H I

SHIVAJI UNIVERSITY, KOLHAPUR

CENTRE FOR DISTANCE AND ONLINE EDUCATION

Partial Differential Equations

(Mathematics)

For

M. Sc.-I Sem. II

(In accordance with National Education Policy 2020)

(Academic Year 2022-23 onwards)

K J
Copyright © Registrar,
Shivaji University,
Kolhapur. (Maharashtra)
First Edition 2009
Revised Edition 2023

Prescribed for M. Sc. Part-II

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 : 500

Published by:
Dr. V. N. Shinde
Ag. Registrar,
Shivaji University,
Kolhapur-416 004

Printed by :
Shri. B. P. Patil
Superintendent,
Shivaji University Press,
Kolhapur-416 004

Cover Design by :
Pratik Printing Services, Kolhapur.

ISBN- 978-81-8486-302-4

H Further information about the Centre for Distance and Online Education & Shivaji University
may be obtained from the University Office at Vidyanagar, Kolhapur-416 004, India.

(ii)
 Centre for Distance and Online Education
Shivaji University, Kolhapur
n ADVISORY COMMITTEE n

Prof. (Dr.) D. T. Shirke Prof. (Dr.) M. S. Deshmukh

Honourable Vice Chancellor, Dean, Faculty of Humanities,
Shivaji University, Kolhapur Shivaji University, Kolhapur

Prof. (Dr.) P. S. Patil Prof. (Dr.) S. S. Mahajan

Honourable Pro-Vice Chancellor, Dean, Faculty of Commerce and
Shivaji University, Kolhapur Management, Shivaji University, Kolhapur

Prof. (Dr.) Prakash Pawar Prof. (Dr.) Smt. S. H. Thakar

Department of Political Science, I/c. Dean, Faculty of Science and
Shivaji University, Kolhapur Technology, Shivaji University, Kolhapur

Prof. (Dr.) S. Vidyashankar Prin. (Dr.) Smt. M. V. Gulavani

Vice-Chancellor, KSOU, I/c. Dean, Faculty of Inter-disciplinary
Mukthagangotri, Mysuru, Karnataka Studies, Shivaji University, Kolhapur

Dr. Rajendra Kankariya Dr. V. N. Shinde

G-2/121, Indira Park, Ag. Registrar,
Chinchwadgaon, Pune Shivaji University, Kolhapur

Prof. (Dr.) Smt. Cima Yeole Dr. A. N. Jadhav

Git Govind, Flat No. 2, Director, Board of Examinations and
1139 Sykes Extension, Kolhapur Evaluation, Shivaji University, Kolhapur

Dr. Sanjay Ratnaparkhi Shri. A. B. Chougule

D-16, Teachers Colony, I/c. Finance and Accounts Officer,
Vidhyanagari, Mumbai University, Shivaji University, Kolhapur
Santacruz (E), Mumbai
Prof. (Dr.) D. K. More
Prof. (Dr.) Smt. Kavita Oza (Member Secretary) Director,
Department of Computer Science, Centre for Distance and Online Education,
Shivaji University, Kolhapur Shivaji University, Kolhapur.

Prof. (Dr.) Chetan Awati

Department of Technology,
Shivaji University, Kolhapur

(iii)
 Centre for Distance and Online Education
Shivaji University, Kolhapur

n MEMBERS OF B.O.S. IN MATHEMATICS n

Chairman- Prof. Dr. Kishor Kucche
Department of Mathematics, Shivaji University, Kolhapur

l Dr. Mrs. S. H. Thakar l Dr. Sanjay Anant Morye

Head, Dept. of Mathematics, Rajaram College, Vidyanagar,
Shivaji University, Kolhapur. Kolhapur

l Dr. Girish Dhondiram Shelake l Dr. Hambirrao Tatyasaheb Dinde

Willingdon College, Sangli Shri Shiv Shahu Mahavidyalaya,
Sarud, Tal. Shahuwadi, Dist.
l Dr. Smt. Bebitai Annaso Sajane Kolhapur
Smt. Kasturbai Walchand College of
Arts & Science, Sangli l Dr. Santaji Shrikant Khopade
Karmaveer Hire Arts, Science,
l Dr. Dadasaheb Rajaram Commerce and Education College,
Phadatare Gargoti, Tal. Bhudaragad, Dist.
Balasaheb Desai College, Patan, Kolhapur
Tal. Karad, Dist. Satara

l Dr. Santosh Bhaurao Joshi

Walchand College of Engineering,
Sangli

(iv)
 Preface
Large number of students appears for M.A./M. Sc. examinations externally every year. In
view of this, Shivaji University has introduced the Distance and Online Education Mode for
external students from the year 2008-09, and entrust the task to us to prepare the Self
Instructional Material (SIM) for aspirants. An objective of the SIM is to provide students the
material on the subject from which they can prepare for examination on their own without the help
of a tutor. Today we are extremely happy to present the book on "Partial Differential Equations"
for M.A./M.Sc. Semester-II students as a SIM prepared by well devoted expert Dr. L. N. Katkar.
We hope that the exposition of the material in the book will meet the needs of all aspirants.

The mathematical formulation of the real world problems in science and engineering involves
partial differential equations. In order to understand the physical behaviour of these real world
problems, it is necessary to have some knowledge about the solutions of the governing partial
differential equations. For example, transverse vibrations of an elastic string are governed by the
wave equation; the temperature distribution in a homogeneous isotropic rod is governed by the
heat equation etc. The wave equation; the heat equation and the Laplace equation have been
derived by taking into account certain physical situations.

Partial differential equations of first order and various methods such as Charpit's method,
Jacobi method of finding their complete integral; general integraol; singular integral and Cauchy
integral surfaces are dealt in the first four units. The classification of second order partial
differential equations and their canonical forms are given in the unit 5. Boundary value problems
such as Dirichlet and Neumann boundary value problems are discussed in the subsequent units,
besides maximum-minimum principle and families of equipotential surfaces. The well-known
mathematical techniques namely, the most powerful method of separable of variables, Fourier
transform techniques and Green's function approach are applied to solve various boundary value
problems involving parabolic, elliptic and hyperbolic partial differential equations.

An attempt has been made to make the presentation of the various units comprehensive,
rigorous and yet simple. One of the features of this book is that in all units numerous examples
have been solved for the use of students working independently of a teacher. Although the book
is aimed to M. Sc. Distance Education Students, even SET/NET aspirants and students of
physics and engineering would find it useful.

We owe a deep sense of gratitude to the Ag. Vice-Chancellor who has given impetus to go
ahead with ambitious projects like the present one. Dr. L. N. Katkar, of the Department of
Mathematics, Shivaji University, has to be profusely thanked for the ovation he has poured to
prepare the SIM on Partial Differential Equations. We also thank Prof. S. H. Thakar, Head,
Department of Mathematics, Director of Distance Education Mode Prof. Cima Yeole, Shivaji
University, for their help and keen interest in completion of the SIM. Thanks are also due to Mr.
Sachin Kadam for computerizing the manuscript neatly and correctly.

Prof. K. D. Kucche
Chairman
BOS in Mathematics
Shivaji University, Kolhapur-416004.

(v)
Centre for Distance and Online Education Partial Differential Equations
Shivaji University,
Kolhapur.

Writing Team Unit No.

Dr. L. N. Katkar All

Dept. of Mathematics,
Shivaji University, Kolhapur.

n Editor n

Dr. L. N. Katkar
Department of Mathematics,
Shivaji University, Kolhapur.
Maharashtra.

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

Contents
Unit-1 : First Order Partial Differential Equations 1

Unit-2 : Linear Equations of the first Order 29

Unit-3 : Compatible Systems of First Order 60

Partial Differential Equations

Unit-4 : The Cauchy Problem 114

Unit-5 : Second Order Partial Differential Equations 173

Unit-6 : Heat Conduction Problem 223

Unit-7 : Laplace Equation 248

Unit-8 : Riemann's Method of Solution of

Linear Hyperbolic Equations 294

(vii)
Each Unit begins with the section objectives -

Objectives are directive and indicative of :

1. what has been presented in the unit and

2. what is expected from you

3. what you are expected to know pertaining to the specific unit,

once you have completed working on the unit.

The exercises at the end of each unit are not to be submitted to

us for evaluation. They have been provided to you as study tools to
keep you in the right track as you study the unit.

Dear Students

The SIM is simply a supporting material for the study of this paper.
It is also advised to see the new syllabus 2022-23 and study the
reference books & other related material for the detailed study of the
paper.

(viii)
 UNIT - I

FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS

Introduction :
The mathematical formulation of the real situations in science and engineering involves partial
differential equations. In order to understand the physical behaviour of the real world situations, it is
necessary to have some knowledge about the properties and the solutions of the governing partial
differential equation. A partial differential equation is one involving more than one independent variable,
a dependent variable and its partial derivatives with respect to the independent variables. In general,
partial differential equations arise in physics in problems involving electric fields, fluid dynamics, wave
motion etc. These equations are called Heat equations, Laplace equations wave equations. Each is
profoundly significant in theoretical physics and their study is stimulated in the development of many
mathematical ideas.
The basic concepts from solid geometry play important roles in the study of partial differential
equations and it is essential that they should be understood thoroughly before the study of partial
differential equations is begun. Hence we define some basic concepts from geometry.
Curves and Surfaces :
Curves in Space : Let I be an interval on the real line R and t a continuous variable which varies in I.
If f1, f2, f3 are continuous functions of t, then the equations.
x  f1 (t ), y  f 2 (t ), z  f 3 (t ) , ... (1.1)
represent the parametric equations of a curve in three dimensional space.
Note : (i) t is called the parameter of the curve.
(ii) The standard parameter is the arc length of the curve measured from some fixed point on the
curve to any current point, it is denoted by s instead of t.
(iii) The square of an infinitesimal arc length between two neighbouring points on the curve in 3-
dim. space is given by
ds 2  dx 2  dy 2  dz 2 ,
2 2 2
 dx   dy   dz 
        1.
 ds   ds   ds 
It follows from equation (1.1) that the condition that the parameter t be the arc length of the curve is
that

f1'2  f 2'2  f 3'2  1 .

1
Examples :
(i) The simplest example of a curve in space is a straight line with direction cosines (  ,m,n)
passing through a point (x0, y0, z0) and is given by
x  x0  s, y  y0  ms, z  z0  ns , ... (1.2)
where s is the parameter.
(ii) A right circular helix lying on a circular cylinder is a space curve and is given by the parametric
equations
x  a cos t , y  a sin t , z  kt , ... (1.3)
where a, k, w, are constants.
Surface : Let (x, y, z) be the cartesian co-ordinates of a point in a 3-dimensional space. Then the
functional relation between these variables x, y, z given by the equation
F (x, y, z) = 0 ... (1.4)
is called a surface.
If F is linear, then equation (1.4) can be solved for one of the variables and it can be expressed in terms
of the other two independent variables and we are left with only two degrees of freedom. Hence a
surface is defined as the locus of a point moving in space with two degrees of freedom.

Parametric Equations of a Surface :

A set of those points of a 3-dimensional space which are expressed as a function of two
parameters is called a surface. Thus a set of relations of the form.
x  F1 (u, v), y  F2 (u , v), z  F3 (u, v) ... (1.5)
determines a surface.
Exaplanation : Solving the first pair of equations (1.5)
viz., x  F1 (u, v), y  F2 (u , v) ,
for u and v as functions of x and y, we obtain
Say u   ( x, y ), v   ( x, y ) .
This shows that once x and y are known, then u and v are determined. Then corresponding value of z
is obtained by substituting these values of u and v in the third equation z  F3 (u, v) . In otherwards the
value of z is determined once the values of x and y are known. Symbolically,
z  F3   ( x, y),  ( x, y )  . ... (1.6)
Which is a functional relation between the coordinates x, y and z. Thus any point (x, y, z) determined
from equations (1.5) always lies on a surface. The equations (1.5) therefore are called the parametric
equations of the surface.
Note : Not every point in space corresponds to a pair of values of u and v.
2
   F1, F2 
For that 0
 (u, v)
Note : Parametric equations of a curve and a surface are not unique.

Examples :
(1) The parametric equations of a surface of a sphere of radius ‘a’ are given by
x = asinucosv,
y = asinusinv, a = constant ... (1.7)
x = acosu.
The same surface is also represented by the set of equations

1  v 2 
xa cos u ,
1  v 2 

1  v 2 
ya sin u , ... (1.8)
1  v 2 
2av , a = constant
z
1  v 2 
(2) The parametric equations of a cone x 2  y 2  z 2 tan 2  are given by

x  r sin  cos  ,
y  r sin  sin  ,  = constant ... (1.9)

z  r cos  .
or x  r cos  , y  r sin  , z  r cot  . ... (1.10)

A Curve Through Surfaces :

Consider a surface f (x, y, z) = 0, ... (1.11)
and a plane z = k. ... (1.12)
A point whose co-ordinates satisfy equation (1.11) and which lies in the plane (1.12) has its co-
ordinates satisfying the equations

z  k , f ( x, y , k )  0 , ... (1.13)
which represents a curve in the plane z = k.

3
For example : Let S be a sphere with equation z

x2  y2  z 2  a2 .
Then the points of S with z = k have z=k
C
z=k and
x2  y2  a2  k 2 ,
y
which is a curve C and the curve is a circle of radius

a 2  k 2 for k < a.
x
Thus a curve can be thought of as the intersection of the surface (1.11) and the plane (1.12).
In general, the common points to the surfaces,
S1 : f ( x, y, z )  0 ,

and S 2 : g ( x, y , z )  0 ,
lie on the curve C.
Thus, the locus of a point whose co-ordinates satisfies a pair of relations S1 = 0 and S2 = 0 is a
curve in space.

Direction Cosines of a line passing through two points :

Consider a line through the points P  x1 , y1 , z1  and Q  x2 , y 2 , z 2  . The vector PQ is defined by

PQ   x2 , y2 , z 2    x1, y1 , z1  ,

PQ   x2  x1, y2  y1 , z2  z1  ,

 PQ   x2  x1  i   y2  y1  j   z 2  z1  k . ... (1.14)

Direction cosines of a line PQ are the cosines of the angles made by the line PQ with
co-ordinate axes. Let cos  , cos  , cos  be the direction cosines of the line PQ , then we have
from equations (1.14)

PQ  i  PQ cos    x2  x1 

 cos  
 x2  x1  .
PQ

Similarly, cos  
 y2  y1  , ... (1.15)
PQ

4
 cos  
 z2  z1  .
PQ

 x2  x1 y2  y1 z 2  z1 
Thus  PQ , PQ , PQ  are the direction cosines of the line through the points P  x1 , y1 , z1 
 
and Q  x2 , y 2 , z 2  .
Note : From equations (1.15), we see that x2 – x1, y2 – y1, z2 – z1 are proportional to the direction
cosines of the line and hence they represent the direction ratios of the line PQ .

The Direction Cosines of the tangent to the curve :

Example 1 : Show that the direction cosines of the tangent to the curve
x = x(s), y = y(s), z = z(s),
where s is the arc length of the curve measured from the fixed point p0 on the curve to any point P on
 dx dy dz 
the curve are  , ,  .
 ds ds ds 
Solution : Consider a curve in 3-dimensional space given by
x = x(s), y = y(s), z = z(s), ... (1.16)
where s is the arc length measured from the fixed point p0 to any point P on the curve.
z
Thus s = P0P. s Q
P
Let Q be any other point at a distance  s from P,, c

 P0Q  s   s . P0
O y
Consequently, the co-ordinates of the point Q are

Q   x  s   s  , y  s   y  , z  s   s  . x

Since  s is measured along the curve from P to Q and is therefore greater than the length  c of the
chord PQ.
In the limiting case as  s 
 0 , we have
s
lim 1. ... (1.17)
 s 0  c

The direction cosines of the chord PQ are given by

 x  s   s   x( s) y  s   s   y (s) z  s   s   z ( s) 
 , , .
 c c c 

5
By Taylor’s series expansion, we have

 dx  2
x  s   s   x ( s )   s    O  s  .
 ds 
Hence direction cosines of the chord PQ are given by

  s  dx   s  dy   s  dz 
   O( s)  ,   O( s )  ,   O( s )  
  c  ds   c  ds   c  ds 

In the limiting case as  s 

 0 , the point Q tends towards the point P and the chord PQ takes up
to the tangent to the curve at P. Thus as  s 
 0 , the direction cosines of the tangent to the curve
(1.16) at apoint P are

 dx dy dz 
 , , .
 ds ds ds 
Direction ratios of the normal to the surface :
Let us consider that the curve C (defined in equation
S (1.16)) lies on the surface S : F (x, y, z) = 0.
Any point (x(s), y(s), z(s)) on the curve lies on this surface
P satisfies the equation
T
F  x ( s ), y ( s ), z ( s )   0 . ... (1.18)
C
If the curve lies entirely on the surface, then equation
(1.18) is an identity for all values of s. Differentiating
equation (1.18) with respect to s, we get

F dx F dy F dz
      0, ... (1.19)
x ds y ds z ds

 dx dy dz 
where  , ,  are the direction cosines of the tangent to the curve C at the point P. Equation
 ds ds ds 
(1.19) shows that the tangent to the curve C at the point P is perpendicular to the line whose direction
 F F F 
ratios are  , ,
z 
.
 x y

 F F F 
 , ,  are the direction ratios of the normal to the surfaces S at the point P..
 x y z 
Example 2 : Find the direction cosines of the normal to the surface S of the form z = f (x, y).
Solution : Let the surface S : z = f (x, y)
 F ( x, y , z )  f ( x, y )  z ... (1.20)

6
  F F F 
The direction ratios of the normal to the surface S : F ( x, y, z )  0 are  , , ,
 x y z 
where from equation (1.20), we have

F f z
  ,
x x x

F f z
  , and
y y y

F
 1 .
z

z z
Let us introduce the notations  p and q
x y

 F F F 
Thus  x , y , z   ( p, q, 1) .
 

 Direction ratios of the normal to the surface (1.20) are (p, q, -1). Hence the direction cosines of
the normal to the surface S at the point P are

1
( p, q, 1) .
p2  q2 1

Equation of a line when two surfaces are given :

Let S1 : F (x, y, z) = 0 and S2 : G (x, y, z) = 0 be two surfaces. Then the equations of the
tangent planes  1 and  2 at point P (x, y, z) to the surfaces S1 = 0 and S2 = 0 are given by

 X  x  F  Y  y  F   Z  z  F  0 , ... (1.21)
x y z

and  X  x  G  Y  y  G   Z  z  G  0 , ... (1.22)

x y z
where (x, y, z) are the co-ordinates of any other point on tangent plane. Let C be the locus of the
intersection of two surfaces S1 and S2, and L the intersection of two planes  1 and  2 . We see that
the intersection L of the planes  1 and  2 is the tangent at P to the curve C, which is the intersection
of the surfaces S1 and S2.

7
 S2=0

L
2
C

P S 1= 0
1

It follows from equations (1.21), (1.22) that the equation of the line L is

X x Yy Zz
 
Fy Fz Fz Fx Fx Fy
Gy Gz Gz Gx Gx Gy

or X x Yy Zz . ... (1.23)

 
( F , G ) ( F , G) ( F , G )
 ( y, z )  ( z , x)  ( x, y )

  ( F , G) ( F , G )  ( F , G) 
This is the equation of line whose direction ratios are  , ,  , when two surfaces
  ( y , z )  ( z , x )  ( x, y ) 
S1 and S2 are given.
2) Partial Differential Equations :
A partial differential equation is one involving more than one independent variables x, y, t, ...,
one dependent variable   C n in some domain D and its partial derivatives  x ,  y ,...,  xx ,  xt ,...,
such as,

f  x, y, t ,..., , x , y ,..., xt ,...  0 , ... (2.1)

where C n denotes a set of functions possessing continuous partial derivatives of order n.

Order of a partial differential equation :
The order of a partial differential equation is the order of the derivative of the highest order
occuring in the equation.
In this unit and in the next three units, we shall consider partial differential equations of the first
order with one dependent variable z and two independent variables x and y. Then the most general first
order partial differential equation is given by

8
 f (x, y, z, p, q) = 0, ... (2.2)

where z z .
p ,q 
x y
Partial differential equations arise in a large variety of subjects in geometry, physics, mathematics etc.

Origin of first order Partial Differential Equations :

We shall examine in the following the intersecting question of how the first order partial differential
equations arise.
Example 3 : Find the first order partial differential equation which represents the set of all spheres
with centres on the z-axis and of radius a.
Solution : The set of all spheres with centres on the z-axis and of radius a is given by

x 2  y 2  ( z  c) 2  a 2 , ... (2.3)
where a and c are constants.
Differentiating equations (2.3) with respect to x and y we get,
x + p (z –c) = 0 ,
and y + q (z – c) = 0 .
Eliminating the arbitrary constant c from the equations we obtain,
yp – xq = 0,
which is the required first order partial differential equation.
Example 4 : Find the partial differential equation which represents the set of all right circular cones
with z-axis as the axis of symmetry.
Solution : The set of all right circular cones with z-axis as the axis of symmetry is given by the equation

x 2  y 2  ( z  c ) 2 tan 2  , ... (2.5)

where c is a constant and  is a constant semi-vertical angle of the cone.
Differentiating equation (2.5) with respect to x and y we get

x  p ( z  c) tan 2  ,

y  q ( z  c) tan 2  .
Now eliminating c and  from the above equations we get
yp  xq  0 . ... (2.6)
Thus the set of cones, vertex on the z-axis with semi-vertical angle  is characterized by the first order
partial differential equation (2.6).

9
Example 5 : Find the partial differential equation which represents all surfaces of revolution with
z-axis as the axis of revolution.
Solution : All surfaces of revolution with z-axis as the axis of revolution are of the form
z  F (r ) , ... (2.7)

where r  x 2  y 2 and F is an arbitrary function of class C1 on some domain D.

On differentiating equation (2.7) first with respect to x and then with respect to y we get respectively

r r
p  F ( r ) , q  F (r )
x y

r x r y
where  ,  .
y r y r

 x
 p    F '(r ) ,
r

 y
and q    F '(r ) .
r
Eliminating the arbitrary function F from the above equation we get
yp – xq = 0, ... (2.8)
which is a partial differential equation of first order satisfied by all surfaces of revolution.
Note : We see from examples (3), (4) and (5) that the surfaces spheres, cones and in general all
surfaces of revolution with z-axis as the axis revolution give rise to the same first order partial differential
equation. What is common in all surfaces is that all surfaces of revolution have the z-axis as the axis of
symmetry.
The obvious generlization of the surfaces of revolution with z-axis as the axis of symmetry is
the relation between x, y and z of the form F (u, v) = 0, where u and v are functions of x, y, z. Hence
we shall now generalize the above argument slightly in the following.
Example 6 : Find the partial differential equation satisfied by all surfaces of the form,
F (u, v) = 0 ,
where u = u (x, y, z) and v = v (x, y, z) are known functions of x, y and z and F is the arbitrary function
of u and v.
Solution : The equations of all surfaces ingeneral is given by the equation.
F (u, v) = 0, ... (2.9)
where u = u (x, y, z) and v = v (x, y, z) are known functions of x, y and z.
Differentiating equation (2.9) with respect to x and y respectively, we get

Fu  u x  u z p   Fv  vx  v z p   0 , ... (2.10)
10
and Fu  u y  u z q   Fv  v y  v z q   0 . ... (2.11)

Eliminating Fu and Fv between equations (2.10) and (2.11) we get

 v x  pvz    v y  qvz  ,
 u x  pu z   u y  qu z 
 p  u y vz  u z v y   q  u z vx  u x vz    u y vx  u x v y   0 ,

  u, v    u, v    u, v  . ... (2.12)
p q 
  y, z    z , x    x, y 

This is the partial differential equation of first order satisfied by all surfaces of the form
F ( u, v) = 0,

  u, v 
where  uxv y  u yvx ,
  x, y 

is the Jacobian of u, v with respect to x and y.

Theorem : A necessary and sufficient condition that there exists between two functions u (x, y) and
v (x, y) a relation F (u, v) = 0 or u = H (v) not involving x or y explicitly is that

  u, v 
0.
  x, y 

Proof : The necessary condition

Let there exist between two functions u (x, y) and v (x, y) a relation of the type
F (u, v) = 0 ... (2.13)
not involving x and y explicitly.
Differentiating equation (2.13) with respect to x and then with respect to y we get
Fu u x  Fv v x  0 , ... (2.14)

and Fu u y  Fv v y  0 . ... (2.15)

Eliminating Fu and Fv between (2.14) and (2.15) we get

ux vx
 0.
uy vy

 u xv y  u yvx  0 ,

11
   u, v 
 0. ... (2.16)
  x, y 

v
The sufficient condition : Let u (x, y) and v (x, y) be two functions of x and y such that  0 and
y

  u, v 
if  0 then we claim that there exists a relation F (u, v) = 0 not involving x and y explicitly..
  x, y 

Eliminating y between the fuunctions u (x, y) and v (x, y) we get a relation

F (u, v, x) = 0 . ... (2.17)
Differentiating (2.17) with respect to x and y respectively we get
Fx  Fu u x  Fv v x  0 , ... (2.18)

and Fu u y  Fv v y  0 . ... (2.19)

Eliminating Fv from these equations we get

 uy 
Fx  Fu  u x  v x   0 ,
 v y 


 Fx v y  Fu  u x v y  u y v x   0 ,

  u, v 
 Fx v y  Fu  0. ... (2.20)
  x, y 

  u, v 
Since 0,
  x, y 

 Fx v y  0 ,

 Fx  0 as v y  0 .

 the function F does not contain the variable x explicitly. Hence from the relation (2.17) we have
F (u, v) = 0.
Hence the condition is sufficient.

Remark : We have obtained partial differential equation of first order by eliminating arbitrary constants.
(Refer examples (3), (4)).
Now consider two parameter family of surfaces given by the equation.
f (x, y, z, a, b) = 0 .

12
Solving for z, we get
z = F (x, y, a, b), ... (2.21)
where a and b are arbitrary constants. Differentiating equation (2.21) with respect to x and then with
respect to y, we get

p  Fx and q  Fy . ... (2.22)

The set of equations (2.21) and (2.22) constitute three equations involving two arbitrary constants ‘a’
and ‘b’. Now eliminating ‘a’ and ‘b’ from these equations we obtain a relation of the type.
f (x, y, z, p, q) = 0 ... (2.23)
which is a partial differential equation of the first order. In general equation (2.23) need not be linear.
Example 7 : Obtain the partial differential equation of first order by eliminating arbitrary constants
from the relation

 x  a 2   y  b 2  z 2  1 .

Solution : We are given two parameter family of surfaces

 x  a 2   y  b 2  z 2  1 . ... (2.24)

Equation (2.24) represents a set of all spheres of unit radius with centre in the xy plane. Differentiating
equation (2.24) with respect to x and y we get respectively.
 x  a   zp  0 ,

 zp    x  a  , ... (2.25)

and  y  b   zq  0 ,
 zq    y  b  . ... (2.26)
Eliminating the constants ‘a’ and ‘b’ from equation (2.24) we obtain

z 2  p 2  q 2  1  1 . ... (2.27)

This is the first order non-linear partial differential equation.

Example 8 : Obtain the partial differential equation of first order by eliminating arbitrary constants
from the relation.

z 2 1  a 3   8  x  ay  b  .
3

Solution : Two parameters family of surfaces are given by the equation

z 2 1  a 3   8  x  ay  b  .
3
... (2.28)

Differentiating equation (2.28) with respect to x and y we get respectively.

13
 z  a 3  1 p  12  x  ay  b  ,
2

z  a 3  1 q  12a  x  ay  b  .
2
and

12
 p  x  ay  b 2 ,
z  a 3  1

12a
and q  x  ay  b  2 .
z  a  1
3

Consider

(12) 3
p3  q3   x  ay  b 6 1  a 3  ,
z 3  a 3  1
3

(12) 3  x  ay  b  ,
6

z 3  a 3  1
2

(12) 3 z  x  ay  b  ,
6
 2
 z  a  1 
2 3

(12) 3 z  x  ay  b  ,
6
 by equation (2.28)
(8) 2  x  ay  b 
6

 p 3  q 3  27 z . ... (2.29)
This is the required first partial differential equation.
Example 9 : Eliminate the arbitrary functions from the following equations and find the corresponding
partial differential equations.

(i) z  xy  F  x 2  y 2  ,

(ii) F  x  y, x  z  ,

(iii) z  f  x  ct   g  x  ct  .
Solution :
(i) The equation of the surface is given by

z  xy  F  x 2  y 2  , ... (2.30)

where F is arbitrary function. Differentiating equation (2.30) with respect to x and y we get respectively
p  y  2 xF ' ,
14
and q  x  2 yF ' .
Eliminating F' between these equations we obtain

 p  y y  q  x x ,
 x 2  y 2  qx  py  0 .
This is the required partial differential equation.
(ii) The equation of the surface is given by the equation

F  x  y, x  z   0, ... (2.31)

where F is arbitray
Let u = x + y and v = x – z. ... (2.32)
Hence equation (2.31) becomes
F ( u, v) = 0 . ... (2.33)
Differentiating equation (2.33) with respect to x and y respectively we get

Fu  u x  u z p   Fv  v x  vz p   0 and

Fu  u y  u z q   Fv  v y  v z q   0 ,

where from equation (2.32), we find

1 ,
u x  1, u y  1, v x  1, v z  
2 z

 1 
 Fu  Fv 1  p  0, ... (2.34)
 2 z 

 1 
and Fu  Fv  q  0 . ... (2.35)
2 z 
Eliminating Fu and Fv between equations (2.34) and (2.35) we get

pq  2 z . ... (2.36)

This is the required partial differential equation
(iii) Here z  f  x  ct   g  x  ct  ,

 z x  f '  x  ct   g '  x  ct  ,

z xx  f ''  x  ct   g ''  x  ct  ,

zt  cf '  x  ct   cg '  x  ct  ,

15
 ztt  c 2 f ''  x  ct   c 2 g ''  x  ct  ,

 c 2 z xx  ztt ... (2.37)

This is the required second order partial differential equation.
Exercise :
1. Obtain the partial differential equation of first order by eliminating arbitrary constants from the
relations

(i) z  x  ax 2 y 2  b, (iii) z   x  a   y  b  ,

(ii) 2 z   ax  y  2  b, (iv) z  ax  by .

2. Obtain the partial differential equation by eliminating arbitrary functions from the following
relations.

(i) z  x  y  F  xy  , (ii) F  x  z , y  z   0 ,

 xy  x
(iii) z  F   , (iv) z  F   ,
 z   y

(v) z  F ( x  y ) , (vi) f  x 2  y 2  z 2 , z 2  2 xy   0 .

3. Classification of First Order Partial Differential Equations :

1. Linear Equation : A first order partial differential equation is said to be a linear equation if it
is linear in p, q and z. It is represented in the form
P ( x , y ) p  Q ( x, y ) q  R ( x , y ) z  S ( x , y ) , ... (3.1)
e.g. yp  xq  xyz  x .
2. Semi-linear Equation : A first order partial differential equation is said to be a semi-linear
equation if it is linear in p and q and the coefficients of p and q are functions of x and y only. It is
represented in the form.
P ( x , y ) p  Q ( x, y ) q  R ( x , y , z ) , ... (3.2)

e.g. xp  yxq  xz 2 .
3. Quasi-linear Equation : A first order partial differential equation is said to be a quasi-linear
equation if it is linear in p and q.
The equation of the type
P ( x , y , z ) p  Q ( x , y , z ) q  R ( x, y , z ) ... (3.3)

16
is called quasi-linear equation.

e.g.  x 2  z 2  p  xyq  z 2 x  y 2 .
4. Non-linear Equation : The partial differential equations of the form f (x, y, z, p, q) = 0 which
do not come under the above three types are said to be non-linear equations.
eg. pq = z
This is a non-linear partial differential equation of first order.
Note : We observe that by eliminating arbitrary functions, we always produce quasi-linear partial
differential equations only. However, we obtain both quasi-linear as well as non-linear partial differential
equations when we eliminate arbitrary constants. If further, the number of constants to be eliminated
from the given relation is just equal to the number of independent variables then the partial differential
equation obtained by eliminating these constants is an equation of first order. However, if the number of
constants to be eliminated is greater than the number of independent variables, the equation of second
order will arise.
Classification of Integrals :
Consider a first order partial differential equation
f (x, y, z, p, q) = 0 ... (3.4)
A solution of equation (3.4) in a region D     is given by z = z (x, y) as a continuously differentiable
function of x and y for  x, y   D such that the value of p and q obtained from the relation z = z (x, y)
must satisfy the equation (3.4). A solution z = z (x, y) of the first order partial differential equation
represents a surface in 3-dimensional space. This surface in 3-dimensional space will be called an
integral surface of the partial differential equation.
There are different types of solutions (integral surfaces) for the first order partial differential
equation (3.4).
1. Complete Integral : A complete integral of partial differential equation (3.4) is a relation
between the variables involving as many arbitrary constants as there are independent variables, provided
the value of p and q obtained from it satisfies equation (3.4). Geometrically it represents doubly infinite
system of surfaces.
Alternately, it is also defined as follows :
A two parameter family of solutions
z = F (x, y, a, b) ... (3.5)
is called a complete integral of the first order partial differential equation (3.4) if in the region considered,
the rank of the matrix

 Fa Fxa Fya 
M  
 Fb Fxb Fyb 

is two.
17
2. General Integral : A solution of a partial differential equation (3.4) of the form

  u, v   0 , ... (3.6)

where u = u (x, y, z) and v = v (x, y, z) and  is an arbitrary function, is called the general integral.
The complete integral (3.5) can be used in the derivation of general integral. Let the complete
integral of the partial differential equation (3.4) by given by two parameter family of surfaces of the
form.
z = F (x, y, a, b). ... (3.7)
If we choose b   (a ) we get one-parameter family of solution of equation (3.4) of the form

z  F  x, y , a,  ( a )  . ... (3.8)
This is a sub-family of the two parameter family (3.7). The envelope of (3.8) if it exists and is obtained
by eliminating ‘a’ between (3.8) and
Fa  Fb '(a)  0 . ... (3.9)
Solving this equation for ‘a’ we get
a = a (x, y).
Substituting this in equation (3.8), we obtain an integral surface of (3.4) as

z  F  x, y , a ( x, y ),  ( a ( x, y ))  ... (3.10)
If the function a (x, y) is arbitrary, then such a solution is called a general integral (general solution) of
(3.4). Geometrically it represents the envelope of one parameter family of surfaces.
Note : When  ( a ) is a particular function, then we obtain a particular solution of the partial differential
equation. Thus different choices of  may give different particular solution of the partial differential
equation.
Characteristic Curve :
Consider one-parameter family of surfaces
f (x, y, z, a) = 0. ... (3.11)
For slightly different value of ‘a’ say a   a , the system of surfaces becomes
f ( x, y , z , a   a )  0 . ... (3.12)
These two surfaces will intersect in a curve given by the equation

f  x, y, z , a   0, f  x, y, z , a   a   0 . ... (3.13)
Similarly, we can easily see that the curve may also be considered to be the intersection of the surface
(3.11) with the surface whose equation is

1
 f  x, y , z , a   a   f  x, y , z , a    0 .
a 
18
As  a  0 , we see that this curve of intersection is given by the equations


f  x, y , z , a   0, f  x, y , z , a   0 . ... (3.14)
a
This limiting curve is called the ‘characteristic curve’ of equation (3.11). Geometrically, it is the curve
on the surface (3.11) approached by the intersection of (3.11) and (3.12) as  a  0 .
Envelope of the one-parameter family f (x, y, z, a) = 0 :
Consider a characteristic curve


f ( x, t , a )  0, f  x, y , z , a   0 , ... (3.15)
a
where ‘a’ is a parameter. As the parameter ‘a’ varies, the characteristic curve (3.15) will trace out a
surface whose equation is obtained by eliminating ‘a’ between equations (3.15). Let this surface be
given by

g  x, y , z   0 . ... (3.16)
This surface is called the envelope of the one-parameter system f (x, y, z, a) = 0.
e.g. Consider one parameter family of surfaces
3
x2  y 2   z  a  1. ... (3.17)

It represents the family of spheres of unit radius with centres on the z-axis.

f  x2  y 2   z  a  1 ,
2
If

then f a  2  3  a  .
z
The characteristic curve is given by

x 2  y 2  z  a   1
2
x2  y2   z  a   1
2

(0,0,a) Characteristic Curve and z  a ... (3.18)

z  a, x 2  y 2  z  a   1
2
Eliminating ‘a’ between equations
y
(3.18) we get
x2  y 2  1
envelope x2  y2  1 ... (3.19)
x

Which represents the envelope of the family and is the cylinder.

We shall show that the envelope of one-parameter family of surfaces if it exists is a solution of
the given partial differential equation.

19
Result : Let z  F  x, y, a  be a one-parameter family of solutions of the first order partial differential
equation f (x, y, z, p, q) = 0. Then show that the envelope of this one parameter family, if it exists, is
also a solution of the partial differential equation.
Proof : Consider the first order partial differential equation,

f  x , y , z , p, q   0 . ... (3.20)
The one-parameter family of solutions of (3.20) is given by

z  F  x, y , a  . ... (3.21)
Differentiating equation (3.21) with respect to a we get

Fa  x, y, a   0 . ... (3.22)
Now the envelope of the family of one-parameter is obtained by eliminating ‘a’ between the equations
(3.21) and (3.22). Let the envelope be given by

z  G  x, y   F  x, y, a ( x, y )  , ... (3.23)
where a (x, y) is obtained from equations (3.22) by solving for ‘a’ in terms of x and y.
Now we prove that the envelope (3.23) is a solution of equation (3.20). Hence differentiating
equation (3.23) with respect to x and y we get

p  G x  Fx  Fa a x and q  G y  Fy  Fa a y .
Using equation (3.22), we get
p  G x  Fx , q  G y  Fy ... (3.24)
This shows that the envelope will have the same partial derivatives as those of a member of the
family. Since p = Fx and q = Fy satisfy the equation (3.20). This implies that p = Gx and q = Gy also
satisfy the partial differential equation (3.20). This proves that the envelope of one parameter family of
surfaces is also a solution of a partial differential equation.
Envelope of the two-parameter family of surfaces f (x, y, z, a, b) = 0 :
Consider the two-parameter system of surfaces defined by the equation
f (x, y, z, a, b) = 0, ... (3.25)
where ‘a’ and ‘b’ are parameter.
Let b   (a ) . ... (3.26)
Differentiating equation (3.25) with respect to a we get

f f b
 0 ... (3.27)
a b a
The envelope is obtained by eliminating ‘a’ and ‘b’ from equations (3.25), (3.26) and (3.27).

20
3. The Singular Integral :
The envelope of the two-parameter family of surfaces z = F (x, y, a, b), which is obtained by
eliminating ‘a’ and ‘b’ from the equations
z = F (x, y, a, b),
Fa = 0, Fb = 0,
is called the singular integral of the first order partial differential equation.
Note : This solution cannot be obtained by giving any values to the constants a and b and hence is not
contained in the complete integral.
Result : Prove that singular integral is also a solution of the first order partial differential equation.
Proof : Let a two-parameter family of solutions
z = F (a, y, a, b) ... (3.28)
be a complete integral of the first-order partial differential equation
f (x, y, z, p, q) = 0. ... (3.29)
The singular solution of (3.29) is the envelope of (3.28). We will show that the envelope of this
two parameter family (3.28), if it exists, is also a solution of (3.29). Hence differentiating equation
(3.28) with respect to ‘a’ and ‘b’ we get respectively
Fa (x, y, a, b) = 0 ... (3.30)
and Fb (x, y, a, b) = 0 ... (3.31)
Now eliminating the parameters ‘a’ and ‘b’ between equations (3.28), (3.30) and (3.31) we obtain the
envelope
z = G (x, y) = F (x, y, a(x,y), b(x,y)), ... (3.32)
where a(x,y) and b(x,y) are obtained from equations (3.30), (3.31) by solving for ‘a’ and ‘b’ in terms
of x and y.
Differentiating equation (3.32) with respect to x and then with respect to y we get respectively
p  Gx  Fx  Fa a x  Fbbx ,

and q  G y  Fy  Fa a y  Fbb y .

By virtue of (3.30) and (3.31) we have

p  Gx  Fx and q  G y  Fy .
This shows that the envelope will have the some partial derivatives as those of a member of the
family. As the two-parameter family is the complete integral of the first order partial differential equation
(3.29). Hence p = Gx and q = Gy also satisfy the equation (3.29). This proves that the envelope of
the two parameter family (singular integral) is also a solution of the first order partial differential equation.
Note : The singular integral can also be found from the given partial differential equation without
knowing the complete integral.

21
Result : Prove that singular solution is obtained by eliminating p and q from the equations.
f (x, y, z, p, q) = 0, fp (x, y, z, p, q) = 0, fq (x, y, z, p, q) = 0.
Proof : Consider the first order partial differential equation given by
f (x, y, z, p, q) = 0. ... (3.33)
The complete integral of (3.33) is given by
z = F (x, y, a, b). ... (3.34)
Differentiating (3.34) with respect to x and y we get respectively

p  Fx  x, y, a, b  , ... (3.35)

and q  Fy  x, y, a, b  . ... (3.36)

Substituting equations (3.34), (3.35) and (3.36) in equation (3.33) we obtain

 
f x, y, F  x, y, a, b  , Fx  x, y, a, b  , Fy  x, y, a, b   0 . ... (3.37)

This holds identically for all ‘a’ and ‘b’. Now we shall show that equation (3.37) satisfies
fp = 0 and fq = 0.
Differentiating equation (3.37) with respect to ‘a’ and ‘b’ we get respectively.
f z Fa  f p Fxa  f q Fya  0 , ... (3.38)

and f z Fb  f p Fxb  f q Fyb  0 . ... (3.39)

However, on the singular solution, we have
Fa = 0 and Fb = 0.
Hence equations (3.38) and (3.39) reduce to
f p Fxa  f q Fya  0 , ... (3.40)

f p Fxb  f q Fyb  0 . ... (3.41)

Multiplying equation (3.40) by Fyb and equation (3.41) by Fya and subtracting we get

 
f p Fxa Fyb  Fxb Fya  0 . ... (3.42)
Since on the two-parameter family of surface (3.34)
Fxa Fyb  Fxb Fya  0 .
If Fxa Fyb  Fxb Fya  0

Fxa Fya
then  0,
Fxb Fyb

22
  Fa Fxa Fya 
and hence the matrix  F Fxa

Fyb  will not have rank two (Since Fa = 0, Fb = 0), which contradicts
 b
the fact that z = F (x, y, a, b) is a complete integral. Hence from equation (3.42) we have
fp = 0.
Similarly, we prove fq = 0.
This proves the result.
4. The Special Integral : Usually (but not always) the three integrals viz., the complete integral,
the general integral and the singular integral include all the integrals of the first order partial differential
equation f (x, y, z, p, q) = 0. However, there are some solutions of certain first order partial differential
equations which do not fall under any of the three classes. Such solutions are called “sepcial integrals”.
Example 1 : Show that z = ax + by + a2 + b2 is a complete integral of z = px + qy + p2 + q2.

By taking (i) b  1  a 2 , (ii) b = a, find the envelope of the sub-family. Further find the singular
integral.
Solution : Let

z  F  x, y, a, b   ax  by  a 2  b 2 . ... (3.43)

To prove z = F (x, y, a, b) is a complete integral of equation

z  px  qy  p 2  q 2 . ... (3.44)

 Fa Fxa Fya 
We prove the rank of the matrix  F Fxb

Fyb  is two.
 b
Thus from equation (3.43) we find
Fa  x  2a , Fb  y  2b, Fxa  1, Fya  0, Fya  1, Fxb  0

Hence the above matrix becomes

 Fa Fxa Fya   x  2a 1 0 
  
 Fb Fxb Fyb   y  2b 0 1 .

Obviously, the rank of the matrix is 2. Hence equation (3.43) is a complete integral of (3.44).

Case 1 : Take b  1  a 2

Then the one-parameter sub-family is given by

 
z  F x, y, a 1  a 2  ax  1  a 2 y  1 . ... (3.45)

Differentiating equation (3.45) with respect to ‘a’ we get

23
 ay
Fa  x   0. ... (3.46)
1 a2
From (3.45) we find

 1 a2 
z 1  a  x  y, ... (3.47)
 a 
 
where from (3.46) we find
2
a x a2  x
  2
  ,
1 a2 y 1 a  y


 a2 y2  x2  x2 , 
x2
 a2  .
x2  y2
Consequently, eliminating ‘a’ from equation (3.47) we obtain
2
 x2   y2 
 z  1 2
  2 
2  x   ,
 x  y  x 

  z  1  x 2  y 2 .
2
... (3.48)

This is the envelope (particular solution) of the equation (3.44).

Case : If b = a, then the one parameter sub-family of surfaces is given by

z  a ( x  y )  2a 2 . ... (3.49)
Differentiating this with respect to a we get
Fa  0  x  y  4a  0 ,

 x  y  4 a .
Substituting this value in equation (3.49) we get

 x y
z  
 x  y 2

2 ,
 4  16

 8z    x  y  .
2
... (3.50)

This is another envelope (particular solution) of the given partial differential equation (3.44).

24
Now to find singular integral of (3.44), we differentiate equation (3.43), with respect to ‘a’ and then
with respect to ‘b’, we get respectively
Fa  x  2a  0 , ... (3.51)

and Fb  y  2b  0 . ... (3.52)

Eliminating ‘a’ and ‘b’ between equations (3.43), (3.51) and (3.52) we get

x2 y 2 x2 y 2
z   
2 2 4 4


 4z   x2  y2 ,  ... (3.53)

which is a singular integral of (3.44).

Note : The singular integral of equation (3.44) can also be obtained directly by eliminating p and q
between equations (3.44) and
x
fp  x 2p  0  p   ,
2
y
and f q  y  2q  0  q   .
2
Substituting these in equation (3.44) we get


4z   x 2  y 2 
as the singular integral of equation (3.44)
Example 2 : Show that

 x  a 2   y  b 2  z 2  1
is a complete integral of

 
z 2 1  p2  q2  1.

By taking (i) b = 2a, (ii) b = a, show that the envelopes of the subfamily are respectively.

 y  2 x  2  5 z 2  5 and  x  y  2  2 z 2  2 ,
which are particular integrals. Show further that z  1 are the singular integrals.
Solution : Let

 
f  x, y , z , p, q   z 2 1  p 2  q 2  1  0 ... (3.54)

be the given partial differential equation.

25
 F  x , y , z , a, b    x  a    y  b   z 2  1
2 2
Let ... (3.55)

be the two-parameter family of surfaces.

Differentiating (3.55) with respect to a, b etc., we find
Fa  2( x  a ), Fb  2( y  b), Fxa  2, Fxb  0, Fxb  0, Fyb  2

Hence the matrix

 Fa Fxa Fya   2( x  a ) 2 0 

  
 Fb Fxb Fyb   2( y  b) 0 2 

has rank 2.

  x  a    y  b  z2  1 .
2 2

is a complete integral of equation (3.54)

Case 1 : Take b = 2a.
Hence from (3.55) the one-parameter sub-family of surfaces becomes.

 x  a  2   y  2a  2  z 2  1 . ... (3.56)

Differentiating (3.56) with respect to ‘a’ we get

2  x  a   4  y  2a   0 ,

 x  2 y  5a  0 , ... (3.57)

x  2y
a .
5
Substituting this in equation (3.56) we get
2 2
 x  2y   2x  4 y  2
x   y   z  1,
 5   5 

  y  2 x   5z 2  5 .
2
... (3.58)
This is the envelope of one parameter sub-family.
Case 2 : If we take b = a, then the one parameter family of sub-system becomes

 x  a 2   y  a 2  z 2  1 . ... (3.59)
Differentiating this equation with respect to ‘a’ we get
x  y  2a  0 ,

x y
a . ... (3.60)
2
26
Substituting the value of ‘a’ in equation (3.59) we get
2 2
 x y  x y 2
x   y    z 1,
 2   2 

  x  y   2z 2  2 .
2
... (3.61)

This is the envelope of one parameter family. Equations (3.58) and (3.61) are particular solutions of
(3.54).
Now to find singular integral of equation (3.54), we differentiate 2-parameter family of surfaces (3.55)
to get

Fa  x, y, z , a, b   x  a  0 , ... (3.62)

Fb  x, y, z , a, b   y  b  0 . ... (3.63)
The envelope is obtain by eliminating ‘a’ and ‘b’ between (3.55), (3.62) and (3.63). Thus
z  1 .
This shows that the envelope consists of the pair of planes z  1 . These planes are integral surfaces
of the equations (3.55). It is the singular integral of the equation.
Note : The characteristic curve of the two-parameter system (3.55) is the locus of points of intersection
of (3.55) with the plane (3.57). Since this plane passes through the centre of the sphere (a, 2a, 0),
hence the characteristic curve of the system is the great circle.

y
Example 3 : Show that z  ax   b is a complete integral of pq = 1. This problem has no singular
a
integral. Find the particular solution corresponding to the sub-family b = a.
Solution : Let the partial differential equation is given by

f  x, y, z, p, q   pq  1  0 . ... (3.64)
Let also the two parameters family of surfaces be given by

y
z  F  x, y , a, b   ax  b. ... (3.65)
b
We find

y
Fa  x  ,
a2

1
Fb  1, Fx  a, Fy 
a

1
 Fxa  1, Fxb  0, Fya   , Fyb  0 .
a2

27
  Fa Fxa Fya   x  y 1 
1 
  a2 a 2 
Hence the matrix
 Fb Fxb Fyb  
 1 0 0 
has rank two.
y
This proves that z  ax   b is a complete integral of equation (3.64).
a
Now if b = a, then from equation (3.65) we get one parameter sub-system as
y
z  ax  a. ... (3.66)
a
Differentiating this with respect to ‘a’ we get
y
0 x 1,
a2
y
 a2  . ... (3.67)
x 1
To eliminate ‘a’ from equation (3.66), we first write it as
za  a 2  x  1  y ,
z 2 a 2  a 4 ( x  1) 2  y 2  2a 2  x  1 y .
Putting the value of a2, we get

 y  2  x  1 2y
  x  1  y
2
z2    ( x  1) y ,
 x 1  y ( x  1)

 z 2  2( x  1) y  2 y 2 . ... (3.68)
This is the envelope of one parameter family and is the required particular solution of equation (3.64).
Now differentiating equation (3.65) with respect to ‘a’ and then with respect to ‘b’ we get respectively
y
Fa  x  , Fb  1
a2
y
Fa  0  x  2  0, Fb  0  1  0 .
a
This is not true.  the equation (3.64) has no singular integral.
Note : It is always possible to obtain different complete integrals which are not equivalent to each
other. These are not obtained from one another merely by a change in the choice of arbitrary constants.
Exercise :
Show that 2z   ax  y   b is a complete integral of px  qy  q 2  0 .
2
1.

