Applied Mathematics and Computation 219 (2013) 8790–8802
Contents lists available at SciVerse ScienceDirect
Applied Mathematics and Computation journal homepage: www.elsevier.com/locate/amc
Structure relations for the bivariate big q-Jacobi polynomials Stanisław Lewanowicz ⇑, Paweł Woz´ ny, Rafał Nowak Institute of Computer Science, University of Wrocław, ul. F. Joliot-Curie 15, 50-383 Wrocław, Poland
a r t i c l e
i n f o
Keywords: Bivariate big q-Jacobi polynomial Structure relation
a b s t r a c t We give structure relations for the orthogonal polynomials in two variables, defined by Lewanowicz and Woz´ ny [S. Lewanowicz, P. Woz´ ny, J. Comput. Appl. Math. 233 (2010) 1554–1561]
P n;k ðx; y; a; b; c; d; qÞ :¼ Pnk ðy; a; bcq
2kþ1
k
; dq ; qÞyk ðdq=y; qÞk P k ðx=y; c; b; d=y; qÞðn
P 0; k ¼ 0; 1; . . . ; nÞ where q 2 ð0; 1Þ; 0 < aq; bq; cq < 1; d < 0, and P m ðt; a; b; c; qÞ are univariate big q-Jacobi polynomials. We discuss in full detail the case of the polynomials Pn;k ðx; y; a; b; c; 0; qÞ, which are closely related to Dunkl’s bivariate (little) q-Jacobi polynomials [C.F. Dunkl, SIAM J. Algebra. Discr. Methods 1 (1980) 137–151]. Ó 2013 Elsevier Inc. All rights reserved.
1. Introduction One of the many characteristic properties of the classical orthogonal polynomials (Hermite, Laguerre, Jacobi, and Bessel polynomials) is the existence of the structure relation [1] (see also [19])
/ðxÞp0n ðxÞ ¼ an pnþ1 ðxÞ þ bn pn ðxÞ þ cn pn1 ðxÞ;
ð1:1Þ
where / is a fixed polynomial of degree at most 2. Maroni [20,21] introduced semi-classical orthogonal polynomials, satisfying the more general structure relation
/ðxÞp0n ðxÞ ¼
nþt X
anj pj ðxÞ;
j¼ns
where / is a polynomial, and s; t are independent of n. Later on, it was shown that relation (1.1) also characterizes discrete classical orthogonal polynomials (Hahn, Krawtchouk, Meixner, and Charlier polynomials) [12] as well as q-classical orthogonal polynomials of the q-Hahn class [22], with the derivative being replaced by the difference operator Df ðxÞ ¼ f ðx þ 1Þ f ðxÞ and the q-derivative operator Dq f ðxÞ ¼ ðf ðqxÞ f ðxÞÞ=ððq 1ÞxÞ, respectively. Recently, Koornwinder [17] gave an explicit structure relation for Askey–Wilson polynomials. For bivariate orthogonal polynomials, structure relations were studied in [9–11] (classical case) and [2,3,8] (semi-classical case). Structure relations for bivariate discrete classical orthogonal polynomials on the uniform lattice are obtained in [24]; the case of bivariate q-classical orthogonal polynomials was studied by Rodal [23] (see our discussion in Remark 1). The aim of the present paper is to go further in the study of the bivariate big q-Jacobi polynomials giving structure relations for them. These polynomials were introduced by the first two authors in [18] by ⇑ Corresponding author. E-mail addresses:
[email protected] (S. Lewanowicz),
[email protected] (P. Woz´ ny),
[email protected] (R. Nowak). 0096-3003/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amc.2013.02.059
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802 2kþ1
Pn;k ðx; y; a; b; c; d; qÞ :¼ Pnk ðy; a; bcq
k
; dq ; qÞyk ðdq=y; qÞk P k ðx=y; c; b; d=y; qÞ
ðn P 0; k ¼ 0; 1; . . . ; nÞ;
8791
ð1:2Þ
where q 2 ð0; 1Þ; 0 < aq; bq; cq < 1; d < 0, and
! qm ; abqmþ1 ; t ðm P 0Þ q; q aq; cq
Pm ðt; a; b; c; qÞ :¼ 3 /2
ð1:3Þ
are univariate big q-Jacobi polynomials (see, e.g. [4, Section 7.3], or [15, Section 14.5]). For the notation used, see Section 2. In [18], we gave some basic properties of polynomials (1.2). First, we showed that they form an orthogonal system with respect to the linear functional u defined by
hu; pi :¼
Z
aq
dq
Z
cqy
ðdq=y; x=ðcyÞ; x=d; y=a; y=d; qÞ1 pðx; yÞdq xdq y; yðd=ðcyÞ; cqy=d; x=y; bx=d; y; qÞ1
dq
where p is any two-variable polynomial. Second, we proved that the following three-term relations hold:
x Pn ¼ An;1 Pnþ1 þ Bn;1 Pn þ C n;1 Pn1 ; y Pn ¼ An;2 Pnþ1 þ Bn;2 Pn þ C n;2 Pn1 ;
ðn P 0Þ;
ð1:4Þ
where we used the notation Pn :¼ ½Pn;0 ; Pn;1 ; . . . ; P n;n T , and An;i ; Bn;i and C n;i are tridiagonal (i ¼ 1) or diagonal (i ¼ 2) matrices of appropriate dimensions. Last, we showed that for any n P 0, the polynomial vector Pn satisfies the linear second-order partial q-difference equation
þ
l11 Dq;x Dq1 ;x Pn þ l22 Dq;y Dq1 ;y Pn þ l12 Dq1 ;x Dq1 ;y Pn þ l12 Dq;x Dq;y Pn þ m1 Dq;x Pn þ m2 Dq;y Pn ¼ kn Pn ;
ð1:5Þ
with
l11 ðxÞ :¼ ðx dqÞðx acq2 Þ;
l22 ðyÞ :¼ ðy aqÞðy dqÞ;
þ
l12 ðx; yÞ :¼ q1 ðx dqÞðy aqÞ; l12 ðx; yÞ :¼ acq3 ðbx dÞðy 1Þ; n o 3 m1 ðxÞ :¼ ðabcq 1Þðx 1Þ ðacq2 1Þðdq 1Þ =ðq 1Þ; n o 3 m2 ðyÞ :¼ ðabcq 1Þðy 1Þ ðaq 1Þðdq 1Þ =ðq 1Þ; nþ2
kn :¼ ½nq q1n ðabcq
1Þ=ðq 1Þ;
where Dq1 ;x and Dq1 ;y are partial q-derivative operators (see Section 2), and we let
Dq1 ;z Pn :¼ ½Dq1 ;z Pn;0 ; Dq1 ;z Pn;1 ; . . . ; Dq1 ;z Pn;n T
ðz ¼ x; yÞ:
The main result of the present paper gives structure relations in the form
r1 Dq1 ;x Pn ¼ F n;1 Pnþ1 þ Gn;1 Pn þ Hn;1 Pn1 ; r2 Dq1 ;y Pn ¼ F n;2 Pnþ1 þ Gn;2 Pn þ Hn;2 Pn1 ; where r i are certain polynomials of total degree 2, and F n;i ; Gn;i and Hn;i (i ¼ 1; 2) are tridiagonal matrices of appropriate dimensions. A general result is given in Section 3, while in Section 4 we discuss in detail the case of polynomials (1.2) with d ¼ 0, which are closely related to Dunkl’s bivariate (little) q-Jacobi polynomials [6].
