这一节通过Theorem 11和Theorem 12 将linear transformation和矩阵建立了一对一的关联,此外Theorem 13说明linear transformation的composition与矩阵乘法也有天然的联系。Theorem 14则说明在linear operator的情况下,同一个linear transformation不同basis的矩阵存在相似(similar)的关系,且这个导致相似的矩阵实质上就是一组基在另一组基下的矩阵。
Exercises
1.Let
T
T
T be the linear operator on
C
2
C^2
C2 defined by
T
(
x
1
,
x
2
)
=
(
x
1
,
0
)
T(x_1,x_2)=(x_1,0)
T(x1,x2)=(x1,0). Let
B
\mathfrak B
B be the stantard ordered basis for
C
2
C^2
C2 and let
B
′
=
{
α
1
,
α
2
}
\mathfrak B'=\{\alpha_1,\alpha_2\}
B′={α1,α2} be the ordered basis defined by
α
1
=
(
1
,
i
)
,
α
2
=
(
−
i
,
2
)
\alpha_1=(1,i),\alpha_2=(-i,2)
α1=(1,i),α2=(−i,2).
( a ) What is the matrix of
T
T
T relative to the pair
B
,
B
′
\mathfrak B,\mathfrak B'
B,B′?
( b ) What is the matrix of
T
T
T relative to the pair
B
′
,
B
\mathfrak B',\mathfrak B
B′,B?
( c ) What is the matrix of
T
T
T in the ordered basis
B
′
\mathfrak B'
B′?
( d ) What is the matrix of
T
T
T in the ordered basis
{
α
2
,
α
1
}
\{\alpha_2,\alpha_1\}
{α2,α1}?
Solution:
( a )
T
ϵ
1
=
2
α
1
−
i
α
2
,
T
ϵ
2
=
0
α
1
+
0
α
2
Tϵ_1=2α_1-iα_2,Tϵ_2=0α_1+0α_2
Tϵ1=2α1−iα2,Tϵ2=0α1+0α2, thus the matrix of
T
T
T relative to the pair
B
,
B
′
\mathfrak B,\mathfrak B'
B,B′ is
[
2
0
−
i
0
]
\begin{bmatrix}2&0\\-i&0\end{bmatrix}
[2−i00].
( b )
T
α
1
=
ϵ
1
,
T
α
2
=
−
i
ϵ
1
Tα_1=ϵ_1,Tα_2=-iϵ_1
Tα1=ϵ1,Tα2=−iϵ1, thus the matrix of
T
T
T relative to the pair
B
′
,
B
\mathfrak B',\mathfrak B
B′,B is
[
1
−
i
0
0
]
\begin{bmatrix}1&-i\\0&0\end{bmatrix}
[10−i0].
( c )
T
α
1
=
2
α
1
−
i
α
2
,
T
α
2
=
−
2
i
α
1
−
α
2
Tα_1=2α_1-iα_2,Tα_2=-2iα_1-α_2
Tα1=2α1−iα2,Tα2=−2iα1−α2, thus
[
T
]
B
′
=
[
2
−
2
i
−
i
−
1
]
[T]_{\mathfrak B'}=\begin{bmatrix}2&-2i\\-i&-1\end{bmatrix}
[T]B′=[2−i−2i−1].
( d )
T
α
2
=
−
α
2
−
2
i
α
1
,
T
α
1
=
−
i
α
2
+
2
α
1
Tα_2=-α_2-2iα_1,Tα_1=-iα_2+2α_1
Tα2=−α2−2iα1,Tα1=−iα2+2α1, thus
[
T
]
{
α
2
,
α
1
}
=
[
−
1
−
i
−
2
i
2
]
[T]_{\{α_2,α_1\}}=\begin{bmatrix}-1&-i\\-2i&2\end{bmatrix}
[T]{α2,α1}=[−1−2i−i2].
2.Let
T
T
T be the linear transformation from
R
3
R^3
R3 into
R
2
R^2
R2 defined by
T
(
x
1
,
x
2
,
x
3
)
=
(
x
1
+
x
2
,
2
x
3
−
x
1
)
T(x_1,x_2,x_3)=(x_1+x_2,2x_3-x_1)
T(x1,x2,x3)=(x1+x2,2x3−x1)
( a ) If
B
\mathfrak B
B is the standard ordered basis for
R
3
R^3
R3 and
B
′
\mathfrak B'
B′ is the standard ordered basis for
R
2
R^2
R2, what is the matrix of
T
T
T relative to the pair
B
,
B
′
\mathfrak B,\mathfrak B'
B,B′?
( b ) If
B
=
{
α
1
,
α
2
,
α
3
}
\mathfrak B=\{\alpha_1,\alpha_2,\alpha_3\}
B={α1,α2,α3} and
B
′
=
{
β
1
,
β
2
}
\mathfrak B'=\{\beta_1,\beta_2\}
B′={β1,β2}, where
α
1
=
(
1
,
0
,
−
1
)
,
α
2
=
(
1
,
1
,
1
)
,
α
3
=
(
1
,
0
,
0
)
,
β
1
=
(
0
,
1
)
,
β
2
=
(
1
,
0
)
\alpha_1=(1,0,-1),\quad\alpha_2=(1,1,1),\quad\alpha_3=(1,0,0),\\ \beta_1=(0,1),\quad\beta_2=(1,0)
α1=(1,0,−1),α2=(1,1,1),α3=(1,0,0),β1=(0,1),β2=(1,0)
what is the matrix of
T
T
T relative to the pair
B
,
B
′
\mathfrak B,\mathfrak B'
B,B′?
Solution:
( a ) We let
ϵ
1
,
ϵ
2
,
ϵ
3
ϵ_1,ϵ_2,ϵ_3
ϵ1,ϵ2,ϵ3 be the standard ordered basis for
R
3
R^3
R3 and
ϵ
1
′
,
ϵ
2
′
ϵ_1',ϵ_2'
ϵ1′,ϵ2′ be the standard ordered basis for
R
2
R^2
R2, then
T
ϵ
1
=
(
1
,
−
1
)
,
T
ϵ
2
=
(
1
,
0
)
,
T
ϵ
3
=
(
0
,
2
)
Tϵ_1=(1,-1),Tϵ_2=(1,0),Tϵ_3=(0,2)
Tϵ1=(1,−1),Tϵ2=(1,0),Tϵ3=(0,2)
thus the matrix of
T
T
T relative to the pair
B
′
,
B
\mathfrak B',\mathfrak B
B′,B is
A
=
[
1
1
0
−
1
0
2
]
A=\begin{bmatrix}1&1&0\\-1&0&2\end{bmatrix}
A=[1−11002]
( b ) A direct calculation shows
T
α
1
=
(
1
,
−
3
)
=
−
3
β
1
+
β
2
T
α
2
=
(
2
,
1
)
=
β
1
+
2
β
2
,
T
α
3
=
(
1
,
−
1
)
=
−
β
1
+
β
2
Tα_1=(1,-3)=-3β_1+β_2\quad Tα_2=(2,1)=β_1+2β_2,\quad Tα_3=(1,-1)=-β_1+β_2
Tα1=(1,−3)=−3β1+β2Tα2=(2,1)=β1+2β2,Tα3=(1,−1)=−β1+β2
thus the matrix of
T
T
T relative to the pair
B
′
,
B
\mathfrak B',\mathfrak B
B′,B is
A
=
[
−
3
1
−
1
1
2
1
]
A=\begin{bmatrix}-3&1&-1\\1&2&1\end{bmatrix}
A=[−3112−11]
3.Let
T
T
T be a linear operator on
F
n
F^n
Fn, let
A
A
A be the matrix of
T
T
T in the standard ordered basis for
F
n
F^n
Fn, and let
W
W
W be the subspace of
F
n
F^n
Fn spanned by the column vectors of
A
A
A. What does
W
W
W have to do with
T
T
T?
