6.

12 Linear Algebra
(b) Show that the mapping T : Mnn  M nn given by T (A) = A – AT is a linear operator
on Mn n.

5. Let P be a fixed non-singular matrix in M nn. Show that the mapping T : M nn  M n n given
by T (A) = P –1 AP is a linear operator.

6. Let V and W be vector spaces. Show that a function T : V  W is a linear transformation if

and only if T ( v1 +  v 2) =  T(v1) +  T(v2), for all v1, v2  V and all   .

7. Let T1, T2 : V  W be linear transformations. Define

T1 + T2 : V  W by (T1 + T2)(v) = T1(v) + T2(v), v  V
Also, define
c T1 : V  W by (cT1)(v) = c (T1(v)), v  V
Show that T1 + T2 and cT are linear transformations.

ANSWERS
1. (a) Yes (b) No (c) No (d) No (e) Yes (f ) No (g) No

6.2 THE MATRIX OF A LINEAR TRANSFORMATION

In this section we will show that a linear transformation between finite-dimensional vector spaces
is uniquely determined if we know its action on an ordered basis for the domain. We will also show
that every linear transformation between finite-dimensional vector spaces has a unique matrix ABC
with respect to the ordered bases B and C chosen for the domain and codomain, respectively.

A Linear Transformation is Determined by its Action on a Basis

One of the most useful properties of linear transformations is that, if we know how a linear map
T : V  W acts on a basis of V, then we know how it acts on the whole of V.

THEOREM 6.4 Let B = {v1, v2, ..., vn} be an ordered basis for a vector space V. Let W be a
vector space, and let w1, w2, ..., wn be any n (not necessarily distinct) vectors in W. Then there
is one and only one linear transformation T : V  W satisfying T(v1) = w1, T(v2) = w2, ..., T(vn) = wn.
In other words, a linear transformation is determined by its action on a basis.

Proof Let v be any vector in V. Since B = {v1, v2, ..., vn} is an ordered basis for V, there exist
unique scalars a1, a2, ..., an in  such that v = a1 v1 + a2 v2 + ... + an vn.
Define a function T : V  W by
T (v) = a1 w1 + a2 w2 + ... + an wn
Since the scalars ai ’s are unique, T is well-defined. We will show that T is a linear transformation.
Let x and y be two vectors in V. Then
x = b1 v1 + b2 v2 + ... + bn vn
Linear Transformations 6.13
and y = c1 v1 + c2 v2 + ... + cn vn
for some unique bi ’s and ci ’s in . Then, by definition of T, we have
T (x) = b1 w1 + b2 w2 + ... + bn wn
T (y) = c1 w1 + c2 w2 + ... + cn wn
 T(x) + T(y) = (b1 w1 + b2 w2 + ... + bn wn) + (c1 w1 + c2 w2 + ... + cn wn)
= (b1 + c1)w1 + (b2 + c2)w2 + ... + (bn + cn)wn
However, x + y = (b1 v1 + b2 v2 + ... + bn vn) + (c1 v1 + c2 v2 + ... + cn vn)
= (b1 + c1)v1 + (b2 + c2)v2 + ... + (bn + cn)vn
 T (x + y) = (b1 + c1)w1 + (b2 + c2)w2 + ... + (bn + cn)wn,
again by definition of T. Hence, T(x + y) = T(x) + T(y). Next, for any scalar c ,
c x = c (b1 v1 + b2 v2 + ... + bn vn) = (c b1)v1 + (c b2) v2 + ... + (c bn)vn
 T(c x) = (c b1)w1 + (c b2)w2 + ... + (c bn)wn
= c(b1 w1) + c (b2 w2) + ... +c(bn wn)
= c(b1 w1 + b2 w2 + ... + bn wn)
= cT (x)
Hence T is a linear transformation.

To prove the uniqueness, let L : V  W be another linear transformation satisfying

L (v1) = w1, L (v2) = w2, ..., L (vn) = wn

If v  V, then v = a1 v1 + a2 v2 + ... + an vn, for unique scalars a1, a2, ..., an  . But then

L (v) = L(a1 v1 + a2 v2 + ... + an vn)

= a1 L (v1) + a2 L (v2) + ... + an L (vn) ( L is a L.T.)
= a1 w1 + a2 w2 + ... + an wn = T (v)
 L = T and hence T is uniquely determined.

