q-Difference Systems for the Jackson Integral of Symmetric Selberg Type

We provide an explicit expression for the first-order -difference system for the Jackson integral of symmetric Selberg type. The q-difference system gives a generalization of the -analog of contiguous relations for the Gauss hypergeometric function. As a basis of the system, we use a set of symmet...

Повний опис

Збережено в:
Бібліографічні деталі
Опубліковано в: :Symmetry, Integrability and Geometry: Methods and Applications
Дата:2020
Автор: Ito, Masahiko
Формат: Стаття
Мова:Англійська
Опубліковано: Інститут математики НАН України 2020
Онлайн доступ:https://nasplib.isofts.kiev.ua/handle/123456789/211007
Теги: Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Цитувати:q-Difference Systems for the Jackson Integral of Symmetric Selberg Type. Masahiko Ito. SIGMA 16 (2020), 113, 31 pages

Репозитарії

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859802899163381760
author Ito, Masahiko
author_facet Ito, Masahiko
citation_txt q-Difference Systems for the Jackson Integral of Symmetric Selberg Type. Masahiko Ito. SIGMA 16 (2020), 113, 31 pages
collection DSpace DC
container_title Symmetry, Integrability and Geometry: Methods and Applications
description We provide an explicit expression for the first-order -difference system for the Jackson integral of symmetric Selberg type. The q-difference system gives a generalization of the -analog of contiguous relations for the Gauss hypergeometric function. As a basis of the system, we use a set of symmetric polynomials introduced by Matsuo in his study of the -KZ equation. Our main result is an explicit expression for the coefficient matrix of the -difference system in terms of its Gauss matrix decomposition. We introduce a class of symmetric polynomials called interpolation polynomials, which includes Matsuo's polynomials. By repeated use of three-term relations among the interpolation polynomials, we compute the coefficient matrix.
first_indexed 2026-03-16T07:30:58Z
format Article
fulltext Symmetry, Integrability and Geometry: Methods and Applications SIGMA 16 (2020), 113, 31 pages q-Difference Systems for the Jackson Integral of Symmetric Selberg Type Masahiko ITO Department of Mathematical Sciences, University of the Ryukyus, Okinawa 903-0213, Japan E-mail: mito@sci.u-ryukyu.ac.jp Received April 29, 2020, in final form October 29, 2020; Published online November 08, 2020 https://doi.org/10.3842/SIGMA.2020.113 Abstract. We provide an explicit expression for the first order q-difference system for the Jackson integral of symmetric Selberg type. The q-difference system gives a generalization of q-analog of contiguous relations for the Gauss hypergeometric function. As a basis of the system we use a set of the symmetric polynomials introduced by Matsuo in his study of the q-KZ equation. Our main result is an explicit expression for the coefficient matrix of the q-difference system in terms of its Gauss matrix decomposition. We introduce a class of symmetric polynomials called interpolation polynomials, which includes Matsuo’s poly- nomials. By repeated use of three-term relations among the interpolation polynomials we compute the coefficient matrix. Key words: q-difference equations; Selberg type integral; contiguous relations; Gauss de- composition 2020 Mathematics Subject Classification: 33D60; 39A13 1 Introduction The Gauss hypergeometric function 2F1 ( a, b c ;x ) = Γ(c) Γ(a)Γ(c− a) ∫ 1 0 za−1(1− z)c−a−1(1− xz)−b dz, where Re c > Re a > 0 and |x| < 1, satisfies the contiguous relations 2F1 ( a, b c ;x ) = 2F1 ( a, b+ 1 c+ 1 ;x ) − xa(c− b) c(c+ 1) 2F1 ( a+ 1, b+ 1 c+ 2 ;x ) (1.1) and 2F1 ( a, b c ;x ) = 2F1 ( a+ 1, b c+ 1 ;x ) − xb(c− a) c(c+ 1) 2F1 ( a+ 1, b+ 1 c+ 2 ;x ) . (1.2) These contiguous relations for the Gauss hypergeometric function are extended to a difference system for a function defined by multivariable integral with respect to the Selberg type kernel [34] Ψ(z) := n∏ i=1 zα−1i (1− zi)β−1(x− zi)γ−1 ∏ 1≤j<k≤n |zj − zk|2τ . For the integral 〈ei〉 := ∫ C ei(z)Ψ(z) dz1 · · · dzn, i = 0, 1, . . . , n, This paper is a contribution to the Special Issue on Elliptic Integrable Systems, Special Functions and Quan- tum Field Theory. The full collection is available at https://www.emis.de/journals/SIGMA/elliptic-integrable- systems.html mailto:mito@sci.u-ryukyu.ac.jp https://doi.org/10.3842/SIGMA.2020.113 https://www.emis.de/journals/SIGMA/elliptic-integrable-systems.html https://www.emis.de/journals/SIGMA/elliptic-integrable-systems.html 2 M. Ito where ei(z) is the function specified by ei(z) := n−i∏ j=1 (x− zj) n∏ k=n−i+1 (1− zk) and C is some suitable region, the (n + 1)-tuple (〈e0〉, 〈e1〉, . . . , 〈en〉) satisfies the following difference system. Let δij be the symbol of Kronecker’s delta. Proposition 1.1 ([14, Theorem 2.2]). Let Tα be the shift operator with respect to α → α + 1, i.e., Tαf(α) = f(α+ 1) for an arbitrary function f : C→ C. Then Tα(〈e0〉, 〈e1〉, . . . , 〈en〉) = (〈e0〉, 〈e1〉, . . . , 〈en〉)M, (1.3) where the (n+ 1)× (n+ 1) matrix M is written in terms of its Gauss matrix decomposition as M = LDU = U ′D′L′. (1.4) Here L = (lij)0≤i,j≤n, D = (djδij)0≤i,j≤n, U = (uij)0≤i,j≤n are the lower triangular, diagonal, upper triangular matrices, respectively, given by lij = (−x)i−j ( n− j n− i ) (γ + jτ ; τ)i−j (α+ γ + 2jτ ; τ)i−j , dj = xj(α; τ)j(α+ γ + 2jτ ; τ)n−j (α+ γ + (j − 1)τ ; τ)j(α+ β + γ + (n+ j − 1)τ ; τ)n−j , uij = (−1)j−i ( j i ) (β + (n− j)τ ; τ)j−i (α+ γ + 2iτ ; τ)j−i , and U ′ = (u′ij)0≤i,j≤n, D′ = (d′jδij)0≤i,j≤n, L′ = (l′ij)0≤i,j≤n are the upper triangular, diagonal, lower triangular matrices, respectively, given by u′ij = (−x−1)j−i ( j i ) (β + (n− j)τ ; τ)j−i (α+ β + 2(n− j)τ ; τ)j−i , d′j = xj(α+ β + 2(n− j)τ ; τ)j(α; τ)n−j (α+ β + γ + (2n− j − 1)τ ; τ)j(α+ β + (n− j − 1)τ ; τ)n−j , l′ij = (−1)i−j ( n− j n− i ) (γ + jτ ; τ)i−j (α+ β + 2(n− i)τ ; τ)i−j , where (x; τ)0 := 1 and (x; τ)i := x(x+ τ)(x+ 2τ) · · · (x+ (i− 1)τ) for i = 1, 2, . . .. In particular, when n = 1 the system (1.3) is given by Tα(〈e0〉, 〈e1〉) = (〈e0〉, 〈e1〉) ( α+ γ 0 −xγ xα )( α+ β + γ β 0 α+ γ )−1 , (1.5a) Tα(〈e0〉, 〈e1〉) = (〈e0〉, 〈e1〉) ( α −β 0 x(α+ β) )( α+ β 0 γ α+ β + γ )−1 . (1.5b) The system (1.5b) can be rewritten as the following system of three-term equations (α+ β + γ)Tα〈e1〉 = −β〈e0〉+ x(α+ β)〈e1〉, (1.6) and γTα〈e1〉+ (α+ β)Tα〈e0〉 = α〈e0〉. (1.7) q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 3 Since, for n = 1 〈e0〉 = ∫ 1 0 zα−1(1− z)β−1(x− z)γ dz and 〈e1〉 = ∫ 1 0 zα−1(1− z)β(x− z)γ−1 dz, under the conditions Reα > 0, Reβ > 0 and |x| > 1 we see that the equations (1.6) and (1.7) exactly coincide with the contiguous relations (1.1) and (1.2), respectively, after the substitutions α→ a, β → c−a, γ → −b and x→ 1/x. Therefore the difference system (1.3) expressed in terms of Gauss matrix decomposition (1.4) can be regarded as a natural extension of the contiguous relations (1.1) and (1.2). For further applications of the difference system (1.3) in random matrix theory, see [14]. In the discussion of the result being a generalization of contiguous relations of the Gauss hypergeometric function, the analog of the equation (1.1) for the Selberg integral can be found in [15, 21]. Next we would like to discuss a q-analogue of the difference system (1.3) in Proposition 1.1. This is one of the aims of this paper. For an arbitrary c ∈ C∗ we use the c-shifted factorial for x ∈ C (x; c)i =  (1− x)(1− cx) · · · ( 1− ci−1x ) if i = 1, 2, . . . , 1 if i = 0, 1( 1− c−1x )( 1− c−2x ) · · · ( 1− cix ) if i = −1,−2, . . . , and the c-binomial coefficient[ i j ] c = (c; c)i (c; c)i−j(c; c)j . We also use the symbol (x; c)∞ := ∏∞ i=0 ( 1 − cix ) for |c| < 1. Throughout this paper we fix q ∈ C∗ with |q| < 1. For a point ξ = (ξ1, . . . , ξn) ∈ (C∗)n and a function f(z) = f(z1, . . . , zn) on (C∗)n we define the following sum over the lattice Zn by∫ ξ∞ 0 f(z) dqz1 z1 ∧ · · · ∧ dqzn zn := (1− q)n ∑ (ν1,...,νn)∈Zn f ( ξ1q ν1 , . . . , ξnq νn ) , (1.8) if it converges. We call it the Jackson integral of f(z). By definition the Jackson integral (1.8) is invariant under the q-shift ξi → qξi (i = 1, . . . , n). Let Φn,m(z) and ∆(z) be the functions on (C∗)n specified by Φn,m(z) := n∏ i=1 { zαi m∏ r=1 ( qa−1r zi; q ) ∞ (brzi; q)∞ } ∏ 1≤j<k≤n z2τ−1j ( qt−1zk/zj ; q ) ∞ (tzk/zj ; q)∞ , (1.9) ∆(z) := ∏ 1≤i<j≤n (zi − zj), where t = qτ . For a point ξ = (ξ1, . . . , ξn) ∈ (C∗)n and an arbitrary symmetric function φ(z) = φ(z1, . . . , zn) on (C∗)n we set 〈φ, ξ〉 := ∫ ξ∞ 0 φ(z)Φn,m(z)∆(z) dqz1 z1 ∧ · · · ∧ dqzn zn , which we call the Jackson integral of symmetric Selberg type. In the study of q-difference de Rham cohomology associated with Jackson integrals [4, 7], Aomoto and Kato [8] showed that the Jackson integral of symmetric Selberg type satisfies q-difference systems of rank ( n+m−1 m−1 ) 4 M. Ito when the parameters are generic. When m = 1 the Jackson integral of symmetric Selberg type is equivalent to the q-Selberg integral defined by Askey [11] and proved by others, see [6, 13, 16, 20] for instance. See also recent references [15, Section 2.3] and [18]. q-Selberg integral is a very active area of research with important connections to special functions, combinatorics, mathematical physics and orthogonal polynomials (see [1, 23, 24, 25, 32, 35] and [17, Section 5]). Using the Jackson integral of symmetric Selberg type for m = 2, Matsuo [28, 29] constructed a set of solutions of the q-KZ equation. Varchenko [33] extended Matsuo’s construction to more general setting of the q-KZ equation using the Jackson integral of symmetric Selberg type for general m. Writing ar = xr, br = qβrx−1r in (1.9), the q-KZ equation they studied can be regarded