28
 UNIT - II

LINEAR EQUATIONS OF THE FIRST ORDER

Introduction :
In this unit we study a method of finding a general integral of a quasi linear equation.
Theorem : The general solution of the Lagrange’s equation (quasi-linear equation).

P  x, y, z  p  Q  x, y, z  q  R  x, y, z  ,

where P, Q and R are given continuously differentiable functions of x, y, and z (and not vanishing
simultaneously) is F (u, v) = 0, where F is an arbitrary function of u and v and

u  x, y, z   C1 , v  x, y, z   C2

are the solutions of the system

dx dy dz
 
P  x, y , z  Q  x, y , z  R  x , y , z  .
Proof : Given that

u  x, y, z   C1 and v  x, y, z   C2 ... (1.1)

are the solutions of the system of differential equations

dx dy dz . ... (1.2)
 
P Q R
This implies that equation (1.1) satisfy equations (1.2),
 u x dx  u y dy  u z dz  0 and ... (1.3)

 v x dx  v y dy  v z dz  0 . ... (1.4)
This shows that the equations (1.3) and (1.4) must be consistent equations. Hence we have

u x P  u yQ  u z R  0 , ... (1.5)

and vx P  v yQ  vz R  0 . ... (1.6)

Solving equations (1.5) and (1.6) for P, Q and R we obtain

29
 P Q R
 
uy uz uz ux ux uy
vy vz vz vx vx vy

P Q R ,
  
 u y vz  u z v y  u v  u v 
 z x x z  uxv y  u yvx 
P Q R . ... (1.7)
  
  u, v    u, v    u, v 
  y , z    z , x    x, y 

Now we shall show that F (u, v) = 0 is a solution of Pp  Qq  R .

Consider the relation F (u, v) = 0. Differentiating this partially with respect to x and y we get

Fu  u x  pu z   Fv  v x  pv z   0 , ... (1.8)

and   
Fu u y  qu z  Fv v y  qv z  0 .  ... (1.9)

Eliminating Fu and Fv from equations (1.8) and (1.19) we get

 u x  pu z   v y  qvz    u y  qu z   vx  pv z  ,

  
 p u y vz  u z v y  q  u z vx  u xvz   u xv y  u y vx 
  u, v    u, v    u, v 
p q  . ... (1.10)
  y, z    z , x    x, y 

From equations (1.7) and (1.8) we find

Pp  Qq  R . ... (1.11)
This shows that F (u, v) = 0 is a solution of equation (1.11), where u (x, y,z) = C1 and
v (x, y, z) = C2 are solution of (1.2). This equivalently means that any surface F (u, v) = 0 generated by
the integral curves (1.2) is a solution of (1.11).
General Case :

Theorem : If ui  x1 , x2 ,..., xn , z   Ci , i = 1, 2, ...., n are independent solutions of the euqtions

dx1 dx2 dx dz
  ...  n  ,
P1 P2 Pn R

where P1, P2, ..., Pn and R are continuous differentiable functions of x1, x2, ..., xn and z not simultaneously
zero, then the relation

30
   u1 , u2 ,..., un   0,

where  is arbitrary, is a general solution of the quasi-linear partial differential equation

z z z
P1  P2  ...  Pn  R.
x1 x2 xn

Proof : We are given that

ui  x1 , x2 ,..., xn , z   Ci , i = 1, 2, ..., n ... (1.12)

are independent solutions of the equations

dx1 dx2 dx dz
  ...  n  ... (1.13)
P1 P2 Pn R

Differentiating equation (1.12) we get

u i u u u
dx1  i dx2  ...  i dxn  i dz  0 . ... (1.14)
x1 x 2 xn z

This shows that the equations (1.13) and (1.14) must be compatible (consistent).
n
ui u
 Pj  i R  0 , i = 1, 2, ..., n. ... (1.15)
j 1 x j z

For each i, we have n-equations. Solving these n-equations for P1, P2, ..., Pn and R we get

Pi R , ... (1.16)

  u1, u2 ,..., un    u1 , u2 ,..., un 
  x1 , x2 ,..., xi 1, z, xi 1,..., xn    x1, x2 ,..., xn 

u1 u1 u1


x1 x2 xn
u2 u 2 un
  u1 , u 2 ,..., un 
where  xi x1 xn
  x1 , x2 ,..., xn 

un u n u n
x1 x2 xn

is the Jacobian of the transformation. Now we shall show that the relation

F  u1 , u2 ,..., un   0 ... (1.17)

is a solution of quasi-linear partial differential equation

31
 z z z
P1  P2  ...  Pn R
x1 x2 xn

Differentiating equation (1.17) partially with respect to xi we get

 F u j u j z 
  u    0.
z xi  ... (1.18)
j  j xi

F F F
Eliminating u , u ,..., u from these n equations we get
1 2 n

z   u1 , u 2 ,..., u n    u1 , u 2 ,..., u n 
 x  . ... (1.19)
j j 
 x1 , x2 ,..., x j 1 , z , x j 1 ,..., xn    x1 , x2 ,..., xn 

From equations (1.16) and (1.19) we obtain

n
z
 x Pj  R ... (1.20)
j 1 j

This proves that if u1, u2, ..., un are independent solutions of (1.13) then F (u1, u2, ..., un) is a solution
of equations (1.20). This proves the theorem.
Example 1 : Find the general integral of

z  xp  yq   y 2  x 2 .

Solution : The given partial differential equations is

zxp  zyq  y 2  x 2 . ... (1.21)

The integral surface of the equation (1.21) is generated by the integral curves of the auxilary euqtion

dx dy dz . ... (1.22)
  2
zx  yz y  x 2

Consider the first two ratios of the equation

dx dy dx dy .
  
xz  yz x y
Integrating we get
log x   log y  log C1 ,

 xy  C1 . ... (1.23)
Now we consider each ratio of the equation (1.22)

32
 xdx  ydy  zdz ,

x z  y 2 z  zy 2  zx 2
2

 xdx  ydy  zdz  0 .

Integrating the equation we get

x 2  y 2  z 2  C2 . ... (1.24)
The curves given by equations (1.23) and (1.24) generate the integral surface


F xy, x 2  y 2  z 2  0, 
which is the general integral of equation (1.21).

Example 2 : Find the general integral of the partial differential equation

  
2x y  z 2 p  y 2 y  z 2 q  z3 . 
Solution : The given partial differential equation is

  
2x y  z 2 p  y 2 y  z 2 q  z3 .  ... (1.25)

The integral surface of equation (1.25) is generated by the integral curves of the auxiliary equation

dx dy dz
  . ... (1.26)

2x y  z 2  
y 2y  z2  z3

The first integral curve is obtained by considering each ratio of the equation (1.26) as

dx  dy  dz
x y z
,
2 y  2z  2 y  z  z 2
2 2

dx dy dz
    0.
x y z
Integrating gives
log x  log y  log z  log C1 ,

x
  C1 . ... (1.27)
yz
Now to find the second integral curve, consider the ratios

dy dz ,


y 2y  z 2
 z3

33
  z 3dy  2 y 2 dz  z 2 ydz ,

 z 2  ydz  zdy   2 y 2 dz ,

ydz  zdy 2dz

 2
 2 ,
y z

 z dz
d   2 2 .
 y z
Integration gives

z 2
  C2 . ... (1.28)
y z
Thus the curves given by equations (1.27) and (1.28) generate the integral surface

 x z2  2 y 
F  ,   0.
 yz yz 

Example 3 : Find the general integral of the partial differential equation

x ( x  y ) p  y ( x  y ) q  ( x  y )(2 x  2 y  z ) .
Solution : The integral surface of the equation
x ( x  y ) p  y ( x  y ) q  ( x  y )(2 x  2 y  z ) ... (1.29)
is generated by the integral curves of the auxilary equation

dx dy dz . ... (1.30)
 
x( x  y )  y ( x  y ) ( x  y )(2 x  2 y  z )
To find the first integral curve, consider the ratio

dx dy dx dy .
  
x( x  y )  y ( x  y ) x y
Integration yields
log x   log y  log C1 ,

 xy  C1 . ... (1.31)
Similarly, to find the second integral curve, each ratio of equation (1.30) is

dx  dy dx  dy  dz ,

( x  y )( x  y ) ( x  y )( x  y  z )

34
 dx  dy dx  dy  dz .
 
x y  x  y  z

Integration results in
log( x  y )   log( x  y  z )  log C2 .

 ( x  y )( x  y  z )  C2 , ... (1.32)
These curves (1.31) and (1.32) generate the integral surface

F  xy, ( x  y )( x  y  z )   0 .

Example 4 : Find the general integral of

x 2

 y 2 p  2 xyq  ( x  y ) z .

Solution : To find the integral surface of the equation

x 2

 y 2 p  2 xyq  ( x  y ) z , ... (1.33)

we first find the integral curves of the auxiliary equation

dx dy dz . ... (1.34)
 
2
x y 2
2 xy ( x  y ) z

To get the integral curve, we consider the ratios

dx  dy dz dx  y dz .
  
 x  y 2
( x  y) z x y z

Integration of which gives

log( x  y )  log z  log C1 ,

x y
  C1 . ... (1.35)
z
Similarly, the other integral curve is obtained by consider the ratios
dx
2
x y 2

dx
2 xy

 x 2  y 2 dy  2 xydx ,
 y 2 dy   x 2 dy  2 xydx ,
2xydx  x 2 dy
dy  ,
y2

 x2 
 dy  d   .
 y 

35
Integration results in

x2
y  C2 . ... (1.36)
y
Hence the integral surface generated by the curves (1.35) and (1.36) is given by

 x  y y2  x2 
F  , 0.
 z y 

Example 5 : Find the general integral of the partial differential equation

 xy 3
   
 2 x 4 p  2 y 4  x3 y 3 q  9 z x3  y 3 . 
Solution : The general solution of the equation

 xy 3
   
 2 x 4 p  2 y 4  x3 y 3 q  9 z x3  y 3  ... (1.37)

is the integral surface generated by the integral curves of the auxilliary equation

dx dy dz . ... (1.38)
 

xy 3  2 x 4 2 y 4  x 3 y 9 z x 3  y 3 
To find the integral curve, we first consider the ratios of the equation (1.38) as

dx dy dz

x y z
 ,
3 3 3
y  2x  2 y  x 3
9 x  y3
3
 
dx dy dz .
  
x y 3z
Integration of which gives

1
log x  log y   log z  log C1 ,
3

 x 3 y 3 z  C1 ... (1.39)
Now consider the ratios

dx dy
3 4
 4 ,
xy  2 x 2 y  x3 y

   
 2 y 4  x 3 y dx  xy 3  2 x 4 dy  0 .

Dividing by x3y3 we get

36
 2y dx dy 2 x
3
dx  2  2  3 dy  0 ,
x y x y

x
 xdy  2 ydx   y  ydx  2 xdy   0
,
x4 y4

 y   x 
 d  2  d  2   0.
x  y 
Integration yields

y x
2
 2  C2 . ... (1.40)
x y
Hence the general integral generated by the curves (1.39) and (1.40) is given by

 y x 
F  x 3 y 3 z, 2  2   0 .
 x y 

Example 6 : Find the general integral of

x 3
   
 3 xy 2 p  y 3  3 x 2 y q  2 x 2  y 2 z . 
Solution : The general solution of the equation

x 3
   
 3 xy 2 p  y 3  3 x 2 y q  2 x 2  y 2 z  ... (1.41)

is the integral surface generated by the integral curves of the auxilliary equations

dx dy dz
 3  . ... (1.42)
3
x  3 xy 2 2
y  3x y 2 x 2  y 2 z  
The first integral curve of (1.42) is obtained by considering the ratio

dx  dy  2 dz
x y z ,
2 2 2 2
x  3 y  y  3x  4 x  y  2 2

dx dy 2dz
    0.
x y z
Integrating we obtain
log x  log y  2 log z  log C1 ,

37
 xy
  C1 . ... (1.43)
z2
The second integral curve of (1.42) is obtained by considering the ratio

dz
xdx  ydy

1 z  2  xdx  ydy   dz .
4 2 2 4 2 2
x  3x y  y 3x y 2 x  y2
2
x2  y 2 z

Integrating we get

 
log x 2  y 2  log z  log C2 ,

 x2  y2 
    C2 . .... (1.44)
 z 

Hence the general integral is given by

 xy x 2  y 2 
F  2 ,   0 .
z z 
Exercise :
Find the general integral of the following partial differential equations.
1. ( y  1) p  ( x  1)q  z

2. z 2

 2 yz  y 2 p  x  y  z  q  x  y  z 

3. yzp  xzq  x  y

4. y 2 p  xyq  x  z  2 y 

5. xzp  yzq  xy

6. x 2
  
 yz p  y 2  zx q  z 2  xy

7. pq  2 z
Answers :

1. 
F x 2  y 2  2 x  2 y, z ( x  y )  0 
2. 
F x 2  y 2  z 2 , y 2  2 yz  z 2  0 
3. 
F x 2  y 2 , z 2  2( x  y )  0

38
4.  
F x 2  y 2 , y( y  z )  0

x 
5. F  , xy  z 2   0
y 

 x y yz
6. F , 0
 yz zx

7. F  x  y, x  z   0
2. Pfaffian Differential Equations :
Introduction : In this section, we introduce a Pfaffian difference equations. There is a fundamental
difference between Pfaffian differential equations in two variables and those in higher number of variables.
A Pfaffian differential equation in two variables always possesses an integrating factor. However, a
Pfaffian differential equation in more than two variables may not be integrable in general. We shall
derive in the following a necessary and sufficient condition for the integrability of a Pfaffian differential
equation in three variables.
Definition : A Pfaffian differential equation is a differential equation of the form
n
 Fi  x1, x2 ,..., xn  dxi  0 , ... (2.1)
i 1

where Fi  i  1, 2,..., n  are continuous functions of some or all of the n-independent variables

x1, x2, ..., xn , is called a Pfaffian differential equation, and the expression  Fi  x1 , x2 ,..., xn  dxi is
i
called a Pfaffian differential form.
Definition : A Pfaffian differential form is said to be exact if we can find a continuously difefrentiable
function u  x1, x2 ,..., xn  such that

du  F1  x1 , x2 ,..., xn  dx1  F2  x1, x2 ,..., xn  dx2  ...  Fn  x1 , x2 ,..., xn  dxn .

Definition : A Pfaffian differential equation is said to be integrable, if there exists a non-zero differentiable
function   x1 , x2 ,..., xn  such that the Pfaffian differential form

  F1  x1 , x2 ,..., xn  dx1  ...  Fn  x1 , x2 ,..., xn  dxn 

is exact. In this case the function   x1 , x2 ,..., xn  is called the integrating factor of the Pfaffian differential

equation and u  x1 , x2 ,..., xn   C , where C is an arbitrary constant, is called the integral of the
corresponding Pfaffian differential equation.

39
Note : Since the integral u  x1 , x2 ,..., xn   C of the Pfaffian differential equation (2.1) represents a
surface in  n , then it follows from Pfaffian differential equation that, at every point of the integral the
normal has direction ratios (F1, F2, ..., Fn).
Result : A Pfaffian differential equation in two variables always possesses an integrating factor.
Proof : Consider a Pfaffian differential equation in two variables x and y in the form
P ( x, y ) dx  Q ( x, y )dy  0 . ... (2.2)
If Q ( x, y )  0 , then we write this as

dy P ( x, y )
  f  x, y  , ... (2.3)
dx Q ( x, y )
where P (x, y) and Q (x, y) are known functions of x and y, so that f (x,y) is defined uniquely at each
point of the xy plane, at which the functions P (x, y) and Q (x, y) are defined. From the existence
theorem for a first order ordinary differential equation the equation (2.3) has a solution
F ( x, y )  C1 ... (2.4)

Result : If X  ( P, Q, R ) is a vector such that X  curl X  0 and  is an arbitrary differentiable

function of x, y and z then prove that

 
 X  curl  X  0

Proof : Let X   P, Q, R   Pi  Qj  Rk , ... (2.5)

be a vector, where i, j, k are unit vectors in the positive x, y and z directions respectively, such that

X  curl X  0 . ... (2.6)

By definition, we have

i j k
  
 
curl  X 
x y z
 P Q  R

  R  Q    P  R    Q  P 
 
 curl  X  i 
 y

z 
 j
 z

x 
k
 x

y 
,

  R  Q    P  R    Q  P 
   
  X  curl  X   P 
 y

z 
  Q 
 z

x 
  R
 x

y 
,

40
   R Q   P R   Q P  
   
  X  curl  X   2  P    Q    R   
  y z   z x   x y  

       
   PR  PQ  PQ  QR  QR  PR  .
 y z z x x y 

  R Q   P R   Q P  
   
  X  curl  X   2  P 
 y

z
Q
  z

x
  R
  x
  .
y  
... (2.7)

This can also be written as

  X   curl   X    2  P  Ry  Qz  . ... (2.8)

x, y , z

i.e.   X   curl   X    2  X  curl X  .

By virture of equation (2.6) we have

  X   curl   X   0 .
   
Conversely, let  X  curl  X  0 , for   0 then it follow from the definition

X  curl X  0 .
Note : The condition X  curl X  0 is equivalent to

 Q R   R P   P Q 
P   Q    R  0.
 z y   x z   y x 

Criteria of Integrability of a Pfaffian Differential Equation :

Note that all Pfaffian differential equations do not possesses integral. If however, the equation
is such that there exists a function   x, y, z  with the property that   Pdx  Qdy  Rdz  is an exact
differential d  , then the equation is said to be integrable. The function  is called the primitive of the
differential equation. In the following theorem we find a necessary and sufficient condition that a Pfaffian
differential equation is integrable.
Theorem : A necessary and sufficient condition that the Pfaffian differential equation X  dr  0 is
integrable is that X  curl X  0 , where X  ( P, Q, R ) is a vector..
Proof : Consider a Pfaffian differential equation in three variables x, y, z given by
P ( x, y, z )dx  Q ( x, y , z )dy  R ( x, y , z )dz  0 . ... (2.9)

41
If X  ( P, Q, R ) is a vector and r   x, y, z  ,  dr   dx, dy, dz  , then equation (2.9) can also be
written as

X dr  Pdx  Qdy  Rdz  0 ,

 X dr  0 . ... (2.10)
Let us assume that the equation (2.10) is integrable. We claim that

X  curl X  0 .

Since the equation (2.9) is integrable. This implies that there exist differentiable functions   x, y, z  and

u  x, y, z  such that.

du    x, y, z   Pdx  Qdy   Rdz  ,   0 ... (2.11)

where u  u ( x, y, z )

 du  u x dx  u y dy  u z dz . ... (2.12)
From equations (2.11) and (2.12) we find
 P  u x , Q  u y ,  R  u z

 u xi  u y j  u z k    Pi  Qj  Rk 

 u   X .
Taking the curl of the equation we get

 
curl     curl  X .

Since the identity curl     0

 
 curl  X  0 ,

 
  X curl  X  0 , 
 Xcurl X  0 .
 the equation X dr  0 is integrable if Xcurl X  0 .

Conversely, assume that Xcurl X  0 .

We prove that the Pfaffian differential equation Pdx  Qdy  Rdz  0 is integrable.
Let us assume that one of the variables say z is a constant. Hence the Pfaffian differential
equation becomes,

42
 P ( x, y, z )dx  Q( x, y , z )dy  0 . ... (2.13)
This is a Pfaffian differential equation in two variables, hence it is always integrable. This implies that
there exists a function U and the integrating factor  such that

dU    Pdx  Qdy  ,

 U x dx  U y dy    Pdx  Qdy  ,

 U x   P, U y  Q . ... (2.14)
Substituting the values of P and Q in equation (2.9) we get

U x dx  U y dy  U z dz    R  U z  dz  0 .

This is equivalent to
dU  Kdz  0 , ... (2.15)

where K  R U z . ... (2.16)

We are given that X  curl X  0 ,

  X  curl  X  0, ... (2.17)

where  X    P,  Q,  R  ,

 X    P,  Q, U z  K  , due to (2.16)

 
  X  U x , U y , U z   0,0, K  ,

 X  U   0, 0, K  .
Taking the curl of this equation and using the identity, curl of gradU = 0, we readily get

K K
curl  X  i j,
y x

 K K 
 curl  X   , ,0 . ... (2.18)
 y x 

Thus   X   curl   X    U x ,U y ,U z  K    Ky ,  Kx , 0  ,

U K U K
  ,
x y y x

43
   X   curl   X   ((Ux,, Ky)) .
Thus the condition (2.17) implies

 (U , K )
 0, ... (2.19)
 ( x, y )

 there exists a relation between U and K not involving x and y explicitly. Hence K can be
expressed as a function of U and z alone. Therefore equation (2.15) becomes,

dU
 K U , z   0 . ... (2.20)
dz
This is a first order ordinary differential equation, it possesses a solution. Let

 U , z   C ,
where C is an arbitrary constant, be a solution of equation (2.20). On replacing U by its expression in
terms of x, y and z we obtain the solution in the form
U ( x , y, z )  C .
Hence the Pfaffian differential equation (2.9) is integrable.
Note : The Pfaffian differential equation (2.9) is in fact, exact if and only if curl X  0 .
Show that the following Pfaffian differential equations are integrable and hence find the corresponding
integrals.

Example 1 : y 2
   
 yz dx  xz  z 2 dy  y 2  xy dz  0 .  ... (2.21)

Solution : Here P  y 2  yz , Q  xz  z 2 , R  y 2  xy .

Hence the vector X becomes


X  y 2  yz, xz  z 2 , y 2  xy , 
i j k
  
 curl X 
x y z ,
y 2  yz xz  z 2 y 2  xy

 curl X   2  y  x  z  , 2 y, 2 y  ,

 
 X  curl X  y 2  yz, xz  z 2 , y 2  xy  2  y  x  z  , 2 y , 2 y  ,

44
  2  y ( y  z )( y  x  z )  y ( xz  z 2 )  y 2 ( y  x)  ,


 2 y 3  xy 2  y 2 z  y 2 z  xyz  yz 2  xyz  z 2 y  y 3  xy 2 , 
 X  curl X  0 .
This proves that the Pfaffian differential equation (2.21) is integrable. Now to find the integral of (2.21)
we assume z = constant  dz = 0. Hence equation (2.21) becomes

y 2
 
 yz dx  xz  z 2 dy  0 . ... (2.22)

We write this equation as

dx dy
 0,
z( x  z ) y( y  z )

dx zdy
  0,
x  z y( y  z )

dx dy dy
   0.
x z y y z
Since z is a constant. On integrating we get
log( x  z )  log y  log( y  z )  log C1 ,

y( x  z)
  C1 ,
yz
where C1 is a constant, may be function of z. Let the integral of (2.22) be denoted by U.

y( x  z)
U  . ... (2.23)
yz
Hence there exist a function  such that

1  y ,
U x  P     
P yz

1 , ... (2.24)

 y  z 2

and U satisfies the equation

dU
 K  0, ... (2.25)
dz

45
where K  R U z

y  y  x y y x  z
K   ,
( y  z) 2
y  z  y  z 2

K = 0.
Hence equation (2.25) becomes

dU
0,
dz
 dU  0 ,
 U constant (independent of z),
 y( x  z)  C ( y  z) .
This is the required integral of (2.21).
Example 2 : yzdx  xzdy  xydz  0 .
Solution : The Pfaffian differential equation is
yzdx  xzdy  xydz  0 , ... (2.26)

where the vector X  ( P , Q, R )  ( yz , xz , xy ) .

We see that

i j k
curl X      i ( x  x)  j ( y  y )  k ( z  z )  0
x y z
yz xz xy

 curl X  0 ,

 X  curl X  0 . ... (2.27)

This shows that the equation (2.26) is integrable.
Now to find the integral of (2.26) we treat z = constant
 dz  0 .
Hence equation (2.26) reduces to
yzdx  xzdy  0

z  ydx  xdy   0 or d  xy   0

 xy  C1 , ... (2.28)
where C1 may be function of z.
46
Let U = xy. ... (2.29)

U
There must exist a function  such that  P
x

1 U 1 1
  ( y)    . ... (2.30)
P x yz z
The function U in (2.29) therefore satisfies the equation

dU
 K  0, ... (2.31)
dz

where U .
K  R 
z
Thus 1 xy .
K xy  0  K 
z z

U
K As U  xy ... (2.32)
z
Hence equation (2.31) becomes

dU U
  0,
dz z

dU dz
  0.
U z
Integrating we get
log U  log z  log C or Uz  C ,

 xyz  C , ... (2.33)

which is the required integral.

2
 2
 
Example 3 : yzdx  x y  zx dy  x z  xy dz  0 
Solution : The Pfaffian differential equation is given by

   
yzdx  x 2 y  zx dy  x 2 z  xy dz  0 ... (2.34)

where the vector

 
X  ( P, Q, R)  yz, x 2 y  zx, x 2 z  xy .

Therefore, we find

47
 i j k
curl X      i   x  x   j  2 xz  y  y   k  2 xy  z  z 
x y z
yz x 2 y  zx x 2 z  xy

 curl X   0,   2 xz  2 y  , 2 xy  2 z  .

Therefore, we see that

   
X  curl X  0  x 2 y  zx  2 y  2 xz    2 xy  2 z  x 2 z  xy 
 X  curl X  2  x 2 y 2  xyz  x 3 yz  x 2 z 2  x 3 yz  x 2 y 2  x 2 z 2  xyz 
 
X  curl X  0 . ... (2.35)
This shows that the equation (2.34) is integrable. Now to find the integral of (2.34) we treat
z  constant  dz  0 .
Hence equation (2.34) reduces to


yzdx  x 2 y  zx dy  0 ,
 yzdx  zxdy  x 2 ydy  0 ,

or z  ydx  xdy    x 2 ydy ,

 ydx  xdy 
or z    ydy ,
 x2 

 xdy  ydx   y
or z 2   ydy  zd    ydy .
 x  x
Integrating we get
2
 y y
z    C1 ,
x 2

zy y 2
or   C1 ,
x 2

y  2 z  xy 
  C1 ,
2x

48
 y  2 z  xy 
Let U . ... (2.36)
2x

U
There must exist a function  such that  P
x

1 U 1  yz 
or      2 ,
P x yz  x 

1 . ... (2.37)
 
x2
Also the function U in (2.36) satisfies the equation

dU
 K  0, ... (2.38)
dz
where

U
K  R 
z
1
  y
 K   2 x 2 z  xy    ,
x x

y y
 K  z   ,
x x
 K  z . ... (2.39)
Hence equation (2.38) becomes

dU
z0 or dU  zdz  0 .
dz
Integrating we get

z2
U C .
2

yz y 2 z 2
i.e.   C,
x 2 2

or
yz 1 2

 y  z2  C ,
x 2

or  
2 yz  x y 2  z 2  2 xC , ... (2.40)

which is the required integral.

49
Example 4 :  6 x  yz  dx   xz  2 y  dy   xy  2 z  dz  0
Solution : The Pfaffian differential equation is given by

 6 x  yz  dx   xz  2 y  dy   xy  2 z  dz  0 ,
... (2.41)
where the vector

X   6 x  yz, xz  2 y, xy  2 z  .

We find

i j k
Curl X      i( x  x)  j ( y  y )  k ( z  z )
x y z
6 x  yz xz  2 y xy  2 z

Curl X  0  X  curl X  0 ... (2.42)

The equation (2.41) is integrable. To find the integral of (2.42) we treat
z  constant  dz  0 .
Hence equation (2.41) becomes

 6 x  yz  dx   xz  2 y  dy  0
 6 xdx  yzdx  xzdy  2 ydy  0

 6 xdx  z  ydx  xdy   2 ydy  0

or 6 xdx  zd ( xy )  2 ydy  0 .
Integrating we get

3x 2  zxy  y 2  C1 , ... (2.43)

where C1 may involved z.

Let U  3 x 2  xyz  y 2 .
There must exist a function  such that
U
 P
x
1 U

P x
1
  6 x  yz     1 .
6 x  yz

50
Also the function U in (2.44) must satisfy the equation

dU
 K  0, ... (2.45)
dz
where

U
K  R 
z
 ( xy  2 z )  xy

 K  2z .
Hence equation (2.45) becomes

dU
 2 z  0 or dU  2 zdz  0 .
dz
Integarting we get

U  z 2  C2 ,

ie. 3x 2  xyz  y 2  z 2  C2 , ... (2.46)

which is the required integral of (2.41).

2
 3
 2 3
Example 5 : x z  y dx  3 xy dy  x dz  0 .

Solution : To test the integrability of the equation (2.47) we note that


X  x 2 z  y 3 ,3 xy 2 , x 3 . 
So that

i j k
curl X  
x

y

z   
 i (0  0)  j 3x 2  x 2  k 3 y 2  3 y 2 
x 2 z  y 3 3 xy 2 x3


Curl X  0, 2 x 2 , 6 y 2 
Therefore, 
X  Curl X  0  6 x 3 y 2  6 x 3 y 2  0 
 X  Curl X  0 ... (2.48)
 The equation (2.47) is integrable. Now to find the integral of (2.47) we treat

51
 z  constant  dz  0 .
Hence equation (2.47) reduces to

 x z  y  dx  3xy dy  0 .
2 3 2

We write this equation as

zx 2 dx  y 3 dx  3 xy 2 dy  0

 3 x 2 dx  3 xy 2 dy  y 3 dx  0

3xy 2 dy  y 3dx
 zdx  0
x2

 y3 
 zdx  d    0 . z = constant
 x 
Integrating we get

y3
zx   C1 . ... (2.49)
x

y3
Let U  zx  . ... (2.50)
x
There must exist a function  such that

U 1 U
  P or  
x P x

1  z  y3  1
  2     2 . ... (2.51)
x z  y 
2 3
 x  x

The function U in (2.50) also satisfies the equation

dU
 K  0, ... (2.52)
dz

U 1
where K  R   K  2 x3  x  K  0
z x
Therefore equation (2.52) becomes
 dU  0  U  C
i.e. x 2 z  y 3  Cx , ... (2.53)
which is required integral.
52
Example 6 : 1  yz  dx  x  z  x  dy  1  xy  dz  0 .
Solution : The Pfaffian differential equation is given by

1  yz  dx  x  z  x  dy  1  xy  dz  0 , ... (2.54)

where the vector

X  1  yz, x( z  x), (1  xy )  .

i j k
 curl X      i x  x  j  y  y  k  z  2x  z 
x y z
1  yz x( z  x) 1  xy

 curl X   2 x, 2 y  z , 2 x  .

We seet that

 
X  curl X  2 x 1  yz   2 y xz  x 2  2 x 1  xy 

 2 x  2 xyz  2 xyz  2 x 2 y  2 x  2 x 2 y

X  curl X  0 ... (2.55)

 the equation (2.54) is integrable. Now to find the integral of (2.54) we treat
z  constant  dz  0 .
Therefore equation (2.54) reduces to

1  yz  dx  x  z  x  dy  0 .
We write this as

dx dy
  0,
x  z  x  1  yz

zdx 3dy
or   0,
x  z  x  1  yz

dx dx dy
  0. z = constant
x zx y 1
z
Integrating we get


log x  log( z  x)  log y  1
z   log C ,1

53
  1
x y  
 z
 C1
 z  x

x  yz  1
or  C1 .
z  z  x

Let x  yz  1 . ... (2.56)

U
z  z  x

Therefore, there must exist a function  such that

U
 P
x

1 U 1   yz  1 z  x   x  yz  1 
or    
P x z 1  yz    z  x  2



1  z  x  x 1
  . ... (2.57)
z  z  x 2
 z  x  2

The function U in (2.56) therefore satisfies the equation

dU
 K  0, ... (2.58)
dz
where

U
K  R 
z

1  z  z  x  xy  x  yz  1 2 z  x  
K   1  xy     
 z  x 2  z2  z  x
2


K
z 2
1
 z  x 2  
  z 2 (1  xy )  z 2 xy  zx 2 y  2 xyz 2  2 zx  x 2 yz  x 2 
 



 z 2  2 xz  x 2     z  x 2

z2  z  x z2  z  x
2 2

1
K  . ... (2.59)
z2

54
Hence equation (2.58) becomes

dU 1 dz
 2  0 or dU  2  0 .
dz z z
Integrating we get

1
U  C2
z

x  yz  1 1
or   C2
z  z  x z

 1  xy   C2  z  x  .

Example 7 :  2x  y 2

 2 xz dx  2 xydy  x 2 dz  0 .

Solution : The Pfaffian differential equation is given by

 2x  y 2

 2 xz dx  2 xydy  x 2 dz  0 , ... (2.60)

where the vector


X  2 x  y 2  2 xz, 2 xy, x 2 
i j k
 curl X    
x y z
2
2 x  y  2 xz 2 xy x2

curl X  i (0  0)  j (2 x  2 x )  k (2 y  2 y )

curl X  0  X  curl X  0 ... (2.61)

 the equation (2.60) is integrable. To find the integral of (2.60) we treat
x  constant  dx  0 .
Thus equation (2.60) reduces to

2 xydy  x 2 dz  0 .
Integrating we get

y2
2x  x 2 z  C1
2

or xy 2  x 2 z  C1 .

55
Let U  xy 2  x 2 z . ... (2.62)
There must exist a function  such that

U U
Q  , R 
y z

1 U 1
  2 xy    1 . ... (2.63)
Q y 2 xy
Also the function U satisfies the equation

dU
 K  0, ... (2.64)
dx

U
where K  P 
x


K  2 x  y 2  2 xz  y 2  2 xz 
 K  2x . ... (2.65)
Hence equation (2.65) becomes

dU
 2x  0 .
dx
Integrating we get

U  x 2  C2

or xy 2  x 2 z  x 2  C2 ... (2.66)
which is the solution of equation (2.60).

Example 8 : 1  yz  dx  z  z  x  dy  1  xy  dz  0 ... (2.67)

Solution : Here X  1  yz, z ( z  x), (1  xy ) 

i j k
 curl X    
x y z
1  yz z ( z  x) (1  xy )

curl X  i   x  2 z  x   j   y  y   k   z  z 

curl X  2 zi  2 yj  2 zk

56
 curl X   2 z , 2 y, 2 z 

 
 X  curl X  1  yz, z 2  xz , 1  xy  2 z , 2 y , 2 z 

X  curl X  0 . ... (2.68)

 the given equation (2.67) is integrable. Consider,,
x  constant  dx  0 .
Therefore, given equation becomes

z  z  x  dy  1  xy  dz  0 ,

dy dz
   0,
1  xy z  z  x 

xdy xdz
   0,
1  xy z  z  x 

dy dz dz
   0. x = constant
 1 z zx
y 
 x

Integrating we get

 1
log  y    log z  log  z  x   log C1 ,
 x

 1
z y  
 
x
 C1 ,
( z  x)

z  yx  1
or  C1 .
x( z  x)

z  yx  1
Let U . ... (2.69)
x  z  x

There must exist a function  such that

U U ,
Q  or  R 
y z

1 U 1  xz 
   ,
Q y z  z  x   x  z  x  
57
 xz 1 . ... (2.70)
 
xz  z  x   z  x
2 2

The function U satisfies the equation

dU
 K  0, ... (2.71)
dx

where

U
K  P 
x

1  x  z  x  yz  z  xy  1 z  2 x  
2
K 1  yz    
 z  x  x2  z  x
2

On simplifying we get

1 1
K  z  x 2  K 
x 2
 z  x 2
x2 . ... (2.72)

dU 1 dx
 2  0 or dU  2  0 .
dx x x

Integrating we get

1
U  C2
x

z  xy  1 1
   C2
x  z  x x

 z  xy  1   z  x   C2 x  z  x 

 z 1  xy    z  x 1  C2 x  . ... (2.73)

58
Exercise :
Show that the following Pfaffian differential equations are integrable and hence find the corresponding
integrals.

1. z  z  y  dx  z  x  z  dy  x  x  y  dz  0

2. yzdx  2 xzdy  3 xydz  0

3. ydx  xdy  2 zdz  0

4.  yz  xyz  dx   zx  xyz  dy   xy  xyz  dz  0

Answers :
1. z  x  y  C  x  z

2. xy 2  Cz 2

3. xy  z 2  C

59
 UNIT - III

COMPATIBLE SYSTEMS OF FIRST ORDER PARTIAL

DIFFERENTIAL EQUATIONS

Introduction :
In this unit we introduce a system of first order partial differential equations and find the conditions
that the system has common solution. As discussed in the Unit 2, the method of finding the general
integral of Lagrange’s equation, in this unit we introduce methods due to Chanpits and Jacobi to find
the complete integral of the first order partial differential equations.
Definition : Two first order partial differential equations

f  x , y , z , p, q   0 ... (1.1)

and g  x, y , z , p , q   0 ... (1.2)

are said to be compatible (they have a common solution) on a domain D, if and only if

 f , g
J 0 on D ... (1.3)
  p, q 

and the equation

dz    x, y, z  da   x, y, z  dy ... (1.4)

is integrable, where p    x, y, z  and q    x, y, z  are obtained by solving (1.1) and (1.2).

Theorem : A necessary and sufficient condition for the two partial differential equations

f  x, y, z, p, q   0 and g  x, y, z , p, q   0

to be compatible is that
 f , g  f ,g  f , g  f , g
 f , g  p  q 0.
  x, p    z, p    y, q    z, q 
Proof : Consider two first order partial differential equations

f  x, y, z, p, q   0 ... (1.5)

and g  x, y, z , p, q   0 . ... (1.6)

By definition, equations (1.5) and (1.6) are said to compatible iff

 f , g
J 0 ... (1.7)
  p, q 

60
and p    x, y, z  and q    x, y, z 

obtainable from (1.5) and (1.6) render the equation

dz    x, y, z  dx   x, y, z  dy ... (1.8)

integrable. We write equation (1.8) as

  x, y, z  dx   x, y, z  dy  dz  0 ... (1.9)
We know the condition that the equation (1.9) is integrable iff

X  curl X  0 ,
where

X   , , 1

i j k
  
 curl X 
x y z
  1


 curl X  i   z    z j  k  x   y 

 curl X   z ,  z , x   y . 
Thus the condition X  curl X  0 becomes

 , , 1   z , z , x   y   0
  x   z   y   z . ... (1.10)

Substituting  and  for p and q respectively in equation (1.5) we get

f  x, y, z,  ,   0 ... (1.11)

Differentiating equation (1.11) with respect to x and z we get

f x  f p x  f q x  0 ... (1.12)

and f z  f p z  f q z  0 . ... (1.13)

Multiplying equation (1.13) by  and adding it to the equation (1.12) we get

f x   f z  f p  x   z   f q  x   z   0 . ... (1.14)
Similarly, from equation (1.6) we obtain

61
 g x   g z  g p  x   z   g q  x   z   0 . ... (1.15)
Multiplying equation (1.14) by gp and (1.15) by fp and subtracting we get

   
f x g p  f p g x   f z g p  f p g z   x   z  f q g p  f p g q  0

    
 f p g q  f q g p  x   z   f x g p  f p g x   f z g p  f p g z 
 f ,g  f ,g  f , g
  x   z   
  p, q    x, p    z, p 

1   f ,g  f ,g
or  x   z    , ... (1.16)
J    x, p    z , p  

 f , g
where J  0.
  p, q 

Similarly, differentiating equation (1.11) with respect to y and z we obtain, after similar analysis, the
equation

1    f , g   f , g 
 y  z    . ... (1.17)
J    y, q    z , q  

Now substituting equations (1.16) and (1.17) in the equation (1.10) we obtain

 f , g  f ,g  f , g  f , g
   .
  x, p    z, p    y, q    z, q 

Replacing  and  by p and q respectively, we get

 f , g  f ,g  f , g  f ,g
p  q 0. ... (1.18)
  x, p    z , p    y, q    z, q 

This is the desired compatibility condition. This condition can also be written as

 f , g  0.
Example 1 : Show that the equations
2 2

f  p 2  q 2  1  0 and g  p  q x  pz  0 
are compatible and find the one parameter family of common solutions.
Solution : Let the partial differential equations be given by

f  x, y , z , p , q   p 2  q 2  1  0 ... (1.19)

62
  
g  x, y, z , p, q   p 2  q 2 x  pz  0 . ... (1.20)

We know the condition that the equations (1.19) and (1.20) are compatible iff

 f , g  0
 f , g  f ,g  f , g  f ,g
i.e. p  q 0, ... (1.21)
  x, p    z , p    y, q    z, q 

where from equations (1.19) and (1.20) we readily obtain

x, ... (1.22)
pz  x  p 
z
2
 x 1 2
and q 2  1  p 2  q 2  1    or q  z  x2 . ... (1.23)
z z
We find from equations (1.19) and (1.20) that

f x  0, f p  2 p, g x  p 2  q 2 , g p  2 px  z

 f ,g  f , g
Therefore,
  x, p 
 
 f x g p  f p g x  2 p p 2  q 2 
  x, p 
x
 2  2 p .
z

Similarly,
f z  0, g z   p

 f , g  f ,g x2
Therefore,  f z g p  f p g z  2 p 2   2 p2  2 2 .
  z, p    z, p  z

Similarly,
f y  0, f q  2q, g y  0, g q  2qx

 f , g
Therefore,  f y gq  fq g y  0 .
  y, q 

Next we find

 f , g
 f z g q  f q g z  2q   p   2 pq
  z, q 

 f , g 2x
  2 pq  2 z 2  x 2 .
  z, q  z

Substituting these in equation (1.21) we get

63
  f , g   2 p  p  2 p 2   0  q  2 pq 


 2 p  2 p p 2  q 2 
 2 p  2 p As p 2  q 2  1 by ... (1.19)
=0

  f , g  0 . ... (1.24)
This shows that the equations (1.19) and (1.20) are compatible.
Now to solve these equations we have
dz  pdx  qdy

x 1 2
 dx  z  x 2 dy
z z

 zdz  xdx  z 2  x 2 dy


zdz  xdx
2
z x 2
 dy  d  
z 2  x 2  dy

Integrating we get

z2  x2  y  c

z 2  x2   y  c
2
or

z 2  x2   y  c ,
2
or ... (1.25)

which is a required one parameter family of common solution.

Example 2 : Show that the equations

xp  yq, z  xp  yq   2 xy

are compatible and find a one parameter family of common solution.

Solution : Let the partial differential equation be given by

f  x, y, z, p, q   xp  qy  0 , ... (1.26)

g  x, y, z , p, q   z  xp  yq   2 xy  0 . ... (1.27)
We know the condition that the equations (1.26) and (1.27) are compatible iff

 f , g  0,

64
where

 f , g  f , g  f , g  f , g
 f , g  p  q . ... (1.28)
  x, p    z, p    y, q    z, q 
From equations (1.26) and (1.27) we find
f x  p, f p  x, g x  zp  2 y , g p  zx

 f , g
Therefore,  f x g p  f p g x  2 xy .
  x, p 

Now we find
f z  0, g z  xp  yq

 f , g  f , g
Thus,  f z g p  f p g z   x  xp  yq     x 2 p  xyq .
  z, p    z, p 

Similarly,
f y   q, f q   y , g y  zq  2 x, g q  zy

 f , g
Hence,  f y g q  f q g y  2 yx ,
  y, q 

and

 f ,g  f , g
 f z g q  f q g z  y  xp  yq    xyp  y 2 q .
  z, q    z, q 

Substituting these values in equation (1.28) we get

 f , g     p 2 x 2  y 2q 2   0
  f , g  0

 Equations (1.26) and (1.27) are compatible. Now to solve these equations, we find from (1.26)
and (1.27) that

y x
p , q .
z z
Substituting this in dz  pdx  qdy we get

y x
dz  dx  dy
z z
 zdz  ydx  xdy

65
 zdz  d ( xy ) .
Integrating we get

z2
 xy  C1
2

or z 2  2 xy  C .
Which is the required one parameter family of common solution.

Example 3 : Show that the equation z = px + qy is compatible with any equation f  x, y, z, p, q   0

that is homogeneous in x, y and z.
Solution : Since f is a homogeneous in x, y, z, therefore it can be written as

x y 
f  z n  , , p, q  . ... (1.29)
z z 

x y
Put u ,v  .
z z
Therefore, the given equations reduce to

g  x, y, z, p, q   px  qy  z  0 ... (1.30)

and h  x, y, z, p, q     u, v, p, q   0 . ... (1.31)

We know equations (1.30) and (1.31) are compatible iff

 g , h  0 ,
where we know

  g, h   g, h   g, h   g, h
 g , h  p  q . ... (1.32)
  x, p    z, p    y, q    z, q 

Therefore, we have

  g, h
 g x h p  g p hx
  x, p 

 p p  x x

 p p  xu u x

  g, h x
 p p  u .
  x, p  z

66
Similarly,

  g, h
 g z hp  g pqz
  z, p 

  p  x z

  g, h
   p  x z
  z, p 

  g, h
 g y hq  g q h y
  y, q 

 qq  yv v y

  g, h y
 qq  v ,
  y, q  z

and

  g, h
 g z hq  g q hz
  z, q 

  g, h
 q  y z .
  z, q 

Substituting these in [ g, h ] we get

x y
 g , h  p p  u  p p  px z  qq  v  qq  q y z
z z

 uu  vv   px  qy   z

 uu  vv  z u  u z  v v z  by ... (1.30)

 x y 
 uu  vv  z  u 2  v 2 
 z z 

x y
 uu  vv  u  v
z z

 uu  vv  uu  vv

 g , h  0
 the given equations are compatible.

67
Exercise :
1. Show that the following first order partial differential equations are compatible and find
a one-parameter family of common solution.
xp  yq  x,

x 2 p  q  xz.

2. Show that the equations f  x, y, p, q   0 and g  x, y, p, q   0 are compatible if

 f , g  f , g
  0.
  x, p    y , q 

2. Charpit’s Method :
In this section, we present a method of finding complete integral of a first order p.d.e.
f  x, y, z, p, q   0 due to Charpit. Method is based on the concept of the last section viz. the concept
of compatibility.
Definition : Let a first order p.d.e. be given by

f  x, y, z, p, q   0 . ... (2.1)

A one-parameter family of p.d. equations given by

g  x, y, z , p, q, a   0 , a is a parameter, ... (2.2)

is said to be compatible with (2.1) if (2.2) is compatible with (2.1) for each value of a.
Result : Describe Charpit’s Method of solving a first order partial differential equation

f  x , y , z , p, q   0 .

Proof : Let the first order p.d.e. whose complete integral is to be determined be given by

f  x , y , z , p, q   0 . ... (2.3)
The fundamental idea in Charpit’s method is the introduction of a second partial differential equation of
the first order

g  x, y , z , p , q , a   0 ... (2.4)
which contain an arbitrary constant ‘a’ and which is such that

(i) equations (2.3) and (2.4) can be solved for p  p  x, y, z  and q  q  x, y, z  and

(ii) the equation dz  p  x, y, z  dx  q  x, y, z  dy is integrable.

