Chapter 4: Linear Transformations

1 Let A be an m × n matrix with real entries.

2 Then A “acts” on the n-dimensional space Rn by left multiplication : If
v ∈ Rn then Av ∈ Rm .
3 In other words, A defines a function

TA : Rn −→ Rm , TA (v ) = Av .

4 By properties of matrix multiplication, TA satisfies the following conditions:

i. TA (v + w ) = TA (v ) + TA (w )
ii. TA (cv ) = cTA (v )
where c ∈ R and v , w ∈ Rn .
5 We say that TA respects the two operations in the vector space Rn .
6 In this lecture we study such maps between the vector spaces.
1 / 24
Linear Transformations
Definition
Let V , W be vector spaces over F. A linear transformation T : V −→ W is a
function satisfying

T (v + w ) = T (v ) + T (w ) and T (cv ) = cT (v )

where v , w ∈ V and c ∈ F.

1 If T : V → W is a linear transformation, then T (0) = 0 (why?)

2 Examples:

i. For any pair of vector spaces V , W over F, the “zero map” T0 : V → W

defined as T0 (v ) = 0 for all v ∈ V , is clearly a linear transformation.
3 Can you now think of another linear map from a vector space V to itself?

ii. The identity map I : V → V defined as I (v ) = v for all v ∈ V , is clearly a

linear map.
2 / 24
Linear Transformations: Examples
iii. Let c ∈ R, V = W = R2 . Define T : R2 −→ R2 by
         
x c 0 x cx x
T = = =c .
y 0 c y cy y
It follows that T is a linear transformation (why?) since
T (v + w ) = c(v + w ) = cv + cw = T (v ) + T (w )
T (dv ) = c(dv ) = d(cv ) = dT (v ), for v , w ∈ R2 , d ∈ R.

iv. Rotation: Fix θ and define T : R2 −→ R2 by

       
x cos θ − sin θ x x cos θ − y sin θ
T = = .
y sin θ cos θ y x sin θ + y cos θ

1 Then T (e1 ) = (cos θ, sin θ)t and T (e2 ) = (− sin θ, cos θ)t .
2 Thus T rotates the whole space by θ. (Draw a picture to convince yourself of
this. Another way is to identify the vector (x, y )t with the complex number
z = x + iy . Then we can write T (z) = ze iθ ).
3 / 24
Linear Transformations: Examples

v. Let D be the vector space of differentiable functions f : R −→ R such that

f (n) exists for all n. Define D : D −→ D by
0
D(f ) = f .

1 Then D is a linear transformation (why?) since

0 0
D(af + bg ) = af + bg = aD(f ) + bD(g ), for f , g ∈ D and a, b ∈ R.

vi. Define I : D −→ D by Z x
I(f )(x) = f (t) dt.
0
2 By properties of integration, I is a linear transformation.

vii. The map T : R → R given by T (x) = x 2 is not linear (why?).

viii. Let V = Mn×n (F) be the vector space of all n × n matrices over F. Fix
A ∈ V . The map T : V → V given by T (N) = AN is linear (why?).

4 / 24
Linear Transformations: Rank and Nullity
1 Let T : V → W be a linear transformation of vector spaces. There are two
important subspaces associated with T .
Nullspace of T = N (T ) = {v ∈ V | T (v ) = 0}.
Image of T = Im (T ) = {T (v ) | v ∈ V }.

2 Let V be a finite dimensional vector space. Suppose that α, β are scalars. If

v , w ∈ N (T ) then T (αv + βw ) = αT (v ) + βT (w ) = 0. Hence
αv + βw ∈ N (T ).
3 Thus N (T ) is a subspace of V . The dimension of N (T ) is called the nullity
of T and it is denoted by nullity (T ).
4 Suppose that v , w ∈ V . Then
αT (v ) + βT (w ) = T (αv + βw ) ∈ Im (T ).

5 Thus Im (T ) is a subspace of W . The dimension of Im (T ), denoted by

rank(T ), is called the rank of T .
5 / 24
Linear Transformations: Rank and Nullity
Proposition
Let T : V → W be a linear map of vector spaces. Then T is 1-1 if and only if
N (T ) = {0}.

1 Proof: (⇐=) T (u) = T (v ) =⇒ T (u − v ) = 0 =⇒ u = v .

(=⇒) v ∈ N (T ) =⇒ T (v ) = 0 = T (0) =⇒ v = 0.

