Generalized Eigenvectors and Jordan Form

We have seen that an n n matrix A is diagonalizable precisely when the dimensions of its
eigenspaces sum to n. So if A is not diagonalizable, there is at least one eigenvalue with a
geometric multiplicity (dimension of its eigenspace) which is strictly less than its algebraic
multiplicity. In this handout, we will discuss how one can make up for this deficiency of
eigenvectors by finding what are called generalized eigenvectors, which can in turn be used
to find the Jordan form of the matrix A.
First consider the following non-diagonalizable system.
Example 1. 3 The matrix

3 1
A=
0 3

2
has characteristic polynomial ( 3)
, so
it has only one eigenvalue = 3, and the cor1
responding eigenspace is E3 = span
. Since dim(E3 ) = 1 < 3, the matrix A is not
0
diagonalizable. Nevertheless, it is still possible to solve the system

dy
= Ay
dt
without much difficulty. Writing out the two equations
dy1
= 3y1 + y2
dt
dy2
= 3y2
dt
we see that the second equation has general solution y2 = c1 e3t . Plugging this into the first
equation gives
dy1
= 3y1 + c1 e3t
dt
This is a first order linear equation that can be solved by the method of integrating factors.
Its solution is y1 = c1 te3t + c2 e3t . Thus the general solution of the system is

y1
c1 te3t + c2 e3t
1
0
3t
3t 1
y=
=
= c1 e
t
+
+ c2 e
y2
c2 e3t
0
1
0
3
The matrix A in the previous example is said to be in Jordan form.
Definition 1. A 2 2 matrix of the form

1
J=
0

for some number is called a Jordan form matrix.

1

The following theorem generalizes the calculations of the previous example.

Theorem 1. The general solution of the system

dy
= Jy,
dt
is

1
J=
0

1
0
t 1
y = c1 e
t
+
+ c2 e
0
0
1
t

In general, suppose A is a 22 matrix with a single repeated eigenvalue with dim(E ) =

1. Then, as we have seen before, making the change of variables y = Cx transforms the
system
dy
= Ay
dt
into the system
dx
= C 1 ACx.
dt
We now want to find a matrix C such that the coefficient matrix C 1 AC for the new system
is a Jordan form matrix J. That is, we want AC = CJ. Writing

|
|

1
C = v1 v2 ,
J=
,
0
|
|
we have

|
|
CJ = v1 v2 + v1
|
|

|
|
AC = Av1 Av2 ,
|
|

Therefore the columns of C must satisfy

Av1 = v1
Av2 = v2 + v1
Thus the vector v1 is an eigenvector with eigenvalue . Rewriting these equations
(A I)v1 = 0
(A I)v2 = v1
it follows that (A I)2 v2 = (A I)v1 = 0. Thus the vector v2 must be in ker(A I)2 .

Definition 2. A nonzero vector v which satisfies (AI)p v = 0 for some positive integer
p is called a generalized eigenvector of A with eigenvalue .
The vectors v1 and v2 form a generalized eigenvector chain, as the following diagram
illustrates:
v2 v1 0
AI

AI

Therefore, to find the columns of the matrix C that puts A in Jordan form, we must find
a chain of generalized eigenvectors, as follows:
Find a nonzero vector v2 in ker(A I)2 that is not in ker(A I).
Set v1 = (A I)v2 .
Having found such vectors, we can then write down the solution of the system
From above, we know that the solution of the system dx
= Jx is
dt


1
0
t
t 1
x = c1 e
t
+
+ c2 e
0
1
0

dy
dt

= Ay.

so multiplying this by C we get

y = Cx = c1 et (tv1 + v2 ) + c2 et v1 .
In summary, we have the following result.
Theorem 2. Let A be a 2 2 matrix and suppose v2 v1 0 is a chain of generalized
eigenvectors of A with eigenvalue . Then the general solution of the system dy
= Ay is
dt
y(t) = c1 et (tv1 + v2 ) + c2 et v1 .
Example 2. 3 Let

2 1
A=
1 4

The characteristic polynomial of A is 2 6 + 9 = ( 3)2 , so = 3 is the only eigenvalue

of A. Next, we compute

1 1
0 0
2
A 3I =
,
(A 3I) =
1 1
0 0
Now we choose v2 to be any vector in ker(A 3I)2 that is not in ker(A
3I). One such
0
1
vector is v2 =
. With this choice, we than have v1 = (A 3I)v2 =
. The solution of
1
1
the system dy
= Ay is therefore
dt


1
0
3t
3t 1
y = c1 e
t
+
+ c2 e
.
1
1
1
3
3

For larger systems it turns out that any chain of generalized eigenvectors of the coefficient
matrix can be used to write down a linearly independent set of solutions.
Theorem 3. Suppose {v1 , v2 , . . . , vk } is a chain of generalized eigenvectors of A with
eigenvalue . That is,
vk

AI

vk1

AI

vk2

AI

AI

v2

AI

v1

AI

0.

