Vous êtes sur la page 1sur 197

H I

SHIVAJI UNIVERSITY, KOLHAPUR

CENTRE FOR DISTANCE EDUCATION

Differential Equations
(Mathematics)

For

M. Sc. Part-I
K J
Copyright © Registrar,
Shivaji University,
Kolhapur. (Maharashtra)
First Edition 2008
Second Edition 2010

Prescribed for M. Sc. Part-I

All rights reserved, No part of this work may be reproduced in any form by mimeography or
any other means without permission in writing from the Shivaji University, Kolhapur (MS)

Copies : 1000

Published by:
Dr. D. V. Muley
Registrar,
Shivaji University,
Kolhapur-416 004

Printed by :
Shri. A. S. Mane,
I/c. Superintendent,
Shivaji University Press,
Kolhapur-416 004

ISBN- 978-81-8486-012-2

H Further information about the Centre for Distance Education & Shivaji University may be
obtained from the University Office at Vidyanagar, Kolhapur-416 004, India.
H This material has been produced with the developmental grant from DEC-IGNOU, New
Delhi.

(ii)
Centre for Distance Education
Shivaji University, Kolhapur
n EXPERT COMMITTEE n
Prof. (Dr.) N. J. Pawar Dr. D. V. Muley
Vice-Chancellor, Registrar,
Shivaji University, Kolhapur Shivaji University, Kolhapur.

n ADVISORY COMMITTEE n

Prof. (Dr.) N. J. Pawar Dr. D. V. Muley


Vice-Chancellor, Registrar,
Shivaji University, Kolhapur. Shivaji University, Kolhapur.
Dr. A. B. Rajge Shri. B. S. Patil
Director BCUD, Finance and Accounts Officer,
Shivaji University, Kolhapur. Shivaji University, Kolhapur.
Dr. B. M. Hirdekar Prof. (Dr.) U. B. Bhoite
Controller of Examination Lal Bahadur Shastri Marg,
Shivaji University, Kolhapur. Bharati Vidyapeeth, Pune.
Dr. (Smt.) Vasanti Rasam Prof. (Dr.) A. N. Joshi
Dean, Faculty of Social Sciences, Director, School of Education,
Shivaji University, Kolhapur. Y. C. M. O. U. Nashik.
Prof. (Dr.) B. S. Sawant Shri. J. R. Jadhav
Dean, Faculty of Commerce, Dean, Faculty of Arts & Fine Arts,
Shivaji University, Kolhapur. Shivaji University, Kolhapur.

Dr. T. B. Jagtap Prof. (Dr.) S. A. Bari


Dean, Faculty of Science, Director, Distance Education,
Shivaji University, Kolhapur. Kuvempu University, Karnataka.
Prof. Dr. (Smt.) Cima Yeole
Dr. K. N. Sangale
(Member Secretary)
Dean, Faculty of Education,
Director, Centre for Distance Education,
Shivaji University, Kolhapur.
Shivaji University, Kolhapur.

n B. O. S. MEMBERS OF MATHEMATICS n
Chairman- Prof. S. R. Bhosale
P.D.V.P. Mahavidyalaya, Tasgaon, Dist. Sangli.
l Dr. L. N. Katkar l Dr. A. D. Lokhande
Head, Dept. of Mathematics, Yashavantrao Chavan Warana Mahavidyalaya,
Shivaji University, Kolhapur. Warananagar.
l Dr. H. T. Dinde l Prof. S. P. Patankar
Karmveer Bhaurao Patil College, Urun-Islampur, Vivekanand College, Kolhapur.
Tal. Walwa, Dist. Sangli. l Prof. V. P. Rathod
l Dr. T. B. Jagtap Dept. of Mathematics, Gulbarga University,
Yashwantrao Chavan Institute of Science, Satara. Gulbarga, (Karnataka State.)
l Shri. L. B. Jamale l Prof. S. S. Benchalli
Krishna Mahavidyalaya, Rethare Bk., Karad, Dept. of Mathematics,Karnataka University, Dharwad.
Dist. Satara. l Shri. Santosh Pawar,
1012, 'A' Ward Sadashiv Jadhav, Housing
Society, Radhanagari Road, Kolhapur-416 012.

(iii)
Centre for Distance Education Differential Equations
Shivaji University,
Kolhapur.

Writing Team Unit No.

Dr. (Mrs.) Sarita Thakar All


Department of Mathematics
Shivaji University, Kolhapur.
Maharashtra

n Editor n

Dr. (Mrs.) Sarita Thakar


Department of Mathematics,
Shivaji University, Kolhapur.
Maharashtra.

(iv)
Preface
Large numbers of students appear for M.A./M. Sc. Examinations externally every
year. In view of this, Shivaji University has introduced the Distance Education Mode for
external students from the year 2007-2008, and entrusted the task to us to prepare the
Self Instructional Material (SIM) for aspirants.

It is hoped that students must learn Mathematics not only to become competent
mathematicians but also skilled users of Mathematics in the solution of problems in the
real world. They must learn how to use their Mathematical knowledge in solving the
problems of the real world. Differential equations usually are description of physical
systems. This book on Differential Equations consists of four chapters. Chapter one
contains the complete discussion of linear equations with constant coefficients, including
the uniqueness theorem. In chapter two linear equations with variable coefficients are
trea. Equations with analytic coefficients are introduced and series solutions are
obtained by a simple formal process. A detailed treatment of linear equations with
regular singular points is discussed in chapter four. Classification of regular singular
points and regular singular points at infinity is studied. In chapter five existence and
uniqueness of solutions of first order initial value problem are established. The
innumerable examples and exercises are given at the end of each unit.

The book introduces the students to some of the abstract topics that pervade
modern analysis. The first chapter deals with the Riemann Stieltjes integration. The
problems in Physics and Chemistry which involve mass distribution that are partly
discrete and partly continuous can be solved by using Riemann Stietjes integrations.
The Chapter 2 deals with convergence and uniform convergence of sequences of
functions and series where as the Chapter 3 consists of multidimensional calculus.
The Chapter 4 deals with implicit functions and extremum problems which have wide
applications in optimization theory. Line integrals, surface integrals and Volume integrals
are the subject matter of Chapter 5. This provides sufficient background to study the
Gauss divergence Theorem and Stokes Theorem.

(v)
We owe a deep sense of gratitude to the Vice-Chancellor Dr. N. J. Pawar who has
given impetus to go ahead with ambitious projects like the present one. Dr. Sarita
Thakar, Reader, Department of Mathematics, Shivaji University has to be profusely
thanked for the ovations he has poured to prepare the SIM on Differential Equations.
We also thank Prof. M. S. Chaudhary, Head, Department of Mathematics, Shivaji
University, Director of Centre for Distance Education Mrs. Cima Yeole and Deputy
Director Shri. Raj Patil for their help and keen interest in completion of the SIM.

Prof. S. R. Bhosale
Chairman BOS in Mathematics
Shivaji University, Kolhapur-416004.

(vi)
M. Sc. (Mathematics)
Differential Equations

Contents
Chapter 1 : Linear Equations with Constant Coefficients 1

Chapter 2 : Linear Equations with Variable Coefficients 53

Chapter 3 : Linear Equations with Regular Singular Points 100

Chapter 4 : Existence and Uniqueness of Solution to 159


First Order Equations

(vii)
p p
M. Sc. (Mathematics)
Paper III

Differential Equations

❁ D r. Sarita Thakar ❁

Department of Mathematics
Shivaji University, Kolhapur (M.S.)

p p
Differential Equations
Chapter
1
Linear Equations with
Constant Coefficients

Contents :
Unit 1 : Initial value problems for second order equations.
Unit 2 : Linear dependence and independencce
Unit 3 : The homogenous equation of order n
Unit 4 : The non-homogeneous equation of order n

Introduction :
We live in a world of interrelated changing entities. The position of the earth changes with
time, the velocity of falling body changes with distance, the bending of a beam changes with the
weight of the load placed on it, the area of circle changes with the size of the radius, the path of
projectile changes with the velocity and angle at which it is fired.
In the language of mathematics changing entities are called variables and the rate of change
of one variable with respect to another is called derivative. Equations which express a relation
among these variables and their derivatives are called differential equations.
A Linear differential equation of order n with constant coefficients is an equation of the
form
a0 y ( n ) + a1 y ( n –1) + a2 y ( n –2) + ⋅⋅⋅ + an y = b( x),
where, a0 ≠ 0, a1, a2 , ⋅⋅⋅, an are complex constants
and b is complex valued function on an interval I : a < x < b .
The operator L defined by

L(φ ) ( x ) = φ (n ) ( x ) + a1 φ (n –1) ( x ) + a2 φ (n –2) ( x ) + .... + an φ ( x ) is called as


differential operator of order n with constant coefficients.
The equation L(y) = b(x) is called non-homogenous equation. If b(x) = 0 for all x in I the
corresponding equation L(y) = 0 is called a homogenous equation.

Differential Equations (1)


Unit 1 : Initial Value Problems for Second Order Equations
Here, we are concerned with the equation
L( y ) = y ′′ + a1 y ′ + a2 y = 0
where a1 and a2 are constants.

Theorem 1.1.1
Let, a1, a2 be constants and consider the equation L(y) = y¢¢ + a1y¢ + a2 y = 0
1. If r1, r2 are distinct roots of the characteristic polynomial
p(r) = r2 + a1r + a2
then the functions φ1 ( x ) = e r1x and φ2 ( x ) = e r2 x are solutions of L(y) = 0.

2. If r1 is a repeated root of the characteristic polynomial p(r), then the functions φ1 ( x) = e r1x
and φ2 ( x ) = xe r1x are solutions of L(y) = 0.
Proof : Let f (x) = erx be a solutions of L(y) = 0.
L(e rx ) = (e rx )′′ + a1 (e rx )′ + a2 e rx
= (r 2 + a1r + a2 )e rx
L (erx) = 0 if and only if p(r) = r2 + a1r + a2 = 0.
1. If r 1 and r 2 are distinct roots of p(r) then L(e r1x ) = L(e r2 x ) = 0 and φ1 ( x ) = e r1x and
φ2 ( x) = e r2 x are solutions of L(y) = 0.
2. If r1 is a repeated root of p(r) then
P(r ) = (r – r1 ) 2 and p′(r ) = 2(r – r1 )

L(e r x ) = P(r )er x for all r & x.


∂ ∂
L(er x ) =  P (r )erx 
∂r ∂r  
⇒ L( xe r x ) = [P′(r ) + xP(r ) ]e rx .
At r = r1, P(r1) = P¢(r1) = 0.

i.e. L( xer1 x ) = 0 thus, showing that xer1 x is a solution of L(y) = 0.


Thus if r1 is a repeated root of the characteristic polynomial P(r), then φ1 ( x) = e r1x and

φ2 ( x ) = xer1x are solutions of L(y) = 0.

Theorem 1.1.2 :
If f1 and f2 are two solutions of L(y) = 0 then C1 f1 + Cf2 is also a solution of L(y) = 0.
Where, C1 and C2 are any two constants.
Proof : Let f1 and f2 be two solutions of L(y) = 0
L (φ1 ) = φ1′′ + a1 φ1′ + a2 φ1 = 0
Differential Equations (2)
L(φ2 ) = φ2′′ + a1 φ2′ + a2 φ1 = 0
Suppose C1 and C2 are any two constants then the function f defined by f = C1 f1 + C2 f2
is also a solution of L(y) = 0.
L(φ ) = (aφ1 + c2φ 2 )′′ + a1 (aφ1 + c2φ 2 ) + a2 (aφ1 + c2φ 2 )

= c1 (φ1′′ + a1φ1′ + a2φ1 ) + c2 (φ 2′′ + a1φ 2′ + a2φ 2 )

= c1L(φ1 ) + c2 L (φ2 )
=0
The function f which is zero for all x is also a solution called the trivial solution
of L(y) = 0.
The results of above two theorems allow us to solve all homogeneous linear second order
differential equations with constant coefficients.

Definition 1.1 :
An initial value problem L(y) = 0 is a problem of finding a solution f satisfying
φ ( x0 ) = α 0 and φ ′( x0 ) = β 0 where, x0 is some real number and a0, b0 are given constants.

Theorem 1.1.3 : (Existence Theorem)


For any real x0 and constants a, b , there exists a solution f of the initial value problem

L( y ) = y ′′ + a1 y ′ + a2 y = 0, y ( x0 ) = α , y ′( x0 ) β , – ∞ < x < ∞.
Proof : By theorem 1.1.1 there exist two solutions f1 and f2 that satisfy L(f1) = L(f2) = 0. From
theorem 1.1.2 we know that c1 f1 + c2 f2 is a solution of L(y) = 0. We show that there are
unique constants c1, c2 such that φ = c1φ1 + c2φ2 satisfies φ ( x0 ) = α and φ ′( x0 ) = β .

φ ( x0 ) = c1φ1 ( x0 ) + c2φ 2 ( x0 ) = α
φ ′( x0 ) = c1φ1′ ( x0 ) + c2φ 2′ ( x0 ) = β
Above system of equations will have a unique solution c1, c2 if the determinant
φ1 ( x0 ) φ 2 ( x0 )
∆= = φ1 ( x0 ) φ2′ ( x0 ) – φ2 ( x0 ) φ1′ ( x0 ) ≠ 0.
φ1′ ( x0 ) φ 2′ ( x0 )
By theorem 1.1.1 (1), φ1 ( x ) e r1 x and φ2 ( x) e r2 , x are two solution of L( y ) = 0 for r1 ≠ r2
and
∆ = e r1 x0 r2 e r2 x0 – e r2 x0 r1 e r1x0
= (r2 – r1 )e(r1 +r2 ) x0 ≠ 0.

By theorem 1.1.1 (2), φ1 ( x ) = e r1 x and φ2 ( x ) xe r1, x are solutions of L(y) = 0 and

∆ = e r1 x0 e r1x0 + x0r1e r1x0  – x0e r1x0 r1e r1x0


 
= e 2 r1x0 ≠ 0
Differential Equations (3)
Thus, the determinant condition is satisfied in both the cases. Therefore, c1, c2 are uniquely
determined. The function f = c1 f 1 + c2 f 2 is a desired solution of the initial value problems.

Defination 1.2 :
A solution of a differential equation will be called a particular solution if it satisfies the
equation and does not contain arbitrary constants.

Theorem 1.1.4 :
Let, f be any solution of
L( y ) = y′′ + a1 y ′ + a2 y = 0
on an interval I containing a point x0, Then for all x in I.
|| φ ( x0 ) || e – k | x – x0 | ≤ || φ ( x) || ≤ || φ ( x0 ) || e k | x – x0 |
Where,
1
φ ( x ) = | φ ( x ) |2 + | φ ′( x ) |2  and k = 1+ | a1 | + | a2 | .
2
 
Proof : Let,
u ( x) = || φ ( x) ||2
= | φ ( x ) |2 + | φ ′( x ) |2
= φ ( x) φ ( x) + φ ′( x) φ ′( x)
Then, u ′( x) = φ ′( x) φ ( x) + φ ( x) φ ′( x ) + φ ′′( x) φ ′( x) + φ ′( x) φ ′′( x)
and | u ′( x) | ≤ 2 | φ ( x ) | | φ ′( x) | +2 | φ ′( x ) | | φ ′′( x ) |
as | φ ( x ) | = | φ ( x ) |
Since f is a solution of L(y) = 0, L(φ ) = φ ′′ + a1φ ′ + a2φ = 0

i.e. φ ′′( x ) = – a1φ ′( x) – a2φ ( x ) and the above inequality becomes


| u ′( x) | ≤ 2| φ ( x) | | φ ′( x) | + 2 | φ ′( x) | [ | a1 || φ ′( x) | + | a2 || φ ( x ) | ]
≤ 2 [1 + | a2 |] | φ ( x ) || φ ′( x) | +2 | a1 || φ ′( x ) |2

But, 2| φ ( x) || φ ′( x) | ≤ | φ ( x) |2 + | φ ′( x) |2
Therefore,
| u ′( x) | ≤ 2 (1+ | a1 | + | a2 | )| φ ′( x ) |2 + 2 (1 + | a2 | )| φ ( x ) |2
≤ 2 (1+ | a1 | + | a2 |) | φ ′( x) |2 + | φ ( x) |2 
 
≤ 2k u ( x)
Thus, we get
–2u ( x) ≤ u ′( x) ≤ 2ku ( x )
u ′( x) ≤ 2ku ( x) is equivalent to u′( x ) – 2k u ( x) ≤ 0 since exponential functions are positive
on multiplying above inequality by e–2kx we get

Differential Equations (4)


( ′
e –2kx (u′( x ) – 2ku ( x) ) = e –2kx u ( x) ≤ 0. )
Integrating above inequality between the limits x0 to x for x > x0 yields.
e –2kxu ( x) – e –2kx0 u ( x0 ) ≤ 0
u ( x ) ≤ e 2 k ( x – x0 ) u ( x0 )

Thus, || φ ( x) ||2 ≤ e 2 k ( x – x0 ) || φ ( x0 ) ||2


Similarly, for x > x0 the inequality –2 k u (x) £ u¢ (x) implies
|| φ ( x0 ) ||2 e –2 k ( x – x0 ) ≤ || φ ( x ) ||2
Therefore for x > x0 we get
|| φ ( x0 ) ||2 e –2k ( x – x0 ) ≤ || φ ( x) ||2 ≤ e 2k ( x – x0 ) || φ ( x0 ) ||2 .......... (1.1.1)
For x < x0, the sign of above inequality will get changed
|| φ ( x0 ) ||2 e –2k ( x – x0 ) ≥ || φ ( x) ||2 ≥ e 2k ( x – x0 ) || φ ( x0 ) ||2
This inequality can be written as
e 2k ( x – x0 ) || φ ( x0 ) ||2 ≤ || φ ( x0 ) ||2 ≤ || φ ( x0 ) ||2 e –2k ( x – x0 )
since x < x0, x0 – x > 0.
e –2k ( x0 – x ) || φ ( x0 ) ||2 ≤ || φ ( x) ||2 ≤ || φ ( x0 ) ||2 e 2 k ( x0 – x ) .......... (1.1.2)
Equation (1.1.1) and (1.1.2) together can be put in the form
e –2k |x0 – x| || φ ( x0 ) ||2 ≤ || φ ( x) ||2 ≤ || φ ( x0 ) ||2 e 2k |x0 – x|
Since all the terms in above inequality are positive the square root of each term results
into the required inequality.

Theorem 1.1.5 (Uniqueness Theorem)


Let a, b be any two constants and let x0 be any real number. On any interval I containing
x0 there exists at most one solution f of the initial value problem
L( y ) = y′′ + a1 y′ + a2 y = 0, y ( x0 ) = α , y′( x0 ) = β
Proof : Suppose f and y are two solutions.
Let θ = φ – ψ . Since L (φ ) = L (ψ ) = 0,
L(θ ) = L(θ – ψ ) = L(θ ) – L(ψ ) = 0
Since φ ( x0 ) = ψ ( x0 ) = α and φ ′( x0 ) = ψ ′( x0 ) = β ,
θ ( x0 ) = φ ( x0 ) – ψ ( x0 ) = 0 and θ ′( x0 ) = φ ( x0 ) – ψ ′( x0 ) = 0

Thus, L(θ ) = 0, θ ( x0 ) = 0 and θ ′( x0 ) = 0.

|| θ ( x0 ) ||2 = | θ ( x0 ) |2 + | θ ′( x0 ) 2 | = 0
 

Differential Equations (5)


By theorem (1.1.4) we see that
|| θ ( x) ||= | θ ( x) |2 + | θ ′( x) 2 | = 0 for all x in I
 
This implies q (x) = 0 for all x in I.
But θ ( x ) = θ ( x ) – ψ ( x ) = 0 i.e. φ ( x) ≡ φ ( x ) .

Theorem 1.1.6 :
Let f 1, f 2 be two solutions of L(y) = 0 given by theorem 1.1.1. If c1, c2 are any two
constants the function f = c1 f 1 + c2 f 2 is a solution of L(y) = 0 on – ¥ < x <¥.
Conversely, if f is any solution of L(y) = 0 on – ¥ < x <¥, then there are unique constants
C1 and C2 such that f = C1 f 1+ C2 f 2.
Proof : First part of the theorem follows from theorem 1.1.2.
Conversely suppose f is any solution of L(y) = 0. Let φ ( x0 ) = α and φ ′( x0 ) = β for
some constants a and b. In the proof of existence theorem 1.1.3 we showed that there is a
solution y of L(y) = 0 satisfying.
ψ ( x0 ) = α , ψ ′( x0 ) = β of the form
ψ ( x ) = c1 φ1 ( x ) + c2φ2 ( x )1 where c1 and c2 are uniquely determined by a and b . By
uniqueness theorem 1.1.5 φ = ψ , for all x.

Examples :
1. Find all solutions of the following equations.
(a) y²– 4 y = 0 (b) y²+ 2 i y¢ + y = 0 (c) y²– 4 y¢ + 5y = 0
Answer :
(a) The characteristic polynomial is p(r) = r2 – 4. r1 = 2 and r2 = – 2 are two distinct roots of
p (r) = 0.
Therefore φ1 ( s ) = e 2 x and φ2 ( x) = e –2 x are two solutions. For any constants c1 and
c2, c1e2x + c2e–2x is a solution. Thus the general solution is φ1 ( x ) = c1e 2 x + c2e –2 x .
(b) The characteristic polynomial p(r) = r2 + 2ir + 1
1
p( r ) = 0 ⇒ r= –2i ± (2i) 2 – 4 

2 
1
= –2i ± –8 
2
= –i ± 2 i

(
= –1 ± 2 i )

Differential Equations (6)


( ) ( )
Thus r1 = –1 + 2 i and r2 = –1 – 2 i are two district roots of p(r) = 0.

Therefore φ1 ( x) = e
( –1+ 2 )ix and φ2 ( x) = e
( –1– 2 )ix are two solutions. Thus, for any

constants c1 and c2, φ ( x ) = c1e ( ) ( –1– 2 )ix


–1+ 2 ix
+ c2 e is a general solution.

(c) The characteristic polynomial p(r) = r2 – 4r + 5. p(r) = 0 gives r 1 = 2 + i and


r2 = 2 – i as two distinct roots. f1(x) = e (2 + i) x and f2(x) = e (2 – i) x are two solutions of the
differential equation. For any constants c1 and c2, f (x) = c1e (2 – i) x+ c2e (2 + i) x is a general
1
solution. In particular for c1 = c2 = we get,
2
 e + e  2x
– i x i x
φ ( x) = e 2 x   = e cos x. and for
 2 
–1 1
c1 = and c2 = we get
2i 2i
 ei x – e – i x  2 x
φ ( x) = e  2x
 = e sin x
 2i 
Thus, f (x) = A e cos x + B e sin x is a solution of the differential equation for any
2x 2x

constants A & B.

2. Find the solutions f of the following initial value problems.

(a) φ ′′ + φ ′ – 6 φ = 0, φ (0) = 1, φ ′(0) = 0

π
(b) φ ′′ + φ = 0, φ (0) = 1, φ   = 0
2

(c) φ ′′ + kφ = 0, k is any constant, φ (0) = 0, φ (π ) = 0


(d) φ ′′ – 2φ ′ – 3φ = 0, φ (0) = 0, φ ′(0) = 1

Answer :
(a) The characteristic polynomial p(r) = r2 + r – 6. r1 = 2 and r2 = – 3 are distinct roots
φ ( x ) = c1e 2 x + c2e –3x is a general solution.
φ (0) = 1 ⇒ c1 + c2 = 1 .......... (1)
φ ′ (0) = 0 ⇒ φ ′ ( x) = 2c1e 2 x – 3 c2e –3 x at x = 0, gives φ ′ (0) = 2c1 – 3c2 = 0 .......... (2)
solving equation (1) and (2) for c1 and c2 we get c1 = 3/5 and c2 = +2/5.

3e 2 x 2 e3 x
Thus, the required solution is φ ( x ) = + .
5 5

Differential Equations (7)


(b) The characteristic polynomial is p (r) = r2 + 1. r1 = i and r2 = – i are distinct roots
φ ( x ) = c1 cos x + c2 sin x is a general solution.
φ (0) = 1 ⇒ c1 cos 0 + c2 sin = 1 gives c1 = 1
φ (π 2 ) = 2 ⇒ c1 cos π 2 + c2 sin π 2 = 2 . gives c2 = 2.
Thus, f (x) = cos x + 2 sin x is the required solution.
(c) The characteristic polynomial is p (r) = r2 + k since k is any constants, k can be positive,
negative or zero.

Case 1. k > 0

Then r1 = k i and r2 = – k i ; are distinct roots.

∴ φ ( x) = c1 e k ix
+ c2 e – k ix is a general solution
In general φ ( x) = A cos k x + B sin k x is a solution.
φ (0) = 0 ⇒ A cos 0 + B sin 0 = 0 i.e. A = 0
φ (π ) = 0 ⇒ A cos π + B sin π = 0 i.e. A = 0

Thus, f (x) = B sin k x is a solution where B is any constant.

Case 2. k = 0
p(r ) = r 2 = 0 ⇒ r = 0 a repeated root.

∴ φ ( x ) = c1e 0 + c2 xe 0 = c1 + c2 x is a solution
φ (0) = 0 ⇒ c1 = 0
φ (π ) = 0 ⇒ c1 + c2π = 0 ⇒ c2 = 0
Therefore there is no nontrivial solution corresponding to k = 0.

Case 3. k < 0
for k = 0, p (r) = r2 + k has distinct roots
r1 = – k & r2 = – – k ( Since k < 0, – k > 0)

φ (x) = c1e –k x
+ c 2e – –k x

φ (0) = c1 + c2 = 0
–k π –k π
φ (π ) = c1e + c2e – =0
Simultaneous evaluation of above two equations give c1 = c2 = 0.
Thus, there is no non-trival solution corresponding to k < 0.
The only non-trivial solution for the given equation is φ ( x ) = B sin k x.
(d) The characteristic polynomial p(r) = r2 – 2r – 3
r1 = 3, r2 = 1 are two distinct roots.
Differential Equations (8)
\ φ ( x) = c1e3 x + c2e – x is a general solution
φ (0) = 0 ⇒ φ (0) = c1 + c2 = 0
φ ′ ( x) = 3c1 e3 x – c2 e – x
φ ′ (0) = 1 ⇒ φ ′ (0) = 1 = 3c1 – c2
Thus, c1 + c2 = 0 and 3c1 – c2 = 1 gives
1 1
c1 =
and c2 = –
4 4
1 3x 1 – x
Therefore φ ( x) = e – e is the required solution.
4 4

EXERCISES

1. Fill in the blanks.


(i) If r1, r 2 are distinct roots of characteristic polynomial p (r ) = r 2 + a1 r + a2 then
φ1 ( x ) = .............. and φ2 ( x) = .............. are solutions of the differential equation
y ′′ + a1 y ′ + a2 y = 0

(ii) If p (r ) = (r – r1 )2 is a characteristic polynomial then φ1 ( x ) = .............. and


φ2 ( x) = .............. are two solutions of the differential equation y ′′ – 2r1 y ′ + r12 y = 0.
(iii) Uniqueness theorem states that ....................
(iv) Solution of y ′′ – 2 y ′ + 4 y = 0 are φ1 ( x ) = ............. and φ2 ( x) = ............. .
(v) The general solution of y ′′ – 3 y ′ + 2 y = 0 is.....
2. Find the gental solution of each of the following equation.
(i) y ′′ + 4 y ′ = 0 (ii) y ′′ – y = 0 (iii) y′′ + y′ – 6 y = 0

(iv) y ′′ + 4ky ′ –12k 2 y = 0 (v) y ′′ – 2ay ′ + a 2 y = 0 (vi) y ′′ – 4 y ′ + 20 y = 0


3. Find the solution of the following initial value problems :
(i) y ′′ = 0, y (1) = 2, y ′(1) = –1
(ii) y ′′ + 4 y ′ + 4 y = 0, y (0) = 1, y ′(0) = 1
(iii) y ′′ – 2 y ′ + 5 y = 0, y (0) = 2, y ′(0) = 4
(iv) y ′′ – 4 y ′ + 20 y = 0, y (π 2 ) = 0, y ′ (π 2 ) = 1
Answers :
1. (i) φ1 ( x ) = e r1x , φ2 ( x ) = e r2 x (ii) φ1 ( x ) = e r1x , φ2 ( x ) = xe r1x

(iii) theorem 1.1.5 (iv) φ1 ( x ) = e 2 x , φ2 ( x ) = x e 2 x


(v) c1e 2 x + c2e x
Differential Equations (9)
2. (i) c1 + c2 e –4 x (ii) c1e x + c2e – x

(iii) c1e 2 x + c2 e –3 x (iv) c1e –6 kx + c2e 2 kx

(v) (c1 + c2 x )e ax (vi) e2 x (c1 cos 4 x + c2 sin 4 x)

3. (i) 3 – x (ii) (1 + 3x) e–2x


1 2 x –π
(iii) e x (2cos 2 x + sin 2 x) (iv) e sin 4 x
4

Unit 2 : Linear Dependence and Independence


Every solution of the equation L (y) = 0 is a linear combination of two solutions obtained
in theorem 1.1.1. Therefore these two solutions span the solution space of the differential equation
L(y) = 0.
Defination 1.3 : A set of n real or complex functions f1, f2, f3,......, fn defined on an interval (a,
b) is said to be linearly independent when c1 f1 ( x ) + c2 f 2 ( x) + c3 f3 ( x) + ⋅⋅⋅ + cn f n ( x) = 0
for every x in (a, b) implies c1 = c2 = c3 = ⋅⋅⋅ = cn = 0 .

Defination 1.4 : Given the functions f1, f 2, f3, ⋅⋅⋅, f n if constants c1, c2 , c3 , ⋅⋅⋅, cn not all zero
exist such c1 f1 ( x ) + c2 f 2 ( x) + c3 f 3 ( x) + ⋅⋅⋅ + cn f n ( x) = 0 for every x in (a, b), then these
functions are linearly dependent.
A set which is not linearly independent is said to be linearly dependent.
There are two notions of linear independence, according as we allow the coefficients
ck, k = 1, 2, 3, ...., n to assume only real values or also complex values. In the first case, one says
that the functions are linearly independent over the field of reals; in the second case, that they
are linearly independent over the complex field.
Lemma 1.2.1 : A set of real valued functions on an interval (a, b) is linearly independent over
the complex field if and only if it is linearly independent over the real field.
Proof : If the set of real valued functions on an interval (a, b) is linearly independent over
the complex field then it is linearly independent over the field of reals.
Conversely suppose the set is linearly independent over the real field. Therefore for
n
α j ∈ R, ∑ α j f j ( x ) = α1 f1 ( x ) + α 2 f 2 ( x ) + α3 f 3 ( x ) + ⋅⋅⋅ + α n f n ( x ) = 0 for all x in (a, b)
j =1
n
implies aj = 0 for all j = 1, 2, 3...., n. Let ∑ c j f j ( x ) = 0 for all x in (a, b) and for some
j =1
c j ∈ C , j = 1, 2,3 ⋅⋅⋅, n. Since the function f j are real valued and ∑ c j f j ( x) = 0,
 c j – c j* 
( )
* n n
 ∑ c j f j ( x)  = 0 . implies ∑ c j f j ( x ) = 0 . Thus, ∑ 
*  f j ( x ) = 0 . But c j – c j* / i
  j =1 j =1  i 
 
are all real and the set is linearly independent over the real field therefore c j = c j*. But then cj’s
Differential Equations (10)
n
are all real therefore ∑ c j f j ( x) = 0 implies cj = 0 for j = 1, 2,....n.
j =1
A set of functions which is linearly dependent on a given domain may become linearly
independent when the functions are extended to a larger domain. However, a linearly independent
set of functions clearly remain linearly independent on the restricted domain.
Illustration 1 : The functions f1 and f2 define by f1(x) = Cos x and f2(x) = Sin x are linearly
independet on the real line IR and therefore are linearly independent on (0, 2 p).
Illustration 2 : The functions f1 and f2 define by f1(x) = x, f2(x) = | x | are linearly indepent on
the interval (–1, 1) but is not linearly independent on the interval (0, 1) as on the interval
(0, 1), f1(x) = f2(x).

Theorem 1.2.1 :
Let a1, a2 be constants and consider the equation L( y ) = y ′′ + a1 y ′ + a2 y = 0. The two
solutions of L (y) = 0 given in the theorem 1.1.1 are linearly independent on any interval I.
Proof : Let r1, r2 be the roots of characteristic polynomial p(r) = r2 + a1 r + a2.

Case 1.

If r1 ¹ r2, then φ1 ( x) = e r1 x and φ2 ( x) = e r2 x


are two solutions of the equation L(y) = 0 on
an interval I.
Suppose c1e r1, x + c2e r2 x = 0 for all x in I.
Then c1 + c2 e ( r2 – r1) x = 0 for all x in I.

Differentiation of above equation with respect to x gives c2 (r2 – r1 )e(r2 – r1 ) x = 0 for all x in
I.
Since, r2 ¹ r1 and exponential function in non-zero, c2 is zero. But if c2 is zero then
c1 + c2 e( r2 – r1 ) x = 0 implies c1 is zero. Thus, c1e r1x + c2 e r2 x = 0 implies c1 = c2 = 0.
Therefore φ1 ( x ) = e r1 x and φ2 ( x ) = e r2 x are linearly independent.

Case 2.

If r1 = r2, then φ1 ( x) = e r1 x and φ2 ( x) = xe r1 x are two solutions of the equation L(y) = 0 on


an interval I.

Suppose c1e e1 x + c2 xe r2 x = 0 then c1 + c2 x = 0 for all x in I. Therefore c1 = c2 = 0 . Thus,


f 1 and f 2 are linearly independent
Thus, in both cases the two solutions f 1 and f 2 of L(y) = 0 are linearly independent.

Defination 1.5 : Assume that each of the functions f1 ( x ), f 2 ( x), f3 ( x ), ⋅⋅⋅, f n ( x ) are
differentiable atleast (n – 1) times in the interval (a, b). Then the determinant
Differential Equations (11)
f1 ( x ) f 2 ( x) f3 ( x) L fn ( x)
f1′ ( x) f 2′ ( x ) f3′ ( x) L f n′ ( x )
f1′′ ( x) f 2′′ ( x) f3′′ ( x) L f n′′ ( x)
M M M M
f1(n –1) ( x ) f 2(n –1) ( x ) f 3(n –1) ( x ) L f n (n –1) ( x )

denoted by W( f1, f 2, f 3, ...., f n ) ( x) is called the wronskian of the n functions f1, f 2, f 3, ...., f n .

Theorem 1.2.2 :
Two solutions f1, f2 of L (y) = 0 are linearly independent on an interval I if and only if
W(φ1, φ2 ) ( x) ≠ 0 for all x in I.

Proof : Suppose W(φ1, φ2 ) (x) ≠ 0 for all x in I


Let c1, c2 be constants such that
c1 f1 (x) + c2 f2 (x) = 0 for all x in I. Then
c1 f1¢ (x) + c2 f2¢ (x) = 0 for all x in I.
Above two equations can be written as
 φ1 ( x ) φ 2 ( x )   c1   0
  c  =  0
φ1′ ( x ) φ2′ ( x )   2  

Since, W(φ1 , φ2 ) ( x) ≠ 0 for all x in I, the coefficient matrix is invertible. On premultiplying


the inverse of the coefficient matrix results in c1 = c2 = 0. This proves that f1 and f2 are linearly
independent on I.
Conversely, assume that f1, f2 are linearly independent on I. Suppose that there is a point
x0 in I such that W(φ1, φ1) ( x0 ) = 0. Then the system of equations

 φ1 ( x0 ) φ 2 ( x0 )   c1   0 
   = 
φ1′ ( x0 ) φ2′ ( x0 )   c2   0 
has a solution c1, c2 where at least one of these numbers is not zero. Let c1, c2, be such a solution
and consider the function ψ ( x ) = c1 φ1 ( x) + c2 φ2 ( x ). Now L(ψ ) = 0 and ψ ( x0 ) = 0, ψ ′( x0 ) = 0.
1
Therefore || ψ ( x0 ) ||= | ψ ( x0 ) |2 + | ψ ′( x0 ) |2  2 = 0. By theorem 1.1.4 || ψ ( x) || = 0. But
 
|| ψ ( x) ||= | ψ ( x) |2 + | ψ ′( x) |2  = 0. Therefore ψ ( x ) = 0 for all x in I and thus
 
c1 φ1( x) + c2 φ 2 ( x) = 0 for all x in I. But then f1 and f2 are linearly dependent. Thus, the
supposition W(φ1, φ2 ) ( x0 ) = 0 must be false and therefore W(φ1, φ 2 ) ( x) ≠ 0 for all x in I.

In the next theorem we will prove that we need to compute W(φ1, φ 2 ) at only one point to
test the linear independence of the solutions f1 and f2 .
Differential Equations (12)
Theorem 1.2.3 :
Let f1, f2 be two solution of L(y) = 0 on an interval I and let x0 be any point in I. Then two
solutions f1 and f2 are linearly independent on I if and only if W(φ1, φ2 ) ( x0 ) ≠ 0.

Proof : If f1 and f2 are linearly independent on I then by theorem 1.2.2, W(φ1, φ 2 ) ( x) ≠ 0 for
all x in I. In particular W(φ1, φ2 ) ( x0 ) ≠ 0 conversely, suppose W(φ1, φ2 ) ( x0 ) ≠ 0 and
suppose c1, c2 are constants such that c1 φ1 ( x ) + c2 φ 2 ( x ) = 0 for all x in I. Then
c1 φ1 ( x0 ) + c2 φ 2 ( x0 ) = 0 and c1 φ1′ ( x0 ) + c2 φ 2′ ( x0 ) = 0.
 φ1 ( x0 ) φ 2 ( x0 )   c1   0
i.e.   c  =  0
φ1′ ( x0 ) φ2′ ( x0 )   2  

But since the determinant of the coefficient is W(φ1, φ2 ) ( x0 ) ≠ 0 we obtain c1 = c2 = 0.


Thus f1, f2 are linearly independent on I.
In the next theorem we show that the knowledge of two linearly independent solutions of
L(y) = 0 is sufficient to generate all solutions of L(y) = 0.

Theorem 1.2.4 :
Let f1, f2 be any two linearly independent solutions of L(y) = 0 on an interval I. Every
solution f of L(y) = 0 can be written uniquely as
φ = c1 φ1 + c2 φ2 where c1, c2 are constants.

Proof : Let x0 be a point in I. Let φ ( x0 ) = α , φ ′ ( x0 ) = β . Since f1, f2 are linearly


independent on I we know that W (φ1 , φ 2 )( x0 ) ≠ 0 . Consider the two equations.

 φ1 ( x0 ) φ 2 ( x0 )   c1  α 
  c  =  β 
φ1′ ( x0 ) φ 2′ ( x0 )   2  

Since W(φ1, φ2 ) ( x0 ) ≠ 0, above system of equations has a unique solution c1, c2 . For
this choice of c1, c2 the functionψ ( x ) = c1 φ1 ( x ) + c2 φ2 ( x ) satisfies ψ ( x0 ) = c1 φ1 ( x0 ) + c2 φ2 ( x0 )
= α = φ ( x0 ) i.e. ψ ( x0 ) = φ ( x0 ) similarly ψ ′( x0 ) = φ ′( x0 ) and L(ψ ) = 0 . From the uniqueness
theorem 1.1.5 it follows that ψ = φ on I i.e. φ = c1 φ1 + c1 φ2 .

Examples :
Q1. Show that the functions ex, e2x, e3x are linearly independent.
Ans. :
Method 1 :
Let c1e x + c2e 2 x + c3e3 x = 0
then c1 + c2 e x + c3e 2 x = 0 .......... (1)
Differential Equations (13)
Differentiate above equation (1) with respect to x then c2ex + 2 c3 e2x = 0 implies
c2 + 2c3e x = 0 .......... (2)

By differentiating equation (2) with respect to x we get 2c3e x = 0 therefore c3 = 0.

But then by equation (2) c2 = 0 and by equation (1) we get c1 = 0. Thus c1 = c2 = c3 = 0.


Therefore the functions ex, e2x, e3x are linearly independent.
Method 2 :

Let φ1 ( x) = e x , φ2 ( x) = e 2 x , φ3 ( x) = e3x

ex e2 x e3 x 1 1 1
x
W(φ1, φ2 , φ3 ) = e x
2e 2x
3e 3x
=e e e 2x 3x
1 2 3
ex 4e 2 x 9e3 x 1 4 9

= e6x [1(18–12) – 1 (9–3) + 1 (4–2)]


= 2 e6x ¹ 0.
by theorem 1.2.2 f1, f2, f3 are linearly independent.
Q2. : The functions f1, f2 are defined on – ¥ < x < ¥ . Determine whether they are linearly
dependent or independent there.
(i) φ1 ( x) = x, φ2 ( x) = e rx , r is a complex constant
(ii) φ1 ( x ) = x 2 , φ2 ( x) = 5 x 2
(iii) φ1 ( x) = x, φ2 ( x) =| x |
(iv) φ1 ( x) = cos x, φ2 ( x) = sin x
Ans. (i) :
Method 1 :
Let c1 φ1 ( x ) + c2 φ 2 ( x ) = 0

i.e. c1 x + c2 e rx = 0 .......... (1)


if r = 0, c1x + c2 = 0 for all x ∈ R implies
c1 = 0 and c2 = 0. \ f1, f2 are linearly independent if r ¹ 0, differentiate
equation (1) with respect to x then c1 + rc2e rx = 0

Again differentiate above equation with respect to x then r 2c2e rx = 0. But r ≠ 0 and e rx ≠ 0
therefore c2 = 0 and from equation (1) we get c1 = 0. Thus f1, f2 are linearly independent.
Method 2 :

x e rx x 1
W(φ1, φ2 ) = = e rx
1 re rx 1 r

Differential Equations (14)


= erx (r x –1) ≠ 0 for x ∈ IR
\ f1, f2 are linearly independent
Method 3 :
0 1
W(φ1, φ2 ) (0) = = 1 ≠ 0 therefore by theorem 1.2.3 f , f are linearly
1 r 1 2
independent.

Ans. (ii) :
Let c1 φ1 + c2 φ 2 = 0

i.e. c1 x 2 + c2 5 x 2 = 0
if (c1 + 5c2 ) x 2 = 0

If we choose c1 = – 5c2 ¹ 0 then the linear combination c1 φ1 + c2 φ 2 = 0 therefore by


definition 1.4, f1, f2 are linearly dependent.

Ans. (iii) :
For x > 0 c1 φ1 + c2 φ 2 = (c1 + c2 ) x as | x |= x
and for x < 0 c1 φ1 + c2 φ 2 = (c1 – c2 ) x as | x |= – x
Thus, c1 φ1 + c2 φ 2 = 0 for x ∈ R
⇒ (c1 + c2 ) x = 0 and (c1 – c2 ) x = 0
for every x ∈ R above two equations hold true if and only if c1 = c2 = 0. Thus f1, f2 defined by
φ1 ( x ) = x and φ 2 ( x ) =| x | are linearly independent.

Ans. (iv) :
φ1 ( x) = cos x; φ2 ( x) = sin x
cos x sin x
W(φ1, φ2 ) ( x) = =1
– sin x cos x
Q W (φ1 , φ2 ) ( x) = 1 ≠ 0, φ1 , φ 2 are linearly independent.

Q3. : Let fn be any function satisfying the boundary value problem

y ′′ + n 2 y = 0, y (0) = y (2π ), y ′(0) = y ′(2π ), n = 0,1, 2,3,.....



show that ∫ φ n ( x) φ m ( x) dx = 0 if n ≠ m.
0

Ans. :
The characteristic polynomial p(r ) = r 2 + n2 has roots r1 = in, r2 = –in and therefore the

Differential Equations (15)


general solution φn ( x ) = cn cos nx + d n sin nx
From the given boundary conditions.
φn (0) = cn and φn (2π ) = cn ⇒ φn (0) = φn (2π )
and φn′ (0) = nd n and φn′ (2π ) = nd n ⇒ φn′ (0) = φn′ (2π )
Thus, φn ( x ) = cn cos nx + d n sin nx satisfies the given boundary conditions.

The solution fn satisfies φn′′ ( x ) + n φn ( x) = 0 where as φm′′ ( x) + m 2φm ( x) = 0 holds.


2

Thus, ( n 2 – m 2 ) φn ( x) φm ( x ) = φn′′ ( x ) φm ( x) – φn ( x) φm′′ ( x )



= φn′ ( x ) φm ( x) – φn ( x ) φm′ ( x )
 
Integrating above equation from 0 to 2p
We get,
2π 2π ′
(n 2
–m 2
) ∫ φ ( x) φ
0
n m ( x) dx = ∫ φn′ ( x) φm ( x) – φn ( x) φm′ ( x)  dx
0
 

= φn′ ( x ) φm ( x ) – φn ( x ) φm′ ( x )
 0

But φn (0) = cn, φn (2π ) = cn ; φn′ (0) = ndn, φn′ (2π ) = ndn

Similarly, φm (0) = cm, φm (2π ) = cm ; φm′ (0) = mdm = φm′ (2π )



Thus, (n 2
)
– m2 ∫ φn ( x) φm ( x) dx = [ndn cm – cn mdm ] – [ndn cm – cn mdm ]
0
=0

Since, n ≠ m, ∫ φn ( x) φm ( x) dx = 0.
0

Q4. (a) : Show that fn (x) = Sin nx satisfies the boundary value problem y² + n2y = 0,
y ( 0 ) = 0 , y ( p ) = 0, n = 1, 2.....

(b) : Using (a) show that


π
∫ sin nx sin mx dx = 0 if n ≠ m
0

Ans. 4(a) :
Method 1 :

The characteristic polynomial p( r ) = r 2 + n 2 has roots r = ± in and therefore the general


solution

Differential Equations (16)


φn ( x ) = cn cos nx + d n sin nx
y (0) = φn (0) = 0 ⇒ φn (0) = cn = 0
y (π ) = φn (π ) = 0 ⇒ φ n (π ) = cn (–1) n = 0.
Thus, φn ( x ) = sin nx is a solution for n = 1, 2, 3,....

Method 2 :
φn ( x ) = sin nx, φn′ ( x ) = n cos nx

φn′′ ( x ) = – n 2 sin nx
Thus, φn′′ ( x ) + n 2φn ( x ) = – n 2 sin nx + n 2 sin nx = 0 .

Since, φn ( x ) = sin nx satisfies φn′′ ( x ) + n 2φn ( x) = 0


and φn (0) = 0, φn (π ) = 0
φn ( x ) = sin nx is a solution of y′′ + n2 y = 0, y (0) = y(π ) = 0.

Ans. 4(b) :
Working on the similar line as in example 2 we get,
π π
(n – m ) ∫ φn ( x) φm ( x) dx = (n2 – m2 ) ∫ sin nx sin mx dx
2 2

0 0
π
= [sin nx (– m cos mx) – sin mx(– n cos nx)]0
= 0 (as sin 0 = sin np = 0)
π
Since n ≠ m, ∫ φ n ( x) φ m ( x) dx = 0.
0

Q5 : Suppose f1, f2 are linearly independent solutions of the constant coefficient equation
y ′′ + a1 y ′ + a2 y = 0, Let W (f1, f2 ) be abbreviated to W. Show that W is constant if and
only if a1 = 0.

Ans. :
φ1 φ2
W = W(φ1, φ2 ) =
φ1′ φ2′
(
= φ1 φ2′ – φ2 φ1′ )
Then (
W ′ = φ1 φ 2′ – φ 2 φ1′

)
= φ1 φ2′′ + φ1′ φ2′ – φ2′ φ1′ – φ2 φ1′′

= φ1 φ2′′ – φ2 φ1′′
But f1 and f2 are solutions of y ′′ + a1 y ′ + a2 y = 0.

Differential Equations (17)


Therefore φ1′′ + a1 φ1′ + a2 φ1 = 0 ⇒ φ1′′ = – a1 φ1′ – a2φ1

Similarly, φ2′′ = – a1 φ2′ – a2φ2

Thus, W ′ = φ1 (– a1 φ2′ – a2 φ2 ) – φ 2 (– a1 φ1′ – a2 φ1 )

= – a1 ( φ1 φ 2′ – φ 2 φ1′ )
= – a1W
Thus, W ′ = 0 iff a1 = 0
Therefore W = constant if and only if a1 = 0

Q6 : Let f1, f2 be two different function on an interval I, which are not necessarily
solutions of an equation L(y) = 0. Prove the following
(a) If f1, f2 are linearly dependent on I then W(f1, f2 ) (x) = 0 for all x in I
(b) If W(f1, f2 ) (x0) ¹ 0 for some x0 in I, then f1, f2 are linearly independent on I.
(c) W(f1, f2 )(x) = 0 for all x in I does not imply that f1, f2 are linearly dependent on I.
(d) W(f1, f2 ) (x) = 0 for all x in I and f2 (x) ¹ 0 on I, imply that are f1, f2 linearly
dependent.

Ans. 6(a) :
Suppose f1, f2 are linearly dependent on I then c1 φ1 ( x ) + c2 φ 2 ( x ) = 0 for some non-zero
c1 and c2.
c2
i.e. φ1 ( x) = – φ2 ( x).
c1
φ1 φ2
W(φ1, φ2 ) ( x) = = φ1 ( x) φ2′ ( x) – φ2 ( x) φ1′ ( x)
φ1′ φ2′

 c   c 
∴ W(φ1 , φ2 ) ( x) =  – 2 φ2 ( x )  φ2′ ( x) – φ 2 ( x)  – 2 φ2′ ( x )  = 0
 c1   c1 
∴ W(φ1 , φ2 ) ( x ) = 0 for all x ∈ I.

Ans. 6(b) :
Suppose c1 φ1 ( x ) + c2 φ 2 ( x ) = 0 then

c1 φ1′ ( x) + c2 φ 2′ ( x) = 0
Thus we have a system of equation
 φ1 ( x ) φ 2 ( x )   c1   0
  c  =  0
φ1′ ( x ) φ2′ ( x )   2  

Differential Equations (18)


Therefore at x = x0

 φ1 ( x0 ) φ 2 ( x0 )   c1   0 
  c  = 0
φ1′ ( x0 ) φ2′ ( x0 )   2  
Thus, c1 = c2 = 0 if and only if the coefficient matrix is invertible i.e. the determinant of
coefficient matrix is non-zero

 φ1 ( x0 ) φ 2 ( x0 ) 
But   = W(φ1 , φ2 ) ( x0 ) ≠ 0
φ1′ ( x0 ) φ2′ ( x0 )
Since, W(φ1, φ2 ) ( x0 ) ≠ 0 ⇒ c1 = c2 = 0
∴ c1 φ1 ( x ) + c2 φ 2 ( x ) = 0 ⇒ c1 = c2 = 0.
Hence f1 and f2 are linearly independent on I.

Ans. 6(c) :

Define φ1 ( x) = x 2 , φ2 ( x) = x | x |

for x > 0, | x |= x ∴ φ1 ( x) = x 2 , φ2 ( x) = x 2

x2 x2
∴ W(φ1, φ2 ) = = 0.
2x 2x

for x = 0, φ1 ( x ) = φ 2 ( x ) = 0 ∴ W(φ1 , φ 2 ) = 0

for x < 0, | x |= – x ⇒ φ1 ( x) = x 2 and φ2 ( x) = – x 2

x2 – x2
∴ W(φ1, φ 2 ) = = 0.
2 x –2 x

Thus W(φ1, φ2 ) ( x) = 0 for – ∞ < x < ∞

Let c1 φ1 ( x ) + c2 φ 2 ( x ) = 0

for x > 0, c1 φ1 ( x ) + c2 φ2 ( x ) = (c1 + c2 ) x 2 = 0.

⇒ c1 + c2 = 0 .......... (i)

for x < 0, c1 φ1 ( x ) + c2 φ2 ( x ) = c1 x 2 – c2 x 2 = 0.

⇒ c1 – c2 = 0 .......... (ii)
But c1 + c2 = 0 and c1 – c2 = 0 ⇒ c1 = c2 = 0
Thus, c1 φ1 + c2 φ 2 = 0 ⇒ c1 = c2 = 0
Therefore f1, f2 are linearly independent.

Differential Equations (19)


Ans. 6(d) :
φ1 ( x ) φ2 ( x)
W(φ1, φ2 ) ( x) = 0 ⇒ W(φ1, φ2 ) ( x) = =0
φ1′ ( x) φ2′ ( x)

⇒ φ1 ( x ) φ2′ ( x ) – φ2 ( x ) φ1′ ( x ) = 0

⇒ φ2 ( x ) φ1′ ( x ) – φ1 ( x ) φ2′ ( x ) = 0
Since φ 2 ( x ) ≠ 0 ∀x ∈ I

φ2 ( x) φ1′ ( x) – φ1 ( x) φ2′ ( x)
∴ =0
φ2 2 ( x )

 φ ′ φ
⇒  1  = 0 ⇒ 1 = constant = k (say)
 φ2  φ2

Therefore φ1 ( x ) = kφ 2 ( x ) and hence φ1, φ2 are linearly dependent.

Q7 : If f1, f2 are two solution of L(y) = 0 on an interval I containing a point x0, then
W(φ1 , φ 2 )( x ) = e – a1 ( x – x0 ) W(φ1 , φ 2 )( x0 ).

