41 The Method of Variation of Parameters

2.2 The Method of Variation of Parameters

Consider the problem

We try to find a particular
L[y] = −f (2.56)
solution. You could instead,
use the integration factor where
approach or the method
L[y] = a0 (x)y (n) + a1 (x)y (n−1) + · · · + an−1 (x)y (1) + an (x)y
of undetermined coefficients;
those methods involve some aj (x) are continuous for x ∈ (a, b) and a0 (x) 6= 0 for all x ∈ (a, b). Assume that there exists
guess work. a particular solution of the form
n
X
yp (x) = µj (x)fj (x) (2.57)
j=1

where µj (x) are as yet undetermined and {fj (x)} is a fundamental set of L[y] = 0. Differenti-
µj (x) cannot all be con- ating 2.57 gives
stants because then we are n n
dyp (1)
X (1)
X (1)
back to finding a solution for = yp = µj fj + µj fj
dx j=1 j=1
L[y] = 0 instead of L[y] =
−f . If we set
n
(1)
X
µj fj = 0 (2.58)
Watch out for notation: f j=1
refers to the nonhomoge-
then we get
neous term whereas fj refers n
(1)
X
to a fundamental solution. yp(1) = µj fj (2.59)
j=1

(2) Pn (1) (1) Pn (2)

Differentiating once again gives yp = j=1 µj fj + j=1 µj fj . Now setting
n
(1) (1)
X
µj fj =0 (2.60)
j=1

gives
n
(2)
X
yp(2) = µj fj (2.61)
j=1

In similar fashion,
(3) Pn (1) (2) Pn (3) Pn (1) (2)
yp = j=1 µj fj + j=1 µj fj and set j=1 µj fj =0 (2.62)
..
.
(n−1) Pn (1) (n−2) Pn (n−2) Pn (n−1) Pn (1) (n−2)
yp = j=1 µj fj + j=1 µj fj + j=1 µj fj and set j=1 µj fj = 0 (2.63)

and
n n
(1) (n−1) (n)
X X
yp(n) = µj fj + µj fj (2.64)
j=1 j=1

Pn (1) (n−1)
For this last equation, we cannot set j=1 µj fj = 0 as we did for lower-order derivatives:
Eqns.2.58, 2.60, 2.62 and 2.63 between them form a set of n algebraic equations in the n
(1)
unknowns µj . The determinant of this set of equations would be the Wronskian W (x) which
(1)
is nonzero, and hence that would force a unique solution µj = 0, which in turn implies that
µj s would be constants which is not permissible.
 42 Differential Equations

Pn (1) (n−1) (j)

We cannot set j=1 µj fj = 0 but by substituting the yp into L[yp ] = −f we get an
expression for that sum as follows.

L[yp ] = ao yp(n) + a1 yp(n−1) + · · · + an−1 yp(1) + an yp

 
n n n n n
(1) (n−1) (n) (n−1) (1)
X X X X X
= a0 µj fj + a0 µj fj + a1 µj fj + · · · + an−1 µj fj + an µj fj 
j=1 j=1 j=1 j=1 j=1
 
n n
(1) (n−1) (n) (n−1) (1)
X X
= a0 µj fj + µj {a0 fj + a1 fj + · · · + an−1 fj + an fj }
j=1 j=1
n n 0
(1) (n−1)
X X
= a0 µj fj + L[f
µj ]
j
*
j=1 j=1
n
(1) (n−1)
X
= a0 µj fj (2.65)
j=1

Since L[yp ] = −f , we can rearrange this to

n
X (1) (n−1) f (x)
µj fj =− (2.66)
j=1
a0 (x)

Eqs.2.58, 2.60, 2.62, 2.63, and 2.66 form a set of n linear algebraic equations in n unknowns
(1)
µj (x):
(1) (1)
µ1 f1 + µ2 f2 + · · · + µ(1)
n fn = 0
(1) (1) (1) (1)
µ1 f1 + µ2 f2 + · · · + µ(1) (1)
n fn = 0
..
.
(1) (n−2) (1) (n−2)
µ1 f1 + µ2 f2 + · · · + µ(1)
n fn
(n−2)
=0
(1) (n−1) (1) (n−1) f (x)
µ1 f1 + µ2 f2 + · · · + µ(1)
n fn
(n−1)
=−
a0 (x)
From Cramer’s rule, the solution is
 
 f1
 f2 ··· fj−1 0 fj+1 ··· fn 

 f (1) f2
(2)
···
(1)
fj−1 0
(1)
fj+1 ··· fn
(1) 
(1) 1  1 
µj (x) = . 
W (x)  .. 