68
i.e. we need only to seek an equation

g  x, y , z , p , q , a   0

Compatible with the given equation

f  x , y , z , p, q   0

We know equations (2.3) and (2.4) are compatible iff

 f , g  0
 f , g  f ,g  f , g  f ,g
i.e. p  q 0. ... (2.5)
  x, p    z , p    y, q    z, q 

      
 f x g p  f p g x  p f z g p  f p g z  f y gq  f q g y  q f z gq  fq g z  0 . 
We write this as

g g g g g
 fp
x
 fq
y

 pf p  qf q
z

  f x  pf z   f y  qf z
p q
0 
g g g g g
or fp
x
 fq
y

 pf p  qf q
z

  f x  pf z   f y  qf z
p q
 0.  ... (2.6)

This is a quasi-linear first order partial differential equation for g with x, y, z, p and q as the
independent variables.
Thus our problem of finding a one-parameter family of p.d. equations (2.4) which is compatible
with the given p.d.e. (2.3) is equivalent to find a solution of equation (2.6) in as simple form as possible
involving p or q or both and an arbitrary constant a.
This we do by finding an integral of the following subsidiary equations involving and arbitrary
constant.

dx d y dz dp dq . ... (2.7)
   
f p f q pf p  qf q f x  pf z f y  qf z

Once an integral g  x, y, z , p, q, a  of this kind has been found, solving the p.d.e. (2.3) and
the integral thus obtained for p and q, we get

p    x, y , z , a  , q    x , y , z , a  .

Then
dz   dx   dy ... (2.8)
is integrable by virtue of the fact that the equations (2.1) and (2.2) are compatible.
Let the integral of (2.8) be of the form

69
 F  x, y, z, a, b   0 . ... (2.9)
This is a two-parameter family of solutions of (2.3), it is a complete integral of (2.3).

Example 1 : Find the complete integral of z 2  pqxy  0 by Charpit’s method.

Solution : Let f  x, y, z , p, q   z 2  pqxy  0 ... (2.10)

To find a one-parameter family of p.d.e. which is compatible with (2.10), we know the auxiliary
equations are

dx dy dz dp dq , ... (2.11)
   
f p f q pf p  qf q f x  pf z f y  qf z

where from equation (2.10) we have

f p   qxy , f q   pxy , f x   pqy , f y   pqx, f z  2 z .

Hence the equations (2.11) become

dx dy dz dp dq .
   
 qxy  pxy  pqxy  pqxy  pqy  2 pz  pqx  2qz

dx dy dz dp dq
or     ... (2.12)
qxy pxy 2 pqxy 2 pz  pqy 2qz  pqx
Each ratio of (2.13) is also equal to

pdx  qdy  xdp  ydq


pqxy  pqxy  2 pxz  pqxy  2qzy  pqxy

each ratio pdx  qdy  xdp  ydq .


2 z  px  qy 

Consider

dz pdx  qdy  xdp  ydq


2 pqxy 2 z  px  qy 

Since from equation (2.10) we have

z2
pq 
xy

dz pdx  qdy  xdp  ydq

Hence, 
2z 2
2 z  px  qy 

70
 dz pdx  qdy  xdp  ydq
 
z px  qy

dz d  xp  yq 
 .
z px  qy
Integrating we get

log z  log  px  qy   log a

 z  a  px  qy  , ... (2.13)
where a is an arbitrary constant.

Let g  x, y, z, p, q, a   z  a  px  qy  . ... (2.14)

Thus equations (2.10) and (2.14) are compatible.
Solving equations (2.10) and (2.14) for p and q we obtain

cz
p
z
cx
and q 
y
1

with a c  c  1 . 
Hence the equation dz  pdx  qdy becomes

z cz
dz  dx  dy
cx y

dz 1 dx 1
   c dy .
z c x y
Integrating we get

1
log z  log x  c log y  log b
c
1
z  bx c yc ... (2.15)

1
 F  x, y, z, b, c   z  bx c yc ... (2.16)

which is the complete integral of the first order p.d.e. (2.10). This is a two-parameter family of solutions
of equation (2.10) and is the required complete integral.

Example 2 : Find the complete integral of

p 2

 q 2 y  qz  0 .

71
Solution : Let the p.d.e. be given by

 
f  x, y, z , p, q   p 2  q 2 y  qz  0 . ... (2.17)

From equation (2.17) we have

f p  2 py, f q  2qy  z , f x  0, f y  p 2  q 2 , f z  q .

Hence the auxiliary equations (2.7) becomes

i.e. dx dy dz dp dq
    2 . ... (2.18)
2 py 2qy  z 2 p y  2q y  qz pq
2 2
p
Consider the ratio

dp dq
 ,
q p

 pdp   qdq .
Integrating we get

p2  q2  a2 , ... (2.19)
where ‘a’ is a constant.

Let g  x, y , z , p, q, a   p 2  q 2  a 2  0 ... (2.20)

which is compatible with (2.17). Now to find the complete integral of (2.17) we solve equations (2.17)
and (2.20) for p and q.
Hence we write from (2.17) and (2.20) that

p 2

 q 2 y  qz ,

and p 2

 q 2  a 2  qz  a 2 y

or a2 y .
q
z

y2
Hence p 2 y  a 2 y  ya 4
z2

y2 a 2
or p2  a2  a4 2
 p z  a2 y2 .
z z
Hence the equation
dz  pdx  qdy
72
becomes

a 2 2 2 a2 y
dz  z  a y dx  dy
z z

zdz  a 2 ydy
or  adx
z2  a2 y2

i.e. d  
z 2  a 2 y 2  adx .

Integrating we get

z 2  a 2 y 2  ax  b

z 2  a 2 y 2   ax  b  .
2
or

Hence the equired complete integral is

z 2  a 2 y 2   ax  b  .
2
... (2.21)

Example 3 : Find the complete integral of the p.d.e.


p 2q 2  x 2 y 2  x 2q 2 x 2  y 2 
Solution : Let the given p.d.e. be denoted by

 
f  x, y , z , p , q   p 2 q 2  x 2 y 2  x 2 q 2 x 2  y 2  0 . ... (2.22)

From equation (2.22) we have

f p  2 pq 2 , f q  2qp 2  2 x 2 q  x 2  y 2  , f z  0,

f x  2 xy 2  q 2  4 x 3  2 xy 2  , f y  2 yx 2  2 yx 2 q 2 .

Hence the auxiliary equations (2.7) become

dx dy dz
  
2 pq 2 2qp 2  2 x 2 q  x 2  y 2  2 p 2 q 2  2q 2 p 2  2 x 2 q 2  x 2  y 2 

 dp  dq .
 
2 xy 2  2q 2  2 x 3  xy 2 
2 2 2
2 yx  2 x yq
Consider the ratios

73
 qdy  ydq

2 2
2 p q  2x q 2 2
x 2
y 2
 2 x y  2 x 2 y 2q 2
2 2

qdy  ydq
 
2 x y 1  q 2 
2 2
2 x y 2 2 by ... (2.22)

dy dq dy  1 1 1  .
      dq
y q 1  q 
2 y  q 2 1  q  2(1  q) 

Integrating we get

1 1
log y  log q  log 1  q   log 1  q   log a
2 2

2 log y  2 log q  log 1  q   log 1  q   2 log a

 q2  2
 log y 2  log  2 
a
 1  q 

q2
 y2  a2
1 q2

or 1  q 2  y 2  a 2 y 2  q 2  y 2  a 2   y 2
y . ... (2.23)
q 1
 y 2  a2  2

Substituting this in equation (2.22) we get

 y2  y2
p2  2 2
  x 2 2
y  x 2
 x 2
 y 2

y a  y2  a2

x2 y 2   y 2  a 2   x 2  y 2 
  
y 2
a 2

 p2  x2  x2  a2 
1
p  x  x2  a2  2. ... (2.24)

Substituting these values in the equation

dz  pdx  qdy

74
 1 y
dz  x  x 2  a 2  2
dx  1
dy
y 2
a 2
 2

Integrating we get

y
z   x x 2  a 2 dx   dy  b
y2  a2
3 1
x  a2  2   y2  a2  2  b .
1 2
z ... (2.25)
3
Example 4 : Find the complete integral of

px 5  4q 3 x 2  6 x 2 z  2  0
by Charpit’s method.
Solution : Let

f  x, y, z , p, q   px 5  4q 3 x 2  6 x 2 z  2  0 . ... (2.26)

From equation (2.26) we find

f p  x 5 , f q  12q 2 x 2 , f x  5 px 4  8q 3 x  12 xz , f y  0, f z  6 x 2 .

Hence the auxiliary equations (2.7) become

dx dy dz  dp dq
5
 2 2
 5 3 2
 4 3 2
 2 . ... (2.27)
x 12q x px  12q x 5 px  8q x  12 xz  6 px 6x q
Consider the ratios

dx dq
5
 2
x 6x q

dx dq
6 3
 .
x q
Integrating we get

 1 
6   2    log q  log a
 2x 

3
3  q  2
 2  log    q  a  e x
x  C1 

9
3
 q  a e 3 x2 .

75
Substituting in (2.26) we get
9
9
2 x2
3 x2 4a e 6z 2 .
px 5  4a e x2  6x2 z  2  p   
x3 x3 x5
Hence the equation dz  pdx  qdy becomes

 9  3
 4a 3e x 2 6 z 2  2
dz   3
 3  5  dx  ae x dy
 x x x 

 9  3
6z  3
4a e x 2
2 2
dz  3 dx   3
 5  dx  ae x dy
x  x x 

3  6

3 
6 z   4a e x 
 3 2 2
or x2  2e x
e  dz  dx      dx  ady
 x3   x3 x5 

 6

3 
  3   3 x2 2 
 2  4a e 2e x
 d  ze x    dx  ady .
 x3 x5 
Integrating we get
6 3
3 

ex2 e x2
ze x2  4a 3  dx  2 5 dx  ay  b ... (2.28)
3
x x

6
2
ex
Consider I  dx
x3

6
Put t
x2

12 dx  dt
 3
dx  dt  3 
x x 12
6
1 1 2
Hence, I    e t dt   e x
12 2

6
6
e a3 2 x2
 4a  3 dx   e x
3
x 3
76
 3

e x2
Now consider 2 dx
x5

3 1 t 6
Put  2
 t  2    3 dx  dt
x x 3 x
3

e x2 dt  t  1 tet
 2 dx  2 e t
9  dt
    
x5 6  3
1 t
 t  e  et 
9
3 6

x2
a3 2
Thus ze   e x  ay  b
3
3 9 3 3 3

a3 2 1 2 3 2
 z   ay  b 
2 2
ex  ex  ex  2 ex e x
3 9 x

3 9
a3 2 1 1
z   ay  b  e x2  ex  2  . ... (2.29)
3 3x 9
Example 5 : Find the complete integral of the p.d.e.

2 z  p 2  qy  2 y 2  0
by Charpit’s method.
Solution : Let

f  x, y, z, p, q   2 z  p 2  qy  2 y 2  0 . ... (2.30)
From equation (2.30) we have
f p  2 p, f q  y , f x  0, f y  q  4 y , f z  2

Hence equations (2.7) become

dx dy dz dp dq .
   
2p 2
y 2 p  qy 2p q  4 y  2q

Consider the ratio

dx dp
  dx   dp .
2p 2p
Integrating we get

77
 x  pa ... (2.31)
or pax.
Substituting this in (2.30) we get
2
2 z   a  x   qy  2 y 2  0 .

1
 2 z   a  x   2 y  .
2 2
q ... (2.32)
y

Substituting these equation in dz  pdx  qdy we get

1
 2 z   a  x   2 y  dy
2
dz   a  x  dx  2
y

2
2z a  x
 dz  dy   a  x  dx  dy  2 ydy
y y

2
ydz  2 zdy  a  x  ydx   a  x  dy
  2 ydy
y y

2
ydz  2 zdy   a  x  ydx   a  x  dy  2 y 2 dy
Multiplying this equation by 2y we get
2
2 y 2 dz  4 zydy  2 y 2  a  x  dx  2 y  a  x  dy  4 y 3dy

or  
d  2 zy 2     d y 2  a  x    d  y 4  .
2

Integrating we get
2
2zy 2  y 2  a  x   y 4  b

y 2  2 z   a  x   y 2   b .
2
or ... (2.33)

Which is the required complete integral.

Example 6 : Find the complete integral of

2  z  xp  yq   yp 2
by Charpit’s method
Solution : Let

f  x, y, z, p, q   2  z  xp  yq   yp 2  0 ... (2.34)

where from equation (2.34)

78
 f p  2 x  2 yp, f q  2 y, f x  2 p, f y  2q  p 2 , f z  2 .

Hence the auxiliary equations (2.7) become

dx dy dz dp dq .
    
2  x  yp  2 y 2  xp  yq   2 yp 2
4 p 4q  p 2

Consider the ratios

dy dp 2dy dp
 or  .
y 2p y p
Integrating we get
2 log y  log p  log a ,

a
 y 2 p  a or p  2 . ... (2.35)
y
Substituting this in (2.34) we get

 a  a2 ,
2  z  x 2  yq   y 4
 y  y

a2  a ,
 2 yq   2  z  x
y3  y 2 

or a 2  z xa  . ... (2.36)
q  
2 y 4  y y 3 

Substituting in dz  pdx  qdy we get

a  a 2  z ax  
dz  dx   4    3   dy ,
y2  2 y  y y  
a ax a2 z
 2
dx  3
dy  4
dy  dy ,
y y 2y y
z a ax a2
 dz  dy  2 dx  3 dy  4 dy ,
y y y 2y

ydz  zdy a ax a2
  2 dx  3 dy  4 dy ,
y y y 2y

a ax a2
d ( yz )  dx  2 dy  3 dy ,
y y 2y
79
 2
 ydx  xdy  a
 a   2 y 3 dy
 y2 

 x  a2
d ( yz )  a   dy .
 y  2 y 3

Integrating we get
2
x a  y 
2
yz  a     b
 y  2  2 

ax b a 2
or z  
y2 y 4 y3

ax b a 2
z  
y2 y 4 y3
which is the required complete integral.
Example 7 : Find the complete integral of the p.d.e.

z  p 2  q 2   px  qy  0 .
Solution : Let the given p.d.e. be denoted by

f  x, y, z , p, q   z  p 2  q 2   px  qy  0 , ... (2.37)

 f p  2 pz  x, f q  2qz  y , f z   p 2  q 2  , f x  p, f y  q
Hence the auxiliary equations (2.7) reduce to

dx dy dz  dp  dq
   
2 pz  x 2qz  y 2 p z  px  2q z  qy p  p  p  q  q  q  p 2  q 2 
2 2 2 2

Consider the ratios

qdp pdq qdp  pdq qdp  pdq

    0
pq  pq  p  q
2 2
 pq  pq  p  q
2 2
 0 q2

 p
d 0
q

Integrating we get

80
 p
a
q
or p  aq . ... (2.38)
Substituting this in (2.37) we get

z  a 2 q 2  q 2   aqx  qy  0

q 2  z  a 2  1   q  ax  y   0

q  qz  a 2  1   ax  y    0

 ax  y 
for q 0q  ... (2.39)
z  a 2  1
Substituting these values in dz  pdx  qdy we get
dz  aqdx  qdy

 ax  y 
  adx  dy 
z  a 2  1

1
zdz    ax  y  adx  y  
or  a 2  1 
1
 zdz   2  ax  y  adx  dy 
2  a 2  1

1
d  ax  y  .
2
zdz  
2  a  1
2

Integrating we get

z2 1
  ax  y  2  b
2 2  a  1
2

2  ax  y  2
z  b
 a 2  1

 ax  y  2
or z2  b. ... (2.40)
a2 1
Example 8 : Find the complete integral of the equation

81
 xp  3 yq  2  z  x 2 q 2 
by Charpit’s method.
Solution : Let

f  x, y, z , p, q   xp  3 yq  2  z  x 2 q 2   0 . ... (2.41)

From equation (2.41) we obtain

f p  x, f q  3 y  4 x 2 q, f x  p  4 xq 2 , f z  2 .

Hence the auxiliary equations (2.7) become

dx dy dz  dp  dq
    .
2 2 2
x 3 y  4 x q px  3 yq  4q x p  4 xq  2 p 3q  2q
2

Consider the ratio

dx dq .

x q
Integrating we get
log x  log q  log a  xq  a ,

a
or q . ... (2.42)
x
Substituting (2.42) in (2.41) we get

a  x 2a 2 
xp  3 y  2 z  2   0 ,
x  x 

3ay
 px  2 z  2a 2  ,
x

2  z  q 2  3qy
or p  2 . ... (2.43)
x x
Substituting these values in the equation dz  pdx  qdy we get

 2  z  q 2  3ay  a
dz    2  dx  dy ,
 x x  x

1   a
dz  2 
2 x z  a 2   3ay  dx  dy
x x

82
  x 2 dz  2 x  x  a 2  dx 
or x4    3aydx  axdy ,
 x2 

 z  a2  
x 4d    3aydx  axdy ,
 x2 

 z  a2  a 3ay
or d  2   3 dy  4 dx ,
 x  x x

 z  a2   ay 
d 2 d  3.
 x  x 
Integrating we get

z  a 2 ay
 3 b
x2 x

or x  z  a 2   ay  bx 3 ... (2.44)

which is the required complete integral.

Example 9 : Find the complete integral of the p.d.e.
pxy  pq  qy  yz
by Charpit’s method.
Solution : Let

f  x, y, z , p, q   pxy  pq  qy  yz  0 . ... (2.45)

From equation (2.45) we have

f p  xy  q, f q  p  y , f x  py , f y  px  q  z , f z   y .

Hence the auxiliary equations (2.7) reduce to

dx dy dz dp  dq
   
xy  q p  y pxy  pq  pq  qy py  py px  q  z  qy
we see that
dp  0 ,
 p a. ... (2.46)
Substituting in equation (2.45) we get
axy  aq  qy  yz  0 ,

83
 q  a  y   yz  axy

y  z  ax  . ... (2.47)
q
a  y
Substituting these values in dz  pdx  qdy we get

 z  ax 
dz  adx  y dy
a  y .
We write this as

 z  ax 
 dz  adx  y dy
a y

dz  adx ydy dz  adx  a 

    1  dy  .
z  ax a y z  ax  a y 
Integrating we get

log  z  ax   y  a log( y  a )  log b

 z  ax    y  a  a
 log y
b

  z  ax   y  a   be y
a

a
z  ax  be y  y  a  .

This is required complete integral.

Example 10 : Find the complete integral of the p.d.e.

x 2 p 2  y 2q 2  4  0
Solution : Let the p.d.e. be given by

f  x, y, z, p, q   x 2 p 2  y 2 q 2  4  0 . ... (2.48)

 f p  2 px 2 , f q  2qy 2 , f x  2 xp 2 , f y  2 yq 2 , f z  0

Hence auxiliary equations (2.7) become

dx dy dz dp dq . ... (2.49)
2
 2
 2 2 2 2
 2

2 px 2qy 2 p x  2q y 2 xp 2 yq 2

84
Now consider the ratios

dx dp dx dp
2
 2
  .
2 px 2 xp x p

Integrating we get
log x   log p  log a
 xp  a .

Let g  x, y, z , p, q, a   xp  a  0 ... (2.50)

be the one-parameter family of p.d.c. which is compatible with (2.48).
Now to find the complete integral of (2.48) we solve

x 2 p 2  y 2 q 2  4 and xp  a for p and q to get

a 4  a2
p and q 
x y
Hence the equation
dz  pdx  qdy

a dy
becomes dz  dx  4  a 2 .
x y
Integrating we get

z  a log x  4  a 2 log y  b . ... (2.51)

This involves two arbitrary constants and hence it is called the complete integral of (2.48).
If however, we choose the ratio

dy dq dy dq
2
 2
 
2 y q 2 yq y q

Integrating we get
log y   log q  log a
 yq  a

Let g  x, y , z , p, q   yq  a  0 ... (2.52)

This is a one-parameter family of p.d.e. compatible with (2.48). Solving we get

a.
q
y

85
Also from equation (2.48) we get

4  a2
x2 p2  a2  4  p  .
x
Hence the equation dz  pdx  qdy becomes

dx a
dz  4  a 2  dy .
x y
Integrating we get

z  4  a 2 log x  a log y  b . ... (2.53)

This equation involves two auxiliary constants and hence is called a complete integral of (2.48).
Note : Equations (2.51) and (2.53) are not different.
Example 11 : Find the complete integral of

p2x  q2 y  z
by Charpit’s method.
Solution : Let

f  x, y, z, p, q   p 2 x  q 2 y  z  0 ... (2.54)

From this equation we find

f p  2 px, f q  2qy, f x  p 2 , f q  q 2 , f z  1 .

Hence the auxiliary equations (2.7) become

dx dy dz dp dq .
  2 2
 2  2
2 px 2qy 2 p x  2q y p p q q
Consider the ratios

p 2 dx  2 pxdp q 2 dy  2qydq

2 p3 x  2 p3 x  2 p 2 x 2q 3 y  2q 3 y  2q 2 y

p 2 dx  2 pxdp q 2 dy  2qydq .
 
p2x q2 y
Integrating we get

log  p 2 x   log  q 2 y   log a

 p 2 x  aq 2 y ... (2.55)

86
Using this in (2.54) we get

z
aq 2 y  q 2 y  z  q 2 y 
a 1
1
 z  2
q  . ... (2.56)
 (a  1) y 
Hence from equation (2.55) we get
1

p  
az  2

 x  z  1 
Substituting these values in
dz  pdx  qdy
1 1
 az  2  z  2
We get dz   dx   dy
 x  a  1  
 ( a  1) y 

dz dx dy .
 1 a  a 
z x y

Integrating we get

1  a  z  ax  y  b

1 1 1
 a  1 z  2   ax  2  y 2  b .
This is the required complete Integral.
Example 12 : Solve the p.d.c. by Charpit’s method

p   z  qy  .
2

Solution : Let

f  x, y, z , p, q   p   z  qy   0
2
... (2.57)

be the given non-linear p.d.e.

Where from equation (2.57) we find

f p  1, f q  2 y  z  qy  , f x  0, f y  2q  z  qy  , f z  2  z  qy  .

Hence the auxiliary equation (2.7) becomes

dx dy dz dp dq
     2q  z  qy  ... (2.58)
1 2y  z  qy  p  2qy  z  qy  2p  z  qy  2q  z  qy 

87
Consider the ratios

dy dp dy dp
   .
2y  z  qy  2p  z  qy  y p

Integrating we get
log y  log p  log a
 yp  a ... (2.59)

a
or p .
y
Putting this value in (2.57) we get

a a
  z  qy  or z  qy 
2
y y

or 1 a 
q   z . ... (2.60)
y y 
Substituting these values in
dz  pdx  qdy ,
we get

a 1 a 
dz  dx    z  dy ,
y y y 

a z a
dz  dx  dy  3 dy ,
y y y 2

y
 ydz  zdy  adx  a  3
dy ,
y 2

1
d  yz   adx  a dy .
y

Integrating we get

yz  ax  2 ay  b . ... (2.61)
Example 13 : Find the complete integral of the p.d.e.

z 2  p2z2  q 2   1 .

88
Solution : Let

f  x, y, z, p, q   z 2  p 2 z 2  q 2   1  0 , ... (2.62)

where

f p  2pz 4 , f q  2qz 2 , f x  0, f y  0, f z  4z 3 p 2  2zq 2 .

Hence the auxiliary equation (2.7) become

dx dy dz  dp dq .
4
 2  2 4 2 2
 3 3 2
 3 2 3
2pz 2z q 2p z  2q z 4z p  2zpq 4z p q  2zq
Consider the ratios

dp dq ,

p  4z p  2zq
3 2 2
 q  4z p  2zq 2 
3 2

dp dq .
 
p q
On integrating we get
log p  log q  log a  p  aq . ... (2.63)
Substituting this value in equation (2.62) we get

z 2  a 2q 2 z 2  q 2   1  q 2 z 2 1  a 2 z 2   1 ,

1 ,
q2 
z 1  a 2 z 2 
2

or 1 . ... (2.64)
q
2 2
z 1 a z

Substituting in dz  pdx  qdy we get

z 1  a 2 z 2 dz  adx  dy .

On integrating we get
3
 a 2z 2  1 2
 ax  y  b . ... (2.65)
3a 2

89
Example 14 : Find the complete integral of the p.d.e.

2x  z 2q 2  1  pz
by Charpit’s method
Solution : Let

f  x, y, z, p, q   2x  z 2 q 2  1  pz  0 ... (2.66)

 f p   z, f q  4xz 2 q, f x  2  z 2 q 2  1 , f y  0, f z  4xzq 2  p .

Hence the auxiliary equations (2.7) become

dx dy dz dp dq
   
 z 4xz q  zp  4q xz 2  z q  1  4xzq p  p
2 2 2 2 2 2 2
4xzq 3  pq

Consider the ratios

dz  dq
z q

 p  4xq z  p  4xzq 2
2

dz dq
  .
z q
Integrating we get
log z   log q  log a
 zq  a

a.
q
z
Substituting this in (2.66) we get

2x  a 2  1  pz  0

x.
 p  2  a 2  1
z
Substituting this in dz  pdx  qdy we get

x a
dz  2  a 2  1 dx  dy
z z

 zdz  2  a 2  1 xdx  ady

Integrating we get

90
 z2  2  2
 a  1 x  ay  b ,
2

or z 2  2  a 2  1 x 2  2ay  b .
This is the complete integral of (2.66).
Example 15 : Obtain the complete integral of the p.d.e.

z 2 1  p 2  q 2   1 .
Solution : Let

f  x, y, z, p, q   z 2 1  p 2  q 2   1  0 ... (2.67)

be the given p.d.e. Where from equation (2.67) we obtain

f p  2pz 2 , f q  2qz 2 , f x  0, f y  0, f z  2z 1  p 2  q 2  .

Hence the auxiliary equations (2.7) become

dx dy dz dp dq . ... (2.68)

  2 2  
2pz 2
2qz 2 2 2
2p z  2q z  2
2zp 1  p  q 2
  2
2zq 1  p  q 2

Consider the ratios

dp dq ,

2zp 1  p 2
q 2
 2zq 1  p 2  q 2 

dp dq .
 
p q
Integrating we get
log p  log q  log a ,
 p  aq .

Let g  x, y, z, p, q, a   p  aq  0 ... (2.69)

be the one-parameter family of p.d.e. compatible with (2.67). Solving equations (2.67) and (2.69) we
get

z 2 1  a 2 q 2  q 2   1

 z 2q 2  a 2  1  1  z 2 .

1
 1   1 z 
2
2 1  1  2
q   2  2   q   2  1 .
 a 1  z  a 1  z
2 

91
Therefore, the equation dz  pdx  qdy becomes

1 1
 1a  2  1
1  2
dz   2  1 dx   2  1 dy
a 1  z
2  a 1  z
2 

dz a 1
 1
 dx  dy
 1  2 a 2 1 a2 1
 2  1
z 

zdz a 1
or
1
 dx  dy .
1  z 2  2 a 2 1 a2 1

Integrating we get

a 1
  1 z2  x yb
a 2 1 a 2 1

  a 2  1 1  z 2    y  ax  b  .
2

This is the required complete integral.

Note : A first order p.d.e. can have several complete integrals. Note however that the two complete
integrals are equivalent, in the sense that one can be obtained from another merely by changing the
arbitrary constants.
Remark : However, when one complete integral has been obtained, every other solution, including
every other complete integral can be obtained. We shall explain the procedure in the next section.

3. Some Standard Types of p.d.e.

Type (I) : This type of equation is of the form

f  p, q   0 ... (3.1)
i.e. The given partial differential equation does not involve x, y and z.

Hence f x  0, f y  0,f z  0 .

From auxiliary equations (2.7) we have

dx dy dz dp dq .
   
f p f q pf p  qf q 0 0

Solving the last equation we get either p = a or q = a.

Putting this in (3.1) we get

92
 f  a, q   0 or f  p, a   0

 q  Q(a) or p  P(a) .
Therefore, putting this in dz  pdx  qdy we get
dz  adx  Q(a)dy .
Integrating we get
z  ax  Q(a)y  b
or z  P(a)x  ay  b .
Type (II) : This type of equation is of the form

f  z, p, q   0 ... (3.2)
i.e. the given p.d.e. does not involve x and y explicitly.
The auxiliary equations become

dx dy dz dp dq
    ... (3.3)
f p f q pf q  qf q  pf z qf z

Consider the last two ratios

dp dq
  p  aq ... (3.4)
p q
Substituting in (3.2) we get

f  z, aq, q   0 or q  Q(a, z) ... (3.5)

Therefore, the equation dz  pdx  qdy becomes

dz  aQ(a, z)dx  Q(a, z)dy

1
or dz  adx  dy
Q(a, z)
Integrating we get

dz
 Q(a, z)  ax  y  b . ... (3.6)

Type (III) : This type of equations is of the form

g(x, p)  h(y, q) (separable type) ... (3.7)
and not z is involved.
The auxiliary equations are

93
 dx dy dz dp dq
   
g p  hq pg p  qh q g x h y

Consider the ratios

dx dx

gp gx

 g x dx  g p dp

 g x dx  g p dp  0 .

Integrating we get

g  x, p   a ... (3.8)

 from (1) that

h  y, q   a . ... (3.9)
Solving these equations (3.8) and (3.9) for p and q we get
p  G(a, x), q  H(a, y) .
Therefore, the equation dz  pdx  qdy becomes
dz  G(a, x)dx  H(a, y)dy .
Integrating we get

z   G  a, x  dx   H  a, y  dy  b ,

which is the complete integral.

Type (IV) : This type of equation is of the form

z  px  qy  f  p, q  . ... (3.10)
This is called Clairaut form of partial differential equation.
The auxiliary equations are

dx dy dz  dp  dq
   
 x  g p  y  g q  xp  pg p  yq  qg q  p  p q  q

 dp  o  p  a and q  b
Substituting this in equation (3.10), we obtain its complete integral in the form

z  ax  by  f  a, b  . ... (3.11)
e.g. Find the complete integral of the p.d.e.

94
 pqz  p 2  xq  p 2   q 2  yp  q 2  .
Solution : The given partial differential equation is in Clairaut form, hence its complete integral is given
by

a 4  b4
z  ax  by  .
ab
4. Jacobi’s Method
Introduction : Let a partial differential equation be

F  x, y , z , p , q   0 ... (4.1)

and u  x, y , z   0 ... (4.2)

be the solution of (4.1)
Differentiating (4.2) w.r.t. y we get

uy
u y  uzq  0  q   .
uz

Substituting the values for p and q in equation (4.1), let the equation (4.1) reduce to

f  x, y , z , u x , u y , u z   0 . ... (4.3)

This is a p.d.e., in which x, y and z are the independent variables and the dependent variable u does not
appear explicitly in the equation.

Complete integral of f  x, y, z , u x , u y , u z   0 :

A function u  x, y, z , a, b, c  is said to be a complete integral of (4.3) if it satisfies the p.d.e.

and the associated matrix

 Fa Fax Fay Faz 

 
 Fb Fbx Fby Fbz 
F Fcx Fcy Fcz 
 c

is of rank three.

Theorem : Let f  x, y , z , u x , u y , u z   0 ... (4.4)

be a p.d.e. Show that any function h given by

h  x, y , z , u x , u y , u z   0

is compatible with (4.4) is

95
   f , h   f , h   f , h
  0.
  x, u x    y , u y    z , u z 

Proof : Let f  x, y , z , u x , u y , u z   0 ... (4.5)

be a given p.d.e. in which x, y, z are independent variables. Differentiating (4.5) w.r.t. x, y and z we get

f f f
fx  u xx  u yx  u zx  0 , ... (4.6)
u x u y u z

f f f
fy  u xy  u yy  u zy  0 , ... (4.7)
u x u y u z

f f f
fz  u xz  u yz  u zz  0 . ... (4.8)
u x u y u z

Consider

h  x, y , z , u x , u y , u z   0 , ... (4.9)

where h  hi , i = 1, 2, ...
On differentiating equation (4.9) w.r.t. x, y and z we obtain

h h h
hx  u xx  u yx  u zx  0 , ... (4.10)
u x u y u z

h h h
hy  u xy  u yy  u zy  0 , ... (4.11)
u x u y u z

h h h
and hz  u xz  u yz  u zz  0 . ... (4.12)
u x u y u z

h f
Multiply equation (4.6) by u and (4.12) by u and subtracting we get
x x

h f  f h f h   f h f h  . ... (4.13)
fx  hx  u xy       u xz  u  u  u  u   0
u x u x  u y u x u x u y   z x x z 

h f
Now multiplying equation (4.7) by u and (4.11) by u and subtracting we get
y y

96
 h f  f h f h   f h f h  . ... (4.14)
fy  hy  u xy       u yz     0
u y u y  u x u y u y u x   u z u y u y u z 

h f
Similarly, on multiplying equation (4.8) by u and (4.12) by u and subtracting we obtain
z z

h f  f h f h   f h f h 
fz  hz  u xz       u yz       0. ... (4.15)
u z u z  u x u z u z u x   u y u z u z u y 
Adding equations (4.13), (4.14) and (4.15) we get

 h f   h f   h f 
 f x u  u hx    f y u  u hy    f z u  u hz   0 . ... (4.16)
 x x   y y   z z 

  f , h   f , h   f , h
or   0. ... (4.17)
  x, u x    y , u y    z , u z 

Equation (4.16) can also be written as

h h h h h h
fux  fu y  fuz  fx  fy  fz 0 ... (4.18)
x y z u x u y u z

which is the required result.

Jacobi’s Method :
Result : Describe Jacobi’s Method of solving the first order partial differential equation of the form

f  x, y , z , u x , u y , u z   0 .

Proof : Let the first order partial differential equation whose complete integral is to be determined be
given by the equation.

f  x, y , z , u x , u y , u z   0 , ... (4.19)

where x, y, z are independent variables.

The fundamental idea of Jacobi’s method is the introduction of two partial differential equations
of the first order

h1  x, y, z , u x , u y , u z , a   0 , ... (4.20)

h2  x, y, z , u x , u y , u z , b   0 , ... (4.21)

each involving one arbitrary constant ‘a’ and ‘b’ such that

97
   f , h1 , h2 
(i)  0 on D and
 ux , u y , uz 

(ii) the Pfaffian equation

du  u x  x, y, z  dx  u y  x, y, z  dy  u z  x, y, z  dz

is integrable, where u x , u y , u z are obtained by solving equations (4.19), (4.20) and (4.21)
i.e. we seek functions h1, and h2 such that the equations (4.20) and (4.21) are compatible with (4.19).
We know that any h ( = hi , i = 1,2 ) compatible with equation (4.19) is given by

  f , h   f , h   f , h
  0
  x, u x    y , u y    z , u z  ... (4.22)

h h h h h h
 fx  fux  fy  fu y  fz  fu z  0.
u x x u y y u z z

We write this as

h h h h h h
fux  fu y  fuz  fx  fy  fz  0. ... (4.23)
x y z u x u y u z

This is the first order partial differential equation for h with x, y, z, ux, uy and uz as the independent
variables.
Hence the subsidiary equations of (4.23) are

dx dy dz du x du y du z
     . ... (4.24)
fux fu y fuz  f x  f y  f z

From equation (4.24) we find two integrals involving arbitrary constants ‘a’ and ‘b’ of the form

h1  x, y, z , u x , u y , u z , a   0 , ... (4.25)

and h2  x, y, z , u x , u y , u z , b   0 . ... (4.26)

These integrals are such that, equations (4.19), (4.25) and (4.26) can be solved for ux, uy, uz. These
values of ux, uy and uz are then substituted in
du  u x dx  u y dy  u z dz ... (4.27)
which is integrable. The integral satisfying (4.19) is of the form

  x , y , z , a , b, c   0 . ... (4.28)
This is the required complete integral of equation (4.19).

98
Remark : The conditions for the equation
du  u x dx  u y dy  u z dz

to be exact are

u x u y u y u z u z u x
 ,  , 
y x z y x z
These conditions are obviously true. Hence the equation is either exact or not integrable at all.
Example 1 : Solve the equation

z 2  zu z  u x 2  u y 2  0

by Jacobi’s method.
Solution : Let

f  x, y, z , u x , u y , u z   z 2  zu z  u x 2  u y 2  0 ... (4.29)

be the first order partial differential equation.

Where from equation (4.29) we find

f u x  2u x , f u y  2u y , f u z  z , f x  0, f y  0, f z  2 z  u z

Hence the auxiliary equations (4.24) become

dx dy dz du x du y du z . ... (4.30)
    
2u x 2u y z 0 0 2 z  u z

From this we obtain two independent solutions from the ratios

du x du y

0 0

 du x  0  u x  a ,

 du y  0  u y  b ,

which are the two integral of (4.29). Substituting these in the equation (4.29) we obtain

z 2  zu z  a 2  b 2  0 ,

a2  b2  z 2
or uz  .
z
Substituting these values in the equation
du  u x dx  u y dy  u y dy

99
we get

 a 2  b2  z 2 
du  adx  bdy    dz ,
 z 

dz
du  adx  bdy   a 2  b 2   zdz .
z
Integrating we get

z2
u  ax  by   a 2  b 2  log z  C .
2
This is the required complete integral of (4.29).

Example 2 : Solve the equation

z  2u z   u x  u z   0
2

by Jacobi’s method.
Solution : Let

f  x, y , z , u x , u y , u z  z  2 y z   u x  u z   0 .
2
... (4.31)

We first find two one-parameter family of p.d.e. which are compatible with (4.31)
From equation (4.31) we find

f u x  2  u x  u y  , f u y  2  u x  u y  , f u z  2, f x  0, f y  0, f z  1 .

Hence the auxiliary equations (4.24) reduce to

dx dy dz du x du y du z .
    
2  u x  u y  2  u x  u y  0 0 0 1

From which we obtain two independent integrals by considering the ratios

du x du y
 ,
0 0

 du x  0  u x  a ,

du y  0  u y  b .

Let h1  x, y, z , u x , u y , u z , a   u x  a  0 , ... (4.33)

and h2  x, y, z , u x , u y , u z , b   u y  b  0 , ... (4.34)

100
which are compatible with (4.31). Substituting this in (4.31) we get
2
z  2u z   a  b   0 ,

1
or uz   a  b 2  z  .
2
Substituting these values in the equation
du  u x d x  u y d y  u z d z ,

we get

1
du  adx  bdy   a  b  2  z  dz .
2
Integrating we get

1 2 z2
u  ax  by   a  b  z   C
2 4
which is the required complete integral of (4.31).

Example 3 : Solve the equation

u x x 2  u y 2  au z 2  0

by Jacobi’s method.
Solution : Let

f  x, y, z , u x , u y , u z   u x x 2  u y 2  au z 2  0 ... (4.35)

be a given partial differential equation. From equation (4.35) we find

f u x  x 2 , f u y  2u y , f u z  2au z , f x  2 xu x , f y  0, f z  0 .

Hence, the auxiliary equations (4.24) become

dx dy dz du x du y du
     z ... (4.36)
x 2
2u y 2au z 2 xu x 0 0

From which we obtain two integrals by considering the ratios

du y du z

0 0

 uy  0  uy  b ,

101
and du z  0  u z  c .
Let these be denoted by

h1  x, y, z , u x , u y , u z , a   u y  b  0 , ... (4.37)

and h2  x, y, z , u x , u y , u z , b   u z  c  0 . ... (4.38)

Substituting these in (4.35) we obtain

b 2  ac 2
u x x 2  b 2  ac 2  u x  . ... (4.39)
x2
Substituting these values in the equation
du  u x dx  u y dy  u z dz

we get

 b 2  ac 2 
du  dx  bdy  cdz .
x2
Integrating

 b 2  ac 2 
u  by  cz  d ... (4.40)
x
Example 4 : Solve the equation by Jacobi’s method

ux 2  u y 2  uz  1

Solution : Let

f  x , y, z , u x , u y , u z   u x 2  u y 2  u z  1  0 ... (4.41)

be a given p.d.e. Where from equation (4.41) we find

f u x  2u x , f u y  2u y , f u z  1, f x  0, f y  0, f z  0 .

Substituting these in auxiliary equations (4.24) we find

dx dy dz du du y du
   x   z . ... (4.42)
2u x 2u y 1 0 0 0

Consider the ratios

du x du y
  du x  0
0 0

ux  a ,

102
and du y  0  u y  b .

Let these be denoted by

h1  x, y, z , u x , u y , u z , a   u x  a  0 , ... (4.43)

h2  x, y, z , u x , u y , u z , b   u y  b  0 . ... (4.44)

Substituting these in (4.41) we get

a2  b2  uz 1  0

 uz  1   a 2  b2  . ... (4.45)

Substituting these values in

du  u x dx  u y dy  u z dz ,

we get

du  adx  bdy  1   a 2  b 2   dz .

Integrating we get

u  ax  by  1  a 2  b 2  z  c .
Example 5 : Solve the equation by Jacobi’s method

xu x  yu y  u z 2 .

Solution : Let

f  x, y, z , u x , u y , u z   xu x  yu y  u z 2  0 . ... (4.46)

We find two one-parameter family of p.d.e. which are compatible with (4.46).
From equation (4.46) we have
f u x  x, f u y  y , f u z  2u z , f x  u x , f y  u y , f z  0 .

Hence the auxiliary equations (4.24) become

dx dy dz du du y du
   x   z .
x y 2u z ux uy 0

From which we obtain two independent integrals by considering

 du z  0  u z  a , ... (4.47)

dx du
and   x  log x   log u a  log b
x ux

103
 xu x  b . ... (4.48)
Equations (4.47) and (4.48) are compatible with (4.46). Therefore, substituting (4.47) and (4.48) in
(4.46) we get

1 2
b  yu y  a 2  u y  a  b . ... (4.49)
y
Substituting the values from equations (4.47), (4.48) and (4.49) in equation
du  u x dx  u y dy  u z dz ,

we get

dx  2 1
du  b  a  b  dy  adz .
x y
Integrating we get

u  b log x   a 2  b  log y  ax  C .
This is the required complete integral of (4.46)

Example 6 : Solve the equation by Jacobi’s method

xu x  yu y  u z 2 .

Solution : Let

f  x, y, z , u x , u y , u z   xu x  yu y  u z 2  0 ... (4.50)

be a given p.d.e. Where from equation (4.50) we find

f u x  x, f u y  y , f u z  2u z , f x  u x , f y  u y , f z  0 ... (4.51)

Hence, the auxiliary equations (4.24) become

dx dy dz du du y du
   x   z . ... (4.52)
x y 2u z ux uy 0

From which we find

du z  0  u z  C , ... (4.53)

dx du
and   x  log x   log u x  log a  xu x  a
x ux

a
 ux  . ... (4.54)
x

104
Substituting these in (4.50) we get

 C2  a 
a  yu y  C 2  0  u y   . ... (4.55)
 y 
Substituting these values in
du  u x dx  u y dy  u z dz ,

we get

dx  2
du  a  C  a  dy  cdz  b .
x
Integrating we get

u  a log x   C 2  a  log y  Cz  b
This is the required complete integral of equation (4.50).
Example 7 : Solve

z 2u x 2 u y 2u z 2  u x 2 u y 2  u z 2  0 .

Solution : Let

f  x , y , z , u x , u y , u z   z 2u x 2 u y 2 u z 2  u x 2u y 2  u z 2  0 ... (4.56)

We first find two one-parameter family of p.d.e., which are compatible with (4.56).
From equation (4.56) we have

f u x  2u x z 2u y 2u z 2  2u x u y 2 , f u y  2u y z 2u x 2u z 2  2u x 2u y ,

f u z  2u z z 2u x 2u y 2  2u z , f x  0  f y , f z  2 zu x 2u y 2u z 2 .

Hence the the auxiliary equations (4.24) become

dx dy dz
2 2 2 2
 2 2 2 2
 2 2 2

2u x z u y u z  2u x u y 2u y z u x u z  2u x u y 2u z z u x u y  2u z

du x du y du z
   .
2 2 2
0 0 2 zu x u y u z

Therefore, the equation du x  0 and du y  0 give

u x  a and u y  b .
Substituting these values in equation (4.56) we get

105
 a 2 b 2 z 2 u z 2  a 2b 2  u z 2  0

ab
 u z 2 1  a 2b 2 z 2   a 2b 2  u z  .
1  a 2b 2 z 2
Hence, the equation
du  u x dx  u y dy  u z dz

reduces to

ab
du  adx  bdy  dz .
1  a 2b 2 z 2
Integrating we get

u  ax  by  sin 1  abz   C .
Example 8 : Solve

uz z ux  u y   x  y  0 .

Solution : Let

f  x, y , z , u x , u y , u z   u z z  u x  u y   x  y  0 . ... (4.57)

 f u x  zu z , f u y  zu z , f u z  z  u x  u y  , f x  1, f y  1, f z  u z  u x  u y  .

Hence, the auxiliary equations are

dx dy dz du du y du z
   x   . ... (4.58)
zu z zu z z  u x  u y  1 1 uz u x  u y 

Consider the ratios

dz du z dz du
   z .
z ux  u y  uz ux  u y  z uz

Integrating we get
log z   log u z  log a  zu z  a

a. ... (4.59)
uz 
z
Now consider the ratios
du x  du y .

Integrating we ger

106
 ux  u y  b . ... (4.60)
Now from (4.57) and (4.59) we have

x y
ux  u y  . ... (4.61)
a
Solving (4.60) and (4.61) we get

b x y
 ux   , ... (4.62)
2 2a

b x y
and uy    . ... (4.63)
2 2a

Substituting for u x , u y , u z in

du  u x dx  u y dy  u z dz

b x y b x y a
we get du     dx     dy  dz
 2 2a   2 2a  z

b 1 b 1 1 a
du  dx  xdx  dy  ydy   ydx  xdy   dz .
2 2a 2 2a 2a z

b 1 2 b y2 1 a
Integrating we get u x x  y  xy  2  C
2 4a 2 4a 2a z

b 1 1 a
i.e. u  x  y    x 2  y 2   xy  2  C .
2 4a 2a z
Jacobi’s Method to solve a non-linear p.d.e. in two variables :
Consider the following non-linear partial differential equation in the form

f  x, y , z , p , q   0 . ... (4.64)
The solution of (4.64) is a relation between x, y and z. Let this relation be

u  x, y , z   C ... (4.65)
Then we have from (4.65) on differentiating w.r.t. x and y

ux uy
p , q .
uz uz

On substituting these values in equation (4.64) we obtain a relation of the form

g  x, y , z , u x , u y , u z   0 . ... (4.66)

This can be solved by Jacobi’s method discussed earlier, which yields

107
 u  f  x , y , z , a , b, c  ... (4.67)
In this if we choose u = c, we get a complete integral of (4.64).

Example 9 : Find a complete integral of the equation

p2x  q2 y  z
by Jacobi’s method.
Solution : Let

f  x, y , z , p , q   z  p 2 x  q 2 y  0 ... (4.68)
be a given non-linear partial differential equation.
Let

u  x, y , z   C ... (4.69)
be the solution of equation (4.68). Then on differentiating (4.69) w.r.t. x and then w.r.t. y we get
respectively,

ux uy
p , q .
uz uz

Substituting this in equation (4.68) we get

2 2
u   uy 
x x   y   z .
 uz   uz 

or xu x 2  yu y 2  zu z 2 .

Let

f  x, y, z , u x , u y , u z   xu x 2  yu y 2  zu z 2  0 . ... (4.70)

The auxiliary equations are

dx dy dz du du y du
   x   z . ... (4.71)
fux fu y fuz fx fy fz

From equation (4.70) we find

f u x  2 xu x , f u y  2 yu y , f u z  2 zu z , f x  u x 2 , f y  u y 2 , f z  u z 2 .

Therefore, equations (4.71) become

108
 dx dy dz du du y du
    2x   2   2z . ... (4.72)
2 xu x 2 yu y 2 zu z ux uy uz

The two solutions of these equations are obtained by considering the ratios

dx du dx du
  2x   2 x  0 ,
2 xu x ux x ux

dy du y dy du y
and  2  2  0.
2 yu y uy y uy

Integrating we get

log x  2log u x  log a and log y  2 log u y  log b ,

i.e. xu x 2  a and yu y 2  b .

Let h1  x, y , z , u x , u y , u z , a   xu x 2  a  0 , ... (4.73)

and h2  x, y, z , u x , u y , u z , b   yu y 2  b  0 . ... (4.74)

Which are compatible with (4.70). Solving (4.73) and (4.74) we obtain

a b
ux 2  , uy2  .
x y
Substituting these values in the equation (4.70) we get

a b
x    y    zu z 2
 x  y

1
 uz 2  a  b . ... (4.75)
z
Consequently the equation
du  u x dx  u y dy  u z dz

reduces to

a b ab
du  dx  dy  dz .
x y z

Integrating we get
1 1 1
u  2  ax  2  2  by  2  2  ( a  b) z  2 C .

