Eigen values & Eigen vectors

Definition: Let A =   aij n n be any real or complex square matrix.

Consider the matrix equation AX =X or (A −I X) =O . A value of  for

which AX =X has a solution X 0 is called an eigen value or characteristic

value of the matrix

A. The corresponding solution X  0 to the homogeneous system

(A −I X) =O of linear equations is called an eigen vector or characteristic

vector of the matrix A.

Remarks: Let A =   aij n n be any real or complex square matrix. If

det(A −I)  0, then the homogeneous system (A −I X) =O of linear

equations has a unique solution X O= . The homogeneous system (A −I

X) =O has infinitely many non-trivial solutions provided det(A −I) = 0

Definition: Let A =   aij n n be any real or complex square matrix.

Consider the system AX =X or (A −I X) =O . The equation det(A −I) =

0 is called a

characteristic equation of the matrix A.

The characteristic polynomial det(A −I) is an n th degree polynomial in 

and its roots are called eigen values of the matrix A. The set of all eigen
values of the matrix A is the spectrum of A. The largest of the absolute
values of the eigen values of A is called the spectral radius of A.
Definition: Two square matrices A and B are said to be similar matrices if
there exists a non-singular matrix P such that B P AP= −1
.
Problem: Find the eigen values and corresponding eigen vectors of

−5 2
A= 

1
  2 −2

−5 2 − −5  2 
= =
Solution: Given A  . Then A −I  

 2 −2 2 − −2 
The characteristic polynomial of A is det(A −I) = 0
 − −( 5 )(− − − =2 )4 0

 (5+ )(2+ + =) 4 0

2 +7+ =6 0

 +( 1)(+ =6) 0

 =− 1 or −6. Put 1 =−1 and 2 =−6

Then 1 =−1 and 2 =−6 are eigen values of the given matrix A.

First we find eigen vectors corresponding to 1 =−1. For this, consider

(A −I X) =O . Put  = 1 = −1

 +(A I X) =O

−4 2 x1  0  x1


  2 1−   x2 =  0 , where x =   x2

−4 2
Now (A + I ) = 
 2 −1

Apply R2 → 2R2 +R1

−4 2
 0
(A + I  )
0

Take x2 =k, where k is a constant

−4 2
 0
From the first row of (A +  I ), we have −4x1 + 2x2 =
0 0

2
 k
−2x1 + x2 = 0  x1 =
2

 k
 x1  2 is a general solution of (A + I X) =O
X =   x2 =   k

 1 −1  2
 X1 =   , X2 =   , X3 =   ,... are eigen vectors corresponding to 1 =−1
 2 −2  4

Now we find eigen vectors corresponding to 2 =−6

Consider (A −I X) =O and put  = =−2 6

 (A + 6I X) =O

1 2 x1  0  x1

  2 4  x2 =   0 , where x =   x2

 1 2
Take (A + I ) = 
2 4

Apply R2 → R2 −2R1

1 2
 0
(A +6 I  )
0

Choose x2 =k, where k is a constant

1 2
 0
From the first row of (A +6  I ), we have x 1 + 2x2 = 0
0

 x1 = −2k

 x1 −2k
 X =  x2 =  k  is a general solution of (A + 6I X) =O


3
 −2  2 −4
 X1 =   , X2 =   , X3 =   ,... are eigen vectors corresponding to
 1 −1  2
2 =−6.

Problem: Find the eigen values and corresponding eigen vectors of

 2 2 1
 
A= 1 3 1 .

 1 2 2

2 2 1  2− 2 1
 
Solution: Given A = 1 3 1 . Then A −I =  1 3− 1

   

 1 2 2  1 2 2−

The characteristic equation of A is det(A −I) = 0.

 
 2− 2 1   det 13− 1 =0
 
 1 2 2−

 (2−)(3− )(2− )−2 −2(2−) −1 + 2 − (3−) = 0

   3 −7 2 +11 − =50
 ( −5) ( −1)2 = 0

 = 1,1 or 5

 = 1 or 5 are eigen values of A. Put 1 =1 and 2 = 5

First we find eigen vectors corresponding to 1 =1. For this, consider

(A −I X) =O . Put  = =1 1

 −(A I X) =O

1 2 1x1  0 x1
  1 21x2 =   0 , where x = x2

4
  
 1 2 1x3  0  x3 

1 2 1
 
Consider A − =I 1 2 1
 
 1 2 1

Apply R2 →R R and R2 − 1 3 →R R3 − 1

1 2 1

A− I00
0
 

0 0 0

Choose x3 =k1 and x2 =k2 , where k1 and k2 are constants

1 2 1
0 
From the first row of A  − I00 , we have x1 +2x2 + x3 = 0

0 0 0
 x1 =−(k1 + 2k2)

−(k +2k )

 x1  1 2 

 k 
X =   x2 =  k12  is a general
solution of (A − I X) =O

  x3
 
−3  3  1
  
 X1 =  1 , X2 = − 1  , X3 = −  1 ,... are eigen vectors
corresponding to
 1 −1   1
1 =1.
Now we find eigen vectors corresponding to 2 = 5

Consider (A −I X) =O and put  = 2 = 5

5
 (A −5I X) =O

−3 2 1x1  0 x1

   
  1 −21  x2  =  0 , where x = x2
 1 2 −3x3  0 x3

−3 2 1
 
Consider A −5I = 1 −2 1
 
 1 2 −3

Apply R2 →3R2 +R1 and R3 →3R3 +R1

−3 2 1

A −5I−4 4
0
 

0 8 −8
Apply R3 →R3 +2R2

−3 2 1

A −5I−4 4
0
 

0 0 0

R2
Apply R2 →
4

−3 2 1

A −5I−1 1
0
 

0 0 0

From the second row, we have −x2 + x3 = 0  x3 = x2

Choose x3 =k , where k is a constant

 x2 =k and from the first row we have

−3x1 +2x2 + x3 = 0 −3x1 + 3k = 0  x1 =k

x1  k1
    is a general solution of (A − 5I X) =O

6
 X = x2 =  k1
x3  k1
 1  2
   
 X1 = 1 , X2 = 2 ,... are eigen vectors corresponding to 2 = 5.
   