EXAMPLE 14 Suppose L :  3   2 is a linear transformation with

L([1, –1, 0]) = [2, 1], L([0, 1, –1]) = [–1, 3] and L([0, 1, 0]) = [0, 1].
Find L([–1, 1, 2]). Also, give a formula for L([x, y, z]), for any [x, y, z]  3.
[Delhi Univ. GE-2, 2017]

SOLUTION To find L([–1, 1, 2]), we need to express the vector v = [–1, 1, 2] as a linear
combination of vectors v1 = [1, –1, 0], v2 = [0, 1, –1] and v3 = [0, 1, 0]. That is, we need to find
constants a1, a2 and a3 such that
v = a 1 v 1 + a 2 v 2 + a 3 v 3,
which leads to the linear system whose augmented matrix is
6.14 Linear Algebra
 1 0 0 1 
 
 1 1 1 1 
 0 1 0 2 
 
We transform this matrix to reduced row echelon form :
 1 0 0 1   1 0 0 1 
  R2  R2  R1  
 1 1 1 1   0 1 1 0 
 0 1 0 2   0 1 0 2 
   
 1 0 0 1 
R3  R3  R2  
  0 1 1 0 
0 0 1 2 
 
1 0 0 1 

R2  R2  R3 
  0 1 0 2 
 
 0 0 1 2 
This gives a1 = –1, a2 = –2, and a3 = 2. So,
v = – v1 – 2v2 + 2 v3
 L(v) = L(– v1 – 2v2 + 2 v3)
= L(– v1) – 2L(v2) + 2L(v3)
= –[2, 1] – 2[–1, 3] + 2[0, 1] = [0, –5]
i.e., L([–1, 1, 2]) = [0, –5]
To find L([x, y, z]) for any [x, y, z]   3, we row reduce
 1 0 0 x 1 0 0 x
   
 1 1 1 y  to obtain  0 1 0  z
 0 1 0 z  0 0 1 x  y  z 
   
Thus, [x, y, z] = x v1 – z v2 + (x + y + z) v3
 L([x, y, z]) = L(x v1 – z v2 + (x + y + z) v3)
= x L(v1) – z L(v2) + (x + y + z)L(v3)
= x[2, 1] – z[–1, 3] + (x + y + z) [0, 1]
= [2x + z, 2x + y – 2z].

EXAMPLE 15 Suppose L :  2   2 is a linear operator and L([1, 1]) = [1, –3] and
L([–2, 3]) = [–4, 2]. Express L([1, 0]) and L([0, 1]) as linear combinations of the vectors
[1, 0] and [0, 1]. [Delhi Univ. GE-2, 2019]

SOLUTION To find L([1, 0]) and L([0, 1]), we first express the vectors v 1 = [1, 0] and
v2 = [0, 1] as linear combinations of vectors w1 = [1, 1] and w2 = [–2, 3]. To do this, we row
reduce the augmented matrix
Linear Transformations 6.15

1 2 1 0 
[w 1 w2 | v1 v2 ] =  
1 3 0 1
Thus, we row reduce
1 2 1 0  R 2  R 2  R1  1 2 1 0 
     
1 3 0 1 0 5 1 1 
1
R2  R2  1 2 1 0

5   
0 1 1/ 5 1/ 5 

R1  R1  2 R 2  1 0 3/ 5 2 / 5 
   
 0 1 1/ 5 1/ 5 
3 1 2 1
 v1 = w – w and v 2 = w1 + w 2
5 1 5 2 5 5
3 1
This gives L(v 1) = L (w1) – L (w2 )
5 5
3 1
= L([1, 1]) – L([–2, 3])
5 5
3 1  7 11 7 11
= [1, –3] – [–4, 2] =  ,   [1, 0]  [0, 1]
5 5  5 5  5 5
2 1
and L(v 2) = L (w1) + L (w2 )
5 5
2 1
= L([1, 1]) + L([–2, 3])
5 5
2 1 2 4 2 4
[1, –3] + [–4, 2] =  ,   [1, 0]  [0, 1].
5 
=
5 5  5 5 5

The Matrix of a Linear Transformation

We now show that any linear transformation on a finite-dimensional vector space can be expressed
as a matrix multiplication. This will enable us to find the effect of any linear transformation by
simply using matrix multiplication.
Let V and W be non-trivial vector spaces, with dim V = n and dim W = m. Let B = {v1, v2, ..., vn}
and C = {w1, w2, ..., wm} be ordered bases for V and W, respectively. Let T : V  W be a linear
transformation. For each v in V, the coordinate vectors for v and T (v) with respect to ordered
bases B and C are [v]B and [T (v)]C , respectively. Our goal is to find an m × n matrix A = (aij )
(1  i  m ; 1  j  n ) such that
A[v]B = [T(v)]C ...(1)
6.16 Linear Algebra
holds for all vectors v in V. Since Equation (1) must hold for all vectors in V, it must hold, in
particular, for the basis vectors in B, that is,
A[v1]B = [T(v1)]C , A[v2]B = [T(v2)]C , ...., A[vn]B = [T(vn)]C ...(2)