Solution: By the condition given, if we write
A
=
[
A
11
⋯
A
1
n
⋮
⋱
⋮
A
n
1
⋯
A
n
n
]
A=\begin{bmatrix}A_{11}&\cdots&A_{1n}\\\vdots&\ddots&\vdots\\A_{n1}&\cdots&A_{nn}\end{bmatrix}
A=⎣⎢⎡A11⋮An1⋯⋱⋯A1n⋮Ann⎦⎥⎤, then
T
ϵ
j
=
∑
i
=
1
n
A
i
j
ϵ
i
=
A
j
Tϵ_j=∑_{i=1}^nA_{ij} ϵ_i=A_j
Tϵj=∑i=1nAijϵi=Aj, the
j
j
j-th column vector of
A
A
A, thus if
W
W
W is spanned by
A
1
,
…
,
A
n
A_1,\dots,A_n
A1,…,An, it is easy to see
W
=
range
T
W=\text{range }T
W=range T.
4.Let
V
V
V be a two-dimensional vector space over the field
F
F
F, and let
B
\mathfrak B
B be an ordered basis for
V
V
V. If
T
T
T is a linear operator on
V
V
V and
[
T
]
B
=
[
a
b
c
d
]
[T]_{\mathfrak B}=\begin{bmatrix}a&b\\c&d\end{bmatrix}
[T]B=[acbd]
prove that
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
=
0
T^2-(a+d)T+(ad-bc)I=0
T2−(a+d)T+(ad−bc)I=0.
Solution: We write
B
=
{
α
1
,
α
2
}
\mathfrak B=\{α_1,α_2\}
B={α1,α2}, then for any
α
∈
V
α∈V
α∈V, we have
α
=
x
1
α
1
+
x
2
α
2
,
x
1
,
x
2
∈
F
α=x_1 α_1+x_2 α_2,x_1,x_2∈F
α=x1α1+x2α2,x1,x2∈F, notice that
[
T
α
1
]
B
=
[
T
]
B
[
α
1
]
B
=
[
a
b
c
d
]
[
1
0
]
=
[
a
c
]
,
[Tα_1 ]_{\mathfrak B}=[T]_{\mathfrak B} [α_1 ]_{\mathfrak B}=\begin{bmatrix}a&b\\c&d\end{bmatrix}\begin{bmatrix}1\\0\end{bmatrix}=\begin{bmatrix}a\\c\end{bmatrix},
[Tα1]B=[T]B[α1]B=[acbd][10]=[ac],
[
T
α
2
]
B
=
[
T
]
B
[
α
2
]
B
=
[
a
b
c
d
]
[
0
1
]
=
[
b
d
]
,
[Tα_2 ]_{\mathfrak B}=[T]_{\mathfrak B} [α_2 ]_{\mathfrak B}=\begin{bmatrix}a&b\\c&d\end{bmatrix}\begin{bmatrix}0\\1\end{bmatrix}=\begin{bmatrix}b\\d\end{bmatrix},
[Tα2]B=[T]B[α2]B=[acbd][01]=[bd],
[
T
2
α
1
]
B
=
[
T
]
B
[
T
α
1
]
B
=
[
a
b
c
d
]
[
a
c
]
=
[
a
2
+
b
c
a
c
+
c
d
]
,
[T^2 α_1 ]_{\mathfrak B}=[T]_{\mathfrak B} [Tα_1 ]_{\mathfrak B}=\begin{bmatrix}a&b\\c&d\end{bmatrix}\begin{bmatrix}a\\c\end{bmatrix}=\begin{bmatrix}a^2+bc\\ac+cd\end{bmatrix},
[T2α1]B=[T]B[Tα1]B=[acbd][ac]=[a2+bcac+cd],
[
T
2
α
2
]
B
=
[
T
]
B
[
T
α
2
]
B
=
[
a
b
c
d
]
[
b
d
]
=
[
a
b
+
b
d
b
c
+
d
2
]
[T^2 α_2 ]_{\mathfrak B}=[T]_{\mathfrak B} [Tα_2 ]_{\mathfrak B}=\begin{bmatrix}a&b\\c&d\end{bmatrix}\begin{bmatrix}b\\d\end{bmatrix}=\begin{bmatrix}ab+bd\\bc+d^2\end{bmatrix}
[T2α2]B=[T]B[Tα2]B=[acbd][bd]=[ab+bdbc+d2]
thus
(
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
)
α
1
=
T
2
α
1
−
(
a
+
d
)
T
α
1
+
(
a
d
−
b
c
)
α
1
=
(
a
2
+
b
c
)
α
1
+
(
a
c
+
c
d
)
α
2
−
(
a
+
d
)
(
a
α
1
+
c
α
2
)
+
(
a
d
−
b
c
)
α
1
=
(
a
2
+
b
c
−
a
2
−
a
d
+
a
d
−
b
c
)
α
1
+
(
a
c
+
c
d
−
a
c
−
c
d
)
α
2
=
0
\begin{aligned}&\quad(T^2-(a+d)T+(ad-bc)I) α_1\\&=T^2 α_1-(a+d)Tα_1+(ad-bc) α_1\\&=(a^2+bc) α_1+(ac+cd) α_2-(a+d)(aα_1+cα_2 )+(ad-bc) α_1\\&=(a^2+bc-a^2-ad+ad-bc) α_1+(ac+cd-ac-cd) α_2\\&=0\end{aligned}
(T2−(a+d)T+(ad−bc)I)α1=T2α1−(a+d)Tα1+(ad−bc)α1=(a2+bc)α1+(ac+cd)α2−(a+d)(aα1+cα2)+(ad−bc)α1=(a2+bc−a2−ad+ad−bc)α1+(ac+cd−ac−cd)α2=0
(
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
)
α
2
=
T
2
α
2
−
(
a
+
d
)
T
α
2
+
(
a
d
−
b
c
)
α
2
=
(
a
b
+
b
d
)
α
1
+
(
b
c
+
d
2
)
α
2
−
(
a
+
d
)
(
b
α
1
+
d
α
2
)
+
(
a
d
−
b
c
)
α
2
=
(
a
b
+
b
d
−
a
b
−
b
d
)
α
1
+
(
b
c
+
d
2
−
a
d
−
d
2
+
a
d
−
b
c
)
α
2
=
0
\begin{aligned}&\quad (T^2-(a+d)T+(ad-bc)I) α_2\\&=T^2 α_2-(a+d)Tα_2+(ad-bc) α_2\\&=(ab+bd) α_1+(bc+d^2 ) α_2-(a+d)(bα_1+dα_2 )+(ad-bc) α_2\\&=(ab+bd-ab-bd) α_1+(bc+d^2-ad-d^2+ad-bc) α_2\\&=0\end{aligned}
(T2−(a+d)T+(ad−bc)I)α2=T2α2−(a+d)Tα2+(ad−bc)α2=(ab+bd)α1+(bc+d2)α2−(a+d)(bα1+dα2)+(ad−bc)α2=(ab+bd−ab−bd)α1+(bc+d2−ad−d2+ad−bc)α2=0
Now we have
(
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
)
α
1
=
(
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
)
α
2
=
0
(T^2-(a+d)T+(ad-bc)I) α_1=(T^2-(a+d)T+(ad-bc)I) α_2=0
(T2−(a+d)T+(ad−bc)I)α1=(T2−(a+d)T+(ad−bc)I)α2=0, and so
(
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
)
α
=
(
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
)
(
x
1
α
1
+
x
2
α
2
)
=
x
1
(
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
)
α
1
+
x
2
(
T
2
−
(
a
+
d
)
T
+
(
a
d
−
b
c
)
I
)
α
2
=
0
\begin{aligned}&\quad (T^2-(a+d)T+(ad-bc)I)α\\&=(T^2-(a+d)T+(ad-bc)I)(x_1 α_1+x_2 α_2 )\\&=x_1 (T^2-(a+d)T+(ad-bc)I) α_1+x_2 (T^2-(a+d)T+(ad-bc)I) α_2\\&=0\end{aligned}
(T2−(a+d)T+(ad−bc)I)α=(T2−(a+d)T+(ad−bc)I)(x1α1+x2α2)=x1(T2−(a+d)T+(ad−bc)I)α1+x2(T2−(a+d)T+(ad−bc)I)α2=0
5.Let
T
T
T be the linear operator on
R
3
R^3
R3, the matrix of which in the standard ordered basis is
A
=
[
1
2
1
0
1
1
−
1
3
4
]
A=\begin{bmatrix}1&2&1\\0&1&1\\-1&3&4\end{bmatrix}
A=⎣⎡10−1213114⎦⎤
Find a basis for the range of
T
T
T and a basis for the null space of
T
T
T.