Ans. :
Since f1, f2 are solution of L(y) = 0,
φ1′′ + a1 φ1′ + a2φ1 = 0

φ2′′ + a1 φ2′ + a2 φ2 = 0
On multiplying the first equation by –f2, second equation by f1 and adding we obtain
φ1 φ2′′ – φ2 φ1′′ + a1 (φ1 φ2′ – φ2 φ1′ ) + a2 (φ1 φ 2 – φ2 φ1 ) = 0
(φ1 φ2′′ – φ2 φ1′′ ) + a1 (φ1 φ2′ – φ2 φ1′ ) = 0 .......... (i)
φ1 ( x) φ2 ( x)
Let W = W (φ1, φ2 ) ( x ) =
φ1′ ( x ) φ 2′ ( x )
Then W = φ1( x) φ 2′ ( x) – φ 2 ( x) φ1′ ( x)
and W ′ = φ1 ( x) φ2′′ ( x) + φ1′ ( x) φ2′ ( x) – φ 2′ ( x) φ1′ ( x) – φ2 ( x) φ1′ ( x)

= φ1 ( x ) φ 2′′ ( x ) – φ2 ( x ) φ1′′ ( x )
Thus, equation (i) becomes
W ′ +a1W = 0.
Thus W satisfies the first order differential equation
W ′ +a1W = 0

Differential Equations (20)


Hence, W( x) = c ⋅ e – a1 x where c is constant of integration. At x = x0 we get
W( x0 ) = c ⋅ e – a1, x0 i.e. c = e a1x0 W( x0 )
Thus, W( x) = e a1x0 W( x0 )e – a1x

= e – a1 ( x – x0 ) W( x0 )
Therefore W(φ1, φ2 ) ( x) = e – a1 ( x – x0 ) W(φ1 , φ 2 ) ( x0 )

EXERCISES

1. The functions f1, f2 are defined on –∞ < x < ∞


Determine whether they are linearly dependent or independent there.
(i) φ1 ( x) = cos x, φ2 ( x) = sin x

(ii) φ1 ( x ) = sin x, φ2 ( x ) = e ix

(iii) φ1 ( x) = sin nx, φ2 ( x) = cos nx

(iv) φ1 ( x) = 1, φ2 ( x) = cos x

(v) φ1 ( x) = sin 2 x, φ2 ( x) = cos 2 x

(vi) φ1 ( x ) = 1, φ2 ( x) = sin 2 x, φ3 ( x ) = cos 2 x

(vii) φ1 ( x) = cos x, φ2 ( x) = ei x + e – i x
2. State whether the following statements are true or false.
(a) If f1, f2 are linearly independent functions on an interval I, they are linearly independent
on any interval J contained inside I.
(b) If f1, f2 are linearly dependent on an internal I, they are linearly dependent on any internal
J contained inside I.
(c) If f1, f2 are linearly independent solutions of L (y) = 0 on an internal I, they are linearly
independent an any internal J contained inside I.
(d) If f1, f2 are linearly dependent solutions of L (y) = 0 on an interval I, they are linearly
dependent on any internal J contained inside I.
Ans. : 1.
(i) independent (ii) independent (iii) independent
(iv) independent (v) independent (vi) dependent
(vii) dependent.
Ans. : 2.
(a) false (b) true (c) true (d) true

S
Differential Equations (21)
Unit 3 : The Homogeneous Equation of Order n
Everything we have done for the second order equation can be carried over to the case of
the equation of order n. Here, we are concerned with the equation
L( y ) = y (n ) + a1 y (n –1) + a2 y (n –2) + ⋅⋅⋅ + an y = 0,
where, a1, a1, a3 ,......, an are constants.

Theorem 1.3.1 :
Let r 1 , r 2 , r 3 ,......., r s be the distinct roots of the characteristic polynomial
p(r ) = r n + a1r n –1 + a2 r n –2 + ⋅⋅⋅ + an and suppose r i has multiplicity mi (m1 + m2 + m3 + ⋅⋅⋅
+ ms = n). Then n functions

e r1x , xe r1x ,...., x m1 –1e r1x ; e r2 x , xe r2 x ,...., x m2 –1e r2 x ;.....;


e rs x , xe rs , x , x 2 e rs x ,...., x ms –1e rs x

are solutions of L( y ) = y (n ) + a1 y (n –1) + a2 y (n –2) + ⋅⋅⋅ + an y = 0

Proof : Suppose ri is a root of p(r) of multiplicity mi. Then p (r ) = (r – ri ) mi q (r ) where q is a


polynomial of degree n – mi. On differentiating p(r), (mi – 1) times we get,
p′(r ) = (r – ri ) mi q′( r ) + mi (r – ri ) mi –1 q ( r )
= (r – ri ) mi –1 [q′(r )(r – ri ) + mi q (r )]

p′′(r ) = (r – ri )mi q′′(r ) + 2mi (r – ri )mi –1 q′ (r ) + mi (mi –1) (r – ri )mi –2 q(r )

= (r – ri )mi –2  (r – ri ) 2 q′′(r ) + 2mi (r – ri )q′(r ) + mi (mi –1)q (r ) 


 
= (r – ri ) mi –2 [Polynomial of order n – mi ]
and so on

[Polynomial of order n – mi ]
mi –( mi –1)
p ( mi –1) (r ) = (r – ri )
= (r – ri ) [Polynomial of order n – mi ]
Therefore, p( ri ) = p′( ri ) = p′′( ri ) = ⋅⋅⋅ = p ( mi –1) ( ri ) = 0.

Let erx be a solution of L(y) = 0. We see that L(e rx ) = p (r )e rx where p(r ) = r n + a1r n –1
+ a2r n –2 + ⋅⋅⋅ + an.

Therefore L(e ri x ) = p (ri )e ri x = 0. Thus e ri x is a solution of L(y) = 0.

If we differentiate L(e rx ) = p(r )e rx k times with respect to r we obtain

 ∂ k rx 
∂k
∂r k
L ( e rx
) = L  k e =L x e
 ∂r 
k rx
( )
Differential Equations (22)
 k ( k –1) ( k –2) 
=  p ( k ) (r ) + kp ( k –1) (r ) x + p (r ) x 2 + ⋅⋅⋅ + p (r ) x k  e rx
 2! 
Thus for r = r i and k = 0, 1, 2,.....m i – 1 we get L( x k e ri x ) = 0. Therefore
x k e ri x , k = 0,1, 2,......mi –1 , are solutions of L(y) = 0. This is true for every characteristic root ri
with multiplicity mi. i.e. x k e ri x , k = 0,1, 2,....mi –1, i = 1, 2,3,....s are solutions of L(y) = 0 and
the result follows.

Theorem 1.3.2 :
The n solutions of L(y) = 0 given in theorem 1.3.1 are linearly independent on any
interval I.
Proof : We prove that functions given in theorem 1.3.1 satisfy the condition given in defination
1.3.

Suppose we have n constants cij, i = 1, 2....s, j = 0,.....mi –1


Such that
(c e + c x e + c x e + .... + c x
10
r1x
11
r1 x
12
2 r1 x
1( m1 –1)
m1 –1 r1x
e )
+ (c e + c x e + c x e + .... + c )
r2 x r2 x 2 r2 x m2 –1 r2 x
20 21 22 2( m2 –1) x e

+.... + (c e + c x e + c x e + .... + c
s0
rs x
s1
rs x
s2
2 rs x
s ( ms –1) x
ms –1 rs x
e ) = 0.
Define pi ( x) = ci 0 + ci1x + ci 2 x 2 + .... + ci (mi –1) x mi –1

Then p1 ( x)e r1x + p2 ( x)er2 x + p3 ( x )e r2 x + ..... + ps ( x)e rs x = 0.


Assume that not all constants cij are zero. Then there will be at least one of the polynomials
pi which is not identically zero on I. Suppose ps(x) is not identically zero on I. On dividing above
equation by e r1x we get
p1 ( x) + p2 ( x)e (r2 – r1 ) x + p3 ( x)e (r3 – r1 ) x + .... + ps ( x )e (rs – r1 ) x = 0.
Upon differentiating above equation sufficiently many (at most mi) times, we obtain the
expression of the form
Q2 ( x )e(r2 – r1 ) x + Q3 ( x )e( r3 – r1 ) x + .... + Qs ( x )e( rs – r1 ) x = 0
i.e. Q2 ( x ) + Q3 ( x)e ( r3 – r2 ) x + .... + Qs ( x )e( rs – r2 ) x = 0
where the Qi’s are polynomials, degree of Qi is equal to degree of Pi and Qs does not vanish
identically.
Continuing this process we finally arrive at a situation where,
Rs ( x)e rs x = 0,
on I and Rs is a polynomial, degree of Rs is equal to degree of Ps, which does not vanish
identically on I. But Rs ( x )e rs x = 0 implies Rs ( x ) = 0 is a contradiction. Therefore our supposition
that Ps ( x) is not identically zero is not true. Thus Ps ( x ) = 0 for all x in I.
Differential Equations (23)
Thus all constants Cij = 0 proving that the n solutions given in theorem 3.1 are linearly
independent on an interval I.
* Initial value problem for nth order equations.
The problem of finding a solution f of
L( y ) = y ( n) + a1 y ( n –1) + a2 y ( n –2) + .... + an y = 0 satisfying
φ ( x0 ) = α1, φ ′( x0 ) = α 2 ,......., φ n –1 ( x) = α n where a1, a2 , a3 ,......., an and
α1 ,α 2 ,α3 ,.......,α n are constants is denoted by
L( y ) = 0, y ( x0 ) = α1 , y′( x0 ) = α 2 ,....., y ( n –1) ( x0 ) = α n
and is called an initial value problem.

Theorem 1.3.3 :
Let f be any solution of
L( y ) = y ( n ) + a1 y ( n –1) + a2 y ( n –2) + .... + an y = 0
on an interval I containing a point x0. Then for all x in I
|| φ ( x0 ) || e – k|x – x0| ≤ || φ ( x) || ≤ || φ ( x0 ) || e k ( x – x0 )
where, k = 1+ | a1 | + | a2 | + | a3 | +....+ | an |
1
and || φ ( x) || = | φ ( x) | + | φ ′ ( x) | +....+ | φ
2 2 ( n –1)
( x) | 22
 
Proof : This proof is similar to the proof of theorem 1.1.4.
Let u ( x) = || φ ( x) ||2
= | φ |2 + | φ ′ |2 +....+ | φ (n –1) |2
= φ φ + φ ′ φ ′ + .... + φ (n –1)φ (n –1)
Hence u ′( x) = φ ′φ + φ φ ′ + φ ′′ φ ′ + φ ′φ ′′ + .... + φ ( n –1) φ ( n) + φ ( n) φ ( n –1)
Therefore | u ′( x) | ≤ 2 | φ ( x) | | φ ′( x) | + 2 | φ ′ | | φ ′′ | +.... + 2 | φ ( n –1) | | φ ( n) |
Since f is solution of L(y) = 0, L(f ) = 0 and therefore
φ ( n ) = – a1 φ ( n –1) – a2 φ ( n –2) – a3 φ ( n –3) – .... – an φ
On substituting the expression for φ (n) we get

| u′( x) | ≤ 2 | φ | | φ ′ | +2 | φ ′ | | φ ′′ | +..... + 2 | φ (n –2) | | φ (n –1) |


+2 | a1 | | φ ( n –1) |2 + 2 | a2 | | φ ( n–1) | | φ ( n –2) | + ..... + 2 | an | | φ ( n –1) | | φ |
(| a | – | b |) 2 ≥ 0 ⇒ | a |2 + | b |2 ≥ 2 | a | | b |
 
| u ′( x) | ≤ (| φ |2 + | φ ′ |2 ) + (| φ ′ |2 + | φ ′′ |2 ) + .... + (| φ ( n –2) |2 + | φ ( n –1) |2 )
+ | a1 | (| φ ( n –1) |2 + | φ ( n –1) |2 ) + ....+ | an | (| φ ( n –1) |2 + | φ |2 )
Differential Equations (24)
≤ (1+ | an |) | φ |2 +(1 + 1+ | an –1 |) | φ ′ |2 + (2+ | an –2 |) | φ ′′ |2
+.... + ( 2+ | a2 |) | φ (n –2) |2 +(1 + 2 | a1 | + | a2 | +....+ | an |) | φ (n –1) |2
Since each coefficient on the right hand side is less than 2k we have
| u′( x) | ≤ 2k (| φ |2 + | φ ′ |2 +....+ | φ ( n –1) |2 )
= 2k || φ ( x ) ||2 = 2k u ( x )
Therefore | u ′( x ) | ≤ 2ku ( x )
Thus, we get –2 k u ( x ) ≤ u ′( x) ≤ 2ku ( x)
u ′ – 2 ku ( x ) ≤ 0 implies (e –2kxu ( x ))′ ≤ 0
Integrating above inequality between the limits x0 to x for x > x0 yields
e –2 kx u ( x ) – e –2kx0 u ( x0 ) ≤ 0
i.e. u ( x) ≤ e 2k ( x – x0 ) u ( x0 )
Thus, || φ ( x) || ≤ e k ( x – x0 ) || φ ( x0 ) ||
Similarly for x > x0 the inequality
–2 ku ( x) ≤ u′( x) implies
|| φ ( x0 ) || e – k ( x – x0 ) ≤ || φ ( x) ||
Combining the above two inequalities we get the required result for x > x0.
For x < x0 interchange the role of x and x0
We get || φ ( x0 ) || ≤ ek ( x0 – x ) || φ ( x) || ⇒ || φ ( x0 ) || e – k ( x0 – x ) ≤ || φ ( x) ||
and || φ ( x ) || e – k ( x0 – x ) ≤ || φ ( x0 ) || ⇒ || φ ( x) || ≤ e k ( x0 – x) || φ ( x0 ) ||
Thus, || φ ( x0 ) || e – k ( x0 – x ) ≤ || φ ( x) || ≤ e k ( x0 – x ) || φ ( x0 ) ||, ( x < x0 )
which is the required result for x < x0

Theorem 1.3.4 (Uniqueness theorem)


Let α1, α 2 , α 3 ,...., α n be any n constants and let x0 be any real number. On any interval I
containing x0 there exists at most one solution f of L (y) = 0 satisfying φ ( x0 ) = α1 , φ ′ ( x0 ) = α 2 ,
,......., φ ( n –1) ( x0 ) = α n
Proof : Suppose f and y were two solutions of L (y) = 0 on I satisfying the above conditions at
x = x0. i.e.
φ ( x0 ) = ψ ( x0 ) = α1, φ ′( x0 ) = ψ ′( x0 ) = α 2 , ...., φ ( n –1) ( x0 ) = ψ ( n–1) ( x0 ) = α n
Define θ = φ – ψ . Since f and y satisfy L (φ ) = L (ψ ) therefore L (θ ) = 0 and
θ ( x0 ) = φ ( x0 ) – ψ ( x0 ) = 0, θ ′( x0 ) = 0,....,θ (n –1) ( x0 ) = 0.
1
Thus || θ ( x0 ) || = | θ ( x0 ) |2 + | θ ′ ( x0 ) |2 +....+ | θ ( n–1) ( x0 ) |2  2 = 0
 

Differential Equations (25)


Applying theorem 1.3.3 we obtain || θ ( x) || = 0 for all x in I. This implies θ ( x) = 0 for
all x in I.
i.e. φ ( x ) = ψ ( x ) for all x in I.

Theorem 1.3.5
If φ1, φ2 , φ3 ,....φn , are n solutions of L(y) = 0 on an interval I, they are linearly independent
if and only if W(φ1, φ2 , φ3 ,....φn ) ( x) ≠ 0 for all x in I. (definition 1.5)
Proof : The proof is entirely similar to the proof of theorem 1.2.2
Suppose W(φ1, φ2 , φ3 ,....φn ) ( x) ≠ 0 for all x in I. Let c1, c2, c3,...., cn be constants such

that c1φ1( x) + c2 φ2 ( x) + .... + cn φn ( x) = 0 for all x in I.


By differentiating above equation (n – 1) times we get a system of equations as follows.
 φ1 ( x ) φ2 ( x ) φ3 ( x ) L φn ( x)  c
   1   0

 φ1 ( x) ′
φ 2 ( x) ′
φ3 ( x) L ′
φ n ( x)  c2  0
 
    
 1 φ ′′ ( x ) φ ′′ ( x ) φ ′′ ( x ) L φ ′′ ( x )   c3 =  0

2 3 n
M  

M M M M
  M 
 0
φ1( n –1) ( x ) φ2( n –1) ( x ) φ3( n –1) ( x ) φ n( n –1) ( x )   cn 
The coefficient matrix is invertible because the determinant of coefficient matrix is
(definition 1.5) W(φ1, φ2 , φ3 ,....φ n ) ( x ) ≠ 0. On premultiplying the inverse of the coefficient
matrix we get, c1 = c2 = c3..... = cn = 0. This proves that φ1, φ2 , φ3 ,....φn are linearly independent.

Conversely, assume that φ1, φ2 ,....φ n are linearly independent on I. Suppose there is a
point x0 in I such that W(φ1, φ2 , φ3 ,....φn ) ( x0 ) = 0. Then the system of equations

 φ1 ( x0 ) φ2 ( x0 ) φ3 ( x0 ) L φn ( x0 ) 
 c1 
  0 
 φ1′ ( x0 ) φ2′ ( x0 ) φ3′ ( x0 ) L φn′ ( x0 )    0 
   c2   
 φ1′′ ( x0 ) φ 2′′ ( x0 ) φ3′′ ( x0 ) L φ n′′ ( x0 )   c3  = 0 
     

M M M M
 M M 
φ1(n –1) ( x0 ) φ2 (n –1) ( x0 ) φ3(n –1) ( x0 )  cn  0 
φn (n –1) ( x0 ) 

has a solution c1 , c2 , c3, ....., cn where at least one of these numbers is not zero. Let c1, c2 ,....., cn
be such a solution and consider a function
ψ ( x ) = c1 φ1 ( x ) + c2 φ2 ( x ) + .... + cn φ n ( x ).
Now L(ψ ) = 0 and ψ ′( x0 ) = ψ ′′( x0 ) = .... = ψ ( n –1) ( x0 ) = 0.

Therefore || ψ ( x0 ) || = 0 . But then by theorem 1.3.3 , || ψ ( x) || = 0 , for all x in I. Therefore

Differential Equations (26)


by defination of || ψ ( x) || , ψ ( x ) = 0 for all x in I. But then φ1, φ2 , φ3 ,....φ n are linearly dependent.
Thus the supposition W(φ1, φ 2 , φ3 ,....φ n ) ( x0 ) = 0 must be false. Therefore
W(φ1, φ2 , φ3 ,....φn ) ( x) ≠ 0 for all x in I.

Theorem 1.3.6 (Existence Theorem)


Let α1 , α 2 , α3 ,...., α n be any n constants and let x0 be any real number. There exists a
solution f of L(y) = 0 on – ¥ < x < ¥ satisfying
φ ( x0 ) = α1, φ ′( x0 ) = α 2 , φ ′′( x0 ) = α 3 ,...., φ (n –1) ( x0 ) = α n
Proof : Let φ1, φ2 , φ3 ,....φn be any set of n linearly independent solutions of L(y) = 0 on
– ¥ < x < ¥. We will show that there exist unique constants c1, c2 , c3 ,....., cn such that
φ = c1 φ1 + c2 φ 2 + c3 φ3 + .... + cn φn

is a solution of L(y) = 0 satisfying the given initial conditions φ (i ) ( x0 ) = αi , i = 0,1, 2,...., n –1.
These constants c1, c2 , c3 ,....., cn would have to sartisfy
 φ1 ( x0 ) φ 2 ( x0 ) φ3 ( x0 ) L φn ( x0 ) 
 c1   α1 
    α 
 φ1′ ( x0 ) φ 2′ ( x0 ) φ3′ ( x0 ) L φn′ ( x0 ) 
 c2   2
   c3  = α3 
 φ1′′ ( x0 ) φ 2′′ ( x0 ) φ3′′ ( x0 ) L φ n′′ ( x0 ) 
   
 M M M M  M  M 
   cn  α n 
φ1(n –1) ( x0 ) φ 2(n –1) ( x0 ) φ3(n –1) ( x0 ) φn ( n –1)
( x0 )

Since φ1, φ2 , φ3 ,....φ n are linearly independent, by theorem 1.3.5, the determinant of the
coefficients i.e. W(φ1, φ 2 , φ3 ,....φ n ) ( x0 ) ≠ 0. Thus the coefficient matrix is invertible. Therefore
there is a unique set of constants c1, c2 , c3 ,....., cn satisfying above system of equations. For this
choice of c1, c2 , c3 ,....., cn the function
φ ( x ) = c1 φ1 ( x ) + c2 φ 2 ( x ) + c3 φ3 ( x ) + .... + cn φ n ( x )
will be the desired solution.

Theorem 1.3.7 :
Let φ1, φ2 , φ3 ,....φ n be n linearly independent solutions of L(y) = 0 on an interval I. If
c1, c2 , c3 ,....., cn are any constants
φ ( x ) = c1 φ1 ( x ) + c2 φ 2 ( x ) + c3 φ3 ( x ) + .... + cn φ n ( x )
is a solution and every solution may be represented in this form.
Proof : Since φi , i = 1, 2, 3.....n is solution of L(y) = 0, L(φi ) = 0, i = 1, 2, 3.....n .

Therefore L(φ ) = c1 L(φ1 ) + c2 L(φ 2 ) + c3 L(φ3 ) + .... + cn L(φn ) = 0 and


φ = c1 φ1 + c2 φ2 + c3 φ3 + .... + cn φn is a solution of L(f ) = 0.
Differential Equations (27)
Let f be any solution of L(y) = 0 and x0 be in I.
Suppose φ ( x0 ) = α1 , φ ′( x0 ) = α 2 , φ ′′( x0 ) = α3 ,...., φ ( n –1) ( x0 ) = α n .
By existence theorem 1.3.6 there exist unique constants c1, c2 , c3 ,....., cn such that
ψ = c1 φ1 + c2 φ2 + c3 φ3 + .... + cn φ n
is a solution of L(y) = 0 on I satisfying
ψ ( x0 ) = α1,ψ ′( x0 ) = α 2 ,ψ ′′( x0 ) = α 3 ,....,ψ (n –1) ( x0 ) = α n
The uniqueness theorem 1.3.4 implies that f = y. Thus φ = c1 φ1 + c2 φ2 + c3 φ3 + .... + cn φn .

Theorem 1.3.8
Let φ1, φ2 , φ3 ,....φn be n solutions of L(y) = 0 on an interval I constaining a point x0. Then

W(φ1, φ2 , φ3 ,....φn ) ( x) = e – a1 ( x – x0 ) W(φ1, φ2 , φ3 ,....φn ) ( x0 )


Proof :
φ1 ( x ) φ 2 ( x) φ3 ( x ) L φ n ( x)
φ1′ ( x) φ 2′ ( x) φ3′ ( x) L φ n′ ( x)
W(φ1, φ2 , φ3 ,...., φn ) ( x) = φ1′′ ( x) φ 2′′ ( x) φ3′′ ( x) L φn′′ ( x)
M M M M
φ1( n –1) ( x ) φ 2( n –1) ( x ) φ3( n –1) ( x ) φ n( n –1) ( x )
By differentiating above determinant row-wise we get,

φ1′ φ2′ φ3′ L φn′


φ1′ φ2′ φ3′ L φn′
W ′(φ1, φ 2 , φ3 ,....,φ n ) ( x) = φ1′′ φ2′′ φ3′′ L φn′′
M M M M
φ1(n –1) φ 2(n –1) φ3(n –1) φn (n –1)
φ1 φ2 φ3 L φn φ1 φ2 φ3 L φn
φ1′′ φ 2′′ φ3′′ L φ n′′ φ1′ φ2′ φ3′ L φ n′
+ φ1′′ φ 2′′ φ3′′ L φ n′′ + .... + φ1′′ φ2′′ φ3′′ L φ n′′
M M M M M M M M
φ1(n –1) φ 2(n –1) φ3(n –1) L φ n (n –1) φ1(n ) φ 2(n) φ3(n ) L φn (n )
Since two rows are identical the value of first (n – 1) determinants is zero. Therefore
φ1 φ2 φ3 L φn
φ1′ φ 2′ φ3′ L φ n′
W ′(φ1, φ2 , φ3 ,...., φn ) ( x) = φ1′′ φ 2′′ φ3′′ L φ n′′
M M M M
φ1( n) φ 2( n) φ3( n) φ n ( n)
Differential Equations (28)
Since each φi , i = 1, 2,3,...., n is a solution of L(y) = 0 φi ( n) = – (a1 φi ( n–1) + a2 φi ( n–2)

+ a3 φi ( n –3) .... + an φi ). Hence,

φ1 ( x ) φ2 ( x) φ3 ( x ) L φn ( x)
φ1′ ( x) φ 2′ ( x ) φ3′ ( x) L φ n′ ( x )
W ′(φ1, φ2 , φ3 ,..., φn ) ( x) = M M M L M
φ1(n –2) ( x) φ2 (n –2) ( x) φ3(n –2) ( x) φ n (n –2) ( x)
– a1 φ1(n –1) ( x) – a1 φ 2(n –1) ( x) – a1 φ3(n –1) ( x) L – a1 φ n (n –1) ( x)

Since, φ1 ( x ) φ2 ( x) φ3 ( x ) L φn ( x)
φ1′ ( x ) φ 2′ ( x ) φ3′ ( x ) L φ n′ ( x )
M M M L M =0
φ1( n –2) ( x) φ2( n –2) ( x) φ3( n –2) ( x) φ n( n –2) ( x)
– ak φ1(n – k ) ( x ) – ak φ2 (n –k ) ( x ) – ak φ3(n – k ) ( x ) L – ak φ n (n – k ) ( x )

for k = 2, 3, 4,....n, as two rows of the determinant are constant multiplies of each other are
Thus,
φ1 ( x) φ 2 ( x) φ3 ( x) L φ n ( x)
φ1′ ( x ) φ 2′ ( x) φ3′ ( x) L φ n′ ( x )
W ′(φ1, φ2 , φ3 ,..., φn ) ( x) = – a1 φ1′′ ( x ) φ 2′′ ( x) φ3′′ ( x) L φ n′′ ( x )
M M M M
φ1(n –1) ( x ) φ 2 (n –1) ( x) φ3(n –1) ( x) L φ n (n –1) ( x )

= – a1 W(φ1, φ 2 , φ3 ,..., φ n ) ( x)

Thus W ′ + a1W = 0. On integrating this equation between the limits x0 to x we get ,

e a1 x W ( x ) = e a1x0 W ( x0 )

or W ( x) = e – a1 ( x – x0 ) W ( x0 )

Thus W(φ1 , φ2 , φ3 ,..., φn ) ( x) = e – a1 ( x – x0 ) W(φ1 , φ2 , φ3 ,..., φn ) ( x0 )

Theorem 1.3.9
Let φ1, φ2 , φ3 ,....φn be n solutions of L(y) = 0 on an interval I containing x0. Then they are
linearly independent on I if and only if W(φ1, φ 2 , φ3 ,..., φ n ) ( x0 ) ≠ 0

Differential Equations (29)


Proof : By theorem 1.3.5 the solutions φ1, φ2 , φ3 ,....φn of L(y) = 0 are linearly independent
on an interval I if and only if W(φ1, φ2 , φ3 ,..., φn ) ( x) ≠ 0 for all x in I.

But W(φ1 , φ2 , φ3 ,..., φ n ) ( x) = e – a1( x – x0 ) W(φ1 , φ2 , φ3 ,..., φ n ) ( x0 ) (by theorem 1.3.8.)


Therefore W(φ1, φ2 , φ3 ,..., φn ) ( x) ≠ 0 if and only if W(φ1, φ 2 , φ3 ,..., φ n ) ( x0 ) ≠ 0 and the result
follows.

EXAMPLES

Q.1. Consider the equation


y (5) – y (4) – y′ + y = 0
(a) Compute five linearly independent solutions.
(b) Compute the wronkian of the solutions found in (a).
(c) Find that solution f satisfying
φ (0) = 1, φ ′ (0) = φ ′′ (0) = φ ′′′ (0) = φ (4) (0) = 0.
Ans (a) :
The characteristic equation
p( r ) = r 5 – r 4 – r + 1
= r 4 (r –1) – ( r –1)
= (r 4 –1) (r –1)
= (r 2 –1) (r 2 + 1) (r –1)
= (r + 1) (r –1) (r 2 + 1) (r – 1)
Thus the characteristic roots are 1, 1, –1, i, – i

Therefore φ1 ( x ) = e x , φ2 ( x) = x e x , φ3 ( x ) = e – x , φ4 ( x ) = sin x φ5 ( x ) = cos x are solutions


of the given differential equation.
Ans (b) :
W(φ1, φ2 , φ3 , φ4 , φ5 ) ( x) = e – a ( x – x0 ) W(φ1, φ2 , φ3 , φ4 , φ5 ) ( x0 )
For the given equation a1 = – 1. Let x0 = 0 then

W(φ1, φ2 , φ3 , φ4 , φ5 ) ( x) = e x W(φ1, φ2 , φ3 , φ 4 , φ5 ) (0).

ex x ex e– x sin x cos x
ex (1 + x) e x –e– x cos x – sin x
W (φ1 , φ 2 , φ3 , φ 4 , φ5 ) ( x) = e x (2 + x) e x e– x – sin x – cos x
ex (3 + x) e x –e– x – cos x sin x
ex (4 + x) e x e– x sin x cos x
Differential Equations (30)
1 0 1 0 1
1 1 –1 1 0
W (φ1, φ2 , φ3 , φ4 , φ5 ) (0) = 1 2 1 0 –1
1 3 –1 –1 0
1 4 1 0 1
The row transformations
R2 – R1 , R3 – R1 , R4 – R1 , R5 – R1 gives
1 0 1 0 1
0 1 –2 1 –1
W(φ1, φ2 , φ3 , φ4 , φ5 ) (0) = 0 2 0 0 –2
0 3 –2 –1 –1
0 4 0 0 0
1 –2 1 –1
2 0 0 –2
=
3 –2 –1 –1
4 0 0 0

0 0 –2 2 0 –2 2 0 0
= –2 –1 –1 + 2 3 –1 –1 + 3 –2 –1
0 0 0 4 0 0 4 0 0

2 0 0
+ 3 –2 –1 = – 32
4 0 0

Thus, W(φ1, φ2 , φ3 , φ4 , φ5 ) = e x W(φ1, φ2 , φ3 , φ4 , φ5 ) (0) = –32e x

Ans (c) :
The general solution f is φ ( x) = c1e x + c2 x e x + c3 e – x + c4 sin x + c5 cos x
The initial conditions φ (0) = 1, φ ′ (0) = φ ′′(0) = φ ′′′(0) = φ (iv) (0) = 0 gives the following
system of equations.

1 0 +1 1 0  c1  1 
1 1 –1 1 0   c   0
  2  
1 2 1 0 –1  c3  =  0
     
1 3 –1 –1 0   c4   0
1 4 1 0 1   c5   0

The row transformation R2 – R1 , R3 – R1 , R4 – R1 , R5 – R1 gives

Differential Equations (31)


1 0 +1 10  c1   1 
0 1 –2 1 –1  c   –1
  2  
0 2 0 0 –2   c3  =  –1
     
0 3 –2 –1 –1  c4   –1
 0 4 0 0 0   c5   –1
Solving the above system of equations simultaneously we get the values of c1, c2,
c3, c4, c5.
1
From last equation we get 4c2 = – 1 gives c2 = –
4
From the third row of the above system we get,
1
2c2 – 2c5 = –1 gives c5 =
4
From second and fourth row we get,
c2 – 2c3 + c4 – c5 = –1
3c2 – 2c3 – c4 – c5 = –1
Substitution of c2 and c5 in above equations give
1
–2c3 + c4 = –
2
–2c3 – c4 = 0
1 1
Thus, c3 = , c4 = –
8 4
5
From first row we get, c1 =
8
Thus, φ ( x) = c1e x + c2 x e x + c3 e – x + c4 sin x + c5 cos x
5 1 1 1 1
= e x – x e x + e – x – sin x + cos x
8 4 8 4 4
is the required solution.

Q.2. Find all solutions of the following equations.


(a) y ′′′ – 8 y = 0 (b) y (4) + 16 y = 0 (c) y ′′′ – 5 y ′′ + 6 y ′ = 0

(d) y (iv) –16 y = 0 (e) y ′′′ – 3 y ′ – 2 y = 0 (f) y (4) + 5 y′′ + 4 y = 0

Ans. (a) :

The characteristic polynomial is p(r ) = r 3 – 8 and its roots are 2, – 1 + 3 i, – 1 – 3 i


Thus, three linearly independent solutions are given by e 2 x , e(–1+ 3 i) x
, e (–1– 3 i) x and any
solution f has the form φ ( x) = c1 e2 x + c2e(–1+ 3 i) x
+ c3e(–1– 3 i) x where c1, c2, c3 are any
constants.
Differential Equations (32)
Ans. (b) : The characteristic polynomial is p (r ) = r 4 + 16

p( r ) = r 4 – (2 i ) 4 = ( r 2 + (2 i ) 2 ) (r 2 – (2 i ) 2 )
= (r 2 – i 2 (2 i )2 ) (r 2 – ( i 2) 2 )
= ( r + 2i i ) ( r – 2i i ) ( r + 2 i ) ( r – 2 i )
Thus, p ( r ) = ( r + 2i i ) ( r – 2i i ) ( r + 2 i ) ( r – 2 i )
π
π π i
i = cos + i sin = e 2
2 2
1
 iπ 2 π
2  = e 4 = cos π + i sin π
i
\ i =  e
  4 4
 
1+ i i(1 + i) –1 + i
Therefore i= , i i= =
2 2 2
The roots of characteristic polynomial are – 2(–1 + i), 2(–1 + i ), 2(1 + i ), – 2(1 + i)
Thus four linearly independent solutions are
2 + 2 i) x 2 +i 2 ) x
e( 2 –i 2 ) x
, e(– , e( , e(– 2 –i 2 ) x

and every solution f has the form


2 +i 2 ) x 2 +i 2) x
φ ( x) = c1e( 2 –i 2 ) x
+ c2 e(– + c3 e( + c4e(– 2 –i 2 ) x

Ans. (c) : The characteristic polynomial is p(r ) = r – 5r + 6r and its roots are 0, 3, 2. Thus
3 2

three linearly independent solutions are given by 1, e3x, e2x and any solution f has the
form φ ( x) = c1 e3 x + c2 e2 x + c3

Ans. (d) : The characteristic polynomial is p(r ) = r 4 –16 = ( r 2 + 4) ( r 2 – 4) = (r + 2i) ( r – 2i)


(r + 2) (r – 2) and its roots are 2, – 2, 2i, –2i. Thus four linearly independent solutions are
given by e 2 x , e –2 x , cos 2 x, sin 2 x and every solution f has the form

φ ( x) = c1e 2 x + c2e –2 x + c3 cos 2 x + c4 sin 2 x


Ans. (e) : The characteristic polynomial is
p( r ) = r 3 – 3r – 2 = ( r + 1) ) ( r 2 – r – 2)
1+ 5 1 – 5
and its roots are –1, , .
2 2  1– 5 
x  x
(1+ 5 )
Thus, three linearly independent solutions are e – x , e 2 , e 2  , and every solution
f has the form
1+ 5 (
1– 5
)x
( )x 2
φ ( x) = c1e –x
+ c2 e 2 + c3e
Differential Equations (33)
Ans. (f) : The characteristic polynomial is
p(r ) = r 4 + 5r 2 + 4 = (r 2 + 4) (r 2 + 1)
and its roots are 2i, –2i, i, –i. Thus four linearly independent solutions are
cos 2 x, sin 2 x , cos x, sin x and every solution f has the form
φ ( x) = c1 cos 2 x, + c2 sin 2 x + c3 cos x + c4 sin x.
Q.3. Consider the equation y ′′′ – 4 y ′ = 0
(a) Compute three linearly independent solutions.
(b) Compute the wronkian of the solutions found in (a).
(c) Find the solution f satisfying
φ (0) = 0, φ ′(0) = 1, φ ′′(0) = 0

Ans. (a) : The characteristic polynomial p (r ) = r 3 – 4r and its roots are 0, 2, –2. Thus, three
linearly independent solution are e° = 1, e 2 x , e –2 x and every solution f has the form

φ ( x) = c1 + c2e 2 x + c3e –2 x
Ans. (b) :
W(φ1, φ2 , φ3 ) ( x) = e0( x –0) W(φ1, φ 2 , φ3 ) (0)

1 e2x e –2 x
W(φ1, φ2 , φ3 ) ( x) = 0 2e 2 x –2e –2 x
0 4e 2 x 4e –2 x

1 1 1
W(φ1, φ2 , φ3 ) (0) = 0 2 –2
0 4 4
Thus, W(φ1, φ2 , φ3 ) ( x) = 16 .

Ans. (c) :
φ (0) = 0, φ ′ (0) = 1, φ ′′ (0) = 0,
φ ( x) = c1 + c2e 2 x + c3e –2 x , φ (0) = c1 + c2 + c3 = 0 and so on
1 1 1   c1  0 
 0 2 –2  c  = 1 
   2  
 0 4 4   c3   0
R3 – 2 R2 gives
1 1 1   c1   0 
 0 2 –2  c  =  1 
   2  
 0 0 8   c3   –2

Differential Equations (34)


1 1 1
Therefore c3 = – , 2c2 – 2c3 = 1 ⇒ c2 – c3 = ⇒ c2 =
4 2 4
c1 + c2 + c3 = 0 ⇒ c1 = 0

Thus, φ ( x ) = c1 + c2e 2 x + c3e –2 x =


4
(
1 2x
)
e – e –2 x is the required solution.

EXERCISE

1. Are the following statements true or false ?


(a) If φ1, φ2 , φ3 ,...., φn are linearly independent functions on an interval I, then any subset of
them forms a linearly independent set of functions on I.
(b) If φ1, φ2 , φ3 ,...., φn are linearly dependent functions on an interval I, then any subset of
them forms a linearly dependent set of functions on I.
2. Are the following sets of functions defined on – ¥ < x < ¥ linearly independent or
dependent ? why ?
(a) φ1 ( x ) = 1, φ2 ( x) = x, φ3 ( x ) = x 2

(b) φ1 ( x ) = ei x , φ2 ( x) = sin x, φ3 ( x ) = 2 cos x

(c) φ1 ( x ) = x, φ2 ( x ) = e 2 x , φ3 ( x) = | x |

3. Find a basis of solutions of the differential equations.


(a) y ′′ + 5 y ′ + 4 = 0 (b) y ′′′ + 6 y ′′ + 12 y ′ + 8 y = 0
(c) y (4) – y = 0

4. Find the general solution of each of the following equations.


x 4x
(i) 6 y ′′ –11 y ′ + 4 y = 0 (Ans. y ( x ) = c1 e 2 + c2 e3)

(ii) y′′ + 2 y′ – y = 0 (Ans. y ( x) = c1 e(–1+ 2)x


+ c2 e(–1– 2)x
)

(iii) y ′′′ + y ′′ – 6 y ′ = 0 (Ans. y ( x ) = c1 + c2 e 2 x + c3 e –3 x )

(iv) y (4) – 2 y ′′ = 0 (Ans. y ( x ) = c1 + c2 x + c3 e 2x


+ c4 e – 2x
)

(v) y′′′ + 8 y = 0 (Ans. y ( x) = c1 e –2 x + c2e2 x + c3 x e 2 x )


5. For each of the following equations find a particular solution which satisfies the given
initial conditions.
(i) y ′′ = 0, y (1) = 2, y ′(1) = –1
(ii) y ′′ + 4 y ′ + 4 y = 0, y (0) = 1, y ′(0) = 1
Differential Equations (35)
(iii) y ′′ – 2 y ′ + 5 y = 0, y (0) = 2, y ′(0) = 4
(iv) y′′ – 4 y′ + 20 y = 0, y (π 2 ) = 0, y′ (π 2 ) = 1
(v) 3 y′′′ + 5 y′′ + y′ – y = 0, y (0) = 0, y′(0) = 1, y′′(0) = –1

[Ans. : (i) y ( x ) = 3 – x, (ii) y ( x) = (1 + 3x) e –2 x (iii) y ( x) = e x (2cos 2 x + sin 2 x)


x
1 2 x –π 9 3  x 9  –x
(iv) e sin 4 x (v) y = e +  –  e .]
4 16  4 16 

Ans. 1 :
(a) True (b) false

Ans. 2 :
(a) independent (b) dependent (iii) independent

Ans. 3 :

(a) φ1 ( x ) = e –4 x , φ2 ( x ) = e – x

(b) φ1 ( x) = e –2 x , φ2 ( x) = xe –2 x , φ3 ( x) = x 2 e –2 x

(c) φ1 ( x ) = e x , φ2 ( x ) = e – x , φ3 ( x ) = cos x, φ4 ( x) = sin x

Unit 4 : The Non-Homogeneous Equation of Order n


We now return to the nth order non-homogeneous linear differential equation with constant
coefficients. In the first part we will discuss the method of finding all solutions of the second
order non-homogeneous equation.
L( y ) = y ′′ + a1 y ′ + a2 y = b( x),
Where b is some continuous function on an interval I. The general solution of the above
equation is
y ( x ) = yc ( x) + y p ( x ),
where, yc(x), the complementary function is the general solution of the related homogenous
equation and yp(x) is a particular solution of the equation.
Suppose we know that yp is a particular solution of the equation L(y) = b(x) and let y be
any other solution. Then,
L(ψ – ψ p ) = L(ψ ) – L(ψ p ) = b ( x ) – b ( x) = 0
on I. This shows that y –yp is a solution of the homogenous equation L(y) = 0. Therefore if f 1,
f 2 are linearly independent solutions of L(y) = 0, there are unique constants c1, c2 such that

Differential Equations (36)


ψ – ψ p = c1 φ1 + c2 φ2
In other words every solution y of L(y) = b (x) can be written in the form
ψ = ψ p + c1 φ1 + c2 φ2
The problem of finding all solutions of L(y) = b (x) reduces to finding a particular
solution yp.

Theorem 1.4.1
Let b(x) be continuous on an interval I. Every solution y of L(y) = b (x) on I can be
written as ψ = ψ p + c1 φ1 + c2 φ2 .

Where yp is a particular solution, f1, f2 are two linearly independent solutions of L(y) = 0
and c1, c2 are constants. A particular solution yp is given by
x
[φ1 (t ) φ 2 ( x) – φ1 ( x) φ 2 (t )] b (t )
ψ p ( x) = ∫ dt.
x0 W(φ1 , φ2 ) (t )
Conversely every such y is a solutions of L(y) = b (x)
Proof :
Let y and yp be two solutions of
L( y ) = y′′ + a1 y ′ + a2 y = b

Then L(ψ – ψ p ) = L(ψ ) – L(ψ p ) = 0

This shows that ψ – ψ p is a solution of a homogeneous equation L(y) = 0. By theorem


1.1.1 there exist two linearly independent solutions f1, f2 and every solution of L(y) = 0 is of the
form c1 φ1 + c2 φ 2 where c1 and c2 are constants. Such a function c1 φ1 + c2 φ2 cannot be a solution
of L(y) = b(x) unless b(x) = 0 on I.
Suppose φ ( x ) = u1 ( x) φ1 ( x ) + u2 ( x ) φ2 ( x ) is a solution of L(y) = b(x) on I.
(This procedure is called as the variation of constants.)
Then
(u1 φ1 + u2 φ2 )′′ + a1 (u1 φ1 + u2 φ2 )′ + a2 (u1 φ1 + u2 φ2 ) = b( x )
i.e. a2 (u1 φ1 + u2 φ2 ) + a1 (u1′ φ1 + u1 φ1′ + u2′ φ2 + u2 φ2′ )

+ (u1′′ φ1 + 2u1′ φ1′ + u1 φ1′′ + u 2′′ φ 2 + 2u1′ φ1′ + u 2 φ 2′′ ) = b ( x )


Therefore
u1 (φ1′′ + a1 φ1′ + a2φ1 ) + u 2 (φ 2′′ + a1 φ 2′ + a2φ 2 )

+ (φ1u1′′ + φ1u2′′ ) + 2(φ1′u1′ + φ2′u2′ ) + a1 (φ1u1′ + φ2u2′ ) = b( x )


i.e. (φ1u1′′ + φ2u2′′ ) + 2(φ1′u1′ + φ2′u2′ ) + a1 (φ1u1′ + φ 2u2′ ) = b( x)

Differential Equations (37)


Observe that if
φ1u1′ + φ2u 2′ = 0
then (φ1u1′ + φ2u2′ )′ = (φ1′u1′ + φ 2′u2′ ) + (φ1u1′′ + φ 2u2′′ )

and φ1′u1′ + φ2′u2′ = b( x)


Thus if we can find two functions u1(x) and u2(x) such that
φ1u1′ + φ2u2′ = 0

φ1′u1′ + φ 2′u 2′ = b( x )
Then u1φ1 + u2φ2 will satisfy L(y) = b(x).

On solving above two equations for u1′ and u 2′


we get,
–φ 2 b φ1 b
u1′ ( x ) = , u2′ ( x ) = ,
W(φ1 , φ 2 ) W(φ1 , φ 2 )
Integration of above equation between the limits x0 to x provides
x
φ 2 ( t ) b( t )
u1 ( x ) = – ∫ dt + u1 ( x0 )
x0 W(φ1 , φ2 ) (t )
x
φ1 (t ) b(t )
and u2 ( x ) = ∫ dt + u2 ( x0 ).
x0 W(φ1 , φ2 ) (t )

The solution u1φ1 + u2φ2 takes the form

 x φ (t ) b(t ) 
φ ( x ) = φ1 ( x )  – ∫ 2
dt + u1 ( x0 ) 
 x0 W(φ1 , φ2 ) (t ) 
 x φ (t ) b(t ) 
+ φ2 ( x)  + ∫ 1
dt + u2 ( x0 ) 
 x0 W(φ1, φ2 ) (t ) 

The term φ1 ( x ) u1 ( x0 ) + φ 2 ( x ) u2 ( x0 ) is a complementary function or the solution of


corresponding homogeneous equation L(y) = 0 and the particular solution takes the form
x
φ 2 (t ) b(t ) x
φ1 (t ) b(t )
ψ p ( x ) = –φ1 ( x ) ∫ dt + φ2 ( x ) ∫ dt
x0 W(φ1 , φ2 ) (t ) x W(φ1 , φ2 ) (t )
0

x
[φ1 (t ) φ 2 ( x ) – φ 2 (t ) φ1( x)] b(t )
ψ p ( x) = ∫ dt
x0 W(φ1, φ 2 ) (t )
The function yp(x) is a solution of L(y) = b (x).
Theorem 1.4.1 provides a method to find a solution of second order non-homogeneous
differential equation with constant coefficients. The same procedure can be generalized for the
non-homogeneous equation of order n.

Differential Equations (38)


Theorem 1.4.2
Let b be continuous on an interval I and let φ1, φ2 , φ3 ,...., φn be n linearly independent
solutions of L( y ) = y ( n ) + a1 y ( n –1) + a2 y ( n –2) + .... + an y = 0 on I. Every solution y of L(y) =
b(x) can be written as
ψ = ψ p + c1φ1 + c2φ2 + c3φ3 + .... + cnφ n
Where yp is a particular solution of L(y) = b(x) and c1, c2 , c3 ,......, cn are constants. Every
such y is a solution of L(y) = b(x). A particular solution yp is given by
n x
Wk (t )b (t )
ψ p ( x) = ∑ φk ( x) ∫ dt.
k =1 x0 W(φ1 , φ2 , φ3 ,...., φn ) (t )
Proof : The proof is similar to the proof of theorem 1.4.1 Let b be continuous function on an
interval I. Consider the differential equation
L( y ) = y ( n ) + a1 y ( n –1) + a2 y ( n –2) + .... + an y = b( x)
where, a1, a2, a3,..., an are constants. If yp is a particular solution of L(y) = b(x) and y is any
other solution of L(y) = b(x), then
L(ψ – ψ p ) = L(ψ ) – L(ψ p ) = b( x ) – b( x ) = 0
and y – yp is a solution of corresponding homogeneous equation L(y) = 0. (is called subtraction
principle).
Thus any solution y of L(y) = b(x) can be written in the form
ψ = ψ p + c1 φ1 + c2 φ2 + c3 φ3 + .... + cn φn
where, yp is a particular solution of L(y) = b(x), the functions φ1, φ2 , φ3 ,...., φn are n linearly
independent solutions of L(y) = 0 (determined in theorem 1.3.1) and c1, c2 , c3 ,..., cn are constants.
To find a particular solution yp we use the variation of constants method. Suppose
ψ p = u1 ( x ) φ1 ( x ) + u2 ( x ) φ2 ( x ) + u3 ( x ) φ3 ( x) + .... + un ( x) φn ( x)
is a solution of L(y) = b(x). Since yp is a solution it satisfies the equation i.e L(yp) = b(x).
ψ p = u1φ1 + u2φ2 + u3φ3 + .... + unφn
n
= ∑ ui φi
i =1

Then, ψ p′ = u1φ1′ + u1′φ1 + u2φ2′ + u2′φ2 + .... + u nφn′ + un′φn

= (u1 φ1′ + u 2 φ 2′ + u3 φ3′ + .... + un φ n′ ) + (u1′φ1 + u2′φ 2 + .... + un′φ n )


n n
= ∑ ui φi′ + ∑ ui′ φi
i =1 i =1

Let ∑ ui′φi = 0 then ψ p′ = ∑ ui φi′

We have ψ p′′ = ∑ ui φi′′ + ∑ ui′ φi′

Differential Equations (39)


Suppose ∑ ui ′φi ′ = 0 then ψ p′′ = ∑ ui φi′′
Continuing the same assumptions we get,
∑ ui′φi = 0 ; ψ p′ = ∑ ui φi′
∑ ui′φi′ = 0 ; ψ p′′ = ∑ ui φi′′

∑ ui′φi′′ = 0 ; ψ p′′′ = ∑ ui φi′′′


M
∑ ui′φi ( n –2) = 0 ; ψ p ( n –1) = ∑ ui φi ( n –1)

ψ p (n ) = ∑ ui′φi (n –1) + ∑ ui φi (n )

If ∑ ui ′φi ( n –1) = b( x ) then ψ p (n ) = ∑ uiφi (n ) + b ( x) and L(ψ p ) becomes

L(ψ p ) =  ∑ uiφi (n ) + b ( x )  + a1 ∑ uiφi (n –1) + a2 ∑ uiφi (n –2) + .... + an ∑ uiφi .


 
= b ( x) + ∑ ui φi (n ) + a1φi (n –1) + ..... + anφi 
 
Thus, L(ψ p ) = b( x) and therefore yp is a solution of L(y) = b (x). Therefore the problem
is now reduced to solving the system given below for the functions u1, u2, u3,..., un.
∑ ui′φi = 0
∑ ui′φi′ = 0
∑ ui′φi′′ = 0
M
∑ ui′φi (n –2) = 0
∑ ui′φi (n –1) = b( x)
Thus, we have system of equations
 φ1 φ2 φ3 L φn  u ′
   1   0 
 φ1′ φ 2′ φ3′ L φ n′  u2′  0 
   ′   
 φ1′′ φ 2′′ φ3′′ L φ n  u3 =
′′  0 
 M  M   
    M 
′ b( x )
φ1( n –1) φ2( n –1) φ3( n –1) L φn( n –1)  un 

By solving above system of equations by Cramer’s rule we get,


Wk ( x) b( x )
uk′ ( x ) = , k = 1, 2,3,....., n
W (φ1 , φ2 , φ3 ,...., φn ) ( x )
Where Wk ( x) is the determinant obtained from W[φ1, φ 2 , φ3 ,...., φ n ]( x) by replacing the
kth column i.e. [φk φk ′ φk ′′ .....φk ( n –1) ]T by [0 0 0....0 1]T .
If x0 is any point in I, we can integrate u¢k and the functions uk can be written as
Differential Equations (40)
x
W (t ) b(t ) dt
uk ( x) = ∫ W(φ , φ , φ ,...., φ ) (t ) k = 1, 2,3,...., n.
k
x0 1 2 3 n
The particular solution yp now takes the form
x
Wk (t ) b(t ) dt
ψ p ( x ) = ∑ φk ( x ) ∫
x0 W(φ1 , φ2 , φ3 ,...., φn ) (t )
Now we are in a position to find out a solution of the non-homogenous equation of
order n.
Observe that a particular solution yp satisfies

ψ p ( x0 ) = ψ p′ ( x0 ) = ψ p′′ ( x0 ) = ...... = ψ p (n –1) ( x0 ) = 0.

