Cayley-Hamilton Theorem

This theorem was first established by Hamilton in 1883 for a particular type of matrices and
was later on stated by Cayley in 1885.

Matrix Polynomials
Definition An expression of the form

F ( λ )  A0  A1 λ  A2 λ 2  ...  Am1 λ m1  Am λ m , (1)

where A0, A1, A2...,Am all are square matrices of the same order is called a Matrix Polynomial
of degree m provided Am is not a null matrix. The symbol λ is called indeterminate. If the
order of each of the matrix coefficient is n, then we say that the matrix polynomial is n-
rowed.

Two matrix polynomials are equal if and only if the coefficients of like powers of 
are the same.

Example 1 Let A be square matrix of order 3 given by

0 1 1
A  1 0 1 
1 1 0 

Let us compute adj (A- I ) for this matrix

2  1 λ 1  1 
 
adj ( A  I )    1 λ 1
2
 1  .
  1 λ 1 2  1


The highest power of  occurring in any elements of adj (A- I ) is 2. Rewriting each element
as a quadratic in  , supplying missing coefficient with zero, we get

2  0  1 02  λ  1 02  λ  1 

 
adj ( A  I )  02    1 λ 2  0  1 02  λ  1  .
02    1 02  λ  1 λ 2  0  1


Obviously adj (A- I ) can be written as the matrix polynomial

1 0 0 0 1 1  1 1 1
adj (A- I )   0 1 0    1
 2
0 1    1
 1 1

 0 0 1  1 1 0  1 1 - 

Remark: If A is a square matrix of order n, then adj (A- I ) is a matrix polynomial of degree
at most ( n – 1 ).
CAYLEY- HAMILTON THEOREM: Every square matrix satisfies its own characteristic
equation i.e. if for a square matrix A of order n,


| A -  In |  (1) n n  p1 n 1  p 2 n  2    p n ,  (2)

then the characteristic equation | A -  In | = 0 is satisfied by A i.e.

An  p1 An1  p2 An2    pn I n  0 (3)

In other words if  is replaced by A in the characteristic equation, then it is satisfied.

Proof The elements of the matrix (A -  In) are of at the most first degree in  . Hence
adj (A-  In) may be expressed as a matrix polynomial of degree at most (n-1) in powers of 
and has the form given as:

adj (A-  In) = B0  n-1 + B 1  n-2

+...+ Bn-2  + Bn-1 , (4)

where B0 , B1 , ..., Bn 1 are all n×n matrices whose elements are functions of aij’s elements of
matrix A. To understand this read the example 1given before this theorem in which a matrix
A of order 3, which is a function of  ,is expressed in powers of  whose all coefficient are
matrices of order 3.

We know that A (adj. A) = |A| In

Hence  A  I n  adj. ( A  I n )  | A  I n | I n


or  A  I n  ( B0 n1  B1 n2    Bn2   Bn1 )   1 n  p1 n1    pn1   pn I n
n

Comparing the coefficients of like powers of  in both sides of above, we get

 B0   1 I n
n

A B0  B1   1 p1 I n
n

A B1  B2   1 p 2 I n
n

...................................
ABn  2  Bn 1   1 p n1 I n
n

ABn 1   1 p n I n
n

Pre-multiplying the above equation successively by A n , A n 1 , ,I n and adding we get

0   1
n
A n
 p1 A n 1  p 2 A n  2    p n 1 A  p n I n 
or A n  p1 A n 1  p 2 A n 2    p n 1 A  p n I n  0

Corollary If A be a non-singular matrix,|A|  0 and pn  0, then multiplying Eq. (3) by A-1,

we get

A 1  
1
pn

A n 1  p1 A n  2    p n  2 A  p n 1 I n . 
Example 2 Obtain the characteristic equation of the matrix

1 0 2
A   0 2 1 
 2 0 3 

and verify that it is satisfied by A and hence find its inverse.