 1  2

Exercise
1. Find eigen values and corresponding eigen vectors of the following real
matrices
 3 5 4−−   5 5 −4 0 1 0 0 1 0
       
a. −5 6 5−− b. 45 −4 c. 1 0 1 d. 0 0 1
       
−4 5 3− −1 1 2−  1 1 2 0 0 0
0 1 0 0 8 3− −6 −6 2 2
   −4 2 2
0010 4 1−
1 −4 g. 0 3 172 3 11 h. −1 
e.   f.   13 0
0 0 0 1 3 5− −4
 
   1   
0 1 2 0 0 2
−4 0 0 −2 −1
2 2 0
 
i. 2 5 0
 
 0 0 3
2. Find eigen values and corresponding eigen vectors of the following
complex matrices

       
a.  4 1 3− i  b.  3i 2+i  c.  2 3+i  d.  3i 2+i 
1 3+ i 7  − +2 i −i  3−4i 2 − +2 i −i 

 2i 3i
d.    3i −i


7
Properties of Eigen Values
Theorem-1: If  is an eigen value of a square matrix A then +k is an eigen
value of A kI+ .

Proof: Let  be an eigen value of a square matrix A. Suppose that X is an

eigen vector corresponding to  of A and k is a constant.

Since X is an eigen vector corresponding to  of A, AX =X

Now (A +kI X) = AX +kIX = X +kX = +( k X)

 (A +kI X) = (+k X)

 +k is an eigen value of A kI+ .

Theorem-2: If  is an eigen value of a square matrix A then k is an eigen

value of kA.

Proof: Let  be an eigen value of a square matrix A. Suppose that X is an

eigen vector corresponding to  of A and k is a constant.

Since X is an eigen vector corresponding to  of A, AX =X

Now (kA X) =k AX( ) =k(X) =(k) X (kA

X) = (k) X

 k is an eigen value ofkA.

Theorem-3: A and AT have the same eigen values.

Proof: Let A be any square matrix and AT denote its transpose. The eigen
values of A are the roots of its characteristic polynomial det(A −I) = 0.

T T

We have (A −I ) = A −I

8
 ( )
Since det(A−I) =det(A−I)T =det AT −I , the characteristic

polynomials of A and AT are the same. Hence the eigen values of A and AT
are same.

Theorem-4: If  is an eigen value of a square matrix A, then n is an eigen

value of An .

Proof: Let  be an eigen value of a square matrix A. Suppose that X is an

eigen vector corresponding to  of A. Then AX =X

 AAX =A X
 A X2 =AX = ( X) = 2X

 A X2 = 2X
2 is an eigen value of A2

Suppose that n−1 is an eigen value An−1

 An−1X = n−1 X

 A X A An( n−1 ) = A n−1 X

 A Xn= A A( n−1
X ) = A (n−1 X ) =n−1(AX ) =  n−1( X ) = nX

 A Xn =n X

 n is an eigen value of An .

Theorem-5: The product of eigen values of square matrix A is the

determinant of A and the sum of eigen values is the trace of A.

Proof: Let Abe any square matrix of order n n .

Case 1:

a a1112 
Suppose that n =2 and A =  
a a21222 2

9
 a11 − a12 
=
 A −I  
a21 a22 −

 det(A −I) =(a11−)(a22 − −) a12 21a

=a11 22a −(a11 +a22) + −2 a12 21a

= −2 (a11 +a22)+(a11 22a −a12 21a )

=2 −trace A+ det A

The characteristic equation of A is det(A −I) = 0.

 2 −trace A+ det A = 0 → (1)

Comparing this equation with    2 −( 1+ 2 ) + 12 = 0, where  1 and 2 are

roots of the characteristic equation of A we get  1 + 2 =trace A &1 2 = det

 The sum of eigen values of A is the trace of A and the product of eigen

values of A is the determinant of A.

Case 2:

Let A be an n n square matrix. Suppose that  1 2 3 , , ,...,n are eigen

values of A. The characteristic equation of A is det(A −I) = 0

a11 a12 a13... a1n 

 a a a ... a
Let A =  21 22 23 2n 

......................... 
 
an1 an2 an3...ann 

a11 − a12 a13... a1n 

 
A −I = a21 a22 − a23... a2n 

10
 ......................... 
 
an1 an2 an3... ann −

n n+1

det(A −I ) = −( 1) n + −( 1)trace A n−1 +...+ det A

Now det(A −I) = 0

 (−1)nn + −( 1)n+1trace A n−1 +...+ det A = 0 → (2)

Since  1 2 3, , ,...,n are eigen values of A, they are roots of the

characteristic equation det(A −I) = 0. So we can write

(−1) n n
+ −( 1)n+1(    1 + 2 + 3 +... )
n
n−1
+...+ ( 1 2 3... )
n =0
→ (3)
Comparing equations (1) & (2), we get

  1 + 2 + 3 +...n =trace A

 1 2 3... n = det A

 The sum of eigen values of A is the trace of A and the product of eigen

values of A is the determinant of A.

