Nonhomogeneous Linear Systems of Differential Equations

with Constant Coefficients

Objective: Solve
d~x
= A~x + ~f (t),
dt

f1 (t)
where A is an nn constant coefficient matrix A and ~f (t) = ... is a given vector function.

fn (t)

x1 (t)
The unknown is ~x(t) = ... .
xn (t)

Solution Formula Using Fundamental Matrix: Suppose that M(t) is a fundamental matrix
solution of the corresponding homogeneous system ~x (t) = A~x(t); in other words,
M(t) satisfies M (t) = AM(t); that is, every column of M(t) solves the homogeneous system ~x (t) = A~x(t);
M(t) is an invertible matrix for every t; that is, the n columns of M(t) are
linearly independent.
The general solutions of the nonhomogeneous system d~x/dt = A~x + ~f(t) are

Z t
C1
.
~x(t) = M(t) .. +
M(t)M(s)1 ~f(s)ds,
Cn
where C1 , , Cn are arbitrary constants.
The solution of the initial value problem
d~x
= A~x + ~f (t),
dt
is given by
~x(t) = M(t)M(t0 )

~x0 +

~x(t0 ) = ~x0
t

M(t)M(s)1 ~f(s)ds.

t0

Solution Formula Using Matrix Exponential: The general solutions of the nonhomogeneous
system d~x/dt = A~x + ~f(t) are

C1
Z t
tA ..
~x(t) = e
+
e(ts)A ~f (s)ds,
.
Cn
where C1 , , Cn are arbitrary constants.
The solution of the initial value problem
d~x
= A~x + ~f (t),
dt
1

~x(t0 ) = ~x0

is given by
(tt0 )A

~x(t) = e

~x0 +

e(ts)A ~f (s)ds.

t0

Solution Method by Decoupling: If A is diagonalizable (i.e., A = P DP 1 with an invertible

P and a diagonal D), then the system can be decoupled by setting ~x(t) = P ~u(t). The system
for ~u(t) becomes
d~u
= D~u + P 1~f(t).
dt
EXAMPLE. Solve the initial value probelm




 2t 
d~x
6/7 15/14
4
e
~x + t , ~x(0) =
=
.
5/7 37/14
e
1
dt
Solution 1 (Use a fundamental matrix): First find eigenvalues and eigenvectors of A.
The eigenvalues
 of A are 1 = 1/2, 2 = 3.
3
Vector v1 =
is an eigenvector associated with 1 = 1/2.
1 
1
is an eigenvector associated with 2 = 3.
Vector v2 =
2
Thus,

 t/2
3e
e3t
1 t
2 t
M(t) = [ e v1
e v2 ] =
et/2
2e3t
is a fundamental matrix for the homogeneous system ~x (t) = A~x(t).
The solution to the initial value problem is given by
Z t
1
~x(t) = M(t)M(0) ~x(0) +
M(t)M(s)1 ~f (s)ds
0
  s/2

1 
 Z t  t/2
 t/2
3t
3e
3
1
4
3e
e3t
3e
e
+
=
t/2
3t
es/2
1
2
1
e
2e
et/2
2e3t
0

 
 Z t  t/2
 t/2
3e
e3t 1 2e3s/2 + e3s/2
3e + e3t
ds
+
= t/2
et/2
2e3t 7 es + 3e4s
e 2e3t
0
4 3t/2
t/2
t/2

2 3t/2
3e
e3t
3e + e3t
(e

1)

(e

1)
3
3
1
+

=
7
3 4t
t/2
3t
t/2
3t
t
e
2e
e 2e
(e 1) 4 (e 1)
3
5 t 2 t/2
1 3t
2t

 t/2
e

e
+
e
3e + e3t
28
7
28

7
+
= t/2

e 2e3t
2 t/2
1 3t
10 2t 13 t
e e e e
21
42
21
14

3
5
19
29
e2t et + et/2 + e3t

7
28
7
28
=
.
10 2t 13 t 19 t/2 29 3t
e e + e e
21
42
21
14
2

e3s
2e3s

1 


e2s
ds
es

Solution 2 (Use decoupling): The geiven coefficient matrix A is diagonalizable: A =