109
Writing u = C we get

   ax  2   by  2 
1 1 1
 (a  b) z  2

 ax 1 2 by  2 
1
1   
or z 2       
 a  b   a  b  

2
 ax 1 2 1 
    by  2 
or z     . ... (4.76)
 a  b   a  b  
Which is the complete integral of the equation (4.68).
Example 10 : Solve the p.d.e. by Jacobi’s Method

z 3  pqxy .
Solution : Let

f  x, y, z, p, q   z 3  pqxy  0 ... (4.77)

be a given non-linear partial differential equation.

Let u  x, y, z   C be its solution. ... (4.78)

Therefore, differentiating (4.78) w.r.t. x and y we get respectively

ux uy
p ,q   .
uz uz

Substituting these in equation (4.77) we get

f  x, y, z , u x , u y , u z   u z 2 z 3  u x u y xy  0 . ... (4.79)

 f u x   xyu y , f u y   xyu x , f u z  2u z z 3 , f x  u x u y y, f y  u x u y x, f z  2 z 3u z 2 . ... (4.80)

Therefore, the auxiliary equations (4.71) become

dx dy dz du x du y du

     2 z2 . ... (4.81)
 xyu y  xyu x 2u z z 3
 yu xu y u xu y x 3z u z

Consider the ratios

dx du x dx du
   x .
 xyu y u xu y y x ux

Integrating we get

log  xu x   log a

110
 xu x  a or u x  a . ... (4.82)
x
Now consider the ratios

dy du y dy du y
   .
 xyu x xu x u y y uy

Integrating we get

log  yu y   log b

 yu y  b or u y  b . ... (4.83)
y
Using these values in equation (4.83) we get

or ab . ... (4.84)
uz  3
z 2

Now substituting these values in the equation

du  u x d x  u y dy  u z dz

we get

dx dy dz
du  a  b  ab 3 .
x y z 2
Integrating we get

 2 
u  a log x  b log y  ab   C .
 z

Taking u  x, y, z   C we get

 log  x a  y b  .
2 ab
z

 ab 
x a y b  exp  2 . ... (4.85)
 z 
Which is the required complete integral.
Example 11 : Solve by Jacobi method the equation
pq  xz .

111
Solution : Let

f  x, y, z , p, q   pq  xz  0 , ... (4.86)
be a given non-linear partial differential equations.

Let u  x, y, z   C be its solution. ... (3.87)

Therefore, differentiating (4.87) w.r.t. x and y we get respectively

u x and uy
p q .
uz uz

Substituting these values in (4.86) we get

f  x, y, z , u x , u y , u z   u x u y  xzu z 2  0 . ... (4.88)

From equation (4.88) we find

f u x  u y , f u y  u x , f u z  2 xzu z , f x  zu z 2 , f y  0, f z   xu z 2 . ... (4.89)

Hence the Jacobi’s auxiliary equations (4.71) reduce to

dx dy dz du x du y du z
     .
u y u x 2 xzu z zu z 2
0  xu z 2

Now the equation

du y  0  u y  a ... (4.90)
The ratios

dz du z dz du z
   ,
2 xzu z  xu z 2
2 z u z

 log z  2 log u z  log b ,

 zu z 2  b . ... (4.91)
Using (4.90) and (4.91) in (4.88) we get

b
au x  xb  0  u x    x . ... (4.92)
a
Substituting these values in the equation
du  u x d x  u y dy  u z dz
we get
b 1
du    xdx  ady  b dz .
a 2
112
Integrtaing we get
2
bx
u     ay  2 b z  C .
a 2

b 2
u  x  ay  2 bz  C .
2a
Which is the required complete integral of (4.88). Writing u = C we get
2
1  b 2 
z  x  ay 
4b  2a 
as a complete integral of equation (4.86).

Exercise :
1. Solve the partial differential equation by Jacobi method.

 p 2  q 2  y  qz .
2. Solve by Jacobi method.

p   z  qy  .
2

113
 UNIT - IV

THE CAUCHY PROBLEM

Introduction :
Given a partial differential equation and a curve in space, the Cauchy problem is to find an
integral surface of the equation which contains the given curve.
Let a partial differentiable equation and a curve be given by

f  x, y , z , p , q   0 , ... (1.1)

x  x0 ( s ), y  y0 ( s ), z  z0 ( s) . s   a, b  ... (1.2)
Then the Cauchy problem is to find a solution

z  z  x, y 
of the partial differential equation (1.1) such that

z0 (s)  z  x0 ( s), y0 ( s)  . s   a, b  .
In the unit 4, we find the integral surfaces through a given curve for a
1) linear partial differential equations,
2) non-linear partial differential equations,
3) and quasi-linear equations.

1. Integral Surfaces through a given curve for a Linear Partial Differential Equations.
Result : Discuss how a general solution may be used to determine the integral surface, which passes
through a given curve.
Proof : Consider a linear partial differential equation in the form
Pp  Qq  R . ... (1.3)
The general solution of the equation (1.3) is given by

F  u, v   0 , ... (1.4)

where F is an arbitrary function and u  x, y, z   C1 , v  x, y, z   C 2 are solutions of the equation

dx dy dz
  . ... (1.5)
P Q R

114
This solution is a two parameter family of curves.
Let C be a given curve whose parametric equations are given by
x  x0 ( s ), y  y0 ( s ), z  z0 ( s) , ... (1.6)
where s is a parameter (not necessarily the arc length) of the curve. Our aim is to find F such that the
integral surface F  u, v   0 contains the given curve C.
To obtain the integral surface containing the curve C, let us assume that, we can drive from
equation (1.5) two relations of the form

u  x, y , z   C1 and v  x, y, z   C 2 ... (1.7)

involving two arbitrary constants C1 and C2. Substituting x  x0 ( s ), y  y0 ( s ), z  z0 ( s) in these

equations, we get

u  x0 ( s), y0 (s), z0 ( s )   C1 ,

and v  x0 ( s), y0 ( s), z0 ( s)   C2 . ... (1.8)

From this particular solutions, we can eliminate the parameter s to obtain the relation between C1 and
C2 of the type

F  C1 , C2   0 . ... (1.9)

Then the required integral surface z  z  x, y  is obtained by eliminating C1 and C2 between equations
(1.8) and (1.9).
Note : Sometimes the solution can also be obtained by assuming v  G  u  and determining G..

Example 1 : Find the integral surface of the p.d.e.

 x  y  y 2 p   y  x  x 2q   x 2  y 2  z
through the curve xz  a 2 , y  0 .
Solution : Given p.d.e. is

 x  y  y 2 p   y  x  x 2q   x 2  y 2  z ... (1.10)

The integral surface of the equation (1.10) is generated by the integral curves of the auxiliary equations

dx dy dz . ... (1.11)
  2
 x  y y  y  x x  x  y  z
2 2 2

Consider the ratios

dx dy
  x 2 dx  y 2 dy  0
 x  y y  y  x x
2 2

115
Integrating we get

x 3  y 3  C1 .

Let u  x, y, z   x 3  y 3  C1 . ... (1.12)

Now consider the ratios

dx  dy dx dx  dy dz .
  
 x  y x 2
y 2
 x 2
y 2
z x y z

Integrating we get

log  x  y   log z  log C 2 ,

i.e. x  y  zC2 ,

x y
or  C2 .
z

x y
Let v  x, y , z    C2 . ... (1.13)
z
It is given that the general surface represented by (1.12) and (1.13) passes through the curve

xz  a 2 , y  0 .
The parametric representation of these equations are

a
x  as , y = 0 and z  . ... (1.14)
s
Substituting this in (1.12) and (1.13) we get

a 3 s 3  C1 ,

 as  0  2
and   s  C2  s  C 2 .
 a 

 a 6C2 3  C12 . ... (1.15)

From equations (1.12), (1.13) and (1.15) we have
3
6 x y
x 3
y 
3 2
a   ,
 z 

 z 3  x3  y 3   a 6  x  y  .
2 3
... (1.16)

This is the required integral surface of (1.10).

116
Example 2 : Find the integral surface of the differential equation

x  z  2  p   xz  2 yz  2 y  q  z  z  1
passing through the curve
x0  s, y0  0 and z 0  2 s .
Solution : Given p.d.e. is

x  z  2  p   xz  2 yz  2 y  q  z  z  1 . ... (1.17)
The integral surface of the equation (1.17) is generated by the integral curves of the auxiliary equations

dx dy dz . ... (1.18)
 

x z2   xz  2 yz  2 y  z z  1


Each ratio of (1.18) is equal to

dx  dy dx  dy ,

xz  2 x  xz  2 yz  2 y 2 xz  2 yz  2  x  y 

dx  dy .

2 x  y z 1
 

Therefore, consider the ratios

dx  dy dz
 2 .
x y z

Integrating we get log( x  y )  2 log z  log C1 ,

 x y
  2   C1 . ... (1.19)
 z 

x y
Let u  x, y , z    C1 . ... (1.20)
z2
Now consider the ratios

dx dz

x  z  2  z  z  1

dx  z  2  dz
 
x z  z  1

117
 dx  2 1 
    dz .
x  z z 1
Integrating we get

log x  2 log z  log  z  1  log C 2

x  z  1
  C2 . ... (1.21)
z2

x  z  1
Let v  x, y , z    C2 .
z2
It is given that the general surface represented by (1.19) and (1.21) passes through the curve
x0  s, y0  0, z 0  2 s .
Hence equations (1.19) and (1.21) become

s 1
2
 C1   C1
4s 4s

1
and  2 s  1  C2 .
4s

C2 C
 C1  2 s  1  C2  2 s  1   2s  2  1
C1 C1

 1  C2  C1 1
2    C2  C1  ... (1.22)
 4C1  C1 2

x  z  1 x  y 1
  2 
z2 z 2

1 1
or 2
 zx  y  
z 2

or 2  xz  y   z 2 .
Example 3 : Find the integral surface of

x 2 p  y 2q  z 2  0
which passes through the parabola
xy  x  y , z  1 .
Solution : Given p.d.e. is

x 2 p  y 2q   z 2 . ... (1.23)
118
The integral surface of the equation (1.23) is generated by the integral curves of the auxiliary equation

dx dy dz
2
 2  2. ... (1.24)
x y z
Consider the ratios

dx dy .

x2 y2
Integrating we get

1 1
  C1 ... (1.25)
x y
Similarly, by considering the ratios

dy  dz
 2 ,
y2 z
we obtain

1 1
  C2 . ... (1.26)
y z
Given that the general surface represented by (1.25) and (1.26) passes through the curve
xy  x  y , z  1 ,
whose parametric equations are

x
y  x  1  x  y 
x 1

s
x  s, y  , z  1. ... (1.27)
s 1
Substituting these in (1.25) and (1.26) we get

1 s 1 2s 2
  C1   C1  s 
s s s 1  C1

4
1
s 1 2s  1 1  C1
 1  C2   C2   C2
s s 1
1  C1

4  1  C1
i.e.  C2 or 2C2  C1  3 . ... (1.28)
2

119
Using equations (1.25) and (1.26) in the equation (1.28) we get

 1 1 1 1
2      3
 y z x y

1 1 2
or    3. ... (1.29)
y x z
This is the required integral surface.
Example 4 : Find the equation of the integral surface of the equation

x 3 p  y  3x 2  y  q  z  2 x 2  y  ... (1.30)

which passes through the curve x0  1 , y0  s, z 0  s 1  s  .

Solution : The integral surface of the equation (1.30) is generated by the integral curves of the auxiliary
equations

dx dy dz . ... (1.31)
 
y  3x  y  z  2 x 2  y 
3 2
x

1 1 1
 dx  dy  dz
Each ratio of the equation  x y z
 x  3x  y  2 x 2  y
2 2

1 1 1
 dx  dy  dz
x y z

0

dx dy dz
   0.
x y z
Integrating we get
 log x  log y  log z  log C1

y
  C1 .
xz

y
Let u  C1 . ... (1.32)
xz
Now consider the ratios

dx dy

y  3x 2  y 
3
x

120
 
 3x 2  y  dx  dy   3x 2  y  dx  dy
... (1.33)
x3 y x3  y
Each ratio of (1.33)


 3x 2  y  dx  dy  xdy .
x 3  y  xy
Consider

dy  3x 2  y  dx  dy  xdy

y x 3  y  xy

3x 2 dx  dy   ydx  xdy 

x 3  y  xy

dy 3x 2 dx  dy  d  xy 

y x 3  y  xy
Integrating we get

log y  log  x 3  y  xy   log C2

x 3  y  xy
or  C2 .
y

x 3  y  xy
Let v  C2 . ... (1.34)
y
Given that the general surface (1.32) and (1.34) passes through the curve

x0  1, y0  s, z0  s 1  s  .

s
  C1
s 1  s 

1  C1
1  C1 1  s   s 
C1

1 s  s
and  C2  1  s 2  C2 s
s

or 1   C2  2  s

121
 1 1 C 1
s  1 s  1  2
C2  2 C2  2 C2  2

Substituting in 1  C1 1  s  we get

 C 1 
1  C1  2 
 C2  2 

 C2  2  C1C2  C1

or C1C2  C1  C2  2  0 . ... (1.35)

From equations (1.32), (1.34) and (1.35) we obtain

x 3  xy 2 y  x 3  y  xy
   0.
xz y

x 2  y y  x 3  xy
  0
z y

  x 2  y  y  z  y  x 3  xy   0

or  x 2  y  y  xz  x 2  y   yz  0
  x 2  y   y  xz   yz  0

or  x 2  y   xz  y   yz  0
or yz   x 2  y   xz  y  . ... (1.36)

This is the required integral surface.

Example 4 : Find the equation of the integral surface of the differential equation
2 y ( z  3) p  (2 x  z ) q  y (2 x  3)
which passes through the circle.

z  0, x 2  y 2  2 x .
Solution : Given p.d.e. is

2 y  z  3 p   2 x  z  q  y  2 x  3 . ... (1.37)
The integral surface of the given equation (1.37) is generated by the integral curves of the auxiliary
equations

dx dy dz
  ... (1.38)
2 y ( z  3) 2 x  z y (2 x  3)
122
 ydy  dz
Each ratio 
2 yx  yz  2 xy  3 y

ydy  dz
 .
 y ( z  3)
Therefore, consider the ratios

dx ydy  dz ,

2 y ( z  3)  y  z  3

 dx  2  ydy  dz  .
Integrating we get

x   y 2  2 z  C1 .

or x  y 2  2 z  C1 .

Let u  x  y 2  2 z  C1 . ... (1.39)

Now consider the ratios

dx dz dx dz
  
2 y  z  3 y  2 x  3 2  z  3 2 x  3

  2 x  3 dx  2  z  3 dz
Integrating we get

x 2  3 x  z 2  6 z  C2

or x 2  z 2  3 x  6 z  C2 .

Let v  x 2  z 2  3 x  6 z  C2 . ... (1.40)

Given that the general surface represented by (1.39) and (1.40) passes through the circle

z  0, x 2  y 2  2 x ,

or z  0, x 2  2 x  y 2  0 ,
2
i.e. z  0,  x  1  y 2  1 .
The parametric representations of these equations are
x  1  cos  , y  sin  , z  0 .
Therefore, substituting these in (1.39) and (1.40) we get

123
 1  cos   sin 2   C1 . ... (1.42)

and 1  cos   2  3 1  cos    C2

or  cos 2   cos   2  C2 . ... (1.43)

Adding equations (1.42) and (1.43) we get

C1  C2  0 ... (1.44)
Thus eliminating C1, C2 between (1.39), (1.40) and (1.44) we get

x  y 2  2 z  x 2  z 2  3x  6 z  0

or x2  y 2  z 2  2x  4z  0 . ... (1.45)
This is the required integral surface.
Example 5 : Find the integral surface of the linear partial differential equation

x  y2  z  p  y  x2  z  q   x2  y 2  z
which contains the straight line
x  y  0, z  1 .
Solution : The linear partial differential equation is given by

x  y2  z  p  y  x2  z  q   x2  y 2  z . ... (1.46)

The integral surface of the equation (1.46) is generated by the integral curves of the auxiliary equations

dx dy dz . ... (1.47)
 
x y 2
 z y  x 2
 z x 2
y 2
z
Each ratio of (1.47)

yzdx  xzdy  xydz



xyz y 2  z  x 2  z  x 2  y 2 
 d  xyz   0 .
Integrating we get
xyz  C1 .
Let u  xyz  C1 . ... (1.48)
Now each ratio of the equation (1.47)

xdx  ydy  dz

x 2
y 2
 z   y2  x2  z   z  x2  y2 

124
 xdx  ydy  dz

0
 xdx  ydy  dz  0 .
Integrating we get

or x 2  y 2  2 z  C2 .

Let v  x 2  y 2  2 z  C2 . ... (1.49)

It is given that the general surface given in (1.48) and (1.49) contains the straight line
x  y  0, z  1 .
The parametric equations of the straight line are
x  t , y  t , z  1 .
Therefore, substituting these in equations (1.48) and (1.49) we get

t 2  C1 ,

2  t 2  1  C2 .
Eliminating the parameter t, we get
2C1  C2  2  0 . ... (1.50)
Thus the required integral surface is obtained by eliminating C1 and C2 from (1.50). Elimination gives

2 xyz  x 2  y 2  2 z  2  0 ... (1.51)

Example 6 : Find the integral surface of the equation

 x  a  p  2 yq  2 z
passing through the initial data curve
x0  1, y0  s, z0  s .
Solution : The linear p.d.e. is given by
 x  2  p  2 yq  2 z . ... (1.52)
The integral surface of the equation (1.52) is generated by the integral curves of the auxiliary equations

dx dy dz . ... (1.53)
 
x  2 2 y 2z
Consider the ratios

dx dy .

x  2 2y
125
Integrating we get

1
log( x  2)  log y  log C1
2
2
  x  2   yC1 . ... (1.54)

 x  2 2
Let u  C1 . ... (1.55)
y
Now consider the ratio

dy dz

2 y 2z
Integrating we get
log y  log z  log C2

or y  zC 2 .

y
Let v  C2 . ... (1.56)
z
It is given that the general surfaces (1.55) and (1.56) passes through the initial data curve given by

x0  1, y0  s and z0  s . ... (1.57)

Therefore, substituting in (1.55) and (1.56) we get

1
 C1 and s  C2
s

1 . ... (1.58)
 C2 2 
C1

Eliminating C1 and C2 from (1.58) we obtain

2
 y y
  
z  x  2 2

z  y  x  2 . ... (1.59)
This is the required integral surface.
Example 7 : Find the equation of the integral surface of the equation

 x 2  y 2  p  2 xyq   x  y  z ... (1.60)

which passes through the curve

126
 x0  0, y0  s 2 , z0  s
Solution : The general solution of equation (1.60) is obtained in the example (4) of unit 2 in the form.

x y
 C1 , ... (1.61)
z
and

y 2  x2
 C2 . ... (1.62)
y
This surface passes through the given curve

x0  0, y0  s 2 , z0   s . ... (1.63)
Substituting these values in equations (1.61) and (1.62) we get
s  C1 ,

s 2  C2 ,

 C2  C12 .
Consequently, from equations (1.61) and (1.62) we obtain

z2  y2  x2   y  x  y .
2
... (1.64)

Given that the general surface (1.62) and (1.63) passes through the curve

x0  0, y0  s 2 , z0   s .
Substituting these in equations (1.62) and (1.63) we get

C1   s, C2  s 2 ,

 C2  C12 .
Eliminating C1 and C2 we get

2
x2  x  y 
 y   ,
y  z 

z2  y2  x2   y  x  y .
2
... (1.64)

Example 8 : Find integral surface of

2x  y  z 2  p  y  2 y  z 2  q  z3
which passes through the curve

127
 x0  s 2 , y0  s, z0  1 .
Solution : We have obtained the general solution of the equation in the form (Refer example (2) of
Unit 2)

x
u  C1 , ... (1.65)
yz

z2  2y
and v  C2 . ... (1.66)
yz
Given that the general surface (1.65) and (1.66) passes through the curve

x0  s 2 , y0  s, z0  1 .
Substituting in (1.65) and (1.66) we get

1  2s
s  C1 and  C2
s

 C1C2  2C1  1 .
Using equations (1.65) and (1.66) we deminate C1 and C2 to get

yz  2 x 
x 2
 z  2y
yz

  yz  2 x  yz  x  z 2  2 y  . ... (1.67)

Example 9 : Find the integral surface passing through the circle z = 1, x2 + y2 = 1 of the partial
differential equation

x  y p   y  x  zq  z .
Solution : Let the linear partial differential equation be given by

x  y p   y  x  zq  z . ... (1.68)

The auxiliary equations are

dx dy dz
  . ... (1.69)
x y yxz z
Each ratio of (1.69)

dx  dy dz
 
x y yxz z

 dx  dy   dz .
Integrating we get
128
 x  y  z  C1 .

Let u  x  y  z  C1 . ... (1.70)

Now each ratio of equation (1.69)

dx  dy  dz
 ,
x y y x z z

dx  dy  dz .

2 x  y  z

Consider the ratios

dz dx  dy  dz .

z 2x  y  z

Integrating we get

2 log z  log  x  y  z   log C 2 ,

or x  y  z  C2 z 2 . ... (1.71)
It is given that the general surface represented by (1.70) and (1.71) passes through the curve (circle).

z  1, x 2  y 2  1 ,
whose parametric equations are
x  cos t , y  sin t , z  1 . ... (1.72)
Substituting this in (1.70) and (1.71) we get
cos t  sin t  1  C1 and cos t  sin t  1  C2

C1  C2  2 and C  C2
 cos t  sin t  1 .
2 2

1
Hence cos 2 t  sin 2 t  1    C1  C2  2  2   C1  C2  2   1
4

 C12  C2 2  2  C1  C2   0 . ... (1.73)

Substituting the values of C1 and C2 from equations (1.70) and (1.71) we get

2
x yz x yz
 x  y  z  2   2

  2 x  y  z    0,
 z   z2 

 z 4  x  y  z    x  y  z   2 z 2  x  y  z  z 2  x  y  z   0 ,
2 2

129
 z 4  x  y  z    x  y  z   2 z 2  x  y  z   z 2  x  y  z    0 .
2 2
i.e. ... (1.74)

This is the required integral surface (particular solution) through the given circle.
Example 10 : Find the integral surface of the linear partial differential equation
xp  yq  z
which contains the circle

x 2  y 2  z 2  4, x  y  z  2 .
Solution : The given partial differential equation is
xp  yq  z . ... (1.75)
The integral surface of the equation (1.75) is generated by the integral curves of the auxiliary equations

dx dy dz
  .
x y z
Consider the ratios

dx dy
 ,
x y
and integrating we get
log x  log y  log C1

x
  C1 . ... (1.76)
y
Similarly, by considering the last two ratios we get

y
 C2 . ... (1.77)
z
Thus the integral surface of the equation (1.75) is

 x y
F  ,   0. ... (1.78)
y z
It is given that this integral surface passes through the given curve

x2  y2  z2  4 , ... (1.79)

x y z  2. ... (1.80)
From equations (1.76) and (1.77) we find

x and x .
y z
C1 C1C2

130
Substituting this in equations (1.79) and (1.80) we get

 1 1 
x 2 1  2  2 2   4 , ... (1.81)
 C1 C1 C2 

 1 1 
and x 1   2 . ... (1.82)
 C1 C C
1 2

From equations (1.81) and (1.82) we find

2
1 1  1 1 
1   1  
C12 C12C2 2  C1 C1C2 


1 1 1
   2 0
C1 C1C2 C1 C2

 C1C2  C1  1  0 .

x y
Now replacing C1  and C 2  we get
y z
xy  xz  yz  0 . ... (1.83)
This is the required integral surface of the given partial differential equation.

Exercise :
1. Find the integral surface of the equation

 2 xy  1 p   z  2 x 2  q  2  x  yz 
which passes through the line x0 ( s )  1, y0 ( s )  0 and z0 ( s )  0 .

2. Integral surfaces through a given curve for a non-linear Partial

Differential Equations
Result : Discuss the method of finding the integral surface of a non-linear partial differential equation.

Proof : Let f  x, y , z , p , q   0 ... (2.1)

be a given non-linear partial differential equation. By usual (Charpit’s) method we find its complete
integral. Let

F  x, y, z , a, b   0 ... (2.2)
be a complete integral of equation (2.1), which involves two arbitrary constants ‘a’ and ‘b’.

131
Let C be a given curve whose parametric equations are given by
x  x0 ( s ), y  y0 ( s ), z  z0 ( s) , ... (2.3)
where s is aparameter of the curve. Our aim is to find the integral surface of the given partial differential
equation (2.1) which contains the given curve (2.3).
We expect that, this solution to be an envelope of one parameter subfamily of (2.2). This
envelope contains the curve C. This requires that

F  x0 (s ), y0 (s ), z0 ( s), a, b   0 . ... (2.4)

Differentiating this with respect to s we get

F
 x0 (s), y0 ( s ), z0 ( s), a, b   0 . ... (2.5)
s
Thus we have two relations (2.4) and (2.5) from which we eliminate s to obtain the relation between
‘a’ and ‘b’ such as

  a, b   0 . ... (2.6)
Factorizing this we get

b  1  a  , b   2 ( a ) ... (2.7)
Each one of the relations (2.7) defines a one parameter subfamily of the complete integral
(2.2). The envelope of each of these subfamilies if it exists, is an integral surface of the equation (2.1).
Note : The solution may not be unique.

Example 1 : Find a complete integral of the equation

 p 2  q 2  x  pz
and the integral surface which passes through the curve

C : x0  0, y0  s 2 , z0  2s .
Solution : Let

f  x, y, z , p, q    p 2  q 2  x  pz  0 . ... (2.8)

be a given non-linear p.d.e. To find its complete integral, we know the Charpit’s auxiliary equations are

dx dy dz dp dq . ... (2.9)
   
f p f q pf p  qf q f x  pf z f y  qf z

Where from equation (2.8) we find

f p  2 px  z , f q  2qx, f x  p 2  q 2 , f y  0, f z   p .

132
Consequently, equation (2.9) becomes

dx dy dz dp dq . ... (2.10)
   2 
2 px  z 2qx pz q pq

Considering the ratios

dp dq
 
q 2 pq

 pdp  qdq  0 .
Integrating we get

p2 q2
  Constant
2 2

or p2  q2  a2 . ... (2.11)
Substituting this in (2.8) we get

2 a2x
a x  pz  p  .
z
The equation (2.11) gives
2
2  a2 x 
2
q  a  
 z 

a 2
q z  a2 x2 . ... (2.12)
z
Substituting these in equation
dz  pdx  qdy
we get

a2x a 2
dz  dx  z  a 2 x 2 dy
z z

 zdz  a 2 xdx  a z 2  a 2 x 2 dy

 ady  d  z 2  a 2 x 2   ady .
zdz  a 2 xdx

2 2 2
z a x
Integrating we get

z 2  a 2 x 2  ay  b

133
 z 2  a 2 x 2   ay  b 
2
or

 z 2  a 2 x 2   ay  b  .
2
... (2.13)

This is the required complete integral. Given that this complete integral passes through the curve

C : x0  0, y0  s 2 , z0  2s .

 4s 2   as 2  b  .
2
... (2.14)

Differentiating (2.14) w.r.t. s we get

8s  2  as 2  b   2 sa

 2  a  as 2  b  . ... (2.15)
Eliminating s between (2.14) and (2.15) we get

4 1
4s 2  2
 s2  2
a a

 1 
 2   a 2  2  ba 
 a 

1
b ... (2.16)
a
Substituting this in (2.13) we obtain one-parameter subfamily of the complete integral in the form
2
2 
2 2 1
z  a x   ay  
 a

a 2 z 2  a 4 x 2   a 2 y  1
2

 a 2 z 2  a 4  x 2  y 2   2a 2 y  1

 a4  x2  y 2   a2 2 y  z2  1  0 . ... (2.17)

To find the envelope of (2.17) differentiate (2.17) w.r.t. ‘a’ we get

2a 2  x 2  y 2    2 y  z 2   0 . ... (2.18)

Eliminating ‘a’ between (2.17) and (2.18) we get the required envelope of one-parameter subfamily as

2 y  z 2   x2  y 2   2 y  z 2  2 y  z 2   1  0
2

4  x2  y2  2  x2  y2 
2

134
 2 y  z2   2 2 y  z2   4  x2  y2   0 ,
2 2
or

 z2  2y  4  x2  y 2  .
2
or


 z2  2 y  x2  y2 , 

 z 2  2 y  x2  y2 .  ... (2.19)

(by discarding the negative sign as x2  y2  y )

This is the required integral surface of (2.8).
Example 2 : Find a complete integral of the equation

 p 2  q 2  x  pz
and the integral surface passing through the parabola x  0, z 2  4 y .
Solution : The complete integrate of the p.d.e

 p 2  q 2  x  pz ... (2.20)

is given by (refer earlier example)

z 2  a 2 x 2   ay  b  .
2
... (2.21)

Given that this passes through the parabola C

x  0, z 2  4 y ,
whose parametric equations are

x  0, y  t , z 2  4t
2
 4t   at  b 

 a 2t 2   2ab  4  t  b 2  0
For real roots we must have

b 2  4aC  0
2
  2ab  4   4a 2b 2  0

or  ab  2  2  a 2b 2  a 2b 2  4ab  4  a 2b 2

 ab  1 ... (2.22)

135
 1
or b .
a
Substituting this in (2.21) we obtain the equation of the required integral surface in the form


z2  2 y  x2  y2  or 2 y  z2 
2
 4  x2  y2  .

Exercise 3 : Find the integral surface of z  p 2  q 2 which passes through the curve

4 z  x 2  0, y  0 .
Solution : Let

f  x, y, z, p, q   p 2  q 2  z  0 ... (2.23)

be the given non-linear partial differential equation. To find the complete integral we have from Charpit’s
auxiliary equations

dx dy dz dp  dq .
   
2 p 2q 2 p  2q
2 2
0  p q

Consider the ratios

dx dp
  dx  2dp
2p p
Integrating we get

1
x  2 p  a or p   x  a . ... (2.24)
2
Substituting this in (2.23) we get
2
 xa
q2  p2  z  q2    z
 2 

 x  a 2  4 z 1
or q q  x  a 2  4 z . ... (2.25)
2 2
Substituting these values in the equation
dz  pdx  qdy
we get

 xa 1  2
dz    dx  x  a   4 z  dy
 2  2

136
  xa 1  2
 dz    dx  x  a   4 z  dy
 2  2

2dz   x  a  dx
  dy
 x  a 2  4 z

  d   x  a   4 z   dy
2

Integrating we get

 x  a 2  4z  y  b

or  x  a 2  4 z   y  b 2 . ... (2.26)

Thisis the required complete integral. It is given that this integral passes through the curve

s2
x0  s, y0  0, z 0   .
4
Hence equation (2.26) gives

 s  a 2  s 2  b 2 ... (2.27)
Differentiating (2.27) w.r.t. s we get

2  s  a   2s  0
 a  2s . ... (2.28)
Eliminating s between (2.27) and (2.28), we obtain

a  2b .
Substituting this in (2.26) we get

x  2b   4 z   2 s  b  ,
2 2 ... (2.29)

which is the one parameter of subfamily of complete integral.

Differentiating (2.29) w.r.t. b we get

2  x  2b    2   2  y  b  ,

 y b 
  2. ... (2.30)
 x  2b 
Eliminating b between (2.29) and (2.30) we get

137
 4 z
2,
x  2b 
2

 2 z   x  2 y  .
2
... (2.31)

Example 4 : Find the complete integral of the equation

p 2 x  qy  z  0
and derive the equation of the integral surface containing the line y = 1 and x + z = 0.
Solution : Let

f  x, y, z , p, q   p 2 x  qy  z  0 ... (2.32)
be a given non-linear p.d.e.
To find its complete integral, the Charpit’s auxiliary equations give

dx dy dz dp dq . ... (2.33)
   
2
2 px y 2 p x  qy  p  p  1 0

From which we obtain

dq  0 or q  a ... (2.34)
Using (2.34) in (2.32) we get
1
2  z  ay  2
xp  z  ay  p    . ... (2.35)
 x 
Substituting (2.34) and (2.35) in the equation
dz  pdx  qdy
we get
1
 z  ay  2
dz    dx  ady
 x 
dz  ady dx .
 
z  ay x

On integrating we get

z  ay  2  b .
Squaring we get

z  ay  x  b  2 xb

138
or z  ay  x  b  2 xb .
Squaring we get

 ay  z  x  b  2  4 xb . ... (2.36)

Which is the required complete integral.

Given that this complete integral passes through the curve

C : y  1, x  z  0 .

i.e. y  1, x  t , z  t .
On substituting this in (2.36) we get

 a  t  t  b  2  4bt

 a  b  2t  2  4bt . ... (2.37)

Differentiating (2.37) w.r.t. t we get

4  a  b  2t   4b

 a  b  2t  b

 a  2t and b  4t
 b  2a
or 2a  b  0 . ... (2.38)
On substituting b = – 2a in equation (2.36) we get one-parameter subfamily of the complete integral of
p.d.e. (2.32) in the form

 ay  z  x  2a  2  4 x  2a 

 ay  z  x  2a  2  8ax . ... (2.39)

Differentiate equation (2.39) w.r.t. a we get

2  ay  z  x  2 a  y  2   8 x

or  y  2  ay  z  x  2a   4 x . ... (2.40)
The envelope is obtained by eliminating a between equations (2.39) and (2.40) we get

xy  z  y  2  . ... (2.41)
Which is the required integral surface of equation (2.32).

139
Exercise :
1. Find the integral surface of

x 2 p  y 2q  z 2  0
which passes through the hyperbola
xy  x  y , z  1 .

3. Integral Surfaces through a given curve by a method of Characteristics

In this section we shall discuss the method of characteristics to find the integral surface of a
semi-linear and quasi-linear partial differential equations.
(a) Semi-linear Partial Differential Equations :
Consider a semi-linear partial differential equation given by

P  x, y  p  Q  x, y  q  R  x, y , z  . ... (3.1)
The expression on the left hand side of the equation (3.1) is called the directional derivative of z (x, y)
in the direction of  P  x, y  , Q  x, y   at the point (x, y).

The one parameter family of curves in the xy plane is characterized by the ordinary differential equation

dy Q  x, y  . ... (3.2)

dx P  x, y 

Or the system of ordinary differential equations

dx dy
 P  x , y  ,  Q  x, y  . ... (3.3)
dt dt
These curves have the property that along them z (x, y) will satisfy the ordinary differential equation

dz dy
 zx  z y
dx dx

Q  x, y 
 zx  zy
P  x, y 

dz z x P  x, y   z y Q  x, y 
  ... (3.4)
dx P  x, y 

dz dx dy
or  zx  zy
dt dt dt

dz
  z x P  x, y   z y Q  x, y  ... (3.5)
dt
140
 dz
  P  x, y  p  Q  x, y  q ,
dt

where p  z z and q  z y

dz
  R  x, y , z  . ... (3.6)
dt
The one parameter family of curves defined by equation (3.3) are called the characteristics curves of
the partial differential equation (3.1) and the equation is called characteristic equation.

Let  x0 , y0  be a point in the xy plane. By the existence and uniqueness of the solution of the
initial value problem for the ordinary differential equation, (3.3) will define a unique characteristic curve
(say)

x  t   x  x0 , y0 , t  , y  t   y  x0 , y0 , t  , ... (3.7)

such that x(0)  x0 and y (0)  y0 .

If z 0 is the value for z  x, y  at  x0 , y0  then the equation (3.7) determines a unique solution z as

z  z  x0 , y0 , t  .

Thus z (x, y) is uniquely determined along the characteristic passing through the point  x0 , y0  if we

know z (x, y) at  x0 , y0  .
Example 1 : Solve the equation
xz y  yz x  z

with the initial condition

z  x,0   f  x  , x  0 .
Solution : Let the equation be
xz y  yz x  z . ... (3.8)
We know the characteristic curves of (3.8) are given by the equation

dy Q  x, y 
 ... (3.9)
dx P  x, y 

where

P  x, y    y , Q  x, y   x , R  x , y , z   z .
Hence equation (3.9) gives

141
 dy x
 . ... (3.10)
dx y
This equation has solution given by

x2  y2  C 2 . ... (3.11)
These curves have the property that along them the function z (x, y) satisfies the ordinary differential
equation

dz dy
 zx  z y
dx dx

 x
 zx  z y    by equation (3.10)
 y

yz x  xz y

y

dz z
 . by equation (3.8)
dx y

dz z
i.e.  by equation (3.11)
dx c2  x2

dz dx .
 
z 2
c x 2

Integrating we get

dx
log z     log k
c2  x2

x
log z   sin 1    log k
c

x
 sin 1 
or z k (c )e c , ... (3.12)

where the constant of integration k may depend on c.

Therefore we write the general solution of the equation as
x
 sin 1 
z  kx  y2 2
 e c. ... (3.13)

Now applying the initial conditions, we get

142
 z  x, 0   f  x  , x  0
we have

 x  x 
sin 1    sin 1  
c  x2  y2
 

x 
 sin 1    sin 1 1 
x 2


 z  x, 0   f  x   k  x 2  e 2


 k  x2   f  x  e 2


or k  x  f  x e 2 ... (3.14)

Hence the general solution (3.13) becomes

 sin 1 x 
z  x, y   f  e c. ... (3.15)
2 2 2
x y

Example 2 : Solve the equations

zx  z y  z 2

with the initial condition z  x, 0   f  x  .

Solution : We are given that

zx  z y  z 2 . ... (3.16)

We know the characteristic curves of the p.d.e (3.16) are given by the equation

dy Q  x, y  , ... (3.17)

dx P  x, y 

where

Q  x, y   1, P  x, y   1 . ... (3.18)

dy
  1.
dx
whose solution is
x y C . ... (3.19)
These one parameter family of curves have the property that along them the function z (x, y) must
satisfy the ordinary differential equation

143
 dz dy
 zx  z y
dx dx

 z x  z y 1 by equation (3.18)

dz
 z2 . by equation (3.16)
dx

dz
  dx
z2
Integrating we get

1 1 , .... (3.20)
  xk  z 
z x  k c
where the constant of integration k may be a function of C.

1
 z  x, y    . ... (3.21)
x  k  x  y

Now applying the initial condition

z  x, 0   f  x 

1
we get  z  x, 0   f  x   
x  k  x

 f  x   x  k  x    1 ,

xf  x   f  x  k  x   1

 f  x   k  x   1  xf  x 

 k  x  
1  xf  x  
f  x

1   x  y  f  x  y  
 k  x  y    .
f x  y

Substituting this in (3.21) we get

1
z  x, y  
1   x  y  f  x  y  
x
f  x  y

Simplifying we get

144
 f  x  y .
 z  x, y  
1  yf  x  y 

Which is the required solution.

(b) Quasi-linear Equation :

Consider a quasi-linear p.d.e. given by

P  x, y, z  z x  Q  x, y, z  z y  R  x, yz  . ... (3.22)

We know its solution defines an integral surface z  z  x, y  in the x, y, z space.

We know the direction ratios of the normal to this surface are given by  z x , z y , 1 .

Hence equation (3.22) states that the integral surface is such that at each point the line with
direction ratios (P, Q, R) is tangent to the surface at that point. (Infact, any surface z = z (x, y) has the
property that it is an integral surface iff the tangent plane contains the characteristic direction (P, Q, R)
defined by the p.d.e at each point).
In the case of quasi-linear equation, the characteristic curves are a family of space curves
whose tangent at each point coincides with the characteristic direction (P, Q, R) at that point. These
are given by the following system of ordinary differential equations

dx dy dz
   dt (say). ... (3.23)
P  x , y , z  Q  x, y , z  R  x, y , z 

dx dy dz
or  P  x, y , z  ,  Q  x, y , z  ,  R  x, y , z  . ... (3.24)
dt dt dt
By the existence and uniqueness of the solution of IVP of a system of ordinary differential equations
there passes a characteristic curve

x  x  x0 , y0 , z0 , t  , y  y  x0 , y0 , z0 , t  , z  z  x0 , y0 , z0 , t  ... (3.25)

through each point  x0 , y0 , z0  .

Hence there is a two parameter family of characteristic curves. (Two parameter family of
characteristics are nothing but the curves of intersection of the surfaces u  x, y, z   C1 and
v  x, y, z   C 2 ).
Eliminating s and t from (3.25) we get the required integral surface.
Result : Every surface generated by a one parameter family of characteristics is an integral surface.

Proof : Let z  z  x, y  be an integral surface. Take P  x, y , z  be any point on the surface. Then the
tangent to the characteristic curve passing through that point lies on the plane to the surface. Thus the

145
tangent plane to the surface at each point contains the line with direction ratios (P, Q, R). Hence the
surface is an integral surface.
Conversely : We prove that every integral surface is generated by a family of characteristic curves.
Consider an integral surface

z  z  x, y  . ... (3.26)

Let x  x (t ), y  y (t ) be the solutions of the equations

dx
 P  x, y , z  x, y   ,
dt

dy
 Q  x, y , z  x, y   , ... (3.27)
dt

with the initial conditions x  x0 , y  y0 at t = 0. The corresponding curve in 3-dimension is

x  x (t ), y  y (t ), z   x (t ), y (t )  ... (3.28)
We see that this curve lies on the given integral surface (3.26).
Further,

dz dx dy
 zx  zy
dt dt dt

 P  x, y , z  z x  Q  x, y , z  z y

 R  x, y , z 

 The curve satisfies equations for characteristic curves. viz.

dx dy dz
 P  x, y , z  ,  Q  x, y , z  ,  R  x , y , z  .
dt dt dt
Therefore, integral surface (3.26) is generated by the characteristic curves.
Theorem : Consider the first order quasi-linear partial differential equation

P  x, y , z  z x  Q  x, y , z  z y  R  x, y , z 

where P, Q and R have continuous partial derivatives with respect to x, y and z and they do not vanish
simultaneously. Let z  z0 ( s ) be prescribed along the initial curve given by

 0 : x  x0 ( s ), y  y0 ( s )

x0 , y0 and z 0 being continuously differentiable functions. Further, for a  s  b , if

dy0 dx
P  x0 ( s ), y0 ( s), z 0 ( s )   0 Q  x0 ( s ), y0 ( s ), z 0 ( s )   0 ,
ds ds

146
then there exists a unique solution z (x, y) defined in some neighbourhood of the initial curve  0 , which
satisfies the p.d.e. and the initial condition

z  x0 ( s), y0 ( s)   z 0 ( s ) .

Proof : Consider the p.d.e

P  x, y , z  z x  Q  x, y , z  z y  R  x, y , z  ... (3.29)
where P, Q, R are continuous differentiable functions of x, y, z and do not vanish simultaneously.
Let x  x0 ( s ), y  y0 ( s ) ... (3.30)
be the initial data curve and

dy0 dx
P  x0 ( s ), y0 ( s), z 0 ( s )   0 Q  x0 ( s ), y0 ( s ), z 0 ( s )   0 . ... (3.31)
ds ds
Claim : We prove that z (x, y) is a unique solution of p.d.e. (3.29) satisfying

z  x0 ( s), y0 ( s)   z0 ( s) .

We know the integral surface of (3.29) are the family of space curves and are given by the system of
ordinary differential equations

dx dy dz
 P  x, y , z  ,  Q  x, y , z  ,  R  x, y , z  . ... (3.32)
dt dt dt

We solve these equations to find a unique family of characteristics (through  x0 , y0 , z0  )

x  x  x 0 , y0 , z 0 , t   x  s , t  ,

y  y  x0 , y0 , z0 , t   y  s, t  ,

z  z  x0 , y0 , z0 , t   z  s, t  , ... (3.33)
where x, y, z have continuous derivatives w.r.t. the parameters s and t satisfying the initial conditions

x  s, o   x0 ( s ), y  s, o   y0 ( s ) and z  s, o   z 0 ( s )
We see from equations (3.33) that

  x, y  x xt
 s   xs yt  y s xt 
  s, t  t 0 y s yt t 0
t 0

  x s Q  y x P t  0 by equation (3.32)

 0. due to admissibility conditions (3.31)

Now solving equations (3.33) for s and t, we obtain the relation

147
Say   x, y   z  s  x, y  , t  x , y   ... (3.34)

At t = 0 we get

  x 0 , y0   z  s , o   z 0 ( s )

This implies that   x, y  satisfies the initial condition. To prove   x, y  also satisfies the equation
(3.29).
We consider,

P x  Q y  P  z s s x  zt t x   Q  z s s y  zt t y 

 z s  Ps x  Qs y   zt  Pt x  Qt y 

P x  Q y  z s  s x xt  s y yt   zt  t x xt  t y yt  . due to equation (3.32)

However, by chain rule, we have

s x xt  s y yt  st  0 (s and t are independent parameters)

and t x xt  t y yt  t t  1

Hence the above equation becomes,

P x  Q y  zt  R  x, y, z  by equation (3.32)

i.e. P x  Q y  R  x, y, z 

This shows that x    x, y  satisfies the p.d.e. (3.29). Thus z    x, y  is a solution of p.d.e. (3.29).

Uniqueness : Let   x, y  be not unique solution of (3.29). This means that there are two surfaces
which intersect along the given initial curve. Through each point on the initial curve, there passes one
and only one characteristic curve. Therefore, this characteristic curve has to be on both the surfaces.
Hence the same family of characteristic curves which passes through each point of the initial curve lie
on both the surfaces.
Hence both the surfaces must coincide as both are generated by the same family of characteristics
curces.
This proves the uniqueness.
Example 1 : Solve the initial value problem for the quasi-linear equation
zz x  z y  1

with the initial conditions

1
x  s , y  s, z  s for 0  s  1 .
2

148
Solution : Given p.d.e. i.e. a quasi-linear p.d.e. given by
zz x  z y  1 , ... (3.35)

where P  x, y , z   z , Q  1, R  1 , ... (3.36)

subject to the initial conditions

1
x  x0 ( s )  s, y  y 0 ( s )  s , z  z 0 ( s )  s , 0  s  1. ... (3.37)
2
We observe that

dy0 dx
P  x0 ( s ), y0 ( s), z 0 ( s )   0 Q  x0 ( s ), y0 ( s ), z 0 ( s ) 
ds ds

1
 s 1
2
 0 for 0  s  1
 from the above theorem that there exists unique solution z (x,y) satisfies the p.d.e. and the initial
condition. Hence we solve the equations.
We know the family of characteristic curves which generate the surface are the solution of the equations

dx dy dz
 P,  Q,  R.
dt dt dt

dx dy dz
i.e.  z,  1,  1, ... (3.38)
dt dt dt
with the initial conditions

1
x  s, 0   s , y  s , 0   s, z  s , 0   s. ... (3.39)
2
From equation (3.38) we find
z  t  C1 , y  t  C2

dx
and  t  C1 gives
dt

t2
x  C1t  tC3 .
2
Hence the family of characteristic curves through the initial data are found to be

t2 1
x  st  s ,
2 2

149
 y t s,

s. ... (3.39)
z t
2
Solving these equaqtions for s and t interms of x and y we obtain (from the first two equations of 3.39)

t2 1
x  t  y t  y t
2 2

1 
 x  y  t  y  1
2 

or yx . ... (3.40)

t
y
1
2
Substituting this in y = t + s we get

yx
s y
y
1
2

y2
x
s 2 . ... (3.41)
y
1
2

s
Substituting the values of t and s from equations (3.40) and (3.41) in z  t  we get
2

 y2 
2 y  x   x  
z  2 . ... (3.42)
2  y
This is the required solution of integral surface.
Example 2 : Solve the Chauchy problem for
2 z x  yz y  z

when the initial data curve is

C : x0  s, y0  s 2 , z 0  s , 1  s  2 .
Solution : The partial differential equation is given by
2 z x  yz y  z , ... (3.43)

150
subject to the initial conditions

x0  s, y0  s 2 , z0  s, 1  s  2 . ... (3.44)
Here P = 2, Q = y, R = z .

Hence P  x0 , y0 , z 0   2, Q  x0 , y0 , z0   s 2 , R  x0 , y0 , z0   s .

