Linear Transformation

T :V !W
Def: Let V and W be vector spaces over the same …eld F . We call a function
T : V ! W a linear transformation from V to W if, for all x; y 2 V and c 2 F ,
we have
a. T (x + y) = T (x) + T (y), preserving vector addition
b. T (cx) = cT (x), preserving scalar multiplication

Properties of a linear function T

1. If T is linear, then T (0v ) = 0w .
2. T is linear i¤ T (cx + y) = cT (x) + T (y) for all x; y 2 V and c 2 F .
3. If T is linear, then T (x y) = T (x) T (y) for all x; y 2 V .
4. T is linear i¤, for x1 ; :::; xn 2 V and a1 ; :::; an 2 F ,
Pn P
n
T ( ai xi ) = ai T (xi )
i=1 i=1

Ex: De…ne T : R2 ! R2 by T (a1 ; a2 ) = (2a1 + a2 ; a1 )

To show that T is linear, let c 2 R, and x; y 2 R2
where x = (b1 ; b2 ) and y = (d1 ; d2 )
cx + y = c(b1 ; b2 ) + (d1 ; d2 ) = (cb1 + d1 ; cb2 + d2 )

T (cx + y) = (2cb1 + 2d1 + cb2 + d2 ; cb1 + d1 )

Also
cT (x) + T (y)
= cT (b1 ; b2 ) + T (d1 ; d2 )
= c(2b1 + b2 ; b1 ) + (2d1 + d2 ; d1 )
= (2cb1 + cb2 ; cb1 ) + (2d1 + d2 ; d1 )
= (2cb1 + cb2 + 2d1 + d2 ; cb1 + d1 )

Ex: For any angle , de…ne T : R2 ! R2 by the rule: T (a1 ; a2 ) is the

vector obtained by rotating (a1 ; a2 ) counterclockwise by if (a1 ; a2 ) 6= (0; 0),
and T (0; 0) = (0; 0). Then T : R2 ! R2 is a linear transformation that is
called rotation by .

Determine an explicit formula for T . Fix a nonzero vector (a1 ; a2 ) 2 R2 .

1
is the angle (a1 ; a2 ) makes with the positive x-axis, and let r = (a21 + a22 ) 2 .
Then a1 = r cos and a2 = r sin .
T (a1 ; a2 ) has length r and makes an angle + with the positive x-axis.
(a01 ; a02 ) = T (a1 ; a2 ) = [r cos( + ); r sin( + )]
= (r cos cos r sin sin ; r cos sin + r sin cos )
= (a1 cos a2 sin ; a1 sin + a2 cos )
a0 a cos a2 sin
T (a1 ; a2 ) = 10 = 1
a2 a1 sin + a2 cos
a01 cos sin a1
=
a02 sin cos a2

1
 This is the matrix representation of rotation by an angle in R2 .

(a1 ; a2 ); (b1 ; b2 )
show T (c(a1 ; a2 ) + (b1 ; b2 )) = cT (a1 ; a2 ) + T (b1 ; b2 )
c(a1 ; a2 ) + (b1 ; b2 ) = (ca1 ; ca2 ) + (b1 ; b2 )
= (ca1 + b1 ; ca2 + b2 )
T (c(a1 ; a2 ) + (b1 ; b2 )) = T (ca1 + b1 ; ca2 + b2 )
= ((ca1 + b1 ) cos (ca2 + b2 ) sin ;
(ca1 + b1 ) sin + (ca2 + b2 ) cos )

according to the transformation

T (a1 ; a2 ) = (a1 cos a2 sin ; a1 sin + a2 cos )

cT (a1 ; a2 ) = c(a1 cos a2 sin ; a1 sin + a2 cos )

= (ca1 cos ca2 sin ; ca1 sin + ca2 cos )

T (b1 ; b2 ) = (b1 cos b2 sin ; b1 sin + b2 cos )

cT (a1 ; a2 ) + T (b1 ; b2 )
= (ca1 cos ca2 sin ; ca1 sin + ca2 cos )
+(b1 cos b2 sin ; b1 sin + b2 cos )
= (ca1 cos ca2 sin + b1 cos b2 sin ;
ca1 sin + ca2 cos + b1 sin + b2 cos )
= ((ca1 + b1 ) cos (ca2 + b2 ) sin ;
(ca2 + b2 ) cos + (ca1 + b1 ) sin )

((ca1 + b1 ) cos (ca2 + b2 ) sin ;

(ca1 + b1 ) sin + (ca2 + b2 ) cos )

Ex: De…ne T : R2 ! R2 by T (a1 ; a2 ) = (a1 ; a2 )

Re‡ection about x-axis.
1 0 a1 a1
=
0 1 a2 a2

Ex: De…ne T : R2 ! R2 by T (a1 ; a2 ) = (a1 ; 0). T i called projection on the

x-axis.