 1 0  0 
0   1 0 
But [v1]B =   , [v2]B =   , ...., [vn]B =   ,
   
     
 0   0   1
 a11 a12  a1n  1   a11 
a  a2 n  0  a 
21 a22
 A[v1]B =    21 
          
    
 am1 am 2  amn  0   am1 

 a11 a12  a1n   0  a12 

a  
 a2n  1   a 
21 a22
A[v2]B =    22 
          
    
 am1 am 2  amn   0  am 2 

 a11 a12  a1n   0  a1n 

a a  a2n   0 a 
A[vn]B =  21 22   2n 
          
    
 am1 am 2  amn  1   amn 
Substituting these results into (2), we obtain

 a11   a12   a1n 

a  a  a 
 21  = [T(v )] ,  22  = [T(v )] , ......,  2n  = [T(v )] ,
   1 C    2 C    n C
     
 am1   am2   amn 
This shows that the successive columns of A are the coordinate vectors of T(v1), T(v2), ..., T(vn)
with respect to the ordered basis C. Thus, the matrix A is given by
A = [[T(v1)]C [T(v2)]C [T(vn)]C ]
We will call this matrix as the matrix of T relative to the bases B and C and will denote it by the
symbol ABC or [T ]BC . Thus,
ABC = [[T(v1)]C [T(v2)]C ... T(vn)]C ]
From (1), the matrix ABC satisfies the property
ABC [v]B = [T(v)]C for all v  V.
We have thus proved :
Linear Transformations 6.17

THEOREM 6.5 Let V and W be non-trivial vector spaces, with dim(V) = n and dim(W) = m.
Let B = {v1, v2, ..., vn} and C = {w1, w2, ..., wm} be ordered bases for V and W, respectively. Let
T : V  W be a linear transformation. Then there is a unique m × n matrix ABC such that
ABC[v]B = [T [v]]C, for all v  V. (That is ABC times the coordinatization of v with respect to B
gives the coordinatization of T (v) with respect to C).
Furthermore, for 1  i  n, the ith column of ABC = [T [vi]]C.

EXAMPLE 16 Let T :  3   3 be the linear operator given by T ([x1, x2, x3]) = [3x1 + x2,
x1 + x3, x1 – x3]. Find the matrix for T with respect to the standard basis for  3.

SOLUTION The standard basis for  3 is B = {e1 = [1, 0, 0], e2 = [0, 1, 0], e3 = [0, 0, 1]}.
Substituting each standard basis vector into the given formula for T shows that
T(e1) = [3, 1, 1], T(e2) = [1, 0, 0], T(e3) = [0, 1, –1]
Since the coordinate vector of any element [x1, x2, x3] in  3 with respect to the standard basis
 x1 
 
{e1, e2, e3} is  x2  , we have
 x3 

 3 1   0
 
[T(e1)]B = 1 , [T(e2)]B =  0  , [T(e3)]B =  1 
1   0    1 

Thus, the matrix ABB for T with respect to the standard basis is :

3 1 0 
 
ABB = [[T(e1)]B [T(e2)]B [T(e3)] B ] =  1 0 1  .
 1 0 1 

EXAMPLE 17 Let T : P3   3 be the linear transformation given by T (a x 3 + bx2 + cx + d )

= [4a – b + 3c + 3d, a + 3b – c + 5d, –2a – 7b + 5c – d]. Find the matrix for T with respect to the
standard bases B = {x3, x2, x, 1} for P3 and C = {e1, e2, e3} for  3.

SOLUTION Substituting each standard basis vector in B into the given formula for T shows that
T(x3) = [4, 1, –2], T (x2) = [–1, 3, –7], T (x) = [3, –1, 5] and T (1) = [3, 5, –1]. Since we are using
the standard basis C for  3,
 4  1  3  3
     
[T (x3)] C =  1 , [T (x2)]C =  3 , [T (x)]C =  1 , [T (1)]C =  5
 2  7   5  1
Thus, the matrix of T with respect to the bases B and C is:
 4 1 3 3
 
ABC = [ [T (x3)] C [T (x2)] C [T (x)]C [T (1)]C ] =  1 3 1 5 .
 2 7 5 1
6.18 Linear Algebra
EXAMPLE 18 Let T : P3  P2 be the linear transformation given by T (p) = p, where p  P3.
Find the matrix for T with respect to the standard bases for P3 and P2. Use this matrix to calculate
T(4x3 – 5x2 + 6x – 7) by matrix multiplication.