as the q-difference system with respect to the q-shift xr → qxr (r = 1, . . . ,m). In another context, writing ar = qx−1r , br = qµrxr in (1.9), Kaneko [22] showed an explicit expression for the q-difference system with respect to the q-shift xr → qxr (r = 2, . . . ,m) satisfied by the Jackson integral of symmetric Selberg type for general m with special constraints µ2 = · · · = µm = 1 or µ2 = · · · = µm = −τ . With these constraints the q-difference system degenerates to be very simple and it can also be regarded as a generalization of the second order q-difference equation satisfied by Heine’s 2φ1 q-hypergeometric function. In this paper, we fix m = 2 for (1.9), and study two types of q-difference systems for the Jackson integral of symmetric Selberg type for Φ(z) = Φn,2(z). One is the q-difference system with respect to the shift α → α + 1, and the other is the system with respect to the q-shifts ai → qai and bi → q−1bi simultaneously. For these purposes, we define the set of symmetric polynomials {ei(a, b; z) | i = 0, 1, . . . , n}, where ei(a, b; z) := 1 ∆(z) ×A n−i∏ j=1 (1− bzj) n∏ j=n−i+1 ( 1− a−1zj ) ∏ 1≤k<l≤n ( zk − t−1zl ) , (1.10) which we call Matsuo’s polynomials. The symbol A means the skew-symmetrization (see the definition (2.1) of A in Section 2). With these symmetric polynomials, we denote 〈ei(a, b), ξ〉 := ∫ ξ∞ 0 ei(a, b; z)Φ(z)∆(z) dqz1 z1 ∧ · · · ∧ dqzn zn . (1.11) We assume that∣∣qa−11 a−12 b−11 b−12 ∣∣ < ∣∣qα∣∣ < 1 and ∣∣qa−11 a−12 b−11 b−12 ∣∣ < ∣∣qαt2n−2∣∣ < 1 for convergence of the Jackson integrals (1.11). (See [19, Lemma 3.1] for details of convergence.) For the polynomials (1.10) let R be the (n+ 1)× (n+ 1) matrix defined by( en(a2, b1; z), en−1(a2, b1; z), . . . , e0(a2, b1; z) ) = ( e0(a1, b2; z), e1(a1, b2; z), . . . , en(a1, b2; z) ) R. (1.12) The transition matrix R is called the R-matrix in the context of [29]. Matsuo [29] gave the q-difference system with respect to the q-shifts ai → qai and bi → q−1bi simultaneously, using Matsuo’s polynomials as follows. Proposition 1.2 (Matsuo). Let Tq,u be the q-shift operator with respect to u→ qu, and T−1q,bi Tq,ai (i = 1, 2) denote the q-shift operator with respect to ai → qai and bi → q−1bi simultaneously. Then, the Jackson integrals of symmetric Selberg type satisfy the q-difference system with respect to T−1q,bi Tq,ai (i = 1, 2) given by T−1q,bi Tq,ai ( 〈en(a2, b1), ξ〉, 〈en−1(a2, b1), ξ〉, . . . , 〈e0(a2, b1), ξ〉 ) q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 5 = ( 〈en(a2, b1), ξ〉, 〈en−1(a2, b1), ξ〉, . . . , 〈e0(a2, b1), ξ〉 ) Ki, (1.13) whose coefficient matrices Ki are expressed as K1 = R−1D1 and K2 = D2 ( T−1q,b2 Tq,a2R ) , where R is the (n+ 1)× (n+ 1) matrix given by (1.12), and D1, D2 are the diagonal matrices given by D1 = (( qαtn−1 )n−i δij ) 0≤i,j≤n, D2 = (( qαtn−1 )i δij ) 0≤i,j≤n. Remark 1.3. If we replace ai and bi as ai = xi and bi = qβix−1i , respectively, then (1.13) simplifies to the case considered by Matsuo, and then T−1q,bi Tq,ai in (1.13) becomes the single q- shift operator Tq,xi , and the system (1.13) coincides with the q-KZ equation (see [28, 29, 33]). For the problem of finding the explicit form of the coefficient matrix Ki for the system (1.13) Aomoto and Kato used the information of a connection matrix [9] between two kinds of fundamental solutions of (1.13) specified by their asymptotic behaviors. Based on Birkhoff’s classical theory they introduced a way to derive the explicit form of the coefficient matrix for a linear ordinary q-difference system from its connection matrix. They call their method the Riemann–Hilbert approach for q-difference equations from connection matrices [5], and they presentedKi explicitly when n = 1 and 2 as an example of their method (see [10, p. 272, examples]). The problem of finding the explicit form of the coefficient matrix for the equivalent q-difference system with respect to Tq,xi was also studied by Mimachi [30, 31]. As a basis of the system Mimachi [31] introduced a family of Schur polynomials different from Matsuo’s polynomials, and he calculated the entries of the coefficient matrix explicitly when n = 1, 2 and 3. From Proposition 1.2, if we want to know the coefficient matrices Ki of the above q-difference systems, it suffices to give the explicit expression for the transition matrix R or its inverse R−1. Theorem 1.4 ([12, 26]). The matrix R is written in terms of its Gauss matrix decomposition as R = LRDR UR = U ′RD ′ R L ′ R, (1.14) where LR = ( lRij ) 0≤i,j≤n, DR = ( dR j δij ) 0≤i,j≤n, UR = ( uR ij ) 0≤i,j≤n are the lower triangular, diagonal, upper triangular matrices, respectively, given by lRij = [ n− j n− i ] t−1 (−1)i−jt−(i−j 2 )(a2b2tj ; t)i−j( a−11 a2t−(n−2j−1); t ) i−j , (1.15a) dR j = ( a1a −1 2 t−j ; t ) n−j(a2b1; t)j (a1b2; t)n−j ( a−11 a2t−(n−j); t ) j , (1.15b) uR ij = [ j i ] t−1 ( a1b1t n−j ; t ) j−i( a1a −1 2 tn−i−j ; t ) j−i , (1.15c) and U ′R = ( uR ′ ij ) 0≤i,j≤n, D′R = ( dR ′ j δij ) 0≤i,j≤n, L′R = ( lR ′ij ) 0≤i,j≤n are the upper triangular, diagonal, lower triangular matrices, respectively, given by uR ′ ij = [ j i ] t (−1)j−it( j−i 2 )(a−11 b−11 t−(n−i−1); t ) j−i( b−11 b2ti+j−n; t ) j−i , (1.16a) dR ′ j = ( b1b −1 2 tn−2j+1; t ) j ( a−12 b−11 t−(n−j−1); t ) n−j( a−11 b−12 t−(j−1); t ) j ( b−11 b2t−(n−2j−1); t ) n−j , (1.16b) lR ′ij = [ n− j n− i ] t ( a−12 b−12 t−(i−1); t ) i−j( b1b −1 2 tn−2i+1; t ) i−j . (1.16c) 6 M. Ito One of the main aims of this paper is to give a proof of the above result, which we will do in Section 6. The explicit expression for R−1 in terms of its Gauss matrix decomposition is also presented as Corollary 6.1 in Section 6. Remark 1.5. After completing of earlier version of this paper, the author was informed that Theorem 1.4 previously appeared implicitly in [12, Section 5] and [26]. Our proof of the Theorem is, however, very different to that of [12] and [26]. From Theorem 1.4 we immediately obtain a closed-form expression for the determinant of R (or Ki). Corollary 1.6. The determinant of the transition matrix R evaluates as detR = dR 0 d R 1 · · · dR n = ( −a1a−12 )(n+1 2 ) n∏ i=1 (a2b1; t)i (a1b2; t)i . The determinants of the coefficient matrices K1 and K2 given in (1.13) evaluate as detK1 = det ( R−1D1 ) = ( −a2a−11 qαtn−1 )(n+1 2 ) n∏ i=1 (a1b2; t)i (a2b1; t)i and detK2 = det ( D2 ( T−1q,b2 Tq,a2R )) = ( −a1a−12 qα−1tn−1 )(n+1 2 ) n∏ i=1 (qa2b1; t)i( q−1a1b2; t ) i . Next, we focus on the q-difference system with respect to the shift α→ α+ 1 for the Jackson integral of symmetric Selberg type. Using Matsuo’s polynomials {ei(a1, b2; z) | i = 0, 1, . . . , n}, this q-difference system is given explicitly in terms of its Gauss matrix decomposition. Theorem 1.7. Let Tα be the shift operator with respect to α→ α+ 1, i.e., Tαf(α) = f(α+ 1) for an arbitrary function f(α) of α ∈ C. Then Tα ( 〈e0(a1, b2), ξ〉, 〈e1(a1, b2), ξ〉, . . . , 〈en(a1, b2), ξ〉 ) = ( 〈e0(a1, b2), ξ〉, 〈e1(a1, b2), ξ〉, . . . , 〈en(a1, b2), ξ〉 ) A, (1.17) where the coefficient matrix A is written in terms of its Gauss matrix decomposition as A = LADA UA = U ′AD ′ A L ′ A. Here LA = ( lAij ) 0≤i,j≤n, DA = ( dA j δij ) 0≤i,j≤n, UA = ( uA ij ) 0≤i,j≤n are the lower triangular, diag- onal, upper triangular matrices, respectively, given by lAij = (−1)i−jt( n−i 2 )−(n−j 2 ) [ n− j n− i ] t ( a2b2t j ; t ) i−j( qαa2b2t2j ; t ) i−j , (1.18a) dA j = an−j1 aj2t (j2)+(n−j 2 ) ( qα; t ) j ( qαa2b2t 2j ; t ) n−j( qαa2b2tj−1; t ) j ( qαa1a2b1b2tn+j−1; t ) n−j , (1.18b) uA ij = ( −qαa−11 a2 )j−i t( j 2)−(i 2) [ j i ] t ( a1b1t n−j ; t ) j−i( qαa2b2t2i; t ) j−i , (1.18c) q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 7 and U ′A = ( uA ′ ij ) 0≤i,j≤n, D′A = ( dA ′ j δij ) 0≤i,j≤n, L′A = ( lA ′ij ) 0≤i,j≤n are the upper triangular, diagonal, lower triangular matrices, respectively, given by uA ′ ij = ( −qα )j−i t( n−i 2 )−(n−j 2 ) [ j i ] t ( a1b1t n−j ; t ) j−i( qαa1b1t2(n−j); t ) j−i , (1.19a) dA ′ j = an−j1 aj2t (j2)+(n−j 2 ) ( qαa1b1t 2(n−j); t ) j ( qα; t ) n−j( qαa1a2b1b2t2n−j−1; t ) j ( qαa1b1tn−j−1; t ) n−j , (1.19b) lA ′ij = ( −a1a−12 )i−j t( j 2)−(i 2) [ n− j n− i ] t ( a2b2t j ; t ) i−j( qαa1b1t2(n−i); t ) i−j . (1.19c) The first part of Theorem 1.7 will be proved in Section 5, while the latter part of Theorem 1.7 will be explained in the Appendix. Note that, from this theorem we immediately have the following. Corollary 1.8. The determinant of the coefficient matrix A evaluates as detA = dA 0 d A 1 · · · dA n = (a1a2) (n+1 2 )t2( n+1 3 ) n∏ i=1 ( qα; t ) i( qαa1a2b1b2t2n−i−1; t ) i . Remark 1.9. The expression (1.17) for the q-difference equation is equivalent to Tα ( 〈e0(a1, b2), ξ〉, 〈e1(a1, b2), ξ〉, . . . , 〈en(a1, b2), ξ〉 ) U−1A = ( 〈e0(a1, b2), ξ〉, 〈e1(a1, b2), ξ〉, . . . , 〈en(a1, b2), ξ〉 ) LADA or Tα ( 〈e0(a1, b2), ξ〉, 〈e1(a1, b2), ξ〉, . . . , 〈en(a1, b2), ξ〉 ) L′−1A = ( 〈e0(a1, b2), ξ〉, 〈e1(a1, b2), ξ〉, . . . , 〈en(a1, b2), ξ〉 ) U ′AD ′ A. Here the entries of U−1A and L′−1A are also factorized into binomials like UA and L′A, respectively. We will see the explicit expression for U−1A in Section 4 as Proposition 4.4. For the explicit expression for L′A, see Proposition A.4. Remark 1.10. If we consider the q → 1 limit after replacing ai and bi as a1 = 1, a2 = x, b1 = qβ and b2 = qγx−1 on Theorem 1.7, then ei ( 1, qγx−1; z ) Φ(z)∆(z) dqz1 z1 · · · dqznzn tends to e′i(z)Ψ ′(z)dz1 · · · dzn, where Ψ′(z) = n∏ i=1 zα−1i (1− zi)β−1(1− zi/x)γ−1 ∏ 1≤j<k≤n (zj − zk)2τ , e′i(z) = A n−i∏ j=1 (1− zj/x) n∏ j=n−i+1 (1− zj)  . This confirms that Theorem 1.7 in the q → 1 limit is consistent with the result presented in Proposition 1.1. The paper is organized as follows. After defining some basic terminology in Section 2, we characterize in Section 3 Matsuo’s