ORDINARY DIFFERENTIAL EQUATION

1 Introduction
Equation involving derovatoves of 1 or more dependent variables with respect to one or more
independent variables is called a differential equation.
Classifications: ODE: differential equations involving variables with respect to single inde-
pendent variable is called ODE.
PDE: A differential equation involving parital derivatives of one or more dependent variables
with respect to two or more independent variables.
Order and degree of DE: The order of a differential equation is the order of highest derivative
present in that equation.
The degree of the differential equation is the degree of the highest ordered derivative present
in that equation.
Example:
d2 y
• = xy, this is a second order first degree ODE.
dx2
∂u ∂u
• x +y = xy, first order first degree PDE.
∂x ∂y
∂2y ∂2y ∂2y
• + = , second order PDE.
∂x2 ∂t2 ∂z 2
If the independent variable is not appering explicity in a DE then it is said to be homogeneous
differential equations otherwise differential equation is non-homogeneous.
Example:
∂2y
• = xy,Homogeneous DE.
∂x2
00 0
• y + xy + y = 2, non-homogenous DE.
∂2y
• + y − sin2 x = 0, non-homogeneous DE.
∂x2
Linear and non-linear differential equations: A differential equation is said to be linear of
the dependent variable y and its derivatives occur to the first degree only and that no product
of dependent variable and that are its derivatives are non-transendental functions of dependent
variable and are its derivatives occur.
Example:
∂2y dy
• +2 − 4y = 0, LinearODE, homogeneous De.
∂x2 dx
d3 y 2
3 d + x4 dy = xe2 , Linear ODE.
• + x
dx3 dx2 dx

d2
+ 5y 3 = 0, non − linear ODE, (1)
dx2
d2 dy
2
+ 4( )2 + 6y = 0, non − linear ODE, (2)
dx dx
d2 dy
2
+y + 10y = 0, non − linear ODE (3)
dx dx

1
 dy dy
becuase of the terms y 2 , ( )2 , y in the equations (1, 2, and 3) respectively.
dx dx
Examples of Differential equations:
dy
• + xy = sinx, firest oder linear non-homogenous ODE,
dx
dy du
• x +y = xy, first order linear PDE,
dx dy
d2 dy
• 2
+ ( )2 + y = ex , second order non-;inear ODE
dx dx
dy
• + siny = 0, first order non-linear ODE.
dx
nth orderLinearODE : Let

f (x, y, y 0 , ..., y (n−1) , y (n) ) = 0, (4)

nth orderlinear ODE where f is a real function of n + 2 arguments.

dn y d(n−1) dy
x, y, y 0 , ..., y (n−1) , y (n) = a0 (x) n
+ a1 (x) (n−1)
+ ... + a(n−a) (x) + an (x)y = b(x), (5)
dx dx dx

eqution (5) general form of nth order linear ODE.

If dx 6= 0 then equation (5) is called non-Homogenous differential equations. Using simming
equations (5) can be written as
n
X dn−r
ar (x) n−r = b(x) (6)
dx
r=0

Pn dn−r
Ln y = b(x), where Ln = r=0 ar (x) is an nth order differential operator in particular
dxn−r
n = 2, 3

d2 dy
L2 y = b(x)a0 (x) + a1 (x) + a2 (x) = b(x)andL3 = b(x) (7)
dx2 dx
d3 y dy dy
a0 (x) 3 + a1 (x) + a2 (x) + a3 (x)y = b(x) (8)
dx dx dx
a solution of equation (4) is a real function if defined on an interva I having derivatives upto nth
order such that f (x), f (x0 ), f 0 (x), f 00 (x), ..., f (n−1) (x), f (n) (x) = 0, for all x ∈ I.
Initial value and Boundary value problems: The problem of finding solution of a differential
equation ssubjected to the conditons prescried at only one point is called an intial value problem.
Example:
dy
• dx + y = x2 , y(x0 ) = y0
d2 y dy
• dx2
+ x dx + xy = ex · x

A problem of finding the solution of the differential equation subjected to the conditions
prescribed at more than one point is called a boundary value problem.
Example:
d2 y dy
+ x2 + xy = xex , y(x0 ) = y0 , y 0 (x1 ) = y1
dx2 dx
d y 3 d2 y
Example 1. Solve x dx3 − dx2
=0

2
 d3 y d2 y
x3 − =0
dx3 dx2
Put y = ex and z = log x

dz 1
=
dx x
dy dy dz 1 dy
= · =
dx dz dx x dz
dy dy
x =
dx dz
dy d
x = Dy, where D =
dx dz
Differentiate with respect to x.

d2 y
 
dy d dy
2
+ =
dx dx dx dz
 
d dy dz
= ·
dz dz dx

d2 y 1
= ·
dz 2 x

d2 y dy d2
x2 +x = D2 y, where D2 =
dx2 dx dz 2
Differentiate with respect to x:

d3 y d2 y d2 y d2 y
 
dy d dz
x2 3 + 2x 2 + x 2 + = 2
·
dx dx dx dx dz dz dx

d3 y d2 y d2 y d d2 y
 
dy 1
x2 3 + 2x 2 + x 2 + = 2
·
dx dx dx dx dz dz x
3 2 2 d d2 y
 
3d y 2d y 2d y dy
x + 2x +x +x = ·
dx3 dx2 dx2 dx dz dz 2
3 2 d d2 y
 
3d y 2d y dy
x + 3x +x =
dx3 dx2 dx dz dz 2

d3 y 2
2d y dy
x3 = D 3
y − 3x −x
dx3 dx2 dx

d3 y
x3 = D3 y − 3x2 D2 y − xDy
dx3

= D3 y − 3(D − 1)y − Dy

= (D3 − 3D)y

= (D3 − 3D2 + 2D)y

3
 (D3 − 3D2 + 2D)y − D(D − 1)y =

(D3 − 4D2 + 3D)y = 0

Axilary equation is egiven by
m3 − 4m2 + 3m = 0

m(m2 − 4m + 3) = 0

⇒ m = 0, 1, 3
Solution of the equation:
y = c1 + c2 ez + c3 e3z
φ1 (x) = 1, φ2 (x) = x, φ3 (x) = x3
are the solutions of the equation (2):

d3 y d2 y
x 3
− 2 =0
dx dx
d3 d2
φ1 (x) = 1 ⇒ x · 3
(1) − 2 (1) = 0
dx dx

d3 d2
φ2 (x) = x ⇒ x · (x) − (x) = 0
dx3 dx2

d3 3 d2 3
φ3 (x) = x3 ⇒ x · (x ) − (x ) = 6x − 6x = 0
dx3 dx2
φ(x) = c1 φ1 (x) + c2 φ2 (x) + c3 φ3 (x)

⇒ φ(x) = c1 + c2 x + c3 x3
Superposition Principle:
Any linear combination of solutions of a linear homogeneous ODE is also a solution of that
equation.
This superposition principle does not hold for:
- Non-homogeneous ODEs - Non-linear ODEs
Examples:
1. y 00 − 3y 0 + 2y = e3x (Non-homogeneous)
2. y · y 00 − (y 0 )2 = 0 (Non-linear)

d3 y d2 y
x − =0 (1)
dx3 dx2
This implies
φ(x0 ) = 0 = φ0 (x0 ) = φ00 (x0 ).
Subjected to
y(1) = 0 = y 0 (1) = y 00 (1).

φ1 (x) = 1, φ2 (x) = x, φ3 (x) = x3 .

4
 y(x) = c1 + c2 x + c3 x3 (2)
General solution of (1). Put x = 1,

y(1) = c1 + c2 + c3 .

Differentiate equation (2) with respect to x:

y 0 (x) = c2 + 3c3 x2 .

Put x = 1,
y 0 (1) = c2 + 3c3 .
Again differentiate equation (2) with respect to x:

y 00 (x) = 6c3 x.

Put x = 1,
0 = y 00 (1) = 6c3 .
Thus,
y(0) = 0, for all x.
Given problem has trivial solution.
In general, if φ(x) is an n-th solution of an n-th ordered linear ODE

Ln y = 0,

subjected to initial conditions

φ(r−1) (x0 ) = 0, r = 1, 2, 3, . . . , n,

for some x0 ∈ I, then φ(x) will be the trivial solution.

Note: a superposition is not true for non-homogeneous linear ODE and also for non-linear
homogeneous ODE.

Example: Consider
y 00 − 3y 0 + 2y = e3x . (1)
Auxiliary equation:
m2 − 3m + 2 = 0.

(m − 2)(m − 1) = 0.
The roots are
m = 1 and m = 2.
The solutions are
ex and e2x .
General solution:
y = c1 ex + c2 e2x .
Particular solution:
1
yp = Ae3x , A = e3x .
2
1
y1 (x) = ex + e3x .
2

5
 1
φ2 (x) = ex + e2x . (2)
2

y(x) = 2φ1 (x) + φ2 (x)

is not a solution of equation (1).
From equation (2),

y 00 − 3y 0 + 2y = e3x .

φ00 (x) − 3φ0 (x) + 2φ(x) = e3x .

Example 2:
y · y 00 − (y 0 )2 = 0
is a nonlinear equation.
Given
y · y 00 − (y 0 )2 = 0, (1)

φ1 (x) = ex , φ2 (x) = e2x

are the solutions of equation (1).
But
φ1 (x) + φ2 (x)
is not a solution of equation (1).
Put
dy
= y 0 = u.
dx
Then
du du dy du
y 00 = u0 = = · =u .
dx dy dx dy
du
y·u − u2 = 0.
dy
du du
y − u = 0 or y = u.
dy dy
Z Z
du dy
= .
u y

log u = log y + log c.

u = cy.

dy
= cy.
dx
dy
or = c dx.
y

y = Aecx .

6
 The solution that superposition principle exists. Certainly! HereâĂŹs your dictated text
formatted precisely in LaTeX:
“ ‘latex
Note: let V be a vector space over the field F . The set of vectors v1 , v2 , v3 , . . . , vn ∈ V is
said to be linearly independent if

c1 v1 + c2 v2 + · · · + cn vn = 0,

for scalars c1 , c2 , . . . , cn ∈ F implies

c1 = c2 = · · · = cn = 0.

Linear independent solutions: Let f1 (x), f2 (x), f3 (x), . . . , fn (x) for i = 1, 2, 3, . . . , n be n

functions defined on the interval I. The functions fi (x) are linearly independent over I if

c1 f1 (x) + c2 f2 (x) + · · · + cn fn (x) = 0

only when
c1 = c2 = · · · = cn = 0.

If fi (x) are n solutions of

Ln y = 0,
then the set of all fi (x) for i = 1 to n is a set of linear independent solutions of

Ln y = 0,

that is the fundamental set of

Ln y = 0.
Example 1, x and 2x are not linearly independent.

c1 x + c2 (2x) = 0.
For c1 = −2, c2 = 1.

Example 2: x and x2 are linearly independent on the closed interval [0, 1].
Suppose
c1 x + c2 x2 = 0.
Differentiating,
c1 + 2c2 x = 0.
At some point,
2c2 = 0.
Hence,
c2 = 0 and c1 = 0.
Example 3. sin x, − sin x, 3 sin x are linearly dependent.

c1 sin x + c2 (− sin x) + c3 (3 sin x) = 0,

for c1 = −2, c2 = 1, c3 = 1.
Example 4: sin x and cos x are linearly independent.
C1 sin x + C2 cos x = 0.

7
Therefore, when C1 = C2 = 0.
sin x and cos x are two solutions of
d2 y
+ y = 0.
dx2
Label this equation as (1). Since the equation is linear and homogeneous, by the superposition
principle, C1 sin x + C2 cos x is also a solution of the given differential equation.
Any solution of√ equation (1) can be written as a linear combination of sin x and cos x.
sin x + π6 = 23 sin x + 12 cos x.

Set of all sin x, cos x is a fundamental set of equation (1).

In general, any linear combination of f1 (x), f2 (x), . . . , fn (x) is a solution of Ln y = 0.
Conversely, any solution of Ln y = 0 can be written as a linear combination of f1 to fn , that is,
φ(x) = c1 f1 + c2 f2 + . . . + cn fn
for some choice of cn ’s.
Definition: A set of all f1 (x), f2 (x), . . . , fn (x) of n solutions of Ln [y] = 0 is called a funda-
mental set of Ln [y] = 0 if the functions f1 (x), f2 (x), . . . , fn (x) are linearly independent.
Let f1 , f2 , . . . , fn be real differentiable functions, each with an (n − 1)th derivative on a real
interval a ≤ x ≤ b, then the determinant

f1 (x) f2 (x) ··· fn (x)

f10 (x) f20 (x) ··· fn0 (x)
.. .. .. ..
. . . .
(n−1) (n−1) (n−1)
f1 (x) f2 (x) · · · fn (x)
is called the Wronskian. The Wronskian of n functions f1 , f2 , . . . , fn is itself a real function
of x defined on the interval I or [a, b]. It is denoted by W [f1 , f2 , . . . , fn ](x) or W [f1 (x), f2 (x), . . . , fn (x)]
or simply W (x).
Example: Consider the set {sin x, cos x} of two solutions of
d2 y
+ y = 0.
dx2
sin x and cos x are linearly independent.
sin x cos x
W (sin x, cos x) =
cos x − sin x
= − sin2 x − cos2 x = −1 6= 0.
Example 2
Consider a set of solution of equation
d3 y d2 y dy
3
− 2 · 2
− + 2y = 0. (1)
dx dx dx
solution:

m3 − 2m2 − m2 − m − 2 = 0
m2 − m − 2 = 0
m2 − 2m − 2 − m − 2 = 0
m−m−2−1−m−2=0
m − 2, m−1=0
Roots are:
m = 1, −1, 2
The functions ex , e−x , e2x
are solutions of equation (1).
x −x 2x
c1 e + c2 e + c3 e = 0, holds only when c1 = c2 = c3 = 0.
The functions ex , e−x , e2x are linearly independent.

8
Examples
1) x and 2x are not L.I.
c1 x + c2 (2x) = 0
for c1 = −2, c2 = 1.

2) x and x2 are L.I. on [0, 1].

Suppose, c1 x + c2 x2 = 0. diff
c1 + 2c2 x = 0.
2c2 = 0
c1 = 0
therefore, c1 = 0, c2 = 0

3) sin x, − sin x, 3 sin x are linearly dependent.

c1 sin x + c2 (− sin x) + c3 3 sin x = 0.

for c1 = 2, c2 = c3 = 1.

4) sin x and cos x are L.I.

c1 sin x + c2 cos x = 0.
Only when c1 = c2 = 0. sin x and cos x are two solutions of

d2 y
+ y = 0. (1)
dx2
Since the equation is linear and homogeneous, by the superposition principle, c1 sin x + c2 cos x
is also a solution of the given differential equation. Any solution of eq. (1) can be written as a
linear combination of sin x and cos x.
√
3 1
sin(x + π/6) = sin x + cos x
2 2
{sin x, cos x} is a fundamental set of solutions of (1).

In general, any linear combination now f1 (x), f2 (x), ..., fn (X) is a solution of Ln y = 0 dif-
ferential equation. Conversely, any solution of Ln y = 0 can be written as a linear combination
of f1 to fn , i.e., f (x) = C1 f1 + C2 f2 + · · · + Cn fn for some choice of Cn0 s.
Defn : A set {f1 (x), f2 (x), . . . , fn (x)} of n solutions of Ln y = 0 is called a fundamental set
of ny = 0, if the functions f1 (x), . . . , fn (x) are linearly independent.
Wronskian:
Let f1 , f2 , . . . , fn be a n real differential functions, each of which has (n − 1)st derivative on
a real interval a ≤ x ≤ b. Then the determinant

f1 (x) f2 (x) ... fn (x)

f10 (x) f20 (x) ... fn0 (x)
.. .. .. ..
. . . .
(n−1) (n−1) (n−1)
f1 (x) f2 (x) . . . fn (x)
is called the Wronskian. If W (f1 , f2 , . . . , fn ) 6= 0, then f1 , f2 , . . . , fn are linearly independent.
The Wronskian of n functions f1 , f2 , . . . , fn is itself a real function of x defined on the interval
[a, b]. It is denoted by W [f1 (x), f2 (x), . . . , fn (x)] or simply W [x].

9
 1. Consider {sin x, cos x} of two solutions of:

d2 y
+ y = 0.
dx2

sin x, cos x are L.I. (linearly independent).

sin x cos x
W (sin x, cos x) = = − sin2 x − cos2 x = −1 6= 0.
cos x − sin x

Consider a set of solutions of the equation:

d3 y d2 y dy
3
− 2 2
− + 2y = 0 (1)
dx dx dx
Solution:
Solve the characteristic equation:

m3 − 2m2 − m + 2 = 0

Using synthetic division:

1 −2 −1 2
1 1 −1 1
1 −1 0 0
Then:
m2 − m − 2 = 0

m2 − 3m + m − 2 = 0 ⇒ m(m − 3) + 1(m − 2) = 0

(m − 2)(m + 1) = 0
Roots are m = 1, −1, 2.
The functions ex , e−x , e2x are solutions of (1).
c1 e + c2 e−x + c3 e2x = 0 holds only when c1 = c2 = c3 = 0.
x

The functions ex , e−x , e2x are linearly independent.

ex e−x e2x
W [ex , e−x , e2x ] = ex −e−x 2e2x
ex e−x 4e2x

= ex −4e−x − 2ex − (−e−x ) 4ex − 2e3x + e2x (1 + 1)

= −6ex e−x − e−x (2e3x ) + 2e2x

= −8e2x 6= 0
Above expression shows that the solutions are linearly independent if the Wronskian of the
solutions is not equal to zero.

Theorem
The set {φj (x) | j = 1, 2, . . . , n} forms a fundamental set of solutions of the equation Ln y = 0
on I if and only if the Wronskian W (φ1 , φ2 , . . . , φn ) 6= 0 on I.

10
Proof
Let {φj (x) | j = 1, 2, . . . , n} be a fundamental set of solutions and

Ln y = 0. (1)

Let c1 φ1 (x) + c2 φ2 (x) + · · · + cn φn (x) = 0 for some x0 ∈ I. Suppose W (φj (x)) = 0, j =

1, 2, . . . , n − 1. Consider the system of equations:

c1 φ1 (x0 ) + c2 φ2 (x0 ) + · · · + cn φn (x0 ) = 0

c1 φ01 (x0 ) + c2 φ02 (x0 ) + · · · + cn φ0n (x0 ) = 0
..
.
(n−1) (n−1)
c1 φ1 (x0 ) + c2 φ2 (x0 ) + · · · + cn φ(n−1)
n (x0 ) = 0, (2)

Since the Wronskian is zero, the system (2), has a non-trivial solution for the unknowns
c1 , c2 , . . . , cn . Now, becuase φ1 (x), φ2 (x), · · · , φn (x) are the solutions of the equation (1). the
contribution let

φ(x) = c1 φ1 (x) + c2 φ2 (x) + · · · + cn φn (x) 6= 0, (3)

is also a solution of Ln y = 0. for all x ∈ I

From (2) and (3):

φ(x0 ) = 0 = φ0 (x0 ) = φ00 (x0 ) = · · · = φ(n−1) (x0 ),

i.e., φ(x) = 0 for every x ∈ I.

Thus,

c1 φ1 (x) + c2 φ2 (x) + · · · + cn φn (x) = 0 (4)

for every x ∈ I. The c1 , c2 , . . . , cn are not all zero since the solution for the unknowns c1 , c2 , . . . , cn
is non-trivial.
Form equation (4), suppose φ1 (x), φ2 (x) are linearly dependent on I.
This is a contradiction (1) shows that our assumption is wrong. Therefore Wronskian
W {φi (x) : i = 1, 2, . . . , n} =
6 0 for any x ∈ I.
Conversely, Suppose Wronskian W 0 (φj (x) : i = 1, 2, . . . , n) 6= 0, ∀x ∈ I.
Suppose the functions φj (x) are linearly dependent, then ∃ scalars C1 , C2 , . . . , Cn not all
zero such that C1 φ1 + C2 φ2 + · · · + Cn φn = 0.
Differentiating with respect to x, (n − 1) times we get,

0 0 0
C1 φ1 + C2 φ2 + · · · + Cn φn = 0
..
.
(n−1) (n−1)
C1 φ1 + C2 φ 2 + · · · + Cn φ(n−1)
n =0 (5)

Since W {φj (x), j = 1 . . . n} =

6 0.
The system φ has a trivial solution for the unknowns C1 , C2 , . . . , Cn , i.e., C1 = C2 = · · · =
Cn = 0.
This contradiction shows that the functions φj (x) are linearly independent, or

{φj (x) : i = 1, 2, . . . , n}
is a fundamental set of Ln y = 0.

11
Example
1. If the Wronskian of φ1 (x) and φ2 (x) is 3e4x and if of φ2 (x) = e2x then find φ2 (x).

φ1 (x) φ2 (x)
W (φ1 (x), φ2 (x)) = = 3e4x
φ01 (x) φ02 (x)
and φ2 (x) = e2x .

φ1 (x)φ02 (x) − φ01 (x)φ2 (x) = 3e4x

1.
e2x φ02 (x) − 2e2x φ2 (x) = 3e4x
e2x (φ02 (x) − 2φ2 (x)) = 3e4x
φ02 (x) − 2φ2 (x) = 3e2x
y 0 − 2y = 3e2x
m − 2 = 0, n = 2
C.F. = Ce2x

1
P.I. = 3e2x = 3x e3x
D−2

φ2 (x) = C e2x + 3xe2x = (C + 3x)e2x

2) Find the Wronskian of independent solutions of the homogeneous differential equation on

(−∞, ∞) whose roots are (1, ±1, ±i).
3) Find the Wronskian of independent solutions of

y5 − y4 − y0 + y = 0

Sol:

y 5 − y 4 − y 0 + y = 0 ⇒ m5 − m4 − m + 1 = 0 (1)
1 −1 0 0 −1 1
1 0 1 0 0 0 −1
1 0 0 0 −1 0
For m = ±1:
m4 − 1 = 0
(m2 )2 − (i2 )2 = 0
(m2 − 12 ) (m2 + 12 ) = 0
m = ±1 m = ±i
The roots are 1, ±1, ±i.
The solutions are ex , xex , e−x , cos x, sin x.
To verify that the solutions are linearly independent, we compute the Wronskian W (ex , xex , e−x , cos x, sin x).
5. Show that the functions f1 (x) = x2 and f2 (x) = x|x|, x ∈ R, x 6= 0, are linearly
independent.
Proof

12
 Suppose α1 f1 (x) + α2 f2 (x) = 0 for all x.
Then

α1 x2 + α2 x|x| = 0, (1)

for x ∈ R, x 6= 0.
If x > 0, then:
x|x| = x2 ⇒ α1 x2 + α2 x2 = 0.
Hence, α1 + α2 = 0 (2).
If x < 0, then x|x| = −x2 . Thus:

α1 x2 − α2 x2 = 0 ⇒ α1 − α2 = 0 (3)

since x 6= 0.
Solving equations (2) and (3):
α1 + α2 = 0,
α1 − α2 = 0.
by adding and substracting, we get α1 = 0 and α2 = 0.
Thus, the functions x2 and x|x| are linearly independent. The Wronskian of x2 and x|x| is:

x2 x|x|
W (x2 , x|x|) = = x2 (2|x|) − x(2x|x|) = 2x3 − 2x3 = 0
2x 2|x|
Note: Linear independence is verified by a non-zero Wronskian on I.

Liouville’s Formula or Abel’s Identity

Let {φi (x); i = 1, 2, . . . , n} be n solutions of Ln y = 0 on some interval I.
Then
R a1 (t)
dt
W (φi (x) : j = 1, . . . , n) = W (φi (x0 ) : j = 1, . . . , n)e a0 (t) (for any x0 , x ∈ I)

Proof
Given {φj (x)}j=1,2,...,n as solutions of Ln y = 0:

dn y dn−1 y dy
a0 n
+ a1 n−1
+ · · · + an−1 + an y = 0
dx dx dx
on I, and we assume that a0 (x) 6= 0 for x ∈ I.
Let us consider the case for n = 3

a0 (x) y (3) (x) + a1 (x) y 00 (x) + a2 (x) y 0 (x) + a3 (x) y(x) = 0

Since φ1 (x), φ2 (x), φ3 (x) are the solutions of equation (1) we have
(3)
a0 (x) φ1 (x) + a1 (x) φ001 (x) + a2 (x) φ01 (x) + a3 (x) φ1 (x) = 0

(3)
a0 (x) φ2 (x) + a1 (x) φ002 (x) + a2 (x) φ02 (x) + a3 (x) φ2 (x) = 0

(3)
a0 (x) φ3 (x) + a1 (x) φ003 (x) + a2 (x) φ03 (x) + a3 (x) φ3 (x) = 0

13
 ⇒

(3) a1 (x) φ001 (x) + a2 (x) φ01 (x) + a3 (x) φ1 (x)

φ1 (x) = −
a0 (x)

(3) a1 (x) φ002 (x) + a2 (x) φ02 (x) + a3 (x) φ2 (x)

φ2 (x) = −
a0 (x)

(3) a1 (x) φ003 (x) + a2 (x) φ03 (x) + a3 (x) φ3 (x)

φ3 (x) = −
a0 (x)
Divide by W (x) = W [φ1 (x), φ2 (x), φ3 (x)] so the wronskian W (x) is

φ1 (x) φ2 (x) φ3 (x)

W (x) = φ01 (x) φ02 (x) φ03 (x)
φ001 (x) φ002 (x) φ003 (x)
Differentiate with respect to x

φ01 (x) φ02 (x) φ03 (x) φ1 (x) φ2 (x) φ3 (x) φ1 (x) φ2 (x) φ3 (x)
dW (x) 0 (x) 0 (x)
0 0 0 00 00 00
= φ1 (x) φ2 (x) φ3 (x) + φ1 (x) φ2 (x) φ3 (x) + 1 φ φ 2 φ03 (x)
dx (3) (3) (3)
φ001 (x) φ002 (x) φ003 (x) φ001 (x) φ002 (x) φ003 (x) φ1 (x) φ2 (x) φ3 (x)

φ1 (x) φ2 (x) φ3 (x)

dW (x)
= φ01 (x) φ02 (x) φ03 (x)
dx a1 00 a1 00 a1 00
− a0 φ1 − aa02 φ1 − a3
a0 φ1 − a0 φ2 − aa20 φ2 − a3
a0 φ2 − a0 φ3 − aa02 φ3 − a3
a0 φ3

φ (x) φ2 (x) φ3 (x) φ (x) φ2 (x) φ3 (x) φ (x) φ2 (x) φ3 (x)

dW (x) a1 10 0 0 a2 10 0 0 a3 10
=− φ (x) φ2 (x) φ3 (x) − φ (x) φ2 (x) φ3 (x) − φ (x) φ02 (x) φ03 (x)
dx a0 001 00 00 a0 001 00 00 a0 001
φ1 (x) φ2 (x) φ3 (x) φ1 (x) φ2 (x) φ3 (x) φ1 (x) φ002 (x) φ003 (x)

dW (x) −a1 (x)

⇒ = · W (x) (4)
dx a0 (x)
(Wronskian formula)
This can be written as
Separating the variables and integrating between x0 to x, we have

−a1 (t)
Z Z
dW (x)
= dt
W (x) a0 (t)
Z
x a1 (t)
log W (t) x0 } = − dt
{ a0 (t)
Z
a1 (t)
⇒ log W (x) − log W (x0 ) = − dt
a0 (t)
W (x) −
R a1 (t)
dt
⇒ =e a0 (t)
W (x0 )
R a1 (t)
− dt
⇒ W (x) = W (x0 ) e a0 (t) (5)
Eqn (4) and (5) holds for finite n.

