Vector Spaces and Matrices
Uploaded by
shreekant vaderVector Spaces and Matrices
Uploaded by
shreekant vaderGraduate Course
B.A. (PROGRAMME) 1 YEAR
ALGEBRA AND CALCULUS
SM – 1 PART – A: (ALGEBRA)
VECTOR SPACES AND MATRICES
CONTENTS
Lesson 1 : Vector Spaces
Lesson 2 : Matrices : Basic Concepts
Lesson 3 : Elementary Operations on a Matrix and Inverse of a Matrix
Lesson 4 : Rank of a Matrix
Lesson 5 : Systems of Linear Equations
Lesson 6 : The Characteristic Equation of a Matrix
Editor:
Dr. S.K. Verma
SCHOOL OF OPEN LEARNING
UNIVERSITY OF DELHI
5, Cavalry Lane, Delhi-110007
1
Session 2012-2013 (1500 Copies)
© School of Open Learning
Published By: Executive Director, School of Open Learning, 5, Cavalary Lane, Delhi-110007
Laser Composing By : M/s Computek System (2012-13)
2
LESSON 1
VECTOR SPACES
1.1 Introduction
You are already familier with several algebric structures such as groups, rings, integral domains and
fields. In this lesson we shall tell you about another equally important algebric structures, namely, a
vector space.
Let V be a non-emply set and let F be a field. Let us agree to call elements of V vectors and
elements of F scalars.
A maping from V × V to V will be called addition in V and a mapping from F × V to V will be
called multiplication by a scalar multiplication, V is said to a vector space over F if addition and scalar
multiplication satisfy certain properties. Of course, these conditions are to be chosen in such a manner
that the resulting algebric structure is rich enough to be useful. Before presenting the definition of a vector
space, let us note that addition in V is denoted by the symbol ‘+’, and scalar multiplication is denoted by
juxtaposition, i.e., if x ∈ V, y ∈ V, and α ∈ F, the x + y denoted the sum of x and y, and αx denotes the
scalar multiople of x by α.
Definition 1. A non-empty set V is said to be a vector space over a field F with respect to addition
and scalar multiplication if the following properties hold.
V 1 Addition in V is associative, i.e.,
x + (y + z) = (x + y) + z, for all x, y, z, ∈ V
V2 There exists of natural element for addition in V, i.e., there exists an element 0 Î V such that
x + 0 = 0 + x = x, for all x ∈ V
V3 Every element of V possesses a negative (or addition inverse), i.e., for each x ∈ V, there
exists an element y ∈ V such that
x + y = y + x = 0.
V4 Addition in V is commulative, i.e., for all elements x, y ∈ V,
x+y= y+x
V5 Associtiavity of scalar multiplication, i.e.,
i.e., α (βx) = (α β) x, for all α, β, ∈ F and x ∈ V
V6 Property of 1. For all x∈V,
1x = x, where 1 is the multiplicative identity of F.
V7 Distributivity properties for all α, β ∈ F and x, y ∈ V.
(α + β)x = αx + βx
α(x + y) = αx + αy
Remarks 1. The first of the two distributivity properties stated in V 7 above is generally called
distributivity of scalar multiplication over addition in F, and the second of the two distributibity
properties is called distributivity of scalar multiplication over addition in V.
3
2. We generally refer to properties V 1 – V 7 above by saying that (V, +) is a vector space over F. If
the underlying field F is fixed, we simply say that (V, +,) is a vector space, and do not make an explicit
reference to F.
In case, the two vector space compositions are known, we denote a vectors space over a field’ F by
the symbol V(F). If there is no chance of confusion about the underlying field, then we simply talk of ‘the
vector space V’.
3. You might have observed that the axions V 1 to V 4 simply assert the V is an abelian group for
the composition ‘+’. In view of the we can re-state the definition of a vector space as follows:
2. Definnition and Exampls of a Vector Space
Defintion 2. A triple (V, +,) is said to be a vector space over a field F if (V, +) is an abelian group,
and the following properties are satisfied :
α(βx) = (αβ)x, ∀α, β ∈ F and ∀ x ∈ V
1x = x, ∀x, β ∈ V, where 1 is the multiplicative identity of F
(α + β)x = αx + βx, ∀α, β ∈ F, and ∀x, y ∈ V
α (x + y) = αx + αy, ∀α ∈ F and ∀ x, y ∈ V
We shall now consider some examples of vector spaces.
Example 1. Let R be the set of number (R, +) is vector space over R. The addition is addition in R
and scalar multiplication is simply multiplication of real numbers.
It is easy to verify that all the vector space axioms are verified. In fact, V1-4 are satisfied because R
is an abelian group with respect to addition, V5 is nothing but the associative property of multipication,
V6 is the property of the multiplicative identity in (R, +,) and the properties listed in V7 are nothing but
the distributivity of multiplication over addition.
Example 2. (C, +, ) is a vector space over C
Example 3. (Q, +, ) is a vector space over Q.
Example 4. Let F be any field. F is a vector space over itself for the usual compositions of addition
and multipication (to be called scalar multiplication) in F.
Example 5. C is a vector space over R, and R is a vector space over Q.
Example 6. R is not a vector space over C. Observe that if α ∈ C and x ∈ R, the αx is not in R.
Therefore the multiplication composition in R fails to give rise to the scalar multiplication composition.
The examples considered above are in a way re-labelling of the field properties C, R or Q. We shall
now consider some examples of a different type.
Example 7. Let V be the set of all vectors in a plane. You know that addition of two vectors is a
vector, and that V is a group with respect to sum of vectors. Let us take addition of vectors as the first
compostion for the purpose of our example. Also, we know that if d be any vector and k be any real
numbers, then k d is a vector. Let us take R as the underlying field and multiplication of vector by a
scalar as the second vector space composition. It is easy to see that V is vector space over R for these two
compositons.
Example 8. Let R3 be the set
{( x1 , x2 , x3 ) : x1 , x2, x3 ∈ R}
and the addition and scalar multiplication R3 be defined as follows:
4
If x ∈ R3 and y ∈ R3 let
x + y = ( x1 + y1 , x2 + y2 , x3 + y3 ).
Also if x ∈ R3 and c ∈ R, let
cx = (cx1 , cx2 , cx3 ).
It can be seen that R3 is a vector space over R for the two compositions—addition and scalar
multiplication, as defined above.
We may note before passing on to the next example that the vector space being considered here is
nothing but the space of the vectors (in space) with addition and scalar multiplication as the composition.
This example is of special interest because it was in fact motivation for the present terminology of vector
spaces.
The next three examples are a little abstract in nature but are quite important.
Example 9. Let Rn be the set of ordered n-tuples of real numbers, so that a typical element of Rn is
We shall denote this element by x (printed as a bold-face better), and write
x = ( x1 , x2 , ..., xn )
Let us take R to be the underlying field and define addition and scalar multiplication in Rn by setting
x + y = ( x1 + y1 , x2 + y2 , ..., xn + yn ),
where x = ( x1 , x2 , ... xn ), y = ( y1 , y2 , ... yn )
and αx = ( αx1, αx2 , ... αxn ) , ∀ α ∈ R
Let us first of all see that addition scalar multiplication as defined above are meaningful in the sense
that they define the two compositions that we need for making Rn vector space.
Since x1 , x2 , x3 , xn and y1 , y2 , ... yn are all real numbers, therefore x1 + y1 , x2 y2 , ..., xn + yn are all
real numbers and therefore ( x1 + y1 , x2 + y2 , ..., xn + yn ) is an ordered n-type of real numbers and
consequently it is in Rn. Again, since α is a real number and ( x1, x2 , ..., xn ) are also real numbers,
therefore αx1, αx2 , ..., αxn are also real numbers, and consequently (αx1, αx2 , ..., αxn ) is an n-type of real
numbers and so is Rn.
Having defined addition and scalar multiplication in Rn, let us see in some detail that all the
properties needed for Rn to be a vector space are actually satisfied.
1. Let x = ( x1, x2 , ..., xn ), y = ( y1, y2 , ..., yn ), z = ( z1, z2 , ..., zn ) be any three elements of Rn.
Then (x + y) + z = ( x1 + y1 , x2 + y2 , ..., xn + yn ) + ( z1 , z2 , K, zn )
= [( x1 + y1 ) + z1 , ( x2 + y2 ) + z2 , ..., ( xn + yn ) + zn ]
= [ x1 + ( y1 + z1 ), x2 + ( y2 + z2 ), K, xn + ( yn + zn )]
= ( x1 , x2 , x3 , K, xn ) + ( y1 + z1 , y2 + z2 , ... yn + zn )
= x + (y + z).
2. Let o = (0, 0, ..., 0). so that o ∈ Rn and x + 0 = o + x = x for all x ∈ Rn
3. Let x = ( x1 , x2 , ..., xn ) be any element of Rn. If y = (–x1, –xn, ..., –xn) then y ∈ Rn and y is an
element of Rn such that
x+y= y+x=0
4. If x = ( x1 , x2 , K, xn ), y = ( x1 , y2 , ..., yn ) are any two elements of Rn then
5
x + y = ( x1 , x2 , K, xn ), + ( y1 , y2 , ..., yn )
= ( x1 + y1 , x2 + y2 , K, xn + yn )
= ( y1 + x1 , y2 + x2 , K, yn + xn )
= ( y1 , y2 , ... yn ) + ( x1 , x2 , ..., xn )
= y + x.
5. If x = ( x1 , x2 , ... xn ), be any element of Rn and p, q be any real numbers, then
(pq)x = ( pq) ( x1 , x2 , ... xn )
= [( pq) x1 , ( pq) x2 , ..., ( pq) xn ]
= [ p(qx1 ), p(qx2 ), ... p(qxn )]
= p (qx1 , qx2 , ... qxn )
= p(qx)
6. If x = (x1, ..., xn) be any element of Rn and p, q be any real numbers, then
(p + q) x = [( p + q) x1 , ( p + q) x2 , ..., ( p + q) xn ]
= ( px1 , px2 , ..., pxn ) + (qx1 , qx2 , ..., pxn )
= px + qx
7. If x = ( x1 , x2 , ..., xn ) and y = ( y1 , y2 , ..., yn ) be any two elements of Rn, and p be any real
number, then
p(x + y) = p(x1 + y1, x2 + y2, ..., xn + yn)
= [p(x1 + y1), p(x2 + x2), ..., p(xn + yn)]
= (px1, px2, ..., pxn) + (py1, py2, ..., pyn)
= p(x1, x2, ..., xn) + p(y1, y2, ..., yn)
= px + py.
8. If x = ( x1 , x2 , ..., xn ) be any element of Rn, then
1x = 1( x1 , x2 , ..., xn )
= (1 ⋅ x1 ,1 ⋅ x2 , K,1 ⋅ xn )
= ( x1 , x2 , ... xn )
= x.
From 1-8 above we find that Rn is a vector space over R with co-ordinatewise addition and co-
ordinatewise scalar multiplication as the two vector space compositions.
The use of the word co-otherwise, is due to the fact that if x = ( x1 , x2 , ..., xn ) be any element of Rn
then x1 , x2 , ..., xn are called the co-ordinates of x, and while defining x + y, we add the corresponding co-
ordinates of x and y, and while defining cx we multiply the co-ordinates of x by c.
We may note that the space in example 9 is only a spacing case of the example 9 for n = 3.
Example 10. The set Cn of all ordered n-tuples of complex number is a vector space over C for co-
ordinatewise addition and co-ordinatewise scalar multiplication as the two vector space compositions.
6
Example 11. Let F be any field. The set Fn of all ordered n-tuples of elements of F is a vector space
over F with co-ordinatewise addition and co-ordinatewise scalar multiplication as the two vector space
compositions.
Example 12. Let Mmn is a vector space over C with respect to matrices over C. Mmn is a vector space
over C with respect to matrix addition and multiplication of a matrix by a scalar, for
1. The sum of two m × n matices with complex entries is an m × n matrix with complex entries.
2. Addition of matrices is associative.
3. The m × n zero-matrix is a natural element for addtion.
4. It A be an m × n matrix with complex entries, then –A is also an m × n with complex entries
such that (–A) + A + (–A) = 0
5. Addition of matrices is commutative.
6. If A ∈ Mmn and c be any complex number, then –A is also m × n matrix with complex entries
and so cA ∈ Mmn
7. If p, q be any complex numbers and A, B be any two m × n matrices with complex entries, then
p(q A) = (pq) A,
(p + q) A = pA + qA,
p(A + B) = pA + pB,
1 A = A, for all A ∈ Mmn
a b
Example 13. The set S of all matrices of the form , where a, b are any complex numbers, is
−b a
a vector space over C with respect to matrix addition and multiplication by a scalar for the following
reasons:
1. If A, B, ∈ S, the A + B ∈ S.
a b c d
A= c
For, if ,
−b a − d
a+c b+d p q
A+B= = , where p = a + c, q = b + d .
a + c −q p
then
− (b + d )
2. Matrix addition is associative.
o o
3. The matrix O = ∈ S , and A + O = O + A = A, for all A ∈ S.
o o
p q − p −q
4. If A = ∈ S , then the matrix B = ∈ S , and is such that A + B = B + A = 0.
−q p q −p
5. Addition f matrices is commulative.
a b
6. If c be any complex number, and A = ∈ S, then cA is the matix given by
−b a
ca cb
cA = .
−cb ca
It is obvious that cA ∈ S.
7
a b
7. If p, q be any complex numbers, and A, B be any two matrices in S, say A = ,
−b a
c d
B= , then
−d c
a b ( pq) a ( pq )b p (qA) p(qb)
(i) (pq)A = pq
− = ( pq)(−b) ( pq )a = p (−qb) p (qa)
= p (qA).
b a
(ii) (p + q) A = pA + qA
(iii) p(A + B) = pA + pB
(iv) 1 A = A.
x y
Example 14. The set of all matrices of the form , where x, y, z, ∈ C, is a vector space over
z o
C with respect to matrix addition and multiplication of a matrix by a scalar.
The verification of the vector space axioms is straight forward.
Example 15. The set s of all hermitian matrices of order n is a vector space over R with respect to
matrix addition and multiplication of matrix by a scalar.
To verify that (S, +, ) is a vector space over C, we proceed as follows:
1. Let A, B be two hermitian matrices of order n. Then A + B is a matrix of order n. It is hermitian
because (A + B)θ = Aθ + Bθ = A + B, since Aθ = A, Bθ = B.
2. Addition of matrices is associative.
3. The n-rowed, zero matrix O is a matrix is a hermitian matrix such that A + 0 = 0 + A = A
4. It A ∈ S, so that At = A, then (–A)t = –At = –A, so that –4A ∈ S, and A + (–A) = (–A) + A = 0.
5. Matrix addition is commutative
6. If C ∈ R, and A ∈ S, then (CA)t = CAt = CA, so that A ∈ S.
7. If p, q, ∈ R and and A B ∈ S, then
(i) (p, q)A = p(qA)
(ii) (p + q)A = pA + qA
(iii) p(A + B) = pA + pB
(iv) 1A= A
In view of the above properties it follows that (S, +, ) is a vector space over R.
Example 16. The set S of all real symmetric matrices of order n is a vector space over R with
respect to matrix addition and multiplication of a matrix by a scalar.
In order to convince ourselves that (S, +, ) is a vector space over R, let us note the following:
1. A, B be n- rowed real symmetric matrices, then At = A, Bt = A. It follows that
(A + B)t = At + Bt = A + B
2. Addition of matrices is associative, so that for all A, B, C ∈ S,
A + (B + C) = (A + B) + C.
3. The n-rowed zero matrices o is a real symmetric matrix, and therefore it is in S.
Also, A+0= 0+A=A
for all A in S.
4. If A ∈ S so that At = A, then (–A)t = –At = –A, so that—17 A ∈ S. Also, A + (–A) = (–A) + A = 0.
8
5. Matrix addition is commulatative.
6. If C ∈ R and A ∈ S, then (cA)t = cAt = cA, so that A ∈ S.
7. If p, q ∈ R, and A, B ∈ S, then
(i) (p, q) A = p(qA)
(ii) (p + q)A = pA + qA
(iii) P(A + B) = pA + pB
(iv) 1A= A
3. Some Direct Consequents of Vector Space Axioms
We shall now state and prove some elementary consequences of the vector space axioms. These will
help us in dealing with vector in a convenient way in many situations.
Theorem. Let V be a vector over a field F. Then for all a ∈ F and x ∈ V,
(i) α0 = 0,
(ii) 0x = 0,
(iii) (–α) x = –ax
(iv) (–α) (–x) = αx
(v) αx = 0 if either α = 0 or x = 0.
Proof.
(i) α0 = α(0 + 0), by the property of 0 ∈ V
= α0 + α0, by distributivity of scalar multiplication over addition in V
Since α0 + α0 = α0 = α0 + 0, by the property of 0 in V by cancellation law in (V, +), it
follows that α0 = 0
(ii) 0 x = (0 + 0) x, by the property of 0 in F
= 0x + 0x, by distributivity of scalar multiplication over addition in F.
Since 0x + 0x = 0x + 0, by, the property of 0 in V, therefore by cancellation law in (V, +),
it follows thats
0x = 0
(iii) 0 = a + (–α) ⇒ 0x = (α + (–α))x
⇒ 0 = αx + (–α) x, since ox = 0.
Now αx, (–α)x are two elements of V such that ax + (–α)x = 0, therefore (–α) x is the negative of.
αx, i.e, (–α) x = –(αx).
(iv) Now –α(0) = –α (x + (–x))
⇒ 0 = –αx + ((–α) (–x)), since –α(0) = 0, by (i) above
⇒ (–α) (–x) is the negative of –αx in V
⇒ (–α) (–x) = –(–αx) –αx,
because –αx ∈ V and therefore negative of –αx in V is simply αx.
(v) Let us first suppose that ax = 0. If α = 0, we are done. If α ≠ 0, then α–1 ∈ F, because α ∈ F
and F is a field
Therefore 0 = α–1 0
9
= α–1 (αx), because αx = 0 by hypothesis
= (α–1 α)x, by associativity of scalar multiplication
= 1x
= x.
Thus αx = 0 ⇒ either α = 0 or x = 0. Conversely, let us assume taht either α = 0 or x = 0
In case α = 0, by (ii) above αx = 0x = 0.
In case x = 0, by (i) above αx = α0 = 0.