2. Preliminaries The hypergeometric series is defined by (see, e.g. [4, Section 2.1])
2F1
1 X a; b ðaÞk ðbÞk k z :¼ z; k!ðcÞ c k¼0
k
where a; b; c; z 2 C, and ðaÞk is the shifted factorial symbol, defined for any a 2 C by
ðaÞ0 :¼ 1;
ðaÞk :¼ aða þ 1Þ ða þ k 1Þ
The q-shifted factorial is defined for any a 2 C by
ða; qÞk :¼
k1 Y ð1 a qj Þ ðk P 0Þ: j¼0
Assuming that 0 < q < 1, we also put
ða; qÞ1 :¼
1 Y j¼0
ð1 a qj Þ:
ðk P 1Þ:
8792
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802
In the sequel, we make use of the convention
ða1 ; a2 ; . . . ; am ; qÞk :¼
m Y
aj ; q
k
ðk ¼ 0; 1; . . . or 1Þ:
j¼1
The q-binomial coefficient is given by
n k
:¼
q
ðq; qÞn : ðq; qÞk ðq; qÞnk
For a 2 C, we define the q-number ½aq by
½aq :¼
qa 1 : q1
The q-derivative operator is defined for q 2 C n f1g by
Dq f ðxÞ :¼
f ðqxÞ f ðxÞ ; ðq 1Þx
while the partial q-derivative operators are defined by
Dq;x f ðx; yÞ :¼
f ðqx; yÞ f ðx; yÞ ; ðq 1Þx
Dq;y f ðx; yÞ :¼
f ðx; qyÞ f ðx; yÞ : ðq 1Þy
The q-integral is defined by
Z
b
f ðxÞ dq x :¼
a
Z
b
f ðxÞdq x 0
Z
a
f ðxÞdq x; 0
where
Z
a
1 X f ðxÞdq x :¼ að1 qÞ f ðaqk Þ qk :
0
k¼0
The basic hypergeometric series is defined by (see, e.g. [4, Section 10.9])
r /s
1 X a1 ; . . . ; ar ða1 ; . . . ; ar ; qÞk q; z :¼ ðq; b1 ; . . . ; bs ; qÞ b ;...;b 1
s
k¼0
k
k
ð1Þ q
!1þsr
2
zk ;
k
where r; s 2 Zþ and a1 ; . . . ; ar ; b1 ; . . . ; bs ; z 2 C. We denote the set of all polynomials in one independent variable x by P1 and the set of all polynomials in two independent variables x and y by P2 . The q-gradient (rq ) and the q-divergence (divq ) operators are defined by
rq p :¼
Dq;x p Dq;y p
;
divq
p :¼ Dq;x p þ Dq;y q ðp; q 2 P2 Þ: q
For i ¼ 1; 2, a linear mapping u : Pi ! C is called a moment functional. For a one-variable moment functional u and r 2 P1 , we let Dq u, the q-derivative of u, and ru, multiplication of u by a polynomial, be those moment functionals defined by
hDq u; pi :¼ hu; Dq pi;
hru; pi :¼ hu; rpi ðp 2 P1 Þ:
In the two-variable case, we define the first partial q-derivatives of u, as moment functionals, by
hDq;z u; pi :¼ hu; Dq;z pi ðp 2 P2 ; z ¼ x; yÞ and we define multiplication ru, where r 2 P2 , by hru; pi :¼ hu; rpi (p 2 P2 ). Let A ¼ ½aij and B ¼ ½bij be polynomial matrices of dimension m n and m r, respectively. We define (see, e.g. [3,8,10]) the action of u over A by hu; Ai ¼ C, where C ¼ ½hu; aij i is the real matrix of dimension m n; the left multiplication of u by A as hAu; Bi ¼ hu; AT Bi. The distributional q-divergence operator is defined in the following way:
Dq;x p hdivq ðMuÞ; pi ¼ hMu; rq pi ¼ u; M T ðp 2 P2 Þ; Dq;y p where M is a polynomial matrix of dimension 2 n.