1
Theorem-6: If is an eigen value of a non-singular matrix A then is an


eigen value of A−1.

Proof: Suppose that A is a non-singular matrix

 det A  0 and A−1 exists

Since the product of eigen values of A is the determinant of A, the eigen

values of a non-singular matrix cannot be zero. Let  be an eigen value of A

Since A is non-singular,  0. Let X be an eigen vector of A corresponding

to .

11
 AX =X

 A−1 (AX) = A−1 (X)

 X =(A−1 X )
−1
1
A X= X


1 −1

 is an eigen value of A .

det A
Theorem-7: If is an eigen value of a non-singular matrix A, then

is an eigen value of adj A.

Proof: Suppose that A is a non-singular matrix

−1
1
 det A  0 and A= adj A exists det
A

 adj A =(det A A) −1
→ (1)

Since the product of eigen values of A is the determinant of A, the eigen

values of a non-singular matrix cannot be zero. Let  be an eigen value of A

Since A is non-singular,  0. Let X be an eigen vector of A corresponding

to .

 AX =X

 A−1 (AX) = A−1 (X)

 X =(A−1 X )
−1
1

12
A X= X → (2)


Now, (adj A) X = (det A) (A −1

X )
 
1
) X
= (det A  
 

det A 
= X


 (adj A) X = det A  X



det A
Hence is an eigen value of adj A.

Theorem-8: The eigen values of a diagonal matrix are its principal diagonal
entries.

Proof: Let A be a diagonal matrix of order n n s follows

a11 0 0... 0 
 
0 a 0... 0 
22

A=
.........................  0 0 0 ...ann

 
a11 − 0 0... 0 0 a22− 0... 0 
......................................   
 
 A −I = 

0 0 0 ... ann − 

 det (A −I) =(a11−)(a22 −)(a33 −)...(ann −)

13
The characteristic equation of A is det(A −I) = 0

 (a11 −)(a22 −)(a33 −)...(ann − =) 0

 the roots are a11, a22, a33,… and ann

 the eigen values of A are a11, a22, a33 … and ann.

Theore-9: The eigen values of a triangular matrix are its principal diagonal
entries.

Proof: Let A be a triangular matrix of order n n s follows

a11 a12 a13 ... a1n 
 
0 a22 a23 ... a2n 
A = 0 0 a33 ... a3n 
 
......................... 
 
0 0 0 ... ann 
a11 − a12 a13 ... a1n 
 
0 a22− a23 ... a2n
 A − I =  0 0 a33 − ... a3n 
 
...................................... 
 0 0 0 ... ann −

 det (A −I) =(a11−)(a22 −)(a33 −)...(ann −)

The characteristic equation of A is det(A −I) = 0

 (a11 −)(a22 −)(a33 −)...(ann − =) 0

 the roots are a11, a22, a33,… and ann

 the eigen values of A are a11, a22, a33 … and ann.

Theorem-10: Similar matrices have the same eigen values.

14
Proof: Suppose that A and B are any two similar matrices. Then there
exists a non-singular matrix P such that B P AP= −1
→ (1)

We have det(B −I) = det (P AP −1

−P IP−1 )

= det P−1(A −I P) 

= detP−1 det(A −I) detP

= detP P−1 det(A −I)

= detI det(A −I)

= det(A −I)

 det(B −I)= det(A −I)

Thus the characteristic equations B and A are the same.

 the eigen values of B and A are the same.

Theorem-11: If A and B are any two square matrices and if A is invertible

then A B−1 and BA−1 have the same eigen values.

Proof: Suppose that A and B are any two square matrices and A is

invertible. Since BA−1 and P−1(BA−1)P are similar matrices, they have
same eigen values. Here P is any non-singular matrix. If we take P A=
then

P−1(BA−1)P = A−1(BA−1)A = A−1 B

 A B−1 and BA−1 have the same eigen values.

Theorem-12: The eigen values of a Hermitian matrix are real numbers.

Proof: Let A be a Hermitian matrix. That is, AT = A

Suppose that X is an eigen vector of A corresponding to an eigen value .

15
 AX =X

 (AX)T = (X)T

 X AT T =XT

 X AT T =XT

 X AT =XT

 A XT =X XT

 XT  X = X XT

 X XT =X XT

 X XT −X XT =0

 ( − ) X XT = 0

Since XT X  0, we have  − = 0

 =
 is a real number.

Hence The eigen values of a Hermitian matrix are real numbers.

Theorem-13: The eigen values of a skew-Hermitian matrix are purely

imaginary or zero.

Proof: Let A be a skew- Hermitian matrix. That is, AT = − A

Suppose that X is an eigen vector of A corresponding to an eigen value .

 AX =X
 (AX)T = (X)T
 X AT T =XT

16
 X AT T =XT

 −X AT =XT

 − X A XT =X XT

 − XT  X = X XT

 −X XT =X XT

 X XT +X XT =0

 ( + ) X XT = 0

Since XT X  0, we have  + =0

Re=0 where Re denotes the real part of  

= 0 or  is purely imaginary.

Hence The eigen values of a skew-Hermitian matrix are purely imaginary or

zero.

Theorem-14: The eigen values of a Unitary matrix have absolute value one.
Proof: Let A be a Unitary matrix. That is, A AT = A AT = I

Suppose that X is an eigen vector of A corresponding to an eigen value .