Solution: Using Exercise 3 we know the range of
T
T
T is spanned by the column vectors of
A
A
A, using elementary column operations we have
A
=
[
1
2
1
0
1
1
−
1
3
4
]
→
[
1
0
0
0
1
1
−
1
5
5
]
→
[
1
0
0
0
1
0
−
1
5
0
]
A=\begin{bmatrix}1&2&1\\0&1&1\\-1&3&4\end{bmatrix}→\begin{bmatrix}1&0&0\\0&1&1\\-1&5&5\end{bmatrix}→\begin{bmatrix}1&0&0\\0&1&0\\-1&5&0\end{bmatrix}
A=⎣⎡10−1213114⎦⎤→⎣⎡10−1015015⎦⎤→⎣⎡10−1015000⎦⎤
so a basis for the range of
T
T
T is
(
1
,
0
,
−
1
)
,
(
0
,
1
,
5
)
(1,0,-1),(0,1,5)
(1,0,−1),(0,1,5), notice that
T
ϵ
3
−
T
ϵ
1
=
T
ϵ
2
−
2
T
ϵ
1
Tϵ_3-Tϵ_1=Tϵ_2-2Tϵ_1
Tϵ3−Tϵ1=Tϵ2−2Tϵ1, thus
T
(
ϵ
3
+
ϵ
1
−
ϵ
2
)
=
T
(
1
,
−
1
,
1
)
=
0
T(ϵ_3+ϵ_1-ϵ_2 )=T(1,-1,1)=0
T(ϵ3+ϵ1−ϵ2)=T(1,−1,1)=0
and the dimension of the null space of
T
T
T is
1
1
1, so a basis for the null space of
T
T
T is
(
1
,
−
1
,
1
)
(1,-1,1)
(1,−1,1).
6.Let
T
T
T be the linear operator on
R
2
R^2
R2 defined by
T
(
x
1
,
x
2
)
=
(
−
x
2
,
x
1
)
T(x_1,x_2)=(-x_2,x_1)
T(x1,x2)=(−x2,x1)
( a ) What is the matrix of
T
T
T in the standard ordered basis for
R
2
R^2
R2?
( b ) What is the matrix of
T
T
T in the ordered basis
B
=
{
α
1
,
α
2
}
\mathfrak B=\{\alpha_1,\alpha_2\}
B={α1,α2}, where
α
1
=
(
1
,
2
)
\alpha_1=(1,2)
α1=(1,2) and
α
2
=
(
1
,
−
1
)
\alpha_2=(1,-1)
α2=(1,−1)?
( c ) Prove that for every real number
c
c
c the operator
(
T
−
c
I
)
(T-cI)
(T−cI) is invertible.
( d ) Prove that if
B
\mathfrak B
B is any ordered basis for
R
2
R^2
R2 and
[
T
]
B
=
A
[T]_{\mathfrak B}=A
[T]B=A, then
A
12
A
21
≠
0
A_{12}A_{21}\neq 0
A12A21=0.
Solution:
( a )
T
ϵ
1
=
(
0
,
1
)
=
ϵ
2
,
T
ϵ
2
=
(
−
1
,
0
)
=
−
ϵ
1
Tϵ_1=(0,1)=ϵ_2,Tϵ_2=(-1,0)=-ϵ_1
Tϵ1=(0,1)=ϵ2,Tϵ2=(−1,0)=−ϵ1, so the matrix of
T
T
T in the standard ordered basis for
R
2
R^2
R2 is
[
0
−
1
1
0
]
\begin{bmatrix}0&-1\\1&0\end{bmatrix}
[01−10].
( b ) We have
T
α
1
=
(
−
2
,
1
)
=
−
1
3
α
1
−
5
3
α
2
,
T
α
2
=
(
1
,
1
)
=
2
3
α
1
+
1
3
α
2
Tα_1=(-2,1)=-\frac{1}{3} α_1-\frac{5}{3} α_2,\quad Tα_2=(1,1)=\frac{2}{3}α_1+\frac{1}{3}α_2
Tα1=(−2,1)=−31α1−35α2,Tα2=(1,1)=32α1+31α2
thus
[
T
]
B
=
[
−
1
/
3
2
/
3
−
5
/
3
1
/
3
]
[T]_{\mathfrak B}=\begin{bmatrix}-1/3&2/3\\-5/3&1/3\end{bmatrix}
[T]B=[−1/3−5/32/31/3]
( c ) The matrix of
T
−
c
I
T-cI
T−cI in the standard ordered basis for
R
2
R^2
R2 is
[
−
c
−
1
1
−
c
]
\begin{bmatrix}-c&-1\\1&-c\end{bmatrix}
[−c1−1−c], thus
(
T
−
c
I
)
ϵ
1
=
(
−
c
,
1
)
,
(
T
−
c
I
)
ϵ
2
=
(
−
1
,
−
c
)
(T-cI) ϵ_1=(-c,1),\quad (T-cI) ϵ_2=(-1,-c)
(T−cI)ϵ1=(−c,1),(T−cI)ϵ2=(−1,−c)
since
(
−
c
,
1
)
,
(
−
1
,
−
c
)
(-c,1),(-1,-c)
(−c,1),(−1,−c) are linearly independent for any
c
∈
R
c∈R
c∈R, thus a basis of
R
2
R^2
R2, this means
T
−
c
I
T-cI
T−cI is invertible.
( d ) Let
B
′
=
{
ϵ
1
,
ϵ
2
}
\mathfrak B'=\{ϵ_1,ϵ_2\}
B′={ϵ1,ϵ2}, then
[
T
]
B
′
=
[
0
−
1
1
0
]
[T]_{\mathfrak B' }=\begin{bmatrix}0&-1\\1&0\end{bmatrix}
[T]B′=[01−10], given any
B
\mathfrak B
B, we can find an invertible
P
P
P s.t.
[
T
]
B
=
A
=
P
[
T
]
B
′
P
−
1
[T]_{\mathfrak B}=A=P[T]_{\mathfrak B'} P^{-1}
[T]B=A=P[T]B′P−1
we can assume
P
=
[
a
b
c
d
]
P=\begin{bmatrix}a&b\\c&d\end{bmatrix}
P=[acbd], then
P
−
1
=
1
a
d
−
b
c
[
d
−
b
−
c
a
]
P^{-1}=\frac{1}{ad-bc} \begin{bmatrix}d&-b\\-c&a\end{bmatrix}
P−1=ad−bc1[d−c−ba], obviously
a
,
b
,
c
,
d
a,b,c,d
a,b,c,d cannot be all zero.
A
=
1
a
d
−
b
c
[
a
b
c
d
]
[
0
−
1
1
0
]
[
d
−
b
−
c
a
]
=
1
a
d
−
b
c
[
b
−
a
d
−
c
]
[
d
−
b
−
c
a
]
=
1
a
d
−
b
c
[
b
d
+
a
c
−
b
2
−
a
2
d
2
+
c
2
−
b
d
−
a
c
]
\begin{aligned}A&=\frac{1}{ad-bc}\begin{bmatrix}a&b\\c&d\end{bmatrix}\begin{bmatrix}0&-1\\1&0\end{bmatrix}\begin{bmatrix}d&-b\\-c&a\end{bmatrix}\\&=\frac{1}{ad-bc}\begin{bmatrix}b&-a\\d&-c\end{bmatrix}\begin{bmatrix}d&-b\\-c&a\end{bmatrix}\\&=\frac{1}{ad-bc}\begin{bmatrix}bd+ac&-b^2-a^2\\d^2+c^2&-bd-ac\end{bmatrix}\end{aligned}
A=ad−bc1[acbd][01−10][d−c−ba]=ad−bc1[bd−a−c][d−c−ba]=ad−bc1[bd+acd2+c2−b2−a2−bd−ac]
thus
A
12
A
21
=
−
(
a
2
+
b
2
)
(
c
2
+
d
2
)
a
d
−
b
c
≠
0
A_{12} A_{21}=-\frac{(a^2+b^2 )(c^2+d^2)}{ad-bc}\neq 0
A12A21=−ad−bc(a2+b2)(c2+d2)=0.
