Dr.

Satish Shukla 1 of 39

Engg. Math. II (Unit-I)

Calculus of Matrices

Syllabus. Systems of linear equations and their solutions. Matrices, determi-

nants, rank and inverse. Linear transformations. Range space and rank, null
space and nullity. Eigenvalues and eigenvectors. Similarity transformations. Di-
agonalization of Hermitian matrices. Bilinear and quadratic forms.

Vectors and their linear combination and generated space. We consider the three
dimensional Euclidian space and denote it by R3 . Each vector of this space can be

represented by its position vector. For example, if O is the origin and OP is the vector

with tail O and head P , where coordinates of P are (x1 , x2 , x3 ), then OP = x1 i + x2 j +
x3 k is completely described by its coordinates (x1 , x2 , x3 ), i.e., all the information

about OP is contained in the coordinates (x1 , x2 , x3 ). Therefore, all the vectors of R3
are represented by their coordinates and we write R3 = {(x1 , x2 , x3 ) : x1 , x2 , x3 R}.
x1
In further discussion, we represent a vector by X = x2 , since it same as (x1 , x2 , x3 )
x3
in thesense that both the notations give same information about vector. The vector
x1
X = x2 is also called a column vector or column matrix. Similarly, we can define
x3
a row vector of row matrix
byX 0
= [ x1 , x2 , x3 ].
1 1
Suppose, X1 = 0 , X2 = 1 are two three dimensional vectors and a1 , a2 R.

0 0
Then, the expression

1 1 a1 + a2
X = a1 X 1 + a2 X 2 = a1 0 + a2 1 = a2
0 0 0
is called a linear combination of the vectors X1 , X2 . Obviously, by changing the values
of a1 and a2 we can find infinitely many linear combinations of X1 and X2 . The set of
all linear combinations of X1 and X2 is called the space generated by the vectors X1 and
X2 . These notions can be generalized for arbitrary number of vectors of n-dimensional
vectors.
Linear independence and dependence of vectors. Vectors X1 , X2 , . . . , Xn are called
linearly dependent if any one of them is a linear combination of other vectors, otherwise
vectors are called linearly independent.
Echelon form of a matrix. We can always reduce the given matrix into a matrix which
is in the following form:
(1) All nonzero rows are above any zero row;
(2) the pivot (first nonzero entry from the left) of any row is always strictly to the
right of the pivot of the row above it.
Such a form of matrix is called the echelon form.
a11 a12 a1n
a21 a22 a2n
Rank of a Matrix. Suppose A = .. = [X1 X2 . . . Xn ], where

.. ..
. . .
am1 am2 amn
Dr. Satish Shukla 2 of 39

a11 a21 a1n
a21 a22 a2n
X1 = , X2 = . . ., Xn = . Then the number of linearly

.. .. ..
. . .
am1 am2 amn
independent column vectors in X1 , X2 , . . . , Xn iscalledthe column rank of matrix A.
X10
X0
2
Similarly, the matrix A can be written as A = .. where X10 = [a11 a12 a1n ],
.
0
Xm
X10 = [a21 a22 a2n ] , . . . , Xm
0
= [am1 am2 amn ]. Then, the number of linearly in-
dependent row vectors in X1 , X20 , . . . , Xm
0 0
is called the row rank of matrix A. An
interesting property of matrices says that the row and column ranks of a matrix are
always equal and this common value is called the rank of matrix and denoted by (A).
How to find the rank of a matrix. We use the following two methods of finding
the rank of matrices:
(I). Method of determinants. The rank of a matrix can also be calculated using
determinants. The rank of a matrix is the order of the largest square submatrix of
the given matrix with nonzero determinant. To find this, we search a submatrix
with nonzero determinant, and we start with the largest possible submatrix of
the given matrix. If this largest submatrix has a nonzero determinant, then
the order of the submetrix is the rank of the given matrix. If the determinant
of largest is zero, then we go to the submatrix of order less than 1 from the
largest submetrix, and repeat this process till we get a submetrix with nonzero
determinant.
(II). Method of Echelon form. In this method, we first reduce the given metric
into echelon form by applying the elementary transformations, and then, the
number of nonzero rows in the reduced echelon form is the rank of the given
matrix.

Example 1. Find
one nonzero minor of highest order of the matrix A =
1 2 3
2 4 1 , hence find its rank.
1 2 7

Sol. We start with the highest ordr minor, i.e., the minor of order 3
 
 1 2 3 
 
|A| =  2 4 1 
 1 2 7 
= 1(28 + 2) + 2(14 1) + 3(4 + 4)
= 30 30 = 0.

Therefore, the highest minor of order 3 is zero. We next consider the second highest
order minor, i.e., the minor of order 2:
 
 2 3 
=   = 2 12 = 10 6= 0.
4 1 

Thus, the highest order nonzero minor of A is of order 2, and so, (A) = 2.
Dr. Satish Shukla 3 of 39

1 1 1
Example 2. In each case, find the rank of the matrix A = b + c c + a a + b
bc ca ab

Sol. Again, we start with the highest order minor of A, then:

 
 1 1 1 
 
|A| =  b + c c + a a + b 
 bc ca ab 
 
 1 0 0 
 
=  b + c a b a c  (applying C2 C2 C1 , C3 C3 C1 )
 bc c(a b) b(a c) 
 
 ab a c 
=  
c(a b) b(a c) 
= (a b)(b c)(c a).

We consider the following cases:

Case I. When a = b = c. In this case |A| = 0, and so, (A) < 3. Also, for a = b = c
 1 1 1 
 
we have A =  2a 2a 2a  . Clearly, all the minors of order 2 of A are zero, and so,
 a2 a2 a2 
(A) < 2. Now, clearly the minor of order 1 of A is nonzero , therefore (A) = 1.
Case II. When a = b 6= c. In this
 case |A| = 0, and so, (A) < 3. Also, for a = b 6= c
 1 1 1   
   1 1 
we have A =  a + c a + c 2a  , and a minor of order 2 =   = a c 6= 0.
 ac 2 a + c 2a 
ac a 
Therefore, (A) = 2. Since A is symmetric in a, b, c, therefore, if any two of a, b, c are
equal, and the remaining is not equal to the first two, the rank of A remains 2.
Case III. When a 6= b 6= c. In this case |A| =
6 0, and so, (A) = 3.

Nullity of a square Matrix. The nullity of a matrix is the excess of the order of matrix
over its rank, and it is denoted by (A). If A is a square matrix of order n and (A) = r,
the (A) = n (A) = n r.

Example 3. Determine the rank and nullity of the following matrices:

2 2 2 2
1 2 3 4
22 32 42 52
A= 32 42 52 62 .

42 52 62 72

1 4 9 16
4 9 16 25
Sol. Given matrix is: A =
9 16 25
.
36
16 25 36 49
Applying R4 R4 R3 , R3 R3 R2 ; R2 R2 R1 we obtain:

1 4 9 16
3 5 7 9
A 5
.
7 9 11
7 9 11 13
Dr. Satish Shukla 4 of 39

Applying R4 R4 R3 , R3 R3 R2 ; R2 R2 R1 we obtain:

1 4 9 16
2 1 2 7
A 2 2 2
.
2
2 2 2 2

Applying R3 21 R3 ; R3 R1 we obtain:

1 1 1 1
2 1 2 7
A
1
.
4 9 16
2 2 2 2

Applying R4 R4 2R1 , R3 R3 R1 ; R2 R2 2R1 we obtain:

1 1 1 1
0 1 4 9
A 0 3
.
8 15
0 0 0 0

Applying R3 R3 + 3R2 we obtain:

1 1 1 1
0 1 4 9
A 0 0 4 12
.

0 0 0 0

The above matrix is in the echelon form, therefore, the rank of matrix A;

(A) = no. of nonzero rows in the echelon form= 3 and (A) = 4 3 = 1.

Example 4. Determine the rank and nullity of the following matrices:

0 1 3 1
1 2 3 1 0 1 1
(i) 1 4 2 (ii)
3 1 0

2
2 6 5
1 1 2 0

1 2 3
Sol. (i). Let A = 1 4 2 . Applying the transformation R2 R2 R1 , R3
2 6 5
R3 2R1 we have:
1 2 3
A 0 2 1 .
0 2 1
Applying the transformation R3 R3 R2 we have:

1 2 3
A 0 2 1 .
0 0 0

The above matrix is in the echelon form, therefore, the rank of matrix A;
Dr. Satish Shukla 5 of 39

(A) = no. of nonzero rows in the echelon form= 2 and (A) = 3 2 = 1.

0 1 3 1
1 0 1 1
(II). Let A =
3 1 0
. Applying the transformation R1 R2 we have
2
1 1 2 0

1 0 1 1
0 1 3 1
A 3 1 0
.
2
1 1 2 0
Applying R3 R3 3R1 , R4 R4 R1 we have:

1 0 1 1
0 1 3 1
A 0 1 3 1
.