Thus in both cases we find that αx = 0.
It is now obvious that αx = 0 if either x = 0 or α = 0.
Exercise
1. Show that the set
C2 = {(x1, x2) : x1 ∈ C, x2 ∈ C}
is a vector space over C with respect to co-ordinateswise addition and scalar multiplication.
2. Show that the set of all 2 × 2 matrices over C is a vector space over C with respect to matrix
addition and multiplication of a matrix by a scalar.
3. Let V = {(a1, a2, a3, a4) : a1, a2, a3, a4 are integers}
Is V a vector space over R with respect to co-ordinatewise addition and scalar multiplication?
Justify your answer.
4. Let V = {(x1, x2, x3) : (x1, x2, x3, are complex numbers, and x1, x2, = 0}
Is V a vector space over C with respect to co-ordinateswise addition and scalar multiplication?
Justify your answer.
x o
5. Show that the set of all matrices of the form , where y ∈ C is a vector space over C with
o y
respect to matrix addition and multiplication of a matrix by a scalar.
a b
6. Show that the set of all matrices of the form , wher a, b ∈ C is vector space over C
−b a
with respect to matrix addition an multiplication of matrix by a scalar.
10
LESSON 2
MATRICES
(BASIC CONCEPTS)
1. Introduction
You are already familier with addition and multiplication of matrices. We shall now talk about some
important types of matrices such as symmetric and skew-symmetric matrices, hermitian and skew-
hermitian matrices etc., elementary operations on a matrix inverse of a matrix, rank of a matrix, and
characteristic equation of a matrix. In the end we shall apply some of these concepts to solutions of
systems of lineare equations. However, before we do so, we shall briefly recapitalate the main facts about
addition and multiplication of matrices.
2. Defintion of a Matrix
Let S be any set. A set of mn elements arranged in a rectangular array of m rows and n column as
a11 a12 K a1n
a21 a22 K a2 n
K K K K
am1 am 2 K amn
is called an m × n (“m by n”) matrix over S. A matrix may be represented by the symbols || aij ||, [aij], [aij]
or by a single letter such as A. The aij’s in a matrix are called the element of the matrix. The indices is and
j of an element indicate respectively the row and the column in which the elements aij is located.
Since we shall be dealing only with matrices over the set of complex number therefore, we shall use
the word “matrix” so as to mean “matrix over C” throughtout, unless we state to the contrary.
Thes 1 × n matrics are called row vectors and the m × l matrices are called column vectors. The m ×
n matrix whose elements are 0 is called the null matrix (or zero matrix) of the type m × n. It is usually
denoted by Om × n or simply by O if there is no possibility of confusion.
If the number of rows and the number of columns of a matrix are equal (say each equal to n) the
matrix is said to be a square matrix of order n or an n-row square matrix. The elements a11 a22, ... amn of a
square matrix A are said to constitute the main doagonal of A. A square matrix in which all the off
diagonal elements are zero is called a diagonal matrix. Thus an n-rowed square matrix [aij] is a diagonal
matrix if aij = 0 whenever i + j. An n-rowed diagonal matrix is often written as
dia. [a11, a22, ..., amn]
A diagonal matrix in whic all the diagronal elements are equal is called a scalar matrix. In other
words, an n-rowed square matrix [aij] is a scalar matrix if for some number k.
aij = {
k , when i = j ,
0, when i ≠ j.
A scalar matrix in which each diagonal element is unity, is called a unit matrix. Thus, an n-rowed
square matrix [aij] is called a unit matrix if
aij = {
1, whenever i = j,
0, whenever i ≠ j.
11
The n-rowed unit matrix is usually denoted by In (or simply by I if there is no possibility of
confustion).
The matrix of elements which remain after deleting any number of rows and columns of a matrix A
is called a sub matrix of A.
Illustrations :
0 0 0 0
1. 0 0 0 0 is the 3 × 4 null matrix.
0 0 0 0
3 1 2
2. 5 4 7 is a 3-rowed square matrix. 3, 4, 8 constitute the main diagonal of this matrix.
−1 2 8
1 0 0
3. 0 7 0 is a 3-rowed diagonal matrix.
0 0 −2
1 0 0
4. 0 7 0 is a 3-rowed scalar matrix.
0 0 −2
1 0 0
5. 0 1 0 is the 3-rowed unit matrix. We denote it by I .
0 0 1 3
1 8 7
3 4
6. The matrix is submatrix of −2 3 4 because it can be obtained from the latter by
6 −5 1 6 −5
deleting the first row and the first columne.
3. Equality of Matrices
Two matrices A = [aij] and B = [bij] are said to be equal if (i) they are comparable, i.e., the number
of rows in B is the same as the number or rows A, and the number of columns in B is the same as the
number of columns in A; (ii) aij = bij = for every pair of subscripts i and j. Thus, for example, the matrices
3 7 1 5 4 −1 2 3 4 3 6
8 9 and 3 6 2 are not comparable; the matrices 3 1 0 and 1 8 9 are comparable but
2 4 7 4 4 7
not equal; the matrices and are equal.
6 3 −1 2.3 9 −1
From the definition of equality of matrices, it can be easily verified that if A, B, and C be any
matrices, then
(i) A = A (reflexivity)
(ii) A = B ⇒ B = A (symmetry)
(iii) if A = B and B = C, the A = (transitivity)
The above statements (i)—(iii) can be summed up by saying that the relation of equality in the set of
all matrices is an equivlance relation.
4. Addition of Matrices
12
If A = [aij], and B = [bij] be two matrices of the same type, say m × n, their sum is the m × n matrix C
= [cij], where cij = aij + bij for every pair of subscripts i and j. In other words,
a11 a12 K a1n b11 b12 K b1n
a a22 K a2 n b b22 K b2 n
If A = 21 B = 21
K K K K K K K K
a m1 am 2 K amn bm1 bm 2 K bmn
a11 + b11 a12 + b12 K a1n + a1n
then A + B = a21 + b21 a22 + b22 K a2 n + b2 n
am1 + am1 am 2 + bm 2 K amn + bmn
−2 1 3 4 2 5
Illustrations. If A=
− and B = 3 0 6
4 2 1
−2 + 4 1 + 2 3 + 5 2 3 8
then A+B= =
4 + 3 2 + 0 −1 + 6 7 2 5
Properties of matrix addition
Addtion of matrices, has the following properties:
(i) Addition of matrices is associative. That is, if A, B, and C be matrices of the same type, then A
+ (B + C) = (A + B) + C
(ii) Addition of matrices is commutative. That is, if A and B be matrics of the same type, then A +
B = B + A.
(iii) Property of zero matrix. If A be an m × n matrix and 0 denotes the m × n zero matrix, then A +
0 = 0 + A = A.
(iv) Negative of a matrix. If a be an m × n matrix, there exists an m × n matrix B, called the
negative of the matrix A, such that A + B = B + A = 0.
5. Multplication of a Matrix by a Solar
If A = [aij] be an m × n matrix, and k be any complex number, then kA is defined to be the m × b
matrix whose (i, j)th the elements is k aij. The matrix kA is called the scalar multiple of A by k. The
following properties of scalar multiplication are worthnothing :
(i) If A and B are comparable matrices, and k is any complex number, then
k(A + B) = kA + kB.
(ii) If A be any matrix, and k and l be any two complex numbers, then
(k + 1) A = kA + lA.
(iii) If A be any matrix, and k and l be any two complex numbers, then
k(lA) = (kl) A.
(iv) For every matrix A,
1 A =A
6. Multiplication of Matrices
Definition 1. Let A = [aij] and B = [bij] be m × n and n × p matrices respectively. The m × p matrix
[cij], where
13
n
cij = aij, bij + ai2 + b2j + ain bnj = ∑ aij bkj ,
k =1
is called the product of the matrices A and B and is denoted by AB.
The above definition expresses two facts:
(i) We can talk of the product AB if two matrices of and only if the number of columns of A is
equal to the number of rows of B. In the case this condition is satisfied, we say that A and B are
conformable to multiplication.
(ii) If A and B are conformable to multiplication, the (i, j)th element of the matrix AB is obtained by
multiplying the elements of the ith row of A by the corresponding elements of the jth column of
B and adding the products. The sum so obtained is the desired (ij)th elements of AB.
Properties of matrix multiplication
The following are same of the important properties of matrix mulplication :
(i) Matrix multiplication is associative. That is, if A, B, and C be of suitable sizes for the products
A(BC) and (A + B) C to exist, then A(BC) = (AB) C.
(ii) Matrix multiplication is not commutative. That is, given two matrices A and B, AB = BA is not
always true. It is important to note here that for pair of matrices A and B several different
possibilities arise.
(a) Neither of the products AB and BA exits;
(b) only one of the products AB and BA exist and the other one does not exist;
(c) both AB as well as BA exist but they are of different type;
(d) both AB as well as BA exist and are of the same type, but are not equal
(e) AB = BA.
All the above possibilities do exist for certain pairs of matrices.
The important thing to note is that the phase ‘matrix multiplication is not cummutative’ means
that AB is not always equal to BA. It does not exclude the possibility of AB and BA being
equal in some cases.
(iii) Multiplication of matrices is distributive with respect to addition, i.e.,
A(B + C) = AB + BC
and (B + C)D = BD + CD
where A, B, C and D are of the suitable sizes for the above relations to be meaningful.
(iv) Multiplication by the unit matrix. If A be any m × n matrix then Im A = A = A In
7. Positive Integral Powers of Sources Matrix
If A be an n-rowed square matrix, and n be a positive integer, then An is defined by setting A1 = A,
Ak + 1 = AkA. By the principle of finite induction this defines An for all positive integer An.
If A be an n-rowed square matrix and p and q be positive integers, it can be easily shown that
Ap · Aq = Ap + q, (Ap)q = Apq.
8. Transpose of a Matrix
Consider the matrices
14
3 6
3 1 −4
A= , B = 1 0 .
6 0 7 −4 7
The matrix A is a 2 × 3 matrix, and the matrix B is a 3 × 2 matrix. Also, the first column of B is the
same as the first row of A, and the second columne of B is the same as the second row of A. In other
words, B is the matrix obtained from A by writing the row of A as columns. We say that the matrix B is
the transpose of A.
Defintion 2. If A = [aij] be an m × n matrix, then the n × m matrix B = [bij], such that bij = aji is
called the transpose of A and is deonted by At.
From the above defintion we find that
(i) the transpose of an m × n matrix is an n × m matrix ;
(ii) the (i, j)th element of At is the (j, i)th element of A.
3 −1 2 −1 3 5
Example 1. Let A = and B = .
0 4 7 6 2 1
Compute At, (At)t, Bt, (A + B)t, At + Bt, (3A)t, 3At.
3 0
3 −1 2
Solution. At = −1 4 (At)t = 0 4 7
2 7
−1 6
Bt = 3 2 .
5 1
2 6 2 6
2 2 7 2 6 .
A+B= , ( A + B ) t
= 2 6 , At
+ B t
=
6 6 8 7 8
7 8
9 0 9 0
9 −3 6 −3 12 .
3A = , (3 A) t
= −3 12 , 3 At
=
0 12 21 6 21
6 21
Remark. In the above example we find that ( At )t = A, ( A + B)t = At + Bt , and (3A)t = 3At. These
results are only special cases of the following theorem :
Theorem 1. If At and Bt are transposes of A and B respectively, then
(i) (At)t = A
(ii) C A + Bt = At + Bt, if A and B are comparable.
(iii) kAt = kAt, k being any complex number
Proof. (i) Let A = [aij] be an m × n matrix. Then At is an n × m matrix and (At)t is an m × n matrix.
The matrices (At)t and A are, therefore, comparable.
Also, (i, j)th element of (At)t
= (j, i)th element of At
= (i, j)th element of A
Since the matrices (At)t and A are comparable and their (i, j)th elements are equal, threfore, (At)t = A.
15
(ii) Let A = [aij] and B = [bij] be m × n matrices. Since A and B are both m × n matrices, therefore A
+ B exists and is an m × n matrix. Consequently (A + B)t is an n × m matrix, therefore
Again, At and Bt are both n × m matrices, so that At + Bt also exists and is an n × m matrix.
The matrices (A + B)t and At + Bt are both of the type n × m, and are therefore comparable.
Also, (i, j)th element (A + B)t = (j, i)th element of A + B
= aji + bji
= (j, i)th element of At + (i, j)th element Bt
= (j, i)th element of (A + B)t
Thus the matrices (A, B)t and At + Bt are comparable, and their corresponding elements are equal.
Hence (A + B)t + At + Bt .
(iii) Let A = [aij] be an m × n matrix. kA is an m × n matrix and therefore (kA)t is an n × m matrix.
Also At being an n × m matrix, kAt is an n × m matrix. The matrices (kA)t and kAt are both of type n × m,
and are, therefore, comparable. Also (i, j)th element of (kA)t = (j, i)th element of kA
= kaji
= k[(i, j)th elements of At]
Since the matrices (kA)t and kAt are comparable and their (i, j)th elements are equal, therefore
(kAt) = kAt.
Remark. If At = B, the Bt = (At)t = A, i.e., if B is the transpose of A, then A is the transpose of B.
1 3 4
3 −1 2
Example 2. If A = and B = −2 1 −1 ,
1 0 −3 0 −4 2
Compute (AB)t and BtAt.
1 3 4
3 −1 2
Solution. AB = −2 1 −1
1 0 −3 0 −4 2
5 0 17
= ,
1 15 −2
5 1
so that (AB)t = 0 15
17 −2
1 −2 0 3 1
Also, Bt = 3 1 −4 , At = −1 0 .
4 −1 2 2 −3
5 1
Therefore Bt At = 0 15 .
17 −2
Remark. In the above example (AB)t = BtAt. This is of course, only a particular case of the general
result which we state and prove in the following theorem.
Theorem 2. If A and B be matrices conformable to multiplication, then
(AB)t = BtAt
16
Proof. Let A = [aij] and B[bij] be m × n and n × p matrices particularly.
The At = [cij], where cij = aji, is an n × m matrix.
Bt = [cij], where dij = bji, is an p × n matrix. The matrices (AB)t and BtAt are
both of type of p × m, and are therefore comparable.
Also (i, j)th elements of (AB)t
= (j, i)th element of (AB)t
n
= ∑ a jk bki
k =1
n
= ∑ a jk bki
k =1
n
= ∑ dij ckj
k =1
= (i, j)th element of BtAt
Since the matrices (AB)t and BtAt are of the same type, and their (ij)th elements are equal, therefore,
(AB) = BtAt
t
9. Symmetric and Skew-Symmetric Matrices
Consider the matrices
3 0 2 0 1 3
A = 0 4 −1 , B = −1 0 −4
2 −1 5 −3 4 0
In matrix A, (1, 2)th element is equal to (2, 1)th element, (1, 3)th element is equal to (3, 1)th element,
and (2, 3)th element is equal to (3, 2)th element. Because of these properties we say that the matrix A is
symmetric.
In matrix B, (2, 1)th element is the negative of (1, 2)th element, (3, 1)th element is the negative of the
(1, 3)th element, (3, 2)th element is the negative of the (2, 3)th element, and (1, 1)th element, (2, 2)th
element, and (3, 3)th element are own negatives, (i.e., they are all zero). Because of these properties we
say that the matrix B is skew-symmetric.
Symmetric and skew-symmetric matrices play useful (an important) roles in the theory of matrices.
Definition 3. A square matrix A = [aij] is said to be symmetric if aij = aji for all i and j.
Illustrations 1. The matrices
4 1 − i 2 a h g 1 0 0
1 − i 3 7 h b f and 0 1 0 are all symmetric.
2 7 i g f c
0 0 1
2. The matrices
0 i 1 + i 0 3 6 − i
−i 0 −3 and −3 0 −4 are both skew-symmetric
−1 − i 3 0
0 −6 + i 4
3. The matrices
17
1 i 1 + i 0 1 −2
−i 2 6i and 1 0
i are neither symmetric nor skew-symmetric.
−1 − i 6i 0
3 2 −i
In the following theorem we state and prove some basic facts about symmetric and skew-symmetric
matrices.
Theorem 3.
(i) A necessary and sufficient condition for a matrix A to be symmetric is that At = A.
(ii) A necessary and sufficient condition for a matrix A to be skew-symmetric is that At = –A.
(iii) The diagonal elements of a skew-symmetric matrix are all zero.
Proof. (i) Necessity. Let A = [aij] be a symmetric matrix. Since A is symmetric, it must be square
matrix, say of order n, At is then also of order n, so that At and A are comparable. Also, (i, j)th element of
At = aji = aij = (i, j)th elements of A.
Therefore At = A.
Sufficiently. Let A = [aij] be an m × n matrix such that At = A. Since A is an m × n matrix., therefore
At is an m × n matrix. Since At and A are equal matrices, they are comparable, so that n = m, and
consequently A is a square matrix. Also, as given, (i, j)th element of At = (i, j)th, which gives aji = aij.
Since A a square matrix such that aij = aij for all i and j, therefore A is symmetric.
(ii) Necessity. Let A = [aij] be a skew-symmetric matrix. Since A is skew-symmetric, it must be a
square matrix, say of order n. At is then also of order n, so that At and A are comparable. Also, (i, j)th
element of At = aji = –aij = (i, j)th element of –A threfore At = A.
Sufficiently. Let A = [aij] b an m × n matrix such that At = –A. Since A is m × n matrix, therefore At is
an n × m matrix and –A is an m × n matrix. Since the matries At and –A are equal, they are comparable, so
that n = m, and consequently A is a square matrix. Also (i, j)th element of At = –[(i, j)th element of A],
which gives aji = –Aij.
We shall now state and prove a theorem which assures us that every square matrix can be expressed
uniquely as a sum of a symmetric and skew-symmetric matrix.
Theorem 4. Every square matrix can be expressed uniquely as the sum of a symmetric and a skew-
symmetric matrix.
Proof. Let A be an n-rowed square matrix.
Let A = X + Y,
where X is an n-rowed symmetric, and Y is an n-rowed skew-symmetric matrix. Taking the
transpose of both sides of (i), we have
At = ( X + Y )t = X t = Y t = X − Y ... (2)
since X is symmetric and Y is skew-symmetric so that Xt = X and Yt = –Y
From (1) and (2), we get
1
X= ( A + At ), ... (3)
2
1
Y= ( A − At ) ... (4)
2
We have shown that if A is expressible as the sum of a symmetric matrix X and a skew-symmetric
matrix Y, then X and Y must be given by (3) and (4). This establishes the uniqueness part. To demonstrate
18
the existence of a symmetric matrix, X and a skew-symmetric matrix Y such that A = X + Y, we have only
to see that if we write.
