A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications
We present an explicit expression for the q-difference system, which the BC1-type Jackson integral (q-series) satisfies, as first order simultaneous q-difference equations with a concrete basis. As an application, we give a simple proof for the hypergeometric summation formula introduced by Gustafso...
Збережено в:
| Опубліковано в: : | Symmetry, Integrability and Geometry: Methods and Applications |
|---|---|
| Дата: | 2009 |
| Автор: | |
| Формат: | Стаття |
| Мова: | Англійська |
| Опубліковано: |
Інститут математики НАН України
2009
|
| Онлайн доступ: | https://nasplib.isofts.kiev.ua/handle/123456789/149165 |
| 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: | A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications / M. Ito // Symmetry, Integrability and Geometry: Methods and Applications. — 2009. — Т. 5. — Бібліогр.: 17 назв. — англ. |
Репозитарії
Digital Library of Periodicals of National Academy of Sciences of Ukraine| _version_ | 1859479430393495552 |
|---|---|
| author | Ito, M. |
| author_facet | Ito, M. |
| citation_txt | A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications / M. Ito // Symmetry, Integrability and Geometry: Methods and Applications. — 2009. — Т. 5. — Бібліогр.: 17 назв. — англ. |
| collection | DSpace DC |
| container_title | Symmetry, Integrability and Geometry: Methods and Applications |
| description | We present an explicit expression for the q-difference system, which the BC1-type Jackson integral (q-series) satisfies, as first order simultaneous q-difference equations with a concrete basis. As an application, we give a simple proof for the hypergeometric summation formula introduced by Gustafson and the product formula of the q-integral introduced by Nassrallah-Rahman and Gustafson.
|
| first_indexed | 2025-11-24T12:51:25Z |
| format | Article |
| fulltext |
Symmetry, Integrability and Geometry: Methods and Applications SIGMA 5 (2009), 041, 14 pages
A First Order q-Difference System for the BC1-Type
Jackson Integral and Its Applications?
Masahiko ITO
Department of Physics and Mathematics, Aoyama Gakuin University,
Kanagawa 229-8558, Japan
E-mail: mito@gem.aoyama.ac.jp
Received December 01, 2008, in final form March 18, 2009; Published online April 03, 2009
doi:10.3842/SIGMA.2009.041
Abstract. We present an explicit expression for the q-difference system, which the BC1-
type Jackson integral (q-series) satisfies, as first order simultaneous q-difference equations
with a concrete basis. As an application, we give a simple proof for the hypergeometric
summation formula introduced by Gustafson and the product formula of the q-integral
introduced by Nassrallah–Rahman and Gustafson.
Key words: q-difference equations; Jackson integral of type BC1; Gustafson’s Cn-type sum;
Nassrallah–Rahman integral
2000 Mathematics Subject Classification: 33D15; 33D67; 39A13
1 Introduction
A lot of summation and transformation formulae for basic hypergeometric series have been
found to date. The BC1-type Jackson integral, which is the main subject of interest in this
paper, is a q-series which can be written as a basic hypergeometric series in a class of so
called very-well-poised-balanced 2rψ2r. A key reason to consider the BC1-type Jackson inte-
grals is to give an explanation of these hypergeometric series from the view points of the Weyl
group symmetry and the q-difference equations of the BC1-type Jackson integrals. In [15], we
showed that Slater’s transformation formula for a very-well-poised-balanced 2rψ2r series can
be regarded as a connection formula for the solutions of q-difference equations of the BC1-
type Jackson integral, i.e., the Jackson integral as a general solution of q-difference system is
written as a linear combination of particular solutions. As a consequence we gave a simple
proof of Slater’s transformation formula. (See [15] for details. Also see [13] for a connec-
tion formula for the BCn-type Jackson integral, which is a multisum generalization of that of
type BC1.)
The aim of this paper is to present an explicit form of the q-difference system as first order
simultaneous q-difference equations for the BC1-type Jackson integral with generic condition
on the parameters. We give the Gauss decomposition of the coefficient matrix of the system
with a concrete basis (see Theorem 4.1). Each entry of the decomposed matrices is written
as a product of binomials and, as a consequence, the determinant of the coefficient matrix is
easy to calculate explicitly. As an application we give a simple proof of the product formula
for Gustafson’s multiple Cn-type sum [10]. We also present an explicit form of the q-difference
system for the BC1-type Jackson integral with a balancing condition on the parameters. We
finally give a simple proof of the product formula for the q-integral of Nassrallah–Rahman [16]
and Gustafson [9]. A recent work of Rains and Spiridonov [17] contains results for the elliptic
?This paper is a contribution to the Proceedings of the Workshop “Elliptic Integrable Systems, Isomonodromy
Problems, and Hypergeometric Functions” (July 21–25, 2008, MPIM, Bonn, Germany). The full collection is
available at http://www.emis.de/journals/SIGMA/Elliptic-Integrable-Systems.html
mailto:mito@gem.aoyama.ac.jp
http://dx.doi.org/10.3842/SIGMA.2009.041
http://www.emis.de/journals/SIGMA/Elliptic-Integrable-Systems.html
2 M. Ito
hypergeometric integral of a similar type to those contained for the BC1-type Jackson integral
obtained here.
2 BC1-type Jackson integral
Throughout this paper, we assume 0 < q < 1 and denote the q-shifted factorial for all integers N
by (x)∞ :=
∞∏
i=0
(1− qix) and (x)N := (x; q)∞/(qNx; q)∞.
Let O(C∗) be the set of holomorphic functions on the complex multiplicative group C∗.