Ex: De…ne T : Mm n (F ) ! Mn m (F ) by T (A) = AT , where AT is trans-

pose of A.

Ex: Let V denote the set of all real-valued functions de…ned on the real line
that have derivative of every order.
De…ne T : V ! V by T (f ) = f 0 , the derivative of f . Let g; h 2 V and a 2 R.
Now T (ag + h) = (ag + h)0 = ag 0 + h0 = aT (g) + T (h)

Ex: Let V = C(R) the vector space of continuous real valued functions on
Rb
R. Let a; b 2 R, a < b. De…ne T : V ! R by T (f ) = a f (t)dt for all f 2 V .

2
 T is a linear transformation because the de…nite integral of a linear com-
bination of functions is the same as the linear transformation of the de…nite
integrals of the functions.
Rb Rb Rb
a
[cg(t) + h(t)]dt = c a g(t)dt + a h(t)dt

For vector spaces V and W over F

Def:
identity transformation IV : V ! V by IV (x) = x for all x 2 V .
zero transformation T0 : V ! W by T0 (x) = 0 for all x 2 V .

Def: Let V and W be vector spaces, and let T : V ! W be linear. Let the
Kernel (or Nullspace) N (T ) or ker(T ) of T to be the set of all vectors x in V
such that T (x) = 0; that is, ker(T ) = fx 2 V : T (x) = 0g.

Linear system ! homogeneous solution

Coe¢ cient matrix ! nullspace
Linear transformation ! kernel

Def: The range (or image or codomain) R(T ) or image(T ) of T to be

the subset of W consisting of all images (under T ) of vectors in V ; that is,
image(T ) = fT (x) : x 2 V g.
fx + y + z = 0g

1.0
-4 0.5 -4
-2 -2
z 0.0
0 0
-0.5
y-1.0 2
2 x
4 4

Ex: Let V and W be vector spaces, and let I : V ! V and T0 : V ! W

be the identity and zero transformations, respectively. Then ker(I) = f0g.
image(I) = V , ker(T0 ) = V , and image(T0 ) = f0g.

3
 Ex: T 2: R2 ! 3
3 R de…ned by
1 2
T x = 43 45 x
5 26 3 2 3
1 2 x1 + 2x2
x x
T 1 = 43 45 1 = 43x1 + 4x2 5
x2 x2
5 6 5x1 + 6x2
Check on preservation of vector addition
x x0
x = 1 ; x0 = 10
x2 x2
2 3 2 0 3
0 1 2 0 x1 + 2x02 + x1 + 2x2
x + x1 x + x1
T (x + x0 ) = T 1 = 43 45 1 = 43x01 + 4x02 + 3x1 + 4x2 5
x2 + x02 x2 + x02
2 0 3 5 6 5x01 + 6x02 + 5x1 + 6x2
0
x1 + x1 + 2x2 + 2x2
= 43x01 + 3x1 + 4x02 + 4x2 5
5x201 + 5x1 + 6x02 + 6x2 3
(x01 + x1 ) + 2(x02 + x2 )
= 43(x01 + x1 ) + 4(x02 + x2 )5
5(x01 + x1 ) + 6(x02 + x2 )
2 3 2 3 2 3
1 2 1 2 x01 + 2x02 + x1 + 2x2
x x0
T x + T x0 = 43 45 1 + 43 45 10 = 43x01 + 4x02 + 3x1 + 4x2 5 =
x2 x2
2 0 5 6 5 6 5x01 + 6x02 + 5x1 + 6x2
3
x1 + x1 + 2x02 + 2x2
43x01 + 3x1 + 4x02 + 4x2 5
5x01 + 5x1 + 6x02 + 6x2