1 1
X= ( A + At ), Y = ( A − At ),
2 2
t
1
then Xt = ( A + At ) ,
2
t
1
= ( A + At ) ,
2
1 t
= ( A + A),
2
= X,
t
1
Yt = ( A − At )
2
t
1
= ( A − At )
2
1 t
= ( A − A)
2
= –Y ,
so that X is an n × n symmetric matrix and Y is an n × n skew-symmetric matrix. Furthermore, X + Y
= A, which completes the proof.
Example 3. Express the matrix
2 6 5
A = 3 1 4
9 −1 7
as the sume of symmetric and a skew-symmetric matrix.
Solution. Let
2 6 5
3 1 4 = X + Y,
9 −1 7
where X is a 3-rowed symmetric matrix and Y is a 3-rowed skew-symmetric matrix.
Taking transposes of both sides (1), and using the facts that (X + Y)t, = , we have
2 3 6
6 1 −1 = X – Y.
5 4 7
From (1) and (3) we find that
2 3 6 0 3 −4
2X = 6 1 −1 2Y = −3 0 5 ,
5 4 7 4 −5 0
so that
19
9 3
2 7 0 −2
2 2
9 3 3 5
X= 1 , Y = − 0
2 2 2 2
3 5
7 7 2 − 0
2 2
Verification. Since Xt = X, Yt = Y, therefore that matrix X is symmetric and the matrix Y is skew-
symmetric. Also, by actual addition we find that X + Y = A.
10. Hermitian and Skew-Hermitian Matrice
Let A = [aij] be an m × n matrix. The m × n matrix B = [bij] such that bij = is called the conjugate of
A and is denoted by A. The transpose of A is called the tranposed conjugate or transjugate of A and is
enoted byAθ.
For example
3 + 2i 4i
A=
4i
If ,
− 7 3 −
3 − 2i − 4i
then = ,
− 7 3 + 4i
3 − 2i − 7
Aθ =
− 4i 3 + 4i
It can be easily the seen that :
(i) if A be any matrix, then (Aθ)θ = A.
(ii) if and B be two matrices conformable to addition, then
(A + B)θ = Aθ + Bθ
(iii) if A be any matrix and k be any complex number, then
(kA)θ = kAθ .
(iv) if A and B be two matrix conformable to multiplication, then
(AB)θ = BθAθ.
We have the proofs of (i), (ii) to the reader and prove only (iii).
Let A = [aij] be an m × n matrix and let B = [bij] be an n × p matrix.
The matrices (AB)θ Bθ, Aθ are respectively of types p × m, p × n, and n × m, so that BθAθ is a p × m
matrix.
Since (AB)θ and BθAθ are both the type p × m, therefore they are comparable. Also (ij)th element of
n n
(AB)θ = (j, i)th element of ( AB) = ∑ a jk bki = ∑ a jk bki
k =1
k =1
n
Also, (i, j)th element of BθAθ = ∑ [(ik )th element of Bθ × (kj )th element of Aθ ]
k =1
n
= ∑ [(ki)th element of B θ × ( jk )th element of A]
k =1
20
n
= ∑ bki a jk
k =1
n
= ∑ akl bki ,
k =1
which is the same as the (i, j)th element of (AB)θ.
Since (i) (AB)θ and BθAθ are matrices of the same type, and (ii) (ij)th elements of (A, B)θ and BθAθ
are equal for all i and j, therefore (AB)θ = BθAθ.
Defintion 4. A square matrix A = [aij] is said to be Hermitian if for all i and j A square matrix A
[aij] is said to be skew-hermitina if for all i and j.
Illustration 1. The matrices
2 −i 1 3 + 4i
i 3 , 3 − 4i − 4
are Hermitian.
2. The matrices
2i 1 + 3i 1 5 − 7i
−1 + 3i , −5 − 7i 3i
are skew-hermitian.
5 i
i 4 + i
3. The matrix
1
is neither Hermitian nor skew-hermitian.
4 − i
In the following theorem we prove the important facts about hermitian and skew-hermitina matrices.
Theorem 5. Let A be an n-rowed square matrix.
(i) A necessary and sufficient condition for A to be hermitina is that Aθ = A.
(ii) A necessary and sufficient condition for A to be skew-hermitian is that Aθ = –A.
(iii) The diagonal elements of a hermitian matrix are all real, and the diagonal elements of a skew
hermitian matrix are either pure imagineries of zero.
Proof. (i) First let us assume that A is hermitian. Since A is an n-rowed square matrix, therefore Aθ
(= ( A)t ) is also an n-rowed square matrix. Consequently A and Aθ are matrices of the same type. Also, (i,
j)th element of Aθ (j, i)th element of A = a ji = aij = (i , j ) th element of A. Hence Aθ = A. Conversely, let us
suppose that Aθ = A ⋅ (i, j ) th element of Aθ = a ji . Since Aθ = A, therefore a ji = aij , i.e., aij = (a ji ) and
consequently A is hermitian.
(ii) First lets assume that A is skew-hermitian. Since A is an n-rowed square matrix, therefore, Aθ
and –A are both n-rowed square matrices, and as such they are comparable. Also (i, j)th element of Aθ = (j,
i)th element of A = a ji , and (i, j ) th element of –A = –aij. Since A is skew-hermitian, therefore a ji = − aij .
Hence (i, j)th element of Aθ = (i, j)th element of –A.
Since Aθ and –A are comparable, and their (i, j)th elements are equal for all i and j, therefore Aθ = –
A.
Conversely, let us suppose that Αθ = –A then (i, j)th elements of Aθ and –A must be equal for all i and
j. This gives a ji = − aij . Taking conjugates of both sides, we have a ji = − aij , i.e., aij – a ji , so that A is
skew-hermitian.
21
(iii) Let A be a hermitain matrix. then aij = a ji for all i and j. For a diagonal elements j = i, which
gives aij = a ji for all i, i.e., aij thus real. Thus the diagonal elements of A are all real.
If A is skew-hermitian, aij = − a ji for all i, j. For j = i this gives aij = − a ji . If aij = p + iq, then aji = p
– iq and so we have (p – iq) = –(p + iq) i.e., 2p = 0 or p = 0. The real part of aij is, therefore, zero and
consequently aji is either pure imaginary (if q ≠ 0).
We shall now state and prove a decomposition theorem which tells us that every square matrix can
be uniquely decomposed as the sum of the hermitian and a skew-hermitian matrix.
Theorem 6. Every square matrix can be express uniquely in the form X + iY, where X and Y are
hermitian matrices.
Proof. Let A be an n-rowed squrae matrix.
Let A = aij = − a ji .
= X θ + (iY )θ
= X θ − iY θ
= X − iY ,
Since Xθ = X and Yθ = Y.
From (1) and (2), we have
1 1
X=( A + Aθ ), Y = ( A − Aθ )
2 2i
We have thus shown that if A is expressivle as X + iY, where X and Y, are hermitian, the XZ and Y
must be given by (3). Thus establishes the uniqueness part.
To demonstrate the existence hermitain matrices X and Y such that A = X + iY, we have only to see
that if we write
1 1
X= ( A + Aθ ), Y = ( A − Aθ )
2 2i
θ
1
then Xθ = ( A + Aθ )
2
1
= ( A + Aθθ )
2
1 θ
= ( A + A), since Aθθ = + A.
2
= X,
θ
1
Yθ = ( A − Aθ )
2i
1 θ
=− ( A − Aθθ )
2i
1 θ
=− ( A − A), since Aθθ = A.
2i
= Y,
22
so that X and Y are hermitian.
Also, X + Y = A, which completes the proof.
Exampe 4. Express the matrix.
1 7 1− i
A = 1 + i i 2
− 3 1 4 + 2i
in the form X + iY, where X and Y are hermitian matrices.
Solution. We know that every square matrix A can be uniquely expressed in the form X + iY, whre
X and Y are hermitian matrices given by
1 1
X= ( A + Aθ ), Y = ( A − Aθ )
2 2i
1 1− i − 3
Here Aθ = 7 −i 1
1 + i 2 4 − 2i
1 1
1 4 − i −1 − i
2 2
1 1 3
θ
Therefore Xθ = (A + A ) = 4 + i 0
2 2 2
1 3
−1 + i 4
2 2
1 1
0 − 3i − − 2i
2 2
1 1 1
θ
Y= ( A − A ) = + 3i 1 − i
2i 2 2
1 1
− + 2i i 2
2 2
θ θ
Verification. Since X = X and Y = Y, the matrices X and Y are hermitian. Also X + iY = A.
Exercise
1. For each of the following matrices A, verify that At = A :
1 2 −1 −1 6 −7 a h g
2 0 −4 , 6 3 8 , h b f
−1 −4 3 −7 8 −5 g f c
2. For each of the following matrices A, verify that At = –A.
0 i 1 0 3 1 + i 0 −3 2i
−i 0 −2 , − 3 0 − 4 , 3 0 4
−1 2 0 −1 − i 4 0 −2i −4 0
1 2 −2
1
3. If A = 2 1 2 , verify that A At = At = A = I3.
3 2 −2 −1
4. If A be any square matrix, verify that the matrix A + At is symmetric and the matrix A —At is
skew-symmetric.
5. If A be any square matrix, prove that A + Aθ, AAθ, Aθ A are the hermitian and A – Aθ is
skew-hermitian.
23
3 1 −7
6. Express the matrix 2 4 8 as X + Y where X is symmetric and Y is skew-symmetric.
6 −1 2
7. Find hermitian matrices A and B such that
1 1+ i 3
A + iB = 4 2 1 − i
−3 2 + 3i i
24
LESSON 3
ELEMENTARY OPERATIONS ON A MATRIX
AND INVERSE OF A MATRIX
1. Introduction
In this lesson we shall study two important concepts—namely elementary operations on a matrix,
and inverse of a matrix. Both are devices for producing new matrices out of a given matrix. However,
there is one difference. Elementary operations can be applied to a matrix of any type but the process of
finding the inverse can be applied only to some special types of matrices. Furthermore, while it is possible
to produce as many new matrices as we like from a given matrix by means of elementary operations, the
operation of finding the inverse, if applicable, can produce only one matrix from a given matrix.
Both the types of operations that we are going to study have one things in common. Both can be
used for solving a system of linear equations. There is one important connection between the two types of
operations. Elementary operations can be used to find the inverse of a matrix—and that is one reason for
studying them together in this lesson.
2. Elementary Operations
Consider the matrices
−2 3 4 5 0 −6 −6 9 12 −2 3 4
S = , A = −2 3 4 , B = 5 0 −6 , C = 1 6 2 ,
5 0 −6
−2 −6 4 −2 3 16
E = ,F = .
5 0 −6 5 0 −6
The matrices A, B, C, D, E and F are related to S in as much as :
A can be obtained from S by interchanging the first and second rows.
B can be obtained from S by multiplying the first row of S by 3.
C can be obtained from S by adding 2 times the first row to the second row.
D can be obtained from S by interchanging the first and third columns.
E can be obtained from S by multiplying the second column of S by –2.
F can be obtained from S by adding 4 times the second column of S to the third column.
The operations described above are only examples of operations known as elementary operations.
Definition 1. An elementary operation is an operation of the following types:
Type I. Interchange of two rows of columns
Type II. Multiplication of a row or a column by a non-zero number
Type III. Addition of a multiple of one row or column to another row or column.
An elementary operation is called a row operation or a column operation according as it applies to
rows or columns.
It can be easily seen that if a matrix B is obatined from a matrix A by an E-operation then the matrix
A can be obtained form B by an E-operation of the same type.
It is convenient (and usual) to use the following notation of E-operations :
25
We shall denote the interchange of ith and jth rows (columns) by R1 ↔ Rj (Ci ↔ Cj); multiplication
of ith row (column) by a number k ≠ 0 by Ri → kRj (Ci → + kCi); and addition of k times the jth row
(column) to the ith row (column) by Ri → Ri + k Rj(Ci → Ci + kCj).
If a matrix B is obtained form a matrix A by a finite chain of E-operations, we say that A is
equivalent to B and write it is as A ~ B.
Elementary Matrices. A matrix obtained form a unit matrix by a single E-operation is called an
elemenatary matrix of an E-matrix. For example, the matrices.
0 1 0 1 0 0 1 0 0
1 0 0, 0 3 0, 0 1 0
0 0 1 0 0 1 3 0 1
are the E-matrices obtained from I3 by the E-operations R1 ↔ R2 (or C1 ↔ C2 ), R2 → 3R2 ) or (C2
→ 3C2) and R3 → R3 + 3R1 (or C1 → C1 + 3C3) respectively.
Remark. It can be easily see that the operations Ri ↔ Rj and Ci ↔ Cj have the same effect on In; Ri
→ kRi and Ci → kCi have the same effect on In; Ri → Ri + kRj and Cj → Cj + kCi have the same effect
on In. It an E-matrix is obtained from In by an E-operation τ, we say that it is the E-matrix corresponding
to the operation τ.
Effect of an elementary operation on the product of two matrices.
Before we consider the effect of an E-operation on the product of two matrices, let us consider the
following example.
3 1 −2 1 2
Example 1. Let A = 1 0 −4 and B = −1 4 .
−3 5 6 3 −5
Let C and D be the matrices obtained form A and AB respectively by the E-row operation R1 ↔ R2
Compute C, D and CB and show that D = CB.
Solution.
3 1 −2 1 2
AB = 1 0 −4 −1 4
−3 5 6 3 −5
−4 20
= −11 22 ,
10 −16
1 0 −4
C = 3 1 −2 ,
−3 5 6
−11 22
D = −4 20 ,
10 −16
1 0 −4 1 2
CB = 3 1 −2 −1 4
−3 5 6 3 −5
26
−11 22
= −4 20
10 −16
= D.
In the above example we find that the E-row operation R1 ↔ R2 on the product AB is equivalent to
the same E-operation on the pre-factor A. in other words.
Whether we apply the E-row operation R1 ↔ R2 to the matrix A and then post-multiply the resulting
matrix by B, or first multiply the matrices A and B, and then apply the E-row operation R1 ↔ R2 to the
product we get the same result.
The above example suggests the following theorem :
Theorem 1. An elementary row operation on the product of two matrices is equivalent to the same
elementary row operation on the pre-factor.
Proof. Let A [aij] and B = [bij] be m × n and n × p matrices respectively. We shall show that if A* be
the matrix obtained form A by an E-row operation, and (AB)* be the matrix obtained from AB by the same
E-row operations, then A * B = (AB)*.
We shall consider the three types of E-rows operations one by one and prove the result in each case.
Type I. Consider the E-row operation Ri ↔ Rj
Let A * and (AB)* be the matrices obtained from A and B respectivelyss by the E-row operation Ri
↔ Rj since A is of type m × n, therefore A* is also of type m × n, and consequently A * B is of type m ×
p. Also since AB is of type m × p, (AB)* is also of type m × p. The matrices A * B and (AB)* are, therefore
of the same type.
The matrices A* and A differ form each other in the ith and jth rows only, therefore, it follows that A
* B and AB differ form each other in the ith and jth row only. Also (AB)* and AB differ from each other in
the ith and jth row only.
Consequently, the matrices A* B and (AB)* can differ at the most in ith and jth rows only. It follows
that in order to prove the equality of A * B and (AB)* it is enough to show that the ith (as well as jth) row of
the matrices A * B and (AB)* are identical.
Now (i, k)th element of A* B
n
= ∑ a jr brk
r =1
= ( j , k ) th element of AB
= (i, k ) th element of ( AB)*;
showing that the ith rows of A * B and (AB)* are identical.
Similarly the jth rows of (AB)* and A * B are identical.
Hence A * B = (AB)*
Type II. Consider the E-row operations Ri ↔ kRi (k ≠ 0)
Let A* and (AB)* be the matrices obtained from A and AB respectively by the E-row operation
Ri ↔ kRi (k ≠ 0).
27
As in the case of the Type I, (i) the matrices A * B and (AB)* are of the same type (ii) A * B and
(AB)* can differ in the ith row only at the most.
n
= ∑ {(i, p) th element of A* ⋅ (p ⋅ l )th element of B}
p=1
n
= ∑ kaip bpl
p=1
n
= k ∑ aip b pl
p=1
= k ⋅ {(i, l ) th element of AB}.
= (i, l ) th element of ( AB)*,
so that the ith rows of A * B and (AB)* are identical.
Hence A * B = (AB)*
Type III. Let A* and (AB)* be the matrices obtained from A to AB respectively by the E-row
operation Ri → Ri + kRj. As in the case of Type I, (i) the matrices A * B and (AB)* are of the same type.
(ii) A * B and (AB)* can differ in the ith row at the most. Therefore in order to complete the proof it
is enough to show that the ith rows of A * B and (AB)* are identical.
Now (i, j)th element A * B
n
= ∑ {(i, r )th element of A* ⋅ (r ⋅ l ) th element of B}
i =1
n
= ∑ (air + ka jr ) brl
r =1
n n
= ∑ air brl + k ∑ a jr brl
r =1 r =1
= (i, l)th element of AB + k times = (j, l)th element of AB
= (i, l)ss element of (AB)*,
showing that the ith rows of A * B and (AB)* are identical.
Hence A * B = (AB)*
From the above we find that if A* and (AB)* be the matrices obtained from A and AB respectively
by the same E-row operation of any one of the three types, then A * B = (AB)*.
A theorem similar to the above holds for E-column operations.
Theorem 2. An elementary column operation on the product of two matrices is equivalent to the
same column operation on the post-factor.
Proof. Given a matrix M and an E-column operation c, let us denote by c(M) the matrix obtained for
M by the E-column operation c. Also, let us denote by c* (M) the matrix obtained from M by the
corresponding E-row operation.
Let A and B be two matrices of types m × n and n × p respectively. Denoting by c(A) and c(AB) the
matrices obtained from A and AB respectively by the E-column operation c, we have
c(AB) = [c * ( AB)t ]t ,
28
= [c * ( Bt At ]t ,
= [c * ( Bt ) At ]t
= ( At )t [c * ( Bt )]t
= Ac(B).