A function f on C∗ is said to be symmetric or skew-symmetric under the Weyl group action
z → z−1 if f satisfies f(z) = f(z−1) or f(z) = −f(z−1), respectively. For ξ ∈ C∗ and a function
f on C∗, we define the sum over the lattice Z∫ ξ∞
0
f(z)
dqz
z
:= (1− q)
∞∑
ν=−∞
f(qνξ),
which, provided the integral converges, we call the Jackson integral. For an arbitrary positive
integer s, we define the function Φ and the skew-symmetric function ∆ on C∗ as follows:
Φ(z) :=
2s+2∏
m=1
z
1
2
−αm
(qz/am)∞
(zam)∞
, ∆(z) := z−1 − z, (2.1)
where am = qαm . For a symmetric function ϕ on C∗ and a point ξ ∈ C∗, we define the following
sum over the lattice Z:∫ ξ∞
0
ϕ(z)Φ(z)∆(z)
dqz
z
,
which we call the Jackson integral of type BC1 and is simply denoted by 〈ϕ, ξ〉. By definition
the sum 〈ϕ, ξ〉 is invariant under the shift ξ → qνξ for ν ∈ Z.
Let Θ(z) be the function on C∗ defined by
Θ(z) :=
zs−α1−···−α2s+2θ(z2)
2s+2∏
m=1
θ(amz)
,
where θ(z) denotes the function (z)∞(q/z)∞, which satisfies
θ(qz) = −θ(z)/z and θ(q/z) = θ(z). (2.2)
For a symmetric function ϕ ∈ O(C∗), we denote the function 〈ϕ, z〉/Θ(z) by 〈〈ϕ, z〉〉. We call
〈〈ϕ, z〉〉 the regularized Jackson integral of type BC1, which satisfies the following:
Lemma 2.1. Assume α1 + α2 + · · · + α2s+2 6∈ 1
2 + Z. If ϕ ∈ O(C∗) is symmetric, then the
function 〈〈ϕ, z〉〉 is symmetric and holomorphic on C∗.
Proof. See [15, Proposition 2.2]. �
For an arbitrary meromorphic function ϕ on C∗ we define the function ∇ϕ on C∗ by
∇ϕ(z) := ϕ(z)− Φ(qz)
Φ(z)
ϕ(qz).
A First Order q-Difference System for the BC1-Type Jackson Integral 3
In particular, from (2.1), the function Φ(qz)/Φ(z) is the rational function
Φ(qz)
Φ(z)
= qs+1
2s+2∏
m=1
1− amz
am − qz
.
The following proposition will be used for the proof of the key equation (Theorem 3.1):
Lemma 2.2. If
∫ ξ∞
0
Φ(z)ϕ(z)
dqz
z
is convergent for ϕ ∈ O(C∗), then
∫ ξ∞
0
Φ(z)∇ϕ(z)
dqz
z
= 0.
Proof. See [11, Lemma 5.1] for instance. �
3 Key equation
In this section, we will present a key equation to construct the difference equations for the
BC1-type Jackson integral. Before we state it, we introduce the function e(x; y) defined by
e(x; y) := x+ x−1 −
(
y + y−1
)
,
which is expressed by the product form
e(x; y) =
(y − x)(1− xy)
xy
.
The basic properties of e(x; y) are the following:
• e(x; z) = e(x; y) + e(y; z), (3.1)
• e(x; y) = −e(y;x), e(x; y) = e
(
x−1; y
)
, (3.2)
• e(x; y)e(z;w)− e(x; z)e(y;w) + e(x;w)e(y; z) = 0. (3.3)
Remark 3.1. As we will see later, equation (3.3) is ignorable in the case a1a2 · · · a2s+2 6= 1,
while equation (3.1) is ignorable in the case a1a2 · · · a2s+2 = 1.
For functions f , g on C∗, the function fg on C∗ is defined by
(fg)(z) := f(z)g(z) for z ∈ C∗.
Set ei(z) := e(z; ai) and (ei1ei2 · · ·eis)(z) := ei1(z)ei2(z)· · ·eis(z). The symbol (ei1 · · ·êik · · ·eis)(z)
is equal to (ei1 · · · eik−1
eik+1
· · · eis)(z). The key equation is the following:
Theorem 3.1. Suppose ai 6= aj if i 6= j. If {i1, i2, . . . , is} ⊂ {1, 2, . . . , 2s+ 2}, then
C0〈ei1ei2 · · · eis , ξ〉+
s∑
k=1
Cik〈ei1 · · · êik · · · eis , ξ〉 = 0,
where the coefficients C0 and Cik (1 ≤ k ≤ s) are given by
C0 = 1− a1a2 · · · a2s+2 and Cik =
2s+2∏
m=1
(1− aikam)
as
ik
(1− a2
ik
)
∏
1≤`≤s
` 6=k
e(aik ; ai`)
.
4 M. Ito
Proof. Without loss of generality, it suffices to show that
C0〈e1e2 · · · es, ξ〉+
s∑
i=1
Ci 〈e1 · · · êi · · · es, ξ〉 = 0, (3.4)
where the coefficients C0 and Ci are given by
C0 = 1− a1a2 · · · a2s+2 and Ci =
2s+2∏
m=1
(1− aiam)
as
i (1− a2
i )
∏
1≤k≤s
k 6=i
e(ai; ak)
. (3.5)
Set F (z) =
2s+2∏
m=1
(am − z) and G(z) =
2s+2∏
m=1
(1− amz). Then, from Lemma 2.2, it follows that
∫ ξ∞
0
Φ(z)∇
(
F (z)
zs+1
)
dqz
z
= 0, where ∇
(
F (z)
zs+1
)
=
F (z)−G(z)
zs+1
. (3.6)
Since (F (z) − G(z))/zs+1 is skew-symmetric under the reflection z → z−1, it is divisible by
z − z−1, and we can expand it as
F (z)−G(z)
zs+1(z − z−1)
= C0 e(z; a1)e(z; a2) · · · e(z; as) +
s∑
i=1
Ci e(z; a1) · · · ê(z; ai) · · · e(z; as), (3.7)
where the coefficients Ci will be determined below. We obtain C0 = 1− a1a2 · · · a2s+2 from the
principal term of asymptotic behavior of (3.7) as z → +∞. If we put z = ai (1 ≤ i ≤ s), then
we have
−F (ai)−G(ai)
as
i (1− a2
i )
= Ci
∏
1≤k≤s
k 6=i
e(ai; ak).