−f (x)
 (n−1) (n−1) (n−1) (n−1) (n−1) 
f
1 f2 ··· fj−1 a0 (x) fj+1 ··· fn
Dj (x)
= (2.67)
W (x)
which on integration gives
Z x
Dj (t)
µj (x) = dt (2.68)
x0 W (t)
where µj (x0 ) is arbitrarily set to zero. We introduce the following notation
A nonzero value for µj (x0 )  
in eq. 2.57 results in a lin-
 f1
 f2 · · · fj−1 0 fj+1 ··· fn 

 (1) (2) (1) (1) (1)
 f1 f2 · · · fj−1 0 fj+1 ··· fn

ear combination of the fun- 
Wj (t) =  .
  (2.69)
damental set of L[y] = 0.  ..


 (n−1) (n−1) (n−1) (n−1) (n−1) 

f
1 f2 · · · fj−1 1 fj+1 ··· fn

which modifies eq. 2.68 to

x
−f (t) Wj (t)
Z
µj (x) = dt (2.70)
x0 a0 (t) W (t)
 43 The Method of Variation of Parameters

From eq. 2.57 the particular solution is

The integration and sum- n n x
−f (t) Wj (t)
X X Z
mation operations are ex- yp (x) = µj (x)fj (x) = fj (x) dt
changed. j=1 j=1 x0 a0 (x) W (t)
x
Pn
−f (t) fj (x)Wj (t)
Z
j=1
= dt
x0 a0 (t) W (t)

Noting that
 
 f1 (t)
 (1) f2 (t) ··· fn (t)

(1) (1)
n
 f (t)
 1 f2 (t) ··· fn (t) 

∼
X  .. 
fj (x)Wj (t) = W (x, t) =  .

 (2.71)
j=1  (n−2) (n−2) (n−2) 
f1
 (t) f2 (t) ··· fn (t)
 f (x)
1 f 2 (x) ··· fn (x) 

which gives
∼
x
−f (t) W (x, t)
Z
yp (x) = dt (2.72)
x0 a0 (t) W (t)

which is the particular desired solution of L[y] = −f .

∼
The term W (x, t)/a0 (t)W (t) is called the one-sided Green’s function of the operator L[ ]. It
is a function only of L[ ] and not of the nonhomogeneous term −f (x). Once it is determined,
the particular solution can be found using eq. 2.72 for an arbitrary f (x).

In generating the particular solution, we arbitrarily assumed eqs. 2.58, 2.60, etc. We need to
confirm that this is indeed a particular solution given that we assumed things along the way.
We do this by first stating the following theorem.

Theorem 2.15. The particular solution in eq. 2.72 satisfies the homogeneous ICs

yp (x0 ) = 0, yp(1) (x0 ) = 0, · · · , yp(n−1) (x0 ) = 0 (2.73)

Proof yp (x0 ) = 0 is easily observed (integration limits are the same in eq. 2.72. We need
to use the Leibnitz formula to handle the higher derivatives. Given
(
Z α2 x)
I(x) = F (x, t)dt (2.74)
α1 (x)

its derivative is
(
Z α2 x)  
dI ∂F (x, t) dα2 dα1
= dt + F (x, α2 (x)) − F (x, α1 (x)) (2.75)
dx α1 (x) ∂x dx dx

Differentiating eq. 2.72 gives

∼  ∼
x 
−f (t) ∂ W (x, t) −f (x) W (x, x)
Z
yp(1) (x) = dt +
x0 a0 (t)W (t) ∂x a0 (x) W (x)

By definition,
 
 f1 (t)
 (1) f2 (t) ··· fn (t) 

(1) (1)
 f (t)
 1 f2 (t) ··· fn (t) 

∼  .. 
W (x, t) =  .


 (n−2) (n−2) (n−2) 
f1
 (t) f2 (t) ··· fn (t)
 f (x)
1 f2 (x) ··· fn (x) 
 44 Differential Equations

∼
and hence W (x, x) = 0 since the first and last rows become identical. This results in
∼
x
−f (t) ∂ W (x, t)
Z
yp(1) (x) = dt (2.76)
x0 a0 (t)W (t) ∂x
(1)
and hence yp (x0 ) = 0. From the definition of the derivative of a determinant,
 
 f (t)
 1 f2 (t) ··· fn (t) 
 (1) (1) (1)
···

∼  f1 (t) f2 (t) fn (t) 
∂ j W (x, t) 
..

