0.

1 Matrix of linear transformation

Let V and W be two finite-dimensional linear spaces with bases B and

B ′ , respectively, B = {u1 , u2 , . . . , un } and B ′ = {v1 , v2 , . . . , vm }. T : V → W is a
linear transformation such that

T (u1 ) = a11 v1 + a21 v2 + · · · + am1 vm

T (u2 ) = a12 v1 + a22 v2 + · · · + am2 vm
(0.1.1)
······
T (un ) = a1n v1 + a2n v2 + · · · + amn vm .

The coordinates of [T (ui )] under the basis B ′ are:

2 3 2 3 2 3
66 a11 77 66 a12 77 66 a1n 77
6 77 6 77 6 77
=6 =6 77 , . . . , [T (un)]B = 666
a21 a22 a2n
[T (u1 )]B ′ 66 .. 77 , [T (u2)]B 66 .. .. 77
64 75 64 75 64 75
′
. . .
am1 am2 amn

Let 0 1
BB a11 a12 · · · a1n CC
B CC
A=B
a21 a22 · · · a2n
BB CC
B@ ··· ··· ··· ··· CA
am1 am2 · · · amn

which is an m×n matrix whose n columns correspond to [T (ui )]B ′ . The m×n
matrix A is called the coordinate matrix of T ∈ L(U, V) with respect to the
pair (B, B ′ ) . We can denote it as [T ]BB′ .

When T is a linear operator on U, and when there is only one basis

involved, [T]B is used in place of [T]B,B to denote the (necessarily square)
coordinate matrix of T with respect to B, the matrix A is called the matrix
of T relative to the basis B. Borrow the notation of matrix multiplication,
we could rewrite equation 0.1.1 into matrix form:

(T (u1 ), T (u2 ), · · · , T (un )) = (v1 , v2 , · · · , vm )A (0.1.2)

Example 0.1.1. Given a linear transformation in P3 (t),

T (a0 + a1 t + a2 t2 + a3 t3 ) = (a0 − a2 ) + (a1 − a3 )t + (a2 − a0 )t2 + (a3 − a1 )t3

Compute the matrix of T relative to the basis B.

Compute

T (1) = 1 − t2 , T (t) = t − t3 , T (t2 ) = −1 + t2 , T (t3 ) = −1 + t3

write the matrix of T relative to the natural basis 1, t, t2 , t3 directly

0 1
B
B
1 0 −1 CC
0
B 0 −1 C
A=B CC .
1 0
B
B
B 0 C CA
@ −1 0 1
0 −1 0 1

Theorem 0.1.1. Let T ∈ L(U, V), and let B = u1 , u2 , · · · , un be basis for U and
B ′ = (v1 , v2 , · · · , vm ) be basis for V. The action of T on u ∈ U is given by matrix
multiplication between their coordinates. That is

[T(u)]B′ = [T]BB′ [u]B

Proof. Let B = {u1 , u2 , . . . , un } and B ′ = {v1 , v2 , . . . , vm } . If u =

Pn xj uj and
P
j=1

T (u ) = m
j i=1 αij vi , then
0 1 0 1
B x1 C BB α11 α12 ··· α1n CC
B
B C
C B CC
[u]B = B C =B
x2 α21 α22 ··· α2n
B
B .. C
C and [T]BB′ BB .. .. .. CC
B
@ . C
A B@ . .
...
. CA
xn αm1 αm2 · · · αmn

so, according to (4.7.3)

„n Ž
X X
m X
T(u) = αij xj vi = αij xj vi
i,j i=1 j=1

In other words, the coordinates of T(u) with respect to B ′ are the terms

–2–
学位论文

Pn αij xj for i = 1, 2, . . . , m, and therefore

j=1

0P 1 0 10 1
B j α1j xj C BB α11 α12 ··· α1n CC BB x1 CC
B P
B j α2j xj C
C BB CC BB CC
=B C α21 α22 ··· α2n x2
[T(u)]B′ B
B .. C
C = BB .. .. .. CC BB .. CC = [T]B,B [u]B .
′

B
@P . C
A B@ . .
...
. CA B@ . CA
j αmj xj αm1 αm2 · · · αmn xn

After the basis for linear space V and W are fixed, the coordinate matrix
of linear transformation T is uniquely determined. Through this way, we
have established a one-to-one correspondence between matrix and linear
transformation. Meanwhile, we can use matrix A to study linear transfor-
mation T .
If the matrix of linear transformation T under basis α1 , α2 , · · · , αn is A,
then we have the following conclusion：

(1) The coordinates of the basis of R(T ) correspond to the maximal linearly
independent set of column vector sets of A.

