Chapter 5

Linear Transformations

1
Wen-Hsien Fang Engineering Mathematics (II) 1/53
 Organization

• Basic
concepts
What Where
(5.1)? (5.3)?
Linear
Transformation

• Applications
How
• Characterization (5.2)?
• One-to-one &
onto

2
Wen-Hsien Fang Engineering Mathematics (II) 2/53
 5.1 WHAT are Linear
Transformations
A mapping T from a vector space V into a vector space W
will be denoted by
T: V  W

Definition
A mapping T from a vector space V into a vector Space
W is said to be a linear transformation if
T(v1 + v2) = T(v1) + T(v2)
for all v1, v2  V and for all scalars  and .

3
Wen-Hsien Fang Engineering Mathematics (II) 3/53
5.1 WHAT are Linear Transformations

Remarks:
1. If T is a linear transformation mapping a vector in a
vector space V into another vector in W.
2. The linearity condition can also be expressed as
T(v1 + v2) = T(v1) + T(v2) ( =  = 1) (1)
T(v) = T(v) (v1 = v,  = 0) (2)
i.e. Preservation of linear operation
3. If the vector space V and W are the same, the linear
transformation T: V  V is referred to as a linear
operator on V.

4
Wen-Hsien Fang Engineering Mathematics (II) 4/53
5.1 WHAT are Linear Transformations

Example 5.1-1
2 2
Linear Transformation from R to R

stretching 90o rotation reflection projection

  x    cx    x    y    x    y    x    x 
T      T      T      T     
  y    cy    y    x    y    x    y    0

5
Wen-Hsien Fang Engineering Mathematics (II) 5/53
5.1 WHAT are Linear Transformations

Example 5.1-2
x  y 
 x   
T : R  R by T      x  2 y 
2 3
 
  y    3x 
 
This is a linear transformation as

  x1   x2     x1   x2  
T          T   
y
  1 y
 2 
 1 y   y 2

( x1   x2 )  ( y1   y2 )  ( x1  y1 )  ( x2  y2 ) 
 ( x1   x2 )  2( y1   y2 )    ( x1  2 y1 )    ( x2  2 y2 ) 
     
 3( x1   x2 )   3 x1   3 x2 
 
T ( 1 ) T ( 2 )
x x
 y1   y2 

6
Wen-Hsien Fang Engineering Mathematics (II) 6/53
5.1 WHAT are Linear Transformations

Example 5.1-3 (Geometric Interpretation of Matrix-

Vector Multiplication as Linear Transformation)
If A is any m  n matrix, we can define a mapping
TA : R  R by TA  v   Av for all (column) vector v in R .
n m n

Ax

Then TA is a linear transformation (matrix transformation).

Proof: Let v1 , v 2 be any two vectors in R n , and r1 , r2  R
TA  r1v1  r2 v 2   A  r1v1  r2 v 2   r1Av1  r2 Av 2
1 A  v1   r2TA  v 2   TA    is a linear trasformation
 rT
7
Wen-Hsien Fang Engineering Mathematics (II) 7/53
5.1 WHAT are Linear Transformations

Example 5-1-4 (Some Linear Transformations)

 Identity transformation on V IV : V  V where IV ( v) for all v in V

 Zero transformation V  W 0 : V  W where 0( v )  0 for all v in V

 Scalar transformation V  V a : V  V where a( v)  av for all v in V
Here a is any real number

8
Wen-Hsien Fang Engineering Mathematics (II) 8/53
5.1 WHAT are Linear Transformations

Example 5.1-5 (Differentiation/ Integration)

D : Pn  Pn1 where D  p ( x )   p '  x  for all p ( x ) in Pn

x
I : Pn  Pn1 where I  p ( x )    p (t )dt for all p ( x ) in Pn
0

are also linear transformations on function spaces.