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802
8793
3. The general form of the structure relations Let u be the moment functional defined by
hu; pi :¼
Z
aq
dq
Z
cqy
Wðx; yÞpðx; yÞdq xdq y ðp 2 P2 Þ;
ð3:1Þ
dq
where
Wðx; yÞ Wðx; y; a; b; c; d; qÞ :¼
ðdq=y; x=ðcyÞ; x=d; y=a; y=d; qÞ1 : yðd=ðcyÞ; cqy=d; x=y; bx=d; y; qÞ1
ð3:2Þ
Remember that polynomials Pn;k ðx; yÞ ¼ Pn;k ðx; y; a; b; c; d; qÞ, defined by (1.2), are orthogonal with respect to this functional: hu; Pn;k Pml i ¼ Kn;k dnm dk;l for some constants Kn;k > 0 [18]. We need the following properties of the weight (3.2). Lemma 3.1. The weight function W satisfies the system of equations
Dq;x Dq;y
r1 W ¼ s1 W; r2 W ¼ s2 W
ð3:3Þ
ð3:4Þ
or, in the equivalent form,
Dq1 ;x Dq1 ;y
rþ1 W ¼ q s1 W; rþ2 W ¼ q s2 W;
ð3:5Þ
ð3:6Þ
where the bivariate polynomials
ri of total degree two, and si of total degree one, are given by
r1 ðx; yÞ :¼ ðx dqÞðx cqyÞ; rþ1 ðx; yÞ :¼ cq2 ðx yÞðbx dÞ; rþ ðx; yÞ r1 ðx; yÞ s1 ðx; yÞ :¼ 1 ; ðq 1Þx
r2 ðx; yÞ :¼ ðx yÞðy aqÞ; rþ2 ðx; yÞ :¼ aqðy 1Þðx cqyÞ; rþ ðx; yÞ r2 ðx; yÞ s2 ðx; yÞ :¼ 2 : ðq 1Þy
Proof. The general theory of bivariate q-hypergeometric orthogonal polynomials developed by Rodal [23, Section 5.5] yields equations
r1 ðqx; yÞ Wðqx; yÞ ¼ rþ1 ðx; yÞ Wðx; yÞ; r2 ðx; qyÞ Wðx; qyÞ ¼ rþ2 ðx; yÞ Wðx; yÞ [23, Eqs. (5.5.39) and (5.5.40)], which are equivalent to (3.3) and (3.4), respectively. Those two equations, together with identities
r1 W ¼ q1 Dq1 ;x rþ1 W ; Dq;y r2 W ¼ q1 Dq1 ;y rþ2 W Dq;x
[23, Eq. (5.5.49)], imply the remaining Eqs. (3.5) and (3.6). h Using the above results, we obtain the following properties of the functional (3.1). Lemma 3.2. The functional u defined by (3.1) satisfies the matrix q-Pearson equation
divq ðUþ uÞ ¼ WT u;
ð3:7Þ
or, equivalently,
divq1 ðU uÞ ¼ qWT u;
ð3:8Þ
where
U :¼
r1
0
0
r
2
;
W :¼
s1 : s2
8794
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802
Proof. Using the distributional operations defined in Section 2, we obtain
#
þ " r1 Dq;x p hDq;x rþ1 u ; pi T Dq;x p ¼ u; ¼ ; hdivq ðUþ uÞ; pi ¼ hUþ u; rq pi ¼ u; Uþ Dq;y p rþ2 Dq;y p hDq;y rþ2 u ; pi
s1 p hs1 u; pi ¼ : hWT u; pi ¼ h½s1 ; s2 u; pi ¼ u; hs2 u; pi s2 p Hence, Eq. (3.7) is equivalent to the system
Dq;x
Dq;y
ð3:9Þ
ð3:10Þ
rþ1 u ¼ s1 u; rþ2 u ¼ s2 u:
In a similar way, we show that Eq. (3.8) means
Dq1 ;x Dq1 ;y
ð3:11Þ
ð3:12Þ
r1 u ¼ qs1 u;
r2 u ¼ qs2 u:
We prove Eq. (3.9); the remaining equations can be proved in a similar way. We have
hDq;x
rþ1 u ; pi ¼ hu; rþ1 Dq;x pi ¼
Z
Z
aq
dq
cqy
Wðx; yÞ rþ1 ðx; yÞ Dq;x pðx; yÞdq xdq y:
dq
ð3:13Þ
Using the identity (see, e.g. [14, p. 74])
Z
b
Dq f ðxÞ gðxÞdq x ¼ f ðxÞgðx=qÞjba q1
a
Z a
b
f ðxÞ Dq1 gðxÞdq x;
we write the inner integral in (3.13) as
Z
1 Wðx=q; yÞ rþ1 ðx=q; yÞ pðx; yÞjx¼cqy x¼dq q
cqy
pðx; yÞ Dq1 ;x
dq
rþ1 ðx; yÞ Wðx; yÞ dq x:
Now, the first term vanishes, while the second one can be, by (3.5), written as
Z
cqy
pðx; yÞ s1 ðx; yÞ Wðx; yÞdq x; dq
so that we obtain
hDq;x
rþ1 u ; pi ¼
Z
aq dq
Z
cqy
pðx; yÞ s1 ðx; yÞ Wðx; yÞdq xdq y ¼ hu; s1 pi ¼ hs1 u; pi:
dq
Hence, Eq. (3.9) follows. h Now we can prove the main result of this section. We introduce the following column vector notation:
Pn :¼ ½P n;0 ; Pn;1 ; . . . ; Pn;n T ;
ð3:14Þ
Dq1 ;z Pn :¼ ½Dq1 ;z Pn;0 ; Dq1 ;z Pn;1 ; . . . ; Dq1 ;z Pn;n T
ðz ¼ x; yÞ;
where P n;k :¼ Pn;k ðx; y; a; b; c; d; qÞ. Theorem 3.3. For n P 0, there exist unique tridiagonal matrices ðn þ 1Þ ðn þ 2Þ; ðn þ 1Þ ðn þ 1Þ and ðn þ 1Þ n, respectively, of the form
2
F n;i
f0;0 6f 6 1;0 6 6 :¼ 6 6 6 4 2
Gn;i
g 0;0
6g 6 1;0 6 6 :¼ 6 6 6 4
f1;2 .. .
..
fn1;n2
fn1;n1
fn1;n
fn;n1
fn;n
.
H n;i
(i ¼ 1; 2)
of
the
size
7 7 7 7 7; 7 7 0 5
ð3:15Þ
fn;nþ1 3
g 0;1 g 1;1 .. .
and
3
f0;1 f1;1 .. .
F n;i ; Gn;i
g 1;2 .. .
