Chapter 7

Legendre Polynomials, Chebyshev

Polynomials, Bessel Functions and
Sturm-Liouville Problem
7.1 Introduction

fa many problems of mathemaical physics and engineering. we come across second order line
diferential equations whose solutions give rise to special functions, which are so calledi as they are
difierent from the standard functions like sine, cosine, exponential, logarithmic functions etc.
In this chapter, we obtain the solutions of this important class of linear differential equations and
define the special functions.

1.2 Legendre Differential Equation and Legendre Polynomials

One of the important differential equations which gives rise to special functions is the Legendre
differential equation

(1-)-2ry+nn+1)y =0 (7.1)
where n is a real constant. In most applications, n takes positive integral values. We can also write this
equation as

C1-)yT+ nn + 1)y =0. (7.2)

The singularities of this equation are x =t 1. Ifwe now seek a power series solution about x =0, of

the form y(x) = Cmx, then this series solution is convergent in | x |< 1, since the distance
m=0
between x = 0 and the nearest singularity is 1. Substituting

y(x) = E cn". y(x)= m c , a n d ylx)= Z m{m - 1)cm (7.3)

in Eq. (7.1l), we obtain

 7.2 Engineering Mathematics

Lm(m -I)cm " 2 m(m- 1Dcm-2 2 mcmx"

+n(n + ) 2 cm "n =0
m2 m

or 2c2 +ocz +2 m(m -1)cux 2 m(m -1)cm"-2c1x

m=4 m-

-2 2 mcm + n(n+ IMCo + cjx) + n{n + 1) 2 Cmx" = 0.

m=
m=2

Substituting m - 2 =1 in the first sum, we get

2c2+n(n + 1)co] + 16c^ -2c1 +n(n+ 1)eiJr+ 2(+ 2(t + 1)c,2x'

m(m
m=2
-

1) + 2m -n(n + 1)]cm = 0.

Since t is a dummy variable, we can combine the third and fourth terms of this
equation. Equating the
coefficients of various powers of x to zero, we get
2c2 +n(n + 1)c% = 0, 6c3-2c1 +n{n + 1)c = 0,

and m+2m + 1)Cm+2-(m*+m-n*-n)em = 0, m =2,3,...

Therefore, C a n + 1cg, eg =2-n-nle = - n

2.
- n +2)0

(n- n)m +n + 1)
Cm+2 m22.
(m +2(m+ I) Cm» (7.4)
We have

C4--2)(n +3
_(n-2(n +3)
-

4.3
,C2 =2)n(n +
=
c n
=
-2)n(n +10n +3)¢o
4 3 2!
cs-
4 = -3)(n- 1(n +2%n +4)¢1.
Substituting in the power series solution, we obtain

yx)=co-nn+1)x+Tn-2)n(n+I0n +3)x*-..
+ a 1Xn +2)x+ zn -3n - a+ 2Xn +4)x-..
=
Co Yor) +C[y1t). (7.5)
where
yo(x)= 1 -7 n(n+ 1)x3n-2)n(n+ 1Xn +3)x*-..
(r) =x- - 1m + 2)x° (n -3(n -

1(n +2n + 4)x*-... (7.6)

 Legendre Polmomials, Chebyshev P'olnomials, Bessel Funco s 7.3
The series for yolx),
y (t) converge for |x|< 1. We note that
and )contains
odd powers
of yolt) contains Cven
e powers ofx only
solutions of the
x only. Hence, the two solutions yl), yt) yil are the linearly
ndependent
Legendre'
As n takes the value zero and even
s cquation.
positive integral values, we obtain for
n=0: yolt) =l, n=2: yot) =1 -3?, n=4: yo(t) =1-
10x+',(7.6a)
hat yolr)
This implies that reduces to an even polynomial (polynomial of evenpowers) aS n takes even
sitive integral alues, whereas y1(r) remains an infinite series.
p o s i t i

As ntakes odd positive integral values, we obtain for

n=1: )=X, n=3:y(x) =x-r, n=5:y (t) =x - + , (7.6b)

Therefore, y1) reduces to a polynomial of odd powers (odd polynomial) as n takes odd positive
integral values, whereas yo(r) remains an infinite series. These polynomials multiplied by suitable
constants are called Legendre polynomials. The Legendre polynomials are denoted by PA) where n
denotes the order of the polynomial. Therefore, when n takes integral values, one of the linearly
independent solutions of Eq. (7.1) is a Legendre polynomial and the second independent solutionis
an infinite series. The second solution is denoted by Q,Ce). In order to explicitly write the expressions
of the
for the Legendre polynomials, we need to evaluate the multiplicative constants. Thethevalues recurrence
multiplicative constants are obtained by setting P) 1. It is too cumbersome to use
=

relation given in Eq. (7.4) or the polynomials given in Eqs. (7.6a), (7.6b) to determine the multiplicative
find the expressions for the
constants. It is easy to use the Rodrigue's formula (see next section) to
Legendre polynomials.