Generally, any m n matrix gives rise to a linear transformation from Rn

to Rm
2 3
5 0 0
Ex: T : R3 ! R3 given by T x = 5x corresponds to the matrix 40 5 05
0 0 5

Ex: permutation as a linear transformation. Let us take R4 and consider

the operation of switching the …rst and third components
T (x1 ; x2 ; x3 ; x34 ) = (x3 ; x2 ; x1 ; x4 ) 2 3
2
0 0 1 0 x3
60 1 0 07 6 7
6 7 T 6x2 7
41 0 0 05 (x1 ; x2 ; x3 ; x4 ) = 4x1 5
0 0 0 1 x4

Ex: The right-shift operation as a linear transformation

R1 is the space of all sequences, and it contains (x1 ; x2 ; x3 ; x4 ; :::)
De…ne the right-shift operator U : R1 ! R1 by
U (x1 ; x2 ; x3 ; x4 ; :::) = (0; x1 ; x2 ; x3 ; x4 ; :::)

4
 Lemma: Let T : V ! W be a linear transformation. Then T is one-to-one
i¤ (if and only if) ker(T ) = f0g.
Proof:
Suppose that T is one-to-one, we need to show that ker(T ) = f0g. 0 2
ker(T ), because T 0v = 0w . Now show there is no other elements in ker(T ).
Prove by contradiction, there was a nonzero vector v 2 V such that v 2 ker(T ),
T v = 0 = T 0. Since T is one-to-one, this has to force v = 0, then we have a
contradiction now.

Now suppose that ker(T ) = f0g; we have to show that T is one-to-one. If

T v = T v 0 , then we must have v = v 0 . Suppose T v = T v 0 , then T v T v 0 = 0,
T (v v 0 ) = 0. v v 0 = 0 ! v = v 0 , this will imply that T is one-to-one.

Def: The rank rank(T ) of a linear transformation T : V ! W is de…ned to

be the dimension of image(T ), thus rank(T ) = dim(image(T )).

Ex: The zero transformation has rank 0.

Ex: (x1 ; x2 ; x3 ; x4 ; x5 ) := (x1 ; x2 ; x3 ; 0; 0)

image( ) = f(x1 ; x2 ; x3 ; 0; 0) : x1 ; x2 ; x3 2 Rg
rank( ) = 3

The rank measures how much information or degree of freedom is retained

by the transformation T .

The Replacement Theorem

Let V be a vector space that is generated by a set G containing exactly n
vectors, and let L be a linearly independent subset of V containing exactly m
vectors. Then m n and there exists a subset H of G containing n m vectors
such that L [ H generates V .
Proof: By induction on m.
Base case m = 0; L = ;, and so taking H = G gives the desired result.
Suppose the theorem is true for m 0, show it is true for m + 1.
Let L = fv1 ; :::; vm+1 g be a linearly independent subset of V containing
m + 1 vectors.
Use the following theorem:
Let V be a vector space, and let S1 S2 V . If S1 is linearly (in)dependent,
then S2 is linearly (in)dependent.
We may apply the induction hypothesis to conclude that m n and tat there
is a subset fu1 ; :::un m g of G such that fv1 ; :::; vm g [ fu1 ; :::; un m g generates
V . Thus there exist scalars a1 ; :::; am ; b1 ; :::; bn m such that
a1 v1 + ::: + am vm + b1 u1 + ::: + bn m un m = vm+1
n m > 0, let vm+1 be a linear combination of v1 ; :::; vm , by the following
theorem
Let S be a linearly independent subset of a vector space V , and let v be a
vector in V that is not in S. Then S [ fvg is linearly dependent if and only if
v 2 span(S).

5
 Contradicts the assumption that L is linearly independent. Hence n > m;
that is, n m + 1. Say b is nonzero
u1 = ( ab11 ) + ( ab12 )v2 + ::: + ( abm
1
)vm b11 vm+1
+( bb21 )u2 + ::: + ( bnb1m )un m
Let H = fu2 ; :::; un m g. Then u1 2 span(L[H), and because v1 ; :::; vm ; u2 ; :::; un m
are in the span(L [ H), it follows that
fv1 ; :::; vm ; u1 ; :::; un m g span(L [ H)
because fv1 ; :::; vm ; u1 ; :::; un m g generates V , by the following theorem
The span of any subset S of a vector space V is a subspace of V that contains
S.

implies that span(L [ H) = V . Since H is a subset of G that contains

(n m) 1 = n (m + 1) vectors, the theorem is true for m + 1.