..
g n1;n2
g n1;n1
.
g n;n1
7 7 7 7 7; 7 7 g n1;n 5 g n;n
ð3:16Þ
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802
2
Hn;i
h0;0 6h 6 1;0 6 6 :¼ 6 6 6 4
8795
3
h0;1 h1;1 .. .
h1;2 .. . hn1;n2
7 7 7 7 .. ; . 7 7 7 hn1;n1 5 hn;n1
ð3:17Þ
such that the following structure relations hold:
r1 Dq1 ;x Pn ¼ F n;1 Pnþ1 þ Gn;1 Pn þ Hn;1 Pn1 ; r2 Dq1 ;y Pn ¼ F n;2 Pnþ1 þ Gn;2 Pn þ Hn;2 Pn1 ;
ð3:18Þ ð3:19Þ
where we define P1 :¼ 0. Proof. We prove the ‘‘plus’’ identity of (3.18) (the remaining identities can be proved in a similar way). Obviously, we have
rþ1 Dq;x Pn;k ¼
nþ1 X l X sþl;m Pl;m l¼0 m¼0
for some coefficients sþ l;m . A. We show that sþ ¼ 0 for l < n 1. From the orthogonality of fP n;k g with respect to u, we find that l;m
sþl;m ¼
hu; rþ1 Dq;x Pn;k Pl;m i hu; P2l;m i
:
Using the rule (see, e.g. [14, p. 3])
Dq ðf ðxÞgðxÞÞ ¼ f ðxÞDq gðxÞ þ gðqxÞDq f ðxÞ and (3.9), we obtain
hu; P2l;m i sþl;m ¼ hu; rþ1 Dq;x Pn;k Pl;m i ¼ ¼ ¼
ð3:20Þ
h þ1 u; Dq;x ðP n;k P l;m Þ Pn;k Dq;x P l;m i hDq;x ð þ1 uÞ; P n;k P l;m i hu; þ1 Pn;k Dq;x P l;m i h 1 u; Pn;k P l;m i hu; þ1 P n;k Dq;x P l;m i
r
r
r
s
r
ð3:21Þ
with both terms vanishing when l < n 1. Here P l;m ðx; yÞ ¼ P l;m ðx=q; yÞ. B. We show that sþ ¼ 0 for m R fk 1; k; k þ 1g. First, observe that equation l;m
Wðx; y; a; b; c; d; qÞ ¼ wðx=y; c; b; d=y; qÞ v ðy; a; c; d; qÞ
ð3:22Þ
holds, where we use the notation of (3.2), and
wðx; a; b; c; qÞ :¼
ðx=a; x=c; qÞ1 ; ðx; bx=c; qÞ1
v ðy; k; l; m; qÞ :¼
ðmq=y; y=k; y=m; qÞ1 : y ðm=ðlyÞ; lqy=m; y; qÞ1
Let us recall that the univariate big q-Jacobi polynomials Pm ðt; a; b; c; qÞ (cf. (1.3)) are orthogonal with respect to the functional
hv ; pi :¼
Z
aq
wðx; a; b; cÞpðxÞdq x ðp 2 P1 Þ:
ð3:23Þ
cq
Also, we need the formula kþ1
Dq;x P n;k ðx; y; a; b; c; d; qÞ :¼
q1k ½kq ð1 bcq Þ k 2kþ1 k ; dq ; qÞPk1 ðqx=y; cq; bq; dq=y; qÞ: y ðdq=y; qÞk P nk ðy; a; bcq ð1 cqÞð1 dq=yÞ
B1. We show that sþ ¼ 0 for m > k þ 1. Using (3.22) and (3.24) in the second member of (3.20), we obtain l;m
ð3:24Þ
8796
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802
hu; P 2l;m i sþl;m ¼ hu; rþ1 Dq;x Pn;k Pl;m i Z aq Z cqy Wðx; y; a; b; c; d; qÞ rþ1 ðx; yÞ Dq;x Pn;k ðx; y; a; b; c; d; qÞPl;m ðx; y; a; b; c; d; qÞdq xdq y ¼ dq
¼
Z
dq
kþ1
aq
v ðy; a; c; d; qÞ
dq 2kþ1
Pnk ðy; a; bcq
q1k ½kq ð1 bcq Þ kþm y ðdq=y; qÞk ðdq=y; qÞj ð1 cqÞð1 dq=yÞ k
2jþ1
; dq ; qÞP lm ðy; a; bcq
j
; dq ; qÞ Ik;m ðyÞdq y;
with
Ik;m ðyÞ :¼
Z
cqy
wðx=y; c; b; d=y; qÞ rþ1 ðx; yÞ Pk1 ðqx=y; cq; bq; dq=y; qÞP m ðx=y; c; b; d=y; qÞdq x
dq
¼y
Z
cq
wðt; c; b; d=y; qÞ Q k ðt; yÞ Pm ðt; c; b; d=y; qÞdq t; dq=y
where Q k ðt; yÞ :¼ rþ 1 ðyt; yÞ P k1 ðqt; cq; bq; dq=y; qÞ is a polynomial in t of degree k þ 1. Notice that the last integral is the result of action of the functional (3.23) with a ¼ c; b ¼ b; c ¼ d=y on the product pðtÞ :¼ Q k ðt; yÞ P m ðt; c; b; d=y; qÞ; y being a parameter. Hence, Ik;m ðyÞ as well as sþ l;m vanishes for m > k þ 1. B2. In the remaining part of the proof, we use Eq. (3.21),
hu; P 2l;m i sþl;m ¼ hs1 u; P n;k P l;m i hu; rþ1 Pn;k Dq;x P l;m i; to show that sþ ¼ 0 for m < k 1: l;m First, by reversing the roles of Pn;k and Pl;m in the argument of the part B1 of the proof, we easily conclude that
hu; rþ1 Pn;k Dq;x P l;m i ¼ 0 for m < k 1: Next, we obtain
hs1 u; Pn;k P l;m i ¼
Z
aq
dq
v ðy; a; c; d; qÞykþm ðdq=y; qÞk ðdq=y; qÞj Pnk ðy; a; bcq2kþ1 ; dqk ; qÞPlm ðy; a; bcq2jþ1 ; dqj ; qÞ Jk;m ðyÞdq y;
where
J k;m ðyÞ :¼ y
Z
cq
wðt; c; b; d=y; qÞ Rm ðt; yÞ P k ðt; c; b; d=y; qÞdq t;
dq=y
with Rm ðt; yÞ :¼ s1 ðyt; yÞ P m ðt=q; c; b; d=y; qÞ being a polynomial in t of degree m þ 1. Obviously, J k;m ðyÞ as well as hs1 u; Pn;k P l;m i vanishes for m < k 1. Summing up, we have shown that the following identity holds:
rþ1 Dq;x Pn;k ¼
nþ1 X kþ1 X
sþl;m P l;m ;
ð3:25Þ
l¼n1m¼k1
which is equivalent to the ‘‘plus’’ identity of (3.18) with the obvious identification
fk;m ¼ sþnþ1;m ðn; kÞ;
g k;m ¼ sþn;m ðn; kÞ;
hk;m ¼ sþn1;m ðn; kÞ:
Remark 1. Notice that in [23, Thm 5.5.2], the structure relations are given in a weaker form
arþ1 Dq;x Pn drþ2 Dq;y Pn ¼ W n Pnþ1 þ Y n Pn þ Z n Pn1 ; ~ n Pnþ1 þ Y~ n Pn þ Z~ n Pn1 ; ar D 1 Pn dr D 1 Pn ¼ W 1
q
;x
2
q
;y
ð3:26Þ ð3:27Þ
~ n; Y ~ n and Z ~ n are matrices of appropriate dimensions. Obviously, such identities can be easily obtained where W n ; Y n ; Z n ; W using (3.18) and (3.19).
