1

Chapter 1 Introduction to Matrices

1.1 Definitions and Notation

A matrix is a rectangular array of numbers such as

1 2 2 2
0  1 i
 1 2 2  3 −1 2   
(1.1.1)
0 0 1 2
, 5 16 −6  , or  i −1 .
  0 1
 
0 0 0 1

In general if m and n are positive integers and aij is a number for all integers i and j with
1  i  m and 1  j  n , then

 a11 a12 a13  a1n 

a a2 n 
 21 a22 a23 
(1.1.2)  a31 a32 a33  a3n 
 
  
 am1 am 2 am3  amn 

is called an m  n matrix. We say that the size of this matrix is m  n ; it has m rows (numbered
from top to bottom) and n columns (numbered from left to right). The matrices in (1.1.1) have
sizes 4  4, 2  3, and 3  2, respectively. If the number of rows equals the number of columns
we say that the matrix is square.

Each number aij in (1.1.2) is called an entry of the matrix, the ij -entry. The first number in the
subscript refers to the row number of the entry and the second number refers to the column
number. Thus, a32 is the entry in row 3 and column 2. The notation  aij  is used to refer to
m n

the matrix in (1.1.2). Or we may denote it simply as  aij  if the size is clear from context. As
another example of this notation, suppose we use b jk  to refer to the first matrix in (1.1.1).
4 4

Then

1 if j = k

b jk = 2 if j  k
0 if j  k

 2

We use upper-case letters to name matrices, such as

1 0 0 
(1.1.3) A = 0 2 0  .
0 0 3

Given a square matrix A =  aij  , the main diagonal consists of the entries aii for
n n
i = 1, 2,  , n. For example, the main diagonal entries of A in (1.1.3) are 1, 2, and 3.

Matrices of size m1 for some m are referred to as column vectors, and matrices of size 1 n
are referred to as row vectors. Column and row vectors are named with lower-case, bold-faced
letters, or, when written by hand, with lower-case letters with arrows above them. For example,

1 
  
b = b = 2
 3 

is a column vector.

We say that the matrices A =  aij  and B = bij  are equal if and only if m = m, n = n,
m n m n

and aij = bij for 1  i  m, 1  j  n. That is, two matrices are equal if and only if they have the
same size and equal corresponding entries.

1.2 Sums and Scalar Multiples

Given A =  aij  and B = bij  we define the sum of these matrices by

m n m n

A + B = aij + bij  .
m n

In other words, we add matrices of the same size by adding their corresponding entries. We say
that the addition is done entrywise (or componentwise). For example,

 −1 1 2   1 2 0  0 3 2 
 2 −3 4  +  −1 2 3 =  1 −1 7  .
     

If matrices are not the same size, they may not be added.
 3

Numbers are referred to as scalars when distinguishing them from matrices and vectors. Given a
scalar r and a matrix A =  aij  , we define the scalar multiple
m n

rA =  raij  .
m n

That is, to multiply a matrix by a scalar we simply multiply each entry by the scalar. For
example,

1 2  3 6 
3 = .
3 4  9 12 

This definition is motivated by the desire to have, for example,

3  aij  =  aij  +  aij  +  aij  = 3aij  .

We define the m  n zero matrix to be the matrix 0m n =  aij  , where aij = 0 for 1  i  m ,
m n
1  j  n . For example,

0 0 
022 =  .
0 0 

Such a matrix serves as an additive identity. We denote a zero matrix by simply 0 if the size is
clear from context.

Given a matrix, A, we define − A = ( −1) A, which serves as an additive inverse. Finally, given
matrices, A and B, of the same size, we define the difference A − B = A + ( − B ) .

1.3 Properties of Sums and Scalar Multiples

Since matrix addition and scalar multiplication are entrywise operations, matrices obey the same
properties that the complex or real number systems do with respect to these operations. In the
following proposition, we use C to denote the set of complex numbers.