7.Let
T
T
T be the linear operator on
R
3
R^3
R3 defined by
T
(
x
1
,
x
2
,
x
3
)
=
(
3
x
1
+
x
3
,
−
2
x
1
+
x
2
,
−
x
1
+
2
x
2
+
4
x
3
)
T(x_1,x_2,x_3)=(3x_1+x_3,-2x_1+x_2,-x_1+2x_2+4x_3)
T(x1,x2,x3)=(3x1+x3,−2x1+x2,−x1+2x2+4x3).
( a ) What is the matrix of
T
T
T in the standard ordered basis for
R
3
R^3
R3?
( b ) What is the matrix of
T
T
T in the ordered basis
{
α
1
,
α
2
,
α
3
}
\{\alpha_1,\alpha_2,\alpha_3\}
{α1,α2,α3}, where
α
1
=
(
1
,
0
,
1
)
,
α
2
=
(
−
1
,
2
,
1
)
,
α
3
=
(
2
,
1
,
1
)
\alpha_1=(1,0,1),\alpha_2=(-1,2,1),\alpha_3=(2,1,1)
α1=(1,0,1),α2=(−1,2,1),α3=(2,1,1)?
( c ) Prove that
T
T
T is invertible and give a rule for
T
−
1
T^{-1}
T−1 like the one which defines
T
T
T.
Solution:
( a ) The matrix of
T
T
T in the standard ordered basis for
R
3
R^3
R3 is
[
3
0
1
−
2
1
0
−
1
2
4
]
\begin{bmatrix}3&0&1\\-2&1&0\\-1&2&4\end{bmatrix}
⎣⎡3−2−1012104⎦⎤.
(b) We form
P
P
P where
P
j
=
[
α
j
]
{
ϵ
1
,
ϵ
2
,
ϵ
3
}
P_j=[α_j ]_{\{ϵ_1,ϵ_2,ϵ_3\}}
Pj=[αj]{ϵ1,ϵ2,ϵ3}, it is easy to see
P
=
[
1
−
1
2
0
2
1
1
1
1
]
P=\begin{bmatrix}1&-1&2\\0&2&1\\1&1&1\end{bmatrix}
P=⎣⎡101−121211⎦⎤, and we perform
[
1
−
1
2
1
0
0
0
2
1
0
1
0
1
1
1
0
0
1
]
→
[
1
−
1
2
1
0
0
0
2
1
0
1
0
0
2
−
1
−
1
0
1
]
→
[
1
−
1
2
1
0
0
0
2
1
0
1
0
0
0
−
2
−
1
−
1
1
]
→
[
1
−
1
0
0
−
1
1
0
2
0
−
1
/
2
1
/
2
1
/
2
0
0
1
1
/
2
1
/
2
−
1
/
2
]
→
[
1
0
0
−
1
/
4
−
3
/
4
5
/
4
0
1
0
−
1
/
4
1
/
4
1
/
4
0
0
1
1
/
2
1
/
2
−
1
/
2
]
\begin{aligned}\begin{bmatrix}1&-1&2&1&0&0\\0&2&1&0&1&0\\1&1&1&0&0&1\end{bmatrix}&→\begin{bmatrix}1&-1&2&1&0&0\\0&2&1&0&1&0\\0&2&-1&-1&0&1\end{bmatrix}\\&→\begin{bmatrix}1&-1&2&1&0&0\\0&2&1&0&1&0\\0&0&-2&-1&-1&1\end{bmatrix}\\&→\begin{bmatrix}1&-1&0&0&-1&1\\0&2&0&-1/2&1/2&1/2\\0&0&1&1/2&1/2&-1/2\end{bmatrix}\\&→\begin{bmatrix}1&0&0&-1/4&-3/4&5/4\\0&1&0&-1/4&1/4&1/4\\0&0&1&1/2&1/2&-1/2\end{bmatrix}\end{aligned}
⎣⎡101−121211100010001⎦⎤→⎣⎡100−12221−110−1010001⎦⎤→⎣⎡100−12021−210−101−1001⎦⎤→⎣⎡100−1200010−1/21/2−11/21/211/2−1/2⎦⎤→⎣⎡100010001−1/4−1/41/2−3/41/41/25/41/4−1/2⎦⎤
thus
P
−
1
=
1
/
4
[
−
1
−
3
5
−
1
1
1
2
2
−
2
]
P^{-1}=1/4 \begin{bmatrix}-1&-3&5\\-1&1&1\\2&2&-2\end{bmatrix}
P−1=1/4⎣⎡−1−12−31251−2⎦⎤
and then
[
T
]
{
α
1
,
α
2
,
α
3
}
=
P
−
1
[
T
]
{
ϵ
1
,
ϵ
2
,
ϵ
3
}
P
=
1
4
[
−
1
−
3
5
−
1
1
1
2
2
−
2
]
[
3
0
1
−
2
1
0
−
1
2
4
]
[
1
−
1
2
0
2
1
1
1
1
]
=
1
4
[
−
2
7
19
−
6
3
3
4
−
2
−
6
]
[
1
−
1
2
0
2
1
1
1
1
]
=
1
4
[
17
35
22
−
3
15
−
6
−
2
−
14
0
]
\begin{aligned}[T]_{\{α_1,α_2,α_3\}} &=P^{-1} [T]_{\{ϵ_1,ϵ_2,ϵ_3\}} P\\&=\frac{1}{4} \begin{bmatrix}-1&-3&5\\-1&1&1\\2&2&-2\end{bmatrix}\begin{bmatrix}3&0&1\\-2&1&0\\-1&2&4\end{bmatrix}\begin{bmatrix}1&-1&2\\0&2&1\\1&1&1\end{bmatrix}\\&=\frac{1}{4} \begin{bmatrix}-2&7&19\\-6&3&3\\4&-2&-6\end{bmatrix}\begin{bmatrix}1&-1&2\\0&2&1\\1&1&1\end{bmatrix}\\&=\frac{1}{4}\begin{bmatrix}17&35&22\\-3&15&-6\\-2&-14&0\end{bmatrix}\end{aligned}
[T]{α1,α2,α3}=P−1[T]{ϵ1,ϵ2,ϵ3}P=41⎣⎡−1−12−31251−2⎦⎤⎣⎡3−2−1012104⎦⎤⎣⎡101−121211⎦⎤=41⎣⎡−2−6473−2193−6⎦⎤⎣⎡101−121211⎦⎤=41⎣⎡17−3−23515−1422−60⎦⎤
( c ) It is enough to prove
[
T
]
{
ϵ
1
,
ϵ
2
,
ϵ
3
}
[T]_{\{ϵ_1,ϵ_2,ϵ_3\}}
[T]{ϵ1,ϵ2,ϵ3} is invertible, this is true since
[
3
0
1
−
2
1
0
−
1
2
4
]
−
1
=
[
4
/
9
2
/
9
−
1
/
9
8
/
9
13
/
9
−
2
/
9
−
1
/
3
−
2
/
3
1
/
3
]
\begin{bmatrix}3&0&1\\-2&1&0\\-1&2&4\end{bmatrix}^{-1}=\begin{bmatrix}4/9&2/9&-1/9\\8/9&13/9&-2/9\\-1/3&-2/3&1/3\end{bmatrix}
⎣⎡3−2−1012104⎦⎤−1=⎣⎡4/98/9−1/32/913/9−2/3−1/9−2/91/3⎦⎤
since we know
[
T
−
1
]
{
ϵ
1
,
ϵ
2
,
ϵ
3
}
=
(
[
T
]
{
ϵ
1
,
ϵ
2
,
ϵ
3
}
)
−
1
[T^{-1}]_{\{ϵ_1,ϵ_2,ϵ_3\}}=([T]_{\{ϵ_1,ϵ_2,ϵ_3\}})^{-1}
[T−1]{ϵ1,ϵ2,ϵ3}=([T]{ϵ1,ϵ2,ϵ3})−1, it’s able to describe
T
−
1
(
x
1
,
x
2
,
x
3
)
=
(
4
9
x
1
+
2
9
x
2
−
1
9
x
3
,
8
9
x
1
+
13
9
x
2
−
2
9
x
3
