The Rahman Polynomials Are Bispectral
In a very recent paper, M. Rahman introduced a remarkable family of polynomials in two variables as the eigenfunctions of the transition matrix for a nontrivial Markov chain due to M. Hoare and M. Rahman. I indicate here that these polynomials are bispectral. This should be just one of the many rema...
Saved in:
| Published in: | Symmetry, Integrability and Geometry: Methods and Applications |
|---|---|
| Date: | 2007 |
| Main Author: | |
| Format: | Article |
| Language: | English |
| Published: |
Інститут математики НАН України
2007
|
| Online Access: | https://nasplib.isofts.kiev.ua/handle/123456789/147372 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Journal Title: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Cite this: | The Rahman Polynomials Are Bispectral / F.A. Grünbaum // Symmetry, Integrability and Geometry: Methods and Applications. — 2007. — Т. 3. — Бібліогр.: 34 назв. — англ. |
Institution
Digital Library of Periodicals of National Academy of Sciences of Ukraine| _version_ | 1860022354350964736 |
|---|---|
| author | Grünbaum, F.A. |
| author_facet | Grünbaum, F.A. |
| citation_txt | The Rahman Polynomials Are Bispectral / F.A. Grünbaum // Symmetry, Integrability and Geometry: Methods and Applications. — 2007. — Т. 3. — Бібліогр.: 34 назв. — англ. |
| collection | DSpace DC |
| container_title | Symmetry, Integrability and Geometry: Methods and Applications |
| description | In a very recent paper, M. Rahman introduced a remarkable family of polynomials in two variables as the eigenfunctions of the transition matrix for a nontrivial Markov chain due to M. Hoare and M. Rahman. I indicate here that these polynomials are bispectral. This should be just one of the many remarkable properties enjoyed by these polynomials. For several challenges, including finding a general proof of some of the facts displayed here the reader should look at the last section of this paper.
|
| first_indexed | 2025-12-07T16:48:00Z |
| format | Article |
| fulltext |
Symmetry, Integrability and Geometry: Methods and Applications SIGMA 3 (2007), 065, 11 pages
The Rahman Polynomials Are Bispectral?
F. Alberto GRÜNBAUM
Department of Mathematics, University of California, Berkeley, CA 94720, USA
E-mail: grunbaum@math.berkeley.edu
Received February 01, 2007, in final form April 22, 2007; Published online May 03, 2007
Original article is available at http://www.emis.de/journals/SIGMA/2007/065/
Abstract. In a very recent paper, M. Rahman introduced a remarkable family of polyno-
mials in two variables as the eigenfunctions of the transition matrix for a nontrivial Markov
chain due to M. Hoare and M. Rahman. I indicate here that these polynomials are bispectral.
This should be just one of the many remarkable properties enjoyed by these polynomials.
For several challenges, including finding a general proof of some of the facts displayed here
the reader should look at the last section of this paper.
Key words: bispectral property; multivariable polynomials; rings of commuting difference
operators
2000 Mathematics Subject Classification: 33C45; 22E45
Dedicated to the memory of Vadim Kuznetsov
I first met Vadim at a meeting in Esterel, Canada back in 1994. I only saw him a few more
times after that, in Amsterdam, in Leeds, in Louvain-la-Neuve, and the last time in Leganes.
Although we never wrote any papers together we shared a number of common interests. I always
found Vadim extremely helpful and willing to share his ideas and his vast knowledge. He was
also willing to put up with people whose work was not in the mainstream and he usually would
have some useful remark or a suggestion to make. I like to think that he would have enjoyed
seeing this paper which builds on some remarkable work of another common friend I met for the
first time at the same Esterel meeting.
1 Preliminaries
Starting about thirty years ago, M. Hoare and M. Rahman, published a number of papers
studying a class of statistical models that were nicely associated with “classical” orthogonal
polynomials in one variable. The reader can consult [19, 20, 21, 4].
In a very recent paper [22], which I have seen before publication thanks to the kindness
of M. Rahman, they take a huge step and consider the case of several variables. I propose
to call these statistical models with the name Hoare–Rahman, and to call the multivariable
polynomials that arise with the name Rahman polynomials. I am thankful to Prof. Rahman for
several inspiring conversations on the contents of [22], in different coffee houses and restaurants
in Fremont, California on the last days of 2006.