The Dimension Theorem

Let V be a …nite-dimensional space, and let T : V ! W be a linear trans-
formation. Then
nullity(T ) + rank(T ) = dim(T )
corresponds to the rank theorem (version 1)
(number of free var)+(number of pivots)=(number of total var)
Proof: By hypothesis, dim(V ) is …nite; n := dim(V ). Since ker(T ) is a
subspace of V , and has dim k, then k := dim(ker(T )) = nullity(T ). 0 k n.
Now need to show k + rank(T ) = n, or dim(image(T )) = n k.

ker(T ) has a basis fv1 ; :::; vk g of k elements. Since this basis is in ker(T ),
so it is also in V , and the basis vectors are linearly independent, therefore it
must be part of a basis of V , which must have n = dim(V ). A basis for V is
fv1 ; :::; vn g; we can add n k extra basis vectors vk+1 ; :::; vn .

Since vk+1 ; :::; vn lie in V , the elements T vk+1 ; :::; T vn lie in image(T ). Now
we can claim that fT vk+1 ; :::; T vn g are a basis for image(T ), so image(T ) has
dimension n k.

Verify that fT vk+1 ; :::T vn g form a basis, we need to show that they span
image(T ) and are linearly independent. Every vector in image(T ) is a linear
combination of fT vk+1 ; :::T vn g. We pick a vector w from image(T ), and show
w is linear combination of fT vk+1 ; :::; T vn g. By de…nition of image, w = T v.

v = a1 v1 + ::: + an vn for scalars a1 ; :::; an

Apply T
w = T v = (a1 T v1 + a2 T v2 + ak T vk ) + ak+1 T vk+1 + ::: + an T vn

v1 ; :::; vk are in ker(T ), so T v1 = ::: = T vk = 0,

w = T v = ak+1 T vk+1 + ::: + an T vn now we have w as a linear combination
of the basis fvk+1 ; :::; vn g.

6
 Now we show that fvk+1 ; :::; vn g is linearly independent. Prove by contra-
diction that they are dependent,
ak+1 T vk+1 + ::: + an T vn = 0 for some scalars ak+1 ; :::; an which are not all
zeros.
T (ak+1 vk+1 + ::: + an vn ) = 0
because ak+1 vk+1 + ::: + an vn 2 ker(T ) and it is spanned by fv1 ; :::; vk g, we
have
ak+1 vk+1 + ::: + an vn = a1 v1 + ::: + ak vk for some scalars a1 ; :::; ak
a1 v1 ::: ak vk + ak+1 vk+1 + ::: + an vn = 0
since fv1 ; :::; vn g is linearly independent, that means all coe¢ cients ai must
be zero. This will lead to a contradiction with the hypothesis. Thus fT vk+1 ; :::; T vn g
must be linearly independent.

Ex: Let T : R2 ! R2 denote the linear transformation

T (x; y) = (x + y; 2x + 2y) = (x + y; 2(x + y))
Find the ker(T ) and image(T ).
ker(T ) = f(x; y) 2 R2 : x + y = 0)
dim(ker(T )) = 1
image(T ) = f(t; 2t) : t 2 Rg
dim(image(T )) = 1
dim(ker(T )) + dim(image(T )) = 1 + 1 = 2 = dim(T )

Lemma: Let V and W be …nite-dimensional vector spaces of the same di-

mension, dim(V ) = dim(W ), and let T : V ! W be a linear transformation
from V to W . Then T is one-to-one i¤ T is onto.

Proof: If T is one-to-one, then nullity(T ) = 0 from dimension theorem.

Since dim(V ) = dim(W ), we have dim(image(T )) = dim(W ). Since image(T )
is a subspace of W , image(T ) = W , thus T is onto.

Theorem: If T : V ! W be a linear transformation, and let fv1 ; :::; vn g span

V , then fT v1 ; :::; T vn g span image(T ).
Proof: we choose any vector w in image(T ), and show that w can be written
as a linear combination of T v1 ; :::; T vn .
w = T v for some v 2 V . Since fv1 ; :::; vn g span V , v = a1 v1 + ::: + an vn .
Apply T to both sides, w = T v = T (a1 v1 + ::: + an vn ) = a1 T v1 + ::: + an T vn
w is now written as a linear combination of T v1 ; :::; T vn .