(2) The coordinates of the basis of N (T ) correspond to the basic solution

system of Ax = 0.

(3)
dim(R(T )) = rank(A),

dim(N (T )) = dim(V ) − rank(A).

… linear transformation of R : T (X) = M X −

2×2
Example 0.1.2. Given the
1 2
XM, (∀X ∈ R2×2 )，M = , find the base and dimension of R(T )，
0 3
N (T ).

Solution:Choose the natural basis of R2×2 : E11 , E12 , E21 , E22 , then
…  …  …  … 
0 −2 0 −2 2 0 0 2
T (E11 ) = , T (E12 ) = , T (E21 ) = , T (E22 ) =
0 0 0 0 2 −2 0 0

–3–
So the matrix of T in E11 , E12 , E21 , E22 is
0 1 0 1
B
B
0 0 2 0 CC BB 1 1 0 −1 CC
B −2 2 CCC ∼ BBB 0 CC
A=B
−2 0 0 1 0
B
B CC
B 0 C CA BB@ 0 CA
@ 0 0 2 0 0 0
0 0 −2 0 0 0 0 0

You can get , rank(A) = 2，(0, −2, 0, 0)T ，(2, 0, 2, −2)T is a maximal…linearly 
in-
0 −2
dependent group of column vector group of A, so dim(R(T )) = 2, ，
0 0
… 
2 0
is a set of basis of R(T ). The fundamental solution of Ax = 0 is
2 −2
…  … 
−1 1 1 0
(−1, 1, 0, 0)T ，(1, 0, 0, 1)T , so dim(N (T )) = 2, ， is a set of
0 0 0 1
basis of N (T ).

Theorem 0.1.2. (1) If T, L ∈ L(U, V), and if B and B ′ are bases for U and V,
then
[kT]BB′ = k[T]BB′ for scalars k.

[T + L]BB′ = [T]BB′ + [L]BB′

(2) If T ∈ L(U, V) and L ∈ L(V, W), and if B, B ′ , and B ′′ are bases for U, V,
and W, respectively, then LT ∈ L(U, W), and

[LT]BB′′ = [L]B′ B′′ [T]BB′ .

(3) If T ∈ L(U, U) is invertible in the sense that TT−1 = T−1 T = I for some
T−1 ∈ L(U, U), then for every basis B of U

” —
T−1 B
= [T]B−1

Example 0.1.3. W2 = {a0 + a1 cos x + b1 sin x + a2 cos 2x + b2 sin 2x|a0 , a1 , a2 , b1 , b2 }

Define linear transformation T : W2 → W2 as

T (f (x)) = f ′′ (x) + f (0).

–4–
学位论文

Prove that T is a reversible linear transformation on W2 .

Solution:For

T (f (x)+g(x)) = (f +g)′′ (x)+f (0)+g(0) = f ′′ (x)+f (0)+g ′′ (x)+g(0) = T (f (x))+T (g(x))

T (kf (x)) = k(f ′′ (x) + f (0))

Therefore, T is a linear transformation on W2 .

Take the base of W2 : 1, cos x, sin x, cos 2x, sin 2x, then

T (1) = 1, T (cos x) = − cos x + 1, T (sin x) = − sin x,

T (cos 2x) = −4 cos 2x + 1, T (sin 2x) = −4 sin 2x

The matrix of T under basis 1, cos x, sin x, cos 2x, sin 2x is

0 1
BB 1 1 0 1 0 CC
BB 0 −1 0 0 0 CC
B CC
A=B BB 0 0 −1 0 0 CC
BB 0 −4 CC
@ 0 0 0
A
0 0 0 0 −4

For |A| = 16 ̸= 0, A is a reversible matrix, T is a reversible transformation.

