
426
THEORY AND APPLICATIONS OF SPECIAL FUNCTIONS
We have seen that the difference equation (DFA) (x)
=
E(A)FA(x) for
generic x
E
i
R
is valid under the extra assumption
a,
P
E
2Z>o. For
a,
p
E
this difference equation has been proved by Ismail and Rah-
man (Ismail and Rahman, 1991) using the explicit expression of FA(x)
as very-well-poised
8(P7
series.
0
Using the explicit expressions of FA(x), the conditions on x,
X
and the
four parameters
a,
p,
p,
a
can be relaxed by meromorphic continuation.
The resulting function
FA
is a meromorphic, ~-'-~eriodic eigenfunction
of the Askey-Wilson second order difference operator
D
=
D~*~*~,~
with
eigenvalue
E
(A)
=
E
(A;
P)
.
Remark
4.7.
Some special cases of the matrix coeficients
(PA
were ex-
plicitly expressed
in
terms of very-well-poised
S(P7
series
in
(Koelink and
Stokman, 2001b) using the realization of the representation
.rrx
on the
representation space l2(2). In this approach the basic hypergeometric se-
ries manipulations are much harder, since one needs a highly nontrivial
evaluation of a non-symmetric Poisson type kernel involving nontermi-
nating 2pl-series which is due to Rahman, see the appendix of (Koelink
and Stokman, 2001 b) and (Koelink and Rosengren, 2002).
Remark
4.8.
The explicit
897
expression
(4.13)
of
FA(x)
and its an-
alytic continuation was named the Askey- Wilson function in (Koelink
and Stokman, 2001a). Suslov (Suslov, 1997), (Suslov, 2002) established
Fourier-Bessel type orthogonality relations for the Aslcey- Wilson func-
tion. Koelink and the author (Koelink and Stokman, 2001b), (Koelink
and Stokman, 2OOla) defined a generalized Fourier transform involv-
ing the Askey- Wilson function as the integral kernel, and established its
Plancherel and inversion formula. This transform, called the Askey-
Wilson function transform, arises as Fourier transform on the noncom-
pact quantum group
SUq
(1,l)
(see (Koelink and Stokman, ,2001 b)), and
may thus be seen as a natural analogue of the Jacobi function transform.
5.
The expansion formula and the elliptic cosine
kernel
In this section we still assume that
T
E
ilW>o, so
0
<
q
=
q,
=
exp(2ri~)
<
1. To keep contact with the conventions of the previous
section, we keep working in base
q2.