Lecture 24

Jordan-Canonical Form

We know that not every matrix is similar to a diagonal matrix. Here, we discuss the simplest matrix to
which a square matrix is similar. This simplest matrix coincides with a diagonal matrix if the matrix is
diagonalizable.

Definition 1. A square matrix A is called block diagonal if A has the form

0 1
A1 0 . . . 0
B 0 A2 . . . 0 C
B C
B .. .. .. C ,
@ . . ... . A
0 0 . . . Ak

where Ai is a square matrix and the diagonal entries of Ai lie on the diagonal of A.
HA3
Definition 2. Let 2 C. A Jordan block J( ) is an upper triangular matrix whose all diagonal entries
are , all entries of the superdiagonal (entries just above the diagonal) are 1 and other entries are zero.
Therefore, 0 1
1 0 ... 0

Eiiii
B0 1 . . . 0C
B C
B C
J( ) = B ... ... . . . ... C.
B C
@0 0 . . . 1A
0 0 ... 0
Definition 3. A Jordan form or Jordan-Canonical form is a block diagonal matrix whose each block
is a Jordan block, that is, Jordan form is a matrix of the following form
0
J1 0 . . . 0
1
PAP J
B 0 J2 . . . 0 C
B C
B .. .. .. C .
@ . . ... . A
0 0 . . . Jk

Definition 4. Let T : V ! V be a linear transformation and 2 C. A non-zero vector v 2 V is called

a generalized eigenvector of T corresponding to ⇤ if (T I)p (v) = 0 for some positive integer p.

The generalized eigenspace of T corresponding to , denoted by K , is the subset of V defined by

K = {v 2 V | (T I)2
p
(v) = 0 for some natural number p}.

Remark 5. 1. If v 2 V is a generalized eigenvector of a linear transformation T corresponding to 2 C,

then is an eigenvalue of T.
2. The generalized eigenspace K is a subspace of V and T x 2 K for all x 2 K .
3. Let E be the eigenspace corresponding to . Them E ⇢ K .

Theorem 6. Let J be an m ⇥ m Jordan block with eigenvalue . Then characteristic polynomial of J is

equal to its minimal polynomial, that is pJ (x) = (x )m = mJ (x).
0
1
5 2 7 0
G
G 1

a d x 2 135
 y
mini x 1 x d b x 13 A 1 A d A 13
m
5
Proof. Note that J is an upper-triangular matrix, hence the characteristic polynomial is (x ) and
the minimal polynomial is0(x )k for some11  k  m. Here,0we claim that (J 1 I)k 6= 0 for k < m.
0 1 0 ... 0 0 1 0 ... 0
B0 0 1 . . . 0 C B0 0 1 . . . 0 C
B C B C
B C B C
Observe that, J I = B ... ... . . . ... C, (J I)2 = B ... ... . . . ... C and (J I)m 1 =
B C B C
@0 0 . . . 0 1 A @0 0 . . . 0 0 A
0 0 ... 0 0 0 0 ... 0 0
0 1
0 1 0 ··· 0
B0 0 0 · · · 0 C
B C
B .. .. . .C
B . . · · · .. .. C so that the minimal polynomial of J is (x )m .
poly
B C
@0 0 · · · 0 0 A
β A 0 where P is ch
0 0 ··· 0 0
0 where m is minimal poly
Remark 7. 1. If a matrixMalAI
A is similar to a Jordan block of order m with eigenvalue , then there exist
an invertible matrix P such that P 1 AP = J. Let Xi be the i-th column of P . Then {X1 , X2 , . . . , Xm }
is a basis of Rm , which is called Jordan basis.
2. The vector X1 is an eigenvector corresponding to and Xj 1 = (A )Xj for j = 2, . . . , m.