D(f + g) = f’ + g’ = D(f) + D(g)

b
I ( f   g )   ( f   g )( x)dx
a
b b
   f ( x) dx    g ( x)dx   I ( f )   I ( g )
a a

9
Wen-Hsien Fang Engineering Mathematics (II) 9/53
5.1 WHAT are Linear Transformations

Example 5.1-6
Consider the mapping M defined by M(x) = (x12 + x22)1/2.

Solution
Since: M (αx)  M  α  x1    M   αx1    (αx1 ) 2  (αx 2 ) 2 1/ 2
 x    αx  
  2  2
 | α | ( x12  x22 )1/2
αM (x)  α ( x12  x22 )1/2
 M(x)  M(x) whenever  < 0
 M is not a linear transformation.

10
Wen-Hsien Fang Engineering Mathematics (II) 10/53
5.1 WHAT are Linear Transformations

SOME USEFUL PROPERTIES

If T is a linear transformation mapping a vector space V
into a vector space W, then
(1) T(0V) = 0W (where 0V and 0W are zero vectors in V
and W)
(2) T(1v1 + 2v2 + … + nvn)
= 1T(v1) + 2T(v2) + … + nT(vn)
(3) T(-v) = -T(v) for all v  V.
Proof:
(1) & (3) take r=0 and -1 respectively in T(rv)=rT(v)
(2) Prove by mathematical induction
11
Wen-Hsien Fang Engineering Mathematics (II) 11/53
5.1 WHAT are Linear Transformations

Example 5.1-7

If T  v  3v1  =w and T  2 v  v1  =w1 , find T  v  & T  v1 

in terms of w & w1

Solution
 1
T  v   3T  v1   w T  v   5  3w1  w 
 
2T  v   T  v1   w1 T  v   1  w  2w 
 1
5
1

12
Wen-Hsien Fang Engineering Mathematics (II) 12/53
5.1 WHAT are Linear Transformations

Example 5.1-8
T : R  R is a linear transformation and T  3, 1, 2  =5,
3

T 1,0,1  2, compute T  1,1,0  .

Solution
   1,1,0 
 
T  1,1,0   T    3, 1, 2   2 1, 0,1 
 
 
=  T  3, 1, 2   2T 1,0,1  5  2  2  1
13
Wen-Hsien Fang Engineering Mathematics (II) 13/53
 5.2 HOW to Specify Linear
Transformations
Definition:
Two linear transformations T:V  W and S:V  W are
called equal (written T=S) if they have the same action ;
that is T(v)=S(v) for all v in V.

Since every point in V can be expressed as a linear combination

of the basis, if two linear transformations have the same effect on
a basis, then they are equal. This leads to the following theorem.

14
Wen-Hsien Fang Engineering Mathematics (II) 14/53
5.2 HOW to Specify Linear Transformations

Theorem
Consider a linear transformation T : V  W. Suppose
{e1 ,e2 ,en } is a basis of V and T  ei  = w i , w i  W, i  1,2, n.
Then for every v in V, T  v  is uniquely determined by
w1 ,w 2 , w n  they needn't be distinct  . More specifically,
if v  1e1   2e2     nen in V, then
T  v   T 1e1   2e2     nen   1w1   2 w 2     n w n

Proof:  T is linear,
 T( v)  T 1e1   2e 2     ne n 
=1T  e1 )   2T (e 2 )     nT (e n 
=1w1   2 w 2     n w n
15
Wen-Hsien Fang Engineering Mathematics (II) 15/53
5.2 HOW to Specify Linear Transformations

Remarks:
The significance of the above theorem is that a linear
transformation defined on V is completely determined
by the mapping of the basis vectors, i.e.

T  e1  , T  e 2  ,T  en 

Therefore, deciding whether two linear transformations

are equal comes down to determining whether they have
the same effect on the basis vectors.