EXAMPLES

Q.1. Compute the solution y of y ′′′ + y ′′ + y ′ + y = 1 which satisfies ψ (0) = 0, ψ ′(0) = 1,


ψ ′′(0) = 0 .
Ans. : The characteristic polynomial of the corresponding homogeneous equation is
p( r ) = r 3 + r 2 + r + 1. The characteristic roots are i, –i, 1. The basic solutions of the
corresponding homogeneous equation are
φ1 ( x) = cos x φ 2 ( x) = sin x φ3 ( x) = e – x
To obtain the particular solution of the form
ψ p = u1 φ1 + u2 φ2 + u3 φ3
We have to find W (f1, f2 , f3 ) (x) and Wk (t) for k = 1, 2, 3.

cos x sin x e– x
W(φ1, φ2 , φ3 ) ( x) = – sin x cos x – e – x
– cos x – sin x e– x

W(φ1, φ2 , φ3 ) ( x) = e – a1 ( x – x0 ) W(φ1, φ2 , φ3 ) ( x0 )

= e – x W(φ1 , φ 2 , φ3 ) (0)

1 0 1
W(φ1, φ2 , φ3 ) (0) = 0 1 –1
–1 0 1
= 1[1 – 0] + 1 [0 + 1] = 2

Thus W(φ1, φ2 , φ3 ) ( x) = 2e – x

Differential Equations (41)


0 sin x e– x
W1 ( x) = 0 cos x – e – x = (–1)3+1[– e – x cos x – e – x sin x] = – e – x (cos x + sin x)
–1 – sin x e– x

cos x 0 e– x
W2 ( x) = – sin x 0 – e – x = e – x (cos x – sin x)
– cos x 1 e– x

cos x sin x 0
W3 ( x) = – sin x cos x 0 = 1
– cos x – sin x 1
x
W1 (t ) b(t ) x
e – t (cos t + sin t )
u1 ( x ) = ∫ dt = ∫ – dt
W(φ1, φ2 , φ3 ) (t ) 2e – t
1
= – [ + sin x – cos x]
2
1
Thus, u1 ( x ) = [cos x – sin x]
2
x x –t
W2 (t ) b(t ) e (cos t – sin t )
u2 ( x ) = ∫ dt = ∫ dt
W(φ1, φ2 , φ3 ) (t ) 2e – t
1
= [+ sin x + cos x ]
2
x x
W3 (t ) b(t ) dt dt 1
u3 ( x) = ∫ =∫ – t = e x
W(φ1, φ2 ,φ3 )(t ) 2e 2
Therefore a particular solution is given by
ψ p = u1 ( x ) φ1 ( x ) + u2 ( x ) φ2 ( x ) + u3 ( x ) φ3 ( x)
1 1 1
= – (cos x – sin x) cos x + (cos x + sin x )sin x + e x e – x = 1
2 2 2
The most general solution is
ψ ( x ) = ψ p + c1 φ1 + c2 φ 2 + c3 φ3

= 1 + c1 cos x + c2 sin x + c3 e – x
ψ (0) = 0 ⇒ 1 + c1 + c3 = 0
ψ ′(0) = 1 ⇒ ψ ′( x ) = – c1 sin x + c2 cos x – c3 e – x
Thus, y¢ (0) = c2 – c3 = 1

ψ ′′( x ) = – c1 cos x – c2 sin x + c3 e – x

Differential Equations (42)


ψ ′′(0) = – c1 + c3 = 0
Solving the system of equations
1 + c1 + c3 = 0
c2 – c3 = 1
– c1 + c3 = 0
1 1 1
We get, c1 = – , c2 = , c3 = –
2 2 2
Therefore the solution of our problem is given by
1
ψ ( x ) = 1 + (sin x – cos x – e – x )
2
Q.2. Find all solutions y of the following equations
(a) y′′′ – y′ = x
(b) y ′′ – 3 y ′ + 2 y = sin e – x
(c) y′′ + 4 y′ + 4 y = 3x e –2 x

Ans. (a) : The characteristic polynimial p (r ) = r 3 – r has roots 0, 1, –1 and the linearly
independent solution of the related homogeneous equation are φ1 ( x) = 1, φ2 ( x) = e x ,
φ3 ( x ) = e – x
Let ψ p = u1 ( x ) φ1 ( x ) + u2 ( x ) φ2 ( x ) + u3 ( x ) φ3 ( x)
x
Wk (t ) b (t ) ds
uk ( x ) = ∫ k = 1, 2,3.
W(φ1, φ2 ,φ3 ) (t )

1 ex e– x
W(φ1, φ2 , φ3 ) ( x) = 0 e x –e – x = 2
0 ex e– x

0 ex e– x
b ( x)W1 ( x ) = 0 e x – e – x = –2 x
x ex e– x

1 0 e– x
b ( x)W2 ( x) = 0 0 – e x = xe – x
0 x e–x

1 ex 0
b ( x )W3 ( x) = 0 e x 0 = xe x
0 ex x

Differential Equations (43)


x
–2t dt x2
u1 ( x) = ∫ =–
2 2
x
te – t 1 x 1
u2 ( x ) = ∫ dt = ∫ t e – t dt = – (1 + x)e – x
2 2 2
x
tet dt 1 x t 1
u3 ( x ) = ∫ = ∫ t e dt = ( x – 1)e x
2 2 2
x2 1 1 x2
ψ p = u1 φ1 + u2 φ2 + u3 φ3 = –
– (1 + x ) + ( x – 1) = – –1 is the
2 2 2 2
required particular integral and the solution ψ = c1 + c2 e x + c3 e – x + ψ p .
(b) : The characteristic polynomial p(r ) = r 2 – 3r + 2 has roots +2, +1 and therefore the two
linearly independent solution of the corresponding homogeneous equation are
φ1 ( x) = e x and φ2 ( x) = e 2 x
Let ψ p = u1 ( x ) φ1 ( x) + u 2 ( x) φ2 ( x) be a particular integral of the given differential equation
then by method of separation of parameters we get,
W ( x) b( x) dx W ( x) b( x) dx
u1 ( x ) = ∫ 1 and u2 ( x ) = ∫ 2
W(φ1, φ2 ) ( x) W(φ1, φ2 ) ( x)

ex e2 x
where, W1 (φ1, φ2 ) ( x) = = e3 x ,
x 2x
e 2e

0 e2x
b ( x)W1 ( x) = = – e 2 x sin e – x ,
–x 2x
sin e 2e

ex 0
b ( x)W2 ( x ) = = e x sin e – x .
x –x
e sin e

Thus, u1 ( x ) = ∫ – e – x sin e – x dx = – cos e – x

and u2 ( x ) = ∫ + e –2 x sin e – x dx = – sin e – x + e – x cos e – x

[ Integrate above equation with the substitution t = e–x].


Then the general solution
ψ = c1 φ1 + c2 φ2 + ψ p

= c1 e x + c2 e 2 x + (– cos e – x )e x + (– sin e – x + e – x cos e – x ) e 2 x


= c1 e x + c2 e 2 x – e 2 x sin e – x .
The characteristic polynomial p(r ) = r + 4r + 4 has roots –2, –2 and therefore the two
2
(c) :
linearly independent solution of the corresponding homogeneous equation are
φ1 ( x ) = e –2 x , φ2 ( x ) = x e –2 x
Differential Equations (44)
Let ψ p = u1 ( x ) φ1 ( x) + u 2 ( x) φ2 ( x) be a particular integral of the given differential equation
then by method of separation of parameters we get,

W1 ( x) b ( x) dx W2 ( x) b( x ) dx
u1 ( x) = ∫ , u2 ( x ) = ∫
W(φ1, φ2 )( x) W(φ1 , φ2 ) ( x)

e –2 x xe –2 x
where, b ( x) = 3 xe –2 x , W1 (φ1, φ2 ) ( x) = = e –4 x ,
–2 x –2 x
–2e (1 – 2 x )e

0 xe –2 x
b ( x)W1 ( x) = = –3 x 2 e –4 x ,
–2 x –2 x
3xe (1 – 2 x )e

–e 2 x 0
b ( x)W2 ( x) = = 3 xe –4 x .
–2 x –2 x
–2e 3 xe

3x 2e –4 x
Thus, u1 ( x) = ∫ – –4 x
dx = – x3 and
e
3x e –4 x 3 2
u2 ( x ) = ∫ + –4 x
dx = x
e 2
3 3 –2 x
Therefore ψ p = u1 φ1 + u2 φ 2 = – x 3e –2 x + xe
2
1 3 –2 x
= x e
2
The general solution
1 3 –2 x
ψ = c1e –2 x + c2 xe –2 x + xe
2
Q.3. Find the general solution of
π π
y ′′ + y = tan x , – < x< .
2 2
Ans : The characteristic polynimial p( r ) = r + 1 has roots +i, –i and the two linearly independent
2

solutions are
φ1 ( x ) = cos x and φ2 ( x ) = sin x

Let ψ p ( x) = u1 ( x) φ1 ( x) + u2 ( x) φ2 ( x) be a particular integral of the given differential equation


then by method of separation of parameters we get,
W1 ( x ) b ( x ) dx W2 ( x ) b ( x ) dx
u1 ( x ) = ∫ , u2 ( x ) = ∫
W(φ1, φ2 ) ( x) W(φ1 , φ2 ) ( x )
where, b (x) = tan x

Differential Equations (45)


cos x sin x 0 sin x sin 2 x
W(φ1 , φ2 ) ( x) = = 1, b( x)W1 ( x) = =– ,
– sin x cos x tan x cos x cos x
cos x 0
b ( x)W2 ( x) = = sin x.
– sin x tan x
sin 2 x
Therefore u1 ( x) = ∫ – dx = – log (sec x + tan x) + sin x
cos x

u2 ( x ) = ∫ sin x d x = – cos x

π
and ψ p = – cos x log (sec x + tan x ), – < x<π2
2
The general solution

y ( x) = c1 cos x + c2 sin x – cos x log(sec x + tan x), – π 2 < x < π 2

Note. The formula for a particular a solution yp of L(y) = b(x) makes sense for some
discontinuous functions b (x). Then yp will be a solution of L(y) = b(x) at the continuity
points of b .
Q 4. Find a particular solution of the equation.
y ′′ + y = b ( x ),
Where, b( x ) = –1 (– Π ≤ x < 0) ,
=1 (0 ≤ x ≤ Π ) ,
=0 (1x1 > Π ).
Ans : Let us find out the particular solution of y ′′ + y = α where a is a constant.
The characteristic polynomial is p( r ) = r 2 + 1 and has roots +i, –i. Therefore the basic
solutions (linearly independent solutions) are
φ1 ( x) = cos x, φ2 ( x) = sin x
Let ψ p = u1 ( x) φ1 ( x) + u2 φ2 ( x) be a particular solution of the equation y ′′ + y = α . By
method of separation of parameters we get,
W1 ( x) b ( x) dx W2 ( x) b( x ) dx
u1 ( x) = ∫ , u2 ( x ) = ∫
W(φ1, φ2 )( x) W(φ1 , φ2 ) ( x)
cos x sin x
where, W(φ1, φ2 ) ( x) = =1
– sin x cos x
0 sin x cos x 0
W1 ( x ) = = – sin x, W2 ( x ) = = cos x
1 cos x – sin x 1

W1 ( x) b( x) dx –α sin x
Then, u1 ( x ) = ∫ =∫ dx = α cos x
W(φ1 , φ 2 ) ( x) 1

Differential Equations (46)


W2 ( x) b( x) α cos x
u2 ( x ) = ∫ dx = ∫ dx = α sin x
W(φ1, φ2 )( x) 1
The particular solution
ψ p = u1φ1 + u2 φ2 = α cos 2 x + α sin 2 x = α
Thus the general solution of y ′′ + y = α is
ψ = c1 cos x + c2 sin x + α
The general solution on the real line becomes
ψ ( x ) = c1 cos x + c2 sin x ; – ∞ < x < –π
= c3 cos x + c4 sin x –1 ; –π ≤ x < 0
= c5 cos x + c6 sin x + 1 ; 0 ≤ x ≤π
= c7 cos x + c8 sin x ; π <x<∞
The continuity of y at x = –π , 0, π gives – c1 = – c3 –1, c3 –1 = c5 + 1, – c5 + 1 = – c7
Since we have three equations in 4 unknown, the particular solution yp will not be unique
e.g. choose c3 = c1* + 1 and c2 = c4 = c6 = c8 = c
Then c1* cos x + c sin x is a complementary function or the solution of corresponding
homogenous equation y ′′ + y = 0 and particular equation will be determined as follows.

If c3 = c1* + 1 then c1 = c3 + 1 = c1* + 2

c3 = c1* + 1
c5 = c3 – 2 = c1* + 1 – 2 = c1* –1
c7 = c5 –1 = c1* –1–1 = c1* – 2
Thus, the particular solution becomes
ψ ( x ) = 2 cos x –∞ < x < π
= cos x –1 –π ≤ x < 0
= – cos x + 1 0 ≤ x ≤π
= –2 cos x π <x<∞
If we choose c3 = c1* + 2 then c1 = c3 + 1 = c1* + 3,
c3 = c1* + 2,
c5 = c3 – 2 = c1* ,
c7 = c5 –1 = c1* –1,
and the particular solution becomes
ψ ( x ) = 3cos x ; – ∞ < x < –π
= 2cos x –1 ; –π ≤ x < 0
=1 ; 0 ≤ x ≤π
= – cos x ; π < x < ∞.
Differential Equations (47)
Thus, we can generate infinitely many particular solutions that are piecewise continuous.
Method of undetermined coefficients :
The method described sofar is called the method of variation of parameters. Although this
method yields a solution of the non-homogeneous equation it sometimes require more labor
than necessary. We now explain a method which is often faster than a method of variation of
parameters. This method is useful to solve the non-homogeneous equation L (y) = b (x), when
b(x) is a solution of some homogeneous equation with constant coefficients. The procedure we
are about to describe is called the method of undetermined coefficients.
For the given different equation L (y) = b (x), suppose b(x) is a solution of some
homogeneous equation M (y) = 0 with constant coefficients. Then M (b(x)) = 0. If y is a solution
of L (y) = b (x) and M (b) = 0 then
M [ L(ψ )] = M (b) = 0.
Therefore y is a solution of the homogeneous equation M (L(y)) = 0 with constant
coefficients. If the order of differential operator Lis n and that of M is m then M(L(y)) = 0 is a
homogeneous differential equation of order m + n and therefore there are m + n linearly
independent solutions of M (L(x)) = 0. Since b(x) is a particular solution of M (y) = 0 every
linear combination of these n + m linearly independent solution will not be a solution of
L (y) = b (x). Thus, to fine the solution of L (y) = b (x) we substitute the linear combination of
solutions into L (y) = b (x) and determine the set of coefficients other than the coefficients of the
solutions corresponding to the homogeneous equation L (y) = 0.
We give an example to show the usefulness of this method. Suppose we consider
L( y) = y′′ – 3 y′ + 2 y = x 2
Since ( x 2 )′′′ = 0, x2 is a solution of M ( y ) = y′′′ = 0.
Every solution y of L (y) = x2 is a solution of
M ( L( y)) = M ( y′′′ – 3 y′ + 2 y) = M ( x 2 ) = 0.
But M ( y′′′ – 3 y ′ + 2 y ) = ( y′′′ – 3 y ′ + 2 y )′′′ = 0
i.e. y (v) – 3 y (iv) + 2 y′′′ = 0.
The characteristic polynomial of this equation is p(r ) = r – 3 y + 2r (just the product
5 4 3

of characteristics polynomials of L and M ). The roots of p (r) are 0, 0, 0, 1, 2 and hence y must
have the form ψ = c0 + c1 x + c2 x 2 + c3e x + c4e 2 x observe that c3e x + c4e 2 x is a solution
of L (y) = 0.

Since we are interested only in particular solution y p of L( y ) = x 2 , we can assume

ψ p = c0 + c1 x + c2 x 2
Since y p is a solution, it should satisfy the differential equation L (y) = x2.
L(ψ p ) = ψ p′′ – 3ψ p′ + 2ψ p = 2 c0 + c1x + c2 x 2  – 3[c1 + 2c2 x] + [2c2 ]

L(ψ p ) = x 2 gives 2c2 x 2 + (2c1 – 6c2 ) x + (2c0 – 3c1 + 2c2 ) = x 2

Differential Equations (48)


Since the above equation should hold for all values of x, on equating the coefficients of
equal powers of x we get ,
2c2 = 1, 2c1 – 6c2 = 0, 2c0 – 3c1 + 2c2 = 0.
By solving these equations simultaneously we get,
1 3 7
c2 = , c1 = , c0 = .
2 2 4
7 3 1 2
Therefore, ψ p = + x + x
4 2 2
1
= (2 x 2 + 6 x + 7)
4
is a particular solution of L ( y ) = x 2
and ψ = ψ p + c3e x + c4e 2 x
1
ψ = (2 x 2 + 6 x + 7) + c3e x + c4 e 2 x is a general solution of y′′ – 3 y′ +
4
2 y = x2.
This method is also called as annihilator method since to solve L (y) = b (x), we find the
operator M which annihilates b(x). i.e. M (b (x)) = 0.
Once M has been found the problem becomes algebraic in nature.

EXAMPLES

Exp. 1. Using the annihilator method find a particular solution of each of the following
equations.
(a) y ′′ + 4 y = cos x (b) y′′ – 4 y = 3e2 x + 4e – x

(c) y ′′ – y ′ – 2 y = x 2 + cos x

Ans. (a) : cos x is a solution of y ′′ + y = 0 therefore


M ( y ) = y ′′ + y

L( y ) = y ′′ + 4 y therefore M ( L( y )) = [L( y )]′′ + [ L( y ) ]


M [L( y )] = ( y′′ + 4 y )′′ + y′′ + 4 y

= y (iv ) + 5 y′′ + 4 y

Thus, M [ L( y )] = 0 implies y (iv) + 5 y ′′ + 4 y = 0.


The characteristic polynomial of the above equation is
p(r ) = r 4 + 5r 2 + 4 = (r 2 + 4) (r 2 + 1)
The root of p(r) are i, –i, +2i, – 2i and hence the solutions y have the form

Differential Equations (49)


ψ = c1 cos x + c2 sin x + c3 cos 2 x + c4 sin 2 x
observe that c3 cos 2 x + c4 sin 2 x is a solution of L( y ) = y ′′ + 4 y = 0. Since we are interested only
in particular solution yp of L (y) = cos x, we can assume
ψ p ( x) = c1 cos x + c2 sin x

Since yp is a solution it should satisfy the differential equation y ′′ + 4 y = cos x.


ψ p′′ + 4ψ p = – c1 cos x – c2 sin x + 4(c1 cos x + c2 sin x) = cos x
On equation the coefficients of cos x and sin x we get
1
3c1 = 1, 3c2 = 0 i.e. c1 = and c2 = 0
3
1
Thus, particular solution ψ p = cos x
3

Ans. (b) : 3e 2 x + 4e – x is a solution of (D – 2) (D – 1) = 0 i.e. y ′′ – y ′ – 2 y = 0.


Thus, M ( y ) = y ′′ – y ′ – 2 y . since L( y ) = y ′′ – 4 y , M ( L( y )) = y (iv ) – 4 y ′′′
– 6 y′′ + 4 y′ + 8 y .
The differential equation
y (iv) – 4 y′′′ – 6 y′′ + 4 y′ + 8 y = 0
has a characteristic polynomial p( r ) = r 4 – 4r 3 – 6 r 2 + 4r + 8. The roots of characteristic
polynomial are 2, 2, –1, –2.
The solution y has the form
ψ ( x ) = c1e 2 x + c2 xe 2 x + c3e – x + c4e –2 x
Observe that c1 e 2 x + c4 e –2 x is a solution of the homogeneous equation L(y) = 0.
Since we are only interested in particular solution assume the solution
ψ p = c1 e 2 x + c2 x e 2 x + c3e – x
Since yp is a particular solution is should satisfy the equation y′′ – 4 y = 3e2 x + 4e – x

ψ p′ = [2c1 + (1 + 2 x) c2 ]e 2 x – c3e – x

ψ p′′ = [4c1 + (4 + 4 x) c2 ] e 2 x + c3e – x

ψ p′′ – 4ψ p = [4c1 + (4 + 4 x) c2 – 4c1 – 4 xc2 ] e 2 x – 3c3 e – x

But yp satisfies y ′′ – 4 y = 3e 2 x + 4e – x

Therefore [4c2 ]e 2 x – 3c3e – x = 3e 2 x + 4e – x


By comparing coefficient of e2x and e–x we get 4c2 = 3 and –3c3 = 4

Differential Equations (50)


3 2x 4 – x
Thus, ψ = c1e 2 x + x e – e + c4 e –2 x and particular integral
4 3
3 4
ψ p = x e2 x – e – x .
4 3

Ans. (c) : x 2 + cos x is a solution of D3 ( D 2 + 1) y = 0 i.e. y (v) + y′′′ = 0. Thus M ( y) = y (v ) + y′′′


L( y ) = y′′ – y′ – 2 y

Therefore M [ L( y )] = [ y ′′ – y ′ – 2 y ](v) + [ y ′′ – y ′ – 2 y ]′′′

= y (7) – y (6) – y (5) – y (4) – 2 y′′′ = 0.


The differential equation M [ L( y )] = 0 has a characteristic polynomial p(r ) = r 7 – r 6
– r 5 – r 4 – 2r 3 .
1+ 5 1– 5
The roots of characteristic polynomial are 0, 0, 0, i, – i, and .
2 2
The solution y must have the form
 1+ 5   1– 5 
  x   x
 2   2 
ψ ( x ) = c0 + c1x + c2 x 2 + c3 cos x + c4 sin x + c5e + c6e
 1+ 5   1– 5 
  x   x
 2   2 
The expression c5 e + c6 e is a solution of the homogeneous equation
L( y ) = y′′ – y ′ – 2 y = 0.
Since we are interested in particular solution assume the solution
ψ p = c0 + c1 x + c2 x 2 + c3 cos x + c4 sin x

The problem is to determine the constants c0, c1, c2, c3, c4 so that L(ψ p ) = x 2 + cos x.

ψ p′ = c1 + 2c2 x – c3 sin x + c4 cos x

ψ p′′ = 2c2 – c3 cos x – c4 sin x

L(ψ p ) = ψ p′′ – ψ p′ – 2ψ p

= (2c2 – c1 – 2c0 ) – (2c1 + 2c2 ) x – 2c2 x 2 – (3c3 + c4 ) cos x


+ (c3 – 3c4 ) sin x
= x 2 + cos x
Thus, 2c2 – c1 – 2c0 = 0, 2c1 + 2c2 = 0, – 2c2 = 1, 3c3 + c4 = –1 and c3 – 3c4 = 0
Simulataneous evaluation of above equation gives
1 1 3 1 3
c2 = – , c1 = , c0 = – , c3 = 3c4 , c4 = – , c3 = –
2 2 4 10 10
3 1 1 3 1
Therefore ψ p = – + x – x 2 – cos x – sin x
4 2 2 10 10
Differential Equations (51)
EXERCISE

Exp. 1. Use the method of variation of parameters and find the general solution of each of
the following equation.

(a) y′′ – y = sin 2 x (b) y ′′ + y = 4 x sin x

(c) y ′′ + 3 y ′ + 2 y = 12e x (d) y ′′ + 2 y ′ + y = x 2e – x

(e) y ′′ + 4 y = cos x (f) y ′′ + 9 y = sin 3x

(g) y ′′ – 7 y ′ + 6 y = sin x (h) 4 y′′ – y = e x

(i) 6 y ′′ + 5 y ′ – 6 y = x

Exp. 2. Find the particular solution of each of the following equation using the method of
undetermined coefficients.

(a) y′′ + 4 y′ + 4 y = 4 x 2 + 6e x (b) y′′ – 3 y′ + 2 y = 2 xe3 x + 3sin x

(c) y ′′ + 4 y ′ + 4 y = 3xe –2 x (d) y ′′ – 3 y ′ + 2 y = 6e – x

–x 1 2
Ans.(1): (a) y = c1e + c2e – sin x –
x 2
(b) c1 cos x + c2 sin x – x 2 cos x + x sin x
5 5
x 4e – x
(c) c1e –2 x + c2c – x + 2e x (d) c1e – x + c2 xe – x +
12
1 1
(e) c1 cos 2 x + c2 sin 2 x + cos x (f) c1 cos 3x + c2 sin 3 x – x cos 3 x
3 6
x x
1 – 1
(g) c1e
6x
+ c2e + (7 cos x + 5sin x )
x
(b) c1e 2 + c2e 2 + ex
74 3
2x 3x
– 1 5
(i) c1e 3 + c2 e 2 – x– .
6 36

3 2 x 3 3x 3 9
Ans.(2): (a) x – 2 x + + e (b) xe – e + sin x + cos x
2 3x
2 3 2 10 10
1 3 –2 x
(c) x e (e) e – x
2

S
Differential Equations (52)
Chapter
2
Linear Equations with
Variable Coefficients

Contents :
Unit 1 : Homogenous equations with variable coefficients.
(a) Initial value problems for the homogeneous equation.
(b) Solution’s of homogenous equation
(c) Reduction of an order of a homogeneous equation
Unit 2 : Basis
(a) Linear independence and Wronskian
(b) Solution of non-homogeneous equations
Unit 3 : Homogenous equations with analytic coefficients.

Introduction
Solutions to linear equations with variables coefficients are necessary to analysis most of
the situations in science and technology. In the last chapter we have studied linear equations
with constant coefficients. In this chapter we are going to study linear equations with variable
coefficients. There is no standard procedure to find all possible solutions of a given equation.
However it is possible to construct series solution if the coefficient functions and the control
function are analytic on some open set.

Unit 1 : Homogeneous equations with variable coefficients.


A linear differential equation of order n with variable coefficients is an equation of the
form a0 ( x) y ( n ) + a1 ( x) y ( n –1) + a2 ( x) y ( n –2) + ..... + an ( x) y = b( x), where a0 , a1, a2 , ...., an , b
are complex valued functions defined on some interval I ⊂ R. Points where a0 (x) = 0 for x in
I are called singular points. In this chapter we assume that a0 (x) ¹ 0 on I. Since a0 is non-zero
we can divide the equation by a0 and rename functions ai (x) / a0 (x) by new ai(x) and b (x) /
a0 (x) as new b(x). Then above equation can be written as
y ( n) + a1 y ( n–1) + a2 y ( n –2) + ..... + an y = b ( x )
In this chapter we denote the left hand side of the above equation by an operator L. Thus,
L( y ) = y (n ) + a1 ( x) y (n –1) + a2 ( x ) y (n –2) + .... + an ( x ) y and the equation becomes L(y) = b (x).
If b(x) = 0 for all x in I we call equation L(y) = 0 a, homogeneous equation whereas if

Differential Equations (53)


b(x) ¹ 0 for some x in I, the equation is called a non-homogeneous equation.
A function f is a solution of L(y) = 0 on I if f is n times differentiable and satisfies
L(f ) = 0 for all x in I.
Most of the results we developed in chapter I are valid in more general case we are now
considering. The major difficulty with linear equations with variable coefficients, from a practical
point of view, is that there are very few types of equations whose solutions can be expressed in
terms of elementary functions and for which standard method of obtaining them, if they do
exist, are available. However, in case a1, a2, a3,..... an have convergent power series expansions
the solutions will have this property also and the series solutions can be obtained by a simple
formal procedure. But there is no analogue of the theorem 1.3.1 of chapter I, which gives a
procedure to find all possible solutions of given equation.

A. Initial value problems for the homogeneous equation


Although in many cases it is not possible to find the solution, we can prove that if the
functions ai ( x), i = 1, 2, 3,..... n are continous functions then there is a solution to L(y) = 0.
Moreover if we know the initial values of the solution and its derivatives then the solution is
unique.

Theorem 2.1.1 :
Let b1, b2 , b3 , ...., bn be non-negative constants such that for all x in I
| ai ( x) | ≤ bi i = 1, 2, 3, ...., n and define k by
k = 1 + b1 + b2 + b3 + .... + bn .
If x0 is a point in I and f is a solution of L(y) = 0 on I then
|| φ ( x0 ) || e – k |x – x0 | ≤ || φ ( x) || ≤ || φ ( x0 ) || e k |x – x0 | for all x in I.
Proof : The proof of this theorem is similar to the proof of theorem 1.3.3.
Let u ( x) = || φ ( x) ||2 =| φ |2 + | φ ′( x) |2 + | φ ′′ |2 +....+ | φ (n –1) |2
= φ φ + φ ′φ ′ + φ ′′ φ ′′ + .... + φ ( n)φ ( n–1) + φ ( n –1)φ ( n –1)
Hence u ′( x ) = φ ′ φ + φ φ ′ + φ ′′φ ′ + φ ′ φ ′′ + .... + φ ( n)φ ( n –1) + φ ( n –1)φ ( n)

Therefore | u ′( x) | ≤ 2 | φ || φ ′ | +2 | φ ′ || φ ′′ | +2 | φ ′′ || φ ′′′ | +.... + 2 | φ || φ n |


( n –1)

(for any complex variable z, | z | =| z | )


Since f is solution of L(y) = 0,
φ (n ) = – a1 ( x) φ (n –1) – a2 ( x ) φ (n –2) – a3 φ (n –3) – ..... – anφ
| φ ( n) ( x) | ≤ | a1 ( x) || φ ( n –1) ( x) | + | a2 ( x) || φ ( n –2) | + a3 ( x ) | φ ( n –3) | +....+ | an || φ |
For all x in I, | ai ( x) | ≤ bi , i = 1, 2, 3,....., n and therefore
| φ ( n ) ( x ) | ≤ b1 | φ ( n –1) | + b2 | φ ( n –2) | + b3 | φ ( n –3) | +.... + bn | φ |
and | u ′ | ≤ 2 | φ || φ ′ | +2 | φ ′ || φ ′′ | +2 | φ ′′ || φ ′′′ | +.....

Differential Equations (54)


+2 | φ (n –2) || φ (n –1) | +2 | φ (n –1) | b1 | φ (n –1) | + b2 |φ (n –2) | +.... + bn | φ |
 
The rest of the proof is on the same lines as that of theorem 1.3.3.
 (| a | – | b |) 2 ≥ 0 ⇒ 2 | a || b | ≤ | a |2 + | b |2 
 
| u′( x) | ≤ (| φ |2 + | φ ′ |2 ) + (| φ ′ |2 + | φ ′′ |2 ) + ..... + (| φ (n –2) |2 + | φ (n –1) |2 )
+ b1 (| φ ( n –1) |2 + | φ ( n –1) |2 ) + b2 (| φ ( n –1) |2 + | φ ( n –2) |2 ) + ....
+ bn (| φ (n –1) |2 + | φ |2 )
≤ (1 + bn ) | φ |2 + (2 + bn –1 ) | φ ′ |2 +.... + (2 + b2 ) | φ ( n –2) |2
+ (1 + 2b1 + b2 + b3 + .... + bn )| φ ( n –1) |2
Since each coefficient on the right hand side is less than 2 k we have
| u ′( x ) | ≤ 2k u ( x )
Consider the right inequality which can be written as
u ′( x ) – 2k u ( x) ≤ 0.
Integrate above inequality from x0 to x with x > x0.
e –2kxu ( x) – e –2k x0 u ( x0 ) ≤ 0
or u ( x ) ≤ e 2 k ( x – x0 ) u ( x0 )
|| φ ( x) ||2 ≤ e 2 k ( x – x0 ) || φ ( x0 ) ||2
i.e. || φ ( x) || ≤ e k ( x – x0 ) || φ ( x0 ) ||
Similarly –2k u ( x ) ≤ u ′( x ) gives
| φ ( x0 ) || e – k ( x – x0 ) ≤ || φ ( x) || , ( x > x0 )
and therefore
|| φ ( x0 ) || e – k ( x – x0 ) ≤ || φ ( x) || ≤ e k ( x – x0 ) || φ ( x0 ) ||, ( x > x0 )
If x < x0 repeat the same procedure and integrate the inequality from x to x0. We get
|| φ ( x0 ) || e – k ( x – x0 ) ≤ || φ ( x) || ≤ || φ ( x0 ) || e – k ( x – x0 ) ( x < x0 )
which is the required inequality for x < x0.
Observe that if interval I is closed and bounded interval and if ai(x) are continuous functions
on I then these functions are bounded. [continuous function on closed and bounded intervals is
bounded and the function attains it bounds ]. Since aj (x) are bounded functions on I, there
always exist finite constants bj such that | a j ( x) | ≤ b j for j = 1, 2, 3, ...., n.

Theorem 2.2.1 : (Uniqueness theorem)


Let x0 be in I and let α1 , α 2 , α3 ,...., α n be any n constants. There is at most one solution f
of L(y) = 0 on I satisfying
φ ( x0 ) = α1, φ ′ ( x0 ) = α 2 , φ ′′ ( x0 ) = α3 ,....., φ (n –1) ( x0 ) = α n .
Differential Equations (55)
Proof : Let x be any point in I other than x0. Let J be closed and bounded interval in I containing
x0 and x. On the interval J continuous functions aj (x) are bounded, that is,
| a j ( x) | ≤ b j ( j = 1, 2, 3,...., n),
for some constants bj (These constants bj may depend on the choice of J ⊂ I ).
Suppose f and y are two solutions of L(y) = 0 on J satisfying the given initial conditions
i.e. φ ( x0 ) = ψ ( x0 ) = α1, φ ′( x0 ) = ψ ′( x0 ) = α 2 ,......, φ ( n –1) ( x0 ) = ψ ( n –1) ( x0 ) = α n . Define
q = f – y in J. Since f and y satisfy L(y) = 0. θ ( x0 ) = φ ( x0 ) – ψ ( x0 ) = 0, and L(φ ) = L(ψ ) = 0
therefore by linearity L(q) = 0. q (x0) = f (x0) – y (x0) = 0similarly θ ′( x0 ) = θ ′′( x0 ) = .....

= θ (n –1) ( x0 ) = 0. but || θ ( x0 ) ||2 =| θ ( x0 ) |2 + | θ ′( x0 ) |2 + | θ ′′( x0 ) |2 + ....+ | θ (n –1) ( x0 ) |2 = 0.


Applying theorem 2.1.1 we obtain || θ ( x) || = 0 for all x in J. In particular q (x) = 0 for all x in J
⊂ I. But x is any point in I and therefore q (x) = 0 for every x in I. This proves that φ ( x) = θ ( x)
for every x in I.
Here we state existence theorem without proof.

Theorem 2.1.3 : (Existence Theorem)


Let a1 ( x ), a2 ( x), a3 ( x),....., an ( x) be continuous functions on an interval I containing the
point x0. If α1, α 2 ,...., α n are any n constants, there exists a solution f of

L( y ) = y (n ) + a1 ( x) y (n –1) + a2 ( x ) y (n –2) + ..... + an ( x) y = 0 on I satisfying

φ ( x0 ) = α1, φ ′ ( x0 ) = α 2 , φ ′′ ( x0 ) = α 3 ,......, φ ( n –1) ( x0 ) = α n .

(B) Solutions of homogeneous equation


Superposition principle :
If φ1 , φ2 , φ3 ,...., φm are any m solutions of the L(y) = 0 on an interval I and c1, c2,c3,..., cm
are any m constants then c1φ1 + c2 φ2 + c3φ3 + ..... + cmφm is also a solution of L(y) = 0.
The trivial solution is a function which is identically zero on I.

Theorem 2.1.4
There exist n linearly independent solutions (definition 1.3) of L(y) = 0 on I.
Proof : Let x0 be a point in I. According to theorem 2.1.3 and theorem 2.1.2, there is a unique
solution of L(y) = 0 satisfying given initial conditions at x0.
Let f 1 be a solution of L(y) = 0 satisfying
φ1 ( x0 ) = 0, φ1′ ( x0 ) = 0, φ1′′ ( x0 ) = 0,......, φ1( n –1) ( x0 ) = 0
Let f 2 be a solution of L(y) = 0 satisfying
φ2 ( x0 ) = 0, φ2′ ( x0 ) = 1, φ2′′ ( x0 ) = 0,......, φ2 ( n –1) ( x0 ) = 0
In general Let f i be a solution of L(y) = 0 with

Differential Equations (56)


φi (i –1) ( x0 ) = 1, and φi ( x0 ) = φi′ ( x0 ) = ...... = φi (i –2) ( x0 ) = φi (i ) ( x0 ) = ......φi ( n –1) ( x0 ) = 0
i.e. φi ( i –1) ( x0 ) = 1, and φi ( k ) ( x0 ) = 0, k = 1, 2, 3,....., n –1, k ≠ i –1.

We will prove that these solutions φ1, φ2 , φ3 ,...., φn are linearly independent on I. Suppose
there are constants c1, c2, c3,..... cn such that
c1 φ1 ( x ) + c2 φ 2 ( x) + c3 φ3 ( x) + ..... + cn φ n ( x ) = 0 for all x in I.
Differentiating above equation (n – 1) times we get,
c1 φ1′ ( x) + c2 φ 2′ ( x) + c3 φ3′ ( x) + ..... + cn φ n′ ( x) = 0

c1 φ1′′ ( x ) + c2 φ 2′′ ( x) + c3 φ3′′ ( x) + ..... + cn φ n′′ ( x ) = 0


M M M M
c1 φ1( n –1) ( x ) + c2 φ 2 ( n –1) ( x) + c3 φ3( n –1) ( x) + ..... + cn φ n( n –1) ( x) = 0
Above equations hold for all values of x in I.
In particular these equations are true for x = x0.

Since φi ( j –1) ( x0 ) = 0 for j = 1, 2, 3,....., n, j ≠ i and φi(i –1) ( x0 ) = 1 for j = i we get

c1.1 + c2.0 + c3.0 + .... + cn .0 = 0


c1.0 + c2.1 + c3.0 + .... + cn .0 = 0
In general
c1.0 + c2.0 + .... + ci –1.0 + ci .1 + ci +1.0 + ..... + cn .0 = 0
Thus, ci = 0 for i = 0, 1, 2, 3,...., n and therefore solutions φ1, φ2 , φ3 ,...., φn are linearly
independent.
(C) Reduction of order of a homogeneous equation

Suppose we have found one solution of the equation L( y ) = y ′′ + a1 ( x) y′ + a2 ( x) y = 0


then by using the variation of constants method we can reduce L(y) = 0 into a linear differential
equation of order one and obtain the second solution of the differential equation.

Theorem 2.1.5
If f1(x) is a solution of L( y ) = y ′′ + a1 ( x) y′ + a2 ( x) y = 0 on an interval I and f1(x) ¹ 0 on
I, the second solution f2 (x) is given by
x
1  s 
φ2 ( x) = φ1 ( x) ∫ exp  – ∫ 1
a ( t ) dt  ds.
x0 [ φ1 ( s )]
2
 x0 
The functions f1 and f2 are linearly independent.
Proof : Since f1 is a solution of L(y) = 0, L(f1) = 0.
Let φ2 ( x ) = u ( x ) φ1 ( x ) be second solution of L(y) = 0.
L(φ 2 ) = L(u φ1 ) = (u φ1 )′′ + a1 ( x) (u φ1 )′ + a2 ( x) (u φ1) = 0.
i.e. u ′′( x ) φ1 ( x ) + 2u ′( x ) φ1′ ( x ) + u ( x ) φ1′′ ( x )
Differential Equations (57)
+ a1 ( x ) u′ ( x) φ1 ( x ) + u ( x) φ1′ ( x)  + a2 ( x ) (u ( x )φ1 ( x)) = 0.
 
Since L(φ1 ) = φ1′′ + a1 ( x ) φ1′ + a2 ( x )φ1 = 0,

u′′( x) φ1 ( x) + 2u′( x) φ1′ ( x) + a1 ( x) [u′( x)φ1 ( x)] = 0

Thus, φ1 ( x)u ′′( x) +  2φ1′ ( x) + a1 ( x)φ1 ( x)  u′( x) = 0.


 
If v = u¢ then above equation is linear equation of order one and can always be solved
explicitly provided φ1 ( x) ≠ 0 on I.

φ1 ( x )v′( x ) +  2φ1′ ( x) + a1 ( x ) φ1 ( x )  v ( x ) = 0
 

v′( x)  2φ1′ ( x) 
+ + a1 ( x)  = 0
v ( x )  φ1 ( x ) 
On integrating above equation between the limits x0 to x, we get
x  2φ ′ (t ) 
log v ( x) – log v ( x0 ) + ∫  φ (t ) + a1 (t )  dt = 0
1
x0 
 1 
x
log v( x) – log v ( x0 ) + 2 [log φ1 ( x ) – log φ1 ( x0 )] + ∫ a1 (t ) dt = 0.
x0

v ( x )φ12 ( x ) x
log = – ∫ a1 (t ) dt
v ( x0 )φ12 ( x0 ) x0
x

v( x)φ12 ( x)
– ∫ a1 ( t ) dt  x 
=e x0
= exp  – ∫ a1 (t ) dt 
v( x0 )φ12 ( x0 )  x0 

v ( x0 )φ12 ( x0 )  x 
i.e. v( x) = exp  – ∫ a1 (t ) dt 
φ12 ( x)  x0 

But v ( x0 ) φ12 ( x0 ) are the values of v ( x ) φ12 ( x ) evaluated at point x0 and therefore is constant

Let c = v( x0 ) φ12 ( x0 ), then

c  x 
v ( x) = exp  – ∫ 1
a (t ) dt 
φ12 ( x )  x0 
But v ( x ) = u′( x ) and therefore
x
c  s 
u ( x) = ∫ 2 exp  – ∫ a1 (t ) dt  ds.
φ1 ( s )  x0 
Since, φ2 ( x ) = u ( x ) φ1 ( x ) we get the required result.
We can generalize above theorem for linear differential equation
Differential Equations (58)
L( y ) = y (n ) + a1 ( x ) y (n –1) + a2 ( x ) y (n –2) + ..... + an ( x) y = 0.

Theorem 2.1.6 :
Let φ1 be a solution of L(y) = 0 on an interval I and suppose φ1 ( x) ≠ 0 on I. Then we can
reduce the order of equation L(y) = 0 by one. If v2, v3,...., vn are linearly independent solutions of
the reduced differential equation of order n – 1 and if vk = u′k , k = 1, 2, 3,..., n, then
φ1, u1φ1, u2 φ 2 ,...., un φ n are linearly independent solutions of L(y) = 0 on I.
Proof : Let φ1 be solution of L(y) = 0 on I. we try to find a solution f of L(y) = 0 of the form
φ = u ( x ) φ1 ( x ), where u(x) is n times differentiable function defined on an interval I. If
φ ( x ) = u ( x) φ1 ( x) is a solution of L(y) = 0 then L(u ( x ) φ1 ( x )) = 0 .

L(u φ1 ) = (u φ1 )(n ) + a1 ( x) (u φ1 )(n –1) + a2 ( x) (u φ1) (n –2) + ..... + an (u φ1 ) = 0

= u ( n )φ1 + n u ( n –1)φ1′ + ..... + u φ1( n)

+ a1 ( x) u ( n –1)φ1 + ( n – 1) u ( n –2)φ1′ + .... + uφ1( n –1) 


 
+ a2 ( x) u ( n –2)φ1 + (n – 2) u ( n –3)φ1′ + .... + u φ1( n –2) 
 
+......
+ an –1 u′φ1 + u1 φ1′  + an ⋅ u φ1 = 0
 
The coefficient of u in the above expression is φ1( n ) + a1 ( x ) φ1( n –1) + a2 ( x) φ2( n–1) +
.... + φ1 = L(φ1 ) = 0. Therefore the right hand side of above equation consists of u ′, u ′′, u ′′′,...., u (n )
Therefore if we substitute v = u¢ then the above equation becomes a linear homogeneous equation
of order n – 1 in v.

φ1u ( n ) +  n φ1′ + a1 ( x ) φ1  u ( n –1) + .... +  n φ1( n –1) + (n –1) a1 ( x) φ1( n –2) + .... + an –1 φ1  u ′ = 0
   
Since v( x) = u ′( x) we get

φ1v ( n –1) +  n φ1′ + a1 ( x ) φ1  v ( n –2) + .... +  n φ1( n –1) + (n –1) a1 ( x ) φ1( n –2) + .... + an –1 φ1  v = 0
   
Since, φ1 ( x) ≠ 0 on I we can divide above equation byf1. Thus, we can reduce the order of
differential equation by one. Suppose v2, v3, v4,..., vn are linearly independent solutions of the
differential equation in v of order n – 1. Then

φ1vk ( n –1) +  nφ1′ + a1 φ1  vk ( n –2) + .... +  n φ1( n –1) + ( n – 1) a1 φ1( n –2) + .... + an –1φ1  vk = 0
   
But then vk ( x ) = uk ′ ( x ) for k = 2,3, 4,...., n
x
and uk ( x ) = ∫ vk (t ) dt k = 2, 3, 4,...., n
x0

Differential Equations (59)


But then by assumption uk ( x) φ1 ( x) is a solution of L(y) = 0. Thus the functions
φ1, u2 φ1, u3 φ1,....., un φ1 are solutions of L(y) = 0. These functions are linearly independent.
Suppose we have constants c1, c2, c3,...., cn such that
c1 φ1 + c2 u2 φ1 + c3 u3 φ1 + ..... + cn unφ1 = 0
Since, φ1 ( x ) ≠ 0 on I this implies
c1 + c2u 2 + c3u3 + .... + cn un = 0
Differentation above equation and substituting
uk′ = vk for k = 2, 3,...., n we get
c2v2 + c3v3 + c4v4 + .... + cn vn = 0.
Since v2, v3, v4,....., vn are linearly independent by definition 1.3 we get c2 = c3 = c4 =
..... = cn = 0 and therefore c1 = 0. Thus φ1, u2φ1, u3 φ1,....., un φ1 are linearly independent solutions.

EXAMPLES

Q. 1. Consider the equation


1 ′ 1
y′′ + y – 2 y=0 for x > 0.
x x
(a) Show that there is a solution of the form xr, where r is a constant.
(b) Find two linearly independent solutions for x > 0 and prove that they are linearly
independent.
(c) Find the two solutions f1, f2 satisfying
φ1 (1) = 1 ; φ2 (1) = 0
φ1′ (1) = 0 ; φ2′ (1) = 1
Ans (a) :

1 1
Let φ ( x ) = x r be a solution to L( y ) = y ′′ + y ′ – 2 y = 0 Since f is a solution
x x
L(y) = 0.
Therefore r ( r –1) x r –2 + r x r –2 – x r –2 = 0
that is (r 2 –1) x r –2 = 0 for x>0
Thus, r 2 –1 = 0 or r = +1, –1.
1
Therefore φ1 ( x) = x and φ2 ( x) = are two solutions of L(y) = 0.
x
Ans (b) :
c2
Let c1φ1 + c2φ 2 = 0 then c1x + = 0 . Differentiate this equation twice with respect to
x
Differential Equations (60)
2c2
x we get = 0 implies c2 = 0 and therefore c1 = 0. Thus, f1 and f2 are linearly independent.
x3
Ans (c) :
1
φ1 ( x) = c1 x + c2
x
φ1 (1) = 1 and φ1′ (1) = 0 gives
1
c1 x + c2 = 1 at x = 1 i.e. c1 + c2 = 1
x
c2
c1 x – = 0 at x = 1 i.e. c1 – c 2 = 0
x2

and φ1 ( x) =  x + 
1 1 1
Thus, c1 = c2 =
2 2 x
1
Let, φ2 ( x ) = d1x + d 2
x
φ2 (1) = 0 and φ2′ (1) = 1 gives
1 1
d1 + d 2 = 0 and d1 – d 2 = 1. Then d1 = and d2 = –
2 2
φ 2 ( x ) =  x –  .
1 1
and
2 x
Q. 2. Find two linearly independent solutions of the equation
1
(3 x –1)2 y′′ + (9 x – 3) y′ – 9 y = 0 for x>
2
dy dy dt
Ans. : Put t = 3 x –1 then = ⋅ = y& ⋅ 3
dx dt dx
where . represents derivative with respect to t.
d2y d dy d d dt
2
= = y& ⋅ 3 = (3 y& ) = 3&&y ⋅ 3 = 9 &&
y
dx dx dx dx dt dx
Therefore 9t 2 &&
y + 9t y& – 9 y = 0
or t 2 &&
y + ty& – y = 0
Let y = tr be a solution then
r ( r –1)t r + rt + r – t r = 0 implies (r 2 –1)t r = 0.
But t > 0 therefore r = +1, –1
1
and φ1 (t ) = t and φ2 (t ) = are solutions
t
But t = 3x – 1 and therefore
1
φ1 ( x ) = 3 x –1 and φ2 ( x) = are two solutions of given equation.
3x –1

Differential Equations (61)


c2
c1 φ1 ( x ) + c2 φ 2 ( x ) = 0 implies c1 (3 x –1) + =0
3x –1
18c2
On differentiating this equation two times with respect to x we get – = 0 and
(3 x –1)2
therefore c2 = 0. Since 3x – 1 ¹ 0, c1 = 0.
Thus f1 and f2 are linearly independent.
Q. 3. A differential equation and a function f1 are given in each of the following. Verify
that the function f1 satisfies the equation and find a second independent solution.
(a) x 2 y ′′ – 7 xy′ + 15 y = 0, φ1 ( x) = x3 , ( x > 0)
(b) xy ′′ – ( x + 1) y ′ + y = 0, φ1 ( x ) = e x , ( x > 0)
(c) (1 – x 2 ) y′′ – 2 xy′ + 2 y = 0, φ1 ( x ) = x, (0 < x < 1)

Ans (a) : φ1 ( x ) = x 3 , φ1′ ( x ) = 3 x 2 , φ1′′ ( x ) = 6 x

L(φ1 ) = x 2φ1′′ – 7 xφ1′ + 15φ1 = x 2 (6 x) – 7 x (3 x 2 ) + 15 x 3 = 0.

Since L(φ1 ) = 0, φ1 is a solution of L(y) = 0


To determine the second solution, since x > 0, we can divide the given equation by x2.
Consider y′′ – 7 y′ + 15 y = 0.
x x2
7 3 15 3
Let φ ( x ) = u ( x) φ1 ( x ) = x 3u ( x ) be a solution. Then ( x u( x))′′ – ( x u)′ + 2 ( x u) = 0
3
x x
7 15
gives (u′′x3 + 6 x 2u ′ + 6 xu ) – ( x 3u′ + 3x 2u ) + 2 ( x 3u ) = 0 or u ′′x + 6 x u ′ – 7 x u ′ = 0
3 2 2
x x
u ′′ 1
i.e. u′′x – u′ = 0 ⇒ = (Integrate with respect to x)
u ′ x
log u ′ = log x + log k ⇒ u′ = k x

x2
But then u = k Let k = 2 Then φ ( x ) = u ( x ) φ1 ( x ) = x 2 ( x 3 ) = x 5 is the
2
second solution independent of f1 as c1x 3 + c2 x 5 = 0 implies c1 = c2 = 0.

1 1
Ans (b) : φ1 ( x) = e x = φ1′ ( x) = φ1′′ ( x) Let L( y ) = y ′′ – (1 + ) y ′ + y = 0 . (We can divide the
x x
given equation by x as x > 0.)
1 1
L(φ1 ) = e x – (1 + ) e x + e x = 0 ∴ φ1 is a solution.
x x
To determine second solution, let φ ( x ) = u ( x) φ1 ( x) be a solution then by theorem 2.1.5

exp  – ∫ a1 (t ) dt  dx
1 x
u ( x) = ∫
φ12 ( x )  

Differential Equations (62)


 
= ∫ e –2 x exp  + ∫ 1 +  dx  dx
1
  x 
= ∫ e –2 x exp[ x + log x ] dx

= ∫ e –2 x xe x dx = ∫ xe – x dx = – (1 + x) e–x
Thus, f2 (x) = – (1 + x) is a second solution of the equation L(y) = 0
c1φ1 ( x) + c2φ 2 ( x) = c1e x + c2 (–1) (1 + x) = 0 ⇒ c1 + c2 (–1) (1 + x)e – x = 0
But then c2 = c1 = 0 therefore f1 and f2 are linearly independent solutions.

2x 2y
Ans (c) : φ1 ( x) = x, φ1′ ( x) = 1, φ1′′ ( x) = 0. Let L( y ) = y ′′ – 2
y′ + = 0.
1– x 1– x 2
2x 2x
L(φ1 ) = 0 – 1 + = 0. Therefore f1 is a solition L(y) = 0. To determine
1 – x 2 1– x 2

second solution, let f2(x) = u (x)f1(x) be a solution of L(y) = 0.

By Theorem 2.1.5.

exp  – ∫ a1 (t ) dt  dx
1 x
u ( x) = ∫
φ12  

1  x –2t dt  1 1
=∫ exp  – ∫ 2 
dx = ∫ 2 dx
x2
 1– t  x 1– x 2
dx dx 1 dx 1 dx
=∫ =∫ 2 + ∫ + ∫
2 2
x (1– x ) x 2 1– x 2 1 + x

1+ x 
= – x –1 + log 
1
.
2  1– x 
1+ x 
Then φ2 ( x ) = φ1 ( x ) u ( x ) = –1 + log 
x
 is a second solution.
2  1– x 

Q. 4. One solution of x 3 y′′′ – 3x 2 y′′ – 6 xy′ – 6 y = 0 for x > 0 is f1 (x) = x find the remaining
two independent solutions for x > 0.
Ans : Let f = xu be a solution of L (y) = 0. Then φ ′ = xu ′ + u , φ ′′ = xu ′′ + 2u ′, φ ′′′ = xu ′′′ + 3u ′′ .

L(φ ) = x 3 ( xu ′′′ + 3u ′′) – 3x 2 ( xu ′′ + 2u ′) + 6 x ( xu ′ + u ) – 6 xu = 0 implies x 4u′′′ = 0. Since

x ≠ 0, u ′′′ = 0 gives u = c1x + c2 x 2 . Thus u = x and u = x2 are two linear independent


solutions of u ′′′ = 0 . But f = xu is a solution. Therefore φ2 ( x ) = x 2 and φ3 ( x ) = x3 are
remaining two linearly independent solutions.

Differential Equations (63)


Q. 5. Consider the equation L( y ) = y′′′ + a1 ( x) y′′ + a2 ( x) y′ + a3 ( x) y = 0. Suppose f1 and f2
are given linearly independent solutions of L (y) = 0. Let f = uf1 and compute the
 φ2 ′
solution of order two satisfied by u¢ in order that L (f ) = 0. Show that   is a
 φ1 
solution of this equation of order two.
Ans (c) : Let f = uf1 be a solution of L(y) = 0.

φ ′ = u ′φ1 + u φ1′ , φ ′′ = u ′′φ1 + 2u ′φ1′ + uφ1′′ , φ ′′′ = u ′′′φ1 + 3u ′′φ1′ + +3u ′φ1′′ + uφ1′′′

L( y ) = u ′′′φ1 + 3u′′φ1′ + 3u ′φ1′′ + u φ1′′′  + a1 ( x ) u ′′φ1 + 2u ′φ1′ + u φ1′′ 


   

+ a 2 ( x ) u ′φ1 + uφ1′  + a3 ( x ) u φ1 = 0.
 
Since f1 is a solution φ1′′′ + a1φ1′′ + a2φ1′ + a3φ1 = 0.

L( y ) = u ′′′φ1 + 3u′′φ1′ + 3u′φ1′′ + a1 ( x) u ′′φ1 + 2u ′φ1′  + a2 ( x ) u′φ1 = 0.


 

= φ1 u ′′′ + 3φ1′ + a1 ( x)φ1  u ′′ + 3φ1′′ + 2a1φ1′ + a2φ1  u ′ = 0


   
φ  ′
Thus, L(y) = 0 is an equation of order two in u ′.  2  is a solution of L(v) = φ1v′′
 φ1 
+ 3φ1′ + a1φ1  v′ + 3φ1′′ + 2 a1φ1′ + a2φ1  v = 0 if it satisfies the equation, L(v) = 0.
   

 φ ′ φ ′ φ φ ′ φ ′′ φ ′ φ ′ φ ′ φ ′ φ φ ′′ 2φ φ ′2
v =  2  = 2 – 2 21 , v′ = 2 – 1 22 – 1 22 – 2 21 + 2 31
 φ1  φ1 φ1 φ1 φ1 φ1 φ1 φ1

φ 2′′ 2φ1′ φ 2′ φ 2 φ1′′ 2φ 2 φ1′ 2


v′ = – – +
φ1 φ12 φ12 φ13

φ 2′′′ 3φ 2′′ φ1′ 3φ1′′ φ 2′ 6φ1′ 2 φ 2′ 6φ 2 φ1′φ1′′ 6φ 2 φ1′ 3 φ 2 φ1′′′


v '' = – – + + – –
φ1 φ12 φ12 φ13 φ13 φ14 φ12

L (v ) = φ1v′′ + 3φ1′ + a1φ1  v′ + 3φ1′′ + 2 a1φ1′ + a2φ1  v


   

( φ
) (
= φ2′′′ + a1φ2′′ + a2φ2′ – 2 φ1′′′ + a1φ1′′ + a2φ1′
φ1
)
Since f 2 and f 1 are solutions, φ2′′′ + a1φ2′′ + a2φ2′ + a3φ2 = 0 and φ1′′′ + a1φ1′′ + a2φ1′
+ a3φ1 = 0 and therefore
φ2
L (v ) = – a3 φ 2 – ( – a3φ1 ) = 0.
φ1
 φ2 ′
Thus, v =   is a solution of reduced equation.
 φ1 
Differential Equations (64)
EXERCISE

Use the reduction of order method and find the general solution of each of the following
equations. Verify that f1 satisfies the equation.