14
 In general,
R a1 (t)
− dt
W {φj (x), j = 1, 2, . . . n} = W {φj (x0 ), j := 1, 2, . . . n}e a0 (t) (6)

.
⇒
Using (6) we derive . . . if a1 (t) = 0 then . . .

W {φj (x) : j = 1, 2, . . . , n} = W {φj (x0 ) : j = 1, 2, . . . , n}

The Wronskian is invariant throughout the interval if a1 (x) = 0.

W {φj (x) : j = 1, 2, . . . , n} = 0in eq. (6)

W {φj (x) : j = 1, 2, . . . , n} = 0andand vice versa.

This shows that if the Wronskian vanishes at some point x ∈ I, then it vanishes throughout
the interval.

Example
Find the Wronskian solution of:

y 000 + y 00 − y 0 + y = 0

2x2 y 00 + 7xy 0 + 3y = 0
Verify Hermite’s formula:

x2 y 00 − 7xy 0 + 15y = 0 (1)

Put x = ez or z = log x:

dy dy dx 1 dy
= = (or)
dx dz dz x dz
dy dy d
x = = Dy D=
dx dz dz

d2 y dy d2 y
x2 + x = x = Dy
dx2 dx dz 2
d2
D2 = 2
dz

d2 y
x2 = D(D − 1)y
dx2
Substitute this in eq. (1):

xD2 − Dy − 7Dy + 15y = 0 (2)

D2 y = 8Dy + 45y = 0

m2 − 8m + 15 = 0

15
 m2 − 5m − 3m + 15 = 0

m(m − 5) − 3(m − 5) = 0
(m − 5)(m − 3) = 0
The roots are m = 3, 5.
The solutions of (2) are e3x and e5x .
The solutions of (1) are x3 and x5 .
Here,
x3 x5
W (x3 , x5 ) =
3x2 5x4

= 5x7 − 3x7 = 2x7

dW (x)
= 14x6
dx
Then a0 (x) = x2 , a1 (x) = −7x, a2 (x) = 15.
Consider,
a1 (x) −7x × x7
− x(x) = = 14x6
a0 (x) x2

d log(x)
2. Verify Liouville’s formula for

x3 y 00 − 3x2 y 0 + 6xy 0 − 6y = 0 · · · (1)

⇒ put x = ez , α = z, z = log x
dy dy dz 1 dy
= =
dx dz dx x dz
dy dy d
x= = Dy, D =
dx dz dz

d2 y d2 y
 
dy d dy
x2 2 + x = = D2 y, D2 = 2
dx dx dz dz dz

d2 y
x2
dx2

d2 y
x2 = D2 y − Dy
dx2

d2 y
x2= D(D − 1)y
dx2

d2 y d3 y 1 d2 dy dz
 
2x 2 + x2 3 = 2 2
dx dx x dz dz dx

d3 y 2 2
2d d y d3 y
x3 + 2x = = D3 y
dx3 dx2 dz 3
(D3 − D2 − 2D2 + 2D − 3D2 + 3D + 6D − 6)y = 0

16
 (D3 − 6D2 + 11D − 6)y = 0
1 −6 11 −6
0 1 −5 6
1 −5 6 0

m2 − 5m + 6 = 0
m2 − 3m − 2m + 6 = 0
m(m − 3) − 2(m − 3) = 0
m = 1, 2, 3

The solution of eq (2) is ez , e2z , e3z .

The solution of equation (1) are x, x2 and x3 .

x x 2 x3
2 3
These work: W (x, x , x ) = 1 2x 3x2
0 2 6x

= x(12x2 − 6x2 ) − x2 (6x − 0) + x3 (2 − 0)

= 6x3 − 6x3 + 2x3 = 2x3

dW
(x) = 6x2
dx
Thus, a0 (x) = x3 , a1 (x) = −3x2

a2 (x) = 6x,
−a1 (x) 3x2 dW (x)
· W (x) = + 3 · 2x3 = 6 · x2 =
a0 (x) x dx

Theorem
Let {φj (x) : j = 1 to n} be a fundamental set of Ln y = 0 on I. Then if {ψj (x) : j = 1 to n}
is another fundamental set of Ln y = 0 on I, there exists a non-singular constant matrix C of
order n such that ψ = Cφ, where
 T
ψ = ψ1 (x) ψ2 (x) · · · ψn (x)
 T
φ = φ1 (x) φ2 (x) · · · φn (x)
Further, if W {ψj (x) : j = 1 to n} = |c|W {φj (x) : j = 1 to n}.

Proof
Given {φj (x) : j = 1 · · · n} is a fundamental set of Ln y = 0 on (1), linear homogenous ODE.
Then, ψ1 (x) = C11 φ1 (x) + C12 φ2 (x) + · · · + C1n φn (x) (2)
where C11 , C12 , · · · , C1n are constants not all zero.
The function

ψ2 (x) = C21 φ1 (x) + C22 φ2 (x) + · · · + C2n φn (x) (3)

17
where C21 , C22 , · · · are constants not all zero.

ψn (x) = Cn1 φ1 (x) + Cn2 φ2 (x) + · · · + Cnn φn (x) (4)

are also solutions of (1).
System of equations (2), (3), (4) can be written in matrix form as:
    
ψ1 (x) C11 C12 · · · C1n φ1 (x)
 ψ2 (x)   C21 C22 · · · C2n   φ2 (x) 
 ..  =  .. (5)
    
.. .. ..   .. 
 .   . . . .   . 
ψn (x) Cn1 Cn2 · · · Cnn φn (x)

i.e., ψ = Cφ (6)
where  T
ψ = ψ1 · · · ψn (7)
,
 T
φ = φ1 · · · φn (8)
and C = [Cij ]1≤i,j≤n , C is a constant matrix. From the previous discussion, it can be proven
that
φ = Bψ (9)
where φ and ψ are in (7) and (8), if B is a constant matrix.
From (6) and (9),
φ = BCφ ⇒ (I − BC)φ = 0.
Since φ 6= 0,
I − BC = 0 ⇒ BC = I (8)
It means that B = C −1 .
Inverse of C exists or, C is non-singular.
Conversely, suppose there is a non-singular matrix C such that ψ = Cφ.
We have to show that {ψj (x) : j = 1ton} is a fundamental set of Ln y = 0 on I.
Since ψ = Cφ, where ψ and φ are in (7) and (8):

ψ1 (x) = C11 φ1 (x) + · · · + C1n φn (x),

ψ2 (x) = C21 φ1 (x) + · · · + C2n φn (x),
ψn (x) = Cn1 φ1 (x) + · · · + Cnn φn (x) (10)
So by the superposition principle, ψ1 (x), ψ2 (x), . . . , ψn (x) are solutions of Ln y = 0 on I.
Now we prove the solutions are linearly independent. Let us assume that

α1 ψ1 (x) + α2 ψ2 (x) + · · · + αn ψn (x) = 0 (11)

From (10) and (11), we have

α1 (C11 φ1 (x)+· · ·+C1n φn (x))+α2 (C21 φ1 (x)+· · ·+C2n φn (x))+· · ·+αn (Cn1 φ1 (x)+· · ·+Cnn φn (x)) = 0

(α1 C11 + α2 C21 + · · · + αn Cn1 )φ1 (x)+

(α1 C12 + α2 C22 + · · · + αn Cn2 )φ2 (x)+
· · · + (α1 C1n + α2 C2n + · · · + αn Cnn )φn (x) = 0
Since φ1 (x), φ2 (x), . . . , φn (x) are linearly independent, their coefficients

18
 α1 C11 + α2 C21 + · · · + αn Cn1 = 0,
α1 C12 + α2 C22 + · · · + αn Cn2 = 0,
··· ,
α1 C1n + α2 C2n + · · · + αn Cnn = 0.
In matrix form:
  
C11 C21 ··· Cn1 α1
 C12 C22 ··· Cn2   α2 
..   ..  = 0
  
 .. .. ..
 . . . .  . 
C1n C2n · · · Cnn αn
By our assumption, the matrix C = (Cij )1≤i,j≤n is a non-singular matrix. Therefore, α1 =
α2 = · · · = αn = 0.
This shows that the solutions ψ1 (x), ψ2 (x), . . . , ψn (x) are linearly independent on I and
hence form a fundamental set of solutions for Ln y = 0.
Normalized Differential Equations:
Where differential equation in which the coefficient of the highest order derivative is unity is
called a normalized differential equation.
Example: Find the normalized D.E. having {1, x, x3 } as its fundamental set.
Solution: The required normalized D.E. is given by:

W [1, x, x3 , y]
= 0 — (1)
W [1, x, x3 ]

We first compute W [1, x, x3 , y].

1 6x
= = 6xy 000 − 6y 00
6 y 00

w(1, x, x2 , y) = 6xy 000 − 6y 00

we next compute w[1, x, x3 ]
1 x x3
= 0 1 3x2 = x · 6
0 0 6x
6xy 00 − 6y 00 xy 00 − y 00
normalized D.E is = = 0 =⇒ =0
6x x
2. find normalized D.E which has {x, xex } as its fundamental set
The required normalized D.E is given by
w[x, xex , y]
= 0 (1)
w[x, xex ]

x xex
w[x, xex , y] = = x2 e2x − xe2x = xex (x + 1)
xe xe2x
x

we have compute w(x, xex , y) = 6xy 000 − 6y 00

we next compute w(x, xex , y)

x xex y
= xe x xe + ex y 0
2x

0 xe2x + 2ex y 00

19
 = x x(ex (x + 1))y 00 − y 0 (ex (x + 2)) − xex (y 00 ) + y(ex (x + 2))

x xex
w[x, xex ] = = x(ex (x + 1)) − xex
1 xex + ex

normalized D.E:
x2 ex y 00
− x(x + 2)ex y 0 + (x + 2)xex y
=0
xex
   
00 2 0 1 2
=⇒ y + 1 + y + + y=0
x x x
If φ1 (x) is a solution of

y 00 (x) + a1 (x)y 0 (x) + a2 (x)y(x) = 0 on I

φ2 (x) = φ1 (x) is another solution of the equation provided y 0 (x) is a solution to the equation
2
(φ1 y) + φ1 (x)φ1 x = 0 and also we have that solutions φ1 (x), φ2 (x) are L.I.
Solution: Suppose φ1 (x) is a solution of:

y 00 (x) + a1 (x)y 0 (x) + a2 (x)y(x) = 0 on I

Then φ001 (x) + a1 (x)φ01 (x) + a2 (x)φ1 (x) = 0
Now let φ2 (x) = φ1 (x)f (x) as the another solution of (1)

φ2 (x) = φ001 (x) + a1 (x)φ01 (x) + a2 (x)φ1 (x) = 0

(φ1 f )00 + a1 (φ1 f )0 + a2 (φ1 f ) = 0

differentiate with respect to x

(φ1 f 0 + φ1 f )0 + a1 (φ1 f 0 + f φ01 ) + a2 (φ1 f ) = 0

multiply by φ1 (x)

[φ1 (x)]2 f 00 (x) + [2φ01 (x)φ(x) + a1 (x)φ1 (x)2 ]f 0 (x) = 0

⇒ (φ1 (x))2 f 00 (x) + ([2φ01 (x)φ(x))f 0 (x) + a1 (x)φ1 (x)2 f 0 (x) = 0

⇒ (φ21 f 0 )0 + a1 φ21 f 0 = 0 - (3) (9)

⇒ (φ21 y)0 + a1 (φ1 y) = 0 - (4) (10)

From (3), we have:

[φ21 (x)f 0 (x)]0
Z
−a1 (x) =
φ21 (x)f 0 (x)
Integrating on both sides:
Z
log(φ21 (x)f 0 (x)) = −a1 (x) dx

R
φ21 (x)f 0 (x) = e −a1 (x) dx

20
 Z Z
0 1 R
−a1 (x) dx
f (x) dx = e
φ21 (x)
On integration: Z
1 R
−a1 (x) dx
f (x) = e dx
φ21 (x)
Thus we have: Z
1 R
−a1 (x) dx
φ2 (x) = φ1 (x)f (x) = φ1 (x) e dx
φ21 (x)
In another solution of equation (1). Now, to show that these two solutions are linearly
independent. We consider:
c1 φ1 (x) + c2 φ2 (x) = 0
i.e., Z
1 R
−a1 (x) dx
c1 φ1 (x) + c2 φ1 (x) e dx = 0
φ21 (x)
Since φ1 (x) 6= 0, we have:
Z
1 R
−a1 (x) dx
c1 + c2 e dx = 0
φ21 (x)

But this holds only when c1 = c2 = 0. In other words, the solutions φ1 (x) and φ2 (x) are
linearly independent.
Eg:

x2 y 00 − xy 0 + y = 0 (x 6= 0)
Find another solution & the Wronskian of the solution.
Solution: The given equation is x2 y 00 − xy 0 + y = 0. Divide the equation by x2 :
1 0 y
y 00 − y + 2 =0
x x
Here, a1 (x) = − x1 .
Given that φ1 (x) is the solution of the equation (1) = x,
Another solution is given by:
Z
1 R
φ2 (x) = φ1 (x) 2 e− a1 (x) dx dx
φ1 (x)
Z Z
1 − 1
=x e − dx dx
x2 x
Z
1 log x
=x e dx
x2
Z
1
=x x dx
x2
Z
1
=x dx
x

= x log x

φ2 (x) = x log x

21
 Wronskian:

x x log x
W {x, x log x} = 1
1 x x + log x

= x(1 + log x) − x log x

=x
Given that x 6= 0 =⇒ x log x 6= 0
The solutions are linearly independent.
Example:
If sin x is one solution of y 00 + 2 tan x · y 0 − y = 0, find the other solution and also find the
Wronskian of these solutions.
Given equation y 00 + 2 tan x · y 0 − y = 0 (1) is in a normalized form.
Given φ1 (x) = sin x is one of (1).
The another solution of (1) is given by
Z
1 R
− a1 (x) dx
φ2 (x) = φ1 (x) e dx
φ21 (x)
Z
1 R
− 2 tan x dx
= sin x e dx
sin2 x
Z
1
= sin x e−2 log(cos x) dx
sin2 x

cos2 x
Z
= sin x dx
sin2 x

1 − sin2 x
Z
= sin x dx
sin2 x
Z
csc2 x − 1 dx


= sin x

= sin x (− cot x − x)

φ2 (x) = − cos x − x sin x

The Wronskian of the two solutions sin x and − cos x − x sin x is

sin x − cos x −x sin x

cos x sin x − cos x − x sin x

= −x sin x cos x + cos2 x = cos2 x

Theorem:
Let y1 (x), y2 (x), . . . , yn (x) be n solutions of

Ln y = 0 on I

satisfying the initial conditions

Aij (i−1) (x0 ) = [Aij ], (i.e i, j = 1, 2, ≤ n).

22
Aij are conditions x0 then the neccessary and sufficient condition for yj to form a fundamental
set if |Aij | =
6 0.
proof: Let {yi (x) : i = 1, 2, . . . , n} be a fundamental set of

Ln y = 0 (1)
(i−1)
on I. given yj (x0 ) = Aij x0 ∈ Iall 1 ≤ i, j ≤ n.
Suppose |Aij | =
6 0.
Consider

c1 A11 + c2 A12 + . . . + cn A1n = 0

c1 A21 + c2 A22 + . . . + cn A2n = 0
..
.
c1 An1 + c2 An2 + . . . + cn Ann = 0 (1)

In matrix form:   
A11 . . . A1n c1
 A21 . . . A2n   c2 
..   ..  = 0
  
 ..
 . ... .  . 
An1 . . . Ann cn
6 0, c = (c1 , c2 , . . . , cn )T has a non-trivial solution if not all ci are zero.
Since |Aij | =
Consider y(x) = c1 y1 (x) + c2 y2 (x) + . . . + cn yn (x). By the superposition principle, y(x) is
also a solution of Ln y = 0.
Now, by the initial conditions and from equation (1),

y(x0 ) = 0, y 0 (x0 ) = 0, ..., y (n−1) (x0 ) = 0

y(x) = 0 ∀x ∈ I
i.e.
c1 y1 (x) + c2 y2 (x) + . . . + cn yn (x) = 0 ∀x ∈ I
where not all ci are zero.
y1 (x) and y2 (x) are L.D then it is a contradiction.
Thus, the contradiction proves that our assumption |Aij | = 0 is wrong. Therefore, |A| =
6 0.
Conversely, suppose that |Aij | =
6 0. Consider

y(x) = c1 y1 (x) + c2 y2 (x) + · · · + cn yn (x).

Differentiating with respect to (x) successively (i − 1) times for i = 1, 2, . . . , n and putting

x = x0 , we get a system of equations when they are in the form

c1 A11 + c2 A12 + · · · + cn A1n = 0,

c1 A21 + c2 A22 + · · · + cn A2n = 0, (1)
c1 An1 + c2 An2 + · · · + cn Ann = 0.

i.e.,     
A11 · · · A1n c1 0
 .. .. ..   ..  =  .. 
 . . .   .  .
An1 . . . Ann cn 0
Since |Aij | =
6 0, the system of equations en has a trivial solution, i.e., c1 = c2 = · · · = cn = 0.
The functions y1 (x) and y2 (x) form a fundamental set.

23
Adjoint Differential Equation:
Consider linear homogeneous n-th order D.E:

dn y dn−1 y
a0 (x) + a1 (x) + · · · + an (x)y = 0.
dxn dxn−1
or
n
X dn−r y
ar (x) (1)
dxn−r
r=0

where a0 (x) 6= 0 on I. If ar (x) is continuous on I and ar (x) has continuous derivatives of order
n − r, then differential equation
n
X dn−r
(−1)n−r n−r (ar (x)y) = 0 (2)
dx
r=1

is called the adjoint D.E to equation (1).

dn n−1 d
n−1 d
(−1n ) n
(a0 (x)y) + (−1) n−1
(a1 (x)y) + · · · + (−1) (an−1 (x)y) + an (x)y = 0.
dx dx dx
Consider second order D.E. (differential equation) linear homogeneous ordinary:

a0 (x)y 00 + a1 (x)y 0 + a2 (x)y = 0 (3)

The adjoint D.E. to (3) is, here n = 2:

d2
 
2 d
(−1) (a0 (x)y) + (−1) a1 (x) y + a2 (x)y = 0
dx2 dx

d2 y

dy dy
a0 (x) 2 + 2a00 (x) + a000 (x)y(x) − a01 (x)y(x) − a1 (x) + a2 (x)y = 0}
dx dx dx

d2 y 0

dy 00 0


a0 (x) + 2a 0 (x) − a 1 (x) + a0 (x) − a 1 (x) + a 2 (x) y=0 (3)
dx2 dx
NOTE:

• The adjoint D.E. is also a linear homogeneous D.E.

• Adjoint D.E. to Lny = 0 is denoted by Ln∗ y = 0 which is also a D.E. of order n.

• The adjoint eq. to adjoint eq. of Ln y = 0 is the original eq. Ln y = 0.

• Ln∗ y = 0 implies same as Ln y = 0.

• The adjoint of D.E. Ln∗ 6= Ln .

– If Ln = Ln ∗ , if general.
But if Ln = Ln ∗ they can say that Ln is self adjoint.
d 2 d
• If L2 = a0 (x) dx 2 + a1 (x) dx + a2 (x), then

d2
d
L∗2 = a0 (x) + 2a00 (x) − a1 (x) + a000 (x) − a01 (x) + a2 (x)


dx 2 dx

24
 • L∗∗
n = Ln
[Ln 0 + Ln 2 ] = Ln + Ln 2∗
(k2n)∗ = (k2n)∗

Theorem:
A necessary and sufficient condition for second order L.D.E:

d2 y dy
a0 (x) 2
+ a1 (x) + a2 (x)y = 0
dx dx
to be self-adjoint if

a00 (x) = a1 (x) on a ≤ x ≤ 0

Given

d2 y dy
a0 (x) 2
+ a1 (x) + a2 (x)y = 0 (1)
dx dx
The adjoint of D.E is:

L∗2 y = a0 (x)y 00 + (2a0 (x)y 00 − a1 (x)y 0 + a1 (x))y 0 + (a000 (x)y − a1 (x) + a2 (x))y = 0 (2)

If a10 (x) = a1 (x) then (2) takes the form:

a0 (x)y 00 + a1 (x)y 0 + a2 (x)y = 0 which is same as (2)

i.e. L2 = L∗2 , if a00 = a1 ,
i.e. L2 is self adjoint if a00 = a1 .
Conversely suppose L2 = L∗2 , let the equation (1) and (2) are identical:

a1 (x) = 2a10 (x) − a1 (x)

a2 (x) = a000 (x) − a1 (x) + a2 (x)

⇒ a000 (x) = a01 (x)

⇒ a00 (x) = a1 (x) + c and the first equation ⇒ c = 0

i.e. a00 (x) = a1 (x).
Corollary:
If the second order equation

d2 y dy
a0 (x) 2
+ a1 (x) + a2 (x)y = 0
dx dx
is self-adjoint then it can be written in the form
 
d dy
a0 (x) + a2 (x)y = 0
dx dx
Proof:
From the above theorem

d2 y dy
a0 (x) 2
+ a1 (x) + a2 (x)y = 0 (1)
dx dx

25
 if a00 (x) = a1 (x).
Equation (1) takes
d2 y dy
a0 (x) 2
+ a1 (x) + a2 (x)y = 0
dx dx
which implies  
d dy
a0 (x) + a2 (x)y = 0
dx dx

Example 1
Legendre’s D.E.
d2 y dy
(1 − x2 ) − 2x + n(n + 1)y = 0
dx2 dx
Here

a0 (x) = (1 − x2 )
a1 (x) = −2x
a2 (x) = n(n + 1)
0
a0 (x) = −2x = a1 (x)
Equation is self-adjoint.

Theorem
The D.E.
d2 y dy
2
+ a1 (x)
a0 (x) + a2 (x)y = 0
dx dx
is not self-adjoint. This D.E. can be transformed into an equivalent self-adjoint D.E. by multi-
plying the equation throughout by
1 R aa1 (x) dx
e 0 (x)
a0 (x)
Given
d2 y dy
2
a0 (x)
+ a1 (x) + a2 (x)y = 0 (1)
dx dx
Multiply Equation (1) throughout by

1 R a1 (x)
dx
e a0 (x)
a0 (x)

We obtain an equation of the form

2
dx d y a1 (x) R dx dy a2 (x) R
R a1 (x) a1 (x) a1 (x)
dx
e a0 (x) + e a0 (x) + e a0 (x) y=0
dx2 a0 (x) dx a0 (x)
where
b0 (x)y 00 + b1 (x)y 0 + b2 (x)y = 0
and
a1 (x) R a1 (x)
dx
b00 (x) = e a0 (x) = b1 (x)
a0 (x)
Since b00 (x) = b1 (x), equation (2) if self-adjoint. But it is not original equation it is equivalent
D.E. of eq (1).

26
Problems 1. x2 y 00 + 7xy 0 + 8y = 0 — (1)

d2 y dy
a0 (x) 2
+ a1 (x) + a2 (x)y = 0
dx
00 0
dx
a0 (x)y + a1 (x)y + a2 (x)y = 0 — (2)
The adjoint D.E to (2) is if

a0 (x)y 00 + (2a10 (x) − a1 (x))y 0 + (a000 − a01 + a2 )y = 0

Here a0 (x) = x2 , a1 (x) = 7x, a2 (x) = 8 — (3)

For the given eq. adjoint is:

x2 y 00 + (4x − 7x)y 0 + (2 − 7 + 8)y = 0

i.e., x2 y 00 − 3xy 0 + 3y = 0 — (4)

2. x2 y 00 − 3xy 0 + 3y = 0 — (1)
Given eq. is in the standardized form

a0 (x)y 00 + a1 (x)y 0 + a2 (x)y = 0 − − − (2)

a0 (x) = x2 , a1 (x) = −3x, a2 (x) = 3

The adjoint D.E to (1) is if

a0 (x)y 00 + (2a00 − a1 (x))y 0 + (a000 − a01 + a2 )y = 0

For the given eq. adjoint is:

x2 y 00 + (2(2x) + 3x)y 0 + (2 + 3 + 3)y = 0

x2 y 00 + 7xy 0 + 8y = 0
Transform both (1) & (2) into equivalent self-adjoint D.E:

x2 y 00 + 7xy 0 + 8y = 0

a0 (x) = x2 , a1 (x) = 7x, a2 (x) = 8

R 7
1 dx x7
multiply by 2 e x = 2 = x5
x x
x7 y 00 − 7x6 y 0 + 8x5 y = 0

self disjoint differential equation (2) in other form

x2 y 00 − 3xy 0 + 3y = 0

a0 (x) = x2 , a1 (x) = −3x, a2 (x) = 3

− 3x2
R
1 dx
Multiply by x2
e x :
1 R − 3x2 dx
= e x
x2
1 −3 R 1
dx 1 −3logx
= e x = e
x2 x2

27
 1 log x− 3
= e
x2

x− 3 1
= = 5
x2 x
We transform the equation:

x2 00 3xy 0 3y
y − 5 + 5 =0
x5 x x
Simplified to:
1 00 3 3y
y − 4 y0 + 5 = 0
x3 x x
Where:
1 −3 3
a0 (x) = , a1 (x) = , a2 (x) =
x3 x4 x5
−3
= a1 (x) a0 (x) =
x4
Green function, Green formula and Lagrange’s identity
Let f (x) and g(x) be two functions having continuous derivatives up to an order n on [a, b].
Consider:
Z b
g(x)Ln (f (x)) dx
a

dn dn−1
Ln = a0 (x)+ a1 (x) + · · · + an (x)
dxn dxn−1
Z b Z b
dn dn−1
 
g(x)Ln (f (x)) dx = g(x) a0 (x) n + a1 (x) n−1 + · · · + an (x)f (x)
a a dx dx
Z b 
= f (n) (x) [a0 g] + f ( n − 1) [a1 g] · · · + f (x)(an g) dx (1)
a
Expect the last term on RHS of (1), integrate all other terms on RHS by parts till no derivative
of f (x) occurs inside the integral.