16
Wen-Hsien Fang Engineering Mathematics (II) 16/53
5.2 HOW to Specify Linear Transformations

T1
e1   w1
e2   w 2 =w 3
e3  
 
 wn
en 
V T2 W
e1   w1
e2   w2
e3   w3
  
en   wn
 
17
Wen-Hsien Fang Engineering Mathematics (II) 17/53
5.2 HOW to Specify Linear Transformations

Example 5.2-1
Find a linear transformation T : R 3  R 2 such that
 1    1    0  
    2    1      0
T  1     T 0     T 1    
   1      1    0 
 0   1    1  
     
Solution
 1  1  0   x
       3  y
 1  ,  0  , 1   is a basis for R an arbitrary vector  
  0  1  1    z 
      
in R 3 can be expressed as
18
Wen-Hsien Fang Engineering Mathematics (II) 18/53
5.2 HOW to Specify Linear Transformations

 x 1  1  0
 y   1 ( x  y  z ) 1   1 ( x  y  z )  0   1 ( x  y  z ) 1 
  2   2   2  
 z  0  1  1 
Same
weighting
 x
  1  2 1 1 1 0 
T ( y )  ( x  y  z )    ( x  y  z )    ( x  y  z )  
  2  1 2  1 2  0
 z 
1 3 x  y  z 
 
2  2( y  z ) 

19
Wen-Hsien Fang Engineering Mathematics (II) 19/53
 5.3 Where - Kernel and Image

Let us first consider the following two subspaces which can help
us decide if a linear transformation is one-to-one or onto.
Definition
Let T: V  W be a linear transformation. The kernel of
T, denoted ker(T), is define by
ker(T) = {v  V | T(v) = 0W}

Ker(T)
T 0W
V W
● ker(T) is the region in V which T maps to zero 0w , i.e.
without taking any operation on the vectors in ker(T). 20
Wen-Hsien Fang Engineering Mathematics (II) 20/53
5.3 Where- Kernel and images

Definition:
Let T: V  W be a linear transformation and the image
(or range) of T, denoted Im(T), is defined by
Im(T) = {T(v) | for all v  V}

Range
of T
Im(T)
V T
W

● Im (T) is the region in W which the range of T takes on.

Other region in W is unreachable by this mapping on W.
21
Wen-Hsien Fang Engineering Mathematics (II) 21/53
5.3 Where- Kernel and images

ker(T) is a subspace of V and Im(T) is a subspace of W.

Proof:
(1) If v  ker(T) and  is a scalar  T(v) = 0W
T(v) =  T(v) = 0W = 0W  v  ker(T)
Also, if v1 and v2  ker(T)  T (v1) = T (v2) = 0W, then
T(v1 + v2) = T (v1) + T (v2) = 0W + 0W = 0W  v1 + v2  ker(T)
(2) If w  Im(T),w = T(v) for some v  V. w = T(v) =T(v).
Since V is a subspace  v  V w  Im(T)
If w1 and w2  Im(T), then there exist v1 and v2  V such that
T(v1) = w1 and T(v2) = w2
 w1+w2 = T(v1) + T(v2) = T(v1+v2)  w1+w2 Im(T)
22
Wen-Hsien Fang Engineering Mathematics (II) 22/53
5.3 Where- Kernel and images

Example 5.3-1
T : R  R is denoted by T ( x, y , z )  ( x  y, z , y  x )
3 3

Find ker(T ) & Im(T ).

Solution

(i ) ker(T )  {(x, y, z )|(x  y, z , y  x)  (0, 0, 0)}=span{(1, 1, 0)}

(ii ) Im(T ): (x  y, z , y  x )  x(1, 0,  1)  y (1, 0, 1)  z (0, 1, 0)

 Im(T )  span{(1, 0,  1), (0, 1, 0)}  dim( Im(T ))  2

23
Wen-Hsien Fang Engineering Mathematics (II) 23/53
5.3 Where- Kernel and images

Definition
Let T: V  W be a linear transformation.
(i)T is said to be onto if Im(T) = W
(ii)T is said to be one-to-one if T(v1) =T(v2) implies v1=v2

onto
T  x
x
 x
V V One-to-one
W W
24
Wen-Hsien Fang Engineering Mathematics (II) 24/53
5.3 Where- Kernel and images

