expected to contribute to tau decay into @xmath11 : @xmath12 and @xmath13 @xcite . these are shown in [ eventshape](a ) and ( b ) , respectively . in this study , we present new measurements of the branching ratios for @xmath5 , @xmath0 , @xmath6 , and @xmath7 decays ( unless specified , the charge conjugation decay is also implied throughout this paper ) . the measurements of these kaon containing 3prong decay modes and @xmath5 decay are correlated due to particle misidentification , so they should be treated simultaneously . in the center of mass system , where one @xmath14 ( @xmath15 ) decays to the signal mode ( @xmath16 ) and the other @xmath14 ( @xmath17 ) decays to the pure leptonic modes ( @xmath18 or @xmath19 ) . each figure shows one of two possible intermediate resonance states in the signal mode . , scaledwidth=60.0% ] this study uses a huge amount of tau decays collected with the belle detector at the kekb asymmetric energy @xmath20 collider @xcite . since 1999 , the belle experiment has accumulated approximately 850 @xmath2 data until the summer of 2008 . the belle detector is a large - solid - angle magnetic spectrometer that consists of a silicon vertex detector ( svd ) , a 50-layer central drift chamber ( cdc ) , an array of aerogel threshold cherenkov counters ( acc ) , a barrel - like arrangement of time - of - flight scintillation counters ( tof ) , and an electromagnetic calorimeter comprised of csi(tl ) crystals located inside a super - conducting solenoid coil that provides a 1.5 t magnetic field . an iron flux - return located outside of the coil is instrumented to detect @xmath21 mesons and to identify muons ( klm ) . the detector is described in detail elsewhere @xcite . in this study , we used 668.9 @xmath2 data collected on the @xmath22(4s ) resonance , 10.58 gev , and 60 mev below it ( off resonance ) , which corresponds to the production of @xmath23 @xmath14pairs . to select @xmath0 and other 3 prong decays , @xmath5 , @xmath6 and @xmath7 , first , tau pair events are selected . we require the number of tracks to be four and the sum of charges of these tracks to be zero , when the transverse momentum of a track in the laboratory frame is greater than 0.1 gev/@xmath24 , and the track extrapolates to the interaction point within @xmath251 cm transversely and @xmath255 cm along the beam direction . the sum of the reconstructed momenta in the center of mass ( cm ) frame is required to be less than 10 gev/@xmath24 , and the sum of energy deposited in the calorimeter is required to be less than 10 gev . the maximum transverse momentum is required to be greater than 0.5 gev/@xmath24 , to reject two photon events which include many low transverse momentum tracks . to reject beam related background , we cut on the position of the reconstructed event vertex , requiring it to be closer to the interaction point @xmath250.5 cm transversely and @xmath253 cm along the beam direction . the missing mass @xmath26 and in the cm frame are efficient variables for rejecting two photon and bhabha backgrounds . in the definition of missing mass , @xmath27 and @xmath28 are the four momenta of measured tracks and photons , respectively , and @xmath29 is the initial cm frame momentum of the @xmath20 beams . we require that the missing mass should be greater than 1 gev/@xmath30 and less than 7 gev/@xmath30 , and the polar angle with respect to the beam direction in the cm frame should be greater than 30@xmath31 and less than 150@xmath31 . particle identification is performed to select tau events containing one lepton ( electron or muon ) and 3 hadrons ( pions or kaons ) . the magnitude of thrust is evaluated and is required to be greater than 0.9 , to suppress two photon and @xmath32 backgrounds , where the thrust is defined by the maximum of @xmath33 when @xmath34 is the momentum of @xmath35th track and where @xmath36 is the unit vector in the direction of the thrust axis the direction maximizing the thrust . also the invariant mass of charged tracks and gamma clusters for each side are required to be less than the tau mass . finally we require that there are no , @xmath37 , and energetic @xmath38 on the signal side . [ cutstatus ] shows the performance of the criteria used to select the 13 prong sample , the events finally selected are candidates for @xmath39 , where @xmath40 is a hadron identified as either pion or kaon . at this stage , the reconstruction efficiency of @xmath0 decay is @xmath41 , while the dominant background modes , @xmath42 , @xmath43 , and @xmath32 are reduced to 6.0% , 1.8% , and 0.004% , respectively . the reconstruction efficiency of two photon background is so small ( @xmath44 % ) that it is negligible . signal mc , solid lines show two photon background , and dotted lines show @xmath32 background . , scaledwidth=60.0% ] one of the most important issues in this analysis is the separation of kaons and pions . in the belle experiment , @xmath45 information from the cdc , hit information of the acc , and time of flight from the tof are used to construct the . [ kidlikelihoodratio ] shows the likelihood ratio of kaon identification , @xmath46 , for the charged particles versus their momenta . a clean separation between kaon and pion is observed in the whole momentum region relevant for this analysis , . as is discussed later , we choose a relatively tight particle identification ( pid ) condition for the kaon ( @xmath47 ) and a relatively loose condition for the pion . with these criteria , the efficiency of kaon identification is @xmath48 , and the fake rate of pion mis identification to kaon is @xmath49 . the calibration of the kaon and pion identification efficiencies and their fake rates are carried out by the data using the kaon and pion tracks in the @xmath50 decays , @xmath51 sample , where one knows which tracks are kaons and pions from kinematics and the charges of the tracks . we evaluate the efficiencies and fake rates for this calibration sample and compare them to the monte carlo ( mc ) expectations . from this comparison , we prepare a correction table as a function of track momenta and polar angles , and apply them to the mc . , versus the momentum of the particles . the red filled circles and blue empty circles represent the real kaons and pions , respectively.,scaledwidth=60.0% ] this kaon identification criterion is determined by maximizing the figure of merit ( fom ) , . the figure of merit is defined as : @xmath52 . the result of fom is shown in fig . [ pidstudy1 ] , where one can see that the fom is maximal with the particle identification criteria used in this analysis . for the case of electron and muon identification , the probabilities to be an electron or muon are evaluated using the information from ecl , klm , and other particle identification detectors . also the efficiencies and the systematic uncertainties of lepton identification are evaluated by using the control sample @xmath53 . to reduce the feed - down background which contain and @xmath37 , we reconstruct and @xmath37 signals explicitly . is reconstructed from two charged pion tracks having invariant mass @xmath54 within @xmath2513.5 mev/@xmath30 of the mass . to improve the purity of , the point of closest approach to the interaction point along the extrapolation of each track is required to be greater than 0.3 cm transverse to the beam direction . the azimuthal angle between the momentum vector and the decay vertex vector of reconstructed is required to be less than 0.1 rad . the distance between the two daughter pion tracks at their interception point is required to be less than 1.8 cm , and the flight length of transversely to the beam direction is required to be greater than 0.08 cm . finally we required that the invariant mass of a candidate should not be located in the @xmath55 , @xmath56 , and @xmath38 mass range , with the daughter tracks assumed to be electrons or pions and protons or anti protons as appropriate . also the @xmath37s are reconstructed from two gamma clusters . the energy of each gamma is required to exceed 50 mev for the candidates in the barrel part of the calorimeter , and 100 mev for the candidates in the endcap part of the calorimeter . also we applied a requirement on the invariant mass of two gammas @xmath57 to be within @xmath2516.2 mev/@xmath30 of the @xmath37 mass . an additional selection that there should be no energetic gamma whose energy is greater than 0.3 gev , is applied to reject remaining events containing @xmath37 . we generated @xmath15 signal mode mc data with the tauola @xcite based kkmc @xcite generator , to estimate signal efficiency . also the detector response of all mc data sets is simulated with the geant @xcite simulator . the tauola generator uses a decay model for @xmath0 , where the process is mediated by @xmath58 and @xmath59 resonances and their interference ( we will call this decay model as `` normal '' decay model , in the following ) . to check for a possible bias of efficiency due to the specific decay model used , events assuming phase space decay of @xmath0 were also generated using the kkmc program . the evaluated efficiencies for @xmath60 for both cases are compared in fig . [ efficiencycurve ] . the relative difference of between is around 1 % , so small that it is negligible . decay as a function of @xmath60 . squares and triangular points represent the normal decay , and the phase space decay , respectively.,scaledwidth=60.0% ] the average efficiencies and the of the background for all 3prong decays are summarized in table [ efficiencytable ] , where we used the normal decay model for the determination of the efficiency . one can see that the fake rate of cross feed from @xmath5 to @xmath0 , is very small , but since the branching ratio of @xmath5 is large , there is a substantial contamination in the reconstruction of @xmath0 events coming from misidentified @xmath5 events . [ cols="<,^,^,^,^ " , ] [ branchingratio ] we present results from the measurement of the branching ratio for the @xmath0 decay mode , as well as those for the @xmath5 , @xmath6 , and @xmath7 decay modes . the result for the @xmath0 decay mode is compared with previous measurements in fig . [ kpipibranchingratiohistory ] . the statistical and systematic uncertainties are compatible with a recent precision measurement from babar , although the branching ratio of this study is more consistent with world average . we thank the kekb group for the excellent operation of the accelerator , the kek cryogenics group for the efficient operation of the solenoid , and the kek computer group and the national institute of informatics for valuable computing and sinet3 network support . we acknowledge support from the ministry of education , culture , sports , science , and technology of japan and the japan society for the promotion of science ; the australian research council and the australian department of education , science and training ; the national natural science foundation of china under contract no . 10575109 and 10775142 ; the department of science and technology of india ; the bk21 program of the ministry of education of korea , the chep src program and basic research program ( grant no . r01 - 2005 - 000 - 10089 - 0 ) of the korea science and engineering foundation , and the pure basic research group program of the korea research foundation ; the polish state committee for scientific research ; the ministry of education and science of the russian federation and the russian federal agency for atomic energy ; the slovenian research agency ; the swiss national science foundation ; the national science council and the ministry of education of taiwan ; and the u.s.department of energy . 99 e. gmiz , _ et al . . lett . * 94 * , 011803 ( 2005 ) . r. barate , _ et al . _ , eur . phys . j. c * 4 * , 29 ( 1998 ) . g. abbiendi , _ et al . j. c * 35 * , 437 ( 2004 ) . r. a. briere , _ et al . _ , phys . rev . * 90 * , 181802 ( 2003 ) . b. aubert , _ et al . lett . * 100 * , 011801 ( 2008 ) . d. m. asner , _ et al . d * 62 * , 072006 ( 2000 ) . m. finkemeier and e. mirkes , z. phys . c * 69 * , 243 ( 1996 ) . s. kurokawa and e. kikutani , nucl . instr . and . a499 , 1 ( 2003 ) , and other papers included in this volume . a. abashian _ et al . _ ( belle collab . ) , nucl . instr . and meth . a * 479 * , 117 ( 2002 ) . r. brun _ et al . _ , geant 3.21 , cern report dd / ee/84 - 1 , ( 1984 ) . d. epifanov _ et al . _ ( belle collab . ) , phys . b * 654 * , 65 ( 2007 ) . s. banerjee , b. pietrzyk , j. m. roney and z. was , phys . d * 77 * , 054012 ( 2008 ) .

we present a study of @xmath0 decay using @xmath1669 @xmath2 data , collected with the belle detector at the kekb asymmetric energy @xmath3 collider . the data is recorded at a center of mass energy 10.58 gev . the result for the branching ratio is : @xmath4 we also present results of the precise measurement of the branching ratio of other 3prong decay modes , @xmath5 , @xmath6 , and @xmath7 . .

spin - triplet superconductivity has been one of the major issues in the field of condensed - matter physics . nearly all the conventional and unconventional superconductors known to date are spin - singlet paired . the best - known example of triplet pairing is not a superconductor but a superfluid @xmath4he , where atomic cooper pairs are formed in spin - triplet channel.@xcite only a few materials of spin - triplet superconductivity have so far been confirmed in strongly correlated electron systems , which iclude ruthenium oxide sr@xmath5ruo@xmath6@xcite and some heavy - fermion compounds such as upt@xmath7.@xcite here , some questions will naturally arise . one is whether the electron correlation can take an essential part in superconductivity carried by spin - triplet pairs . another is how the behavior differs from that of spin - singlet superconductivity . in this manner , research on spin - triplet superconductivity may offer an opportunity to expose unknown physical phenomena . quite recently , a new mechanism of the spin - triplet superconductivity has been proposed in a fairly simple correlated electron system.@xcite the model consistes of two hubbard chains coupled with zigzag bonds and has a unique structure of hopping integrals : sign of the hopping integrals changes alternately along the zigzag bonds connecting two chains , while the sign along the one - dimensional ( 1d ) chain is always negative . [ a model where all the hopping integrals are taken to be positive is equivalent under canonical transformation . ] under this sign rule of the hopping integrals , the ring - exchange mechanism@xcite yields ferromagnetic spin correlations , and accordingly , attractive interaction between electrons is derived . in ref . , the argument was developed on the basis of only the static properties such as pair binding energy , spin excitation gap , and pair correlation function , as well as spin correlation function . therefore , further investigations including dynamical properties have been desired . as for real materials , this mechanism@xcite may be of possible relevance to superconductivity observed in quasi-1d organic conductor ( tmtsf)@xmath5x [ x = pf@xmath8 , clo@xmath6 , etc . ] , the so - called bechgaard salts.@xcite the system exhibits a rich phase diagram upon variation of the pressure and temperature . at low temperatures , the phase changes , in the order , as the spin - peierls insulator , antiferromagnetic insulator , spin - density - wave ( sdw ) insulator , superconductivity , and paramagnetic metal , with increasing pressure . so far , experimental evidences that the superconducting state occurs in the triplet channel have been piled up,@xcite although not much is known on the nature of the pairing . a newly synthesized copper - oxide compound pr@xmath5ba@xmath6cu@xmath9o@xmath10@xcite may also be a relating system . this material consists of both the single cuo chains ( as in prba@xmath5cu@xmath7o@xmath9 ) and the double cuo chains ( as in prba@xmath5cu@xmath6o@xmath11 ) , and those chains are separated by insulating cuo@xmath5 plains . it has been reported that the double chains turn into a superconducting state below @xmath12 k.@xcite although the signs of the hopping integrals seem not to satisfy the ferromagnetic sign rule , the structure of the double cuo chains bears a certain similarity to our model.@xcite the purpose of the present study is therefore to build up understanding of the ring - exchange superconducting mechanism by calculating dynamical quantities for the same model proposed in ref . . to see the excitations in the spin and charge degrees of freedom separately , we here calculate the momentum - dependent dynamical spin - spin and density - density correlation functions . we use the dynamical density - matrix renormalization group ( ddmrg ) method for calculating dynamical quantities,@xcite which is an extension of the standard dmrg method.@xcite the obtained results with high resolutions enable us to discuss details of the fundamental properties on the spin and charge excitations . thus , we can find some interesting features in the low - energy physics of our model . we will show that , although the spin and charge excitation spectra are basically the same as those of the two weakly - coupled 1d hubbard ( and @xmath0@xmath13@xmath3 ) chains at quarter filling , the small hopping integral between the chains plays a crucial role in the short - range correlations and low - energy excitaions ; i.e. , the ferromagnetic spin correlations between electrons on the neighboring sites is enhanced and the spin - triplet pairing correlations between the electrons is induced , of which the consequences are clearly seen in the calculated spin and charge excitation spectra at low energies . we hope that the present investigation will provide deeper insight into the mechanism of the spin - triplet superconductivity . our paper is organized as follows . in sec . ii , we define the 1d @xmath0-@xmath1-@xmath14 model and introduce the physical quantities of interest , namely , spin and charge excitation spectra . in sec . iii , exact solution of the spin ( and charge ) excitation spectrum in the noninteracting case is presented , and by comparing our result with the exact one , we evaluate the performance of our ddmrg method . we then study the spectra with onsite coulomb interaction and discuss relevance to the spin - triplet superconductivity . we close with a summary in sec . we consider the 1d hubbard model defined by the hamiltonian @xmath15 where @xmath16 ( @xmath17 ) is the creation ( anihilation ) operator of an electron with spin @xmath18 @xmath19 at site @xmath20 , @xmath21 is the number operator , @xmath1 and @xmath0 are the nearest - neighbor and next - nearest - neighbor hopping integrals respectively , and @xmath14 is the on - site coulomb interaction . we choose the signs of the hopping integrals @xmath0 and @xmath1 to be positive , so that the two spins on a triangle of the lattice align ferromagnetically when @xmath14 is large.@xcite the lattice structure of our model is shown schematically in fig . [ fig0](a ) . we call the chain along the @xmath0 ( @xmath1 ) hopping integral the @xmath0-chain ( @xmath1-chain ) . the dispersion relation is given by @xmath22 where @xmath23 is the lattice constant along the @xmath0-chain ( we set @xmath24 hereafter ) . we consider the case where there are four fermi momenta @xmath25 and @xmath26 ( @xmath27 ) ; i.e. , the case where there are two branches in the noninteracting band dispersion [ see fig . [ fig0](b ) ] . this case occurs when the hopping integrals satisfy the condition @xmath28 } { \sin^2\big[(2-\rho)\pi\big]}\ ] ] where @xmath29 is the band filling . in this paper , we restrict ourselves to the case where @xmath1 is a few times as large as @xmath0 and the system is at quarter filling , @xmath30 . hence , the model can be regarded as a double @xmath1-chain hubbard model weakly coupled by the @xmath0-chain . because @xmath0 is much smaller than @xmath1 , it would be very useful to allow a case of @xmath31 for familiarization with our results . in the limit of @xmath32 , the system is equivalent to the two independent 1d hubbard chains at quarter filling since electrons are distributed equally to the two chains . the noninteracting band dispersion reads @xmath33 and there are four fermi momenta @xmath34 and @xmath35 in the original brillouin zone defined for the @xmath0-chain . we define @xmath36 as the ` nesting ' vector in our model at @xmath37 . we use this definition of the brillouin zone throughout the paper . we calculate the spin excitation spectrum @xmath38 defined with @xmath39 , and the charge excitation spectrum @xmath40 defined with @xmath41 . here , @xmath42 and @xmath43 are , respectively , the wavefunction and energy of the ground state of the hamiltonian eq . ( [ hamiltonian ] ) . the ddmrg technique is applied to calculate the excitation spectra . we here use the open - end boundary condition ( obc ) for accurate calculation , because the system is relatively hard to deal with by the dmrg method due to large long - range hopping integrals.@xcite when the obc is used , we need to use the quasimomenta @xmath44 for integers @xmath45 on a chain with @xmath46 sites in order to express the momentum - dependent operators @xmath47 and @xmath48.@xcite [ @xmath49 for @xmath50 and @xmath51 in the noninteracting case ( @xmath52 ) . broadening of the spectrum @xmath53 is introduced . five momenta indicated with arrows in the left side denote gapless points . inset : exact spin structure factor @xmath54.,height=453 ] first , let us consider the noninteracting case , @xmath52 , where the model is exactly solvable . in this case , an excitation corresponds to creating a particle - hole pair for the ground state and therefore we can obtain the exact spectrum of spin excitations : @xmath55 where small @xmath56 is introduced to regularize the poles at particular frequencies @xmath57 . note that the spectrum of charge excitations at @xmath52 is exactly twice as large as that of the spin excitations , i.e. , @xmath58 , because @xmath59 is just a sum over both up and down spins . [ @xmath49 in the noninteracting case ( @xmath52 ) calculated by the ddmrg method . the same set of the parameter values as in fig . [ fig1 ] is used . right ( left ) panel shows the result for @xmath60 and @xmath61 ( @xmath62 and @xmath63 ) . inset : spin structure factor @xmath54 obtained from @xmath57-integration of @xmath64.,height=453 ] in fig . [ fig1 ] , we show the exact noninteracting spin excitation spectrum @xmath64 given in eq . ( [ exactspec ] ) . for small @xmath65 @xmath66 , the spectrum contains two predominant features : ( i ) large - weighted structure consisting of the double sine curve , whose dispersions are approimately written as @xmath67 , and ( ii ) small - weighted continuum structure at low frequencies , which arises from excitations between different branches of the noninteracting bands . a zero - energy excitation is caused by the creation of a particle - hole pair just at the fermi level @xmath68 , so that the gap closes at five momenta @xmath69 , @xmath70 @xmath71 , @xmath72 , @xmath72 , and @xmath73 . now , using the ddmrg method , we attempt to reproduce the noninteracting spin excitation spectrum . the result is shown in fig . [ fig1 ] . since the noninteracting model poses a nontrivial problem to the ddmrg technique , it gives us a relevant accuracy test . when carrying out the ddmrg calculation , we have to take into account the required cpu time . usually , the ddmrg method takes much longer cpu time than the standard dmrg method because the excited states must be obtained and an asked quantity must be calculated ( almost ) individually for each frequency . additionally , a required cpu time @xmath74 increases rapidly with frequency @xmath57 and system size @xmath46 in the ddmrg calculation , which is estimated approximately as @xmath75 @xmath76 and as @xmath77 if we keep other conditions . hence , it would be efficient to take a relatively small system for obtaining an overview of the spectrum and a larger system for studying the detailed structure at low frequencies . let us then check our ddmrg result with the exact spectrum . in the right panel of fig . [ fig2 ] , we show the spin excitation spectrum @xmath64 at @xmath52 calculated with the ddmrg method in a chain with @xmath60 sites . the structure of the double sine curve can be clearly seen . however , it is hard to see dispersive structures in the continuum spectra at low frequenciese because only discrete peaks can be obtained due to the finite - size effect . we need to take larger systems to resolve this problem since the resolution of spectrum can be improved in proportion to the system size . we therefore choose to double the system size , @xmath62 , and consider the low - energy excitations . the result is shown in the left panel of fig . the resolution is obviously improved and we can now confirm the five momenta at which the zero - energy excitations occur . moreover , we can see good agreement in the spin structure factor @xmath54 @xmath78 $ ] between the ddmrg and exact results , as shown in the inset of fig . thus , we are confident that our ddmrg method indeed enable us to study the detailed structures of the relatively complicated spectrum . for the information , we keep @xmath79 ( @xmath80 ) density - matrix eigenstates to obtain the spectrum for @xmath60 ( @xmath81 ) sites . note that a larger @xmath82 value should be necessary to get the true ground state and excited state @xmath83 ( or @xmath84 ) of the system . we therefore set @xmath85 in the first @xmath86 ddmrg sweeps . next , let us see how the spin excitation spectrum is modified by the inclusion of the onsite coulomb interaction @xmath14 . in fig . [ fig3 ] , we show our ddmrg result for the spin excitation spectrum @xmath64 at @xmath50 , @xmath51 , and @xmath87 , where we use the clusters with @xmath60 ( right panel ) and @xmath62 ( left panel ) . roughly speaking , the lower edge of the spectrum consists of three sine curves with four nodes at @xmath88 , @xmath70 @xmath71 , @xmath89 @xmath90 , and @xmath91 . the excitation gap seems to close around these nodes . this is consistent with previous theoretical studies,@xcite which have suggested that no spin gap exists in the strong coupling regime . on the other hand , the higher edge of the spectrum is approximately represented as a sine curve @xmath92 as in the case of @xmath52 , which comes from the creation of the particle - hole pairs within the same branch . let us now take a closer look at the spectrm . as far as the spin degrees of freedom are concerned , the model eq . ( [ hamiltonian ] ) for large @xmath14 may be mapped onto a two - chain @xmath0@xmath13@xmath3 model coupled with zigzag bonds . for small @xmath65 , the antiferromagnetic interaction along the @xmath1-chain , @xmath93 , must be much larger than that along the @xmath0-chain , @xmath3 , if we assume that the exchange interaction comes from the second - order perturbation of the hopping integrals with a fixed @xmath14 , i.e. , @xmath93 @xmath94 @xmath95 @xmath3 @xmath96 . consequently , the features of the spectrum can be basically interpreted as those of the 1d quarter - filled @xmath0@xmath13@xmath3 model.@xcite the nodes at the lower edge of the ddmrg spectrum occur at the momenta @xmath69 , @xmath97 , @xmath98 , and @xmath91 . also , we can see considerable enhancement of spectral intensities around @xmath99 @xmath100 in comparison with the noninteracting spectrum . this indicates that the onsite coulomb interaction enhances the antiferromagnetic correlation with a period of @xmath101 times the lattice constant along the @xmath1-chain , which can be easily expected from the fact that the @xmath97-sdw correlation is the most dominant for small @xmath3 in the 1d @xmath0@xmath13@xmath3 model at quarter filling . this result may be compared with the @xmath102-sdw state observed experimentally in ( tmtsf)@xmath5x.@xcite we then examine the effects of a small hopping integral @xmath0 , which leads to the antiferromagnetic interaction @xmath3 along the @xmath0-chain as mentioned above . from the viewpoint of the spin degrees of freedom , magnetic frustration must be brought because triangular lattices are formed of only the antiferromagnetic interactions . although @xmath3 is much smaller than @xmath93 , we can clearly see the influence in our ddmrg spectrum ; there are two nodes around @xmath103 , i.e. , @xmath104 ( @xmath105 ) and @xmath106 ( @xmath107 ) . these nodes are collected into a single node at @xmath103 when @xmath37 . this split actually signifies a tendency to a formation of the @xmath102-sdw state along the @xmath0-chain as well as to a collapse of the @xmath102-sdw state along the @xmath1-chain . with increasing @xmath65 , the node at @xmath108 approaches @xmath109 and the adjacent spectral weight increases , whereas the node at @xmath110 goes away from @xmath103 and the weight goes to zero . in other wards , the hopping term @xmath0 weakens the @xmath102-sdw oscillation along the @xmath1-chain since the competing antiferromagnetic correlation along the @xmath0-chain is enhanced . in fact , the spectral weight around @xmath99 will certainly diminish as @xmath65 increases . another noticeable feature is that the spectral weight around @xmath111 is obviously smaller than that around @xmath99 , while the spectrum should be symmetrical about @xmath109 in the case of @xmath37 . this result implies the presence of ferromagetic spin fluctuations at low energies . for clearer understanding , we study the spin structure factor @xmath54 . as shown in the inset of fig . [ fig3 ] , it is evident that @xmath54 around @xmath99 is greater than that around @xmath111 ; otherwise , @xmath54 should be almost symmetrical about @xmath112 . this result indicates an enhancement of the ferromagnetic spin correlation between the neighbouring sites along the @xmath0-chain . we may explain this as follows . a real - space behavior of the spin - spin correlation may be derived from the fourier transform of @xmath54 , @xmath113 with @xmath114 . we find that the spin correlation along the @xmath1-chain is not affected so much by a small @xmath0 value but the decay length of the @xmath102-sdw oscillation is slightly shortened . on the contrary , the @xmath0 hopping term plays a prominent role in spin correlation between the @xmath1-chains , which may be estimated as @xmath115\times{\rm ( decaying~term)},\ ] ] where @xmath116 is an odd number . for @xmath117 , we obtain @xmath118 , which indicates the presence of ferromagnetic correlation between two electrons at the neighboring sites . this result is consistent with a scenario of spin - triplet superconductivity where the pairing of electrons occurs between the inter @xmath1-chain nearest - neighbor sites , which is proposed in ref . . finally , we study the charge correlation function in the presence of the onsite coulomb interaction @xmath14 . in fig . [ fig4 ] , we show the charge excitation spectrum @xmath59 at @xmath50 , @xmath51 , and @xmath87 , calculated by the ddmrg method for the chains with @xmath60 ( right panel ) and with @xmath62 ( left panel ) . the outline of the spectrum , which is roughly expressed as @xmath119 , seems to be similar to that in the noninteracting spectrum . this result reflects the fact that the overall dispersion is hardly affected by the onsite coulomb interaction . however , we notice that two distinct features emerge in the low - frequency range of the spectrum , which are discussed below . calculated by the ddmrg method for the same set of the parameter values as in fig . we use the clusters @xmath60 with broadening @xmath61 ( right panel ) and @xmath62 with @xmath63 ( left panel ) . inset : charge structure factor @xmath120 obtained from @xmath57-integration of @xmath59.,width=264 ] one is the increase in the low - energy spectral weight around @xmath109 in the excitation spectrum . the excitation gap seems to close there , so that the lower edge of the spectrum consistes of the two sine curves with three nodes at @xmath88 , @xmath121 , and @xmath91 . we may thus have one gapless charge mode . this result reflects a tendency toward the peierls instability , i.e. , the formation of the @xmath122-charge - density - wave ( cdw ) along the @xmath1-chain . we find that the spectral weight for the @xmath102-cdw correlation is instead much reduced . this feature would be more evident if we observe the momentum - dependent charge structure factor @xmath120 , which is shown in the inset of fig . we find that @xmath120 takes the maximum value at @xmath109 and goes down along the practically straight lines to @xmath69 and @xmath91 . this dependence is consistent with results in the 1d quarter - filled hubbard chain at large @xmath14.@xcite ; in fact , the result is almost equivalent to the result for noninteracting spinless fermions at half filling , as far as the charge degrees of freedom are concerned . the other is the apperance of large - weighted sharp peaks around @xmath69 at @xmath123 . the point is as follows ; the spectral weight of the peaks around @xmath69 is larger than that around @xmath124 in the low - frequency range , and they are also gathered at lower frequencies . this implies that the electrons tend to come in the neighboring sites along the @xmath0-chain , which is associated with the pairing of two electrons between the inter @xmath1-chain sites ; accordingly , the pairs tend to be arranged alternately along the @xmath1-chain . note that the two momenta @xmath69 and @xmath91 should be equivalent when @xmath37 , so that the present result implies that the small hopping integral @xmath0 enhances the pairing fluctuations at low energies . additionally , we can estimate the so - called tomonaga - luttinger liquid parameter @xmath125 from the derivative of @xmath120 at @xmath69 , @xmath126 whereby we find the value @xmath127 . because , in the presence of one gapless charge mode , the pairing correlation is dominant in comparison with the @xmath122-cdw correlation when @xmath128,@xcite our estimated value of @xmath125 is consistent with the occurence of spin - triplet superconductivity proposed in ref . . we have calculated the spin and charge excitation spectra of the two - chain zigzag - bond hubbard model at quarter filling in order to seek for consequences of the spin - triplet pairing induced by the ring - exchange mechanism . the model is topologically equivalent to the 1d hubbard model with nearest - neighbor @xmath0 @xmath129 and next - nearest - neighbor @xmath1 @xmath129 hopping integrals . we here have considered the case at @xmath130 . we have used the ddmrg technique to calculate the excitation spectra . we have first demonstrated the accuracy of the ddmrg method by reproducing the noninteracting exact spectrum . we have suggested that , for practical calculations , it is neccesary to adopt a relatively small system for obtaining an overview of the spectrum and a larger system for investigating detailed structures of the spectrum at low frequencies because the required cpu time increases rapidly with increasing the frequency and/or system size . thus , the ddmrg method enables us to study the details of relatively complicated structures of the spectra . then , we have investigated how the spectra are deformed when the strong onsite coulomb interaction @xmath14 sets in . we find that the spin and charge excitation spectra are basically the same as those of the 1d hubbard ( and @xmath0-@xmath3 ) chain at quarter filling ; i.e. , the spectra come from the two nearly independent 1d chains where the @xmath102-sdw and @xmath122-cdw correlations along the @xmath1-chains are enhanced . however , we find that the hopping integral @xmath0 plays a crucial role in the short - range spin and charge correlations ; i.e. , the ferromagnetic spin correlations between electrons on the neighboring sites is enhanced and the spin - triplet pairing correlations between the electrons is induced , of which the consequences are clearly seen in the calculated spin and charge excitation spectra at low energies . our ddmrg calculations for the spin and charge excitation spectra have thus supported the ring - exchange mechanism for spin - triplet superconductivity where the pairing of electrons occurs between the two chains . we thank t. takimoto for useful discussions . this work was supported in part by grants - in - aid for scientific research ( nos . 18028008 , 18043006 , and 18540338 ) from the ministry of education , science , sports , and culture of japan . a part of computations was carried out at the research center for computational science , okazaki research facilities , and the institute for solid state physics , university of tokyo . 99 a. j. leggett , rev . phys . * 47 * , 331 ( 1975 ) . y. maeno , h. hashimoto , k. yoshida , s. nishizaki , t. fujita , j. g. bednorz , and f. lichtenberg , nature * 372 * , 532 ( 1994 ) . r. joynt and l. taillefer , rev . phys . * 74 * , 235 ( 2002 ) . y. ohta , s. nishimoto , t. shirakara , and y. yamaguchi , , 012503 ( 2005 ) . p. fazekas , _ lecture notes on electron correlation and magnetism _ ( world scientific , singapore , 1999 ) , p. 435 . d. jerme , a. mazaud , m. ribault , and k. bechgaard , j. phys . ( france ) lett . * 41 * , l95 ( 1980 ) . k. bechgaard , k. carneiro , m. olsen , f. b. rasmussen , and c. s. jacobsen , , 852 ( 1981 ) . i. j. lee , s. e. brown , and j. naughton , j. phys . soc . jpn . * 75 * , 051011 ( 2006 ) and references therein . m. matsukawa , y. yamada , m. chiba , h. ogasawara , t. shibata , a. matsushita , and y. takano , physica c * 411 * , 101 ( 2004 ) . s. sasaki , s. watanabe , y. yamada , f. ishikawa , k. fukuda , s. sekiya , cond - mat/0603067 . k. sano , y. no , and y. yamada , j. phys . soc . jpn . * 74 * , 2885 ( 2005 ) . e. jeckelmann , , 045114 ( 2002 ) . s. r. white , , 2863 ( 1992 ) ; , 10345 ( 1993 ) . h. benthien , f. gebhard , and e. jeckelmann , , 256401 ( 2004 ) . s. daul and r. m. noack , , 2635 ( 1998 ) . for a review article , see t. ishiguro , k. yamaji , and g. saito , _ organic superconductors _ ( springer - verlag , berlin , 1990 ) . p. a. bares and g. blatter , , 2567 ( 1990 ) . t tohyama , p. horsch and s. maekawa , , 980 ( 1995 ) . y. saiga and y. kuramoto , j. phys . . jpn . * 68 * , 3631 ( 1999 ) . j. e. hirsch and d. j. scalapino , , 7169 ( 1983 ) . n. nagaosa , solid state commun . * 94 * , 495 ( 1995 ) . h. j. schulz , , r2959 ( 1996 ) ; in _ correlated fermions and transport in mesoscopic systems _ , edited by t. martin , g. montambaux , and t. trn thanh vn ( editions frontires , gif - sur - yvette , france , 1996 ) , p. 81

the dynamical density - matrix renormalization group technique is used to calculate spin and charge excitation spectra in the one - dimensional ( 1d ) hubbard model at quarter filling with nearest - neighbor @xmath0 and next - nearest - neighbor @xmath1 hopping integrals . we consider a case where @xmath0 ( @xmath2 ) is much smaller than @xmath1 ( @xmath2 ) . we find that the spin and charge excitation spectra come from the two nearly independent @xmath1-chains and are basically the same as those of the 1d hubbard ( and @xmath0-@xmath3 ) chain at quarter filling . however , we find that the hopping integral @xmath0 plays a crucial role in the short - range spin and charge correlations ; i.e. , the ferromagnetic spin correlations between electrons on the neighboring sites is enhanced and simultaneously the spin - triplet pairing correlations is induced , of which the consequences are clearly seen in the calculated spin and charge excitation spectra at low energies .

recently many three - body @xmath0 decay modes , for example , @xmath14 , @xmath15 , @xmath16 , @xmath17 , @xmath18 , @xmath19 and @xmath20 , have been observed at @xmath0 factories with branching ratios of order @xmath21 . the dalitz plot analysis of 3-body @xmath0 decays provides a nice methodology for extracting information on the unitarity triangle in the standard model . the three - body meson decays are generally dominated by intermediate vector and scalar resonances , namely , they proceed via quasi - two - body decays containing a resonance state and a pseudoscalar meson . indeed , most of the quasi - two @xmath0 decays are extracted from the analysis of three - body @xmath0 decays using the dalitz plot technique . three - body hadronic @xmath0 decays involving a vector meson or charmed meson also have been observed at @xmath0 factories . in this talk i ll focus on charmless 3-body @xmath0 decays . experimentally , there are two striking features for 3-body hadronic @xmath0 decays : @lcc@ decay & babar & belle + @xmath22 & @xmath23 & @xmath24 + @xmath25 & @xmath26 & + @xmath27 & & @xmath28 + @xmath6 & @xmath29 & @xmath30 + @xmath31 & @xmath32 & @xmath33 + @xmath34 & @xmath35 & + [ tab : brexpt ] 0.2 cm ( i ) large noresonant fractions in peguin - dominated modes 0.2 cm it is known that the nonresonant signal in charm decays is small , less than 10% @xcite . in the past few years , some of the charmless @xmath0 to 3-body decay modes have been measured at @xmath0 factories and studied using the dalitz plot analysis . we see from table [ tab : brexpt ] that the nonresonant fraction is about @xmath36 90% in @xmath2 decays , @xmath37 in @xmath38 decays ( smaller in the @xmath39 decay ) , and @xmath36 14% in the @xmath40 decay . hence , the nonresonant 3-body decays play an essential role in penguin - dominated @xmath0 decays . while this is a surprise in view of the rather small nonresonant contributions in 3-body charm decays , it is not entirely unexpected because the energy release scale in weak @xmath0 decays is of order 5 gev , whereas the major resonances lie in the energy region of 0.77 to 1.6 gev . consequently , it is likely that 3-body @xmath0 decays may receive sizable nonresonant contributions . it is important to understand and identify the underlying mechanism for nonresonant decays . nonresonant amplitudes in charm decays are usually assumed to be uniform in phase space . however , this is no longer true in @xmath0 decays due to the large energy release in weak @xmath0 decays . while both babar and belle have adopted the parametrization @xmath41 to describe the non - resonant @xmath2 amplitudes , they differ in the analysis of the nonresonant component in @xmath38 decays . belle still employed the above exponential parametrization to analyze the nonresonant contribution , but babar used the lass parametrization to describe the @xmath42 @xmath43-wave and the nonresonant component by a single amplitude suggested by the lass collaboration to describe the scalar amplitude in elastic @xmath42 scattering . since the babar and belle definitions of the @xmath44 and nonresonant differ , the branching fractions and phases are not directly comparable . we will come this point more in sec . experimentally , it is hard to measure the direct 3-body decays as the interference between nonresonant and quasi - two - body amplitudes makes it difficult to disentangle these two distinct contributions and extract the nonresonant one . 0.2 cm ( ii ) new broad scalar resonances @xmath45 and @xmath46 0.2 cm a broad scalar resonance @xmath47 ( or @xmath48 denoted by babar ) has been seen in @xmath49 , @xmath50 and @xmath51 decays at energies around 1.5 gev . however , it can not be identified with the well known scaler meson @xmath52 . this is because @xmath52 decays into @xmath53 about five times more frequently than to @xmath54 . identification of @xmath47 with @xmath52 will imply that the @xmath54 peak at 1.5 gev will be accompanied by a peak in @xmath53 , which is not seen experimentally . hence , the nature of @xmath47 is not clear . moreover , there exists a production puzzle for @xmath47 . both babar and belle have seen a large fraction from @xmath47 in the decay @xmath22 : @xmath55 by babar @xcite and @xmath56 by belle @xcite , whereas it is only about 4% seen by babar in @xmath57 @xcite . the puzzle is that why @xmath47 behaves so dramatically different in charged and neutral @xmath0 decays to 3 kaons . it is not clear whether the large production of @xmath47 is a genuine effect or just a statistical fluctuation . anyway , this issue should be clarified soon . notice that belle actually found two solutions for the fraction of @xmath58 in @xmath59 @xcite : @xmath56 and @xmath60 . the first solution is preferred by belle . it is probably worth of re - examining the small solution . in analog to two - body decays of heavy mesons which can be analyzed using the model - independent quark diagrammatic approach , three - body decays of the heavy mesons can be expressed in terms of some quark - graph amplitudes @xcite ( see fig . [ qd ] ) : @xmath61 and @xmath62 , the color - allowed external @xmath63-emission tree diagrams ; @xmath64 and @xmath65 , the color - suppressed internal @xmath63-emission diagrams ; @xmath66 , the @xmath63-exchange diagram ; @xmath67 , the @xmath63-annihilation diagram ; @xmath68 and @xmath69 , the penguin diagrams , and @xmath70 , the penguin - induced annihilation diagram . the quark - graph amplitudes of various 3-body @xmath0 decays @xmath71 and @xmath72 are summarized in table i of @xcite . as mentioned in @xcite , the use of the quark - diagram amplitudes for three - body decays are in general momentum dependent . this means that unless its momentum dependence is known , the quark - diagram amplitudes of direct 3-body decays can not be extracted from experiment without making further assumptions . moreover , the momentum dependence of each quark - diagram amplitude varies from channel to channel . denotes a heavy quark.,width=215 ] we take the decay @xmath25 as an illustration . under the factorization approach , its decay amplitude consists of three distinct factorizable terms : ( i ) the transition process induced by @xmath73 penguins , @xmath74 , ( ii ) the current - induced process through the tree @xmath75 transition , @xmath76 , and ( iii ) the annihilation process @xmath77 , where @xmath78 denotes a @xmath79 transition matrix element . for the transition process , the general expression of the nonresonant contribution has the form @xmath80 , \end{aligned}\ ] ] where @xmath81 . in principle , one can apply heavy meson chiral perturbation theory ( hmchpt ) to evaluate the form factors @xmath82 and @xmath83 ( for previous studies , see @xcite ) . however , this will lead to too large decay rates in disagreement with experiment @xcite . a direct calculation indicates that the branching ratio of @xmath25 arising from the transition process alone is already at the level of @xmath84 which exceeds the measured total branching ratio @xcite of @xmath85 . the issue has to do with the applicability of hmchpt . in order to apply this approach , two of the final - state pseudoscalars ( @xmath86 and @xmath87 in this example ) have to be soft . the momentum of the soft pseudoscalar should be smaller than the chiral symmetry breaking scale @xmath88 of order @xmath89 gev . for 3-body charmless @xmath0 decays , the available phase space where chiral perturbation theory is applicable is only a small fraction of the whole dalitz plot . therefore , it is not justified to apply chiral and heavy quark symmetries to a certain kinematic region and then generalize it to the region beyond its validity . if the soft meson result is assumed to be the same in the whole dalitz plot , the decay rate will be greatly overestimated . recently we have proposed to parametrize the @xmath75 trasnition - induced nonresonant amplitude given by eq . ( [ eq : ahmchpt ] ) as @xcite @xmath90 so that the hmchpt results are recovered in the chiral limit @xmath91 . that is , the nonresonant amplitude in the soft meson region is described by hmchpt , but its energy dependence beyond the chiral limit is governed by the exponential term @xmath92 . the unknown parameter @xmath93 can be determined from the data of the tree - dominated decay @xmath34 . in addition to the @xmath75 tree transition , we need to consider the nonresonant contributions to the @xmath73 penguin amplitude @xmath94 the 2-kaon creation matrix elements can be expressed in terms of time - like kaon current form factors as @xmath95 the weak vector form factors @xmath96 and @xmath97 can be related to the kaon electromagnetic ( e.m . ) form factors @xmath98 and @xmath99 for the charged and neutral kaons , respectively . phenomenologically , the e.m . form factors receive resonant and nonresonant contributions @xmath100 the resonant and nonresonant terms in eq . ( [ eq : kkemff ] ) can be determined from a fit to the kaon e.m . the non - resonant contribution to the matrix element @xmath101 is given by @xmath102 the nonresonant @xmath103 term is introduced for the following reason . although the nonresonant contributions to @xmath104 and @xmath105 are related through the equation of motion , the resonant ones are different and not related _ a priori_. as stressed in @xcite , to apply the equation of motion , the form factors should be away from the resonant region . in the presence of the resonances , we thus need to introduce a nonresonant @xmath103 term which can be constrained by the measured @xmath106 rate and the @xmath54 mass spectrum @xcite . vector meson and scalar resonances contribute to the two - body matrix elements @xmath107 and @xmath108 , respectively . they can also contribute to the three - body matrix element @xmath109 . resonant effects are described in terms of the usual breit - wigner formalism . more precisely , @xmath110 where @xmath111 and @xmath112 . in this manner we are able to figure out the relevant resonances which contribute to the 3-body decays of interest and compute the rates of @xmath7 and @xmath8 . [ cols= " < , < , < , < " , ] [ tab : cp ] the deviation of the mixing - induced @xmath9 asymmetry in @xmath113 , @xmath114 , @xmath115 and @xmath13 from that measured in @xmath116 , i.e. @xmath117 @xcite , namely , @xmath118 , is shown in table [ tab : cp ] . our calculation indicates the deviation of the mixing - induced @xmath9 asymmetry in @xmath119 from that measured in @xmath120 is very similar to that of the @xmath114 mode as the tree pollution effect in the former is somewhat washed out . nevertheless , direct @xmath9 asymmetry of the former , being of order @xmath121 , is more prominent than the latter . it is important to account for the large nonresonant amplitudes in the study of charmless 3-body baryonic @xmath0 decays . we have identified two distinct sources of nonresonant contributions : a small contribution from the tree transition and a large source of the nonresonant signal in the matrix elements of scalar densities , e.g. @xmath1 , induced from the penguin transition . this explains the dominance of the nonresonant background in @xmath2 decays , the sizable nonresonant fraction in @xmath3 and @xmath4 modes and the smallness of nonresonant rates in @xmath5 decays . the seemingly huge discrepancy between babar and belle for the nonresonant contribution in the decay @xmath6 is now relieved . since penguin contributions to charm decays are gim suppressed , hence nonresonant signals in @xmath122 decays are always small . we have computed the resonant and nonresonant contributions to charmless 3-body decays and determined the rates for the quasi - two - body decays @xmath7 and @xmath8 . time - dependent @xmath9 asymmetries @xmath10 and @xmath11 in @xmath12 and @xmath13 modes are estimated . since we have a realistic model for resonant and nonresonant contributions , our estimation of @xmath10 for 3-body @xmath0 decays should be more reliable and trustworthy . i m grateful to chun - khiang chua and amarjit soni for fruitful collaboration , to tim gershon and jim smith for discussions and to paoti chang and hsiang - nan li for organizing this stimulating conference .

charmless 3-body decays of @xmath0 mesons are studied in a simple model based on the factorization approach . we have identified two distinct sources of nonresonant contributions : a small contribution from the tree transition and a large source of the nonresonant signal in the matrix elements of scalar densities , e.g. , @xmath1 , induced from the penguin transition . this explains the dominance of the nonresonant background in @xmath2 decays , the sizable nonresonant fraction in @xmath3 and @xmath4 modes and the smallness of nonresonant rates in @xmath5 decays . the seemingly huge discrepancy between babar and belle for the nonresonant contribution in the decay @xmath6 is now relieved . we have computed the resonant and nonresonant contributions to charmless 3-body decays and determined the rates for the quasi - two - body decays @xmath7 and @xmath8 . time - dependent @xmath9 asymmetries @xmath10 and @xmath11 in @xmath12 and @xmath13 modes are estimated .

a neutron - star low - mass x - ray binary ( ns - lmxb , or hereafter simply lmxb ) consists of a mass accreting neutron star ( ns ) and a mass - donating low - mass star which fills the binary roche lobe . it is well known that these objects show bimodality in x - ray spectral states , namely soft and hard states ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in the soft state when the mass accretion rate is high , mass from the companion spirals onto the ns through a standard accretion disk @xcite which is optically thick and geometrically thin . the spectrum is soft , which we can describe by the canonical model of @xcite : a disk blackbody with a temperature of @xmath51 kev , plus a blackbody from the ns surface with a temperature of @xmath5 2 kev . the latter component is often found to be weakly comptonized ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? over the past three decades , this model has been reinforced by a number of studies including , e.g. , @xcite , @xcite , @xcite , @xcite and @xcite . when the accretion rate is low and the source is in the hard state , the inner accretion disk transites into an optically thin and geometrically thick flow around the ns , which is further characterized by a large radial velocity component , a nearly free - fall ion temperature , and possibly a significant difference between the ion and electron temperatures . the hot electrons therein strongly comptonize the ns blackbody emission and harden the spectrum to @xmath7 100 kev @xcite . in contrast to these normal " objects , a special kind of lmxbs with high inclination angles , namely dipping lmxbs ( also called dippers ) showing periodic dips in their x - ray intensity , seldom have their spectral states investigated . for a long time , their dips were thought to be caused by progressive covering of a largely extended scattering region called accretion disk corona ( adc , * ? ? ? * ) by the companion star ( e.g. , * ? ? ? * ; * ? ? ? * ) . however , new studies revealed that the dips can be explained simply by obscuration of the central compact x - ray source by an ionized and extended absorber located at some outer parts of the accretion disk @xcite , and the adc is no longer required . based on the new scenario , it is reasonable to consider dipping lmxbs essentially the same objects as non - dipping lmxbs , except for their higher inclinations . hence it is worthwhile to examine whether we can understand dippers in the same spectral classification scheme as for non - dipping lmxbs , possibly with some modifications . so far , broad - band spectra of dipping lmxbs were studied extensively , e.g. , with bepposax . examples include relatively luminous dippers 4u 1915 - 05 @xcite , xb 1254 - 690 @xcite and xb 1658 - 298 @xcite , as well as dimmer ones such as xb 1323 - 619 @xcite and exo 0748 - 676 @xcite . the reported spectra are generally rather hard , often extending to @xmath5 100 kev , which are reminiscent of the hard state of non - dipping lmxbs . however , these previous works did not clearly specify whether the objects were in the hard state or not . since dippers , like the other lmxbs , must be distributed over the two states , we presume that some of them actually reside in the soft state , and their broad - band spectra appear harder because , e.g. , their higher inclination leads to a stronger comptonization . to examine this possibility , we need to analyze broad - band non - dipping spectra of some appropriate dippers in a unified way which is consistent with the study of non - dipping objects . the available publications are clearly inadequate for this purpose , because different authors employed different modelings . for the above purpose , we need to use an x - ray instrument with a high spectral resolution and broad - band coverage . the most suited one is the japanese satellite suzaku @xcite , which has a soft x - ray energy resolution of @xmath8 ev realized by the x - ray imaging spectrometer ( xis ; * ? ? ? * ) , and an energy coverage up to @xmath9 100 kev achieved by the hard x - ray detector ( hxd ; * ? ? ? so far , five dipping lmxbs have been observed in total with suzaku , and among them , we selected the most luminous one 4u 1915 - 05 ( see section [ sec : observation ] for details ) . this source is the first dipper discovered in 1982 @xcite , showing stable periodic dips with a period of @xmath10 mins @xcite , and frequent type i x - ray bursts @xcite . it is also known as an ultracompact x - ray binary ( ucxb ) which has a helium rich donor star @xcite . its 0.6 - 10 kev persistent luminosity , was @xmath11 erg s@xmath1 @xcite at an assumed distance of 9.3 kpc ( see below ) , reached the range typical of lmxbs in the soft state . the broadband spectrum of this source was studied by @xcite using bepposax . by fitting a 0.5 - 200 kev spectrum with a cut - off power - law model , these authors reported a rather high cut - off energy of @xmath5 80 kev , which is suggestive of strong comptonization . however , further exploration with other x - ray satellites ( e.g. , rosat , chandra , xmm - newton ) did not give much improved knowledge of this issue due to their limited energy coverage @xcite . furthermore , few of these studies attempted spectral classification of this object in comparison with non - dipping lmxbs . in the present paper , we utilized an archival suzaku data set of 4u 1915 - 05 acquired on 2006 november 8 . we focus on its non - dip broadband spectral properties , while the x - ray dips are out of the scope of our study . throughout the paper , the source distance is assumed to be 9.3 kpc , as determined by @xcite with an uncertainty of @xmath12% , using the x - ray bursts as a standard candle . using the same method , other studies gave similar values of 8.4 - 10.8 kpc @xcite and 7 - 9 kpc @xcite ; thus , the distance uncertainty is typically 20% . since the source shows x - ray dips , but no x - ray eclipse , its inclination angle is estimated to be @xmath13 @xcite , with a @xmath14 uncertainty @xcite . table [ tab : fourdippers ] summarizes the suzaku observations of four dipping lmxbs among the five sources in the archive , excluding 4u 1822 - 37 which has been found to have strong magnetic field ( @xmath5 12 g ; * ? ? ? we estimated their 1 - 50 kev luminosities , by fitting their background subtracted spectra with an absorbed cut - off power - law model , and then removing the absorption . the bolometric corrections , which depend on the cutoff energy , are typically smaller than 1.3 . thus , among the four dippers observed with suzaku , 4u 1915 - 05 is the most luminous , and hence is suited to our purpose of identifying a dipping lmxb in the soft state . @xmath15 the 0.8 - 10 kev xis spectrum and 15 - 40 kev hxd - pin spectrum were fitted simultaneously . the quoted errors refer to 90% confidence limits . + @xmath16 another edge was found around 3.4 kev , which is probably an ionized argon k - edge . + @xmath17 @xmath18 was corrected for the inclination factor @xmath19 , in which @xmath20 was assumed . + @xmath21 definition of @xmath22 , @xmath23 and @xmath24 are the same as in table [ tab : totalspectra ] , however the unit is @xmath25 erg @xmath26 s@xmath1 .