Since F (ai) = 0 and G(ai) =
2s+2∏
m=1
(1 − aiam) by definition, the above equation implies (3.5).
From (3.6) and (3.7), we obtain (3.4), which completes the proof. �
4 The case a1a2 · · · a2s+2 6= 1
4.1 q-difference equation
Set
vk(z) :=
{
ei1ei2 · · · eis−1(z) if k = 0,
ei1 · · · êik · · · eis−1(z) if 1 ≤ k ≤ s− 1,
(4.1)
where the hat symbol denotes the term to be omitted.
Let Taj be the difference operator corresponding to the q-shift aj → qaj .
Theorem 4.1. Suppose a1a2· · ·a2s+2 6=1. For the BC1-type Jackson integrals, if {i1, i2, . . . , is−1}
⊂ {1, 2, . . . , 2s + 2} and j 6∈ {i1, i2, . . . , is−1}, then the first order vector-valued q-difference
equation with respect to the basis {v0, v1, . . . , vs−1} defined by (4.1) is given by
Taj (〈v0, ξ〉, . . . , 〈vs−1, ξ〉) = (〈v0, ξ〉, . . . , 〈vs−1, ξ〉)B, (4.2)
A First Order q-Difference System for the BC1-Type Jackson Integral 5
where B = UL. Here U and L are the s× s matrices defined by
U =
c0 1 1 · · · 1
c1
c2
. . .
cs−1
, L =
1
d1 1
d2 1
...
. . .
ds−1 1
,
where
c0 =
2s+2∏
m=1
(1− ajam)
(−aj)s(1− a1a2 · · · a2s+2)
(
1− a2
j
) s−1∏
`=1
e(ai` ; aj)
and
ck = e(aik ; aj), dk =
2s+2∏
m=1
(1− aikam)
(−aik)s(1−a1a2 · · · a2s+2)(1−a2
ik
)e(aj ; aik)
∏
1≤`≤s−1
` 6=k
e(ai` ; aik)
(4.3)
for k = 1, 2, . . . , s− 1. Moreover,
detB =
2s+2∏
m=1
(1− ajam)
(−aj)s(1− a1a2 · · · a2s+2)
(
1− a2
j
) .
Proof. Equation (4.2) is rewritten as Taj (〈v0, ξ〉, . . . , 〈vs−1, ξ〉)L−1 = (〈v0, ξ〉, . . . , 〈vs−1, ξ〉)U ,
where
L−1 =
1
−d1 1
−d2 1
...
. . .
−ds−1 1
.
Since Taj 〈vi, ξ〉 = 〈ejvi, ξ〉, the above equation is equivalent to
〈ejv0, ξ〉 −
s−1∑
k=1
dk〈ejvk, ξ〉 = c0〈v0, ξ〉 (4.4)
and
〈ejvk, ξ〉 = 〈v0, ξ〉+ ck〈vk, ξ〉 for k = 1, 2, . . . , s− 1, (4.5)
which are to be proved. Equation (4.4) is a direct consequence of (3.2) and Theorem 3.1 if
a1a2 · · · a2s+2 6= 1. Equation (4.5) is trivial using e(z; aj) = e(z; aik) + e(aik ; aj) from (3.1).
Lastly detB = detU detL = c0c1 · · · cs−1, which completes the proof. �
Since the function Θ(z) satisfies TajΘ(z) = −ajΘ(z), we immediately have the following from
Theorem 4.1:
6 M. Ito
Corollary 4.1. Suppose a1a2 · · · a2s+2 6= 1. For the regularized BC1-type Jackson integrals, if
{i1, i2, . . . , is−1} ⊂ {1, 2, . . . , 2s+ 2} and j 6∈ {i1, i2, . . . , is−1}, then the first order vector-valued
q-difference equation with respect to the basis {v0, v1, . . . , vs−1} is given by
Taj (〈〈v0, ξ〉〉, . . . , 〈〈vs−1, ξ〉〉) = (〈〈v0, ξ〉〉, . . . , 〈〈vs−1, ξ〉〉)B̄, (4.6)
where
B̄ =
−1
aj
c0 1 1 · · · 1
c1
c2
. . .
cs−1
1
d1 1
d2 1
...
. . .
ds−1 1
and ci and di are given by (4.3). In particular, the diagonal entries of the upper triangular part
are written as
− c0
aj
=
2s+2∏
m=1
(
1− a−1
j a−1
m
)
(
1− a−1
1 a−1
2 · · · a−1
2s+2
)(
1− a−2
j
) s−1∏
`=1
(
1− ai`a
−1
j
)(
1− a−1
i`
a−1
j
)
and
−ck
aj
=
(
1− aika
−1
j
)(
1− a−1
ik
a−1
j
)
for i = 1, 2, . . . , s− 1.
Moreover,
det B̄ =
2s+2∏
m=1
(
1− a−1
j a−1
m
)
(
1− a−2
j
)(
1− a−1
1 a−1
2 · · · a−1
2s+2
) . (4.7)
Remark 4.1. The q-difference system for the BCn-type Jackson integral is discussed in [3] for
its rank, and in [4, 5] for the explicit expression of the determinant of the coefficient matrix
of the system. On the other hand, though it is only for the BC1-type Jackson integral, the
coefficient matrix in its Gauss decomposition form is obtained explicitly only in the present
paper.
4.2 Application
The aim of this subsection is to give a simple proof of Gustafson’s multiple Cn-type summation
formula (Corollary 4.2). The point of the proof is to obtain a recurrence relation of Gustafson’s
multiple series of Cn-type. Before we state the recurrence relation, we first give the definition
of the multiple series of Cn-type 〈〈1, x〉〉G.