Z b n ob
g(x) ln f (x) dx = f (n−1) (x)[a0 g] + (−1)f (n−2) (x)[a0 g]0 + . . . + (−1)n−1 f (x)[a0 g](n−1)
a a

Z b
+(−1 ) n
f (x)[a0 g](n) + f (n−2) (x)[a1 g] + (−1)f (n−3) (x)[a1 g]0 + . . .
a
Z b
(−1)(n−2) f (x)[a1 g](n−2) |ba + (−1) + [an−1 g]dx
a
Z bh i
= (−1)n [a0 g]n f (x) + (−1)(n−1) [a1 g](n−1) f (x) + . . . + (−1)1 [an g]0 f (x) + [an g]f (x) dx
a
Z bh i
= (−1)n [a0 g]n + (−1)(n−1) [a1 g](n−1) + . . . + (−1)1 [an g]0 + [an g] f (x)dx
a

28
 n o
integrated terms ≡ [f, g](x)ba

Where
n n−1
X X
{f, g} = (−1)r−s−1 f s (x)[an−r g]r−s−1
r=1 s=0
n
X dn−r
Ln f (x)dx = ar (x) f (x)
dxn−r
r=0

therefore
n
X dn−r
Ln ∗ f (x)dx = (−1)n−r f (x)
dxn−r
r=0

Here, eq (3) is called Green’s formula which provides a fundamental relation between ln and
Ln ∗ . Eq (3) can also be written as:
Z b
d
g(x)Ln f (x) − Ln ∗ g(x)f (x) − [f.g](x)dx = 0
a dx
d
⇒ g(x)Ln f (x) − Ln ∗ g(x)f (x) = [f.g](x)
dx
Eq (4) is called Lagrange’s identity.
Note We observed that [f, g](x) = c, a constant, is a differentiable eqn of order (n − 1) and n
given by
n n−1
X X
(−1)r−s−1 f s (x)[an−r g]r−s−1 = c
r=1 s=0

Equation (5) takes the form for n = 2:

n n−1
X X
(−1)r−s−1 f s (x)[a2−r g]r−s−1 = c
r=1 s=0

Theorem: Let f (x), g(x) be two functions having continuous derivatives of order n on [a, b].
Now the function f (x) is a solution of Lny = 0 on [a, b] if it is a solution of

[f, g](x) = c
where c is a constant, g(x) is a solution of Lny = 0. Proof: Suppose f (x) is a solution of

Ln ∗ y = 0 (-1)

We have to prove that f (x) is a solution of [f, g](x) = c.

Since g(x) is a solution Ln ∗ y = 0. where congruence identity

d[f.g]
g(x)Ln f (x) − f (x)Ln ∗ g(x) = (x) (1)
dx
Then
d
[f, g](x) = 0 (integration this)
dx
[f, g](x) = c (constant)
This means that f satisfies the equation (1).
Conversely, suppose that f (x) is a solution of [y, g](x) = c. i.e. [f, g](x) = c = 0.

29
 From equation (2) we get

g(x)Ln f (x) − f (x)Ln ∗ g(x) == 0

For a function g(x) solution of Ln ∗ y = 0 we have g(x)Ln t(x) = 0 ⇒ g(x) 6= 0, Ln t(x) = 0.

t(x) is a solution of Ln y = 0.
Example: h i
dk dk
P.T. the Operator L = dx k p(x) dxk
where p(x) is a real valued differentiable function, is
self-adjoint.
By Green’s formula the fundamental relation LandL∗ is given by
Z b Z b
g(x)Ln f (x) = f (x)Ln ∗ g(x)
a a
Z b Z b
gLn f dx = f Ln ∗ g + c([f.g](x))ba
a a
where f (x), g(x) have continuous derivative of order n on [a, b].
If Nn = L∗n then Ln is self-adjoint.
Now consider Z b Z b
dk dk
 
g(x)Ln f (x)dx = [g(x) k p(x) k t(x) ]dx
a a dx dx
Z b h ik
= [g(x) p(x)f (k) (x) ]
a

Integrating the right-hand side k th time by parts, we get

h ib
k−1 0 k k−2
g(x)[(p(x)f (x))] − (g (x)p(x) f (x) ) . . .
a
Z b
(−1) (g k (x)p(x)f (x)k)dx
a
Z b Z b
= (−1)k p(x)g(x)k f (x)k dx + integratingterms
a a

Integrating the 1st term on RHS on k times by parts, we get

h ib
(−1)k p(x)g(x)k f (x)k−1 − (p(x)g(x)k )0 f (x)k−2 + . . . +
a
Z b
(−1)k (g k (x)p(x)f (x)k)dx
a
Z b Z b
= (−1)k (−1)k f (x)[p(x)g(x)k ]k dx + integrating terms
a a
Z bn o
= (p(x)q(x)) (g(x))k dx + integrated term
a
Z b Z b
g(x)f (x)dx = f (x)Lg(x) + integratingterms
a a
From the above equation, we observed that differential operators involved in the integral of
RHS, if L is self-adjoint, would have been otherwise L∗ .

L∗ = L

30
 h i
dk dk
L is self adjoint. L = dx k p(x) dxk

Example: Verify the Lagrange’s identity for

xy = 3y 4 + 3y = 0
For n = 2. Lagrange’s identity is
 
d [f.g]
g(x)L2 [f (x) − f (x)L∗2 [g(x)] = (x)
dx dx
Z b 
=f g(x)
a

where
[f, g] (x) = a0 (f 0 g − f g 0 ) + (a1 , a00 ) f g
L2 y = a0 y (4) + a1 y 00 + a2 y
00 )
L∗2 y = a0 y ( + (2a2 − a1 )y 0 + (a0 − a1 + a2 )y
given
00 )
L2 y = a0 y ( + a1 y 00 + a2 y
where a0 = x2 , a1 = −3x, a2 = 3.
00 )
L2 y = x2 y ( + (−3x)y 0 + 3y
00 )
L∗2 y = x2 y ( + [2(3) − (−3x)] y 0 + [2(3) − 3 + 3] y
00 )
= x2 y ( + 7xy 0 + 8y

d
g(x)L2 y − f (x)L∗2 y = {f, g} [x]
dx
LHS of Lagrange’s identity:
h 00
i h 00
i
g x2 f ( ) − 3xf 0 + 3f − f x2 g ( ) + 7xg 0 + 8g =

00 ) 00 ) d
x2 gy ( − 3xf 0 g + 3f g − x2 f g ( − 7xf g 0 + 8f g = {f, g} [x]
dx
Next, compute the RHS of Lagrange’s identity:
d d
= {f, g}[x] = (a0 {f 0 g − f g 0 } + (a1 − a0 )f g)
dx dx

d  2 0
x {f g + f g 0 − f g}

=
dx
d  2 0
x {f g + f g 0 } − 5xf g

=
dx
h 00 00
i
= 2x{f 0 g + f g 0 } + x2 f ( ) g + f g ( ) − f 0 g 0 − f g 00 − 5x{f 0 g + f g 0 } − 5f g

= x2 f 00 g − f g 00 − 3xf 0 g − 7xf g 0 − 5f g
 

Thus, verify the Lagrange’s identity proof.

1. x2 y 00 + 9xy 0 + 12y = 0

31
 g(x)[L2 (x)] + f (x)L∗2 (x)g(x) = dx d
[f, g](x)
[f, g](x) = a0 [f g − f g ] + a1 (f − a00 )f g
0 0

L2 y = a0 y 00 + a1 y 0 + a2 y
L∗2 y = a0 y 00 + (2a00 − a1 )y 0 + (a000 − a01 + a2 )y
Given L2 y = a0 y 00 + a1 y 0 + a2 y

a0 = x2 ,
a1 = 9x,
a2 = 12y
L2 y = x2 y 00 + 9xy 0 + 12y
L∗2 y = x2 y 00 + [2(2x) − 9x]y 0 + [2 − 9 + 12]y
= x2 y 00 − 5xy 0 + 5y = 0
LHS of Lagrange’s identity:

g[x2 f 00 + 9xf 0 + 12f ] − [x2 g 00 + 5xg 0 + 5g]

x2 f g 00 + 9xy 0 + 12ygf 0 + 12f g − x2 f g 00 + 5xf g 0 − 5f g
Next compute the RHS of Lagrange’s identity:

d d
[f − g](x) = [a0 (f 0 g − f g 0 ) + (a1 − a00 )f g]
dx dx
d 2 0
= [x (f g − f g 0 ) + 7xf g]
dx
d x2
   
0 0 0 7
[f g − f g ] + 2xq − 2x g 0 + 7x2 y
dx dx dx
= x2 {gf 00 + f g 00 } + 2xf 0 g − 2xf g 0 + 7xf 0 g + 7xf g 0 + 7f g
= x2 {f 00 g − f g 00 } + 9xf 0 g + 5f g 0 + 7f g this Verify of Lagrange’s Identity 3. Verify Lagrange’s
identity by finding the solution of the adjoint equation:

x2 y 00 + 7xy 0 + 8y = 0 (11)

Given:
x2 y 00 + 7xy 0 + 8y = 0 (1)
Find its adjoint equation:
Assume a0 = x2 , a1 = 7x, a2 = 8.
The adjoint differential equation of (1) is:

x2 y 00 + [2(2x) − 7x]y 0 + [2 − 7 + 8]y = 0

Simplifies to:
x2 y 00 − 3xy 0 + 3y = 0 (2)
It is a 2nd order differential equation with variable coefficient.
dy
Put x = ez or z = log x, then dx = dy dz 1
dz . dx . x .
Using these relations:

d2 y dy d2 z
 
d dy dz dz
= +
dx2 dz dz dx dx dz dx2
And:
dy dy
x =
dz dz

32
 Differentiate with respect to x:

d2 y dy
 
d dy dz dz
x 2+ =
dx dz dz dz dx dx
Thus, we find:
d2 y dy 1
x 2
+ = D2 y
dx dz x
Rewrite it:

d2 y dy d2 y
x2 + + 2
dx2 dz dz
Then:

d2 y
x2 = D2 = Dy
dx2
Substitute these into (2) we obtain:

(D − 3D + 3)y = 0
Or:

(D − 1)(D − 3)y = 0 (3)

With m = 1, 3, the solution of (3) is:

ez and ce3z

The solution (3) are x and x3 .

Solution of adjoint to given D.E.
It is known that the solution of (1)is a solution of [y, g](x) = c where c is constant, provided
g(x) is a solution of adjoint eqn (3)

To find the solution (1)we need to find the solution

[y, g](x) = c for g(x) = x, x3

Case (i): g(x) = x

Write D.E[y, g](x) = c in adj. D.E:

a0 (y 0 g − yg 0 ) + (a0 − a00 )yg = 0

x2 y 0 x − y(1) + (7x − 2x)xy = 0



x3 y 0 − x2 y + 5x2 y = 0

x3 y 0 + 4x2 y = c
d 4
[x y] = x (product rule)
dx
Integrating, we get

cx2
x4 y = +d
2

33
 c d
y= 2
+ 4
2x x
Solution of given equation.

Case (ii): g(x) = x3

Write D.E [y, g](x) = c.

a0 {y 0 g − 4yg 0 } + (a0 − a00 )yg = c

x2 [y 0 x3 − y3x2 ] + [(7x − 2x)4y]x3 = 0}

x5 y 0 − 3x4 y + 5x4 y = c

x5 y 0 + 2x4 y = c

2
y0 + y=c
x
dy 2y c
⇒ + = 5
dx x x
R
1. I.F. = e p dx

2
R
dx
I.F. = e x

= e2 log x
2
= elog x
= x2
Z
y · (I.F.) = (I.F.) · Q(x) dx + c1

x2 c
Z
2
yx = dx + c1
x5
Z
2 c
yx = dx + c1
x3

c
yx2 = − + c1
2x2

−c c1
y= +
2x4 x2

A B
y= +
x4 x2
−c
where A = 2 and B = c1 .

34
 2. Find general solution of x2 y 00 + 2xy 0 + 12y = 0 by finding the solution of it and adjoint
equation.
Given x2 y 00 + 2xy 0 + 12y = 0 (1)
Adjoint equation if:
a0 y 00 + (a0 a1 )y 0 + (a000 − a01 + a2 )y = 0

x2 y 00 − 5xy 0 + 5y = 0 (2)

The solution of eqn (1) is given by

Put x = ez

z = log x

dz 1
=
dx x

dy dy dz dy 1 1 dy
= · = · =
dx dz dx dz x x dz

Let Dy = x dy
dz

again Diff w.r.t x:

d2 y
 
dy dz dy
x 2+ =
dx dx dx dz
  
d dy dz
=
dz dz dx

d2 y 1
= ·
dz 2 x

d2 y dy d2 y
x2 + x =
dx2 dx dz 2

d2 y d2 y dy
⇒ x2 = −x from (3)
dx2 dz 2 dx

d2 y
x2 = Dy 2 − Dy (4)
dx2
Sub (4 and (3) in (2)

D2 y − Dy − 5Dy + 5y = 0

(D2 − D + 5D + 5)y = 0

D2 − 6D + 5 = 0

(D − 5D + 1)(D − 5)

35
 D(D) − 5D − D + 5 = 0

(D − 5)(D − 1) ⇒ D = 5, 1

The soln are ez , e5 z

The soln are x, 5x
Known that soln {y, g}x = c where g(x) is a soln of adjoint D.E.
Try g(x) = x, x5 to find soln.
i) g(x) = x, sub in Eq (1)

(y[g])(x) = c

a0 (y 0 g − yg0 ) + (a1 − a00 )yg = c

R
p dx
I.F. = e
2
R
dx
=e x

= e2 log x
2
= elog x
= x2
Z
y(I.F.) = (I.F.)θ(x) dx + c1
Z
c
yx = x2 5 dx + c1
2
x
Z
2 c
yx = dx + c1
x3
c
yx2 = − 2 + c1
2x
c c1
y=− 4 + 2
2x x
A B
y= 4+ 2
x x
c
where A = − , B = c1
2
4. Find general solution of x2 y 00 + 9xy 0 + 12y = 0 by finding the solution of it and adjoint
equation.
Given x2 y 00 + 9xy 0 + 12y = 0 (1)
Adjoint equation is

a0 y 00 + (a0 − a1 )y 0 + (a000 − a01 + a2 )y = 0

x2 y 00 − 5xy 0 + 5y = 0 (2)
The solution of eq. (1) is also given by

Put x = ez

36
 z = log x
dz 1
=
dx x
dy dy dz dy 1 dy dy
= · = · =x =D
dx dz dx dz x dx 3
again Ddff w.r.t x:
 
d2y dy dz dy
x 2 + =
dx dx dx dz
  
d dy dz
=
dz dz dx

d2 y 1
= ·
dz 2 x

d2 y dy d2 y
x2 + x = −
dx2 dx dz 2

d2 y d2 y dy
x2 2
= 2
−x
dx dz dx

d2 y
x2 = D2 y − Dy 4
dz 2
Sub (4) and (3) in (2):

D2 y − Dy + 5Dy + 5y = 0

(D2 − D − 5D + 5)y = 0

D2 − 6D + 5 = 0

(D − 5)(D − 1)y = 0

D(D − 5) − 1(D − 5)

(D − 5)(D − 1) =⇒ D = 5, 1
Therefore the soln are e5z , ez
The poln are x, x5
Known that soln {[y.g]x = c} where g(x) is a soln of the adjoint D.E.
Let g(x) = x, x5 to find soln .

i) g(x) = x sub in eq (1).

{y[g](x) = c}

a0 (y 0 g − gy 0 ) + (a1 − a00 )yg = c

37
 x2 (xy 0 − y) + (9x − 2x)xy = c

x3 y 0 − x2 y + 7x2 y = c

x3 y 0 6x2 y c
3
+ 3 = 3
x x x

6 c
y0 + y= 3
x x

dy 6 c
+ y= 3
dx x x

6
R R
p dx dx
I.F = e =e x = e6 log x = x6

Z
c
y(I.F ) = (I.F ) · dx + c1
x3
Z
c
yx6 = x6 dx + c1
x3
Z
= x3 c dx + c1

c
yx6 = x4 + c1
4

c c1
y= +
4x2 x6
where
A B
y= 2
+ 6
x x

c
A= and B = c1
4

(ii) g(x) = x5

[y · g](x) = c

a0 (y 0 g − gy 0 ) + (a1 − a00 )yg = c

x2 y 0 x5 − 5x4 y + (ax − 2x)yx5 = c

x7 y 0 − 5x6 y + 7x6 y = c

38
 x7 y 0 + 2x6 y = c

2 c
y0 + y= 7
x x
I.F
R R
p dx 2 log x
e =e = x2
d d


Show that operation L = dx p(x) dx is self adjoint where p(x) is a real valued function.
Consider Z b Z b  
d d
g(x)Lf (x) dx = g(x) p(x) f (x) dx
a a dx dx
Z b
0
g(x) p(x)f 0 (x) dx

=
a
Integrating RHS by parts:
b Z b
= g(x) p(x)f 0 (x) g 0 (x) p(x)f 0 (x) dx
   
−
a a
b Z b Z b
= g(x) p(x)f 0 (x) f (x)[p(x)g 0 (x)] dx
  
− [f (x)Lg(x)] dx +
a a a
Z b Z b
g(x)Lf (x) dx = [f (x)Lg(x)] dx + constant terms
a a
By above equation, left integral involves no derivatives of g(x) if right integral involves no
derivative of f (x) but the diff operation involved on RHS is L which would have been L∗ .

L∗ = L =⇒ L is self adjoint.
Theorem: Solution of non-homogenous D.E using the method of variation of
parameters.
Consider nth order linear non-homogeneous ODE

a0 (x)y (n) + a1 (x)y (n−1) + · · · + an (x)y = b(x) (1)

where a0 (x), a1 (x), . . . , an (x), b(x) are continuous functions on I with a0 (x) 6= 0 when the
complimentary function of eqn (1) is known as the complete solution of eqn (1) may be obtained
by quadrature. The general solution ψ(x) of (1) is of the form
n
X
ψ(x) = cr φr (x) + ψp (x) (2)
n=1

where c1 , c2 , . . . , cn are constants, {φ1 (x), . . . , φn (x)} is a fundamental set of (2). Let Ln y =
b(x) and ψp (x) is the particular solution of (1) if we replace the constant.
in (2) by the functions Cr (x) whose values are to be determined i.e ψp (x) = nr=1 Cr0 (x)φr (x)
P
(3)

Diff. w.r.t x

39
 n
X
ψp0 (x) = Cr0 (x)φn (x) + Cr (x)φ0p (x)
r=1
n
X n
X
ψp0 (x) = Cr (x)φn (x) + Cr (x)φ0p (x)
r=1 r=1
n
X
Cr0 (x)φr (x) = 0 (4)
r=1
So we have
n
0
X
ψp (x) = Cr (x)φ0r (x)(5)
r=1

Diff cont x

n
X
ψp00 (x) = Cr (x)φ00r (x) + Cr0 (x)φ0p (x)
r=1
n
X n
X
ψp00 (x) = Cr0 (x)φ0r (x) + Cr (x)φ00r (x)
r=1 r=1

As a II condition on ci for i = 1 to n
assume that
n
0
X
Cr0 (x)φr (x) = 0
r=1
So we have
n
X
ψp00 (x) = Cr (x)φ00r (x) (6)
r=1
continuing the procedure
n
00
X
ψp( ) (x) = Cr (x)φ(n−1)
n (x)(7)
r=1
provided
n
X
Cr0 (x)φ(n−2)
r (x) = 0 (8)
i=1
Now diff 7 we get
n
X
0
ψp(n) x = Cr (x)φ(n) n−1
r (x) + Cr (x)φr (9)
i=1

ψp (x), if a particular solution of eq. 1 we have

a0 ψp + a1 ψpn−1 + a2 ψpn−2 (x) + . . . + an ψp (x) = b(x) (10)

(i)
Sub for ψp (x) for i = 1, . . . , n in eqn (11)

40
 n n
" # " n # " n
#
X X X X
⇒ a0 c0r (x)φnr (x) + c0r (x)φn−1
r (x) +a1 cr (x)φn−1
r (x) +. . .+an cr (x)φr (x) = b(x)
r=1 r=1 r=1 r=1

n
X
a0 cr (x)φnr (x) + a1 c0r (x)φn−1
 
⇒ r (x) + . . . + an cr (x)φr (x) = b(x) (11)
r=1

φ1 to φn form a fundamental set of eqn (10).

Peqn (11) is ⇒ 0.
The first of
we get a0 nr=1 c0r (x)φn−1
r (x)x = b(x) (12)
(12)
Thus it is observed that ψp (x) is a particular solution of eqn (1)
n
X n
X
c0r (x)φr (x) = 0, (c0r (x)φ0r (x) = 0, c00r (x)φ00r = 0,
r=1 r=1

In homogeneous linear ODE using the method of variation of parameter.

1
y 00 + y =
1 + sin x
The given function is y 00 + y = 1
1+sin x .
A.E: m2 + 1 = 0

m2 = −1
m = ±i
solutions are cos x + sin x.
φ1 (x) = cos x, φ2 (x) = sin x.
C.F. = cos x + sin x.
By the method of variation of parameter of solution y,

y(x) = c1 φ1 (x) + c2 φ2 (x) + c1 (x)φ1 (x) + c2 (x)φ2 (x)

= c1 cos x + c2 sin x + c1 (x) cos x + c2 (x) sin x
where
W1 (φ1 (x), φ2 (x))b0 (x)
c1 (x) = dx
W (φ1 (x), φ2 (x))a0 (x)

cos x sin x
W (φ1 (x), φ2 (x)) = = 1 6= 0
− sin x cos x

0 sin x
w1 = = − sin x
g(x) cos x
− sin x 1 − sin x
Z Z
1
c1 (x) = − sin dx = · dx
1 + sin x 1 + sin x 1 − sin x

− sin x + sin2 x
Z
= dx
1 − cos2 x sin x
Z Z
= − tan x sec x dx tan2 x dx

41
 = − sec x + tan x + x

n n
X X b(x)
cr (x)φ00r (x) = 0, . . . , c0r (x)φm−1
r (x) ·
a0 (x)
r=1 r=1

We have a system of eqn equating summation

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

c1 (x)φ01 (x) + c2 (x)φ02 (x) + . . . + cn (x)φ0n (x) = 0

c1 (x)φn−1
1 (x) + c2 (x)φ2n−1 (x) + . . . + cn (x)φn−1
n (x) = 0 (13)

In matrix form eqn (13) written as:

    
φ1 (x) φ2 (x) ... φn (x) c1 0
 φ0 (x) φ02 (x) ... φ0n (x)   c2  b(x) 0

 1
  ..  =
  
.. .. .. .. .
  .  a0 (x)  .. 

 . . . .
n−1 n−1
φ1 (x) φ2 (x) . . . φnn−1 (x) cn 0
(14) Solving (14) by Cramer’s Rule, we obtain:

w1 {φj (x) : j = 1, . . . , n} b(x)

ci (x) =
w {φj (x) : j = 1, . . . , n} a0 (x)
where w1 {φj (x) : j = 1, . . . , n} is the determinant formed from w {φj (x) : j = 1, 2, . . . , n} by
(n−1)
replacing the 1st column elements φ1 , . . . , φ1 by (0, 0, . . . , 0)T .

wr {φj (x) : j = 1, 2, . . . , n} b(x)

cr (x) = (15)
w {φj (x) : j = 1, 2, . . . , n} a0 (x)
where r = 1, 2, . . . , n.
wr {φj (x) : j = 1, . . . , n} is obtained from w {φj (x) : j = 1, 2, . . . , n} by replacing the r-th
column by (0, 0, . . . , 1).
Now integrating (15) with respect to x, we obtain:

wr {φj (x) : j = 1, . . . , n} b(x)

Z
cr (x) = dx (16)
w {φj (x) : j = 1, . . . , n} a(x)
for r = 1, 2, . . . , n.
Thus, the general solution of equation (1) is:
n
X n
X
ψ(x) = cr φr (x) + cr (x)φr (x)
r=1 r=1

where cr (x) are given by as in (16).