P DP 1 with




3 1
1/2 0
P =
,
D=
.
1 2
0 3
Set ~x(t) = P ~u(t). The system for ~u(t) becomes



  2t 
d~u
1/2 0
2/7 1/7
e
1~
~u +
= D~u + P f(t) =
t ,
0
3
e
1/7
3/7
dt

~u(0) = P


4
.
1

or, equivalently,
2
1
1
u1 = u1 + e2t + et , u1 (0) = 1,
2
7
7
1
3
u2 = 3u2 e2t + et , u2 (0) = 1.
7
7
These two equations can be solved separately (the method of integrating factor and the method
of undetermined coefficients both work in this case). The solutions are
2
19
4 2t
e et + et/2 ,
21
21
21
1 2t
3 t 29 3t
u2 (t) = e e e .
7
28
28

u1 (t) =

Finally, the solution to the original problem is given by



u1 (t)
~x(t) = P ~u(t) = P
u2 (t)

4
2
19


e2t et + et/2
3 1 21
21
21

1 2
3 t 29 3t
1 2t
e e e
7
28
28
3

5
19
29
e2t et + et/2 + e3t
28
7
28
7

=
.
10 2t 13 t 19 t/2 29 3t
e e + e e
21
42
21
14
Solution 3 (Use matrix exponential): First find etA using a fundamental matrix: etA =
M(t)M(0)1 . Or, one can also find etA by diagonalization:
6
1
3 t/2 3 3t



  t/2
et/2 + e3t
e e
2/7 1/7
3 1
e
0
7
7
7
7
tA
tD 1
=
e = Pe P =
.
3t
1/7 3/7
1 2
0 e
1 t/2 6 3t
2 t/2 2 3t
e e
e + e
7
7
7
7
The solution to the initial value problem is given by
Z t
tA
~x(t) = e ~x(0) +
e(ts)A~f (s)ds
0

1
3 t/2 3 3t


et/2 + e3t
e e
7
7
7 4
7
=

1
1 t/2 6 3t
2 t/2 2 3t
e e
e + e
7
7
7
7
6
1
3 (ts)/2 3 3(ts)


Z t
e(ts)/2 + e3(ts)
e
e
7
7
7
e2s
7
+
s ds

e
1 (ts)/2 6 3(ts)
2 (ts)/2 2 3(ts)
0
e
e
e
+ e
7
7
7
7
6
1 3ts 3 t/23s/2 3 3t4s
t/2+3s/2

 t/2
Z
e
+ e
+ e
e
t
3e + e3t
7
7
7
7

+
= t/2

ds
e 2e3t
2
1
6
2
0
t/2+3s/2
3ts
t/23s/2
3t4s
e
e
+ e
+ e
7
7
7
7

 t/2
3e + e3t
= t/2
e 2e3t
6
2
1
3
2
3
et/2 (e3t/2 1) + e3t (1 et ) + et/2 (1 e3t/2 ) e3t
3
7
7
3
7
7
+
2
1
2
6
2 t/2 2 3t/2
e (e
1) e3t (1 et ) + et/2 (1 e3t/2 ) + e3t
7
3
7
7
3
7
3

5
2
1

 t/2
e2t et et/2 + e3t
3t
3e + e
7
28
7
28

+
= t/2

e 2e3t
2 t/2
1 3t
10 2t 13 t
e e e e
21
42
21
14
3

5
19
29
e2t et + et/2 + e3t
28
7
28
7

=
.
10 2t 13 t 19 t/2 29 3t
e e + e e
21
42
21
14
6

1
(1 e4t )
4

1
4t
(1 e )
4

EXERCISES
[1] Solve
x1 = 5x1 3x2 + 8,
x2 = x1 + x2 + 32t,
x1 (0) = 2, x2 (0) = 0.
[2] Solve
x1 = 7x1 9x2 + 9x3 + et ,
x2 = 3x1 + 5x2 3x3 + 2et + et ,
x3 = 3x1 3x2 + 5x3 + 3et ,
x1 (0) = 1, x2 (0) = 0, x3 (0) = 0.
[3] Solve
x1 = 5x1 8x2 + 4x3 ,
x2 = 2x1 + 3x2 2x3 + et ,
x3 = 6x1 + 14x2 5x3 + 9t,
x1 (0) = 2, x2 (0) = 1, x3 (0) = 1.

Answers:
[1] x1 (t) = 10 12t + 3e4t + 9e2t ,
x2 (t) = 10 20t + e4t + 9e2t
[2] x1 (t) = 9et + 34 e2t +
x2 (t) =

26 t
e + 9tet ,
3
2et + 35 e2t 11
et 3tet ,
3
t
2t
t
t

x3 (t) = 6e + 3e + 3e

+ 3te

[3] x1 (t) = 38 + 4t + (4 + 6t)et + ( 32 18t)e3t ,

x2 (t) =
x3 (t) =

4
3
1
3

2t 2tet + ( 31 + 9t)e3t ,
+ t + ( 11
+ 2t)et + ( 29
+ 9t)e3t
2
6
5