Theorem: If T : V ! W is a linear transformation which is one-to-one, and

fv1 ; :::; vn g is linearly independent, then fT v1 ; :::; T vn g is also linearly indepen-
dent.
Proof: suppose 0 = a1 T v1 + ::: + an T vn
we need to show that a1 ; :::; an are all zeros.
0 = T (a1 v1 + ::: + an vn )
Since T is one-to-one, ker(T ) = f0g, and thus

7
 0 = a1 v1 + ::: + an vn
But since fv1 ; :::; vn g is linearly independent, implies a1 ; :::; an are all zero.

Corollary: If T : V ! W is both one-to-one and onto, and fv1 ; :::; vn g is

a basis for V , then fT v1 ; :::; T vn g is a basis for W . In particular, dim(V ) =
dim(W ).
Proof: Since fv1 ; :::; vn g is a basis for V , it spans V . Then fT v1 ; :::; T vn g
spans image(T ).
T is onto, image(T ) = W .
fv1 ; :::; vn g is linearly independent and T is one-to-one, then fT v1 ; :::; T vn g
are linearly independent. Now the two requirements for basis are both satis…ed.

The converse is also true: if T : V ! W is one-to-one, and fT v1 ; :::; T vn g is

a basis, then fv1 ; :::; vn g is also a basis.

Theorem: Let V be a …nite-dimensional vector space, and let fv1 ; :::; vn g be

a basis for V . Let W be another vector space, and let w1 ; :::; wn be some vectors
in W . Then there exists exactly one linear transformation T : V ! W such
that T vj = wj for each j = 1; :::; n.

De…nition: Let V be a …nite dimensional vector space. An ordered basis of

V is an ordered sequence (v1 ; :::; vn ) of vectors in V such that the set fv1 ; :::; vn g
is a basis.

Ex: The sequence ((1; 0; 0); (0; 1; 0); (0; 0; 1)) is an ordered basis for R3 ; the
sequence ((0; 1; 0); (1; 0; 0); (0; 0; 1)) is a di¤erent ordered basis for R3 .

Ex: v :=
2 (3;
3 4; 5). If := ((1; 0; 0); (0; 1; 0); (0; 0; 1)), then
3
[v] = 445
5
and (3; 4; 5) = 3(1; 0; 0) + 4(0; 1; 0) + 5(0; 0; 1)

2
Ex: P2 (R),
2 3 and let f = 3x + 4x + 6. If := (1; x; x2 ), then
6
[f ] = 445
3

Def: Let V and W be vector spaces, and let S : V ! W and T : V ! W be

two linear transformations from V to W . We de…ne the sum of S + T of these
transformations to be a third transformation S + T : V ! W de…ned by
(S + T )(v) := Sv + T v

Def: Let T : V ! W be a linear transformation, and let c be a scalar.

We de…ne the scalar multiplication cT of c and T to be the transformation
cT : V ! W , de…ned by

8
 (cT )(v) = c(T v)

Def: Let L(V; W ) be the space of linear transformations from V to W .

Lemma: The space L(V; W ) is a subspace of F (V; W ), the space of all func-
tions from V to W . In particular, L(V; W ) is a vector space.

Def: Let U; V; W be vector spaces. Let S : V ! W be a linear transformation

from V to W , and let T : U ! V be a linear transformation from U to V . Then
we de…ne the product or composition ST : U ! W to be the transformation
ST (u) := S(T (u))

Ex: Let U : R1 ! R1 be the right shift operator

U (x1 ; x2 ; :::) := (0; x1 ; x2 ; :::)
Then the operator U U = U 2 given by
U 2 (x1 ; x2 ; :::) = U (U (x1 ; x2 ; :::))
= U (0; x1 ; x2 ; :::) = (0; 0; x1 ; x2 ; :::)

Matrix Representation of Linear Transformations

Let V and W be …nite-dimensional vector spaces, and let := (v1 ; :::; vn ) and
:= (w1 ; :::; wm ) be ordered bases for V and W , respectively; thus fv1 ; :::; vn g
is a basis for V and fw1 ; :::; wm g is a basis for W , so that V is n-dimensional
and W is m-dimensional. Let T be a linear transformation from V to W .