the dipping low - mass x - ray binary 4u 1915 - 05 was observed by suzaku on 2007 november 8 for a net exposure of 39 ksec . it was detected by the xis with a 0.8 - 10 kev signal rate of @xmath0 cts s@xmath1 per camera , and hxd - pin with a 12 - 45 kev signal rate of @xmath2 cts s@xmath1 . after removing the periodic dips and an x - ray burst , the 0.8 45 kev continuum was successfully described by an optically thick disk emission with an inner - disk temperature @xmath3 kev and a neutron - star blackbody emission with a temperature @xmath4 kev , on condition that the blackbody component , or possibly the disk emission too , is significantly comptonized . this successful modeling is consistent with 4u 1915 - 05 being in a high - soft state in this observation , and implies that its broadband spectrum can be interpreted in the same scheme as for many non - dipping low - mass x - ray binaries in the soft state . its bolometric luminosity ( @xmath5 0.02 times the eddington limit ) is relatively low for the soft state , but within a tolerance , if considering the distance and inclination uncertainties . as a high - inclination binary , this source exhibited stronger comptonization effect , with a larger comptonizing @xmath6-parameter , compared to low and medium inclination binaries . this suggests that the comptonizing coronae of these objects in the soft state is in an oblate ( rather than spherical ) shape , extending along the accretion disk plane , because the @xmath6-parameter would not depend on the inclination if the corona were spherical .

quantum - enhanced optical phase estimation through the mach - zehnder interferometer ( mzi ) is important for multiple areas of scientific research @xcite , such as imaging , sensing , and high - precision gravitational waves detection . the mzi - based optical phase estimation consists of three steps ( see e.g. fig . [ fig1](a ) ) . first , a two - mode input state of the light is prepared . second , the light passes successively through a beam splitter , the unknown relative phase shift @xmath6 between the two arms of the mzi , and another beam splitter , and evolves to the output state . third , the output state is measured for many times and the outcomes @xmath7 is processed to construct an unbiased estimator @xmath8 to the unknown parameter @xmath6 @xcite . the estimation precision is quantified by the standard deviation @xmath9 . by using optimal data processing techniques to extract all the information contained in the data , the estimation precision from @xmath10 repeated measurements is given by the cramr - rao lower bound @xcite : @xmath11 , where @xmath12 is the classical fisher information ( cfi ) for the measurement scheme used . given the input state , maximizing @xmath12 over all possible measurement schemes gives the quantum fisher information ( qfi ) @xmath13 and hence the quantum cramr - rao bound @xmath14 @xcite , which sets an ultimate precision for estimating the unknown phase shift @xmath6 . usually , the precision @xmath15 improves with increasing number of photons @xmath3 contained in the input state . using a coherent - state of light as the input , the achievable phase sensitivity per measurement is limited by the classical ( or shot noise ) limit @xmath16 , as the qfi @xmath17 . to improve the precision beyond the classical limit ( @xmath18 ) , it is necessary to employ quantum resources , such as entanglement and squeezing in the input state caves , giovannetti , ma , aasi , dowling15,matthews , luca . in this context , the squeezed states of light play an important role and have been widely studied in the past decades ever since the pioneer work of caves in 1981 @xcite , who shows that by feeding a coherent state @xmath19 into one port of the mzi and a squeezed vacuum @xmath20 into the other port , the unknown phase shift can be estimated with a precision beyond the classical limit . in 2008 , pezz and smerzi @xcite further suggested that the previously used phase estimator based on the averaged relative photon number is not optimal . when the injected fields are phase matched , i.e. , the phases of two light fields @xmath21 and @xmath22 obeying @xmath23 , the qfi can reach the heisenberg scaling @xmath24 for a given mean photon number @xmath25 . more importantly , this qfi can be saturated by the cfi for ideal photon counting measurements . consequently , by using the optimal data processing technique ( such as the maximum - likelihood estimation or bayesian estimation ) to process these measurement outcomes , the phase estimation precision can attain the heisenberg limit @xmath26 . recently , lang and caves @xcite proved that given the total average photon number @xmath3 of the input state , if a coherent light is fed into one input port of the mzi , then the squeezed vacuum is the optimal state to inject into the second input port . et al_. @xcite have analyzed the phase - matching condition ( pmc ) that maximizes the qfi in the squeezed - state interferometer , where a superposition of even or odd number of photons is injected from one port of the interferometer and any input state from another . an important requirement of these theoretical works @xcite is to take into account all the photon - counting events , which in turn requires photon - number - resolving detectors with perfect number resolution seshadreesan . however , on the experimental side , the best detector up to date can only resolve the number of photons up to @xmath27 smerzi07,kardynal . this makes it unclear whether or not the heisenberg limit of the estimation precision can still be achieved by using realistic photon detectors with an upper threshold on the number resolution . to bridge this gap between the theory and experiments , it is of interest to investigate the experimentally achievable estimation precision when the total number of photons being detected is limited , i.e. , @xmath28 , where @xmath29 determines the number resolution by a single photon - counting detector . since the existence of an upper threshold @xmath30 essentially amounts to discarding the information contained in photon - counting events with the number of photons larger than @xmath30 , it is therefore important to investigate the distribution of the qfi and cfi in the @xmath0-photon components of the coherent @xmath31 squeezed vacuum input state and calculate how much the qfi is kept with a finite number resolution . in addition , studying the distribution of the qfi and cfi in the @xmath0-photon components also helps to understand the phase estimation precision in recent post - selection experiments . when the mzi is fed by the coherent @xmath31 squeezed vacuum , the state after the first beam splitter of the mzi contains a small fraction of the path - entangled noon state @xcite , which is a well - known @xmath0-photon non - classical state that allows the phase estimation precision to achieve the heisenberg limit kok , pryde , boto , mitchell , walther , cable , toppel . in the limit latexmath:[$% @xmath2-fold oscillations of the coincidence rates for @xmath0 up to @xmath33 , manifesting the appearance of @xmath0-photon noon states . however , the generation probability of a @xmath0-photon noon state decreases dramatically with increasing @xmath0 , e.g. , the @xmath33-photon count rate @xmath34 per @xmath35 second @xcite . therefore , it is desirable to study the overall estimation precision when such small generation probabilities are included , since there are general conclusions that the generated state under postselection can not improve the precision for estimating a single parameter when the _ total _ number of input photons are included ( see e.g. , refs . combes , pang , haine ) . in this paper , we investigate the distribution of the qfi and cfi in the different @xmath0-photon components of the coherent @xmath1 squeezed vacuum input state and provide the achievable estimation precision by using imperfect photon counters with an upper threshold @xmath30 for the photon number resolution . under the pmc @xmath36 , we show that the cfi always saturates the qfi for each _ individual _ @xmath0-photon component . consequently , when the detectable number of photons is upper bounded by @xmath30 , the phase estimation precision @xmath37 is always equal to @xmath38 and both of them are determined by the sum of the cfi or equivalently the qfi for each @xmath0-photon component with @xmath0 up to @xmath30 . for the commonly used optimal input state with @xmath39 ) gives the cfi or the qfi @xmath40 , leading to the heisenberg limit of the estimation precision smerzi08,lang , liu . for finite photon number resolution , we provide an approximate analytical expression that quantifies the amount of lost information , which predicts that over @xmath4 percent of the ideal qfi can be achieved as long as @xmath41 . compared with the ideal case ( i.e. , @xmath42 finite @xmath0-photon state under postselection ------------------------------------------------------- as illustrated schematically by fig . [ fig1](a ) , a post - selection scheme for creating path - entangled noon states has been proposed by injecting a coherent state of light and a squeezed vacuum into mach - zehnder interferometer @xcite . this scheme has been demonstrated by afek _ et al . _ @xcite in the limit @xmath43 . however , the generated @xmath0-photon state in postselection can not improve the precision for estimating an unknown phase shift , since the cfi is weighted by the generation probability @xcite . it is therefore important to investigate whether or not a sum of each @xmath0-component for @xmath0 up to a finite number can beat the shot - noise scaling @xmath44 . to answer this question , in this section , we first derive explicit form of the @xmath0-photon state generated by postselection next , we calculate ( quantum ) fisher information of the @xmath0-photon state , which determines the ultimate precision on the phase estimation . and a squeezed vacuum @xmath20 , and the @xmath0-photon state @xmath45 , post - selected by the number of photons being detected @xmath46 . ( b ) for a given @xmath47 , the probability distribution @xmath48 against @xmath49 $ ] and @xmath50 . at a certain value of the ratio @xmath51 ( indicated by the blue line ) , the distribution shows two symmetric peaks at @xmath52 , indicating the appearance of a path - entangled noon state . ( c ) the fidelity between the @xmath0-photon state and an ideal noon state , and ( d ) the qfi of @xmath0-photon state , with their values calculated at @xmath53 ( blue solid line ) and @xmath54 ( red line with crosses ) , see text . the inset in ( c ) indicates @xmath55 . ] without any loss and additional reference beams , the input state can be expressed as a superposition of @xmath0-photon states @xcite , i.e. , @xmath56 , where @xmath57 denotes the generation probability of a finite @xmath2-photon state , and @xmath58 is the number of photons post - selected by the photon counting events @xmath59 . in fock basis , the @xmath0-photon state is given by @xmath60}c_{n-2k}(\theta _ { a})s_{2k}(\theta _ { b})|n-2k,2k\rangle _ { a , b } , \label{psi_n}\]]where @xmath61 , and the sum over @xmath62 is up to @xmath63=(n-1)/2 $ ] ( for odd @xmath0 ) , or @xmath64 ( for even @xmath2 ) , because of even number of photons that injected from the port @xmath65 . note that the probability amplitudes of the coherent state and the squeezed vacuum @xmath66 and @xmath67 depend explicitly on the phases of two input light fields @xmath68 and @xmath69 ( see the append . furthermore , the generation probability @xmath57 is also the normalization factor and is given by @xmath70}\left\vert c_{n-2k}s_{2k}\right\vert ^{2}=\frac{% e^{-\left\vert \alpha \right\vert ^{2}}}{\cosh |\xi |}\left ( \frac{\tanh ratio @xmath71 , and a polynomial @xmath72}\frac{(2k)!}{\left ( n-2k\right ) ! ( k!)^{2}}% ( 2x)^{n-2k}. \label{rn}\]]which obeys @xmath73^{2}$ ] for even @xmath0 , and @xmath74 for odd @xmath0 , similar to the hermite polynomials at @xmath75 . in the limit latexmath:[$|\alpha latexmath:[$x\sim |\alpha the coherent state divided by that of the squeezed vacuum @xcite . explicit form of the @xmath0-photon state crucially depends on the relative phase difference between the squeezing parameter @xmath77 and the coherent - state amplitude @xmath78 . following refs . @xcite , we consider the pmc , i.e. , @xmath36 , for which eq . ( [ psi_n ] ) can be reexpressed as @xmath79 , where @xmath80 denotes the @xmath0-photon states with real amplitudes ( for details , please see the append . after the first beam splitter , the @xmath0-photon state becomes @xmath81 , i.e. , @xmath82 and @xmath83 $ ] . under the pmc , the probability amplitudes of @xmath84 can be written as @xmath85which depends solely on the phase of the coherent - state light @xmath68 , and the probability distribution ( see the append . a ) @xmath86}d_{\mu , j-2k}^{j}\left ( \frac{\pi } { 2}% \right ) \frac{\sqrt{(2k)!}}{k!\sqrt{(n-2k)!}}\left ( 2x\right ) ^{n/2-k}\right ) ^{2 } , \label{pmu}\end{aligned}\ ] ] where @xmath87 are the elements of wigner s d - matrix tajima , feng . it is interesting to note that for a given @xmath0 , the probability distribution depends only on the introduced ratio @xmath88 . figure [ fig1](b ) shows the probability distribution as a function of @xmath89 for a large enough @xmath0 . at @xmath75 , i.e. , a pure squeezed vacuum being injected , the probability distribution is almost a gaussian , due to @xmath90^{2}\varpropto \exp ( -\mu ^{2}/j)$ ] . as @xmath91 increases , the @xmath0-photon state always shows symmetric probability distribution ( i.e. , @xmath92 ) . one can see this directly from eq . ( [ pmu ] ) , where @xmath93 ; see e.g. , refs . physically , the symmetric probability distribution arises from the fact that the @xmath0-photon state @xmath94 contains only even number of photons in mode @xmath65 , i.e. , @xmath95 , which in turn leads to @xmath96and hence @xmath92 . this symmetry enables us to write down explicit expression of the @xmath0-photon state @xmath97which is indeed a superposition of the path - entangled states @xmath98 , where the relative phase @xmath99 comes from eq . ( [ ampbs ] ) . for a certain value of @xmath91 , the probability distribution @xmath100 reaches its maximum at @xmath101 , indicating @xmath102 , with the fidelity given by @xmath103clearly , the fidelity depends on the ratio @xmath91 and the number of photons being detected @xmath0 ( @xmath104 ) . for a given @xmath0 , maximizing the fidelity with respect to @xmath91 , one can obtain the optimal value of the ratio , denoted hereinafter as @xmath105 . for small @xmath0 s , it has been obtained @xmath106 ( for @xmath107 , @xmath108 ) , @xmath109 ( @xmath110 ) , and @xmath111 ( @xmath112 ) ; see ref . when @xmath113 , the optimal value of @xmath91 is about @xmath64 , for which @xmath114 ( see ref . hofmann , and also table [ optimal ] ) . [ fig1](c ) , we show the optimal value of the fidelity @xmath115 as a function of @xmath0 ( the blue solid line ) , which coincides with afek _ et al _ @xcite . from eqs . ( psibs ) and ( [ fidelity ] ) , one can also see that before the first beam splitter , @xmath116 itself at @xmath117 approaches the noon state @xmath118 , which shows the polarization along @xmath119 . we now investigate the cfi of the @xmath0-photon state in the photon - counting measurements and show that it always equals to the qfi under the pmc . to this end , we first calculate the qfi of the phase - encoded state @xmath120 , where the unitary operator represents sequent actions of the first beam splitter , the phase - shift accumulation in the path , and the second 50:50 beam splitter at the output ports , as illustrated in fig . [ fig1](a ) . due to @xmath121 , it is easy to obtain the qfi braunstein , braunstein96,luo , smerzi09,donner : @xmath122where @xmath123 denotes the @xmath0-photon state after the first beam splitter and its probability distribution @xmath100 has been given by eq . ( [ pmu ] ) . similar to the fidelity , one can see that the qfi depends on the ratio @xmath91 and the number of photons @xmath0 . for the cases @xmath124 , @xmath108 , and @xmath27 , both of them reach maximum at @xmath125 . this is because of the relation : @xmath126where @xmath127 , @xmath128 , and @xmath129 are vanishing at @xmath130 . when @xmath0 @xmath131 , however , @xmath132 with @xmath133 reaches its maximum at @xmath134 , which is slightly smaller than @xmath125 ( see table i ) . in fig fig1(d ) , we plot maximum of the qfi as a function of @xmath0 and find @xmath135 @xmath136 . next , we consider the photon counting measurements over the phase - encoded state @xmath137 and calculate the cfi . again , we consider the pmc and rewrite the @xmath0-photon state as @xmath138 , where @xmath80 is given by eq . ( [ psi_n ] ) with @xmath139 . note that the probability amplitudes of @xmath140 are real , which result in the conditional probabilities ( see the append . a ) : @xmath141 ^{2 } , \label{pnmu}\]]where @xmath142 $ ] and @xmath143 . obviously , for a given @xmath0 , there are @xmath144 outcomes with their probabilities satisfying the normalization condition @xmath145 , we further obtain @xmath146 indicating that @xmath147 is purely imaginary for each @xmath148 . this is the key point to obtain the cfi : @xmath149 ^{2}}{p_{n}(\mu |\varphi ) } = -4\sum_{\mu = -j}^{+j}% \left [ \langle j,\mu |j_{y}e^{-i\varphi j_{y}}|\tilde{\psi}_{n}\rangle % \right ] ^{2}=4\langle \tilde{\psi}_{n}|j_{y}^{2}|\tilde{\psi}_{n}\rangle = f_{q , n } , \label{fqn2}\]]where @xmath135 is the qfi of the phase - encoded state @xmath150 under the pmc , given by eq . ( [ fqn ] ) . .for a given @xmath0 , the fidelity @xmath151 and the qfi @xmath135 depend solely on the ratio @xmath152 , and reach maximum at @xmath153 and @xmath134 , respectively . for @xmath107 , @xmath108 , @xmath154 at @xmath155 ; while for @xmath110 , @xmath156 at @xmath157 . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] as one of main results of this work , eq . ( [ fqn2 ] ) indicates that as the input " state , @xmath158 at @xmath159 could provide a global phase estimation at the heisenberg scaling @xcite , as @xmath160 . however , this scaling is defined with respect to the number of photons being detected @xmath0 . furthermore , @xmath158 is post - selected by the @xmath0-photon detection events with the generation probability @xmath57 , which is usually very small as @xmath113 ( see fig . [ fig2](a ) ) . indeed , purely with the @xmath0-photon detection events ( i.e. , totally @xmath144 outcomes with a definite @xmath0 ) , one can not improve the accuracy for estimating an unknown phase shift , since the cfi is weighted by the generation probability @xcite , i.e. , @xmath161 . for the input @xmath162 with a given mean photon number @xmath163 , one can see that the weighted cfi for different values of @xmath0 can only reach the classical limit @xmath44 , as depicted by fig . [ fig2](c ) , where we considered the special case @xmath164 , for which the pmc is naturally fulfilled and therefore @xmath165 . in order to improve the estimation precision , all the detection evens @xmath59 have to be taken into account in the photon - counting measurements , which gives ideal result of the cfi braunstein , braunstein96,luo , smerzi09,donner : @xmath166 ^{2}}{p(j,\mu we have reexpressed the input state as @xmath167 , so we have @xmath168and @xmath169 , given by eq . ( [ pnmu ] ) . note that the total cfi is indeed a sum of each @xmath0-component contribution @xmath170 weighted by @xmath57 . with only the @xmath0-photon detection events , the fisher information is simply given by @xmath171 , as mentioned above . similar to eq . ( [ fqn ] ) , we further calculate the total qfi of the output state @xmath172 , which is independent from any specific measurement scheme and is given by @xmath173 @xcite . for the input state @xmath174 , we obtain @xmath175 , and hence ideal result of the qfi @xmath176where @xmath135 is the qfi of the @xmath0-photon component . under the pmc , we have show that for each @xmath0-component @xmath177 , which naturally results in a global phase estimation @xmath178 @xcite . according to refs . smerzi08,lang , liu , one can obtain explicit form of the qfi by directly calculating @xmath179 ( see also the append . b ) , namely latexmath:[\[f^{\mathrm{(id)}}(\varphi ) = f_{q}^{\mathrm{(id)}}=|\alpha |^{2}e^{2|\xi photon number @xmath3 , maximum of the qfi was found to achieve the heisenberg scaling @xmath181 @xcite , provided @xmath182(d)-(f ) . however , such a scaling is only possible with exactly perfect photon - number - resolving detectors @xcite , which enable us to record infinite number of the photon - counting events ; see also eq . ( [ cfi_cv ] ) . usually , a single number - resolving detector can only register the number of photons up to @xmath27 @xcite . it is therefore important to investigate the cfi of each @xmath0-component for @xmath0 up to a finite number of photons being resolvable @xmath30 . for brevity , we consider the input fields with the real amplitudes and large enough mean photon number ( i.e. , @xmath183 ) . since the pmc is naturally fulfilled , the cfi is still a sum of each @xmath0-component with the weight @xmath184 and equals to the qfi : @xmath185}\left [ n+4k(n-2k)+\frac{% 4k\alpha ^{2}}{\tanh \xi } \right ] \left [ c_{n-2k}(0)s_{2k}(0)\right ] ^{2 } , \label{qfires}\end{aligned}\ ] ] where @xmath158 is the @xmath0-photon state and @xmath186 denotes its generation probability , given by eqs . ( [ psi_n ] ) and ( [ gn ] ) . obviously , the qfi considered here depends on three variables @xmath187 , or equivalently , @xmath188 for a given @xmath189 . when @xmath190 , the ideal result of the qfi is recovered ( see the append . b ) . , ( b ) the qfi of each @xmath0-photon state @xmath135 , ( c ) the weighted qfi @xmath191 , and ( d ) the total qfi @xmath192 , for @xmath193 fixed and @xmath194 ( solid circles ) , @xmath5 ( squares ) , @xmath195 ( open circles ) , and @xmath196 ( red solid lines ) . the last case is given by eq . ( [ qfi ] ) , which predicts @xmath197 and @xmath198 . for each a given @xmath199 $ ] , using the same values of @xmath30 and maximizing the total qfi with respect to @xmath200 to obtain : ( e ) @xmath201 , and ( f ) the associated qfi @xmath202 . in ( c ) , the peak height of the weighted qfi is about @xmath203 . the vertical lines in ( d ) : the optimal value of @xmath200 for different values of @xmath30 . the dashed line in ( f ) : the classical limit @xmath204 . ] the heisenberg scaling of the qfi can be maintained for large enough @xmath205 , provided that all the nonvanishing @xmath206 are included . to obtain the minimum value of @xmath30 , we show @xmath57 , @xmath135 , and @xmath207 against @xmath0 and @xmath208 under a constraint on @xmath3 . from fig . [ fig2](b ) , one can see that @xmath135 increases quadratically with @xmath0 . this is because the qfi reaches its maximum @xmath209 when @xmath210 ( see table i ) , which corresponds to @xmath211 , i.e. , the classical light being dominant for a given @xmath212 . on the other hand , the generation probability shows a little complex behavior on @xmath0 ; see fig . [ fig2](a ) . at @xmath213 , @xmath184 is nonvanishing at even number of @xmath0 and decreases monotonically with the increase of @xmath0 . when @xmath214 ( i.e. , @xmath215 ) , it reaches maximum at a certain value of @xmath0 and then decreases . similar to @xmath57 , the weighted qfi @xmath207 reaches maximum at @xmath216 , and then decreases with the increase of @xmath0 . as depicted in fig . [ fig2](c ) , one can also see that the values of @xmath207 tend to vanishing as @xmath217 , implying @xmath218 . to confirm the above result , we maximize eq . ( [ qfires ] ) with respect to @xmath219 for given @xmath3 and @xmath30 . figure [ fig2](d ) shows @xmath13 as a function of @xmath208 for a fixed @xmath193 , where @xmath194 ( the solid circles ) , @xmath5 ( the squares ) , and @xmath195 ( the open circles ) . when @xmath220 ( the red solid line ) , the ideal result of the qfi is recovered and is given by eq . ( [ qfi ] ) , which reaches the heisenberg scaling @xmath221 @xcite . one can see that the qfi with @xmath222 almost follow the ideal result . in fig . [ fig2](e ) and ( f ) , we show optimal value of the ratio @xmath223 and the associated qfi @xmath224 for each a given value of @xmath225 $ ] , where we take the number resolution @xmath30 the same to fig . fig2(d ) . from fig . [ fig2](e ) , one can see that when @xmath226 , the optimal input state contains more coherent light photons than that of the squeezed vacuum . the heisenberg scaling of the qfi is attainable with @xmath41 , as depicted by fig . [ fig2](f ) . figure [ fig3 ] shows @xmath227 as a function of @xmath228 for the increase of @xmath3 from @xmath229 up to @xmath230 . for each a given @xmath3 , we first maximize the ideal qfi with respect to @xmath208 to obtain @xmath231 and @xmath232 , as depicted by the red lines of fig . fig2(d)-(f ) , and then calculate the qfi of eq . ( [ qfires ] ) using the same input state . our numerical results show that @xmath227 increases with @xmath30 and approaches to @xmath233 as @xmath234 . to quantify how much phase information is kept by a finite cutoff @xmath205 , we try to find analytical result of @xmath235 in the limit @xmath236 . to this end , we first separate the qfi into two terms @xmath237 , where @xmath238 denotes the qfi being lost . this expression is the same to eq . ( [ qfires ] ) , except the sum over @xmath239 . next , we note that the photon number distribution of the coherent state is much narrow than that of the squeezed vacuum , which enable us to obtain an approximate result of @xmath240 ( see appendix b ) . furthermore , the ideal result of the qfi can reach its maximum at the optimal condition @xmath241 @xcite . using the same input , we obtain@xmath242where @xmath243 and @xmath244 denotes the error function . our analytical result shows a good agreement with the numerical results ; see the solid lines of fig . when @xmath41 , it predicts that over @xmath245 of the ideal qfi can be kept ; while for @xmath246 , most of the phase information is lost . finally , it should be mentioned that coherent - state interferometry has been demonstrated using two visible light photon counters with @xmath247 @xcite . this number resolution is large enough to realize the global phase estimation for the coherent - state input @xmath248 . based upon a bayesian protocol @xcite , the achievable phase sensitivity was found almost saturating quantum cramr - rao bound over a wide phase interval , in agreement with the theoretical prediction @xmath249 . to realize higher - precision optical metrology , it requires a bright nonclassical light source with larger mean photon number smerzi08 , low photon loss @xcite and low noise qasimi , teklu , ycliu , brivio , genoni , genoni12,escher , zhong , bardhan , feng2014,vidrighin , ygao , as well as the photon counters with high detection efficiency @xcite and large enough number resolution . as a function of @xmath228 for given values of @xmath3 , using the optimal condition that maximizes eq . ( [ qfi ] ) . the solid line is given by our asymptotic result , i.e. , eq . ( [ ratio ] ) , and the dashed line is a fitting result for the case @xmath250 . both of them indicate that @xmath245 of the ideal qfi can be obtained as long as @xmath251 ( the vertical dashed lines ) . ] in summary , we have investigated optical phase estimation with coherent @xmath252 squeezed vacuum light by using imperfect photon counters with an upper threshold @xmath30 for the photon number resolution . we show that both the cfi and the qfi are the sum of the contributions from individual @xmath0-photon components , and the cfi always saturates the qfi for each individual @xmath0-photon component . for ideal photon - counting detectors with @xmath253 , the cfi or the qfi attains its maximum @xmath254 when @xmath255 , leading to the heisenberg limit of the estimation precision . for the detectors with large enough number resolution @xmath256 , we find that the optimal input state contains more coherent light photons than that of the squeezed vacuum . we present an analytical result that quantifies the amount of lost information and show that over @xmath4 percent of ideal qfi can be attained as long as @xmath41 ; while for @xmath246 , most of the phase information is lost . our results highlight the important influence of the finite number resolution of photon - counting detectors on optical phase estimation . it is also interesting to explore the performance of other continuous - variable input states , e.g. , a product of two squeezed vacuum @xmath257 @xcite , when realistic photon counters are used . the first two authors p. liu and p. wang contribute equally to this work . we would like to thank s. rosen and y. silberberg for kindly response to our questions . this work has been supported by the nsfc ( grant nos . 11421063 , 11534002 , 11274036 , and 11322542 ) , the national 973 program ( grant nos . 2012cb922104 , 2014cb921403 , and 2014cb848700 ) . g.r.j . also acknowledges support from the major research plan of the nsfc ( grant no . 91636108 ) . formally , a single - mode squeezed vacuum of light is defined by @xmath258 , with the squeeze operator @xcite : @xmath259 = \exp \left ( -e^{i\theta _ { b}}\frac{\tanh |\xi |}{2}b^{\dagger 2}\right ) \left ( \frac{1}{\cosh |\xi |}\right ) ^{b^{\dagger } b+\frac{1}{2}}\exp \left ( e^{-i\theta _ { b}}\frac{\tanh |\xi |}{2}b^{2}\right ) , \label{disentangling}\]]where @xmath260 denotes complex amplitude of the squeezed vacuum . in fock basis , using @xmath261 , the squeezed vacuum can be expressed as @xmath262where @xmath263 denote the probability amplitudes of the squeezed vacuum , given by @xmath264with the hermite polynomials @xmath265 and @xmath266 . note that one can obtain explicit form of the squeezed vacuum using the disentangling formula @xcite , as done in eq . ( disentangling ) , or alternatively , directly solving the eigenvalue equation @xmath267 @xcite . the single - mode squeezed vacuum contains only even number of photons and has been generated in experiments slusher1985,lawu1986,lawu1987,slusher1987,breitenbach97,vahlbruch08,vahlbruch16 . we now consider the interferometer fed with the squeezed vacuum from one input port and a coherent - state light from another port . the coherent state is given by @xmath268 , with the probability amplitudes@xmath269where @xmath270 denotes the complex amplitude of the coherent light . ( [ ss ] ) and ( [ cs ] ) , we have written down explicitly the phase dependence of the probability amplitudes , purely for later use . under the phase - matching condition ( pmc ) : @xmath271 , we now calculate the probability amplitudes of the @xmath0-photon states @xmath116 as @xmath272where we have used explicit form of @xmath57 , given by eq . ( [ gn ] ) , and the condition @xmath273=+1 $ ] for integers @xmath62 . note that eq . ( [ psi_n ] ) can be rewritten as @xmath274 , where @xmath80 denotes the @xmath0-photon states with real amplitudes ( i.e. , @xmath275 ) . finally , we consider a unitary operation @xmath276 on the @xmath0-photon states @xmath116 , with @xmath277 , to obtain eqs . ( [ ampbs ] ) and ( [ pmu ] ) in main text . under the pmc , we obtain @xmath278}e^{i\eta ( j-2k)}c_{n-2k}(0)s_{2k}(0)|j , j-2k\rangle , \end{aligned}\]]where , in the second step , we have used the relation @xmath279 , and eq . ( [ psi_n ] ) with @xmath280 for @xmath80 , which is expressed in terms of the states @xmath281}e^{i\eta ( j-2k)}c_{n-2k}(0)s_{2k}(0)\langle j,\mu |e^{-i\varphi j_{x}}|j , j-2k\rangle \notag \\ & = & \frac{e^{in\theta _ { a}}}{\sqrt{r_{n}(x)}}e^{i(\frac{\pi } { 2}-\eta ) ( \mu -j)}\sum_{k=0}^{[n/2]}e^{-i2k\eta } \frac{\sqrt{(2k)!}}{k!\sqrt{(n-2k)!}}% \left ( 2x\right ) ^{n/2-k}d_{\mu , j-2k}^{j}(\varphi ) , \label{amplitude}\end{aligned}\]]where , in the last step , we have introduced wigner s d - matrix @xmath282 . obviously , for the special case @xmath283 and @xmath284 , we obtain the @xmath0-photon state after the first 50:50 beam splitter @xmath285 and its probability distributions ; see eqs . ( [ ampbs ] ) and ( [ pmu ] ) . for @xmath286 and arbitrary @xmath287 , we can obtain the output state @xmath288 and its probabilities @xmath289 . in a lossless and noiseless interferometer , the qfi of a pure phase - encoded state @xmath290 is simply given by @xmath291 braunstein , braunstein96,luo , smerzi09,donner , where @xmath292 is a hermitian operator . for the squeezed - state interferometer , as illustrated in fig . fig1(a ) , the input state is the product of a coherent state and a squeezed vacuum , i.e. , @xmath293 , and the phase shifter is given by @xmath294 , or @xmath295 , where , for brevity , we have introduced the angular - momentum operators @xmath296 and @xmath297 , with the bosonic operators of two light fields @xmath298 and @xmath65 . according to ref . @xcite , the qfi of the output state @xmath299 is optimal when the two injected light fields are phase matched , i.e. , the pmc @xmath300 . recently , liu _ et al_. @xcite have derived a more general form of the pmc for the interferometer @xmath301 , where a superposition of even or odd number of photons is injected from one port and an arbitrary state from another port . to show it clearly , we focus on the pmc @xmath36 and calculate the qfi of @xmath302 , namely @xmath303where @xmath304 denotes the hermitian conjugate . there are two contributions to the qfi . first , it is easy to obtain @xmath305with @xmath306 and @xmath307 being mean photon number of light fields from two input ports . second , using the relation @xmath308 , we obtain @xmath309therefore , the ideal result of the qfi is given by@xmath310 + \bar{n}% _ { b}\leq \bar{n}_{a}\left [ 1 + 2\bar{n}_{b}+2\sqrt{\bar{n}_{b}(1+\bar{n}_{b})}% \right ] + \bar{n}_{b } , \label{idealqfi}\]]where the equality holds when the pmc is fulfilled , i.e. , @xmath300 . similarly , one can note that the pmc @xmath311 is a good choice for the output state @xmath312 , e.g. , the phases of the two light fields @xmath313 @xcite and @xmath314 @xcite . furthermore , one can simplify the ideal result of the qfi as eq . ( [ qfi ] ) , using the relation @xmath315 . with a finite number resolution @xmath30 , we have shown that the cfi and the qfi are the same and is given by eqs . ( [ qfires ] ) , which can be rewritten as@xmath316 \left [ c_{n_{a}}(0)s_{n_{b}}(0)\right ] ^{2 } \notag \\ & \approx & \bar{n}_{a}\sum_{n_{b}=0}^{n_{\mathrm{res}}-\bar{n}_{a}}\left [ s_{n_{b}}(0)\right ] ^{2}+\left ( 1 + 2\bar{n}_{a}+\frac{2\bar{n}_{a}}{\tanh \xi } \right ) \sum_{n_{b}=0}^{n_{\mathrm{res}}-\bar{n}_{a}}n_{b}\left [ s_{n_{b}}(0)\right ] ^{2 } , \label{qfi2}\end{aligned}\]]where , for brevity , we consider the two light fields with real amplitudes , i.e. , @xmath317 , and the probability amplitudes @xmath318 and @xmath319 are given by eqs . ( [ ss ] ) and ( [ cs ] ) . in the above result , we made an approximation @xmath320 ^{2}\approx\sum_{n_{a}=0}^{\infty}f(n_{a})\left [ c_{n_{a}}(0)\right ] ^{2 } \sum_{n_{b}=0}^{n_{\mathrm{res}}-\bar{n}_{a}}g(n_{b})\left[s_{n_{b}}(0)\right ] ^{2},\]]where @xmath306 and the sum over the mode @xmath65 is still kept , since the photon number distribution of the squeezed vacuum is usually wider than that of the coherent state ( even for @xmath321 ) @xcite . for a finite @xmath322 and @xmath323 , it is easy to obtain the ideal result of the qfi as@xmath324where @xmath325 and @xmath326 is given by eq . ( [ idealqfi ] ) . finally , we consider a finite number resolution with large enough @xmath205 ( @xmath327 ) , and derive analytical result of the qfi . to this end , we first rewrite eq . ( [ qfi2 ] ) as @xmath328 , where @xmath329 quantifies the lost phase information caused by the finite number resolution , given by @xmath330 ^{2}+\left ( 1 + 2\bar{n}_{a}+\frac{2\bar{n}_{a}}{\tanh \xi } % \right ) \sum_{k = n_{\mathrm{res}}-\bar{n}_{a}}^{\infty } k\left [ s_{k}(0)% \right ] ^{2 } \notag \\ & \approx & \left ( 2\bar{n}_{a}+\frac{2\bar{n}_{a}}{\tanh \xi } \right ) \int_{n_{\mathrm{res}}-\bar{n}_{a}}^{\infty } kdk\frac{\left ( \tanh \xi \right ) ^{k}}{\sqrt{2\pi k}\cosh \xi } , \label{qfilost0}\end{aligned}\]]where , in the last step , we only keep the terms @xmath331 . in addition , we replace the sum over @xmath62 by an integral and use the stirling s formula @xmath332 . when @xmath333 , it is easy to find @xmath334 and hence the achievable qfi @xmath335 or @xmath336 , corresponding to almost complete loss of the phase information , or the ultimate estimation precision in the classical limit . to enlarge the qfi , we take @xmath337 and obtain @xmath338 , \label{qfilost1}\]]where @xmath339 , @xmath340 $ ] , and @xmath341 denotes the complementary error function . our analytical result coincides with the numerical results in figs . [ fig2](d)-(f ) . in the limit @xmath342 , we obtain @xmath343 and hence @xmath344 , \]]where @xmath345 . this result gives eq . ( ratio ) in main text . r. demkowicz - dobrzaski , u. dorner , b. j. smith , j. s. lundeen , w. wasilewski , k. banaszek , and i. a. walmsley , phys . a * 80 * , 013825 ( 2009 ) ; u. dorner , r. demkowicz - dobrzaski , b. j. smith , j. s. lundeen , w. wasilewski , k. banaszek , and i. a. walmsley , phys . rev . lett . * 102 * , 040403 ( 2009 ) . y. gao and r. m. wang , phys . a * 93 * , 013809 ( 2016 ) . b. calkins , p. l. mennea , a. e. lita , b. j. metcalf , w. s. kolthammer , a. lamas - linares , j. b. spring , p. c. humphreys , r. p. mirin , j. c. gates , p. g. r. smith , i. a. walmsley , t. gerrits , and s. w. nam , opt . express * 21 * , 22657 ( 2013 ) .

squeezed - state interferometry plays an important role in quantum - enhanced optical phase estimation , as it allows the estimation precision to be improved up to the heisenberg limit by using ideal photon - number - resolving detectors at the output ports . here we show that for each individual @xmath0-photon component of the phase - matched coherent @xmath1 squeezed vacuum input state , the classical fisher information always saturates the quantum fisher information . moreover , the total fisher information is the sum of the contributions from each individual @xmath0-photon components , where the largest @xmath2 is limited by the finite number resolution of available photon counters . based on this observation , we provide an approximate analytical formula that quantifies the amount of lost information due to the finite photon number resolution , e.g. , given the mean photon number @xmath3 in the input state , over @xmath4 percent of the heisenberg limit can be achieved with the number resolution larger than @xmath5 .

in this note , we consider the @xmath0 integrals [ intnp ] du u_i_1j_1u_i_pj_p_k_1l_1_k_nl_n , with @xmath5 the normalized haar measure , or their generating functions [ genfnznp]z_p , n(j,)=du ( ( u))^p ( ( j ) ) ^n , where @xmath6 and @xmath7 are arbitrary @xmath8 matrices . such integrals , mainly over the group u@xmath9 , have been the object of numerous publications in the past , in the context of lattice gauge theories @xcite , or in the large @xmath4 limit @xcite , or for their connections with combinatorics @xcite . integrals over @xmath0 seem to have received less attention , see , however , @xcite . recent work by rossi and veneziano @xcite has prompted this new investigation . let us first recall the physical motivations for studying the @xmath0 integrals ( [ intnp ] ) or ( [ genfnznp ] ) . such integrals appear in the context of lattice calculations of baryon spectrum . indeed , consider a su@xmath9 lattice gauge theory , with link variables denoted @xmath10 , wilson action @xmath11 , and lattice averages @xmath12 . following @xcite , introduce the baryonic wilson loop or book observable " @xmath13 where the ordered products @xmath14 , @xmath15 , stand for @xmath4 ( static ) quark lines joining two points a and b , a distance @xmath16 apart , see fig . 1 . @xmath17 represents a baryon made of @xmath4 quarks , created at a and annihilated at b. to lowest order in a small @xmath18 ( strong coupling ) expansion , @xmath19 , with @xmath20 the total area of the @xmath4 sheets " of the book , and the last @xmath21-integration is carried out on the junction " of these sheets and is given by @xmath22 . higher order corrections in @xmath18 may involve some other integrals @xmath23 . .,width=226 ] by @xmath24 invariance , it is clear that the above integrals vanish if [ non - van - cond ] p - n0 n. from a representation theoretic point of view , the number of independent terms in ( [ genfnznp ] ) , ( of independent tensors with the right symmetries in ( [ intnp ] ) ) , is given by the number of invariants in @xmath25 , where @xmath9 and @xmath26 denote the fundamental @xmath4-dimensional representation of @xmath0 and its complex conjugate . @xmath27 is a well known function of ( traces of powers of ) @xmath28 , ( weingarten s function " @xcite ) , at least for @xmath29 , and one may collect all @xmath30 s into [ defzw ] z_w(j,)=_n0 z_n , n(j , k).for the convenience of the reader , a certain number of known results on these integrals and their generating function are recalled in the appendix . this case @xmath31 in eqs . ( [ intnp ] ) , ( [ genfnznp ] ) will be referred to as the ordinary case " . we now turn to the determination of the @xmath23 . first , for @xmath32 , @xmath33 is given by the only invariant in @xmath34 , namely by the totally antisymmetric tensor product , ( the determinant of @xmath35 , equal to 1 in @xmath0 ) . hence @xmath36 with a constant @xmath37 determined by contraction with @xmath38 and use of @xmath39 , hence @xmath40 , and @xmath41 . thus [ int0 ] du u_i_1j_1u_i_nj_n = _ i_1i_n_j_1j_n.equivalently , z_n,0(j,)=. more generally , for @xmath29 , @xmath42 is the product of @xmath43 by a polynomial of @xmath44 , invariant under @xmath45 , for any @xmath46 , and homogeneous of degree @xmath47 in @xmath48 , hence a polynomial that may be expanded on traces of powers of @xmath48 [ znnn ] z_n+n , n = du ( ( u))^n+n ( ( j))^n= _ n _ t _ . with a sum over partitions of @xmath47 denoted @xmath49 $ ] , @xmath50 , and with the notations t_:= _ q t_q^_q , t_q:= ( j)^q.the coefficients @xmath51 are determined through recursion formulae resulting from a contraction of ( [ intnp ] ) with a kronecker delta : [ diff - eqn ] _ jk z_n+n , n=(n+n ) n _ ilz_n+n-1,n-1 . + the first coefficients are readily determined n=1 & _ [ 1]=1 + n=2 & _ [ 2]=-_[1 ^ 2]= + [ coeff123 ] n=3 & _ [ 3]= _ [ 1,2]=- _ [ 1 ^ 3]= + n=4 & _ [ 4]=- _ [ 1,3]= _ [ 2 ^ 2]= + & _ [ 1 ^ 2,2]=- _ [ 1 ^ 4]= + these coefficients are in fact simply related to the analogous coefficients @xmath52 in the expansion of the ordinary generating function , see appendix , eq . ( [ zalphadef ] ) . if @xmath52 , @xmath53 , is written in the form [ zgener ] _ = with @xmath54 a polynomial of @xmath4 , then [ dgener]_=. this follows from the comparison between the two systems of recursion formulae , see below sect . [ recurr ] , and may be verified on the first coefficients ( [ coeff123 ] ) and ( [ z1234 ] ) . note that the coefficients @xmath51 decrease more slowly than the @xmath52 , for fixed @xmath47 , as @xmath4 grows . this behavior has consequences on the large @xmath4 limit of the generating function @xmath55 defined by [ defzd ] z_d=_n0 z_n+n , n = du e^ uj ( u)^n = : z_d .note also the different summations in ( [ defzw ] ) and ( [ defzd ] ) : here @xmath56 is a homogeneity parameter for the eigenvalues of @xmath57 while in ( [ defzw ] ) , it is @xmath58 that plays that role . in both cases , however , we take @xmath56 of order @xmath4 and set @xmath59 . then while in the ordinary case , a non trivial limit is obtained by taking the traces @xmath60 also of order @xmath4 , resulting in an exponentation @xmath61 , here one finds that the traces @xmath60 have to be kept of order 1 and then @xmath62 , with @xmath63 a non trivial function of the @xmath60 s . indeed one finds for the first terms _ n = t_1 + ( t_1 ^ 2-t_2 ) + ( t_1 ^ 3 - 3t_1t_2 + 2t_3 ) + ( t_1 ^ 4 - 6t_2t_1 ^ 2 + 2t_2 ^ 2 + 8t_3t_1 - 5t_4)+ which is just the beginning of a simple formula in the large @xmath4 limit : [ simple - form ] w_d:=_n = _ n1^n _ n ( -1)^n-_q _ p in terms of the catalan numbers @xmath64 . this will be proved below in sect . [ expon ] . let us return to ( [ diff - eqn ] ) and look at the action of the differential operator @xmath65 on a typical term @xmath66 of ( [ znnn ] ) . making use of the identities & = & ( cof ) _ ji + ( cof ) _ ji ( x)_jl & = & ( ) x_il one may write for each term in the expansion of @xmath67 & & ( t _ ) = _ q=1^p q _ q ( ( j)^q-1)_jl t _ + [ recurs ] & = & _ q=1^p q _ q ( ( n+1 ) ( j)^q-1_il + _ s=1^q-1 t_s ( j)^q - s-1_il ) t _ + & + & _ q=1^p q^2 _ q ( _ q-1 ) ( j)^2q-1_il t _ + 2 _ 1q < rp qr_q_r ( j)^q+r-1_il t _ with @xmath68 , @xmath69 , and @xmath70 . in this way , ( [ diff - eqn ] ) yields an ( overdetermined ) system of relations between the coefficients @xmath51 at ranks @xmath47 and @xmath71 , namely [ recurd ] & & _ n _ \ { _ q=1^p q _ q ( ( j)^q-1_il + _ s=1^q-1 t_s ( j)^q - s-1_il ) t _ ( ) . + . & + & _ q=1^p q^2 _ q ( _ q-1 ) ( j)^2q-1_il t _ ( ) + 2 _ 1q < rp qr_q_r ( j)^q+r-1_il t _ ( ) } + & & = n _ il _ n-1 _ t_ ( ) . compare these equations with those satisfied by the coefficients @xmath52 in the similar expansion of @xmath27 in the ordinary case , see ( [ recurz ] ) . their structure is the same , except for changes in the terms of ( [ recurd ] ) marked in red . in the linear system on the @xmath51 , the parameter @xmath4 in the lhs of ( [ recurz ] ) has been changed into @xmath72 while the right hand side is multiplied by @xmath73 . as a result the solutions of the @xmath51 linear system are obtained from those of the @xmath52 one by @xmath74 if @xmath52 is written as in ( [ zgener ] ) , it follows that @xmath51 has the form ( [ dgener ] ) , qed . + in particular , this relation implies that @xmath52 and @xmath51 have the same overall sign , namely @xmath75 . the differential equation ( [ diff - eqn ] ) carries over to the generating function @xmath55 of ( [ defzd ] ) in the form _ jk z_d=_il ( ( n+1 ) + ^2 ) z_d . we write @xmath76 , in which @xmath77 is a function of @xmath78 invariant under @xmath79 for any @xmath46 , and we rewrite the differential equation as m_jk + ( n+1 ) = _ il ( ( n+1 ) + ^2 ) z_d.this may be reexpressed as a differential equation wrt the eigenvalues @xmath80 of @xmath48 ( generically @xmath48 is diagonalizable ) . this is a standard procedure @xcite with the result that for any @xmath81 , @xmath82 _ i + ( n+1 ) + _ ji _ j = ( ( n+1 ) + ^2 ) z_d . finally we write @xmath62 , which results in _ i ( n + n^2 ( ) ^2 ) + ( n+1)n + n_ji _ j = ( n+1 ) + n^2 . in the large @xmath4 limit , we rescale @xmath59 , keeping all @xmath83 of order 1 , and after dropping the subdominant terms , we get for @xmath84 the equation _ i w_i^2 + w_i = + ^2 . ( note that this is in contrast with the ordinary case " where the term @xmath85 contributes in the large @xmath4 limit @xcite . ) now @xmath86 is just an homogeneity variable of the @xmath87 s , and we may thus substitute @xmath88 . the equation finally reduces to a system of _ algebraic _ equations for the @xmath89 s [ ] _ i w_i^2 + w_i = ( 1+_j _ j w_j ) . assuming @xmath90 known , one finds [ wi - eqn ] _ i w_i^2 + w_i = ( 1+w ) + w_i= and @xmath91 is thus the root of [ w - eqn]1+w=1- + _ i . equation ( [ wi - eqn ] ) should be compared with that of the generating function of catalan numbers @xmath64 , namely @xmath92 , @xmath93 . we find that @xmath94 , hence w_i= ( 1+w)_n0 ( n ) ( -(1+w)_i)^n and [ w - eqn2 ] w=_i _ i w_i= -_n1 ( n-1 ) ( -(1+w))^n t_n . let @xmath95 , @xmath96 for @xmath97 , @xmath98 , then ( [ w - eqn2 ] ) reads @xmath99 , whose solution is given by lagrange formula y=(1+w ) & = & _ n0 ( ) ^n ( _ m0 f_m z^m ) ^n+1|_z=0 + & = & + _ n1 _ n n ! + w=_n1 m_n ^n & = & _ n1 ^n _ n f_where the multinomial coefficient appears naturally in the expansion of the @xmath100-th power , and the @xmath101 results from the @xmath47-th derivative of @xmath102 . upon integration we get @xmath103 which establishes ( [ simple - form ] ) . note the similarity of this calculation with the relation between ordinary moments @xmath104 and non crossing ( or free ) cumulants @xmath105 of a given distribution . the combinatorial or diagrammatical interpretation of ( [ simple - form ] ) remains to be found . it is a pleasure to thank gabriele veneziano for inspiring discussions . 99 d. weingarten , _ asymptotic behavior of group integrals in the limit of infinite rank _ , j. math . phys * 19 * ( 1978 ) 9991001 m. creutz , _ on invariant integration over su(@xmath4 ) _ , j. math . phys . * 19 * ( 1978 ) 20432046 i. bars and f. green , _ complete integration of u(@xmath4 ) lattice gauge theory in the large @xmath4 limit _ , phys rev * d20 * ( 1979 ) 33113330 ; + s. samuel , _ u(@xmath4 ) integrals , 1/@xmath4 and de witt hooft anomalies _ , j. math * 21 * ( 1980 ) 26952703 ; + i. bars , _ u(@xmath4 ) integral for the generating functional in lattice gauge theories _ , j. math * 21 * ( 1980 ) 26782681 j .- m . drouffe and j .- b . zuber , _ strong coupling and mean field methods in lattice gauge theories _ , physics reports * 102 * ( 1983 ) 1119 . brzin and d. gross , _ the external field problem in the large @xmath4 limit of qcd _ , phys . lett . * 97b * ( 1980 ) 120124 k.h . obrien and j .- b . zuber , _ a note on u(@xmath4 ) integrals in the large @xmath4 limit _ , phys . lett . * 144b * ( 1984 ) 407408 b. collins , _ moments and cumulants of polynomial random variables on unitary groups , the itzykson - zuber integral and free probability _ , imrn * 17 * ( 2003 ) 952982 ; + b. collins and p. niady , _ integration with respect to the haar measure of unitary , orthogonal and symplectic group _ , * 264 * ( 2006 ) 773795 j. novak , _ complete homogeneous symmetric polynomials in jucys murphy elements and the weingarten function _ , banach center publications , * 89 * ( 2010 ) 231235 , http://arxiv.org/abs/0811.3595 ; + p. zinn - justin , _ jucys murphy elements and weingarten matrices _ , lett . math . * 91 * ( 2 ) ( 2010 ) , 119127 , http://arxiv.org/abs/0907.2719 j. carlsson , _ integrals over su(@xmath4 ) _ , http://arxiv.org/abs/0802.3409 g.c . rossi and g. veneziano , _ the string - junction picture of multiquark states : an update _ , http://arxiv.org/abs/1603.05830 in this appendix , we recall a certain number of results on the weingarten s functions @xmath27 and their generating function @xmath106 of ( [ defzw ] ) . the @xmath30 are well known functions of ( traces of powers of ) @xmath28 , at least for @xmath29 , see also @xcite . for @xmath107 , the traces are no longer independent , or equivalently the tensors @xmath108 , @xmath109 , are no longer independent , and we have to deal with a pseudo - inverse " of the gram matrix , see @xcite . consider the integrals on u@xmath9 ( or su@xmath9 , this is irrelevant here , assuming @xmath29 ) @xcite du u_i_1j_1u_i_nj_n _ k_1_1 _ k_n_n&= & _ , s_n c ( [ ] ) _ a=1^n _ i_a_(a ) _ j_a k_(a ) + & = & _ , s_n _ y young diagr.|y|=n ( ^()(1))^2 ^()([])n!^2 s_(i ) _ q=1^p _ i_q_(q ) _ j_q k_(q ) where @xmath110)$ ] is the character of the symmetric group @xmath111 associated with the young diagram @xmath112 , ( a function of the class @xmath113 $ ] of @xmath114 ) ; thus @xmath115 is the dimension of that representation ; @xmath116 is the character of the linear group gl(@xmath4 ) associated with young diagram @xmath112 , that is a schur function when expressed in terms of the eigenvalues of @xmath117 ; @xmath118 is thus the dimension of that representation . finally the coefficient @xmath119)$ ] will be determined below . alternatively , in terms of generating functions with sources @xmath6 and @xmath120 [ weing ] z_n , n(j,)=du ( u)^n ( j ) ^n = _ n n ! || c ( [ ] ) t _ + where @xmath121 is the cardinal of class @xmath122 $ ] in @xmath111 , thus @xmath123 . in ( [ weing ] ) , the factor @xmath101 comes from the sum over @xmath124 and the factor @xmath121 from that over the elements @xmath125 $ ] . thus [ zalphadef ] z_w(j,):= _ u(n ) du = _ n=0^ _ n _ t _ + with @xmath126)$ ] . by the same argument as in sect . 2 , the @xmath52 satisfy the linear system of recursion relations [ recurz ] & & _ n _ \ { _ q=1^p q _ q ( n ( j)^q-1_il + _ s=1^q-1 t_s ( j)^q - s-1_il ) t _ . + . & & + _ q=1^p q^2 _ q ( _ q-1 ) ( j)^2q-1_il t _ + 2 _ 1q < rp qr_q_r ( j)^q+r-1_il t _ } + & & = n _ il _ n-1 _ t_ . explicitly , the first coefficients @xmath127)$ ] read n=1 & _ [ 1]= + n=2 & _ [ 2]=- , _ [ 1 ^ 2]= + [ z1234 ] n=3 & _ [ 3]= , _ [ 1,2]=- , _ [ 1 ^ 3 ] = + n=4 & _ [ 4]=- , _ [ 1,3]= + & _ [ 2 ^ 2 ] = , _ [ 1 ^ 2,2 ] = - , + & _ [ 1 ^ 4 ] = + the overall sign of @xmath52 is @xmath128 and its large @xmath4 behavior @xmath129 . to study the large @xmath4 limit , we take @xmath130 and @xmath131 , with @xmath132 and @xmath133 of order 1 . then brzin and gross @xcite have shown that w_w(j):= _ n z_w(j , ; n ) exists and satisfies the coupled equations [ bgsystem ] w_w & = & _ i ( ^2 _ i+c)^- _ i , j - c- + with c&= & the solution has two determinations , in a strong coupling and in a weak coupling phase . here we are concerned with the strong coupling regime in which we may expand w_w=_n1 ^2n _ n w__with the first terms w_w=^2 _ 1 + ( _ 1 ^ 2 -_2 ) + ( 2 _ 1 ^ 3 - 3 _ 1 _ 2 + _ 3 ) + ( 24 _ 1 ^ 4 - 48 _ 1 ^ 2 _ 2 + 9 _ 2 ^ 2 + 20 _ 1 _ 3 - 5 _ 4)+ and the general term given in @xcite [ obzres ] w_= ( -1)^n _ q ( -)^_q .

