Recursive construction for a class of radial functions. II. Superspace

Recursive construction for a class of radial functions. II. SuperspaceThomas Guhr and Heiner Kohler Citation: Journal of Mathematical Physics 43, 2741 (2002); doi: 10.1063/1.1463218 View online: http://dx.doi.org/10.1063/1.1463218 View Table of Contents: http://scitation.aip.org/content/aip/journal/jmp/43/5?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Integration of Grassmann variables over invariant functions on flat superspaces J. Math. Phys. 50, 013528 (2009); 10.1063/1.3049630 Evaluation of Fluctuation Coefficients for Three Consecutive Term Recursive Basis Functions AIP Conf. Proc. 1048, 231 (2008); 10.1063/1.2990899 Spherical functions on homogeneous superspaces J. Math. Phys. 46, 043513 (2005); 10.1063/1.1868859 Associated Bessel functions and the discrete approximation of the free-particle time evolution operator incylindrical coordinates J. Math. Phys. 45, 1988 (2004); 10.1063/1.1695601 Recursive construction for a class of radial functions. I. Ordinary space J. Math. Phys. 43, 2707 (2002); 10.1063/1.1463709

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

http://scitation.aip.org/content/aip/journal/jmp?ver=pdfcov

http://oasc12039.247realmedia.com/RealMedia/ads/click_lx.ads/www.aip.org/pt/adcenter/pdfcover_test/L-37/1218966096/x01/AIP-PT/CiSE_JMPArticleDL_120314/Awareness_LibraryF.jpg/47344656396c504a5a37344142416b75?x

http://scitation.aip.org/search?value1=Thomas+Guhr&option1=author

http://scitation.aip.org/search?value1=Heiner+Kohler&option1=author

http://scitation.aip.org/content/aip/journal/jmp?ver=pdfcov

http://dx.doi.org/10.1063/1.1463218

http://scitation.aip.org/content/aip/journal/jmp/43/5?ver=pdfcov

http://scitation.aip.org/content/aip?ver=pdfcov

http://scitation.aip.org/content/aip/journal/jmp/50/1/10.1063/1.3049630?ver=pdfcov

http://scitation.aip.org/content/aip/proceeding/aipcp/10.1063/1.2990899?ver=pdfcov





Recursive construction for a class of radial functions.II. Superspace

Thomas Guhra)

Matematisk Fysik, LTH, Lunds Universitet, Box 118, 22100 Lund, Swedenand Max Planck Institut fu¨r Kernphysik, Postfach 103980, 69029 Heidelberg, Germany

Heiner Kohlerb)

Departamento de Fı´sica, Universidad Auto´noma de Madrid, Madrid, Spainand Max Planck Institut fu¨r Kernphysik, Postfach 103980, 69029 Heidelberg, Germany

~Received 19 October 2001; accepted for publication 1 February 2002!

We extend the recursion formula for matrix Bessel functions, which we obtainedpreviously, to superspace. It is sufficient to do this for the unitary orthosymplecticsupergroup. By direct computations, we show that fairly explicit results can beobtained, at least up to dimension 838 for the supermatrices. Since we introducea new technique, we discuss various of its aspects in some detail. ©2002 Ameri-can Institute of Physics.@DOI: 10.1063/1.1463218#

I. INTRODUCTION

In a previous work, we studied properties of matrix Bessel functions in ordinary space.1 Here,we generalize these investigations to superspace. For the introductory remarks and the mathemati-cal and physical background relevant for the ordinary space, and also relevant as the basis for thepresent study, we refer the reader to Ref. 1.

In mathematics, supersymmetry was pioneered by Berezin2 and, in particular group theoreticalaspects, by Kac.3,4 The theory of nonlinears models in spaces of supermatrix fields was devel-oped in physics of disordered systems by Efetov.5,6 Verbaarschot, Weidenmu¨ller, and Zirnbauer7,8

used his approach to study models in random matrix theory. In Ref. 9, the first supersymmetricgeneralization of the Itzykson–Zuber integral10 was given. In Ref. 11, Gelfand–Tzetlincoordinates12 were constructed for the unitary supergroup. Extending Shatashvili’s13 method, thesupersymmetric Itzykson–Zuber integral was also rederived in Ref. 11 in its most general form.Using the techniques of Ref. 9, such a calculation was also performed in Ref. 14.

From a mathematical viewpoint, Efetov’s work5 is the basis for a harmonic analysis in certainsupersymmetric coset spaces, the Efetov spaces, which are relevant for the nonlinears models. Inthe full superspaces, a technique involving convolution integrals and ingredients of the corre-sponding harmonic analysis was introduced in Ref. 15. In the Efetov spaces, the theory of har-monic analysis, in both its mathematical and physical aspects, was developed by Zirnbauer16 andwas applied to disordered systems in Refs. 17 and 18. In the present contribution, we do not focuson the Efetov spaces, rather we address the full supergroup spaces. The supersymmetric Itzykson–Zuber integral9 and its application in Ref. 15 is the simplest example of a supermatrix Besselfunction appearing in this kind of harmonic analysis.

The matrix Bessel functions in superspace find direct application in random matrix theory. Forgeneral reviews, see Refs. 19–21. In Ref. 22 it was shown that they are the kernels for thesupersymmetric analog of Dyson’s Brownian motion.

The paper is organized as follows: In Sec. II, we introduce the supermatrix Bessel functionsand collect basic definitions and notations. In Sec. III, we extend the recursion formula of Ref. 1to superspaces. Since it is one of our goals to demonstrate that explicit results for supermatrix

a!Electronic mail: [email protected]!Electronic mail: [email protected]

JOURNAL OF MATHEMATICAL PHYSICS VOLUME 43, NUMBER 5 MAY 2002

27410022-2488/2002/43(5)/2741/29/$19.00 © 2002 American Institute of Physics


129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

Bessel functions can indeed be obtained, we present, in some detail, such calculations for certainsupermatrix Bessel functions in Secs. IV and V, respectively. The asymptotics and the normaliza-tion are discussed in Sec. VI. We briefly comment on applications in Sec. VII and we summarizeand conclude in Sec. VIII. Various calculations are shifted to the appendices.

II. SUPERMATRIX BESSEL FUNCTIONS

Similar to ordinary spaces,1 the superunitary case, i.e., integration over the supergroupU(k1 /k2), is the simplest one. Since this was already discussed in detail in Refs. 9, 11, and 22, werefrain from reconsidering it here. Thus, it turns out that we may restrict ourselves to the super-matrix Bessel function of the unitary orthosymplectic group UOSp(k1/2k2). As discussed byKac,3,4 the supergroups U(k1 /k2) and UOSp(k1/2k2) exhaust almost all classical compact super-groups, apart from some exotic exceptions which are of little relevance for applications. Hence,the integral we have to deal with is given by

Fk12k2~s,r !5E

uPUOSp(k1/2k2)exp~ i trgu21sur!dm~u!, ~2.1!

where dm(u) is the invariant measure. The arguments of the function~2.1! are the diagonalmatricess5diag(Acs1 ,A2cs2! and r 5diag(Acr1 ,A2cr2!. Here, we use Wegner’s notation23

and introduce the labelc561 to distinguish the two possible forms. We will return to this issue.The matricess1 , s2 and r 1 , r 2 are given by

s15diag~s11,s21,...,sk11!, s25diag~s1212 ,...,sk2212!,

~2.2!r 15diag~r 11,r 21,...,r k11!, r 25diag~r 1212 ,...,r k2212!.

There is a twofold degeneracy ins2 andr 2 , because the matrixu21su or, equivalently,uru21 hasto be areal Hermitiansupermatrix23 of the form

s5FAcs (R) s (A)†

s (A) A2cs (HSd)G , c561. ~2.3!

The matricess (R) and s (HSd) have ordinary commuting entries, i.e., bosons, they are real sym-metric and Hermitian self-dual, respectively. The matrixs (A) has anticommuting or Grassmannentries, i.e., fermions, and is of the form

s (A)5@s1(A) , ...,sk1

(A)#, s i(A)5F s1i

(A)

s1i(A)*]

sk2i(A)

sk2i(A)*

G . ~2.4!

We can now appreciate the meaning of the parameterc which enters the definition~2.3! of the realHermitian matrices. Forc51, it yields the real symmetric and forc521 the Hermitian self-dualmatrix as boson–boson block, and vice versa for the fermion–fermion block. In the framework ofrandom matrix theory, we find the supermatrices corresponding to the Gaussian orthogonal en-semble~GOE! for c511 and those for the Gaussian symplectic ensemble~GSE! for c521.

The infinitesimal volume element is given by

d@s#5)i 51

k1

)j 51

k2

ds i j(A)* ds i j

(A))i , j

ds i j(R))

i 51

k1

ds i i(R))

i , jd@s i j

(HSd)#)i 51

k2

ds i i(HSd) , ~2.5!

2742 J. Math. Phys., Vol. 43, No. 5, May 2002 T. Guhr and H. Kohler


129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

where d@s i j(HSd)# is the product of the differentials of all independent elements of the quaternion

s i j(HSd) .

The supermatrix Bessel functions~2.1! are eigenfunctions of a wave equation in the curvedspace of the eigenvaluess or r . As in the ordinary case, a supermatrix gradient]/]s is introducedand the Laplace operator is defined by

D5trgS ]

]s D 2

. ~2.6!

The plane waves exp(i trgsr) are the eigenfunctions, i.e., we have

D exp~ i trgsr!52trgr2 exp~ i trgsr!. ~2.7!

Here, boths andr are real Hermitian. As in ordinary space, the supermatrix Bessel functions areobtained by averaging over the angular coordinates, i.e., over the diagonalizing group. The La-placian commutes with the average and we arrive at the differential equation

DsFk12k2~s,r !52trg r 2Fk12k2

~s,r !, ~2.8!

where the radial part of of the Laplacian~2.6! reads

Ds51

Bk1k2

(c) ~s!S (

p51

k1 ]

]sp1

Bk1k2

(c) ~s!]

]sp1

11

2(p51

k2 ]

]sp2

Bk1k2

(c) ~s!]

]sp2D . ~2.9!

The Jacobian or Berezinian is given by22

Bk1k2

(1) ~s!5uDk1

~s1!uDk2

4 ~ is2!

)p51k1 )q51

k2 ~sp12 isq2!2 , Bk1k2

(21)~s!5uDk1

~ is1!uDk2

4 ~s2!

)p51k1 )q51

k2 ~ isp12sq2!2 . ~2.10!

One easily convinces oneself thatDs depends onc only through a factorAc. Thus, without lossof generality, we setc51 and omit the indexc.

