Unit 4 Special Functions-I

UNIT 4
SPECIAL FUNCTIONS-I
Structure

4.1 Introduction 4.3 Spherical Harmonics

Expected Learning Outcomes 4.4 Hypergeometric Functions
4.2 Legendre Polynomials 4.5 Summary
Generating Function 4.6 Terminal Questions
Recurrence Relations 4.7 Solutions and Answers
Orthogonality Relations
Rodrigues’s Formula

4.1 INTRODUCTION
In this unit, we discuss Legendre polynomials, spherical harmonics and
hypergeometric functions. We will solve respective differential equations,
write their generating functions and obtain recurrence relations for each
special function.
In the next unit, we discuss Hermite and Laguerre Polynomials, the Sturm-
Liouville problem, and explain how to expand a given function in terms of
orthogonal functions.
Expected Learning Outcomes
After studying this unit, you should be able to:
 obtain Legendre polynomials by solving Legendre’s differential equation as
well as from the generating function and Rodrigues’s formula;
 derive the recurrence relations, orthogonality relation for Legendre
polynomials and use them to solve problems in physics;
 obtain spherical harmonics by solving the ODE as well as from their
generating function and derive their recurrence relations; and
 obtain Hypergeometric functions by solving the ODE as well as from their
generating function and derive their recurrence relations.

4.2 LEGENDRE POLYNOMIALS

In Unit 1, you have learnt how to solve Laplace’s equation in spherical polar
co-ordinates (r, , ) and seen that the solutions are expressed in terms of
Legendre polynomials. Similarly, the solution of the (, ) part of the
Schrӧdinger equation for an electron is expressed in terms of Legendre
polynomials. In your UG courses, you must have learnt how to solve
Legendre’s differential equation using the power series method. Its solutions
75
Block 1 Mathematical Methods in Physics
are called Legendre polynomials, which are new functions with very interesting
properties.

In this section, we revisit the solution of Legendre’s differential equation and

obtain the Legendre polynomials in two different ways: By solving the
differential equation and from the generating function. (You should refresh
your knowledge by studying Unit 3 of the course PHE-05 entitled
Mathematical Methods in Physics-II.) Then we discuss the properties of
Legendre polynomials; especially the orthogonality property. We have
discussed many applications of Legendre polynomials. We expect you to
study these examples carefully and link them up with the relevant topics. It is
important to study Legendre’s associated differential equation (refer to
Appendix of Unit 14 of the course PHE-14) but we shall not go into the
mathematical rigor of the properties of the associated Legendre polynomials.

You have learnt about Legendre’s differential equation in your UG courses

(see Examples 1 and 3 of Unit 3, Block 1, PHE-05):

(1  x 2 )y   2xy   n(n  1)y  0 (4.1)

You can obtain the solution of this equation by solving SAQ 1. It is:

n(n  1) 2 n(n  2)(n  1)(n  3) 4

y  a0 1  x  x  ...
 2! 4! 

(n  1)(n  2) 3 (n  1)(n  3)(n  2)(n  4) 5

 a1  x  x  x  ...
 3! 3! 
(4.2)

SAQ 1

Solve Eq. (4.1) and obtain its general solution given by Eq. (4.2).

Note from Eq. (4.2) that for an even integer n  0 , the first bracketed term in
this series (with even powers of x) terminates leading to a polynomial solution.
For an odd integer (n > 0), the latter term in the series (with odd powers of x)
terminates and gives a polynomial solution. So, for any integer n  0 ,
Legendre’s equation has a polynomial solution. For n  0,1, 2, ..., Eq. (4.2)
leads to the following solutions of Eq. (4.1):

y  a0 , n0 (4.3a)

y  a1 x, n 1 (4.3b)

y  a0 (1  3 x 2 ) , n2 (4.3c)

 3x  5x3 
y  a1  , n  3 (4.3d)
 2 
 

These expressions (of y) are, apart from multiplication constant, the

Legendre polynomials Pn (x ). The multiplicative constant is chosen so that
76 Pn (1)  1.
Unit 4 Special Functions-I
We can write Eq. (4.2) as:
y  a0 y 1( x )  a1y 2 ( x ) (4.4)

where y1( x ) and y 2 ( x ) are linearly independent. You should note that
Eq. (4.4) does not give the general solution of Legendre’s differential
equation. To understand the general solution we have to reconsider the
recursion formula arrived at while solving Legendre’s equation (solution of
SAQ 1):
(n  k )(n  k  1)
ak 2   ak ; n  0,1, 2, ..., (4.5)
(k  1)(k  2)
Note that the coefficients ak  2 will vanish when (i) n  k and/or (ii)
n   (k  1). ak  2  0 implies that the series terminates with ak as the last
non-zero coefficient. That is how we get polynomial solutions. While obtaining
the polynomial solutions Pn (x ), we considered the polynomial solutions with
n  k only.
When we take n  (k  1), the series obtained diverges for x  1. This
solution is unbounded. It is called a Legendre function of second kind and
denoted by Qn (x ) . Thus, the most general solution of Legendre’s differential
equation can be written as:
y  A1 Pn ( x )  A2 Qn ( x ) (4.6)

Here we shall restrict ourselves to Legendre polynomials of the first kind,

that is Pn (x ) . Let us now derive the expressions for Pn (x ) .

From Eq. (4.5), we note that if k  n, an  2  0 and, by induction,

a n  4  0  a n  6 ...

The continue our discussion, we invert Eq. (4.5) to obtain

(k  1)(k  2)
ak   ak 2
(n  k )  (n  k  1)
By taking k  n  2, n  4,..., we get:
n(n  1)
a n 2   an
2(n  1)
(n  2)(n  3) n(n  1)(n  2)(n  3)
an 4   an 2  an
4(2n  3) 2  4(2n  1)(2n  3)

This yields the polynomial solution

 n(n  1) n 2 n(n  1)(n  2)(n  3) n 4 
y  an  x n  x  x  ...
 2(2n  1) 2  4(2n  1)(2n  3) 
(4.7)
The Legendre polynomials Pn (x ) are defined by choosing

(2n  1)(2n  3)...3! (2n )!

an  
n! 2 (n! ) 2
n

This choice of the coefficients an ensures that Pn (1)  1 when n  1. Thus, 77

 Block 1 Mathematical Methods in Physics
(2n )!  n n(n  1) n 2
Pn ( x )   x  x
2 n (n! ) 2  2(2n  1)

n(n  1)(n  2)(n  3) n 4 

 x  ... (4.8)
2  4(2n  1)(2n  3) 

We now work out a simple example to enable you to fix these concepts

To show the equivalence Example 4.1

of expressions in
Eq. (4.8), we note that Starting from Eq. (4.8), prove that
even terms are not
1
occurring in the numerator P0 ( x )  1, P1( x )  x and P2 ( x )  (3 x 2  1)
defining the coefficient an . 2
So, we multiply the
Solution : From Eq. (4.8), we have
numerator and
denominator by (2n )!  n n(n  1) n 2
Pn  x    x  2(2n  1) x
2n(2n  2) (2n  4)...4  2 . 
2 n (n! ) 2
This gives
(2n ) (2n  1) (2n  2) n (n  1)(n  2)(n  3) n 4 
 x  ...
2  4(2n  1)(2n  3) 
(2n  3)..4  3  2  1
an 
n! (2n ) (2n  2) For n = 0, substituting 0! = 1 and x 0  1, we readily obtain P0 ( x )  1 .
(2n  4)...  4  2
For n  1,
2n!

n! 2n(2n  2)...  4  2 2! 2x
P1( x )  x x
Now (2n )( 2n  2) ...  4  2
21  (1! ) 2 2

 (2n )  2(n  1)... 4!  2 2  2  1 0 

P2 ( x )   x  2  ( 4  1) x 
 2(2)  2(1) 2 2  (2! ) 2  

 2 n n! 24  2 1  1
  x    (3x 2  1)
So that the expression for 4 4  3 2
a n reduces to
2n! In this way, you can obtain expressions for higher order Legendre
an 
2 n (n! ) 2 polynomials. For practice, you may like to solve SAQ 2.

SAQ 2
Determine P3 ( x ).