SOLUTION The standard basis for P3 is B = {x3, x2, x, 1}, and the standard basis for P2 is
C = {x2, x, 1}. Computing the derivative of each polynomial in the standard bases B for P3 shows
that
T(x3) = 3x 2, T(x2) = 2x, T(x) = 1, and T(1) = 0.
We convert these resulting polynomials in P2 to vectors in  3 :
3x2  [3, 0, 0] ; 2x  [0, 2, 0] ; 1  [0, 0, 1] ; and 0  [0, 0, 0].
Using each of these vectors as columns yields

3 0 0 0
 
ABC = 0 2 0 0  .
0 0 1 0 
We will compute T (4x3 – 5x2 + 6x – 7) using this matrix. Now,

 4
 5 
3 2  
[4x – 5x + 6x – 7]B =  6 
 
 7 
Hence,

 4
 3 0 0 0     12 
  5  
[T(4x3 – 5x2 + 6x – 7)]C = ABC [4x3 – 5x2 + 6x – 7]B = 0 2 0 0      10  .
6
0 0 1 0     6 
 7 
Converting back from C-coordinates to polynomials gives
T (4x3 – 5x2 + 6x – 7) = 12x2 – 10x + 6.

EXAMPLE 19 Let T : 3  2 be the linear transformation given by T ([x1, x2, x3]) = [–2x1 + 3x3,
x1 + 2x2 – x3]. Find the matrix for T with respect to the ordered bases B = {[1, –3, 2], [–4, 13, –3],
[2, –3, 20]} for  3 and C = {[–2, –1], [5, 3]} for  2. [Delhi Univ. GE-2, 2019(Modified)]
SOLUTION By definition, the matrix ABC of T with respect to the ordered bases B and C is given
by ABC = [[T(v1)]C [T(v2)]C [T(v3)]C ], where v1 = [1, –3, 2], v2 = [–4, 13, – 3], and v3 = [2, –3, 20]
are the basis vectors in B. Substituting each basis vector in B into the given formula for T shows that
T(v1) = [4, –7], T(v2) = [–1, 25], T(v3) = [56, –24]
Next, we must find the coordinate vector of each of these images in  2 with respect to the C
basis. To do this, we use the Coordinatization Method. Thus, we must row reduce matrix
[w1 w2 | T (v1) T (v2) T (v3)],
where w1 = [–2, –1], w2 = [5, 3] are the basis vectors in C. Thus, we row reduce
Linear Transformations 6.19
2 5 4 1 56   1 0 47 128 288 
  to obtain  
 1 3 7 25 24  0 1 18 51 104 

 47  128  288

Hence [T (v1)]C =  18 , [T (v2)]C =   , [T (v3)]C =  
   51  104 
 47 128 288
 The matrix of T with respect to the bases B and C is ABC =  18 51 104  .


Finding the New Matrix for a Linear Transformation After a Change of Basis
We now state a theorem (proof omitted) which helps us in computing the matrix for a linear
transformation when we change the bases for the domain and codomain.

THEOREM 6.6 Let V and W be two non-trivial finite-dimensional vector spaces with ordered
bases B and C, respectively. Let T : V  W be a linear transformation with matrix ABC with
respect to bases B and C. Suppose that D and E are other ordered bases for V and W, respectively.
Let P be the transition matrix from B to D, and let Q be the transition matrix from C to E. Then
the matrix ADE for T with respect to bases D and E is given by ADE = QABC P–1.

EXAMPLE 20 Let T :  3   3 be the linear operator given by T [(a, b, c)] = [–2a + b,

– b – c, a + 3c].

(a) Find the matrix ABB for T with respect to the standard basis B = {e1 = [1, 0, 0], e2 = [0, 1, 0],
e3 = [0, 0, 1]} for  3.

(b) Use part (a) to find the matrix ADE with respect to the standard bases D = {[15, – 6, 4],
[2, 0, 1], [3, –1, 1]} and E = {[1, – 3, 1], [0, 3, –1], [2, –2, 1]}.