The model of interest here captures very well the idea of a game of chance where the player
takes a risk that can never improve his/her winnings but is then given a second opportunity
to improve his/her lot. I am convinced that models of this kind should be of great interest
in several areas of applied mathematics not only in physics (where they were conceived) but
also in biology, population dynamics, evolutionary models, etc. On top of their potential for
applications the polynomials discovered by M. Rahman are a piece of beauty. I hope that this
?This paper is a contribution to the Vadim Kuznetsov Memorial Issue ‘Integrable Systems and Related Topics’.
The full collection is available at http://www.emis.de/journals/SIGMA/kuznetsov.html
mailto:grunbaum@math.berkeley.edu
http://www.emis.de/journals/SIGMA/2007/065/
http://www.emis.de/journals/SIGMA/kuznetsov.html
2 F.A. Grünbaum
paper will incite other people to look around for the gems contained in [22]. I only recall the
points from this paper that are needed in the present context, urging the reader to go back to
the source for inspiration.
For a systematic study of polynomials in several variables one should consult [8] as well as
the work on Macdonald polynomials of various kinds, see [31]. A look at the pioneering work of
Tom Koornwinder, see for instance [28], is always a very good idea.
2 The Rahman polynomials
Given generic values of the parameters p1, p2, p3, p4 (these could even be complex numbers)
define t, u, v, w by means of
t =
(p1 + p2)(p1 + p3)
p1(p1 + p2 + p3 + p4)
, u =
(p1 + p3)(p4 + p3)
p3(p1 + p2 + p3 + p4)
,
v =
(p1 + p2)(p2 + p4)
p2(p1 + p2 + p3 + p4)
, w =
(p4 + p2)(p4 + p3)
p4(p1 + p2 + p3 + p4)
.
Now we can define with M. Rahman the polynomials in x, y given by
Pm,n(x, y) =
∑
i
∑
j
∑
k
∑
`
(−m)i+j(−n)k+`(−x)i+k(−y)j+`
i!j!k!`!(−N)i+j+k+`
tiujvkw`.
I will not try to explain M. Rahman’s path to these polynomials, but the reader should con-
sult [22]. In particular one finds in [22] a discussion of the connection with the 9 − j symbols
used in quantum mechanics.
3 The Hoare–Rahman model and the kernel K
Denote by
b(k, N, α) =
(
N
k
)
αk(1− α)N−k
the usual binomial distribution giving the probability of k successes in N repetitions of an ex-
periment with two possible outcomes: success (with probability α) and failure (with probability
1−α). Likewise denote by b2(i1, i2, N ;β1, β2) the trinomial distribution with parameters β1, β2.
This means that that for i1, i2 nonnegative and such that i1 + i2 ≤ N we have
b2(i1, i2, N ;β1, β2) = βi1
1 βi2
2 (1− β1 − β2)N−i1−i2
(
N !
i1!i2!(N − i1 − i2)!
)
.
Define now a transition probability kernel
K(j1, j2; i1, i2)
in the space of non-negative integer physical variables i1, i2 and j1, j2 such that
i1 + i2 ≤ N, j1 + j2 ≤ N
by means of the expression
K(j1, j2; i1, i2) =
min(i1,j1)∑
k1=0
min(i2,j2)∑
k2=0
b(k1, i1;α1)b(k2, i2;α2)
The Rahman Polynomials Are Bispectral 3
× b2(j1 − k1, j2 − k2, N − k1 − k2;β1, β2).
The expression above is written so as to agree with the one in [22]. Notice that its interpre-
tation is that of the probability of a transition from the state (i1, i2) to the new sate (j1, j2) in
one step (consisting of three tosses) of our process, as will be explained in the next section.
At this point the parameters α1, α2, β1, β2 are free to lie anywhere in the interval [0, 1], with
the sum of β1 and β2 not exceeding the value 1.
The corresponding Markov chain is not a “nearest neighbour one” in the sense that the
classical Ehrenfest urn model, see [9, 10, 25, 32], or all other classical models are. Here the
transition probability matrix has all its entries positive and therefore there is no sense in which
a one-step transition moves you only to your immediate neighbours. I find it remarkable that
the complete spectral problem can be solved for such a kernel, but this is exactly what the
Rahman polynomials do, at least under certain extra conditions.
4 Poker dice
A concrete example of the model is provided by a gambling game known as Poker dice. The
version considered by M. Hoare and M. Rahman goes as follows. You have N identical dice
with n faces each. There are three kinds of faces: red, black and “blanks”. The probabilities of
getting a red or a black face when rolling one dice are α1 and α2, respectively. The state of the
system is given by (i1, i2), the number of red and black faces among the N dice. Think of the
remaining N − i1 − i2 dice as failures.