in this note , i revisit integrals over @xmath0 of the form @xmath1 . while the case @xmath2 is well known , it seems that explicit expressions for @xmath3 had not appeared in the literature . similarities and differences , in particular in the large @xmath4 limit , between the two cases are discussed .

recent work@xcite has shown that thermodynamic properties can be computed for large-@xmath0 supersymmetric theories . the approach is based on light - cone coordinates@xcite and the numerical technique of supersymmetric discrete light - cone quantization ( sdlcq).@xcite here we consider two - dimensional supersymmetric quantum chromodynamics with a chern simons term ( sqcd - cs),@xcite dimensionally reduced from three dimensions . light - cone coordinates@xcite are defined by the time variable , @xmath1 , and spatial components , @xmath2 , where @xmath3 and @xmath4 . the light - cone energy and momentum are given by @xmath5 and @xmath6 , respectively . for field theories quantized in terms of these coordinates , the standard numerical technique is discrete light - cone quantization ( dlcq).@xcite space is restricted to a light - cone box @xmath7 , @xmath8 with periodic boundary conditions . momentum is then discretized as @xmath9 , @xmath10 with @xmath11 , @xmath12 and @xmath13 all integers . the limit @xmath14 is exchanged for a limit in terms of the integer ( harmonic ) resolution@xcite @xmath15 for fixed total momentum @xmath16 . because the @xmath11 are positive , the number of particles is limited to no more than @xmath17 . integrals are replaced by discrete sums . sdlcq@xcite is a special form of dlcq that preserves at least part of the supersymmetry algebra \{q^+,q^+}=2p^+ , \{q^-,q^-}=2p^- , \{q^+,q^-}=-4p_. instead of discretizing the hamiltonian @xmath18 directly , the supercharge @xmath19 is discretized , and @xmath18 is computed from the algebra as p_sdlcq^-=\{q^-,q^- } p_dlcq^- . for ordinary dlcq , one recovers supersymmetry only in the infinite resolution limit . after a brief summary of the sqcd - cs theory , we show how thermodynamic quantities can be constructed from the partition function . this requires knowledge of the spectrum of the theory , which we obtain numerically with an iterative lanczos algorithm . some preliminary results are presented and future work discussed . we consider a dimensional reduction from 2 + 1 to 1 + 1 dimensions of @xmath20 supersymmetric quantum chromodynamics with a chern simons term ( sqcd - cs).@xcite the action is [ action ] s&=&d^3x\{-f_f^ + d_^d^+i| d_^. + & & -g+^d _ + & & .+^ + | } , where the adjoint fields are the gauge boson @xmath21 ( gluons ) and a majorana fermion @xmath22 ( gluinos ) and the fundamental fields are a dirac fermion @xmath23 ( quarks ) and a complex scalar @xmath24 ( squarks ) . the chern simons coupling , @xmath25 , has the effect of providing a mass for the adjoint fields . the covariant derivatives are d_&=&_+ig[a_,],d_=_+iga _ , + d_&=&_+iga_. the fields transform according to & & a_=_,=f_^ , + & & = , = -^d_. we reduce to 1 + 1 dimensions by assuming the fields to be independent of the transverse coordinate @xmath26 . we define fermion components by = ( , ) ^t,=(,)^t , q=(q^+,q^-)^t . there are constraints , which in light - cone gauge ( @xmath27 ) are written _ -&=&- ( [ a^2,]+i^-i^ ) , + _ -&=&-a^2 + -/ , [ fcurrent ] ^2_-a^-=gj , with [ current ] ji[a^2,_-a^2]+ \ { , } + _ - a^2 -ih_-^+ i_-^+^. the reduced supercharge is q^-&=&gdx^-\ { 2 ^ 3/4(i[a^2,_-a^2]-_-a^2 + \ { , } ) . + & & -(i _ -^-i_-^+2^ ) + & & .-2(^a^2+^a^2)}. in the large-@xmath0 approximation , there are only single - trace fock states , the mesons ^_i_1(k_1 ) a^_i_1i_2(k_2) b^_i_ni_n+1(k_n-1) f^_i_p(k_n)|0and glueballs [ a^_i_1i_2(k_1) b^_i_ni_n+1(k_n)]|0 , where @xmath28 and @xmath29 create fundamental partons and @xmath30 and @xmath31 create adjoint partons . either type of state could be a boson or a fermion . the theory possesses a useful @xmath32 symmetry@xcite a_ij(k , n^)-a_ji(k , n^ ) , b_ij(k , n^)-b_ji(k , n^ ) , which further divides the fock space between states with even and odd numbers of gluons . we then diagonalize in each sector separately . from the partition function , @xmath33 , we compute the bosonic free energy _ b=_n=1^_m_n^dp_0 ( 1- e^- p_0/t ) . and the fermionic free energy _ f= -_n=1^ _ m_n^dp_0 ( 1 + e^- p_0/t ) . the total free energy , once expanded in logarithms and @xmath34 integrals are performed , is given by ( t , v)=- vt^2-_n=1^_l=0^ m_n . the sum over @xmath35 is well approximated by the first few terms . we represent the sum over @xmath36 as an integral over a density of states @xmath37 : @xmath38 . the density is approximated by a continuous function , and the integral @xmath39 is computed by standard numerical techniques . the discrete density of states is @xmath40 , where @xmath41 is the degeneracy of the mass eigenvalue @xmath42 . the density can be written in the form of a trace over the evolution operator @xmath43 : ( m^2)=_-^e^im^2x^+/2p^+ tre^-ip^-x^+ dx^+ . we approximate the trace as an average over a random sample of vectors@xcite ( m^2)1s_s=1^s _ s(m^2 ) , with @xmath44 a local density for a single vector @xmath45 , defined by _ s(m^2)=_-^e^im^2x^+/2p^+ s|e^-ip^-x^+|sdx^+ . the sample vectors @xmath45 can be chosen as random phase vectors;@xcite the coefficient of each fock state in the basis is a random number of modulus one . we approximate the matrix element @xmath46 by lanczos iterations.@xcite let @xmath47 be the length of @xmath45 , and define @xmath48 as the initial lanczos vector . the matrix element @xmath49 can be approximated by the @xmath50 element of the exponentiation of the lanczos tridiagonalization of @xmath18 . let @xmath51 be the tridiagonal lanczos matrix . it can be exponentiated by first diagonalizing it : p_s^-_j^s=_j^s , such that @xmath52 , with @xmath53 and @xmath54 . the @xmath50 element is given by ( e^-ip_s^-x^+)_11=_n|(c_n^s)_1|^2e^-im_sn^2 x^+/2p^+ . the local density can now be estimated by _ s(m^2)_n w_sn ( m^2-m_sn^2 ) , where @xmath55 is the weight of each lanczos eigenvalue . only the extreme lanczos eigenvalues are good approximations to eigenvalues of the original @xmath18 . the other lanczos eigenvalues provide a smeared representation of the full spectrum . from this density of states , we compute the cumulative distribution function ( cdf ) , @xmath56 as an average n(m^2)1s_s n_s(m^2 ) , of local cdfs n_s(m^2)^m^2 dm^2_s(m^2)_n w_sn(m^2-m_sn^2 ) . the convergence of the approximation is dependent on the number of lanczos iterations per sample , as well as the number @xmath57 of samples . test runs indicate that taking 20 samples is sufficient . the number of lanczos iterations needs to be on the order of 1000 per sample ; using only 100 leaves errors on the order of 1 - 2% . some preliminary results are presented in the accompanying figures . figure 1 compares the numerical results for the cdf to analytic results , which can be obtained when the yang - mills coupling is zero . the numerical results are quite good , with only one noticeable deviation , at large @xmath58 , a region where the discrete spectrum is sparse . figure 2 shows the free energy at particular values of the yang mills coupling . in the analytically solvable case of zero yang mills coupling . the numerical and analytic solutions are compared.,width=377 ] as a function of temperature @xmath59 for resolution @xmath60.,width=377 ] additional work is in progress to complete this study of finite temperature properties of two - dimensional sqcd . beyond this particular effort , one can consider finite-@xmath0 effects , with baryons and mixing of mesons and glueballs , and the full three - dimensional theory . as these techniques mature , analysis of four - dimensional theories can be considered . the work reported here was done in collaboration with s. pinsky , y. proestos , n. salwen , and u. trittmann and supported in part by the us department of energy and the minnesota supercomputing institute . 0 j. r. hiller , y. proestos , s. pinsky , and n. salwen , _ phys . d _ * 70 * , 065012 ( 2004 ) . a. m. dirac , _ rev . phys . _ * 21 * , 392 ( 1949 ) . y. matsumura , n. sakai , and t. sakai , _ phys . d _ * 52 * , 2446 ( 1995 ) . o. lunin and s. s. pinsky , aip conf . proc . * 494 * , 140 ( 1999 ) . j. r. hiller , s. s. pinsky , and u. trittmann , _ nucl . b _ * 661 * , 99 ( 2003 ) ; _ phys . rev . d _ * 67 * , 115005 ( 2003 ) pauli and s. j. brodsky , _ phys . d _ * 32 * , 1993 ( 1985 ) ; 2001 ( 1985 ) . s. j. brodsky , h .- c . pauli , and s. s. pinsky , _ phys . _ * 301 * , 299 ( 1997 ) . d. kutasov , _ nucl . b _ * 414 * , 33 ( 1994 ) . r. alben , m. blume , h. krakauer , and l. schwartz , _ phys . b _ * 12 * , 4090 ( 1975 ) ; a. hams and h. de raedt , _ phys . rev . e _ * 62 * , 4365 ( 2000 ) . t. iitaka and t. ebisuzaki , _ phys . rev . e _ * 69 * , 057701 ( 2004 ) . j. jakli and p. prelovek , _ phys . b _ * 49 * , 5065(r ) ( 1994 ) ; m. aichhorn , m. daghofer , h. g. evertz , and w. von der linden , _ phys . b _ * 67 * , 161103(r ) ( 2003 ) .

light - cone coordinates and supersymmetric discrete light - cone quantization are used to analyze the thermodynamics of two - dimensional supersymmetric quantum chromodynamics with a chern simons term in the large-@xmath0 approximation . this requires estimation of the entire spectrum of the theory , which is done with a new algorithm based on lanczos iterations . although this work is still in progress , some preliminary results are presented .

scalar fields are often used in general relativity , in particular to discuss the cosmological observations @xcite . in theoretical models of a dynamical dark energy the `` cosmological '' scalar fields provide an inflationary stage of the early universe ( see , e.g. , @xcite for a review and references therein ) . one might think that effects of such fields on galactic scales are negligible at present epoch . nevertheless , since the early works @xcite , it is well known that any presence of a scalar field can cause a cardinal change of the space - time topology in the vicinity of a compact object . the test body motion is probably a principal mean that can signal , by means of electromagnetic radiation , about the presence or absence of `` non - canonical '' effects related to scalar fields . the circular orbits around compact objects with scalar fields has been studied elsewhere @xcite . there is an interesting property of the space - time geodesic structure discovered in presence of the linear massless scalar field @xcite : in this case there exist disconnected regions of stable circular orbits . discontinuous structure of the orbits arises also owing to the presence of a quadrupole source @xcite . one can suppose that such separated rings of orbits near relativistic compact objects is a widespread phenomenon in presence of the scalar fields ( and maybe , in presence of the other fields ) . if such a separation really exists , this could affect considerably the form of the relativistic fluorescent iron lines generated in accretion disks , which are observed in active galactic nuclei ( agns ) @xcite . this may be important for observations in order to test gravitational theories . the present paper deals with test - body orbits in case of a family of spherically symmetric space - times of general relativity , which describe static configurations in presence of a self - interacting scalar field . we derived exact special solutions of einstein s equations using a known method ( see , e.g. , @xcite ) that allows one to generate asymptotically flat space - time metrics along with the corresponding scalar field potentials ( section [ sec : setup ] ) . we concentrate on solutions , which describe either a naked singularity ( ns ) or a black hole ( bh ) having a positive mass . then we consider a test body motion in a static region ( section [ orbits ] ) to find conditions for the existence of stable circular orbits . in particular , our numerical investigation shows that for some set of parameters of the problem there are disconnected regions of stable circular orbits , which are separated by a region where a stable motion is impossible . the action functional in presence of a self - interacting minimally coupled scalar field @xmath0 is @xmath1 \quad , \ ] ] where @xmath2 is the standard gravitational action of general relativity ( @xmath3 ) , @xmath4 is a self - interaction potential to be specified below . the metric of a spherically symmetric space - time can be written as @xmath5 the einstein s equations for this metric following from action ( [ eq1 ] ) yield ( see , e.g. , @xcite,@xcite,@xcite ) : @xmath6 @xmath7 , @xmath8 variation of action ( [ eq1 ] ) with respect to @xmath0 yields one one more equation , which is not independent from equations ( [ eq3][eq3c ] ) ; therefore , we do not write it . equation ( [ eq3c ] ) yields @xmath9 where @xmath10 is a constant . in view of ( [ eq3]),([eq3b ] ) we have @xmath11 and @xmath12 the system of equations ( [ eq4][eq5a ] ) is equivalent to ( [ eq3][eq3c ] ) . on account of ( [ eq4]),([eq5 ] ) we can use the `` inverse '' method @xcite to generate special solutions of this system : instead of looking for functions @xmath13 for some given potential @xmath4 , we may fix any of these functions , for example @xmath14 , and look for @xmath15 and @xmath4 . this problem is solved by quadratures according to ( [ eq4]),([eq5 ] ) , ( [ eq5a ] ) ; @xmath4 is defined parametrically under condition @xmath16 . now we specify the problem to consider a family of solutions with @xmath17,\quad n > 1.\ ] ] obviously , for @xmath18 we have @xmath19 , @xmath20 , that is the conditions to apply the inverse method are fulfilled . we look for solutions that describe configurations with positive mass @xmath21 , i.e. having an asymptotic at large distances from the center @xmath22 . using expansions of ( [ eq4 ] ) in powers of @xmath23 ( see ( [ a_at_infty ] ) below ) , we have then @xmath24 . integration in ( [ eq4 ] ) for @xmath18 yields @xmath25 in terms of the hypergeometric function @xmath26 ^ 2\left\ { { f\left [ { 4,\frac{2}{n},1 + \frac{2}{n},\left ( { \frac{x_0 } { x } } \right)^n } \right ] - \frac{c}{3x}f\left [ { 4,\frac{3}{n},1 + \frac{3}{n},\left ( { \frac{x_0 } { x } } \right)^n } \right ] } \right\ } \quad .\ ] ] in view of ( [ eq5 ] ) we have @xmath27.\ ] ] formulas ( [ eq5a],[eq7],[eq8 ] ) define the potential @xmath28 parametrically for @xmath29 and correspondingly @xmath30 . the potential @xmath28 is explicitly derived using the functions @xmath31 in ( [ eq5],[eq5a ] ) , then we get @xmath4 by a substitution of @xmath32 into @xmath28 . thus , we get a spherically - symmetric solutions of einstein s equations with the scalar field in case of this potential ; the solutions are given by ( [ eq6],[eq7],[eq8 ] ) . asymptotic relations at the spatial infinity are as follows @xmath33 @xmath34 the scalar field self - interaction potential near the origin is similar for the ns and bh cases ( see fig . [ sf_potentials ] , left panel ) : @xmath35 here we are dealing with massles scalar fields . using ( [ eq4 ] ) and ( [ eq6 ] ) we have in the neighborhood of @xmath36 : the metric coefficient @xmath37+o(x - x_0),\ ] ] the potential @xmath38 and the ricci scalar @xmath39 if @xmath40 , then for @xmath30 the integrand of ( [ eq4 ] ) is positive and @xmath41 . it is easy to show that the time needed for signals from the center @xmath42 to reach an external observer , is finite . therefore we have ns at @xmath43 . in this case the scalar field potential is not bounded from below . if @xmath44 , then @xmath45 for @xmath46 , so @xmath25 changes its sign at some point @xmath47 . according to @xcite this is a simple horizon of bh . the qualitative behavior of the solutions is consistent with the general results of @xcite , @xcite . in both cases the potentials @xmath48 are negative in some region ( see examples in fig . [ sf_potentials ] ) , so the solutions do not contradict to the no - scalar - hair theorems @xcite ) . in case of ns we have @xmath49 . in case of bh @xmath50 for @xmath51 , here we have something like the `` mexican hat '' potential ( fig . [ sf_potentials ] ) . . right panel : the cases of bh for larger @xmath0 showing the asymptotic behavior when @xmath50 ; the details near the origin are not visible on this scale.,title="fig:",width=302 ] . right panel : the cases of bh for larger @xmath0 showing the asymptotic behavior when @xmath50 ; the details near the origin are not visible on this scale.,title="fig:",width=302 ] -3 mm further we use notations according to fig . [ proba-0 ] : the radial coordinates @xmath52 correspond to the outer region of circular orbits , and @xmath53 correspond to the inner region ( if it exists ) , which is separated by a prohibited area @xmath54 from the outer region . -3 mm in case of spherically symmetric space - time metrics ( [ eq2 ] ) the integrals for the test body geodesic motion in the equatorial plane are as follows @xmath55 @xmath56 is the canonical parameter on the time - like geodesics . these equations together with the normalization integral yield the equation for the radial variable @xmath57 where @xmath58 equation ( [ eq10 ] ) formally describes a classical particle motion with energy @xmath59 in a field with effective potential @xmath60 . minima of @xmath61 for some fixed @xmath62 correspond to stable circular orbits ( with some fixed angular momenta @xmath63 ) , and the maxima to unstable circular orbits that are limiting cycles for the test body trajectories . at the point of the minimum @xmath64 in case of the schwarzschild metric ( @xmath65 ) there may be either one minimum @xmath66 of @xmath67 and one maximum @xmath68 , or there is none . here we shall look for situations when there may be two minima of @xmath67 . let us consider when additional minima can appear . the bifurcation parameter is @xmath69 . at the bifurcation point ( e.g. , when the minimum / maximum appears or disappears ) in addition to ( [ eq12 ] ) we have @xmath70 it follows from ( [ eq12]),([eq13 ] ) that the bifurcation values @xmath71 ( if they exist ) must be determined from the formula @xmath72 where @xmath73 is a root of the function @xmath74 the explicit form of this function on account of ( [ eq4 ] ) is @xmath75a(x ) - \frac{2}{r^2(x ) } + \left [ { \frac{2r(x){r}'(x)}{2x - c}a(x ) - 1 } \right ] \cdot \left [ { \frac{4{r}'(x)}{r^3(x)}(2x - c ) - \frac{2}{r^2(x ) } } \right].\ ] ] for @xmath76 on account of ( [ eq9 ] ) and ( [ eq6 ] ) we have the asymptotic behavior : @xmath77 for @xmath78 we have @xmath79 . for numerical estimates we choose further the units of length so as to have @xmath80 . using zeros of @xmath81 one can look for parameters @xmath82 that lead to an appearence or disappearance of minima of @xmath61 for some @xmath83 . for three values of @xmath84 in the bh case . the dashed curve shows @xmath85 for @xmath86 , which defines the lower boundary of the stable circular orbit radii @xmath87 for different @xmath88 . the case of @xmath89 ( solid ) represents a three - valued function between @xmath90 and @xmath91 : the branch between points @xmath92 and @xmath93 represents @xmath94 , between @xmath93 and @xmath95 @xmath96 and between @xmath95 and @xmath97 @xmath98 . the latter branch extends to infinity for @xmath99 . the case of @xmath100 ( dotted line ) is intermediate ; it shows how the curves transform as @xmath84 changes . , title="fig:",width=377 ] -3 mm for @xmath101 ( solid ) , @xmath102 ( dashed ) : @xmath103 ( dotted ) . the lower curves @xmath104 are derived from the condition of minimum of @xmath25.,title="fig:",width=377 ] -3 mm as a function of @xmath84 for different @xmath88 in the neigborhood of a qualitative transition at @xmath105 : 1 @xmath106 , 2 @xmath107 , 3 @xmath108 , 4 @xmath109 . the lower curve for @xmath110 ( the ns case ) is actually made up of four different branches ( 1,2,3,4 ) of the boundary radii @xmath111 , which are almost merged together.,title="fig:",width=377 ] -3 mm after determination of roots @xmath112 of @xmath81 for fixed @xmath113 we determine limiting values of angular momentum @xmath114 according to ( [ eq13 ] ) when additional minima of @xmath67 begin to appear . however , equation @xmath115 ( [ eq14 ] ) gives us a necessary ( not sufficient ) condition . in this view , we carried out a direct numerical check of the occurrence of these minima between different bifurcation points . further we present main results of the numerical investigation ; they look different for bh ( @xmath116 ) and ns ( @xmath117 ) . the dependences of boundary radii @xmath118 upon @xmath88 for some fixed values of @xmath84 are shown in figs . [ xk_of_n_bh ] , [ xk_of_n_ns ] . [ xk_of_x0 ] represents the dependence @xmath119 for some fixed values of @xmath88 . \(i ) in case of bh @xmath120 at the horizon @xmath121 ; function @xmath122 is increasing in some neighborhood of @xmath121 . the type of monotony for @xmath123 ( and the occurrence of minima of @xmath122 ) depends on @xmath124 . we studied this using roots @xmath112 of @xmath81 . we show in fig . [ xk_of_n_bh ] how curves @xmath125 are transformed as @xmath84 changes . for sufficiently small @xmath84 a typical situation is represented by the dashed curve @xmath85 of this figure ; here the dependence @xmath85 single - valued . for every @xmath88 there is the only root @xmath85 that defines a lower boundary of stable orbit radii and a corresponding value of the limiting angular momentum @xmath126 . in this case there is the only connected region of stable circular orbits having different angular momenta . this region contains non - relativistic orbits with large @xmath127 and correspondingly large @xmath63 . for larger @xmath84 the dependence @xmath85 can be many - valued . typical picture is represented by the solid curve of fig.[xk_of_n_bh ] @xmath128 for fixed @xmath89 . for @xmath129 we have the only branch @xmath94 . for @xmath130 there are three branches : @xmath131 that are merged together at the points `` b '' and `` c '' . for any fixed @xmath132 these dependencies define two regions of the stable circular orbits : @xmath133 , @xmath134 , and the intermediate region @xmath135 is prohibited ( cf . [ proba-0 ] ) . for @xmath99 there is again the only root @xmath98 . to clarify the situation we show in fig . [ proba1 ] typical examples of effective potentials @xmath67 in case of bh ( @xmath136 ) and their derivatives @xmath137 . the figure illustrates how minima of @xmath67 emerge for different @xmath63 and explicitly demonstrate that two minima of @xmath67 are indeed possible for some @xmath63 . for all @xmath138 , both bounded and unbounded motions are possible . in case of effective potentials of fig . [ proba1 ] we have three critical values of @xmath63 ; there are stable circular orbits with radii @xmath139 $ ] and @xmath140 , @xmath141 . . in the left panel , effective potentials @xmath67 for critical values of the angular momentum : @xmath142 ( dash - dot ) , @xmath143 ( dotted ) , @xmath144 ( dashed ) when these graphs have inflection points . right panel : the derivatives of these potentials ; here the curves in interval @xmath145 ( to the left of the break ) are shown at a smaller scale . for @xmath146 there is no minima of @xmath61 ; for @xmath147 there is one minimum only . for @xmath148 an additional right minimum appears and both minima exist for @xmath149 $ ] , this is illustrated by the solid curves . for @xmath150 the left minimum disappears . here @xmath151 corresponds to minimal radius of stable circular orbits @xmath152 ( see fig . [ proba-0 ] for the notations ) , @xmath153 to @xmath154 , @xmath155 to @xmath156 . , title="fig:",width=302 ] . in the left panel , effective potentials @xmath67 for critical values of the angular momentum : @xmath142 ( dash - dot ) , @xmath143 ( dotted ) , @xmath144 ( dashed ) when these graphs have inflection points . right panel : the derivatives of these potentials ; here the curves in interval @xmath145 ( to the left of the break ) are shown at a smaller scale . for @xmath146 there is no minima of @xmath61 ; for @xmath147 there is one minimum only . for @xmath148 an additional right minimum appears and both minima exist for @xmath149 $ ] , this is illustrated by the solid curves . for @xmath150 the left minimum disappears . here @xmath151 corresponds to minimal radius of stable circular orbits @xmath152 ( see fig . [ proba-0 ] for the notations ) , @xmath153 to @xmath154 , @xmath155 to @xmath156 . , title="fig:",width=302 ] -3 mm \(ii ) in case of ns ( @xmath117 ) there also can be disconnected regions of the stable circular orbits . first of all we note that @xmath157 for @xmath158 is monotonically increasing and taking into account asymptotic relation ( [ eq9 ] ) we see that @xmath67 is decreasing near the center . therefore , in case of ns there is always some minimum of @xmath67 for any @xmath124 . in view of ( [ eq10 ] , [ eq11 ] ) @xmath159 . analogously , @xmath25 always has a minimum at some @xmath160 , which is a radial coordinate of a stationary particle that hangs at rest over the singularity . our analysis shows that minima of @xmath161 for @xmath162 are shifted to larger @xmath127 values . therefore , @xmath160 is the lower boundary of stable circular orbit radii for all @xmath63 . on the other hand , the left root @xmath94 of @xmath81 is unphysical ( it leads to @xmath163 ) . typical dependences @xmath164 in the ns case are presented in fig . [ xk_of_n_ns ] . for @xmath88 less than some @xmath165 this is the only branch of the limiting radii that extends to infinity . starting from @xmath165 additional two branches appear , which can be calculated as roots of @xmath81 ( @xmath166 and @xmath167 ) ; these roots are physical and they indeed define boundaries of the stable orbits region for fixed @xmath113 . as we see , the separated regions of orbits arise not for all parameters of the family . . left panel : effective potentials @xmath67 for two critical values of @xmath63 : @xmath168 ( dotted ) @xmath169 ( dashed ) and for the intermediate value ( solid , @xmath170 ) . right panel : the derivatives . the left minima of @xmath67 exist for angular momenta @xmath171 $ ] . the right minima exist for @xmath172 $ ] . there is a common interval @xmath173 $ ] where two minima exist ; the solid curve represents such an example . correspondingly , for the boundary radii of the circular orbit regions we have @xmath174,title="fig:",width=302 ] . left panel : effective potentials @xmath67 for two critical values of @xmath63 : @xmath168 ( dotted ) @xmath169 ( dashed ) and for the intermediate value ( solid , @xmath170 ) . right panel : the derivatives . the left minima of @xmath67 exist for angular momenta @xmath171 $ ] . the right minima exist for @xmath172 $ ] . there is a common interval @xmath173 $ ] where two minima exist ; the solid curve represents such an example . correspondingly , for the boundary radii of the circular orbit regions we have @xmath174,title="fig:",width=302 ] -3 mm as for dependences @xmath119 plotted in fig . [ xk_of_x0 ] , they also represent qualitatively different cases . for larger @xmath88 there is an upper branch @xmath175 that extends to all @xmath84 , and a continuous lower curve that consists of two parts , which are connected at the point @xmath176 , where there is a discontinuity of the derivative . for @xmath177 this curve is broken into two parts ; the smaller @xmath88 , the farther away is the right upper arm of the curve in the region of large @xmath84 . in the same way as in fig . [ xk_of_n_ns ] , the lower curve in fig . [ xk_of_x0 ] to the right of the break point also represents @xmath178 , not @xmath179 . the other curves in this figure are obtained as the roots @xmath180 of @xmath81 . in fig . [ proba3 ] we show a typical example of effective potentials with two minima in case of ns ( @xmath181 ) , including the case with two minima for some @xmath63 . the radii of the circular orbits that correspond to left and right minima belong to different intervals @xmath182 and @xmath183 ( see fig . [ proba-0 ] for the notations ) . we derived a family of special solutions of the joint system of einstein equations and massless scalar field equations with nonzero self - interaction potentials . we concentrated on the solutions that describe static spherically symmetric configurations having a positive total mass and corresponding asymptotically flat metric at spatial infinity . some of these solutions deal with either ns in the center of the configuration , or with bh . the family includes the schwarzschild metric as limiting case @xmath184 . the main outcome of this paper is that , for the solutions in question , separate ( disconnected ) ring - like structures of stable circular orbits can be formed . analogous feature was first revealed in case of the massless scalar field without self - interaction @xcite . we show that a similar result takes place for a massless nonlinear scalar field in case of the above family of solutions , including the bh case . we remind that the well - known no - hair theorem @xcite does not work , if the scalar field potential is not positively definite ; this is just the case of our solutions . evidently , our results dealing with a special family solutions do not describe the most general situation with the scalar field . nevertheless , the results support the idea that the existence of separated ring - like regions in relativistic objects can be rather a common phenomenon in presence of the scalar field and/or some other fields of alternative gravitation theories . the presence of such ring - like structures in accretion disks around supermassive bhs can testify models of dark energy by means of the fluorescent iron lines ( especially , fe k@xmath83 ) in the x - ray spectra observed in agns @xcite . the appearance of the ring - like structures in astrophysical objects could be detected on account of unusual deformations of the fe k@xmath83 lines . of course , the scalar field effects , if they exist , will be obscured by less exotic ones , e.g. , due to accretion disk warps in agns , inhomogeneous distributions of matter in the disks , reflection of x - rays from dust tori , and possible existence of bh companions ( see , e.g. , @xcite ) . some additional information , that can reduce the number of unknowns in this problem , can be obtained in agns of gravitational lens systems @xcite . on the other hand , there may be deformations in case of another kind of exotic objects @xcite . this stimulates further observations and theoretical investigations of astrophysical compact objects in order to look for a smoking gun of the general relativity modifications . this work has been fulfilled in the framework of the scientific research program `` astronomy and space physics '' of taras shevchenko national university of kyiv . authors acknowledge the partial support of the state foundation for fundamental research in ukraine ( project @xmath18564/45 - 2016 ) . b. novosyadlyi , v. pelykh , yu . shtanov , a. zhuk . dark energy and dark matter of the universe : in three volumes . v. shulga . vol . 1 : dark matter : observational evidence and theoretical models ( kiev , akademperiodyka , 2013 ) . chowdhury , m. patil , d. malafarina , p.s . circular geodesics and accretion disks in the janis - newman - winicour and gamma metric spacetimes . d * 85 * , i d . 104031 ( 2012 ) [ arxiv:1112.2522 ] .

we study test - body orbits in the gravitational field of a static spherically symmetric object in presence of a minimally coupled nonlinear scalar field . we generated a two - parametric family of scalar field potentials , which allow finding solutions of einstein s equations in an analytic form . the results are presented by means of hypergeometric functions ; they describe either a naked singularity ( ns ) or a black hole ( bh ) . our numerical investigation shows that in both cases the stable circular orbits can form separated ( non - connected ) regions around the configuration . we found existence conditions for such separated regions and present examples for some family parameters in case of ns and bh . the results may be of interest for testing models of the dynamical dark energy .

an important way to understand the interactions among particles is via the study of scattering . in related experiments , the initial status of a scattering state is required to be precisely controlled . this is relatively easy for charged incident particles because their initial momentum can be tuned by adjusting the electromagnetic forces imposing on them . for neutral incident particles , the initial momentum is in general difficult to control precisely . however , the recent progress in the technology of trapping neutral atoms via optical trap might open a new way for studying the scattering of neutral particles with precisely controllable initial status @xcite . in this paper a model of a device is proposed and related theoretical calculation is performed to show how the mentioned scattering is realized . it turns out that in this device , as we shall see , the collisions among particles occur regularly and repeatedly . thereby the effect of each individual collision can be accumulated . this would be helpful for the understanding of the very weak interactions among neutral atoms ( molecules ) . traditionally , the scatterings would at most have two incident channels ( say , in the experiments with head - on colliders ) . however , in the following device , three or more incident channels can be realized . as an example , a three - body scattering with three incident channels is chosen to be studied . this is a generalization of a previous work on a two - body scattering in a trap with two incident channels @xcite . it is assumed that , in the beginning , there are three narrow optical traps located at the three vertexes of a regular triangle , and each optical trap provides a harmonic confinement . the total potential is @xmath0 , where @xmath1 points from the origin to the @xmath2 vertexes , and @xmath3 is the mass of a particle . when the center of the triangle is placed at the origin , @xmath4 @xmath5 . it is further assumed that each trap contains an atom in the ground state of a harmonic oscillator , the three atoms are identical bosons with spin zero , and @xmath6 is large enough so that the atoms are well localized initially . suddenly the three narrow traps are cancelled . instead , a broader new trap located at the origin @xmath7 is created , @xmath8 . then , the system begins to evolve . the evolution is affected not only by @xmath9 but also by the atom - atom interaction @xmath10 . in what follows the details of the evolution is studied , three - body head - on collisions occurring repeatedly are found , and the effect of interaction is demonstrated . we shall use @xmath11 and @xmath12 as units of energy and length . the symmetrized and normalized initial state @xmath13 where @xmath14 implies a summation over the permutations @xmath15 , and @xmath16 . without loss of generality , @xmath17 is given lying along the z - axis , while the triangle is given lying on the x - z plane . for convenience , three sets of jacobi coordinates denoted by @xmath18 , @xmath19 , and @xmath20 , respectively , are introduced . the coordinates of the @xmath21 set are defined as , @xmath22 , @xmath23 , and @xmath24 . the other two sets can be obtained from the @xmath18 set by cyclic permutations . in terms of the @xmath18 set , we introduce the harmonic oscillator ( h.o . ) states @xmath25 as basis functions , where @xmath26 , @xmath27 , or @xmath28 . they are normalized eigenstates of the hamiltonian @xmath29 with the eigenenergy @xmath30 and with the angular momentum @xmath31 and its z - component @xmath32 , where @xmath33 , @xmath34 , and @xmath35 when @xmath26 , @xmath27 , and @xmath28 , respectively . then the initial state can be expanded as @xmath36 where the first factor is for the c.m . motion which is completely separated from the internal motion , @xmath37 the notation @xmath38 denotes a set of quantum numbers @xmath39 and @xmath40 , and @xmath41 is the parity , @xmath42 \label{e4_psijmpq}\ ] ] where @xmath31 and @xmath40 are coupled to @xmath43 and @xmath32 . @xmath44\end{aligned}\ ] ] where @xmath45 @xmath46 @xmath47 is the clebsch - gordan coefficients , @xmath48 is for the set @xmath49 , @xmath50 , @xmath51 and @xmath52 to be summed up , @xmath53 and @xmath54 are the spherical polar coordinates of @xmath55 , and so on . @xmath56_{j } \ | \ [ \phi _ { n'l'}^{(1/2)}(\mathbf{r}^{\beta } ) \phi _ { n'l'}^{(2/3)}(\mathbf{r}^{\beta } ) ] _ { j } \ \rangle\ ] ] is the bracket of transformation between the @xmath57 and @xmath58sets ( the superscript @xmath21 is usually ignored ) , which is called the talmi - moshinsky ( t - m ) coefficients . their analytical expression can be found in @xcite . in general , the symmetrization would cause the appearance of all three set of coordinates . however , by using the t - m coefficients , eq . ( [ e2_psii ] ) contains only the @xmath59set so as to facilitate greatly the calculation . incidentally , since @xmath60is symmetrized , @xmath31 ( included in @xmath38 ) of eq . ( [ e2_psii ] ) must be even and therefore @xmath61 . in principle , the right side of eq . ( [ e2_psii ] ) should contain infinite terms . however , the overlap between @xmath60 and higher h.o . states are very small . say , if , @xmath62 , @xmath63 , @xmath64 are all smaller than 12 and @xmath31 , @xmath40 , @xmath65 are all smaller than @xmath66 , with the parameters specified below , the overlap of the right side of eq . ( [ e2_psii ] ) with itself is equal to @xmath67 . this implies that higher h.o . states can be safely ignored . the evolution is governed by the hamiltonian @xmath68 containing @xmath9 and the interaction . when the jacobi coordinates are used it can be separated as @xmath69 where @xmath70 describes the c.m . motion , and @xmath71 describes the internal motion . since the eigenstates of @xmath72 are well known , if the eigenstates of @xmath73 are also know , the evolution starting from any initial state can be understood . in what follows , the symmetrized eigenstates are obtained via a diagonalization of @xmath73 in a limited space . in this way , only approximate solutions can be obtained . then , we increase the dimension of the space until a better convergency is achieved . when @xmath74 as defined in eq . ( [ e4_psijmpq ] ) are used as basis functions ( where @xmath31 is restricted to be even as mentioned ) , the matrix elements of @xmath73 is @xmath75\end{aligned}\ ] ] where @xmath76 denotes the set @xmath77 , and the implication of @xmath78 is alike , @xmath79 to control the size of the space , a number @xmath80 is introduced and @xmath81 is required for all the basis functions . there are two choices to obtain symmetrized eigenstates of @xmath73 . in the first choice , the set @xmath82 is firstly symmetrized and orthonormalized before carrying on the diagonalization . however , this procedure is complicated . therefore we make the second choice , in which the set @xmath83 is simply used without symmetrization but with the requirement that @xmath31 must be even and all the basis functions satisfying @xmath81 are included without missing . this requirement assures that the space is close under permutations , and no basis functions that will contribute to the symmetrized eigenstates would be missed , unless they are too high to have @xmath84 . however , in this choice , a number of unphysical eigenstates with confused symmetry will emerge together with those with correct symmetry . therefore a discrimination is needed as shown below . let an eigenstate be denoted as @xmath85 where @xmath86 is a serial number of the @xmath87series . expanding in terms of the basis functions , @xmath88 where the @xmath31 in @xmath38 must be even , and the coefficients @xmath89 can be directly known from the diagonalization of @xmath73 . if @xmath85 is correctly symmetrized , the coefficients would obey @xmath90 and @xmath91 in the summations of eqs . ( [ e14_baeven ] ) and ( [ e15_baodd ] ) , @xmath31 ( in @xmath38 ) is restricted to be even . with the help of eqs . ( [ e14_baeven ] ) and ( [ e15_baodd ] ) , the states with confused symmetry can be discriminated and dropped , and all the symmetrized eigenstates under the restriction caused by @xmath80 can be extracted without missing . they are one - to - one identical to those obtained via the first choice if the same @xmath80 are used . in what follows @xmath85 denotes only the symmetrized eigenstate , and the associated energy is denoted by @xmath92 . with the eigenstates it is straight forward to obtain the time - dependent solution of @xmath68 as @xmath93 @xmath94 @xmath95 where @xmath96 and @xmath97 . obviously , eqs . ( [ e16_psit ] ) to ( [ e18_psiinrrt ] ) give only an approximate solution because the set @xmath85 obtained via diagonalization each would deviate more or less from the corresponding exact eigenstate , and because only finite number of @xmath85 are used in the expansion . however , it is believed that , when the number of basis functions becomes larger and larger , the above @xmath98 would be closer and closer to the exact solution . the crucial point is the convergency . in this paper the interaction is assumed to be weak . it turns out that , in this case , the convergency is satisfying as shown below . we shall demonstrate that the evolution is a repeating 3-body collisions , and the effect of interaction will be also shown . for these purposes , we extract the following quantities from @xmath98 . \(i ) the density @xmath99 where the integration covers all the degrees of freedom except @xmath100 . therefore @xmath101 obviously , @xmath102 is the probability density that the inter - distance is @xmath103 . \(ii ) the density @xmath104 fulfilling @xmath105 \(iii ) the density @xmath106 fulfilling @xmath107 , where @xmath108 is the angle between @xmath55 and @xmath27 . \(iv ) the density @xmath109 fulfilling @xmath110 . it was found that the c.m . is distributed very close to the origin , therefore @xmath27 is approximately proportional to @xmath111 . therefore , the behavior of the particle 3 can be roughly understood via @xmath109 . incidentally , the behaviors of all the particles are the same due to the symmetrization . the analytical expression of these densities are given in the appendix . in order to have numerical results , as an example , the interaction is assumed to be a repulsive core as @xmath112 if @xmath113 , or zero if @xmath114 , where @xmath115 is a constant to be given . the other parameters are chosen as @xmath116 , @xmath117 , and @xmath118 . to show the initial localization of the particles , we define the one - body density of the initial state as @xmath119 . this density is plotted in fig.1 . starting from @xmath60 , the details of evolution are given as follows . @xmath104 , @xmath102 , and @xmath106 against @xmath120 in the earliest stage of evolution is plotted in fig2a to 2c . where @xmath121 is assumed , and @xmath122 is from @xmath123 to @xmath124 ( say , if @xmath125 , then @xmath120 is from @xmath123 to @xmath126 ) . when @xmath127 , the inter - distances among the particles are about @xmath128 as shown in fig.1 and 2b . when the evolution begins , the peaks of @xmath129 and @xmath130 move inward synchronously , while @xmath131 remains to peak at @xmath132 . it implies a contraction of the regular triangle . accordingly , the three particles rush towards the center leading to a 3-body head - on collision . when @xmath122 is close to @xmath133 , the inter - distances become much shorter , and @xmath131 becomes nearly uniform . it implies that the geometric character ( i.e. , the regular triangle ) will be spoiled when the particles are close to each other . after the first collision , the particles move outward . when @xmath134 , each particle will be close to the opposite end of its initial position as shown by the solid curve of fig.3a . where @xmath109 is given at the point @xmath135 and @xmath122 is given in the interval @xmath136 $ ] . the sharp peak of the solid curve demonstrates clearly that a particle arrives at the opposite end when @xmath137 . due to the symmetrization of the wave function , this is also true for other two particles . then , the above process begins to reverse . when @xmath138 , the system recovers its initial status nearly . if the interaction is removed , the recovery is complete and the system will undergo an exact periodic motion with the period @xmath139 . the motion is characterized by the repeating head - on 3-body collisions . each occurs once within the interval @xmath140 . however , due to the interaction , the recovery is not exact . when the time goes on the effect of the weak interaction will accumulate and gradually emerge . this is shown by the dash - curve of fig.3a which describes the behavior of @xmath141 after @xmath142 rounds of head - on collisions . where the peak is lower than that of the solid curve . it implies that the density is diffusing . furthermore , the peak of the dash - curve has shifted a little right implying that the arrival is a little delayed . the diffusion and the delay will become more explicit when @xmath120 is larger as shown by the dash - dot - dot curve . when the interaction is attractive , the above repeating 3-body collisions remain , and the densities remain to be diffusing . however , instead of a delay , the peak will arrive at the end earlier as shown in fig.3b , where @xmath143 . due to the diffusion of the density as shown in fig.3 , the phenomenon of repeating 3-body collisions will become ambiguous when @xmath120 is sufficiently large . when @xmath115 is given at a number of values and @xmath27 is fixed at @xmath144 , @xmath109 against @xmath122 is plotted in fig.4 , in which @xmath122 varies in an interval close to @xmath145 . in this figure both the locations and the heights of the peaks depend on the strength @xmath115 . in general , a larger @xmath146 will cause a stronger diffusion and therefore a shorter life of the phenomenon of clear repeating 3-body collisions , and a more positive ( negative ) @xmath115 will cause a larger shift of the peak to the right ( left ) . in addition to @xmath115 , the evolution depends also on the range of interaction . therefore , by observing the time - dependent densities , information on the interaction can be obtained . the accuracy of the above numerical results depends on @xmath80 , the number controlling the total number of basis functions . as an example the values of @xmath141 are listed in table 1 to show the dependence . .the values of @xmath109 with @xmath147 and with three choices of @xmath80 . the other parameters are the same as those for fig.2 [ cols="<,<,<,<",options="header " , ] the convergency shown in this table is satisfying . thus we conclude that the numerical results obtained by using @xmath118 are accurate enough in qualitative sense . a model of a device containing repeating 3-body collisions is proposed to study the weak interactions among neutral atoms and molecules . the advantage is twofold . ( i ) the initial status can be precisely controlled . ( ii ) the weak effect of the interaction can be accumulated and therefore easier to be detected . numerical results support that information on interaction can be thereby extracted . a crucial point in the device is the initial localization of the particles . if they are better localized ( by increasing @xmath6 ) and/or they are more separated from each other initially ( by increasing @xmath148 ) , the repeating 3-body collisions would become more explicit and be maintained longer . furthermore , if @xmath68 has a larger @xmath149 , the interval between two successive collisions ( @xmath150 ) becomes shorter , therefore the evolution would proceed swifter . although the three particles are assumed to be identical bosons , a generalization to fermions and/or distinguishable particles is straight forward . in principle , the above device could also be used to study the three - body forces if they exist . the support from the nsfc under the grant 10874249 is appreciated . the analytical expressions of the time - dependent densities are as follows . @xmath151^{\ast } \ \mathcal{d}_{q}^{jm\pi } ( t ) \nonumber \\ & & \times r^{2 } f_{n'l}^{(2/3)}(r ) f_{nl}^{(2/3)}(r)\end{aligned}\ ] ] @xmath152^{\ast } \ \mathcal{d}_{q}^{jm\pi } ( t ) \nonumber \\ & & \times r^{2 } f_{n'l}^{(1/2)}(r ) f_{nl}^{(1/2)}(r)\end{aligned}\ ] ] ( iii ) for the derivation of @xmath106 , a transformation is made so that @xmath55 and @xmath27 are transformed to @xmath153 and the three euler angles specifying the orientation of the triangle formed by the three particles . then , integrating all the degrees of freedom except @xmath154 , we have @xmath155^{\ast } \ \mathcal{d}_{q}^{jm\pi } ( t)\ ( -1)^{l+l'-j } \nonumber \\ & & \times [ ( 2l^{\prime } + 1)(2l+1)(2l^{\prime } + 1)(2l+1)]^{1/2 } \ [ \int r^{2}dr\ f_{n'l'}^{(1/2)}(r)\ f_{nl}^{(1/2)}(r ) ] \nonumber \\ & & \times [ \int r^{2}dr\ f_{n'l'}^{(2/3)}(r)\ f_{nl}^{(2/3)}(r ) ] \nonumber \\ & & \times \sum_{\lambda } w(ll'll';\lambda j ) \sqrt{\frac{\pi}{2\lambda+1 } } c_{l0,l'0}^{\lambda 0 } c_{l0,l'0}^{\lambda 0 } y_{\lambda 0}(\theta , 0)\end{aligned}\ ] ] @xmath156^{\ast } \ \mathcal{d}_{q}^{jm\pi } ( t ) \nonumber \\ & & \times \sum_{k } [ c_{l'k , l'm'-k}^{j'm ' } c_{lk , lm - k}^{jm } y_{l',m'-k}^{\ast}(\widehat{r } ) y_{l , m - k}(\widehat{r } ) ] \nonumber \\ & & \times f_{n'l'}^{(2/3)}(r)\ f_{nl}^{(2/3)}(r)\end{aligned}\ ] ]

a model of a device is proposed and related theoretical calculation is performed to study the weak interactions among neutral atoms and molecules . in this model 3-body collisions among the neutral particles occur repeatedly in a trap . results of calculation demonstrate that information on interaction can be obtained by observing the time - dependent densities of the system .