Theorem 8. An m ⇥ m matrix A is similar to an m ⇥ m Jordan block J with eigenvalue if and

only if there exist m independent vectors X1 , X2 , . . . , Xm such that (A I)X1 = 0, (A I)X2 =
X1 , . . . , (A I)Xm = Xm 1 . 2
✓
3 1
◆ trace Al 0 detA
Example 9. Consider A = . Then the characteristic polynomial and minimal polynomial are
1 1
the same which is (x 2)2 . Hence the matrix in not diagonalizable. Here, (1, 1) is an eigenvector
corresponding to 2. If J is the Jordan form of A, then we have a basis {X1 , X2 } with respect to which
the matrix representation of A is J. By previous theorem X1 is an eigenvector and X ✓2 can◆be found by
1 1
solving (A 2I)X2 = X1 . Set X1 = (1, 1), then X2 = (1, 0). Now construct P = and verify
✓ ◆ 1 0
XI X2 I
CA
I
2 1
that P 1 AP = J, where J = .

4
0 2
Him
Theorem 10. Let A be an n ⇥ n matrix with the characteristic polynomial (x 1)
r1
· · · (x k)
rk
, where
i ’s are distinct. Then A is similar to a matrix of the following form
0 1 various o
J1 0 . . . 0
B 0 J2 . . . 0 C
B
B .. ..
@ . . . . . .. A
C
. C, P
0 0 . . . Jk
p A P 82
where J1 , J2 , . . . , Jk are Jordan blocks. The matrix J is unique except for the order of the blocks J1 , J2 , . . . , Jk .

Remark 11. 1. The sum of orders of the blocks corresponding to i is ri (the A.M.( i )).
2. The order of the largest block associated to i is si , the exponent of x i in the minimal polynomial
of A.
3. The number of blocks associated with the eigenvalue i is equal to the GM ( i ).
4. Knowing the characteristic polynomial and the minimal polynomial and the geometric multiplicity of
each eigenvalue i need not be sufficient to determine Jordan form of a matrix.

3 2
Ch poly x 2 min 2 x 2
 6M 2 I
Example 12. Let A be a matrix with characteristic polynomial (x 1)43 (x 2)2 and minimal polynomial
(x 1)2 (x 2). Then we can find the Jordan form J of A by using above remarks,
(i) The eigenvalue 1 appears on the diagonal 3 times, and 2 appears 2 times.
(ii) The largest Jordan block corresponding to = 1 is of order 2 (exponent of (x 1) in the minimal
polynomial), and the largest Jordan block corresponding to = 2 is of order 1.
(iii) The number number of Jordan blocks corresponding to = 1 is 2 where one block is of order 2 and
other is of order 1. (iv) The number number of Jordan blocks corresponding to = 2 is 2 where both the
blocks are of order 1. Therefore, the Jordan form of A is
0✓ ◆ 1
1 1
B 0 1 C
c
B C
B (1) C.
B C
@ (2) A
(2)

Example 13. Let A be a matrix with characteristic polynomial (x 1)3 (x 2)2 and minimal polynomial
(x 1)3 (x 2)2 . Then
(i) The eigenvalue 1 appears on the diagonal 3 times, and 2 appears 2 times.
(ii) The largest Jordan block corresponding to = 1 is of order 3 (exponent of (x 1) in the minimal
polynomial), and the largest Jordan block corresponding to = 2 is of order 2.
(iii) The number number of Jordan blocks corresponding to = 1 is 1. (iv) The number number of Jordan
blocks corresponding to = 2 is 1. Therefore, the Jordan form of A is
00 1 1
1 1 0
B@0 1 1 A C
B C
B 0 0 1 C
B ✓ ◆C .
@ 2 1 A
0 2

Example 14. minimal and characteristic are not always sufficient Let A be a matrix with char-
acteristic polynomial (x 1)4 and minimal polynomial (x 1)2 . Then
(i) The eigenvalue 1 appears on the diagonal 4 times.
(ii) The largest Jordan block corresponding to = 1 is of order 2 (exponent of (x 1) in the minimal
polynomial).
(iii) The number number of Jordan blocks corresponding to = 1 is GM(1) which is not known. Note
that GM (1)  4 as minimal polynomial confirms that A is not diagonalizable. Also, GM (1) 6= 1, if
GM (1) = 1, the the Jordan matrix has only one block corresponding to = 1 which must be of order 4,
which is not true.
(iv) Thus GM (1) = 2 or 3.