(a) x 2 y′′ – xy ′ + y = 0, φ1 = x (Ans. y = c1x + c2 x log x )

2 ′ 2
(b) y′′ – y + 2 y = 0, φ1 = x (Ans. y = c1x + c2 x 2 )
x x

(c) (2 x 2 + 1) y′′ – 4 xy′ + 4 y = 0, φ1 = x (Ans. y = c1x + c2 (2 x 2 –1))

x3 x 2
– +
e 3 2
(d) y ′′ + ( x 2 – x ) y ′ – ( x –1) y = 0, φ1 = x (Ans. y = c1x + c2 x ∫ dx )
x2

– x2
x 1
(e) y ′′ +  –  y ′ – y = 0, φ1 = x
2 e 4
2 x (Ans. y = c1 x 2 + c2 x 2 ∫ dx)
x3

1
(f) 2 x y ′′ + 3xy′ – y = 0, φ1 = x
2 12
(Ans. y = c1 x2 + c2 x –1 )

Unit 2 : Basis
In the course on linear algebra we learn about a vector space also called as linear space
and the basis of a linear space. Suppose S is a set of functions with the following property.
If f1, f2 Î S, c1f1 + c2f2 Î S for any two constants c1, c2. Then the set S is called a linear
space of functions. If a linear space of functions S contains n functions φ1, φ2 , φ3 ,...., φn which
are linearly independent and every function from S can be represented as a linear combination
of these functions φ1, φ2 , φ3 ,...., φn then the set { φ1, φ2 , φ3 ,...., φn } is called a basis for the linear
space S. The number n is called dimension of S.
For a given linear differential equation L( y ) = y ( n ) + a1 ( x ) y ( n –1) + a2 ( x) y ( n –1) + ⋅⋅⋅
+ an ( x ) y = 0, the collection of all solutions denoted by S of L(y) = 0 is a linear space. Every
basis of S contains n linearly independent functions and therefore dimension of solution space S
is n.
To check the linear independence of functions φ1, φ2 , φ3 ,...., φn , we consider the wronskian
W( φ1, φ2 , φ3 ,...., φn ). There is a relation between the linear indepence of functions and the
Wronskian W( φ1, φ2 , φ3 ,...., φn ). In chapter I we have proved this result for the linear differential
equation with constant coefficients.
Differential Equations (65)
A. Linear Independence and Wronskian

In section 1(B) we have seen that for the differential equation L( y ) = y ( n ) + a1 ( x) y ( n –1)
+ a2 ( x ) y ( n –2) + ⋅⋅⋅ + an y = 0 there are n linearly independent solutions φ1, φ2 , φ3 ,...., φn satisfying
the initial conditions φi(i –1) ( x0 ) = 1, φi( j –1) ( x0 ) = 0, j ≠ i. These linearly independent solutions
is a basis of solution space of L(y) = 0. Every solution of L(y) = 0 can be represented as a linear
combination of these functions φ1, φ2 , φ3 ,...., φn .

Theorem 2.2.1
Let φ1, φ2 , φ3 ,...., φn be n solutions of L(y) = 0 on I satisfying the initial conditions.

φi(i –1) ( x0 ) = 1, φi( j –1) ( x0 ) = 0, j ≠ i, x0 ∈ I


If f is any solution of L(y) = 0 on I, there are n constants c1, c2, c3,....,cn such that
φ = c1φ1 + c2φ2 + ⋅⋅⋅ + cnφn
Proof :

Let f is any solution of L(y) = 0 on I. Let φ ( x0 ) = α1, φ ′( x0 ) = α 2, φ ′′( x0 ) = α3,

⋅⋅⋅ φ ( n –1) ( x0 ) = α n for some constants α1, α 2, α 3, ⋅⋅⋅ α n .

Consider a function ψ = α1 φ1 + α 2 φ2 + α 3 φ3 + ⋅⋅⋅ + α n φn

Since φ1, φ2 ,...., φ n are solutions of L(y) = 0, by superposition principle (chapter 2 unit
1(B)) y is also a solution of L(y) = 0 and clearly
ψ ( x0 ) = α1 φ1 ( x0 ) + α 2 φ2 ( x0 ) + α3 φ3 ( x0 ) + ⋅⋅⋅ + α n φn ( x0 ) = α1
as φ1 ( x0 ) = 1 and φi ( x0 ) = 0 for i = 2,3, 4,..., n.

ψ ′( x0 ) = α1 φ1′ ( x0 ) + α 2 φ2′ ( x0 ) + α3 φ3′ ( x0 ) + ⋅⋅⋅ + α n φ n′ ( x0 ) = α 2


Since, φ1′ ( x0 ) = 0, φ2′ ( x0 ) = 1, φ3′ ( x0 ) = 0 ⋅⋅⋅ φ n′ ( x0 ) = 0.

ψ ′′( x0 ) = α1 φ1′′ ( x0 ) + α 2 φ2′′ ( x0 ) + α 3 φ3′′ ( x0 ) + ⋅⋅⋅ + α n φ n′′ ( x0 ) = α 3

Since, φ1′′ ( x0 ) = 0, φ2′′ ( x0 ) = 1, φ3′′ ( x0 ) = 0 ⋅⋅⋅ φ n′′ ( x0 ) = 0.


In general ψ i ( x0 ) = αi i = 3, 4,5,...., n –1
Thus, we see that
ψ ( x0 ) = α1 , ψ ′( x0 ) = α 2 , ψ ′′( x0 ) = α3 ,...,ψ ( n –1) ( x0 ) = α n .
Thus, y is a solution of L(y) = 0 having the same initial conditions at x0 as that of f . By
uniqueness theorem (chapter II Unit I theorem 2.1.2) we must have y = f .
i.e. φ = α1 φ1 + α 2 φ 2 + α3 φ3 + .... + α n φn

Differential Equations (66)


Thus, every solution of L(y) = 0, can be uniquely represented as a linear combination of
φ1, φ2 ,...., φ n . Since φ1, φ2 ,...., φ n are linearly independent the set { φ1, φ2 , φ3 ,...., φn } is a basis
for the solutions L(y) = 0.

Recall that the Wronkian of n functions φ1, φ2 , φ3 ,...., φn is defined as the determinant

φ1 φ2 φ3 L φn
φ1′ φ 2′ φ3′ L φ n′
W (φ1, φ 2 , φ3 ,...., φn ) = φ1′′ φ 2′′ φ3′′ L φ n′′
M M M M
φ1(n –1) φ 2(n –1) φ3(n –1) φ n (n –1)

Theorem 2.2.2 :
If φ1, φ2 , φ3 ,...., φn are n solutions of L(y) = 0 where L( y ) = y ( n ) + a1 ( x ) y ( n –1) +
a2 ( x) y ( n –2) + ⋅⋅⋅ + an ( x) y, on an interval I, then they are linearly independent on I if and only
if W (φ1, φ 2 , φ3 ,...., φn ) ( x) ≠ 0 for all x in I.
Proof :

Suppose W (φ1, φ 2 , φ3 ,...., φn ) ( x) ≠ 0 for all x in I. We show that φ1, φ2 , φ3 ,...., φn are
n
linearly independent on I. i.e. ∑ ci φi = 0 ⇒ c1 = c2 = c3 = .... = cn = 0
i=1
If there are constants c1, c2, c3,....,cn such that
∑ ci φi ( x) = c1 φ1 ( x) + c2 φ2 ( x) + c3 φ3 ( x) + .... + cn φ n ( x) = 0 for all x in I then clearly,

c1 φ1′ ( x) + c2 φ 2′ ( x) + c3 φ3′ ( x) + .... + cn φ n′ ( x) = 0

c1 φ1′′ ( x) + c2 φ2′′ ( x) + c3 φ3′′ ( x) + .... + cn φn′′ ( x) = 0


M
c1 φ1( n –1) ( x ) + c2 φ 2 ( n –1) ( x ) + c3 φ3( n –1) ( x) + .... + cn φ n ( n –1) ( x ) = 0

Thus we get a system of linear equations

 φ1 ( x ) φ2 ( x ) φ3 ( x ) L φn ( x ) 
 c1 
   0
 φ1′ ( x ) φ 2′ ( x ) φ3′ ( x ) L φ n′ ( x )     
   c2   0
 φ1′′ ( x ) φ 2′′ ( x ) φ3′′ ( x ) L φ n′′ ( x )   c3  =  0
     

M M M M
 M M 
φ1(n –1) ( x) φ2 (n –1) ( x) φ3(n –1) ( x)  cn   0
φn (n –1) ( x) 

Differential Equations (67)


The above system can be represented by Ax = 0. If A is invertible then we can premultiply
–1
by A and we get x = 0. The square matrix is invertible if it is non-singular i.e. determinant of
A is non-zero.
The determinant of the matrix A is W (φ1, φ 2 , φ3 ,...., φn ) ( x) . Therefore if
n
W (φ1, φ 2 , φ3 ,...., φn ) ( x) ≠ 0 then c1= c 2 = c 3 =.... = cn = 0. Since ∑ ci φi = 0 ⇒ ci = 0,
i=1
i = 1, 2,3,...., n, φ1, φ2 , φ3 ,..., φn are linearly independent.
Conversely, suppose φ1, φ2 , φ3 ,...., φn are linearly independent solutions of L(y) = 0 defined
on I. Suppose there is an x0 in I such that W (φ1, φ 2 , φ3 ,...., φ n ) ( x0 ) = 0.
Then system of n linear equations
 φ1 ( x0 ) φ2 ( x0 ) φ3 ( x0 ) L φ n ( x0 ) 
 c1 
  0 
 φ1′ ( x0 ) φ 2′ ( x0 ) φ3′ ( x0 ) L φ n′ ( x0 )     
   c2  0 
 φ1′′ ( x0 ) φ 2′′ ( x0 ) φ3′′ ( x0 ) L φ n′′ ( x0 )   c3  = 0 
     

M M M M
 M M 
φ1( n –1) ( x0 ) φ2( n –1) ( x0 ) φ3( n –1) ( x0 ) cn  0 
φ n( n –1) ( x0 ) 
has a solution c 1, c2, c 3,....,cn, where not all the constants c1, c 2, c3,...,cn are zero. Let
c1, c2, c3,....,cn be a non-zero solution of above system of equations and consider
ψ ( x ) = c1 φ1 ( x ) + c2 φ2 ( x ) + c3 φ3 ( x ) + ⋅⋅⋅ + cn φ n ( x ).
Since, φi , i = 1, 2,3....n are solution of L(y) = 0, y is also a solution of equation L(y) = 0.
Now L(y) = 0 and from above system of equations we get
ψ ( x0 ) = c1 φ1 ( x0 ) + c2 φ2 ( x0 ) + c3 φ3 ( x0 ) + ⋅⋅⋅ + cn φ n ( x0 ) = 0 .

ψ ′( x0 ) = c1 φ1′ ( x0 ) + c2 φ2′ ( x0 ) + c3 φ3′ ( x0 ) + ⋅⋅⋅ + cn φn′ ( x0 ) = 0


In general
ψ (i ) ( x0 ) = c1 φ1(i ) ( x0 ) + c2 φ 2 (i ) ( x0 ) + c3 φ3(i ) ( x0 ) + ⋅⋅⋅ + cn φ n (i ) ( x0 ) = 0
for i = 1, 2, 3, 4,...., n –1.

Thus, ψ ( x0 ) = ψ ′( x0 ) = ψ ′′( x0 ) = .... = ψ (n –1) ( x0 ) = 0.

From theorem 2.1.1 it follows that ψ ( x ) ≡ 0 on I.

Therefore, c1 φ1 ( x ) + c2 φ 2 ( x ) + c3 φ3 ( x ) + ⋅⋅⋅ + cn φ n ( x ) = 0 for all x in I. Thus, we have c1,


c2, c3,..... cn not all zero such that c1 φ1 ( x ) + c2 φ 2 ( x ) + c3 φ3 ( x ) + ⋅⋅⋅ + cn φ n ( x ) = 0 for all x in I.
Therefore the set φ1, φ2 , φ3 ,...., φn is not linearly independent on I. But this contradicts the fact
that φ1, φ2 , φ3 ,...., φn are linearly independent on I. Therefore the assumption that there was a
point x0 in I such that W (φ1, φ 2 , φ3 ,...., φ n ) ( x0 ) = 0. must be false i.e. W (φ1, φ 2 , φ3 ,...., φn ) ( x) ≠ 0
for all x in I.

Differential Equations (68)


Theorem 2.2.3
Let φ1, φ2 , φ3 ,...., φn be n linearly independent solutions of L( y ) = y ( n ) + a1 ( x) y ( n –1)

+ a2 ( x ) y ( n –2) + ⋅⋅⋅ + an ( x ) y = 0 on an interval I. If f is any solution of L(y) = 0 on I, it can be


represented in the form φ = c1 φ1 + c2 φ 2 + c3 φ3 + .... + cn φn , where c1, c2, c3,..... cn are constants.
Thus any set of n linearly independent solutions of L(y) = 0 on I is a basis for the solution space
of L(y) = 0 on I.
Proof :

Let x0 be a point in I. Suppose f is any solution of L(y) = 0. Let φ ( x0 ) = α1, φ ′( x0 ) = α 2,


φ ′′( x0 ) = α3, ⋅⋅⋅ φ ( n –1) ( x0 ) = α n We show that there exist unique constants c1, c2, c3,..... cn such
that ψ = c1 φ1 ( x ) + c2 φ 2 ( x) + c3 φ3 ( x) + ⋅⋅⋅ + cn φ n ( x ) is a solution of L(y) = 0 satisfying

ψ ( x0 ) = α1 , ψ ′( x0 ) = α 2 , ψ ′′( x0 ) = α3 ,....,ψ ( n –1) ( x0 ) = α n . These initial conditions are


equivalent to the following equations for c1, c2, c3,..... cn (e.g. ψ ( x0 ) = c1φ1 ( x0 ) + c2φ2 ( x0 ) +
.... + cnφ n ( x0 ) = α1 )

 φ1 ( x0 ) φ 2 ( x0 ) φ3 ( x0 ) L φn ( x0 ) 
 c1   α1 
    α 
 φ1′ ( x0 ) φ 2′ ( x0 ) φ3′ ( x0 ) L φn′ ( x0 ) 
 c2   2
   c3  = α3 
 φ1′′ ( x0 ) φ 2′′ ( x0 ) φ3′′ ( x0 ) L φ n′′ ( x0 ) 
   
 M M M M  M  M 
   cn  α n 
φ1(n –1) ( x0 ) φ 2(n –1) ( x0 ) φ3(n –1) ( x0 ) φn (n –1) ( x0 )

Since φ1, φ2 , φ3 ,...., φn are linearly independent by theorem 2.2.2, W (φ1, φ 2 , φ3 ,...., φ n )
( x0 ) ≠ 0. Therefore the coefficient matrix is inversible and there is a unique solution
c1, c2, c3,..... cn of the above system of equations.
Thus we have a unique solution
ψ = c1 φ1 ( x ) + c2 φ 2 ( x) + c3 φ3 ( x) + ⋅⋅⋅ + cn φ n ( x )

Satisfying ψ ( x0 ) = α1 , ψ ′( x0 ) = α 2 , ψ ′′( x0 ) = α3 ,....,ψ ( n –1) ( x0 ) = α n . But f is a solution


with identical initial conditions. Therefore by uniqueness theorem we have f (x) = y (x) on I.
Thus φ ( x) = c1 φ1 ( x) + c2 φ2 ( x) + c3 φ3 ( x) + ⋅⋅⋅ + cn φ n ( x) on I and any solution of L(y) = 0 can
be represented as a linear combination of n linearly independent solutions φ1, φ2 , φ3 ,...., φn .

In theorem 2.2.2 we have seen that the function φ1, φ2 , φ3 ,...., φn are linearly independent
solutions of L(y) = 0 if and only if the Wronskian W (φ1, φ 2 , φ3 ,...., φn ) ( x) ≠ 0 for all x in I. In the
next theorem we show that it is sufficient to calculate the Wronskian W (φ1, φ 2 , φ3 ,...., φ n ) at
some point x0 in I.

Differential Equations (69)


Theorem 2.2.4
Let φ1, φ2 , φ3 ,...., φn be n solutions of L(y) = 0 on an interval I and let x0 be any point in I.
Then
 x 
W(φ1, φ2 , φ3 ,..., φn ) ( x) = exp  – ∫ a1(t ) dt  W(φ1, φ 2 , φ3 ,..., φ n ) ( x0 )
 x0 
Note that since exponential function is non-zero function,
W(φ1, φ 2 , φ3 ,..., φ n ) ( x0 ) ≠ 0 implies W(φ1, φ2 , φ3 ,..., φn ) ( x) ≠ 0 for all x in I.
Proof :

φ1 φ2 φ3 L φn
φ1′ φ 2′ φ3′ L φ n′
Let W = W (φ1, φ 2 , φ3 ,...., φn ) = φ1′′ φ 2′′ φ3′′ L φ n′′
M M M M
φ1(n –1) φ 2(n –1) φ3(n –1) φ n (n –1)

On differentiating W row wise we get

 φ′ φ 2′ φ3′ L φ n′   φ1 φ2 φ3 L φn 
 1   
 φ1′ φ 2′ φ3′ L φ n′   φ1′′ φ 2′′ φ3′′ L φ n′′ 
   
W ′ =  φ ′′ φ ′′ φ ′′ L φ ′′  +  φ ′′ φ ′′ φ ′′ L φ ′′  + ...
1 2 3 n 1 2 3 n
   M
 M M M M  M M M 
 
 ( n –1) ( n –1)   ( n –1) 
φ1 φ2 φ3 φn
( n –1) ( n –1) ( n –1)
φ1 φ2 φ3 φn
( n–1) ( n–1)
 

 φ1 φ2 φ3 L φn 
 
 φ1′ φ 2′ φ3′ L φ n′ 
+  M M M L M 

φ (n –2) φ (n –2) φ (n –2) φn ( n –2) 
 1 2 3

 φ1(n ) φ 2( n ) φ3(n ) φ n (n ) 
= V1 + V2 + V3 + ..... + Vn (say)
Where Vk differs from W only in its kth row and the kth row of Vk is obtained by
differentiating the kth row of W. The first n – 1 determinants are all zero, since they each have
two identical rows. Observe that Vk has kth and (k + 1)th row identical.
Since φ1, φ2 ,...., φ n are solution of L(y) = 0, we have

φi( n ) = – a1 φi( n –1) – a2 φi( n –2) – a3 φi( n –3) .... – an φi (i = 1, 2, 3,..., n)


Differential Equations (70)
n –1
= – ∑ an – j φi( j )
j =0
Therefore
 φ1 φ2 φ3 L φn 
 
 φ1′ φ2 φ3 L φn′ 
 
 φ1′′ φ2′′ φ3′′ L φn′′ 
W ′ = Vn =  M M M M

 
 φ1( n –2)
φ2 (n –2) φ3(n –2) φn ( n –2) 
 
 n –1 n –1 n –1 n –1 
 – ∑ an – j φ1 – ∑ an – j φ 2( j ) – ∑ an – j φ3( j ) L – ∑ an – j φ n( j ) 
( j)

 j =0 j =0 j =0 j =0 
Elimentary row transformations do not change the value of the determinant. Perform the
transformation
Rn + an R1 + an –1R3 + an –3 R3 + ... + a2 Rn –1. we get

 φ1 φ2 φ3 L φn 
 
 φ1′ φ2′ φ3′ L φn′ 
W ′ =  M M M M 

 φ ( n –2) φ2( n –2) φ3( n –2) L φ1( n–2) 
 1 
 – a1φ1( n –1) – a1φ2( n –1) – a1φ3( n –1) L – a1φn( n –1) 

 φ1 φ2 φ3 L φn 
 
 φ1′ φ 2′ φ3′ L φ n′ 
 
= – a1  φ1′′ φ 2′′ φ3′′ L φ n′′  = – a1W(φ1, φ2 , φ3 ,...., φ n )
 M M M M 
 
φ1(n –1) φ 2(n –1) φ3(n –1) φ n (n –1) 
Therefore W¢ + a1 W = 0 and we get,
x
– ∫ a1 ( t ) dt
W( x) = e x0
W( x0 )
 x 
i.e. W(φ , φ
1 2 3, φ ,...., φ n ) ( x ) = exp  – ∫ 1
a (t ) dt  W(φ1, φ2 , φ3 ,...., φn ) ( x0 ).
 x0 
Corollatory : If the coefficient a1 is constant then
W(φ1 , φ2 , φ3 ,...., φ n ) ( x) = e – a1 ( x – x0 ) W(φ1, φ 2 , φ3 ,...., φ n ) ( x0 ).
From theorem 2.2.2 and theorem 2.2.4 it follows that n solutions φ1, φ2 , φ3 ,...., φn of
L(y) = 0 on I are linearly independent if and only if W(φ1, φ 2 , φ3 ,..., φ n ) ( x0 ) ≠ 0 for some point
x0 in I.
Differential Equations (71)
B. Solutions of non-homogeneous equation

The equation L( y ) = y ( n) + a1 ( x ) y ( n –1) + a2 ( x ) y ( n–2) + ... + an y = b( x) where a1, a2, a3,....,


an, b are continuous functions on an interval I is a non-homogeneous linear equation of order n
with variable coefficients. The solutions of this equation can be determined by the variations of
constant method.

Theorem 2.2.5
Let b (x) be a continuous function on an interval I and let φ1, φ2 , φ3 ,...., φn be a basis for the
solutions of L(y) = 0 on I. Every solution y of L( y ) = y ( n ) + a1 ( x) y ( n –1) + a2 ( x) y ( n –2) +

... + an ( x) y = b( x) can be written as ψ = ψ p + c1φ1 + c2φ2 + c3φ3 + ... + cnφn where is a yp is a


particular solution of L(y) = b (x) and c1, c2, c3,..... cn are constants. Every such y is a solution of
L(y) = b(x). A particular solution yp is given by
n x
Wk (t ) b(t )
ψ p = ∑ φk ( x ) ∫ dt
k =1 x0 W (φ1 , φ 2 , φ3 ,..., φ n ) (t )

where W(φ1, φ 2 , φ3 ,..., φ n ) is a wronkian of φ1, φ2 , φ3 ,...., φn and Wk is the determinant obtained

from W(φ1, φ 2 , φ3 ,..., φ n ) by replacing the kth column (φk φk ′ φk ′′ ....φk


( n –1) T
) by (0, 0, 0..., 0, 1)T.
Proof :
If yp is a particular solution of L(y) = b(x) and y is any other solution of L(y) = b(x), then
L(ψ – ψ p ) = L(ψ ) – L(ψ p ) = b ( x) – b ( x) = 0.
Therefore y – yp is a solution of corresponding homogeneous equation L(y) = 0. Since
φ1, φ2 , φ3 ,...., φn is a basis for the solution of L(y) = 0 on I, every solution of L(y) = 0 can be
expressed as a linear combination of φ1, φ2 , φ3 ,...., φn .
ψ – ψ p = c1φ1 + c2φ2 + c3φ3 + ... + cnφn
ψ = ψ p + c1φ1 + c2φ2 + c3φ3 + ... + cnφn.
A particular solution yp can be found by variation of constants method. Let yp be of the
form
ψ p = u1 ( x ) φ1 ( x ) + u2 ( x ) φ2 ( x ) + u3 ( x) φ3 ( x) + ... + un ( x) φn ( x )
Since yp is a solution, L(ψ p ) = b( x) .
ψ p′ = u1′φ1 + u2′ φ2 + u3′ φ3 + ... + u ′n φn + u1φ1′ + u2 φ2′ + u3 φ3′ + ... + un φn′

Choose u1, u2, u3,....un such that u1′φ1 + u2′ φ 2 + u3′ φ3 + ... + un′ φn = 0

Then ψ p′ = u1φ1′ + u2 φ2′ + u3 φ3′ + ... + un φn′


ψ p′′ = u1′φ1′ + u2′ φ2′ + u3′ φ3′ + ... + u′n φn′ + u1φ1′′ + u2 φ2′′ + u3 φ3′′ + ... + un φn′′

Let u1′φ1′ + u2′ φ 2′ + u3′ φ3′ + ... + un′ φn′ = 0 then


Differential Equations (72)
ψ p′′ = u1φ1′′ + u 2 φ2′′ + u3 φ3′′ + ... + un φn′′

In general choose u1′φ1( k ) + u2′ φ2 ( k ) + u3′ φ3( k ) + ... + un′ φn ( k ) = 0

Then ψ p (k +1) = u1φ1(k +1) + u2 φ2 (k +1) + u3 φ3(k +1) + ... + un φn (k +1)

and ψ p (n ) = u1′φ1(n –1) + u2′ φ2(n –1) + u3′ φ3(n –1) + ... + un′ φn (n –1) + u1φ1(n ) + u2 φ 2(n ) + ... + un φn (n)

If we choose u1′φ1( n –1) + u2′ φ2( n –1) + u3′ φ3( n –1) + ... + un′ φn( n –1) = b( x ) . Then
ψ p (n ) = u1φ1(n ) + u2 φ2 (n) + u3 φ3(n) + ... + un φn (n ) + b ( x)
Thus we have the following equations
ψ p = u1φ1 + u2 φ2 + u3 φ3 + ... + un φn

ψ p′ = u1φ1′ + u2 φ2′ + u3 φ3′ + ... + un φn′ ; u1′φ1 + u2′ φ 2 + u3′ φ3 + ... + u′n φn = 0

ψ p′′ = u1φ1′′ + u2 φ 2′′ + u3 φ3′′ + ... + un φn′′ ; u1′φ1′ + u2′ φ 2′ + u3′ φ3′ + ... + un′ φn′ = 0

ψ p′′′ = u1φ1′′′ + u2 φ 2′′′ + u3 φ3′′′ + ... + un φn′′′ ; u1′φ1′′ + u2′ φ 2′′ + u3′ φ3′′ + ... + un′ φn′′ = 0
M
ψ p (n –1) = u1φ1(n –1) + u 2 φ2 (n –1) + ... + un φn (n –1) ; u1′φ1(n –2) + u2′ φ2(n –2)

+ u3′ φ3( n –2) + ... + un′ φ n ( n –2) = 0

ψ p (n ) = u1φ1(n ) + u2 φ2(n ) + u3 φ3(n) + ... + un φn (n ) + b ( x) ; u1′φ1(n –1) + u2′ φ 2(n –1)

+u3′ φ3( n –1) + ... + u n′ φ n ( n –1) = b ( x)


Adding the terms columwise on left we get
L(ψ p ) = u1L(φ1 ) + u2 L(φ2 ) + u3 L(φ3 ) + ... + un L(φn ) + b ( x )
Since φ1, φ2 , φ3 ,...., φn are n solutions of homogeneous equation L(y) = 0, L(φ1 ) = L(φ 2 ) =
L(φ3 ) = ... = L(φ n ) = 0 and L(ψ p ) = b( x).
The right hand side equations are the following system of linear equations.

 φ1 φ2 φ3 L φn  u ′ 
   1 0 
 φ1′ φ2′ φ3′ L φn′  u2′  0 
     
 φ1′′ φ2′′ φ3′′ L φn′′   u3′  = 0 
 M    
M M M   M M 
 
φ1(n –1) φ 2(n –1) φ3(n –1)  ′ b 
φn (n –1) 
 n
u
We solve the above system of equations by Cramer’s rule.
Wk
Thus, uk′ =
W(φ1, φ 2 , φ3 ,..., φ n )
Where W(φ1, φ 2 , φ3 ,..., φ n ) is a Wronksian of φ1, φ2 , φ3 ,..., φn and Wk is the determinant
Differential Equations (73)
obtained from W(φ1, φ 2 , φ3 ,..., φ n ) by replacing kth column by (0, 0, 0,....., 0, b )T . Thus
φ1 φ2 L φ k –1 0 φ k +1 L φn
φ1′ φ 2′ φk′ –1 0 φk′ +1 L φ n′
Wk = φ1′′ φ 2′′ φk′′–1 0 φk′′+1 L φ n′′
M M M M M M
φ1( n –1) φ 2( n–1) φ k –1( n –1) b ( x ) φ k(n+1–1) φ n( n–1)

φ1 φ2 L φ k –1 0 φ k +1 L φn
φ1′ φ 2′ φ k′ –1 0 φ k′ +1 L φ n′
= b ( x) φ1′′ φ 2′′ φ k′′–1 0 φ k′′+1 L φ n′′ = b( x ) Wk
M M M M M M
φ1( n –1) φ2( n–1) φ k –1( n –1) 1 φk(n+1–1) φ n( n–1)

x
b ( x )Wk b(t ) Wk (t )
Thus, uk′ = i.e. uk = ∫ dt.
W(φ1 , φ 2 , φ3 ,..., φ n ) x0 W (φ1 , φ 2 , φ3 ,..., φ n ) (t )
n
and ψ p = u1φ1 + u2 φ 2 + u3 φ3 + ... + un φn = ∑ φk uk
k =1
n x
b(t )Wk (t )
= ∑ φk ( x) ∫ dt.
k =1 x0 W (φ1 , φ 2 , φ3 ,..., φ n ) (t )

EXAMPLES

Q. 1. Consider the equation L( y ) = y ′′ + a1 ( x ) y ′ + a2 ( x ) y = 0, where a1, a2 are continuous on


some interval I. Let f1, f2 and y1, y2 be two bases for the solution L(y) = 0. Show that
there is a non-zero constant k. Such that W (ψ1, ψ 2 ) ( x) = k W (φ1 , φ2 ) ( x)
Ans. : Since f1, f2 is bases for the solutions of L(y) = 0 and y1, y2 are solutions of L(y) = 0.
ψ 1 = c1φ1 + c2φ2 and ψ 2 = d1φ1 + d 2φ 2
for some constants c1, c2, d1, d2.
c1φ1 + c2φ 2 d1φ1 + d 2φ 2
W(ψ1,ψ 2 ) ( x) =
c1φ1′ + c2φ 2′ d1φ1′ + d 2φ 2′

c1φ1 d1φ1 + d 2φ 2 c2φ2 d1φ1 + d 2φ2


= +
c1φ1′ d1φ1′ + d 2φ 2′ c2φ2′ d1φ1′ + d 2φ 2′

 φ1 d1φ1 φ1 d 2φ2   φ2 d1φ1 φ2 d 2φ 2 


= c1  +  + c2  + 
 φ1′ d1φ1′ φ1′ d 2φ 2′   φ 2′ d1φ1′ φ 2′ d 2φ 2′ 

Differential Equations (74)


 φ1 φ 2 φ 2 φ1 
= c1d 2  + c2 d1 
′ ′ φ2′ φ1′ 
 φ1 φ2
φ1 φ2
= (c1d 2 – c2d1 ) = (c1d 2 – c2 d1 ) W(φ1, φ 2 ) ( x)
φ1′ φ2′
Thus, W(ψ 1,ψ 2 ) ( x) = (c1d 2 – c2 d1 ) W(φ1, φ 2 ) ( x)
Since ψ 1,ψ 2 are independent c1d 2 – c2 d1 ≠ 0.
Therefore there is a non-zero constant k = c1 d2 – c2 d1 such that W(ψ 1,ψ 2 ) ( x) =
k W(φ1 , φ2 ) ( x ).

Q. 2. Consider L( y ) = y′′ + a1 ( x ) y′ + a2 ( x ) y = 0 . Show that a1 and a2 are uniquely determined


by any basis f1, f2 for the solutions of L(y) = 0. Show that
φ1 φ2 φ1′ φ 2′
φ1′′ φ2′′ φ ′′ φ 2′′
a1 = – , a2 = – 1
W(φ1, φ2 ) W(φ1, φ 2 )
Ans. : Since f1, f2 is basis for solutions of L(y) = 0 , f1, f2 are solutions of L(y) = 0.
L (φ1 ) = φ1′′ + a1 φ1′ + a2 φ1 = 0

L(φ2 ) = φ2′′ + a1 φ2′ + a2 φ2 = 0


Solving above two equations for a1 and a2 by Cramers rule, we get.
– φ1′′ φ1 φ1 φ2
– φ 2′′ φ 2 φ1′′ φ 2′′
a1 = – =
φ1′ φ1 W(φ1, φ2 )

φ 2′ φ 2

+ φ1′ –φ1′′ φ1′ φ 2′


+ φ2′ –φ2′′ φ1′′ φ2′′
a2 = – =
φ1′ φ1 W(φ1, φ 2 )

φ 2′ φ 2
[ We use the elementory properties of determinants det A = det AT and if we interchange
row / column, the value of det change its sign.]

Q. 3. Consider the equation y ′′ + α ( x ) y = 0 where a is a continuous function on – ∞ < x < ∞ .


Let, f1, f2 be the basis for the solutions satisfying
φ1 (0) = 1, φ2 (0) = 0, φ1′ (0) = 1, φ2′ (0) = 1,
Show that W(φ1, φ2 ) ( x) = 1 for all x

Differential Equations (75)


Ans. : For the differential equation
y ′′ + a1 ( x ) y ′ + a2 ( x ) y = 0, if f1 and f2 are two solutions then
x
– ∫ a1 ( t ) dt
W(φ1, φ2 )( x) = e x0
W(φ1, φ2 ) ( x0 )
For y ′′ + α ( x ) y = 0, a1 = 0 . Therefore W(φ1, φ 2 ) ( x) = W(φ1, φ 2 ) (0)
φ1 (0) φ 2 (0) 1 0
= = =1
φ1′ (0) φ2′ (0) 0 1
Thus, W(φ1, φ2 ) ( x) = 1 for all x.

2
Q. 4. Find a general solution of y′′ – y = x (0 < x < a)
x2
2
Ans. : Assume that the solution of homogeneous equation L( y ) = y ′′ – y = 0 is of the form
x2
xr. Then y = xr implies L( x r ) = r (r –1) x r –2 – 2 x r –2 = 0 gives r (r – 1) – 2 = 0. Then
1
r2 – r – 2 = 0 implies r = 2 and r = – 1. Thus, f1(x) = x2 and φ 2 ( x ) = are solutions of
x
2
homogeneous equation L( y ) = y ′′ – 2 y = 0.
x
A solution yp of the non-homogeneous equation has the form
ψ p = u1 ( x) φ1 ( x) + u2 ( x) φ2 ( x)
1
= u1 ( x ) x 2 + u 2 ( x )
x
b( x ) W1 b( x ) W2
Where, u1′ = and u2′ =
W(φ1, φ 2 ) W(φ1, φ 2 )
Here b (x) = x
1
0
x 1 x2 0
W1 = =– , W2 = = x2
1 x 2x 1
1 – 2
x
1
φ1 φ2 x2
x
W(φ1, φ2 ) = = = –1 – 2 = –3 and we find that
φ1′ φ2′ 1
2x – 2
x
x (– 1x) 1 b( x ) W2 x( x 2 ) x3
u1′ = = and u2′ = = =–
–3 3 W –3 3
x x4
We may take u1 = and u2 = – . We skip the constants of integration as they correspond
3 12
to the solution of corresponding homogeneous equation.

Differential Equations (76)


Thus, the solution of non-homogeneous equation becomes
x 2 x 4 1 x3
ψp = ⋅x – ⋅ = .
3 12 x 4
Every solution f of L(y) = x has the form
x3 c
φ ( x) = ψ p + c1 φ1( x) + c2 φ2 ( x) = + c1x 2 + 2
4 x
where, c1 and c2 are constants.
Q. 5. One solution of x 2 y′′ – 2 y = 0 on 0 < x < ¥ is φ1 ( x ) = x 2 . Find all solutions of
x 2 y′′ – 2 y = 2 x – 1 on 0 < x < ¥.
Ans. : φ1 ( x ) = x 2 Let φ2 ( x ) = u ( x ) φ1 ( x ) be a solution of x 2 y′′ – 2 y = 0 . Then

L (φ 2 ) = x 2 u ′′( x) φ1 ( x) + 2u ′( x)φ1′ ( x) + u ( x )φ1′′ ( x )  – 2u ( x ) φ1 ( x ) = 0


 
= x 2 u′′x 2 + 2u′( x) ⋅ 2 x + u ( x) ⋅ 2  – 2u ( x) x 2 = 0
 
u ′′ 4
L(φ 2 ) = 0 gives u′′x 4 + 4u′x 3 = 0 i.e. =–
u ′ x
1
log u′ = –4 log x and u ′ = x –4 or u = – 3
3x
1 1
Therefore φ2 ( x) = u ( x) φ1 ( x) = – 3 ⋅ x = –
2
3x 3x
1
Since L(y) is a linear differential operator φ 2 ( x ) = is a second solution.
x
1
Thus, φ1 ( x ) = x 2 and φ2 ( x) = are solutions of the homogenous equation
x
2
L( y ) = x 2 y ′′ – 2 y = 0 or y ′′ – y = 0.
x
Equation x 2 y′′ – 2 y = 2 x – 1 is the given differential equation. To reduce the equation in
standard form we have to divide the given equation by x2 we can do so since x is positive.
2 2 1
Therefore consider the eqution y ′′ – y= – . Solution of this equation will be a solution
x2 x x2
of given equation.
2 2 1
A solution yp of a non-homogeneous equation y ′′ – y= – has the form
2 x x2
x
1
ψ p = u1 ( x ) φ1 ( x) + u 2 ( x) φ2 ( x ) = u1 ( x ) x 2 + u2 ( x )
x
b ( x )W1 b ( x )W2
Where, u1′ ( x ) = and u 2′ ( x ) =
W(φ1 , φ 2 ) W(φ1 , φ 2 )
2 1 x2 1
b( x ) = – 2 , W(φ1, φ 2 )= 1
x
= –3
x x 2x – 2
x
Differential Equations (77)
1
0 x 1 x2 0
W1 = = – , W2 = = x2
1 – 1 2 x 2x 1
x

 2 1  1 
 x – 2  – x  2 1 1 1 
u1 ( x ) = 
′ x   2 1
= 2– 3 and u1 ( x ) =  –  –  – 2 
–3 3x 3x 3  x  3  2x 
2 1  2
 x – 2  (x )
u2′ ( x ) =  x  2 1 1 1
= – x+ and u2 ( x) = – x 2 + x
–3 3 3 3 3
Thus ψ p = u1 ( x ) φ1 ( x ) + u2 ( x ) φ2 ( x )

 2 1   x2 x  1 1
=  – + 2  x2 +  – +  = –x + .
 3x 6 x   3 3 x 2
1 c
The general solution of given non-homogeneous equation is ψ = – x + + c1x 2 + 2 , where
2 x
c1 and c2 are constants.

Q. 6. One solution of x y′′ – xy′ + y = 0 ( x > 0) is φ1 ( x ) = x . Find the solutionsy of


2

x 2 y′′ – xy′ + y = x 2 satisfying ψ (1) = 1, ψ ′(1) = 0.

Ans. : The given non-homogeneous equation is y ′′ – 1 y ′ + 1 y = 1. (We can divide the equation
2 2
x x2
by x as x is positive)
Let φ2 ( x) = u ( x) φ1 ( x) = u ( x) x be an other solution.
c1
L(φ 2 ) = [u ′′x + 2u ′] – [u ′x + u ] + 2 u ( x) x = 0 gives u′′x + u′ = 0. Therefore u ′ = x
1 1
x x
and u ( x) = c1 log x.
φ2 ( x) = u ( x) φ1 ( x) = c1 x log x is second solution. Without loss of generality we choose
c1 = 1.
Thus, φ1 ( x ) = x and φ2 ( x ) = x log x are two solutions of homogeneous equation
1 ′ 1
y′′ –
y + 2 y = 0.
x x
A solution yp of a non-homogeneous equation
1 ′ 1
y′′ – y + 2 y = 1 has the form
x x
ψ p = u1 ( x ) φ1 ( x ) + u2 ( x ) φ2 ( x )
b( x )W1 b( x)W2
Then u1′ ( x ) = , u2′ ( x ) =
W(φ1 , φ2 ) W(φ1 , φ2 )

Differential Equations (78)


x x log x
where b( x ) = 1, W(φ1, φ 2 ) = =x
1 1 + log x
0 x log x x 0
W1 = = – x log x, W2 = =x
1 1 + log x 1 1
– x log x
u1′ ( x ) = = – log x and u1 ( x ) = – ( x log x – x )
x
x
u2′ ( x ) = = 1 and u2 (x) = x
x
Therefore ψ p = ( x – x log x) x + x ⋅ x log x = x 2
The general solution of given non-homogeneous equation is
ψ = x 2 + c1x + c2 x log x
Since ψ (1) = 1 and ψ ′(1) = 0, 1 + c1 = 1 and c1 = 0
ψ ′( x ) = 2 x + C2 (log x + 1), ψ ′(1) = 2 + c2 = 0 and c2 = –2
Therefore the solution satisfying given initial condition is
ψ ( x) = x 2 – 2 x log x.

Q. 7.

(a) Show that there is a basis f1, f2 for the solutions of x 2 y ′′ + 4 xy ′ + (2 + x 2 ) y = 0 ( x > 0)
of the form
ψ1 ( x) ψ 2 ( x)
φ1 ( x) = , φ 2 ( x) =
x2 x2
(b) Find all solutions of
x 2 y′′ + 4 xy′ + (2 + x 2 ) y = x 2 for x > 0 .
Ans. :

v
(a) Let φ = be a solution of the given homogeneous equation.
x2
4  2 
L ( y ) = y ′′ + y ′ +  2 + 1 y = 0
x x 
v′ 2v v′′ 4v′ 6v
Then, φ′ = 2
– 3 , φ ′′ = 2 – 3 + 4 and
x x x x x
 v′′ 4v′ 6v  4  v′ 2v   2  v
L( y ) =  2 – 3 + 4  +  2 – 3  +  2 + 1 2 = 0.
x x x  x x x  x x
Therefore L(y) = 0 implies v′′ + v = 0.
ψ1 ( x ) = cos x and ψ 2 ( x ) = sin x are two linearly independent solutions of v′′ + v = 0.

Differential Equations (79)


cos x sin x
Thus, φ1 ( x) = 2
and φ2 ( x) = 2 are two linearly independent solutions of given
x x
equation.
cos x sin x
(b) φ1 ( x) = 2
, φ2 ( x) = 2
x x
4  2 
L( y ) = y ′′ +
y′ +  2 + 1 y = 1
x x 
A solution yp of L(y) = 1 has the form

ψ p = u1 ( x) φ1 ( x) + u2 ( x) φ2 ( x).

b ( x )W1 b ( x)W2
Then u1′ ( x ) = , u 2′ ( x ) =
W(φ1 , φ2 ) W(φ1 , φ2 )
where b(x) = 1
cos x sin x
x2 x2 1
W(φ1, φ 2 ) = = 4
sin x 2 cos x cos x 2sin x x
– 2 – –
x x3 x2 x3
sin x cos x
0 2
0
x sin x x2 cos x
W1 = = – 2 , W2 = =
1
cos x 2sin x
– x sin x 2cos x
– 2 – 1 x2
2 3
x x x x3
– sin x
u1′ ( x ) = x2 = – x 2 sin x, u1 ( x ) = x 2 cos x – 2 x sin x – 2 cos x
1
x4
cos x
u2′ ( x ) = x2 = + x 2 cos x, u2 ( x ) = x 2 sin x + 2 x cos x – 2sin x
1
x4

ψ p = u1 ( x) φ1 ( x) + u2 ( x) φ2 ( x)
cos x sin x
= ( x 2 cos x – 2 x sin x – 2 cos x ) 2
+ ( x 2 sin x + 2 x cos x – 2sin x )
x x2
2
= 1–
x2
Therefore the general solution of non-homogeneous equation is
2 cos x sin x
ψ = ψ p + c1 φ1 + c2 φ2 = 1– 2
+ c1 2
+ c2 .
x x x2

Q. 8. Consider the equation y ′′ + y = b ( x ) where b is a continuous function on 1 ≤ x < ∞



satisfying ∫ | b(t ) | dt < ∞. show that particular solution y p is given by
1

Differential Equations (80)


x
ψ p ( x) = ∫ sin ( x – t )b(t ) dt
1
Ans. : The homogeneous equation y ′′ + y = 0 has two solutions φ1 ( x ) = cos x, φ2 ( x ) = sin x.
The particular solution yp has the form
ψ p ( x) = u1 ( x) φ1 ( x) + u2 ( x) φ2 ( x) where

b( x )W1 b( x )W2
u1′ ( x ) = , u2′ ( x ) =
W1 (φ1, φ2 ) W1 (φ1 , φ 2 )
cos x sin x
W (φ1, φ2 ) = =1
– sin x cos x
0 sin x cos x 0
W1 = = – sin x, W2 = = cos x
1 cos x – sin x 1
x x
b(t )W1 (t ) – b(t ) sin t
u1 ( x ) = ∫ dt = ∫ dt
1 W(φ1, φ 2 ) (t ) 1 1
x x
b(t )W2 (t ) b(t ) cos t
u2 ( x ) = ∫ dt = ∫ dt
1 W(φ ,
1 2φ ) ( t ) 1 1
x x
ψ p = – cos x ∫ b (t ) sin t dt + sin x ∫ b(t ) cos t dt
1 1
x
= ∫ b(t ) [sin x cos t – cos x sin t ] dt
1
x
= ∫ b(t ) sin ( x – t ) dt
1

EXERCISE

1. Consider the equation y ′′ + a1 ( x ) y ′ + a2 ( x ) y = 0 where a1(x) and a2(x) are continuous


functions on – ¥ < x < ¥ and are periodic with period q > 0 i.e. a1 ( x + θ ) = a1 ( x), a2 ( x + θ )
= a2 ( x) for all x. Let f be a non-trival solution and let ψ ( x ) = φ ( x + θ ) . Shown that y is
also a solution.
2. Consider the equation y ′′ + α ( x ) y = 0 where a is a continuous functions on – ¥ < x < ¥
which is of period q > 0. Let f1, f2 be the basis for solution satisfying
φ1 (0) = 1, φ 2 (0) = 0
φ1′ (0) = 0, φ2′ (0) = 1
Show that there is at least one non-trival solution f of period q if and only if φ1 (θ ) + φ2′ (θ ) = 2.

Differential Equations (81)


1 1
3. One solution of L( y) = y′′ + y = 0 for x > 0 is φ1 ( x ) = x 2 show that there is another
4x2
solution y of the form ψ = u φ where u is some function.

4. Use the method of variation of parameter and find the particular solution of the following
equations where the solutions for the related homogeneous equation are given.

(a) y ′′ –
2 2
y ′ + 2 y = x log x, φ1 ( x ) = x, φ2 ( x) = x 2  Ans. : ψ = 1 x3 log x – 3 x3 
x x  p
2 4 
1  Ans. : ψ p = x3 
(b) x y′′ + xy′ – 4 y = x , φ1 = x , φ2 =
2 3 2
x2  5

(c) x 2 y ′′ + xy′ – y = x 2e – x , φ1 = x, φ2 = 1  Ans. : ψ p = e – x (1 + x –1 ) 


x  
1
(d) 2 x 2 y ′′ + 3xy′ – y = x –1 ; φ1 = x 2 , φ2 = x –1  Ans. : ψ p = – 1 3 x –1 log x 
 

Unit 3 : Homogeneous equations with analytic coefficients


So far we have shown how to construct solutions of various special types of differential
equations using the exponential function, polynomials and the fundamental theorem of
calculus - that is how to reduce the integration of these differential equations to one or more
quadratures. The major difficulty with linear equations with variable coefficients, from a practical
point of view, is that it is rare that we can solve equations in terms of elementary functions, such
as exponential and trignometric functions. . However in case the coefficients a1, a2, a3, ...., an,
and b have convergent power series expansions the solutions will have this property also and
these series solutions can be obtained by a simple formal process.

An infinite series of the form ∑ an ( z – z0 ) is called a power series in z – z0. Here an, z,
n
n =0
z0 are complex numbers. With every power series there is associated a disk, called the disk of
convergence such that a series converges absoulately for every z interior to this disk. The center
of the disk is at z0 and its radius is called the radius of convergence of the power series.

Given a power series ∑ an ( z – z0 ) , let λ = nLim
n 1
–1∞
sup n
| an | , r = (where r = 0 if
n =0 λ
λ = +∞ and r = ¥ if λ = 0 ). The series converges absoulately if | z – z0 |< r and diverges if
| z – z0 |> r .


If x0, x and an are real numbers the series ∑ an ( x – x0 ) n is called a real power series. Its
n =0
disk of convergence intersects the real axis in an interval (x0 – r, x0 + r) called the interval of
convergence.
If g is a function defined on an interval I containing point x0 we say that g is analytic at x0
if g can be expanded in a power series about x0 which has a positive radius of convergence. Thus
g is analytic at x0 if it can be representd in the form
Differential Equations (82)

g ( x) = ∑ an ( x – x0 ) n
n =0

Where an are constants and the series converges for | x – x0 | < r , r > 0. If g has a power
series expansion then all the derivatives of g exist on | x – x0 |< r and they may be computed by
differentating the series term by term that is
∞ ∞
g ′( x ) = ∑ n an ( x – x0 ) n –1 , g ′′( x) = ∑ n (n –1) an ( x – x0 ) n –2 etc.
n =0 n =0
The differentiated series converges on |x – x0 | < r.
In calculus there are certain tests by which one could determine an interval of converge of
a real power series. A simple one and one which is frequently used is known as ratio test.

un+1
The series ∑ ui converges absolutely if = k < 1.
Lt
n→∞
n =0 un

xn
Example 1 : For the power series ∑ ,
n =0 n

un+1 x n +1
n+1 n
Lt
n→∞ = Lt
n→∞ = Lim
n→∞ ⋅ x =| x |
un xn
n
n +1
Hence the series converges absoulately if | x | < |.

Example 2 : For the power series

x 2 x 4 x6 (–1) n –1 2n –2
1– + – + .... + x + ....,
2! 4! 6! (2n – 2)!

(–1) n –1 2 n –2 (–1) n 2 n
un = x and un+1 = x
(2n – 2)! (2 n)!
Therefore
un+1 (–1) n x 2 n (2n – 2)! x2
Lim
n→∞ = Lim
n→∞ × = Lim
n→∞ = 0 for each x
un (2n )! (–1) n –1 x 2 n –2 2n(2n –1)
Hence the series converges absoulately for all x. Its interval of convergence is the entire
real axis.

Theorem 2.3.1 : (Existence theorem)


Let x0 be a real number and suppose that the coefficients a1, a2, a3,..... an in
L( y ) = y (n ) + a1 ( x) y (n –1) + a2 ( x ) y (n –2) + .... + an ( x ) y

have convergent power series expansions in powers of (x – x0) on an interval | x – x0 |< r , r > 0.
If a1, a2, a3,.....an are any n constants, there exists a solution f of the problem
L( y ) = 0, y ( x0 ) = α1 , y′( x0 ) = α 2 ,...., y (n –1) ( x0 ) = α n
Differential Equations (83)
with a power series expansion

φ ( x) = ∑ ck ( x – x0 ) k
k =0
is convergent for | x – x0 |< r . We have

k ! ck = α k +1 (k = 0,1, 2,3,...., n –1), and ck for k ³ n may be computed in terms of


c0, c1, c2, c3,..... cn – 1 by substituting the series into L(y) = 0.
If the coefficients a1, a2, a3,..... an are analytic at x0 then the solutions are also analytic.
The solutions can be computed by a formal algebraic process.
Illustration :
L( y ) = y′′ – xy = 0
Here a1 ( x) = 0, a2 ( x ) = – x are analytic for all real x.
Let the solution of the equation L(y) = 0 be f defined by

φ ( x) = a0 + a1 x + a2 x 2 + a3 x 3 + .... = ∑ an x n
k =0

Then φ ′( x ) = a1 + 2a2 x + 3a3 x 2 + 4a4 x 3 + .... = ∑ n an x n –1
n=1

φ ′′( x ) = 2a2 + 6a3 x + 12a4 x 2 + .... = ∑ n (n –1)an x n –2
n=0
∞ ∞
φ ′′( x ) – x φ ( x ) = ∑ n (n –1)an x n –2 – x ∑ an x n
n =2 n =0

= 2a2 + 6a3 x + 12a4 x 2 + 20a5 x 3 + ... – {a0 x + a1x 2 + a2 x 3 + a3 x 4 + ...}


= 2a2 + (6a3 – a0 ) x + (12a4 – a1 ) x 2 + (20a5 – a2 ) x 3 + ...

= 2a2 + ∑ [(n + 2) (n + 1) an+ 2 – an –1 ] x n
n=1
f is a solutin of L( y ) = y′′ – xy = 0 if φ ′′ – x φ = 0 or

2a2 + ∑ [(n + 2) (n + 1) an +2 – an –1 ] x n = 0
n =1

Above equation is true only if all the coefficients of the power series of x are zero. Thus,
2a2 = 0, (n + 2) (n + 1) an+ 2 – an –1 = 0, n = 1, 2, 3....
This gives an infinite set of equations, and can be solved for an. Thus, for n = 1 we have
a0
(3) ⋅ (2) ⋅ a3 = a0 or a3 =
(3) ⋅ (2)
For n = 2 we find
a1
(4) ⋅ (3) a4 = a1 or a4 =
(4) ⋅ (3)
Differential Equations (84)
Continuing in this way we see that
a0; a1 a2 =0
a a1 a2
a3 = 0 ; a4 = ; a5 = =0
(3)⋅(2) (4)⋅(3) (5)⋅(4)
a a0 a a1
a6 = 3 = ; a7 = 4 = ; a8 =0
(6)⋅(5) (6)⋅(5)⋅(3)⋅(2) (7)⋅(6) (7)⋅(6)⋅(4)⋅(3)
a a0 a a1
a9 = 6 = ; a10 = 7 = ; a11 =0.
(9)⋅(8) (9)⋅(8)⋅(6)⋅(5)⋅(3)⋅ (2) (10)⋅(9) (10)⋅(9)⋅(7)⋅ (6)⋅ (4)⋅ (3)
In general
a0
a3m = ;
(2) ⋅ (3) ⋅ (5) ⋅ (6) ⋅ (8) ⋅ (9) ⋅⋅⋅ (3m – 1) (3m )
a1
a3m +1 = ; a3m + 2 = 0
(3) ⋅ (4) ⋅ (6) ⋅ (7) ⋅ (9) ⋅ (10) ⋅⋅⋅ (3m) (3m + 1)
Thus all the constants are determined in terms of a0 and a1. Collecting together terms
containing a0 and a1 as a factor we have
 x3 x6   x4 x7 
φ ( x ) = a0 1 + + + ... + a1  x + + + ...
 (3) ⋅ (2) (2) ⋅ (3) ⋅ (5) ⋅ (6)   (4) ⋅ (3) (3) ⋅ (4) ⋅ (6) ⋅ (7) 
Let f1 and f2 represent the two series in the brackets.
∞ x 3m
Thus, φ1 ( x ) = 1 + ∑ ,
m =1 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9....(3m –1) (3m)

∞ x 3m+1
φ2 ( x ) = x + ∑ .
m =1 3 ⋅ 4 ⋅ 6 ⋅ 7 ⋅ 9 ⋅ 10....(3m) (3m + 1)

We have shown, in a formal way that f satisfies y ′′ – xy = 0 for any two contants a0 and a1
In particular the choice a0 = 0 and a1 = 1 implies f2 (x) satisfies the equation and a0 = 1, a1
= 0 implies f1 (x) satisfies the equation.
The only question that remains is about the convergence of the series, defining f1 (x) and
f2 (x).
∞ x 3m
φ1 ( x ) = 1 + ∑ d m ( x ) = 1 + ∑
m =1 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9....(3m –1) (3m)

d m +1 x 3m + 3 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9 ⋅⋅⋅ (3m –1) (3m)


= ×
dm 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9 ⋅⋅⋅ (3m) (3m + 2) (3m + 3) x 3m
x3
=
(3m + 2) (3m + 3)
1 1
Lt sup
m →∞ = Lt inf
m →∞ = 0.
(3m + 2) (3m + 3) (3m + 2) (3m + 3)
The series converges if | x | < ¥.
Similarly f2 (x) is convergent series.