Proposition 1.3.1 Let A =  aij  , B = bij  , C = cij  , r C , and s C . Then

m n m n m n

(1) A + B = B + A (commutativity)
(2) ( A + B ) + C = A + ( B + C ) (associativity)
 4

(3) A + 0m n = A (additive identity)

(4) A + ( − A) = 0m n (additive inverses)
(5) r ( sA) = ( rs ) A (associativity)
(6) r ( A + B ) = rA + rB (distributivity)
(7) ( r + s ) A = rA + sA (distributivity)

Proof of (2) ( A + B ) + C = aij + bij  + cij  = ( aij + bij ) + cij  = aij + (bij + cij )
=  aij  + bij + cij  = A + ( B + C )


( )
Proof of (6) r ( A + B ) = r aij  + bij  = r aij + bij  = r ( aij + bij ) = raij + rbij 
 
=  raij  +  rbij  = r  aij  + r bij  = rA + rB.


The proofs of the other parts are similar and left to the reader.

1.4 The Transpose

Suppose A =  aij  . Then we define the transpose of A to be the matrix

m n

AT =  aij  , where aij = a ji .

n m

 1 4
1 2 3
For example, if A =   then A =  2 5 .
T

 4 5 6  3 6 

Note that the ith row of AT contains the same entries (in order) as the ith column of A, and the jth
column of AT contains the same entries (in order) as the jth row of A. Further note that it is
improper to write AT =  a ji  because that notation would imply that in AT the entry a ji is in
n m
the jth row and ith column when it is actually in the ith row and jth column.

Proposition 1.4.1 Suppose A =  aij  , B = bij  , and r C. Then

m n m n

(1) (A )T T
=A
 5

( A + B)
T
(2) = AT + B T

( rA) = r ( AT )
T
(3)

Proof of (2) First note that each side of the equation is an n  m matrix. So, to establish equality
we must show that corresponding entries are equal. Let AT =  aij  , B T = bij  ,

( A + B)
T
= cij  , and AT + B T =  dij  . Since A + B =  aij + bij  , we have cij = a ji + b ji . Also,
dij = aij + bij = a ji + b ji . So, cij = dij and hence ( A + B ) = AT + B T .
T

The proofs of the other parts are similar and left to the reader.

A square matrix A is called symmetric if AT = A and skew-symmetric if AT = − A. For example,

let

 1 5 10   0 1 2
A =  5 2 25 and B =  −1 0 −3 .
 
10 25 3   −2 3 0 

Then A is symmetric and B is skew-symmetric. Note that a symmetric matrix is so named

because its entries are symmetric with respect to its main diagonal.

1.5 Products

Given A =  aij  and B = b jk  we define the product AB by AB = C =  cik m p where
m n n p

n
cik =  aij b jk = ai1b1k + ai 2b2 k + ai 3b3k +  + ainbnk .
j =1

In other words, to find the ik-entry of C we take the entries in the ith row of A and multiply them
by the corresponding entries in the kth column of B, and then sum these products. In order for
this to be possible, the number of entries in each row of A must equal the number of entries in
each column of B, or, in other words, the number of columns of A must equal the number of rows
of B. For example, consider the following product.
 6

a b c
 f   2a + 3d + 4 g + 5 j 2b + 3e + 4h + 5k 2c + 3 f + 4i + 5l 
 2 3 4 5  d e
6 7 8 9  g h =
  i  6a + 7d + 8 g + 9 j 6b + 7e + 8h + 9k 6c + 7 f + 8i + 9l 
 
j k l

A common type of matrix product we will use is the product of an m  n matrix and an n1
column vector. Proposition 1.5.1 gives us another way of viewing such a product.

Proposition 1.5.1 Suppose A =  aij  is a matrix with columns a1 , a2 ,  , an and suppose

m n

b = b1 b2  bn  . Then Ab = b1a1 + b2a2 +  + bnan .