Start form the state (i1, i2) and get ready to roll three times. The resulting number of reds
and blacks after these three rools will be denoted by (j1, j2). This will constitute one time step
in the Markov chain to be described now.
The i1 dice that show red are tossed and yield k1 red faces. The i2 dice that showed a black
face are tossed and yield k2 blacks. Now you take the initial N − i1 − i2 blanks and combine
them with the i1 − k1 dice that failed to reproduce their original red color and the i2 − k2 dice
that failed to reproduce their black color. The total number of “failures” is now N − k1 − k2.
These dice are thrown by themselves: they yield a certain number of red and black faces which
are added to the k1 red faced dice and k2 black faced dice that were saved earlier. This total
number of reds is now j1 and the total number of blacks is j2. To obtain full generality the
probabilities of getting a red or a black for each individual dice on this last toss can be taken to
be β1 and β2, not necessarily the old values α1 and α2.
The case of N = 5 is referred to as standard “poker dice” in [22] and will be used to illustrate
later on the main point in this paper, namely the bispectral property of the Rahman polynomials.
We have verified this property for many other values of N , but we do not have a nice general
expression for the operator in question, or a nice proof of its existence for general N .
5 The orthogonality weight
In [22] one sees that the Rahman polynomials are orthogonal with respect to the trinomial
distribution
b2(x, y;N, η1, η2),
where the parameters η1, η2 are given in terms of p1, p2, p3, p4 by means of
η1 =
p1p2(p1 + p2 + p3 + p4)
(p1 + p2)(p1 + p3)(p2 + p4)
, η2 =
p3p4(p1 + p2 + p3 + p4)
(p1 + p3)(p4 + p2)(p4 + p3)
.
4 F.A. Grünbaum
Given this weight, for an arbitrary choice of the parameters η1, η2, one could blindly decide
to “order” the set of monomials
xiyj
and then apply the Gram–Schmidt process to obtain orthogonal polynomials. Of course there
is no unique way to order these monomials and each ordering will produce a different family of
polynomials.
One of the features of the work of M. Rahman is that it automatically gives one set of
polynomials that are already orthogonal and, as we see in [22] and here, enjoy remarkable
properties. Moreover he produces a four parameter family of these polynomials, having started
from a the two parameters η1, η2. There is clearly a lot of inspiration behind this construction,
and the reader should consult [22].
6 The bispectral property
Recall that K depends on one set of four parameters and the Rahman polynomials on a different
set of four parameters.
As shown in [22], the Rahman polynomials satisfy the eigenvalue relation
KPm,n(x, y) = λm,nPm,n(x, y)
for (x, y) in “physical space”, i.e. x, y are integers satisfying x ≥ 0, y ≥ 0, x + y ≤ N , with
λm,n =
{
1− α1η1 − α2η2 − η1t(1− α1)− η2u(1− α2)
1− α1η1 − α2η2
}m
×
{
1− α1η1 − α2η2 − η1v(1− α1)− η2w(1− α2)
1− α1η1 − α2η2
}n
= (1− β1t− β2u)m(1− β1v − β2w)n.
More explicitly, with the definitions made above this reads,∑
j1,j2
K(j1, j2; i1, i2)Pm,n(j1, j2) = λm,nPm,n(i1, i2).
For this to be true we need to impose some relations connecting the parameters p1, p2, p3, p4
and α1, α2, β1, β2. This issue has been addressed in [22].
In the considerations above we are thinking of these polynomials as functions of the physical
space variables x, y for fixed values of the frequency parameters m, n.
In the spirit of [5] one could ask if they are the eigenvectors of another spectral problem
when considered, for fixed x, y, as functions of m, n. If this were true we would have an analog
of the three-term recursion relation satisfied by orthogonal polynomials of one variable. In the
continuous-continuous case this question was raised in [5] for a second order Schrödinger type
differential operator. For many ramifications of this problem the reader can see [18].
We give below the simplest recursion that we have been able to find. To be very concrete we
display the result in the case of N = 5 which corresponds to the Poker dice example described
above. The operator in question consists of multiplication by the function (p1+p2)x−(p3+p4)y
and it gives a “five-point” formula, where most points in the “spectral space” have four neigh-
bours. There are, as indicated above, similar recursions for other values of N but we do not
have a general expression that would cover all cases.