Example 0.1.4. Give two linear transformations T, S in R2×2 :

…  … 
1 0 1 1
T (X) = XN, S(X) = M X, ∀X ∈ R2×2 , M = ,N = ,
−2 0 1 −1

Find (1) The matrices of T + S，T S under the basis E11 , E12 , E21 , E22 .

(2) Are T and S reversible? If reversible, find its inverse transformation.

Solution:(1) The matrices of T and S under the basis E11 , E12 , E21 , E22 can be

–5–
obtained as follows:
0 1 0 1
B
B
1 1 0 0 CC BB 1 0 0 0 CC
B1 CC B 0 0 C
A=B CC , B = BBB CC
−1 0 0 1 0
B
B
B CA B@ −2 0 C CA
@0 0 1 1 0 0
0 0 1 −1 0 −2 0 0

So, the matrices of T + S and T S under the basis E11 , E12 , E21 , E22 are:
0 1 0 1
B
B
2 1 0 0 CC BB 1 1 0 0 CC
B 1 CC B 1 0 C
A+B =B CC , AB = BBB CC
0 0 0 −1 0
B
B
B CA B@ −2 0 C CA
@ −2 0 1 1 −2 0
0 −2 1 −1 −2 2 0 0

(2)For det(A) = 4，det(B) = 0，So T is reversible, while S is irreversible.

The matrix of T −1 under the basis E11 , E12 , E21 , E22 is
0 1
B
B
1/2 1/2 0 0 CC
B 1/2 CC
A−1 = B
−1/2 0 0
B
B CC
B
@ 0 0 1/2 1/2 CA
0 0 1/2 −1/2
… 
x1 x2
Take X = ∈ R2×2 , then
x3 x4

0 1
BB x1 CC
B x2 CC
T −1 (X) = (E11 , E12 , E21 , E22 )A−1 BBB CC
B@ x3 CA
x4
x1 + x2 x1 − x2 x3 + x4 x3 − x4
= E11 + E12 + E21 +
…2 2  …2  …2 E22 
1 x1 + x2 x1 − x2 x1 x2 1 1 1
= = .
2 x3 + x4 x3 − x4 x3 x4 2 1 −1

Further, it is known that a linear space can take different basis. What is
the relationship between the matrix of linear transformation under different
bases? We will discuss in the following.

–6–
学位论文

Theorem 0.1.3. Let P be the transition matrix from basis α1 , α2 , · · · , αn to ba-

sis β1 , β2 , · · · , βn in linear space V , the coordinate matrices of linear transfor-
mation T in V under the two basis are A and B respectively, then B = P −1 AP .

Proof. From the known conditions, we have

(β1 , β2 , · · · , βn ) = (α1 , α2 , · · · , αn )P,

and
T (α1 , α2 , · · · , αn ) = (α1 , α2 , · · · , αn )A,

T (β1 , β2 , · · · , βn ) = (β1 , β2 , · · · , βn )B,

For β1 , β2 , · · · , βn is linearly independent, P is reversible. Therefore,

(α1 , α2 , · · · , αn ) = (β1 , β2 , · · · , βn )P −1 ,

So

(β1 , β2 , · · · , βn )B = T (β1 , β2 , · · · , βn ) = T (α1 , α2 , · · · , αn )P

= (α1 , α2 , · · · , αn )AP
= (β1 , β2 , · · · , βn )P −1 AP

We obtain
B = P −1 AP.

Theorem 0.1.3 shows that the matrices of linear transformation T over

V are similar under different bases.

Example 0.1.5. In three-dimensional space R3 , ε1 = (1, 0, 0) ，ε2 = (0, 1, 0)，

ε3 = (0, 0, 1) is the natural basis of R3 .