1.2.1 Rodrigue's Formula

the
the binomial theorem to (- 1)" and differentiating n times,
We shall now stow that by applying differential
of n, is a solution of the Legendre
resulting expression, which is a polynomial degree that P(1) 1.
We can then determine the expression for P,t) by requiring
=

equation.
Letu= (x*- 1)". Differentiating, get
we

2nxu

u where u.
(1 -)u + 2nxu 0,
=
=

or
times by using the Leibniz theorem, we obtain
Differentiating this equation (n 1)
+

+2n[xu,4 + (n + 1IJu,] =0
(-xuy2 + (n + 10-2x)491 *n + Dn(-2)4,
0
[2n(n + 1) -n(n + DJu,
=

2nx]Mp+1 +
or (1-u2 +-2+ 1)x+
(1-M-2x(u +m(n + IXtu) = 0
or

which is the Legendre differential equation

for y =
u
Therefore,
Legendre's equation.
series solution of the
However, Pz(x) is the finite
a u a is a constant
Pnx) =

(7.7)
=a -1.
 7.4 Engineering Mathematics

I. Now, denoting
D =
didx, we have
nc constant a is determined by setting P.(1) =

P,0)= I =a -1)"(+ 1"|

=1

=
a "C, D'(r -

1)" D"-l(x +
1)"|
x=l
= aln! (r + 1" + terms containing product of various powers of (x - 1) and (r + 1

= aln! 2"]
Since all the terms containing positive powers of (r - 1) vanish at x = 1. Therefore,

and ) = - 1)".
(7.8)
n12
This equality is called the Rodrigue's formula. Using the Binomial theorem, we get

-1 " C , ( x * ) - ) =r r ! ( n - r)!*

Differentiating n times, we get from Eq. (7.8)

P,(x) rn-r)!d
1 (-1)'2n-2r)!x-2
r!(n- r)!(n - 2r)! (7.9)
where N= n/2 or (n - 1)/2 whichever is an integer. In notation, we usually write as N = Ln/2]
Substituting n = 0, 1, 2,... in Eq. (7.9), we get the Legendre polynomials as

Ps)=1, Pir) =z, B(r)=50x*-1)

B()= 6x'-3x), P,(x)=(35-30x+3),. 7.10)
The graphs ofPolr), Pi(), Patt), Patr) and Pacx) are given in Fig. 7.1. We note that | PAT) |S 1,
xe -1, 1]
The Legendre polynomials belong to an important class of orthogonal polynomials, which we
shall discuss in a later section.

Example 7.1 Express P(x) = 3Pstx) + 2P2x) + 4Pi) + 5P%t) as a polynomial in x, where Pm)
is the Legendre polynomial of order m.

Solution Substituting the expressions for Po. Pi» Pz and P3, We obtain

P(x)= x -3x) + (3x - 1)+4x+ 5

(15x 9x +6x -2+8x + 10) (15r =

+ 6x2 -x +8).
 Legendre Polynomials, Chebyshev Polynomials, Bessel Functions .

PAx)
0, 1)

P'
P

(-1,0) X (1,0)

(0,-1)
-

Fig. 7.1. Legendre polynomials.

wample 7.2 Express f ) =**2x - 6x* + Sx - 3 in terms of Legendre polynomials.

Solution Writing various powers of x in terms of Legendre polynomials, we get

I= Polc). x =Pi), x*=

2P+1)=2n+ B),
x'=(2P +3x) =(2P +3R).
x= (8P +30x 3) =
8 P 10(2 P
+ + P)-3 Pl

5(8P+20P + 7P).
Therefore, f() = 8P+20P+ 7Ph)+ 2 P +3A)-2(2P + P) + 5R - 3/P

8P+ 28P 120P +217R 168 Ph).