polynomials by their vanishing property (Proposition 3.1), and define a family of symmetric polynomials of higher degree, which includes Matsuo’s polyno- mials. We call such polynomials the interpolation polynomials, which are inspired from Aomoto’s method [2, Section 8], [3], which is a technique to obtain difference equations for the Selberg 8 M. Ito integrals (see also [18] for a q-analogue of Aomoto’s method). We state several vanishing prop- erties for the interpolation polynomials, which are used in subsequent sections. In Section 4 we present three-term relations (Lemma 4.1) among the interpolation polynomials. These are key equations for obtaining the coefficient matrix of the q-difference system with respect to the shift α → α + 1. By repeated use of these three-term relations we obtain a proof of Theorem 1.7. Section 5 is devoted to the proof of Lemma 4.1. In Section 6 we explain the Gauss decom- position of the transition matrix R. For this purpose, we introduce another set of symmetric polynomials called the Lagrange interpolation polynomials of type A in [19], which are different from Matsuo’s polynomials. Both upper and lower triangular matrices in the decomposition can be understood as a transition matrix between Matsuo’s polynomials and the other polynomials. In the Appendix we explain the proof of the latter part of Theorem 1.7. Finally we would like to make some remarks about the original motivation for the current paper. Although the author already knew the results of this paper before publishing [14], many years have passed since then. The author recently learned of an interesting application of the q- difference systems of this paper in collaboration with Yasuhiko Yamada. They intend to publish the detail in a forthcoming paper. 2 Notation Let Sn be the symmetric group on {1, 2, . . . , n}. For a function f : (C∗)n → C we define an action of the symmetric group Sn on f by (σf)(z) := f ( σ−1(z) ) = f(zσ(1), zσ(2), . . . , zσ(n)) for σ ∈ Sn. We say that a function f(z) on (C∗)n is symmetric or skew-symmetric if σf(z) = f(z) or σf(z) = (sgnσ)f(z) for all σ ∈ Sn, respectively. We denote by Af(z) the alternating sum over Sn defined by (Af)(z) := ∑ σ∈Sn (sgnσ)(σf)(z), (2.1) which is skew-symmetric. Let P be the set of partitions defined by P := { (λ1, λ2, . . . , λn) ∈ Zn |λ1 ≥ λ2 ≥ · · · ≥ λn ≥ 0 } . We define the lexicographic ordering < on P as follows. For λ, µ ∈ P , we denote λ < µ if there exists a positive integer k such that λi = µi for all i < k and λk < µk. For λ = (λ1, λ2, . . . , λn) ∈ Zn, we denote by zλ the monomial zλ11 zλ22 · · · zλnn . For λ ∈ P the monomial symmetric polynomials mλ(z) are defined by mλ(z) := ∑ µ∈Snλ zµ, where Snλ := {σλ |σ ∈ Sn} is the Sn-orbit of λ. For λ ∈ P , we denote by mi the multiplicity of i in λ, i.e., mi = #{j |λj = i}, see [27] for instance. It is convenient to use the notation λ = ( 1m12m2 · · · rmr · · · ) : for example, ( 1322 ) = (2, 2, 1, 1, 1, 0) and z(1 322) = z21z 2 2z3z4z5. 3 Interpolation polynomials In this section we define a family of symmetric functions which extends Matsuo’s polynomials. For a, b ∈ C∗ and z = (z1, . . . , zn) ∈ (C∗)n let Ek,i(a, b; z) (k, i = 0, 1, . . . , n) be functions q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 9 specified by Ek,i(a, b; z) := z1z2 · · · zk∆(t; z) n−i∏ j=1 (1− bzj) n∏ j=n−i+1 ( 1− a−1zj ) , (3.1) where ∆(t; z) := ∏ 1≤i<j≤n ( zi − t−1zj ) = t−(n2) ∏ 1≤i<j≤n (tzi − zj), and let Ẽk,i(a, b; z) (k, i = 0, 1, . . . , n) be the symmetric functions of z ∈ (C∗)n specified by Ẽk,i(a, b; z) := AEk,i(a, b; z) ∆(z) , (3.2) which, in particular, satisfy Ẽ0,i(a, b; z) = ei(a, b; z) and Ẽn,i(a, b; z) = z1z2 · · · znei(a, b; z), as special cases. We sometimes abbreviate Ẽk,i(a, b; z) to Ẽk,i(z). The leading term of the symmetric polynomial Ẽk,i(z) is m(1n−k2k)(z), i.e., Ẽk,i(z) = Ckim(1n−k2k)(z) + lower order terms, where the coefficient Cki of the monomial m(1n−k2k)(z) is expressed as Cki = (−1)n ( t−1; t−1 ) k ( t−1; t−1 ) n−k aib−(n−i) ( 1− t−1 )n . For arbitrary x, y ∈ C∗, we set ζj(x, y) := ( yt−(j−1), yt−(j−2), . . . , yt−1, y︸ ︷︷ ︸ j , x, xt, xt2, . . . , xtn−j−1︸ ︷︷ ︸ n−j ) ∈ (C∗)n. (3.3) The following gives another characterization of Matsuo’s polynomials ei(a, b; z) = Ẽ0,i(z). Proposition 3.1. The leading term of the function Ẽ0,i(z) is m(1n)(z) up to a multiplicative constant. The functions Ẽ0,i(z), i = 0, 1, . . . , n, satisfy Ẽ0,i ( ζj ( a, b−1 )) = ciδij , (3.4) where the constant ci is given by ci = ( abti; t ) n−i ( a−1b−1t−(i−1); t ) i (t; t)i(t; t)n−i t( n 2)(1− t)n (3.5a) = (ab; t)n−i ( a−1b−1t−(n−1); t ) i ( t−1; t−1 ) i ( t−1; t−1 ) n−i( 1− t−1 )n . (3.5b) Remark 3.2. The set of symmetric functions { Ẽ0,i(z) | i = 0, 1, . . . , n } forms a basis of the linear space spanned by {mλ(z) |λ ≤ (1n)}. Conversely such basis satisfying the condition (3.4) is uniquely determined. Thus we can take Proposition 3.1 as a definition of Matsuo’s polynomials, instead of (1.10). 10 M. Ito Proof. By definition Ẽ0,i(ζj(a, b −1)) = 0 if i 6= j. Ẽ0,i(ζi(a, b −1)) evaluates as Ẽ0,i(ζi(a, b −1)) = E0,i ( atn−i−1, . . . , at2, at, a, b−1, b−1t−1, . . . , b−1t−(i−1) ) ∆ ( atn−i−1, . . . , at2, at, a, b−1, b−1t−1, . . . , b−1t−(i−1) ) , which coincides with (3.5a) and (3.5b). � Lemma 3.3 (triangularity). Suppose ξj := ( b−1t−(j−1), b−1t−(j−2), . . . , b−1t−1, b−1︸ ︷︷ ︸ j , z1, z2, . . . , zn−j ) ∈ (C∗)n. Then Ẽk,i(ξj) = 0 if 0 ≤ i < j ≤ n. (3.6) Moreover, Ẽ0,i(ξi) evaluates as Ẽ0,i(ξi) = t−i(n−i) ( t−1; t−1 ) i ( t−1; t−1 ) n−i( 1− t−1 )n ( a−1b−1t−(i−1); t ) i n−i∏ j=1 ( 1− zjbti ) = (t; t)i(t; t)n−i t( n 2)(1− t)n ( a−1b−1t−(i−1); t ) i n−i∏ j=1 ( 1− zjbti ) . (3.7) On the other hand, if ηj := ( z1, z2, . . . , zj , a, at, at 2, . . . , atn−j−1︸ ︷︷ ︸ n−j ) ∈ (C∗)n, then Ẽk,i(ηj) = 0 if 0 ≤ j < i ≤ n. (3.8) Moreover, Ẽ0,i(ηi) evaluates as Ẽ0,i(ηi) = ( t−1; t−1 ) i ( t−1; t−1 ) n−i( 1− t−1 )n (ab; t)n−i i∏ j=1 ( 1− zj/atn−i ) . (3.9) Proof. By the definition (3.1) of Ẽk,i(z) it immediately follows that Ẽk,i(ξj) = 0 if i < j, and Ẽk,i(ηj) = 0 if j < i. If we put zj = b−1t−i (j = 1, 2, . . . , n− i) in the polynomial Ẽk,i(ξi), then we have Ẽk,i(ξi) = 0 because Ẽk,i(ξi) satisfies the condition of (3.6). This implies that Ẽk,i(ξi) is divisible by ∏n−i j=1(1− zjbti) up to a constant. Thus we write Ẽk,i(ξi) = c ∏n−i j=1(1− zjbti), where c is some constant independent of z1, . . . , zn−i. Next we determine the explicit form of c. If we put z1 = a, z2 = at, . . . , zn−i = atn−i−1 in Ẽk,i(ξi), then Ẽk,i(ξi) = Ẽ0,i ( ζi ( a, b−1 )) . From (3.4), we have ci = c(abti; t)n−i, where ci is given by (3.5b). Therefore the constant c is evaluated as c = ci/(abt i; t)n−i, i.e., we obtain the expression (3.7) for Ẽk,i(ξi). The evaluation (3.9) is carried out in the same way as above. � Lemma 3.4. For 0 ≤ j ≤ n, let ζj ( x, b−1 ) ∈ (C∗)n be the point specified by (3.3) with y = b−1. Then Ẽ0,i(ζj(x, b −1)) = ( xbti; t ) n−i ( xa−1; t ) i−j ( a−1b−1t−(j−1); t ) j (t; t)n t( n 2)(1− t)n [ i j ] t[ n j ] t . (3.10) Moreover, if i+ k ≤ n (i.e., k ≤ n− i ≤ n− j), then Ẽk,i ( ζj ( x, b−1 )) = xkt(n−j)k−(k+1 2 )Ẽ0,i ( ζj ( x, b−1 )) . (3.11) q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 11 Proof. If i ≤ j, Ẽk,i ( ζj ( x, b−1 )) = 0 is a special case of (3.6). Suppose j ≤ i. If we put x = b−1t−i−k (k = 0, 1, . . . , n−i−1), then the polynomial Ẽ0,i ( ζj ( x, b−1 )) satisfies the condition (3.6) of Lemma 3.3, so that it is equal to zero, which implies Ẽ0,i ( ζj ( x, b−1 )) is divisible by (xbti; t)n−i. If we put x = at−k (k = 0, 1, . . . , i − j − 1), then the polynomial Ẽ0,i ( ζj ( x, b−1 )) satisfies the condition (3.8) of Lemma 3.3, so that it is also equal to zero, which implies Ẽ0,i ( ζj ( x, b−1 )) is divisible by ( xa−1; t ) i−j . Therefore we have Ẽ0,i ( ζj ( x, b−1 )) = c ( xbti; t ) n−i ( xa−1; t ) i−j , (3.12) where c is some constant independent of x. Next we determine the explicit form of c. Put x = b−1t−(i−1) in (3.12). Then, using (3.7), the left-hand side of (3.12) is written as Ẽ0,i ( ζj ( x, b−1 ))∣∣∣ x=b−1t−(i−1) = Ẽ0,i ( ζi ( x, b−1 ))∣∣∣ x=b−1t−(j−1) = (t; t)i(t; t)n−i(t; t)n−j t( n 2)(1− t)n(t; t)i−j ( a−1b−1t−(i−1); t ) i , (3.13) while the right-hand side of (3.12) is c ( xbti; t ) n−i ( xa−1; t ) i−j ∣∣∣ x=b−1t−(i−1) = c(t; t)n−i ( a−1b−1t−(i−1); t ) i−j . (3.14) Comparing with (3.13) and (3.14), we have c = (t; t)i(t; t)n−j t( n 2)(1− t)n(t; t)i−j ( a−1b−1t−(j−1); t ) j . Therefore we obtain (3.10). Moreover, if i+ k ≤ n, by definition we have Ẽk,i ( ζj ( x, b−1 )) = ( xt(n−j−1) )( xt(n−j−2) ) · · · ( xt(n−j−k) ) Ẽ0,i ( ζj ( x, b−1 )) , which coincides with (3.11). � As a counterpart of Lemma 3.4, we have the following. Lemma 3.5. For 0 ≤ j ≤ n, let ζj(a, y) ∈ (C∗)n be the point specified by (3.3) with x = a. Then Ẽ0,i(ζj(a, y)) = ( ybt−(j−i−1); t ) j−i ( y/atn−1; t ) i (ab; t)n−j ( t−1; t−1 ) n( 1− t−1 )n [ n− i n− j ] t−1[ n j ] t−1 . (3.15) Moreover, if n ≤ i+ k (i.e., n− j ≤ n− i ≤ k), then Ẽk,i(ζj(a, y)) = yk+j−nan−jt( n−j 2 )−(k+j−n 2 )Ẽ0,i(ζj(a, y)). Proof. This lemma can be proved in the same way as Lemma 3.4 using Lemma 3.3. � 12 M. Ito 4 Three-term relations In this section we fix Ẽk,i(z) = Ẽk,i(a1, b2; z). In particular we have ei(a1, b2; z) = Ẽ0,i(a1, b2; z). The following lemma is a technical key for computing the coefficient matrix A of (1.17) in Theorem 1.7. We abbreviate 〈 Ẽk,i, x 〉 to 〈 Ẽk,i 〉 throughout this section. Lemma 4.1 (three-term relations). Suppose i+ k ≤ n. Then, 1− qαa1a2b1b2t2n−k−1 a1tn−i−k 〈 Ẽk,i 〉 = ( 1− qαa2b2tn+i−k )〈 Ẽk−1,i 〉 − ( 1− a2b2ti )〈 Ẽk−1,i+1 〉 .