For z = (z1, . . . , zn) ∈ (C∗)n, we set
ΦG(z) :=
n∏
i=1
2s+2∏
m=1
z
1/2−αm
i
(qa−1
m zi)∞
(amzi)∞
,
∆Cn(z) :=
n∏
i=1
1− z2
i
zi
∏
1≤j<k≤n
(1− zj/zk)(1− zjzk)
zj
,
A First Order q-Difference System for the BC1-Type Jackson Integral 7
where qαm = am. For an arbitrary ξ = (ξ1, . . . , ξn) ∈ (C∗)n, we define the q-shift ξ → qνξ
by a lattice point ν = (ν1, . . . , νn) ∈ Zn, where qνξ := (qν1ξ1, . . . , q
νnξn) ∈ (C∗)n. For ξ =
(ξ1, . . . , ξn) ∈ (C∗)n we define the sum over the lattice Zn by
〈1, ξ〉G := (1− q)n
∑
ν∈Zn
ΦG(qνξ)∆Cn(qνξ),
which we call the BCn-type Jackson integral. Moreover we set 〈〈1, ξ〉〉G := 〈1, ξ〉G/ΘG(ξ), where
ΘG(ξ) :=
n∏
i=1
ξ
i−α1−α2−···−α2s+2
i θ(ξ2i )
2s+2∏
m=1
θ(amξi)
∏
1≤j<k≤n
θ(ξj/ξk)θ(ξjξk).
By definition, it can be confirmed that 〈〈1, ξ〉〉G is holomorphic on (C∗)n (see [4, Proposition 3.7]),
and we call it the regularized BCn-type Jackson integral. In particular, if we assume s = n we call
〈〈1, ξ〉〉G the regularized Jackson integral of Gustafson’s Cn-type, which is, in particular, a constant
not depending on ξ ∈ (C∗)n.
Remark 4.2. For further results on BCn-type Jackson integrals, see [2, 4, 3, 5, 13, 14], for
instance.
Now we state the recurrence relation for Gustafson’s sum 〈〈1, ξ〉〉G.
Proposition 4.1. Suppose s = n and x = (x1, . . . , xn) ∈ (C∗)n. The sum 〈〈1, x〉〉G satisfies
Taj 〈〈1, x〉〉G = 〈〈1, x〉〉G
2n+2∏
m=1
(1− a−1
j a−1
m )(
1− a−2
j
)(
1− a−1
1 a−1
2 · · · a−1
2n+2
) for j = 1, 2, . . . , 2n+ 2.
Proof. We assume s = n for the basis {v0, v1, . . . , vn−1} of the BC1-type Jackson integral.
Let P be the transition matrix from the basis {v0, v1, . . . , vn−1} to {χ(n−1), χ(n−2), . . . , χ(0)}:
(χ(n−1), χ(n−2), . . . , χ(0)) = (v0, v1, . . . , vn−1)P,
where χ(i) is the irreducible character of type C1 defined by
χ(i)(z) =
zi+1 − z−i−1
z − z−1
for i = 0, 1, 2, . . . . (4.8)
From (4.6) it follows that
Taj (〈〈χ(n−1), ξ〉〉, . . . , 〈〈χ(0), ξ〉〉) = (〈〈χ(n−1), ξ〉〉, . . . , 〈〈χ(0), ξ〉〉)P−1B̄P, (4.9)
so that
Taj det
(
〈〈χ(n−i), xj〉〉
)
1≤i,j≤n
= det
(
〈〈χ(n−i), xj〉〉
)
1≤i,j≤n
det B̄ (4.10)
for x = (x1, . . . , xn) ∈ (C∗)n. By definition, the relation between the determinant of the BC1-
type Jackson integrals and the Jackson integral of Gustafson’s Cn-type itself is given as
det
(
〈〈χ(n−i), xj〉〉
)
1≤i,j≤n
= 〈〈1, x〉〉G
∏
1≤j<k≤n
θ(xj/xk)θ(xjxk)
xj
, (4.11)
which is also referred to in [14]. From (4.10) and (4.11), we obtain Taj 〈〈1, x〉〉G = 〈〈1, x〉〉G det B̄,
where det B̄ has already been given in (4.7). �
8 M. Ito
Remark 4.3. The explicit form of the coefficient matrix of the system (4.9) is given in [1]
or [3].
Corollary 4.2 (Gustafson [10]). Suppose s = n and x = (x1, . . . , xn) ∈ (C∗)n. Then the sum
〈〈1, x〉〉G is written as
〈〈1, x〉〉G = (1− q)n
(q)n
∞
∏
1≤i<j≤2n+2
(
qa−1
i a−1
j
)
∞(
qa−1
1 a−1
2 · · · a−1
2n+2
)
∞
.
Proof. By repeated use of the recurrence relation in Proposition 4.1, using the asymptotic be-
havior of the Jackson integral as the boundary condition of the recurrence relation, we eventually
obtain Corollary 4.2. See [12] for further details about the proof. �
Remark 4.4. From (4.11) and Corollary 4.2, we see
det
(
〈〈χ(s−i), xj〉〉
)
1≤i,j≤s
= (1− q)s
(q)s
∞
∏
1≤i<j≤2s+2
(
qa−1
i a−1
j
)
∞(
qa−1
1 a−1
2 · · · a−1
2s+2
)
∞
∏
1≤j<k≤s
θ(xj/xk)θ(xjxk)
xj
,
which is non-degenerate under generic condition. This indicates that the set {χ(s−1), χ(s−2), . . .,
χ(0)} is linearly independent. And we eventually know the rank of the q-difference system with
respect to this basis is s, so are the ranks of the systems (4.2) and (4.6).
5 The case a1a2 · · · a2s+2 = 1
5.1 Reflection equation
Theorem 5.1. Suppose a1a2 · · · a2s+2 = 1. Let vk(z), k = 1, 2, . . . , s−1, be the functions defined
by (4.1) for the fixed indices i1, i2, . . . , is−1 ∈ {1, 2, . . . , 2s+ 2}. If j1, j2 6∈ {i1, i2, . . . , is−1}, then
(〈ej1v1, ξ〉, . . . , 〈ej1vs−1, ξ〉) = (〈ej2v1, ξ〉, . . . , 〈ej2vs−1, ξ〉)M,
where M = Mj2NM
−1
j1
. Here Mj and N are the matrices defined by
Mj =
γ1,j
γ2,j 1
γ3,j 1
...