Solution We have

1 λ 0 2
|A  λI|  0 2 λ 1
2 0 3 λ

 1    2    3     2 [0  22   ]  2  λ  [ 1  λ  3  λ   4]

 
 2    2  4  1   3  62  7  2  
 The characteristic equation of A is

3  62  7  2  0 (i)

By Cayley – Hamilton theorem

A3  6 A2  7 A  2 I  0 (ii)

Verification of the above equation

1 0 2  1 0 2  5 0 8 
A 2   0 2 1    0 2 1    2 4 5 
 2 0 3   2 0 3  8 0 13 

1 0 2 5 0 8   21 0 34 
A   0
3
2 1  2
 4 5   12
 8 23 
 2 0 3   8 0 13   34 0 55 
  21 0 34 
Now A  6 A  7 A  2 I  12
3 2
8 23 
 34 0 55 

5 0 8  1 0 2 1 0 0
 6  2 4 5   7  0 2 1   2  0 1 0 
 8 0 13   2 0 3   0 0 1 

 30 0 48   30 0 48  0 0 0
 12 24 30   12 24 30    0 0 0 
 48 0 78   48 0 78   0 0 0 

Hence Cayley – Hamilton theorem is verified. Now we shall complete A-1 .Multiplying (ii) by
A-1, we get

A1  
1 2
2

A  6 A  7I 
5 0 8  1 0 2 1 0 0
1  7 
  2 4 5   3  0
 2 
1  0 1 0 
2 2 
8 0 13   2 0 3   0 0 1 

 3 0 2 
 1 1 
  1
 2 2 
2 0  1


1 2
Example 3 If A    , express A  4 A  8 A  12 A  14 A as a linear polynomial
6 5 4 3 2

  1 3 
in A. Also verify Cayley- Hamilton Theorem.

Solution The characteristic equation is

1  2
| A  I |  2  4  5  0
1 3

By Cayley-Hamilton theorem

A2  4 A  5I  0 .

This gives A2  4 A  5I and A3  A( 4 A  5I )  4 A2  5 A  11A  20 I

Similarly A4  24 A  55I , A5  41 A  120 I , A6  44 A  205 I

Now A6  4 A5  8 A 4  12 A3  14 A 2  44 A  205I  4( 41 A  120 I ) +8 (24 A  55 I )

  12( 11A  20 I )+14 (4 A  5 I )

  4 A  5I .

Verification of Cayley-Hamilton Theorem

1 2 1 2  1 8
A2     ,
 1 3   1


3  4 7

 1 8 1 2 1 0 0 0
A2  4 A  5 I    4   5   .
 4 7  1 3 0 1 0 0

Hence Cayley – Hamilton theorem is verified.

Example 4 Find the characteristic equation of the matrix

2 1 1 
 1 2  1 .

1 1 2 

Show that the equation is satisfied by A and hence find A-1.

Solution The characteristic equation of A is

2 1 1
| A  I |  1 2 1  0
1 1 2

or 3  62  9  4  0. (i)

To verity Cayley- Hamilton theorem, we have to show that

A3 – 6A2 + 9A – 4 I = 0

2 1 1  2 1 1  6 5 5 
Now A =  1
2
2 1  1 2 1    5 6 5
 
1 1 2  1 1 2  5 5 6 

6 5 5  2 1 1  22 21 21 
3  5 6 5  1 2 1   21 22 21
A =  
5 5 6  1 1 2  21 21 22 

22 21 21  6 5 5 
 A  6 A  9 A  4 I   21
3 2
22 21  6  5
 6 5
21 21 22  5 5 6 
 2 1 1  1 0 1  0 0 0
 9   1 2 1  4 0
 1 0    0 0 0 
1 1 2   0 0 1   0 0 0 

Now pre-multiply the equation (i) by A-1, we have

A2  6 A  9 I  4 A1  0

or 4 A1  A2  6 A  9 I

6 5 5  2 1 1  1 0 0
   5 6  5   6  1
 2 1  9  0
 1 0 
5 5 6  1 1 2   0 0 1 

3 1 1 3 1 1
1 
 1 3  1
1 , or A  1 3 1 
4 
 1 1 3   1 1 3 

2 1 1
Example 5 Obtain the characteristic equation of the matrix A= 0 1 0  .

1 1 2

and hence compute the matrix

A8  5 A 7  7 A 6  3 A 5  A 4  5 A 3  8 A 2  2 A  I .

