MA/MSc MT-03

Vardhaman Mahaveer Open University, Kota

Differential Equations, Calculus of Variations

and Special Functions
 Course Development Committee
Chairman
Prof. (Dr.) Naresh Dadhich
Vice-Chancellor
Vardhaman Mahaveer Open University, Kota
Co-ordinator/Convener and Members
Subject Convener Co-ordinator
Prof. D.S. Chauhan Dr. Anuradha Sharma
Department of Mathematics Assistant Professor
University of Rajasthan, Jaipur Department of Botany, V.M.O.U., Kota
Members :
1. Prof. V.P. Saxena 4. Prof. S.P. Goyal 7. Dr. Paresh Vyas
Ex Vice-Chancellor Emeritus Scientist (CSIR) Assistant Professor
Jiwaji University, Deptt. of Mathematics Deptt. of Mathematics
Gwalior (MP) University of Rajasthan, Jaipur University of Rajasthan, Jaipur
2. Prof. S.C. Rajvanshi 5. Dr. A.K. Mathur 8. Dr. Vimlesh Soni
Deptt. of Mathematics Associate Prof. (Retired) Lecturer
Institute of Eng. & Tech. Deptt. of Mathematics Deptt. of Mathematics
Bhaddal, Ropar (Punjab) University of Rajasthan, Jaipur Govt. PG College, Kota (Raj.)
3. Prof. P.K. Banerjee 6. Dr. K.N. Singh 9. Dr. K.K. Mishra
Emeritus Fellow (UGC) Associate Prof. (Retired) Lecturer
Deptt. of Mathematics Deptt. of Mathematics Deptt. of Mathematics
J.N.V. University, Jodhpur University of Rajasthan, Jaipur M.S.J. Collage, Bharatpur (Raj.)
10. Dr. K.S. Shekhawat
Lecturer, Deptt. of Mathematics
Govt. Shri Kalyan College, Sikar (Raj.)

Editing and Course Writing

Editor Writers
Prof. S.P. Goyal 1. Prof. Kantesh Gupta 2. Dr. Atul Garg
Emeritus Scientist (CSIR) Deptt. of Mathematics Lecturer, Deptt. of Mathematics
Deptt. of Mathematics MNIT, Jaipur Seth Moti Lal PG College, Jhunjhunu
University of Rajasthan, Jaipur 3. Dr. K.S. Shekhawat 4. Dr. Chandra Shekhar
Lecturer, Deptt. of Mathematics Assistant Professor
Govt. Sri Kalyan College, Sikar Mathematics Group
5. Mr. Pranay Goswamy Birla Institute of Technology
Lecturer, Deptt. of Mathematics and Science, Pilani
Amity University, Jaipur

Academic and Administrative Management

Prof. (Dr.) Naresh Dadhich Prof. M.K. Ghadoliya Mr. Yogendra Goyal
Vice-Chancellor Director (Academic) Incharge
Vardhaman Mahveer Open University, Vardhaman Mahveer Open University, Material Production and
Kota Kota Distribution Department

Course Material Production

Mr. Yogendra Goyal
Assistant Production Officer
Vardhaman Mahaveer Open University, Kota
 MA/MSc MT-03

Vardhaman Mahaveer Open University, Kota

Differential Equations, Calculus of Variations and Special Functions

Unit No. Units Page No.

1. Non-Linear Ordinary Differential Equations of Particular

Forms and Riccati’s Equation 1–24
2. Total Differential Equations 25–45
3. Partial Differential Equations of Second order, Monge’s Method 46–64
4. Classification of Linear PDE of Second Order, Cauchy Problem and
Method of Separation of Variables 65–80
5. Laplace, Wave and Diffusion Equations And Canonical Forms 81–100
6. Eigenvalues, Eigenfunctions and Sturm-Liouville
Boundary Value Probleon 101–120
7. Variational Problems with Fixed Boundaries and
Euler-Lagrange Equation 121–138
8. Functionals Dependent on Higher Order Derivatives and
Variational Problems in Parametric Form 139–156
9. Series Solution of Second Order Linear Differential Equation 157–181
10. Gauss Hypergeometric Function: its Properties and
Integral Representation 182–200
11. Gauss and Confluent Hypergeometric Functions 201–217
12. Legendre’s Polynomials and Functions Pn(x) and Qn (x) 218–246
13. Bessel’s Functions 247–268
14. Hermite Polynomials 269–285
15. Laguerre Polynomials 286–303
 Reference Books 304
 PREFACE

The Present book entitled ‘‘Differential Equations, Calculus of Variations

and Special Functions’’ has been designed so as to cover the unit-wise syllabus of

Mathematics-Third paper for M.A./M.Sc. (Previous) students of Vardhaman

Mahaveer Open University, Kota. It can also be used for competitive examinations.

The basic principles and theory have been explained in a simple, concise and lucid

manner. Adequate number of illustrative examples and exercises have also been

included to enable the students to grasp the subject easily. The units have been

written by various experts in the field. The unit writers have consulted various

standard books on the subject and they are thankful to the authors of these refer-

ence books.
Unit 1 : Non-Linear Ordinary Differential Equations of
Particular Forms and Riccati’s Equation
Structure of the Unit

1.0 Objective

1.1 Introduction

1.2 Exact Non-Linear Differential Equation

1.3 Riccati’s Equation

1.3.1 General solution of Riccati’s equation

1.3.2 Theorem

1.3.3 Method of solution of Riccati’s equation when one particular solution is known

1.3.4 Method of solution of Riccati’s equation when two particular solutions are known

1.3.5 Method of solution of Riccati’s equation when three particular solutions are known
d2y
1.4 Equation of the form  f  y
dx 2
1.5 Equations not containing y directly

1.6 Equation not containing x directly

1.7 Equations in which y appears in only two derivatives whose orders differ by two

1.8 Equations in which y appears in only two derivatives whose orders differ by unity

1.9 Homogeneous Equation

1.10 Summary

1.11 Answers of Self-Learning Exercise

1.12 Exercise

1.0 Objective

The purpose of this unit is to discuss various methods for solving some particular forms of sec-
ond and higher order non-linear differential equations. The methods for solving exact non-linear differen-
tial equations and Riccati’s equation are also discussed.

1
1.1 Introduction

In earlier classes we studied a great deal about linear differential equations of second and higher
orders when coefficient may or may not be constant. It is a known fact that due to superimposition of
linearly independent solutions, it is easy to solve linear differential equation and we have well established
theories for such types of equations.
On the other hand, the non-linear differential equations are difficult to handle. In the case of some
first order equations, we have well established methods. However, there is no known general method
for solving second and higher order non linear differential equations. It is only some particular forms that
may be reduced to linear equations by suitable transformation and integrated to yield compact results.
The aim of this unit is to study those easily integrable non-linear equations.
Next we shall discuss the general solution of Riccati’s equation. The solution of this equation
when one, two or three particular solutions are known will also be discussed.

1.2 Exact Non-linear Differential Equations

There is no simple method for testing the exactness of non-linear differential equations as in the
case of linear equations. One possible method is that if the terms of the equation be grouped, by inspec-
tion, in such a way that they become perfect differential and their integrals may be written directly. The
other method of obtaining the integral of an exact differential equation, which is applicable both for linear
and non-linear equations is explained below.
Let s = f (x) be a differential equation of nth order. If it is an exact deferential equation it should

dny
be derived merely by differentiation, so as to contain in the first degree. Now we write the equa-
dx n

d n 1 y
tion in the form sdx = f (x) dx and will integrates assuming that as if were the only variable in the
dx n 1

dny
differential equation and is its differential coefficient.
dx n
Denoting the result by s1 then sdx – ds1 will contain differential coefficients at the most upto

d n 1 y
(n – 1 )th order. Restriction of taking as the only variable should be removed while finding ds1.
dx n 1
Repeating the above process as many times as necessary, we shall finally get
sdx – ds1 – ds2 – ... = 0
or ds1 + ds2 + ... = sdx
On integration, we get
s1 + s2 ... = sdx = f (x) dx

2
 Ex.1. Show that the differential equation
3
dy  dy   dy  d 2 y
y + 3 x + 2 y   +  x 2 + 2 y2  =0
dx  dx   dx  dx
is an exact equation and find its first integral.
Sol. The given equation can be written as
 d2y 3 
dy d 2 y  dy  dy
sdx   x 2 2  2 y 2 2
 2 y    3x  y  dx  0
 dx dx dx  dx  dx 
Now here the first three terms are the differentiation of
2
dy  dy 
x2  y2  
dx  dx 

 2 dy 2
2  dy  
So putting s1   x y   
 dx  dx  
On differentiation, we get
 2 d 2 y 3 2 
dy  dy  2 dy d y 
ds1   x 2
 2x  2 y    2 y  dx
 dx dx  dx  dx dx 2 

 dy 
Thus sdx  ds1   y  x  dx .....(1)
 dx 
Again the terms on R.H.S. are the differentiation of xy, so putting
s2 = xy
On differentiation, we get
 dy 
ds2   x  y  dx .....(2)
 dx 
From (1) and (2), we finally get
sdx – ds1 – ds2 = 0
which on integration gives
s1 + s2 = constant
This relation shows that the given equation is exact and the first integral will be given by
2
dy
2  dy 
x  y 2    xy  c.
dx  dx 
Ex.2. Solve the following differential equation :

d2y dy dy
2sin x 2
+ 2cos x + 2sin x + 2 y cos x = cos x
dx dx dx
Sol. We can writte the given equation as
 d2y dy dy 
sdx   2 sin x 2  2 cos x  2sin x  2 y cos x  dx  cos x dx
 dx dx dx 
3
 dy
Here first term of above equation will arise from the differentiation of 2sin x , so putting
dx
dy
s1  2sin x
dx

 d2y dy 
which implies that ds1   2 sin x 2  2 cos x  dx
 dx dx 

 dy 
Thus sdx  ds1   2 sin x  2 y cos x  dx
 dx 
Again putting
s2 = 2y sin x
On differentiation, we get
 dy 
ds2   2 sin x  2 y cos x  dx
 dx 
 sdx – ds1 – ds2 = 0
This shows that the given equation is exact and on integrating, we get
s1 + s2 = sdx = cos x dx
dy
or 2sin x  2 y sin x  sin x  2c1
dx
dy 1
or  y   c1cosec x
dx 2
This is a linear differential equation of first order whose integrating factor (I.F.) is ex
Thus its solution is
1 
y.  I .F .     c1cosec x   I .F . dx  c2
2 

1 x
or y ex  e  c1  e x cosec x dx  c2
2
2
2 d2y 2  dy  dy
Ex.3. Solve 2 x cos y 2  2 x sin y   + x cos y  sin y = log x
dx  dx  dx
Sol. The given equation is

 2 2 
d2y 2  dy  dy
sdx   2 x cos y 2  2 x sin y    x cos y  sin y  dx  log x dx .....(3)
 dx  dx  dx 

dy
Let s1  2 x 2 cos y
dx

 d2y  dy  dy 
So that ds1   2 x 2 cos y 2  2 x 2 sin y    4 x cos y  dx
 dx  dx  dx 

4
  dy 
 sdx  ds1   3 x cos y  sin y  dx
 dx 
Again let s2 = – 3x sin y
 dy 
So that ds2   3 x cos y  3sin y  dx
 dx 
 s dx – ds1 – ds2 = 2sin y dx
Hence the equation is not exact.
So dividing the given equation (3) by x2, we get
 2 
d2y  dy  1 dy 1 log x
sdx   2 cos y 2  2sin y    cos y  2 sin y  dx  2 dx
 dx  dx  x dx x  x

dy
Now let s1  2 cos y
dx

 2
d2y  dy  
so that ds1   2 cos y  2sin y    dx
 dx  dx  

1 dy 1 
 sdx  ds1   cos y  2 sin y  dx
x dx x 

1
Again let s2  sin y
x

1 dy 1 
So that ds2   cos y  2 sin y  dx
x dx x 
 sdx – ds1 – ds2 = 0
Hence the equation is exact, and
log x
ds1  ds2  sdx  dx
x2
Integrating we get
1
s1  s2   log x dx  c1
x2
dy 1 1
2 cos y  sin y    log x  1  c1 .....(4)
dx x x
Let sin y = u. Then
dy du
cos y 
dx dx
 (4) reduces to
du u 1 c
    log x  1  1 .....(5)
dx 2 x 2x 2
5
which is linear with
1 1
 dx
I .F .  e2 x  x
Hence the solution of (5) is

1  log x  1 x c
u x   dx  1  x dx  c2
2 x 2
1 w2 c
or x sin y     w  1 e dw  1 x 3 2  c2 , where w = log x
2 3
c1 3 2
   w  1 e w 2  2e w 2  x  c2
3
c1 3 2
   log x  1 x  2 x  x  c2
3
c1
or sin y   log x  1  x  c 2 x 1 2
3
which is the required solution.
2
2d 2 y  dy 
Ex.4. Solve x y 2  x  y  0
dx  dx 
Sol. The given equation is
2
 2 d2y 2  dy  dy 
sdx   x y 2  x    2 xy  y 2  dx  0 .....(6)
 dx  dx  dx 

dy
Let s1  x 2 y
dx
2
 d2y  dy  dy 
 ds1   x 2 y 2  x 2    2 xy  dx
 dx  dx  dx 

 dy 
So that sdx  ds1   4 xy  y 2  dx
 dx 
Again let s2 = – 2xy2
 dy 
So that ds2   4 xy  2 y 2  dx
 dx 
 sdx – ds1 – ds2 = 3y2 dx
Hence the equation is not exact.
Therefore dividing the given equation (6) by x2, we get

 d 2 y  dy 2 2 y dy y 2 
sdx   y 2       dx  0
 dx  dx  x dx x 2 

dy
Now let s1  y
dx
6
  d 2 y  dy 2 
Then ds1   y 2     dx
 dx  dx  

 2 y dy y 2 
So that sdx  ds1     2  dx
 x dx x 

y2  2 y dy y 2 
Let s2   so that ds2     2  dx
x  x dx x 
Hence sdx – ds1 – ds2 = 0
or ds1 + ds2 = sdx = 0
or s1 + s2 = c1
dy y 2
or y   c1 .....(7)
dx x

y2 dy du
Let u so that y 
2 dx dx
Hence equation (7) becomes
du 2
 u  c1 .....(8)
dx x
 2 x  dx 1
which is linear with I .F .  e 
x2
Thus solution of (8) is
u c y2
2
  1  c2 or  x  c1  c2 x 
x x 2
or y2 = x(Ax – B),
where A and B are arlitrary corstants.

1.3 Riccati’s Equation

Originally, the name Riccati’s equation was given to the differential equation
dy
 by 2  cx m .....(1)
dx
where b are c are constants. Equation (1) can be written in the form
y1 + by2 = cxm .....(2)
where suffixes denotes differentiation w.r.t. x
The more general form of (2) is
xy1 – ay + by2 = cxm .....(3)
which can be easily reduced to the form
du b 2 c  m a  2
 u  z .....(4)
dt a a
by using the substitution t = xa and then changing the variable y to u by substitution y = ut.
7
 The Equation (4) can be easily written in the form
y1 = P + Qy + Ry2 .....(5)
where P, Q and R are function of x.
The equation (5) is known as the generalised Riccati’s equation.
French Mathematician Liouville, in 1841, proved that equation (5) is one of the simplest differ-
ential equation of the first order and first degree that can not, in general be integrated by quadratures.
Due to historical and theoretical importance and its usefulness in Differential Geometry, the study of Riccati’s
equation becomes quite useful.

1.3.1 General solution of Riccati’s equation

Equation (5) can be reduced to a second order linear differential equation by introducing an-
other dependent variable S such that
S1 1
y   S1  RS  .....(6)
RS
On differentiation, we get
y1 = – S2(RS)–1 + S1(RS)–2 [R1S + RS1] .....(7)
where a subscript denote differentiation with respect to x.
Substituting (6) and (7) in (5), we get
S2 R1 S1 S12  S1   S12 
  2   P  Q 
 RS   R  2 2
RS R S RS 2   R S 
or –RS2 + R1S1 = PR2S – QS1R
or RS2 –(QR + R1) S1 + PR2S = 0 .....(8)
This is linear differential equation of second order. We know that the general solution of (8) is of
the form
S = Af (x) + Bg(x) .....(9)
where A and B are arbitrary constants and f (x), g(x) are two linearly independent integrals.
Now, from (6) and (9), we get

y
 Af1  Bg1     A / B  f1  g1
R  Af  Bg  R  A / B  f  g 
which is of the form
cf1  x   g1  x 
y .....(10)
R  cf  x   g  x  
where c = A/B is an arbitrary constant. Hence the general solution of (5) is (10).
1.3.2 Theorem : The cross ratio of any four particular integrals of a Riccati’s equation
is independent of x
Proof : We know that the general solution of Riccati’s equation
y1 = P + Qy + Ry2 .....(11)
8
 cf1  g1
is of the form y .....(12)
R cf  g 
where f1, g1, f, g are appropriate functions of x and c is an arbitrary constant.
Let p(x), q(x), r(x) and s(x) are four particular solutions of (11) obtained from (12) by giving
four different values of c, say , , , .

p  x  
 f1  g1 
Then R f  g 

q  x  
f1  g1 
R  f  g 

r  x  
 f1  g1 
R  f  g 

s  x  
f1  g1 
R  f  g 

      fg1  f1 g 
Then p q 
R  f  g f  g 

      fg1  f1 g 
rs
R  f  g f  g 

      fg1  f1 g 
ps
R  f  g  f  g 

     fg1  f1 g 
r q 
R  f  g f  g 

 p  q  r  s            k say
Thus  
 p  s  r  q         
when k is independent of x. This shows that the cross-ratio of any four particular solutions of a Riccati’s
equation is independent of x.
1.3.3 Method of solution of Riccati’s equation when one particular solutions is known
Let p(x) be the known particular solution of Riccati’s’s equation
y1 = P + Qy + Ry2 .....(13)
So that p1 = P + Qp + Rp2
Let u be the another dependent variable such that
1
y  p  x  .....(14)
u
then equation (13) reduces
u1  1  2p 1 
p1  2
 P  Q  p    R  p2    .....(15)
u  u  u u2 
9
 Using (14) and (15) in (13), we get
u1 Q 2p 1 
   R  2
 u u 
2
u u
or u1 + (Q + 2pR) u = – R
which is a linear differential equation of first order and first degree in u and x. Its integrating factor is
given by
I.F. = e (Q + 2Rq) dx
and hence the required general solution is
ue (Q + 2Rq) dx = Re (Q + 2Rq) dxdx + c
where c is an arbitrary constant.

1.3.4 Method of solution of Riccati’s equation when two particular solutions are known
Let p(x) and q(x) be the two know particular solutions of Riccati’s equation
y1 = P + Qy + Ry2 .....(16)
so that p1 = P + Qp + Rp2 .....(17)
q1 = P + Qq + Rq2 .....(18)
From (16) and (17), we get
y1 – p1 = (y – p) Q + (y2 – p2)R
or y1 – p1 = (y – p) [Q + (y + p)R]
y1  p1
or  Q   y  p R .....(19)
y p
Similarly from (16) and (18), we get
y1  q1
 Q   y  q R .....(20)
yq
From (19) and (20), we get
y1  p1 y1  q1
   p  q R
y p yq
On integration, we get
log (y – p) – log (y – q) = c + (p – q) Rdx
which is the required general solution.
1.3.5 Method of solution of Riccati’s equation when three particular solutions are known
Let p(x). q(x) and r(x) be the three known particular solutions of Riccali’s equation
y1 = P + Qy + Ry2
and the corresponding values of c be ,  and . Then by Theorem 1.3.2, we can write

p
 f1  g1 
R f  g 

q
f1  g1 
R f  g 

r
 f1  g1 
R  f  g 
10
  p  q  r  y   k constant
then, we have  
 r  q  p  y 
where k is independent of x. This is the required solution of Riccati’s equation when three particular
solutions are known.
Ex.1. solve y1 = cos x – y sin x + y2
Sol. Taking y = sin x so that y1 = cos x. Substituting these in the given equation, we get
cos x = cos x – sin2 x + sin2 x
This shows that y = sin x is a particular solution of given equation.
1 u1
Now taking y  sin x  so that y  cos x 
u u2
Using these in given equation, we get
2
u1  1  1
cos x  2
 cos x  sin x  sin x     sin x  
u  u  u
u1 sin x 1
or    2
u2 u u
du
or  u sin x  1 .....(21)
dx
Equation (21) is a linear equation of first order whose integrating factor is
I.F. = esin x dx = e–cos x and hence the solution of (21) is
u. e–cos x = c – e–cos x dx .....(22)
Now putting the value of
1
u
 y  sin x 
in equation (22), we get
e cos x
 c   e  cos x dx
y  sin x
which is the required solution of given equation.
Ex.2.Find the general solution of the Riccati’s equation
dy
= 2  2 y + y2
dx
whose one particular solution is (1 + tan x).
Sol. The given equation is
dy
 2  2 y  y2 .....(23)
dx
Since (1 + tan x) is a given particular solution then taking
1 1 du
y  1  tan x   so that y1  sec 2 x  2 .....(24)
u u dx
Putting (24) in (23), we get
1 du 1 2 tan x
  
u 2 dx u 2 u

11
 du
or   2 tan x  u  1
dx
It is a linear differential equation of first order having integrating factor
I.F. = e(2tan x)dx = e2log sec x = sec2 x
Hence the solution is
u sec2 x = c – sec2 x dx = c – tan x .....(25)
From (24) and (25), the required general solution is
sec 2 x
y  1  tan x 
c  tan x
k
Ex.3. Show that there are two values of the constant for which is an integral of
x
x2 (y1 + y2) = 2, and hence obtain the general solution.
Sol. Rewriting the given equation in the standard Riccati’s form as
y1 = P + Qy + Ry2 .....(26)
 2 
y1   2   y 2 .....(27)
x 
Let p(x) and q(x) are two particular integrals of (26), than by §1.3.4, we have
 y  p
log    c    p  q  Rdx .....(28)
  y  q  
k k
Now let y so that y1   2
x x
Substituting these in (27), we get
k 2 k2
   or k 2  k  2  0 so that. k  2, 1
x 2 x2 x2
2 1
Hence and  are two particular solutions of (27)
x x
Now taking
2 1
p  x  and q  x    .....(29)
x x
On comparing (26) and (27), we get R = – 1 .....(30)
Using (29) and (30) in (28), we get
xy  2 2 1
log  log k       1 dx, taking c  log k
xy  1  x x
xy  2
or log  log k  3log x
xy  1

 xy  2  3
or  x  k
 xy  1 
or x3(xy –2) = k (xy + 1), where k is an arbitrary constant.
12
 Ex.4. Show that 1, x, x2 are three particular integrals of x (x2 – 1) y1 + x2 – (x2 – 1)
y – y2 = 0, and hence obtain the general solution y (x + k) = x + kx2, k being an
arbitrary constant.
Sol. Re writing the given equation in the standard Riccati’s form as
x 1 1
y1    y y2 .....(31)
x  x  1
2 2
x 1 x

Now putting y = 1 (one of the three given integrals) so that y1 = 0, and we get
x 1 1
0   0
2
x  1 x x x2  1  
This show that y = 1 is an particular integral of (1). Similarly we can prove that y = x and y = x2
are also particular integrals of (31).
Now taking p(x) = 1, q(x) = x, r(x) = x2 and using § 1.3.5, we get

1  x   x 2  y  1
 (say)
 x2  x  1  y  k
1  x   x 2  y  1
or 
 x 1  x 1  y  k
or k(x2 – y) = – x(1 – y)
or y(k + x) = x + kx2
which is the required solution.

d2 y
1.4 Equation of the Form 2 = f  y 
dx

dy
To find the solution of above equation, we multiply both side by 2 , then we get
dx
dy d 2 y dy
2 2
 2 f  y
dx dx dx
On integration we obtain
2
 dy 
   2  f  y  dy  a
 dx 
dy
1/ 2
 dx
or 2   f  y  dy  a 
 
Again integrating, we finally obtain
dy
 1/ 2
 xb
2   f  y  dy  a 
 

13
 d2 y
Ex.1. Solve sin3 y = cos y
dx 2
Sol. We can write the given equation as

d2y
2
 cosec2 y cot y
dx
dy
Now multiplying both sides by 2 and integrating, we get
dx
2
 dy  2 a sin 2 y  cos 2 y
   a  cot y 
 dx  sin 2 y

sin y dy
 dx
or 2
a  (1  a) cos y
Again integrating, we get the required solution as

1  1  a 
 sin 1  cos y   x  c
1 a  a 

d2 y
Ex.2. Solve y3 =c
dx 2
Sol. We can write the given equation as
d2y c

dx 2 y 3
dy
Now multiplying both side by 2 and integrating, we get
dx
2
 dy  c
 dx    2  a
  y

y dy
 dx
or 2
ay  c
Again integrating, we get the required solution as
ay2 = c + (ax + b)2
where a and b are two constants.

1.5 Equation not Containing y Directly

In this case general equation is given in the form

 d n y d n 1 y dy 
f  n , n 1 ,....., , x  0 .....(1)
 dx dx dx 

14
 To solve it, the order of equation is depressed by assuming the lowest differential coefficient
present in the equation as a dependent variable. So let

dy d 2 y dp d n y d n1 p
 p, 2  , ...., n  n1
dx dx dx dx dx
therefore equation (1) reduces to

 d n 1 p d n  2 p 
f  n 1 , n  2 , ..., p,x   0
 dx dx 
which may be possibly solved for p.
dy
Let p   x
dx
then the solution is
y     x  dx  c .
2
 d3y  d3y d2y
Ex.1. Solve  3  + x 3  2 = 0
 dx  dx dx
Sol. The given equation does not contain y directly. Here the lowest differential coefficient is
d2y
. So putting
dx 2

d2y d 3 y dp
 p and  .
dx 2 dx3 dx
We get from the given equation
2
 dp  dp
  x p0
 dx  dx

2
dp  dp 
or p  x   [Clairaut`s form y = px + f (p)]
dx  dx 
So its solution is
p = cx + c2

d2y
or 2
 cx  c 2
dx

dy x2 2
on integration,  c  c x  c1
dx 2
Again integrating, we get the general solution as
x3 2 x 2
y c c  c1 x  c2
6 2

15
 2
d 2 y  dy 
Ex.2. Solve 2 2    + 4= 0
dx  dx 
Sol. The given equation does not contain y directly. Here the lowest differential coefficient is
dy
. So putting
dx
dy d 2 y dp
 p and  .
dx dx 2 dx
We get from the given equation
dp
2  p2  4  0
dx
2dp
or  dx
p2  4
Integrating
1 p2
log xa
2 p2
or (p – 2) = (p – 2) be2x, where b = e2a.

dy  2be 2 x 
or p  2 1  2x 
dx  1  be 
On integration, we get the general solution as
y = 2x – 2 log (1 – be2x) + c.

1.6 Equation not Containing x Directly

In this case general equation is given in the form

 d n y d n1 y dy 
f  n , n 1 , ... , , y0
 dx dx dx 

dy d 2 y dp dp dy dp
Now putting  p,     p
dx dx 2 dx dy dx dy

d 3 y d  d 2 y  d  dp  dy
Similarly     p 
dx 3 dx  dx 2  dy  dy  dx

 d 2 p  dp 2 
p 2    p
 dy  dy  

2
d2 p
2  dp 
` p  p  
dy 2  dy 

16
Hence the given equation reduces to
 d n 1 p 
f  n 1 , ..., p , y   0
 dy 
which may be possibly solved for p.
dy
Let p    y .
dx
Then the solution is
dy
  y  x  c
3
d2 y dy  dy 
Ex.1. Solve 2
 2 4  0
dx dx  dx 
Sol. The given equation does not contain x directly, so substituting
dy d2y dp
 p , 2  p , we get
dx dx dy
dp
p  2 p  4 p3  0
dy
dp
or   2 dy
1  2 p2
On integration, we get
1
2

tan 1 p 2   2 y  a 
or  
tan 1 p 2  b  2 2 y , where b  2a

or 
2 cot b  2 2 y dy  dx . 
Again integrating, we get the general solution as


log sin b  2 2 y   2 x  log c 
or 
sin b  2 2 y  ce2 x 
2
d2y  dy 
Ex.2. Solve y 1  log y  2  1  log y    = 0
dx  dx 
Sol. The given equation does not contain x directly, so substituting

dy d2y dp
 p, 2  p , we get
dx dx dy
dp
y 1  log y  p  1  log y  p 2  0
dy
17
 dp 1  log y 
or  dy  0.
p y 1  log y 
On integration, we get by substituting log y = t
log p  log y  2 log  log y  1  constant
dy 2
or p  ay  log y  1
dx
dy
or 2
 a dx
y  log y  1
Again integrating , we get the general solution as
1
  ax  b
 log y  1
1
or 1  log y  
ax  b

1.7 Equation in which y Appears in only Two Derivatives Whose Orders Differ by
Two.

In this case general equation is given in the form

 d n y d n2 y 
f  n , n  2 , x   0
 dx dx 
d n 2 y
Now putting p
dx n 2

dny d2 p
so that 
dx n dx 2
then the given equation becomes
 d2p 
f  2 , p , x   0
 dx 
d n2 y
which gives p  n 2    x  .
dx
By successive integration, we can find the value of y.

d5 y 3
2 d y
Ex.1. Solve 5
n 3
= e ax
dx dx
Sol.  In the given equation y appears in two derivatives whose order differs by two. Now sub-

d3y
stituting = p . So the given equation transforms to
dx 3
18
 d2 p 2
2
 n p = e ax
dx
whose solution will be

d3y nx  nx e ax
p  3  c1 e  c2 e 
dx a 2  n2  
On integration, we get

d 2 y c1 nx c2  nx e ax
 e  e   c3
dx 2 n n a a 2  n2  
Again integrating

dy c1 nx c2  nx eax
 2e  2e   c3 x  c4
dx n n a 2 a 2  n2  
which on integration gives the general solution as

c1 nx c2  nx e ax x2
y e  e   c3  c4 x  c5
n3 n3 a3 a 2  x 2 2 
1.8 Equation in which y Appears in only Two Derivatives Whose Orders Differ by
Unity
In this case general equation is given in the form
 d n y d n 1 y 
f  n , n 1 , x   0
 dx dx 

d n1 y
Now putting p
dx n1

d n y dp
so that  .
dx n dx
Hence the given equation reduces to

 dp 
f  , p , x0
 dx 
This is an equation of first order. We can here easily find the value of p in terms of x as

d n1 y
p    x .
dx n1
By successive integration, we get the general solution.

19
 1
2 2
d y   dy 
2
Ex.1. Solve a = 1 +   
dx 2   dx  

Sol.  In the given equation y appears in two derivatives whose order differs by unity. Now sub-
stituting

dy d 2 y dp
 p, 
dx dx 2 dx
so the given equation transforms to
1
dp
a  1 p2
dx
  2

dp 1
 dx
or a
1 p2

x
Integrating sin h 1 p   c1
a
dy x 
p  sin h   c1 
dx a 
Again integrating, we get the general solution as
x 
y  a cos h   c1   c2
a 

1.9 Homogeneous Equation

We mean by homogeneous equation that an equation in which all the terms will be of the same
dimensions.
Dimention of a differential equation is calculated as given under
2
d 2 y  dy  dy
x 2    2
dx  dx  dx

 d2y   y 
Now Dim  x 2   Dim  x 2   Dim y1 x 1  
 dx   x 

  dy 2    y 2 
Dim      Dim      Dim y 2 x 2
  dx    x    
   

 dy 
Dim    Dim y1 x 1
 dx 
 
Dim  2   0

Hence the given equation has the 0 dimension

20
 Note :
(a) Derivative in a differential equation does not alter the dimension of the variables x and y.
(b) The dimension of x is invariably taken as unity.
In such cases suitable transformations are made to lower the order of the equation
2
d2y  dy 
Ex.1. Solve nx 3 2
=yx 
dx  dx 
Sol. Here x and y both of dimension unity. There for the given equation is homogeneous of
dimension 2. Substituting y = zx and x = e, we get
2
2 dz d 2 z    dz  
n e   2    xz  x  z  
 d d    d  

 dz d 2 z   dz  2
or n   2    
 d d   d 
dz
Now if we put   , then above equation becomes
d
 d  2
n   
 d  

 1 1
or    n    d   d
 

1 n
on integrating log    constant
n 
dz
Now substituting   and then integrating , we get the general solution as
d
 c 
y  n x log  c1  2 
 x

1.10 Summary

In this unit, you studied the exactness of differential equation and the method by which we can
solve exact equations. Methods for solution of the standard Riccati’s equation of first order, with one,
two or three known particular solutions were discussed. The methods have been illustrated with the
help of examples.

Self-Learning exercise

1. What do you mean by exact equation ?

2. Write down the Riccati’s equation of first order.
3. Riccati’s equation is a non-linear differential equation. Is it true ?
21
1.11 Answers of Self-Learning Exercise

1. A differential equation which is integrable directly.

dy
2.  P  Qy  Ry 2 , where P, Q, R are functions of x or constants.
dx
3. True

1.12 Exercise

1. Solve the following differential equations :

2
2 d 2 y  dy  2 5 2c1
(a) x y 2
  x  y   3 y2  0 [Ans. xy  c2 x  ]
dx  dx  5

d2y dy  dy 
(b)  2 y  x  2
 2 1    0 [Ans. y 2  xy  c1x  c2 ]
dx dx  dx 

2 2
d2y  dy  dy  x  1 x
(c) cos y 2  sin y    cos y  x 1 [Ans. sin y   xc 1  c2 e ]
dx  dx  dx 2
2. Solve the following differential equations :

x4  x 2 x3
 3
 2 2 2
(a) x 1  x y1  x  y  2 xy , x is an integral [Ans.
y  x2
 c 
3
]

dy
(b)  1  y 2 , tan x is an integral [Ans. y  c  tan x   c tan x  1 ]
dx

(c) x3 y1  x 2 y  y 2  x 2 [Ans. y (ce2/x – 1) = x + cxe2/x]

2 y
x 2
c 
(d) x  x  1 y1   2 x  1 y  yh  2 x  0 , x is a solution [Ans. ]
 x  c
3. Solve :

d2y 1 12
(a)
dx 2

ay
[Ans. 3 x  2a1 4  y  2c1  y  c1   c2 ]

d 2 y a2 1
(b) 
dx 2 y 2
0 [Ans. c1 y 2  y 
c1
log  
c1 y  1  c1 y  ac1 2 x  c2 ]

4. Solve :

2
d2y  dy 
(a) 2
 1   [Ans. y  cos h  x  c1   c2 ]
dx  dx 

22
 2
d2y  dy 

(b) 1  x
2
 dx 2
1    0
 dx 
 2

[Ans. y  cx  1  c log  x  c   c1 ]

3
d 2 y dy  dy 
(c)     0
dx 2 dx  dx 
1 x

[Ans. y   sin c1e  c2 ] 
2 52
d3y d2 y  d2y  2 4  c1 x  a 
(d) 2 x 3  2   2   a [Ans. y   c2 x  c3 ]
dx dx  dx  15c12

5. Solve :
2
d 2 y  dy  2
(a) y 2     y [Ans. y 2  c1 sin h 2 x  c2 ]
dx dx
 
2
d2y  dy 
(b) 2
 a   0 [Ans. e ay  c1x  c2 ]
dx  dx 
2
d 2 y  dy 
(c) y    1  0 [Ans. y 2  x 2  c1 x  c2  0 ]
dx 2  dx 

2
d 2 y  dy 
(d) y 2
    y 2 log y [Ans. log y  c1e x  c2e x ]
dx  dx 

2 12
2  dy 2
2 2
 dy  d y   2d y  12
(e)  dx   y 2  n  dx   a  2  
  dx    dx  

[Ans. cy  n 1  a 2 c 2   c2 ecx ]

6. Solve :

d4y 2
2 d y
(a)  a 0 [Ans. y  c1eax  c2e  ax  c3 x  c4 ]
dx 4 dx 2

2 d4y 2
2 d y 52 2 c4 
1  4a 2 
(b) x  a 0 [Ans. y  c1  c2 x  x  c3 x 1  4a 
dx 4 dx 2  x 

1 52 1 2 x 1
when a < and y  c1  c2 x  c3 x cos  2 4a  1 log c  when a > ]
2  4  2
7. Solve :

d 2 y dy
(a) x  0 [Ans. y  c1 log x  c2 ]
dx 2 dx

d3y d2 y 52
(b) 3  2  2 [Ans. 15 y  8  x  c1   c2 x  c3 ]
dx dx
23
 8. Solve :
2
d2y  dy  dy
(a) xy 2
 x    3y [Ans. y 2  y 4  c1 x 4  c2 x 4 ]
dx  dx  dx
2
d2y  dy  dy
(b)  2 y  x  2  2    2 2 [Ans. xy  y 2  x 2  c1x  c2 ]
dx  dx  dx
9. By reduction to a linear equation show that the solution of the Riccati’s equation
dy
x2  2  2 xy  x 2 y 2  0 is
dx

 
y x 2  c1 x  2 x  c1
10. Show that tan x is one integral of the equation
y1  1  y 2
and hence obtain the general solution in the form
y  c1  tan x   c1 tan x  1
where c1 is a constant.
11. Determine the curve whose radius of curvature varies as the cube of the length of the normal
intercepted the curve and x-axis. [Ans. c3  c1 y 2   c1 x  4  ]



24
Unit 2 : Total Differential Equations
Structure of the Unit

2.0 Objective

2.1 Introduction

2.2. Necessary and Sufficient Condition for Integrability of the Total Differential Equation

2.2.1 Theorem

2.3 Methods of Solving Total Differential Equations

2.3.1 Method of Inspection

2.3.2 Method for Homogeneous Equations

2.3.3 Working Rule for Solving Homogenous Equations

2.3.4 Method of Auxiliary Equations

2.3.5 General Method

2.4 Geometrical meaning of Pdx + Qdy + Rdz = 0

2.5 Equations Containing More Than Three Variables

2.6 Method for Obtaining Solution Involving Four Variables

2.7 Total Differential Equation of Second Order

2.8 Summary

2.9 Answers of Self Learning Exercise

2.10 Exercise

2.0 Objective

In this unit, you will learn various methods for solving different types of total differential equa-
tions. Some of the methods are : Method of inspection, method for homogeneous equations, method of
Auxiliary equations and general method. You will also study the geometrical meaning and method for
solving total differential equations involving three or four variables.

2.1 Introduction

In this unit, we propose to discuss differential equations with one independent variable and more
than one dependent variables.
25
 n
The expression  ui dxi , where u , i = 1, 2 .... n are, in general, functions of some or all of n
i
i 1

independent variables x1, x2 .... xn is called a total differential forms in n variables and the equation
n
 ui dxi  0 .....(1)
i 1

is called a total differential equation in n variables x1, x2 .... xn. It is also known as Pfaffian differen-
tial equation.
In the case of two variables, equation (1) may be written as
M  x, y  dx  N  x, y  dy  0 .....(2)
It is a differential equation of first order and first degree. The necessary and sufficient condition
for its exactness (integrability) is
M N
 .....(3)
y x
In the case of three variables x, y, z the total differential equation (1) may be written as
Pdx + Qdy + Rdz = 0 .....(4)
where P, Q and R are functions of x, y and z. In vector notations, equation (4) may be written as
X dr = 0 where X = (P, Q, R) and dr = (dx, dy, dz).
It is not always possible to integrate equation (4) directly. If however, the equation is such that
there exist a function u (x, y, z) whose total differential du is equal to the left hand side of (4), then only it
is integrated directly. In other cases equations (4) may or may not be integrable.
Now we proceed to find the condition which P, Q, R must satisfy, so that equation (4) is inte-
grable. This is also known as condition of integrability.

2.2 Necessary and Sufficient Condition for integrability of the Total Differential

Equation Pdx + Qdy + Rdz = 0.

2.2.1. Theorem :
The necessary and sufficient condition for the total differential equation Pdx + Qdy +
Rdz = 0 to be integrable is

 R Q   R P   Q P 
P   Q    R  0
 y z   x z   x y 
or X  curl X = 0, where X = (P, Q, R)

P Q R
  
0
or x y z
P Q R
26
 Proof : Condition is necessary :
Let u (x, y, z) = C .....(1)
be an integral of total differential equation
Pdx + Qdy + Rdz = 0 .....(2)
Then total differential du of (1), must be equal to Pdx + Qdy + Rdz, or it multiplied by a factor.
But we know the differentiation of (1) is

 u   u   u 
du    dx    dy    dz .....(3)
 x   y   z 

u u u
Since (1) is an integral of (2), therefore P, Q, R must be proportional to , and .
x y z
u x u y u z
So,      x, y, z   say 
P Q R
u u u
 P  , Q  ,  R  .....(4)
x y z
From the first two parts of (4), we get

  2u  2u   u  
 P         Q 
y yx xy x  y  x
P  Q 
or  P  Q
y y x x

 P Q   
or   Q P .....(5)
 y x  x y
Similarly, we can write
 Q R   
  R Q .....(6)
 z y  y z

 R P   
and   P R .....(7)
 x z  z x
Multiplying (5), (6) and (7) by R, P and Q respectively and adding, we get
 R Q   R P   Q P 
P   Q    R  0 .....(8)
 y z   x z   x y 
This is the condition for the integrability of total differential equation (2).

Sufficient Condition :

Now we prove that if the condition (8) is satisfied, then the equation (2) will have a solution of
the form (1).
Now if the condition (8) is satisfied for P, Q, R of the equation (2) then it can be easily verified
that the same condition will hold for the coefficients of

27
  Pdx  Qdy   Rdz  0
where is any function of x, y, z and replacing P, Q, R by P, Q, R respectively.
Here, if we treat variable z as constant then the differential equation (2) becomes
Pdx+Qdy = 0.
Now Pdx+ Qdy may be regarded as an exact differential. For if it not so, then an integrating
factor can be found to make it exact. Thus there is no loss of generality in regarding Pdx + Qdy as an
exact differential. Therefore
(Pdx + Qdy) = V (say). .....(9)
It follows that
V V
P and Q 
x y
Differentiating (9), we get
V V
Pdx  Qdy  dx  dy .....(10)
x y
Substituting these values in the given condition (8), we find that

V  R  2V  V  R  2V 
        0
x  y zy  y  x zx 

V   V  V   V 
or    R    R  0
x y  z  y x  z 

V   V 
 R
x x  z 
0
or V   V 
  R
y y  z 

 V 
This shows that a relation independent of x and y exists between V and   R  . Conse-
 z 
V
quently  R can be expressed as a function of z and V. That is we can take
z
V
 R    z, V  .....(11)
z
V V  V 
Hence Pdx  Qdy  Rdz  dx  dy      dz
x y  z 
 V V V 
 dx  dy  dz    dz
 x y z 
 dV  dz
Thus (2) may be written as dV – dz = 0 which is a first order equation in two variables hence
integrable will give equation in two variables.
28
 Suppose the integral is U(z, V) = c, then subestituting the value of V from (9), we get the solu-
tion in the form given by (1).
Thus the condition is sufficient.
2.1.1 Theorem : Prove that the necessary condition for integrability of the total dif-
ferential equation X  dr = Pdx + Qdy + Qdz = 0 is X  curl X = 0.
Proof : Let r = xi + yj + zk, so that
dr = dxi + dyj + dzk
and X = Pi + Qj + Rk
Then we have
X dr = Pdx + Qdy + Rdz .....(12)
Then we see that (12) is satisfied by usual rule of dot product of two vectors X and dr.
Now, we know that
 R Q   R P   Q P 
Curl X =   i     j  R   k.
 y z   x z   x y 
Now by usual rule of dot product of two vectors, we get
 R Q   R P   Q P 
X  Curl X =    Q   R  
 y z   x z   x y 
which is equal to zero. So the necessary condition is X  curl X = 0

2.3 Methods of Solving Total Differential Equation Pdx + Qdy + Rdz = 0

If the following condition of integrability

 R Q   R P   Q P 
P   Q    R  0
 y z   x z   x y 
is satisfied, then the total differential equation may be solved by several methods as given below.
2.3.1 Method of Inspection
If the condition of integrability is satisfied, then sometimes it will be possible to rearrange the
terms of the given equation, by dividing or multiplying by a suitable function, so that it can be integrated
directly.
The following list will help to rewrite the given equation in the form of exact differential.
x dy  y dx  y
(i) x dy + y dx = d (xy) (ii) 2
d 
x x
x dy  y dx  y x dy  y dx  y
(iii)  d  log  (iv) 2 2
 d  tan 1 
xy  x x y  x

x dy  y dx x dy  y dx 1 
(v)
xy
 d  log  xy   (vi) 2
x y 2
2
 
 d  log x 2  y 2 


2xy dy  y 2 dx  y2  ye x dx  e x dy  ex 
(vii)  d   (viii)  d  
xy  x  y2  y 
29
 Ex.1. Show that (2x + y2 + 2xz) dx + 2xy dy + x2dz = 0 is integrable (i.e., condition of
integrability is satisfied).
Sol. Comparing the given equation with Pdx + Qdy + Rdz = 0
We get, P = 2x + y2 + 2xz ; Q = 2xy ; R = x2
Now the condition of integrability is
 R Q   R P   Q P 
P   Q    R  0
 y z   x z   x y 
Substituting the values of P, Q, R in it, we get
(2x + y2 + 2xz) (0 – 0) – 2xy (2x – 2x) + x2 (2y – 2y) = 0
Showing that the condition of integrability is satisfied and hence the given equation is integrable.
Ex.2. Solve (yz + xyz) dx + (zx + xyz) dy + (xy + xyz) dz = 0
Sol. Comparing the given equation with Pdx + Qdy + Rdz = 0
We get P = yz + xyz ; Q = zx + xyz ; R = xy + xyz
Now the condition of integrability is

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

 yz 1  x   x  xz    x  xy   zx 1  y   y  yz    y  xy 
 xy 1  z   z  yz    z  xz 

 yz 1  x  x  z  y   zx 1  y  y  z  x   xy 1  z  z  y  x 

 xyz  z  y    z  x    y  x    x  z  y   y  z  x   z  y  x 

 xyz  0  0  0
This shows that the given equation is integrable.
Now dividing the whole equation by xyz, then given equation becomes

1  1  1 
  1 dx    1 dy    1 dz  0
x  y  z 
On integration, we get
log x + x + log y + y + log z + z = C
or log (xyz) + x + y + z = C
which is the required general solution, C being an arbitrary constant.
Ex.3. Solve (y2 + z2 – x2) dx – 2xy dy – 2xz dz = 0
Sol. As usual, we see that the condition of integrability is satisfied. Now rearranging the terms
of the given equation as

x 2

 y 2  z 2 dx  2 x 2 dx  2 xy dy  2 x zdz

or x 2
 y2  z2  dx  2 x  xdx  ydy  zdz 
30
 dx 2 xdx  2 ydy  2 zdz
or 
x x2  y 2  z 2
On integration, we get


log x  log c  log x 2  y 2  z 2 
or x 2  y 2  z 2  cx
is the required general solution.
Ex.4. Solve (2x2y + 2xy2 + 2xyz + 1) dx + (x3 + x2y + x2z + 2xyz + 2y2z + 2yz2 + 1) dy
+ (xy2 + y3 + y2z + 1) dz = 0
Sol. As usual, it may be verified that the condition of integrability is satisfied. Now rearranging
the terms of the given equation as

2 xy  x  y  z   1 dx   x 2  x  y  z   2 yz  x  y  z   1
 
dy  y 2  x  y  z   1 dz  0

or  x  y  z   2 xy dx  x 2 dy  2 y zdy  y 2dz   dx  dy  dz  0

or  2 xy dx  x dy    2 yzdy  y dz    dxxdyy zdz   0

2 2

On integration, we get
x2y + y2z + log (x + y + z) = C
This is the required general solution.
yz xz  y
Ex.5. Solve 2 2
dx  2 2
dy  tan 1   dz  0
x y x y  x
Sol. It can be easily verified that the condition of integrability is satisfied. Arranging the terms of
the given equation as
ydx  xdy dz

 y z
x 2

 y 2 tan 1  
x
.....(13)

 y xdy  ydx
Taking tan 1    s , so that  ds . Then equation (13) becomes
x 2 y2 
x 1  2 
 x 

ds dz
or  
s z
Integrating  log s  log z  log c
1
or s
cz
 y 1
i.e. tan 1   
 x  cz
31
  y 1
which gives    tan  
x  cz 
This is the required general solution.
2.3.2 Method for Homogeneous Equations
The equation Pdx + Qdy + Rdz = 0 is called a homogeneous equation if P, Q, R are homoge-
neous functions of x, y, z of the same degree. In such a case one variable is separated from the other
two by the substitution
x = uz, y = vz .....(14)
then dx = udz + zdu, dy = vdz + zdv .....(15)
Further, let
P  z n f1  u , v  , Q  z n f 2  u, v  and R  z n f3  u, v  .....(16)
Hence the given equation Pdx + Qdy + Rdz = 0 becomes
z n 1  f1  u, v  du  f 2  u , v  dv  z n uf1  u , v   vf 2  u , v   f 3  u , v  dz  0
On multiplying by z, we get
z n  2  f1  u , v  du  f 2  u , v  dv  z n1 uf1  u , v   vf 2  u , v   f 3  u , v  dz  0 .....(17)
Now following two cases arise :
Case I : Px + Qy + Rz = 0
If Px + Qy + Rz = 0 that is by substituting the values of x, y from (14) and P, Q, R from (16) in
it, we find
z n 1 uf1  u, v   vf 2  u , v   f3  u , v   0
then the coefficient of dz in equation (17) will become zero and hence it reduces to
f1  u , v  du  f 2  u , v  dv  0 .....(18)
which can be integrated easily.
Case II : Px + Qy + Rz  0
In this case the coefficient of dz will not be zero and therefore equation (17) may be written as.

f1  u , v  du  f 2  u , v  dv dz
 0 ...(19)
uf1  u, v   vf2  u, v   f3  u, v  z
Now since the given equation Pdx + Qdy + Rdz = 0 is integrable so equation (19) will be an
exact differential and hence this equation may be integrated easily.
2.3.3 Working Rule for Solving Homogeneous Equations
(i) First of all verify the condition of integrability.
(ii) If Px + Qy + Rz = 0, then substitute x = uz, y = vz and solve
1
(iii) If Px + Qy + Rz  0 then will be an integrating factor of the homogeneous
Px  Qy  Rz
equation Pdx + Qdy + Rdz = 0. After multiplying this equation by this integrating factor and rearranging
the terms we can integrate the equation by inspection.
32
   
Ex.6. Solve z 2 dx  z 2  2 yz dy  2 y 2  yz  xz dz  0 
Sol. Comparing the given equation with the standard equation Pdx + Qdy + Rdz = 0, we get
P = z2, Q = z2 – 2yz, R = 2y2 – yz – xz
The given equation is homogeneous of degree 2. Now first of all we test the condition of inte-
grability

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

   
 z 2  4 y  z  2 z  2 y   z 2  2 yz   z  2 z   2 y 2  yz  xz  0  0 

 6 yz 2  3z 3  3z 3  6 yz 2  0
Hence the condition of integrability is satisfied
Further, Px  Qy  Rz  xz 2  yz 2  2 y 2 z  2 y 2 z  yz 2  xz 2  0
Therefore, we substitute
x = uz, y = vz
Hence dx = udz + zdu, dy = vdz + zdv
and the given equation reduces to

 
z 2  udz  zdu   z 2 1  2v  vdz  zdv   z 2 2v 2  v  u dz  0

or du + (1 – 2v) dv = 0
Integrating, we get
u + v – v2 = C
or xz + yz – y2 = cz2
This is the required general solution.

Ex.7. Solve  yz  z  dx  xzdy  xydz  0

2

Sol. On comparing the given equation with Pdx + Qdy + Rdz = 0,

we have P = yz + z2, Q = – xz, R = xy
Here the given equation is homogeneous of degree 2 and the condition of integrability is satisfied
(do your self)
Now Let D = Px + Qy + Rz
= x (yz + z2) – xyz + xyz = xz (y + z)  0
Multiplying the given equation by integrating factor 1/D, we get

 yz  z  dx  xz dy  xy dz  0
2
.....(17)
D
Now d  D   d  xz  y  z     z dx  xdz  y  z   xz  dy  dz 

or d  D   z  y  z  dx  x  y  2 z  dz  xz dy

33
 Now rewriting the numerator of (17) as

 
d  D   d  D   yz  z 2 dx  xzdy  xydz  d  D   2 xz  dy  dz 
 Equation (17) becomes
d  D  2 xz  dy  dz 
 0
D D

d  D 2 xz  dy  dz 
or  0
D xz  y  z 

Integrating, log D – 2 log (y + z) = log C

or D = C (y + z)2
or xz (y + z) = C (y + z)2
or xz = C (y + z)
which is the required general solution, C being an arbitrary constant.

Ex.8. Solve  2 xz  yz  dx   2 yz  xz  dy   x 2  xy  y 2  dz  0
Sol. First of we verify the condition of integrability (do yourself). Since the given equation is
homogeneous, so putting
x = uz, y = vz so that dx = udz + zdu, dy= zdv + vdz .....(18)
Now using these values in given equation, we get

 2uz 2
    
 vz 2  udz  zdu   2vz 2  uz 2  vdz  zdv   u 2 z 2  4vz 2  v 2 z 2 dz  0

or  2u  v  udz  zdu    2v  u  vdz  zdv    u 2  uv  v 2  dz  0

or   
z  2u  v  du   2v  u  dv   u  2u  v   v  2v  u   u 2  uv  v 2  dz  0


or 
z  2udu   udv  vdu   2vdv   u 2  uv  v 2 dz  0
or  
z  du 2  d  uv   dv 2   u 2  uv  v 2 dz  0

or

d u 2  uv  v 2   dz  0
2 2
u  uv  v z
On integration, we get
log (u2 – uv + v2) + log z = log C
or  
z u 2  uv  v 2  C

 x2 x y y 2 
or z  2    2   C
z z z z 

or x2 – xy + y2 = cz
which is the required general solution.
34
 Ex.9. Solve yz  y  z  dx  zx  x  z  dy  xy  x  y  dz  0
Sol. First of all verify the condition of integrability (do your self). Since the given equation is
homogeneous, we put
x = uz, y = vz so that dx = zdu + udz, dy = zdv + vdz .....(19)
Substituting these in the given equation, we get
v (v + 1) z3 (zdu + udz) + u (u + 1) z3 (zdv + vdz) + uv (u + v) z3dz = 0
or [v (v + 1) du + u (u + 1) dv] z4 + [uv(v + 1) + uv (u + 1) + uv (u + v)] z3dz = 0
or [v (v + 1) du + u (u + 1) dv] z4 + 2uv (u + v + 1) z3dz = 0
Dividing above equation by uv (u + v + 1) z4, we get
 v  1 du   u  1 dv  2 dz  0
u  u  v  1 v  u  v  1 z

1 1  1 1  dz
or    du     dv  2  0
 u u  v 1   v u  v 1 z
du dv du  dv dz
or   2 0
u v u  v 1 z
On integration, we get
log u  log v  log  u  v  1  2log z  log C
or uvz2 = C (u + v + 1)
 x  y  2 x y 
or    z  C    1 by using (9)
 z  z  z z 
or xyz = C (x + y + z)
this is the required general solution.
2.3.4 Method of Auxiliary Equations
Let Pdx + Qdy + Rdz = 0 .....(20)
by the given equation. Its condition of integrability is
 R Q   R P   Q P 
P   Q    R    0. .....(21)
 y z   x z   x y 
On comparing (20) and (21), we obtain simultaneous equations, known as auxiliary equations.
dx dy dz
 
R Q  R P   Q P 
     .....(22)
y z  x z   x y 
For solving (22) let u = c1 and v = c2 be their two integrals. After finding the value of Adu + Bdv
= 0 and comparing it with the given equation, the values of A and B will be obtained. Integration of Adn
+ Bdv = 0, will give the required solution.
R Q R P Q P
This method will fail if  ,  and 
y z x z x y
35
 Ex.10. Solve xz3dx – zdy + 2ydz = 0
Sol. Here the condition of integrability is satisfied (do your self) now given equation is
xz3dx – zdy + 2ydz = 0 .....(23)
Comparing it with Pdx + Qdy + Rdz = 0, we have
P = xz3, Q = – z, R = 2y
The auxiliary equations of the given equation are
dx dy dz
 
 R Q   R P   Q P 
 y  z   x  z   x  y 
   
dx dy dz
or  2

2  1 3xz 0
dx dy dz
or  
1 xz 2 0
Taking last two terms, we get
dz = 0 so that z = c1 = u (say) ......(24)
Taking first two terms, we get
xz2dx – dy = 0
or 2xu2dx – 2dy = 0 [by using (23)]
Integrating, x2u2 – 2y = c2 = v (say)
or x2z2 – 2y = v [by using (23)] .....(25)
Substituting the values of u and v from (24) and (25) in Adu + Bdv = 0, we get
Adz + Bd (x2z2–2y) = 0
or Adz + B (2xz2dx + 2x2zdz – 2dy) = 0
or 2Bxz2dx – 2Bdy + (A + 2Bx2z) dz = 0 .....(26)
Comparing (23) and (26), we have
xz 3  2 Bxz 2 ,  z  2 B

1
and 2 y  A  2 Bx 2 z  B    z and A  2 y  2 Bx 2 z  2 y  x 2 z 2
2
1
or B    u and A = – v, [by using (24) and (25)]. Substituting these values
2
of A and B in Adu + Bdv = 0, we get
1
vdu    udv  0
2
1 1
or dv  2   du
v u
On integration, we get
log v  2 log u  log c
v = cu2 .....(27)
36
 Putting the values of u and v from (24) and (25) in (27), we get
x 2 z 2  2 y  cz 2
which is the required general solution.
2.3.5 General Method
Step I : Let the condition of integrability is satisfied for the given equation
Pdx  Qdy  Rdz  0 .....(28)
Step II : Treating one of the variables of (28), say z, as a constant then dz = 0 and the given
equation is reduced to
Pdx + Qdy = 0
Integrating it, keeping z as constant. If necessary the help of an integrating factor may be taken.
Let the result so obtained be
u (x, y, z) = f (z) .....(29)
where f (z) is a function of z alone. This is possible because the arbitrary function f (z) is con-
stant with respect to x and y.
Step III : Now we differentiate (29) totally with respect to x, y, z and then compare the result
with the given equation (28). We will get a relation between df and dz. If the of df and dz involve func-
tions of x and y, it would be possible to eliminate them with the help of (22). Thus we shall get an equa-
tion in df and dz which will be independent of x and y.
Step IV : The values of f (z) will be obtained by integrating the above equation. After sustituting
it in (32), we get the complete solution.
Remark : General method, for solving the total differential equation of the type
Pdx + Qdy + Rdz = 0
should be adopted only when the equations are non-homogeneous and the method of inspection
fails.
Ex.11. Solve 3x2dx + 3y2dy – (x3 + y3 + e2z) dz = 0
Sol. Here, the condition of integrability is satisfied. Let us treat z as constant, so that dz = 0.
Then the given equation become
3x2dx + 3y2dy = 0
On integration, we get
x3 + y3 = f (z) (say) .....(30)
where the constant of integration has been taken as a function f (z) as we have treated z as
constant.
Now differentiating (30), we have
3 x 2 dx  3 y 2 dy  f   z  dz  0 .....(31)
Comparing (31) with the given equation, we get
f   z   x 3  y 3  e 2z

or f   z   f  z   e2 z [by using (30)]

37
 df
or  f  e 2z , which is a linear equation having integrating factor as
dz
 1 dz
IF  e   e  z . Hence the solution is

 
f  z  e  z   e 2 z e  z dz  c  e z  c

or f  z   e 2 z  ce z

or x3  y 3  e 2z  ce z [by using (30)]

Which is the required general, C being an arbitrary constant.

Ex.12. e x
   
y  e z dx  e y z  e x dy  e y  e x y  e y z dz  0 
Sol. Here, the condition of integrability is satisfied. Let us treat z as constant so that dz = 0.
Then the given equation becomes

 e y dx  e dy    e zdy  e dx   0
x z y z

On integration. we get
ex y  e y z  ez x  f  z  .....(32)
Now differentiating equation (32), we obtain

 e y  e  dx   e z  e  dy   e
x z y x y

 e z x dz  f   z  dz .....(33)
Comparing (33) with the given equation, we get
e y  ez x  f  z   e y  ex y  e y z
which gives
f  z   ex y  e y z  ez x  f  z  (by 32)
df
or  f
dz
Integrating, we get
f (z) = cez
Putting the value of f (z) from equation (32), we get the required general solution as
exy + eyz + ezx = cez
Ex.13. Solve y 2 z  x cos x  sin x  dx  x 2 z  y cos y  sin y  dy

 xy  y sin x  x sin y  xy cos z  dz  0

Sol. Here, the condition of integrability, is satisfied. Let us treat z as constant so that dz = 0.
Then the given equation becomes
y 2 z  x cos x  sin x  dx  x 2 z  y cos y  sin y  dy  0
x cos x  sin x y cos y  sin y
or dx  dy  0
x2 y2
38
  sin x   sin y 
or d d 0
 x   y 
On integration, we get
sin x sin y
  f  z .....(34)
x y
where the constant of integration has been taken as a function f (z) as we have treated z as
constant.
Now differentiating (34), we get
x cos x  sin x y cos y  sin y
dx  dy  f   z  dz
x2 y2

or zy 2  x cos x  sin x  dx  zx 2  y cos y  sin y  dy  x 2 y 2 z f   z  dz  0 .....(35)

Comparing (35) with the given equation, we have
 x 2 y 2 z f   z   xy  y sin x  x sin y  xy cos z 

sin x sin y
or z f   z     cos z  f  z   cos z [by using (34)]
x y

df 1 cos z
or  f  , which is a linear equation having integrating factor (IF)
dz z z
as
1 z  dz
IF  e   elog z  z and the solution is
  cos z 
z f z   z  dz  c   sin z  c
 z 

 sin x sin y 
or z    c  sin z [by using (34)]
 x y 
which is the required general solution, c being an arbitrary constant.