4. Special case: bivariate little q-Jacobi polynomials Notice that
Pn;k ðx; y; a; b; c; 0; qÞ ¼ cn;k p n;k ðx; y; a; b; cj qÞ with
ð4:1Þ
8797
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802 k nk þ þn 2 2
n nk k
cn;k :¼ ð1Þ a
c q
2kþ2 ;q ðbq; qÞk bcq nk ; ðcq; qÞk ðaq; qÞnk
where the polynomials 2kþ1
p n;k ðx; y; a; b; cj qÞ :¼ p nk ðy; bcq
; ajqÞ yk p k ðx=y; b; cjqÞ
are closely related to Dunkl’s bivariate p m ðx; a; bj qÞ :¼ pm ðx=ðbqÞ; a; bj qÞ, where mþ1
qm ; abq aq
pm ðx; a; bj qÞ :¼ 2 /1
(little)
ð4:2Þ
q-Jacobi
polynomials
[6].
Here
we
use
! q; qx
the
notation
ð4:3Þ
are little q-Jacobi polynomials of one variable (see, e.g. [13, p. 182], or [15, Section 14.12]). From (4.2) and (4.3), the following power representation can be deduced:
p n;k ðx; y; a; b; cj qÞ ¼
k X ni X
bi;j ðn; kÞ xi yj ;
ð4:4Þ
i¼0 j¼ki
where
kþ1 nþkþ2 qk ; bcq ; q qkn ; abcq ;q i
iþjk : bi;j ðn; kÞ :¼ 2kþ2 ajþik ci ðq; bq; qÞi q; bcq ;q
ð4:5Þ
iþjk
The following inverse representation:
xi y j ¼
n minði;kÞ X X k¼0
ak;l ði; jÞ p k;l ðx; y; a; b; cj qÞ;
ð4:6Þ
l¼0
where n :¼ i þ j and k
l kl
ak;l ði; jÞ :¼ ð1Þ q
þ
2
2
ð1 abcq2kþ2 Þðbq; qÞi bcqiþlþ2 ; q nl j iþjl i i
ðaqÞ ðcqÞ : lþ1 kþlþ2 l q kl q bcq ; q abcq ;q l
ð4:7Þ
nlþ1
can be proved by induction on n, using the recurrence relations (1.4). Theorem 4.1. The following structure relations hold:
xðx yÞDq;x p n;k ðx; y; a; b; cj qÞ ¼
nþ1 kþ1 X X
sþl;m p l;m ðx; y; a; b; cj qÞ;
ð4:8Þ
l¼n1 m¼k1 nþ1 kþ1 X X
ðy 1Þðx cqyÞDq;y p n;k ðx; y; a; b; cj qÞ ¼
r þl;m p l;m ðx; y; a; b; cj qÞ;
ð4:9Þ
l¼n1 m¼k1
where kþ1
sþn1;k1 ¼ a2 q2n2kþ2 ½kq
ð1 cqk Þð1 bcq
2nþ1 abcq ;q
Þ
;
2
sþn1;k
¼
k
acqknþ3 ½kq
½k 1q ð1 bq Þ
½k þ 1q ð1 bq
kþ1
Þ
! 1 þ cq
kþnþ1 2k 2kþ2 1 bcq 1 bcq 1 abcq ! 2n nþkþ1 nk n k þ 1 ð1 abcq2n Þð1 bcqnþkþ2 Þ Þ kþn 2 ð1 abcq Þð1 bcq q ð1 bcq Þ ½n kq þ ; 2nþ1 2nþ2 2 2 1 abcq 1 abcq q q
nk
kþ1
kþ1
ð1 bq Þð1 bcq Þ
2kþ1 2kþ2 2 bcq ; bcq ;q q
2
1 0 0 1 nþkþ1 nþkþ2
½2q bcq ;q bcq ;q 2n nþk 2 2 A q bcq ; q A; @ð1 abcq Þ@ 2nþ1 2nþ2 2 1 abcq 1 abcq
sþn1;kþ1 ¼cq3k2nþ4 ½kq
8798
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802 kþ1
sþn;k1 ¼ a2 q2n2kþ2 ½kq ðq þ 1Þ 0 sþn;k ¼ acq2 ½kq @ 0 @
ð1 cqk Þð1 bcq 2nþ1
ð1 abcq
k ½k 1q 1 bq 2k
1 bcq
nþkþ1 ½n kq 1 bcq 2nþ1
1 abcq
kþ1 ½k þ 1q 1 bq
ð1 bq
Þ
;
1 1A þ cq
1
2kþ2
1 bcq nþkþ2 ½n k þ 1q 1 bcq
A;
2nþ3
1 abcq
kþ1
sþn;kþ1 ¼ cq2knþ2 ½kq ½n kq
Þ 2nþ3
Þð1 abcq
kþ1
nþkþ2
Þð1 bcq Þð1 abcq
2kþ1 2kþ2 bcq ; bcq ;q
1 0 nþkþ2 nþkþ1 bcq ; q bcq ; q Þ@ 2 2A ; 2nþ3 2nþ1 1 abcq 1 abcq
2
kþ1
sþnþ1;k1 ¼ a2 q2n2kþ3 ½kq
ð1 cqk Þð1 bcq
2nþ2 abcq ;q
Þ
;
2 kþ1
2kþ1
sþnþ1;k ¼ aqnkþ1 ½kq ð1 cqk cqkþ1 þ bcq
sþnþ1;kþ1
Þ
ð1 bcq
nþkþ2
Þð1 bcq
2k
2kþ2
ð1 bcq Þð1 bcq
nþkþ2
Þð1 abcq Þ
; 2nþ2 Þ abcq ;q 2
kþ1 kþ1 nþkþ2 nþkþ2 ð1 bq Þð1 bcq Þ bcq ; abcq ;q 2
¼ c qkþ1 ½kq 2kþ1 2kþ2 2nþ2 bcq ; bcq ; abcq ;q 2
and
rþn1;k1 ¼ abc qnþ2 ½kq
ð1 cqk Þð1 aqnkþ1 Þ
; 2nþ1 abcq ;q 2
2 n
bcq ð½kq ½nq Þ aqn qk
ðc þ 1Þ abcqnþkþ2 bcq2kþ1 1 rþn1;k ¼ 2k 2kþ2 2 nþ1 1 bcq abcq ;q 1 bcq 2
o 2 nþ2kþ2 2 2 nþkþ1 k nþ1 abc q þ acqnkþ1 ; þðb þ 1Þ cq ð1 þ qÞ acq þ ab c3 qnþ3kþ3 abc qnþkþ3 þ abcq
r þn1;kþ1 ¼
nk 2
2
kþ1
kþ1
bc q4knþ4 ð1 q2 Þð1 bq Þð1 bcq Þð1 abcq
2kþ1 2kþ2 2nþ1 a bcq ; bcq ; abcq ;q q
kþnþ2
Þ aqnk1 ; q 2
2
rþn;k1 ¼ abc qnþ2 ½kq
2nþ2 ð1 cqk Þ ð1 abcq Þ ð1 aqnkþ1 Þ ð1 aqnkþ2 Þ 2nþ1
ð1 abcq nþkþ1
rþn;k
2
¼acq
1 ½n kq ð1 bcq þ 2nþ1 aq 1 abcq
Þ
nþkþ2
2nþ3
2k
1 bcq