So far we have obtained expressions for Legendre polynomials by solving

Legendre’s differential equation. These expressions can also be generated
from a function of two variables, say x and t. When expanded in powers of t,
the x-dependent coefficients define the Legendre polynomials. This forms the
subject matter of discussion for the following section.

4.2.1 Generating Function

Consider the function

g ( x, t )  (1  2xt  t 2 ) 1/ 2   Pn ( x )t n (4.9)
78 n 0
Unit 4 Special Functions-I
1/ 2
We expand (1  2 xt  t 2 ) 1/ 2  1  (2 x  t )t  in powers of t for t  1 using
binomial expansion:
 1 1  1  3 
[1  t (2x  t )]1/ 2  1    t (2x  t )     t 2 (2x  t )2
2 2!  2  2 
t (2x  t )  1 3  2  1 3  5  t 3 (2x  t )3  ...
 1  t (2x  t ) 2   
2 2 4  2 46
1  1  3  5  3
     t (2x  t )  ...
3!  2  2  2 
1 3
 1 (2xt  t 2 )  t 2 ( 4 x 2  t 2  4 xt )
2 8
5 3
 t (8 x 3  t 3  12x 2t  6 xt 2 )  ...
16
1 2 3 2 2 3 4 3 3
 1  xt  t  x t  t  xt
2 2 8 2
5 3 3 5 6 15 4 2
 x t  t  t x  ...
2 16 4
1 1
 1  xt  (3 x 2  1) t 2  (5 x 3  3 x ) t 3  ... (4.10)
2 2
On equating the coefficients of t n for n  0,1, 2 and 3 in Eqs. (4.9) and (4.10),
we find that the first few Legendre polynomials are given by
P0 ( x )  1
P1( x )  x
1
P2 ( x )  (3 x 2  1)
2
1
P3 ( x )  (5 x 3  3 x )
2
You should note that these expressions are the same as obtained in
Example 4.1 and SAQ 2. The coefficient of t n in the above expansion will be:
1 3  5...  (2n  1)
(2x ) n
2  4  6...2n
1 3  5...  (2n  3) (n  1)
 . (2x )n 2
2  4  6...  (2n  2) 1!
1 3  5...(2n  5) (n  2)(n  3)
 . (2x ) n 4
2  4  6...(2n  4) 2!
This may be rewritten as:
1 3  5...(2n  1)  n n(n  1) n 2
x  x
n!  2(2n  1)
n (n  1) (n  2) (n  3) n 4 
 x  ...
2  4(2n  1) (2n  3) 

You can readily identify this with Eq. (4.8). So we may conclude that Eq. (4.9)
with t  1 signifies the generating relation for Legendre polynomials and 79
 Block 1 Mathematical Methods in Physics
1
g ( x, t )  is called the generating function for Legendre
1  2 xt  t 2
polynomials.

For x = 1, Eq. (4.9) gives 1  2t  t 2 
1/ 2
  Pn (1) t n
n 0

The left hand side now reduces to (1  t ) 1, which has series representation as

 t n . Then the above expression takes the form:
n 0

 
 tn   Pn (1) t n
n 0 n 0

On comparing the coefficients of t n , we get Pn (1)  1 so that the expression of

the generating function gets validated.
a
The generating function is useful in solving physical problems involving the
potential associated with any inverse square force. To illustrate this we
O q z
consider an electric charge q placed on the z-axis at z = a (see Fig. 4.1). From
Fig. 4.1: An electric
UG courses, you will recall that the electrostatic potential at a non-axial point
charge q placed at due to this charge at a distance r1 from it is given by:
z = a.
1 q
V  (4.11)
4 0 r1

From the properties of a triangle, we can write:

r1  r 2  a 2  2ar cos 

On inserting this expression in Eq. (4.11), we get

q 1
V 
4 0 r 2  a 2  2ar cos 

q 1
 (4.12)
4  0 r a 2
1     2  cos 
a
r  r 

For r  a, the expression under the radical sign may be written as

a
(1  2xt  t 2 )1/ 2 where x  cos  and t  ; t  1. From Eq. (4.10) we note
r
2 1/ 2
that when (1  2xt  t ) is expanded in powers of t for t  1, the coefficient
of t n can be identified with Pn (x ) .

On inserting this result in Eq. (4.12), we get


a n
 Pn (cos ) 
q
V  (4.13)
4   0 r n 0 r 
80
Unit 4 Special Functions-I

Example 4.2
Calculate the values of P2n (0) and P2n 1(0 ).

Solution : Putting x = 0 in Eq. (4.9), we get


(1  t 2 ) 1/ 2   Pn (0)t n
n 0

Using binomial expansion, we can write:

1 2 3 4
(1  t 2 ) 1 / 2  1  t  t  ...
2 8
1 3  5  ...  (2n  1) 2n
 ( 1)n t  ...
2n n!

 P0 (0)  P1(0)t  P2 (0)t 2  ...  P2n (0)t 2n

On comparing the powers of t 2n in the above equation, we get:

1  3  5...  (2n  1)
P2n (0)  ( 1) n
2 n n!

Since in the expansion of (1  t 2 )1/ 2 we only get even power of t, we have:

P2n 1(0)  0

You may now like to solve SAQ 3 on generating functions.

SAQ 3

Prove that
a) Pn ( 1)  ( 1) n and
b) Pn ( x )  (1) n Pn ( x )

We shall now use the generating function to obtain the recurrence relations or
recursion relations for Legendre polynomials. This nomenclature stems from
the fact that expressions for higher order polynomial can be derived from
knowledge of expressions for lower order polynomials.
4.2.2 Recurrence Relations
To obtain two primary recurrence relations, we differentiate g ( x, t ) partially
with respect to t and x, respectively.
You have learnt that the generating function [Eq. (4.9)] for Legendre
polynomials is:

 Pn ( x ) t n
1
g ( x, t )  
1  2tx  t 2 n 0

Differentiating the above expression partially with respect to t, we get:


g  1   2x  2t 
   
t  2  (1  2 xt  t 2 ) 3 / 2 n 0

nPn ( x ) t n 1
81
Block 1 Mathematical Methods in Physics

x t
or
(1  2xt  t 2 ) 3 / 2
  nPn ( x ) t n1
n 0

We rewrite this result as:


(x  t )
 1  2 xt  t 2   nPn ( x )t n 1
1  2 xt  t 2 ) n 0

and use the generating function to write:

 
(x  t )  Pn ( x ) t n  1  2 xt  t 2   nPn ( x ) t n1
n 0 n 0

or
 
1  2xt  t 2  nPn ( x )t n 1  (t  x )  Pn ( x ) t n 0
n 0 n 0

On re-arrangement of terms, we obtain:


 (n  1)Pn ( x ) t n1  (2n  1)xPn ( x ) t n  nPn ( x ) t n1  0
n 0

To collect the coefficient of t n , we replace n by n – 1 in the first term and by

n + 1 in the last term. Then equating the resulting expression to zero, we get:

n Pn 1( x )  (2n  1)x Pn ( x )  (n  1) Pn 1( x )  0

(2n  1)x Pn ( x )  (n  1) Pn 1( x )  nPn 1( x ) (4.14)

This is one of the two important recurrence relations and correlates any
Legendre polynomial with its adjoining polynomials. For example, by putting
n  1 in Eq. (4.14), we get:

3 x P1( x )  2 P2 ( x )  P0 ( x )

But we know that P0 ( x )  1and P1( x )  x. Hence, we find that:

3 x 2  2P2 ( x )  1

1
or P2 ( x )  (3 x 2  1)
2

Similarly, P3 ( x ) can be obtained using expressions for P1( x ) and P2 ( x ); P4 ( x )

can be obtained using expressions for P2 ( x ) and P3 ( x ), and so on.