(4.1) On the other hand, if i+ k ≥ n, then, tn−k ( 1− qαtn−k )〈 Ẽk−1,i+1 〉 = a−11 qαtn+i−k ( 1− a1b1tn−i−1 )〈 Ẽk,i 〉 + a−12 ( 1− qαa2b2tn+i−k )〈 Ẽk,i+1 〉 . (4.2) The proof of this lemma will be given in Section 5. The rest of this section is devoted to computing the Gauss matrix decomposition of A in Theorem 1.7 using Lemma 4.1. By repeated use of Lemma 4.1, we have the following. Corollary 4.2. Suppose i+ k ≤ n. For 0 ≤ l ≤ k, 〈 Ẽk,i 〉 is expressed as 〈 Ẽk,i 〉 = l∑ j=0 Lk,ik−l,i+j 〈 Ẽk−l,i+j 〉 , (4.3) where the coefficients Lk,ik−l,i+j is expressed as Lk,ik−l,i+j = [ l j ] t (−1)j ( a1t n−i−k)lt(l−j 2 )(a2b2ti; t)j(qαa2b2tn+i+j−k; t)l−j( qαa1a2b1b2t2n−k−1; t ) l . On the other hand, if i+ k ≥ n, then, 〈 Ẽk,i 〉 = l∑ j=0 Uk,ik−l+j,i−j 〈 Ẽk−l+j,i−j 〉 , (4.4) where the coefficient Uk,ik−l+j,i−j is expressed as Uk,ik−l+j,i−j = [ l j ] t ( −qαa−11 ti−l )j( a2t n−k)lt(l 2) ( a1b1t n−i; t ) j ( qαtn−k; t ) l−j( qαa2b2tn+i−k−j−1; t ) l−j ( qαa2b2tn+i−k−2j+l; t ) j . Proof. (4.3) and (4.4) follow by double induction on l and j using (4.1) and (4.2), respective- ly. � In particular, we immediately have the following as a special case of Corollary 4.2. Corollary 4.3. For 0 ≤ j ≤ n, 〈En−j,j〉 is expressed as 〈 Ẽn−j,j 〉 = n∑ i=j l̃ij 〈 Ẽ0,i 〉 , (4.5) q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 13 where the coefficients l̃ij is expressed as l̃ij = Ln−j,j0,i = [ n− j n− i ] t (−1)i−jan−j1 t( n−i 2 )(a2b2tj ; t)i−j(qαa2b2ti+j ; t)n−i( qαa1a2b1b2tn+j−1; t ) n−j , (4.6) while, for 0 ≤ j ≤ n, 〈En,j〉 is expressed as 〈 Ẽn,j 〉 = j∑ i=0 ũij 〈 Ẽn−i,i 〉 , (4.7) where the coefficients ũij is expressed as ũij = Un,jn−i,i = [ j i ] t ( −qαa−11 )j−i aj2t (j2) ( qα; t ) i ( a1b1t n−j ; t ) j−i( qαa2b2ti−1; t ) i ( qαa2b2t2i; t ) j−i . (4.8) We now give the explicit expression for A in terms of Gauss decomposition. Proof of Theorem 1.7. From (4.5), we have(〈 Ẽn,0 〉 , 〈 Ẽn−1,1 〉 , . . . , 〈 Ẽ1,n−1 〉 , 〈 Ẽ0,n 〉) = (〈 Ẽ0,0 〉 , 〈 Ẽ0,1 〉 , . . . , 〈 Ẽ0,n−1 〉 , 〈 Ẽ0,n 〉) L̃, where the matrix L̃ = ( l̃ij ) 0≤i,j≤n is defined by (4.6). Moreover, from (4.7) we have(〈 Ẽn,0 〉 , 〈 Ẽn,1 〉 , . . . , 〈 Ẽn,n−1 〉 , 〈 Ẽn,n 〉) = (〈 Ẽn,0 〉 , 〈 Ẽn−1,1 〉 , . . . , 〈 Ẽ1,n−1 〉 , 〈 Ẽ0,n 〉) Ũ (4.9a) = (〈 Ẽ0,0 〉 , 〈 Ẽ0,1 〉 , . . . , 〈 Ẽ0,n−1 〉 , 〈 Ẽ0,n 〉) L̃Ũ , (4.9b) where the matrix Ũ = ( ũij ) 0≤i,j≤n is defined by (4.8). Since TαΦ(z) = z1z2 · · · znΦ(z) and z1z2 · · · znẼ0,i(z) = Ẽn,i(z), we have Tα 〈 Ẽ0,i 〉 = 〈 Ẽn,i 〉 , i.e., Tα (〈 Ẽ0,0 〉 , 〈 Ẽ0,1 〉 , . . . , 〈 Ẽ0,n−1 〉 , 〈 Ẽ0,n 〉) = (〈 Ẽn,0 〉 , 〈 Ẽn,1 〉 , . . . , 〈 Ẽn,n−1 〉 , 〈 Ẽn,n 〉) . (4.10) From (4.9b) and (4.10), we obtain the difference system Tα (〈 Ẽ0,0 〉 , 〈 Ẽ0,1 〉 , . . . , 〈 Ẽ0,n−1 〉 , 〈 Ẽ0,n 〉) = (〈 Ẽ0,0 〉 , 〈 Ẽ0,1 〉 , . . . , 〈 Ẽ0,n−1 〉 , 〈 Ẽ0,n 〉) L̃Ũ . Comparing this with (1.17), we therefore obtain A = L̃Ũ = LADAUA, i.e., lAij = l̃ij l̃jj , dA j = l̃jj ũjj , uA ij = ũij ũii . Corollary 4.2 implies that lAij , d A j and uA ij coincide with (1.18a), (1.18b) and (1.18c), respectively. The Gauss decomposition of A in the opposite direction, i.e., A = U ′AD ′ AL ′ A can also be given in the same way as above. This will be explained in the Appendix. � Finally we state the explicit forms for U−1A . Proposition 4.4. The inverse matrix U−1A = ( uA∗ ij ) 0≤i,j≤n is upper triangular, and is written as uA∗ ij = ( qαa−11 a2t j−1)j−i [j i ] t ( a1b1t n−j ; t ) j−i( qαa2b2tj+i−1; t ) j−i . (4.11) 14 M. Ito Proof. Since A = L̃Ũ = LADAUA, we have Ũ = DŨUA, where DŨ is the diagonal matrix defined by the diagonal elements of Ũ = ( ũij ) 0≤i,j≤n, i.e., DŨ = ( ũiiδij ) 0≤i,j≤n, where ũij is given by (4.8). We first compute the explicit expression for Ũ−1. From (4.9a), Ũ−1 is regarded as the transition matrix (〈En,0〉, 〈En−1,1〉, . . . , 〈E1,n−1〉, 〈E0,n〉) = (〈En,0〉, 〈En,1〉, . . . , 〈En,n−1〉, 〈En,n〉)Ũ−1, (4.12) namely, if we write Ũ−1 = ( ṽij ) 0≤i,j≤n, then (4.12) is equivalent to 〈 Ẽn−j,j 〉 = ∑j i=0 ṽij 〈 Ẽn,i 〉 . Similar to Corollary 4.2, by repeated use of the three-term relation (4.4) inductively, 〈Ẽk,i〉 is generally expressed as 〈 Ẽk,i 〉 = l∑ j=0 V k,i k+l,i−j 〈 Ẽk+l,i−j 〉 , where the coefficients V k,i k+l,i−j is given by V k,i k+l,i−j = [ l j ] t ( qαa−11 ti−1 )j t( l−j 2 )−(j2) ( a1b1t n−i; t ) j ( qαa2b2t n+i−k−l−1; t ) l−j( a2tn−k−1 )l−j( qαtn−k−l; t ) l . In particular, the entries ṽij of Ũ−1 are explicitly expressed as ṽij = V n−j,j n,i = [ j i ] t ( qαa−11 tj−1 )j−i t( i 2)−(j−i 2 )(a1b1tn−j ; t)j−i(qαa2b2tj−1; t)i( a2tj−1 )i( qα; t ) j . (4.13) Next we compute U−1A = ( uA∗ ij ) 0≤i,j≤n. Since U−1A is expressed as U−1A = Ũ−1DŨ , using (4.8) and (4.13), we obtain uA∗ ij = ṽij ũjj = [ j i ] t ( qαa−11 tj−1 )j−i t( i 2)−(j−i 2 )(a1b1tn−j ; t)j−i(qαa2b2tj−1; t)i( a2tj−1 )i( qα; t ) j aj2t (j2) ( qα; t ) j( qαa2b2tj−1; t ) j , which coincides with (4.11). � 5 Proof of Lemma 4.1 The aim of this section is to give a proof of Lemma 4.1. Throughout this section we fix Ẽk,i(z) = Ẽk,i(a1, b2; z). For Φ(z) = Φn,2, let ∇ be the operator specified by (∇ϕ)(z) := ϕ(z)− Tq,z1Φ(z) Φ(z) Tq,z1ϕ(z), where Tq,z1 is the q-shift operator with respect to z1 → qz1, i.e., Tq,z1f(z1, z2, . . . , zn) = f(qz1, z2, . . . , zn) for an arbitrary function f(z1, z2, . . . , zn). Here the ratio Tq,z1Φ(z)/Φ(z) is expressed explicitly as Tq,z1Φ(z) Φ(z) = qαt2(n−1)(1− b1z1)(1− b2z1)( 1− qa−11 z1 )( 1− qa−12 z1 ) n∏ j=2 z1 − t−1zj qz1 − tzj = G1(z) Tz1F1(z) , q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 15 where F1(z) = ( 1− a−11 z1 )( 1− a−12 z1 ) n∏ j=2 (z1 − tzj), G1(z) = qαt2(n−1)(1− b1z1)(1− b2z1) n∏ j=2 ( z1 − t−1zj ) . Lemma 5.1. Suppose that ∫ x∞ 0 Φ(z)ϕ(z)$q converges for a meromorphic function ϕ(z), then∫ x∞ 0 Φ(z)∇ϕ(z)$q = 0. Moreover,∫ x∞ 0 Φ(z)A∇ϕ(z)$q = 0. Proof. See Lemma 5.3 in [19]. � The rest of this section is devoted to the proof of Lemma 4.1. We show a further lemma before proving Lemma 4.1. For this purpose we abbreviate Ẽk,i(a1, b2; z) to Ẽk,i(z). When we need to specify the number of variables z1, . . . , zn, we use the notation Ẽ (n) k,i (z) = Ẽk,i(z) and ∆(n)(z) = ∆(z). We set ϕk,i(z) := F1(z)E (n−1) k−1,i (z2, . . . , zn). Then ∇ϕk,i(z) = (F1(z)−G1(z))E (n−1) k−1,i (z2, . . . , zn). Let ϕ̃k,i(z) be the skew-symmetrization of ∇ϕk,i(z), i.e., ϕ̃k,i(z) := A∇ϕk,i(z) = n∑ j=1 (−1)j−1(Fj(z)−Gj(z))Ẽ(n−1) k−1,i (ẑj)∆ (n−1)(ẑj), (5.1) where ẑj := (z1, . . . , zj−1, zj+1, . . . , zn) ∈ (C∗)n−1 for j = 1, . . . , n, and Fi(z) = ( 1− a−11 zi )( 1− a−12 zi ) ∏ 1≤k≤n k 6=i (zi − tzk), Gi(z) = qαt2(n−1)(1− b1zi)(1− b2zi) ∏ 1≤k≤n k 6=i ( zi − t−1zk ) , which satisfy the following vanishing property at the point z = ζj ( x, b−12 ) or z = ζj(a1, y). Lemma 5.2. If i 6= 0 and i 6= j, then Fi+1 ( ζj ( x, b−12 )) = 0. If i 6= n, then Gi ( ζj ( x, b−12 )) = 0. Otherwise, F1 ( ζj ( x, b−12 )) = ( 1− a−11 b−12 t−(j−1) )( 1− a−12 b−12 t−(j−1) )( b2tj−1 )n−1 (1− t) (t; t)j ( xb2t j ; t ) n−j , (5.2a) Fj+1 ( ζj ( x, b−12 )) = (−1)j ( 1− a−11 x )( 1− a−12 x ) xn−j−1 bj2t (j−1 2 )−1(1− t) (t, t)n−j ( xb2t −1; t ) j , (5.2b) Gn ( ζj ( x, b−12 )) = qαt2(n−1) ( 1− b1xtn−j−1 )( 1− b2xtn−j−1 ) 16 M. Ito × ( −b−12 )j t−(j+1 2 )(xbtn−j ; t) j ( xtn−j−1 )n−j−1 (t−1; t−1)n−j 1− t−1 , (5.2c) while, if i 6= 1, then Fi(ζj(a1, y)) = 0. If i 6= n and i 6= j, then Gi(ζj(a1, y)) = 0. Otherwise, F1(ζj(a1, y)) = ( 1− ya−12 t−(j−1) ) (−yt)j−1(−a1t)n−jt( n−j 2 )−(j−1 2 ) × ( ya−11 t−(n−1); t ) n−j+1 ( t−1; t−1 ) j 1− t−1 , (5.3a) Gj(ζj(a1, y)) = qαt2(n−1)(1− yb1)(1− yb2) × yj−1(−a1t−1)n−jt( n−j 2 )(ya−11 t−(n−j−2); t ) n−j ( t−1; t−1 ) j 1− t−1 , (5.3b) Gn(ζj(a1, y)) = qαt2(n−1) ( 1− a1b1tn−j−1 )( 1− a1b2tn−j−1 ) × ( a1t n−j−1)n−1(ya−11 t−(n−1); t ) j ( t−1; t−1 ) n−j 1− t−1 . (5.3c) Proof. The proof follows by direct computation and we omit the details. � Since the leading term of the symmetric polynomial ϕ̃k,i(z)/∆ (n)(z) is equal to m(1n−k2k)(z) up to a multiplicative constant, ϕ̃k,i(z)/∆ (n)(z) is expressed as the linear combination of the symmetric polynomials Ẽ (n) l,j (z) in the following two ways: ϕ̃k,i(z) ∆(n)(z) = k∑ l=0 n−l∑ j=0 cljẼ (n) l,j (z) = k∑ l=0 n∑ j=n−l dljẼ (n) l,j (z), (5.4) where clj and dlj are some coefficients. Lemma 5.3. Suppose i+ k ≤ n. Then, (5.4) is written as ϕ̃k,i(z) ∆(n)(z) = ck,iẼ (n) k,i (z) + ck−1,iẼ (n) k−1,i(z) + ck−1,i+1Ẽ (n) k−1,i+1(z), (5.5) where ck,i = −a−11 a−12 b−12 tk−1 ( 1− qαa1a2b1b2t2n−k−1 ) = qαb1t 2n−2 − a−11 a−12 b−12 tk−1, (5.6a) ck−1,i = a−12 b−12 tn−i−1 ( 1− qαa2b2tn+i−k ) = a−12 b−12 tn−i−1 − qαt2n−k−1, (5.6b) ck−1,i+1 = −a−12 b−12 tn−i−1 ( 1− a2b2ti ) = tn−1 ( 1− a−12 b−12 t−i ) . (5.6c) Suppose i+ k ≥ n. Then, (5.4) is written as ϕ̃k,i(z) ∆(n)(z) = dk,iẼ (n) k,i (z) + dk,i+1Ẽ (n) k,i+1(z) + dk−1,i+1Ẽ (n) k−1,i+1(z), (5.7) where dk,i = −a−11 qαtn+i−1 ( 1− a1b1tn−i−1 ) = qαb1t 2n−2 − qαa−11 tn+i−1, (5.8a) dk,i+1 = −a−12 tk−1 ( 1− qαa2b2tn+i−k ) = qαb2t n+i−1 − a−12 tk−1, (5.8b) dk−1,i+1 = tn−1 ( 1− qαtn−k ) = tn−1 − qαt2n−k−1. (5.8c) Remark 5.4. Given Lemma 5.3, Lemma 4.1 immediately follows by Lemma 5.1. Instead of proving Lemma 4.1 it thus suffices to prove Lemma 5.3. q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 17 Before proving Lemma 5.3 we show it holds for the following specific cases. Lemma 5.5. If i + k ≤ n, then the equation (5.5) holds for the points z = ζj ( x, b−12 ) (j = 0, 1, . . . , n), while if i + k ≥ n, then the equation (5.7) holds for the points z = ζj(a1, y) (j = 0, 1, . . . , n). Proof. Suppose i+ k ≤ n. If z = ζj ( x, b−12 ) , then the right-hand side of (5.5) with coefficients given by (5.6a)–(5.6c) can be written as ck,iẼ (n) k,i ( ζj ( x, b−12 )) + ck−1,iẼ (n) k−1,i ( ζj ( x, b−12 )) + ck−1,i+1Ẽ (n) k−1,i+1 ( ζj ( x, b−12 )) = ( ck,ixt n−j−k + ck−1,i ) Ẽ (n) k−1,i ( ζj ( x, b−12 )) + ck−1,i+1Ẽ (n) k−1,i+1 ( ζj ( x, b−12 )) = [( 1− a−11 xti−j ) a−12 b−12 tn−i−1 − qαt2n−k−1 ( 1− xb1tn−j−1 )] Ẽ (n) k−1,i ( ζj ( x, b−12 )) + tn−1 ( 1− a−12 b−12 t−i ) Ẽ (n) k−1,i+1 ( ζj ( x, b−12 )) = [( 1− xb2ti )( 1− ti−j+1 ) a2b2ti ( 1− ti+1 ) + ( 1− a−12 b−12 t−i )] tn−1Ẽ (n) k−1,i+1 ( ζj ( x, b−12 )) − qαt2n−k−1 ( 1− xb1tn−j−1 ) Ẽ (n) k−1,i ( ζj ( x, b−12 )) . (5.9) The final equality follows from the relation( 1− a−11 xti−j )( 1− ti+1 ) Ẽ (n) k−1,i ( ζj ( x, b−12 )) = ( 1− xb2ti )( 1− ti−j+1 ) Ẽ (n) k−1,i+1 ( ζj ( x, b−12 )) , which follows from (3.10) and (3.11). On the other hand, using (5.1) and (5.2a)–(5.2c), the left-hand side of (5.5) at z = ζj ( x, b−12 ) can be written as ϕ̃k,i ( ζj ( x, b−1 )) ∆(n) ( ζ (n) j ( x, b−1 )) = F1 ( ζ (n) j ( x, b−1 )) Ẽ (n−1) k−1,i ( ζ (n−1) j−1 ( x, b−1 ))∆(n−1)(ζ(n−1)j−1 ( x, b−1 )) ∆(n) ( ζ (n) j ( x, b−1 )) + (−1)jFj+1 ( ζ (n) j ( x, b−1 )) Ẽ (n−1) k−1,i ( ζ (n−1) j ( xt, b−1 ))∆(n−1)(ζ(n−1)j ( xt, b−1 )) ∆(n) ( ζ (n) j ( x, b−1 )) + (−1)nGn ( ζ (n) j ( x, b−1 )) Ẽ (n−1) k−1,i ( ζ (n−1) j ( x, b−1 ))∆(n−1)(ζ(n−1)j ( x, b−1 )) ∆(n) ( ζ (n) j ( x, b−1 )) . (5.10) Since we can compute Ẽ (n−1) k−1,i ( ζ (n−1) j−1 ( x, b−1 )) = tn−1(1− t)(1− xbti)Ẽ(n) k−1,i+1 ( ζ (n) j ( x, b−1 ))( 1− xbtn−1 )( 1− a−1b−1t−(j−1) )( 1− ti+1 ) , (5.11a) Ẽ (n−1) k−1,i ( ζ (n−1) j ( xt, b−1 )) = tn−1(1− t)(1− ti−j+1)Ẽ (n) k−1,i+1 ( ζ (n) j ( x, b−1 ))( 1− xa−1 )( 1− ti+1 )( 1− tn−j ) , (5.11b) Ẽ (n−1) k−1,i ( ζ (n−1) j ( x, b−1 )) = tn−k(1− t)Ẽ(n) k−1,i ( ζ (n) j ( x, b−1 ))( 1− xbtn−1 )( 1− tn−j ) , (5.11c) and ∆(n−1)(ζ(n−1)j−1 ( x, b−1 )) ∆(n) ( ζ (n) j ( x, b−1 )) = ( b2t j−1)n−1 (t; t)j−1 ( xbtj−1; t ) n−j , (5.11d) 18 M. Ito ∆(n−1)(ζ(n−1)j ( xt, b−1 )) ∆(n) ( ζ (n) j ( x, b−1 )) = x−(n−j−1)bj2t (j2) (t; t)n−j−1(xb; t)j (5.11e) ∆(n−1)(ζ(n−1)j ( xb−1 )) ∆(n) ( ζ (n) j ( x, b−1 )) = x−(n−j−1)bj2t (j2)−(n−j−1 2 ) (t; t)n−j−1 ( xbtn−j−1; t ) j , (5.11f) applying (5.11a)–(5.11f) to (5.10), the left-hand side of (5.5) at z = ζj ( x, b−12 ) can be expressed as ϕ̃k,i ( ζj ( x, b−12 )) ∆(n) ( ζ (n) j ( x, b−12 )) = [( 1− a−12 b−12 t−(j−1) )( 1− tj )( 1− xb2ti )( 1− xb2tj−1 )( 1− ti+1 ) + ( 1− a−12 x )( 1− xb2t−1 )( 1− ti−j+1 )( 1− xb2tj−1 )( 1− ti+1 ) tj ] tn−1Ẽ (n) k−1,i+1 ( ζ (n) j ( x, b−12 )) − qαt2n−k−1 ( 1− xb1tn−j−1 ) Ẽ (n) k−1,i ( ζ (n) j ( x, b−12 )) . (5.12) Comparing (5.9) with (5.12), the claim of the lemma is proved if we can check the identity( 1− xb2ti )( 1− ti−j+1 ) a2b2ti(1− ti+1) + ( 1− a−12 b−12 t−i ) = ( 1− a−12 b−12 t−(j−1) )( 1− tj )( 1− xb2ti )( 1− xb2tj−1 )( 1− ti+1 ) + ( 1− a−12 x )( 1− xb2t−1 )( 1− ti−j+1 )( 1− xb2tj−1 )( 1− ti+1 ) tj , which is confirmed by direct computation. Next suppose i + k ≥ n. If z = ζj(a1, y), then the right-hand side of (5.7) with coefficients given by (5.8a)–(5.8c) can be written as dk,iẼ (n) k,i (ζj(a1, y)) + dk,i+1Ẽ (n) k,i+1(ζj(a1, y)) + dk−1,i+1Ẽ (n) k−1,i+1(ζj(a1, y)) = dk,iẼ (n) k,i (ζj(a1, y)) + [ dk,i+1yt −(k+j−n−1) + dk−1,i+1 ] Ẽ (n) k−1,i+1(ζj(a1, y)) = −qαa−11 tn+i−1 ( 1− a1b1tn−i−1 ) Ẽ (n) k,i (ζj(a1, y)) + [ tn−1 ( 1− ya−12 t−(j−1) ) − qαt2n−k−1 ( 1− yb2t−(j−i−1) )] Ẽ (n) k−1,i+1(ζj(a1, y)) = −qαtn+i−1 [ a−11 ( 1− a1b1tn−i−1 ) + ( 1− ya−11 t−(n−i−1) )( 1− t−(j−i) ) yt−(j−i−1) ( 1− t−(n−i) ) ] Ẽ (n) k,i (ζj(a1, y)) + tn−1 ( 1− ya−12 t−(j−1) ) Ẽ (n) k−1,i+1(ζj(a1, y)) = −qαtn+j−2 [ 1− a1b1tn−i−1 a1tj−i−1 + ( 1− ya−11 t−(n−i−1) )( 1− t−(j−i) ) y ( 1− t−(n−i) ) ] Ẽ (n) k,i (ζj(a1, y)) + tn−1 ( 1− ya−12 t−(j−1) ) Ẽ (n) k−1,i+1(ζj(a1, y)). (5.13) On the other hand, using (5.1) and (5.3a)–(5.3c), the left-hand side of (5.7) at z = ζj(a1, y) can be written as ϕ̃k,i(ζj(a1, y)) ∆(n) ( ζ (n) j (a1, y) ) = F1(ζ (n) j (a1, y))Ẽ (n−1) k−1,i ( ζ (n−1) j−1 (a1, y) )∆(n−1)(ζ(n−1)j−1 (a1, y) ) ∆(n) ( ζ (n) j (a1, y) ) − (−1)j−1Gj ( ζ (n) j (a1, y) ) Ẽ (n−1) k−1,i ( ζ (n−1) j−1 ( a1, yt −1))∆(n−1)(ζ(n−1)j−1 ( a1, yt −1)) ∆(n) ( ζ (n) j (a1, y) ) q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 19 − (−1)n−1Gn ( ζ (n) j (a1, y) ) Ẽ (n−1) k−1,i ( ζ (n−1) j (a1, y) )∆(n−1)(ζ(n−1)j (a1, y) ) ∆(n) ( ζ (n) j (a1, y) ) . (5.14) Since we can compute Ẽ (n−1) k−1,i ( ζ (n−1) j−1 (a, y) ) = ( 1− t−1 ) Ẽ (n) k−1,i+1 ( ζ (n) j (a, y) )( 1− t−j )( 1− ya−1t−(n−1) ) , (5.15a) Ẽ (n−1) k−1,i ( ζ (n−1) j−1 ( a, yt−1 )) = ( 1− t−1 )( 1− t−(j−i) ) Ẽ (n) k,i ( ζ (n) j (a, y) )( 1− t−j )( 1− t−(n−i) ) (1− yb) , (5.15b) Ẽ (n−1) k−1,i ( ζ (n−1) j (a, y) ) = ( 1− t−1 )( 1− ya−1t−(n−i−1) ) Ẽ (n) k,i ( ζ (n) j (a, y) ) atn−j−1 ( 1− t−(n−i) )( 1− ya−1t−(n−1) )( 1− abtn−j−1 ) , (5.15c) and ∆(n−1)(ζ(n−1)j−1 (a1, y) ) ∆(n) ( ζ (n) j (a1, y) ) = (−1)n−j yj−1an−j1 t( n−j 2 )−(j−1 2 )(ya−11 t−(n−2); t ) n−j ( t−1; t−1 ) j−1 , (5.15d) ∆(n−1)(ζ(n−1)j−1 (a1, yt −1) ) ∆(n) ( ζ (n) j (a1, y) ) = (−1)n−1 yj−1an−j1 t( n−j 2 )(ya−11 t−(n−j−1); t ) n−j ( t−1; t−1 ) j−1 , (5.15e) ∆(n−1)(ζ(n−1)j (a1, y) ) ∆(n) ( ζ (n) j (a1, y) ) = (−1)n−1( a1tn−j−1 )n−1( ya−11 t−(n−2); t ) j ( t−1; t−1 ) n−j−1 , (5.15f) applying (5.15a)–(5.15f) to (5.14), the left-hand side of (5.7) at z = ζj(a, y) can be expressed as ϕ̃k,i(ζj(a1, y)) ∆(n) ( ζ (n) j (a1, y) ) = tn−1 ( 1− ya−12 t−(j−1) ) Ẽ (n) k−1,i+1(ζj(a1, y)) − qαtn+j−2 [ (1− yb1) ( 1− ya−1t )( 1− t−(j−i) ) y ( 1− ya−1t−(n−j−1) )( 1− t−(n−i) ) + ( 1− a1b1tn−j−1 ) t(1− ya−1t−(n−i−1))(1− t−(n−j)) a ( 1− ya−1t−(n−j−1) )( 1− t−(n−i) )]Ẽ(n) k,i (ζj(a1, y)). (5.16) Comparing with (5.13) and (5.16), the claim of the lemma is proved if we can check the identity 1− a1b1tn−i−1 a1tj−i−1 + ( 1− ya−11 t−(n−i−1) )( 1− t−(j−i) ) y ( 1− t−(n−i) ) = (1− yb1) ( 1− ya−1t )( 1− t−(j−i) ) y ( 1− ya−1t−(n−j−1) )( 1− t−(n−i) ) + ( 1− a1b1tn−j−1 ) t(1− ya−1t−(n−i−1))(1− t−(n−j)) a ( 1− ya−1t−(n−j−1) )( 1− t−(n−i) ) , which follows from direct computation. � Proof of Lemma 5.3. Set Dj = { (l, i) ∈ Z2 | j ≤ i, 0 ≤ l, i + l ≤ n } , which satisfies D0 ⊃ D1 ⊃ · · · ⊃ Dn = {(0, n)}. The set { Ẽk,i(z) | (k, i) ∈ D0 } forms a basis for the linear space spanned by {mλ(z) |λ ≤ (2n)}. If we put ψ(z) := ϕ̃k,i(z) ∆(z) − ( ck,iẼk,i(z) + ck−1,iẼk−1,i(z) + ck−1,i+1Ẽk−1,i+1(z) ) , 20 M. Ito where ck,i, ck−1,i and ck−1,i+1 are specified by (5.6a)–(5.6c), then the symmetric polynomial ψ(z) is expressed as a linear combination of Ẽk,i(z), (k, i) ∈ D0, i.e., ψ(z) = ∑ (l,m)∈D0 c′lmẼl,m(z), (5.17) where the coefficients c′lm are some constants. We now prove ψ(z) = 0 identically, i.e., c′lm = 0 for all (l,m) ∈ D0 inductively. Namely, we prove that, if c′lm = 0 for (l,m) ∈ Dj+1, then c′lm = 0 for (l,m) ∈ Dj . First we show that c′0n = 0 as the starting point of induction. Using Lemma 3.4 for (5.17) at z = ζn ( x, b−12 ) we have ψ ( ζn ( x, b−12 )) = c′0nẼ0,n ( ζn ( x, b−12 )) . From Lemma 5.5 we have ψ ( ζn(x, b−12 )) = 0, while Ẽ0,n ( ζn ( x, b−12 )) 6= 0. Therefore c′0n = 0. Next suppose that c′lm = 0 for (l,m) ∈ Dj+1. Then using Lemma 3.4 for (5.17) at z = ζj ( x, b−12 ) we have ψ ( ζj ( x, b−12 )) = n−j∑ l=0 c′ljẼl,j ( ζj ( x, b−12 )) = ( n−j∑ l=0 c′ljx ltl(n−j)−(l+1 2 ) ) Ẽ0,j ( ζj ( x, b−12 )) . From Lemma 5.5 ψ ( ζj ( x, b−12 )) vanishes as a function of x, while Ẽ0,j ( ζj ( x, b−12 )) 6= 0. Thus,∑n−j l=0 c ′ ljx ltl(n−j)−(l+1 2 ) = 0, i.e., the coefficient c′ljt l(n−j)−(l+1 2 ) of xl vanishes for 0 ≤ l ≤ n − j. Therefore c′lj = 0 for 0 ≤ l ≤ n− j. This implies c′lm = 0 for (l,m) ∈ Dj . On the other hand, we prove (5.7) of Lemma 5.3. Set D′j = { (l, i) ∈ Z2 |n ≤ i+ l, 0 ≤ l ≤ n, 0 ≤ i ≤ j } , which satisfies {(n, 0)} = D′0 ⊂ D′1 ⊂ · · · ⊂ D′n. The set {Ẽk,i(z) | (k, i) ∈ D′n} also forms a basis for the linear space spanned by {mλ(z) |λ ≤ (2n)}. If we put ψ′(z) := ϕ̃k,i(z) ∆(z) − ( dk,iẼk,i(z) + dk,i+1Ẽk,i+1(z) + dk−1,i+1Ẽk−1,i+1(z) ) , where dk,i, dk,i+1 and dk−1,i+1 are specified by (5.8a)–(5.8c), then the symmetric polynomial ψ′(z) is expressed as a linear combination of Ẽk,i(z), (k, i) ∈ D′n, i.e., ψ′(z) = ∑ (l,m)∈D′n d′lmẼl,m(z), (5.18) where the coefficients d′lm are some constants. We now prove ψ′(z) = 0 identically, i.e., d′lm = 0 for all (l,m) ∈ D′n inductively. Namely, we prove that, if d′lm = 0 for (l,m) ∈ D′j−1, then d′lm = 0 for (l,m) ∈ D′j . First we show that d′n0 = 0 as the starting point of induction. Using Lemma 3.5 for (5.18) at z = ζ0((a1, y)) we have ψ′(ζ0(a1, y)) = d′n0Ẽn,0(ζ0(a1, y)). From Lemma 5.5 we have ψ′(ζ0(a1, y)) = 0, while Ẽn,0(ζ0(a1, y)) 6= 0. Therefore d′n0 = 0. Next suppose that d′lm = 0 for (l,m) ∈ D′j−1. Then using Lemma 3.5 for (5.18) at z = ζj(a1, y) we have ψ′(ζj(a1, y)) = n∑ l=n−j d′ljẼl,j(ζj(a1, y)) =  n∑ l=n−j d′ljy l+j−nan−j1 t( n−j 2 )−(l+j−n 2 )  Ẽ0,j(ζj(a1, y)). From Lemma 5.5 ψ′(ζj(a1, y)) vanishes as a function of y, while Ẽ0,j(ζj(a1, y)) 6= 0. Thus,∑n l=n−j d ′ ljy l+j−nan−j1 t( n−j 2 )−(l+j−n 2 ) = 0, i.e., the coefficient d′lja n−j 1 t( n−j 2 )−(l+j−n 2 ) of yl+j−n va- nishes for n − j ≤ l ≤ n. Therefore d′lj = 0 for n − j ≤ l ≤ n. This implies d′lm = 0 for (l,m) ∈ D′j . � q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 21 6 The transition matrix R In this section we give a proof of Theorem 1.4. Before proving Theorem 1.4, we will show the results deduced from Theorem 1.4. By the definition (1.12) of the transition matrix R, we have R−1 = JR̄J, where the symbol R̄ is the matrix R after the interchange (a1, b1)↔ (a2, b2) and J is the matrix specified by J =  1 1 · · · 1  . The explicit form of the inverse matrix of R is given by Corollary 6.1. The inverse matrix R−1 is written as Gauss matrix decomposition R−1 = U−1R D−1R L−1R = L′−1R D′−1R U ′−1R , where the inverse matrices L−1R = ( lR∗ij ) 0≤i,j≤n, D−1R = ( dR∗ j δij ) 0≤i,j≤n, U−1R = ( uR∗ ij ) 0≤i,j≤n are lower triangular, diagonal, upper triangular, respectively, given by lR∗ij = uR n−i,n−j = [ n− j n− i ] t−1 ( a2b2t j ; t ) i−j( a2a −1 1 ti+j−n; t ) i−j , (6.1a) dR∗ j = dR n−j = ( a2a −1 1 t−(n−j); t ) j (a1b2; t)n−j (a2b1; t)j ( a−12 a1t−j ; t ) n−j , (6.1b) uR∗ ij = lRn−i,n−j = (−1)j−it−(j−i 2 ) [ j i ] t−1 ( a1b1t n−j ; t ) j−i( a−12 a1tn−2j+1; t ) j−i , (6.1c) and the inverse matrices U ′−1R = ( uR ′∗ ij ) 