Theorem
If T : V  W is a linear transformation, then T is
one-to-one iff ker(T )  {0V }
Proof :
(i ) one-to-one  ker(T )  {0V }
(ii ) Suppose two vectors v1 , v 2 in V have T ( v1 )  T ( v 2 )
 0W  T (v1 )  T ( v 2 )  T ( v1  v 2 )
 ker(T )  {0V }  v1  v 2

25
Wen-Hsien Fang Engineering Mathematics (II) 25/53
5.3 Where- Kernel and images

Example 5.3-2

S , T are two linear transformations

S : R 3  R 2 by S ( x, y , z )  ( x  y , x  y )
T : R  R by T ( x, y )  ( x  y , x  y, x)
2 3

Solution
S is onto but not one-to-one

T is one-to-one but not onto

26
Wen-Hsien Fang Engineering Mathematics (II) 26/53
5.3 Where- Kernel and images

• Dimension Theory for Linear Transformation

Theorem
Let T : V  W be any linear transformation, and assume
that both ker(T ) and Im(T ) are finite dimensional. The
dimension of V is also finite and
dim(V )  dim(ker(T ))  dim(Im(T ))
i.e. dim(V )  nullity(T )  rank(T )

27
Wen-Hsien Fang Engineering Mathematics (II) 27/53
5.3 Where- Kernel and images

The following block diagram shows this relationship:

{e k 1 ,  , e n } {T (e k 1 ), , T (e n )}
nk nk
basis a basis for Im(T )
n
{e1 ,  ,e k }
k {0W }
basis for ker(T )

V W
T
dim(V )  n dim(W )  m

28
Wen-Hsien Fang Engineering Mathematics (II) 28/53
5.3 Where- Kernel and images

Proof :
Assume {e1 ,,ek } is a basis for ker(T ). This basis
can be extened to {e1 ,,ek , e k 1 ,,e n }. It suffices to
show that B ={T (e k 1 ),T (ek  2 ),,T (e n )} is a basis for Im(T )

(i ) B spans Im(T ): every vector v in V can be expressed as

v  r1e1    rne n
1 (e1 )    rk T (e k )  rk 1T (e k 1 )    rnT (e n )
T ( v )  rT
 rk 1T (e k 1 )    rnT (e n )

29
Wen-Hsien Fang Engineering Mathematics (II) 29/53
5.3 Where- Kernel and images

(ii ) B is linearly independent:

n n
Suppose  rT (e )  0
i  k 1
i i W  T (  riei )  0W
i  k 1
n k n
  riei  i ei (   riei ker(T ) )
i  k 1 i 1 i  k 1

 1e1 + + k e k  rk 1e k 1    rne n  0

 1   = k  rk 1    rn =0
{e1 e k , e k 1 en } is a basis (thus linearly independent)

 dim(ker(T ))  dim(Im(T ))  k  (n  k )  n  dim(V )

 
k nk
30
Wen-Hsien Fang Engineering Mathematics (II) 30/53
5.3 Where- Kernel and images

Example 5-3-3

Let D: Pn→Pn-1 be the differentiation operator, defined

by D[p(x)] = p’(x) . Compute ker (D) and hence
conclude what D is onto.
Solution
p’(x) = 0 → p(x) is constant → dim(ker(D)) = 1.
Because dim(Pn) = n ,the dimension theorem gives
dim(Im(D)) = n - dim(ker(D)) = n-1 = dim(Pn-1)
→ Im(D) = Pn-1, so D is onto.
31
Wen-Hsien Fang Engineering Mathematics (II) 31/53
5.3 Where- Kernel and images

• Dimension Theorem for Matrices

ker(TA )  {x  R : TA  0}  {x : Ax  0}  N ( A)
n

Im(TA )  {TA (x) : x  R }  {Ax : x  R }  R( A )

n n

dim  r R(A) dim R( A)  r

dim  m  r
dim N( A)
N(A)
 nr Rm

Rn
dim( R )  dim(N( A))  dim(R( A ))  (n  r )  r  n
n

32
Wen-Hsien Fang Engineering Mathematics (II) 32/53
5.3 Where- Kernel and images

Theorem
Let T: V W be a linear transformation and assume that dim(V)=n
and dim(W)=m. Then:
(a) If m>n, then T is not onto.
(b) If m<n, then T is not one-to-one.
(c) If m=n, then T is onto iff T is one-to-one.