,
−
1
3
x
1
−
2
3
x
2
+
1
3
x
3
)
T^{-1} (x_1,x_2,x_3 )=\left(\frac{4}{9} x_1+\frac{2}{9} x_2-\frac{1}{9} x_3,\frac{8}{9} x_1+\frac{13}{9} x_2-\frac{2}{9} x_3,-\frac{1}{3} x_1-\frac{2}{3} x_2+\frac{1}{3} x_3\right)
T−1(x1,x2,x3)=(94x1+92x2−91x3,98x1+913x2−92x3,−31x1−32x2+31x3)
8.Let
θ
\theta
θ be a real number. Prove that the following two matrices are similar over the field of complex numbers:
[
cos
θ
−
sin
θ
sin
θ
cos
θ
]
,
[
e
i
θ
0
0
e
−
i
θ
]
\begin{bmatrix}\cos{\theta}&-\sin{\theta}\\\sin{\theta}&\cos{\theta}\end{bmatrix},\quad \begin{bmatrix}e^{i{\theta}}&0\\0&e^{-i\theta}\end{bmatrix}
[cosθsinθ−sinθcosθ],[eiθ00e−iθ]
Solution: Let
T
T
T be the linear operator on
C
2
C^2
C2 which is represented by the first matrix in the standard ordered basis, then
T
(
1
,
0
)
=
(
cos
θ
,
sin
θ
)
,
T
(
0
,
1
)
=
(
−
sin
θ
,
cos
θ
)
T(1,0)=(\cosθ,\sinθ), T(0,1)=(-\sinθ,\cosθ)
T(1,0)=(cosθ,sinθ),T(0,1)=(−sinθ,cosθ), let
α
1
=
(
i
,
1
)
,
α
2
=
(
1
,
i
)
α_1=(i,1),α_2=(1,i)
α1=(i,1),α2=(1,i), then
T
α
1
=
i
(
cos
θ
,
sin
θ
)
+
(
−
sin
θ
,
cos
θ
)
=
(
i
cos
θ
−
sin
θ
,
i
sin
θ
+
cos
θ
)
=
e
i
θ
(
i
,
1
)
=
e
i
θ
α
1
Tα_1=i(\cosθ,\sinθ )+(-\sinθ,\cosθ )=(i \cosθ-\sinθ,i \sinθ+\cosθ )=e^{iθ} (i,1)=e^{iθ} α_1
Tα1=i(cosθ,sinθ)+(−sinθ,cosθ)=(icosθ−sinθ,isinθ+cosθ)=eiθ(i,1)=eiθα1
and similarly we can see
T
α
2
=
e
−
i
θ
α
2
Tα_2=e^{-iθ} α_2
Tα2=e−iθα2, it is easy to see
{
α
1
,
α
2
}
\{α_1,α_2 \}
{α1,α2} are linearly independent, thus a basis of
C
2
C^2
C2, since
[
T
]
{
α
1
,
α
2
}
=
[
e
i
θ
0
0
e
−
i
θ
]
[T]_{\{α_1,α_2\}} =\begin{bmatrix}e^{iθ}&0\\0&e^{-iθ} \end{bmatrix}
[T]{α1,α2}=[eiθ00e−iθ], the two matrices are similar, with
P
=
[
i
1
1
i
]
P=\begin{bmatrix}i&1\\1&i\end{bmatrix}
P=[i11i] and
[
T
]
{
α
1
,
α
2
}
=
P
−
1
[
cos
θ
−
sin
θ
sin
θ
cos
θ
]
P
[T]_{\{α_1,α_2\}} =P^{-1} \begin{bmatrix}\cosθ&-\sinθ\\\sinθ&\cosθ \end{bmatrix}P
[T]{α1,α2}=P−1[cosθsinθ−sinθcosθ]P
9.Let
V
V
V be a finite-dimensional vector space over the field
F
F
F and let
S
S
S and
T
T
T be linear operators on
V
V
V. We ask: When do there exist ordered bases
B
\mathfrak B
B and
B
′
\mathfrak B'
B′ for
V
V
V such that
[
S
]
B
=
[
T
]
B
′
[S]_{\mathfrak B}=[T]_{\mathfrak B'}
[S]B=[T]B′? Prove that such bases exist if and only if there is an invertible linear operator
U
U
U on
V
V
V such that
T
=
U
S
U
−
1
T=USU^{-1}
T=USU−1.
Solution: If
[
S
]
B
=
[
T
]
B
′
[S]_{\mathfrak B}=[T]_{\mathfrak B'}
[S]B=[T]B′, then let
U
U
U be the linear operator that carries
B
{\mathfrak B}
B onto
B
′
{\mathfrak B'}
B′, i.e., if we let
B
=
{
a
1
,
…
,
a
n
}
,
B
′
=
{
b
1
,
…
,
b
n
}
{\mathfrak B}=\{a_1,\dots,a_n \},\quad {\mathfrak B'}=\{b_1,\dots,b_n \}
B={a1,…,an},B′={b1,…,bn}
and define
U
U
U by
U
a
i
=
b
i
,
i
=
1
,
…
,
n
Ua_i=b_i,i=1,\dots,n
Uai=bi,i=1,…,n. Then
U
U
U is invertible since it carries a basis onto another basis, and
U
−
1
b
i
=
a
i
,
i
=
1
,
…
,
n
U^{-1} b_i=a_i,i=1,\dots,n
U−1bi=ai,i=1,…,n. If we denote
[
S
]
B
=
[
T
]
B
′
=
A
=
(
A
i
j
)
[S]_{\mathfrak B}=[T]_{\mathfrak B'}=A=(A_{ij})
[S]B=[T]B′=A=(Aij), then by definition we have
S
a
j
=
∑
i
=
1
n
A
i
j
a
i
,
T
b
j
=
∑
i
=
1
n
A
i
j
b
i
Sa_j=∑_{i=1}^nA_{ij} a_i ,Tb_j=∑_{i=1}^nA_{ij} b_i
Saj=∑i=1nAijai,Tbj=∑i=1nAijbi, so
U
S
U
−
1
(
b
j
)
=
U
S
a
j
=
U
(
∑
i
=
1
n
A
i
j
a
i
)
=
∑
i
=
1
n
A
i
j
U
a
i
=
∑
i
=
1
n
A
i
j
b
i
=
T
b
j
,
j
=
1
,
…
,
n
\begin{aligned}USU^{-1}(b_j )&=USa_j=U\left(∑_{i=1}^nA_{ij} a_i \right)\\&=∑_{i=1}^nA_{ij} Ua_i=∑_{i=1}^nA_{ij}b_i\\&=Tb_j,\quad j=1,\dots,n\end{aligned}
USU−1(bj)=USaj=U(i=1∑nAijai)=i=1∑nAijUai=i=1∑nAijbi=Tbj,j=1,…,n
Since
U
S
U
−
1
USU^{-1}
USU−1 and
T
T
T are equal on a basis of
V
V
V, we have
T
=
U
S
U
−
1
T=USU^{-1}
T=USU−1.
