Linear Transformations

(Chapter 5)
Definition :
Let U and V be real vector spaces. A
map L: U V from U to V is called a
linear map, or Linear transformation
if the following holds:

(i ) L u v L u L v for all u , v U
(ii) L(ku) =kL(u) for all u U
and all real numbers k
 Example of L.T.

Reflection about x-axis L(x,y)=

(x,-y)

Projection on y-axis
L(x,y)=(0,y)
These are L.T.
Example: Let L: C [-1, 1] C [-1, 1] be
defined by
L( g(x) ) = x g (x) for all g(x) C[-1, 1].
show that L is a linear map.
Sol.: Let h(x), g(x) C[-1,1]
For condn.(i): L( h (x) + g(x))
= x ( h(x)+g(x) )'
= x(h '(x)+g'(x) )
=xh'(x)+xg'(x)=L(h(x) )+ L(g(x))
For condn.(ii):
L(g(x)) = x [g (x)]' =x(g '(x))= L(g(x))
Hence L is linear transformation
Example: Let a map L: R2R3 be defined by

L(x, y) = (x, y , x y). Check whether L is linear.

Sol.: For any real number r:
L (r(x,y)) = L(rx, ry)
= (rx, ry , rx ry )
= (rx, ry , r2xy)
= r(x, y, rxy) r(x, y ,x y).
Hence L is not linear.
 Trivial Example:

Zero linear operator :

z(u)=0U,uU, is linear

Identity linear operator from

(U to U):
I(u) = u is linear u U,
Thm

 Theorem :

Let L: UV be a linear map, then

L(r1u1++rnun)=
r1L(u1)++ rnL(un)
for any vectors u1, , un in U and
any scalars r1,, rn
 Theorem :
Let L: UV be a linear map, then
i. L(0U)=0V, ii. L(-u)=-L(u)

ii. L(u-v) = L(u)-L(v)

Proof-(i):
L(ku)=kL(u)
for any scalar k since L is L.T
Take k=0:
 L(ku)=kL(u)
implies L(0u)=0L(u)
Implies L(0u)=0v
since 0u=0u (element of U)
and 0L(u)=0v (element of V)
Proof-(ii):
Similarly prove others.
Discussion of (i):

Let L: UV be a linear map, then

i. L(0U)=0V,
In other words, a linear map L transforms
the 0U into the 0V .
Converse is not true.
Exam. L:R3 R1 defined by
L(x,y,z)=x2+y2+z2 is not L.T.
Corollary:
Let L: UV be a function, if
L(0U) 0V , then L is not L.T.
This can be used for showing L is
not L.T
 Theorem :

Let L: UV be a L.T. of an n-
dimensional vector space U
into a vector space V. Also let
S ={u1, u2, , un} be a basis
for U. If u is any element of U,
L(u) is completely determined
by {L(u1), L(u2), , L(un)}
 Proof.:
Then any vector uU can be
expressed uniquely in the
form
u = 1u1+ 2u2++ nun.
So, if L:UV is a linear map,
then
L(u) =L(1u1+2u2++ nun)
= 1L(u1)+ 2L(u2)++ nL(un).

Thus L(u) is completely determined by

L(u1), L(u2),L(un)
i.e, L(u) is known as soon as
L(u1), L(u2),L(un) are known,
Ex. Let L : R2R2 be a L.T.,
L(1,1)=(3,0),
L(-1,1)= (0,1) what is L(x,y)?
Sol. {(1,1), (-1,1)} is linearly
independent. hence {(1,1), (-1,1)}
is basis of R2

To find linear transformation we

express (x,y) as linear combination of
{(1,1), (-1,1)}.
(x,y) = (1,1)+(-1,1)
= (-,+)
-=x & +=y
Solving we get

=(x+y)/2 & = (y-x)/2

= ( (3x+3y)/2, (y-x)/2 )
 Composite Linear Transformations

Definition:
Let V1 , V2 and V3 are vector
spaces with dimension n, m & p,
respectively.
Let L1: V1 V2 and L2 : V2 V3
be linear transformations.
 Definition (Cont.)
The function L2 o L1 : V1 V3
defined by
( L2 o L1) (u) = L2(L1(u)) for u in V1
is called the composite of L2 with
L1.
 Theorem

Let : L1 : V1 V2 and L2 : V2 V3
be linear transformations.
Then L2 o L1 : V1 V3 is a linear
transformation
Proof. ( not required):
 Example
Let L1 : P2 P2 and L2 : P2 P2
be defined by
L1 ( at2 + bt + c ) = 2at + b
L2 ( at2 + bt + c ) = 2at2 + bt

Compute (i) L2 o L1 (ii) L1 o L2

 Solution
(i) (L2 o L1) (at2 + bt + c )
= L2( L1( at2 + bt + c))
= L2( 2at + b ) = 2at
(ii) (L1 o L2 ) ( at2 + bt + c)
= 4 at + b

Note : L2 o L1 L1 o L2
Dimensional Theorem

(Section 5.3)
 Definition of Kernel:
Let L: UV be a L.T.

