Second Order Linear Differential equations

In this lecture
• We define a second order Linear DE
• State the existence and uniqueness of solutions
of second order Initial Value problems
• We find the general solution of a second order
Homogeneous Linear DE
• Define the Wronskian of two functions and
give a criterion for LI of two functions
Definition Recall that a DE is said to be
linear if the dependent variable and its
derivatives occur only in the first degree.
Thus a second order linear DE in the standard
form is
2
d y dy
2
 P ( x )  Q ( x ) y  R ( x ) …. (*)
dx dx
where P(x), Q(x), and R(x) are continuous
functions of x on some interval [a, b ].
Existence and uniqueness of solutions of
a second order Initial value problem
Theorem Let P(x), Q(x), and R(x) be
continuous functions of x on a closed interval
[a, b]. Let x0 be any point in the interval [a, b]
and let y0, y0 be two real numbers. Then there
exists a unique solution y = y(x) of the second
order LDE
2
d y dy
2
 P ( x)  Q( x) y  R( x)
dx dx
such that when x = x0, y = y0, y = y0 .
Problem: (8, p 87) Show that y = 0 and
y = x2sin x are both solutions of

x y  4 xy  ( x  6) y  0,
2 2

and that both satisfy the conditions

y(0)=0 and y’(0)=0. Does this contradict
Theorem A? If not, why not?
Henceforth whenever we will deal with
a second order LDE, we shall assume
that the hypotheses of Theorem A are
satisfied.

That is to say, every equation we deal

with will have a solution.
Second Order Homogeneous L.D.E.s
A second order homogeneous LDE is an
equation of the form
2
d y dy
2
 P ( x)  Q( x) y  0 … (**)
dx dx
If y = u(x) and y = v(x) are two solutions of
the eqn. (**), then any scalar linear
combination y =  u(x) +  v(x) is also a
solution of (**). Here ,  are two real
numbers.
Follows from the fact that the derivative of
sum is sum of the derivatives, the derivative of
constant times a function is the same constant
times the derivative and that 0 + 0 = 0.
or  u ( x)   v( x)   P( x)  u ( x)   v( x) 
 Q ( x )  u ( x )   v ( x ) 
   u( x)  P ( x)u ( x)  Q( x)u ( x) 
   v( x)  P ( x)v( x)  Q( x)v( x) 
  .0   .0
Theorem If y = yp(x) is a “particular solution”
of
2
d y dy
 P ( x)  Q( x) y  R( x)
dx 2
dx …. (*)

and if y = yh(x) is the general solution of the

“associated homogeneous equation” (**),
then the general solution of (*) is
y = yh(x) + yp(x)
Thus from the previous theorem, we find
that to solve a general second order LDE we
have to find
• the general solution of the “associated
homogeneous equation”
• a particular solution of the given second
order linear differential equation
• and add the two to get the general
solution of the given second order linear
differential equation
We let C[a, b] as the set of all continuous real
valued functions on the interval [a, b]. We
define C[a, b] as the set of all twice
continuously differentiable real valued
functions on the interval [a, b]. We know from
Linear Algebra that C[a, b] and C[a, b] are
both real vector spaces under pointwise
addition and scalar multiplication of functions.
i.e. If f, g are two functions, f + g, α f are
defined by (f+g)(x) = f(x) + g(x)
(α f )(x) = α f (x) for all points x.
The second order LDE
2
d y dy
2
 P ( x)  Q( x) y  R( x)
dx dx
can be written in the compact form L y  R ( x )
2
d d
where L  2  P( x)  Q( x)
dx dx
 D  P( x) D  Q( x)
2

2
d d
(D  , D  2 )
2

dx dx
 We thus have a Linear transformation
L : C[a, b]  C[a, b]

