4.

2 Matrix Representations of Linear Transformations

In the last section, we have learned that for any m  n matrix A , the following operator:

LA ( x)  Ax

from Rn  Rm for each x  Rn is a linear operator.

In this section, we will see that for every linear operator L mapping from Rn  Rm , there is an m  n
matrix A such that

L( x)  Ax

We will also see how any linear operator between finite dimensional spaces can be represented by a
matrix.

Theorem 1: If L is a linear operator mapping from Rn  Rm , there is an m  n matrix A such that

L( x)  Ax

for each x  Rn . In fact, the jth column vector of A is given by

a j  L(e j ), j  1, 2, , n.

Proof: Define a j  (a1 j , a2 j , , anj )T  L(e j ), j  1, 2, , n. Let A  (aij )  (a1 , a2 , , an ) .

For any x  ( x1 , x2 , , xn )T  Rn ( x  x1e1  x2e2   xn en ),

L( x)  L( x1e1  x2 e2   xn en )  x1L(e1 )  x2 L(e2 )   xn L(en )

 x1a1  x2 a2   xn an
 x1 
 
x
 (a1 a2 an )  2 
 
 
 xn 
 Ax

Example 1: Define the operator L : R2  R3 by

L( x)  (2 x1  3x2 , x1  x2 , 2 x1 )T

Find the matrix A such that L( x)  Ax for each x  R 2 .

Solution: We use the above theorem.

  2  0  2  0  3   3 
1      0    
L(e1 )  L( )   1  0    1  , L(e2 )  L( )   0  1    1  .
0  2   2 1  0   0 
       

 2 3 
Thus, we have A  ( L(e1 ), L(e2 ))   1 1 
2 0 
 

Theorem 2: If S  {s1, s2 , , sn } and R  {r1, r2 , , rn } are ordered bases for vector spaces V and W
respectively, then corresponding to each linear transformation L from V  W , there is an m  n matrix
A such that

[ L(v)]R  A[v]S

for each v V . A is the matrix representing L relative to the ordered bases S and R . In fact, the jth
column vector of A is given by

a j  [ L(s j )]R , j  1, 2, , n.

Example 2: Define the operator L : R3  R2 by

1  1
L( x)  (2 x1  3x3 )r1  ( x1  x2 )r2 for each x  R3 , where r1    , r2   

 
2 4

Find the matrix A representing L relative to the ordered bases [e1 ,e2 , e3 ] and [r1 , r2 ] .

Solution:

L( x)  (2 x1  3 x3 )r1  ( x1  x2 )r2 for each x  R 3 , where

1  1
r1    , r2   
 2  4
 2
L(e1 )  (2  0)r1  (1  0)r2  2r1  r2 , the coordinate vector relative to R=[r1 , r2 ] is [L (e1 )]R   
1
0
L(e2 )  (0  0)r1  (0  1) r2   r2 , the coordinate vector relative to R=[r1 , r2 ] is [L(e1 )]R   
 1
 3 
L(e3 )  (0  3)r1  (0  0) r2  3r1 , the coordinate vector relative to R=[r1 , r2 ] is [L (e1 )]R   
0
 2 0 3 
Thus, A   
 1 1 0 
Example 3: The linear operator D defined by D( p)  p ' maps P3 to P2 . Given the ordered bases
S  [ x2 , x,1] and R=[ x,1] for P3 and P2 , respectively, we wish to determine a matrix representation for D .

 2
D( x 2 )  2 x, the coordinate vector relative to ordered basis R=[x,1] is [D(x 2 )]R   
0
0
D( x)  1, the coordinate vector relative to ordered basis R=[x,1] is [D(x)]R   
1
0
D(1)  0, the coordinate vector relative to ordered basis R=[x,1] is [D(1)]R   
0

The matrix representation for D is

 2 0 0
A 
 0 1 0

Theorem 3: Given S  {s1, s2 , , sn } and R  {r1, r2 , , rn } are ordered bases for vector spaces Rn and Rm
respectively. If L is a linear transformation from Rn  Rm , and m  n matrix A is the matrix
representing L with respect to S and R , then

a j  R1L(s j ), j  1, 2, , n.

where R  (r1 , r2 , , rn ) .

Remark: This result can be obtained by Theorem 2:

a j  [ L(s j )]R , j  1, 2, , n, and change of basis result: [ L(s j )]E  R[ L(s j )]R  [ L(s j )]R  R1[ L(s j )]E

Thus we have a j  R1[ L(s j )]E  R1L(s j ), j  1, 2, , n.

Example 4: Define the operator L : R3  R2 by

L( x)  (2 x1  3x3 , x1  x2 )T for each x  R3.

Find the matrix A representing L relative to the ordered bases [s1 ,s2 , s3 ] and [r1 , r2 ] ,
where
1  1 1
     
s1 =  0  ,s2 = 1 , s3 =  0  , and
1  1  0
     
1  1
r1    , r2   

 2 3

Solution:
  1 1 1  3 1
R R  
 2 3   2 1

where
 2  3   1   2  3   1   2  0  2
L( s1 )=   =   ,L(s2 )=   =   , L( s3 )=  = 
 1  0    1  1  1   0  1 0   1
 3 1 1  2 
a1  R 1 L( s1 )=      
 2 1 1   1 
3 1 1  3 
a2  R 1 L( s2 )=      
2 1 0   2 
3 1 2   7 
a3  R 1 L( s3 )=      
2 1  1   5 

The matrix A representing L relative to the ordered bases [s1 ,s2 , s3 ] and [r1 , r2 ] ,

 2 3 7 
A 
 1 2 5 

HW: 2(a),(b),3, 4,7,16.