. . .
γs−1,j 1
, N =
1 σ2 σ3 · · · σs−1
τ2
τ3
. . .
τs−1
,
where the entries of the above matrices are given by
σk =
e(aj1 ; aj2)
e(aik ; aj2)
, τk =
e(aj1 ; aik)
e(aj2 ; aik)
,
γk,j =
as
j
(
1− a2
j
)
as
ik
(
1− a2
ik
) 2s+2∏
m=1
1− aikam
1− ajam
∏
1≤`≤s−1
` 6=k
e(aj ; ai`)
e(aik ; ai`)
.
Moreover,
detM =
as
j2
(
1− a2
j2
)
as
j1
(
1− a2
j1
) 2s+2∏
m=1
1− aj1am
1− aj2am
. (5.1)
A First Order q-Difference System for the BC1-Type Jackson Integral 9
Proof. First we will prove the following:
(〈ej1v1, ξ〉, 〈ej1v2, ξ〉, . . . , 〈ej1vs−1, ξ〉)Mj1 = (〈v0, ξ〉, 〈ej2v2, ξ〉, . . . , 〈ej2vs−1, ξ〉)N, (5.2)
(〈ej2v1, ξ〉, 〈ej2v2, ξ〉, . . . , 〈ej2vs−1, ξ〉)Mj2 = (〈v0, ξ〉, 〈ej2v2, ξ〉, . . . , 〈ej2vs−1, ξ〉), (5.3)
which are equivalent to
s−1∑
k=1
γk,j〈ejvk, ξ〉 = 〈v0, ξ〉 (5.4)
and
〈ej1vk, ξ〉 = σk〈v0, ξ〉+ τk〈ej2vk, ξ〉, k = 2, . . . , s− 1. (5.5)
Under the condition a1a2 · · · a2s+2 = 1, Equation (5.4) is a direct consequence of Theorem 3.1.
Equation (5.5) is trivial from the equation
e(z; ai) = e(z; aj)
e(ai; ak)
e(aj ; ak)
+ e(z; ak)
e(ai; aj)
e(ak; aj)
,
which was given in (3.3). From (5.2) and (5.3), it follows M = Mj2NM
−1
j1
. Moreover, we obtain
detM =
detMj1 detN
detMj2
=
γ1,j1τ2 · · · τs−1
γ1,j2
=
as
j2
(
1− a2
j2
)
as
j1
(
1− a2
j1
) 2s+2∏
m=1
1− aj1am
1− aj2am
,
which completes the proof. �
Corollary 5.1. Suppose s = 2 and the condition a6 = q
a1a2a3a4a5
. The recurrence relation for
the BC1-type Jackson integral 〈1, ξ〉 is
Taj 〈1, ξ〉 = 〈1, ξ〉 q
aja6
∏
1≤`≤5
` 6=j
1− aja`
1− qa−1
6 a−1
`
for j = 1, 2, . . . , 5.
Proof. Without loss of generality, it suffices to show that
Ta1〈1, ξ〉 = 〈1, ξ〉 q
a1a6
5∏
`=2
1− a1a`
1− qa−1
6 a−1
`
. (5.6)
Set J(a1, a2, a3, a4, a5, a6; ξ) := 〈1, ξ〉. Under the condition a1a2a3a4a5a6 = 1, we have
J(qa1, a2, a3, a4, a5, a6; ξ) = J(a1, a2, a3, a4, a5, qa6; ξ)
a2
6
a2
1
5∏
`=2
1− a1a`
1− a6a`
from Theorem 5.1 by setting j1 = 1 and j2 = 6. We now replace a6 by q−1a6 in the above
equation. Then, under the condition a1a2a3a4a5(q−1a6) = 1, we have
J
(
qa1, a2, a3, a4, a5, q
−1a6; ξ
)
= J(a1, a2, a3, a4, a5, a6; ξ)
q
a1a6
5∏
`=2
1− a1a`
1− qa−1
6 a−1
`
.
Since Ta1〈1, ξ〉 = J(qa1, a2, a3, a4, a5, q
−1a6; ξ) under this condition a6 = q(a1a2a3a4a5)−1, we
obtain (5.6), which completes the proof. �
10 M. Ito
Corollary 5.2. Suppose s = n+ 1 and the condition a2n+4 = q
a1a2···a2n+3
. Then the recurrence
relation for Gustafson’s sum 〈1, x〉G where x = (x1, . . . , xn) ∈ (C∗)n is given by
Taj 〈1, x〉G = 〈1, x〉G
q
aja2n+4
∏
1≤`≤2n+3
` 6=j
1− aja`
1− qa−1
` a−1
2n+4
for j = 1, 2, . . . , 2n+ 3.
Proof. Fix s = n+ 1. For the BC1-type Jackson integral, we first set
J(a1, a2, . . . , a2n+4;x) := det
(
〈χ(n−i), xj〉
)
1≤i,j≤n
,
where χ(i) is defined in (4.8), under no condition on a1, a2, . . . , a2n+4. By the definition of Φ,
we have
J(qa1, a2, . . . , a2n+4;x) = det
(
〈e1χ(n−i), xj〉
)
1≤i,j≤n
.
Let Q be the transition matrix from the basis {v1, v2, . . . , vn} to {χ(n−1), χ(n−2), . . . , χ(0)}, i.e.,
(χ(n−1), χ(n−2), . . . , χ(0)) = (v1, v2, . . . , vn)Q.
