Linear Transformations

1.Introduction to Linear Transformations

2.The Kernel and Range of a Linear
Transformation
3.Matrices for Linear Transformations

Rank of a linear transformation T:VW :

rank(T ) the dimension of the range of T dim(range(T ))
range(T) = CS(A), so dim(range(T)) = dim(CS(A))

Nullity of a linear transformation T:VW:

nullity(T ) the dimension of the kernel of T dim(ker(T ))
ker(T) = NS(A), so dim(ker(T)) = dim(NS(A))

Note:
If T : R n R m is a linear transformation given by T (x) Ax, then
rank(T ) dim(range(T )) dim(CS ( A)) rank( A)
nullity(T ) dim(ker(T )) dim( NS ( A)) nullity( A)

The dimension of the row (or column) space of a matrix A is called the rank of A
The dimension of the nullspace of A (
) is called the nullity
NS ( A) {x | Ax 0}
of A

Theorem 5: Sum of rank and nullity

Let T: V W be a linear transformation from an n-dimensional
vector space V (i.e. the dim (domain of T) is n) into a vector
space W. Then
rank(T ) nullity(T ) n
(i.e. dim(range of T ) dim(kernel of T ) dim(domain of T ))
You can image that the dim(domain of T)
should equals the dim(range of T)
originally
But some dimensions of the domain of T
is absorbed by the zero vector in W
So the dim(range of T) is smaller than
the dim(domain of T) by the number of
how many dimensions of the domain of
T are absorbed by the zero vector,
which is exactly the dim(kernel of T)

Pf:

Let T be represented by an m n matrix A, and assume rank( A) r

(1) rank(T ) dim(range of T ) dim(column space of A) rank( A) r

(2) nullity(T ) dim(kernel of T ) dim(null space of A) n r

rank (T ) nullity(T ) r (n r ) n

according to Thm. 4.17 where

rank(A) + nullity(A) = n

Here we only consider that T is represented by an mn matrix A. In

the next section, we will prove that any linear transformation from an
n-dimensional space to an m-dimensional space can be represented
by mn matrix

Ex 8: Finding the rank and nullity of a linear transformation

Find the rank and nullity of the linear transformation T : R 3 R 3
define by
1 0 2
A 0 1 1
0 0 0

Sol:

rank (T ) rank ( A) 2
nullity(T ) dim( domain of T ) rank (T ) 3 2 1
The rank is determined by the number of leading 1s, and the
nullity by the number of free variables (columns without leading
1s)

Ex 9: Finding the rank and nullity of a linear transformation

Let T : R 5 R 7 be a linear transformation

(a) Find the dimension of the kernel of T if the dimension
of the range of T is 2
(b) Find the rank of T if the nullity of T is 4
(c) Find the rank of T if ker(T ) {0}
Sol:
(a) dim(domain of T ) n 5
dim(kernel of T ) n dim(range of T ) 5 2 3

(b) rank(T ) n nullity(T ) 5 4 1

(c) rank(T ) n nullity(T ) 5 0 5

One-to-one:
A function T : V W is called one-to-one if the preimage of
every w in the range consists of a single vector. This is equivalent
to saying that T is one-to-one iff for all u and v in V , T (u) T ( v)
implies that u v

one-to-one

not one-to-one

One to one
Not onto

Onto
Not one to one

One to one
Onto
(one to one
correspondance)

Theorem 6: One-to-one linear transformation

Let T : V W be a linear transformation. Then
T is one-to-one iff ker(T ) {0}

Pf:
() Suppose T is one-to-one
Then T ( v) 0 can have only one solution : v 0

Due to the fact

that T(0) = 0

i.e. ker(T ) {0}

() Suppose ker(T )={0} and T (u)=T (v)
T (u v) T (u) T ( v) 0
T is a linear transformation, see Property 3 in Thm. 1

u v ker(T ) u v 0 u v
T is one-to-one (because T (u) T ( v) implies that u v)

Ex 10: One-to-one and not one-to-one linear transformation

(a) The linear transformation T : M mn M nm given by T ( A) AT

is one-to-one
because its kernel consists of only the mn zero matrix

(b) The zero transformation T : R3 R3 is not one-to-one

because its kernel is all of R3

Onto:
A function T : V W is said to be onto if every element
in W has a preimage in V

(T is onto W when W is equal to the range of T)

Theorem 7: Onto linear transformations

Let T: V W be a linear transformation, where W is finite
dimensional. Then T is onto if and only if the rank of T is equal
to the dimension of W
rank (T ) dim( range of T ) dim(W )
The definition of
the rank of a linear
transformation