0≤i,j≤n, D′−1R = ( dR ′∗ j δij ) 0≤i,j≤n, L′−1R = ( lR ′∗ij ) 0≤i,j≤n are upper triangular, diagonal, lower triangular, respectively, given by uR ′∗ ij = lR ′n−i,n−j = [ j i ] t ( a−11 b−11 t−(n−i−1); t ) j−i( b2b −1 1 t−(n−2i−1); t ) j−i , (6.2a) dR ′∗ j = dR ′ n−j = ( b2b −1 1 t−(n−2j−1); t ) n−j ( a−11 b−12 t−(j−1); t ) j( a−12 b−11 t−(n−j−1); t ) n−j ( b−12 b1tn−2j+1; t ) j , (6.2b) lR ′∗ij = uR ′ n−i,n−j = (−1)i−jt( i−j 2 ) [ n− j n− i ] t ( a−12 b−12 t−(i−1); t ) i−j( b−12 b1tn−i−j ; t ) i−j . (6.2c) Proof. Since R−1 = JR̄J , we have R−1 = U−1R D−1R L−1R , where L−1R = JURJ , D−1R = JDRJ and U−1R = JLRJ . Thus we immediately have the expressions lR∗ij = uR n−i,n−j , d R∗ j = dR n−j and uR∗ ij = lRn−i,n−j . From Theorem 1.4 this gives the explicit forms (6.1a), (6.1b) and (6.1c). On the other hand, we also have R−1 = L′−1R D′−1R U ′−1R , where U ′−1R = JL′RJ , D′−1R = JD′RJ and L′−1R = JU ′RJ . Therefore uR ′∗ ij = lR ′n−i,n−j , d R ′∗ j = dR ′ n−j , and lR ′∗ij = uR ′ n−i,n−j . Thus we obtain the expressions (6.2a), (6.2b) and (6.2c). � 22 M. Ito The rest of this section is devoted to the proof of Theorem 1.4. For this purpose we introduce another set of symmetric polynomials different from Matsuo’s polynomials. For 0 ≤ r ≤ n, let fr(a1, a2; t; z) be (symmetric) polynomials specified by fr(a1, a2; t; z) := ∑ I⊆{1,...,n} |I|=r r∏ k=1 zik − a2tik−k a1tk−1 − a2tik−k n−r∏ l=1 zjl − a1tjl−l a2tl−1 − a1tjl−l , (6.3) where the summation is over all r-subsets I of {1, . . . , n}, and I = {i1 < · · · < ir}, J = {1, . . . , n}\I = {j1 < · · · < jn−r}. In particular, f0(a1, a2; t; z) = n∏ i=1 zi − a1 a2ti−1 − a1 , fn(a1, a2; t; z) = n∏ i=1 zi − a2 a1ti−1 − a2 . We remark that the polynomials fr(a1, a2; t; z) are called the Lagrange interpolation polynomials of type A and their properties are discussed in [19, Appendix B]. By definition the polynomial fi(a1, a2; t; z) satisfies fi(a1, a2; t; z) = fn−i(a2, a1; t; z). (6.4) When we need to specify the number of variables z1, . . . , zn, we use the notation f (n) i (a1, a2; t; z) = fi(a1, a2; t; z). Lemma 6.2 (recurrence relation). The polynomials (6.3) satisfy the following recurrence rela- tions: f (n) i (a1, a2; t; z) = zn − a2tn−i a1ti−1 − a2tn−i f (n−1) i−1 (a1, a2; t; ẑn) + zn − a1ti a2tn−i−1 − a1ti f (n−1) i (a1, a2; t; ẑn) for i = 0, 1, . . . , n, where ẑn = (z1, . . . , zn−1) ∈ (C∗)n−1. Proof. The lemma follows from a direct computation and we omit the detail. � For arbitrary x, y ∈ C∗ we define ξj(x, y; t) := ( x, xt, . . . , xtj−1︸ ︷︷ ︸ j , y, yt, . . . , ytn−j−1︸ ︷︷ ︸ n−j ) ∈ (C∗)n (6.5) for j = 0, 1, . . . , n. Proposition 6.3. The polynomial fi(a1, a2; t; z) is symmetric in the variables z = (z1, . . . , zn). The leading term of fi(a1, a2; t; z) is m(1n)(z) up to a multiplicative constant. The functions fi(a1, a2; t; z) (i = 0, 1, . . . , n) satisfy fi(a1, a2; t; ξj(a1, a2; t)) = δij . (6.6) Proof. See [19, Example 4.3 and equation (4.7)]. Otherwise, using Lemma 6.2 we can also prove this proposition directly by induction on n. � Remark 6.4. The set of symmetric polynomials {fi(a1, a2; t; z) | i = 0, 1, . . . , n} forms a basis of the linear space spanned by {mλ(z) |λ ≤ (1n)}. Conversely such basis satisfying the con- dition (6.6) is uniquely determined. Thus we can take Proposition 6.3 as a definition of the polynomials fi(a1, a2; t; z), instead of (6.3). q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 23 Lemma 6.5 (triangularity). Suppose that ξj(a1) := ( a1, a1t, . . . , a1t j−1︸ ︷︷ ︸ j , z1, z2, . . . , zn−j ) ∈ (C∗)n. If i < j, then fi(a1, a2; t; ξj(a1)) = 0. (6.7) Moreover, fi(a1, a2; t; ξi(a1)) evaluates as fi(a1, a2; t; ξi(a1)) = n−i∏ l=1 zl − a1ti a2tl−1 − a1ti = ∏n−i l=1 ( 1− zla−11 t−i )( a2a −1 1 t−i; t ) n−i . (6.8) On the other hand, suppose that ηj(a2) := ( z1, z2, . . . , zj , a2, a2t, . . . , a2t n−j−1︸ ︷︷ ︸ n−j ) ∈ (C∗)n. If i > j, then fi(a1, a2; t; ηj(a2)) = 0. (6.9) Moreover, fi(a1, a2; t; ηi(a2)) evaluates as fi(a1, a2; t; ηi(a2)) = i∏ l=1 zl − a2tn−i a1tl−1 − a2tn−i = ∏i l=1 ( 1− zla−12 t−(n−i) )( a1a −1 2 t−(n−i); t ) i . (6.10) Proof. First we show (6.9) by induction on n. For simplicity we write ηi(a2) as ηi. Suppose i > j. Using Lemma 6.2 we have f (n) i (a1, a2; t; ηj) = a2t n−j−1 − a2tn−i a1ti−1 − a2tn−i f (n−1) i−1 ( a1, a2; t; η (n−1) j ) + a2t n−j−1 − a1ti a2tn−i−1 − a1ti f (n−1) i ( a1, a2; t; η (n−1) j ) , where η (n−1) j = ( z1, z2, . . . , zj , a2, a2t, . . . , a2t n−j−2) ∈ (C∗)n−1. Since f (n−1) i ( a1, a2; t; η (n−1) i ) = 0 by the induction hypothesis, we have f (n) i (a1, a2; t; ηj) = a2t n−j−1 − a2tn−i a1ti−1 − a2tn−i f (n−1) i−1 ( a1, a2; t; η (n−1) j ) . If i−1 > j, then f (n−1) i−1 ( a1, a2; t; η (n−1) j ) = 0 by the induction hypothesis, while if i−1 = j, then a2t n−j−1 − a2tn−i = 0. In any case we obtain f (n) i (a1, a2; t; ηi) = 0, which is the claim of (6.9). Next we show (6.10). If we put zl = a2t n−i for l ∈ {1, . . . , i} in the polynomial fi(a1, a2; t; ηi(a2)) of z1, . . . , zi, then we have fi(a1, a2; t; ηi(a2)) = 0 because fi(a1, a2; t; ηi(a2))|zk=a2tn−i satisfies the condition of (6.9). This implies fi(a1, a2; t; ηi(a2)) is divisible by ∏ l=1 ( zl−a2tn−i ) , so that we have fi(a1, a2; t; ηi(a2)) = c ∏i l=1 ( zl − a2tn−i ) , where c is some constant. Thus fi(a1, a2; t; ηi(a2)) ∣∣∣ (z1,...,zi)=(a1,a1t,...,a1ti−1) = c i∏ l=1 ( a1t l−1 − a2tn−i ) . 24 M. Ito On the other hand, (6.6) implies that fi(a1, a2; t; ηi(a2)) ∣∣∣ (z1,...,zi)=(a1,a1t,...,a1ti−1) = fi(a1, a2; t; ξi(a1, a2; t)) = 1. We therefore obtain c = 1/ ∏i l=1 ( a1t l−1 − a2tn−i ) , which implies (6.10). Finally we show (6.7) and (6.8). From (6.4) we have fi(a1, a2; t; ξj(a1)) = fn−i(a2, a1; t; ξj(a1)) = fn−i(a2, a1; t; ηn−j(a1)). If i < j (i.e., n − i > n − j), then using (6.9) we see that the right-hand side of the above is equal to zero. Moreover, using (6.10) we obtain fi(a1, a2; t; ξi(a1)) = fn−i(a2, a1; t; ηn−i(a1)) = n−i∏ l=1 zl − a1ti a2tl−1 − a1ti = ∏n−i l=1 ( 1− zla−11 t−i )( a2a −1 1 t−i; t ) n−i , which completes the proof. � Corollary 6.6. Let ξj(x, a2; t) ∈ (C∗)n be the point specified by (6.5) with y = a2. Then fi(a1, a2; t; ξj(x, a2; t)) evaluates as fi(a1, a2; t; ξj(x, a2; t)) = [ j i ] t−1 ( xa−11 ; t ) j−i ( xa−12 t−(n−j); t ) i( a−11 a2tn−j−i; t ) j−i ( a1a −1 2 t−(n−i); t ) i . (6.11) Proof. If i > j, then fi(a1, a2; t; ξj(x, a2; t)) = 0 is a special case of (6.9) in Lemma 6.5. Now we assume that i ≤ j. If we put x = a2t n−j−k (k = 0, 1, . . . , i − 1), then from (6.9) we have fi(a1, a2; t; ξj(x, a2; t)) = 0. If we put x = a1t −k (k = 0, 1, . . . , j − i− 1), then from (6.7) we also have fi(a1, a2; t; ξj(x, a2; t)) = 0. This implies that fi(a1, a2; t; ξj(x, a2; t)) as a polynomial of x is divisible by (xa1)j−i ( xa−12 t−(n−j) ) i . Since the degree of fi(a1, a2; t; ξj(x, a2; t)) as a function of x is equal to j, the function fi(a1, a2; t; ξj(x, a2; t)) can be expressed as fi(a1, a2; t; ξj(x, a2; t)) = c ( xa−11 ; t ) j−i ( xa−12 t−(n−j); t ) i , where c is some constant. In order to determine the constant c, we put x = a2t n−j−i in the above equation. Then fi(a1, a2; t; ξj(x, a2; t)) ∣∣∣ x=a2tn−j−i = c ( a−11 a2t n−j−i; t ) j−i ( t−i; t ) i , while, from (6.10) we have fi(a1, a2; t; ξj(x, a2; t)) ∣∣∣ x=a2tn−j−i = fi(a1, a2; t; ξi(x, a2; t)) ∣∣∣ x=a2tn−j−i = ( t−j ; t ) i( a1a −1 2 t−(n−i); t ) i . The constant c can be explicitly computed as c = ( t−j ; t ) i( a−11 a2tn−j−i; t ) j−i ( a1a −1 2 t−(n−i); t ) i ( t−i; t ) i = ( t−1; t−1 ) j( a−11 a2tn−j−i; t ) j−i ( a1a −1 2 t−(n−i); t ) i ( t−1; t−1 ) j−i ( t−1; t−1 ) i . We therefore obtain (6.11). � q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 25 Lemma 6.7. Suppose that ŨR is the (n+ 1)× (n+ 1) matrix satisfying( en(a2, b1; z), en−1(a2, b1; z), . . . , e0(a2, b1; z) ) = ( fn(a1, a2; t; z), fn−1(a1, a2; t; z), . . . , f0(a1, a2; t; z) ) ŨR. (6.12) Then ŨR = ( ũR ij ) 0≤i,j≤n is an upper triangular matrix with entries given by ũR ij = ( a1b1t n−j ; t ) j−i ( a1a −1 2 t−i; t ) n−j(a2b1; t)i ( t−1; t−1 ) n( 1− t−1 )n [ j i ] t−1[ n i ] t−1 . (6.13) Suppose that L̃R is the (n+ 1)× (n+ 1) matrix satisfying( fn(a1, a2; t; z), fn−1(a1, a2; t; z), . . . , f0(a1, a2; t; z) ) = ( e0(a1, b2; z), e1(a1, b2; z), . . . , en(a1, b2; z) ) L̃R. (6.14) Then L̃R = ( l̃Rij ) 0≤i,j≤n is a lower triangular matrix with entries given by l̃Rij = (−1)i−jt−(i−j 2 )(a2b2tj ; t)i−j(1− t−1)n( a−11 a2t−(n−2j−1); t ) i−j(a1b2; t)n−j ( a−11 a2t−(n−j); t ) j ( t−1; t−1 ) n [ n i ] t−1 [ i j ] t−1 . (6.15) Proof. Since both {ei(a2, b1; z) | i = 0, 1, . . . , n} and {fi(a1, a2; t; z) | i = 0, 1, . . . , n} form bases of the linear space spanned by {mλ(z) |λ < (1n)}, the polynomial ei(a2, b1; z) is expressed as a linear combination of fi(a1, a2; t; z) (i = 0, 1, . . . , n), i.e., en−j(a2, b1; z) = n∑ i=0 ũR ijfn−i(a1, a2; t; z), where ũR ij are some constants. From the vanishing property (6.6), the coefficient ũR ij is given by ũR ij = en−j(a2, b1; ξn−i(a1, a2; t)) = en−j(a2, b1; ζn−i(a2, y)) ∣∣∣ y=a1tn−i−1 . (6.16) From (3.15) in Lemma 3.5, en−j(a2, b1; ζn−i(a2, y)) evaluates as en−j(a2, b1; ζn−i(a2, y)) = ( yb1t −(j−i−1); t ) j−i ( ya−12 t−(n−1); t ) n−j(a2b1; t)i ( t−1; t−1 ) n−i ( t−1; t−1 ) j( t−1; t−1 ) j−i ( 1− t−1 )n . Combining this and (6.16), we obtain the expression (6.13). Since {ei(a1, b2; z) | i = 0, 1, . . . , n} is also a basis of the linear space spanned by {mλ(z) |λ < (1n)}, the polynomial fi(a1, a2; t; z) is expressed as a linear combination of ei(a1, b2; z) (i = 0, 1, . . . , n), i.e., fn−j(a1, a2; t; z) = n∑ i=0 l̃Rijei(a1, b2; z), where l̃Rij are some constants. From (3.4), the coefficient l̃Rij is written as l̃Rij = fn−j ( a1, a2; t; ζi ( a1, b −1 2 )) ci = c−1i fn−j(a1, a2; t; ξn−i(a1, x)) ∣∣∣ x=b−1 2 t−(i−1) , (6.17) 26 M. Ito where ci is given explicitly in (3.5b) as ci = (a1b2; t)n−i ( a−11 b−12 t−(n−1); t ) i ( t−1; t−1 ) i ( t−1; t−1 ) n−i( 1− t−1 )n . (6.18) Using (6.11) in Corollary 6.6, we have fn−j(a1, a2; t; ξn−i(a1, x)) = fj(a2, a1; t; ξn−i(a1, x)) = fj(a2, a1; t; ξi(x, a1)) = [ i j ] t−1 ( xa−12 ; t ) i−j ( xa−11 t−(n−i); t ) j( a−12 a1tn−i−j ; t ) i−j ( a2a −1 1 t−(n−j); t ) j . Combining this, (6.17) and (6.18), we therefore obtain the expression (6.15). � Lemma 6.8. Suppose that L̃′R is the (n+ 1)× (n+ 1) matrix satisfying( en(a2, b1; z), en−1(a2, b1; z), . . . , e0(a2, b1; z) ) = ( fn ( b−11 , b−12 ; t−1; z ) , fn−1 ( b−11 , b−12 ; t−1; z ) , . . . , f0 ( b−11 , b−12 ; t−1; z )) L̃′R. (6.19) Then L̃′R = ( l̃R ′ij ) 0≤i,j≤n is a lower triangular matrix with entries given by l̃R ′ij = ( a−12 b−12 t−(i−1); t ) i−j ( a−12 b−11 t−(n−i−1); t ) n−i ( b1b −1 2 tn−i−j+1; t ) j (t; t)n t( n 2)(1− t)n [ i j ] t[ n j ] t . (6.20) Suppose that Ũ ′R is the (n+ 1)× (n+ 1) matrix satisfying( fn ( b−11 , b−12 ; t−1; z ) , fn−1 ( b−11 , b−12 ; t−1; z ) , . . . , f0 ( b−11 , b−12 ; t−1; z )) = ( e0(a1, b2; z), e1(a1, b2; z), . . . , en(a1, b2; z) ) Ũ ′R, (6.21) then Ũ ′R = ( ũR ′ ij ) 0≤i,j≤n is an upper triangular matrix with entries given by ũR ′ ij = (−1)j−it( j−i 2 )+(n2) ( a−11 b−11 t−(n−i−1); t ) j−i(1− t) n( b−11 b2t−(n−i−j); t ) j−i ( b−11 b2t−(n−2j−1); t ) n−j ( a−11 b−12 t−(j−1); t ) j (t; t)n × [ n j ] t [ j i ] t . (6.22) Proof. Since both {ei(a2, b1; z) | i = 0, 1, . . . , n} and { fi ( b−11 , b−12 ; t−1; z ) | i = 0, 1, . . . , n } form bases of the linear space spanned by {mλ(z) |λ < (1n)}, the polynomial ei(a2, b1; z) is expressed as a linear combination of fi ( b−11 , b−12 ; t−1; z ) (i = 0, 1, . . . , n), i.e., en−j(a2, b1; z) = n∑ i=0 l̃R ′ij fn−i ( b−11 , b−12 ; t−1; z ) , where l̃R ′ij are some constants. From (6.6) we have l̃R ′ij = en−j ( a2, b1; ξn−i ( b−11 , b−12 ; t−1 )) = en−j ( a2, b1; ζn−i ( x, b−11 ))∣∣∣ x=b−1 2 t−(i−1) . (6.23) Using (3.10) in Lemma 3.4 we have en−j ( a2, b1; ζn−i ( x, b−11 )) = ( xb1t n−j ; t ) j ( xa−12 ; t ) i−j ( a−12 b−11 t−(n−i−1); t ) n−i q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 27 × (t; t)n−j(t; t)i t( n 2)(1− t)n(t; t)i−j . Combining this and (6.23), we obtain (6.20). On the other hand, since {ei(a1, b2; z) | i = 0, 1, . . . , n} is also a basis of the linear space spanned by {mλ(z) |λ < (1n)}, the polynomial fi ( b−11 , b−12 ; t−1; z ) is expressed as a linear com- bination of ei(a1, b2; z) (i = 0, 1, . . . , n), i.e., fn−j ( b−11 , b−12 ; t−1; z ) = n∑ i=0 ũR ′ ij ei(a1, b2; z), where ũR ′ ij are some constants. From (3.4) we have ũR ′ ij = fn−j ( b−11 , b−12 ; t−1; ζi ( a1, b −1 2 )) ci = fn−j ( b−11 , b−12 ; t−1; ξn−i ( x, b−12 ; t−1 )) ci ∣∣∣ x=a1tn−i−1 , (6.24) where ci is the constant given in (3.5a) as ci = ( a1b2t i; t ) n−i ( a−11 b−12 t−(i−1); t ) i (t; t)i(t; t)n−i t( n 2)(1− t)n . (6.25) From (6.11) in Corollary 6.6 we have fn−j ( b−11 , b−12 ; t−1; ξn−i ( x, b−12 ; t−1 )) = ( xb1; t −1) j−i ( xb2t i; t−1 ) n−j(t; t)n−i( b1b −1 2 t−(i+j−n); t−1 ) j−i ( b−11 b2tj ; t−1 ) n−j(t; t)j−i(t; t)n−j . Combining this, (6.24) and (6.25), we therefore obtain the expression (6.22). � Proof of Theorem 1.4. From (6.12) and (6.14) in Lemma 6.7, we have( en(a2, b1; z), en−1(a2, b1; z), . . . , e0(a2, b1; z) ) = ( e0(a1, b2; z), e1(a1, b2; z), . . . , en(a1, b2; z) ) L̃RŨR, (6.26) where L̃R = ( l̃Rij ) 0≤i,j≤n and ŨR = ( ũR ij ) 0≤i,j≤n are the matrices given by (6.13) and (6.15), respectively. Comparing (6.26) with (1.14), we obtain R = L̃RŨR = LRDRUR, i.e., lRij = l̃Rij l̃Rjj , dR j = l̃Rjj ũ R jj , uR ij = ũR ij ũR ii . Lemma 6.7 implies that lRij , d R j and uR ij above coincide with (1.15a), (1.15b) and (1.15c), respec- tively. On the other hand, from (6.19) and (6.21) in Lemma 6.8, we have( en(a2, b1; z), en−1(a2, b1; z), . . . , e0(a2, b1; z) ) = ( e0(a1, b2; z), e1(a1, b2; z), . . . , en(a1, b2; z) ) Ũ ′RL̃ ′ R, (6.27) where Ũ ′R = ( ũR ′ ij ) 0≤i,j≤n and L̃′R = ( l̃R ′ij ) 0≤i,j≤n are the matrices given by (6.20) and (6.22), respectively. Comparing (6.27) with (1.14), we obtain R = Ũ ′RL̃ ′ R = U ′RD ′ RL ′ R, i.e., uR ′ ij = ũR ′ ij ũR ′ jj , dR ′ j = ũR ′ jj l̃ R ′ jj , lR ′ij = l̃R ′ij l̃R ′ii . Lemma 6.8 implies that uR ′ ij , dR ′ j and lR ′ij above coincide with (1.16a), (1.16b) and (1.16c), respectively. � 28 M. Ito A Appendix In this appendix we consider the Gauss decomposition part A = U ′AD ′ AL ′ A of Theorem 1.7. Since the method to compute A = U ′AD ′ AL ′ A is almost the same as that to compute A = LADAUA, we only give the outline of the proof. For this purpose, we define another family of interpolation polynomials Ẽ′k,i(a, b; z) slightly different from (3.2). Set Ẽ′k,i(z) = Ẽ′k,i(a, b; z) := AE′k,i(a, b; z)/∆(z), where E′k,i(a, b; z) := zn−k+1zn−k+2 · · · zn︸ ︷︷ ︸ k ∆(t; z) n−i∏ j=1 (1− bzj) n∏ j=n−i+1 ( 1− a−1zj ) . We now specify a = a1, b = b2, i.e., we set Ẽ′k,i(z) = Ẽ′k,i(a1, b2; z) throughout this section. Lemma A.1 (three-term relations). Suppose k ≤ i. Then, a−12 ( 1− qαa1a2b1b2t2n−k−1 )〈 Ẽ′k,i 〉 = tk−1 ( 1− qαa1b1t2n−k−i )〈 Ẽ′k−1,i 〉 − qαtn−1 ( 1− a1b1tn−i )〈 Ẽ′k−1,i−1 〉 . (A.1) On the other hand, if k ≥ i, then, a−11 ( 1− qαa1b1t2n−k−i )〈 Ẽ′k,i−1 〉 = tk−i ( 1− qαtn−k )〈 Ẽ′k−1,i−1 〉 − a−12 t−(i−1) ( 1− a2b2ti−1 )〈 Ẽ′k,i 〉 . (A.2) Proof. Put ϕ̃′k,i−1(z) := A∇1ϕ′k,i−1(z), where ϕ′k,i−1(z) := ( 1− a−11 z1 )( 1− a−12 z1 ) n∏ j=2 (z1 − tzj)× E(n−1) k−1,i−1(z2, . . . , zn). Then, by a similar argument to that used in the proof of Lemma 5.3, it follows that, if k ≤ i, the polynomial ϕ̃′k,i−1(z) satisfies ϕ̃′k,i−1(z) ∆(z) = c′k,iẼ ′ k,i(z) + c′k−1,iẼ ′ k−1,i(z) + c′k−1,i−1Ẽ ′ k−1,i−1(z), (A.3) where c′k,i = −a−12 tn−1 ( 1− qαa1a2b1b2t2n−k−1 ) , c′k−1,i = tn+k−2 ( 1− qαa1b1t2n−k−i ) , c′k−1,i−1 = −qαt2n−2 ( 1− a1b1tn−i ) , while, if i ≤ k, then ϕ̃′k,i−1(z) satisfies ϕ̃′k,i−1(z) ∆(z) = d′k,iẼ ′ k,i(z) + d′k,i−1Ẽ ′ k,i−1(z) + d′k−1,i−1Ẽ ′ k−1,i−1(z), (A.4) where d′k,i = −a−12 tn−1 ( 1− a2b2ti−1 ) , d′k,i−1 = −a−11 tn+i−2 ( 1− qαa1b1t2n−k−i ) , d′k−1,i−1 = tn+k−2 ( 1− qαtn−k ) . Using the expressions (A.3) and (A.4) for ϕ̃k,i−1(z), (A.1) and (A.2) follow by application of Lemma 5.1. � q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 29 By repeated use of the three-term relations (A.1) and (A.2), we obtain the following. Lemma A.2. If k ≤ i, then 〈 Ẽ′k,i 〉 = l∑ j=0 U ′k,ik−l,i−j 〈 Ẽ′k−l,i−j 〉 , where U ′k,ik−l,i−j = ( −qαtn−k+l−1 )j( a2t k−l)lt(l−j 2 ) [ l j ] t ( a1b1t n−i; t ) j ( qαa1b1t 2n−k−i+j ; t ) l−j( qαa1a2b1b2t2n−k−1; t ) l , while, if k ≥ i, then 〈 Ẽ′k,i 〉 = l∑ j=0 L′k,ik−l+j,i+j 〈 Ẽ′k−l+j,i+j 〉 , where L′k,ik−l+j,i+j = [ l j ] t ( −a−12 t−(k−1) )j( a1t k−i−1)lt−(l 2) ( qαtn−k; t ) l−j ( a2b2t i; t ) j( qαa1b1t2n−k−i−j−1; t ) l−j ( qαa1b1t2n−k−i−2j+l; t ) j . As a special case of the above lemma we immediately have the following. Lemma A.3. For 0 ≤ j ≤ n, 〈 Ẽ′j,j 〉 is expressed as 〈 Ẽ′j,j 〉 = j∑ i=0 ũ′ij 〈 Ẽ′0,i 〉 , (A.5) where ũ′ij = U ′j,j0,i = ( −qαtn−1 )j−i aj2t (i 2) [ j i ] t ( a1b1t n−j ; t ) j−i ( qαa1b1t 2n−i−j ; t ) i( qαa1a2b1b2t2n−j−1; t ) j , (A.6) while, for 0 ≤ j ≤ n, 〈 Ẽ′n,j 〉 is expressed as 〈 Ẽ′n,j 〉 = n∑ i=j l̃′ij 〈 Ẽ′i,i 〉 , (A.7) where l̃′ij = L′n,ji,i = (−1)i−j [ n− j n− i ] t an−j1 a −(i−j) 2 t( n−i 2 )+(j2)−(i 2) ( qα; t ) n−i ( a2b2t j ; t ) i−j( qαa1b1tn−i−1; t ) n−i ( qαa1b1t2(n−i); t ) i−j . (A.8) Proof of (1.19a)–(1.19c) in Theorem 1.7. From (A.5), we have(〈 Ẽ′0,0 〉 , 〈 Ẽ′1,1 〉 , . . . , 〈 Ẽ′n−1,n−1 〉 , 〈 Ẽ′n,n 〉) = (〈 Ẽ′0,0 〉 , 〈 Ẽ′0,1 〉 , . . . , 〈 Ẽ′0,n−1 〉 , 〈 Ẽ′0,n 〉) Ũ ′, where the matrix Ũ ′ = ( ũ′ij ) 0≤i,j≤n is defined by (A.6). Moreover, from (A.7) we have(〈 Ẽ′n,0 〉 , 〈 Ẽ′n,1 〉 , . . . , 〈 Ẽ′n,n−1 〉 , 〈 Ẽ′n,n 〉) = (〈 Ẽ′0,0 〉 , 〈 Ẽ′1,1 〉 , . . . , 〈 Ẽ′n−1,n−1 〉 , 〈 Ẽ′n,n 〉) L̃′ = (〈 Ẽ′0,0 〉 , 〈 Ẽ′0,1 〉 , . . . , 〈 Ẽ′0,n−1 〉 , 〈 Ẽ′0,n 〉) Ũ ′L̃′, (A.9) 30 M. Ito where the matrix L̃′ = ( l̃′ij ) 0≤i,j≤n is defined by (A.8). Since TαΦ(z) = z1z2 · · · znΦ(z) and z1z2 · · · znẼ′0,i(z) = Ẽ′n,i(z), we have Tα 〈 Ẽ′0,i 〉 = 〈 Ẽ′n,i 〉 , i.e., Tα (〈 Ẽ′0,0 〉 , 〈 Ẽ′0,1 〉 , . . . , 〈 Ẽ′0,n−1 〉 , 〈 Ẽ′0,n 〉) = (〈 Ẽ′n,0 〉 , 〈 Ẽ′n,1 〉 , . . . , 〈 Ẽ′n,n−1 〉 , 〈 Ẽ′n,n 〉) . (A.10) From (A.9) and (A.10), we obtain the difference system Tα (〈 Ẽ′0,0 〉 , 〈 Ẽ′0,1 〉 , . . . , 〈 Ẽ′0,n−1 〉 , 〈 Ẽ′0,n 〉) = (〈 Ẽ′0,0 〉 , 〈 Ẽ′0,1 〉 , . . . , 〈 Ẽ′0,n−1 〉 , 〈 Ẽ′0,n 〉) Ũ ′L̃′. Comparing this with (1.17), we therefore obtain A = Ũ ′L̃′ = U ′AD ′ AL ′ A, i.e., uA ′ ij = ũ′ij ũ′jj , dA ′ j = ũ′jj l̃ ′ jj , lA ′ij = l̃′ij l̃′ii . Lemma A.3 implies that uA ′ ij , dA ′ j and lA ′ij above coincide with (1.19a), (1.19b) and (1.19c), respectively, which completes the proof. � Finally we give an explicit forms for L′A −1. Proposition A.4. The inverse matrix L′−1A = ( lA ′∗ij ) 0≤i,j≤n is lower triangular and is written as lA ′∗ij = [ n− j n− i ] t ( a1a −1 2 t−j )i−j( a2b2t j ; t ) i−j( qαa1b1t2n−i−j−1; t ) i−j . Proof. Using (A.2) we can calculate the entries lA ′∗ij of the lower triangular matrix L′−1A by completely the same way as Proposition 4.4. We omit the details. � Acknowledgements The author would like to thank the anonymous referees who kindly provided many careful comments and suggestions for improving his manuscript. He would also like to express his gratitude to Professor Peter J. Forrester for providing considerable encouragement from the early stage of this research. The comments and suggestions by Professor Yasuhiko Yamada on the preliminary version of this manuscript are most appreciated. This work was supported by JSPS KAKENHI Grant Number (C)18K03339. References [1] Albion S.P., Rains E.M., Warnaar S.O., AFLT-type Selberg integrals, arXiv:2001.05637. [2] Andrews G.E., Askey R., Roy R., Special functions, Encyclopedia of Mathematics and its Applications, Vol. 71, Cambridge University Press, Cambridge, 1999. [3] Aomoto K., Jacobi polynomials associated with Selberg integrals, SIAM J. Math. Anal. 18 (1987), 545–549. [4] Aomoto K., q-analogue of de Rham cohomology associated with Jackson integrals. I, Proc. Japan Acad. Ser. A Math. Sci. 66 (1990), 161–164. [5] Aomoto K., Connection matrices and Riemann–Hilbert problem for q-difference equations, in Structure of Solutions of Differential Equations (Katata/Kyoto, 1995), World Sci. Publ., River Edge, NJ, 1996, 51–69. [6] Aomoto K., On elliptic product formulas for Jackson integrals associated with reduced root systems, J. Al- gebraic Combin. 