33
Wen-Hsien Fang Engineering Mathematics (II) 33/53
5.3 Where- Kernel and images

Proof
(a) n T
m

V W

(b) T
n m

34
Wen-Hsien Fang Engineering Mathematics (II) 34/53
5.3 Where- Kernel and images

V W
x
x
x
no ker(T) no unachievable
region

35
Wen-Hsien Fang Engineering Mathematics (II) 35/53
 5.4 Matrix Representation of
Linear Transformations

Recall that given an mn matrix A, the transformation

TA(x) = Ax is a linear transformation,
The converse is also true: For each linear transformation
TA mapping Rn into Rm there is an mn matrix A such
that
TA(x) = Ax
for each x  Rn.

36
Wen-Hsien Fang Engineering Mathematics (II) 36/53
5.4 Matrix Representation

Theorem
If T is a linear transformation mapping to , there is an
matrix A such that
T(v)=Av
For each . In fact, the jth column vector of A is given by

, j=1, 2,…, n
A=[T( ), T( ),…, T( )]

37
Wen-Hsien Fang Engineering Mathematics (II) 37/53
5.4 Matrix Representation

Proof:
T(v) = T(v1e1 + v2e2 + … + vnen)
= v1T(e1) + v2T(e2) + … + vnT(en)
= v1a1 + v2a2 + … + vnan
 v1 
v 
 a1 | a 2 |  | a n   2
   
A  
 vn 
v

38
Wen-Hsien Fang Engineering Mathematics (II) 38/53
5.4 Matrix Representation

Remarks：
(i) Linear transformation is uniquely specified if we
choose a basis for Rn and their corresponding
matching points.
We choose the basis {e1,e2, …,en},and the
corresponding points are{a1,a2, …,an} e T a1
1

A
  
a1 a2  a n  x  Ax en
T
an
n m
(ii) The A determined above is referred to as the
standard matrix representation of T
39
Wen-Hsien Fang Engineering Mathematics (II) 39/53
5.4 Matrix Representation

Example 5.4-1
The standard matrix of T: R3  R2 defined by
  x 
    x  2y  z
T y 
     x  z 
z
Solution
 
  1    0    0   
              1 2 1 
The desire matrix =  T 0 T 1 T 0 
        1 0 1
  0    0 
 
 1   
   

 x  x
    1 2 1     x  2 y  z 
so T y =  y 
    1 0 1    x  z 

  z    
  z 
  A
40
Wen-Hsien Fang Engineering Mathematics (II) 40/53
5.4 Matrix Representation

Example 5.4-2

Find the corresponding standard matrix for the linear

transformation T: R2  R2
(a) Rotation about the origin counterclockwise through
the angle   R
y
x
 y
 

x
41
Wen-Hsien Fang Engineering Mathematics (II) 41/53
5.4 Matrix Representation

 x
(b) Projection on the -line  y
 
 P

(c) Reflection on the -line y

 RE x
 y
 
x
(d)In (a), rotation followed by a projection on x axis

42
Wen-Hsien Fang Engineering Mathematics (II) 42/53
5.4 Matrix Representation

Solution

(a) Rotation:
• The desired matrix is

cos θ sin θ 
A   R  (e1 ) R  (e 2 )    
 sin θ cos θ 
  x    cos θ sin θ   x 
 R         y
y sin θ
       cos θ
A
Rotation by θ
43
Wen-Hsien Fang Engineering Mathematics (II) 43/53
5.4 Matrix Representation