½n kq ð1 bcq
2nþ1
1 abcq
kþ1
kþ1
½k þ 1q ½n kq ð1 bq
!
2kþ2
k
1þ
Þ
1 bcq
q½kq ð1 bq Þ 1 bcq nþkþ2
2k
kþ1
kþ1
½k þ 1q ð1 bq 2kþ2
1 bcq
nþkþ2
Þð1 bcq Þð1 bq Þð1 abcq
2kþ1 2kþ2 bcq ; bcq ;q 2 ! kþ3þn kþ1þn ½n kq ð1 bcq Þ ½n k 1q ð1 bcq Þ 1 ; 2nþ3 2nþ1 aq 1 abcq 1 bcaq
rþn;kþ1 ¼ cq2knþ2 ½n kq
ð1 bcq
Þ
! Þ
1 abcq
½kq ½n k þ 1q ð1 bq Þ nþkþ1
acqnkþ1
;
Þ
½n k þ 1q ð1 bcq k
½n kq þ
2nþ3
Þð1 abcq
Þ
! Þ
;
;
8799
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802 nþkþ1
rþnþ1;k1 ¼ a2 bc q2nkþ4 ½kq
ð1 cqk Þð1 bcq
2nþ2 abcq ;q
Þ
;
2
nþkþ2
rþnþ1;k
nkþ2
¼ ac q
nþkþ2
ð1 bcq Þð1 abcq
2nþ2 abcq ;q
Þ
k
½n kq þ
½kq ½n k þ 1q ð1 bq Þ 1 bcq
2k
½k þ 1q ½n kq ð1 bq
kþ1
2kþ2
1 bcq
! Þ
;
2
kþ1
rþnþ1;kþ1
¼ ½n kq cq
kþ1
ð1 bq
kþ1 nþkþ2 nþkþ2 Þð1 bcq Þ bcq ;q abcq ;q 2 2
: 2kþ1 2kþ2 2nþ2 bcq ; bcq ; abcq ;q 2
Proof. Formulas (4.8) and (4.9) as well as (4.10) and (4.11), mentioned below, are obtained in a way, which we explain in case of the structure relation (4.8). Using the power representation (4.4) of p n;k ðx; y; a; b; cjqÞ and the q-differentiation rule Dq;x xi ¼ ½iq xi1 , we compute the coefficients ci;j in
xðx yÞDq;x p n;k ðx; y; a; b; cj qÞ ¼
kþ1 nkþ1 X X i¼0
ci;j xi ykþji :
j¼0
Then we use the inverse representation formula (4.6) to obtain
xðx yÞDq;x p n;k ðx; y; a; b; cj qÞ ¼
kþ1 nkþ1 X X i¼0
¼
kþj minði;lÞ X X
al;m ði; k þ j iÞp l;m ðx; y; a; b; c jqÞ
l¼0 m¼0
j¼0
kþ1 nkþ1 X X i¼0
ci;j ci;j
nþ1 X nþ1 X al;m ði; k þ j iÞp l;m ðx; y; a; b; cj qÞ: l¼0 m¼0
j¼0
Changing the order of summation gives
xðx
yÞDq;x p n;k ðx; y; a; b; cj qÞ
¼
nþ1 X nþ1 X kþ1 nkþ1 X X l¼0 m¼0
i¼0
! ci;j al;m ði; k þ j iÞ p l;m ðx; y; a; b; cj qÞ:
j¼0
By Theorem 3.3 (in particular, see (3.25)), we can reduce the range of summation of the first two sums, so that
xðx yÞDq;x p n;k ðx; y; a; b; cj qÞ ¼
nþ1 kþ1 X X
sþl;m p l;m ðx; y; a; b; cj qÞ;
l¼n1 m¼k1
where
sþl;m :¼
kþ1 nkþ1 X X i¼0
ci;j al;m ði; k þ j iÞ ¼
j¼0
kþ1 nkþ1 X X
ci;j al;m ði; k þ j iÞ:
i¼m j¼lk
In the last step, the range of summation is reduced thanks to the fact that the coefficients ak;l ði; jÞ vanish for l > i or k > i þ j (cf. (4.7)); the resulting expression is a sum of at most nine terms. Now, the coefficients sþ can be computed using a coml;m puter algebra system. Actually, we have used MapleTM system to produce formulas given in the theorem. h Remark 2. On the Web page http://www.ii.uni.wroc.pl/ pwo/programs.html, one can find a MapleTM worksheet to produce þ explicit expressions for the coefficients sþ l;m and r l;m in Eqs. (4.8) and (4.9) as well as the coefficients sl;m and r l;m in two other structure relations,
xðx cqyÞDq1 ;x p n;k ðx; y; a; b; cj qÞ ¼
nþ1 kþ1 X X
sl;m p l;m ðx; y; a; b; cj qÞ;
ð4:10Þ
l¼n1 m¼k1
ðx yÞðaq yÞDq1 ;y p n;k ðx; y; a; b; cj qÞ ¼
nþ1 kþ1 X X
r l;m p l;m ðx; y; a; b; cj qÞ:
l¼n1 m¼k1
ð4:11Þ
8800
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802
4.1. Limit form: bivariate Jacobi polynomials Notice that
lim p n;k ðx; y; qa ; qb ; qc j qÞ ¼ q!1
ða;b;cÞ
where P n;k
ða;b;cÞ
Pn;k
ð1Þk k!ðn kÞ! ðaþ1=2;bþ1=2;cþ1=2Þ P ð1 y; xÞ; ðb þ 1Þk ð2k þ b þ c þ 2Þnk n;k
ðx; yÞ are the triangle Jacobi polynomials (see [7, p. 86], or [16]), ð2kþbþc;a12Þ
ðx; yÞ :¼ Rnk
ðc12;b12Þ
ðxÞ ð1 xÞk Rk
y ; 1x
where a; b; c > 12, and
Rðml;mÞ ðtÞ :¼
m; m þ l þ m þ 1 ðl þ 1Þm 1 t 2F1 m! lþ1