At this point, an important comment is in order. The normalization in ordinary space accord-ing to Eq.~3.17! in Ref. 1,FN

(b)(x,0)51 andFN(b)(0,k)51, does not carry over to the supersym-

metric case. This is due to the fact that the volume of some supergroups is zero2 resulting in thevanishing ofFk12k2

(0,s) for certain values ofk1 andk2 . This collides with the normalization ofthe plane waves~2.7! to unity at the origin. The reason for this contradiction is a well-knownphenomenon in superanalysis. In going from Cartesian to angle eigenvalue coordinates, one has toadd additional terms to the measure to preserve the symmetries of the original integral. These arecalled Efetov–Wegner–Parisi–Sourlas terms in the physics literature. A full-fledged mathematicaltheory of these boundary terms was given by Rothstein.24

To solve this normalization problem, we use the following strategy. First, we evaluate thesupermatrix Bessel functions without taking care of the normalization. We just multiply the inte-grals with a normalization constantGk12k2

. Having done the integrals, we determine the normal-ization by comparing the asymptotics of the supermatrix Bessel function for large arguments withthe Gaussian integral.

III. SUPERSYMMETRIC RECURSION FORMULA

We extend the recursion formula in ordinary space1 to superspace. After stating the result inSec. III A, we present the derivation and the calculation of the invariant measure in Secs. III B andIII C, respectively.

2743J. Math. Phys., Vol. 43, No. 5, May 2002 Recursive construction. II


129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

A. Statement of the result

Let Fk12k2(s,r ) be defined through the group integral in Eq.~2.1!. It has two diagonal

matrices defined as in Eq.~2.2! as arguments. It can be calculated iteratively by therecursionformula

Fk12k2~s,r !5Gk12k2

E dm~s8,s! exp~ i ~ trgs2trgs8!r 11!F (k121)2k2~s8, r !, ~3.1!

where F (k121)2k2(s8, r ) is the group integral~2.1! over UOSp((k121)/2k2)) and Gk12k2

is anormalization constant, see Secs. II and VI. As in the ordinary case,1 the coordinatess8 areradialGelfand–Tzetlin coordinates. Again, they are different from theangular Gelfand–Tzetlin coordi-nates, which will be discussed elsewhere.25 We also introduced the diagonal matrix

r 5diag~r 21,...,r k11 ,ir 2!5diag~ r 1 ,i r 2! ~3.2!

such thatr 5diag(r11, r ) and the diagonal matrix

s85diag~s118 ,...,s(k121)18 ,is28!5diag~s18 ,is28!. ~3.3!

The invariant measure reads

dm~s8,s!52k211mB~s18 ,s1!mF~s28 ,s2!mBF~s8,s!d@j8#d@s18#,

mB~s18 ,s1!5Dk1

~s18!

A2)p51k1 )q51

k121~sp12sq18 !

,

~3.4!

mF~s28 ,s2!5Dk2

4 ~ is28!

)p51k2 )q51

k2 ~ isp22 isq28 !2 ,

mBF~s8,s!5)p51

k1 ) l 51k2 )q51

k121~ isl28 2sp1!~ isl22sq18 !

)p51k121

) l 51k2 ~ isl28 2sp18 !2 .

Here, we have introduced

d@j8#5 )p51

k2

djp8* djp8 , d@s18#5 )p51

k121

dsp18 . ~3.5!

The domain of integration for the bosonic variables is compact and given by

sp1<sp18 <s(p11)1, p51,...,~k121!. ~3.6!

The fermionic eigenvaluesisp28 are related to Grassmann variablesjp8 andjp8* through

ujp8u25 isp28 2 isp2 . ~3.7!

The Jacobian or Berezinian consists of three parts. One of them,mB(s18 ,s1), depends only onbosonic eigenvalues and one,mF(s18 ,s1), only on fermionic eigenvalues, i.e., only on Grassmannvariables. The third part mixes commuting and anticommuting integration variables. To underlineonce more the difference between radial and angular Gelfand–Tzetlin coordinates which is alsopresent in superspace, we mention that the radial measure~3.4! is quite different from the angularone.25



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

As in ordinary space, the recursion formula is an exact map of the group integration onto aniteration exclusively in the radial space. Having done the iteration on the firstk1 levels, we havetreated all Grassmann variables. Thus, in the integrand, we are left with the matrix Bessel functionFk2

(4)(2 i2s2(k121) ,r 2) for USp(2k2) in ordinary space.1 We notice the occurrence of the factor

2 i2 in the argument ofFk2

(4)(2 i2s2(k121) ,r 2). Collecting everything, we arrive at

Fk12k2~s,r !5E )

n51

k121

dm~s(n),s(n21)!exp~ i ~ trgs(n21)2trgs(n)!r n1!

3exp~ is11(k121)r k11!Fk2

(4)~2 i2s2(k121) ,r 2!. ~3.8!

We have sets5s(0) and s85s(1). It is worthwhile to notice that the radial Gelfand–Tzetlincoordinates have a highly appreciated and valuable property: The Grassmann variables only ap-pear as moduli squared in the integrand. Thus, the number of integrals over anticommutingvariables is onlyhalf the number of the independent Grassmann variables. Moreover, advanta-geously, the exponential is a simple function of the integration variables. Thus, we may concludethat the radial Gelfand–Tzetlin coordinates are the natural coordinates of the matrix and thesupermatrix Bessel functions, because their intrinsic features are reflected.

B. Derivation

All crucial steps needed for the derivation of the supersymmetric recursion formula~3.1! carryover from the ordinary recursion formula in Ref. 1. We order the columns of the matrixuPUOSp(k1/2k2) in the form u5@u1 u2 ¯uk1

uk111¯uk11k2#. We also introduce a rectangular

matrix b5@u2¯uk1uk111¯uk11k2

# such thatu5@u1b#. Analogous to the ordinary case, we have

b†b51(k121)2k2,

~3.9!

bb†5 (p52

k1

upup†1 (

p5k111

k11k2

upup†51k12k2

2u1u1† .

We define the square matrixs5b†sb and rewrite the trace in the exponent as

trgu†sur5trg s r 1s11r 11, ~3.10!

with s115u1†su1 . Similar to the ordinary case, the first term on the right-hand side of Eq.~3.10!

depends on the lastk1211k2 columnsup collected inb and the second term depends only onu1 .Thus, it is useful to decompose the invariant measure,

dm~u!5dm~b!dm~u1!, ~3.11!

and to write Eq.~2.1! in the form

Fk12k2~s,r !5E dm~u1!exp~ is11r 11!E dm~b!exp~ i trg s r !. ~3.12!

Since the coordinatesb are locally orthogonal tou1 , the measure dm(b) also depends onu1 .We now generalize the radial Gelfand–Tzetlin coordinates introduced in Ref. 1 for the ordi-

nary spaces to the superspace. Naturally, the projector reads (1k12k22u1u1

†) and we have thedefining equation

~1k12k22u1u1

†!s~1k12k22u1u1

†!ep85sp8ep8 , p51,...,k121,k111,...,k11k2 ~3.13!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

for the (k1211k2) radial Gelfand–Tzetlin coordinatessp8 and the corresponding vectorsep8 aseigenvalues and eigenvectors of the matrix (1k12k2

2u1u1†)s(1k12k2

2u1u1†) which has the gener-

alized rankk1211k2 . Due tou1†ep850, we find

~1k12k22u1u1

†!sep85sp8ep8 , p51,...,k121,k111,...,k2 . ~3.14!

As in Ref. 11, the eigenvaluessp8 are calculated from the characteristic function

z~sp8!5detg~~1k12k22u1u1

†!s2sp8!52sp8 detg~s2sp8!u1†

1k12k2

s2sp8u1 , ~3.15!

which has to be discussed in the limits

z~sp8!→H 0 for p51,...,k121

` for p5k111,...,k11k2 .~3.16!

Thus, together with the normalizationu1†u151, these arek11k2 equations for the elements ofu1 .

The two parts of the integral~3.12! have to be expressed in terms of the radial Gelfand–Tzetlin coordinatessp8 . In a calculation fully analogous to the ordinary case, we find

s115trgs2trgs8. ~3.17!

The eigenvaluestp , p51,...,k121,k111,...,k11k2 of s obtain from the characteristic function

w~ tp!5detg~ s2tp!521

tpdetg~~1k12k2

2u1u1†!s2tp!. ~3.18!

Comparison with Eq.~3.15! shows that the characteristic functionsw(tp) and z(sp8) are, apartfrom the nonzero factor2tp , identical. This impliestp[sp8 , p51,...,k121,k111,...,k11k2 .Thus, by introducing the square matrixu which diagonalizess, we may write

s5b†sb5u†s8u. ~3.19!

By construction,u must be in the group UOSp(k121/2k2), becauses and s share the samesymmetries.

These intermediate results allow us to transform Eq.~3.12! into

Fk12k2~s,r !5E dm~s8,s!exp~ i ~ trgs2trgs8!r 11!E dm~b!exp~ i trg u†s8ur !, ~3.20!

where dm(s8,s) is, apart from phase angles, the invariant measure dm(u1), expressed in the radialGelfand–Tzetlin coordinatess8. To do the integration overb, we view, for the moment, the vectoru1 as fixed and observe that the measure dm(b) is the invariant measure of the group UOSp(k1

21/2k2) under the constraint thatb is locally orthogonal tou1 . The matrix uPUOSp(k1

21/2k2) is constructed fromb under the same constraint. Thus, sinceb and u cover the samemanifold, the integral overb in Eq. ~3.20! must yield the supermatrix Bessel functionF (k121)2k2

(s8, r ) and we arrive at the supersymmetric recursion formula~3.1!. In the last step, weused a line of arguing slightly different from the derivation in ordinary space. In this way weavoided a discussion related to the ill-defined supergroup volume. The invariance of the measureis the crucial property we need for the proof and this holds both in superspace and in ordinaryspace.



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

C. Invariant measure

In order to evaluate the invariant measure, we have to solve the system of equations~3.15! foruvp

(1)u25uup1u2, p51,...,k1 anduap(1)u25uu(k112p)1u21uu(k112p21)1u2, p51,...,k2 in terms of the

bosonic eigenvaluessp85sp18 , p51,...,k121 and the fermionic eigenvaluessk112p8 5sk112p218

5 isp28 , p51,...,k2 ,

15 (p51

k1

uvp(1)u21 (

p51

k2

uap(1)u2, ~3.21!

05 (q51

k1 uvq(1)u2

sq12sp181 (

q51

k2 uaq(1)u2

isq22sp18, p51,...,k121, ~3.22!

zp5 isp28)q51

k1 ~sq12 isp28 !