0 1 3 1
Applying R3 R3 R2 , R4 R4 R2 we have:

1 0 1 1
0 1 3 1
A 0
.
0 0 0
0 0 0 0
The above matrix is in the echelon form, therefore, the rank of matrix A;
(A) = no. of nonzero rows in the echelon form= 2 and (A) = 4 2 = 2.
Normal form of a Matrix. By elementary row and column transformation every matrix
can be reduced into one of the following forms:
   
Ir 0 Ir  
; ; Ir 0 ; [Ir ]
0 0 0
where r is the rank of the matrix. The above four forms are called the normal form of
the matrix.
Example 5. Reduce the matrix into the normal form and find its rank and nullity:

2 3 1 1
1 1 2 4
A= .
3 1 3 2
6 3 0 7

Sol. Applying R1 R2 :
1 1 2 4
2 3 1 1
A .
3 1 3 2
6 3 0 7
Applying R2 R2 2R1 , R3 R3 3R1 , R4 R4 6R1 :

1 1 2 4
0 5 3 7
A 0 4
.
9 10
0 9 12 17
Dr. Satish Shukla 6 of 39

Applying R2 R2 R3 :

1 1 2 4
0 1 6 3
A
0 4
.
9 10
0 9 12 17

Applying R3 R3 4R2 , R4 R4 9R2 :

1 1 2 4
0 1 6 3
A 0 0
.
33 22
0 0 66 44

Applying R4 R4 2R3 :

1 1 2 4
0 1 6 3
A
0 0 33 22 .

0 0 0 0

Applying the following series of operations in order:

C2 C2 + C1 , C3 C3 + 2C1 , C4 C4 + 4C1 ;
2 1
C3 C3 + 6C2 , C4 C4 + 3C2 ; C4 C4 C3 ; C3 C3 :
3 33

1 0 0 0  
0 1 0 0 I3 0
A 0 0 1
= .
0 0 0
0 0 0 0
It is the required normal form of the matrix and:

(A) = 3 and (A) = 4 3 = 1.

Note. Given a matrix of rank r, there exist two non-singular matrices P and Q
such that P AQ is in the normal form, and if A is of order m n, then P and Q
are square matrices of order m and n respectively.

Procedure to find P and Q. Suppose Amn be the given matrix. Write A =

Im AIn , where Im (the pre-multiplier) and In (the post-multiplier) are the identity
matrix of order m and n respectively. Now apply the row and column operations
on A to reduce it into a normal form, and apply the same row operations on Im
and column operations on In and get normal form of A = P AQ.

1 1 2
Example 6. For the matrix A = 1 2
3 , find two non-singular matrices
0 1 1
P and Q such that P AQ is in the normal form, hence find the rank and nullity
of A.
Dr. Satish Shukla 7 of 39

1 1 2 1 0 0 1 0 0
Sol. We write A = I3 AI3 , i.e., 1 2 3 = 0 1 0 A 0 1 0 .
0 1 1 0 0 1 0 0 1
Applying R2 R2 R1 (note that this row operation is applied to the pre-multiplier
as well):
1 1 2 1 0 0 1 0 0
0 1 1 = 1 1 0 A 0 1 0 .
0 1 1 0 0 1 0 0 1
Applying R3 R3 + R2 :

1 1 2 1 0 0 1 0 0
0 1 1 = 1 1 0 A 0 1 0 .
0 0 0 1 1 1 0 0 1

Applying C2 C2 C1 , C3 C3 2C1 (note that these column operations are applied

to the post-multiplier as well):

1 0 0 1 0 0 1 1 2
0 1 1 = 1 1 0 A 0 1 0 .
0 0 0 1 1 1 0 0 1
Applying C3 C3 C2

1 0 0 1 0 0 1 1 1
0 1 0 = 1 1 0 A 0 1 1
0 0 0 1 1 1 0 0 1

  1 0 0 1 1 1
I2 0
i.e., = P AQ, where P = 1 1 0 , Q = 0 1 1 .
0 0
1 1 1 0 0 1

Thus, (A) = 2 and (A) = 3 2 = 1.

Inverse of matrix by elementary operations. Suppose A be an invertible matrix of

order n. Then, to fixed the inverse of A by elementary row operations, write

A = In A (1)

where In is the unit matrix of order n. Now we convert this equation into the form
In = BA by applying a sequence of elementary row operation on equation (1). In
the new equation B is the inverse of A. To find B, we apply same sequence of row
operations on In (of R.H.S. of equation (1)) which we apply on A (of L.H.S. of (1)),
so that In of RH.S. is converted into B. If we want to apply elementary column
operations we write In = AIn , now we repeat whole process keeping in mind that the
sequence of column operations is now applied to the In .

1 1 3
Example 7. Find the inverse of the matrix A = 1 3 3 .
2 4 4

Sol. We have A = IA, i.e.,

1 1 3 1 0 0
1 3 3 = 0 1 0 A.
2 4 4 0 0 1
Dr. Satish Shukla 8 of 39

Applying R2 R2 R1 , R3 R3 + 2R1

1 1 3 1 0 0
0 2 6 = 1 1 0 A.
0 2 2 2 0 1

Applying R3 R3 + R2

1 1 3 1 0 0
0 2 6 = 1 1 0 A.
0 0 4 1 1 1

Applying R3 14 R3

1 1 3 1 0 0
0 2 6 = 1 1 0 A.
0 0 1 1/4 1/4 1/4

Applying R1 R1 3R3 , R2 R2 + 6R3

1 1 0 7/4 3/4 3/4
0 2 0 = 5/2 1/2 3/2 A.
0 0 1 1/4 1/4 1/4

Applying R2 21 R2

1 1 0 7/4 3/4 3/4
0 1 0 = 5/4 1/4 3/4 A.
0 0 1 1/4 1/4 1/4

Applying R1 R1 R2

1 0 0 3 1 3/2
0 1 0 = 5/4 1/4 3/4 A
0 0 1 1/4 1/4 1/4

3 1 3/2
i.e., I = BA, where B = 5/4 1/4 3/4 is the inverse of A.
1/4 1/4 1/4

Exercise (Assignment)

1 2 3 2
(Q.1) Find the rank and nullity of the matrix A = 2 3 5 1 . Ans. (A) = 2
1 3 4 5
and nullity is not defined.

1 1 1
(Q.2) Find the rank and nullity of the matrix A = a b c .
a3 b 3 c 3
Ans. (i) If a = b = c, then (A) = 1 (ii) if a = b 6= c, then (A) = 2 (iii) if
a 6= b 6= c, but a + b + c = 0, then (A) = 2 (iv) if a 6= b 6= c, but a + b + c 6= 0,
then (A) = 3.
Dr. Satish Shukla 9 of 39

(Q.3) Find the rank and nullity of the matrix for zero and nonzero values of a, b, c

a b c
A = ab
bc ca .
a2 b2 c 2
Ans. (i) If a = b = c = 0, then (A) = 0 (ii) if any one of a, b, c is zero and
other two are nonzero, then (A) = 2 (iii) if any two of a, b, c are zero and the
remaining one is nonzero, then (A) = 1.
(Q.4) Reduce the matrix into normal form and find the rank and nullity

2 3 4 5
3 8 1 3 6
4 5 6 (ii) A = 0
(i) A =
4 3 2 1 .
5 6 7
8 1 3 4
9 10 11 12
Ans. (i) (2), (A) = 2 (ii) (3), (A) =not defined.
(Q.5) Find two non-singular matrices P and Q such that P AQ is the normal form,
where
1 2 3
3 2 1 1 1
1
(i) A =
1 3
(ii) A = 1 1 1 .
2
3 1 1
2 1 3

(Q.6) Use row (or column) transformation and find the inverse of the matrix:

2 3 4 5 2 4
(i) A = 4 3 1 (ii) A = 2 1 1 .
1 2 4 4 1 0

Solution of System of Linear Equations. Consider a system of m linear equations in

n variables:
a11 x1 + a12 x2 + + a1n xn = b1

a21 x1 + a22 x2 + + a2n xn = b2


.. (2)
.


am1 x1 + am2 x2 + + amn xn = bm .

a11 a12 a1n
a12 a22 a2n
In matrix form this system can be written as AX = B, where A =

.. .. ... ..
. . .
am1 am2 amn

x1 b1
x2 b2
is the coefficient matrix, X = is the variable vector and b = .. is the con-

..
. .
xn bn
stant vector.  
We denote the augmented matrix by A B and

a11 a12 a1n b1
  a
12 a22 a2n b2
A B = .. .

.. ... .. ..
. . . .
am1 am2 amn bn
Dr. Satish Shukla 10 of 39

Then, we follow the following procedure to test the consistency and obtain the solution
of the system (2):
 

(A) If A B 6= (A), then system is inconsistent, and has no solution.
 

(B) If A B = (A) = n, then system is consistent, and has a unique solution.
 

(C) If A B = (A) < n, then system is consistent, and has infinity many
solutions with n (A) independent variables.

Example 8. Show that the following system is consistent and solve it:

x+y+z = 3
x + 2y + 3z = 4
x + 4y + 9z = 6.

1 1 1 x 3
Sol. Write the system as: AX = B, A = 1 2 3 , X = y , B = 4 . Now,

1 4 9 z 6
augmented matrix will be

  1 1 1 3
A B = 1 2 3 4 .
1 4 9 6

Applying R2 R2 R1 , R3 R3 R1

  1 1 1 3
A B 0 1 2 1 .
0 3 8 3

Applying R3 R3 3R2

  1 1 1 3
A B 0 1 2 1 .
0 0 2 0
 

It is the echelon form of the augmented matrix and it is clear that A B =
(A) = 3, which is equal to the number of unknown variables. Therefore, the given
system is consistent and has a unique solution. From the echelon form we have the
equations:
x+y+z = 3
y + 2z = 1
2z = 0.
Therefore, the solution will be:

x = 2, y = 1, z = 0.

