Notes On Generating Functions of Polynomials

194 RECORDS OF PROCEEDINGS AT MEETINGS.
On Cauchy's problem in non-linear differential equations of para-

bolic type: M. R. Siddiqi.
The asymptotic expansion of a function whose operational equivalent
is known: W. G. L. Sutton.
A proof of a theorem of Watson: E. C. Titchmarsh.
On positive harmonic functions; a contribution to ths algebra
of Fourier series: S. Verblunsky.
On an integral inequality: A. Zygmund.
NOTES ON GENERATING FUNCTIONS OF POLYNOMIALS:

(2) HERMITE POLYNOMIALS
G. N. WATSON*
The object of this note is to give proofs of the theorem concerning

Hermite polynomials which is the analogue of the theorem concerning
Laguerre polynomials contained in the first notef of this series.
It is well known J that the Hermite polynomials
HH(x)= (-)***?£?•
n(n-l)(n-2)(n-3) 4
c -f- 4_8 x
are made orthogonal for the interval (— oo, oo) by multiplying them by
the function e~ixi; and that, since
e-x!{Hn(x)}2dx=2nn\
0
the functions *pn(x), defined by the formula
are both orthogonal and normal for the interval (—00,00). The generating
function to be discussed is consequently
K(x, y,t)= L tniffn(x)ifin{y) (Ul < 1).
• Received 4 March, 1933; read 16 March, 1933.

f Supra, 189-192.
J C. Hermite, Camptes rendus, 58 1864), 93-97, 266-273. Cf. Whittaker and Watson,
Modern analysis (1927;, 350-351.
NOTES ON GENERATING FUNCTIONS OF POLYNOMIALS: (2). 195
As Hille has remarked*, nearly everybody who has studied the properties
of Hermite polynomials has discovered t h a t |
1 4a;y
KIX v t)- cxpr '~ ( a ; 2 + y 2 ) ( 1 +' 2 ) l
exp
*(*, y, t) - v{7T{i-t*)} L 20^) J-
Extant proofs of Mehler's formula
nZo
- V{7r( i_* 2) jexp[_ 2(1=^5

are, for the most part, somewhat indirect and elaborate; in fact, it was the
complexity of the investigation given by Wiener J that suggested to me
the desirability of compiling this set of notes.
I now give three proofs of Mehler's formula; the first is due to Hille §,
who has pointed out that, since
H2n(x) = (-)"2 2 »7i! U-»{x*), H2n+1(x) = (-)«2«»+ 1 »! *2#> (**),
Mehler's formula must be deducible from the corresponding result for
Laguerre polynomials; in fact
2»n!
n-o
n-0
rf2) J
Aim r y ( + y ) ( + ) 1
VHl-« 2 )} P
L 2(l-*«) J'
this is Mehler's formula.
In view of the analysis contained in note 1, this proof of Mehler's
formula can be regarded as elementary. The second proof which I give
* Cf. Hardy, Journal London Math. Soc, 7 (1932), 192.

•f It seems that the original discoverer of this result was Mehler, Journal fur Math., 66
(1866), 161-170 (174).
J The Fourier integral (Cambridge, 1933), 57-62.
§ Proc. Nat. Acad. of Sci., 12 (1926), 264.
o2
196 G. N. WATSON
is the elementary proof abbreviated to the utmost by the avoidance

of the introduction of Laguerre polynomials; this proof involves only
(i) rearrangement of absolutely convergent multiple (quadruple) series,
(ii) the use of the formula of Saalschiitz for generalized hypergeometric
functions, and (iii) a (removable) appeal to continuity.
Omitting the trivial factor TT~} exp{—l(x2-\-y2)} fromMehler's formula,
we see that we have to prove that, when \t\ < 1,
~ n r 0 *! L7 2 a -r!(w-2r)!jLT2 28 5!
r! (n-2a)!j'
where the summations with respect to r and s extend over such integral
values as do not give rise to negative factorials in the denominators.
Now, when | <| < p < 1 and|z| ^.A <oo, \y\ < B < oo, the expansion of
which is obtained by writing it in the form

" (2xyt—x2t2-y*t2)N
and then expanding the numerators by the multinomial theorem and the
expressions (1—t2)-N-* by the binomial theorem, is dominated by the
corresponding convergent expansion of
It is therefore permissible to expand the function

T2xyt-(x2+y2)t2l
2 P
<) L i-«2 J
into the quadruple series just described and then to rearrange the terms of
the quadruple series in any convenient manner.
Writing a(a+l) ...(a+m—1) = (a)m for brevity, we find that
{N+\)
£> M$N-M)l ml
M oo / )
(N-M)\ (M-B)\ Rl ml
NOTES ON GENERATING FUNCTIONS OF POLYNOMIALS : (2). 197
Now effect a rearrangement in the summations with respect to M, N,

and R by writing
M = r+s—2m,
N = n—r—s,
R = r—m.
The summations with respect to m, n, r, s then range over all such
integral values as do not give rise to negative factorials in the denominators,
and we have
= s vvv {-)r+8{n-r-s+$m2n+2m-*r-i8tnxn-iryn-'i3
~ n-o r t m (n+2m—2r—2s)! (s—m)\ (r—ra)! m!
_ • y y (—)r+8 2ra-2r-28 P a;n-2r y"-23 _, / — r, —s, n—r—s+l \
~n-o y 3 2
n-o rr .. (n-2r-2s)\
(n-2r-2s)\ r\ r\ a\a\ ^ ^ n + | - r - 5 , J n + l - r - « / '
ow, so long
Now, lon as n is not an integer, it follows from the formula of
Saalschiitz that
, -s, n—
r(w-2r-2s+l)
X
When we make n tend to an integral value, we get