This proves the theorem.
Remark. Instead of deducing the theroem 2 from theorem 1, we can prove it by imitating the proof
of theorem 1.
As direct consequences of the above theorems we can prove the following important theorem.
Theorem 3. Every elementary row (resp. column) operation on a matrix is equivalent to pre-
multiplication (resp. post-multiplication) by the elementary matrix corresponding to that theorem.
Proof. Let A = [aij] be an m × n matrix. With the same notation as in the proof of theorem 2, we
have
A* = (ImA)* = I*m A,
from which we find that A* can be obtained from A by pre-multiplying A by the elementary matrix
Again, writing
A = AIn, we have
c(A) = c(Aln)
= Ac(ln),
From which we find that c(A) can be obtained from A by post-multiplying A by the elementary
matrix c(In). Hence the theorem.
Example 2. Reduce the matrix.
–1 2 1 8
A = 3 1 −1 0
4 2 1 7
to triangular from by E-row operations.
Solution.
Step 1. By suitable E-row operations on A well shall reduce A to a matrix in which all elements in
the first column except the first element are zero. We can do this by the E-row operations, R2 → R2 + 3R1
and R3 → R3 + 4R1. These two operationa yield
−1 2 1 8
A ~ 0 7 −2 24 = B, say
0 10 5 39
Step 2. By suitable E-row operations, we shall reduce B to a matrix in which all the elements in the
second column except the first two are zero. Since we would like that in this process the elements of the
10
first column remain unaltered, therefore we apply the E-row operations R3 → R3 – R2 to B. This will
7
reduce the (3, 2)th element to zero and will not alter any element of the first column.
We thus get
29
−1 2 1 8
B ~ 0 7 −2 24 = C , say.
0 0 55 33
7 7
The matrix C is a triangular matrix. We thus find that the E-row operations R2 → R2 + 3R1, R3 → R3
10
– 4R1, R3 → R3 – R2 reduce A to the triangular matrix C.
7
Remarks. 1. If a matrix has more than three rows, then we shall have several steps which would be
similar to step 2 of the above example.
2. If one of the elements in the first column is either 1 or — 1, it is convenient but not essential to
bring it ot the (1, 1)th place as the initial step in reducing a matrix to triangular form.
3. If every element in the first column of the given matrix happens to be zero then step 1 of the
above example is not required and we have to start with step 2.
4. If the (1, 1)th element is zero, but there is at least one element in the first column different from
zero, we apply a suitable E-row operation and bring this non-zero element to (1, 1)th place. For example,
in order to reduce the matrix
0 3 −1
−2 −4 −5
3 1 9
to triangular form we apply the E-row operation R1 ↔ R2 so as to obtain the matrix.
−2 −4 −5
0 3 −1
3 1 9
in which the (1, 1)th element is different from zero. We can now effect triangular reduction of this
matrix in the same way as the in the above example.
5. In the illustration in remark 4 above, we could as well have applied the E-row operation R1 ↔ R3
to the given matrix to obtain a matrix in which the (1, 1)th element is difficult from zero.
6. Triangular reduction of a matrix is not unique. In fact, if we apply the E-row operation R3 → 7R3
on the matrix C in the above example, then
We get the matrix
−1 2 1 8
0 7 −2 24
0 0 55 33
which is also a traingular matrix.
Theorem 4. Every matrix can be reduced to a triangular form by elementary row operations.
Solution. We shall prove the theorem by inductive on the number of rows.
Let A = [aij] be an m × n matrix.
The theorem is trivially true when m = 1, for every matrix having only one row is a triangular
matrix.
30
Let us assume that the theorem holds for all matrices having (m – 1) rows, i.e., every matrix having
(m – 1) rows can be reduced to triangular form by E-row operations. We shall show that A can be reduced
to triangular form by E-row operations.
Three different cases arise according as (i) a11 ≠ 0, (ii) a11 = 0 but a11 ≠ 0 for some i, (iii) a11 = 0 for
all i. We shall consider these cases onle by one.
Case (i). Let a11 ≠ 0. The E-row operation R1 → a11–1 R1 reduces A to an m × n matrix B = [bij] in
which b11 = 1. The E-row operations Rf → Rf – bf1 R1 (f = 1, 2, ..., m) reduce B to an m + n matrix C in
which Cf1 = 0 whenever f > 1. The matrix C is of the form
1 c12 c12 K c1n
0 c22 c23 K c2 n
K K K K K
0 cm 2 cm3 K cmn
By our hypothesis, the (m – 1) rowed matrix
c22 c23 K c2 n
K K K K
c cm3 K cmn
m2
can be reduced to triangular from by E-row operations. If we apply the corresponding E-row
operations to C, it will be reduced to triangular form.
Case (ii). Let a11 = 0, but ali ≠ 0 for some f such that 1 ≤ f ≤ m. By applying the E-row operation R1
→ R1, to A we get a matrix D = [dij] in which d11 = a11 ≠ 0.
We can now proceed as in case (i) and reduce D to triangular form by E-row operations.
Case (iii). If ail = 0 for all i such that l ≤ i ≤ m, (i.e., all the elements in the first column are zero),
then A is of the form
0 a12 a13 K a1n
0 a22 a23 K a2 n
K K K K K
0 am 2 am 3 K amn
By hypothesis we can reduce the (m – 1) - rowed matrix
a22 a23 K a2 n
K K K K
a am3 K amn
m2
to triangular form by E-row operations. When we apply the corresponding E-row operations to A, it will
be reduced to triangular form.
From the above we find that in all the three cases the matrix A can be reduced to trinagular form by
E-row operations.
The proof is now complete by the principle of finite induction.
Exercise 1
1. Apply the elementary operation R2 ↔ R3 to the matrix
31
1 −1 3
4 2 −6
5 8 9
2. Apply the elementary operation C2 → 2C2 to the matrixss
−1 3 7 6
5 −1 4 −2
3. Write down the elementary matrix obtained by applying R3 → R3 – 4R1 to l3.
4. Reduce the matrix
1 2 −3 4
3 −1 2 0
2 1 −1 5
to triangular form by applying E-row operations
5. Reduce the matrix
1 −1 −1
4 1 0
8 1 1
to I3 by E-row operations.
6. Verify that the E-row equation R1 → R1 – R3 on the matrix AB is equivalent to the same E-row
operation on A, where
1 −1 2 2 3 −5
A = −3 1 4 , B = 1 −2 6 .
0 1 3 1 1
3
3. Inverse of a Matrix
Consider the matrices
1 2 −1 4 −1 1
A = −4 −7 −4 , B = 4 1 0 .
−4 −9 5 8 1 1
It can be easily seen that
AB = BA = I3.
Because of this relationship between A and B we say that the matrix B is an inverse of the matrix A.
In fact, we have the following definition.
Definition 2. If A = [aij] be an n-rowed square matrix, then a square matrix B = [bij] is said to be an
inverse of A if AB = BA = In.
Since the relation AB = BA remains unaltered when we interchange A and B, it follows that if a
matrix B is an inverse of a matrix, A then A must be the inverse of B. Furthermore, it is obvious from the
defintion of an inverse of a matrix that if a matrix A has an inverse, then it must be a square matrix. In
fact, if A be an m × n matrix and B be a p × q matrix such that AB and BA both exist, we must have p = n,
and g = m. If, AB and BA are to he comparable (which is a necessary condition for AB and BA to be
equal) we must also have m = n, i.e., A must be a square matrix.
A matrix a having an inverse is called an invertible matrix.
It is quite natural to ask the following questions regarding inverses of matrices :
Question 1. Does every square matrix have an inverse?
32
Question 2. In case the answer to question 1 is ‘No’, how can we test as two whether a given matrix
in invertible?
Questions 3. Can a matrix have more than one inverse?
Questions 4. In case the answer to question 2 is ‘No’ how can we proceeed to determine the inverse
of invertible matrix?
We shall try to answer to above quetions one by one.
Answer to question 1. Let us consider the matrix.
0 0
A= .
0 0
If be any 2-rowed square matrix, we find that
AB = BA = 0.
We thus find that there cannot be any matrix B for which AB and BA are both equal to I*. Therefore
A is not invertible. We, therefore, find that a given square matrix need not have an inverse.
Before taking up question 2, we shall try to answer question 3.
Answer to questions 3. As the following theorem shows, if a matrix possesses an inverse, it must
be unique.
Theorem 5. If B and VC be inverses of a square matrix, A then B = C.
Proof. Let B and C be inverses of a square matrix A. Since B is an inverse of A, therefore
AB = BA = I ... (1)
Again, since C is an inverse of A, therefore
AC = CA = I ... (2)
From (1) we find that
C(AB) = CI = I A... (3)
Also, from (2) we find that
(CA)B = IB = B ... (4)
Since C(AB) = (CA)B , it follows from (3) and (4), that
C= B
In view of the above theorem it is only proper to talk of the inverse of a square matrix rather than
taking of an inverse.
The inverse of an inversible matrix is denoted by A–1.
Answer to question 2. Suppose a square matrix A possesses an inverse B. Then
AB = BA = I.
Since the determinant of the product of two matrices is equal to the product of their determinantsm,
we find that
| AB | = | I |
i.e., | A | × | B |= 1
From the above relation we find that | A | ≠ 0. Thus a necessary condition for a square matrix to
have an inverse is that | A | ≠ 0. We shall now show that this conidtion is sufficient as well. In order to do
1
so, let us consider the matrix C = adj. A.
| A|
By using the idendity
33
A(adj. A) = (adj. A) A = | A | I,
1 1
We find that AC = ( A adj. A) = ⋅| A | I = I,
| A| | A|
1 1 1
and CA = adj. A A = (adj. A) A = ⋅ | A | I = I.
| A| | A| | A|
Since AC = CA = I,
it follows that C is the inverse of A.
From the above discussion we find that a square matrix possess an inverse if and only if
| A | ≠ 0.
This answers questgion 2. In order to test whether a square matrix possesses an inverse, we have
simply to calculate | A |. If | A | = 0 then A does not possesses an inverse but if | A | ≠ 0 then A possesses in
inverse.
Answer to question 4. White trying to answer quetions 2, we saw that if | A | ≠ 0, then the matrix
1
adj. A is the inverse of A. This provides one possible answer to question 4.
| A|
If a square matrix A possesses an inverse A, then in order to find the inverse we compute adj A, and
1
multiply it by the scalar . The resulting matrix is the desired inverse of A.
| A|
Example 3. Find the inver of the matrix
2 5 0
0 1 1 by first computing its adjoint.
−1 0 3
Solution. The given matrix be denoted by A.
The co-factors of the elements of the first row of A are 3, – 1, and 1 respectively.
The co-factors of the elements of the second row of A are –15, 6, and –5 respectively.
The co-factors of the elements of the third row are 5, –2, and 2 respectively.
3 −1 1
Therefore adj. A = −15 6 −5
5 −2 2
Also | A | = 2.3 + 5(–1) + 0.1 = 1
3 −1 1
Therefore A–1 = −15 6 −5
5 −2 2
4. Use of Elementary Operations to Compute the Inverse of a Matrix
We have already described one method for finding the inverse of an invertible matrix. We shall now
describe another method for the same, namely by applying elementary row operations.
Suppose we wish to find the inverse of an n × n matrix A. We consider the identity
IA = A.
We reduce the matrix A on the right hand side to triangular from by E-row operations, and apply the
same E-row operations to the pre-factor I on the left hand side. In this manner we get the identity
34
PA = Q,
where P and Q are some triangular matrices. As our next step, we apply E-row operations on Q and
reduce it to the unit matrix I. The same E-row operations are, of course, applied to P. We get the identity
BA = I.
where B is obtained from P by E-row operations. The matrix B is the desired inverse of A. We shall
illustrate the procedure by considering a few examples.
Example 4. Find the inver of the matrix.
3 −1 1
A = −15 6 −5
5 −2 2
Solution. Consider the identity
1 0 0 2 3 −5 2 3 −5
0 1 0 −3 −5 9 = −3 −5 9
0 0 1 1 1 −2 1 1 −2
3
By performing the E-row operation R2 → R2 + R1 on the matrix on right as well as on the pre-
2
factor on the left, we have
1 0 0 2 3 −5
3 2 3 −5 1 3
1 0 −3 −5 9 = −3 −
2 1 1 −2 2 2
0 0 1 1 1 −2
1
Performing R3 → R3 – R1 on the matrix on the right as well as on the pre-factor on the left, we
2
have
1 0 0 2 3 −5
3 2 3 −5 1 3
1 0 −3 −5 9 = 0 −
2 1 1 −2 2 2
− 1 0 0 0 − 1 1
2 2 2
Performing R3 → R3 – R2 on the matrix on the right as well as on the pre-factor on the left, we have
1 0 0 2 3 −5
3 2 3 −5 1 3
1 0 −3 −5 9 = 0 −
2 1 1 −2 2 2
−2 −1 1 0 0 −1
Performing R1 → R1 – 5R3, on the matrix on the right as well as on the pre-factor on the left, we
have
11 5 −5 2 3 0
3 2 3 −5
1 3 1
− − −3 −5 9 = 0 − 0
2 2 2 1 1 −2 2
− 2 − 1 1 0 0 −1
Performing R1 → R1 + 6R2 on the matrix on the right as well as on the pre-factor on the left, we have
35
2 2 4 2 0 0
3 2 3 −5
1 3 1
− − −3 − 5 9 = 0 − 0
2 2 2 1 1 −2 2
− 2 − 1 1 0
0 −1
1
Performing R1 , R2 → − 2 R2 , and R3 → − R3 on the matrix on the right as well as on the pre-factor
2
on the left, we have
1 1 2 2 3 −5 1 0 0
3 1 −3 −3 −5 9 = 0 1 0
2 1 −1 1 1 −2 0 0 1
From the above identity we find that the desired inverse is
1 1 2
B = 3 1 −3
2 1 −1
Verification. By actual multiplication we find that
AB = BA = I,
The above working an often be arranged in a more convenient and compact from as in the following
example.
Example 5. Find the invers of the matrix
4 −1 −4
3 0 −4 .
3 −1 −3
Solution. We shall first reduce the given matrix to I3 by E-row operations
4 −1 −4
4 −1 −4
3 0 −4 ~ 0 3 3 3
−1 R2 → R2 − R1 , R3 → R3 − R1
3 −1 −3 4 4 4
1
0 − 0
4
4 −1 −4
3 1
~ 0 −1 R3 → R3 + R2
4 3
0 0 −1
3
4 −1 0
3
~ 0 0 [ R1 → R1 − 12 R3 , R2 → R2 − 3R3 ]
4
0 0 −1
3
36
4 0 0
3 4
~ 0 0 R1 → R1 + R2
4 3
0 1
0 −
3
1 0 0
1 4
~ 0 1 0 R1 → R1 , R2 → R2 , R3 → − 3R3
0 0 1 4 3
3 3
The given matrix has been reduced to I3 by the E-row operation R2 → R2 − R1 , R3 → R3 − R2 ,
4 4
1 4 1 4
R3 → R3 + R2 , R1 → R1 − 12 R3 , R2 → R2 − 3R3 , R1 → R1 + R2 , R1 → R1 , R2 → R2, and R3 → 3R3.
3 3 4 3
We shall now perform these operations on l3 in this order. We then have
1 0 0
1 0 0
0 1 0 ~ − 3 1 0 R2 → R2 − 3 R1 , R3 − 3 R1
0 0 1 4 4
4
− 3 0 1
4
1 0 0
3 1
~ − 1 0 R3 → R3 + R2
4 3
−1 1 1
3
13 −4 −12
9
~ 0 −3 [ R1 → R1 − 12 R3 , R2 → R2 − 3R3 ]
7
−1 1
1
3
16 −4 −10
9 4
~ 0 −3 R1 → R1 + R2
4 3
−1 1
1
3
4 −1 −4
1 4
~ 3 0 −4 R1 → R1 , R2 → R2 , R3 → − 3R3
3 −1 −3 4 3
The last matrix is the desired invers.
The following theorem is sometimes useful for conputing inverses:
Theorem 6. If A and B are n-rowed invertible matrices, then At and AB are both invertible and
(a) (At)–1 = (A–1)t.
(b) (AB)–1 = B–1A–1.
Proof. (a). Since AA–1 = A–1A = In, therefore by the reversal law for transposes we have
37
(AA–1)t = (A–1A)t = Itn,
i.e., (A–1)t At = A(A–1)t = In.
By the definition of the inverse of a matrix it follows that the matrix At is invertible and its inverse is
the matrix (A–1)t.
(b) Since A and B are both n-rowed invertible matrices, therefore A–1 and B–1 exist and are both n-
rowed square matrices, and consequently B–1A–1 is also an n-rowed square matrix. Let us denote the
matrix B–1A–1 C. We wish to show that B–1A–1 is the inverse of the n-rowed square matrix AB. In order to
achieve our aim we have to verify that
(AB) C = C(AB) = In,
where C denotes the matrix B–1A–1.
Now (AB)C = ( AB) ( B −1 A−1 )
= A( BB –1 ) A−1
= AIn A–1
= AA–1
= In
Also, C(AB) = (B–1 A–1) AB
= B–1(A–1 A)B
= B–1In B
= B–1B
= In.
From the above relations we find that (AB) C = (AB) = In.
Hence it follows that the matrix AB is invertible and its inverse is the matrix, i.e.,
(AB)–1 = C = B–1A–1.
Remark. Result (b) of the above theorem is often referred to as the reversal law of inverses.
The following example illustrates how the reversal law for inverse can be used for computing the
inverses of the product of two invertible matrices.
Example 6. Obtain the invers of the matrices
1 p 0 1 0 0
A = 0 1 p , B = q 1 0
0 0 1 0 q 1
and hence that of the matrix
1 p 0 1 0 0
C = 0 1 p , B = q 1 0
0 0 1 0 q 1
Solution. In the very begining let us observe the following two facts which will be very useful :
(i) B is simply the matrix At with p replaced by q every where in At.
(ii) C = AB.
Step 1. Compute A–1
To compute A–1 we reduce A to I3, by E-row operations.