(v) If GM (1) = 2 the Jordan form of A is

0✓ ◆ 1
1 1
B 0 1 C
B ✓ ◆C .
@ 1 1 A
0 1

3
(vi) If GM (1) = 3, the Jordan form of A is
0✓ ◆ 1
1 1
B 0 1 C
B C.
@ (1) A
(1)

Example 15. Possible Jordan forms for a given characteristic polynomial Let A be a matrix
with characteristic polynomial (x 1)3 (x 2)2 . Then choices of minimal polynomials are
(i)(x 1)(x 2), then Jordan form is the diagonal matrix.
(ii) (x 1)2 (x 2), Example 12. 00 1 1
1 1 0
B@0 1 1 A C
B C
3 B
(iii) (x 1) (x 2), the Jordan form is B 0 0 1 C.
C
@ (2) A
(2)
0 1
(1)
B (1) C
2 B
B C
(iv) (x 1)(x 2) , B (1) ✓ C
◆C .
@ 2 1 A
0 2
0✓ ◆ 1
1 1
B 0 1 C
B C
(v) (x 1)2 (x 2)2 , B B (1) ✓
C
◆C .
@ 2 1 A
0 2
3 2
(vi) (x 1) (x 2) , Example 13.

Example 16. (minimal, characteristic and GM ( ) are not always sufficient) Let A be a matrix
with characteristic polynomial (x 1)7 and minimal polynomial (x 1)3 and GM (1) = 3. Then there are
two possible Jordan forms (write the corresponding Jordan forms yourself !):
(i) One Jordan block of order 3 and other two blocks of order 2.
(ii) Two Jordan blocks of order 3 and one of order 1.

0 1
1 1 0
Example 17. Find a Jordan basis Let A = @1 1 1A. The characteristic polynomial of A is
0 1 0 1 1
0 1 0
(x 1) and A I = 1 0 1A . Thus nullity(A I) is 1=GM(1). Therefore, the Jordan form of
3 @
0 1 0 1 0
1 1 0
A is J = 0 1 1A. The problem is to find a Jordan basis or a matrix P such that P 1 AP = J.
@
0 0 1
P = [X1 X2 X3 ], where (A I)X1 = X1 , (A I)X2 = X1 , (A I)X3 = X2 . On 0solving, we 1 get
1 1 1
X1 = (1, 0, 1), X2 = (1, 1, 1) or ( 1, 1, 1) and X3 = (1, 1, 0) or (0, 1, 1). Hence, P = @ 0 1 1A .
1 1 0

4
Example 18. (Finding a Jordan basis is not always straight forward)
0 1 0 1
3 1 0 1 1 0
Let A = @ 1 1 0A. The characteristic polynomial of A is (x 2)3 and A 2I = @ 1 1 0A .
1 1 2 0✓ ◆1 1 1 0
2 1
Thus nullity(A I2) is 2=GM(2). Therefore, the Jordan form of A is J = @ 0 2 A. The
(2)
1
problem is to find a Jordan basis or a matrix P such that P AP = J. P = [X1 X2 X3 ]. Here, we
get an eigenvector (x, y, z) satisfies x + y = 0, two independent eigenvectors are (0, 0, 1) and ( 1, 1, 0).
Note that each eigenvector corresponds to a Jordan 0 block. Thus,
1 0set 1 X1 0
= (0,
1 0, 1), (A 2I)X2 = X1 ,
1 1 0 x 0
X3 = ( 1, 1, 0) or . But, (A 2I)X2 = X1 ) @ 1 1 0A @y A = @0A which is an inconsistent
1 1 0 z 1
system. Similarly, (A 2I)X2 = X1 , where X1 = ( 1, 1, 0) is inconsistent. For finding a Jordan basis,
we will
0 change the eigenvector,
1 let X1 = ( 1, 1, 1), then X2 = (0, 1, 0), and X3 = ( 1, 1, 0) Hence,
1 0 1
P =@ 1 1 1 A.
1 0 0