Solution The characteristic equation of the matrix A is

2λ 1 1
0 1 λ 0  0
1 1 2 λ

or 3  52  7   3  0

According to Cayley – Hamilton theorem, we have

A3  5 A2  7 A  3I  0 .

Using this result the given expression is

A8  5 A 7  7 A 6  3 A 5  A 4  5 A 3  8 A 2  2 A  I

 A5 ( A3  5 A2  7 A  3I )  A ( A3  5 A2  7 A  3I )  A2  A  I

 A2  A  I
 2 1 1 2 1 1  2 1 1  1 0 0
 0 1 0 0
 1 0  0 1 0  0 1 0
1 1 2 1 1 2 1 1 2 0 0 1 

5 4 4 2 1 1  1 0 0  8 5 5
  0 1 0    0
 1 0    0 1 0    0 3 0  .
 4 4 5  1 1 2   0 0 1   5 5 8 

Example 6 Show that the matrix A satisfies the Cayley – Hamilton theorem where

0 c b 
A   c 0 a  .
 b a 0 

Solution The characteristic determinant is

0 λ c b
c 0 λ a 
  λ λ 2  a 2 )  c ( cλ  ab )  b ac  bλ  
b a 0 λ


  3   a 2  b 2  c 2 . 
The characteristic equation of the matrix A is

3   a 2  b 2  c 2   0 .

We have to verify that


A3  a 2  b 2  c 2 A  0 . 
0 c b  0 c b  0 c b 
A 3    c 0 a   c
 0 a   c
 0 a 
 b a 0   b a 0   b a 0 

 c 2  b 2 ab ac  0 c b 
   c
  ab c  a2 2
bc   0 a 
 ac bc 2
b  a 
2  b a 0 


0 c 3  b 2c  a 2c bc 2  b 3  a 2 b 
 
 c 3  a 2 c  b 2 c 0  ab 2  ac 2  a 3 
  bc 2  b 3  a 2 b ac 2  ab 2  a 3 0 
 
 0 c b

= - a b c 2 2 2
  c
 0 a    (a 2  b 2  c 2 ) A
 b a 0 

Hence A3+ (a2 + b2 + c2) A = 0, and the matrix satisfies Cayley-Hamilton theorem.

Diagonalization of a Matrix
We shall recall the definition and some properties of similar matrices.

Let A and B be two square matrices of order n. Then B is said to be similar to A if there exists
a non-singular matrix P such that

B  P 1 AP (5)

Properties of similar matrices are

(i) Similarity of matrices is an equivalence relation.

(ii) Similar matrices have the same determinant.

(iii) Similar matrices have the same characteristic polynomial and hence the same eigen
values.

Definition A matrix A is said to be diagonalizable if it is similar to a diagonal matrix.

Thus a matrix A is diagonalizable if there exists an invertible matrix P such that P 1 AP  D ,

where D is a diagonal matrix given by

 λ1 0 0 ... 0 
 
D = 0 λ2 0 ... 0  , where λs are eigen values of A.
 ... ... ... ... 
 
 0 0 0 ... λ n 

THEOREM An n × n matrix is diagonalizable if and only if it possesses n linearly

independent eigen vectors.

Proof Suppose A is diagonalizable. Then A is similar to matrix D= diag[ λ1 , λ2 , ,λn ] .


Therefore, there exists an invertible matrix P  X 1 , X 2 ..., X n such that 
P 1 AP  D
or
AP  PD

   
i.e. A X 1 , X 2 , ,X n  X 1 , X 2 , ,X n diag λ1, , λ2 , ,λn  
  
or AX 1 , AX 2 ,...,AX n  λ1 X 1 , λ2 X 2 ...,λn X n 
or AX 1  λ1 X 1 , AX 2  λ2 X 2 ,..., AX n  λn X n .

Therefore X1, X2,..,Xn are eigen vectors of A corresponding to the eigen values 1 , 2 ,  , n
respectively. Since the matrix P is non-singular so vectors X1, X2,...,.Xn are linearly
independent.