0 
1 
 

(b) Projection : 1 
0 
 
 cos 2 θ 
P (e1 )   
 cosθsinθ 
0 
cosθsinθ  1  c 
P (e 2 )   2 
  θ c 
 sin θ  s 
1 
c
The corresponding matrix is s  0 
 
s 
 cos 2 θ cos θsin θ 
A 2 
cosθsin θ sin θ 

44
Wen-Hsien Fang Engineering Mathematics (II) 44/53
5.4 Matrix Representation

(c) Reflection :
(assume  <45,
same result can be derived for   45)

 cos(2θ)   2c  1
2
RE (e1 )     
 sin(2θ)   2cs 
1 
0 
 

45
Wen-Hsien Fang Engineering Mathematics (II) 45/53
5.4 Matrix Representation

 cos(2(90  θ)  90)   sin(2θ)   2cs 

RE (e 2 )        2 2
 sin(2(90  θ)  90)    cos(2θ)   c  s 

 2c  1 2cs  2
A 2
 2cs c  s 
2

2(90   )  90

c   s 
(d) Now T(e1 )    ,T(e 2 )   
0  0
c s 
 corresponding matrix A =  
0 0 
46
Wen-Hsien Fang Engineering Mathematics (II) 46/53
5.4 Matrix Representation

Application: Computer Graphics and Animation

A picture with n vertices can be stored in a 2n matrix.

The x coordinates of the vertices are stored in the first
row and the y coordinates in the second. Each
successive pair of points is connected by a straight line.

 x1 x2  xn x1 
y y2  yn y1 
 1

47
Wen-Hsien Fang Engineering Mathematics (II) 47/53
5.4 Matrix Representation

An example: a triangle with 3 vertices (0, 0), (1, 1) and

(1, -1) are stored in a matrix:

0 1 1 0 
T 
0 1  1 0 

48
Wen-Hsien Fang Engineering Mathematics (II) 48/53
5.4 Matrix Representation

Dilations and Contractions

The transformation T is represented by the matrix A = cI,

where I is the 22 identity matrix

 c 0   0 1 1 0  0 c c 0
AT       
 0 c   0 1  1 0  0 c  c 0 
49
Wen-Hsien Fang Engineering Mathematics (II) 49/53
5.4 Matrix Representation

Reflection about Y-axis:

Similarly, if Ty is a linear transformation that reflects a
vector about the y-axis, then Ty is represented by a 22
matrix A.

 1 0   1 0   0 1 1 0  0 1 1 0 
A   AT       
 0 1   0 1   0 1  1 0   0 1  1 0 
50
Wen-Hsien Fang Engineering Mathematics (II) 50/53
5.4 Matrix Representation

Rotation

cos θ  sin θ  0 1 1 0  0 cos θ  sin θ cos θ  sin θ 0 

AT       
 sin θ cos θ   0 1  1 0   0 sin θ  cos θ sin θ  cos θ 0 
51
Wen-Hsien Fang Engineering Mathematics (II) 51/53
5.4 Matrix Representation

Remarks:
• Translation:
A translation by a vector a is a transformation of the form
T(x) = x + a
If a  0, T is not a linear transformation and hence T cannot
be represented by a 22 matrix.

 x1 
 x1   
 x    x2 
 2
 1 
52
Wen-Hsien Fang Engineering Mathematics (II) 52/53
5.4 Matrix Representation

Homogeneous Coordinates Systems (for R2 ):

 x1 
 x1   
 x    x2 
 2
 1 
 a1 
If T(x) = x + a (   a  ), the matrix representation can be
 2
formed by replacing the first two entries in the last column
of the identity matrix I with the entries of a:

1 0 0 1 0 a1  1 0 a1   x1   x1  a1 
 0 1 0   0 1 a   0 1 a   x    x  a 
   2
  2
  2  2 2

0 0 1 0 0 1  0 0 1   1   1 
53
Wen-Hsien Fang Engineering Mathematics (II) 53/53