)q51k2 ~ isq22 isp28 !2 S (

q51

k1 uvq(1)u2

sq12 isp281 (

q51

k2 uaq(1)u2

isq22 isp28D , zp→`, p51,...,k2 .

~3.23!

In Appendix A, we sketch the solution of this system for small dimensions. Inspired by thesesolutions one can conjecture the general solutions and verify them by plugging them directly intoEqs.~3.21!–~3.23!; one finds

uvp(1)u25

)q51k121

~sp12sq18 !)q51k2 ~sp12 isq2!2

)q51k2 ~sp12 isq28 !2)q51,qÞp

k1 ~sp12sq1!, p51,...,k1 ,

~3.24!

uap(1)u252~ isp28 2 isp2!

)q51k121

~ isp22sq18 !)q51,qÞpk2 ~ isp22 isq2!2

)q51,qÞpk2 ~ isp22 isq28 !2)q51

k1 ~ isp22sq1!, p51,...,k2 .

These expressions are reminiscent of the ones derived in Ref. 11 for unitary matrices. However,importantly, all products in~3.24! involving fermionic eigenvalues are squared. This reflects thedegeneracy ofs in the fermion–fermion block. We have introduced new anticommuting variablesjp8 ,jp8* with ujp8u

25 isp28 2 isp2 according to definition~3.7!.From this point on, the invariant measure can be calculated in the same way as for the angular

Gelfand–Tzetlin coordinates, see Ref. 11 for details. The result is summarized in Eq.~3.4!.

IV. THE FUNCTION F22„s ,r …

We use the recursion formula~3.1! to calculate the supermatrix Bessel function forUOSp(2/2). To avoid the imaginary unit in the exponent, we studyF22(2 is,r ). The recursionformula reads

F22~2 is,r !5G22E dm~s8,s!exp~~ trgs2trgs8!r 11!F12~2 is8, r !. ~4.1!

The functionF12(2 is8, r ) is easily found to be

F12~2 is8, r !5G12~122~r 212 ir 12!~s118 2 is128 !!exp~2r 12s12!. ~4.2!

The measure of the coset UOSp(2/2)/UOSp(1/2) is according to formula~3.4! given by

dm~s8,s!5~ is122s118 !)n51

2 ~ is128 2sn1!

A2)n512 ~s118 2sn1!~ is128 2s118 !2

ds118 dj18* dj18 . ~4.3!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

We do the Grassmann integration and find

F22~2 is,r !5G22exp~r 11~s111s21!22is12ir 12!

3Es11

s21mB~s8,s!)

q51

2

~ is122sq1!S 4)j 51

2

~ ir 122r j 1!22~ ir 122r 11!

3 (q51

21

is122sq112M11~s18 ,s1!D exp~s118 ~r 212r 11!!ds118 , ~4.4!

where we have introduced the operator

Mm j~s18 ,s1!51

~ ism22sj 18 ! S 1

2 (n51

k1 1

ism22sn12

1

ism22sj 182 (

n51nÞ j

k1 1

sj 18 2sn182

]

]sj 18 D . ~4.5!

For later purposes, we introduced general indicesm and j . Obviously, the Grassmann integrationyielded eigenvalues in the denominator. This is somewhat surprising because of the followingobservation: we can always parametrize the group elementuPUOSp(2/2) in a noncanonical cosetparametrization in the spirit of an Euler parametrization in ordinary space. Inserting this param-etrization into the defining equation of the supermatrix Bessel function~2.1! one can expand thetrace in all Grassmann variables. The expansion coefficients are polynomials in the commutingintegration variables and—more important—in the matrix elements ofs and r . The invariantmeasure can be expanded in the Grassmann variables as well. It does not depend onr and s.Although this procedure gets rapidly out of hand even for small groups, it is clear that the outcomeof this expansion will be polynomial in the eigenvalues ofs and r . In other words: eigenvaluescan only appear in the denominator by an integration over commuting variables and never by aGrassmann integration. Therefore, before performing any integral over commuting variables, theremust exist a form ofF22(2 is,r ), which is polynomial in the eigenvalues ofs and r .

To remove the denominators and to obtain such a polynomial expression, we use the followingresult. Let f (s18) be an analytic, symmetric function ins1i8 , i 51,...,k1 . Furthermore, define theoperator

Lm~s!5(j 51

k1 1

ism22sj 1

]

]sj 1. ~4.6!

Then the action of the operator on the integral over the bosonic part of the measure is given by

Lm~s!Es11

s21¯E

s(k121)1

sk11mB~s8,s! f ~s18!d@s18#

52Es11

s21¯E

s(k121)1

sk11mB~s8,s! (

j 51

k121

Mm j~s18 ,s1! f ~s18!d@s18#. ~4.7!

This formula is derived in Appendix B.We now setf (s18)5exp(2s118 (r 212r 11)) and insert Eq.~4.7! into Eq. ~4.4!, we arrive at

F22~2 is,r !5G22exp~22is12ir 12!S 4)j 51

2

~ ir 122r j 1!~ is122sj 1!

22(q51

2

~ is122sq1!S ir 122r 212r 112]

]sq1D DF2

(1)~2 is1 ,r 1!, ~4.8!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

whereF2(1)(s1 ,r 1) is the matrix Bessel function of the orthogonal group O~2! in ordinary space as

defined in Ref. 1. Although this can already be taken as the result, we underline the symmetrybetweens and r by using the explicit form~3.21! of Ref. 1 forF2(s1 ,r 1),

F22~2 is,r !5G22expS trg rs2z

2D3S 4)

j 51

2

~ ir 122r j 1!~ is122sj 1!2 (q51

2

~ is122sq1! (p51

2

~ir122rp1!2zd

dzD 2pI 0~z/2!,

~4.9!

where we have introducedz5(s112s21)(r 112r 21) and the modified Bessel functionI 0 as definedin Ref. 26.

The result~4.7! was crucial in the derivation ofF22(2 is,r ). By means of this formula, thedenominator problem was overcome in one step. Because of its importance, we want to gain moreinsight into this problem: In Appendix C, we rederiveF22(2 is,r ) in two other ways. It is clearthat the methods of Appendix C cannot be used for higher dimensionsk1 and 2k2 , but it will helpto understand the mechanisms needed when working with radial Gelfand–Tzetlin coordinates.

V. THE SERIES OF FUNCTIONS Fk 14„s ,r …

We calculate iteratively the four supermatrix Bessel functionsFk14(s,r ) for k151,2,3,4. Wedo this in Secs. V A–V D, respectively.

A. First level k 1Ä1

According to the recursion formulas~3.1! and ~3.8!, the starting point is the matrix Besselfunction for the unitary symplectic groupF2

(4)(s2 ,r 2), which was already calculated in Ref. 1. Upto a normalization, we have

F2(4)~ i2s2 ,r 2!5 (

vPS2S 1

D22~ is2!D2

2~v~ ir 2!!2

1

D23~ is2!D2

3~v~ ir 2!! Dexp~2tris2v~ ir 2!!.

~5.1!

Since the subgroup O~1! of UOSp(1/4) is trivial, no commuting integral has to be performed toderive F14(2 is,r ). Inserting the measure~3.4! into the recursion formula and performing theGrassmann integrations yields straightforwardly

F14~2 is,r !5G14exp~ trg rs!S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D ~2~ is212s11!~ ir 212r 11!21!

3~2~ is222s11!~ ir 222r 11!21!2G14

exp~ trg~rs!!

D23~ ir 2!D2

3~ is2!1~ ir 12↔ ir 22!. ~5.2!

The exchange term (ir 12↔ ir 22) accounts for the permutation groupS2 in Eq. ~5.1!. Anticipatingthat the structure ofF14(2 is,r ) will, remarkably, survive on all levels up toF44(2 is,r ), we statethat F14(2 is,r ) essentially consists of two parts. A comparison with Eqs.~4.2!, and~5.1! showsthat the first part ofF14(2 is,r ) is a product of an exponential with three other terms. The firstone,

S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D , ~5.3!

stems from the integral over the USp~4! subgroup. The other two terms can be identified with thesupermatrix Bessel functions



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

F12~2 is,r ! with s5diag~s11,is12,is12!, r 5diag~r 11,ir 12,ir 12! ~5.4!

and

F12~2 is,r ! with s5diag~s11,is22,is22!, r 5diag~r 11,ir 22,ir 22!. ~5.5!

The second part can be considered as a correction term, which destroys the product structure ofF14(2 is,r ). We may identify the different parts of the product with the integrations over thecorresponding subsets of the group. Thus,F2

(4)(2 is2 ,r 2) arises from the integration over theUSp~4! subgroup, the O~1! integration yields unity, and the other two factors come from theintegration over the coset UOSp(1/4)/(USp(4) O(1)).

B. Second level k 1Ä2

We now have to do one integration over a commuting variable. After the Grassmann integra-tion, we are left with a considerable amount of terms. To arrange them in a convenient way, weintroduce the following notation for the product of two operatorsD1(s)D2(s) acting on a functionf (s), we define

@D1→~s!D2~s!# f ~s!5D1~s!D2~s! f ~s!2~D1~s!D2~s!! f ~s!. ~5.6!

This means, an operator with an arrow only acts on the terms outside the squared bracket. Withthis notation we can write

F24~2 is,r !5G24exp~ tr~r 2s2!1r 21~s111s21!!Es11

s21dmB~s18 ,s1!F)

i 51

2

)j 51

2

~ isi22sj 1!

3S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D3S 4)

i 51

2

~ ir 122r i1!12(k51

2r 212 ir 12

is122sk112M

11~s18 ,s1!D

3S 4)i 51

2

~ ir 222r i1!12(k51

2r 212 ir 22

is222sk112M21~s18 ,s1!D

1S 1

D22~ ir 2!D3

2~ is2!1

1

D23~ ir 2!D2

4~ is2! D3S 4

is122s118M21~s18 ,s1!2

2

is222s118M11~s18 ,s1! D

12

D23~ ir 2!D2

4~ is2!S 2 tr r 12tr ir 21(

i 51

21

is222si1D M11~s18 ,s1!

22

D23~ ir 2!D2

4~ is2!S 2 tr r 12tr ir 21(

i 51

21

is122si1D M21~s18 ,s1!

24

D23~ ir 2!D2

3~ is2! (k51

2

)j 51

2r 212 ir j 2

sk12 isj 2Gexp~s118 ~r 112r 21!!1~ ir 12↔ ir 22!.

~5.7!