Self Learning Exercise-I

1. Write down pfaffian differential equation in n variables.
2. Write the condition when an equation of the type Mdx + Ndy = 0 become exact.
3. What is the condition of integrability for the equation Pdx + Qdy + Rdz = 0 ?
4. Which equations are called homogeneous ?

2.4 Geometrical Meaning of Pdx + Qdy + Rdz = 0

We know that direction cosines of the tangent at a point (x, y, z) on a curve are proportional to
dx, dy, dz. Therefore, the differential equation Pdx + Qdy + Rdz = 0 .....(1)
signifies that the tangent to a curve at the point (x, y, z) is perpendicular to a line, whose direction co-
sines are proportional to P, Q, R.
39
 Whereas the simultaneous equations
dx dy dz
  .....(2)
P Q R
express that the tangent to a curve at a point (x, y, z) is parallel to a line with direction cosines propor-
tional to P, Q, R.
We thus have two sets of curve, and if they intersect, they intersect at right angle. Now we dis-
cuss two cases.
Case I : If the equation Pdx + Qdy + Rdz = 0 is integrable, it means that family of surfaces can
be obtained such that all curves on it are perpendicular to the curves represented by the equation (2) at
all points where curves cut the surface. Since the solution of equation (1) will be of the form (x, y, z) =
C and that of (2) will be of the form f1 (x, y, z) = C1 and f2 (x, y, z) = C2 , it means that in this case an
infinite number of surfaces can be drawn to cut orthogonally a doubly infinite set of curves.
dx dy dz
Case II : If equation (1) is not integrable than the curves represented by   may
P Q R
not admit of such a family of orthogonal surfaces.
Ex.1. Solve Find the system of curves satisfying the differential equating.

x2 y2
xdx  ydy  c 1   dz  0 ....(3)
a 2 b2
which lie on the surface
x 2 y2 z2
1 2  2  2 ....(4)
a b c
Sol. Equation of the given surface can be written as
x2 y 2 z 2
  1 ....(5)
a2 b2 c 2
with the help of (3), the given equation can be written as
xdx+ydy +zdz=0
on Integration, we get
x2 + y2 + z2 = k ....(6)
Hence the required system of curves will be given by the intersection of (5) and (6).
Ex.2. Find the differential equation of the family of twisted cubic curves y = ax2 ,
y2 = bzx. Show that all these curves cut orthogonally the family of ellipsoids
x2 + 2y2 + 3z2 = c2.
Sol. Family of twisted cubic curves as given in question is
y = ax2 .....(7)
y2 = bzx ....(8)
On differentiating (7), we get
dy = 2ax dx

40
 y
or dy  2 dx [by using (7)]
x
or 2ydx - xdy = 0 ....(9)
Now similarly, differentiating (8), we obtain
2ydy = b(zdx + xdz)
y2
or 2 ydy  ( zdx  xdz ) [by using (8)]
zx
or yz dx - 2zx dy + xy dz = 0 ....(10)
From (9) and (10), we get
dx dy dz
2
 2

 x y 2 xy ( 2 zx )2 y  (  x )yz
dx dy dz
or 2
 2

 x y 2 xy 3xyz
dx dy dz
or  
x 2 y 3z
which are the required differential equations of the family of curves.
The differential equations of the surfaces which are cut orthogonally by the given curves is
x dx + 2y dy + 3z dz = 0
Integrating, we get
x2 + 2y2 + 3z2 = k = c2 (say)

2.5 Equations Containing More than Three Variables

Let us consider an equation of the form

Pdx + Qdy + Rdz + Tdt = 0 ....(1)
Treating t as constant, so that dt = 0, then equation (1) becomes
Pdx + Qdy + Rdz = 0 ....(2)
Condition of integrability for equation (2) will be
 R Q   R P   Q P 
P   Q    R  0 ....(3)
 y z   x z   x y 
Similarly If we take z, x and y as constant, then we get dz = 0, dx = 0, dy = 0. The condition of
integrability in these cases will be
 T Q   T P   Q P 
P   Q   T   0 ....(4)
 y t   x t   x y 

 T R   T Q   R Q 
Q    R   T   0 ....(5)
 z t   y t   y z 
 T P   T R   P R 
and R    P  T   0 ....(6)
 x t   z t   z x 
41
 Hence we see that in the case of more than three variables, the condition of integrability must be
satisfied for the coefficients of all the terms taken three at a time.
Here we note that only three of the relations (3), (4), (5) and (6) are independent and the fourth
one can be derived from the remaining three.

2.6 Method for Obtaing Solution Involving Four Variables

If the condition of integrability is satisfied, then the solution the total differential equation can be
obtained by two methods.
Method 1. By Inspection : In this method we can arrange the coefficients in such way that the
given equation is directly integrable.
Method 2. In this method, we take any two of the four variables constant. The equation is inte-
grated and the constant of integration is taken as the function of those variables which were kept con-
stant. The result is compared with the given equation after obtaining its differential and in such a way the
values of constants of integration are obtained. This will give the complete solution.
Ex.1. Solve (2x + y2 + 2xz) dx + 2xy dy + x2dz = dt.
Sol. We can write the given equation as
(2x + y2 + 2xz) dx + 2xy dy + x2dz – dt = 0
we can easily verify the condition of integrability as given by equations (3), (4), (5) and (6) of
§2.5.
Now the given equation can be written as 2xdx + (y2dx + 2xy dy) + (2xzdx + x2dz) - dt = 0.
Which on integration gives the complete solution as x2 + xy2 + x2z - t = c.
Ex.2. Solve z (y + z) dx + z (t - x) dy + y (x - t) dz + y (y + z) dt = 0
Sol. On comparing the given question by the standard equation Pdx + Qdy + Rdz + Tdt = 0,
we get
P = z(y + z), Q = z(t – x), R = y(x – t), T = y(y + z)
Here we can easily show that the conditions of integrability (equations (3), (4), (5) and (6) of
§2.5) are satisfied.
Now we solve the given question by treating two variables as constant. Treating y and z as con-
stants so that dy = 0 and dz = 0. Then the given equation reduces to
z(y + z) dx + y(y + z) dt = 0
or zdx + ydt = 0
On integration, we get
zx + yt = f (x, z) ( say) .....(7)
Now on differentiation (7), we get
zdx + tdy + xdz + ydt = df
or (y + z) (zdx + tdy + xdz + ydt) = (y + z) df
42
 or z(y + z) dx + t(y + z) dy + x(y + z) dz + y(y + z) dt = (y + z) df .....(8)
Comparing (8) with the given equation, we have
t(y + z) dy + x(y + z) dz – (y + z) df = z(t – x) dy + y(x – t) dz
or (ty + xz) dy + (ty + xz) dz = (y + z) df
or (ty + xz) (dy + dz) = (y + z) df
or f (dy + dz) = (y + z) df [by using (7)]
dt dy  dz
or  .....(9)
f yz
Integration of (9) yields
log f = log (y + z) + log c
or f = c(y + z)
or zx + yt = c(y + z) [by using (7)]

2.7 Total Differential Equation of Second Degree

It the given equation be of the form

Adx2 + Bdy2 + Cdz2 + 2Ddydz + 2Edzdx + 2Fdxdy = 0
where A, B, C, D, E and F are functions of x, y, and z then it can be easily resolved into factors, if
ABC + 2DEF – AD2 – BE2 – CF2 = 0
Let the two factors be
Pdx + Qdy + Rdz = 0
and Pdx + Qdy + Rdz = 0
The solutions of either of these may be obtained by the methods discussed earlier. The two gen-
eral solutions taken together constitute the complete solution.
Ex.1. Solve (xdx + ydy + zdz)2z = {(z2x2y2) (xdx + ydy + zdz) dz}
Sol. We can factorize the given equation as
(xdx + ydy + zdz) {z(xdx + ydy + zdz) – (z2 – x2 – y2) dz} = 0
i.e., xdx + ydy + zdz = 0 .....(1)
and z(xdx + ydy + zdz) – z2dz + (x2 + y2) dz = 0 .....(2)
On integration of (1), we get
x2 + y2 + z2 = c1 .....(3)
To obtain the integral of (2), the equation may be written as
z(xdx + ydy) + (x2 + y2) dz = 0
or z2(2xdx + 2ydy) + (x2 + y2) 2zdz = 0
On integration, we get
z2(x2 + y2) = c2 .....(4)
43
 Hence the required solution is
(x2 + y2 + z2 – c1) (z2x2 + z2y2 – c2) = 0

Self Learning Exercise-II

1. The direction cosines of the tangent at a point (x, y, z) on a curve are proportional to _, _, _.
2. What is the equation of family of twisted cubic curves ?

2.8 Summary

In this unit, you studied about the condition of integrability of total differential equation and vari-
ous methods for solving it. Now you must be knowing about the geometrical meaning of Pdx + Qdy +
Rdz = 0 and methods of finding solution of total differential equation containing three or more than three
variables

2.9 Answers of Self Learning Exercises

Exercise I
n
1.  ui dxi  0 , where u (i = 1, 2 .......... n) are n functions of some or all of n independent vari-
i
i 1
ables x1, x2 ,....., xn.
M N
2. 
y x

 R Q   R P   Q P 
3. P    Q    R  0
 y z   x z   x y 
4. Equation Pdx + Qdy + Rdz = 0 is called homogeneous if P, Q, R are homogenous functions of
x, y, z of the same degree.

Exercise II
1. dx, dy, dz
2. y = ax2, y2 = bzx

2.10 Exercise

Solve the following differential equations

2
1. (2x2 + 2xy + 2xz2 + 1) dx + dy + 2zdz = 0 [Ans. e x  x  y  z 2   c ]
y
2. xdy – ydx + 2x2z dz = 0 [Ans.  z2  c ]
x
3. (y + a)2 dx + zdy – (y + a)dz = 0 [Ans. z = (x + c) (y + a)]

44
 4. yzdx + zxdy + xy dz = 0 [Ans. xyz = c]
5. (ydx + xdy) (a – z) + xydz = 0 [Ans. xy = c(a – z)]
6. zdz + (x – a) dx = {h2 – z2 – (x – a)2 }1/2 dy [Ans. h2 – z2 – (x – a)2 = (y – c)2]
7. zydx = zxdy + y2dz [Ans. x – cy – y log z = 0]
8. yz2(x2 – yz) dx + x2z(y2 – xz) dy + xy2(z2 – xy) dz = 0 [Ans. x2z + yz2 + xy2 = cxyz]
9. (y2 + yz + z2) dx + (x2 + xz + z2) dy + (x2 + xy + y2) dz = 0
[Ans. xy + yz + zx = c(x + y + z)]
10. (x – y – z + 2xy + 2xz ) dx + (y – z – x + 2yz + 2yx ) dy + (z2 – x2 – y2 + 2zx + 2zy )
2 2 2 2 2 2

dz = 0 [Ans. x2 + y2 + z2 = c (x + y + z)]
11. 2(y + z) dx – (x + z) dy + (2y – x + z) = 0 [Ans. (x + z)2 = c(y + z)]
12. z(z – y) dx + (z + x)zdy + x(x + y)dz = 0 [Ans. z(x + y) = c(x + z)]
x y z z
13. (x2y – y3 – y2z) dx + (xy2 – x2z – x3) dy + (xy2 + x2y) dz = 0 [Ans. y  x  x  y  c ]

14. (y2 + yz) dx + (xz + z2) dy + (y2 – xy) dz = 0 [Ans. y(x + z) = c(y + z)]
2 2 2 2 2 2
15. (y + z + 2xy + 2xz) dx + (x + z + 2xy + 2yz) dy + (x + y + 2xz + 2yz) dz = 0
[Ans. x(y2 + z2) + y(z2 + x2) + z(x2 + y2) = c]
16. (2xy + z2) dx + (x2 + 2yz) dy + (y2 + 2xz) dz = 0 [Ans. x2y + y2z + z2x = c]
nx  lz
17. (mz – ny) dx + (nx – lz) dy + (ly – mx) dz = 0 [Ans. c]
mz  ny
18. (cos x + exy) dx + (ex + eyz) dy + ey dz = 0 [Ans. eyy + eyz + sin x = c]
x2  2z c
19. 2xz(y – z) dx + z(x2 + 2z) dy + y(x2 + 2y) dz = 0 [Ans. y  z  z  2 ]

20. xdy – ydx – 2x2zdz = 0 [Ans. y = x (c – z2)]

21. (z + z2) cos x dx – (z + z2) dy + (1 – z2) (y – sin x) dz = 0 [Ans. y = sin x – cze–z)]
22. y sin dx + x sin dy – xy sin dz – xy cos d = 0 [Ans. xy = c sin ez)]
23. yzdx + 2xzdy – 3xydz = 0 [Ans. xy2 = cz3]
24. (2y2 + 4az2x2) xdx + [3y + 2x2 + (y2 + z2)–1/2] ydy + [4z2 + 2ax4 + (y2 + z2)–1/2] zdz = 0

[Ans. x 2 y 2  ax 4 z 2  y 2  z 2  y 2
 z 2   c. ]

25. Find the equation of the curve that passes through the point (3, 2, 1) and cut orthogonally the
family of surfaces x + yz = c
[Ans. y2 – z2 = 3, y + z = 3ex–3]



45
Unit 3 : Partial Differential Equations of Second order,
Monge’s Method
Structure of the Unit
3.0 Objective

3.1 Introduction

3.2 Solution of P.D.E. of Second order by Inspection.

3.3 Exercise – I

3.4 Monge’s Method for Solving Equation of the Type Rr + Ss + Tt = V

3.5 Monge’s Method for Solving Equation of the Type Rr + Ss + Tt + U(rt – s2) = V

3.6 Summary

3.7 Answers of self-Learning Exercises

3.8 Exercise – II

3.0 Objective

The purpose of this unit is to discuss partial differential equations of order two with variable co-
efficients. Here you will learn how a large class of second order partial differential equations may be
solved by using the methods applicable for solving ordinary differential equations ? You will also study
Monge’s method for solution of some special type of second order partial differential equations.

3.1 Introduction

A partial differential equation (P.D.E) is said to be of order two, if it involves at least one of the
differential coefficients r, s, t and none of order higher than two. The general form of a second order
partial differential equation in two independent variables x, y is given as
as F(x, y, z, p, q, r, s, t) = 0 ;

z z 2 z 2z 2 z
where p , q , r 2, s , t 2
x y x xy y
The most general linear partial differential equation of second order in two independent variable
x and y with variable coefficient is given as
Rr + Ss + Tt + Pp + Qq + Zz = F
where R, S, T, P, Q, Z, F are functions of x and y only and not all R, S, T are zero.
46
3.2 Solution of P.D.E. of Second Order by Inspection

Before taking up the general equation of second degree P.D.E., we discuss the solution of simple
problems which can be integrated merely by inspection. On two successive integral of given P.D.E., we
get the general solution which is a relation in x, y, z. To understand this, we discuss the following prob-
lems.
Ex.1. Solve t + s + q = 0
Sol. We can write the given problem as

 2 z  2 z z
  0
y 2 xy y
Integrating with respect to y, treating x as constant, we get
z z
  z  f  x or p + q = f (x) – z
y x
which is the form of standard Lagrange’s linear equation Pp + Qq = R, so the auxiliary equation will be
dx dy dz
 
1 1 f  x  z
from first two terms, we obtain
x – y = c1 (constant) .....(1)
and from first and last terms, we have
dz
 z  f x .....(2)
dx
which is linear differential equation of first order having integrative factor ex.
Hence the solution of (2) will be
z ex = f (x) exdx + c2 (constant)
Therefore the required solution of given equation will be (by using (1)]
zex – (x) =  (x – y)
where c2 is a function of c1 or of (x – y).
Ex.2. Solve t – qx = x2
Sol. We can write the given problem as
q
 qx  x 2 .....(3)
y
which is linear in q and y having integrating factor e–xdy = e–xy. Therefore the solution of (3) is
q e–xy = x2 e–xy dy + f (x) (as x is constant)
or q e–xy = –xe–xy + f (x)
z
or   x  f  x  e xy
y
Again integrating with respect to y (treating x as constant), we get.
1
z   xy  f  x  e xy    x  .
x
47
 x
Ex.3. Solve xt =
y2
Sol. We can write the given problem as

2 z 2 z x
 2  2
xy y y
Integrating with respect to y (treating x as constant), we get
z z x
    f  x
x y y
x
or pq    f  x
y
which is the form of standard Lagrange’s linear equation Pp + Qq = R, so the auxiliary equation will be
dx dy dz
 
1 1  x y  f  x 
From first two terms, we obtain
x + y = c1 (constant) .....(4)
and from first and last terms, we have
x
dz  dx  f  x  dx
y
x
or dz  dx  f  x  dx [by using (4)]
c1  x
 c 
or dz  1   dx  f  x  dx
 c1  x 
On integrating, we get
z  x  c1 log  c1  x    f  x  dx  c2

or z  x  c1 log y    x   F  x  y 
where c2 is a function of c1 or of (x + y).
Ex.4. Solve rx = (n – 1) p
Sol. We can write the given problem as

2 z z
x   n  1
x 2 x

2 z
x 2  n  1
or z x
x
Now integrating both sides with respect to x treating y as constant, we get
 z 
log     n  1 log x  log f1  y 
 x 
48
 z
or  x n 1 f1  y 
x
Again integrating w.r.t. x treating y as constant, we obtain
xn
z f1  y   f 2  y 
n
Ex.5. Solve 2yq + y2t = 1
Sol. We can write the given problem as
q
2 yq  y 2 1
y
 2
or
y
 
y q 1

Now integrating both side with respect to y treating x as constant, we get

y2q = f1(x)
z 1
or q  f1  x 
y y 2
Again integrating with respect to y, we obtain
1
z     f1  x   f 2  x  .
 y
Ex.6. Show that a surface passing through the circle z = 0, x2 + y2 = 1 and satisfying
the differential equation s = 8xy is z = (x2 + y2)2 – 1
Sol. We can write the given differential equation as

2 z   z 
    8 xy
xy x  y 
Integrating with respect to x, we get
z
 4x2 y  f  y 
y
Again integrating with respect to y, we obtain
z  2 x 2 y 2   f  y  dy  1  x 

or z  2 x 2 y 2  2  y   1  x  .....(5)

where 2  y    f  y  dy
where 1 and 2 are two arbitrary functions.
Now given circle is
x2 + y2 = 1, z = 0
Putting z = 0 in (5), we get
2x2y2 + 2 (y) + 1 (x) = 0 .....(6)
2 2 2 2 2 2
Now, x +y = 1 (x + y ) = 1
49
 or 2x2y2 + x4 +y4 = 1 .....(7)
On comparing (6) with (7), we get
2  y   1  x   x 4  y 4  1
Substituting this in (5), we obtain
z  2 x2 y 2  x 4  y 4  1
2
or 
z  x2  y2  1
Hence the result.

Self-Learning Exercise-I

1. What is the general form of a second order p.d.e. in two independent variables x and y ?
2. The most general linear p.d.e. of second order in two independent variables x and y is ....... .
3. The solution of r = 6x is .......

3.3 Exercise-1

Solve the following partial differential equations :

1 3
1. ar = xy [Ans. az  x y  x f  y  F  y ]
6
2. r = 2y2 [Ans. z  x 2 y 2  x f  y   F  y  ]

3. s – t = x/y2 [Ans. z  ( x  y ) log y  f  x   F  x  y  ]

4. xr + p = 9x2y2 [Ans. z  x3 y 3  log a f  y   F  y  ]

1 2 1 y2
5. yt – q = xy [Ans. z  xy log y  xy 2 f  x  F  x ]
2 4 2
6. log s = x + y [Ans. z  e x  y  f  y   F  x  ]

7. p + r + s = 1
y

y y

[Ans. z  x  e  y e  F  y   e f  x  y  ]

8. ys + p = cos (x + y) – y sin (x + y) [Ans. yz  y sin  x  y   f  x   F  y  ]

x2
9. s = x/y + a [Ans. z  log y  axy  f  x   F  y  ]
2
It may he noted here that a p.d.e.f (x, y, z, p, q, r, s, t) = 0 can he integrated only in special
cases. The most important method of solution, due to Monge, is applicable to a wide class of
such equations but not to all equations.

3.4 Monge’s Method for Solving Equation of the Type Rr + Ss +Tt = V

Monge’s gives a method for solving p.d.e. of second order of the type
Rr + Ss + Tt = V .....(1)
50
 where R, S, T and V are, in general, functions of x, y, z, p and q. Indeed this a equation of first
degree in r, s and t. To solve such type of equations, first we determine the intermediate integrals. For
this we have
p p
dp  dx  dy
x y
or dp  rdx  sdy ....(2)
dp  sdy
hence r ....(3)
dx
q q
Similarly dq  dx  dy
x y
or dq  sdx  tdy ....(4)
dq  sdx
hence t ....(5)
dy
Now, r and t are eliminated from equation (1) with the help of (3) and (5). Thus we get an
equation in s as
 dp  sdy   dq  sdx 
R   Ss  T   V
 dx   dy 
or  Rdpdy  Tdqdx  Vdydx   s  Rdy 2  Sdydx  Tdx2   0 ....(6)
Equation (6) will be identically satisfied if we take
Rdpdy  Tdqdx  Vdydx  0 ....(7)
and Rdy 2  Sdydx  Tdx 2  0 ....(8)
which are called Monge’s subsidiary equations and will provide us the intermediate integrals. Here
we note that the equation (8) is quadratic for the ratio dy : dx and therefore can be decomposed into
two linear equations in dx and dy of the form
dy  m1dx  0 and dy  m2 dx  0
Now combining equations dy – m1dx = 0 and (7) with dz = pdx + qdy, two integrals u1 = u1(x,
y, z, p, q) and v1 = v1(x, y, z, p, q) can be obtained. Then we get u1 = f1(v1) as the first intermediate
integral. Similarly on combining equations dy – m2dx = 0 and (7) with dz = pdx + qdy, and following the
above procedure, the second intermediate integral u2 = f2 (v2) can be obtained.
From these two intermediate integrals, the values of p and q may be obtained in terms of x and
y and then substituting them in dz = pdx + qdy and integrating it, the complete integral of (1) is obtained.
Ex.1. Solve r = a2t by Monge’s method.
Sol. Comparing the given equation with Rr + Ss + Tt = V, we get R = 1, S = 0, T = – a2,V = 0.
The Monge’s subsidiary equations are given by
Rdpdy  Tdqdx  Vdydx  0
and Rdy 2  Sdydx  Tdx 2  0
51
 Substituting the values of R, S, T and V, the subsidiary equations will be
dpdy  a 2 dqdx  0 ....(9)

dy 2  a 2 dx 2  0 ....(10)
Equation (10) may be factorised as
 dy  adx   0 ....(11)

and  dy  adx   0 ....(12)

Combining equation (11) with subsidiary equation (9), we get
dp  adx   a 2 dqdx  0
or dp  adq  0 ( dx = 0, gives trivial solution) ....(13)
Now from (11) and (13) we obtain
y  ax  c1 , p  aq  c2
therefore the first intermediate integral is
 p  aq   f1  y  ax  ....(14)
Similarly combining (dy + adx) = 0 with subsidiary equation (9), we get the second intermediate
integral as
 p  aq   f 2  y  ax  ....(15)
Now from above two intermediate integrals (14) and (15) we deduce the value of p and q as.
1
p  f1  y  ax   f 2  y  ax  
2
1
q  f 2  y  ax   f1  y  ax  
2a 
Substituting these values of p and q in dz = pdx + qdy, we get
 dy  adx   dy  adx 
dz    f 2  y  ax     f1  y  ax 
 2a   2a 
On integration, we have
1 1
z 2  y  ax   1  y  ax 
2a 2a
Hence the required solution is
z  F1  y  ax   F2  y  ax 
Ex.2. Solve r + (a + b) s + abt = xy by Monge’s method.
Sol. Comparing the given equation with Rr + Ss + Tt = V, we have R = 1, S = a + b,T = ab, V
= xy. Here Monge’s subsidiary equations
Rdpdy + Tdqdx – Vdydx = 0
Rdy2 – Sdxdy + Tdx2 = 0
become dpdy + abdqdx – xydxdy = 0 .....(16)
and dy2 – (a + b) dxdy + ab dx2 = 0 .....(17)

52
Equation (17) may be factorised as
(dy – bdx) = 0 .....(18)
and (dy – adx) = 0 .....(19)
On integration y – bx = c1 .....(20)
y – ax = c2 .....(21)
Combining equation (18) with subsidiary equation (16), we get
dp (bdx) + abdqdx – xydx (bdx) = 0
or dp + adq – xy dx = 0
or dp + adq – x (c1 + bx) dx = 0 [by using (20)]
On integration, we get

c  b
p  aq   1  x 2    x 3  c3
2  3

x2 b
or p  aq   y  bx     x3  c 3 [by using (20)]
2  3

1 1
or p  aq    yx 2  bx3  c3
2 6
Therefore the first intermediate integral is
1 2 1 3
p  aq  yx  bx  f1  y  bx  .....(22)
2 6
Similarly, the second intermediate integral corresponding to equation (19) is
1 2 1 3
yx  ax  f 2  y  ax 
p  bq  .....(23)
2 6
Now from above two intermediate integrals (22) and (23), we deduce the values of p and q as
1 2 1 1
p x y   a  b  x3   a f 2  y  ax   b f1  y  bx  
2 6 ab 
1 3  1 
and q x    f1  y  bx   f 2  y  ax  
6  a b 
Substituting these values of p and q in dz = pdx + qdy, we get

1 2 1 1
dz  x ydx   a  b  x 3dx   af  y  ax dx  bf1  y  bx  dx  
2 6 a  b  2
1 1
 x 3dy   f  y  bx  dy  f 2  y  ax  dy 
6 a  b  1

1 1 1
or dz 
6
 6

3 x 2 ydx  x3 dy   a  b  x 3dx   af  y  ax  dx  bf1  y  bx  dx 
b  a   2
1
  f  y  bx  dy  f 2  y  ax  dy 
b  a   1
53
 1 1 1
or dz 
6
 
d x3 y   a  b  x3dx 
6
f  y  ax  dy  adx 
b  a  2
1
 f  y  bx  dy  bdx 
b  a  1
Integrating, we get the required solution as
1 3 1
z x y   a  b  x 4    y  ax   1  y  bx 
6 24
Ex.3. Solve x r + 2xy s + y2t = 0 by Monge’s method.
2

Sol. Comparing the given equation with Rr + Ss + Tt = V, we have R = x2, S = 2xy, T = y2,
and V = 0. Hence Monge’s subsidiary equations
Rdpdy  Tdqdx  Vdydx  0
Rdy 2  Sdx dy  Tdx 2  0
become
x 2 dpdy  y 2 dqdx  0 .....(24)
and x 2 dy 2  2 xy dy  y 2 dx 2  0 .....(25)
Equation (25) may be factorised as

 xdy  ydx 2  0
or  xdy  ydx   0 .....(26)
Combining it with the equation (24), we get
xdp (ydx) + y2dq dx = 0
or xdp + ydq = 0
or xdp + pdx + qdy + ydq = pdx + qdy
or d (xp) + d (yq) = dz
On integration, we get
px + qy = z + c1
Now equation (26) gives
y
 c2
x
Thus the intermediate integral will be
px + qy = z + f (c2)
which is of Lagrange’s form having the subsidiary equations
dx dy dz
 
x y z  f  c2 
First two terms gives
y
 c2
x
and the last two terms gives z  f  c2   cy
54
 Hence required solution is
 y  y
z  yf1    f2  
 z x
Ex.4. Solve (x – y) (xr – xs – ys – yt) = (x + y) (p – q) by Monge’s method.
Sol. Monge’s subsidiary equations in this case will be
x  x  y  dpdy  y  x  y  dqdx   x  y  p  q  dxdy  0 .....(27)
2 2
and x  dy    x  y  dxdy  y  dx   0 .....(28)
Factors of equation (28) are
xdy + ydx = 0,
which on integration gives xy = c1
and dx + dy = 0,
which on integration gives x + y = c2 . Combining equation (27) with (xdy + ydx) = 0, we get
 x  y  dp  dq    p  q  dx  dy 
On integration, we obtain
pq
 constant.
x y
Therefore the intermediate integral is
(p – q) = (x – y) f (xy)
for which the Lagrange’s subsidiary equation will be
dx dy dz f  xy  ydx  xdy   dz
  
1 1  x  y  f  xy  0
From first two terms, we get
From the last two relations, we get x + y = c2
dz + f (xy) d (xy) = 0
On integration
z = F1 (xy) + constant
Hence required solution is
z = F1 (xy) + F2 (x + y)
Ex.5. Solve q2r – 2pqs + p2t = 0 by Monge’s method.
Sol. Monge’s subsidiary equations in this case will be
q2 dpdy + p2dqdx=0 .....(29)
and q 2 dy 2  2 pq dx dy  p 2 dx 2  0 .....(30)
Factors of equation (30) are
 qdy  pdx 2  0
or qdy  pdx  0
which on integration gives (after putting in dz = pdx + qdy)
dz = 0  z  c1  constant 
55
 Now substituting qdy = – pdx in (29), we get

qdp   pdx   p 2 dqdx  0

or qdp – pdq = 0 [dx = 0 will give the trivial solution]
On integration, we get
p
b  constant 
q
Therefore the intermediate integral is
p
 f  z
q

or p  q f  z  0
For which the Lagrange’s subsidiary equation will be
dx dy dz
 
1  f z 0
from first two terms, we get
y  x f  z  c
and from last two terms, we get
z = c1
Hence the required solution is
y  x f  z  F  z as c = F (z)
Ex.6. Solve t – r sec4y = 2q tan y by Monge’s method.
Sol. Monge’s subsidiary equations in this case will be
 sec4 ydpdy  dqdx  2q tan y dxdy  0 .....(31)

and  sec4 y dy 2  dx 2  0 .....(32)

Factors of equation (32) are
dx – sec2y dy = 0, .....(33)
which on integration gives x – tan y = constant
and dx + sec2y dy = 0 .....(34)
which on integration gives
x + tan y = constant
Now combining (34) with equation (31), we get
sec 2 y dp  dq  2q tan y dy  0
On integration, we get
p  q cos 2 y  constant  f1  x  tan y  .....(35)
Similarly, when (33) is combined with (31) , and integrated gives
p  q cos 2 y  f 2  x  tan y  .....(36)
56
 On solving (35) and (36), we get the values of p and q as
1
p  f1  x  tan y   f 2  x  tan y  
2
1
q  sec 2 y  f1  x  tan y   f 2  x  tan y  
2
Substituting these values in dz = pdx + qdy, we obtain
1 1
dz   f1  x  tan y   f 2  x  tan y   dx   f1  x  tan y   f 2  x  tan y  sec 2 y dy 
2 2

or   
2dz  f1  x  tan y  dx  sec2 y dy  f 2  x  tan y  dx  sec2 y dy 
which on integration gives the required solution as
2 z  F1  x  tan y   F2  x  tan y 

3.5 Monge’s Method for Solving Equation of the Type Rr + Ss + Tt + U (rt – s2) = V

Prof G. Monge gave a method for solving equation

Rr + Ss + Tt + U (rt – s2) = V .....(1)
where R, S, T, U and V are, in general, functions of x, y, z, p and q.
We know that
p p
dp  dx  dy
x y
or dp  rdx  sdy
dp  sdy
or r .....(2)
dx
Similarly
q q
dq  dx  dy
x y
dq  Sdx
therefore t ......(3)
dy
Putting the values of r and t from (2) and (3) in (1), we get
 dp  sdy   dq  sdx   dp  sdy dq  sdx 2 
R   Ss  T   U   s  V
 dx   dy   dx dy 
or (Rdpdy + Tdqdx + Udpdq – Vdxdy) – s (Rdy2 – Sdxdy + Tdx2
+ Udpdx + Udqdy) = 0 .....(4)
Equation (4) will be identically satisfied if we take
Rdpdy + Tdqdx + Udpdq – Vdxdy = 0 .....(5)
and Rdy 2  S dxdy  Tdx 2  Udpdx  Udqdy  0 .....(6)
These simultaneous equations (5) and (6) are known as Monge’s subsidiary equations.
Here the equation (6) can not be factorized. So we will try to factorize
57
  Rdy 2
 S dxdy  Tdx 2  Udpdx  Udqdy  
.....(7)
  Rdpdy  Tdq dx  Udpdq  Vdxdy   0
where is some multiple and is determined later.
Let us suppose that the factors of (7) are

 1  
 Rdy  m1Tdx  m2Udp   dy  dx  dq   0 .....(8)
 m1 m2 
On comparing (7) with (8), we obtain

R R
 m1T    S  V  , m2  m1 , U .....(9)
m1 m2

R
The last two relations gives m1  . Putting this in the first relation of (9), we obtain
U
 2 UV  RT   SU  U 2  0 .....(10)
This equation is called –equation, where , in general, is a function of x, y, z, p and q.
Now since equation (10) is quadratic in so suppose that it is satisfied by two values of say
1 and 2 then the factors corresponding to these values will be

 R1  U U 
 Rdy  U Tdx  R1dp   dy  R dx  R dq   0
  1 
R1
as m1  m2 
U
or Udy  1Tdx  1Udp Udx  1Rdy  1Udq   0 .....(11)
Similarly corresponding to 2, we can obtain
Udy   2Tdx   2Udp Udx   2 Rdy   2Udq   0 .....(12)
Now one factor from (11) and one from (12) will be combined in pairs to get intermediate inte-
grals in the form u = f (v). We can combine factors as
Udy  1Tdx  1 Udp  0
Udx   2 Rdy   2 Udp  0
and Udx  1Rdy  1 Udp  0
Udy   2Tdx   2 Udp  0
These two pairs will give intermediate integrals provided these total differential equations are
integrable, from which the values of p and q can be determined. Substituting these values of p and q in
dz = pdx + qdy, we get the general solution on integration.
Ex.1. Solve 3r + 4s + t + (rt – s2) = 1
Sol. Comparing the given equation with Rr + Ss + Tt + U (rt – S2) = V, we have R = 3, S = 4,
T = 1, U = 1, V = 1. Then – quadratic equation
58
  2 UV  RT   SU  U 2  0
becomes 4 2  4  1  0
1
or  2  12  0  1   2  
2
Hence there is only one intermediate integral given by the equations
Udy  1Tdx  1Udp  0
and Udx   2 Rdy   2Udq  0
On putting above values, we get
 1  1
dy     dx     dp  0
 2  2
 1  1
and dx     3dy     dq  0
 2  2
or 2dy  dx  dp  0
and 3dy  2dx  dq  0
On integration, we obtain
2 y  x  p  c1 .....(13)
and 3 y  2 x  q  c2 .....(14)
Hence the only intermediate integral is
2 y  x  p  f  3 y  2 x  q 
where f is any arbitrary function
Now solving (13) and (14) for p and q, we get
p  2 y  x  c1
q  3 y  2 x  c2
Putting these values of p and q in dz = pdx + qdy, we get
dz   2 y  x  c1  dx   3 y  2 x  c2  dy

or dz  2  ydx  xdy   xdx  3 ydy  c1dx  c2 dy

On integrating, we obtain the general solution as
1 2 3 2
z  2 xy  x  y  c1 x  c2 y  c3
2 2
where c1, c2, c3 are arbitrary constants.
Ex.2. Solve 2s + (rt – s2) = 1
Sol. Comparing the given equation with Rr + Ss + Tt + U (rt – S2) = V, we have R = 0, S = 2,
T = 0, U = 1, V = 1.
Then the quadratic equation
 2 UV  RT   SU  U 2  0
becomes  2  2  1  0
59
 giving 1   2  1
Hence there is only one intermediate integral given by the equations
Udy  1Tdx  1Udp  0
and Udx   2 Rdy   2Udq  0
On putting above values, we get
 dy  dp   0 and dx  dq  0
Integrating, we obtain
y – p = c1 and x – q = c2 .....(15)
Hence the only intermediate integral is
 x  q  f  y  p
where f is any arbitrary function.
Now putting the values of p and q from (15) in dz = pdx + qdy, we get
dz   y  c1  dx   x  c2  dy

or dz   y dx  x dy   c1dx  c2 dy
On integrating, we get the general solution as
z  xy  c1 x  c2 y  c3 .
Ex.3. Solve 2r + (p + x) S + yt + y (rt – s2) + q = 0
Sol. Comparing the given equation with standard equation we have R = q, S = (p + x), T = y, U
= y and V = – q. Then equation
 2 UV  RT   SU  U 2  0

becomes  2   yq  yq     p  x  y  y 2  0

Which gives  y 
1      0 and  2  
 p  x 
Hence the two intermediate integrals are given as
y2 y2
y dy  dx  dp  0
 p  x  p  x
 1 
and 0  q dy  y dq  0  as  0
 2 
which gives
 px
   c1 and qy  c2
 y 
Hence the intermediate integral will be given by
 px
qy  f   .....(16)
 y 
Similarly, the second intermediate integral obtained as
p  x  c3 .....(17)
60
 Substituting the values of p and q from (16) and (17) in dz = pdx + qdy, we get

1  px
dz   c3  x  dx  f  dy
y  y 

1  c3 
or dz   c3  x  dx  f   dy
y  y
On integration, we get the general solution as
1 c 
z  c3 x  x 2  F  3   G  c3 
2  y
Ex.4. Solve (rt – s2) – s (sinx + sin y) = sinx siny
Sol. Comparing the giving equation with standard equation we have R = 0, S = – (sinx + siny),
T = 0, U = 1, and V = sinx siny. Then -equation is
 2 UV  RT   U  U 2  0

becomes  2  sin x sin y     sin x  sin y   1  0

which gives 1  cosce x and  2  cosce y

The first intermediate integral is given by
sin x dy + dp = 0, siny dx + dq = 0
which are not integrable.The other intermediate integrable is given by
sin ydy + dp = 0, sin xdx + dq = 0
On integration, we get
p – cos y = c1 and q – cos x = c2
Hence the intermediate integral will be given by
 p  cos y   f  q  cos x 
This can not be integrated further unless we know f. Therefore, let us suppose that the arbitrary
function f is linear, i.e.,
 p  cos y     q  cos x    .....(18)
where and are constants.
Lagrange’s subsidiary equations for (18) will be
dx dy dz
 
1  cos y   cos x  
From first two terms, we get
y  x  c3
and from the first and last term, we obtain
dz   cos  c3  x    cos x    dx
On integration, we get the general solution as
z  sin y   2 sin x  x  c4
61
 Ex.5. Solve z (1 + q2) r – 2pqzs (1 + p2) t + z2 (rt – s2) + 1 + p2 + q2 = 0
Sol. Comparing the given equation with the standard equation, we have R = z (1 + q2),


2 2
 2 2
 
S = – 2pqz, T  1  p ,U  z and V  1  p  q . Then the -equation is

 2 p 2 q 2  2pqz  z 2  0

or  pq  z 2  0
z
which gives 
pq
Putting the value of in
Udy  T dx  U dp  0
and U dx  R dy  U dq  0
we get
and pqdy + (1 + p2) dx + zdp = 0 .....(19)
pqdx + (1 + q2) dy + zdp = 0 .....(20)
dz  p dx  q dy .....(21)
Combining (19) and (21), and on integration, we obtain
x + zp = c1 .....(22)
Similarly by combining (20) and (21), and on integration, we obtain
y + zq = c2 .....(23)
Putting the values of p and q obtained from (22) and (23) in dz = p dx + qdy, we get
c x  c2  y 
dz   1  dx    dy
 z   z 
2 2
Integrating z 2   c1  x    y  c2   c3
which is the required solution.
Ex.6. Solve 5r + 6s + 3t + 2 (rt – s2) + 3 = 0
Sol. Comparing the given equation with the standard equation, we have R = 5, S = 6, U = 2,
and V = – 3. Then the equation will be
9 2  12  4  0
or  3  2 2  0
2
which gives 1   2  
3
There is only one intermediate integral given by the equations
 2  2
2dy      3 dx      2dp  0
 3  3
 2  2
and 2dx      5 dy      2dq  0
 3  3
or 3 dy  3 dx  2dp  0
62
 and 3 dx  5 dy  2dq  0
Integrating,we get 3y – 3x – 2p = c1 .....(24)
and 3x – 5y – 2q = c2 .....(25)
Hence the only intermediate integral is
3y – 3x – 2p = f (3x – 5y – 2q) .....(26)
where f is an arbitrary function.
Solving (24) and (25) for p and q, we get
1 1
p 3 y  3 x  c1  and q  3x  5 y  c2 
2 2
Putting p and q in dz = p dx + q dy, we get
1 1
dz   3 y  3 x  c1  dx   3x  5 y  c2  dy
2 2
or 2dz = 3 (ydx + xdy) – 3xdx – 5ydy – c1dx – c2dy
Integrating, we get
3 5
2 z  3 xy    x 2    y 2  c1 x  c2 y  c3
2 2
which is the required solution. c1, c2 and c3 are arbitrary constants.

Self-Learning Exercise-II

1. For p.d.e. R r + Ss + Tt = V1 the Monge’s subsidiary equations are ................ and ..... .... ..
.
2. The Monge’s subsidiary equations for p.d.e. r = kt are .... ..... and ..... ..... .
3. The -equation in Monge’s method for solving p.d.e. r + 3s + t + (rt – s2) = 1 is .... .... ....

3.6 Summary

In this unit, you learn about partial differential equations of second order and their solution. You
also studied the solution of two types of P.D.E. by Monge’s method.

3.7 Answers of Self–Learning Exercise

Exercise-I
1. F  x, y , z, p, q, r , s, t   0
2. Rr  Ss  Tt  Pp  Qq  Zz  F
3. z  x 3  x f  y     y 

Exercise-II

1. R dpdy  T dq dx  V dy dx  0
R dy 2  S dy dx  Tdx 2  0
63
 2. dp dy  R dq dx  0 and dy 2  r dx 2  0
3. 2 2  3  1  0

3.8 Exercise-II

Solve the following P.D.E by Monge’s method :

1. pt – qs = q3 [Ans. y = xz + f (z) + F (x)]
2. y2r – 2ys + t = p + 6y [Ans. z = y – y f (y2 + 2x) + F (y2 +2x)]
3

3. 2 x 2 r  5 xys  2 y 2t  2  px  qy   0 [Ans. z = f (x2 y) + F (xy2)]

2 2
4. 1  q  r  2 1  p  q  pq  s  1  p  t  0 [Ans. y = f (x + y + z) + x F (x + y + z)]
5. (q + 1) s = (p + 1) t [Ans. z = f(x) +(x + y + z) or y – (x + y + z = (x)]
6. r – t cos2 x + p tan x = 0 [Ans. 2z = f (y + sin x) – F (y – sin x)]
2 2
7. s – rt = a [Ans. z = x f1 (q – ax) + qy + (q – ax)]
8. ar + bs + ct + e (rt – s2) = b, where a, b, c, e, and h are constants
 x 2 ay 2 
 Ans. ez  xf1  ay  eq  m2 x     y  ay  eq   constant 
 2 2 

9. 2 pr + 2qt – 4pq (rt – s2) = 1  Ans. 3z   2  c  x 3 2  2  c  y 3 2  c 

 1 2 3

Solve the following partial differential equations :

10. 2r + tex – (rt – s2) = 2ex [Ans. z = ex + bx + y2 – ay + c]
 x2 3y 2 
2
11. 3r + s + t + (rt – s ) + 9 = 0  Ans. z  cy  2 xy    f  c  5x   F  c 
 2 2 

 1 2 
12. r + 3s + t + (rt – s2) = 1  Ans. z   2  x  y   F1     F2     f 2      f1    
 
 1 
13. (rt – s2) + 3s = 2  Ans. x  2      ; y  f     g    ; z   xy            y
 
 1 
14. qxr+(x + y)s + pyt + xy(rt – s2)=1 – pq
m
 Ans. z  m y  mx  n log x  f x y 
 
 



64
Unit 4 : Classification of Linear PDE of Second Order,
Cauchy Problem and Method of Separation
of Variables
Structure of the Unit

4.0 Objective

4.1 Introduction

4.2 Classification of PDE of Second Order

4.3 Classification of Second Order PDE in More Than Two Independent Variables

4.4 Cauchy Problem

4.5 Method of Separation of Variables

4.6 Summary

4.7 Answers to Self-Learning Exercises

4.8 Exercise

4.0 Objective

Partial differential equations generally occur in the problems of physics and engineering. After
studying this unit, you should be able to identify and classify partial differential equations (PDE). You will
have an idea of Cauchy problem. At last you will get knowledge of how to solve the partial differential
equations by method of separation of variables.

4.1 Introduction

The importance of partial differential equations among the topis of applied mathematics has been
recognized for many years. However, the increasing complexity of today’s technology is demanding of the
mathematician, the engineer and the scientists, an understanding of the subject previously attained only by
specialists. This unit of partial differential equations (PDE) comprises identification and classification of
PDE. It also presents the principal technique method of separation of variables for constructing solution to
partial differential equation problems. The solved and supplementary problems have the vital role of apply-
ing reinforcing and sometimes expanding the theoretical concepts.

4.2 Classification of PDE of Second Order

Consider the second order partial differential equation

Rr + Ss + Tt + f (x, y, z, p, q) = 0 ....(1)

65
where R, S and T are continuous functions of x and y only possessing continuous partial derivatives. The
PDE can be classified into three categories depending on nature of values of the discriminant S2 – 4RT.
Thus (1) is known as
(i) Hyperbolic if S2 – 4RT > 0
(ii) Parabolic if S2 – 4RT = 0
(iii) Elliptic if S2 – 4RT < 0

2 z 2 z
Ex. 1 : Consider the one dimensional Laplace’s equation 2  = 0 i.e. r + t = 0. Compar--
x y 2
ing it with equation (1), we have R = 1, S = T = 0. Hence S2 – 4RT = 0 and so given equation is parabolic.

2 z z
Ex. 2 : Consider the one dimensional diffusion equation  i.e. r – q = 0. Comparing it
2 y
x
2
with equation (1), we have R = 1, S = 0 and T = – 1. Hence S – 4RT = 4 > 0 and so given equation is
hyperbolic.

4.3 Classification of a Second Order PDE in More Than Two Independent Variables

A linear second order partial differential equation having more than two independent variables can
suitably be reduced, in general, to a canonical form only when the coefficients are constants. Let x, x2,...,
xn be n independent variables and u be the dependent variable, then such a second order PDE may be
written as
n n n
 2u u
  aij xi x j
  bi
x
 cu  0 ....(1)
i 1 j 1 j 1 i

where aij, bi and c are constants and aij = aji. Now we consider a one-one transformation
i = i (x1, x2, ....., xn), i = 1, 2, ....., n ....(2)
Then the equation (1) transforms to
n n
  A kl u   k l
 
 F 1 ,  2 ,....,  n ;u , u 1 , u  2 ,..., u n  0 .....(3)
k 1 l 1

where Akl  aij   k  x  i  x ...(4)

i j

The characteristic quadratic Q() associated with equation (1) in this case is
n n
Q() =   aij  i j ...(5)
i 1 j 1
The associated real symmetric matrix in this case will be

 a a ..... a 
 11 12 1n 
 a a22 ..... a2 n 
M   21     ....(6)
 
 an 1 an 2 ..... ann 

66
and the characteristic roots “eigenvalues” will be given by
 | M –  | = 0 ...(7)
and their nature and signs will determine the type of the given PDE.
Case I : Elliptic PDE : If all the eigenvalues are nonzero and of the same sign, the given PDE is
of elliptic type.
Case II : Hyperbolic PDE : If all the eigenvalues are nonzero and have the same sign except
precisely one of them, the given PDE is of hyperbolic type.
Case III : Ultra Hyperbolic PDE (n 4) : If all the eigenvalues are nonzero and atleast two of
them are positive and two negative then the given PDE is of ultra hyperbolic type.
Case IV : Parabolic PDE : If any of the eigenvalues is zero, the given PDE is of parabolic type.
Note : As an alternative of finding the eigenvalues of matrix M, which sometimes may be cumber
-some, the classification can be made with the help of by expressing the quadratic form (5) as a sum of
squares. The number of positive and negative squares will be the same as the number of positive and
negative eigenvalues of the associated matrix. Either of the methods, as per convenience, may be chosen
for the classification of partial differential equation.
Ex. 1. Determine the nature of following PDE
2 z 2 2z
=x
x 2 y 2

2 z 2 2 z
Sol. x 0
x 2 y 2
Comparing with standard second order PDE, we have
R = 1, S = 0, T = – x2
S2 – 4RT = 0 – 4 (–x2) = 4x2
Since x2 > 0, therefore given PDE is hyperbolic.
Ex. 2. Classify the following PDE as hyperbolic, parabolic or elliptic :
2z
2z 2z
2  0
x 2 x y y 2
Sol. On comparing it with equation (1), we have
R = 1, S = 2, T = 1
Hence the value of discriminant S2 – 4RT = 0
Therefore given PDE is parabolic in nature.
Ex. 3. Find the nature of following PDE
2z 2z 2z z
3 2 2 5 2  x =0
x x y y y
Sol. On comparing given equation with standard PDE, we have
R = 3, S = 2, T = 5
So S2 – 4RT = 1 – 15 = – 14 < 0
then given PDE is elliptic in nature.
67
 2 z 2 z 2
2  z
Ex. 4. Show that the equation 2
x
 2 x
x 2 y
 1  y
y 2
0 
is elliptic for values of x and in the region x2 + y2 < 1, parabolic on the boundary
and hyperbolic outside this region.
Sol. Given equation is

2 z 2 z 2
2  z
x 2
+ 2 x
xy
+ 1 y
y 2
=0

Obviously R = 1, S = 2x, T = 1 – y2
Now discriminant is
S2 – 4RT = 4x2 – 4(1 – y2) = 4(x2 + y2 – 1)
Given equation is elliptic in nature if
S2 – 4RT < 0
or 4(x2 + y2 – 1) < 0  x2 + y2 < 1 (inside boundary)
Given equation is parabolic in nature if
S2 – 4RT = 0
or 4(x2 + y2 – 1) = 0  x2 + y2 = 1 (on boundary)
Given equation is hyperbolic in nature if
S2 – 4RT >0
or 4(x2 + y2 – 1) > 0  x2 + y2 > 1 (outside the boundary)
Ex. 5. Classify the following differential equation as to type in the second quadrant
of xy-plane
 2u  2u  2u
y2  x2 2  2  x  y   y2  x2 2  0
x x y y

Sol. : Here R y 2  x2 , S  2  x  y  , T  y 2  x2
Now S2– 4RT = 4(x – y)2 – 4(x2 – y2)
= 4(x2 + y2 – 2xy – y2 – x2)
= – 8xy
In second quadrant, y is positive while x is – negative,therefore
S2 – 4RT = +ve > 0
Hence given PDE is hyperbolic in nature.
Ex. 6. Classify the equations :

 2u  2u  2u  2u  2u
(a)  2   2  2
x 2 y 2 z 2 x y yz

 2u 2 u  2u 1  2u
(b) + + =
x 2 y 2 z 2 c 2 t 2
68
 Sol. (a) The given PDE can be written as
 2u  2u  2u  2u  2u
 2   2  2 0
x 2 y 2 z 2 xy yz
Here a11 = 1, a22 = 2, a33 = 1,
a12 = a21 = –1, a23 = a32 = –1, a13 = a31 = 0,
therefore the quadratic form
Q () = aij i j
becomes Q    12  2 22   32  21 2  2 2 3
2 2 2
 1   2    2  3    0 
here the two shares are positive and one is zero therefore the given PDE is of parabolic type.
Aliter : The associated matrix is
1 1 0 
M  1 2 1 
 0 1 1 
The eigenvalues of the matrix are given by
|M – I | = 0
(1 – ) ( – 3) = 0 i.e.= 0, = 1,  = 3
Since one of the eigenvalues is zero, the given PDE is a parabolic type
(b) The given equation can be written as
 2u 2u  2u 1  2u
   0
x 2 y 2 z 2 c 2 t 2
Here a11=1, a22 = 1, a33 = 1, a44 =  1 and aij = aji = 0, i  j
Hence the quadratic form c2
Q () =aij i j
2
1 
becomes Q    12   22  32
  4 
c 
This shows that the three shares are positive and only one is negative and therefore the given PDE
is of hyperbolic type.
Ex. 7. Classify the equations
 2u  2u  2u  2u  2u  2u
 3  84  28  16  2 =0
x 2 y 2 z 2 y z z x x y
Sol. Here, a11 = 1, a22 = 3, a33 = 84
a12 = a22 = 1, a23 = a32 = 14, a31 = a13 = 8.
The associated matrix is
1 1 8 
M  1 3 14 
8 14 84 
69
 The eigenvalues of the matrix are given by
| M – I | = 0
  – 98+ 78 – 4 = 0
By Descarte’s rule of signs, The given equation has all the three positive roots and therefore the
given PDE is of elliptic type.
Aliter : The quadratic form
Q () = aij i j

becomes Q() = 12  3 22  84 32  21 2  161 3  28 2 3

2
2
 1   2  83    2  2  3 3 
2
 2 3 
Here all the three squares are positive the given PDE is of elliptic type.

Self -Learning Exercise–1

1. Mark the correct alternative :
(i) The second order PDE Rr +Ss +Tt +f (x, y, z, p, q) = 0 is parabolic if
(a) S2 – 4RT > 0 (b) S2 – 4RT = 0 (c) S2 – 4RT < 0 (d) none of these

 2u  2u  2u
(ii)The PDE + + = 0 is
x 2 x y y 2
(a) hyperbolic (b) parabolic (c) elliptic (d) none of these

 2u  2u 2u
(iii) In the region x2 > 4y the PDE + x + y = 0 is
x 2 xy y 2
(a) hyperbolic (b) parabolic (c) elliptic (d) none of these

2u  2u 2u
(iv) The differential equation 4  16 + 9 = 0 is
x 2 x y y 2
(a) hyperbolic (b) parabolic (c) elliptic (d) none of these
2. Write the condition under which a second order PDE in more than two independent variables is
elliptic.
3. The region in which the equation (x log y) r + 4yt = 0 is hyperbolic is...