Example 9. Test for consistency and solve (if consistent):

5x + 3y + 7z = 4
3x + 26y + 2z = 9
7x + 2y + 10z = 5.
Dr. Satish Shukla 11 of 39

5 3 7 x 4
Sol. Write the system as: AX = B, A = 3 26 2 , X = y , B = 9 .
7 2 10 z 5
Now, augmented matrix will be

  5 3 7 4
A B = 3 26 2 9 .
7 2 10 5
Applying R1 R1 2R2

  1 49 3 14
A B 3 26 2 9 .
7 2 10 5
Applying R2 R2 + 3R1 , R3 R3 + 7R1

  1 49 3 14
A B 0 121 11 33 .
0 341 31 93
31
Applying R3 R3 R
11 2

  1 49 3 14
A B 0 121 11 33 .
0 0 0 0
 

It is the echelon form of the augmented matrix and it is clear that A B =
(A) = 2, which is less than the number of unknown variables (3). Therefore, the given
system is consistent and has infinitely many solution. Since n (A) = 3 2 = 1, so
one variable in the given system id independent. From the echelon form we have two
equations:
x 49y + 3z = 14
121y + 11z = 33.
Since one variable is independent, let z = k, then from the above equations we get
7 16 3+k
x= k, y = , z = k.
11 11 11

Example 10. Show that the following system is inconsistent:

x 2y + z w = 1
3x 2z + 3w = 4
5x 4y + w = 3.

Sol. The augmented matrix of given system is:

  1 2 1 1 1
A B = 3 0 2 3 4 .
5 4 0 1 3
Applying R2 R2 3R1 , R3 R3 5R1

  1 2 1 1 1
A B 0 6 5 6 1 .
0 6 5 6 2
Dr. Satish Shukla 12 of 39

Applying R3 R3 R2

  1 2 1 1 1
A B 0 6 5 6 1 .
0 0 0 0 3
 

It is the echelon form of the augmented matrix and it is clear that A B = 3 6=
(A) = 2. Therefore, the given system is inconsistent and has no solution.

Example 11. Investigate the values of and so that the equations:

x+y+z = 6
x + 2y + 3z = 10
x + 2y + z = ,

have (i) no solution and (ii) a unique solution (iii) an infinite number of solu-
tions.

1 1 1 x
Sol. Write the system as: AX = B, where A = 1 2 3 ,X =
y ,B =
1 2 z
6
10 . Now, augmented matrix will be


  1 1 1 6
A B = 1 2 3 10 .
1 2

Applying R2 R2 R1 , R3 R3 R1

  1 1 1 6
A B 0 1 2 4 .
0 1 1 6

Applying R3 R3 R2

  1 1 1 6
A B 0 1 2 4 .
0 0 3 10

It is the echelon form of the augmented matrix. Now, we consider the following cases:
 

(i) System has no solution: It is possible only when A B 6= (A), i.e., when
= 3 and 6= 10.
 

(ii) System has a unique solution: It is possible only when A B = (A) = 3,
i.e., 6= 3 and R.
 

(iii) System has infinitely many solutions: It is possible only when A B =
(A) < 3, i.e., = 3 and = 10.
Dr. Satish Shukla 13 of 39

Example 12. For what value(s) of k the equations:

x+y+z = 1
x + 2y + 4z = k
x + 4y + 10z = k 2 ,

have a solution and solve completely in each case.

Sol. The augmented matrix of given system is:

  1 1 1 1
A B = 1 2 4 k .
1 4 10 k 2
Applying R2 R2 R1 , R3 R3 R1

  1 1 1 1
A B 0 1 3 k 1 .
0 3 9 k2 1
Applying R3 R3 3R2

  1 1 1 1
A B 0 1 3 k1 . (3)
2
0 0 0 k 3k + 2
 

The system of equations will have a solution if it is consistent, i.e., if A B =
2
(A), and it is possible
 
when k 3k + 2 = 0, i.e., k = 1 or k = 2. In both the
only
cases we have A B = (A) = 2 < 3 (no. of variables). Therefore, in both the
cases we have 3 2 = 1 independent variable and infinitely many solutions of the given
system. We consider the following cases:
Case I. If k = 1, then from (3) we have the following equations:
x+y+z = 1
y + 3z = 0.
Since one variable is independent, choose z = a we have the following solution:
x = 1 + 2a, y = 3a, z = a.
Case I. If k = 2, then from (3) we have the following equations:
x+y+z = 1
y + 3z = 1.
Since one variable is independent, choose z = b we have the following solution:
x = 2b, y = 1 3b, z = b.

Example 13. For what value of k the equations:

2x 3y + 6z 5t = 3
y 4z + t = 1
4x 5y + 8z 9t = k,

(i) have no solution (ii) have infinitely many solutions.

Dr. Satish Shukla 14 of 39

Sol. The augmented matrix of given system is:

  2 3 6 5 3
A B = 0 1 4 1 1 .
4 5 8 9 k

Applying R3 R3 2R1

  2 3 6 5 3
A B 0 1 4 1 1 .
0 1 4 1 k 6

Applying R3 R3 R2

  2 3 6 5 3
A B 0 1 4 1 1 .
0 0 0 0 k7
 

The system of equations will have a solution if it is consistent, i.e., if A B =
(A), it is possible only when k 7 = 0, i.e., k = 7. In this case we have
 and 

A B = (A) = 2 < 4 (no. of variables). Therefore, we have 4 2 = 1
independent variables and infinitely many solutions of the given system. Now, for
k=7, the above system reduces to:

2x 3y + 6z 5t = 3
y 4z + t = 1.

Since two variables are independent, choose z = a and t = b we have the following
solution:
x = 3 + 3a + b, y = 1 + 4a b, z = a, t = b.
Homogeneous system of equations. We consider the following system of equations:

a11 x1 + a12 x2 + + a1n xn = 0
a21 x1 + a22 x2 + + a2n xn = 0

.. (4)
.


am1 x1 + am2 x2 + + amn xn = 0.

Note that, in the above system all the constant of R.H.S. are zero and such a system is
called the homogeneous system. In matrix form it can be written as AX = 0, where A
is the coefficient
 matrix
  and X is the variable vector. In such systems the augmented
matrix is A B = A 0 . Note that,  in any
 case the rank of coefficient matrix
A and the rank of augmented matrix A B are equal. Therefore, homogeneous
systems are always consistent and have a solution, namely, x1 = x2 = = xn = 0 is
always a solution of system (4) and it is called the zero solution or trivial solution.

Note. (A) If (A) = n =number of variables, then system (4) has only the trivial
solution.
(B) If (A) < n =number of variables, then system (4) has a nontrivial solution.
 
(C) For homogeneous system, we find the rank of only A, not of A B .

Example 14. Solve the equations: x+3y-2z=0, 2x-y+4z=0, x-11y+14z=0.

Dr. Satish Shukla 15 of 39

Sol. The coefficient matrix of the given system is:

1 3 2
A = 2 1 4
1 11 14

Applying R2 R2 2R1 , R3 R3 R1

1 3 2
A 0 7 8 .
0 14 16

Applying R3 R3 2R2
1 3 2
A 0 7 8 .
0 0 0
It is the echelon form. Clearly, (A) = 2 < 3 =number of variables. Therefore, system
has a nontrivial solution and n (A) = 3 2 = 1 variable is independent. By the
echelon form we have the following equations:

x + 3y 2z = 0
7y + 8z = 0.

Since one variable is independent, choose z = k, we obtain from the above equations:
10k 8k
x= , y= , z = k.
7 7

Example 15. For which value of b the following system:

2x + y + 2z = 0
x + y + 3z = 0
4x + 3y + bz = 0

has (i) trivial solution (ii) nontrivial solution. Find the nontrivial solution.

Sol. The coefficient matrix of the given system is:

2 1 2
A= 1 1
3
4 3 b

Applying R1 R2 ; then R2 R2 2R1 , R3 R3 4R1

1 1 3
A 0 1 4 .
0 1 b 12

Applying R3 R3 R2
1 1 3
A 0 1 4 .
0 0 b8
It is the echelon form. We consider the following cases:
(i) If b 6= 8, then obviously (A) = 3 =number of variables and so system has a trivial
Dr. Satish Shukla 16 of 39

solution.
(ii) If b = 8, then obviously (A) = 2 < 3 =number of variables and so system has
nontrivial solution. In this case n (A) = 3 2 = 1 variable will be independent. By
the echelon form we have the following equations:
x+y+z = 0
y 4z = 0.
Since one variable is independent, choose z = k, we obtain from the above equations:

x = k, y = 4k, z = k.

Exercise (Assignment)

(Q.1) Investigate the values of and so that the equations:

2x + 3y + 5z = 9
7x + 3y 2z = 8
2x + 3y + z = ,

have (i) no solution and (ii) a unique solution (iii) an infinite number of solutions.
Ans. (i) = 5, 6= 9 (ii) 6= 5, R (iii) = 5, = 9.
(Q.2) Test the consistency of the system:

x + 2y z = 3
2x 2y + 3z = 2
3x y + 2z = 1
xy+z = 1.
Ans. Consistent and has a unique solution x = 1, y = 4, z = 4.
(Q.3) Show that the following system is consistent and sole it:

x + 2y 5z = 9
3x y + 2z = 5
2x + 3y z = 3.

Ans. Consistent and has a unique solution x = 1/2, y = 3/2, z = 5/2.