As in Sec. IV, a denominator problem occurs. It becomes obvious in the productM 11

→ (s18 ,s1)M21(s18 ,s1). Thus, we expect an identity similar to formula~4.7!. This identity shouldmap a product of operatorsL1(s)L2(s) acting on the integral onto a product of operators



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

M11(s18 ,s1)M21(s18 ,s1) acting under the integral. Neither the outer operators,Lm(s), nor the innerones,Mm j(s), commute. Hence, the desired identity must be a nontrivial one. It is given by thefollowing result.

We have the same conditions as in formula~4.7!, furthermore we define

@Lm→~s!Ll~s!#5 (

n51

k1

(q51

k1 1

~ ism22sn1!~ isl22sq1!

]2

]sn1]sq1. ~5.8!

Then the following formula holds

@Lm→~s!Ll~s!#E

s11

s21¯E

s(k121)1

sk11mB~s8,s! d@s18# f ~s18!

5Es11

s21¯E

s(k121)1

sk11mB~s8,s!F (

j 51

k121

(k51

k121

Mm j

~s18 ,s1!Mlk~s18 ,s1!

21

isl22 ism2(j 51

k121 S 1

ism22sj 18Ml j ~s18 ,s1!2

1

isl22sj 18Mm j~s18 ,s1! D

21

2 (kÞ j

k1211

~ ism22sk18 !~ ism22sj 18 !~ isl22sk18 !~ isl22sj 18 !G f ~s18! d@s18#. ~5.9!

The derivation is along the same lines as the one for formula~4.7!, it also involves formula~4.7!.With the identities~5.9! and~4.7! the denominator problem is again solved in one step. After somefurther manipulations we arrive at

F24~2 is,r !52pG24expS trg rs2z

2D S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D3F S 4)

i 51

2

~r i12 ir 12!~si12 is12!2(i 51j 51

2

~sj 12 is12!~r i12 ir 12!2z]→

]z D3S 4)

i 51

2

~r i12 ir 22!~si12 is22!2(i 51j 51

2

~sj 12 is22!~r i12 ir 22!2z]

]zD G I 0~z/2!

22pG24expS trg rs2z

2D 2

D23~ ir 2!D2

3~ is2!

3 (i 51k51

2

)j 51

2

~si12 isj 2!~r k12 ir j 2!I 0~z/2!22pG24expS trg rs2z

2D3

1

D23~ ir 2!D2

3~ is2!~~ trgs!~ trg r !21!z

]

]zI 0~z/2!1~ ir 12↔ ir 22!. ~5.10!

As in Sec. IV, we used the composite variablez5(s112s12)(r 112r 12). A comparison with Eqs.~4.8! and~5.2! shows the similarity in the structures ofF24(2 is,r ) andF14(2 is,r ). The formeralso decomposes into two parts. The first part is a product, whose factors can be assigned to theintegrations over the different submanifolds of the group in the same way as in the case ofF14(2 is,r ). The other one can be interpreted as a correction term due to the noncommutativity ofthe operatorsLm in formula ~5.9!.



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

C. Third level k 1Ä3

This structure ofFk14(2 is,r ) emerging in the previous calculations is likely to also bepresent for arbitraryk1 . However, fork1.2, we have so far not been able to treat the general case.Fortunately, in important physics applications, one matrix argument of the supermatrix Besselfunction has an additional twofold degeneracy in the boson–boson block. In this case, it ispossible to carry on the recursion up toF44(2 is,r ) by extending the techniques developed fork151 andk252. Thus, from now on, we restrict ourselves to this case.

At first sight, one might hope to achieve some simplification by applying the projectionprocedure onto the degenerate matrix, because this results in a considerable simplification of theinvariant measure. However, it turned out that the integrations are easier if one does the recursionwith the nondegenerate coordinates. Hence, we use the measure as it stands in Eq.~3.4!. WeconsiderF34(2 is,r ) in the case that

r 15diag~r 11,r 21,r 21!. ~5.11!

Having performed the Grassmann integral, one can arrange the terms in a way similar to Eq.~5.7!.The complete expression and further details are given in Appendix D. We then can use formula~5.9! and find after some further algebra

F34~2 is,r !54 G34exp~ tr r 2s2!~r 212 ir 12!~r 212 ir 22!F S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D3S 4)

k51

2

~r k12 ir 12!)k51

3

~sk12 is12!12(k51

3

)j Þk

3

~sj 12 is12!

3S r 111r 212 ir 122]→

]sk1D D

3S 4)k51

2

~r k12 ir 22!)k51

3

~sk12 is22!12(k51

3

)j Þk

3

~sj 12 is22!

3S r 111r 212 ir 222]

]sk1D D 2

4

D23~ ir 2!D2

3~ is2! (i 51

3

)j 51j Þ i

3

~sj 12 is12!~sj 12 is22!

3S r 112 1r 21

2 1r 11r 212~r 111r 21!~ ir 121 ir 22!1 ir 12ir 22

2~r 111r 212 ir 122 ir 22!]

]si1D2

2

D23~ ir 2!D2

4~ is2!

3(i ,k

3

)j 51j Þ i

3

)l 51lÞk

3

~sj 12 is12!~sl12 is22!S ]

]si12

]

]sk1D G

3F3(1)~2 is1 ,r 1!, 1~ ir 12↔ ir 22!, ~5.12!

whereF3(1)(s1 ,r 1) is the matrix Bessel function of the orthogonal group O~3!. We notice that the

structure ofF14(2 is,r ) andF24(2 is,r ) reappears inF34(2 is,r ).

D. Fourth level k 1Ä4

In the calculation ofF44(2 is,r ), we again consider the case that the matrixr is degenerate,

r 15diag~r 11,r 11,r 21,r 21!. ~5.13!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

The main problem is to find a convenient representation for the matrix Bessel functionF3(1)

(2 is18 , r 1) appearing on the third level in Eq.~5.12!. It turns out that the representation derived inAppendix B of Ref. 1 is very well suited to our purpose. Due to the degeneracy inr 1 , the originalthreefold integral can be reduced to an integral over just one single variable

F3(1)~2 is18 , r 1!5

exp~r 21 tr s18!

Aur 112r 21uE

2`

1`

dtexp~ i ~r 112r 21!t !

) i 513 Asi18 2 i t

. ~5.14!

Here, we again neglected the normalization because we want to fix it afterwards as explainedpreviously. Similarly,F4

(1)(2 is1 ,r 1) can be written as a double integral,

F4(1)~2 is1 ,r 1!5

exp~r 21 tr s1!

ur 112r 21uE

2`

1`

dt1E2`

1`

dt2ut12t2uexp~ i ~r 112r 21!~ t11t2!!

) i 514 )n51

2 Asi12 i t n

. ~5.15!

Singularities have to be taken care of appropriately. After inserting Eq.~5.10! into the recursionformula and performing the Grassmann integration, one can arrange the terms in a similar way asin the case ofF34(2 is,r ). At this point, we notice that formulas~4.7! and ~5.9! need to besupplemented by further identities. We state the most important one in the following.

The same conditions as for formula~4.7! apply. Moreover, we define the operator

Lm~s!5 (q51

k1 1

ism22sq1

]2

]sq12 1

1

2 (qÞn

1

~ ism22sq1!~sq12sn1! S ]

]sq12

]

]sn1D . ~5.16!

Then we have

Lm~s!Es11

s11¯E

s(k121)1

sk11mB~s8,s!d@s18# f ~s18!

52Es11

s21¯E

s(k121)1

sk11mB~s8,s!F(

j 51Mm j

→ ~s18 ,s1!]

]sj 18G f ~s18! d@s18#. ~5.17!

Again, the proof is along the same lines as the proof of formula~4.7! and the proof of formula~5.9! in Ref. 1.

Thus, there is a family of rules to transform operators symmetric insi1 Lm(s),Lm(s) actingonto an integral into an operator acting under the integral. We need one more such transformationrule which tells us how the product@Lm(s) L l(s)# transforms into operators acting under theintegral. This formula and further details are given in Appendix E. Collecting everything, wefinally arrive at

F44~2 is,r !54G44exp~2tr ~r 2s2!!)i , j

~r i12 ir j 2!F S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D3S 8)

i 51

2

~r i12 ir 12!)j 51

4

~sj 12 is12!14(i 51

4

)j Þ i

4

~sj 12 is12!S r 111r 212 ir 122]→

]si1D D

3S 8)i 51

2

~r i12 ir 22!)j 51

4

~sj 12 is22!14(i 51

4

)j Þ i

4

~sj 12 is22!S r 111r 212 ir 222]

]si1D D

216

D23~ ir 2!D2

3~ is2! (i 51

4

)j Þ i

4

~sj 12 is12!~sj 12 is22!S r 112 1r 21

2 1r 11r 212~ ir 121 ir 22!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

3~r 111r 21!1 ir 12ir 2221

2trg r

]

]si1D2

8

D23~ ir 2!D2

4~ is2! (i 51j 51

4

)lÞ i

4

~sl12 is12!

3)lÞ j

4

~sl12 is22!S ]

]si12

]

]sj 1D GF4

(1)~2 is1 ,r 1!1~ ir 12↔ ir 22!. ~5.18!

We mention that in the derivation of this result we frequently used properties of the matrix BesselfunctionsF3

(1)(s1 ,r 1) and F4(1)(s1 ,r 1) that only hold for the case that one matrix has an addi-

tional degeneracy.

VI. ASYMPTOTICS AND NORMALIZATION

The asymptotic behavior of the supermatrix Bessel functions calculated in the previous sec-tions is a useful check which also allows us to fix the normalization constants. We find from theexpressions in Eqs.~4.1!, ~4.8! and in Eqs.~5.2!, ~5.10!, ~5.12!, and~5.18!,

lims→`r→`

Fk12k2~2 is,r !52k1k2Gk12k2

) l 51k1 )m51

k2 ~sl12 ism2!~r l12 ir m2!

Dk2

2 ~ is2!Dk2

2 ~ ir 2!

3det@exp~2si2r j 2!# i , j 51¯k2lim

s1→`r 1→`

Fk1

(1)~2 is1 ,r 1!. ~6.1!

In the degenerate case, each degenerate eigenvalue contributes according to its multiplicity. Theasymptotics of the matrix Bessel functions of the orthogonal group is given by27,28

limt→0

Fk1

(1)~2 is1 /t,r 1!5C(k1)t (k121)k1/4det@exp~sn1r m1 /t !#n,m51,...,k1

uDk1~s1!Dk1

~r 1!u1/2 , ~6.2!

where the constant can be found in Muirhead’s book,28

C(k1)5G~k1/2!

k1!pk1

2/22k1/4. ~6.3!

