这一节继续复习线性代数知识,回顾了初等矩阵、求逆、行列式、代数余子式、Cramer法则等相关内容,总体上也易于理解。
Exercises
Exercise 1. Consider the matrix
A
=
[
1
2
1
−
1
0
1
]
.
A=\begin{bmatrix}1&2\\1&-1\\0&1\end{bmatrix}.
A=⎣⎡1102−11⎦⎤.
( a ) Find two different left inverses for
A
A
A.
( b ) Show that
A
A
A has no right inverse.
Solution:
(a) The two left inverses for
A
A
A may be
B
1
=
[
1
0
−
2
0
0
1
]
,
B
2
=
[
0
1
1
0
0
1
]
B_1=\begin{bmatrix}1&0&-2\\0&0&1\end{bmatrix},\quad B_2=\begin{bmatrix}0&1&1\\0&0&1\end{bmatrix}
B1=[1000−21],B2=[001011]
(b) Assume
A
A
A has a right inverse
B
=
[
a
b
c
d
e
f
]
,
A
B
=
I
3
B=\begin{bmatrix}a&b&c\\d&e&f\end{bmatrix},\quad AB=I_3
B=[adbecf],AB=I3
then
A
B
=
[
1
2
1
−
1
0
1
]
[
a
b
c
d
e
f
]
=
[
a
+
2
d
b
+
2
e
c
+
2
f
a
−
d
b
−
e
e
−
f
d
e
f
]
=
[
1
0
0
0
1
0
0
0
1
]
AB=\begin{bmatrix}1&2\\1&-1\\0&1\end{bmatrix}\begin{bmatrix}a&b&c\\d&e&f\end{bmatrix}=\begin{bmatrix}a+2d&b+2e&c+2f\\a-d&b-e&e-f\\d&e&f\end{bmatrix}=\begin{bmatrix}1&0&0\\0&1&0\\0&0&1\end{bmatrix}
AB=⎣⎡1102−11⎦⎤[adbecf]=⎣⎡a+2da−ddb+2eb−eec+2fe−ff⎦⎤=⎣⎡100010001⎦⎤
from the bottom row we have
f
=
1
,
d
=
e
=
0
f=1,d=e=0
f=1,d=e=0, so
e
−
f
=
−
1
e-f=-1
e−f=−1, but from the second row we know
e
−
f
=
0
e-f=0
e−f=0, a contradiction.
Exercise 2. Let
A
A
A be an
n
n
n by
m
m
m matrix with
n
≠
m
n\neq m
n=m.
( a ) If
rank
A
=
m
\text{rank }A=m
rank A=m, show there exists a matrix
D
D
D that is a product of elementary matrices such that
D
⋅
A
=
[
I
m
0
]
.
D\cdot A=\begin{bmatrix}I_m\\0\end{bmatrix}.
D⋅A=[Im0].
( b ) Show that
A
A
A has a left inverse if and only if
rank
A
=
m
\text{rank }A=m
rank A=m.
( c ) Show that
A
A
A has a right inverse if and only if
rank
A
=
n
\text{rank }A=n
rank A=n.
Solution:
( a ) If
rank
A
=
m
\text{rank }A=m
rank A=m, then
A
A
A has at least
m
m
m rows, but
n
≠
m
n\neq m
n=m, so
n
>
m
n>m
n>m. Since we can use elementary row operations to reduce
A
A
A to
[
I
m
0
]
\begin{bmatrix}I_m\\0\end{bmatrix}
[Im0],
D
D
D is the product of the corresponding elementary matrices.
( b ) If
rank
A
=
m
\text{rank }A=m
rank A=m, then by (a) we can find
D
D
D such that
D
⋅
A
=
[
I
m
0
]
D\cdot A= \begin{bmatrix}I_m\\0\end{bmatrix}
D⋅A=[Im0], let the
m
×
n
m\times n
m×n matrix
E
E
E be
E
=
[
I
m
0
]
E=\begin{bmatrix}I_m&0\end{bmatrix}
E=[Im0], then
E
D
A
=
[
I
m
0
]
[
I
m
0
]
=
I
m
EDA=\begin{bmatrix}I_m&0\end{bmatrix}\begin{bmatrix}I_m\\0\end{bmatrix}=I_m
EDA=[Im0][Im0]=Im
thus
E
D
ED
ED is a left inverse of
A
A
A. Conversely, if
A
A
A has a left inverse, then there is
B
B
B s.t.