Proposition
Let V , W be vector spaces. Assume V is finite dimensional with {v1 , . . . , vn } as a
basis. Let (w1 , . . . , wn ) (these wj ’s need not be distinct) be an arbitrary sequence
of vectors in W . Then there is a unique linear map T : V → W with T (vi ) = wi ,
for all i = 1, . . . , n.
2 Proof: (uniqueness) Given v ∈ V we can write (uniquely)
v = a1 v1 + · · · + an vn , for scalars ai . Then
T (v ) = a1 T (v1 ) + · · · + an T (vn ) = a1 w1 + · · · + an wn . So T is determined
by (w1 , . . . , wn ).
6 / 24
Linear Transformations: Rank and Nullity
1 (existence) Define T as follows.
2 Given v ∈ V write (uniquely) v = a1 v1 + · · · + an vn , for scalars ai and then
define T (v ) = a1 w1 + · · · + an wn .
3 Show that T is linear (exercise).

Theorem (The Rank-Nullity Theorem)

Let T : V → W be a linear transformation of vector spaces where V is finite
dimensional. Then
rank(T ) + nullity (T ) = dim V .

4 Proof: Suppose dim V = n. Let B = {v1 , v2 , . . . , vk } be a basis of N (T ). We

can extend B to a basis C = {v1 , v2 , . . . , vk , w1 , w2 , . . . , wn−k } of V .
5 We show that
D = {T (w1 ), T (w2 ), . . . , T (wn−k )}
is a basis of Im (T ).
7 / 24
Rank-Nullity Theorem: Proof continues...
1 Note that any v ∈ V can be expressed uniquely as
v = α1 v1 + α2 v2 + · · · + αk vk + β1 w1 + · · · + βn−k wn−k .

2 This implies that

T (v ) = α1 T (v1 ) + · · · + αk T (vk ) + β1 T (w1 ) + · · · + βn−k T (wn−k )
= β1 T (w1 ) + · · · + βn−k T (wn−k ).

3 Hence D spans Im T .
4 We now show that D is linearly independent. For, suppose
β1 T (w1 ) + · · · + βn−k T (wn−k ) = T (β1 w1 + · · · + βn−k wn−k ) = 0.

5 Then β1 w1 + · · · + βn−k wn−k ∈ N (T ) and hence there are scalars

α1 , α2 , . . . , αk such that
β1 w1 + β2 w2 + · · · + βn−k wn−k = α1 v1 + α2 v2 + · · · + αk vk
=⇒ β1 = β2 = · · · = βn−k = 0.
8 / 24
Rank-Nullity Theorem: Proof continues...
1 Hence D is a basis of Im T . Thus
rank(T ) = n − k = dim V − dim N (T ).

2 Coordinate Vectors:
3 Let V be a finite dimensional vector space (fdvs) of dimension n over F. By
an ordered basis of V we mean a sequence (v1 , v2 , . . . , vn ) of distinct vectors
of V such that the set {v1 , . . . , vn } is linearly independent.
4 Let u ∈ V . Write uniquely (why?)
u = a1 v1 + a2 v2 + · · · + an vn , ai ∈ F.

5 Define the coordinate vector of u with respect to (wrt) the ordered basis B by
 t
[u]B = a1 a2 · · · an .

6 Note that (why?) for vectors u, v ∈ V and scalar a ∈ F, we have

[u + v ]B = [u]B + [v ]B , [av ]B = a[v ]B .
9 / 24
Coordinate Vectors: Change of Basis

1 Suppose C = (u1 , . . . , un ) is another ordered basis of V .

2 Given u ∈ V , what is the relation between [u]B and [u]C ?
3 Define MBC , the transition matrix from C to B, to be the n × n matrix whose
jth column is [uj ]B :

MBC = [[u1 ]B [u2 ]B · · · [un ]B ] .

Proposition
Set M = MBC . Then, for all u ∈ V , we have

[u]B = M[u]C .

4 Proof: Let  t
[u]C = a1 a2 ··· an .

10 / 24
Coordinate Vectors: Change of Basis

1 Then u = a1 u1 + a2 u2 + · · · + an un and we have

[u]B = [a1 u1 + · · · + an un ]B

= a1 [u1 ]B + · · · + an [un ]B
 
a1

 a2 

= [[u1 ]B [u2 ]B · · · [un ]B ] 
 . 

 . 
an

= M[u]C .