is the mth shifted Jacobi polynomial in one variable [15, Section 9.8]. Corollary 4.2. The following structure relations hold:
9 nþ1 kþ1 X X > @ ða;b;cÞ Hl;m > ða;b;cÞ > xðx þ y 1Þ P n;k ðx; yÞ ¼ sl;m Pl;m ðx; yÞ; > > = @x H n;k l¼n1 m¼k1
ð4:12Þ
nþ1 kþ1 > X X > @ ða;b;cÞ Hl;m ða;b;cÞ > Pn;k ðx; yÞ ¼ rl;m Pl;m ðx; yÞ; > yðx þ y 1Þ > ; @y H n;k l¼n1 m¼k1
where
Hi;j :¼
ð1Þj j!ði jÞ! ; b þ 12 j ð2j þ d þ 1Þij
sn1;k1
k k þ c 12 n k þ a þ 12 ¼ ; 2n þ k 12 2
sn;k1 ¼
k k þ c 12 2k þ d a 32 ; 2n þ k 12 2n þ k þ 32
snþ1;k1 ¼
sn1;k
sn;k
k k þ c 12 ðn þ k þ dÞ ; 2n þ k þ 12 2
¼ kð2n þ a þ 2c þ k þ 1Þðk þ dÞ þ ðd 1Þ c þ 12 n þ k þ 12
ðk nÞ n k þ a 12 ; ð2k þ d 1Þð2k þ d þ 1Þ 2n þ k 12 2
( ) ðn kÞðn þ k þ dÞ ðn k þ 1Þðn þ k þ d þ 1Þ ¼ þ1 2n þ k 12 2n þ k þ 32 kðn k þ 1Þ k þ b 12 ðk þ 1Þðn kÞ k þ b þ 12 nkþ 2k þ d 1 2k þ d þ 1 ðk þ 1Þ k þ b þ 12 ðn kÞðn þ k þ dÞ k k þ b 12 þ 1 ; 2k þ d 1 2k þ d þ 1 2n þ k 12
snþ1;k ¼
ðn þ k þ d þ 1Þ n þ k þ k þ 12 kðn k þ 1Þ k þ b 12 ðk þ 1Þðn kÞ k þ b þ 12 ; n k þ 2k þ d 1 2k þ d þ 1 2n þ k þ 12 2
sn1;kþ1 ¼
sn;kþ1 ¼
ðn k 1Þ2 k þ b þ 12 ðk þ dÞ k þ n þ k þ 12 n k þ a 32 2 ; ð2k þ dÞ2 ð2k þ d þ 1Þ2 2n þ k 12 2
ðk nÞðk þ dÞðn þ k þ d þ 1Þ k þ b þ 12 n þ k þ k þ 12 2k þ k þ 32 n k þ a 12 ; ð2k þ dÞ2 ð2k þ d þ 1Þ2 2n þ k 12 2n þ k þ 32
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802
snþ1;kþ1 ¼
8801
ðk nÞ k þ b þ 12 ðk þ dÞðn þ k þ d þ 1Þ2 n þ k þ k þ 12 2 ð2k þ dÞ2 ð2k þ d þ 1Þ2 2n þ k þ 12 2
and
rn1;k1
k k þ c 12 ðk þ dÞ ¼ ; 2n þ k 12 2
rn;k1 ¼
rnþ1;k1 ¼
rn1;k ¼
rn;k ¼
2k k þ c 12 ðk þ dÞ ; 2n þ k 12 2n þ k þ 32
k k þ c 12 ðk þ dÞ ; 2n þ k þ 12 2
kðk þ dÞðc bÞðn kÞ n k þ a 12 ; ð2k þ d 1Þð2k þ d þ 1Þ 2n þ k 12 2
! ðc bÞkðk þ dÞ ðn kÞðn þ k þ dÞ ðn k þ 1Þðn þ k þ d þ 1Þ ; ð2k þ d 1Þð2k þ d þ 1Þ 2n þ k 12 2n þ k þ 32
rnþ1;k ¼
ðc bÞkðk þ dÞðn þ k þ d þ 1Þ n þ k þ k þ 12 ; ð2k þ d 1Þð2k þ d þ 1Þ 2n þ k þ 12 2
rn1;kþ1 ¼
rn;kþ1 ¼
ðn k 1Þ2 k k þ b þ 12 ðk þ dÞ n k þ a 32 2 ; ð2k þ dÞ2 ð2k þ d þ 1Þ2 2n þ k 12 2
2kðn kÞ k þ b þ 12 ðk þ dÞ n þ k þ k þ 12 ðn þ k þ d þ 1Þ n k þ a 12 ; ð2k þ dÞ2 ð2k þ d þ 1Þ2 2n þ k 12 2n þ k þ 32
rnþ1;kþ1 ¼
kðk þ b þ 12Þðk þ dÞðn þ k þ d þ 1Þ2 n þ k þ k þ 12 2 ð2k þ dÞ2 ð2k þ d þ 1Þ2 ð2n þ k þ 12 Þ2
;
with k :¼ a þ b þ c; d :¼ b þ c. Remark 3. Structure relations (4.12) for the triangle Jacobi polynomials are in agreement with general results on classical orthogonal polynomials in two variables [5,10]. To our best knowledge, however, they were not given in full detail before. Acknowledgments The authors thank the referee for the valuable critical remarks and for some references which had not been considered in the first version of the paper. References [1] W.A. Al-Salam, T.S. Chihara, Another characterization of the classical orthogonal polynomials, SIAM J. Math. Anal. 3 (1972) 65–70. [2] M. Álvarez de Morales, L. Fernández, T.E. Pérez, M.A. Piñar, A semiclassical perspective on multivariate orthogonal polynomials, J. Comput. Appl. Math. 214 (2008) 447–456. [3] M. Álvarez de Morales, L. Fernández, T.E. Pérez, M.A. Piñar, Semiclassical orthogonal polynomials in two variables, J. Comput. Appl. Math. 207 (2007) 323–330. [4] G.E. Andrews, R. Askey, R. Roy, Special Functions, Encyclopedia of Mathematics and its Applications, vol. 71, Cambridge University Press, Cambridge, 1999. [5] I. Area, E. Godoy, A. Ronveaux, A. Zarzo, Bivariate second-order linear partial differential equations and