Mathematics and Statistics II - Linear Algebra

Week 5: Linear Transformations

Andrés Perea

Maastricht University

Period 2

Andrés Perea (Maastricht University) Linear Transformations Period 2 1 / 26

Linear Transformations

Consider a vector space U and a (possibly di¤erent) vector space V .

A function T : U ! V is a linear transformation if

T (û + v̂ ) = T (û ) + T (v̂ ) for all û, v̂ 2 U, and
T (λ û ) = λ T (û ) for all û 2 U and all λ 2 R.
Or, equivalently, T (λ û + µ v̂ ) = λ T (û ) + µ T (v̂ )
for all û, v̂ 2 U and all λ, µ 2 R.
By L(U, V ) we denote the set of all linear transformations from U to
V.

Andrés Perea (Maastricht University) Linear Transformations Period 2 2 / 26

 Example: Consider the function T : R3 ! R2 where
02 31
x
2x + 3y z
T @4 y 5A = .
x + 4y
z
Check
0 whether T is a 1
linear transformation:

B2 3 2 0 3C
B x x C
B C
T B4 y 5 + 4 y 0 5C =
B 0 C
@ z z A
| {z } | {z }
02 û 31v̂
x + x0
2(x + x 0 ) + 3(y + y 0 ) (z + z 0 )
T @4 y + y 0 5A =
(x + x 0 ) + 4(y + y 0 )
z + z0
2x + 3y z 2x 0 + 3y 0 z 0
= + = T (û ) + T (v̂ ).
x + 4y x 0 + 4y 0
Try to show that T (λ û ) = λ T (û ) for every û 2 R3 and every
λ 2 R. Hence, T is a linear transformation.
Andrés Perea (Maastricht University) Linear Transformations Period 2 3 / 26
 Example: Let Pn be the set of polynomials of degree at most n.
d
Consider the function T : Pn ! Pn 1 where T (p ) = dx p.

Check whether T is a linear transformation:

d d d
T (p + q ) = dx (p + q) = dx p + dx q = T (p ) + T (q ).
d d
T (λ p ) = dx ( λ p) = λ dx p = λ T (p ).
Hence, T is a linear transformation.

Example: Function T : R2 ! R2 where

x x y
T = is not a linear transformation.
y x2+ y2
Try to show this.

Andrés Perea (Maastricht University) Linear Transformations Period 2 4 / 26

Theorem
Let U and V be vector spaces, and T : U ! V a linear transformation.
Then:
(a) T (0̂U ) = 0̂V , and
(b) T ( û ) = T (û ) for all û 2 U.

Proof: (a) T (0̂U ) = T (0 û ) = 0 T (û ) = 0̂V .

(b) T ( û ) + T (û ) = T (( û ) + û ) = T (0̂U ) = 0̂V , and hence
T ( û ) = T (û ).
Example: Transformation T : R2 ! R3 with
2 3
2x + y
x
T = 4 x + 3y + 2 5 is not a linear transformation, since
y
3x + 2y
2 3 2 3
0 0
0
T = 2 6= 0 5 .
4 5 4
0
0 0
Andrés Perea (Maastricht University) Linear Transformations Period 2 5 / 26
Inverse Transformation

Consider a linear transformation T : U ! V .

Then, T is one-to-one if for every û, v̂ 2 U with û 6= v̂ we have that

T (û ) 6= T (v̂ ).

T is onto if for every v̂ 2 V there is some û 2 U with T (û ) = v̂ .

02 31
x
x y
Example: T : R3 ! R2 with T @4 y 5A = is onto
x +z
z
(check
02 this),31 but not 0one-to-one,
2 31 since
1 0
0
T @4 1 5A = T @4 0 5A = .
0
1 0

Andrés Perea (Maastricht University) Linear Transformations Period 2 6 / 26

 Consider a linear transformation T : U ! V .

Then, T is one-to-one if for every û, v̂ 2 U with û 6= v̂ we have that

T (û ) 6= T (v̂ ).

T is onto if for every v̂ 2 V there is some û 2 U with T (û ) = v̂ .

3 2
2x
x
Example: T : R2 ! R3 with T = 4 2x 5 is
y
x + 4y
2
one-to-one 3 this), but not onto, since there is no û 2 R with
2 (check
1
T (û ) = 2 5 .
4
0

Andrés Perea (Maastricht University) Linear Transformations Period 2 7 / 26

 Then, T is one-to-one if for every û, v̂ 2 U with û 6= v̂ we have that
T (û ) 6= T (v̂ ).
T is onto if for every v̂ 2 V there is some û 2 U with T (û ) = v̂ .