8 (1998), 115–126. [7] Aomoto K., Kato Y., A q-analogue of de Rham cohomology associated with Jackson integrals, in Special Functions (Okayama, 1990), ICM-90 Satell. Conf. Proc., Springer, Tokyo, 1991, 30–62. https://arxiv.org/abs/2001.05637 https://doi.org/10.1017/CBO9781107325937 https://doi.org/10.1137/0518042 https://doi.org/10.3792/pjaa.66.161 https://doi.org/10.3792/pjaa.66.161 https://doi.org/10.1023/A:1008629309210 https://doi.org/10.1023/A:1008629309210 https://doi.org/10.1007/978-4-431-68170-0_2 q-Difference Systems for the Jackson Integral of Symmetric Selberg Type 31 [8] Aomoto K., Kato Y., Gauss decomposition of connection matrices and application to Yang–Baxter equa- tion. I, Proc. Japan Acad. Ser. A Math. Sci. 69 (1993), 238–242. [9] Aomoto K., Kato Y., Gauss decomposition of connection matrices for symmetric A-type Jackson integrals, Selecta Math. (N.S.) 1 (1995), 623–666. [10] Aomoto K., Kato Y., Derivation of q-difference equation from connection matrix for Selberg type Jackson integrals, J. Differ. Equations Appl. 4 (1998), 247–278. [11] Askey R., Some basic hypergeometric extensions of integrals of Selberg and Andrews, SIAM J. Math. Anal. 11 (1980), 938–951. [12] Bosnjak G., Mangazeev V.V., Construction of R-matrices for symmetric tensor representations related to Uq(ŝln), J. Phys. A: Math. Theor. 49 (2016), 495204, 19 pages, arXiv:1607.07968. [13] Evans R.J., Multidimensional q-beta integrals, SIAM J. Math. Anal. 23 (1992), 758–765. [14] Forrester P.J., Ito M., Difference system for Selberg correlation integrals, J. Phys. A: Math. Theor. 43 (2010), 175202, 19 pages, arXiv:1011.1650. [15] Forrester P.J., Warnaar S.O., The importance of the Selberg integral, Bull. Amer. Math. Soc. (N.S.) 45 (2008), 489–534, arXiv:0710.3981. [16] Habsieger L., Une q-intégrale de Selberg et Askey, SIAM J. Math. Anal. 19 (1988), 1475–1489. [17] Ishikawa M., Zeng J., Hankel hyperpfaffian calculations and Selberg integrals, arXiv:2008.09776. [18] Ito M., Forrester P.J., A bilateral extension of the q-Selberg integral, Trans. Amer. Math. Soc. 369 (2017), 2843–2878, arXiv:1309.0001. [19] Ito M., Noumi M., Connection formula for the Jackson integral of type An and elliptic Lagrange interpola- tion, SIGMA 14 (2018), 077, 42 pages, arXiv:1801.07041. [20] Kadell K.W.J., A proof of Askey’s conjectured q-analogue of Selberg’s integral and a conjecture of Morris, SIAM J. Math. Anal. 19 (1988), 969–986. [21] Kaneko J., Selberg integrals and hypergeometric functions associated with Jack polynomials, SIAM J. Math. Anal. 24 (1993), 1086–1110. [22] Kaneko J., q-Selberg integrals and Macdonald polynomials, Ann. Sci. École Norm. Sup. (4) 29 (1996), 583–637. [23] Kim J.S., Okada S., A new q-Selberg integral, Schur functions, and Young books, Ramanujan J. 42 (2017), 43–57, arXiv:1412.7914. [24] Kim J.S., Stanton D., On q-integrals over order polytopes, Adv. Math. 308 (2017), 1269–1317, arXiv:1608.03342. [25] Kim J.S., Yoo M., Hook length property of d-complete posets via q-integrals, J. Combin. Theory Ser. A 162 (2019), 167–221, arXiv:1708.09109. [26] Kuniba A., Okado M., Yoneyama A., Matrix product solution to the reflection equation associated with a coideal subalgebra of Uq ( A (1) n−1 ) , Lett. Math. Phys. 109 (2019), 2049–2067, arXiv:1812.03767. [27] Macdonald I.G., Symmetric functions and Hall polynomials, 2nd ed., Oxford Mathematical Monographs, The Clarendon Press, Oxford University Press, New York, 1995. [28] Matsuo A., Jackson integrals of Jordan–Pochhammer type and quantum Knizhnik–Zamolodchikov equa- tions, Comm. Math. Phys. 151 (1993), 263–273. [29] Matsuo A., Quantum algebra structure of certain Jackson integrals, Comm. Math. Phys. 157 (1993), 479– 498. [30] Mimachi K., Connection problem in holonomic q-difference system associated with a Jackson integral of Jordan–Pochhammer type, Nagoya Math. J. 116 (1989), 149–161. [31] Mimachi K., Holonomic q-difference system of the first order associated with a Jackson integral of Selberg type, Duke Math. J. 73 (1994), 453–468. [32] Rimányi R., Tarasov V., Varchenko A., Zinn-Justin P., Extended Joseph polynomials, quantized conformal blocks, and a q-Selberg type integral, J. Geom. Phys. 62 (2012), 2188–2207, arXiv:1110.2187. [33] Varchenko A., Quantized Knizhnik–Zamolodchikov equations, quantum Yang–Baxter equation, and differ- ence equations for q-hypergeometric functions, Comm. Math. Phys. 162 (1994), 499–528. [34] Varchenko A., Special functions, KZ type equations, and representation theory, CBMS Regional Conference Series in Mathematics, Vol. 98, Amer. Math. Soc., Providence, RI, 2003. [35] Warnaar S.O., q-Selberg integrals and Macdonald polynomials, Ramanujan J. 10 (2005), 237–268. https://doi.org/10.3792/pjaa.69.238 https://doi.org/10.1007/BF01587906 https://doi.org/10.1080/10236199808808141 https://doi.org/10.1137/0511084 https://doi.org/10.1088/1751-8113/49/49/495204 https://arxiv.org/abs/1607.07968 https://doi.org/10.1137/0523039 https://doi.org/10.1088/1751-8113/43/17/175202 https://arxiv.org/abs/1011.1650 https://doi.org/10.1090/S0273-0979-08-01221-4 https://arxiv.org/abs/0710.3981 https://doi.org/10.1137/0519111 https://arxiv.org/abs/2008.09776 https://doi.org/10.1090/tran/6851 https://arxiv.org/abs/1309.0001 https://doi.org/10.3842/SIGMA.2018.077 https://arxiv.org/abs/1801.07041 https://doi.org/10.1137/0519067 https://doi.org/10.1137/0524064 https://doi.org/10.1137/0524064 https://doi.org/10.24033/asens.1749 https://doi.org/10.1007/s11139-015-9721-9 https://arxiv.org/abs/1412.7914 https://doi.org/10.1016/j.aim.2017.01.001 https://arxiv.org/abs/1608.03342 https://doi.org/10.1016/j.jcta.2018.11.003 https://arxiv.org/abs/1708.09109 https://doi.org/10.1007/s11005-019-01175-x https://arxiv.org/abs/1812.03767 https://doi.org/10.1007/BF02096769 https://doi.org/10.1007/BF02096880 https://doi.org/10.1017/S0027763000001744 https://doi.org/10.1215/S0012-7094-94-07319-5 https://doi.org/10.1016/j.geomphys.2012.06.008 https://arxiv.org/abs/1110.2187 https://doi.org/10.1007/BF02101745 https://doi.org/10.1090/cbms/098 https://doi.org/10.1007/s11139-005-4849-7 1 Introduction 2 Notation 3 Interpolation polynomials 4 Three-term relations 5 Proof of Lemma 4.1 6 The transition matrix R A Appendix References
id nasplib_isofts_kiev_ua-123456789-211007
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1815-0659
language English
last_indexed 2026-03-16T07:30:58Z
publishDate 2020
publisher Інститут математики НАН України
record_format dspace
spelling Ito, Masahiko
2025-12-22T09:25:02Z
2020
q-Difference Systems for the Jackson Integral of Symmetric Selberg Type. Masahiko Ito. SIGMA 16 (2020), 113, 31 pages
1815-0659
2020 Mathematics Subject Classification: 33D60; 39A13
arXiv:1910.08393
https://nasplib.isofts.kiev.ua/handle/123456789/211007
https://doi.org/10.3842/SIGMA.2020.113
We provide an explicit expression for the first-order -difference system for the Jackson integral of symmetric Selberg type. The q-difference system gives a generalization of the -analog of contiguous relations for the Gauss hypergeometric function. As a basis of the system, we use a set of symmetric polynomials introduced by Matsuo in his study of the -KZ equation. Our main result is an explicit expression for the coefficient matrix of the -difference system in terms of its Gauss matrix decomposition. We introduce a class of symmetric polynomials called interpolation polynomials, which includes Matsuo's polynomials. By repeated use of three-term relations among the interpolation polynomials, we compute the coefficient matrix.
The author would like to thank the anonymous referees who kindly provided many careful comments and suggestions for improving his manuscript. He would also like to express his gratitude to Professor Peter J. Forrester for providing considerable encouragement from the early stage of this research. The comments and suggestions by Professor Yasuhiko Yamada on the preliminary version of this manuscript are most appreciated. This work was supported by JSPS KAKENHI Grant Number (C)18K03339.
en
Інститут математики НАН України
Symmetry, Integrability and Geometry: Methods and Applications
q-Difference Systems for the Jackson Integral of Symmetric Selberg Type
Article
published earlier
spellingShingle q-Difference Systems for the Jackson Integral of Symmetric Selberg Type
Ito, Masahiko
title q-Difference Systems for the Jackson Integral of Symmetric Selberg Type
title_full q-Difference Systems for the Jackson Integral of Symmetric Selberg Type
title_fullStr q-Difference Systems for the Jackson Integral of Symmetric Selberg Type
title_full_unstemmed q-Difference Systems for the Jackson Integral of Symmetric Selberg Type
title_short q-Difference Systems for the Jackson Integral of Symmetric Selberg Type
title_sort q-difference systems for the jackson integral of symmetric selberg type
url https://nasplib.isofts.kiev.ua/handle/123456789/211007
work_keys_str_mv AT itomasahiko qdifferencesystemsforthejacksonintegralofsymmetricselbergtype