The definition
of onto linear
transformations

Theorem 8: One-to-one and onto linear transformations

Let T : V W be a linear transformation with vector space V and W

both of dimension n. Then T is one-to-one if and only if it is onto
Pf:

() If T is one-to-one, then ker(T ) {0} and dim(ker(T )) 0

Thm. 5

dim( range(T )) n dim(ker(T )) n dim(W )

Consequently, T is onto

According to the definition of dimension

that if a vector space V consists of the
zero vector alone, the dimension of V is
defined as zero

() If T is onto, then dim( range of T ) dim(W ) n

Thm. 5

dim(ker(T )) n dim( range of T ) n n 0 ker(T ) {0}

Therefore, T is one-to-one

Ex 11:

The linear transformation T : R n R m is given by T (x) Ax. Find the nullity

and rank of T and determine whether T is one-to-one, onto, or neither
1 2 0
(a) A 0 1 1
0 0 1

1 2
(b) A 0 1
0 0
= dim(Rn)
=n

Sol:
T:RnRm

1 2 0
(c) A

0 1 1

= dim(range
of T)
= # of
leading 1s

= (1) (2) =
dim(ker(T))

dim(domain of T)
rank(T) (2)
(1)

1 2 0
(d) A 0 1 1
0 0 0
If nullity(T)
= dim(ker(T))
=0

If rank(T) =
dim(Rm) =
m

nullity(T)

1-1

onto

(a) T:R3R3

Yes

Yes

(b) T:R2R3

Yes

No

(c) T:R3R2

No

Yes

(d) T:R3R3

No

No

Isomorphism :
A linear transformation T : V W that is one to one and onto
is called an isomorphism. M oreover, if V and W are vector spaces
such that there exists an isomorphism from V to W , then V and W
are said to be isomorphic to each other

Theorem 9: Isomorphic spaces and dimension

Two finite-dimensional vector space V and W are isomorphic
if and only if they are of the same dimension
Pf:
() Assume that V is isomorphic toW , where V has dimension n

There exists a L.T.T : V W that is one to one and onto

T is one-to-one
dim(V) = n
dim(ker(T )) 0

dim( range of T ) dim( domain of T ) dim(ker(T )) n 0 n

T is onto
dim( range of T ) dim(W ) n
Thus dim(V ) dim(W ) n

() Assume that V and W both have dimension n

Let B v1 , v 2 , ,v n be a basis of V and

B ' w1 , w 2 ,

,w n be a basis of W

Then an arbitrary vector in V can be represented as

v c1v1 c2 v 2 cn v n
and you can define a L.T. T : V W as follows
w T ( v) c1w1 c2 w 2 cn w n (by defining T (vi ) w i )

Since B' is a basis for V, {w1, w2,wn} is linearly

independent, and the only solution for w=0 is c1=c2==cn=0
So with w=0, the corresponding v is 0, i.e., ker(T) = {0}

T is one - to - one
By Theorem5, we can derive that dim(range of T) =
dim(domain of T) dim(ker(T)) = n 0 = n = dim(W)

T is onto
Since this linear transformation is both one-to-one and onto,
then V and W are isomorphic

3. Matrices for Linear Transformations

Two representations of the linear transformation T:R3R3 :

(1) T ( x1 , x2 , x3 ) (2 x1 x2 x3 , x1 3x2 2 x3 ,3x2 4 x3 )

2 1 1 x1
(2) T (x) Ax 1 3 2 x2
0 3 4 x3

Three reasons for matrix representation of a linear transformation:

It is simpler to write

It is simpler to read

It is more easily adapted for computer use

Theorem 10: Standard matrix for a linear transformation

Let T : Rn Rm be a linear trtansformation such that
a11
a12
a
a
T (e1 ) 21 , T (e2 ) 22 ,




a
m1
am 2

where {e1 , e2 ,

a1n
a
, T (e n ) 2 n ,


amn

, en } is a standard basis for R n . Then the m n

matrix whose n columns correspond to T (ei ),

A T (e1 ) T (e 2 )

a11
a
T (e n ) 21

am1

a1n
a2 n
,

amn

a12
a22
am 2

is such that T(v) Av for every v in R , A is called the

standard matrix for T
n

Pf:

v1
1
0
v
0
1
v 2 v1 v2






v
0

0
n

0
0
vn v1e1 v2e 2


1

T is a linear transformation T ( v) T (v1e1 v2e2

If A T (e1 ) T (e2 )

a11
a21
Av

am1

a12
a22

am 2

vne n

vne n )

T (v1e1 ) T (v2e2 )