Then
y1 (t) = et v1
y2 (t) = et (v2 + tv1 )

t2
t
y3 (t) = e v3 + tv2 + v1
2!
..
.

t2
tk2
tk1
t
yk (t) = e vk + tvk1 + vk2 + +
v2 +
v1
2!
(k 2)!
(k 1)!
is a linearly independent set of solutions of

dy
dt

= Ay.

Proof. We show that yk is a solution. The proof that the others are solutions is similar.
Differentiating gives

dyk
t2
t
= e (vk + vk1 ) + t(vk1 + vk2 ) + (vk2 + vk3 ) +
dt
2!

k2
k1
t
t
+
(v2 + v1 ) +
(v1 ) .
(k 2)!
(k 1)!
Multiplying by A gives

t2
tk2
tk1
t
Ayk = e (Avk ) + t(Avk1 ) + (Avk2 ) + +
(Av2 ) +
(Av1 ) .
2!
(k 2)!
(k 1)!
The two expressions above are equal, since Av1 = v1 , Av2 = v2 + v1 , and so on.
Next we show that the chain {v1 , v2 , . . . , vk } is necessarily linearly independent. Suppose
c1 v1 + c2 v2 + + ck vk = 0.

(1)

The multiplying by (A I)k1 sends v1 through vk to zero, and vk to v1 , so we are left

with ck v1 = 0. Since v1 is nonzero, this implies ck = 0. Similar reasoning shows that
the remaining coefficients must also be zero. Thus the chain of generalized eigenvectors is
linearly independent. Now suppose
c1 y1 (t) + c2 y2 (t) + + cn yn (t) = 0
for all t. Then since yj (0) = vj , at t = 0 this equation is precisely equation (1) above, which
we just showed implies c1 = c2 = = ck = 0.
4

The next theorem guarantees that we can always find enough solutions of this form to
generate a fundamental set of solutions.

Theorem 4. Let A be an nn matrix, and suppose is an eigenvalue of A with algebraic

multiplicity m. Then there is some integer p m such that dim(ker(A I)p ) = m.

Example 3. 3 Let

5 1 4
A = 4 3 5
3 1 2

The characteristic polynomial of A is pA () = ( 2)3 . Since

3 1 4
1 0 1
A 2I = 4 1 5 0 1 1 ,
rref
3 1 4
0 0 0
the eigenspace E2 = ker(A 2I) has dimension 1. Next, since

1 0 1
1 0 1
(A 2I)2 = 1 0 1 0 0 0 ,
rref
0 0 0
1 0 1
the space ker(A 2I)2 has dimension 2. Finally, since

0 0 0
(A 2I)3 = 0 0 0 ,
0 0 0
the space ker(A 2I)3 has dimension 3, which matches the algebraic multiplicity of the
eigenvalue = 2.
We can now form a chain of 3 generalized eigenvectors by choosing a vector v3 in ker(A
2I)3 and defining v2 = (A 2I)v3 and v1 = (A 2I)v2 = (A 2I)2 v3 . To ensure that v2
and v1 are both non-zero, we need v3 to not be in ker(A 2I)2 (which in turn implies that
v3 is not in ker(A 2I)). Since ker(A 2I)3 = R3 , we can choose v3 to be any vector not
in ker(A 2I)2 . Since the first column of (A 2I)2 is non-zero, we may choose



1
3
1
v3 = 0
=
v2 = (A 2I)v3 = 4
=
v1 = (A 2I)v2 = 1 .
0
3
1

Then

1
2t
y1 (t) = e 1
1

1
3
2t

y2 (t) = e
t 1 + 4
1
3


3
1
2 1
2t t

1 + t 4 + 0
y3 (t) = e
2!
1
3
0

is a fundamental set of solutions of

dy
dt

= Ay.

