Repeated Eigenvalues

Occasionally when we have repeated eigenvalues, we are still able to nd the correct number
of linearly independent eigenvectors. Take for example

3 1 2
3 1 6
2 2 2

One can verify that the eigenvalues of this matrix are = 2, 2, 4. We can nd the eigenvalue
corresponding to = 4 using the usual methods, and nd u
4
=

1
3
2

. For the = 2 case,

we must solve the system

3 2 2 2
3 1 2 6
2 2 2 2

x
y
z

0
0
0

i.e.

1 2 2
3 3 6
2 2 4

x
y
z

0
0
0

The second and third equations are just multiples of the rst equation, so we expect to have 2
parameters in our solution. Note that this will not always be the case for a 3x3 matrix.
Let z = s and y = t. Then x = y 2z = s 2t, from the rst equation. So our eigenvector is

s 2t
s
t

= s

1
1
0

+ t

2
0
1

We can see that there are two linearly independent vectors here, and each will be an eigen-
vector for = 2. In this case the solution to the dierential equation
x =

3 1 2
3 1 6
2 2 2

x
is
C
1
e
4t

1
3
2

+ e
2t

C
2

1
1
0

+ C
3

2
0
1

However, sometimes we cant nd two linearly independent eigenvectors in this way. Consider
the dierential equation

x
y

10 2
18 2

x
y

1
Here we nd a repeated eigenvalue of = 4. To nd the eigenvector(s), we set up the system

6 2
18 6

x
y

0
0

These equations are multiples of each other, so we can set x = t and get y = 3t. So there is
only one linearly independent eigenvector,

1
3

. This will give us one solution to the dierential

equation, but we need to nd another one. Instead of just multiplying our previous solution by
t, we guess (without much justication) that a solution will be

x
y

= te
4t

1
3

+ e
4t

A
B

Where now we need to nd A and B. We can nd these constants by plugging in our solution to
the dierential equation, and solving the system that we get. Our nal solution to the problem
will be

x
y

= C
1

te
4t

1
3

+ e
4t

A
B

+ C
2
e
4t

1
3