Chapter 8

Hermite Polynomials

by

Sandip Banerjee

Department of Mathematics
Indian Institute of Technology Roorkee
Roorkee 247667, Uttarakhand.
E-mail: [email protected]
Module 1: Solution of Hermite equation.

1. Hermite Polynomials
In mathematics, the Hermite polynomials fall in the category of classical or-
thogonal polynomial sequence which occurs in
• probability, such as the Edgeworth series;
• combinatorics, as an example of an Appell sequence, obeying the umbral
calculus;
• in physics, where they give rise to the eigenstates of the quantum harmonic
oscillator.
Hermite polynomials are named in honor of Charles Hermite. However, they
were first defined by Laplace in 1810 and studied in detail by Chebyshev in
1859. Unfortunately, Chebyshev’s work was overlooked. In 1864, Charles Her-
mite wrote the polynomials, stating them as new and hence the name Hermite
polynomial in his honor.

2. Hermite equation and its solution.

Hermite’s equation of order n is defined as
d2 y dy
2
− 2x + 2ny = 0 (1)
dx dx
We use Frobenius method to solve (1). Let
∞
X
y= cm xk+m , c0 6= 0 (2)
m=0

dy d2 y
be the series solution of (1). We differentiate (2) to get dx , dx2 and substitute
them in (1) to obtain
∞
X ∞
X
Cm (k + m)(k + m − 1)xk+m−2 − 2xCm (k + m)xk+m−1 + 2n Cm xk+m = 0
m=0 m=0
∞
X ∞
X
⇒ Cm (k + m)(k + m − 1)xk+m−2 − 2 Cm (k + m − n)xk+m = 0 (3)
m=0 m=0

The indicial equation is (obtained by equating to zero the smallest power of

x in (7), namely xk−2 )
c0 k(k − 1) = 0 ⇒ k = 0, 1 (since c0 6= 0);

1
which are distinct and differ by an integer. We now equate to zero the coefficient
of xk−1 in (7) to obtain

C1 k(k + 1) = 0 (4)

From (4), it is clear that C1 becomes indeterminate when k = 0. Hence, we can

take C0 and C1 as arbitrary constants. We next equate to zero the coefficient
of xk+m−2 from (7) to obtain

Cm (k + m)(k + m − 1) − 2Cm−2 (k + m − 2 − n) = 0
2(k + m − 2 − n)
Cm = Cm−2 (5)
(k + m)(k + m − 1)

When k = 0, (5) becomes

2(m − 2 − n)
Cm = Cm−2 (6)
m(m − 1)

Now, substituting putting m=2,4,6,....2m, in (6), we get

2n 2n (−1)1 .21 .n
C2 = − C0 = − C0 = − C0
2.1 2! 2!
2(2 − n) (−1)2 .2(2 − n) 2n (−1)2 22 n(n − 2)
C4 = C2 = × C0 = C0 ,
4.3 4.3 2! 4!
... ... ... ... ...

and
(−1)m 2m .n(n − 2)...(n − 2m + 2)
C2m = C0
(2m)!

Again, substituting m = 3, 5, 7, .....2m + 1, in (6), we obtain

2(1 − n) 2(n − 1) (−1)1 21 (n − 1)

C3 = C1 = − C1 = C1
3.2 3! 3!
2(3 − n) (−1)2 22 (n − 1)(n − 3)
C5 = C3 = C1
5.4 5!
... ... ... ... ... ...

and
(−1)m 2m .(n − 1)(n − 3)...(n − 2m + 1)
C2m+1 = C1
(2m + 1)!
.