Example
Using the method of variation of parameters, solve the equation x2 y 00 − xy 0 − 3y = x3 .
Consider x2 y 00 − xy 0 − 3y = 0 (1).
Let:
y = xm
y 0 = mxm−1

42
 y 00 = m(m − 1)xm−2
Substituting, we get:

x2 (m(m − 1)xm−2 ) − xmxm−1 − 3xm = 0

⇒ m(m − 1)xm − mxm − 3xm = 0

⇒ xm [m(m − 1) − m − 3] = 0
m2 − 2m − 3 = 0
m2 − 3m + 1m − 3 = 0
m(m − 3) + 1(m − 3) = 0

m = −1, 3

The solutions are e−z , e3z

The solutions are 1/z, z 3

1
y1 (x) = 1/x φ2 (x) = x3 ⇒ C.F. = c1 x3 + c2
x
By the method of variation of parameters, the general solution ψ is given by:

ψ(x) = c1 (1/x) + c2 x3 + c1 (x)φ1 (x) (1)

where Z
w{φ1 (x), φ2 (x)}b(x)
c1 (x) = dx
w{φ1 (x), φ2 (x)}a0 (x)

1/x x3
W [φ1 (x), φ2 (x)] = = 3x − x2 = 2x 6= 0
−1/x 3x2
2

0 x3
w1 = = −x3
1 3x2

−x3 x3
Z
c1 (x) = dx = · · ·
4x × x2

−x3
Z
c1 (x) = dx
4

−x4
=
16
w2 {φ1 (x), φ2 (x)}b(x)
Z
c2 (x) = dx
w{φ1 (x), φ2 (x)}b(x)
1
ω2 =
x

1/xx3
Z
c2 (x) = dx
4xx2

43
 Z
1
= dx
4x
1
=
log x
4
 
cos x 0
w2 = = cos x
− sin x 1
Z Z
cos x cos x
c2 (x) = dx = dx
1 + sin x 1 + sin x
Multiplying and dividing by 1 − sin x,

cos x · (1 − sin x)
Z
= dx
1 + sin x · (1 − sin x)

cos x · dx
Z Z Z Z
sin x dx 1
= − = dx + tan x dx
cos2 x cos x cos x
Thus,
c2 (x) = log(1 + sin x)
y(x) = c1 cos x + c2 sin x + C
Find the general solution of

(x2 + 1)y 00 − 2xy 0 + 2y = 6(x2 + 1)

Given that y = x is the solution of the corresponding homogeneous equation, consider

(x2 + 1)y 00 − 2xy 0 + 2y = 0

The given equation is the normalized form:
2x 2
(y 00 ) − (y 0 ) + 2 y=0
x2 +1 x +1
Where
φ1 (x) = x
is one solution of (1). Another solution is:
R
Z
e a(x) dx
φ2 (x) = φ1 (x) dx
φ21 (x)
Z
1 R x22x+1 dx
a. x e dx
x2
Z
1 log(x2 +1)
=x e dx
x2
Z 2
x +1
=x dx
x2
Z  
1
=x 1 + 2 dx
x
 
1
=x x −
x

44
 φ2 (x) = x2 − 1
Therefore, the general solution of (1) is

ψ(x) = c1 x + c2 (x2 − 1)x + c3 (x)x + c4 (x)(x2 − 1)

where
w1 (x, x2 − 1) 6(x2 + 1)2
Z
c1 (x) = · dx
w(x, x2 − 1) x2 + 1

0 x2 − 1
1 2x 6(x2 + 1)2
Z
= dx
2
x x − 1 x2 + 1
1 2x
−x2 + 1
= · 6(x2 + 1) dx
2x − x2 + 1
 3 
−x
=6 +x
3

x3
 
=6 x−
3

w2 (x, x2 − 1) 6(x2 + 1)2

Z
c2 (x) = · dx
w(x, x2 − 1) x2 + 1
Z Z
x
6 (x2 + 1) dx
2x2 − x2 + 1
Z
x
= 2
· 6 (x2 + 1) dx
x +1
Z
= 6 x dx
 2
x
c2 = 6
2
c2 = 3x2

x3
y(x) = c1 x + c2 (x2 − 1)x + 6(x − )x + 3x2 (x2 − 1)
3
4.
sin2 x y 00 − sin 2x y 0 + (1 + cos2 x)y = sin3 x given
that y = sin x is one solution of corresponding homogeneous eqn
consider
x sin2 x y 00 − sin 2x y 0 + (1 + cos2 x)y = 0 (1)
The given eqn is normalized form is divided by sin2 x

1 + cos2 x
 
sin 2x 0
y 00 − y + y = 0 (2)
sin2 x sin2 x
Given φ1 (x) = sin x, if solution of (2)

45
 Z
1 R
φ2 (x) = φ2 (x) e− a(x) dx dx
φ1 (x)
Z
1 R sin x
dx·dx
= sin x e sin2 x
sin2 x
Z
1 R 8 sin x·dx
dx·dx
= sin x e sin x
sin2 x
Z
1
= sin x e2 log sin x dx
sin2 x
Z
1
= sin x sin2 x dx
sin2 x
Z
1
= sin x sin2 x dx
sin2 x

φ2 (x) = x sin x
The general solution of (1) if y(x) = c1 sin x + c2 x sin x + c1 (x) sin x + c2 (x) sin x

0 x sin x
1 x cos x + sin x sin3 x
Z
c1 = c2 (x) = · dx
sin x x sin x sin2 x
cos x x cos x + sin x
Z
sin x
= −x sin x dx
x sin x + x sin x cos x2 − x sin x cos xx
−x sin x
Z
= sin x dx
x sin2 x

−x3
Z
c1 (x) = −x2 dx =
3

sin x 0
cos x 1 sin3 x
Z
c2 (x) = · dx
sin x x sin x sin2 x
cos x x cos x + sin x
Z
sin x
= sin x dx = x
sinx
 3
−x
y(x) = c1 sin x + c2 sin x + sin x + x2 sin x
3
Solve y 00 + 4y = csc 2x
The given eqn y 00 + 4y = 0
A.E: m2 + 4 = 0
m = ±2i
Therefore, the solutions are C1 cos 2x + C2 sin 2x.

y = C1 cos 2x + C2 sin 2x + C1 (x) cos 2x + C2 (x) sin 2x

46
 0 sin 2x
1 2 cos 2x
Z
c2 (x) = · csc 2x dx
cos 2x sin 2x
−2 sin 2x 2 cos 2x
Z
csc 2x
= (− sin 2x) · cos 2x dx
2 cos2 2x + 2 sin2 2x
1
dx = − x2
R
C1(x) = − 2

cos 2x 0
−2 sin 2x 1
Z
C2 (x) = · cos 2x dx
cos 2x sin 2x
−2 sin 2x 2 csc 2x
Z
cos 2x
= · csc 2x dx
2 cos2 2x + sin2 2x
1
= log(sin 2x)
4
x 1
y(x) = C1 cos 2x + C2 sin 2x − cos 2x + log(sin 2x) sin 2x
2 4
Problem 6

y 00 + y = tan x
Auxiliary Equation (A.E):

m2 + 1 = 0 ⇒ m = ±i
The solutions are cos x + sin x.

y(x) = C1 cos x + C2 sin x + c1 (x) cos x + c2 (x) sin x

Where

0 sin x
1 cos x − sin2 x
Z Z
c1 (x) = · tan x dx = dx
cos x sin x cos x
− sin x cos x
sin2 x cos2 x − 1
Z Z
= dx = dx
cos x cos x

c1 (x) = sin x − log(csc x + tan x)

And

cos x 0
− sin x 1
Z Z
cos x
c2 (x) = · tan x dx = sin x dx
cos x sin x cos x
− sin x cos x

c2 (x) = − cos x

47
 Thus

y(x) = C1 cos x + C2 sin x + (sin x − log(sec x + tan x)) cos x − cos x sin x

Zeros of Solutions of Linear ODE:

Theorem A:
Ln y = 0.
Linear n-th order ODE, a0 (x) 6= 0. If a ≤ x ≤ b let x0 ∈ I and (c0 , c1 , . . . , cn−1 ) are constants.
Then there exists a φ defined on I of Ln y = 0 such that

φ(x0 ) = c0 , φ0 (x0 ) = c1 , ..., φ(n−1) (x0 ) = cn−1 .

Corollary A:
If φ is a solution of Ln y = 0 such that

φ(x0 ) = 0, φ0 (x0 ) = 0, ..., φ(n−1) (x0 ) = 0

for x0 ∈ [a, b] with the coefficients a0 , . . . are constants on [a, b], a0 (x) 6= 0, then φ(x) ≡ 0 for all
x ∈ [a, b], i.e. a trivial solution.

Liouville’s Formula:
Let f1 , f2 , . . . , fn be n solutions of Ln y = 0 on [a, b] and x0 ∈ [a, b]. Then the
Z x
− an (t)
w{f1 , . . . , fn }(x) = w{f1 , . . . , fn }(x0 ) e dt ∀ x ∈ [a, b].
x0 a0 (t)

We know that self-adjoint second order equation may be written in the form

(p(x)y 0 )0 + q(x)y = 0

where p(x) is continuous derivative p(x) > 0 on [a, b].

Theorem B:
Let φ be a solution of
(p(x)y 0 )0 + q(x)y = 0
having a first order derivative φ0 [a, b]. If φ has an infinite number of zeros on [a, b], then φ(x) ≡ 0
for x ∈ [a, b].

Proof:
Given
(p(x)y 0 )0 + q(x)y = 0
is a self-adjoint second order ODE. Suppose the solution φ(x) has infinetly many zeros on [a, b]
By Bolzano-Weierstrass Property, the set of zeros of φ(x) has a limit point say x0 ∈ [a, b]. There
is a sequence {xn } of points on the set of zeros of φ(x) such that xn → x0 .
Since φ is continuous:
i.e., limn→∞ φ(xn ) = φ(xn ) → φ(x0 ) = 0,
φ(x0 ) = 0, since φ(xn ) = 0 for all xn .

48
 Since φ0 (x0 ) exists:
i.e., limx→x0 φ(x)−φ(x
x−x0
0)
= φ0 (x0 ) exists.
We approach x0 by the sequence {xn }:
Then limn→∞ φ(xxnn)−φ(x −x0
0)
= φ0 (x0 ),
Since φ(xn ) = 0 = φ(x0 ),
we have φ0 (x0 ) = 0.
Then we have P.T. there is q, s so that:
0
(p(x)y )0 + q(x)y = 0)
such that for some x0 ∈ [a, b], φ(x0 ) = 0, φ0 (x0 ) = 0.
From the corollary (if φ is a solution of Ln y = 0
such that φ(x0 ) = 0, φ0 (x0 ) = 0, ..., φ(n) (x0 ) = 0 for x ∈ [a, b] with the coefficients
a0 , a1 , . . . , an are constant on [a, b], a0 (x) 6= 0. Then φ(x) ≡ 0 ∀x ∈ [a, b] is a trivial solu-
tion.)
Corollary:
0
A non-trivial solution (p(x)y )0 + q(x)y = 0)
has only a finite number of zeros in any bounded closed sub-interval of [a, b].
Suppose φ(x) has infinitely many zeros in a closed bounded sub-interval of I. Then φ(x)
becomes a trivial solution (by previous theorem).
Therefore, φ is a non-trivial solution of φ and has only finitely many zeros in a closed bounded
interval.

Abel’s Formula
Let φ1 , φ2 be any two solutions of:

(p(x)y 0 )0 + q(x)y = 0 on [a, b]

Then ∀x ∈ [a, b]:

p(x) φ1 (x)φ02 (x) − φ01 (x)φ2 (x) = K, a constant.

Proof
Given φ1 (x), φ2 (x) are solutions of (p(x)y 0 )0 + q(x)y = 0,

(p(x)φ01 (x))0 + q(x)φ1 (x) = 0 — (2)

(p(x)φ02 (x))0 + q(x)φ2 (x) = 0 — (3)

To eliminate q(x) on (2) and (3), multiply equation (2) by φ2 (x) and equation (3) by φ1 (x),
and subtract equation (3) by equation (2), we get:

(p(x)φ01 (x))φ2 (x) − (p(x)φ02 (x))φ1 (x) = 0

Integrating by parts from a to x:
Z x Z x
x x
0
(p(x)φ1 (x)φ2 (x)) − p(t)φ01 (t)φ2 (t)dt − (p(x)φ02 (x)φ1 (x)) + p(t)φ02 (t)φ1 (t)dt = 0
a a a a

[p(x)φ01 (x)φ2 (x) − p(a)φ1 (a)φ2 (a)] − [p(x)φ02 (x)φ1 (x) − p(a)φ02 (a)φ1 (a)]+ = 0
Given:

P (x) φ01 (x)φ2 (x) − P (a)φ01 (a)φ2 (a) − p(x)φ02 (x)φ1 (x) − P (a) φ02 (a)φ1 (a)
   

=0

49
 P (x) φ1 (x)φ2 (x) − φ02 (x)φ1 (x) = P (a) φ2 (a)φ01 (a) − φ02 (a)φ1 (a)
   

P (x) φ1 (x)φ2 (x) − φ02 (x)φ1 (x) = k a constant where

 

k = P (a)(φ2 (a)φ01 (a) − φ02 (a)φ1 (a))

Theorem:
Let φ1 and φ2 be two solutions of (p(x)y 0 )0 + q(x)y = 0, such that φ1 and φ2 have a common
zero, then φ1 & φ2 are linearly dependent on [a, b].

1. By Abel’s formula

P (x) [φ1 (x)φ2 (x) − φ2 (x)φ1 (x)] = k, x ∈ [a, b]

If x0 ∈ [a, b] is a common zero of φ1 (x) and φ2 (x), then φ1 (x0 ) = 0 = φ2 (x0 ). From (1),
k = 0.

Since P (x) > 0

φ1 (x)φ02 (x) − φ2 (x)φ01 (x) = 0, x ∈ [a, b]

φ1 (x) φ2 (x)
W = =0
φ01 (x) φ02 (x)

w [φ1 , φ2 ] (x) = 0 ∀x ∈ [a, b]

φ2 (x) and φ1 (x) are linearly dependent

2. Since φ2 (x) and φ1 (x) are linearly dependent on (a, b), ∃ constant c1 , c2 , not both zero

such that c1 φ1 (x) + c2 φ2 (x) = 0, x ∈ [a, b]

By hypothesis neither φ1 nor φ2 is zero on [a, b].

If c1 = 0 Then c2 = 0
If c2 = 0 then c1 = 0, this is contradiction.

therfore c1 6= 0 then c2 = 0 if φ1 (x0 ) for some x0 in [a, b]

Sturm’s Separation Theorem

Let φ1 (x) and φ2 (x) be linearly independent solutions of the self-adjoint equation

(p(x)y 0 )0 + q(x)y = 0 on [a, b].

Then, between the consecutive zeros of φ1 (x) there is precisely one zero of φ2 (x) and vice versa.

50
Proof:
Since φ1 (x) and φ2 (x) are solutions of

(p(x)y 0 )0 + q(x)y = 0, (1)

we have
(p(x)φ1 (x)0 )0 + q(x)φ1 (x) = 0 (2)
and
(p(x)φ2 (x)0 )0 + q(x)φ2 (x) = 0. (3)
Multiply (2) by φ2 (x), (3) by φ1 (x), and subtract (3) from (2):

φ2 (x)(p(x)φ01 (x)0 ) − φ1 (x)(p(x)φ2 (x)0 ) = 0.

Integrating the above equation from a to x by parts:
Z x Z x
x x
− (p(x)φ1 (x)φ2 (x))a − p(x)φ1 (x)φ2 (x) dx − p(x)φ01 (x)φ2 (x) a
+ p(x)φ02 (x)φ01 (x) dx = 0.
a a

= p(x)φ01 (x)φ2 (x) − p(x)φ1 (x)φ02 (x) + P (a)φ01 (a)φ2 (a) + P (a)φ1 (a)φ02 (a)
⇒ P (x)(φ01 (x)φ2 (x) − φ02 (x)φ1 (x)) = k. (4)
Where k = P (a)(φ01 (a)φ2 (a) − φ02 (a)φ1 (a)) = α (constant).
Equation (4) holds for all values of x ∈ [a, b]. Suppose x1 and x2 are two consecutive zeros
of φ1 (x) in [a, b] with x1 < x2 .
From (4)
P (x1 ) φ01 (x1 )φ2 (x1 ) − φ02 (x1 )φ1 (x1 ) = k

where
P (x2 ) φ01 (x2 )φ2 (x2 ) − P (x2 ) φ02 (x2 )φ1 (x2 ) = k

So,
P (x1 ) φ01 (x1 )φ2 (x1 ) = P (x2 ) φ01 (x2 )φ2 (x2 )



(5)

P (x2 ) (φ01 (x2 )φ2 (x2 ))

⇒ φ2 (x1 ) =
P (x1 )φ01 (x1 )
We observe that either φ01 (x2 ) < 0 and φ02 (x2 ) > 0 or φ01 (x1 ) > 0 and φ02 (x2 ) < 0
⇒ That φ2 (x1 ) and φ2 (x2 ) have opposite signs or in other words φ(x) has a zero between
say ξ between x1 and x2 such that φ2 has at least one zero between two consecutive zeros x1
and x2 of φ(x).
To prove the uniqueness of ξ,
Suppose η is another zero of φ2 (x) between x1 and x2 . Then, φ2 (ξ) = 0 and φ2 (η) = 0.
Repeating the above discussion procedure for φ we get that φ has a zero between ξ and η.
This contradicts the fact that x1 and x2 are two consecutive zeros of φ.
Thus, there is no zero of φ2 (x) other than ξ between x1 and x2 .
Example:
Illustrative example:
Consider y 00 + y = 0 on (−∞, ∞).

(D2 + 1) = 0 ⇒ b = ±i
m2 + 1 = 0 ⇒ m = ±i
The function on φ1 (x) = cos x, φ2 (x) = sin x are two solutions of (1) and are linearly
independent.

51
 Zeroes of φ1 (x) are π2 + nπ, n = 0, 1, 2, . . .
Zeroes of φ2 (x) are nπ for n = 0, 1, 2, . . .
π
0 and φ are two consecutive zeroes of φ2 (x) = sin x and there is exactly one zero 2 of
φ1 (x) = cos x.
Strum’s Fundamental Comparison Theorem:
Let φ1 (x) be a solution of
(p(x)y 0 )0 + q(x)y = 0
having consecutive zero at x = a and x = b where a < b and φ2 (x) is a solution of

(p(x)y 0 )0 + q(x)y = 0

where q2 (x) > q1 (x), x ∈ [a, b].

Then φ2 (x) has a zero at some point x = c below a and b.
proof: Suppose on the contrary, the solution φ2 (x) has no such zero at some point below a and
b without loss of generality. We assumed that φ1 (x) > 0, φ2 (x) > 0 and x ∈ (a, b).
Since φ1 (x) and φ2 (x) are solutions of

(p(x)y 0 )0 + q(x)1 y = 0 (1)

and
(p(x)y 0 )00 + q2 (x)y = 0 (2)
respectively, we have
(P (x)φ01 (x))0 + q1 (x)φ1 (x) = 0 (3)
(P (x)φ02 (x))0 + q2 (x)φ2 (x) = 0 (4)
Multiply equation (3) by φ2 (x) and (4) by φ1 (x):

q1 (x)φ1 (x)φ2 (x) − q2 (x)φ2 (x)φ1 (x)+

(P (x)φ1 (x))0 φ2 (x) − (P (x)φ2 (x))0 φ1 (x) = 0 (5)

Now the above equation can be written as

d 
P (x)(φ01 (x)φ2 (x)) − φ01 (x)φ1 (x) = q2 (x)φ2 (x)φ1 (x) − q1 (x)φ1 (x)φ2 (x)

dx
Integrating the above equation w.r.t. x between a and b, we obtain:

b
Z b
P (x)φ1 (x)φ2 (x) − φ01 (x)φ1 (x) a =

[q2 (x)φ1 (x)φ2 (x) − q1 (x)φ1 (x)φ2 (x)]dx
a

[P (b)(φ1 (b)φ02 (b) − φ1 (b)φ02 (b))] − [P (a)(φ1 (a)φ02 (a) − φ1 (a)φ02 (a))] > 0 (6)
Hence
q2 (x) > q1 (x). φ1 (x) > φ2 (x) > 0 on (a, b)
Since φ1 (a) = 0 = φ1 (b)
P (b)φ01 (b)φ2 (b) − P (a)φ01 (a)φ2 (a) > 0 (6a)
Also φ01 (a) > 0 and φ01 (b) < 0

P (b)φ01 (b)φ2 (b) − P (a)φ1 (a)φ2 (a) < 0 (7)

Thus we arrived at two inequalities contradicting each other.

φ2 (x) has at least one zero at some point x = c between a and b.

52
Example: Illustrative
d2 y
Consider the equation dx2
+ A2 y ≡ 0 Consider

d2 y
+ A2 y = 0 (1)
dx2
d2 y
+ B2y = 0 (2)
dx2
where A and B are constant s.t. B > A > 0.

Solution:
Let φ1 (x) = sin Ax, φ2 (x) = sin Bx Then φ1 (x) and φ2 (x) are solutions of (1) and (2) respec-
tively.
φ1 (π/A) = sin Aπ/A = sin π = 0
x = π/A
The solution φ1 (x) as a solution of (1) has zero at x = 0 and x = π/A.
Now the solution φ2 (x) of (2) has a zero at π/B and also we have
π π
B>A⇒ <
B A

Oscillatory and Non-Oscillatory Differential Equations

The D.E. (p(x)y 0 )0 + q(x)y = 0 is said to be

• Oscillatory in I if it is a solution with infinitely many zeroes in I.

• Non-oscillatory in I if all the solutions have only a finite number of zeroes in I.

Example
1. Solve the D.E. y 00 + xk2 y = 0, x > 0. Where k is constant and x > 0 is arbitrary. The equation
is non-oscillatory according to:
1 1
k> , 0<k≤
4 4
Given equation:
k
y 00 + 2 y = 0
x
Rewriting:
x2 y 00 + ky = 0 (1)
Put x = ez , Then z = log x, Differentiating:
dz 1
=
dx x
Therefore,
dy dy dz
= ·
dx dz dx
Thus, we rewrite the differential equation with respect to z.
dy dy
x = = Dy
dx dz

53
 d2 y dy d2 y 1
x + = ,
dx2 dx dz 2 x
d2 y dy d
x2 2 + x = d2 x (D(D-1)),
dx dx dy
d2 y
x2 2 = D2 y − Dy,
dx
d2 y
− dy = (D(D − 1))y,
dz 2
k 2 y − ky + ky = 0,
(D(D − 1) + k)y = 0,
(D2 − D + k)y = 0,
a2 = m2 = m2 + k = 0,
√
1 − 4k
m=1±
 m2
2 (z)
m = e = xm ,
1
m is real if 1 − 4k ≥ 0 and k ≤ ,
4
φ(x) = xm , now x ∈ [x, ∞],
1
This eq(1) is non-oscillatory in [x, ∞] for k ≤ .
4
Now suppose k > 14 . Then m is complex, say α + iβ, we have a real-valued solution
n o n o n o
φ(x) = Re xα+iβ = Re xα xiβ = Re xα eiβ log x ,
= Re {xα (cos(β log x) + i sin(β log x))} ,
= xα cos(β log x).

φn (x) = 0 if x2 cos(β log x) = 0.

We have a solution of equation (2) which has infinitely many zeros.
The D.E is oscillatory in (x, ∞) with k > 41

Existence or Uniqueness of Solutions of ODE

A D.E may or may not have a solution even if it has a solution, it may not be a unique solution.
Consider the equation:
dy
= f (x, y), y(x0 ) = y0 (1)
dx
Check the condition of f (x, y) such that the initial value problem (IVP) solution of (1) is
guaranteed to be unique.

Example 1
Consider the IVP (
dy y(1 − 2x), x ≥ 0
= f (x, y) = , y(0) = 1 (2)
dx y(2x − 1), x < 0
Now we solve by the variable separation method:
dy
= (1 − 2x)dx, x > 0 (2)
y

54
 dy
= (2x − 1)dx, x < 0 (3)
y
Integrating both (2) and (3):
log y = (x − x2 ) + log C1
2
⇒ y = C1 ex−x , x > 0
2 −x
y = C2 ex ,x<0
From the initial condition at x = 0:

y(0) = 1 ⇒ log y = 0 + log C1

y
log = x − x2
C1
2
C1 = 1, y = ex−x , x ≥ 0
∂y
Here ∂x is not continuous along the y-axis, even then if a unique solution which is continuous.
Consider the IVP: ( 3
4x y
dy x 4 +y 2 , (x, y) 6= (0, 0),
= f (x, y) =
dx 0, (x, y) = (0, 0)
with y(0) = 0. We have:
dy x3 y
= 4 , (x, y) 6= (0, 0)
dx x + y2
(2) can be written as:
(x4 + y 2 )dy = 4x3 y dx

x4 dy − x3 y dx
+ dy = 0 (÷y 2 )
y2
 4
x
d − + dy = 0
y
Integrating we get:

−x4
+ y = 2c2
y
y 2 − x4 = 2cy
y 2 − 2c2 y − x4 = 0.
√
−b ± b2 − 4ac
y=
2a √
2c2 ± 4c4 + 4x4
⇒y=
p 2
2
⇒ y = c ± c4 + x4
⇒ (y − c2 )2 = c4 + x4

for any value of c, y(0) = 0.

Observation: here we note that f (x, y) is continuous and they are infinitely many solutions are
given initial value problem.
dy √
1. Consider the I.V.P dx = f (x, y) = y, y(0) 6= 0 (1)

55
 2. Equation (1) can be written as

dy √
= y y > 0 (2)
dx
dy √
and = −y, y < 0 (3)
dx

Now, separating the variables in Equation (2):

dy
√ = dx
y
Z Z
dy
⇒ √ = dx
y

Integrating, we obtain:
√
2 y = x + 2c
√ x
⇒ y = +c
2
x 2
y= +c is the solution of 2
2

From the initial condition y0 = 0, we have c = 0.

x2
y= is the solution of (1)
4
when y > 0.
For y < 0, we put −y = z.
dy dz
=−
dx dx
dz √
(3) ⇒ dx = z
dz
− √ = dx
z
Integrating, we get: √
−2 z = x + 2c
√ x
− z = +c
2
x 2
z= +c
2
x 2
y=− +c This is a solution of (B)
2
Apply y(0) = 0, we have c = 0.

x2
y=− is the solution of IVP (1)
4
when y < 0. This function f is continuous, but the solution y is not unique. The f (x, y) has
to be continuous on [a, b] if the solution of IVP to exist; but for uniqueness, something more is
needed – some other condition is required.

56
Definition 1.1 The function f (x, y) is said to satisfy the Lipschitz condition with respect to y
in the region D of the xy-plane, if there exists a constant k > 0 such that

|f (x, y1 ) − f (x, y2 )| ≤ k|y1 − y2 |

for all y1 , y2 ∈ D and the constant k is called the Lipschitz constant.

If the function satisfies the Lipschitz condition, then it is continuous with respect to y. That is,

|f (x, y1 ) − f (x, y2 )| ≤ k|y1 − y2 |

for all y1 , y2 ∈ D.
To show that f (x, y) is continuous for any t > 0, we choose δ = k , and then if ky1 − y2 k < δ,
we have:
|f (x, y1 ) − f (x, y2 )| ≤ k|y1 − y2 | < kδ < 
.
Example 1: Let f (x, y) = x2 +2y, and D = R2 .
y

=⇒ |f (x, y1 ) − f (x, y2 )| = |(x2 + 2y1 ) − (x2 + 2y2 )|

= |2y1 − 2y2 | x
= 2|y1 − y2 |.
R2
Therefore, f (x, y) satisfies the lipschitz condi-
tion with respect to y.
Example 2: Let f (x, y) = xy 2 , with the domain |x| ≤ 1 and |y| < ∞.

=⇒ |f (x, y1 ) − f (x, y2 )| = |xy12 − xy22 |

= |x(y12 − y22 )|
= |x(y1 − y2 )(y1 + y2 )|
= |x| · |y1 − y2 | · |y1 + y2 |
≤ |x| · |y1 | + |y2 | · |y1 + y2 |
→ ∞ as y1 , y2 → ∞.

Therefore, f (x, y) is unbounded, and the function does not satisfy the Lipschitz condition in the
given domain D.
Example 3: Let f (x, y) = y 2/3 , with the domain |x| ≤ 1 and |y| ≤ 1.

2/3 2/3
=⇒ |f (x, y1 ) − f (x, y2 )| = |y1 − y2 |
2/3
= |y1 − 02/3 | (if y2 = 0)
2/3
|f (x, y1 ) − f (x, y2 )| |y |
⇒ = 1
|y1 − y2 | |y1 |
1
=
|y1 |1/3
→ ∞ as |y1 | → 0.

57
Therefore, f (x, y) does not satisfy the Lipschitz condition in the given domain.

cos x
y + y 2 , with the domain |y| ≤ 1 and |x − 1| ≤ 12 .


Example 4: Let f (x, y) = x2

cos x cos x
consider, |f (x, y1 ) − f (x, y2 )| = (y1 + y12 ) − 2 (y2 + y22 )
x2 x
| cos x|
= y1 + y12 − y2 + y22
|x2 |
1
≤ 2 |(y1 − y2 )| + |(y1 − y2 )(y1 + y2 )|)
|x |
1
≤ 2 |(y1 − y2 )(1 + y1 + y2 )|
|x |
|y1 − y2 |
≤ · |y1 + y2 + 1|
|x2 |
1
≤ ( )2 · 3 · |y1 − y2 |
2
3
≤ · |y1 − y2 |.
4
Therefore, f (x, y) satisfies the Lipschitz condition in the given region. Therefore, k is 34 .