Differential Equations (85)


EXAMPLES

1. Find two linearly independent power series solutions (in powers of x) of the following
equations.
(a) y ′′ − xy ′ + y = 0 (b) y ′′ + 3x 2 y ′ – xy = 0
(c) y′′ – x 2 y = 0 (d) y′′ + 3x 3 y′ + x 2 y = 0

Ans. (a) : Let φ ( x ) = a0 + a1x + a2 x 2 + a3 x 3 + ... + an x n + ... = ∑ an x n be a solution of
n =0
L ( y ) = y′′ – xy′ + y = 0 . Since it is a solution it satisfies the equation L(f ) = 0.

φ ( x) = a0 + a1x + a2 x 2 + a3 x3 + ... + an x n + ... = ∑ an x n
n=0

Then φ ′ ( x) = a1 + 2a2 x + 3a3 x 2 + ... + n an x n –1 + ... = ∑ n an x n –1
n =1

φ ′′ ( x ) = 2a2 + 3 ⋅ 2a3 x + ... + n (n –1) an x n –2 + ... = ∑ n (n –1) an x n –2
n=2
∞ ∞ ∞
Thus, L(φ ) = ∑ n (n –1) an x n –2 – x ∑ n an x n –1 + ∑ an x n = 0.
n=2 n=1 n =0
∞ ∞ ∞
= ∑ (n + 2) (n + 1) an+ 2 x n – ∑ n an x n + ∑ an x n = 0
n =0 n=1 n =0

= (2 a2 + a0 ) + ∑ {(n + 2) (n + 1) an+2 – nan + an }x n = 0
n=1

L(φ ) = (2a2 + a0 ) + ∑ {(n + 2) (n + 1) an+ 2 – (n –1)an }x n = 0
n=1
We see that L(f ) = 0 if and only if 2a2 + a0 = 0 and (n + 2) (n + 1) an+ 2 – (n –1) an = 0 for
1 (n –1) an
n = 1, 2,3,.... a2 = – a0 ; an + 2 = is called recurrence relation.
2 (n + 2) (n + 1)
a0 ; a1
– a0
a2 = ; ; a3 = 0.a1
2 ⋅1
a a 2
a4 = 2 = – 0 ; a5 = .0 = 0
4⋅3 2⋅3⋅ 4 5⋅ 4
3a 3a0
a6 = 4 = – ; a7 = 0
6⋅5 2 ⋅ 3 ⋅ 4 ⋅ 5⋅ 6
3 ⋅ 5 a0
a8 = – ; a9 = 0
2 ⋅ 3⋅ 4 ⋅ 5⋅ 6⋅ 7 ⋅ 8
3 ⋅ 5 ⋅ 7 a0
a10 = – ; a11 = 0
10!
Differential Equations (86)
In general a2n+1 = 0 n = 1, 2,3,....
3 ⋅ 5 ⋅ 7 ⋅ 9 ⋅⋅⋅ (2n – 3) a0
a2 n = –
(2n)!
2 ⋅ 3 ⋅ 4 ⋅ 5 ⋅ 6 ⋅ 7 ⋅ 8 ⋅ 9 ⋅⋅⋅ (2n – 3) (2n – 2) (2n –1) (2n)
=– a0
2 ⋅ 4 ⋅ 6 ⋅ 8 ⋅⋅⋅⋅ 2n (2n –1) (2n)!
(2n) !
=– n
a0
2 n !(2n – 1) (2n)!
a0
=– n
2 n !(2n –1)
∞ ∞
φ ( x ) = ∑ a2n x 2n + ∑ a2n+1 x 2n+1
n =0 n =0
∞ a0
= a0 – ∑ n
x 2 n + a1x
n =1 2 n !(2n –1)
 ∞
x 2n 
= a0 1 – ∑ n  + a1 x
 n =1 2 n !(2n –1) 
∞ x 2n
φ1 ( x) = 1– ∑ and φ2 ( x) = x are two solutions of the equation
n=1 2 n n !(2n –1)

Let φ1 ( x ) = ∑ d m ( x )
m =0

x 2(n+1)
d n +1 2n+1 (n + 1)!(2n + 1) x 2 (2n –1)
= =
dn x 2n 2(n + 1) (2n + 1)
2 n n !(2n –1)
d n +1 Lt (2 n – 1)
Lt
n→∞ = n→∞ =0
dn 2(n + 1) (2n + 1)
Redius of convergence = ¥
The series converges if | x | < ¥ i.e. all values of x. Both the solutions are convergent for
all values of x.


Ans. (b) : Let φ ( x ) = ∑ an x n be a solution.
n =0

∞ ∞
φ ′ ( x) = ∑ nan x n –1 , φ ′′( x) = ∑ n (n – 1)an x n –2
n=1 n=2

L(φ ) = φ ′′ + 3 x φ ′ – xφ
2

Differential Equations (87)


∞ ∞ ∞
= ∑ n(n –1)an x n –2 +3x 2 ∑ n an x n –1 – x ∑ an x n
n=2 n=1 n =0
L(φ ) = 0 implies
2 ⋅ 1a2 + 3 ⋅ 2a3 x + 4 ⋅ 3a4 x 2 + ⋅⋅⋅ + 3  a1x 2 + 2a2 x 3 + 3a3 x 4 + 4a4 x 5 + ....
 
–  a0 x + a1 x 2 + a2 x3 + a3 x 4 + .... = 0.

2 ⋅ 1a2 + (3 ⋅ 2 a3 – a0 ) x + [4 ⋅ 3a4 + (3 – 1)a1 ] x 2 + [5 ⋅ 4a5 + (3(2) – 1)a2 ] x 3

– [6.5a6 + (3(3) –1)a3 ] x 4 + ⋅⋅⋅⋅ = 0



2 ⋅1a2 + (3 ⋅ 2a3 – a0 ) x + ∑ [(n + 3) (n + 2) an+3 + (3n –1) an ] x n+1 = 0
n=1
a0 (3n –1) an
Then a2 = 0 ; a3 = ; an +3 = –
2⋅3 (n + 3) (n + 2)
a0 ; a1 ; a2 = 0
a0 2 a1
a3 = – ; a4 = – ; a5 = 0
2⋅3 4⋅3
8 a3 8 a0 11 a4 11 ⋅ 2 a1
a6 = – =+ ; a7 = – = ; a8 = 0
6⋅5 2 ⋅ 3⋅ 5⋅ 6 7 ⋅ 6 7 ⋅ 6 ⋅ 4⋅ 3
(18 –1) (9 –1)a0 20 ⋅11 ⋅ 2 a1
a9 = – ; a10 = – ; a11 = 0.
9 ⋅8 ⋅ 6 ⋅ 5 ⋅ 3⋅ 2 10 ⋅ 9 ⋅ 7 ⋅ 6 ⋅ 4 ⋅ 3
The solution
∞ (–1) m (–1)8 ⋅17 ⋅⋅⋅ (9m – 1)
φ1 ( x) = 1 + ∑ x 3m
m =1 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9 ⋅⋅⋅⋅ (3m –1) (3m)

∞ (–1) m 2 ⋅11 ⋅ 20 ⋅⋅⋅ (3(3m – 2) –1) 3m+1


φ2 ( x) = x + ∑ x
m =1 3 ⋅ 4 ⋅ 6 ⋅ 7 ⋅ 9 ⋅10 ⋅⋅⋅⋅ (3m) (3m + 1)


Ans. (c) : Let φ ( x) = ∑ an x be a solution of y′′ – x 2 y = 0. Since it is a solution f (x) satisfies
n
n= 0

L(φ ) = φ ′′ – x 2φ = 0.
∞ ∞ ∞
φ ( x) = ∑ an x n , φ ′ ( x) = ∑ n an x n –1 , φ ′′ ( x) = ∑ n (n –1) an x n –2
n= 0 n =1 n= 2

∞ ∞
L(φ ) = ∑ n(n –1) an x n –2 – x 2 ∑ an x n = 0.
n= 2 n= 0

∞ ∞
2 ⋅1a2 + 3 ⋅ 2 a3 x + ∑ n (n –1)an x n –2 – ∑ an x n+2 = 0
n= 4 n= 0

Differential Equations (88)


∞ ∞
2 ⋅1a2 + 3 ⋅ 2 a3 x + ∑ (n + 4) (n + 3) an+ 4 x n+ 2 – ∑ an x n+ 2 = 0
n= 0 n= 0

Here we have replaced n by n + 4 and therefore the sum is from 0 to ¥.



2 ⋅1 a2 + 3 ⋅ 2 a3 x + ∑ [(n + 4) (n + 3) an+ 4 – an ] xn+ 2 = 0.
n= 0

an
Thus, a2 = 0, a3 = 0 and an + 4 = .
(n + 4) (n + 3)
a0 a1 a2 = 0 a3 = 0
a0 a1
a4 = ; a5 = ; ; a6 = 0 ; a7 = 0
3⋅ 4 5⋅4
a a0 a a1
a8 = 4 = ; a9 = 5 = ; a10 = 0 ; a11 = 0
8 ⋅ 7 3⋅ 4 ⋅7 ⋅8 9 ⋅8 4 ⋅5 ⋅8 ⋅9
a0 a1
a12 = ; a13 = ; a14 = 0 ; a15 = 0
3 ⋅ 4 ⋅ 7 ⋅ 8 ⋅ 11⋅ 12 4 ⋅ 5 ⋅ 8 ⋅ 9 ⋅ 12 ⋅ 13
M M M M
Thus all the coefficients an’s are determined in terms of a0 and a1 since a2 = a3 = 0 implies
a4m+2 and a4m+3 = 0 for m = 0, 1, 2, 3,..... Therefore
∞ ∞
φ ( x) = ∑ a4m x m + ∑ a4m +1 x 4 m+1
m=0 m=0

 ∞
x 4m 
= a0 1 + ∑ 
 m = 1 3 ⋅ 4 ⋅ 7 ⋅11 ⋅12 ⋅⋅⋅ (4m –1) (4m) 
 ∞
x 4m +1 
+ a1  x + ∑ 
 m = 1 4 ⋅ 5 ⋅ 8 ⋅ 9 ⋅12 ⋅ 13 ⋅⋅⋅ (4m) (4m + 1) 
Therefore two linearly independent solutions are

x 4m
φ1 ( x ) = 1 + ∑ and
m = 1 3 ⋅ 4 ⋅ 7 ⋅11⋅12 ⋅13 ⋅⋅⋅ (4m –1) (4m)

∞ x 4 m+1
φ2 ( x ) = x + ∑ .
m = 1 4 ⋅ 5 ⋅ 8 ⋅ 9 ⋅12 ⋅13 ⋅⋅⋅ (4m) (4m + 1)


Ans. (d) : Let φ2 ( x) = ∑ an xn be a solution of L( y ) = y′′ + x3 y′ + x 2 y = 0. Therefore
n =0
∞ ∞ ∞
L(φ ) = ∑ n (n –1) an x n –2 + x3 ∑ n an x n –1 + x 2 ∑ an xn = 0
n=2 n =1 n=0

∞ ∞ ∞
that is ∑ n (n –1) an x n –2 + ∑ n an x n + 2 + ∑ an x n+ 2 = 0.
n=2 n =1 n=0

Differential Equations (89)


The first term starts from x0 where as last two series start from x3 and x2 respectively. To
get the common base we write the expansion in the following form
∞ ∞ ∞
2 ⋅1a2 + 3 ⋅ 2a3 x + 4 ⋅ 3a4 x 2 + ∑ n (n –1) an x n –2 + ∑ n an x n+ 2 + a0 x 2 + ∑ an x n+2 = 0.
n=5 n =1 n =1
Therefore
∞ ∞
2 ⋅1a2 + 3 ⋅ 2a3 x + (4 ⋅ 3a4 + a0 ) x 2 + ∑ n (n –1) an x n –2 + ∑ (n + 1) an x n+ 2 = 0
n=5 n =1

We replace n by n + 4 in the first series.



2 ⋅1a2 + 3 ⋅ 2a3 x + (4 ⋅ 3a4 + a0 ) x 2 + ∑ [(n + 4) (n + 3) an+4 + (n + 1) an ] x n+ 2 = 0.
n =1

Above equation is true for all values of x and therefore


a2 = 0; a3 = 0; 4 ⋅ 3 a4 + a0 = 0; (n + 4) (n + 3) an+4 + (n + 1) an = 0

a0 a1 a2 = 0 a3 = 0
– a0 –2 a1
a4 = ; ; a5 = ; a6 = 0 ; a7 = 0
3⋅ 4 5⋅ 4
5 a0 6 ⋅ 2 a1
a8 = + ; a9 = ; a10 = 0 ; a11 = 0
3⋅ 4 ⋅ 7 ⋅8 4 ⋅5 ⋅8 ⋅9
9 ⋅ 5 a0 10 ⋅ 6 ⋅ 2 a1
a12 = – ; a13 = – ; a14 = 0 ; a15 = 0
3 ⋅ 4 ⋅ 7 ⋅ 8 ⋅ 11⋅ 12 4 ⋅ 5 ⋅ 8 ⋅ 9 ⋅ 12 ⋅ 13
M M M M
(–1) m 5 ⋅ 9 ⋅ 13 ⋅⋅⋅ (4m – 3)
a4m = ;
3 ⋅ 4 ⋅ 7 ⋅ 8 ⋅ 11⋅ 12 ⋅⋅⋅ (4m –1) (4m)
(–1) m 2 ⋅ 6 ⋅ 10 ⋅⋅⋅ (4m – 2)
a4m +1 = ; a4 m + 2 = a4m +3 = 0.
4 ⋅ 5 ⋅ 8 ⋅ 9 ⋅⋅⋅⋅(4 m) (4m + 1)
Therefore two linearly independent solutions are
∞ (–1) m 5 ⋅ 9 ⋅ 13 ⋅⋅⋅ (4m – 3)
φ1 ( x ) = 1 + ∑ x 4m
m =1 3 ⋅ 4 ⋅ 7 ⋅ 8 ⋅ 11⋅ 12 ⋅⋅⋅ (4m –1) (4m)
∞ (–1) m 2 ⋅ 6 ⋅ 10 ⋅⋅⋅ (4m – 2) 4 m+1
φ2 ( x ) = x + ∑ x
m =1 4 ⋅ 5 ⋅ 8 ⋅ 9 ⋅⋅⋅ (4m) (4m + 1)


2. Find the solution f of y ′′ + ( x –1) 2 y ′ – ( x –1) y = 0 in the form φ ( x) = ∑ ak ( x – 1) k which
k =0
satsfies φ (1) = 1, φ ′(1) = 0.


Ans. : Let φ ( x) = ∑ ak ( x –1) k be a solution of L( y ) = y′′ + ( x –1) 2 y′ – ( x –1) y = 0.
k =0

Differential Equations (90)


∞ ∞ ∞
L(φ ) = ∑ k (k –1) ak ( x –1) k –2 + ( x –1) 2 ∑ k ak ( x –1) k –1 – ∑ ak ( x –1) k +1 = 0.
k =2 k =1 k =0
∞ ∞ ∞
that is ∑ k (k –1) ak ( x –1) k –2 + ∑ k ak ( x – 1) k +1 – ∑ ak ( x –1)k +1 = 0.
k =2 k =1 k =0
∞ ∞
2 ⋅1a2 + 3 ⋅ 2a3 ( x –1) + ∑ k (k –1) ak ( x –1)k –2 + ∑ k ak ( x –1) k +1
k =4 k =1

– a0 ( x –1) – ∑ ak ( x –1) k +1 = 0
k =1

In the third term replace k by k + 3 we get



2 ⋅1a2 + [3 ⋅ 2a3 – a0 ] ( x –1) + ∑ [(k + 3) (k + 2) ak +3 + (k –1) ak ] x k +1 = 0.
k =1
Thus,
a0 ; a1 ; a2 = 0
a0
a3 = ; a4 = 0 ; a5 = 0
3⋅ 2
2 a3 –2 a0
a6 = – = ; a7 = 0 ; a8 = 0
6 ⋅5 2 ⋅3 ⋅5⋅ 6
–5a6 5 ⋅ 2 a0
a9 = = ; a10 = 0 ; a11 = 0.
9 ⋅ 8 2 ⋅ 3⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9
M M M
(–1) m 2 ⋅ 5 ⋅ 8 ⋅⋅⋅ (3m –1) a0
a3m = ; a3m+1 = 0 for m = 1, 2... ; a3m+ 2 = 0.
2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9 ⋅⋅⋅ (3m –1) (3m)
Corresponding to the coefficients a0 and a1 we get the following two linearly independent
solutions.
∞ (–1) m 2 ⋅ 5 ⋅ 8 ⋅⋅⋅ (3m – 1)
φ1 ( x ) = 1 + ∑ ( x – 1)3m and
m =1 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9 ⋅⋅⋅ (3m –1) (3m)
φ1 ( x) = ( x –1)
The general solution f is
 ∞ (–1) m 2 ⋅ 5 ⋅ 8 ⋅⋅⋅ (3m –1) 
φ ( x) = a0 1 + ∑ ( x –1)3m  + a1 ( x –1)
 m = 1 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅⋅⋅ (3m –1) (3m) 
φ1 (1) = 1 give a0 = 1
∞ (–1) m 2 ⋅ 5 ⋅ 8 ⋅⋅⋅ (3m –1) (3m)
φ ′ ( x) = a0 ∑ ( x –1)3m –1 + a1
m =1 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅⋅⋅ (3m –1) (3m)

φ ′(1) = 0 give a1 = 0.

Differential Equations (91)


Thus, the required solution is
∞ (–1) m 2 ⋅ 5 ⋅ 8 ⋅⋅⋅ (3m –1)
φ ( x) = 1 + ∑ ( x – 1)3m .
m =1 2 ⋅ 3 ⋅ 5 ⋅ 6 ⋅⋅⋅ (3m –1) (3m )

3. Compute the solution f of y ′′′ – xy = 0 which satisfies φ (0) = 1, φ ′(0) = 0, φ ′′(0) = 0.


Ans. : Let φ ( x) = ∑ an x n be a solution of L( y ) = y′′′ – xy = 0. Then
k =0
∞ ∞
L(φ ) = ∑ n (n –1) (n – 2) an x n –3 – x ∑ an xn = 0
n= 3 n= 0
∞ ∞
Then 3 ⋅ 2 ⋅1 a3 + ∑ n ( n –1) (n – 2) an xn –3 – ∑ an xn +1 = 0
n= 4 n= 0
In the first sum replace n by n + 4, then

3 ⋅ 2 ⋅ 1 a3 + ∑ [ ( n + 4) ( n + 3) ( n + 2) an+4 – an ] x n +1 = 0
n= 0

an
Thus, a3 = 0 and an + 4 = .
(n + 4) (n + 3) (n + 2)
a0 a1 ; a2 ; a3 = 0
a0 a1 a2
a4 = ; a5 = ; a6 = ; a7 = 0
4 ⋅ 3⋅ 2 5⋅ 4⋅ 3 6⋅ 5⋅ 4
a0 a1 a2
a8 = ; a9 = ; a10 = ; a11 = 0
8⋅ 7 ⋅ 6 ⋅ 4 ⋅3⋅ 2 9 ⋅8 ⋅7 ⋅5 ⋅ 4 ⋅3 10 ⋅ 9 ⋅ 8 ⋅ 6 ⋅ 5 ⋅ 4
a0 a1 a2
a12 = ; a13 = ; a14 =
2 ⋅ 3 ⋅ 4 ⋅ 6 ⋅ 7 ⋅ 8 ⋅10 ⋅11 ⋅12 3 ⋅ 4 ⋅ 5 ⋅ 7 ⋅ 8 ⋅ 9 ⋅11 ⋅12 ⋅13 4 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9 ⋅ 10 ⋅ 12 ⋅13 ⋅ 14
;a15 = 0
M M M M
a0 a1
a4 m = ; a4m +1 =
2 ⋅ 3 ⋅ 4 ⋅ 6 ⋅ 7 ⋅ 8 ⋅⋅⋅ (4m – 2) (4m –1) (4m) 3 ⋅ 4 ⋅ 5 ⋅⋅⋅ (4m –1) (4m) (4m + 1)
a2
; a4 m + 2 = ; a4 m +3 = 0
4 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9 ⋅ 10 ⋅⋅⋅ (4m) (4m + 1) (4m + 2)
The general solution f (x) of the given equation contains three parameters a0, a1, a2.
The solution f (x) becomes
 ∞ x 4m 
φ ( x) = a0 1 + ∑ 
 m = 1 2 ⋅ 3 ⋅ 4 ⋅ 6 ⋅ 7 ⋅ 8 ⋅⋅⋅ (4 m – 2) (4 m –1) (4 m) 
 ∞
x 4 m+1 
+ a1  x + ∑ 
 m = 1 3 ⋅ 4 ⋅ 5 ⋅ 7 ⋅ 8 ⋅ 9 ⋅⋅⋅ (4m –1) (4m) (4m + 1) 

Differential Equations (92)


 ∞
x 4 m+ 2 
+ a2  x 2 + ∑ 
 m = 1 4 ⋅ 5 ⋅ 6 ⋅ 8 ⋅ 9 ⋅ 10 ⋅⋅⋅ (4m) (4m + 1) (4m + 2) 

φ (0) = 1 gives a0 = 1
∞ 4mx (4 m –1)
φ ′ ( x) = a0 ∑
m =1 2 ⋅ 3 ⋅ 4 ⋅ 6 ⋅ 7 ⋅ 8 ⋅⋅⋅ (4m – 2) (4m –1) (4m)
 ∞ (4m + 1) x 4 m 
+ a1 1 + ∑ 
 m=1 3 ⋅ 4 ⋅ 5 ⋅ 7 ⋅ 8 ⋅ 9 ⋅⋅⋅ (4m –1) (4m ) (4m + 1) 
 ∞ (4m + 2) x 4 m+1 
+ a2  2 x + ∑ 
 m =1 3 ⋅ 4 ⋅ 5 ⋅ 7 ⋅ 8 ⋅ 9 ⋅⋅⋅ (4m –1) (4m) (4m + 1) 

φ ′ (0) = 0 gives a1 = 0.
Similarly φ ′′ (0) = 0 gives a2 = 0.
Thus, the required solution is
∞ x 4m
φ ( x) = 1 + ∑ .
m =1 2 ⋅ 3 ⋅ 4 ⋅ 6 ⋅ 7 ⋅ 8 ⋅⋅⋅ (4m – 2) (4m –1) (4m)

4. Legendre equation is an important differential equation occur in physical problems. The


equation
L( y ) = (1– x 2 ) y ′′ – 2 xy′ + α (α + 1) y = 0
where a is constant is called Legendre equation.
If we write this equation as
2x α (α + 1)
y ′′ – y ′+ y = 0,
1– x 2 1 – x2
we see that a1, a2 are given by
2x α (α + 1)
a1 ( x) = 2
and a2 ( x) = .
1– x 1 – x2
1 ∞
Both these functions are analytic at x = 0. Indeed, 2
= ∑ x 2 k and the series converges
1– x k =0
for | x | < |.
Thus, a1(x) and a2(x) have the series expansions. Both there series converge for | x | < | .
Thus by existence theorem the solution L(y) = 0 on | x | < | have convergent power series
expansions.
Let f be any solution of L(y) = 0 on | x | < |.

Suppose φ ( x) = ∑ an xn then
k =0
∞ ∞ ∞
L(φ ) = (1 – x 2 ) ∑ n (n –1)an x n –2 – 2 x ∑ nan x n –1 + α (α + 1) ∑ an x n
n=2 n =1 n=0

Differential Equations (93)


∞ ∞ ∞ ∞
= ∑ n (n –1)an x n –2 – ∑ n (n –1) an x n – 2 ∑ nan x n + α (α + 1) ∑ an x n
n=2 n=2 n =1 n=0

= ∑ [ (n + 2) (n + 1)an+2 – {n (n –1) + 2n – α (α + 1)} an ] x n
n=0

For f to satisfy L(f ) = 0 we must have all the coefficients of the powers of x equal to zero.

Hence, (n + 2) (n + 1) an +2 – [n (n + 1) – α (α + 1)] an = 0, n = 0,1, 2,3,...


This is a recurrance relation which gives an +2 in terms of an.
n (n + 1) – α (α + 1)
an + 2 = an
(n + 1) (n + 2)
– (α + n + 1) (α – n)
= an .
( n + 1) ( n + 2)
for n = 0 we get
α (α + 1)
a2 = – a0
2
(α + 2) (α –1)
a3 = – a1
2⋅3
Similarly,
(α + 3) (α – 2) (α + 4) (α – 3)
a4 = – a2 ; a5 = – a3
3⋅ 4 4⋅5
α (α + 1) (α + 3)(α – 2) (α + 4) (α + 2) (α –1) (α – 3)
=+ a0 =+ a1
2 ⋅ 3⋅ 4 2 ⋅ 3⋅ 4 ⋅ 5
In general
(α + 2m –1) (α + 2m – 3) ⋅⋅⋅ (α + 1)α (α – 2) ⋅⋅⋅ (α – 2m + 2)
a2 m = (–1) m a0
(2m)!
(α + 2m) (α + 2m – 2) ⋅ ⋅ ⋅ (α + 2) (α – 1) (α – 3) ⋅ ⋅ ⋅ (α – 2m + 1)
a2m +1 = (–1) m a1
(2m + 1)!
All the coefficients are determined in terms of a0 and a1 and we have
φ ( x ) = a0 φ1 ( x) + a1 φ2 ( x)
∞ (–1) m (α + 2m –1)(α + 2m – 3) ⋅⋅⋅α (α – 2) ⋅⋅⋅ (α – 2m + 2) 2 m
where, φ1 ( x ) = 1 + ∑ x
m =1 (2m)!
∞ (–1) m (α + 2m)(α + 2m – 2) ⋅⋅⋅ (α + 2) (α –1)(α – 3) ⋅⋅⋅ (α – 2m + 1) 2m+1
φ2 ( x ) = x + ∑ x
m =1 (2m + 1)!
Both f1 and f2 are solutions of Legendre equation, corresponding to the choices
c0 = 1, c1 = 0 and c0 = 0, c1 = 1,
respectively. They form a basis for the solutions, since
φ1 (0) = 1, φ 2 (0) = 0 ; φ1′ (0) = 0, φ2′ (0) = 1

Differential Equations (94)


φ1 (0) φ2 (0)
∴ W(φ1, φ2 ) (0) = = φ1 (0) φ2′ (0) – φ2 (0) φ1′ (0) = 1
φ1′ (0) φ2′ (0)

Since Wronkian W(φ1, φ2 ) ≠ (0), φ1, φ2 are linearly independent and therefore forms a basis.
If a is a non-negativ even integer a = 2n, then f1 has only a finite number of non-zero
terms. In this case f1 is a polynomial of degree 2n containing only even powers of x. for example,
α = 0, φ1 ( x ) = 1 = p0 ( x )
(–1) (2 + 1)
α = 2, φ1 ( x) = 1 + α x 2 = 1– 3 x 2 = p2 ( x)
2
or the recurrance relation

n( n + 1) – α (α + 1)
an+ 2 = an implies
(n + 1) (n + 2)

0(0) – 2 (3)
a2 = a0 = –3a0
1⋅ 2
2(3) – 2 (3)
a4 = a2 = 0
3⋅ 4
with a0 = 1 we get φ1 ( x ) = 1 – 3 x 2 = p2 ( x )

0(0) – 4(5)
for a = 4, a2 = a0 = –10a0
1⋅ 2
2(3) – 4(5)
a4 = a2
3⋅ 4
6 – 20
= (–10 a0 )
12
140
=+ a0
12
35
= a0
3
4(5) – 4 (5)
a6 =
5⋅ 6
=0
 3 5 4
φ1 ( x) = a0 1 – 10 x 2 + x  with a0 = 1
 3 
35 4
φ1 ( x) = 1–10 x 2 + x = p4 ( x)
3
The solution f2 is not a polynomial in this case since none of the coefficients in the series
of f2 vanish.
A similar situation occurs when a is a positive odd integer n. Then f2 is a polynomial of
degree n having only odd powers of x and f1 is not a polynomial.

Differential Equations (95)


for example
n ( n + 1) – α (α + 1)
(a = 1) an + 2 = an
( n + 1) ( n + 2)
1(2) –1(2)
a3 = a1 = 0
2⋅3
φ2 ( x) = x = p1 ( x) (say)
1(2) – 3(4)
(a = 3) a3 = a1
2⋅3
5
= – a1
3
3(4) – 3(4)
a5 = a3 = 0
4⋅5
5
φ2 ( x) = x – x 3 = p3 ( x) (say)
3
1(2) – 5 ⋅ 6
(a = 5) a3 = a1
2 ⋅3
14
= – a1
3
3(4) – 5 ⋅ 6
a5 = a3
4⋅5
18  14 
=–  –  a1
20  3 
21
= a1
5
5⋅6 – 5⋅6
a7 = a5 = 0.
6⋅7
14 21
φ2 ( x) = x – x3 + x 5 = p5 ( x) (say)
3 5
Definition : 2.1.3

A polynomial solution pn of degree n of (1 – x ) y ′′ – 2 xy ′ + n (n + 1) = 0,


2

Satisfying Pn (1) = 1 is called the nth Legendre polynomial and the differential equation is
called Legendre equation.
Let f be a polynomial of degree n defined by
dn 2
φ ( x) = n
( x –1) n
dx
u ( x ) = ( x 2 –1)n implies u ′( x) = n ( x 2 –1) n –1 2 x gives
( x 2 –1) u′( x) – 2n x u ( x) = 0

Differential Equations (96)


Differentiate this equation (n + 1) times.
First differentiation gives
( x 2 –1) u ′′( x ) + 2 x (1 – n) u ′( x ) – 2n u ( x ) = 0
Second differentiation gives
( x 2 –1) u ′′′( x ) + 2 x [(1 + 1) – n ]u ′′ + 2 [(1 – n ) + (0 – n ) ]u′( x ) = 0
Third differentiation gives
( x 2 –1) u ( iv ) + 2 x [(1 + 1 + 1) – n ]u ′′′ + 2 [(2 – n) + (1– n) + (0 – n)]u ′′ = 0.

i.e. ( x 2 –1) u ( iv ) + 2 x (3 – n) u ′′′ – 2 [(n – 2) + (n –1) + n ]u ′′ = 0.


In general (n + 1)th differentiation gives

( x 2 –1) u ( n + 2) + 2 x ((n + 1) – n) u ( n +1) – 2 [1 + 2 + 3 + 4 + ... + (n – 1) + n ]u ( n ) = 0.

i.e. ( x 2 –1) u ( n+ 2) + 2 x u ( n+1) – n (n + 1) u (n ) = 0. or

(1 – x 2 )u (n+ 2) – 2 x u (n+1) + n (n + 1) u (n ) = 0.
dn 2 dn
Since φ ( x) = n
( x –1) n
= n
u ( x) = u ( n) ( x),
dx dx
(1– x 2 ) φ ′′ – 2 x φ ′( x) + n (n + 1) φ ( x) = 0
dn 2
Thus the function φ ( x) = n
( x – 1) n is a solution of Legendre equation.
dx
dn 2 dn 
( x –1) n
= n 
( x + 1) n ( x –1) n 
dx n
dx 

 d n   d n –1  d
=  n  ( x –1)n   ( x + 1)n +  n –1 ( x – 1)n  ( x + 1)n + ...
 
 dx   dx  dx
= n (n –1) (n – 2)....2 ⋅1( x + 1) n + terms containing (x – 1) as factor.
= n ! (x + 1)n + terms containing (x – 1) as factor.
Thus, at x = 1,
dn 2
n
( x –1) n = 2 n ⋅ n !
dx
1 1 dn 2
Define Pn ( x) = n
φ ( x ) = n n
( x –1) n then Pn (x) is a solution of Legendre
2 n! 2 n ! dx
1
equation witha = n Pn (1) = n  2 n ⋅ n ! = 1. Thus, Pn(x) is a Legendre polynomial of degree n.
2 n!  

Supposey is a polynomial solution of Legendre equation with a = n. Since f1 and f2 are


basic solutions of Legendre equation ψ = c1 φ1 + c2 φ2 on | x | < | for some constants c1 and c2 is
a solution. If n is even f1 is polynomial solution and f2 is not a polynomial ψ – c1 φ1 is polynomial

Differential Equations (97)


where as c2 f2 is not a polynomial and therefore c2 = 0. In particular the function Pn satisfies
Pn ( x ) = c1 φ1 ( x ) for some constant c1 if n is even. Since Pn (1) = 1 = c1 φ1 (1) therefore φ1 (1) ≠ 0.
Thus no nontrivial polynomial solution of Legendre equation can be zero at x = 1. A similar
result is valid for n odd.
The formula
dn 2 1
Pn ( x ) =
n n
( x –1) n
2 n ! dx
is known as Rodriges formula. This expression can be used to prove properties of Legendre
polynomials.

EXERCISES

1. The equation (1 – x 2 ) y ′′ – xy′ + α 2 y = 0 where a is a constant is called the Chebyshev


equation.
(a) Compute two linearly independent series solutions for | x | < |.
(b) Show that for every non negative integer a = n there is a polynomial solution of
degree n.

2. The equation y ′′ – 2 xy ′ + 2α y = 0, where a is a constant, is called the Hermite equation.


(a) Find two linearly independent solutions on – ¥ < x < ¥.
(b) Show that there is a polynomial solution of degree n, in case a = n is a non-negative
integer.

3. Find the general solution valid near the origin

(i) y ′′ + 3xy ′ + 3 y = 0 (ii) (1 + 4 x 2 ) y′′ – 8 y = 0

(iii) (1 + x 2 ) y′′ – 4 xy′ + 6 y = 0 (iv) 2 y′′ + xy′ – 4 y = 0

(v) y ′′ + x 2 y = 0

Answers :

∞ (–α 2 ) (22 – α 2 ) ⋅⋅⋅  (2m – 2)2 – α 2 


1. (a) φ1 ( x ) = 1 + ∑   x 2m
m =1 (2m)!

∞ (12 – α 2 ) (32 – α 2 ) ⋅⋅⋅ (2m –1) 2 – α 2 


φ2 ( x ) = x + ∑   x 2m +1
m =1 (2m + 1)!

(b) f1 is a polynomial if a is an even integer,


f2 is a polynomial if a is an odd integer.

Differential Equations (98)


∞ 2m (–α ) (2 – α ) ⋅⋅⋅ (2m – 2 – α ) 2m
2. (a) φ1 ( x ) = 1 + ∑ x
m =1 (2m)!

∞ 2m (1– α ) (3 – α ) ⋅⋅⋅ (2m –1 – α ) 2m+1


φ2 ( x ) = x + ∑ x
m =1 (2m + 1)!
(b) f1 is a polynomial if a is an even integer, f2 is a polynomial if a is an odd integer.

 ∞ (–3) k x 2k 
3. (i) y ( x ) = a0 1 + ∑ k 
 k = 1 2 k ! 

 ∞ (–3)k x 2k +1 
+ a1  x + ∑ 
 m =1 3 ⋅ 5 ⋅ 7, ⋅⋅⋅(2k + 1) 

∞ (–1) k +1 2 2k x 2 k +1
(ii) y = a0 (1 + 4 x 2 ) + a1 ∑
k =0 4k 2 –1

 x3 
(iii) y ( x ) = a0 (1– 3 x 2 ) + a1  x – 
 3 

 1  ∞ 3(–1)k x 2 k +1
(iv) y ( x) = a0 1 + x 2 + x 4  + a1 ∑ 2 k
 12  k = 0 2 k !(2k – 3) (2k –1) (2k + 1)

 ∞ (–1) k x 4k 
(v) y ( x ) = a0 1 + ∑ 2 k 
 k = 1 2 k ! 3 ⋅ 7 ⋅ 11⋅⋅⋅ (4k –1) 

 ∞ (–1) k x 4 k +1 
+ a1  x + ∑ 2k 
 k = 1 2 k ! 5 ⋅ 9 ⋅13 ⋅⋅⋅ (4k + 1) 

Differential Equations (99)


Chapter
3
Linear Equations with
Regular Singular Points

Contents :
Unit 1 : Euler equation
Unit 2 : Second order equations with regular singular points
Unit 3 : The Bessel equation
Unit 4 : Regular singular points at infinity

Introduction
For a linear differential equation a0 ( x) y ( n) + a1 ( x) y ( n –1) + a2 ( x) y ( n–2) + .... + an ( x) y = 0,
where the coefficient functions a0, a1, a2, a3,..... an are analytic at some point x0, the point x0 is
called an ordinary point of the equation if a0 (x0) ¹ 0. In the last chapter we have obtain power
series solutions valid near an ordinary point of a linear equation.
A singular point of the above linear equation is any point x = x1 for which a0 (x1) = 0. In
this chapter we shall get power series solutions valid near a certain kind of singular points of the
equation. It is usually difficult to determine the nature of the solutions in the vicinity of singular
points. However there is a large class of equations for which the singularity is rather weak in the
sense that slight modification of the methods used for solving equations with analytic coefficients
discussed in chapter II unit 3, serve to yield solutions near the singularities.
Definition 3.1.1 (a)
A point x = x0 is a regular singular point of L( y ) = a0 ( x ) y ( n ) + a1 ( x) y ( n –1) + a2 ( x) y ( n –2)
+ ⋅⋅⋅ + an ( x) y = 0 if the equation can be written in the form L( y ) = ( x – x0 ) n y ( n) + b1 ( x)
( x – x0 )( n –1) y ( n –1) + ⋅⋅⋅ + bn ( x ) y = 0 where b1, b2, b3,..., bn are analytic at x0.
If the functions b1, b2, b3,..., bn can be written in the form
bk ( x ) = ( x – x0 ) k β k ( x ) k = 1, 2, 3,.......,n
Where b1, b2, b3,..., bn are analytic at x0 then L(y) = 0 becomes

y ( n ) + β1 ( x ) y ( n –1) + β 2 ( x) y ( n –2) + ..... + β n ( x ) y = 0

Differential Equations (100)


Definition 3.1.1 (b)
A equation of the form c0 ( x) ( x – x0 ) n y ( n ) + c1 ( x) ( x – x0 )n –1 y ( y –1) + c2 ( x) ( x – x0 )( n –2)
y ( n –2) + .... + cn ( x) y = 0 has a regular singular point at x0 if c0, c1, c2, c3,..... cn are analytic at
x = x0 and c0(x0) ¹ 0.
Definition 3.1.2
If x = x0 is a singular point but is not a regular singular point, then it is called irregular
singular point. For example, consider the equation
3
x 2 y ′′ – y′ –
y = 0.
4
The origin x = 0is a singular point but not regular therefore x = 0 is irregular singular
point. The coefficient of y¢ is not of the form xb1(x) where b1(x) analytic.
In the first unit we study the differential equation that has a regular singular point at origin
and all the analytic functions b1, b2, b3,..., bn are constants.

Unit 1 : The Euler Equation


The simplest example of a second order equation that follows defination 3.1.1(a) is the
Euler equation
L( y) = x 2 y′′ + a x y′ + b y = 0
where a, b are constants.

Theorem 3.1.1
Consider the second order Euler equation
L( y) = x 2 y′′ + a x y′ + b y = 0 (a, b constants),
and the polynomial q given by
q ( r ) = r ( r –1) + ar + b
A basis for the solutions of the Euler equation on any interval not containing x = 0 is given
by
φ1 ( x ) = | x |r1 , φ2 ( x ) = | x |r2 ,
in case r1, r2 are distinct roots of q and by
φ1 ( x) = | x |r1 , φ 2 ( x) = | x |r1 log| x |,
if r1 is a root of equation q of multiplicity two.
Proof :
Case 1 : r1 ¹ r2
(a) We first consider the equation for x > 0. Let xr be a solution of Euler equation
L( y) = x 2 y′′ + a x y′ + b y = 0
L( x r ) = x 2  r (r – 1) x r –2  + a x  r x r –1  + bx r = 0
   

Differential Equations (101)


implies [r (r –1) + ar + b]  x r  = 0
q is a polynomial defined by q ( r ) = r ( r – 1) + a r + b
Thus, we have
L ( x r ) = q (r ) x r
If r1 is a root of q(x) then q(r1) = 0 and therefore L( x r1 ) = 0. i.e. φ1 ( x) = x r1 is a solution of
L(y) = 0. If r2 is another root of q and r2 ¹ r1 then φ2 ( x ) = x r2 is another solution of
L(y) = 0. Thus, φ1 ( x) = x r1 and φ2 ( x) = x r2 is a basis for the solution of the Euler equation
as f1 and f2 are linearly independent.
(b) If x < 0, Let (–x)r be a solution (if x < 0, –x > 0).

 (– x )r  ′ = – r (– x )r –1 ,  (– x) r  ′′ = r (r –1) (– x )r –2
   

x  (– x )r  = r (– x ) (– x )r –1 = r (– x )r and
 
L( y) = r (r –1)(– x) r + a r (– x)r + b (– x) r = q (r )(– x) r
if r1 ¹ r2 then

φ1 ( x) = (– x) r1 , φ2 ( x) = (– x) r2 are solutions of L(y) = 0. If r1 and r2 are complex


roots of q(r) = 0, we define xr for r complex by

x r = e r log x ( x > 0)
r r
then ( x r )′ = (r ⋅ log x )′ ⋅ e r log x = ⋅ x = r x r –1 and the result follows on the same lines
x
for complex roots also.
Thus, we have proved that if (x > 0) r1 ¹ r2 φ1 ( x) = x r1 and φ2 ( x) = x r2 are solutions of
L(y) = 0 and for x < 0, r1 ¹ r2 we have φ1 ( x ) = (– x) r1 and φ2 ( x ) = (– x )r2 are solution of L(y) = 0.
Since | x | = x for x > 0 and | x | = – x for x < 0 φ1 ( x) =| x |r1 and φ2 ( x ) =| x |r2 are solutions of
L(y) = 0 if r1, r2 are distinct roots of q(r) = 0.
We prove that f1 and f2 are linearly independent.

Let c1φ1 + c2 φ 2 = 0 i.e. c1 | x |r1 + c2 | x |r2 = 0 then c1 + c2 | x |r2 – r1 = 0 for every x ÎR.
Differentiating above equation w.r.t. x for x > 0 or x < 0 we get,
c2 (r2 – r1 )| x |r2 – r1 = 0
But r1 ¹ r2 and x ¹ 0 therefore c2 = 0 and c1φ1 + c2 φ 2 = 0 for all x implies c1 = 0 since
c2 = 0 and φ1 ( x) ≠ 0 .
Thus, f1 and f2 are linearly independent solutions.

Therefore if r1 and r2 are distinct roots of q(x) = 0, then φ1 ( x ) =| x |r1 and φ2 ( x ) =| x |r2
forms a basis for the solutions of L(y) = 0.
Differential Equations (102)
Case 2 : r1 = r2
(a) x > 0 : If r1 = r2 then q(r1) = 0 and q¢ (r1) = 0. We have proved that if r1 is a root of q(x) = 0
then φ1 ( x) = x r is a solution. To construct second solution consider
∂ ∂
L( x r ) =  q (r ) x r 
∂r ∂r  
= [q′(r ) + q (r ) log x ] x r
∂ r
Since, x = x r log x
∂r
But if r1 = r2 = r then q (r) = 0 and q¢ (r) = 0 and we have
∂ 
L( x r )  = 0 .
∂r  
∂   ∂ 
L( x r ) = L  x r  = L( x r log x)
∂r    ∂r 
Thus, L( x r log x) = 0 implies xr log x is a solution of L(y) = 0.
If r1 is a root of q (r) = 0 of multiplicity two then φ1 ( x) = x r1 and φ2 ( x) = x r1 log x are two
solutions of L(y) = 0.

(b) x < 0 : If x < 0, then – x > 0 and φ1 ( x ) = (– x ) r1 and φ2 ( x ) = (– x )r1 log (– x ) are solution of
L(y) = 0.
Thus φ1 ( x ) = | x |r1 and φ2 ( x ) = | x |r1 log | – x | are two solution of L(y) = 0.
c1φ1 + c2φ 2 = 0 implies c1 + c2 log | x | = 0 for all x and therefore c1 = c2 = 0 and f1, f2 are
linearly independent.
Thus if r1 is a repeated root of q(r) = 0 then φ1 ( x ) =| x |r1 and φ2 ( x) =| x |r1 log | x | is a
basis for solutions of the Euler equation L( y) = x 2 y′′ + a xy′ + by = 0.
Illustration :
x 2 y ′′ + xy′ + y = 0 for x ¹ 0 is Euler equation with a = b = 1.

The polynomial q(r ) = r ( r –1) + r + 1 = r 2 + 1 and r = +i, – i are roots of q(r). A basis for
the solutions by theorem 3.1.1 are
φ1 ( x) =| x |i and φ2 ( x) =| x |– i ( x ≠ 0)

| x |i = ei log|x| = cos(log | x |) + i sin(log | x |)


Thus ψ 1 ( x ) = cos (log | x |) and ψ 2 ( x ) = sin (log | x |) is another basis for solution of
L ( y ) = x 2 y ′′ + xy ′ + y = 0.

Theorem 3.1.2
Consider the Euler equation of order n.
L( y ) = x n y (n ) + a1x n –1 y (n –1) + a2 x n –2 y (n –2) + ... + an y = 0,
Differential Equations (103)
where a1, a2, a3,..., an are constants. Let r1, r2,..., rs be distinct roots of the indicial polynomial
q( r ) = r ( r – 1) (r – 2)...( r – n + 1) + a1 r ( r –1)...( r – n + 2) + ... + an and suppose
ri has multiplicity mi. Then the n functions
| x |r1 , | x |r1 log | x |,....,| x |r1 (log | x |) m1 –1; | x |r2 ,| x |r2 log | x |,...,| x |r2 (log | x |) m2 –1;...;
| x |rs , | x |rs log | x |,...., | x |rs (log | x |) ms –1
form a basis for the solution of L(y) = 0 on any interval not containing zero.
Proof : Let | x |r be a solution of L(y) = 0.
(| x |r )′ = r | x |r –1 , (| x |r )′′ = r (r –1) | x |r –2 ,....
(| x |r )( n) = r (r – 1) ( r – 2)...(r – n + 1)| x |r – n
Hence, L(| x |r ) = r (r – 1) (r – 2)...(r – n + 1)| x |r + a1r (r –1) (r – 2)
...(r – n + 2) | x |r +... + an | x |r
= q ( r ) | x |r
where q (r ) = (r ) (r – 1) (r – 2)...(r – n + 1) + r (r –1) (r – 2)...(r – n + 2)a1 + ... + an .
The polynomial q(r) is called indicial polynomial. Thus, | x |r is a solution of L(y) = 0 if
q (r) = 0 i.e. if r is a root of indicial polynomial then | x |r is a solution of L(y) = 0.
Differentiating L(| x |r ) = q(r ) | x |r with respect to ‘r’ we get
∂  ∂ 
L (| x |r ) = L  | x |r 
∂r  ∂r 
= ( q ′( r ) + q ( r ) log | x | ) | x |r
In general k times differentiation gives
∂k  ∂k r
L (| x |r
) = L  | x | 
∂r k  ∂r
k

=  q( k ) (r ) + kq ( r –1) (r ) log | x | +k (k –1)q( k –2) ( x) (log | x |)2 + ... + q (r ) (log | x |)k  | x |r .
 
If r is a root of q (r) with multiplicity (k + 1) then q ( r ) = 0, q′( r ) = 0, q ′′( r ) = 0...,
 ∂i r
q(k ) (r ) = 0 and therefore L  i | x |  = 0 for i = 1, 2,3,...., k .
 ∂r 
∂ i
Thus φ ( x) = i | x | , i = 1, 2,3,...., k. are solution of L(y) = 0.
r
∂r
If r1 is a root of q(r) of multiplicity m1 then
∂ ∂2 ∂ m1 –1
| x |r1 , | x |r1 =| x |r1 log | x |, 2 | x |r1 =| x |r1 (log| x |) 2 ,..., m –1 | x |r1
∂r1 ∂r1 ∂r1 1

= | x |r1 (log | x |) m1 –1 are solutions of L(y) = 0.


Repeating this process for each root of q(r) we obtain all the solution and the result follows.
All these solutions are linearly independent and therefore form a basis for the solutions of
L(y) = 0 on any interval not containing zero.
Differential Equations (104)
EXAMPLES

Q. 1. Find all solutions of the following equations for x > 0


(a) x 2 y′′ + 2 xy′ – 6 y = 0 (b) 2 x 2 y′′ + xy ′ – y = 0
(c) x 2 y′′ + xy ′ – 4 y = 0 (d) x 2 y′′ – 5 xy ′ + 9 y = x3
(e) x3 y′′′ + 2 x 2 y′′ – xy '+ y = 0.
Ans.:
(a) The indicial equation
q( r ) = r (r –1) + 2 r – 6 = r 2 + r – 6 has root r = 3, – 2.
Therefore φ1 ( x) = x 3 and φ2 ( x ) = x –2 are basic solutions and φ ( x) = c1 x3 + c2 x –2
is general solution for constants c1, c2.
(b) The indicial equation 1

q( r ) = 2r (r –1) + r –1 = 2 r – r –1 has root r = 1, –
2 1
2
and φ1 ( x ) = x, φ2 ( x ) = x 2
1

are basic solution, φ ( x) = c1 x + c2 x 2 is general solution for constants c1, c2.

(c) The indicial equation


q( r ) = r (r – 1) + r – 4 = r 2 – 4 has root 2, – 2 Then φ1 ( x) = x 2 and φ2 ( x ) = x –2 are
basic solution, φ ( x) = c1 x 2 + c2 x –2 is general solution.

(d) The indicial equation


q( r ) = r (r –1) – 5r + 9 = r 2 – 6r + 9 has root 3, 3. Since the root 3 is repeated root
of multiplicity two φ1 ( x) = x 3 and φ2 ( x) = x3 log x are basic solution of
corresponding homogeneous equation x 2 y′′ – 5xy′ + 9 y = 0.
The particular solution will be determined by using variation of constant method.
Let ψ = u1 ( x ) φ1 ( x) + u2 ( x ) φ2 ( x) be a solution of equation
5 9
y ′′ – y ′ + 2 y = x then
x x
Wk (t ) b (t ) dt
uk ( x ) = ∫ Here b(t) = t,
W(φ1 , φ 2 )

x3 x 3 log x
W (φ1, φ2 ) = = x5 ,
3x 2
x + 3x log x
2 2

0 x 3 log x
W1 = = – x 3 log x,
1 x + 3x log x
2 2

Differential Equations (105)


x3 0
W2 = = + x3,
2
3x 1

– x3 log x ⋅ x log x 1
u1 ( x ) = ∫ 5
dx = – ∫ dx = – (log x ) 2
x x 2

x3 ⋅ x dx dx
u2 ( x ) = ∫ 5
= ∫ = log x
x x
ψ ( x ) = u1 ( x ) φ1 ( x ) + u2 ( x ) φ2 ( x )
1 1
= – (log x) 2 x 3 + (log x) x 3 log x = x 3 (log x) 2
2 2
The general solution
1
φ = c1φ1 + c2φ2 + ψ = c1 x 3 + c2 x 3 log x + x 3 (log x) 2 .
2
(e) The indicial equation
q( r ) = r (r –1) (r – 2) + 2r ( r – 1) – r + 1

= (r –1)  r 2 – 2r + 2r –1 = (r –1) (r 2 –1) has root 1, 1, –1.


 
Since one is a root of multiplicity two, φ1 ( x) = x, φ2 ( x) = x log x and corresponding
to –1, φ3 ( x ) = x –1 .
The general solution
φ ( x) = c1x + c2 x log x + c3 x –1.

Q. 2. Find all solutions of the following equations for | x | > 0.

(a) x 2 y ′′ + xy′ + 4 y = 1 (b) x 2 y′′ – 3xy ′ + 5 y = 0


(c) x 2 y′′ + xy′ – 4π y = x
Ans.:
(a) The indicial equation q(r ) = r ( r –1) + r + 4 has root r = ± 2 i .

Since both the roots are distinct, φ1 ( x) =| x |2i and φ2 ( x) =| x | –2i . The general
solution of homogeneous equation is
φ ( x) = c1 | x |2i + c2 | x |–2i
The particular solution will be calculated by variation of constant method.
Case 1 : x > 0,
If x > 0 then | x | = x
φ1 ( x ) = x 2i and φ2 ( x ) = x –2i
Let ψ ( x ) = u1 ( x ) φ1 ( x ) + u 2 ( x) φ2 ( x ) be a solution of

Differential Equations (106)


W1 ( x ) b ( x) dx
x 2 y ′′ + xy′ + 4 y = 1. then u1 ( x ) = ∫ and
W(φ1 , φ 2 )
W2 ( x) b ( x) 1
u2 ( x ) = ∫ dx where b ( x) = 2 ,
W(φ1 , φ 2 ) x

x 2i x –2i 4i
W(φ1, φ2 ) = = –2i x –1 – 2i x –1 = –
2i x 2i –1 –2ix –2i –1 x

0 x –2i x 2i 0
W1 ( x) = =–x –2i
; W2 = = x 2i
1 –2ix –2i –1 2i x 2i –1 1

 1 
– x –2i  2  –2i –2+1
 x  dx = x x –2 i x –2i
u1 ( x ) = ∫ ∫ 4i dx = =
 4i  4i (–2i) 8
 – 
 x
 1 
x 2i  2  2i –2+1
 x  dx = x x2i x 2i
u2 ( x ) = ∫ ∫ – 4i dx = = .
 4i  – 4i (+2i ) 8
– 
 x
x –2i 2i x 2i –2i 1
Thus, ψ ( x ) = u1 ( x ) φ1 ( x ) + u2 ( x )φ 2 ( x ) = ⋅x + ⋅ x = . For x > 0,
8 8 4
1
φ ( x ) = c1x 2i + c2 x +
–2i
is a solution of given equation.
4
Case 2 : x < 0

If x < 0 then | x |= – x and φ1 ( x) = (– x) 2i and φ2 ( x) = (– x) –2i


Let ψ ( x ) = u1 ( x ) φ1 ( x ) + u 2 ( x) φ2 ( x ) be a solution of the given differential equation.