If T is one-to-one and onto, then for every v̂ 2 V there is a exactly

one û 2 U with T (û ) = v̂ .
In this case, inverse transformation T 1 : V ! U is given by
1
T (v̂ ) = û, where û 2 U and T (û ) = v̂ .
x x +y
Example: T : R2 ! R2 with T = is one-to-one
y x y
and onto.
Inverse 1 : R2 ! R2 with
0 transformation:
1 T
B C 1
1 B x C 2x + 12 y
T B C= 1 1 .
@ y A 2x 2y
| {z } | {z }
v̂ û
Andrés Perea (Maastricht University) Linear Transformations Period 2 8 / 26
Theorem
If T : U ! V is a linear transformation that is one-to-one and onto, then
T 1 : V ! U is a linear transformation also.

Proof: Show that T 1 (û + v̂ ) = T 1 (û ) + T 1 (v̂ ) for all û, v̂ 2 V .

We have: T (T 1 (û + v̂ )) = û + v̂ .
Also, T (T 1 (û ) + T 1 (v̂ )) = T (T 1 (û )) + T (T 1 (v̂ )) = û + v̂ .
Hence, T (T 1 (û + v̂ )) = T (T 1 (û ) + T 1 (v̂ )).

Since T is one-to-one, T 1 (û + v̂ ) = T 1 (û ) + T 1 (v̂ ).

Try to show that T 1 (λ û ) = λ T 1 (û ).

Andrés Perea (Maastricht University) Linear Transformations Period 2 9 / 26

Nullspace and Range

Consider a linear transformation T : U ! V .

Nullspace (or kernel): N (T ) = fû 2 U j T (û ) = 0̂V g.

Range (or image):

R(T ) = fv̂ 2 V j there is û 2 U with T (û ) = v̂ g.

Theorem
Let T : U ! V be a linear transformation. Then,
(a) N (T ) is a subspace of U, and
(b) R(T ) is a subspace of V .

Andrés Perea (Maastricht University) Linear Transformations Period 2 10 / 26

Theorem
Let T : U ! V be a linear transformation. Then,
(a) N (T ) is a subspace of U, and
(b) R(T ) is a subspace of V .

Proof: (a) Show that N (T ) is closed under addition and scalar

multiplication:
Take û, v̂ 2 N (T ). Then, T (û ) = 0̂V and T (v̂ ) = 0̂V . Hence,
T (û + v̂ ) = T (û ) + T (v̂ ) = 0̂V + 0̂V = 0̂V .
Hence, û + v̂ 2 N (T ).
Try to show that N (T ) is closed under scalar multiplication.
(b) Take v̂1 , v̂2 2 R(T ). Then, there are û1 , û2 2 U with T (û1 ) = v̂1
and T (û2 ) = v̂2 .
Hence, v̂1 + v̂2 = T (û1 ) + T (û2 ) = T (û1 + û2 ) 2 R(T ).
Try to show that R(T ) is closed under scalar multiplication.
Andrés Perea (Maastricht University) Linear Transformations Period 2 11 / 26
Representation with Respect to Standard Basis

Consider the linear

2 transformation
3 T : R2 ! R3 given by
2x + y
x 4
T = x + 3y 5 .
y
3x + 2y
2 3
2 1
x x
Then, T =4 1 3 5 .
y y
3 2
| {z }
A
1
Moreover, …rst column of A is T , and second column of A
0
0
is T .
1

That is, A = [T (ê1 ), T (ê2 )].

Andrés Perea (Maastricht University) Linear Transformations Period 2 12 / 26

Theorem
Let T : Rn ! Rm be a linear transformation, and let fê1 , ..., ên g be the
standard basis for Rn . Then,
T (û ) = A û, where
A is the m n matrix given by A = [T (ê1 ), ..., T (ên )].

3 2
Example:
02 1 T : R ! R be given by
3Let
x
2x 4y
T @4 y 5A = .
3x + 5z
z
2 4 0
Then, T (ê1 ) = , T (ê2 ) = and T (ê3 ) = .
3 0 5
02 31 2 3
x x
2 4 0
Hence, T @4 y 5A = 4 y 5.
3 0 5
z z

Andrés Perea (Maastricht University) Linear Transformations Period 2 13 / 26

Theorem
Let T : Rn ! Rm be a linear transformation, and let fê1 , ..., ên g be the
standard basis for Rn . Then,
T (û ) = A û, where
A is the m n matrix given by A = [T (ê1 ), ..., T (ên )].