Example 5: If f (x, y) is continuously differentiable with respect to y on some closed region

R, then f (x, y) is continuous with respect to y on R.
solution: Given f is continuously differentiable on R with respect to y.
Let
∂f (x, y)
K = max
(x,y)∈R ∂y
Consider
Z y1
∂f (x, y)
∀(x, y1 ), (x, y2 ) ∈ R, |f (x, y1 ) − f (x, y2 )| = dy
y2 ∂y
Z y1
∂f (x, y)
≤ dy
y2 ∂y
Since
∂f (x, y)
≤K for all (x, y) ∈ R,
∂y
we get Z y1
∂f (x, y)
dy ≤ K|y1 − y2 |
y2 ∂y
K is a positive constant such that

∂f (x, y)
≤ K ⇒ f (x, y) is Lipschitz with respect to y in the region D.
∂y

Example 6: Consider the IVP:

dy
= x2 + y 2 , y(1) = 3 any D ⊂ R2.
dx

58
 Here, f (x, y) = x2 + y 2 ,
2
∂f ∂2f
= 2y, = 2.
∂y ∂y
f (x, y) is differentiable as its partials are continuous.
From the above result, f (x, y) is Lipschitz w.r.t. y.

Existence and Uniqueness Theorem

Let M be an upper bound of a continuous function f (x, y) in the domain D = {|x − x0 | ≤
a, |y − y0 | ≤ b}, where a and b are real constants and (x0 , y0 ) ∈ D. Further, if f (x, y) satisfies
a Lipschitz condition w.r.t. y and domain D0 : |x − x0 | ≤ h, |y − y0 | ≤ bwhere h = min (a, Mb ),
then there exist a unique solution to the IVP:
dy
= f (x, y), y(x0 ) = y0 , x over |x − x0 | ≤ h.
dx
Ex.1: Consider the IVP:
dy
= x2 + y, y(0) = 1
dx
and the domain of interest is D = {(x, y) : |x| ≤ 1, |y − 1| ≤ 1}.
solution: Here, f (x, y) = x2 + y, and the domain is

−1 ≤ x ≤ 1, 0≤y≤2
Here, W.K.T f (x, y) is continuous over the given domain and

M = max f (x, y) = max(x2 + y)

M = 12 + 2 = |3| = 3.
We have,
|f (x, y1 ) − f (x, y2 )| = |(x2 + y1 ) − (x2 + y2 )| = |y1 − y2 |

⇒ f (x, y) satisfies the Lipschitz condition over the given domain.

Let h = min a, Mb

 
1 1
⇒ h = min 1, = .
3 3
Hence from the existence and uniqueness theorem, the given IVP has a unique solution in
the interval |x| ≤ 13 .

dy
Example 2. Consider the IVP = 5y 4/5 , y(0) = 0. The domain of interest is |x| ≤ 1,
dx
|y| ≤ 1.
solution: Here, f (x, y) = 5y 4/5 and the domain is

−1 ≤ x ≤ 1, −1 ≤ y ≤ 1

f (x, y) is continuous over the given domain D, and

M = max |f (x, y)| = max |5y 4/5 | = 5.

The existence of the solution is guaranteed, but for uniqueness we have to check the Lipschitz
condition.

59
 4/5 4/5
|f (x, y1 ) − f (x, y2 )| |5y1 − 5y2 |
=
|y1 − y2 | |y1 − y2 |
Let y2 = 0 and y1 = δ > 0, then:

|f (x, y1 ) − f (x, y2 )| = |5δ 4/5 − 0| = 5δ 4/5

|y1 − y2 | = δ

|f (x, y1 ) − f (x, y2 )| 5δ 4/5 5

⇒ = = 1/5 → ∞ as δ → 0.
|y1 − y2 | δ δ
Therefore, f (x, y) is unbounded in the given domain D.
It does not satisfy the Lipschitz condition in the given domain.
This indicates that the given IVP has no unique solution.
(or)

dy
= 5y 4/5
dx
⇒ y −4/5 dy = 5dx
Z Z
−4/5
⇒ y dy = 5dx
5
⇒ y 1/5 = 5x + C
1
⇒ y 1/5 = x + C
⇒ y = (x + C)5 is the solution of (1)

When C = 0:
y = x5 so this is also a solution of (1).
Also, y(x) ≡ 0 is also a solution of IVP. Therefore, the solution is not unique.
Example 3: Consider the IVP: y 0 = y 2 , y(1) = −1, the domain of interest is |x − 1| ≤
a, |y − 1| ≤ b, where a, b are real constants.
solution: Here,
f (x, y) = y 2 and domain: − a < x − 1 < a
Since f (x, y) is continuous in the domain −a + 1 ≤ x ≤ a + 1, −b + 1 ≤ y ≤ b + 1, the
given domain D, and

M = max{|f (x, y)|} = max |y 2 | = max |y|2 = max| − (1 + b)2 | = (1 + b)2

So, the solution of the IVP is guaranteed.

Now, we have:

|f (x, y1 ) − f (x, y2 )| = |(y12 − y22 )| = |(y1 − y2 )(y1 + y2 )|

≤ 2(b − 1)|y1 − y2 |.
So f (x, y) satisfies the Lipschitz condition.
Thus, the given IVP has a unique solution on the interval |x − 1| ≤ h, where
   
b b
h = min a, = min a, .
M (1 + b)2

60
Let
b
g(b) =
(1 + b)2

(1 + b)2 · 1 − b · 2(1 + b)
g 0 (b) =
(1 + b)4
(1 + b) [(1 + b) − 2b]
=
(1 + b)4
(1 + b)(1 − b)
=
(1 + b)4
1−b
=
(1 + b)3
g 0 (b) = 0 at b = 1

So, g attains its max and min at b = 1

Now,
(1 + b)3 (−1) − (1 − b) · 3(1 + b)2
g 00 (b) =
(1 + b)6
−(1 + b)3 − 3(1 − b)(1 + b)2
=
(1 + b)6
−(1 + b)2 [(1 + b) + 3(1 − b)]
=
(1 + b)6
2
−(1 + b) [1 + b + 3 − 3b]
=
(1 + b)6
−(1 + b)2 (4 − 2b)
=
(1 + b)6
2b − 4
=
(1 + b)4
−2 −1 −1
g 00 (1) =
4
= 3 = < 0.
2 2 8
Thus, g attains a maximum at b = 1. The maximum value of g is:
1
g(1) =
4
If a ≥ 14 , then:
b
≤ a.
(1 + b)2
 
b 1
h = min a, ≤ .
(1 + b)2 4
If a ≤ 14 , then h < 14
And for b = 1, min(a, 14 ) = 1
4

1 3 5
⇒ The given IVP has a unique solution on the interval |x − 1| < (rest of < x < ).
4 4 4
dy
= y2
dx

61
 dy
= dx
y2
Z Z
dy
= dx
y2
1
− =x+C
y
−1
y= (Equation A)
x+C
From the initial condition (I.C.) y(1) = −1:
−1
y= , put x = 1
x+C
−1
−1 = ⇒ −1(1 + C) = −1 ⇒ −C = 0 ⇒ C = 0
1+C
−1
therefore y = (is the exact solution of equation (1))
x
3 5
It is the unique solution in <x< .
4 4

2 Picard’s Method of Successive Approximations

Consider the IVP:
dy
= f (x, y), (12)
dx
with y(x0 ) = y0 , where f (x, y) is continuous in the domain of interval.
Equation (12) can be written as:

dy = f (x, y) dx

Then integrate both sides: Z y Z x

dy = f (t, y(t)) dt
y0 x0
Z x
y = y0 + f (t, y(t)) dt (13)
x0

Equations (12) and (13) are equivalent, but equation (13) has the advantage over (12) because
it produces successive approximations more easily.
As a start of the solution, take y0 (x) = y0 , so that equation (13) becomes:
Z x
0
y (x) = y0 + f (t, y0 ) dt (14)
x0

Thus, y1 is a better approximation of the solution y(x) than y0 . At the next step, replace y0 by
y1 in equation (3). We get a new approximation y2 (x):
Z x
y2 (x) = y0 + f (t, y1 (t)) dt
x0

At the nth step, we get:

Z x
yn (x) = y0 + f (t, yn−1 (t)) dt
x0

62
 So the functions yn (x), for n = 1, 2, 3, . . . are called “Picard’s approximations", and the
procedure is called “Picard’s method of successive approximations".

Now, the sufficient condition for the sequence of approximations yn (x) to converge to the
exact solution y(x) is given in the following theorem:

Picard’s Existence and Uniqueness Theorem

Let M be an upper bound of continuous function f (x, y) in the domain D : |x−x0 | ≤ a, |y−y0 | ≤
b, where a, b are real constants, (x0 , y0 ) ∈ D. Further, if f (x, y) satisfies the “Lipschitz condition"
w.r.t. y in the domain |x − x0 | ≤ h, |y − y0 | ≤ b, where
 
b
h = min a,
M
then the successive approximation
Z x
yn (x) = y0 + f (t, yn−1 (t)) dt
x0

for n = 1, 2, 3, . . . converges to the unique solution of the IVP.

dy
= f (x, y), y(x0 ) = y0 ,
dx
over the interval |x − x0 | ≤ h.

Example 1: Find the Picard’s 1st, 2nd and 3rd approximations for the IVP:
dy
= x2 + y, y(0) = 1
dx
on the domain |x| ≤ 1, |y − 1| ≤ 1. Also, find a local interval of existence of the solution.

Solution:
Given,
dy
= x2 + y (1)
dx
with I.C.: y(0) = 1
Domain D : −1 ≤ x ≤ 1, 0 ≤ y ≤ 2
Here, f (x, y) = x2 + y

⇒ M = max |f (x, y)| = max |x2 + y| = 3

Since f (x, y) is continuous function, so solution exists.
Equation (1) can be written as:

dy = (x2 + y) dx
Z y Z x
dy = (t2 + y(t)) dt
y0 x0
Z x
⇒ y = y0 + (t2 + y(t)) dt
x0

Successive approximation by Picard method is:

Z x
yn (x) = y0 + (t2 + yn−1 (t)) dt
x0

63
 Given x0 = 0, y0 = 1

1st approximation: Z x
y1 (x) = 1 + (t2 + 1) dt
0
x
t3 x3

y1 (x) = 1 + +t =1+ +x
3 0 3

2nd approximation: Z x
t2 + y1 (t) dt


y2 (x) = y0 +
0
x
t3
Z 
2
=1+ t + + t dt
0 3
x
t3 t4 t2

= t+ + +
3 12 2 0
x3 x4 x2
=1+ +x+ + .
3 12 2
3rd Approximation:
Z x x
t3 t4 t2
 3
t4 t2 t5 t3

2 t
y3 (x) = y0 + t +1+ +t+ + dt = 1 + +t+ + + +
0 3 12 2 3 12 2 60 6 0
Simplifying:
x3 x4 x2 x5 x3
y3 (x) = 1 + +x+ + + +
3 12 2 60 6
Local interval of existence:
 
b 1 1
h = min a, = min(1, ) =
M 3 3

So the interval is |x − x0 | ≤ h ⇒ |x| ≤ 31

Example 2:
dy √
= 2 y, y(0) = 0
dx
√
solution: Here, f (x, y) = 2 y, |x| ≤ 1, |y| ≤ 1,
I.C.: y(0) = 0 f (x, y) is continuous.
Let
√
M = max |f (x, y)| = max |2 y| = 2
f is continuous and has an upper bound.
Check Lipschitz condition:
√ √ √ √
|f (x, y1 ) − f (x, y2 )| |2 y1 − 2 y2 | 2| y1 − y2 | 2
= = √ √ √ √ = √ √ −→ ∞as y1 , y2 −→ 0.
|y1 − y2 | |y1 − y2 | | y1 − y2 |.| y1 + y2 | | y1 + y2 |

Domain contains 0 and y1 , y2 −→ 0.

f (x, y) is unbounded. So, this does not satisfy a Lipschitz condition in any domain including
the origin.
Picard’s successive approximation:
Z x
yn (x) = y0 + f (t, yn−1 (t))dt, n = 1, 2, 3, ....
x0

64
Given x0 = 0, y0 = 0
1st Approximation:
x
√
Z
y1 (x) = y0 + 2 ydt = 0
x0
Z x √
y1 (x) = 0 + 2 0dt = 0
0

2nd Approximation: Z x
y2 (x) = 0 + f (t, 0)dt = 0
0
Similarly,
yn (x) → 0 as n → ∞
So the solution of IVP is zero.
But solving:
dy √
= 2 y ⇒ y −1/2 dy = 2dx ⇒ 2y 1/2 = 2x + 2C ⇒ y 1/2 = x + C ⇒ y = (x + C)2
dx
Applying I.C. y(0) = 0 ⇒ C = 0

y = x2 is a solution of the given IVP.

So solution of given IVP is not unique since we have 2 solutions.

This shows that when the Lipschitz condition fails for the given DE, a unique solution does
not exist.

Eigenvalue Problem
If the problem of determinent y(x) satisfies a differential equation:

a0 (x)y 00 (x) + a1 (x)y 0 (x) + a2 (x)y(x) = λy(x) in [a, b], or Ly = λy,

with ai (x) 6= 0, ∀x ∈ [a, b] and ai (x) (i = 0, 1, 2) are continuous on [a, b], subject to the boundary
conditions:
α1 y(a) + α2 y 0 (a) = 0,
β1 y(b) + β2 y 0 (b) = 0, |α1 | + |α2 | =
6 0, |β1 | + |β2 | =
6 0,
where λ is a parameter, α1 , α2 , β1 , β2 are real function, then it is called an eigenvalue problem.
Note: The eigenvalue problem is a special type of boundary value problem. y(x) = 0
is always a solution to the eigenvalue problem, but it is a trivial solution and therefore it is
discarded.

Sturm-Liouville’s Eigenvalue Problem

If the differential equation L2 y = 0 is in the form:

(p(x)y 0 )0 + (q(x) + λr(x))y = 0,

where p(x), q(x), r(x) are real-valued continuous functions on [a, b] with p(x) > 0, r(x) > 0,
∀x ∈ [a, b], and p(x) is differentiable, with the supplementary boundary conditions:

A1 y(a) + A2 y 0 (a) = 0, B1 y(b) + B2 y 0 (b) = 0,

65
such that real constants A1 , A2 are not both zero, and real constants B1 , B2 are not both zero.
It is said to be a Sturm-Liouville eigenvalue problem.
Now, the values of λ for which non-trivial solution exists to the eigenvalue problem are called
the eigenvalues λ; the corresponding solutions of the eigenvalue problem are called eigenfunc-
tions.
If there are two linearly independent eigenfunctions corresponding to an eigenvalue λ, then
λ is called a double eigenvalue or eigenvalue of multiplicity 2. Otherwise, the eigenvalue λ is
called a simple eigenvalue.
Special cases of supplementary conditions are

y(a) = 0, y(b) = 0, or y 0 (a) = 0, y 0 (b) = 0.

Example 1: Find the Eigenvalues and Eigenfunctions for:

y 00 + λy = 0, y(0) = 0, y(π) = 0. (1)

Given: (p(x)y 0 )0 + (q(x) + λr(x))y = 0 where p(x) = 1, q(x) = 0, r(x) = 1.

Given problem is Sturm-Liouville Problem.
For (1), the A.E. is:
m2 + λ = 0 ⇒ m2 = −λ.
The roots are √
m = ± λi.
. Let us consider the following three cases:
Case I: λ = 0
Then m = 0, i.e., 0 is a double root.

⇒ y(x) = (C1 + C2 x)e0x = (C1 + C2 x),

Applying boundary conditions:

y(0) = 0 ⇒ C1 = 0, y(π) = 0 ⇒ C2 π = 0 ⇒ C2 = 0.

⇒ y(x) = 0 ⇒ trivial solution, so λ = 0 is not an eigenvalue.

Case II: λ < 0
Put λ = −α2 , then the A.E. becomes:

m2 + λ = 0

⇒ m2 − α2 = 0 ⇒ m = ±α.
The solution:
y(x) = C1 eαx + C2 e−αx .
Applying boundary conditions:

y(0) = 0 ⇒ C1 + C2 = 0 ⇒ C1 = −C2 .

y(π) = 0 ⇒ C1 (eαπ − e−απ ) = 0.

⇒ 2C1 sinh(απ) 6= 0
Since α 6= 0, sinh(απ) 6= 0,
⇒ C1 = 0.
The solution is
y(x) = C2 e−απ

66
B.C. y(π) = 0
⇒ 0 = C2 e−απ
since e−απ = 0, we have C2 = 0.

⇒ y(x) = 0 ⇒ trivial solution, λ is not an eigenvalue.

Case III: λ > 0 √

For this case, the roots of m2 + λ = 0 are m = ±i λ = ±in
The solution: √ √
y(x) = C1 cos( λx) + C2 sin( λx).
Applying B.C.:
y(0) = 0 ⇒ C1 = 0.
√
y(π) = 0 ⇒ C2 sin( λπ) = 0.
Assume C2 6= 0, then:
√ √ √
sin( λπ) = 0 ⇒ λπ = nπ ⇒ λ = n ⇒ λ = n2 , n = 1, 2, 3, ...

therefore y(x) = A sin(nx), are the eigenvalues and eigenfunctions.

Therefore, the eigenvalues are called simple (Asin(x), Asin(2x), · · · ).
Ex 2: Find the eigenvalues and eigenfunctions of
 
d 2 dy y
(x + 1) +λ 2 = 0, y(0) = 0, y(1) = 0
dx dx x +1

Given:  
d 2 dy y
(x + 1) +λ =0 (1)
dx dx x2 + 1
y(0) = 0, y(1) = 0 (2)
Equations (1) and (2) constitute a Sturm-Liouville problem.
Here,
1
p(x) = x2 + 1, q(x) = 0, r(x) =
x2 +1
Put t = tan−1 x ⇒ x = tan t
Then,
dy dy dt dy 1 dy
= · = · 2
= · cos2 t
dx dt dx dt 1 + x dt
dy dy
(1 + x2 ) =
dx dt
Substituting this in equation (1),


d dy y
+λ 2 =0
dx dt x +1
 
d dy dt y
⇒ +λ 2 =0
dt dt dx x +1
1 d2 y λ
⇒ + 2 y=0
x2 + 1 dt2 x +1

67
Multiply by (1 + x2 )
d2 y
⇒ + λy = 0
dt2
This is a second-order linear differential equation. Its auxiliary equation is:
√
m2 + λ = 0 ⇒ m = ±i λ

Now, we consider three cases as follows:

Case I: λ = 0 ⇒ m = 0 (a double root)
Then, the solution of (3) is:

y(t) = (C1 + C2 t) ⇒ y(x) = C1 + C2 tan−1 x

Applying B.C: y(0) = 0 ⇒ C1 = 0

Then,
y(x) = C2 tan−1 x
Apply 2nd B.C: y(1) = 0 ⇒ 0 = C2 tan−1 1 = C2 · π
4 ⇒ C2 = 0

⇒ y(x) = 0 (trivial solution)

Case II: λ < 0 ⇒ λ = −α2 , α 6= 0
Then the roots are m = ±α
Solution of (3) is:
−1 −1
y(t) = C1 eαt + C2 e−αt ⇒ y(x) = C1 eα tan x
+ C2 e−α tan x

Apply B.C: y(0) = 0 ⇒ C1 + C2 = 0 ⇒ C2 = −C1

So,  
−1 −1
y(x) = C1 eα tan x − e−α tan x = 2C1 sinh(α tan−1 x)

Apply 2nd B.C: y(1) = 0

π π
⇒ 0 = C1 eα 4 + C2 e−α 4
 π π

⇒ 0 = C1 eα 4 − e−α 4
π
⇒ 0 = 2C1 sinh(α )
4
π
Since α 6= 0 and 4 6= 0 ⇒ C1 = 0. Then with C2 = −C1 gives C2 = 0.
Hence, C1 = 0 ⇒ y(x) = 0 (trivial solution)
Therefore λ is not an eigenvalue
Case III: λ > 0
Then, √
m = ±i λ ⇒ roots are imaginary
The solution of (3) is:
√ √ √ √
y(t) = C1 cos( λt) + C2 sin( λt) ⇒ y(x) = C1 cos( λ tan−1 x) + C2 sin( λ tan−1 x)

Apply B.C: y(0) = 0 ⇒ C1 = 0

Then,
y(x) = C2 sin(µ tan−1 x)
2nd boundary conditions:
y(1) = 0,

68
 Assume the solution: √
y(x) = c2 sin( λ tan−1 x)
√
0 = c2 sin( λ tan−1 1)
√ π 
y(1) = 0 ⇒ 0 = c2 sin λ
4
To get non-trivial solutions, assume c2 6= 0. So:
√ π  √ π √
sin λ = 0 ⇒ λ = nπ ⇒ λ = 4n ⇒ λ = 16n2 , n = 1, 2, 3, . . .
4 4
Hence, λ = 16n2 are the eigenvalues and the eigenfunctions are:

y(x) = A sin(4n tan−1 x), n = 1, 2, 3, . . .

Ex 3: Find the eigenvalues and eigenfunctions of

 
d 1 dy
+ λ(3x2 + 1)y = 0
dx (3x2 + 1) dx
with boundary conditions:
y(0) = 0, y(π) = 0
solution: Given:  
d 1 dy
+ λ(3x2 + 1)y = 0 (1)
dx (3x2 + 1) dx
with boundary conditions:
y(0) = 0, y(π) = 0 (2)
Put
dt
t = x3 + x ⇒ = 3x2 + 1
dx
Then:
dy dy dt dy 1 dy dy
= · = (3x2 + 1) ⇒ 2 =
dx dt dx dt 3x + 1 dx dt
The differential equation becomes:
   
d dy 2 d dy dt
+ λ(3x + 1)y = 0 ⇒ + λ(3x2 + 1)y = 0
dx dt dt dt dx

d2 y
⇒ (3x2 + 1) + λ(3x2 + 1)y = 0
dt2
÷(3x2 + 1), we get
d2 y
⇒ + λy = 0 (3)
dt2
Solving: √
Auxillary equation: m2 + λ = 0 ⇒ m = ±i λ
√
the roots are: m = ±i λ
we consider 3 cases:

69
Case I: λ = 0
Then m = 0, a double root:

y(t) = (c1 + c2 t)e0 = c1 + c2 t ⇒ y(x) = c1 + c2 (x3 + x)

Applying BCs:
1. y(0) = 0 ⇒ c1 + c2 (0) = 0 ⇒ c1 = 0

The solution is : y(x) = c2 (x3 + x)

2. y(π) = 0 ⇒ c2 (π 3 + π) = 0 ⇒ c2 = 0

The solution is : y(x) = 0

So the only solution is trivial. Hence, λ = 0 is not an eigenvalue.

Case II: λ < 0

Let λ = −α2 , α 6= 0, so:
The roots are: m2 = α2 , m = ±α
3 +x) 3 +x)
The solution of (3) is ⇒ y(t) = c1 eαt + c2 e−αt ⇒ y(x) = c1 eα(x + c2 e−α(x
Apply 1st boundary condition:

y(0) = 0 ⇒ c1 e0 + c2 e0 = c1 + c2 = 0 ⇒ c2 = −c1

Apply 2nd boundary condition:

3 +π) 3 +π)
 3 3

y(π) = 0 ⇒ c1 eα(π − c1 e−α(π = 0 ⇒ c1 eα(π +π) − e−α(π +π) = 0

Case III: λ > 0

Auxiliary equation (A.E.) is:
m2 + λ = 0,
The roots are: √
⇒ m = ± λi
The solution of the equation (3) is t is:
√ √
y(t) = c1 cos( λt) + c2 sin( λt)

The solution of equation (1) is:

√ √
y(x) = c1 cos( λ(x3 + x)) + c2 sin( λ(x3 + x))

Applying the first boundary condition:

y(0) = 0 ⇒ c1 cos(0) + c2 sin(0) = c1 · 1 + c2 · 0 = 0 ⇒ c1 = 0

√
⇒ y(x) = c2 sin( λ(x3 + x))
Apply the second boundary condition:
√
y(π) = 0 ⇒ c2 sin( λ(π 3 + π)) = 0

70
 To get a non-trivial solution, assume c2 6= 0, then:
√ √
sin( λ(π 3 + π)) = 0 ⇒ λ(π 3 + π) = nπ, n = 1, 2, 3, . . .

2
√ n2

nπ n n
λ= 3 = 2 ⇒λ= 2
= , n = 1, 2, . . .
π +π π +1 π +1 (π 2 + 1)2
Thus, we get a non-trivial solution. So, λ is an eigenvalue, and the corresponding eigenfunc-
tions are:  
n 3
φn (x) = yn (x) = A sin (x + x) , n = 1, 2, . . .
π2 + 1
Corresponding to the eigenvalue:

n2
λ= .
(π 2 + 1)2

Ex 4:
 
d dy λ
x + y = 0 with boundary conditions : y(1) = 0, y(eπ ) = 0
dx dx x

solution: Given the differential equation:

 
d dy λ
x + y=0 (1)
dx dx x

with boundary conditions:

y(1) = 0, y(eπ ) = 0 (2)
dt
Let t = log x, then x = et and dx = x1 . Then:

dy dy dt 1 dy
= · =
dx dt dx x dt
dy dy
x = ,
dx dt
 
d dy λ
(1) ⇒ + y=0
dx dt x
 
d dy dt λ
( + y=0
dt dt dx x

1 d2 y λ
+ y=0
x dt2 x
÷x, we get

d2 y
+ λy = 0 (3)
dt2
The auxiliary equation is:
m2 + λ = 0
The roots are: √
m = ±i λ.
We consider 3 cases:

71
Case I: λ = 0
Then roots are m = 0, double root. The solution of (3) is:

y(t) = c1 + c2 t ⇒ y(x) = c1 + c2 log x

Apply the boundary condition:

y(1) = 0 ⇒ c1 = 0, y(eπ ) = 0 ⇒ c2 (logeπ ) = 0 ⇒ c2 = 0

⇒ y(x) = 0 is a trivial solution

Case II: λ < 0, put λ = −α2 , α 6= 0

Then roots are m = ±α, and the solution of (3) is:

y(t) = c1 eαt + c2 e−αt

The solution of (1) is:

y(x) = c1 eαlogx + c2 e−αlogx
Apply 1st boundary condition:

y(1) = 0 ⇒ c1 + c2 = 0 ⇒ c2 = −c1

Apply 2nd boundary condition:

y(eπ ) = 0 ⇒ c1 eαπ − e−απ = 0 ⇒ c1 sinh(απ) = 0

Since α 6= 0, sinh(απ) 6= 0 ⇒ c1 = 0 ⇒ c2 = 0.

⇒ y(x) = 0 is a trivial solution, hence λ is not an eigenvalue.