(Q.4) Examine the consistency of the following system:

5x + 3y + 14z = 4
y + 2z = 1
x y + 2z = 0
2x + y + 6z = 2.

Ans. Inconsistent (has no solution).

(Q.5) Show that the system of equations as given below is consistent if and only if
a + c = 2b and find the solution(s) when exists:

3x + 4y + 5z = a
4x + 5y + 6z = b
5x + 6y + 7z = c.
Dr. Satish Shukla 17 of 39

(Q.6) Find the values of so that the system of equations has non trivial solution and
hence find the non trivial solution:
( 1)x + (3 + 1)y + 2z = 0
( 1)x + (4 2)y + ( + 3)z = 0
2x + (3 + 1)y + 3( 1)z = 0.
Ans. = 0, 1, 3.
(Q.7) Solve the following system of equations completely:
2w + 3x y z = 0
4w 6x 2y + 2z = 0
6w + 12x + 3y 4z = 0.
k1 k2
Ans. w = 2
, x= 2
, y = k1 , z = k2 .

Eigenvalues (characteristic
value) and eigenvectors
(characteristic vectors)
of a ma-
a11 a12 a1n x1
a12 a22 a2n x2
trix. Suppose A = .. be a quare matrix and X = .. and

.. .. ..
. . . . .
am1 am2 amn xn
we want to find nonzero vectors X such that A transforms X into itself or a scalar
multiple of itself, i.e., we want to find vectors X such that
AX = X or (A I)X = 0. (5)
for some scalar . The equation (5) is called the eigenvalue or characteristic value
problem of A, X is called the eigenvector or characteristic vector of A and the scaler
is called the corresponding eigenvalue characteristic value of A.
Note that, the equation (5) is equivalent to the following homogeneous system of
equations:
(a11 )x1 + a12 x2 + + a1n xn = 0

a21 x1 + (a22 )x2 + + a2n xn = 0

..
.


am1 x1 + am2 x2 + + (ann )xn = 0
and we know that the above homogeneous system will have a nonzero (nontrivial)
a11 a12 a1n
a12 a 22 a 2n

solution if the matrix A = = A I has rank less

.. .. . . ..
. . . .
am1 am2 ann
that n, i.e., A I is singular, i.e.,
|A I| = 0. (6)
Equation (6) is called the eigen equation of A.

Procedure to find the eigenvalues and eigenvectors:

(A). First find the eigenvalues of A, then eigenvectors.
(B). To find the eigenvalues of A solve the eigen equation of A.
(C). To find the eigenvectors of A solve the eigenvalue problem of A.
Dr. Satish Shukla 18 of 39

 
5 4
Example 16. Find the eigenvalues and eigenvectors of the matrix A = .
1 2

Sol. First we find the eigenvalues of A. then the eigen equation of A will be
 
 5 4 
|A I| = 0, i.e.,   = 0.
1 2 

Solving the above equation we get 2 7 + 6 = 0, i.e., = 1, 6. These are the

eigenvalues of A.
We now find the corresponding eigenvectors of A. Then:
(I) For = 1, the eigenvalue problem will be (A I)X = 0, i.e.,
    
4 4 x1 0
(A 1 I)X = 0 = = .
1 1 x2 0

Applying R2 R2 41 R1 we get
    
4 4 x1 0
= .
0 0 x2 0
It is the echelon form. Clearly, the rank of matrix is 1 < 2 =no. of variables. Therefore,
21 = 1 variable will be independent. Now from the echelon form we have the equation:
4x1 +4x2 =0, and one variable is independent choosing x2 = 1 we get x1 = 1. Thus,
1
X= is the corresponding eigenvector.
1
(II) For = 6, the eigenvalue problem will be (A I)X = 0, i.e.,
    
1 4 x1 0
(A 6 I)X = 0 = = .
1 4 x2 0
Applying R2 R2 + R1 we get
    
1 4 x1 0
= .
0 0 x2 0
It is the echelon form. Clearly, the rank of matrix is 1 < 2 =no. of variables. Therefore,
21 = 1 variable will be independent. Now from the echelon form we have the equation:
x1 +  4x2 = 0, and one variable is independent choosing x2 = 1 we get x1 = 4. Thus,
4
X= is the corresponding eigenvector.
1

1 1 3
Example 17. Find the latent values and latent vectors of the matrix 1 5 1 .
3 1 1

Sol. The eigen equation of the given matrix is: |A I| = 0, i.e.,

 
 1 1 3 
 
 1
 5 1  = 0 = 3 72 + 36 = 0.
 3 1 1 

Since the aboive equation satisfied by = 2, therefore, we write it: ( + 2)(2

9 + 18) = 0 or ( + 2)( 3)( 6) = 0. So, there are three eigenvalues of the given
Dr. Satish Shukla 19 of 39

matrix, = 3, 6, 2.
Now we find the corresponding eigenvectors, then the eigenvalue problem of A will be
(A I)X = 0, i.e.,

1 1 3 x1
1 5 1 x2 = 0. (7)
3 1 1 x3

(I) For = 3. Putting = 3 in (7) we obtain:

2 1 3 x1
1 2 1 x2 = 0.
3 1 2 x3

Applying R1 R2
1 2 1 x1
2 1 3 x2 = 0.
3 1 2 x3
Applying R2 R2 + 2R1 , R3 R3 3R1

1 2 1 x1
0 5 5 x2 = 0.
0 5 5 x3

Applying R3 R3 R3 + R2

1 2 1 x1
0 5 5 x2 = 0.
0 0 0 x3

It is the echelon form. Clearly, the rank of matrix is 2 < 3 =no. of variables. Therefore,
3 2 = 1 variable will be independent. Now from the echelon form we have the
equations: x1 + 2x2 + x3 = 0; and 5x2 + 5x3 = 0; andone variable is independent
1
choosing x3 = 1 we get x2 = 1, x1 = 1. Thus, X = 1 is the corresponding
1
eigenvector.
(II) For = 6. Putting = 6 in (7) we obtain:

5 1 3 x1
1 1 1 x2 = 0.
3 1 5 x3

Applying R1 R2
1 1 1 x1
5 1 3 x2 = 0.
3 1 5 x3
Applying R2 R2 + 5R1 , R3 R3 3R1

1 1 1 x1
0 4 8 x2 = 0.
0 4 8 x3
Dr. Satish Shukla 20 of 39

Applying R3 R3 R3 + R2

1 1 1 x1
0 4 8 x2 = 0.
0 0 0 x3

It is the echelon form. Clearly, the rank of matrix is 2 < 3 =no. of variables. Therefore,
3 2 = 1 variable will be independent. Now from the echelon form we have the
equations: x1 x2 + x3 = 0; and 4x2 + 8x3 = 0; and one
variable is independent
1
choosing x3 = 1 we get x2 = 2, x1 = 1. Thus, X = 2 is the corresponding
1
eigenvector.
(III) For = 2. Putting = 2 in (7) we obtain:

3 1 3 x1
1 7 1 x2 = 0.
3 1 3 x3

Applying R1 R2
1 7 1 x1
3 1 3 x2 = 0.
3 1 3 x3
Applying R2 R2 3R1 , R3 R3 3R1

1 7 1 x1
0 20 0 x2 = 0.
0 20 0 x3

Applying R3 R3 R3 R2

1 7 1 x1
0 20 0 x2 = 0.
0 0 0 x3

It is the echelon form. Clearly, the rank of matrix is 2 < 3 =no. of variables. Therefore,
3 2 = 1 variable will be independent. Now from the echelon form we have the
equations: x1 + 7x2 + x3 = 0; and 20x2 = 0, i.e., x2 = 0; and one variable is
1
independent choosing x3 = 1 we get x1 = 1. Thus, X = 0 is the corresponding
1
eigenvector.

3 1 4
Example 18. Find the eigenvalues and eigenvectors of the matrix 0 2 6 .
0 0 5

Sol. The eigen equation of the given matrix is: |A I| = 0, i.e.,

 
 3 1 4 
 
 0
 2 6  = 0 = (3 )(2 )(5 ) = 0.
 0 0 5 
Dr. Satish Shukla 21 of 39

So, there are three eigenvalues of the given matrix, = 2, 3, 5.

Now we find the corresponding eigenvectors, then the eigenvalue problem of A will be
(A I)X = 0, i.e.,

3 1 4 x1
0 2 6 x2 = 0. (8)
0 0 5 x3

(I) For = 2. Putting = 2 in (8) we obtain:

1 1 4 x1
0 0 6 x2 = 0.
0 0 3 x3

Applying R3 R3 21 R2

1 1 4 x1
0 0 6 x2 = 0.
0 0 0 x3
It is the echelon form. Clearly, the rank of matrix is 2 < 3 =no. of variables. Therefore,
3 2 = 1 variable will be independent. Now from the echelon form we have the
equations: x1 + x2 + 4x3 = 0; and 6x3 = 0, i.e.,
x3 =0; and one variable is independent
1
choosing x2 = 1 we get x1 = 1. Thus, X = 1 is the corresponding eigenvector.
0
(II) For = 3. Putting = 3 in (8) we obtain:

0 1 4 x1
0 1 6 x2 = 0.
0 0 2 x3

Applying R2 R2 + R1
0 1 4 x1
0 0 10 x2 = 0.
0 0 2 x3
Applying R3 R3 15 R2

0 1 4 x1
0 0 10 x2 = 0.
0 0 0 x3

It is the echelon form. Clearly, the rank of matrix is 2 < 3 =no. of variables. Therefore,
3 2 = 1 variable will be independent. Now from the echelon form we have the
equations: x2 + 4x3 = 0; and 10x3 = 0, i.e., x3= 0 and so x2 = 0; and one variable is
1
independent choosing x1 = 1. Thus, X = 0 is the corresponding eigenvector.