In the case of N = 5 the frequency space (and also the physical space) looks as follows
The Rahman Polynomials Are Bispectral 5
If we declare two points to be adjacent when they can be obtained by going one step either
in the vertical or horizontal directions, and we order the 21 states as follows
(0, 0), (1, 0), (2, 0), (3, 0), (4, 0), (5, 0), (0, 1), (1, 1), (2, 1), (3, 1), (4, 1),
(0, 2), (1, 2), (2, 2), (3, 2), (0, 3), (1, 3), (2, 3), (0, 4), (1, 4), (0, 5)
then one can work out easily the adjacency matrix.
The matrix B to be detailed below will have nonzero entries exactly where the adjacency
matrix has a 1. It is therefore as much of a local matrix as one can get in this setup. It will
correspond to the analog of the three term recursion relation of the usual situation when one
deals with orthogonal polynomials in one variable.
We claim that the Rahman polynomials, as functions of m, n satisfy the difference equation
BPm,n(x, y) = ((p1 + p2)x− (p3 + p4)y)Pm,n(x, y)
as long as (x, y) lie in physical space.
Here the matrix B is made up of rows as specified one at a time below. Notice that in each
row there is an element denoted by si for which we do not give an expression. This is not needed
since the sum of all the entries in each row equals the value zero.
[s1, a1, 0, 0, 0, 0, b1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
with
a1 =
5p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
, b1 = −5p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[a2, s2, c2, 0, 0, 0, 0, d2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
with
a2 =
p1p4 − p2p3
p3 + p1
, c2 =
4p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
d2 = −4p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, a3, s3, c3, 0, 0, 0, 0, d3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
with
a3 =
2(p1p4 − p2p3)
p3 + p1
, c3 =
3p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
6 F.A. Grünbaum
d3 = −3p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, a4, s4, c4, 0, 0, 0, 0, d4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
with
a4 =
3(p1p4 − p2p3)
p3 + p1
, c4 =
2p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
d4 = −2p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, 0, a5, s5, c5, 0, 0, 0, 0, e5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
with
a5 =
4(p1p4 − p2p3)
p3 + p1
, c5 =
p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
e5 = −p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, 0, 0, a6, s6, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
with
a6 =
5(p1p4 − p2p3)
p3 + p1
;
[d7, 0, 0, 0, 0, 0, s7, c7, 0, 0, 0, a7, 0, 0, 0, 0, 0, 0, 0, 0, 0]
with
a7 = −4p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
,
c7 =
4p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
, d7 = −p1p4 − p2p3
p4 + p2
;
[0, a8, 0, 0, 0, 0, b8, s8, d8, 0, 0, 0, e8, 0, 0, 0, 0, 0, 0, 0, 0]
with
a8 = −p1p4 − p2p3
p4 + p2
, b8 =
p1p4 − p2p3
p3 + p1
,
d8 =
3p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
, e8 = −3p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, e9, 0, 0, 0, 0, c9, s9, a9, 0, 0, 0, d9, 0, 0, 0, 0, 0, 0, 0]
with
a9 =
2p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
, c9 =
2(p1p4 − p2p3)
p3 + p1
,
d9 = −2p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
, e9 = −p1p4 − p2p3
p4 + p2
;
[0, 0, 0, d10, 0, 0, 0, 0, a10, s10, c10, 0, 0, 0, e10, 0, 0, 0, 0, 0, 0]
with
a10 =
3(p1p4 − p2p3)
p3 + p1
, c10 =
p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
The Rahman Polynomials Are Bispectral 7
d10 = −p1p4 − p2p3
p4 + p2
, e10 = −p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, 0, 0, b11, 0, 0, 0, 0, a11, s11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
with
a11 =
4(p1p4 − p2p3)
p3 + p1
, b11 = −p1p4 − p2p3
p4 + p2
;
[0, 0, 0, 0, 0, 0, a12, 0, 0, 0, 0, s12, d12, 0, 0, c12, 0, 0, 0, 0, 0]
with
a12 = −2(p1p4 − p2p3)
p4 + p2
, c12 = −3p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
,
d12 =
3p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
;
[0, 0, 0, 0, 0, 0, 0, c13, 0, 0, 0, a13, s13, d13, 0, 0, e13, 0, 0, 0, 0]