Ex: Let V = P3 (R), W = p2 (R), and T : V ! W be the di¤erentiation map

T f := f 0 . We use the standard order basis := (1; x; x2 ; x3 ) for V , and the
standard ordered basis := (1; x; x2 ) for W .

2
Ex: pick
2 a3 v 2 P3 (R), v = 3x + 7x + 5
5
677
[v] = 6435
7

0
d
Then we have T v = dx v = 6x + 7
so that 2 3
7
[T v] = 465
0

General base
v = x1 v1 + ::: + xn vn
Let us apply T to both sides
T v = x1 T v1 + ::: + xn T vn
from the formula [T v] we have

9
 T v = y1 w1 + ::: + ym wm
The vectors T v1 ; :::; T vn lie in W , and so they are linear combinations of
w1 ; :::; wm
T v1 = a11 w1 + a21 w2 + ::: + am1 wm
T v2 = a12 w1 + a22 w2 + ::: + am2 wm
:::
T vn = a1n w1 + a2n w2 + ::: + amn wm

Lemma: Let V; W be …nite-dimensional spaces with ordered bases , re-

spectively. Let S : V ! W and T : V ! W be linear transformations from V
to W , and let c be a scalar. Then
[S + T ] = [S] + [T ]
[cT ] = c[T ]
Proof on the scalar multiplication part:
Let = (v1 ; :::; vn ) and = (w1 ; :::; wm ), and denote the matrix [T ] by
2 3
a11 a12 ::: a1n
6 a12 a22 ::: a2n 7
[T ] := 64
7
5
:::
am1 am2 ::: amn
T v1 = a11 w1 + a21 w2 + ::: + am1 wm
T v2 = a12 w1 + a22 w2 + ::: + am2 wm
:::
T vn = a1n w1 + a2n w2 + ::: + amn wm
multiply by c
(cT )v1 = ca11 w1 + ca21 w2 + ::: + cam1 wm
(cT )v2 = ca12 w1 + ca22 w2 + ::: + cam2 wm
:::
(cT )vn = ca1n w1 + ca2n w2 + ::: + camn wm
2 3
ca11 ca12 ::: ca1n
6 ca21 ca22 ::: ca2n 7
[cT ] = 6
4
7
5
:::
cam1 cam2 ::: camn
[cT ] = c[T ]

Def: Let T : V ! W be a linear transformation. We say that a linear

transformation S : W ! V is the inverse of T if T S = IW amd ST = IV .
We say that T is invertible if it has an inverse, and call the inverse T 1 ; thus
T T 1 = IW and T 1 T = IV .

x1 x x2 y 2y1
Ex: T = 1 ;S 1 =
x2 x1 + x2 y2 y2
S T [a]
x x x2 2(x1 x2 )
S T 1 =S 1 =
x2 x1 + x2 (x1 + x2 )

10
 2 2 x1
=
1 1 x2
1 1 2 0
T = ;S
1 1 0 1
2 0 1 1 2 2
ST = =
0 1 1 1 1 1
Lemma: Let T : V ! W be a linear transformation, and let S : W ! V
and S 0 : W ! V both be inverses of T . Then S = S 0 .
Proof: S = SIW = S(T S 0 ) = (ST )S 0 = IV S 0 = S 0

Lemma: If T : V ! W has an inverse S : W ! V , then T must be one-to-

one and onto.
Proof: First show that T is one-to-one. Suppose that T v = T v 0 ; we have to
show that v = v 0 . Apply S on both sides we get ST v = ST v 0 ! IV v = IV v 0 ,
thus v = v 0 as desired. Now we show that T is onto. Let w 2 W ; we have to
…nd v such that T v = w. But w = IW w = T Sw = T (Sw), so if we let v := Sw
then we have T v = w as desired.
The converse of this lemma is also true.