C[ a , b ]
C [a, b] y Ly

defined by y  Ly
The set of all solutions of the “homogeneous
linear DE” d 2 y dy
2
 P( x)  Q( x) y  0
dx dx
equals  y  C[a, b]| Ly  0
and hence is the null space of the linear
transformation L and hence is a subspace of
C [a, b]. We now show that it is a
two-dimensional subspace of C [a, b].
Theorem Let P(x), Q(x), and R(x) be
continuous functions of x on a closed interval
[a, b]. The set, S, of all solutions of the second
order homogeneous LDE
2
d y dy
2
 P( x)  Q( x) y  0 … (**)
dx dx
is a two-dimensional subspace of C [a, b].
Proof Fix a point x0 in the interval [a,b]. By
the existence and uniqueness theorem of a
second order Initial-value problem,
there exists a unique solution y = u(x) of (**)
satisfying the initial conditions:
u(x0) = 1, u (x0) = 0.
Again by the same theorem, there exists a
unique solution y = v(x) of (**) satisfying the
initial conditions: v(x0) = 0, v (x0) = 1.
We show
(i) {u(x), v(x) } is LI.
(ii) u(x), v(x) span S.
Proof of (i) Let α u(x) +  v(x) = 0 …..(1)
Differentiating (1) w.r.t. x, we get
α u(x) +  v(x) = 0 …..(2)
Put x = x0 in (1). We get α × 1 +  × 0 = 0
i.e. α = 0
Put x = x0 in (2). We get α × 0 +  × 1 = 0
i.e.  = 0
Thus {u(x), v(x) } is LI.
Proof of (ii)
Let y = w(x) be a solution of the homogeneous
equation (**). Suppose w(x0) = , w(x0) = .
Now y =  u(x) +  v(x) is also a solution
of (**) satisfying the initial conditions:
when x = x0, y = , y = .
Hence by the uniqueness of the initial value
problem, w(x) =  u(x) +  v(x) or u(x), v(x)
span S.
Thus S has a basis consisting of two vectors.
Or S is two-dimensional. Q.E.D.
 Example: The 2nd order equation
2
d y
2
 y0 (1)
dx
Its solution space is a 2-dimensional subspace of
C(-, ).
It is easy to show that the functions
y1 ( x)  cosh x, y2 ( x)  sinh x

are solutions of (1) on (-, ), and

since
 y1 (0)  1, y1 (0)  0,
y2 (0)  0, y2 (0)  1,
since the vectors (1,0) and (0, 1) are LI ,
coshx and sinhx also form a basis for the
solution space of the equation. It follows
that the general solution of (1) may also
be written
y  c1 cosh x  c2 sinh x,
where c1 and c2 are arbt. Constants.
 x
The function y1 ( x)  e , y2 ( x)  e
x

provide second pair of solutions of (1).

In this case y1 (0)  1, y1 (0)  1,
y2 (0)  0, y2 (0)  1,
since the vectors (1,1) and (1, -1) are LI ,
ex and e-x also form a basis for the
solution space of the equation. It follows
that the general solution of (1) may also
x
be written y  c1e  c2e
x
Wronskian test for Linear Independence
Definition Let y1 ( x), y2 ( x) be two
differentiable functions. Their Wronskian is
defined as the determinant
y1 ( x) y2 ( x )
W ( x) 
y1 ( x) y2 ( x)
Note that the Wronskian is also a function of x.
We may also write W ( x)  W [ y1, y2 ]( x)
to show its dependence on y1 ( x), y2 ( x).
Examples
1. Let y1 ( x)  1, y2 ( x)  x
1 x
Their Wronskian is  1.
0 1
2. Let y1 ( x)  e , y2 ( x)  e
2x 3x

2x 3x
Their Wronskian is e e
2x 3x
 e .
5x

2e 3e
3. Let y1 ( x)  e , y2 ( x)  xe
x x

Their Wronskian is e 2 x .
Fact 1. Let y1 ( x), y2 ( x) be two functions in
C [a, b]. If their Wronskian is not zero
at some point x0 in [a, b], then
{ y1 ( x), y2 ( x)} is LI.

Proof Let c1 y1 ( x)  c2 y2 ( x)  0 …. (i)

Differentiating w.r.t. x, we get
c1 y1 ( x)  c2 y2 ( x)  0 …. (ii)
Substituting x = x0 in equations (i), (ii), we get
 c1 y1 ( x0 )  c2 y2 ( x0 )  0
c1 y1 ( x0 )  c2 y2 ( x0 )  0
These are two homogeneous linear
equations in the two unknowns c1, c2.
Determinant of the coefficient matrix is
y1 ( x0 ) y2 ( x0 )
 W ( x0 )  0
y1 ( x0 ) y2 ( x0 )
Hence c1= 0, c2 =0.
or { y1 ( x), y2 ( x)} is LI. Q.E.D.
 Remark. The converse is not true.
3 3
We can show that {x ,| x | } is LI in C (, ).

But their Wronskian is zero.