with
a13 =
p1p4 − p2p3
p3 + p1
, c13 = −2(p1p4 − p2p3)
p4 + p2
,
d13 =
2p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
, e13 = −2p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, 0, 0, 0, 0, 0, 0, d14, 0, 0, 0, a14, s14, c14, 0, 0, e14, 0, 0, 0]
with
a14 =
2(p1p4 − p2p3)
p3 + p1
, c14 =
p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
d14 = −2(p1p4 − p2p3)
p4 + p2
, e14 = −p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, 0, 0, 0, 0, 0, 0, 0, a15, 0, 0, 0, c15, s15, 0, 0, 0, 0, 0, 0]
with
a15 = −2(p1p4 − p2p3)
p4 + p2
, c15 =
3(p1p4 − p2p3)
p3 + p1
;
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, a16, 0, 0, 0, s16, c16, 0, d16, 0, 0]
with
a16 = −3(p1p4 − p2p3)
p4 + p2
, c16 =
2p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
d16 = −2p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, e17, 0, 0, a17, s17, c17, 0, d17, 0]
with
a17 =
p1p4 − p2p3
p3 + p1
, c17 =
p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
d17 = −p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
, e17 = −3(p1p4 − p2p3)
p4 + p2
;
8 F.A. Grünbaum
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, a18, 0, 0, c18, s18, 0, 0, 0]
with
a18 = −3(p1p4 − p2p3)
p4 + p2
, c18 =
2(p1p4 − p2p3)
p3 + p1
;
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, c19, 0, 0, s19, a19, d19]
with
a19 =
p1p3(p4 + p2)(p4 + p3 + p2 + p1)
(p3 + p1)(p1p4 − p2p3)
,
c19 = −4(p1p4 − p2p3)
p4 + p2
, d19 = −p2(p3 + p1)p4(p4 + p3 + p2 + p1)
(p4 + p2)(p1p4 − p2p3)
;
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, a20, 0, c20, s20, 0]
with
a20 = −4(p1p4 − p2p3)
p4 + p2
, c20 =
p1p4 − p2p3
p3 + p1
;
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, a21, 0, s21]
with
a21 = −5(p1p4 − p2p3)
p4 + p2
.
We note that these relations hold for arbitrary (generic) values of p1, p2, p3, p4.
If one considers multiplication by x or by y we get a “seven-point” formula. Of course it
would be interesting, in the spirit of [3] to consider the full algebra of difference operators going
along with the Rahman polynomials.
Thinking along these lines one could also consider the full algebra of difference operators that
go along with the Rahman polynomials when thought of as functions of x, y. This simply means
the algebra of all matrices that commute with the kernel K. While the original kernel K has
a very nice probabilistic interpretation there is no reason why other operators in this algebra
should allow such an interpretation. More on this in the next section.
7 Global vs. local
One should naturally think of the kernel K given earlier as a global operator, one that in this
discrete case is represented by a full matrix and that in the continuous case would be represented
by an integral operator. At the core of many considerations leading up to the bispectral problem
is the issue of the existence of a local operator that would commute with the global one. This
remarkable piece of work is forever associated with the names of D. Slepian, H. Landau and
H. Pollack. This is discussed for instance in [13, 14].
Here, as an illustration we show that this question has a positive answer in the setup of Hoare
and Rahman. Take N = 3 in which case we can see, for arbitrary choice of the free parameters
α1, α2, β1, β2 that the matrix given below is essentially the only one that commutes with the
kernel K. Once again, we have checked that this holds for other values of N , and this case is
used here only for illustration purposes.
In the spirit of the work just alluded to one should consider K as the analog of the operator
of time-band-time limiting and the operator below as the analog of the second order differential
The Rahman Polynomials Are Bispectral 9
operator that comes from separation of variables for the Laplacian in R3 in prolate spheroidal
coordinates.
In agreement with the case N = 5 considered as an illustration earlier, we order the states
in physical space as follows
(0, 0), (1, 0), (2, 0), (3, 0), (0, 1), (1, 1), (2, 1), (0, 2), (1, 2), (0, 3).