Thus we find

limt→0

Fk12k2~2 is/t,r !52k1k2t ((k122k2)21(k122k2))/4C(k1)Gk12k2

3det@exp~sn1r m1 /t !#n,m51,...,k1

det@exp~2si2r j 2 /t !# i , j 51¯k2

ABk1k2~s!Bk1k2

~r !~6.4!

for the asymptotic behavior.On the other hand, the supermatrix Bessel function relates to the kernel of Dyson’s Brownian

motion in superspace.22 Due to the normalization of the Gaussian integral,

S p

2t D2((k122k2)21(k122k2))/4

2k222k22k1/2E d@s#expS 2

1

t~s2r! D51, ~6.5!

the kernel



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

Gk1k2~s,r ,t !5S p

2t D2((k122k2)21(k122k2))/4

2k222k22k1/2E

uPUOSp(k1/2k2)dm~u!expS 2

1

t~s2r! D

~6.6!

is also normalized. Since it is obviously connected with to the supermatrix Bessel function by

Gk1k2~s,r ,t !5S p

2t D2((k122k2)21(k122k2))/4

2k222k22k1/2expS 2

1

t~ trgs21trg r 2! DFk12k2

~2 is/t,r !

~6.7!

we can fix the normalization by using the asymptotic behavior

limt→0

Gk1k2~s,r ,t !5S p

2D 2((k122k2)21(k122k2))/42k2

22k22k1/2

k1! k2!

3det@d~si12r j 1!# i , j 51¯k1

det@d~si22r j 2!# i , j 51¯k2

ABk1k2~s!Bk1k2

~r !~6.8!

of the kernel. Comparing Eq.~6.4! with Eq. ~6.8!, we find

Gk12k25

23k2(k22k1)1k12/425k2/22k1/2

p ((k122k2)212k1222k2)/4k2!G~k1/2!

. ~6.9!

We mention that this calculation also shows that the diffusion kernels of the one-point functionand of the two-point function of Dyson’s Brownian motion,22 i.e., the functionG (2k)k(s,r ,t) whichwas denoted byGk(s,r ,t) in Ref. 22, indeed satisfy the proper initial condition.

VII. APPLICATIONS

Although we focus in this contribution on the mathematical aspects, we now briefly commenton a particular kind of application. As the reader will realize, our results derived in Sec. VI are, insome sense, more general than what we need in those applications on which we focus here. Wetake this as an indication that explicit results for even more complex supermatrix Bessel functionscan also be obtained. The results of the previous sections yield the kernels of the supersymmetricanalog of Dyson’s Brownian Motion for the GOE and the GSE in the casesk51 andk52. We donot present the physics background here. The reader interested in these applications is asked toconsult Refs. 19–21 for generalities and Ref. 22, in particular Sec. 4.2, for the issue discussedhere. In the present contribution, we use the same notations and conventions. We restrict ourselvesto the transition toward the GOE and suppress the indexc. The corresponding formulas for thetransition toward the GSE are derived accordingly. We treat the one- and two-point functions inSecs. VII A and VII B, respectively.

Forrester and Nagao29 derived expressions for generalized one-point functions of Dyson’sBrownian motion model with Poissonian initial conditions. They used an expansion of the Greenfunction in terms of Jack polynomials. Datta and Kunz30 employed a supersymmetric technique toaddress the two-level correlation function of the Poisson GOE transition. They arrive at a finitenumber of ordinary and Grassmannian integrals which are still to be performed. In our approach,we also arrive at a representation of the correlation function in terms of a finite number ofintegrals. However, since we managed to integrate over almost all angular integrals in the previoussections, our result contains considerably less integrals, in particular, no Grassmannian ones. It hasa clear structure due to the fact that, apart from two integrals, all others are eigenvalue integrals,i.e., live in the curved eigenvalue space of Dyson’s Brownian motion. Moreover, since our for-mulas for the kernel are valid on all scales, our result is also exact for finite level number.



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

A. Level density

We use the result~4.9!, derived in Sec. IV, for the supermatrix Bessel functionF22

(2 is,r ). Using the replacementr→(x1J) and s→s/t and the relation~6.7!, we obtain thediffusion kernel for the level density

G1~s,x1J,t !5~2p!21/2J1

2texpS 2

1

t~s112x12J1!22

1

t~s212x12J1!21

2

t~ is122x11J1!2D

3S 22J1

t )j 51

2

~ is122sj 1!1 (q51

2

~ is122sq1!D . ~7.1!

We take the derivative with respect to the source variableJ1 and arrive at the level density

R1~x1 ,t !51

~2p!3/2t E expS 21

t~s112x1!22

1

t~s212x1!21

2

t~ is122x1!2D

3~~ is122s11!1~ is122s21!!B21~s! Z1(0)~s! d@s#, ~7.2!

where the Berezinian is given by Eq.~2.10! for k152 and k251. This result is exact for anarbitrary initial condition and for arbitraryN. In the case of a diagonal matrixH (0) as the initialcondition, we have for arbitraryk,

Zk(0)~s!5E d@H (0)#P~H (0)!)

n51

N) j 51

k ~ isj 22Hnn(0)!

) j 512k ~sj 11 i«2Hnn

(0)!1/2 ~7.3!

and analogously for the GSE. This has to be used in Eq.~7.2! for k51.In the limit t→` the stationary distribution of classical Gaussian random matrix theory is

recovered. This can be seen by rewriting Eq.~7.1! for the rescaled energyx15x1 /t and therescaled source variableJ15J1 /t, see also Ref. 22. In this limit the initial condition yields unityand we arrive at an integral representation of the one-point correlation function of the pure GOE,

R1~x1!51

~2p!3/2IE exp~2~s112x1!22~s212x1!21~ is122x1!2!

3us112s21u

~ is122s11!~ is122s21!S 1

is122s111

1

is122s21D ~ is12!

N

~s111 i«!N/2~s211 i«!N/2 d@s#

~7.4!

where the symbolI denotes the imaginary part. Equation.~7.4! is equivalent to the classicalexpressions for the one-point functions as in Mehta’s book.19

Finally, we state an integral expression for the one-point function for the case of Poissonianinitial conditions, see Eq.~5.1! of Ref. 22. We have

Zk(0)~s!5S E dzp~z!

) j 51k ~ isj 22z!

) j 512k ~sj 11 i«2z!1/2D N

. ~7.5!

Inserting this initial condition fork51 into Eq. ~7.2! yields the level density of a transitionensemble between Poisson regularity and GOE in terms of a fourfold integral. A further analysiswill be published elsewhere.



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

B. Two-point function

The result~5.18!, derived in Sec. V, for the supermatrix Bessel functionF44(2 is,r ) gives,with the replacementr→(x1J) ands→s/t and according to Eq.~6.7!, the diffusion kernel for thetwo-point function

Gk~s,x1J,t !5expS 21

t~ trgs21trg ~x1J!2! DF44~22is/t,x1J!. ~7.6!

The derivative with respect to the source terms leaves us with the two-point correlation function

R2~x1 ,x2 ,t !526G44

t2p2 IE SAB42~s!AuD4~s1!uZ2(0)~s!expS 2

1

t~ trs1

212x1212x2

222~ is122x1!2

22~ is222x2!2! D(k, j

F 1

~ is122sk1!~ is222sj 1! S 2x1

t2

]

]sj 1D S 2x2

t2

]

]sk1D

1t

2~x12x2!~ is122 is22!~ is122sk1!~ is222sj 1! S 2x1

t2

]

]sj 1D S 2x2

t2

]

]sk1D

321

4

t

~x12x2!~ is122sk1!~ is222sj 1!~ is122 is22!2 S ]

]sj 12

]

]sk1D G

3F4(1)~22is1 /t,x! D d@s#1~x1↔x2!. ~7.7!

The last line indicates that the integral withx1 and x2 interchanged has to be added. Since theterms antisymmetric inx1 andx2 are antisymmetric insi2 ands22 as well this yields just a factor2 in Eq. ~7.7!. The symbolI denotes a certain linear combination ofR2(x1 ,x2 ,t) as explained inRef. 22. The normalization constant obtains from Eq.~6.9! and is given byG4452(2p)24. Thisresult is an exact expression for the two-point function of Dyson’s Brownian motion for everyinitial condition. Plugging in the initial condition of Eq.~7.5! for k52, we find an integralrepresentation of the two-point function for the transition toward the GOE. We notice thatF4

(1)(22is1 /t,x) is, according to Eq.~5.15!, given as a double integral.In the previous discussion, we referred to properties of the kernels which can be seen from the

explicit formulas. In Ref. 22, only the explicit form of the kernel forb52 was available. How-ever, some of the general properties of the kernels forb51 andb54 could be anticipated in Ref.22 from scaling relations of the supersymmetric version of Dyson’s Brownian motion. The explicitformulas derived in the present contribution allow one to derive the integral representations~7.2!and ~7.7! for the one-point and for the two-point function. Moreover, we emphasize that thekernels for the supersymmetric version of Dyson’s Brownian motion are the same on all energyscales.22 Thus, the integral representation derived here for the two-level correlation function is,apart from the initial condition, the same on the so-called unfolded scale which is relevant forphysics applications. The initial condition on the unfolded scale is found along the lines given inRef. 22.

VIII. SUMMARY AND CONCLUSION

We extended the recursion formula of Ref. 1 to superspace. Due to the group structures insuperspace, we could restrict ourselves to the unitary orthosymplectic supergroup. As in theordinary case, the recursion formula is an exact map of a group integration onto an iteration in theradial space. We used it to calculate explicit expression for certain supermatrix Bessel functions.



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

In ordinary space, we saw that the matrix Bessel functions are only special cases of the radialfunctions.1 We have not yet studied this further, but in our opinion a similar generalization is likelyto also exist in superspace.

It is a major advantage of the radial Gelfand–Tzetlin coordinates in superspace that theGrassmann variables appear only as moduli squared. Thus, the number of Grassmann integrals isa priori reduced by half. As we showed in detail, this is a highly welcome feature for explicitcalculations. As a remarkable consequence of this recursive way to proceed, the structure of thesupermatrix Bessel functions is only very little influenced by the matrix dimension. We also sawthat the basic structures of the supermatrix Bessel function for smaller matrix dimensions surviveduring the iteration to higher ones. The matrix Bessel functions in ordinary space show similarfeatures. There, the structure of the matrix Bessel functions is much more influenced by the groupparameterb than by the matrix dimension. However, as in ordinary space, it remains a challengeto find the structure of these functions for arbitrary matrix dimension.

An interesting feature occurred which sheds light on the general properties of the recursion.Total derivatives showed up in the integral over the commuting variables after having done theGrassmann integration. Since similar terms already occurred in ordinary space, they are likely tobe an intrinsic property of the recursion formula. Here, we succeeded in constructing a series ofoperator identities to remove them. This was a crucial step for the application of the recursionformula. A deeper understanding of these identities is highly desired.