T

Proof Note that

 a11 a12  a1n   b1   a11b1 + a12b2 + + a1nbn 

a a22  a2 n  b2   a21b1 + a22b2 + + a2 nbn 
Ab =  21
= ,
     
    
 am1 am 2  amn  bn   am1b1 + am 2b2 + + amnbn 

and

 a11   a12   a1n   a11b1 + a12b2 +  + a1nbn 

a  a  a   a b + a b + + a b 
b1a1 + b2a 2 +  + bna n = b1  21 
+ b2  22  + + bn  2 n  =  21 1 22 2 2n n 
.
       
       
 am1   am 2   amn   am1b1 + am 2b2 +  + amnbn 


For example,

a b c   2 a  b   c   2a   3b   4c   2a + 3b + 4c 
d e f   3 = 2  d  + 3  e  + 4  f  =  2d  +  3e  +  4 f  =  2d + 3e + 4 f  .
      

 g h i   4  g   h   i   2 g  3h   4i   2 g + 3h + 4i 

If the columns of a matrix A are a1 , a2 ,  , an then we may denote matrix A by

a1 a2  a n .
 7

Proposition 1.5.2 Suppose A =  aij  and B = b jk  with columns b1 , b2 ,  , b p . Then

m n n p

AB =  Ab1 Ab 2  Ab p  .

Proof Note that the kth column of AB is

 a11b1k + a12b2 k +  + a1nbnk 

 a b + a b + + a b 
 21 1k 22 2 k 2 n nk 
,
  
 
 am1b1k + am 2b2 k +  + amnbnk 

found by multiplying each row of A by the kth column of B. But this is exactly the result of the
product Abk. So each column of AB equals the corresponding column of  Ab1 Ab 2  Ab p 
and hence AB =  Ab1 Ab 2  Ab p  .


Proposition 1.5.2 shows that matrix multiplication is a natural generalization of multiplication of

a matrix and a column vector.

1.6 Properties of Products

While the additive structure of matrices is analogous to that of the complex or real number
systems, matrices do not obey all multiplicative properties of those number systems. For
example,

1 2   5 6  19 22   5 6  1 2   23 34 
3 4  7 8  =  43 50  but 7 8  3 4  =  31 46 .
         

Hence, matrix multiplication is not commutative. In fact, it may be the case that only one of the
two products AB and BA is defined. For example, if A is a 3  2 matrix and B is a 2  4 matrix,
then AB is defined but BA is not. There are other differences between matrix multiplication and
the multiplication of complex or real numbers; you will explore some of those differences in the
exercises.
 8

Proposition 1.6.1 Suppose matrices A, B, and C have sizes for which the indicated operations
are defined, and r C. Then
(1) ( AB ) C = A ( BC ) (associativity)
(2) A ( B + C ) = AB + AC (distributivity)
(3) ( B + C ) A = BA + CA (distributivity)
(4) r ( AB ) = ( rA) B = A ( rB )
( AB ) = B T AT
T
(5)

Proof of (1) Suppose A =  aij  , B = b jk  , and C =  ckr  p  q . First note that both
m n n p
n
( AB ) C and A ( BC ) are m  q matrices. Then note the ik-entry of AB is  aij b jk and so the
j =1

p
 n 
ir-entry of ( AB ) C is    aij b jk  ckr , which equals
k =1  j =1 

p
 n 
(1.6.1)    aijb jk ckr .
k =1  j =1 

p n
 p

Similarly, the jr-entry of BC is  b jk ckr and so the ir-entry of A ( BC ) is  aij   b jk ckr  ,
k =1 j =1  k =1 
which equals

n
 p 
(1.6.2)    aij b jk ckr .
j =1  k =1 

Since the order of summation does not matter in a finite sum, the sums in (1.6.1) and (1.6.2) are
equal and so ( AB ) C = A ( BC ) since their corresponding entries are equal.


Proof of (3) Suppose B = bij  , C = cij  , and A =  a jk  . First note that both
m n m n n p

( B + C ) A and BA + CA are m  p matrices. Then note B + C = bij + cij 

m n
and so the ik-entry

of ( B + C ) A is

n
(1.6.3)  ( bij + cij ) a jk .
j =1
 9

n n
The ik-entry of BA is  bij a jk
j =1
and the ik-entry of CA is  cij a jk . So, the ik-entry of BA + CA is
j =1

n n
(1.6.4)  bij a jk +  cij a jk .
j =1 j =1

n n n n
Now note that  ( bij + cij ) a jk =  (bij a jk + cij a jk ) =  bij a jk +  cij a jk , i.e. (1.6.3) equals
j =1 j =1 j =1 j =1

(1.6.4). Therefore, ( B + C ) A = BA + CA since their corresponding entries are equal.