 AX =X

 (AX)T = (X)T

 X AT T
=XT

 X AT T
=XT

17
 X AT T
A =X AT

 X IT =X AT

 XT =X AT

 X XT =X AXT

 X XT = XT X

 X XT =X XT

 X XT − X XT =0

 (−1) X XT = 0

Since XT X  0, we have −1 = 0

=1
2

  =1

 =1

Hence The eigen values of a Unitary matrix have absolute value one.

Properties of Eigen Vectors

Theorem-1: If X1 and X2 are any two eigen vectors of a square matrix A

corresponding to an eigen value  then X1 + X2 is also an eigen vector of

A corresponding to .

Proof: Suppose that X1 and X2 are any two eigen vectors of a square matrix

A corresponding to an eigen value . Then AX1 =X1 → (1)

18
 AX2 =X2 → (2)

 A X( 1 + X2) = AX1 + AX2

= X1 + X2

=(X1 + X2)

 A X( 1 +X2) =(X1 +X2)

 X1 + X2 is an eigen vector of A corresponding to .

Theorem-2: If k is a constant and X is an eigen vector of a square matrix

A corresponding to eigen value  then kX is also eigen vector of A
corresponding to .

Proof: Suppose that k is a constant and X is an eigen vector of a square

matrix A corresponding to eigen value . Then AX =X  kAX =k X

 A kX( ) =(kX)
 kX is eigen vector of A corresponding to .

Theorem-3: Two eigen vectors of a square matrix A corresponding to two

different eigen values are linearly independent.

Proof: Suppose that X1 and X2 are any two eigen vectors of a square matrix
A corresponding to two different eigen values 1 and 2 respectively.

 
 AX1 = 11 X and AX2 = X
22

Suppose that c X1 1 +c X2 2 =O → (1) where c1 and c2 are constants

 A c X( 11 +c X2 2) =O
c1 ( AX1) + c2 ( AX2)=O

c1 (1 1X ) + c2 (2 2X )=O → (2)

Solving (1) and (2)

c1 1( − )
2 X1 =O

19
 c 1= 0

Similarly c2 ( 2 − )
1 X2 =O

 c 2= 0

Hence X1 and X2 are linearly independent.

Definition: Two eigen vectors X1 and X2 are said to be orthogonal if

X X1T 2 =0 or X X2T 1 =0.

Theorem-4: Two eigen vectors of a symmetric matrix A corresponding to

two different eigen values are orthogonal.

Proof: Suppose that X1 and X2 are any two eigen vectors of a symmetric
matrix A corresponding to two different eigen values 1 and 2
respectively.

 
 AX1 = 11 X and AX2 = 22 XX


AX2T 1 = X2 1 1TX and AX2 = X
22

T T T
(
X AX2T1) (X )
= X
 T 2 1 1T  and X AX12 = X1 2 2X

 X A X1T T 2 =1 X X1T 2 and X AX1T 2 =2 1 2 X XT

Since A is symmetric, AT =A.

Consider X A X1T T 2 =1 X X1T 2 and X AX1T2 =2 1 2X XT

 X AX1T 2 =1 X X1T 2 and X AX1T2 =2 1 2X XT

 1 X X1T 2 =2 1 2X XT

 1 X X1T 2 −2 1 2X XT= 0

 ( 1 − )
2 X X1T 2 =0

20
  1 − 2 = 0 or X X1T 2 =0

Since  1  2 ,  1 − 2  0

 X X1T 2 =0

 X1 and X2 are orthogonal.

*****

Cayley – Hamilton theorem

Statement: Every square matrix satisfies its own characteristic equation.

212
 
Problem: Verify Cayley- Hamilton theorem for the matrixA = 533 .
 
−1 0 −2

21
 212
 
Solution: Given A = 5 3 3
 
−1 0 −2

 2− 1 2 
 
 A −I = 5 3− 3
 
−1 0 − −2 

 det(A −I) = (2−) (3−)(− −2 ) − 5(− −2 ) +

3 + 2 3 −
   3 −3 2 −7 −1

Then the characteristic equation of A is det(A−I)

=0  3 −32 −7− =1 0 → (1)

212212 7 5 3
    
Now A2 = 533 533 = 22 14 13
    
−1 0 −2  −1 0 −2  0 −1 2 

7 5 3212  36 22 23
    
 A3 = 22 14 13 5 3 3 = 101 64 60
    
 0 −1 2   −1 0 −2  −7 −3 −7

Then
 36 22 23 21 15 9  14 7 14 1 0 0
      1 0
LHS = 101 64 60 − 66 42 39 − 35 21 21 − 
0
       1
 −7 −3 −7 0 −3 6   −7 0 −14 
0

 15 7 14  14 7 14  1 0 0

22
      
= 35 22 21 − 35 21 21 − 0 1 0
     
−7 0 −13  −7 0 −14  0 0 1
1 0 0  1 0 0
   
= 0 1 0 − 0 1 0
   
 0 0 1  0 0 1

0 0 0
 
= 0 0 0 =O
 
 0 0 0

= RHS

   3 −3 2 −7 − =1 O .

 A satisfies its characteristic equation.

1 3
2 
4 by using Cayley – Hamilton

Problem: Find A−1 of a matrix A = 2 
−1 −1

 3
1
theorem.