√
Case III: λ > 0, A.E is m2 + λ = 0, the roots are m = ± λi
Solution of (3) is: √ √
y(t) = c1 cos( λt) + c2 sin( λt)
Solution of (1) is: √ √
y(x) = c1 cos( λ log x) + c2 sin( λ log x)
Apply 1st boundary condition:
√
y(1) = 0 ⇒ c1 = 0 ⇒ y(x) = c2 sin( λ log x)

Apply 2nd boundary condition:

√ √ √
y(eπ ) = 0 ⇒ sin( λπ) = 0 ⇒ λπ = nπ ⇒ λ = n, n = 1, 2, . . .

⇒ λ = n2 are the eigenvalues.

and the corresponding eigenfunctions are:

y(x) = φ(x) = A sin(n log x).

Solve the following:

Ex 1: Show that√ the eigenvalues

√ of y 00 + y = 0; y(1) = 0, y(0) + y 0 (0) = 0 are the solution
of the equation tan λ = λ, (λ > 0).

72
 Ex 2: Show that the eigenvalues of y 00 + (x2 +1)
λ
2 y = 0; y(0) = 0, y(1) = 0.
3 0 0
Ex 3: Show that the eigenvalues of (x y ) + λxy = 0; y(1) = 0, y(e) = 0.
Ex 4: Show that the eigenvalues of y 00 + λx y = 0; y(1) = 0, y(eπ ) = 0.

Theorem: The eigenvalues of a SturmâĂŞLiouville problem are simple.

Proof: Consider the SturmâĂŞLiouville eigenvalue problem:
 
d dy
p(x) + q(x)y + λr(x)y = 0 (1)
dx dx

on [a, b] subject to boundary conditions:

α1 y(a) + α2 y 0 (a) = 0 with |α1 | + |α2 | =

6 0

β1 y(b) + β2 y 0 (b) = 0 with |β1 | + |β2 | =

6 0 (2)
Suppose φ1 (x) and φ2 (x) are two eigenfunctions corresponding to the same eigenvalue λ of
(1)-(2).
Since φ1 (x) and φ2 (x) satisfy the boundary conditions:

α1 φ1 (a) + α2 φ01 (a) = 0

α1 φ2 (b) + α2 φ02 (b) = 0

    
α1 φ1 (a) φ2 (a) 0
=
α2 φ01 (a) φ02 (a) 0
Since α1 and α2 are not both zero:

φ1 (a) φ2 (a)
= 0 ⇒ W [φ1 , φ2 ](a) = 0
φ01 (a) φ02 (a)

⇒ W [φ1 , φ2 ](x) = 0 on [a, b]

⇒ φ1 and φ2 are linearly dependent.
Hence, the eigenvalues of the SturmâĂŞLiouville problem are simple.

Self-Adjoint Eigenvalue Problem

Consider the eigenvalue problem:

Ly = λy or a0 (x)y 00 (x) + a1 (x)y 0 (x) + a2 (x)y(x) = λy(x) (1)

where a0 (x) 6= 0, ∀x ∈ [a, b], ai (x), i = 0, 1, 2 and a0 (x), a1 (x), a2 (x) are continuous on
[a, b] and real-valued functions and λ is a parameter.
Subject to boundary conditions:

α1 y(a) + α2 y 0 (a) = 0, |α1 | + |α2 | =

6 0

β1 y(b) + β2 y 0 (b) = 0, |β1 | + |β2 | =

6 0 (2)
This eigenvalue problem is said to be self-adjoint if the relation:
Z b Z b
ḡ(x)Lf (x)dx = f (x)Lḡ(x)dx (3)
a a

holds for all φ continuously differentiable functions in f (x)and g(x) on [a, b], where ḡ denotes
the complex conjugate.

73
 From GreenâĂŹs formula:
Z b Z b
ḡ(x)Lf (x) dx = f (x)L¯∗ g(x) dx + [f, g](b) − [f, g](a) (4)
a a

Eqn (4) reduces to Eqn (3) if

(i) L∗ = L i.e., L is self-adjoint.
(ii) [f, g](b) = [f, g](a)
This 2 conditions are satisfied if:
 
d d
L= p(x) + q(x) (L is self adjoint) (6)
dx dx

i.e.,
Ly = p(x)y 00 (x) + p0 (x)y 0 (x) + q(x)y
and also a0 (x) = p(x), a1 (x) = p0 (x)

[f, g](b) = p(x)(f 0 g − f¯g 0 ) + (p0 (x) − p0 (x))f ḡ

[f, g](a) = p(x)(f 0 g − f¯g 0 ) + (p0 (x) − p0 (x))f ḡ
Hence f (x) and g(x) satisfy the B.C., we have:

α1 f (a) + α2 f 0 (a) = 0

α1 ḡ(a) + α2 ḡ 0 (a) = 0
Since α1 and α2 (6= 0) are not simultaneously zero,

f (a) f 0 (a)
= 0 ⇒ f (a)ḡ 0 (a) − f 0 (a)ḡ(a) = 0
ḡ(a) ḡ 0 (a)
and

f (b)ḡ 0 (b) − f 0 (b)ḡ(b) = 0

This two examples implies that

[f, g](b) = 0 = [f, g](a)

i.e., L in this form is self-adjoint.
Theorem: The eigenvalues of a self-adjoint eigenvalue problem are trivial.
Proof: The eigenvalue problem is self-adjoint if
Z b Z b
ḡ(x)Lf (x) dx = f (x)Lḡ(x) dx (1)
a a
This holds for all φ is continuous and differentiable functions f (x) and g(x) in [a, b].
Let us take f (x) = φ(x), g(x) = φ(x), where φ(x) is an eigenfunction corresponding to the
eigenvalue λ.
Then we have:
Lφ(x) = λφ(x)
Now (1) becomes:
Z b Z b
φ̄(x)Lφ(x) dx = φ(x)L̄φ̄(x) dx
a a

74
 Z b Z b
⇒ φ̄(x)λφ(x) dx = φ(x)λ̄φ̄(x) dx
a a
Z b Z b
⇒λ |φ(x)|2 dx = λ̄ |φ(x)|2 dx
a a
Z b
⇒ (λ − λ̄) |φ(x)|2 dx = 0
a
Rb
Since a |φ(x)|2 dx 6= 0 (as φ(x) is an eigenfunction corresponding to eigenvalue λ), we
conclude:

λ = λ̄
i.e., λ is real.
Hence, all eigenvalues are self-adjoint. This proves that eigenvalues of a self-adjoint eigen-
value problem are real.

Orthogonal Functions
Let f (x), g(x) be the continuous functions defined over [a, b]. Then f (x) and g(x) are said to be
orthogonal over [a, b] if
Z b
f (x)g(x) dx = 0.
a
A function f (x) is said to be normalized over [a, b] if
Z b
|f (x)|2 dx = 1.
a

Orthonormal Set
A set of continuous functions {fn } is said to be orthonormal if every pair of functions is
orthogonal over [a, b] and, furthermore, if every function is normalized over [a, b]; i.e.,
Z b (
0, m 6= n (orthogonality condition)
fm (x)fn (x) dx =
a 1, m = n (normalization condition)

This can be written as: Z b

fm (x)fn (x) dx = δmn .
a
Theorem: The eigenfunctions corresponding to distinct (not necessarily real) eigenvalues
of a self-adjoint eigenvalue problem are orthogonal over the relevant set [a, b].

Proof:
Let φ1 (x), φ2 (x) be two eigenfunctions corresponding to the eigenvalues λ1 and λ2 of an eigen-
value problem over the interval [a, b], with λ1 6= λ2 . Then we have:

Lφ1 (x) = λ1 φ1 (x) (15)

Lφ2 (x) = λ2 φ2 (x) (1)

75
 Now consider the eigenvalue problem to be self-adjoint in [a, b]. If
Z b Z b
g(x)Lf (x) dx = f (x)Lg(x) dx (2)
a a

holds and for all φ are continuously differentiable on [a, b], then since φ1 (x), φ2 (x) are piecewise
continuously differentiable on [a, b], taking f (x) = φ1 (x), g(x) = φ2 (x) in equation (2), we get
Z b Z b
ϕ̄2 (x)Lϕ1 (x) dx = ϕ1 (x)L̄ϕ̄2 (x) dx
a a

Z b Z b
⇒ ϕ̄2 (x)λ1 ϕ1 (x) dx = ϕ1 (x)λ̄2 ϕ̄2 (x) dx
a a
Z b Z b
⇒ λ1 ϕ1 (x)ϕ̄2 (x) dx = λ̄2 ϕ1 (x)ϕ̄2 (x) dx
a a

Z b
⇒ (λ1 − λ2 ) ϕ1 (x)ϕ̄2 (x) dx = 0 (eigenvalues of selfadjoint eigenvalue problem are real)
a

Since λ1 6= λ2 , we have:
Z b
ϕ1 (x)ϕ̄2 (x) dx = 0
a
i.e., the eigenfunctions ϕ1 (x)and ϕ2 (x) are orthogonal to each other.

Eigenfunction Expansion Formula

Let {φn (x) : n = 1, 2, 3, . . .} be a set of eigenfunctions of a self-adjoint eigenvalue problem
defined over [a, b]. Evidently, {φn (x)} is an orthogonal set but need not be normalized.
Let ϕn (x) = kn φn (x), n = 1, 2, 3, . . . be normalized for some constants kn . Then we note
that {ϕn (x)} is an orthonormal set of eigenfunctions.
Now the constants kn can be determined as follows:
Z b Z b Z b
ϕ2n (x) dx = kn2 φ2n (x) dx =1⇒ kn2 φ2n (x) dx = 1
a a a
!
Z b
1
⇒ kn = p dx , n = 1, 2, 3, ...
a φ2n (x)
Let f (x) be any arbitrarily integrable function satisfying suitable conditions on [a, b]. Then
f (x) can be expressed as:
∞
X Z b
f (x) = cn ϕn (x), where cn = f (x)ϕn (x) dx
n=0 a

where {ϕn (x) : n = 1, 2, 3, . . .} is an orthonormal set of eigenfunctions of a self-adjoint

eigenvalue problem, and cn s are Fourier coefficients of f (x), which can be determined as follows.
Now, multiply Eq. (1) by ϕk (x) and integrate between a and b with respect to x, to get:
∞ ∞
Z b Z b X ! Z b
X
f (x)ϕk (x) dx = cn ϕn (x) ϕk (x) dx = cn ϕn (x)ϕk (x) dx
a a n=0 n=0 a

76
 Since Z b
ϕn (x)ϕk (x) dx = ck
a
Thus, we have:
∞ Z
X b 
f (x) = f (t)ϕn (t) dt ϕn (x)
n=0 a

Hence, f (x) can be expressed in terms of orthonormal ϕn ’s.

Example 1: Find the eigenvalues and eigenfunctions of

y 00 + λy = 0 (λ > 0), y(0) = 0, y(π) = 0

Expand the function f in terms of orthonormal eigenfunctions of the eigenvalue problem.

Solution: Given:
y 00 + λy = 0 (1)

y(0) = 0, y(π) = 0 (2)

Auxiliary equation:
m2 + λ = 0 ⇒ m2 = −λ
The roots are: √
m = ±i λ
So the general solution of (1) is:
√ √
y(x) = c1 cos λx + c2 sin λx

Applying 1s t boundary conditions:

y(0) = 0 ⇒ 0 = c1 cos 0 + c2 sin 0 ⇒ c1 = 0

√ √ √
y(π) = 0 ⇒ 0 = c2 sin λπ ⇒ sin λπ = 0 ⇒ λ = n, n = 1, 2, 3, . . .
Let λ be the eigenvalue of the boundary value problem
√ √
sin( λπ) = 0 ⇒ λ = n ⇒ λ = n2 π 2

The eigenvalues are λ = n2 , n = 1, 2, 3, . . . and the corresponding eigenfunctions are

φn (x) = sin(nx), n = 1, 2, 3, . . .

Let ψn (x) = kφn (x), where k is a constant such that ψn becomes orthonormal eigen functions
of the given eigen value problem.
The constant k can be determined by:
1
k = qR , with a = 0, b = π
b
a φ2n (x) dx

1
k = qR
π
0 sin2 (nx) dx

1 1 1
k = qR =q
= q1
π 1−cos2nx 1 sin2nx π
0 2 dx 2 x− 2n 0 2 ((π − 0) − (0 − 0))

77
 r
1 2
k = pπ =
2
π
Hence, r
2
ψn (x) = sin(nx), n = 1, 2, 3, . . .
π
is the orthonormal eigenfunction set of the given eigenvalue problem.
Let
∞
X
f (x) = cn ψn (x)
n=1

Then the coefficients are given by:

Z π
cn = f (t)ψn (t) dt
0

Given f (x) = x,P

we get:
Let x = f (x) = ∞n=1 cn ψn (x)
r Z π
2
where cn = t sin(nt) dt
π 0

Using integration by parts:

Z π  π Z π
t 1
t sin(nt) dt = − cos(nt) + cos(nt) dt
0 n 0 0 n

So,
π
√
r 
2 π(−1)n+1 cosnπ sinnt (−1)n+1
cn = + = 2π ·
π n n2 0 n
Thus,
∞ √
√ X (−1)n+1 2
x= 2π · sinnx
n π
n=1
∞
X (−1)n+1
x=2 sin(nx)
n
n=1

This is the required orthonormal eigen function expansion of x.

Example 2: Expand πx − x2 in terms of orthonormal eigenfunctions of the eigen-
value problem:

y 00 + λy = 0, y(0) = 0, y(π) = 0
A.E. is: m2 + λ = 0. √
The roots are: m = ±i λ.
The general solution is:
√ √
y(x) = C1 cos( λx) + C2 sin( λx)

Applying boundary conditions:

√
y(0) = 0 ⇒ C1 = 0 ⇒ y(x) = C2 sin( λx)

78
 √ √
y(π) = 0 ⇒ C2 sin( λπ) = 0 ⇒ λ = n ⇒ λ = n2
So the eigenfunctions are:

φn (x) = sin(nx), n = 1, 2, 3, . . .

Now constant k can be determined by

r r
1 1 2 2
k = qR dx = qR = ⇒ ψn (x) = sin(nx),
π
φ2n (x)
π
sin2 (nx) dx π π
0 0

are the orthonormal eigen functions of the given eigen value problem.
Now let us assume:
X∞
2
πx − x = cn ψn (x)
n=1

where r Z π
Z π
2
cn = f (t)ψn (t) dt = (πt − t2 ) sin(nt) dt
0 π 0
r  Z π Z π 
2
= π t sin nt dt − t2 sin nt dt
π 0 0

r  π    π Z π  
−cosnt 2 −cosnt
Z
2 π cosnt
= π(t cos nt) − ( (1) dt) − t − 2t dt
π 0 0 n n 0 0 n

sinnt π sinnt π
r  Z π
2 2 (−1)(−1)n n
  
2 (−1)(−1) sinnt
= π +π − π − 2t − dt
π n n 0 n n 0 0 n2

cosnt π
r   
2
= −2
π n2 0
r 
2 (−1)n

1
= −2 − 3
π n3 n
r  √
2 2(−1)n+1

2 2 2
= + 3 = 3 √ ((−1)n+1 + 1)
π n3 n n π
∞ √
2
X ((−1)n+1 + 1) 2 sin nx
⇒ πx − x = 2 √
n=1
n3 π
∞
4 X ((−1)n+1 + 1)
=√ sin nx
π n=1 n3
Example 3: Expand sin x in terms of orthonormal eigenfunctions of the eigenvalue problem:

y 00 + λy = 0, y(0) = 0, y(π) = 0

solution:
⇒ y 00 + λy = 0

79
 A.E. is:
⇒ m2 + λ = 0.
The roots are: √
⇒ m = ±i λ.
General solution is √ √
y(x) = C1 cos λx + C2 sin λx
Applying boundary conditions:
√
y(0) = 0 ⇒ C1 = 0, y(π) = 0 ⇒ C2 sin λπ = 0
λ is the eigenvalue if √
⇒ λπ = nπ ⇒ λ = n2

φn (x) = sin nx, n = 1, 2, 3, . . .

Now constant K can be determined by:
r
1 2
K = qR =
π
φ2n (x) dx π
0

r
2
φn (x) = sin nx are the orthonormal eigenfunctions of the given eigen value problem.
π
Thus, the function sin x can be expressed in terms of orthonormal eigenfunctions φn (x):
∞
X
sin x = cn φn (x)
n=1

Where
r !
Z b Z π
2
cn = f (t)φn (t)dt = sin t sin nt dt
a 0 π
Using identity:
1
cos(A) cos(B) = [cos(A − B) + cos(A + B)]
2
Z π
1 1
=√ [cos((1 − n)t) − cos((1 + n)t)]dt
2π 0 2
sin((1 − n)t) sin((1 + n)t) π
 
1
=√ −
2π 1−n 1+n 0
 
1 sin((1 − n)π) sin((1 + n)π)
cn = √ −
2π 1−n 1+n
∞  
1 X sin((1 − n)π) sin((1 + n)π)
⇒ sin x = − sin x
π 1−n 1+n
n=1

80
Green’s Function Method
Consider the eigenvalue problem:

y 00 + (λ − q(x))y = 0 on a ≤ x ≤ b (1)
with

α1 y(a) + α2 y 0 (a) = 0, |α1 | + |α2 | =

6 0

β1 y(b) + β2 y 0 (b) = 0, |β1 | + |β2 | =

6 0 (2)
where q(x) is a real-valued continuous function on [a, b], λ is a real or complex parameter,
α1 , α2 , β1 , β2 are real constants.
Let φ1 (x, λ) and φ2 (x, λ) be solutions of equation (1) satisfying boundary conditions.

φ1 (a, λ) = 1, φ01 (a, λ) = 0, φ2 (a, λ) = 0, φ02 (a, λ) = 1

Now Green’s function method apply when we try to find the solution φ(x, λ) for the non-
homogeneous differential equation y 00 + (λ − q(x))y = b(x), where b(x) is continuous function
satisfying the boundary conditions.
Now
Z b
φ(x, λ) = G(x, ξ, λ)b(ξ) dξ
a
where the Green’s function is given by:
( −θ
1 (x,λ)θ2 (ξ,λ)
g(λ) , a≤x≤ξ
G(x, ξ, λ) = −θ2 (x,λ)θ1 (ξ,λ)
g(λ) , ξ<x≤b
Here, θ1 (x, λ) = α2 φ1 (x, λ) − α1 φ2 (x, λ), and
θ2 (x, λ) = β2 φ1 (x, λ) − β1 φ2 (x, λ)

α1 α2
g(λ) = = α1 β2 − α2 β1
β1 β2
with

B1 = β1 φ1 (b, λ) + β2 φ01 (b, λ), B2 = β1 φ2 (b, λ) + β2 φ02 (b, λ).

Example 1: Construct the Green’s function for

y 00 + λy = g(x), y(0) = 0, y(π) = 0

Solution: Given:
y 00 + λy = 0, y(0) = 0, y(π) = 0
Here, α1 = 1, α2 = 0, β1 = 1, β2 = 0 √
Auxiliary Equation: m2 + λ = 0 ⇒ m = ±i λ
The solution is: √ √
y = c1 cos λx + c2 sin λx
Let us take: √
√ sin λx
φ1 (x, λ) = cos λx, φ2 (x, λ) = √
λ

81
 Then the two solutions satisfy:

φ1 (0, λ) = 1, φ01 (0, λ) = 0, φ2 (0, λ) = 0, φ02 (0, λ) = 1

To construct the Green’s function, first compute:

√
B1 = β1 φ1 (π, λ) + β2 φ01 (π, λ) = cos λπ

√
sin λπ
B2 = β1 φ2 (π, λ) + β2 φ02 (π, λ)
= √
λ
√
sin λx
θ1 (x, λ) = α2 φ1 (x, λ) − α1 φ2 (x, λ) = − √
λ
√
1 √ √ sin λ(π − x)
θ2 (x, λ) = β2 φ1 (x, λ) − β1 φ2 (x, λ) = √ (sin( λπ − λx)) = − √
λ λ
√
α1 α2 1 0√ sin λπ
g(λ) = = √ sin√ λπ = √
B1 B2 cos λπ λ
λ

Thus:
√
sin λx 1 √
θ1 (x, λ) = − √ , θ2 (x, λ) = √ sin λ(π − x)
λ λ
So, the Green’s function is:
( θ (x,λ)θ (ξ,λ)
− 1 g(λ)2 , 0≤x≤ξ
G(x, ξ, λ) = θ2 (x,λ)θ1 (ξ,λ)
− g(λ) , ξ<x≤π
 sin √λx sin √λ(π−ξ)
√ √
 λ√1 sin √λπ λ
, 0≤x≤ξ


= sin
√ λ √
λ(π−x) sin λξ
 √ √

 λ √ λ
√1 sin λπ
, ξ<x≤π
λ
 √ √
 sin( √
λx) sin( λ(π−ξ))
√ , 0≤x≤ξ
λ sin( λπ)
√ √
G(x, ξ, λ) =
 sin( √
λ(π−x)) sin( λξ)
√ , ξ<x≤π
λ sin( λπ)

is the required Green’s function.

Example 2: Construct the Green’s function for y 00 + λy = x, with boundary conditions:
y(0) = 0, y(1) = 0. Hence, find its solution.
solution:
y 00 + λy = 0 with y(0) = 0, y(1) = 0
General solution of the homogeneous equation:
Auxiliary equation: √
m2 + λ = 0 ⇒ m = ±i λ
The solution is: √ √
y(x) = c1 cos( λx) + c2 sin( λx)
Let us take the fundamental solutions:
√
φ1 (x, λ) = cos( λx)

82
 √
sin( λx)
φ2 (x, λ) = √
λ
Then these satisfy:

φ1 (0, λ) = 1, φ01 (0, λ) = 0, φ2 (0, λ) = 0, φ02 (0, λ) = 1

To construct the Green’s function, let us compute:

√
B1 = β1 φ1 (1, λ) + β2 φ01 (1, λ) = cos( λ)
√
sin( λ)
B2 = β1 φ2 (1, λ) + β2 φ02 (1, λ) = √
λ
1 0√ √
α1 α2 sin( λ)
g(λ) = = √ sin( λ) = √
B1 B2 cos( λ) √ λ
λ
θ1 (x; λ) = α2 φ1 (x; λ) − α1 φ2 (x; λ)
√
sin λ
=− √ x
λ
θ2 (x; λ) = β2 φ1 (x; λ) − β1 φ2 (x; λ)
√ √
sin λ √ √ sin λ x
= √ cos λ x − cos λ · √
λ λ
1 √ √ 1 √
= √ sin( λ − λ x) = √ sin λ(1 − x)
λ λ
The Green’s function is
 sin √λx sin √λ(1−ξ)
 λ 1 · √λ
√ √
, 0≤x≤ξ

√ sin λ

G(x, ξ, λ) = sin
√ λ √
λ(1−x) sin λξ
 √ √

 λ √ λ
√1 sin λ
, ξ<x≤1
λ
Z 1
φ(x, λ) = G(x, ξ, λ) · b(ξ) dξ
0

ξ
√ √
sin λ x sin λ(1 − ξ)
Z
= √ √ ξ dξ
0 λ sin λ
Z 1 √ √
sin λ(1 − x) sin λ ξ
+ √ √ ξ dξ
ξ λ sin λ
√ √
sin λ x
Z ξ √ sin λ (1 − x) 1
Z √
=√ √ sin λ(1 − x) · t dξ + √ √ tsin λt dξ
λ sin λ 0 λ sin λ ξ
√ " √ ! Z ξ √ #
sin λ x cos λ(1 − t) ξ − cos λ(1 − t)
=√ √ t √ |0 − (1) · √ dξ
λ sin λ λ 0 − λ
√  √ !1 Z 1 √ 
sin λ(1 − x)  − cos λt − cos λt 
+ √ √ t √ − (1) · √ dξ
λ sin λ λ ξ
ξ − λ
√ √ √ !ξ
sin λx cos λ(1 − t) sin λ(1 − t)
= √ t √ − √
λ sin λ λ −( λ)2 0

83
 √ √ √ !1
sin λ(1 − x) t(− cos λt) sin λt
+ √ √ + √
λ λ ( λ)2 ξ
√ √ √ !
sin λx cos λ(1 − ξ) sin λ(1 − ξ)
= √ ξ √ +
λ sin λ λ λ
√ " √ √ ! √ √ !#
sin λ(1 − x) cos λ ξ(cos λξ) sin λ sin λξ
+ √ − √ + √ + −
λ λ λ λ λ
√
sin λx x
φ(x, λ) = √ − .
λ sin λ λ
Example 3:Construct the Green’s function for y 00 + 14 y = sin 2x
with y(0) = 0, y(π) = 0.
solution: Given:
1
y 00 + y = 0 (1)
4
with y(0) = 0, y(π) = 0, α1 = 1, α2 = 0, β1 = 1, β2 = 0
The A.E. is:
1 1
m2 + = 0 ⇒ m = ± i
4 2
x x
The solution is: cos 2 , sin 2 are solutions of (1). Let us take:
x x
φ1 (x, λ) = cos , φ2 (x, λ) = sin
2 2
To construct the GreenâĂŹs function, we compute B1 , B2 , g(λ), Θ1 (x, λ), Θ2 (x, λ):

B1 = β1 φ1 (π, λ) + β2 φ01 (π, λ)

π
B1 = cos =0
2
π π
B2 = β1 φ02 (π, λ) + β2 φ2 (π, λ) = 1 · 2 sin = 1 ⇒ 2 sin = 2
2 2
α1 α2 1 0
g(λ) = = =2
B1 B2 0 2
x x
θ1 (x, λ) = α2 φ1 (x, λ) − α1 φ2 (x, λ) = 0 − 1 · 2 sin = − sin
2 2
x x
θ2 (x, λ) = B1 φ1 (x, λ) − B2 φ2 (x, λ) = 1 · 2 cos = 2 cos
2 2
Therefore, Green’s function is
( θ (ξ,λ)θ (ξ,λ)
− 1 g(λ)2 , 0≤x<ξ
G(x, ξ, λ) =
− θ2 (x,λ)θ1 (ξ,λ)
g(λ) , ξ≤x<b
 x
 sin 2 cos 2ξ , 0≤x<ξ
= 2
ξ
 2cos x2 ·2sin 2 , ξ ≤ x < b
2
Exercise: Construct the Green’s function for the eigenvalue problem:
x00 + k 2 x = 0, x(0) = 0, x(π) = 0
where k is a constant.

84
Series Solution of Linear ODEs
Types: Ordinary point, Singular point, Regular singular point
Consider a second order linear ODE:
d2 y dy
2
+ P (x) + Q(x)y = 0 (1)
dx dx
A point x = x0 is said to be an ordinary point of (1) if both P (x) and Q(x) are analytic
at x0 . Otherwise, x = x0 is a singular point.
A singular point of (1), x = x0 , is said to be a regular singular point if both

(x − x0 )P (x) and (x − x0 )2 Q(x)

are analytic at x0 . Otherwise, x = x0 is called an irregular singular point.