It should be emphasized that the total derivatives are no boundary terms in the sense ofRothstein. We showed in detail that such terms cannot occur because we always work in acompact space. Thus, according to a theorem due to Berezin,2 the transformation of the invariantmeasure to our radial Gelfand–Tzetlin coordinates cannot yield Rothstein boundary terms. How-ever, if further integration over the eigenvalues is required, such terms can emerge.

As an application, we worked out some kernels for the supersymmetric analog of Dyson’sBrownian motion.

The radial Gelfand–Tzetlin coordinates are the natural coordinate system for the matrixBessel functions in superspace. This parametrization represents the appropriate tool for the recur-sive integration of Grassmann variables. Once the particular features of this parametrization arebetter understood, they may allow for the evaluation of higher dimensional group integrals.

ACKNOWLEDGMENTS

We thank B. Balantekin and Z. Pluhar for useful discussions. We acknowledge financialsupport from the Deutsche Forschungsgemeinschaft—T. G. for a Heisenberg fellowship and H. K.for a doctoral grant. H. K. also thanks the Max-Planck-Institute for financial support.

APPENDIX A: RADIAL GELFAND–TZETLIN COORDINATES FOR THE UNITARYORTHOSYMPLECTIC GROUP UOSp „k 1Õ2k 2…

We wish to express the moduli squared of the elements of an orthogonal (k1/2k2) dimensionalunit supervector in radial Gelfand–Tzetlin coordinates. To illustrate the mechanism, we start withthe smallest nontrivial case, the group UOSp(2/4). We notice that there are at first sight minor, yetcrucial, differences to the calculation in Ref. 11 where we also started with the smallest nontrivialcase. The set of solutions of the Gelfand–Tzetlin equations~3.23! involves one bosonic and twofermionic eigenvalues. The eigenvalue equations read

15 (p51

2

~ uvp(1)u21uap

(1)u2!, ~A1!

05 (q51

2 S uvq(1)u2

sq12s1(1) 1

uaq(1)u2

isq22s1(1)D , ~A2!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

z15 is2(1))

q51

2~sq12 is2

(1)!

~ isq22 is2(1)!2 (

q51

2 S uvq(1)u2

sq12 is2(1) 1

uaq(1)u2

isq22 is2(1)D , ~A3!

where the last equation has to be solved in the limitz1→`. The bosonic equation~A2! has aunique solutions1

(1)5s118 . Takings118 as new parameter, Eqs.~A1! and ~A2! can be solved,

uvp(1)u25

sp12s118

sp12sq1S 12 (

k51

2isk22sq1

isk22s118uak

(1)u2D , p51,2. ~A4!

We insert these relations in Eq.~A3! and obtain

z15 is2(1)~s118 2 is2

(1)!)q51

2~sq12 is2

(1)!

~ isq22 is2(1)!2 S 11 (

k51

2ck

isk22 is2(1) uak

(1)u2D ~A5!

with z1→`. Here, we have defined the commuting variables

ck5)q51

2 ~ isk22sq1!

isk22s118, k51,2. ~A6!

It remains to determine the set of solutions of the fermionic eigenvalue equation~A5!. To this end,both sides are inverted

05 )q51

2

~ isq22 is2(1)!2S 12 (

k51

2ck

isk22 is2(1) uak

(1)u212)k51

2ck

isk22 is2(1) uak

(1)u2D . ~A7!

We can now take the square root on both sides

05 )q51

2

~ isq22 is2(1)!S 12

1

2 (k51

2ck

isk22 is2(1) uak

(1)u213

4 )k51

2ck

isk22 is2(1) uak

(1)u2D . ~A8!

The most general form of the fermionic eigenvalue is

is2(1)5a01 (

k51

2

akuak(1)u21a12)

k51

2

uak(1)u2. ~A9!

After inserting this ansatz in Eq.~A8!, we obtain two sets of solutions for the coefficientsai0 ,ai12

andai j with i 51,2,j 51,2

is128 5 is121S c11c1c2

is122 is22ua2

(1)u2D ua1(1)u2

2,

~A10!

is228 5 is221S c21c1c2

is222 is12ua1

(1)u2D ua2(1)u2

2.

Remarkably, we havea125a2150. This allows us to write the nilpotent part ofisk28 as themodulus squared of a new anticommuting coordinate,

isk28 5 isk21ujk8u2. ~A11!

We solve Eq.~A10! for uap(1)u2, insert the results in Eq.~A4! and arrive at



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

uvp(1)u25

~sp12s118 !)n512 ~sp12 isn2!2

~sp12sq1!)n512 ~sp12 isn28 !2,

~A12!

uap(1)u252~ isp28 2 isp2!

~ isp22s118 !~ isp22 isq2!2

~ isp22 isq28 !2)n512 ~ isp22sn1!

, p,q51,2 , qÞp.

The structure of Eq.~A12! indicates the form of the solutions for groups of higher order as theywere stated in Eq.~3.24!. They are checked by inserting them directly into the Gelfand–Tzetlinequations~3.23!. The algebra needed is, although tedious, straightforward and similar to the onehere.

APPENDIX B: DERIVATION OF FORMULA „4.7…

The technique we use is an extension of the one developed in Appendix D of Ref. 1. First, werewrite the integral in terms ofQ functions. The left hand side reads

Lm~s!E mB~s8,s! f ~s18!d@s18#)k< l

Q~sk12sl18 !)l ,n

Q~sl18 2sn1!. ~B1!

The integration domain is now the real axis for all variables. The action ofLm(s) on the integralyields:

E SmB~s8,s! (i 51

k1

(j 51

k1211

2

21

~ is122si1!~si12sj 18 !f ~s18!)

k< lQ~sk12sl18 !)

l ,nQ~sl18 2sn1!D d@s18#

1E mB~s8,s! (i 51

k1 1

ism22si1

]

]si1)k< l

Q~sk12sl18 !)l ,n

Q~sl18 2sn1!d@s18#. ~B2!

We decompose the first term in partial fractions and find

E SmB~s8,s! (i 51

k1

(j 51

k121 1

2

21

~ is122si1!~ is122sj 18 !f ~s18!2Dk1

~s18! f ~s18!

3 (j 51

k121 1

is122sj 18

]

]sj 18

1

A2) i 51k1 ~si12sj 18 !

)k< l

Q~sk12sl18 !)l ,n

Q~sl18 2sn1!d@s18#

1E mB~s8,s! f ~s8!(i 51

k1 1

ism22si1

]

]si1)k< l

Q~sk12sl18 !)l ,n

Q~sl18 2sn1!d@s18#. ~B3!

An integration by parts yields

E mB~s8,s! (j 51

k121

~2Mm j~s18 ,s1!!)k< l

Q~sk12sl18 !)l ,n

Q~sl18 2sn1!d@s18#1E mB~s8,s! f ~s8!

3S (i 51

k1 1

ism22si1

]

]si11 (

j 51

k1211

ism22sj 18

]

]sj 18D)

k< lQ~sk12sl18 !)

l ,nQ~sl18 2sn1!d@s18#.

~B4!

The derivatives of theQ functions yieldd distributions. Upon integration of thed distribution thetwo terms in the last integral cancel each other. Hence the last term vanishes identically. Thiscompletes the proof.



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

APPENDIX C: ALTERNATIVE DERIVATIONS OF F22„s ,r …

We present two different alternative derivations. We do this in some detail, becausethe calculations give helpful informations on the roˆle played by the radial Gelfand–Tzetlincoordinates.

First, we use an angular parametrization of the coset UOSp(2/2)/UOSp(1/2) by writing thefirst column ofuPUOSp(2/2) as

u153A12uau2 cosq

A12uau2 sinq

1

&a

1

&a*

4 . ~C1!

This is a canonical way to parametrize the supersphereS1u2 that is isomorphic to the cosetUOSp(2/2)/UOSp(1/2), see Ref. 31. It coincides with a special choice of theangular Gelfand–Tzetlin coordinates. The invariant measure is in these coordinates simply dm(u1)5da* dadq.Thus, one directly obtains the the volumeV(S1u2)50, see Ref. 31. In the parametrization of themeasure byradial Gelfand–Tzetlin coordinates~4.3!, one has to perform the Grassmann integra-tion and to apply formula~4.7! to achieve this result.

Although we use a different coordinate system, we still take advantage of the recursionformula ~3.1!. To use it in the parametrization~C1!, one has to solve the Gelfand–Tzetlin equa-tions ~3.21!–~3.23! for the eigenvalues. The unique solution of the bosonic equation~3.22! is

s118 5a01) i 51

2 ~si12a0!

is122a0uau2, a05

s111s21

22

s112s12

2cos 2q. ~C2!

The fermionic equation yields

is128 5 is121) i 51

2 ~si12 is12!

is122a0uau2. ~C3!

After inserting Eqs.~C2! and ~C3! and the measure dm(u1) into the recursion formula~3.1!, theGrassmann integration can be performed. Remarkably, we arrive at the denominator–free expres-sion

F22~s,r !5G22E0

2p

dqexpS trg rs2z

21

z

2cos 2q D

3F S )i 51

2

~r i12 ir 12!~si12 is12!11

2 (i 51j 51

2

~sj 12 is12!~r i12 ir 12!D2

1

2 S ir 1221

2~r 111r 21! D ~s112s21!cos 2q2

z

8~ ir 122r 21!~s112s21!sin2 2qG .

~C4!

To make contact with Eq.~4.9! one has to realize that in Eq.~C4! an additional total derivativeappears in the integrand. This becomes obvious if one adds and subtractsz/4 cos 2q in the squarebracket of Eq.~C4!,



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

F22~s,r !5G22E0

2p

dq expS trg rs2z

21

z

2cos 2q D

3S )i 51

2

~r i12 ir 12!~si12 is12!11

2 (i 51j 51

2

~sj 12 is12!~r i12 ir 12!2z

4cos 2q D

1ir 122r 21

r 212r 11E

0

2p

dq]2

]~2q!2 expS trg rs2z

2~12cos 2q! D . ~C5!