In this case:

N (T ) = fû 2 U j T (û ) = 0̂V g = fû 2 U j A û = 0̂V g

= nullspace of A.
Also,

R(T ) = fv̂ 2 V j there is û 2 U with T (û ) = v̂ g

= fv̂ 2 V j there is û 2 U with A û = v̂ g
= column space of A.
Hence, dim(N (T )) + dim(R(T )) = dim(U ).
Andrés Perea (Maastricht University) Linear Transformations Period 2 14 / 26
 Example:
02 3Consider
1 the linear transformation T : R3 ! R2 with
x
2x 4y
T @4 y 5A = .
4x + 6z
z

Find a basis for N (T ) and R(T ).

02 31 2 3
x x
2 4 0
Note that T @4 y 5A = 4 y 5.
4 0 6
z | {z } z
A
Hence, N (T ) = nullspace of A and R(T ) = column space of A.

Transform A into row-reduced form:

2 4 0 1 2 0
! 3 .
4 0 6 0 1 4

Andrés Perea (Maastricht University) Linear Transformations Period 2 15 / 26

 Transform A into row-reduced form:
2 4 0 1 2 0
! 3 .
4 0 6 0 1 4

Basis for N (T ) is basis for nullspace of A.

3 3
Second equation: y + 4z = 0. Choose z = α. Then, y = 4 α, and
x = 32 α.
2 3 2 3
3 2 3
3
x 2α 2
Hence, 4 y 5 = 4 3
4α
5=α 4 3
4
5.
z α 1
82 3
39
< 2 =
So, basis for N (T ) is 4 3
4
5 and dim(N (T )) = 1.
: ;
1

Andrés Perea (Maastricht University) Linear Transformations Period 2 16 / 26

 Transform A into row-reduced form:
2 4 0 1 2 0
! 3 .
4 0 6 0 1 4

Basis for R(T ) is basis for column space of A.

Hence, basis for R(T ) are the …rst two columns in A, that is,
2 4
, .
4 0

As such, dim(R(T )) = 2.

Andrés Perea (Maastricht University) Linear Transformations Period 2 17 / 26

Theorem
Let T : Rn ! Rm be a linear transformation, and let fê1 , ..., ên g be the
standard basis for Rn . Then,
T (û ) = A û, where
A is the m n matrix given by A = [T (ê1 ), ..., T (ên )].

Suppose that T : Rn ! Rn is invertible, that is, one-to-one and onto.

Let B be the matrix that corresponds to T 1.

Then,
1 1
û = T (T (û )) = T (A û ) = B A û,
and hence B A = In .

Thus, B = A 1.

The matrix that corresponds to T 1 is A 1.

Andrés Perea (Maastricht University) Linear Transformations Period 2 18 / 26

Representation with Respect to Di¤erent Basis
Example: Consider linear transformation T : R2 ! R3 given by
2 3
x + 2y
x
T = 4 3x 4y 5 .
y
y
If C is the standard basis for R2 and D is the standard basis for R3 ,
then the associated matrix is
2 3
1 2
[T ]D
C = [T (ê1 ), T (ê2 )] =
4 3 4 5.
0 1

Here, [T ]D
C describes how the standard basis vectors in C are mapped
by T into linear combinations of the standard basis vectors in D.
We have seen: T (û ) = [T ]D
C û.

Or, equivalently, [T (û )]D = [T ]D

C [û ]C .
Andrés Perea (Maastricht University) Linear Transformations Period 2 19 / 26
 Let B = fb̂1 , b̂2 g be a di¤erent basis for R2 , and E = fê1 , ê2 , ê3 g a
di¤erent basis for R3 .

Then, [T ]EB = T (b̂1 E , T (b̂2 E describes how the basis vectors

in B are mapped by T into linear combinations of basis vectors in E .

We have: [T (û )]E = [T ]EB [û ]B .

Let PB = b̂1 , b̂2 be the transition matrix from B to C , and

PE = [ê1 , ê2 , ê3 ] the transition matrix from E to D.

Then: [T (û )]D = [T ]D D

C [û ]C = [T ]C PB [û ]B .

Also: [T (û )]D = PE [T (û )]E = PE [T ]EB [û ]B .

Hence, [T ]D E
C PB [û ]B = PE [T ]B [û ]B .

Thus, [T ]D E E
C PB = PE [T ]B , and hence [T ]B = PE
1
[T ]D
C PB .