T (vnen )

v1T (e1 ) v2T (e2 )

vnT (en )

T (en ) , then

a1n v1 a11v1 a12v2 a1n vn

a2 n v2 a21v1 a22v2 a2 n vn

amn vn am1v1 am 2 v2 amn vn

 a11
a12
a1n
a
a
a
v1 21 v2 22 vn 2 n






am1
am 2
amn
v1T (e1 ) v2T (e2 ) vnT (e n )
Therefore, T ( v) Av for each v in R n

Note

Theorem 10 tells us that once we know the image of every

vector in the standard basis (that is T(ei)), you can use the
properties of linear transformations to determine T(v) for any
v in V

Ex 1: Finding the standard matrix of a linear transformation

Find the standard matrix for the L.T. T : R3 R2 defined by
T ( x, y, z) ( x 2 y, 2 x y)

Sol:
Vector Notation
T (e1 ) T (1, 0, 0) (1, 2)

T (e2 ) T (0, 1, 0) (2, 1)

T (e3 ) T (0, 0, 1) (0, 0)

Matrix Notation
1
1

T (e1 ) T ( 0 )
2
0
0
2

T (e2 ) T ( 1 )
1
0
0
0

T (e3 ) T ( 0 )
0

6.21

A T (e1 ) T (e2 ) T (e3 )

1 2 0

2
1
0

Check:
x
x
1 2 0 x 2 y
A y
y

2 1 0 2 x y
z
z
i.e., T ( x, y, z) ( x 2 y, 2 x y)

Note: a more direct way to construct the standard matrix

1 2 0 1x 2 y 0 z
A
2 1 0 2 x 1y 0 z

The first (second) row actually

represents the linear transformation
function to generate the first (second)
component of the target vector

Ex 2: Finding the standard matrix of a linear transformation

The linear transformation T : R 2 R 2 is given by projecting
each point in R 2 onto the x - axis. Find the standard matrix for T

Sol:
T ( x, y) ( x, 0)

1 0
A T (e1 ) T (e2 ) T (1, 0) T (0, 1)

0
0

Notes:
(1) The standard matrix for the zero transformation from Rn into Rm
is the mn zero matrix
(2) The standard matrix for the identity transformation from Rn into
Rn is the nn identity matrix In

Composition of T1:RnRm with T2:RmRp :

T ( v) T2 (T1 ( v)), v R n
This composition is denoted by T T2 T1

Theorem11: Composition of linear transformations :

Let T1 : R n R m and T2 : R m R p be linear transformations
with standard matrices A1 and A2 ,then
(1) The composition T : R n R p , defined by T ( v) T2 (T1 ( v)),
is still a linear transformation
(2) The standard matrix A for T is given by the matrix product
A A2 A1

Pf:
(1) (T is a linear transformation)
Let u and v be vectors in R n and let c be any scalar. Then
T (u v) T2 (T1 (u v)) T2 (T1 (u) T1 ( v))
T2 (T1 (u)) T2 (T1 ( v)) T (u) T ( v)
T (cv) T2 (T1 (cv)) T2 (cT1 ( v)) cT2 (T1 ( v)) cT ( v)

(2) ( A2 A1 is the standard matrix for T )

T ( v) T2 (T1 ( v)) T2 ( A1v) A2 A1v ( A2 A1 ) v

Note:
T1 T2 T2 T1

Ex 3: The standard matrix of a composition

Let T1 and T2 be linear transformations from R3 into R3 s.t.
T1 ( x, y, z) (2 x y, 0, x z)
T2 ( x, y, z) ( x y, z, y)
Find the standard matrices for the compositions T T2 T1
and T ' T1 T2

Sol:
2
A1 0
1
1
A2 0
0

1 0
0 0 (standard matrix for T1 )
0 1
1 0
0 1 (standard matrix for T2 )
1 0

The standard matrix for T T2 T1

1 1 0 2 1 0 2 1 0
A A2 A1 0 0 1 0 0 0 1 0 1

0
1
0
1
0
1
0
0
0

The standard matrix for T ' T1 T2

2 1 0 1 1 0 2 2 1
A' A1 A2 0 0 0 0 0 1 0 0 0
1 0 1 0 1 0 1 0 0

Inverse linear transformation :

If T1 : Rn Rn and T2 : Rn Rn are L.T. s.t. for every v in Rn
T2 (T1 ( v)) v and T1 (T2 ( v)) v

Then T2 is called the inverse of T1 and T1 is said to be invertible

Note:

If the transformation T is invertible, then the inverse is

unique and denoted by T1

Theorem 12: Existence of an inverse transformation

Let T : R n R n be a linear transformation with standard matrix A,
Then the following condition are equivalent
For (2) (1), you can imagine that
since T is one-to-one and onto, for every
w in the codomain of T, there is only one
(2) T is an isomorphism
preimage v, which implies that T-1(w) =v
can be a L.T. and well-defined, so we can
(3) A is invertible
infer that T is invertible
On the contrary, if there are many
preimages for each w, it is impossible to
find a L.T to represent T-1 (because for a
L.T, there is always one input and one
Note:
output), so T cannot be invertible

(1) T is invertible

If T is invertible with standard matrix A, then the standard

matrix for T1 is A1

Ex 4: Finding the inverse of a linear transformation

The linear transformation T : R3 R3 is defined by

T ( x1 , x2 , x3 ) (2 x1 3x2 x3 , 3x1 3x2 x3 , 2 x1 4 x2 x3 )

Show that T is invertible, and find its inverse

Sol:
The standard matrix for T
2 3 1
A 3 3 1
2 4 1

2 x1 3 x2 x3
3 x1 3 x2 x3
2 x1 4 x2 x3

2 3 1 1 0 0
A I 3 3 3 1 0 1 0

2
4
1
0
0
1

1 0 0 1 1 0
G.-J. E.

0 1 0 1 0 1 I
0 0 1 6 2 3

A1

Therefore T is invertible and the standard matrix for T 1 is A1

0
1 1
A1 1 0
1

0 x1 x1 x2
1 1
T 1 ( v) A1 v 1 0
1 x2 x1 x3

3
x
6
x

2
x

3
x
2
3

3 1
In other words,
T 1 ( x1 , x2 , x3 ) ( x1 x2 , x1 x3 , 6 x1 2 x2 3x3 )
Check T-1(T(2, 3, 4)) = T-1(17, 19, 20) = (2, 3, 4)

The matrix of T relative to the bases B and B:

T :V W
B {v1 , v 2 ,

, vn}

(a linear transformation)
(a nonstandard basis for V )

The coordinate matrix of any v relative to B is denoted by [v]B

if v can be represeted as c v c v
1 1
2 2

c1
c
cn v n , then [v]B 2


cn

A matrix A can represent T if the result of A multiplied by a

coordinate matrix of v relative to B is a coordinate matrix of v
relative to B, where B is a basis for W. That is,

T ( v)B ' A[ v]B ,

where A is called the matrix of T relative to the bases B and B

Transformation matrix for nonstandard bases (the

generalization of Theorem 10, in which standard bases are
considered) :
Let V and W be finite - dimensional vector spaces with bases B and B ',
respectively, where B {v1 , v 2 , , v n }
If T : V W is a linear transformation s.t.

T ( v1 )B '

a11
a
21 ,


am1

T ( v 2 )B '

a12
a
22 ,


am 2

T ( v n )B '

a1n
a
2n


amn

then the m n matrix who se n columns correspond to T (vi )B '

A [T ( v1 )]B [T ( v 2 )]B

a11
a
[T ( v n )]B 21

am1

a12
a22
am 2

a1n
a2 n

amn

is such that T ( v)B ' A[ v]B for every v in V

The above result state that the coordinate of T(v) relative to the basis B
equals the multiplication of A defined above and the coordinate of v
relative to the basis B.
Comparing to the result in Thm. 10 (T(v) = Av), it can infer that the
linear transformation and the basis change can be achieved in one step
through multiplying the matrix A defined above.

Ex 5: Finding a matrix relative to nonstandard bases

Let T : R2 R2 be a linear transformation defined by
T ( x1 , x2 ) ( x1 x2 , 2 x1 x2 )

Find the matrix of T relative to the basis B {(1, 2), (1, 1)}
and B ' {(1, 0), (0, 1)}

Sol:
T (1, 2) (3, 0) 3(1, 0) 0(0, 1)
T (1, 1) (0, 3) 0(1, 0) 3(0, 1)
T (1, 2)B ' 03, T (1, 1)B ' 03


the matrix for T relative to B and B'

A T (1, 2)B '

3 0
T (1, 2)B'

Ex 6:
For the L.T. T : R 2 R 2 given in Example 5, use the matrix A
to find T ( v), where v (2, 1)
Sol:
v (2, 1) 1(1, 2) 1(1, 1)
B {(1, 2), (1, 1)}
1
v B
1
3 0 1 3
T ( v)B ' Av B

0 3 1 3
T ( v) 3(1, 0) 3(0, 1) (3, 3)
B' {(1, 0), (0, 1)}

Check:
T (2, 1) (2 1, 2(2) 1) (3, 3)