Working rule: Write the characteristic eq. | A  I | = 0. Find Eigen values. Let 1 2 , 3 be
the eigen values of a third order square matrix and (a1, b1, c1 )' , (a2 , b2 , c2 )' , (a3 , b3 , c3 )' be
the corresponding eigen vectors, then matrix P will be

a1 a2 a3 
 
P  b1 b2 b3  , and calculate P-1 .
c c2 c3 
1

Then A is diagonalized by using the formula (5)

Powers of a Matrix A

Consider the relation D = P-1 AP

Multiply this relation by D, we get

  
D 2  P 1 AP P 1 AP  P 1 A ( PP 1 ) AP

 P 1 AI AP  P 1 A2 P . (6)

Similarly we get D 3  P 1 A3 P .

Proceeding this way we get D n  P 1 An P

or An  PD n P 1 . (7)

This is the formula for calculating the nth power of A.

EXAMPLE 1 Diagonalize the following matrix A

3 1 1 
A    1 5  1
1 1 3 

Solution Characteristic equation of A is

3 1 1
1 5 1 0
1 1 3
or (3   ){(5   )(3   )  1}  1(  3  1)  1(1  5   )  0

or (3   )(14  8  2 )    2  4    0

or 42  24  32  14  82  3  2  6  0

or 3  112  36  36  0    2,3,6 .

Therefore, the eigen values of A are 2, 3, 6 which are all distinct. Corresponding to eigen
value 2 , we have [ A -2 I ] X = 0 , i.e.

1 1 1   x1  0 
 1
 3 1  x2   0  , Apply R21(1) and R31(-1)
1 1 1   x3  0 

1 1 1   x1  0
0
 2 0  x2   0
0 0 0  x3  0

This gives x1  x2  x3  0 and x2  0 . Let x1  k . Then x3  k .

Therefore X 1  k [1 0  1]' is an eigen vector of A corresponding to eigen value 2.

For eigen value 3, [ A -3 I ] X = 0 gives

0 1 1   x1  0 
 1
 2 1  x2   0  , Apply R13
1 1 0   x3  0 

1 1 0   x1  0 
 1
 2 1  x2   0  , Apply R21(1) and then R32(1)
0 1 1   x3  0 

1 1 0   x1  0 
0
 1 1  x2   0 
0 0 0   x3  0 

This gives X 2  k [1 1 1]' .

For eigen value 6, [ A -6 I ] X = 0 gives

 3 1 1   x1  0 
 1
 1 1  x2   0 
1 1 3  x3  0 

Apply R13 and then R21(1) and R31(3)

1 1 3  x1  0
0
 2 4  x2   0
0 4 8  x3  0

Apply R32(-2)

1 1 3  x1  0
0
 2 4  x2   0
0 0 0   x3  0

This gives x1 - x2 -3 x3 = 0 , x2 +2 x3 = 0. Let x3 = k. Then x2 = -2k and x1 = k. Therefore an

eigen vector is given by X 3  k [1 2 1]' . So the required matrix P is

1 1
2 0 
1 1 1  2
 
1 1 1 
P   0 1  2  , P 1  
3 3 3 
  1 1 1   
1 
1 1 
 6 3 6 

2 0 0
Thus P AP  0
1
3 0
0 0 6

The inverse of the matrix P may be calculated by the Gauss Jordan method as:

1 1 1  1 0 0
0 1 2  0
 1 0 A, Apply R13 (-1), we get

 1 1 1  0 0 1 

2 0 0  1 0 1
0 1 2  0
 1 0  A, Apply R32 (-1)

 1 1 1  0 0 1 

2 0 0  1 0 1
0 1
 1 2  0
 1 0  A, Apply R31  
2
 1 0 3  0 1 1 
  
2 0 0  1 0  1
  2
0 1 
2   0 1 0  A , Apply R23  3 
  
 0 0 3  1 1 
 1 
2 2 

 
1 0  1
2 0 0   1 1
0 1 1 1  A, Apply R1   and R3  
 1 0    2  3
3 3 3 
 0 0 3  1 1 
 1 
2 2 

1 1
2 0 
1 0 0 2
 
0 1 1 1  A.
 1 0   
3 3 3 
 0 0 1   
1 
1 1 
 6 3 6 