1 2 3
 
Solution: Given A = 2 −1 4
 
 3 1 −1

 1− 2 3
 
 A −I = 2 − −1  4
 
 3 1 − −1 

 det(A −I) = (1−)(− −1 )2 − 4 −2 2 1 (− −)−12 −3 2 3

23
1 − (− −)
= 3 + 2 −18− 40

The characteristic equation of A is det(A −I) =0

  3 + 2 −18− 40 = 0 → (1)

By the Cayley – Hamilton theorem, A satisfies its characteristic equation

 A3 + A2 −18A − 40 I =O → (2)

Multiplying with A−1, we get A2 + A −18I − 40 A−1 =O

 40 A−1 = A2 + A −18I
1 2 31 23   14 3 8
    
Now A2 = 2 −1 2 −1 4 = 12
4 9 −2
    
−1  2 4 14
 3 1 −1   3 1

From equation (2), we have

 14 3 8   1 1 0 23  0
     1 0
40 A−1 = 12 9 −2 + 2 −1 4 −18 0
0 
     1
−1  0
 2 4 14  3 1

 −3 5 11
 
= 14 −10 2
 
 5 5 −5

 −3 5 11
1
 
 A−1 = 14 −10 2
40    55
−5
7 2−2
 
Problem: Find A4 of a matrix A = − 6 −1 2 by using Cayley – Hamilton
 
−1

24
  6 2
theorem.

7 2 −2
 
Solution: Given A = − 6 −1 2
 
 6 2 −1

 7− 2 −2 
 
 A −I = − 6 − −1  2
 
 6 2 − −1 

 det(A −I) = (7−)(− −1 ) − 4 −2 6 1− − −( )−12 −2 12 6

1− − (− −)

=  3 −5 2 + 7 − 3

The characteristic equation of A is det(A−I) =0

 3 −52 + 7− =3 0

By the Cayley – Hamilton theorem, A satisfies its characteristic equation

 A3 −5A2 + 7A − 3 I =O
Multiplying withA, we get A4 −5 A3 + 7A2 − 3A =O

 A4 = 5 A3 − 7A2 + 3A

7 2 −2  7 2 −2  25 8 −8 
    
Now A2 = − 6 −1 2 −6 −1 2 = − 24 −7 8
    
 6 2 −1   6 2 −1  24 8 −7

 25 8 −8   7 2 −2  79 26 −26
    
A3 = − 24 −7 8 −6 −1 2 = − 78 −25 26
    

25
  24 8 −7   6 2 −1  78 26 −25

Then

 79 26 −26  25 8 −8 7 2 −2

     
LHS = 5 −78 −25 26 − 7 − 24 −7 8 +3 −6 −1 2
     
 78 26 −25  24 8 −7  6 2 −1
 395 130 −130  175 56−56   21 6 −6
    
= − 240 −125 130 − − 168 −49 56 + −18 −3 6
    
 390 130 −125  168 56 −49   18 6 −3

 241 80 −80 
 
= − 240 −79 80
 
 240 80 −79

Exercise
1. Verify the Cayley-Hamilton theorem for the following matrices.

3 1 4  1 0 a. 0 2 −1 1 −2 2 1 2 3 
      4 5
6 b. 1 2 1 c. 1 − 2 3 d. 2
   5 
    6
0 0 5 2 23 2 3
 0 −1
2. Using Cayley – Hamilton theorem, find the inverse of the following
matrices.

7 −1 3 4 6 6 2 −1 1
     
a. 6 1 4 b. 1 3 2 c. −1 2 −1 d.
     
2 4 8  −1 −4 −3  1 −1 2
1 1 2

26
  3 1
1
 
 2 3 1

*****

Diagonalization
Definition: A square matrix A is said to be diagonalizable, if there exists a
non – singular matrix P such that P AP−1 is a diagonal matrix. That is, A is
said to be diagonalizable if it is singular to a diagonal matrix. The non –
singular matrix P which diagonalizes A is called a modal matrix and the
diagonal matrix D = P A P−1 is called a spectral matrix.

Remarks: Let A be a square matrix of order n.

1. If A has n distinct eigen values then A is diagonalizable.

2. If A has exactly n linearly independent eigen vectors then A is
diagonalizable.
3. Every symmetric matrix is diagonalizable.
27
 −5 2
Problem: Diagonalize the matrix A =  
 2 −2

−5 2 − −5  2 
= =
Solution: Given A  . Then A −I  

 2 −2 2 − −2 
The characteristic polynomial of A is det(A −I) = 0

 − −( 5 )(− − − =2 )4 0

 (5+ )(2+ + =) 4 0

2 +7+ =6 0

 +( 1)(+ =6) 0

 =− 1 or −6. Put 1 =−1 and 2 =−6

Then 1 =−1 and 2 =−6 are eigen values of the given matrix A.

First we find eigen vectors corresponding to 1 =−1. For this, consider

(A −I X) =O . Put  = 1 = −1

 +(A I X) =O

−4 2 x1  0  x1

  2 1−   x2 =   0 , where x =   x2

−4 2
Now (A + I ) = 
 2 −1

Apply R2 → 2R2 +R1

2
−4
(A + I  0

)
0

Take x2 =k, where k is a constant

−4 2
 0

28
From the first row of (A + I ), we have −4x1 + 2x2 = 0 0

k
−2x1 + x2 = 0  x1
=
 k 2
 x  2
1

X =   x2 =  is a general solution of (A + I X) =O

 k
 X1 =   , X2 =   , X3 =   ,... are eigen vectors corresponding to 1 =−1
 2 −2  4

Now we find eigen vectors corresponding to 2 =−6

Consider (A −I X) =O and put  = =−2 6

 (A + 6I X) =O

1 2 x1  0  x1

=
  2 4  x2 =   0 , where x  x2

 1 2
Take (A + I ) = 
2 4

Apply R2 → R2 −2R1

1 2
 0
(A +6 I  )
0

Choose x2 =k, where k is a constant

1 2
 0
From the first row of (A +6  I ), we have x 1 + 2x2 = 0
0

 x1 = −2k

 x1 −2k
Y==

29
 x2  k  is a general solution of (A + 6I X) =O
 1 −1  2

30
 −2  2 −4
 Y1 =   , Y2 =   , Y3 =   ,... are eigen vectors corresponding to
 1 −1  2

2 =−6.