To derive the second recurrence relation, we differentiate g ( x, t ) partially with

respect to x:

  1   2t 
  
g  2
 
x 1  2 xt  t 2 3 / 2 n 0

Pn ( x ) t n

82
Unit 4 Special Functions-I
so that

 Pn ( x ) t n
t

(1  2 xt  t 2 ) 3 / 2 n 0

or

 Pn ( x ) t n
t
 (1  2 xt  t 2 )
1  2 xt  t 2 n 0

On combining this result with Eq. (4.9), we get:

 
t  Pn ( x ) t n  1  2 xt  t 2   Pn ( x ) t n
n 0 n 0

On rearranging terms, we obtain:

 
 2x Pn ( x )  Pn ( x ) t n 1   1  t 2 Pn ( x ) t n
n 0 n 0

 
  t n Pn ( x )   Pn ( x ) t n  2
n 0 n 0

As before, on collecting coefficients of t n 1 from both sides and equating

them, we get the second recurrence relation:
2 x Pn ( x )  Pn ( x )  Pn 1( x )  Pn 1( x ) (4.15)

Another very useful recurrence relation is obtained by combining Eqs. (4.14)

and (4.15). To this end, we first differentiate Eq. (4.14) with respect to x and
multiply the result by 2. This gives

2(2n  1) Pn ( x )  2(2n  1)x Pn ( x )  2(n  1) Pn 1( x )  2n Pn 1( x )

We rewrite it as:

2(2n  1)x Pn ( x )  2(n  1) Pn 1( x )  2nPn 1( x )  2(2n  1) Pn ( x )

Next, we multiply Eq. (4.15) by (2n+1). This leads to:

2(2n  1)x Pn ( x )  (2n  1) Pn ( x )  (2n  1) Pn 1( x )  (2n  1) Pn 1( x )

Substituting for 2(2n  1)x Pn ( x ) in this relation, we obtain:

2(n  1)Pn 1( x )  2n Pn 1( x )  2(2n  1) Pn ( x )

 2(2n  1) Pn ( x )  (2n  1) Pn 1( x )  (2n  1) Pn 1( x )

On simplification, we obtain:
Pn 1( x )  Pn 1( x )  (2n  1) Pn ( x ) (4.16)

Eqs. (4.14) and (4.16) are the prime recurrence relations for the Legendre
polynomials. Using these recurrence relations and Legendre’s differential
equation, you can obtain the following relations:
83
Block 1 Mathematical Methods in Physics

Pn 1( x )  (n  1) Pn ( x )  x Pn ( x ) (4.17a)

Pn 1( x )  x Pn ( x )  n Pn ( x ) (4.17b)

(1  x 2 ) Pn ( x )  n Pn 1( x )  nx Pn ( x ) (4.17c)

and (1  x 2 ) Pn ( x )  (n  1) x Pn ( x )  (n  1) Pn 1( x ) (4.17d)

You should note that recurrence relations are identities in x and simplify proofs
and derivations.
Two functions A(x) and
B(x) are said to be SAQ 4
orthogonal on the Prove Eqs. (4.17a to d).
interval (a, b) if they
satisfy the relation
b 4.2.3 Orthogonality Relations
 A( x ) B( x ) dx  0
One of the important characteristics of Legendre polynomials is that they are
a
From your UG courses orthogonal. This property enables us to express a given function defined on
(refer to Block 2 of the interval (–1,1) in a series of Legendre polynomials. It is therefore important
PHE-05), you may for you to master its applications. The mth and the nth order Legendre
recall that sine and polynomials Pm (x ) and Pn (x ), respectively, satisfy the equations:
cosine functions are
orthogonal in the   2x Pm
(1  x 2 ) Pm   m(m  1) Pm ( x )  0 (4.18a)
interval (1, 1). A
function whose norm is and (1  x 2 ) Pn  2x Pn  n (n  1) Pn ( x )  0 (4.18b)
unity, i.e.,
We multiply the first equation by Pn (x ) and the second equation by Pm (x ) .
 f 2 dx  1 is said to Next we subtract the latter product from the former. This leads to the relation:
be normalised. A
  Pm Pn )  2x(Pn Pm
(1  x 2 ) (Pn Pm   Pm Pn )
system of normalised
functions which are  [n(n  1)  m(m  1)] Pm Pn
orthogonal is said to
be orthonormal. You can verify quite easily that the left hand side is equal to
d
(1  x 2 ) (Pn Pm  Pm Pn ) so that we can write
dx
d
(1  x 2 ) (Pn Pm  Pm Pn )   [n(n  1)  m (m  1)] Pm Pn
dx
On integrating both sides over x from x = –1 to x = +1, we obtain:
1


[n(n  1)  m(m  1)] Pm ( x ) Pn ( x ) dx
1

  Pm Pn ] 11
 (1  x 2 ) [Pn Pm

You should note that (1  x 2 ) vanishes for both the limits ( x  1) implying that
the right hand side will be zero always. Further, when m  n, for the above
relation to hold we must have:
1

 Pm ( x )Pn ( x ) dx  0 (4.19)
84 1
Unit 4 Special Functions-I
which means that the scalar product of Legendre polynomials of different
orders (in the range  1  x  1) is zero. Eq. (4.19) constitutes the
orthogonality relation for Legendre polynomials. Let us now evaluate the
1
integral  Pm ( x )Pn ( x ) dx for m  n . In other words, we have to evaluate the
1
1
integral  [Pn ( x )]2 dx .
1

 Pn ( x ) t n
1
From the generating relation we recall that 
1  2t x  t2 m 0


 Pm ( x ) s m
1
We can also write: 
1  2 sx  s 2 m 0

Multiplying these equations and integrating with respect to x, between x  1

and x  1, we get
1
dx
 
1  2t x  t 2  1  2sx  s 2 
1

 1 
 
    Pm ( x ) Pn ( x ) dx  s m t n
m  0 n  0
1 


From Eq. (4.19) we understand that the RHS survives only for terms for which
m  n. We also observe that when we consider terms for which m  n, the
notations t and s become identical and we can write:
1 1 1 
  2n
 1 2 t x  t 2  
dx 2
  [ Pn ( x )] dx t
1 1 
1 


To evaluate the integral on the LHS, we put 1  2 t x  t 2  u so that

du
 2t dx  du or dx   . When x  1, the limit of integration changes to
2t
u  1  2t  t 2  (1  t )2. Similarly, when x  1, we have u  1  2t  t 2  (1  t )2 .
Hence,
(1t )2 (1t )2
du 1 du 1
   ln u (1t )2
2
1  
2 t u 2t u 2t (1t )
(1t ) 2 (1t ) 2

2  1 t  1  1 t 
 ln   ln  
2t  1  t  t  1  t 

We now recall the series expansion of a logarithmic function:

t2 t3 t4
ln (1  t )  t     ...
2 3 4
and
t2 t3 t4
ln (1  t )  t     ...
2 3 4
85
Block 1 Mathematical Methods in Physics
so that

1  1 t  1
ln    ln (1  t )  ln (1  t )
t  1 t  t

1 2t 3 2t 5 
  2t    ...
t  3 5 


 t2 t4  
t 2n

 2 1 
 3 5

 ...  2
 
2n  1
  n 0

  1  
   2  t 2n
Thus,    [Pn ( x )] 2 dx  t 2n    
2n  1 
n 0 
1 
 n 0 

On equating the coefficient of t 2n , we get

1
2
 [Pn ( x )]2 dx  2n  1 (4.20)
1

We can combine Eqs. (4.19) and (4.20) to write

1
2
 Pm ( x ) Pn ( x )  2n  1  mn (4.21)
1

where mn is the Kronechers’s delta, defined as

 m n  0, when m  n 


 1, when m  n 

Eq. (4.21) tells us that the scalar product of a Legendre polynomial of a

particular order with that of another order is zero, whereas it is non-zero for the
product of a Legendre polynomial with itself (in the range  1 x  1) . This
suggests that Legendre polynomials of different orders are orthogonal.

Completeness of Legendre polynomials

In vector analysis, we define a set of orthogonal basis vectors as complete if

there is no other vector orthogonal to them all in the number of dimensions
under consideration. By analogy, we define a set of orthogonal functions as
complete if there is no other function orthogonal to all of them. From UG
courses, you know how to use an infinite series of sine and cosine terms to
express a function, say temperature distribution, in a Fourier series on ( , )
(see Unit 7, Block 2 of PHE-05).