11 / 24
Coordinate Vectors: Change of Basis
1 Example: Let V = R3 and let
           
1 0 0 1 0 0
v1 =  1  , v2 =  1  , v3 =  0  , u1 =  0  , u2 =  1  , u3 =  0  .
1 1 1 0 0 1

2 Consider the ordered bases B = (v1 , v2 , v3 ) and C = (u1 , u2 , u3 ). We have

(why?)  
1 0 0
M = MBC =  −1 1 0 .
0 −1 1
   
2 2
3 Let u =  3 . So (why?) [u]C =  3 .
4 4
4 Then     
1 0 0 2 2
[u]B =  −1 1 0  3  =  1 .
0 −1 1 4 1
12 / 24
Coordinate Vectors: Change of Basis
1 Check that        
2 1 0 0
 3  = 2 1  +  1  +  0 .
4 1 1 1

Proposition
Let V be a finite dimensional vector space and B and C be two ordered bases.
Then
MBC = (MCB )−1 .
2 Proof: Put M = MCB and N = MBC . We need to show that MN = NM = I .
3 We have, for all u ∈ V , [u]B = N[u]C , [u]C = M[u]B .
4 It follows that, for all u ∈ V ,
[u]B = N[u]C = NM[u]B
[u]C = M[u]B = MN[u]C .

5 Thus (why?) MN = NM = I .
13 / 24
Coordinate Vectors: Change of Basis
1 Example: Let M be the (n + 1) × (n + 1) matrix, with rows and columns
indexed by
{0, 1, . . . , n}, and with entry in row i and column j, 0 ≤ i, j ≤ n,
given by ji . We show that M is invertible and find the inverse explicitly.
2 Consider the vector space Pn (R) of real polynomials of degree ≤ n. Then
B = (1, x, x 2 , . . . , x n ) and C = (1, x − 1, (x − 1)2 , . . . , (x − 1)n ) are both
ordered bases (why?).
3 We claim that M = MCB . To see this note the following computation. For
0 ≤ j ≤ n we have
xj = (1 + (x − 1))j
j  
X j
= (x − 1)i
i
i=0
n  
X j
= (x − 1)i ,
i
i=0

j


where in the last step we have used the fact that i = 0 for i > j.
14 / 24
Coordinate Vectors: Change of Basis
h
i
j
1 Thus M = i = MCB and hence it is invertible.

2 Since M −1 = (MCB )−1 = MBC , its entries are given by the following
computation.
3 For 0 ≤ j ≤ n, we have
j 
j−i j
X
j
(x − 1) = (−1) xi
i
i=0
n 
j−i j
X
= (−1) xi.
i
i=0

−1 j
j−i


4 Thus theh entry in row
i i and column j of M is (−1) i , that is,
M −1 = (−1)j−i ji .

15 / 24
Matrices and Linear Transformations
1 Let V and W be finite dimensional vector spaces with dim V = n and
dim W = m. Suppose E = (v1 , v2 , . . . , vn ) is an ordered basis for V and
F = (w1 , w2 , . . . , wm ) is an ordered basis for W .
2 Let T : V −→ W be a linear transformation.
3 We define MFE (T ), the matrix of T with respect to the ordered bases E and
F , to be the m × n matrix whose jth column is [T (vj )]F :

MFE (T ) = [[T (v1 )]F [T (v2 )]F · · · [T (vn )]F ] .

4 Please do the following important exercise.

5 Exercise: Let A be an m × n matrix over F and consider the linear map
TA : Fn → Fm given by TA (v ) = Av , for v ∈ Fn (we are considering column
vectors here).
Considering the ordered basis E = (e1 , . . . , en ) and F = (e1 , . . . , em ) of Fn
and Fm respectively, show that MFE (TA ) = A.
16 / 24
Matrices and Linear Transformations
1 Let L(V , W ) denote the set of all linear transformations from V to W .
Suppose S, T ∈ L(V , W ) and c is a scalar.
2 Define S + T and cS as follows :
(S + T )(x) = S(x) + T (x)
(cS)(x) = cS(x)
for all x ∈ V .
3 It is easy to show that L(V , W ) is a vector space under these operations.

Proposition
Fix ordered bases E and F of V and W respectively. For all S, T ∈ L(V , W ) and
scalar c we have
i. MFE (S + T ) = MFE (S) + MFE (T )
ii. MFE (cS) = cMFE (S)
iii. MFE (S) = MFE (T ) ⇔ S = T .

4 Proof: Exercise.
17 / 24
Matrices and Linear Transformations
Proposition
Suppose V , W are vector spaces of dimensions n, m respectively. Suppose
T : V −→ W is a linear transformation. Suppose
E = (v1 , . . . , vn ), F = (w1 , . . . , wm ) are ordered bases of V , W respectively. Then

[T (v )]F = MFE (T )[v ]E , v ∈ V .