0
(III) For = 5. Putting = 5 in (8) we obtain:

2 1 4 x1
0 3 6 x2 = 0.
0 0 0 x3
Dr. Satish Shukla 22 of 39

It is the echelon form. Clearly, the rank of matrix is 2 < 3 =no. of variables. Therefore,
3 2 = 1 variable will be independent. Now from the echelon form we have the
equations: 2x1 + x2 + 4x3 = 0; and 3x2 + 6x3 = 0; andone variable is independent
3
choosing x3 = 1 we get x2 = 2, x1 = 3. Thus, X = 2 is the corresponding
1
eigenvector.

2 2 1
Example 19. Find the eigenvalues and eigenvectors of the matrix 1 3 1 .
1 2 2

Sol. The eigen equation of the given matrix is: |A I| = 0, i.e.,

 
 2 2 1 

 1
 3 1  = 0 = 3 72 + 11 5 = 0.
 1 2 2 

Since the above equation is satisfied by = 1, therefore, we write it: ( 1)(2 6 +

5) = 0 or ( 1)( 1)( 5) = 0. So, there are three eigenvalues of the given matrix,

= 1, 1, 5.

Now we find the corresponding eigenvectors, then the eigenvalue problem of A will
be (A I)X = 0, i.e.,

2 2 1 x1
1 3 1 x2 = 0. (9)
1 2 2 x3

(I) For = 5. Putting = 3 in (9) we obtain:

3 2 1 x1
1 2 1 x2 = 0.
1 2 3 x3

Applying R1 R2
1 2 1 x1
3 2 1 x2 = 0.
1 2 3 x3
Applying R2 R2 + 3R1 , R3 R3 R1

1 2 1 x1
0 4 4 x2 = 0.
0 4 4 x3

Applying R3 R3 R3 + R2

1 2 1 x1
0 4 4 x2 = 0.
0 0 0 x3

It is the echelon form. Clearly, the rank of matrix is 2 < 3 =no. of variables. Therefore,
3 2 = 1 variable will be independent. Now from the echelon form we have the
Dr. Satish Shukla 23 of 39

equations: x1 2x2 + x3 = 0; and 4x2 + 4x3 = 0; and one

variable is independent
1
choosing x3 = 1 we get x2 = 1, x1 = 1. Thus, X = 1 is the corresponding
1
eigenvector.
(II) For = 1. Putting = 1 in (9) we obtain:

1 2 1 x1
1 2 1 x2 = 0.
1 2 1 x3

Applying R2 R2 R1 , R3 R3 R1

1 2 1 x1
0 0 0 x2 = 0.
0 0 0 x3

It is the echelon form. Clearly, the rank of matrix is 1 < 3 =no. of variables. Therefore,
3 1 = 2 variable will be independent. Now from the echelon form we have only one
equation: x1 + 2x2 + x3 = 0; and two variables are independent choosing: (i)x3 = 1,

1 2
x2 = 0 we get x1 = 1 (ii) x3 = 0, x2 = 1 we get x1 = 2 Thus, X = 0 , 1
1 0
are two linearly independent eigenvectors corresponding eigenvalue = 1.

Example 20. Find

  the latent values and latent vectors of the matrix
cos sin
.
sin cos

Sol. The eigen equation of the given matrix is: |A I| = 0, i.e.,

 
 cos sin 
  = 0 = 2 1 = 0.
 sin cos 

Since the above equation is satisfied by = 1, therefore, we write it: ( 1)(2 6 +

5) = 0 or ( 1)( 1)( 5) = 0. So, there are two latent values of the given matrix,
= 1, 1.
Now we find the corresponding eigenvectors, then the eigenvalue problem of A will be
(A I)X = 0, i.e.,
  
cos sin x1
= 0. (10)
sin cos x2

(I) For = 1. Putting = 1 in (10) we obtain:

     
cos 1 sin x1 sin (/2) cos (/2) x1
= 0 = = 0.
sin cos 1 x2 sin (/2) cos (/2) x2

Applying R2 R2 + R1
  
sin (/2) cos (/2) x1
= 0.
0 0 x2

It is the echelon form. Clearly, the rank of matrix is 1 < 2 =no. of variables. There-
fore, 2 1 = 1 variable will be independent. Now from the echelon form we have the
Dr. Satish Shukla 24 of 39

equations: sin (/2) x1 + cos (/2) x2 = 0;

 and one variable
 is independent choosing
cot (/2)
x2 = 1 we get x1 = cot (/2) . Thus, X = is the corresponding eigenvec-
1
tor.
 
tan (/2)
(I) For = 1. X = is the corresponding eigenvector.
1

Some properties of eigenvalues and eigenvectors.

1
Property I. If be an eigenvalue of a non-singular matrix A, show that
is an eigen-
value of the matrix A1 .
Proof. Since is an eigenvalue of A, we have:

|A I| = 0
= |A1 ||A I| = 0
= |A1 (A I)| = 0
= |I A1 | = 0
 
 1 1 
= A I  = 0


 
 1 1 
= A I  = 0.
 

It shows that 1 = 1

is an eigenvalue of A1 .

Property II. If be an eigenvalue of a non-singular matrix A, show that |A|

is an
eigenvalue of the matrix adjA.
adjA
Proof. We know that 1 is an eigenvalue of A1 and A1 = |A| , therefore we have
 
 1 1 
= A I  = 0

 
 adjA 1 
=  I  = 0
|A|
 
1  |A| 
= adjA I =0
|A|  
 
 |A| 
= adjA
 I = 0.

Property III. The eigenvalue of an idempotent matrix are either zero or unity.
Proof. If A is idempotent, then A2 = A. If is eigenvalue of A. then we have:

AX = X
= A(AX) = A(X)
= A2 X = AX
= X = (X) (since A2 = A and AX = X)
= (2 )X = 0
= = 0 or 1.
Dr. Satish Shukla 25 of 39

Exercise (Assignment)


8 6 2
(Q.1) Find the eigenvalues and eigenvectors of the matrix 6 7 4 .
2 4 3
Sol. = 0, 3, 15. For = 0, X = [1 2 2]T , for = 3, X = [2 1 2]T , for = 15,
X = [2 2 1]T .

2 2 3
(Q.2) Find the eigenvalues and eigenvectors of the matrix 2 1 6 .
1 2 0
Sol. = 3, 3, 5. For = 5, X = [1 2 1]T , for = 3, X = [3 0 1]T , [2 1 0]T .

6 2 2
(Q.3) Find the eigenvalues and eigenvectors of the matrix 2 3 1 .
2 1 3
Sol. = 2, 2, 8. For = 8, X = [2 1 1]T , for = 2, X = [1 0 2]T , [1 2 0]T .

2 1 1
(Q.4) Find the eigenvalues and eigenvectors of the matrix 1 2 1 .
0 0 1
Sol. = 1, 1, 3. For = 3, X = [1 1 0]T , for = 1, X = [1 1 0]T , [1 0 1]T .

1 6 4
(Q.5) Find the eigenvalues and eigenvectors of the matrix 0 4 2 .
0 6 3
Sol. = 0, 1, 1. For = 0, X = [2 1 2]T , for = 1, X = [1 0 0]T , [0 1 3/2]T .

(Q.6) Prove that: (i) The eigenvalues of A and its transpose A0 or AT are same;
(ii) the sum of eigenvalues of a matrix is its trace, i.e., the sum of its principle
diagonal;
(iii) The product of the eigenvalues is equal to its determinant
(iv) If is an eigenvalues of A, then m is an eigenvalue of Am .

Linear transformations.
Y Y
X1 (x, y) BX1 = X2 (y, x) X1 (x, y)

O X O X

AX1 = X2 (x, y)

Reflection (A) about x-axis Rotation (B) about origin by an angle /2

Dr. Satish Shukla 26 of 39

2
Consider the two dimensional
 vector
 space R , i.e., the cartesian plane or xy-plane.
1 0 0 1
Let A = ,B = be two matrices. We note that, when the matrix
0 1 1 0  
x
A is multiplied to a vector X1 = , then it transform to the vector X1 into the
  y
x
vector X2 = AX1 = , i.e., the multiplication of matrix A to any vector X1 is
y
equivalent to the reflection of that vector about x-axis, as shown in the figure. In this
case, we say that A is a transformation which is a reflection. Similarly, the matrix B
is a rotation and multiplication of B to any vector is equivalent to the rotation of that
vector about origin by an angle /2.
In general, multiplication of every square matrix A of order n to a vector of n-
dimensional vector space represents a transformation to that vector, and so, every
square matrix is called a transformation. We notice that if X, Y are two vectors and
a, b are two numbers (scaler), then

A(aX + bY ) = aAX + bAY.

This property is called the linearity property and since every square matrix satisfy it,
therefore every square matrix is called a linear transformation.

The range and null spaces of a matrix or transformation. Suppose, T is a linear

transformation with matrix A of order m n. The set of all linear combinations of the
column vectors of A is called the range of T , and the dimension of range space of T ,
i.e., the no. of linearly independent column vectors of A is called the rank of T . The
set of all column vectors X satisfying the equation AX = 0 is called the null space of
T and the dimension of null space is called the nullity of T .