38
1 p 0 1 p 0
0 1 p ~ 0 1 0 [ R2 → R2 − pR3 ]
0 0 1 0 0 1
1 0 0
~ 0 1 0 [ R1 → R1 − pR2 ]
0 0 1
We find that A can be reduced to I3 by applying the E-row operations R2 → R2 – pR3 and R1 → R1 –
pR2. We now apply these E-row operations to I3 in this very order. We than get the matrix.
1 0 0
I3 ~ 0 1 − p [ R2 → R2 − pR3 ]
0 0 1
1 − p p2
= 0 1 − p [ R1 → R1 − pR2 ]
0 0 1
1 − p p2
Thus A–1 = 0 1 −p .... (1)
0 0 1
From (i) we find that
1 0 0
( At ) −1 = ( A−1 )t = − p 1 0
2
p − p 1
Replacing p and q in the above identity throughout and using (i), we have
1 0 0
B–1 = −q 1 0 ... (2)
2
q − q 1
By applying the reversal law to (ii) we have
C–1 = (AB)–1 = B–1A–1
1 0 0 1 − p p2
= −q 1 0 0 1 −p
2
q −q 1 0 0 1
1 −p p2
= −q 1 + pq − p2q − p .
q − pq 2 − q
2
p 2 q 2 + pq + 1
Exercise 2
1. Find the invers of the matrix.
3 −15 5
−1 6 −2
1 −5 2
39
by calculating its adjoint.
2. Show that the matrix.
1 −2 3
0 −1 4
−2 2 1
possesses an inverse.
1 1 0
3. Given that A = −2 −1 0 , calculate A2 and A–1.
0 0 2
1 2 −1 2 1 2
4. If A = −4 −7 4 , B = 2 2 1 , verify that (AB)–1 = B–1A–1
−4 −9 5 1 2 2
5. By applying elementary row operations find the inverse of the matrix
1 1 2
3 1 −3 .
2 1 −1
40
LESSON 4
RANK OF A MATRIX
1. Introduction
In this lesson we shall introduce the important concept of the rank of a matrix. We shall show that
the rank of matrix remains unaltered by elementary operations. This will give us a neat method for finding
the rank of matrix.
In the next lesson we shall use the concept of rank to determine whether a given system of linear
equation possesses a solution, and if so, then how many linearly independent solutions are there.
2. Rank of Matrices
Consider the matrix
2 −3 −1 4
A = 7 0 6 −5 .
1 4 −3 2
A is a 3 × 4 matrix. By delecting any one column of A we can obtain a 3 × 3 matrix. Let us recall
that such a matrix is called a sub-matrix of A. The determinant of such a matrix is called a minor of A.
Thus for example,
2 −3 −1 2 −1 4 −3 −1 4
7 0 6 , 7 6 −5 , 0 6 −5
1 4 −3 1 −3 2 4 −3 2
are all minor of A. Since each of these determinants has 3 rows, these minors are called 3-rowed
minors.
By detecting any one row and any two columns of A we can obtain sub-matrices of the type 2 × 2.
The determinant of any such sub-matrices is called a 2-rowed minor of A. Thus
2 −3 −1 4 0 6
, ,
7 0 6 −5 4 −3
are all 2-rowed minors of A.
In general, we have the following definitions:
Definition 1. Let A be an m × n matrix. The determinant of any s × s submatrix of A, obtained by
deleting m-s rows and n-s columns of A is called and s-rowed minor of A.
Let us now consider the matrix
2 3 −5
A = 1 −7 8 .
9 −19 9
The matrix A has only one 3-rowed minor, namely
2 3 −5
1 −7 8 ,
9 −16 9
and this can be seen to be zero.
41
A has 9 two-rowed minors (why ?) These are not all zero. In fact,
2 3
1 −7
is a 2-rowed minor which has the value 2. (–7) – 3.1 = –17.
The fact that every 3-rowed minor of A is zero and there is at least one 2-rowed minor which is not
zero, is usally expressed bny saying that the rank of the matrix A is 2.
Let us now consider the matrix
1 0 0
I = 0 1 0 .
0 0 1
It has only one 3-rowed minor, namely
1 0 0
0 1 0 ,
0 0 1
and this minor has the value 1. Since I a 3-rowed non-zero minor, and there are not minors having more
than three rows, we say that I is of rank 3.
In general, we have the following definition :
Definition 2. (1) A number r is said to be the rank of mark of a matrix A if
(i) A possesses at least one r-rowed minor which is different from zero.
(ii) A does not possess any non-zero (r + 1)-rowed minor
(2) The rank of every zero matrix is, by definition zero.
The rank of a matrix A is usually denoted by p(A).
Illustrations 1. If A In, then p(A) = n.
0 0 0 0
2. If A = , then ρ( A) = 0.
0 0 0 0
From the definition of the rank of a matrix we have the following useful criteria for determining the
rank of a matrix;
(i) If all the (r + 1)-rowed minors of a matrix vanish or if the matrix does not possess any (r + 1)-
rowed minor, then the rank of the matrix must be either less than or at the most equal to r.
(ii) It atleast one r-rowed minor of a matrix does not vanish, then the rank of the matrix must be
either grether than or at least equal to r.
It is important (and useful) to note that if every (r + 1)-rowed minor of a matrix is zero, then every
higher order minor is automatically zero (because we can expand every higher order minor ultimately in
terms of (r + 1)-rowed minors.
Example 1. Find the rank of the matrix.
2 −4 5 6
A= .
3 8 7 −1
Solution. The given matrix has at least one non-zero minor of order 2, namely
2 −4
3 8
42
(In fact this minor = 2.8 – (–4).3 = 28) Actually, A has 6 2-rowed minors all the of which are non-
zero.
Since A has a non-zero 2-rowed minor, of therefore ρ(A) ≥ 2.
Again, since A is a 2 × 4 matrix, therefore it does not have any 3-rowed minors. Consequently ρ(A)
≤ 2.
From the relations ρ(A) ≥ 2 and ρ(A) ≤ 2 we have ρ(A) = 2.
Example 2. Find the rank of the matrix
1 1 1
A = a b c ,
2 2
2
a b c
all of a, b, and c being real numbers.
Solution.
1 1 1
|A|= a b c
a2 b2 c2
= (b – c) (c – a) (a – b).
The following different cases arise:
Case I. a, b, and c are all different. If a, b, and c all the different from each other, then | A | ≠ 0.
Therefore A has a non-zero three-rowed minor, namely
1 1 1
a b c ,
a2 b2 c2
so that ρ(A) ≥ 3.
Also, A has no higher order minors.
Therefore ρ(A) ≤ 3.
Hence ρ(A) = 3.
Case II. Two of the numbers a, b, and c are equal but a, b, and c are not all equal. For the sake of
definiteness assume that a = b ≠ c.
In this case | A | = 0 since | A | is the only 3-rowed minor and this minor vanishes, therefore
ρ| A | ≤ 2. Also, A has atleast one non-zero 2-rowd minor, namely
1 1
b c
Therefore ρ(A) ≥ 2.
Hence ρ(A) = 2.
In the same way as above we find that ρ(A) = 2 in the cases b = c ≠ a or c = a ≠ b.
Case III. a = b = c,
43
1 1 1
Since a = b = c, therefore | A | = a a a = 0 so that the only 3-rowed minor is zero. Also all
a2 a2 a2
the 2-rowed minors of A vanish. In fact all three columns of A are identical in this case and therefore all
the nine 2-rowed minors of A vanish. Therefore, ρ(A) ≤ 1.
Also, A has a non-zero of 1-rowed monor. (In fact these are at least three 1-rowed non-zero minors,
namely those which are obtained by deleting the 2nd and 3rd rows and any tow columns of A). Therefore
ρ(A) ≥ 1.
Hence ρ(A) = 1.
3. Rank and Elementary Operations
In the two examples discussed above, we has obtained the ranks of the matrices being considered by
a straight forward consideration of all the minors. This may not be so easy in all the cases. It we have to
find the rank of A 4 × 4 matrix, we may have to consider a fourth order determinants, 16 third order
determinants, and perhaps some second order determinants.
By using elementary operations we can conveniently handle the problem of determining the rank of
a matrix. The following theorem gives the relation between elementary operations on a matrix and its
rank.
Theorem 1. The rank of a matrix remains unaltered by elementary operations.
Proof. We shall prove the theorem for elementary row operations only. The proof for elementary
column operations is similar.
Let A = [aij] be an m × n matrix of rank r.
Case I. Rp ↔ Rq.
Let B be the matrix obtained from A by the E-row operation Rp ↔ Rq and let the rank of B be s.
In order to verify that ρ(B) = r, we shall show that all (r + 1)-rowed minors of B vanish and atleast
one r-rowed minor does not vanish.
Let B* be any (r + 1) - rowed squae sub-matrix of B. Since B has been obtained from A by the
operation Rp ↔ Rq, therefore every row of B is also a row of A (and every row of A is a row of B). This
implies that | B* | = ± | A* | for some (r + 1) -rowed submatrix A* or A. But every (r + 1)-rowed minor of A
is zero (because ρ(A) = r). Therefore | B* | = 0, and consequently every (r + 1)-rowed minor of B is zero.
This implies that ρ(B) ≤ r, i.e., s ≤ r.
Since B has been obtained from A by interchaning the pth and qth rows, therefore we can get A from
B by applying Rp ↔ Rq to B. By interchanging the roles of A and B in the above proof it follows that r ≤ s.
The relations s ≤ r and r ≤ s together yield r = s.
Case II. Rp → cRp (C ≠ 0)
Let us denote by B the matrix obtained from A by the opeation Rp → cRp, and let the rank of B be s.
Let B* be any (r + 1) - rowed sub-matrix of B and let A* be the correspondingly placed sub-matrix of A.
Let us compare | B* | and | A* |. Two possiblities arise :
(i) pth row of B is one of the row struck off to obtain B* from B.
In this case B* is identical with A* and | B* | = | A* |.
44
(ii) pth row of B is not among the rows struck off to obtain B* from B. In this case | B |* =
c | A* |, for B* and A* agree with each other in all rows except one row, and every element of this
paritcular row of B* is c times the corresponding element of A*.
Since A is of rank r, therefore every (r + 1) -rowed minor of A is zero. In particular | A* | = 0 and in
each of the two cases (i) and (ii) above, we have | B* | = 0.
Since every (r + 1) - rowed minor of B is zero, therefore P(B) ≤ ρ, i.e., s ≤ r. Also, since A can be
obtained from B by the E- operation Rp → C–1Rp, therefore by interchanging the roots of A and B it
follows that r ≤ s. The relations s ≤ r and r ≤ s together yield r = s.
Case III. Rp → Rp + kRq (k ≠ 0).
Let us denote by B the matrix obtained from A by the E-operation Rp → kRq, and the rank of B be s.
Let B* be any (r + 1) - rowed sub-matrix of B and let A* be the corresponding sub-matrix of A.
Three different possibities arise :
(i) Neither the pth row nor qth row of B has been struck off while obtaining B*. In this case A* and
B* differ in only one row (the rows which correspond to pth row of A) and have all the other rows
identical. We have | B *| = | A* |. Since every (r + 1) - rowed minor of A is zero, threfore is particular | A*
| = 0, and consequently | B* | = 0.
(ii) The pth row of B has not been struck off while obtaining B* from B but the qth row has been
struck off.
Let us denote by C* the (r + 1) - rowed matrix obtained from A* by replacing apj by aqj. By the
property of determinants we have
| B* | = | A* | + k| C* |.
Observe that C* can be obtained from A by first performing the E-operation Rp → Rq and then
striking off those very rows and columns of the new matrix thus obtained as are strucke off for obtaining
B* from B. This implies that | C* | = ±1 times some (r + 1) - rowed minor of A. Since ρ(A) = r,s therefore
every (r + 1) - rowed minor of A is zero, and so | A* | = 0, and | C* | = 0. From (1) it follows that | B* | =
0.
(iii) The pth row of B is one of those rows which have been struck off while obtaining B*. Since A
and B differ only in the pth row, therefore in this case B* = A* an so | B* | = | A* |. Since A is of rank r,
therefore | A* | = 0 and consequently | B* | = 0. Since | B* | = 0, in all the cases (i) - (iii), therefore every
(r + 1) - rowed minor of B is zero and so ρ(B) ≤ r, i.e., s ≤ r. Also since A can be obtained from B by the
E-operation Rp → Rp – kRq, therefore by interchanging the roles of A and B it follows that r ≤ s. Hence r
= s.
We have thus shown that the rank of a matrix remains unaltered by E-row operations, and so
remarked in the beginning this complest the proof.
Example 3. Reduce the matrix
3 −1 1 3
−1 −4 −2 −7
2 1 3 0
−1 −2 3 0
to triangular form by E-row operations and hence determine its rank.
Solution. Let us denote the given matrix by A. We shall first of all interchange R1 and R4 so that we
get –1 as the (1, 1) the entry. This will ultimately simplify the entire working.
45
3 −1 1 3
−1 −4 −2 −7
A=
2 1 3 0
−1 −2 3 0
−1 −2 3 0
−1 −4 −2 −7
~ , by R1 ↔ R4
2 1 3 0
3 −1 1 3
−1 −2 3 0
0 −2 −5 −7
~ , by R2 → R2 − R1 , R3 → R3 + 2 R1 , R4 → R4 + 3R1
0 −3 9 0
−7 10 2
0
−1 −2 3 0
0 −2 −5 −7
33 21 3 7
~ 0 0 , by R3 → R3 − R2 , R4 → R4 − R2
2 2 2 2
55 55
0 0
2 2
−1 −2 3 0
0 −2 −5 −7
~ 33 21 , by R → R − 5 R
0 0 4 4
3
3
2 2
0 0 0 10
If we denote the above matrix by B, then we see that B = 330 ≠ 0, so that rank of B = 4.
But A and B have the same rank.
Therefore A is of rank 4.
2 3 4 5 6
3 4 5 6 7
Example 4. Find the rank of the matrix 4 5 6 7 8 .
9 10 11 12 13
14 15 16 17 18
Solution. Let us denote the given matrix by A. Observe that if we subtract the lement of the first row
from the corresponding elements of the second row, then the second row will consist of ones only. In fact
1 1
by R2 → R2 − R1 , R3 → R − R1 , R4 → R4 − R1 , R5 → R5 − R1 , and then R3 → R3 , R4 → R4 ,
2 7
1
R5 → R12 , we find that
12
2 3 4 5 6
1 1 1 1 1
A ~ 1 1 1 1 1
1 1 1 1 1
1 1 1 1 1
46
2 1 2 4
3
1 0 0 0
0
~ 1 0 0 0 , by C 2 → C2 − C1 , C3 → C3 − C1 , C4 → C4 − C1 , C5 → C5 − C1 ]
0
1 0 0 0
0
1 0
0 0 0
2 1 0 0 0
1 0 0 0 0
~ 1 0 0 0 0 [by C3 → C3 − 2C2 , C4 − 3C2 , C5 → C5 − 4C2 ]
1 0 0 0 0
1 0 0 0 0
Denoting the last matrix by B, we find that every 3-rowed minor of B is zero, and B has atleast one
2 1
non-zero two-rowed minor, namely .
1 0
There ρ(B) = 2.
But ρ(A) = ρ(B), so that ρ(A) = 2.
Remark. Observe that in the above example we did not proceed in the normal way for triangular
reduction.
Example 5. If x, y, z are all unequal, show by using elementary operations or otherwise that the
matrix
0 x − y x − z y + z
y − x 0 y − z x + z
A=
z − x z − y 0 x + y
y + z z + z x+ y 0
is of rank 2.
Solution.
0 x− y x−z y+ z
y − x 0 y−z x+ z
A=
z − x z−y 0 x+ y
y + z z+x x+ y 0
x−y x− y x− y y − x
y− x 0 y− z x + z
~ [by R1 → R1 − R2 ]
z − x z−y 0 x + y
y + z z+x x+ y 0
1 1 1 −1
y − x 0 y − z x + z 1
~ , by R1 → R1
z − x z − y 0 x + y x−y
y + z x− y x+ y 0
1 0 0 0
y − x x − y x−z y + z
~ , [by C 2 → C2 − C1 , C3 → C3 − C1 , C4 → C4 + C1 ]
z − x x − y x−z y + z
y + z x− y x−z y + z
47
1 0 0 0
0 x − y x−z y + z
~ , [by R2 → R2 − ( y − z ) R1 , R3 →
0 x − y x−z y + z
0 x − y x−z y + z
Two different cases arise according as y + z ≠ 0 or y + z = 0.
Case I. y + z ≠ 0.
1 0 0 0
0 1 1 1 1 1 1
A~ , by C2 → C2 , C3 → C3 , C4 →
1
C4
0 1 1 x− y x−z y+ z
0 1 1 1
1 0 0 0
0 1 0 0
~ , [by C3 → C3 − C2 , C4 → C4 − C2 ]
0 1 0 0
0 1 0 0
1 0 0 0
0 1 0 0
~ , [by R3 → R3 − R2 , R4 → R4 ~ R2 ]
0 0 0 0
0 0 0 0
1 0 0 0
0 1 0 0
Since the matrix is of rank 2, therefore it follows that ρ(A) = 2
0 0 0 0
0 0 0 0
Case II. y + z = 0
1 0 0 0
0 x − y x − z 0
A~
0 x − y x − z 0
0 x − y x − z 0
1 0 0 0
0 1 1 0 1 1
~ , by C2 → C2 , C3 →
0
C3
0 1 1 x−y x−z
0 1 1 0
1 0 0 0
0 1 1 0
~ , [by R3 → R3 − R2 , R4 → R4 − R2 ]
0 0 0 0
0 0 0 0
1 0 0 0
0 1 0 0
~ , [by C3 → C3 − C2 ]
0 0 0 0
0 0 0 0
which is of rank 2.
Hence ρ(A) = 2 in this case also.
48
Thus ρ(A) = 2
Example 6. Find the rank of the matrix
0 r −q x
−r 0 p y
q −p 0 z
,
−x − y −z 0
where px + qy + rz = 0, and all of p, q, and r are positive real numbers.s
Solution. Let the given matrix be denoted by A.
