Variation of Parameters for Second

Order Linear Differential Equations

The solution of non-homogeneous equations is possible when a particular solution, yp , of the

equation can be found. If yc is the general solution of the associated homogeneous equation,
then we know that y = yc + yp is the general solution of the non-homogeneous equation. The
method of variation of parameters is a powerful general method that can be used to find such a
particular solution. To be more specific (and focus on second order linear equations) we would
like to find a particular solution for,
′′ ′
y + P (x)y + Q(x)y = R(x) (1)

with R(x) any well-behaved function. To apply the method we need to know the general solution
of the associated homogeneous equation, i.e.
′′ ′
y + P (x)y + Q(x)y = 0 (2)

Let us assume that the general solution is yc = C1 y1 + C2 y2 , with C1 and C2 two arbitrary
constants and y1 and y2 two linearly independent solutions of equation (2). In the method we
are described the following form is postulated for a particular solution:

yp = v1 (x)y1 (x) + v2 (x)y2 (x) (3)

i.e. the arbitrary constants become two variable functions to be found. In order to do so we
simply compute first and second derivatives, The first derivative yields:
′ ′ ′ ′ ′ ′ ′ ′ ′
yp = v1 y1 + v1 y1 + v2 y2 + v2 y2 = (v1 y1 + v2 y2 ) + (v1 y1 + v2 y2 )

We have collected terms in the above way to avoid having to handle second derivatives for v1
and v2 . Thus, we fix the content in the second parenthesis equal to 0:
′ ′
v1 y1 + v2 y2 = 0 (4)
′
The first derivative, yp , becomes, thus,
′ ′ ′
yp = v1 y1 + v2 y2 (5)

From (5) we carry on, computing the second derivative:

′′ ′ ′ ′′ ′ ′ ′′
yp = v1 y1 + v1 y1 + v2 y2 + v2 y2 (6)

Next, let us substitute (3), (5), (6) in the differential equation (1). The result is:
′ ′ ′′ ′ ′ ′′ ′ ′
(v1 y1 + v1 y1 + v2 y2 + v2 y2 ) + P (x)(v1 y1 + v2 y2 ) + Q(x)(v1 y1 + v2 y2 ) = R(x)

1
 ⇓
′′ ′ ′′ ′ ′ ′ ′ ′
v1 [y1 + P (x)y1 + Q(x)y1 ] + v2 [y2 + P (x)y2 + Q(x)y2 ] + v1 y1 + v2 y2 = R(x)
Quantities inside the square brackets are null, as both y1 and y2 obey equation (2). The above
expression becomes, thus:
′ ′ ′ ′
v1 y1 + v2 y2 = R(x) (7)
′ ′
Together, equations (4) and (7) form a system for the independent quantities v1 and v2 . To solve
this system we first compute the following determinant:
 
 y1 y2  ′ ′
 y y ′  = y1 y2 − y2 y1 ≡ W (y1 , y2 )
 ′ 
1 2
The above determinant has been indicated as W (y1 , y2 ) because is the wronskian of the two
solutions y1 and y2 . This determinant is always different from zero, as y1 and y2 are linearly
independent. We need two more determinants to find the system’s solution:
   
 0 y2   y1 0 
 R(x) y ′  = −y2 R(x) ,  y ′ R(x)  = y1 R(x)
 
2 1
The solutions are, therefore,
′ y2 R(x) ′ y1 R(x)
v1 = − , v2 = (8)
W (y1 , y2 ) W (y1 , y2 )
These can, successively, be easily integrated to provide v1 and v2 .

EXAMPLE 1.
Find the general solution of,
′′ ′
y − y − 2y = 4x2
Solution.
The associated homogeneous equation has got the following two independent solutions:
y1 = exp(2x) , y2 = exp(−x)
Their derivatives are:
′ ′
y1 = 2 exp(2x) , y2 = − exp(−x)
Thus, the wronskian is given by:
 2x 
 e e−x 
  = −3ex
 2e2x −e−x 
We can find v1 and v2 using formula (8), from which:
′−4x2 e−x 4 2 −2x ′ 4x2 e2x 4
v1 = = x e , v2 = = − x2 ex
−3e x 3 −3e x 3
Eventually, after repeated integration by parts, we find:
1 4
v1 = − e−2x (2x2 + 2x + 1) , v2 = − ex (x2 − 2x + 2)
3 3
The particular solution is, therefore, given by:
   
1 −2x 2 2x 4 x 2
yp = v1 y1 + v2 y2 = − e (2x + 2x + 1) e + − e (x − 2x + 2) e−x
3 3
or, after calculations,
yp = −2x2 + 2x − 3
The general solution of our equation is, in the end,
y = C1 exp(2x) + C2 exp(−x) − 2x2 + 2x − 3