Scribd
Upload
76 views

Hermite Polynomials and Hermite Functions

The document summarizes properties of Hermite polynomials and Hermite functions. It defines Hermite polynomials using power series expansions. It then derives key properties like orthogonality, differential equations satisfied, and recursion formulas. Hermite polynomials form a complete orthonormal basis for the Hilbert space L2(R,γ) where γ is the Gaussian measure. Hermite functions are defined from Hermite polynomials and also form a complete orthonormal basis, but for the Hilbert space L2(R). Estimates are provided on the decay rates of Hermite polynomials and functions.

Uploaded by

Nirmal Kumar
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
76 views

Hermite Polynomials and Hermite Functions

The document summarizes properties of Hermite polynomials and Hermite functions. It defines Hermite polynomials using power series expansions. It then derives key properties like orthogonality, differential equations satisfied, and recursion formulas. Hermite polynomials form a complete orthonormal basis for the Hilbert space L2(R,γ) where γ is the Gaussian measure. Hermite functions are defined from Hermite polynomials and also form a complete orthonormal basis, but for the Hilbert space L2(R). Estimates are provided on the decay rates of Hermite polynomials and functions.

Uploaded by

Nirmal Kumar
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Appendix A

Hermite polynomials and Hermite functions

Real Hermite polynomials are defined to be

which are coefficients in expansion of power series for exp{ tu - t 2 /2} as function
of t:
(A.2)

By this expansion formula we have:


Theorem A.1 Hermite polynomials have the following expression:

(A.3)

Conversely,
[n/2]
n _ ,'" H n - 2k (U) (A.4)
u - n. L...J 2k k!(n _ 2k)!' n E IN o.
k=O

{Hn' n E IN} satisfy the following differential equations

H~(u) = nHn- 1 (u), n:2: 1, (A.5)


H~(u) - uH~(u) + nHn(u) = 0, n:2:0 (A.6)

and recursion formula:

Ho(u):=I, H 1 (u)=u,
Hn+l(U) = uHn(u) - nHn _ 1 (u), n:2:1, (A.7)

as well as multiplication formula:

(A.8)
Appendix A Hermite polynomials and Hermite functions 253
Moreover, for any), E IR it holds that

[n/2] (),2 _I)k ),n-2k


Hn(>\u) = n! L
k=O
kk'( _ k)' Hn- 2k(U).
2 .n 2 .
(A.9)

Proof. Replacing the power series of etu and e- t2 / 2 with respect to t


into eq. (A.2) and comparing the coefficients of t n on both sides, we obtain
eqs. (A.3) and (AA). Differentiating eq. (A.2) with respect to u and comparing
the coefficients of power series we get (A.5) and (A.6). Again from eq. (A.2) we
know

= ~ (s+t)j H.( )~ sktk


6 ., J U 6 k'
j=O J. k=O'

= ~ Hj(u) ~ (j)i+kt j - 1+k


.6 j!k! 6 l .
J,k=O 1=0
Letting l + k = m, j - l +k = n in the last expression, we have

~!n! L k! 7 ~ H m+n- 2k.


n
L00

m,n=O
mt mAn ( ) ( )
k=O
The multiplication formula (A.8) is obtained by comparing the coefficients of
smtn. In particular, the recursion formula (A.7) is obtained by letting m = 1 in
eq. (A.8). Finally, it follows from eq. (A.2) that

Letting j + 2k = n in the last expression, we obtain

00 [n/2] (),2 _ I)k ),n-2k


~ tn ~ 2kk!(n _ 2k)! H n - 2k (U),

by comparing the coefficients of tn, we then have eq. (A.9).


Considering the Gaussian measure on 1R:
•
'Y(du) = (27r)-1/2exp{ -u2 /2}du
254 Appendix A Hermite polynomials and Hermite functions

and the Hilbert space L 2(lR,'Y), we have


TheoreIYl A.2 Hermite polynomials constitute an orthogonal system in
L2(lR,'Y) :

m,n E INa. (A. 10)

Denote i = A. Then

n E INa, (A.ll)

moreover,

n E INa. (A.12)

When t 2 < 1, we have

(A.13)

Proof. It follows from eq. (A.2) that

r { + t)u -
= 1lR exp (s S2 + t }
-
2
2 - 'Y(du)

=exp{_s2;t2 +(s~t)2}=est

=~ (st)n.
L..J n!
n=O

Comparing the coefficients of smt n we obtain eq. (A.lO). Using contour integra-
tion we have

1m. exp{t(u ± iv)h(dv) = exp { tu - ~}.


By expansion in power series of t (using eq. (A.2) for right-hand side) and com-
paring the coefficients of t n we prove eq. (A.ll). From eq. (A.ll) we know

Hn(u + v) = IlR (u + + iyt'Y(dy)


V

= t
k=O
(~)uk hlR
(v+iy)n-k'Y(dy),
Appendix A Hermite polynomials and Hermite functions 255
which implies eq. (A.12). Again by eq. (A.ll) we have

= iJRL exp{t(u + ix)(v + iY)h(dx)"'((dy).


A direct computation of the integral yields eq. (A.13). •
It follows from eq. (A.4) and multiplication formula (A.8) that Hermite poly-
nomials constitute a linear base of polynomial ring. In view of eq. (A.lO) and
density of polynomials in L2(JR, ,), we know that {(n!)-1/2 Hn} is an orthonor-
mal base of L2(JR, ,). Now consider the Hilbert space L2(JR) = L2(JR, du), where
du is Lebesgue measure. For f E L2(JR), define

(A.14)

Then

Moreover,
(A.15)
Hence J : L2(JR) --t L2(JR,,) is an isomorphism for Hilbert spaces. Let

hn(u) == (n!)-1/2J- 1H n (u)


= (n!)-1/2 7r -l/4 e -u 2
/2 Hn( vlzu). (A.16)

Then {h n , n E IN o} constitute an orthonormal base of L2 (JR). They are called


Hermite functions. By definition and properties of Hermite polynomials we have

n?1. (A.17)

In addition, the following estimates are very useful, for the proof see Hille-
Phillips[l] or G.Szego[I].
Theorem A.3 For any fixed u E JR, we have

hn(u) = O(n-l/4), (A.18)

ioU hn(v)dv = O(n- 3 / 4 ). (A.19)

Moreover,

IlhnilLOO == sup Ihn(u)1 = O(n-l/12), (A.20)


uEJR
IIhnllL' == L Ihn(u)ldu = O(n 1/ 4). (A.21)
256 Appendix A Hermite polynomials and Hermite functions

(A.22)

More precisely, we may take c = 1.2 in the above inequality and (A.22) is then
called Cramer's estimate (cf. Erdelyi[1],p.208).

You might also like