B
A
=
I
m
BA=I_m
BA=Im, from the proof of Theorem 2.3 we know
rank
A
=
m
\text{rank }A=m
rank A=m.
( c ) From step 3 of the proof of Theorem 2.3 we know if
A
A
A has a right inverse, then
rank
A
=
n
\text{rank }A=n
rank A=n. Conversely if
rank
A
=
n
\text{rank }A=n
rank A=n, then
rank
A
t
r
=
n
\text{rank }A^{tr}=n
rank Atr=n, by (b),
A
t
r
A^{tr}
Atr has a left inverse
B
B
B, so
B
A
t
r
=
I
n
BA^{tr}=I_n
BAtr=In, thus
(
B
A
t
r
)
t
r
=
A
B
t
r
=
I
n
(BA^{tr} )^{tr}=AB^{tr}=I_n
(BAtr)tr=ABtr=In.
Exercise 3. Verify that the functions defined in Example 1 satisfy the axioms for the determinant function.
Solution: For 1 by 1 matrices, we have
det
[
a
]
=
a
\det[a]=a
det[a]=a, the verification is
(1) there is no way to exchange rows.
(2) we have
det
[
a
]
=
a
=
a
⋅
1
=
a
det
[
1
]
\det[a]=a=a\cdot 1=a \det[1]
det[a]=a=a⋅1=adet[1].
(3)
det
[
1
]
=
1
\det[1]=1
det[1]=1.
For 2 by 2 matrices, we have
det
[
a
b
c
d
]
=
a
d
−
b
c
\det\begin{bmatrix}a&b\\c&d\end{bmatrix}=ad-bc
det[acbd]=ad−bc, the verification is
(1)
det
[
c
d
a
b
]
=
c
b
−
a
d
=
−
(
a
d
−
b
c
)
=
−
det
[
a
b
c
d
]
\det\begin{bmatrix}c&d\\a&b\end{bmatrix}=cb-ad=-(ad-bc)=-\det\begin{bmatrix}a&b\\c&d\end{bmatrix}
det[cadb]=cb−ad=−(ad−bc)=−det[acbd]
(2) Let
x
=
(
x
1
,
x
2
)
,
y
=
(
y
1
,
y
2
)
\mathrm x=(x_1,x_2 ),\mathrm{y}=(y_1,y_2 )
x=(x1,x2),y=(y1,y2), then
A
1
(
x
)
=
[
x
1
x
2
c
d
]
,
A
2
(
x
)
=
[
a
b
x
1
x
2
]
,
A
1
(
y
)
=
[
y
1
y
2
c
d
]
,
A
2
(
y
)
=
[
a
b
y
1
y
2
]
A_1 (x)=\begin{bmatrix}x_1&x_2\\c&d\end{bmatrix},A_2 (x)=\begin{bmatrix}a&b\\x_1&x_2 \end{bmatrix},A_1 (y)=\begin{bmatrix}y_1&y_2\\c&d\end{bmatrix},A_2 (y)=\begin{bmatrix}a&b\\y_1&y_2 \end{bmatrix}
A1(x)=[x1cx2d],A2(x)=[ax1bx2],A1(y)=[y1cy2d],A2(y)=[ay1by2]
so
det
A
1
(
m
x
+
n
y
)
=
det
[
m
x
1
+
n
y
1
m
x
2
+
n
y
2
c
d
]
=
(
m
x
1
+
n
y
1
)
d
−
(
m
x
2
+
n
y
2
)
c
=
m
(
x
1
d
−
x
2
c
)
+
n
(
y
1
d
−
y
2
c
)
=
m
det
A
1
(
x
)
+
n
det
A
1
(
y
)
\begin{aligned}\det A_1 (m\mathrm{x}+n\mathrm{y})&=\det\begin{bmatrix}mx_1+ny_1&mx_2+ny_2\\c&d\end{bmatrix}\\&=(mx_1+ny_1 )d-(mx_2+ny_2 )c\\&=m(x_1 d-x_2 c)+n(y_1 d-y_2 c)\\&=m \detA_1 (x)+n \det A_1 (y)\end{aligned}
detA1(mx+ny)=det[mx1+ny1cmx2+ny2d]=(mx1+ny1)d−(mx2+ny2)c=m(x1d−x2c)+n(y1d−y2c)=mdetA1(x)+ndetA1(y)
Similarly
det
A
2
(
m
x
+
n
y
)
=
m
det
A
2
(
x
)
+
n
det
A
2
(
y
)
\det A_2 (mx+ny)=m \det A_2 (x)+n \det A_2 (y)
detA2(mx+ny)=mdetA2(x)+ndetA2(y).