Now you will learn how to expand a function in a series of Legendre

polynomials, which form a complete orthogonal set on (1,1) . The
completeness means that any well-behaved function f(x) can be approximated
to any desired accuracy by a series of Pk (x ) through the relation:

f (x)   Ak Pk ( x )  1  x  1 (4.22)
86 k 0
Unit 4 Special Functions-I
To obtain the coefficient AK , we multiply both sides by Pm (x ) and integrate the
resultant expression in the range  1 to  1:
1  1

 Pm ( x ) f ( x ) dx   Ak  Pm ( x ) Pk ( x ) dx
1 k 0 1

Using orthogonality relation [Eq. (4.21)], we can write:

1 
 Ak  2m  1  km 
2
1 Pm ( x ) f ( x ) dx 
k 0



2 2 Am
 Ak  km 
2m  1 k 0 2m  1

since  km  0 except for k  m and mn  1.

1
2m  1
Hence, Am 
2 
Pm ( x ) f ( x ) dx
1

1
2k  1
or Ak 
2 
Pk ( x ) f ( x ) dx (4.23)
1

We shall now illustrate this with the help of an example.

Example 4.3

1, 0  x 1

Expand the function f ( x )   in a series of the form
0, 1 x  0

 Ak Pk ( x ) dx .
k 0

Solution : From Eq. (4.23), we have

1
2k  1
Ak 
2 
Pk ( x ) f ( x ) dx
1

2k  1  
0 1

2   
 Pk ( x ) f ( x ) dx  Pk ( x ) f ( x ) dx 

1 0 

We are given that

0 for  1  x  0

f (x)  
1 for 0  x  1

On inserting these values in the expression for Ak , we get

1
2k  1
Ak 
2 
Pk ( x ) dx
87
0
Block 1 Mathematical Methods in Physics
Now, refer to Example 4.1 and SAQ 2. We recall that P0 ( x )  1, P1( x )  x,

(3x 2  1) 5x 3  3x
P2 ( x )  , P3 ( x )  , and so on. Therefore,
2 2
1 1
1 1 1
A0 
2 
P0 ( x ) dx 
2
dx 2
0 0

1 1
3 3 3
A1 
2 
P1( x ) dx 
2
x dx 4
0 0

1 1
5 5
A2 
2 
P2 ( x ) dx 
4 
(3 x 2  1) dx  0
0 0

and
1 1
7 7 7
A3 
2 
P3 ( x ) dx 
4 
(5 x 3  3 x ) dx  
16
0 0

11
Proceeding in this way, you will find that A4  0, A5  , and so on. Thus,
32
we can write:
1 3 7 11
f (x)  P0 ( x )  P1( x )  P3 ( x )  P5 ( x )  ...
2 4 16 32
You may now like to solve SAQ 5.

SAQ 5

a) In Example 4.3 you must have observed that Ak  0 for even k  0 . Re-
establish this result using the recurrence relation given by Eq. (4.16).

b) Expand f ( x )  x 2 in a series of the form  Ak Pk ( x ) .
k 0

You have learnt how to obtain Legendre polynomials from the generating
function. There is another simple way of arriving at the Legendre polynomials.
This is through Rodrigues’ Formula, which we now discuss.

4.2.4 Rodrigues’s Formula

Let us consider the function

v  ( x 2  1) n

and differentiate it with respect to x. This gives

dv
 n( x 2  1) n 1  2x  2nx( x 2  1) n 1
dx

so that
dv
( x 2  1)  2nx( x 2  1) n  2nxv
88 dx
Unit 4 Special Functions-I
That is, v satisfies the differential equation
dv
(1  x 2 )  2nx v  0
dx
Differentiating it again with respect to x, we get:
d 2v dv dv
(1  x 2 )  2x  2nx  2nv  0
dx 2 dx dx
or
d 2v dv
(1  x 2 )  2 (n  1)x  2nv  0 (4.24)
dx 2 dx

If you differentiate it again with respect to x, you can write the resultant
expression as:
d 21v d 11v dv
(1  x 2 )  2x  2(n  1)
dx 21 dx 11 dx

d 11v dv
 2(n  1)x  2n  0.
dx 11 dx

This can be re-arranged and written in a compact form as:

d 21v d 11v dv
(1  x 2 )  2x(n  1  1)  (1  1) (2n  1) 0
dx 21 dx 11 dx

After r differentiations, you will get

d 2 r d 1r v drv
(1  x 2 )  2 x(n  r  1)  (r  1) (2n  r ) 0
dx 2r dx 1r dx r
When r  n, we get the nth derivative of Eq. (4.24):

d n 2v d n 1 d nv
(1  x 2 )  2x  n(n  1) 0
dx n 2 dx n 1 dx n
This equation can be rewritten as:
d 2  d nv  d  d nv  d nv
(1  x 2 )    2 x    n(n  1) 0
dx 2  dx n  dx  dx n  dx n

On comparing it with Legendre’s differential equation

d 2y dy
(1  x 2 )  2x  n(n  1) y  0
dx 2 dx

you will note that d nv / dx n satisfies the Legendre’s equation. So can write

d nv
Pn ( x )  C
dx n
d nv
That is, Pn (x ) is a constant multiplier times . On substituting for v, we
dx n
obtain
dn
Pn ( x )  C ( x 2  1) n
dx n 89
Block 1 Mathematical Methods in Physics
where C is a constant. To determine this constant, we have to consider terms
with the highest power of x on both sides. From Eq. (4.8) we recall that the
(2n )!
term with the highest power of x it the expression for Pn (x ) is xn.
2 (n! ) 2
n

Hence,
(2n )! dn
xn  C x 2n  C.2n(2n  1) (2n  2)...[2n(n  1)] x n
2 n (n! ) 2 dx n

(2n )! n
C x
n!

On comparing the coefficients of x n on both sides, we get:

1
C
(2 n ) n!

Hence, we can write:

1 dn
Pn ( x )  ( x 2  1) n (4.25)
(2 n )n! dx n

This is known as the Rodrigues’ formula for Pn (x ) . We now consider an

application of this formula.
Example 4.5
Obtain the value of P3 ( x ) using Rodrigues’s formula.

Solution
From Eq. (4.25), we can write:

P3 ( x ) 
1 d3  1  d 3 ( x 6  3 x 4  3 x 2  1)
( x 3  1) 3   
(2 3 )3! dx 3  48  dx 3

Differentiating the expression in the small brackets with respect to x, you will
get:
d
( x 6  3 x 4  3 x 2  1)  6 x 5  12x 3  6 x
dx
We differentiate again the resultant expression with respect to x and get:
d2 d
( x 6  3 x 4  3 x 2  1)  (6 x 5  12x 3  6 x )
dx 2 dx
 30 x 4  36 x 2  6

and
d3 d
( x 6  3 x 4  3 x 2  1)  (30x 4  36x 2  6)
dx 3 dx
 120 x 3  72 x

P3 ( x )   .24 (5x 3  3x )  (5x 3  3x )

1 1

 48  2
90
Unit 4 Special Functions-I
You may now like to solve SAQ 6.

SAQ 6
Obtain P4 ( x ) using Eq. (4.25).

We have discussed the basic operations involving Legendre polynomials. We

now intend to discuss their applications in physics. The most instructive
applications arise while solving Laplace’s equation in spherical polar
coordinates for potential and temperature related problems. You should go
through the following examples carefully as you can learn a lot of good
physics.
Example 4.6
z
A conducting sphere is placed in a uniform electric field of strength E0 , as
shown in Fig. 4.2. Calculate the electrostatic potential at a point outside the
sphere. E0
Solution : We note that due to spherical symmetry, the potential function will
be independent of the azimuthal angle  . From Unit 1, recall that Laplace’s V=0
equation can be split into the following differential equations: Fig. 4.2: A conducting
d 2R dR sphere placed in a
r2  2r  l (l  1)R  0 (i) uniform electric field.
dr 2 dr
and
1 d  d 
 sin    l (l  1)   0 (ii)
sin  d  d 
where l (l  1) is the separation constant. You have seen that Eq. (i) of these
equations is the radial part of Laplace’s equation and has solutions of the
form:
Bl
Rt (r )  Al r l 
rl 1

Note that Eq. (ii) is Legendre’s equation, which has Legendre polynomials as
solutions. Hence, the solution of Laplace’s equation with azimuthal symmetry
(no  dependence) is given by:

  Al r l  r l 1  Pl (cos )
B
V ( r , )  (iii)
t 0

To determine the constants An and Bn , we have to impose boundary

conditions. Since the original uniform electric field (before the sphere is placed
in electric field) is E0 , we must have In this case, only the
potential outside the
V (r   )  E 0 z  E 0 r cos  (z  r cos ) sphere is relevant as
or potential inside a
V  E 0 r P1 (cos ) (P1( x )  x ) (iv) conducting sphere is
zero.
As r  , the second term in the bracket on the right side of Eq. (iii) will
disappear. Thus, on comparing Eqs. (iii) and (iv), we get:
A0  0, An  0 for all n > 1 91
Block 1 Mathematical Methods in Physics
and

A1  E 0 (v)

For the second boundary condition, we choose the surface of the sphere to be
at zero potential. Thus, from Eq. (iii) we have:

 B  P (cos )
V (r  a )   A0  0    1  E 0 a  P1(cos )  
B
Bt t 0
 a  a 2  l 1 a l 1

If this equality is to hold for all values of , each coefficient of Pl (cos) must
vanish. Hence, we have:

A0  B0  0, Bn  0 for n  2

and B1  E 0 a 3

Inserting these results in the above expression, we find that the electrostatic
potential is given by:

E0a 3
V  E 0 r P1(cos )  P1(cos )
r2
 a 
3
 E 0 r cos  1  
 r3 

You should now solve SAQ 7 for practice.