 1 −2
= =
Consider X1   and Y1  . These two vectors are linearly independent.
 2  1

1 −2  −1 1  1 2
=
Let P 2 1 . This matrix P is non – singular and P= 5 −2 1 .

−1 0 −1

=
Then D = P A P  .
 0 −6

This diagonal matrix D is similar to A and A diagonalizable.

−9 4 4
 
Problem: Diagonalize A = − 83 4 .
 
−16 8 7

−9 4 4
 
Solution: Given matrix is A = − 83 4
 
−16 8 7

 − −9  4 4
 
Then A −I = −8 3− 1
 
 −16 8 7−

The characteristic equation of A is det(A −I) = 0.

  3 − 2 −5− =3 0

31
 (+1)2 (−3) = 0

 =− − 1, 1 or 3

1 = −1 and 2 = 3 are eigen values of A.

To find eigen vectors corresponding to1 =−1, consider(A −1I X)=O .

−8 4 4
 4
Now A −1I = A + =IA = −  8 
4 8
−16 8
Apply R2 →R2 −R1 and R3 →R3 −2R1

−8 4 4

A+ I00
 0
 
0
 0 0

R
Apply R1 →
−2 1 1

A+ I00
 0
 
0
 0 0

Take x2 =k1 and x3 =k2 where k1 and k2 are constants

From the first row, we have −2x1 + x2 + x3 = 0

k k1 + 2

 x1 =
2
Then

k1 +k2
 2 
 
 k1  is a general solution of (A − I X) =O
1

 k2 

32
 

 1  0
   
Choose the eigen vectors X1 =  1  and X2 = −  1  corresponding to1
=−1.
 1  1

Now we find eigen vectors corresponding to 2 = 3

Consider (A−2I X) =O

−12 4 4
 
Then A −2I = A −3I = A = − 80 4 
−16 8 4
R R R
Apply R1 → , R2 → 1
and R3 →
2 3

4 4 4
−3 1 1
 
A −3I −2 0 1

−42 1
Apply R3 →R3 −2R2

−3 1 1
 
A −3I −2 0 1
 
0 2 −1

Apply R2 → 3R2 −2R1

−3 1 1
 0 −2
 
A −3I1


0 2 −1

Apply R3 → R3 + R2 
A −3I1
−3 1 1 
 0 −2

33

 0 0 0

Take x3 = k where k is a constant

Then −2x2 + x3 = 0

k
 x2 = 2

k
−3x x x1 + + =  =2 3 0 x1

Then
 1
 k 
  Choose the eigen vector
 2   
X3 =  1 
 k 
 corresponding to2 = 3.

 (A−2I X) =O is a general solution of
 
2
 
k
 
 2

Suppose that c X1 1 +c X2 2 +c X3 3 =O

 1  0  1  0
 c1  1 +c2 −1+c3  1 =  0
 1  1  2  0

 c1 +c3 = 0, c1 −c2 +c3 = 0 and c1 +c2 + 2c3 = 0

 c1 = 0, c2 = 0,c3 = 0

 1  0  1
     
1 , −1 and 1 are linearly independent vectors
     
 1  1  2
0 1 1
 −1 1
The modal matrix is given by P = 1 
1


34
  1 2
3 −1 −1

 P−1 = 1 −10
 
−2 1 1
−1 0 0
 
 P AP−1 = 0−1 0
 
 0 0 3
 A is diagonalizable and D P AP= −1
is a diagonal matrix similar to A

0 −6 −4
 
Example: The matrix A = 5 −11 −6  is not diagonalizable.

−6 9 4 

Exercise
1. Diagonalize the following matrices.

2 1 −1 0 1 0 
   
a. 1 4 3 b. 100
   
−1 3 4 0 0 0

2 0 0 0 0 1
   
c. 0 1 1 d. 001
   
0 1 1 1 1 1

*****

35
Quadratic Forms

Definition: A quadratic form in n variables (x x1 2,,...,xn ) is defined by

x1 
 
x
2



X AXT where A =  aijn n and X = .  .
.
 
xn

The matrix A =   aij n n is called a matrix of the quadratic form

Problem: Write a quadratic form in two variables x1 and x2 .

a a
=
Solution: Let A  11 12  and X =   x1
a21 a22  x2

36
 a a
Then X AXT = x1 x2     x1
11 12

a21 a22  x2

=a x11 12 +a12 +a21x x1 2 +a x22 22

 X AXT= a x11 12 +a12 +a21 x x1 2 +a x22 22

If we take a12 =a21, then X AX a xT= 11 1

2
+ 2a xx a x12 1 2 + 22 2
2

a a12
In this case a12 =a21, A =  11  and then A is symmetric.
a12 a22
Problem: Write a quadratic form in three variables x1 , x2 and x3 .