The rank-nullity theorem. The sum of rank and nullity of a linear transformation
(matrix) is equal to the number of columns in a matrix. If A is a matrix of order m n
then:
rank(T )+nullity(T ) = m.

Example
 21. Let a linear
 transformation T : R3 R2 is given by the matrix
1 2 6
A= , then find the range, null space, rank and nullity of A and
2 4 12
verify the rank-nullity theorem.

Sol. To find the rank of A we find the number of linearly independent column vectors
 the matrix into echelon form. Applying R2 R2 + 2R1 we
of A. For this we reduce
1 2 6
get A . Therefore, the rank of A is 1, i.e., the number of linearly
0 0 0
independent rows (which is also, the the number of linearly independent columns) is
equal to 1. Therefore

rank(T ) = 1, Range of T = set of L.C. of columns of A = {aC}

where a R and C is any column of A. For nullity, consider the system AX = 0,

x1
where X = x2 . Then, since rank of the matrix is 1, this system is equivalent to

x3
the single equation
x1 + 2x2 6x3 = 0
Dr. Satish Shukla 27 of 39

and we have 3 1 = 2 independent variables in the system AX = 0. Suppose x2 =

x2 = 1, x + 3 = 0 = x1 = 2, and so, the solution vectors
0, x3 = 1 = x1 = 6 and
6 2
are S1 = 0 , S2 =
1 . Therefore
1 0

Nullity(T ) = 2, Null space of T = set of L.C. of S1 , S2 = {aS1 + bS2 }

where a, b R.

Similarity Transformation. Suppose, A, B be are two square matrices or two linear

transformations. Suppose there exists an invertible matrix C such that B = C 1 AC.
Then the matrix A and B are called similar matrices and the transformation which
maps the matrix A into the matrix B is called the similarity transformation.

Theorem 1. Similar matrices have same eigenvalues.

Proof. Suppose A and B are similar. Then, there exists an invertible matrix C such
that B = C 1 AC. We know that

|B I| = |C 1 AC I| = |C 1 AC C 1 IC|
= |C 1 ||A I||C|
= |A I|.

It shows that A and B have same eigen equations, and so, same eigenvalues.

Diagonalizable matrix. An n n square matrix A is said to be diagonalizable if there

exists an invertible matrix C such that C 1 AC is a diagonal matrix. In other words,
a matrix A is said to be diagonalizable if it is similar to a diagonal matrix.

Remark 1. If A is diagonalizable, then there exists an invertible matrix C such

that C 1 AC = , where is a diagonal matrix. Therefore, and A are similar,
and so, have same eigenvalues. But the eigenvalues of a diagonal matrix are the
entries of the diagonal of , therefore, we conclude that if A is diagonalizable,
then the diagonal entries of are the eigenvalues of A.

We note that not every matrix is diagonalizable.

 
1 2
Example 22. Show that the matrix A = is not diagonalizable.
0 1

Sol. Suppose A is diagonalizable, then there exists an invertible matrix C such that
C 1 AC = , where is a digonal matrix with eigenvalues of A as the diagonal entries,
which are obviously = 1, 1. Therefore,
 
1 1 0
C AC = = .
0 1

On multiplying by C from left and by C 1 from right we obtain:

 
1 0
A=C C 1 = I.
0 1
Dr. Satish Shukla 28 of 39

This contradiction shows that A is not diagonalizable.

Next, we consider a particular type of matrix which is always diagonalizable and useful
in the study of quadratic forms.

Hermitian and skew-Hermitian matrices. A complex square matrix A is called Her-

mitian if it is equal to its conjugate transpose, i.e., if

A = A = AT .

where bar represents the complex conjugate. A is said to be skew-Hermitian if

A = A.

Properties of Hermitian Matrices.

The entries of the main diagonal of a Hermitian matrix are always real.

If A = [aij ]nn is Hermitian, the aij = aji .

Every real symmetric matrix is Hermitian.

Every Hermitian matrix is digonalizable.

Theorem 2. Eigenvalues of a Hermitian matrix are real.

Proof. Suppose A is Hermitian, an eigenvalue of A and X be the corresponding

eigenvector. Then we have

AX = X = X AX = X X.

Taking conjugate transpose we obtain

(X AX) = (X X) = X A (X ) = X (X )
= X AX = X X (since A = A, (X ) = X)
= X X = X X (since AX = X)

= ( )XX = 0
= = .

Therefore, is real.

Theorem 3. Every square matrix A can be expressed as H1 + iH2 , where H1 , H2

are two Hermitian matrix.

Proof. Suppose A be a square matrix. Then we write:

A + A A A
A = +i = H1 + iH2
2 2i
A+A
where H1 = 2
,and H2 = AA2i
. Note that

A + A A + (A ) A + A


H1 = = = = H1 .
2 2 2

A A A (A ) A A


H2 = = = = H2 .
2i 2i 2i
Dr. Satish Shukla 29 of 39

Therefore, H1 and H2 are Hermitian.

Diagonalization of Hermitian matrix. There is a beautiful property about the Hermi-

tian matrix that, if it is of order n then it has n linearly independent eigenvectors. We
see that due to this property, every Hermitian matrix is diagonalizable.
Suppose
a11 a12 a13
A = a21 a22 a23
a31 a32 a33
be a Hermitian matrix with eigenvalues 1 , 2 and 3 and corresponding linearly inde-
pendent eigenvectors C1 , C2 and C3 respectively. Then AC1 = 1 C1 , AC2 = 2 C2 and
AC3 = 3 C3 . Let  
C = C1 C2 C3
be the matrix with the eigenvectors as its columns. Then C is invertible (since its
columns are L.I. and so it has rank 3). Now
   
AC = A C1 C2 C3 = AC1 AC2 AC3
 
= 1 C 1 2 C 2 3 C 3
 
= C1 C2 C3
= C.

Multiplying by C 1 from the left we get C 1 AC = . Therefore, A is diagonalizable.

Process for orthogonal diagonalization of Hermitian matrices. (i) Find the

eigenvalues 1 , 2 , . . . , n of Hermitian matrix.
(ii) Find the eigenvectors C1 , C2 , . . . , Cn for each eigenvalue.
(iii) Construct the transformation
C = [C1 C2 Cn ].
1 0 0
0 2 0
(iv) Write C 1 AC = = .. .. = diag(1 , 2 , . . . , n ).

.. . .
. . . .
0 0 n

2 1 1
Example 23. Diagonalize the following matrix: A = 1 2 1 .
1 1 2

Sol. The eigen equation of A is: |A I| = 0, i.e.,

 
 2 1 1 
 
 1 2 1  = 0 = 3 62 + 9 4 = 0.
 
 1 1 2 

On solving we get:
= 1, 1, 4.
Now we find the corresponding eigenvectors, then the eigenvalue problem of A will be
(A I)X = 0, i.e.,

2 1 1 x1
1 2 1 x2 = 0. (11)
1 1 2 x3
Dr. Satish Shukla 30 of 39

(I) For = 1. Putting = 1 in (11) we obtain:

1 1 1 x1
1 1 1 x2 = 0.
1 1 1 x3

Applying R2 R2 + R1 , R3 R3 R1

1 1 1 x1
0 0 0 x2 = 0.
0 0 0 x3

It is the echelon form. Clearly, the rank of matrix is 1 < 3 =no. of variables. There-
fore, 3 1 = 2 variable will be independent. Now from the echelon form we have only
one equation: x1 x2 + x3 = 0; and two variables are independent choosing:

1
(i) x3 = 1, x2 = 0 we get x1 = 1 and the first eigenvector is C1 = 0 .
1
1
(ii) x3 = 0, x2 = 1 we get x1 = 1 and the second eigenvector is C2 = 1 .

0
1
(II) For = 4. The third eigenvector is C3 = 1 . Thus, the required transfor-

1
1 1 1
mation will be C = [C1 C2 C3 ] = 0 1 1 and
1 0 1

1 0 0
C 1 AC = = 0 1 0 .
0 0 4

3 1 1
Example 24. Diagonalize the following matrix: A = 1 3 1 .
1 1 3

Sol. The eigen equation of A is: |A I| = 0, i.e.,

 
 3 1 1 
 
 1 3 1  = 0 = 3 92 + 24 16 = 0.
 
 1 1 3 

On solving we get:
= 4, 4, 1.
Now we find the corresponding eigenvectors, then the eigenvalue problem of A will be
(A I)X = 0, i.e.,

3 1 1 x1
1 3 1 x2 = 0. (12)
1 1 3 x3
Dr. Satish Shukla 31 of 39

(I) For = 4. Putting = 1 in (11) we obtain:

1 1 1 x1
1 1 1 x2 = 0.
1 1 1 x3

Applying R2 R2 + R1 , R3 R3 + R1

1 1 1 x1
0 0 0 x2 = 0.
0 0 0 x3

It is the echelon form. Clearly, the rank of matrix is 1 < 3 =no. of variables. Therefore,
3 1 = 2 variable will be independent. Now from the echelon form we have only one
equation: x1 + x2 + x3 = 0; and two variables are independent choosing:

1
(i) x3 = 1, x2 = 0 we get x1 = 1 and the first eigenvector is C1 = 0 .