Under the condition a1a2 · · · a2n+4 = 1, from Theorem 5.1 with j1 = 1 and j2 = 2n+4, it follows
that
(〈e1v1, ξ〉, . . . , 〈e1vn, ξ〉) = (〈e2n+4v1, ξ〉, . . . , 〈e2n+4vn, ξ〉)M,
so that
(〈e1χ(n−1), ξ〉, . . . , 〈e1χ(0), ξ〉) = (〈e2n+4χ(n−1), ξ〉, . . . , 〈e2n+4χ(0), ξ〉)Q−1MQ.
This indicates that
det
(
〈e1χ(n−i), xj〉
)
1≤i,j≤n
= det
(
〈e2n+4χ(n−i), xj〉
)
1≤i,j≤n
detM.
From (5.1) and the above equation we have
J(qa1, a2, . . . , a2n+4;x) = J(a1, a2, . . . , qa2n+4;x)
(
a2n+4
a1
)n+1 2n+3∏
`=2
1− a`a1
1− a`a2n+4
,
under the condition a1a2 · · · a2n+4 = 1. We now replace a2n+4 by q−1a2n+4 in the above equation.
Then we have
J
(
qa1, a2, . . . , q
−1a2n+4;x
)
= J(a1, a2, . . . , a2n+4;x)
q
a1a2n+4
2n+3∏
`=2
1− a`a1
1− qa−1
` a−1
2n+4
,
under the condition a1a2 · · · (q−1a2n+4) = 1. Since
Ta1 det
(
〈χ(n−i), xj〉
)
1≤i,j≤n
= J
(
qa1, a2, . . . , q
−1a2n+4;x
)
if a1a2 · · · a2n+4 = q, we have
Ta1 det
(
〈χ(n−i), xj〉
)
1≤i,j≤n
= det
(
〈χ(n−i), xj〉
)
1≤i,j≤n
q
a1a2n+4
2n+3∏
`=2
1− a`a1
1− qa−1
` a−1
2n+4
.
A First Order q-Difference System for the BC1-Type Jackson Integral 11
On the other hand, if x = (x1, . . . , xn) ∈ (C∗)n, then, by definition we have
det
(
〈χ(n−i), xj〉
)
1≤i,j≤n
= 〈1, x〉G,
which is also referred to in [14]. Therefore, under the condition a2n+4 = q(a1a2 · · · a2n+3)−1 we
obtain
Ta1〈1, x〉G = 〈1, x〉G
q
a1a2n+4
2n+3∏
`=2
1− a`a1
1− qa−1
` a−1
2n+4
.
Since the same argument holds for parameters a2, . . . , a2n+3, we can conclude Corollary 5.2. �
Remark 5.1. If we take ξ = ai, i = 1, . . . , 6, and add the terminating condition a1a2 = q−N ,
N = 1, 2, . . . , to the assumptions of Corollary 5.1, then the finite product expression of 〈1, ξ〉,
which is equivalent to Jackson’s formula for terminating 8φ7 series [8, p. 43, equation (2.6.2)],
is obtained from finite repeated use of Corollary 5.1. In the same way, if we take a suitable x
and add the terminating condition to the assumptions of Corollary 5.2, then the finite product
expression of 〈1, x〉G, which is equivalent to the Jackson type formula for terminating multiple 8φ7
series (see [7, Theorem 4] or [6, p. 231, equation (4.4)], for instance), is obtained from finite
repeated use of Corollary 5.2.
5.2 Application
The aim of this subsection is to give a simple proof of the following propositions proved by
Nassrallah and Rahman [16] and Gustafson [9].
Proposition 5.1 (Nassrallah–Rahman). Assume |ai| < 1 for 1 ≤ i ≤ 5. If a6 = q
a1a2a3a4a5
,
then
1
2π
√
−1
∫
T
(
qa−1
6 z
)
∞
(
qa−1
6 z−1
)
∞
(
z2
)
∞
(
z−2
)
∞
5∏
i=1
(aiz)∞
(
aiz−1
)
∞
dz
z
=
2
5∏
k=1
(
qa−1
6 a−1
k
)
∞
(q)∞
∏
1≤i<j≤5
(aiaj)∞
, (5.7)
where T is the unit circle taken in the positive direction.
Proof. We denote the left-hand side of (5.7) by I(a1, a2, a3, a4, a5). By residue calculation,
I(a1, a2, a3, a4, a5) =
5∑
k=1
∞∑
ν=0
Res
z=akqν
θ
(
qa−1
6 z−1
)
θ
(
z−2
)
z
5∏
m=1
θ
(
amz−1
) z(1− z2)
6∏
m=1
(
qa−1
m z
)
∞
(amz)∞
dz
z
(5.8)
=
5∑
k=1
Res
z=ak
θ
(
qa−1
6 z−1
)
θ
(
z−2
)
z
5∏
m=1
θ
(
amz−1
) dz
z
∫ ak∞
0
z(1− z2)
6∏
m=1
(
qa−1
m z
)
∞
(amz)∞
dqz
z
=
5∑
k=1
Rk〈1, ak〉, (5.9)
where
Rk := Res
z=ak
θ
(
qa−1
6 z−1
)
θ
(
z−2
)
z
5∏
m=1
θ
(
amz−1
) dz
z
=
θ
(
qa−1
6 a−1
k
)
θ
(
a−2
k
)
(q)2∞ak
∏
1≤m≤5
m6=k
θ
(
ama
−1
k
) ,
12 M. Ito
whose recurrence relation is
TajRk =
(
q−1aja6
)
Rk (5.10)
for 1 ≤ j, k ≤ 5, which is obtained using (2.2). From (5.9), (5.10) and Corollary 5.1, we obtain
the recurrence relation for I(a1, a2, a3, a4, a5) as
TajI(a1, a2, a3, a4, a5) = I(a1, a2, a3, a4, a5)
∏
1≤`≤5
` 6=j
1− aja`
1− qa−1
6 a−1
`
.