Then, up to trivial scaling and shifts the only local matrix that commutes with K is given by
x1 x2 0 0 x3 0 0 0 0 0
x4 x5 x6 0 0 x7 0 0 0 0
0 x8 x9 x10 0 0 x11 0 0 0
0 0 x12 x13 0 0 0 0 0 0
x14 0 0 0 x15 x16 0 x17 0 0
0 x18 0 0 x19 x20 x21 0 x22 0
0 0 x23 0 0 x24 x25 0 0 0
0 0 0 0 x26 0 0 x27 x28 x29
0 0 0 0 0 x30 0 x31 x32 0
0 0 0 0 0 0 0 x33 0 x34
as long as we choose its entries as follows
x1 = 3(α1α2β2 − 2α1β2 + α1α2β1 − α2β1 − α1β1 − α1α2 + α1)/(α1β2),
x2 =
3α2β1
α1β2
x3 = 3, x4 =
(α1 − 1)α2(β2 + β1 − 1)
α1β2
,
x5 = (2α1α2β2 + α2β2 − 5α1β2 + 2α1α2β1 − α2β1 − 3α1β1 − 2α1α2 − α2 + 3α1)/(α1β2),
x6 =
2α2β1
α1β2
, x7 = 2, x8 =
2(α1 − 1)α2(β2 + β1 − 1)
α1β2
,
x9 = (α1α2β2 + 2α2β2 − 4α1β2 + α1α2β1 + α2β1 − 3α1β1 − α1α2 − 2α2 + 3α1)/(α1β2),
x10 =
α2β1
α1β2
, x11 = 1, x12 =
3(α1 − 1)α2(β2 + β1 − 1)
α1β2
,
x13 =
3(α2 − α1)(β2 + β1 − 1)
α1β2
, x14 =
(α2 − 1)(β2 + β1 − 1)
β2
,
x15 = 2(α1α2β2 − 2α1β2 + α1α2β1 − α2β1 − α1β1 − α1α2 + α1)/(α1β2),
x16 =
2α2β1
α1β2
, x17 = 2, x18 =
(a2 − 1)(β2 + b1 − 1)
β2
,
x19 =
(α1 − 1)α2(β2 + β1 − 1)
α1β2
,
x20 = (α1α2β2 + α2β2 − 3α1β2 + α1α2β1 − 2α1β1 − α1α2 − α2 + 2α1)/(α1β2),
x21 =
α2β1
a1β2
, x22 = 1, x23 =
(α2 − 1)(β2 + β1 − 1)
β2
,
x24 =
2(α1 − 1)α2(β2 + β1 − 1)
α1β2
, x25 =
2(α2 − α1)(β2 + β1 − 1)
α1β2
,
x26 =
2(α2 − 1)(β2 + β1 − 1)
β2
,
x27 = (α1α2β2 − 2α1β2 + α1α2β1 − α2β1 − α1β1 − α1α2 + α1)/(α1β2),
x28 =
α2β1
α1β2
, x29 = 1, x30 =
2(α2 − 1)(β2 + β1 − 1)
β2
,
10 F.A. Grünbaum
x31 =
(α1 − 1)α2(β2 + β1 − 1)
α1β2
, x32 =
(α2 − α1)(β2 + β1 − 1)
α1β2
,
x33 =
3(α2 − 1)(β2 + β1 − 1)
β2
, x34 = 0.
8 Some closing challenges
Here is a challenge that is motivated by recent work on matrix valued orthogonal polynomials.
This subject was initiated by M.G. Krein a long time ago, see [29, 30]. There has been quite
a bit of work in terms of producing explicit examples of these polynomials. For a representative
sample see [3, 6, 7, 11, 15, 16, 17, 33, 34]. Most of this work deals with looking for examples
that would enjoy the appropriate bispectral property. On the surface of it, there is little relation
between these polynomials and the topics discussed here. But appearances can be deceiving.
In a recent paper [12] I noticed how the example in [11] has a probabilistic interpretation (still
waiting to be unravelled) and leads to nonnearest neighbours transitions on the nonnegative
integers. This goes beyond the usual birth-and-death processes for which there is a well known
relation to orthogonal polynomials, see [1, 2, 23, 24, 26, 27]. In the same paper [12] one finds
examples arising in the theory of matrix valued orthogonal polynomials (in one variable) that
give rise to Markov chains in two dimensional discrete space.
Another challenge would be to find the appropriate concrete algebraic geometry that lies
behind the relation among the variables α1, α2, β1, β2 and p1, p2, p3, p4. In particular this
could have some bearing on the situation discussed in [22] where the spectrum of K is degenerate,
see (5.15).
One challenge mentioned already a couple of times consists in writing down the general
formula for a five-term recursion formula that is displayed here only for N = 5.