 2u  2u  2u
4. Classify the following PDE 4 + 4 + =0
x 2 xy y 2

 2u
 2u  2u
5. Classify the PDE 5 2  9 +4 2 =0
x xy y

 2u
 2u  2u u u
6. Classify the PDE 2 + t +x
2
+2 + + 6u = 0
t x t x t x

70
4.4 Cauchy Problem

The Cauchy problem is a boundary value problem dealing with the unique solution of a second
order quasi–linear PDE when its initial value and slope are specified.
Statement : Determine the solution z = z(x, y) of the PDE
Rr + Ss + Tt + f(x, y, z, p, q) = 0 ...(1)
where R, S and T are in general functions of x, y, z, p and q such that the solution takes on a given
space curve C, having the parametric equation
x = x(t), y = y(t), z = z(t) ...(2)
z
prescribed value of z and , where n is the distance measured along the normal to the curve.
n
The latter set of boundary conditions is equivalent to assuming that the values of x,y, z, p, q are
determined on the curve, but it be noted that the values of p and q can not be assigned arbitrarily along the
curve.
Method of solution : The solution of eq. (1) will be some surface, called integral surface,
passing through C. Hence at each point of C, by relations (2) we have
z0  px  qy0 .....(3)
which shows that p0 and q0 are not independent.
Thus, the Cauchy problem finds the solution of (1) passing through the integral strip of the first
order formed by the planar elements (x0,y0, z0, p0, q0) of the curve C. At every point of the integral strip
p0 = p0(t), q0 = q0(t), so that of we differentiate these equation w.r.t. ‘t’ we find that
p 0  rx0  sy0 , q0  sx0  ty0 ....(4)
Knowing R, S, T, f, x0 , y0 , p0 , q0 , p 0 , q0 at each point of C, we may regard equations (1) and
(4) as linear simultaneous equations for the unknowns r, s, t at each point of C. Solving by Cramer’s’s rule,
we get
r s t 1
   .....(5)
1  2  3 

S T f R T f
1  y0 0  p 0  2  x0 0  p 0
where ,
x0 y0 q0 0 y0 q0

R S f
3  x0 y0  p 0
.....(6)
0 x0 q0

R S T
  x0 y0 0
.....(7)
0 x0 y0

71
 If  0, we can easily calculate the expressions for second order derivatives r0, s0 and t0 along C.
The third order partial differential coefficient of z can similarly be calculated at every point of C by
differentiating (1) w.r.t. x and y respectively, making use of
r0  zxxx x0  z xxy y0 .....(8)
etc. and solving as above.
Proceeding in this manner, we can calculate partial derivatives of every order of the points of C.
The values of the function z at neighbouring points, can be obtained by using Taylor’s Theorem for func-
tions of two independent variables. Thus the Cauchy problem possesses a solution as long as
  0. In the elliptic case 4RT – S2 > 0, so that   0 always holds and the derivatives, of all orders, of z
are uniquely determined.
If  = 0, then the Cauchy’s method of solution breaks down. This critical case leads to the
condition
Ry 2  Sxy
  Tx 2  0
or Rdy 2  Sdydx  Tdx 2  0 ....(9)
At each point (x,y, 0) of (which is orthogonal projection of the curve C on the plane z = 0) the
eq. (9) would give a pair of directions along which  = 0. These directions are called as characteristics.
Thus curves known as characteristic base curves, may be drawn through every point (x,y, 0) of
the base curve  simply by approximating them by straight line segments whose directions are taken to
coincide with those of the tangents given by the roots of (9), viz.

dy S  S 2  4 RT
 .....(10)
dx 2R
Thus a curve  in the xy plane satisfying (10) is called a characteristic base curve of the PDE (1),
and the curve C of which it is the projection is called a characteristics curve of the same equation.
Note that characteristics are :
(i) Real and distinct if S 2 – 4RT > 0
(ii) Coincident if S2 – 4RT = 0 and
(iii) Imaginary if S2 – 4RT < 0
Hence these are two families of characteristics if the given PDE is hyperbolic, one family if it is
parabolic and none if it is elliptic. Thus the Cauchy problem will fail to have unique solution if an arc element
of the base curve  coincides with the characteristics. Consequently, the condition  0 is both necessary
and sufficient to solve the Cauchy problem.7
Characteristic equations :
Corresponding to (1), consider –quadratic
R2 + S + T = 0 ....(11)
2
when S – 4RT  0, eq. (11) has real roots. Then, the ordinary differential equation
dy
   x, y   0
dx
are called the characteristic equations.
Again the solution of (11) will be characteristic curves or simply the characteristic of the second
order PDE (1).
72
4.5 Method of Separation of Variables

For given linear second order partial differential equation

Rr +Ss +Tt +Pp +Qq +Zz = F (x,y) ....(1)
where R, S, T, P, Q, Z and F are functions of independent variables x and y only. Let Z(x,y) be solution of
(1).
The method of separation of variables for this problem is a powerful tool and begins with assump-
tion that Z(x,y) is of the form X(x) . Y(y) i.e.
Z(x,y) = X(x) .Y(y) ....(2)
where X is function of independent variables x only and Y is function of independent variables y only.
On substituting (2) in (1) we have
1 1
f  D  X = g  D  Y ....(3)
X Y
 
where f (D) and g(D) are quadratic functions of D = and D = respectively. This has the effect of
x y
replacing the single PDE with two second order linear ordinary differential equations since LHS of (3) is
function of x alone and the RHS is function of y alone. Since x and y are independent variables, the two
sides of (3) will be equal only if each side is a constant (say ) be
1 1
f  D  X  g  D  Y  
X Y
or f  D  X  X and g  D  Y  Y .....(4)
which can be solved by the methods of ordinary differential equation.
The theory of eigenfunction expansions enters into the treatment of any in homogenous aspect of
the problem. The general solution of equation (4) will depend on the choice of  positive or negative or
zero. In practical problems, the nature of the boundary conditions determine the nature of  and it
becomes an eigenvalue problem.
The method of separation of variables can be employed in a similar manner for more than two
independent variables also.
In the application of ordinary linear differential equation, we first find the general solution and then
determine the arbitrary constant from the initial values, But the same method is not applicable to problem
involving PDE In method of separation of variables right from the beginning we try to find the particular
solution of PDE which satisfy all or some of the boundary conditions and then the remaining conditions are
also satisfied. The combination of these particular solutions gives the solution of the problem.
Ex. 1. Find the characteristics of
y2r – x2t = 0.
Sol. Given y2r – x2t = 0 ....(5)
Comparing (5) with
Rr + Ss + Tt + f (x, y, z, p, q) = 0, we have .....(6)
73
 R = y2, S = 0 and T = –x2
Hence S2 – 4RT = 0 – 4y2(–x2) = 4x2y2 > 0
and thus (1) is hyperbolic everywhere except on the coordinate axes x = 0 and y = 0. The  quadratic is
R2 + S + T = 0 or y22 – x2 = 0 ....(7)
x x
Solving (7), we get   ,  (two district real roots)
y y
Corresponding characteristic equations are
dy x dy x
  0 and  0
dx y dx y
or xdx + ydy = 0 and xdx – ydy = 0
Integrating, we get
x2 + y2 = C1 and x2 – y2 = C1
which are required families of characteristics.
Here these are families of circles and hyperbolas respectively.
Ex. 2. Find the characteristics of
x2r + 2xys + y2t = 0 .....(8)
Sol. Comparing (8) with (6) we have
R = x2, S = 2xy and T = y2
Hence S2 – 4RT = 0
and hence (3) is parabolic everywhere. The  quadratic is
2x2 + 2xy + y2 = 0
Solving it we get
y y
 x  y  2  0 or    ,  (two equal roots)
x x
The characteristic equations is
dy x 1 1
  0 or dy  dx  0
dx y y x
Integrating, we get
y
 c and y  c1x ....(9)
x 1
which is the required family of characteristics. (9) represents a family of straight lines passing through the
origin.
Ex. 3. Solve the followings P.D.E.
2z z
 , 0  x  , y  0
x 2 y
satisfying the boundary conditions
(i) z = 0 when x = 0
(ii) z = 0 when x = 
(iii) z = sin 3x when y = 0
74
 Sol. Let z(x, y) be solution of given PDE Assume that
z(x, y) = X(x)Y(y)
where X and Y be function of only x and y respectively.
On substituting the value of z(x, y) = X(x)Y(y) in given PDE, we have
1 d 2 X 1 dY
   n 2  say 
X dx 2 Y dy
d2X dY
then  n2 X  0,  n2Y  0
dx 2 dy
2
Hence X  a sin  nx    , and y  be  n y

where a, b,  are arbitrary constants

2
Thus z = X(x)Y(y) = A sin  nx   e n y , A = ab .....(10)
According to conditions (i) and (ii) given with the problem, from (10), we get
2 n 2
0  A sin  e  n y and 0  A  1 sin  e  n y . Thus = 0 as A  0
2
Hence z  A sin nx e  n y .....(11)
Also by condition (iii), from (11), we get
sin 3x = A sin nx  A= 1, n = 3
Hence z  sin 3xe 9 y
be required solution of given PDE under specified boundary conditions.
Ex. 4. Use the method of separation of variables to solve the equation
u u
 2  u given that u  x , 0   6e 3 x
x t
Sol. Let u(x, t) = X(x)T(t) be solution of given PDE where X is a function of x only and T is a
function of t only.
u dX u dT
Now T and X
x dx t dt
On substituting these values in given PDE, we get
dX dT
T  2X  XT
dx dt
Dividing by XT, we have
X  2T 
  1   n2 (say)
X T
Now we have two ordinary differential equations.
X 2T 
  n 2 and  1  n2
X T

dX 2 T  n2  1 
or  n X  0, and   
dx T  2 
 
75
 Sloving these equations, we find that
 n 2 1 
2
 t
2 
X  c1e n x and T  c2 e 

 n 2 1 
 n 2 x  t
Hence  2 
u  x, t   X  x  T  t   c1c2 e
2
Under given condition we get 6e3x  c1c2e n x
 c1c2 = 6 and n2 = 3
Thus the required solution of the problem is u(x, t) = 6 e–3x–2t
Ex. 5. Use the method of separation of variables to solve the PDE
 2u u u
2  0
2 x y
x
Sol. : Let u(x, y) be solution of given PDE. For method of separation of variables, we assume
u(x, y) =X(x) Y(y) .....(12)
where X is function of x only and Y is function of y only.

u dX du dY  2u d2X
Now we have Y , X , Y
x dx dt dy x 2 dx 2
On substituting these values in given problem, we get

d2X dX dY
Y  2Y X 0
2 dx dy
dx
On dividing by XY, we have
X  2 X  Y 
  0
X X Y
X   2 X  Y 
or  0
X Y
X   2 X  Y
or     p 2 (say)
X Y
From above equalities, we have two ordinary differential equation.
X   2 X   p 2 X  0 and Y   p 2Y  0
Now consider first differential equation from the above pair of equations i.e.
X– 2X + p2X = 0 .....(13)
Now auxiliary equation for (13) is
m2 – 2m + p2 = 0

2  4  4 p2
m  1  1  p2
2

Therefore CF  c1e
1 1 p x  c2e1 1 p x
2 2

76
 i.e. X  c1e
1 1 p x  c2e1 1 p x
2 2
.....(14)

dY dY
Again  p 2Y   p 2 dy
dy Y

log Y  p 2 y  log c3
2
Y  c3e p y .....(15)
Substituting the values of X and Y from equation (14) and (15) respectively in (12), we get


 1 1P 2 x
u  x, y   X  x  Y  y    c1e

 c2e
 
1 1 p 2 x  2
 c3e p y
 


 1 1 p 2 x
u (x,y)   Ae

 Be

1 1 p 2 x  p 2 y
e

Thus
 
where A = c1c3 and B = c2c3.
Ex. 6. Solve by the method of separation of variables the PDE
u u
4   3u , given that u = 3e–x – e–5x when t = 0
t x
Sol. Let u(x, t) =X(x)T(t) be solution of given PDE where X is a function of x only and T is a
function of only t.
On substituting the value of u(x, t) in the given PDE and dividing by XT, we get
4T  X 
 3
T X
4T  X 
3   p 2 (say)
T X
4T  X
So we have  p 2  3 and   p2
T X

4T  dT  3  p 2 
Now  p2  3    dt
T T  4 

 3  p2 
 log T    t  log c1
 4 
 

or T  c1e
 p 3t
2
4

X  dX
Again  p2    p 2 dx
X X
or log X   p 2 x  log c2
77
 2
or X  c2e p x

Hence u  x , t   X T  c1c 2e
 
 p 2 x  p 2 3 t 4
 bn e
 
 p 2 x  p 2 3 t 4

The general solution is


u  x, t    bn e
 
 p 2 x  p 2 3 t 4

n 1
By the condition given in the problem, for t = 0 we have,
 2
u  x, 0   3e  x  e 5 x   bn e p x
n 1
2 2
So we have, p = 1, b1 = 3 or p = 5, b2 = – 1
Hence the general solution is
u  x, t   3e x t  e5 x 2t
which required solution of given PDE under specified condition.

Self Learning Exercise–II

1. The equation 4r + 5s + t + p + q – 2 = 0
has ......... real characteristic family of curves.
2. For one family of characteristic of PDE
Rr + Ss + Tt + f (x, y, z, p, q) = 0
S2 – 4RT should be ................
3. If S2 – 4RT < 0 for PDE
Rr + Ss + Tt + f (x, y, z, p, q) = 0
then it has ................ real characteristics.
4. If PDE Rr + Ss + Tt + f (x, y, z, p, q) = 0
is hyperbolic the number of real characteristics will be ...............
5. By the method of separation of variables to solve the one dimensional wave equation

2 z 1  2 z , z (x, t) = ..................

x 2 c 2 t 2

4.6 Summary

In this unit, we get an idea and importance of partial differential equation for physical and practical
problems. We have learnt how we can classify the nature of different equations. Cauchy problem is physi-
cal roblem arise in analysis of physical and mathematical problem . A very powerful tool ‘The method of
separation of variables’ is also introduced in this unit. At last for concrete depth in PDE, we have included
the self- learning exercises, illustrative Ex.s and questions for practice.
78
4.7 Answers to Self-Learning Exercises

Exercise – I
1. (i) b (ii) c (iii) a (iv) a
2. See § 4 3. xy log y < 0
4. Parabolic 5. Hyperbolic
6. Elliptic if t2 – 4x < 0,Hyperbolic if t2 – 4x > 0 and Parabolic if t2 – 4x = 0
Exercise – II
1. 2 2. S2 – 4RT = 0 3. Zero
4. 2 5. X (x) T (t)

4.8 Exercise

1. Find the characteristics of

x
(i) 4r + 5s + t + p + q – 2 = 0 [Ans. y  x  c1 , and y   c2 ]
y
(ii) (sin2 x)r + (2cos x)s – t = 0 [Ans. y + cosec x – cot x = c1 ; y + cosec x + cot x = c2]
2. Show that the equation uxx + xuyy + uy = 0 is elliptic for x > 0 and hyperbolic for x < 0.
3. Find whther the following PDE are parabolic or elliptic

 2u  2u
(i) x 2  u  0 [Ans. Hyperbolic if x > 0, parabolic if x = 0 and elliptic if x < 0 ]
t 2 x 2

 2u  2u  2u u
(ii) t 2   0
t 2 x t x 2 x
[Ans. tx < 1 for hyperbolic, tx > 1 for elliptic and tx = 1 for parabolic]
4. Solve by the method of separation of variables :
u u
4 ; u  0, y   8e 3 y [Ans. u (x, y) = 8e–3y-12x]
x y
5. Solve by the method of separation of variables :
u u
  2u ; u  x, 0   10e x  6e4 x [Ans. u (x, t) = 10e–(x+3t) – 6e–2(2x+3t)]
t x
 kx

6. Solve 2uxx – uy = 0 by separation of variables. [Ans. u  x, y    Ae  Be kx  e 2 ky ]
 
7. Solve the following PDE by the method of separation of variables,
(i) 4ux + uy = 3u and u (0, y) = e–5y [Ans. u (x, y) = e2x – 5y]

 2u u 
 p2 t 2 ]
(ii) 2 ; u (x, 0) = x (a – x) [Ans. u (x, y) = x (a – x) e
2 t
x

(iii) y3ux + x2uy = 0 [Ans. u  x, y   ce   

k x3 3  y 4 4  ]
79
 8. Use the method of separation of variables to solve the equation,

 2u u
2
  2u [Ans. u  x, y    A cos px  B sin px  e  p 2 2  y ]
x y
9. Solve the method of separation of variables,
u u
3  2  0 ; u  x, 0   4e  x [Ans. u  x, y   4e  x 3 2  y ]
x y
10. Solve by method of separation of variables,
u u
4   3u ; u  0, y   4e  y  e 5 y [Ans. u  x, y   4e x  y  e 2 x5 y ]
x y
11. Solve by method of separation of variables,
u u
 2  u = 0 when u (x, 0) = 6e–3x [Ans. u  x, y   6e 3 x 2t ]
x t

  

80
Unit 5 : Laplace, Wave and Diffusion Equations And
Canonical Forms
Structure of the Unit
5.0 Objective
5.1 Introduction
5.2 Laplace, Wave and Diffusion Equations
5.2.1 Laplace Equations
5.2.2 Wave Equations
5.2.3 Diffusion Equations
5.3 Canonical Forms
5.4 Summary
5.5 Answers to Self-Learning Exercises
5.6 Exercise

5.0 Objective

After studying this unit, you should be able to know application of partial differential equations. You
will get an idea of wave, diffusion and Laplace equations in different coordinate system and their solutions.
You will also study the reduction of the second order P.D.E’s to canonical forms.

5.1 Introduction

In physical and engineering application, PDE’s of second order are of utmost significance. These
equations arise in the modelling of vibration of string and membranes, theory of hydraulics, gravitational
and potential problems and so on. Since a comprehensive treatment of the subject is not possible in this
unit, we restrict our study to a consideration of some special types of equations.

5.2 Laplace, Wave and Diffusion Equations

In applied mathematics and theoretical physics three types of equations occur frequently. These
are
(i) Laplace Equation
(ii) Wave Equation and
(iii) Diffusion Equation.
In many practical problems the solution of these equations may be obtained with the help of sepa-
ration of variables.
81
5.2.1 Laplace Equation

One of the most important PDE appearing in theoretical physics is Laplace’s equation. It is
usually written as
2u = 0 .....(1)
2
where the operator  , known as Laplacian depends on the coordinate system chosen. It is an elliptic
PDE.
(i) in three dimensions, this equation in Cartesian system of coordinates (x, y, z) is written as

 2u  2u  2u
  0 .....(2)
x 2 y 2 z 2
(ii) in cylindrical polar coordinates (r, , z), eq. (2) becomes

 2u 1 u 1  2u  2u
   0 .....(3)
r 2 r r r 2 2 z 2
(iii) in antisymmetric case i.e. u is independent of , therefore equation (3) reduces to

 2u 1 u  2u
  0 .....(4)
r 2 r r z 2
(iv) in spherical polar coordinates (r, , ), eq. (2) reduces to

 2u 2 u 1  2u cot  u 1  2u
  2 2 2  2 2 0 .....(5)
r 2 r r r  r  r sin  2

(v) when u is independent of the azimuthal angle , (5) reduces to

 2u 2 u 1  2u cot  u
   2 0
r 2 r r r 2 2 r 

  2 u  1   u 
or r   sin    0 .....(6)
r  r  sin     

(vi) in two dimensions, Laplace equation is

 2u  2u
 0 .....(7)
x 2 y 2
in Cartesian coordinates (x, y) and

 2u 1 u 1  2u
  0 .....(8)
r 2 r r r 2 2

in polar coordinates (r, ).

Equation (7) is also known as Harmonic equation.
82
5.2.2 Wave Equation

The wave equation is

2 1  2u
 u .....(9)
c 2 t 2
It is hyperbolic PDE.  is a Laplacian operator which depends on the coordinate system cho-
sen.
(i) Three dimensional wave equation (sound waves in space) in Cartesian coordinates is

 2u  2u  2u 1  2u
   .....(10)
x 2 y 2 z 2 c 2 t 2
(ii) Transverse vibrations of a membrane are governed by two dimensional wave equation

 2u  2u 1  2u
  .....(11)
x 2 y 2 c 2 t 2
(iii) Transverse vibrations of a string are governed by the one dimensional wave equation

 2u 1  2u
 .....(12)
x 2 c 2 t 2

5.2.3 Diffusion Equation or Heat Conduction Equation

The diffusion equation or heat conduction equation in general, is written as

1 u
 2u  .....(13)
k t
where u is interpreted as temperature. It is parabolic PDE.
The one dimensional diffusion equation, which is very much used, may be written as

 2u 1 u
2
 .....(14)
x k t
Ex. 1. Find the most general functions X(x) and T(t), each of one is variable, such
that u(x, y) = XT satisfies the PDE.
 2u 1 u
=
2 k t
x
Also obtain a solution of the above equation for k =1 and which satisfies the boundary
conditions u = 0 when x = 0 or 
u = sin 3x when t = 0 and 0 < x < 
Sol. The given differential equation is
 2u 1 u
= .....(15)
x 2 k t
Let the solution of eq. (15) by method of separation of variables is of the form
u (x, t) = X(x) T(t) .....(16)
83
 Substituting (16) in (15), we get
1 d2X 1 dT
 ....(17)
X dx 2 kT dt
The expression on LHS of eq. (17) is a function of independent variable x while on RHS, it is
function of independent variable t only. Both are equal if both are constant and equal to either – n2, 0 or n2.
Hence three cases arise as follows :

d2X dT
Case I : 0 and 0
2 dt
dx
The solution will be X = Ax + B and T=C

d2X dT
Case II :  n2 x  0 and  n 2 kt
dx 2 dt
2
The solution will be X = Aenx + Be–nx and T  Ce n kt

d2X dT
Case III :  n2 x  0 and   n 2 kt
2 dt
dx
2
The solution is X = A sin (nx + ) T  Be  n ktand
where A, B, C and are arbitrary constants. Since when t , u (x, t)  0, hence case III is most
appropriate solution of eq. (15). Hence
2
u  x, t   Ae  n kt sin  nx   
is the most general solution of given problem
Special case : u (x, t) = 0 when x = 0 or  gives  = 0
Further u (x, t) = sin 3x when t = 0 gives
sin 3x = A sin nx  A =1 and n = 3
Also k=1

 2u u
Hence solution of  is given by
2 t
x
u (x. t) = e–9t sin 3x
Example 2 : Solve the two dimensional heat conduction equation
 2u  2u 1 u
2
 2
 .....(18)
x y k t
by the method of separation of variables.
Sol. : Let the solution of (18) is
u(x, y, t) = X(x) Y(y) T(t) ....(19)
Substituting (19) in (18), we get

1 d 2 X 1 d 2Y 1 dT
  ....(20)
X dx 2 Y dy 2 kT dt
84
 The RHS of (20) is a function of independent variable ‘t’ only whereas LHS is a function of two
independent variables x and y. They are equal if both are constant only. If RHS of (20) is a constant and
sum of two functions of two independent variables then both are constants also. Now three cases arise.

1 d2X 1 d 2Y 1 dT
Case I : 2
 0, 2
 0 and 0
X dx Y dy kT dt
The solution of these relations will give
X = ax + b, Y = cy + d and T = e
where a, b, c, d and e are arbitrary constants.

1 d2X 2 1 d 2Y 2 1 dT
Case II : 2
 m , 2
 n and  p2
X dx Y dy kT dt

d2X d 2Y dT
or 2
 m 2 X  0, 2
 n 2Y  0 and  p 2kT
dx dy dt
where m2 + n2 = p2
On solving these equations, we get
2
X = a1emx + b1e–mx, Y = a2exnx + b2e–nx and T  a3e p kt

1 d2X 2 1 d 2Y 2 1 dT
Case III :   m ,   n and   p2
X dx 2 Y dy 2 kT dt

d2X d 2Y dT
or 2
 m2 X  0, 2
 n 2
Y  0 and   p 2 kT
dx dy dt
where m2 + n2 = p2
Solving these equations, we get

X  c1 cos  mx  cm  , Y  c3 cos  ny  cn  and T  c5e

 
 m 2  n 2 kt

Since u(x, y, t)  0 as t , therefore case III is most appropriate. Hence solution of (18) which
is linear can be written as
 
 
k m2 n2 t
u  x, y , t     cmn cos  mx  cm  cos  ny  cn  e
m1 n1
Ex. 3. A thin rectangular plate whose surface is impervious to heat flows has at
t = 0 an arbitrary function f (x, y). Its four edges x = 0, x = a, y = 0, y = b are kept at zero
temperature. Determine the temperature at a point of the plate as ‘t’ increases.
Sol. Here the temperature U(x, y, t) in the plate is governed by the two dimensional heat equation

 2U  2U 1 U
  ....(21)
x 2 y 2 k t
with boundary conditions
85
 U  0, y , t   0, U  a, y , t   0, U  x, 0, t   0, U  x, b, t   0 .....(22)
and initial condition is
U (x, y, t) = f (x, y) .....(23)
Proceeding similarily to Ex.2, we find that if solution of (21) may be assumed as
U (x, y, 0) = X(x) Y(y)T(t)
then X = c1 cos (mx + cm) = A1cos mx+ B1sin mx,
Y = A2 cos nx + B2 sin nx

and T  A3e

 k m2 n2 t 
Using boundary conditions (22), we find that
A1 = 0, B1 sin ma = 0, A2 = 0, B2 sin nb = 0
 A1 = 0 = A2, sin ma = sin u and sin nb = sin v (u, = 1, 2, 3 ....) as
B1  0 and B2  0
u v
Thus A1 = 0 = A2, m  and n 
a b
Hence the general solution of (21) will be
 u 2 2 
   k  2  2  2t
u x y a b 
U  x, y, t     Fuv sin sin e 
u1 1
a b
Now under initial condition (23), we have
 
u x y
U  x, y, o   f  x, y     Fu sin sin .....(24)
u 1 1 a b
which is a double Fourier series of f (x, y).
a b
4 u x v y
Hence Fuv    f  x, y  sin sin dxdy .....(25)
ab a b
x0 y0

Thus (24) is a general solution of (21) under boundary and initial condition (22) and (23) where
constant Fuv as given by (25).
Ex. 4. By separating the variables, show that the one dimensional wave equation
 2u 1  2u
 .....(26)
x 2 c 2 t 2
has solution of the form A exp   in x  in ct  where A and n are constants.
Sol. Let the solution of (26) is
u(x, t) = X(x) T(t) ....(27)
Substituting (27) in (26), we get

1 d2X 1 d 2T
  n2  say 
X dx 2 2 2
c T dt

86
 d2X d 2T
  n 2 X  0 and  n 2c 2T  0
2 2
dx dt
Solving these we get
X  c1e  in x and T  c2 e in ct .....(28)
Hence from (27) and (28), we get the solution of (26) as
u(x, t) = A exp ( in x  in ct)
Ex. 5. A tightly stretched sting which has fixed end points x = 0 and x = l is initially
in a position given by y = k sin3 (x/l). It is released from rest from this position. Find the
displacement y (x, t).
Sol. Since the string is tightly stretched initially between two fixed points and released from rest, it
will make transverse vibrations in (x, y) plane. The displacement y(x, t) of any point on it will be a gov-
erned by the following wave equation

2 y 1 2 y
 .....(29)
x 2 c 2 t 2

with the boundary conditions

t > 0 : y(0,t) = 0 = y(l, t) .....(30)
and the initial condition
t = 0 : y(x,0) = k sin3(x/l) .....(31)
 y 
which also implies  t   0
 t 0
Applying the method of separation of variables if solution of (29) is of the form X(x)T(t) we
find that
X = A cos x + B sin x
and T = C cos ct + D sin ct ....(32)
Using boundary condition (30), we get
n
A=0 and B sin   0     B  0  n  1, 2,3....

Hence Xn(x) = An sin (n/l) ....(33)

 y 
Under initial condition    0 , we get D = 0 from (32). Therefore
 t t 0
Tn(t) = Bn cos (nct/l) ....(34)
Hence (33) and (34), we get the general solution of (29) as
nx nct
yn  x, t   Cn sin cos ,n N
 
where Cn = An Bn is an arbitrary constant
 
nx nct
Hence y  x, t    yn  x, t    Cn sin cos .....(35)
n1 n1  

87
 To determine the constant Cn we apply the condition (31) on (35), we get

 x   nx 
k sin 3     Cn sin  
   n1   


k  x   3x    nx 
or  3sin    sin      Cn sin  
4        n 1   

3 k
 C1  k , C3   and c2  c4  c5  c6  ....0
4 4
Hence the required solution is
k  x   ct   3x   3ct  
y  x, t    3sin   cos    sin   cos  
4            
Ex. 6. Solve the harmonic equation
 2u  2u
 0 .....(36)
x 2 y 2
satisfying the conditions

 x  
u  x , 0   0, u  x , a   sin   
   .....(37)
u  0, y   u  l , y   0 

Sol. Let the solution of (36) is
u(x, y) = X(x) Y(y)
Substituting in (36) we get

1 d 2 X 1 d 2Y
 0
X dx 2 Y dy 2

1 d2X 1 d 2Y
 2
  2
  2  say 
X dx Y dy

d2X
Now   2 X  0  X  x   A cos x  B sin x
2
dx
Applying u (0, y) = X(0) = 0 and u (l, y)= X(l) = 0
n
we get A = 0 and   n or   ,n  N

 nx 
thus X n  x   Bn sin  
 l 

d 2Y
Again 2
  2Y  0  Y  y   C cos h y  D sin hy
dy
Now u (x, 0) = y(0) = 0 gives C = 0
88
  ny 
thus Yn  y   Dn sin  
 l 
 nx   ny 
Hence we have un  x, y   X n  x  Yn  y   Fn sin   sin h  
 l   l 
where Fn is arbitrary constant. Therefore

 nx   ny 
u  x, y    Fn sin   sin h  
n1  l   l 
Now applying the boundary condition
u (x, a) = sin (x/l)

sin x   nx   na 

We find that   Fn sin   sin h   x
l n 1  l   l 
Equating coefficients of like terms, we get
F1 sinh (a/l) = 1 and F2 = F3 = .... = 0
Hence, the required solution is
 a   x   y 
u  x, y   cosech   sin   sin h  
 l   l   l 

5.3 Canonical Forms

Let us consider the equations of the type

Rr + Ss + Tt + f (x, y, z, p, q) = 0 ....(1)
where R, S, T are continuous functions of x and y possessing continuous partial derivatives of as high as
order as necessary. It is a typical class of semi-linear equations of the type of
Rr + Ss + Tt = V
Changing the independent variables x, y to ,  such that
 =  (x, y), = (x, y) .....(2)
z = z () ....(3)
Here it is assumed that  are doubly differentiable and the transformation from (x, y)–plane
to ( )–plane is locally one to one. This requires that the Jacobian of the transformation is
nonzero, i.e.
  ,      
J   0
  x, y  x y y x
Now from (2) and (3), we get
z z  z       
p     z
x  x  x  x  x  

z z  z       
q     z
y  y  y  y  y  
89
 2 z        z  z 
r 
2
   
x  x  x    x  x  

2 2 2 2 2   2   z   2  z
    z      z     z
  2     2 2   2 2 
 x  2  x x    x  2  x    x  
   

 2 z         z  z       2 z
s             
xy  x  x    y  y     x y  2

        2 z      2 z  2  z  2 z
           
 x y x y    x y  2 x y   x y 

2
2 z
       z  z      2 z
and t  2             
y  y  y    y  y     y  2
2
     2 z     2 z   2   z   2  z
2      2   2  .....(4)
 y y    y    y     y  
2

Now substituting these values in (1), it takes the form

2 z 2 z 2z  z z 
A  2 B  C  F  , , z , , 0 .....(5)
2  2    

2 2
        
where A  R   S T   .....(6)
 x  x y  y 

  1        
BR   S     T  .....(7)
x x 2  x y x y  y y

2 2
        
C  R   S  T   .....(8)
 x  x y  y 

 z z 
and F  , , z , , 
   
is obtained from the transformed form of f (x, y, z, p, q) and the remaining terms containing first order
partial derivatives of transformed Rr, Ss, and Tt.
One of the relations satisfied by A, B, C and R, S, T which can be easily seen, is
1
AC  B 2 
4
 
4 RT  S 2 J 2 .....(9)

We shall now determine the functional relationship [equations (2)] of  with x and y so that the
transformed equation (5) takes the simplest possible form.
90
 The procedure is simple when the discriminant S2 – 4RT of the quadratic equation (called 
equation)
Q () = R2 + S + T = 0 .....(10)
is either positive, negative or zero everywhere. We shall discuss these cases separately. It may be noted
that Q () is called the ‘characteristic quadratic form’ and the discriminant of the quadratic will determine
the nature of P.D.E. This will depend on the characteristic roots of the associated real symmetric metric.
 R S 2
M   .....(11)
S 2 T 
Case I : S2 – 4RT > 0.
In this case the roots  and  of equation (10), which are in general functions of x and y, would
be real and distinct.
 
Let us take  1 .....(12)
x y
 
and  2 ....(13)
x y
then from (6) and (8), we find that
2
  
A  R 12 
 S 1  T    0
 y 
.....(14)

2
  
and C  R 22 
 S2  T    0
 y 
.....(15)

where  and  are roots of (10).

The equation (5) reduces to

2 z   z z 
2B  F  , , z, ,   0 .....(16)
    
Equation (12) is a Lagrange’s linear equation of first order, whose subsidary equations are
dx dy d
 
1 1 0
which gives = constant,
dy
and  1  0 ....(17)
dx
Let f1(x, y) = constant be the solution of equation (17) then the general solution of equation (12)
will be
 = f1(x, y) ....(18)
In a similar manner the general solution of equation (13) will be
 = f2(x, y) .....(19)
where f1 = constant and f2 = constant are the solution of differential equations
dy dy
 1  0 ,  2  0 .....(20)
dx dx
91
respectively. Relations (18) and (19) are the desired transformations for independent variables which
reduce the given equations (1) to the form (16).
Now from (9), we have
2
1      
4

AC  B  4 RT  S 2 
2
  
 x y y x 
.....(21)

This shows that the sign of (AC – B2) is the same as of (4RT – S2) i.e. it is invariant under
transformation.
Therefore, when A = C = 0, from (21), we have
2
     
2

4 B  S  4 RT 2
  
 x y y x 
.....(22)

Since we have assumed that S2 > 4RT, it implies from (22) that B2 > 0 i.e. B  0 and therefore we
may devide both sides of equation (16) by it and write it finally as

2 z  z z 
 1  , , z, ,  .....(23)
    
which is the canonical form of equation (1) when S2 – 4RT > 0.
Case II : S2 – 4RT = 0.
In this case the two roots of the quadratic equation (10) are equal i.e.  =  Therefore one of the
functions, say  will be defined by equation (18) of case I. We may now take to be any suitable function
of x and y which should be independent of  . Therefore, as before, A = 0 but C  0. Further, from (21),
since S2 – 4RT = 0 we have
B=0
Hence equation (5) reduces to

2 z  z z 
C  F  , , z , ,   0
2
    

2 z  z z 
or  2  , , z , , 
2 .....(24)
    
which is the canonical form of the equation (1) when S2 – 4RT = 0.
Case III : S2 – 4RT < 0.
This is particularly the same as case I except that the roots of the quadratic equation (10) in this
case are complex. If we proceed in the same manner as we did in case I, we shall arrive at equation (21)
but in this case the variables are not real and in fact complex conjugates. To get a real cononical form we
transform the independent varialdes  and  again be the following relations.
1 1
     ,   i     ....(25)
2 2
92
 z z  z  1    
then       i  z
     2    

z z  z  1   
      i z
     2    

2 z 1    z z  1  2 z 2 z 
Hence    i   i      .....(26)
 4        4   2  2 
and therefore the relation (23) reduces to

2 z 2 z  z z 
2
 2
 3   ,  , z , ,  ....(27)
     
which is the Canonical form of equation (1) when S2 – 4RT < 0.
Ex. 1. Reduce the equation

2 2z 2
2n  z z
 n  1 2
 y 2
 ny 2 n1
x y y
to canonical form and find its general solution.
Sol. Comparing the given equation with the standard form
Rr + Ss + Tt + f (x, y, z, p, q) = 0, we get
2 z
R   n  1 , S  0, T   y 2n , f  ny 2n 1
y
Here, S2 – 4RT = 4(n – 1)2 y2n > 0 provided n  1.
Hence the given differential equation is hyperbolic differential equation. The roots of the -equa-
tion Rd2 + S + T = 0
or (n–1)22 – y2n = 0

yn  yn
are 1  and  2 
n 1 n 1
Changing the independent variables from x, y to  such that  = f1(x, y),  = f2(x, y) where
f1 = constant and f2 = constant are the solution of the differential equations
dy dy
 1  0 and   2  0 respectively..
dx dx
These gives f1(x, y) = y1–n – x = constant
and f2(x, y) = y1–n + x = constant
Hence  = y1–n – x and = y1–n + x
z z z
Now, p  
x  
z  z z 
q  1  n  y  n   
y    
93
 2 z 2 z 2 z 2 z
r   2 
x 2 2  2

2 z  2 z
2 2 n  2 z 2 z   z z
t   
n  1 y  2  2  2   n  n  1 y  n 1     
2   
y      
Therefore, the given equation reduces to
2
2
 2 z 2 z 
z  2 z
2 2 z 2 z 
 n  1  2  2     n  1  2  2  
   2     2 

 z z   z z 
n  n  1 y n 1     n  n  1 y n1   
       

2 2 z
or 4  n  1 0

2 z
or 0

which is the required canonical form if n 1.
The general solution of the above equation may be easily obtained as
z  1     1  
where  and  arbitrary functions of and respectively. Changing to original variables we get finally

 
z  1 y1n  x  2 y1n  x  
Note : If n = 1, the character of the given differential equation changes. It becomes a parabolic
equation, viz.

2 z z
y 2
 0
y y
whose general solution is
z  1  x  log y  2  x  .
Ex.2. Reduce the equation
2z 2z 2z
2  0
x 2 xy y 2
to canonical form and hence solve it.
Sol. Comparing the given equation with the standard form Rr + Ss + Tt + f (x, y, z, p, q) = 0,
we get, R = 1, S = 2, T = 1, f = 0
Here S2 – 4RT = 4 – 4 = 0
Hence the given equation is a parabolic differential equation.
The roots of the -equation
R2 + S + T = 0
94
 or 2 + 2 + 1 = 0
are = – 1, – 1
Changing the independent variables x, y to ,  where = f1 (x, y), such that fi = const. is the
solution of the differential equation

dy
 1  0
dx

dy
or 1  0 which gives x – y = const.
dx
Hence = x– y
We may now take  to be any suitable function of x and y which should be independent of . Let
=x+y
z z  z  z z
Now,      
x  x  x  

z z  z  z z
     
y  y  y  

2 z     z z   2 z 2 z  2 z
     2 
x 2         2  2

2z     z z  2 z  2 z
      2  2
xy         

2z     z z 
       
2
y       

2 z 2 z  2 z
  2 
 2  2
Therefore the given equation reduces to
2 z
0
2
which is the required canonical form.
The general solution of equation may be easily obtained as

z       2   
where 1 and 2 are arbitrary functions of .
Changing to the original variables, we get finally
z   x  y  1  x  y   2  x  y 

95
  2z 2z
Ex. 3. Reduce  x2 to canonical form.
x 2 y 2
Sol. Comparing the given equation with
Rr + Ss + Tt + f (x, y, z, p, q) = 0,
We get R = 1, S = 0, T = –x2
Now the roots of the -equation
R2 + S + T = 0
or 2 – x2 = 0
are =  x
Changing the independent variables x, y to , where
= f1 (x, y) and = f2 (x, y)
such that f1 = const. and f2 = const. are the solutions of the differential equations.
dy dy
Hence  1  0 and  2  0
dx dx
dy dy
becomes  x  0 and x0
dx dx
x2 x2
Integrating y = const. and y  = const.
2 2
x2 x2
Hence  y and   y 
2 2
z z  z  z z
Now,     x x
x  x  x  
z z  z  z z
     
y  y  y  

2 z     z z  2
2  z  2 z  2 z  z z
     x  2 2   
x 2           2   

2z 2 y 2 z 2z
 2 
y 2 2  2
Therefore the given equation reduces to

2 2z  2 z  2 z  z z 2
2  z 2 z 2 z 
x  2 2    x  2 2  0
   2       2 

2 z 1  z z 
or  2  
 4 x    

2 z 1  z z 
 
or  4         
which is the required canonical form.
96
Ex.4. Reduce the equation

  
xyr  x 2  y 2 s  xyt  py  qx  2 x 2  y 2 
to canonical form and hence solve it.
Sol. Comparing the given equation with standard form Rr + Ss + Tt + f (x, y, z, p, q) = 0, we get
R = xy, S = – (x2 – y2), T = – xy
So -equation R 2 + S + T = 0
becomes xy 2 – (x2 – y2) – xy = 0
y x
or  ,
x y

dy dy
Hence  a1  0 and  a2  0
dx dx

dy y dy x
becomes   0 and  0
dx x dx y

y
Integrating,  c1 , x 2  y 2  c2
x
y
Now, we take  ,   x2  y 2
x

z z  z  y z z
Then      2  2x
x  x  x x  

z z  z  1 z z
      2y
y  y  y x  
2
2 z  y   2 z 2
 y   z
2
2  z 2 y z z
2
 
 2 2
 2   2 
2 x   4 x 2
 3 2
x  x    x    x  

 2 z  y  1   2 z   y  1  2z  2 z 1 z
   2   2   2 y   2   2 x    4 xy 2  2
xy  x  x     x  x    x 
2
2z 1  z
2
1 2z 2
2  z z
2
   2
 2    
2 y  4 y 2
2
y  x   x   

Therefore the given equation reduces to

2 2 y
 x2  y 2  

 y 2  x2 x2 
2 y


y 2  x 2 x2


2  1
or 2 2 .....(28)


x2  y 2 
2  1  
97
 Integrating (28) w.r.t. , we get
z 2  1
 2
d     
 2
 1 
d d
 2
 2 2
   
 1
   1
2

 
d   1 d 
  2      
2  1  2 1 1  2  1 2 1 1   2  1 
 
 

z 
     
 2
 1 
Integrating it, we get


z=   1     2   
   1 2

 y
or z =  xy    x 1
2

 y 2  2  
x
Self Learning Exercise
1. The Harmonic equation is ....

 2u  2u 1 u
2. 2
 2
 is two-dimensional ........... equation.
x y k t
3. Write general Laplace’s equation.
4. Write wave equation.
5. Give a common method for solving Laplace, wave and diffusion equations.

5.4 Summary
In this unit, we have covered nature and types of Laplace, wave and diffusion equations and their
solutions under different boundary and initial conditions, with illustrative examples. We have also presented
the canonical form of PDE and its general solution also for hyperbolic, parabolic and elliptic equations.

5.5 Answers of Self-Learning Exercises

 2u  2u
1.  0 2. Diffusion
x 2 y 2

 2u 1  2u
3.  u  02 4. 
x 2 c 2 t 2
5. Separation of variables
98
5.6 Exercise
1. A string is stretched between the fixed point O (x = 0) and A (x = 1) and released at rest from the
position U(x, 0) = A sin x. Find the formula for its subsequent displacement U(x ,t)
[Ans : U (x, t) = A cos ct cos x]
2. A string is stretched between the fixed points (0, 0) and (l, 0). If it is released at rest from the initial

 2k l
 l x ; 0  x
2
deflection f  x  
 2k  l  x  ; l  x  l
 l 2
where ‘k’ is arbitrary constant. Find the expression of deflection of the string at any instant ‘t’.
8k  sin x cos ct 1 3x 3ct 
[Ans : U  x, t    sin cos  ....
2 
  l l 9 l l 
3. A tightly string stretched string with fixed end points x = 0 and x = is initially at rest in its
equilibrium position. If it is set vibrating by giving to each of its points an initial velocity
 U 
   0.03sin x  0.04 sin 3 x
 t t  0
then find the displacement U (x, t) at any point x and at any instant t.
1
[Ans. U  x, t    0.03sin x sin ct  0.01333sin 3x sin 3ct  ]
c
 y 
4. Solve ytt = 4yxx, y (5, t) = 0 = y (5, t), y (x, 0) = 0 and    f  x   5sin x
 t t 0
5
[Ans. y  x , t   sin x sin 2t ]
2

u  2u
5. Solve diffusion equation  , 0 < x < l, t > 0
t x 2
u(x, 0) = 3 sin n x, u (0, t) = 0, u(l, t) = 0.
 2 2
[Ans. u  x, t   3 e  n  t sin nx ]
n1

6. The temperature distribution in a bar of length which is perfectly insulated at ends x = 0 and x =
is governed by PDE
u  2u

t x 2
Assuming the initial temperature distribution as u(x, 0) = cos 2x. Find the temperature distribution
at any instant of time. [Ans. u(x, t) = e–4t cos2x]
7. A homogeneous rod of conducting material of length 100 cm has its ends kept at zero temperature
and the temperature initially is
99
  x, 0  x  50
u  x, 0   
100  x, 50  x  100
Find the temperature u(x ,t) at any time.
2
  2 n 1c 
  t
[Ans. u  x, t  
400 
 1n sin  2n  1 xe  100 
2
 2 100
]
n  0  2 n  1

8. Solve ut = a2uxx under the conditions ux(0, t ) = 0 = ux(, t), u(x, 0) = x2, 0 < x < , t > 0.
n
3  1
  2 2
[Ans. u  x, t    4  2 cos n xe  a n t ]
3 n 1 n

 2u  2u
9. Solve   0 ; 0  x  , 0  y  
x 2 y 2
which satisfies the conditions u  0, y   u  , y   u  x,    0 and u  x, 0   sin 2 x .

8  sin  2n  1 x sin h  2n  1   y 

[Ans. u  x, t    ]
 n 1  2n  1  2n  12  y  sin h  2n  1 
 

2z z
10. Reduce the equation y   0 to canonical form and find its general solution.
2 y
y
[Ans. z  1  x  log y  2  x  ]
11. Reduce the equation
2 z 2z  2 z y 2 z x 2 z
y2  2 xy  x2 2  
x 2 xy y x x y y

to canonical form and hence solve it.

2
2 2
 
2 2 2
[Ans. z  x  y 1 x  y  2 x  y ]   
2 z 2 2 z
12. Reduce the equation x to canonical form .
x 2 y 2
Also state the nature of the equation.

2 z 1  z z  x2 x2
[Ans.   ;   y  ,   y  , hyperbolic.]
 4         2 2
2 z 2 2z
13. Reduce the equation x  0 to canonical form.
x 2 y 2

 2 z 2 z 1 z x2
Also find its nature. [Ans.   ,   y,   , elliptic]
 2  2 2  2

  

100
Unit 6 : Eigenvalues, Eigenfunctions and Sturm-Liouville
Boundary Value Probleon
Structure of the Unit
6.0 Objective

6.1 Introduction

6.2 Linear Homogeneous Boundary Value Problem

6.3 Eigenvalues and Eigenfunctions

6.3.1 Eigenvalue

6.3.2 Eigenfunction

6.4 Sturm-Liouville Problem

6.5 Orbnogonality of Eigenfunctions

6.6 Important Theorems for Sturm-Liouville System

6.6.1 Theorem 1

6.6.2 Theorem 2

6.6.3 Theorem 3

6.6.4 Theorem 4

6.7 Summary

6.8 Answer to Self-Learning Exercise

6.9 Exercise

6.0 Objective

After completing the present unit, you will get a basic knowledge about eigenvalue and
eigenfunction of boundary value problems. You will study special boundary value problem known as Sturm-
Liouvelle problem and properties of eigenfunctions in later part of unit. The knowledge which you gain
here, can be used to study various special functions that arise in physical and engineering problems.

6.1 Introduction

In the eighteenth century much attention was given to the problem of determing the mathematical
laws governing the notion of a vibrating string with fixed end points. We wish to motivate the physics of
vibrating string. In the last unit, we dealt the wave equation in detail with some other physical problems
where we had derived boundary value problems for seeking non-trivial solution of partial differental equa-
101
tions involved in formulating physical problems. In this unit we study the condition of parameter involved
in boundary value problem and corresponding non-trivial solution. We will also see special boundary
value problem, known as Sturm-Liouville problem in detail which helps in studying regular boundary
value problem and special functions in future.

6.2 Linear Homogeneous Boundary Value Problems

In previous unit, we have noticed that most important application of the idea is in boundary value
problems of any type. For second order linear differential equation, boundary value problem is defined
as
Ly = h .....(1)
where L is a second order linear differential operator defined on a finite interval [a, b] and h is a function
in [a, b] and pair of homogeneous boundary conditions of the form
1 y (a) + 2 y (b) + 3 y(a) + 4 y(b) = 1 .....(2)
1 y (a) + 2 y (b) + 3 y (a) + 4 y (b) = 2 .....(3)
where i, i and i for i = 1, 2 are constants. The problem (1) with boundary conditions (2) and (3) is
known as linear homogeneous boundary value problem. In this problem, we seek all non-trivial func-
tions of y(x) in [a, b] which simultaneously satisfy differential equation (1) and boundary conditions (2)
to (3).
For example, y + y = 0 .....(4)
with boundary conditions
y(0) = 0 and y() = 0 .....(5)
is a boundary value problem of above type on the interval [a, b]. The parameter ‘’ in (4) is free to
assume any real value.
The situation with boundary conditions is quite different from that for initial condition. The initial
value problem is a sophisticated variation of the fundamental theorem of calcalus. The boundary value
problem is rather more subtle.

6.3 Eigenvalues and Eigenfunctions

In previous study, we have considered initial value problem, in which the solution of second or-
der differential equation is sought that satisfies two conditions at a single value of the independent vari-
able. Here we have absolutely different situation for we wish to satsfy one condition at each of two
distinct values of independent variable x. The part of our task is to discover the values of s for which
problem can be solved for getting non-trivial solution. The solution of given problem in (4) with bound-
ary conditions (5) is not difficult to find. We simply apply the boundary conditions to the general solu-
tion. But we have to analyse the solution for all possible values of s. So, three cases arise as follows.
Case I : is negative or < 0
Let  = –m2
The given problem (4) with (5) becomes
y – m2y = 0 .....(1)
102
 and y(0) = 0 and y() = 0
so, the general solution is
y(x) = c1emx + c2e–mx
Now y(0) = 0  c1 + c2 = 0 .....(2)
and y() = 0  c1em + c2e–m = 0 .....(3)
Equations (2) and (3) give
c1sinh m = 0  c1 = 0 as sinh m  0 for m  0
Hence c1 = c2 = 0. Thus we get only one trivial solution exists.
Case II : = 0
The given problem (4) with (5) becomes
y = 0
and y(0) = 0 and y() = 0
Hence the general solution is
y(x) = c1x + c2
When y(0) = 0, we have c2 = 0
So y(x) = c1 x
When y() = 0, we have c1 = 0
Under given boundary conditions, c1 = c2 = 0
i.e. we have trivial solution for given problem for this value of or y  0
Thus, we are restricted to the case in which is postive for seeking non-trivial solution.
Case III : > 0
Let  = m2
The given problem (4) with (5) reduces to
y + m2y = 0 .....(9)
and y(0)= 0 and y() = 0
so, the general solution is
y(x)= c1 sin mx + c2 cos mx
for y(0) = 0, we have c2 = 0
Hence y(x) = c1sin mx
and for y() = 0, 0 = c1 = sin m
Since c1  0 for seeking non-trivial solution, we must have
sin m = 0  sin m = n; for some positive integer
or m = n; n = 1, 2, 3, ....
or m=n
2
Hence n = n ;n = 1, 2, 3, ........ which is known as eigenvalues and corresponding solution is
yn(x) = c1 sin nx; n = 1, 2, 3, ....
which is called as eigenfunction.
6.3.1 Eigenvalue or Characteristic Value
The values s, for which given boundary value problem has non-trivial solutions, are called eigen-
values of given problem.
For example  = 1, 4, 9, ......n2 are eigenvalues of problem (4)
103
 6.3.2 Eigenfunction or Characteristic Function
The non-trivial solution of given boundary value problem corresponding to particular eigenvalues
is termed as eigenfunction.
For example yn(x) = sin x, sin 2x, ......, sin nx, ..... are eigenfunctions for eigenvalues in = 1, 4, 9,
..... n2, ...... respectivoly for problem in (4)
It is to be noted here that the eigenvalues are uniqely determined by the problem but the
eigenfunctions are not. Any non-zero sealar multiple of eigenfunction is also a eigenfunction.
From the above study, we have three important conclusions for eigenvalues and eigenfunctions
as follows
(i) The eigenvalues form an increasing seauence of positive numbers that approaches  i.e.
1 < 2 < 3 ....... n .........
and n  as n 
For example, 1 < 4 < 9 ...... < n2 < ......... in above problem
(ii) The nth eigenfunction vanishes at the end points of the interval and has exactly n – 1 zeros
inside this interval.
For example, for n = n2, yn = sin nx vanisheos at the end points of the interval [0, ] and has
exactly n – 1 zeros inside this interval (0, ) in above problem in (4).
(iii) If yn(x) is an eigenfunction for eigenvalue for given problem, then cyn(x) is also eigenfunction
where c is arbitrary constant for same eigenvalue. Hence the eigenfunction corresponding to each eigen-
value is unique except for a multiple of an arbitrary constant factor.
The problems of heat, wave and Laplace in previous unit or many other physical or applied math-
ematical problems are boundary value problems. In solution procedure by seperation of variables for
any problem, notice that we have calculated eigenvalues and corresponding eigenfunctions also.
Ex.1. Find the eigenvalues s and corresponding eigenfunctions yn(x) for the equa-
tion y + y = 0 under the boundary condition y(0) = 0 and y(/2) = 0
Sol. We have three cases.
Case I : is negatve or < 0
Let  = – m2
The given differential equation becomes
y – m2y = 0
whose general solution is
y(x) = c1emx + c2e–mx
Now y(0) = 0  c1 + c2 = 0
and y(/2) = 0  c1em/2 + c2e–m/2 = 0
The above thwo equations give us
c1 sinh (m2) = 0  c1 = 0 ( m  0)
Thus we get only one trivial solution i.e. y(x) = 0
104
 Case II : when = 0
The given problem reduces to
y = 0
Hence the general solution is
y(x) = c1x + c2
So, under given boundary conditions, c1 = c2 = 0
which gives trivial solution only i.e. y  0 for = 0
Thus  0 are not eigenvalues for given problem.
Case III : when is positive or > 0
Let  = m2
Then problem becomes
y + m2y = 0 .....(6)
and y(0) = 0 and y(/2) = 0
The general solution is y(x) = c1 sin mx + c2 cos mx
When y(0) = 0, c2 = 0 and hence y(x) = c1 sin mx
When y(/2) = 0, 0 = c1 sinn /2
For seeking non-trivial solution, we should have c1  0 then sinn /2 = 0
or sin m/2 = n; for some positive integer n
 m/2 = n; n = 1, 2, 3, ....
 m = 2n; n = 1, 2, 3, .....
Therefore n = m2 = 4n2 ; n = 1, 2, 3, .....
Hence n = 4, 16, 36, ........, 4n2 ...... are the increasing sequence of eigenvalues. The corre-
sponding eigenfunctions are
yn(x) = sin 2nx; n = 1, 2, 3, ....
Ex.2. Find the eigenvalues and eigenfunctions for the boundary value problem y +
y = 0 under the boundary condition y(a) = 0 and y(b) = 0, 0 < a < b; a, b are arbitrary real
constants.
Case I :  < 0 or= – 2
Given problem reduces to
y – 2y = 0
with y(a) = 0 and y(b) = 0
The general solution is
y(x) = c1ex + c2e–x
When y(a) = 0, c1ea + c2e–a = 0  –c1e2a = c2
y(b) = 0, c1eb + c2e–b = 0  –c1e2b = c2
Hence, –c1e2a = –c1e2b
 c1(e2a – e2b) = 0
Since a  b, c1 = 0
and hence c2 = 0
which implies y  0 i.e. only trivial solution exists.
105
 Case II : If = 0
Given problem reduces to
y = 0
with y(a) = 0 and y(b) = 0
The general solution is y(x) = c1x + c2
For y(a) = 0, c1a + c2 = 0 and for y(b) = 0, c1b + c2 = 0
On subtracting we have c1 (a – b) = 0
Since a  b, c1 = 0
and hence c2 = 0 and y  0 i.e. we get only trivial solution.
Case III : When > 0 or = 2
Given problem becomes, y + 2y = 0
With y(a) = 0 and y(b) = 0
The general solution is y(x) = c1 cos x + c2 sin x
For y(a) = 0, 0 = c1 cos a + c2 sin a
For y(b) = 0, 0 = c1 cos b + c2 sin b
Non-trivial solution for c1 and c2 in above system of equation may exist only when we have
cos a sin a
0
cos b sin b
i.e. sin (b – a) = 0
or sin (b – a) = sinn ; for n = 1, 2, 3, .....
or (b – a) = n
n
or  ; n  1, 2,3,......
ba
Hence the eigenvalues are

2 n 2 2
n    ; n  1, 2,3,......
 b  a 2
and correspondingeigenfunctions are
n n
yn  x   c1 cos x  c2 sin x
ba ba
nb nb
If we suppose that c1  sin and c2  cos
ba ba
n
then eigenfurctions are yn  x   sin b  x 
ba
Ex.3. Find the eigenvalues and eigenfunction for the boundary value problem
y – 2y + y = 0; y(0) = 0, y() = 0
mx
Sol. Put y = e
Auxillary equation is m2 – 2m + = 0
m  1  1 
106
Case I : If 1 –  > 0 or< 1
The general solution is

y  x   c1e
1 
1 x 1
 c2e

1 x

Under given boundary conditions, y(0) = y () = 0, we have

c1 = c2 = 0 or y  0
So, only trival solution exist i.e. < 1 does not give any eigenvalue.
Case II : If 1 – = 0 or = 1
The general solution of given problem is
y(x) = (c1x + c2) ex
On applying boundary conditions,
y(0) = 0 and y() = 0 we have c1 + c2 = 0
Hence, only trivial solution exists and therefore = 1 is not an eigenvalue.
Case III : If 1– < 0 or > 1
The general solution is
y   A cos   1x  B sin   1x  e x

When y(0) = 0, we have A = 0 or y(x)  B sin   1xe x

For y() = 0, sin   1  0

since e  0 and B  0 for seeking non-zero solutions.
Hence sin   1  = 0 = sin n, n = 1, 2, 3, ....
  – 1 = n2
or n = n2 + 1; n = 1, 2, 3, .....
are required eigenvalues and corresponding eigenfunctions are
yn(x) = exsin nx n = 1, 2, 3, .....
Ex.4. Find the eigenvalues and eigenfunctions for the following boundary value prob-
lem
y – 4y + (4 – 9)y = 0, y(0) = 0, y(a) = 0,
where ‘a’ is a positive real constant.
Sol. The auxillary equation of a given problem is
m2 + 4m + (4 – 9) = 0
m  4  16  4  4  9   2  3 
Case I : when  = 0
The general solution of given problem is
y(x) = e–2x (c1 + c2x)
When y(0) = 0, c1 = 0
or y(x) = c2xe–2x
Also when y(a) = 0, c2ae–2a = 0
Since a > 0, therefore c2 = 0
Hence, y  0 i.e. only trivial solution exists.
107
 Case II : When > 0
The general solution is


y  x   e2 x c1e3 x  c2e3 x 
On applying boundary condition
y(0) = 0, c1 + c2 = 0 or c2 = – c1

 
y  x   c1e2 x e3 x  e3 x 
Again 
y  a   0, gives c1e2 a e3 a  e 3 a  0 
 c1 = 0  c2 = 0, y  0, only trival solution exists.
For  0, the given problem has no non-zero eigenfunction.
Case III : When < 0
The general solution of given differential equation is

  
y  x   e2 x c1 sin 3 x  c2 cos 3 x  
Now y(0) = 0 gives c2 = 0

 y  x   e2 x sin 3 x 

Also y  a   0 gives c1e 2 a sin  3 a  0 
For non-trivial solution, we have c1  0, then


sin 3 a  0 
or sin  3 a   sin n; n  1, 2,3, ....

n
  
3a

n22
or  
9a 2

 n 2 2
Hence n  ; n  1, 2, 3.....
9a 2
are the required eigenvalues for given problem. Hence the corresponding eigenfunctions are
 n x 
y n  x   e 2x sin   ; n  1, 2,3.....
 a 
Ex.5. Find the eigenvalues and eigenfunctions for the following boundary value prob-
lem
y – 3y + 2(1 + ) y = 0, y(0) = 0, y(1) = 0
Sol. Auxillary equation for given differential equation is
m2 – 3m + 2(1 + ) = 0
108
 3  9  4  2  1   
Solving, we get m
2

3 1  8
 
2 2
Thus, three cases arise
1
Case I : When 1 – 8 = 0 or λ =
8
The general solution of equation is
y(x) = e(3/2)x (c1 + c2x)
Now y(0) = 0 gives c1 = 0.
Therefore y(x) = c2x(3/2)x
Again y(1) = 0 gives c2 = 0
Hence, y  0 is the only trivial solution of the given problem.
1
Case II : when 1 – 8> 0 or λ <
8
The solution of given equation is

y  x   e
3 2 x  1 2   c2e
18x 1 2  18x 
 c1e 
 
when, y(0) = 0  c1 + c2 = 0 or c2 = – c1

y  x   c1e
3 2  x  1 2  18x
 e
 1 2  18x 
e 
 

 1  8 
y  x   2c1e
3 2 x
or sin h   x
 2 
 1  8 
y 1  0  y 1  2c1e
3 2
Again sin h    0
 2 
 c1 = 0
Therefore c2 = 0. Hence y(x)  0,
1
Thus for   , only trivial solution exists.
8
1
Case III : when 1 – 8 < 0 or λ >
8
The solution is

 8  1 8  1 
y  x   e
3 2x
c1 sin x  c2 cos x
 2 2 
Now for y(0) = 0, we have c2 = 0

109
 8  1
y  x   c1e
3 2x
sin x
2
8  1
y 1  0  c1e
3 2x
Also sin 0
2
For seeking non-trivial solution, we have c1  0
8  1
therefore sin 0
2
8  1
or sin  sin n; for positive integral n
2

8  1 4 n 2 2  1
  n  8  1  2n   n  ; n  1, 2,3,....
2 8
are required eigenvalues and corresponding eigenfunctions are yn(x) = e(3/2)x sin nx (n  N)

6.4 Sturm-Liouville Problem

A boundary value problem consisting of second order homogeneous linear differential equation
of the form
d  dy 
 p  x     q  x   r  x   y  0 .....(1)
dx  dx 
where p, q and r are continuous real valued functions defined on a  x  b such that p has a continuous
derivative, p(x) > 0 and q(x) > 0 and  is a parameher independent of x and two homogenous bound-
ary conditions
A1 y(a) + A2 y(a) = 0 .....(2)
B1 y(b) + A2 y(b) = 0 .....(3)
where A1, A2, B1 and B2 are real constants such that A1 and A2 are not both zero and B1 and B2 are not
both zero simultaneously, is called Sturm-Liouville problem. All the problems we have discussed in pre-
vious section are Sturm-Liouville problems.
Ex.1. Check whether the boundary value problem
y – y = 0 with y(0) = 0 = y()
is Sturm-Liouville problem or not
Sol. On comparing with stanard form of Sturm-Liouville problem, we have
p(x) = 1, q(x) = 1, r(x) = 0, a = 0 and b = ;
A1 = B1 = 1 and A2 = B2 = 0
Hence given problem is Sturm-Liouville problem.
Ex.2. Check whether the following boundary value problem
xy + y + (x2 + 1 + ) y = 0
y(0) = 0 and y(L) = 0, L is constant such that L > 1
is Sturm-Liouville problem or not.
110
 Sol. xy + y + (x2 + 1 + ) y = 0
(xy) + (x2 + 1 + ) y = 0
 p(x) = x, q(x) = 1, r(x) = 1 + x2,a = 0 and B = L;
A1 = 1, B1 = 0, A2 = 0 and B2 = 1
Since p(x) > 0 for 0  x  L
Given boundary value problem is Sturm-Liouville problem
Ex.3. Find the eigenvalues and eigenfunctions of the following Sturm-Liouville prob-
lem
d  2x dy  2x
e  +  λ + 1 e y = 0;
dx  dx 
y(0) = 0 = y()
Sol. Transform dependent variable from y to u by using transformation
y = e–xu
dy du  x
  e x e u
dx dx
Therefore given differential equation reduces to
d  2 x   x du  x  
 e e  e u       1 e2 x e  x u  0
dx   dx 

 du  x 
 2e 2 x  e  x e u
 dx 

 du  x d 2u  x  x du

e 2 x  e  x e  e u  e 
 dx dx 2 dx 
 
2x –x 2x –x
+ e  e u + e e u = 0
 d 2u 
or e x  2  u   0
 dx 
i.e. u + u = 0
and boundary conditions reduce to
u(0) = 0 = u() since e–x  0  x R
we know that n = n2; n = 1, 2, 3, ....
are the eigenvalues for reduced problem and corresponding eigenfunctions are un(x) = sin nx (see §6.3)
Hence n = n2; n = 1, 2, 3, .... are the eigenvalues for given problem and corresponding
eigenfunctions are
yn(x) e–x sin nx ; n N
Ex.4. Solve the following Sturm-Liouville problem
y + y = 0; y(–) = 0, y() = 0
Sol. Let < 0 i.e. = –2
Then given problem becomes
y – 2y = 0; y(–) = 0, y() = 0
111
 The general solution is y(x) = c1 ex + c2e–x
 y(x) = c1ex – c2e–x
Now y(–) = 0  c1e–– c2e= 0
and y() = 0 c1e– c2e–= 0
For non-trivial solution for c1 and c2 for above system of equations, the coffecient determinant
must vanish. Hence

e   e
0
e  e 
 – e– + e–= 0
which is not possible. Hence c1= c2 = 0
Therefore only trivial solution exists that is y = 0
When = 0.
The general solution is y = c1 x + c2
So, y = c1
For boundary condition y(–) = 0 and y() = 0, c1= 0
Hence y (x) = c2 is solution
When > 0. Let = 2
Then given problem becomes
y + y = 0
The general solution of the differential equation is
y(x) = c1cos x + c2sin x
y(x) = – c1sin x + c2cos x
An appling boundary condition y(–) = 0, we have c2 = 0
 y(x) = – c1sin x
Again for y() = 0; –c1sin  = 0
Since c1  0; therefore sin = 0, i.e. sin = sin n; n =1,2,3, .....
or = n ; n = 1,2,3, .....
 n= n2 ; n = 1,2,3, .....
are the required eigenvalues and corresponding eigenfunctions are yn(x) = cos nx
Hence from Case II and Case III, the eigenvalues for given problem are n = 0,1,4,9, .....n2 .....
and corresponding eigenfunctions are yn(x) = 1, cos x, cos2x, cos3x, .....cos nx, .....