(– x )2i (– x ) –2i 2i 2i 4i
W(φ1, φ2 ) = = + =
–2i(– x) 2i –1
+2i(– x) –2i –1 (– x ) (– x ) – x

0 (– x) –2i (– x) 2i 0
W1 = = –(– x) –2 i
, W2 = = (– x) +2 i ,
–2i –1 2i –1
1 2i (– x ) –2i (– x ) 1

1
b ( x) =
x2
1
–(– x ) –2 i ⋅
W1 ( x) b ( x) (+ x ) 2 1 (– x ) –2i (– x ) –2i
u1 ( x ) = ∫ dx = ∫ dx = – = .
W(φ1, φ2 ) 4i 4i 2i 8
(– x )

Differential Equations (107)


(– x) 2 i
W ( x) b ( x) (– x )2 1 (– x )+2i (– x )+2 i
u2 ( x ) = ∫ 2 dx = ∫ dx = =
W(φ1, φ2 ) 4i 4 i –2 i 8
(– x)
ψ ( x ) = u1 ( x ) φ1 ( x ) + u 2 ( x) φ2 ( x )
(– x ) –2i (– x ) 2i 1
ψ ( x) = ⋅ (– x ) 2i + (– x ) –2i = ⋅
8 8 4
1
for x < 0, φ ( x) = c1 (– x) 2i + c2 (– x) –2i + is a solution of the given differential equation.
4
1
Thus φ ( x) = c1 | x |2i + c2 | x | –2i + is a solution of the given differential equation if x ¹ 0.
4
(b) The indical equation q ( r ) = r ( r – 1) – 3r + 5 has roots 2 + i, 2 – i. Since both the roots are

distinct φ1 ( x ) =| x |2+ i and φ2 ( x ) =| x |2–i are two independent solutions. The general
solution

(
φ = c1φ1 ( x ) + c2 φ2 ( x) = c1 | x |2+i + c2 | x |2–i = x 2 c1 | x |i +c2 | x |– i . )
(c) The indical equation q ( r ) = r ( r – 1) + r – 4π has roots 2 π and –2 π . Since both the
π π
roots are distinct φ1 ( x) =| x |2 and φ2 ( x) =| x |–2 are two solutions. The general
π π
solution of corresponding homogeneous equation is φ ( x) = c1 | x |2 + c2 | x |–2 .
We solve the non-homogeneous equation using the variation of constants method.
Case 1 : x > 0
π π π π
If x > 0 then | x |2 = x2 , | x | –2 = x –2
Let ψ ( x ) = u1 ( x ) φ1 ( x ) + u 2 ( x) φ2 ( x ) be a solution of given equation then
π π
x2 x –2 –2 π 2 π – 4 π
W(φ1, φ2 ) = = – =
π –1 π –1 x x x
2 π x2 –2 π x –2
π π
0 x –2 –2 π x2 0 π
W1 ( x) = = –x , W2 ( x ) = = x2
–2 π –1 +2 π –1
1 –2 π x +2 π x 1

, since the given equation is y ′′ + 1 y ′ – 4π y = 1 .


1
b ( x) =
x x x x
–2 π  1 
–x  
W ( x) b ( x)  x  dx = + 1 x –2 π –1+1dx = + 1 x –2 π +1
u1 ( x ) = ∫ 1 dx = ∫ ∫
W(φ1 , φ2 ) –4 π 4 π 4 π (–2 π + 1)
x
x –2 π +1
Thus, u1 ( x ) = +
4 π (1– 2 π )
Differential Equations (108)
1 π
x2 ⋅2 π +1
W ( x) b ( x) x dx = – 1 x
u2 ( x ) = ∫ 2 dx = ∫
W(φ1, φ 2 ) –4 π ⋅
1 4 π 2 π +1
x
x –2 π +1 π 1 x 2 π +1
ψ ( x) = + ⋅ x2 –
4 π (1 – 2 π ) 4 π 2 π +1
x  1 1  x 4 π x
=+  – = ⋅ =
4 π 1 – 2 π 2 π + 1  4 π 1 – 4π 1– 4π
π π x
For x > 0, φ ( x) = c1 x + c2 x –2 +
2
is a solution of given equation. For x < 0
1– 4π
x
also we get ψ ( x ) = .
1– 4π
π x π
Thus the general solution of the given equation is φ ( x) = c1 | x |2 + c2 | x |–2
. +
1– 4π
Till now we have considered Euler equation having a regular singular point at origin. At
the beginning of this chapter we defined singular points, regular singular points and irregular
singular points. We present some definitions of singularities which can be used to classify the
singularities of the given differential equation.
Definition 3.1.3 (a) :
A second order differential equation
y ′′ + p ( z ) y ′ + q ( z ) y = 0,
analytic for 0 < | z – z0 | < r, has a regular singular point at z0 when p(z) has at worst a simple pole
at z = z0 and q(z) has at worst a double pole at z = z0.
Definition 3.1.3 (b) :
For a second order differential equation
y ′′ + p ( x ) y ′ + q ( x ) y = 0,
if x = x0 is a singular point and if the denominator of p (x) does not contain the factor (x – x0) to
a power higher than one and if the denominator of q (x) does not contain the factor (x – x0) to a
power higher than two, then x – x0 is called a regular singular point.

EXAMPLES

Q. 1. Classify the singular points, in the finite plane, of the equation


x ( x –1)2 ( x + 2) y′′ + x 2 y′ – ( x3 + 2 x –1) y = 0
Ans.: a0 ( x) = x( x –1) 2 ( x + 2) = 0 gives x = 0, 1, –2
Thus the singular points in a finite plane are at x = 0, 1, –2.
Given equation can be written as
x2 ( x3 + 2 x –1)
y′′ + y ′ – y =0
x( x –1) 2 ( x + 2) x( x –1) 2 ( x + 2)
Differential Equations (109)
x –( x3 + 2 x –1)
therefore p ( x) = and q ( x ) =
( x –1)2 ( x + 2) x ( x –1)2 ( x + 2)
Since the denominator of p(x) does not contain the factor (x – 0) and the denominator of
q (x) does not contain a factor (x – 0)p for p > 2 . Hence, x = 0 is a regular singular point. Now
consider x = 1. Since the denominator of p(x) contains the factor (x – 1)p where p = 2 > 1
therefore x = 1 is not a regular singular point i.e. x = 1 is irregular singular point.
At x = –2, the factor (x + 2) appears to the first power in the denominator of p(x) which is
not higher than 1 and the factor (x + 2) appears to the first power in the denominator of q(x)
which is not higher than 2. so x = – 2 is a regular singular point.

Q. 2. Classify the singular points in the finite plane for the equation
x 4 ( x 2 + 1) ( x –1) 2 y′′ + 4 x3 ( x –1) y′ + ( x + 1) y = 0

Ans.: a0 ( x) = x 4 ( x 2 + 1)( x –1) 2 = 0 gives x = 0, x = ± i, x = 1 are roots of a0 (x) = 0.


Thus, the singular points in a finite plane are at x = 0, + i, – i, 1.
Given equation is of the form
4 ( x + 1)
y ′′ + y′ + 4 2 y = 0.
x( x + 1) ( x – 1)
2
x ( x + 1) ( x – 1) 2
4 x +1
Here p ( x) = and q ( x ) = 4 2
x ( x + 1) ( x –1)
2
x ( x + 1) ( x –1)2
(i) x = 0
The denominator of p(x) contains a factor (x – 0)r where r = 1 >/ 1 and the denominator
of q(x) contains a factor (x – 0)r where r = 4 > 2. Therefore x = 0 is an irregular
singular point.
(ii) x = i
The denominator of p(x) contains a factor (x – i)r where r = 1 >/ 1 and the denominator
of q(x) contains a factor (x – i)r where r = 1 >/ 2. Therefore x = i is a regular singular
point.
(iii) x = – i
The denominator of p(x) contains a factor (x + i)r where r = 1 >/ 1 and the denominator
of q(x) contains a factor (x + i)r where r = 1 >/ 2. Therefore x = –i is a regular
singular point.
(iv) x = 1
The denominator of p(x) contains a factor (x – 1)r where r = 1 >/ 1 and the denominator
of q(x) contains a factor (x – 1)r where r = 2 >/ 2. Therefore x = 1 is a regular singular
point.
Thus x = i, – i, 1 are regular singular points and x = 0 is an irregular singular point.

Differential Equations (110)


Q. 3. For each equation, locate and classify all its singular points in the finite plane.
(a) x 3 ( x –1) y′′ + ( x –1) y′ + 4 xy = 0. (b) x 2 ( x 2 – 4) y′′ + 2 x3 y′ + 3 y = 0
(c) y′′ + xy = 0 (d) x 2 ( x – 4) 2 y ′′ + 3x y ′ – ( x – 4) y = 0
Ans.:

(a) a0 ( x) = x 3 ( x –1), a0 ( x) = 0 give x = 0, x = 1. Therefore x = 0 and x = 1 are singularities.


Given equation can be put in the form
1 4
y ′′ + 3
y′ + 2 y =0
x x ( x –1)
for x = 0, denominator of p (x) contains a factor xr where r = 3 > 1 and therefore x = 0
is irregular singular point. For x = 1, denominator of p(x) contains a factor (x – 1)r
where r = 0 >/ 1 and the denominator of q(x) contains a factor (x – 1)r where r = 1 >/ 2.
Therefore x = 1 is a regular singular point.

(b) a0 ( x) = x 2 ( x 2 – 4) = x 2 ( x + 2) ( x – 2) . a0 (x) = 0 gives x = 0, 2, –2. Therefore 0, 2, –2


are singular points. Given equation is
2x 3
y ′′ + y′ + 2 y = 0.
( x + 2) ( x – 2) x ( x + 2) ( x – 2)
For x = 0, the denominator of p (x) contains a factor xr where r = 0 >/ 1 and denominator
of q(x) contains a factor xr for r = 2 >/ 2. Therefor x = 0 is a regular singular point.
For x = 2, the denominator of p (x) contains a factor ( x – 2)r for r = 1 >/ 1 and the
denominator of q(x) contains a factor ( x – 2)r for r = 1 >/ 2. Therefor x = 1 is a regular
singular point.
For x = – 2, the denominator of p (x) contains a factor ( x + 2)r for r = 1 >/ 1 and the
denominator of q(x) contains a factor ( x + 2)r for r = 1 >/ 2. Therefor x = –2 is a regular
singular point.
Thus, all the singular points are regular.
(c) a0 (x) = 1 ¹ 0 for any x therefore equation do not have any finite singular point.

(d) a0 ( x) = x 2 ( x – 4)2 . a0 ( x) = 0 gives x = 0, 4. x = 0, 4 are singular point of the given


equation. Given equation is
3 1
y ′′ + y ′– y=0
x ( x – 4) 2 x 2 ( x – 4)
For x = 0, the denominator of p (x) contains a factor xr for r = 1 >/ 1 and the denominator
of q(x) contains a factor xr for r = 2 >/ 2. Therefor x = 0 is a regular singular point.
For x = 4, the denominator of p (x) contains a factor ( x – 4)r for r = 2 > 1 therefor x =
4 is not a regular singular point.
Thus x = 0 is regular and x = 4 is irregular singular point.

Differential Equations (111)


EXERCISE

1. For each equation, locate and classify all its singular points in the finite plane
(a) x 2 y′′ + y = 0 (Ans.: x = 0 is regular, no irregular)

(b) ( x 2 + 1)( x – 4)2 y′′ + ( x – 4) 2 y′ + y = 0 (Ans.: x = i, – i regular, x = 4 irregular)

(c) x 2 ( x – 2) y′′ + 3( x – 2) y′ + y = 0 (Ans.: x = 2 is regular, x = 0 irregular)


i
(d) (1 + 4 x 2 ) 2 y ′′ + 6 x (1 + 4 x 2 ) y ′ – 9 y = 0 (Ans.: x = ± are regular)
2
2. Find all solutions of the following equations.
(a) x 2 y′′ + 2 xy′ – 12 y = 0 (Ans.: y = c1x 3 + c2 x –4 )

(b) x 2 y′′ + xy′ – 9 y = 0 (Ans.: y = c1x 3 + c2 x –3 )


x
(Ans.: x = c1x + c2 x –
2 –2
(c) x 2 y′′ + xy′ – 4 y = x )
3
(d) x 2 y′′ – 3 xy′ + 4 y = 0 (Ans.: y = x 2 (c1 + c2 ln x ) )

(e) x 2 y′′ + 5 xy′ + 5 y = 0 (Ans.: y = x –2 [c1 cos(ln x ) + c2 sin(ln x ) ] )

3. Find all solutions of the following equations.


(a) x 2 y′′ – 5xy′ + 9 y = 0 (Ans.: y =| x |3 (c1 + c2 ln | x |) )
1 2
(b) 9 x y ′′ + 2 y = 0
2 (Ans.: y = c1 | x |3 + c2 |x| 3 )
1
(c) 2 x 2 y′′ – 3xy′ + 2 y = 0 (Ans.: y = c1 | x |2 + c2 | x | 2 )

Unit 2 : Second order equation with Regular Singular Points


A second order equation with a regular singular point at x0 has the form
L( y ) = ( x – x0 ) 2 y ′′ + ( x – x0 ) a ( x) y′ + b ( x ) y = 0,
where a(x), b(x) are analytic functions at x0 i.e. they have power series expansions
∞ ∞
a( x) = ∑ α k ( x – x0 )k and b( x) = ∑ β k ( x – x0 )k
k=0 k=0
which are convergent on some interval |x – x0| < r0 for some r0 > 0.
Without loss of generality we assume x0 = 0. Then
L( y ) = x 2 y ′′ + x a ( x) y′ + b( x ) y = 0 and
∞ ∞
a( x) = ∑ α k xk , b( x) = ∑ β k xk which are convergent on an interval
k=0 k=0
| x | < r0 , r0 > 0. The Euler equation is a particular case of L(y) = 0 with a, b constants.
Differential Equations (112)
A second order equation with regular singular point has a power series solution. If functions
a (x), b (x) have power series expansion on some interval | x | < r0 then the power series solution
converges on the interval | x | < r0.

Theorem 3.2.1
Consider the equation
x 2 y′′ + a ( x ) xy′ + b ( x ) y = 0,
where a and b have convergent power series expansions for | x | < r0, r0 > 0. Let r1, r2
(Re r1 ³ Re r2 ) be the roots of the indicial polynomial
q ( r ) = r ( r –1) + a (0) r + b (0)
for 0 < | x | < r0 there is a solution f1 of the form

φ1 ( x ) = | x |r1 ∑ ck x k (c0 = 1),
k =0

where the series converges for | x | < r0. If r1 – r2 is not zero or a positiove integer, there is a
second solution f2 for 0 < | x | < r0 of the form
∞ : :
φ2 ( x) = | x |r2 ∑ ck xk (c0 = 1),
k =0
where the series converge for | x | < r0.
:
The coefficients ck , ck can be obtained by substitution of the solution into the differential
equation.
Proof :
Suppose we have a solution f of the form

φ ( x ) = x r ∑ ck x k (c0 ≠ 0, x > 0)
k =0

for the equation L( y ) = x 2 y′′ + a ( x) xy ′ + b ( x ) y = 0.


∞ ∞
Where a ( x) = ∑ α k x k and b ( x) = ∑ β k x k for | x | < r0. Then
k =0 k =0
∞ ∞
φ ( x ) = x r ∑ ck x k = ∑ ck x k + r ,
k =0 k =0
∞ ∞
φ ′( x) = ∑ (k + r ) ck x k +r –1 = x r –1 ∑ (k + r )ck x k ,
k =0 k =0
∞ ∞
φ ′′( x ) = ∑ (k + r ) (k + r –1) ck x k +r –2 = x r –2 ∑ (k + r ) (k + r –1)ck x k
k =0 k =0

 ∞  ∞ 
b ( x ) φ ( x ) =  ∑ β k x k   x r ∑ ck x k 
 k =0   k =0 
  
Differential Equations (113)
∞ : : k
= x r ∑ β k x k where β k = ∑ c j β k – j
k =0 j =0

 ∞  ∞ 
xa ( x ) φ ′( x) = x  ∑ α k x k   x r –1 ∑ (k + r ) ck x k 
 k =0  k =0 
 ∞  ∞ 
= x r  ∑ α k x k   ∑ ( k + r ) ck x k 
 k =0   k =0 
  
∞ : : k
= xr ∑ α k x k where α k = ∑ ( j + r ) c j α k – j
k =0 j =0

x 2φ ′′( x) = x r ∑ (k + r ) (k + r –1) ck x k .
k =0
∞ ∞ : ∞ :
Thus, L(φ ) ( x) = x
r
∑ (k + r ) (k + r –1) ck xk + x r ∑ α k x k + x r ∑ β k x k
k =0 k =0 k =0
∞  : : 
= xr ∑ (k + r ) ( k + r –1) ck + α k + β k  x k
k =0  
 : : 
L(f ) = 0 implies [ ]k =  ( k + r ) ( k + r –1) ck + α k + β k ) = 0
 
: : k = 0, 1, 2, 3,....
Using the definitions of α k , β k we can write [ ]k as
k k
[ ]k = (k + r ) (k + r –1) ck + ∑ ( j + r) c j αk– j + ∑ c j βk– j
j =0 j =0
k –1
= [(k + r )(k + r –1) + (k + r ) α 0 + β 0 ]ck + ∑ ( j + r ) α k – j + β k – j  c j
j =0
for k = 0 we must have
r ( r –1) + r α 0 + β 0 = 0.
Since c0 ¹ 0 the second degree polynomial q given by
q (r ) = r (r –1) + r α 0 + β 0
is called the indicial polynomial and the only admissible values of r are the roots of q.
[ ]k = q ( r + k )c k + d k = 0 (k = 1, 2,3,....) ......(3.2.1)
k –1
where dk = ∑ ( j + r )α k – j + β k – j  c j (k = 1, 2,3,....) ......(3.2.2)
j =0

Note that dk is a linear combination of c0, c1, c2,..... ck–1 with coefficients involving the
known functions a, b and r. Leaving r and c0 indeterminant for the moment we solve equations
(3.2.1) and (3.2.2) successively in terms of c0 and r. The solutions we denote by Ck (r) and the
corresponding dk by Dk(r). Thus,
D (r)
D1 (r ) = (r α1 + β1 )c0 , C1 (r ) = – 1 ,
q (r + 1)
Differential Equations (114)
and in general
k –1 D (r )
Dk (r ) = ∑ ( j + r ) α k – j + β k – j  C j (r ), Ck (r ) = – k (k = 1, 2, 3,...)
j =0 q (r + k )
The Ck thus, determined are rational functions of r, and the only points where they cease
to exist are the points r for which the denominator q(r + k) = 0 for some k = 1, 2, 3,.... Only two
such possible points exist.
Define F by

Φ ( x, r ) = c0 x r + x r ∑ Ck ( r ) x k ....(3.2.3)
k =0
If the series converges for 0 < x < r0, then clearly
L(Φ ) ( x, r ) = c0 q (r ) x r ,
since Ck (r ) satisfies equation 3.2.1 for every k = 1, 2, 3....

Thus if the function φ = x
r
∑ Ck x k is a solution of L(y) = 0 then r must be a root of the
k =0
indicial polynomial
q ( r ) = r (r –1) + r α 0 + β 0
and ck (k ³ 1) are determined uniquely in terms of r and c0 given by equation (3.2.2), provided
q(r + k) ¹ 0 k = 1, 2, 3,... Conversely if r is a root of q and if Ck(r) can be determined then the
function f given by equation (3.2.3) is a solution of L(y) = 0 for any choice of c0, provided the
series in equation (3.2.3) is convergent.
Let r1, r2 be two roots of q and suppose Re r1 ³ Re r2. Then q(r1 + k) ¹ 0 for all k = 1, 2, 3,...
Thus, Ck(r1) exists for all k = 1, 2, 3,... and for c0 = C0 (r) = 1 we get a solution.

φ1 ( x ) = Φ ( x, r1 ) = x r1 ∑ C k (r1 ) x k (C0 (r ) = 1),
k =0
is a solution of L(y) = 0, provided the series converges.
If r2 is a root of q distinct from r1 and q(r2 + k) ¹ 0 for k = 1, 2, 3,..., then clearly Ck(r2) is
defined for k = 1, 2, 3,.... and the function F2 defined by

Φ 2 ( x) = Φ ( x, r2 ) = x r2 ∑ Ck ( r2 ) x k (C0 (r2 ) = 1)
k =0

is another solution of L(y) = 0, provided the series is convergent. The condition q(r2 + k) ¹ 0 for
k = 1, 2, ... is same as r2 + k ¹ r1 for any k = 1, 2, 3,.... or r1 – r2 ¹ k i.e. r1 – r2 is not a positive
integer and the result follows.
Illustration :
Consider the equation
3
L( y ) = x 2 y ′′ +
xy ′ + xy = 0
2
As per theorem 3.2.1 we assume the solution f of the equation L(y) = 0 as

Differential Equations (115)



φ ( x) = x r ∑ ck x k
k =0

φ ′ ( x) = x r –1 ∑ (k + r ) ck x k
k =0

and φ ′′ ( x) = x r –2 ∑ (k + r ) (k + r –1) ck x k
k =0
∞ 3 ∞ ∞
L(φ ) = x r ∑ (k + r ) (k + r –1) ck x k + x r ∑ (k + r ) ck x k + x r +1 ∑ ck x k
k =0 2 k =0 k =0

 
=  r (r –1) + r  c0 x r +   (r + 1) (r ) + (r + 1)  c1 + c0  x r +1
3 3
 2   2  
 
+  (r + 2) (r + 1) + (r + 2)  c2 + c1  x r + 2 + ....
3
 2  
3
q( r ) = r ( r –1) + r is the indicial polynomial
2
L(φ ) = q (r )c0 x r + [q (r + 1) c1 + c0 ] x r +1 + [q (r + 2) c2 + c1 ] x r + 2 + .....

= q (r )c0 x r + x r ∑ [q (r + k ) ck + ck –1 ] x k
k =1

L(φ ) = 0 implies q(r) = 0 and q (r + k ) ck + ck –1 = 0

q( r ) = r ( r –1) + r = r  r +  = 0 implies r = 0, –
3 1 1
2  2 2
1
(Re r1 > Re r2 ) Define r1 = 0, r2 = –
2
ck –1
q( r + k ) ck + ck –1 = 0 gives ck = – , (k = 1, 2,3,...)
q( r + k )
 1  1  1   1 
Thus, ck =  –  –  –  ....  –  c0
 q( r + k )   q(r + k –1)   q(r + k – 2)   q( r + 1) 
In the above expression ck–1 is written in terms of ck–2 , ck–2 is expressed in terms of
ck–3 and so on.
(–1) k
ck = , k = 1, 2, 3.....
q (r + k ) q (r + k –1) q (r + k – 2).....q (r + 1)
1
Since r1 = 0, r2 = – , r1 – r2 is non zero and is not an integer. Therefore we apply theorem
2
3.2.1. For r = r1 = 0, c0 = 1 we get
∞ (–1) k x k
φ1 ( x) = 1 + ∑
k =1 q (k ) q (k –1) q (k – 2).....q(1)

Differential Equations (116)


and for c0 = 1, r = r2 = – 1/2 we obtain another solution
1 1 ∞
– – (–1) k x k
φ2 ( x) = x ∑ 2 +x 2 .
k = 1 q  k – 1  q  k – 3  q  k – 5  ..... q  1 
       
 2  2  2  2
These functions f1, f2 will be solutions provided the series converge on some interval
containing 0.

Let φ1 ( x) = ∑ d k ( x).
k =0
Using the ratio test we obtain
d k +1 ( x ) | x| | x|
= = → 0 as k → ∞ provided | x | < ¥. Thus
d k ( x) | q (k + 1) |
 3
(k + 1)  k + 
 2
the series defining f1 is convergent for all finite x. The same is true for f2.
r
To obtain solutions for x < 0, all the above calculations are valid if xr replaced by | x | ,
where | x | = e
r r log| x|

Thus two solutions which are valid for all x ¹ 0 are


∞ (–1) k x k
φ1 ( x ) = 1 + ∑ and
k =1 q( k ) q(k –1) q( k – 2)....q(1)
 

1  ∞ (–1) x k k 
φ2 ( x) = | x | 2 1 + ∑ ,
 k =1  – 1   – 3  ....  1  
qk  qk  q 
  2  2  2 
1
where | x | 2 is a positive square root of | x |.
Thus we have seen that if the roots of indicial polynomials are distinct and the difference
between these two roots is not an integer then the solutions of L(y) = 0 will be constructed by
using power series method.
In the next theorem we prove that if the roots are identical or the difference between the
roots is an integer still the power series solution exist.

Theorem 3.2.2
Consider the equation
L( y ) = x 2 y′′ + a( x) xy′ + b( x ) y = 0,
where a, b have power series expansions which are convergent for | x | < r0, r0 > 0 . Let r1, r2
(Re r1 ³ Re r2) be the roots of the indicial polynomial
q ( r ) = r (r –1) + a (0) r + b(0).
If r1 = r2 there are two linearly independent solutions f1, f2 for 0 < | x | < r0 of the form
φ1 ( x ) = | x |r1 σ 1 ( x ), φ 2 ( x ) =| x |r1+1 σ 2 ( x ) + (log | x |) φ1 ( x ),
Differential Equations (117)
where s1, s2 have power series expansions which are convergent for | x | < r0 and s1(0) ¹ 0.
If r1 – r2 is a positive integer there are two linearly independent solutions f1, f2 for
0 < | x | < r0 of the form
φ1 ( x ) = | x |r1 σ 1 ( x),
φ2 ( x) = | x |r2 σ 2 ( x) + c(log | x |) φ1 ( x),
where s1, s2 have power series expansions which are convergent for | x |< r0 , σ 1(0) ≠ 0,
σ 2 (0) ≠ 0, and c is a constant. It may happen that c = 0.
Proof :
For x > 0, suppose we have a solution f of the form

φ ( x ) = x r ∑ ck x k .
k =0

∞  : : 
L(φ ) ( x ) = x r ∑ (k + r ) (k + r –1) ck + α k + β k  x k
k =0 
: k : k
where α k = ∑ ( j + r ) c j α k – j and β k = ∑ c j β k – j
j=0 j=0
∞ ∞
a( x) = ∑ α k x k , b( x) = ∑ β k x k .
k =0 k =0

L (φ ) ( x ) = 0 implies
 : : 
[ ]k = (k + r ) (k + r –1) ck + α k + β k  = 0, k = 0,1, 2,3,...
 
 k k 
=  (k + r ) (k + r –1) ck + ∑ ( j + r) c j αk– j + ∑ c j βk – j 
 j =0 j =0 
= [(k + r ) (k + r – 1) ck + (k + r ) α 0 + β 0 ] ck
k –1
+ ∑ ( j + r ) α k – j + β k – j  c j
j =0
For k = 0 we must have
q(r ) = r ( r –1) + r α 0 + β 0 = 0
Then
[ ]k = q (r + k ) ck + d k = 0 .... 3.2.4
k –1
where dk = ∑ ( j + r ) α k – j + β k – j  c j .... 3.2.5
j =0

Here, we are going to consider two cases according as the roots r1, r2 (Re r1 ³ Re r2) of the
indicial polynomial q(r) satisfy.

Differential Equations (118)


Case (i) r1 = r2
Case (ii) r1 – r2 is a positive integer.
Since, Re r1 ³ Re r2, q(r1 + k) ¹ 0 for k = 1, 2, 3,.... and we can solve equation (3.2.4) and
(3.2.5) for ck and dk. Let the solutions of ck be denoted by Ck(r) and solution for dk be denoted by
Dk(r) . Then
L(Φ ) ( x, r ) = c0 q(r ) x r .....3.2.6
where F is given by

Φ ( x, r ) = c0 x r + x r ∑ Ck (r ) x k
k =0
The Ck(r) are determined recursively by the formulas
C0 (r ) = c0 ≠ 0,
q( r + k ) Ck ( r ) = – Dk ( r )
k –1
Dk (r ) = ∑ ( j + r ) α k – j + β k – j  C j (r ), k = 1, 2,3,....
j=0

In case (i) i.e. r1 = r2 , q( r1) = 0, q′ (r1 ) = 0.


On differentiating equation (3.2.6) with respect to r
We get
∂  ∂Φ 
L (Φ ) ( x , r ) = L   ( x, r )
∂r  ∂r 
= C0 [q′(r ) + (log x ) q (r )] x r
and we see that if r = r1 = r2 and C0 = 1, then
 ∂Φ 
 ( x, r1 ) = c0 [q ′( r1 ) + (log x) q ( r1 ) ] x
r1
L
 ∂r 
= 0.
 ∂Φ   ∂Φ 
Since L   ( x, r1 ) = 0,   ( x, r1 ) is a solution of L(f ) = 0. Thus the term by term
 ∂r   ∂r 
differentiation of equation (3.2.3) gives the second solution
∞ ∞
φ2 ( x ) = x r1 ∑ C k′ (r1 ) x k + (log x ) x r1 ∑ Ck (r1 ) x k
k =0 k =0

= x r1 ∑ Ck′ (r1 ) x k + (log x) φ1 ( x ).


where f1 is the solution already obtained in (3.2.3)

φ1 ( x ) = x r1 ∑ Ck (r1 ) x k , (Ck (0) = 1)
k =0
Case (ii) : Suppose r1 = r2 + m, where m is a positive integer. If C0 is given,
C1 (r2 ), C2 (r2 ),....., Cm –1 (r )

Differential Equations (119)


all exist as finite numbers, but since
q ( r + m) Cm ( r ) = – Dm ( r ), the coefficient of Cm(r) becomes zero at r = r2.

q(r ) = (r – r1 ) ( r – r2 )
and hence,
q ( r + m) = ( r + m – r1 ) ( r + m – r2 )
= (r + m – r2 – m) (r + m – r2 )
= (r – r2 ) (r + m – r2 )
If Dm(r) also has (r1 – r2) as a factor (i.e. Dm(r2) = 0 ), then it will get cancel from both the
sides of equation q(r + m) Cm(r) = – Dm(r) and would give Cm(r2) as a finite number. Then
Cm +1 (r2 ), Cm + 2 (r2 ),.....
all exist. In this special situation we will have a solution f2 of the form

φ2 ( x) = x r2 ∑ Ck (r2 ) x k (C0 (r2 ) = 1)
k =0

If we choose C0(r) = r – r2 then Dm(r2) = 0, as Dm(r) is linear homogeneous in C0(r) ,


C1(r).... Cm – 1(r) and hence Dm(r) has C0 (r) as a factor.
Let

ψ ( x, r ) = x r ∑ Ck (r ) x k (C0 (r ) = (r – r2 )),
k =0

L(ψ ) ( x, r ) = (r – r2 ) q (r ) x r
Therefore L(ψ ) ( x, r2 ) = 0 and
ψ ( x ) = ψ ( x, r2 )
is the second solution of L(y) = 0
Since C0 (r2 ) = C1 (r2 ) = ..... = Cm –1 (r2 ) = 0, the series y actually starts with the m-th power
in x.
To get a solution associated with r2 differentiate
L(ψ ) ( x, r ) = (r – r2 ) q (r ) x r
with respect to r then
∂ ∂ψ
[L(ψ ) ( x, r ) ] = L  
 ( x, r )
∂r  ∂r 
= q(r ) x r + (r – r2 ) [q′( x) + (log x) q( r )] x r
 ∂ψ 
and L  = 0 at r = r2
 ∂r 
∂ψ
and φ2 ( x) = ( x, r2 )
∂r

Differential Equations (120)


is a solution provided the series involved is convergent and
∞ ∞
φ2 ( x ) = x r2 ∑ Ck ′ (r2 ) x k + (log x ) x r2 ∑ Ck (r2 ) x k
k =0 k =0
where C0 (r) = (r – r2) and
C0 (r2 ) = C1 (r2 ) = C2 ( r2 ) = .... = Cm –1 ( r2 ) = 0
Thus,
∞ ∞
φ2 ( x ) = x r2 ∑ Ck′ (r2 ) x k + (log x ) x r2 ∑ Ck (r2 ) x k
k =0 k =m

∞ ∞
= x r2 ∑ Ck′ (r2 ) x k + (log x) x r2 ∑ Ck +m (r2 ) x k +m
k =0 m=0

∞ ∞
= x r2 ∑ Ck′ (r2 ) x k + (log x) x r2 + k ∑ Ck +m (r – m) x m
k =0 m=0

∞ ∞
= x r2 ∑ Ck′ (r2 ) x k + (log x) x r1 C ∑ Cm (r1 ) x m
k =0 m=0


φ2 ( x) = x r2 ∑ Ck′ (r2 ) x k + (log x) ⋅ c ⋅ φ1 ( x)
k =0

Where c is constant.
For x < 0, we replace x r1 , x r2 , log x everywhere by | x |r1 ,| x |r2 , log |x| respectively and the result
follows.
The method used in the theorem 3.2.2 is called the Frobenius method. The solutions f1, f2
are linearly independent. Thus, if the roots are equal or they differ by an integer then theorem
3.2.2 gives two linearly independent solutions of the differential equation
L( y ) = x 2 y′′ + x a( x) y′ + b( x) y = 0.

EXAMPLES

Q. 1. Find all solutions f of the form



Φ ( x ) = | x |r ∑ Ck x k (| x | > 0),
k =0
for the following equations.
(a) 3 x 2 y′′ + 5 x y′ + 3 x y = 0
(b) 2 x y′′ + (1 + x) y′ – 2 y = 0
Test each of the series involved for convergence
Answer (a) : For x > 0 suppose we have a solution f of the form

Differential Equations (121)



φ ( x) = x r ∑ ck xk , c0 ≠ 0
k =0

∞ ∞
then φ ′( x) = x r –1 ∑ (k + r ) ck x k and φ ′′( x ) = x
r –2
∑ (k + r ) (k + r –1) ck x k
k =0 k =0

Let L( y) = 3x 2 y′′ + 5 x y′ + 3x y therefore


∞ ∞
L(φ ) ( x ) = x r ∑ [3(k + r ) (k + r –1) + 5(k + r )]ck x k + x r ⋅ 3 ∑ ck x k +1
k =0 k =0

= [3r (r –1) + 5r ]c0 x r + { [3(r + 1) r + 5(r + 1)]c1 + 3 c0 } x r +1

+ { [3(r + 2) (r + 1) + 5(r + 2) ]c2 + 3 c1} x r +2 + ...


Let q (r ) = 3r (r –1) + 5r = r (3r + 2) then
L (φ ) ( x, r ) = q ( r )c0 x r + [q ( r + 1) c1 + 3c0 ] x r +1 + [q ( r + 2) c2 + 3c1 ] x r + 2 + ....

= q (r ) c0 x r + ∑ [q(r + k ) ck + 3 ck –1 ] x r + k
k =1

L(φ ) ( x, r ) = 0 only if q(r) = 0 and


q( r + k ) ck + 3 ck –1 = 0 for k = 1, 2, 3,.....
2 2
The indicial equation q (r) = 0 implies r (3r + 2) = 0 that is r = 0, – . Let r1 = 0, r2 = – .
3 3
(By choice r1 > r2)
Since q( r ) = r (3r + 2), q( r + k ) = ( r + k ) (3( r + k ) + 2) = ( r + k ) (3r + 3k + 2)
q( r + k ) ck + 3ck –1 = 0 gives
–3ck –1
ck = , k = 1, 2,3...
q (r + k )
(–3)k c0
=
q (r + k ) q(r + k –1) q (r + k – 2)..... q(r + 1)
Case 1 : r1 = 0
(–3) k c0
For c0 = 1 we obtain ck =
q (k ) q (k –1)..... q(1)
(–3)k
=
k (3k + 2) (k –1) (3k –1) (k – 2) (3k – 4).....1 ⋅ 5
(–3) k
=
k ! 5 ⋅ 8 ⋅11 ⋅⋅⋅⋅(3k – 4) (3k –1) (3k + 2)
Thus,
∞ (–3)k x k
φ1 ( x) = 1 + ∑
k =1 k ! 5 ⋅ 8 ⋅11⋅⋅⋅⋅(3k – 4) (3k –1) (3k + 2)

Differential Equations (122)


2
Case 2 : r2 = –
3
(–3)k c0
For c0 = 1 we obtain ck =
 2  5  7 1
q  k –  q  k –  q  k –  ..... q  
 3  3  3 3
Since q(r) = r (3r + 2)
(–3)k c0
ck =
 – 2  (3 )  – 5  (3 – 3)  – 8  (3 – 6) ..... 1 ⋅ 3
k  k k  k k  k
 3  3  3 3
(–3)k c0
=
k ! 1⋅ 4 ⋅ 7 ⋅⋅⋅ (3k – 8) (3k – 5) (3k – 2)
Thus,

2
 ∞ (–3) k x k 
φ2 ( x) = x 3
 ∑
1 + 
 k = 1 k ! 1 ⋅ 4 ⋅ 7 ⋅⋅⋅⋅ (3k – 8) (3k – 5) (3k – 2) 
To obtain solutions for x < 0, we replace xr by | x |r. Thus,
∞ (–3) k x k
φ1 ( x ) = 1 + ∑
k =1 k ! 5 ⋅ 8 ⋅11⋅⋅⋅⋅ (3k – 4) (3k –1) (3k + 2)
2
 – (–3)k x k 
and φ2 ( x ) = | x |
1 + ∑ 3

 k ! 1 ⋅ 4 ⋅ 7 ⋅⋅⋅⋅ (3k – 8) (3k – 5) (3k – 2) 
These functions f1 and f2 will be solutions for x ¹ 0, provided both the series converges
on some interval containing x = 0.

Let φ1 ( x) = ∑ dk ( x)
k =0
Using ratio test we obtain
d k +1 ( x ) (–3) x 3| x |
= = → 0
dk ( x) (k + 1) (3k + 5) (k + 1) (3k + 5)
as k ® ¥ provided | x | < ¥. Thus, series defining f1 is convergent for all finite x.

Let φ2 ( x) = ∑ d k ( x)
k =1
Using ratio test we obtain

d k +1 ( x ) –3 x 3| x|
= = → 0
d k ( x)  + 1 + (3k + 1) ( k + 1)
3 k  (k 1)
 3
as k ® ¥ provided | x | < ¥. Thus, series defining f2 is convergent for all finite x.
Thus f1, f2 are solutions of the given equation.
Differential Equations (123)
(b) Suppose for x > 0 we have a solution f of the form

φ ( x) = x r ∑ Ck x k , c0 ≠ 0
k =0

Let L( y ) = 2 xy′′ + (1 + x ) y′ – 2 y = 0 then


∞ ∞
L(φ ) ( x, r ) = ∑ 2(k + r ) (k + r –1) Ck x k +r –1 + ∑ (k + r ) Ck x k +r –1
k =0 k =0
∞ ∞
+ ∑ (k + r ) Ck x k +r – 2 ∑ Ck xk +r
k =0 k =0
∞ ∞
= ∑ [2(k + r ) (k + r –1) + (k + r )]Ck xk +r –1 + ∑ (k + r – 2) Ck x k +r
k =0 k =0

= [2r (r –1) + r ]c0 x r –1 + {[2(r + 1) (r ) + (r + 1)]c1 + (r – 2) c0 } x r

+ {[2(r + 2) (r + 1) + (r + 2) ]c2 + (r –1) c1} x r +1 + ⋅⋅⋅⋅



= q (r ) c0 x r –1 + ∑ [q (r + k ) Ck + (r + k – 3)Ck –1 ] xk +r
k =1
1
The indicial equation q(r) = 0 implies 2r2 – 2r + r = 0, r (2r – 1) = 0 gives r = 0,
2
1
Let r1 = and r2 = 0.
2
Observe that r1 ¹ r2 and r1 – r2 is not a positive integer. L(f ) (x, r) = 0 if and only if
q(r) = 0 and
q (r + k ) Ck + (r + k – 3) Ck –1 = 0 or
(r + k – 3) Ck –1
Ck = – for k = 1, 2,3
q (r + k )
Since q ( r ) = r (2r – 1) therefore q ( r + k ) = ( r + k ) (2 r + 2 k –1) and
 (r + k – 3)   (r + k – 4)   (r + k – 5)   (r – 2) 
Ck = +  –  –  –  ..........  –  C0
 q (r + k )   q (r + k –1)   q (r + k – 2)   q (r + 1) 
(–1) k ( r + k – 3) ( r + k – 4) ( r + k – 5) ⋅⋅⋅ ( r – 2) C0
=
(r + k ) (2r + 2k –1) (r + k –1) (2r + 2k – 3).....(r + 1) (2r + 1)
1
Case 1 : r1 = , C0 = 1
2
(–1) k  k –   k – 
5 7 9  3
 k –  ⋅⋅⋅⋅  – 
Ck =  2  2  2  2
 +1  1  3 3
k  (2k )  k –  (2k – 2)  k –  (2k – 4) ⋅⋅⋅⋅ (2)
 2  2  2 2

(–1) k (2k – 5) (2k – 7) (2k – 9) ⋅⋅⋅⋅3 (–1) (–3)


=
2 k k !(2k + 1) (2k –1) (2k – 3) ⋅⋅⋅⋅(3)
Differential Equations (124)
(–1) k (3)
=
2k k !(2k + 1) (2k –1) (2k – 3)
and
1
 ∞ 3(–1) k x k 
φ1 x = x  + ∑ = k
( ) 2 1 .
 k = 1 2 k !(2 k + 1) (2 k –1) (2 k – 3) 
Case 2 : r2 = 0, C0 = 1

(–1) k (k – 3) (k – 4) (k – 5) ⋅⋅⋅⋅(–1) (–2)


Ck =
k (2k –1) ( k –1) (2k – 3) ⋅⋅⋅⋅(1)

for k = 3, C3 = 0 therefore Ck = 0 for k = 1, 2, 3,...


(–2) C0
C1 = – =2
(1)
(–1) C1 2 1
C2 = – = = .
2⋅3 6 3
and φ2 ( x) = c0 + c1 x + c2 x 2
1
= 1 + 2x + x2
3
Thus, for x ¹ 0 we get two solutions
1
 ∞ 3(–1) k x k 
φ1 ( x ) = | x |2 1 + ∑ k 
 k = 1 2 k !(2k + 1) (2k –1) (2k – 3) 
x2
φ2 ( x ) = 1 + 2 x +
3

Check that series in the first solution is convergent Let φ1 ( x ) = ∑ d k ( x).
k =0
Using ratio test
d k +1 x | x|
= = → 0
dk 2( k + 1) (2k + 3) 2 ( k + 1) (2k + 3)
as k ® ¥ if | x | < ¥. The series convergent for finite x.

Q. 2. Obtain two linearly independent solutions of the following equations which are
valid near x = 0.
(a) x 2 y′′ + 3 x y′ + (1 + x ) y = 0

(b) x 2 y′′ + 2 x 2 y ′ – 2 y = 0
Ans. :
(a): For x > 0 suppose we have a solution f of the form

Differential Equations (125)



φ ( x) = x r ∑ Ck xk , C0 ≠ 0
k =0

Let L( y ) = x 2 y′′ + 3 xy′ + (1 + x ) y

L(φ ) ( x, r ) = x r ∑ (k + r ) (k + r –1) Ck x k + 3 x r ∑ (k + r ) Ck x k + ∑ Ck x k + r + x r ∑ Ck x k +1

∞ ∞
= ∑ [(k + r )(k + r –1) + 3(k + r ) + 1) ]Ck x k +r + ∑ ck x k +r +1
k =0 k =0

L (φ ) ( x, r ) = 0 implies
∞ ∞
∑ [(k + r )(k + r –1) + 3(k + r ) + 1]Ck x k + r + ∑ Ck x k + r +1 = 0
k =0 k =0

[r (r –1) + 3r + 1]C0 x r + {[(r + 1)(r ) + 3(r +1) + 1]C1 + C0} x r +1


+ {[(r + 2) (r + 1) + 3(r + 2) + 1]C2 + C1} x r +2 + ⋅⋅⋅ = 0
q ( r ) = r ( r –1) + 3r + 1 = 0 is indicial equation.
q(r ) = r 2 – r + 3r + 1 = (r + 1) 2
L (φ ) ( x ) = q( r ) C0 x r + [q ( r + 1)C1 + C0 ] x r +1 + [q ( r + 2)C2 + C1 ] q r + 2 + ⋅⋅⋅ = 0

= q (r ) C0 x r + ∑ [q (r + k ) Ck + Ck –1 ] x r + k = 0
k =1

L(φ ) ( x ) = 0 if and only if q(r) = 0 and


q(r + k )Ck + Ck –1 = 0 for k = 1, 2, 3,...
q(r) = 0 implies (r + 1)2 = 0 that is r = – 1 is a repeated root. Here r = r1 = r2 = – 1.
q (r + k ) Ck + Ck –1 = 0 for k = 1, 2, 3,...

Since, q( r ) = ( r + 1) 2 , q( r + k ) = (r + k + 1) 2 and
– Ck –1
Ck =
(r + k + 1) 2
 –1   –1   –1  –1   –1 
=   
2  
2 2
   ⋅⋅⋅  2

 (r + k + 1)   (r + k )   (r + k –1)   (r + k – 2)   (r + 2) 
2

(–1) k C0
=
[(r + 2) (r + 3) ⋅⋅⋅ (r + k – 2) (r + k –1) (r + k ) (r + k + 1)] 2
The first solution will be constructed by substituting Ck’s at r = – 1 in the series. Ck at
r = – 1 is
(–1)k C0
Ck =
k !2

Differential Equations (126)



φ1 ( x ) = x r ∑ Ck x k
k =0
∞ (–1) k
= x –1 ∑ 2
xk
k! k =0
The series converges for all finite x.
Since r = – 1 is the root of multiplicity 2, i.e. r = r1 = r2 = – 1, the second solution

φ2 ( x) = xr1 ∑ Ck′ ( r1 ) x k + (log x) φ1 ( x)
k =0

(–1) k C0
Ck = .
[(r + 2) (r + 3) (r + 4) ⋅⋅⋅ (r + k + 1) ] 2

Define D = (r + 2) (r + 3) (r + 4) ..... (r + k + 1) then


(–2)  1 1 1 1 
Ck' = (–1) k C0 + +
2 r + 2 r +3 r + 4
+ ⋅⋅⋅ +
D  r + k + 1 
(–2) (–1) k 1 1 1 1
Ck′ (–1) =  + + + ⋅⋅⋅ + 
k! 2
1 2 3 k
∞ (–2) (–1) k 1 + 1 + 1 + ⋅⋅⋅ + 1  k + (log ) φ ( )
φ2 ( x) = x –1 ∑  2 3 x x 1 x
k =0 k !2 k 
To obtain solution for x < 0 we replace x by | x |.
Thus, the two solutions are
∞ (–1) k x k
φ1 ( x ) = | x | –1

k =0 k !2
∞ (–2) (–1) k 1 + 1 + 1 + ⋅⋅⋅ + 1  k
φ2 ( x ) = | x |–1 ∑  2 3 x
k =0 k !2 k 
∞ (–1) k x k
+ (log | x |) | x |–1 ∑ .
k =0 k !2
Check that series in both the solutions converge.
(b): For x > 0 suppose we have a solution f of the form

φ ( x) = x r ∑ Ck xk (C0 ≠ 0)
k =0

Let L( y ) = x 2 y ′′ + 2 x 2 y ′ – 2 y then

L(φ ) ( x, r ) = ∑ (k + r ) (k + r –1) Ck x k + r + 2 (k + r )Ck x k + r +1 – 2 Ck x k + r 
k =0
∞ ∞
= ∑ [(k + r ) (k + r –1) – 2]Ck x k + r + ∑ 2(k + r ) Ck x k + r +1
k =0 k =0
Differential Equations (127)
= [r (r –1) – 2]C0 x r + {[(r + 1)(r ) – 2]C1 + 2 r C0 } x r +1

+ {[(r + 2) (r + 1) – 2]C2 + 2 (r + 1) C1} xr + 2 + ⋅⋅⋅



= q (r ) C0 xr + ∑ [q (r + k ) Ck + 2(r + k –1) Ck –1 ] xr +k
k =1

Indicial equation is q (r ) = r (r – 1) – 2 = 0 gives r 2 – r – 2 = (r – 2) ( r + 1) = 0 r = – 1, 2.


r1 = 2, r2 = –1 and r1 – r2 = 3 a positive integer.
If r1 – r2 is a positive integer, we try a series using the smallest root. If c0 and c3 both turn
out to be arbitrary, we obtain the general solution by this method. Otherwise the general solution
will involve a logarithm as it did in the case of equal roots. That logarithmic case is treated in
theorem 3.2.2.

Let us consider the series solution as φ ( x ) = ∑ ck x k –1 (r2 − smallest root = –1)
k =0

L (φ ) = ∑ (k –1) (k – 2) ck x k –1 + 2∑ (k –1) ck x k – 2 ∑ ck x k –1
k =0
∞ ∞
= ∑ [(k –1) (k – 2) – 2 ]ck x k –1 + 2 ∑ (k –1) ck x k
k =0 k =0

= (2 – 2) c0 x –1 + ∑ {[(k –1) (k – 2) – 2] ck + 2(k – 2) ck –1} x k –1
k =1
Since f is a solution L(f ) = 0 i.e.
0.c0 = 0 i.e. c0 is arbitary.

[(k –1)(k – 2) – 2] ck + 2 (k – 2) ck –1 = 0 k = 1, 2,3,...


k =1 – 2 c1 – 2c0 = 0 i.e. c1 = – c0
k=2 – 2 c2 = 0 i.e. c2 = 0
k =3 (2 – 2) c3 + 2 c2 = 0 i.e. 0 ⋅ c3 = 0 ⇒ c3 is arbitary.

[(k –1)(k – 2) – 2] ck + 2(k – 2) ck –1 = 0 k = 4,5, 6,...

(k 2 – 3 k ) ck + 2 (k – 2) ck –1 = 0
k (k – 3) ck + 2(k – 2) ck –1 = 0
2 (k – 2) ck –1
ck = – , k = 4,5, 6, 7,....
k (k – 3)

 –2 (k – 2)   –2 (k – 3)   –2 (k – 4)   –2(2) 
ck =     ⋅⋅⋅   c3
 k ( k – 3)   ( k –1) (k – 4)   ( k – 2) ( k – 5)   4 ⋅1 

(–2) k –3 (k – 2) ( k – 3) (k – 4) ⋅⋅⋅ (2)


= c3
k (k – 3) ( k –1) (k – 4) ( k – 2) (k – 5) ⋅⋅⋅ 4 ⋅1

Differential Equations (128)


(–2) k –3 (k – 2) 6
= c3
k!
Thus, we get a solution
∞ (–2) k –3 (k – 2) 6
φ ( x ) = c0 x –1 – c0 x 0 + 0 ⋅ x1 + c3 x 2 + ∑ c3 x k –1
k =4 k !

 ∞ (–2) k –3 ( k – 2) 6 
= c0 ( x –1 –1) + c3  x 2 + ∑ x k –1 
 k=4 k! 

 ∞ (–2) k ( k + 1) 6 
= c0 x –1 (1 – x ) + c3  x 2 + ∑ x k +2 
 k =1 (k + 3)! 
 ∞ (–2) k ( k + 1) 6 
= c0 x –1 (1– x) + c3 x 2 1 + ∑ xk 
 k = 1 (k + 3)! 
Thus, we get two solutions
 ∞ (–2) k (k + 1) 6 
φ1 ( x ) = x –1 (1– x) and φ2 ( x ) = x 2 1 + ∑ xk 
 k = 1 (k + 3)! 
These are two solutions for x > 0 for x < 0 replace x by | x | we get,
 ∞ (–2) k ( k + 1) 6 
φ1 ( x ) = | x |–1 (1– x) and φ 2 ( x ) =| x |2 1 + ∑ xk 
 k = 1 (k + 3)! 
Check that series appearing in f2 is convergent series.

EXERCISE

1. Compute indicial polynomials and their roots for the following equations.
(a) x 2 y′′ + ( x + x 2 ) y ′ – y = 0

(b) x 2 y ′′ + x y′ +  x 2 –  y = 0
1
 4
(c) 4 x 2 y′′ + (4 x 4 – 5 x) y′ + ( x 2 + 2) y = 0

(d) x 2 y′′ + ( x – 3x 2 ) y′ + e x y = 0

2. Find a solutions f of the form



φ ( x) = x r ∑ Ck xk , ( x > 0)
k =0
for the following equations.

Differential Equations (129)


(a) 2 x 2 y′′ + ( x 2 – x) y′ + y = 0

(b) x 2 y′′ + ( x – x 2 ) y′ + y = 0
3. For each equation obtain two linearly independent solutions valid near origin
(a) 2 x ( x –1) y′′ + 3( x –1) y′ – y = 0
(b) 2 x y′′ + 5(1 + 2 x ) y′ + 5 y = 0
(c) 3 x y′′ + (2 – x) y′ – 2 y = 0
(d) 2 xy′′ + (1 – 2 x 2 ) y′ – 4 xy = 0

4. Consider the following equation near x = 0


(a) 2 x 2 y′′ + (5x + x 2 ) y′ + ( x 2 – 2) y = 0
(b) 4 x 2 y′′ – 4 xe x y′ + 3(cos x ) y = 0
Compute the roots r1, r2 of the indicial equation for each relative to x = 0.

5. Obtain two linearly independent solutions of the following equations which are valid
near x = 0.
(a) x 2 y′′ – 2 x ( x + 1) y ′ + 2 ( x + 1) y = 0

(b) x 2 y′′ – 2 x 2 y′ + (4x – 2) y = 0


(c) x y ′′ – (4 + x) y ′ + 2 y = 0

(d) x 2 y′′ + 2 x ( x – 2) y ′ + 2(2 – 3x ) y = 0 .

Answers :

1. (a) q( r ) = r 2 –1 ; r1 = 1, r2 = –1

1 1 1
(b) q( r ) = r – r1 = , r2 = –
2
;
4 2 2

2 9 1 1
(c) q( r ) = r – r + ; r1 = 2, r2 =
4 2 4

(d) q(r ) = r 2 + 1 ; r1 = i, r2 = – i

∞ (–1) k x k 1 ∞ (–1)k x k 1 –x
2. (a) y1( x) = x ∑ , y 2 ( x) = x 2 ∑ = x 2e 2
k =0 1⋅ 3 ⋅ 5 ⋅ 7 ⋅⋅⋅⋅(2k + 1) k =0 2k k !