We show that the converse is true if
y1 ( x), y2 ( x) are two solutions of a second
order homogeneous linear DE
2
d y dy
2
 P ( x)  Q( x) y  0
dx dx
Abel’s formula for the Wronskian

Theorem Let y1 ( x), y2 ( x) be two solutions of

of a second order homogeneous linear DE
2
d y dy
2
 P ( x )  Q ( x ) y  0 …(**)
dx dx
Then their Wronskian is given by

W ( x)  Ce   P ( x ) dx
for some constant C.
Proof Since y1 ( x), y2 ( x) are solutions of (**)
2
d y1 dy1
2
 P ( x )  Q ( x ) y1  0 …(i)
dx dx
2
d y2 dy2
2
 P ( x )  Q ( x ) y 2  0 …(ii)
dx dx
(i)×y2 – (ii)× y1 gives
 dy1 dy2 
2 2
d y1 d y2
y2 2  y1 2  P( x)  y2  y1   0
dx dx  dx dx 
 The Wronskian of y1 ( x), y2 ( x) is
y1 ( x) y2 ( x )
W ( x) 
y1 ( x) y2 ( x)

 y1 ( x) y2 ( x)  y2 ( x) y1 ( x)
dW
Hence  y1 ( x) y2 ( x)  y1 ( x) y2 ( x)
dx
 ( y2 ( x) y1 ( x)  y2 ( x) y1 ( x))

 y1 ( x) y2 ( x)  y2 ( x) y1 ( x)
 dW
Thus we get  P( x)W  0
dx
Separating the variables and integrating we
get

W ( x)  Ce   P ( x ) dx
C, an arbitrary constant

Corollary The Wronskian of two solutions of

a homogeneous second order linear DE is
identically zero or is never zero.
Theorem Let the Wronskian of two
solutions y1 ( x), y2 ( x) of a second order
homogeneous linear DE be zero. Then
{ y1 ( x), y2 ( x)} is LD.
Proof Case (i) y1(x) is the zero function

Hence { y1 ( x), y2 ( x)} is LD.

Case (ii) y1(x) is not zero

d  y2  y1 ( x) y2 ( x)  y2 ( x) y1 ( x)
  2
dx  y1  y1
W ( x)
 2
= 0.
y1
y2
Hence  c, a constant.
y1

or y2  cy1 i.e. { y1 ( x), y2 ( x)} is LD.

We now give an alternative proof.
Theorem Let the Wronskian of two solutions
y1 ( x), y2 ( x) of a second order homogeneous
linear DE be zero at some point x0 in [a, b].
Then y1 ( x), y2 ( x) is LD.

Proof Given W ( x0 )  y1 ( x0 ) y2 ( x0 )
 0.
y1 ( x0 ) y2 ( x0 )
Consider the two homogeneous algebraic
linear equations in two unknowns x1, x2 :
x1 y1 ( x0 )  x2 y2 ( x0 )  0
x1 y1 ( x0 )  x2 y2 ( x0 )  0
The determinant of the coefficient matrix is
W ( x0 )  0.
Hence the above system has a nontrivial
solution (c1, c2). That is we can find c1, c2
(both not zero , i.e. at least one of them 0)
such that c1 y1 ( x0 )  c2 y2 ( x0 )  0

c1 y1 ( x0 )  c2 y2 ( x0 )  0

Now y  c1 y1 ( x)  c2 y2 ( x)
is a solution of the homogeneous second order
LDE (**) such that y(x0) = 0, y(x0) = 0.
But y = 0 is also a solution of the
homogeneous second order LDE (**) such
that y(x0) = 0, y(x0) = 0. Hence by the
uniqueness theorem, c1 y1 ( x)  c2 y2 ( x)  0
Thus we have found constants c1, c2

(not both zero) such that

c1 y1 ( x)  c2 y2 ( x)  0

Thus { y1 ( x), y2 ( x)} is LD.

Q.E.D.
Examples 1. y1 ( x), y2 ( x) are two solutions
of the second order linear d.e.
y  4 y  4 y  0
Find their Wronskian W(x).
By Abel’s formula, W ( x)  Ce  P ( x ) dx
 Ce  4 dx
 Ce , C an arbitrary constant
4x