6.5 Orthogonality of Eigenfunctions

From previous section, it is very much clear that the Sturm-Liouville problem is advanced boundary
value problem and have non-trivial solution if function p(x) and q(x) are restricted for p (x) > 0 and
q (x) > 0 on [a, b] and iff the parameter takes a certain specific value. These are termed as eigenval-
ues of boundary value problem. They are real numbers that can be arranged in an increasing sequence :
112
 1 < 2 < ....... <n < n+1< ...... ....(1)
and furthurmore n  as n 
This ordering is desirable to arrange corresponding eigenfunctions
y1 (x), y2 (x), ....., yn (x), yn+1 (x), ... .....(2)
in their own natural order. The eigenfunctions are not unique, but with the boundary conditions, they are
determined up to a non-zero constant factor.
Now, we introduce a new concept in broader context that will assist to understand the property
of various special functions that generally arise in various physical and engineering modelling.
A sequence of eigenfunctions yn (x) in (2) having the property
b 0 if m  n
y
a m n x  y  x  dx  
n  0 if m  n
is said to be orthogonal on the interval [a, b].
If n = 1,  n, the function yn (x) are said to be normalized and sequence of eigenfunctions is
known as orthonormal sequence.
If sequence of eigenfunctions yn (x) have the following general property
b 0, mn
a q  x  ym  x  yn  x  dx   n  0, mn
then, this sequence is said to be orthogonal with respect to a weight function q (x).

6.6 Important Theorems of Sturm-Liouville Systems

6.6.1 Theorem 1. The eigenvalues of Sturm-Liouville system are real

Proof. We have
d  dy 
 p  x     q  x    r  x    0 .....(1)
dx  dx 
where a1 y (a) + a2 y(a) = 0, and b1y (b) + b2 y (b) = 0 ......(2)
Suppose the p(x), q(x), r(x), a1, a2, b1 and b2 are real, while and y may be complex. Let 
and y denote complex conjugates of and y respectively. Now we have from (1) and (2)
d  d y
p  x  q  x   r  x    0 .....(3)
dx  dx  
where a 1 y  a   a2 y  a   0 , and b1 y b   b2 y  b   0 . .....(4)
Multiplying (1) by y and (3) by y and then subtracting we find that
d
 p  x  yy   yy        r  x  yy .....(5)
dx 
Integrating it from a to b and using boundary conditions (2) and (4), we find that
b
     a r  x  yy dx  0 .....(6)
Since r(x) is a non-negative and r(x)  0 for a  x  b, therefore (6) gives
    0       is real.
113
 6.6.2 Theorem 2 : Let  m and  n be two distinct eigenvalues of the Sturm-Liouville
problem
d  dy 
 p  x     q  x   r  x   y  0 .....(6)
dx  dx 
and ym(x) and yn(x) be their corresponding eigenfunctions. Then ym(x) and yn(x) are
orthogonal with respect to the weight function q(x) on the interval a  x  b.
Proof : If m and n are eigenvalues of given Sturm-Liouville problem

 p  x  y  x     q  x   r  x   y  x   0 .....(7)
then we have

 p  x  ym  x      m q  x   r  x  ym  x   0 .....(8)

and

 p  x  yn  x      n q  x   r  x   yn  x   0 .....(9)

On multiplying by (8) by yn and (9) ym respectively and on subtracting we get.

yn  x   p  x  ym  x    ym  x   p  x  yn  x     m   n  q  x  ym  x  yn  x   0

   m   n  q  x  ym  x  yn  x   ym  x   p  x  yn  x    yn  x   p  x  yn  x 

On integrating writh respect to x between a and b, we have

b b  b
  m   n  a q  x  ym  x  yn  x  dx  a ym  x   p  x  yn  x   dx   yn  x   p  x  ym  x   dx
  a

b b b
   m   n   a q  x  ym  x  yn  x  dx   ym  x  p  x  yn  x   a  a ym  x  p  x  yn  x  dx

b b
  yn  x  p  x  ym  x   a   yn  x  p  x  ym  x  dx
a

b
   m   n   a q  x  ym  x  yn  x  dx  ym  b  p  b  yn  b   ym  a  p  a  yn  a 
 yn  b  p  b  ym  b   yn  a  p  a  ym  a 
b
   m   n   a q  x  ym  x  yn  x  dx  p  b   ym  b  yn  b   yn  b  ym  b 
 p  a   ym  a  yn  a   yn  a  ym  a   
Now define w(x), a Wronskian determinant of the solution or eigenfunctions ym (x) and yn (x) as
ym  x  ym  x 
w x   ym  x  yn  x   yn  x  ym  x 
yn  x  yn  x 
So, expression (10) can be written as
b
  m   n  a q  x  ym  x  yn  x  dx  p  b  w  b   p  a  w  a  .....(11)
114
 For obtaining the orthogonality property
b
a q  x  ym  x  yn  x  dx  0 for m # n
We seek right hand side of (10) or (11) to vanish, that is
p(b) w(b) – p(a) w(a) = 0
This will certainly happen if the boundary conditions required for a non-trival solution of (7) are

y  a   0 and y  b   0 

or  .....(12)
y   a   0 and y   b   0 

Above boundary conditions are special cases of more general boundary conditions.
c1y(a) + c2y(a) = 0 and d1y(b) + d2y(b) = 0 .....(13)
where c1 and c2 donot vanish simultaneously and simillary d1 and d2 do not vanish simultaneously. To
verify that the general boundary condition in (13) really vanishes the right hand side of (11), Let
eigenfunction ym(x) and yn(x) also satify boundary condition (13) i.e.
c1 ym  a   c2 ym  a   0

c1 yn  a   c2 yn  a   0
For non-trivial solution of c1 and c2 in above system of equations, the determinant
ym  a  ym  a 
 wa
yn  a  yn  a 

must vanish. Hence w (a) = 0. Simillarly w(b) = 0.

So right hand side of (11) definitely vanishes and orthogonality of eigenfunctions is validated un-
der suitable boundary condition (13) which are homogeneous in nature. The problem (7) with boundary
condition (13) is known as Sturm-Liouville problem.
The significance of orthogonality property of eigenfunctions of Sturm-Liouville problem is to rep-
resent series expansions of function f (x) in terms of eigenfunctions yn (x) as
f (x) = a1 y1 (x) + a2y2 (x) + ...... + an yn (x) + ...... .
where the cofficient a1, a2 .... an, ..... can be derived using orthogonality property of eigenfunctions.
6.6.3 Theorem 3 :To every eigenvalue of a Sturm-Liouville system there corresponds
only one linearly independent eigenfunction.
Proof. Let if possible, y1 (x) and y2 (x) be two distinct eigenfunctions of the systems, corresponding
to same eigenvalue . In order to prove the linear independence of y1(x) and y2 (x), it is sufficient to
prove that the wronskian

y1  x  y2  x 
w (x) = y  x
1  y2  x  is identically zero.
By definition,
w (x) = y1 y2  y2 y1
115
 We have w  x   y1 y2  y2 y1
and from the given boundary conditions
w (a) = w (b) = 0 .....(14)
Since y1(x) and y2(x) are solutions of Sturm-Liouville’s problem, therefore

 p y1    q  r  y1  0
and  p y2    q  r  y2  0
Eliminating (q + r), we get
 y2 y1  y1y2  p  x    y2 y1  y1 y2  p  x   0
or p  x  w  x   p  x  w  x   0
C
or d  p  x  w  x    0  w  x  
p  x
Since p(x)  0, the boundary condition (14) gives C = 0 for all x. Hence w(x)  0 in [a, b],
which means that, the eigenfunction y1(x) and y2(x) corresponding to same eigenvalue  are linearly in-
dependent.
6.6.4 Theorem 4 : (Expansion of a function in terms of eigenfunctions of Sturm-Liouville
system). If {n(x)} be a set of eigenfunctions of Sturm-Liouville system, then

 Ann  x  converges uniformly to a function f(x) in [a, b] such that
n1


f  x   An n  x  , a xb .....(15)
n 1

b
 a r  x  f  x  m  x  dx
where Am  , mN .....(16)
b
a r  x  2m  x  dx

Proof. Without taking the proof of convergence, let f(x) is given by (15). Multiplying both sides
of (15) by r(x) m(x), integrating from a to b and changing the order of integration and summation (which
is justified due to uniform convengence of the series) we find that

b  b
a r  x  f  x  m  x  dx   An  r  x  n  x  m  x  dx .....(17)
a
n 1
Since the set of eigenfunctions of Sturm-Liouville system are orthogenal in [a, b] w.r.t weight
function r(x), therefore relation (17) reduces to
b b 2
 a r  x  f  x  m  x  dx  Am  a r  x  m  x  dx
which gives Am given by (16).
116
Ex.1. Compute the eigenvalues and eigenfunctions for boundary value problem
y  2 y   1    y  0; y   0   0 and y(1) = 0
Also prove that the set of eigenfunctions for the given problem is an orthogonal set.
Sol.  The auxillary equation is m2 + 2m + (1 – ) = 0

2  4  4 1   
or m  1  
2
Now, three cases arise
Case I : When > 0 or = 2
The general solution of the given differential equation in this case will be
 1   x  c e 1   x
Now y (x) = c1e 2
For y (0) = 0  c1 + c2 = 0 or c2 = – c1

  1   e 1   x 
 x
 y  x   c1  e 
 
  1    e 1     0
Now y (1) = 0 gives c1 e 
 
 c1 = 0
Hence c2 = 0 = c1  y(x)  0 i.e. only trivial solution exists.
Case II : When = 0 :
The general solution is y(x) = e–x (c1 + c2 x)
For y (0) = 0, we get c1 = 0. Hence y (x) = c2xe–x
When y (1) = 0, we get c2 e–1 = 0 c2 = 0.
Thus c1 = c2 = 0, which gives y  0 i.e. only trivial solution exists.
Case III : When < 0 or = – 2
Then general solution is
x
y (x) = e  c1 cos  x  c2 sin  x 
For y (0) = 0, we have c1 = 0
So y  x   c2 e  x sin  x

Now, for y(1) = 0, we have c2e 1 sin   0

For seeking non-trivial solution of given problem, we have c2  0, so sin   0
or sin   sin n; n is positive integer
   n 
 – = n22
 l n= – n22 ; n = 1, 2, 3, ...... .
Hence, corresponding eigenfunctions are
yn (x) = e–x sin nx
117
 Let ym (x) = e–x sin mx and yn (x) = e–x sin nx are two eigenfunctions corresponding to eigen-
values m = – m22 and n = – n2 2 resectively. Then the integral
1 2x 1
0 e ym  x  yn  x    e 2 x e x sin mx e  x sin nx dx
0

1
  sin mx sin nx dx
0

11
 cos  m  n  x  cos  m  n  x  dx
0 2
1
1  sin  m  n  x sin  m  n  x 
   
2  m  n  m  n  0
=0
prompts that ym (x) and yn (x) are orthogonal in [0, 1] with respect to weight function e2x.

Self-Learning Exercise
1. Classify the following problem as boundary value problem or initial value problem
(a) y – y = 0, y(0) = 0 and y(1) = 0
(b) y + 2y + 2y = 0, y(0) = 1
(c) (xy) + (9 + 4) y = 0, y(a) = 0, and y(b) = 0, a, b are constonts
(d) 3y + 4y + 2y = 0, y(2) = 5, y(2) = 6
2. Find the eigenvalues n and eigenfunctions yn(x) for y + y = 0 in each of the following bound-
ary conditions
(a) y(0) = 0, y(1) = 0
(b) y(–2) = 0, y(2) = 0
(c) y(–3) = 0, y(0) = 0
(d) y(1) = 0, y(4) = 0
3. Check whether following boundary value problems are Sturm-Liouville problem or not
(a) exy + exy + y = 0; y(0) = 0, y(1) = 0
(b) y + (1 +x)y = 0; y(0) = 0, y(2) + y(2) = 0

 1 
(c)  y  + (x + ) y = 0; y(0) + 3y(0) = 0, y(1) = 0
x 
(d) (xy) + (x2+ 1 – x2)y = 0; y(0) = 0; y(0) + 3y(0) = 0, y(1) + y(1) = 0
(e) (xy) + (x2+ 1 + ex)y = 0; y(1) = 0; y(1) + 2y(1) = 0; y(2) – 3y(2) = 0
4. Find eigeavalues and corresponding eigenfunction of the following Sturm-Liouville problems.
(a) y+ y = 0; y (0) = 0 and y() = 0
(b) y+ y = 0; y (0) = 0 and y(L) = 0
(c) y+ y = 0; y (–) = 0 and y() = 0
118
6.7 Summary

In this unit, we introduced a special type of boundary value problem known as Sturm-Liouville
problem which gives fundamental basics for important concepts like eigenvalue, eigenfunction, orthogo-
nality and Fourier series. These concepts directly involved in solving practical problems arise in physical
and engineering challenges.

6.8 Answer to Self-Learning Exercise

1. (a) Boundary value problem

(b) Initial value problem
(c) Boundary value problem
(d) Initial value problem
2. (a)  n  n 2 2 ; n  1, 2,3...... yn  x   sin n  x

n22 n
(b)  n  ; n  1, 2, 3...... yn  x   sin  x  L 
16 4
n 2 2 nx
(c)  n  ; n  1, 2,3...... yn  x   sin
9 3
n22 n
(d)  n  ; n  1, 2,3...... yn  x   sin  4  x 
9 3
3. (a) Yes (b) Yes
(c) No, since p(x) is not continuous in [0, 1]
(d) No, since q(x) < 0 in [0, 1]
(d) Yes

4. (a)  n 
 2n  12 ; n  ,1, 2,3....., yn  x   sin
2n  1
x
4 2
n 2 2 cos n x
(b)  n  2
; n  0,1, 2, 3....., yn  x  
L L
(c)  n  n 2 ; n  1, 2,3....., yn  x   cos nx

6.9 Exercise

1. Find the eigenvalues n and eigenfunction yn (x) for the following boundary value problem
y   y  0 in each of the following boundary conditions :

n2 nx
(a) y (0) = 0, y (2) = 0 [Ans.  n  ; n  1, 2,3,......, yn  x   sin ]
4 2
(b) y (0) = 0, y (L) = 0 ; L > 0, L is positive constant
n 2 2 nx
[Ans.  n  2
; n  1, 2,3,......, yn  x   sin ]
L L
119
 (c) y (–L) = 0, y (L) = 0 ; L > 0, L is positive constant
n 22 n  x  L 
[Ans.  n  2
; n  1, 2, 3,......, yn  x   sin ]
4L 2L
2. Solve the following Sturm-Liouville problem
d  dy  
(a) 
 x   y  0; y 1  0, y e  0
dx  dx  x
 
[Ans.  n  n 2 ; n  1, 2,3,......, yn  x   sin  n ln | x | ]

d  2 dy  
(b) 
 x 1
dx 
 2
dx  x  1
y  0; y(0) = 0 and y(1) = 0 (Hint put x = tan t)


[Ans.  n  16n 2 ; n  1, 2,3,......, yn  x   sin 4n tan 1 x ] 
3. Compute the eigenvalues and eigenfunctions for boundary value problem and determine Euclid-
ean space in which a complete set of eigenfunctions for the given problem is an orthogonal set
(a) y   1    y  0; y  0   0, y    0

[Ans.  n  n 2  1; n = 1, 2, 3, ....., yn (x) = sin nx orthogonal in [0,]

(b) 4 y   4 y  1    y  0, y  1  0, y 1  0

 x2 nx
e sin 2 ; n  2, 4, 6......
[Ans.  n  n 22 , yn (x) = 
e x 2 cos nx ; n  1,3,5......
 2
Orthogonal in (–1, 1) with resepect to function e–x ]
(c) y   2 y  1    y  0; y  0   0 and y      0
[Ans. n = – n2 ; n = 1, 2, 3, ..... ., yn = e–x (n cos nx + sin nx), n = 1, yn = 1. Orthogonal in
[0, ] with respect to weight function e2x ]
4. Find the real eigenvalues and eigenfunctions for the boundary value problem y   y  0 ;
y (0) = 0, y  1  0 [Ans.  0  0; y 0  x   1;  n  n 2 2 , y n  x   cos n x , n  N ]
1 
5. Find the solution of Sturm-Liouville problem y   y  2 y  0, 1  x  2
x x

 log x 
with boundary conditions y 1  0  y  2  [Ans. y   Bn sin  n ]
n 1  log 2 
7. Determine the normalized eigenfunctions of the problem y   y  0 , y(0) = 0, y  1  y 1  0 .
Hence expand the function f (x) = x, 0  x  1, in terms of these normalized eigenfunctions.
12
 2   4sin  n
[Ans. y n  x      
sin x  n , n  N , x   
sin x  n ] 
2
1  cos  n  n 1 
 n 1  cos 2  n 
  
120
Unit 7 : Variational Problems with Fixed Boundaries and
Euler-Lagrange Equation
Structure of the Unit

7.0 Objective

7.1 Introduction

7.2 Definitions and Fundamental Problems

7.2.1 Functionals

7.2.2 Linear Functionals

7.2.3 Brachistochrone Problem

7.2.4 Problem of Geodesics

7.2 Isoperimetric Problem

7.3 Euler-Lagrange Equation

7.3.1 Basic Lemma

7.3.2 Euler-Lagrange Equation

7.4 Some Elementary Cases of Integrability of Euler-Lagrange Equation

7.4.1 F is independent of y

7.4.2 F is independent of x and y.

7.4.3 F is independent of only y.

7.4.4 F is a linear function of y.

7.4.5 F is independent of only x

7.5 Variational Problems for Functionals Involving Several Dependent Variables and Their First

Order Derivatives.

7.6 Summary

7.7 Answers to Self-Learning Exercise

7.8 Exercise
121
7.0 Objective

In this unit you will study the methods of finding curves connecting two given points which either
maximizes or minimizes some given integral. You will also know about Euler-Lagrange equation for an
extremal. Variational problems involving several independent variables will also be discussed.

7.1 Introduction

Calculus of variations is a field of mathematics that deals with extremizing functionals as opposed
to ordinary calculus which deals with functions. The origin of calculus of variations was based on famous
“Brachistochrone problem or quickest path problem.” In calculus of variation, we generally encoumter
with the problems where one has to find the maximal and minimal value that is extreme value of special
quantities called functionals.

7.2 Definitions and Fundamental Problems

7.2.1 Functionals : Functionals are variable quantities whose values are determined by choice
of one or several functions. In short, we may say that functionals are functions of functions.
Ex.1. Let the parametric equations of the plane curve be x = x(t), y = y(t), t being the param-
eter. The arc length of the plane curve from P(t0) to Q(t) is given by
t
s [x(t),y(t)] =  t0 x 2  y 2 dt

Q (t)
y
s

P (t 0)

o x

Fig. 7.1
where x and y represent the differentiation of x and y with respect to ‘t ’ respectively..
Here s is a functional which is function of functions x(t) and y(t).
7.2.2 Linear Functionals : A functional L [y(x)] satisfying the conditions.
(i) L [cy(x)] = cL [y(x)]
(ii) L [y1(x) + y2(x)] = L [y1(x)] + L[y2(x)]
where c is a arbitrary constant is known as linear functional.
122
 x
 dy 
Ex.2. L  y  x     a  x  dx  b  x  y dx, is a linear functional.
x0

The calculus of variations provides a method for determining maximal and minimal values of
functionals. Such problems are known as variational problems.
Now we deal with three problems of historical importance which influenced the development of
this subject.
7.2.3 Brachistochrone Problem
Suppose P and Q are two points in the plane but not in the same vertical line. Imagine, there is a
thin flexible wire connecting those two points. Suppose P is above Q, and we let a frictionless bead
travel under gravity from P to Q. The Bachistochrone problem (or quickest discent problem) is con-
cerned with determining the path of the bead when it reaches the point Q in the least possible time. This
problem was first introduced by J. Bernaulli in the mid of 17th century and was first solved by Sir Isaac
Newton.
7.2.4 Problem of Geodesics
In general relativity, a geodesic generalizes the concept of straight line to curve spacetime. For
example : Find the curve of shortest length connecting two points in space. If there is no constraints the
solution obviously is a straight line joining the points. However, if the curve is constrained and is to lie on
a surface, then in space, the solution is less obvious and possibly many solutions may exist.
The solutions are called geodesics. In other words a geodesics on a surface is a curve along
which the distance between two points on the surface is a minimum. To find the geodesics on a surface
is a variational involving conditional extremum.
7.2.5 Isoperimetric Problem
In this problem, we required to find a closed plane curve of a given length l bounding a maximal
area S. Let the parametric equation of the plane curve be x = x(t) , y = y(t), and the curve is traversed
once in anti-clockwise as t increases from t0 to t1, then length l of given curve is
t1
2 2
l   x  t     y  t   dt ......(1)
t0
which is a constant, and enclosed area is given by
t
1
S    xy  yx  dt .....(2)
2t
0
The problem is to maximize the functional S, given by (2) subject to the condition that the length
l of the curve given by (1) must have a constant value.

7.3 Euler-Lagrange Equation

7.3.1 Basic Lemma : Let M(x) be a continuous function on the internal [a, b]. Suppose
b
that for any continuous function h(x), we have  M  x  h  x  dx  0 then M(x)  0 on the inter-
a
val [a,b].
123
 Proof : Let M(x)  0 (say positive) at a point x where a  x  b. Since M(x) is continuous
on [a, b], it follows that if M(x)  0. Then M(x) maintains its sign in a certain neighbourhood x0  x  x1
of the point x .
Since h(x) is arbitrary continuous function, we may choose h(x) s.t. h(x) remains positive in
x0  x  x1 while it vanishes outside the interval. Hence, we obtain.
b x1

 h  x  M  x  dx   h  x  M  x  dx  0 .....(1)
a x0

Since the product h(x)M(x) remains positive in [x0, x1] and vanishes outside this interval.
b

By the hypothesis  h  x  M  x  dx  0 .....(2)

a

which contradicts (1). This contradiction shows that our assumption M(x)  0 at some point x
must be wrong and so M(x)  0 on [a, b].
7.3.2 Euler-Lagrange Equation : If y(x) is a curve in interval [a, b] which is a twice
differentiable and satisfying the conditions y(a) = y1 and y(b) = y2 and minimizes the functional.
b
F  y  x     f  x , y , y  dx .....(3)
a

dy
where y   .
dx
Then the following differential equation must be satisfied
f d  f 
  0 .....(4)
y dx  y  
Proof. Suppose y  y(x) is a curve which minimizes the functional F. That is, for any permissible
curve y = g(x), F[y(x)]  F[g(x)]. We have to construct a function of one real variable satisfying follow-
ing properties.
1. H() is a differentiable near = 0
2. H(0) is a local minimum for H.
We begin by constructing a variation of y(x). Let  be a small real number (positive or
negative). s.t.
y  x   y  x    h  x 
where h(x) is a continuous function in [a, b] and h(a) = h(b) = 0.

o x=a x=b x
Fig. 7.2
124
 We can define a function H to be
H  F  y  x  
Since y(x) minimizes F(y(x)), it follows that it minimizes H(). Since H(0) is minimum value of
H, we know that from ordinary calculus that H(0) = 0.
The function H can be differentiated by using Leibnitz rule, that is
b
d d  
H     f  x, y, y  dx 
d d   a 
b

  f  x, y, y  dx  .....(5)
a
 

Now applying chain rule within the integral, we abtain

f f x f y f y
     
  x   y   y  
f y f y
   
y   y  
f f
 h  x  h  x 
y y
f
Substituting the value of in the equation (5), we have

b
dH   f f 
  h x  h  x   dx
d y
a 
y 
Using H(0) = 0, we find that
b
 f f 
H   0    h  x   h  x   dx  0
a
y y  
Integrating by parts, we get
b b b
f  f d  f  
H   0   h  x  dx   h  x    h  x  dx  0
a
y  y a
dx  y   a

b b
 f d  f    f 
      h  x  dx   h  x    0
a  y dx  y    y  a
b
 f d  f  
  
   h  x  dx  0 [Using h(a) = h(b) = 0]
a
y dx  y 
By using lemma, we conclude that
f d  f 
  0 .....(6)
y dx  y  
125
 This equation is called Euler-Lagrange equation.
7.3.5 Remark : The statement of the lemma and its proof donot change if restriction h(a) =
h(b) = 0 is imposed on the function h(x).

7.4 Some Elementary Cases of the Integrability of the Euler-Lagrange Equation

7.4.1. f is independent of y : If f is independent of y, then f is function of (x, y) only. There-

f
fore  0 . Thus the Euler-Lagrange equation reduces to following form :
y
f
0 .....(1)
y
Now integrating (1), with respect to y, we obtain a arbitrary curve f = g(x), without any con-
stant and in general, does not satisfy boundary conditions y(a) = y1 and y(b) = y2. Thus this type of
equation does not posses a solution.
7.4.2. f is indpendent of x and y : In this case,

f f 2 f 2 f
   0 .....(2)
x y xy yy
From Euler-Lagrange equation
f d  f 
    0 , we get
y dx  y  

f  2 f 2 f 2 f
  y  y  2  0  f  f  x, y , y   
y xy  yy y 
From equation (2), we have
2 f
 y 0 .....(3)
y  2

2 f
This implies that either y = 0 or 0
 y 2
Now y = 0
 y = Ax + B .....(4)
2 f
where A and B are arbitrary constants, which is a two parameter family of straight lines. But if 0
 y 2
has one or several real roots y = Kn, then y = Knx + c
which is one parameter family of straight line contained in two parameter family of straight lines. Thus
extremals are all possible straight lines.
7.4.3. f is indpendent of only y : Here f  f(x, y ), therefore Euler-Lagrange equation can be
written as
126
 d  f  f
  0 as 0
dx  y  y

f
 c .....(5)
y
where c is a constant. Since this relation is independent of y it can be solved for y as a function of x.
Another integration leads to a solution involving two arblitraray constants which can be obtained by
using given boundary conditions.
7.4.4. f is a linear function of y or f is binearly dependent on y such that f(x, y, y) =

p(x, y) + q(x,y)y

Forming the Euler-Lagrange equation for this particular f, we have

p q dq p q  q q  p q
 y    y      y    0 .....(7)
y y dx y y  x y  y x
for all x and y.
Solution of this problem, in general, not possible because solution does not satisfy given bound-
p q
ary conditions. But if we consider   0 , then the expression pdx + qdy becomes exact differen-
y x
tial equation whose solution does not depend on path of extremal and therefore variational becomes
meaningless.
Ex.1. Test for an extremum of the functional
1
F  y  x      x 2 y 2  x 2 y  dx , y  0   0, y  1  1 .....(8)
0

Sol. Clearly we see that

f  x, y , y   x 2 y 2  x 2 y
is a linear function of y. Now from case 7.4.4, we have p(x,y) = x2y2, q(x,y) = x2
Hence from equation (7), we find that
p q
 0
y x
 2x2y – 2x = 0
 2x (xy-1) = 0
 xy =1 or x=0
Obviously first boundary condition is satisfied by only x = 0, by and second boundary condition
is satisfied by only xy = 1. Both boundary conditions are not satisfied by the curves x = 0 and xy =
1.Thus no solution exist for this problem.
127
 Ex.2. Test for extremum of the functional
b
F  y  x      cos y  xy sin y  dx .....(9)
a

with boundary conditions y(a) = y0, y(b) = y1

Sol. For this problem, Euler-Lagrange equation is given by

d
 sin y  xy  cos y    x sin y   0
dx
or  sin y  xy  cos y  sin y  xy  cos y  0
Thus, integrand being an exact differential equation. Therefore variational problem becomes mean-
ingless
f
7.4.5. f is independent of x : In this case,  0 , therefore Euler-Lagrange equation re-
x
duces to
d  f 
  f  y 0
dx  y 
f
Hence Euler-Lagrange equation has its integral as f  y c
y 

where c is arbitrary constant

Ex.1. Test for extremum of the functional
1


F y x =   1  y 2 dy , y(0) = 0, y(1) = 2
0

Sol. Using Euler-Lagrange equation, we get

d  2 y 
  0
dx  1  y 2 

Integrating with respect to ‘x’, we get
y
 c , where c is arbitray constant
1  y 2

c2
 y    A  say 
1  c2
Again integrating with respect to ‘x’
y = Ax + B
y(0) = 0 and y(1) = 2, implies that B = 0, A = 2
Thus y = 2x which is a straight line.
128
 Ex.2. Test for extremum of the functional
1
F  y  x      y 2  x 2  dx , y  0   1, y  1  2
0

Sol. Using Euler-Lagrange equation, we get

f d  f 
  0
y dx  y  
d
   2 y  0
dx
 y
Integrating two times we get y = Ax + B
Using y(0) = 1, y(1) = 2, we get A = B = 1.
Thus solution is y = x + 1.
Ex.5. (Brachistochrone problem or quickest descent problem)
Find the shape of the curve on which a bead is sliding from rest and accelerated by
gravity will ship (without friction) in least time from one point to another.
Sol. Let us consider a particle P descending from A(0,0) to B(a,b) under gravity along some
curve. We have to determine shape of the curve which gives minimum possible time to descent. Let
P(x,y) be the position of the particle at any time t and having actual arc length s from a point A.
A (0, 0 ) x

P (x, y )
y
B (a, b )

Fig. 7.3
Under the gravity, the motion of particle is given by
ds
  2 gy
dt
ds
 dt 
2y g
Hence time T of descent is (from A to B).
a
ds
 T  .....(10)
0 2y g
But we know that
2
ds  dy 
 1  
dx  dx 

129
 dy
 ds  1  y2 dx where y  
dx
Putting the value of ds in equation (10), we obtain
a
1  y 2
T  dx
0 2 gy

1  y2
Here f(x, y, y ) = f(x)
2 gy
Now from case (7.4.5), we have
f
y  f  c1
y

1 y 1  y 2
 y     c1
2 gy 1 y  2 2 gy

1
 c2
(where c2   2 g c1 )


y 1  y 2 
or 
y 1  y2  c3   where c 3  1 c22 
c3
Now putting y   cot   y  c3 sin 2   1  cos 2 
2

dy dy
Since  y  dx 
dx y

2c3 cos  sin  d

 dx 
cot 

 dx  2c3 sin 2  d  c3 1  cos 2 

Integrating we get

 sin 2  c3
x  c3      c4   2  sin 2   c4
 2  2

c3
and y 1  cos 2
2
If we substitute 2 = , and using intial condition (that is at A(0,0)), we have
c3 c
c4 = 0 ; and x    sin   , and y  23 1  cos  
2
c3
which is equation of the cycloid with radius of rolling circle and c3 can be obtained by using
2
appropriate boundary condition.

130
 Ex.4. (The minimal surface of revolution problem)
Find the curve with fixed boundary revolves such that its rotation about x-axis gener-
ate mininal surface area.
Sol. We know that, surface area of the relvolution is given by
b
S  y  x     2 y ds
a

b
  2 y 1  y 2 dx
a

Here f(x,y, y ) = 2y 1  y2 f(x)

From case (7.4.5), the first integral of Euler’s equation is
f
f  y  c1
y 

2 yy 2

2 y 1  y2   c1
1  y2

y c1
 c2
 (where c2  )
1  y 2 2

dy
dx 

y 2  c22
Integrating with respect to ‘y’ we get

 y
x  c2 cosh 1    c3
 c2 

 x  c3 
y  c2 cosh  
 c2 
where c2 and c3 are arbitrary contestants, which is a equation of the “catenary” and the corre-
sponding surface of revolution is called “centroid” of revolution.

7.5 Functionals Involving Several Dependent Variables and Their First Order De-
rivatives.

We now proceed to derive the differential equations that must be satisfied by the twice differen-
tiable functions x1(t), x2(t), ...., xn(t) that extremize the integral
t2
I  f  x1, x2 ,.....xn , x1, x2 ,.....xn , t  dt 
t1

131
with respect to those functions of x1, x2, ...., xn which achieve prescribed values at the fixed limits of
integration t1 and t2, where t1 < t2. The superior dot represents ordinary differentiation with respect to
the independent variable t.
We denote the set of actual extremizing functions by x1(t), x2(t), ...., xn(t) and proceed to form
the one-parameter family of comparison functions
X1(t) = x1(t) +1(t), X2(t) = x2(t) +2(t) ,....., Xn(t) = xn(t) +n(t) .....(2)
where 12nare arbitrary differentiable functions for which
1(t1) = 1(t2) = 2(t1) = 2(t2) = ....= n(t1) = n(t2) = 0 .....(3)
and is the parameter of the family. The condition (3) assures us that every member of each compari-
son family satisfies the required prescribed end point conditions. We see, moreover, that no matter what
the choice of 12n, the set of extremizing functions x1(t), x2(t), ...., xn(t) is a member of each
comparison family for the penameter value = 0. Thus if we form the integral.
t2
I    f  X1, X 2 ,....., X n , X 1, X 2 ,....., X n , t  dt .....(4)
t1

by replacing x1, x2, ...., xn etc, in (4) by X1, X2, ...., Xn etc., respectively, we have that I(0) is the extre-
mum value sought. We therefore conclude that
I  0  0 .....(5)
It follows from (2) that
X 1  x1   1 , X 2  X n   2 , ...., X n  xn   n .....(6)
Now differentiate (4) with respect to ‘’, we have
2 t
dI  f f  f f  f f  
 1  1  2   2  .....  n   n  dt , .....(7)
d  t  X 1 X 1 X 2 X 2 X n X n 
1

 X   X 
where we use (2) and (6) to derive the sequence of substitution  1   1 ,.....,  n   n .
     
It is clear from (2) and (6) that setting = 0 is equivalent to replaeing X1,X2, ....., Xn, X 1 , X 2 ,....., X n
by x2, x2, ...., xn, x1 , x2 ,.....xn respectively. Thus because of (5), we abtain from (7) on setting = 0
t2
 f f f f f f 
I 0    x1 1  x1 1  x2  2  x2 2  .....  xn  n  xn n dt  0 .....(8)
t1

This last relation holds for all choices of the functions 1(t),2(t), ...., n(t). In particular, it holds
for the special choice in which 2, ...., nare identically zero, but for which 1(t) is still arbitrary, consis-
tent with (3). With this selection of 1,2, ...., nwe integrate by parts the second term of the second
member of (8) to obtain, since 1(t1) = 1(t2) = 0,
t2
 f d  f  
 x dt  x1  1dt  0
  .....(9)
t1  1 
132
 Since (9) holds for all, 1 we conclude by applying the basic Lemma that

f d  f 
  0 .....(10)
x1 dt  x1 
Through similar treatment of the successive pairs of terms of the second member of (8) we de-
rive like equations, with x1 replaced by x2, ...., xn, Joining these equations with (10), we have

f d  f  f d  f  f d  f 
    0,     0,.....,   0 .....(11)
x1 dt  x1  x2 dt  x2  xn dt  xn 
for the system of simultaneaes Euler-Lagrange equations which must be satisfied by the functions x1(t),
x2(t), ....., xn(t) which render the integral (1) an extremum.
Ex.1. Find the extremals of the functional
π
2
I  y, z  =   y
2
+ z 2 + 2 yz  dt
0

with the boundary conditions y(0) = 0, y() = –1 ; z(0) = 0, z() = 1

    y 2  z 2  2 yz
Sol. Here f  y,z, y,z,t
Then from equation (11), we can see that
f d  f 
 0
y dt  y 
f d  f 
 0
z dt  z 
or 
y  z  0 and 
zy0 .....(12)
Eliminating ‘z’ from this system, we get
d4y
y    y  0 or
iv y0
dt 4
Its solution is given by
y = c1 et + c2 e–t + c3 cost + c4 sint .....(13)
where c1, c2, c3 and c4 are arbitrary constants. Now from equation (12) we have
y  c1et  c2e t  c3 cos t  c4 sin t
z   .....(14)
Applying the given boundary conditions
y(0) = 0, y() = –1, z(0) = 0, z() = 1, we find that
c1 = c2 = c3 = 0, c4 = –1.
Hence the extremal curve is the intersection of the surfaces
y = – sin t, z = sin t.
Ex.2. (a) Find the extremum of the function
1

F  y  x   
x2
 1  y2  2

 x
dx
x1

133
(b) Show that the curve through (1,0) and (2,1) which minimize
12
2
 1  y  2

 x
dx is a circle.
1

x2
Sol. (a) Comparing the given functional with  f  x, y, y  dx, we get
x1

where f  x, y , y   
 1  y2  2
.....(15)
x
Euler-Lagrange equation is
f d  f 
  0 .....(16)
y dx  y 
From (15), we have
f f y
 0 and  1
.....(17)
y y
x 1  y 2   2

f d  f 
Since  0, (16) reduces to  0
y dx  y 
f
Integrating it, we get c
y 
y
c
or
x 1  y 2
1
Thus, 
y   cx 1  y  2  2 .....(18)

dy
Now let  y   tan 
dx
Then (18) yields tan = cx sec
 x = c1 sin where c1 = 1/c
Now dy = tan dx = c1 tancosd = c1 sind
Integrating it, we get y = –c1 cos + c2
Thus x = c1 sin and y – c2 = – c1cos or x2 + (y – c2)2 = c12 .....(19)
which is a family of circle with center at axis.
(b) Proceed exactly as in part (a) upto (19). In the present problem, using the boundary condi-
tions x = 1, y = 0 and x = 2, y = 1, (19) yields
2
1  c22  c12 and 4  1  c2   c12 giving c1  5, c2  2 .
Hence from (19) the required curve is the circle x2 + (y – 2)2 = 5.
134
Ex.3. Obtain the Euler-Lagrange equation for the extremals of the functional
x2
2
  y  yy + y2  dx
x1

x2

Sol. Comparing the given functional with  f  x, y, y  dx , we get

x1

f  x, y, y   y 2  yy   y 2 .....(20)
Euler-Lagramge’s equation is
f d  f 
  0 .....(21)
y dx  y  

f f
From (10), we get  2 y  y ,   y  2 y
y y 

d  f 
and     y   2 y 
dx  y 
Using these values, the required Euler-Lagrange equation (21), becomes
2 y  y    y  2 y   0 or y   y  0
Ex.4. Test for an extremal of the functional

2
 
F  y  x      y  y 0  0, y    1
2
 y 2 dx ,
0 2

2
Sol. Comparing the given functional with  f  x, y, y dx , we get
0

f  x, y, y    y 2  y 2 .....(22)
Euler-Lagrange’s equation is
f d  f 
  0 .....(23)
y dx  y  

f f d  f 
From (22), we have  2 y,  2 y and    2 y .
y y  dx  y 
Using these values, (23) reduces to
–2y – (2y) = 0 or y+ y = 0
d
or (D2 + 1) y = 0 where D  .....(24)
dx
y = c1 cos x + c2 sin x .....(25)
Using boundary condition, we get
c1 = 0 and c2 = 1.
Hence, from (25), an extremun can be attained only on the curve y = sin x
135
 Self-Learning Exercise

x
2
1. Is L  y  x    y dx is linear ?  Yes / No.
x0

x
d2y 
2. Is L  y  x   
  dx 2  c  x  y  dx is linear ?  Yes / No.
x0
3. As extremal of the functional
b
F  y  x     f  x, y , y   dx, y(a) = y1 , y(b) = y2 satisfies Euler-Lagrange equation,
a

which in general is a
(a) linear second order ODE
(b) admits a unique solution
(c) non-linear ODE of order greater than two.
(d) may not admit a solution.
4. The curve of shortest distance between two fixed points is
(a) straight line
(b) circle
(c) parabola
(d) none of these
b
5. The Euler-Lagrange equation for a functional of the form  f  x, y  dx is
a

(a) fy = c1
(b) fy – yfy= c1
(c) fy = c1
(d) none of these
6. The extremizing curve of the brachistochrone problem is a
(a) circle
(b) catenary
(c) cycloid
(d) straight line.

7.6 Summary

The caluclus of variation, which plays an important role in both pure and applied mathematics,
dates from the time of Newton. Development of the subject started mainly with the work of Euler and
Lagrange. In this unit we have solved a number of problem of engineering and physics with the help of
Euler-Lagrange equations.
136
7.7 Answers to Self-Learning Exercises

(1) No (2) Yes

(3) (d) (4) (a)
(5) (d) (6) (a)

7.8 Exercise

1
1  y2
1. Find the extemal of the function I  y  x     y dx, throught the origin and the point (1, 1).
0
[Ans. y = tan(x/4)]
2. (a) Show that if y satisfies the Euler-Lagrange’s equation associated with the integral
x2

p y2  q 2 y 2 dx

2
I
x1

x2
2
where p(x) and q(x) are known functions, then I has the value  p yy  x   1

(b) Show that, it y satisfies the Euler-Lagrange’s equation associated with part (a) and if z(x) is an
arbitrary differentiable function for which z(x1) = z(x2) = 0
x2

then I    p 2 y z   q 2 yz dx  0
x1

b
12
3. Prove that the extremal of 
 y 1  y
2
 is the catenary y = a cos h(ax + b)
a

2
y 2
4. Prove that the extremal of  dx with y(0) = 0 and y(2) = 1 is a parabola.
0
x
5. Prove that the extremals of
x2

I   u  x  y 2   x  y 2  dx
x1

subject to the condition that

x2

J     x  y 2 dx  k (a constant)
x1

are the solution of Sturm-Liouville equation

d  dy 
 u  x      x     x   y  0, with y(x1) = y(x2) = 0
dx  dx 

137
 6. Show that the extremum of the functional
x2

I    y 2  y2  2 y sin x  dx,

x1

1
is given by y  c1e x  c2 e  x  sin x
2
7. Show that the Euler’s equation for the functional
b
f y
I   f  x, y  1  y 2 dx has the form f y  f x ,y  0
a 1  y 2
8. Find an extremal to
2
1  y2
I  dx, y 1  0 , y  2  1
1 x
[Ans. x2 + (y – 2)2 = 5]
9. Find the curve y = (x) which corresponds to the extreme value of
b 2
 dy 
F  y  x     x n   dx
a  dx 

c1 x1 n
[Ans. y   c2 , n  1  c1logx  c2 , n  1 ]
1 n
10. Show that the curve of shortest distance (geodesic) on a right circular cylinder is a Helix or a
generator.
b
11. Find the extremals of the functional F  y  x  ,z  x      2 yz  2 y  y2  z 2  dx
2

Deduce the extremals if a = 0, b =  ; y  0   0, y     1, z  0   0, z    1 .

[Ans. y   c1 x  c2  cosx   c3  c4  sinx z   c1 x  c2  2c3  cosx   c3  c4  2c1  sinx ]



138
Unit 8 : Functionals Dependent on Higher Order Derivatives
and Variational Problems in Parametric Form
Structure of the Unit
8.0 Objective

8.1 Introduction

8.2 Variational Problems Involving Several Higher Order Derivatives

8.3 Variational Problems Involving Functionals Dependent on the Functions of Several Independent

Variables and Dependent Variable

8.4 Variational Problems in Parameteric Form

8.5 Isoperimetric Problem

8.6 Summary

8.7 Answers to self-learning Exercise

8.8 Exercise

8.0 Objective

This unit deals with the functionals dependent on higher order derivatives and functions of more
than one independent variable. The variational problems in parametric form are also included in the present
unit.

8.1 Introduction

In the previous unit, we have discussed the Euler-Lagrange’s equation and various variational
problems having their first order derivatives. In this unit, we will disuss the variational problem with func-
tional dependent on higher order derivatives, several independent variables and variational problem in
parametric form.

8.2 Varitional Problems Involving Several Higher Order Derivatives

Theorem : If the function f contains higher order derivatives, say upto any order n, then


f  f x , y , y , ...., y  
n
 .....(1)

and we need to extremize the integral

x2
I 
x1
 
f x , y , y , ... y   dx
n
.....(2)

139
where we consider the funciton f is differentiable (n + 2)- times with respect to ‘x’. and also
assume that the boundary conditions are given by
y  x1   y1 , y   x1   y1 , y  x1   y1, ....., y  n   x1   y1 n

y  x 2   y2 , y  x2   y2 , y   x 2   y2 , ....., y  n  x 2   y 2 n  .....(3)

n
f d  f  n d  f 
Then I extremized by     ......   1 n  0 .....(4)
y dx  y  dx  y n 
Proof : Let the extremum is attained on the curve y = y(x) and y = y (x) be comparison curve
to extremizing curve y = y (x), and let both of these be 2n times differentiable.
Now we consider
y  x   y  x     x  , .....(4)
where   x1     x2      x1      x2   .....    n   x1     n   x2   0
Obviously y (x,0) = y(x), the extremizing curve.
Now substituting it in equation (1), we get
x2

 f  x, y , y,....., y  dx
 
n
I   .....(5)
x1

Since setting = 0 has the effect of replacing y , y ,....., y  n  in (5) by t he

y, y, y ,....., y   ,I  must take extreme value when = 0. This happens no matter what particular value
n

function (x) is involved in (4) and (5). But by elementary calculus, a necessary condition of extremum is
given by I () = 0 ....(6)
Using Leibniz’s rule of differentiation under integral sign, (6) gives.
x2
d
  n
I     d  f x, y , y,....., y dx
x1

Now using the chain rule for differentiating functions of several variables, we get

f y  
n
d f x f y
d

f x, y , y ,....., y  n    
x   y  
 .....   n  
y 
.....(7)

By using (4), we have

f f f
L.H.S. of (7)    x      x   ........   n    n   x  .....(8)
y y y
From (8), we get
x2
 f f f  n  

I   y
    x      x   ........    x   dx  0
y  
n
x1 
y 
which, upon setting = 0 and making use of (6), gives
x2

f f f   n 
  y   x   y    x   ........  y    x  dx  0
x1
n .....(9)

140
 where we have used the fact that when = 0, y  y, y   y ,....., y  n   y  n  .
Now integrating by parts, we have

x2 x2 x
f  f  2
d  f 
x y    y    x dx  y   x dx
  x   x 
1 x1 1

x2
d  f 
     x dx
x1
dx  y 
x2 x
x2
f  f  2
d  f 
and x y    x  dx =  y   x   x1  x1 dx  y    x 

1

x2
d  f 
    x 
x1
dx  y   .....(10)

Again integrating with respect to ‘x’ we get

x2
  d  f   x2
 d 2  f   
x2 y
    x  dx        x     2    x  dx 
x1 y    dx  y   x1 x1  dx  y   
x2 x
f 2
d2  f 
Thus, x y    x dx  x dx2  y    x  dx
 
1 1

x2 x2
f  n  n dn  f 
Similarly x y  n  dx   1 x dx n   n     x  dx
 y 
...(11)
1 1

Using it in equation (9), we obtain

x2
f d  f  d   f  n d  f 
n

 
x y dx  y  dx2  y   ..........    n  n   x  0
1
1 
dx  y 
which gives
f d  f  d   f  n d
n
 f 
    2   ...   1 n  0 .....(12)
y dx  y  dx  y  dx  y  n 

8.3 Variational Problem Involving Functionals Dependent on the Functions of

Several Independent Variables and Dependent Variables

In this section, we will discuss the variational problems which is dependent on several depen-
dent and independant variables.
Theorem : If z is a curve which is dependent on x, y and is twice differentiable in its
domain D, and extremize the functional
141
 

i  z  x , y     F  x , y , p, q  dx dy .....(1)
D

Then following differential equaiton must be satisfied

F   F    F 
    0 .....(2)
z x  p  y  q 
z z
where p , and q 
x y
Proof : Take some admissible surface z  z  x, y  close to z = z(x,y) and include the surfaces z

= z(x,y) and z  z  x, y  in a one-parameter family of surfaces

z  x, y,   z  x, y    z

where  z  z  x, y   z  x, y 

For = 0, we get the surface z = z(x,y), for = 1, we have z  z  x, y  z is called the varia-
tion of the fucntion z(x,y).
On fucntions of the family z = z(x,y,), the functional I reduces to the fucntion of  which has
an extremum for = 0. Hence, we have
  
  I  z  x , y ,      0
 0

The derivative of I [z(x,y,)] with respect to , for = 0 is known as the variation of the func-
tion and is denoted by I. Accordingly, we have

  
I  
 
 F  x, y,z  x, y,  , p  x, y,  ,q  x, y,   dx dy 
D  0

or     Fz z  Fp p  Fq q dx dy .....(3)

D

where z(x,y,) = z(x,y) +z

z  x, y, 
p  x, y,    p  x, y    p
x
z  x, y, 
and q  x, y,    q  x, y    q
y
Now, we have


 Fp z  
Fp
 z  Fp p  Fp p 

 Fp  z   Fp  z
x x x x


 Fq z  
Fq
 z  Fq q  Fq q 

 Fq  z   Fq  z
and
y y y y
Using above two results, we have
142
   
  Fp  p  Fq q   dx dy   
x

Fp z 
y
 
Fq z  dx dy 
D D  

 F p Fq 
      zdx dy .....(4)
D
x y 
where Fp/x is known as total partial derivative with respect to ‘x’. While computing it, y is assumed to
be fixed, but depenence of z, p and q upon x is taken account. Therefore, we have
Fp z p q
 Fpx  Fpz  Fpp  Fpq
x x x x
Fq z p q
Similarly  Fqy  Fqz  Fqp  Fqq
y x y y
Using the well-known Green’s theorem. We have
  
  x  Fp z   y  Fq z dx dy  C  Fp dy  Fq dx   z  0 .....(5)
D

The last imtegral is equal to zero, since on the contour C the variation z = 0 because all per-
missible surfaces pass through one and same spatial cantour C. Using (5), (4) reduces to
   
  Fp p  Fq q  dx dy    Fp  Fq  z dx dy
x y  .....(6)
D D
Using (6) in (3), it gives
   
 I   Fz z dx dy    Fp  Fq  z dx dy
D D
x y 
Hecne the neccessary condtion for I = 0 for an extremum of the functional (2) takes from
   
  Fz  x Fp  y Fq   z dx dy  0
D
Since the variation z is arbitrary and the factor is continuous, it follows from the fundamental
lemma of the calculus of variation that on extemizing surface z = z (x,y), we must have
  F   F    F 
Fz  Fp  Fq  0 that is     0 .....(7)
x y z x  p  y  q 
Remark . For the functional
I  z  x1 ,x2 ,....., xn      F  x 1 , x 2 ,.....x n , z , p1 , p 2 ,....., p n  dx 1 dx 2 .....dx n
D

z
where pi  , in exactly similar way, we get from the basic necessery conditon for extremum
xi
I = 0, the following equation
n

Fz   Fpi  0
i 1 xi

which the function z = z(x1,x2, ...., xn) extremizing the functional I must satisfy.
143
8.4 Variational Problems in Parameteric Ferm

In some problems, the requirement of single valuedness is excessively restrictive; for it turns out
that Euler-Lagrange’s equation-derived under assumption that the extremizing function is single valued-
may have for the solution which satisfies the given end point conditions, a relationship in which depen-
dent variable is not a single valued function of the independent variable. One cannot, without further
justificaiton, accept such a solution as valid.
We proceed to show, that the extremizing relationship between a pair of variables x and y is the
same, whether the solution is derived under the assumption that y is a single valued function of x or that
a more general parametric representation is required to express the relation between x and y. We do this
by showing that the solution of Euler-Lagrange equation derived on the basis of the assumption of the
single valuedness of y as a funciton x satisfies also the system of Euler-Lagrange’s equations derived on
the basis of the parametric relationship between x and y.
Under the assumption that y is a single valued funciton of x, the integral to be extremized is given
as
x2
I  f  x, y, y  dx .....(1)
x1

where y is required to have values y1 and y2 at x = x1 and x = x2. If instead, we use the parametric
representation x = x(t), y = y(t) where x(tj) = xj and y(tj) = yj for j = 1,2, the integral (1) transformed
to through the relationships
dy y
y   and dx = x dt .....(2)
dx x
where the supirior dot represents differntiation with respect ot ‘t’ .
t2
 y 
Therefore I   f  x, y, x  x dt .....(3)
t1

The Euler-Lagrange’s equation for (1) is

f d  f 
  0 .....(4)
y dx  y  
According to § 7.5, the system of Euler-Lagrange’s equation assiociated with (3) can be written
as
g d  g  g d  g 
     0,   0
x dt  x  y dt  y 

where  y   f  x, y, y  x
g  x, y,x, .....(5)
From (5), we obtain
g f g f y f
 
 x,  f  x  2  f  y .....(6)
x x x y  x y

144
 With the aid of second relation of (2), we obtain
d  g  d  f    f d  f   f 
   
x  f  y    
x  y       .....(7)
dt  x  dx  y    y dx  y    x 
Further, we obtain from (5)
g f g f 1 f
 
x,  x  .....(8)
y y y y  x y 
According to the second relation of (5), we have
d  g  d  f 
    x    .....(9)
dt  y  dx  y 
Combining this last result with the first of (8), we obtain the pair of equations
g d  g   g d  g  
     y     ,
x dt  x   y dx  y   
g d  g   g d  g  
     x     .....(10)
y dt  y   y dx  y   
From this result, we conclude that any relationship, single-valued or not, that satisfies the Euler-
Lagrange’s equaiton (4), derived on the basis of an assumed single valued solution y = y(x), satisfies
also the system (5), whose derivation requires no assumption of single-valuedness of y as function of x.

8.5 Isoperimetric Problem

In this section, we seek to derive the differentiable equation which must be satisfied by the funciton
which renders the integral
b
I   f  x, y, y   dx .....(1)
a

an extremum with respect to continuously differentiable functions y = y(x) for which the second integral.
b
J   g  x, y, y   dx .....(2)
a

possesses a given prescribed value, and with y(a) = y1, y(b) = y2 both prescribed boundary conditions
The given functions f and g are twice differentiable with respect to x.
To solve this type of problem, we will use the method of Lagrange’s multiplier. But first of all,
we need to choose suitable extremizing function for this problem. If we choose Y(x) = y(x) + (x)
which is a function of one perameter family. Then it yields the problem, because any change of the value
of the single perameter would in general alter the value of J, whose constancy must be maintained as
prescribed. For this reason we introduce the two perameter family
Y(x) = y(x) + (x) + (x) .....(3)
in which , and are arbitrary differentiable function for which (a) = (a) = 0 and (b) = (b) =
0. These conditions ensures that Y(a) = y(a) = y1 and Y(b) = y(b) = y2 as prescribed, for all values of
parameters and 
145
 We replace y by Y(x), given by (3), in both equations (1) and (2) so as to form respectively
b
I 1 ,2    f  x,Y ,Y   dx .....(4)
a

and J 1 ,2  =  g  x,Y ,Y   dx .....(5)

a

Clearly, the parameters and are not independent, because J is to be maintained at a con-
stant value, it is clear from (5) that there is a functional relation between them-namely,
J 1 ,2  = constant (prescribed) .....(6)
Now using, method of Lagranges multipliers, we introduce the function for , 
b
I 
 I 1 ,2    J 1 ,2    f   x,Y ,Y   dx .....(7)
a

where, according to (1) and (2),

f   f  g .....(8)
The constant  is the undetermined multiplier whose value remains to be determined by condi-
tions of each individual problem to which the method is applied. Thus for extemizing the value of I  , we
have

I  I 
  0, when  =   
 1  2  

From (7), with the help of (3), it follows that

b
I   f  Y f  Y  
    dx
 j 
a
Y   j Y    
j

b
 f  f   
  j   j  dx .....(10)
a  Y Y  
(j = 1,2)
Setting = = 0, so that according to (3), Y ,Y   is replaced by  y, y   , we thus have that
b
I   f  f   
  j   j  dx  0 (j = 1,2), .....(11)
 j a y y  
0

Note that the symbol 0 indicates that the setting of = = 0. Integrating by parts the sec-
ond term of the integrand of (11), we obtain with aid of boundary conditions that
b  f  d  f   
  j  y  dx  y  dx  0 (j = 1,2) .....(12)
a1   
146
 Now using basic lemma, we obtain the differential equation

f  d  f  
  0 .....(13)
y dx  y 
as the Euler-Lagrange’s equation which must be satisfied by the function y(x) which extrimizes (1) under
the restriction that (2) be maintained at a prescribed value.
Ex.1. Find the extremal of the functional
1

 1  y  dx ,
2
I =
0

under the conditions y  0  = 0 , y   0  = 1 , y 1 = 1 , y  1 = 1

Sol. In this problem,
f  x, y , y, y   1  y2
Therefore, the extremal function is given by solution of the following differential equation
f d  f  d 2  f 
    0
y dx  y   dx 2  y 

d2
 00  2 y  0
dx 2

d4y
 0 .....(14)
dx 4
The solution of differential equation (14) is
y = c1 + c2x + c3x2 + c4x3
Using the given conditions we easily obtain y = x
Thus extremal curve is a straight line.
Ex.2 . Find the extremal of the functional.
π
2
2
I  y  x   =   y - y 2 + x 2  dx,
0

 
y  0  = 1 , y   0  = 0 , y π 2 = 0 , y  π 2 = 1 .  
Sol. Comparing the given functional with
π
2
I  y  x   =  f  x, y, y, y dx
0

we get f  x, y, y, y    y 2  y 2  x 2 .....(15)

Using equation
f d  f  d 2  f 
  0 .....(16)
y dx  y  dx 2  y  
147
From (15) we get
f f f
 2 y,  0,  2 y
y y  y
So (16) reduces to
d2 d2y
2 y   2 y    0 or y0
dx 2 dx 4
d
or (D4–1)y = 0 where D  .....(17)
dx
The auxilliay equation of (17) is
m4 – 1 = 0 m = 1, i
Thus solution of (17) is y = c1ex + c2e–x + c3cosx + c4 sin x .....(18)
Using boundary conditions y(0) = 1, y(2) = 0, we get
c1+ c2+ c3 = 1 .....(19)
 
and c1e 2
 c2e 2
 c4  0 .....(20)

 
Since y   0   0 and y   2  1 therefore we find that
c1– c2+ c4 = 0 .....(21)
 
c1e 2  c2e 2
 c3  1 .....(22)
Adding (19) and (22), we get


c1 1  e

2
  c 1  e
2

2 
  0
and subtracting (20) from (21), we get


c1 1  e

2
  c 1  e   0
2

2

Above two relations give c1 = c2 = 0 and using it in (19) and (21), we get
c 4 = 0, c3 = 1
Hence extremum can be attained only on the curve y = cosx
Ex.3 . Find the extremal equation for the following functional

 z  2  z  2 
I  z  x1 , x 2   = 
D  x1    x2   dx1 dx2
 
Sol. Here the integrand f is a function of two independent variables x1 and x2 , i.e.
2 2
 z z   z   z 
F  z, , , x1 ,x2       .....(23)
 x1 x2   x1   x2 
Therefore, using the result

  F    F  F
     0 .....(24)
x1  p1  x2  p2  z
148
 z z
where p1  , and p2  ,
x1 x2

  z    z 
From (23) and (24), we obtain x  2 x   x  2 x   0
1  1 2  2

2 z  2 z
  0,
x12 x22
which is the familiar Laplace equation.
Ex.4 . Obtain the surface of minimum area, stretched over a given closed curve C, en-
closing the domain D in the xy plane.
Sol. From calculus, we know that the required given problem reduces to find the extremal of the
functional
1
  z  2  z 2  2

I  z  x, y     1        dx dy
x y
D       
1\2
Now we have F (x, y, z, p1, p2)  1  p12  p22   .....(25)

F   F    F 
      0, .....(26)
z x1  p1  x2  p2 

z z
where p1   zx and p2   zy
x y
(25) implies
F F 1 F 1
 0,  p1 1  p12  p22  2 ,  p2 1  p12  p22  2
z p1 p2
From (26), we have

   
  p1    p2 0

x  1  p 2  p 2 1 2  y  1 
 2  
2 2 2
1   1  p1  p 2 

   
  zx    zy 0
or x  1  z 2  z 2 1 2  y  1  .....(27)
 y  
2 2 2
x   1  z x  z y 
From (27), we get
1 1 3
z xx 1  z x2  z y2  2
 z x 1  z x2  z y2  2  2  z x z xx  z y z yx 
2
1 1 3
 z yy 1  zx2  z y2  2
 z y 1  zx2  z y2  2  2  zx zxy  2z y z yy   0
2
149
    z y2 
 1 z x2   z yy  1 
or z xx  
 1  z 2  z 2 1 2 1  z 2  z 2 3 2   1  z 2  z 2 1 2 1  z 2  z 2 3 2 
 x y x y   x y x y 

2z x z y z xy
 32
0
1  z 2
x  z y2 

or z xx 1  z y2   z yy 1  z x2   2 z x z y z xy  0

2 2
 2 z   z    2 z   z   z z  2 z
That is 1 
     1 
     2 0
x 2   y   y 2   x   x y xy

whose solution will yield the desired minimal surface.

Ex.5 . Find the closed convex curve of length L that encloses greatest possible area.
Sol. We know that the area of the closed plane curve is given by the integral
b
1 .....(27)
I   xy  yx  dt
2 a

dx dy
where x  dt , y  dt .
The total length of the curve is, given by
b 1
L    x 2  y 2  2 dt .....(28)
a

has the same value L where L is the length of the plane curve. Now the question is to maximize (extremize)
(27) under the restriction (28), We will use the equation (13) of (§ 8.5), which is given below :

f  d  f   f  d  f  
    0 ,   0 .....(29)
x dt  x  y dt  y 

1
where f   xy  yx    x 2  y 2 .....(30)
2
From (29) and (30), we have

1 d 1  x 
y    y  0
2 dt  2 x 2  y 2 


1 d 1  y 
 x   x  0
2 dt  2 x 2  y 2 

From which we obtain, by direct integration with respect to ‘t’,
 x  y
y  c1 , x   c2 ......(30)
x 2  y 2 x 2  y 2
150
 From these, we have
2 2  x 2  y 2 
 y  c1    x  c2    2  2 2
 2
 x  y 
Thus we have the well-known result “that the closed curve of given perimeter for which the
enclosed area is a maximum is a circle.”
Ex.6 . (Shape of hanging rope). Find the shape assumed by a uniform rope when sus-
pended by its end from two points. at equal heights.