(3)
det
[
1
0
0
1
]
=
1
⋅
1
−
0
⋅
0
=
1
\det\begin{bmatrix}1&0\\0&1\end{bmatrix}=1\cdot 1-0\cdot 0=1
det[1001]=1⋅1−0⋅0=1
The verification for 3 by 3 matrices is of the same logic and omitted.
Exercise 4. ( a ) Let
A
A
A be an
n
n
n by
n
n
n matrix of rank
n
n
n. By applying elementary row operations to
A
A
A, one can reduce
A
A
A to the identity matrix. Show that by applying the same operations, in the same order, to
I
n
I_n
In, one obtains the matrix
A
−
1
A^{-1}
A−1.
( b ) Let
A
=
[
1
2
3
0
1
2
1
2
1
]
.
A=\begin{bmatrix}1&2&3\\0&1&2\\1&2&1\end{bmatrix}.
A=⎣⎡101212321⎦⎤.
Calculate
A
−
1
A^{-1}
A−1 by using the algorithm suggested in (a).
( c ) Calculate
A
−
1
A^{-1}
A−1 using the formula involving determinants.
Solution:
( a ) We suppose, by Theorem 2.1, that there are elementary matrices
E
1
,
…
,
E
k
E_1,\dots,E_k
E1,…,Ek such that
E
k
⋯
E
1
A
=
I
n
E_k{\cdots}E_1 A=I_n
Ek⋯E1A=In, let
B
=
E
k
⋯
E
1
B=E_k{\cdots}E_1
B=Ek⋯E1, then
B
B
B is a left inverse for
A
A
A, by Theorem 2.5,
B
B
B is also a right inverse for
A
A
A, thus
B
=
A
−
1
B=A^{-1}
B=A−1, further we have
E
k
…
E
1
I
n
=
B
I
n
=
B
=
A
−
1
E_k{\dots}E_1 I_n=BI_n=B=A^{-1}
Ek…E1In=BIn=B=A−1.
( b ) We have
[
A
I
3
]
=
[
1
2
3
1
0
0
0
1
2
0
1
0
1
2
1
0
0
1
]
→
[
1
2
3
1
0
0
0
1
2
0
1
0
0
0
−
2
−
1
0
1
]
→
[
1
2
0
−
1
/
2
0
3
/
2
0
1
0
−
1
1
1
0
0
1
1
/
2
0
−
1
/
2
]
→
[
1
0
0
3
/
2
−
2
−
1
/
2
0
1
0
−
1
1
1
0
0
1
1
/
2
0
−
1
/
2
]
\begin{aligned}\begin{bmatrix}A&I_3 \end{bmatrix}&=\begin{bmatrix}1&2&3&1&0&0\\0&1&2&0&1&0\\1&2&1&0&0&1\end{bmatrix}{\rightarrow}\begin{bmatrix}1&2&3&1&0&0\\0&1&2&0&1&0\\0&0&-2&-1&0&1\end{bmatrix}\\&{\rightarrow}\begin{bmatrix}1&2&0&-1/2&0&3/2\\0&1&0&-1&1&1\\0&0&1&1/2&0&-1/2\end{bmatrix}{\rightarrow}\begin{bmatrix}1&0&0&3/2&-2&-1/2\\0&1&0&-1&1&1\\0&0&1&1/2&0&-1/2\end{bmatrix}\end{aligned}
[AI3]=⎣⎡101212321100010001⎦⎤→⎣⎡10021032−210−1010001⎦⎤→⎣⎡100210001−1/2−11/20103/21−1/2⎦⎤→⎣⎡1000100013/2−11/2−210−1/21−1/2⎦⎤
thus
A
−
1
=
[
3
/
2
−
2
−
1
/
2
−
1
1
1
1
/
2
0
−
1
/
2
]
A^{-1}=\begin{bmatrix}3/2&-2&-1/2\\-1&1&1\\1/2&0&-1/2\end{bmatrix}
A−1=⎣⎡3/2−11/2−210−1/21−1/2⎦⎤.