While the first integral reproduces Eq.~4.9!, the second one vanishes identically. In general, inperforming Grassmann integrations, one has to take care of boundary contributions.2,24 Thesecontributions can appear whenever even coordinates are shifted by nilpotents and the function oneintegrates does not have compact support.2 However, in our case the basis space is always givenby an n dimensional sphere, i.e., by a compact manifoldwithout boundary. Thus in a properlychosen coordinate system, no boundary terms should appear. With regard to Eq.~C5! this means:the fact that the last term vanishes is a direct consequence of the compactness of the circle and ofthe analyticity of the function, that we integrate. However, in the radial Gelfand–Tzetlin coordi-nates, only the moduli squared of the vectoru1 are determined. Therefore, not the whole sphere,but only a (2n11)th segment of it is covered by Eq.~3.24!. In our case, not the circle but only aquarter of it is parametrized. This is allowed since the supermatrix Bessel functions depend onlyon the moduli squareduui1u2. Nevertheless, one has to ensure that the introduction of theseartificial boundaries does not alter the result. To this end we use the following integration formula.

Let s11,s118 ,s21 be real and letj8,j8* be anticommuting. Furthermore, define

f ~s118 ,j,j* !5 f 0~s118 !1 f 1~s118 !uju2, ~C6!

with two analytic functionsf 0(s118 ), f 1(s118 ). Then the integral

I 5Es11

s21ds118 dj* dj f ~s118 ,j,j* ! ~C7!

transforms under a shift ofs118 by nilpotents

s118 5y1g~y!uju2 ~C8!

in the following way:

I 5Es11

s21dy dj* dj

]s118

]yf ~y~s118 !,j,j* !2@ f 0~s21!g~s21!2 f 0~s11!g~s11!#. ~C9!

The proof is by direct calculation. The second term in Eq.~C9! is often referred to as boundaryterm. It can be viewed as the integral of a total derivative, i.e., an exact one-form, that has to beadded to the integration measure for functions with noncompact support.24 For functions of anarbitrary number of commuting and anticommuting arguments, a similar integral formula holdswith additional boundary terms.2 In going from the canonical coordinates (q,a,a* ) to the radialones (s118 ,j18 ,j18* ), in principle boundary terms can arise, since the bosonic Gelfand–Tzetlineigenvalue~C2! contains nilpotents. However, the crucial quantity isg(y) in formula ~C9! which,in our case, is given by

g~s118 !5) i 51

2 ~si12s118 !

is122s118. ~C10!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

Thus,g(s118 ) causes the boundary term to vanish ats11 ands21. It is the product structure of theleft-hand side of Eq.~3.24! which always guarantees the vanishing of the boundary terms, whenone goes from the Cartesian set of coordinates to the radial Gelfand–Tzetlin coordinates.

Therefore, one may think of the denominators, arising in Eqs.~4.4! and~4.5!, as belonging tototal derivatives of functions, which vanish at the boundaries. Keeping this in mind we derive Eq.~4.9! in yet another way. We expand the product

)q51

k1

~ ism22sq1!5 (n50

k1 1

n!~ ism22sj 18 !n

]n

]~sj 18 !n )q51

k1

~sj 18 2sq1!, ~C11!

and insert it into the integral

Es11

s21¯E

s(k121)1

sk11mB~s18 ,s1!Km j~s18 ,s1! f ~s18!d@s18#

5Es11

s21¯E

s(k121)1

sk11mB~s18 ,s1!)

n51

k1

~ ism22sn1!Mm j~s18 ,s1! f ~s18! d@s18#. ~C12!

We can remove the term proportional to (ism22sj 18 )22 in the integrand by an integration by parts.Through the expansion~C11!, the vanishing of the boundary terms is assured. We arrive at

Km j~s18 ,s1!52 (n52

k1 1

n!~ ism22sj 18 !n22

]n

]~sj 18 !n )q51

k1

~sj 18 2sq1!

1)q51

k1 ~ ism22sq1!2)q51k1 ~sj 18 2sq1!

ism22sj 18

3S 1

2 (q51

k1 1

ism22sq12

1

2 (q51

k1 1

sj 18 2sq12 (

q51qÞ j

k1 1

sj 18 2sq182

]

]sj 18 D . ~C13!

We notice that in the new operatorKm j(s18 ,s1) all denominators of the type (ism22sj 18 )21 havedisappeared. Fork152, we calculate

K1152~ is121s118 2s112s21!]

]sq18, ~C14!

which can be inserted into Eq.~4.4! by using the definition~C12!. Finally, the result~4.9! isreproduced by the substitution

s118 5s111s21

22

s112s12

2cos 2q, ~C15!

see Eq.~C2!. In other words, we have seen that the result of this procedure is summarized informula ~4.7!.

Finally, some remarks are in order: First, from this discussion, one might conclude that theradial Gelfand–Tzetlin coordinates are less adapted to the problem than the canonical parametri-zation ~C1!, because, in the latter, no denominators appear. We stress that this is not true. Cer-tainly, the denominators appear due to the shift of the bosonic variable by nilpotents in Eq.~C2!.However, the difficulty in deriving Eq.~4.9! is the identification of the different parts of theintegrand after the Grassmann integration. Some of them belong to total derivatives and thisproblem exists in both parametrizations. Second, we emphasize that the appearance of total de-rivatives in the integrand is not a peculiarity of supersymmetry. Already in Ref. 1 where the matrix



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

Bessel functions in ordinary space were treated we had to solve a similar problem. The appearanceof these total derivatives is an intrinsic property of the recursion formula. A geometrical interpre-tation of this phenomenon is highly desired.

APPENDIX D: DETAILS FOR THE DERIVATION OF F34„À is ,r …

We always consider the case that one matrix has an additional degeneracy according to Eqs.~5.11! and ~5.13!. We introduce the notation

Si j 5~si12 isj 2!, Ri j 5~r i12 ir j 2!. ~D1!

Due to the degeneracy,F24(2 is, r ) simplifies enormously as compared to the general result~5.10!. We insert it into the recursion formula~3.1! and do the trivial integral over the O~2!subgroup. After performing the Grassmann integrals we arrive at an expression similar to Eq.~5.7!,

F34~2 is,r !54G34exp~ tr r 2s21r 11~s111s21!!E dmB~s18 ,s1!)i 51

2

R1i )j 51

3

Sji

3F S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D S 4)i 51

2

Ri122(k51

3

R21Sk121

12(j 51

2

M1 j

~s18 ,s1!D S 4)i 51

2

Ri222(k51

3

R22Sk22112(

j 51

2

M2 j~s18 ,s1!D1S 1

D22~ ir 2!D2

3~ is2!1

1

D23~ ir 2!D2

4~ is2! D3(

j 51

2 S 4

is122sj 18M2 j~s18 ,s1!2

4

is222sj 18M1 j~s18 ,s1! D

2S 1

D22~ ir 2!D2

3~ is2!1

1

D23~ ir 2!D2

3~ is2! D )k51j 51

22

isk22sj 181

2

D23~ ir 2!D2

4~ is2!

3S trg r 1r 112(i 51

2

Si221D (

j 51

2

M1 j~s18 ,s1!22

D23~ ir 2!D2

4~ is2!

3S trg r 1r 112(i 51

2

Si121D (

j 51

2

M2 j~s18 ,s1!

24

D23~ ir 2!D2

3~ is2! (k51

3

)j 51

2

R2 jSk j21Gexp~~s218 1s118 !~r 212r 11!!1~ ir 12↔ ir 22!.

~D2!

Formulas~5.9! and~4.7! are needed to remove the denominators, in a way similar as forF24(s,r ).A single sum( j 51

2 M1 j (s18 ,s1) transforms according to formula~4.7!. Moreover, we observe thatparts of Eq.~D2! together with the product( j 51

2 M1 j

(s18 ,s1)(k512 M2k(s18 ,s1) yield exactly the

integrand of formula~5.9!. Thus, it can be transformed accordingly. After rearranging terms, wearrive at the result~5.12!.



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

APPENDIX E: DETAILS FOR THE DERIVATION OF F44„À is ,r …

For the recursion, we needF34(s,r ) with degenerater 5diag(r11,r 21,r 21) according to Eq.~5.11!. Using the integral representation~5.14! for F3

(1)(2 is18 , r 1) we find the helpful identity

]

]si18

]

]sj 18exp~2r 21 tr s18!F3

(1)~2 is18 , r 1!51

2

1

si18 2sj 18S ]

]si182

]

]sj 18Dexp~2r 21trs18!F3

(1)~2 is18 , r 1!.

~E1!

We stress that this relation, which is crucial in the derivation, only holds, because of the degen-eracy in the matrixr 1 . Employing Eq.~5.14! and another identity,

(i 51

3]

]si18exp~2r 21tr s18!F3

(1)~2 is18 , r 1!5~r 112r 21!exp~2r 21 tr s18!F3(1)~2 is18 , r 1!. ~E2!

We insertF34(s8, r ) into the recursion formula~3.1!. We then can arrange the terms emergingfrom the Grassmann integration in a way similar to the former cases. We obtain

F44~2 is,r !54G44exp~ tr r 2s21r 11 tr s1!E dmB~s18 ,s1!)i 51

2

R2i)j 51

4

Sji

3F S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D3S 8R21R11

2 24R11R21(k51

4

Sk12114(

j 51

3 S R212]→

]sj 18D M1 j

→~s18 ,s1!

3S 8R22R122 24R12R22(

k51

4

Sk22114(

j 51

3 S R222]→

]sj 18D M2 j~s18 ,s1!D

116

D23~ ir 2!D2

4~ is2! (j 51

3

M1 j→~s18 ,s1!S 1

2R11R12S trg r 2(

i 51

4

Si121D

1S r 212r 112]

]sj 18D S R22R121R11R121R11R211

1

2~R121R22!(

i 51

4

Si121D D

216

D23~ ir 2!D2

4~ is2! (j 51

3

M2 j→~s18 ,s1!S 1

2R11R12S trg r 2(

i 51

4

Si221D

1S r 212r 112]

]sj 18D S R11R211R11R121R12R221

1

2~R121R22!(

i 51

4

Si221D D

216

D23~ ir 2!D2

3~ is2! (k51

4

)i , j

2

Ri j Sk j21exp~2r 11 tr s18!F3

(1)~2 is18 , r 1!1C~s,r !

1~ ir 12↔ ir 22!. ~E3!

Again, all operators with an arrow are understood to act only onto the term outside the squaredbracket, i.e., onto exp(2r11 tr s18)F3

(1)(2 is18 , r 1). In the functionC(s,r ), we summarized theterms that are expected to arise due to noncommutativity of some operators acting on the integraland some operators acting under the integral. The last two lines in formula~5.9! are examples ofsuch terms



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

C~s,r !54G44exp~ tr r 2s21r 11trs1!E dmB~s18 ,s1!)i 51

2

R2i)j 51

4

Sji F S 1

D22~ ir 2!D2

3~ is2!