1
1
(ii) x3 = 0, x2 = 1 we get x1 = 1 and the second eigenvector is C2 = 1 .
0
1
(II) For = 1. For = 1 the eigen vector is C3 = 1 .
1
1 1 1
Thus, the required transformation will be C = [C1 C2 C3 ] = 0 1 1 and
1 0 1

4 0 0
C 1 AC = = 0 4 0 .
0 0 1

Bilinear and quadratic forms. Consider the following expression:

q = a11 x1 y1 +a12 x1 y2 +a13 x1 y3 +a21 x2 y1 +a22 x2 y2 +a23 x2 y3 +a31 x3 y1 +a32 x3 y2 +a33 x3 y3 .

The above expression is called a bilinear form and we can express this form in matrix
notaions as follows:

 a11 a12 a13 y1
q = q(X, Y ) = x1 x2 x3 a21 a22 a23 y2 = X T AY

(13)
a31 a32 a33 y3

x1 y1 a11 a12 a13
where X = x2 , Y = y2 and A = a21 a22 a23 . In the above expression,
x3 y3 a31 a32 a33
since, in each term the degree of x and y are one and it contains two vectors X and Y ,
therefore, it is called a bilinear form.
If we put X = Y in the bilinear form q(X, Y ), i.e., in equation (13), then it reduces
into the following form:

a 11 a 12 a 13 x 1
q = q(X, X) = x1 x2 x3 a21 a22 a23 x2 = X T AX.
 
(14)
a31 a32 a33 x3
Dr. Satish Shukla 32 of 39

In expanded form:
q(X, X) = a11 x21 + a22 x22 + a33 x23 + (a12 + a21 )x1 x2 + (a13 + a31 )x1 x3 + (a23 + a32 )x2 x3 .
The above expression has the second order term of x, and it is called a quadratic form.

Remark 2. The matrix A in the quadratic form q(X, X) can always be taken as
a symmetric matrix by choosing suitable values of entries of A.

Example 25. Find the matrix of the following quadratic form hence writ it into
matrix form:
3x2 + 5y 2 + 3z 2 2yz + 2zx 2xy.

3 1 1
Sol. The matrix of given quadratic form is A = 1 5 1 . Therefore, in
1 1 3
matrix notation, the given quadratic form is:

3 1 1 x
T
 
q(X, X) = X AX = x y z 1 5 1 y .
1 1 3 z
Hermitian form: An expression of the following form:
q = a11 x1 x1 +a12 x1 x2 +a13 x1 x3 +a21 x2 x1 +a22 x2 x2 +a23 x2 x3 +a31 x3 x1 +a32 x3 x2 +a33 x3 x3
or

a11 a12 a13 x1
a21 a22 a23 x2 = X AY = X T AX
 
q = q(X, X) = x1 x2 x3 (15)
a31 a32 a33 x3

is called a Hermitian form. Note that, it can be obtained by replacing X by X and Y

by X in the bilinear form (13).

Classification of the quadratic form and definiteness.

Consider the following quadratic form:
q(X, X) = x21 + (2x2 )2 + x23
where X T = [x1 x2 x3 ] R3 . What we can say about the sign of q(X, X)? Definitely,
q(X, X) > 0 for all x1 , x2 , x3 6= 0, because, q(X, X) is a sum of squares. Again consider
the following form:
q(X, X) = x21 (2x2 )2 x23
where X T = [x1 x2 x3 ] R3 . Now q(X, X) < 0 for all x1 , x2 , x3 6= 0, because, q(X, X)
is the negative sum of squares. Again consider the following form:
q(X, X) = x21 + x22 2x1 x2 + x23 .
where X T = [x1 x2 x3 ] R3 . Obviously, q(X, X) is not positive for all x1 , x2 , x3 6= 0,
but it may be equal to 0 for some x1 , x2 , x3 6= 0, i.e., q(X, X) 0 for all x1 , x2 , x3 6= 0.
Consider the following form:
q(X, X) = x21 x22 2x1 x3 x23 .
Dr. Satish Shukla 33 of 39

where X T = [x1 x2 x3 ] R3 . Obviously, q(X, X) is not negative for all x1 , x2 , x3 6= 0,

but it may be equal to 0 for some x1 , x2 , x3 6= 0, i.e., q(X, X) 0 for all x1 , x2 , x3 6= 0.
Finally, consider the following form:

q(X, X) = x21 x22 + x23 .

It is clear that q(X, X) is sometimes positive, negative, as well as, sometimes it becomes
zero, i.e., this quadratic form has no definite sign.
On the basis of the above discussion, we classify the quadratic forms into the fol-
lowing way: a quadratic form is said to be

1. positive definite, if q(X, X) > 0 for all X 6= 0;

2. positive semi-definite, if q(X, X) 0 for all X 6= 0;

3. negative definite, if q(X, X) < 0 for all X 6= 0;

4. negative semi-definite, if q(X, X) 0 for all X 6= 0;

5. otherwise it is called indefinite.

Reduction of quadratic form into canonical form. To decide that whether a quadratic
form is definite or indefinite, we must reduce it into a form which is sum of squares.
A quadratic form which is sum of squares is called a canonical form. There are two
transformations to reduce a quadratic form into canonical form.
(i) The Congruent Transformation. Suppose A (of order 3 3), the symmetric matrix
of quadratic form. Then we find a nonsigular matrix P such that

P T AP =diag(a1 , a2 , a3 ).

x1 y1
If we use the substitution X = P Y , where X = x2 , Y = y2 , then the
x3 y3
quadratic form reduces into the following form:

q(X, X) = X T AX = (P Y )T A(P Y ) = Y T (P T AP )Y = Y T diag(a1 , a2 , a3 )Y

= a1 y12 + a2 y22 + a3 y32 .

The matrix P is called the transformation matrix. To find P we write

A = IAI. (16)

Now by applying row and column operation on A the it is transformed into a diagonal
matrix, and all the row operations are applied to pre-multiplier I all the column op-
eration on post-multiplier I simultaneously, so that, the equation (16) is reduced into
the form
diag(a1 , a2 , a3 ) = P T AP.
Therefore, the post-multiplier I is now reduced into P .
(ii) The Orthogonal Transformation. If there exists a nonsingular orthogonal matrix
P (i.e., P 1 = P T ) such that

P 1 AP =diag(1 , 2 , 3 ).
Dr. Satish Shukla 34 of 39

Then, the given quadratic form can be reduced into canonical form by the substitution
X = P Y . To find P we find the eigenvalues 1 , 2 , 3 of A, and then the corresponding
eigenvectors C1 , C2 , C3 such that these vectors are orthogonal to each other and are
normalized, i.e., C1 C2 = C1 C3 = C2 C3 = 0 and all these vectors are unit. Then, P is
given by:
P = [C1 C2 C3 ].
Rank, Index and Signature of a quadratic form. The number p of positive terms in
the canonical form is called the index and the number r of total terms is the rank of
the quadratic form. The number s of positive terms minus the number of negative
terms, s = p (r p) = 2p r, is called the signature of the quadratic form.

Remark 3. If p, r and s for a quadratic form are known, then the form is:

1. positive definite, if r = p = n;

2. positive semi-definite, if r = p < n;

3. negative definite, if p = 0, r = n;

4. negative semi-definite, if p = 0, r < n;

5. otherwise it is indefinite.

Example 26. Reduce the following quadratic form into canonical form, hence find
its rank, signature and index and definiteness:

q = 2x21 + 4x1 x2 + 6x1 x3 + 3x22 + x2 x3 6x23 .

2 2 3
Sol. The symmetric matrix of the given form is A = 2 3 1/2 . Now
3 1/2 6

2 2 3 1 0 0 1 0 0
A = IAI = 2 3 1/2 = 0 1 0 A 0
1 0
3 1/2 6 0 0 1 0 0 1
Applying R2 R2 R1 , R3 R3 23 R1 ; and C2 C2 C1 , C3 C3 23 C1

2 0 0 1 0 0 1 1 3/2
0 1 5/2 = 1 1 0 A 0 1 0
0 5/2 21/2 3/2 0 1 0 0 1
Applying R3 R3 + 25 R2 ; and C3 C3 + 25 C2

2 0 0 1 0 0 1 1 4
0 1 0 = 1 1 0 A 0 1 5/2 .
0 0 67/4 4 5/2 1 0 0 1

1 1 4
T
Therefore diag(2, 1, 67/4) = P AP. Here P = 0 1 5/2 , and so, the trans-

0 0 1
formation will be

x1 1 1 4 y1
X = P Y = x2 = 0 1 5/2 y2
x3 0 0 1 y3
Dr. Satish Shukla 35 of 39

x1 = y1 y2 4y3
5
x2 = y2 + y
2 3
x3 = y3 .
Now the given quadratic form will be:

q = X T AX = Y T (P T AP )Y = Y T diag(2, 1, 67/4)Y
67
= 2y12 + y22 y32 .
4
Therefore, the rank of quadratic form r = 3, the index p = 2 and the signature
s = 2p r = 1, and so, the given quadratic form is indefinite.

Example 27. Reduce the following quadratic form into canonical form, hence find
its rank, signature and index and definiteness:

q = 4x2 + 10y 2 + 11z 2 4xy + 12xz 12yz + 4.