Therefore, we observe that

dy0 dx
P  x0 , y 0 , z 0   0 Q  x 0 , y 0 , z 0   4 s  s 2 ,
ds ds
 0 for 1  s  2 . ... (3.45)
The admissibility condition (3.45) implies that there exists unique solution z (x, y) satisfies the p.d.e.
and the initial conditions.
We know the family of characteristic curves which generate the surface are the solutions of the equations

dx dy dz
 2,  y, z, ... (3.46)
dt dt dt

such that x0  s, y0  s 2 , z 0  s .
Solving the equations (3.46) we obtain

x  2t  C1 , log y  t  log C2  y  C2et ,

and log z  t  log C3  z  C3et . ... (3.46a)

We have from equations (3.44) and (3.46a)

C1  s, C2  s 2 , C3  s .
Thus the family of characteristic curves is found to be
x  s  2t , ... (3.47)

y  s 2et , ... (3.48)

z  set . ... (3.49)

The solution is obtained by eliminating s and t between equations (3.47), (3.48) and (3.49).
Therefore, we see that

xz  y  s 2e t  2 stet  s 2 e t ,

xz  y  2 ste t ,

xz  y
t.
2z

151
Substituting in (3.47) we get
 xz  y 
x  2   s,
 2z 
y.
s
z
Therefore the equation z  se t becomes

 y  xz  y 
z   exp  ,
 z  2z 

 xz  y 
or z 2  yexp  . ... (3.50)
 2z 
This is the required solution (integral surface).
Example 3 : Find the solution of the initial value problem for the quasi-linear equation
z x  zz y  z  0 y and x > 0

for the initial data curve

C : x0  0, y0  s, z0  2 s ,   s   .
Solution : The quasi-linear p.d.e. is given by
z x  zz y   z , ... (3.51)
with the initial data curve
C : x0  0, y0  s, z0  2 s .
Here P =1, Q = -z and R = – z.

Therefore P  x0 , y0 , z0   1, Q  x0 , y0 , z0   2s, R  x0 , y0 , z0   2s .

We observe from the admissibility condition that

dy0 dy
P  0 Q 1 0. s ... (3.52)
ds ds
This shows that there exists unique solution z (x, y) satisfies the p.d.e. and the initial conditions.
We know the family of characteristic curves which generate surface are the solutions of the equations

dx dy dz
 P,  Q, R.
dt dt dt

dx dy dz
i.e.  1,   z,  z , ... (3.53)
dt dt dt

152
such that x t 0  x0  0 , y t 0  s , z t 0  2s .
Solving these equations we get

x  t  C1 , log z  t  log C2  z  C2e t ,

dy
and  C2 e t  y  C2  e t dt  C3 ,
dt

 y  C2e t  C3 .
Hence the family of characteristic curves through the initial data curve is found to be

x  t, 

y  2 se t  3s, and 
... (3.54)

z  2 se t . 

The solution is obtain by eliminating t and s from equation (3.54). Thus we have

y
s x ,
2e 3
Substituting this value in z  2 se t we get

2y
z x
ex ,
2e  3

2 y 2
z for log    x  0 ... (3.55)
3e x  2 3

This is the required solution (integral surface).

2
Note : The solution breaks down at x  log .
3
Example 4 : Find the integral surface for the differential equation

z  xz x  yz y   y 2  x 2

passing through the initial data curve (2s, s, s).

Solution : The quasi-linear p.d.e is given by

zxz x  zyz y  y 2  x 2 , ... (3.56)

subject to the initial conditions (2s, s, s)

i.e. x0  2 s, y0  s, z 0  s . ... (3.57)

153
Here P  x, y, z   xz, Q  x, y, z    zy, R  x, y, z   y 2  x 2 .

Therefore P  x0 , y0 , z0   2 s 2 , Q  x0 , y0 , z0    s 2 , R  x0 , y0 , z0   s 2  4 s 2  3s 2 .

We observe from the admissibility condition that

dy0 dx 2
 P  0  Q  2 s 2  2   s   4 s 2  0 , s  0 . ... (3.58)
ds ds
This shows that there exists unique solution z (x, y) satisfies the p.d.e. and the initial conditions.
We know the family of characteristic curves which generate surface are the solutions of the equations

dx dy dz
 P,  Q,  R.
dt dt dt

dx dy dz
i.e.  xz ,   zy ,  y2  x2 , ... (3.59)
dt dt dt
Satisfying the initial data (3.57). To solve equations (3.59) we write these equation as

dx dy dz
  2  dt . ... (3.59a)
xz  yz y  x 2

dy y .
 
dx x
Integrating we get
xy  C1 .

xdx  ydy  zdz

Now each ratio of equation (3.59a) 
0
 xdx  ydy  zdz  0 .
Integrating we get

x 2  y 2  z 2  C2 .
Hence the family of characteristic curves through the initial data (3.57) is given by

xy  2 s 2 ,

x 2  y 2  z 2  6s 2 .
If we choose

2s 2
xt y
t

154
and z 2  6 s 2  t 2  4 s 2t 2 .
Eliminating t and s we get the required integral surface

z 2  6 s 2  x 2  x 2 y 2 x 2 ,

 z 2  3 xy  x 2  y 2 .
This is a required integral surface.

Example 5 : Find the integral surface passing through x = 1, z  y 2  y of the equation

x3 z x  y 3x 2  y  z y  z  2 x 2  y  .

Solution : The given p.d.e. is

x3 z x  y 3x 2  y  z y  z  2 x 2  y  , ... (3.60)

subject to the initial conditions

x  1, z  y 2  y . ... (3.61)
Comparing (3.60) with the standard equation we have

P  x3 , Q  y 3x 2  y  , R  z  2 x 2  y  . ... (3.62)

We choose the parameter y0  s

 z0  s 2  s  z0  s  s  1 .
Therefore, the initial data is

x0  1, y0  s, z0  s  s  1 . ... (3.63)
Hence we have

P  x0 , y0 , z0   1, Q  x0 , y0 , z 0   s  3  s  , R  x0 , y0 , z0   s  s  1 2  s  .

We observe from the admissibility condition that

dy0 dx
P  0 Q  1 0  0 .
ds ds
This shows that there exists unique solution z (x, y) satisfying the p.d.e. and the initial conditions.
Thus the family of characteristic curves which generate surface are the solution of the equations

dx dy dz
   dt
y  3x  y  z  2 x 2  y 
x 3 2 ... (3.64)

dx dy dz
  x3 ,  y 3x 2  y  ,  z  2 x 2  y  , ... (3.65)
dt dt dt

155
satisfying the initial data (3.63). To solve equations (3.65) consider the each ratio of equation (3.64)

1 1 1 1 1 1
 dx  dy  dz  dx  dy  dz
x y z x y z
 2 2 2

 x  3x  y  2 x  y 0

dx dy dz
   0.
x y z
Integrating we get
 log x  log y  log z  log C1

y
  C1 . ... (3.66)
xz
Now consider the ratios

dx dy

y  3x 2  y 
3
x

  3x 2  y 
dx dy

x3 y

 3 x 2 ydx  y 2 dx  x 3dy

 3 x 2 ydx  x 3 dy   y 2 dx

3x 2 ydx  x 3dy  x3 
   dx   dx  d  .
y2  y 
Integrating we get

x3
x   C2
y

x3
or  x  C2 . ... (3.67)
y
Using the initial data (3.63) in equations (3.66) and (3.67) we get

1
C1  , ... (3.68)
s 1

1
C2  1 . ... (3.69)
s

156
Concequently, the family of characteristic curve is obtain by eliminating C1 and C2 from equations
(3.66) and (3.67).

This gives y 1 , ... (3.70)


xz s  1

x3 1
and  x  1 . ... (3.71)
y s
If we choose x=t, ... (3.72)

t3 1
we get  t  1 ,
y s

t3 1 1  s 1  t 
 1 t  ,
y s s

y s st 3
  y  . ... (3.73)
t 3 1  s 1  t  1  s 1  t 
Consequently, we have from equation (3.70)

st 3 z s  s  1 t 2
 z . ... (3.74)
1  s 1  t  t s  1 1  s 1  t 
Eliminating t and s between (3.72), (3.73), (3.74) we get

sx 3
y  y 1  s 1  x   sx3 ,
1  s 1  x 

y  s  y  xy   sx 3 ,

y  s  x 3  y  xy  .

y
or 3
 s. ... (3.75)
x  y  xy
Using (3.72) and (3.75) in (3.74) we get

 y  y 
x2  3  3
 1 
x  y  xy  x  y  xy  
z 
y
1  1  x  3
x  y  xy

z  x 3  y  xy   y 2  x 2 y

157
 x 3 z  x 2 y  xyz  y 2  yz

x 2  xz  y   y  xz  y   yz

 x 2  y   xz  y   yz
This is the required integral surface.

(c) Non-Linear First Order Partial Differential Equations :

In this section we shall consider a method of finding integral surfaces of a non-linear partial
differential equations of first order, which is based largely on geometrical ideas. The method was first
developed by Cauchy and is called Cauchy’s Method of Characteristics.
The method involves the following steps :

Step 1 : Let f  x, y , z , p , q   0 ... (3.76)

be a given p.d.e. and the initial data curve be
x  x0 ( s ), y  y0 ( s ), z  z0 ( s) . ... (3.77)

Using (3.76) and (3.77) determine the functions p0 ( s ) and q0 ( s ) such that

f  x0 ( s), y0 ( s), z0 ( s), p0 ( s), q0 ( s)   0 , ... (3.78)

dz 0 dx dy
and  p0 0  q0 0 (These are called strip conditions) ... (3.79)
ds ds ds

Note : There could be several choices for p0 ( s ) and q0 ( s ) . One can find unique solution for each
such choice.
Step 2 : Once a choice for p0 ( s ) and q0 ( s ) is made (i.e. the initial strip is chosen) we can solve the
following Cauchy characteristic equations

dx dy dz
 f p,  fq ,  pf p  qf q ,
dt dt dt

dp dq
  fx  fz p ,   f y  fzq , ... (3.80)
dt dt
subject to the initial conditions
x  x0 ( s ), y  y0 ( s ), z  z0 ( s ), p  p0 ( s) and q  q0 ( s ) at t = 0. ... (3.81)

The corresponding characteristic curves x  x ( s, t ), y  y ( s, t ), z  z ( s, t ) generate the required

integral surface after eliminating s and t. Let the solution surface be in the z = z (x, y).
The method is illustrated in the following example.

158
Example 1 : Determine the characteristics of the equation

z  p2  q2
and find the integral surface which passes through the curve

s2
x0  s, y0  0, z 0   . (the parabola 4 z  x 2  0, y  0 ).
4
Solution : Let the given p.d.e. be denoted by

f  x, y , z , p , q   p 2  q 2  z  0 . ... (3.82)
The initial data curve is

s2 . ... (3.83)
x0  s, y0  0, z 0  
4

To determine the function p0 ( s ) and q0 ( s ) we have the strip conditions.

s2
f  x 0 , y 0 , z 0 , p0 , q 0   0  p 0 2  q 0 2   0,
4

dz 0 dx dy s
and  p0 0  q 0 0    p 0 . ... (3.84)
ds ds ds 2

s2 s2 s2
 q0 2    q0 2 
4 4 2

or s . ... (3.85)
q0 
2
Now the Cauchy characteristic equations (3.80) become

dx
 2p,
dt

dy
 2q ,
dt

dz
 2 p 2  2q 2 ,
dt

dp
 p,
dt

dq
and  q. ... (3.86)
dt

159
Thus we have

dx dy dz dp dq
     dt
2 p 2 q 2 p  2 q
2 2
p q

dx dp
The ratios  give
2p p

 x  2 p  C1 . ... (3.87)
Now consider the ratios

dy dq
  y  2 q  C 2 . ... (3.88)
2 q q
The conditions

dp
 dt  log p  t  log C3 ,
p

 p  C3e t . ... (3.89)

dq
Similarly,  dt  log q  t  log C4 ,
q

 q  C4 e t . ... (3.90)

dz
Now the equation  dt
2 p  2q 2
2

gives dz  2  C3 2 e 2t  C 4 2 e 2t  dt ,

dz  2  C3 2  C 4 2  e 2t dt ,

 z   C3 2  C 4 2  e 2 t  C 5 . ... (3.91)

Now on using the initial data (3.83) to (3.85) we have from equations (3.87) to (3.91) that
s s
C1  s 2, C2  2 s, C3   , C4  , C5  0
2 2
Eliminating these constants, we have finally from equations (3.87) to (3.91)

x   set  2s , ... (3.92)

y  2  s 1  e t  , ... (3.93)

s
p   et , ... (3.94)
2
160
 s t
q e , ... (3.95)
2

s 2 2t
and z e . ... (3.96)
4
Solving (3.92) and (3.93) for s and et we get

 y 
s x 
 2

x  2y
and et 
y
x
2
Substituting in (3.96) we get
2
1 y   x  2y 
z   x   
4 2   x y 
 
 2 

1
 x  2y .
2
z ... (3.97)
4
This is the required integral surface.
Example 2 : Find by the method of characteristics, the integral surface of
pq  xy
which passes through the curve z = x, y = 0.
Solution : Let the given p.d.e. be

f  x, y, z , p, q   pq  xy  0 , ... (3.98)
and the initial data curve be
x0 ( s )  s, y0 ( s )  0, z0 ( s )  s . ... (3.99)

Hence the equation f  x0 , y0 , z0 , p0 , q0   0 becomes

p0 q0  x0 y0  0

 p0 q0  0  0

 p0 ( s )  q0 ( s )  0 ... (3.100)
Now the equation

161
 dz 0 dx dy
 p0 0  q0 0
ds ds ds

becomes 1  p0 1  q0  0 ... (3.101)

 p0  1 . ( unique initial data)

From equation (3.100)
 q0  0 . ... (3.102)
Now the Cauchy characteristics equations (3.80) become

dx dy dz dp dq
 q,  p,  2 pq,  y,  x. ... (3.103)
dt dt dt dt dt
Thus from equations

dx dq
 q and  x,
dt dt

d 2 x dq
we have  x
dt 2 dt

d 2x
 2  x  0,
dt
which has solution

x  aet  bet . ... (3.104)

dx
q  aet  bet . ... (3.105)
dt
Similarly, from equations

dy dp
 p,  y,
dt dt

d 2 y dp
we have  y
dt 2 dt

d2y
 2
 y  0  y  ce t  det . ... (3.106)
dt

dy
Hence, p  ce t  det ... (3.107)
dt

162
 dz
Therefore,  dt
2 pq

 dz  2  ce t  det  ae t  be  t  dt

dz  2  ace 2t  bc  ad  bde 2t  dt

 z  ace 2t  bde 2t  2  bc  ad  t  e . ... (3.108)

Using the initial data (3.99) to (3.102) at t = 0 we get

s
ab , ... (3.109)
2

1
c  d  , ...(3.100)
2
and
s  ac  bd  e
s s s
s  e  e  ... (3.111)
4 4 2
Finally, we have

 et  e t 
x  s   x  s cosh t , ... (3.112)
 2 

et  e t
y  y  sinh t , ... (3.113)
2

s 2t s 2t s s s
z e  e  2  t  ,
4 4 4 4 2

s  e 2t  e 2t  s
z  
2 2  2

s
 z   cosh 2t  1
2

s
z cosh 2 t  sinh 2 t  1
2

s
 cosh 2 t  cosh 2 t 
2

z  s cosh 2 t . ... (3.114)

163
 1 t
Now p e  e  t   p  cosh t . ... (3.115)
2

 et  e t 
q  s   q  s sinh t . ... (3.116)
 2 
Now eliminating s and t from (3.112), (3.113) and (3.114) we get

z 2  s 2 cosh 4 t
2
  s cosh t  cosh 2 t

 x 2 1  sinh 2 t 

z 2  x 2 1  y 2  . ... (3.117)

This is the required integral surface through the given initial data curve.
Example 3 : Find the characteristics of the equation
pq  z
and determine the integral surface which passes through the parabola x = 0, y2 = z.
Solution : Let the given p.d.e. be denoted by

f  x, y, z , p, q   pq  z  0 . ... (3.118)
The initial data curve is

x0  0, y0  s, z0  s 2 . ... (3.119)

To determine the functions p0 ( s ) and q0 ( s ) , we have the strip conditions

f  x0 , y0 , z0 , p0 , q0   0

 p0 q0  z0  0  p0 q0  s 2  0 , ... (3.20)

dz 0 dx dy
and  p0 0  q0 0
ds ds ds

 2s  p0 (0)  q0  q0  2s . ... (3.121)

s
Equation (3.120)  p0  . ... (3.122)
2
Now the Cauchy characteristic equations (3.80) become

dx
 q  dx  C2 et dt  x  C2 e t  C3 .
dt

164
 dy
 p  dy  C1e t dt  y  C1e t  C 4 ,
dt

dz dz
 2 pq   2t  log z  t 2  log C5 ,
dt z
2
z  C5 e t ,

dp
 p  p  C1e t ,
dt

dq
 q  q  C 2e t . ... (3.123)
dt
Thus we have

dx dy dz dp dq
     dt .
q p 2 pq p q
The ratios

dx dq
  x  q  C1 , ... (3.124)
q q

dy dp
and   y  p  C2 . ... (3.125)
p p

dp
Also  dt  log p  t  log C3
p

 p  C3e t , ... (3.126)

dq
and  dt  log q  t  log C4  q  C4e t . ... (3.127)
q

dz
Now  dt  dz  2C3C4 e 2t  dt
2 pq

z  C3C 4e 2t  C5 . ... (3.128)

Now using the initial data curve (3.119), (3.121) and (3.122) at t = 0, we have from (3.124)
0  2s  C1  C1  2s ,

s s
s  C2  C 2  ,
2 2

165
 s
Equation (3.126) gives  C3 .
2

Equation (3.127) gives 2s  C4 ,

s
and equation (3.128) gives s2  2 s  C 5  C5  0 .
2
Finally we have on substituting the values of these constants in equations (3.124) to (3.128)

x  2 se t  2s , ... (3.129)

s t s s
y e   y   e t  1 , ... (3.130)
2 2 2

s
p  et , ... (3.131)
2

q  2 se t , ... (3.132)

and z  s 2 e 2t . ... (3.133)

Now eliminating s and t between (3.129), (3.130) and (3.133) we have

x 
 et  1 
2s 
 adding we get
2y
 e  1
t
s 

 x  4y
et  .
4s
Putting this in (3.129) we get

 x  4y 
x  2s   1
 4s 

 x  4 y  4s 
x  2s  
 4s 
 2 x  x  4 y  4s
or 4s   x  4 y

4y  x
Therefore, s
4

Hence x  4y .
et 
4y  x
166
Substituting this in (3.133) we get
2 2
 4y  x   x  4y 
z 
 4   4 y  x 

2
 x  4y 
z  .
 4 
This is the required integral surface.
Example 4 : Find by the method of characteristics the integral surface of the equation

p 2 x  qy  z  0
which passes through the initial data y = 1, x + z = 0.
Solution : Let the given p.d.e. be denoted by

f  x, y, z , p, q   p 2 x  qy  z  0 . ... (3.134)
The initial data curve is
x0  s, y0  1, z0   s . ... (3.135)

To determine the functions p0 ( s ) and q0 ( s ) , we have the strip conditions

f  x0 , y0 , z 0 , p0 , q0   0  p0 2 s  q0  s  0 , ... (3.136)

dz0 dx dy
and  p0 0  q0 0  1  p0  q0 (0)  p0  1 . ... (3.137)
ds ds ds
Therefore equation (3.136) give
s  q0  s  0  q0  2 s . ... (3.138)
Now the Cauchy characteristic equations (3.80) reduce to

dx
 2 px ,
dt

dy
 y,
dt

dz
 2 p 2 x  qy ,
dt

dp
  p2  p ,
dt

dq dq
 q  q   0. ... (3.139)
dt dt

167
 dy
The equation  y gives
dt

 y  C1e t ... (3.140)

dp dp
The equation   p  p  1    dt
dt p  p  1

1 1 
 dp     dt
 p p 1 
Integrating we get

log p  log  p  1  t  log C 2

p C2 et
  C2 et  p  . ... (3.141)
 p  1  C2et  1
dq
The equation  0  q  C3 . ... (3.142)
dt

dx 2 px dx 2x
The equation   
dp  p  p  1 dp  p  1

or dx 2dp .

x p 1
Integrating we get

log x  2 log  p  1  log C 4

2
 x  C 4  p  1 . ... (3.143)

C2 e t C2 e t
From (3.141) p  p 1  1
C2 e t  1 C2 e t  1

1
 
2 2
 p 1    p  1  C 2 e t
 1
C2e t  1

x  C4  C2e t  1 .
2
Hence ... (3.144)

dz
Now the equation  2 p 2 x  qy becomes
dt

168
 2
 C et 
  C 2 e  1  C3C1e
dz t 2 t
 2C4  2t
dt  C2e  1 

dz
 2C 2 2C 4 e 2t  C1C3e t
dt

 dz  2C2 2C4  e 2t dt   C1C3  e t dt 

Integrating we get

z  C2 2C4 e 2t  C1C3et  C5 . ... (3.145)

Now using the initial data curve
x0  s, y0  1, z0   s, p0  1 and q0  2 s ,

we have from equation (3.140)  C1  1 .

C2
From equation (3.141) we have 1   C2  C 2  1 .
C2  1

From equation (3.142) we find C 3  2 s .

2 C4
From equation (3.143) we have s  C4  1  1 s  C4  4 s .
4
From equation (3.145) we find C5 = 0.
Thus the family of characteristic curves are given by

x  C4  C2e t  1
2

 x  s  et  2  ,
2

y  e t , z  se 2t  2se t , .... (3.146)

1 t
e
2 et
p  p and q  2 s .
1 t  et  2
 e  1
2 

Solving the equations (3.146) for s and t we get

x
s and e t  y .
 y  2 2

Putting this in the expression for z we get

169
 x 2 xy xy
z y2  or  z  .
 y  2 2
y2 x y

This is the required integral surface.

Example 5 : Find the integral surface of the equation

z  p 2  3q 2

passing through C : x0  s, y0  0, z0  s 2 .

Show that there are two possible initial strips p0  2 s , q0   s .

Solution : Let the given non-linear p.d.e. be given by

f  x, y, z , p, q   p 2  3q 2  z  0 , ... (3.147)
with the initial conditions

x0  s, y0  0, z0  s 2 . ... (3.148)

To determine initial strip, we have the strip conditions

f  x0 , y0 , z0 , p0 , q0   0 ,

 p0 2  3q0 2  z0  0  p0 2  3q0 2  s 2 , ... (3.149)

dz 0 dx dy
and  p0 0  q0 0
ds ds ds

 p0  2 s . ... (3.150)

Hence equation (3.149) gives q0   s .

Hence the initial strip is

x0  s, y0  0, z0  s 2 , p0  2s, q0   s . ... (3.151)

Now the Cauchy characteristic equations (3.80) are given by

dx
 2p,
dt

dy
 6q ,
dt

dz
 2 p 2  6q 2 ,
dt

dp
 p,
dt
170
 dq
 q.
dt

dx
  2.
dp

Integrating we get x  2 p  C1 . ... (3.152)

dy
Similarly,  6  dy  6dq ... (3.153)
dq

 y  6q  C2

dp
 p  log p  t  log C3
dt

 p  C3e t
Hence equation (3.152) gives

x  2C3et  C1 . ... (3.154)

dq
Now  q  q  C 4e t . ... (3.155)
dt
Consequently, equation (3.153) gives

y  6C4et  C2 .

dz
Now  2 p 2  6q 2  2 z ,
dt

dz
  2dt .
z
Integrating we get

 z  C5e 2t . ... (3.156)

Now using the initial data curve (3.151) we have from above equations (3.152) to (3.156)
s  4 s  C1  C1  3s ,

0  6s  C2  C2  6s ,

2 s  C 3  C3  2 s ,

s  C4  C4  s ,

s 2  C5  C5  s 2 .
171
Therefore, the characteristic curves are given by

x  4se t  3s  x  4 s  et  1  s, 


y  6se  6 s  y  6 s e  1 , 
t t 

z  s 2 e 2t ,  .... (3.157)
t 
p  2se , 
q  se .t 


The integral surface is obtained by eliminating s and t from (3.157) we get

2 2
 3x  2t   y 
z  1  2  3x  2 y  
 3   

9  2x  2 y   2x  y 
2 2
z z . ... (3.158)
36 4
Exercise :
1. Find the characteristics of the equation pq = z and determine the integral surface which passes
through the straight line
x = 1, z = y.
2. Find the characteristics of the equation

p2  q2  2
and determine the integral surface which passes through x = 0, z = y.

172
 UNIT - V

SECOND ORDER PARTIAL DIFFERENTIAL EQUATIONS

Introduction :
Partial differential equations of second order describe the physical behaviour of many practical
situations in science and engineering. We will see how such second order partial differential equations
arise in physics and engineering mathematics. Further, in many situations a given partial differential
equation of second order is difficult to solve, hence in this unit we classify the second order partial
differential equation in to elliptic, parabolic and hyperbolic forms by transforming it into canonical form.
The idea of reducing the given partial differential equation to a canonical form is that the transformed
equation assumes a simple form so that the subsequent analysis of solving the equation is easy. We also
discuss the methods of separation of variables of solving second order partial differential equation.
Definition : A semi-linear second order partial differential equation is expressed in the form.
Ru xx  Su xy  Tu yy  g  x, y, u , u x , u y   0 , ... (1.1)

where R, S, T are continuous functions of x and y only and R 2  S 2  T 2  0 , u is a dependent

variable. Equation (1.1) can also be written as

Rr  Ss  Tt  g  x, y, u, u x , u y   0 , ... (1.1a)

where r  u xx , s  u xy , t  u yy .
Solution of the Equation :

Definition : A function u  f  x, y  is said to be a regular solution of equation (1.1) in a domain

D     iff f  x, y   on D and the function and its derivatives satisfy equation (1.1) identically..

Origin of Partial Differential Equation :

One dimensional wave equation :
Result : Derive an equation governing small transverse
vibrations of an elastic string. T2
Q 2
Proof : Let an elastic string be stretched to a length 
P s
and then fixed at the end points. Let the string be distorted
1
and further let at time t = 0 it be released and allowed to T1
vibrate. Our aim is to obtain the equation which governs x  x
the deflection y  x, t  at any point x after any time t > 0.

173
 Let y = y (x, t) be the displacement from the mean position (x-axis) of a string at time t at point
x. Let  s be the small portion of the string between two points P amd Q. We assume that the string is
homogeneous (i.e. mass per unit length is constant) perfectly elastic (i.e. does not offer any resistence
on bending) and weight of the string is neglected (i.e. action of the gravitational force on the string is
neglected).
In order to find the differential equation which describes the motion of string, we consider the
forces acting on the portion  s . Let T1 and T2 be the tensions at points P and Q respectively acting
along the tangential direction. Since there is no motion of the string in the horizontal direction, therefore,
the horizontal components of the tensions will be constants.
 T1 cos 1  T2 cos 2  constant  T (say). ... (1.2)
The resultant vertical force acting on the portion PQ is
T2 sin 2  T1 sin 1 .
Hence, the equation of motion is given by
Force = Mass • Acceleration

2 y
 T2 sin 2  T1 sin 1  9 s  , ... (1.3)
t 2

2 y
where 9 is the density of the string and 9 s is the mass of the portion PQ and is the acceleration
t 2
in the vertical direction.
We write from equation (1.3) that

T2 sin 2  T1 sin 1  s 2 y
9  2
T T t

T2 sin 2 T1 sin 1  s 2 y
  9  2
T2 cos 2 T1 cos 1 T t

 s 2 y .
or tan 2  tan 1  9  ... (1.4)
T t 2

Since tan 1 and tan 2 are the slopes of the curve of the string at points P and Q respectively,,
therefore we have by definition.

 y   y 
tan 1       and
 x  P  x  x

 y   y 
tan 2      
 x Q  x  x  x .

174
Hence equation (1.4) beomes

 y   y   s 2 y .
     9
 x  x  x  x  x T t 2

In the limiting case as  x  0 i.e. Q  P , we have

s x,
therefore, we write

  y   y  
  x    
x  x  x  x  9 2 y ,
lim  
 x 0  x  T t 2

2 y 9 2 y
  ,
x 2 T t 2

2 y 1 2 y , ... (1.5)
 
x 2 c 2 t 2

2 T
where c  , and c represents the speed of the wave propagation. Equation (1.5) is called the one
9
dimensional wave equation.
Heat Conduction Equation :
Result : Derive the second order partial differential equation which describes the temperature distribution
in a homogeneous isotropic solid.
Note : Homogeneous means distribution of material is uniform, isotropic means the material properties
are the same in all directions.
Specific heat of the solid means the amount of heat absorbed by the matter per unit mass per unit rise
in temperature.
Density of the solid means mass per unit volume.
Proof : Consider a homogeneous isotropic solid and V be any arbitrary volume inside the solid bounded
by a surface S. Let  V be a volume element. We denote
c : the specific heat of the solid,
9 : the density of the solid and,
u : the temperature which is a function of position and time.
Hence the heat energy stored in the volume element  V is equal to c9u V .

Hence the total heat energy in the volume V is given by    c9udV ... (1.6)
V

175
If  S is a surface element, then the heat flow across  S  k un S ... (1.7)
where n is the outward normal to the surface S,
k - the thermal conductivity of the solid.

Hence the total flux across S    k undS . ... (1.8)

S

Using the Gauss-Divergence theorem, we write

Total flux across S        k u  dV ... (1.9)

V

Since the rate of change of heat energy in V is equal to the flux of heat energy across S. Therefore from
equations (1.6) and (1.9) we have


c9udV        k u  dV ,
t  V  V

 
      c9u      k u  dV  0 . ... (1.10)
V
 t 

Since V is an arbitrary volume, we have therefore


 c9u     k u   0 ,
t

u
 c9    k u   0 .
t
If the thermal conductivity k is constant through out the body, then we have

u
c9  k  2u  0 .
t

u
or  K  2u , ... (1.11)
t

k
where K  represents the heat conductivity, and
c9

2  2u  2u  2u
 u 2  2  2 .
x y z
Equation (1.11) is the required heat conduction equation.
Note : One dimensional heat equation is given by

u  2u
K 2 . ... (1.12)
t x
176
There are some equations arise in physics. One of the most important partial differential equations in
Physics is the Laplace equation given by

 2u  0

 2u  2u  2u
i.e.    0. ... (1.13)
x 2 y 2 z 2
Note : The heat equation (1.11) reduces to Laplace equation when the temperature u does not change
with time t.

u
i.e.  0 then equation (1.11) becomes
t

 2u  0 .
Note : 2-dimensional Laplace equation is given by

 2u  2 u
  0. ... (1.14)
x 2 y 2
Classification of second order Partial Differential Equation :
Result : By a suitable change of the independent variables, show that a second order partial differential
equation

Rr  Ss  Tt  g  x, y, u , u x , u y   0

can be reduced to one of the canonical forms on the basis of

S 2  4 RT  0, S 2  4 RT  0, S 2  4 RT  0 .
Proof : A semi-linear second order partial differential equation can also be written as

Lu  g  x, y , u , u x , u y   0 , ... (1.15)

where 2 2 2 , ... (1.16)

LR  S  T
x 2 xy y 2
(x, y) are independent variables and u the dependent. We change the independent variables x, y to new
independent variables  ,  by means of the transformation equations

    x, y  ,     x, y  , ... (1.17)

  ,   x  y
where J    x y   y x  0 .
  x, y   x  y

Then by using the chain rule of partial differentiation we obtain,

177
 u x  u  x  u x ,

u y  u  y  u y ,

u xy  u  x y  u  x y  u  xy  u  y x  u x y  u xy .

Similarly, we find
u xx  u  x x  u  x x  u  xx  u  x x  u x x  u xx ,

u yy  u  y y  u  y y  u  yy  u  y y  u y y  u yy .

Hence the operator (1.16) becomes

Lu  R u  x2  2u  x x  u x2  u  xx  u xx  

 S u  x y  u  x y   y x   u x y  u  xy  u xy  

T u  y2  2u y y  u y2  u  yy  u yy  .

We write this equation as

 
Lu  u R x2  S x y  T  y2  u  2 R x x  S  x y   y x   2T  y y  

 
u R x2  S x y  T y2  R  u  xx  u xx  

 S  u  xy  u xy   T  u  yy  u yy 

Substituting this in equation (1.15) we get

A  x ,  y  u  2 B  x ,  y ; x , y  u  A  x , y  u  G  , , u, u , u  , ... (1.18)

where, A  u , v   Ru 2  Suv  Tv 2 , ... (1.19)

1
B  u1 , v1; u2 , v2   Ru1u2  S  u1v2  u 2v1   Tv1v1 , ... (1.20)
2
and A, B satisfy the equation

1
A  x ,  y   A  x , y   B 2  x ,  y ; x , y   4 RT  S 2   x y   y x  .
2
... (1.21)
4
We see that the transformed equation (1.18) has the same form as that of the original equation (1.15)
under the transformation (1.17). Since the classification of (1.15) depends on S 2  4 RT ; therefore
we choose the new independent variables  and  so that the equation (1.18) takes the simplest
possible form. Thus the equation (1.18) will reduce to its simplest integrable form if the discriminant
S 2  4 RT of the quadratic equation

R 2  S   T  0 ... (1.22)
178
is either positive, zero or negative every where.
Case (i) : Let S 2  4 RT  0 .

In this case the roots 1 , 2 of the equation (1.22) will be real and distinct. Thus we choose  and 
such that
 x  1 y , ... (1.23)

and  x  2 y . ... (1.24)

These are the first order partial differential equations for  and  .
Solving equation (1.23) by Lagrange’s method, we have

dx  dy dt
 
1 1 0

dy
  1  x, y   0 . ... (1.25)
dx
Similarly, from equation (1.24) we find

dy
  2  x, y   0 . ... (1.26)
dx

If f1 ( x, y )  C1 and f 2 ( x, y )  C2 are the solutions of the ordinary differential equations (1.25) and
(1.26) respectively, and are called the characteristic curves of the equation (1.15), then we choose

  f1  x, y  ,

and   f 2  x, y  . ... (1.27)

The variables  , are called the characteristic variables. For this choice of  and  we have

A  x ,  y   R x2  S x y  T  y2

 R12 y2  S 1 y2  T  y2

 A  x ,  y    R12  S 1  T   y2 .

As 1 is a root of equation (1.22), we have therefore

A  x ,  y   0 .

Similarly, we show that

A  x , y    y2  R22  S 2  T   0 ,

 A0.
179
Consequently, equation (1.18) reduces to

2 B  x ,  y ; x , y  u  G  , , u , u , u  . ... (1.28)

Since A = 0 and S 2  4 RT  0 , then from equation (1.21) we have

B2  0  B  0 .
Thus we have from equation (1.28)

G
u     , , u , u , u  ,
2B

or u    , , u , u , u  . ... (1.29)

This is the desired canonical form of the equation (1.15). This form (1.29) is called hyperbolic form of
equation (1.15).
Case (ii) : Let S 2  4 RT  0

In this case the roots of the equation (1.22) are equal say 1  2  

We choose  such that

 x   y ,

and  to be any arbitrary function of x and y independent of  . This is the Lagranges form of the
equation, solving we obtain

  f  x, y  ,

where f  x, y   C is a solution of the equation

dy
   x, y   0 .
dx

Since A  x ,  y    y2  R 2  S   T   0 , due to (1.22)

 A  x ,  y   0 ,

and A  x , y   0 , otherwise  would be function of  . Hence from equation (1.21) we have

B = 0. Putting these values in equation (1.18) we get

A  x , y  u  G  , , u, u , u 

or u    , , u, u , u  . ... (1.30)

This is the desired canonical form of the equation (1.15) and is called the parabolic form of (1.15).

180
Case (iii) : Let S 2  4 RT  0

In this case the roots of the quadratic equation (1.22) are complex. We choose  and  as in the
case (i), so that

A  x ,  y   0  A  x , y  ,

and equation (1.18) reduces to

u    , , u , u , u  . ... (1.31)

This is similar to equation (1.29) except that the variables  , are not real but are the complex
conjugates. Hence to obtain a real canonical form we make the transformation

1
      ,
2

1
and        . ... (1.32)
2
Hence by using the chain rule of partial differentiation we have,
u  u   u   .

Using (1.32) we find

1
 u 
2
 u  iu  

and u   u      u  
 

1 i
u 
4
 u  iu     u  iu  
4

1
 u 
4
 u  u 
Hence equation (1.31) becomes

u  u     ,  , u , u , u   . ... (1.33)

This is the required real canonical form and is called an elliptic form of partial differential equation.
Thus we define the three types of canonical forms as follows :
Definition : A partial differential equation of second order viz.

Rr  Ss  Tt  g  x, y, u, u x , u y   0

is said to

181
(i) hyperbolic if S 2  4 RT  0 and the corresponding canonical form is given by

u    , , u , u , u  ,

(ii) parabolic if S 2  4 RT  0 , and the corresponding canonical form is

u    , , u, u , u  ,

(iii) elliptic if S 2  4 RT  0 and the corresponding canonical form is

u  u     ,  , u , u , u   .

Example 1 : Show that

1
A  x ,  y  A  x , y   B 2  x ,  y ; x , y   4 RT  S 2   x y   y x  ,
2
4

where A  x ,  y   R x2  S x y  T  y2 , A  x , y   R x2  S x y  T y2 ,

1
and B  x ,  y ; x , y   R x x  S  x y   y x   T  y y .
2
Solution : Consider

A  x ,  y  A  x , y   B 2  x ,  y ; x , y   R 2 x2 x2  RS x2  x y  RT  x2 y2  RS y2 x y 

 S 2 x y x y  ST  x y y2  RT x2 y2  ST  y2 x y

1

T 2 y2 y2  R 2 x2 x2  S 2  x2 y2   y2 x2  2 x y x y 
4

T 2 y2 y2  RS x x  x y   y x   2 RT  x y x y 

 ST  y y  x y   y x 

1

 A  x ,  y  A  x , y   B 2  x ,  y ; x , y    S 2  x2 y2   y2 x2  2 x y x y 
4



 RT  x2 y2   x2 y2  2 x x y y 
1 2
 A  x ,  y  A  x , y   B 2  x ,  y ; x , y    S  4 RT   x y   y x 
2
4

182
Example 2 : Reduce the equation u xx  x 2u yy  0 to a canonical form.

Solution : The equation u xx  x 2u yy  0

can be written as r  x 2t  0 . ... (1.34)

Comparing this with the standard form we have

R  1, S  0, T   x 2 .
Hence we see that

S 2  4 RT  4 x 2  0 .
 Equation (1.34) is hyperbolic. Therefore, the quadratic equation

R 2  S   T  0 ,
become  2  x2  0
   x .
Let 1  x , and 2   x be its roots. Hence the ordinary differential equations

dy dy
 1  0 and  2  0 ,
dx dx

dy dy
become  x  0 and  x  0.
dx dx
Integrating we get

x2 x2
y  C1 and y   C2 .
2 2
Therefore, we choose the new independent variables  and  in the form

x2
  y and ... (1.35)
2

x2
  y . ... (1.36)
2
Now by changing the independent variables x, y as new independent variables  and  , we obtain by
using the chain rule of partial differention
u x  u  x  u x ,

u y  u  y  u y ,

183
and u xx  u  x y  2u  x x  u x x  u  xx  u xx ,

u yy  u  y y  2u  y y  u y y  u  yy  u yy ,

where from equations (1.35) and (1.36) we obtain

 x  x,  y  1,  xx  1,  xx  1,  x   x,  y  1,  xy  0,  yy  0 ,

u xx  u x 2  2u   x   u  x 2   u (1)  u (1)

2

 u xx  x 2  u  2u  u   u  u ,

and u yy  u  2u  u

Substituting this in equation (1.34) we get

x 2 u  2x 2u  x 2 u  u  u  x 2 u  2x 2u  x 2 u  0

4x 2u    u  u 

1
or u 
4    
 u  u  , for x 2     .

This is a required hyperbolic canonical form.

Example 3 : Reduce the equation

y2 x2
y 2u xx  2 xyu xy  x 2u yy  ux  u y
x y
into canonical form and hence solve it.
Solution : Given equation can be written as

y2 x2
y 2 r  2 xys  x 2t  ux  u y . ... (1.37)
x y
Comparing this with the standard form, we have

R  y 2 , S  2 xy , T  x 2 .
We observe that

S 2  4 RT  4 x 2 y 2  4 x 2 y 2  0 .
Hence equation (1.37) is parabolic.
Hence the roots of the quadratic equation R 2  S   T  0 become

y 2  2  2 xy  x 2  0

184
   y  x   0
2

x
 
y. twice

Hence the solution of the ordinary differential equation

dy dy x
   x, y   0   0
dx dx y
is given by

y2  x2  c2 . ... (1.38)

Now we choose the new independent variable  such that

  x2  y2 .

Let us choose   x 2  y 2 (choice is arbitrary)

Therefore, we have
 x  2 x ,  y  2 y ,  x  2 x ,  y  2 y ,

 xx  2,  yy  2,  xx  2,  yy  2 ,

and  xy  0   xy .

Thus by changing the independent variables (x, y) to  ,  we obtain

u x  2 x  u  un  ,

u y  2 y  u  un  ,

 u xy  2 x  u 2 y  u  2 y   u 2 y  u  2 y   ,

and u xx  u  4 x 2   2u  4 x 2   u  4 x 2   u (2)  u (2) ,

u yy  u  4 y 2   2u  4 y 2   u  4 y 2   u (2)  u (2) .

Hence equation (1.37) becomes.

 
4x 2 y 2 u  2 u  u  2 y 2  u  u  

   
2 xy  4 xy  u  u  4 x 2 y 2 u  2u  u  2 x 2  u  u  =

y2 x2
 2 x  u  u   2 y  u  u 
x y
185
  16 x 2 y 2u  2u   y 2  x 2  y 2  x 2   2u   y 2  x 2  y 2  x 2 

 u  0 . ... (1.39)

Which is required parabolic canonical form. This is a homogeneous second order p.d.e. with constant
coefficients.
Integrating w.r.t.  we get

u
 f  


 u  f    g  

or u  x, y   f  x 2  y 2  x 2  y 2   g  x 2  y 2  ,
where f and g are arbitrary.
Example 4 : Reduce the equation

u xx  x 2u yy  0

to a canonical form.
Solution : Given second order p.d.e. can also be written as

r  x 2t  0 ... (1.40)
Comparing this equation with the standard form we get

R  1, S  0, T  x 2 .
We notice that

S 2  4 RT  4 x 2  0 .
Hence the equation (1.40) is elliptic. Hence the quadratic equation R 2  S   T  0 becomes

 2  x2  0
It has roots
  ix
Let 1  ix and 2  ix be the complex roots. Hence the solutions of the first order ordinary
differential equations

dy dy
 1  x, y   0 and   2  x, y   0
dx dx

dy dy
i.e.  ix  0 and  ix  0 ,
dx dx
are given by
186
 x2 x2
yi  constant and y  i  constant .
2 2
We write this as

x2 x2
iy   C1 and iy   C2 . ... (1.41)
2 2
Therefore we choose the independent variables  and  such that

x2
  iy  ,
2

x2
and   iy  .
2
To obtain the real canonical form, further we make the transformations

1 x2
         x  x,  y  0 ,
2 2

1
and   i      y   x  0,  y  1 ,
2
 xx  1,  yy  0 .
Hence we obtain
u x  u  x  u   x  u x  u  (0)

u x  xu .

u y  u  y  u   y  u  0   u  1

u y  u .

 u xx  x u  x  u  x   u

 x  xu   u

u xx  x 2u  u .

Similarly, u yy  u   y  u   y  u 

u yy  u  .

Hence the equation (1.40) becomes

x 2u  u  x 2u   0

187
 1
or u  u    u ... (1.42)
2

x 2  2 .
Which is the desired elliptic canonical form.
Example 5 : Reduce the equation

 n  1 2 u xx  y 2 nu yy  ny 2n1u y

to a canonical form and hence find its general solution

Solution : The given second order p.d.e. is

 n  1 2 u xx  y 2 nu yy  ny 2n1u y , ... (1.43)

where in this case

2
R   n  1 , S  0, T   y 2 n .

Case (i) When n= 1, we see that S 2  4 RT  0 . Hence equation (1.43) reduces to

1
u yy   u y , ... (1.44)
y
which is in the parabolic canonical form.
Case (ii) n > 1.

Then we see that S 2  4 RT  4  n  1 2 y 2 n  0 . Hence equation (1.43) is hyperbolic.

Hence the quadratic equation R 2  S   T  0

i.e.  n  1 2  2  y 2n  0
has roots

yn
 .
n 1

yn yn
Let 1  , and 2   be roots of the equation. Hence the first order ordinary differential
n 1 n 1
equations

dy dy
 1  x, y   0 and   2  x, y   0 ,
dx dx
become

188
 dy yn dy y n
  0 and   0.
dx  n  1 dx n  1
We write these equations as

dy dy
(n  1)  dx  0 and (n  1) n  dx  0 .
yn y

The solutions of these equations are given by

 y  n1  x  C1 and y  n1  x  C2 .

These are called the characteristic curves of the equation. Therefore we choose the independent variables
 and  (which are called characteristic variable) such that

1 (n  1) ,
  x n 1
  x  1,  y 
y yn

1 (n  1)
and   x n 1
  x  1,  y   ,
y yn

and  xx  0,  xy  0,  xy  0,  xx  0 ,

1 1 .
 yy   n  n  1 n 1
,  yy  n  n  1
y y n 1
Hence we obtain
u x  u  u ,

1   n  1   n  1
u y  u  n  1 n  u   y n   u y  y n u  u  ,
y  

u xx  u  u  u  u ,

 n  1 2   n  1 2 
and u yy  u  u    2n 

y 2n  y 

n  n  1 2  n  1 2 n( n  1)
 u (n  1) n 1
 u  u  u ,
y y 2n
y 2n
y n1

 n  1 2 n  n  1
 u yy 
y 2n u  2u  u  
y n 1
 u  u  .
Substituting these values in equation (1.43) we get

189
  2 
 n  1 2 u  2u  u   y 2 n  n  1 u  2u  u   
  2n  
 y 

y 2n 2 n 1  n  1
 n  n  1
y n1
 u  u   ny
yn
 u  u 
2
 4  n  1 u  0

 u  0 for n  1 .

 2u
i.e.  0. ... (1.45)

This is the required hyperbolic canonical form. To find its solution, we integrate equation (1.45)
w.r.t.  to get

u
 1   ,


where    is a function of  . Integrating again w.r.t.  , we get

u   1   d   g   .

We write this as

u  x, y   f    g   ,

where f     1   d  .

i.e. u  x, y   f  x  y1 n   g  x  y1 n  , ... (1.46)

where f and g are arbitrary. This is the required general solution of the equation (1.43).
Example 6 : Classify the equation

u xx  2sin xu xy  cos 2 xu yy  cos xu y  0 .

Reduce it to the canonical form and obtain its general solution.

Solution : The given partial differential equation is

u xx  2sin xu xy  cos 2 xxu yy  cos xu y  0 , ... (1.47)

where R  1, S  2 sin x, T   cos 2 x .

We see that

S 2  4 RT  4sin 2 x  4cos 2 x  4  0

190
 The equation (1.47) is hyperbolic. Hence the quadratic equation R 2  S   T  0 becomes

 2  2sin x  cos 2 x  0
It has roots   sin x  1 . Let 1  sin x  1 and 2  sin x  1 . Hence the first order ordinary
equation.

dy dy
 1  x, y   0 and   2  x, y   0 ,
dx dx

dy dy
become  sin x  1  0 and  sin x  1  0 .
dx dx
Solutions of these equations are obtained by integrating
y  x  cos x  C1 , y  cos x  x  C2 .

So that we choose the independent variables  and  such that

  y  cos x  x ,   y  cos x  x .

From this we find  x  sin x  1,  y  1,  xx  cos x,  xy  0,  yy  0 ,

 x  sin x  1,  y  1,  xx  cos x,  xy  0   yy .
Hence we obtain

u x  u 1  sin x   u  sin x  1 ,

2
u xx  u 1  sin x   1  sin x  u  sin x  1  u cos x 

2
  sin x  1 u  sin x  1   sin x  1 u  u cos x ,

u xx  1  sin x  u  2  sin 2 x  1 u   sin x  1 u  cos x  u  u  ,

2 2

 u xy  u 1  sin x   u 2sin x   sin x  1 u

and u y  u  u

u yy  u  u  u  u  u yy  u  2u  u .

Substituting these values in equation (1.47) we get

1  sin x  2 u  2  sin 2  1 u   sin x  1 2 u  cos x  u  u  

191
 2sin x 1  sin x  u  2 sin xu   sin x  1 u  

 cos 2 x u  2u  u   cos x  u  u   0

On simplifying we obtain
u  0

 2u
or  0. ... (1.48)

This is the required hyperbolic canonical form of the given p.d.e. Clearly its solution is

u  x, y   f    g  

or u  x, y   f  y  cos x  x   g  y  cos x  x  . ... (1.49)

This is the required general solution of (1.47).
Example 7 : Reduce the equation

x 2u xx  y 2u yy  0

into canonical form

Solution : Given equation is

x 2u xx  y 2u yy  0 , ... (1.50)

where R  x 2 , S  0, T   y 2 .