a11 a12 a13  x1

 
= a22 a23 and X = x2
Solution: Let A a21
a32 a33 x3
a31
a11 a12 a13   x1

x3 a21 a22a23  

x2
a31 a32a33  
Then X AX T
= x 1 x2x3

=a x11 12 +a12 +a21 x x1 2 +a x22 22 +a23 +a32x x2 3 +a13 +a31x x1 3 +a x33

32

 X AXT= a x11 12 +a x22 22 +a x33 32 +a12 +a21x x1 2 +a23 +a32x x2 3 +a13

+a31x x1 3

If we take a12 =a21, a13 =a31 and a23 =a32 then

X AXT= a x11 12 +a x22 22 +a x33 32 +2a x x12 1 2 +2a x x23 2 3 +2a x x13 1 3
a11 a12 a13 
 a22 
In this case, A = a12 a23 a23 and then A is symmetric.
a13 a33

37
Problem: Write a symmetric matrix corresponding to the quadratic form

2x12 − x22 + 6xx1 2

Solution: Given quadratic form is 2x12 − x22 + 6x x1 2

Comparing with X AXT = a x11 12 + 2a x x12 1 2 +a x22 22, we get

a11 = 2,a22 = −1 and a12 =a21 = 3

2 3
Then the symmetric matrix of the quadratic form is A =   . 3
−1
2 1 2
Problem: Write a quadratic form corresponding to the matrix A = 3 1 .
 2
1 
 2
1

2 2 1 x1
   
Solution: Given A = 1 3 1 . Let X = x2 
 
1 2 2 x3
 X AXT = a x11 12 +a x22 22 +a x33 32 +a12 +a21x x1 2 +a23 +a32x x2 3 +a13

+a31x x1 3  X AXT = 2x12 +3x22 +2x32 +3x x1 2 +3x x2 3 +2x x1 3 .

Problem: Write a symmetric matrix corresponding to the quadratic form

2x12 +4x22 +4x32 +2x x1 2 +6x x2 3 −2x x1 3

Solution: Given quadratic form is 2x12 +4x22 +4x32 +2x x1 2 +6x x2 3 −2x x1 3

Comparing with a x11 12 +a x22 22 +a x33 32 + 2a x x12 1 2 + 2a x x23 2 3 + 2a x x13 1 3 ,

we get

a11 = 2,a22 = 4, a33 = 4, a12 =a21 =1, a23 =a23 = 3 and a13 =a31 = −1

Then the symmetric matrix corresponding to the given quadratic form is

38
 2 1 −1
14 3
 
−1 3 4 

Definition: The normal form or canonical form of a quadratic form X AXT

   
in n variables is defined by x 2+
11 x 2+
22 x 2 +...+
33 x 2.
nn

Definition: Let X AXT be a quadratic form in n variables and let A be the

symmetric matrix corresponding to X AXT . Then

1. The rank of the quadratic form X AXT is defined to be the rank of

A. It is denoted by (A) =r .
2. The number of positive terms in the normal form of X AXT is called

the index of the quadratic form. It is denoted by s .

3. The signature of X AXT is defined by 2s r− .

Definition: A quadratic form X AXT in n variables is said to be

1. Positive definite if r = =s n or if all the eigen values of A are

positive
2. Positive semidefinite if r s n=  or if the eigen values of A are zero or
positive
3. Negative definite if r n= and s =0 or if all the eigen values of A are
negative

4. Negative semidefinite if r n and s =0 or if the eigen values of A are

zero or negative.
Problem: Reduce the quadratic form

8x12 + 7x22 + 3x32 −12x x1 2 − 8x x2 3 + 4x x1 3 into normal form and hence

find its rank, signature and index.

Solution: Given quadratic form is

X AXT =8x12 +7x22 +3x32 −12xx1 2 −8x x2 3 +4xx1 3

39
 Comparing with

X AXT = a x11 12 +a x22 22 +a x33 32 +2a x x12 1 2 +2a x x23 2 3 +2a x x13 1 3 ,

we get a11 = 8, a22 = 7, a33 = 3, a12 =a21 = −6, a13 =a31 = 2 and a23 =a32 =

−4

Now the symmetric matrix corresponding to the given quadratic form is

8 −6 2
 
A=− 6 7 −4
 
 2 −4 3

 8− 2
−4 

 A −I = −6 −6 
 7− 3−
 2 −4 
The characteristic equation of A is   3 −18 2 +45 =0

 ( −3)( −15) =0

Hence 1 = 0,2 = 3 and 3 =15 are the eigen values of A.

To find the eigen vectors corresponding to 1 =0 , consider (A−1I X) =O

 AX =O

8 2
 
Consider A = − 6 −6 −4
 7 
 2 −4 3
Apply R2 → 4R2 +3R1 and R3 → 4R3 −R1
 8 −6 2

A10 −10
 0
 
0
 −10 10
Apply R3 →R3 +R2

8 −6 2

A10 −10

40
  0
 
0
 0 0 

Take x3 = k, where k is a constant

Now 10x2 −10x3 = 0  x2 =k

k
Also 8x1 −6x2 +2x3 =0  x1 = 2

 k
 2

Therefore  k is a general solution of (A−1I X) =O
 k



 1

Choose the eigen vector X1 =  2 corresponding to 1 =0.
  2

To find eigen vectors corresponding to 2 = 3, consider (A−2I X)=O

 5 −6 2
 4 −4
We have A −2I = A
= −6 −4 
−3I
 0
 2
Apply R2 →5R2 +6R1 and R3 → 5R3 −2R1