1) Define linear transformation T (x1 , x2 , x3 ) = (2x1 − x2 , x2 + x3 , x1 ), find

the matrix of T under the basis ε1 ，ε2 ，ε3 ;

2) It is known that the matrix of linear transformation T under the basis

–7–
 ˆ ’
1 0 1
η1 = (−1, 1, 1)，η2 = (1, 0, −1)，η3 = (0, 1, 1) is 1 1 0 ，find the matrix
−1 2 1
of T under the basis ε1 ，ε2 ，ε3 ;

3) Known a basis α1 = (−1, 0, 2)，α2 = (0, 1, 1)，α3 = (3, −1, 0), T (α1 ) =
(−5, 0, 3), and T (α2 ) = (0, −1, 6)，T (α3 ) = (−5, −1, 9), find matrices of T under
ε1 ，ε2 ，ε3 and α1 ，α2 ，α3 .

Solution:1) T (x1 , x2 , x3 ) = (2x1 − x2 , x2 + x3 , x1 ), so

T (ε1 ) = T (1, 0, 0) = (2, 0, 1), T (ε2 ) = T (0, 1, 0) = (−1, 1, 0), T (ε3 ) = T (0, 0, 1) = (0, 1, 0).

Because the coordinates of the three-dimensional vector under the basis ε1 ，

ε2 ，ε3 are its components,
ˆ ’
2 −1 0
T (ε1 , ε2 , ε3 ) = (ε1 , ε2 , ε3 ) 0 1 1 ,
1 0 0
ˆ ’
2 −1 0
So the matrix of T under the basis ε1 , ε2 , ε3 is 0 1 1 。
1 0 0
ˆ ’
1 0 1
2) ∵ T (η1 , η2 , η3 ) = (η1 , η2 , η3 ) 1 1 0 ，
−1 2 1
ˆ ’
−1 1 0
And (η1 , η2 , η3 ) = (ε1 , ε2 , ε3 ) 1 0 1 ，
1 −1 1
ˆ ’−1 ˆ ’
−1 1 0 −1 1 −1
∴ (ε1 , ε2 , ε3 ) = (η1 , η2 , η3 ) 1 0 1 = (η1 , η2 , η3 ) 0 1 −1 。
1 −1 1 1 0 1

–8–
学位论文

So we have
ˆ ’ˆ ’ˆ ’
−1 1 0 1 0 1 −1 1 −1
T (ε1 , ε2 , ε3 ) = (ε1 , ε2 , ε3 ) 1 0 1 1 1 0 0 1 −1
−1 1 −1 2 1
ˆ 1
’ 1 0 1
−1 1 −2
= (ε1 , ε2 , ε3 ) 2 2 0 .
3 0 2
ˆ ’
−1 1 −2
The matrix of T under the basis ε1 , ε2 , ε3 is 2 2 0 。
3 0 2

ˆ ’
−5 0 −5
3) For T (α1 , α2 , α3 ) = (ε1 , ε2 , ε3 ) 0 −1 −1 ，
3 6 9

ˆ ’
−1 0 3
(α1 , α2 , α3 ) = (ε1 , ε2 , ε3 ) 0 1 −1 ，
2 1 0

So
ˆ ’−1
−1 0 3
T (ε1 , ε2 , ε3 ) = T (α1 , α2 , α3 ) 0 1 −1
2 1 0
ˆ ’ˆ ’−1
−5 0 −5 −1 0 3
= (ε1 , ε2 , ε3 ) 0 −1 −1 0 1 −1

ˆ 3 6 9
’ˆ 2 1 0
’
−5 0 −5 −1/7 −3/7 3/7
= (ε1 , ε2 , ε3 ) 0 −1 −1 2/7 6/7 1/7
−1/7 1/7
ˆ3 6 9
’ 2/7
−5 20 −20
1
= (ε1 , ε2 , ε3 ) −4 −5 −2 ,
7
27 18 24

–9–
The matrix of T under the basis ε1 , ε2 , ε3 is
ˆ ’
−5 20 −20
1
−4 −5 −2 .
7
27 18 24