2
 From (2), with k = 0, we get

y = C0 + C1 x + C2 x2 + C3 x3 + ....
⇒ y = (C0 + C2 x2 + C4 x4 ....) + (C1 x + C3 x3 + C5 x5 ....)
⇒ y = C0 u(x) + C1 v(x), where (7)

2n 2 22 n(n − 2) 4 (−2)m n(n − 2)....(n − 2m + 2) 2m

u(x) = 1− x + x .... + x + ...
2! 4! (2m)!
2(n − 1) 3 22 (n − 1)(n − 3) 5
v(x) = x − x + x + ...
3! 5!
m
(−2) (n − 1)(n − 3)...(n − 2m + 1) 2m+1
+ x + ...
(2m + 1)!

Clearly, u(x)
v(x) is not a constant, which implies, u(x) and v(x) form a linearly
independent set of solutions of (1). Therefore, we conclude (7) to be the most
general solution of (1), C0 and C1 being two arbitrary constants.

To derive Hermite polynomials as a solution of (1), we look into the practical

requirement, that is, the solution should be finite for all finite values of x and
exp.(1/2x2 )y(x) → 0 as x → ∞. The second condition is not satisfied with the
current form of the solution given by (7). We need a terminating series for such
requirement. To satisfy such requirement, we shall now find the series solution
of (1) in descending powers of x, by assuming n to be a non-negative integer.
Therefore, we rewrite (2) for k = 0 as

y = Cn xn + Cn−2 xn − 2 + Cn−4 xn−4 + ... (8)

From (6) we get,

(m + 1)(m + 2)
Cm = Cm+2 (replacing m by m+2) (9)
2(n − m)

We put m = n − 2, n − 4, ....., in (9) to obtain

(n − 1)n n(n − 1)
Cn−2 = − Cn = − Cn
2(n − n + 2) 2.2
(n − 3)(n − 2) n(n − 1)(n − 2)(n − 3)
Cn−4 = − Cn−2 = − Cn
2(n − n + 4) 22 .2.4
.... ................................... and so on.

3
 Substituting these values in (8), we get (after replacing Cn by An )
 
n n(n − 1) n−2 n(n − 1)(n − 2)(n − 3) n−4 r n(n − 1)(n − 2r + 1) n−2r
y = An x − x + x + ... + (−1) x + ...
2.2 22 .2.4 2r .2.4...2r
[n/2]
X n(n − 1)(n − 2r + 1) n−2r
= An (−1)r x
r=0
2r .2.4...2r
[n/2]
X n!
= An (−1)r xn−2r .
r=0
22r r!(n − 2r)!

where 

 n/2 if n is even

[n/2] =

 (n − 1)/2 if n is odd


We take An = 2n and get the solution of (1) in the form, known as the
Hermite polynomial of order n, defined by

[n/2]
X n!
Hn (x) = (−1)r (2x)n−2r
r=0
r!(n − 2r)!

where, 

 n/2 if n is even

[n/2] =

 (n − 1)/2 if n is odd


Putting n = 0, 1, 2, 3, 4 respectively in the expression for Hn (x), we obtain,

0
X 0!
H0 (x) = (−1)r (2x)−2r
r=0
r!(−2r)!
1
= (−1)0 (2x)0 = 1.
0! 0!

0
X 1!
H1 (x) = (−1)r (2x)1−2r
r=0
r!(1 − 2r)!
1
= (−1)0 (2x)1 = 2x.
0! 1!

1
X 2!
H2 (x) = (−1)r (2x)2−2r
r=0
r!(2 − 2r)!
2! 2!
= (−1)0 (2x)2 + (−1)1 (2x)0 = 4x2 − 2.
0! 2! 1! 0!

4
 1
X 3!
H3 (x) = (−1)r (2x)3−2r
r=0
r!(3 − 2r)!
3! 3!
= (−1)0 (2x)3 + (−1)1 (2x)1 = 8x3 − 12x.
0! 3! 1! 1!

2
X 4!
H4 (x) = (−1)r (2x)4−2r
r=0
r!(4 − 2r)!
4! 4! 4!
= (−1)0 (2x)4 + (−1)1 (2x)2 + (−1)2 (2x)0
0! 4! 1! 2! 2! 0!
= 16x4 − 48x2 + 12.