Q(a, 0)
x

Fig. 8.1
Sol. Let the rope of lenght 2L be suspended between two points P(–a,0) and Q(a,0) in the
same straight line, as points are at equal heights.
Thus if  denotes the constant mass per unit length of rope, the potential energy of an element of
length ds at (x, – y) is given by (–gy ds) where g is the constant acceleration due to gravrity. Accord-
ingly, the total potential energy of the rope in the arbitrary configuration y = y(x) is given by

a a
I    g y ds   g  y 1  y 2 dx .....(31)
a a
where prime represents the differentiation with respect to ‘x’. and taking absolute value.
According to minimum energy principle the equillibrium configuration is supplied by particular
relation y = y(x) for which (31) is a minimum with respect to functions y(x) for which y(a) = 0, y(–a) =
0, and for which the total length of arc
a

J   1  y2 dx  2 L .....(32)
a
We may therefore apply the Euler-Lagrange equation to the integrand function
f    gy 1  y2   1  y2 .....(33)
formed from (31) and (32). Since f  is explicitly independent of the variable x, however, we may use
Euler-Lagrange equation and so substitute (30) into (13) (§ 8.5), we easily abtain.

 y2 
 gy      1  y2   c1
2 
 1  y 
151
 1 c1

 
2
1  y  gy  
2
 gy   
 1  y  2

c12
2
 gy   
 y 2 1
c12
2
 gy     c12
 y 
c12

c1 dy
 dx
 2 2
 gy    c 1

c1 1   gy   
Putting g y += c1cos ht and integrating, we find that  g cos h  c   c2  x
 1 
Solving we get
 c  g  x  c2 
y   1 cosh .....(34)
g g c1
where c2 is an arbitrary constant of integration.
Thus, according to (34), the shape of a hanging rope is that of a catanary with vertical axis. By
specifying that catenry passing through (–a,0) and (a,0) and that arc included between these points have
length 2L, we may assign value to constants c1, c2, . appearing in (34).
Ex.7 . Determine the curve of prescribed lengh 2l which joins the points (–a,b) and (a,b)
and has its centre of gravity as low as possible.
Sol. Let P1 P2 be an are joining the given paints (–a,b) and (a,b). The y-coordinate of the centre
of gravity of the required curve is given by
a

 y ds 1
a 1
I  a
a

2l a
y 1  y2   2
dx
 ds
a

where we have used the given constraint; namely

a 1
a a 1
 ds   1  y 
2
1
2
dx  2l, that is 
2l  a
1  y 2   2
dx  1 .....(35)
a a
The boundary conditions are y(–a) = b, and y(a) = b
y 1  1
Let F  x, y, y  
2l

1  y2  2

2l

1  y2  2

152
  y   1

2l
1  y  2 2

where  is the Lagrange’s multiplier. Since F does not contain x, thus from Euler-Lagrange’s equation
 F 
F  y   c (a constant)
 y  
12
 y    1  y2   y    y2
 c
or 2l 2l 12 .....(36)

1  y2 
 y    c1
or 12
1  y 2 
where c1 = 2cl. Re-writing the above equation we have
2 2
2
dy   y     c1 
2
2 
 y  

1  y or 2 
c12 dx 
 c1 
Separating variables and then integrating, we get

dy y
x  c1   c2 1
12 or x  c1cosh  c2
 y    2
 c12  c1

 x  c2 
So that y  c1cosh   .....(37)
 c1 
which is a complete solution of equation (36) on [–a,a] and boundary condition will be satisfied by this
solution if and only if
b  a  c2  b  a  c2 
 cosh   and  cosh  
c1  c1  c1  c1 
that is to say if and only if (a + c2)/c1 = (a – c2)/c1
x
Hence c2 = 0. Thus equation (37) reduces to y  c1cosh  c    .....(38)
 1
This shows curve must be symmetric with respect to y-axis. Thus, we get.
  c1cosh a/c1  b .....(39)
Using (38) in (35), we get
a
1 12

2l  a

1  sinh 2  x c1   dx  1

or  cos h  x c1  dx  2l
a

or 2c1 sin h(a c1 )  2l  l  c1 sin (a c1 ) 

153
 From (39), we have
12
 2
  c1 1  sinh  a/c1   b

12
 l 2 
 c1 1  2   b (using (40))
 c1 

12

 c12  l 2  b

Thus equation of the curve is given by

x 12
2
y  cosh    c1  l
c
 1

2
 b

Self-Learning Exercise

1.The possible value of  for which the functional

1
2
I  y  x     3 y  3 y   dy, y    1
0

can be extremized ?
(a) –1,0 (b) 0,1(c) –1,1 (d) –1,0,1
2. Find Euler-Lagrange’s equation for
x2

 F  x, y,z, y,z, y,z,....., y 

 k  ,z  k  dx
I 
x1

8.6 Summary

In this chapter, we obtain solution of some variabtional problems involving higher order deriva-
tives, some functional dependent on some dependent and independent variables. A number of problems
are included to illustrate various concepts of calculus of variation.

8.7 Answer to of Self-Leanning Exercise

(i) (b)

F d  F  d 2  F  k
k d  F 
(ii)       .......   1  0
y dx  y   dx 2  y  dx k  y  k  

F d  F  d 2  F  k
k d  F 
      .......   1  0
z dx  z   dx 2  z   dx k  z  k  

154
8.8 Exercise

1. Show that the Euler’s equation for the surface area functional
2 2
I  u    1  u x  uy dx dy


is 1  u  u  2u u u  1  u  u

y
2
xx x y xy x
2
yy 0
2. Find the Euler’s equation for the functional.
2 2
I   u x  u y  2 f  x, y  u  x, y   dx xy


where is a closed region in the xy-plane and u has continuous partial derivatives.

2 z  2 z
[Ans :   f  x, y  ]
x 2 y 2
3. Find the general solution of the extremals

 p 2 xy qx 2 y 2 
(i)   2  2  dx dy
D

  xyz  ypq  xp  dx dy
2
(ii)
D

where p  z x , q  z y [Ans : (i) z = c1(y) log x + c2 (y) + (x3/9)

(ii) z = c1(y) – {–c2(y)/2x2} + (yx2/15)]

4. Find the extremal for the functional

t2

  x 
12
I  x  t  , y  t    2
 y 2   a 2  xy  yx  dt
t1

where a being a constant. [Ans : circles]

5. Find the extremal of the functional
 4

  xy 
2
I  x  t  , y  t      2 x  2 y 2 dt,
0


subject to the initial conditions at t = 0, x = y = 0; at t  x = y = 1.
4
sin h 2t
[Ans. x  y  sin h  2 ]

6. Find the curve of length L that join the paints (0, 0) and (1, 0) lie above the x-axis, and encloses
the maxinmm area between itself and x-axis.

155
 12
1  1
[Ans.  x  c1  2 2
  y  c1    2
where c1  ,c2    2  
2  4

1  L 
and  is the solution of  sin  ]
2  2 
7. Find the extremals of the isoperimetric problem
1 1
 
I  y  x     y2  x 2 dx, given that y
2
dx  2 ; y  0   0 , y 1  0 .
0 0

[Ans. y = sin mx , m = 1,2,3]

8. Find the curve joining two points (x1,y1) and (x2,y2) that yields a surface of revolution of station-
ary area when revolved about the x-axis. [Ans. a circle]



156
Unit 9 : Series Solution of Second Order Linear Differential
Equation
Structure of the Unit
9.0 Objective

9.1 Introduction

9.2 Power Series Method

9.2.1 Validity of the Power Series Method

9.2.2 Definitions

9.2.3 Radius of Convergence

9.3 Series Solution Near an Ordinary Point

9.4 Series Solution Near a Regular Singular Point

9.4.1 Working Rule

9.5 Series Solution in Descending Powers of the Independent Variable

9.6 Self-Learning Exercise

9.7 Summary

9.8 Answers to Self-Learning Exercise

9.9 Exercise

9.0 Objective

The main object of this unit is to find the solution of a linear differential equation of second order
with variable coefficients in terms of a series near ordinary and singular points with special reference to
Gauss hypergeometric equation and Legendre equation.

9.1 Introduction

We know about the methods of solving linear differential equations of second order with con-
stant coefficients and in certain cases with variable coefficients. But sometimes, in case of variable coef-
ficients the problem becomes intricate and we are not able to find the solution in a closed form. Under
such situation, we can find a power series in terms of the independent variable x satisfying certain condi-
tions. This method is called the method of solution in series or integration in series. Legendre’s equa-
tion, Hypergeometric equation and Bessel’s equation are the examples whose solutions have been ex-
pressed in the form of a infinite power series eg. the general solution of y + y = 0 is y = a cos x + b
sin x and this may be rewritten as
157
  x 2 x 4   x3 x 5 
y  a 1    ...  b  x    ...
 2 4   3 5 
This shows that the general solution of the linear differential equation may be expressed by the
superposition of a pair of infinite series.

9.2 Power Series Method

The basic concept of power series method is simple and we will apply this technique to the so-
lution of some second order differential equations.
Let us consider the differential equation

d2y dy
P  x 2  Q  x  R  x y  0 .....(1)
dx dx
where P(x), Q(x) and R(x) are polynomial in x and P(x)  0.
The above equation may be written as
d2y dy
 p1  x   p2  x  y  0 .....(2)
dx 2 dx
Q  x R  x
where p1(x)  , and p2  x  
P  x P  x
To find the solution of the equation (1), we assume a series for y of the form

y = a0 + a1x + a2x2 + ...   ar x r .....(3)
r 0

dy d2y
Now substituting the values of y, and in equation (2) and rearranging the terms of
dx dx 2
different powers of x, we get an algebraic equation of the type
0 + 1x + 2x2 + ... = 0 .....(4)
Since equation (4) holds good for all values of x, identically, we obtain
0 = 0 , 1 = 0, 2 = 0 ,..., n = 0 ...
From these equations, we can determine the coefficients a0, a1, a2 ... etc. Putting the values of
a0, a1, a2, ... in the equation (3), we get the required solution which will be clear from the following
example.
Ex.1. Solve in series
d2y dy
 1  x  dx
2
2
 2x
dx
 2y  0

Sol. Let the solution of the equation be

y = a0 + a1x + a2x2 + ... .....(5)
dy
 = a1 + 2a2x + 3a3x2 + ...
dx
158
 d2y
and = 2a2 + 6a3x + 12a4x2 + ...
2
dx

dy d2y
Substituting the values of y, and 2 in the given differential equation and simplifying, we
dx dx
get
2a0 + 2a2 + 6a3x + (12a4 – 4a2)x2 + (20a5 – 10a3)x3 + ... = 0
Equating to zero, the coefficients of various powers of x, we obtain
a2 a
a2 = – a0, a3 = 0, a4    0 , a5  0
3 3
Substituting for as in equation (5), we get
 x4 
y  a1 x  a0 1  x 2    ....
 3 

which is the required solution.
9.2.1 Validity of The Power Series Method

In general an infinite series of the form


r 2
 ar  x  x0   a0  a1  x  x0   a2  x  x0   ....
r 0

is called a power aseries

Let us consider a differential equation

d2y dy
x2
dx 2 
 x2  x
dx

 2y  0

If we assume a solution of the form

y = a0 + a1x + a2x2 + ...
and solve the equation by the above method, we find that
a0 = 0, a1 = 0, a2 = 0, ...
This shows that the above equation has no series solution and if it is not so then what should be
the conditions under which the above equation admits of the series solution.

9.2.2 Definitions

The following definitions will help us in establishing the validity of the series methods.
(a) Ordinary and singular points
If P(x0)  0, then x = x0 is called an ordinary point of (1), otherwise a singular point. If
P(x0) = 0, then P1(x) and/or P2(x) become unbounbed as x0  0, such a point is called singular point of eq.
(1). For example, in the Legendre equation

d2y dy
1  x2  dx 2
 2 x
dx
 n  n  1 y  0,

the point x0 = 0 is an ordinary point because

159
 P  x0   1  x02  0 at x0 = 0,
while x0 =  1 are the singular points of the Legendre equation.
2
In Bessel’s equation x 2 d y  x dy  x 2  n2 y  0 clearly, x0 = 0 is a singular point and all
 
2 dx
dx
other points are ordinary points.
It is found that every solution of the eq. (1) at the ordinary point is analytic.
(b) Regular singular point
A singular point x = x0 of (1) is called regular if the following conditions are satisfied

Q  x
lim  x  x0   lim  x  x0  p1  x  = finite
x  x0 P  x x x0

2 R  x 2
and lim  x  x0   lim  x  x0  p2  x  = finite
x  x0 P  x x x0

For more general functions than polynomials, x0 is a regular singular point of equation (1) if the

Q  x 2 R  x
expressions  x  x0  and  x  x0  are analytic at x = x0, i.e., they have convergent Taylor’s
’s
P  x P  x
series expansion about x0.
where P(x), Q(x) and R(x) are polynomials in x and p1(x), p2(x) are defined by eq. (2).
(c) Irregular singular point
Any singular point of the equation (1) which is not a regular singular point is called an irregular
singular point. For example
(i) the differential equation

2 d2y dy
x  x  1 2
 2 x   x  1 y  0 has the singular points x0 = 0, x0 = 1 . It can be easily
dx dx
seen that x0 = 0 is a regular singular point as
2x
lim  x  0 p1  x   lim  x  0 2
0
x0 x 0 x  x  1
2  x  1
and
2
lim  x  0  p2  x   lim  x  0 2
0
x 0 x 0 x  x  1
whereas x0 = 1 is an irregular singular point, since

2x  2 
lim  x  1 p1  x   lim  x  1   lim 
2  does not exist.
x 1 x 1 x  x  1 x1  x  1 

(ii) the point x0 = 1 is a regular singular point of the Legendre equation

2
1  x  ddx y  2 x dy
2
2 dx
 n  n  1 y  0, since

 2 x 
lim  x  1 p1  x   lim  x  1 1
x 1 x1

1  x2 
160
 2 2 n  n  1
and lim  x  1 p2  x   lim  x  1 0
x 1 x1
1  x 
2

In a similar manner, it can be shown that x0 = –1 is also a regular singular point of the Legendre
equation.
(c) Radius of convergence
Whether x0 is ordinary or singular point, the power series method for solving the differential equa-
tion (1) is based on the idea of expressing y as intinite series in powers of (x – x0). Here note that only
convergent series will yield desired solutions, if it exist.

r
A power series  ar  x  x0  is said to converge at a point x, if
r 0

m
r
lim
m
 ar  x  x0  exists
r 0
Obviously if the series converges for x = x0 it may converge for all x or only for some values of x
for which the convergence tests studied in Real analysis may be used.

r
If there exists a number R  0, such that  ar  x  x0  converges absolutely for | x – x0 | < R
r 0

and diverges for | x – x0 | > R, the number R is caled the Radius of convergence of the series.
For a series that converges no where except at x0, the radius of convergence is said to be zero. If

ar
it converges for all x, we say that radius of convergence is infinite. Also note that R  lim , pro-
r  ar 1
vided the limit exists.

9.3 Series Solution Near an Ordinary Point

If x = x0 is an ordinary point of the equation (1), then each solution can be expressed in the form

y   an  x  x0 n  a0 y1  x   a1 y2  x  ,
n0
where a0 and a1 are arbitrary constants and y1 and y2 are linearly independent series solutions which are
analytic at x0.
Following examples will make the method more clear.

d2 y dy
Ex.1. Solve in series 2  x2
 dx 2
 2x
dx
 2 y  0.


Sol. Since x0 = 0 is an ordinary point i  e P  x0   2  x02  0 at x0  0 , we assume the solu- 
tion in the form
 
y   ar  x  0  x r   ar x r
r 0 r 0

161
 dy d2y
Substituting for y, and 2 in the given equation, we have
dx dx
     
 
2  x 2   ar r  r  1 x r  2   2 x   ar rx r 1   2   ar x r   0
 r 0   r 0   r 0 
 
or 2  ar r  r  1 x r 2   ar  r  1 r  2  x r  0
r 0 r 0

Equating to zero, the coefficient of the smallest power of x i.e. xr – 2, we get

2 ar r(r – 1) – ar – 2(r – 3) (r – 4) = 0

or ar 
 r  3 r  4  a , r  2
r 2
2r  r  1

This gives a 2  a0 , a3 = 0; a4 = 0; a5 = 0; a6 = 0, .....

2
This shows that all the coefficients beyond a2 are zero.
Hence the solution of the given equation is given by
y = a0 + a1x + a2x2
 x2 
or y  a0 1    a1 x.
 2 

Ex.3. Solve the Legendre’s equation

d2y dy
 1  x2  dx 2
 2x
dx
 n  n  1 y  0.

 
Sol. Since x0 = 0 is an ordinary point i.e. P  x0   1  x02  0 at x0  0 , therefore we may as-

sume the solution in the form

 
r
y  ar  x  0    ar x r .....(1)
r 0 r 0
 2 
dy r 1 and d y
so that   ar r x   ar r  r  1 x r 2
dx r 0 2
dx r 0
Putting these values, in the given equation, we get
      
1  x    a
2

r 0
r r  r  1 x r 2   2 x   ar r x r 1   n  n  1   ar x r   0
  r 0   r 0 
 
r 2
or  ar r  r  1 x   ar  r  n  r  n  1 x r  0
r 0 r 0

Equating to zero, the coefficient of xr the recurrence relation is given by

ar + 2(r + 2)(r + 1) – ar (r – n) (r + n + 1) = 0

162
 or ar + 2 
 r  n  r  n  1 a , where r = 0, 1, 2 ..... .....(2)
 r  1 r  2  r
The relation (2) gives even and odd coefficients in terms of the one immediately preceding it,
except for a1 and a2 which are arbitrary.
From (1), we find that

a2  n  n  1 a
0
2 1

a4   2  n  2  n  1 a
2
3 4

or a4   n  2  n  n  1 n  3 a
0
4  3  2 1

and a3    n  1 n  2  a
1
3 2

a5    n  3 n  4  a
3
5 4

a5  
n  1 n  3 n  2  n  4 
or a1
5  4  3  2 1
and so on.
Putting these coefficients in (1), the solution of the given equation can be written as

 n  n  1 2  n  2  n  n  1 n  3 4 
y  a0 1  x  x  ...
 2 4 

  n  1 n  2  3  n  1 n  3 n  2  n  4  5 

a1  x  x  x  ...
 3 5 
 y = a0 y1(x) + a1 y2(x).

9.4 Series Solution Near a Regular Singular Point

If x = x0 is a regular singularity of the equation (1) (§9.2), then at least one of the solutions can
be expressed as
 
m r m r
y   x  x0   ar  x  x0    ar  x  x0  .....(1)
r 0 r 0

where ‘m’ may be a positive or negative integer or a fraction and is called the index of the series solu-
tion. This method of solution was suggested by George Frobenius (1849–1917) and is called Frobenius
method. We now discuss the method of solving equation (1) in the neighbourhood of a regular singular
point x = x0. Without loss of generality, we can take x0 = 0. If x0  0, we can transform the equation by
letting x = x0 = z.
163
 Since x0 = 0 is a regular singular point of the equation (1), its solution can be expressed in the
following form
 
y  x m  ar x r   ar x m r , where a0  0 .....(2)
r 0 r 0

9.4.1 Working Rule :

dy d2y
(i) Substitute the value of y, and 2 in the given differential equation.
dx dx
(ii) Rearrange the terms in powers of x and equate to zero the coefficient of lowest power of x.
This gives us a quadratic equation in m which is called the indicial equation.
(iii) Solve the indicial equation. The following cases arise :
(a) The roots of the indicial equation are different and not differing by an integer.
(b) The roots of the indicial equation are equal.
(c) The roots of the indicial equation are different, differing by an integer and also making a
coefficient of y infinite.
(d) The roots of the indicial equation are different, differing by an integer and making a
coefficient of y indeterminate.
(iv) We equate to zero the coefficient of general power of x (e g. xm + r or xm + r – 1 whichever
may be the lowest) in the equation obtained in step (ii). The equation so obtained will be
called the recurrence relation, because it connects together the coefficients am, am – 2 or
am, am – 1 etc.
(v) If the recurrence relation connects am and am – 2, then we, in general, determine a1 by equat-
ing to zero the coefficient of the next higher power. On the other hand, if the recurrence rela-
tion connects am, am – 1, this step may be omitted.
(vi) With the help of the recurrence relation all the as are determined in terms of a0 and these as
will be put in eq. (2). Then replacing m by m1 and m2 and a0 by a and b respectively, we shall
obtain two independent solutions, say au and bv. Therefore the complete solution of the given
differential equation is given by
y = au + bv, where a and b are arbitrary constants.
The method is illustrated with the help of following examples %
Case I. When the roots m1, m2 of the indicial equation are different and not differing by
an integer, the complete solution is
y  c1  y  m  c2  y  m
1 2

where c1 and c2 are arbitrary constants

d2 y dy
Ex.1. Solve in series 2 x 2
dx 2
x
dx
 
 1  x 2 y  0.

Sol. Here x0 = 0 is a regular singular point as

164
 Q  x   x  1
lim  x  0   lim  x  0  p1  x   lim  x  0   2    finite
x 0 P  x  x 0 x 0  2x  2
R  x 2
2 2 2  1 x  1
and lim  x  0   lim  x  0  p2  x   lim  x  0      finite
x 0 P  x  x 0 x 0  2x  2
2
 
therefore we assume the series solution in the form

y   ar x m r , a0  0 .....(3)
r 0

dy d2y
Substituting for y, and 2 in the given equation, we find that
dx dx
     
  r 0 
 
2 x 2   ar  m  r  m  r  1 x mr 2   x   ar  m  r  x m r 1   1  x 2   ar x m r   0
 r 0  r 0 
 
m r
or  ar  m  r  1 2m  2r  1  x   ar x m r 2  0 .....(4)
r 0 r 0

which is an identity. Now equating to zero, the coefficient of smallest power x i.e. xm (put r = 0 in the
first summation) then the equation (4) gives the indicial equation or quadratic equation in m as
a0(m – 1) (2m – 1) = 0
which implies that m = 1, 1/2 as a0  0
so the roots of the indicial equal are different and not differing by an integer.
To obtain the recurrence relation, we equate to zero the coefficient of xm + r and obtain
1
ar  a .....(5)
 m  r  1 2m  2r  1 r 2
This formula connects ar with ar – 2. Now we proceed to find a1 as explained in step (v) of
§ 9.4.1. For this purpose, we equate to zero, the coefficient of next higher power of x i.e. xm + 1
(put r = 1 in the first summation), we get
a1[m(2m + 1)] = 0
 1
Since the quantity within the bracket is not zero for any above values of m  1 or  , this gives
 2
a1 = 0
Since a1 = 0, then from (5), we have a3 = a5 = ... = 0.
Also taking r = 2, in (5), we get

1
a2  a .....(6)
 m  1 2m  3 0
Next taking r = 4, in (5) and using (6), we obtain

1
a4  a
 m  1 m  3 2m  3 2m  7  0
and so on.
165
 Putting these values in (3), i.e. y = xm[a0 + a1x + a2x2 + a3x3 + ...] gives
 x2 x4 
y  a0 x m 1    ... .....(7)
  m  1 2m  3  m  1 m  3 2m  3  2m  7  
Putting m = 1, and replacing a0 by a in (7), we get
 1 2 1 
y  ax 1  x  x 4  ...  au (say)
 25 2  459 
Next putting m = 1/2, and replacing a0 by b, we obtain
 x2 x412 
y  bx 1    ...  bv (say)
 23 234 7 
Therefore the complete solution is given by
y = au + bv,
where a and b are arbitrary constants.
Ex.2. Solve the Gauss hypergeometric equation

d2y dy
x 1  x      1      x   y  0
2 dx
dx
in series in the neighbourhood of the regular singular point (i) x = 0 (ii) x = 1 and (iii) x = .
Sol. Given
d2y dy
x 1  x     1      x   y  0 .....(8)
2 dx
dx
Dividing by x(1 – x), we get

d2y

  1      x dy


y0
dx 2 x 1  x  dx x 1  x 
Comparing it with y + p1(x) y + p2(x) y = 0, we have

p1(x) 
  1      x
x 1  x 

and p2(x) 
x 1  x 
Since x p1(x) and x2 p2(x) both tends to a finite value at x = 0, so x = 0 is regular singular point
of (8).
Case I. Solution in the neighbourhood of x = 0.
We assume that the given equation (8) has the solution of the form

y  ar x m  r , a0  0 .....(9)
r 0
Substituting the values of y, yand yin the given equation (8), we get
 
 
x  x 2   ar  m  r  m  r  1 x m r 2  
 r 0 
166
  
  1      x   ar  m  r  xmr 1     ar x mr   0
 r 0 
 
or  ar  m  r  m  r  1    x mr 1   ar  m  r    m  r    x mr  0 ....(10)
r 0 r 0
which is an identity. Equating to zero, the coefficient of the smallest power of x i.e. xm – 1 (put
r = 0 in the first summation), we get the indicial equation as
a0m(m – 1 + ) = 0, a0  0
This gives m = 0, 1 – 
To obtain the recurrence relation, we equate to zero the coefficient of x m + r – 1. Then we have
ar(m + r) (m + r –1 + ) – ar –1(m + r –1 + ) (m + r –1 + ) = 0

or ar 
 m  r  1    m  r  1    ar 1 .....(11)
 m  r  m  r  1   
For the solution corresponding to m = 0, the recurrence relation (11) reduces to

ar 
 r  1    r  1    ar 1
r r 1  
from which it follows that
 
a1  a0 ,
1 
1   1    a   1     1    a
a2  2  1  
  1 1  2 1   
0

and so on.
Putting these values and m = 0 and replacing a0 by a in (2) gives
    1     1    2 
y  a 1  x x  .... .....(12)
 1  1  2 1    
If we take a = 1 in (12), the series on the right hand side of (12) is called the hypergeometric
series and is represented by 2F1 (, , ; x). Thus we see that 2F1 (, , ; x) is a solution of (8).
For the solution corresponding to m = 1 – , when 1 –  is neither zero nor an integer, the
recurrence relation (11) reduces to.

ar 
1    r  1   1    r  1    a
r 1
1    r 1    r  1   
   r  1    r  1 a
or ar  r 1 .....(13)
r     r  1
where = 1 –  +,  = 1 –  + ,  = 2 –  .....(14)
Replacing r = 1, 2, 3, .... successively in (13), we have
  
a1  a0
1  
167
 a2 
   1    1 a       1       1 a etc.
1 0
2     1 1  2        1
Hence putting m = 1 – ..., using the above values of a1, a2 ... in (9) and replacing a0 by b gives
        1       1 2 
y  bx1 1     x  x  .... .....(15)
 1   1  2      1 
If we take b = 1 in (15), the series on the right hand side of (15) would be
x1 – 2F1 (, ; ; x) i.e. x1 –  2F1 (1 –  + ,1 –  +; 2 –  ; x)
which is another independent solution of (8).
Hence the general solution of (8) is
y = a 2F1 (, ; ; x) + bx1– 2F1 (1 –  + ,1 –  +; 2 –  ; x) ....(16)
which a and b are arbitrary conatants.
Case II. Solution in the neighbourhood of x = 1.
It can be easily see that

lim  x  1 p1  x   lim  x  1
  1      x 
finite value
x1 x1 x 1  x 

2 2    0 
and lim  x  1 p2  x   lim  x  1 finite value
x1 x1 x 1  x 
so x = 1 is also a regular singular point of (8).
If we substitute  = 1 – x in the equation (8), it reduces to

d2y dy
 1    2
       1      1     y  0 .....(17)
d d
On comparing (8) and (17), we find that (17) is the same as (8) except that  is replaced by
+  – +1 and x by .
Hence the solution (16) of (8) near x = 0 will be valid for (17) near  = 0, i.e. near x = 1.
Hence in this case, the required solution will be
y = A2F1 (, ; +  –  + 1; 1 – x)
+ B(1 – x)– +  2F1 (  – , –; – – + 1; 1 – x) .....(18)
where  –– is neither zero nor an integer
Case III. Solution in the neighbourhood of x = .
To find the solution of the given hypergeometric differential equation (8) for large values of the
independent variable i.e. about x = , we change the independent variable from x to t with the help of
the following transformation x = 1/t i.e., t = 1/x .....(19)
Clearly large values of x correspond to small values of t. Using the above equation (19), we
rewrite (8) and obtain the transformed equation near t = 0, say

d2y dy
 p1  t   p2  t  y  0 .....(20)
dx 2 dx
168
 Then the given equation (8) is said to have a regular singular point at x =  if the transformed
equation (20) has regular singular point at t = 0.
1 1 dt 1
For x  or t  ,  .....(21)
t x dx x 2

dy dy dt dy  1  dy
and y     2   t 2 .....(22)
dx dt dx dt  x  dt

d2y d  dy  d  dy  dt 2
4d y dy
Also y         t  2t 3 .....(23)
dx 2 dx  dx  dt  dx  dx dt 2 dt
Using (21), (22) and (23), the given equation (8) transforms to

2 d2y dy
t  t  1  2  t  1   t      1 t   y  0 .....(24)
2 dt
dt
To solve (24), let its series solution be

y  ar t mr , a0  0 .....(25)
r 0

dy  d2y  m r 2
so that   ar  m  r  t m  r 1 and 2   ar  m  r  m  r  1 t
dt r 0 dt r 0

dy d2y
Putting these values of y, and 2 in (24), we get
dt dt


 t3  t 2  r0 ar  m  r  m  r 1 t mr 2
 
 2  t  1   t      1 t  ar  m  r  t mr 1    ar t mr  0
r 0 r 0

 
mr
or  ar  m  r    m  r    t   ar  m  r  m  r  1    t m r 1  0 .....(26)
r 0 r 0
which is an identity. Equating to zero, the coefficient of the smallest power of t (put r = 0, in the first
summation), we get
a0(m –) (m – ) = 0 m = ,  as a0  0
Next equating to zero, the coefficient of tm + r + 1 in (26), we find that

ar + 1 
 m  r  m  r  1    a .....(27)
 m  r  1    m  r  1    r
For the solution, corresponding to m = , the recurrence relation (27) reduces to

ar + 1 
  r   r  1    a
 r  1  r  1    r
169
    1   
from which it follows that a1  a0
1    1   

a2 
  1  2    a     1  1     2    a
1 0
2    2    1  2   1     2   
and so on.
Putting these values and replacing a0 by A in (25), gives

  1         11     1      1 2 
y  At  1  t t  ...
 1  1      1  2 1     1      1 

   1      tk
 At  1k     k
k 0 k
k

 k
 1     k 1      k 1 1
or y  A    
 x  k 0 1      k k  x

1
or y  Ax  2 F1   ,1     ; 1     ;  .....(28)
 x
By symmetry for m = , we get

1
y  Bx   2 F1   ,1     ;1     ;  .....(29)
 x
Therefore the complete solution of the Gauss hypergeometric equation when  –  is neither
zero nor an integer, is given by
 1  1
y  Ax  2 F1   ,1     ;1     ;   Bx   2 F1   ,1     ;1     ; 
 x  x
Case II. When the roots m 1, m 2 of the indicial equation are equal, the complete

 y 
solution is y  c1  y  m1  c2   .
 m  m1
This case is illustrated in the following example :

d2y dy
Ex.3. Solve in series x  1  x   1  5 x  4y  0
2 dx
dx
Sol. Since x0 = 0 is a regular singular point therefore we assume that the solution is of the form

y  ar x m  r , a0  0 .....(30)
r 0

dy d2y
Putting the values for y, and in the given equation and rearranging the terms, we get
dx dx 2
170
  
2 2
 ar  m  r  xmr 1   ar  m  r  2  x mr  0 .....(31)
r 0 r 0
Equating to zero, the coefficients of lowest power of x, the indicial equation gives
a0m2= 0 m = 0, 0 as a0  0.
Since both the values of m are equal so it gives us only one independent solution. Equating to
zero, the coefficient of xm + r, we find that
2
mr 2
ar + 1    ar .....(32)
 m  r 1 
2
Which gives a1   m  2  a0
 m 1 
2 2
a2   m  3  a1   m  3  a0
m2  m 1 
and so on.
Hence the solution is given by

 2 2 2 
m2  m3 2  m 4 3
y  a0 x m 1    x    x    x  ... .....(33)
  m  1   m 1   m 1  
Putting m = 0 and replacing a0 by a in (33) gives
y = a [1 + 22x + 32x2 + 42x3 + ...] = au (say) .....(34)
To get the second solution, we procecd as follows :
Rewriting (33)

 m  m  2 2 m1  m  3  2 m2 
y  a0  x    x   x  ....
  m 1   m 1  
which on differentiation with respect to x gives

dy  m1  m  2 2 m  m3
2
m 1

 a0  mx    m  1 x     m  2 x  .... and
dx   m 1   m 1  

2 2
d2y  m 2  m  2  m 1  m  3  m

 a0  m  m  1 x     m  1 mx     m  2  m  1 x  ....
dx 2   m 1   m 1  

dy d2y
Putting the values of y, and in the left hand side of the given equation, we get
dx dx 2

 2 2 
m 2  m  2  m1  m  3 
x  x 
2
a0  m  m  1 x


 m 1 
 m  m  1 x   
 m 1 
 m  1 m  2  x m
 ...


171
  2 2 
m2 m  m3
 1  5 x  a0  mx m1     m  1 x     m  2  x m1
 ...
  m 1   m 1  

 m  m  2  2 m1  m  3 2 m2 
4a0  x    x    x  ...  a0m 2 x m1
  m  1   m  1  

The coefficient of remaining powers of x being zero, it can be easily verified by considering the
coefficients one by one.
Thus we may write

d2y
 x  x2  dx 2  1  5x  dydx  4 y  a0 m2 xm1
which on partial differentiation with respect to m, gives
2
  2 d d 
m 

 xx
dx 2
1  5 x 
dx
 4  y  2a0 mx

m 1
 a 0m 2 x m1 log x

Since the operators are commutative, therefore the above relation may be rewritten as
 2  y
2 d d


 x x
dx 2 
 1  5 x 
dx
 4 
  m
 2a0 mx m1  a 0 m 2 x m1 log x

Putting m = 0, we get

 2   y 
2 d d


 xx
dx 2
1  5 x   4  
dx   m m0

 y 
which shows that   is a second solution of the given differential equation.
 m  m  0
Hence differentiating (33) partially with respect to m, we get

y   m  2 2  m3 2
2 
 a 0 x m log x 1    x    x  ...
m   m  1   m 1  

 m2  1  m  2    m  3  2 
   m  3   1
 a0 x m  2     x  ...  2     x  ...
  m  1    m  1  m  12   m  1    m  1  m  12  
     
Putting m = 0 and replacing a0 by b gives

 y 
 m   b log x 1  22 x  32 x 2  ...  2b  2 1  2  x  3 1  3 x 2  ....
 m 0    

 y 
  m 
  m 0   
 b u log x  2 1 2 x  2  3x 2  ...   bv,

(say)

172
 Thus the required solution is
y = au + bv,
where a and b are arbitrary constants.
Case III. When the roots m1, m2 (m1 > m2) of the indicial equation are different and
differing by an integer and also making a coefficient of y infinite.
Working Rule. If the indicial equation has unequal roots, say m1 and m2 (m1 > m2) differing by
an integer and if some of the coefficients of y become infinite when m = m2, we modify the form of y by
replacing a0 by d0(m – m2) where d0  0. Then two independent solutions can be obtained by putting
y
m = m2 in the modified form of y and . In this case the solution by putting m = m1 in y is rejected
m
because it only gives a numerical multiple of the solution obtained by putting m = m2 in modified y. Thus
the complete solution is
 y 
y  c1  y m  c2  
2  m  m2

d2 y dy
Ex.4. Solve x2
dx 2
x
dx
 
 x 2  1 y  0 in series.

dyd2y
Sol. Given
dx
x2
dx
 x2 1 y  0
2
x   .....(35)

Since x = 0 is a regular singular point as x p1(x) and x2 p2(x) tends to a finite limit as x  0,
therefore we assume the solution of the given equation (35) in the form

y  ar x m  r , a0  0
r 0

 
then y   ar  m  r  x m r 1, y   ar  m  r  m  r  1 x m r 2
r 0 r 0
Substituting for y, y and y in (35), then it gives
  
x 2
 ar  m  r  m  r  1 x
r 0
mr 2
 x  ar  m  r  x
r 0
m  r 1
 2
 x 1  r0 ar xmr  0
 
or  ar  m  r  m  r  1   m  r   1 x m r   ar x mr 2  0
r 0 r 0

 
mr
or  ar  m  r  1 m  r  1 x   ar x m  r  2  0 .....(36)
r 0 r 0

which is an identity. Equating to zero, the coefficients of the smallest power of x, namely xm (put r = 0 in
the first summation), gives the indicial equation
a0(m + 1) (m – 1) = 0
so that m= 1, – 1 as a0  0 .....(37)
173
 The roots given by (37) are different and differing by an integer.
To obtain the recurrence relation, we equate to zero, the coefficient of xm + r and obtain
ar(m + r +1) (m + r –1) + ar – 2 = 0
1
or ar  a .....(38)
 m  r  1 m  r  1 r 2
[Since (38) gives the relationship between ar and ar – 2, we proceed to find a1 as explained in
step (v) of § 9.4.1]
Equating to zero, the coefficient of xm + 1 in (36) (put r = 1 in the first summation), we find that
a1(m + 2)m = 0, giving a1 = 0
Since the quantity within the bracket is not zero for any above values of m.
From (38) and a1 = 0, we have
a1 = a3 = a5 = a7 = ... = 0
Further, taking r = 2 in (38), we get
1
or a2   a0 .....(39)
 m  3 m  1
For r = 4, in (38) and using (39), we find that
1 1
a4   a2  a0
 m  5 m  3  m  1 m  32  m  5 

Putting these values in y   ar x m r , we get
r 0

 1 1 
y  a0 x m 1  x2  x 4  ... .....(40)
  m  1 m  3  m  1 m  32  m  5  

Since the factor (m + 1) appears in the denominator, the coefficient of y will be infinite for
m = –1.
To overcome this difficulty, we put a0 = d0(m + 1), of course the condition a0  0 is now vio-
lated, therefore we assume in its place d0  0. The above equation (40) becomes

 x2 x4 
y  d0 x m  m  1    ... .....(41)


 m  3  m  32  m  5  
Putting m = –1 and replacing d0 by a, we get

 1 1 
y  ax 1   x 2  2 x 4  ...  au (say) .....(42)
 2 2 4 

 y 
The obtain another solution, m = –1 will be substituted in   obtained from (41).
 m 

174
 y  x2 x4 
Now  d0 x m log x  m  1    ...
m 

 m  3  m  32  m  5  

 x2  2 1  
 d 0 x m 1      x 4
 ...
  m  3 2   m  3 3  m  5   m  3 2  m  52  
   
Putting m = –1, replacing d0 by b, the second solution will be obtained as

 y   1 1 
 m   bx 1 log x   x 2  2 x 4  ...
  m1  2 2 4 

 1 1 2 1 
bx 1 1  2 x 2  2    x 4  ...
 2 2 4  2 4  

1
 x2 5 4

 bu log x  bx 1  2  2 2 x  ...
 2 2 4 

 2
 x 5 
 b u log x  x 1 1  2  2 2 x 4  ... .....(43)
  2 2 4 

= b v (say)
Hence the complete solution of the given differential equation is
y = au + bv.
1
Note : If we substitute m = 1 and d0  in (41), we get
2

 1 2 x4 
y  x 1  x   ... 
 2  4 2  42  6 

 x3 x5 
y  x    .   2u
2
 2  4 2  4  6 
which gives no new independent solution.
Case IV. When the roots m1, m2 of the indicial equation are different and differing by
an integer and also making a coefficient of y indeterminate.
Working Rule. If the indicial equation has two different roots say m1, m2 (m1 > m2) differing by
an integer and if one of the coefficients of y become indeterminate when m = m2, the complete solution
is given by putting m = m2 in y, which contains two arbitrary constants. In this case, the solution obtained
by putting m = m1 in y is rejected because it only gives a numerical multiple of one of the series con-
tained in the first solution.
175
 d2 y
Ex.5. Solve x 2
dx 2

 x  x2  dxdy   x  9  y  0 in series.
Sol. Since x0 = 0 is a regular singular point, we assume the solution in the form

y  ar x m  r , a0  0
r 0

   
then
 r 0 

x 2   ar  m  r  m  r  1 x m r 2   x  x 2    ar  m  r  x
 r 0
m  r 1



 
  x  9    ar x m  r   0
 r 0 
 
or  ar  m  r  m  r  1   m  r   9 x mr   ar  m  r  1 x mr 1  0
r 0 r 0

 
or  ar  m  r  3 m  r  3 xmr   ar  m  r  1 x mr 1  0
r 0 r 0
which is an identity. Equating to zero, the coefficient of the smallest power of x, namely xm
(putting r = 0 in the first summation), we get
a0(m – 3) (m + 3) = 0, m = 3, –3 ( a0  0)
The roots of the equation are different and differing by an integer. To obtain the recurrence rela-
tion, we equate to zero, the coefficient of the general term i.e. xm + r, we get
ar(m + r + 3) (m + r – 3) + ar – 1(m + r) = 0

m  r 
or ar  ar 1 .....(44)
 m  r  3 m  r  3
  r  3
Taking m = –3, we get ar  ar 1
r r  6

2
Thus for r = 1, we have a1  a0 and for r = 2, 3, 4, 5, and 6 we have
5
1 2 1
a2   a1   a0
8 5 8
a3 = 0, a4 = 0, a5 = 0 and

  6  3 0
a6  a5  (inderminate)
6  6  6 0
and may be taken as a free constant
4 5 45
Also a7  a6 and a8  a7  a6
7 16 7  16
and so on.
176
 
 y   ar x m r  x m a0  a1x  a2 x 2  a3 x 3  ...
r 0

y  x 3 a0  a1x  a2 x 2  a6 x 6  a7 x 7  a8 x8  ...

 

 2 1 2   4 4 5 
 x 3  a0  a0 x   a0 x 2   x 3  a6 x 6  a6 x 7  a6 x8  ...
 5 8 5   7 7 16 

 2 2 1   4 4 5 2 
 y  a0 x 3 1  x   x 2   a6 x3 1  x  x  ...
 5 5 8   7 7 16 
This contains two arbitrary constants a0 and a6 and therefore may be taken as the complete so-
lution
Note. If we put m = 3 in (44), we get a series solution
 4 4 5 2 
y  a0 x3 1  x  x  ...
 7 7 16 
which gives no new independent solution.

9.5 Series Solution in Descending Powers of the Independent Variable

Till now we have obtained series solutions in ascending powers of the independent variable. How-
ever, the following cases may arise.

(i) There exists no solution of the form  ar x m  r .
r 0
(ii) The usual Frobenius method breaks down.
(iii) The series solution obtained by earlier methods does not converge.
In such cases we obtain the series solution in descending powers of the independent variable.
Sometimes, the series solution in descending powers are desirable and are more useful in practice.
Working Rule

(i) We assume a solution of the form y   ar x mr , a0  0
r 0
(ii) For indicial equation, we equate to zero the coefficient of the highest power of x in the iden-
tity.
(iii) For recurrence relation, the coefficient of the higher power, in general, in the identity is equated
to zero.
To illustrate the method we consider following examples :
Ex.1. Integrate in descending series the Legendre’s equation or determine the solution of
Legendre’s equation.
Sol. The differential equation of the form
(1 – x2) y – 2x y + n(n + 1) y = 0 .....(1)
177
is called the Legendre’s equation, where n  N. Let the series solution of (1) be of the form

y  ar x mr , a0  0 .....(2)
r 0
Substituting the values of y, y and y in the given equation, we get
  

1  x2  r 0
 ar  m  r  m  r  1 xmr 2  2 x  ar  m  r  x mr 1  n  n  1  ar x mr  0
r 0 r 0

 
or  ar  m  r  m  r  1 xmr 2   ar  m  r  n  m  r  n  1 x mr  0 .....(3)
r 0 r 0
which is an identity. Equating to zero, the coefficient of the highest power of x, namely xm, (put
r = 0 in the second summation), we get the indicial equation
a0(m – n) (m + n + 1) = 0
Since a0  0  m = n, – (n + 1)
which shows that the roots are different.
To obtain the recurrence relation, we equate to zero the coefficient of xm – r and obtain
ar –2(m – r + 2) (m – r + 1) – ar(m – r – n) (m – r + n + 1) = 0

or ar 
 m  r  2  m  r  1 a .....(4)
 m  r  n  m  r  n  1 r 2
Here we need to evaluate a1. It can be done by equating to zero, the coefficient of the next
lower power of x i.e. xm – 1, which gives
a1(m – 1 – n) (m + n) = 0
 a1 = 0, since the quantity within the bracket is not zero for any above values of m
Since a1 = 0, then from (4), we have a3 = a5 = .... = 0

m  m  1
Also a2  a0
 m  n  2  m  n  1

a4 
 m  2  m  1 a
 m  n  4  m  n  3 2
m  m  1 m  2  m  3
a4  a0
 m  n  2  m  n  4  m  n  1 m  n  3
Putting these values in (2), the solution is.

m m  m  1 x 2 m  m  1 m  2  m  3 x 4 
y  a0 x 1    ...
  m  n  2  m  n  1  m  n  2  m  n  4  m  n  1 m  n  3 
.....(5)
When m = n, replacing a0 by a, in (5) one of the solution is
 n  n  1 2 n  n  1 n  2  n  3 4 
y  ax n 1  x  x  ...  au (say) .....(6)
 2  2n  1 2  4  2n  1 2n  3 
178
 When y = –(n +1) and replacing a0 by b, in (5) the other solution is

  n  1 n  2  2  n  1 n  2  n  3 n  4  4 

y  bx  n1 1  x  x  ...  bv (say) .....(7)
 2   2n  3  2  4  2n  3 2n  5 
Hence the complete solution is
y = au + bv,
where a and b are arbitrary constants.

Self Learning Exercise

Fill up the blanks :

(1) The ordinary point of (x2 – 1) y + xy – y = 0 is ... ...
(2) For differential equation 2x2y + 7x(x + 1) y – 3y = 0, x = 0 is a ... ... singular point.
(3) The regular and irregular singular points of the differential equation
x2(x + 1)2 y + (x2 – 1) y + 2y = 0
are ........... and ............... respectively
(4) The nature of the point x = 0 for the equation xy + y sin x = 0 is ..... .

9.7 Summary

In this unit you studied the Frobenius method for finding the solution of a linear differential equa-
tion of second order with variable coefficient near ordinary and regular singular points. Various cases of
this important method were discussed and illustrated with the help of examples.

9.8 Answers to Self Learning Exercise

(1) x = 0 (2) Regular (3) x = 0 and x = – 1 (4) Regular singular

9.9 Exercise

Solve the following differential equations in series :

1. (1 – x2) y2 – xy1 + 4y = 0
 1 3 1 5 1 7 
 2

[Ans. y  a0 1  2 x  a1  x  x  x  x ...  ]
 2 8 16 
2
2. (1 – x ) y2 + 2xy1 + y = 0
 1 2 1 4 1 6   1 3 1 5 3 7 
[Ans. y  a0 1  x  x  x ...   a1  x  x  x  x  ...  ]
 2 8 18   2 40 560 
3. y2 + x2y = 0
 1 4 1   1 3 1 
[Ans. y  a0 1  x  x8  ...   a1  x  x  x 5 ...  ]
 3 4 3 4 7 8   45 4 5 8 9 
4. (2 + x2) y2 + xy1 + (1 + x) y = 0
 1 2 1 3 5 4   1 1 4 1 5 
[Ans. y  a0 1  x  x  x ...   a1  x  x 3  x  x  ...  ]
 4 12 96   6 24 24 
179
 5. 2x(1 – x) y2 + (1 – x) y1 + 3y = 0

 3 2 3 3 3 4 
[Ans. y  a 1  3 x  x  x  x  ...   bx1 2 1  x  ]
 1 3 3.5 57 
6. x2y2 + xy1 + (x2 – n2) y = 0, when n is not an integer.

 1 1 
[Ans. y  ax n 1  x2  x 4  ...
 4 1  n  4  8 1  n  2  n  

 1 1 
bx  n 1  x2  x 4  ... ]
 4 1  n  4  8 1  n  2  n  
7. (2x + x3) y2 – y1 – 6xy = 0

 2 3 4 3 1 6  3 2 3 2 3 1 4 3  1 5 6 
[Ans. y  a 1  3x  x  x  ...  bx 1  x  x  x  ...  ]
 5 59   8 8 16 8 16  24 
8. 9x(1 – x) y2 – 12y1 + 4y = 0

 1 1 4 2 1 4  7 3 
[Ans. y  a  1  x  x  x  ... 
 3 3.6 369 

 8 8  11 2 8 11 14 3 
bx 7 3  1  x  x  x  ...  ]
 10 10 13 10 13 16 
9. 4xy2 + 2y1 + y = 0

 x x2  12 x x2 
[Ans. y  a 1   ...   bx 1   ...  ]
 2 4   3 5 

10. x(1 – x) y2 + 3y1 + 2y = 0

 1 2 1 4   x3 1 4 
[Ans. y  a  1  x  x  ...  b
  x   x  ...  ]
 2 24   6 12 

11. xy2 + y1 + xy = 0
 1 1 
[Ans. y = ay1 + by2, where y1  1  2 x2  2 2 x4  ... and
 2 2 4 
 1 1  1 1  1 1 
y2  y1 log x   2 x 2  2 2 1   x 4  2 2 2  1    x 6  ...  ]
2 2 4  2  2 4 6  2 3 
12. (x – x2) y2 + (1 – x)y1 – y = 0
 2 2 5 3 
[Ans. y   a  b log x   1  x  x 2 
 4 49 

x  ...   b 2 x  x 2  ... ] 
13. xy2 + (1 + x)y1 + 2y = 0

 3 2 4 3    1 3 1 1 2 
[Ans. y   a  b log x   1  2 x  x  x  ...   b 2  2   x   2    x  ... ]
 2 3    2 2 2 3 
180
14. x(1 – x2) y2 + (1 – 3x2)y1 – xy = 0
 12 12  32  1 21 4 
[Ans. y   a  b log x   1  2 x 2  2 2 x 4  ...   b  x  x  ...  ]
 2 2 4  4 128 
15. x(1 – x) y2 – (1 + 3x)y1 – y = 0

   
[Ans. y   a  b log x  1 2 x 2  2  3 x3 ...  b 1  x  5 x 2 ... ]
16. xy2 + xy1 + (x2 – 4)y = 0
 1 1 1 
[Ans. y   a  b log x  x 2  2 x 4  3 x6  3 2 x8  ...
 2 4 2 46 2  4 68 
 1 1 1 
bx 2 1  2 x 2  2 2 x 4  2 2 2 x 6  ...  ]
 2 2 4 2 4 6 
17. x(1 – x) y2 – 3xy1 – y = 0

 2 3
  2
[Ans. y   a  b log x  x  2 x  3 x  ...  b 1  x  x  ... ] 
18. x2y2 + x(1 + 2x)y1 – 4y = 0
 4 2   4 4 
[Ans. y  a0 x 2  1  x  x 2   a4 x 2 1  x  x 2  ...  ]
 3 3   5 10 
19. (1 – x2)y2 + 2xy1 + y = 0

 1 2 1 4   x3 1 5 
[Ans. y  a0 1  x  x  ...  a
 1  x   x  ...  ]
 2 8  2 40 
 

  

181
Unit 10 : Gauss Hypergeometric Function: its Properties And
Integral Representation

Structure of the Unit

10.0 Objective

10.1 Introduction

10.2 Convergence of the Series

10.3 Special Cases of the Gauss Function

10.4 Integral Representation

10.4.1 Deductions

10.5 Gauss Hypergeometric Differential Equation and Its Solution

10.6 Two Summation Theorems

10.6.1 Theorem 1

10.6.2 Theorem 2

10.7 Summary

10.8 Answers of Self -Learning Exercise

10.9 Exercise

10.0 Objective

The aim of this unit is to study a special function known as Gauss hypergeometric function. Also
its special cases, properties, convergence conditions and summation theorems such as Gauss’s theorem,
Kummer’s theorem and Vandermonde’s theorem are obtained.

10.1 Introduction

The series

ab a  a  1 b  b  1 z 2 a  a  1 a  2  b  b  1 b  2  z 3
1 z   ... .....(1)
c c  c  1 2 c  c  1 c  2  3

is called the Gauss series or the Ordinary hypergeometric series. It is usually represented by the symbol
F1 (a, b ; c ; z), The three quantities a, b and c are called the parameters and z is the variable of the
2

series. All these four quantities may be any number, real or complex. In the notation 2F1 (.), the left suffix
182
2 and the right suffix 1 indicate the number of parameters in the numerator and denominator respec-
tively. If either of the parameters a or b (or both) is a negative integer, the series terminates i.e. it has
only a finite number of terms and becomes in fact a polynomial. Also when c is zero or a negative inte-
ger, the series is not defined.
C.F. Gauss carried out an exhaustive study of this function in a systematic way and Euler dis-
covered many properties of the function.
The function has its importance because of its application in solving various problems arising in
physical and engineering sciences. It is interesting to note that apart from the elementary functions such
as exponential function, logarithmic function, sine and cosine functions etc., it is also possible to derive
Bessel’s functions, Kummer’s confluent hypergeometric function, Bessel polynomials, Hermite polyno-
mials, Jacobi polynomials etc. either as a limiting case or as a special case of this function.
If we introduce the conventional notation (Pochammer symbol)

 a  n
  n       1   2  ...    n  1 , n  1 .....(2)
 
and   0  1,   0 ,
then the equation (1) can be written in the contracted form

 a, b    a  b  n
n n z
F
2 1
(a,b; c; z) or F
2 1 ; z  
 .....(3)
 c  n 0  c  n n
As pointed out earlier, in general a, b, and c are complex parameters and z is a complex vari-
able. If a or b is a negative integer then series terminates. Also c is neither zero nor a negative integer i.e.
c  0,–1,–2, .....
From (1), it follows easily that
(i) 2F1 (a, b; c; 0) =1
(ii) 2F1 (a, b; c; z) = 2F1(b, a; c; z)
The last property indicates that the hypergeometric function is symmetric in the upper param-
eters a and b.

10.2 Convergence of the Series in (3)

To test the convergence of the series in (3), let us apply the D’ Alembert’s ratio test. We see
that

un1  a n1  b n1 z n1  c n n

lim  lim 
n un n  c n1 n  1  a n  b n z n

 lim
 a  n  b  n   z
n c  n n 1

 z,
so long as non of a, b, c is zero or a negative interer.
183
 Therefore, the series converges absolutely within the circle of convergence if | z | <1 and di-
verges outside the circle of convergence i.e. | z | >1, provided that c is neither zero nor a negative inte-
ger. If either or both of a and b is zero or a negative integer, the series terminates, and convergence does
not enter the discussion.
For | z | =1, i.e. on the circle of convergence, the test fails. In this ease, let us compare this
series with the series

1
 vn   1 ,
n 1 n
where 2  = Re (c – a – b) > 0.

un  a n  b n 1
Since lim  n
n vn  c n n

 a n n  1  nc n  1 n1
 b n
 lim 
n n  1  n a n  1  nb  c  n n n c  a b

  a  n
1  b  n  1 n  1 nc   c  n  1 n1
 lim      
n n  1 n a   a  n  1  nb   b   c  n n nc  a b

n 1 nz
But we know that lim 1
n    z  n 

un  c  1
therefore lim v   a  b  lim n c a b   0,
n n     n 
because Re (c – a – b –  ) = 2  –  >0, therefore the series in (3) is absolutely convergent on
| z | = 1 when Re (c – a – b) > 0.
To summarise, we conclude that the hypergeometric series (3) or (1) is
(a) absolutely convergent within the circle of convergence | z | < 1
(b) divergent outside the circle of convergence | z | > 1.
(c) for | z | = 1 i.e. on the circle of convergence, it converges absolutely if Re (c – a – b) > 0. It
also converges conditionally for z = –1 if – 1 < Re (c – a – b)  0, and divergent if Re (c – a – b)

 – 1.

10.3 Special cases of the Gauss function

When a = 1, b = c, the R.H.S. of (1) reduces to

1
1  z  z 2  .....  , | z | 1
1 z
which is simply a geometric series. This is why (1) called the hypergeometric series.
184
 Most of the elementary functions which occur in Mathematical Physics, can be expressed in terms
of the Gauss function. For example,

  a n n

  z n
(i) 2 F1 (a, b; b; z )   n
z    a  a  1 ....  a  n  1 n
n 0 n 0

or F1 (a, b; b; z) = (1 – z)–a
2

This is simply a statement of the Binomial theorem for | z | < 1.


  z n
1
 log 1  z 
(ii) 2 F1 (1,1; 2;  z )   1 n z
n 0

(iii)
1 3 1
For | z | < 1, 2 F1 ( ,1; ; z 2 )  log
1  z 
2 2 2z 1  z 
 n
z  1  2    n  1   z
(iv) F
Since 2 1 (1; b;1;
b
)  1    b 
1 

1 
b
....
  
 
1 
 b
 ,
 n
n1

  z   z n
therefore, lim  2 F1  1, b;1;      ez
b 0   b   n0 n

 1 1 3 2  1 1
(v) 2 F1  , ; ; z   sin z
2 2 2  z

1 3  1
(vi) 2 F1  ,1; ;  z 2   tan 1 z
2 2  z

The Legendre polynomial Pn (x) is defined as the coefficient of zn in the expansion, in ascending

powers of z, of (1–2xz + z2)–1/2. By direct expansion, we can prove that the coefficient is in fact

 1 1 
2 F1   n,1  n;1;
 x = Pn (x). This result is known as Murphy’s formula.
 2 2 
Other elementary special cases are
 1 1  1 2 a 1 2 a
2 F1  a, a 

; ; z  1 z
2 2  2

 1 z
2
  
12 a
 1   1 1 
2 F1  a  , a; 2a; z     1 z 
 2  2 2 
2a 1
and 2 F1  2a, a  1; a; z   1  z  / 1  z 

10.4 Integral Representation

If | z | < 1 and if Re (c)> Re (b) > 0, then

185
 1
c  b 1 a
B (b, c – b) 2F1 (a, b ; c ; z)   t 1  t  1  zt 
b 1
dt
0

 c 1
c  b 1 a
F (a,b;c;z)  t 1  t  b 1
1  zt  dt
 b   c  b 
or 2 1
.....(1)
0

1
c  b 1 a
I   t 1  t  1  zt 
b 1
Proof. Let dt
0

r
1
b 1 c b 1

 a r  zt 
 t 1  t   dt
0 r 0 r
Now interchanging the order of integration and summation, we see that

 a r z r 1 b  r 1 c  b 1
I  t 1  t  dt
r 0 r 0


  b  r    c  b   a r z r
 
r 0  c  r  r

 b 
 a r  b r z r
  c  b  
 c r 0   c  r r

= B (b, c – b) 2F1 (a, b ; c ; z)

10.4.1 Deductions from integral representation

As a consequence of equation (1), we derive the Gauss’s theorem which gives rise to

Vandermonde’s theorem of the hypergeometric function. Kummer’s theorem is also derived. These

theorems are of great importance in the study of various special functions of mathematical physics.