We see that S 2  4 RT  4 x 2 y 2  0 .
 The p.d.e. of second order (1.50) is hyperbolic. Now the quadratic equation

R 2  S   T  0
becomes x 2 2  y 2  0 .

y
It has roots  .
x

y y
Let 1  and 2   . Consider the first order ordinary differential equations
x x

dy dy
 1  x, y   0 and   2  x, y   0 .
dx dx

dy y dy y
   0 and  0,
dx x dx x

192
 dy dx dy dx
or   0 and   0.
y x y x
Integrating we get
log y  log x  log C1 and log y  log x  log C2

y
i.e. xy  C1 and  C2 .
x
Hence we choose the independent variables  and  such that

y
  xy and     x  y,  y  x,  xx  0   yy ,
x

y 1 2y 1
x     ,  xx   3 ,  yy  0 ,  yx   2 .
2 , y
x x x x
Hence we obtain

 y 
u x  u y  u   2  , u y  u x  u 1 ,
 x  x

 y  y   y 
u xx  y 2u  yu   2   2  u y  u   2   ,
 x  x   x 

2 2y2 y2  2y 
u xx  y u  2 u  4 u  u  3  ,
x x x 

1
and u yy  x 2u  2u  u .
x2
Hence the given p.d.e. (1.50) becomes

2y
4 y 2u  u  0 .
x

1
or u  u  0 .
2 xy

1
i.e. u  u  0 . ... (1.51)
2
Which is required hyperbolic canonical form.

193
Example 8 : Reduce the equation
4u xx  4u xy  5u yy  0 to canonical form.

Solution : Let 4u xx  4u xy  5u yy  0 , ... (1.52)

where S = – 4, R = 4, T = 5.
Therefore, S 2  4 Rt  16  80  0
Therefore equation (1.52) is elliptic. The quadratic equation R 2  S   T  0 becomes

4 2  4  5  0 .
This has roots

1
  1  2i  .
2

1 1
We choose 1   i and 2   i .
2 2

dy dy
Hence the ordinary differential equations  1  0 and  2  0 become
dx dx

dy  1  dy  1 
   i   0 and  i  0.
dx  2  dx  2 
Integrating we get

y    i  x  C1 and y    i  x  C2 ,
1 1
2  2 

i.e. x  2 y  2ix  C1 ,

and x  2 y  2ix  C2 .

We choose   x  2 y  2ix ,
and   x  2 y  2ix .
Therefore, to obtain real canonical form, we consider the transformation

1 i
      and       .
2 2
   x  2 y and   2x .

u x  u  x  u   x ,

 u x  u  2u  ,

194
 u y  u  y  u   y  u y  2u ,

2
u xx  u  u  2   u  0   u   2   u   2   u   0 

u xx  u  4u  4u  ,

u xy  2u  u , ... (1.53)

u xy  2u  4u ,

and u yy  4u .

Substituting these values in equation (1.52) we get

4u  16u  16u   8u  16u  20u  0 .

 u  u   0 .

This is required elliptic canonical form of the equation.

Example 8 : Reduce the equation

e 2 x u xx  2e x  y u xy  e 2 y u yy  0

into canonical form

Solution : The second order p.d.e. is given by

e 2 x u xx  2e x  y u xy  e 2 y u yy  0 , ... (1.54)

where R  e 2 x , S  2e x  y , T  e 2 y .
We observe that

S 2  4 RT  4e 2 x  y   4e 2 x  y   0 .
 The equation (1.54) is parabolic. Hence the quadratic equation R 2  S   T  0 becomes

i.e. e 2 x  2  2e x  y   e 2 y  0

  e x  e y   0
2

 e x  e y  0 .

   e y  x .
Hence the ordinary differential equation

dy
   x, y   0 ,
dx

195
 dy
becomes  e y x  0 .
dx
We write this equation as

e  y dy  e  x dx  0 .
On integrating we obtain its solution as

e  y  e  x  C1 . ... (1.55)

Now we choose the independent variables  and  such that

  e  x  e  y and  is arbitrary.. ... (1.56)

We choose   e x  e y . ... (1.57)

From these equations we find

 x   e  x ,  y  e  y , x   e  x , y   e  y ,

 xx  e  x ,  yy  e  y , xx  e  x , yy  e  y ,

and  xy  0 ,  xy  0 .
Therefore, we find

u x  u  e  x   u  e  x 

 u x   e  x  u  u  ,

and u y  u e  y  u  e  y   u y  e  y  u  u  .

Now u xx  e  x  u  u   e  x u  e  x   u  e  x   u  e  x   u  e  x  

u xx  e  x  u  u   e 2 x  u  2u  u  ,

u xy  e  ( x  y ) u  u  ,

and u yy  e  y u e  y  u  e  y   u e  y  u e  y   e  y  u  u 

u yy  e 2 y u  2u  u   e  y  u  u  .

Substituting these in given p.d.e. (1.54) we get

196
  
e 2 x e  x  u  u   u  2u  u  2e x  y  e  ( x  y )  u  u  

 
 e 2 y e 2 y u  2u  u  e  y  u  u    0 ,
 

 4u  u  e y  e x   u  e x  e y 

Solving equations (1.56) and (1.57) we find

2 and y 2 .
ex  e 
   
Thus we obtain

4
 u   u   u  .
  2 
2 ... (1.58)

This is the required parabolic canonical form.

Exercise :
1. Reduce the equation
u xx  2u xy  u yy  0

into the canonical form and hence solve it.

2. Reduce the equation

sin 2 xu xx  2 cos xu xy  u yy  0

into canonical form.

3. Find the characteristics of the equation

u xx  2u xy  sin 2 xu yy  u y  0

when it is of hyperbolic form.

4. Reduce the equation to a canonical form

1  x 2  u xx  1  y 2  u yy  xu x  yu y  0 .

197
2. One Dimensional Wave Equation :
1. Vibration of an infinite string (both ends are not fixed)
Result : Obtain DAlembert’s solution of the one dimensional wave equation which describes the
vibrations of an infinite string.
Proof : We know the vibrations of a string is governed by the second order partial differential euqtion
given by

1
y xx  ytt ,   x   , ... (2.1)
c2
where y (x, t) is the deflection of the string.
Since the string is infinite boundaries of the string are not fixed. If f (x) is the initial deflection
(mean position) of the string and g (x) the initial velocity of the string, then the function y (x, t) is
required to satisfy the initial conditions

y  x, 0   f ( x ) , ... (2.2)
(this gives initial position of the string)

and yt  x, 0   g ( x) ,   x   . ... (2.3)

(this gives the initial velocity of the string.)
Thus our problem is to find the solution of the one-dimensional wave equation (2.1) satisfying
the initial conditions (2.2) and (2.3). We first reduce the equation (2.1) into canonical form by changing
the independent variables (x, t) into the new independent variables (Characteristic variables)  and 
by using the transformation equations
  x  ct , ... (2.4)
and   x  ct , ... (2.5)

where  x  1,  t  c,  xx  0,  xt  0,  tt  1 ,

 x  1, t  c,  xx  0,  xt  0, tt  0 .
Also by using the chain rule of partial differentiation, we find,
y x  y  y  ,

yt  c  y  y   ,

 y xx  y  2 y  y ,

and ytt  c 2  y  2 y  y  ,

Substituting these values in euqtion (2.1) we get

y  2 y  y  y  2 y  y

198
  y  0 . ... (2.6)
This is the required canonical form of the equation (2.1).
Now integrating equation (2.6) we obtain

y  x, t   F    G   .

Replacing  and  as defined in (2.4) and (2.5) we get

y  x, t   F  x  ct   G  x  ct  , ... (2.7)
where F and G are arbitrary functions. Equation (2.7) is the general solution of the one dimensional
wave equation. The two terms in equation (2.7) can be interpreted as waves travelling to the right and
left respectively with velocity c.
The solution (2.7) is required to staisfy the initial conditions (2.2) and (2.3). Hence we have

y  x, 0   f  x   F  x   G  x  . ... (2.8)
Now differentiating equation (2.7) with respect to t we get

yt  x, t   cF '  x  ct   cG '  x  ct 

 yt  x, 0   g  x   cF '  x   cG '  x  . ... (2.9)

On integrating equation (2.9) between x0 to x we get
x
1
g  x  dx   F ( x)  G ( x) .
c x ... (2.10)
0

Adding and subtracting equations (2.8) and (2.10) we get respectively

1  
x
G ( x)   cf ( x)   g ( x)dx  ,
2c  
 x0 

1  
x

and F ( x)   cf ( x)   g ( x)dx  .
2c  
 x0 
Substituting these values in equation (2.7) we get

x ct x  ct
1    1  
y  x, t   cf x  ct    g ( s)ds    cf  x  ct    g ( s)ds  ,
2c   2c  
 x0 x0

1  1  x0 x  ct 
 y  x, t    f x  ct   f  x  ct     g ( s )ds   g ( s )ds  ,
2 2c  x ct 
 x0

199
 x  ct
1 1
 y  x, t    f  x  ct   f  x  ct  
2c x ct
g ( s )ds . ... (2.11)
2

where f  c 2 , g  C 1 so that y  x, t   C 2 function. This is called the d’Alembert’s solution which

describes the vibrations of an infinite string at any point x and at any time t.
Note : x – ct = constant and x + ct = constant are called the characteristic curves of one dimensional
wave equation.
Note : If the string is released from rest then g (x) = 0, so that the solution (2.11) becomes

1 
y  x, t    f x  ct   f  x  ct   .
2
Physical Meaning of the solution of the wave equation
We know the general solution of one dimensional wave equation (2.1) is given by

y  x, t   F  x  ct   G  x  ct  . ... (2.12)

Consider u1  x, t   F  x  ct  . ... (2.13)

Hence the initial wave profile (shape) is given by

u1  x, 0   F  x  .

1
Now at time t  , we have from (2.13) that
c

u1  x,   F  x  1 .
1
 c

1
 In time t  , the wave has travelled through a distance of 1-unit. Further, if we put x '  x  1 ,
c
then we have

F  x  1  F  x ' .
This implies that the original shape of the wave is retained even if the origin is shifted by one unit along
the x-axis.

2
Now at time t  , we have from equation (2.13)
c

u1  x,   F  x  2  ,
2
 c
2
 the wave has travelled through a distance of 2 units at time t  .
c
Thus in particular, at t = 1, we have from equation (2.13),
200
 u1  x,1  F  x  c 

 in one unit of time the profile has moved c units to the right.
 c is the speed of propagation.
Similarly, we conclude that the equation

u 2  x, t   G  x  ct 
represents a wave profile travelling to the left with speed c along x-axis. Thus the general solution
(2.12) of the one dimensional wave equation represents the superposition of two arbitrary wave profiles,
both of which are travelling with a common speed but in the opposite direction along the x-axis.

3. Vibrations of a Semi-infinite String (one end point is fixed)

Result : Obtain d’Alembert’s solution of the one dimensional wave equation which describes the
vibrations of a semi-infinite string.
Proof : The vibration of a string is governed by the second order one dimensional wave equation

1
y xx  ytt , 0  x   , t  0 , ... (3.1)
c2
where y (x, t) represents the deflection of the string at any point x and at any time t. Since the string is
semi-infinite i.e. one end of the string x = 0 is kept fixed for all time. If u (x) and v (x) are the initial
deflection and the initial velocity of the string, then the function y (x, t) is required to satisfy the initial
conditions.

y  x, 0   u  x  , x  0 . ... (3.2)
This equation describes initial position of the string and

yt  x, 0   v  x  . ... (3.3)
This describes initial velocity at point x. The deflection y (x, t) has to satisfy the boundary conditions

y  0, t   0 , t  0 . ... (3.4)
This shows there is no deflection at fixed point x = 0 at any time t,

and yt  0, t   0 , ... (3.5)

showing that velocity at fixed point x = 0 is zero.
Our aim is to find the solution of equation (3.1) satisfying the conditions (3.2) to (3.5).
We know the d Alembert’s solution of one dimensional wave equation is given by
x  ct
1 1
y  x, t   u  x  ct   u  x  ct  
2c x ct
v( x)ds . ... (3.6)
2

201
However, this solution cannot be used for the given initial value problem, since u  x  ct  has no
x
meaning for values t  . Therefore we modify our semi-infinite string problem to an infinite string
c
problem. Thus our problem is to find deflection of an infinite string subject to the initial conditions.

y  x, 0   U  x 

yt  x , 0   V  x  ,   x   ,

u ( x) if x  0,
where U  x   ... (3.7)
u ( x) if x  0,

v( x) if x  0,
and V  x   ... (3.8)
 v( x) if x  0.
We notice that U and V are odd functions of x. Thus the solution of equation (3.1) subject to the
conditions (3.7) and (3.8) is given by d Alembert’s solution
x  ct
1 1
y  x, t   U  x  ct   U  x  ct  
2c x ct
V ( s )ds ... (3.9)
2

Now we will show that the solution (3.9) is also a solution of equation (3.1) subject to the conditions
(3.2) to (3.5). For this, we simply prove that the solution (3.9) satisfies the initial and boundary conditions
(3.2) to (3.5). Therefore, from equation (3.9) we have at t = 0 and x > 0.

x
1 1
y  x, 0   U  x   U  x    V ( s )ds , x  0
2 2c x

 y ( x, 0)  u ( x ) , x > 0, due to (3.7). ... (3.10)

Now from equation (3.9) we find after differentiating (3.9) with respect to t.

x  ct
1 1 
yt  x, t   cU '  x  ct   cU '  x  ct    V ( s)ds .
2 2c t x  ct

We use the formula

x  ct x  ct
     
 V  s  ds   V s ds  V  x  ct   x  ct   V  c  ct   x  ct  .
t x  ct x  ct
t t t

202
 1 1
 yt  x, t   cU '  x  ct   cU '  x  ct    cV  x  ct   cV  x  ct  . ... (3.11)
2 2c
At t = 0, this gives

1
yt  x, 0    cU '( x)  cU '( x)  1 V ( x)  V ( x) 
2 2

yt  x, 0   V ( x ) . for x > 0 ... (3.12)

Now from equation (3.9) we have at x = 0
ct
1  
y  0, t   U ct  U (ct )  1  V ( s)ds .
2 2c  ct

ct

Since V is an odd function, this implies the integral  V (s)ds vanishes.

 ct

1
 y  0, t   U (ct )  U (ct )  .
2
Using equation (3.7) we get

1
y  0, t    u (ct )  u (ct ) ,
2

 y  0, t   0 . ... (3.13)
Now from equation (3.11) we find for x = 0, and t > 0

1
yt  0, t    cU '(ct )  cU '(ct )  1 V (ct )  V (ct ) .
2 2
Using equations (3.7) and (3.8) we find for x  0
U '( x)  u '( x)(1) ,
 U '( x)  u '(  x ) ,
 U '( ct )  u '(ct ) .
Similarly, for x  0 , U '( x )  u '( x )
 U '(ct )  u '(ct ) .
Hence we get

1
yt  0, t    cu '(ct )  cu '(ct )  1 v(ct )  v(ct ) .
2 2

 yt  0, t   0 . ... (3.14)

203
Thus we have proved that the d’Alembert’s solution (3.9) also satisfies the initial and the boundary
conditions (3.2) to (3.5). This proves the D’Alembert’s solution (3.9) is the desired solution of the one
dimensional wave equation (3.1) subject to the conditions (3.2) to (3.5).
Note : In particular, if the string is released from rest i.e. v(x) = 0 then the solution (3.9) becomes

1 
y  x, t   u x  ct   u  x  ct   , for x  ct ,
2

1 1 
 u x  ct   u  ct  x   , for x  ct .

2 2 

4. Vibrations of a string of finite length

Result : Show that the d’Alembert’s solution of the one domensional wave equation which describes
the vibrations of a finite length string is given by
 
y  x, t    un sin 
n x   n ct    vn
sin 
n x   n ct 
  
 cos   
    c n1 n
 sin 
     
.
n 1

Proof : Let a string be of length  . The vibrations of a string is goverened by the second order partial
differential equation given by

y xx  c 2 ytt , 0  x   , t > 0. ... (4.1)

Since the string is finite, hence both the ends of the string are fixed for all time. Therefore the function
y (x, t) must satisfy the initial conditions

y  x, 0   u ( x ) , ... (4.2)

yt  x, 0   v ( x ) , 0  x   ,
where u (x) represents the initial position of the string and v (x) represents the initial velocity of the
string. The deflection of the string y (x, t) also satisfy the boundary conditions.

y  0, t   y  , t   0 , ... (4.3)

 there is no deflection at the end points of the string at any time t > 0,
and yt  0, t   yt  , t   0 . ... (4.4)
This shows the velocity of the string at end points at any time t is zero.
The d’Alembert’s solution of equation (4.1) is given by
x  ct
1
y  x, t   u ( x  ct )  u ( x  ct )  1  v(s)ds . ... (4.5)
2 2c x ct

However, this solution cannot be used for given initial value problem as u (x – ct) has no meaning for
204
 x
values t  . Hence we convert our problem into a problem of vibrations of an infinite string by
c
extending our data.
Thus we consider the vibrations of an infinite string subject to the initial conditions.

y  x, 0   U ( x ) ,

and yt  x, 0   V ( x ) , ... (4.6)

u ( x), if 0  x  
where U ( x)   ... (4.7)
u ( x), if    x  0

and U  x  2r    U ( x) , if   x   , r  1, 2,... ,

v( x), if 0  x  
and V ( x)   ... (4.8)
v( x), if    x  0

and V  x  2r    V ( x) ,   x   , r  1, 2,...

This shows that U (x) and V (x) are odd functions of x and are periodic with period 2  .
Hence the deflection of the string given in (4.5) subject to the conditions (4.6) to (4.8) becomes
x  ct
1 1
y  x, t   U  x  ct   U  x  ct  
2c x ct
V ( s )ds . ... (4.9)
2

We assume U (x) and V (x) can be expanded into a Fourier series in  ,   . Since U (x) and V (x)
are odd functions, it contains only sine terms.
Thus we have,

n x 
U ( x)   un sin  , ... (4.10)
n 1   


n x 
and V ( x)   vn sin  , ... (4.11)
n 1   

where the Fourier constants u n and vn are given by


n s 
u n   u ( s )sin 
2
ds
0    , ... (4.12)


n s 
v ( s ) sin 
2
 0
and vn  ds
   . ... (4.13)

205
Using equation (4.10) we find

1
U ( x  ct )  U ( x  ct )  1  un sin  n ( x  ct )   sin  n ( x  ct )   .
2 2 n 1       

A B   A B 
Since sin A  sin B  2sin    cos  
 2   2 


1
U ( x  ct )  U ( x  ct )   un sin  n x   cos  n ct  . ... (4.14)
2 n 1      

Similarly, on using (4.11) we find

x  ct x  ct 
n
 vn sin   s  ds
1 1
2c x ct 2c x ct n 1
V ( s ) ds 


 x  ct
n
vn  sin  s  ds
1
 
2c n 1 x ct   

   x  ct 
1  n 
  vn   cos  s 
2c n 1  n    x ct 


 v  n n 
  n
 cos  ( x  ct   cos  ( x  ct   .
2 c n  
n 1     

Using the formula

A B   B A
cos A  cos B  2sin    sin  
 2   2 

x  ct 
 vn   n n  
x   sin 
1
we find, 
2c x ct
V ( s )ds  
2 c n 1 n sin   
ct 
    ... (4.15)

Using equations (4.14) and (4.15) in equation (4.9), we readily obtain

 
n  n ct   vn n   n ct 
y  x, t    un sin  x   cos    sin  x  sin  . ... (4.16)
n 1        c n 1 n      

Differentiating equation (4.16) with respect to t we get

 
c n   n ct       c   n x  cos  n ct 
yt ( x, t )    nun sin  x  sin         vn sin    .
 n 1         c     n 1      

206
  
c n n ct   n n ct 
 yt ( x , t )    nun sin  x  sin     vn sin  x  cos   . ... (4.17)
 n 1       n 1      

Now we easily check that the d’Alembert’s solution (4.16) satisfies the initial and boundary conditions.
Thus from equation (4.16) we find at t = 0

n 
y  x, 0    un sin  x   U ( x) , by (4.10) ... (4.18)
n 1   

and y  0, t   0  y  , t  . as sin 0  sin n  0

Now from equation (4.17), we find at t = 0

n
yt  x, 0    vn sin  x   V ( x) , due to (4.11)
n 1   

and yt  0, t   0  yt  , t  . ... (4.19)

Thus we see from equations (4.18) and (4.19) that the d’Alembert’s solution (4.16) satisfies the initial
and the boundary conditions identically. Hence equation (4.16) is the desired solution of the one-
dimensional wave equation (4.1).

5. Vibrations of a string of finite length (Method of Separation of Variables)

Introduction : Among the many methods that are available for the solutions of a second order partial
differential equation, the method of separation of variables is a powerful method which is applicable in
certain circumstances. We will apply the method to find the solutions of one dimensional wave equation.
The method will also be used to solve Heat and Laplace equations in the Units 6 and 7 below :
Result : By separable variable method find the solution of

ytt  c 2 y xx , 0  x   , t > 0. ... (5.1)

subject to the conditions that

y  x, 0   f ( x), 0  x  
... (5.2)
yt  x, 0   g ( x), 0  x  
and y  0, t   y  , t   0 , ... (5.3)
where f (x) and g (x) are initial displacement and velocity of the string.
Proof : We assume the method of separation of variables to find the vibration in a string which is
governed by the equation (5.1).
Therefore, let y  x, t   X ( x)T (t ) ... (5.4)
be the solution of equation (5.1)

207
  y x  X '( x)T (t ) , y xx  X ''( x)T (t ) .

Similarly, ytt  X ( x)T ''(t ) .

Substituting this in equation (5.1) we get

XT ''  c 2 X '' T ,
X '' T ''
  2 . ... (5.5)
X cT
We see that the left handside is a function of x and the right hand side is a function t alone.
Equation (5.5) shows that each side must be a constant say  .

 X ''   X and T ''  c 2T

or X ''  X  0 and T '' c 2T  0 , ... (5.6)

where  may be zero, positive or negative. From the boundary condition

y  0, t   0  X (0)T (t )  0
 X (0)  0 . as T (t )  0

Similarly, y  , t   0  X     0 .
Thus our problem is
X ''  X  0 ... (5.7)

such that X (0)  X     0 .

Case (i) :   0
The solution of the equation (5.7) in this case is
X  Ax  B ... (5.8)
The boundary conditions X (0)  0 and X     0 give

A  0 and B  0 .
Consequently, we get X (x) = 0 as the solution of equation (5.7). This is a trivial solution hence we
drop it.
Case (ii) : Let   0
Let    2 , where  is positive or negative. In this case the solution of equation (5.7) is given by

X ( x)  A  e  x  B e x . ... (5.9)

To determine the constants A and B, we use the boundary conditions

208
 X (0)  0  A  B  0 ,

and X ()  0  Ae    Be   0

 A  Be 2   0 .
or B 1  e 2    0
 B 0 A0
Hence for   0 and   0 , the solutions (5.8) and (5.9) do not constitute the solution of the wave
equation (5.1).
Case (iii) : Let   0
Let    2
In this case the solution of equation (5.7) is given by
X ( x )  A cos  x  B sin  x . ... (5.10)
Now the boundary conditions
X (0)  0  0  A
 A0
and X     0  0  B sin   
Now if B = 0 then we have y  x, t   0 is again a trivial solution of equation (5.1).
Therefore we assume B  0

 sin     0
    n , n = 1, 2, 3, ...,
n
or  , n = 1, 2, ...,

For each value of n = 1, 2, ..., let    n .

n
Thus n  , n = 1, 2, ... ... (5.11)


 n x 
These  n are called eigen values of the equation (5.1) and the corresponding functions sin  
  
are called eigen functions. Hence the solution (5.10) can be denoted by

n 
X n  Bn sin  x  , n = 1, 2, ..... ... (5.12)
  

209
 n
Similarly, for  n  , the solution of other equation, T '' c 2 2T  0 is given by


n x  n c 
Tn (t )  Cn cos  t   Dn sin  t, ... (5.13)
     
where Cn and Dn are arbitrary constants. Hence, the solution (5.4) becomes

 n ct   n ct   sin  n 
y n  x, t    an cos    bn sin    , ... (5.14)
          

where an  Cn Bn and bn  Dn Bn .
By the principle of superposition, the series
 
 n ct   n ct   sin  n 
y  x, t    yn  x, t     an cos    bn sin     ... (5.15)
n 1 n 1           

if it converges, is also a solution of equation (5.1) satisfying the boundary conditions (5.3). We choose
an and bn such that y (x, t) in (5.15) satisfies the initial conditions (5.2).
Therefore the initial condition y  x, 0   f ( x) gives


n 
f ( x)   an sin   , 0 x ... (5.16)
n 1   

Now differentiating equation (5.15) with respect to t we get


c  n  n ct    n
yt  x, t  

  nan sin  ct   nbn cos 
 
 sin 
     
x  ,

... (5.17)
n 1

Thus at t = 0, we have
yt ( x, 0)  g ( x) ,


n n   n 
 g ( x)   bn   sin  x,
      0 x
. ... (5.18)
n 1

Equations (5.16) and (5.18) show that f (x) and g (x) are expanded in a half range sine series.
Therefore an and bn are coefficients of the half range sine series of f (x) and g (x) respectively.

n x 
f ( x) sin 
2
 an 
0    ,
 dx ... (5.19)


2  n x  dx
n c 
and bn  g ( x ) sin  
0
   . ... (5.20)

210
Thus the solution of one dimensional wave equation (5.1) subject to the conditions (5.2) and (5.3) is
given by the equation (5.15) with the coefficients an and bn given in equations (5.19) and (5.20)
respectively.
Note : When initial velocity of the string is zero. i.e. if g (x) = 0, then we have bn = 0. In this case the
solution (5.15) becomes

n ct   n x 
y  x, y    an cos    sin  , ... (5.21)
n 1      


n x 
f ( x ) sin 
2
with an 
0    .
dx ... (5.22)

Example 1 : A tightly stretched string with fixed end points x = 0 and x = 1 is initially in a position
given by

y  x, 0   x 1  x 
It is released from rest from this position. Find the displacement y (x, t) at any time t.
Solution : We know the vibration of a string is governed by the second order p.d.e. given by

ytt  c 2 y xx  0 , 0  x  1 , t  0 , ... (5.23)

subject to the initial conditions

y  0, t   y 1, t   0 , t ... (5.24)

and y  x, 0   x 1  x  . ... (5.25)

Also the initial velocity of the string is given by

yt  x, 0   0 . ... (5.26)
By variable separable method, the solution of (5.23) is given by

y  x, t    A cos  x   B sin  x    C cos  ct   D sin  ct   . ... (5.27)

The boundary conditions

y  0, t   0  0  A . for C  0, D  0
Also from (5.27) we have

yt  x, t    c  A cos  x   B sin  x   C sin  ct   D cos  ct 

Therefore, the condition yt  x, 0   0  D  0 .

Therefore, equation (5.27) becomes

y  x, t   B sin  x  C cos  ct  . ... (5.28)

Now the condition y 1, t   0 gives

211
 0  B sin    C cos  ct 
 B  0 or sin   0 for C  0
If B = 0 then we have only trivial solution of (5.23).
Therefore, we assume B  0
 sin   0    n , n = 1, 2, ......
Let  n  n , for n = 1, 2, 3, .....
Therefore, corresponding to each n, the solution (5.28) becomes

yn  x, t   an sin  n x  cos  n ct  , for an  BnCn . ... (5.29)

By superposition principle, the most general solution of equation (5.23) is given by
 
y  x, t    yn  x, t    an sin  n x  cos  n ct  , ... (5.30)
n 1 n 1

where the constant an is determined by using the condition

y  x, 0   x 1  x  .
Therefore, from (5.30) we have

y  x, 0   x 1  x    an sin  n x  . 0  x 1 ... (5.31)
n 1

We see from equation (5.31) that f ( x )  x (1  x ) is expressed in the Fourier sine series. Hence the
corresponding Fourier constant an is given by
1
an  2  x (1  x) sin  n x  dx
0

1 1 
 2   x sin  n x  dx   x 2 sin  n x  dx  .
 0 0 

Integrating by parts we get

 1 1 2
1 1 
x 1 x 2x
an  2   cos  n x    cos  n x  dx  cos  n x    cos  n x  dx 
 n 0 n n n 
 0 0 0 

4 cos  n x  1 
an  2 2   
n  n 0

212
 4   n 
an   3 3  1  1
n

8
an  3 3 , for n is odd,
n
an = 0 for n even.
Substituting this in (5.30) we get

8 1
y  x, t  
 3  n3 sin  n x   cos  n ct  . ... (5.32)
n 1

Example 2 : Solve

ytt  c 2 y xx  0 , o  x  1, t  0 , ... (5.33)

y  0, t   y 1, t   0 , ... (5.34)

y  x,0   0 , 0  x  1, ... (5.35)

yt  x,0   x 2 , 0  x  1. ... (5.36)

Solution : Let y  x, t   X ( x)T (t ) ...(5.37)

be a solution of equation (5.33) and is given by

y  x, t    A cos  x   B sin  x    C cos  ct   D sin  ct   . ... (5.38)

The boundary condition y  0, t   0  A  0 for C  0, D  0 .

Also the condition

y  x, 0   0  0  B sin  x  C   C  0 for B  0 .
Therefore, the solution (5.38) implies

y  x, t   B sin  x   D sin  ct   . ... (5.39)

Now the condition

y 1, t   0  0  B sin    D sin  ct 

 B  0 or sin   0 for D  0
If B = 0 we have trivial solution of equation (5.33).
Therefore we assume
B  0  sin   0    n , n = 1, 2, ....
Let  n  n , n = 1, 2, ...
Therefore, corresponding to each n the solution (5.39) becomes
213
 y n  x, t   an sin  n x  sin  n ct  . ... (5.40)
Therefore, by superposition principle, the most general solution of (5.33) is given by
 
y  x, t    yn  x, t    an sin  n x  sin  n ct  , ... (5.41)
n 1 n 1

where the constant an is to be determined.

From equation (5.41) we have

yt  x, t    an sin  n x    n c   cos  n ct  . ... (5.42)
n 1

Therefore, the condition yt  x,0   x 2 gives


x 2   c  an n sin  n x  , 0  x  1. ... (5.43)
n 1

This shows that x2 is expressed in the Fourier sine series. Hence the corresponding Fourier constant
nan is given by

1
 cnan  2  x 2 sin  n x dx .
0

Integrating by parts we get

 x2 1
1
1 
 cnan  2  cos  n x    
n 0
2 x cos n x dx ,
 n 0 

 1 2  x 1 1
1 
 2  cos  n    sin  n x    sin  n x  dx  ,
 n n  n 0 0 n 

 1 n 2  cos  n x   
1
 2   1     ,
 n  n  2  n  0 

2   n 
 2   1 
1 n
1  1 ,
 n  n 3 
 
n
2  1
 3 3   1  1 ,
4 n
 cnan 
n n
n
2  1
 4 4   1  1 .
4 n
an  2 2 ... (5.44)
n  c cn 

214
Substituting this in (5.41) we get
  
2(1) n
y  x, t     2 2  4 4  (1) n  1  sin  n x  sin  n ct  .
4
... (5.45)
n 1  n  c n c 
This is the required solution.

Uniqueness of Solution of Wave Equation :

Theorem : Show that the solution u (x, t) of the equation

utt  c 2u xx  F  x, t  , 0  x  , t  0 ,

u  x, 0   f ( x), 0  x   ,

ut  x, 0   g ( x ), 0  x   ,

u  0, t   u (, t )  0, t  0 ,
if it exists, is unique.

Proof : Let there be two solutions u1  x, t  and u 2  x, t  of the equation

utt  c 2u xx  F  x, t  , 0  x   , t > 0, ... (5.46)

satisfing the conditions

u  x, 0   f ( x ) , 0 x, ... (5.47)

ut  x, 0   g ( x ) ,

and u  0, t   u  , t   0 . t0 ... (5.48)

 2u1 2  2u1
 c  F  x, t  ,
t 2 x 2

 2u 2 2
2  u2
and  c  F  x, t 
t 2 x 2
Subtracting these equations we get

 2  u1  u 2  2
 c2  u1  u2   0 .
t 2 x 2

Also u1  x, 0   f ( x ) and u 2  x, 0   f ( x )

  u1  u 2   x, 0   0

215
 u1 u
Also  x, 0   g ( x) and 2  x, 0   g ( x)
t t


  u1  u2   x, 0   0 .
t
This shows that the function v  u1  u2 satisfies the corresponding partial differential equation

 2v 2  2v
c 0, 0 x,t>0 ... (5.49)
t 2 x 2
subject to the conditions

v  x, 0   0 , o x, ... (5.50)

vt  x, 0   0 , o x, ... (5.51)

and v  0, t   v  , t   0 , t  0.

Claim : We prove v = 0 i.e. u1  x, t   u 2  x, t  .

Therefore, consider

E t  
1
20  c 2 v x2  vt2  dx ... (5.52)

Since v (x, t) is twice differentiable, we see that E (t) is a differentiable function of t.


  2  c 2 v x v xt  vt vtt  dx
dE 1
Hence dt 2 0

 
dE
  vt vtt dx   c 2v x v xt dx .
dt 0 0

Integrating the second integral by parts we get

 

  vt vtt dx   c v x vt  0   c 2 vt v xx dx .
2

0 0

However, from equation (5.51) we have

v  0, t   0  vt  0, t   0 t , t  0 ,

and v  , t   0  v t   , t   0  t  0
Hence we have

216
 
 vt  vtt  c 2 v xx dx .
dE
dt 0

From equation (5.49) we have

vtt  c 2vxx  0 .

dE
 0.
dt
 E  Constant . ... (5.53)

However, v  x, 0   0  v x  x, 0   0 and vt  x, 0   0 , therefore, from equation (5.52) we have


E (0)    c 2v x2  x, 0   vt2  x, 0   dx
1
20 ... (5.54)

E (0)  0 ,
Therefore E  0.
Hence from equation (5.52) we have

v x  x, t   0, vt  x, t   0  t  0, 0  x   .
This is possible only if v (x, t) = constant. The condition (5.50) gives

v  x, 0   0  constant  0

 v  x, t   0

 u1  x, t   u 2  x, t  ,
which proves the uniqueness of the solution of the wave equation.
Remarks : The solution of the problem of vibrations of a string of finite length is also unique, as it is a
special case of the problem when F  x, t   0 .
Example 3 : A tightly stretched string with fixed end points x = 0 and x =  is initially in a position
given by

y  x, 0   y0 sin 3 
x
.
  
It is released from rest from this position find the displacement y(x, t) at any time t.
Solution : We know vibrations in a string are governed by the second order p.d.e. given by

2 y 2
2  y
 c , 0  x   , t > 0, ... (5.55)
t 2 x 2

217
such that y  0, t   0  y  , t  , t , ... (5.56)

y  x, 0   y0 sin 3 
x
and . ... (5.57)
  
It is also given that the initial velocity of the string is zero.

 yt  x , 0   0 . ... (5.58)
We know by separable variable method the solution of equation (5.55) is given by

y  x, t    A cos  x   B sin  x   C cos  ct   D sin  ct   . ... (5.59)

The boundary condition

y  0, t   0  A  0 .
Also from (5.59) we find

yt  x, t    c  A cos  x   B sin  x   C sin  ct   D cos  ct 

Hence, yt  x, 0   0  D  0 .
Hence solution (5.59) becomes

y  x, t   BC sin  x   cos  ct  .
Now the condition

y  , t   0  0  B sin     C cos  ct  ... (5.60)

 B 0, for C  0 ,

or sin     0 , for C  0 ,

If B = 0 we have trivial solution y (x, t) = 0. Therefore, we assume B  0  sin     0 ,

    n , n = 1, 2, 3, .....

n 
Let  n    , n = 1, 2, 3, .... ... (5.61)
  
These are called the eigen values of the equation. Hence the solution (5.60) becomes

n x   n ct 
yn  x, t   an sin    cos  . ... (5.62)
     

By the superposition principle, the most general solution of (5.55) is given by

218
  
n x   n ct 
y  x, t    yn  x, t    an sin    cos  , ... (5.63)
n 1 n 1      

where the constant an is determined by using the condition that

x
y  x, 0   y0 sin 3  .
  
 From equations (5.57) and (5.63) we have

x  n x 
y  x, 0   y0 sin 3     an sin  . ... (5.64)
   n 1   

We know sin 3  3sin   4sin 3 

3sin 
 x   sin  3 x 
  
 sin 3 
x      .

   4
Therefore, equation (5.64) becomes

 x  3 x  
 3sin     sin       n x  .
y0     an sin  
 4  n1   

Comparing corresponding coefficients on both sides we get

3 1
a1  y0 , a2  0, a3   y0 , a4  0....
4 4
Therefore, the solution (5.63) becomes

 x    ct  y0  3 x   3 ct 
y0 sin 
3
y  x, t    cos    sin   cos  .
4       4      

y  x, t  
y0   x    ct   3 x   3 ct  
or
4 3sin     cos     sin    cos     . ... (5.65)

Example 4 : By separating the variables, show that one dimensional wave equation

2 y 1 2 y

x 2 c 2 t 2

has solution solution of the form A exp   in x  in ct  , where A and n are constants. Hence show
that the functions of the form

219
  r ct   r ct   sin  r x 
y  x, t     Ar cos    Br    
r   a   a    a 

where Ar and Br are constants, satisfy the wave equation and the boundary conditions.

y  0, t   0, y  a, t   0 t .

Solution : One dimension wave equation is given by

2 y 1 2 y
 , 0 xa ... (5.66)
x 2 c 2 t 2
where the deflection y (x, t) satisfies the conditions

y  0, t   0  y  a, t  t ... (5.67)

Let y  x, t   X ( x)T (t ) ... (5.68)

be its solution. Therefore, we have
y xx  X ''( x)T (t ) and ytt  X ( x)T ''(t ) .
Therefore, equation (5.66) becomes

1
X '' T  X ( x )T ''
c2

X '' T ''
or  2   n 2 (say) ... (5.69)
X c T

 X '' n 2 X  0 , ... (5.70)

and T '' c 2 n 2T  0 . ... (5.71)

Solving equations (5.70) and (5.71) we have

X  e  inx ,

T  e  inct .
Hence y  x, t   A  e  inx inct , ... (5.72)
is a solution of equation (5.73), where A = constant. We can also write the solution of (5.70) and
(5.71) as

X  A cos  nx   B sin  nx  and T  C cos  nct   D sin  nct  .

Therefore, the solution of (5.66) is given by

y  x, t    A cos(nx )  B sin(nx) C cos( nct )  D sin(nct )  . ... (5.73)

220
Now applying the initial condition

y  0, t   0  0  A ,

and y  a , t   0  0  B sin( an) C cos( nct )  D sin(nct ) 

 B  0 or sin( an )  0 for C  0 , D  0 as T (t )  0
If B = 0, we have trivial solution. Therefore, we assume B  0  sin( an)  0

 an  r , r = 1, 2, .....

r
or n    , r = 1, 2, .....
 a 
Therefore, solution for each value of r, we have

r x     rct   D sin   rct  

yr  x, t   Br sin    C cos     . ... (5.74)
 a    a   a  

By superposition principle, the most general solution is given by

 
  rct    rct   sin   rx 
y  x, t    yr  x, t     Ar cos    Br sin     , ... (5.75)
r 1 r 1   a   a    a 

where Ar and Br are constants.

Exercise :
1. Obtain the solution of the wave equation

utt  c 2u xx
under the following conditions

u  0, t   u  2, t   0 ,

 
u  x, 0   sin 3  x ,
2 

ut  x, 0   0 .

221
2. The vibrations of an elastic string is governed by the partial differential equation
utt  u xx
The length of the string is  and the ends are fixed. The initial velocity is zero and initial
deflection is u  x, 0   2  sin x  sin 3x  . Find the deflection u (x, t) of the vibrating string for
t >0.
3. A string is fixed at two points  apart and is stretched. The motion takes place by displacing
x
the string in the form y  a sin   from which it is released at time t. Show that the
  
displacement of any point at a distance x from one end at time t is

 x    ct 
y  x, t   a sin   cos  .
     

222
 UNIT - VI

HEAT CONDUCTION PROBLEM

In this unit we consider heat conduction problem in a rod with the following assumptions.
1. The rod is homogeneous.
2. It is sufficiently thin so that the heat is uniformly distributed over it cross section at a given time t.
3. The surface of the rod is insulated to prevent any loss of heat through the boundary.
4. u ( x, t ) is the temperature at the point x at time t.
We know the temperature u ( x, t ) in a rod is governed by the second order one dimensional.
p.d..e.

u  2u
k ,
t x 2
satisfying some initial and boundary conditions.
Case 1 : Heat conduction - Finite rod.
Result : By separable variable method, find the temperature distribution in a rod of length  satisfying
the boundary conditions

u  0, t   u  , t   0 , t  0 ( end points of the rod are kept at 0° C.)

The initial temperature is u ( x, 0)  f ( x ) , 0  x  .
The rod and its ends are perfectly insulated.
Or

u  2u
By separable variable method, find the solution of the equation k , 0  x  , t  0
t x 2
satisfying the conditions
u (0, t )  u (, t )  0 , t0

u ( x, 0)  f ( x ). 0 x
Solution : Let u ( x, t ) be the temperature in a rod of length  . We know the temperature distribution
in rod is governed by the second order partial differential equation given by

223
 u  2u
k , 0 x, t 0 ... (1.1)
t x 2
satisfying the boundary conditions
u (0, t )  u (, t )  0 , t 0 ... (1.2)
and u ( x, 0)  f ( x ) , 0 x, ... (1.3)
where f ( x ) is the initial temperature in the rod.
To find temperature in the rod at any instant t, by separable variable method, we assume the solution of
the equation (1.1) in the form
u ( x , t )  X ( x )  (t ) ... (1.4)

 ut ( x, t )  X ( x)' (t ) ,

and u x ( x, t )  X ' ( x)(t ) ,

 u xx ( x, t )  X '' ( x)(t ) .
Hence equation ( 1.1 ) becomes

X ( x)' (t )  kX '' ( x )(t )

X '' '
   (say), ... (1.5)
X k
where  is a constant may be zero, positive or negative

 X" X  0, ... (1.6)

and '   k   0 . ... (1.7)

Case (i ) If   0 , then solutions of (1.6) and ( 1.7) are given by
X  Ax  B, T  
The conditions (1.2)  X (0)  X ()  0  A  0  B.
Consequently, u ( x, t )  0 , which is a trivial solution of equation (1.1).

Case ( ii ) If   0 say    2
Therefore, solutions of equations (1.6) and (1.7) are
2kt
X ( x)  Ae x  Be  x and (t )  Ce .

Now the conditions ( 1.2 )  X (0)  X ()  0 ,

224
  A B  0 
and     A  B  0.
Ae  Be  0 

 u ( x, t )  0 .
Thus for   0 and   0 we have trivial solutions. Therefore, we assume

Case ( iii ) :   0 say    2 ,   0.

Therfore, equations ( 1.6 ) and ( 1.7 ) become

X "   2 X  0 and '   2 k   0 ,

which have solutions
X ( x )  A cos  x  B sin  x , ... (1.8)
2kt
and   Ce  a . ... (1.9)

Therefore, solution (1.4) becomes

2kt
u  x, t   ( A cos  x  B sin  x)C.e  . ... (1.10)

The boundary conditions (1.2) viz.

u  0, t   u  , t   0  X (0)  0  X () ,

X (0)  0  A  0 ,

and X     0  B sin ( )=0 .

If B  0 yields only trivial solutions. Therefore we assume B  0

 sin    0 ,
 (  )  n , n  1, 2, 3,....
Let for each value of n=1,2,....

n
n  , n  1, 2,... ... (1.11)

These are called eigen values of the differential equation. Hence the solutions of (1.6) and (1.7) are
respectively

 n x 
X n ( x)  Bn sin  , ... (1.12)
  

 n2 2 
n  t   Cn exp  kt  . ... (1.13)
  
 
225
These are called the corresponding eigen functions of the equations. Therefore we write from (1.4)

 n 2 2 kt   n x 
un  x, t   Bn Cn exp   sin 
      .
 

 n2 2 kt   n 
un  x, t   an exp   sin x  , for a = B C
or   2     n n n ... (1.14)
 
Thus by the principle of superposition, we have

u ( x, t )   u n  x , t 
n 1

is also solution of ( 1.1 ).

  n 2 2 kt   n 
 u ( x, t )   an exp   sin x ,
  2    
... (1.15)
n 1  
if it converges, is also a solution of (1.1) satisfying the boundary conditions. That the initial temperature
in the rod is given by
u ( x, 0)  f ( x ) .

  n 
 f ( x)   an sin  x, 0  x   . ... (1.16)
n1   
This is a Fourier series expansion of f (x), where the Fourier constant an is given by

2  n x 
 o
an  f ( x) sin  dx . ... (1.17)
  
Thus equation (1.15) is a solution of the equation (1.1), where the constant an is given in equation
(1.17).
Example 1 : Solve ut  u xx , 0  x  , t  0 ,

u (0, t )  u (, t )  0,
u ( x, 0)  x(, t ), 0 x.
OR

A heat flow in a rod of length 10 cm of homogeneous material is governed by the p.d.e ut  c 2u xx .

The ends of the rod are kept at temperature O 0c and initial temperature is u  x, 0   x 10  x  . Find
the temperature in the rod at any instant.

226
Solution : we are given that
ut  u xx , 0  x  , t  0 , ... (1.18)

satisfying u (0, t )  u (, t )  0 , ... (1.19)

and u ( x, 0)  x(  x )  0, 0 x. ... (1.20)
We assume u ( x, t )  X ( x )T (t ) ... (1.21)

 ut  X ( x)' (t ) and u xx  X "( x)(t )

Hence equation (1.18) becomes

X ( x)' (t )  X " ( x)(t )

X" '
or    ( say ) .
X 
We have if  is zero or positive, the equation (1.18) has trivial solution. Therefore, we assume  is
negative. We choose

   2
 X " 2X  0 , ... (1.22)
and '   2  0 . ... (1.23)
The solutions of ( 1.22 )and ( 1.24 ) are respectively given by
X ( x )  A cos  x   sin  x , ... (1.24)

and (t )  C exp( 2t ) . ... (1.25)

The boundary conditions (1.19) give
X (0)  0  X () .
From equation (1.24) we have for X (0)  0

 A0,
and X ()  0  0  B sin   .
If B  0 we have only trivial solution of (1.18). Hence we assume B  0 . In this case, we have
sin(  )  0 ,

    n for n  1, 2, 3...

n
Let n  , n  1, 2,3... ... (1.26)


227
Hence the corresponding solutions of (1.22) and (1.23) are

 n 
X n ( x )  Bn sin  x , and ... (1.27)
  

 n 2 2 
n (t )  Cn exp   2 t  . ... (1.28)
  
 
These are called the corresponding eigen functions. Thus the solution of (1.18) can be written as

 n 2 2   n 
un ( x, t )  BnCn exp   2 t  sin  x
     
 

 n 2 2t   n 
or un ( x, t )  an exp   2  sin  x. ... (1.29)

     

The by the principle of superposition, the solution of equation (1.18) is given by


u ( x , t )   un ( x , t ) .
n 1

  n2 a 2   n 
That is u ( x, t )   an exp  2 t  sin  x ... (1.30)
     
n 1  
if it converges. However, it is given that the initial temperature of the rod is
u ( x, o )  x (   x ) , 0 x
  n 
 x    x    an sin  x , 0 x ... (1.31)
n 1   
which is the Fourier sine series of f ( x )  x (  x ). Hence the Fourier constant an is given by

2  n 

 0
an  x (  x ) sin  x dx ,
  

2   n  
 n  
   x sin  x  dx   x 2 sin  x dx  ,
  0    0    

228
 
2     n   
  n  
an     x cos  x    cos  x  dx  
   n    0 0 n    



2 2   n  
  n  

 x cos  x   2 x cos  x  dx 
 n   0 n    
 0


2  3 3 3  n  
an    n
( 1)  n
(1)  2 cos  x 
  n n (n ) 3
  0 


4 2
an  ( 1)n  1 ... (1.32)
 
 n  3

8 2
 an  , for n is odd,
n 3 3
an = 0, for n is even. ... (1.33)
Hence equation (1.30) is the required solution of equation of (1.18) with

8 2
an  .
n3 3
Uniqueness of the Solution :
Theorem : Show that the solution u ( x, t ) of the differential equation

ut  ku xx  F ( x, t ), 0  x  , t  0 , ... (1.34)
satisfying the initial condition
u ( x, 0)  f ( x ) , 0 x, ... (1.35)
and the boundary conditions
u (0, t )  u (, t )  0 , t0 ... (1.36)
is unique.
Proof. Let u1 ( x, t ) and u2 ( x, t ) be two solutions of the equation (1.34) subject to the conditions
(1.35) and (1.36).

u1  2u1
 k  F ( x, t ) , 0 x   t  0, ... (1.37)
t x 2

229
 u2  2 u2
and k  F ( x, t ) , 0  x  , t  0 , ... (1.38)
t x 2
satisfying the conditions
u1 ( x, 0)  f ( x) , 0 x ... (1.39)
u2 ( x,0)  f ( x) , ... (1.40)

and u1 (0, t )  u , (, t )  0 , ... (1.41)

u2 (0, t )  u2 (, t )  0 . ... (1.42)

Subtracting (1.38) from (1.37) we get

  2  u1  u2 
 u1  u2   k  0,
t x 2

satisfying u1 ( x,0)  u2 ( x, 0)  0 ,

and u1 (0, t )  u2 (0, t )  0  u1 (, t )  u2 (, t ) .