Acknowledgments
I am very thankful to a couple of referees who read the paper with great care and pointed out
typos as well as ways to improve the presentation. The author was supported in part by NSF
Grant # 0603901.
References
[1] Andrews G., Askey R., Roy R., Special functions, Encyclopedia of Mathematics and Its Applications,
Cambridge University Press, 1999.
[2] Askey R., Wilson J., Some basic hypergeometric orthogonal polynomials that generalize Jacobi polynomials,
Mem. Amer. Math. Soc. 54 (1985), no. 319.
[3] Castro M., Grünbaum F.A., The algebra of matrix valued differential operators associated to a given family
of matrix valued orthogonal polynomials: five instructive examples, Int. Math. Res. Not. 2006 (2006).
Article ID 47602, 33 pages.
[4] Cooper R., Hoare M., Rahman M., Stochastic processes and special functions: on the probabilistic origin
of some positive kernels associated with classical orthogonal polynomials, J. Math. Anal. Appl. 61 (1977),
262–291.
[5] Duistermaat J.J., Grünbaum F.A., Differential equations in the spectral parameter, Comm. Math. Phys.
103 (1986), 177–240.
[6] Duran A.J., Grünbaum F.A., Orthogonal matrix polynomials satisfying second order differential equations,
Int. Math. Res. Not. 2004 (2004), no. 10, 461–484.
[7] Duran A., van Assche W., Orthogonal matrix polynomials and higher order recurrence relations, Linear
Algebra Appl. 219 (1995), 261–280, math.CA/9310220.
[8] Dunkl C., Xu Y., Orthogonal polynomials of several variables, Cambridge University Press, 2001.
http://arxiv.org/abs/math.CA/9310220
The Rahman Polynomials Are Bispectral 11
[9] Ehrenfest P., Eherenfest T., Über zwei bekannte Einwände gegen das Boltzmannsche H-Theorem, Physikalis-
che Zeitschrift 8 (1907), 311–314.
[10] Feller W., An introduction to probability theory and its applications, Vol. 1, 3rd ed., Wiley, 1967.
[11] Grünbaum F.A., Matrix valued Jacobi polynomials, Bull. Sci. Math. 127 (2003), 207–214.
[12] Grünbaum F.A., Random walks and orthogonal polynomials: some challenges, math.PR/0703375.
[13] Grünbaum F.A., The bispectral problem: an overview, in Special Functions 2000: Current Perspective and
Future Directions, Eds. J. Bustoz et al., 2001, 129–140.
[14] Grünbaum F.A., Some bispectral musings, in The Bispectral Problem (Montreal, 1997), CRM Proc. Lecture
Notes, Vol. 14, Amer. Math. Soc., Providence, RI, 1998, 11–30.
[15] Grünbaum F.A., Pacharoni I., Tirao J.A., Matrix valued spherical functions associated to the complex
projective plane, J. Funct. Anal. 188 (2002), 350–441, math.RT/0108042.
[16] Grünbaum F.A., Pacharoni I., Tirao J.A., Matrix valued orthogonal polynomials of the Jacobi type, Indag.
Mathem. 14 (2003), 353–366.
[17] Grünbaum F.A., Pacharoni I., Tirao J.A., Matrix valued orthogonal polynomials of the Jacobi type: the
role of group representation theory, Ann. Inst. Fourier (Grenoble) 55 (2005), 2051–2068.
[18] Harnard J., Kasman A. (Editors), The bispectral problem (Montreal, 1997), CRM Proc. Lecture Notes,
Vol. 14, Amer. Math. Soc., Providence, RI, 1998.
[19] Hoare M., Rahman M., Distributive processes in discrete systems, Phys. A 97 (1979), 1–41.
[20] Hoare M., Rahman M., Cumultive Bernoulli trials and Krawtchouk processes, Stochastic Process. Appl. 16
(1983), 113–139.
[21] Hoare M., Rahman M., Cumultive hypergeometric processes: a statistical role for the nFn−1 functions,
J. Math. Anal. Appl. 135 (1988), 615–626.
[22] Hoare M.R., Rahman M., A probabilistic origin for a new class of bivariate polynomials, SIGMA 4 (2008),
089, 18 pages, arXiv:0812.3879.
[23] Ismail M.E.H., Classical and quantum orthogonal polynomials in one variable, Cambridge University Press,
2005.