∞ i(1 + i) ⋅⋅⋅⋅ (k –1 + i) ∞ (–i) (1– i) ⋅⋅⋅⋅ (k –1– i)


(b) y1 ( x) = x ∑ y2 ( x) = x – i ∑
i
xk ; xK
k =0 (1 + 2i) (2 + 2i) ⋅⋅⋅⋅ (k + 2i) k ! k =0 k ! (1 + 2i ) (2 – 2i ) ⋅⋅⋅⋅ (k – 2i)

Differential Equations (130)



xn − 12 1
3. (a) y1 = 1– ∑ ; y2 = x –x 2
n=1 4n 2 –1

3
3(–5) x n – – 12
(b) y1 = 1 + ∑ ; y2 = x 2 –10 x
n !(2n + 1) (2n + 3)

∞ n + 13 ∞
(3 n + 4) x ( n + 1) x n
(c) y1 = ∑ ; y2 = 1 + ∑
n =0 4 ⋅ 3n n ! n =1 2 ⋅ 5 ⋅ 8 ⋅ ⋅ ⋅ ⋅(3 n – 1)


x 2 k +1 ∞ 2k x 2k
(d) y1 = ∑ ; y2 = 1 + ∑
n =0 2k k ! n =1 3 ⋅ 7 ⋅11 ⋅⋅⋅⋅(4k –1)

1
4. (a) r1 = ; r2 = –2
2

3 1
(b) r1 = ; r2 =
2 2

5. (a) y1 ( x ) = x ; y2 ( x ) = x (e 2 x – 1)

 ∞
2k xk 
(b) y1(x) = x y2(x) = x–1 1+ 3x + 6x2 –3 ∑  − 4x log| x |
2 2
;
 k =4 (k –3) k !

1 1 ∞ 60 x n
(c) y1 = 1 + x + x2 ; y2 = x 5 + ∑
2 12 n = 6 ( n – 5)! n (n –1) ( n – 2)

∞ 6(–2) n –3 x n+1
(d) y1 = x – 2 x + 2 x y2 = x 4 + ∑
2 3
;
n=4 n!

Unit 3 : The Bessel equation


If a is a constant, Re a ³ 0 the Bessel equation of order a is the equation
x 2 y′′ + xy ′ + ( x 2 – α 2 ) y = 0
This has the form
1 x2 – α 2
y ′′ + y′ + y=0
x x2
1 x2 – α 2
where p ( x) = and q ( x ) = . x = 0 is a singular point. Since the denominator of p(x)
x x2
does not contain x to a power higher than one and the denominator of q(x) (i.e. x2) does not
contain the factor x to a power higher then 2, x = 0 is a regular singular point. Therefore the
power series solution f will have the form
Differential Equations (131)

φ ( x) = x r ∑ Ck x k
k =0

Let L( y ) = x 2 y ′′ + xy′ + ( x 2 – α 2 ) y.

L(φ ) ( x, r ) = ∑ (k + r ) (k + r –1) C k x k +r + ∑ (k + r ) C k x k +r + ( x 2 – α 2 ) ∑ C k x k +r

∞ ∞
= ∑ (k + r )(k + r –1) + (k + r ) – α 2  Ck x k + r + ∑ Ck x k + r + 2
k =0
  k =0

=  r ( r –1) + r – α 2  C0 x r + ( r + 1) r + ( r + 1) – α 2  C1x r +1
   

{ }

+ ∑ (k + r ) 2 – α 2  Ck + Ck –2 x k +r
k =2
 
The indicial equation is
q( r ) = r ( r –1) + r – α 2 = 0
q( r ) = r 2 – α 2 = 0.
The indicial polynomial q(r) has two roots r1 = a and r2 = –a . We shall construct solutions
for x > 0. We consider three cases namely a = 0, 2a is not a positive integer and 2a is a positive
integer.
Case 1 : a = 0
Since the roots are both equal to zero by theorem 3.2.2, there are two solutions f1, f2 of
the form
φ1 ( x) = σ 1 ( x) and φ2 ( x) = x σ 2 ( x) + (log x) φ1 ( x),
Where s1(x), s2(x) have power series expansions which converge for all finite x.
Since a = 0,

L( y ) = x 2 y′′ + x y′ + x 2 y
∞ ∞
Suppose σ 1 ( x ) = xα ∑ Ck x k = ∑ Ck x k (C0 ≠ 0) be a solution of L(y) = 0. Then
k =0 k =0
∞ ∞ ∞
L(σ 1 ) ( x ) = ∑ k (k –1) Ck x k + ∑ k Ck x k + ∑ Ck x k +2
k =2 k =1 k =0
∞ ∞ ∞
= ∑ k (k –1) Ck x k + C1x + ∑ k Ck x k + ∑ Ck –2 x k
k =2 k =2 k =2

= C1x + ∑ {[k (k –1) + k ]Ck + Ck –2 } x k
k=2

Since s1 is a solution L (s1) = 0 for all x. Therefore C1 = 0 and [k (k –1) + k ]Ck + Ck –2 = 0,


k = 2, 3, 4
Thus, C1 = 0 and
Differential Equations (132)
k 2Ck = –Ck –2 for k = 2,3, 4,.....
The recurrance relation becomes
Ck –2
Ck = – , k = 2,3, 4,.....
k2
Since C1 = 0, C3 = C5 = C7 = ..... = C2n +1 = .... = 0.
The choice C0 = 1 implies
C0 1 C2  1   1  (–1) 2
C2 = – =– , C 4 = – =–  –  = ,.......
22 22 42  42   22  22 ⋅ 42
In general
(–1) m (–1) m
C2m = = , m = 1, 2,3,.....
22 ⋅ 42 ⋅ 62 ⋅⋅⋅⋅(2m) 2 22m m !2
Thus s1 contains only even powers of x and we get
∞ (–1) m x 2m
σ1( x) = ∑
m=0 2 2m m !2
The function defined by this series is called the Bessel function of zero order of the first
kind and is denoted by J0. Thus
∞ (–1) m x 2m
J 0 ( x) = ∑
m=0 22 m m !2

Let J 0 ( x) = ∑ d k ( x)
k =0

Using ratio test


d k +1 x2 x2
= 2 = →0
dk 2 (k + 1) 2 4(k + 1) 2
as k ® ¥ if | x | < ¥. Thus, the series converges for | x | < ¥ and J0 (x) is the first solution of
Bessel equation with a = 0.
Now we determine a second solution f2 for the Bessel equation of order zero (i.e. a = 0).
Let φ1 ( x ) = J 0 ( x ) then the solution f2 has the form
φ2 ( x) = x σ 2 ( x) + (log x) φ1 ( x)

= ∑ ck x k + (log x) φ1 ( x), (C0 = 0).
k =0
Since f2 is second solution

L(φ 2 ) ( x) = x 2φ 2′′ + x φ 2′ + x 2φ 2 = 0
∞ φ ( x)
φ2′ ( x) = ∑ k ck x k –1 + 1 + (log x) φ1′ ( x)
k =1 x

Differential Equations (133)


∞ 2 φ1′ ( x) φ1
φ2′′ ( x ) = ∑ k (k –1) ck x k –2 + – 2 + (log x ) φ1′′
k =2 x x

L(φ 2 ) ( x) = ∑ k (k –1) ck x k + 2 x φ1′ ( x) – φ1 ( x)
k =2

+ x 2 (log x) φ1′′ + ∑ k ck x k + φ1( x) + x (log x) φ1′ ( x)
k =1

+ x2 ∑ ck x k + x 2 (log x) φ1 ( x)
k =0
∞ ∞ ∞
= ∑ k (k –1) ck x k + ∑ k ck x k + ∑ ck x k + 2 + 2 x φ1′ ( x)
k =2 k =1 k =0

+ (log x ) ( x 2φ1′′ + x φ1′ + x 2φ1 )

Since L (φ1 ) = x 2φ1′′ + x φ1′ + φ1 = 0,


∞ ∞ ∞
L(φ2 )( x) = ∑ k (k –1) ck x k + c1 x + ∑ k ck x k + ∑ ck –2 x k + 2 x φ1′ ( x)
k =2 k =2 k =2

Since L(f2 ) should be zero and φ1′ ( x ) = J 0′ ( x ),


(–1) m 2m x 2m –1
φ1′ ( x) = J 0′ ( x) = ∑ ,
22 m (m!) 2
∞ ∞ ∞ (–1) m 2m x 2m
Thus ∑ k (k –1) ck x k + c1 x + ∑ k ck x k + ∑ ck –2 x k = –2 ∑
k =2 k =2 k =2 2 2m (m !) 2
∞ (–1) m 2m x 2m
c1x + ∑ {[k (k –1) + k ]ck + ck –2} x k = –2 ∑
k =2 2 2m (m !) 2
Since the series on right has only even powers of x, all odd terms on the left hand side
should be zero c1 = c3 = c5 = c7 = .... = c2n+1 = .....= 0.
The relation for the other coefficients that is for k even (let k = 2m) is

(–1) m 2m
[2m (2m –1) + 2m]c2m + c2m–2 = –2
22m (m!) 2

(–1) m+1 m
[4m 2 ]c2m + c2m –2 = , m = 1, 2,3, 4,.....
22m –2 (m!) 2

1  (–1) m+1 m 
c2 m = 2  2 m –2
– c2 m –2  , m = 1, 2,3, 4....
(2 m)  2 (m !) 2

c0 = 0

Differential Equations (134)


1 1  1 1 1 1 
c2 = , c4 = – –  = –  + 1
22 42  2.2 2 2 2  2 24 2  2 
1 1 1  1  1  1 + 1 + 1
c6 =  2 2 + 2 2  + 1 = + 2 2 2  
 2 4 3 2 4  2  
62 2 4 6 3 2 
and it can be shown by induction that
(–1) m  1
+ ⋅⋅⋅ + + + 1  , m = 1, 2,3,....
1 1 1 1
c2m = + +
2  m m –1 m – 2
2 (m !) 
2m 3 2 
The solution thus determined is called a Bessel function of zero order of second kind, and
is denoted by K0. Hence,
∞ (–1) m
2m
 + 1 + 1 + ⋅⋅⋅ + 1  x 
K 0 ( x) = – ∑ 1   + (log x ) J 0 ( x)
m =1 (m !)2  2 3 m  2 
Using the ratio test we can check that the series on the right is convergent for all finite
values of x.
Now we compute solution for Bessel equation of order a, where a ¹ 0 and Re a ³ 0.
L( y ) = x 2 y ′′ + x y ′ + ( x 2 – α 2 ) y = 0
Let x > 0. The roots of indicial equation are
r1 = a, r2 = – a.
Let us find out the solution corresponding to r1 = a. A solution f1 (x) has the form

φ1 ( x) = xα ∑ ck x k , (c0 ≠ 0)
k =0
Then
∞ ∞
L(φ1 )( x) = ∑ (k + α )(k + α –1) ck x k +α + ∑ (k + α ) ck x k +α
k =0 k =0

∞ ∞
+ ∑ ck xk +α + 2 – α 2 ∑ ck xk +α
k =0 k =0

∞ ∞
= ∑ (k + α ) (k + α –1) + (k + α ) – α 2  ck x k +α + ∑ ck x k +α + 2
k =0 k =0

∞ ∞
= ∑ (k + α ) 2 – α 2  ck x k +α + ∑ ck x k +α +2
k =0 k =0


= 0 ⋅ c0 xα +  (α + 1) 2 – α 2  c1 xα +1 + ∑  (k + α ) 2 – α 2  ck x k +α
   
k =2


+ ∑ ck –2 x k +α
k =2

Differential Equations (135)



 k =2
 ( )
= 0 ⋅ c0 xα + (α + 1) 2 – α 2  c1 xα +1 + ∑  (k + α 2 ) – α 2 ck + ck –2  x k +α
 

Thus, L(φ1 ) ( x) = 0 implies


0 ⋅ c0 = 0, c0 is arbitrary.
c1 = 0,
 (α + k ) 2 – α 2  ck + ck –2 = 0 k = 2, 3, 4,.....
 
Since (α + k ) 2 – α 2 = k (2α + k ) ≠ 0 for k = 2,3,......,
k (2α + k )ck + ck –2 = 0 gives
ck –2
ck = – .
k (2α + k )
Since c1 = 0, c2k +1 = 0 for k = 0,1, 2, 3,.....
that is all odd terms are zero.
We find
c0 –c
c2 = – = 2 0
2(2α + 2) 2 (α + 1)
c2 c0
c4 = – = 2
4(2α + 4) 2 (α + 1) (α + 2)

– c4 – c4 – c4 – c0
c6 = = = = 6
6(2α + 6) 12(α + 3) 2 ⋅ 3!(α + 3) 2 ⋅ 3!(α + 1) (α + 2) (α + 3)

– c6 – c0
c8 = = 8
8(2α + 8) 2 4!(α + 1) (α + 2) (α + 3)(α + 4)

In general,
(–1) m c0
c2m = .
22m m!(α + 1)(α + 2) (α + 3) ⋅⋅⋅ (α + m)
Thus the solution f1 becomes

φ1 ( x) = xα ∑ ck xk (c0 ≠ 0)
k =0

∞ (–1) m x 2 m
= c0 xα + xα ∑
m =1 22 m m !(α + 1) (α + 2) ⋅⋅⋅ (α + m)
for a = 0 and c0 = 1, f1(x) becomes J0 (x). Before going for the second solution let us define
gamma function and study some properties of gamma function.

Differential Equations (136)


Definition 3.3.1
The gamma function is defined by

Γ( z ) = ∫ e – x x z –1dx, (Re z > 0)
0
Lemma 1 : Γ( z + 1) = z Γ( z )

Proof : Γ( z + 1) = ∫ e – x x z dx
0

= TLim
→∞ ∫ e
–x z
x dx
0

 z e– x T T – x 
= TLim
→∞  x | – ∫ – e ⋅ ze z –1dx 
 –1 x = 0 0 
T
= TLim
→∞ z ∫ e – x x z –1dx
0

= z ∫ e – x x z –1dx
0
= z Γ( z )
T
Observe that Lim
T →∞ x ze – x | = Lim
T →∞ T z ⋅ e –T – 0 = 0
0

By definition 3.3.1

Γ(1) = ∫ e – x dx = 1
0
Thus, if z is a positive integer n,
Γ(n + 1) = n Γ(n) = n (n –1) (n – 2)
= ( n ) ( n –1) ( n – 2) ( n – 3) ⋅⋅⋅⋅Γ (1)
=n!
Thus, gamma function is an extension of the factorial function to numbers which are not
integers.
Suppose Re z < 0 and z is not a negative integer then there is a natural number N such that
– N < Re z < – N + 1
But then Re (z + N) > 0 and therefore we can define
Γ( z + N ) = ( z + N –1) ( z + N – 2) ⋅⋅⋅⋅ ( z + 1) z Γ( z ). Then
Γ( z + N )
Γ( z ) = , (Re z < 0)
( z + N –1) ( z + N – 2) ( z + N – 3) ⋅⋅⋅⋅ ( z + 1) z
The gamma function is not defined at 0, –1, –2, –3,.......
We have a solution f1 (x) as

Differential Equations (137)


∞ (–1) m x 2m
φ1 ( x ) = c0 x α + c0 xα ∑
m =1 22m m !(α + 1) (α + 2) ⋅⋅⋅⋅ (α + m )
1
Now choose c0 = α
. we obtain a solution of the Bessel equation of order a
2 Γ(α + 1)
which is denoted by Ja and is called the Bessel function of order a of the first kind.
α ∞ 2m
x (–1) m x
J α ( x) =   ∑  
2 m=0 m !(α + m) (α + m – 1) ⋅⋅⋅⋅ (α + 1) Γ(α + 1)  2 
α ∞ 2m
(–1) m
=  
x  x
∑   , (Re z > 0)
 2 m=0 m ! Γ (α + m + 1)  2 
Observe that this formula for Ja reduces to J0 when a = 0 as G(m + 1) = m!. Ja (x) is one
solution of Bessel equation with a ¹ 0 and Re a ³ 0.
To determine second solution we have to consider two situations. Either 2a is not a positive
integer or 2a is a positive integer. We determine second solution for both the situations.

Case 2 : 2a is not a positive integer


If 2a is not a positive integer there is another solution f2 (x) corresponding to the root
r2 = – a of the form

φ2 ( x) = x –α ∑ ck x k .
k =0

On repeating the same calculations we have carried out for the root a, (replace a by – a
everywhere)
We get the second solution
–α ∞ 2m
x (–1) m x
J –α ( x ) =   ∑   .
 2 m=0 m ! Γ (m – α + 1)  2 
Observe that G (m – a + 1) exists for all m = 0, 1, 2, 3,..... since a is not a positive integer.

Case 3 : 2a is a positive integer


(a) a is not a positive integer .
If a is not a positive integer G (m – a + 1) exists and the function J–a (x) is the second
solution of the Bessel equation.
(b) a is a positive integer.
Suppose a = n. According to theorem 3.2.2 there is a solution f2 of the form

φ2 ( x) = x – n ∑ ck x k + c (log x) J n ( x).
k =0

L(φ 2 ) ( x ) = x 2φ 2′′ ( x) + x φ 2′ ( x ) + ( x 2 – n 2 ) φ 2 ( x) = 0 implies

Differential Equations (138)


 ∞ 2c c 
x 2  ∑ (k – n) (k – n –1) ck x k – n –2 + J n′ – 2 J n + (c log x ) J n′′ 
 k = 0 x x 
 ∞ c 
+ x  ∑ (k – n ) ck x k – n –1 + J n + (c log x ) J n′ 
 k = 0 x 
 ∞ 
+ ( x 2 – n 2 )  ∑ ck x k – n + (c log x) J n ( x )  = 0.
 k = 0 
Since Jn (x) is the first solution corresponding to a = n, x 2 J n′′ + x J n′ + ( x 2 – n 2 ) J n ( x) = 0,
and above equation gets reduce to
∞ ∞
∑ (k – n)(k – n –1) ck x k – n + 2c x J n′ – c J n + ∑ (k – n) ck x k – n + c J n
k =0 k =0
∞ ∞
+ ∑ ck x k – n+ 2 – n 2 ∑ ck x k – n = 0
k =0 k =0
Therefore
∞ ∞
∑ (k – n) (k – n –1) + (k – n) – n 2  ck x k – n + ∑ ck x k – n+2 + 2 c x J n′ ( x) = 0.
k =0 k =0

∞ ∞
or ∑ (k – n) 2 – n 2  ck x k – n + ∑ ck x k – n +2 + 2 c x J n′ ( x) = 0
k =0 k =0

∞ ∞
that is 0 ⋅ c0 x – n +  (1– n)2 – n 2  c x1– n + ∑  (k – n)2 – n 2  ck x k – n + ∑ ck x k – n +2
   
k =2 k =0

+2 c x J n′ ( x) = 0.

Since ∑ ck x k – n+ 2 = c0 x – n+ 2 + c1 x1– n+ 2 + c2 x 2– n+ 2 +c3 x3– n+ 2 + ⋅⋅⋅⋅
k =0

= ∑ ck –2 x k – n , we get
k =2

{ }

0 ⋅ c0 x – n + (1– n)2 – n2  c1 x1– n + ∑ (k – n)2 – n 2  ck + ck –2 x k – n + c x J n′ = 0.
  k =2
 
On multiplying by xn we have

0 ⋅ c0 + (1– 2n) c1 x + ∑ [k (k – 2n) ck + ck –2 ] xk = –2 c x J n′ ( x) ⋅ x n .............. (3.3.1)
k =2

Since the first solution is Ja (x) with a = n and for a = n, G (m + a + 1) = G (m + n + 1)


= (m + n) ! we have
∞ 2m + n
(–1) m  x 
J n ( x) = ∑  
m=0 m ! (m + n )!  2 

Differential Equations (139)


Therefore
2 m + n –1
(2m + n) (–1) m  x  1
J n′ ( x) = ∑    
m ! (m + n)!  2   2
Thus equation (3.3.1) becomes

∞ ∞ (2m + n ) (–1)m x 2m + 2 n
0 ⋅ c0 + (1– 2n) c1 x + ∑ [k (k – 2n) cn + ck –2 ] x k = – c ∑
k =2 m=0 m ! (m + n)! 22m + n –1
.............. (3.3.2)
The series on the right side begin with x2n and since n is positive integer, the right side do
not contain any odd terms. Therefore c1 = 0 c2k+1 = 0 for k = 1, 2, 3,.... and if n > 1 then
k ( k – 2n) ck + ck –2 = 0, for k = 2,3, 4,...., 2n –1.

Since c1 = 0, c3 = c5 = c7 = .... = c2n –1 = 0.


+ ck –2
Whereas ck = gives
k (2n – k )
c0 c2 c0
c2 = , c4 = = 4
2
2 (n –1) 4 ⋅ 2 (n – 2) 2 ⋅ 2 ⋅ (n – 2) (n –1)
c4 c0
c6 = =
6 ⋅ 2 (n – 3) 3!26 (n – 3) (n – 2) (n –1)
c0
in general c2 j = , j = 1, 2,3,....., n –1 ...(3.3.3)
2 2j
j !(n –1) (n – 2) (n – 3) ⋅⋅⋅⋅ (n – j )
In particular
c0
c2n –2 = 2n –2
2 (n – 1)!(n –1)!
On comparing the coefficients of x2n in equation (3.3.2) we get
n c
c2n –2 = – c n –1
=–
n !2 (n –1)!2n –1
c0 c
Thus c2n –2 = 2n –2 2
=– n –1
2 (n –1)! 2 (n –1)!
c0
and therefore c=– n –1
. .....( 3.3.4 )
2 (n –1)!
Since c2 n–2 is used to find c, c2n remains undetermined, but the remaining coefficients
c2n+2, c2n + 4, c2n + 6,..... can be obtained by comparing the coefficients of x2(n + j) in equation
(3.3.2).
(–1) j (2 j + n) 1
(2n + 2 j ) (2n + 2 j – 2n) c2n+ 2 j + c2n+ 2 j –2 = – c j = 1, 2,3.....
j !(n + j )! 2 +n –1
2 j

Differential Equations (140)


(–1) j (2 j + n) 1
4 j (n + j ) c2 n+2 j + c2 n+2 j –2 = – c j = 1, 2,3.....
j !(n + j )! 2 + n –1
2 j

for j = 1 we have
(–1) (n + 2)
4(n + 1) c2 n+2 + c2 n = – c
(n + 1)! 2n+1
c  + 1  c2n
c2n+ 2 =  1  –
4 ⋅ 2 n+1 ⋅ (n + 1)!  n + 1  4(n + 1)
c2n –c  + 1 + 1 + 1 + ⋅⋅⋅⋅ + 1 
Choose = + 1 
4(n + 1) 4 ⋅ 2 ⋅ (n + 1)!
n 1
 2 3 4 n
–c  1 1 1 1
i.e. c2 n = n +1  1 + + + + ⋅⋅⋅⋅ + 
2 n!  2 3 4 n
+c  + + 1 + 1 + 1 + ⋅⋅⋅⋅ + 1 + 1 
Then c2n+ 2 = n+1 1 1 
4 ⋅ 2 (n + 1)!  2 3 4 n n +1
for j = 2 we have
c (–1) 2 ( n + 4)
4 ⋅ 2 (n + 2) c2 n + 4 = – – c2 n+2
2!(n + 2)!2n +3
–c  2  1 c  1 1 1 
c2 n + 4 = n+ 3 
1+ – n+1 1 + 1 + + + ⋅⋅⋅⋅ + 
4 ⋅ 2 ⋅ 2!(n + 2)! 2  n + 2  4 ⋅ 2 ( n + 2) 4 ⋅ 2 (n + 1)!  2 3 n +1 
–c  1 + 1 + + + 1 + 1 + ⋅⋅⋅⋅ + 1 + 1 
= n+2 
1 1 
4 (n + 2)! 2
2  2 n+2 2 3 n n +1
–c  1   1 1 1 1 
= n + 2 
1 +  + 1 + + + ⋅⋅⋅⋅ + + 
4 (n + 2)! 2
2
 2   2 3 n n + 1  

– (–1) 2 c  1   1 1 1 1 
= 1 + 2  + 1 + 2 + 3 + ⋅⋅⋅⋅ + n + n + 1 
2 ⋅ 2!(n + 2)! 2n+4  
It can be shown by induction that

– (–1) m c  1 1 1  1 1 1 
c2n + 2m =  1 + 2 + 3 + ⋅⋅⋅⋅ + m  +  1 + 2 + 3 + ⋅⋅⋅⋅ + n + m   ,
2 ⋅ m!(n + m)! 2 n+2 m  
m = 1, 2,3,....
Finally we get a solution f2

φ2 ( x ) = x – n ∑ ck x k + (c log x ) J n ( x )
k =0
2n –1 ∞
= x – n ∑ ck xk + x – nc2n –1x2n−1 + x – nc2n x2 n + x – n ∑ ck xk + (c log x) J n ( x)
k =0 k = 2n–1
Since all odd terms c2k+1 = 0, k = 1, 2, 3,....., we get,

Differential Equations (141)


n –1 x 2 j c0 c 1 + 1 + 1 + ⋅⋅⋅⋅ + 1  n
φ2 ( x ) = x c0 + x
–n –n
∑ – n+1  x
j =1 2 2j
j !(n –1) (n – 2) ⋅⋅⋅⋅(n – j ) 2 n! 2 3 n
∞ (–1) m c x 2 m+ n  1 1 1  1 1 1 
– ∑ 1 + 2 + 3 + ⋅⋅⋅⋅ + m  + 1 + 2 + 3 + ⋅⋅⋅⋅ + n + m  
m =1 2 ⋅ m !(n + m)! 2n +2m  
+ (c log x) J n ( x).
Where c0 and c are constants related by equation (3.3.4) when c = 1, the resulting solution
f2 is often denoted by Kn. If c = 1 then c0 = – 2n–1 (n – 1) ! and
n –1 (–1) 2 n –1 (n –1)! x 2 j
φ2 ( x ) = – x – n 2 n –1 (n –1)!+ x – n ∑
j =1 22 j j !(n –1) (n – 2) ⋅⋅⋅⋅(n – j )
m n+2 m
xn
 1 1 1  n –1 (–1)  x
– n+1 1 + + + ⋅⋅⋅⋅ +  – ∑  
2 n!  2 3 n  m =1 2 ⋅ m !( n + m)!  2 
 1 1 1  1 1 1 
1 + 2 + 3 + ⋅⋅⋅⋅ + m  + 1 + 2 + 3 + ⋅⋅⋅⋅ + n + m   + (c log x) J n ( x)
 
–n –n 2j
= –  
x 1  x 1 n –1 – (n – j –1)!  x 
(n – 1)!+   ⋅ ∑  
 2 2  2 2 j =1 j!  2
n+2 m
1 ∞
m
xn
 1 1 (–1)  x
– n+1 1 + + + ⋅⋅⋅⋅ +  – ∑  
2 n!  2 3 n  m =1 2 ⋅ m ! ( n + m)!  2 
 1 1 1  1 1 1 
 1 + 2 + 3 + ⋅⋅⋅⋅ + m  +  1 + 2 + 3 + ⋅⋅⋅⋅ + n + m   + (c log x ) J n .
 
The function f2 when c = 1 is denoted by Kn . Thus
– n n –1 2j n
1x (n – j –1)!  x  1 1 1 1 1x
K n ( x) = –   ∑   – ⋅  1 + + + ⋅⋅⋅⋅ +   
22 j =0 j! 2 2 n!  2 3 n  2
n ∞ 2m
1x (–1) m  x   1 1 1  1 1 1 
–   ∑    1 + 2 + 3 + ⋅⋅⋅⋅ + m  + 1 + 2 + 3 + ⋅⋅⋅⋅ + m + n  
2 2 m =1 m !( n + m)!  2   
+ (log x ) J n ( x ). .......... (3.3.5)
The function Kn is called a Bessel function of order n of second kind.
In this section we have derived all kinds of Bessel functions. We list all these functions
here.
(1) Bessel function of zero order of the first kind denoted by J0(x) and defined by
∞ (–1) m x 2m
J 0 ( x) = ∑
m=0 2 2 m m !2
(2) Bessel function of zero order of second kind denoted by K0(x) is
∞ 2m
(–1) m  + 1 + 1 + ⋅⋅⋅⋅ + 1  x 
K 0 ( x) = – ∑ 1   + (log x ) J 0 ( x )
m =1 m !2  2 3 m  2 

Differential Equations (142)


(3) Bessel function of order a of first kind denoted by Ja (x) is defined by
α ∞ 2m
x (–1) m x
Jα ( x) =   ∑   , (Re α > 0)
2 m=0 m !Γ(α + m + 1)  2 
(4) Bessel function of order n of second kind is defined by equation (3.3.5)
J0(x) is a solution of Bessel equation with a = 0 K0(x) is a second solution of Bessel
equation with a = 0 obtained according to theorem 3.2.2 where the roots of indicial equation
r1 = r2 = a = 0. Ja (x) is the first solution of Bessel equation where 2a is not a positive integer
and Kn(x) is the second solution of Bessel equation where a = n a positive integer.
Depending upon the situation choose a and then find the required Bessel function.

EXAMPLES

1
Ex. 1. Suppose f is any solution of x y′′ + xy ′ + x y = 0 for x > 0 and let ψ ( x ) = x 2 φ ( x ) .
2 2

2 1
show that y satisfies the equation x y′′ + ( x + ) y = 0 for x > 0.
2
4

Ans. : Since f is a solution, x 2φ ′′ + x φ ′ + x 2φ = 0


1 1 1
1 –
Let ψ ( x) = x2 φ ( x ) then ψ ′( x ) = x 2 φ + x 2 φ ′
2
3 1 1
1 – –
and ψ ′′( x ) = – x 2 φ + x 2 φ ′ + x 2φ ′′
4
1 3 5
1
x ψ ′′( x) = – x 2 φ + x 2 φ ′ + x 2φ ′′
2
4
1 1
1
= – x 2 φ + x 2 ( x φ ′ + x 2φ ′′)
4
1 1
1
= – x 2 φ + x 2 (– x 2φ ) (Since x 2φ ′′ + x φ ′ + x 2φ = 0)
4
1
1 1
= – ( + x 2 ) x 2 φ = – ( + x 2 )ψ ( x)
4 4
1
Thus, x 2ψ ′′( x) + ( x 2 + )ψ ( x) = 0
4
2 1
and y satisfies the equation x y ′′ + ( x + ) y = 0 for x > 0.
2
4
Ex. 2. Let f be a real valued non-trivial solution of y ′′ + α ( x ) y = 0 on a < x < b
Let y be a real valued non-trivial solution of y ′′ + β ( x ) y = 0 on a < x < b
Here a, b are real valued continuous functions.

Differential Equations (143)


Suppose that β ( x) > α ( x), (a < x < b) Show that if x1 and x2 are successive zeros of f on
a < x < b, then y must vanish at sone point r1, x1 < r1 < x2.
Ans. : Suppose ψ ( x ) ≠ 0 for x1 < x < x2 then either ψ ( x ) > 0 ∀ x ∈ ( x1 , x2 ) or ψ ( x ) < 0
∀ x ∈ ( x1 , x2 ) suppose ψ ( x ) > 0 for x1 < x < x2 .
Since x1 and x2 are successive zeros either φ ( x) > 0 on ( x1, x2 ) or φ ( x) < 0 on ( x1, x2 )
suppose φ ( x) > 0 on ( x1, x2 ) . Then
(ψ φ ′ – φψ ′)′ = ψ φ ′′ – φψ ′′ = –α ( x )ψ φ + β ψ φ
= (β – α )φ ψ

[Since f is solution of y ′′ + α y = 0, φ ′′ + α ( x) φ = 0 similarly ψ ′′( x ) + β ( x )ψ = 0 ]

Thus, (ψ φ ′ – φψ ′)′ = ( β – α ) φ ψ > 0 ( β > α , φ ,ψ > 0) Integration of above inequality


between x1 and x2 gives
[ψ ( x2 ) φ ′( x2 ) – φ ( x2 )ψ ′( x2 )] – [ψ ( x1) φ ′( x1 ) – φ ( x1)ψ ′( x1 )] > 0
But x1 and x2 are zeros of f therefore φ ( x1 ) = φ ( x2 ) = 0 and above inequality becomes
ψ ( x2 ) φ ′ ( x2 ) – ψ ( x1 ) φ ′( x1 ) > 0.
Since φ ( x ) > 0 for x1 < x < x2 and φ ( x1 ) = 0, φ ( x1 – h) < 0 for h > 0. Therefore
φ ( x1 ) – φ ( x1 – h)
φ ′( x1 ) = hLim
→0 >0
h
Similarly φ ′( x2 ) < 0
Let φ ′( x2 ) = – L1 and φ ′( x1 ) = L2 then L1 , L2 > 0, –ψ ( x2 ) L1 – ψ ( x1 ) L2 > 0
i.e. L1ψ ( x2 ) + L2ψ ( x1 ) < 0
But ψ ( x ) > 0 for x1 < x < x2 and L1 , L2 > 0.
This is a contradiction to our assumption that ψ ( x ) > 0 for x1 < x < x2 . Therefore y takes
both positive and negative values in the interval (x1, x2) and hence ∃ there exists r1 ∈ ( x1, x2 )
such that ψ (r1 ) = 0.

Ex. 3. Show that J0 has an infinity of positive zeros.


Ans. : J0 (x) is a solution of differential equation x 2 y′′ + x y′ + x 2 y = 0
1
If ψ ( x ) = x2 J 0 ( x) then by example 1, y satisfies

 1 
y ′′ + 1 + 2  y = 0, ( x > 0)
 4x 
The function satisfies f ( x ) = sin x satisfies y ′′ + y = 0
1
Since 1 + > 1 and sin x = 0 has infinitely many zeroes x = n π , n = 0,1, 2,3,....
4x 2

Differential Equations (144)


 1 
By above example  β ( x ) = 1 + and α ( x ) = 1
 4x 2

1
ψ ( x) = 0 ( x ) has a zero between np and (n + 1) p for n = 0, 1, 2,..... Thus, J0 (x) has
x2 J
infinite number of positive zeros.
1
1
Ex. 4. (a) If λ > 0 and φλ ( x ) = x2J 0 (λ x ) shows that φλ ′′ + φ = – λ 2φλ .
2 λ
4x

(b) If l, m are positive constants, show that


1
(λ 2 – µ 2 ) ∫ φλ ( x ) φ µ ( x ) d x = φλ (1) φ µ ′ (1) – φ µ (1) φλ ′ (1).
0

(c) If λ ≠ µ and J 0 (λ ) = 0, J 0 ( µ ) = 0, show that

1 1
∫ φλ ( x) φ µ ( x) d x = ∫ x J 0 (λ x) J 0 ( µ x) dx = 0
0 0

Ans. (a) : J 0 (x) is solution of x 2 y′′ + xy′ + x 2 y = 0 therefore J 0 (l x) is solution of

y + λ x y& + λ 2 x 2 y = 0 where ‘.’ represents differentiation with respect to l x.


λ 2 x 2 &&
1 1 1
1 –
If φλ ( x) = x 2 J 0 (λ x) then φλ ′ ( x) = x 2J 2 ′
0 ( λ x) + x λ ⋅ J 0 ( λ x)
2
3 1 1
1 – –
φλ ′′ ( x) = – x 2 J 0 (λ x) + x 2 λ J 0′ (λ x) + x 2 λ 2 J 0′′ (λ x)
4
1 1
x J 0 (λ x) + x 2  λ 2 x 2 J 0′′ (λ x) + λ xJ 0′′ (λ x) 
1 2
x 2φλ ′′ ( x) = –
4  
1 1
1
= – x 2 J 0 (λ x ) + x 2  – λ 2 x 2 J 0 (λ x ) 
4  
1
1 
= –x2J 0 (λ x )  + λ 2 x2 
4 
1 
Therefore x 2φλ ′′ ( x ) +  + λ 2 x 2  φλ ( x ) = 0.
4 
1
φλ ′′ ( x ) + 2
φλ ( x ) = – λ 2φλ
4x

1
Ans. (b) : φ µ′′ ( x) + φ ( x) = – µ 2φµ
2 µ
4x
1
φλ ′′ ( x ) + 2 φλ ( x ) = – λ 2φλ .
4x
Differential Equations (145)
Multiply first equation by fl and second equation by fm and subtract these equations.
φ µ ′′ φλ – φλ ′′ φ µ = – µ 2φ µ φλ + λ 2φ µ φλ = (λ 2 – µ 2 )φλ φ µ .

Thus, (
(λ 2 – µ 2 )φλ φ µ = φ µ ′ φλ – φλ ′ φ µ )′
Integrate above equation between 0 to 1. Since
1 1
φλ ( x) = x2J 0 (λ x), φλ (0) = 0 and φ µ ( x) = x2J 0 (µ x), φ µ (0) = 0.
Therefore
1
(λ 2 – µ 2 ) ∫ φλ ( x) φ µ ( x) d x = φ µ ′ (1) φ λ (1) – φλ ′ (1) φ µ (1)
0
1
Ans. (c) : (λ 2 – µ 2 ) ∫ φλ ( x) φ µ ( x) d x = φ µ ′ (1) φ λ (1) – φλ ′ (1) φ µ (1)
0

Since, J 0 (λ ) = 0, φλ (1) = 0 and J 0 ( µ ) = 0 ⇒ φ µ (1) = 0.


1
\ (λ 2 – µ 2 ) ∫ φ λ ( x ) φ µ ( x ) d x = 0
0

1 1 1 1
i.e. ∫ φλ ( x) φ µ ( x) dx = ∫ x2J 0 (λ x) x 2 J 0 (µ x )d x = 0
0 0
1
i.e. ∫ x J 0 (λ x) J 0 ( µ x) dx = 0.
0

Ex. 5. Show that J0¢ (x) = – J1 (x).

∞ (–1) m  x 
2m
Ans. : J 0 ( x) = ∑ 2  
m = 0 m!  2 

∞ (–1) m 2m x 2 m –1
J 0′ ( x ) = ∑ 2
m = 1 m! 2 2m
∞ (–1) m m x 2 m –1
= ∑ 2
m = 1 m! 2 2 m –1
∞ (–1) m +1 (m + 1) x 2 m +1
= ∑ (Replace m by m + 1)
m = 0 ( m + 1)!
2
2 2m +1

x ∞
2m
= –   ∑
(–1) m  x 
 
 2  m = 0 m !( m + 1)!  2 
= – J1 ( x)

Differential Equations (146)


1
Ex. 6. Define, when k is a non-positive integer to be zero. Show that if n is a positive
(k)
integer the formula for J–n(x) gives.
J – n ( x ) = (–1) n J n ( x)

–n ∞ 2m
 x (–1)m  x
Ans. : J – n ( x) =   ∑  
 2 m=0 m! (m – n + 1) 2
–n ∞ 2m
(–1)m  x 
=  
x 1
∑   (As = 0 for k ≤ 0)
2 m=n m! (m – n)!  2  (k )
2 m+ 2 m
∞ (–1) m+ n  x 
–n
=  
x
∑   (Replace m by m + n)
 2 m=0 (m + n)! m !  2 
n ∞ (–1) m
2m
= (–1) n  
x x
∑  
2 m=0 m ! ( m – n)!  2 

= (–1) n J n ( x ).

Ex. 7. Show that


(a) ( xα Jα )′ ( x ) = xα Jα –1 ( x )

(b) ( x –α Jα )′ ( x ) = – x –α Jα +1 ( x )

α ∞ 2m
 x (–1)m  x
Ans. (a) : Jα ( x) =   ∑  
 2 m =0 m! (m + α + 1) 2

x 2α ∞ 2m
(–1)m  x
\ xα Jα ( x) = ∑  
2α m=0 m! (m + α + 1)  2 

1 ∞ (–1)m x 2 m+ 2 α
= ∑
2α m =0 m! (m + α + 1) 22m

1 ∞ (–1) m (2m + 2 α ) x 2m +2α –1


( xα Jα )′ = ∑
2α m=0 m ! (m + α + 1) 2 2m
α –1 ∞ 2m
(–1) m
= x  
x
α x
∑  
2 m=0 m ! (m + α )  2 

= xα Jα –1 ( x )

α 2m
 x (–1) m  x
Ans. (b) : Jα ( x) =   ∑  
 2 m! (m + α + 1)  2 

Differential Equations (147)


∞ 2m
1 (–1)m  x
x –α Jα = ∑  
2α m=0 m! (m + α + 1)  2 

( x –α Jα )′
=
1


m =1
(–1)m 2m x 2m–1
m ! (m + α + 1) 22m

1 ∞ (–1)m+1 2(m + 1) x 2m+1


= ∑
2α m=0 (m + 1)! (m + α + 2) 22m+2

x ∞ (–1)m x2m
=– ∑
2α +1 m = 0 m! (m + α + 2) 2m
α +1 ∞ 2m
 x (–1)m  x
But Jα +1 ( x) =   ∑  
 2 m=0 m! (m + α + 2)  2 
∞ 2m
x (–1)m x
\ – x –α Jα +1 = – ∑  
2α +1 m = 0 m! (m + α + 2)  2 

(
Thus x –α Jα )′ = – x –α Jα +1
8. Show that
(a) Jα –1 ( x ) – Jα +1 ( x) = 2 Jα ′ ( x)
(b) Jα –1 ( x ) + Jα +1( x) = 2 α x –1Jα ( x)

Ans. xα Jα –1 ( x ) = ( xα Jα )′ = α xα –1 Jα + xα Jα′ .......... (i)


– x –α Jα +1 ( x) = ( x –α Jα )′ = –α x –α –1Jα + x –α Jα′ .......... (ii)

(a) Multiply equation (i) by x–a and equation (ii) by xa and add.
Jα –1 ( x ) – Jα +1 ( x ) = (α x –1 – α x –1 ) Jα + Jα ′ + Jα ′ = 2 Jα ′ ( x )

Thus, Jα –1 ( x) – Jα +1 ( x) = 2 Jα ′ ( x)

(b) Multiply equation (i) by x–a and equation (ii) by xa and subtract.
Jα –1 ( x ) + Jα +1( x) = 2 α x –1Jα ( x)
Thus, using results of example 7 we have proved the required result.

Ex. 9. Show that K 0′ ( x) = – K1 ( x)

∞ (–1) m  1 1 1   x
2m
Ans. : K 0 ( x ) = – ∑ 2 
1 + + + ⋅⋅⋅⋅ +    + log x J 0 ( x )
m =1 m!  2 3 m  2

Differential Equations (148)


∞ (–1) m  1 1 1  2m x 2m –1 1
K 0′ ( x) = – ∑  1 + + + ⋅⋅⋅⋅ +  + J 0 ( x) + log x J 0′ ( x)
m =1 m ! 2
 2 3 m  2 2m x
∞ 1  m x 2m –1 1  ∞ (–1) m  x  
2m
(–1) m  1 1
=– ∑ 1 + + + ⋅⋅⋅⋅ +  2 m –1 +  ∑    + log x J 0′
m =1 m! 2  2 3 m 2 x  m = 0 m !2  2  

x ∞ (–1) m
 + 1 + 1 + ⋅⋅⋅⋅ + 1  m x
2 m –1
1  x 2  ∞ (–1) m x 2 m –1
= – ∑  1  + 1 –  + ∑
2 m = 2 m! 2  2 3 m  2 2 m –1 x  4  m = 2 m !2 2 2 m

+ log x J 0′ ( x)
∞ (–1) m x 2 m –1  1
1 x  + 1 + 1 + ⋅⋅⋅⋅ + 1   +
= + + ∑  – m  1  log x J 0′ ( x)
x 4 m = 2 m ! 2 22 m –1  2  2 3 m  

1 x  x  ∞ (–1) m x 2m –2  1  1 1 1 
= + +  ∑  – – m  1 + + + ⋅⋅⋅⋅ +   + log x J 0′
x 4  2  m = 2 m! 2
2 2 m –2
 2  2 3 m –1 

Replace m by m + 1 and use the result J 0′ = – J1 then

1 x  x  ∞ (–1) m+1  x   1 
2m
′ 1 (m + 1)  1 1
K 0 ( x) = + +  ∑   – – 1 + + + ⋅⋅⋅⋅ +  
x 4  2  m = 1 (m + 1)!m !  2   2 (m + 1) (m + 1)  2 3 m 
– log x ⋅ J1 ( x )
2m
(–1) m  x   1 1 
= + +  
1 x 1 x  1 1
∑    m + 1 + 2 1 + 2 + 3 + ⋅⋅⋅⋅ + m  
x 4 2 2 m !( m + 1)!  2   
– log x ⋅ J1 ( x )
= – K1 ( x)

Thus, K 0′ ( x ) = – K1 ( x )

EXAMPLES

1. Let f be any solution for x > 0 of the Bessel equation of order a


x 2 y′′ + xy ′ + ( x 2 – α 2 ) y = 0
1
Put ψ ( x ) = x 2φ ( x ). show that y satisfies equation
 1 –α2 
 
y ′′ + 1 + 4 2  y = 0
 x 
 
2. Show that if a > 0 then Ja has an infinite number of positive zeros.
3. Show that J0¢ satisfies the Bessel equation of order one

Differential Equations (149)


x 2 y′′ + x y′ + ( x 2 –1) y = 0.

1
4. For a fixed a > 0 and l > 0 let φλ ( x) = x 2 Jα (λ x) show that

 1 α2 
 – 
φλ ′′ +  4 2  φλ = – λ 2 φλ
 x 
 
5. If l, m are positive show that
1
(λ 2 – µ 2 ) ∫ φλ ( x ) φ µ ( x )d x = φ λ (1) φ µ′ (1) – φ µ (1) φλ′ (1)
0

Unit 4 : Regular singular points at infinity

At the beginning of chapter 3 we have defined singular points of linear differential equation
of order n on the domain | x | < ¥. In unit 2 of chapter 3 we have discussed the power series
solutions of second order differential equation with regular singular points. These singular points
lie in a finite plane | x | < ¥. Often it is necessary to investigate solution of the differential
equation for large values of | x |. A simple way of doing this is to change the independent variable
1
by its reciprocal x = and study the solution of the resulting equation near t = 0. If the resulting
t
equation possesses the regular singular point of t = 0. We say that the original equation has a
regular singular point at infinity. The results on analytic solution and equations with regular
singular point at t = 0 can be applied to the transformed equation. Analysis of equation at t = 0
gives the analysis of given equation for infinite x.
Let us consider the second order differential equation
L( y ) = y ′′ + a1 ( x) y′ + a2 ( x) y = 0
for large values of | x |.
Suppose f is a solution of L(y) = 0 for | x | > r0 for some r0 > 0.
1
and let φ% (t ) = φ   , a%1 (t ) = a1   , a% 2 (t ) = a2   . These functions will
1 1 1
Define t =
x t t t
1
exist for | t |< .
x0
dφ% (t ) d dx  –1  d φ%
= φ ( x) ⋅ = φ ′( x)  2  ⇒ φ ′( x) = – t 2 .
dt dx dt t  dt

d 2φ% (t )  1  d dx d  –1 
=  – 2  (φ ′ ( x) ) + φ ′( x)  2 
dt 2
 t  dx dt dt  t 

1 ′′ 2
= 4
φ ( x) + 3 φ ′ ( x)
t t
Therefore

Differential Equations (150)


d 2φ% (t )
φ ′′( x) = t 4 – 2t φ ′ ( x)
dt 2
d 2φ%  2 d φ% 
= t4 – 2t  –t 
dt 2  dt 
and
L(φ ) = L% (φ% ) = t 4φ%′′(t ) + 2t 3φ%′ (t ) + a%1 (t ) (– t 2 ) φ%′ (t ) + a%2 (t ) φ% (t )

= t 4 φ%′′(t ) +  2t 3 – a%1 (t ) t 2  φ%′(t ) + a%2 (t ) φ% (t )


 
L(φ ) = 0 gives L% ( φ% ) = 0. Thus, φ% satisfies

L% ( y ) = t 4 y′′ +  2t 3 – a%1 (t ) t 2  y ′ + a%2 (t ) y = 0 .......(3.4.1)


 
Where the prime denotes differentiation with respect to t.
Convertsely if φ% satisfies L% ( y) = 0 the function f will satisfy L(y) = 0. The equation
1
(3.4.1) is called the induce equation associated with L(y) = 0 and the substitution x = .
t
Definition 3.4.1 :
We say that infinity is a regular singular point for L ( y ) = y ′′( x ) + a1 ( x ) y′( x )
+ a2 ( x ) y ( x ) = 0 if t = 0 is a regular singular point of

L% ( y ) = t 4 y ′′(t ) + 2t 3 – t 2a%1 (t )  y ′(t ) + a%2 (t ) y (t ) = 0


 
L% ( y ) is equivalent to the equation
2t – a%1(t ) a% (t )
y′′(t ) + y ′(t ) + 2 y (t ) = 0
t2 t4
On comparing this equation with the equation in definition 3.1.3(b) we see that
2t – a%1 (t ) a%2 (t )
p (t ) = and q (t ) =
t2 t4
∞ ∞ 1
If a%1 (t ) = t ∑ αk t k and a%2 (t ) = t ∑ β k t where the series converge for | t |<, r0 > 0,
2 k
k =0 k =0 r0
then the denomiantor of p(t) will not contain a factor t to a power higher than one and the
denominator of q(t) will not contain a factor t to a power higher than two. By definition
3.1.3(b) t = 0 is a regular singular point of L% ( y ) = 0 and therefore infinity is a regular singular
point of L(y) = 0.

EXAMPLES

Ex. 1. Check whether infinity is regular singular point of x 2 y′′ + a x y′ + b y = 0, where a, b


are constants.
1
Ans. : Put x= .
t
Differential Equations (151)
d y dt
y ′( x) = ⋅ = – t 2 y& (t ) where .dot (.) represents differentiation with respect to t.
dt d x
d dt
y ′′ ( x ) = [ y( x)] = d [ y′( x) ] =  –t 2 &&y (t ) – 2 t y& (t )  (–t 2 )
dx dt dx
= t 4 &&
y (t ) + 2 t 3 y& (t )
1  4 && & (t ) + a ⋅ 1  –t 2 y& (t ) + by
x 2 y′′( x) + ax y′ + by = t y (t ) + 2t 3
y
t2   t 
y (t ) + [2 – a ] t y& (t ) + b y
= t 2 &&
L [ y ( x)] = 0 implies
2–a b
&&
y (t ) + y (t ) + 2 y (t ) = 0.
t t
This equation is of form y ′′ + p y ′ + q y = 0. Since denomination of p contains a factor

t r , r >/ 1 and denominator of q contain a factor t r , r >/ 2, t = 0 is a regular singular point. Thus,
infinity is a regular singular point of the given differential equation.

Ex. 2. Show that infinity is not a regular singular point for the equation
y ′′ + a y ′ + b y = 0
where a, b are constants, not both zero.

Ans. : y ′( x) = – t 2 y& (t )

y ′′( x ) = t 4 &&
y (t ) + 2 t 3 y& (t )
and y ′′( x ) + a y ′( x ) + b y ( x ) = 0 gives
t 4 &&
y (t ) + 2 t 3 y& (t ) – a t 2 y& (t ) + b y (t ) = 0.
Therefore
2t – a b
y (t ) + 2 y& (t ) + 4 y (t ) = 0
&&
t t
2t – a b
Here p (t ) = 2 and q (t ) = 4
t t
2
[If a = b = 0 then p (t ) =and q (t) = 0. Since denominator of p(t) contain a factor t r , r >/ 1
t
and denominator of q(t) contains a factor t r , r = 0 >/ 2 t = 0 is a regular singular point and
infinity is regular singular point of the equation.]
2t – a
Since either a or b is non-zero, p (t ) = 2
contains the determinator tr with r = 2 > 1 or
t
b
q (t ) = 4
contains the denominator tr with r = 4 > 2. Therefore t = 0 is an irregular singular
t
point of the transformed equation and infinity is an irregular singular point of the given equation.

Differential Equations (152)


Ex. 3. Show that infinity is a regular singular point for the Legendre equation
(1– x 2 ) y′′ – 2 x y ′ + α (α + 1) y = 0
where a, b are constants, not both zero.

1
Ans. : y ′( x ) = – t 2 y& (t ), y ′′( x) = t 4 &&
y (t ) + 2t 3 y& (t ), x =
t
\ Legendre equation becomes,
 1 4
 1 – 2  t &&
 t 
1
( )
y + 2 t 3 y&  – 2 –t 2 y& + α (α + 1) y = 0
 t
y + 2(t 2 – 1) t y& + 2 t y& + α (α + 1) y = 0
(t 2 – 1) t 2 &&
2t α (α + 1)
&&
y+ y& + 2 2 y=0
2
(t –1) t (t –1)
2t
Here p (t ) = 2
contains a factor t r in the denominator with r = 0 >/ 1 and
(t –1)
α (α + 1)
q (t ) = 2 2
contains a factor tr in the denominator with r = 2 >/ 2.
t (t –1)
Therefore by definition 3.1.3(b) t = 0 is a regular singular point of the transformed equation
and infinity is a regular singular point of a given equation.