There may in general be more than one chain of generalized eigenvectors corresponding
to a given eigenvalue. Since the last vector in each chain is an eigenvector, the number of
chains corresponding to an eigenvalue is equal to the dimension of the eigenspace E .
Example 4. 3 Let

4 1 1
A = 2 1 2 .
1 1 4

The characteristic polynomial of A is pA () = ( 3)3 . Since

1 1 1
1
1
1
A 3I = 2 2 2 0 0 0 ,
rref
0 0 0
1
1
1
the eigenspace E3 = ker(A 3I) has dimension 2,

0
(A 3I)2 = 0
0

so there will be two chains. Next, since

0 0
0 0 ,
0 0

the space ker(A 3I)2 has dimension 3, which matches the algebraic multiplicity of = 3.
Thus one of the chains will have length 2, so the other must have length 1.
We now form a chain of 2 generalized eigenvectors by choosing v2 in ker(A 3I)2 such
that v2 is not in ker(A 3I). Since every vector is in ker(A 3I)2 and the first column of
A 3I is non-zero,we may again choose


1
1

v2 = 0
=
v1 = (A 3I)v2 = 2
0
1
To form a basis for R3 , we now need one additional chain of one generalized eigenvector.
This vector must be an eigenvector independent from v1 . Since

1
1

1 , 0 ,
E3 = span
0
1
6

and neither of these spanning vectors is itself a scalar multiple of v1 , we may choose either
one of them. So let

1

w1 = 1
0
Then we have the two chains
v2

v1
w1

0
0.

The first chain generates the two solutions

1
y1 (t) = e3t 2
1

1
1
3t

y2 (t) = e
t 2 + 0
1
0

and the second chain generates a third solution

1
y3 (t) = e3t 1
0
and together y1 , y2 and y3 form a fundamental set of solutions of
Example 5. 3 Let

5
0
A=
0
0
The characteristic polynomial of A is

0 1
0 0
A 5I =
0 0
0 0

1
5
0
0

dy
dt

= Ay.

3 2
0 3

5 1
0 5

( 5)4 . Since

3 2
0 3

0 1 rref
0 0

the eigenspace E5 = ker(A 5I) has dimension 2, and

0 0 0

0 0 0
(A 5I)2 =
0 0 0
0 0 0

0
0

0
0

1
0
0
0

3
0
0
0

0
1
,
0
0

there are two chains. Since

0
0

0
0

the dimension of the space ker(A 5I)2 is 4, matching the algebraic multiplicity of = 5.
7

To form a chain of length 2, we first choose a vector v2 in ker(A 5I)2 which is not in
ker(A 5I). Let


0
1
1
0

v2 =
=
v1 = (A 5I)v2 =
0
0 .
0
0
We now need one more chain of length 2. To find a second chain, we need to choose another
vector w2 in ker(A 5I)2 which is not in ker(A 5I), in such a way that w2 is independent
of v2 and w1 = (A 5I)w2 is independent of v1 . One such choice is


2
0
3
0

w2 =
=
w1 = (A 5I)w2 =
1 .
0
1
0
Thus

y1 (t) = e5t
0
0

2

y3 (t) = e5t
1
0

is a fundamental set of solutions of

1
0

0 + 1
y2 (t) = e5t
t
0 0
0
0

2
0

3 + 0
y4 (t) = e5t
t
1 0
0
1
dy
dt

= Ay.

Jordan Form for n n Matrices

Definition 3. A matrix of the form

1
1

... ...

...

consisting of in each entry along the main diagonal, 1 in each entry directly above the
main diagonal, and 0 elsewhere, is called an elementary Jordan block. A matrix of the
form

J1

J2

.
.

.
Jk
8

where each Ji is an elementary Jordan block (with possibly different values of ) is called a
Jordan form matrix.
Example 6. 3 The following are

2 1 0
2
0 2 0 , 0
0 0 3
0

examples of Jordan form matrices:

1 0
3 1
0
2 0 0
2 1 , 0 3 0 , 0 1 0
0 2
0
0 3
0 0 3
3

Theorem 5. For any n n matrix A, there is a matrix C and a Jordan form matrix J such
that C 1 AC = J.
The Jordan matrix J is determined by the number and length of the generalized eigenvector chains for A.
Example 7. 3 Let A be the matrix in Example 3. Since
Av1 = 2v1
Av2 = v1 + 2v2
Av2 = v2 + 2v3 ,
letting

implies

|
|
|
C = v1 v2 v3
|
|
|

|
|
|
|
|
|
2 1 0
AC = 2v1 v1 + 2v2 v2 + 2v3 = v1 v2 v3 0 2 1 = CJ,
|
|
|
|
|
|
0 0 2

so C 1 AC = J consists of a single elementary Jordan block.

Example 8. 3 Let A be the matrix in Example 4. Since

Av1 = 3v1
Av2 = v1 + 3v2
Aw1 = 3w1 ,
letting

implies

|
|
|
C = v1 v2 w1
|
|
|

|
|
|
|
|
|
3 1 0
AC = 3v1 v1 + 3v2 w1 = v1 v2 w1 0 3 0 = CJ,
|
|
|
|
|
|
0 0 3

so C 1 AC = J consists of 2 elementary Jordan blocks, one 2 2 and one 1 1.

9