0 r −q x
−r 0 p y
A=
z
,
q − p 0
− x − y − z 0
0 r −q px
−r 0 p py
~ , [by C4 → pC4 , since p ≠ 0]
q − p 0 py
q − p −2 0
0 r −q px + qy
−r 0 p 0
~ , [by C4 → C4 − yC3 ]
q − p 0 pz
− x − y − z yz
0 r −q px + qy + rz
−r 0 p 0
~ , [by C4 → C4 + zC2 ]
q − p 0 0
− x − y − z 0
0 r −q 0
−r 0 p 0
~ , since px + qy + rz = 0.
q − p 0 0
−x − y −z 0
0 r − qr 0
−r 0 pr 0
~ , [by C3 → rC3 ]
q − p 0 0
− x − y − rz 0
0 r 0 0
−r 0 pr 0
~ , [C → C2 + qC2 ]
q − p − pq 0 3
− x − y − qy − rz 0
0 r 0 0
−r 0 0 0
~ , [by C3 → C3 + qC2 ]
q − p 0 0
− x − y −( px + qy + rz ) 0
49
0 r 0 0
−r 0 0 0
~ , since px + qy + rz = 0.
q − p 0 0
− x − y 0 0
= B(say)
Now every 3-rowed minor of the matrix B vanishes, and there is atleast one non-zero 2-rowed minor
0 r
namely , therefore the rank of B is 2. Since A ~ B, therefore ρ(A) = ρ(B) = 2.
−r 0
Example 7. If the number a, b, and c are all different from each other, and the numbers, x, y,z are
also all different from each other slow that the matrix
1 a x ax
A = 1 b y by
1 c z cz
is of rank 3.
Solution.
1 a x ax
A = 1 b y by
1 c z cz
1 a x ax
~ 0 b − a y − x by − ax , [by R2 → R2 − R1 , R3 → R3 − R1 ]
0 c − a z − x cz − ax
1 a 0 0
~ 0 b − a y − x by − ax , [by C2 → C2 − aC1 , C3 → C3 − xC1 , C4 → C4 − axC1 ]
0 c − a z − x cz − ax
1 0 0 0
~ 0 b − a y − x b( y − x) , [by C4 → C4 − xC2 ]
0 c − a z − x c( z − x)
1 0 0 0
~ 0 b − a y−x 0 , [by C4 → C4 − bC3 ]
0 c − a z − x (c − b) ( z − x)
= B(say).
Let us first of all consider the rank of B. since B is a 3 × 4 matrix, ρ(B) ≤ 3. Also, since B has a 3-
1 0 0
rowed non-zero minor, namely 0 y − x 0 , therefore r(B) ≤ 3.
0 z − x (c − b)( z − x)
From the inequalities ρ(B) ≤ 3 and ρ(B) ≥ 3 it follows that ρ(B) = 3.
Also, since the rank of a matrix remains unaltered by elementary operations, therefore ρ(A) = ρ(B) =
3.
Hence the rank of the given matrix is 3.
Exercise
1. Find the rank of the matrix
50
3 −1 2 −6
6 −2 4 −9 .
2. Find the rank of the matrix
−1 2 3
0 3 4 .
0 0 0
3. Find the rank of the matrix.
1 1 1
a b c .
3 3
3
a b c
4. Reduce the matrix
1 −1 2 −3
4 1 0 2
0 3 1 4
.
0 1 0 2
to triangular form and hence find its rank.
5. Find the rank of the matrix.
12 22 32 42
2
2 32 42 52
.
32 42 52 62
2
4 52 62 7 2
51
LESSON 5
SYSTEMS OF LINEAR EQUATIONS
1. Introduction
In this lesson we shall apply our knowledge of matrices to obtain solutions of systems of linear
equations.
Consider a system of m linear equations.
a11 x1 + a12 x2 + K + a1n xn = b1 ,
a21 x1 + a22 x2 + K + a2 n xn = b2
... (1)
KKKKKKKKKKKKK
am1 x1 + am 2 x2 + K + amn xn = bm ,
in n unknows x1, x2, ..., xn.
We can write the above system of equation as
x1 b1
x2 b2
a11 a12 K a1n
. .
a21 a22 K a2 n
. = . ... (2)
K K K K
. .
am1 am 2 K amn
. .
x b
n n
If we write
x1
a11 a12 K a1n x2
.
a a22 K a2 n
A = 21 , X = .
K K K K
.
am1 am 2 K amn
.
x
n
b1
b2
.
and B = , so that A is an m × n
.
.
b
n
matrix, X is an n × 1 matrix, and B is an m × 1 matrix, then we can write (2) as
AX = B
52
The matrix A is called the co-efficient matrix, and [A, B] is called the augmental matrix, where [A,
B] stands for the m × (n + 1) matrix obtained by adjoining one column, namely (b1, b2) ... bm)t to the m ×
n matrix A.
We shall confine out study to the following aspects :
I. Application of the inverse of a matrix to study a system of n equations AX = B in n-unknowns. We
shall show that if the matrix A possesses an inverse, then the system has the unique solution which is
given by X = A–1B. As a corollary of this result we shall show that if the matrix A is invertible, then the
system of equations AX = 0 does not possess any non-zero solution.
II. We shall use elementary row operations to put AX = B in a simplified form from which the
solution of the system can be written easily.
III. We shall apply the concept of rank of a matrix to discuss the existence and uniqueness of
solutions of AX = B. We shall show that if A be m × n matrix of rank r, then (i) the general solution of the
system of homogeneous equations AX = 0 contains n–r arbitrary constants, and (ii) the system of non-
homogeneous equations AX = B posses a solution if and only if the matrix [A, B] is also of rank r. In case
this condition is satisfied, the solution contains n–r arbitrary constant.
2. Use of Inverse of Matrix to Solve a System of Linear Equations
Consider a system of n linear equations
a11 x1 + a12 x2 + K + a1n xn = b1 ,
a21 x1 + a22 x2 + K + a2 n xn = b2
KKKKKKKKKKKKK
... (1)
an1 x1 + an 2 x2 + K + ann xn = bn ,
in n unknowns x1, x2, ... xn.
We can write the above system of equation as
x1 b1
x b
a11 a12 K a1n 2 2
a21 . .
a22 K a2 n
. = .
K
... (2)
K K K
. .
an1 an 2 K ann
. .
x b
n n
If we write
x1
x2
a11 a12 K a1n .
a a22 K a2 n
A = 21 , X = . ,
K K K K
.
an1 an 2 K ann .
x
n
b1
b2
.
and B = ,
.
.
b
n
53
so that A is an n × n matrix, X is an n × 1 matrix, and B is an n × 1 matrix then we can write (2) as
AX = B ... (3)
If A is a non-singular matrix, so that A–1 exists, we pre-multiply (3) throughout by A–1 , as that
A–1 (AX) = A–1 B
⇒ (A–1A)X = A–1B
⇒ IX = A–1B
⇒ = A–1B. ... (4)
From (4) we find that if the system of equations (3) has a solution, it is given by (4). By substituting
X = A–1B in (3) we find that
AX = A(A–1B) = (AA–1) B = IB = B, showing that X = A–1B is indeed a solution. Thus we have the
following result.
Corrollary. If A is an n-rowed non-singular matrix, the system of homogeneous linear equations.
AX = 0 has X = 0 as its only solution.
Example 1. Show that the system of equations
4x + 3y + 3z = 1
–x – z = 2
–4x – 4y – 3z = 3
has a unique solution, and obtain it by first computing the inverse of the co-efficient matrix.
Solution. The given system of equations can be written as
4 3 3 x 1
−1 0 −1 y = 2 ,
−4 −4 −3 x 3
or as AX = B.
4 3 3 x 1
where A = −1 0 −1 , X = y , B = 2 .
−4 −4 −3 z
3
4 3 3
|A|= −1 0 −1
−4 −4 −3
0 −1 −1 −1 −1 0
= −3 +3
−4 −3 −4 −3 −4 −4
= 4(–4) – 3. (–1) + 3.4
= –1.
Since | A | ≠ 0, the matrix is invertible. Since the co-efficient matrix is invertible, therefore the given
system of equations possesses a unique solution.
We shall now find A–1.
−4 −3 −3
Now adj. A= 1 0 1
4 4 3
54
1
A–1 = adj. A
| A|
4 3 3
= −1 0 −1
−4 −4 −3
X = A–1B
4 3 3 1
= −1 0 −1 2 ,
−4 −4 −3 3
19
= −4 .
−21
Therefore x = 19, y = –4, z = –21 is the desired solution.
Verification. When x = 19, y = –4, z = –21,
4x + 3y + 3z = 4.19 + 3(–4) + 3 (–21) = 1,
–x –z = –19 + 21 = 2
–4x – 4y –3z = –4 (19) –4 (–4) –3 (–21) = 3.
Example 2. Show, by considering the co-efficient matrix, that the system of equations
x + 2y – z = 0,
–4x –7y + 4z = 0,
–4x – 9y + 5z = 0,
does not possess a non-zero soltion.
Solution. The co-efficient matrix of the given system of equation of given by
1 2 −1
A = −4 −7 4
−4 −9 5
−7 4 −4 4 −4 −7
|A|= −2 −
−9 5 −4 5 −4 −9
= 1 – 2(–4) –8
= 1.
Since | A | ≠ 0, therefore A is invertible. Since A–1 exists, the given system of equations does not
possess a non-zero solution.
3. Use of Elementary Row Operation to Solve a System of Linear Equatons
Consider the system of m linear equations
a11 x1 + a12 x2 + K + a1n xn = b1 ,
a21 x1 + a22 x2 + K + a2 n xn = b2
KKKKKKKKKKKKK
s ...(1)
am1 x1 + am 2 x2 + K + amn xn = bm ,
55
By writing
a11 a12 K a1n
a21 a22 K a2 n
A=
K K K K
am1 am 2 K amn
x1 b1
x2 b2
. .
X = , and B =
. .
. .
x b
n n
we can write (1) as
AX = B,
where A = [aij] is an m × n matrix, X = ( x1 , x2 , K xn )t is an n × 1 matrix, and B = (b1 , b2 , K bn )t is an m ×
1 matrix.
The system of equations (1) remains unaltered by the following operations :
(i) Interchange of any two equations;
(ii) Multiplication of both sides of an equation by non-zero number k;
(iii) addition of k times an equation to another equation.
The above equations are equivalent to the following elementary row operations on the matrices A
and B.
(i) interchange of two rows.
(ii) multiplying a row of A and the corresponding row of B by a non-zero number k;
(iii) adding k times an equation to another equation.
Let us recall that :
(a) A matrix can be reduced to triangular form by E-row operations.
(b) An E-row operation on the product of two matrices is equivalent to the same E-row operation
on the pre-factor.
In view of the all the facts stated above we find that in order to solve a given system of linear
equation we may proceed as follows :
(1) Write the given system in matrix notation as AX = B.
(2) Reduce A to a triangular matrix say A* by E-row operations.
(3) Keep on applying the E-row operations in (2) above to the matrix B in the same order as in (2)
above.
(4) Suppose the result of (2) and (3) gives us the equation A * X = B* where A* is a triangular
matrix. By writing A * X = B* as a system of linear equations, we shall arrive at a system of
equations which can be solved easily. Of course, it can be happen that the system does not
possess a solution. This will happen if one of the equations in A * B = B* is of the form
0 · x1 + 0 · x2 + ... + 0 · xn = b,
for some non-zero number b.
We shall illustrate the method with the help of examples.
56
Example 3. Solve the system of equations
2x – 5y + 7z = 6,
x – 3y + 4z = 3,
3x – 8y + 11z = 11.
2 −5 7 6
Solution. Let A = 1 −3 4 , B = 3
3 −8 11
11
x
and assume that there exists a matrix X = y , such that AX = B
z
2 −5 7 x 6
Now 1 −3 4 y = 3
3 −8 11 z 11
1 −3 4 x 3
⇒ 2 −5 7 y = 6 , by R1 ↔ R2
3 −8 11 z 11
1 −3 4 x 3
⇒ 0 1 −1 y = 0 , by R2 → R2 − 2 R1 , R3 → R3 − 3R1
0 1 −1 z 2
1 −3 4 x 3
⇒ 0 1 −1 y = 0 , by R3 → R3 − R2
0 0 0 z 2
⇒ x – 3y + 4z = 3,
y – z = 0,
0 = 2.
Since the conclusion 0 = 2 is false, therefore our assumption that for some X, AX = B, is false.
Consequently there is no X for which AX = B, that is, the given system of equations has no solution.
Remark. A system of equations having no solution is said to be inconsistent.
Example 4. Solve the system of linear equations
x – 2y + 3z = 6,
3x + y – 4z = –7,
5x – 3y + 2z = 5.
1 −2 3 6 x
Solution. Let A = 3 1 −4 , B = −7 , and assume that there exists a matrix X = y , such
5 −3 2 z
5
that AX = B.
1 −2 3 x 6
Then 3 1 −4 y = −7
5 −3 2 z 5
57
1 −2 3 x 6
⇒ 0 7 −13 y = −25 , by R2 → R2 − 3R1 , R3 → R3 − 5 R1.
0 7 13 z −25
1 −2 3 x 6
⇒ 0 7 −13 y = −25 , by R3 → R3 − R2
0 7 0 z 0
⇒ { x − 2 y + 3 z = 6,
7 y − 13 z = − 25
13 25
Form the second of above equations, we have y = z − . Substituting this value of y in the first
7 7
13 25 5 8
equation, we have x = 2y – 3z + 6 = z − − 3z + 6 = z − .
7 7 7 7
Therefore we have
5 8
x= z−
7 7
13 25
y= z− ,
7 7
z = z,
5 8 13 25
i.e., x = k − , y = k − , z = k , where k is any real number. By actual substitution we find
7 7 7 7
that these values satisfy the given equations.
Hence all the solutions of the given system of equations are
5 8 13 25
x= k − , y = k − , z = k,
7 7 7 7
where k is any real number
Example 5. Solve the following system of linear equations :
2x – 3y + z = 0,
x + 2y – 3z = 0,
4x – y –2z = 0.
2 −3 1 x
Solution. Let 1 2 −3 , and assume that there exists X = y , such that AX = O, where O
4 −1 −2 z
stand for the zero-matrix of type 3 × 1.
2 −3 1 x
Then 1 2 −3 y = O
4 −1 −2 z
1 2 −3 x
⇒ 2 −3 1 y = O, by R1 ↔ R2
4 −1 −2 z
58
1 2 −3 x
⇒ 0 −7 7 y = O, by R ↔ R – 2R , R → R – 4R
0 −9 10 z 2 2 1 3 3 1
1 2 −3 x
⇒ 0 −7 7 y = O, R →
0 −9 10 z 2
1 2 −3 x
⇒ 0 −7 7 y = O, R → R + 9R
0 −9 10 z 3 3 2
⇒ x + 2y – 3z = 0
y – z =0
z =0
⇒ x = 0, y = 0, z = 0.
Thus we find that if the given system possesses a solution, it must be given by
x = 0, y = 0, z = 0.
Also, by actual substitution it follows that x = y = z = 0 is a solution.
Hence x = y = z = 0 = is the only solution of the given system.
Example 6. Solve the following system of equations :
x + 3y + 4y + 7w = 0,
2x + 4y + 5z + 8w = 0,
3x + y + 3z + 3w = 0,
x
1 3 4 7 y
Solution. Let A = 2 4 5 8 and assume that there exists a matrix X = such that AX = O,
3 1 2 3 z
w
where O stands for the zero-matrix of type 4 × 1.
x
1 3 4 7
Then 2 4 5 8 y = O
3 1 2 3 z
w
x
1 3 4 7
⇒ 0 −2 − 3 − 6 y = O, by R → R – 2R , R → R – 3R
0 −8 −10 −18 z 2 2 1 3 3 1
w
x
1 3 4 7
⇒ 0 −2 −3 −6 y = 0, R → R – 4R
0 0 2 6 z 3 3 2
w
⇒ x + 3y – + 4z + 7w = 0
–2y – 3z – 6w = 0
2z + 6w = 0
59
3 1
⇒ z = −3w, y = w, x = w
2 2
1 3
⇒ x = k , y = k , z = − 3k , w = k ,
2 2
where k is any real number. Thus we find that if the given system possesses a solution it must be
given by
1 3
x = k , y = k , z = − 3k , w = k , where k is any real number.
2 2
By actual substitution we find that (1) is indeed a solution of the given system.
1 3
Hence all the solutions of the given system of equations are given by x = k, y = k,
2 2
z = − 3k , w = k , where k is any real number.
Example 7. Solve the following system of equations :
x – 3y + 2z = 0,
7x – 21y + 14z = 0,
–3x + 9y – 6z = 0.
1 −3 2 x
Solution. Let A = 7 −21 17 and assume that there exists a matrix X = y , such that AX = O,
−3 9 −6 z
where O is the zero-metrix of type 3 × 1.
1 −3 2 x
Then 7 −21 14 y = O
−3 9 −6 z
1 −3 2 x
⇒ 0 0 0 y = 0, by R → R – 7R , R → R + R
0 0 0 z 2 2 2 3 3 1
⇒ x – 3y + 2z = 0
⇒ x = 3y – 2z, y = y, z = z
⇒ x = 3k1 – 2k2, y = k1, z = k2, ... (1)
where k1, k2 are any real numbers.
Thus we find that if there is a solution of the given system, it must be (1).
Again, by actually substituting x = 3k1 – 2k2, y = k1, z = k2 in the given system of equations, we find
that
x – 3y + 2z = (3k1 – 2k2) – k1 + 2k2 = 0,
7x – 21y + 14z = 7(3k1 – 2k2) – 21k1 + 12 k2 = 0,
–3x + 9y – 6z = – 3(3k1 – 2k2) + 9k1 – 6k2 = 0,
so that (1) is indeed a solution. Therefore we conclude that
x = 3k1 – 2k2, y = k1, 3 = k2, where k1, k2 are real numbers, are all the solutions of the given system of
equations.
4. Application of Rank of a Matrix to a System of Linear Equations
60
Let us consider a system of m linear equations in n variables:
AX = B,
where A is an m × n matrix, B is an m × 1 matrix, and X is an n × 1 matrix. It can be shown that the above
system of equtions is consistent if and only if the matrices A and [A, B] have the same rank. Further more,
if the common rank be r, the general solution contains n–r arbitrary constants.
We shall not prove the about result, but will contnet ourselves with the followig illustrative
examples.
Example 8. Show, without actually solving, that the following system of equations is inconsistent.
x – y + z = – 1,
2x + y – 4z = –1,
6x – 7y + 8z = 7.