By repeated use of the above relation, we obtain
I(a1, a2, a3, a4, a5) =
5∏
k=1
(
qa−1
6 a−1
k
)
2N∏
1≤i<j≤5
(aiaj)2N
I
(
qNa1, q
Na2, q
Na3, q
Na4, q
Na5
)
=
5∏
k=1
(
qa−1
6 a−1
k
)
∞∏
1≤i<j≤5
(aiaj)∞
lim
N→∞
I
(
qNa1, q
Na2, q
Na3, q
Na4, q
Na5
)
and
lim
N→∞
I
(
qNa1, q
Na2, q
Na3, q
Na4, q
Na5
)
=
1
2π
√
−1
∫
T
(
z2
)
∞
(
z−2
)
∞
dz
z
=
2
(q)∞
.
This completes the proof. �
Remark 5.2. Strictly speaking, the residue calculation (5.8) requires that
Iε :=
1
2π
√
−1
∫
|z|=ε
(
qa−1
6 z
)
∞
(
qa−1
6 z−1
)
∞
(
z2
)
∞
(
z−2
)
∞
5∏
i=1
(aiz)∞
(
aiz−1
)
∞
dz
z
→ 0 if ε→ 0, (5.11)
which can be shown in the following way. We first take ε = qNε′ for ε′ > 0 and positive
integer N . If we put
F (z) :=
(
qa−1
6 z
)
∞
(
qa−1
6 z−1
)
∞
(
z2
)
∞
(
z−2
)
∞
5∏
i=1
(aiz)∞
(
aiz−1
)
∞
,
then we have F (z) = zG1(z)G2(z), where
G1(z) =
θ
(
qa−1
6 z−1
)
θ
(
z−2
)
z
5∏
i=1
θ
(
aiz−1
) , G2(z) =
(
1− z2
) 6∏
i=1
(
qa−1
i z
)
∞
(aiz)∞
.
Since G1(z) is a continuous function on the compact set |z| = ε′ and is invariant under the q-shift
z → qz under the condition a6 = q(a1a2a3a4a5)−1, |G1(z)| is bounded on |z| = qNε′. |G2(z)| is
also bounded because G2(z) → 1 if z → 0. Thus there exists C > 0 such that |F (z)| < C|z|. If
we put z = εe2π
√
−1τ , then
|Iε| <
∫ 1
0
|F (εe2π
√
−1τ )|dτ < C
∫ 1
0
|εe2π
√
−1τ |dτ = Cε→ 0, ε→ 0,
which proves (5.11).
A First Order q-Difference System for the BC1-Type Jackson Integral 13
Proposition 5.2 (Gustafson [9]). Assume |ai| < 1 for 1 ≤ i ≤ 2n+3. If a2n+4 = q
a1a2···a2n+3
,
then (
1
2π
√
−1
)n ∫
Tn
n∏
i=1
(
qa−1
2n+4zi
)
∞
(
qa−1
2n+4z
−1
i
)
∞
(
z2
i
)
∞
(
z−2
i
)
∞
2n+3∏
k=1
(akzi)∞
(
akz
−1
i
)
∞
×
∏
1≤i<j≤n
(zizj)∞
(
ziz
−1
j
)
∞
(
z−1
i zj
)
∞
(
z−1
i z−1
j
)
∞
dz1
z1
∧ · · · ∧ dzn
zn
=
2nn!
2n+3∏
k=1
(
qa−1
2s+4a
−1
k
)
∞
(q)n
∞
∏
1≤i<j≤2s+3
(aiaj)∞
, (5.12)
where Tn is the n-fold direct product of the unit circle traversed in the positive direction.
The proof below is based on an idea using residue computation due to Gustafson [10], which
is done for the case of the hypergeometric integral under no balancing condition. Here we will
show that his residue method is still effective even for the integral under the balancing condition
a1a2 · · · a2n+4 = q. In particular, this is different from his proof in [9].
Proof. Let L be the set of indices defined by
L := {λ = (λ1, . . . , λn); 1 ≤ λ1 < λ2 < · · · < λn ≤ 2n+ 3}.
Set a(µ) := (aµ1 , . . . , aµn) ∈ (C∗)n for µ = (µ1, . . . , µn) ∈ L. We denote the left-hand side
of (5.12) by I(a1, a2, . . . , a2n+3). By residue calculation, we have
I(a1, a2, . . . , a2n+3) =
∑
µ∈L
Rµ〈1, a(µ)〉G, (5.13)
where the coefficients Rµ, µ ∈ L, are
Rµ := Res
z1=aµ1···
zn=aµn
n∏
i=1
θ
(
qa−1
2n+4z
−1
i
)
θ
(
z−2
i
)
zi
2n+3∏
m=1
θ
(
amz
−1
i
) ∏
1≤j<k≤n
θ
(
z−1
j zk
)
θ
(
z−1
j z−1
k
)
dz1
z1
∧ · · · ∧ dzn
zn
.
The recurrence relation for Rµ is
TajRµ =
(
q−1aja2n+4
)
Rµ. (5.14)
From (5.13), (5.14) and Corollary 5.2, we obtain the recurrence relation for I(a1, a2, . . . , a2n+3)
as
TajI(a1, a2, . . . , a2n+3) = I(a1, a2, . . . , a2n+3)
∏
1≤`≤2n+3
` 6=j
1− aja`
1− qa−1
2n+4a
−1
`
.
By repeated use of the above relation, we obtain
I(a1, a2, . . . , a2n+3) =
2n+3∏
k=1
(
qa−1
2n+4a
−1
k
)
2N∏
1≤i<j≤2n+3
(aiaj)2N
I
(
qNa1, q
Na2, . . . , q
Na2n+3
)
14 M. Ito
=
2n+3∏
k=1
(
qa−1
2n+4a
−1
k
)
∞∏
1≤i<j≤2n+3
(aiaj)∞
lim
N→∞
I
(
qNa1, q
Na2, . . . , q
Na2n+3
)
and
lim
N→∞
I
(
qNa1, q
Na2, . . . , q
Na2n+3
)
=
(
1
2π
√
−1
)n ∫
Tn
n∏
i=1
(
z2
i
)
∞
(
z−2
i
)
∞
×
∏
1≤i<j≤n
(zizj)∞
(
ziz
−1
j
)
∞
(
z−1
i zj
)
∞
(
z−1
i z−1
j
)
∞
dz1
z1
∧ · · · ∧ dzn
zn
=
2nn!