Conversely, if
T
=
U
S
U
−
1
T=USU^{-1}
T=USU−1 for some invertible
U
U
U, we let
B
=
{
a
1
,
…
,
a
n
}
B=\{a_1,\dots,a_n \}
B={a1,…,an} be an ordered basis for
V
V
V, and
B
′
=
{
U
a
1
,
…
,
U
a
n
}
\mathfrak B'=\{Ua_1,\dots,Ua_n \}
B′={Ua1,…,Uan}, since
U
U
U is invertible,
B
′
\mathfrak B'
B′ is a basis for
V
V
V. Notice that if
α
∈
V
α∈V
α∈V, then
α
=
k
1
a
1
+
⋯
+
k
n
a
n
,
k
1
,
…
,
k
n
∈
F
α=k_1 a_1+\dots+k_n a_n,\quad k_1,\dots,k_n∈F
α=k1a1+⋯+knan,k1,…,kn∈F
thus
[
α
]
B
=
[
k
1
⋮
k
n
]
[α]_{\mathfrak B}=\begin{bmatrix}k_1\\\vdots\\k_n\end{bmatrix}
[α]B=⎣⎢⎡k1⋮kn⎦⎥⎤, and
U
α
=
k
1
U
a
1
+
⋯
+
k
n
U
a
n
Uα=k_1 Ua_1+\dots+k_n Ua_n
Uα=k1Ua1+⋯+knUan, so
[
U
α
]
B
′
=
[
k
1
⋮
k
n
]
[Uα]_{\mathfrak B'}=\begin{bmatrix}k_1\\\vdots\\k_n\end{bmatrix}
[Uα]B′=⎣⎢⎡k1⋮kn⎦⎥⎤, so we have
[
α
]
B
=
[
U
α
]
B
′
[α]_{\mathfrak B}=[Uα]_{\mathfrak B'}
[α]B=[Uα]B′, from this and the fact that
T
U
=
U
S
TU=US
TU=US we can have
[
S
]
B
[
α
]
B
=
[
S
α
]
B
=
[
U
S
α
]
B
′
=
[
T
U
α
]
B
′
=
[
T
]
B
′
[
U
α
]
B
′
=
[
T
]
B
′
[
α
]
B
[S]_{\mathfrak B} [α]_{\mathfrak B}=[Sα]_{\mathfrak B}=[USα]_{\mathfrak B'}=[TUα]_{\mathfrak B'}=[T]_{\mathfrak B'} [Uα]_{\mathfrak B'}=[T]_{\mathfrak B'} [α]_{\mathfrak B}
[S]B[α]B=[Sα]B=[USα]B′=[TUα]B′=[T]B′[Uα]B′=[T]B′[α]B
and it must follow that
[
S
]
B
=
[
T
]
B
′
[S]_{\mathfrak B}=[T]_{\mathfrak B'}
[S]B=[T]B′.
[Alternatively, one easier proof is using Theorem 14 and the consequence of Theorem 13, since in this case we have
[
T
]
B
′
=
[
U
−
1
]
B
[
T
]
B
[
U
]
B
=
[
U
−
1
T
U
]
B
=
[
U
−
1
(
U
S
U
−
1
)
U
]
B
=
[
S
]
B
[T]_{\mathfrak B'}=[U^{-1}]_{\mathfrak B} [T]_{\mathfrak B} [U]_{\mathfrak B}=[U^{-1} TU]_{\mathfrak B}=[U^{-1}(USU^{-1})U]_{\mathfrak B}=[S]_{\mathfrak B}
[T]B′=[U−1]B[T]B[U]B=[U−1TU]B=[U−1(USU−1)U]B=[S]B ].
10.We have seen that the linear operator
T
T
T on
R
2
R^2
R2 defined by
T
(
x
1
,
x
2
)
=
(
x
1
,
0
)
T(x_1,x_2)=(x_1,0)
T(x1,x2)=(x1,0) is represented in the standard ordered basis by the matrix
A
=
[
1
0
0
0
]
.
A=\begin{bmatrix}1&0\\0&0\end{bmatrix}.
A=[1000].
This operator satisfies
T
2
=
T
T^2=T
T2=T. Prove that if
S
S
S is a linear operator on
R
2
R^2
R2 such that
S
2
=
S
S^2=S
S2=S, then
S
=
0
S=0
S=0,or
S
=
I
S=I
S=I, or there is an ordered basis
B
\mathfrak B
B for
R
2
R^2
R2 such that
[
S
]
B
=
A
[S]_{\mathfrak B}=A
[S]B=A (above).
Solution: If
S
=
0
S=0
S=0 or
S
=
I
S=I
S=I, we obviously have
S
2
=
S
S^2=S
S2=S, now suppose
S
≠
0
,
S
≠
I
S\neq 0,S\neq I
S=0,S=I, but
S
2
=
S
S^2=S
S2=S, then it is able to find
α
,
β
∈
R
2
α,β∈R^2
α,β∈R2, s.t.
S
α
≠
0
,
S
β
≠
β
Sα\neq 0,Sβ\neq β
Sα=0,Sβ=β, since
S
α
∈
range
S
Sα∈\text{range }S
Sα∈range S, we have
dim
range
S
≥
1
\dim \text{range }S≥1
dimrange S≥1, also since
S
(
S
β
−
β
)
=
S
2
β
−
S
β
=
S
β
−
S
β
=
0
S(Sβ-β)=S^2 β-Sβ=Sβ-Sβ=0
S(Sβ−β)=S2β−Sβ=Sβ−Sβ=0, we know
S
β
−
β
∈
null
S
Sβ-β∈\text{null }S
Sβ−β∈null S, and
dim
null
S
≥
1
\dim \text{null }S≥1
dimnull S≥1, as we discuss in
R
2
R^2
R2,
dim
range
S
+
dim
null
S
=
2
\dim \text{range }S+\dim \text{null }S=2
dimrange S+dimnull S=2, thus
dim
range
S
=
dim
null
S
=
1
\dim \text{range }S=\dim \text{null }S=1
dimrange S=dimnull S=1. Now if we let
a
=
S
α
,
b
=
S
β
−
β
a=Sα,b=Sβ-β
a=Sα,b=Sβ−β, and
B
=
{
a
,
b
}
{\mathfrak B}=\{a,b\}
B={a,b}, then
B
{\mathfrak B}
B is an ordered basis for
R
2
R^2
R2 and
[
S
]
B
=
A
[S]_{\mathfrak B}=A
[S]B=A.
11.Let
W
W
W be the space of all
n
×
1
n\times 1
n×1 column matrices over a field
F
F
F. If
A
A
A is an
n
×
n
n\times n
n×n matrix over
F
F
F, then
A
A
A defines a linear operator
L
A
L_A
LA on
W
W
W through left multiplicaition:
L
A
(
X
)
=
A
X
L_A(X)=AX
LA(X)=AX. Prove that every linear operator on
W
W
W is left multiplication by some
n
×
n
n\times n
n×n matrix, i.e., is
L
A
L_A
LA for some
A
A
A.
Now suppose
V
V
V is an
n
n
n-dimensional vector space over the field
F
F
F, and let
B
\mathfrak B
B be an ordered basis for
V
V
V. For each
α
\alpha
α in
V
V
V, define
U
α
=
[
α
]
B
U\alpha =[\alpha]_{\mathfrak B}
Uα=[α]B. Prove that
U
U
U is an isomorphism of
V
V
V onto
W
W
W. If
T
T
T is a linear operator on
V
V
V, then
U
T
U
−
1
UTU^{-1}
UTU−1 is a linear operator on
W
W
W. Accordingly,
U
T
U
−
1
UTU^{-1}
UTU−1 is left multiplication by some
n
×
n
n\times n
n×n matrix
A
A
A. What is
A
A
A?
Solution: If
T
T
T is a linear operator on
W
W
W, let
B
′
=
{
ϵ
1
,
…
,
ϵ
n
}
\mathfrak B'=\{ϵ_1,\dots,ϵ_n \}
B′={ϵ1,…,ϵn} be the standard basis on
W
W
W, for each
X
=
[
x
1
⋮
x
n
]
X=\begin{bmatrix}x_1\\\vdots\\x_n \end{bmatrix}
X=⎣⎢⎡x1⋮xn⎦⎥⎤, we have
X
=
∑
j
=
1
n
x
j
ϵ
j
X=∑_{j=1}^nx_j ϵ_j
X=∑j=1nxjϵj, and if we define
A
:
=
[
T
]
B
′
A:=[T]_{\mathfrak B'}
A:=[T]B′, then
T
(
X
)
=
T
(
∑
j
=
1
n
x
j
ϵ
j
)
=
∑
j
=
1
n
x
j
T
ϵ
j
=
∑
j
=
1
n
x
j
∑
i
=
1
n
A
i
j
ϵ
i
=
∑
i
=
1
n
(
∑
j
=
1
n
A
i
j
x
j
)
ϵ
i
=
A
X
T(X)=T(∑_{j=1}^nx_j ϵ_j )=∑_{j=1}^nx_j Tϵ_j =∑_{j=1}^nx_j ∑_{i=1}^nA_{ij} ϵ_i =∑_{i=1}^n(∑_{j=1}^nA_{ij} x_j )ϵ_i =AX
T(X)=T(∑j=1nxjϵj)=∑j=1nxjTϵj=∑j=1nxj∑i=1nAijϵi=∑i=1n(∑j=1nAijxj)ϵi=AX, thus
T
=
L
A
T=L_A
T=LA.