Lemma: If T : V ! W is one-to-one and onto linear transformation, then it

has an inverse S : W ! V , which is also a linear transformation.
Proof: Let T : V ! W be one-to-one and onto. Let w be an element of W .
Since T is onto, we have w = T v for some v in V ; since T is one-to-one; this v
is unique. De…ne Sw as equal to this v, thus S is a transformation from W to
V . For any w 2 W , we have w = T v and Sw = v for some v 2 V , and hence
T Sw = w; thus T S is the identity IW :
Now show that ST = IV , that for every v 2 V , we have ST v = v. Since
we already know that T S = IW , we have that T Sw = w for all w 2 W . In
particular we have T ST v = T v, since T v 2 W . But since T is invertible, this
implies that ST v = v as desired.
Finally we show that S is linear, that it preserves addition and scalar multi-
plication. We will just show that it preserves addition. Let w; w0 2 W ; we need
to show that S(w + w0 ) = Sw + Sw0 .
T (S(w + w0 ) = T S(w + w0 ) = IW (w + w0 ) = IW w + IW w0
= T Sw + T Sw0 = T (Sw + Sw0 );
Since T is one-to-one, this implies that S(w + w0 ) = Sw + Sw0 as desired.
Thus a linear transformation is invertible if and only if it is one-to-one.
Invertible linear transformations are also known as isomorphisms.

Def: Two vector spaces V and W are said to be isomorphic if there is an

invertible linear transformation T : V ! W from one space to another.

Example: The map T : R3 ! P2 (R) de…ned by

T (a; b; c) := ax2 + bx + c
is linear, one-to-one, and onto, and hence an isomorphism. Thus R3 and
P2 (R) are isomorphic.

11
 Lemma: Two …nite-dimensional spaces V and W are isomorphic if and only
if dim(V ) = dim(W ).
Proof: if V and W are isomorphic, then there is an invertible linear trans-
formation T : V ! W from V to W , which by the lemma on one-to-one and
onto properties of inverse linear transformations, is one-to-one and onto. Since
T is one-to-one, nullity(T ) = 0. Sine T is onto, rank(T ) = dim(W ). By the
dimension theorem we have dim(V ) = dim(W ).

Now suppose that dim(V ) and dim(W ) are equal; let’s say that dim(V ) =
dim(W ) = n. Then V has a basis (v1 ; :::; vn ), and W has a basis fw1 ; ::; wn g.
We can …n a linear transformation T : V ! W such that T v1 = w1 ; :::; T vn =
wn . fw1 ; :::; wn g must span image(T ). But since w1 ; :::; wn span W , we have
image(T ) = W , T is onto, and therefore T is one-to-one, and hence is an
isomorphism. Thus V and W are isomorphic.
2 3 21 3
2 0 0 2 0 0
Ex: If A = 40 3 05 ; A 1
= 40 1
3 05
1
0 0 4 0 0 4

Theorem: Let V be a vector space with …nite ordered basis , and let W
be a vector space with …nite ordered basis . Then a linear transformation
T : V ! W is invertible if and only if the matrix [T ] is invertible. Furthermore
([T ] ) 1 = [T 1 ]

Proof: Suppose that V is n-dimensional and W is m-dimensional; this makes

[T ] an m n matrix.
First suppose that T : V ! W has an inverse T 1 : W ! V . Then
[T ] [T 1 ] = [T T 1 ] = [IW ] = Im
[T 1 ] [T ] = [T 1 T ] = [IV ] = In
Thus [T 1 ] is the inverse of [T ] and so [T ] is invertible.

Now suppose that [T ] is invertible, with inverse B. We will prove that there
exists a linear transformation S : W ! V with [S] = B.
Assume that we have
[ST ] = [S] [T ] = B[T ] = [IV ]
and hence ST = IV . A similar argument gives T S = IW , and so S is the
inverse of T and so T is invertbile.
It remains to show that we can …nd a transformation S : W ! V with
[S] = B. Write = (v1; :::; vn ) and = (w1 ; :::; wm ). Then we want a linear
transformation S : W ! V such that
Sw1 = B11 v1 + ::: + B1n vn
Sw2 = B21 v1 + ::: + B2n vn
:::
Swm = Bm1 v1 + ::: + Bmn vn

12
 Corollary: An m n matrix A is invertible if and only if the linear transfor-
mation LA : Rn ! Rm is invertible. Furthermore, the inverse of LA is LA 1 .
Proof: If is the standard basis for Rn and is the standard basis for Rm ,
then
[LA ] = A
By the theorem above, A is invertible if and only if LA is. Also, from the
above theorem we have
[LA1 ] = ([LA ] ) 1 = A 1 = [LA 1 ]
and hence LA1 = LA 1 as desired.

Corollary: In order for a matrix A to be invertible, it must be square.

13