IMPORTANT FACTS ABOUT THE

FUNDAMENTAL MATRIX
Since a solution matrix X(t) is a fundamental matrix for the linear homogeneous
system ẋ = A(t)x provided det X(t) 6= 0, it is easy to see that if C is any n × n
non-singular matrix then X(t) C is also a fundamental matrix. Indeed, if

X(t) = col (v1 (t), . . . , vn (t)) ,

then the columns of X(t) C are linear combinations of the columns of X(t).
But since the columns are solutions, so is any linear combination of columns
(superposition!) and so X(t) C is certainly a solution matrix. To see that it is a
fundamental matrix, we note that

det (X(t) C) = det (X(t)) det (C) 6= 0 ,

since we have both determinants on the right non-zero.

It is not surprising to see that, in fact, two fundamental matrices must differ
from one another by a non-singlar matrix. Specifically, if X(t) and Y (t) are
two fundamental matrices for the linear system ẋ = A(t)x then if uj is the jth
column of Y (t) then it must be a linear combination of the columns of X(t) or,
otherwise said, uj = X(t)cj , j = 1, 2, . . . n, where cj are suitable constant vectors.
Therefore, if we define C as the constant matrix whose columns are the vectors
cj , we have at once the y(t) = X(t) C. Since det Y (t) = det X(t) det C, and since
the determinants of both fundamental matrices are non-zero, we have that C has
non-zero determinant and hence is non- singular.

In the case of constant coefficients, the fundamental matrix eA t is therefore related

to any other fundamental matrix X(t) by

exp (A t) = X(t) C.
From this we see, evaluating at t = 0 that I = X(0)C so that we must have

exp (A t) = X(t) X −1 (0).

Let us give an example of a case in which we can compute exp (At) explicitly:
Example 1: Consider the system
!
1 1
ẋ = x.
0 1
Then we can write
! ! !
1 1 1 0 0 1
A = = + = I +N,
0 1 0 1 0 0

1
so that

exp (At) = exp (It + N t) = exp (It) exp (N t) .

Now, since N 2 = 0, we have from the definition of the exponential of a matrix as

a series, that
!
1 t
exp N t = I + N t =
0 1
and so
! ! !
et 0 1 t et tet
exp At = exp It exp N t = =
0 et 0 1 0 et
Note that if one evaluates this latter matrix at t = 0 it reduces to the identity
matrix. You should check that this matrix actually IS a solution of the original
system.

Let us suppose that the matrix X(t) is a principal fundamental matrix at to for
the linear system. Then if the initial condition is x(to ) = xo then the solution
of the non-homogeneous system ẋ = A(t)x + f is given by the Variation of
Constants Formula
Zt
x(t) = X(t)xo + X(t) X −1 (s) f (s) ds.
to

As we have seen, in the constant coefficient case the principal fundamental matrix
at 0 is exactly the exponential matrix exp (At) whose inverse is just exp (−At).
Hence if the initial condition is given for to = 0 the Variation of Constants Formula
takes the form
Zt
At
x(t) = e xo + e At
e−As f (s) ds
0

This last formula is an exact analog of the formula for a single first order equation.

Let us look at a simple computation.

Example 2: For the matrix A of the preceeding example we have computed the
exponential matrix
!
et tet
exp (At) =
0 et

Let f (t) = (e−t , t)> and let the initial condition be xo = (1, 1)> . Then the
Variation of Constants formula gives us:

2
 Zt
x(t) = exp (At) xo + exp (At) exp (−As) f (s) ds
0
 
t t
! !
t t
! Zt
−s −s
! s
e te 1 e te e −se
= +  e  ds
 
0 et 1 0 et 0 e−s
0 s
! ! !
et (1 + t) et tet −te−t
= +
et 0 et t
! ! !
et (1 + t) t2 et − t et (1 + t) + t2 et − t
= + =
et tet (1 + t)et

One of the principle problems in the application of these ideas is to find ways to
compute the exponential solution matrix eA t, the Principal Fundamental Matrix
at t = 0. One way to proceed is to solve a series of initial value problems with
the canonical unit vectors E j = (0, 0, . . . , 0, 1, 0, . . . 0)> where the 1 appears in
the j th coordinate position and there are zeros elsewhere. Here is an example:

Example 3 Given
!
−1 6
A = ,
1 −2
find the PFM at t = 0 and then solve the initial value problem ẋ = A x , x(1) =
(2, 3)> .

A standard calculation shows that the eigenvalues are λ1 = −4 and λ2 = 1 with

corressponding eigenvectors v (1) = (−2, 1)> and v (2) = (3, 1)> respectively. Hence
the general solution of the system is
! !
−2 e−4t 3 et
x(t) = c1 + c2
e−4 t et

To obtain the first column of eAt we must solve the system

−2 c1 + 3 c3 = 1
c1 + c2 = 0

which has solution c1 = −1/5, c2 = 1/5. Hence the first column for the PFM is
2
e−4t + 53 et
 
5
ϕ(1) (t) =   .
− 15 e−4t + 51 et

To get the second column, we must solve

3
 −2 c1 + 3 c3 = 0
c1 + c2 = 1

which yields c1 = 3/5 , c2 = 2/5 and so the second column is

− 56 e−4t + 56 et
 

ϕ(2) (t) =   .
3 −4t 2 t
5
e + e
5

Hence Φ(t) = col (ϕ(1) , ϕ(2) ) and so

2
e−4t + 53 et , − 56 e−4t + 65 et
 
5
Φ(t) = eA t =   .
− 51 e−4t + 51 et , 3
5
e−4t + 25 et

Note that, we want the solution at to = 1 and hence we must use the Principal
Fundamental Matrix at t = 1 in order to write down the solution. But this is easy
to obtain as Φ(t − 1) = eA(t−1) . Then the solution to the initial value problem is
simply

! !
x1 (t) 2
= Φ(t − 1)
x2 (t) −3
! !
2
5
e−4t + 35 et , − 65 e−4t + 65 et 2
=
− 15 e−4t + 15 et , 3 −4t
5
e + 25 et −3
!
1 22 e−4(t−1) − 12 e(t−1)
= .
5 −11 e−4(t−1) − 4 e(t−1)