For the second question, let
B
=
{
a
1
,
…
,
a
n
}
\mathfrak B=\{a_1,\dots,a_n \}
B={a1,…,an}, if
[
α
]
B
=
[
β
]
B
=
[
x
1
⋮
x
n
]
[α]_{\mathfrak B}=[β]_{\mathfrak B}=\begin{bmatrix}x_1\\\vdots \\x_n\end{bmatrix}
[α]B=[β]B=⎣⎢⎡x1⋮xn⎦⎥⎤, then
α
=
β
=
∑
j
=
1
n
x
j
a
j
α=β=∑_{j=1}^nx_j a_j
α=β=∑j=1nxjaj , this shows
(
U
α
=
U
β
)
⇒
(
α
=
β
)
(Uα=Uβ)⇒(α=β)
(Uα=Uβ)⇒(α=β), so
U
U
U is injective. Also for any
X
=
[
x
1
⋮
x
n
]
∈
W
X=\begin{bmatrix}x_1\\\vdots\\x_n \end{bmatrix}∈W
X=⎣⎢⎡x1⋮xn⎦⎥⎤∈W, define
α
=
∑
j
=
1
n
x
j
a
j
α=∑_{j=1}^nx_j a_j
α=∑j=1nxjaj, then
U
α
=
X
Uα=X
Uα=X, thus
U
U
U is surjective, combined we show
U
U
U is an isomorphism.
If
T
T
T is a linear operator on
V
V
V, then let
X
=
[
x
1
⋮
x
n
]
,
Y
=
[
y
1
⋮
y
n
]
∈
W
X=\begin{bmatrix}x_1\\\vdots\\x_n\end{bmatrix},Y=\begin{bmatrix}y_1\\\vdots\\y_n \end{bmatrix}∈W
X=⎣⎢⎡x1⋮xn⎦⎥⎤,Y=⎣⎢⎡y1⋮yn⎦⎥⎤∈W, by definition we have
U
T
U
−
1
(
c
X
+
Y
)
=
U
T
(
U
−
1
(
c
X
+
Y
)
)
=
U
T
(
∑
j
=
1
n
(
c
x
j
+
y
j
)
a
j
)
=
U
(
T
(
∑
j
=
1
n
(
c
x
j
+
y
j
)
a
j
)
)
=
U
(
∑
j
=
1
n
(
c
x
j
+
y
j
)
T
a
j
)
=
∑
j
=
1
n
(
c
x
j
+
y
j
)
U
T
a
j
=
c
∑
j
=
1
n
x
j
U
T
a
j
+
∑
j
=
1
n
y
j
U
T
a
j
=
c
(
U
T
∑
j
=
1
n
x
j
a
j
)
+
U
T
(
∑
j
=
1
n
y
j
a
j
)
=
c
(
U
T
(
U
−
1
(
X
)
)
)
+
(
U
T
(
U
−
1
(
Y
)
)
)
=
c
(
U
T
U
−
1
)
(
X
)
+
(
U
T
U
−
1
)
(
Y
)
\begin{aligned}UTU^{-1} (cX+Y)&=UT(U^{-1} (cX+Y))\\&=UT\left(∑_{j=1}^n(cx_j+y_j)a_j \right)=U\left(T\Big(∑_{j=1}^n(cx_j+y_j)a_j \Big)\right)\\&=U\left(∑_{j=1}^n(cx_j+y_j)Ta_j \right)=∑_{j=1}^n(cx_j+y_j)UTa_j \\&=c∑_{j=1}^nx_j UTa_j +∑_{j=1}^ny_j UTa_j =c\left(UT∑_{j=1}^nx_j a_j \right)+UT\left(∑_{j=1}^ny_j a_j \right)\\&=c(UT(U^{-1}(X)))+(UT(U^{-1}(Y)))\\&=c(UTU^{-1})(X)+(UTU^{-1})(Y)\end{aligned}
UTU−1(cX+Y)=UT(U−1(cX+Y))=UT(j=1∑n(cxj+yj)aj)=U(T(j=1∑n(cxj+yj)aj))=U(j=1∑n(cxj+yj)Taj)=j=1∑n(cxj+yj)UTaj=cj=1∑nxjUTaj+j=1∑nyjUTaj=c(UTj=1∑nxjaj)+UT(j=1∑nyjaj)=c(UT(U−1(X)))+(UT(U−1(Y)))=c(UTU−1)(X)+(UTU−1)(Y)
thus
U
T
U
−
1
UTU^{-1}
UTU−1 is a linear operator on
W
W
W.
If
T
T
T is a linear operator on
V
V
V, then let
C
=
[
T
]
B
C=[T]_{\mathfrak B}
C=[T]B, i.e.
T
a
j
=
∑
j
=
1
n
C
i
j
a
i
Ta_j=∑_{j=1}^nC_{ij} a_i
Taj=∑j=1nCijai, in the first part we proved
A
=
[
U
T
U
−
1
]
B
′
A=[UTU^{-1}]_{\mathfrak B'}
A=[UTU−1]B′, to compute
A
A
A, we see that
U
T
U
−
1
(
ϵ
j
)
=
U
T
a
j
=
U
(
∑
j
=
1
n
C
i
j
a
i
)
=
∑
j
=
1
n
C
i
j
U
a
i
=
∑
j
=
1
n
C
i
j
ϵ
j
UTU^{-1}(ϵ_j )=UTa_j=U(∑_{j=1}^nC_{ij}a_i)=∑_{j=1}^nC_{ij}Ua_i =∑_{j=1}^nC_{ij}ϵ_j
UTU−1(ϵj)=UTaj=U(∑j=1nCijai)=∑j=1nCijUai=∑j=1nCijϵj , thus
[
U
T
U
−
1
]
B
′
=
C
=
[
T
]
B
[UTU^{-1}]_{\mathfrak B'}=C=[T]_{\mathfrak B}
[UTU−1]B′=C=[T]B, or
A
=
[
T
]
B
A=[T]_{\mathfrak B}
A=[T]B.
12.Let
V
V
V be an
n
n
n-dimensional vector space over the field
F
F
F, and let
B
=
{
α
1
,
…
,
α
n
}
\mathfrak B=\{\alpha_1,\dots,\alpha_n\}
B={α1,…,αn} be an ordered basis for
V
V
V.
( a ) According to Theorem 1, there is a unique linear operator
T
T
T on
V
V
V such that
T
α
j
=
α
j
+
1
,
j
=
1
,
…
,
n
−
1
,
T
α
n
=
0.
T{\alpha}_j={\alpha}_{j+1},\qquad j=1,\dots,n-1,\qquad T{\alpha}_n=0.
Tαj=αj+1,j=1,…,n−1,Tαn=0.
What is the matrix
A
A
A of
T
T
T in the ordered basis
B
\mathfrak B
B?
( b ) Prove that
T
n
=
0
T^n=0
Tn=0 but
T
n
−
1
≠
0
T^{n-1}\neq 0
Tn−1=0.
( c ) Let
S
S
S be any linear operator on
V
V
V such that
S
n
=
0
S^n=0
Sn=0 but
S
n
−
1
≠
0
S^{n-1}\neq 0
Sn−1=0. Prove that there is an ordered basis
B
′
\mathfrak B'
B′ for
V
V
V such that the matrix of
S
S
S in the ordered basis
B
′
\mathfrak B'
B′ is the matrix
A
A
A of part (a).