=

∂xj

 . 

f (n−2) (t) f (n−2) (t) · · · f (n−2) (t)
 1 2 n 
 (j) (j) (j) 
 f1 (x) f2 (x) · · · fn (x) 
which implies that
because the (j+1)th and last ∼
∂ W (x, x)
rows are identical. = 0, for j = 0, 1, . . . , (n − 2) (2.77)
∂xj
Differentiating eq. 2.76, we get
∼  ∼
x
−f (t) ∂ 2 W (x, t)

−f (x)
Z
∂ W (x, x)
yp(2) (x) = 2
dt +
x0 a0 (t)W (t) ∂x a0 (x)W (x) ∂x
which simplifies to
∼
x
−f (t) ∂ 2 W (x, t)
Z
yp(2) = dt (2.78)
x0 a0 (t)W (t) ∂x2
(2)
and therefore yp (x0 ) = 0. Applying this procedure successively results in
∼  n−2 ∼
x
−f (t) ∂ n−1 W (x, t)

−f (x)
Z
∂ W (x, x)
yp(n−1) (x) = n−1
dt +
x0 a0 (t)W (t) ∂x a0 (x)W (x) ∂xn−2
The last term is zero and therefore
∼
x
−f (t) ∂ n−1 W (x, t)
Z
yp(n−1) (x) = dt (2.79)
x0 a0 (t)W (t) ∂xn−1
(n−1)
which implies that yp (x0 ) = 0 and confirms eq. 2.73.

To prove that eq. 2.72 satisfies L[y] = −f , first differentiate eq. 2.79. Since
∼
n−1 n−1
∂ W (x, x)/∂x = W (x),
∼
x
−f (t) ∂ n W (x, t)
 
−f (x)
Z
yp(n) (x) = dt + (2.80)
x0 a0 (t)W (t) ∂xn a0 (t)
(j)
Using the integral expressions for yp (x) (eqs. 2.72, 2.76, 2.78 to 2.80), we get

L[yp ] = a0 (x)yp(n) + a1 (x)yp(n−1) + · · · + an−1 (x)yp(1) + an (x)yp

∼ ∼ ∼
 
Z x
−f (t)  ∂ n W (x, t) ∂ n−1 W (x, t) ∂ W (x, t) ∼
= −f (x) + a0 (x) + a1 (x) + · · · + an−1 + an (x)W (x, t) dt
x0 a0 (t)W (t) ∂xn ∂xn−1 ∂x

The term in the square brackets can be seen to be

 
 f1 (t)
 (1) f2 (t) ··· fn (t) 

(1) (1)
 f (t)
 1 f2 (t) ··· fn (t) 

 .. 
.
 
 
 (n−2) (n−2) (n−2) 
 f1
 (t) f2 (t) · · · fn (t)
 L[f ] L[f2 ] ··· L[fn ] 
1
 45 Separation of Variables

which is zero because each term in the last row is zero, since {fj } is a fundamental set of
solutions of the homogeneous equation L[y] = 0. Hence

L[yp ] = −f (2.81)

Consider the case where {fj } is the fundamental set of solutions of L[y] = 0: they satisfy the
ICs
The Kronecker delta δij is 1 (n−1)
fj (x0 ) = δij
for i = j and 0 for i 6= j.
Then from Theorems 2.7 and 2.12 it follows that
n Z x ∼
X −f (t) W (x, t)
y(x) = αj fj (x) + dt (2.82)
j=1 x0 a0 (t) W (t)

is the unique solution of the ODE L[y] = −f with ICs y (j−1) (x0 ) = αj , j = 1, 2, . . . , n, where
αj are constants.

Example 2.16. Using the one-sided Green’s function find a particular solution of
d3 y dy
L[y] = 3
− = y (3) − y (1) = x (2.83)
dx dx
From eq. 2.72, a particular solution of L[y] = −f is given by
∼
x
−f (t) W (x, t)
Z
yp = dt
x0 a0 (t) W (t)
where W (x) is the Wronskian determinant corresponding to a fundamental solution set {fj }
of the homogeneous solution L[y] = 0. First we therefore find the fundamental set solving
y (3) − y (1) = 0. Using y = emx , the equation becomes m3 − m = 0 which gives m = 0, 1, −1.
Hence f1 = 1, f2 = ex , and f3 = e−x . The Wronskian determinant is nonzero and confirms
that this is a fundamental set:
   
 f1 (x)
 f2 (x) f3 (x)  1 ex e−x 
W (x) = f1(1) (x) f2(1) (x) (x) 
f3  = 0 ex −e−x  = 2
  
 (2)
f (x) f (2) (x) f (2) (x) 0 ex e−x 
  
1 2 3

Also,
   
∼
 f1 (t)
 f2 (t) f3 (t)  1 et e−t 
W (x, t) = f1(1) (t) f2(1) (t) f3(1) (t) = 0 et −e−t  = −2 + et−x + e−t+x
   
ex e−x 
   
 f1 (x) f2 (x) f3 (x)  0

The particular solution then is

Z x
t 1
yp (x) = (−2+et−x +e−t+x )dt = [−2+(x20 −x2 )+(1−x0 )ex0 −x +(1+x0 )ex−x0 ] (2.84)
x0 2 2

2.3 Separation of Variables

Introduction

Consider the transient conduction equation for T (x, t). The PDE is
 
∂ ∂T ∂
k = (Cp ρT )
∂x ∂x ∂t