the higgs mechanism @xcite provides a general framework to explain the observed masses of the w and z gauge bosons through electroweak symmetry breaking . within the standard model ( sm ) , this mechanism posits the existence of a scalar boson , the higgs boson , with an a priori unknown mass ( @xmath2 ) . the direct searches for the sm higgs boson at the lep experiments have excluded @xmath6 gev @xcite , while the searches at the tevatron exclude @xmath7 gev @xcite . however , a global fit of the electroweak measurements performed at lep , sld and the tevatron experiments , predicts a sm higgs boson mass of @xmath8 gev . the sm higgs searches in atlas have been performed over a wide range of higgs boson masses ( 110 - 600 gev ) considering different sm production mechanisms ( gluon fusion , vector boson fusion and vector boson associated production ) and their subsequent decay modes . the detector resolution for the reconstructed higgs boson mass plays a crucial role in classifying the searches into numerous channels having different selection criteria as detailed below . the sm higgs boson signal events have been simulated using powheg and pythia generators , while the background contributions have been estimated either using simulation , or directly from data as appropriate . searches in these channels are focused on high mass higgs boson searches over a typical higgs mass range of 200 - 600 gev . in the zz channels the events are required to have a lepton pair ( @xmath12 or @xmath13 ) with reconstructed @xmath14 close to the z boson mass . the selected events for the @xmath15 @xcite search are further classified into two subcategories considering the pile - up effects ( ` low ' and ` high ' ) on the reconstructed missing transverse energies ( @xmath16 ) , while the @xmath17 @xcite search considers ` tagged ' ( 2 b - tagged jets ) and ` untagged ' ( @xmath18 b - tagged jets ) events separately . figures [ fig : searchchannels ] ( a ) and [ fig : searchchannels ] ( b ) show the transverse mass distribution for @xmath15 candidates in the ` low ' pile - up data , and @xmath19 distribution for @xmath17 in ` tagged ' selection respectively . the selection criteria for the @xmath20 search @xcite are optimized over a higgs boson mass range of 300 - 600 gev considering ` h+0jet ' , ` h+1jet ' and ` h+2jets ' ( mostly from the vector boson fusion processes ) . the search in this channel @xcite consists of event categories with different lepton flavor combinations , while the sm @xmath22 production processes remain irreducible background at the final level of event selection . full reconstruction of the higgs boson mass is possible for this channel with excellent mass resolution ( 2% and 1.5% for 4e and 4@xmath23 at @xmath24 gev ) . the 4-lepton invariant mass ( @xmath25 ) distribution is displayed in figure [ fig : searchchannels ] ( c ) . although @xmath27 decays have small branching ratio ( about 0.2% ) , search in this channel @xcite has the potential to discover the higgs boson in the low mass range ( 110 - 150 gev ) . the analysis is split into nine independent sub - channels based on the photon pseudorapidity , conversion status , and the momentum component of the diphoton system transverse to the thrust axis ( @xmath28 ) . the background distribution is obtained by fitting the @xmath29 distribution in data with a smoothly falling exponential function , while the atlas @xmath29 mass resolution is approximately 1.4% for @xmath30 gev . the diphoton invariant mass distribution is shown in figure [ fig : searchchannels ] ( d ) . [ cols="^,^,^ " , ] the searches in this channel @xcite cover a wide range of @xmath4 gev consisting of different number of jets ( 0 , 1 and 2 jets ) and lepton flavor combinations ( ee , @xmath32 and @xmath33 ) in the final states . the reconstructed ww transverse mass ( @xmath34 ) has been utilized in this analysis as a discriminating variable as shown in figure [ fig : searchchannels](e ) for the ` h+0jet ' sub - channel . the searches in @xmath36 channels @xcite are optimized for @xmath2 of 110 - 130 gev . the main advantage with these channels is due to the large @xmath37 at low @xmath2 , along with the possibility to fully reconstruct the higgs boson mass through @xmath38 . the final states with one or two charged leptons are sub - categorized depending on the transverse momentum of the reconstructed vector boson , and the lepton flavors . figure [ fig : searchchannels ] ( f ) shows the @xmath38 distribution for the @xmath39 sub - channel . @xmath43 searches @xcite consist of all possible leptonic and hadronic ( @xmath44 ) decay modes of the @xmath45-leptons originating from the higgs boson decay . for the @xmath46 and @xmath47 channels , invariant mass of the @xmath48 system is reconstructed assuming collinear approximation . the searches in @xmath49 channel reconstruct the @xmath50 using the missing mass calculator techniques @xcite where the full event topology is reconstructed using the kinematics of the @xmath45-lepton decay products . the results from the aforesaid search channels have been utilized to set an upper limit on the sm higgs boson production cross section as a function of the @xmath2 . the limits are conveniently expressed in terms of the signal strength ( @xmath51 ) , the ratio of a given higgs boson production cross section ( @xmath52 ) to its sm predicted value ( @xmath53 ) . the @xmath54 prescription@xcite with a profile likelihood ratio test statistic , @xmath55 @xcite has been utilized here to derive the exclusion limits . figure [ fig : combination ] shows the exclusion limits on @xmath23 at 95% confidence level ( cl ) for individual search channels , along with the combined ones from all search channels over the @xmath2 range of 110 - 150 gev . a small excess of events near @xmath3 gev is observed in @xmath26 and @xmath21 search channels , both of which fully reconstruct the higgs boson mass with high resolution . range of 110 - 150 gev ( right).,title="fig:",width=192,height=144 ] range of 110 - 150 gev ( right).,title="fig:",width=192,height=144 ] atlas has performed extensive searches for the sm higgs boson utilizing the full 4.7 - 4.9 @xmath0 dataset recorded during the 2011 lhc operations . the atlas combination @xcite from numerous search channels excludes the sm higgs boson mass in ranges of 110.0 - 117.5 gev , 118.5 - 122.5 gev , and 129 - 539 gev at the 95% cl , while an exclusion of 120 gev@xmath56555 gev is expected in the absence of higgs signal . furthermore , the exclusion limits have been recalculated at 99% cl and the sm higgs boson over a mass range of 130 - 486 gev has been excluded . the excess in the observed data has a local significance of 2.5@xmath57 , where the expected significance in the presence of a sm higgs boson with @xmath58 gev is 2.9@xmath52 . the global probability for such an excess to occur across the entire sm higgs mass range ( 110 - 600 gev ) is estimated to be 30% . however , the said probability reduces to nearly 10% if a 110 - 146 gev mass range for the sm higgs boson is considered . 99 p. higgs , phys . * 12 * 132 ( 1964 ) ; p. higgs , phys . * 13 * 508 ( 1964 ) ; p. higgs , phys . rev . * 145 * 1156 ( 1966 ) ; f. englert and r. brout , phys . lett . * 13 * 321 ( 1964 ) . the lep working group for higgs boson searches , phys . b * 565 * 61 ( 2003 ) . the tevnph working group , arxiv:1107.5518v2 [ hep - ex ] ( 2011 ) . atlas collaboration , atlas - conf-2012 - 016 , http://cdsweb.cern.ch/record/1429665 . atlas collaboration , atlas - conf-2012 - 017 , http://cdsweb.cern.ch/record/1429666 . atlas collaboration , atlas - conf-2012 - 018 , http://cdsweb.cern.ch/record/1429667 . atlas collaboration , phys . b * 710 * 383 ( 2012 ) . atlas collaboration , phys . * 108 * 111803 ( 2012 ) . atlas collaboration , atlas - conf-2012 - 012 , http://cdsweb.cern.ch/record/1429660 . atlas collaboration , atlas - conf-2012 - 015 , http://cdsweb.cern.ch/record/1429664 . atlas collaboration , atlas - conf-2012 - 014 , http://cdsweb.cern.ch/record/1429662 . a. elagin _ et al . _ , a654 * 481 ( 2011 ) , arxiv:1012.4686 [ hep - ex ] . a. read , j. phys . * g28 * 2693 ( 2002 ) . g. cowan _ et al . _ , j. * c71 * 1554 ( 2011 ) . atlas collaboration , atlas - conf-2012 - 019 , http://cdsweb.cern.ch/record/1430033 .

the search for the standard model ( sm ) higgs boson based on 4.7 - 4.9 @xmath0 of pp collision data at @xmath1 tev recorded with the atlas detector is presented here . the combined atlas results exclude the sm higgs boson masses ( @xmath2 ) of 110.0 - 117.5 , 118.5 - 122.5 and 129 - 539 gev at 95% confidence level . an excess of events has been observed around @xmath3 gev with the global probability of 30% ( 10% ) to occur anywhere in @xmath4 ( @xmath5 ) gev .

the design and management of green wireless networks @xcite has become increasingly important for modern wireless networks , in particular , to manage operating costs . futuristic ( beyond 5 g ) cellular networks face the dual challenges of being able to respond to the explosion of data rates and also to manage network energy consumption . due to the limited spectrum and large number of active users in modern networks , energy - efficient distributed power control is an important issue . sensor networks , which have multiple sensors sending information to a common receiver with a limited energy , capacity have also recently surged in popularity . energy minimization in sensor networks has been analysed in many recent works @xcite . + several of the above described systems have some common features : 1 . multiple transmitters connected to a common receiver . 2 . lack of centralization or coordination , i.e. , a distributed and de - centralized network . 3 . relevance of minimizing energy consumption or maximizing energy - efficiency ( ee ) . transmitters that have arbitrary data transmission . these features are present in many modern systems like a sensor network which has multiple sensors with limited energy connected in a distributed manner to a common receiver . these sensors do nt always have information to transmit , resulting in sporadic data transmission . another example would be several mobile devices connected to a hot - spot ( via wifi or even bluetooth ) . due to these features of the network , inter - transmitter communication is not possible and the transmitters are independent decision makers . therefore , implementing frequency or time division multiple access becomes harder and a mac protocol ( with single carrier ) is often the preferred or natural method of channel access . in many existing works , both network - centric and user - centric approaches have been studied . in a network - centric approach , the global energy - efficiency ( gee ) is defined as the ratio between the system benefit ( sum - throughput or sum - rate ) over the total cost in terms of consumed power @xcite . however , when targeting an efficient solution in an user - centric problem , the gee becomes not ideal as it has no significance to any of the decision makers . in this case , other metrics are required to reflect the individual interest of each decision maker . therefore , we redefine the gee to be the sum over individual energy - efficiencies as a suitable metric of interest @xcite . + the major novelty of this work is in improving the * sum of energy - efficiencies * for a communication system with * all the listed features above*. in such a * decentralized and distributed network * , as each transmitter operates independently , implementing a frequency division or a time division multiple access is not trivial . therefore , we are interested in looking at a * mac system * where all transmitters operate on the same band . additionally , * ee * will be our preferred metric due to its relevance . this metric has been defined in @xcite as the ratio between the average net data rate and the transmitted power . in @xcite , the total power consumed by the transmitter was taken into account in the ee expression to design distributed power control which is one of the most well known techniques for improving ee . however , many of the works available on energy - efficient power control consider the ee defined in @xcite where the possible presence of a queue at the transmitter is ignored . in contrast with the existing works , we consider a new generalized ee based on a cross - layer approach developed recently in @xcite . this approach is important since it takes into account : 1 ) a fixed cost in terms of power namely , a cost which does not depend on the radiated power ; and 2 ) * the presence of a finite packet buffer and sporadic packet arrival * at the transmitter ( which corresponds to including the 4th feature mentioned above ) . although providing a more general model , the distributed system in @xcite may operate at a point which is energy - inefficient . indeed , the point at which the system operates is a nash equilibrium ( ne ) of a certain non - cooperative static game . the present work aims at filling this gap by not only considering a cross - layer approach of energy - efficient power control but also improving the system performance in terms of sum of energy - efficiencies . nash bargaining ( nb ) solution in a cooperative game can provide a possible efficient solution concept for the problem of interest as it is pareto - efficient . however , it generally requires global channel state information ( csi ) @xcite . therefore , we are interested in improving the average performance of the system by considering long - term utilities . we focus then on repeated games ( rg ) where repetition allows efficient equilibrium points to be implemented . unlike static games which are played in one shot , rg are a special case of dynamic games which consider a cooperation plan and consist in repeating at each step the same static game and the utilities result from averaging the static game utilities over time @xcite . there are two relevant dynamic rg models : finite ( frg ) and discounted ( drg ) . the frg is defined when the number of stages during which the players interact is finite . for the drg model , the discount factor is seen as the stopping probability at each stage @xcite . the power control problem using the classic ee developed by goodman et al in @xcite has been solved with rg only in @xcite where authors developed an operating point ( op ) relying on individual csi and showed that rg lead to efficient distributed solution . here , we investigate the power control problem of a mac system by referring to rg ( finite and discounted ) where the utility function is based on a cross - layer approach . accordingly , we contribute to : 1 . determine the closed - form expressions of the minimum number of stages for the frg and the maximum discount factor for the drg . these two parameters identify the two considered rg . 2 . determine a distributed solution pareto - dominating the ne and improving the system performances in terms of powers and utilities compared not only to the ne but also to the nb solution even for high number of users . 3 . show that the rg formulation when using the new ee and the new op leads to significant gains in terms of social welfare ( sum of utilities of all the users ) compared to the ne . 4 . show that the following aspects of the cross - layer model improve considerably the system performances when comparing to the goodman model even for large number of users : * the minimum number of stages in the cross - layer ee model can always be shorter than the minimum number of stages in the goodman ee formulation . * the social welfare for the drg in the cross - layer model decreases slightly when the number of users increases while it decreases considerably in the goodman model . 5 . show that in real systems with random packet arrivals , the cross - layer power control algorithm outperforms the goodman algorithm and then the new op with the cross - layer approach is more efficient . this paper is structured as follows . in section [ sec : problemstatement ] , we define the system model under study , introduce the generalized ee metric and define the non - cooperative static game . this is followed ( section [ sec : nashbargainingsolution ] ) by the study of the nb solution . in section [ sec : repeatedgamesformulation ] , we introduce the new op , give the formulation of both rg models ( frg and drg ) and determine the closed - form expressions of the minimum number of stages and the maximum discount factor as well . numerical results are presented in section [ sec : numericalresults ] and finally we draw several concluding remarks . we consider a mac system composed of @xmath0 small transmitters communicating with a receiver . the @xmath1 transmitter transmits a signal @xmath2 with a power @xmath3 $ ] where @xmath4 is the maximum transmit power assumed identical for all users ( @xmath5 ) . the additive noise , which is the same for all users , is an additive white gaussian noise denoted as @xmath6 with zero mean and variance @xmath7 . we assume that the users transmit their data over block fading channels . the channel gain between user @xmath8 and the receiver is given by @xmath9 . thus , the baseband signal received at the receiver is written as : @xmath10 therefore , the resulting sinr @xmath11 corresponding to the @xmath1 transmitter is given by @xcite : @xmath12 where @xmath13 defines the power vector of all users and can be written as @xmath14 with @xmath15 . the purpose of this work is to determine how each user is going to control its power in an optimum way . game theory , as a powerful mathematical tool , helps to solve such an optimization problem where the utility function is the ee which is a function of the users powers . since the system under study has multiple agencies each with individual interest , the sum over individual energy - efficiencies will be considered as the performance metric reflecting the individual interest of each decision maker . the ee is defined in @xcite as a ratio of the net data rate to the transmit power level and is given by : @xmath16 where @xmath17 is the transmission rate ( in bit / s ) while @xmath18 $ ] denotes the efficiency function which is sigmoidal and corresponds to the packet success rate verifying @xmath19 and @xmath20 . authors of @xcite were the first to consider a total transmission cost of the type _ radiated power _ ( @xmath21 ) @xmath22 _ consumed power _ ( @xmath23 ) to design distributed power control strategies for multiple access channels @xcite as follows : @xmath24 in @xcite , a more generalized ee metric has been developed by considering a packet arrival process following a bernoulli process with a constant probability @xmath25 and a finite memory buffer of size @xmath26 . the new ee expression is given by : @xmath27 where the function @xmath28 identifies the packet loss due to both bad channel conditions and the finiteness of the packet buffer and is expressed as follows : @xmath29 where @xmath30 is the stationary probability that the buffer is full and is given by : @xmath31 with : @xmath32 it is important to highlight that this new generalized ee given by ( [ eqeegen ] ) includes the conventional case of ( [ eqchi ] ) when making @xmath33 . + the static cross - layer power control game is a non - cooperative game which can be defined as a strategic form game @xcite . _ the game is defined by the ordered triplet @xmath34 where @xmath35 is the set of players ( the @xmath0 transmitters ) , @xmath36 are the corresponding sets of strategies with @xmath37 $ ] and @xmath38 are the utility functions given by : _ @xmath39 _ where @xmath40 is given by equation ( [ eqeegen ] ) . _ in a non - cooperative game , each user ( player ) seeks to maximize selfishly its individual utility function . the optimum solution results then by setting @xmath41 to zero as follows : @xmath42=0 , \label{9}\ ] ] where @xmath43 , @xmath44 and @xmath45 . + authors in @xcite proved that such equation has a unique best response . in the game @xmath46 , this best response defines the ne and is denoted as @xmath47 . however , the ne solution is not always pareto - efficient for many scenarios . we highlight in fig . [ fig : achievableregion ] that the ne is not on the pareto frontier ( the outer boundary of the achievable utilities region ) . therefore , we are motivated to design a more efficient solution than the ne . for this , as a first step we investigate the nb solution . due to the inefficiency of the ne , a pareto - efficient solution can be achieved by introducing the cooperation between the players . the resulting solution is called nb solution whose determination requires two elements @xcite : * the region of achievable utilities formed by the set of the feasible utilities of all the players should be compact and convex @xcite ; * the threat point is defined by the ne of the one - shot game @xcite . we denote @xmath48 the achievable utilities region defined as follows : @xmath49^n\}. \label{19}\ ] ] as the strategies sets @xmath36 are compact since @xmath50 $ ] and the utility function @xmath51 is continuous , the region @xmath48 is compact for a given channel configuration @xcite . since it is generally not convex , time - sharing has been a solution to convexify it . in order to illustrate the main idea of this technique applied to our problem , let us consider a system of 2 users @xcite . during a time fraction @xmath52 , the users use the powers @xmath53 to have utilities @xmath54 . during a time fraction @xmath55 , they use another combination of powers @xmath56 to have @xmath57 @xcite . thus , the new achievable utilities region ( for the 2-users system ) is : @xmath58 we define @xmath59 the pareto boundary ( the outer frontier ) of the convex hull of @xmath60 . [ fig : achievableregion ] shows the convexified achievable utilities region with the ne point , the nb solution and the nash curve ( both will be defined next ) . let @xmath61 define the improvement region of utilities versus the ne and it is given by : @xmath62\}.\ ] ] the nb solution belongs to the region @xmath61 . here , in the power control game @xmath46 , there exists a unique nb solution denoted as @xmath63 and is given by @xcite : @xmath64 } } \prod_{i=1}^{n}{(u_{i}-u_{i}^{ne } ) } , \label{30}\ ] ] since the ne can always be reached and the achievable utility region is a compact convex set , the nb solution exists . it is unique since it verifies certain axioms : individual rationality and feasibility , independence of irrelevant alternatives , symmetry , pareto optimality ( efficiency ) and independence of linear transformations @xcite . the nb solution results from the intersection of the pareto boundary ( @xmath59 ) with the nash curve whose form is @xmath65 where @xmath66 is a constant chosen such that there is precisely one intersection point @xcite ( see fig . [ fig : achievableregion ] ) . although the nb solution is pareto - efficient , it generally requires global csi at the transmitters due to the nash product @xmath67 introducing all the users utilities @xcite . for this reason , we are looking for another efficient solution through the study of the dynamic rg . rg consist in their standard formulation , in repeating the same static game at every time instance and the players seek to maximize their utility averaged over the whole game duration @xcite . repetition allows efficient equilibrium points to be implemented and which can be predicted from the one - shot static game according to the folk theorem , which provides the set of possible nash equilibria of the repeated game @xcite . in a repeated game , certain agreements between players on a common cooperation plan and a punishment policy can be implemented to punish the deviators @xcite . in what follows , we introduce the new op and characterize the two rg models . the new op consists in setting @xmath68 to a constant @xmath69 which is unique when maximizing the expected sum utility over all the channel states . it is given by @xcite : @xmath70 . \label{14}\ ] ] the power of the @xmath1 player is then deduced as follows : @xmath71 the new op pareto - dominates the ne and relies on individual csi at the transmitter . in order to implement a cooperation plan between the players , we assume in addition to the individual csi assumption , that every player is able to know the power of the received signal at each game stage , which is denoted by @xcite : @xmath72 when assuming that @xmath73 is set to the constant @xmath69 , the received signal power can be written as : @xmath74 accordingly , each transmitter needs only its individual sinr and the constant @xmath69 ( depending only on @xmath21 and @xmath75 ) to establish the received signal power @xmath76 . we assume that the data transmission is over block fading channels and that channel gains @xmath77 lie in a compact set @xmath78 $ ] @xcite . thus , the interval to which the received signal power belongs , is @xmath79}$ ] . since the players detect a variation of the received signal power , a deviation from the cooperation plan has occurred . indeed , when playing at the new op , the received signal power is constant and equal to @xmath80 . consequently , when any player deviates from the new op , the latter quantity changes and the deviation is then detected @xcite . a rg is a long - term interaction game where players react to past experience by taking into account what happened in all previous stages and make decisions about their future choices @xcite . the resulting payoff is an average over all the stage payoffs . we denote by @xmath81 , the game stage which corresponds to the instant in which all players choose their actions . accordingly , a profile of actions can be defined for all players as @xmath82 . a history @xmath83 of player @xmath8 at time @xmath81 is the pair of vectors @xmath84 and which lies in the set @xmath85 with @xmath86 = [ 0,p^{\max}]$ ] ( as all the users have the same maximum power ) @xcite . histories are fundamental in rg as they allow players to coordinate their behavior at each stage so that previous histories are known by all the players @xcite . we denote @xmath87 the pure strategy of the @xmath1 player . it defines the action to select after each history @xcite : @xmath88\\ \mathbf{h}(t ) & \mapsto & p_i(t ) \end{array } \right . \label{8}\ ] ] in rg literature , there are two important models @xcite : * the finite rg where the number of stages of the game ( denoted as @xmath89 ) during which the players interact is finite ; * the discounted rg where the discount factor ( denoted as @xmath900,1[$ ] ) is seen as the stopping probability at each stage . the utility function of each player results from averaging over the instantaneous utilities over all the game stages in the frg while it is a geometric average of the instantaneous utilities during the game stages in the drg @xcite . we denote @xmath91 the joint strategy of all players . _ a joint strategy @xmath92 satisfies the equilibrium condition for the repeated game defined by @xmath93 if @xmath94 , @xmath95 , @xmath96 with @xmath97 for the frg or @xmath98 for the drg such that : _ @xmath99 in rg with complete information and full monitoring , the folk theorem characterizes the set of possible equilibrium utilities . it ensures that the set of ne in a rg is precisely the set of feasible and individually rational outcomes of the one - shot game @xcite . a cooperation / punishment plan is established between the players before playing @xcite . the players cooperate by always transmitting at the new op with powers @xmath100 . when the power of the received signal changes , a deviation is then detected and the players punish the deviator by transmitting with their maximum transmit power @xmath101 in the frg and by playing at the one - shot game in the drg . in what follows , we give the equilibrium solution of each repeated game model and mention the corresponding algorithm @xcite . it is important to note that in contrast with iterative algorithms ( e.g. , iterative water - filling type algorithms ) , there is no convergence problem in repeated games ( frg and drg ) . indeed , the transmitters implement an equilibrium strategy ( referred to as the operating point ) at every stage of the repeated game . the frg is characterized by the minimum number of stages ( @xmath102 ) . if the number of stages in the game @xmath103 verifies @xmath104 , a more efficient equilibrium point can be reached . however , if it is less than @xmath102 , the ne is then played . assuming that channel gains @xmath77 lie in a compact set @xmath78 $ ] @xcite , we have the following proposition @xcite : : when supposing the following condition is met : @xmath105 with : @xmath106 such that : @xmath107 then , the ne corresponding to the @xmath103-stage frg is given by the following action plan for any @xmath108 and @xmath109 : @xmath110 the quantities @xmath111 , @xmath112 , @xmath113 , @xmath114 , @xmath115 , @xmath116 , @xmath117 and @xmath118 are defined in app . a and @xmath119 is the sinr at the ne while @xmath120 and @xmath121 are the sinrs related to the maximal utility and the utility min - max respectively ( the proof of this proposition is detailed in @xcite ) . the corresponding algorithm is as follows . + 1l + @xmath122 each user transmits at the new op with power @xmath100 during the first phase + of the game @xmath123 . + @xmath124 in the second phase @xmath125 , each user plays the ne . + as the frg has a finite number of stages , this phase ensures the + punishment of the deviator for two reasons @xcite : + @xmath126 if it deviates at the last stage , it can not therefore be punished ; + @xmath126 if it deviates earlier , the punishment can be not sufficiently severe . + @xmath127 the power of the received signal is assumed to be constant during the + first phase . when it changes , a deviation is then detected . + @xmath128 the deviator is punished by other transmitters by playing at their + maximum transmit power @xmath101 . + in the drg , the probability that the game stops at stage @xmath81 is @xmath129 with @xmath900,1[$ ] defines the discount factor @xcite . accordingly , we can express the analytic form of the maximum discount factor in a drg when assuming that channel gains @xmath77 lie in a compact set @xmath78 $ ] @xcite . : when assuming the following condition is met : @xmath130 with : @xmath131 then , the ne corresponding to the drg is given by the following action plan @xmath109 : @xmath132 for the proof , see app . a. the corresponding algorithm is as follows . + 1l + @xmath122 each user transmits at the new op with power @xmath100 . + @xmath124 when the power of the received signal changes , a deviation is detected . + @xmath127 the other transmitters punish the deviator by transmitting at the one - shot + game with power @xmath133 . in this section , we consider the efficiency function @xmath134 with @xmath135 . it has be proven in @xcite that such a function is sigmoidal as it is convex on the open interval @xmath136 $ ] and concave on @xmath137 . the throughput @xmath17 and the used bandwidth @xmath138 are equal to 1 mbps and 1 mhz respectively . the maximum power @xmath139 is set to 0.1 watt while the noise variance is set to @xmath140 watt . the buffer size @xmath26 , the packet arrival rate @xmath25 and the consumed power @xmath23 are fixed to 10 , 0.5 and @xmath141 watt respectively . we consider rayleigh fading channels and a spreading factor @xmath142 introducing an interference processing ( @xmath143 ) in the interference term of the sinr . in fig . [ reg2 ] , we present the achievable utility region , the new op , the ne and the nb solution . we stress that the new op and the nb solution dominate both the ne in the sense of pareto . the region between the pareto frontier and the min - max level is the possible set of equilibrium utilities of the rg according to the folk theorem . ) . ] in order to study the efficiency of the new op versus the nb solution and the ne , we are interested in comparing powers and utilities of the three equilibria by averaging over channel gains for different scenarios ( different number of users @xmath0 in the system ) . in fig . [ reg4 ] , we plot the power and the utility that a user ( in a system of @xmath0 users ) can reach for each equilibrium . thus , we highlight that the new op and the nb solution have better performances than the ne as they pareto - dominate it . when @xmath144 , we notice that the new op and the nb solution are more efficient than the ne . it is clear that the nb solution requires less power and provides higher utility compared to the new op , but it is important to stress that values , in terms of powers and utilities , are slightly different for both equilibria ( new op and nb solution ) . when @xmath145 , we highlight that lower powers are provided with the new op which leads also to higher values of the utilities . thus , we notice that the new op gives better performances than the ne and the nb solution . therefore , the new op contributes not only to improve the system performances better than the ne for any given scenario but also enables important gains in terms of powers and utilities when compared to the nb solution for a system with a large number of users ( @xmath145 ) . . ] we are interested in studying the performances of the social welfare ( @xmath146 ) according to the frg versus the ne in a multi - users system . the corresponding expression is given by : @xmath147 in fig . [ reg5 ] , we present the ratio of the social welfare corresponding to the frg ( @xmath148 ) vs the ne social welfare ( @xmath149 ) . we proceed by averaging over channel gains lying in a compact set such that @xmath150 . we highlight that the social welfare of the frg reaches higher values than the ne ( @xmath151 ) . in addition , we notice that the social welfare ratio increases with the number of users for both models ( goodman and cross - layer ) . the minimum number of stages @xmath102 according to the cross - layer model is much lower compared to the one related to the goodman model . to illustrate this , when @xmath152 , @xmath102 for the goodman model is equal to 4600 while it is 3700 for the cross - layer model . this difference becomes considerable with the increase of the number of users . indeed , when @xmath153 , the minimum number of stages for the goodman ee is 14300 while it is equal to 10900 for the cross - layer approach . ( @xmath154 ) . ] we are interested in plotting the minimum number of stages as a function of the consumed power @xmath23 and the packet arrival rate @xmath25 according to both ee models . results , obtained by averaging over channel realizations , are drawn in figures [ reg6 ] and [ reg11 ] . according to fig . [ reg6 ] , we stress that @xmath102 increases with the number of users while it decreases with the spreading factor . it is clear that for any values of @xmath0 and @xmath142 , it exists a consumed power @xmath155 for which @xmath102 is less than @xmath102 when @xmath156 . thus , a good choice of the fixed consumed power leads to a lower minimum number of stages for the cross - layer model compared to the goodman model . for the cross - layer model ( @xmath155 ) lower than @xmath102 of goodman model ( @xmath156 ) . ] in fig . [ reg11 ] , we highlight that the minimum number of stages is an increasing function of the packet arrival rate @xmath25 according to the cross - layer model while it is a constant function for the goodman model since the latter does not take into account the packet arrival process . one can confirm that the minimum number of stages is an increase function of the number of users as deduced previously . simulations show that it exists a packet arrival rate @xmath157 before which @xmath102 of the cross - layer model is much lower than @xmath102 of the goodman model for different number of users . simulations show that @xmath158 and for @xmath159 , @xmath102 of the cross - layer model converges to @xmath102 corresponding to the goodman model . it is important to highlight that when @xmath152 and @xmath159 , @xmath102 of the cross - layer model takes higher values than @xmath102 corresponding to the goodman model but values are quite similar . with the increase of the number of users , the difference between the minimum number of stages for both models becomes noticeable . according to figures [ reg6 ] and [ reg11 ] , one can conclude that the cross - layer model can be exploited for short games . of the cross - layer model when comparing to goodman model ( @xmath154 ) . ] for the drg model , we plot in a first step the improvement of the social welfare ( @xmath160 ) versus the one - shot game ( @xmath149 ) for goodman and cross - layer models ( @xmath156 and @xmath161 respectively ) as a function of the spectral efficiency @xmath162 . we simulated our algorithm by averaging over channel gains for different number of users . results are given in fig . it is important to highlight that the drg social welfare reaches higher values than the ne social welfare ( @xmath163 ) . for low values of the spectral efficiency , the social welfare ratio is quite similar for both models while the difference becomes noticeable when the spectral efficiency takes higher values . the social welfare ratio increases with the number of users for both ee models . for each model , when @xmath0 takes high values , the social welfare ratios become closer ( for the cross - layer model , the curves corresponding to @xmath152 and @xmath153 are closer than with the curve of @xmath144 ) . for different number of users @xmath0 . ] for this reason , we studied the variation of @xmath164 as a function of @xmath165 and @xmath25 for both ee models and for different number of users . results are given in figures [ reg8 ] and [ reg10 ] . according to fig . [ reg8 ] , we deduce how @xmath164 decreases with the number of users for both ee models . in addition , we stress that the values reached by @xmath164 becomes closer when @xmath0 takes higher values . this can explain fig . [ reg7 ] . for goodman and cross - layer models as a function of the spectral efficiency @xmath165 with different number of users @xmath0 . ] the study of the variation of @xmath164 versus the packet arrival rate @xmath25 ( in fig . [ reg10 ] ) shows that the maximum discount factor @xmath164 decreases with the number of users and with the packet arrival rate @xmath25 as well . simulations show that it exists a packet arrival rate @xmath166 before which the @xmath164 corresponding to the cross - layer model takes higher values than the maximum discount factor of the goodman model for different number of users . we notice that starting from @xmath166 , the maximum discount factor of the cross - layer model converges to @xmath164 corresponding to the goodman model . as a function of the packet arrival rate @xmath25 ( @xmath167 ) . ] in a second step , we plotted in fig . [ reg9 ] the variation of the drg social welfare as a function of @xmath168 . we notice that @xmath160 is an increase function of @xmath169 . thus , when @xmath170 , @xmath160 reaches highest value . however , we stress that @xmath160 decreases with the number of users especially for the goodman model while it is quite similar for the cross - layer model . this confirms that the proposed new op is still quite efficient and can be utilized for games with high number of users . as a function of @xmath169 ( @xmath167 ) . ] finally , we plot for both rg models ( frg and drg ) the social welfare when using the cross - layer approach against the constant power @xmath23 for two different values of the packet arrival rate @xmath25 ( @xmath171 and @xmath172 ) . the considered system is composed of @xmath173 users and the spreading factor @xmath142 is fixed to @xmath174 . the idea consists in studying the efficiency of the cross - layer approach regarding the goodman power control algorithm . accordingly , for each packet arrival rate , we plot the social welfare with the cross - layer approach ( powers at the equilibrium are determined normally according to @xmath25 ) and the social welfare with the cross - layer power control but when powers at the equilibrium are determined by the goodman algorithm @xmath175)$ ] . indeed , the packet arrival rate is assumed constant in the goodman model and equal to 1 ( packets arrive with probability @xmath176 ) . for both rg models , we stress that the cross - layer power control approach outperforms the goodman algorithm for both values of the packet arrival rate @xmath25 . important ( relative ) gains are reached . to illustrate this , for @xmath177 and @xmath178 watt the relative gain is higher than @xmath179 in the frg and the drg as well . therefore , we conclude that the op with the cross - layer approach provides better performances and is more efficient than the op with the goodman power control approach . for @xmath177 and @xmath180 : the cross - layer power control approach outperforms the goodman algorithm . ] for @xmath177 and @xmath180 : the cross - layer approach improves the power control when compared to the goodman algorithm . ] in this paper , we have investigated rg for distributed power control in a mac system . as the ne is not always energy - efficient , the nb solution might be a possible efficient solution since it is pareto - efficient . however , the latter , in general , requires global csi at each transmitter node . thus , we were motivated to investigate using the repeated game formulation and develop a new op , that simultaneously is both more efficient than the ne and achievable with only individual csi being required at the transmitter . also , we consider a new ee metric taking into account the presence of a queue at the transmitter with an arbitrary packet arrivals . cooperation plans are proposed where the new op is considered and closed - form expressions of the minimum number of stages for the frg and the maximum discount factor for the drg have been established . the study of the social welfare ( sum of utilities of all the users ) shows that considerable gains are reached compared to the ne ( for the frg and drg ) . moreover , our model proves that even with a high number of users , the frg can always be played with a minimum number of stages shorter than when using the goodman model . in addition , the social welfare in the drg decreases slightly with the number of users with the cross - layer approach while it decreases considerably with the goodman model . finally , the comparison of the cross - layer algorithm versus the goodman algorithm , shows that in real systems with random packet arrivals , the cross - layer power control algorithm outperforms the goodman algorithm . thus , the new op with the cross - layer approach is more efficient . an interesting extension to this work would be to consider the interference channel instead of the mac channel and generalize the framework applied here . another possible extension would be to consider the multi - carrier case and the resulting repeated game . let us determine the maximal utility that a player can get and which is denoted as follows : @xmath182 we denote @xmath183 the power maximizing the utility function @xmath184 and which is the solution of the following equation : @xmath185=0 } , \label{a3}\ ] ] with @xmath186 , @xmath187 and @xmath188 . therefore , the expression of the maximum utility function writes as : @xmath189 with : @xmath190 we have to study then the behavior of @xmath191 regarding @xmath192 for @xmath193 and then we determine the sign of @xmath194 which is given by : @xmath195 we are interested to study the sign of the numerator : @xmath196 with : @xmath197 the next step would be to determine the sign of the expression + @xmath198 . it is obvious that @xmath199 since @xmath200 is an increasing function of the sinr . therefore , we need to determine the sign of @xmath201 . we have : @xmath202 the sign of the first term is negative while the sign of the second term is the same as @xmath203 since @xmath204 and we have : @xmath205 however @xmath206 and then : @xmath207 as shown in @xcite , we have : @xmath208 the latter quantity can be expressed as : @xmath209 consequently , we have : @xmath210>0 . \label{13}\ ] ] therefore , @xmath211 and hence @xmath212 . in particular , we have @xmath213 . thus , we have @xmath214 and finally @xmath215 . we deduce then that @xmath216 is a decreasing function of @xmath192 . it reaches its maximum when @xmath217 and it is minimum when @xmath218 ( for all @xmath193 ) . when substituting @xmath217 in the sinr expression , this allows the determination of the optimal power : @xmath219=0 , \label{17}\ ] ] with : @xmath220 . + the latter equation is a function of the sinr . we determine then the solution in terms of sinr which we denote @xmath120 and for which the optimal power is @xmath221 . this sinr exists due to the quasi - concavity of @xmath184 in @xmath222 @xcite . then , we have : @xmath223 the sinr @xmath224 refers to the sinr when playing the new op while @xmath119 , @xmath120 and @xmath121 are the sinrs at the ne , at the maximal utility and at the utility min - max respectively . in order to simplify expressions , we define the following notations : @xmath225 at a stage @xmath81 , the equilibrium condition is @xcite : @xmath226 } \\ \leq \lambda \tilde{u}_i(\mathbf{p}(t ) ) + \sum_{s\geq t+1}{\lambda ( 1-\lambda)^{s - t}\mathbb{e}_{g}[\tilde{u}_i(\mathbf{p}(s ) ) ] } \end{array}\ ] ] knowing that @xmath227 , we have : @xmath228 \leq \lambda \tilde{u}_i+(1-\lambda)\mathbb{e}_{g}[\tilde{u}_i]\ ] ] @xmath229 \\ \leq \lambda { \frac{g|g_i|^2}{{b|g_i|^2+\tilde{\alpha}h}}}+(1-\lambda)\mathbb{e}_{g}\left [ { \frac{g|g_i|^2}{{b|g_i|^2+\tilde{\alpha}h}}}\right ] \end{array}\ ] ] @xmath230 \\ \leq ( 1-\lambda)\left[{\frac{g\nu_i^{\min}}{{b\nu_i^{\max}+\tilde{\alpha}h } } - \frac{e\nu_i^{\min}}{{b\nu_i^{\max}+\gamma_i^{\ast}\left(\sigma^2+\sum_{j \neq i}{p_j^{\ast } \nu_i^{\max}}\right)f}}}\right ] . \end{array}\ ] ] let @xmath231 and @xmath232 define the following quantities : @xmath131 thus : @xmath233 the authors declare that they have no competing interests . s. bandyopadhyay and e. j. coyle , an energy efficient hierarchical clustering algorithm for wireless sensor networks , infocom 2003 . 22nd annual joint conference of the ieee computer and communications . ieee societies , 3:1713 - 1723 , ( 2003 ) . m. cardei , m. t. thai , y. li and w. wu , energy - efficient target coverage in wireless sensor networks , infocom 2005 . 24th annual joint conference of the ieee computer and communications societies . proceedings ieee , 3:1976 - 1984 , ( mar . 2005 ) . c. isheden , z. chong , e. jorswieck and g. fettweis , framework for link - level energy efficiency optimization with informed transmitter , ieee transactions on wireless communications , 11(8):2946 - 2957 , ( aug . 2012 ) . s. m. betz and h. v. poor , energy efficient communications in cdma networks : a game theoretic analysis considering operating costs , ieee transactions on signal processing , 56(10):5181 - 5190 , ( sep . 2008 ) . a. zappone , z. chong , e. jorswieck and s. buzzi , energy - aware competitive power control in relay - assisted interference wireless networks , ieee transactions on wireless communications , 12(4):1860 - 1871 , ( apr . 2013 ) . v. s. varma , s. lasaulce , y. hayel and s. e. elayoubi , a cross - layer approach for distributed energy - efficient power control in interference networks , ieee transactions on vehicular technology , ( aug . 2014 ) . m. mhiri , k. cheikhrouhou , a. samet , f. mriaux and samson lasaulce , energy - efficient spectrum sharing in relay - assisted cognitive radio systems , ieee proceedings of the 6th international conference on network games , control and optimization ( netgcoop ) , 86 - 91 , ( nov . 2012 ) . s. lasaulce , m. debbah and e. altman , methodologies for analyzing equilibria in wireless games : a look at pure , mixed , and correlated equilibria , ieee signal processing magazine , 26(5):41 - 52 , ( sep . 2009 ) . m. mhiri , v. s. varma , m. le treust , s. lasaulce and a. samet , on the benefits of repeated game models for green cross - layer power control in small cells , 1st international black sea conference on communications and networking ( blackseacom ) , 137 - 141 , ( jul . 2013 ) . m. abidi and v. t. vakili , a game theoretic approach for sinr - constrained power control in 3 g cellular cdma communication systems , ieee 18th international symposium on personal , indoor and mobile radio communications ( pimrc ) , 1 - 5 , ( sep . 2007 ) . y. xu , j. wang , q. wu , a. anpalagan and y. d. yao , opportunistic spectrum access in unknown dynamic environment : a game - theoretic stochastic learning solution , ieee transactions on wireless communications , 11(4):1380 - 1391 , ( apr . 2012 ) . y. song , s. h. y. wong and k. w. lee , optimal gateway selection in multidomain wireless networks : a potential game perspective , mobicom11 proceedings of the 17th annual international conference on mobile computing and networking , 325 - 336 , ( 2011 ) . e. v. belmega and s. lasaulce , an information - theoretic look at mimo energy - efficient communications , valuetools09 proceedings of the 4th international icst conference on performance evaluation methodologies and tools , ( oct .

the main objective of this work is to improve the energy - efficiency ( ee ) of a multiple access channel ( mac ) system , through power control , in a distributed manner . in contrast with many existing works on energy - efficient power control , which ignore the possible presence of a queue at the transmitter , we consider a new generalized cross - layer ee metric . this approach is relevant when the transmitters have a non - zero energy cost even when the radiated power is zero and takes into account the presence of a finite packet buffer and packet arrival at the transmitter . as the nash equilibrium ( ne ) is an energy - inefficient solution , the present work aims at overcoming this deficit by improving the global energy - efficiency . indeed , as the considered system has multiple agencies each with their own interest , the performance metric reflecting the individual interest of each decision maker is the global energy - efficiency defined then as the sum over individual energy - efficiencies . repeated games ( rg ) are investigated through the study of two dynamic games ( finite rg and discounted rg ) , whose equilibrium is defined when introducing a new operating point ( op ) , pareto - dominating the ne and relying only on individual channel state information ( csi ) . accordingly , closed - form expressions of the minimum number of stages of the game for finite rg ( frg ) and the maximum discount factor of the discounted rg ( drg ) were established . our contributions consist of improving the system performances in terms of powers and utilities when using the new op compared to the ne and the nash bargaining ( nb ) solution . moreover , the cross - layer model in the rg formulation leads to achieving a shorter minimum number of stages in the frg even for higher number of users . in addition , the social welfare ( sum of utilities ) in the drg decreases slightly with the cross - layer model when the number of users increases while it is reduced considerably with the goodman model . finally , we show that in real systems with random packet arrivals , the cross - layer power control algorithm outperforms the goodman algorithm .