These equations show that, v  u1  u2 satisfies the corresponding equation

v d 2v
k 0, 0  x  , t  0 ... (1.43)
t x 2
v(0, t )  v(, t )  0 , t  0 ... (1.44)
v ( x, 0)  0, 0  x  . ... (1.45)

Claim : We prove that u1 ( x, t )  u2 ( x, t )

Let us define a function E ( t ), such that

1
v
2
E (t )  ( x, t ) dx . ... (1.46)
2k 0

Since the integrand is positive definite  E  0 ... (1.47)

Differentiating (1.46) w.r.t. t we get

dE 1 v
dt

k  2v t
dx ,
0

1

k  2vk uxx dx by equations (1.43)
0

230
 
dE
 2  v vxx dx .
dt 0

Integrating the r.h.s. by parts we get

  
dE 
 2 v vx 0   vx2 dx 
dt  0 
Therefore , by boundary conditions (1.44) and (1.45) we have
v(0, t )  v(, t )  0 .


dE dE
  2  vx2 dx  0  0. ... (1.48)
dt 0
dt

This shows that E ( t ) is decreasing function of t. From the condition v( x,0)  0 we have from (1.46)
E(0) = 0. Therefore we have
E (t )  0, t  0. ... (1.49)
But, by definition (1.46) E (t) is non-negative.
 E (t )  0, t  0
 v ( x, t )  0 on 0  x  , t  0

 u1 ( x, t )  u2 ( x, t ) .
Hence the solution is unique.
Example 2 : The temperature u ( x, t ) in a rod of length  in governed by the p.d.e.

ut  c 2u xx .

The initial temperature is u ( x, 0)  f ( x ) . The rod and its ends are perfectly insulated

u x (0, t )  0 and u x (, t )  0 .

Find the temperature distribution in the rod.
Solution : Let u ( x, t ) be the temperature in a rod of length  . We know it is governed by the p.d.e.

ut  c 2u xx 0  x  , t  0 . ... (1.50)
Given that initial temperature is
u ( x, 0)  f ( x ) , ... (1.51)

and u x (0, t )  u x (, t )  0 . ... (1.52)

We assume u ( x, t )  X ( x )T (t ) ... (1.53)

231
be the solution of equation (1.50)
 u x  X '( x)T (t )  u xx  X "( x)T (t ) ,

and ut  X ( x)T '(t )

Hence equation (1.50) becomes

X ( x)T '(t )  c 2 X "( x )T (t )

X " T '(t)
   (say).
X c2T

 X "  X  0 , ... (1.54)

and T ' c 2T  0 , ... (1.55)
where  is either zero or positive or negative. If   0 and   0 we know that it has trivial solution.
Therefore, we choose   0 Let    2 . Hence equations (1.54) and (1.55) become

X "( x )   2 X  0 , ... (1.56)

and T '   2c 2T  0 . ... (1.57)

Solving equations (1.56) and (1.57) we get
X  A cos  x  B sin  x , ... (1.58)

and T  C exp(  2 c 2t ), C  0 . ... (1.59)

Thus the temperature distribution in the rod is given by

u ( x, t )  ( A cos  x  B sin  x ).C exp(  2 c 2t ) . ... (1.60)

To find the constants, we use the given conditions (1.52). From equation (1.60) we find

u x ( x, t )  a( A sin  x  B cos  x)C.exp( 2e 2t )

Thus u x  0, t   0  0  B for C  0 .

Hence equation (1.60) reduces to

u ( x, t )  A cos  x.C exp( 2 c 2t ) . ... (1.61)

Also u x (, t )  0  0  A.sin( ).C exp( 2c 2t ) .

 A  0 for C  0 ,
or sin(  )  0 for C  0 .
If A  0 then we have trivial solution. Therefore, we assume A  0 for C  0 .

232
  sin(  )  0 ,
    n , for n  1, 2,3,.....

n
 n  , n  1, 2,3,..... ... (1.62)

These are called the eigen values. Substituting this in equation (1.61) we get

 n   n 2 2c 2 
u ( x, t )  AC cos  x  exp  t.
   
  


 n    n2 2c 2 
or un ( x, t )  an cos  x  exp  t . ... (1.63)
   
  

By the superposition principle, the most general solution of equation (1.50) is given by

   n    n 2 2c 2 
u ( x, t )   un ( x, t )  an cos  x  exp  t. ... (1.64)
n 1 n 1    
  

Given that the initial temperature in the rod is
u ( x, 0)  f ( x ) .
Therefore, from (1.64) we have
  n 
u ( x, 0)  f ( x)   an cos  x . ... (1.65)
n 1   
This represents the expansion of f (x) in the Fourier cosine series. Consequently , the Fourier constant
an is given by

2  n 
 0
an  f ( x ) cos  x dx . ... (1.66)
  

2    n   n   n 2 2c 2 
 u ( x, t )  
 n1 0
f ( x ) cos  x  dx  cos  x   exp  t. ... (1.67)
      
  


2. Heat conduction - Infinite Rod.

Result : Find the temperature distribution in a rod of infinite length satisfying the initial conditions
u ( x, 0)  f ( x ) ,   x  

233
Solutions : Consider a homogeneous sufficiently thin rod of infinite length such that its surface is
insulated. If u ( x, t ) is the temperature in the rod, then the temperature distribution in the rod is
governed by the second order partial differential equation
ut  ku xx ,   x  , t  0 , ... (2.1)

satisfying u ( x, 0)  f ( x ) ,   x   . ... (2.2)

We use the Fourier transform method to solve the equation. Therefore let the Fourier transform of
u ( x, t ) be U ( , t )

i.e. F  u ( x, 1)   U ( , t )
Thus by definition of Fourier transform, we have

1
F  u ( x, 1)   U ( , t )   u ( x, t )e
i t
dt . ... (2.3)
2 

Also we know the formula for Fourier transform of derivative as

F  f n ( x)   (i )n F  f ( x)  . ... (2.4)

Hence taking of the Fourier transform of equation (2.1) and using the formula (2.4) we get

U t  k (i )2 F (u( x, t )) ,

U t   k 2U ( , t ) ,

 U t  k 2U  0 . ... (2.5)

This is the first order differential equation, whose solution is obtain by integrating equation (2.5)
2kt
 U ( , t )  A( )e , ... (2.6)

where A( ) is an arbitrary function to be determined from the initial conditions.

From the definition (2.3) we obtain

1
F  u ( x, 0)    u ( x, 0)e
i x
dx
2 


1
 U ( , 0)   f ( x )ei x dx by equation (2.2).
2 

But from equation (2.6), we have

U ( , 0)  A( )

 A( )  F  f ( x)   F ( ) .
234
Hence equation (2.6) becomes
2kt
U ( , t )  F ( )e  . ... (2.7)

Taking the inverse Fourier transform of equation (2.7) we get

F 1
F  u ( x  1)    F 1
F  f ( x )  F ( g )  .... for F ( g )  e  2kt

 u ( x, t )  F 1
F  f * g   ,

u ( x, t )  f * g , ... (2.8)

where f * g is the convolution of f ( x ) and g ( x ) over the interval  ,   and is defined by


1
 f * g  ( x)   f ( x   ) g ( ) d  .
2 

Thus we have

1
u ( x, t ) 
2
 f ( ) g ( x   ) d  , ... (2.9)


2kt
where g F 1
(e )

1  x2 
 g ( x)  exp   
2  kt  4kt 
 

 ( x   )2 
1
 g(x   )  exp   .
2  kt  4 kt 
 
Hence equation (2.9) becomes

1 1    x   2 
u ( x, t ) 
2 2  kt
 f ( ) exp 
 4kt  .
 d ... (2.10)
  
If k  1 and

0 when x<0
f ( x)  
a When x>0,
then we have

235
  ( x   )2 
a 8

 2t 0
u ( x, t )  exp    d
 4t  .

x 
Put  ,
2 t

 d   2 td

x
when   0,   
2 t
and as   ,    .
Thus we have

a 2
u ( x, t )   e  2 td
 2t  x
2 t


a 2
u ( x, t )  2  e d
 x
2 t

We write this as

 0  
a   2  2 
u ( x, t )  2   e d   e d 
 . ... (2.11)
  x 2 t 0 

Now consider the integral

0
2
 e d
 x
2 t

Put    y  d   dy

x x
When  y
2 t 2 t
and  0  y0
x
0 2 t
 2 2
 e d   e  d
Thus we have x 0
... (2.12)
2 t

236
 
 2
Also consider e d .
0

Put  2  t  2 d  dt ,

1
 d  dt .
2 t
When  0t 0,
and  t .

  1 1
 2 1 1 1
e d   e
2 0
t t 2
Thus t 2 dt  e t dt
0 0
2


2 1 1 
  e d   . ... (2.13)
0
2 2 2

Using (2.12) and (2.13) in equation (2.11) we get

 x 
a 2 t 2  
u ( x, t )  2   e d  
  0 2 
 

 x 
a  2 2 t
2 
 1   e d 
2   0 
 

a   x 
 u ( x, t ) 1  erf   , ... (2.14)
2  2 t 
where

2  2
erf ( x ) 
 e d is the error function.
0

237
3. Families of Equipotential surfaces :
Definition :
Let f ( x, y , z )  c be a one-parameter family of surfaces. We say that this family of surfaces
is equipotential if these exists a potential function  ( which is a solution of Laplace equation  2  0 )
such that  is constant whenever f ( x, y , z ) is constant.
Note : Not every one parameter family of surfaces f ( x, y , z )  c is a family of equipotential surfaces.
Result : Find the condition that a one parameter family of surfaces form a family of equipotential
surfaces.
Proof : Let f ( x, y , z )  c be a one parameter family of surfaces. By definition, equation (3.1) will be
a family of equipotential surfaces if the potential function  (Which is a solution of the Laplace equation
 2  0 ) is constant whenever f ( x, y , z ) is constant.
This means that there must exist a functional relation of the type

  F  f ( x, y, z ) ... (3.2)

between the function  and F such that  = constant whenever f ( x, y , z ) = constant.

Differentiating (3.2) partially w.r.t. x we obtain

 dF f
 . ,
x df x
... (3.3)
2
 2 d 2 F  f  dF  2 f
and  2 .   . . ... (3.4)
x 2 df  x  df x 2

2
 2 d 2 F  f  dF  2 f
Similarly,     . , ... (3.5)
y 2 df 2  y  df y 2

2
 2 d 2 F  f  dF  2 f
and     . . ... (3.6)
z 2 df 2  z  df z 2

Therefore, consider

2  2  2  2
    ,
y 2 y 2 z 2

238
 d 2 F  f   f   f   dF   2 f  2 f  2 f
2 2 2

2
               
df 2  x   y   z   df  x 2 y 2 z 2 

 

 2  F '' ( f ) ( gradf ) 2  F ' ( f ) 2 f . ... (3.7)

Since  satisfies the Laplace equation in free space

  2  0

F '' ( f ) 2 f
  . ... (3.8)
F'( f )  gradf 2
This shows that, the condition that the surfaces (3.1) form a family of equipotential surfaces is that the

2 f
quantity 2 is a function of f alone. We denote this function by  ( f ) . Hence equation (3.8)
gradf
can be written as

F '' ( f )
  ( f ) ,
F '( f )

d 2F df
 2
 ( f )  0. ... (3.9)
df df

Integrating we get

dF
 A.e 
  ( f ) df
... (3.10)
df
where A is a constant. Integrating (3.10) w.r.t. f we get

  F ( f )  A e 
  ( f ) df
df  B ... (3.11)

where B is a constant. This is the general form of the corresponding potential function.
This is the necessary condition that the one-parameter family of surfaces f ( x, y , z )  c is a family of
equipotential surfaces.

Example 1 : Show the surfaces

x2  y 2  z2  r 2 , r0
form a family of equipotential surfaces and find the general form a the corresponding potential function.

239
Solution : Let

f  x, y , z   x 2  y 2  z 2  r 2 ... (3.12)
be the one parameter family of surfaces. To show that this family forms a family of equipotential
surfaces, we find the potential  s.t.

 f f f 
grad f   x , dy , z   2( x, y, z ), ... (3.13)
 

2 f 2 f 2 f
2 f     222  6 ... (3.14)
x 2 y 2 z 2

 2 f  6 .
2
 f  4( x 2  y 2  z 2 ) ... (3.15)

2
 f 4f .

Therefore, the equation

2 f 3
2
  ( f ) ... (3.16)
f 2f

F '' ( f )
Therefore, the equation   ( f )
F '( f )

  A e 
  ( f ) df
has solution df  B

3 1 3
2 f
 df
2
   A e .df  B    A elog( f ) df  B.

3

  A f 2 df B,

This gives

2A
 1
B
.
f 2

2A
   B.
r

240
Example 2 : Show that the surfaces
2
2 2 2
x  y  z  cx 3
can form an equipotential of surfaces, and find the general form of the potential function.
Solution : One parameter family of surfaces is given by
2
2 2 2
x  y  z  c x3

2
i.e. 
x 3 x2  y 2  z 2  c . 
2
Let 
f  x, y, x   x 3 x 2  y 2  z 2  c  ... (3.17)

To show this family forms a family of equipotential surfaces we find

 f f f 
f   , , 
 x dy dz 

 4 1 2 5 2 
f   x 3  x 3  y 2  z 2  , 2 yx 3 , 2 zx 3 
2

3 3 

f 
2 53
x  2 x 2  y 2  z 2 ,3xy,3xz  . ... (3.18)
3

 2 f 4  2 3 10  8 3 2
Now 2
 x  x  y  z2  ,
x 9 9

2 f 2
2
3 ,  f  2x 3
2
2
 2 x 2 .
y z

 2 f 2 f 2 f 
Hence  2 f   2  2  2  becomes
 x y z 

 y  z 2   2 x  3  2 xx  3 ,
4  2 3 10  8 3 2 2 2
2 f  x  x
9 9

 x  x  y  z2 
40  2 3 10  8 3 2
9 9

Thus, 2 f 
10  8 3
x 4x2  y2  z 2  .
9

241
 2 2 2
 f   f   f 
Now  f  2
      ... (3.19)
 x   y   z 

becomes

4 10 3 
 2 x 2  y 2  z 2   9 x 2 y 2  9 x 2 z 2  ,
2 2
grand f  x 
9

4 x   y 2  z 2   4 x 2  y 2  z 2   9 x 2  y 2  z 2   ,
4 10 3  4 2
 x 
9

4 x   y 2  z 2   5 x 2  y 2  z 2   ,
4 10 3  4 2
 x 
9


4 10 3  4
x  4 x  4 x 2
 y 2
 z 2
   y 2
 z   x 2  y 2  z 2  ,
2 2
9

 4 x  x  y  z    y  z  y  z  x   ,
4 10 3  2 2 2 2 2 2 2 2 2 
 x
9

f
2

4 10 3 
x   4 x 2  y 2  z 2    x 2  y 2  z 2  . ... (3.20)
9

 f 2
10  8 3
x 4x2  y2  z 2 
 9
So that,
 4 x 2  y 2  z 2  x 2  y 2  z 2  ,
2 4 10 3
f x
9

2
5 x 3

2  x2  y2  z 2 

5
2 f 2 2 f 5
Therefore,      (t ) ... (3.21)
 x2  y 2  z 2 
2 2 2
f x 3 f 2f

This shows that the given set of surfaces forms a family of equipotential surfaces.
Now to find the general form of the corresponding potantial function, we know it is given by

  A e 
  ( f ) df
df  B ,
5 1
2 f
 df
  A e df  B ,

5
2
  A e log( f ) df  B ,

242
 5
  A f 2 df B,

2 3
   Af 2 B,
3

3
   A  x 3  x 2  y 2  z 2 
2 2 2
or B,
3

 3
2
   A  x  x 2  y 2  z 2 
2
  B .
3
This is the required potential function.
Example 3 : Show that the family of right circular cones x2 + y2 = cz2, where c is a parameter, forms
a set of equipotential surfaces and show that the corresponding potential function is of the form

A log tan  B , where A and B are constants and  is the usual polar angle.
2
Solution : The family of right circular cones is given by

x 2  y 2  cz 2

i.e. z 2  x 2  y 2   c

Let f  x, y, z   z 2  x 2  y 2   c . ... (3.22)

To show that, this surfaces form an equipotential surfaces, we find

 f f f 

f   , ,   2 xz 2 , 2 yz 2 , 2 z 3  x 2  y 2  
 x y z 


f  2 xz 2 , yz 2 ,  z 3  x 2  y 2  .  ... (3.23)

Therefore, f
2

 4 x 2 z 4  y 2 z 4  z 6  x 2  y 2 
2
,

 4z 6 x 2 z 2  y 2 z 2   x 2  y 2 
2
,
 4 z 6  x 2  y 2  x 2  y 2  z 2  .
2
f ... (3.24)

2 f 2 f 2 f
Next 2 f    gives
x 2 y 2 z 2

 2 f  2 z 2  2 z 2  6 z 4  x 2  y 2  ,

243
  2 f  4 z 2  6 z 4  x 2  y 2  ,

 2 f  2 z 4  3 x 2  3 y 2  2 z 2  ,


 2 f  2 z 4 2  x 2  y 2  z 2    x 2  y 2  .  ... (3.25)

2 f 2 z 4  2  x 2  y 2  z 2    x 2  y 2 
Therefore, 
4 z 6  x 2  y 2  x 2  y 2  z 2 
2
f

1 2 2 1 
 z  2
 2
2 x  y 2 2 2
x y z 

1 1
  2 2
z 2
 x  y  2z  x  y 2   2
2 2

2 f 1 1
or    f  .... (3.26)
f
2
f 2  f  1

This shows that the one parameter family of surfaces (3.22) forms an equipotential surfaces.
To find the corresponding potential function, we know it is given by

   f  df
  A e  df  B ,

   
1 1 
df
 1 
  A B,
 f 2 f
e df

  A e
 1
  log f  log( f 1) 2 df  df  B ,

  A   ( f  1) 2
1 1
 df  B ,
f 

  A  log f  2 f  1   B . ... (3.27)

This is the required potential function. Now we show that potential function is given by


  A tan  B .
2
We consider the transformation
x  r sin  cos  , y  r sin  sin  , r  r cos 

244
  x 2  y 2  r 2 sin 2 

Hence  f  z 2  x 2  y 2   tan 2   df  2 tan  sec 2  d

 f  1  tan 2   1  sec  .

1 1
  f  
f 2  f  1

1 1
becomes  f   .
tan  2 sec 2 
2

Integrating we get

 1 1 
   f  df    tan 2   2sec 2   2 tan  sec
2
 d ,

2sec 2 
 d   tan  d ,
tan 

   f  df  2 log tan   log cos  ,

tan 2   sin 2  
   f  df  log cos 
 log 
 cos3  
.

   f  df
Therefore   A e  df  B

 cos3  
log 

2
becomes   A e  sin    2 tan  sec 2  d  B ,

cos 3 
  2 A 2
 tan  sec 2  d  B ,
sin 

  2 A cosec d  B ,


  A log  tan   B .
 2
This is the required family of equipotential surfaces.
Example 4 : Show that surfaces

 x2  y2   2a 2  x 2  y 2   a 4  c
2

can form a family of equipotential surfaces and find the general form of the corresponding potential
function.
245
Solution : The one-parameter family of surfaces is given by

f  x , y , z    x 2  y 2   2a 2  x 2  y 2   a 4  c .
2
... (3.28)

To show this surfaces form an equipotential surfaces, we find

 f f f 

f   , ,   4 x  x 2  y 2   4 a 2 x , 4 y  x 2  y 2   4 a 2 y , 0 , 
 x y z 


 f  4 x  x 2  y 2   4a 2 x, 4 y  x 2  y 2   4a 2 y, 0 .  ... (3.29)

2 2 2
2  f   f   f 
Therefore f        ,
 x   y   z 

 16 x 2  x 2  y 2  a 2    16 y 2  x 2  y 2  a 2   ,
2 2 2
becomes f


 16  x 2

 x 2  y 2   a 4  2a 2  x 2  y 2  
2

 y2   
x 2  y 2   a 4  2a 2  x 2  y 2   ,
2

 16  x 2  y 2   a 4  x 2  y 2   2a 2  x 2  y 2  x 2  y 2   ,
3

 16  x 2  y 2   x 2  y 2   a 4  2a 2  x 2  y 2  .
2 2
f ... (3.30)

2 2 f 2 f 2 f
Now  f  2  2  2
x y z

becomes  2 f  16  x 2  y 2  ... (3.31)

2 f 16  x 2  y 2  1
 
16  x  y f
So that 2 2 2 f
f

2 f 1
Hence 2
   f . ... (3.32)
f f

This shows that the given set of surfaces (3.28) forms a family of equipotential surfaces.
Now to find the general form of the potential function, we know
   f  df
  A e  df  B ,

246
 1
 df
  A e f
 df  B ,

1
  A elog f df  B ,

  A 1 df  B ,
f

  A log f  B ,

  A log  x 2  y 2   2a 2  x 2  y 2   a 4   B ,
2
or ... (3.33)

where A, B are constants. This is a required equipotential function.

247
 UNIT - VII

LAPLACE EQUATION

1. Introduction :
Various physical phenomena are goverened by the Laplace equation. In this unit we derive the
Laplace equation and discuss the method of its solution. Various boundary value problems for the
Laplace equation viz., the Dirichlet problem and Neumann problem for certain specified regions are
the subject matter of this unit.
Result : Derive Laplace equation.
Proof : Consider two particles m and m1 at Q and P respectively separated by a distance r. Then by
Newton’s law of gravitation, the magnitude of the force is directly proportional to the product of the
masses and inversely proportional to the square of the distance between them.

Gmm1
F  , ... (1.1)
r2
where the negative sign indicates the force is attractive. Here G is the gravitational constant. Assuming
the unit mass at Q and G = 1, the force at Q due to the mass m1 at P is given by

m1
F  ,
r2

  m1 
F  . ... (1.2)
r  r 
Let the particle of unit mass move under the attraction of the particle of mass m1 at P from infinity upto
Q, then the work done by the force F is given by
r r
  m1 
 Fdr   r
  dr ,
 r 
 

r
m 
 d 1 ,

 r 

r
m1
  Fdr 

r . ... (1.3)

The gravitational potential is defined to be the amount of work which must be done against gravitational
force. Hence the potential V at Q due to a particle at P is given by
248
 m1
V  . ... (1.4)
r
From equations (1.2) the intensity of the force at P is given by

F  V . ... (1.5)

Now if we consider a system of particles of masses m1, m2, ...., mn which are at distance r1, r2, ..., rn
respectively, then the force of attraction at Q due to the system of particles is given by
n
m 
F   i  ,
i 1  ri 

 n m 
 F   i  . ... (1.6)
 i 1 ri 
The work done by the force acting on the particle is
r n
mi
 Fdr   r
i 1 i
 V , ... (1.7)


 n m 
  2V   2   i  , , ri  0
 i 1 ri 
n
m 
  V    2  i  ,
2

i 1  ri 

where ri  xi i  yi j  zi k ,
1
 ri  ri   xi  yi  zi  2

m   2 2 2  2 1
Thus 2  i   mi 2
  
 i x  yi
2
 z i 
2  2
,
 ri   x y 2 z 2 

1  xi
where
 2
x
 xi  yi2  zi2  2  3 ,
 xi2  yi2  zi2  2


2 2
 x  y 2
 z 
1
2  2

 2 xi2  yi2  zi2  .
i i i
x 2 5
 xi2  yi2  zi2  2

2 2
 x  y 2
 z 
1
2  2

 2 yi2  zi2  xi2 
i i i
Similarly, y 2 5 ,
 xi2  yi2  zi2  2
249
 2 2
 i i i
x  y 2
 z
1
2  2

 2 zi2  xi2  yi2 
and z 2 5 ,
 xi2  yi2  zi2  2
 2 2 2   m 
  2  2  2  i   0,
 x y z   ri 

m 
 2  i 0,
 ri 

  2V  0 . ... (1.8)
This is called the Laplace equation.
Note : In 2-dimensions, the Laplace’s equation is given by

 2u  2u
 2u    0. ... (1.9)
x 2 y 2
A solution u (x, y) of equation (1.9) is called 2-dimensional harmonic function.

Solution of Laplace Equation :

Example 1 : Obtain the solution of the two-dimensional Laplace equations  2u  0 by the method
of separation of variables.
Solution : Consider the two-dimensional Laplace equation

 2u  0 ,

 2u  2 u
i.e.   0. ... (1.10)
x 2 y 2
To find the solution of (1.10) we assume

u  x, y   X ( x )Y ( y ) . ... (1.11)

 2u  2u
  X ''( x )Y ( y ) and  X ( x)Y ''( y )
x 2 y 2
Therefore, equation (1.10) becomes
X '' Y  XY ''( y )  0 ,

X '' Y ''
or   k (say), ... (1.12)
X Y
where k is called the separation constant, and k may be positive, zero or negative.

250
Case (i) k > 0. Take k   2 ,  is real.
Therefore, we get from equations (1.12)
2
X ''  2 x  0 and Y ''  y  0 ... (1.13)
Solutions of these equations are respectively given by

X  C1e x  C2 e  x and Y  C3 cos  y   C 4 sin  y  . ... (1.14)

Hence the solution of equation (1.10) becomes

u  x, y    C1e x  C2 e  x   C3 cos  y  C4 sin  y  . ... (1.15)

Case (ii) If k = 0, then from equations (1.12) we have

X ''  0 and Y ''  0 .
Which provide us
X  C5 x  C6 and Y  C7 y  C8 .
Hence the solution of (1.10) becomes

u  x, y    C5 x  C6  C7 y  C8  . ... (1.16)

Case (iii) Let k < 0. Take k   2

Hence the equations (1.12) become
2
X ''  2 x  0 and Y ''  y  0 ,
which have solutions

X   C9 cos  x  C10 sin  x  and Y  C11e y  C12e  y .

Hence the general solution of (1.12) is given by

u  x, y    C9 cos  x  C10 sin  x   C11e y  C12 e  y  . ... (1.17)

In all these solutions Ci (i = 1, 2, ...,12) are constants of integration and are to be calculated by using
the boundary conditions.
Laplace Equation in Polar Form :

Result : Show that in polar-coordinates r,  , the two-dimensional Laplace equation u xx  u yy  0

takes the form

1 1
u rr  u r  2 u  0
r r
Proof : In Cartesian co-ordinates the two-dimensional Laplace equation is given by

 2u  0 ,
251
  u xx  u yy  0 . ... (1.18)
We know the relations between the Cartesian co-ordinates and polar co-ordinates are given by
x  r cos  , y  r sin  ,

1  y 
 r 2  x 2  y 2 and   tan   ,
x

r x
where   cos  ,
x r

r y
  sin  ,
y r

 y  sin 
and  2 
x x  y 2
r ,

 x cos 
 2 .
y r r
We have by Chain rule of partial differentiation

u u r u  x 1  y 
     ur  u  2 
x r x  x r y2  x ,
1 2
x

 y 
 u x  u r cos   u  2 2 
x y 

sin  
 u x  u r cos   u   ... (1.19)
 r 

y  x2   1 
and u y  ur  u  2 2   ,
r  x  y  x 

cos  
u y  u r sin   u  . ... (1.20)
 r 

r 
Similarly, we find u xx   u x  x   u x r   u x  ,
x x

 sin    cos   u cos   u  sin      sin  

u xx  ur cos   u    r    ,
  r  r  r    r 

252
  sin    u sin   cos  
u xx  u rr cos   u r    2 
  r  r 

 sin    u cos     sin  

 u r cos   ur sin   u     ,
  r  r   r 

sin  cos  sin 2  sin 2  sin  cos  .

u xx  urr cos 2   2ur  u 2
 u r  2u ... (1.21)
r r r r2
Similarly, we find

r  ,
u yy   u y    u y   u y 
y r y  y

cos   cos    cos  

 u r sin   u sin   u r sin   u  
 r r 
 r   r 

cos  cos 
  urr sin   u r  u 2  sin  
 r r 

cos  sin  cos 

  ur sin   u  u  ur cos   ,
 r r  r

sin  cos  cos 2  sin  cos  cos 2  . ... (1.22)

u yy  u rr sin 2   2u r  u  2u  u r
r r2 r2 r
Adding equations (1.21) and (1.22) we get

1 1
u xx  u yy  u rr  u  ur .
2 
r r

Thus u xx  u yy  0

1 1
 u rr  u r  2 u  0 . ... (1.23)
r r
This is the polar form of the 2-dimensional Laplace equation.
Example 2 : Show that the two-dimensional Laplace equation  2u  0 in polar co-ordinates r, 

has the solution of the form   Ar n  Br n e in , where A and B are constants.
n

Solution : The two-dimensional Laplace equation in plane polar co-ordinates is given by

 2u 1 u 1  2u
   0. ... (1.24)
r 2 r r r 2  2
Let u  R ( r ) ( ) ... (1.25)
253
be the solution of equation (1.24)
 urr  R ''(r )( ) and ur  R '(r ) ,

u  R ( r ) ''  
Hence equation (1.25) becomes

1 1
R ''( r )    R '( r )    2 R ( r ) ''    0 .
r r

Dividing this equation by R(r )   we get

R '' 1 R '(r ) 1  ''

  2  0,
R r R r 

1 2  ''
or r R '' rR '     n 2 . (say) ... (1.26)
R 

  '' n 2  0 ,

d 2
i.e. 2
 n 2  0 , ... (1.27)
d

d 2R dR
and r2 2
r  n2R  0 . ... (1.28)
dr dr
Equation (1.27) provides

  e  in ,
or   C cos n  D sin n . ... (1.29)
Let R  r m be the solution of equation (1.28). Hence the equation (1.28) becomes

r 2  m  m  1 r m2   rmr m1  n 2 r m  0 ,

 m2  n2  r m  0 ,
 m  n .
Hence the solution of (1.28) is given by

R  Ar n  Br  n . ... (1.29)
Therefore the solution of equation (1.24) becomes

u  r ,     An r n  Bn r  n  e  in , ... (1.30)

n

or u  r ,     An r n  Bn r  n   Cn cos  n   Dn sin  n   . ... (1.31)

n

Which is the required result.

254
Laplace Equation in Spherical Polar Co-ordinates :
Result : Show that in spherical polar co-ordinates r,  ,  the Laplace equation  2u  0 takes the
form

  2 u  1   u  1  2u
r   sin   2 0.
r  r  sin      sin   2

Proof : In Cartersian co-ordinates the Laplace equation is given by

 2u  u xx  u yy  u zz  0 . ... (1.32)

To transform equation (1.32) in to spherical polar co-ordinates, we have

x  r sin  cos  , y  r sin  sin  , z  r cos  ... (1.33)

1 x2  y2  y
2 2 2
 r  x  y  z ,   tan 2
,   tan 1   ... (1.34)
z x

x y z
where rx   sin  cos  , ry   sin  sin  , rz   cos  , ... (1.35)
y r r

and xz cos  cos  ,

x   x 
r 2 x2  y2 r

cos  sin  sin 

y  , z  ... (1.36)
r r
Similarly, we find

sin  cos 
x   , y  , z  0 ... (1.37)
r sin  r sin 
Now by using the chain rule of partial differentiation we write
u x  u r rx  u  x  u x .

Using equations (1.35), (1.36) and (1.37) we get

cos  cos  sin 

u x  sin  cos  u r  u  u . ... (1.38)
r r sin 
In the same way, we find

cos  sin  cos 

u y  sin  sin  u r  u  u , ... (1.39)
r r sin 

sin 
and u z  cos  u r  u . ... (1.40)
r

255
Now to find the second order derivative, we again use the chain rule and write

u xx   u x  r rx   u x   x   u x   x

 cos  cos  sin  

  sin  cos  u r  u  u    sin  cos   
 r r sin   r

 cos  cos  sin    cos  cos  

  sin  cos  u r  u  u   
 r r sin    r 

cos  cos  sin  sin 

  sin  cos  ur  u  u     
,
 r r sin    r sin  

cos 2  cos 2  sin 2 

 u xx  u rr sin 2  cos 2   u  u 
r2 r 2 sin 2 

 2 sin  cos  cos 2    2sin  cos  

u r    ur   
 r   r 

2 cos  cos  sin    cos 2  cos 2  sin 2  

u   
 ru  
 r 2 sin    r r 

 sin  cos  cos 2  cos  sin  sin  cos  

u    2 2 
 r2 r 2 sin 2  r sin  

 cos  sin 2  2 cos  sin  cos 2  

u  2  . ... (1.41)
 r sin  r2 
Similarly, the second order derivative

u yy   u y  ry   u y   y   u y   y gives
r  

cos 2  sin 2  cos 2 

u yy  u rr sin 2  sin 2   u  u 2 
r2 r sin 2 

 2sin  cos  sin 2    2 cos  sin    u  2 cos  cos  sin   

u r    u r     
 r   r   r 2 sin  

 cos 2  sin 2  cos 2    2sin  cos  sin 2  cos  cos 2  

u r     
 u  
 r r   r2 r 2 sin  

 sin  cos  sin  cos  cos 2  sin  cos  

u    2 2  . ... (1.42)
 r2 r sin  r 2 sin 2  

256
and u zz   u z  r rz   u z   z   u z   z ,

sin   sin   sin 

 u zz   u r cos   u 
 cos    u r cos   u  ,
 r r  r  r

2sin  cos  sin 2  sin 2  cos  sin 

 u zz  u rr cos 2   u r  u 2
 u r  u . ... (1.43)
r r r r2
Adding equations (1.41), (1.42) and (1.43) we obtain

 2u  0
1 1 2 cos 
 u rr  u  2 2 u  u r  2
2 
u  0 . ... (1.44)
r r sin  r r sin 
This can also be written as

1   2 u  1   u  1  2u
 r    sin     0,
r 2 r  r  r 2 sin      r 2 sin 2   2

  2 u  1   u  1  2u
  r    sin    0. ... (1.45)
r  r  sin      sin 2   2

This is the required Laplace equation in spherical polar co-ordinates.

Example 3 : Show by using the method of separation of variables that the general solution of Laplace’s
equation in (r,  ,  ) co-ordinates is

u  r , ,       n r n   n n1  S n  ,   ,
1
n0  r 

n
where S n  ,     Pnm     Anm cos m  Bnm sin m  ,
m 0

  cos  , and Pnm    is the associated Legendre function and  n ,  n ,

Anm and Bnm are constants.
Solution : We know the Laplace equation in spherical polar form is given by

  2 u  1   u  1  2u
r   sin   2 0 ... (1.46)
r  r  sin      sin   2

This can be written as

 2u u 1   u  1  2u
r2  2 r   sin    0 ... (1.47)
r 2 r sin      sin 2   2

257
Since u is a function of r,  ,  we assume

u  r ,  ,    R ( r )      ... (1.48)
is the solution of equation (1.47). Therefore, we find from (1.48)

urr  R '' ,  r  R ' ,    R '  and u  R '' . ... (1.49)
Substituting these in equation (1.47) we get

1  1
r 2 R ''   2rR '   sin  R ''   cos  R '    R ''  0 .
sin  sin 2 
Dividing throughout by R we get

R '' R '  '' ' 1  ''

r2  2r   cot    0,
R R   sin 2  

r 2 d 2 R 2r dR 1 d 2 cot  d  1 d 2
i.e.      0,
R dr 2 R dr  d 2  d  sin 2  d 2

r 2 d 2 R 2r dR 1 d    1 d 2
   sin    0,
R dr 2 R dr  sin  d     sin 2  d 2

 r 2 d 2 R 2r dR 1 d     2 1 d 2
or R    sin    sin     m 2 .(say) ... (1.50)
 dr 2
R dr  sin  d      d 2

Now consider the r.h.s. equations of (1.50)

d 2
2
 m 2  0 , ... (1.51)
d
which has solution

    Ce  im . ... (1.52)

Now the l.h.s. equation of (1.50) becomes

r 2 d 2 R 2r dR 1 d  d  m2
   sin    ... (1.53)
R dr 2 R dr  sin  d  d  sin 2 
This can also be written as

r 2 d 2 R 2r dR 1 d  d  m2
    sin     k (say) ... (1.54)
R dr 2 R dr  sin  d  d  sin 2 
Consider

258
 r 2 d 2 R 2r dR
 k,
R dr 2 R dr

d 2R dR
 r2 2
 2r  kR .
dr dr

Take for convenience k  n  n  1

d 2R dR
Therefore r2 2
 2r  n  n  1 R  0 , ... (1.55)
dr dr

1 d  d   m 2
and  sin    n  n  1   0
sin  d  d  sin 2 

1 d  d     m2 
  sin    n n  1    0 ... (1.56)
sin  d  d   sin 2  
Equation (1.55) is a homogeneous linear equation of second order. We put r = ez (changing the
independent variable r to z), hence we find

dr dz 1
r 
dz dr r

dR dR dz 1 dR dR dR
    r 
dr dz dr r dz dr dz

d 2 R d  1 dR  1 d 2 R dz 1 dR
Now     
dr 2 dr  r dz  r dz 2 dr r 2 dz

1  d 2 R dR 
   ,
r 2  dz 2 dz 

d 2 R d 2 R dR
 r2  2  .
dr 2 dz dz

d
If   then we have
dz

d
r  ,
dr

d2
and r2     1 .
dr 2
Hence equation (1.55) becomes

259
    1  2  n(n  1)  R  0 ,
   n   n  1 R  0 . ... (1.57)
This is a differential equation with constant coefficients whose auxiliary equation is
  n   n  1  0 ,

which has roots   n,     n  1 .

Hence solution of (1.57) is given by

R  C1e nz  C2e  ( n1) z . ... (1.58)

Consequently, the solution of equation (1.55) is

R  C1r n  C2r  ( n1) . ... (1.59)

Now to find the solution of equation (1.56) we put

d
  cos  ,    sin  ,
d

d d d  d d
We write       sin  ,
d d  d d d

d 2 d d 2 d 
  cos   sin   ,
d 2 d d  2 d

d 2 d 2
2 d 
and    cos   sin  .
d 2 d d2
From equation (1.56) we have

1  d 2 d   m2 
 sin   cos     n ( n  1)    0 ... (1.60)
sin   d 2 d   sin 2  
Using the above expressions in equation (1.60) we get

d d 2  cos    2 
 cos   sin 2    sin   d    n( n  1)  m    0 ,
d d  2 sin  d  sin 2  

1   2  d
2
 d  m2 
 2   n(n  1)  0.
i.e. d2 d  1   2   ... (1.61)

This is called as associated Legendre’s equation whose solution is given by

  APnm     APnm  cos   . ... (1.62)

260
Using equations (1.52), (1.59) and (1.62) in equation (1.48) we obtain the general solution of given
equation (1.46) in the form

u  r , ,     A1r n  A2 r ( n1)  Pnm  cos    e  im .

By supperposition principle, the general solution can also be written as

  
u  r , ,       n r n  n n1  S n  ,   ,
n 0  r 

n
where S n  ,     Pnm  cos   Anm cos m  Bnm sin m  .
m 0

Kelvin’s Inversion Theorem :

Theorem : If u  u  r ,  ,   is a harmonic function, where  r ,  ,   are the spherical polar co-ordinates,
then show that

a2  a2 
u u  , ,  
r  r 
is also a harmonic function, where ‘a’ is a constant.

Proof : Given that u  r ,  ,   is a harmonic function.  it satisfies Laplace equation.

  2 u  1   u  1  2u
 r   sin   0. ... (1.63)
r  r  sin      sin 2   2

Claim : We prove that

a2  a2 
u u  , ,   is also a harmonic function.
r  r 

a2
Let R ,
r

 u  Ru  R,  ,   .

Since u  R,  ,   satisfies the equation (1.63) as it is harmonic.

  2 u  1   u  1  2u
  R    sin    0. ... (1.64)
R  R  sin      sin 2   2

We claim that

261
 2
  2 u   1   sin  u  1  u 0
r    . ... (1.65)
r  r  sin      sin 2   2

Therefore, consider

u 
r2  r 2  Ru  R, ,    ,
r r

u R 2 R
 r 2R  r  u  R,  ,   ,
R r r

u  a 2  a 2  u  a2 
 r2  r2    2   r 2   2  u  R,  ,   ,
r  r   r  R  r 

u a 4 u
 r2   a 2u  R ,  ,   . ... (1.66)
r r R
Differentiating this with respect to r we get

  2 u  a 4 u a 4  2u R u R
r  2   a2 ,
r  r  r R r R r 2
R r

2a 4 u a 6  2u
  ,
r 2 R r 3 R 2
2
  2 u   2 R 2 u  R 3  u
 r  ,
r  r  R R 2

  2 u    2 u 
or r R R . ... (1.67)
r  r  R  R 
Similarly, consider

u 
sin   sin   Ru  R, ,    ,
 

u u
sin   R sin  . as R  R  
 
Differentiating this with respect to  we obtain

  u    u 
 sin    R sin ,
       

u  2u
 R cos   R sin  ,
  2

262
 1   u  R  u  2u 
  sin    cos   sin  2 
sin      sin     

1   u  R   u 
  sin    sin  . ... (1.68)
sin      sin     
Next consider

1  2u 1 2
  Ru  R, ,    ,
sin 2   2 sin 2   2

1  2u R  2u
 . ... (1.69)
sin 2   2 sin 2   2
Adding equations (1.67), (1.68) and (1.69) we get
2
  2 u   1   sin  u  1  u
r    =
r  r  sin      sin 2   2
  2 u  R   u  R  2u
R  R    sin    ,
R  R  sin      sin 2   2

 u  1   u  1  2u 
 R   R 2    sin   
 R  R  sin      sin 2   2  ,

=0 by virtue of equation (1.64)

  2 u  1   u  1  2u
 r   sin   2 0.
r  r  sin      sin   2

a2  a2 
This proves that u  u  , ,   is also harmonic.
r  r 

2. Boundary Value Problems :

Any problem of determining a function u (x, y) satisfying Laplace’s equation within certain
region D and satisfying certain conditions on the boundary B of the region D is called boundary value
problem, for the Laplace equation.
There are mainly three types of boundary value problems for Laplace equation viz.
1. The first boundary value problem, called The Dirichlet problem.
2. The second boundary value problem called The Neumann problem.
3. The third boundary value problem called the Mixed Boundary Value problem.

263
Dirichlet Problem :
There are two types of Dirichlet peoblems -
(i) Interior Dirichlet Problem and
(ii) Exterior Dirichlet Problem.
Interior Dirichlet Problem :
If f is a continuous function on the boundary B of some finite region D, then the problem of
determining a function u (x, y) such that

(i)  2u ( x, y )  0 with D (i.e. u (x, y) is harmonic inside D) and

(ii) u (x, y) = f on B (i.e. u coincides with f on the boundary B)
is called Interior Dirichlet Problem.
Exterior Dirichlet Problem :
If f is a continuous function prescribed on the boundary B of a finite simply connected region
D, then the problem of determining a function u (x, y) such that

(i)  2u ( x, y )  0 outside D and

(ii) u (x, u) = f on the boundary B
is called Exterior Dirichlet Problem.
The Neumann Problem :
Interior Neumann Problem :
If f is a continuous function defined uniquely at each point of the boundary B of a finite region D, then
the problem of determining a fuunction u (x, y) such that

(i)  2u ( x, y )  0 in D (i.e. u is harmonic inside D) and

u 
(ii) Satisfies  f ( s ) on the boundary B, where is the directional derivative along the
n n
u
outward normal (i.e. normal derivative coincides with f at every points of B)
n
is called the interior Neumann problem.
Exterior Neumann Problem :
If f is a continuous function prescribed at each point of the smooth boundary B of a bounded
simply connected region D. Then finding a function u (x, y) satisfying

(i)  2u ( x, y )  0 outside D and

u
(ii)  f on the boundary B
n
is called an exterior Neiumann Problem.
264
 
Note : If  is the temperature, is the heat flux representing the amount of heat crossing per unit
n
volume per unit time along the normal direction.
The Third Voundary Value Problem :
The problem of finding a function u (x, y) which is harmonic in D and satisfies the condition
u
 h( s )u ( s )  0 on B where h( s)  0 and h( s)  0 .
n
The Fourth Boundary Value Problems (Mixed Boundary Value Problem) :
The Robin Problem :
The problem of finding a function u (x, y) which is harmonic in D and satisfies different boundary
u
conditions on different portions of the boundary B, such as u  f1 ( s ) as B1  f 2 ( s ) on B2, where
n
B1  B2  B , is called Robin Problem.
Maximum and Minimum Principle :
Theorem : Let D be a region bounded by a simple, closed, piecewise smooth curve B. Let u (x, y) be
a function which is continuous in a closed region D  D  B and satisfy the Laplace equation  2u  0
(i.e. harmonic in D) in the interior of D. If u is not constant everywhere on D , then the maximum and
minimum values of u (x, y) must occur only on the boundary B of D.
Proof : Let D be a region bounded by B inside which the function u (x, y) is harmonic.
i.e.  2u  0 in D

 2u  2 u
i.e.   0 in D. ... (2.1)
x 2 y 2
Let the maximum value of u (x, y) on B be M. Let the theorem be not true. Therefore we assume that
the function u (x,y) attains its maximum at some interior point  x0 , y0  in D and not at any point on the
boundary B of D.

If M 0  u  x0 , y0  then M 0  M .

Say M 0  u  x0 , y0   M  . ... (2.2)

Let us construct an auxiliary function

M0  M  2
 x  x0    y  y0   ,
2
v ( x , y )  u ( x, y )  2 ... (2.3)
4R

where  x, y   D and R is the radius of the circle with centre  x0 , y0  containing D. Since D is
bounded as R exists. We observe from equation (2.3) that

265
 v  x0 , y0   u  x0 , y0   M 0 . ... (2.4)

We show that v (x, y) like u (x, y) attains its maximum at a point  x0 , y0  in D. However, on B we
have

M M   x  x 2   y  y 2 
0 0
v ( x, y )  M  0 ,   1 
4  R2 

M0  M
 v  x, y   M   M0.
4

 the function v (x, y), like u (x, y) must attain its maximum at a point  x0 , y0  in D.

 v xx  0, v yy  0 at some point in D.
D
 v xx  v yy  0 at some point in D.
R
B
However, in D we have from equation (2.3)
x0 , y0 
M M
v xx  v yy  u xx  u yy  0 2 (2  2) ,
4R

M0  M
v xx  v yy  u xx  u yy  ,
R2
Since u is harmonic in D
 u xx  u yy  0 .

M0  M
 v xx  v yy  0, Since M 0  M
R2
 v xx  v yy  0 .

This is a contracdiction.
 the maximum of u must be attained on the boundary B.
Green Identity :
Let B be a closed surface in the space and D denote the bounded region enclosed by B.
Let F be a vector  C 1 in D and continuous on D. Then we know the Gauss divergence theorem is
given by

  Fnds
ˆ       F  dV
, ... (2.5)
B D

where dV is an element of volume, ds is an element of surface area and n̂ is the outward normal.
Green’s identity is obtained from (2.5).
266
Let F  f  , where f is a vector function and  is a scalar function. Then from equation (2.5), we
have

      f   dV    nˆ  f  ds
D B

we know    f    f     f

     f     f  dV    nˆ  f  ds ,
D B

    f  dV    nˆ  f  ds       fdV .
D B D

We choose the vector function

f   .
Therefore, the above equation yields

     dV     nˆ  ds       dV .
2
... (2.6)
D B D

Since n̂  is the derivative of  in the direction of n̂ . We denote this directional derivative by


n̂   .
n
Therefore, equation (2.6) reduces to


     dV     n
ds      2 dV . ... (2.7)
D B D

This equation is known as Green’s first identity.

Now interchanging the role of  and  , we obtain from (2.7) the equation


     dV     n ds      dV
2
... (2.8)
D B D

Now substracting (2.8) from (2.7) we get

     ds
        dV     
2 2
 ... (2.9)
D B
n n 

This is known as Green’s Second identity.