7.2.2 Generating Function for Legendre Polynomials

Legendre's polynomials can also be obtained by using a generating function. We shall prove that

= 2 P(x), t* 1. (7.11)
I-2xt+12 n=0

The function on the left hand side is called the generating function of the Legendre polynomials. We
have

(1-2xt+ 2)-1/2 = [1 - (2x- t)t]-2 = [1 - ul-/2

where u=
(2x- t)t.
 7.6 Engineering Mathematics

Expanding in a binomial series, we get

(-u)- =1+u4 (1/2)3/2)

2!
, (/2)3/215/2),44..
3!

l
2!
2nu+
1323"*21)2" "(31 2 . . .t n?22
4! 6
Substituting u = (2x- t)t, we get
4!
(1-2xt+12)-2 =1 C1)223 -1t* 2x -1)2

(2-2)-(2x - 1)*t**+... + (2n)(2x-t"t"

2 r - t)"1"4t..
(n) 22
Now, the term in t", in the product (2x - 1)** is given by

2n-2r
[-C(-1)y (2x)-2r|= i 2 1 ) ' 1 " , r!(n 2 r ) - ) " r - 2

Therefore, the contribution of this term to r", in the series is

(2n-2r)!
(2-2r)!_(-r)!2- -r= "'yr
2 ( n - r ) ! r !( n - 2 , 1 0 ' " * 2
l(n-)!?22n-27
Summing for r = 0, 1,... , N, where N = n/2 or (n-I/2, whichever is an integer, we get on

using Eq. (79)

(-1(2n- 2r)!* =
Pz (x)r".
ro 2r!(n- r)!(n -2/)!
Therefore, (1 -2xt+)-U2 =
Z P,(x)",
n=0
t#1.

This proves that P,(x) can be obtained by the generatingg function.

Example 7.3 Show that (a) P,(1) = 1, (b) Pa(-x) = (-1'P,t). Hence find P,-1).

Solution
(a) Substituting x = 1 in the generating function, we get

P , P, (1)
I-2+ 7=0

Or P,(1) =(1-1) =1 +t+1?+... +t"

n=0

Comparing the coefficient of " onboth sides, we get P,(1) =1,n=0, l,..
we
(6) Replacing x and t by -

x and -

t respectively in the generating function given in Eq. (7.11),

get
=
P,-x-i" (-1)"P,(-x)"
n=0
=

1-2xt+t
 Legendre Polynomials, Chebyshev Polynomials, Bessel Funcluons
Since the left hand side is 2 P(x)", we get
n=0

PCx)=2-1" P,-x)"
n=0
n=0

Comaring the coefficients of r" on both sides, we get

PAx) (-1"P,(-x)
=
or
P(-x) =(-1"Pax).
Setting x = l1, we get Pr-1) = (-1'P,(1) = ( - I

Recurrence Relations for Legendre

7.2.3 Polynomials
many applications, use of recurrence relations between Legendre polynomials simplifies the
ion of a given problem. We prove the following recurrence relations.
() (+1)Pa1) (2n + 1)xP,*) - nPn-1).
=
(7.12)
This formula is also called the Bonnet's recurrence relation.)

(b) nPax) =
xPE(x) - Pi-1(x). (7.13)
Pa+1(x)-xP{(x) = (n + 1)P,(x). (7.14)
Differentiating the generating function

(1-2xt+ 2)-/2 =
2 Pa (x)r" (7.15)
n=0

partially with respect to 1, we get

- 1 2xt+ t2)-2 (-2x+21) E nP, (x)rn- (7.16)

n=0

Or (x-01 -2xt+1)-i2 =(1 -2x1 +12) nP,(x)r" n=0

or (x-1) Pa(x)r" = (1 - 2xt + ) E nP,(x)"-1,

n=0 n=0

Comparing the coefficient of t" on both sides, we obtain

xPp)-Pr-1) = (n + 1)Pa+1)-2nxP,)+ (n - 1)P-1)

(n + 1)P+1) = (2n + 1)xP,) - n P - )

This proves (a).

Diterentiating Eq. (7.11) partially with respect to x, we obtain

-1-2xt+ t2)-32(-2) = E P{(x)r". (7.17)

n=
7.8 Engineering Mathematics

Dividing the expressions in Eqs. (7.16) and (7.17), we get

2nP, (x)n-1

(x-1) PKx)N" = 2 nP,(x)!".

Comparing the coefficient of on both sides, we get

xP (x) -

Par(x) =
nP,(x).
This proves (b). Similarly, (c) is proved (see Problem 19.in Exercise 7.1).