11

D23~ ir 2!D2

4~ is2!D (

j 51

3 S R11R121~R111R12!~r 212r 11!2~R111R12!]→

]sj 18D

3S 16

is122sj 18M2 j~s18 ,s1!2

16

is222sj 18M1 j~s18 ,s1!D

2S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2!D

3)k51

2

)j 51

3 S R11R121~R111R12!~r 212r 11!2~R111R12!]→

]sj 18D

38

isk22sj 182S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2!D

3(j ,k

3 S ~r 212r 11!2]→

]sj 18D S ~r 212r 11!2

]→

]sk18D M1 j

→M2k28

D22~ ir 2!D2

3~ is2!

3(j 51

3 S ~r 212r 11!2]→

]sj 18D S (

k51

4

Sk221M j 1

→2 (k51

4

Sk121M j 2D

18

D22~ ir 2!D2

3~ is2!S (

iÞ jM j 1M j 2

→S ]→

]si182

]→

]sj 18D D Gexp~2r 11trs18!F3

(1)~2 is18 , r 1!.

~E4!

In order to evaluate Eqs.~E3! and ~E4! we need some more properties of the matrix Besselfunction F4

(1)(2 is1 ,r 1). We investigate the action ofLk on F4(1)(2 is1 ,r 1) using the integral

representation~5.15!.After a straightforward calculation involving an integration by parts we find

Lk exp~2r 11tr s1!F4(1)~2 is1 ,r 1!

5(i 51

41

isk22si1S ~r 212r 11!

21~r 212r 11!]

]si1Dexp~2r 11 tr s1!F4

(1)~2 is1 ,r 1!. ~E5!

Now Eqs.~E3! and ~E4! can be enormously simplified by the observation that

~~r 212r 11!Lk2Lk!exp~2r 11trs1!F4(1)~2 is1 ,r 1!50, ~E6!

which follows directly from Eq.~E5!. We find for Eq.~E3!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

F44~2 is,r !54G44exp~ tr r 2s21r 11trs1!E dmB~s18 ,s1!)i 51

2

R1i )k51

4

Ski

3F S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D S R12R11S 8R12R2224R22(k51

4

Sk221D

3S 8R11R2124R21(k51

4

Sk12114(

j 51

3

M1 j→~s18 ,s1!D

1R12R11S 8R11R2124R21(k51

4

Sk121D

3S 8R22R1224R22(k51

4

Sk22114(

j 51

3

M2 j~s18 ,s1!D1(

j ,iR11S r 212r 112

]→

]sj 18D M1i

→~s18 ,s1!M2 j~s18 ,s1!

1(j ,i

R12S r 212r 112]→

]si18D M1i

→~s18 ,s1!M2 j~s18 ,s1!

1(j ,i

R21R22M1 j→~s18 ,s1!M2i~s18 ,s1!1

8

D23~ ir 2!D2

4~ is2!R11R12

3S trg r 2(i 51

4

Si121D (

j 51

3

~M1 j~s18 ,s1!2M2 j~s18 ,s1!!

216

D23~ ir 2!D2

3~ is2! (k51

4

)i , j

2

Ri j Sk j21G

3exp~2r 11tr s18!F3(1)~2 is18 , r 1!1C~s,r !1~ ir 12↔ ir 22!. ~E7!

The terms contained in Eq.~E4! simplify, too. We arrive at

C~s,r !54G44exp~ tr r 2s21r 11 tr s1!E dmB~s18 ,s1!)i 51

2

R1i)j 51

4

Sji

3F S 1

D22~ ir 2!D2

3~ is2!1

1

D23~ ir 2!D2

4~ is2! D3(

j 51

3 S R11R121~R111R12!~r 212r 11!2~R111R12!]→

]sj 18D

3S 16

is122sj 18M2 j~s18 ,s1!2

16

is222sj 18M1 j~s18 ,s1! D

2S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D3)

k51j 51

2 S R11R121~R111R12!~r 212r 11!2~R111R12!]→

]sj 18D

38

isk22sj 181

8

D22~ ir 2!D2

3~ is2! S (iÞ jM j 1

→~s18 ,s1!M j 2~s18 ,s1!S ]→

]si182

]→

]sj 18D D G

3exp~2r 11tr s18!F3(1)~2 is18 , r 1!. ~E8!



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

To further evaluate the expressions, we can now invoke a symmetry argument between the eigen-valuesr 11 and r 21, respectively. Since the productR11R12 appears as a prefactor in front of theintegral~E7!, R21R22 must also appear as a prefactor in the final result. Thus, all terms in Eqs.~E7!and~E8! which do not containR21R22 as a factor must yield zero. The remaining terms which areproportional toR21R22 can again be treated using formulas~4.7! and ~5.9!. However, we want toshow explicitly that this line of arguing is correct and that the other terms indeed vanish. To thisend, we need an additional identity to treat the operator product

(j 51

2]→

]sj 18M

1 j~s18 ,s1!(

k51

2

M2k~s18 ,s1!. ~E9!

The required identity is given by the following formula: The same conditions as for formula~4.7!apply, furthermore we define

Lm→~s!Ln~s!5(

i , j

1

~ ism22si1!~ isn22sj 1!

]3

]si1]sj 12 1

1

2 (i , j

1

~ ism22si1!~ isn22sj 1!

3]

]si1(kÞ j

k1 1

sj 12sk1S ]

]sj 12

]

]sk1D . ~E10!

Then we have

Lm→~s!Ln~s!E

s11

s21¯E

s(k121)1

sk11mB~s8,s!d@s18# f ~s18!

5Es11

s21¯E

s(k121)1

sk11 F (j 51

k121

(i 51

k121

Mmi→ ~s18 ,s1!

]→

]sj 18Mn j~s18 ,s1! f ~s18!2

1

isn22 ism2

3 (i 51

k121 S 1

ism22si18

]→

]si18Mni~s18 ,s1!2

1

isn22si18

]→

]si18Mmi~s18 ,s1! D

21

2 (kÞ l

1

~ ism22sk18 !~ isn22sk18 !~sk18 2sl18 !2

]→

]sk1

11

2 (kÞ l

1

~ ism22sk18 !~ isn22sl18 !~sk18 2sl18 !2

]→

]sk1G f ~s18!mB~s8,s!d@s18#. ~E11!

The proof is similar to the one of formula~4.7!. We notice that the arrow in Eq.~E10! is usedslightly differently than previously. The operatorLm

→(s) acts also on a part ofLn(s). This is notconsistent with the definition in Eq.~5.6!. However, since this is obvious where it occurs, we stilluse the same arrow. We can now translate the left-hand side of Eq.~E8! into an expression in termsof F4

(1)(2 is,r ). After some further manipulations involving the identities in Eqs.~E6!, ~E1!, and~E2! we arrive at



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

F44~2 is,r !5G44exp~ tr r 2s21r 11 tr s1!)i , j

2

Rji )k51

4

SkiF S 1

D22~ ir 2!D2

2~ is2!1

1

D23~ ir 2!D2

3~ is2! D3S 8R12R2224R22(

k51

4

Sk22124L2~s!D S 8R11R2124R21(

k51

4

Sk12124L1~s!D

28

D23~ ir 2!D2

4~ is2!S trg r 2(

i 51

4

Si121D ~L1~s!2L2~s!!

216

D23~ ir 2!D2

3~ is2! (k51

4

)j 51

2

R2 jSk j21Gexp~2r 11tr s18!F4

(1)~s18 , r 1!~ ir 12↔ ir 22!.

~E12!

After rearranging terms this yields the result~5.18! for F44(2 is,r ).

1T. Guhr and H. Kohler, J. Math. Phys.43, 2707~2002!, preceding paper; math-ph/0011007.2F. A. Berezin,Introduction to Superanalysis, ~Reidel, Dordrecht, 1987!, Vol. 9.3V. C. Kac, Commun. Math. Phys.53, 31 ~1977!.4V. C. Kac, Adv. Math.26, 8 ~1977!.5K. B. Efetov, Adv. Phys.32, 53 ~1983!.6K. B. Efetov,Supersymmetry in Disorder and Chaos, ~Cambridge University Press, Cambridge, 1997!.7J. J. M. Verbaarschot and M. R. Zirnbauer, Ann. Phys.~N.Y.! 158, 78 ~1984!.8J. J. M. Verbaarschot, H. A. Weidenmu¨ller, and M. R. Zirnbauer, Phys. Rep.129, 367 ~1985!.9T. Guhr, J. Math. Phys.32, 336 ~1991!.

10C. Itzykson and J. B. Zuber, J. Math. Phys.21, 411 ~1980!.11T. Guhr, Commun. Math. Phys.176, 555 ~1996!.12I. M. Gelfand and M. L. Tzetlin, Dokl. Akad. Nauk SSSR71, 825 ~1950!.13S. L. Shatashvili, Commun. Math. Phys.154, 421 ~1993!.14J. Alfaro, R. Medina, and L. Urrutia, J. Math. Phys.36, 3085~1995!.15T. Guhr and H. A. Weidenmu¨ller, Ann. Phys.~N.Y.! 199, 412 ~1990!.16M. R. Zirnbauer, Commun. Math. Phys.141, 503 ~1991!.17M. R. Zirnbauer, Phys. Rev. Lett.69, 1584~1992!.18A. D. Mirlin, A. Mu ller-Groeling, and M. R. Zirnbauer, Ann. Phys.~N.Y.! 236, 325 ~1994!.19M. L. Mehta,Random Matrices, 2nd ed.~Academic, San Diego, 1991!.20F. Haake,Quantum Signatures of Chaos, 2nd ed.~Springer, Berlin, 2001!.21T. Guhr, A. Muller-Groeling, and H. A. Weidenmu¨ller, Phys. Rep.299, 189 ~1998!.22T. Guhr, Ann. Phys.~N.Y.! 250, 145 ~1996!.23F. J. Wegner, Z. Phys. B: Condens. Matter49, 297 ~1983!.24M. J. Rothstein, Trans. Am. Math. Soc.299, 387 ~1987!.25T. Guhr and H. Kohler~unpublished!.26M. Abramowitz and I. A. Stegun,Handbook of Mathematical Functions, 9th ed.~Dover, New York, 1970!.27 Harish-Chandra, Am. J. Math.80, 241 ~1958!.28R. J. Muirhead,Aspects of Multivariate Statistical Theory, ~Wiley, New York, 1982!.29P. J. Forrester and T. Nagao, Nucl. Phys. B532, 733 ~1998!.30N. Datta and H. Kunz, cond-mat/0006488.31M. R. Zirnbauer, J. Math. Phys.37, 4986~1996!.



129.12.233.45 On: Mon, 08 Dec 2014 14:17:53

Documents

Recursive construction for a class of radial functions. II. Superspace