we begin with the definition of our basic objects of interest . recall that a * contact form * on a @xmath0-manifold is a differential @xmath4-form @xmath5 such that @xmath6 . a * contact sphere * is a triple of @xmath4-forms @xmath7 on a @xmath0-manifold such that any non - trivial linear combination of these forms ( with constant coefficients ) is a contact form . in other words , we require that the @xmath0-form @xmath8 be nowhere zero , i.e. a volume form , for any @xmath9 with @xmath10 . the name ` contact sphere ' derives from the fact that it suffices to check this condition for points @xmath11 on the unit sphere @xmath12 . a contact sphere @xmath13 is called * taut * if the contact form @xmath14 defines the same volume form for all @xmath15 . the requirement for a contact sphere to be taut is equivalent to the system of equations ( for @xmath16 ) @xmath17 a straightforward calculation shows that one can then find a @xmath4-form @xmath18 and a nowhere zero function @xmath19 such that @xmath20 where @xmath21 runs over the cyclic permutations of @xmath22 . notice that @xmath19 is defined by @xmath23 . the analogous structure of a ( taut ) contact circle , defined in terms of two contact forms @xmath24 , was studied in our previous papers @xcite , @xcite , @xcite . in @xcite we gave a complete classification of the closed , orientable @xmath0-manifolds that admit a taut contact circle or a taut contact sphere : [ thm : classold ] let @xmath25 be a closed @xmath0-manifold . * @xmath25 admits a taut contact circle if and only if @xmath25 is diffeomorphic to a quotient of the lie group @xmath26 under a discrete subgroup @xmath27 acting by left multiplication , where @xmath26 is one of the following . * * @xmath28 , the universal cover of @xmath29 . * * @xmath30 , the universal cover of @xmath31 . * * @xmath32 , the universal cover of the euclidean group ( that is , orientation preserving isometries of @xmath33 ) . * @xmath25 admits a taut contact sphere if and only if it is diffeomorphic to a left - quotient of @xmath2 . in the course of this paper we shall present a new proof , more self - contained than the one given in @xcite , of the fact that the universal cover of a closed @xmath0-manifold admitting a taut contact sphere is diffeomorphic to @xmath34 . in @xcite we showed that every closed , orientable @xmath0-manifold admits a ( non - taut ) contact circle , and we gave examples of contact spheres . for instance , @xmath35 , described in terms of coordinates @xmath36 , does not admit any taut contact circles by theorem [ thm : classold ] , but it admits the contact sphere @xmath37 in @xcite we described deformation spaces for taut contact circles and gave a complete classification of taut contact circles . the present paper achieves the corresponding classification for taut contact spheres . the investigation of taut contact spheres amounts to a systematic study of hyperkhler metrics with a homothety , in a sense made precise below . constructions of complete hyperkhler metrics with translational _ invariance _ have played a prominent role in general relativity and supersymmetric field theories , beginning with the gibbons - hawking ansatz @xcite . this ansatz will be discussed in section [ section : gh ] in the context of explicit constructions of taut contact spheres with nontrivial local geometry . see @xcite for a fairly recent discussion of several constructions related to the gibbons - hawking ansatz ( eguchi - hanson metric , taub - nut metric , atiyah - hitchin metric ) . hyperkhler metrics with a _ homothety _ are equally important in physical applications , see @xcite and @xcite . the latter pays particular attention to homotheties that are hypersurface orthogonal ( such homotheties are called _ dilatations _ in @xcite ) . this is equivalent to saying that the hyperkhler metric on @xmath38 , where @xmath39 is some @xmath0-manifold , is the cone metric over a @xmath0-sasakian metric on @xmath39 , cf . @xcite . for more general information on sasakian and in particular @xmath0-sasakian geometry see the definitive survey @xcite or the monograph @xcite ; some of the definitions will be recalled in section [ section : sasakian ] . there we use our methods to recover the classification of the closed @xmath0-manifolds admitting @xmath0-sasakian structures . in contrast with @xmath0-sasakian structures , taut contact spheres do not , in general , give rise to a cone metric on @xmath38 ( in other words , the relevant homothety is not a dilatation ; such general homotheties also appear in @xcite ) . this implies that taut contact spheres are definitely more general than @xmath0-sasakian structures . we elaborate on this point in section [ section : sasakian ] . by comparison , khler metrics on @xmath38 admitting a dilatation correspond to a sasakian structure on @xmath39 . ( for a classification of the closed @xmath0-dimensional manifolds admitting sasakian structures see @xcite . ) in @xcite it is shown that if the sasakian analogue of the khler potential satisfies a monge - ampre equation , then the metric on @xmath39 is sasakian - einstein . this happens in particular if @xmath38 is ricci - flat , cf . @xcite . a taut contact sphere on a closed manifold @xmath25 always gives rise to a _ flat _ hyperkhler metric on @xmath40 ( theorem [ thm : flat ] ) . this must be read as a global rigidity phenomenon , because the theorem fails for open @xmath0-manifolds . in section [ section : helmholtz ] we use a monge - ampre equation to construct a taut contact sphere on an open subset @xmath39 of @xmath3 giving rise to a non - flat hyperkhler metric on @xmath41 . in the appendix we use a contact transformation to relate this construction to the helmholtz equation on the @xmath1-sphere . in section [ section : gh ] we use the gibbons - hawking ansatz to give an even simpler construction of a non - flat example . in section [ section : bernstein ] we discuss the question , known as a _ bernstein problem _ , whether such non - flat examples can give rise to complete metrics . the answer to this question depends on the choice of one of the two natural metrics associated with a taut contact sphere ( see definition [ defn : short - long ] , theorem [ thm : bernstein ] , and the comments following it ) . this paper supersedes our preprint `` contact spheres and quaternionic structures '' . we now describe in outline some of the main results of the present paper . our notational convention throughout will be that @xmath25 denotes a closed , orientable @xmath0-manifold ; @xmath39 will denote a @xmath0-manifold ( without boundary ) that need not be compact . the relation @xmath42 is easily seen to be an equivalence relation within the set of ( taut ) contact spheres . two ( taut ) contact spheres are * conformally equivalent * if one is obtained from the other by multiplying each contact form by the same positive function . we call a contact sphere * naturally ordered * if @xmath43 is a positive multiple of @xmath44 . throughout this paper we shall assume our contact spheres to satisfy this condition . since @xmath43 and @xmath45 scale with the second and third power of @xmath46 , respectively , it is obvious that every conformal equivalence class of naturally ordered taut contact spheres contains , for any @xmath47 , a unique representative satisfying @xmath48 we call a taut contact sphere @xmath13 * @xmath49-normalised * if it satisfies equation ( [ eqn : normalised ] ) . this condition is equivalent to @xmath50 in equation ( [ eqn : structure ] ) ; beware that @xmath18 in that equation is not an invariant of the conformal equivalence class . it is implicit in @xcite and follows by a simple extension of the ideas from @xcite that a taut contact sphere on @xmath39 gives rise to a hyperkhler structure on @xmath38 . in section [ section : hyper ] we analyse this situation a little more carefully . one of the results proved there is the following . [ prop : hyper1 ] a contact sphere on @xmath39 determines an oriented conformal structure on @xmath38 . a naturally ordered taut contact sphere on @xmath39 determines a hyperkhler structure on @xmath38 . conformally equivalent ( taut ) contact spheres determine isomorphic conformal ( resp . hyperkhler ) structures . as we shall see in section [ section : hyper ] , the hyperkhler structure @xmath51 on @xmath52 induced by a taut contact sphere @xmath13 on @xmath39 is given by the equations @xmath53 where @xmath54 denotes the @xmath55-coordinate , and the complex structures @xmath56 are @xmath57-invariant . often we write the hyperkhler structure as the triple @xmath58 of symplectic forms . these symplectic forms are homogeneous of degree @xmath4 with respect to the vector field @xmath57 , that is , @xmath59 . the hyperkhler metric @xmath60 has the same homogeneity . for reversing this construction , it is useful to make the following definition . a vector field @xmath61 on a hyperkhler manifold is called * tri - liouville * if it is a liouville vector field for every parallel self - dual @xmath1-form @xmath62 , that is , @xmath63 . notice that a tri - liouville vector field is automatically homothetic for the hyperkhler metric , but the converse is not true . the construction of taut contact spheres in this paper uses two ingredients : a hyperkhler metric @xmath60 and a nowhere zero tri - liouville vector field @xmath61 . for any conformal basis @xmath58 of parallel self - dual @xmath1-forms , the corresponding taut contact sphere is defined on any transversal to the flow of @xmath61 by restricting the triple of @xmath4-forms @xmath64 to that transversal . for the formal statement see proposition [ prop : hyper2 ] . we also show that for a naturally ordered taut contact sphere @xmath65 on @xmath39 there is the following _ pointwise _ model on @xmath38 for the triple of symplectic forms @xmath58 , expressed in quaternionic notation : at any point @xmath66 , there is a quaternionic coordinate @xmath67 for the tangent space @xmath68 such that @xmath69 the key to the classification of taut contact spheres is then the following statement . [ thm : flat ] the hyperkhler metric on @xmath70 induced by a taut contact sphere on a closed @xmath0-manifold @xmath25 is flat . hyperkhler metrics are always ricci flat @xcite , and any ricci flat khler manifold of complex dimension @xmath1 is anti - self - dual @xcite , that is , the self - dual part @xmath71 of the weyl tensor of the metric @xmath60 vanishes . since the weyl tensor is an invariant of the conformal class of a metric @xcite , @xmath71 also vanishes for the metric @xmath72 , which descends to the quotient @xmath73 of @xmath70 under the map @xmath74 , say . ( in fact , the hyperhermitian structure @xmath75 descends to that quotient , and one may also appeal to a result of boyer @xcite saying that a hyperhermitian metric on a @xmath76-manifold is anti - self - dual . ) then the signature formula for @xmath77 yields @xmath78 but @xmath79 for purely topological reasons . so the weyl tensor @xmath80 vanishes for @xmath72 . again appealing to the conformal invariance of the weyl tensor , we deduce that it also vanishes for @xmath60 . for ricci flat metrics ( thus , in particular , for the hyperkhler metric @xmath60 ) the full curvature tensor equals its weyl part @xcite . this proves the theorem . this theorem implies that on a closed manifold the _ pointwise _ model above for the triple of symplectic forms @xmath81 , @xmath82 , coming from a taut contact sphere , is actually a _ local _ model , since the three symplectic forms and the hypercomplex structure @xmath83 are all parallel with respect to the flat hyperkhler metric @xmath60 . it is then not very difficult , using properties of the @xmath57-flow , to derive the following classification statement . the proof will be given in section [ section : class ] . [ thm : class ] if @xmath25 is diffeomorphic to a lens space @xmath84 , including the @xmath0-sphere @xmath85 , then the ( naturally ordered ) taut contact spheres on @xmath25 , up to diffeomorphism and conformal equivalence , are given by the following family of @xmath86-invariant quaternionic @xmath4-forms on @xmath87 , @xmath88 where @xmath89 and @xmath90 determine equivalent structures . if @xmath25 is diffeomorphic to @xmath91 with @xmath92 a non - abelian group , there is a unique equivalence class of taut contact spheres on @xmath25 , described by the formula above with @xmath93 . all these taut contact spheres are homogeneous under a natural @xmath29-action . in particular , all great circles in a given taut contact sphere are isomorphic taut contact circles . note that the manifolds listed in this theorem exhaust all the possible left - quotients of @xmath2 , cf . the @xmath86-action on @xmath34 that produces the quotient @xmath84 is generated by right multiplication with @xmath94 . as we shall see in the construction of the moduli space of taut contact spheres , all taut contact spheres on a given closed @xmath0-dimensional manifold yield the same @xmath76-dimensional metric @xmath60 up to global isometry . so it is not this induced metric which determines the modulus , but in fact the different possible homothetic vector fields @xmath57 . in the cases where non - trivial moduli exist , neither the vector field @xmath57 nor the function @xmath95 ( which turns out to equal @xmath96 ) are unique for that @xmath60 . with our coordinate conventions in section [ section : hyper ] below , which seem natural in that context , _ left _ multiplication on @xmath97 by @xmath98 where @xmath99 , @xmath100 , @xmath101 , corresponds to _ right _ multiplication on @xmath102 by the unit quaternion @xmath103 . the quaternionic @xmath4-form @xmath104 is invariant under this right multiplication @xmath105 by unit quaternions @xmath106 , and therefore descends to all left - quotients of @xmath2 . the @xmath4-form @xmath107 is invariant under right multiplication by unit quaternions of the special form @xmath108 , hence it descends to all _ abelian _ quotients . here are some details about what we mean by ` homogeneity under a natural @xmath29-action ' in theorem [ thm : class ] . the action of @xmath29 on @xmath109 rotates the contact forms and is given by conjugating the quaternionic @xmath4-form by elements @xmath110 . this action is induced from the @xmath34-action @xmath111 on @xmath102 , and that latter action can not be replaced by an @xmath29-action . this amounts to a spinor phenomenon , see also the proof of proposition [ prop : planes ] . since this left multiplication by unit quaternions commutes with all quaternionic right multiplications , it descends to all left - quotients of @xmath2 . with @xmath112 we obtain the following real expression for the taut contact sphere in theorem [ thm : class ] corresponding to @xmath93 , where @xmath21 runs over the cyclic permutations of @xmath22 : @xmath113 notice that it satisfies ( [ eqn : structure ] ) with @xmath114 and @xmath115 . in particular , it is a @xmath1-normalised taut contact sphere . restricting this to the hyperplane @xmath116 , we obtain a simple expression for a taut contact sphere on @xmath3 : @xmath117 see proposition [ prop : hyper2 ] for the principle behind this observation . one might suspect that taut contact spheres constitute such a rigid structure that theorem [ thm : flat ] would also hold locally and for open manifolds . however , this turns out to be false , even conformally . [ thm : nonflat ] there are examples of taut contact spheres on open domains @xmath39 in @xmath3 inducing a metric on @xmath38 that is not conformally flat . this theorem will be proved in section [ section : nonflat ] , where we present two methods for constructing such examples . our first construction , in section [ section : helmholtz ] , starts from the observation that ( locally ) the conditions for a taut contact sphere lead to a complex monge - ampre equation . after imposing an additional homogeneity amounting to the existence of a tri - hamiltonian vector field , we are able to find concrete solutions of that equation whose associated @xmath76-dimensional metric is non - flat . in fact , subject to this extra homogeneity , the monge - ampre equation can be linearised to yield the helmholtz equation @xmath118 for the laplacian @xmath119 on the @xmath1-sphere . solutions of that helmholtz equation then give rise to taut contact spheres with a tri - hamiltonian symmetry . our second construction , in section [ section : gh ] , starts from the gibbons - hawking ansatz , which is essentially a construction of an @xmath55-invariant hyperkhler metric on @xmath120 , starting from a harmonic function on an open subset @xmath121 of @xmath3 . for an appropriate choice of such a harmonic function , one obtains a non - flat hyperkhler metric giving rise to a taut contact sphere on a suitable hypersurface @xmath122 . beware that the @xmath55-factor in this splitting @xmath120 is not the one corresponding to the tri - liouville vector field . taut contact spheres come associated with two natural metrics on the @xmath0-manifold . in order to describe these , we observe that the @xmath1-forms @xmath123 can be written with the help of the structure equation ( [ eqn : structure ] ) as follows : @xmath124 so the hyperkhler metric @xmath60 is given by @xmath125 in particular , if the contact sphere is @xmath4-normalised , we have @xmath126 this motivates the following definition . [ defn : short - long ] the * short metric * associated with a @xmath4-normalised taut contact sphere @xmath13 on @xmath39 is the metric @xmath127 the * long metric * associated with this contact sphere is @xmath128 observe that @xmath129 is simply the restriction of @xmath60 to @xmath130 , so from the viewpoint of hyperkhler geometry this is the more natural metric to consider . in either of the above - mentioned constructions of taut contact spheres on @xmath39 giving rise to a non - flat hyperkhler metric on @xmath41 , the induced long metric @xmath129 on @xmath39 is incomplete , and so is , _ a fortiori _ , the short metric @xmath131 . in section [ section : bernstein ] we raise the question whether one can find such examples where @xmath129 , at least , is complete . this type of question is known as a bernstein problem @xcite . concerning @xmath131 , we provide a partial answer to this problem for contact spheres with additional symmetries . for @xmath129 there is a positive answer to the bernstein problem , even subject to the additional symmetry requirement : [ thm : bernstein ] let @xmath13 be a @xmath4-normalised taut contact sphere on @xmath39 giving rise to the short metric @xmath131 and the long metric @xmath129 on @xmath39 . \(a ) if @xmath131 is complete and admits a non - trivial killing vector field that preserves @xmath13 , then @xmath39 is necessarily compact , and hence @xmath13 belongs to the family described in theorem [ thm : class ] ( and in particular gives rise to a flat hyperkhler metric ) . \(b ) there are examples of @xmath132-invariant taut contact spheres on @xmath133 , where @xmath134 is the open unit disc , for which @xmath129 is complete . the induced hyperkhler metrics on @xmath41 are not flat . part ( a ) will be proved in section [ section : bernstein ] . theorem [ thm : bernstein - killing ] in that section is a more explicit reformulation of this part . we reserve the proof of part ( b ) for a forthcoming paper . it turns out that in the @xmath55- or @xmath132-invariant context one can use the gibbons - hawking ansatz in order to develop a complete theory of such contact circles . an infinite - dimensional family of examples giving rise to complete long metrics is then found with the help of blaschke products on @xmath135 . in this section we discuss the linear algebraic aspects of contact spheres , leading to a proof of proposition [ prop : hyper1 ] . this prepares the ground for the proof of theorem [ thm : class ] . let @xmath13 be a contact sphere on a @xmath0-manifold @xmath39 . this gives rise to the symplectic forms @xmath123 , @xmath82 , on @xmath38 . at any point @xmath136 of @xmath38 , these symplectic forms span a definite @xmath0-plane in the space @xmath137 of skew - symmetric bilinear forms on the tangent space @xmath68 . if the contact sphere is taut , we have in addition the identities ( for @xmath16 ) @xmath138 first we are going to study the linear algebra of this situation . thus , let @xmath139 be a @xmath76-dimensional real vector space and write @xmath140 . consider the quadratic form @xmath141 of signature @xmath142 . we call a triple @xmath143 of elements of @xmath144 a * symplectic triple * on @xmath139 if it spans a definite @xmath0-plane for @xmath145 in @xmath144 , and a * conformal symplectic triple * if the stronger condition @xmath146 is satisfied for @xmath16 , cf . @xcite . the same terminology will be used for triples of symplectic forms @xmath58 on a @xmath76-manifold as described above . in the sequel , any equation ( or other statement ) involving the indices @xmath147 is meant to be read as three equations , with @xmath21 ranging over the cyclic permutations of @xmath22 . we write bold face @xmath148 for @xmath149 , and bold face @xmath150 for the standard quaternionic units with @xmath151 . the relation between real , complex , and quaternionic coordinates will be given by @xmath152 @xmath153 for @xmath154 a complex structure on @xmath139 , we denote by @xmath155 the space of exterior forms of type @xmath156 on @xmath139 . as shown in @xcite , a conformal symplectic couple @xmath157 determines a unique complex structure @xmath158 on @xmath139 for which @xmath159 . thus , a conformal symplectic triple @xmath143 induces three complex structures @xmath160 . in @xcite it was shown that these complex structures satisfy the quaternionic identities as well as the relations @xmath161 for a unique definite symmetric bilinear form @xmath60 . we call a conformal symplectic triple * naturally ordered * if this @xmath60 is _ positive _ definite . notice that the sign of @xmath60 is well defined , since the @xmath162 do not satisfy the quaternionic identities . ( in particular , hyperkhler structures are always naturally ordered . ) an alternative way to define this definite bilinear form @xmath60 is via the identity @xmath163 this is an obvious consequence of the @xmath29-homogeneous normal form for conformal symplectic triples discussed in the next two propositions . [ prop : pointmodel ] let @xmath143 be a naturally ordered conformal symplectic triple on @xmath139 . then there are real linear coordinates @xmath164 on @xmath139 such that @xmath165 in terms of the corresponding complex and quaternionic coordinates we have @xmath166 and @xmath167 because of the non - commutativity of @xmath102 some care is necessary in interpreting the wedge product of @xmath102-valued @xmath4-forms @xmath168 on a vector space @xmath169 . our convention is to read it as @xmath170 this ensures @xmath171 , and that @xmath172 is always purely imaginary . from @xmath173 and @xmath174 we conclude that the @xmath175-part of @xmath176 with respect to @xmath177 is zero . the form @xmath176 being real , its @xmath178-part must also vanish , hence @xmath179 . let @xmath180 be a basis for @xmath181 and set @xmath182 . then we can write @xmath183 with @xmath184 a hermitian @xmath185-matrix . by our assumption on @xmath143 being naturally ordered , the matrix @xmath184 is positive definite . hence there is a matrix @xmath186 such that @xmath187 and so in terms of the basis @xmath188 for @xmath181 defined by @xmath189 we have @xmath190 and @xmath191 for some @xmath192 . we then find @xmath193 from which we conclude @xmath194 . the linear complex coordinates @xmath195 corresponding to @xmath196 then give the desired complex normal form . the real and quaternionic normal forms can be derived easily from the complex one . [ rem : metric ] ( 1 ) notice that in terms of these pointwise coordinates we have @xmath197 and @xmath198 . moreover , we recognise the @xmath0-plane in @xmath144 spanned by @xmath199 as the space of self - dual @xmath1-forms for the metric @xmath60 and the orientation of @xmath139 defined by @xmath200 . the length of the @xmath201 equals @xmath202 . \(2 ) here is another characterisation of taut contact spheres that can be read off from the preceding proposition : the purely imaginary @xmath4-form @xmath203 defines a taut contact sphere if and only if @xmath204 is a contact form and at each point @xmath136 of the manifold there is an @xmath102-valued linear form @xmath205 on the tangent space at @xmath136 such that @xmath206 . the case @xmath207 corresponds to @xmath13 being naturally ordered . the following proposition will be an important ingredient in the proof of proposition [ prop : hyper1 ] , while the last part of its proof is used in the proof of theorem [ thm : class ] . [ prop : planes ] there is a one - to - one correspondence between oriented conformal structures on @xmath139 and definite @xmath0-planes @xmath208 in @xmath144 ( with respect to @xmath145 ) . given an oriented conformal structure on @xmath139 , define @xmath208 as the corresponding space of self - dual @xmath1-forms on @xmath139 . for the converse , we recall some quaternionic linear algebra . under the identification of the purely imaginary quaternions with @xmath3 , any element @xmath209 of @xmath29 can be written as quaternionic conjugation @xmath210 , @xmath211 with some unit quaternion @xmath212 . the map @xmath213 is the standard double covering @xmath214 . given a definite @xmath0-plane in @xmath144 , choose a naturally ordered conformal symplectic triple @xmath143 spanning it . an orientation on @xmath139 is then given by @xmath200 . let @xmath215 be a quaternionic coordinate for @xmath139 as in proposition [ prop : pointmodel ] , and @xmath197 the inner product on @xmath139 determined by @xmath143 . suppose that @xmath216 is another naturally ordered conformal symplectic triple spanning the same @xmath0-plane @xmath208 and satisfying @xmath217 . notice that @xmath145 defines an inner product on @xmath208 for which @xmath143 and @xmath218 are orthogonal bases consisting of vectors of equal length , and defining the same orientation . hence there is an element @xmath219 such that @xmath220 by the preceding discussion this can be written as @xmath221 so a quaternionic coordinate corresponding to @xmath218 is given by @xmath222 , which gives rise to the same inner product @xmath60 on @xmath139 since left multiplication on @xmath102 by a unit quaternion is an isometry for @xmath223 . if the conformal symplectic triple @xmath143 is replaced by @xmath224 , @xmath225 , then the induced inner product changes to @xmath226 . this proves the proposition . a ( taut ) contact sphere @xmath13 on @xmath39 gives rise to a ( conformal ) symplectic triple @xmath58 on @xmath41 as described at the beginning of this section . the complex structures @xmath160 defined on each tangent space @xmath227 depend smoothly on the point @xmath136 , and thus define almost complex structures on @xmath38 , which are integrable in the taut case , see @xcite . so the statement concerning contact spheres and conformal structures is immediate from the preceding proposition . notice that the orientation and conformal class of the metric on @xmath38 are completely characterised as the unique ones for which the @xmath228 are self - dual . the statement about taut contact spheres and hyperkhler structures follows similarly ; see @xcite for an explicit argument . non - taut _ contact sphere determines at each point a linear @xmath1-sphere worth of complex structures , since pointwise the symplectic triple @xmath58 can be replaced by a conformal symplectic triple spanning the same @xmath0-plane of skew - symmetric forms ; this does not change the space of corresponding almost complex structures . although this replacement can be done globally , leading again to triples @xmath83 satisfying the quaternionic identities , there does not seem to be a canonical choice for doing it . if two contact spheres are related by multiplication by the function @xmath229 , the induced structures on @xmath38 are related by the diffeomorphism given by @xmath230 . the following proposition explains how to go back from a hyperkhler structure and a tri - liouville vector field to a taut contact sphere . [ prop : hyper2 ] let @xmath58 be a hyperkhler structure on a @xmath76-manifold @xmath231 with a nowhere zero tri - liouville vector field @xmath61 . then the equations @xmath232 define a naturally ordered taut contact sphere @xmath13 on any transversal of @xmath61 . shifting the points of the transversal along the orbits of @xmath61 will change the taut contact sphere within its conformal class . the proof is a straightforward computation , using @xmath233 and @xmath234 for @xmath16 , cf . @xcite , as well as the identities @xmath235 , cf.identity ( [ eqn : cubic ] ) . if the hyperkhler structure comes from a taut contact sphere , then the above construction with @xmath236 recovers that contact sphere . this section is largely devoted to the proof of theorem [ thm : class ] . as promised , this includes a new proof that the universal cover of a closed @xmath0-manifold carrying a taut contact sphere is diffeomorphic to @xmath34 . the strategy for the proof is as follows . from theorem [ thm : flat ] we know that a taut contact sphere on a closed @xmath0-manifold @xmath25 induces a flat metric on @xmath40 . with the help of the developing map of this metric we are led to study hyperkhler structures on euclidean @xmath76-space @xmath237 , with the hyperkhler metric being the flat euclidean metric . parallel @xmath1-forms for this metric have constant coefficients in euclidean coordinates . the classification of taut contact spheres is thus reduced to a straightforward problem of determining the possible tri - liouville vector fields . let @xmath25 be a closed @xmath0-manifold with a taut contact sphere , and let @xmath13 be the lifted taut contact sphere on its universal cover @xmath238 . write @xmath60 for the induced metric on @xmath239 . the fact that @xmath60 is flat and @xmath239 simply - connected implies that we have a developing map @xmath240 for this metric which is a local isometry . if @xmath241 is a domain on which @xmath242 restricts to a diffeomorphism , then @xmath243 is a vector field @xmath244 on the domain @xmath245 generating a @xmath4-parameter group of homotheties of the euclidean metric ( because @xmath246 by the construction of @xmath60 ) . since homothetic transformations of a riemannian manifold are affine transformations ( this is easy to see for the euclidean metric ) , @xmath244 is the restriction @xmath247 of a homothetic vector field @xmath61 defined on all of @xmath237 . then @xmath57 and @xmath248 are homothetic vector fields for @xmath60 that coincide on the open set @xmath231 , which forces @xmath249 on all of @xmath239 . a homothetic vector field on @xmath237 vanishes at a single point , and without loss of generality we may assume that @xmath61 vanishes at @xmath250 . let @xmath251 be the projection onto the orbit space of @xmath61 ( which we can identify with the unit sphere @xmath252 , for @xmath61 is a genuinely expanding homothetic vector field and hence transverse to any sphere centred at @xmath250 ) . then the composition @xmath253 is a local diffeomorphism and therefore , @xmath238 being simply - connected , a diffeomorphism . so we have proved @xmath254 . moreover , the property @xmath255 implies that @xmath242 is a diffeomorphism from @xmath256 to @xmath257 , hence a global isometry . from now on we write @xmath34 instead of @xmath238 . now @xmath242 sends @xmath258 to some transversal of @xmath61 in @xmath237 , and by proposition [ prop : hyper2 ] the original contact sphere is equivalent to the one induced on @xmath259 . to simplify notation , we continue to write @xmath260 , @xmath228 , @xmath56 for the push - forwards of these objects to @xmath261 , and we identify @xmath57 with @xmath61 . thus @xmath83 defines a hyperkhler structure with respect to the euclidean metric @xmath262 , and @xmath58 are the corresponding khler forms . in particular , the @xmath56 and @xmath228 are parallel with respect to @xmath262 , and thus have constant coefficients in any linear coordinate system for @xmath237 . as a consequence , there are linear coordinates @xmath263 on @xmath237 with respect to which the formulae of proposition [ prop : pointmodel ] hold ( with @xmath201 replaced by @xmath228 ) . observe that this forces @xmath263 to be an orthonormal coordinate system with respect to @xmath262 . write @xmath264 for the flow of @xmath61 . this flow commutes with the @xmath56 and satisfies @xmath265 and @xmath266 , in particular @xmath267 . since the flow of @xmath61 preserves @xmath177 , we have a holomorphic vector field @xmath268 on @xmath269 with @xmath270 . as a homothetic vector field vanishing at zero , @xmath61 can be represented as a linear map , and this implies that @xmath268 corresponds to a complex linear map @xmath271 with respect to the coordinate @xmath272 . a straightforward calculation shows that the condition @xmath273 translates into the matrix @xmath274 being of the form @xmath275 with @xmath276 a skew - hermitian matrix ; the condition @xmath277 forces @xmath276 to have zero trace . after a special unitary change of coordinates ( which does not change the expressions for the @xmath228 ) we may assume that @xmath276 is in diagonal form , i.e. @xmath278 with @xmath279 . in the usual notation for vector fields this means @xmath280 we conclude @xmath281 this analysis can be carried out locally . as a result , the quaternionic formula in theorem [ thm : class ] is a universal local model for ( naturally ordered ) taut contact spheres inducing a _ flat _ hyperkhler metric . to translate the preceding equations into quaternionic notation , we observe @xmath282 and @xmath283 ( using @xmath284 for @xmath285 and @xmath286 ) , hence @xmath287 we also observe @xmath288 putting all this together , we find @xmath289 this shows that , up to conformal equivalence and diffeomorphism , the taut contact spheres on @xmath34 are as described in theorem [ thm : class ] . the fact that different non - negative values of @xmath89 give non - isomorphic contact spheres follows from the corresponding classification of taut contact circles in @xcite , @xcite . we shall be a bit more explicit about this point below , where we give a pictorial description of the moduli spaces in question . this will include a synthetic method for determining the modulus , independent of our previous papers . next we want to show that @xmath89 and @xmath90 correspond to equivalent structures . the diffeomorphism of @xmath34 given by @xmath290 is isotopic to the identity and pulls the quaternionic @xmath4-form with parameter value @xmath89 to that with value @xmath90 , because @xmath291 anticommutes with @xmath148 . after having determined the possible lifted taut contact spheres on @xmath254 , we now consider the taut contact sphere on @xmath25 itself . from theorem [ thm : classold ] we know that @xmath25 has to be a left - quotient of @xmath2 . the induced flat hyperkhler structure on @xmath70 lifts to just such a structure on @xmath256 , invariant under the deck transformation group @xmath27 . as was already argued in @xcite for taut contact circles , this implies that in the complex coordinates @xmath292 which give a normal form as described above , one has @xmath293 . moreover , again as in @xcite , the parameter @xmath89 is forced to be zero for non - abelian @xmath27 , and it can take any value for the group @xmath294 generated by @xmath295 where @xmath296 is some @xmath297th root of unity . now we address the homogeneity issue in the statement of theorem [ thm : class ] . for any unit quaternion @xmath106 we have @xmath298 therefore , if two taut contact spheres are related by an element @xmath299 of @xmath29 as in the proof of proposition [ prop : planes ] , then one is the pull - back of the other under a diffeomorphism of @xmath300 induced by the map @xmath111 . in particular , this @xmath29-action on taut contact spheres shows that any taut contact sphere can be swept out by great circles , all defining isomorphic taut contact circles . this concludes the proof of theorem [ thm : class ] . this theorem allows us to define a map from the space of taut contact spheres to that of taut contact circles : simply pick any great circle . [ thm : moduli ] let @xmath25 be any left - quotient of @xmath2 . the map sending a taut contact sphere on @xmath25 to any of its great circles induces an embedding of the moduli space of taut contact spheres on @xmath25 into the moduli space of taut contact circles . recall the classification ( up to diffeomorphism and homothety , i.e.conformal equivalence and rotation ) of taut contact circles @xmath301 on left - quotients of @xmath34 , see @xcite , @xcite . on the lens spaces @xmath84 we have the continuous family of taut contact circles induced by the @xmath86-invariant complex @xmath4-form @xmath302 ( restricted to @xmath303 ) with @xmath304 , @xmath305 , modulo replacing @xmath306 by @xmath307 , which corresponds to the diffeomorphism defined by @xmath308 and changing from @xmath301 to @xmath309 . so from theorem [ thm : class ] we see that * only the taut contact circles with @xmath306 purely imaginary extend to taut contact spheres , * this extension is unique up to automorphisms of @xmath310 , and * two taut contact circles giving rise to isomorphic taut contact spheres must be isomorphic . this proves the theorem in the given case . for the abelian left - quotients the map on moduli spaces amounts to the inclusion @xmath311 the moduli space of taut contact circles on @xmath84 also includes a discrete family described by @xmath312 where @xmath313 ranges over the natural numbers congruent @xmath314 mod @xmath297 these contact circles , however , do not extend to any taut contact sphere , so they are not being considered here . the moduli spaces described in the foregoing proof ( without the discrete family ) are illustrated on the left - hand side of figure [ figure : moduli ] . the moduli space of taut contact circles is the orbifold @xmath315 the half - line @xmath316 constituting the moduli space of taut contact spheres is shown as a dashed line . the real part @xmath317 of this moduli space , shown in bold , corresponds to so - called cartan structures . the origin represents the unique @xmath0-sasakian structure . these structures will be discussed in section [ section : sasakian ] . under the mapping @xmath318 , the moduli space of taut contact circles is mapped bijectively to the interior of the parabola @xmath319 shown on the right - hand side of figure [ figure : moduli ] . so this singular mapping flattens the cone point and yields a representation of the moduli space as an open subset of @xmath320 . here is the promised synthetic characterisation of the modulus @xmath89 of a taut contact sphere on @xmath34 . one can define a _ canonical slice _ inside the hyperkhler manifold @xmath321 as the subset @xmath322 . a straightforward calculation shows that for the taut contact sphere of modulus @xmath323 we have latexmath:[\[\|\partial_t\|^2=\bigl(\frac{1}{4}+\nu^2\bigr)\cdot\bigl ( @xmath0-sphere of radius @xmath325 . the canonical slice of maximal radius corresponds to the unique @xmath0-sasakian structure . the canonical slice has the property that the @xmath4-forms @xmath260 induce on it a @xmath4-normalised taut contact sphere . we conclude that the family in theorem [ thm : class ] induces on any euclidean sphere a @xmath49-normalised taut contact sphere , for some constant @xmath49 . recall now definition [ defn : short - long ] . once the contact sphere is @xmath4-normalised , the canonical slice is given by @xmath326 and we see that the hyperkhler metric induces the long metric @xmath129 on it . thus , on a closed @xmath0-manifold the long metric is always spherical . the short metric can be written as @xmath327 . we claim that @xmath18 is invariant under the maps @xmath111 with @xmath328 , which shows that @xmath131 is a _ berger metric . _ to see this invariance property , notice that the system of structure equations ( [ eqn : structure ] ) is invariant under rotations of the triple @xmath13 , and we have seen that left multiplication by unit quaternions @xmath106 induces such rotations . the contact sphere with @xmath93 in theorem [ thm : class ] is invariant under right multiplication by unit quaternions , which makes the corresponding @xmath18 bi - invariant , but recall that in this case @xmath114 and so there is no contradiction here . we briefly expand on the point in the proof of theorem [ thm : moduli ] concerning the automorphisms of @xmath329 . from the arguments in ( * section 5.2 ) it follows that automorphisms @xmath209 of @xmath329 are @xmath320-linear maps of @xmath97 , in fact , elements of @xmath330 . given an @xmath204 extending @xmath301 to a taut contact sphere , the other extensions are given by @xmath331 . for @xmath34 and @xmath93 , any element of @xmath330 defines an automorphism of @xmath329 , so the possible extensions are parametrised by @xmath332 . for @xmath34 and @xmath333 , the condition that @xmath209 preserve @xmath329 forces it to be a diagonal map @xmath334 . the same is true for the lens spaces @xmath335 other than @xmath34 , even in the case @xmath93 ; here the condition for @xmath209 to be diagonal follows from the fact that it has to lie in the normaliser of @xmath86 . in these two cases , the possible extensions are parametrised by @xmath336 , since @xmath337 finally , for the quotients of @xmath300 with @xmath27 non - abelian , the fact that @xmath209 lies in the normaliser of @xmath27 forces it to be plus or minus the identity map . thus the extension @xmath204 is unique . in this section we describe two constructions of taut contact spheres giving rise to non - flat hyperkhler metrics . these examples constitute a proof of theorem [ thm : nonflat ] , since a non - flat hyperkhler metric is not even conformally flat ( cf . the proof of theorem [ thm : flat ] ) . we have seen in proposition [ prop : hyper2 ] how to construct a taut contact sphere corresponding to a suitable hyperkhler structure @xmath58 . notice that in terms of the holomorphic structure given by @xmath177 , an equivalent description of this hyperkhler structure is given by a holomorphic symplectic form @xmath338 and a closed real @xmath339-form @xmath340 with @xmath341 . we now make the following ansatz : let @xmath195 be complex coordinates on @xmath97 and identify @xmath97 with @xmath342 by equating the @xmath55-direction with the real part of @xmath343 , so that @xmath344 and @xmath345 . set @xmath346 with @xmath347 a nowhere zero holomorphic function , and @xmath348 with @xmath349 a real - valued function . to satisfy the conditions @xmath350 it is sufficient to have @xmath351 further , the identity @xmath352 is equivalent to the complex monge - ampre equation @xmath353 if all these conditions are met , proposition [ prop : hyper2 ] tells us how to recover the corresponding taut contact sphere . we satisfy condition ( [ eqn : homogeneity ] ) by taking @xmath354 and simplifying the ansatz further to @xmath355 with @xmath356 a function of two real variables . then , writing @xmath357 for the partial derivatives @xmath358 etc . , we have @xmath359 and equation ( [ eqn : ma ] ) becomes @xmath360 the reader may check directly that the function @xmath361 satisfies this equation . in the appendix we shall explain how to derive this sample solution , and verify that it leads to a non - flat hyperkhler metric . we arrive at this example after proving a general result that relates solutions of ( [ eqn : ma2 ] ) to solutions of the helmholtz equation on the @xmath1-sphere . the gibbons - hawking ansatz @xcite starts with a positive function @xmath169 in three variables @xmath362 ( locally on @xmath3 ) and a triple of functions @xmath363 in the same variables , satisfying the condition @xmath364 so that in particular @xmath365 i.e. @xmath169 is harmonic . set @xmath366 and consider the triple of @xmath1-forms on @xmath367 ( with @xmath368 denoting the @xmath55-coordinate ) defined by @xmath369 the relation between @xmath169 and @xmath18 implies that these @xmath1-forms are closed . writing @xmath370 we have @xmath371 which shows that @xmath58 is a conformal symplectic triple . the corresponding hyperkhler metric is @xmath372 observe the formal similarity with the metric in ( [ eqn : metric ] ) . the translational invariance of this metric in @xmath368-direction is obvious . ; this is the usual form of the gibbons - hawking ansatz . ] a homothety can be built into this ansatz by choosing @xmath18 appropriately . here is an example . as domain @xmath373 we take the half - space given by the condition @xmath374 , and @xmath375 is our positive function on that domain . set @xmath376 , that is , @xmath377 . then @xmath378 i.e. condition ( [ eqn : curl ] ) is satisfied . then the gibbons - hawking ansatz yields the hyperkhler structure to @xmath379 is meant to emphasise that we want this to be an @xmath55- , not an @xmath132-coordinate . ] @xmath380 with corresponding hyperkhler metric @xmath381 a tri - liouville vector field for this hyperkhler structure is @xmath382 by proposition [ prop : hyper2 ] , the equations @xmath232 , @xmath82 , define a taut contact sphere on any transversal @xmath39 to @xmath61 . by construction , @xmath60 is in turn the hyperkhler metric on @xmath41 induced by this taut contact sphere , with the @xmath55-factor now corresponding to the flow lines of @xmath61 . the surface @xmath383 is totally geodesic for the metric @xmath60 , since it is the fixed point set of the isometric involution @xmath384 . the metric on @xmath385 induced by @xmath60 is @xmath386 and , from the well - known formula for computing the gau curvature of a metric in diagonal form , one obtains @xmath387 . this proves that @xmath60 is non - flat . we observe that our ansatz from section [ section : helmholtz ] can also be related to a harmonic function on @xmath3 : write @xmath388 for the laplacian on the unit @xmath1-sphere in @xmath3 , and @xmath389 for the laplacian on @xmath3 . moreover , let @xmath390 be the radial coordinate on @xmath3 . given a differentiable function @xmath391 , we have the relation @xmath392 the last summand reflects the fact that both principal curvatures of @xmath393 are equal to @xmath4 . if @xmath394 is of the form @xmath395 then this last equation simplifies to @xmath396 in that particular case , @xmath394 is homogeneous of degree @xmath4 in @xmath390 , hence @xmath397 is homogeneous of degree @xmath314 . this implies that the vanishing of @xmath398 is sufficient for the vanishing of @xmath397 on all @xmath3 . in conclusion , solutions to our monge - ampre equation @xmath399 are by the preceding discussion and proposition [ prop : helmholtz ] in the appendix in direct correspondence with harmonic functions on @xmath3 that are homogeneous of degree @xmath4 in @xmath390 . finally , notice that both examples admit a tri - hamiltonian vector field : in the example of section [ section : helmholtz ] , this is the vector field @xmath400 , with @xmath401 ; in the example of section [ section : gh ] , it is @xmath402 . as we plan to show in a forthcoming paper , all taut contact spheres with such a symmetry can be related to a gibbons - hawking ansatz , although this relation , in the helmholtz case , is far from straightforward . are there any @xmath4-normalised taut contact spheres on a suitable open domain @xmath39 inducing a non - flat metric @xmath60 on @xmath41 such that the long metric @xmath129 ( see definition [ defn : short - long ] ) induced on @xmath130 is complete ? this kind of completeness question is known as a _ bernstein problem _ , see @xcite . in order to appreciate the difficulty of this question , it is helpful to consider how the ansatz in section [ section : helmholtz ] has to be extended if it is to be of any use in providing an answer . first of all , we ensured the homogeneity of @xmath403 by taking @xmath404 to be homogeneous . however , the example @xmath405 where @xmath406 shows that it is perfectly possible for @xmath403 to be homogeneous in @xmath407 without either @xmath404 or @xmath408 having this property . secondly , we took @xmath404 to be independent of @xmath409 , but to discuss the general case one needs to allow the auxiliary function @xmath410 to depend on all three variables @xmath411 , @xmath412 , and @xmath409 . thirdly , while a potential @xmath404 always exists locally , it need not exist globally . the question we just raised has a venerable history . in @xcite calabi describes a construction of _ khler - einstein _ metrics , and in particular ricci - flat khler metrics , on complex tubular domains @xmath413 , where @xmath414 is some connected , open subset of @xmath415 . the construction is based on a real monge - ampre equation ( with a constant on the right - hand side ) , or equivalently , a complex monge - ampre equation for a function depending only on the real parts of @xmath313 complex variables . the resulting metrics are invariant under the group of translations along @xmath416 . earlier results of calabi @xcite allowed him to show that , in the ricci - flat case , metrics obtained via this construction can never be complete , except for the trivial case with @xmath417 and a flat metric . here is a partial answer to this bernstein problem , dealing with contact spheres that possess additional symmetries . in place of the long metric @xmath129 we consider the short metric @xmath131 . completeness of @xmath131 is a stronger condition than completeness of @xmath129 . the following is part ( a ) of theorem [ thm : bernstein ] . [ thm : bernstein - killing ] let @xmath418 be a @xmath4-normalised taut contact sphere on a @xmath0-manifold @xmath39 with the property that the short metric @xmath419 is complete and admits a killing field @xmath420 that preserves each form , i.e. @xmath421 then @xmath39 is compact and hence a left - quotient of @xmath2 , and @xmath418 is isomorphic to one of the taut contact spheres described in theorem [ thm : class ] . the proof of this theorem will take up the remainder of this section . we begin by observing that @xmath422 does not have any zeros , which can be seen as follows . arguing by contradiction , assume that @xmath423 is a point with @xmath424 . then the flow of @xmath422 preserves the distance spheres from @xmath425 , and is then necessarily a rotation about an axis through @xmath425 ( in geodesic normal coordinates ) . this is incompatible with the fact that this flow preserves the coframe @xmath13 . the fact that @xmath422 does not have any zeros means that @xmath426 defines a function @xmath427 . set @xmath428 so that @xmath429 notice that the @xmath430 satisfy the equations ( [ eqn : structure ] ) with that very @xmath19 , and they are likewise invariant under the flow of @xmath422 . consider the map @xmath431 it is clear that the differential of @xmath432 satisfies @xmath433 , i.e. each flow line of @xmath422 is mapped to a single point in @xmath393 . write @xmath434 for the length of tangent vectors to @xmath39 with respect to the short metric @xmath131 , and @xmath435 for the length of tangent vectors to @xmath436 with respect to the standard metric @xmath437 . [ lem:1 ] if @xmath438 is a tangent vector @xmath131-orthogonal to @xmath422 , then @xmath439 the @xmath422-invariance of the @xmath430 gives , with the cartan formula for the lie derivative , @xmath440 all the following computations are made at the single point @xmath425 . rotate the contact sphere so that at that point @xmath425 we have @xmath441 and @xmath442 , i.e.@xmath443 . then , with ( [ eqn : structure ] ) , @xmath444 the condition that @xmath445 be @xmath131-orthogonal to @xmath422 means that @xmath446 . hence @xmath447 from @xmath443 we have @xmath448 . then @xmath449 this lemma implies in particular that @xmath450 has full rank at every point , so @xmath451 is an open subset of @xmath393 . the map @xmath452 is surjective . since @xmath453 is open , it suffices to show that @xmath451 is complete , i.e. that every path @xmath454 of finite length has a limit point inside @xmath451 . let @xmath455 be a maximal lift of such a path , @xmath131-orthogonal to @xmath422 . by the previous lemma , this lift is non - empty and of finite length in the short metric . since @xmath39 is complete , we deduce that @xmath456 and that the lift @xmath457 has a limit point in @xmath39 . therefore @xmath458 has a limit point in @xmath451 . for each @xmath459 , the preimage @xmath460 is a single orbit of @xmath422 . arguing by contradiction , we assume that @xmath459 is a point in the image of two distinct orbits @xmath461 of @xmath422 . let @xmath462\rightarrow u$ ] be a path joining these two orbits ; @xmath463\rightarrow s^2 $ ] is then a loop based at @xmath215 . by the argument in the proof of the preceding lemma we may assume that @xmath457 is orthogonal to @xmath422 . let @xmath464 , @xmath465 $ ] , be a homotopy of @xmath466 rel @xmath467 to the constant path @xmath468 at @xmath215 , and @xmath469 the corresponding homotopy of lifts orthogonal to @xmath422 with initial point @xmath470 for all @xmath465 $ ] . the endpoints @xmath471 form a smooth path in @xmath472 . the completeness of @xmath473 entails that the flow of @xmath422 is complete , and this in turn ensures that @xmath474 for all @xmath475 . but the constant path @xmath468 lifts to the constant path @xmath476 at @xmath477 , hence @xmath478 , which is impossible . as preimages of single points , the orbits of @xmath422 are closed subsets of @xmath39 . this means that every orbit is either periodic or a proper embedding of @xmath55 in @xmath39 . if there is a periodic orbit @xmath479 , all other orbits remain at a bounded distance from @xmath479 , since the flow of @xmath422 is by isometries . this precludes proper embeddings of @xmath55 , i.e. in this case all orbits must be periodic , and @xmath39 is compact , as asserted in theorem [ thm : bernstein - killing ] . it remains to consider the complementary case , when all orbits are proper embeddings of @xmath55 . in this case , the time-@xmath4 map of the flow of @xmath422 will disjoin any sufficiently small compact set @xmath480 from itself , since each single orbit through @xmath480 is proper and the flow is by isometries . it follows that this time-@xmath4 map defines a free and properly discontinuous @xmath481-action on @xmath39 . the quotient manifold under this action is , by the first case , a compact manifold with a taut contact sphere , and hence a left - quotient of @xmath2 . such manifolds do not admit infinite covers . in other words , this second case can not occur . this completes the proof of theorem [ thm : bernstein - killing ] . in this section we consider taut contact spheres @xmath13 on a @xmath0-dimensional domain @xmath39 that satisfy the stronger condition @xmath482 for @xmath16 . the conditions for a naturally ordered taut contact sphere then imply that @xmath483 for some @xmath484 . differentiation of this equation yields @xmath485 it follows that @xmath486 , so @xmath50 for some constant @xmath49 and the contact sphere is @xmath49-normalised . recall that the * reeb vector field * @xmath487 of a contact form @xmath5 is defined by @xmath488 and @xmath489 . we now have the following simple lemma , where @xmath490 denotes the reeb vector field of @xmath430 , and @xmath18 is the @xmath4-form from equation ( [ eqn : structure ] ) . the proof is left to the reader . for a taut contact sphere @xmath13 , the following conditions are equivalent : * @xmath482 for @xmath16 . * the reeb vector fields @xmath491 constitute a frame dual to the coframe @xmath13 . * @xmath492 . * the short metric @xmath131 equals the long metric @xmath129 . if any of these conditions holds and @xmath49 is the normalisation constant , then one has @xmath493=-cr_k$ ] . this implies that the metric @xmath494 has constant curvature equal to @xmath495 . in particular , the metric @xmath496 has curvature identically equal to @xmath497 . in analogy with our terminology in @xcite we call a taut contact sphere satisfying any of the conditions in this lemma a * cartan structure*. of the taut contact spheres in the statement of theorem [ thm : class ] exactly those with @xmath93 are cartan structures . those with @xmath333 are not even conformally equivalent to a cartan structure ; this follows for instance from ( * ? ? ? we now want to relate such cartan structures to @xmath0-sasakian structures . recall the definition of these structures . a metric @xmath498 on a @xmath0-manifold @xmath39 is called * @xmath0-sasakian * if the cone metric @xmath499 on @xmath41 is a hyperkhler metric . given such a cone metric @xmath500 , we observe that @xmath501 is a homothetic vector field and @xmath502 is a closed @xmath4-form . given a riemannian metric @xmath60 and a vector field @xmath445 , the following are equivalent : * @xmath503 is the identity on every tangent space . * @xmath504 and @xmath505 is closed . * @xmath504 , and where @xmath445 is non - vanishing it is orthogonal to a codimension @xmath4 foliation . * on the open set @xmath506 we have local descriptions @xmath507 with @xmath508 and @xmath498 being the metric induced by @xmath60 in a transversal orthogonal to @xmath445 . following @xcite , we call a vector field satisfying either of these conditions a * dilatation*. thus , any particular description of @xmath60 as a cone metric corresponds to a non - vanishing dilatation . a @xmath4-normalised taut contact sphere on a @xmath0-dimensional manifold @xmath39 is a cartan structure if and only if the induced hyperkhler metric @xmath60 on @xmath41 is a cone metric with the induced tri - liouville vector field @xmath57 as dilatation . by formula ( [ eqn : metric - normal ] ) for the metric @xmath60 , the @xmath4-form @xmath509 equals @xmath510 , which is closed if and only if @xmath114 . given any metric @xmath498 on @xmath39 , any endomorphism field on @xmath41 parallel with respect to the corresponding cone metric is invariant under the flow of @xmath501 . let @xmath511 be local coordinates on @xmath39 . computing the levi - civita connection of the cone metric in the coordinates @xmath512 , one finds that @xmath513 are parallel along the radii . the coefficients of a parallel endomorphism field in this frame are constant along the radii . those coefficients are the same in the frame @xmath514 . it is a fact in riemannian geometry that a family of metrics of the form @xmath515 , @xmath516 , in general gives rise to a family of non - isometric cone metrics . we use the methods of this paper to give a simple proof of the following result , well - known in sasakian geometry , which can be read as saying that for @xmath498 having constant positive curvature , only the @xmath517 of curvature equal to @xmath4 gives rise to a hyperkhler cone . a @xmath0-sasakian metric in dimension @xmath0 has constant curvature equal to @xmath4 . therefore , if a @xmath76-dimensional hyperkhler metric admits a dilatation , it must be flat . let @xmath498 be a riemannian metric on a @xmath0-manifold @xmath39 and assume that the cone metric @xmath518 is hyperkhler . by the preceding lemma the flow of the dilatation @xmath519 is tri - holomorphic and also tri - liouville . introduce the new coordinate @xmath520 , so that @xmath521 . by the theory we have developed , there is a @xmath4-normalised taut contact sphere @xmath13 on @xmath39 such that @xmath522 notice that @xmath498 is recovered from the cone metric and the dilatation @xmath501 via the formula @xmath523 . but @xmath524 , so @xmath525 we see that @xmath498 has a @xmath1-normalised cartan structure as an orthonormal frame , making it a metric of constant curvature equal to @xmath4 . then @xmath526 describes the @xmath76-dimensional euclidean metric in spherical coordinates . this result must be read as _ local _ rigidity of @xmath0-sasakian structures . as our constructions of taut contact spheres giving rise to non - flat hyperkhler metrics show , taut contact spheres have much richer local geometry . in particular , their associated hyperkhler metrics admit homotheties , but no dilatations . so they are not cone metrics in any way . as we have mentioned in section [ section : class ] , the @xmath49-normalised taut contact spheres on closed @xmath0-manifolds are orthonormal for berger metrics , which are spherical only in the cartan case . the long metrics , on the other hand , are always spherical . by the discussion in this section , our theorems [ thm : classold ] and [ thm : class ] can be read as a classification of the closed @xmath0-sasakian @xmath0-manifolds : [ cor : sasaki ] the closed @xmath0-sasakian @xmath0-manifolds are precisely the left - quotients of @xmath2 . for the parametric family of taut contact spheres in theorem [ thm : class ] , the induced hyperkhler metric is always the standard euclidean metric on @xmath261 , so it is always a cone metric . but only for the parameter value @xmath93 ( corresponding to the unique class containing cartan structures ) does the tri - liouville vector field @xmath61 point in the radial direction ( i.e. the direction of the only non - vanishing dilatation of @xmath261 ) . in fact , only the tri - liouville vector field changes within this parametric family . the @xmath527 obviously do not depend on @xmath89 , and a little computation shows that neither do the reeb vector fields @xmath490 of the @xmath430 . a proof of corollary [ cor : sasaki ] previous to the one given above was indeed based on the observation that a @xmath0-sasakian @xmath0-manifold is a space of constant curvature @xmath4 ; the classification of @xmath0-sasakian manifolds among the @xmath0-dimensional space forms was achieved by sasaki @xcite ( who still spoke of _ normal contact metric @xmath0-structures _ ) . the following intriguing proposition relates solutions of equation ( [ eqn : ma2 ] ) from section [ section : helmholtz ] to solutions of the helmholtz equation ( cf . @xcite for this terminology ) @xmath118 on the @xmath1-sphere @xmath393 with its metric of constant curvature @xmath4 . in geodesic polar coordinates this metric is given by @xmath528 hence the gradient of a differentiable function @xmath529 is computed via @xmath530 the area element is @xmath531 , so from @xmath532 we see that the _ spherical _ laplacian @xmath533 takes the form @xmath534 [ prop : helmholtz ] let @xmath356 be a solution of equation @xmath399 . set @xmath535 if @xmath536 , then @xmath537 is a function of the independent variables @xmath538 and @xmath368 that solves the spherical helmholtz equation @xmath539 given @xmath356 , let @xmath545 be defined by @xmath546 . the condition @xmath536 is obviously necessary and sufficient for @xmath547 to be an invertible coordinate transformation . from @xmath548 we see that @xmath106 is a function of @xmath538 and @xmath368 with @xmath549 conversely , given @xmath540 , let @xmath550 be defined by @xmath551 . since @xmath552 the condition @xmath542 is necessary and sufficient for @xmath553 to be a coordinate transformation , because @xmath554 . from @xmath555 we infer that @xmath410 is a function of @xmath556 and @xmath557 with @xmath558 it is then a straightforward check that @xmath546 and @xmath551 are indeed inverse transformations of each other . in particular , one finds that @xmath559 which shows that @xmath546 is defined on @xmath410 if and only if @xmath551 is defined on @xmath106 . in order to show that solutions of ( [ eqn : ma2 ] ) correspond to solutions of ( [ eqn : helmholtz ] ) under the transformation @xmath546 , it is convenient to rewrite ( [ eqn : ma2 ] ) as an exterior differential system @xmath560 we then compute , using @xmath551 and @xmath561 , @xmath562 this completes the proof of proposition [ prop : helmholtz ] . since the reader is bound to wonder how we arrived at the transformation @xmath546 , we present , _ in nuce _ , our chain of discovery : first , one can get rid of mixed derivatives in the exterior differential system @xmath563 by introducing @xmath564 as an independent variable . the identity @xmath565 suggests the contact transformation @xmath566 given by @xmath567 with inverse transformation @xmath568 ( here one needs @xmath536 or @xmath569 , respectively , for these to be honest , i.e. invertible , transformations . ) we compute @xmath570 so the contact transformation takes equation ( [ eqn : ma2 ] ) to the following equation , where we now write @xmath571 for @xmath572 etc . : @xmath573 the first order term in this equation can be made to disappear by setting @xmath574 , @xmath575 and @xmath576 . this turns equation @xmath577 into @xmath578 which is the helmholtz equation for the metric @xmath579 the gau curvature of this metric turns out to be identically equal to @xmath4 . indeed , it is the spherical metric in mercator coordinates . therefore , we pass from ( [ eqn : w ] ) to ( [ eqn : helmholtz ] ) by making the substitution @xmath580 we now want to find an explicit solution of @xmath541 . the ansatz @xmath581 leads to @xmath582 which has the solution @xmath583 ; that in turn integrates to @xmath584 the resulting @xmath106 corresponds under @xmath551 to the solution @xmath361 of equation @xmath399 . as domain of definition we may take @xmath585 the khler potential @xmath404 corresponding to this solution gives rise to a hyperkhler metric @xmath60 on @xmath38 inducing a taut contact sphere on @xmath39 , with @xmath586 the relation with the complex coordinates is given by @xmath587 and @xmath588 , say . we claim that @xmath60 is non - flat . the coefficients of this metric are @xmath589 write @xmath590 for the @xmath185-matrix @xmath591 . then the curvature tensor @xmath592 , read for fixed @xmath593 as a @xmath185-matrix indexed by @xmath5 and @xmath594 , is computed by @xmath595 cf . @xcite . we now want to show that @xmath596 is non - zero . in the following computations we write @xmath597 for any matrix entry that is irrelevant for the final result : @xmath598 on the hyperplane @xmath599 , corresponding to the line @xmath600 , we have @xmath601 writing @xmath602 , for short , we have along that same line @xmath603 at @xmath604 we thus find @xmath605 @xmath606 that last equality can be computed from ( [ eqn : ma2 ] ) . hence , at @xmath607 we obtain @xmath608 which yields @xmath609 . thus @xmath60 is indeed non - flat . we are immensely grateful to nigel hitchin for drawing our attention to the similarities between our original treatment of taut contact spheres and the gibbons - hawking ansatz . hitchin s observation inspired a large part of the present paper . we thank martin lbke for useful hints concerning the curvature calculations in the appendix . 99 atiyah , m. f. , hitchin , n. j. , and singer , i. m. : self - duality in four - dimensional riemannian geometry , _ proc . london ser . a _ * 362 * , 425461 ( 1978 ) bakas , i. and sfetsos , k. : toda fields of @xmath29 hyper - khler metrics and free field realizations , _ internat . j. modern phys . a _ * 12 * , 25852611 ( 1997 ) br , c. : real killing spinors and holonomy , _ comm . _ * 154 * , 509521 ( 1993 ) besse , a. : _ einstein manifolds _ , ergeb . berlin : springer - verlag , 1987 boyer , c. p. : a note on hyper - hermitian four - manifolds , _ proc . soc . _ * 102 * , 157164 ( 1988 ) boyer , c. p. and galicki , k. : @xmath0-sasakian manifolds . in : _ surveys in differential geometry : essays on einstein manifolds _ , surv . vi*. boston : int . press , 1999 , pp . 123184 boyer , c. p. and galicki , k. : _ sasakian geometry_. oxford math . oxford : oxford university press , 2008 calabi , e. : improper affine hyperspheres of convex type and a generalization of a theorem by k. jrgens , _ michigan math . j. _ * 5 * , 105126 ( 1958 ) calabi , e. : examples of bernstein problems for some nonlinear equations . in : _ global analysis _ ( berkeley , 1968 ) , proc . sympos . pure math . * 15*. providence : american mathematical society , 1970 , pp . 223230 calabi , e. : a construction of nonhomogeneous einstein metrics , in : _ differential geometry _ ( stanford , 1973 ) , proc . sympos . pure math . * 27 * , part 2 . providence : american mathematical society , 1975 , pp . 1724 chave , t. , tod , k. p. , and valent , g. : @xmath610 and @xmath611 sigma models with a tri - holomorphic killing vector , _ phys . b _ * 383 * , 262270 ( 1996 ) evans , l. c. : _ partial differential equations _ , grad . stud . math . * 19*. providence : american mathematical society , 1998 geiges , h. : symplectic couples on @xmath76-manifolds , _ duke math . j. _ * 85 * , 701711 ( 1996 ) geiges , h. : normal contact structures on @xmath0-manifolds , _ thoku math . j. _ * 49 * , 415422 ( 1997 ) geiges , h. and gonzalo , j. : contact geometry and complex surfaces , _ invent . _ * 121 * , 147209 ( 1995 ) geiges , h. and gonzalo , j. : contact circles on @xmath0-manifolds , _ j. differential geom . _ * 46 * , 236286 ( 1997 ) geiges , h. and gonzalo , j. : moduli of contact circles , _ j. reine angew _ * 551 * , 4185 ( 2002 ) gibbons , g. w. and hawking , s. w. : gravitational multi - instantons , _ phys . b _ * 78 * , 430432 ( 1978 ) gibbons , g. w. and rychenkova , p. : cones , tri - sasakian structures and superconformal invariance , _ phys . b _ * 443 * , 138142 ( 1998 ) godliski , m. , kopczyski , w. , and nurowski , p. : locally sasakian manifolds , _ classical quantum gravity _ * 17 * , l105l115 ( 2000 ) kobayashi , s. and nomizu , k. : _ foundations of differential geometry ii_. new york : interscience , 1969 sasaki , s. : spherical space forms with normal contact metric @xmath0-structure , _ j. differential geom . _ * 6 * , 307315 ( 1972 )

a taut contact sphere on a @xmath0-manifold is a linear @xmath1-sphere of contact forms , all defining the same volume form . in the present paper we completely determine the moduli of taut contact spheres on compact left - quotients of @xmath2 ( the only closed manifolds admitting such structures ) . we also show that the moduli space of taut contact spheres embeds into the moduli space of taut contact circles . this moduli problem leads to a new viewpoint on the gibbons - hawking ansatz in hyperkhler geometry . the classification of taut contact spheres on closed @xmath0-manifolds includes the known classification of @xmath0-sasakian @xmath0-manifolds , but the local riemannian geometry of contact spheres is much richer . we construct two examples of taut contact spheres on open subsets of @xmath3 with nontrivial local geometry ; one from the helmholtz equation on the @xmath1-sphere , and one from the gibbons - hawking ansatz . we address the bernstein problem whether such examples can give rise to complete metrics .