Example 7.4 Using the recurrence relation (7.12), generate the Legendre polynomials Pz, Pa, P,

given that PoCx) = 1 and Pir) = x.

Solution The required recurrence relation is (see Eq. (7.12))

(n+ 1)Pr+1) = (2n + 1)xPatt) - nPn-1(r)
We are given that Polt) = 1, Pi(x) =F.
For n = 1, we get

2Pz(x) 3xPi(x) Polx) 3x(x)- 1,

= - =

or P(r)= Br- 1).

Forn 2, we get
3P3) = 5xPz (x) - 2P1()

=x(3x- 1) -2x=(05x3- 9x)

Bx)= (5x' - 3x).
or

For n 3, we get

4Pat)= 7xPj)-3P;()

x(5s-3t) -3r- )=535-30x +3

or Pa(x)=(35xr* 30x -

+ 3).

Example 7.5 Using the recurrence relation

(n+1)P+1¢) = (2n + 1) xP,(c) - nPp-1(t).

(recursively) evaluate Pa (1.5) and P^(2.1).

Solution Setting n = 1, we get

2P2) = 3xP (x) - Po(«).

 Legendre Polynomials, Chebyshev Polynomials, Bessel Functions *7.9
Polr)
= . for all x and Pi(r) =
x, we get Pi(1.5)
ince,
=
1.5. Thercfore,
A(1.5) 13(1.5X1.5) -

I| =
15.75)= 2.875.
2, we get
Setting n
3P () =
5x P2 (x) 2P1),
-

and 2P2(t) 3xP1 (x) Po(x).

= -

Po(2.1) 1, Pi(2.1) = 2.1.

Now,

Therefore
PA(2.1)= 3(2.1)2.1) 1] = 6.115.

and
B(2.1)=[5(2.1)X6.115)-2(2.1)) =20.0025.
Show that
Example 7.6
Pr)dr = Zn+P (7)- Pa)N

cahution Adding the two recurrence relations given in Eqs. (7.13) and (7.14), we get

Pr+s (x)- P-(x) (2n = +

1)P, (x).
we obtain the result.
Integrating,
and Orthonormal Functions
7.2.4 Orthogonal
is said to be orthogonal on an interval [a, b], with respect to a weight
A set of functions 9;()}
function W), W(x)> 0 in [a, b], if

W(x); (x)6, (x) dx = 0, i*j. (7.18)

a

over [a, b].

If W)= 1, usually say that the functions o,t), 0r), ..are orthogonal
we
definition:
If a discrete data is given, then we have the following
over a set of N+ 1 points x}, with respect
to a
A set of functions {ó{r)} is said to be orthogonal
weight function W(x) > 0, if
(z,) 0, j*k. (7.19)
W(x
i=0
0;(x;)¢e =

The norm 1 o(x) lI of o,;(x) is defined as

l/2
(7.20)
l19(x) =| W(x)9(x) d

It is obvious that W(x)o (x)dx >0

a

N /2
For discrete data, we define
be orthonormal if they are orthogonal on the interval (a, b]
Tunctions o(t),0 (x),.. .are said to l 1 for all i. It is easy to construct orthonormal

l functions have norm 1, that is, 0(x)

Il =
 7.10 Engineering Mathematics

the functions 19,(x)} be orthogonal

unctions from the corresponding orthogonal functions. Let
Define
(7.21)
Then, the functions (f()) are orthonormal. We have
y(x)
wws)dr =W TG)
WRo(a)9r)dr
=
0
1x1ox
b
0 dx
and Wf)ds=|. W) jo a

1.
W)0(x) dx 1o(x) =

1, the functions Ù,(x) = cos ix, i =0, 1, 2,.. . are orthogonal

Example 7.7 Show that with Wx) =

orthonormal set of functions.

over the interval (-7, T]. Construct the corresponding

Solution We have for i*j

j)x + cos(i -jx]d

dr=cosixcosjxds lcos(i
= +
Cxoj (x)
sini +j)x , sin=)x =0.
+*1-.=
cosix dr = (cos 2ix+ 1)dr
Also

=z,i#
2
0.
For i = 0, we have 10C)= | -
dr=
27
Hence, the functions

are orthonormal over the interval -7, 7).

or orthonormal functions is
the series expansion or
An important application of the orthogonal
(suitable) functions in terms of the orthogonal functions.
We first define a complete set oforthogona
functions.