SAQ 7

For a sphere of radius a such that V (r , ) r a  V0 cos 3 , obtain the potential

at a point inside the sphere. Assume that there are no charges at the origin.

We now discuss the spherical harmonics, which forms the basis of many
important concepts in quantum mechanics, especially, those related to the
structure of atom.

4.3 SPHERICAL HARMONICS

In Unit 1, you have encountered Helmholtz equation in Cartesian coordinates
and separated it into three ODEs:

 2 2 2 
    f  x, y , z   k 2 f  x, y , z   0 (4.26a)
 x 2 y 2 z 2 

Let us write this equation in spherical polar coordinates:

1    2 f    f  1  2f 
 sin   r    sin      k 2f
r sin 
2           2
 r r sin 

(4.26b)
92
Unit 4 Special Functions-I
You can separate Eq. (4.26b) into three ODEs by putting f (r, , ) and then
divide by R(r) () (). So, you will get:

1 d  2 dR  1 d  d 
r   sin  
Rr dr 
2 dr  r sin  d 
2 d 

1 d 2
  k2 (4.27a)
r 2 sin2  d2

Multiplying the above equation by r 2 sin2 , we can write:

1 d 2
 r 2 sin 2   k 2 
1 d  2 dR 
r 
 d 2  Rr 2 dr  dr 

1 d  d 
  sin   (4.27b)
r 2 sin  d  d 

The LHS of Eq. (4.27b) is a function of  alone and the RHS, a function of r
and . Since r,  and  are independent variables, we put the LHS and RHS of
Eq. (4.27b) equal to a constant:
d 2 d 2
  m 2 or  m 2  0 (4.28a)
d 2 d 2

The normalized solution of Eq. (4.28a) is:

1
 m ()  e im (4.28b)
2
The function  m ( ) is orthonormal with respect to the variable . Note that we
have used a negative constant because in problems in physics,  appears
mostly as the azimuth angle, which suggests periodic solutions rather than
exponential ones. Substituting Eq. (4.28a) in Eq. (4.27b), we can write:

1 d  2 dR  1 d  d  m2
r   r 2k 2    sin  
R dr  dr   sin  d  d  sin 2 
(4.29a)

You can see that in Eq. (4.29a), the variables are separated. Again, we equate
each side of Eq. (4.29a) to a constant, say C. Then we have:

1 d  d  m 2
 sin    C  0, (4.29b)
sin  d  d  sin 2 
and
1 d  2 dR  CR
r   k 2R  2  0 (4.29c)
r dr 
2 dr  r

Note that Eq. (4.29b) is the same as Eq. (1.28c) and now we will solve it. For
simplicity of calculations, we put C  l ( l  1) as we did in Unit 1 while obtaining
the solution given by Eq. (1.29b). Let us recast Eq. (4.29b)as follows:

d 2 d
sin 2   sin  cos   m 2  l (l  1) sin 2    0 (4.29d)
d 2 d
93
Block 1 Mathematical Methods in Physics
This equation is called the Legendre’s associated differential equation. If
we set m  0, i.e., there is no  dependence, then Eq. (4.29d) reduces to
Legendre’s differential equation. We can transform Eq. (4.29d) by introducing
a change of variables:
x  cos 

so that sin 2   1  x 2 and by the chain rule:

d d dx d
 .   sin 
d dx d dx

d d d
 sin  cos    sin 2  x   x 1  x 2 
d d dx

Similarly, we can write:

d 2 d  d  d  d 
      sin  
d 2 d  d  d  dx 

d d 2  dx
  cos   sin 
dx dx 2 d

d 2 d d 2 d 2 d
so that   cos   sin 2   1  x 2  x
d 2 dx dx 2 dx 2 dx

On inserting these results in Eq. (4.29d), we get

1  x 2  1  x 2  d d 
2 d
x   x 1  x 2 
 dx 2 dx  dx

 m2  l (l  1) 1  x 2   0

Dividing throughout by (1  x 2 ) , we get Legendre’s associated differential

equation in x:
2 d  
1  x 2  d  2x   l (l  1) 
m2
   0 (4.30)
dx 2 dx  1 x 2 

We would like to point out here that solutions of Legendre’s associated

differential equation are labelled by both parameters l and m and written as
Plm (x ) .

For non-negative integral values of m and l, the general solution of Eq. (4.30)
is given by:

  C1Plm ( x )  C 2Q m
l (x)

 C1Plm (cos )  C2 Qlm (cos ) (4.31)

Note that Plm (x ) is finite for  1  x  1(which conforms with x  cos ) and
Qlm (x ) is unbounded for x  1 . For this reason, we consider only Plm (cos)
as an acceptable solution and that too with integral values of m and l.
94
Unit 4 Special Functions-I
The associated Legendre polynomials are connected with Legendre
polynomials through the relation:

Plm ( x )  1  x 2 
m/2 dm
Pl ( x ) (4.32)
dx m

where Pl (x ) is the lth Legendre polynomial. You must note that if m  l ,

Plm ( x )  0.

Orthogonality
Just like Legendre polynomials, the associated Legendre polynomials Plm (x )
are also orthogonal in the range  1  x  1 . Mathematically, we write
1 2 (l  m)!
1 Plm ( x ) Pkm ( x ) dx  2l  1 (l  m)!  kl (4.33)

As before, we can expand f(x) in a series of the form:


f (x)   Ak Pkm (z) (4.34a)
k 0

where
1
(2k  1) (k  m )!
Ak 
2(k  m )! 
f ( x ) Pkm ( x ) dx (4.34b)
1

From Eq. (4.33), we can write the normalized associated Legendre

polynomials as:

2n  1 (n  m )! m
m
n (cos )  Pn (cos )  n  m  n
2 (n  m )!
(4.34c)
We are now ready to define spherical harmonics. Note that the function
 m ( ) is orthonormal with respect to the variable  (the azimuthal angle) and
the function m
n (cos) is orthonormal with respect to the variable  (the polar
angle). We define spherical harmonics as the product of these two functions:

2n  1 (n  m )! m
Ynm (, )  Pn (cos )e im (4.35a)
4 (n  m )!

Thus, spherical harmonics Ynm (, ) are functions of the two angles  and ,
which are orthonormal over the spherical surface. Sometimes, a phase factor
(1)m known as the Condon Shortley phase is used in the expression for
spherical harmonics, which then becomes:

2n  1 (n  m )! m
Ynm (, )  ( 1) m Pn (cos )e im (4.35b)
4 (n  m )!

Eq. (4.35b) is useful in quantum theory of angular momentum. From

Eq. (4.35b), the first few spherical harmonics are:

1
Y00 (, )  (4.36a)
4 95
Block 1 Mathematical Methods in Physics
3
Y10 (, )  cos  (4.36b)
4

3
Y11 (, )   sin  e i (4.36c)
8

3
Y11 (, )   sin  e i (4.36d)
8
and so on.
The orthogonality relation for spherical harmonics is given as:
2 
m1 
  Yn 1
(, )Ynm2 (, ) sin  d d   n1, n2  m1, m2 (4.37)
2
0 0

You may like to know why this nomenclature is used for these functions.
Firstly, these functions are defined over the surface of a sphere with , the
polar angle, and , the azimuth. The term “harmonic” is included in the
nomenclature because solutions of Laplace’s equations were called harmonic
functions and the spherical harmonics represent the angular part of such
solutions. You may like to solve an SAQ now.