If in particular,    in equation (2.7) then we have

267
 
     
2
dV     ds      2 dV . ... (2.10)
D B
n D

Which is a special case of Green’s first identity.

Properties of Harmonic Functions :
Solutions of Laplace equation are called harmonic functions. These functions possess a number of
interesting properties.
Theorem 1 : If a harmonic function vanishes everywhere on the boundary then it is identically zero
everywhere.
Proof : Let  be a harmonic function in D.

  2  0 in D, ... (2.11)

and also   0 on B. ... (2.12)

We shall show that   0 in D  D  B .

We know the Green’s indentity is given by


     
2
dV     ds      2 dV . ... (2.13)
D B
n D

Using (2.11) and (2.12) we have from equation (2.13)

     
2
dV  0 . ... (2.14)
D

Since    2 is positive. It follows that the integral (2.14) will be satisfied only if   0 .

  = constant in D. But  is continuous in D  D  B and   0 on B, it follows from the maximum

and minimum principle that   0 in D.


Theorem : If  is a harmonic function in D and  0 on B, then  is a constant in D .
n
Proof : Let  be a harmonic in D.

  2  0 in D. ... (2.15)


Also  0 on B. ... (2.16)
n
Then we prove that  = constant in D  D  B .
Now by Green’s identity we have

268
 
     
2
dV     ds      2 dV ... (2.17)
D B
n D

Using equations (2.15) and (2.16) we have

     
2
dV  0 .
D

Since    2 is positive, it follows that the integral will be satisfied only if   0 .

  = constant in D.
Since the value of  is not known on the boundary B, but


 0 on B.
n
  = constant on B and hence by the maximum and minimum principle it is constant on D.
This proves the theorem.
Uniqueness Theorem :
Theorem : Prove that the solution of the Dirichlet problem, if it exists, is unique.
Proof : Let us suppose that u1 and u2 are two solutions of the Dirichlet problem.

  2u1  x, y   0 in D and

u1  x, y   f ( s ) on the boundary B, ... (2.18)

where f is a continuous function defined on the boundary B. Similarly, we have

 2u2  x, y   0 in D and

u 2  x, y   f ( s ) on B. ... (2.19)
Since u1 and u2 are harmonic in D, therefore u1 – u2 is also harmonic in D.

  2  u1  u 2   0 in D.
However, from equations (2.18) and (2.19) we have
u1  u2  0 on B ... (2.20)
By the maximum and minimum principle,
u1  u2  0 in D,

 u1  u 2 .

269
(Because if a harmonic function vanishes everywhere on the boundary, then it is identically zero
everywhere). This proves the uniqueness.
Other forms of Green’s Identity :
By Green’s theorem, we know, if u (x, y) and v (x, y) are differentiable functions in D and continuous
on the boundary B of D then

 U V 
  x  y  dS   Udy  Vdx  ... (2.21)
D B

 
Let U  and V  
x y

 U x   x x  xx , V y   y y  yy .
Therefore, equation (2.21) becomes

  x x  xx  y y  yy  dS    x dy   y dx  .
D B


We use  x dy   y dx  dS
n


Hence,   x x  xx  y y  yy  dS   n dS . ... (2.22)
D B

On interchanging  and  in (2.22) we get


  x x   xx   y y   yy  dS    n
dS . ... (2.23)
D B

Substracting (2.23) from (2.22) we get

 
       dS    n  
2 2   dS
 ... (2.24)
D B
n 

The identities (2.22) and (2.24) are called Green’s identities.

Theorem : Show the necessary condition for the existence of the solution of the Neumann problem is
that the integral of f over the boundary B should vanish.
Proof : Let u (x, y) be the solution of the Neumann interior problem.

  2u  0 in D, ... (2.25)
u
and  f ( s ) on B. ... (2.26)
n

270
Then we claim that  f ( s)ds  0 .
B

We know the Green’s identity is given by

  
       dS    n  
2 2 
 dS . ... (2.27)
D B
n 

Put   1 and   u in (2.27) we get

u
  uds   n ds
2

D B

Using equations (2.25) and (2.26) we get

 f ( s)ds  0 . ... (2.28)

B

This proves the result.

Theorem : Show that the solution of the Neumann problem is either unique or it differs from one
another by a constant only (i.e. solution is unique up to the addition of a constant).

Proof : Let u1  x, y  and u 2  x, y  be two solutions of the interior Neumann problem.

 u1 and u 2 are harmonics in D.

i.e.  2u1  0 , in D and  2u2  0 in D, ... (2.29)

u1 u 2
and  f on B and  f on B. ... (2.30)
n n

Then we claim that u1  u2 = constant.

Consider v  u1  u2

Then  2v   2  u1  u2    2u1   2u2

= 0 in D.

  2v  0 in D, ... (2.31)

v  u u
and   u1  u2   1  2
n n n n
=f–f

v
  0 on B. ... (2.32)
n
We know the Green’s identity
271
 
  x x  xx  y y  yy  ds   n ds . ... (2.33)
D B

Put     v in (2.23) we get

v
  vx   vvxx   v y   yy yy  ds   v ds ,
2 2

D B
n

v
  v x2  v 2y  v 2 v  ds   v ds ,
D B
n

v
   v  ds   v   2v  ds   v
2
ds
D D B
n . ... (2.34)

Using equations (2.31) and (2.32) we obtain from equation (2.34)

2
  v  ds  0 . ... (2.35)
D

Since  v  2 is positive. It follows that the integral (2.35) will be satisfied only if v  0 in D.

 v is constant in D.
 u1  u 2 = constant.
This proves that the solution of the Neumann problem differs from one another by a constant.
If constant is zero  the solution is unique.
Stability : A solution is said to be stable if it depends continuously on the initial and/or boundary data.
Stability Theorem : Show that the solution of the Dirichlet problem is stable.
i.e. Show that the solution of the Dirichlet problem depends continuously on the boundary data.
Proof : Let u1 and u2 be two solutions of the Dirichlet problem in D and f1, f2 be given continuous
functions on the boundary B of the region D.

  2u1  0 in D and u1  f1 on B.

Similarly  2u2  0 in D, u 2  f 2 on B.

Let v  u1  u2 .

  2v   2  u1  u 2    2u1   2u2  0 in D

  2v  0 in D
and v  f1  f 2 on B.

 v is a solution of the Dirichlet problem with boundary condition v  f1  f 2 on B.

272
Therefore, by the maximum and minimum principle, the harmonic function v attains the maximum and
minimum values on B.
Equivalently, f1  f 2 has maximum and minimum value on the boundary B. (i.e. f1  f 2 must be
bounded)

Thus if f1  f 2  on B.

Therefore at any interior point in D, we have, for given  0

  vmin  vmax 

 v  in D

 u1  u2  in D.

Hence if f1  f 2  on B then u1  u2  on D.
Thus, small changes in the initial data bring about an arbitrary small change in the solution.
This shows that the solution depends continuously on the boundary data.
i.e. the solution of the Dirichlet problem is stable.
3. Interior Dirichlet Problem for a Circle :
The Dirichlet problem for a circle is defined as follows.
Result : Show that the solution for the Dirichlet Problem for a circle of radius a is given by the Poisson
integral formula.
Example 1 : Find the value of u  r ,  at any point in the interior of the circle (r = a) D in terms of its
values on the boundary B such that u is single valued and continuous within and on a circular region and
satisfies the equation

 2u  0 , 0  r  a , 0    2
subject to u  a,   f   , 0    2 , where f   is continuous function on B.

Solution : Our problem is to solve for u  r ,  satisfying the equation

 2u  0 , 0  r  a , 0    2 , ... (3.1)
subject to the boundary condition

u  a,   f   , o    2 , ... (3.2)

where f   is continuous function on the boundary of circle.

We know the polar form of Laplace equation (3.1) is given by

 2u 1 u 1  2u
  0 ... (3.3)
r 2 r r r 2  2
273
We know the solution of the equation (3.3) is given by (Refer equation 1.31)

u  r ,     Cn r n  Dn r  n   An cos n  Bn sin n  . ... (3.4)
n 0

At r = 0, u  r ,  must be finite. Hence

r  n   as r  0  Dn  0 ,

 u  r ,    r n  An cos n  Bn sin n  , ... (3.5)
n 0

a0  n a
 u  r ,     r  an cos n  bn sin n  . for A0  0 ... (3.6)
2 n 1 2
Now using the boundary condition

u  a,   f   ,
we have

a0  n
f      a  an cos n  bn sin n  .
2 n1

This is the Fourier series expansion of f   , hence Fourier constants are given by

2
1
 a0   f   d ,
 0

2
1
an   f   cos n d ,
 an 0

2
1
and bn   f   sin n d , n = 1, 2, 3, .....
 an 0

Substituting these values in the solution (3.6) we get

2   cos n 2 2 
1 rn sin n
u  r ,    f   d    n   f   cos n d   f   cos n d  .
2 0 n 1 a   0
 0 

Interchanging the order of summation and integrating we get

2 2  n
f      cos n cos n  sin n sin n  d ,
1 1 r
u  r ,  
2  f   d    n 1  a 
0 0

274
 2
1  r n 
f        cos n     d .
1
u  r ,    ... (3.7)
 0  2 n 1  a  
Consider
 n  n
  a  cos n      i sin n         a  ein   ,
r r
n 1 n 1

 n  n
  a  cos n      i sin n         a ei    .
r r
... (3.8)
n 1 n 1

r
Since ra  1 and e i    1 .
a
The expression on the right hand side of the equation (3.8) is a geometric series.

 r  e i  
  
n
 r e i      a 
Therefore,   a   r  i  
.
n 1 1   e
a

Equating the real part on both sides we get

  r  i   
n  e
 r  cos n      Re   a  

  a   r ,
n 1 1  e i   
 a 

  r  i    r i    
  e 1 e
 a  
 Re   a  ,
 r
 1  e      r      

i i
1  e
  a  a  

  r  ei    r  
   
 Re   a  a 
2 ,
1  r  ei    e i     r 
 a a2 

r  r 
 a cos      i sin      a  
 Re  ,

 1  r 2 cos      r 
2

  a      
a  

275
 2
 r  cos       r 
 n    
    cos n       a  a
r .
2
n 1  a   r  r
1   2  cos       
 a a

Substituting this on the r.h.s. of (3.7) we get

 r r2 
2 1 cos       
u  r ,  
1
f     a a 2  d ,
   2 1  2r cos      r 
2
0
 a a 2 

1
2
a2  r2 
u  r ,    f   d . ... (3.9)
 0 2  a 2  2ar cos      r 2 

This is known as Poisson integral formula for a circle, which gives the unique solution for the Dirchlet
problem.
4. The Dirichlet Exterior Problem for a Cicle :
Result : Show that the solution for the exterior Dirichlet problem for a circle of radius a is given by
(Poisson integral formula)

1
2  r 2  a 2  f  
u  r ,    r 2  2ar cos      a 2  d .
2 0  
Solution : Exterior Dirichlet problem is decribed by

 2u  0 for 0    2 , r  a , ... (4.1)

and u  a,   f   , 0    2 , r = a, ... (4.2)

where f   is a continuous function of  on the surface r = a, and u  r ,  must be bounded

as r   . We know by the method of separation of variables, the general solution of (4.1) in polar
co-ordinates is given by

u  r ,     Cn r n  Dn r  n   An cos n  Bn sin n  ... (4.3)
n 0

Now as r   we require that u  r ,  to be bounded

 Cn  0  As r n   as r   
Hence the general solution (4.3) reduces to

276
 
u  r ,    r  n  An cos n  Bn sin n  . ... (4.4)
n0

It can also be written as

a0   n
u  r ,     r  an cos n  bn sin n  . ... (4.5)
2 n 0

Now using the boundary condition

u  a,   f   ,
a0   n
 f      a  an cos n  bn sin n  .
2 n1
Hence the Fourier constants are given by
2
1
a0   f   d ,
 0

2
an
an   f   cos n d ,
 0

2
an
bn   f   sin n d . n = 1, 2, 3, .....
 0

Substituting these values in equation (4.5) we get the solution as

2 2 2
an  

1
u  r ,    f   d     cos n  cos n f   d  sin n  sin n f   d  .
n
r
2 0 n 1   0 0 

Interchanging the order of summation and integration we get

2 2  n
f      cos n cos n  sin n sin n  d .
1 1 a
u  r ,  
2  f   d     n 1  r 
0 0

2
1  1   a n 
u  r ,    f        cos n     d . ... (4.6)
 0  2 n 1  r  
Consider,

 n  n
  r  cos n      i sin        r  ein   .
a a
n 1 n 1

277
 n n
 a  cos n      i sin        a  ei   
  n
i.e.  r  
        ... (4.7)
n 1 n 1  r  
Since r > a

a
  1 and e i    1
r
The expression on the right hand side of the equation (4.7) is a geometric series. Therefore, we have

a i  
 n e
 a  i   
  r  e    a i    ,
r

1  r e
n 1


a i    a i   
r
e 1  r e 
 ,
1  a e i    1  a e i   
 r   r 

a  i   a 
e  
 r  r ,
 a  i   i    a 2 
1  r e e  2
 r 

a
cos      i sin      
a
 n
 a  i    
  r  e   r   a a 2
r.

1  2 r cos      2 
n 1
 r 

Equating the real part on both sides we get

a
cos      
a
 n

  r  cos n       r a r .
a
a2 
n1
1  2 cos       
 r r2 

Substituting this in equation (4.6) we get

 a a2 
2 1 cos      2 
1
u  r ,    f     r r  d .
 0  2 1  2a cos      a
2

 r r2 

278
 2
1 r 2  a2
 u  r ,    f   d ,
 0 2  r 2  2ar cos      a 2 

1
2  r 2  a 2  f  
or u  r ,     r 2  2ar cos      a 2  . ... (4.8)
2 0

This is the required solution of the exterior Dirichlet Problem.

5. Interior Neumann Problem for a Cicle :
The interior Neumann Problem for a circle is described as follows.
Example 1 : Solve

 2u  0 , r < a
subject to the boundary condition

u u
  f   on r = a ,
n r
(Because outward normal to the Ole is along the radius vector)
2

where  f   d  0 .
0

Solution : To find the solution, we solve the equation

 2u 1 u 1  2u
   0, r<a ... (5.1)
r 2 r r r 2  2
subject to the boundary condition

u
 f   on r = a, 0    2 , ... (5.2)
r
where f   is a continuous function of  on the surface r = a.
We know by the method of spearation of variables, the general solution of equation (5.1) is given by

u  r ,     Cn r n  Dn r  n   An cos n  Bn sin n  . ... (5.3)
n 0

Since at r = 0, the solution should be finite hence we must have Dn = 0 ( r  n   as r  0 ).

Hence the solution becomes

u  r ,    r n  An cos n  Bn sin n  .
n 0

279
This can be written as

a0  n
u  r ,     r  an cos n  bn sin n  . ... (5.4)
2 n 1

Differentiating (5.4) w.r.t. r we get


u
  nr n1  an cos n  bn sin n  . ... (5.5)
r n 0

Now using the boundary condition

u
 a,    f   ,
r

we have f     nr n1  an cos  n   bn sin  n   . ... (5.6)
n 0

This is a Fourier series expansion of f   , where the Fourier constants are given by

2
1
an   f   cos  n  d ,
n a n1 0

2
1
bn   f   sin  n  d .
na n 1 0

Substituting these values in (5.4) we get

2
a0  r n
u  r ,     f   cos n cos n  sin n sin n  .
2 n 1 n a n1 0

Interchanging the order of summation and integration, we get

2  n
a0 1
f     cos n     d  .
r a
u  r ,     ... (5.7)
2  0 n 1  a  n

Now consider

 n  n
 r  1
  a  n ein      a  ei    n ,
r 1
n 1 n 1

 r i  
     ... ,
2 3
r i   r i  
 e e e
  a  a  a
 1 2 3 

280
  n
 x 2 x3 
  a  n ein     log 1  a ei    ,
r 1 r
as log 1  x     x    ...
n 1  2 3 

 n
  a  n ein     log 1  a cos      i a sin     .
r 1 r r
... (5.8)
n 1

[ Now to find the real part of log z, for z = x + iy

let W  u  iv  log z ,

 z  x  iy  e u  iv ,

 x  e u cos v, y  e y sin v ,

 e 2u  x 2  y 2

 u  log x 2  y 2 ].
Therefore, equating the real part on both sides of (5.8) we get

 n 2 2
  a  n cos n       log 1  a cos        a sin      ,
r 1 r r
n 1

a 2  2ar cos      r 2
  log .
a2
Substituting this in equation (5.7) we get

2
a a a 2  2ar cos      r 2
u  r ,   0   log  f   d  . ... (5.9)
2  0 a2

Thus the required solution of interior Neumann Problem for a circle can also be written as
2
a a
u  r ,   0   log  a
2
 2ar cos      r 2   f   d . ... (5.10)
2 2 0

The constant factor a2 in the argument of log was eliminated by virtue of the necessary condition for the
Neumann problem.
6. Exterior Neumann Problem for a Circle :
Result : State the exterior Neumann problem and show that its solution for a circle of radius a is given
by
2
a0 a
u  r ,     log  a
2
 2ar cos      r 2   f   d .
2 2 0

281
Proof : The exterior Neumann problem for a circle is described by

 2u  0, r  a, 0    2 , ... (6.1)
subject to the condition

u u
  f   , on the boundary r = a. ... (6.2)
n r
By the method of separation of variables, we know the general solution of (6.1) in polar form is given
by

u  r ,     Cn r n  Dn r  n   An cos  n   Bn sin  n   . ... (6.3)
n 0

Now as r   we require that to u  r ,  be finite (bounded)

 Cn  0 . (as r n   as r   )
Hence the general solution (6.3) reduces to

u  r ,    r  n  An cos n  Bn sin n  . ... (6.4)
n0

Without loss of generality, it can also be written as

a0   n
u  r ,     r  an cos n  bn sin n  . ... (6.5)
2 n 1

Differentiating equation (6.5) w.r.t. r we get


u
    n  r  n 1  an cos n  bn sin n  . ... (6.6)
r n 1

Now using the boundary condition

u
 a,    f   ,
r

we get f       n  a  n1  an cos n  bn sin n  . ... (6.7)
n 1

This is the Fourier series expansion of f   , where the Fourier constants are given by
2
1
an  n  a  n 1   f   cos n d ,
 0
2
a n 1
 an    f   cos n d ,
n 0

282
 2
a n 1
and bn    f   sin n d .
n 0

Substituting these constants in equation (6.5) we get

2
a0  a n 1
u  r ,     r  n  f   cos n cos n  sin n sin n  d .
2 n 1 n 0

Interchanging the order of summation and integration we get

2  n
a0 a
  r   n f   cos n     d ,
a 1
u  r ,    
2  0 n 1

2  n
a
f     
a a 1
u  r ,   0   cos n     d . ... (6.8)
2  0 n 1  r  n

Now consider

 n  n
 a  1
  r  n ein n    r  ei n   n
a 1
n1 n 1

   
 a i n   a  i n  
2
a i   n 
3 
 e   e  e 
 r  r   r
    ...
 1 2 3 

  log 1  ei n  
a  
...  log 1  x     x  x  x  ... 
2 3
 r   
  2 3 
 n
  r  n ein     log 1  r cos      i r sin      .
a 1 a a
n 1

Equating the real part on both sides we get

 n
  r  n cos n       log 1  r cos      i r sin     ,
a 1 a a
n 1

 n
    cos n       log 1  cos        sin      ,
a 1 a 2 a 2

n 1  r  n  r  r 



n
a 1 r 2  2ar cos      a 2
    cos n       log .
n 1  r  n r2

283
Substituting this in equation (6.8) we get

2
a a r 2  2ar cos      a 2
u  r ,   0 
2   f   log r2
d ,
0

2
a a  r 2  2ar cos      a 2 
or u  r ,   0   log   f   d  . .... (6.9)
2 2 0  r2 

This is the required solution.

7. Interior Dirichlet Problem for a Rectangle :
Result : Solve

 2u  0, 0  x  a, 0  y  b , ... (7.1)
subject to the boundary conditions

u  x, b   u  a, y   0 , ... (7.2)

u  0, y   0 , ... (7.3)

u  x, 0   f  x  , 0 xa ... (7.3)

Solution : We assume a variable separable solution of the form
y
u  x, y   X  x  Y ( y ) . ... (7.5)
y=b

Therefore, equation (7.1) becomes u=0

u=0
X '' Y  XY ''  0 , u=0

u (x, 0) = f (x)
X '' Y '' x
   (say), O y=0 x=a ... (7.6)
X Y x=0

where  is a constant, may be positive, zero or negative. For different choices of  we have three
solutions. We have to choose that solution which is consistent with the physical nature of the problem
and the boundary conditions.
Case (i) :   0 , Take    2 .
Then we have the equations

X ''  2 X  0 and Y ''  2Y  0 .

Whose solutions are given by

X  C1e x  C2 e  x , Y  C3 cos  y   C 4 sin  y 

Therefore, the general solution of (7.1) is given by

u  x, y    C1e x  C2 e  x   C3 cos  y   C4 sin  y   . ... (7.7)

284
Now using the boundary condition

u  0, y   0 ,
we get from equation (7.7)

0   C1  C2   C3 cos  y   C4 sin  y    0 ,

 C1  C2  0 . .... (7.8)
Again using the boundary condition

u  a, y   0 ,
we get from equation (7.7)

0   C1e x  C 2 e  x   C3 cos  y   C4 sin  y   ,

 C1e a  C2e  a  0 . ... (7.9)

From equations (7.8) and (7.9) we have C1  0  C2 .

Hence u  x, y   0 is the only possible trivial solution. Hence we neglect the case   0 .
Case (ii) : If   0 , then from equation (7.6) we have
X ''  0 and Y ''  0 .
This provides

X   C5 x  C6  and Y   C7 y  C8  .

Hence the general solution of (7.1) is given by

u  x, y    C5 x  C6  C7 y  C8  . ... (7.10)
Using the boundary conditions

u  0, y   0 and u  a, y   0 ,
we get from (7.10)
C 6  0  C5

 u  x, y   0 is a trivial solution. Hence we discard   0 .

Case (iii) : If   0 , Take    2 .

Hence from equation (7.6) we have

X ''  2 X  0 and Y ''  2Y  0 ,

285
which have solutions

X  C9 cos  x  C10 sin  x and Y  C11e y  C12 e  y .

Therefore, the general solution of (7.1) is given by

u  x, y    C9 cos  x  C10 sin  x   C11e y  C12 e  y  . ... (7.11)

Now using the boundary condition

u  0, u   0 ,

we obtain C9  0 .

Also the boundary condition u  a, y   0 yields

0  C10 sin  a   C11e y  C12 e  y  .

If C10 = 0, we will have again a trivial solution. Therefore, we assume
C10  0

 sin  a  0 ,
  a  n , n  1, 2,...

n ,
or 
a

Take n . n  1, 2,... ... (7.12)

n 
a
These are called the eigen values. Hence the possible non-trivial solution is given by

n  n  n  
u n  x, y   C10 sin  x   an exp  y   bn exp   y
 a   a   a 
By superposition principle, the most general solution of (7.1) is given by

u  x, y    u n ( x , y )
n 1


n   n  n  
Hence u  x, y    sin  x   an exp  y   bn exp   y ... (7.13)
n 1  a   a   a  

Now using the boundary condition

u  x, b   0 ,
we have from equation (7.13)

286
 n  n  n  
0  sin  x   an exp  y   bn exp   y ,
 a   a   a  

n b   n b   0
 an exp    bn exp    ,
 a   a 

n b 
exp  
 bn  an  a  , n = 1, 2, ...
n b 
exp   
 a 

Substituting this in (3.13) we get

n x 
an sin 
   
 
u  x, y     a  exp n ( y  b)  exp  n ( y  b) 
n1 exp  n x   a a  ,
 
 a 

n x 
2an sin 
 
   e x  e x 
  a   sinh n ( y  b) ,
 Since sinh x  
n 1 exp  
n x  a  2 
 
 a 

 

n x  n
 u  x, y    An sin   sinh ( y  b) , for An  2a n
. ... (7.14)
n 1  a  a  n b 
exp   
 a 
Now using the boundary condition

u  x, 0   f ( x ) , 0 xa,
we have from equation (7.14)

n x   n b 
f ( x)   An sin   sinh    ... (7.15)
n 1  a   a 
This is a Fourier sine series, where the Fourier constant is given by
a
n b  2  n x  dx
An sinh      f ( x)sin   ,
 a  a0  a 
a
n  .
f ( x)sin 
2
or An  
n b  0
x  dx ... (7.16)
a sinh   
 a 
 a 
Thus the general solution for the Dirichlet problem for a rectangle is given by

287
  

n n
u  x, y    An sin  x  sinh ( y  b) ,
n 1  a  a

a
n
f ( x)sin  x  dx .
2
where An 
  ... (7.17)
a sinh   0  a 
n b

 a 
8. The Neumann Problem for a rectangle
Result : Solve the equation

 2u  0 0  x  a , 0  y  b , ... (8.1)
subject to the boundary conditions.
u x (0, y )  u x (a, y )  0 , ... (8.2)

u y  x, 0   0 , ... (8.3)

u y  x, b   f ( x) . ... (8.4)
Solution : By variable separable method, we have obtained the general solution of equation (8.1) in
the form

u  x, y    C1 cos  x  C 2 sin  x   C3e y  C 4 e  y  . ... (8.5)

Differentiating equation (8.5) w.r.t. x and y we get

u x ( x, y )   C1 sin  x  C 2 cos  x   C3e y  C4 e  y  . ... (8.6)

Now using the boundary condition (8.2) viz.

u x (0, y )  0 gives

C2  0 .
Therefore, equation (8.5) becomes

u  x, y   C1 cos  x  C3e y  C4 e  y  . ... (8.7)

Now the boundary condition

u x  a, y   0 gives

0   C1 sin  a  C3e y  C 4 e  y  .

If C1 = 0 then we get trivial solution of (8.1). Therefore, for non-trivial solution, we assume C1  0 .

 sin  a   0 ,
  a  n , n = 1, 2, 3, ....
288
 n .
  n = 1, 2, 3, ...
a
Let for each n,

n
n  , n = 1, 2, .... ... (8.8)
a
These are called the eigen values. Thus the possible solution is given by putting  n in equation (8.7).

u  x, y   cos 
n 
x  Ae  n
a
y
 Be

n
a
y
. ... (8.9)
 a 
Differentiating (8.9) w.r.t. y we get
n n
n  n   y  y
u y ( x, y )  cos  x   Ae a  Be a  .
a  a 
Now using the boundary condition
u y ( x, 0)  0

n n 
we have 0 cos  x  ( A  B)
a  a 
 A  B  0  A = B.
Thus the solution (8.9) becomes

n  n  n  
u  x, y   A cos  x   exp  y   exp   y ,
 a   a   a  

n n 
u  x, y   2 A cos  x  cosh  y.
 a   a 

Using the superposition principle and for 2 A  An , we write the general solution of (8.1) as


n n
u  x, y    An cos  x  cosh  h  . ... (8.10)
n 1  a   a 

Finally using the boundary condition (8.4)

u y ( x, b )  f ( x ) ,

we have

n n  sinh  n
f ( x)   An cos  x      b  ,
n 1  a   a   a 

which is the Fourier cosine series. The corresponding Fourier constant is given by

289
 a
 n  A  sinh  n b   2 f ( x) cos  n x  dx
 a  a 0
  n    
 a   a 

a
n
  f ( x) cos  x  dx .
2
 An  ... (8.11)
n
n sinh  b  0  a 
 a 
Hence the required solution of the Neumann problem for a rectangle is given by

n n 
u  x, y    An cos  x  sinh  y,
n 1  a   a 

where the constant An is given by

a
n 
f ( x) cos 
2
An 
 
n b  0  a
x  dx .

... (8.12)
n sinh  
 a 
9. The Dirichlet Problem for the Upper Half Plane
Result : Find the solution of the problem

 2u  0 ,   x  , y  0 ,
u  x, 0   f ( x) ,   x   ,

such that u is bounded as y   , u and ux vanish as x   .

Solution : Given that
u xx  u yy  0 ,   x  , y > 0, ... (9.1)

u ( x, 0)  f ( x ) ,   x   , ... (9.2)

with the conditions that u is bounded as y   and u and ux vanish as x   .

We use the technique of Fourier transform to solve the problem. Let U  , y  be the Fourier transform
of u (x, y) in the variable x. Therefore, by definition we have

1
U  , y   F u ( x, y   u ( x, y )e
i x
dx . ... (9.3)
2 

Now applying the Fourier transform to equation (9.1) we get

 2   2 
F   u2   F   u2   0 .
 x   y 

290
Since Fourier transform for derivative is given by

 
F f ( n ) (n)   i  F  f ( x) .
n

  2u 
  i  F u ( x, y )  F  2   0 .
2

 y 

  2U  , y   U yy  0 ,

i.e. U yy   2U  0 . ... (9.4)

Its solution is given by

U  , y   A   e y  B   e  y . ... (9.5)

Since we require that the solution U  , y  be bounded as y   , therefore, for   0 , we must

have A    0 , and for   0 , B    0 .
Therefore, we have

U  , y   U  , 0  e   y , ... (9.6)

where U  , 0   F u  x, 0 

 F  f ( x) by equation (9.2)

 U  , 0   F   , by definition of Fourier series.

Hence U  , y   F     e   y . ... (9.7)

Also by definition of inverse Fourier transform, we have

F 1
e  
 y 2 y 
  y  x 2  .
 2 ... (9.8)

We write equation (9.7) on using (9.8) as

 
F u  x, y   F  f ( x) F  2   2 2   ,
y
   y  x  

 2  y  
F u  x, y   F  f *  2  Since F  f * g   F ( f )  F ( g ) 
   y  x 2  

Taking inverse Fourier transform on both sides we get

291
 2 y 
u  x, y   f *
  y  x 2  ,
 2


1 2 y 
u  x, y  
2  f     2 
  y  x   
2
d (by convolution theorem.)


f  

y
u  x, y    d . ... (9.9)
  y2   x   
2

This is the required solution of the Dirichlet problem for the upper half plane.
10. The Neumann Problem for the Upper Half Plane :
Result : Find the solution of the problem

 2u  0 ,   x  , y  0 ,

u y ( x, 0)  g ( x ) ,   x   ,

such that u is bounded as y   , u and ux vanish as x   and  g ( x)dx  0 .



Solution : We reformulate the problem by introducing a new variable v (x, y) as

v  x, y   u y  x, y  . ... (10.1)

Then u   v  x,  d ... (10.2)



  2 
Also  2 v  x, y    2u y   u  0,
y

and v  x, 0   u y ( x, 0)  g ( x) .

Thus our problem is reformulated in to the new variable v as

 2 v ( x, y )  0 ,   x   , y > 0, ... (10.3)

v  x, 0   g ( x ) ,   x   . ... (10.4)
Since u is bounded as y    v is also bounded as y   .
Integrating equation (10.1) with respect to y we obtain
y
u  x, y    v( x, )d
a

292
 y
Since, u is bounded as y     v  x, d is bounded as y   .
a

 the integrand v (x, y) is bounded on y   .

Also from equation (10.1) we find
 
v x  x, y   u y  ux
x y

 lim v x  x, y   lim u x ( x, y )  0 .
x  y x 
 v x ( x, y )  0 as x   .
Also from equation (10.2), we have
y
lim u   lim v( x, )d  0
x  x 
a

 v  0 as x   .
However, we know the solution of the problem is given by

g  

y
v ( x, y ) 
    x  2  y 2 d  . ... (10.5)


Hence the solution of the original problem becomes

y
u ( x, y )   v  x, d .


On using (10.5) we get

g  
y 
1
     x  2   2
u ( x, y )   d d
 

y
1 2 1    x  2   2 
y
Consider 
2    x  2   2
d 
2
log   ,

y
 1    x  2  y 2 
 d  log  ,
   x       x    2 
2 2
2 2

1
y    x  2  y 2 
u ( x, y ) 
2  g   log    x 2   2  d . ... (10.6)
  
This determines the solution of the problem.

293
 UNIT - VIII

RIEMANN'S METHOD OF SOLUTION OF LINEAR

HYPERBOLIC EQUATIONS

Introduction :
In this unit we shall discuss Riemann's method of solving a linear second order hyperbolic
partial differential equations which are in the canonical forms. The method is illustrated in the following
theorem.
Theorem : Describe Riemann's method of solving a linear second order hyperbolic equation .
Proof : Let
L[u ]  u xy  au x  bu y  Cu  f ( x, y ) , ... (1.1)
be a linear, second order hyperbolic equation, which is in a canonical form, where a, b, c are functions
of x and y only .
Define another operator M such that
M [u ]  vxy  ( av ) x  (bv ) y  cv, ... (1.2)
where v (x) is a function having continuous second order partial derivatives. The operator M is called
the adjoint operator of L.
Consider

vL u   uM  v   v[u xy  au x  bu y  cu ]  u[vxy  (av) x  (bv) y  cv],

   
 Nu xy  uvxy   vau x  u ( av ) x   vbu y  u (bv) y .

We write
vu xy  uvxy  (vu x ) y  (uv y ) x ,

vau x  (avu ) x  u ( av) x ,

vbu y  (bvu ) y  u (bv ) y .

Therefore

  x   avu  x   bvu  y ,
vL[u ]  uM [v ]   vu x  y  uv y


 avu  uv y  x   bvu  vux  y
 vL[u ]  uM [v]  U x  V y , ... (1.3)
294
where
U  avu  uv y ,

V  bvu  vu x . ... (1.4)

y
Let P ( ,  ) be a point at which the solution is
to be found. Let the characteristics through P intersect
the initial curve  at Q and R. We assume that u , u x , u y Q P  ,  
D
are prescribed along  . Let C be a closed contour
PQRP bounding the region D. We now apply Green's
theorem to this region.
Now from equation ( 1.3) we have
R F
x

  vLu  uMvdxdy   U x  Vy dxdy .

D D

Using Green's theorem we write

  vLu  uMv dxdy   (Udy  Vdx)

D c

R P Q

  vLu  uMvdxdy   (Udy  Vdx)   Udy  Vdx)   (Udy  Vdx) ... (1.5)
D Q R P

Now along PQ, y is constant  dy  0 and along PR , x is constant and hence dx = 0.

Therefore, above equation becomes
R P Q

 (vLu  uMv)   (Udy  Vdx)   Udy   Vdx ... (1.6)

D Q R P

Now consider
Q Q Q

 Vdx   bvudx   vux dx by equation ( 1.4)

P P P

Q R
  bvudx  uv P   uvx dx
Q

P P

Q Q

 Vdx  uv P   u (bv  vx )dx

Q

P P

295
Substituting this in equation (1.6) we get
R P Q

 (vLu  uMv)dxdy   (Udy  Vdx)   Udy  uvQ  uv p   u (bv  vx )dx ,

D Q R p

Q R

 uv  p  uv Q   u  bv  vx  dx   u av  v y dy  
P P

R
  Udy  Vdx    (vLu  uMv)dxdy . ... (1.7)
Q D

The function  is quite arbitrary. We choose the function  such that it is the solution of the adjoint
equation M[v] = 0 satisfying the conditions
 x  b on y   , ( i.e along PQ)

 y  a on x   , (i.e along RP)

and   1 at x   and y   . ( i.e at point P) ... (1.8)

Such a function   x, y,  ,   , if it exists, is called a Riemann function or Green's function.

Hence equation (1.7) reduces to

R
[u ] p  uv Q   (Udy  Vdx)   vLu dxdy ,
Q D

R
u  p  uv Q   (a u  u y )dy  (b u  ux )dx    vf dxdy
Q D

R R
[u ] p  uv Q   uv(ady  bdx)   (uv y dy  vu x dx)   vf dxdy ... (1.9)
Q Q D

Equation (1.9) finds u at p provided u and ux are prescribed along the curve  . However, when u and
uy are prescribed along  , then to find u at p we use the identify
R R

 d (uv)   (uv) x dx  (uv) y dy  , by chain rule

Q Q

R
uv Q  uv R   (uv) x dx  (uv) y dy  . ... (1.10)
Q

296
Substituting this in ( 1.9) we get
R R
u  p  uv R   uv(ady  bdx)    uvx dx  vu y 
dy   vf dxdy ... (1.11)
Q Q D

On adding equations (1.9) and (1.11) we get

R
1
[u ] p  [uv]Q  [uv]R    uv(ady  bdx)   vfdxdy +
2 Q D

1R 1R

2 Q
 uvy dy  vux dx   2   uvx dx  vu y dy  . ... (1.12)
Q

By using any of the equations (1.9), (1.11) and (1.12) whichever is suitable, value of u at p can be
obtained provided values of u, ux or uy are known along the curve  .
Example (1) : Show that for the linear hyperbolic equation

u
u xy   0,
4
the Riemann function is

v  x, y ,  ,    J 0   x    y     ,
where Jo is the Bessel function of the first kind of order zero.

Solution : The linear hyperbolic equation is given by

u
L[u ]  u xy   0. ... (1.13)
4
Comparing this equation with the standard linear hyperbolic equation (1.1) we have

1
a  o, b  o, c  and f ( x, y )  0 .
4
Hence the adjoint operator M is defined by

v
M [v]  vxy  . ... (1.14)
4
We see that
M L
 L is self adjoint.

297
Therefore, we have

vL u   uM v   v  u xy    u  v xy   ,
u v
 4  4

= vu xy  uvxy ,

= (vu x ) y  (uv y ) x ,

vL[u ]  uM [v]  U x  V y . ... (1.15)

We choose the Riemannian function in such a way that
M [v ]  0 .

vx  0 on y,

vy  0 on x  ,
and v 1 at x   and y   .
Let v  v ( ) ,
where  is a single valued differentiable function of x and y .

Let  2   x    y    ,

 2 x  ( y   ) ,

1
 x  (y  ) ,
2

1
and y  (x  ) .
2
Hence, we have

1
vx  v ( y   ) ,
2

1
vy  v ( x   ) .
2

 vx  v x , v y  v y ,

1 1 1 1 
 vxy  v y  y     v  v 2 ( y   ) y  ,
2    

298
 1 1  v
 vxy   v  v  .
4  
vxy   0
4

 the equation is transformed to the ordinary differential equation

1
v  v  v  0 , ... (1.18)

where v( ) satisfies v(o)  1 . Equation ( 1.18) is the Bessel equation of order zero, whose solution is
given by J0 ( ),

 v  Jo  
( x   )( y   ) .

This is the required Riemann function .

Example (2) : Show that the Green's function for the equation u xy  u  o is


v  x, y;  ,    J 0 2 ( x   )( y   ) , 
where J0 is the Bessel's function of first kind and of order zero .
Solution : Here the linear hyperbolic second order partial differential equation is given by
L (u )  u xy  u  0 . ... (1.19)
Comparing this equation with the standard canonical hyperbolic equation we get
a  0, b  0, c  1 and f ( x,y) = 0.
The adjoint operator M is given by
M [v ]  vxy  v  0 . ... (1.20)
We see that M = L, proving L is self adjoint.

Hence vL u   uM  v   v  u xy  u   u  v xy  v  ,

 v u xy  u vxy ,

 vL[u ]  uM [v]  (u u y ) x  (u v x ) y ,

 U x  Vy ,

where U  uu y , V  uvx . ... ( 1.21)

We choose v such that M [v ]  0 ,

vx  0 on y   ,

vy  0 on x  , ... (1.22)

299
and v 1 at x  , y   .
Let v  v ( ),
where  is a single valued differentiable function of x and y..

Let  2  4( x   )( y   )

2 2
 x  y and   y   x   .
 
Therefore

2
vx  v x  v ( y   )


2
and vy  v ( x   )


2 2 2
Thus vxy   y v ( y   )  2 v y ( y   )  v
  

4 4 2
2  
 vxy  v x  (y   )  v ( x   )( y   ) 
3 
v ,
  

1
 vxy  v  v . ... ( 1.23)

Thus the equation
vxy  v  0 ,

is transformed to the equation

1
v  v  v  0 . ... (1.24)

This is a Bessel equation of order zero, whose solution is given by

v  x, y;  ,    J 0  

v ( x, y;  ,  )  J o 2   x    y     . ... (1.25)

Example 3 : Show that

 x  y   2 xy  (   )( x  y)  2 
v ( x, y;  ,  ) 
    3
300
is the Riemann function for the second order p.d.e

2
u xy 
x y
ux  u y   0 .
Hence obtain the solution of the equation in the form


v  2 y 3  3 y 2 x  3 yx 2  2 x3 , 
subject to u 0, ux  3x 2 on y  x .
Solution : A linear second order hyperbolic equation is given by

1
L[u ]  u xy 
x y
 ux  u y   0 , ... (1.26)

2 2
where a , b , c  0, f ( x, y )  0 .
x y x y
The operator M is defined by
 2   2 
M [v ]  vxy   v  v , ... (1.27)
 x  y x  x  y  y
such that
i) M [v ]  0 ,

2
ii) vx  v on y, ... (1.28)
x y

2
iii) vy  v on x  ,
x y

iv) and v  1 at x   y.

Now to show

 x  y   2 xy      x  y   2 
v  x, y;  ,    ... ( 1.29)
    3
in a Riemann's (Green's) function, we simply show that v defined in ( 1.29) must satisfy the equation
(1.27) and the addition ( 1.28). We find

 x  y 1
3
vx  2y      2 xy       y   2  ,
        3

301
 1
vx   2 xy  2 y 2  2 xy      x  y  x  y   2 
 ,
    3

1  4 xy  2 y 2  2 x      2 
vx  . ... (1.30)
    3

1
Next vxy  4 x  4 y
   3
4 x  y
 vxy  . ... (1.31)
    3
1
vy   4 xy  2 x 2  2 y      2 
Also   ... (1.32)
    3

Now consider

2 4
M [v]  vxy  (v x  v y )  v ... (1.33)
x y  x  y 2
On using equations ( 1.30), (1.31), and ( 1.32) in ( 1.33) we get

4 x  y 2 1
M [v ] 
    3

 x  y     3   
8 xy  2 x 2  y 2  2  x  y      4 
 +

4
  x  y   2 xy      x  y   2  ,
 x  y 2    3
4 x  y 4
M [v ]    x  y 2
    3
 x  y     3

 M [v ]  0 .

 v satisfies the condition ( i ) in ( 1.28 ). Now along y   equation (1.30 ) becomes

1
Vx  y    4 x  2 2  2 x      2 
 .
    3

1
Vx  y    2 2  2 x      2 
  ... (1.34)
    3

302
  2  1
Also  v   4 x  2     x     4 
3 
 x  y  y      

1
  2 x      2      4 
    3

 2  1
 v   2  2  2 x      2 
3  . ... (1.35)
 x  y  y      

Equations ( 1.34 ) and ( 1.35 ) show that the condition (ii) of equation (1.28) is satisfied.
Similarly condition (iii) can be verified .
Now consider at x   , y  

   
v
|p   2          2  ,
    3 y
P  ,   y=x
1
v
|p   4   2  2   2  R  ,  
 
    2

v
|p  1
Q  ,  
This shows that the equations ( iv) of ( 1.28 ) also verified.
Hence x

x y
v  x, y;  ,     2 xy      x  y   2 
    3
in the Riemann function of the given p.d.e. (1.26). Now to find it solution , we consider
vL[u ]  uM [v]  U x  V y ,

2
where U vu  uv y , ... (1.36 )
x y

2
and V vu  vu x ... (1.37 )
x y
It is given that along the curve  : y  x that is along QR

u  o and u x  3x 2 . ... (1.38 )

Also along QR the Riemann's function is given by

303
 2x
v  2 x 2  2 
 
    3

v

4 x x 2    . ... (1.39)
    3

We know the value of u at P in given by

R R
u  p  uv Q   uv  ady  bdx     uv y dy  vux dx    vfdxdy ,
Q Q D

where in this carve f ( x, y )  0 . Using ( 1.38 ) and ( 1.39 ) we get

R
[u ] p   vu x dx ,
Q


R

4 x x 2    3x dx 2
,
Q     3


12
x 
5
  x3 dx ,
     3


12  x6 x 4 
     ,
   3  6 4 

1

    3   
 2  6   6  3  4   4 
,  
1
  2 6  2 6  3 5  3 5  
 .
    3

1
u ( ,  )   2 6  2 5   5  2 6  2 5    5  
 ,
    3

1

    3
 2 5      2 5        4   4 
 ,  

304
 1

    3    
 2      5   5    4   4 
. 
  
3 
 2      2  4 2  3  2  2 2  2  3  2 4   3   3   ,
   

    4
 2 
2   3 3  3 3   2 2  2  2 4  ,
   


u(, )     22  22 , 
u  ,    2  3  2 2   2   2   2 2   2 3 ,

u  ,    2  3  3 2  3 2   2 3 .

Thus u at any point ( x,y) is given by

u ( x, y )  2 y 3  3 y 2 x  3 x 2 y  2 x 3 .
Note : We see that the solution of the Cauchy problem at a point ( ,  ) depends only on the Cauchy
data on the curve  . The knowledge of the Riemann's -Green function therefore enables us to solve
the p.d.e with the carve data .

Harnack's Theorem :
Let us prove the following lemma first.

Lemma : Let D be a bounded domain, bounded by a smooth closed curve B. Let un  x, y  be a

sequence of functions each of which is continuous on D  DUB and harmonic in D. If un  x, y 

converges uniformly on B, then u n converges uniformly on D .

Proof : Let un  be a sequence of functions, converges uniformly on the boundary B. Then by
definition, for given  o we can always find N such that

um ( x, y )  un ( x, y)  on B m , n  N .

Since each of un ( x, y ) is harmonic in D.

  2un  0 in D,

and  2um  0 in D.

  2 (um  un )  0 in D,

305
  um  un is harmonic in D.
By the Maximum and minimum principle, we have

um  x, y   un  x, y   on D m, n  N  .

Hence the result.

Harnnack ' Theorem :
Let D be a bounded domain, bounded by a closed smooth curve B. Let un  x, y  be a

sequence of functions, each of which continuous on D and harmonic in D. If un  x, y  converges

uniformly on B, then un converges an D to a limit function which is continuous on D and harmonic
in D.
Proof : Let {un } be a sequence of functions , each of which continuous on D  DUB and harmonic
in D.

  2un  x, y   0 in D ... (1.40)

Let the sequence {un } converge uniformly on the boundary B.

 for given  0 , we can find a number N such that

um  x, y   un  x, y   on B m, n  N ... (1.41)

Since by (1.40) each of un  x, y  is harmonic in D.

  um  un  is harmonic in D.

Then by maximum and minimum principle,

we have um  x, y   un  x, y   on D m,n  N ... (1.42)

i.e. the sequence {un } converges uniformly on D . We know that "On a closed bounded set, a
uniformly convergent sequence of continuous functions converges to a function which is continuous on
that set ."

Since un  x, y  converges uniformly on D , let it converges to u  x, y  . Then u  x, y  is also

continuous on D .

We now show that u  x, y  is hammonic in D.

Let  x, y   D . Since D is open, therefore  a circle with centre at  x, y  and radius 'a'
which is contained in D, whose equations is

306
  X  x  2  Y  y 2  a 2 ,
where , X  x  a cos  , Y  y  a sin  is any point on the circle.

Let un ( )  un ( x  a cos , y  a sin  ) .

By equation (1.40), un  x, y  is harmonic inside the circle and continuous on the circle, then we know

un  x, y  is given by Possion integral formula

un   , n  
1 2
1  9  2

2  1  29 cos(   )  92 un ( )d .
0

We have 9 r a ,
and   x  2   n  y  2  r 2  a 2 .
Hence

u ( , n)  lim un ( , n)  un ( x, y ) converges to u ( x, y )
As
x 

1 2
1  9 2
 lim
x  2  1  29 cos(   )  92 un ( )d ,
0

1 2
1  9 
2

=
2  1  29 cos(   )  92 xlim

un ( ) d .
0

Since the sequence un  x, y  converges uniformly to u ( x, y ) therefore limit and the integral have
been interchanged

1 2
1  9 
2
u ( , n) 
2  1  29 cos(   )  92 u ( )d
0

Hence u is harmonic in the region ( x   ) 2  ( y  n) 2  a 2 for all points ( , n) . Since ( x, y ) is an

arbitrary point of D.
 u is harmonic in D.
This proves the theorem.

307
Exercise:
1. Show that

x 2   2  ( y   )2
v ( x , y;  ,  ) 
2x2
is the Green's function for the second order partial differential equation

2
u xx  u yy  u x  0 .
x

308
REFERENCES :

1. T. Amaranath : An Elementary Course in Partial Differential Equations, Narosa Publishing

House, 2004.
2. I. N. Sneddon : Elements of Partial Differential Equations, International Edition, McGraw-
Hill, Singapore, 1986.
3. K. Sankara Rao : Introduction to Partial Differential Equations, Prentice Hall of India, New
Delhi, 2005.

309