(a) Gauss’s theorem. If Re (c – a – b) > 0, Re (c) > 0, then

  c    c  a  b
F1  a,b;c; 1 
2
  c  a    c  b
Prof. Putting z = 1 in the equation (1), we get
 c 1
c  a  b 1
 t 1  t  b 1
dt
2 F1  a, b; c;1  b   c  b   0

 c  b   c  a  b 
=  b  cb
     c  a 

 c    c  a  b
 F
2 1  a, b; c;1 =  c  a   c  b .....(2)

186
 (b) Vandermonde’s theorem
 c  b n
2 F1   n, b; c;1  =
 c n
Prof. If we make a = – n in eq. (2), where n is a positive integer, then we get
  c    c  b  n n  c  b n
F1  n, b; c;1  
2
 c  n  c  b  c n
(c) Kummer’s Theorem

 b
 1  a  b    1  
 2
2 F1  a , b;1  a  b; 1  
 b  .....(3)
 1  b    1   a 
 2 
Prof. To prove (3), we put z – 1 and c = 1 – a + b in equation (1), we abtain
 1  a  b  1
2 a
F1  a, b;1  a  b; 1  t 1  t  b 1
dt fdgdg .....(4)
  b   1  a  
2
0

Putting t2 = u in the above equation (4), we get

 1  a  b  1
 b 2  1 1 a 1
F1  a, b;1  a  b; 1  u 1  u  du
2
2  b   1  a   0

b
    1  a 
 1  a  b  2
   
2  b   1  a  b 
  1 a 
2 
 1   b / 2    1  a  b 
 2 1 F  a, b;1  a  b; 1 
 1  b   1  a  b / 2 

10.5 Gauss’s Hypergeometric Differential Equation and its Solution

d
Let = z
. Then zn = nzn
dz
Therefore, (+ c – 1) zn = n (n + c – 1) zn.

 a n  b  n n
Now y = 2 F1  a, b; c; z    z
n  0  c n n

 n  n  c  1 a n  b  n
We have    c  1 y   zn
n 0  c n n


 a  n  b n n
 z
n 1  c n 1 n  1

187
 
 a n1  b n1 n 1
 z
n 0  c n n

 a  n  b  n  a n  b n n
 z
n 0  c n n
= z (+ a) (+ b) y
 
 a  n  a n  b n n 
 since   a  
y  z 

 n  0  c  n
n 
Hence y =2F1 (a ,b; c; z) is a solution of differential equation
d
   c  1  z   a   b   y  0,   z 
dz
The above equation can be easily written in the following form
d2y dy
2
 c   1  a  b  z  ab y  0
z 1  z  .....(1)
dz dz
(by employing the relations y = zy and ( – 1) y = z2y) is known as Gauss’s hypergeometric
differential equation.
From the theory of differential equation, it follows that the regular singular points of the above
equation (1) are:
(i) z = 0 with exponents 0, 1 – c
(ii) z =1 with exponents 0, c – a – b
(iii) z =  with exponents a,b.
For details of the solution of the differential equation (1), students are advised to reter Ex. 2 in
§9.4 of the last unit.

10.6 Two summation Theorems

In this section, we discuss two theorems concerning elementary series manipulations which are
important techniques in establishing several transformation formulae, summation formulae and in investi-
gating several other properties of hypergeometric functions, Bessel’s functions and Orthogonal polyno-
mials etc.
10.6.1 Theorem 1.
   m
    n, m 
m  0 n 0
    n, m  n  .....(1)
m  0 n 0
 m  

and.     n, m   
m  0 n 0
   n, m  n  .....(2)
m  0 n 0
Proof. Consider the L.H.S. of the equation (1) in which the term um + n has been inserted for
convenience i.e.
 
m n
    n, m  u
m 0 n 0
.....(3)
188
 Let us collect the powers of u in (3). We introduce new indices of summation s and r by
n = r, m = s – r .....(4)
so that n+m=s .....(5)
The indices n and m now satisfy the inequalities m  0, n  0.
From (4) and (5), it follows that s – r  0, r  0 or 0  r  s
provided that s is restricted to be a non-negative integer. Thus we have
   s
m n s
    n, m  u
m 0 n 0
    r, s  r  u
s  0 r 0

Now putting u = 1 and replacing the dummy indices r and s on the right by n and m respectively,
we get the required result.
In Theorem 1, equation (2) is merely written in reverse order; hence no separate proof is needed.
Theorem 2.
    m 2
    n, m 
m0 n0
    n, m  2 n  .....(6)
m0 n 0

  m 2  

and    n, m   
m0 n0
   n, m  2 n  .....(7)
m  0 n 0

 m 2
where the symbol l 0
indicates that n runs from 0 to the greatest integer less than or equal to m/2.

Proof. If we consider
 
m  2n
    n, m u
m 0 n  0

 
in which u m + 2n
is inserted for convenience, i.e.     n, m  u m  2 n and taking n = r and
m0 n0
m = s – 2r so that m + 2n = s.
Since m  0, n  0, s – 2r  0, r  0 from which 0  2r  s and s  0.
s
Since 0  r  and r is integral, the index r runs from 0 to the greatest integer s/2. Thus we
2
obtain

    s 2
    n, m  u m  2 n      r , s  2 r  u s

m0 n0 s0 r 0

Now putting u = 1 and replacing the dummy indices r and s on the right by n and m respectively,
we get the required result (6). Equation (7) is written in reverse order. If we combine the above two
 m   m 2
theorems, we find that 
m 0 n 0
   n, m       n, m  n 
m 0 n 0

189
 a a 1 1  1 a a
Ex.1. Prove that 2 F1  ,  ; ; z 2   1  z   1  z  
2 2 2 2  2  

1 a a
Sol. Taking R.H.S.  1  z   1  z  
2

1  a  a  1 2 a  a  1 a  2  3 a  a  1 a  2  a  3 4
 1  az  z  z  z
2   2 3 4

a  a  1 a  2  a  3 a  4  5 
 z  ...
5 

 a  a  1 2 a  a  1 a  2  3 a  a  1 a  2  a  3 4
 1  az  z  z  z
 2 3 4

a  a  1 a  2  a  3 a  4  5 
 z  ...
5 

1 2 a  a  1 a  2  a  3 4 
  2  a  a  1 z  z  .... 
2 12 

 a a  a  1 a  2  a  3 4 
 1   a  1 z 2  z  .... 
 2 2 2 2 3 

 aa 1  a  a 1  a  a 3  
 2 2  2      1   
 1    z 2   2  2 2  2  2 2  z 2 2  ....
1 1 3
 
   2 1 
   2 2 
2
 

a a 1 1 
 2 F1  ,  ; ; z 2   L.H.S.
2 2 2 2 
Ex.2. Establish the result
n
 1   1  a  c  n
2 F1   n, a  n ; c ;1 
 c n
Sol. Here L.H.S.  2 F1  n, a  n ; c ;1

  c   c  a 
 (by Gauss’s summation Theorem)
 c  n  c  a  n 

 1  c  a  n 
 (1)n
 1  c  a 
n
 1 1  a  c n
 Hence proved.
 c n
190
Ex.3. Prove that
1
B(  c – ) 2F1 (a, b; c; z)  t 
1
 1  t c 1 2 F1  a , b; ; zt  dt
0
where | z | < 1,  > 0, c –  > 0.
1
1 c 1
Sol. Let I=  t 1  t  2 F1  a, b ;  ; zt  dt
0

1
1 c 1
  a r  b r  zt r
 t 1  t    dt
0 r  0   r r

1

z r  a r  b r c 1
  r     t  r 1 1  t  dt
r 0 r 0

z r  a r  b  r     r    c   

    r  c  r 
r 0 r

      c      a r  b r z r

 c
 
r  0  c r r
= B (c–) 2F1 (a, b; c; z)
Ex.4. Show that if b > 0,
a  2
2 1   z / 2 
F1(a, b ; 2b ; z)  2b 1   sin   2b1  1   cos    a   1   cos    a  d 
2
2 B  b, b  0
z
where   
2 z
Deduce that
a
 1  1 1 1 1 
2 F1  a , b ; 2b ; z   21  z  2 F1  a, a  ; b  ; 2 
 2  2 2 2 2 
Sol. We know that if | z | < 1 and if Re (c) > Re (b) > 0, then
1
b 1 c b 1 a
B (b, c – b) 2F1 (a, b ; c ; z) =  t 1  t  1  tz  dt .....(9)
0
For c = 2b, it reduces to
1
1 b 1 a
F (a, b; 2b; z)   t b 1 1  t  1  tz  dt .....(10)
2 1
B  b, b 
0
2
Putting t = sin , we have
2 a
2
  sin    cos   1  z sin 
2b 1 2b 1 2
F (a, b ; 2b ; z)  d
2 1 B  b, b 
0
191
 2 a
2 2b 1 2b 1   1  cos 2  

B  b, b    sin   cos   1  z 
  2


d
0

2 a
2 2b1 2b 1  2  z  z cos 2 

B  b, b    sin   cos   2  d 
0

a  2 a
2 1   z 2   2b1  z 
 
2b 1
B  b, b    sin   cos   1  2  z cos 2
 
d
0

a  2
2 1   z 2  

2b1   sin 2 2b1 1   cos 2  a d  .....(11)
2 B  b, b  0

z
where   . If we put 2= , then (11) becomes
2 z
a 
1   z 2  
F (a, b; 2b; z)  2b1
2 1   sin  2b1 1   cos  a d  .....(12)
2 B  b, b  0
In the same way, if we substitute t = cos2in (10), we get
a 
1   z 2  
F (a, b; 2b; z)  2b1
2 1   sin  2b1 1   cos  a d  .....(13)
2 B  b, b  0
Adding (12) and (13) and applying the property of the definite integral, viz.

 a
2a
2 f  x  dx, if f  2a  x   f  x  ,
 f  x  dx   0
0 
0, if f  2a  x    f  x 

we obtain the desired result

a
1   z 2    2
F (a, b; 2b; z)  2b1
2 1   sin  2b1 1   cos   a  1   cos   a  d 
2 B  b, b  0

To deduce the second part, we find from example 1 that

1   cos   a  1   cos    a   2  F  a , a  1 ; 1 ;  2 cos 2  

  2 1 
2 2 2 2 
a
4 1   z 2    2 a a 1 1
Hence 2F1 (a, b; 2b; z)  2b 1   sin  2b1 2 F1  ,  ; ; 2 cos2   d 
2 B  b, b  0 2 2 2 2 

Expanding 2F1 (2 cos2) in terms of its series and integrating with the help of beta function
formula, we have
192
 a   a 2   a  1 2 2
4 1   z 2   r
F (a, b; 2b; z)  2b 1  r    sin 2b 1  cos 2 r  d 
2r
2 1
2 B  b, b  r  0 1 2 r r 0

a   a 2   a  1 2
4 1   z 2   r 2r    b    r  1 2 
   r
22b 1 B  b, b  r  0 1 2 r r 2  b  r  1 2 
Applying Legendre’s duplication formula, we get
a 1
a   a 2r   2r
 z  2 r 
 2 1  
 2
  1

r
r 0 b  
 2 r
a
z a a 1 1
 F (a, b; 2b; z)  2  1  
2 1 2 F1 

,  ; b  ; 2 
 2 2 2 2 2 
 
Ex.5. Show that if  x ,
2 2
1 1 1 1 3 2 
sin nx = n sin x 2F1   n,  n ; ; sin x
2 2 2 2 2 

 n n 1 
and cos nx  2 F1  , ; ; sin 2 x.
 2 2 2 
Sol. We know that sin nx and cos nx satisfy the following differential equation

d2y
 n2 y  0 .....(14)
2
dx
Let us transform (14) by the substitution u = sin2 x. Then

du d 2u
 sin 2 x and  2 cos 2 x
dx dx 2
dy dy du dy
Now,    sin 2 x
dx du dx du

d2y d  dy  d  dy 
     sin 2 x  
dx 2 dx  dx  dx  du 

dy d 2 y du
 2 cos 2 x  sin 2 x  
du du 2 dx

dy d2y
 2 cos 2 x  sin 2 2 x
du du 2
2

 2 1  2 sin 2 x  dudy  4sin 2 x cos2 x  ddu 2y
193
 d2y dy d2y
  2 1  2u   4u 1  u 
dx 2 du du 2

d2y
Substituting the value of in (14), it becomes
dx 2

d2y  1  dy n
2
u 1  u 
 u  y0
du 2  2  du 4
The above equation may be written as

d2y  1  n n   dy  n   n 
u 1  u     1    u     y  0
du 2  2  2 2   du  2   2 
n n 1
which is a Gauss’s hypergeometric equation with a  , b  , c  . Hence the general solution of
2 2 2
(14) is given by
 n n 1  1 1 1 1 3 
y  A 2 F1  , ; ; sin 2 x   B sin x 2 F1   n,  n ; ; sin 2 x 
2 2 2  2 2 2 2 2 
Since sin nx is the solution
 n n 1  1 1 1 1 3 
 sin nx  A 2 F1  , ; ; sin 2 x   B sin x 2 F1   n,  n ; ; sin 2 x 
2 2 2  2 2 2 2 2 
.....(15)
For x = 0, equation (15) gives A = 0
Further
sin nx 1 1 1 1 3 
 B 2 F1   n,  n ; ; sin 2 x 
sin x 2 2 2 2 2 
 sin  
Now taking limit of both sides as x  0 , and noting that lim    1,
0   
we get B = n
1 1 1 1 3 
 sin nx = n sin x 2 F1   n,  n ; ; sin 2 x 
2 2 2 2 2 
Again, if y = cos nx, then putting x = 0, we see that A = 1, and on differentiating and putting x = 0,
we get B = 0, which establishes the second part.
Ex.6. Show that

 1 
  a    b  2 F1  a , b ; ; z     e  u cos h 2
 2  00
  u  z ua 1v b1du dv

provided Re (a) > 0 and Re (b) > 0.

 
Sol. R.H.S   e
u 

cosh 2  u  z u a1vb1du dv .....(16)
0 0
194
 
22 r u r v r z r
But we know that cosh 2  
 u  z   2r
r 0
Putting this value in the above integral (16), then it breaks up into product of two integrals, and we
have
  
22 r z r
   e u u a  r 1du  e  v vb  r 1dv
r 0 2r
0 0


22 r
    a  r   b  r   z r
r  0   2r  1


22r  a  r   a  b r   b  r
  2r  2r   1 
z
r 0  
2

1
 22 r 1     a  r   a  b  r   b  z r
  2
 1
r 0 r 22 r 1   r    r  
 2
(applying Legendre’s duplication formula)

  a  r  b r z r
  a   b 
r0 1
r  
 2 r

 1 
   a    b  2 F1  a, b ; ; z   L.H .S
 2 
Ex.7. Prove that

1  a  n  1  b  n  1 n 1
lim 2 F1  a , b ; c ; z   z 2 F1  a  n  1, b  n  1 ; n  2 ; z 
c  n   c  n1

1
Sol. L.H.S = clim 2 F1  a, b ; c ; z 
 n   c 

1   a r  b r z r
 lim
c  n   c 
 c  r
r 0 r

  a r  b  r z r
 lim
c  n
  c  r   r
r 0

  a  r  b r z r
  
r  n 1    n  r  r

195
   a  sn1  b s n1 z s  n1
  (Putting r – n – 1 = s)
s 0   s  1  s  n 1

  a  n  1s  a n 1  b  n  1s  b n1 z s  n1

  s  n  2 s

n 1
s 0

 a n 1  b n1 z n1   a  n  1s  b  n  1s z s


 n  1   n  2 

s
s0 s

 a n1  b n1 n1

 z 2 F1  a  n  1, b  n  1 ; n  2 ; z 
 n  1
= R.H.S
Ex.8. It the complete elliptic integral of first kind being
 2 d
K 
0
1  k 2 sin 2 

 1 1 
Show that K 2 F1  , ; 1 ; k2 
2 2 2 
2
d
Sol. We have K 
0 1  k 2 sin 2 
Putting sin   t
1 1
then cosd = dt
2 t
1 1 1
or d =  dt
2 t 1 t
1
1 1 2
 K 
2  t 1 2 1  t 
1 2
1  k t 
2
dt
0
 By integral representation of 2F1 (a, b ; c ; z), we have
1  1
    1  
1 2  2 1 1 
K  2 F1  , ; 1 ; k2 
2  1 2 2 
 1 1 
 2 F1  , ; 1 ; k 2   R.H .S
2 2 2 
Ex.9. Prove that
1
1 c  b 1
2 F1  a , b ; c ; z    u b 1  1  u  1  zu   a du,
B  b, c , b  0
where c > b > 0. Hence prove that
2 z
2 F1  1,2 ; 3 ; z   log e 1 z 1 z 
196
 Sol. By integral representation of 2F1 (a, b ; c ; z), we have, if | z | < 1 and if Re (c) > Re (b) > 0,
then
1
1 b 1 c b 1 a
F (a, b; c; z) 
B b, cb   u 1  u  1  zu  du
0
1
1 0 1
Now, F (1, 2 ; 3 ; z) 
B  2,1  u 1  u  1  zu  du
0
1
1 u du
  1  zu
B  2,1 0
1
2  1 

z  1  zu  1 du
0

 1 1
2  1
   log 1  zu     u 0 
z   z 0 
2 1 
   log 1  z   log1  1
z z 
2 1 
   log 1  z   1
z z 
2 1z 2 1z
   log 1  z   log e    log e 1  z  
z   z  
2 z

 log e 1  z 
1z
 = R.H.S.
Ex.10. Show that if Re (b) > 0 and if n is a nonnegative integer, then
n
 n n 1 1  2  b n
F
2 1  ,   ; b  ; 1 
 2 2 2 2   2b  n

 n n 1 1 
Sol. L.H.S  2F1   ,   ; b  ; 1
 2 2 2 2 

 1
  b    b    b n
 2

 n  n 1 (by Gauss’s theorem)
b  b   
 2  2 2
Using Legendre’s duplication formula, we have

 n
2 b   1
  2b   2  2
L.H.S =   b n
2b 1   n 
2   2  b  
  2 
 b n
 2n  R.H .S
 2b n
197
 Ex.11. Show that
t
1 2 1  x 2  t  x  2 
1 2 1 1 3 t4 
 x1 2  t  x    dx  t 2 F1  , ; 1 ; 
0
2  4 4 16 

t
2 1 2
Sol. Let
12
I   x t  x 
0
1 2
 1  x2 t  x   dx
n
t
1 2
 1 2 n  x2  t  x 2 
  x1 2  t  x   dx
0 n0 n
t 1
 1 2 n 2 n 2 n
1
   x 2 t  x  2 dx
n0 n 0

Putting x = tu, we have

1 1
 1 2 n 2n 1
I  t 4 n1
 u 2 1  u 2n  2 du
n0 n 0

3  1 
   2n     2n 
 1 2 n 2  2
  t 4 n1  
n0 n   2  4n 

Applying Legendre’s duplication formula for   2  4n  , we find that

1  1
   2n    
 1 2 n 2 2
I=   t 4 n1 4 n1   
n0 n 2   2n  1
Again applying the Legendre’s duplication formula and simplifying, we have

1  3  1
  n  n 
 1 2 n 1 4  4  2
I    t 4 n1  4 n  3 2 
n 2   1 
n 0    n    n  1
2 

1 1 3 3

           4 n 1
 4 n  4   4 n  4  t
   4n  3 2
n0 1n n 2  

1 3
    n
1 
 4 n  4 n  t4  t
 
2 2 n  0 1n n
 
 16  sin   4 

t 1 3 t4 
 F 
2 1 , ; 1 ; 
2 4 4 16 
which completes the solution.
198
 Self-Learning Exercise
1. Define Gauss hypergeometric function in terms of a series.
2. What is circle of convergence for the series representing 2F1 (a, b ; c ; z) ?
3. 2F1 (– n, b ; c ; 1) = .............. .
4. 2F1 (a, b ; 1 – a + b ; – 1) = .............. .
5. 2F1 (a, b ; c ; 1) = .............. .
6. 2F1 (a, b ; b ; z) = .............. .

 z
7. lim 2 F1 1, b ;1;  = ...
b 0  b
8. 2F1 (– n, 1 – b – n ; a ; 1) = .............. .

10.7 Summary

In this unit, the function introduced by C.F. Gauss was studied. The important special cases,
properties and convergence conditions of this function were discussed in detail.

10.8 Answers of Self-Learning Exercise

  a n  b n z n  c  b n
1.  c n 2. | z | < 1 3.
 c n
n0 n

 b
 1  a  b    1  
 2  c    c  a  b
4. 5. 6. (1 – z)– a
 b   c  a   c  b
 1  b   1   a 
 2 
 a  b  12 n
7. ez 8.
 a n  a  b  1n
10.9 Exercise
1. Define hypergeometric function 2F1 (a, b ; c ; z) and state the condition on its elements a, b and c
for its convergence.
2. Find representation of following functions in terms of Gauss hypergeometric function :
(i) (1 + z)n [Ans. 2F1 (– n, 1 ; 1 ; – z)]
1  a 1 a 3 
(ii) 
1  z  a  1  z  a  [Ans. 2 F1   ,  1 ; ; z 2  ]
2az 2 2 2 2 
1
(iii) log 1  z  [Ans. 2F1 (1, 1 ; 2 ; – z)]
z
1 1 z  1 3
(iv) log   [Ans. 2 F1  , 1 ; ; z 2  ]
2z  1 z  2 2 
199
 (v) sin 1 z 1 1 3
[Ans. 2 F1  , ; ; z 2  ]
z 2 2 2 

1 3
(vi) 1 tan 1 z [Ans. 2 F1  ,1 ; ; z 2  ]
z 2 2 
 3 1 
(vii) sin z [Ans. z 0 F1  ; ;  z 2  ]
 2 4 

 1 1 2
(viii)cos z [Ans. 0 F1   ; ;  z  ]
 2 4 
3. Express complete elliptic integral of the second kind in terms of Gauss’s hypergeometric function
1 1 1 
[Ans.  2 F1  ,  ; 1 ; k 2  ]
2 2 2 

d2y
4. By transforming the equation 2
 n2 y  0 to hypergeometric form by the substitution   sin 2 z ,
dx
prove that if 0  z   then,
 n  n n 1   n 
cos nz  cos   2 F1  ,  ; ; cos 2 z   n sin   cos z
 2  2 2 2   2 
1 n 1 n 3 
2 F1   ,  ; ;cos 2 z 
2 2 2 2 2 
 n  n n 1 2   n 
and sin nz  sin   2 F1  ,  ; ; cos z   n cos   cos z
 2  2 2 2   2 
1 n 1 n 3 
2 F1   ,  ; ; cos 2 z 
2 2 2 2 2 
5. Establish the transformation formula
1 1
F1 (2a, 2b; a + b + ; z) = 2F1 {a, b; a + b + ; 4z (1 – z)}
2 2 2
1
provided that a + b + is not zero or a negative integer and if | z | < 1 and | 4z (1 – z) | < 1
2

lim 2 F1 a, b; a  b; z     a  b 
6. Show that
z 1  log 1  z    a   b
2
1 2
 1  k 
2
7. If the complete elliptic integral of the first kind is K   sin 2  d ,
0

  1 1 2
then show that K  2 F1   , ; 1; k  , | k | 1
2  2 2 

  

200
Unit 11 : Gauss and Confluent Hypergeometric Functions
Structure of the Unit
11.0 Objective

11.1 Introduction

11.2 Linear Transformation Formulas for Hypergeometric Function

11.2.1 Applications

11.3 Differentiation Formulas for Hypergeometric Function

11.4 Linear Relation between solutions of Hypergeometric Equations

11.5 Relations of Contiguity for Hypergeometric Function

11.6 Kummer’s Confluent Hypergeometric Function

11.6.1 Convergency

11.6.2 Differentiation Formulas

11.6.3 Integral Representation

11.6.4 Kummer’s First Transformation

11.7 Summary

11.8 Answers of Self-Learning Exercise

11.9 Exercise

11.0 Objective

In the last unit the Gauss hypergeometric function was introduced and some properties, summa-
tion theorems and convergence conditions for this function were discussed. The aim of this unit is to
study further the hypergeometric function. Precisely you will study the linear transformation formulas, con-
tiguous function relations, differentiation formulas and a linear relation between the solutions of hyper-
geometric differential equation. You will also study the kummer’s confluent hypergeometric function and
important formulas concerned with this function.

11.1 Introduction

Here some more results for the Gauss hypergeometric function (introduced in the last unit) will
be established. In fact linear transformation formulas, contiguous function relations, differentiation formu-
las etc. will be discussed in this unit.Next, the Kummer’s confluent hypergeometric function will be intro-
duced and important formulas for this function will also be established.
201
11.2 Linear Transformation Formulas
Result :
z
If z  1 and  1 , then
1 z

a  z 
(i) 2 F1  a , b, c ; z    1  z  F a,
2 1
c  b, c ; .....(1)
 z  1 

b  z 
(ii) 2 F1  a , b, c ; z    1  z  2 F1  c  a , b ; c ; .....(2)
 z  1 
c a b
(iii) 2 F1  a , b, c ; z    1  z  F
2 1  c  a, c  b ; c ; z .....(3)

Proof. (i) We know that by integral representation of 2 F1  a, b, c ; z  , if | z | < 1 and if

Re  c   Re  b   0
1 c b 1
Then
b 1
B  b, c  b  2 F1  a, b ; c ; z    0 t 1  t  1  tz a dt
1
  1  t 
0
b 1
1  1  t cb1 1  z 1  t a  dt
1 b 1
  t c b1 1  t 
0
1  z  tz a dt
a
a 1 cb1 b 1  tz 
 1  z   t 1  t  1   dt
0  z 1
a  z 
 1  z  B  c  b, b  2 F1  a, c  b ; c ; 
 z 1 

a  z 
Thus 2 F1  a, b ; c ; z   1  z  2 F1  a, c  b ; c; 
 z 1 
(ii) Taking L.H.S 2 F1  a, b ; c ; z   2 F1  b, a ; c ; z 

b  z 
 1  z  2 F1  b ;c  a ; c ; 
 z 1 
(by first transformation formula)
b  z 
 1  z  2 F1  c  a ; b ;c 
 z 1
(by symmetric properly)
b  z 
Hence 2 F1  a, b ; c ; z   1  z  2 F1  c  a, b; c ; 
 z 1 

a  z 
(iii) From (1), we have 2 F1  a, b ; c ; z   1  z  2 F1  c  b, a ; c;  .....(4)
 z 1 
202
 z
Putting  y or 1 - y = (1 – z)-1, we have
z 1
 z 
Now 2 F1  c  b, a ; c ;  = F  c, b ; a ; c ; y 
 z 1  2 1

 y 
 1  y    2 F1  c  b, c  a ; c ;
 c b
 .....(5)
 y 1 

 z  c b
2 F1  c  b , a
; c;
z 1  
or   1 z  2 F1  c  b, c  a ; c ; z  ......(6)

Using (6) in (4), we have
c a b
2 F1  a, b ; c ; z   1  z  2 F1  c  a, c  b ; c ; z 
11.2.1 Applications
1
If we set z  in the first transformation formula, then
2
 1
2 F1  a, b ; c ;   2a 2 F1  a, c  b ; c ;  1 .....(7)
 2
The series on the R.H.S. of (7) can be summed in terms of product of gamma functions with the
help of Kummer’s theorem in the following cases :
(i) c  c  a  b  1 that is b  1  a
1 a  b
(ii) c  a   a  b   1 or c 
2
From the first case, we get
 1 a
2 F1  a,1  a ; c ;   2 2 F1  a, c  a  1 ; c ;  1
 2

1 c  a 
2a   c    
 1   2 
2 F1  a,1  a ; c ;   1 c  a 
 2  c  a   
 2 
Further, applying the Legendre’s duplication formula for   c  and   c  a  , then we obtain

 c   1 c 
  
 1 2  2 
F
2 1 a ,1  a ; c ;  
  2  c  a  1 c  a 
  
 2   2 
In the same way, in the second case, we can prove the following result.

 1  1 a  b 
  
 1 a  b 1    2   2 
2 F1  a, b ; ; 
 1 a   1 b 
 2 2   
 2   2 
203
11.3 Differentiation of Hypergeometric Functions
Result : Show that
d ab
(i)  2 F1  a , b ; c ; x    F  a  1, b  1; c  1 ; x  .....(1)
dx c 2 1

dn  a n  b n
(ii)  2 F1  a , b ; c ; x    F  a  n, b  n; c  n ; x  .....(2)
dx n  c n 2 1
Proof of (i), we have
d d    a r  b r xr 
F
2 1  a , b ; c ; x     
dx dx  r 0  c  r r 

 a r  b  r r x r 1

r  0  c r r
  a r  b r r 1
 x
r 1  c r r  1

  a n1  b n1 x n
  c 
n
n 0 n1

Since  a n1  a  a  1 n ,

d  a  a  1 b  b  1 n
 2 F1  a, b ; c ; x     n nx
Therefore
dx n 0 c  c  1n n

d
  2 F1  a, b ; c ; x    ab 2 F1  a  1, b  1 ; c  1 ; x 
dx  c
(ii) We prove the result by the principle of mathematical induction
Since by (1), we have
d ab
 2 F1  a, b ; c ; x    2 F1  a  1, b  1 ; c  1 ; x 
dx c
Therefore the result (2) is true for n = 1
Suppose that (2) is true for n = m (a fixed positive integer) i.e.

dm  a m  b m
 2 F1  a, b ; c ; x    2 F1  a  m, b  m ; c  m ; x 
dx m  c m

d m1 d  dm 
Now,  2 F1  a, b ; c ; x     m  2 F1  a  m, b  m ; c  m ; x 
dx m1 dx  dx 

 a  m  b m d
  2 F1  a  m, b  m ; c  m ; x  
 c m dx 

204 [by equation (4)]

  a m  b m  a  m  b  m 
   a  m  1, b  m  1 ; c  m  1 ; x 
 c m cm 2 F1

 a m1  b m1
 2 F1  a  m  1, b  m  1 ; c  m  1 ; x 
 c m1
Thus result (2) holds for n = m + 1. Hence by P.M.I the result (2) is true for every positive
integer n.

11.4 Linear Relation between the Solutions of Hypergeometric equations

In the unit 9, we have seen that the differential equation

d2y du
z 1  z   c  1  a  b  z  abu  0 .....(1)
2 dz
dz

has the solutions A 2 F1  a, b ; c ; z  and B z1– c 2 F1  a  1  c, b  1  c ; 2  c ; z  which are conver-

gent for | z | < 1 whereas the solutions A 2 F1  a, b ; a  b  1  c ; 1  z  and B 1  z c a b

2 F1 c  a , c  b ; 1  c  a  b ; 1  z  of the hypergeometric differential equation are convergent for

|1 – z | < 1. (Refer Ex.2. §9.4)
Hence there exist an interval (0, 1) in which all the four solutions exist. Since only two solutions
of the second order differential equation are linearly independent, which implies that there may exist a
linear relation between the solutions.
Let the relation be
F  a, b ; c ; z   A 2 F1  a, b, a  b  c  1 ; 1  z  
c a b
B 1  z  2 F1  c  a, c  b ; c  a  b  1 ; 1  z  .....(2)
where A and B are constants.
Putting z = 1 in the above equation (2) and applying the Gauss’s theorem, we have
 c  c  a  b
2 F1  a, b ; c ; 1 = A .....(3)
 c  a  c  b

where R c  a  b  0
Again, if we put z = 0 in (2), then it gives
1 = A 2F1 (a, b; a+b – c+ 1 ; 1) + B 2F1 (c – a; c–b ; c– a – b +1 ; 1)
  a  b  c  1  1  c    c  a  b  1  1  c 
or 1= A B .....(4)
  a  c  1   b  c  1  1  a   1  b 
Putting the value of A from the equation (3) in the equation (4) we obtain
  c    c  a  b    a  b  c  1  1  c    c  a  b  1  1  c 
1 =  c  a  c  b  a  c 1  b  c 1  B
         1  c   1  b 
205
   c   1  c    c  a  b   1  c  a  b   B   c  a  b  1  1  c 
or 1=  1  a   1  b 
  c  a   1  c  a    c  b   1  c  b 


Since   z   1  z   , therefore
sin  z

sin   c  a  sin   c  b    c  a  b  1  1  c 
1= B
sin  c sin   c  a  b   1  a   1  b 

  c  a  b  1  1  c  sin   c  a   sin   c  b 
 B = 1
 1  a   1  b  sin  c  sin   c  a  b 

sin  c  sin   c  a  b   sin   c  a  sin   c  b 


sin  c  sin   c  a  b 

cos   a  b   cos   2c  a  b   cos   b  a   cos   2c  a  b 


2sin  c sin   c  a  b 

cos   a  b   cos   b  a 

2 sin c sin   c  a  b 

sin a sin b

sin c sin   c  a  b 

 1  a   1  b  sin a sin b
 B= 
 1  c    c  a  b  1 sin c sin   a  b  c 


Applying sin z = , we have
  z   1  z 

 1  a   1  b   
B=  
 1  c    c  a  b  1   a   1  a    b   1  b 

 c   1  c    a  b  c   1  a  b  c 

2

 c  a  b  c
B=
  a   b
Substituting these values of A and B in (2), we get the following linear relation :
  c    c  a  b
F (a; b; c; z ) =  c  a  c  b 2 F1  a , b ; a  b  c  1 ; 1  z 
   
  c  a  b  c
  1  z  ca b 2 F1  c  a, c  b ; c  a  b  1 ; 1 z
  a  b
206
11.5 Relations of Contiguity
The functions obtained by increasing or decreasing any one of the parameters of the hypergeo-
metric function 2F1 (a; b; c; z) by unity, are called the functions contiguous to it. In this way, we obtain
the following six functions contiguous to 2F1 (a; b; c; z) :
(i) F (a+) = 2F1 (a+ 1 ; b; c; z)
(ii) F (a–) = 2F1 (a– 1, b; c; z)
(iii) F (b+) = 2F1 (a, b + 1; c; z)
(iv) F (b–) = 2F1 (a, b – 1; c; z)
(v) F (c+) = 2F1 (a b; c + 1; z)
(vi) F (c–) = 2F1 (a b; c – 1; z)
Now we shall see that the function 2F1 can be connected with any two of its contiguous functions
giving rise to fifteen (that is 6C2) relations in this way. These relations were first obtained by Gauss and
are called contiguous function relations.

 a n  b  n z n 
If we write
 c n

n
  n , then clearly F = 2F1   n .....(1)
n0
Now we have
F (a +) = 2F1 (a + 1; b; c; z)
  a  1n  b n z n
 
n 0  c n n

a  n  a n  b n z n

  
n 0 a  c n n


an
  a
 n [using (1)]
n 0

In this way, we obtain the following relations

 
 a  n    a  1
F (a +) =  a
n , F (a –) =   a  1  n  n
n 0 n 0


b  n  
  b  n , F (b –) =  b 1  n n
b 1
F (b +) =
n 0 n 0  

c  
  c  n  n , F (c –) =   c  1  n
c 1  n
F (c +) =
n 0 n 0  
In proving these relations, the formulae
  a  n  1  a  1
  z  1  z   z  and (a – 1)n = =  a n were used.
  a  1  a  n  1
207
 There are fifteen contiguous function relations for the hypergeometric function, which are given
below :
(i) (a – b) F = a F (a +) – b F (b +)
(ii) (a – c + 1) F = a F (a +) – (c – 1) F (c –)
(iii) [(a + ( b – c) z] F=a (1 – z) F (a +) – c –1 (c – a) (c – b) z F (c +)
(iv) (1 – z) F = F (a – ) – c–1 (c – b) z F (c +)
(v) (1 – z) F = F (b – ) – c–1 (c – a) z F (c +)
(vi) [2a – c + (b – a) z] F = a (1 – z) F (a +) – (c – a) F (a –)
(vii) (a + b – c) F = a (1 – z) F (a +) – (c – b) F (b –)
(viii) (c – a – b) F = (c – a) F (a –) – b (1 – z) F (b +)
(ix) (b – a) (1 – z) F = (c – a) F (a –) – (c – b) F (b –)
(x) [1 – a + (c – b – 1) z] F = (c – a) F (a –) – (c – 1) (1 – z) F (c –)
(xi) [2b – c + (a – c) z] F = b (1 – z) F (b +) – (c – b) F (b –)
(xii) [b + (a – c) z] F = b (1 – z) F (b +) – c–1 (c – a) (c – b) z F (c +)
(xiii) (b – c + 1) F = b F (b +) – (c – 1) F (c –)
(xiv) [1 – b + (c – a – 1) z] F = (c – b) F (b –) – (c – 1) (1 – z) F (c –)
(xv) [c – 1 + (a + b + 1 – 2c) z] F = (c – 1) (1 – z) F (c –) – c– 1 (c – a) (c – b) z F (c +)
d n d
Again since z
dz
 
z  n z n , writing   z , we have
dz

 
 z n  nz n and ( + a) zn = (n + a) zn .....(2)


Hence ( + a) F =   n  a  n ......(3)
n0


an
Using the relation F (a +) =   a
 n

n0
( + a) F = a F (a +) .....(4)
 
an bn
Similarly from F (a +) =    n and F  b     
a  b
 n
n0 n0  
( + b) F = b F (b +) .....(5)
and ( + c – 1) = (c – 1) F (c –) ......(6)
Proof. (i) Subtracting (5) from (4), we obtain (i) i.e.,
(+ a) F– (+b) F = aF (a+) – b F(a+)
 (a–b) F = aF (a+) – bF(b+)
(ii) Subtracting (6) from (4), we have
(+a) F– (+ c – 1) F = aF (a+) – (c – 1) F (c –)
 (a– c + 1) F = aF (a+) – (c – 1) F (c –)
208
(iii) We know that (zn) = n zn,
 n  a  n  b n  n  a  n  b n
 F   z n  z zn
n 0  c n n n 1  c n n 1


 a n1  b n1 n 
 a  n  a n  b  n  b n z n
 z z  z 
n 0  c n1 n n 0  c  n  c n n


 a  n  b  n   
 F  z  n .....(7)
n 0 c  n
 a  n  b  n   n  a  b  c 
 c  a  c  b  ...
But .....(8)
c  n c  n
he above equation (7) with the help of (8) is transformed to
 
 c  a  c  b  
c n
F  z  nn   a  b  c  z   n  z cn
n 0 n 0 c n 0

 z F   a  b  c  zF  c 1  c  a  c  b  z F  c   ,

or 1  z   F   a  b  c  zF  c 1  c  a  c  b  z F  c   .....(9)

Also from (4), we have F = –a F +a F(a +)

which implies that (1– z) F = –a (1– z) F +a (1– z) F (a +) .....(10)
From (9) and (10), we have.
 a 1  z    a  b  c  z  F  a 1  z  F  a    c 1  c  a  c  b  z F  c  

or  a   b  c  z  F  a 1  z  F  a    c 1  c  a  c  b  z F  c   . .....(11)

 a  1n  b n

zn 
 a  1n1  b n1 n1
(iv) Consider  F a     z .....(12)
n 1  c n n 1 n  0  c n 1 n
  a  1n 1   a  1 a  n

 a  1 a n  b  n  b n
 F  a    zn
n0  c  n  c n n

  a  1 z 

b  n  
n .....(13)
n  0 c  n

bn
 1
c  b
Since
cn cn
Putting this value in the above relation (13), we get
 
 c  b  
F  a     a  1 z   1  
n0  c  n   n
209
 
  a  1 z   n 
 a  1 c  b  z   c 
c
   c  n    n
n0 n0  
F  a     a  1 z F  c 1  a  1 a  b  z F  c   .....(14)
But in equation (4), if we write (a – 1) in place of a, we get
F  a     a   F   a  1 F  a   .....(15)
Combining the equations (14) and (15), we get the required result (iv).
(v) If we interchange aand b in (iv), we obtain (v).
The remaining ten relations can be deduced by making use of the above five relations.

11.6 Kummer’s Confluent Hypergeometric Function

The hypergeometric differential equation is
d 2u du
z 1  z  2
 c  1  a  b  z  abu  0 .....(1)
dz dz
Replacing z by z /bin (1), we get
2
 z  d u   1  a   du
z  1   2  c   1   z   au  0 .....(2)
 b  dz   b   dz
Now take the limit as b   , the equation (2) reduces to
d 2u du
z 2
  c  z   au  0 .....(3)
dz dz
 z
whose solution is given by lim 2 F1  a, b; c;  .....(4)
b 
 b
The equation (3) is known as the confluent hypergeometric differential equation or Kummer’s
equation.
 b r b  b  1 b  2  .... b  r  1
Now, lim r
 lim
b b b b.b.b.... r times

 1  2   r  1 
 lim  1  1   ... 1   1
b 
 b  b   b 
Hence the solution (4) may be written as
r
 z 
 a r  b  r z
lim 2 F1  a, b; c;   lim   
b r  0  c r r b
b b


 a r z r  b r
 lim
b

r  0  c r r br


 a r z r
   1F1  a; c; z 
r  0  c r r
The function 1F1(a; c; z) is called the confluent hypergeometric function.
210
 Now considering the equation (3), we find that z = 0 is a regular singular point, so if c is neither
zero nor a negative integer, two independent solutions in series of it can be easily found by Frobenius
method described in unit 9
 u1 = 1F1(a; c; z)
u2 = z1–c 1F1(a – c + 1; 2 – c; z)
Hence the general solution of equation (1) is
u  A 1 F1  a; c; z   Bz11 F1  a  c  1; 2  c; z 
where A and B are arbitrary constants.

11.6.1 Convergency of the Confluent hypergeometric function.

If un and un–1 are the nth and (n + 1) th terms of the series representing confluent hypergeometric
function, then

 a n xn  a n1 x n 1
un  and un 1  
 c n n  c n 1 n 1

un1  a n1 n  c n
  x
un  a n n  1  c n 1


 a  n  x
 c  n  n  1
un1
 lim
 a  n  x 0
 lim
n un n  c  n  n  1
un 1
Hence < 1 for all z . Thus the series is always convergent.
un

11.6.2 Differentiation of Confluent hypergeometric function.

Results :
d a
(i) 1 F1 (a; c; x )  1 F1 (a  1; c  1; x )
dx c
dn  a n
(ii) 1 F1 (a; c; x)  1 F1 ( a  n ; c  n ; x )
dx n  c n
The proofs of above formulas are similar to formulas given in §11.3 for Gauss hypergeometric
function.

11.6.3 Integral representation for confluent hypergeometric function

If | z | < 1 and Re (c Re a then
1
a 1 c  a 1 zt
B (ac – a) 1F1 (ac; z)   t 1  t  e dt
0

211
 1
c
a 1 c  a 1 zt
or F (ac z)   a  c  a  t 1  t  e dt
1 1     0

1  zt n
 c a 1 c a 1  
Proof we have R.H.S.   t 1  t   dt
  a   c  a  n0 n
0


 c zn  a  n  c  a
 
  a   c  a  n  0 n  c  n
= 1F1 (a; c; z)

11.6.4 Kummer’s first transformation

Result :
If c is neither zero nor a negative integer, then 1F1 (a c z) = ez 1F1 (cac z).
Proof : By integral representation of confluent hypergeometric function, we have.
1
c a 1 zt
B (acc1F1 (ac z)   t a 1 1  t  e dt
0
Using the property of definite integral, we get
1
e z
t
c  a 1
1  t c1 e zt dt
0
z
= e B c–aa 1F1 (c–a; c; – z)
(

 F (a; c; – z) = ez 1F1 (a–c; a; – z)

1 1

Ex.1. If m is a positive integer, show that

c a
x1 c  1  x  dm
F
2 1   m , a  m ; c; x  
 m  c
 c
dx m
x c  m 1
 1  x a  c  m 
and deduce that
1 1
a 1  a
 a 1 1 1     1
2
   m 2 4
 2 2 d
a 1
m 
2 F1   m , a  m;

 ;   
2 2 2 2  m 1 a  d  m
2
 1 2 2
 
2    m
2 2 

ca r
x1c 1  x  d m   c m1  c  a  m  r x 
Sol. R.H.S.   c  x 
 m  c dx m  r  0 r 

c a
x1c 1  x  d m   c  m  r 1  c  a  m r 
  x 
 c m dx m  r  0 r 

212
 x1c 1  x 
c a 
 c  r m  c  a  m r xc r 1

 c m  r
r 0

ca
  c  m r  c  a  m  r
 x1c 1  x   x c  r 1
r  0  c r r

ca
 1  x  F
2 1 c  m , c  a  m ; c ; x 
But we know by transformation formula
c a b
2
F1 (a,bcz)  1  z  F
2 1 c  a , c  b ; c ; z 
 R.H.S. = 2F1 (– m, a+ m;c x) = L.H.S.
1  1 a
Deduction. Putting x  and c  , we obtain the second part of the question.
2 2
Ex.2. If m is a positive integer, and | x | > 1, show that
m
 1  x m 1 d m  1 
 m 1 m  2 1    .
2 F1  , ;1; 2 
 2 2 x  m dx m  x 2  1 

Sol. We know that

- 1/ 2
1
=
1 1æ ç 1 ö÷
= ç1 + ÷
1/ 2
æ 1 ö÷1/ 2 x çè x 2 ø÷
( 1+ x2 ) x çç1 + 2 ÷
çè x ø÷

1  2 r 1
  1r   x
1  2 r
 
2 1/ 2 r
1  x  r 0

dm    1r  1  2 r 1 
dm 1/ 2
Hence
dx m

1  x2   m
dx

 r 0 r
  x
 2 r




 1r  1  2 r  m 1
    2r  1 2r  2  ......  2r  m  x
r 0 r  2 r

1 2r  m1

  2   2r  1 m x
m r  r
   1
r 0 r

 m 1   m 
2m   r     r   1
  2r  m  1  2   2 
But (2r + 1)m = 
  2r  1  
  r     r  1
 2
Putting the value of (2r + 1)m in the above relation
213
 1 m  m 1  m  2 r m1
m   1m r   2   r      r   1 x
d 1/ 2  2 r  2 2  2 
dx m 
1  x2     1
r 0 r   r     r  1
 2

m 1  m  2 
  1mr     m x 2 r m 1
 2 r  2 r

r 0 r r
(Again applying Legendre’s duplication formula)
m 1  m  2 
 1m x m1 dm   12 m   
r 2 r
  1 x
2 1/ 2  2 r  2 r

m dx m
1  x    1r r
r 0

 m  1 m  2 1 
= 2 F1  , ;1; 2 
 2 2 x 
Ex.3. Prove that
dm
 x  1 m 2 F1  a , b; c; x     a  m x a 1 2 F1  a  m , b; c; x 
m 
dx

d m  a 1 m 
Sol. L.H.S.  m 
x 2 F1  a, b; c; x  
dx

dm    a  b 
  r r
 x a  m  r 1 
dx m  r  0  c r r 

d m a  m  r 1
But
dx m
x     a  m  r  1 a  m  r  2  ....  a  r  x a  r 1

 a  m  r  a  m a  r 1
  a  r  m x a  r 1  x
 a r
 b 
 L.H.S.    c  r r   a  m r  a m x a  r 1
r 0 r

  a m x a 1 2 F1  a  m, b; c; x 
= R.H.S.
Ex4. Prove that If a b c 0, then
 c  a  b  c
lim
x 1
 1  x  a  b c
F  a , b; c; x  
2 1    a    b

Sol. L.H.S.  lim 1  x 

x 1
 a bc
2 F1  a, b; c; x 
214
 Now applying the transformation formula of
c  a b
2 F1  a, b; c; x   1  x  2
F1  c  a, c  b; c; x 

 L.H.S.  lim 1  x 
x 1
 a b c
1  x c a b 2 F1  c  a, c  b; c; x 

 lim  2 F1  c  a, c  b; c; x 
x1

 2 F1  c  a, c  b ; c; L 

 c  a  b  c
 (applying Gauss’s theorem)
  a   b
= R.H.S.

1
1 F1  a , b; z   e  t t a 10 F1  ; b; zt  dt
  a  0
Ex.5. Prove that

 r r
1  t a 1
 z
  t 
Sol. R.H.S.  
 a  0
e t  dt
r 0  b  r r

1 
 z r  t ar 1

 a
  e t dt
r 0  b  r r 0

r
1  z
    a  r  dt
  a  r  0  b r r

 1F1  a, b; z  = L.H.S.

Self-Learning Exercise

d2
1.  2 F1  a, b ; c ; x    .....
dx 2

 x
2. blim 2 F1  a, b ; c ;   .....
  b

 x
3. alim 1 F1  a , c ;    .....
  c
4. Write the Kummer’s first transformation for 1F1

5. aF  a    bF  b    ......

6. xlim
1

1  x  abc
2 F1  a , b ; c ; x   .....

215
11.7 Summary

In this unit we established some important formulae such as differentiation formulas, contiguous
function relations, linear relations etc. for Gauss hypergeometric function introduced in the last unit. We
also introduced and studied Kummer’s confluent hypergeometric function.

11.8 Answers to self-Learning Exercise

a  a  1 b  b  1
1. 2 F1  a  2 , b  2; c  2; x 
c  c  1

2. 1F1  a; c; x 
1 c  / 2  c J
3. x   c1 2 z 
4. 1 F1  a; c; z   e z1F1  c  a; c;  z 

5.  a  b  F

 c    a  b  c 
6.   a   b

11.9 Exercise

d n  x n
1. Prove that  b n n 
e 1F1  a; b; x     1  b  a  n e  x1F1  a; b  n ; x 
dx

 x
2. Show that 1 F1  a; c; x   blim 2 F1  a; b; c; 
  b

d m  x m
3. Show that  c m m 
e 1F1  a; c; x     1  c  a  m e x1F1  a; c  m; x 
dx
(Hint. Use Kummer’s first transformation)
x
4. If incomplete gamma function is defined by   a, x    et t a 1dt , Re  a   0.
0

Show that   a, x   a 1x a1F1  a; a  1;  x  .

5. State Confluent hypergeometric differential equation and explain its solution,
6. Prove that

1 a   
  z        
 1  zt  a    1 zt   2  , Re   0, Re z  0.
 t e F
1 1 a ; ;  dt 
1  a 1  
   
 2 2     
0   
 2   2 
216
 [Hint First replace 1F1 by its integral representation, then change the order of integration, Evalu-
ate the inner integral in terms of the gamma function. Write down the remaining integral in terms
of

   a 1   1 
  
   a 1   2  2
2 F1 
, a; ;1 
 2    1  1
   
 2   2 

a c  b x
7. Prove that F  a, b  1; c  1; x   F  a, b; c; x   F  a  1, b  1; c  2; x 
c  c  1
8. Prove the following relations :
1  b  x F
(i) F  a  1, b  1; c; x   F  a, b  1, c; x    a, b; c  1; x 
b
(ii) aF  a  1, b; c; x    c  1 F  a, b; c  1; x    a  1  c  F  a, b; c; x 
9. Show that
(i) e x  1  x F 1, 2; x 

 x
(ii)  1   e x  F  a  1; a; x 
 a
10. Prove the following relations
(i) bF  a; b; x   bF  a  1; b; x   x F  a; b  1; x 

(ii) aF  a  1; b; x    b  1 F  a; b  1; x    a  b  1 F  a; b; x 



217
Unit 12 : Legendre’s Polynomials and Functions Pn(x) and Qn (x)

Structure of the Unit

12.0 Objective

12.1 Introduction

12.2 Legendre’s equation and its Solution

12.3 Definition

12.3.1 Legendre’s Polynomial of Degree n or Legendre’s Function of First Kind

12.3.2 Legendre’s Function of Second Kind

12.3.3 Values of Pn(x) for n = 0, 1, 2, 3, 4, and 5

12.4 Generating Function for Pn(x)

12.5 Rodrigue’s Formulae for Pn(x)

12.5.1 Alternative form of Rodrigue’s Formula

12.5.2 Application

12.6 Orthogonal Property for Pn(x)

12.7 Recurrence Formulas for Pn(x)

12.8 Cristoffel Expansion

12.8.1 Cristoffel Summation Formula

12.9 Expression for Pn(cos ) in Terms of Cosine Series

12.10 Recurrence Formulae for Qn(x)

12.11 Cristoffel’s Second Summation Formula

12.12 Relations Between Pn(x) and Qn(x)

12.13 Summary

12.14 Answer to Self-Learning Exercise

12.15 Exercise

12.0 Objective

Our aim of this unit is to develop the Legendre Polynomials and to discuss its important

218
properties.

12.1 Introduction

Legendre polynomials may be introduced either through solution of a differential equation or

through a generating funcition. We shall discuss both the methods. Legendre polynomials have many
applications to mathematical physics and these applications depend on a number of special properties
which Legendre polynomials possess.

12.2 Legendre Equation and its Solution

The differential equation of the form

d2y dy

1  x2
dx 2
2 x
dx
 n  n  1 y  0 .....(1)

is called Legendre’s equation, where n is a positive integer. This equation has regular singular points at
x =  1 and x = , whereas all other points are ordinary, one of which be chosen as x = 0 since all
other ordinary points may be transferred at the origin.
The solution of equation (1) in series of descending powers of x can be referred to example
1§ 9.5 of unit 9.
However for sake of completeness we here reproduce the solution of (1).
Let the solution of (1) be

y   ar x k r , a0  0 .....(2)
r 0


dy   a x k r 1  k  r 
then r
dx r 0

d 2 y  a  k  r  k  r  1 x k r 2
and  r
dx 2 r 0

dy d2y
Putting the values of y, and 2 in (1), we get
dx dx
  
1  x   a  k  r  k  r  1 x
2

r 0
r
k r  2
 2 x  ar x k r 1  k  r   n  n  1  ar x k r  0
r 0 r 0

 
k  r 2
or  ar  k  r  k  r  1 x   n  n  1  2  k  r    k  r  k  r  1 ar x k r  0
r 0 r 0

 
or  ar  k  r  k  r  1 x k r 2   ar  n  k  r  n  k  r  1 x k r  0 .....(3)
r 0 r 0

Equating to zero the coefficient of the highest power of x namely xk in (3), we get

219
 a0 (n – k)(n + k + 1) = 0
 k = n, –(n + 1) ( a 0  .....(4)
The next lower power of x is k – 1, so we equate to zero the coefficient of xk – 1 in (3) and
obtain
(n – k + 1) (n + k) a1 = 0 .....(5)
For k = n and –(n + 1), neither (n – k + 1) nor (n + k) in zero. thesefore a1 = 0
Next equating to zero the coefficient of xk – r in (3), we have
(k – r + 2) (k – r + 1) ar –2 + (n – k + r) (n + k – r + 1) ar = 0

ar  
 k  r  2  k  r  1 a .....(6)
 n  k  r  n  k  r  1 r 2
Putting n = 3, 5, 7... in (6) and noting that a1 = 0, we have
a1 = a3 = a5 = a7 = ..... = 0 .....(7)
To obtain a2, a4, a6 .... etc, we consider following two cases
Case I. When k = n then (6) becomes

ar  
 n  r  2  n  r  1 a .....(8)
r 2
r  2n  r  1
Putting r = 2, 4, 6, .... in (8), we have
n  n  1
a2   a0
2  2n  1

a4  
 n  2  n  3 a  n  n  1 n  2  n  3 a
2 0
4  2n  3 2  4  2n  1 2n  3
and so on
Re-writing (2), we have for k = n
y = a0xn + a1xn – 1 + a2xn – 2 + a3xn – 3 + a4xn – 4 + ... ....(9)
Using (7) and the above values of a2, a4, a4, etc in (9) we get

 n  n  1 n 2 n  n  1 n  2  n  3 n4 
y  a0  x n  x  x  .... .....(10)
 2  2n  1 2  4  2n  1 2n  3 
Case II. When k = –( n + 1) then (6) becomes

ar 
 n  r  1 n  r  a .....(11)
r 2
r  2n  r  1
Putting r = 2, 4, 6, .... etc., we get

a2 
 n  1 n  2  a
0
2  2n  3 
 n  3 n  4   n  1 n  2  n  3 n  4  a
a4  4 2n  5 a2  0
  2  4  2n  3 2n  5 
and so on.

220
 For k = – ( n + 1), (2) gives
y = a0x–n – 1 + a1x–n – 2 + a2x–n – 3 + a3x–n – 4 + a4x–n – 5 +... ...(12)
Using (7) and the above values of a2, a4, a4, etc. in (12),we find that


y  a0  x  n1 
 n  1 n  2  x n 3   n  1 n  2  n  3 n  4  x  n5  ...
 ....(13)
 2  2n  3 2  4  2n  3 2n  5  
Thus two independent solutions of (1) are given by (10) and (13). If we take
1  3  5...  2n  1
a0 
n
then solution (10) is denoted by Pn(x) and is called Legendre polynomial of first kind and if we take
n
a0  then solution (13) is denoted by Qn(x) and is called Legender polynomial of
1  3  5... 2n  1
second kind so the general solution of (1) is
y  A Pn  x   B Qn  x 
where A and B are arbitrary constants

12.3 Definition

12.3.1 Legendre’s polynomial of degree n or Legendre’s function of first kind

Legendre’s polynomial of degree n is denoted and defined by

1  3  5...  2n  1  n n  n  1 n 2 n  n  1 n  2  n  3 n4 

Pn  x   x  x  x ...
n!  2  2n  1 2  4  2n  1 2n  3  

n 2  2n  2r 
r
   1 n
x n 2 r , .....(1)
r 0 2 r  n  r   n  2r 

 n    n 2 , if n is even
where  2   n  1 2, if n is odd, .....(2)
  

12.3.2 Legendre’s Function of Second Kind

This is denoted and defined by

n!
Qn  x  
1  3...  2n  1

  n 1  n  1 n  2   n 3  n  1 n  2  n  3 n  4   n 5 

x  x  x  ... .....(3)
 2  2n  3  2  4  2n  3 2n  5  

12.3.3 Values of Pn(x) for n = 0, 1, 2, 3, 4 and 5

221
 Putting n = 0, 1, 2, 3, 4, and 5 in (1), and simplying the expression thus obtained we easily find that

P0(x) = 1, P1(x) = x

1 1
P2(x) 
2
 
3 x 2  1 , P3  x   5 x 2  3 x
2
  .....(4)

1
P4(x) 
8

35 x 4  30 x 2  3 and  .....(5)

1
P5(x) 
8

63 x5  70 x3  15 x  .....(6)

12.4 Generating Function for Pn(x)

1/ 2 
Result. Show that 1  2 xh  h 2
  
n 0
hn Pn  x , | x |  1, | h |  1

or show that Pn(x) is the coefficient of hn in the expansion of the (1 – 2xh + h2) in ascending

powers of h. (1 – 2xh + h2)–1/2 is called generating function for Legendre polynomial Pn(x).