( d ) Prove that if
M
M
M and
N
N
N are
n
×
n
n\times n
n×n matrices over
F
F
F such that
M
n
=
N
n
=
0
M^n=N^n=0
Mn=Nn=0 but
M
n
−
1
≠
0
≠
N
n
−
1
M^{n-1}\neq 0\neq N^{n-1}
Mn−1=0=Nn−1, then
M
M
M and
N
N
N are similar.
Solution:
( a ) A direct computation shows
A
=
[
0
0
…
0
1
0
⋱
⋱
0
1
0
]
A=\begin{bmatrix}0&0&\dots&0\\1&0& & \\ &\ddots&\ddots& \\0& &1&0\end{bmatrix}
A=⎣⎢⎢⎡01000⋱…⋱100⎦⎥⎥⎤.
( b ) We have
T
k
α
n
+
1
−
k
=
0
,
k
=
1
,
…
,
n
T^k α_{n+1-k}=0,k=1,\dots,n
Tkαn+1−k=0,k=1,…,n, thus
T
n
=
0
T^n=0
Tn=0, but
T
n
−
1
α
1
=
α
n
≠
0
T^{n-1}α_1=α_n\neq 0
Tn−1α1=αn=0.
( c ) It is able to choose
α
α
α s.t.
S
n
−
1
α
≠
0
S^{n-1}α\neq 0
Sn−1α=0 but
S
n
α
=
0
S^n α=0
Snα=0, notice
α
≠
0
α\neq 0
α=0, and
{
α
,
S
α
,
…
,
S
n
−
1
α
}
\{α,Sα,\dots,S^{n-1}α\}
{α,Sα,…,Sn−1α} is linearly independent, for if we have
k
1
α
+
k
2
S
α
+
⋯
+
k
n
S
n
−
1
α
=
0
k_1 α+k_2 Sα+\dots+k_n S^{n-1} α=0
k1α+k2Sα+⋯+knSn−1α=0
then
S
n
−
1
(
k
1
α
+
k
2
S
α
+
⋯
+
k
n
S
n
−
1
α
)
=
k
1
S
n
−
1
α
=
0
S^{n-1}(k_1 α+k_2 Sα+\dots+k_n S^{n-1}α)=k_1 S^{n-1}α=0
Sn−1(k1α+k2Sα+⋯+knSn−1α)=k1Sn−1α=0, thus
k
1
=
0
k_1=0
k1=0, the above becomes
k
2
S
α
+
⋯
+
k
n
S
n
−
1
α
=
0
k_2 Sα+\dots+k_n S^{n-1}α=0
k2Sα+⋯+knSn−1α=0
then
S
n
−
2
(
k
2
S
α
+
⋯
+
k
n
S
n
−
1
α
)
=
k
2
S
n
−
1
α
=
0
S^{n-2}(k_2 Sα+\dots+k_n S^{n-1}α)=k_2 S^{n-1}α=0
Sn−2(k2Sα+⋯+knSn−1α)=k2Sn−1α=0, thus
k
2
=
0
k_2=0
k2=0, continue this step we eventually have
k
1
=
⋯
=
k
n
=
0
k_1=\dots=k_n=0
k1=⋯=kn=0. Thus we can define
B
′
=
{
α
,
S
α
,
…
,
S
n
−
1
α
}
\mathfrak B'=\{α,Sα,\dots,S^{n-1} α\}
B′={α,Sα,…,Sn−1α}, and
[
S
]
B
′
=
A
[S]_{\mathfrak B'}=A
[S]B′=A.
( d ) Let
T
T
T and
S
S
S be linear operators which satisfies
[
T
]
B
=
M
,
[
S
]
B
=
N
[T]_{\mathfrak B}=M,[S]_{\mathfrak B}=N
[T]B=M,[S]B=N, in which
B
=
{
ϵ
1
,
…
,
ϵ
n
}
{\mathfrak B}=\{ϵ_1,\dots,ϵ_n \}
B={ϵ1,…,ϵn}. From ( c ) we can find two ordered basis
B
1
{\mathfrak B}_1
B1 and
B
2
{\mathfrak B}_2
B2 s.t.
[
T
]
B
1
=
[
S
]
B
2
=
A
[T]_{\mathfrak B_1}=[S]_{\mathfrak B_2}=A
[T]B1=[S]B2=A, by Theorem 14, let
P
P
P be the
n
×
n
n\times n
n×n matrix with columns
P
j
=
[
ϵ
j
]
B
1
P_j=[ϵ_j ]_{\mathfrak B_1}
Pj=[ϵj]B1, and
Q
Q
Q be the
n
×
n
n\times n
n×n matrix with columns
Q
j
=
[
ϵ
j
]
B
2
Q_j=[ϵ_j ]_{\mathfrak B_2}
Qj=[ϵj]B2, then
M
=
[
T
]
B
=
P
−
1
A
P
,
N
=
[
S
]
B
=
Q
−
1
A
Q
M=[T]_{\mathfrak B}=P^{-1} AP,\quad N=[S]_{\mathfrak B}=Q^{-1}AQ
M=[T]B=P−1AP,N=[S]B=Q−1AQ
thus
M
=
P
−
1
Q
N
Q
−
1
P
=
(
Q
−
1
P
)
−
1
N
(
Q
−
1
P
)
M=P^{-1} QNQ^{-1}P=(Q^{-1}P)^{-1}N(Q^{-1}P)
M=P−1QNQ−1P=(Q−1P)−1N(Q−1P).
13.Let
V
V
V and
W
W
W be finite-dimensional vector spaces over the field
F
F
F and let
T
T
T be a linear transformation from
V
V
V into
W
W
W. If
B
=
{
α
1
,
…
,
α
n
}
and
B
′
=
{
β
1
,
…
,
β
m
}
\mathfrak B=\{\alpha_1,\dots,\alpha_n\}\text{ and }\mathfrak B'=\{\beta_1,\dots,\beta_m\}
B={α1,…,αn} and B′={β1,…,βm}
are ordered base for
V
V
V and
W
W
W, respectively, define the linear transformations
E
p
,
q
E{p,q}
Ep,q as in the proof of Theorem 5:
E
p
,
q
(
α
i
)
=
δ
i
q
β
p
E^{p,q}(\alpha_i)=\delta_{iq}\beta_p
Ep,q(αi)=δiqβp. Then the
E
p
,
q
,
1
≤
p
≤
m
,
1
≤
q
≤
n
E{p,q},1\leq p\leq m,1\leq q\leq n
Ep,q,1≤p≤m,1≤q≤n, form a basis for
L
(
V
,
W
)
L(V,W)
L(V,W), and so
T
=
∑
p
=
1
m
∑
q
=
1
n
A
p
q
E
p
,
q
T=\sum_{p=1}^m\sum_{q=1}^nA_{pq}E^{p,q}
T=p=1∑mq=1∑nApqEp,q
for certain scalars
A
p
q
A_{pq}
Apq (the coordinates of
T
T
T in this basis for
L
(
V
,
W
)
L(V,W)
L(V,W)). Show that the matrix
A
A
A with entries
A
(
p
,
q
)
=
A
p
q
A(p,q)=A_{pq}
A(p,q)=Apq is precisely the matrix of
T
T
T relative to the pair
B
,
B
′
\mathfrak B,\mathfrak B'
B,B′.
Solution:
T
α
j
=
(
∑
p
=
1
m
∑
q
=
1
n
A
p
q
E
p
,
q
)
(
α
j
)
=
∑
p
=
1
m
∑
q
=
1
n
A
p
q
E
p
,
q
(
α
j
)
=
∑
p
=
1
m
∑
q
=
1
n
A
p
q
δ
j
q
β
p
=
∑
p
=
1
m
A
p
j
β
p
Tα_j=(∑_{p=1}^m∑_{q=1}^nA_{pq} E^{p,q})(α_j )=∑_{p=1}^m∑_{q=1}^nA_{pq} E^{p,q}(α_j)=∑_{p=1}^m∑_{q=1}^nA_{pq} δ_{jq} β_p=∑_{p=1}^mA_{pj} β_p
Tαj=(∑p=1m∑q=1nApqEp,q)(αj)=∑p=1m∑q=1nApqEp,q(αj)=∑p=1m∑q=1nApqδjqβp=∑p=1mApjβp, and the conclusion follows.