(q)n
∞
.
This completes the proof. �
References
[1] Aomoto K., A normal form of a holonomic q-difference system and its application to BC1-type, Int. J. Pure
Appl. Math. 50 (2009), 85–95.
[2] Aomoto K., Ito M., On the structure of Jackson integrals of BCn type and holonomic q-difference equations,
Proc. Japan Acad. Ser. A Math. Sci. 81 (2005), 145–150.
[3] Aomoto K., Ito M., Structure of Jackson integrals of BCn type, Tokyo J. Math. 31 (2008), 449–477.
[4] Aomoto K., Ito M., BCn-type Jackson integral generalized from Gustafson’s Cn-type sum, J. Difference
Equ. Appl. 14 (2008), 1059–1097.
[5] Aomoto K., Ito M., A determinant formula for a holonomic q-difference system associated with Jackson
integrals of type BCn, Adv. Math., to appear, doi:10.1016/j.aim.2009.02.003.
[6] van Diejen J.F., Spiridonov V.P., Modular hypergeometric residue sums of elliptic Selberg integrals, Lett.
Math. Phys. 58 (2001), 223–238.
[7] Denis R.Y., Gustafson R.A., An SU(n) q-beta integral transformation and multiple hypergeometric series
identities, SIAM J. Math. Anal. 23 (1992), 552–561.
[8] Gasper G., Rahman M., Basic hypergeometric series, 2nd ed., Encyclopedia of Mathematics and its Appli-
cations, Vol. 96, Cambridge University Press, Cambridge, 2004.
[9] Gustafson R.A., Some q-beta and Mellin–Barnes integrals with many parameters associated to the classical
groups, SIAM J. Math. Anal. 23 (1992), 525–551.
[10] Gustafson R.A., Some q-beta and Mellin–Barnes integrals on compact Lie groups and Lie algebras, Trans.
Amer. Math. Soc. 341 (1994), 69–119.
[11] Ito M., q-difference shift for a BCn type Jackson integral arising from ‘elementary’ symmetric polynomials,
Adv. Math. 204 (2006), 619–646.
[12] Ito M., Another proof of Gustafson’s Cn-type summation formula via ‘elementary’ symmetric polynomials,
Publ. Res. Inst. Math. Sci. 42 (2006), 523–549.
[13] Ito M., A multiple generalization of Slater’s transformation formula for a very-well-poised-balanced 2rψ2r
series, Q. J. Math. 59 (2008), 221–235.
[14] Ito M., Okada S., An application of Cauchy–Sylvester’s theorem on compound determinants to a BCn-type
Jackson integral, in Proceedings of the Conference on Partitions, q-Series and Modular Forms (University
of Florida, March 12–16, 2008), to appear.
[15] Ito M., Sanada Y., On the Sears–Slater basic hypergeometric transformations, Ramanujan J. 17 (2008),
245–257.
[16] Nassrallah B., Rahman M., Projection formulas, a reproducing kernel and a generating function for q-Wilson
polynomials, SIAM J. Math. Anal. 16 (1985), 186–197.
[17] Rains E.M., Spiridonov V.P., Determinants of elliptic hypergeometric integrals, Funct. Anal. Appl., to
appear, arXiv:0712.4253.
http://arxiv.org/abs/0712.4253
1 Introduction
2 BC_1-type Jackson integral
3 Key equation
4 The case a_1a_2 ... a_{2s+2} \not = 1
4.1 q-difference equation
4.2 Application
5 The case a_1a_2 ... a_{2s+2} = 1
5.1 Reflection equation
5.2 Application
References
|
| id | nasplib_isofts_kiev_ua-123456789-149165 |
| institution | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| issn | 1815-0659 |
| language | English |
| last_indexed | 2025-11-24T12:51:25Z |
| publishDate | 2009 |
| publisher | Інститут математики НАН України |
| record_format | dspace |
| spelling | Ito, M. 2019-02-19T17:53:29Z 2019-02-19T17:53:29Z 2009 A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications / M. Ito // Symmetry, Integrability and Geometry: Methods and Applications. — 2009. — Т. 5. — Бібліогр.: 17 назв. — англ. 1815-0659 2000 Mathematics Subject Classification: 33D15; 33D67; 39A13 https://nasplib.isofts.kiev.ua/handle/123456789/149165 We present an explicit expression for the q-difference system, which the BC1-type Jackson integral (q-series) satisfies, as first order simultaneous q-difference equations with a concrete basis. As an application, we give a simple proof for the hypergeometric summation formula introduced by Gustafson and the product formula of the q-integral introduced by Nassrallah-Rahman and Gustafson. This paper is a contribution to the Proceedings of the Workshop “Elliptic Integrable Systems, Isomonodromy Problems, and Hypergeometric Functions” (July 21–25, 2008, MPIM, Bonn, Germany). en Інститут математики НАН України Symmetry, Integrability and Geometry: Methods and Applications A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications Article published earlier |
| spellingShingle | A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications Ito, M. |
| title | A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications |
| title_full | A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications |
| title_fullStr | A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications |
| title_full_unstemmed | A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications |
| title_short | A First Order q-Difference System for the BC₁-Type Jackson Integral and Its Applications |
| title_sort | first order q-difference system for the bc₁-type jackson integral and its applications |
| url | https://nasplib.isofts.kiev.ua/handle/123456789/149165 |
| work_keys_str_mv | AT itom afirstorderqdifferencesystemforthebc1typejacksonintegralanditsapplications AT itom firstorderqdifferencesystemforthebc1typejacksonintegralanditsapplications |