the dynamical structure of the kuiper belt suggests that the outer solar system experienced a phase of planetesimal - driven migration in its early history @xcite . pluto and other kuiper belt objects that are trapped in mean motion resonances ( mmrs ) with neptune are explained by the outward migration of neptune due to interactions with a more massive primordial planetesimal disk in the outer regions of the solar system @xcite . in addition , the so - called the scattered disk of the kuiper belt can also be explained by the outward migration of neptune @xcite , or by the effects of a high eccentricity phase of ice giant planet evolution during the outward migration of neptune @xcite . the basic premise of planetesimal - driven migration is that the giant planets formed in a more compact configuration than we find them today , and that they were surrounded by a massive ( @xmath9 ) disk of unaccreted icy planetesimals that was the progenitor of the currently observed kuiper belt @xcite . when planetesimals are preferentially scattered either inward ( toward the sun ) or outward ( away from the sun ) , net orbital angular momentum is transferred between the disk and the large body , causing a drift in the large body s semimajor axis @xcite . in many simulations of giant planet migration , icy planetesimals are preferentially scattered inward by each of the three outer giant planets ( saturn , uranus , and neptune ) causing these planets to migrate outward . due to jupiter s large mass , planetesimals that encounter jupiter are preferentially ejected out of the solar system , leading to a net loss of mass from the solar system and an inward migration of jupiter . planetesimal - driven giant planet migration has been suggested as a cause of the late heavy bombardment ( lhb ) @xcite , however the link between these two events has yet to be definitively established @xcite . such migration would have enhanced the impact flux of both asteroids and comets onto the terrestrial planets in two ways . first , many of the icy planetesimals scattered by the giant planets would have crossed the orbits of the terrestrial planets . second , as the giant planets migrated , locations of mean motion and secular resonances would have swept across the asteroid belt , raising the eccentricities of asteroids to planet - crossing values . recently , @xcite showed that the patterns of depletion observed in the asteroid belt are consistent with the effects of sweeping of resonances during the migration of the giant planets . the jupiter - facing sides of some of the kirkwood gaps ( regions of the asteroid belt that are nearly empty due to strong jovian mean motion resonances ) are depleted relative to the sun - facing sides , as would be expected due to the inward migration of jupiter and the associated inward sweeping of the jovian mean motion resonances . the region within the inner asteroid belt between semimajor axis range @xmath2@xmath10 also has excess depletion relative to a model asteroid belt that was uniformly populated and then subsequently sculpted by the gravitational perturbations of the planets over @xmath11 , as would be expected due to the outward migration of saturn and the associated inward sweeping of a strong secular resonance , the so - called @xmath0 resonance , as explained below . in our 2009 study , we concluded that the semimajor axis distribution of asteroids in the main belt is consistent with the inward migration of jupiter and outward migration of saturn by amounts proposed in previous studies based on the kuiper belt resonance structure(e.g . , * ? ? ? however , in that study the migration timescale was not strongly constrained , because only the relative depletion of asteroids in nearby semimajor axis bins could be determined , not their overall level of depletion . in the present paper , we explore in more detail the effect that planet migration would have had on the asteroid belt due to asteroid eccentricity excitation by the sweeping of the @xmath0 secular resonance . from the observed eccentricity distribution of main belt asteroids , we find that it is possible to derive constraints on the secular resonance sweeping timescale , and hence on the migration timescale . secular resonances play an important role in the evolution of the main asteroid belt . the inner edge of the belt nearly coincides with the @xmath0 secular resonance which is defined by @xmath12 , where @xmath13 is the rate of precession of the longitude of pericenter , @xmath14 , of an asteroid and @xmath15 is the sixth eigenfrequency of the solar system planets ( approximately the rate of precession of saturn s longitude of pericenter ) . the @xmath0 resonance , is important for the delivery of near earth asteroids ( neas ) to the inner solar system @xcite . @xcite showed that the location of the @xmath0 resonance actually forms surfaces in @xmath16 space , and @xcite showed that those surfaces approximately define the inner edge " of the main asteroid belt . however , as mentioned above , @xcite found that , with the planets in their present configuration , planetary perturbations over the age of the solar system can not fully account for the detailed orbital distribution of the asteroids in the inner asteroid belt . the pattern of excess depletion in inner asteroid belt noted by @xcite is consistent with the effect of the inward sweeping of the @xmath0 secular resonance . in general , the direction of motion of the @xmath0 is anticorrelated with that of saturn , so an inward sweeping of the @xmath0 would be produced by an outwardly migrating saturn . sweeping , or scanning , secular resonances have been analyzed in a number of previous works . sweeping secular resonances due to the changing quadrupole moment of the sun during solar spin - down have been explored as a possible mechanism for explaining the eccentricity and inclination of mercury @xcite . secular resonance sweeping due to the effects of the dissipating solar nebula just after planet formation has also been investigated as a possible mechanism for exciting the orbital eccentricities of mars and of the asteroid belt @xcite . the dissipating massive gaseous solar nebula would have altered the secular frequencies of the solar system planets in a time - dependent way , causing locations of secular resonances to possibly sweep across the inner solar system , thereby exciting asteroids into the eccentric and inclined orbits that are observed today . this mechanism was revisited by @xcite , who incorporated a more sophisticated treatment of the nebular dispersal . however , @xcite have argued that the excitation ( and clearing ) of the primordial asteroid belt was unlikely due to secular resonance sweeping due to the dispersion of the solar nebula . the special case of asteroids on initially circular orbits being swept by the @xmath0 and @xmath17 resonances has been investigated by @xcite . in this paper , we consider the more general case of non - zero initial eccentricities ; our analysis yields qualitatively new results and provides new insights into the dynamical history of the asteroid belt . this extends the work of @xcite and @xcite in developing analytical treatments of the effects of sweeping secular resonances on asteroid orbits . in doing so , we have developed an explicit relationship between the migration rate of the giant planets , the initial eccentricity of the asteroid and its initial longitude of perihelion , and the final eccentricity of the asteroid after the passage of the resonance . we show that for initially non - zero asteroid eccentricity , the sweeping of the @xmath0 resonances can either increase or decrease asteroid eccentricities . examining the orbits of observed main belt asteroids we find evidence for a double - peaked eccentricity distribution ; this supports the case for a history of @xmath0 sweeping . quantitative comparison of our analytical theory with the semimajor axis and eccentricity distribution of asteroids yields new constraints on the timescale of planet migration . we note that although our analysis is carried out in the specific context of the sweeping @xmath0 resonance during the phase of planetesimal - driven migration of jupiter and saturn , the techniques developed here may be extended to other similar problems , for example , the sweeping of the inclination - node @xmath17 resonance in the main asteroid belt , the @xmath18 secular resonance in the kuiper belt , and farther afield , the sweeping of secular resonances in circumstellar or other astrophysical disks . we adopt a simplified model in which a test particle ( asteroid ) is perturbed only by a single resonance , the @xmath0 resonance . we use a system of units where the mass is in solar masses , the semimajor axis is in units of au , and the unit of time is @xmath19y . with this choice , the gravitational constant , @xmath20 , is unity . an asteroid s secular perturbations close to a secular resonance can be described by the following hamiltonian function @xcite : @xmath21 where @xmath22 describes the phase of the p - th eigenmode of the linearized eccentricity - pericenter secular theory for the solar system planets @xcite , @xmath23 is the associated eigenfrequency , @xmath14 is the asteroid s longitude of perihelion , @xmath24 is the canonical generalized momentum which is related to the asteroid s orbital semimajor axis @xmath25 and eccentricity @xmath26 ; @xmath27 and @xmath28 are the canonically conjugate pair of variables in this 1-degree - of - freedom hamiltonian system . the coefficients @xmath29 and @xmath30 are given by : @xmath31 where the subscript @xmath32 refers to a planet , @xmath33 is the amplitude of the @xmath23 mode in the @xmath34 planet s orbit , @xmath35 , @xmath36 is the ratio of the mass of planet @xmath32 to the sun , and @xmath37 and @xmath38 are laplace coefficients ; the sum is over all major planets . the summations in equations ( [ e : g0def])([e : epsdef ] ) are over the 8 major planets , for the greatest accuracy ; however , we will adopt the simpler two - planet model of the sun - jupiter - saturn in [ s : sweeping ] , in which case we sum over only the indices referring to jupiter and saturn ; then @xmath23 is an eigenfrequency of the secular equations of the two planet system . with fixed values of the planetary masses and semimajor axes , @xmath29 , @xmath39 and @xmath30 are constant parameters in the hamiltonian given in equation ( [ e : reshamiltonian ] ) . however , during the epoch of giant planet migration , the planets semimajor axes change secularly with time , so that @xmath29 , @xmath39 and @xmath30 become time - dependent parameters . in the analysis below , we neglect the time - dependence of @xmath29 and @xmath30 , and adopt a simple prescription for the time - dependence of @xmath39 ( see equation [ e : goft ] below ) . this approximation is physically motivated : the fractional variation of @xmath29 and @xmath30 for an individual asteroid is small compared to the effects of the small divisor " @xmath40 during the @xmath0 resonance sweeping event . it is useful to make a canonical transformation to new variables @xmath41 defined by the following generating function , @xmath42 thus , @xmath43 and @xmath44 . the new hamiltonian function is @xmath45 , @xmath46 where we have retained @xmath28 to denote the canonical momentum , since @xmath47 . it is useful to make a second canonical transformation to canonical eccentric variables , @xmath48 where @xmath49 is the canonical coordinate and @xmath50 is the canonically conjugate momentum . the hamiltonian expressed in these variables is @xmath51 as discussed above , during planetary migration , the secular frequency @xmath23 is a slowly varying function of time . we approximate its rate of change , @xmath52 , as a constant , so that @xmath53 we define @xmath54 as the epoch of exact resonance crossing , so that @xmath55 ( cf . * ? ? ? * ) . then , @xmath56 , and the equations of motion from the hamiltonian of equation ( [ e : reshamiltonian3 ] ) can be written as : @xmath57 these equations of motion form a system of linear , nonhomogenous differential equations , whose solution is a linear combination of a homogeneous and a particular solution . the homogeneous solution can be found by inspection , giving : @xmath58 where @xmath59 and @xmath60 are constant coefficients . we use the method of variation of parameters to find the particular solution . accordingly , we replace the constants @xmath59 and @xmath60 in the homogeneous solution with functions @xmath61 and @xmath62 , to seek the particular solution of the form @xmath63 substituting this into the equations of motion we now have : @xmath64 therefore @xmath65 equations ( [ e : dota ] ) and ( [ e : dotb ] ) do not have a simple closed - form solution , but their solution can be expressed in terms of fresnel integrals @xcite . the fresnel integrals are defined as follows : @xmath66 and have the following properties : @xmath67 therefore @xmath68 we denote initial conditions with a subscript @xmath69 , and write the solution to equations ( [ e : dotxsweep ] ) and ( [ e : dotysweep ] ) as @xmath70+y_i\sin\left[\lambda\left(t^2-t_i^2\right)\right]\notag\\ & \qquad+\frac{\varepsilon}{\sqrt{|\lambda|}}\left[\left(s - s_i\right)\cos\lambda t^2-\left(c - c_i\right)\sin\lambda t^2\right]\label{e : xsoln},\\ y(t)&=-x_i\sin\left[\lambda\left(t^2-t_i^2\right)\right]+y_i\cos\left[\lambda\left(t^2-t_i^2\right)\right]\notag\\ & \qquad-\frac{\varepsilon}{\sqrt{|\lambda|}}\left[\left(c - c_i\right)\cos\lambda t^2+\left(s - s_i\right)\sin\lambda t^2\right]\label{e : ysoln}.\end{aligned}\ ] ] because the asteroid is swept over by the secular resonance at time @xmath54 , we can calculate the changes in @xmath71 by letting @xmath72 and evaluating the coefficients @xmath73 far from resonance passage , i.e. , for @xmath74 , by use of equation ( [ e : fresnellimits ] ) . thus we find @xmath75,\label{e : xf}\\ y_f&=y_i- \varepsilon\sqrt{\frac{\pi}{2|\lambda|}}\left[\cos\lambda t_i^2+\sin\lambda t_i^2\right].\label{e : yf}\end{aligned}\ ] ] the final value of @xmath28 long after resonance passage is therefore given by @xmath76\notag\\ & = j_i+\frac{\pi\varepsilon^2}{2|\lambda|}+\varepsilon\sqrt{\frac{2\pi j_i}{|\lambda|}}\cos(\phi_i-\lambda t_i^2-\frac{\pi}{4}).\label{e : jftf}\end{aligned}\ ] ] with a judicious choice of the initial time , @xmath77 , and without loss of generality , the cosine in the last term becomes @xmath78 , and therefore @xmath79 the asteroid s semimajor axis @xmath25 is unchanged by the secular perturbations ; thus , the changes in @xmath28 reflect changes in the asteroid s eccentricity @xmath26 . for asteroids with non - zero initial eccentricity , the phase dependence in equation ( [ e : jf ] ) means that secular resonance sweeping can potentially both excite and damp orbital eccentricities . we also note that the magnitude of eccentricity change is inversely related to the speed of planet migration . in linear secular theory ( equation ( [ e : reshamiltonian ] ) ) , eccentricity and inclination are decoupled , and therefore the effect of the sweeping @xmath0 does not depend on the inclination . however , as @xcite showed , the location of the @xmath0 does depend on inclination , but the dependence is weak for typical inclinations of main belt objects . nevertheless , there are populations of main belt asteroids at high inclination ( such as the hungaria and phocaea families ) , and an analysis of secular resonance sweeping that incorporates coupling between eccentricity and inclination would be valuable for understanding the effects of planet migration on these populations ; we leave this to a future investigation . for small @xmath26 , we can use the approximation @xmath80 . considering all possible values of @xmath81 , an asteroid with initial eccentricity @xmath82 that is swept by the @xmath0 resonance will have a final eccentricity in the range @xmath83 to @xmath84 , where @xmath85 and @xmath86 equations ( [ e : jf])([e : deltaedef ] ) have the following implications : 1 . initially circular orbits become eccentric , with a final eccentricity @xmath87 . an ensemble of orbits with the same @xmath25 and initial non - zero @xmath26 but uniform random orientations of pericenter are transformed into an ensemble that has eccentricities in the range @xmath83 to @xmath84 ; this range is not uniformly distributed because of the @xmath78 dependence in equation ( [ e : jf ] ) , rather the distribution peaks at the extreme values ( see fig . [ f : tpideal ] below ) . an ensemble of asteroids having an initial distribution of eccentricities which is a single - peaked gaussian ( and random orientations of pericenter ) would be transformed into one with a double - peaked eccentricity distribution . in light of the above calculations , it is possible to conclude that if the asteroid belt were initially dynamically cold , that is asteroids were on nearly circular orbits prior to secular resonance sweeping , then the asteroids would be nearly uniformly excited to a narrow range of final eccentricities , the value of which would be determined by the rate of resonance sweeping . because asteroids having eccentricities above planet - crossing values would be unlikely to survive to the present day , it follows that an initially cold asteroid belt which is uniformly excited by the @xmath0 sweeping will either lose all its asteroids or none . on the other hand , an initially excited asteroid belt , that is a belt with asteroids that had non - zero eccentricities prior to the @xmath0 sweeping , would have asteroids final eccentricities bounded by equation ( [ e : finalebounds ] ) , allowing for partial depletion and also broadening of its eccentricity distribution . in this section , we apply our theoretical analysis to the problem of the @xmath0 resonance sweeping through the asteroid belt , and compare the theoretical predictions with the observed eccentricity distribution of asteroids . in order to apply the theory , we must find the location of the @xmath0 resonance as a function of the semimajor axes of the giant planets orbits , and also obtain values for the parameter @xmath30 ( equation [ e : epsdef ] ) , for asteroids with semimajor axis values in the main asteroid belt . the location of the @xmath0 resonance is defined as the semimajor axis , @xmath88 , where the rate , @xmath29 ( equation [ e : g0def ] ) , of pericenter precession of a massless particle ( or asteroid ) is equal to the @xmath15 eigenfrequency of the solar system . in the current solar system , the @xmath15 frequency is associated with the secular mode with the most power in saturn s eccentricity pericenter variations . during the epoch of planetesimal - driven planet migration , jupiter migrated by only a small amount but saturn likely migrated significantly more @xcite , so we expect that the variation in location of the @xmath0 secular resonance is most sensitive to saturn s semimajor axis . we therefore adopt a simple model of planet migration in which jupiter is fixed at @xmath89 and only saturn migrates . we neglect the effects of the ice giants uranus and neptune , as well as secular effects due to the previously more massive kuiper belt and asteroid belt . with these simplifications , the @xmath15 frequency varies with time as saturn migrates , so we parametrize @xmath15 as a function of saturn s semimajor axis . in contrast with the variation of @xmath15 , there is little variation of the asteroid s apsidal precession rate , @xmath29 , as saturn migrates . thus , finding @xmath88 is reduced to calculating the dependence of @xmath15 on saturn s semimajor axis . to calculate @xmath15 as a function of saturn s semimajor axis , we proceed as follows . for fixed planetary semimajor axes , the laplace - lagrange secular theory provides the secular frequencies and orbital element variations of the planets . this is a linear perturbation theory , in which the disturbing function is truncated to secular terms of second order in eccentricity and first order in mass @xcite . in the planar two - planet case , the secular perturbations of planet @xmath32 , where @xmath90 is jupiter and @xmath91 is saturn , are described by the following disturbing function : @xmath92 , \label{e : secham}\ ] ] where @xmath93 is the mean motion , and @xmath94 is a matrix with elements @xmath95 for @xmath96 , @xmath97 , and @xmath98 ; @xmath99 , and @xmath100 the secular motion of the planets is then described by a set of linear differential equations for the eccentricity vectors , @xmath101 , @xmath102 for fixed planetary semimajor axes , the coefficients are constants , and the solution is given by a linear superposition of eigenmodes : @xmath103 where @xmath23 are the eigenfrequencies of the matrix @xmath94 and @xmath33 are the corresponding eigenvectors ; the amplitudes of the eigenvectors and the phases @xmath104 are determined by initial conditions . in our 2-planet model , the secular frequencies @xmath105 and @xmath15 depend on the masses of jupiter , saturn , and the sun and on the semimajor axes of jupiter and saturn . for the current semimajor axes and eccentricities of jupiter and saturn the laplace - lagrance theory gives frequency values @xmath106 and @xmath107 , which are lower than the more accurate values given by @xcite by @xmath108 and @xmath109 , respectively @xcite . @xcite achieved their more accurate solution by incorporating higher order terms in the disturbing function involving @xmath110 , which arise due to jupiter and saturn s proximity to the 5:2 resonance ( the so - called `` great inequality '' ) . by doing an accurate numerical analysis ( described below ) , we found that the effect of the 5:2 resonance is only important over a very narrow range in saturn s semimajor axis . more significant is the perturbation owing to the 2:1 near - resonance of jupiter and saturn . @xcite developed corrections to the laplace - lagrange theory to account for the perturbations from @xmath111 resonances in the context of the uranian satellite system . applying that approach to our problem , we find that the 2:1 near - resonance between jupiter and saturn leads to zeroth order corrections to the elements of the @xmath94 matrix , where @xmath26 is eccentricity and @xmath93 is the order of the resonance . the 5:2 is a third order resonance , so its effect is @xmath112 . the 2:1 is a first order resonance , so that its effect does not depend on @xmath26 . therefore the discrepancy between linear theory and numerical analysis ( or the higher order theory of @xcite ) arising from the great inequality would be much less if jupiter and saturn were on more circular orbits , but the effect due to the 2:1 resonance would remain . ] . including these corrections , we determined the secular frequencies for a range of values of saturn s semimajor axis ; the result for @xmath15 is shown in fig . [ f : g6vsasat]a ( dashed line ) . we also calculated values for the eccentricity - pericenter eigenfrequencies by direct numerical integration of the full equations of motion for the two - planet , planar solar system . in these integrations , jupiter s initial semimajor axis was @xmath89 , saturn s semimajor axis , @xmath113 , was one of 233 values in the range @xmath114@xmath115 , initial eccentricities of jupiter and saturn were @xmath116 , and initial inclinations were zero . the initial longitude of pericenter and mean anomalies of jupiter were @xmath117 and @xmath118 , and saturn were @xmath119 and @xmath120 . in each case , the planets orbits were integrated for 100 myr , and a fourier transform of the time series of the @xmath121 yields their spectrum of secular frequencies . for regular ( non - chaotic ) orbits , the spectral frequencies are well defined and are readily identified with the frequencies of the secular solution . the @xmath15 frequency as a function of saturn s semimajor axis was obtained by this numerical analysis ; the result is shown by the solid line in fig . [ f : g6vsasat]a . the comparison between the numerical analysis and the laplace - lagrange secular theory indicates that the linear secular theory , including the corrections due to the 2:1 near - resonance , is an adequate model for the variation in @xmath15 as a function of @xmath113 . we adopted the latter for its convenience in the needed computations . the value of @xmath88 as a function of saturn s semimajor axis was thus found by solving for the value of asteroid semimajor axis where @xmath122 ; @xmath29 was calculated using equation ( [ e : g0def ] ) and @xmath15 is the eigenfrequency associated with the @xmath123 eigenmode ( at each value of saturn s semimajor axis ) . the result is shown in fig . [ f : g6vsasat]b . we also used the analytical secular theory to calculate the eigenvector components @xmath124 in the secular solution of the 2-planet system , for each value of saturn s semimajor axis . we adopted the same values for the initial conditions of jupiter and saturn as in the direct numerical integrations discussed above . finally , we computed the values of the parameter @xmath30 at each location @xmath88 of the secular resonance . the result is plotted in fig . [ f : epsvsanu6 ] . despite the complexity of the computation , the result shown in fig . [ f : epsvsanu6 ] is approximated well by a simple exponential curve , @xmath125 , in the semimajor axis range @xmath126 . we checked the results of our analytical model against four full numerical simulations of the restricted four - body problem ( the sun - jupiter - saturn system with test particle asteroids ) in which the test particles in the asteroid belt are subjected to the effects of a migrating saturn . the numerical integration was performed with an implementation of a symplectic mapping @xcite , and the integration stepsize was @xmath127 . jupiter and saturn were the only massive planets that were integrated , and their mutual gravitational influence was included . the asteroids were approximated as massless test particles . the current solar system values of the eccentricity of jupiter and saturn were adopted and all inclinations ( planets and test particles ) were set to zero . an external acceleration was applied to saturn to cause it to migrate outward linearly and smoothly starting at @xmath128 at the desired rate . as shown in fig . [ f : g6vsasat]b , the @xmath0 resonance was located at @xmath129 at the beginning of the simulation , and swept inward past the current location of the inner asteroid belt in all simulations . in each of the four simulations , 30 test particles were placed at @xmath130 and given different initial longitudes of pericenter spaced @xmath131 apart . the semimajor axis value of @xmath130 was chosen because it is far away from the complications arising due to strong mean motion resonances . the only parameters varied between each of the four simulations were the initial osculating eccentricities of the test particles , @xmath82 , and the migration speed of saturn , @xmath132 . the parameters explored were : 1 . @xmath133 , @xmath134 ; 2 . @xmath133 , @xmath135 ; 3 . @xmath136 , @xmath134 ; 4 . @xmath137 , @xmath134 . these values were chosen to illustrate the most relevant qualitative features . the migration rates of @xmath138 and @xmath139 are slow enough so that the change in eccentricity is substantial , but not so slow that the non - linear effects at high eccentricity swamp the results . these test cases illustrate both how well the analytical model matches the numerical results , and where the analytical model breaks down . two aspects of the analytical model were checked . first , the perturbative equations of motion , equations ( [ e : dotxsweep ] ) and ( [ e : dotysweep ] ) , were numerically integrated , and their numerical solution compared with the numerical solution from the direct numerical integration of the full equations of motion . for the perturbative solution , we adopted values for @xmath140 that were approximately equivalent to the values of @xmath132 in the full numerical integrations . second , the eccentricity bounds predicted by the analytical theory , equation ( [ e : finalebounds ] ) , were compared with both numerical solutions . the results of these comparisons for the four test cases are shown in fig . [ f : phasetest ] . we find that the analytically predicted values of the maximum and minimum final eccentricities ( shown as horizontal dashed lines ) are in excellent agreement with the final values of the eccentricities found in the numerical solution of the perturbative equations , and in fairly good agreement with those found in the full numerical solution . not surprisingly , we find that the test particles in the full numerical integrations exhibit somewhat more complicated behavior than the perturbative approximation , and equation ( [ e : finalebounds ] ) somewhat underpredicts the maximum final eccentricity : this may be due to higher order terms in the disturbing function that have been neglected in the perturbative analysis and which become more important at high eccentricity ; effects due to close encounters with jupiter also become important at the high eccentricities . does the eccentricity distribution of main belt asteroids retain features corresponding to the effects of the @xmath0 resonance sweeping ? to answer this question , we need to know the main belt eccentricity distribution free of observational bias , and also relatively free of the effects of @xmath141 of collisional evolution subsequent to the effects of planetary migration . we therefore obtained the proper elements of the observationally complete sample of asteroids with absolute magnitude @xmath142 from the astdys online data service @xcite ; we excluded from this set the members of collisional families as identified by @xcite . these same criteria were adopted in @xcite in a study of the long term dynamical evolution of large asteroids . this sample of 931 main belt asteroids is a good approximation to a complete set of large asteroids that have been least perturbed by either dynamical evolution or collisional evolution since the epoch of the last major dynamical event that occurred in this region of the solar system ; therefore this sample likely preserves best the post - migration orbital distribution of the asteroid belt . the proper eccentricity distribution of these asteroids is shown in fig . [ f : mba - big - dist ] . this distribution has usually been described in the literature by simply quoting its mean value ( and sometimes a dispersion ) @xcite . our best fit single gaussian distribution to this data has a mean , @xmath143 and standard deviation , @xmath144 , given by @xmath145 and @xmath146 , and is plotted in fig . [ f : mba - big - dist ] . however , we also note ( by eye ) a possible indication of a double - peak feature in the observed population . our best fit double gaussian distribution ( modelled as two symmetrical gaussians with the same standard deviation , but different mean values ) to the same data has the following parameters : @xmath147 more details of how these fits were obtained are described in appendix [ sec : sweeprate - appendix - fitting ] , where we also discuss goodness - of - fit of the single and double gaussian distributions . a kolmogorov - sinai test shows that there is only a 4.5% probability that the observed eccentricities actually are consistent with the best - fit single gaussian distribution , and a 73% probability that they are consistent with the best - fit double gaussian distribution ; while the double gaussian is apparently a better fit to the data compared to the single gaussian , a k - s probability of only 73% is quite far from a statistically significant level of confidence . a dip test for multi - modality in the observed eccentricity distribution @xcite is also inconclusive ; the likely reason for this is that our data sample is not very large , and the separation between the two putative peaks is too small in relation to the dispersion . nevertheless , we bravely proceed in the next section with the implications of a double - gaussian eccentricity distribution , with the caveat that some of the conclusions we reach are based on this statistically marginal result . by relating the @xmath15 secular frequency to the semimajor axis of saturn , @xmath148 can be related to the migration rate of saturn , @xmath132 . in this section only the effects of the sweeping @xmath0 secular resonance will be considered , and effects due to sweeping jovian mean motion resonances will be ignored . because jupiter is thought to have migrated inward a much smaller distance than saturn migrated outward during planetesimal - driven migration , the effects due to migrating jovian mmrs were likely confined to narrow regions near strong resonances @xcite . in the inner asteroid belt between @xmath2@xmath3 , these would include the 3:1 and 5:2 resonances , currently located at approximately @xmath10 and @xmath149 , respectively . as shown in fig . [ f : g6vsasat]b , plausible parameters for the outward migration of saturn would have allowed the @xmath0 to sweep across the entire inner asteroid belt . therefore the @xmath0 resonance would have been the major excitation mechanism across the @xmath2@xmath3 region of the main belt ( and possibly across the entire main asteroid belt , depending on saturn s pre - migration semimajor axis ) during giant planet migration . we used the results of our analytical model to set limits on the rate of migration of saturn assuming a linear migration profile , with the caveat that many important effects are ignored , such as asteroid - jupiter mean motion resonances , and jupiter - saturn mean motion resonances ( with the exception of the 2:1 resonance ) . we have confined our analysis to only the region of the main belt spanning @xmath2@xmath3 . beyond @xmath3 strong jovian mean motion resonance become more numerous . due to the high probability that the icy planetesimals driving planet migration would be ejected from the solar system by jupiter , jupiter likely migrated inward . the migration of jupiter would have caused strong jovian mean motion resonances to sweep the asteroid belt , causing additional depletion beyond that of the sweeping @xmath0 resonance @xcite . a further complication is that sweeping jovian mean motion resonances may have also trapped icy planetesimals that entered the asteroid belt region from their source region beyond neptune @xcite . the effects of these complications are reduced when we consider only the inner asteroid belt . from fig . [ f : g6vsasat]b , we find that the @xmath0 would have swept the inner asteroid belt region between @xmath2@xmath3 when saturn was between @xmath150@xmath151 . therefore the limits on @xmath152 that we set using the inner asteroid belt as a constraint are only applicable for this particular portion of saturn s migration history . our theoretically estimated final eccentricity as a function of initial asteroid semimajor axis and eccentricity is shown in fig . [ f : sweepratefigs ] for three different adopted migration rates of saturn . the larger the initial asteroid eccentricities , the wider the bounds in their final eccentricities . if we adopt the reasonable criterion that an asteroid is lost from the main belt when it achieves a planet - crossing orbit ( that is , crossing the orbits of either jupiter or mars ) and that initial asteroid eccentricities were therefore confined to @xmath153 , then from fig . [ f : sweepratefigs ] saturn s migration rate must have been @xmath154 when the @xmath0 resonance was sweeping through the inner asteroid belt . our results indicate that if saturn s migration rate had been slower than @xmath155 when it was migrating across @xmath150@xmath151 , then the inner asteroid belt would have been completely swept clear of asteroids by the @xmath0 resonance . in light of our analysis and the observed dispersion of eccentricities in the asteroid belt ( fig . [ f : mba - big - dist ] ) , we can also immediately conclude that the pre - migration asteroid belt between @xmath2 and @xmath3 had significantly non - zero eccentricities . this is because , as discussed at the top of [ s : sweeping ] , an initially cold asteroid belt swept by the @xmath0 resonance would either lose all its asteroids or none , and very low initial eccentricities would result in final asteroid eccentricities in a very narrow range of values ( cf . equation [ e : finalebounds ] ) , in contradiction with the fairly wide eccentricity dispersion that is observed . this conclusion supports recent results from studies of planetesimal accretion and asteroid and planet formation that the asteroids were modestly excited at the end of their formation ( e.g. , * ? ? ? * ) . in appendix [ sec : sweeprate - appendix - fitting ] we show that the double - gaussian distribution is a slightly better fit to the main belt asteroid eccentricity distribution , but the statistical tests do not rule out a single - peaked distribution . we boldly proceed with considering the implications of the double - peaked eccentricity distribution to further constrain the migration rate of saturn , with the caveat that these results can only be said to be consistent with the observations , rather than uniquely constrained by them . if the pre - sweeping asteroid belt had a gaussian eccentricity distribution , then the lower peak of the post - sweeping asteroid belt should be equal to the lower bound of equation ( [ e : finalebounds ] ) . we use the analytical theory to make a rough estimate of the parameter @xmath140 ( and hence @xmath156 ) that would yield a final distribution with lower peak near @xmath157 and upper peak near @xmath158 ( which is similar to the best - fit double gaussian in fig . [ f : mba - big - dist ] ) . applying equation ( [ e : finalebounds ] ) , we see that there are two possible solutions : @xmath159,@xmath160 and @xmath161,@xmath162 . a corresponding migration rate of saturn can be estimated from the value of @xmath87 using equation ( [ e : deltaedef ] ) , and the parameter relationships plotted in figs . [ f : g6vsasat ] and [ f : epsvsanu6 ] . the former solution ( @xmath160 ) requires a migration rate for saturn of @xmath163 . we mention this implausible solution here for completeness , but we will not discuss it any further . the latter solution ( @xmath162 ) requires a migration rate for saturn of @xmath164 . we dub this solution the `` cold belt '' solution . this rate is comparable to the rates of planet migration found in the `` jumping jupiter '' scenario proposed by @xcite . a third solution exists if we consider that eccentricities in the main belt are restricted by the orbits of mars and jupiter on either side , such that stable asteroid orbits do not cross the planetary orbits . this limits asteroid eccentricities to values such that neither the aphelion of the asteroid crosses the perihelion distance of jupiter , nor the perihelion of the asteroid crosses the aphelion distance of mars . maximum asteroid eccentricity is therefore a function of semimajor axis , where @xmath165 , where @xmath166 and @xmath167 are planet aphelion and perihelion , respectively , and @xmath25 is the semimajor axis of the asteroid . in this case , an initial single gaussian eccentricity distribution with a mean greater than @xmath7 would be severely truncated , therefore we need only fit the lower peak of the double gaussian distribution at @xmath168 . applying equation ( [ e : finalebounds ] ) , we find that @xmath169 provides a good fit . the corresponding migration rate of saturn is @xmath170 . we dub this solution the `` hot belt '' solution . we illustrate the two possible solutions for an ensemble of hypothetical asteroids having semimajor axes uniformly distributed randomly in the range @xmath171 to @xmath3 . in fig . [ f : tpdist]a the initial eccentricity distribution is modeled as a gaussian distribution with a mean @xmath161 and a standard deviation of @xmath172 . this initial standard deviation was chosen so that the final standard deviation would be the same as that of the observed main belt . [ f : tpdist]b shows the eccentricity distribution after @xmath0 resonance sweeping has occurred due to the migration of saturn at a rate of @xmath173 . the final distribution was calculated with equation ( [ e : jf ] ) ; we used values of @xmath30 shown in fig . [ f : epsvsanu6 ] , and the value of @xmath140 was calculated with the aid of fig . [ f : g6vsasat ] which relates the value of @xmath23 to the semimajor axis of saturn . as expected , when an ensemble of asteroids with a single - peaked eccentricity distribution is subjected to the sweeping secular resonance , the result is a double - peaked eccentricity distribution . because of the slight bias towards the upper limit of the eccentricity excitation band , proportionally more asteroids are found in the upper peak . in fig . [ f : tpdist]c the initial eccentricity distribution is modeled as a truncated gaussian : a gaussian with mean @xmath174 and standard deviation @xmath175 , but truncated at the semimajor axis dependent mars - crossing value . we used equation ( [ e : jf ] ) to calculate the eccentricity distribution of this hypothetical ensemble after @xmath0 resonance sweeping with @xmath176 . again , allowing that only those asteroids whose final eccentricities are below the mars - crossing value will remain , the resulting post - migration eccentricity distribution is shown in fig . [ f : tpdist]d . in this case , we find the lower peak at the same eccentricity value as the lower peak in the observed main belt distribution ( see fig . [ f : mba - big - dist ] ) . in both cases of possible solutions ( initially cold main belt with @xmath161 and @xmath164 ; and initially hot main belt with @xmath174 and @xmath170 ) , the theoretical models yield an excess of asteroids with eccentricities greater than @xmath177 than in the observed main belt . however , as shown by @xcite , on gigayear timescales , the @xmath178 population of the asteroid belt is dynamically more unstable than the @xmath179 population . thus , both solutions may be consistent with the observations , as post - sweeping dynamical erosion could result in a final eccentricity distribution resembling more closely the observed distribution . the estimates of saturn s migration rate quoted above depend strongly on the eccentricities of the giant planets during their migration . in deriving the above estimates , we adopted the present values of the giant planets orbital eccentricities . the @xmath0 resonance strength coefficient @xmath30 ( equation [ e : epsdef ] ) is proportional to the amplitude of the @xmath123 mode , which is related to the eccentricities of the giant planets ( namely saturn and jupiter ) . from equation ( [ e : hkvector ] ) , and the definition @xmath101 , the value of @xmath33 is a linear combination of the eccentricities of the giant planets . because saturn is the planet with the largest amplitude of the @xmath123 mode , from equation ( [ e : jf ] ) the relationship between the sweep rate and the value of saturn s eccentricity is approximately @xmath180 . therefore , to increase the limiting timescale by a factor of ten would only require that the giant planets eccentricities were @xmath181 their current value ( i.e. , @xmath182 ) . based on the existence of the inner asteroid belt , we conclude that saturn s migration rate must have been @xmath183 as saturn migrated from @xmath184 to @xmath151 ( as the @xmath0 resonance migrated from 2.8 au to 2.1 au ) . migration rates lower than @xmath1 would be inconsistent with the survival of any asteroids in the inner main belt , as the @xmath0 secular resonance would have excited asteroid eccentricities to planet - crossing values . this lower limit for the migration rate of saturn assumes that jupiter and saturn had their current orbital eccentricities ; the migration rate limit is inversely proportional to the square of the amplitude of the @xmath15 secular mode ; if jupiter and saturn s eccentricities were @xmath181 their current value ( i.e. , @xmath182 ) during the planet migration epoch , the limit on the migration rate decreases by a factor of @xmath185 . ( this caveat also applies to the migration rate limits quoted below . ) our analysis of secular resonance sweeping predicts that a single - peaked eccentricity distribution will be transformed into a double - peaked eccentricity by secular resonance sweeping . we find that the observed eccentricity distribution of asteroids may be consistent with a double - peaked distribution function , although we acknowledge that the statistics are poor . we used a double - gaussian function to model the observed asteroid eccentricity distribution to set even tighter , albeit model - dependent , constraints on the migration rate of saturn . we identified two possible migration rates that depend on the pre - migration dynamical state of the main asteroid belt . the first , the `` cold primordial belt '' solution , has an asteroid belt with an initial eccentricity distribution modeled as a gaussian with @xmath161 ; saturn s migration rate of @xmath173 yields a final eccentricity distribution consistent with the observed asteroid belt . the second , the `` hot primordial belt '' solution , has an asteroid belt with an initial eccentricity distribution modeled as a gaussian with @xmath186 , but truncated above the mars - crossing value of eccentricity ; in this case , saturn s migration rate of @xmath187 is generally consistent with the observed asteroid belt . each of these solutions has very different implications for the primordial excitation and depletion of the main asteroid belt . the cold belt solution , with @xmath164 , would lead to little depletion of the asteroid belt during giant planet migration , as the @xmath0 resonance would be unable to raise eccentricities to mars - crossing values . this implies an initial dynamically quite cold asteroid belt with not more than @xmath188 times the mass of the current main belt ; the latter estimate comes from accounting for dynamical erosion over the age of the solar system @xcite . the hot belt solution , with @xmath170 , would lead to loss of asteroids directly due to excitation of eccentricities above planet - crossing values , by about a factor of @xmath188 . in this case , the main asteroid belt was more dynamically excited prior to resonance sweeping than we find it today . this implies much greater loss of asteroids prior to @xmath0-sweeping , as the peak of the eccentricity distribution would be near the mars - crossing value , and subject to strong dynamical erosion @xcite . each of these solutions has different implications for the model that the late heavy bombardment of the inner solar system is linked to the epoch of planetesimal - driven giant planet migration @xcite . these implications will be explored in a future paper . we remind the reader that in order to elucidate the effects of @xmath0 resonance sweeping , we have made a number of simplifying assumptions to arrive at an analytically tractable model . these simplifications include neglecting the effects of planets other than jupiter and saturn , the effects of sweeping jovian mean motion resonances on asteroids , the effects of a presumed massive kuiper belt during the epoch of planet migration , and the self - gravity and collisional interactions of a previously more massive asteroid belt . in addition , our analysis was carried out in the planar approximation , thereby neglecting any eccentricity - inclination coupling effects . these neglected effects can be expected to reduce somewhat the lower limit on saturn s migration speed that we have derived , because in general they would reduce the effectiveness of the @xmath0 in exciting asteroid eccentricities . perhaps more importantly , giant planet migration would also lead to the sweeping of the main asteroid belt by the @xmath17 inclination secular resonance @xcite whose effects could be used to infer additional constraints . recently , @xcite found through numerical simulations with slow rates of planet migration ( e - folding timescales exceeding 5 myr ) that the surviving asteroids in the main belt tend to be clumped around mean motion resonances . the semimajor axis distribution of survivors is found very different from the observed distribution for asteroids , and also that a large proportion of survivors have inclinations above 20@xmath189 , both inconsistent with the current main asteroid belt ; comparisons made with the unbiased main belt asteroid distributions as described in @xcite . this is likely due to the effects of the sweeping @xmath17 inclination - longitude of ascending node secular resonance , analogous to the sweeping @xmath0 eccentricity - pericenter secular resonance that we analyzed in the present study . while the effects of the sweeping @xmath17 resonance are analogous to the @xmath0 , only affecting inclinations instead of eccentricities , a full analysis of the asteroid belt inclinations is beyond the scope of the present work , but will be explored in a future study . a number of other studies have derived limits on the speed of planetesimal - driven giant planet migration . @xcite exclude an @xmath190folding migration timescale @xmath191 to @xmath192 confidence based on the lack of a large observed asymmetry in the population of kuiper belt objects in the two libration centers of the 2:1 neptune mean motion resonance . @xcite exclude @xmath193 based on the observed obliquity of saturn . the latter lower limit on the migration timescale is slightly incompatible with the lower limit on the rate of saturn s migration of @xmath194 we derive based on the existence of the inner asteroid belt . one way these can be reconciled is if saturn s orbital eccentricity were a factor @xmath188 smaller than its present value as it migrated from 8.5 au to 9.2 au ; then , some mechanism would need to have increased saturn s eccentricity up to its present value by the time saturn reached its present semimajor axis of @xmath195 . the authors would like to thank the anonymous reviewer and the editor eric feigelson for useful comments . this research was supported in part by nsf grant no . ast-0806828 and nasa : nessf grant no . nnx08aw25h . the work of david minton was additionally partially supported by nasa nlsi / cloe research grant the binned eccentricity distribution may be modeled as a gaussian probability distribution function , given by : @xmath196 , \label{e : gaussian}\ ] ] where @xmath197 is the standard deviation , @xmath198 is the mean , and @xmath49 is the random variable ; in our case @xmath49 is the eccentricity . with an appropriate scaling factor , equation ( [ e : gaussian ] ) can be used to model the number of asteroids per eccentricity bin . however , rather than fit the binned distribution , we instead perform a least squares fit of the unbinned sample to the gaussian cumulative distribution function given by : @xmath199 for the eccentricities of our sample of 931 main belt asteroids , the best fit parameters are : @xmath200 we also fit the data to a double - gaussian distribution , @xmath201+\exp\left[-\frac{(x-\mu'_2)^2}{2\sigma'^2}\right]\right\}. \label{e : doublegaussian}\ ] ] the cumulative distribution function for equation ( [ e : doublegaussian ] ) is @xmath202 . \label{e : doublecdf}\ ] ] for the eccentricities of our sample of 931 main belt asteroids , we performed a least squares fit to equation ( [ e : doublecdf ] ) and obtained the following best - fit parameters : @xmath147 we evaluated the goodness of fit using the kolmogorov - smirnov ( k - s ) test . the k - s test determines the probability that two distributions are the same , or in our case how well our model distributions fit the observed data @xcite . the k - s test compares the cumulative distribution of the data against the model cumulative distribution function . we found that our asteroid sample has a probability of @xmath203 that it comes from the best fit single gaussian ( equation ( [ e : cdf ] ) ) , but a probability of @xmath204 that it comes from the double - gaussian ( equation ( [ e : doublecdf ] ) ) . therefore , the k - s tests indicate that the double - gaussian is a better fit to the data than the single - gaussian . we performed hartigan s dip test @xcite to test whether the observational data is consistent with a multi - peaked distribution . hartigan s dip test calculates the probability that the distribution being tested has a single peak . applying hartigan s dip test to a given distribution yields in a test statistic ; together with the sample size , the test statistic is matched to a p - value range in a precomputed table provided by @xcite . the p - value is a measure of the probability that the distribution actually has only one peak ( the null - hypothesis , for this problem ) . the smaller the calculated p - value , the less likely is the distribution to have a single peak and the more likely it is to have at least two significant peaks . a p - value of @xmath205 indicates that the null - hypothesis is very unlikely , and that the given distribution has more than one peak . we applied the dip test to the eccentricity distribution of our sample of 931 main belt asteroids ; we determined that the test statistic is @xmath206 . from @xcite , this corresponds to a p - value range of @xmath207 ( based on a sample size of 1000 ) . this indicates that the hartigan dip test does not rule out the null hypothesis , i.e. , a single - peaked distribution can not be ruled out . to further aid the interpretation of this test , we compare this result to that obtained by applying hartigan s dip test to synthetic distributions that were explicitly double - peaked by construction . ten model distributions ( of 1000 eccentricity values ) were generated from the double - gaussian function of equation ( [ e : doublecdf ] ) , with parameter values @xmath208 , @xmath209 , and @xmath210 ( same parameter values as the best - fit for our sample of asteroids ) . only the random seed was varied between each model distribution . applying the hartigan dip test , we find that the test statistic ranged between @xmath211 and @xmath212 , corresponding to p - values between @xmath213 and @xmath214 . this means that , according to the dip test , the null hypothesis i.e. , a single - peaked distribution could not be ruled out for any of the model distributions ( since none had @xmath215 ) , despite the fact that they were each drawn from an explicitly double - peaked distribution by construction . this illustrates that even with a sample size of nearly 1000 , the dispersion in eccentricity around the two peaks is too large in comparison to the distance between the peaks , that the dip test is not sufficiently sensitive to detect the double - peaked underlying distribution . we interpret this to mean that the results of both the k - s test and hartigan s dip test indicate that the main asteroid belt eccentricity distribution is consistent with being drawn from a double - peaked distribution , but that this can not be definitely shown . ) . asteroids swept by the @xmath0 resonance can have a range of final eccentricities depending on their apsidal phase , @xmath216 . the outermost shaded region demarcates the range of final eccentricities for asteroids with an initial eccentricity @xmath217 . the innermost shaded region demarcates the range of final eccentricities for asteroids with an initial eccentricity @xmath133 . the solid line at the center of the shaded regions is the final eccentricity for an asteroid with an initial eccentricity @xmath218 . , title="fig : " ] ) . asteroids swept by the @xmath0 resonance can have a range of final eccentricities depending on their apsidal phase , @xmath216 . the outermost shaded region demarcates the range of final eccentricities for asteroids with an initial eccentricity @xmath217 . the innermost shaded region demarcates the range of final eccentricities for asteroids with an initial eccentricity @xmath133 . the solid line at the center of the shaded regions is the final eccentricity for an asteroid with an initial eccentricity @xmath218 . , title="fig : " ] ) . asteroids swept by the @xmath0 resonance can have a range of final eccentricities depending on their apsidal phase , @xmath216 . the outermost shaded region demarcates the range of final eccentricities for asteroids with an initial eccentricity @xmath217 . the innermost shaded region demarcates the range of final eccentricities for asteroids with an initial eccentricity @xmath133 . the solid line at the center of the shaded regions is the final eccentricity for an asteroid with an initial eccentricity @xmath218 . , title="fig : " ]

we calculate the eccentricity excitation of asteroids produced by the sweeping @xmath0 secular resonance during the epoch of planetesimal - driven giant planet migration in the early history of the solar system . we derive analytical expressions for the magnitude of the eccentricity change and its dependence on the sweep rate and on planetary parameters ; the @xmath0 sweeping leads to either an increase or a decrease of eccentricity depending on an asteroid s initial orbit . based on the slowest rate of @xmath0 sweeping that allows a remnant asteroid belt to survive , we derive a lower limit on saturn s migration speed of @xmath1 during the era that the @xmath0 resonance swept through the inner asteroid belt ( semimajor axis range @xmath2@xmath3 ) . this rate limit is for saturn s current eccentricity , and scales with the square of saturn s eccentricity ; the limit on saturn s migration rate could be lower if saturn s eccentricity were lower during its migration . applied to an ensemble of fictitious asteroids , our calculations show that a prior single - peaked distribution of asteroid eccentricities would be transformed into a double - peaked distribution due to the sweeping of the @xmath0 . examination of the orbital data of main belt asteroids reveals that the proper eccentricities of the known bright ( @xmath4 ) asteroids may be consistent with a double - peaked distribution . if so , our theoretical analysis then yields two possible solutions for the migration rate of saturn and for the dynamical states of the pre - migration asteroid belt : a dynamically cold state ( single - peaked eccentricity distribution with mean of @xmath5 ) linked with saturn s migration speed @xmath6 , or a dynamically hot state ( single - peaked eccentricity distribution with mean of @xmath7 ) linked with saturn s migration speed @xmath8 .