SAQ 8

Write the expressions for the spherical harmonics Y20 (, ), Y21 ( , ) and
Y22 (, ).

In the last section of this unit, we discuss the hypergeometric functions.

4.4 HYPERGEOMETRIC FUNCTIONS

Hypergeometric function is a special function represented by the
hypergeometric series, which is a generalization of a geometric series. The
hypergeometric series has the form of a power series in which the coefficients
are replaced by ratios of rational functions of constants. Hypergeometric
function is the solution of the following differential equation:

x(1  x ) y  ( x )  [c  (a  b  1)x ] y ( x )  ab y ( x )  0 (4.38)

where a, b and c are constants. Eq. (4.38) is also called Gauss’s

hypergeometric equation. You can verify that Eq. (4.38) has regular
singularities at x  0 and x  1. For solving Eq. (4.38), we cast it in the
standard form of a second order linear ODE:

y   p( x )y   q( x )y  0 (4.39a)
[c  (a  b  1)x ] ab
where p( x )  and q( x )   (4.39b)
x(1  x ) x(1  x )
Let us determine the solutions of Eq. (4.39a) for the singularities using the
Frobenius method (refer to Sec. 3.4, Unit 3 of PHE-05, IGNOU B. Sc. Course).
To apply this method, we write Eq. (4.39a) as:

96 x 2 y   x 2 p( x )y   x 2q( x )y  0
Unit 4 Special Functions-I

or x 2 y   xb ( x )y   d ( x )y  0 (4.39c)

where
[c  (a  b  1)x ]
b( x )  xp ( x ) 
(1  x )

abx
and d ( x )  x 2 q( x )   (4.40)
(1  x )

We expand the functions xp(x) and x 2q( x ) in powers of x as follows:

c  (a  b  1)x
b( x )  xp ( x ) 
(1  x )

 [c  (a  b  1)x ] (1  x  x 2  ...)   bn x n (4.41a)
n 0

abx
and d ( x )  x 2 q( x )  
(1  x )

  abx (1  x  x 2  ...)   dn x n (4.41b)
n 0

Let us now solve Eq. (4.39a) for both regular singular points.
Solution for the regular singularity at x  0

We expand y in the following series about x  0, and take its first and second
order derivatives with respect to x:

y (x)   am x mk
m 0


y ( x )   am (m  k )x mk 1
m 0


y ( x )   am (m  k ) (m  k  1)x mk 2
m 0

We substitute y(x) and its derivatives in Eq. (4.39c) and write:


y (x)  x k  (m  k ) (m  k  1) am x m
m 0

    
 xk  
m  0
(m  k ) am x m  
  n  0
bn x n 


    
 xk  
m  0
am x m  
  n  0
dn x n   0

(4.42)

We now determine the indicial equation by equating the coefficient of x k in

Eq. (4.42) to zero. It is the lowest power of x in the equation obtained by
putting m  0 in it. Thus, 97
Block 1 Mathematical Methods in Physics
Coefficient of x k is: k (k  1)  kb0  d 0  0

From Eq. (4.41a), b0  c and d 0  0

Hence, the indicial equation is:

k (k  1)  kc  0  k  0 and k  1  c

k  0 corresponds to a Frobenius series solution if 1 c  0 or the

difference (1  c )  0  1  c is not a positive integer. Note that the second
condition implies the first one, so if 1  c is not a positive integer, then c is
neither zero nor negative integer. Then k  0 corresponds to a Frobenius
series solution of the form

y  x0  an x n  a0  a1x  a2 x 2  ... (4.43)
n 0

where a0  0. Substituting Eq. (4.43) into Eq. (4.38) and equating the
coefficients of x n to zero, we get:
ab (a  1)(b  1) a(a  1)b(b  1)
a1  a0 ; a2  a1  a0
c 2(c  1) 2c(c  1)
(a  n )(b  n )
an1  an (4.44)
(n  1(c  n )
With these coefficients and by letting a0  1, the solution becomes:

ab a(a  1)b(b  1) 2
y  1 x x  ...
c 2c(c  1)

a(a  1)...(a  n  1)b(b  1)...(b  n  1) n
 1  n! c(c  1)...(c  n  1)
x (4.45)
n 1

This is known as the hypergeometric series, and is denoted by the symbol

F (a, b, c, x ). It is called by this name because it generalises the familiar
geometric series as follows:
When a  1 and c  b we obtain:
1
F (1, b, b, x )  1  x  x 2  ...  (4.46)
1 x
If either a or b is either zero or negative integer, the series (4.46) breaks off
and is a polynomial; otherwise the ratio tests shows that it converges for
|x| < 1, since Eq. (4.45) gives:
a n 1x n 1 (a  n ) ( b  n )
 x  x as n  
an xn (n  1) (c  n )

When c is neither zero nor negative integer, F (a, b, c, x ) is an analytic

function called hypergeometric function on the interval x  1.

It is the simplest particular solution of the hypergeometric equation and it

remains unchanged when a and b are interchanged: F (a, b, c, x )  F (b, a, c, x )

98
Next we consider the solution corresponding to k  1  c.
Unit 4 Special Functions-I
If 1  c is neither zero nor negative integer, i.e., c is not a positive integer, then
there is a second independent solution of Eq. (4.38) near x  0 for k  1 c.
We can find this solution by substituting

y  x 1c (a0  a1x  a2 x 2  ...)

in Eq. (4.38) and calculating the coefficients. The other way of finding the
solution is to change the dependent variable in Eq. (4.38) from y to z by writing
y  x 1c z .

After mathematical manipulation, Eq. (4.38) becomes:

x(1  x )z   [(2  c )  ([a  c  1]  [b  c  1]  1)x ]z 
 (a  c  1)(b  c  1)z  0 (4.46)

which is the hypergeometric equation with the constants a, b and c replaced

with a  c  1, b  c  1 and 2  c.

We already know that Eq. (4.46) has the solution:

z  F (a  c  1, b  c  1, 2  c, x )

near the origin, so our desired second solution is:

y  x 1c z  x 1c F (a  c  1, b  c  1, 2  c, x )

So, when c is not an integer, we have two independent Frobenius

series solutions and hence,
y  c1F (a, b, c, x )  c 2 x 1c F (a  c  1, b  c  1, 2  c, x ) (4.47)

is the general solution of the hypergeometric equation (4.38) near the

singular point x  0. Note that the above solution is only valid near the origin.
We now solve Eq. (4.38) near the other singular point x  1.
Solution for the regular singularity at x  1

The simplest way in which we can obtain this solution from the one we have
already determined, is by introducing a new independent variable t  1 x
Hence, x  1  t, dy / dx   dy / dt and d 2 y / dx 2  d 2 y / dt 2 .

Then x  1 corresponds to t  0 and transforms Eq. (4.38) into:

t (1  t ) y   [(a  b  c  1)  (a  b  1)t ] y   aby  0

where the primes denotes the derivatives with respect to t. Since the above
equation is a hypergeometric equation, its general solution near
t  0 can be written down at once from Eq. (4.47) by replacing x by t and c by
a  b  c  1 and then we replace t by 1  x to get the general solution of
Eq. (4.38) near x  1.
y  c1F (a, b, a  b  c  1,1  x )

 c 2 (1  x )c ab F (c  b, c  a, c  a  b  b  1,1  x ) (4.48)

In this case it is necessary to assume that c  a  b is not an integer.
Eqs. (4.47 and 4.48) show that the adaptability of the constants in
Eq. (4.38) makes it possible to express the general solution of this
equation near each of its singular points in terms of the single function F. 99
Block 1 Mathematical Methods in Physics
Any differential equation in which the coefficients of y , y  and y are
polynomials of degree 2, 1 and 0, respectively, and also the first of these
polynomials has distinct real roots, can be brought into the hypergeometric
form by a linear change of the independent variable (TQ 5). Thus, such ODEs
can be solved near their singular points in terms of the hypergeometric
function.
Let us now summarise the unit.