Solution. The given system of equations can be written as
AX = B
1 −1 1 x −1
where A = 2 1 −4 , X = y , B = 1 .
6 −7 8 z 7
Let us find the ranks of the matrices A and [A, B]
1 −1 1
A = 2 1 −4
6 −7 8
1 −1 1
~ 0 3 −6 , by R 2 → R2 − 2 R1 , R3 → R3 − 6 R1
0 −1 2
1 −1 1 1
~ 0 1 −2 , by R2 → R2
0 −1 −2 3
1 −1 1
~ 0 1 −2 , by R3 → R3 + R2
0 0 0
1 −1 1
Since 0 1 −2 , is of rank 2,
0 0 0
therefore ρ(A) = 2
1 −1 1 −1
Also, [A, B] = 2 1 −4 −1
6 −7 8 7
1 −1 1 −1
~ 0 3 −6 1 , by R 2 → R2 − 2 R1 , R3 → R3 − 6 R1
0 −1 2 13
1 −1 1 −1 1
~ 0 3 −6 1 , by R3 → R3 + R2
3
0 0 0 40
3
61
1 −1 1 −1 3
~ 0 3 −6 1 , by R3 → R3
0 0 0 1 40
1 −1 1 0
~ 0 3 −6 1 , by R1 → R1 + R3 , R2 → R2 − R3
0 0 0 1
1 0 0 0
~ 0 3 −6 1 , by C2 → C2 + C1 , C3 → C3 − C1
0 0 0 1
1 0 0 0 1 −1
~ 0 1 1 0 , by C2 → C2 , C3 → C3
0 0 0 1 3 6
1 0 0 0
~ 0 1 0 0 , by C3 → C3 − C2
0 0 0 1
showing that the rank of arugmented matrix is 3.
Since the ranks of the co-efficient matrix and augmented matrix are not equal, the given system of
equations is inconsistent.
3x + 4y – 6 z + w = 7
x – 2y + 3z + 2w = –1
x – 3y + 4z – w = –2
5x – y + z – 2w = 4.
Solution.
A B
3 4 −6 1 : 7
1 −2 3 −2 : −1
?= 1 −3 4 −1 : −2
5 −1 1 −2 : 4
1 −2 3 −2 : −1
3 4 −6 1 : 7
1 , by R1 ↔ R2
−1 : −2
?~
−3 4
5 −1 1 −2 : 4
1 −2 3 −2 : −1
0 10 −15 7 : 10
0 , by R2 → R2 − 3R1 , R3 → R3 − R1 , R4 → R4 − 5 R1
1 : −1
?~
−1 1
0 9 −14 8 : 9
1 −2 3 −2 : −1
0 10 −15 7 : 10
1 17 1 9 2
?~ 0 0 − : 0 , R3 → R3 + R2 , R4 → R4 − R
2 10 10 10
1 17
0 0 − : 0
2 10
62
1 −2 3 −2 : −1
0 10 −15 7 : 10
?~ , by R → R − R .
0 0 − 1 17 : 0 4 4 3
2 10
0 0 : 0
0 0
From the above we find that the ranks of A and [A, B] are both equal to 3. Since the co-efficient
matrix and augment both have the same rank, therefore the given system of equations is consistent.
Remark. Observe that in the above example we have computed the ranks of A and [A, B]
simultaneously. This has meant considerable simplification in workings.
Exericise
1. Does the following system of equations possess a unique common solution?
x + 2y + 3z = 6,
2x + 4y + z = 7,
3x + 2y + 9z = 10.
It so, find it.
2. Show that x = y = z = 0 is the only common solution of the following system of equations:
2x + 3y + 4z = 0,
x – 2y – 3z = 0,
5x + y – 8z = 0.
3. Solve the system of equations :
x + y + z = 3,
3x – 5y + 2z = 8,
5x – 3y + 4z = 14.
4. Solve the followng system of equations :
x – 3y + 2z = 0,
7x – 21y + 14z = 0,
–3x + 9y – 6z = 0.
5. Which of the following system of equations are consistent ?
(a) x – 4y + 7z = 8,
3x + 8y – 2z = 6,
7x – 8y + 26z = 31.
(b) x – y + 2z = 4,
3x + y + 4z = 6,
x + y + z = 1.
63
LESSON 6
THE CHARACTERISTIC EQUATION OF A MATRIX
1. Introduction
In this lesson we shall study the important concept of characteristic equation of a square matrix.
The characterstic equation of a square matrix has many applications. It can be used to compute
power of a square matrix, and inverse of a non-singular matrix. It is also useful in simplifying matrix
polynomials.
We shall begin with some definitions.
Definition. Let A = [Aij] b an n × n matrix. The matrix A – λI, where λ is a scalar and I is the n-
rowed unit matrix, ims called the characteristic matrix of A. The determinant |A – λΙ| is called the
characteristic polynomial of A. The equation |A – λI| = 0 is called the characteristic equation of A and its
roots are called the characteristic roots (or latent roots or eigen values) of A.
Remark. It can be easily seen that the diagonad elements of A are of the first degree in λ and the
off-diagonal elements are independent of λ. |A – λI| is, therefore a polynomial of degree in λ, and the co-
efficient of λn is (–1)n. It follows that |A – λI| = 0 is of degreen n is λ. By the fundamental theorem of
algebra, every polynomial equation of degree n has exactly n roots. Therefore it follows that every square
matrix of order n has exactly n characteristic roots.
Example 1. Find the characteristic roots of the matrix.
1 −1 4
A = 0 3 7
0 0 5
Solution.
1− λ −1 4
| A − λI| = 0 3−λ 7 = (1 − λ )(3 − λ )(5 − λ )
0 0 5−λ
The roots of |A – λl| = 0 are, therefore, 1, 3, and 5.
Hence the characteristic roots of A are 1, 3, and 5.
Example 2. Show that the matrices A and A′ have the same characteristic roots.
Solution. Le A = [aij] be an n × n matrix. The characteristic equation of A is
|A – λI| = 0 (i)
Also the characteristic equation of At is
|At – λI| = 0,
which is the same as
|(A – λI)|t = 0 (ii)
Since the value of a determinant remains unaltered by taking its transpose, therefore
|A – λI| = |(A – λI)t|
Consequently
64
|A – λI| = 0, ⇔ |(A – λI)|t = 0 (iii)
i.e., |A – λI| = 0 and |(A – λI)t| = 0 are the same equation. Since |A – λI| = 0 is the characteristic
equation of A, and |(A – λI)t| = 0 is the characteristic equation of At, it follows that A and At have the same
characteristic roots.
Example 3. Show that the characteristic roots of a triangular matrix are just the diagonal elements
of the matrix.
a11 a12 … … a1n
0 a22 … … a2 n
0 0 a33 … a3n
Solution. Let A = … … … … … be an n × n triangular matrix.
… … … … …
… … … … …
0 0 0 … ann
The characteristic equation of A is
a11 − λ a12 ⋯ a1n
0 a22 − λ ⋯ a2 n
. .
=0
. .
. .
0 0 ann − 1
i.e., (a11 – λ) (a22 – λ) … (ann – λ) = 0. (i)
The roots of (i) are given by
λ = a11, a22, a33, …, and ann which are just the diagonal elements of A.
Hence the charactristic roots of A are just the diagonal elements of A.
Example 4. Show that if λ1 λ2, … λn be the charactertistic roots. of A. then kλ1, kλ2, …, kλn are the
characteristic roots of kA.
Solution. Let A = [aij] be an n-rowed square matrix. The characteristic equation of A is
a11 − λ a12 ⋯ a1n
a21 a22 − λ ⋯ a2 n
. .
=0
. .
. .
an1 an 2 ann − λ
Let the roots of (i) be λ1, λ2, …, λn. to obtain the equation whose roots are kλ1, kλ2, … kλn, we have
to substitute λ* = kλ in (i). The resulting equation in λ* will have roots kλ1, kλ2, …, kλn,
By the transformation λ* = kλ, equation (i) is transformed to
65
λ*
a11 − a12 … a1n
k
λ*
a21 a22 − … a2 n
k
. . .
=0
. . .
. . .
. . .
λ*
an1 an2 ann −
k
ka11 − λ * ka12 … ka1n
n . . .
1
⇔ . . . =0
k
. . .
kan1 kan 2 … kamn − λ *
ka11 − λ * ka12 … kann − λ *
. .
⇔ . ka22 − λ ⋯ . =0
. .
kan1 kan 2 … kann − λ *
⇔ kA − λ* I =0 ...(ii)
Since (ii) is the characteristic equation of kA, therefore it follows that the characteristics roots of (ii)
are k times characteristics roots of (i), i.e., the characteristics roots of kA are kλ1, kλ2, ... kλn.
Example 5. If A and P be square matrices of the same type and if P be invertible, show that the
matrices A and P–1 AP have the same characteristic roots.
Solution. The characteristics matrix of P–1 AP is P–1 AP – λI.
Now P–1 AP – λI = P −1 ( A − P(λI ) P −1 ) P,
= P–1 (A – λI) P, ... (i)
since P (λI) P–1 = (λI) P P–1 = (λI) I = λI.
The characteristic equation of P–1 AP is | P–1 AP – λI) | = 0. Using (i) we find that this equation is
the same is
| P–1 (A – λI) P | = 0,
⇔ | P–1 | | A – λI | | P | = 0, since the determination of the product of two or more matrices
is equal to the product of the determines of the matrices.
⇔ | P–1 | | P | | A – λI | = 0, since | P–1 |, | A – λI |, | P | commute
⇔ | P–1 P | | A – λI | = 0, since the determinant of a product is equal to the product of
determinants
66
⇔ | I | | A – λI |= 0, since P–1 P = 1
⇔ | A – λI | = 0, since | I | = 1
Since |P–1 A P – λ I | = 0 ⇔ |A – λ I | = 0, if follows that the characteristic roots of P–1AP are the
same as the characteristic roots of A.
Exercise
1. Find the characteristics roots of each of the following matrices:
1 2 3 1 −1 −1
(i) 0 −4 2 (ii) 1 −1 0 .
0 0 7 1 0 −1
2. Show that the matrices.
o g f o f h o h g
g o h , f o g , h o f
f h o h g o g f o
have the same characteristic equation.
3. Show that the matrices
a b c b c a c a b
b c a , c a b , a b c
c a b a b c b c a
4. Show that the characteristics roots of A* are the conjugates of the characteristics roots of A.
5. Show that 0 is a characteristics root of a matrix if and only if the matrix is singular.
6. If A and B are n-rowed square matrices and if A be invertible, show that the matrices A–1B and B
A–1 have the same characteristic roots.
2. Cayley-Hamilton Theorem
We shall now establish the relation between a square matrix and its characteristics equation. The
relation, which is known after it discovers Arthur Cayley and W.R. Hamilton is known as Cayley-
Hamilton Theorem. It was established by Hamilton in 1953 for a special class of matrices and was stated
(without proof) by Caylay in 1858.
Theorem. Let | A – λI | = (–1)n = {λ n + a1λ n − 1 + a2λ n − 2 + K + an } be the characteristics polynomial
of an n + n matrix A. Then the matrix equation X n + a1 X n − 1 + a2 X n − 2 + K + an I = O is satisfied by X =
A.
Proof. The elements of A – λI are at most of the first degree in λ. (In fact if we write the matrix in
full, we find that the diagonal elements are of the first degree in λ, and all the off-diagonal elements are
independent of λ). The elements of adj (A – λI) are (n – 1) × (n – 1) determinants, whose elements are at
most of the first degree in λ. Therefore adj. (A – λI) is an n × n matrix whose elements are atmost of
degree n – 1 in λ. The matrix adj (A – λI) can, therefore, be written as a matrix polynomial of degree n – 1
in λ.
Let adj (A – λI) = B0λn–1 + B1λn–2 + ... + Bn–1, where B0, B1, ... Bn–1 are n × n matrices whose
elements are functions of aij’s (the elements of A).
Since (A – λI) adj (A – λI) = | A – λI | I,.
Therefore(A – λI) (B0λn–1 + B1λn–2 + ... + Bn–1) = (–1)n
67
Comparing the co-efficients of like powre of λ on both sides, we hve
– IB0 = (–1)n I,
AB0 – IB1 = (–1)n a1I,
AB1 – IB2 = (–1)n a2I,
....................................
ABn – 1 = (–1)n an I.
Premultiplying the above equations by An, An–1 ... I, respectively and adding the corresponding sides
of the resulting equation, we have
O = (–1)n [An + a1An–1 + ... + anI].
Thus An + a1An–1 + ... + a2An –2 + ... + anI = O, and this proves the theorem.
Remark. Calyey-Hamilton theorem is often expressed by saying that every square matrix satisfies
its characteristics equation.
We shall now illustrate the Cayley-Hamilton therorem with the help of an example.
1 0 2
Example 6. Verify that the matrix A = 6 1 2 satisfies its characteristics equation.
1 −2 0
Soluiton. For the given matrix A,
1 − λ 0 2
| A – λI | = 6 1 − λ 2
1 −2 −λ
= (1 – λ) [{– λ (1 – λ) + 4} + 2 {– 12 – (1 – λ)}
= (1 – λ) (λ2 – λ + 4) + 2 (λ – 13)
= – λ3 + 2λ2 – 3λ – 22
The characteristics equation of A is therefore
– λ3 + 2λ2 – 3λ – 22 = 0
or – λ3 + 2λ2 + 3λ + 22 = 0 ...(i)
6 0 2 6 0 2
Now A2 = 6 1 2 6 1 2
1 −2 0 1 − 0
3 −4 2
= 14 −3 14 ,
−11 −2 −2
3 −4 2 1 0 2
A3 = 14 −3 14 6 1 2
−11 −2 −2 1 −2 0
−19 −8 −2
= 10 −31 22
−25 2 −26
Substituting A for λ in the left-hand side of (i) we have
68
A3 – 2A2 + 3A + 22 I
−19 −8 −2 3 −4 2 1 0 2 1 0 0
= 10 −31 22 − 2 14 −3 14 + 3 6 1 2 + 22 0 1 0
−25 2 −26
−11 −2 −2 1 −2 0 0 0 1
0 0 0
= 0 0 0 = 0.
0 0 0
Thus we have
A3 – A2 + 3A + 22I = 0, showing that the matrix A satisfies its characteristic equation.
3. Application of Cayley-Hamilton Theorem to Compute Powers and Inverse of a given
Square Matrix
Cayley-Hamilton theorem can be used to compute powers of a square matrix and invers of a non-
singular square matrix.
Let the characteristic equation of an n-rowed square matrix A be
λn + a1λn – 1 + a2λn – 2 + ... + an = 0. ...(i)
By Cayley-Hamilton theorem the matrix A satisfies (i).
Therefore we have form (i),
Am + a1An – 1 + a2An – 2 + ... + anAm – n = 0. ...(ii)
Substituting m = n + 1, n + 2, ... n + m in (iii) we have the relations
An + 1 + a1An + a2An – 1 + ... + anA = 0. ...(iii)
An + 2 + a1 An + 1 + a2 An + ... + an A2 = 0.
.....................................................................................
An + k + a1An + k – 1 + a2An + k – 2 + ... + anAk = 0.
By substituting the values of A, A2, ... An in the first of the above equations, we get the value of An +
1. Similarly by substituting the values A2, A3, ... An + 1 in the second of the above equations we can get the
value of An + 2, and so on.
If the matrix A is non-singular, then an ≠ 0.
1 −1
But multiplying (ii) throughout by A we have
an
1 n − 1 a1 n−2
a A + A + K + A−1 = 0
n an
1 a an − 1
or A–1 = − An − 1 + − 1 An − 2 + K − I ...(iv)
an an an
By substituting the values of I, A, A2, ... An – 1 in the right-hand side of (iv) and simplifying we can
get the value of A–1.
Example 7. Find the characteristic equation of the matrix
69
1 2 3
A = 2 3 4,
1 0 −1
and hence compute its cube.
Solution. The characteristic equation of the matrix A is
1−l 2 3
? 2 3−λ 4 = 0.
1 0 −1 − λ
⇔ (1 – λ) (3 – λ) (–1 – λ) – {2 (–1 – λ) – 4} + 3 (–3 + λ) = 0
⇔ λ3 – 3λ2 + 8λ = 0 ...(i)
By Cayley-Hamilton theorem by substituting A for λ in (i), we have
? A3 –3A2 – 8A = 0,
so that A3 = 3A2 + 8A. ...(ii)
1 2 3 1 2 3
Now A2 = 2 3 4 2 3 4
1 0 −1 1 0 −1
8 8 8
? = 12 13 14 ...(iii)
0 2 4
Substituting the values of A2 and A in (ii) we have
8 8 8 1 2 3
A3 = 3 12 13 14 + 8 2 3 4
0 2 4 1 0 −1
32 40 48
= 52 63 74 .
8 6 4
Example 8. Find the characteristics equation of the matrix
1 0 2
A = 0 1 2
1 2 0
and hence compute is inverse.
Solution. The characteristic equation of the matrix A is
1− λ 0 2
0 1− λ 2 =0
1 2 −λ
⇔ (1 – λ) {–λ(1 – λ) – 4} + 2 (–1 + λ) = 0
⇔ λ3 – 2λ2 – 5λ + 6 = 0. ...(i)
By Cayley-Hamilton theorem, A satisfies (i) so that we must have
A3 – 2A2 – 5A + 6I = 0. ...(ii)
Multiplying (ii) throughout by A–1 we have
70
A2 – 2A – 5I + 6A–1 = 0,
so that
6A–1 = –A2 + 2A + 5I
2
1 0 2 1 0 2 1 0 0
= − 0 1 2 + 2 0 1 2 + 5 0 1 0
1 2 0 1 2 0 0 0 1
3 4 2 1 0 2 1 0 0
= − 2 5 2 + 2 0 1 2 + 5 0 1 0
1 2 6 2 0
1 0 0 1
4 −4 2
= −2 2 2
1 2 −1
4 −4 2
1
whence A–1 = −2 2 2 ,
6 1 2 −1
2 2 1
3 −3 3
1 1 1
= −
3 3 3
1 1 1
−
6 3 6
Verification. By actual multiplication it can be easily seen that AA–1 = I.