1 F I —r, —s, n—rs+\ \ n!
(n—2r—2s)\ 3 2 \\n+\—r—s, $n+l—r—s) (n—2r)\(n—2s)\'
Observe that it is unnecessary to go through this limiting process when
n~^2r-\-2s. When n<2r-\-2s, the initial terms of the hypergeometric
series (after division by the factorial) vanish; the limiting process can.
then be avoided by dealing with the truncated series.
We thus get, so long as 11\ < 1,
exp
rn-2ryn-2s n \
= 2J
r\ s\ • (n-2r)\ (n-2s)!'
and this is equivalent to Mehler's formula.
198 G. N. WATSON
The third proof was communicated to me by Prof. Hardy*; it is de-

cidedly simpler than mine, though it is rather less elementary since it
involves a use of absolutely convergent infinite integrals. Its most
striking advantage over my proof is that it does not presuppose a knowledge
of the formula to be proved.
The proof is based on the formula
1 f*
e~xi = —y
77
exp(—u2-\-2ixu)du,
V J-oo
which, on differentiation, gives the known result
HJx) = K- 1T
unexp(-u2+2ixu)du,
V' J-00
so that
2»n!
n.o
The convergence of
(2
^ ) n [ exp(-u*-v2+2Au+2Bv)dvdu
when p < 1 shows that rearrangement of the order of the summation and
the integrations is permissible. Hence, when 11\ ^.p < 1,
- exp{-K»M-y»))
0 2«w! V^ "() "
n_0
r
J_ o
-V{"(!-<*)} L 2 l - < 2 J'

and Mehler's formula follows at once.
* It was included in his lectures on orthogonal functions delivered in the Lent Term,
1933.
NOTES OX GENERATING FUNCTIONS OF POLYNOMIALS: (2). 190
The reader who is familiar with Wigert's work on Laguerre polynomials

will see the similarity of this proof to the corresponding analysis of
Wigert, though the theoretical justification of the various steps is rather
easier since no Bessel functions are involved. Prof. Hardy tells me that
he has not seen his proof in print, though the inevitability of the
successive steps makes him think that it is unlikely to be new.
ON TRIPLE PLANES HAVING BRANCH CURVES OF ORDER

NOT GREATER THAN TWELVE
PATRICK DU VAL*.
Recently I had occasionf to describe a certain triple plane, namely

the canonical regular surface of genus 3 and linear genus 4, which I found
to be characterised by having a branch curve of order 12, with twenty-four
cusps disposed in a particular way. It is known, of course, that not every
curve of a given order can be the branch curve of a triple plane, but
that it will generally be necessary for the curve to have eusps, in
which all three sheets branch together. It is also, of course, possible for
the branch curve of a double plane to have a cusp, or for that of a multiple
plane to have a cusp, in which only two sheets take part in the branching;
such a point is to be regarded as a biplanar node—binode—on the surface,
which is a case of a more general surface not having this singularity; a
cusp in which three sheets branch, on the other hand, cannot be abolished
at all. A most interesting topological method has been devised by
EnriquesJ, and applied in some elementary cases by Zariski§, for deter-
mining whether a given curve can be the branch curve of a multiple
plane; but the results actually obtained are not very copious, owing to the
difficulty of the determination of the Poincare group on which the method
depends. I propose to show in the present and some subsequent papers
how the multiple planes (general cases) having branch curves of low orders
can be discovered and enumerated by comparatively unsophisticated
methods. As a beginning, since attention has already been called to one
• Received 21 April, 1933; read 27 April, 1933.

f Journal London Math. Soc, 8 (193:3), 11.
+ Annali di Mat. (4i, L (1923), 185.
§ American J. of Math., 51 (1929), 305.

Notes On Generating Functions of Polynomials

Uploaded by

Copyright:

Available Formats

Notes On Generating Functions of Polynomials

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Notes On Generating Functions of Polynomials

Uploaded by

Copyright:

Available Formats

194 RECORDS OF PROCEEDINGS AT MEETINGS.

On Cauchy's problem in non-linear differential equations of para-

NOTES ON GENERATING FUNCTIONS OF POLYNOMIALS:

The object of this note is to give proofs of the theorem concerning

the functions *pn(x), defined by the formula

K(x, y,t)= L tniffn(x)ifin{y) (Ul < 1).

• Received 4 March, 1933; read 16 March, 1933.

- V{7r( i_* 2) jexp[_ 2(1=^5

* Cf. Hardy, Journal London Math. Soc, 7 (1932), 192.

is the elementary proof abbreviated to the utmost by the avoidance

which is obtained by writing it in the form

It is therefore permissible to expand the function

Now effect a rearrangement in the summations with respect to M, N,

When we make n tend to an integral value, we get

The third proof was communicated to me by Prof. Hardy*; it is de-

which, on differentiation, gives the known result

-V{"(!-<*)} L 2 l - < 2 J'

The reader who is familiar with Wigert's work on Laguerre polynomials

ON TRIPLE PLANES HAVING BRANCH CURVES OF ORDER

Recently I had occasionf to describe a certain triple plane, namely

• Received 21 April, 1933; read 27 April, 1933.

You might also like