4.5 SUMMARY
In this unit, we have covered the following concepts:
 Legendre equation and its solutions by the Frobenius method.
 Legendre polynomials, generating function, recurrence relations and
Rodrigues’s formula.
 Spherical harmonics.
 Hypergeometric equation, hypergeometric functions.

4.6 TERMINAL QUESTIONS

1. Calculate the values of P2n (0) and P2n 1(0 ).

2. Evaluate the integrals

1
a)  x Pn 1( x ) Pn ( x )dx;
1

1
b)  xPn ( x ) dx
1

3. Determine the steady-state temperature inside a sphere of radius a

given that its upper hemisphere is maintained at a temperature
T  T0 and the lower hemisphere is maintained at temperature
T  T0 .

4. Prove Eq. (4.37).

5. With a suitable change of variable, obtain the solution of the
following ODE near x  0 :
y
(1  e x ) y    ex y  0
2

Hint: Change the independent variable, e x  t.

4.7 SOLUTIONS AND ANSWERS

Self-Assessment Questions
1. We write Eq. (4.1) as follows and use power series method to solve it:

2x n(n  1)
y   y  y 0 (4.1)
(1  x2) (1  x 2 )
100
Unit 4 Special Functions-I
2x n(n  1)
Hence, p( x )   and q( x )  .
(1  x 2 ) (1  x 2 )
The functions p(x ) and q (x ) have regular singularities at the points x  1
and x   1. We use power series method to obtain the solution of this
equation. Substituting

y (x)   am x m
m 0

y ( x )   am mx m1
m1


y ( x )   am m (m  1)x m2
m 2

in Eq. (4.1), we get:


(1  x 2 )  m (m  1)am x m2  2x
m 2
 
 mam x m 1  n(n  1)  am x m  0
m 1 m 0
or
  
 m (m  1)am x m  2   m (m  1)am x m  2  mam x m
m 2 m 2 m 1


 n(n  1)  am x m 0
m 0
th
Collecting the coefficient of the m power of x in the above equation, we
can write:
(m  2) (m  1)am  2  m (m  1)am  2ma m  n(n  1)am  0
or
m (m  1)  2m  n(n  1)
am  2  am
(m  2) (m  1)
m (m  1)  n(n  1)
 am
(m  2) (m  1)
(n  m) (m  n  1)
 am , m  0, 1, 2,...
(m  2) (m  1)

From this recurrence relation for the coefficients an , we can write:

n(n  1) (n  1) (n  2)
a2   a0 , a3   a1,
2! 3!

(n  2) (n  3) (n  2) n (n  1)(n  3)
a4   a2  a0 ,
12 4!
( n  3 ) ( n  4) (n  3) (n  1) (n  2) (n  4)
a5   a3  a1,
20 5!
and so on. 101
Block 1 Mathematical Methods in Physics
Substituting these values in the series for y, we get Eq. (4.2):
n(n  1) 2 n(n  2)(n  1)(n  3) 4
y  a0 1  x  x  ...
 2! 4! 
(n  1)(n  2) 3 (n  1)(n  3)(n  2)(n  4) 5
 a1  x  x  x  ...
 3! 3! 

P3 ( x ) 
6! x3  3  2 x
2.  
2 3 (3! ) 2  25 
1 2  3  4  5   5 x 3  3 x  1
    5 x 3  3 x 
866  5  2

3. a) Putting x  1 in Eq. (4.12), we get


1  2t  t 2    Pn  1 t n
n 0

or 1  t  1   Pn (1)t n
n 0

 1  t  t 2  ...  ( 1) n t n  ...   Pn (1) t n
n 0
 Pn ( 1)  ( 1) n

c) We observe that the generating function remains unchanged even if

we replace x by  x and t by  t. Thus,


g (t, x )  g ( t, x )  1  2(t ) ( x )  (t ) 2 1/ 2
 
  Pn (  x ) ( t ) n   (1) n Pn ( x ) t n
n 0 n 0
 
so that  Pn ( x ) t n   (1) n Pn ( x ) t n
n 0 n 0

 Pn ( x )  (1)n Pn ( x )

or Pn ( x )  (1) n Pn ( x )

4. Upon adding Eqs. (4.15) and (4.16), we get

2Pn 1( x )  2(n  1) Pn ( x )  2 x Pn ( x )

 Pn 1( x )  (n  1) Pn ( x )  x Pn ( x )

which is same as Eq. (4.17a).

On subtracting Eq. (4.16) from Eq. (4.15), we get
2Pn 1( x )  2n Pn ( x )  2 x Pn ( x )

 Pn 1( x )  xPn ( x )  n Pn ( x )

which is identical with Eq. (4.17b).

In Eq. (4.17a), we replace n by (n – 1). This gives
Pn ( x )  nPn  1( x )  x Pn 1( x )
102
Unit 4 Special Functions-I
Now, multiplying Eq. (4.17b) by x, we get

xPn 1( x )  x 2 Pn ( x )  nx Px ( x )

On comparing the above two equations, we get

1  x 2 Pn ( x )  n Pn1( x )  nxPn ( x )

which is Eq. (4.17c).
Upon replacing n by (n  1) in this expression, we get

1  x 2 Pn 1( x )  (n  1)Pn  (n  1)x Pn1( x )

Using Eq. (4.17a), we get

1  x 2  (n  1)Pn ( x )  x1  x 2  Pn ( x )

 (n  1) Pn ( x )  (n  1)x Pn 1( x )

or  (n  1)x 2 Pn ( x )  x 1  x 2 Pn ( x )  (n  1) x Pn 1( x )

Since x  0 (in general), we get

1  x 2 Pn ( x )  (n  1) x Pn ( x )  (n  1) Pn1( x )
which is the result contained in Eq. (4.17d).
2k  1 1
5. a) Ak 
2 0Pk ( x ) dx

Using Eq. (4.16), we can write

1 1 1
Ak  
2 0
Pk 1( x )  Pk 1( x ) dx  Pk 1( x )  Pk 1( x ) 10
2
1
 Pk 1(1)  Pk 1(1)  Pk 1(0)  Pk 1(0)
2
1
 Pk 1(0)  Pk 1(0)
2
Hence, for even k (other than k  0), Ak  0

b) f (x)  x 2   Ak Pk ( x )
k 0

We have to find Ak , k  0, 1, 2, 3,... such that

x 2  A0 P0 ( x )  A1P1( x )  A2 P2 ( x )  A3 P3 ( x )  ...

 3 x 2  1  5x 3  3x 
 A0 (1)  A1( x )  A2    A3    ...
 2   2 
Since the left hand side of the above equation is a polynomial of
degree 2, we must have A3  0, A4  0, A5  0, and so on.

x 2   A0  2   A1x  A2 x 2
A 3

 2  2
A 3
 A0  2  0, A1  0, and A2  1
2 2 103
Block 1 Mathematical Methods in Physics
1 2
Thus, A0  , A1  0 and A2 
3 3
1 2
 x2  P0 ( x )  P2 ( x )
3 3
Alternatively, you can calculate A k s using the relation

2k  1 1 2
Ak 
2 1 
x Pk ( x ) dx

6. P4 ( x ) 
1 d4
x 2  14
2 4 4! dx 4


1 d4
x 8  4x 6  6x 4  4x 2  1
16  24 dx 4


1
8  7  6  5x 4  4  6  5  4  3x 2  6  4  3  2  1
16  24


1
35x 2  30x 2  3
8

7. We consider a sphere of radius a such that V ( r , ) r a  V0 cos 3  and

assume that there are no charges at the origin. Since V must satisfy
Laplace’s equation and the boundary condition has no  dependence, the
solution will be obtained in terms of Legendre Polynomials. We write the
general solution as:

  An r n  t nn1  Pn (cos )
B
V ( r , )  (i)
n 0

From this form of the solution, we note that V can be finite at the origin
only if Bn  0 for all n. Then Eq. (i) reduces to:

V ( r , )   An r n Pn (cos ) (ii)
n 0

On applying the given boundary condition, we have


V (r , )  V0 cos 3    An a n Pn (cos ) (iii)
m 0

To solve for Am we rewrite cos 3  in terms of Legendre polynomials.