Then kernel (null space) of L

is the set
ker L ={uU : L(u)=0v}.
In other words, ker L is the set of all
those elements in U that are mapped
by L into the zero vector of V.
 Definition of Range space:

Let L:UV be a linear map then

Range of L is
range(L)= {v V | L(u) = v for
some uU}
Thus a vector v is in range(L) if
there exists some vector u in U such that
L(u) = v.
If range L = V, we say that L is onto.
Thm. Let L:UV be a L.T.,
then ker (L) is a subspace
of U

Thm. Let L:UV be a L.T., then

range (L) is a subspace of V


 Dimensional Theorem
Let L:UV be a L.T. and U is finite
dimensional, then range(L) is finite
dimensional, and

dim(Ker(L))+dim(range(L))=dim(U)
Definition. Let U and V be vector spaces over the field R
and L:UV be a linear transformation. If U is finite
dimensional, then

rank of L= dim R(L), and

nullity of L=dim N(L)=Ker(L).

Linear Algebra 33
Example : Let L:R3 R4 be a linear map,
defined by
L x, y , z x, y z , x y z , x y z .
Find the basis for Ker(L) and range(L).
L( X ) AX
where
1 0 0
0 1 1
A
1 1 1

1 1 1 RRE of A is
1 0 0
Ker ( L) X R 3 : L( X ) 0 0 1 1
B
= X R : AX 0
3 0 0 0

0 0 0
The equivalent system is
x =0 & y = z .
 ker(L) ={(0,z, z) : z is arbitrary}
={z(0,1,1): z is arbitrary} =span{(0,1,1)}
Basis is ={(0,1,1)}
R(L)=
{ (x, y - z, x -y +z, x +y -z):(x,y, z)R3}

={(x,0,x,x)+(0,y,-y,y)+(0,-z,z,-z): x,y,z R}]

={x(1,0,1,1)+y(0,1,-1,1)+z(0,-1,1,-1): x,y,z
R}]
R(L)={x(1,0,1,1)+y(0,1,-1,1)+z(0,-1,1,-1):
x,y,z R}]

=[{(1,0,1,1), (0,1,-1,1),(0,-1,1,-1)}]
Is the set
S ={(1,0,1,1), (0,1,-1,1),(0,-1,1,-1)} LI?

Basis for R(L) is ={(1,0,1,1), (0,1,-1,1)}

WHY??
 (Section 5.4)

One-to-One and Onto

Linear Transformation
 Definition:
One-to-one:
Let L:UV be a L.T. is said to be
one-to-one if for all u1, u2 in U,
u1 u2 implies L(u1) L(u2).
In other way: L is one-to-one if
for all u1, u2 in U, L(u1)=L(u2)
implies u1=u2
 Theorem
A linear transformation L : U V
is one to-one if and only if
ker(L) = {0U} or, dim (Ker
L)=0.
Proof.: Let L be one-to-one and
u be in ker (L).
Then L(u) = 0V.
Also, L (0U) = 0V since L is L.T
Thus, L(u) = L(0U).
Which implies u = 0U
Since L is one to one,
Hence ker (L) = {0U} .
Conversely, suppose that ker(L)= {0U}.
Let L(u) = L(v), for u and v in U.
Then L(u) L(v) = 0V.
So L(u-v) = 0V
Implies u-v is in ker L .
Therefore , u v = 0U ,
so u = v.
Thus L is one-to-one
Remark 1. Suppose L : U V is a linear transformation and dim U .
Then dim RL dim U , because dim RL dim N L dim U .

Remark 2. If L : U V is a linear transformation and

dim U . Then dim RL min dim U , dim V .

Proof : Since RL is a subspace of V

dim RL dim V .

From Remark 1, dim RL dim U

Hence dim RL min dim U , dim V .
Linear Algebra 44
Remark 3. If L : U V is a linear transformation
and dim U . Then dim RL dim U L is one-one.

Proof: dim RL dim N L dim U

dim RL dim U dim N L 0

N L 0U
L is one one.

Linear Algebra 45
Remark 4. If dim RL dim U , then T is not one-one.
It follows from Remark 3.

Remark 5. L is onto RL V
dim RL dim V

Remark 6. dim RL dim U dim V L is one one

and onto.
Linear Algebra 46
Two more remarks:

(Section 5.5)

Isomorphism
Invertible Linear Transformation

Definition:
A L.T. L : V W is called invertible
if there exists a unique function
L-1 : W V such that
L-1 o L = Iv and L o L-1 = Iw.
The function L-1 is called inverse of
L
Definition
A L.T. L : V W that is both
one-to-one and onto is called
isomorphism from V to W
 Theorem
A L.T. L : V W is isomorphism
if and only if L is invertible.
Moreover
L-1 is a L.T. and (L-1)-1=L
 Theorem
Let L : V W be a L.T. and
dim V = dim W<.
Then L is one-to-one
if and only if L is onto
Proof:

 Exam. Let L:R 3R3 be defined by

x 1 0 1 x

L y 1 1 2 y AX
z 2 1 3 z

(a) Is L onto?
(b) Find ker L
(c) Is L one-to-one?

Conclusion
L is NOT one-one
L is NOT onto

Check that L is one-one and
onto.