4 2 6
Sol. The symmetric matrix of the given form is A = 2 10 6 . Now
6 6 11

4 2 6 1 0 0 1 0 0
A = IAI = 2 10 6 = 0 1 0 A 0 1 0
6 6 11 0 0 1 0 0 1

Applying R2 R2 + 12 R1 , R3 R3 32 R1 ; and C2 C2 + 21 C1 , C3 C3 32 C1

4 0 0 1 0 0 1 1/2 3/2
0 9 3 = 1/2 1 0 A 0 1 0
0 3 2 3/2 0 1 0 0 1

Applying R3 R3 + 31 R2 ; and C3 C3 + 31 C2

4 0 0 1 0 0 1 1/2 4/3
0 9 0 = 1/2 1 0 A 0 1 1/3 .
0 0 1 4/3 1/3 1 0 0 1

1 1/2 4/3
Therefore, diag(4, 9, 1) = P T AP. Here P = 0 1 1/3 , and so, the transfor-
0 0 1
mation will be

x 1 1/2 4/3 u
X = P Y = y = 0 1 1/3 v
z 0 0 1 w
1 3
x = u + 2
v 4
w
1
y = v + 3
w
z = w.
Now the given quadratic form will be:

q = X T AX = Y T (P T AP )Y = Y T diag(4, 9, 1)Y
= 4u2 + 9v 2 + w2 .
Dr. Satish Shukla 36 of 39

Therefore, the rank of quadratic form r = 3, the index p = 3 and the signature
s = 2p r = 3, and so, r = p = n = 3. Therefore, the given quadratic form is positive
definite.
Example 28. Reduce the following quadratic form into canonical form, hence find
its rank, signature and index and definiteness:

q = 2xy + 2yz + 2zx.

0 1 1
Sol. The symmetric matrix of the given form is A = 1 0 1 . Now
1 1 0

0 1 1 1 0 0 1 0 0
A = IAI = 1 0 1 = 0
1 0 A 0 1
0
1 1 0 0 0 1 0 0 1

Applying R1 R1 + R2 ; and C1 C1 + C2

2 1 2 1 1 0 1 0 0
1 0 1 = 0 1 0 A 1 1 0
2 1 0 0 0 1 0 0 1

Applying R2 R2 21 R1 , R3 R3 R1 ; and C2 C2 21 C1 , C3 C3 C1

2 0 0 1 1 0 1 1/2 1
0 1/2 0 = 1/2 1/2 0 A 1 1/2 1
0 0 2 1 1 1 0 0 1

1 1/2 1
Therefore, diag(1, 1/2, 2) = P T AP. Here P = 1 1/2 1 , and so, the trans-
0 0 1
formation will be

x 1 1/2 1 u
X = P Y = y = 1 1/2 1 v
z 0 0 1 w
1
x = u 2
v w
1
y = u + 2
v w
z = w.
Now the given quadratic form will be:

q = X T AX = Y T (P T AP )Y = Y T diag(1, 1/2, 2)Y

1
= u2 v 2 2w2 .
2
Therefore, the rank of quadratic form r = 3, the index p = 1 and the signature
s = 2p r = 1. Therefore, the given quadratic form is indefinite.

Example 29. Examine the definiteness of the following complex quadratic form:

q = x1 x1 + 5x2 x2 + 13x3 x3 + (1 + 2i)x2 x1 + (1 2i)x1 x2 + (2 3i)x3 x1

+(2 + 3i)x1 x3 (4 + 2i)x3 x2 (4 2i)x2 x3 .
Dr. Satish Shukla 37 of 39

1 1 + 2i 2 3i
Sol. The symmetric matrix of the given form is A = 1 2i 5 4 2i .
2 + 3i 4 + 2i 13
Now

1 1 + 2i 2 3i 1 0 0 1 0 0
A = IAI = 1 2i 5 4 2i = 0 1 0 A 0 1 0
2 + 3i 4 + 2i 13 0 0 1 0 0 1

Applying R2 R2 (1 2i)R1 , R3 R3 (2 + 3i)R1 ; and C2 C2 (1 + 2i)C1 , C3

C3 (2 3i)C1

1 0 0 1 0 0 1 2i 1 2 + 3i
0 0 5i = 2i 1 1 0 A 0 1 0
0 5i 0 2 3i 0 1 0 0 1

Applying R2 R2 + iR3 ; and C2 C2 iC3

1 0 0 1 0 0 1 2 2 + 3i
0 10 5i = 2 1 i A 0 1 0
0 5i 0 2 3i 0 1 0 i 1

Applying R3 R3 + 2i R2 ; and C3 C3 2i C2

1 0 0 1 0 0 1 2 2 + 2i
0 10 0 = 2 1 i A 0 1 i/2 .
0 0 5/2 2 2i i/2 1/2 0 i 1/2

1 2 2 + 2i
Therefore diag(1, 10, 5/2) = P AP. Here P = 0 1 i/2 , and so, the
0 i 1/2
transformation will be

x1 1 2 2 + 2i y1
X = P Y = x2 = 0 1 i/2 y2
x3 0 i 1/2 y3

x1 = y1 + 2y2 (2 2i)y3
i
x2 = y2 2 3
y
1
x3 = iy2 + 2 3
y.
Now the given quadratic form will be:

q = X AX = Y (P AP )Y = Y diag(1, 10, 5/2)Y

5
= y1 y1 + 10y2 y2 y3 y3 .
2
Therefore, the given quadratic form is indefinite.

Example 30. Reduce the quadratic form 2x1 x2 +2x1 x3 2x2 x3 to a canonical form
by an orthogonal reduction and discuss its nature. Also find the model matrix.

0 1 1
Sol. The matrix of quadratic form is A = 1 0 1 . The eigenvalues are
1 1 0
= 1, 1, 2.
Dr. Satish Shukla 38 of 39

(II) For = 1. For = 1, the eigenvalue problem will be (A 1 I)X = 0, i.e.,

1 1 1 x1
1 1 1 x2 = 0.
1 1 1 x3
Applying R2 R2 + R1 , R3 R3 + R1

1 1 1 x1
0 0 0 x2 = 0.
0 0 0 x3
It is the echelon form. Clearly, the rank of matrix is 1 < 3 =no. of variables. Therefore,
3 1 = 2 variable will be independent. Now from the echelon form we have only one
equation: x1 + x2 + x3 = 0; and two variables are independent.
1
(i) Choosing x3 = 1, x2 = 0 we get x1 = 1, so the eigen vector will be C1 = 0 .

1
c1
(ii) Suppose, the second eigenvector corresponding to = 1 is C2 = c2 , then C2

c3
must satisfy the equation x1 + x2 + x3 = 0, as well as, it should be orthogonal to C1 ,
i.e., C1 C2 = 0. therefore, we have the following equations:

c1 + c2 + c3 = 0
c1 + c3 = 0.

Choose c3 =1, then c2 = 2 we get c1 = 1. Therefore, second eigenvector will be:

1
C2 = 2 .

1
(II) For = 2. For = 2, the eigenvalue problem will be (A + 2 I)X = 0, i.e.,

2 1 1 x1
1 2 1 x2 = 0.
1 1 2 x3
Applying R1 R2
1 2 1 x1
2 1 1 x2 = 0.
1 1 2 x3
Applying R2 R2 2R1 , R3 R3 R1

1 2 1 x1
0 3 3 x2 = 0.
0 3 3 x3
Applying R3 R3 R2

1 2 1 x1
0 3 3 x2 = 0.
0 0 0 x3
It is the echelon form. Clearly, the rank of matrix is 2 < 3 =no. of variables. Therefore,
3 2 = 1 variable will be independent. Now from the echelon form we have the
Dr. Satish Shukla 39 of 39

equations: x1 + 2x2 x3 = 0; and 3x2 + 3x3 = 0; and one variable

is independent
1
choosing x3 = 1 we get x2 = 1, x1 = 1. Thus, C3 = 1 is the corresponding
1
eigenvector.
1/ 2 1/6
The normalized eigenvectors will be P1 = 0 , P2 = 2/ 6 , P3 =
1/ 2 1/ 6

1/ 3
1/ 3 , and the model matrix will be:

1/ 3

1/ 2 1/6 1/ 3
P = [P1 P2 P3 ] = 0 2/ 6 1/3 .
1/ 2 1/ 6 1/ 3
The orthogonal reduction is:
q = Y T (P T AP )Y = Y T diag (1, 1, 2) Y = y12 + y22 2y32 .
Therefore, the rank of quadratic form r = 3, the index p = 2 and the signature
s = 2p r = 1. Therefore, the given quadratic form is indefinite.

Exercise (Assignment)

(Q.1) Prove that every squre matrix A can be written as the sum of a Hermitian matrix
and skew-Hermitian matrix.

(Hint. Write A = A+A 2
+ AA
2
).
 
2 1
(Q.2) Show that the matrix is not diagonalizable.
0 2

1 1 3
(Q.3) Diagonalize the matrix A = 1 5 1 .
3 1 1
(Ans. = 6, 3, 2.)

2 0 1
(Q.4) Diagonalize the matrix A = 0 2 0 .
1 0 2
(Ans. = 1, 2, 3.)
(Q.5) Reduce the quadratic form into the canonical form and specify the matrix trans-
formation:
3x2 + 5y 2 + 3z 2 2yz + 2zx 2xy.
(Q.6) Reduce the quadratic form into the canonical form, specify the matrix transfor-
mation and find its rank, signature and index:
x21 + 3x22 + 3x23 2x2 x3 .
(Q.7) Reduce the quadratic form into the canonical form by an orthogonal transfor-
mation, find the model matrix and discuss its nature:
3x21 + 3x22 + 3x23 + 2x1 x2 + 2x1 x3 2x2 x3 .