4. Find two linearly independent solutions of the equation (1 – x 2 ) y′′ – 2 x y ′ + 2 y = 0 of



the form x – r ∑ ck x – k valid for | x | > 1
k =0

1
Ans. : Put x = then
t

y ′ ( x ) = –t 2 y& (t ), y ′′( x ) = t 4 &&


y + 2 t 3 y& (t )
Given equation becomes

 1 4
 1– 2  t &&
 t 
1
(
y + 2 t 3 y&  – 2 –t 2 y& + 2 y = 0
 t
)
or (t 2 –1) t 2 &&
y + 2 t y&  + 2 t y& + 2 y = 0

L( y ) = t 2 (t 2 –1) &&
y + 2 t 3 y& + 2 y = 0
From example 2 we observe that x = 0 is a regular singular point of L(y) = 0.
Let f be a solution of L(y) = 0 of the form

φ (t ) = t r ∑ ck t k
k =0

Differential Equations (153)


∞ ∞ ∞
L (φ ) = t 2 (t 2 –1) ∑ ( k + r ) (k + r –1) ck t k +r –2 + 2t 3 ∑ ( k + r ) t k +r –1 + 2 ∑ ck t k +r
k =0 k =0 k =0

∞ ∞
= ∑ (k + r ) (k + r –1) ck t k +r + 2 – ∑ (k + r ) (k + r –1) ck t k +r
k =0 k =0

∞ ∞
+2 ∑ (k + r ) t k + r + 2 + 2 ∑ ck t k + r
k =1 k =0


= ∑ [(k + r )(k + r –1) + 2(k + r )] ck t k + r + 2
k =0


– ∑ [(k + r ) (k + r – 1) – 2 ]ck t k +r
k =0


= ∑ [(k + r ) (k + r + 1)] ck t k +r + 2 – (r (r –1) – 2) c0 t r – ((r + 1) (r ) – 2 ) c1t r +1
k =0


– ∑ [(k + r ) (k + r –1) – 2] ck t k +r
k =2


= –  r 2 – r – 2  c0 t r –  r 2 + r – 2  c1 t r +1 – ∑ [(k + r ) (k + r –1) – 2] ck t k +r
    k =2


+ ∑ [(k + r – 2) (k + r –1) ck –2 ]t k + r
k =2

( ) (
= – r 2 – r – 2 c0 t r – r 2 + r – 2 c1 t r +1)

–∑ {[(k + r) (k + r –1) – 2] ck – [(k + r – 2) (k + r –1)]ck –2}t k +r
k =2

The indicial equation is

q ( r ) = r 2 – r – 2 = 0 gives r = –1, 2 Since r1 – r2 = 2 + 1 = 3 a positive integer we try a


series solution using the smallest root, r = – 1.
At r = – 1, L(f ) = 0 implies
0.c0 = 0, 2c1 = 0 and
[(k –1)(k – 2) – 2]ck – (k – 3) (k – 2) ck –2 = 0 k = 2,3, 4,5,.....
k = 2, – 2c2 = 0 gives c2 = 0
k = 3, 0.c3 = 0 that is c3 is orbitrary
Thus, c0 and c3 are arbitrary whereas c1 = c2 = 0.
Since c2 = 0, all even terms c2k = 0, k = 1, 2, 3,.... and

Differential Equations (154)


(k – 3) (k – 2)
ck = ck –2 k = 4,5,....
(k –1) (k – 2) – 2

In particular
(2k – 2) (2 k – 1)
c2 k +1 = c2k –1 k = 2,3, 4,....
2k (2k –1) – 2
(2k – 2) (2k – 1)
= c2 k –1
(2k – 2) (2k + 1)
2k –1
= c2k –1 , k = 2,3, 4,.....
2k + 1
 2 k –1   2 k – 3   2 k – 5   2 k – 7  1
c2 k +1 =      ⋅⋅⋅⋅⋅ c3
 2k + 1   2k – 1   2k – 3   2k – 5  3
c3
=
2k + 1
Thus we get a solution
φ (t ) = c0t –1 + c3t 2 + c5t 4 + c7t 6 + ....

= c0t –1 + ∑ c2 k +1 t 2 k
k =1
∞ c3
= c0t –1 + ∑ t 2k
k = 1 (2k + 1)


t 2k
= c0t –1 + c3 ∑
k =1 (2k + 1)
Thus, we get two solutions
∞ t 2k
φ1 (t ) = t –1 and φ2 (t ) = ∑
k =1 2k + 1

Let φ2 (t ) = ∑ d k (t ).
k =1
By ratio test
d k +1 (t ) t 2 k + 2 2k + 1 (2k + 1) t 2
= × 2k =
d k (t ) 2k + 3 t (2k + 3)
2k + 1
Since Lt
k →∞ =1
2k + 3
The series converger for | t | < 1.
1
But x = , therefore
t

Differential Equations (155)


∞ x –2k
φ1 ( x ) = x and φ2 ( x) = ∑ are two solutions of given equation. Second
k =1 2k + 1
converges for | x | > 1.
Ex. 5. For each equation locate and classify all its singular points.

(a) x 3 ( x –1) y′′ + ( x –1) y′ + 4 x y = 0

(b) x 2 ( x 2 – 4) y′′ + 2 x 3 y ′ + 3 y = 0
(c) y′′ + xy = 0

(d) x 2 ( x – 4) 2 y′′ + 3 x y′ – ( x – 4) y = 0

Ans. : In chapter 3 Unit I, example 3, we have classified all its singular points in a finite plane.
It remains to check whether infinity is a singular point and whether it is a regular singular
point.
(a) a0 ( x) = x 3 ( x –1) = 0 gives x = 0, x = 1 are singularities x = 0 is irregular singular point
whereas x = 1 is a regular singular point.

put x=
1
t
then
dy dt
y ′( x ) =
dt dx
⋅ = y& (t ) (–t 2 ) and y ′′( x ) =
dt
(
d &
)
y (t ) (–t 2 )
dt
dx
=  –t &&
2
y (t ) – 2 t y& (t )   –t  so y ′′( x ) = t 4 &&
2
y + 2 t 3 y&.
  
L( y ) = x 3 ( x –1) y′′ + ( x –1) y′ + 4 x y
1 1   4
y + 2t 3 y&  +  –1 (–t 2 ) y& + 4 y
1 1
= 3t
–1 t &&
t     t  t

y +  – 2 – t + t 2  y& + y
2 4
= (1 – t ) &&
t  t
2 – (2 + t ) t + t 3 4
L ( y ) = (1 – t ) &&
y+ y& + y
t t
L(y) = 0 can be put in the form
(t – 1) (t 2 – 2) 4
&&
y+ y& + y =0
(1 – t ) t t (1 – t )
2
t –2 4
or &&
y+ y& + y=0
t t (1 – t )
t2 – 2 4
This equation is of the type y ′′ + p (t ) y ′ + Q (t ) y = 0 where p (t ) = and Q (t ) =
t t (1 – t )
Since the denominator of p(t) contains a factor t r , for r = 1 >/ 1 and q(t) contains a
denominator t , for r = 1 >/ 2, t = 0 is a regular singular point.
r

Thus, x = 1 and infinity are regular singular whereas x = 0 is irregular singular point.

Differential Equations (156)


1
(b) Put x = then y ′( x ) = – t 2 y& and y ′′( x ) = t 4 &&
y + 2t 3 y&
t

L( y) = x 2 ( x 2 – 4) y′′ + 2 x3 y′ + 3 y
11  2
= 2 2
– 4  t 4 &&
y + 2t 3 y&  + 3 (– t 2 y& ) + 3 y
t t    t
2 2
= (1 – 4t 2 ) &&
y +  – 8t –  y& + 3 y
t t

= (1– 4t 2 ) &&
y – 8 t y& + 3 y

Since t = 0 is not a singular point of (1 – 4 t ) &&


y – 8t y& + 3 y = 0, infinity is not a singular
2

point of the given equation.

1
(c) y ′′ + x y = (t 4 &&
y + 2t 3 y& ) + y.
t
1
Therefore L( y ) = t &&
y + 2t 3 y& + y = 0 can be written in the form &&
4
y + p (t ) y& + q (t ) y = 0
t
2 1
where p (t ) = and q (t ) = 5 .
t t
Here t = 0 is a singular point but since the denominator of q(t) contains a factor
tr, r = 5 > 2, t = 0 is not a regular singular point.
Since t = 0 is irregular singular point infinity is irregular singular point of the equation
y ′′ + x y = 0.
(d) x = 0 is regular singular point and x = 4 is irregular singular point.
1
Put x = then y′ = –t 2 y& , y′′ = t 4 &&
y + 2 t 3 y&
t
L ( y ) = x 2 ( x – 4) 2 y′′ + x y ′ – ( x – 4) y
2
= 2  – 4  t 4 &&
y + 2t 3 y&  + [–t 2 y& ] –  – 4  y
1 1 1 1
t t    t t 
 2(4t –1) 2 
– t  y& + 
4t –1 
= (4t –1) 2 &&
y+ y
 t   t 

 31t 2 –16t + 1   4t –1  y
= (4t – 1)2 &&
y+  y& +  
 t   t 

31t 2 –16t + 1
y + p (t ) y& + q (t ) y = 0 where p (t ) =
L(y) = 0 can be written in the form && and
(4t –1) 2 ⋅ t

Differential Equations (157)


1
q (t ) =
t (4t –1)
Since t = 0 is a singularity of p(t) and q(y) and is a simple pole by definition 3.1.3(a)
t = 0 is a regular singular point and infinity is regular singular point of the given equation.

Differential Equations (158)


Chapter
4
Existence and Uniqueness of
Solutions to First Order Equations

Contents :
Unit 1 : The method of successive approximations.

Unit 2 : Convergence of the successive approximations.

Introduction :
In the last three chapters we have seen the methods of finding a solution to the given linear
differential equations. For linear differential equation with constant coefficients there is a method
to find all the solutions whereas for linear equations with variable coefficients, there are very
few types of equations whose solutions can be expressed in terms of elementary functions and
therefore we go for power series solutions. All the equations considered so far were linear
differential equations.
In this chapter we consider the general first order equation y ′ = f ( x, y ) where f is some
continuous function (need not be linear in y) Only in special cases it is possible to find explicit
analytic expressions for the solutions of y ′ = f ( x, y ) .
Our main purpose in this chapter is to prove that a wide class of initial value problems
y ′ = f ( x, y ), y ( x0 ) = y0
has a solution. Though it may not be possible to find out the exact solution, it is feasible to
construct a sequence of approximate solutions that may converge to the exact solution.

Unit 1 : Methods of successive approximations


In this unit we study the general problem of finding solutions of the equation
y ′ = f ( x, y ), y ( x0 ) = y0 .......... (4.1.1)
Where f is any continuous real valued function defined on some rectangle
R = {( x, y ) / | x – x0 | ≤ a, | y – y0 | ≤ b, a, b > 0}
in the real (x, y) plane.

A function f is a solution of equation (4.1.1) if φ ( x0 ) = y0 and φ ′ ( x) = f ( x, φ ( x) ).

Differential Equations (159)


Theorem 4.1.1
A function f is a solution of the initial value problem (4.1.1) on an interval I if and only
if it is a solution of the integral equation
x
y = y0 + ∫ f (t , y) d t on I .......... (4.1.2)
x0

Proof : Suppose f is a solution of the initial values problem on I. Then


φ ′ (t ) = f (t , φ (t ) ) and φ ( x0 ) = y0
Since f is continuous on I and f is continuous on R, the function F defined by
F (t ) = f (t , φ (t ))
is continuous on I.
φ ′ (t ) = f (t , φ (t )) and φ ( x0 ) = y0
On integrating above equation between x0 and x we get
x x
∫ φ ′(t ) dt = ∫ f (t , φ (t ) ) dt
x0 x0
x
φ ( x ) – φ ( x0 ) = ∫ f (t , φ (t ) ) dt
x0
x
or φ ( x) = y0 + ∫ f (t , φ (t ) ) dt (as f (x0) = y0)
x0
Thus f is solution of (4.1.2)
Conversely suppose f satisfies (4.1.2) on I that is
x
φ ( x ) = y0 + ∫ f (t , φ (t )) dt
x0

Differentiate this equation with respect to x and use the fundamental theorem of integral
calculus. The integral equation becomes
φ ′( x ) = f ( x, φ ( x )) for all x ∈ I .
From (4.1.2) it is obvious that φ ( x0 ) = y0 .
Thus f is a solution of equation (4.1.1).
Successive approximate solutions
As a first approximation to a solution defined
φ0 ( x ) = y0 .
Then f 0 satisfies an initial condition but does not in general satisfy the differential equation.
Since f 0 is a first approximate solution, substitute y = f0 in equation (4.1.2) to generate
second approximate solution. Call this solution as f1 then
x
φ1 ( x) = y0 + ∫ f (t , φ0 (t )) dt.
x0

Differential Equations (160)


Clearly φ1 ( x0 ) = y0 . Therefore f1 satisfies initial condition.
If we continue the process and define successively
φ0 ( x) = y0 ,
x
φk +1 ( x ) = y0 + ∫ f (t ,φk (t )) dt (k = 0,1, 2,......) .......... (4.1.3)
x0


We get a sequence of functions {φ k }k = 0 . If this sequence converges then it may happen
that the limit function will turn out to be the solution of differential equation (4.1.1).
We now show that there is an interval I containing x0 where all the functions φk , k = 1, 2,.....
exist. Since f is continuous on a compact set R, it is bounded on R, that is there exists a constant
M > 0 such that
| f ( x, y ) | ≤ M for all ( x, y ) ∈ R.

Theorem 4.1.2
The successive approximations defined by (4.1.3) exist and are continuous on
I = {x / | x – x0 | ≤ α where α = min {a, b / M }}
and for x ∈ I , ( x, φ k ( x)) ∈ R.
The function fk satisfy
| φ k ( x) – y0 |≤ M | x – x0 | for all x in I
Proof : We will prove this result by mathematical induction,
(i) Clearly φ0 ( x ) = y0 is continuous on I and
| φ0 ( x) – y0 | = 0
Thus the theorem is true for k = 0.
x
(ii) φ1 ( x) = y0 + ∫ f (t , φ0 (t )) dt
x0
x
= y0 + ∫ f (t , y0 ) d t
x0

Since f is continuous and continuous function is integrable, φ1 ( x ) exist.


x x
φ1 ( x ) – y0 = ∫ f (t , y0 )dt ≤ ∫ f (t , y0 ) dt ≤ M | x – x0 |
x0 x0

Therefore | φ1 ( x) – y0 | ≤ M | x – x0 |
Since f is continuous on R the function F0 defined by
F0 (t ) = f (t , y0 )
is continuous on I. Therefore f1 defined by

Differential Equations (161)


x
φ1 ( x ) = y0 + ∫ F0 (t )dt
x0
is continuous on I.
The theorem is true for k = 1.
(iii) Assume that the theorem is true for fk
(iv) To prove the result for fk+1

We know that (t , φ k (t ) ) ∈ R for t Î I.


Since f is continuous on R and fk is continuous on I,
Fk (t ) = f (t , φk (t ) )
exist for t Î I and Fk is continuous. The function fk+1 given by
x x
φk +1 ( x) = y0 + ∫ f (t , φk ) dt = y0 + ∫ Fk (t )dt
x0 x0
exists and is continuous function on I.
x x
φk +1 ( x) – y0 = ∫ Fk (t ) dt ≤ ∫ Fk (t ) dt ≤ M | x – x0 |
x0 x0

(Since Fk (t ) = f (t , φ k ) ≤ M )
Thus fk+1 exist is continuous and satisfies the required inequality.
Definition : Let f be a function defined for (x, y) in a set S. We say f satisfies a Lipschitz
condition on S if there exists a constant K > 0 such that
| f ( x, y1 ) – f ( x, y2 ) | ≤ K | y1 – y2 |
for all ( x, y1 ), ( x, y2 ) ∈ S . The constant K is called Lipschitz constant.

Theorem 4.1.3 :
Suppose S is either a rectangle
| x – x0 | ≤ a, | y – y0 | ≤ b (a, b > 0);
or a strip
| x – x0 | ≤ α , | y | < ∞ (a > 0)
and that f is real valued function defined on S.
∂f
Such that exists, is continuous on S and
∂y
∂f
( x, y ) ≤ K , for ( x, y ) ∈ S and for some K > 0. Then f satisfies a
∂y
Lipschitz condition on S with Lipschitz constant K.
y1
∂f
Proof : f ( x, y1 ) – f ( x, y2 ) = ∫ ∂ y ( x, t ) d t
y2

Differential Equations (162)


Therefore
y1
∂f
f ( x, y1 ) – f ( x, y2 ) = ∫ ∂ y ( x, t ) d t
y2

y1
∂f
≤ ∫ ( x, t ) d t
y2 ∂ y

y1
≤ ∫ K dt
y2

≤ K y1 – y2
Thus, f ( x, y1 ) – f ( x, y2 ) ≤ K | y1 – y2 | for all (x, y1) (x, y2) in S.

EXAMPLES

1. Consider the initial value problem


y ′ = 3 y + 1, y (0) = 2.
(a) Show that all the successive approximations f0, f1, f2,..... exist for all real x.
(b) Compute the first four approximations f0, f1, f2, f3 to the solution.
(c) Compute exact solution.
(d) Compare exact and approximate solution.
Answer :
(a) We will prove this result by induction on k.
k = 0,
φ0 ( x ) = y0 = 2
f0 exist and is continuous.
Assume that fK exist and is continuous.
x
φk +1 = y0 + ∫ f (t , φk (t )) dt
x0
x
= y0 + ∫ [3φk (t ) + 1] dt
x0
x
= y0 + 3 ∫ φ k (t ) d t + ( x – x0 )
x0

Since fk is continuous, fK is integrable.


Therefore fk+1 exist and is continuous.
Thus, f0, f1, f2,..... exist for all real x.

Differential Equations (163)


(b) φ0 ( x ) = 2
x
φ1 ( x ) = y0 + ∫ f (t , φ0 (t )) d t
x0

Here f (t , y ) = 3 y + 1 . Therefore
x
φ1 ( x) = 2 + ∫ [3 φ0 + 1] d t
0
x
= 2 + ∫ 7dt = 2 + 7 x
0
x
φ2 ( x) = 2 + ∫ [3 φ1 + 1] d t
0
x
= 2 + ∫ [3 (2 + 7 t ) + 1] d t
0
x
= 2 + ∫ (21t + 7) d t
0

21x 2 21x 2
= 2+ + 7x = 2 + 7x + .
2 2
x
φ3 ( x) = 2 + ∫ [3 φ2 (t ) + 1] d t
0
x
 
= 2 + ∫ 3  2 + 7 t + t 2  + 1 d t
21
0  2  
x
 63 
= 2 + ∫ 7 + 21t + t 2  d t
0  2 
x 2 63 x 3
= 2 + 7 x + 21 +
2 2 3
21 21
= 2 + 7 x + x 2 + x3 .
2 2

(c) y′ – 3 y = 1

y = e3 x  ∫ 1e –3 x d x + c1 

 e –3x 
= e3 x  + c1 
 –3 
1
= – + c1 e3 x
3

Differential Equations (164)


1 7
Since at x = 0, y = 2, 2 = – + c1 e 0 i.e. c1 = . and
3 3
1 7
y ( x) = – + e3 x
3 3
1 7 (3 x) 2 
= – + 1 + 3x + + ⋅⋅⋅⋅
3 3 2 
21 2 63 4
= 2 + 7x + x + x + ⋅⋅⋅⋅
2 8
(d) f0, f1, f2 , f3 are respectively first, first 2, first 3 and first 4 terms of the series solution
1 7
y = – + e 3 x.
3 3

2. For each of the following problems compute the first four successive approximations
f0, f1, f2 , f3 .

(a) y ′ = x 2 + y 2 , y (0) = 0 (b) y ′ = 1 + x y , y (0) = 1

(c) y ′ = y 2 , y (0) = 1

Answers :
(a) φ0 ( x ) = y0 = 0, f ( x, y ) = x 2 + y 2
x
φ1 ( x) = y0 + ∫ f (t , φ0 (t ) ) d t
x0

( )
x
= 0 + ∫ t 2 + 02 d t
0

x3
=
3
x
φ2 ( x) = y0 + ∫ f (t , φ1 (t ) ) d t
x0

x 
t3 
= 0 + ∫ f  t,  d t
0  3 

x  t3 
2
= ∫ t +  
2  dt
0 3 

x3 x7
= +
3 63
x 2
2 t t7  
3

φ3 ( x ) = ∫ t +  +  d t
0
  3 63  
Differential Equations (165)
x 
t 6 t14 2 t10 
= ∫ t 2 + + 2 +  dt

0  9 63 3 × 63 

x3 x 7 x15 2 x11
= + + +
3 7.9 15 × 63 × 63 11× 3 × 63

x3 x 7 2 x11 x15
= + + +
3 7.9 11.3.68 15.63.63

(b) φ0 ( x ) = 1 f ( x, y ) = 1 + x y
x
φ1 ( x ) = y0 + ∫ (1 + t φ0 ) d t
x0
x
= 1 + ∫ [1 + t ] d t
0

x2
=1+ x +
2
x  t 2 
φ2 ( x ) = 1 + ∫ 1 + t 1 + t +   d t
0  2  
x 
t3 
= 1 + ∫ 1 + t + t 2 +  d t
0  2

x 2 x3 x 4
=1+ x + + +
2 3 8
x  t 2 t 3 t 4 
φ3 ( x ) = 1 + ∫ 1 + t  1 + t + + +   d t
0  2 3 8  
x 
t3 t 4 t5 
= 1 + ∫ 1 + t + t 2 + + +  d t
0  2 3 8

x 2 x3 x 4 x5 x6
=1+ x + + + + +
2 3 8 15 48

(c) φ0 ( x) = y0 = 1 f ( x, y) = y 2
x
φ1 ( x) = 1 + ∫ [(1) 2 ] d t
0
= 1+ x
x
φ2 ( x) = 1 + ∫ [1 + t ]2 d t
0

Differential Equations (166)


2 x 2 x3
= 1+ x + +
2 3
x3
= 1 + x + x2 +
3
x 2
t3 
φ3 ( x ) = 1 + ∫ 1 + t + t +  d t 2

0 3
x
t6  t3 t 4 t5 
= 1 + ∫ 1 + t 2 + t 4 + + 2  t + t 2 + + t 3 + +   d t
0 9  3 3 3  

x3 x5 x7  x 2 x3 x 4 x 4 x5 x6 
= 1+ x + + + +2  + + + + + 
3 5 7 ⋅9  2 3 12 4 15 18 

2 1 x6 x7
φ3 ( x ) = 1 + x + x 2 + x 3 + x 4 + x 5 + +
3 3 9 63

3. Consider the problem


y′ = x 2 + y 2 y(0) = 0
on R : | x | ≤ 1, | y | ≤1

(a) Compute an upper bound M for f ( x, y) = x 2 + y 2 on R

(b) On what interval containing x = 0 will all the sucdessive approximations exist and be
such that their graphs are in R.
Answers :
(a)
M= sup
R f ( x, y )
= sup
| x|≤1, | y|≤1 ( x
2
+ y2 )
=2
(b) By theorem 4.1.2
 
I =  x / | x – x0 | ≤ α where α = min  a,
b

  M 
Here, x0 = 0, y0 = 0, a = b = 1 and M = 2
 1 
\ I =  x / | x | ≤ α where α = min  1,  
  2 

\ {
I = x/| x| ≤
1
2
.}
Differential Equations (167)
4. By computing appropriate Lipschitz constants show that the following functions satisfy
Lipschilz conditions on the set S.
(a) f ( x, y ) = 4 x 2 + y 2 on S = {( x, y ) / | x | ≤ 1, | y | ≤ 1}

(b) f ( x, y ) = x 2 cos 2 y + y sin 2 x on S = {( x, y ) / | x | ≤ 1, | y | < ∞}

Answers :
(a) f ( x, y ) = 4 x 2 + y 2

f ( x, y1 ) – f ( x, y 2 ) =  4 x 2 + y12  –  4 x 2 + y 22 
   

= y12 – y22

= y1 + y2 y1 – y2
But | y |≤1 ∴ | y1 | ≤ 1 and | y2 | ≤ 1
f ( x, y1 ) – ( f ( x, y2 ) ≤ (| y1 | + | y2 |) | y1 – y2 |
≤ 2 | y1 – y 2 |
Therefore Lipschitz constant K = 2.

(b) f ( x, y ) = x 2 cos 2 y + y sin 2 x

f ( x, y1 ) – f ( x, y2 ) =  x 2 cos 2 y1 + y1 sin 2 x  –  x 2 cos 2 y 2 + y2 sin 2 x 


   

( )
= x 2 cos2 y1 – cos 2 y2 + sin 2 x ( y1 – y2 )

≤ | x |2 cos 2 y1 – cos 2 y2 + sin 2 x y1 – y2

≤ cos 2 y1 – cos 2 y2 + y1 – y2
By mean value theorem f (b) – f (a) = f ′(c ) (b – a )
cos2 y1 – cos 2 y2 = –2cos y sin y ( y1 – y 2 )
Therefore
f ( x, y1 ) – f ( x, y2 ) ≤ 2 cos y sin y y1 – y2 + y1 – y2
≤ 3 y1 – y2
Therefore k = 3 is a Lipschitz constant.
5. (a) Show that the function f given by
f ( x, y ) = x 2 | y |
Satisfies Lipschitz condition on R = {( x, y ) | x | ≤ 1, | y | ≤ 1}
∂f
(b) Show that does not exist at (x, 0) if x ¹ 0.
∂y
Differential Equations (168)
Answer :
(a) f ( x, y1 ) – f ( x, y2 ) = x 2 | y1 | – x 2 | y 2 |

≤ | x 2 | | y1 – y2 |
≤ 1| y1 – y2 |
Thus, function satisfies Lipschitz condition with Lipschitz constant k = 1.

∂f
(b) Since | y | is not differentiable at y = 0, do not exist at (x, 0) unless x = 0 if x = 0 then
∂y
the function itself is zero.

EXERCISE

1. Compute Lipschitz constant for the following functions.


(a) f ( x, y ) = a ( x ) y 2 + b ( x ) y + c ( x) on S = {| x | ≤ 1, | y |< 2} (a, b, c are continuous functions
on | x | ≤ 1 )
(b) f ( x, y ) = a ( x) y + b ( x ) on S = {( x, y ) / | x | ≤ 1, | y | < ∞} (a, b are continuous functions on
| x |≤ 1)

(c) f ( x, y ) = x 3e – x y on S = {( x, y ) / 0 ≤ x ≤ a, | y | < ∞}
2

2. (a) Show that the function f given by


1
f ( x, y ) = y2
does not satisfy Lipschitz condition on
S = {( x, y ) / | x | ≤ 1, 0 ≤ y ≤ 1}
(b) Show that this f satisfies a Lipschitz condition on any rectangle R of the form
R = {( x, y ) / | x | ≤ a, b ≤ y ≤ c, a, b, c > 0}

3. Show that the function f given by


f ( x, y ) = 0, if x = 0, | y | ≤ 1
= 2 x, if 0 < | x | ≤ 1, –1 ≤ y < 0
4y
= 2x – , if 0 < | x | ≤ 1, 0 ≤ y ≤ x 2
x
= –2 x if 0 < | x | ≤ 1, x 2 ≤ y ≤ 1
does not satisfy a Lipschitz condition on R = {( x, y) / | x | ≤ 1, | y | ≤ 1}.

4. Determine the bound for the function given by f ( x, y ) = 1 – 2 x y

{ 1
on S = ( x, y ) / | x | ≤ , | y | ≤ 1 .
2 }
Differential Equations (169)
Unit 2 : Convergence of successive approximations
In the last unit we have found the successive approximate solutions to a differential equation
(4.1.1). In this unit let us prove that this sequence of successive approximate solutions actually
converges to the exact solution of differential equation (4.1.1).

Theorem 4.2.1 : (Existence Theorem)


Let f be a continuous real valued function on the rectangle
R = {( x, y ) / | x – x0 | ≤ a, | y – y0 | ≤ b, a > 0, b > 0}
and let f ( x, y ) ≤ M for all (x, y ) ∈ R.
Suppose f satisfies a Lipschitz condition with Lipschitz constant K in R. Then the successive
approximations.
x
φ ( x0 ) = y0 , φ k +1 ( x) = y0 + ∫ f (t , φ k (t )) d t , k = 0,1, 2,3,.....,
x0

Converge on the interval I = {x / | x – x0 | ≤ α } where α = min a,


the initial value problem (4.1.1)
b
M { } to a solution f of

y ′ = f ( x, y ), y ( x0 ) = y0 on I

Proof (a) : Convergnce of {fk}


Since the function fk can be written as

φk ( x) = φ0 + (φ1 – φ0 ) + (φ2 – φ1 ) + (φ3 – φ2 ) ⋅⋅⋅⋅ + (φk – φk –1 )


k
φk ( x) = φ0 ( x) + ∑ φ p ( x) – φ p –1 ( x) 
p =1

The sequence f k converges, that is Lim


k →∞ φk exists if and only if the series

φ0 ( x) + ∑ φ p ( x) – φ p –1 ( x)  is a convergent series.
p =1

By theorem 4.1.2 the functions fp all exist, each is continuous on I and ( x,φ p ( x)) ∈ R for
x in I.
Moreover φ1 ( x ) – φ0 ( x ) ≤ M | x – x0 | for x in I

 x   x 
φ2 ( x ) – φ1 ( x ) =  y0 + ∫ f (t , φ1 (t )) d t  –  y0 + ∫ f (t , φ0 (t )) d t 
 x0   x0 

x
= ∫ [ f (t , φ1(t )) – f (t, φ 0 (t )) ] dt
x0

Differential Equations (170)


x
Therefore φ2 ( x) – φ1 ( x) ≤ ∫ f (t , φ1 ) – f (t , φ0 ) d t
x0
Since f satisfies Lipschitz condition with constant K
f (t , φ1 ) – f (t , φ 2 ) ≤ K φ1 – φ 2 and we have
x
φ2 ( x) – φ1 ( x) ≤ K ∫ φ1 (t ) – φ0 (t ) d t
x0

But φ1 ( x) – φ0 ( x) ≤ M x – x0 for x in I
x
Therefore φ2 ( x) – φ1 ( x) ≤ K ∫ M t – x0 d t
x0

( x – x0 ) 2
and φ2 ( x) – φ1 ( x) ≤ K M .
2
By mathematical induction we will prove that
| x – x0 | p
φ p ( x ) – φ p –1 ( x ) ≤ M K p –1 for every x in I............ (4.2.1)
p!
We have seen that this inequality is true for p = 1 and p = 2. Let us assume the result for
p = m and we will prove it for p = m + 1.
Without loss of generality assume that x ³ x0.
By definition of fm+1 and fm we get
 x   x 
φm +1 ( x ) – φm ( x ) =  y0 + ∫ f (t , φm (t )) d t  –  y0 + ∫ f (t , φm –1 ) d t 
 x0   x0 
x
= ∫ [ f (t , φ m (t )) – f (t , φ m –1 (t )) ] d t
x0
x
Thus, φm +1 ( x) – φm ( x) ≤ ∫ f (t, φm (t )) – f (t, φm –1 (t )) d t
x0
Since f satisfies Lipschitz condition we get
x
φm +1 ( x) – φm ( x) ≤ K ∫ φm (t ) – φm –1 (t ) d t
x0
m
t – x0
But φm (t ) – φm –1 (t ) ≤ M K m –1 .
m!
x
MK m m
Therefore φm +1 ( x ) – φm ( x ) ≤ ∫ t – x0 d t
m! x0

M K m ( x – x0 )m +1 M ( K | x – x0 |)m +1
≤ =
m! m +1 K ( m + 1)!

Differential Equations (171)


Thus by induction the inequality (4.2.1) is true for p = 1, 2, 3,.....
p p
M K x – x0
Since φ p ( x ) – φ p –1 ( x ) ≤ ,
K p!
p
K p x – x0
( )
k k
M M k |x – x0 |
∑ φ p ( x) – φ p –1 ( x) ≤ ∑ ≤ e –1
p =1 K p =1 p! K
And by weierstrass M-test, left hand series is uniformly convergent. Therefore the series

φ0 ( x ) + ∑ φ p ( x ) – φ p –1 ( x ) 
p =1

is absolutely convergent on I. Let f (x) be a limit function of the series. Then


 k 
Lim
k →∞ φ k ( x ) = kLt→∞ φ0 ( x ) + ∑ φ p ( x ) – φ p –1 ( x )  
 p =1 

(
= φ0 ( x) + ∑ φ p ( x) – φ p –1 ( x) )
p =1

= f (x)
Thus the sequence {fk} of successive approximations is a convergent sequence.

(b) Properties of limit function f.


The limit function f is a continuous function on I.
 x1   x2 
φk +1 ( x1 ) – φk +1 ( x2 ) =  y0 + ∫ f (t , φk (t ) d t  –  y0 + ∫ f (t, φk (t ) d t 
 x0   x0 
x1 x2
= ∫ [ f (t , φk (t )] d t – ∫ [ f (t , φk (t )] d t
x0 x0

x2
= ∫ f (t , φk (t ) d t
x1

Since f is bounded by M, that is,


f ( x, y ) ≤ M for ( x, y ) ∈ R,
φk +1 ( x1 ) – φk +1 ( x2 ) ≤ M | x1 – x2 | ∀x1, x2 ∈ I
By taking limit as k ® ¥ we get
φ ( x1) – φ ( x2 ) ≤ M | x1 – x2 |
Therefore as x2 → x1, φ ( x2 ) → φ ( x1 ), that is, f is continuous on I.
In particular
φ ( x) – φ ( x0 ) ≤ M | x – x0 |, ∀x ∈ I

Differential Equations (172)


Since x Î R, | x – x0 | ≤ α = min a, { }
b
M
and | x – x0 |≤
b
M
implies M | x – x0 | ≤ b.

Therefore | φ ( x) – φ ( x0 ) | ≤ M | x – x0 | ≤ b

Thus, x Î I and | φ ( x) – φ ( x0 ) | ≤ b implies ( x, φ ( x)) ∈ R

(c) Bounds for | φ ( x) – φ k ( x) |



We have φ ( x) = φ0 ( x) + ∑ φ p ( x) – φ p –1 ( x) 
p =1

k
and φk ( x ) = φ0 ( x ) + ∑ φ p ( x ) – φ p –1 ( x) 
p =1

Therefore

φ ( x) – φk ( x) = ∑ φ p ( x) – φ p –1 ( x) 
p = k +1


φ ( x ) – φ k ( x ) ≤ ∑ | φ p ( x ) – φ p –1 ( x ) |
p = k +1
p p
M K | x – x0 |
But φ p ( x) – φ p –1 ( x) ≤ and | x – x0 | < α
K p!

M K pα p
Therefore φ p ( x) – φ p –1 ( x) ≤
K p!
∞ M K pα p
Thus, φ ( x) – φk ( x) ≤ ∑
p = k +1 K p!

M ( K α )k +1 ∞ (K α ) p
≤ ∑
k (k + 1)! p =0 p!
k +1
M (K α )
≤ ek α
K (k + 1)!
for every k we have
k +1
M (K α )
φ ( x) – φ k ( x) ≤ eK α .
K ( k + 1)!
(d) The limit f is a solution
We must show that
x
φ ( x) = y0 + ∫ f (t , φ (t )) d t , for all x in I .
x0

Differential Equations (173)


Since f is continuous on I and f is continuous on R, the function F defined by
F (t ) = f (t , φ (t )) is continuous on I and therefore is integrable.
x
Thus, y0 + ∫ f (t , φ (t )) d t is we defined
x0
x
Now φk +1 ( x ) = y0 + ∫ f (t , φk (t )) d t
x0

Taking limit on both sides we get


x
φ ( x) = Lt
k→ ∞ φk +1 ( x) = y0 + Lt
k→ ∞ ∫ f (t , φk (t )) d t
x0
Therefore it is sufficient to prove that
x x
∫ f (t , φ k (t )) d t = ∫ f (t , φ (t )) d t
Lt
k→ ∞
x0 x0

x x
∫ f (t , φk (t )) d t – ∫ f (t , φ (t )) d t
x0 x0

x
≤ ∫ f (t , φ k (t )) – f (t , φ (t )) d t
x0

x
≤K ∫ φ k (t ) – φ (t ) d t
x0
But by (c)
k +1
M (K α )
φ k (t ) – φ (t ) ≤ eK α
K (k + 1)!
x x
( K α ) k +1 K α
Therefore ∫ f (t , φk (t )) d t – ∫ f (t , φ (t )) d t ≤ M e | x – x0 |
x0 x0 (k + 1)!

( K α ) k +1
Since → 0 as k → ∞,
(k + 1)!
x x
∫ f (t , φk (t )) d t → ∫ f (t , φ (t )) d t that is
x0 x0

x x
Lt
k →∞ ∫ f (t , φ k (t )) d t = ∫ f (t , φ (t )) d t
x0 x0
x
And φ ( x) = y0 + ∫ f (t, φ (t )) d t
x0

Differential Equations (174)


Thus f is a solution of the initial value problem y ′ = f ( x, y ), y ( x0 ) = y0 on I.
In theorem 4.2.1 we have shown the existence of solution of initial value problem 4.1.1.
The solution thus obtained is a uniques solution.
Picard-Lindel of theorem states that if f is continuous function and satisfies Lipschitz condition
on R, then the successive approximations fk exist on | x – x0 | £ a , fk’s are continuous and
converge uniformaly on the interval I to a unique solution passing through (x0, y0) ÎR.

Differential Equations (175)


References

1. Earl A coddington, An introduction to ordinary


differential equations : Prentice Hall of India, New Delhi
1995.

2. Morris Tenenbaum and Harry Pollard, Ordinary


differential Equations : New York 1963

3. Gorett Birkhoff, Gian-Carlo Rota, Ordinary differential


equations : John Wiley and Sons, New York 1978

4. Earl D Rainville, Elementory differential Equations,


Macmillan Company New York 1964

Differential Equations
CHAPTER : 5

GREEN’S FUNCTION AND STURM-LIOUVILLE THEORY

1. Green’s Functions :
Given the inhomogeneous linear second-order differential equation

d 2u du
L(u ) = + p (x ) + q( x )u = r ( x) , .... (1)
dx 2 dx
we will show that there exists an integral operator G defined by,

x
G ( r ) = ∫ G ( x, ξ ) r( ξ ) dξ ... (2)
a

Such that L (G ( r ) ) = r

The Kernel G ( x, ξ ) of equation (2) is called the Green’s function associated with given initial
value problem or boundary value problem.
For the initial value problem of nth order we have following definition for Green’s function.
Definition 1.1 :

d nu d n−1u
Let L(u ) = + p1 ( x ) n−1 + .... + pn ( x )u , be an nth order homogeneous linear differential
dx n dx
operator. By the Green’s function of L (u) for the initial value problem at x = a is meant a function
x
G ( x, ξ ) such that the function defined by u ( x ) = G ( r ) = ∫ G ( x, ξ ) r (ξ ) dξ satisfies L(u ) = r ( x ) ,
a

for any continuous function r ( x ) , and satisfies the initial conditions u (a ) = u '(a ) = .... = u (n−1) ( a ) = 0 .

Theorem 1.1
Let the function G ( t ,τ ) be defined as follows :

(i) G ( t ,τ ) = 0 , for a ≤ t ≤ τ ,

(ii) for each fixed τ ≥ a , and all t > τ , G ( t ,τ ) is the solution of the differential equation.

L(G ) = Gtt + p (t )Gt + q (t )G = 0


which satisfies the initial conditions,

1
Therefore,
t
u′′(t ) = ∫ Gtt ( t ,τ ) t (τ ) dτ + r (t ) ..... (5)
a

Thus using (3), (4) and (5) we have


L(u ) = u ′′(t ) + p( t )u ′( t ) + q( t )u( t)

t t t
= ∫ Gtt ( t ,τ ) r (τ ) dτ + r (t ) + p (t ) ∫G t ( t ,τ ) r (τ ) dτ + q (t ) ∫ G ( t ,τ ) r (τ ) dτ
a a a

t t t
= ∫ Gtt ( t ,τ ) r (τ ) dτ + r (t ) + ∫ p(t )Gt ( t ,τ ) r (τ ) dτ + ∫ q ( t ) G ( t ,τ ) r (τ ) dτ
a a a

t
= ∫ Gtt ( t , τ ) + p (t )Gt ( t ,τ ) + q ( t) G( t ,τ )  dτ + r (t ) .... (6)
a

But by condition in (ii) G ( t ,τ ) satisfy

Gtt + p(t )Gt + q( y )G = 0


Thus equation (6) becomes
L(u ) = r (t )
This proves that the function u defined by (3) is the solution of second order differential equation
(1).
Theorem 1.2 : The solution of inhomogeneous second order differential equation

d 2u du
L(u ) = + p (x ) + q( x )u = r ( x)
dx 2 dx

for the initial conditions u (a ) = u′( a) = 0 , is

t
u (t ) = ∫ G ( t ,τ ) t(τ ) dτ
a

where

f (τ ) g (t ) − g (τ ) f (t )
G(t ,τ ) =
W ( f , g ) (τ )

and W ( f , g ) (τ ) is the Wronskian of two linearly independent solutions f and g of the homogeneous
equation.

3
From (2) and conditions in (4) we obtain

c (τ ) f (τ ) + d (τ ) g (τ ) = 0

and c (τ ) f ′ (τ ) + d (τ ) g ′ (τ ) = 1
Solving these simultaneous linear equations we obtain

g (τ )
0
1 g ′ (τ ) g (τ )
c (τ ) = =−
f (τ ) g (τ ) W ( f , g ) (τ )
f ' (τ ) g′ (τ )

f (τ ) 0
f ′ (τ ) 1 f (τ )
d (τ ) = =
and f (τ ) g (τ ) W ( f , g ) (τ )
f ′ (τ ) g ′ (τ )

Thus from (2) the required Green’s function si,

− g (τ ) f (τ )
G ( t ,τ ) = f (t ) + g (t )
W ( f , g ) (τ ) W ( f , g ) (τ )

f (τ ) g ( t ) − g (τ ) f (t )
= .... (5)
W ( f , g ) f (τ )
Hence the solution of
L(u ) = u′′ + p (x )u ′+ q( x )u = r ( x )

is u (t ) = ∫ G ( t ,τ ) t(τ ) dτ ... (6)


a

where G ( t ,τ ) is given in (5).


Remark : In the elementary theory of differential equations formula (6) is often derived formally by
posing the question : What must be c (τ ) and d (τ ) be in order that the function

G(t ,τ ) = c (τ ) f (t ) + d (τ ) g (t )
when substituted into (6), should give the solution of the homogeneous differential equation

L (u ( t ) ) = r (t )

Since c (τ ) and d (τ ) may be regarded as ‘variable parameters’, which vary with τ , formula
(6) is said to be obtained by the method of variation of parameters.

5
function is given by

f (τ ) g (t ) − g (τ ) f ( t )
G ( t ,τ ) =
W ( f , g ) (τ )

f (τ ) g (t ) − g (τ ) f ( t )
=
f (τ ) g (τ )
f ′ (τ ) g ′ (τ )

cos τ sin t − sin τ cos t


=
cos τ sin τ
− sin τ cos τ

sin ( t − τ )
=
cos 2 τ + sin 2 τ

= sin ( t − τ )
Problem 1.2 : Integrate the following differential equations by constructing Green’s functions.
(a) y′′ + 10 y ′ + 25 y = sin x

(b) 4 y ′′ − 16y ′ + 15 y = x

(c) y′′ − y ′ − 2 y = e− x
Solution :
(a) Given differential equation is
y′′ + 10 y ′ + 25 y = sin x ... (1)
The characteristic polynomial is

p ( r ) = r 2 + 10r + 25 = ( r + 5) 2
and its roots are r = – 5, – 5.

Thus two linearly independent solutions are f ( x ) = e−5 x , g ( x) = xe−5 x . The solution of
differential equation (1) is given by
t
y (t ) = ∫ G ( t ,τ ) r (τ ) dτ ..... (2)

where r (τ ) = sin τ and G ( t ,τ ) is Green’s function defined by

f (τ ) g (t ) − g (τ ) f ( t )
G ( t ,τ ) =
W ( f , g ) (τ )
7
5e −5t t
e −5t
∫e ∫e
5τ 5τ
= sin τ − cosτ dτ
26 26

5e −5t e 5t e−5t e5t


= ( 5sin t − cos t ) − [ 5cos t + sin t ]
26 26 26 25 + 1

5 1
= ( 5sin t − cos t ) − [5cos t + sin t ]
676 676

1
= [ 25sin t − 5cos t − 5cos t − sin t ]
676

1
= [ 24sin t − 10cos t ]
676

12sin t − 5cos t
=
338
is the required solution of (1).

(b) The characteristic polynomial of the equation


4 y ′′ − 16y ′ + 15 y = x .... (1)

is p ( r ) = 4 r 2 − 16r + 15

= 4r 2 − 10r − 6 r + 15
= 2r (2r − 5) − 3(2r − 5)
= (2 r − 3)(2r − 5)
and its roots are

3 5
r= ,
2 2
3x 5x
The two linearly independent solutions are f ( x ) = e 2 and g ( x ) = e 2 .

The solution differential equation (1) is given by


t
y (t ) = ∫ G ( t ,τ ) r (τ ) dτ .... (2)

9
where r (τ ) = e −τ and

f (τ ) g (t ) − g (τ ) f ( t )
G ( t ,τ ) =
W ( f , g ) (τ )

f (τ ) f ( t ) − g (τ ) f (t )
=
f (τ ) g ′ (τ ) − g (τ ) f ′ (τ )

e−τ e 2t − e2τ e −t
=
e −τ (2) e2τ − e 2τ e −τ (−1)

e −τ e 2t − e 2τ e− t
=
2eτ + eτ

=
1
[e −2τ e 2t − eτ e −t ]
3
Thus equation (1) reduces to

t
y (t ) = ∫
1
[ e−2τ e 2t − eτ e − t ] e−τ dτ
3

t
1
[e −3τ 2 t −t ] dτ
3∫
= e −e

1  e −3τ 2t 
=  e − e −t t 
3  ( −3) 

1 −t te− t
=− e −
9 3
is the required solution.

EXERCISE

1. Construct Green’s function for the initial value problem, for the following differential equations.
(a) y′′ + 4 y = 0
(b) y′′ + y = 0

(c) y′′ + 3y ′ + 2 y = 0

11
d  du 
+ q (x )u + λ r (x )u = 0 , a ≤ x ≤ b
dx  dx 
p( x)

Coupled with boundary conditions.


u (a ) = u (b) , u′( a ) = u′(b ) .... (3)
is called a periodic Sturm-Liouville boundary value problem.
Definition 2.5 : A noutrivial solution of Sturm-Liouville boundary value problem (BVP) is called an
eigen function and the corresponding λ is called its eigen value.

For a value of λ if xλ is non trivial solution of Sturm-Liouville BVP then λ is an eigen value
and xλ is eigen function corresponding to λ .
Definition 2.6 : The set of all eigen values of regular Sturm-Liouville BVP is called the spectrum of
the Sturm-Liouville BVP.
Theorem 2.1: Let u and v be two eigen functions of Sturm-Liouville BVP

d  du 
 p( x)  + q (x )u + λ r (x )u = 0 , a ≤ x ≤ b
dx  dx 

α1u ( a) + β1u ′(a ) = 0 ... (2.1)

α 2u (a ) + β2u ′( b) = 0 .... (2.2)

Corresponding to eigen values λ and µ . Then

p [W (u , v ) ]a = 0 ,
b
.... (3)

where W (u, v) is the Wronskian of u and v.


If p (a) = 0 then (3) holds without use of (2.1).
If p (b) = 0 then (3) holds without use of (2.2).
Proof : Let p ( a) ≠ 0 and p (b ) ≠ 0 .
As u and v are the eigen functions of the Sturm-Liouville problem, from (2.1) we have
α1u ( a) + β1u ′(a ) = 0

and α1v (a ) + β1v ′( a ) = 0

By definition of Sturm-Liouville BVP, at least one of α1 and β1 is non-zero.

Let us suppose α1 ≠ 0 .

Elimination of β1 from above two equations leads to

α1 [ u ( a ) v′( a ) − v (a )u ′( a )] = 0

13
Proof : Let u and v be two eigen functions of periodic Strum-Liouville BVP (1) - (2).
Then p (a) = p (b) and in this case we have

[ p( x)W (u(x ), v( x))]ba

=  p( x) ( u ( x ) v′( x ) − v( x) u′( x) ) 
b
a

= p( b) [ u ( b ) v′( b) − v( b) u′( b )] − p (a ) [ u ( a ) v′( a ) − v( a )u′( a) ]

= p( b) [ u ( b ) v′( b) − v ( b) u′(b) − u ( a ) v′( a) + v (a )u ′( a) ]


The expression inside the bracket is zero once we use the periodic conditions.
u (a ) = u (b) = u ′( a ) − u′(b)
and v (a ) = v (b), v ′( a) = v′(b)

Thus [ pW (u , v)]ba = 0
Definition 2.7 : Two integrable real-valued functions f and g are orthogonal with weight function
b

r (t ) > 0 on an interval [ a, b] if and only if ∫ r ( x ) f (x )g (x )dx = 0 .


a

Theorem 2.3 : The eigen functions of a regular Sturm-Liouville BVP corrosponding to distinct eigen
values are orthogonal on an interval [a, b] with weight function r.
Proof : Let u and v be the eigen functions corresponding to distinct eigen values λ and u, of a regular
Sturm-Liouville problem.

d  du 
 p( x)  + q( x )u + λ r (x )u = 0, a ≤ x ≤ b .... (1)
dx  dx 

α1u ( a) + β1u ′(a ) = 0 .... (2.1)

α 2u (b) + β 2u ′(b) = 0 .... (2.2)


Then from (1) we have

d  du 
 p( x)  + q (x )u + λ r (x )u = 0 .... (3)
dx  dx 
and

d  dv 
+ q( x) v + λ r (x )v = 0
dx  dx 
p ( x ) ... (4)

15
Taking complex conjugate of (1), we get

d  du 
+ q (x )u + λ r (x )u = 0
dx  dx 
p ( x ) .... (2)

Multiply the differential equation in (1) by u , and multiply that of (2) by u, and substracting
one from other we get,

d  du  d  du 
− u + ( λ − λ ) r ( x ) uu = 0
dx  dx  dx  dx 
p( x) u p( x)

d   du du  
( λ − λ ) r (x )uu =  p (x )  u −u 
dx   dx dx  

d
=  p( x) W( u( x), u( x) ) 
dx 
Integrating the last equality, we get
b
( λ − λ ) ∫ r ( x )u ( x )u( x ) dx =  p(x )W (u ( x),u ( x) ) b .... (3)
a
a

Since u and u satisfy the boundary conditions (2.1) and (2.2), we have

 p(x )W (u ( x),u ( x) )  = 0
b
a

Thus (3) reduces to

b
( λ − λ ) ∫ r ( x) u ( x ) 2 dx = 0 .... (4)
a

Since u, being an eigen function, is not identically equal to zero and r ( x ) > 0 on [a, b], we
have
b

∫ r (x ) u( x)
2
dx ≠ 0
a

Therefore, from (4), λ = λ . That is, λ is real.


Problem 2.1 : Show that the boundary value problem
u′′( x ) + λu ( x) = 0, u(0) = 0, u ′(1) = 0
is a Sturm-Liouville problem, and find all eigen values and a corrosponding set of
eigen functions.

17
Thus the general solution is
u ( x ) = C1e µ x + C2e − µ x

⇒ u ′( x) = C1µ eµ x − C2 µe− µ x
Applying boundary conditions, we get
0 = u (0) = C1 + C2 ⇒ C2 = − C1
 µ 1 
0 = u′(1) = C1µ e µ + C1µe− µ = 0 ⇒ C1µ  e + µ =0
 e 

µ
1
Since µ ≠ 0 , e + ≠ 0 , we have C1 = 0.

So again the only solution is zero solution, therefore there are no negative eigenvalues.

Case 3 : Let λ > 0 . Assume λ = µ 2 , for some non zero real µ .


The characteristic polynomial becomes
r2 + µ2 = 0
⇒ r = ±iµ
The general solution in this case is
u ( x ) = C1 cos µ x + C2 sin µ x
Using boundary conditions we get
0 = u(0) = C1
Therefore,

u ( x ) = C2 sin µ x

⇒ u ′( x) = C2 µ cos µ x
Using boundary condition, we get

0 = u′(1) = C2 µ cos µ

Since µ ≠ 0 , C2 = 0 again gives trivial solution. Thus we assume C2 ≠ 0 .

This give cos µ = 0 .

( 2n + 1) π
This is true for µn = , n = 0, 1, 2, .....
2
Therefore the eigen values are

 ( 2n +1) π 
2

λn = µ n 2
=  , n = 0, 1, 2, .....
 2 
19
Solving (1) and (2) we get
C1 = 0 , C2 = 0 .

Thus, only soloution in this case is the trivial solution u ( x ) = 0 . Therefore there are no negative
eigen values.

Case 3 : Let λ > 0 . Assume λ = µ 2 , where µ is real nonzero real number. Then the characteristic
polynomial is

r2 + µ2 = 0

⇒ r = ±iµ
Therefore the general solution is
u ( x ) = C1 cos µ x + C2 sin µ x

⇒ u′( x) = −C1µ sin µ x + C2 µ cos µ x

The condition u (0) = u (π ) gives that

C1 = C1 cos µπ + C2 sin µπ

⇒ (1 − cos µπ ) C1 − ( sin µπ ) C2 = 0 .... (3)

Further the condition u′(0) = u′(π ) gives

C2 = −C1 sin µπ + C2 cos µπ

⇒ ( sin µπ ) C1 + (1 − cos µπ ) C2 = 0 .... (4)


The system (3) and (4) has non trivial solution for C1 and C2 if and only if

1 − cos µπ − sin µπ
=0
sin µπ 1 − cos µπ

⇒ (1 − cos µπ ) + sin 2 µπ = 0
2

⇒ 1 − 2cos µπ + cos 2 µπ + sin 2 µπ = 0

⇒ 2 − 2cos µπ = 0
⇒ cos µπ = 1

⇒ µn = ±2n , n = 1, 2, 3, .....
Thus positive eigen values are given by

λn = µ n2 = 4 n2 , n = 1, 2, 3, .....

21
n2π 2
⇒ λn = , n = 1, 2, 3 ....
(log2) 2
are the required eigen values and corrosponding eigen function is

 nπ 
un ( x ) = sin  log x  , n = 1, 2, 3, ......
 log2 
Problem 2.4 : Show that the eigen values of Sturm-Liouville BVP
u′′( x ) + λ u( x) = 0 , 0 ≤ x ≤ L

u (0) = 0 , u ( L) = 0
are all positive.
Solution : The equation Sturm-Liouville equation is
u′′( x ) + λ u( x) = 0
Multiplying this equation by u ( x) and then integrating from 0 to L, we obtain

L
d 2u L

∫ u ( x) dx 2
dx + λ ∫ u 2 ( x) dx = 0
0 0

L
 du  L du du L
⇒ u ( x )  − ∫ dx +λ ∫ u 2 ( x ) dx = 0
 dx  0 0 dx dx 0

As u (0) = 0 and u ( L) = 0 , above equation reduces to

2
 du 
L L
− ∫   dx +λ ∫ u 2 ( x )dx = 0
0
 dx  0

2
L
 du 
∫  dx  dx
⇒λ = L
0

∫u ( x )dx = 0
2

As integrands of both the integrals are non-negative, we must have λ ≥ 0 .

23
Writting Team Chapter Number
Dr. (Mrs.) Sarita Thakar
Department of Mathematics, 1–4
Shivaji University, Kolhapur.
Maharashtra

Dr. Kishor D. Kucche


Department of Mathematics, 5
Shivaji University, Kolhapur.
Maharashtra

25

Vous aimerez peut-être aussi