there has been some interest in the application of on shell perturbation theory to simple scattering problems in one dimension @xcite . those approaches were mainly based on logarithmic perturbation theory@xcite and the dalgarno lewis perturbation theory@xcite . in order to bypass the problem posed by the nodes of real wavefunctions they used complex wavefunctions ( jost functions ) . although the method of dalgarno and lewis was proposed as a device to calculate sum over states@xcite it proves to be an efficient approach for on shell perturbation theory@xcite . this method has recently been applied to solve the differential equation for the density distribution in scattering problems@xcite . the purpose of this paper is to discuss the application of the dalgarno and lewis perturbation theory to potential scattering using real wavefunctions . we also consider the choice of the unperturbed or reference model in order to have convergent perturbation series for the phase shift . in sec . [ sec : model ] we introduce the model . in sec . [ sec : method ] we propose an alternative application of the method of dalgarno and steward and a choice of the unperturbed model potential . in sec . [ sec : example ] we apply the main equations to an illustrative example , and in sec . [ sec : conclusions ] we summarize the main results of the paper and draw some conclusions . for simplicity we consider the one dimensional model @xmath0 and choose the boundary condition @xmath1 in order to mimic the s states of a central field model . the prime denotes differentiation with respect to the coordinate . the potential energy function @xmath2 is negative for @xmath3 and zero for @xmath4 . in order to derive a dimensionless differential equation we define the new independent variable @xmath5 and rewrite eq . ( [ eq : schro ] ) as @xmath6y(x ) \label{eq : y"}\ ] ] where @xmath7 the boundary conditions are @xmath8 where @xmath9 is the phase shift and @xmath10 . in order to avoid ambiguities regarding to multiples of @xmath11 we choose @xmath12 to satisfiy @xmath13 . in order to apply perturbation theory we choose a closely related problem @xmath14y_{0}(x ) \label{eq : y0"}\ ] ] with similar boundary conditions @xmath15 as it is custommary in most applications of perturbation theory we write @xmath16 , where the perturbation parameter @xmath17 is set equal to unity at the end of the calculation , and expand @xmath18 for simplicity we write @xmath19 , and @xmath20 . the function @xmath21 satisfies the differential equation @xmath22 where @xmath23 . since the nodes of @xmath24 and @xmath25 do not coincide one expects @xmath26 to have poles at the nodes of @xmath25 . however , @xmath26 does not appear in the resulting expression for @xmath24 ( see below ) and , consequently , that problem does not arise in practical applications of the method of dalgarno and lewis . notice that @xmath27 is the wronskian of the perturbed and unperturbed solutions . on integrating eq . ( [ eq : diff_eq_f ] ) we have @xmath28 at @xmath29 we substitute the asymptotic forms of @xmath24 and @xmath25 into the wronskian and obtain an expression for the phase shift @xmath30 if we integrate eq . ( [ eq : wronsk2 ] ) between @xmath29 and @xmath31 and multiply the result by @xmath25 we obtain an expression for @xmath24 : @xmath32 where @xmath33 is an integration constant . notice that this expression is free from poles because the function @xmath26 does not appear explicitly in it . when @xmath34 then @xmath35 and we substitute eq . ( [ eq : sin_phase ] ) for the second integral in eq . ( [ eq : y(x ) ] ) . moreover , in order to have the correct asymptotic expression for @xmath24 the integration constant should be @xmath36 that we also expand in a power series : @xmath37 finally , it follows from equations ( [ eq : sin_phase ] ) , ( [ eq : y(x ) ] ) , and ( [ eq : c ] ) that @xmath38 @xmath39 and @xmath40 at the jth perturbation step we first solve for @xmath41 and @xmath42 and then calculate @xmath43 . those expressions are sufficient for present discussion , one can easily derive as many as necessary from the equations above . the number of necessary perturbation corrections depends on the convergence rate of the perturbation series . for that reason it is important to choose a convenient unperturbed or reference potential @xmath44 . it has been argued that in order to have an adequate rate of convercence the unperturbed and perturbed potentials should support the same number of bound states@xcite . this conclusion is based on levinson s theorem than in our case takes the form @xmath45 , where @xmath46 is the number of bound states@xcite . here we investigate the application of the bargmann schwinger upper limit to the number of bound states @xcite and require that at least @xmath47 although this condition does not completely guarantee that both potentials have the same number of bound states , it is a reasonable choice that will prove sound in the example below . the simplest exactly solvable unperturbed model is one with a constant potential @xmath48 in which case @xmath49 the starting point of the calculation are the unperturbed solutions @xmath50 where we choose @xmath51 so that @xmath52 as stated above . in what follows we compare the perturbation results with accurate numerical phase shifts calculated by a straightforward power series method . to this end we assume that the taylor series of the potential around @xmath53 , @xmath54 converges at @xmath29 , and expand the solution in the same way : @xmath55 , where @xmath56 and @xmath57 , @xmath58 . finally , we obtain the phase shift as @xmath59 as an illustrative example , we choose @xmath60 that can be solved in terms of confluent hypergeometric functions@xcite . however , here we resort to the power series approach because it is suitable for more general problems and converges fast in most cases . the power series approach also enables us to obtain the bound state energies from the conditon @xmath61 we choose a well of depth @xmath62 as unperturbed model and the simple condition ( [ eq : bs_limit ] ) gives us @xmath63 . we could obtain analytical expressions for @xmath64 , @xmath65 and @xmath66 but we only show the first perturbation correction because the other ones are too long . @xmath67 \sin \left ( \sqrt{k^{2}+v_{0}}\right ) \cos \left ( \sqrt{k^{2}+v_{0}}\right ) \right . \nonumber \\ & & \left . + \sqrt{k^{2}+v_{0}}\left [ 3a\sin \left ( \sqrt{k^{2}+v_{0}}\right ) ^{2}-a\left ( k^{2}+v_{0}+3\right ) \right . \nonumber \\ & & \left . + 6v_{0}\left ( k^{2}+v_{0}\right ) \right ] \right\ } \label{eq : model_delta1}\end{aligned}\ ] ] in order to test the accuracy of perturbation theory we choose parabolic wells with @xmath68 with no bound states , and @xmath69 with one bound state . figures 1 and 2 show the logarithmic error @xmath70 , where @xmath71 and @xmath72 denote power series and perturbation theory , respectively . we clearly appreciate that the perturbation series converges for both parabolic well strengths , and for all energies . for some @xmath73 values the perturbation series of order @xmath74 may be more accurate than the one of order @xmath75 . we do not know if this behaviour is fortuitous or if there is a mathematical or physical reason behind . table [ tab : err_av ] shows the average relative error @xmath76/\delta _ { ps}(k_{j})| $ ] for some @xmath77values . we clearly appreciate that the perturbation series converges smoothly . one can easily apply the method of dalgarno and steward to more general problems of the form @xmath78 where @xmath79 denotes a differential operation on @xmath24 . we choose a convenient unperturbed problem @xmath80 and proceed as in sec . [ sec : model ] . we obtain the following general equation @xmath81 where we can choose the arbitrary integration limits @xmath82 and @xmath83 and constants @xmath84 , and @xmath85 , conveniently according to the problem . notice that if @xmath86 then @xmath87 is a solution of @xmath88 and @xmath89 . that is to say , @xmath25 and @xmath90 are two independent solutions of the unperturbed model . it is not difficult to verify that the method of dalgarno and lewis applies to velocity dependent problems@xcite . if we choose @xmath91y(x)+\lambda \rho ( x)y^{\prime \prime } ( x ) \nonumber \\ & & -\lambda \left [ y^{\prime } ( x)-\frac{y(x)}{x}\right ] \rho ^{\prime } ( x ) \label{eq : q_jag}\end{aligned}\ ] ] then we have the model treated by jaghoub@xcite . therefore , if @xmath92y_{0}(x ) \label{eq : q0_jag}\ ] ] we have @xmath93 \rho ^{\prime } y_{0}. \label{eq:(qy0-q0y)_jag}\end{aligned}\ ] ] if we expand @xmath24 as in eq . ( [ eq : pt_series ] ) and @xmath94 then we obtain jaghoub s perturbation equations@xcite provided that we choose the arbitrary constants conveniently . in other words : jaghoub s procedure is merely the method of dalgarno and lewis for bound states developed in a different way . by means of the equations displayed above , the reader can easily verify that present implementation of the method of dalgarno and lewis is also suitable for the treatment of scattering problems with velocity dependent interactions@xcite . for @xmath68 . , width=340 ] [ fig : a6 ] for @xmath69 . , width=340 ] [ fig : a18 ] we have shown that the method of dalgarno and lewis is suitable for the calculation of phase shifts by means of perturbation theory . one can use real eigenfunctions because their nodes do not cause the occurrence of poles into the solutions . the constant potential is a suitable unperturbed or reference model for the application of perturbation theory to some problems . a simple prescription for the well depth leads to reasonable convergence rate and enables one to obtain accurate results with perturbation series of low order .

we apply the method of dalgarno and lewis to scattering states and discuss the choice of the unperturbed model in order to have a convergent perturbation series for the phase shift .

an understanding of the structure of the proton in terms of partons , i.e. quarks and gluons which follow qcd dynamics , was established to a large extent in lepton - proton , in particular electron - proton , deep inelastic scattering ( dis ) . at the hera @xmath23 collider with an electron - proton centre - of - mass energy of @xmath24=320 gev , effects of the electroweak sector can also be tested in dis in the region where the squared four momentum transfer @xmath25 or @xmath26 ( @xmath27 and @xmath28 are the @xmath29 and @xmath22 boson masses ) . in addition , signals of new physics beyond the standard model may be expected to arise at the highest @xmath2 where the proton structure is probed at smallest distances . deep inelastic scattering , fig . [ fig1 ] , depends on three variables , @xmath2 , @xmath6 and @xmath30 , where @xmath31 and @xmath32 is the fraction of the proton momentum carried by the struck quark and @xmath33 is the inelasticity of the interaction . the variables are related via @xmath34 . the electron - proton neutral current ( nc ) and charged current ( cc ) cross sections can each be expressed in terms of three proton structure functions . for neutral current interactions mediated by a @xmath35 or @xmath29 boson , the proton structure function @xmath9 dominates the cross section and is proportional to the sum of the quark momentum distributions weighted by the quark charge squared . the longitudinal structure function @xmath10 is directly sensitive to the gluon momentum distribution . the structure function @xmath36 is related to valence quarks only . these three proton structure functions have been measured at hera and confronted with the standard model predictions . the gluon momentum distribution and the strong coupling @xmath12 determined in a qcd analysis of the hera inclusive data have been cross checked in more exclusive channels such as charm and jet production via boson gluon fusion . the propagator mass @xmath28 has been determined from the @xmath2 dependence of the cc interactions and can be compared with direct @xmath28 mass measurements in @xmath37 and @xmath38 interactions . thus , measurements at hera provide tests of both components of the theory , the electroweak sector and the evolution of the parton densities as predicted by perturbative qcd . the first @xmath16 measurements@xcite in @xmath23 interactions at hera , based on a luminosity of 0.025 pb@xmath39 collected in 1992 , revealed at low @xmath6 a pronounced rise of @xmath16 with decreasing @xmath6 which generated a new interest in qcd in dis . since then the collected luminosity has been increased by more than three orders of magnitude . in addition , the h1 and zeus detectors have been continuously improved . in 1995/1996 new detector components were installed by h1 : the backward calorimeter spacal , the backward drift chamber bdc and the backward silicon tracker bst . in 1995 the zeus collaboration installed a beam pipe calorimeter ( bpc ) and in 1997 a beam pipe tracker ( bpt ) was added , covering very small angles of the scattered electron . these improvements and the increase in luminosity made it possible to extend the measured region@xcite to very low @xmath40 gev@xmath5 , covering the transition region from dis to real photoproduction ( @xmath41 ) , as well as very high @xmath42 gev@xmath5 . the variable @xmath6 ranges from @xmath7 to 0.65 in the valence quark region . @xmath43 measurements@xcite at hera together with fixed target experiments are shown in fig . [ f2 ] as a function of @xmath2 for different values of @xmath6 . the h1 and zeus data agree very well in the full range accessible by the hera experiments . at very low @xmath44 , the hera data overlap and agree with the data of the fixed target experiments . in a large part of the phase space the precision of the data is @xmath45 which is comparable with that of fixed target experiments . ( 0,0 ) ( 185,200)(a ) ( 185,75)(b ) ( 250,180)(c ) ( 0,0 ) ( 165,220)(a ) ( 400,100)(b ) the results on the reduced nc cross section , defined as the cross section divided by the kinematical factor @xmath46 with @xmath47 and @xmath48 , the fine structure constant , are shown in fig . [ redsec ] as function of @xmath6 for @xmath49 gev@xmath5 . at very high @xmath30 ( @xmath50 ) , i.e. the lowest x points at fixed @xmath2 , a turn over is visible in the measured cross section which is due to the longitudinal structure function @xmath51 . from these high - y cross sections , @xmath51 was extracted@xcite as shown in fig . this extends the knowledge of the longitudinal structure function to much lower @xmath6 than available from fixed target lepton - proton scattering experiments . the increase of @xmath52 towards low @xmath6 is consistent with the qcd calculation , reflecting the rise of the gluon momentum distribution towards low @xmath6 . another quantity related to the gluon density is the derivative @xmath53 which is shown as a function of @xmath6 in fig . [ deriv]a , b for different @xmath2 . the derivatives show a continuous rise towards low @xmath6 for fixed @xmath2 which is well described by the qcd fit to the h1 data down to @xmath54 gev@xmath5 . the shape of @xmath53 reflects the behaviour of the gluon distribution in the associated kinematic range . the rise of @xmath16 with decreasing @xmath6 can be characterised by @xmath55 where @xmath56 . in fig . [ deriv]c , values of @xmath57 are shown from fits at fixed @xmath2 to the @xmath16 data . the prominent rise of @xmath16 at large @xmath2 values is in agreement with the qcd fit@xcite . for smaller @xmath2 , the rise of @xmath16 becomes less steep and @xmath57 approaches a value of @xmath58 expected from the regge motivated fits@xcite . the high precision of the data allows a simultaneous determination of the gluon distribution and the strong coupling constant @xmath59 in a next - to - leading - order ( nlo ) dglap@xcite qcd fit@xcite by combining the low@xmath6 cross section data of h1 with @xmath60 scattering data of the bcdms collaboration@xcite at large @xmath6 . the gluon distribution from this fit is shown in fig . [ xg ] for @xmath2=5 , 20 , and 200 gev@xmath5 . the inner error band represents the experimental uncertainty of the determination of @xmath61 for @xmath59 fixed . the simultaneous determination of @xmath62 and @xmath59 leads to a small increase of the experimental error of @xmath63 as is illustrated by the middle error band . the full error band includes in addition the uncertainties connected with the fit ansatz ( `` model uncertainty '' ) . the dglap analysis of the data leads to a gluon distribution which rises dramatically at small @xmath6 with increasing @xmath2 . a value of the strong coupling constant @xmath64 = @xmath65 is obtained . the value of @xmath59 changes by about 0.005 , much more than the experimental error , if the renormalisation scale is allowed to vary by a factor of four in the fit . this theoretical uncertainty is expected to be reduced significantly in next - to - next - to - leading - order ( nnlo ) perturbation theory . ( 0,0 ) ( 185,200)(a ) ( 408,200)(b ) exclusive channels such as charm or jet production via boson gluon fusion can also be used for a direct extraction of the gluon density . the charm contribution to @xmath16 , @xmath15 , is determined at hera using measurements of @xmath66 production and account for @xmath67 of @xmath16@xcite as shown in fig . the @xmath15 results are in agreement with expectations calculated using the gluon momentum distribution as determined in the nlo qcd fit to the inclusive cross sections . jet production data@xcite could be used to determine the gluon distribution using @xmath12 from other experiments , or determine the strong coupling constant @xmath12 using parton momentum distributions from global qcd fits . the resulting gluon distribution at the factorisation scale @xmath68 gev is shown in fig . [ f2c]b and consistent with the gluon from the global qcd fits . the results for @xmath12 from jet studies by h1 and zeus collaborations together with @xmath12 determined from scaling violations by h1 are shown in fig . [ alphas ] in comparison with the world average@xcite . ( 0,0 ) ( 193,205)(a ) ( 404,205)(b ) ( 0,0 ) ( 35,200)(a ) ( 270,200)(b ) ( 0,0 ) ( 185,210)(a ) ( 408,210)(b ) purely electroweak effects become observable at hera in a region of very high @xmath2 comparable with the masses squared of the vector bosons @xmath29 and @xmath22 . especially interesting in this respect is the comparison of neutral and charged current cross sections@xcite shown in fig . [ dsdq2 ] . at low @xmath2 , the cross section of the cc process , which is due to exchange of the @xmath22 boson , is suppressed compared to that of the nc process due to the different propagator terms . at high @xmath69 , both cross sections are approaching each other manifesting the unification of the weak and electromagnetic forces . electroweak effects are also visible in the comparison of the @xmath20 and @xmath19 nc cross sections . due to @xmath29 exchange , at very high @xmath2 the nc cross section is predicted to be suppressed in @xmath19 and enhanced in @xmath20 interactions . this effect becomes visible for @xmath701000 gev@xmath5 as shown in fig . [ dsdq2]a , [ redsechix ] and agrees with the standard model prediction . the difference of the nc @xmath19 and @xmath20 cross sections at large @xmath2 has been used to extract the structure function @xmath71 in the range @xmath72 and @xmath73 gev@xmath5 as shown in fig.[xf3 ] . since at hera the dominant contribution to @xmath71 is due to photon-@xmath29 interference , the h1 collaboration evaluated the corresponding structure function @xmath74 . the result is shown in fig . [ xf3gz ] . the integral @xmath75 is found to be consistent within experimental errors with theoretical expectation of 1.11@xcite . the reduced cc cross sections , defined in analogy to the nc case as the cross section divided by @xmath76 with the fermi coupling constant @xmath77 , are shown in fig . [ ccxsec]a . the results are well described by the standard model expectations . at high @xmath6 the @xmath20 and @xmath19 cross sections are dominated by the valence quark contributions , @xmath78 and @xmath79 respectively . for points where the expected @xmath80 contribution to the total cross section is larger than 70% a local extraction method@xcite is used to determine the valence quark densities : @xmath81 where @xmath82 is the measured nc or cc @xmath83 double differential cross sections and the ratio is the theoretical expectation from the h1 97 pdf fit@xcite . the extracted parton densities shown in fig . [ ccxsec]b are rather independent of the theoretical input and are in agreement with global fits . within the standard model , cc interactions are mediated by the @xmath84-channel exchange of a @xmath22 boson , and therefore are sensitive to the @xmath22 mass in the space - like regime . the propagator mass @xmath85 is fitted to the double differential cc cross section data and yields for @xmath19 and @xmath20 respectively@xcite @xmath86 @xmath87 these values are consistent with direct measurements in the time - like domain from lep and tevatron . the present tests of the standard model show good agreement of predictions of perturbative qcd and electroweak theory with the hera data . the luminosity upgrade of the hera collider will bring the opportunity to improve the precision of such tests , especially for the electroweak sector at high @xmath2 where the precision of the measurements is still limited by statistics . h1 collab . , c. adloff et al . , measurement of @xmath66 meson production and the charm contribution to the proton structure @xmath15 in deep inelastic scattering at hera , contrubuted paper to ichep 2000 , osaka , japan , july 2000 ; + zeus collab . ; j.breitweg et al . , eur . j. c12 ( 2000 ) 35 . zeus collab . , measurement of high-@xmath2 neutral current cross section in @xmath20 dis and a first measurement of the structure function @xmath88 at hera ; + zeus collab . , measurement of high-@xmath2 charged current cross section in @xmath20 deep inelastic scattering at hera , contributed papers to ichep 2000 , osaka , japan , july 2000 .

both components of the standard model , qcd and the electroweak sector , are tested in the h1 and zeus experiments at hera . the inclusive @xmath0 and @xmath1 single and double differential cross sections for neutral and charged current processes are measured in the range of four - momentum transfer squared , @xmath2 , between @xmath3 and @xmath4 gev@xmath5 , and bjorken @xmath6 between @xmath7 and @xmath8 . the neutral current double differential cross section , from which the proton structure function @xmath9and the longitudinal structure function @xmath10are extracted , is measured at low @xmath6 with typically 1% statistical and 2 - 3% systematic uncertainties . in a next - to - leading order ( nlo ) dglap qcd analysis using the h1 measurements and the @xmath11 data of the bcdms collaboration the strong coupling constant @xmath12 and the gluon momentum distribution are simultaneously determined . a value of @xmath13 is obtained , with an additional theoretical uncertainty of about @xmath14 , mainly due to the uncertainty of the renormalisation scale . @xmath15 , the charm contribution to @xmath16 , is measured and well described by the boson gluon fusion production mechanism using the gluon distribution from the nlo qcd analysis . the gluon distribution and the strong coupling constant @xmath12 are also determined from jet production . for @xmath17 @xmath18 , an asymmetry between @xmath19 and @xmath20 neutral current scattering is observed and the structure function @xmath21 is extracted . a fit to the charged current data is used to determine a value for the @xmath22 boson propagator mass . all data are found to be in good agreement with standard model predictions . = 14.5pt _ talk given at xv rencontres de physique de la valle daoste , la thuile , italy , + march 4 - 10 , 2001 _ mpi phe/200113 + july 2001

@xcite first reported and characterized penumbral microjets ( pjs ) from high - spatial - resolution , high - cadence h - line chromospheric images of sunspots obtained by the solar optical telescope / filtergraph ( sot / fg : * ? ? ? * ) onboard the _ hinode _ @xcite satellite . the h - line is formed in the chromosphere at a temperature of about 10@xmath0 k or less . pjs are scattered , constantly occurring , transient jet events in sunspot penumbrae , with lifetimes of less than a minute , lengths typically between 1000 and 4000 km , some can be longer up to 10000 km , and with widths of less than 600 km . their apparent speed is more than 100 km s@xmath1 . because they are faint and transient , with an enhanced brightness of 1020% as compared to background penumbra , pjs are more clearly visible in running difference images in comparison to direct intensity images . the fact that pjs were more clearly visible in the limbward side of the penumbra and least visible in the disk - center side of the penumbra , evidently due to foreshortening , led @xcite to conclude that the jets are aligned to the more vertical component of the penumbral magnetic structure . the penumbral magnetic field is a combination of spines ( more vertical field ) and interspines ( more horizontal field ) , first classified by @xcite ; see also @xcite for further details on the two components of sunspot penumbral magnetic field . @xcite observed an increase in the inclination of pjs towards the outer edge of penumbra from an average inclination of pjs of 35@xmath2 ( with respect to the local normal line ) at the umbral - penumbral boundary to 70@xmath2 at the penumbral / quiet sun boundary . this inclination change of pjs is compatible with the average field inclination change with radial distance in sunspots , see e.g. , figure 7 of @xcite for radial variation of field inclination in a sunspot . from a space - time plot , @xcite found some pjs to form near bright penumbral grains @xcite . however , this remained to be established statistically . @xcite found it difficult to locate pjs with respect to penumbral filaments . @xcite proposed that pjs could be produced as a result of magnetic reconnection between the two magnetic components ( spines and interspines ) . in this picture , the jets are oriented in the direction of the spine field , travel upward along it , and are rooted at an edge of a filament , between the field in a spine and the more horizontal field ( along the central axis ) in the filament . recent magnetohydrodynamic ( mhd ) simulations support the idea of magnetic reconnection driving these events , either induced by strong outflows along horizontal flux tubes @xcite , or by assuming the horizontal field in a twisted flux tube @xcite . an alternative is given by @xcite , in which shocks caused by reconnection between neighboring penumbral filaments can produce pjs in a manner that appears consistent with the observations of @xcite . in a recent investigation of internal structure of a sunspot s penumbral filaments by applying a spatially coupled depth - dependent inversion code @xcite on spectropolarimetric data of hinode ( sot / sp ) , @xcite found that the penumbral filaments ( interspines ) resemble elongated convection cells , and behave as stretched magnetized granules . the bright penumbral grains are the heads of penumbral filaments ( the head is the filament s end closer to the umbra ) . strong upflows , with a field of the polarity of that of the umbra , are observed in the heads of penumbral filaments . the upflow continues along the horizontal axis of the filament for more than half its length outward , and weakens with length . a stronger downflow is observed at the tails of filaments , which contain field of opposite polarity to that of the umbra and heads of filaments . weak downflows are also observed at the edges of filaments along the length on the sides of the upflows . these lateral downflows also often contain field of opposite polarity to that of the umbra and heads of filaments . this opposite polarity field was found in one third of the total filaments studied by @xcite . also see , @xcite and @xcite for observational reports on the presence of opposite polarity field in the sides of penumbral filaments , which was also obtained in three - dimensional magnetohydrodynamic ( 3d mhd ) simulations by @xcite . thus , keeping the above picture of a penumbral filament s internal structure in mind , the scenario of formation of pjs ( proposed by * ? ? ? * ) e.g. , by reconnection between two components of penumbral magnetic field , inclined to each other at an acute angle , should be modified . in section [ sec1 ] , we present a modified picture of the magnetic configuration for the production of pjs . the other main issue that we address in this paper is whether pjs have any transition region ( tr ) or coronal signatures . the estimation of chromospheric thermal energy ( 3/2 @xmath3 ) of a pj as estimated by @xcite based on the following numbers : n = 10@xmath4 , k@xmath5 = 1.38@xmath6 , t = 10@xmath0 k , @xmath7 = 2000 km @xmath8 ( 300 km)@xmath9 , returns a value of 2@xmath810@xmath10 j , or 2@xmath810@xmath11 erg , which is of the order of that of a coronal nanoflare @xcite . based on this estimation it was suggested by @xcite that pjs have potential to appear or display some signatures in the corona , and to contribute in some ways to the coronal heating in active regions . to test the above hypothesis , one needs coronal observations of sunspot penumbrae at very high spatial resolution because , as mentioned before , the widths of these pjs are in the range 150 600 km ( 400 km or less according to @xcite ) , close to and at the the resolution limit of the telescope ( sot ) . coronal observation at such a high resolution was not available until the high - resolution coronal imager ( hi - c : * ? ? ? * ) obtained images of an active region ( ar ) in a narrow passband filter centered at 193 at a high spatial resolution of about 150 km , which is also the approximate resolution of hinode ( sot / fg ) . fortunately , as a result of coordinated observations , hi - c and hinode both observed a part of a penumbra of the hi - c ar 11520 for 1.75 minutes . although the part of the observed penumbra was on the disk - center side , which is not the best location for the visibility of pjs , the penumbral field was twisted far from the disk - center direction , so that we could observe several pjs in this part of penumbra within this short time period of observation . we have extended the study for one hour by using different atmospheric imaging assembly channels . we notice a few locations where larger jets are repeatedly produced . to investigate the magnetic structure and origin of these larger penumbral jets , we have also used the co - temporal stokes - v data obtained by sot / fg in the field of view ( fov ) of the sot / fg movie and g - band images of the penumbra . to identify pjs in the penumbra of the noaa ar 11520 ( @xmath12 x @xmath13125 , y @xmath13325 ) observed by hi - c @xcite , we have used h - line broadband filtergraph images ( centered at 3968 ) obtained by sot @xcite onboard _ hinode _ @xcite . we use observations taken between 18:53 and 19:53 ut on 11 july 2012 , which covers 1.75 minutes ( 18:53:44 18:55:30 ut ) overlap with the five minutes observations of hi - c . hi - c observed at high - resolution the corona of ar 11520 in a narrow wavelength range centered at 193 . the cadence of the hi - c 193 images is 5 s. the cadence of the images varies frame to frame from 5 s to 15 s. the spatial resolution of both telescopes is about 150 km . g - band images from sot / fg are used to identify the locations of feet of pjs in the photosphere . heads of filaments appear brighter and tails darker in g - band images . we have also used stokes - v images obtained by sot / fg to examine the magnetic polarity at the locations of some larger jets , discussed later . both the g - band and stokes - v images have a cadence of about 50 s. in addition , we use one - hour - long ( 18:53 19:53 ut ) image series from 1600 , 304 , 171 , 193 , and 94 passbands observed by the atmospheric imaging assembly ( aia : * ? ? ? * ) onboard the _ solar dynamics observatory ( sdo ) _ spacecraft . the pixel size of aia is 0.6 arcsec , so these images detect but do not resolve the jets . the cadence of the aia 1600 channel is 24 s , and the cadence is 12 s for each of the other aia channels used in the present study . all the data used in the paper are calibrated and , whenever applicable , co - registered by using solarsoft routines . the 193 ( aia and hi - c ) and aia 94 bands are both predominantly coronal @xcite . the 193 band particularly detects at about 1.5 mk , but also has some response to the transition region emission of @xmath14 k plasma @xcite . the 94 channel allows mostly hot emissions and is centered on an line ( @xmath15 k ) , but also detects some line emission from fe ions formed at @xmath16 k @xcite ; see also @xcite . there is no known cool tr contamination in the 94 channel . the 1600 aia passband , which primarily passes lower - chromospheric continuum emission , also covers the two lines near 1550 formed at t @xmath17 10@xmath18 k in the transition region ( tr ) . the transient brightenings in the 1600 band are due to these lines , and hence are from the lower tr @xcite . the 304 passband observes the upper chromosphere and lower tr at @xmath12 50,000 k , in emission primarily from . the 171 band observes emission primarily from formed in the upper tr at 6 @xmath19 k. in this section , we first present results on detection of tr signatures but no definite coronal signatures of pjs from the analysis of hinode ( sot / fg ) and hi - c data sets . then we present results from analysis of hinode ( sot / fg ) and sdo / aia data of one hour . we discover larger penumbral jets than normal ones reported earlier by @xcite . we characterize these larger pjs and their magnetic setting . in figure [ f2](a ) , we display a full fov hi - c 193 image with a box on it outlining the fov of the penumbra studied , which is enlarged in panel ( b ) . a co - temporal image of the same fov is shown in panel ( c ) . a movie ` movie1.mp4 ' , connected to this figure , contains four panels with two intensity panels of hi - c and , respectively , and two panels of corresponding running difference images , for 1.75 minutes ( during 18:53:44 18:55:30 ) when hi - c and fg simultaneously looked at the sunspot penumbra . as mentioned before , both instruments have a spatial resolution of about 150 km , which provides a unique opportunity for comparison of the jets at two different heights and temperatures . at least 10 jets can be observed in the movie ` movie1.mp4 ' of 1.75 minutes ; they are especially visible in the running difference movie . most of the pjs have brightness enhancement by 10 20% of penumbral background , however some of the largest jets have enhancement up to 30% . in the hi - c 193 movie and the corresponding running difference movie , one can notice many bright dots ( bds ) that move toward or away the spot center along filaments in the penumbra . these bds are reported and characterized in detail by @xcite . however no signatures of pjs are noticeable in the hi - c 193 movie . in figure [ f3 ] we display two examples of jets from start to end , clearly seen in the running difference images ( middle row ) . to look for any coronal signatures of them , corresponding hi - c images and their running difference images are shown in the two upper rows . we have also displayed intensity and g - band images in the last two rows of the figure [ f3 ] . as noted from the arrows in g - band images it is difficult to identify exactly where pjs are rooted relative to heads of penumbral filaments ; this agrees with @xcite . only two g - band images , corresponding to beginning and decay times of the two pjs , are available . the middle g - band image is the same as in the first column and black arrows on it point to the locations in the penumbra where the example pjs are produced . although many of the pjs appear forming near heads of filaments , as pointed by a black arrow in the middle - panel g - band image in figure [ f3 ] , as an example , it can not be concluded with certainty that most pjs form at heads of filaments . the second jet pointed by the pale green arrow is among one of the larger penumbral jets , which appear to be rooted at the converging tails of a few filaments ( location pointed by a black arrow in the middle g - band image ) , discussed in detail in section [ trc2 ] . both chromospheric penumbral jets displayed in figure [ f3 ] show no signatures in the hi - c running difference images . the same is true for all other pjs seen in the movie ` movie1.mp4 ' . in the rd movie of fg images , we notice a couple of locations where somewhat larger jets repeatedly appear . one such location is pointed by the pale green arrow in figure [ f3 ] . but less than two minutes is too short a time to confirm this repetition . to explore in detail if jets are repeatedly produced at this location for longer time , and if there are other such locations of penumbral larger jets , and whether they have any tr or coronal signatures , we extended our investigation for about an hour of sot s h - line observations of the penumbra . we included sdo / aia data to look for tr and coronal signatures of larger pjs , as described next ( in section [ trc2 ] ) . in the one - hour h - line running difference ( rd ) movie , we noticed at least four locations where multiple jets are produced repeatedly . these jets appear brighter , larger and wider than most pjs . in what follows , we characterize these larger penumbral jets and investigate their signatures in the tr and corona . we also look for the formation mechanism of these larger pjs by studying magnetic field polarity using stokes - v images , which is described in section [ mag ] . first , the aia image of each wavelength ( 1600 , 304 , 171 , 193 , and 94 ) closest in time to each fg h - line image is selected . then we make both intensity and rd movies of each of these sets of images . in figure [ f4 ] , we display one example image of all aia channels used to investigate the signature of penumbral jets in coronal and tr emission . a movie ` movie2.mp4 ' containing six corresponding panels of rd images , in which the jets are better visible , is connected to this figure . the fg rd movie of one hour shows many normal pjs and at least four locations of larger pjs . these four locations of larger pjs are each indicated by an arrow and are numbered in the movie . one of these four locations ( numbered ` 2 ' ) is the same as that pointed by pale green arrow in figure [ f3 ] . the larger pjs have enhanced brightness of 3060% to that of the background penumbra . a clear signature of the example larger pj is visible in 1600 in figure [ f5 ] . the jet appears extending towards north , higher along the magnetic field , and only the front brightens in 193 , 171 and 304 channels . no intensity enhancement is noticeable in 94 . many similar larger jets are observed showing signatures in the four above mentioned wavelengths but no signatures of any of these larger jets were observed in 94 , thus indicating that there is no appreciable mk plasma produced in these penumbral jets , howsoever large they are , and that the brightenings in the other channels is from the emission from transition - region - temperature jet plasma ( see , e.g. , * ? ? ? also noticeable in the movie are some normal pjs , which are neither repetitive nor exceptionally large , with enhanced intensity of 2530% to background , displaying faint signatures in 1600 . however signature of these pjs in 304 , 171 , 193 and of course in 94 is hard to detect if at all . this indicates that some larger - end normal pjs do appear in tr emission , but not in coronal emission , in agreement with the absence of their detection in the hi - c 193 movie of 1.75 minutes described in section [ trc1 ] . we measured widths of many of the normal and larger penumbral jets . an example of width of each of them is shown in figure [ pjw ] . these examples are ones among the largest and widest in both the cases . larger pjs can be much wider than normal pjs , as about double in the presented example . remember the widest normal pjs seen by @xcite was less than 400 km . the larger pj shown here is about 500 km wide . when measured by using the method of fitting a gaussian and taking its full width at half maximum ( fwhm ) , as in figure [ pjw ] , the widths of our largest pj is found to be 600 km . ) , right panel is for a larger pj ( along red cut in figure [ f5 ] ) . black and red colored plots show the original intensity and the fitted gaussian function , respectively . ] the lifetime of larger pjs is however less than a minute , similar to that of normal pjs or a little longer . larger pjs are repeatedly produced at the same locations ( pointed by arrows in the movie ` movie2.mp4 ' ) in the sunspot penumbra . the length of the largest pj is 4200 km , however please note that the location of the sunspot ( close to disk center ) , and particularly of its penumbra ( disk - center side ) does not allow a measurement of the actual length owing to projection . nonetheless , the estimated lengths of our larger pjs fall in the range of that of the longest pjs ( up to 10000 km ) observed earlier by @xcite , indicating larger pjs are of the same category as normal pjs . a rough estimation of the speed of the larger penumbral jet at location ` 4 ' in ` movie2.mov ' between the time 19:40:25 and 19:41:03 gives a value of 250 km s@xmath1 . this speed is much faster than the acoustic speed of 10 km s@xmath1 in the chromosphere . this speed is also larger than that of normal pjs ( 100 - 150 km s@xmath1 ) by a factor of two . the alfvn speed in the penumbral chromosphere is of order 1000 km s@xmath1 . so the speed of the larger pjs is supersonic but sub - alfvnic . because the deduced observed projected speed is less than the true speed , the fastest larger pjs might be nearly alfvnic . the width of penumbral filaments is about 550 km @xcite . normal pjs appear rooted at an edge of a filament head , and are usually 150 - 300 km wide ; the lower limit of their width is not well known owing to the resolution limit of the sot . the size of larger pjs , particularly their widths , seems too large to be produced at the same locations as most pjs are . although it is difficult to detect opposite polarity signal in absence of sp scans , and advanced processing , e.g. , done by @xcite , we plausibly can detect the signals in stokes - v images from sot / fg , especially if larger pjs are produced at a different location having more reversed - polarity flux than the normal pjs . for this purpose , we analyzed corresponding stokes - v images of one hour , as presented in the next section . to investigate the magnetic origin of larger pjs , we looked at the stokes - v data of the same time span as that of the h - line analyzed in this work . the stokes - v images can be considered equivalent to line - of - sight ( los ) magnetograms , albeit with arbitrary units . the fov of stokes - v images is somewhat smaller than that of the h - line images in east - west extent ; it covers all but a western strip of the fov . the cadence of the stokes - v images is about 50 s , which is poorer than that of the images . in figure [ stv_ca ] , we display two images , one of stokes - v on the right and the other a h - line image of nearly the same time and fov , on the left . a movie ` movie3.mp4 ' of stokes - v images is linked to the figure [ stv_ca ] . because of smaller fov only three of the larger pj locations can be viewed in this movie . each of the three locations available in this fov is marked by an arrow . from figure [ stv_ca ] , and ` movie3.mp4 ' , it is clear that there is presence of opposite polarity magnetic field patches at the locations of these pjs in the stokes - v images of fg . the opposite polarity patches present in all the three locations of larger pjs s have different scales , forms and visibility . however , from the g - band images and from the stokes - v movie , each of them appear to be located at or around the tail of a penumbral filament or at a location where tails of several penumbral filaments converge . tails of filaments have opposite polarity field to that of spines @xcite . therefore the opposite polarity fields in spine and tails can reconnect and produce larger pjs , in a repetitive manner due to presence of the large amount of opposite polarity flux at tails . occasionally tails of penumbral filaments converge and form stronger patches of field of opposite polarity to that of spines @xcite , thus increasing the possibility of both the size and the frequency of larger pjs . . please note that the unit of the absolute flux is arbitrary . a vertical dashed line marks the start time of the example larger pj indicated by arrow in figure [ f5 ] . larger pjs are rooted at this location in the box in figure [ stv_ca ] . ] the locations of larger pjs on the photosphere may also contain emerging flux , cancellation of emerging flux or both . we show one example in figure [ stv_ca ] by outlining the area around a location of larger pjs ( numbered as ` 2 ' in the ` movie2.mp4 ' ) by a box . we selected the area within the box , and computed absolute stokes - v signal contained therein . this is equivalent to absolute flux , however , in an arbitrary unit . figure [ flux ] , shows the plot of absolute stokes - v signal . clearly , a trend of decreasing flux is seen for about first 40 minutes ( during which the location ` 2 ' produces larger jets ) that indicates a cancellation of stokes - v signal of opposite polarity field . from the movie ` movie3.mp4 ' , we notice that the existence of patches of positive polarity , which seem to be the locations of convergence of several filaments , emerge and cancel with the existing negative flux , as larger pjs get produced there . compared to the site ` 2 ' , similar or smaller patches of opposite polarity fields are observed in the other two locations of larger pjs . these magnetic field patches that have opposite polarity to that of the sunspot umbra appear to be tails of penumbral filaments surrounded by the opposite polarity field of spines or heads of neighboring filaments or both . the bipolar patches could instead be produced by sea - serpent field lines @xcite that move slowly outward . in either case the reconnection can take place between the field submerging into the photosphere ( tails of filaments ) and the existing opposite - polarity spine field and/or field inside heads of neighboring filaments . please remember that the opposite polarity field at the tails of filaments of inner penumbra may not be so clearly visible owing to the stokes - v signal being dominated by the surrounding spines there . these observations suggest that the larger pjs are produced by magnetic reconnection between impacted opposite polarity fields . the polarity arrangement might or might not be compatible with standard jet models , see , e.g. , @xcite . we have addressed two issues concerning penumbral jets in this paper . first , we have investigated whether pjs have any tr or coronal signatures . during the course of this part of the investigation , we identified locations in the penumbra , apparently near tails of filaments , at which repeatedly occurred pjs larger than the normal pjs . we investigated the magnetic setting of these larger pjs using stokes - v images and also studied their signatures in the tr and corona . second , based on the recent observational results on the internal structure of penumbral filaments @xcite , and the observations of larger pjs repeatedly appearing close to tails of penumbral filaments in the present paper , we propose a modified picture of the formation of pjs , detailed in section [ sec1 ] . taking benefit of high resolution ( @xmath12150 km ) of both the instruments the hinode ( sot / fg ) and hi - c , we compared the h - line and 193 images , searching for any signature of chromospheric pjs in the 193 coronal images . a comparison of running difference images of 1.75 minutes of both hinode ( sot / fg ) h - line and hi - c 193 reveals no significant signature of the jets in 193 images . however , 1.75 minute was too short period for a firm conclusion . an extension of our investigation by combining sdo / aia multiple channels , albeit with lower resolutions , reveals some repeating larger jets at some locations . in the g - band images and stokes - v movie of sot / fg , these locations appear to be tail of a penumbral filament or where tails of a few filaments converge . we also detected some normal ( non - tail , non - repetitive ) pjs clearly displaying signatures in 1600 , but not in 193 and other aia channels studied in this paper . this is in agreement with the recent observations of @xcite , who found tr signatures of pjs in , and k slit - jaw images of iris data . please note that the sensitivity and resolution of aia 1600 might have prevented the detection of response of pjs in the tr . or , the pjs selected by @xcite all being large enough , in the range of ones the largest normal pjs in our case showing the tr response in 1600 , can not be ruled out . thus , we conclude that although most of the normal pjs do not directly contribute to the tr and coronal heating , some of the largest normal pjs do display signatures in the tr . if smaller normal pjs do contribute to significant coronal heating in some form , it could be only by adding energy through increasing stress and braids in coronal loops , or by sending mhd waves up into the corona . we observed at least four locations in the fov of the sunspot penumbra we studied , where larger jets were produced repeatedly . these jets are brighter and larger in size than normal pjs . we find that although the lifetime and length of larger pjs are similar to or at the larger end of the distribution of normal pjs , they are wider , and have higher apparent speeds than the normal pjs . the width of the weakest jets hits the resolution limit of the telescope , sot , whereas the largest pjs are as wide as 600 km , measured by a gaussian fit as shown in figure [ pjw ] , for an example . the opposite polarity patches found at the base of the larger pjs , which apparently are the locations of tails of filaments , sometimes where multiple filaments tails converge @xcite , suggests a magnetic reconnection process to be responsible for their formation too . because at tails of filaments there exists larger patches of opposite polarity field than the other parts ( head and bulk ) , a larger scale magnetic reconnection of the field at tails with spine field is plausible . the flux cancellation as shown in figure [ flux ] for the example location of larger pjs is compatible with the reconnection being responsible for their production . the speeds of all jets , either normal pjs or larger / tail pjs , are faster than sonic but lower than the alfvn speed , consistent with these jets being magnetically driven via reconnection . thus , both the normal and larger / tail penumbral jets appear to be produced by magnetic reconnection process , in some ways similar to the ways other jets , flares and cmes are produced , see , e.g. , @xcite , and references therein . although the velocity of larger / tail pjs and their length are on the large side of that of pjs , we are unable to measure their true values due to projection foreshortening in our penumbra . a future investigation with such high resolution data sets , together with stokes - v images or magnetograms is required to better establish the characteristics of tail pjs . please note that the locations of larger pjs often also produce smaller ( normal ) size pjs . in other words , not all jets formed at the locations of larger pjs are large enough to appear in the hotter aia channels ( e.g. , in 193 ) . for example , the jet marked by green arrow in the h - line rd images of figure [ f3 ] is at a location of larger pjs ( numbered ` 2 ' in the ` movie2.mp4 ' ) but does not show any signature in 193 rd images of hi - c and/or aia . whereas , the jet displayed in figure [ f5 ] at the same location but at a different time clearly displays signatures in aia 1600 , 304 , 171 and 193 channels . the estimated thermal energy of chromospheric normal pjs is of the order of 10@xmath11 erg , and most pjs hardly show any tr or coronal signatures . whereas , some penumbral jets with larger sizes , that are produced at tails of penumbral filaments , have width of 2 - 3 times more than that of most pjs , and thus have energy larger by a factor of four to nine , or more , show signatures in the aia 1600 , 304 , 171 , and 193 channels . because aia 304 , 1600 , 171 are formed in the chromosphere , lower corona or tr , and 193 passes some tr emission , as discussed in section [ data ] , we can not rule out that the tail pjs contribute some weak coronal heating . but what we can definitely say from our current observations is that tail pjs display significant signatures in tr emission . however , because of the sparsity of tail pjs , a significant contribution of tail pjs to coronal heating above sunspot penumbrae is very unlikely . high resolution data from future hi - c flights , solar - c and dkist solar observations should be able to address these questions more closely . [ f1 ] a sunspot penumbra contains penumbral filaments and spines @xcite . the recently explored internal structure of sunspot penumbral filaments by @xcite suggests the following modification of the formation mechanism of pjs proposed by @xcite . the penumbral filaments are horizontally elongated magnetized convective cells in which lateral downflows are present . these downflowing lanes also contain opposite polarity field , which is most clearly present near heads of penumbral filaments but also continues to show at sides along the full length of filaments getting strongest at their tails @xcite . although exact location of pjs has not yet been established @xcite , pjs possibly occur all along penumbral filaments . on each side of a penumbral filament there is spine field that presses against the filament @xcite . the spine field could easily reconnect with the opposite polarity field in the sides of filaments , and produce pjs that travel along the spine field . please note that , with distance from the umbra , the density of spines and field strength in them decrease and their field inclination with respect to vertical increases @xcite . thus , the rate of production of pjs and their strength are expected to decrease in the outer penumbra . in the current work , we discovered several locations in penumbra where repeated larger pjs are formed . these locations appear to be the tails of penumbral filaments . we propose that the reconnection between the spine field and tail field can generate these larger pjs . they are large and repetitive because the opposite polarity flux is larger in strength and area there as compared to that on the sides of penumbral filaments . in figure [ f1 ] , we draw a cartoon diagram of the magnetic configuration resulting in the formation of most / normal pjs and tail / larger pjs via magnetic reconnection discussed above . as displayed in the cartoon , the direction of the pjs is still the same as proposed by @xcite , see also @xcite , along the spine field , the more vertical field lines , which changes inclination with radius getting more inclined outward . this picture fits well with the most recent observations of fine - scale structures of sunspot penumbrae . the earlier picture of @xcite of formation of pjs by reconnection between the two magnetic flux tubes inclined at an acute angle to each other does not fit the recent observations of the convective nature of penumbral filaments . the magnetic field along a filament s axis in the head of the filament and that in the surrounding spines have nearly similar inclination , so reconnection between the spines and opposite polarity field at the edges of penumbral filaments is more likely than it taking place between the field of spines and the same polarity field inside the heads and along the bulk of filaments . in the present picture , at the sides of the penumbral filaments , where the angle between the spine field and the filament field is greatest , most pjs form . depending on the size of tails of filaments , or a convergence of tails of more than one filament , magnetic reconnection can repetitively produce larger / tail pjs , in the way shown in the cartoon . the presence of patches of opposite polarity field in the penumbra at the locations of tail pjs is also compatible with the sea - serpent field ( e.g. , * ? ? ? although the reconnection of the magnetic field in the tails of penumbral filaments with the opposite polarity spine field seems to be the most plausible mechanism of formation of larger pjs , the reconnection of tail fields with the heads of neighboring filaments generating larger pjs can not be ruled out from our observations . neither can the emergence of bipolar field in the penumbra leading to such larger pjs via magnetic reconnection with the existing penumbral field be ruled out . it is also unknown whether most pjs form at sites of flux cancellation or emergence ; both of these processes are appropriate for producing them . our picture needs to be verified by higher resolution observations of future observatories , e.g. , solar - c and dkist . our proposal is supported by the reconnection produced by @xcite in mhd modeling of pjs , see also modeling by @xcite . in their model penumbral filaments are assumed to be one twisted flux tube . we believe that this model corresponds to either side of a penumbral filament ( figure [ f1 ] ) . the reconnection between opposite polarity field at one side of twisted magnetic flux tube can produce jets in a similar way as our proposal for either side of real penumbral filaments . it is important to mention here that only one third of filaments selected by @xcite displayed opposite polarity field in their sides near heads ; the number slighly reduces in the middle ( to before the tail ) of filaments . the fact that this structure is close to the resolution limit of the telescope , and the opposite polarity fields can only be detected by advanced processing of the sp data , e.g. , by spatially coupled inversion code @xcite , or by deconvolution methods @xcite , or can only be seen in the specially processed ground based observations @xcite , the possibility of each of the filaments having opposite polarity field not only near their heads but all along the filament can not be ruled out . in that case , it is possible that the filaments produce pjs all along their sides by the same process as described above , but in some cases we can not detect them for them being very narrow and small , below the limit of resolution of the present telescopes . the normal pjs show at most weak signatures in the transition region ( in aia 1600 images ) and none in the corona : neither in any of the aia coronal channels nor in the high - resolution hi - c images of 193 . weak signatures of largest normal pjs in 1600 suggest that a few largest normal pjs directly power some localized transient tr heating . a few locations of larger penumbral jets are found in the penumbra , which are locations of mixed polarity magnetic flux near tails of filaments in the penumbra . larger / tail pjs are brighter ( up to 60% more than that of the background penumbra ) , wider ( up to 600 km ) and faster ( can be more than 250 km s@xmath1 ) than most / normal pjs , but have lifetimes and apparent lengths similar to or at larger ends of that of most pjs . they flash repeatedly at the same location and clearly display signatures in the transition region ( in aia 1600 , 304 , 171 and 193 channels ) . however , no pure coronal signature in aia 94 is detected . none of the penumbral jets seem to be heated enough to display a signature in the hot corona , however , all of the larger ones do appear in transition region emission . these results should be verified with higher resolution data of future hi - c flights and solar - c mission . in aggregate , most pjs and tail pjs apparently do not directly produce appreciable coronal heating ( normal pjs because of lack of their coronal signatures , larger pjs because of their sparsity and/or their lack of pure coronal signatures ) , but conceivably contribute significantly to coronal heating via braiding of the coronal field rooted in and around them , or by production of alfvn waves . we propose a modified picture of the formation mechanism of pjs considering the recent advancements on the observed magnetic structure of sunspot penumbral filaments . it is more likely that pjs are formed by magnetic reconnection between spines and opposite polarity field at the edges of penumbral filaments along each side of penumbral filaments , as found partly in the mhd simulations of @xcite , as compared to the earlier proposed reconnection between two inclined components of the same polarity field . the larger / tail pjs appear to form as a result of magnetic reconnection between the opposite polarity field of spines and of filament tails . we are grateful to the referee for constructive comments , which resulted in major modification and improvement of the paper . hinode is a japanese mission developed and launched by isas / jaxa , collaborating with naoj as a domestic partner , nasa and stfc ( uk ) as international partners . scientific operation of the hinode mission is conducted by the hinode science team organized at isas / jaxa . this team mainly consists of scientists from institutes in the partner countries . support for the post - launch operation is provided by jaxa and naoj ( japan ) , stfc ( u.k . ) , nasa , esa , and nsc ( norway ) . the aia and hmi data are courtesy of nasa / sdo and the aia and hmi science teams . msfc / nasa led the hi - c mission and partners include the sao in cambridge , mass . ; lmsal in palo alto , calif . ; the uclan in lancashire , england ; and the lpiras in moscow . skt is supported by appointment to the nasa postdoctoral program at the nasa / msfc , administered by orau through a contract with nasa . for this work sea was supported by the national science foundation under grant no . arw and rlm are supported by funding from the lws trt program of the heliophysics division of nasa s smd .