Proof of (5) Suppose A =  aij  and B = b jk  . First note that the size of AB is m  p and
m n n p

so the size of ( AB ) is p  m . One checks that the size of B T AT is also p  m. Now note that
T

n n

 aij b jk and so the ki-entry of ( AB ) is  aij b jk . Next note that the

T
the ik-entry of AB is
j =1 j =1
n
kj-entry of B T is b jk and the ji-entry of AT is aij and so the ki-entry of B T AT is  b jk a . So,
j =1
ij

we see that the entries of ( AB ) are equal to the corresponding entries of B T AT , and thus
T

( AB )
T
= B T AT .


The proofs of parts (2) and (4) of Proposition 1.6.1 are left to the reader.

If A is an n  n matrix then we define A2 = AA and Ak +1 = Ak A for all integers k  2. For

example, let

1 2 
A= .
 2 4

Then

1 2  1 2   5 10 
A2 =   =  = 5A
 2 4   2 4  10 20

and
 10

 5 10  1 2  25 50 
A3 =     =  = 52 A .
10 20  2 4 50 100

We use mathematical induction to prove that

(1.6.5) Ak +1 = 5k A

for all positive integers k. We’ve already established (1.6.5) for k = 1. Suppose (1.6.5) holds for
some positive integer k = n . Then, we have

An + 2 = An +1 A
= ( 5n A ) A by the inductive hypothesis
= 5n ( AA ) by Proposition 1.6.1(4)
= 5n A2
= 5n ( 5 A )
= ( 5n  5 ) A by Proposition 1.3.1(5)
= 5n +1 A

So (1.6.5) holds for k = n + 1 and the induction is complete.

1.7 Exercises

1.7.1A
Compute 1 2 3  4 5 6.
T

1.7.2A
Let A =  aij  such that aij = 1 if i = j and aij = 0 if i  j , and let B = bkr 33 such that
33

bkr = k − r. Compute 9 A − B.

1.7.3A
2 1 2 0 i 
Let A =   and B =   where i 2 = −1 . Find the matrix X such that AAT − 3 X = B 42 .
1 4 1   i 0

1.7.4
Suppose A and B are matrices of the same size. Prove that there is a unique matrix X such that
A + X = B.
 11

1.7.5
Prove that if A, B, and C are matrices of the same size such that A + C = B + C , then A = B.

1.7.6A
Find matrices A, B, and C such that C is nonzero, AC = BC , but A  B.

1.7.7A
Find nonzero matrices A and B such that AB = 0.

1.7.8
Suppose A and B are n  n matrices such that AB = BA. Prove that ( A + B ) = A2 + 2 AB + B 2 .
2

1.7.9
For square matrices of the same size define A  B = AB − BA. Prove that ( A  B )  C = A  ( B  C )
if and only if B  ( A  C ) = 0.

1.7.10A
 25 33
Find a 2  2 matrix B with nonnegative entries such that B 2 =  .
 0 36 

1.7.11A
Prove that if A is a symmetric matrix then An is symmetric for all positive integers n.

1.7.12
Prove that if A and B are symmetric matrices then AB − BA is skew-symmetric.

1.7.13*A
Prove that every square matrix can be written uniquely as the sum of a symmetric and a skew-
symmetric matrix.

1.7.14A
Suppose A and B are square matrices of the same size such that AB = BA. Prove that
( AB )
k
= Ak B k for all positive integers k.

1.7.15
Suppose A, B, and C are square matrices of the same size such that A = B + C , C 2 = 0, and
BC = CB. Prove that Ak +1 = B k ( B + ( k + 1) C ) for all positive integers k.
 12

1.7.16
Define the Fibonacci numbers by F0 = 0 , F1 = 1, and Fk + 2 = Fk +1 + Fk for all nonnegative
0 1 F Fk +1 
integers k. Let A =  . Prove that Ak +1 =  k for all nonnegative integers k.

1 1  Fk +1 Fk + 2 