Proof. Since | h | < 1 and | x |  1, we have

1/ 2
(1 – 2xh + h2)–1/2  1  h  2 x  h  

1 1 3 2 2
 1  h  2x  h  h  2 x  h   ....
2 24
1  3....  2n  3 n 1 1 3....  2n  1 n
 h n1  2 x  h   h n  2 x  h   .... .....(1)
2  4....  2n  2  2  4....  2n 
Now, the coefficient of hn in
1  3....  2n  1 n 1  3  5....  2n  1
hn  2 x  h    2 x n
2  4....  2n  2  4  6....  2n 

1  3  5   2n  1 n
 x .....(2)
n
Again the coefficient of hn in
1  3....  2n  3 n 1 1 3....  2n  3
h n1  2 x  h      n  1 2n2 x n2 
2  4....  2n  2  2n1
1 2  3....  n  1  

1  3....  2n  1 n  n  1
=  .....(3) x n 2
n 2  2n  1
and so on. Using (2), (3), ....., we see that coefficient of hn in expansion of (1 – 2xh + h2)-1/2,
viz. (1) is given by
1  3  5 ....  2n  1  n n  n  1 n 2 n  n  1 n  2  n  3 n 4 
x  x  x  ... = P (x)
n  2  2n  1 2  4  2n  1 2n  3  n

222
 Thus we can say that P1(x), P2(x), .... . Will be coefficientts of h, h2, .... in the expansion of
(1 – 2xh + h2)-1/2. Hence we have
(1 – 2x + h2)-1/2 = 1 + h P1(x) + h2P2(x) + h3P3(x)+ ... + hnPn(x) +...

or (1 – 2xh + h ) 2 -1/2   hn Pn  x 
n0

12.5 Rodrigues Formula for Pn (x)

1 dn n
Result. Show that Pn ( x ) 
2n n dx n
x 2
1 
Proof. Let y = (x2 – 1)n
dy
 = n(x2 – 1)n – 1  2x
dx
Multiplying both sidees by (x2 – 1), we get
dy
= n(x2 – 1)n  2x = 2nxy
(x2 – 1)
dx
Differentiating (n + 1) times both sides of the above equation and using Leibnitz theorem, we
get

d n 2 y d n 1 y dny
 2
x 1 
dx n 2
 n1
c1  2 x  n 1 
dx
n 1
c2  2 n
dx
 d n 1 y dny 
 2n  x n 1  n 1c1 n  1
 dx dx 
Simplifying the above equation, we find that
d n 2 y d n1 y dn y
or 1  x  2
dx n 2
 2x
dx n1
 n  n  1 n  0
dx
.....(1)

dny
Let = z in (4). Then
dx n

d2z dz
1  x2  dx 2
 2x
dx
 n  n  1 z  0 .....(2)

Now (2) is Legendre’s equation and shows that z is a solution to this equation. Hence one of its
solution be

dny
z= = c Pn(x) .....(3)
dx n
where c is constant
To find c, put x = 1 in both sides of (3), therefore

223
 dny 
c Pn(1)   n 
 dx  x1

dny
 c n  [ Pn(1) = 1] .....(4)
 dx  x 1
Again y = (x2 – 1)n = (x – 1)n  (x + 1)n
Differentiating both sides n times by Leibnitz’s theorem, we get
n n n
dn y d n  x  1 d n 1  x  1 d n  x  1
dx n
  x  1
n

dx n
n
dx n 1
n  x  1   ...   x  1
n 1 n

dx n
Now putting x = 1 in both sides of above relation, we see that all the terms in RHS except the
last term vanishes since each term contains the factor (x – 1), and also
n
d n  x  1
n
dx n

 dny  n n
Thus  n   1  1 n  2 n .....(5)
 dx  x 1
Now using (5) in (4), we find that
c  2n n
Substituting the values of y and c in (3), we easily arrive at the Rodrigue’s formula.
12.5.1 Alternative form of Rodrigue’s formula
We have

1 dn
Pn(x)  n
2 n dx
 n  x  1  x  1 
n n

By Leibnitz’s rule we have


1 n n
Pn(x)  n
2 n
 ncr Dnr  x  1 D r  x  1
r 0

 r nr
2 x 1   x  1 

r 0
cn
r   
 2   2 
 .....(6)

12.5.2 Application

tn
Multiplying (6) by and summing from n = 0 to , we get
 n 2
 Pn  x  t n  n 2
 1   x 1   x  1 
r nr
 2    n  r r   2   2  t
n
n0  n n0 r 0  

224
   r n
1  x  1   x  1  n r
  2 2   2 
t
n0 r 0  n 2  r    

  n   
 using   A k, n    A  k , n  k 
 n0 k 0 n0 k 0 

 x 1   x 1 
 0 F1  ;1; t  0 F1  ;1; t .....(7)
 2   2 

12.6 Orthogonal Property for Pn (x)

Result : Prove that

1

(i)  Pm  x  Pn  x  dx  0 if m  n
1

1
2 2
and (ii)   Pn  x  dx 
2n  1
if m  n
1

Proof. The Legendre equation is

d2y dy
 1  x2  dx 2
 2x
dx
 n  n  1 y  0

d  2 dy 
or  1 x
dx  dx 

  n  n  1 y  0  .....(1)

Now since Pm(x) and Pn(x) are solutions of (1), hence

d  2 dPm 
 1 x
dx 

dx 

  m  m  1 Pm  0 .....(2)

d  2 dPn 
and  1 x
dx 

dx 

  n  n  1 Pn  0 .....(3)

Multiplying (2) by Pn and (3) by Pm and substracting, we get

d  2 dPn  d  2 dPm 
Pm  x  
 1 x
dx  dx 

  Pn  1 x
dx  dx 
 
  n  n  1  m  m  1 Pn Pm  0

Integrating above w.r.t. x form –1to 1, we get

1 1 1
d  2 dP  d  2 dP 
   
 Pm  x  dx  1  x dxn  dx   Pn dx  1  x dxm  dx  n  n  1  m  m  1  Pn Pmdx  0
1 1 1

On integration by parts, we get

1 1 1
 2 dPn   dPm  2 dPn    2 dPm 
P
 m

 x  1  x 
dx  1 1  dx 
 1  x 
dx 
dx  P
 n


 x  1  x dx  1
 
225
 1 1
 dP  dPm  
   n  1  x2
 dx  dx


  dx  n  n  1  m  m  1  Pm Pn dx  0
1 1

1

or  n  m  n  m  1  Pm Pn dx  0
1

1

  Pm Pn dx = 0, m  n .....(4)
1

Case II. When m = n. From generating function


n
(1 – 2xh + h2)–1/2   h Pn  x  .....(5)
n0


(1 – 2xh + h2)–1/2   hm Pm  x  .....(6)
m0
Multiplying the corresponding sides of (5) and (6), we get
 
(1 – 2xh + h ) 2 –1   Pn  x  Pm  x  hmn
n0 m0
Integrating both sides of the above with respect to ‘x’ from – 1 to 1,we get
 1  m n 1 1
    Pm  x  Pn  x  dx  h   1  2 xh  h2
n0 m0  
  dx ..(7)
1 1

Making use of (4), (7) reduces to

  1 
2 1  1
  Pn  x  dx  h2n
 
2h 

log 1  2 xh  h 2  
 1
n 0 1  

1
  log 1  h   log 1  h  
h

1  h 2 h3   h 2 h3 
  h    ... 
   h    ...  
h  2 3   2 3  

2 h3 h5  
2
  h    ... 
 h2 n
h 3 5  n 0 2n  1
Equating coefficients of h2n from both sides, we get
1 2 2
1 Pn  x  dx 
2n  1

12.7 Recurrence Formulas for Pn (x)

226
12.7.1 ( n + 1) Pn + 1 (x) = (2n + 1) xPn(x) – nPn – 1(x), n  1,

or (2n + 1) xPn(x) = (n + 1) Pn + 1(x) + n Pn – 1(x)

Proof. We know that

n
(1 – 2xh + h2)–1/2   h Pn  x  .....(1)
n0
Differentiating (1) both sides w.r.t. h, we get

1 3 2 n 1

 1  2 xh  h 2  (– 2x + 2h)   nh Pn  x 
2 n0


or (x – h) (1 – 2xh + h ) 2 –1/2
= (1 – 2xh + h ) 2
 nhn1Pn  x 
n0

 
or (x – h) h n
Pn  x  = (1 – 2xh + h ) 2
 nhn1Pn  x 
n0 n0

Equating coefficients of hn from both sides, we get

or xPn(x) – Pn – 1(x) = (n + 1) Pn + 1(x) – 2xn Pn(x) + (n – 1) Pn–1(x)
or (2n + 1) xPn(x) = (n + 1) Pn + 1(x) + nPn – 1(x)
12.7.2 nPn ( x )  xPn ( x )  Pn1 ( x )
Proof.
Differentiating (1) w.r.t.’h’, we get

1 3 2 n 1

2
1  2 xh  h 2  (– 2x + 2h)   nh Pn  x 
n0


or (x – h) (1 – 2xh + h2)–3/2   nh n 1Pn  x  .....(2)
n0
Again differentiating (1) w.r.t.’x’, we find that

1 3 2

 1  2 xh  h 2  × (–2h)   h n Pn   x 
2 n 0


or h(1 – 2xh + h ) 2 –3/2   h n Pn'  x 
n 0
Multiplying by (x – h) on both sides, we get
3 2 

h(x – h) 1  2xh  h2    x  h   h n Pn'  x 
n 0

227
  
Using (2), we get  nhn Pn  x    x  h   h n Pn'  x 
n0 n 0
n
Equating coefficients of h from both sides of the above equation, we get
nPn (x )  xPn' (x )  Pn' 1 (x )

12.7.3 (2n + 1) Pn(x)  Pn' 1 ( x )  Pn' 1 ( x )n

Proof. From recurrence formulas 12.7.1, we have
(2n + 1) xPn(x) = (n + 1) Pn + 1(x) + nPn – 1(x)
Differentiating it w.r.t.’x’, we get
(2n + 1) Pn(x) + (2n + 1) xPn(x) = (n + 1) Pn + 1(x) + nPn – 1(x) .....(3)
From recurrence 12.7.2, we have
nPn(x) = xPn(x) – Pn – 1(x)
or xPn' ( x )  nPn ( x )  Pn' 1 ( x ) .....(4)
Using (4) in (3), we get
(2n + 1) Pn(x) + (2n + 1) [nPn(x) + Pn – 1(x)] = (n + 1)Pn + 1(x) + nPn – 1(x)
or (2n + 1) (n + 1) Pn(x) = (n + 1) Pn + 1(x) – (n + 1) Pn – 1(x)
or (2n + 1) Pn(x) = P  n + 1(x) – P  n – 1(x)
12.7.4 (n + 1) Pn(x) = [P  n + 1(x) – xP  n(x)]
Proof. From recurrence formulae 12.7.2 and 12.7.3, we have
nPn(x) = xP n(x) – P n – 1(x) .....(5)
(2n + 1) Pn(x) = P n + 1(x) – P n – 1(x) .....(6)
Substracting, we get
(n + 1) Pn(x) = P  n + 1(x) – xP  n(x)
12.7.5 (1 – x2) P  n(x) = x [Pn – 1(x) – xPn(x)]
Proof. From recurrence formulae 12.7.2, we have
nPn(x) = xP n(x) – P n – 1(x)
Multiplying by x, we get nxPn(x) = x2P n(x) – xP n – 1(x) .....(7)
Replacing n by (n – 1) in formula 12.7.4, we have
nPn–1(x) = P n(x) – xP n – 1(x) .....(8)
Substracting (7) from (8), we have
x [Pn – 1(x) – xPn(x)] = (1 – x2)P  n(x)
12.7.6 (1 – x2) P  n(x) = (n + 1) [xPn(x) – Pn + 1(x)]
Proof. From recurrence formula 12.7.1, we have
(2n + 1) xPn(x) = (n + 1) Pn + 1(x) + nPn – 1(x)
or (n + 1) xPn(x) + xnPn(x) = (n + 1) Pn + 1(x) + nPn – 1(x)

228
 or (n + 1) [xPn(x) – Pn + 1(x)] = n[Pn – 1(x) – xPn(x)] .....(9)
From formula 12.7.5 we have
(1 – x2) P n(x) = n[Pn – 1(x) – xPn(x)] .....(10)
From (9) and (10), we easily get
(1 – x2) P  n(x) = (n + 1)[xPn (x) – Pn + 1(x)]
Self-Learning Exercise–I
(1) The solution of Legendre’s differential equation is known as ......

(2) (1 – 2xh + h2)–1/2   h n ......
n0

1
(3)  Pn  x  Pm  x  dx  ..... (if m  n)
1

(4) Pn(1) = ......

(5) Pn(x) is a polynomial of degree ......
n

(6) Pn  x  
1 
d n x2 1  is known as ......
n
2 n dx n
(7) x = ...... is an ordinary point for Legendre differential equations.
(8) The value of P2(x) is .......
(9)  n  1 Pn  x   Pn' 1  x   xPn'  x   .....
(10) if n is even/odd, then Pn(x) is ............ function of x.

12.8 Cristoffel’s Expansion

Result : Prove that

P  n = (2n – 1) Pn – 1+ (2n – 5) Pn – 3+ (2n – 9) Pn – 5+ .....


where P n  Pn(x) and P  n  P  n(x) .....(1)
The last term of the series will be 3P1 or P0 according as n is even or odd.
Proof : Replacing n by n – 1 in recurrence formula 12.7.3, we have
P n = (2n – 1) Pn – 1 + P n – 2 .....(2)
Writing n – 2, n – 4, and so on in place of n in (2), we find that
P n – 2 = (2n – 5) Pn – 3 + P n – 4
P n – 4 = (2n – 9) Pn – 5 + P n – 6
....... ........ ...........
....... ........ ........... .....(A)
P 3 = 5P2 + P 1
P 2 = 3P1 + P 0

229
 When n is even, then adding the relations in (A) and (2), we get
P n = (2n – 1) Pn – 1+ (2n – 5) Pn – 3+ (2n – 9) Pn – 5+ .....+ 3P1 ( P 0(x) = 0)
and when n is odd, then
P n = (2n – 1) Pn – 1+ (2n – 5) Pn – 3+ (2n – 9) Pn – 5+ .....+ 5P2 + P0
( P 1= 1 = P0)
12.8.1 Cristoffel’s Summation Formula
Result : Prove that
n  Pn 1  x  Pn  y   Pn1  y  Pn  x  
  2r  1 Pr  x  Pr  y    n  1  x y
 .....(3)
r 0  
Proof. Prom recurrence formula 12.7.1, we have
(2r + 1) xPr (x) = (r + 1) Pr + 1 (x) + rPr – 1 (x) .....(4)
(2r + 1) yPr (y) = (r + 1) Pr + 1 (y) + rPr – 1 (y) .....(5)
Now multiplying (4) by Pr(y) and (5) by Pr(x) and subtracting, we find that
(2r + 1)(x – y) Pr (x) Pr (y) = (r + 1) [Pr + 1 (x) Pr (y) – Pr + 1 (y) Pr (x)]
+ r[Pr – 1 (x) Pr (y) – Pr (x) Pr – 1 (y)] .....(6)
Taking r = 0, 1, 2 , .... n in (6) and adding the relations column-wise, we get the required result
(3).

12.9 Expression For Pn (cos ) in Terms of Cosine Series

we know that

2 –1/2
(1 – 2xt + t )   Pn  x  t n .....(1)
n0

e i   e  i
Taking x  cos  
2
in (1) we easily get

 Pn  cos   t n = (1 – tei)–1/2(1 – te– i)–1/2 .....(2)
n0

 1 1 3 2 2i 1 3.....  2n  1 n in 

 1  tei  t e  ...  t e  ...
 2 2 4 2  4  6.....2n 

 1 1  3 2 2i 1  3.....  2n  1 n in 

 1  te i  t e  ...  t e  ...
 2 24 2  4  6.....2n 
Now equating coefficients of t n both sides, we get
1  3  5....  2n  1 in 1 3  5....  2n  3  1 i n 2
Pn(cos )  2  4  6.... 2n e  e
 
  in

2

 4  6...  2 n  2 
 e
2
 e  
i n 2 

230
 1 3  5.....  2n  5 1  3 i  n 4 
 
2  4  6....  2n  4  2  4
e   e    .....
i n4 

1  3  5.... 2n  1
or Pn(cos ) =
1  4  6....2n

 n n  n  1 1 3 
×  2 cos n    2 cos  n  2      2 cos  n  4    ...
 2n  1  2n  1 2n  3 1 2 
the above formula is useful in obtaining the integrals involving the products of Pn(cos ) and sine and
cosine multiple of .

1 z 1 
Ex.1. Prove that     Pn  Pn1  z n
z 1  xz  z 2 z n0


Sol. We have RHS    Pn  Pn1  z n
n0

 
1
  z n Pn   z n1Pn1 .....(3)
n0 z n0


Also  z n1Pn1 = zP1 + z2P2 + z3P3 + .. .....(4)
n0


and  z n Pn = P0 + zP1 + z2P2 + z3P3 + .. .....(5)
n0
Substracting (5) from (4), we get
 
 z n1Pn1   z n Pn  P0 .....(6)
n0 n0
Using (6) in (3), we get

1  n
n 
RHS   z Pn    z Pn  P0 
n0 z  n 0 

 1  P
  1    z n Pn  x   0
 z  n0 z

 1 1 2 1
 z

  1   1  2xz  z 2  
z
 P0  1

= L.H.S.
Ex.2. Prove that Pn(1) = 1 and Pn(–1) = (–1)n

Sol. We have  hn Pn  x  = (1 – 2xh + h2)–1/2 .....(7)
n0
For x = 1, we have
231
 
 hn Pn 1 = (1 – h)–1
n0

= 1 + h + h2 + .... hn   h n
n 0

Equating coefficients of hn on both sides, we find that P1(1) = 1

Also for x = – 1, equation (7) gives

 hn Pn  1 = (1 + h)–1 = 1 – h + h2 – .... + (–1)nhn ....+ .....
n0

n
   1 h n
n0
Equating coefficients of h on both sides, we get Pn(–1) = (–1)n
n

Ex.3. Prove that

(2n + 1) (x2 – 1) Pn = n(n + 1) (Pn + 1 – Pn – 1)
and hence deduce that
1
2n  n  1
 x 
2
 1 Pn1  x  Pn  x  dx 
1
 2n  1 2n  3 
Sol. From recurrence relation 12.7.5 and 12.7.6, we have

(1 – x2) Pn' = n (Pn – 1 – xPn) .....(8)

(1 – x2) Pn' = (n + 1) (xPn – Pn + 1) .....(9)

Eliminating xPn from (8) and (9), we get

1  x  P  1  x  P
2
n
' 2
n
'

= Pn – 1 – Pn + 1
n  n  1

 n  11  x 2 Pn'  n 1  x 2  Pn'

or  Pn 1  Pn 1
n  n  1

or (2n + 1) (1 – x2) Pn' = n(n + 1) [Pn–1 – Pn+1]

or (2n + 1) (x2 – 1) Pn' = n(n + 1) [Pn+1 – Pn–1] .....(10)

This result is known as Beltrami’s relation.
Deduction
Multiplying both sides of (10) by Pn+1(x) and integrating w.r.t. ‘x’ from –1 to 1, we find that
1 1
n  n  1
 
x 2  1 Pn 1  x  Pn'  x  dx 
 2n  1 1
Pn 1  x   Pn 1  x   Pn 1  x   dx
1

Using orthogonal property for Legendre’s polynomials, we get the required integral
232
  n
1 x 
Ex.4. Show that Pn  x      x  1 cos   d 
2

 0

where n is a positive integer.
This result is also known that Laplace’s first integral for Pn(x) .
Proof. We know that

d 
 a  b cos  , we have a2 > b2 .....(11)
2 2
0 a b

Taking a = 1 – hx and b = h x 2  1 , then

a2 – b2 = (1 – hx)2 – h2(x2 – 1) = 1 – 2xh + h2

d 
Thus (11) becomes  
0 1  hx   h  x 1 cos 
2 1  2 xh  h 2


2 –1/2 dθ
or (1 – 2xh + h ) 
0 1  hx   h  x 1 cos θ
2

  1

or
n0
n 
π  h Pn  x    1  h x 
0 
   
x 2  1 cos θ  dθ



1 2
  1  ht  d , where t  x  x  1cos 
0



  1  ht  h 2t 2  ...  h nt n  ... d 
0

 
   h n t n d
n0 0
n
Equating coefficients of h from both sides, we get

n
Pn(x)   t d
0

 n
1
or Pn(x)    x  x 2  1 cos   d
 0 

Ex.5. Prove that


2 2n  n  1
 x Pn1 Pn1dx   2n  1 2n  1 2n  3 
0

Proof. From Recurrence formulae 12.7.1 we have

233
 (2n + 1) xPn = (n + 1) Pn + 1+npn – 1
Put (n – 1) and (n + 1) in place of n respectively, we get
(2n – 1) xPn – 1 = nPn +(n – 1) Pn – 2 .....(12)
(2n + 3) xPn + 1 = (n + 2) Pn + 2 +(n + 1) Pn .....(13)
Multiplying (12) and (13), we get
(2n – 1) (2n + 3) x2Pn – 1 Pn + 1 = n(n + 2) Pn Pn + 2 + n(n + 1) P2n
+ (n + 2) (n – 1) Pn + 2 Pn – 2 + (n2 – 1) Pn Pn – 2
Integrating w.r.t. x between limit –1 to +1, we have
1 1
 2n  1 2n  3  x 2 Pn1Pn1dx  n  n  1  Pn2dx
1 1

1

(other integrals on the RHS vanish due to integral  Pm Pn dx  0 if m  n )

1
1
2n  n  1
or  2n  1 2n  3  x 2 Pn1Pn 1dx 
1
 2n  1
1
2n  n  1
or  x 2 Pn1 Pn1dx 
1  2n  1 2n  1 2n  3 
1
2n
Ex.6. Show that  xPn Pn1dx  4n2  1
1

Proof. From Recurrence relation 12.7.1 we have

(2n + 1) xPn = (n + 1) Pn + 1 + nPn – 1 .....(14)
Multiplying (14) by Pn – 1 and then integrating w.r.t. x from –1 to +1, we get.
1 1 1
2
 2n  1  xPn Pn1 dx   n  1  Pn 1 Pn1 dx  n   Pn1  dx
1 1 1
Using orthogonal property for Legendre polynomial, we get
1
2n
 2n  1  xPn Pn 1 dx 
1
 2n  1
1
2n
  xPn Pn 1dx   4n2  1
1


1 d
Ex.7. Prove that Pn  x    n1
.....(15)
0
x x 2
 1 cos  

 

Sol. Taking a = xt – 1 and b  t x 2  1, then a2 – b2 = 1 – 2xt + t2

234
 1 2 
   2x 1 
  1       Pn  x  t  n1 for large t .....(16)
a 2  b2 t  t t2  n 0

1 1
Also

a  b cos  
2

 t x  x  1 cos   1
 
1
 
1
 1 
 t
  2
x  x  1 cos  
 
1  
2
 t x  x  1 cos 
   



t  n 1
 n 1
n0  x  x 2  1cos  .....(17)
 

Now integrating (17) both sides w.r.t.  in (0, ), we get

 
   n1  
t d 
  n 1
d    
0  n 0  x  x 2  1 cos   0
a  b cos  a  b2
2
   

Using (16) in the above expression, we find that

 
  d  n 1 
  n 1
 t    Pn  x  t n1 .....(18)
n0  0  x  x 2  1cos   n0
   

Equating coefficicents of t–n–1 in (18), we get the required integral (15).

Remark. The integral given by (15) is known as Laplace’s second integral.


Ex.8. Evaluate  Pn  cos   cos n d 

0

Sol. By §12.9, we have

1  3  5 ...  2n  1  n
Pn (cos ) 
2  4  5 ... 2n  2cos n   2n  1 2cos  n  2  

n  n  1 1 3 
 2 cos  n  4    ...
 .....(19)
 2n  1 2n  3 1 2 
Multiplying (19) both sides by cos n  and integrating w.r.t  in (0, ) we get

235
 1  3  5 ...  2n  1   2n
I   2 cos 2 n    cos n  cos  n  2   
2  4  5 ... 2n 0  2n  1

2n  n  1 1 3 
  cos n  cos  n  4    ... d 
 2n  1 2n  3 1  2 
Using the following orthogonal property for cosine function

0, m n
 cos m cos n d  
 2, m  n
,
0

1 3 5  2n  1 
  ...  
1  3  5 ...  2n  1  2 2 2  2 
we find that I 2  
2  4  6 ... 2n 2 1 2  3 ... n

 1 1
n   
2 2  1 1
   Bn  , .
  n  1  2 2

12.10 Recurrence Formulae for Qn (x)

We have already defined that

2
2 n  n    n1  n  1 n  2   n 3 
 Qn (x) = x  x  ...... .....(1)
2n  1  2   2n  3 
Again above relation can be written as

n  2r x  
 n 2r 1
2n n 
Qn (x) = 
2n  1 r  0 2r r  2n  3 .....  2n  2r  1
.....(2)

Differentiating (2) with respect x, we get

2n n   n  2r  1 x  
 n 2 r  2

Qn (x) =  
2n  1 r  0 2r r  2n  3 .....  2n  2r  1
.....(3)

Putting n – 1 for n, then we get

2n  n    n  2r  x  
 n 2 r 1

Qn–1(x)   
2n  1 r  0 2r r  2n  3 .....  2n  2r  1
.....(4)

Again putting n + 1 for n in (3), we get

2n n   n  2r  2  x   n  2 r 3

Qn +1(x)   
2n r  0 2r r  2n  1 2n  3 .....  2n  2r  3
.....(5)

12.10.1 Qn 1  Qn 1   2n  1 Qn

236
Proof. Using (1) and (4) above, we get

n  2r x  
 n  2 r 1
2n n 
Qn –1+(2n + 1) Qn = – 
2n  1 r  0 2r r  2n  3 ......  2n  2r  1

n  2r x  
 n 2 r 1
2n n 
  2n  1  
2n  1 r  0 2r r  2n  3 .....  2n  2r  1

n  2r x  
 n  2 r 1
2n n 
   2n  2r  1   2n  1 
2n  1 r  0 2r r  2n  3 ....  2n  2r  1 

n  2r x  
 n 2r 1
2n n 
    2r 
2n  1 r  0 2r r  2n  3 .......  2n  2r  1

n  2r x  
 n 2r 1
2n n 
 
2n  1 r 1 2r 1 r  1 2n  3  ......  2n  2r  1

Putting r – 1 = s  r = s + 1, therefore

n  2s  2 x  
 n  2s  3
2n n 
Qn –1+(2n + 1) Qn   
2n  1 r  0 2s s  2n  3 ......  2n  2s  3

 Q n' 1  x   LHS

12.10.2 nQn 1   n  1 Qn 1   2n  1 x Qn

Proof. Using (1) and (4) above, we get

2 n n  n  2r  1 x  n 2r 2 
(2n + 1)x Qn– (n + 1) Qn-1 
 2 n  1 x  
2n  1 r  0 2r r  2n  3 ......  2n  2r  1

2n n  n  2r x  n  2r 1
  n  1 
2n  1 r  0 2r r  2n  3 ......  2n  2r  1

 1 2n n n  2r x  
  n  2 r 1
 
2n  1 r  0 2r r  2n  3 ......  2n  2r  1

 2n  1 x  n  2r  1 x 1   n  1   2n  2r  1 
 

 1 2n n n  2r  1 x  
  n 2r 1
 
2n  1 r  0 2 r r  2n  3 .......  2n  2r  1
 2nr

237
   n  2n n n  2r x  
  n 2 r 1
 
2n  1 r 1 2r 1 r  1 2n  3 .......  2n  2r  1

  n  2n n  2s  2 x  
  n  2 s 3
n

2n  1
 2s s  2n  3 ......  2n  2s  3
s 0

 nQ n' 1  x   L .H .S .

12.10.3  2n  1 x Qn   n  1 Qn1  nQn1

Proof. Integrateing the recurrence relation 12.10.2 w.r.t. x from x to , we get
 
x  nQ n' 1   n  1 Q n' 1  dx   2n  1  xQ n' dx
  x

    
or  nQn 1   n  1 Qn 1  x   2n  1  xQn  x  x Qn  x  dx 
 


Q n' 1  Q n' 1  dx
  2n  1 
  
xQ n x   2n  1  (by relation 12.10.1)
x  2n  1
  
  2n  1  xQn  x  Qn 1  x   Qn 1  x
The value of Qn –1, Qn or Qn + 1 is zero when x is infinity since they contain only negative integral
power of x, therefore
–nQn+1 – (n + 1) Qn–1 = – (2n + 1) x Qn + Qn + 1 – Qn – 1
Solving it we easily get the required ralation 12.10.3

12.10.4  2n  1  1  x 2  Qn  n  n  1  Qn1  Qn1 

Sol. Since Qn is a solution of Legendre’s equation, namely
d  dy 

dx 
1  x2 
dx 

 n  n  1 y  0

d 
Therefore
dx    
1  x 2 Q n'    n  n  1Q n .....(5)
Integrating w.r.t. x both sides of (5) between the limits,  to x, we have
x
 1  x 2 Q '   n  n  1 x Q dx
 
 n
  n
x
or 1  x Q
2 '
n x   n  n  1  Qn dx

......(6)
Integrating both sides of recursence ralation 12.10.1 between the limit  to x, we get
x
Qn 1  Qn1    2n  1 Qn dx .....(7)

Now, from (6) and (7), we get

238
  Q  x   Qn 1  x  
1  x Q
2 '
n x    n  n  1  n1
 2n  1

 
 2n  1 1  x 2 Q n'  x   n  n  1 Qn1  x   Qn1  x  

12.11 Cristoffel’s Second Summation Formula

n
Result.  y  x    2r  1 Pr  x  Qr  y 
r 1

 1   n  1   Pn 1  x  Qn  y   Pn  x  Qn 1  y   ......(1)
Proof : From recurrence formulas for Pn(x) and Qn(x), we have
(2n + 1) x Pn (x)   n  1 Pn 1  x   nPn1  x  .....(2)

(2n + 1) y Qn (y)   n  1 Qn1  y   nQn 1  y  .....(3)

Multiplying (2) by Qn (y) and (3) by Pn(x) and subtracting, we have
(2n + 1 ) (x – y) Pn (x) Qn (y) + n {Pn–1 (x) Qn–1 (y) – Qn-1 (y)Pn (x)}
  n  1  Pn 1  x  Qn  y   Pn  x  Qn1  y  .....(4)
Taking n = 1, 2, 3 ...... , n in (4) and adding, we get
n
 y  x    2r  1 Pr  x  Qr  y   Q1  x  P0  y   Q0  y  P1  x 
r 1

   n  1  Pn1  x  Qn  y   Pn  x  Qn 1  y  .....(5)
Now since Q1 (y) = y, Q0 (y) = 1, P1 (x) = x, P0 (x) = 1, therefore (5) gives the required
result (1).

12.12 Relations Between Pn (x) and Qn (x)


1
Result. Prove that    2m  1 Pm  x  Qm  y 
y  x m0
and hence deduce that
1 Pm  x 
Qm (y)=  1 dx ,  y  1
y x
1
1 1 x  x x2 xm 
Proof : Let f (x) =  1    y 1  1    .....  m  ...... 
yx y y  y y y 
 
 y 1  x  y 2  x 2 y 3  ......  x m  y  m1  .......
 A0  A1 x  A2 x 2  ....... (Suppose that) ......(1)
where A’s are constants.

239
 
Further suppose that f (x) =  Bm Pm  x ,
m0

1.2.3.......m   m  1 m  2  A  ...... 

then we know that Bm=  Am  m2  ......(2)
1.3.5......  2m  1  2  2m  3 
Comparing (1) and (3) we get
A0 = y 1 , A1  y 2 , ...., Am  y   m 1 ,.....

m!   m 1  m  1 m  2   m3 

 Bm = y  y  ......
1  3  5   2m  1  2  2m  3 
= (2m + 1) Qm (y)

1
Hence    2m  1 Qm  y  Pm  x  .....(3)
yx m0
Now multiplying (3) by Pm (x) and integrating w.r.t x in the interval (–1, 1), we find that

1 1 1   
 1 Pm  x   dx   1
Pm  x     2 m  1 Pm  x  Qm  y   dx
yx  m  0 
1 2  1 
 Qm  y    Pm  x  
1
 2m  1 dx   1 Pm (n) Pn  x  dx  0, m  n 
2  1 2 2 
 Qn  y    2m  1 
2m  1 

1 Pm  x  dn  2m  1
1 1 1
  Pm  x  dx  Qn  y 
2 1 yx
This integral is called the Neumann’s integral for Qn (y).
Ex.1. Prove that x 2

 1  Qn Pn  PnQn   c and deduce that

Qn  dx
(i) 
Pn x 2

x  1 Pn2 
1 x 1
(ii) Q0  x   log
2 x 1
x x 1
(iii) Q0  x   log 1
2 x 1
Sol. The Legendre’s equation is

d2y dy
1  x2  dx 2
 2 x
dx
 n  n  1 y  0
Since Pn (x) and Qn (x) are both the solution of this equation, therefore
d 2 Pn  x  dPn  x 
1  x  2
dx 2
 2x
dx
 n  n  1 Pn  x   0 

240
 d2 d
and  1  x2  dx 2 n 
Q x  2 x Qn  x    n  1 Qn  x   0
dx
.....(5)

Multiplying (2) by Qn (x), and (3) by Pn (x) and then substracting, we get
2
 d2 
 x  1  dxd
2
2
Pn  x  Qn  x   Pn  x 
dx 2 n  
Q x
 

 d d 
2 x Qn  x  Pn  x   Pn  x  Qn  x    0
 dx dx 
d  d d 
that is 
dx 
 
1  x 2  Pn  x   Qn  x   Pn  x  Qn  x    0
 dx dx 
Integrating the above w.r.t x, we get

x 2
 
 1 Pn'  x  Q n  x   Pn  x Q n'  x   c .....(6)

1
c c  1 
Deduction. (i) Pn'  x  Q n  x   Q n'  x  Pn  x     1  
x 2 1 x2 x2
 1 1 1 
 c  2  4  6  ......  .....(7)
x x x 

1.3.... 2n  1   n n  n  1 x n2 
Now Pn (x) = x   ...
n  2  2n  1 

n  n 1  n  1 n  2  x  n 3 
and Qn (x) = x   ......
1.3.....  2n  1  2  2n  3 
Putting these values in (7), we get

1.3..... 2n  1  n 3 
n1 n  n  1 n  2  x n
 nx   .... 
 n  2  2n  1  1.3....  2n  1

  n1  n  1 n  2  x  n3  n

 x   ...... 
 2  2n  3  1.3....  2n  1

 
    n  1 x  n2 
 n  1 n  2  n  3 x  n4  .......

  2  2n  3 

1.3......  2n  1  n n  n  1 x n2  1 1 1 

 x   ......  c  2  4  6  ......
 n  2  2n  1  x x x 
Equating the coefficients of 1/x 2 from both sides, we get
1  3   2n  1 n n  n  1 1.3....  2n  1
n   c
n 1 3   2n  1 1.3....  2n  1 n

241
 c
n

 n  1  1
 2n  1 2n  1
Substituting c = 1 in (6), we get
1
Pn'  x Q n  x   Q n'  x  Pn  x   2
x 1

Pn'  x Q n  x   Q n'  x  Pn  x  1
 
Pn2  x   x  1 P
2
n
2
 x

d  Qn  x   1
    2
dx  Pn  x   
x  1 Pn2  x  
Integrating both sides w.r.t. x between the limit x to , we get

 Qn   dx

 
 Pn  x
x
 x  1 P
2
n
2
 x

Qn  x  Qn  x  dx
or  lim  .....(8)
Pn  x  x  Pn  x  x  x 1 P
2
n
2
 x

dn
n n 
Qn  x  Q x
lim dx
 lim n
Now x  Pn  x  x d
Pn  x 
dx n

n!
1.3.5....  2n  1  
 1n  n  1 n  2  .....2nx  2n 1  .......
 lim
x  1.3.5.....  2n  1
 n!
n!
=0
Thus (8) reduces to

Qn  x   dx

Pn  x 
2
 2
x x  1 Pn  x   .....(9)

(ii) Putting n = 0 in (9) and using P0 (x) = 1, we get

 
dx 1 x  1
Q0 (x) =   log

x2  1 2  x  1  x
x


1 x 1  1  x +1
  log   log  
2 x 1 x 2  x 1 

242
  x 1 
lim log   lim log
1  1 x    0
 
1  1 x 
x   x  1  x 
(iii) Taking n = 1 and using P1 (x) = x in (9), we get
 dx  1 1 
Q1(x) = x   x  2  2  dx
x
x 2
 2
x 1  x  x 1 x 


1 x 1 1 
 x  log 
2 x  1 x  x

x x 1 x x 1
  log  1  log 1
2 x 1 2 x 1

Ex.12. Show that n  PnQn1  Qn Pn1   1

Sol. We know that

(2n + 1) x Pn = (n + 1) Pn + 1 + n Pn–1 .....(10)
and (2n + 1) x Qn = (n + 1) Qn + 1 + n Qn–1 .....(11)
Multiplying (1) by Qn and (2) by Pn and then substracting, we get

or 0 =  n  1  Pn1Qn  Qn1Pn   n  Pn1Qn  Qn1Pn 

or n[Pn Qn –1– Qn Pn –1] = (n + 1)[Pn+1 Qn – Qn+1 Pn]

 f (n + 1) = f (n) .....(12)

where f (n) = n  PnQn1Qn Pn1 

Replacing n by n – 1 in (12), we get

f (n) = f (n - 1)

Similarly f (n – 1) = f (n – 2) = ... = f (1)

Hence f (n + 1) = f (n)= f (n – 1) = ... = f (1)

But f (1) =  P1 Q 0  Q 1P 0   P 0  x   1, P1  x   x 
 
= xQ0 – Q1

= xQ0 – (xQ0 – 1)  Q 1  xQ 0  1
 
=1
Thus f (n) = 1

or n  PnQn1  Qn Pn1   1

243
 Self – Learning Exercise–II

d  dQn 
1. 
dx 

1  x2 
dx 
 ........

2. Q n' 1  Q n' 1  ...........

3. Legendre’s function of second kind is ...........

12.13 Summary

In this unit we studied the Legendre’s differential equation and its solution as Legendre function
of first and second kinds. We also studied the recurrence relation, generating function, orthogo-
nal property, Rodrigues formulae and other important formulas for these functions.

12.14 Answers to Self Learning Exercises

Exercise - I
1. Legendre function of first kind 2. Pn (x)
3. 0 4. 1
5. n 6. Rodrigues formulae
1
7. 0 8.
2

5 x2  3x 
9. 0 10. Even / odd

Exercise - II
1. – n (n + 1) Qn (x) 2. (2n + 1) Qn (x)
3. Qn (x)

12.15 Exercise

1. Prove that Pn (– x) = (– 1)n Pn (x) and Pn (– 1) = (– 1)n.

2. Express P (x) = x4 + 2x3 + 2x2 – x – 3 in terms of Legendre’s polynomial
8 4 40 1 224
[Ans : P  x   P4  x   P3  x   P2  x   p1  x   Po  x  ]
35 5 21 5 105
1
3. Show that  Pn  x  dx  0 except when n = 0 in which case the value of integral is 2.
1

1
  Pn'  x   dx  2n2nn11
2
 1  x
2
4. Prove that
1
244
  2n  1
5. Show that Pn (x) Qn–2 (x) – Qn (x)Pn–2 (x) = x
n  n  1

6. Prove that xQ n'  Q n' 1  nQ n

7. Prove that


n 1

 1 
t 2 x2 1  
 Pn  x  t  1  xt  1F0  ;  ; 
n0

2 1  xt 2 
8. Prove that

 n n 1 1 
Pn  x   x n 2 F1   ,   ; 1 ; 1  2 
 2 2 2 x 
9. Show that

 Pn  x  t n xt


t 2 x2  1   
 n!  e 0 F1   ; 1 ; 4 
n 0  
 
10. Find the values of P2n+1(0), P2n (0), P2n (0) and P2n+1 (0)

 1n 1/ 2 n  1n  3 / 2 n

[Ans. 0, ,0, ]
n! n!
11. Establish the Murphy’s formula

 1 x 
Pn  x   2 F1   n, n  1 ; 1 ;  and deduce that
 2 

n  1 x 
(a) Pn  x    1 2 F1   n, n  1 ; 1 ; 
 2 

n
1 x  x 1 
(b) Pn  x    
 2 F1  n,  n ; 1 ; 
 2   x 1

n
x 1  x 1 
(c) Pn  x    

2 F1   n,  n;1; 
 2   x 1 

 2
(d) Pn  cos    2 F1 n, n  1 ; 1 ; sin   / 2  
12. Prove that

2n 1/ 2 n x n  n 1 n 1 1 
Pn  x   2 F1   ,  ; n ; 
n!  2 2 2 2 x2 
13. Prove that 245
  Pn  x  t n
 n !  e xt J 0  t 1  x 2 
n0

14. Prove that

xPn'  nPn   2n  3  Pn 2   2n  7  Pn  4  .......

and hence or other wise show that

1 ' 2n
(a) 1 xPn Pn dx 
2n  1

1 2n
(b) 1 xPn Pm' dx  0 or
2n  1

1 2 n
15. Show that 1  Pn  x   dx 
n 1

1  1 n1 2  n  1
16. Show that  1 Pn  x  dx  n
2  n  1 2  n  1 2

17. Prove that

1 1
1 Pn  x  dx  0 , n  0 and 1 P0  x  dx  2
18. Prove that
2 2 2

P02  3P12  ......   2n  1 Pn2   n  1 Pn2  1  x 2  Pn  

  

246
Unit 13 : Bessel’s Functions
Structure of the Unit
13.0 Objective

13.1 Introduction

13.2 Definition

13.3 Bessel’s Equation and its solution

13.4 Relation between Jn (x) and J– n(x)

13.5 Generating function

13.6 Recurrence Formulae

13.7 Addition Theorem

13.8 Orthogonal Property

13.9 Integral Representation of Bessel Functions

13.10 An Important Integral

13.11 Summary

13.12 Answers to Self-Learning Exercise

13.13 Exercise

13.0 Objective

In this unit you will learn about Bessel function which besides the solution of the well-known
Bessel’s equation may also be introduced through a generating function. You will also study important
properties for this function.

13.1 Introduction

No other special function have received such detailed treatment in readily available treatises as
have the Bessel functions. These functions were first introduced by F.W. Bessel, who is regarded as the
founder of the modern practical Astronomy. In fact several problems of mathematical physics lead to
Laplace’s equation and in turn converts into Bessel’s equation when there is a cylindrical symmetry. There-
fore Bessel’s function and Bessel’s equation have received great attention.
In this unit, we introduce the Bessel function through the Bessel’s differential equation and gener-

247
ating function. We then discuss the important properties (such as Recurrence formulae, orthogonal property,
Addition theorem, integral representations etc.) for this function.

13.2 Definition

13.2.1 Bessel Differential Equation

The differential equation

d2y dy
2
x
dx 2
dx

 x  x 2  n2 y  0  .....(1)

is called Bessel’s differential equation of order n where n is non-negative real number.

13.2.2 Bessel’s function of the first kind of order n

It denoted by Jn(x) and is defined as

r n2r
Jn  x  

 1 x
  .....(2)
r  0 r !  n  r  1  2 

(where n is any non-negative constant)

 x 2 n  x2 
 0 F1 
 ; n  1;  4  .....(3)
  n  1  
It n is a negative integer, then we put
n
J n  x    1 J  n  x  .....(4)
n
J n   x    1 J n  x  .....(5)
Equations (3) and (4) together define Jn (x) for all finite x and n.
Replacing n by 0 and 1 in (2), we find that

x2 x4 x6
J0  x   1    ... .....(6)
22 22 44 24 44 62

x x3 x5
and J1  x    2  2 2  ... ......(7)
2 2 3 2 4 6

13.3 Bessel’s Equation and its Solution

Bessel differential equation is

d2y dy
x2
dx 2
 x  x 2  n2  0
dx
  .....(1)

The equation (1) has a regular singular point at x = 0, and an irregular singular point at x = ,
while all other points are ordinary points. The solution of equation (1) called Bessel’s function will de-
pend upon n. This index n may be non-integer, a positive integer or zero. We discuss three possibilities :
Case I. Solution of (1) for non-integral values of n
248
 Here the equation (1) is solved in series by using the well-known method of Frobenius.
Let the series solution of (1) be

y   ar x c  r , a0  0 .....(2)
r 0

dy 
From (2), we get   ar  c  r  x c  r 1
dx r 0

d2y 
and 2
  ar  c  r  c  r  1 x c  r  2
dx r 0

dy d 2 y
Substitution for y , , in (1) gives
dx dx 2
  

r 0 r 0

x 2  ar  c  r  c  r  1 x c  r 2  x  ar  c  r  x c  r 1  x 2  n 2  a
r 0
r xc r  0

   
or  ar  c  r  c  r  1 xcr   ar  c  r  xcr   ar xc r 2  n2  ar xc r  0
r 0 r 0 r 0 r 0

 
or   c  r  c  r  1 c  r   n2  ar xc r   ar xcr 2  0
r 0 r 0

 
or   c  r  n  c  r  n  ar xcr   ar xc r 2  0 .....(3)
r 0 r 0

Equating to zero the lowest power x i. e, xr, we get the indical equation as
(c + n) (c – n) a0 = 0
 c = n, –n as a0  0
So roots of the indical equation are c = n, –n.
Now equating to zero, the coefficient of xc + 1, we find that
(c + 1 + n) (c + 1 – n) a1 = 0
so that a1 = 0 for c = n and –n.
Finally equating to zero the coefficient of xc + r, we get
(c + r + n) (c + r – n) ar + ar–2 = 0
1
or ar   ar  2 .....(4)
 c  r  n  c  r  n 
Putting r = 3, 5, 7, ..... in (4) and using a1 = 0 we find that
a1 = a3 = a5 = a7 = ..... = 0 .....(5)
Also putting r = 2, 4, 6, ..... in (4) gives
1
a2   a0
 c  2  n  c  2  n 
249
 1
a4  a0 and so on .
 c  2  n  c  2  n  c  4  n  c  4  n 
Putting these values in (2), we get

y   ar x c  r  a0 x c  a2 x c  2  a4 x c  4  .....  as a1  a3  a5  0
r 0


c x2 x4 ....

or y  a0 x 1    
  c  2  n  c  2  n   c  2  n  c  2  n  c  4  n  c  4  n  
Replacing c by n and – n, we get
 x2 x4 
y  a0 x n 1    ..... .....(5)
 2   2n  2  2  4  2n  2  2n  4  

n  x2 x4 
and y  a 0 x 1    ..... .....(6)
  2n  2   2  2n  2  2n  4   2  4 
The particular solution of the equation (1) obtained from (5) above by taking the arbitrary con-
1
stant a0  is called the Bessel function of the first kind of order n. It will be denoted by
2n   n  1
Jn (x). Thus we have
xn  x2 x4 
Jn  x  n 1    ..... .....(7)
2   n  1  4  n  1 4 8  n  1 n  2  
r
Jn  x  

 1  x  n2r
or  x .....(8)
r  0 r   n  r  1 2

1
Similarly taking a0  n
in (6), we get
2   n  1

J n  x   

 1r x
2 rn

  .....(9)
r 0 r    n  r  1  2 
Let n be non-integral. Since n is not an integer and r is always integral, the factor (–n + r + 1)
in (9) is always finite and non-zero (m) is always finite for m  0 or a negative integer.) Again for
2r < n, (9) shows that J–n(x) contains negative powers of x. On the other hand, (8) shows that Jn(x)
does not contain negative power of x at all. Therefore for x = 0, Jn(x) is finite. While J–n(x) is infinite, and
so one can not be expressed as constant multiple of the other. Thus we conclude that Jn(x) and J–n(x)
are independent solutions of (1) when n is not an integer. Thus general solution of Bessel’s equation (1)
when n is not an integer is
y = AJn(x) + BJ–n(x)
where A and B one arbitrary constants.
Case-II. Solution for positive integral values of n and for n = 0.
It n is a positive integer, then for c = –n, the recurrence relation (4) gives

250
 1
ar  ar  2
r  2n  r 
which breaks when r = 2n.
Also if n = 0, the two roots of the indical equation becomes equal and in that case the afore-
mentioned method is not applicable.
In both the cases, the second solution of (1) can be found by using methods mentioned in unit 9.

13.4 Relation between Jn(x) and J–n(x), n being an integer

n
Result. J  n  x    1  J n  x  .....(1)
Proof. We consider two cases :
Case I. Let n be a positive integer

J n  x   

 1r x
2r n
We have   .....(2)
r0 r   n  r  1  2 
Since n > 0, so (–n + r + 1) is infinite. for r = 0, 1, ..., n – 1, therefore (2) becomes

J n  x   

 1r x 
2 r n

 
r n r    n  r  1  2 


 1mn  x 2m n
   (taking r = m + n)
m 0 m  n   m  1  2 

r n 2 r
  1 
n

 1  x
 
r  0 r   r  n  1  2 

n
  1 J n  x  (by definition)
Case II. Let n < 0.
Putting n = –p, where p is a tive integer
Since P > 0, therefore form Case I, we have
p
J  p  x    1 J p  x 
p
or J p  x    1 J p  x
Putting p = –n, we get the required result.
Hence the relation (1) is true for any integer.

13.5 Generating Function

Theorem. Prove that when n is a positive integer Jn (x) is the coefficient of zn in the
x 1 
expansion of exp   z    in ascending and decending power of z.
z z 
251
 x  1   xz   x 
Proof. We have exp   z     exp    exp   
z  z   2  2z 

  x   x 2 z 2 n n
 x z  x
n 1 n 1
z 
 1    z     ....         ....
  2   2  2 2 n 2 n 1 

 x 2 2 n n n n 1 n 1 
1  x  z  x   1 z  x   1
z    .... ....(1)
 n 1
 1    z     ....     
  2  2 2 2 n  2 n 1 
M ultiplying the R.H.S. of (1) term by term, we find that coefficient of zn is
x n2 n4
1  x 1  x 1  x
      ....
n  2 n  1  2  2 n  2  2 
r n2 r


 1  x
 Jn  x
  .....(2)
r  0 r   n  r  1  2 

Similarly the coefficient of z–n in the expansion (1) is

n n n 1 n2 n2
 1  x   1  x   1  x n4  ....  1 n J x
           n   .....(3)
n 2 n 1  2  2 n2  2
Further, the term independent of z is
x2 x4
 1   ....  J 0  x  .....(4)
22 22  42
Hence relation (1) with help of (2), (3) and (4) may be written as
x  1   1  1 
exp   z     J 0  x    z   J1  x    z 2  2  J 2  x   ....
2  z   z  z 
Since J–n(x) = (–1)n Jn(x), therefore

x  1 
exp   z      J n  x  z n .....(5)
2  z   n 

13.6 Recurrence Formulae for Jn (x)

13.6.1 xJn (x) = nJn (x) – xJn + 1 (x)

Proof. We have
r n 2r

 1 x 
Jn x     
r  0 r   n  r  1  2 

Differentiating above w.r.t. x, we get

r

 1  n  2r   x  n  2 r 1 1
J n  x      
r  0 r   n  r  1  2  2

252
 r n  2 r 1
 n

 1  x 1 x
 

 1r x
n  2 r 1
1
  2r    
r  0 r   n  r  1  2  2 x
r 0 r   n  r  1  2  2
r n  2 r 1
n 
 1r  x
n 2 r 
 1 x

x
 r   n  r  1  2    
r 1  r  1   n  r  1  2 
r 0

s
n 2 s 1
n
 Jn  x   

 1  x
 
x s 0 s  n  s  2   2 

n
 J n  x   J n1  x 
x
Hence xJ n  x   nJ n  x   x J n1  x 

13.6.2 xJn (x) = xJn–1 (x) – nJn (x)

Proof. We have as in formulae 13.6.1
r

 1  n  2r   x  n  2 r 1 1
J n  x      
r 0 r   n  r  1  2  2

r

 1  2n  2r  n   x n  2 r 1 1
   
r 0 r   n  r  1  2  2

r


 1  n  r   x  n  2 r 1 
 1r x
n  2 r 1
 x 1
   n     
r  0 r   n  r  1  2  r 0 r   n  r  1  2   x 2

r


 1  n  r   x n  2 r 1 n 
 
 1r x
n 2 r

   
r 0 r  n  r    n  r   2  x r 0 r   n  r  1  2 

r

 1  n  r   x n  2 r 1 n
    Jn  x
r 0 r   n  r   2  x

n
 J n1  x   Jn  x
x
Hence xJ n  x   x J n1  x   nJ n  x 

13.6.3 2Jn (x) = Jn–1 (x) – Jn + 1 (x)

Proof. Adding recurrence formulae 13.6.1 and 13.6.2, we get the formula 13.6.3.
13.6.4 2n J n  x   x  J n 1  x   J n  1  x  
Proof. Substracting recurrence formula 13.6.2 from 13.6.1, we easily get recurrence formula
13.6.4.
d  n
13.6.5 x J n  x     x  n J n1  x 
dx 
253
Proof. By formulas 13.6.1, we have
x J n  x   n J n  x   x J n1  x 
Multiplying both sides of above by x – n–1, we have
x  n J n  x   nx  n 1 J n  x   x  n J n 1  x 

or x  n J n  x   nx  n 1 J n  x    x  n J n1  x 
d n
or  x J n  x     x  n J n 1  x 
dx  

d  n
13.6.6  x J n  x    x n J n1  x 
dx
Proof. By formula 13.6.2, we have
x J n  x   x J n1  x   n J n  x 
Multiplying both sides of above by x n–1, we have
x n J n  x   x n J n 1  x   nx n1J n  x 

or x n J n  x   nx n 1 J n  x   x n J n1  x 
d  n
or  x J n  x    x n J n 1  x 
dx
2
Ex.1. Prove that J 1 2  x    sin x
x
Sol. We know that
xn  x2 x4 
Jn  x  n 1    ...........
2   n  1  2   2n  2  2  4   2n  2  2n  4  

1
Puting n  and using   3    , we get
2 2 2
2x  x2 x4 
J1 2  x   1    ...
  23 35 2  4 

2  x3 x 5 
  x    ...
x  31 51 

2
 sin x
x
Ex.2. Show that Jn (x) is even and odd function for even n and for odd n respectively.
Sol. Replacing x by – x in the definition for Bessel function, we get
r n2r
Jn x  

 1  x

 
r  0 r   n  r  1  2 

254
 r n 2 r
  1 
n

 1  x n
  1 J n  x 
 
r  0 r   n  r  1  2 

(i) If n is even then J n   x   J n  x  , therefore J n  x  is even.

(ii) If n is odd then J n   x    J n  x  , therefore J n  x  is odd.

Ex.3. By using generating function, for Bessel function, show that
(i) cos  x sin    J 0  2 J 2 cos 2  2 J 4 cos 4  ........

(ii) sin  x sin    2 J1 sin   2J 3 sin 3  ......

(iii) cos x  J 0  2 J 2  2 J 4  ..........

(iv) sin x  2 J1  2 J 3  2J 5  ...........

x  1   n
Sol. We have exp   z      z J n  x 
2  z   n 

 1  1 
 J 0  x    z   J1  x    z 2  2  J 2  x   ...... .....(1)
 z  z 
Let us put z  ei . Then
 1 
z n   n   2i sin 
z 
1
and zn   2cos n
zn
From (1), we have
exp  x i sin    J 0   2i sin   J1  x    2cos 2  J 2  x   .........

 cos  x sin    i sin  x sin    J 0  2 J 2 cos 2  2 J 4 cos 4 

i  2 J1 sin   2 J 3 sin 3  .....
Separating real and imaginary parts, we easily arive at relations (i) and (ii).

Also on putting   in (i) and (ii), we get easily the selations (iii) and (iv).
2
d
Ex.4. Prove that  x J n  x  J n1  x    x  J n2  x   J n21  x   ..... (2)
dx
and deduce that
x  2 J 0 J1  6 J1 J 2  .........  2  n  1 J n J n1  ......

Sol. we have L.H.S of (2)  x J n  x  J n 1  x   x J n  x  J n1  x   J n  x  J n1  x  .....(3)

From recurrence relations 13.6.1 and 13.6.2 we have
x J n  x   n J n  x   x J n1  x  .....(4)

x J n  x   n J n  x   x J n1  x  .....(5)
255
 Putting n as (n + 1) in (5), we get
x J n 1  x     n  1 J n1  x   x J n  x  .....(6)

Substituting the value of x J n  x  and x J n 1  x  from (4) and (6) in (3), we get.

L.H.S of (2) = J n  x     n  1 J n 1  x   x J n  x  

 J n 1  x   n J n  x   x J n 1  x    J n  x  J n 1  x 

 x  J n2  x   J n21  x    R.H.S of (2)

This completes the solution of the problem.

Deduction. Putting n = 0, 1, 2 ..... respectively in (2) and adding after multiplying by 1, 3, 5
res, we get
d
 x  J 0  x  J1  x   3J1  x  J 2  x   5 J 2  x  J 3  x   ......  x .....(7)
dx 
Integrating (7) in the interval (0, x), we get the required result. [after using Ex. 6 (i)]
d n
 J n2  x   J n21  x    2  J n2  x  
 n  1 2 
Ex.5. Prove that   J n 1  x  
dx x x 
d
Sol. We have  J n2  x   J n21  x  
dx  
 2 J n  x  J n  x   2 J n1  x  J n 1  x  .....(8)

n
From recurrence relation 13.6.1, we have J n  x   J n ( x)  J n 1 ( x) .....(9)
x
Replacing n by n + 1 in recurrence relation 13.6.2, we find that

J n 1  x   
 n  1 J
n 1  x   Jn  x  .....(10)
x
Using (9) and (10) in (8), we get
d  2 n   n 1 
 J n  x   J n21  x    2 J n  x   J n  x   J n 1  x    2 J n 1  x    J n 1  x   J n  x 
dx x   x 
n n 1 2 
 2  J n2  x   J n 1  x  
x x 
which completes the solution of the problem.
Ex.6. Prove :
2 2

2 2
(i) J 0 + 2 J1 + J 2 + J 3 + ... = 1 
(ii)   J 0  x   1

(iii) J n  x   21 2 ,  n  1 
Sol. From Ex.5 we have
d  2 n ( n 1) 2 
 J n  J n21   2  J n2  J n 1  .....(11)
dx x x 
Replacing n by 0,1,2,3, ... in (1), we get
256
 d  2  1 
 J 0  J12   2  0  J12 
dx  x 
d  2 1 2 
 J1  J 22   2  J12  J 22 
dx x x 
d  2 2 2 2 3 2
J 2  J 3  2  x J 2  x J3 
dx    
... ... ...
... ... ...
and so on.
Adding column-wise and using nlim J ( x )  0, we get
 n

d
 J 02  2 J12  2 J 22  ....  0 .....(12)
dx  
Integrating the result (12), we get
J 02 ( x )  2  J12 ( x )  J 22 ( x )  ...  c ......(13)
Putting n = 0 in (13) and using
J 0 (0)  1 and J n (0)  0 for n  1,
we obtain 1 + 2(0 + 0 + ......) = c, Thus c = 1
Hence (13) gives J 02  2( J12  J 22  .....)  1 ....(14)
2 2 2
 2
(ii) From (14) we have J 0  1  2 J1  J 2  J 3  ...  ....(15)

Since J12 , J 22 , J 32 ,..... are all positive or zero, (15) gives

J 02  1 so that J 0 ( x )  1
(iii) Also from (14) we have

 
J 02  1  2 J12  J 22  J 32  .....  J n21  J n2  J n21  ...

Solving for J n2 we have

1
J n2 
2
  
1  J 02  J12  J 22  ...  ....(16)

Since J 02 ,J12 ,J 22 ,.... are all positive or zero, therefore

2 1
(16) gives that J n  or J n  x   21 / 2 , where n  1
2

d  J  n  x   2sin nπ
Ex.7. Prove that dx  J x  = 
 n    πxJ n2

2sin nπ
or J n J  n =
πx
Sol. Since J n (x) and J  n (x) are solutions of

257
 d 2 y 1 dy  n2 
  1   y  0,
dx 2 x dx  x 2 

1  n2 
therefore J n  J n  1  2  J n  0 , ...(17)
x  x 

1  n2 
and J n  J  n  1  2  J  n  0 ...(18)
x  x 
Multiplying (17) by J  n and (18) by Jn and substracting, we get
1
 J n J n  J n J n    J  n J n  J n J  n   0 ...(19)
x
Let u  J  n J n  J n J  n .
1 u 1
Then (19) reduces to u   u  0  
x u x
a a
Integrating we get log u = log or u 
x x
where a is arbitrary constant or
a
J  n J n  J n J  n 
x

1  n xn2 x n4 
 x    ....
2  n    n  1  2   2n  2  2.4  2n  2  2n  4  

1  n1  n  2  x n1  n  4  x n 3 
 n  nx    ....
2   n  1  2   2n  2  2.4  2n  2  2n  4  

1  n xn2 x n 4 
 n x    ....
2   n  1  2   2n  2  2.4  2n  2  2n  4  

1   n 1 ( 2  n )x
1 n
 4  n  x 2 n  a
 n   nx    ...  .....(20)
2    n  1  2   2  2n  2.4  2  2n  4  2n   x

1
Comparing the coefficients of on both sides of (20), we get
x
1 2n 2 sin n
a  n   n    
  n  1   n  1 n  n   1  n  


(using   z   1  z   )
sinz
J  n J n  J n J  n 2sin n
Thus 
J n  x J n2
258
 d  Jn  2 sin n 
  
dx  J n  x J n2

Self-Learning Exercise-I

1. J0 (x) is a Bessel’s function of order ....... .

d n
2.  x J n  x   x n ...... .
dx  
3. Write generating function for Bessel function Jn (x) .
4. J–n (x) = (– 1)n ........ .
5. Write differential equation for the Bessel function Jn (x).
6. x  J n 1  x   J n 1  x    ........... .
7. Jn (x) is even function if n is .......

8. lim x  n J n  x   ...........
x 0

13.7 Addition Theorem

Statement : It n is a positive integer, then
n 
r
J n  x  y    J n  x  J n  r  y     1  J r  x  J n r  y   J r  y  J n r  x  ...(1)
r 0 r 1


x  1 
Proof : we have  J n  x  z n  exp   z   
n  2  z 

x 1  y 1 
  J n  x  y  z n  exp   z    exp   z   
n  2  z  2  z 
 
  zr Jr  x  zs Js  y 
r  s 

Now equating the coefficient of z on both sides, keeping in mind that the terms containing zn on
n

R.H.S. are obtained by taking s = n – r and by making r vary from –  to  thus


Jn  x  y   J r  x J n r  y  ...(2)
r 
where n is any integer.
1 n 
or Jn  x  y   J r  x J n r  y    J r  x  J n r  y    J r  x  J n r  y  ...(3)
r  r 0 r  n1

1 1
Now  J r  x  J nr  y    J  p  x  J n p  y  (writing – r = p)
r  p 


p
   1 J p  x  J n p  y 
p 1
259
 
r
   1 J r  x  J n  r  y  (replacing dummy index p by r) ...(4)
r 1

 
Also  J r  x  J n r  y    J n q  x  J  q  y  (taking r = n+q)
r n 1 q 1


q
   1 J n q  x  J q  y 
q 1


r
   1 J n  r  x  J r  y  ....(5)
r 1
Using (4) and (5) in (3), we easily arrive at the addition thorem given by (1).