[24] Ismail M.E.H., Masson D.R., Letessier J., Valent G., Birth and death processes and orthogonal polynomials,
in Orthogonal Polynomials, Editor P. Nevai, NATO Advanced Science Institutes Series C: Mathematical and
Physical Sciences, Vol. 294, Kluwer Acad. Publishers, Dordrecht, 1990, 229–255.
[25] Kac M., Random wak and the theory of Brownian motion, Amer. Math. Monthly 54 (1947), 369–391.
[26] Karlin S., McGregor J., The classification of birth and death processes, Trans. Amer. Math. Soc. 86 (1957),
366–400.
[27] Karlin S., McGregor J., The Hahn polynomials, formulas and an applications, Scripta Math. 26 (1961),
33–46.
[28] Koornwinder T., Two variable analogues of the classical orthogonal polynomials, in Theory and Applic. of
Special Functions, Editor R. Askey, Academic Press, 1975, 435–495.
[29] Krein M.G., Fundamental aspects of the representation theory of Hermitian operators with deficiency index
(m, m), AMS Translations, Series 2, Vol. 97, Providence, Rhode Island, 1971, 75–143.
[30] Krein M.G., Infinite J-matrices and a matrix moment problem, Dokl. Akad. Nauk SSSR 69 (1949), no. 2,
125–128.
[31] Macdonald I., Affine Hecke algebras and orthogonal polynomials, Cambridge University Press, 2003.
[32] Schrödinger E., Kohlrausch F., Das Ehrenfestche Model der H-Kurve, Phys. Zeit. 8 (1907), 311–314.
[33] Sinap A., van Assche W., Orthogonal matrix polynomials and applications, J. Comput. Appl. Math. 66
(1996), 27–52.
[34] Tirao J., The matrix valued hypergeometric equation, Proc. Natl. Acad. Sci. USA 100 (2003), no. 14,
8138–8141.
http://arxiv.org/abs/math.PR/0703375
http://arxiv.org/abs/math.RT/0108042
http://arxiv.org/abs/0812.3879
1 Preliminaries
2 The Rahman polynomials
3 The Hoare-Rahman model and the kernel K
4 Poker dice
5 The orthogonality weight
6 The bispectral property
7 Global vs. local
8 Some closing challenges
References
|
| id | nasplib_isofts_kiev_ua-123456789-147372 |
| institution | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| issn | 1815-0659 |
| language | English |
| last_indexed | 2025-12-07T16:48:00Z |
| publishDate | 2007 |
| publisher | Інститут математики НАН України |
| record_format | dspace |
| spelling | Grünbaum, F.A. 2019-02-14T14:48:36Z 2019-02-14T14:48:36Z 2007 The Rahman Polynomials Are Bispectral / F.A. Grünbaum // Symmetry, Integrability and Geometry: Methods and Applications. — 2007. — Т. 3. — Бібліогр.: 34 назв. — англ. 1815-0659 2000 Mathematics Subject Classification: 33C45; 22E45 https://nasplib.isofts.kiev.ua/handle/123456789/147372 In a very recent paper, M. Rahman introduced a remarkable family of polynomials in two variables as the eigenfunctions of the transition matrix for a nontrivial Markov chain due to M. Hoare and M. Rahman. I indicate here that these polynomials are bispectral. This should be just one of the many remarkable properties enjoyed by these polynomials. For several challenges, including finding a general proof of some of the facts displayed here the reader should look at the last section of this paper. This paper is a contribution to the Vadim Kuznetsov Memorial Issue ‘Integrable Systems and Related Topics’. I am very thankful to a couple of referees who read the paper with great care and pointed out typos as well as ways to improve the presentation. The author was supported in part by NSF Grant # 0603901. en Інститут математики НАН України Symmetry, Integrability and Geometry: Methods and Applications The Rahman Polynomials Are Bispectral Article published earlier |
| spellingShingle | The Rahman Polynomials Are Bispectral Grünbaum, F.A. |
| title | The Rahman Polynomials Are Bispectral |
| title_full | The Rahman Polynomials Are Bispectral |
| title_fullStr | The Rahman Polynomials Are Bispectral |
| title_full_unstemmed | The Rahman Polynomials Are Bispectral |
| title_short | The Rahman Polynomials Are Bispectral |
| title_sort | rahman polynomials are bispectral |
| url | https://nasplib.isofts.kiev.ua/handle/123456789/147372 |
| work_keys_str_mv | AT grunbaumfa therahmanpolynomialsarebispectral AT grunbaumfa rahmanpolynomialsarebispectral |