2. y1 ( x), y2 ( x) are two solutions of the

second order linear d.e. y  2 xy  0
Their Wronskian W(x) is C.
 Use of a known solution to find another.
Suppose y = y1(x) is a known solution of the
second order LDE
2
d y dy … (**)
2
 P ( x )  Q ( x ) y  0
dx dx
We assume a second LI solution as y2  v y1
Now dy2 dy1 dv
v  y1
dx dx dx
2 2 2
d y2 d y1 dv dy1 d v
2
v 2 2  2 y1
dx dx dx dx dx
 Substituting y = y2= vy1 in (**), we get
2 2
d y1 dv dy1 d v
v 2
2  2 y1
dx dx dx dx
 dy1 dv 
 P( x) v  y1   Q( x)vy1  0
 dx dx 
i.e.  d 2 y1 dy1 
v  2  P( x)  Q( x) y1 
 dx dx  This is zero
2
d v dv  dy1 
 2 y1  2  P( x) y1   0
dx dx  dx 
 2
d v dv  dy1 
i.e. y  2
2 1
 P( x) y1   0
dx dx  dx 
dv
Let z 
dx
dz  dy1 
y1  z  2  P( x) y1   0
Hence we get dx  dx 
Separating the variables we get
1 dz  2 dy1 
   P( x) 
z dx  y1 dx 
 Integrating, we get
ln z  2ln y1   P( x)dx
Taking exponentials, we get
dv 1   P ( x ) dx
z  2e
dx y1
1   P ( x ) dx
Again integrating, we get v   2 e dx
y1
We now show that { y1 , y2} is LI, where
y2  v y1
 The Wronskian of y1 ( x), y2 ( x) is
y1 ( x) y2 ( x )
W ( x) 
y1 ( x) y2 ( x)

 y1 ( x) y2 ( x)  y2 ( x) y1 ( x)

 y1 ( x) vy1 ( x)  vy1 ( x)   vy1 ( x) y1 ( x)

 

 vy ( x)  e  P ( x ) dx  0. Hence { y1 , y2} is LI.

2
1
Thus a second LI solution of (**) is
1   P ( x ) dx
y2  v y1, where v   2 e dx
y1
Condition satisfied Solution

1+P(x)+Q(x)=0 y = ex
1-P(x)+Q(x)=0 y = e-x

a2 +aP(x)+Q(x)=0 y = eax

a2 -aP(x)+Q(x)=0 y = e-ax
P(X) + xQ(X)=0 y=x
m(m -1)+mxp(x)+x2Q(x)=0 y = xm
Examples 1. A solution of the d.e.
2
d y dy
2
 4  4y  0
dx dx

is y1  e .
2x
Find the complete solution.
Solution A second LI solution y2 is given by

y2  v y1 where
1   P ( x ) dx
v 2e
y1
1   4 dx
dx   4 x e
e
dx  x.
Hence the complete solution is
y  c1 y1  c2 y2  c1 e 2 x  c2 x e 2 x

where c1, c2 are arbitrary constants.

Problem 9, page 94: Solve the DE
xy  (2 x  1) y  ( x  1) y  0
Solution Since the sum of the coefficients
x  (2 x  1)  ( x  1)  0
y  y1  e is one solution.
x

(2 x  1)
Note P = 
x
Thus a second LI solution y2 is given by
y2  v y1, where
 1   P ( x ) dx
v 2e dx
y1
(2 x 1) 2
1    x
dx   x dx 
dx
  2x e x
e 2
Hence the complete solution is
y  c1 y1  c2 y2  c1 e  c2 x e
x 2 x

where c1, c2 are arbitrary constants.

(Note: We have written c2/2 as a new
constant c2.)
 Problem 5 page 94
The equation

(1-x )y-2xy+2y=0
2
is the special case of Legender’s equation

(1-x )y-2xy+p(p+1)y=o,
2

Corresponding to p = 1, it has y1 = x as
an obvious solution. Find the general
solution.
Solution is given below
( 1 - x )y '' - 2 xy '  2 y  0
2
(1 )
One solution is y1  x (2)
Writing (1) is the standard form
y  P(x)y ' Q(x)y  0. We have,
''

2x ' 2
y 
''
y  y  0 (3)
1 x 2
1 x 2

2 x
here P( x) 
1 x 2
The second linearly independent solution is
given by
y 2  vy1 , where
1   Pdx
v 2e dx
y1
2 x
1   dx
  2e 1 x 2
dx
x
1  ln(1 x2 ) 1 1
 2e dx   2  dx
x x 1 x 2
  1 1  1 1  1 x 
  2  2 
dx    ln  
 x 1 x  x 2.1  1  x 
 1 x 
v  1/ x  1/ 2 ln  
 1  x 
 1 x 
y2  vy1  {1/ x  1/ 2 ln  }x
 1  x 
x  1 x 
y2  ln   1
2  1  x 
Solution is
x  1 x 
y ( x)  c1 x  c2 { ln    1}
2  1  x 