SOLUTION (a) We have T (e1) = [–2, 0, 1], T (e2) = [1, –1, 0], T (e3) = [0, –1, 3]}. Using each
of these vectors as columns yields the matrix ABB :

 2 1 0
 0 1 1
ABB =  
 1 0 3
6.20 Linear Algebra
(b) To find ADE , we make use of the following relationship :
ADE = Q ABB P –1, ...(1)
where P is the transition matrix from B to D and Q is the transition matrix from B to E. Since P –1 is
the transition matrix from D to B and B is the standard basis for  3, it is given by
 15 2 3
 
P –1 =  6 0 1
 4 1 1
To find Q, we first find Q –1, the transition matrix from E to B, which is given by

 1 0 2
 
Q –1 =  3 3 2 
 1 1 1
It can be easily checked that
 1 2 6 
 
Q= (Q –1) –1 =  1 1 4 
0 1 3

Hence, using Eq. (1), we obtain

 1 2 6   2 1 0   15 2 3  202 32 43

 1 1 4   0 1 1  6 0 1  
ADE = Q ABB P –1 =     =  146 23 31 .
0 1 3  1 0 3  4 1 1  83 14 18
EXAMPLE 21 Let T : P3   3 be the linear transformation given by T (ax3 + bx 2 + cx + d )
= [c + d, 2b, a – d].
(a) Find the matrix ABC for T with respect to the standard bases B (for P3) and C (for  3).
(b) Use part (a) to find the matrix ADE for T with respect to the standard bases D = {x3 + x2,
x2 + x, x + 1, 1} for P3 and E = {[–2, 1, –3], [1, –3, 0], [3, –6, 2]} for  3.

SOLUTION (a) To find the matrix ABC for T with respect to the standard bases B = {x3, x2, x, 1}
for P3 and C = {e1 = [1, 0, 0], e2 = [0, 1, 0], e3 = [0, 0, 1]} for  3, we first need to find T (v) for
each v  B. By definition of T, we have
T (x3) = [0, 0, 1], T (x2) = [0, 2, 0], T (x) = [1, 0, 0] and T (1) = [1, 0, –1]
Since we are using the standard basis C for  3, the matrix ABC for T is the matrix whose columns
are these images. Thus
0 0 1 1
 
ABC = 0 2 0 0 
 1 0 0 1
(b) To find ADE, we make use of the following relationship :
ADE = Q ABC P –1 ...(1)
Linear Transformations 6.21
where P is the transition matrix from B to D and Q is the transition matrix C to E. Since P is the
transition matrix from B to D, therefore P –1 is the transition matrix from D to B. To compute
P –1, we need to convert the polynomials in D into vectors in  4. This is done by converting each
polynomial ax 3 + bx 2 + cx + d in D to [a, b, c, d]. Thus
(x3 + x2)  [1, 1, 0, 0] ; (x2 + x)  [0, 1, 1, 0] ; (x + 1)  [0, 0, 1, 1] ; (1)  [0, 0, 0, 1]
Since B is the standard basis for  3, the transition matrix (P –1) from D to B is obtained by using
each of these vectors as columns :

1 0 0 0
1 1 0 0

P –1 = 0 1 1 0
 
0 0 1 1
To find Q, we first find Q –1, the transition matrix from E to C, which is the matrix whose columns
are the vectors in E.

 2 1 3
 1 3 6
Q –1 =  
 3 0 2 

1
 2 1 3  6 2 3
 Q= (Q –1) –1 = 1 3 6
   16 5 9

 
 3 0 2   9 3 5

1 0 0 0
 6 2 3  0 0 1 1   1 10 15 9
   1 1 0 0   1 26
Hence, ADE = QABC P –1 =  16 5 9  0 2 0 0   =  41 25 .
0 1 1 0
 9 3 5  1 0 0 1    1 15 23 14
0 0 1 1

6.3 LINEAR OPERATORS AND SIMILARITY

In this section we will show that any two matrices for the same linear operator (on a finite-dimensional
vector space) with respect to different ordered bases are similar.
Let V be a finite-dimensional vector space with ordered bases B and C, and let T : V  V be a
linear operator. Then we can find two matrices, ABB and ACC , for T with respect to ordered bases
B and C, respectively. We will show that ABB and ACC are similar. To prove this, let P denote the
transition matrix (PCB) from B to C. Then by Theorem 6.6, we have
ACC = P ABB P –1  ABB = P –1 ACC P
This shows that the matrices ABB and ACC are similar. We have thus proved the following:

THEOREM 6.7 Let V be a finite-dimensional vector space with ordered bases B and C. Let T
be a linear operator on V. Then the matrix ABB for T with respect to the basis B is similar to the
matrix ACC for T with respect to the basis C. More specifically, if P is the transition matrix from
B to C, then ABB = P –1 ACC P.