13.8 Orthogonal Property

Result : If i and j are the roots of the equation J n  a   0

0 , if i  j 
  2 
then 0 x J n   i x  J n   i x  dx   a 2 
 J n 1   i a  , if i  j 
2 
Proof : Case I : Let i  j i.e. let i and j are different roots of Jn (a) = 0
 J n  i a   0 and J n  j a  0   ....(1)

Let u  x   J n  i x  and   x   J n  j x   ....(2)

then u and v are Bessel functions satisfying the modified Bessel equation
d2y dy
x2
dx 2 
 x   2 x 2  n2 y  0
dx

or 
x 2 y  xy    2 x 2  n2 y  0  ...(3)

or x u   xu    
2
i x  n u  0
2 2 2
...(4)

or x   x   
2
j x  n   0
2 2 2
...(5)
Multiplying (4) by  and (5) by u and then substracting we get


x 2  u   u   x  u   u   x 2  i 2   j 2 u  0 
or 
x  u   u    u  u   x  j 2   i 2 u 
d
or x
dx

  u   u       u   u    x  j 2   i 2 u  
d
or x x  u  u    x  j 2   i 2 u
dx 
  ...(6)
Integrating (6) w.r.t. x from 0 to a, we get
260
 a a
 j
2
 i 2  0
x u  dx   x  u   u   0 ...(7)

a

2
Using (2), (7) gives  j  i
2
 0
 
xJ n  i x  J n  j x dx

a
  
  x J n  j x J n  i x   J n  i x  J n  j x 
 0  
 
 a  J n  j a J n   i a   J n   i a  J n  j a   
= 0 [using (1)]
Since  i   j the above equation gives
a
 0 xJ n  i x  J n   j x  dx  0 if i  j ....(8)
Case II : Let i = j (equal roots). Multiplying (4) by 2u, we have
2

2 x 2u u   2 x  u    2  i 2 x 2  n2 uu   0 
d  2 2
or x  u   n2 u 2  i 2 x 2u 2   2 i 2 xu 2  0
dx  
d  2 2
or 2i 2 xu 2  x  u    n 2u 2   i 2 x 2u 2  ....(9)
dx  
Integrating (9) w.r.t. x from 0 to a, we get
a 2 a
2i 2  xu 2dx   x 2  u   n 2u 2  i 2 x 2u 2  ....(10)
0  0
Using the relation Jn (0) = 0 and (1) and (2), we have
a 2 2 2 a
2i 2  xJ n2  i x  dx   x 2 J n   i x   n 2 J n   i x    i2 x 2 J n   i x  
0   x 0
a 2
or 2 i 2  xJ n2   i x  dx  a 2  J n   i x   ....(11)
0   at x  a
From recurrence relation 13.6.1, we have
d n
 J n  x    J n  x   J n 1  x  ...(12)
dx x
Replace x by ix in (12), we have
d  J n   i x   n
or  J   x   J n 1  i x 
d  i x   i x  n i
1 d  J n  i x   n
or   J n  i x   J n 1  i x 
i dx i x

n
 J n   i x   J n   i x    i J n 1   i x 
x

261
  n 2
 J    x 2   
 at x  a   x n  i  i n 1  i   
Now  J  x  J x
 n i
  at x  a

2
  0   i J n 1   i a   (by (1))

  i2 J n21  i a  ...(13)
Using it in (11), we get
a a2 2
0 xJ n2   i x  dx  J n 1   i a 
2
Combining these two results we can write
a a2 2
0  
xJ n   i x  J n  j x dx 
2
J n 1   i a  ij

0 ,i  j
where ij = (kronecker delta) =  .
1,i  j
1
Ex.1. Prove that J n  x   cos  n  x sin   d  where n is a positive integer

Sol. We shall use the following results :
   / 2, if m  n
 0 cos m cos nd    0 sin m sin nd    0, if m  n
....(14)

We also proved in Ex. 3(§13.6) that

cos  x sin    J 0  2 J 2 cos 2  2 J 4 cos 4  .... ....(15)

and sin  x sin    2 J1 sin   2 J 3 sin 3  2 J 5 sin 5 + ... ....(16)

Multiplying (15) by cos n and integrating between the limit 0 to , and using (14) we get

 0 cos  x sin  cos nd  0 (if n is odd) ....(17)

   2
and  0 cos  x sin  cos nd   2 J n  0 cos 2
nd   2 J n
 J n (if n is even) ....(18)

Again multiplying (16) by sin n and integrating between the limit 0 to  and using (14), we get

 0 sin  x sin  sin nd   0 (if n is even) ....(19)
 
and  0 sin  x sin  sin n d   2 J n  0 sin 2n d   2 J n   2    J n (if n is odd) ....(20)
Let n be odd. Adding (17) and (20), we get

 0 cos  x sin  cos n  sin  x sin  sin n d    Jn


or  0 cos  n  x sin  d    J n

262
 1 
Jn  cos  n  x sin   d 
 0
or ....(21)

If n is even, then add (18) and (19) to get the required result.
Thus (21) holds for each positive integer n (even as well as odd)
Remark : If n is negative integer so that n = – p, where p is a positive integer. Putting n = – p
in (21) we get
1 
J p  x  cos   p  n sin   d 
 0
....(22)

Let      so that d= – d 

1 0
 R.H.S. of (22) 
 
 
cos   p       x sin         d  

1 
 cos  p  x sin    p d 
 0
1 
 cos  p  x sin   cos p  sin  p  x sin   sin p  d 
 0 
p

 1
cos  p  x sin   d 
 
Thus (22) becomes
p
p  1 
 1 J p  x   0 cos  p  x sin   d 

1 
J n  x   cos   n  x sin   d 
 0
or

Hence the result (22) holds for each integer.

1  1 
Ex.2. Prove that J0  x  
 cos  x sin   d    cos  x cos   d 
 0  0
Sol. From Ex.3(§13.6), we have
cos  x sin    J 0  2 J 2 cos 2  2 J 4 cos 4  ........ .....(23)
Integrating (23)w.r.t ‘’ between the limit 0 to , we get
1 
or J0  x    0 cos  x sin   d 

 
Again replacing  by     in (23) and simplifying,we get
2 
cos (x cos  )  J 0  2 J 2 cos 2  2 J 4 cos 4 ........ ....(24)

Thus  0 cos  x cos  d   π J 0  x 
1 
 J0  x    0 cos  x cos   d 

263
 n dn
Ex.3. Prove that J n  x    2  x n n
J0  x 
d x  2

Sol. Substituting the value of J0 (x) in series in R.H.S, we have

 n    1r  x  2 r 
  2 x 
n  d
R.H.S     
n
r   r  1  2  
 d x  
2
r 0 

n
 dn    1 r  t r  
  2 x     2 r  
 dt n r 0 r   r  1  2  


  2 x 
n

 1r t r  n
r r  n22 r
r 0

  1
n

 1r x
n2r

  
r    n  r  1  2 
r 0

n
  1 J  n  x   J n  x 


 ax 1
Ex.3. If, Prove that e J 0  bx  dx 
0 a 2  b2
Sol. Using series representation for the Bessel function and changing the order of integration and
summation, we find that
r 2r
  ax  1  b 2   2 r  ax

I 
0
e J0  bx  dx   2 0 x e dx
r 0  r
r 2r


 1  b 2    2r  1
2 (using the def. of gamma function)
r 0  r  a 2r 1
Applying gamma duplication formula for   2r  1 and simplifying, we find that
r
1  1 2 r  b2 
I
a
 r
  2 
r 0  a 

1 2
1  b2  1
 1  2  
a a  a  b2
2

13.9 Integral Representation of Bessel Functions

264
Theorem : Prove that
n
 x  1 1 n  1 2   1
 
 2
  n   J n  x    exp  ixt  1  t 2
 2 1
  dt ,  n   
 2
.....(1)

Proof : We have

 i x r 1 n 1 2 
 t 1  t 
r 2
R.H.S. of (1)  dt .....(2)
r 0 r 0

Since the integrand in (2) is even or odd according as r is even or odd respectively, therefore
2k 1
 ix 

 2  t 2 k 1  t 2 
n 1 2 
R.H.S   dt
k 0  2k  0

Putting t2 = u and using the formula

 1
2k    2k  1  22 k 1 2   k  1   k   ,
 2
We get


 1k x 2k  1
n 1 2 
R.H.S. of (1)   2k  u k 1 2 1  u  du .....(3)
k 0 2   k  1   k  1 2  0

Now evaluating the integral by using the well known definition of Beta function, we get
k 2k
 1 
  n  
 1  x
 
R.H.S. of (1)
 2  k 0 k !   k  n  1  2 

n
 x  1
      n   Jn  x 
 2  2
Similarly we have
n 1
 1  x  n 1 2 
 e 1  t 
 ixt 2
   n    Jn  x  dt .....(4)
 2  2  1

Adding (1) and (4) we get

n 1
2 x n 1 2 
Jn  x   
   n  1 2    2   cos xt 1  t
2
  dt ,  n   1 2  .....(5)
0

For t  sin , eq. (5) gives

n /2
2 x
Jn  x     cos  x sin   cos2n  d 
   n  1 2    2  0

 
Replacing  by     in the above relation, we get
2 
265
 n 2
2  x 2n
Jn  x     cos  x cos  sin  d
   n  1 2    2  0

13.10 An Important Integral

Theorem : Prove that

a 4a 2a 2
0  
x a 2  x 2 J 0  kx  dx 
k3
J 1 ak  2 J0  ak 

k
.....(1)

d n
Proof : We know that
dx

x J n  x   x n J n 1  x  
Replacing x by kx, we get
d n
dx
 
x J n  kx   kx n J n 1  kx  .....(2)

Integrating (2) w.r.t. x in the interval (0, a), we get

a
n
x J n 1  kx  dx  a n J n  ka  .....(3)
0

a a a

 x a J
2 2
Now, x 0  ax  dx  a  x J 0  kx  dx   x 3 J 0  kx  dx
2

0 0 0

a3 a x2 d
 J1  ak     x J1  kx   dx
k 0 k dx 

[Using (3) with n = 1 for first integral and (2) with n = 1 for second inte-
gral]

a3 1 a a
 J1  ak    x 2  xJ1  kx    2 x 2 J1  kx  dx 
k k 0 0 

a3 a3 2 a d
 0 dx  x 
2
 J1  ak   J1  ak   2 J 2  kx  dx
k k k

2a 2
 J 2  ax  ....(4)
k2
Also we have the recurrence relation
2nJ n  x   x  J n 1  x   J n 1  x   ....(5)
Taking n = 1 and replacing x by kx in (5), we find that
2
J 2  kx  J1  kx   J 0  kx 
kx
Substituting the value of J2(kx) in (4), we easily get the integral (1).

Self-Learning Exercise-II
266
 2 2
1.  J1 2  x     J 1 2  x   = .............

2. The relation J 01  x    J1  x  is true /false

3.  J1 2  x    .......

4.  0 cos  n  x sin   d   ..........
5. J 0  x   ...., n  1

6. J n  x   ...., n  1

13.11 Summary

In this unit we studied the Bessel’s differential equation and its solution. Also we proved the im-
portant properties such as recurrence relations, generating function, orthogonal property, integrals repre-
sentation for the Bessel function.

13.12 Answers to Self- Learning Exercises

Exercise-1
1. 0 2. Jn-1(x)

x 1  n
3. exp   z    J n  x  z
 4. Jn (x)
2  z   n

d2y dy
5. x
dx
2
2
dx

 x  x 2  n2 y  0  6. 2nJn(x)

1
7. even 8. n
2 n!
Exercise-II
2
1. 2. true
x
2
3. sin x 4.  J n  x 
x
5. 1 6. 21 2

13.13 Exercise

2
1. Prove that J 1 2  x   cos x
x

267
 2

2. Prove that  x J1 2  2 x  dx  1
0

t t
3. Prove that  0 J 0 x  t  x  dx  2 sin 2
4. Prove that
x n 1
(i) 0 x J n  x  dx  x n1 J n1  x  , n  1

x n 1
(ii)  x J n1  x  dx  n
 x n J n  x 
0 2 n
5. Use recurrence relations for Bessel’s functions to show that
J 0  x 
(i) J 2  x     J 0  x 
x
(ii) 4 J 0  x   3J 0  x   J 3  x   0

(iii) 2J 0  x   J 2  x   J 0  x 
6. Using generating function, prove that

Jn  x  y   J r  x  J n r  y 
r  

7. Prove that
2  sin x 
(i) J 3 2  x     cos x 
x  x 

2  cos x 
(ii) J 3 2  x      sin x 
x  x 

2  3  sin x  
(iii) J 5 2  x      cos x   sin x 
x x  x  

2  3  cos x  
(iii) J 5 2  x      sin x   cos x 
x x  x  
2
8. Prove that J n 1  x    nJ n  x    n  2  J n  2  x    n  4  J n  4  x   .....
x
a 1 2 y
2
9. prove that  0 x sin  ky  y  x
2
  dx 
2
J1  ky 

2 n 
10. show that J n  x    J n  x    n  2  J n  2  x    n  4  J n  4  x  .....
x 2 



268
Unit 14 : Hermite Polynomials
Structure of the Unit
14.0 Objective

14.1 Introduction

14.2 Hermite Differential Equation and Its Solution

14.3 Generating Function

14.4 Hypergeometric Form

14.5 Recurrence Formulas

14.6 Rodrigue’s Formula

14.7 Orthogonal Property

14.8 Summary

14.9 Answers to Self-Learning Exercises

14.10 Exercise

14.0 Objective

Here you will study Hermite polynomials its definition and important properties such as recur-
rence relations, generating function, orthogonal property, Rodrigue’s formula etc.

14.1 Introduction

Hermite polynomials occur in the study of wave mechanics and other physical problems. We
start with the Hermite differential equation and its solution. Then we develop and study properties of
Hermite polynomials. We also illustrate the properties with the help of solved problems.

14.2 Hermite Differenential Equation and Its Solution

Hermite’s equation is

d2y dy
 2 x  2 ny  0 .....(1)
dx 2 dx
where n is any integer For solving equation (1), we use Frobenius method.

Let y   ar x k  r , a0  0 .....(2)
r 0

269
 dy d2y
Now obtain and 2 from (2) and substitute in (1), we get
dx dx

 ar  k  r  k  r  1 xk r 2  2  k  r  n  x k r   0 .....(3)
r 0

Equation (3) is an identity. We equate to zero the coefficient of smallest power of x, viz. xk – 2 in
(3) and obtain the indical equation as
a0 k(k –1) = 0
k(k –1) = 0  a0  0 .....(4)
So roots of indical equation are k = 0, 1. They are distinct and differ by an integer.
Again equating to zero the next snallest power of x i.e xk – 1. So we get
a1(k + 1) k = 0 .....(5)
When k = 0, (5) shows that a1 is indeterminate. Hence a0 and a1 can be taken as arbitrary con-
stants.
Equating to zero the coeffcient of xk + r – 2, (3) gives
2  k  r  n  2
ar  ar  2 .....(6)
 k  r  k  r  1
Putting k = 0, we get
2  r  n  2
ar  ar 2 .....(7)
r  r  1
For r = 2, 4, 6, ......., 2 r in (7), we get
1
2n  1  21  n
a2   a0   a0 ,
2 1 2
2
22  n  1 22 n  n  2 
a4   a2  a 0
43 4
.... .... .... ....

a2r  
 1r 2r  n  n  2  .....  n  2r  2  a0
and
2r
Next, putting r = 3, 5, 7, ...., 2r + 1, in (7) we get

a3 
 11 21  n  1 a1
3
2
 1 22  n  1 n  3
a5  a1
5
.... .... .... ....

270
 r
 1 2r  n  1 n  3 ......  n  2r  1
and a2r 1  a1
2r  1
Putting the above values in (2) with k = 0, we get

 2n r 
2 22 n  n  2  4  2  n  n  2  ...  n  2r  2  2 r
y  a0 1  x  x  .....  x  ....
 2 4 2r 

 2  n  1 3 22  n  1 n  3 5  2 r  n  1 n  3 ...  n  2r  1 2r 1 

a1  x  x  x  ...  x  ...
 3 5 2r  1 
....(8)
or y = a0 v + a1w, say .....(9)
Since v or w is not merely a constant, v and w form a fundamental set (i.e. linearly independent)
of solutions of (1). Hence (8) or (9) is the most general solution of (1) with a0 and a1 as two arbitrary
constants.
Remark : In practice we require solution of (1) such that
(i) it is finite for all finite values of x and
(ii) exp (1/2x2) y(x) 0 as x  
The solution (8) does not satisfy the condition (ii). However, if the series terminate then this con-
dition will be satistied. Replacing r by r + 2 in (7), we get
2  r  n
ar  2  ar .....(10)
 r  1 r  2 
If r is a positive integer, then for r = n, ar + 2 = 0 ie the series terminates. We now find the solu-
tion of (1) in descending powers of x for n + (set of positive integers)
For k = 0, the equation (2) becomes
y = anxn + an –2 xn – 2 + an –4 xn – 4 + .... .....(11)

From (10) we get ar  

 r  1 r  2  ar  2
2n  r 
Let r = n – 2, n – 4, ...... Then
n  n  1
an 2   an .
22
n  n  1 n  2  n  3
an  4  an and so on
22  2  4
Putting these values in (11) we find that
 n(n  1) n 2 n(n  1)(n  2)(n  3) n  4
y  an  x n  x  x  ...
 22 22  2  4

271
 ( 1) r n( n  1)...(n  2r  1) n  2 r 
 r
x  ...
2  2  4...2r 
 n 2 1 r n n  1 .... n  2r  1
      x n 2 r
 an  r
2  2  4.....2r
r 0

n
, if n is even
 n   2
Where  2    1
  n  1 , if n is odd
 2

 n 2
r n
Thus y  an   1 2r
x n 2 r
r 0 2 r n  2r
Taking an = 2n, then we get
 n 2
r n!
y  H n  x    1  2 x  n 2 r .....(12)
r 0 r n  2r
where Hn(x) is called the Hermite polynamial of order n.

14.3 Generating function

2

tn
Result. e 2 xt  t   H n  x  valid for all finite x and t.
n0 n
Proof. We have
2 2
e 2xt t  e 2xt  e t

2 s



 2 xt  r   t 
r
 s
r 0 s 0


 
 2 x r  1s t r 2 s
 r s
r 0 s 0

Let r + 2s = n so that r = n – 2s.

Hence the coefficient of t n (for fixed value of s) is given by
s n 2 s

 1  2 x 
n  2s s
The total value of t n is obtained by summing over all admisible value of s, and since r = n – 2s,
r  0.
Now as n – 2s  0 or s  n/2, therefore s goes from 0 to n/2 or from 0 to  n  1 / 2
according as n is even or odd.
So total coefficient of tn in the expansion of exp (2xt – t2) is given by
272
  n 2 1 s 2 x n 2 s
    Hn  x
 n  2s s

n
(From equation (12) of §14.2)
s 0

2
 Hn  x  n
 e2 xt t   t
n 0 n

14.4 Hypergeometric Form


 1s n
We have Hn  x    2 x  n 2 s .....(1)
s 0 s n  2s

n   n  1 2 s   n  2s 
Now    1
n  2 s   n  2s  1   n 

 n   n 1 
2 n  2 s1  1 2     s       s 
  2   2 2 
 n   n 1 
2 n 1  1 2         
 2  2 2

 n  n 1
 22 s       
 2 s  2 2 s
 n 2  1 s x 2 s   n 2    n  1 2 
n s s
Thus H n  x   2x 
s 0 s

n  n 1 n 1 
  2 x  2 F0   , ; ;  2  .....(2)
 2 2 x 

14.5 Recurrence Formulae

14.5.1. 2xHn(x) = 2nHn – 1(x) + Hn+1 (x)

Proof. We know that
2

tn
e2 xt t   Hn  x
n 0 n
Differentiating both sides w.r.t. ‘t’, we have
2

t n 1
or e 2 xt t  2 x  2t    n Hn  x
n 0 n

tn 
t n 1
or 2 x  t   Hn  x   Hn  x
n 0 n n 1 n  1
Equating the coefficients of t n on both sides, we get
2x 2 1
or Hn  x   H n 1  x   H n 1  x 
n n 1 n

273
or 2x Hn(x) – 2n Hn – 1(x) = Hn + 1(x)
or 2x Hn(x) = 2n Hn – 1(x) + Hn + 1(x).
14.5.2. H n(x) = 2nHn – 1(x) (n  1)
2

tn
Proof. We know that e2 xt t   Hn  x
n 0 n
Differentiating both side w.r.t. ‘x’ we have
2

tn
2t  e2 xt t   H n  x 
n 0 n


tn  n
t
or 2t  H n  x    H n  x 
n 0 n n 0 n

Equating the coetticients of t n on both sides, we get

2 1 1
or H n 1  x   H n  x   H n  x 
n 1 n n n 1
or Hn(x) = 2n Hn – 1(x)
14.5.3. Hn(x) = 2xHn – 1(x) – Hn + 1(x)
Proof. Form Recurrence relations 14.5.1 and 14.5.2, we have
2x Hn(x) = 2n Hn – 1(x) + Hn + 1(x) .....(1)
Hn(x) = 2n Hn – 1(x) .....(2)
Shbstracting (2) from (1), we have
or Hn(x) – 2x Hn (x) = –Hn + 1(x)
or Hn(x) = 2x Hn(x) – Hn+1(x)
14.5.4 H n (x) – 2xH n(x) + 2nHn (x) = 0
Proof. Hermite’s differential equation is

d2y dy
2
 2 x  2ny  0
dx dx
 Hn(x) is the solution of above differential equation, therefore.
Hn(x) – 2x Hn (x) + 2n Hn (x) = 0.

Self-Learning Exercise-I

1. H0(x) = ...... 2. H1(x) = ......


tn
3.  H n  x   ......
n 0 n

4. Write down Hermite differential equation.

5. Hn(x) = ...... 6. Hn(–x) = ......

274
Ex.1. Prove that H n (x) = 4n(n – 1) Hn – 2(x)
Sol. From rcurrence relation 14.5.2, we have
Hn(x) = 2n Hn – 1(x)
Differentiating with respect to x, we get
H ''n (x) = 2n H 'n 1 (x)
Again using recurrence relation 14.5.2, we find that
H n (x) = 2n × 2(n – 1) Hn – 2 (x)
= 4n (n – 1) Hn – 2 (x)
Ex.2. Prove that if m < n

d m  H n  x   2m n
 H  x
dx m n  m nm
Sol. We know that
2

tn
e2 x t t   Hn  x
n 0 n

d m  2 x t t 2   t n d m dm
 n m  n  
or e   H x  
dx m  0 n dx dx m
m
m 2 x t t 2

t n d  H n  x   d m
or  2t  e    m
n 0 n dx m dx

 n d m  H  x 
m

tr t  n  dm
or  2t   r H r  x    n  dxm  dxm
r 0 n 0

 n d m  H  x 

t r m t  n 
or 2m  Hr  x    m
r 0 r n 0 n dx
If r+m=n
[Note that r  0  n – m  0 or m  n]
 n d m  H  x 
m

tn t  n 
or 2  H nm  x     m
nm nm n 0 n dx
Equating the coefficient of t n on both sides, we get
m
2m 1 d  H n  x  
H n m  x  
nm n dx m

2m n d m  H n  x  
or H nm  x  
nm dx m
Ex.3. Prove that

275
 2n
(i) H2n (0) = (–1)n 
n
(ii) H2n + 1 (0) = 0
Sol. We have

tn 2
 n H n  x   e2 xt t
n 0
Putting x = 0 in this relation, we get

tn 2
 n H n  0   e t
n 0

n r


  t 
r 0 r



 1r t 2r
r 0 r
Note that R.H.S. contain only the terms of even powers of t. Equating the coefficient of t 2n on
both the sides, we get
n
1
H 2n  0  
 1
2n n

or H 2n  0 
 1  2n   1
n
n 1
 22 n  
n  2 n
Further equating the coefficient of t 2n +1 on both the sides, we obtain
H 2 n1  0   0

3
Ex.4. Prove that H 2 n  0   0 and H 2 n 1  0    1 22 n 1  
n

 2 n
Sol. We have
 n 2  1 s n
 
Hn  x    2 x n  2 s
s 0 s n  2s
Differentiating w.r.t. x, we get
 n1 2 s
2  1 n  2 x 
n  2 s 1
H n  x     n  2s 
s 0 s n  2s

  x   2
H 2n
n 1
 1s 2n  2 x 
2 n  2 s 1

Thus
s 0 s 2n  2 s  1

n  1s  2n  1 2 x 2 n2 s

and   x   2
H 2n+1
s 0 s 2n  2s
276
 Hence   0  0
H 2n

 1n  2n  1
and   0 
H 2n+1
n

n 3
  1 22n 1   (by using gamma duplication formula)
 2 n

14.6 Rodrigues Formula for Hn(x)

To Prove that H n  x    1 n e x

2  
d n e x
2

dx n
Proof. We have

H0  x  H1  x  t Hn  x
f  x, t     .....  t n  .....
0 1 n

2 2  x t 2
where f  x, t   e 2 xt t  e x e  

  n f  x, t   Hn  x
  n   n  Hn  x
 t  t 0 n



Hn  x  

  n e   x  t 2  e x 2

 t n 
  t 0

  ne  x t 2  2
   ex
 t n  .....(1)
  t 0

Let x – t = u that is t = x – u x = u at t = 0

 
Also x – t = u  
t u

  n e xt 2 
 n
   1 
 n e u   2


 t n  u n
 

  n e  xt 2 
 n
   1 
 n e x   2


 t n  x n
  t 0
277
   1 
n
d n e x   2

dx n
2
d n  e  x 
From (1), we get H n  x    1 e x
n 2
 
n
dx

14.7 Orthogonal Property of Hermite Polynomials


 x2
Theorem. Prove that e H n  x  H m  x  dx  2n n   mn where mn is Kronicar delta


 mn
 x2  0 if
or  e H n  x  H m  x  dx   n
   2 n if mn
Proof. We know that
2

tn
e 2 xt t   Hn  x
n 0 n

2 sm

e 2 xs  s   Hm  x
m0 s

2 2

tn 
sm
 e2 xt t  e 2 xs s   Hn  x   Hm  x
n 0 n m0 s

1 1
 H n  x  H m  x  = Coefficent of tnsm in the expansion of
n m
2 2
e2 xt t  e 2 xs  s

 x2
So e H n  x  H m  x  dx  n m times the coefficent of tnsm in the expansion of



 x 2 2 xt  t 2 2 xs  s 2
e e e dx .....(1)


 
2 2 2 2
s2  x 2  2 xt  2 rs
Now  e x e2 xt t e 2 xs  s dx  e t e
 

 2 2
2
s2  x 2  2 x  t  s    t  s   t  s 
 e t e dx


 2
t 2  s 2   x 2  2 x t  s    t  s   2
 e
  t s 
e  e dx

278
  2
  x  t  s  
 e2 st  e   dx




2st u2
e e du [where x – (t + s) = u and hence dx = du]


2 st

 2 st n
e      n
n 0


2n n n
   st
n 0 n

Here the series on right-hend side contains the terms having the equal powers of t and s. There-
fore the coefficent of tn sm, (m  n) will be zero. Equating the coefficent of tn sm on both sides of above
result,we get

2

 e x H n  x  H m  x  dx  0 where m  n


and from (1), we have


x 2 2n 
 e H n  x  H m  x  dx  n m n


 n 2n  , where m = n

x 2
Hence e H n  x  H m  x  dx  2n n   mn


  1 d 2   n
Ex.1. Prove that H n  x   2n exp .  
 4 dx 2   x
   
Sol. We have

   2te
d e2tx
2t x
dx


1d e 
2t x
 te2t x
2 dx
Differentiating w.r.t. x
d  1 d 2t x 
dx  2 dx
e  2 2tx
  2t e


279
 1 d 1 d 
 
2 dx  2 dx 
 
e 2t x   t 2e 2t x

2
 1 d  2t x 2 2t x
   e t e
 2 dx 
Hence by symmetry for n terms, we get
n
 1 d  2t x n 2t x
  e t e .....(2)
 2 dx 
Now,
n
  1 d 2   2t x   1  1 d 2   2t x
       e
exp.    e
2   n 4 dx
  4 dx    n 0   



 1n  1 d  2t x
2n

  e
n 0 n  2 dx 



 1n t 2n e2t x
[from (2)]
n 0 n

2t x

 1n t 2n
e  n
n 0


1 n
 e 2t x 
n 0 n
 
t 2

2 2
 e 2 t x e  t  e 2 x t t
  1 d 2   
1 n

tn
or  exp.   2 
  n  2t x   Hn  x
  4 dx   n 0 n 0 n

Equating the coefficent of tn on both sides we get

  1 d 2   1 n n 1
exp.    2 x  H n  x 
  4 dx   n n

 n  1 d 2   n
 H n  x   2 exp.     x
  4 dx  
which completes the solution of the problem.
Ex.2. Expand xn in a series of Hermite polynomials
Sol. We have
2

tn
e2 xt t   Hn  x
n 0 n

280
 2

tn
e2 xt  et  Hn  x
n 0 n


 2 xt n 
tn   t 2s 
   H n  x   
n 0 n n 0 n  s 0 s 

2n x n n   H
 x  n2s
  n t   n t
 0 n 0 s 0 n s

Put n + 2s = m n = m –2s since m – 2s  0.

 m 2 H m
m 2 s  x   t

 
n 0 s 0 s m  2s
Equating coefficent of tn on both sides

2n xn  n 2 H
  n 2 s
 x
n s 0 s n  2 s

 n 2 n H
n2 s  x 
 xn   2n s n  2s
s 0


2 t2 n
Ex.3. Prove that Pn  x   e t H n  xt  dt
n  0

This result is also known as Curzen’s integral.

 n 2
Sol.We know that H n  x    1s n  2 x n2s
 s n  2s
s 0

 n 2 1 s n 2 xt n 2 s
   
 H n  xt    s n  2s
s 0

2
  n 2  1s n  2 xt n2 s 
t 2 n
Now, RHS = n  e t   dt
0  s0 s n  2s 

 n 2 s
n  1  2 x 
n 2 s 
2 t 2 2 n  2 s

n 
 s n  2s e t dt
s 0 0

1
Put t 2    dt   1 2 d
2

1    1s  2 x n2s  e ns  12 1 d

n2
 R.H.S.  
 s 0 s n  2s 
0

1    1  2 x 
n2 s n2s
 1
 
 s 0 s n  2s
n  s  
 2
281
  1
 n 2 s n 2 s n  s  
 1  2 x   
2   1 2    
  s n  2s  1 2 
 
s 0

 n 2  1s  2 x n 2 s 1 2 n2 s

  s n  2s
s 0
= Pn(x) (by definition of Legendre polynomials)
 H n s  x  t n
Ex.4. Show that 
n 0 n
 
 exp 2 xt  t 2 H s  x  t 

Sol. Consider
  H n s  x  t n s  n H n  x  t n s s
  n s
  ns s
n 0 s 0 n 0 s 0

 n  n s  1s H n  x  t nu s    s
   
n 0 s 0 n s t

s
 Hn  xtn n  n s 
    
n 0 n s 0 s  t

 H n  x  t n   n
 1  t 
n 0 n  

n
 H n  x   t 

n 0 n

2
e    
2x t   t 

2 2
 e 2 xt t  e 2  x t 

2
 H s  x  t  s
 e2 xt t 
s 0 s

s
Comparing the coeffiecient of , we get the required result.
s!
Ex.5. Establish

  c n H n  x  t n c
c c 1 4t 2 
   1  2 xt  2 F0
 ,  ; ;  
2  .....(3)
n 0 n 2 2 2  1  2 xt  


Sol. We have
282
   n 2  1s  c n  2 x n2 s t n
L.H.S. of (3)   
n 0 s 0 s n  2s
Now using a well-known result
  n 2  
  A  k , n     A  k , n  2k  , we get
n 0 k 0 n 0 k 0

   1s  c n 2 s  2 x n t n 2 s
L.H.S. of (3)  
n 0 s 0 s n

  c n  2 s   c  2 s n  c  2 s

  1 s  c 2 s t 2 s   c  2s n  2 xt n
 L.H.S. of (3)   s
 n
s 0 n 0

  1s  c 2 s t 2 s
 1  2 xt c2 s
s 0 s

c   c 1
But  c 2s  22 s     
 2 s  2 2 s

s
c
  c 2s  c  1 2 s  4t 2 
Hence L.H.S. of (3)  1  2 xt    
s 0 s  1  2 xt 2 
 

c
c c 1 4t 2 
 1  2 xt  F
2 0  ,  ;  ;  
2 
2 2 2 1  2 xt 
 
The relation (3) is called the Braf man’s generating function.
x
2 2

Ex.6. Prove that  e  y H n  y  dy  H n 1  0   e  x H n1  x  .....(4)

0

Sol. Using Rodrigue’s formula in the left–hand side of (4), we get

x x x
d n  y2 n d
n 1
 y2 

0
e  y2
H n  y  dy    1
0
n

dy n
e dy  
 1 
 dy
n 1
e 
0
 
x
 2
  e  y H n1  y  0

(Using again the Rodrigue’s formula)

2
 H n1  0   e  x H n1  x 

283
 Self-Learning Exercise-II


2

1.  e x H n  x  H m  x  dx  ............  if m  n 


2. Write down Rodriques formulas for Hn(x).

3. H2n + 1(0) = ...........

14.8 Summary

In this unit, we studied the Hermite differential equation and Hermile polynomials. We also stud-
ied recurrence relation, generating function, Rodrigue, formula and orthogonal property for Hermite poly-
nomials.

14.9 Answers to Self-Learning Exercises

Exercise I
1. e2 xt t 2

d2y dy
2.  2 x  2 y  0
dx 2 dx
3. 2nH n1  x 

Exercise II
1. 0

2. H n x    1 n e x
2  
d n e x
2

dx n
3. 0

14.10 Exercise


2
1. Evaluate  xe  x H m  x  H n  x  dx (m  n) [Ans : 0]


2. Prove that H 5  x   32 x 5  160 x3  120 x

3. Prove that H 2  x   4 x 2  2

4. Express H  x   x 4  2 x3  2 x 2  x  3 in terms of Hermite polynomials.

5. Prove that x H n  n   n H n 1  x   n H n  x 

284
 
2 2  1
6. Prove that  x2e x  H n  x  dx   2n n  n  
  2

n H k  x  Hk  y  H n  x  H n1  y   H n 1  x  H n  y 
7. Show that  k
2 k!

2n 1  y  x  n
k 0

2  t 2 n 1
8. Evaluate 2n 1e x x e t Pn  x t  dt [Ans : Hn(x)]

  x2
9. Evaluate   xe H m  x  H n  x  dt , m  n [Ans : 0]

10. If  n  x   e  x 2 2 Hn  x  , then prove that

 n
(i)   ψm  x  ψn  x  dx  2 n  m,n if m  n  1

 0, if m  n 1
  n 1
(ii)   ψ m  x  ψn  x  dx  2 n , if m  n 1
 n
2 n  1  , if m  n 1

11. Using the expansion of x n in a series of Hermite polynomials, show that

  x2 n n 
  e x H n  2 k  x  dx  22 k
k



285
Unit 15 : Laguerre Polynomials
15.0 Objective

15.1 Introduction

15.2 Definition

15.3 Generating Function for Ln (x)

15.4 Recurrence Formulae for Ln (x)

51.5 Rodrigue’s Formula for Ln (x)

15.6 Orthogonal Property for Ln (x)

15.7 Associated Laguerre Polynomial : Definition

15.8 Generating Function for Associated Laguerre Polynomial

15.9 Recurrence Formulae for Associated Laguerre Polynomial

15.10 Rodrigue’s Formula for Associated Laguerre Polynomial

15.11 Orthogonal Property for Associated Laguerre Polynomial

15.12 Summary

15.13 Answer to Self-Learning Exercises

15.14 Exercise

15.0 Objective

In this unit you will study Laguerre and associated Laguerre polynomials and their important prop-
erties such as generating function, orthogonal property, Rodrigue’s formula, recurrence relations etc.

15.1 Introduction

The purpose of this unit is to introduce and study the Laguerre and associated Laguerre polyno-
mials. We shall state and prove certain important properties associated with these classes of polynomi-
als.

15.2 Laguerre’s Differential Equation and Its Solution

THe Laguerre differential equation of order n is

d2y dy
x 2
 1  x   ny  0, .....(1)
dx dx
where n is a positive integer
286
 Now we apply the method of Frobenius for its solution which is finite for all values of x and
which tends to  no faster than ex/2 as x .
Proceeding on lines similar to explained in the case of Legendre, and Hermite polynomials, we
find that if we assume the solution of (1) in the form

y   ar x r .....(2)
r 0

n
r n
then y  a0   1 2
xr .....(3)
r 0 nr  r
will be solution of equation (1). Taking a0 = 1, the corresponding solution of equation (1) is known as
Laguerre polynomial of order n, and which is denoted by Ln (x). Thus
n
r n
Ln  x     1 2
xr
r 0 nr  r
 1 F1  n ; 1; x  .....(4)
Some times we take a0 as n , then alternative definition of Laguerre polynomials is
2
n
Ln  x     1
 n
xr
r
2 .....(5)
r 0 n r  r

15.3 Generating Function for Ln (x)

Theorem : Show that

xt


1 t
e
  Ln  x   t n
1  t n0
Proof : Using the exponential series we have
xt r
 
e t
1 1 1  x t 

1 t
 r  1  t 
1 t r 0


 1r x r t r 1  t 
 r 1
  r
r 0


 1r x r t r   r  1s t s
  r
 s
r 0 s 0

  1r  r  s  xr t r s

  ....(1)
r 0 s 0  r 2 s
287
 For a fixed r, the coefficient of tn is

r n xr
  1
 r 2  n  r 
Taking n = r + s.
Now s = n – r and s  0, so r  n.
Hence the total coefficient of tn in (1) is
r

 1 n x r
  Ln  x 
2 (By definition)
s 0  r   n  r 

xt
 
Hence e 1t   Ln  x  t n
1 t n 0

15.4 Recurrence Relations for Ln (x)

15.4.1  n  1 Ln1  x    2n  1  x  Ln  x   n Ln1  x 

Proof : From generating function, we have
xt 
1 
n
e 1t  t Ln  x  .....(1)
1  t  n 0

Differentiating (1) w.r.t. ‘t’ we get

xt xt 

n 1 1 
1t 1  x 
 nt Ln  x   2
e  e 1t   2
n 0 1  t  1  t   1  t  
 
1 x
  t n Ln  x   2  t n Ln  x 
1  t  n 0 1  t  n 0

Multiplying both the side by (1 – t)2 we get

  
1  2t  t  
2

n 0
nt n 1
Ln  x   1  t   t Ln  x   x  t n Ln  x 
n 0
n

n 0

  
  nt n1 Ln  x   2  nt n Ln  x    n Ln  x  t n 1
n 0 n 0 n 0

  
n n 1
 t Ln  x    t Ln  x   x  t n Ln  x 
n 0 n 0 n 0

Now equating the coefficent of tn on both sides, we get

 n  1 Ln1  x   2n Ln  x    n  1 Ln1  x   Ln  x   Ln1  x   x Ln  x 
  n  1 Ln1  x    2n  1  x  Ln  x   n Ln1  x 
288
15.4.2 x Ln  x   n Ln  x   n Ln1  x 
Proof : From generating function
 xt
n 1 
t Ln  x   e 1t .....(2)
n 0 1  t 
Differentiating w.r.t. ‘x’ we get
 xt
n
t Ln  x  
1   t 
 1  t 
e 1t  
 1 t 
n 0

 
t
or  t n Ln  x     t n Ln  x 
n 0 1 t n 0

 
or 1  t   t n Ln  x   t  t n Ln  x 
n 0 n 0

  
or  t n Ln  x    t n 1 Ln  x     t n 1 Ln  x 
n 0 n 0 n 0

Equating the coefficients of tn on both sides, we get

Ln  x   Ln 1  x    Ln1  x 

or Ln 1  x   Ln 1  x   Ln 1  x  .....(3)

Differentiating Recurrence relation 15.4.1, we find that
 n  1 Ln1  x    2n  1  x  Ln  x   Ln  x   n Ln1  x  .....(4)
Replacing n by (n + 1) in (3), we obtain
Ln  x   Ln 1  x   Ln  x  .....(5)
Putting the value of Ln 1  x  and Ln 1  x  from (3) and (5) in (4) we get

 n  1  Ln  x   Ln  x    2n  1  x  Ln  x   Ln  x   n  Ln  x   Ln 1  x  

n Ln  x   n Ln  x   Ln  x   Ln  x 

 2nL1n  x   Ln  x   x Ln  x   Ln  x   n Ln  x   n Ln 1  x 

On simplification, we get x Ln  x   n Ln  x   n Ln1  x 

n 1
15.4.3 Ln  x     Lr  x 
r 0
Proof : From generating function
 xt
1 
 t n Ln  x   e 1t .....(6)
n 0 1  t 
Differentiating (6) w.r.t. ‘x’, we get
 xt
n 1   t 
or  t Ln  x   e 1t 
n 0 1  t  1  t 
289
 
1
 t 1  t   Ln  x  t r (using Bionomial theorem)
r 0

 
   Ln  x  t r  s1 .....(7)
r 0 s 0
Taking r + s + 1 = n, we have s = n – s – 1. But s  0 therefore r  n – 1
n 1
So the total coefficient of t R.H.S. is   Lr  x 
n

r 0
n
Now equating coefficient of t on both sides in (7), we arrive at the required recurrence relation
15.4.3.

15.5 Rodrigue’s Formula for Ln (x)

Prove that
ex dn
Ln  x  
n dx n
x ne  x  
Proof : Using Leibnitz’s theorem for n times differentiation, we have

ex n n  x
R.H.S. 
n
D x e  
ex n

n
 n cr Dnr x n  D r e x
r 0

ex n
n
 ncr  n   n  r  xn nr   1
r
 e x
n r 0

n
ex  n 2 xr 1 r e x
 2  
r 0 n  r n  r
r
n
 1 n xr
 2
 Ln  x 
r 0  r  n r

15.6 Orthogonal Property

Prove that

x  0, if mn
e Lm  x  Ln  x  dx   mn  
0  1, if mn
Proof : From generating function, we have
 xt
t n Ln  x   1 e 1t

 .....(1)
n 0 1 t
290
  xs
s m Lm  x   1 e 1 s

 .....(2)
m 0 1 s
Multiplying (1) and (2), we get
   t s 
1 1  x  
n m
 Ln  x  t Lm  x  s   e 1t 1 s 
n 0 m0 1  t  1  s 
    x  n m
     e Lm  x  Ln  x  dx  t s
n 0 m0  0 
  t s 
1  x 
x  1t 1 s 
 e e dx
1  t 1  s  0
  t s 
1  x 1 
 1t 1 s 
 e dx
1  t 1  s  0

  t
 x 1 
s  
1  e  1t 1 s  
  
1  t 1  s   1  t  s  
  1  t 1  s  
0

t s 
  x 1   


1

1  t 1  s   e  1t 1 s  
1  t 1  s  1  t 1  s   t  ts  s  st   
 0

1
   0  1
1  s  t  ts  t  ts  s  st 
1 1
  1  st 
1  st 
2 n
 1  st   st   .....   st   .....
Equating the coefficients of tn sn on both sides, we get

x
e Lm  x  Ln dx  0 if m  n .....(3)
0
and equating the coefficient of tn sm, we get

x  L  x 2  dx = 1
e  n 
0

x
That is e Ln  x  Lm  x  dx  1 (when m = n) .....(4)
0


x
Combining (3) and (4), we get e Lm  x  Ln  x  dx   mn
0
291
Ex.1. Prove that
 n
 st 1 1
 e Ln  t  dt =  1  
0
s s


 st
n
 1r n t r dt
Sol. L.H.S.   e  2
0 r 0  r  n  r

r
n
 1 n 
 2 
e  st t r 11dt
r 0  n  r  r  0

r

n
 1 n   r  1
2
r 0  n  r  r  s r 1

1 n
 1r n  1
  r
s r 0 n  r  r  s

n r n
1 n  1 1 1

s
 cr     1  
 s s s
r 0
= R.H.S.
n  n  1
Ex.2. Prove that (i) Ln (0) = 1, (ii) Ln  0  n and (iii) Ln  0  
2
Sol. We know that
xt 
1 1t 
e  t n Ln  x  ....(5)
1 t n 0
Taking x = 0 in (5), we get

or 1  t  1   t n Ln  0 
n 0

 
or  t n   t n Ln  0 
n 0 n 0
n
Equating coefficients of t on both sides, we get
1  Ln  0 
(ii) From Laguerre differential equation, we have
xy + (1 – x) y + ny = 0
If Ln(x) is the solution of this equation then
x Ln  x   1  x  Ln  x   n Ln  x  = 0
Putting x = 0, we get
Ln  0    n Ln  0 
 n  1 [from (i)]
292
 Thus Ln  0   n
(iii) Differentiating twice w.r.t ‘x’, (1) gives
 xt 1t  2 
e  t 
    Ln  x  t n
1 t  1 t  n 0
Putting x = 0, we get

n
 L  0 t
n 0
n  t2 1 t 
3
.....(6)

Equating the coefficients of tn on both the sides of (6), we find that

Ln  0  = Coeff. of tn in t2 (1 – t)–3
= Coeff. of tn – 2 in (1 – t)–3
 3 3  1 .....3   n  2   1 n2
  1
 n  2
3.4. ..... n n n  n  1
  
 n  2 2 n  2 2

Self-Learning Exercise–1

1. Laguerre’s differential equation is .........


x
2. e Lm  x  Ln dx = ..... if m  n
0


x  L  x 2  dx = ...............
3. e  n 
0

4. Ln (0) = ...............
5. ........ = nLn(x) – nLn–1(x)
6. L0(x) = ...............
7. L1(x) = ...............
8. L2(x) = ...............

15.7 Associated Laguerre Polynomial : Definition

Associated Laguerre polynomials of degree n and order k is denoted and defined as

dk k
 x Lkn   1
Ln k  x  .....(1)
dx k
Now using the series representation for Laguerre polynomials we find that

k dk nk
r n  k 
Lkn  x    1   1 xr
dx k r 0  n  k  r   r 2
293
 k
nk
r n  k  dk r
  1   1 x
.....(2)
r 0  n  k  r   r 2 dx k

 0, if rk
dk r 
Now x   r r k
dx k  r  k x , if rk

n k k 1 n k
Hence breaking  into two sums as  and  , we find that
r 0 r 0 r k

k
nk
r k n  k 
Lkn  x   1   1 x rk
rk  n  k  r  
r r  k 
Let r – k = s, so that r = s + k and when r = k, s = 0 and r = n + k, s = n. Then
n
s 2k n  k 
Lkn  x    1 xs
s 0 n  s s  k  s

n
r n  k 
or Lkn  x    1 xr .....(3)
r 0 n  r  k  r  r

15.8 Generating Function for Associated Laguerre Polynomials

Prove that

1  xt   k n
exp      Ln  x  t
    n 0
k 1
1  t  1  t
Proof : By generating function for Laguerre polynomial, we have

1   xt  n
exp     Ln  x  t .....(1)
1  t  1  t  n0
Differentiation both sides of (1) ‘k’ times w.r.t. ‘x’, gives

1 dk   xt  
n d
k

1  t  dx k   1  t  n
exp     t Ln  x 
0 dx k
k k 1 k  k
1  t   xt  n d n d
or 
1  t   1  t 
exp    t  Ln  x    t k Ln  x 
 1 t  n 0 dx k n k dx

k tk  xt  
n d
k
or  1 exp     0  t Ln  x  .....(2)
1  t k 1  1  t  n k dx k
Here we use that Ln (x) is a polynomial of degres n so that

dk  0 if n  k
k  n  
L x  
dx non - zero if n  k
294
 Multiplying by (2) by (–1)k then we get
tk  xt  k

dk
k 1
exp      1  tn Ln  x 
1  t   1 t  nk dx k

tk  xt  k

dk
 k 1
exp      1  t sk Ls k  x 
1  t   1 t  s 0 dx k
(Taking s as new variable such that n = s + k i.e. s = n – k so when n = k, s = 0 and when n
tends to, s also tends to )

tk  xt  
k dk
 k 1
exp     tk   1 dxk Ls k  x  t s
1  t   1  t  s 0

 k
1  xt  k d
or k 1
exp      1   Ln  k  x  t n
1  t   1  t  dx k
n 0
( The limit remain same so we can change the variable from s to n)

1  xt  k n
exp     Ln  x  t
1  t k 1  1  t  n 0

15.9 Recurrence Relations for Lkn  x 

15.9.1 Lkn1  x   Lkn1  x   Lkn  x 

n
r n  k 
Proof : We know that Lkn  x     1 nr k r r
.....(1)
r 0
Replacing n by (n – 1) in (1), we find that
n 1  1r  n  k  1
r
Lkn 1  x     1  n  r  1 r k  r xr .....(2)
r 0

Replacing k by (k – 1) in (1), we get

n  1r  n  k  1
Lkn1  x    xr .....(3)
r 0 n  r k  r 1 r
Using (2) and (3), we have

n 1  1r  n  k  1 r n  1r  n  k  1
Lkn 1  x  Lkn1  x    n  r  1 k  r r x   xr
r 0 r 0 n  r k  r  1 r

n 1  1r  n  k  1 r n1  1r  n  k  1 r

   n  r  1 k  r r x   x
r 0 r 0 n  r k  r  1 r

 1n  n  k  1 x n

 n  n k  n 1 n
295
 n 1  1r  n  k  1 r nk
n
 1 xn
 x  
r 0  n  r  1  k  r  1 r  k  r  n  r  n


n 1
 1r n  k xr

 n
1 x n
r 0 nr k r r n


n
 1r n  k xr
r 0 nr k r r

 Lkn  x  [by (1)]

15.9.2  n + 1 Lkn+1  x  =  2n + k + 1 - x  Lkn  x  -  n + k  Lkn-1  x 

Proof : From recurrence relation 15.4.1 for Laguerre polynomial we have
 n  1 Ln1  x    2n  1  x  Ln  x   nLn1  x  ...(4)
Replacing n by (n + k) in (4), we get
 n  k  1 Lnk 1  x    2n  2k  1  x  Ln x  x    n  k  Lnk 1  x 
Differentiating k times, the above equation becomes

dk dk
 n  k  1 k  Lnk 1  x    2n  2k  1 k Lnk  x 
dx dx

dk dk
  nk     k Lnk 1  x 
xL x  n  k ...(5)
dx k dx
Using Leibnitz’s theorem, we get
dk dk k d k 1
 n k  
xL x   nk  
L x x  c1 Lnk  x 
dx k dx k dx k 1

dk d k 1
x
Lnk  x   k Lnk  x  ...(6)
dx k dx k 1
Using (6) in (5) and then multiplying both sides by (–1)k, we get

k dk
 1  n  k  1 Ln k 1  x 
dx k

k dk k dk
  1  2n  2k  1  n k    
L x  1 x Lnk  x 
dx k dx k

k 1 d k 1 k dk
  1 k  nk 11       k Lnk 1  x  ....(7)
L x   1 n  k
dx k 1 dx
k
But from definition Lkn  x    1 k d Ln k  x  ....(8)
dx k
Using (8) in (7), we get
296
  n  k  1 Lkn1  x    2n  2k  1 Lkn  x 
 xLkn  x   kLkn11  x    n  k  Lkn 1  x  .....(9)
Replaceing n by n + 1 in 15.9.1, we get
Lkn  x   Lkn11  x   Lkn 1  x 

or Lkn11  x   Lkn 1  x   Lkn  x  .....(10)

Eliminating Lkn11 from (10) and (9), we get

 n  k  1 Lkn1  x    2n  2k  1 Lkn  x   x Lkn  x   k Lkn1  x   Lkn  x    n  k  Lkn1  x 

That is  n  1 Lkn1  x    2n  k  1  x  Lkn  x    n  k  Lkn1  x 
d k
15.9.3 Ln  x    Lkn11  x 
dx
n
 1r n  k xr
Proof : We know that Lkn  x   nr k r r
.....(11)
r 0

Differentiating both side of (11) w.r.t. ‘x’ we get

d k
n  1r  n  k  rxr 1
L.H.S.=
dx
Ln  x    n  r k  r  r
r 0

n
 1r n  k x r 1
  n  r k  r r 1
r 1

n 1
 1s1 n  k xs
  n  s 1 k  s  1 s
(Taking r – 1 = s)
s 0

n 1  1s  n  1  k  1 xs
  1 
s 0 n  s 1 k  s 1 s

d k
Ln  x    Lkn11  x   R.H.S
dx

15.10 Rodrigue’s Formula for Lkn  x 

Theorem : Prove that

e x x k d n
Lkn  x  
n dx n

x n k x  x 
Sol :

e x x k n  x nk
R.H.S. 
n
D e x  
297
 e x x k n

n
 n cr D nr x nk  D r e x (by Leibnitz theorem)
r 0

e x x k n
n n  k x n  k  n r  r
  cr  1 e x
n r 0 n  k  n  r

n
e x x k n n  k x k r r
      1 e  x
r 0 n r nr kr

n
 1r n  k xr
  nr k r r
r 0

 Lkn  x   L.H.S

15.11 Orthogonal Property for Associated Laguerre Polynomial

Theorem : Prove that


nk
 e  x x k Lkn  x  Lkm  x  dx   mn
0
n

Proof : Associated Laguerre differential equations is

d2y dy
x 2  1  x  k   ny  0 .....(1)
dx dx
Multiplying by xk e x we have

k x d2y dy
xx e 2
 1  x  k  x k e  x  ny x k e  x  0
dx dx

d  k 1  x dy 
or  x e   n x k e x y  0 .....(2)
dx  dx 

Since associated Laguerre polynomial Lkm  x  and Lkm  x  satisfy the equation, therefore

d
So  x k 1e  x DLkn  x    nx k e x Lkn  x   0
dx  

d k 1  x
and  x e D Lkm  x    m x k e  x Lkm  x   0 .....(4)
dx  

Multiplying (3) by Lkm  x  and (4) by Lkn  x  and then substracting, we have

d  x k 1 d
Lkm  x  e x D Lkn  x    Lkn  x   e x x k 1 D Lkm  x  
 
dx dx  

  m  n  x k e  x Lkm  x  Lkn  x  x  .....(5)

Integrating both sides of (5) w.r.t. ‘x’ from 0 to , we have
298
  
d   x k 1 k
 m  n  x k e  x Lkm  x  Lkn  x  dx   Lkm  x  e x D Ln  x   dx 
0 0
dx 

d   x k 1 k
 Lkn  x  e x D Lm  x   dx
0
dx 
  k
  Lkm  x  e  x x k 1D Lkn  x     L  x  e  x x k 1D Lkn  x  dx
0 0 m
 
  Lkn  x  e  x x k 1D Lkm  x     Lkn  x  e  x x k 1D Lkn  x  dx .....(6)
0 0
= 0 if m  n

Hence  x k e  x Lkm  x  Lkn  x  dx  0, if m  n.
0
If m = n then we find value of
 
e x x k d n  x nk

k x
x e Lkn Lkn
 x   x  dx   x k e x Lkn  x 
n dx n
e x dx  
0 0


1

n  
Lkn  x  D n e  x x n  k dx 
0

1  k 

n
 Ln  x  D

n 1

x n k e  x 
0 

  D Lkn  x  D n 1 x n  k e  x dx 
0


1
 0   D Lkn  x  D n 1 x n  k e x dx
n0
 
n
 1  n k
 D L  n  x   x nk e x  dx (by symmetry for n terms)
n 0
n 
 1 n
   1 x n k 11e  x dx
n 0


1
  x n  k e  x dx
n0

nk
 .....(7)
n
Combining (6) and (7), we have

nk
 e  x x k Lkn  x  Lkn  x  dx   mn
0
n

299
 
Ex.1. Prove that  e  t Lkn  t  dt  e  x  Lkn  x   Lkn 1  x  
x

Sol. Integrating by parts taking e–t as second function, we get

 

 e t Lkn  t  dt   e t Lkn  t     e t DLkn  t  dt
x
x x


 e  x Lkn  x   e  t DLkn  t  dt
x


 n 1   n 1

 e x Lkn  x    e t    Lkr  t   dt k k
 DLn  t    Lr  t  
x  r 0   r 0 

 n1 
  e t Lkn  t  dt    e t Lkr  t  dt  e x Lkn  x  .....(8)
x r 0 x

n 
or  e  t Lkn  t  dt  e x Lkn  x  .....(9)
r 0 x

Subtracting (9) from (8), we get

n   n1 
or  e t Lkr  t  dt   e t Lkn  t  dt    e t Lkr  t  dt  0
r0 x x r 0 x

 n  n 1 
or  e t Lkn  t  dt    e t Lkr  t  dt    e t Lkr  t  dt
x r 0 x r 0 x

or  e t Lkn  t  dt  e  x Lk  x   e  x Lk [using (9)]

n n 1
x


or  e t Lkn  t  dt  e  x  Lkn  x   Lkn 1  x  
 
x

Ex.4. Establish the generating functions :



(i)   1    x t 
 2

et J n 2 x t    1 1 
n0
Ln  x  t n
n

1  xt    c n 
(ii) F
1 1 c ; 1   ;    Ln  x  t n
1  t c  1  t  n0  1   n
Sol. (i) We have

1
Ln  x  t n  
 n
 1k x k t n
 1   n  k n  k 1    k
n 0 n 0 k 0

300
    
Using  A k, n    A  k , n  k  , we get
n 0 k 0 n 0 k 0


1  
 1k t nk xn
  1    Ln  x  t n   
n 0 n n 0 k 0 k n 1    k


tn 
  x t k
 
n 0 n k 0 k 1   k

 et 0 F1   ;1   ;  x t  .....(10)
We know that
n
 z 2  z2 
Jn  z   F
0 1   ; 1  n ;   .....(11)
  n  1  4
Using (11) in (9) we get the required generating function (i)
(ii) We have
  c n   c n  1k x k t n
n
Ln   x  t n 

 1    
n 0 n n 0 k 0 k n  k 1    k

  c nk  1k xk t nk


 
n 0 k 0 k n 1   k

k n

 c  k   x t    c  k n t
 
k  0 k 1    k n  0 n

 c k   x t k
 1  t c k
k 0 k 1    k

1  xt 
 F c
c 1 1
; 1  ;  
1  t   1 t 

n
Ex.5. Prove that Ln   1  x  y    Lr  x  Ln  r  y 
r0
Sol. We have
xt   yt    x  yt 
1  t 1 exp   1  1   1
 1  t  exp     1  t  exp  
1 t 

 1 t   1 t  
Therefore
  
 Ln   1  x  y  t n   Ln  x  t n  Lr  y  t n
n 0 n 0 r 0
 n
  Ln  r  x  Lr  y  t n
n 0 r 0
n
Comparing the coefficients of t , we get the required result.
301
  1    n  1  y  n r y r Lr  x 
n
Ex.6. Prove that Ln  x y  
r 0 n  r 1   r
Sol. We know that

t
 Ln  xy  t n
e 0 F1   ; 1   ;  xy t  
n 0 1   n
Now,
et 0 F1   ; 1   ;  xy t   e1 y t e yt 0 F1   ; 1   ;  x y t 


1  y n Lr  x  y r t nr

 
n 0 r 0 n 1   r
n r
 Ln  x y  t n 
1  y  t  Lr  x  yt 
  1    
n
 1   
n 0 n n 0 r 0 r

 n
1  y nr t n Lr  x  y r
  n  r 1   
n 0 r 0 r
n
Comparing the coefficients of t we get the required result required.

Self-Learning Exercise-II
1. Associated Laguerre differential equation is .........

2.  e  x x  k Lkm  x  Lkn  x  dx  ........ if m  n.
0
3. Ln + k is a Laguerre polynomial of degree ........
4. Lkn 1  x   Lkn1  x   ........

15.12 Summary

In this unit we studied the Laguerre and asociated Laguerre polynomials. we also studied the
recurrence relation, generating funciton and orthogonal property for these polynomials.

15.13 Answer to Self-Learning Exercises

Exercise-I

1. x y   1  x  y   n y  0 2. 0
3. 1 4. 1
5. x Ln  x  6. 1
1
7. 1 – x 8.
21

2  4 x  x2 
302
 Exercise-II

1. x y   1  x  k  y   n y  0 2. 0

3. n + k 4. Lkn  x 

15.14 Exercise

1. Find the value of


(i)  e x L3  x  L5  x  dx [Ans. 0]
0


2
(ii)  e  x  L4  x   dx [Ans. 1]
0
2. Express 10 – 23x + 10x2 – x3 in terms of Laguerre polynomials.
[Ans. L0(x) + L1 (x) + 2L2 (x) + 6L3 (x) ]

3. Prove that  e y Ln  y  dy  e x  Ln  x   Ln 1  x  
x

t n
 1 H 2 n 1  t 2 
4. Show that  Ln n  t  x  dx 
0 2 2 n  3 2 n

n r
n
 1   k  n  1 x n r
5. Show that Lkn  x    n  1, 2, 3,......
r  0 r n  r   k  n  r  1

6. Prove that
(i) H 2 n  x    1 22 n n Ln1 2  x 2  
n

n
(ii) H 2 n 1  x    1 2 2 n 1 n L x 12
n
2

1
t 2
1 n 1  t2 
7. Show that   x  t  x  2 H 2 n  x  t  x  dx   1  22 n   Ln  
0  2 n  4 

n    s Lns  x 
8. Show that Ln  x  
s 0 s!

2 n  k 
9. Show that  
e  x x k 1 Lkn  x  dx  n
 2n  k  1
0

x
m m n m 1 m 1
10. Prove that   x  t  Ln  t  dt  m  n 1
x Ln  x 
0


303
 Reference Books

1. Differential Equations Vol. II

J.L. Bansal and H.S. Dhami,
Jaipur Publishing House, Jaipur.

2. Advanced Differential Equations

M.D. Raisinghania, S. Chand & Comp.,
New Delhi.

3. Calculus of Variations
Robert Weinstock,
Dover Publications.

4. Special Functions
R.K. Saxena and D.C. Gokhroo,
Jaipur Publishing House, Jaipur.

5. Special Functions
E.D. Rainville, Chelsea Publishing Comp.,
Bronx , New York.

304