orthogonal polynomial solutions, J. Math. Anal. Appl. 387 (2012) 1188–1208. [6] C.F. Dunkl, Orthogonal polynomials in two variables of q-Hahn and q-Jacobi type, SIAM J. Algebra. Discr. Methods 1 (1980) 137–151. [7] C.F. Dunkl, Y. Xu, Orthogonal Polynomials of Several Variables, Encyclopedia of Mathematics and its Applications, vol. 81, Cambridge University Press, Cambridge, 2001. [8] L. Fernández, F. Marcellán, T.E. Pérez, M.A. Piñar, Recent trends on two variable orthogonal polynomials, in: Differential Algebra, Complex Analysis and Orthogonal Polynomials, Contemporary Mathematics, vol. 509, The American Mathematical Society, Providence, RI, 2010, pp. 59–86. [9] L. Fernández, T.E. Pérez, M.A. Piñar, Classical orthogonal polynomials in two variables: a matrix approach, Numer. Algorithms 39 (2005) 131–142. [10] L. Fernández, T.E. Pérez, M.A. Piñar, Weak classical orthogonal polynomials in two variables, J. Comput. Appl. Math. 178 (2005) 191–203. [11] L. Fernández, T.E. Pérez, M.A. Piñar, On Koornwinder classical orthogonal polynomials in two variables, J. Comput. Appl. Math. 236 (2012) 3817–3826. [12] A.G. Garcı`a, F. Marcellán, L. Salto, A distributional study of discrete classical orthogonal polynomials, J. Comput. Appl. Math. 57 (1995) 147–162.
8802
S. Lewanowicz et al. / Applied Mathematics and Computation 219 (2013) 8790–8802
[13] G. Gasper, M. Rahman, Basic hypergeometric series (with a foreword by Richard Askey), Encyclopedia of Mathematics and its Applications, second ed., vol. 96, Cambridge University Press, Cambridge, 2004. [14] V. Kac, P. Cheung, Quantum Calculus, Springer, New York, 2002. [15] R. Koekoek, P.A. Lesky, R.F. Swarttouw, Hypergeometric orthogonal polynomials and their q-analogues (with a foreword by Tom H. Koornwinder), Springer Monographs in Mathematics, Springer-Verlag, Berlin, 2010. [16] T.H. Koornwinder, Two-variable analogues of the classical orthogonal polynomials, in: R.A. Askey (Ed.), Theory and Application of Special Functions, Academic Press, New York, 1975, pp. 435–495. [17] T.H. Koornwinder, The structure relation for the Askey–Wilson polynomials, J. Comput. Appl. Math. 207 (2007) 214–226. [18] S. Lewanowicz, P. Woz´ ny, Two-variable orthogonal polynomials of big q-Jacobi type, J. Comput. Appl. Math. 233 (2010) 1554–1561. [19] F. Marcellán, A. Branquinho, J. Petronilho, Classical orhtogonal polynomials: a functional approach, Acta Appl. Math. 34 (1994) 283–303. [20] P. Maroni, Prolègoménes à l’étude des polynômes orthogonaux classiques, Ann. Mat. Pura Appl. 149 (1987) 165–184. [21] P. Maroni, Une théorie algébrique des polynômes orthogonaux: application aux polynômes orthogonaux semi-classiques, in: C. Brezinski, L. Gori, A. Ronveaux (Eds.), Orthogonal Polynomials and their Applications (Erice, 1990), IMACS Annals on Computing and Applied Mathematics, vol. 9, J.C. Baltzer AG, Basel, 1991, pp. 95–130. [22] J.C. Medem, R. Álvarez-Nodarse, F. Marcellán, On the q-polynomials: a distributional study, J. Comput. Appl. Math. 135 (2001) 157–196. [23] J.A. Rodal, Polinomios ortogonales en varias variables discretas, Ph.D Thesis, Departamento de Matemática Aplicada II, Universidade de Vigo, Vigo, 2008. [24] J. Rodal, I. Area, E. Godoy, Structure relations for monic orthogonal polynomials in two discrete variables, J. Math. Anal. Appl. 340 (2008) 825–844.