1 Proof: Let  t
[v ]E = a1 a2 · · · an .
2 Then v = a1 v1 + a2 v2 + · · · + an vn and hence
T (v ) = a1 T (v1 ) + a2 T (v2 ) + · · · + an T (vn ).
3 We have
[T (v )]F = [a1 T (v1 ) + · · · + an T (vn )]F
= a1 [T (v1 )]F + · · · + an [T (vn )]F
 t
= [[T (v1 )]F [T (v2 )]F · · · [T (vn )]F ] a1 a2 ··· an
= MFE (T )[v ]E .

18 / 24
Matrices and Linear Transformations
Proposition
Suppose U, V , W are vector spaces of dimension n, p, m respectively. Suppose
T : U −→ V and S : V −→ W are linear transformations. Suppose E , F , G are
ordered bases of U, V , W respectively. Then

MGE (SoT ) = MGF (S)MFE (T ).

1 Proof: Let E = (u1 , u2 , . . . , un ). Then, the jth column of MGE (S ◦ T ) is

= [(S ◦ T )(uj )]G = [S(T (uj ))]G .

2 Now the jth column of MGF (S)MFE (T ) is

= MGF (S)(jth column of MFE (T ))
= MGF (S)[T (uj )]F
= [S(T (uj ))]G (since [S(v )]G = MGF (S)[v ]F ).

3 Hence MGE (SoT ) = MGF (S)MFE (T ).

19 / 24
Matrices and Linear Transformations

1 Let V be a finite dimensional vector space. A linear map T : V → V is said

to be a linear operator on V . Let B, C be ordered bases of V .
2 The square matrix MBB (T ) is said to be the matrix of T with respect to the
ordered basis B.
3 Note that the transition matrix MBC from C to B is the matrix MBC (I ) of the
identity map with respect to the bases C and B.
4 Thus it follows that MBC (I ) = MCB (I )−1 .

20 / 24
Matrices and Linear Transformations
Proposition
Let V be a finite dimensional vector space and B, C be a pair of two bases of V .
Then, we have
MBB (T ) = (MCB )−1 MCC (T )MCB .

1 Proof: Consider the sequence of linear operators where the bases used for
computation of matrices of the linear transformations are specified:
I T I
(V , B) →
− (V , C ) −
→ (V , C ) →
− (V , B).

2 Note that the identity map I is just a map from V to V . It is not required
that I maps B to C or C to B. This notation is used just to show that the
mentioned bases are used for the computation of the matrices of the
corresponding linear maps.
3 Then
T =I ◦T ◦I =⇒ MBB (T ) = MBC (I )MCC (T )MCB (I )
=⇒ MBB (T ) = (MCB )−1 MCC (T )MCB .

21 / 24
Matrices and Linear Transformations

1 Example: Consider the linear transformation

T : R2 → R2 , T (e1 ) = e1 , T (e2 ) = e1 + e2 .

2 Let C = (e1 , e2 ) and B = (e1 + e2 , e1 − e2 ) be two ordered bases of R2 .

3 Then
     
1 1 1 1 1/2 1/2
MCC (T ) = , MCB = , MBC = .
0 1 1 −1 1/2 −1/2

4 Since MBB (T ) = (MCB )−1 MCC (T )MCB = MBC MCC (T )MCB , we get
       
B 1/2 1/2 1 1 1 1 1 3 −1
MB (T ) = = .
1/2 −1/2 0 1 1 −1 2 1 1

22 / 24
Sum of two subspaces and its dimension

1 Given subspaces V , W of a vector space U define the sum of V and W ,

denoted V + W , by
V + W = L(V ∪ W ).

Theorem
Let V , W be subspaces of a finite dimensional vector space U. Then
dim(V + W ) = dim V + dim W − dim(V ∩ W ).

23 / 24
Proof of the formula for dim(V + W )
1 Proof: We shall give a sketch of a proof leaving you to fill in the details.
2 Consider the set V × W = {(v , w ) : v ∈ V , w ∈ W }. This set is a vector
space with component wise addition and scalar multiplication.
3 Check that the dimension of this space is dim V + dim W .
4 Define a linear map T : V × W → V + W by T ((v , w )) = v − w .
5 Check that T is onto and that the nullspace of T is {(v , v ) : v ∈ V ∩ W }.
6 The result now follows from the rank nullity theorem for linear maps.
7 Exercise:

i. Let V , W be finite dimensional vector spaces over F with dimensions n, m

respectively. Fix ordered bases E , F for V , W respectively.

Consider the map f : L(V , W ) → Mm×n (F) given by f (T ) = MFE (T ), for

T ∈ L(V , W ). Show that f is linear, 1-1 and onto, which shows that
dim L(V , W ) = mn.
24 / 24