penumbral microjets ( pjs ) are transient narrow bright features in the chromosphere of sunspot penumbrae , first characterized by @xcite using the h - line filter on _ hinode _ s solar optical telescope ( sot ) . it was proposed that the pjs form as a result of reconnection between two magnetic components of penumbra ( spines and interspines ) , and that they could contribute to the transition region ( tr ) and coronal heating above sunspot penumbrae . we propose a modified picture of formation of pjs based on recent results on internal structure of sunspot penumbral filaments . using data of a sunspot from _ hinode_/sot , high resolution coronal imager , and different passbands of the atmospheric imaging assembly ( aia ) onboard the _ solar dynamics observatory _ , we examine whether pjs have signatures in the tr and corona . we find hardly any discernible signature of normal pjs in any aia passbands , except a few of them showing up in the 1600 images . however , we discovered exceptionally stronger jets with similar lifetimes but bigger sizes ( up to 600 km wide ) occurring repeatedly in a few locations in the penumbra , where evidence of patches of opposite polarity fields at the tails of some penumbral filaments is seen in stokes - v images . these large tail pjs do display signatures in the tr . whether they have any coronal - temperature plasma is ambiguous . we infer that none of the pjs , including the large tail pjs , directly heat the corona in ars significantly , but any penumbral jet might drive some coronal heating indirectly via generation of alfvn waves and/or braiding of the coronal field .

the nucleon spin structure function @xmath4 , extracted from the deep inelastic lepton - nucleon scattering with polarization @xcite , if interpreted in the infinite momentum frame , defines a fraction of the nucleon spin carried by quark spins . accordingly , only about 30% of the nucleon spin is carried by quark spins , which is referred to usually as `` spin crisis '' . consequently , a lot of experimental and theoretical efforts for the last 20 years were devoted to search for the rest of the nucleon spin , without obvious success , however . similar results are obtained on the lattice both for nucleons and lowest mesons ( for a review see ref . @xcite ) . in @xcite we have suggested a way to define in a gauge invariant manner and measure the spin content of mesons in the rest frame at different resolution scales , including the infrared ones . the method is based on the unitary transformation from the complete and orthogonal chiral quark - antiquark basis to the complete and orthogonal @xmath1 basis in the rest frame @xcite . using a set of interpolators that transform according to different chiral representations @xcite , one can measure on the lattice a ratio of couplings of different interpolators to a given state in the rest frame . this ratio determines a mixture of different chiral representations in the meson wave functions . then , given this ratio and using a unitary transformation from the chiral basis to the angular momentum basis , one obtains the angular momentum content of a meson in the rest frame . the result was that the angular momentum content of the @xmath2-meson in the rest frame is given approximately by the @xmath3 wave , without obvious trace of the spin crisis . then a natural question is which of the definitions does reflect the spin content of a hadron ? below we overview both principal aspects of the method as well as the numerical results using as an example the @xmath2-meson . the @xmath5 @xmath0 states with unbroken chiral symmetry transform according to two chiral representations , @xmath6 and @xmath7 @xcite . these two representations form a complete and orthogonal basis . the state that transforms as @xmath8 can be created from the vacuum by the vector current , @xmath9 and the state that belongs to the @xmath7 representation can be created by the pseudotensor operator , @xmath10 in the continuum the physical @xmath2-meson with broken chiral symmetry can be created from the vacuum by both operators and the corresponding amplitudes are given as @xmath11 where @xmath12 is the vector meson state with the mass @xmath13 , momentum @xmath14 and polarization @xmath15 . the vector current is conserved , consequently the vector coupling constant @xmath16 is scale - independent . the pseudotensor `` current '' is not conserved and is subject to a nonzero anomalous dimension . consequently the pseudotensor coupling @xmath17 manifestly depends on the resolution scale @xmath18 . the very fact that the @xmath2-meson can be created by the operators that transform according two different chiral representations tells that chiral symmetry is broken and the meson wave function is a superposition of both chiral representations . in the rest frame the ratio @xmath19 determines the ratio of the two allowed chiral representations in the @xmath2-meson wave function at a given resolution scale @xmath18 . such ratio can be measured on the lattice . given a set of operators @xmath20 above we use the variational method @xcite and calculate a cross - correlation matrix at zero spatial momentum ( i.e. , in the rest frame ) , @xmath21 with the coefficients giving the overlap of the operators with the physical state , @xmath22 with a set of operators spanning a complete and orthogonal basis with respect to some symmetry group , these overlaps ( coupling constants ) give the complete information about symmetry breaking . the interpolating composite operators @xmath20 are not normalized on the lattice and consequently the absolute values of the coupling constants @xmath23 can not be obtained . however , a ratio of the couplings is a well defined quantity and can be computed as @xcite @xmath24 here @xmath25 is the cross - correlation matrix from ( [ corr_inf ] ) , a sum is implied for the index @xmath26 on the right - hand side and @xmath27 are the eigenvectors obtained from the generalized eigenvalue problem , @xmath28 with @xmath29 being some normalization point in euclidean time . the ratio ( [ ratio_op_comp ] ) coincides with the ratio of matrix elements ( [ rhov / rhot ] ) with @xmath30 . the chiral basis in the quark - antiquark system is a complete one and can be connected to the complete angular momentum basis in the rest frame via the unitary transformation @xcite @xmath31 with @xmath32 consequently , if we know the mixture of the two allowed chiral representations in a physical state , then applying the unitary transformation ( [ unitary_1 ] ) we are also able to obtain the angular momentum content of this state in the rest frame . the mixture of the chiral representations in the @xmath2-meson wave function at the resolution scale @xmath18 is given by ( [ ratio_op_comp ] ) and ( [ rhov / rhot ] ) and can be measured in lattice simulations . by this we define what is meant under the angular momentum content of the @xmath2-meson in the rest frame and , consequently , we can answer the question whether or not the spin of the @xmath2-meson is carried by spins of its valence quarks in the rest frame . the ratio @xmath33 as well as a partial wave content of a hadron are not the renormalization group invariant quantities . hence they manifestly depend on a resolution scale at which we probe the hadron . if we probe the hadron structure on the lattice with the local interpolators , then we study the hadron decomposition at the scale fixed by the lattice spacing @xmath34 . for a reasonably small @xmath34 this scale is close to the ultraviolet scale . however , we are interested in the hadron content at the infrared scales , where mass is generated . for this purpose we can not use a large @xmath34 , because matching with the continuum qcd will be lost . given a fixed , reasonably small lattice spacing @xmath34 a small resolution scale @xmath35 can be achieved by the gauge - invariant smearing of the point - like interpolators . we smear every quark field in spatial directions with the gaussian profile over the size @xmath36 in physical units such that @xmath37 , see fig . 1 . then even in the continuum limit @xmath38 we probe the hadron content at the resolution scale fixed by @xmath36 . such a definition of the resolution is similar to the experimental one , where an external probe is sensitive only to quark fields ( it is blind to gluonic fields ) at a resolution that is determined by the momentum transfer in spatial directions . definition , title="fig : " ] definition , title="fig : " ] we use three different smearing radii r for the quark fields both in the source and sink . the narrow " smearing width ( index @xmath39 ) varies between 0.33 and 0.36 fm , depending on the set of configurations . the wide " smearing radius ( index @xmath40 ) lies between 0.66 and 0.69 fm and the ultrawide " one is 0.81 - 0.85 fm ( index @xmath41 ) . hence we can study the hadron structure at resolutions 0.33 fm - 0.85 fm . consequently we have the following set of the interpolating operators @xmath42 where @xmath43 is one of the spatial dirac matrices and @xmath44 is the @xmath45-matrix in ( euclidean ) time direction . .[tab : sim ] specification of the data used here ; for the gauge coupling only the leading value @xmath46 is given , @xmath47 denotes the bare mass parameter of the ci action . further details on the action , the simulation and the determination of the lattice spacing and the @xmath48- and @xmath2-masses are found in @xcite . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] the masses of both @xmath2 and @xmath49 states extracted from different @xmath50 and @xmath51 correlation matrices . the crosses indicate the mass values from experiments.,height=302 ] in our study we use chirally improved fermions @xcite and the lscher - weisz gauge action @xcite . the lattice size is @xmath52 . we use dynamical gauge configurations with two mass - degenerate light quarks . with the lattice spacing @xmath53 fm the spatial volume of the lattice is @xmath54 @xmath55 . in our present study we limit ourselves to the @xmath2 and @xmath49 states . for details on the simulation we refer the reader to the table [ tab : sim ] and to @xcite . for the sets a and c we construct @xmath50 correlation matrices ( i.e. , with both vector and pseudotensor interpolators using narrow and wide smearing radii ) , while for the sets b1 and b2 we study the @xmath51 correlation matrix ( with narrow , wide and ultrawide smearings for both vector and pseudotensor operators ) as well as different @xmath50 sub - matrices . in fig . [ fig : masses ] we show masses of both the ground state @xmath2-meson and its first excitation @xmath49 extracted from different correlation matrices . the ratio @xmath56 of the two allowed chiral representations in the @xmath2- and @xmath49-mesons versus the resolution @xmath36 is shown on fig . [ fig : all ] . the results obtained from different @xmath50 and @xmath51 correlation matrices are consistent with each other . for the ground state @xmath2 we observe strong chiral symmetry breaking practically at all resolutions ; only in the deep ultraviolet @xmath57 , where chiral symmetry is unbroken , the tensor `` current '' decouples from the @xmath2-meson . for the @xmath49 the tensor operator decouples much faster towards the ultraviolet than for the @xmath2-meson . this shows that the wave functions of these two states are significantly different and the chiral symmetry breaking is different in both states . at the scale of @xmath58 fm of the hadron size , i.e. , at which the hadron mass is generated , chiral symmetry breaking in the @xmath49 state is less essential than in the ground state . given this ratio and the unitary transformation from the chiral basis to the angular momentum basis ( [ unitary_1 ] ) one can obtain the partial wave content of both states . for the @xmath2 meson it is approximately @xmath59 . hence the ground state in the infrared is practically a pure @xmath3 state with a tiny admixture of the @xmath60 wave . consequently in the rest frame the spin of the @xmath2-meson is almost completely carried by spins of its valence quarks . no trace of the `` spin crisis '' is observed . in this our definition we do not separate contributions of quarks and gluons into hadron spin . in the confining regime , where the gauge invariance and the quark - gluon interaction are of crucial importance , it is not possible to separate both contributions into total spin in a sensible gauge - invariant manner ; after all the quark - gluon interaction is a consequence of the gauge invariance . the spin of the @xmath2-meson in the rest frame is carried by spins of its valence quarks dressed by gluons . the gluonic field is important for the angular momentum generation , because it is this field that provides chiral symmetry breaking and that is responsible for most of the hadron mass . in the excited @xmath2-meson there is a significant contribution of the @xmath60 wave . in the latter case the angular momentum content is between the following two lower and upper bound values . for the lower bound it is @xmath61 and for the upper bound it is @xmath62 . this once again demonstrates that the first excitation of the @xmath2-meson can not be considered a pure radial excitation of the ground state @xmath2 . obviously , both radial and orbital degrees of freedom are excited which reflects yet unknown dynamics of confinement and chiral symmetry breaking . a ratio of the vector to the pseudotensor couplings versus a resolution scale @xmath36 , as extracted from all @xmath50 and @xmath51 correlation matrices . broken lines are drawn only to guide the eye.,height=302 ] we gratefully acknowledge support of the grants p21970-n16 and dk w1203-n08 of the austrian science fund fwf and the dfg project sfb / tr-55 . the calculations have been performed on the sgi altix 4700 of the leibniz - rechenzentrum munich and on the local clusters at the university of graz . 99 -2pt j. ashman _ et al . _ [ european muon collaboration ] , phys . b * 206 * ( 1988 ) 364 ; b. adeva _ et al . _ [ spin muon collaboration ] , phys . d * 58 * ( 1998 ) 112002 . e. s. ageev _ et al . _ [ compass collaboration ] , phys . b * 612 * ( 2005 ) 154 ; v. y. alexakhin _ et al . _ [ compass collaboration ] , phys . b * 647 * ( 2007 ) 8 ; a. airapetian _ et al . _ [ hermes collaboration ] , phys . rev . d * 75 * ( 2007 ) 012007 . hgler , phys . * 490 * ( 2010 ) 49 . l. y. glozman , c. b. lang and m. limmer , phys . * 103 * ( 2009 ) 121601 ; few body syst . * 47 * ( 2010 ) 91 . l. y. glozman , c. b. lang and m. limmer , phys . d * 82 * ( 2010 ) 097501 . l. y. glozman , c. b. lang , m. limmer , phys . lett . * b705 * , 129 - 133 ( 2011 ) . [ arxiv:1106.1010 [ hep - ph ] ] . l. y. glozman and a. v. nefediev , phys . rev . d * 76 * ( 2007 ) 096004 ; phys . d * 80 * ( 2009 ) 057901 . t. d. cohen and x. d. ji , phys . d * 55 * ( 1997 ) 6870 . l. y. glozman , phys . b * 587 * , 69 ( 2004 ) ; phys . rept . * 444 * , 1 ( 2007 ) c. michael , nucl . b * 259 * ( 1985 ) 58 ; m. lscher and u. wolff , nucl . b * 339 * , 222 ( 1990 ) . c. gattringer , phys . d * 63 * ( 2001 ) 114501 ; c. gattringer , i. hip , and c. b. lang , nucl . b * 597 * ( 2001 ) 451 ( 2001 ) . m. lscher and p. weisz , commun . phys . * 97 * 59 ( 1985 ) ; [ erratum - ibid . * 98 * ( 1985 ) 433 ] . c. gattringer , c. hagen , c. b. lang , m. limmer , d. mohler and a. schfer , phys . d * 79 * ( 2009 ) 054501 . g. p. engel , c. b. lang , m. limmer , d. mohler and a. schfer , phys . d * 82 * ( 2010 ) 034505 .

from the parton distributions in the infinite momentum frame one finds that only about 30% of the nucleon spin is carried by spins of the valence quarks , which gave rise to the term `` spin crisis '' . similar results hold for the lowest mesons , as it follows from the lattice simulations . we define the spin content of a meson in the rest frame and use a complete and orthogonal @xmath0 chiral basis and a unitary transformation from the chiral basis to the @xmath1 basis . then , given a mixture of different allowed chiral representations in the meson wave function at a given resolution scale , one can obtain its spin content at this scale . to obtain the mixture of the chiral representations in the meson we measure in dynamical lattice simulations a ratio of couplings of interpolarors with different chiral structure . for the @xmath2 meson we obtain practically the @xmath3 state with no trace of the spin crisis . then a natural question arises : which definition does reflect the spin content of a hadron ?

the neutron skin thickness of nuclei is defined as the difference between the root mean square radius of neutrons and protons , @xmath8 the extraction of neutron radii and neutron skins from the experiment is in general dependent on the shape of the neutron distribution used in the analysis @xcite . the data typically do not indicate unambiguosly if the difference between the peripheral neutron and proton densities is caused by an extended bulk radius of the neutron density , by a modification of the width of the surface , or by some combination of both effects . in particular , the neutron skin @xmath0 of @xmath1pb is nowadays attracting significant interest in both experiment and theory since it has a close relation with the density dependence of the nuclear symmetry energy and with the equation of state of neutron - rich matter @xcite , which have a large impact in diverse problems of nuclear physics and astrophysics where neutron - rich matter is involved @xcite . the analysis of bulk and surface contributions to the neutron skin thickness of a nucleus requires proper definitions of these quantities based on nuclear density distributions @xcite . using the standard definitions of the equivalent sharp radius @xmath9 and surface width parameter @xmath10 of a nucleon density profile , we have shown in refs . @xcite that one can write @xmath11 , where @xmath12 and @xmath13 we recall that @xmath14 stands for the radius of a uniform sharp distribution whose density equals the bulk value of the actual density and that has the same number of particles . therefore , one has a natural splitting of @xmath0 in terms of a bulk part ( [ rb ] ) independent of surface properties , and a part ( [ rs ] ) of surface origin @xcite . nonrelativistic ( nrmf ) and relativistic ( rmf ) mean field models often differ in their predictions for properties of asymmetric nuclear matter . a common example is the value predicted for the density slope @xmath15 of the nuclear symmetry energy at saturation density , which may show large discrepancies in the mf interactions . the @xmath15 parameter is defined as @xmath16 where @xmath17 is the symmetry energy and @xmath18 the nuclear saturation density . linear correlation of @xmath0 of @xmath1pb with the density slope of the nuclear symmetry energy @xmath15 . the dependence on @xmath15 of the bulk and surface contributions defined in eqs . ( [ rb ] ) and ( [ rs ] ) is also displayed . ] in figure [ fig2 ] we show the linear correlation of @xmath0 of @xmath1pb with the parameter @xmath15 and demonstrate that it mainly arises from the bulk part of @xmath0 within a large and representative set of mean field models of very different nature . relativistic ( rmf ) models : g2 , nlc , nl - sh , tm1 , nl - ra1 , nl3 , nl3 * , nl - z , nl1 , dd - me2 and fsugold ; and nonrelativistic ( nrmf ) models : hfb-8 , msk7 , d1s , sgii , d1n , sk - t6 , hfb-17 , sly4 , skm * , sksm * , skmp , ska , sk - rs and sk - t4 . the original references to the different interactions can be found in @xcite for the skyrme models , @xcite for the gogny models , and @xcite and ref . [ 19 ] in @xcite for the rmf models . from this figure , one sees that whereas the bulk contribution to the neutron skin thickness of @xmath2pb changes largely among the different mean - field interactions , the surface contribution remains confined to a narrow band of values . the shape of the neutron density is more uncertain than the proton density in @xmath1pb , and even if the neutron rms radius is determined ( _ e.g. _ in prex @xcite ) , it can correspond to different shapes of the neutron density . as discussed in @xcite , from the study of the two - parameter fermi functions fitted to the self - consistent densities of the mf models , we find that for @xmath2pb the central radii @xmath19 vary within the windows @xmath20 fm in nrmf and @xmath21 fm in rmf , @xmath22 fm in nrmf and @xmath23 fm in rmf , and that the diffuseness parameters @xmath24 vary within the windows @xmath25 fm in nrmf and @xmath26 fm in rmf and @xmath27 fm . from these results , we can conclude that most of the mf models predict a neutron distribution for @xmath1pb between the halo - type limit ( where @xmath28 ) and the skin - type limit ( where @xmath29 ) . the halo - type is supported by models with a very soft symmetry energy . models with a stiffer symmetry energy ( larger @xmath15 values ) have @xmath30 more and more different from zero . however , a pure skin - type is not found in any mean - field model as @xmath31 is always non - vanishing . parity violating asymmetry ( dwba ) in @xmath1pb for 1 gev electrons at @xmath32 scattering angle . mf results and those obtained with the parametrized formula given in the text . ] parity - violating electron scattering ( pves ) probes the neutron density in a nucleus since the @xmath33 boson couples mainly to neutrons @xcite . the prex experiment at jlab @xcite aims to determine the neutron rms radius of @xmath1pb by pves . we have investigated the dependence of pves on the shape of the neutron and proton densities of @xmath1pb within the mf models . to this end , we compute the parity - violating asymmetry @xmath34 at the prex kinematics @xcite . we obtain the differentical cross sections @xmath35 for right and left - handed electrons by performing the exact phase shift analysis ( dbwa ) of the dirac equation @xcite for incident electrons moving in the potentials @xmath36 @xcite . see ref . @xcite for the details of the calculations . we display in fig . [ fig4 ] the results for @xmath37 as a function of @xmath30 and @xmath31 of the two parameter fermi distributions fitted to the @xmath1pb mf densities . we have found that the results can be reasonably parametrized by the formula @xmath38 with @xmath39 , @xmath40 @xmath41 and @xmath42 @xmath41 , which is depicted by the crosses in the figure . recently , in @xcite we have analyzed systematically the correlations of @xmath37 with the neutron skin thickness of @xmath1pb and with the slope of the symmetry energy in the nuclear mf models . we have found that the known linear correlation of @xmath0 of @xmath1pb with the density derivative of the nuclear symmetry energy arises from the bulk part of @xmath0 . this implies that an experimental determination of @xmath43 of @xmath1pb could be as useful for constraining @xmath15 as a determination of @xmath44 . mf models can accomodate the halo - type distribution in @xmath1pb if the symmetry energy is very soft but do not support a purely skin - type distribution . we find a simple parametrization of @xmath37 in terms of @xmath30 and @xmath31 that would provide a new correlation between the central radius and the surface diffuseness of the distribution of neutrons in @xmath1pb assuming the proton density known from experiment . work partially supported by grants csd2007 - 00042 , fis2008 - 01661 and 2009sgr-1289 ( spain ) and n202 231137 ( poland ) .

we study whether the neutron skin thickness @xmath0 of @xmath1pb originates from the bulk or from the surface of the neutron and proton density distributions in mean field models . we find that the size of the bulk contribution to @xmath0 of @xmath1pb strongly depends on the slope of the nuclear symmetry energy , while the surface contribution does not . we note that most mean field models predict a neutron density for @xmath1pb between the halo and skin type limits . we investigate the dependence of parity - violating electron scattering at the kinematics of the prex experiment on the shape of the nucleon densities predicted by the mean field models for @xmath1pb . we find an approximate formula for the parity - violating asymmetry in terms of the central radius and the surface diffuseness of the nucleon densities of @xmath1pb in these models . * study of the neutron skin thickness of @xmath2pb in mean field models * x. roca - maza@xmath3 , m. centelles@xmath4 , x. vias@xmath4 and m. warda@xmath5 * @xmath4departament destructura i constituents de la matria and institut de cincies del cosmos , facultat de fsica , universitat de barcelona , diagonal _ 647 _ , _ 08028 _ barcelona , spain . * @xmath6infn , sezione di milano , via celoria 16 , i-_20133 _ milano , italy . * @xmath7katedra fizyki teoretycznej , uniwersytet marii curie skodowskiej , ul . radziszewskiego 10 , 20 - 031 lublin , poland .

hybrid perovskite photovoltaic ( hppv ) technology @xcite is the most recent rising star in the emerging solar - cell community due to its record increase in power - conversion efficiency ( pce ) during the last four years @xcite . the current state - of - the - art hppv architecture was proposed in 2012 achieving @xmath4 pce @xcite . now the pce of hppv cells has already reached @xmath5 , overtaking the best - performing inorganic - based single - junction thin - film cells such as @xmath6 and copper - indium - gallium - selenide ( cigs ) cells @xcite . the most common photoactive material in hppv cells is methylammonium ( @xmath7 ) lead triiodide ( @xmath0 , shortened as @xmath8 hereafter ) . this hybrid perovskite material exhibits several advantageous features for photovoltaic applications , such as a band gap close to the optimal value for single - junction solar - cell absorbers @xcite , excellent absorption strength in the visible part of the solar spectrum @xcite , and high mobilities for both electron and hole transport @xcite . it can be synthesized in solution at low temperature from common starting materials that have limited harm to the environment . therefore hppv cells are considered as promising candidates that can offer clean , affordable and sustainable energy . apart from the pce improvement , recent experimental and theoretical studies in hppv technology have focused on the origin of the high mobility and low recombination rate @xcite , the observed current - voltage hysteresis @xcite , and the stability of hybrid perovskites materials @xcite . to resolve open questions in hybrid perovskites it is imperative to develop a comprehensive understanding of their atomic structure , which is both fundamental and challenging due to the structure s complexity . taking the prototype hybrid perovskite @xmath8 as an example , the central cation @xmath9 is not spherical ( as , e.g. , @xmath10 in the conventional perovskite @xmath11 ) but exhibits polarity and an orientational preference in the lattice . at low temperatures , @xmath8 assumes a minimal - energy structure with regularly - aligned @xmath9 cations and thereby a regularly - deformed inorganic @xmath12 matrix , resulting in an orthorhombic phase . conversely , at room temperature or above , the @xmath9 cations are randomly oriented due to thermal fluctuations , forming ( dynamically-)disordered structures @xcite . experimental @xcite and theoretical @xcite studies suggest that the activation energy for the reorientation of the @xmath13@xmath14 bond is of the order of @xmath15 ( corresponding to a characteristic thermal energy of @xmath16 ) . the @xmath9 ions therefore do not rotate freely at @xmath17 , but are bound to the inorganic framework by hydrogen bonds @xcite . they form localized patterns @xcite on a small length scale , which makes @xmath8 appear effectively cubic . modeling the orientational disorder of @xmath9 ions in @xmath8 is a challenging task . quantum mechanical first principles techniques are required to correctly describe the hydrogen bonding of @xmath9 ions to the inorganic cage and the corresponding distortions of the cage . however , even density - functional theory ( dft ) , which in local or semi - local approximations is currently the most computationally efficient first principles technique , can not scale up to the required length scales or the large number of candidate structures . the simple primitive - cell model is not representative of @xmath8 s atomic structure , as it effectively describes a system of infinitely many aligned polar @xmath9 ions . the dipole introduced by each @xmath9 unit can be canceled by compensating alignments of @xmath9 ions in an appropriately chosen supercell model . for such compensated models , the atomic and electronic structure , especially in the low - temperature orthorhombic and tetragonal phases , can then be calculated at the dft or beyond level by means of small supercell models such as @xmath18{2}\!\times\!\sqrt[]{2}\!\times\!2 $ ] and @xmath2 @xcite . @xmath2 supercell models have also been adopted to study the distribution of @xmath9 orientations at a finite temperature using _ ab initio _ molecular dynamics ( md ) @xcite . however , to really model disorder we would need to know the structure of @xmath8 on a length scale of a few to a few tens of single ( primitive ) cells . @xmath2 supercell models do not suffice for this purpose because of the periodic boundary conditions , while dft calculations for larger supercell models become computationally very demanding . to our knowledge , the application of dft to large mapbi@xmath19 supercells has mostly been limited to defect modeling @xcite . disordered structures , such as the behavior of domains has so far only been investigated with classical monte carlo simulations @xcite . in our previous work @xcite , we have comprehensively analyzed the atomic structure of hybrid perovskites using the primitive - cell model . we found several stable locations of @xmath9 in the lattice . moreover , our analysis revealed that the stability of hybrid perovskites is closely related to the deformation of the inorganic cage , which acts synergetically with the organic ions analogous to a chicken - and - egg paradox . in this work , we performed dft calculations for a number of different @xmath8 supercell models and focus on the pairs of neighboring @xmath9 ions . we devised a pair - mode description that reduces each @xmath9 ion to a dipole @xcite with discrete orientations that were adopted from our previous primitive - cell results @xcite . we then defined the relative geometry of two nearest individual dipoles as a `` pair mode '' . with the pair - mode concept , we were then able to relate the dependence of certain @xmath8 properties ( e.g. , total energy , band structure ) to the distribution of @xmath9-orientations and to the dipoles alignment . this was done by studying a series of @xmath2 supercell models . we further investigated @xmath3 supercell models focusing on the distribution of pair modes . this distribution tells us for a given dipole which dipole - orientations are preferred in its surrounding , thus providing knowledge of the local structure beyond a single @xmath8 unit cell . the atomic structure of the pair modes , as the physical origin of this pair - mode distribution , will be discussed in detail in a forthcoming paper @xcite . the remainder of this paper is organized as follows . in sec . [ comp ] , we briefly describe the model systems and the computational details of our dft calculations . section [ results ] outlines the concept of pair modes , and uses this concept to discuss the results of the supercell models . finally , sec . [ concl ] concludes with a summary . the supercell models considered in this paper were constructed based on single ( primitive ) cells . in each single cell , the @xmath9 is located close to the centre of the cell , @xmath20 at the corners , and @xmath21 at the edge - centers . we considered a series of @xmath2 and @xmath3 supercell models for different purposes . in @xmath2 supercell models the total dipole moment can be easily canceled with regular alignments of @xmath9 ions . although a @xmath2 supercell model is larger than a @xmath18{2}\!\times\!\sqrt[]{2}\!\times\!1 $ ] model , it has the advantage that it does not introduce artificial differences between the three lattice directions _ a priori_. compared with the @xmath2 counterparts , the larger @xmath3 supercell models contain more @xmath9 cations , thus providing appropriate model systems to mimic the disordered structures by introducing randomly oriented @xmath9 ions . the considered supercell models were initialized with different @xmath9 alignments . in our previous study @xcite we demonstrated that the `` pbe+vdw '' exchange - correlation functional produces the lattice constants of hybrid perovskite systems in good agreement with experiment , and can properly describe the interaction between the organic cations and the inorganic framework . thus we adopted this functional for all dft calculations in this work . specifically , the perdew - burke - ernzerhof ( pbe ) generalized gradient approximation @xcite was used as the exchange - correlation functional , and the long - range van der waals ( vdw ) interactions were described by employing the tkatchenko - scheffler method based on the hirshfeld partitioning of the electron density @xcite . in addition , scalar relativistic effects were included via the zero - order regular approximation ( zora ) @xcite . although spin - orbit coupling @xcite and exact exchange or many - body corrections @xcite have significant impact on the band structure of @xmath8 , we did not include them in the band - structure calculations ( for the @xmath2 supercell models ) , since they are computationally very demanding . for the relative differences of band gaps between different supercell models pbe+vdw+zora is sufficient . all calculations were carried out using the all - electron numeric - atom - centered orbital code fhi - aims @xcite . for the @xmath2 supercell models , we used a @xmath22-centered @xmath23 @xmath24-point mesh and tier 2 basis sets for both structure relaxation and band - structure calculations . we performed direct lattice - vector optimization with the analytical stress tensor implemented in fhi - aims @xcite . for the larger @xmath3 supercells , we have reduced the computational expense by employing a @xmath2 @xmath24-point mesh and tier 1 basis sets . the geometries were optimized for a fixed size of the cubic unit cell , for which the lattice parameter @xmath25 was adopted based on the experimental value @xmath26 of the primitive cell @xcite . the results of all relevant calculations of this work are available from the novel materials discovery ( nomad ) repository @xcite . the essential difference between conventional perovskites such as @xmath27 and hybrid perovskites is the monovalent central cation . metal cations such as @xmath10 are spherically symmetric , while the polar organic @xmath1 cation has a permanent dipole moment pointing from the @xmath13-end ( the methyl group ) to the @xmath14-end ( the ammonium cation group ) . thus , in a primitive - cell model all @xmath9 ions will be aligned parallel . this would result in a large dipole moment in the bulk material , which is not observed experimentally . supercell models allow us to cancel the total dipole moment in the supercell . in this paper , we calculated the total dipole moment within a supercell by the vector sum of individual @xmath9 dipole moments ( or its average per @xmath8 unit ) and represent it in terms of @xmath28 , the permanent dipole moment of an isolate @xmath9 in vacuum . our pbe+vdw / tier 2 result of @xmath28 is @xmath29 , very close to the b3lyp/6 - 31 g * result of @xmath30 @xcite . our previous dft calculations @xcite revealed two stable structures of the cubic primitive - cell model , as shown in fig . [ dipoles](a ) . specifically , in the left structure of fig . [ dipoles](a ) , the @xmath13@xmath14 bond is oriented along the diagonal ( @xmath31 $ ] or equivalent ) direction of the single unit cell , while in the right structure it is oriented along the face - to - face ( @xmath32 $ ] or equivalent ) direction with a small deviation . the face - to - face @xmath9 structure is @xmath33 more stable than the diagonal structure . we attribute this stability to the considerably larger deformation of the inorganic - framework in the face - to - face structure . as a side note , the internal atomic geometry of @xmath9 in @xmath8 is nearly independent of its location in the unit cell . [ cols="^,^,^ " , ] we start from the three most populated modes . table [ f2f ] indicates that the theoretical and actual probabilities are very close for both vertical modes , no . 20 and 22 . they can be understood as `` intrinsically '' dominant modes . a noticeable drop ( @xmath34 , namely @xmath35 ) can be observed in the dft results for the out - of - plane mode ( no . 25 ) . within the least populated linear modes , the comparison between the probabilities in non - interacting systems and actual probabilities are different . the @xmath9-pair interaction results in a @xmath36 ( @xmath37 ) total drop for both the head - to - head mode ( no . 19 ) and the tail - to - tail mode ( no . 21 ) , while an increase of @xmath38 ( @xmath39 ) is observed for the extending mode ( no . 18 ) . finally , the populations of the parallel mode ( no . 23 ) and the antiparallel mode ( no . 24 ) in optimized structures are @xmath40 ( @xmath41 ) lower and @xmath42 ( @xmath43 ) higher than the theoretical values , respectively . this comparison indicates that , in @xmath8 , the interaction between the neighboring @xmath9 ions significantly favors mode no . 18 , which creates a linear alignment of dipoles that will be advantageous for charge transport . conversely , the population - increase in mode no . 24 and decrease in mode no . 23 will oppose linear domain formation . overall this will result in the following domain structure : a linear domain is adjacent to another linear domain containing electric dipoles oriented along the opposite direction . these findings qualitatively agree with the previous monte carlo simulations at @xmath44 that was based on _ ab initio _ md and employed a classical hamiltonian to describe the dynamic motion of dipoles @xcite . on average we expect a domain size ( length ) smaller than that was reported in ref . @xcite . the origin is the prevalent population of the vertical modes no . 20 and 22 , by which a domain is terminated . this discrepancy arises because the md - based monte carlo simulations in ref . @xcite ( i ) were performed for a two - dimensional model and thereby neglect the interaction with the third dimension completely , and ( ii ) only included the direct ( electrostatic ) interaction between dipoles , while our dft calculations also include the indirect interaction via the deformation of the inorganic cage . the pair - mode distribution shown in fig . [ pmall ] enables us to construct larger @xmath8 models that are out of reach of dft . as an example , [ 20x20x20 ] shows the @xmath9-alignment in a layer of the thus - constructed @xmath45 supercell . this @xmath45 model was generated iteratively , applying the pair - mode distribution shown in fig . [ pmall ] . details of the construction and an analysis of the results will be presented in a future publication . since these large - scale models follow the @xmath9-distribution of stable orientations and include @xmath9 nearest - neighbor interactions , they provide good models to study realistic @xmath8 structures under realistic conditions . the improvement of fig . [ pmall ] and the multiscale modeling of these large @xmath8 models are the subjects of future work . we have studied a series of @xmath8 supercell models using dft . to establish a multiscale model , we derived and analyzed the concept of `` pair modes '' , i.e. , the interaction of nearest @xmath9-ion pairs . we first investigated several small @xmath2 supercell models , in which we can cancel out the overall dipole moment by hand using suitable dipole orientations . our dft results indicate that differences in pair modes have a significant effect on the atomic and electronic structure of these models . this finding motivated our exploration into larger @xmath3 @xmath8 supercell models , which we used to simulate disordered @xmath8 structures by randomly initializing the @xmath9 orientations . structural optimization using dft and our pair mode analysis reveal that the final locations of @xmath9 ions is not fully random , but follows certain preferred orientations that depend on the surrounding . we use this dipole distribution to build up a multiscale model to generate the local structure in bulk @xmath8 samples on large length scales , e.g. , of a few tens of single cells . we thank h. levard for fruitful discussions . the generous allocation of computing resources by the csc - it center for science ( via the project no . ay6311 ) and the aalto science - it project are gratefully acknowledged . an award of computer time was provided by the innovative and novel computational impact on theory and experiment ( incite ) program . this research used resources of the argonne leadership computing facility , which is a doe office of science user facility supported under contract de - ac02 - 06ch11357 . this work was supported by the academy of finland through its centres of excellence programme ( 2012 - 2014 and 2015 - 2017 ) under project numbers 251748 and 284621 , as well as its key project funding scheme under project number 305632 . 53 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , * * , ( ) . , , , * * , ( ) . , , , , , , , , , , , * * , ( ) . , , , , , * * , ( ) . , , , , , , * * , ( ) . , , , * * , ( ) . , , , , , , , , , * * , ( ) . , , , , , , , , , * * , ( ) . , , , , , , , , * * , ( ) . , , , , , , , , , , , * * , ( ) . , * * , ( ) . , , , , , , , , , , * * , ( ) . , , , , , , , * * , ( ) . , , , , * * , ( ) . , , , , , * * , ( ) . , , , , , , * * , ( ) . , , , * * , ( ) . , * * , ( ) . , , , , , * * , ( ) . , , , * * , ( ) . , , , * * , ( ) . , , , , , * * , ( ) . , , , , * * , ( ) . , . , * * , ( ) . , * * , ( ) . , , , , , , * * , ( ) . , , , * * , ( ) . , , , , , , , , , , , * * , ( ) . , , , * * , ( ) . , , , , * * , ( ) . , , , , , * * , ( ) . , , , , , , , , , , * * , ( ) . , , , , , , * * , ( ) . , , , * * , ( ) . , , , , * * , ( ) . , , , . , , , * * , ( ) . , * * , ( ) . , , , * * , ( ) . , , , , , * * , ( ) . , , , , * * , ( ) . , , , , * * , ( ) . , , , , , * * , ( ) . , , , , , , , , * * , ( ) . , , , * * , ( ) . , , , , , , , , * * , ( ) . , , , , * * , ( ) . , , , , , , , * * , ( ) , issn . , , , , , * * , ( ) . . , , , , , , , , * * , ( ) . 1.5em supplemental material for * pair modes of organic cations in hybrid perovskites : insight from first - principles calculations of supercell models * jingrui li@xmath46 , jari jrvi@xmath47 and patrick rinke@xmath48 @xmath48__centre of excellence in computational nanoscience ( comp ) and department of applied physics , + aalto university , p.o.box 11100 , fi-00076 aalto , finland _ _ @xmath49__department of physics , university of helsinki , p.o.box 64 , fi-00014 university of helsinki , finland _ _ ' '' '' + @xmath50 : [email protected] in our previous work we found two stable orientations for @xmath9 cations in the cubic primitive - cell model . in the `` diagonal '' structure [ the left structure of fig . 1(a ) in the manuscript ] , the @xmath13@xmath14 bond is oriented along the diagonal direction of the single unit cell . we thus denote the direction of the @xmath9-dipole @xmath31 $ ] , @xmath51 $ ] , and so forth . in the `` face - to - face '' structure [ the right structure of fig . 1(b ) in the manuscript ] , the @xmath13@xmath14 bond is oriented along the face - to - face direction of the single unit cell with a small deviation . this deviation is system- and model - dependent . for example , it is @xmath52 in the primitive - cell model of @xmath8 , @xmath53 on average in structure ii , and @xmath54 on average in structure iii . according to the primitive - cell results , the @xmath13@xmath14 bond along the @xmath32 $ ] direction is located in the symmetry planes @xmath55 $ ] or @xmath56 $ ] . the resulting @xmath13@xmath14 bond directions including the deviation are then @xmath57 $ ] , @xmath58 $ ] , @xmath59 $ ] or @xmath60 $ ] . the six face - to - face @xmath9 dipoles in fig . 1(c ) in the manuscript can be subdivided into twenty - four dipole directions as follows ( thick solid and dashed arrows point out of and into the plane of the paper , respectively ) : ccc without & & + deviation & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .3,.5 ) ( .7,.5 ) ; & ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; & ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .31,.56 ) ( .69,.44 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .31,.47 ) rectangle ( .69,.53 ) ; ( .31,.5 ) ( .69,.5 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .31,.5 ) ( .69,.5 ) ; ( .31,.5 ) ( .69,.5 ) ; + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .7,.5 ) ( .3,.5 ) ; & ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; & ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .69,.44 ) ( .31,.56 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .69,.56 ) ( .31,.44 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .69,.47 ) rectangle ( .31,.53 ) ; ( .69,.5 ) ( .31,.5 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .69,.5 ) ( .31,.5 ) ; ( .69,.5 ) ( .31,.5 ) ; + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.3 ) ( .5,.7 ) ; & ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; & ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .56,.31 ) ( .44,.69 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .47,.31 ) rectangle ( .53,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.31 ) ( .5,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,0 ) ; ( .5,.7 ) ( .5,.3 ) ; & ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; & ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .56,.69 ) ( .44,.31 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,0 ) ; ( .44,.69 ) ( .56,.31 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .53,.69 ) rectangle(.47,.31 ) ; ( .5,.69 ) ( .5,.31 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.69 ) ( .5,.31 ) ; ( .5,.69 ) ( .5,.31 ) ; + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .5,.5 ) circle ( .05 ) ; & ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; & ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .44,.5 ) ( .56,.5 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .56,.5 ) ( .44,.5 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .5,.44 ) ( .5,.56 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .5,.56 ) ( .5,.44 ) ; + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .5,.5)+(45:.15 ) + ( 225:.15 ) ; ( .5,.5)+(135:.15 ) + ( -45:.15 ) ; & ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; & ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .44,.5 ) ( .56,.5 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .56,.5 ) ( .44,.5 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .5,.44 ) ( .5,.56 ) ; ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 1.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 0,1 ) ( 1,1 ) ( 1,0 ) ; ( .5,.5 ) circle ( .15 ) ; ( .5,.56 ) ( .5,.44 ) ; + together with the eight diagonal directions , there are 32 possible directions for an @xmath9 dipole . this results in 86 symmetry inequivalent modes as follows : * pair modes no . 16 : + ccl + at ( 0,0 ) [ left , color = white ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1cm , color = white ] @xmath62 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) [ color = white ] @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.41,.37 ) ( 1.63,.59 ) ( 1.59,.63 ) ( 1.37,.41 ) cycle ; ( 1.39,.39 ) ( 1.61,.61 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 1@xmath65 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.39,.39 ) ( 1.61,.61 ) ; ( 1.39,.39 ) ( 1.61,.61 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 2@xmath65 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.39,.61 ) ( 1.61,.39 ) ; ( 1.39,.61 ) ( 1.61,.39 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 3@xmath65 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.59,.37 ) ( 1.37,.59 ) ( 1.41,.63 ) ( 1.63,.41 ) cycle ; ( 1.61,.39 ) ( 1.39,.61 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 4@xmath65 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.61,.39 ) ( 1.39,.61 ) ; ( 1.61,.39 ) ( 1.39,.61 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 5@xmath65 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.61,.61 ) ( 1.39,.39 ) ; ( 1.61,.61 ) ( 1.39,.39 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 6@xmath65 ; * pair mode no . 7 can be splits into 2 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.3,.5 ) ( 1.7,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 7@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.31,.44 ) ( 1.69,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 7@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.31,.56 ) ( 1.69,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 7@xmath67 ; * pair mode no . 8 splits into 2 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.7,.5 ) ( 1.3,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 8@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.69,.44 ) ( 1.31,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 8@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.69,.56 ) ( 1.31,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 8@xmath67 ; * pair mode no . 9 splits into 4 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.5,.3 ) ( 1.5,.7 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 9@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 9@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 9@xmath67 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 9@xmath68 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 9@xmath69 ; * pair mode no . 10 splits into 4 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.5,.7 ) ( 1.5,.3 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 10@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.44,.69 ) ( 1.56,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 10@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.56,.69 ) ( 1.44,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 10@xmath67 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.47,.69 ) rectangle ( 1.53,.31 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 10@xmath68 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .41,.37 ) ( .63,.59 ) ( .59,.63 ) ( .37,.41 ) cycle ; ( .39,.39 ) ( .61,.61 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 10@xmath69 ; * pair modes no . 1113 : + ccl + at ( 0,0 ) [ left , color = white ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1cm , color = white ] @xmath62 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) [ color = white ] @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.41,.37 ) ( 1.63,.59 ) ( 1.59,.63 ) ( 1.37,.41 ) cycle ; ( 1.39,.39 ) ( 1.61,.61 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 11@xmath65 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.39,.39 ) ( 1.61,.61 ) ; ( 1.39,.39 ) ( 1.61,.61 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 12@xmath65 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.39,.61 ) ( 1.61,.39 ) ; ( 1.39,.61 ) ( 1.61,.39 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 13@xmath65 ; * pair mode no . 14 splits into 2 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.3,.5 ) ( 1.7,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 14@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.31,.44 ) ( 1.69,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 14@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.31,.56 ) ( 1.69,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 14@xmath67 ; * pair mode no . 15 splits into 2 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.7,.5 ) ( 1.3,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 15@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.69,.44 ) ( 1.31,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 15@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.69,.56 ) ( 1.31,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 15@xmath67 ; * pair mode no . 16 splits into 4 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.5,.3 ) ( 1.5,.7 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 16@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 16@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 16@xmath67 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 16@xmath68 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 16@xmath69 ; * pair mode no . 17 splits into 4 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.5,.7 ) ( 1.5,.3 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 17@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.44,.69 ) ( 1.56,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 17@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.56,.69 ) ( 1.44,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 17@xmath67 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.47,.69 ) rectangle ( 1.53,.31 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 17@xmath68 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .59,.37 ) ( .37,.59 ) ( .41,.63 ) ( .63,.41 ) cycle ; ( .61,.39 ) ( .39,.61 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 17@xmath69 ; * pair mode no . 18 splits into 3 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .3,.5 ) ( .7,.5 ) ; ( 1.3,.5 ) ( 1.7,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 18@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.31,.44 ) ( 1.69,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 18@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.31,.56 ) ( 1.69,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 18@xmath67 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.31,.47 ) rectangle ( 1.69,.53 ) ; ( 1.31,.5 ) ( 1.69,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 18@xmath68 ; * pair mode no . 19 splits into 3 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .3,.5 ) ( .7,.5 ) ; ( 1.7,.5 ) ( 1.3,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 19@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.69,.44 ) ( 1.31,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 19@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.69,.56 ) ( 1.31,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 19@xmath67 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.69,.47 ) rectangle ( 1.31,.53 ) ; ( 1.69,.5 ) ( 1.31,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 19@xmath68 ; * pair mode no . 20 splits into 10 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .3,.5 ) ( .7,.5 ) ; ( 1.5,.3 ) ( 1.5,.7 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 20@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath70 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath71 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.44 ) ( .69,.56 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath72 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .31,.56 ) ( .69,.44 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath73 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.56 ) ( .69,.44 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath74 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.56 ) ( .69,.44 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath75 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .31,.47 ) rectangle ( .69,.53 ) ; ( .31,.5 ) ( .69,.5 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath76 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.47 ) rectangle ( .69,.53 ) ; ( .31,.5 ) ( .69,.5 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath77 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.47 ) rectangle ( .69,.53 ) ; ( .31,.5 ) ( .69,.5 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath78 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .31,.47 ) rectangle ( .69,.53 ) ; ( .31,.5 ) ( .69,.5 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 20@xmath79 ; * pair mode no . 21 splits into 3 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .7,.5 ) ( .3,.5 ) ; ( 1.3,.5 ) ( 1.7,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 21@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .69,.44 ) ( .31,.56 ) ; ( 1.31,.44 ) ( 1.69,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 21@xmath66 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.44 ) ( .31,.56 ) ; ( 1.31,.56 ) ( 1.69,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 21@xmath67 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.44 ) ( .31,.56 ) ; ( 1.31,.47 ) rectangle ( 1.69,.53 ) ; ( 1.31,.5 ) ( 1.69,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 21@xmath68 ; * pair mode no . 22 splits into 10 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .7,.5 ) ( .3,.5 ) ; ( 1.5,.3 ) ( 1.5,.7 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 22@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .69,.44 ) ( .31,.56 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath70 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.44 ) ( .31,.56 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath71 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.44 ) ( .31,.56 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath72 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .69,.56 ) ( .31,.44 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath73 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.56 ) ( .31,.44 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath74 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.56 ) ( .31,.44 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath75 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .69,.47 ) rectangle ( .31,.53 ) ; ( .69,.5 ) ( .31,.5 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath76 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.47 ) rectangle ( .31,.53 ) ; ( .69,.5 ) ( .31,.5 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath77 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.47 ) rectangle ( .31,.53 ) ; ( .69,.5 ) ( .31,.5 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath78 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .69,.47 ) rectangle ( .31,.53 ) ; ( .69,.5 ) ( .31,.5 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 22@xmath79 ; * pair mode no . 23 splits into 7 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .5,.3 ) ( .5,.7 ) ; ( 1.5,.3 ) ( 1.5,.7 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 23@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 23@xmath70 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.56,.31 ) ( 1.44,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 23@xmath71 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 23@xmath72 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .56,.31 ) ( .44,.69 ) ; ( 1.44,.31 ) ( 1.56,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 23@xmath73 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .56,.31 ) ( .44,.69 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 23@xmath74 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .47,.31 ) rectangle ( .53,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; ( 1.47,.31 ) rectangle ( 1.53,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 23@xmath76 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .47,.31 ) rectangle ( .53,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; ( 1.5,.31 ) ( 1.5,.69 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 23@xmath77 ; * pair mode no . 24 splits into 7 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .5,.3 ) ( .5,.7 ) ; ( 1.5,.7 ) ( 1.5,.3 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 24@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.44,.69 ) ( 1.56,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 24@xmath70 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.56,.69 ) ( 1.44,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 24@xmath71 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.47,.69 ) rectangle ( 1.53,.31 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 24@xmath72 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .56,.31 ) ( .44,.69 ) ; ( 1.44,.69 ) ( 1.56,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 24@xmath73 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .56,.31 ) ( .44,.69 ) ; ( 1.47,.69 ) rectangle ( 1.53,.31 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 24@xmath74 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .47,.31 ) rectangle ( .53,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; ( 1.47,.69 ) rectangle ( 1.53,.31 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 24@xmath76 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .47,.31 ) rectangle ( .53,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; ( 1.5,.69 ) ( 1.5,.31 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 24@xmath77 ; * pair mode no . 25 splits into 10 variations : + ccl + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .5,.3 ) ( .5,.7 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.5,.5 ) circle ( .05 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] @xmath65no . 25@xmath65 ; + & + ( 0,-0.1 ) ( 0,1.2 ) ; at ( 0,0.5 ) @xmath64 ; + & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.44,.5 ) ( 1.56,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath70 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.56,.5 ) ( 1.44,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath71 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.5,.44 ) ( 1.5,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath72 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .44,.31 ) ( .56,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.5,.56 ) ( 1.5,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath80 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .56,.31 ) ( .44,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.44,.5 ) ( 1.56,.5 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath73 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .56,.31 ) ( .44,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.5,.44 ) ( 1.5,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath74 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .56,.31 ) ( .44,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.5,.56 ) ( 1.5,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath75 ; + + & & + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,1 ) ( 2,0 ) ; ( .47,.31 ) rectangle ( .53,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.5,.44 ) ( 1.5,.56 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath76 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .47,.31 ) rectangle ( .53,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.5,.56 ) ( 1.5,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath77 ; + + ( 0,0 ) circle ( .1 ) ; ( 0,0 ) circle ( .03 ) ; at ( 0,0 ) [ left ] @xmath61 ; ( .1,0 ) ( 2.2,0 ) node [ right , xshift=-.1 cm ] @xmath62 ; ( 0,.1 ) ( 0,1.2 ) node [ above , yshift=-.1 cm ] @xmath63 ; ( 1,0 ) ( 1,1 ) ; ( 0,1 ) ( 2,0 ) ; ( .5,.31 ) ( .5,.69 ) ; ( .5,.31 ) ( .5,.69 ) ; ( 1.5,.5 ) circle ( .15 ) ; ( 1.5,.56 ) ( 1.5,.44 ) ; at ( 1,1.5 ) [ above , yshift=-.1 cm ] no . 25@xmath81 ; \(a ) structure i + \(b ) structure ii + \(c ) structure iii + \(d ) structure iv + * atomic structure : cf . fig . 4 in the manuscript . * lattice parameters : @xmath82 . * band structure ( band gap @xmath83 ) :

we have studied the properties of the prototype hybrid organic - inorganic perovskite @xmath0 using relativistic density functional theory ( dft ) . for our analysis we introduce the concept of @xmath1 `` pair modes '' , that is , characteristic relative orientations of two neighboring @xmath1 cations . in our previous work [ phys . rev . b * 94 * , 045201 ( 2016 ) ] we identified two preferential orientations that a single @xmath1 cation adopts in a unit cell . the total number of relevant pairs can be reduced from the resulting 196 combinations to only 25 by applying symmetry operations . dft results of several @xmath2 supercell models reveal the dependence of the total energy , band gap and band structure on the distribution of @xmath1 cations and the pair modes . we have then analyzed the pair - mode distribution of a series of @xmath3 supercell models with disordered @xmath1 cations . our results show that diagonally - oriented @xmath1 cations are rare in optimized @xmath0 supercell structures . the prevailing pair modes are advantageous for the formation of linear domains . our pair - mode concept provides combined atomistic - statistical insight to understand the disordered structures in bulk hybrid perovskite materials .