For this, we recall that P3 (cos )  (5 cos 3   3 cos ) / 2 and then we write

2 3
cos 3   P3 (cos )  cos 
5 5
Since cos   P1(cos ), we can write:

2 3
cos 2   P3 (cos )  P1(cos )
5 5
Inserting this result in Eq. (iii) we obtain:


1
[2V0 P3 (cos )  3V0 P1 (cos )]  An a n Pn (cos ) (iv)
5 m 0
104
Unit 4 Special Functions-I
Using the orthogonality property of Legendre polynomials, you can see
that A1  3 V0 / 5a and A3  2V0 / 5a 3 . Hence

3 2
V (r , )  V0 (r / a) P1 (cos )  V0 (r / a)3 P3 (cos )
5 5

8. Using Eq. (4.35b), we get:

2n  1 (n  m )! m
Ynm (, )  ( 1) m Pn (cos )e im
4 (n  m )!

5 3 1
Hence, Y20 (, )   cos 2   
4  2 2

5
Y21 (, )   3 sin  cos  e i ,
24

5
and Y22 (, )  3 sin 2  e 2i
96
Terminal Questions
1. Putting x  0, in Eq. (4.12), we get

(1  t 2 ) 1/ 2   Pn (0)t n
n 0

Using binomial expansion, we can write

1 2 3 4
(1  t 2 ) 1/ 2  1  t  t  ...
2 8
1  3  5...  (2n  1) 2n
 ( 1) n t  ...
2 n n!

 P0 (0)  P1(0) t  P2 (0)t 2  ...  P2n (0)t 2n

On comparing the powers of t 2n , we get

1  3  5...  (2n  1)
P2n (0)  ( 1) n
2 n n!

Since in the expansion of (1  t 2 ) 1/ 2 we only get even powers of t, we

have:
P2n 1(0)  0

2. a) Starting from the recurrence relation:

(n  1) Pn 1( x )  (2n  1)xPn ( x )  nPn 1( x )  0

we multiply both sides by Pn1( x ) and integrate the terms with respect
to x between the limits 1 and  1. This yields:
1 1
(n  1) 1Pn1( x ) Pn1( x )dx  (2n  1)1 xPn ( x )Pn1( x ) dx
1
n 1 Pn21( x ) dx  0 105
Block 1 Mathematical Methods in Physics
From Eq. (4.21), we have:
1

 Pn 1( x ) Pn 1 ( x ) dx  0
1

1 2 2
and 1 Pn21( x ) dx  2(n  1)  1  2n  1
1 2n
 0  (2n  1) 1 xPn ( x )Pn1( x ) dx  2n  1  0
On rearranging terms, we get
1 2n
(2n  1) 1 xPn ( x )Pn1( x ) dx  2n  1
1 2n
so that 1 xPn ( x )Pn1( x ) dx  (2n  1) (2n  1)
1
b) To evaluate the integral  x Pn ( x ) dx, we note that x  P1( x ), so that
1
1
1
 x Pn ( x ) dx  1P1( x ) Pn ( x ) dx. Therefore, the orthogonality relation
1
1
for Legendre polynomials [Eq. (4.21)] implies that  x Pn ( x ) dx  0 for
1
2 2
n  1 and  for n = 1.
2 1 1 3
3. We need to solve Laplace’s equation in spherical coordinates subject to
the given boundary conditions. You know the solution of azimuthally
symmetric Laplace’s equation in spherical coordinates. Since temperature
distribution is independent of , we can write

  Am r m  r mm1  Pm (cos )
B
T ( r , )  (i)
m 0

Since temperature will be finite at the centre of the sphere (r = 0), we must
have B m  0, for all m; otherwise the solution will diverge. Hence, Eq. (i)
reduces to:

T ( r , )   Am r m Pm (cos ) (ii)
m 0

As per the given boundary conditions, we can write the temperature

distribution on the surface of the sphere as:
 T0 1 x  0

f (x)   (iii)

 T0 0  x 1
where x  cos  . Applying the boundary conditions to Eq. (ii), we get

f (x)   Am a m Pm ( x ) (iv)
m 0
106
Unit 4 Special Functions-I
To determine the constants Am , we use the orthogonality relation. So, we
multiply both sides of Eq. (iv) by Pl (x ) and integrate the resultant
expression over x in the range –1 to +1. This yields:

 Am a m 1 Pl ( x ) Pm ( x ) dx
1 1
1 Pl ( x ) f ( x ) dx  (v)
m 0

Using the orthogonality relation for the Legendre Polynomials, we get

2l  1 1
Al 
2a l 1 
Pl ( x ) f ( x ) dx

2l  1
T0  Pl ( x ) dx  Pl ( x ) dx 
0 1

2a l  1  0 
2l  1  T0   1
  
1
 2  a   0 
  l   Pl ( x ) dx  Pl ( x ) dx 
0   (vi)

Since Pl (  x )  ( 1)l Pl ( x ), the above expression simplifies to

(2l  1) (T / a l ) 1P ( x ) dx

 0
0
l  l  odd
Al   (vii)


0 l  even

From this you can readily write the values of first few coefficients:
3T0 1 3T 1 3T
A1 
a 0 
P1 ( x ) dx  0
a 0
x dx 
2a 
7T0 1 7T 1  5 x 3  3 x  7T
A3 
2 0 
P3 ( x ) dx  0 
a 3 0  2 

 dx   0
8a 3

Hence, the temperature distribution inside the given sphere is:

 3r 7r 3 
T (r , )  T0  cos   (5 cos 3   3 cos )  ...  (viii)
3
 2a 16a 

2n  1 (n  m)! m
4. Substituting Ynm (, )  Pn (cos )eim from Eq. (4.35a) in
4 (n  m)!
Eq. (4.37), we get:

2n1  1 (n1  m1 )! 2n 2  1 (n 2  m 2 )!
4 (n1  m1 )! 4 (n 2  m 2 )!
2 

  Pn (cos )e im1 Pnm2 (cos )e im2 sin  d d

m1
1 2
0 0

  n1, n2  m1, m2 (i)

Let us first consider the integral over , which is:

2

 e im  e im  d
1 2  2 m1,m2 (ii)
0

as you know from your UG integral calculus. Substituting Eq. (ii) in Eq. (i),
its LHS becomes: 107
Block 1 Mathematical Methods in Physics
(n1  m1)! (n  m2 )!
(2n1  1) (2n2  1) 2
(n1  m1)! (n2  m2 )!


 Pn
m1
(cos ) Pnm2 (cos ) sin  d (iii)
1 2
0

From Eq. (4.33), we have:

1 2 (l  m)!
1 Plm ( x ) Pkm ( x ) dx  2l  1 (l  m)!  kl
With x  cos , the above equation becomes:
 2 (l  m)!
0 Plm (cos ) Pkm (cos ) sin  d  2l  1 (l  m)!  kl (iv)

Using Eq. (iv) in Eq. (iii), we get:

(n1  m1)! (n  m2 )!
(2n1  1) (2n2  1) 2
(n1  m1)! (n2  m2 )!


 Pn
m1
(cos ) Pnm2 (cos ) sin  d   n1, n2
1 2
0

Combining the results of Eqs. (iv) and (ii), we get the orthonormality
relation of the spherical harmonics [Eq. (4.37)].
5. We change the variable as e x  t, and have:
dy dy dt dy dy
  ex t
dx dt dx dt dt
d 2y d  t dy   t dy  t 2 d 2 y
and  t  
dx 2 dt  dt  dt dt 2
The ODE then becomes:
 dy d 2 y  t dy
(1  t )  t  t2    ty  0
 dt dt 2  2 dt

d 2y
   t 
3 dy
or t (1  t )  y 0
dt 2 2  dt
Comparing this equation with the hypergeometric equation [Eq. (4.38)], we
find that:
3
c  , (a  b  1)  1, ab   1  a  1, b   1
2
The general solution as per Eq. (4.48) at t  1 or x  0 is:
y  c1F (a, b, a  b  c  1,1  t )

 c2 (1  t )c a b F (c  b, c  a, c  a  b  1, 1  t )
1
or y  c1F (1,  1,  ,1  e x )
2
5 1 5
 c 2 (1  e x ) 3 2 F ( , , , 1  e x )
2 2 2
108