Andrés Perea (Maastricht University) Linear Transformations Period 2 20 / 26

Theorem
Let T : U ! V be a linear transformation. Let C and D be the standard
bases for U and V , respectively. Let B be an alternative basis for U and E
an alternative basis for V .
Then, for every û 2 U, we have [T (û )]E = [T ]EB [û ]B where
[T ]EB = PE 1 [T ]DC PB .
2 3
x + 2y
x
Example: Consider T : R2 ! R3 with T = 4 3x 4y 5 .
y
y
1 3
Let B = , be a di¤erent basis for R2 , and
2 0
82 3 2 3 2 39
< 1 0 1 =
E = 4 1 , 1 , 0 5 a di¤erent basis for R3 .
5 4 5 4
: ;
0 1 1
2 3 2 3
1 2 1 0 1
1 3
[T ]D
C =
4 3 4 5 , PB = , PE = 4 1 1 0 5 .
2 0
0 1 0 1 1
Andrés Perea (Maastricht University) Linear Transformations Period 2 21 / 26
Theorem
Let T : U ! V be a linear transformation. Let C and D be the standard
bases for U and V , respectively. Let B be an alternative basis for U and E
an alternative basis for V .
Then, for every û 2 U, we have [T (û )]E = [T ]EB [û ]B where
[T ]EB = PE 1 [T ]D
C PB .
2 3 2 3
1 2 1 0 1
D 4 1 3
[T ]C = 3 4 5 , PB = , PE = 4 1 1 0 5 .
2 0
0 1 0 1 1

Hence, [T ]EB = PE 1 [T ]D PB =
2 1 1 1
3 2 C 3 2 3
2 2 2 1 2 1 6
4 1 1 5 4 1 3
2
1
2 2 3 4 5 =4 4 3 5.
1 1 1 2 0
2 2 2 0 1 6 3

Andrés Perea (Maastricht University) Linear Transformations Period 2 22 / 26

Theorem
Let T : U ! V be a linear transformation. Let C and D be the standard
bases for U and V , respectively. Let B be an alternative basis for U and E
an alternative basis for V .
Then, for every û 2 U, we have [T (û )]E = [T ]EB [û ]B where
[T ]EB = PE 1 [T ]D
C PB .
2 3
1 6
We have seen: [T ]EB = 4 4 3 5.
6 3

5
Take û 2 U with [û ]B = .
7
2 3 2 3
1 6 37
5
Then, [T (û )]E = [T ]EB [û ]B = 4 4 3 5 = 4 1 5.
7
6 3 9

Andrés Perea (Maastricht University) Linear Transformations Period 2 23 / 26

Composition of Linear Transformations

Consider a linear transformation T1 : U ! V and a linear

transformation T2 : V ! W .
Then, T2 T1 is the composed (or combined) linear transformation
from U to W given by
(T2 T1 )(û ) = T2 (T1 (û )) for every û 2 U.
Let T1 (û ) = A1 û and T2 (v̂ ) = A2 v̂ .
Then, (T2 T1 )(û ) = T2 (T1 (û )) = A2 T1 (û ) = A2 A1 û.
Hence, the matrix associated with T2 T1 with respect to the
standard bases is A2 A1 .
In general, if B, C and D are (possibly non-standard) bases for U, V
and W , then
[T2 T1 ]D D C
B = [T2 ]C [T1 ]B .
Andrés Perea (Maastricht University) Linear Transformations Period 2 24 / 26
 3 3
Example: 31 2 T1 : R ! R 3with
02 Consider
x 2x 3y + z
T1 @ 4 y 5 A = 4 y + 2z 5 , and
z x 4y
02 31
x
3x y + z
T2 : R3 ! R2 with T2 @4 y 5A = .
x + 2y + 2z
z
What is T2 T1 ?

Let C be the standard basis for R3 and D the standard basis for R2 .
2 3
2 3 1
C 3 1 1
4
Then, [T1 ]C = 0 1 2 5 and [T2 ]DC = .
1 2 2
1 4 0

Andrés Perea (Maastricht University) Linear Transformations Period 2 25 / 26

 2 3
2 3 1
3 1 1
[T1 ]CC = 4 0 1 2 5 and [T2 ]D
C = .
1 2 2
1 4 0
Thus, 2 3
2 3 1
3 1 1
[T2 T1 ]D D C
C = [T2 ]C [T1 ]C =
4 0 1 2 5=
1 2 2
1 4 0
7 14 1
.
0 3 3
02
31
x
7x 14y + z
Hence, (T2 T1 ) @4 y 5A = .
3y + 3z
z

Andrés Perea (Maastricht University) Linear Transformations Period 2 26 / 26