( c ) The calculation is based on Theorem 2.14, and is omitted.
Exercise 5. Let
A
=
[
a
b
c
d
]
A=\begin{bmatrix}a&b\\c&d\end{bmatrix}
A=[acbd], where
a
d
−
b
c
≠
0
ad-bc\neq 0
ad−bc=0. Find
A
−
1
A^{-1}
A−1.
Solution: By Theorem 2.14,
b
11
=
(
−
1
)
1
+
1
d
a
d
−
b
c
,
b
12
=
(
−
1
)
1
+
2
b
a
d
−
b
c
,
b
21
=
(
−
1
)
2
+
1
c
a
d
−
b
c
,
b
22
=
(
−
1
)
2
+
2
a
a
d
−
b
c
b_{11}=\frac{(-1)^{1+1} d}{ad-bc},b_{12}=\frac{(-1)^{1+2}b}{ad-bc},b_{21}=\frac{(-1)^{2+1}c}{ad-bc},b_{22}=\frac{(-1)^{2+2}a}{ad-bc}
b11=ad−bc(−1)1+1d,b12=ad−bc(−1)1+2b,b21=ad−bc(−1)2+1c,b22=ad−bc(−1)2+2a
thus
A
−
1
=
1
a
d
−
b
c
[
d
−
b
−
c
a
]
A^{-1}=\frac{1}{ad-bc}\begin{bmatrix}d&-b\\-c&a\end{bmatrix}
A−1=ad−bc1[d−c−ba]
Exercise 6. Prove the following theorem: Let
A
A
A be a
k
k
k by
k
k
k matrix, let
D
D
D have size
n
n
n by
n
n
n and let
C
C
C have size
n
n
n by
k
k
k. Then
det
[
A
0
C
D
]
=
(
det
A
)
⋅
(
det
D
)
.
\det\begin{bmatrix}A&0\\C&D\end{bmatrix}=(\det A)\cdot(\det D).
det[AC0D]=(detA)⋅(detD).
Solution: We have
[
A
0
0
I
n
]
⋅
[
I
k
0
C
D
]
=
[
A
0
C
D
]
\begin{bmatrix}A&0\\0&I_n \end{bmatrix}⋅\begin{bmatrix}I_k&0\\C&D\end{bmatrix}=\begin{bmatrix}A&0\\C&D\end{bmatrix}
[A00In]⋅[IkC0D]=[AC0D]
and by Lemma 2.12 we have
det
[
A
0
0
I
n
]
=
det
A
,
det
[
I
k
0
C
D
]
=
det
D
\det\begin{bmatrix}A&0\\0&I_n \end{bmatrix}=\det A,\det\begin{bmatrix}I_k&0\\C&D\end{bmatrix}=\det D
det[A00In]=detA,det[IkC0D]=detD
so by Theorem 2.10 we have
det
[
A
0
C
D
]
=
(
det
A
)
⋅
(
det
D
)
\det\begin{bmatrix}A&0\\C&D\end{bmatrix}=(\det A)\cdot(\det D)
det[AC0D]=(detA)⋅(detD)