Also
ˆ ’−1 ˆ ’
−1 0 3 −5 0 −5
T (α1 , α2 , α3 ) = (α1 , α2 , α3 ) 0 1 −1 0 −1 −1

ˆ2 1 0
’ˆ 3 6 9
’
−1 −3 3 −5 0 −5
1
= (α1 , α2 , α3 ) 2 6 1 0 −1 −1
7
−1 1
ˆ 2
’ 3 6 9
2 3 5
= (α1 , α2 , α3 ) −1 0 −1 ,
−1 1 0

The matrix of T under the basis α1 , α2 , α3 is

ˆ ’
2 3 5
−1 0 −1
−1 1 0

Example 0.1.6. Let the matrix of linear transformation T on 3-dimensional

linear space V under bases α1 , α2 , α3 be
ˆ ’
1 2 2
A= 2 1 2 ,
2 2 1

Prove: W = span(−α1 + α2 , −α1 + α3 ) is an invariant subspace of T .

Proof. Get from T (α1 , α2 , α3 ) = (α1 , α2 , α3 )A:

T (α1 ) = α1 + 2α2 + 2α3 , T (α2 ) = 2α1 + α2 + 2α3 , T (α3 ) = 2α1 + 2α2 + α3

– 10 –
学位论文

Take any β ∈ W , have

β = k1 (−α1 + α2 ) + k2 (−α1 + α3 )

and

T (β) = k1 (−T (α1 ) + T (α2 )) + k2 (−T (α1 ) + T (α3 )) = k1 (α1 − α2 ) + k2 (α1 − α3 )

= −k1 (−α1 + α2 ) − k2 (−α1 + α3 ) ∈ W

So W is an invariant subspace of T . 

Theorem 0.1.4. Let T be a linear transformation in n−dimensional linear

space V over the number field R. If and only if V can be decomposed into
the direct summation of several nontrivial invariant subspaces if and only if
when there exists a basis of V , under which the representation matrix of T
under the bases is a block diagonal matrix.

Example 0.1.7. Known matrices

0 1 0 1 0 1
B
B
1 −4 2 5 C
C BB 6 5 −8 6 CC BB 1 0 0 0 CC
B −1 −2 C BB −2 −2 CCC , J = BBB 0 0 C
P =B C CC .
2 0 −1 4 2 0
B
B C BB
1 C 2 C BB 0 1 C
, A =
B
@ 0 0 1 C
A B@ 1 1 0 CA @ 0 2 CA
0 1 0 0 0 0 0 2 0 0 0 2

Satisfy that P −1 AP = J. ξ1 , ξ2 , ξ3 , ξ4 is a set of basis of linear space V , the

matrix of linear transformation T on V under the basis ξ1 , ξ2 , ξ3 , ξ4 is A. Find
all invariant subspaces of linear transformation T .

Solution:Let

e1 = ξ1 − ξ2 , e2 = −4ξ1 + 2ξ2 + ξ4 , e3 = 2ξ1 + ξ3 , e4 = 5ξ1 − 2ξ2 + ξ3

Because P, A, J satisfy P −1 AP = J, it can be known from theorem 0.1.3

that the matrix of linear transformation T under the basis e1 , e2 , e3 , e4 is J,
that is

T (e1 ) = e1 , T (e2 ) = 2e2 , T (e3 ) = 2e3 , T (e4 ) = e3 + 2e4 (0.1.3)

– 11 –
Let

V1 = Span{e1 }, V2 = Span{e2 }, V3 = Span{e3 }, V4 = Span{e3 , e3 + 2e4 }

The formula 0.1.3 shows that V1 , V2 , V3 , V4 are invariant subspaces of lin-

ear transformation T . Since any two invariant subspaces are also invariant
subspaces, the invariant subspaces of linear transformation are:

V0 = {0}, V1 , V2 , V3 , V4 , V1 ⊕V2 , V1 ⊕V3 , V1 ⊕V4 , V2 ⊕V3 , V2 ⊕V4 , V1 ⊕V2 ⊕V3 , V1 ⊕V2 ⊕V4 = V

– 12 –