Transformation
To obtain points at infinity, transform equation (1) using
1 1
ω= or x =
x ω
As ω → 0 ⇔ x → ∞, we can analyze behavior at infinity.

Example 1: Find the ordinary, regular, and irregular singular points of the DE:

y 00 + y = 0

solution: Given equation in the form:

y 00 + P (x)y 0 + Q(x)y = 0

We identify:
d2 y dy
+0· + y = 0 ⇒ P (x) = 0, Q(x) = 1
dx2 dx
Both functions P (x) and Q(x) are analytic at all points on the x-axis.
Hence, all points on the x-axis are ordinary points.

Point at Infinity
Let ω = x1 or x = 1
ω
Then,  
dy dy dω dy 1
= · = · − 2
dx dω dx dω x
differentiate w.r.t. x
d2 y
   
d dy d 1 dy dω
2
= = − 2· ·
dx dx dx dω x dω dx
 
d dy 1
= −ω 2 · · (− 2 )
dω dω x
d2 y
 
dy
= −ω 2 2 + (−2ω) · (−ω 2 )
dω dω
d2 y d2 y dy
2
= 4ω 2 2 + 2ω 3
dx dω dω

85
 Substituting this in equation (1),

d2 y dy
ω4 2
+ 2ω 3 +y =0
dω dω
Divide by ω 4 :
d2 y 2 dy y
2
+ + 4 =0 (2)
dω ω dω ω
Equation (2) is now in the form:

y 00 + P (ω)y 0 + Q(ω)y = 0

Where:
2 1
P (ω) =, Q(ω) = 4
ω ω
Both P (ω) and Q(ω) are not analytic at ω = 0. Hence, ω = 0 (i.e., x = ∞) is a singular
point of (2) or x = ±∞ is a singular point of (2).
To determine whether it is regular or irregular:
 
2
(ω − ω0 )P (ω) = (ω − 0)P (ω) = ω · = 2 (analytic at 0)
ω
1 1
(ω − ω0 )2 Q(ω) = (ω − 0)2 Q(ω) = ω 2 ·4
= 2 (not analytic at 0)
ω ω
Since Q(ω) is not analytic at ω = 0. The point ω = 0 is an irregular singular point of
equation (2) (or) x = ±∞ is an irregular singular point of equation (1). All the values of x are
regular points and x = ±∞ are irregular singular points of equation (1).
Example 2: Find the ordinary, regular and irregular singular points if any of the D.E.

d2 y dy
x + (1 − x) + ny = 0
dx2 dx
where x is a constant.
solution: Given
d2 y dy
x 2
+ (1 − x) + ny = 0 (1)
dx dx
Divide through by x:
d2 y
 
1−x dy n
+ + y=0
dx2 x dx x
This is in standard form:
1−x n
y 00 + P (x)y 0 + Q(x)y = 0, where P (x) = , Q(x) = (2)
x x
Here, P (x) and Q(x) are not analytic at x = 0, but analytic at all other points.
So x = 0 is a singular point of equation (1), and all other values of x are ordinary points.
For regular or irregular singular point:
consider
 
1−x
(x − 0)P (x) = x = 1 − x (analytic at x = 0)
x
n
(x − 0)2 Q(x) = x2 · = nx (analytic at x = 0),
x
the functions are analytic at x = 0. So, x = 0 is a regular singular point of equation (1).
Now consider the point at infinity. Let x = ω1 ⇒ ω = x1

86
 Then:
dy dy dω 1 dy
= · =− 2
dx dω dx x dω

d2 y
   
d 1 dy dω d 1 dy
= − 2 = − 2
dx2 dx x dω dx dω x dω
After conversion, the equation becomes:

d2 y dy dy
ω4 2
+ 2ω 3 − (ω − 1)(ω 2 ) + nωy = 0
dω dω dω
Simplifying:
d2 y 2 dy (ω − 1) dy n
2
+ − 2
+ 3y = 0
dω ω dω ω dω ω
Let:
ω+1 n
P (ω) = , Q(ω) =
ω2 ω3
Check analyticity:
ω+1 ω+1
(ω − 0)P (ω) = ω · 2
= not analytic at ω = 0
ω ω
n n
(ω − 0)2 Q(ω) = ω 2 · 3 = not analytic at ω = 0
ω ω
Thus, ω = 0 is an irregular singular point. Therefore, x = ±∞ are irregular singular points.

Conclusion
x = 0 is a regular singular point. x = ±∞ are irregular singular points. All other points are
ordinary points.
Example 3: Show that x = ±1 and x = ±∞ are regular singular points of Legendre’s D.E.:

(1 − x2 )y 00 − 2xy 0 + n(n + 1)y = 0

where n is a constant. solution: Given

(1 − x2 )y 00 − 2xy 0 + n(n + 1)y = 0

Dividing by 1 − x2 :

2x 0 n(n + 1)
y 00 − y + y=0 (1)
1 − x2 1 − x2
Here,
−2x n(n + 1)
P (x) = 2
, Q(x) =
1−x 1 − x2
These P (x) and Q(x) are not analytic at x = ±1, but they are analytic at all other points.
So x = ±1 is a singular point of Eq. (1), and all other points of x are ordinary points.
Consider at x = 1, the product f · P (x) becomes:

−2x −(x − 1)2x 2x

(x − 1)P (x) = (x − 1) · = =
1 − x2 (1 − x)(1 + x) 1+x

n(n + 1) (x − 1)2 n(n + 1) −(x − 1)n(n + 1)

(x − 1)2 Q(x) = (x − 1)2 · 2
= =
1−x (1 − x)(1 + x) 1+x
The functions are not analytic at x = 1, so x = 1 is an irregular singular point.

87
 At x = −1, both the products f · P (x) and f 2 · Q(x) are analytic.

(x + 1)(−2x) −2x
(x + 1)P (x) = 2
=
1−x 1−x

(x + 1)2 n(n + 1) (x + 1)n(n + 1)

(x + 1)2 Q(x) = 2
=
1−x 1−x
So both the products are analytic at x = −1. Therefore, x = −1 is a regular singular point.

The Point at Infinity

Let w = x1 , i.e., x = 1
w. Then:

dy dy dw dy
= · = −w2
dx dw dx dw

d2 y
   
d 2 dy 2 d dy dy dw
= −w = −w − 2w3
dx2 dx dw dx dw dw dx
Substitute into Eq. (1):

d2 y 3 dy (−w2 ) dy n(n + 1)ω 2

w4 + 2w − 2w · · + y=0
dw2 dw w2 − 1 dw w2 − 1
Simplifying:
d2 y 2 dy 2 dy n(n + 1)
2
+ + 2
+ 2 2 y=0
dw w dw w(w − 1) dw w (w − 1)

d2 y
 
2 2 dy n(n + 1)
2
+ + 2
+ 2 2 y = 0 (2)
dw w w(w − 1) dw w (w − 1)

d2 y 2ω dy n(n + 1)
+ + y=0
dw2 (w2 − 1) dw w2 (w2 − 1)
Since w = 0 is a singular point.
Note:
2w n(n + 1)
P (w) = , w2 Q(w) =
−1 w2 w2 − 1
Both these functions are analytic at w = 0, so w = 0 is a regular singular point of Eq. (2)
(or) x = ±∞ is a regular singular point of Eq. (1).
Therefore, x = ±1 and x = ±∞ are regular singular points of LegendreâĂŹs D.E., and all
other points are ordinary.
Example 4: Find the ordinary, regular, and irregular singular points of D.E.

d2 y dy
(1 − x2 )
2
−x + n2 y = 0
dx dx
where n is a constant (Chebyshev D.E.)
solution: Given:
(1 − x2 )y 00 − xy 0 + n2 y = 0 (1)
In the normalized form, eq (1) is written as:

x n2
y 00 − y 0
+ y=0 (2)
1 − x2 1 − x2
Here,
−x n2
P (x) = , Q(x) =
1 − x2 1 − x2

88
 Both these functions are analytic at all points except x = ±1.
Therefore x = ±1 are singular points of (2)
Take x0 = 1:
−x −x x
(x − 1)P (x) = (x − 1) · 2
= =
1−x (1 − x)(1 + x) (1 + x)

n2 n2 (x − 1)2 n2 (x − 1)2 n2 (1 − x)
(x − 1)2 Q(x) = (x − 1)2 · = = =
1 − x2 (1 − x2 ) (1 − x)(1 + x) 1+x
The above functions are analytic at x = 1.
For x0 = −1:
−x −x(x + 1) x
(x + 1)P (x) = (x + 1) · 2
= =
1−x (1 − x)(1 + x) x−1

n2 n2 (x + 1)2 n2 (x + 1)
(x + 1)2 Q(x) = (x + 1)2 · = =
1 − x2 (1 − x)(1 + x) 1−x
The above functions are analytic at x = −1.
Thereforex = ±1 are regular singular points of eq (2) and all other points are ordinary points.
The point at infinity:
Let x = ω1 or ω = x1
Then,
dy dy d2 y 2
2d y dy
= −ω 2 , 2
= ω 2
+ 2ω 3
dx dω dx dω dω
Substituting this in eq (2):
1
d2 y n2
   
4 3 dy ω 2dy
ω + 2ω − 1 −ω +
y =0
dω 2 dω 1− dω 1 2
ω 1− ω

d2 y 3 dy ω 3 dy n2 ω 2
⇒ ω4 + 2ω + + y=0
dω 2 dω ω 2 − 1 dω ω 2 − 1

d2 y n2
 
2 1 dy
⇒ + + + y=0 (3)
dω 2 ω ω(ω 2 − 1) dω ω 2 − 1
Here,
2ω 2 − 1 n2
P (ω) = , Q(ω) =
ω(ω 2 − 1) ω 2 (ω 2 − 1)
ω = 0 is a singular point of (3). The functions P (ω) and Q(ω) are not analytic at ω = 0.

(2ω 2 − 1) (2ω 2 − 1)
(ω − 0)P (ω) = ω · =
ω(ω 2 − 1) ω2 − 1
n2 n2
(ω − 0)2 Q(ω) = ω 2 · =
ω 2 (ω 2 − 1) ω2 − 1
Both these functions are analytic at ω = 0, so ω = 0 is a regular singular point of (3).
Thus, x = ±∞ are the regular singular points of (1).

⇒ x = ±1 and x = ±∞ are regular singular points of eq (1) and all other points are ordinary points.

Example 5: Find the ordinary, regular singular points (if any) of the equation

x(1 − x)y 00 + [c − (a + b + 1)x]y 0 − aby = 0

89
where a, b, c are constants.
solution:Given:
x(1 − x)y 00 + [c − (a + b + 1)x]y 0 − aby = 0 (1)
This equation is Gauss-Hypergeometric equation.
Divide eq (1) by x(1 − x):

[c − (a + b + 1)x] 0 ab
y 00 + y − y=0 (2)
x(1 − x) x(1 − x)
Here,
[c − (a + b + 1)x] −ab
P (x) = , Q(x) =
x(1 − x) x(1 − x)
Both these functions are analytic at all the points and x = 1 except at x = −1

Therefore x = −1 is a singular points of eq (1) and all other values of x are ordinary points.

These functions are not analytic at x = 0 and x = 1, so x = 0, 1 are singular points of eq (1).
At x = 0:  
c − (a + b + 1)x c − (a + b + 1)x
(x − 0)P (x) = x =
x(1 − x) 1−x
 
−ab −abx
(x − 0)2 Q(x) = x2 =
x(1 − x) 1−x
At x = 0, the functions (x − 0)P (x) and (x − 0)2 Q(x) are analytic.
Therefore x = 0 is a regular singular point of eq (2).
At x = 1:
   
c − (a + b + 1)x c − (a + b + 1)x
(x − 1)P (x) = (x − 1) =−
x(1 − x) x
 
−ab ab(x − 1)
(x − 1)2 Q(x) = (x − 1)2 =
x(1 − x) x
Both these functions are analytic at x = 1. Therefore x = 1 is a regular singular point of eq (2).
The point at infinity
We put x = ω1 or ω = x1
Then,  
dy dy dω dy 1 dy
= · = · − 2 = −ω 2
dx dω dx dω x dω
d2 y
   
d dy d dy dω
2
= = −ω 2 ·
dx dx dx dω dω dx
2
 
dy d y
− ω 2 2 · −ω 2


= −2ω
dω dω
d2 y dy
= ω4 + 2ω 3
dω 2 dω
Substitute this in eq (2):
" # 
1
d2 y c − (a + b + 1) ·

4 3 dy ω 2 dy aby
ω + 2ω + 1 1 · −ω − 1 =0
dω 2 dω ω (1 − ω )
dω ω (1 − ω1 )
Simplifying:
d2 y 3 dy (cω − (a + b + 1)ω) dy abω 2 y
ω4 + 2ω − · − =0
dω 2 dω (ω − 1) dω ω−1

90
 d2 y 2 dy (cω − (a + b + 1)ω) dy aby
+ − · − =0
dω 2 ω dω ω 3 (ω − 1) dω ω 2 (ω − 1)

d2 y
 
2 (cω − (a + b + 1)ω) dy aby
2
+ − 3
− 2 =0
dω ω ω (ω − 1) dω ω (ω − 1)

d2 y −(cω − (a + b + 1)) + 2(ω − 1) dy ab

2
+ − 2 y = 0, (3)
dω ω(ω − 1) dω ω (ω − 1)
(2 − c)ω + (a + b − 1) −ab
here P (ω) = , Q(ω) = 2 .
ω(ω − 1) ω (ω − 1)
Thus, ω = 0 is a singular point of Eq. (3). The functions P (ω) and Q(ω) are singular points
at ω = 0.

ω[(2 − c)ω + (a + b) − 1] [(2 − c)ω + (a + b) − 1]

ωP (ω) = = ,
ω(ω − 1) (ω − 1)
−ab
ω 2 Q(ω) = .
ω−1
The functions P (x) and Q(x) are analytic. Therefore, ω = 0 is a regular singular point of
Eq. (3), (or) x = ±∞ are the regular singular points of Eq. (2). Also, x = 0, 1, ∞ are regular
singular points of the Gauss hypergeometric Eq. (2) and all other points are ordinary points.
Exercise: Find the ordinary & singular points of the following equations

1. y 00 + xy 0 + (x2 + 2)xy = 0.

2. (x − 1)y 00 + xy 0 + xy = 0.

3. (x2 − 1)y 00 + 3xy 0 + xy = 0.

4. x2 y 00 + xy 0 + (x2 − n2 )y = 0.

5. (x − 1)3 y 00 + 3xy 0 + xy = 0.

Power series solution about ordinary point

Given equation y 00 + P (x)y 0 + Q(x)y = 0 · · · (1)
If x = x0 is an ordinary point of (1), then every solution of equation (1) can be represented
by a power series about the point x0 .
∞
X
y(x) = an (x − x0 )n · · · (2)
n=0

This series converges for some interval 0 < |x − x0 | < R, where R > 0 is called the radius of
convergence of the power series about the point x0 .
Example 1: Find the power series solution of the initial value problem

d2 y dy
(x2 − 1) 2
+ 3x + xy = 0 (1)
dx dx
Subject to: y(0) = 4, y 0 (0) = 6 · · · (2)
solution: Given
d2 y 3x dy x
2
+ 2 + 2 y=0
dx x − 1 dx (x − 1)

91
 Here P (x) = x23x−1 , Q(x) = (x2x−1)
These functions are not analytic at x = ±1, i.e., x = ±1 are the singular points and all other
points are ordinary points of equation (1)
Let x0 = 0, an ordinary point of (2).
Assume the power series solution of (2) in the form:
∞
X
y(x) = an (x − xn0 )
n=0
∞
X
y(x) = an xn (3)
n=0
Then the derivatives of (3) are:
∞
X ∞
X
y 0 (x) = nan xn−1 , y 00 (x) = n(n − 1)an xn−2
n=1 n=2

Substituting into the differential equation:

(x2 − 1)y 00 + 3xy 0 + xy = 0
We compute each term:
∞
X ∞
X ∞
X
(x2 − 1) n(n − 1)an xn−2 + 3x nan xn−1 + x an xn = 0
n=2 n=1 n=0
∞
X ∞
X ∞
X ∞
X
n(n − 1)an xn − n(n − 1)an xn−2 + 3 nan xn + an xn+1 = 0
n=2 n=2 n=1 n=0
∞
X ∞
X ∞
X ∞
X
n n n
n(n − 1)an x − (n + 2)(n + 1)x + 3nan x + a0 x + an−1 xn = 0
n=2 n=0 n=1 n=1
∞
X ∞
X
n(n − 1)an xn − 2a2 − 6a3 x − (n + 2)(n + 1)an+2 xn + 3a1 x
n=2 n=2
∞
X ∞
X
n
+ 3nan x + a0 x + an−1 xn = 0
n=2 n=1

∞
X
−2a2 − 6a3 x + 3a1 x + a0 x + [n(n − 1)an − (n + 1)(n + 2)an+2 + 3nan + an−1 ] xn = 0
n=2

Equating the coefficients of like powers of x

−2a2 = 0 ⇒ a2 = 0
−6a3 + 3a1 + a0 = 0 ⇒ 3a1 + a0 = 6a3
3a1 + a0 a1 a0
⇒ a3 = = +
6 2 6
Equating xn terms, we get

n(n − 1)an − (n + 1)(n + 2)an+2 + 3nan + an−1 = 0

(n2 + 2n)an + an−1 − (n + 2)(n + 1)an+2 = 0

(n2 + 2n)an + an−1

⇒ an+2 = , n≥2
(n + 2)(n + 1)

92
Using Initial Conditions
Given: a0 = 4, a1 = 6

a2 = 0
(12 + 2 · 1)a1 + a0 3·6+4 22 11
a3 = = = =
3·2 6 6 3
(22 + 2 · 2)a2 + a1 8·0+6 6 1
a4 = = = =
4·3 12 12 2
(32 + 2 · 3)a3 + a2 15 · 11
3 + 0 55
a5 = = =
5·4 20 4
2
(4 + 2 · 4)a4 + a3 24 · 12 + 11
3 12 + 113
47
47
a6 = = = = 3 =
6·5 30 30 30 90
Substitute these values in equation (3), y = a0 + a1 x + a2 x2 + · · ·

a0 x3 a0 x5 a0 x6
y(x) = a0 + a1 x + + + + ···
 2 6 8 180
x x4 x5 3 a1 x6
+ a1 x + + + a1 x6 + a1 x5 + + ···
6 8 12 8 12
   
1 3 1 5 1 6 1 3 1 4 3 5 1 6
y(x) = a0 1 + x + x + x + · · · + a1 x + x + x + x + x + · · ·
6 8 180 2 12 8 12

Differentiate with respect to x:

 
0 1 2 1 4 1 5
y (x) = a0 (3x ) + (5x ) + (6x ) + · · ·
6 8 180
 
1 2 1 3 3 4 1 5
+ a1 1 + (3x ) + (4x ) + (5x ) + (6x ) + · · ·
2 12 8 12

From the initial conditions: y(0) = 4, y 0 (0) = 6, we have a0 = 4, a1 = 6.

Therefore, the power series solution of the given initial value problem is:
 
1 3 1 5 1 6
y(x) = 4 1 + x + x + x + ···
6 8 180
 
1 3 1 4 3 5 1 6
+ 6 x + x + x + x + x + ···
2 12 8 12

11 3 1 4 11 5 47 6
y(x) = 4 + 6x + x + x + x + x + ···
3 2 4 90
Example 2: Find the series solution of the equation

y 00 + xy 0 + (x2 + 2)y = 0

about the ordinary point x = 0.

solution: Given
y 00 + xy 0 + (x2 + 2)y = 0 (1)
It is of the form: y 00 + P (x)y 0 + Q(x)y = 0 where P (x) = x, Q(x) = x2 + 2

x0 = 0 is an ordinary point of equation (1)

93
 We assume that
∞
X ∞
X
y(x) = an (x − x0 )n = an xn (2)
n=0 n=0
is a power series.
Solution of equation (1):
Differentiate equation (2) with respect to x twice:
∞
X ∞
X
y0 = an nxn−1 , y 00 = an n(n − 1)xn−2
n=1 n=2

Substitute into equation (1):

∞
X ∞
X ∞
X
an n(n − 1)xn−2 + x an nxn−1 + (x2 + 2) an xn = 0
n=2 n=1 n=0

∞
X ∞
X ∞
X ∞
X
an n(n − 1)xn−2 + an nxn + an xn+2 + 2 an xn = 0
n=2 n=1 n=0 n=0
∞
X ∞
X ∞
X ∞
X
(n + 2)(n + 1)an+2 xn + nan xn + an−2 xn + 2 an xn = 0
n=0 n=1 n=2 n=0
∞
X
⇒ 2a2 + 6a3 x + a1 x + 2a0 + 2a1 x + (n + 2)(n + 1)an+2 xn
n=2
∞
X ∞
X ∞
X
n n
+ nan x + an−2 x + 2 an xn = 0
n=2 n=2 n=2

∞
X
⇒ 2(a0 + a2 ) + (3a1 + 6a3 )x + [(n + 1)(n + 2)an+2 + (n + 2)an + an−2 ] xn = 0
n=2

Equating the coefficients of like powers of x

2(a0 + a2 ) = 0 ⇒ a2 = −a0
1
3(a1 + 2a3 ) = 0 ⇒ a3 = − a1
2
Equating xn terms, we get

(n + 1)(n + 2)an+2 + (n + 2)an + an−2 = 0

((n + 2)an + an−2 )

⇒ an+2 = − , n≥2
(n + 1)(n + 2)

94
  
4a2 + a0 1 1 1 1
n = 2, a4 = − =− a0 − a2 = − a0 + a0
12 12 3 12 3
1
= a0 , [a2 = −a0 ]
4

 
5a3 + a1 1 a1
n = 3, a5 = − = − a3 −
20 4 20
1 1 a1 1 a1
= − (− a0 ) − = a1 −
4 2 20 8 20
5a1 − 2a1 3a1
a5 = =
40 40

 
6a4 + a2 1 a2
n = 4, a6 = − = − a4 −
30 5 30
   
1 a0 1 a0
a6 = − (1/4a0 ) + =− a0 −
5 30 20 30
 
−3a0 + 2a0 a0
a6 = =−
60 60

 
7a5 + a3 7 a3
n = 5, a7 = − = − a5 −
42 6 42
   
1 3a1 1
a7 = − + a1
6 40 84
 
3a1 a1
= − +
240 84
1
a7 = − a1
1680

Substituting these in the assumed series solution:

y(x) = a0 + a1 x + a2 x2 + a3 x3 + a4 x4 + . . .
1 1
y(x) = a0 (1 − x2 + x4 − x6 + . . .)
4 60
x3
 
3 5 1 7
+ a1 x − + x − x ...
2 40 1680
y(x) = a0 y0 (x) + a1 y1 (x) (3)
1 1
where y0 (x) = 1 − x2 + x4 − x6 + . . .
4 60
x3 8 5 1 7
y1 (x) = x − + x − x ...
2 40 1680
Eq. (3) is the general solution of given eq.

Example 3: Find the power series of Legendre’s D.E.

solution: The Legendre D.E. is given by:

(1 − x2 )y 00 − 2xy 0 + p(p + 1)y = 0 (1)

where p is a constant.

95
 To find singular points of the equation:
In the normal form, Eq. (1) is:

2x 0 p(p + 1)
y 00 − y + y=0 (2)
1 − x2 1 − x2
x = ±1 are the singular points of (2).
Let x0 = 1,
−2x −2x(x − 1) 2x
(x − 1) · 2
= =
1−x (1 − x)(1 + x) 1+x
is analytic at x0 = 1.
Similarly, for x0 = −1,

−2x −2x(x + 1) 2x
(x + 1) · 2
= =
1−x (1 − x)(1 + x) x−1

is analytic at x = −1.
And
p(p + 1) (x − 1)2 · p(p + 1) (x − 1)2 · p(p + 1) (1 − x) · p(p + 1)
(x − 1)2 · 2
= 2
= =
1−x (1 − x ) (1 − x)(1 + x) 1+x

is analytic at x = 1,

p(p + 1) (1 + x) · p(p + 1)
(x + 1)2 · 2
=
1−x 1−x
is analytic at x = −1.
Hence, x = ±1 are regular singular points of Eq. (1).
We expect a solution of Eq. (1) in the form:
∞
X
y(x) = an xn = a0 + a1 x + a2 x2 + . . . (3)
n=0

About the point x = 0,

∞
X ∞
X
y 0 (x) = nan xn−1 , y 00 (x) = n(n − 1)an xn−2
n=1 n=2

Substituting these into Eq. (1), we get:

∞
X ∞
X ∞
X
2 n−2 n−1
(1 − x ) n(n − 1)an x − 2x nan x + p(p + 1) an xn = 0
n=2 n=1 n=0

Now simplifying:
∞
X ∞
X ∞
X ∞
X
n−2 n n
n(n − 1)an x − n(n − 1)an x − 2nan x + p(p + 1) an xn = 0
n=2 n=2 n=1 n=0

Combine like powers:

∞
X
[(n + 2)(n + 1)an+2 − n(n − 1)an − 2nan + p(p + 1)an ]xn = 0
n=0

⇒ 2a2 + 6a3 x − 2a1 x + p(p + 1)a1 x + p(p + 1)a0

96
 ∞
X ∞
X ∞
X ∞
X
+ (n + 2)(n + 1)an+2 xn − n(n − 1)an xn − 2nan xn + p(p + 1) an xn = 0.
n=2 n=2 n=2 n=2

⇒ 2a2 + (6a3 − 2a1 + p(p + 1)a1 )x + p(p + 1)a0

∞
X
+ [(n + 2)(n + 1)an+2 − n(n − 1)an − 2nan + p(p + 1)an ] xn = 0.
n=2
equating the co-efficients of different power to zero
p(p + 1)
p(p + 1)a0 + 2a2 = 0 ⇒ a2 = − a0
2

6a3 − 2a1 + p(p + 1)a1 = 0

2a1 − p(p + 1)a1 (2 − p(p + 1))a1 −(p − 1)(p + 2)
a3 = = = a1
6 6 6

(n + 1)(n + 2)an+2 − n(n − 1)an − 2nan + p(p + 1)nan = 0.

 
n(n − 1)an − 2nan + p(p + 1)nan
an+2 =− .
(n + 1)(n + 2)
 
n(n + 1)an − p(p + 1)an
=− .
(n + 1)(n + 2)
(p − n)(p + n + 1)
an+2 = − an ; n ≥ 2.
(n + 1)(n + 2)
This recursion formula enables us to write the coefficients in terms of a0 , a1 .
For n = 2,
 
(p − 2)(p + 3) p − 2)(p + 3) P (p + 1)
a4 = − a2 = a0
12 12 2
(p − 2)(p + 3)p(p + 1)
= a0
24
p(p + 1)(p − 2)(p + 3)
a4 = a0 .
4!
For n = 3:
(p − 3)(p + 4) (p − 3)(p + 4)(p − 1)(p + 2)
a5 = − a3 = a1
20 20 · 6
(p − 1)(p + 2)(p − 3)(p + 4)
= a1
5!
For n = 4:
(p − 4)(p + 5) p(p + 1)(p − 2)(p + 3)
a6 = − · a0
30 4!
p(p + 1)(p − 2)(p + 3)(p − 4)(p + 5)
=− a0
6!
For n = 5:
(p − 1)(p − 3)(p − 5)(p + 2)(p + 4)(p + 6)
a7 = − a1
7!
and so on.