 5 −6 2
A −3I  0 −16 −8 

 
 0 −8 −4

R2 R3
Apply R2 → and R3 →
−8 −4
5 −6 2

41
 
A −3I21
 0
 
0
 2 1
Apply R3 →R3 −R2

5 −6 2

A −3I21
 0
 
0
 0 0

Take x3 = k, where k is a constant

k
Now 2x x2 + =3 0 x2 =−
2

Also 5x1 −6x2 +2x3 = 0  x1 = −k

−k 
 k
 −  is a general solution of (A− I X) =O
2

 2
 k 

−2
 
Choose the eigen vector X2 = −  1  corresponding to 2 = 3.
 2

To find eigen vectors corresponding to 3 =15, consider (A−3I X)=O then

−7 −6 2
 
A −15I = − 6 −8 −4
 
 2 −4 −12

R2 R3
Apply R2 → and R3 →
2 2

42
 −7 −6 2
A  −15I−3 −4 −2 


 1 −2 −6

Apply R2 → 7R2 −3R1 and R3 → 7R3 +R1

−7 −6 2
 
A  −15I 0 −10 −20
 
 0 −20 −40

R R
Apply R2 → 2
and R3 →
10
−7 −6 2

A −15I−1 −2
 0
 
0
 −1 −2
Apply R3 →R3 −R2

−7 −6 2

A −15I−1 −2
 0
 
0
 0 0

Take x3 = k, where k is a constant

Now −x2 −2x3 = 0  x2 = −2k

Also −7x1 −6x2 +2x3 = 0  x1 = 2k

 2k 

 − 2k  is a general solution of (A− I X)=O .
3

 k 

 2
 
Choose the eigen vector X3 = −  2  corresponding to 3 =15.

43
  1

 1 −2  2
    
Since the eigen vectors X1 =  2 , X2 = −  1  and X3 = −  2  are

  2  2  1
corresponding to distinct eigen values of a symmetric matrix A, they are
linearly independent and orthogonal. The modal matrix is
1 −2 2
 
P = 2 −1 −2
 
2 2 1
 P is an orthogonal matrix
0 0 0
 0 3
 D = P AP−1 = 0
 0
 15
 0
The canonical form or normal form of the given quadratic form is

  1 1x 2 + 22 x 2+ 33 x 2 , where 1 = 0,2 = 3 and 3 =15

The rank of A is r =2. The index of the given quadratic form is s =2 and
the signature is 2s r− =2. The given quadratic form is positive semidefinite.

Problem: Find the nature of the quadratic form

x12 + 4x22 + x32 − 4x x1 2 +2x x1 3 −4x x2 3 .

Solution: Given quadratic form is

X AXT= x12 + 4x22 + x32 − 4x x1 2 +2x x1 3 −4x x2 3

Comparing with

X AXT = a x11 12 +a x22 22 +a x33 32 + 2a x x12 1 2 + 2a x x23 2 3 + 2a x x13 1 3 ,

we get a11 =1, a22 = 4, a33 =1, a12 =a21 = −2, a13 =a31 =1 and a23 =a32 = −2

Now the symmetric matrix corresponding to the given quadratic form is

1 −2 1
−2

44
  
A=− 2 4
1

 1 −2

 1− 1
−2

 A −I = −2 −2 
 4− 1−
 1 −2 
The characteristic equation of A is det(A −I) =0
 3 −62 = 0

Hence 1 = 0,2 = 0 and 3 = 6 are the eigen values of A.

The given quadratic form is positive semidefinite.

Exercise
1. Write a symmetric matrix corresponding to the following quadratic forms
a. ax2 +by2 +2hxy

b. x2 +2y2 +3z2 + 4xy +5yz +6xz

c. x12 +2x22 −7x32 −4xx1 2 +8xx1 3

d. 2xy +6xz −4yz

e. x12 +6x x1 2 +5x22

2. Find a quadratic form corresponding to the following matrices


a h g a.  h b f
 
 g f c 

0 5 −1

b. 5 1 6
 
 −1 6 2 

1 2 3

c. 2 0 3

45
  
 3 3 1

2 1 5

d. 1 3 −2
 
 5 −2 4

1 5 7

e. 5 4 6
 
 7 6 3 

3. Reduce the following quadratic forms to normal form

a. 6x12 +3x22 +3x32 −4xx1 2 −2x x2 3 +4xx1 3

b. 4x2 +3y2 +z2 −xy −6yz + 4xz

c. 3x2 +2y2 +3z2 −2xy −2yz

d. 3x2 +5y2 +3z2 −2xy −2yz +2xz

e. x12 +3x22 +3x32 −2x x2 3

4. Find rank, signature and index of the following quadratic forms

a. 3x2 +2y2 +3z2 −2xy −2yz

b. 6x2 −4xy +2y2

c. 3x12 +5x22 +3x32 −2x x1 2 −2x x2 3 +2x x1 3

d. 2x2 +2y2 +2z2 −2xy −2yz −2xz

e. x2 +y2 + 4xy

5. Find the nature of the following quadratic forms

a. 2x2 +2y2 +2z2 +2yz

b. x2 +2y2 +3z2 + 4xy +5yz +6xz

c. x12 + 2x22 −7x32 −4x x1 2 +8x x1 3

d. x2 −y2 + 4z2 + 4xy +2yz +6xz

e. x12 +6x x1 2 +5x22

46
 *****

47