ANALYSE MATHÉMATIQUE. - Sur une gé néralisation
des polynomes d'Hermite. Note de M.Krawtchouk, transmise par M.
Émile Borel.
M. Krawtchouk. C.R.Acad. Sci. 1929. T.189, No.17. P.620 - 622.
MATHEMATICAL ANALYSIS. - On one generalization of Hermite polynomials.
Note 1 by M. Krawtchouk,
submitted by M. Émile Borel.
.
Let's assume that the polynomial ym(x) of the mth degree
is defined by the following equalities
|
� � � �
� � �
|
|
u-1 �
i = 0
|
piyl(xi)ym(xi) = 0(l � m), = 1 (l = m) |
|
|
� �
|
xi+1-xi = 1, pi � o, |
u-1 �
i = 0
|
pi = 1 |
� �
|
. |
|
|
|
| (1) |
Then we have
|
� � �
� �
|
xym(x) = m-1ym-1(x)+m0ym(x)+m1ym+1(x) |
|
|
� �
|
mj = |
u-1 �
i = 0
|
pixiym(xi)ym+j(xi) |
� �
|
|
|
|
|
|
and the minimum of the expression
Jk(T0,T1,�,Tk-1) = |
u-1 �
i = 0
|
pi[T0y0(xi)+�+Tk-1yk-1(xi)-y(xi)]2 (k � u) |
| (2) |
is equal to
Jk(A0,A1,�,Ak-1) = |
u-1 �
i = 0
|
piy2(xi)-A02-�-AN-12 |
|
where
Am = |
u-1 �
i = 0
|
piy(xi)ym(xi). |
| (3) |
In the important case p0 = p1 = � = pu-1 which has been studied
by P. Tchebycheff the polynomials ym represent the
generalization of the Legendre polynomials. We want to consider other
important case where
pi = P(i,u;p,q) = |
� �
�
|
u-1
i
|
� �
�
|
piqu-1-i, xi = i (p > 0, q > 0, p+q = 1). |
| (4) |
It is possible to prove that the functions ym have according to
this hypothesis the following simple form:
|
|
|
|
� �
�
|
|
DmP(x-m,u-m;p,q):P(x,u;p,q) |
| (5) | |
|
|
� �
�
|
|
� �
�
|
u-1
m
|
� �
�
|
-1
|
(pq)-m |
|
|
m �
i = o
|
(-1)i |
� �
�
|
u-x-1
m-i
|
� �
�
|
|
� �
�
|
x
i
|
� �
�
|
pm-iqi |
|
| |
|
concerning the limiting cases of the polynomials
const. |
x! ax
|
Dm |
� �
�
|
|
ax-m (x-m)!
|
� �
�
|
(u� �), p(u-1) = a = const.) |
| (6) |
and the Hermite polynomials
const.et2 |
dm dtm
|
( e-t2) |
� �
|
u� �, x = p(u-1)+t |
| _______ �2pq(u-1)
|
� �
|
. |
| (7) |
In accordance with (4) the formulas (*) and (3) are
conversed respectively to
|
|
|
|
| ____________ �(m+1)(u-m-1)pq
|
jm+1(x) |
| |
|
[ p(u-1)+(q-p)m-x] jm(x)- |
| _______ �m(u-m)pq
|
jm-1(x), |
| |
|
|
� �
�
|
|
|
u-1 �
i = 0
|
y(xi)DmP(xi-m, u-m; p,q) |
| (8) | |
|
(-1)m |
� �
�
|
|
|
u-1 �
i = 0
|
P(i-m,u-m;p,q)Dmy(i-m), |
|
| |
|
Appendices. - 1. Calculation of the incomplete generalized
moments
Rm(x) = |
x-1 �
i = 0
|
P(i,u;p,q)jm(i,u;p,q) (m = 1,2,�,k) |
|
of function (4) gives
Rm(x) = |
� �
�
|
|
Dm-1P(x-m,u-m;p,q) (m = 1,2,�,k) |
| (9) |
As to kth incomplete factorial moment
rk(x) = |
x-1 �
i = 0
|
P(i,u;p,q) |
� �
�
|
p( u-1) -x
k
|
� �
�
|
(k = 0,1,2,�), |
|
it is equal to the linear combination of expressions (9) and moment R0(x) [the result is trivial in a limiting case (7)]. Equality by
M.R. Frisch2
follows from the formula (9) as a special case at (m = 1).
- 2. It is necessary to note the following expansion
|
|
|
P(x,u;p,q) |
u-1 �
m = 0
|
|
� �
�
|
|
jm(x,u;p,q) |
| |
|
× |
m �
i = 0
|
(-1)m-i |
� �
�
|
m
i
|
� �
�
|
pm-iqi |
� �
�
|
|
dm[su-m(p1t+q1s)u1] dtidsm-i
|
� �
�
|
s,t = 1
|
|
| |
|
|
| |
|
with the limiting cases which correspond to the polynomials (6) and (7) and are well known.
Footnotes:
1Submitted on September 23, 1929.
2See Ch. Jordan, Statistique math., 1927, p. 85
File translated from
TEX
by
TTH,
version 2.57.
On 28 Mar 2000, 14:22.