97
 Substituting this coefficient in the assumed solution:
X
y(x) = an xn

 
p(p + 1) 2 p(p + 1)(p − 2)(p + 3) 4 (p − 4)(p + 5)p(p + 1)(p − 2)(p + 3) 6
y(x) = a0 1 − x + x − x + ···
2! 4! 6!

(p − 1)(p + 2) 3 (p − 1)(p + 2)(p − 3)(p + 4) 5 (p − 1)(p − 3)(p − 5)(p + 2)(p + 4)(p + 6) 7
+a1 x − x + x − x +·
3! 5! 7!
This is the general solution in the solution about a regular singular point.

The Method of Frobenius

If x = x0 is a regular singular point of a differential equation

y 00 + P (x)y 0 + Q(x)y = 0 (1)

then the DE has at least one solution of the form

∞
X
c
y(x) = (x − x0 ) an (x − x0 )n (2)
n=0

where c is a real constant or a complex constant. Determine a modified series is convergent to

R, 0 < |x − x0 | < R (R > 0) about x0 .
According to this method, the solution of equation (2) is substituted into (1), and the
coefficient of the lowest power of x is equated to zero to obtain the quadratic equation in c,
which is called the indicial equation in c, call the D.E. (1). While equating the hightest power
of x for each value of c to 0, gives the co-efficient an and the values of c can be determined and
nature of the roots gives the solution.
Let c1 and c2 be two roots of the indicial equation of DE (1).
Case i: c1 6= c2 , c1 − c2 6= N , an integer
Then the given DE has linearly independent solutions in the form:
∞
X
c1
y1 (x) = (x − x0 ) an (x − x0 )n (a0 6= 0) (3)
n=0
∞
X
y2 (x) = (x − x0 )c2 a∗n (x − x0 )n (a∗0 6= 0) (4)
n=0
Case ii: c1 6= c2 , c1 − c2 = N , a positive integer
Then the DE has 2 linearly independent solutions of the form:
∞
X
y1 (x) = (x − x0 )c1 an (x − x0 )n (a0 6= 0)
n=0
∞
X
y2 (x) = (x − x0 ) c2
a∗n (x − x0 )n + AyN (x) log(x − x0 ) (a∗0 6= 0), A is a constant
n=0
Case iii: c1 = c2 = c
Then the DE has 2 linearly independent solutions in the form:
∞
X
c
y1 (x) = (x − x0 ) an (x − x0 )n (a0 6= 0)
n=0

98
 ∞
X
y2 (x) = (x − x0 )c+1 a∗n (x − x0 )n + yN (x) log(x − x0 ) (a∗0 6= 0)
n=0

Example 1: Find the series solution of 2x2 y 00 + xy 0 + (x2 − 3)y = 0 about a regular singular
point.
solution: We apply Frobenius method to find the solution.
Given
1 0 x2 − 3
y 00 + y + y=0
2x 2x2
Let x = 0 be a singular point of (1).
1
Consider (x − x0 ) × P (x) = (x − 0) × 2x which is analytic.

x2 − 3 x2 − 3
(x − x0 )2 × Q(x) = x2 × = which is analytic
2x2 2
Therefore, x = 0 is a regular singular point of equation (1). We now assume a Frobenius
series solution of equation (1) in the form:
∞
X
y = (x − x0 )c an (x − x0 )n
n=0

i.e.,
∞
X ∞
X
c n
y(x) = x an x = an xn+c → (2)
n=0 n=0
Now, differentiate the differential equation with respect to x:
∞
X
0
y (x) = an (n + c)xn+c−1
n=0

∞
X
00
y (x) = (n + c)(n + c − 1)an xn+c−2
n=0

Substitute this into differential equation (1):

∞
X ∞
X
2x2 (n + c)(n + c − 1)an xn+c−2 + x (n + c)an xn+c−1
n=0 n=0

∞
X
2
+(x − 3) an xn+c = 0
n=0
∞
X ∞
X
⇒ (n + c)(n + c − 1)an xn+c + (n + c)an xn+c
n=0 n=0
∞
X ∞
X
+ an xn+c+2 − 3 an xn+c = 0
n=0 n=0
∞
X ∞
X
n+c
⇒ [(2(n + c)(n + c − 1) + (n + c) − 3)] an x + an−2 xn+c = 0
n=0 n=2

⇒ (2c(c − 1) + c − 3)a0 xc + (2(c + 1)c + (c − 2)a1 xc+1

 

∞
X
+ [2(n + c)(n + c − 1)an + (n + c) − 3] an + an−2 xn+c = 0 → (3)
n=2

99
 Compare the lowest power of xc .
Now equating to zero coefficient of lowest power of x in (3), we get:

(2c(c − 1)(c − 3))a0 = 0 (c : a0 6= 0)

⇒ 2c2 − 2c + c − 3 = 0

2c2 − c − 3 = 0

(c + 1)(2c − 3) = 0

3
c = −1, c=
2
This values of c are called exponents.
Here the exponents are distinct & do not differ by integer F.M.there exist 2 L.I solution,
corresponding to two distinct roots.
Let c1 = −1 and c2 = 32 .
Equating to zero the coefficient of xc+1 , and xc+n in eq. (3), we get:

(2c(c + 1) + (c − 2))a1 = 0 (4)

(2(c + n) + (c + n − 1) + (n + c − 3))an + an−2 = 0 n ≥ 2 (5)

Case 1: c1 = −1
Eq. (4):

−3a1 = 0 ⇒ a1 = 0
Eq. (5):

2(n − 1)(n − 2) + (n − 4)an + an−2 = 0 (6) n ≥ 2

1
n = 2in (6) : −2a2 + a0 = 0 ⇒ a2 = a0
2
1
n=3: 3a3 + a1 = 0 ⇒ a3 = − a1 = 0
3
1
n=4: 12a4 + a2 = 0 ⇒ a4 = − a0
24
1 1
n=5: 25a5 + a3 = 0 ⇒ a5 = − a3 = − a1 = 0
25 75
P∞ n−1 :
Putting these in y(x) = n=0 an x

y(x) = a0 x−1 + a0 x0 + a1 x1 + a2 x2 + a3 x3 + a4 x4 + a5 x5 . . .

1 x x3
 
y(x) = a0 + − + ...
x 2 24
3
Letting c = 2 in (4) & (5),

100
 3 3 3
(2 × ( + 1) + ( − 2))a1 = 0
2 2 2
⇒ a1 = 0
 
3 3 3
⇒2 + n − 1 ( + n) + ( + n − 3)an = −an−2
2 2 2
 
2n + 1 (2n + 3) (2n − 3)
2 + an = −an−2
2 2 3

(2n + 1)(2n + 3) + (2n − 3))an = −2an − 2

(4n2 + 2n + 6n + 3 + 2n − 3)an = −2an−2

2n(2n + 5)an = −2an−2

(or)
an−2
an = − , n ≥ 2.
n(2n + 5)
For n = 2, in (6),
a0
a2 = −
18
n = 3,
a1
a3 = − = 0 (since a1 = 0)
3 · 11
n = 4:
a2 −a0 a0
a4 = − = =−
4 · (13) 18 × 4 × 13 936
n = 5:
a3 a1
a5 = − =− =0
5 · 15 33 × 5 × 15
n = 6:
a4 −a0 a0
a6 = − = =−
6 · 17 936 × 6 × 17 954722
P∞ n+c , 3
Substitute these in y(x) = n=0 an x c= 2
5
y(x) = a0 x3/2 + a1 x 2 + a2 x7/2 + a3 x9/2 + a4 x1/2 + a5 x13/2 + a6 x15/2
 
3 1 7 1 11 1 15/2
y2 (x) = a0 x 2 − x 2 − x2 − x
18 936 95472
The general solution of (1) is:

   
−1 1 2 1 4 1 6 3 1 2 1 4 1 6
y = c1 x 1+ x − x + x + · · · +c2 x 2 1 − x − x − x + ··· .
2 24 24 × 42 18 936 95472

101
 Example 2: Find the Frobenius solution of (2x + x3 )y 00 − y 0 − 6xy = 0 about the point
x = 0.
solution: Given:

(2x + x3 )y 00 − y 0 − 6xy = 0 (1)

Normalized form is

y0 6y
y 00 − − =0
(2x + x3 ) (2x + x3 )

y0 6y
y 00 − − =0
x(2 + x2 ) x(2 + x2 )
1 6
Here P (x) = − x(2+x 2 ) , P (x) = − (2+x2 )

x = 0 is a singular point of (1).

1 1
(x − 0)P (x) = x × − =− , are analytic at x = 0
x(2 + x2 ) (2 + x2 )

6 6x2
(x − 0)2 Q(x) = x2 × = − , are analytic at x = 0
2 + x2 2 + x2

Therefore, x = 0 is a regular singular point of (1).

Assume the Frobenius series solution of (1) to be:
∞
X ∞
X
y(x) = (x − x0 )c an (x − x0 )n = an xn+c
n=0 n=0

Differentiate with respect to x:

∞
X
y 0 (x) = (n + c)an xn+c−1
n=0

Differentiate again:
∞
X
y 00 (x) = (n + c)(n + c − 1)an xn+c−2
n=0

Substitute in equation (1), we get:

∞
X ∞
X ∞
X
3 n+c−2 n+c−1
(2x + x ) (n + c)(n + c − 1)an x − (n + c)an x − 6x an xn+c = 0
n=0 n=0 n=0

Simplify and equate powers to n + c:

∞
X ∞
X ∞
X ∞
X
2 (n+c)(n+c−1)an xn+c−1 + (n+c)(n+c−1)an xn+c+1 − (n+c)an xn+c−1 −6 an xn+c+1 = 0
n=0 n=0 n=0 n=0

∞
X
⇒ 2c(c−1)a0 xc−1 +2c(c+1)a1 xc + 2(n+c)(n+c−1)an xn+c−1 +c(c−1)a0 xc+1 +c(c+1)a1 xc+2
n=2

102
 ∞
X ∞
X
+ (n + c)(n + c − 1)an xn+c+1 − ca0 xc−1 − (c + 1)a1 xc − (n + c)an xn+c−1 − 6a0 xc+1 = 0
n=2 n=2
∞
X
−6a1 xc+2 − an xn+c+1 = 0.
n=2

⇒ xc [2c(c + 1)a1 − (c + 1)a1 ]+xc−1 [2c(c − 1)a0 − ca0 ]+xc+1 [c(c − 1)a0 − 6a0 ]+xc+2 [c(c + 1)a1 − 6a1 ]
∞
X
2(n + c)(n + c − 1)xn+c−1 + (n + c)(n + c − 1)xn+c+1 − (n + c)xn+c−1 − 6xn+c+1 an = 0
 
+
n=2

Equating lower power of x

2c(c − 1)a0 − ca0 = 0 → coefficient of xc−1

2c2 − 2ca0 − ca0 = 0
2c2 a0 − 3ca0 = 0 ⇒ c(2c − 3) = 0
3
⇒ c = 0, c =
2
Here two exponents are distinct & they do not differ by a non-ve integer.
Equating the coefficient of xc .

2c(c + 1)a1 − (c + 1)a1 = 0

[2c2 + 2c − c + 1]a1 = 0
[2c2 + c + 1]a1 = 0 → (4)
For c = 0, a1 = 0
For c = 32 , "   #
3 2

3
a1 2 + −1 =0
2 2
 
9 3
a1 + −1 =0
2 2
 
9+3−2
a1 =0
2
5
a1 = 0
2
a1 = 0
Hence a1 = 0 for both the values of c.
Equating coefficients of xn+c on b.s

2(n + c)(n + c + 1)an+1 + (n + c + 1)(n + c − 2)an−1 − (n + 1 + c)an+1 − 6an−1 = 0

an+1 [2(n + c)(n + c + 1) − (n + 1 + c)] = an−1 [6 − (n + c − 1)(n + c + 2)]
6 − (n + c − 1)(n + c + 2)
an+1 = → (5)
(n + c + 1) [2(n + c) − 1]

(4) ⇒ xc [2c(c + 1)a1 − (c + 1)a1 ] + xc−1 [2c(c − 1)a0 − ca0 ] + xc+1 [c(c − 1)a0 − 6a0 ]
∞
X
c+2
+x [c(c + 1)a1 − 6a1 ] + 2(n + c + 1)(n + c)an+1 xn+c + (n + c + 1)(n + c + 2)an−1 xn+c
n=2

103
 −(n + c + 1)xn+c an+1 − 6an−1 xn+c = 0 → (6)
Case 1: c = 0.

6 − (n − 1)(n − 2)
(5) → an+1 = an−1
(n + 1)(2n − 1)
When n = 2,
6 6
a3 = a1 ⇒ a1 = 0 ⇒ when n = 1, a2 = a0 = 3a0
9 2
When n = 3,
4 3
a4 = a2 = a0
20 5
When n = 4,
0
a5 = a3 = 0
35
When n = 5,
14 14 3
a6 = − a4 , a4 = − × a0
54 54 5
∞
X
y= an xn+c , Here c = 0.
n=0
8
X
y= an xn
n=0

y = a0 + a1 x + a2 x2 + . . .
 
3 2
y1 (x) = a0 1 + 3x + x4 + x6 + . . .
2
5 45
Case 2: c = 23 .
From equation (5),
6 − (n + 21 )(n − 12 )
an+1 = an−1
(n + 52 )(2n + 2)
6 − n2 + 14
an+1 = 2n+5 an−1
2 × 2(n + 1)

25 − 4n2
an+1 = an−1
4(n + 1)(2n + 5)
When n = 1,
25 − 4 21
a2 = a0 = a0
4(2)(7) 56
When n = 2,
25 − 16
a3 = a1 , a1 = 0 ⇒ a1 = 0
4(3)(9)
When n = 3,
25 − 36 11 11 21 3
a4 = a2 = − a2 = − × a0 = − a0
4(4)(11) 176 176 56 128
When n = 4,
25 − 64
a5 = a3 = 0, since a3 = 0
4(5)(13)

104
 Substitute in equation
∞
X 3 3 3
y2 (x) = an xn+ 2 , since c = ;c =
2 2
n=0
y2 (x) = a0 x3/2 + a1 x5/2 + a2 x7/2 + a3 x9/2 + · · ·
 
3 3 4
y2 (x) = a0 x3/2 1 + x2 − x + ···
8 128
y1 (x) and y2 (x) are linearly independent solutions. The general solution of (1) is
y = Ay1 (x) + By2 (x)
3x4 3x2
   
2 2 6 3/2 3 4
y = A 1 + 3x + − x + · · · + Bx 1+ − x + ··· .
5 45 8 128

Hermite Differential Equation

The DE y 00 − 2xy 0 + 2αy = 0 where α is constant is known as Hermite DE.
Here P (x) = −2x, Q(x) = 2α.
Both the functions are analytic at all points of R.
In particular, the functions are analytic at x = 0.
Therefore x = 0 is an ordinary point of (1).
Assume the series solution of (1) in the form y = ∞ n
P
n=0 an x → (2).

Differentiate (2) with respect to x twice:

∞
X
y0 = nan xn−1
n=1
X∞
y 00 = n(n − 1)an xn−2
n=2

Substitute in equation (1):

∞
X ∞
X ∞
X
n(n − 1)an xn−2 − 2x nan xn−1 + 2α an xn = 0
n=2 n=1 n=0

∞
X ∞
X ∞
X
n n
(n + 2)(n + 1)an+2 x − 2 nan x + 2αan xn = 0
n=0 n=1 n=0
∞
X
[(n + 2)(n + 1)an+2 − 2(n − α)an ] xn = 0
n=0

Equating the coefficients of xn on both sides of (3):

(n + 1)(n + 2)an+2 = 2(n − α)an

2(n − α)
an+2 = an → (4)
(n + 1)(n + 2)
When n = 0, a2 = − 2αa
2! = −αa0 .
0

When n = 1,
2(1 − α) (−2)(α − 1)
a3 = a1 = a1
2×3 3!

105
 When n = 2,
2(2 − α) (−2)(α − 2) (−2)(α − 2) 2αa0
a4 = a2 = a2 = ×−
3×4 3×4 3×4 2!

(−2)2 α(α − 2)
a4 = a0
4!
When n = 3,
2(3 − α) 2(3 − α) (−2)(α − 1)
a5 = a3 = × a1
4×5 4×5 3!
(−2)2 (α − 1)(α − 3)
a5 = a1
5!
When n = 4,
2(4 − α) (−2)(α − 4) (−2)2 (α − 2)
a6 = a4 = × a0
5×6 5×6 4!
(−2)3 (α − 4)(α − 2)
a6 =
6!
Substitute this in equation (2):
∞
X
Yy = an xn
n=0

y(x) = a0 + a1 x + a2 x2 + a3 x3 + · · ·

(−2) (−2)(α − 1) (−2)2 α(α − 2)

y(x) = a0 + a1 x + αa0 x2 + a1 x3 + a0 x4
2! 3! 4!
(−2)2 (α − 1)(α − 3) (−2)3 α(α − 2)(α − 4)
+ a1 x5 + a0 x6
5! 6!
(−2)3 (α − 1)(α − 3)(α − 5)x7
+ + ···
7!

(−2)αx2 (−2)2 α(α − 2)x4 (−2)3 α(α − 2)(α − 4)x6

 
y(x) = a0 1 + + + + ···
2! 4! 6!

(−2)(α − 1) 3 (−2)2 (α − 1)(α − 3) 5 (−2)3 (α − 1)(α − 3)(α − 5)x7

 
a1 x + x + x +
3! 5! 7!

y(x) = a0 y0 (x) + a1 y1 (x) → (5)

where

(−2)αx2 (−2)2 α(α − 2)x4 (−2)3 α(α − 2)(α − 4)x6

y0 = +1+ + + ···
2! 4! 6!

(−2)(α − 1) 3 (−2)2 (α − 1)(α − 3) 5 (−2)3 (α − 1)(α − 3)(α − 5)x7

y1 = x + x + x + + ···
3! 5! 7!
These two solutions y0 (x) and y1 (x) are linearly independent solutions, hence (5) is the
general solution of (1).
Using summation sign, equations, (6) & (7) can be written as:

106
 ∞
X (−2)k α(α − 2)(α − 4) · · · (α − 2k + 2)x2k
y0 (x) = 1 + → (8)
(2k)!
k=1
∞
X (−2)k (α − 1)(α − 3) · · · (α − 2k + 1)x2k+1
y1 (x) = x + → (9)
(2k + 1)!
k=1

We observe that the series 8 & 9 become finite series when α is a positive even integer &
positive odd integer respectively. n 2n n!
If α = 2n then the coefficients x2n+2 , x2n+4 , . . . vanish & the coefficient of x2 n is (−1)
(2n)!
with this, equation 6 becomes

(−1)n n!
y0 (x) = H2n (x) → (10)
(2n)!
where
2n(2n − 1) (−1)n (2n)!
H2n (x) = (2x)2n − (2x)2n−2 + · · · +
1! n!
If α = 2n + 1, the coefficients x2n+3 , x2n+5 , . . . in equation (7) vanish & the coefficient of
n 22n n!
x2n+1 is (−1)
(2n+1)! with this, equation (7) becomes

(−1)n n!
y1 (x) = H2n+1 (x) → (11)
2(2n + 1)!
where

(−1)n (2n + 1)!2x

 
2n+1 2n(2n + 1) 2n−1
H2n+1 (x) = (2x) − (2x) + ··· +
1! n!
Both the polynomials H2n (x) & H2n+1 (x) can be written in a single series as

bX
2c
n

(−1)k n!
Hn (x) = (2x)n−2k → (12)
k!(n − 2k)!
k=0

where (
jnk n/2, n is even
=
2 (n − 1)/2, n is odd
The polynomial Hn (x) in 12 is known as Hermite polynomial of degree n. This is a solution
of equation (1).
Now for Hermite polynomials are
0
X (−1)k 0!
H0 (x) = (2x)0−2k
k!(0 − 2k)!
k=0

(−1)0
Therefore H0 (k) = (2x)0 = 1
0!0!
0
X (−1)k (−1)0
H1 (x) = (2x)1−2k = (2x)1 = 2x
k!(1 − 2k)! 0!1!
k=0
1
X (−1)k 2! (−1)0 2! (−1)1 2!
H2 (x) = (2x)2−2k = (2x)2 + (2x)0 = 4x2 − 2
k!(2 − 2k)! 0!2! 1!0!
k=0

107
and so on · · ·
Now check, since H is a polynomial of degree n, Hn (x) is a solution of D.E. equation (1).
00 0
we have Hn (x) − 2xHn (x) + 2nHn (x) = 0.
Generating function for Hermite polynomials
∞ ∞
2tx−t2 2tx −t2
X (2tx)n X (−t2 )k
e =e e =
s! k!
s=0 k=0
∞ X
∞
X (−1)k 2s ts xs t2k
=
s!k!
s=0 k=0
∞ X
∞
X (−1)k (2x)s t2k+s
=
s!k!
s=0 k=0

Let s + 2k = n, so that s = n − 2k.

Then (1) becomes:
∞ X
∞
2
X (−1)k (2x)n−2k tn
e2tx−t =
k!(n − 2k)!
n=0 k=0

we have, s ≥ 0, Therefore n − 2k ≥ 0or n ≥ 2kor k ≤ n2 .

 n 
∞ [2]
k
(−1) (2x) n−2k n
2
t
X X
e2ts−t = 
k!(n − 2k)! n!
n=0 k=0
∞
X tn
= Hn (x)
n!
n=0

Where k is an integer (
n/2 if n is even
k=
(n − 1)/2 if n is odd

∞
2
X Hn (x)tn
e2xs−t = is the generating function of the Hermite polynomials.
n!
n=0

Orthogonal properties of Hermite polynomials

2
Hermite polynomials are orthogonal over the interval (−∞, ∞) with weight function e−x ,
i.e., Z ∞ √
2
e−x Hn (x)Hm (x) dx = 2n n! π δmn , (1)
−∞
where (
0, m 6= n
δmn =
1, m = n
To prove (1), let us consider the generating function of Hermite polynomials:
∞
2
X Hn (x)tn
e2xt−t =
n!
n=0

∞
2
X Hm (x)sm
e2xs−s =
m!
m=0

108
 ∞ ∞
2 2
X Hn (x)tn X Hm (x)sm
e2xt−t × e2xs−s =
n! m!
n=0 m=0
∞ X
∞
X tn sm 2 2
Hn (x)Hm (x) = e2x(t+s)−(t +s )
n!m!
n=0 m=0

Multiply by e −x2 , to get

∞ X
∞
X 2 tn sm 2 2 2 2
e−x Hn (x)Hm (x) dx = e−x +2x(t+s)−(t +s ) = e−(x−(t+s)) +2+s
n!m!
n=0 m=0

Integrating throught w.r.t x from −∞ to ∞, we get

∞ X
∞ Z ∞ ∞
tn sm
 Z
2 2
X
e−x Hn (x)Hm (x) dx = e2ts e−(x−(t+s)) dx
−∞ n!m! −∞
n=0 m=0

Put z = x − (t + s),
dz = dx, the integral becomes:
∞ X
∞ Z ∞ ∞
tn sm
Z
−x2 2
X
e Hn (x)Hm (x) dx = e2ts e−z dz
n!m!
n=0 m=0 −∞ −∞

R∞ √ P (2ts)n √ P∞ 2n tn sn
we know that e2ts −∞ e−z dz = π ∞
2
n=0 n! = π n=0 n!
m n
Equatinging the coefficients of t s for m 6= n, we get:
Z ∞
1 2
e−x Hn (x)Hm (x) dx = 0
m!n! −∞
Z ∞
2
e−x Hn (x)Hm (x) dx = 0, m 6= n → (4)
−∞
Again, equating the coefficients of tn sm for m = n, on both sides of the equation, we obtain
Z ∞
1 −x2 √ 2n
e Hn (x)H n (x) dx = π
(n!)2 −∞ n!
Z ∞
2 √
e−x Hm (x)Hn (x) dx = π2n n! → (5)
−∞

Eqn (4) & (5) combine to form a single eqn:

Z ∞
2 √
e−x Hn (x)Hm (x) dx = π2n n!δmn ,
−∞
where δmn is 1 when m = n, 0 otherwise.
2
This shows that Hermite polynomial is orthogonal over −∞ to ∞ w.r.t. e−x .

Recurrence Relation of Hermite’s Polynomial

0 0
1. Hn (x) = 2nHn−1 (x), n ≥ 1; H0 (x) = 0
2. Hn+1 (x) = 2xHn (x) − 2nHn−1 (x), n ≥ 1; H1 (x) = 2xH0 (x)
00 0
3. Hn (x) − 2xHn (x) + αHn (x) = 0
Proof:
1). Consider the generating function of Hermite’s polynomials:

109
 ∞
2
X tn
e2tx−t = Hn (x) (1)
n!
n=0

Differentiate equation (1) w.r.t. x:

∞
2tx−t2 ∂ X 0 tn
e · 2t − = Hn (x)
∂t n!
n=0
∞ ∞
X tn X 0 tn
2t Hn (x) = Hn (x)
n! n!
n=0 n=0
∞ ∞
X 0 tn X tn+1
Hn (x) =2 Hn (x)
n! n!
n=0 n=0

Equating coefficients of tn :
0
Hn (x) Hn−1 (x)
=2
n! (n − 1)!
0
⇒ Hn (x) = 2nHn−1 (x)
0
For n = 0, H0 (x) = 0.
2). To prove (2) differentiate (1) partially with respect to t:
∞
2
X ntn−1
(2x − 2t)e2tx−t Hn (x)
n!
n=0
∞ ∞ ∞
X tn X tn X ntn−1
⇒ 2x Hn (x) − 2t Hn (x) = Hn (x)
n! n! n!
n=0 n=0 n=0
∞ ∞ ∞
X tn X tn+1 X tn−1
⇒ 2x Hn (x) = 2 Hn (x) = nHn (x)
n! n! n!
n=0 n=0 n=0

By equating coefficients of tn :

2x 2 Hn+1 (x)(n + 1)
Hn (x) − Hn (x) =
n! (n − 1)! (n + 1)!
multiply equation throughout by n!:

2xHn (x) − 2nHn−1 (x) = Hn+1 (x), n≥1

Put n = 0, H1 (x) ≡ 2xH0 (x).
3). Since Hn (x) is a solution of Hermite DE y 00 − 2xy 0 + αy = 0,
we have Hn00 (x) − 2xHn0 (x) + xHn (x) = 0.

110