4. Application of Cayley-Hamilton Theorem : Simplification of Matrix Polynomials
By using Cayley-Hamilton theorem of an n-rowed square matrix. A, a polynomial in A of degree
greater than n can be reduced to a polynomial in A of degree less than n, i.e., a polynomial of degree
greater than two in a 2 × 2 matrix A can be reduced to a linear polynomial in A, a polynomial of degree
greater than three in a 3 × 3 matrix A can be reduced to a quadratic polynomial in A, and so on. The
following example will illustrate the method.
3 1 5 4 2
Example 9. If A = , express 2A – 3A + A – 4I as a linear polunomial in A.
−1 2
Solution. The characteristic equation of A is
3−λ 1
= 0,
−1 2 − λ
i.e., (3 – λ) (2 – λ) + 1 = 0,
i.e., λ2 – 5λ + 7 = 0.
The given polynomial is f (A), where
f (λ) = 2λ5 – 3λ4 + λ2 – 4.
By division alogrithm for polynomials we can write
f(λ) = (λ2 – 5λ + 7) (2λ3 + 7λ2 + 21λ + 57) + 138λ – 403
∴ f(A) = (A2 – 5A + 7I) (2A3 + 7A2 + 21A + 57) + (138A – 403 I) ...(i)
Since every square matrix satisfies its characteristic equation, therefore
71
A2 – 5A + 7I = 0.
Substituting from (ii) in (i) it follows that f(A) = 138A – 403 I.
Exercise
1. Verify that the matrix
1 2 1
A = −1 0 3 satisfies its characteristic equation.
2 −1 1
2. Obtain the characteristic equation of the matrix
1 0 2
A = 0 2 1 and hence calculate its inverse.
2 0 6
1 2
3. If A = , express
−1 3
A6 – 4A5 + 8A4 – 12A3 + 14A2 as a linear polynomial in A.
4. Evaluate the matrix polynomial A5 – 27A3 + 65A2 as a 3 × 3 matrix, where
0 0 1
A = 3 1 0
−2 1 4
1 0 0
5. If A = 1 0 1
0 1 0
show that A3 – A = A2 – I.
Deduce that for every integer n ≥ 3, An – An – 2 = A2 – I. Hence otherwise determine A50
72
B.A. (Programme) I Year Discipline Course Mathematics
S.R.S. 1
Marks ..............................%
Roll No. .................................... Signature of the Lecturer......................................
When completed send this
Response sheet no :
School of Open Learning Name.......................................................
University of Delhi. Address ...................................................
5, Cavalry Lanes, ..................................................................
Delhi-11007 (India) Pin Code .................................................
Academic Session 2012–2013
Exercise
1. Show that the set
C2 = {(x1, x2) : x1 ∈ C, x2 ∈ C}
is a vector space over C with respect to co-ordinate wise addition and scalar multiplication.
2. Show that the set of all 2 × 2 matrices over C is a vector space over C with respect ot matrix
addition and multiplication of a matrix by a scalar.
3. Let V = {(a1, a2, a3, a4) : a1, a2, a3, a4 are integers}
Is V a vector space over R with respect to co-ordinatewise addition and scalar multiplication ?
Justity your answer.
4. Let V = {(x1, x2, x3,) : x1, x2, x3, are complex numbers, and x1, x2, = 0}
It V a vector space of over C with respect to co-oridnatewise addition and scaler multiplication ?
Justify your answer.
x o
5. Show that the set of all matrices of the form , where y ∈ C is a vector space over C with
o y
respect to matrix addition and multiplication of a matrix by a scalar.
a b
6. Show that the set of all matrices of the form , where a, b ∈ C is a vector space
−b a
over C. With respect to matrix addition and multiplication of a matrix by a scalar.
73
B.A. (Programme) I Year Discipline Course Mathematics
S.R.S. 2
Marks ..............................%
Roll No. .................................... Signature of the Lecturer......................................
When completed send this
Response sheet no :
School of Open Learning Name.......................................................
University of Delhi. Address ...................................................
5, Cavalry Lanes, ..................................................................
Delhi-11007 (India) Pin Code .................................................
Academic Session 2012–2013
Exercise
1. For each of the following matrices A, verify that At = A:
1 2 −1 −1 6 −7 a h g
2 0 −4 , 6 3 8 , h b f
−1 −4 3 −7 8 −5 g f c
2. For each of the following matrices A, verify that At = – A:
0 i 1 0 3 1 + i 0 −3 2i
−i 0 −2 , −3 0 −4 , 3 0 4
−1 2 0 −1 − i 4 0 −2i −4 0
1 2 −2
1
3. If A = 2 1 2 , verify that A At = At = A = I3.
3 2 −2 −1
4. If A be any square matrix, verify that he matrix A + At is symmetric and the matrix A – At is
skew-symmetric.
5. If A be any square matrix, prove that A + Aθ, AAθ, AθA are all hermitian and A – Aθ is
skew-hermitian.
3 1 −7
6. Express the matrix 2 4 8 as X + Y where X is symmetric and Y is skew-symmetric.
6 −1 2
7. Find remain matrices A and B such that
1 1+ i 3
A+ iB = 4 2 1 − i
−3 2 + 3i i
74
B.A. (Programme) I Year Discipline Course Mathematics
S.R.S. 3(a)
Marks ..............................%
Roll No. .................................... Signature of the Lecturer......................................
When completed send this
Response sheet no :
School of Open Learning Name.......................................................
University of Delhi. Address ...................................................
5, Cavalry Lanes, ..................................................................
Delhi-11007 (India) Pin Code .................................................
Academic Session 2012–2013
Exercise
1. Apply the elementary operation R2 ↔ R3 to the matrix
1 −1 3
4 2 −6
5 8 9
2. Apply the elementary operation C2 → 2C2 to the matrix
−1 3 7 6
5 −1 4 −2
3. Write down the elementary matrix obtained by applying R3 → R3 – 4R1 to I3.
4. Reduce the matrix
1 2 −3 4
3 −1 2 0
2 1 −1 5
to triangular from by applying E-row operations
5. Reduce the matrix
1 −1 −1
4 1 0
8 1 1
to I3 by E-row operations
6. Verify that the E-row operation R1 → R1 – R3 on the matrix AB is equivalent to the same E-row
operation on A, where
1 −1 2 2 3 −5
A = −3 1 4 , B = 1 −2 6 .
0 1 3 3 1 1
75
B.A. (Programme) I Year Discipline Course Mathematics
S.R.S. 3(b)
Marks ..............................%
Roll No. .................................... Signature of the Lecturer......................................
When completed send this
Response sheet no :
School of Open Learning Name.......................................................
University of Delhi. Address ...................................................
5, Cavalry Lanes, ..................................................................
Delhi-11007 (India) Pin Code .................................................
Academic Session 2012–2013
Exercise
1. Find the inverse of the matrix
3 −15 5
−1 6 −2
1 −5 2
by calculating its adjoint.
2. Show that the matrix
1 −2 3
0 −1 4
−2 2 1
possesses an inverse.
1 1 0
3. Given that A = −2 −1 0 , calculate A2 and A–1
0 0 2
1 2 −1 2 1 2
4. If A = −4 −7 4 , B = 2 2 1 , verify that (AB)–1 = B–1A–1.
−4 −9 5 1 2 2
5. By applying elementary row operations find the inverse of the matrix
1 1 2
3 1 −3 .
2 1 −1
76
B.A. (Programme) I Year Discipline Course Mathematics
S.R.S. 4
Marks ..............................%
Roll No. .................................... Signature of the Lecturer......................................
When completed send this
Response sheet no :
School of Open Learning Name.......................................................
University of Delhi. Address ...................................................
5, Cavalry Lanes, ..................................................................
Delhi-11007 (India) Pin Code .................................................
Academic Session 2012–2013
Exercise
1. Find the rank of the matrix
3 −1 2 −6
6 −2 4 −9 .
2. Find the rank of the matrix.
−1 2 3
0 3 4 .
0 0 0
3. Find the rank of the matrix
1 1 1
a b c .
3 3
3
a b c
4. Reduce the matrix
1 −1 2 −3
4 1 0 2
0 3 1 4 .
0 1 0 2
to triangular form and hence find rank.
5. Find the rank of the matrix
12 22 32 42
2
2 32 42 52
.
32 42 52 62
2
4 52 62 7 2
77
B.A. (Programme) I Year Discipline Course Mathematics
S.R.S. 5
Marks ..............................%
Roll No. .................................... Signature of the Lecturer......................................
When completed send this
Response sheet no :
School of Open Learning Name.......................................................
University of Delhi. Address ...................................................
5, Cavalry Lanes, ..................................................................
Delhi-11007 (India) Pin Code .................................................
Academic Session 2012–2013
Exercise
1. Does the following system of equations possess a unique common solution?
x + 2y + 3z = 6,
2x + 4y + z = 7,
3x + 2y + 9z = 10.
It so, find it.
2. Show that x = y = z = 0 is the only common solution of the following system of equations:
2x + 3y + 4z = 0,
x – 2y – 3z = 0,
3x + y – 8z = 0.
3. Solve the system of equations:
x + y + z = 3,
3x – 5y + 2z = 8,
5x – 3y + 4z = 14.
4. Solve the following system of equations:
x – 3y + 2z = 0,
7x – 21y + 14z = 0,
–3x + 9y – 6z = 0.
5. Which of the following system of equations are consistent?
( a) x – 4y + 7z = 8,
3x + 8y – 2z = 6,
7x – 8y + 26z = 31.
( b) x – y + 2z = 4,
78
3x + y + 4z = 6,
x+y+z= 1.
B.A. (Programme) I Year Discipline Course Mathematics
S.R.S. 6(a)
Marks ..............................%
Roll No. .................................... Signature of the Lecturer......................................
When completed send this
Response sheet no :
School of Open Learning Name.......................................................
University of Delhi. Address ...................................................
5, Cavalry Lanes, ..................................................................
Delhi-11007 (India) Pin Code .................................................
Academic Session 2012–2013
Exercise
1. Find the characteristics roots of each of the following matrices:
1 2 3 1 −1 −1
(i) 0 −4 2 (ii) 1 −1 0 .
0 0 7 1 0 −1
2. Show that the matrices.
o g f o f h o
g h
g o h , f o g , h
f o
f h o h g o g
o f
have the same characteristics equation.
3. Show that the matrices
a b c b c a c a b
b c a , c a b , a b c
c a b a b c b c a
4. Show that the characteristics roots of A* are the conjugates of the characteristics roots of A.
5. Show that 0 is a characteristics root of a matrix if and only if the matrix is singular.
6. If A and B are n-rowed square matrices and if A be invertible, show that the matrices A–1B and B
A–1 have the same characteristics roots.
79
B.A. (Programme) I Year Discipline Course Mathematics
S.R.S. 6(b)
Marks ..............................%
Roll No. .................................... Signature of the Lecturer......................................
When completed send this
Response sheet no :
School of Open Learning Name.......................................................
University of Delhi. Address ...................................................
5, Cavalry Lanes, ..................................................................
Delhi-11007 (India) Pin Code .................................................
Academic Session 2012–2013
Exercise
1. Verify that the matrix
1 2 1
A = −1 0 3 satisfies its characteristic equation.
2 −1 1
2. Obtain the characteristic equation of the matrix
1 0 2
A = 0 2 1 and hence calculate its inverse.
2 0 6
1 2
3. If A = , express
−1 3
A6 – 4A5 + 8A4 – 12A3 + 14A2 as a linear polynomial in A.
4. Evaluate the matrix polynomial A5 – 27A3 + 65A2 as a 3 × 3 matrix, where
0 0 1
A = 3 1 0
−2 1 4
1 0 0
5. If A = 1 0 1
0 1 0
show that A – A = A2 – I.
3
Deduce that for every integer n ≤ 3, An – An – 2 = A2 – I. Hence otherwise determine A50.
80
You might also like
- 1 Vector Spaces: Closure Property Associative Property Identity PropertyNo ratings yet1 Vector Spaces: Closure Property Associative Property Identity Property31 pages
- Real Vector Spaces and Subspaces (Final)No ratings yetReal Vector Spaces and Subspaces (Final)15 pages
- Course 3: 15.10.2020 Chapter 2 Vector SpacesNo ratings yetCourse 3: 15.10.2020 Chapter 2 Vector Spaces5 pages
- 1.2 Vector Space : Introduction To VectorsNo ratings yet1.2 Vector Space : Introduction To Vectors15 pages
- UGSemsterSyllabus Maths 5Sem517Maths English LINEARALGEBRANo ratings yetUGSemsterSyllabus Maths 5Sem517Maths English LINEARALGEBRA113 pages
- 4.1 - de Nition of Vector Spaces - Mathematics LibreTextsNo ratings yet4.1 - de Nition of Vector Spaces - Mathematics LibreTexts2 pages
- Lecture 9 Row and Column Space 25112024 063451pmNo ratings yetLecture 9 Row and Column Space 25112024 063451pm34 pages
- (A-B) Dictionary of Algebra, Arithmetic, and TrigonometryNo ratings yet(A-B) Dictionary of Algebra, Arithmetic, and Trigonometry20 pages
- Notes 2.1 Day 1 Average and Instantaneous Rate of ChangeNo ratings yetNotes 2.1 Day 1 Average and Instantaneous Rate of Change6 pages
- Derivatives of Polynomials and Exponential FunctionsNo ratings yetDerivatives of Polynomials and Exponential Functions54 pages
- Analytical Tomographic Image Recon Methods - Ch3No ratings yetAnalytical Tomographic Image Recon Methods - Ch352 pages
- Immediate Download Fundamentals of Linear Algebra Textbooks in Mathematics 1st Edition J.S. Chahal Ebooks 2024100% (3)Immediate Download Fundamentals of Linear Algebra Textbooks in Mathematics 1st Edition J.S. Chahal Ebooks 202462 pages
- EEE 201 Engineering Mathematics Prof. Dr. Saffet AYASUN Inst.: Dr. Bülent DağNo ratings yetEEE 201 Engineering Mathematics Prof. Dr. Saffet AYASUN Inst.: Dr. Bülent Dağ36 pages
- MTH6140 Linear Algebra II: Notes 4 1st November 2010No ratings yetMTH6140 Linear Algebra II: Notes 4 1st November 201010 pages
- Unit Iii Applications of Partial Differential Equations (Miscellaneous Problems)No ratings yetUnit Iii Applications of Partial Differential Equations (Miscellaneous Problems)19 pages
- Relation Between The Beta and Gamma FunctionsNo ratings yetRelation Between The Beta and Gamma Functions13 pages
- Wavelets On The Interval and Fast Wavelet Transforms: Albert Cohen, Ingrid Daubechies, Pierre VialNo ratings yetWavelets On The Interval and Fast Wavelet Transforms: Albert Cohen, Ingrid Daubechies, Pierre Vial29 pages
- Notes For Signals and Systems 11.5 Introduction To Laplace Transform Analysis of LTI SystemsNo ratings yetNotes For Signals and Systems 11.5 Introduction To Laplace Transform Analysis of LTI Systems3 pages
- 1954 - Application of The Rayleigh Ritz Method To Variational Problem by IndritzNo ratings yet1954 - Application of The Rayleigh Ritz Method To Variational Problem by Indritz37 pages
- Derivatives: Department of Mathematics Uniglobe Higher Secondary School Kamaladi Ganeshthan, Kathmandu, Nepal100% (1)Derivatives: Department of Mathematics Uniglobe Higher Secondary School Kamaladi Ganeshthan, Kathmandu, Nepal14 pages
- Dead Beat Observer Synthesis: Maria Elena Valcher, Jan C. WillemsNo ratings yetDead Beat Observer Synthesis: Maria Elena Valcher, Jan C. Willems8 pages
- 1 Vector Spaces: Closure Property Associative Property Identity Property1 Vector Spaces: Closure Property Associative Property Identity Property
- UGSemsterSyllabus Maths 5Sem517Maths English LINEARALGEBRAUGSemsterSyllabus Maths 5Sem517Maths English LINEARALGEBRA
- 4.1 - de Nition of Vector Spaces - Mathematics LibreTexts4.1 - de Nition of Vector Spaces - Mathematics LibreTexts
- An Introduction to Linear Algebra and TensorsFrom EverandAn Introduction to Linear Algebra and Tensors
- (A-B) Dictionary of Algebra, Arithmetic, and Trigonometry(A-B) Dictionary of Algebra, Arithmetic, and Trigonometry
- Notes 2.1 Day 1 Average and Instantaneous Rate of ChangeNotes 2.1 Day 1 Average and Instantaneous Rate of Change
- Derivatives of Polynomials and Exponential FunctionsDerivatives of Polynomials and Exponential Functions
- Immediate Download Fundamentals of Linear Algebra Textbooks in Mathematics 1st Edition J.S. Chahal Ebooks 2024Immediate Download Fundamentals of Linear Algebra Textbooks in Mathematics 1st Edition J.S. Chahal Ebooks 2024
- EEE 201 Engineering Mathematics Prof. Dr. Saffet AYASUN Inst.: Dr. Bülent DağEEE 201 Engineering Mathematics Prof. Dr. Saffet AYASUN Inst.: Dr. Bülent Dağ
- MTH6140 Linear Algebra II: Notes 4 1st November 2010MTH6140 Linear Algebra II: Notes 4 1st November 2010
- Unit Iii Applications of Partial Differential Equations (Miscellaneous Problems)Unit Iii Applications of Partial Differential Equations (Miscellaneous Problems)
- Wavelets On The Interval and Fast Wavelet Transforms: Albert Cohen, Ingrid Daubechies, Pierre VialWavelets On The Interval and Fast Wavelet Transforms: Albert Cohen, Ingrid Daubechies, Pierre Vial
- Notes For Signals and Systems 11.5 Introduction To Laplace Transform Analysis of LTI SystemsNotes For Signals and Systems 11.5 Introduction To Laplace Transform Analysis of LTI Systems
- 1954 - Application of The Rayleigh Ritz Method To Variational Problem by Indritz1954 - Application of The Rayleigh Ritz Method To Variational Problem by Indritz
- Derivatives: Department of Mathematics Uniglobe Higher Secondary School Kamaladi Ganeshthan, Kathmandu, NepalDerivatives: Department of Mathematics Uniglobe Higher Secondary School Kamaladi Ganeshthan, Kathmandu, Nepal
- Dead Beat Observer Synthesis: Maria Elena Valcher, Jan C. WillemsDead Beat Observer Synthesis: Maria Elena Valcher, Jan C. Willems