矩阵的Kronecker积的相关结论
矩阵的Kronecker积
矩阵的Kronecker积的定义
设矩阵
A
=
(
a
i
j
)
∈
M
m
,
n
,
B
∈
M
s
,
t
.
A=(a_{ij})\in M_{m,n},B\in M_{s,t}.
A=(aij)∈Mm,n,B∈Ms,t.
A
A
A和
B
B
B的
K
r
o
n
e
c
k
e
r
Kronecker
Kronecker积(也称为张量积)记作
A
⊗
B
A\otimes B
A⊗B,定义为下面的分块矩阵:
A
⊗
B
=
(
a
11
B
a
12
B
⋯
a
1
n
B
a
21
B
a
22
B
⋯
a
2
n
B
⋮
⋮
⋮
a
m
1
B
a
m
2
B
⋯
a
m
n
B
)
∈
M
m
s
,
n
t
.
A\otimes B=\begin{pmatrix} a_{11}B & a_{12}B & \cdots & a_{1n}B\\ a_{21}B & a_{22}B & \cdots & a_{2n}B\\ \vdots & \vdots & & \vdots\\ a_{m1}B & a_{m2}B & \cdots & a_{mn}B \end{pmatrix} \in M_{ms,nt}.
A⊗B=⎝⎜⎜⎜⎛a11Ba21B⋮am1Ba12Ba22B⋮am2B⋯⋯⋯a1nBa2nB⋮amnB⎠⎟⎟⎟⎞∈Mms,nt.
矩阵的Kronecker积的性质
A ∈ M m , n , B ∈ M s , t , C ∈ M p , q , D ∈ M t , r A\in M_{m,n},B\in M_{s,t},C\in M_{p,q},D\in M_{t,r} A∈Mm,n,B∈Ms,t,C∈Mp,q,D∈Mt,r
- ( α A ) ⊗ B = A ⊗ ( α B ) = α ( A ⊗ B ) , α ∈ C (\alpha A)\otimes B=A\otimes (\alpha B)=\alpha(A\otimes B),\alpha\in \mathbf{C} (αA)⊗B=A⊗(αB)=α(A⊗B),α∈C
- ( A ⊗ B ) T = A T ⊗ B T (A \otimes B)^T=A^T\otimes B^T (A⊗B)T=AT⊗BT
- ( A ⊗ B ) H = A H ⊗ B H (A \otimes B)^H=A^H\otimes B^H (A⊗B)H=AH⊗BH
- ( A ⊗ B ) ⊗ C = A ⊗ ( B ⊗ C ) (A \otimes B)\otimes C=A\otimes (B\otimes C) (A⊗B)⊗C=A⊗(B⊗C)
- A ⊗ ( B + C ) = ( A ⊗ B ) + ( A ⊗ C ) A\otimes (B+C)=(A\otimes B)+(A\otimes C) A⊗(B+C)=(A⊗B)+(A⊗C)
- ( A + B ) ⊗ C = ( A ⊗ C ) + ( B ⊗ C ) (A+B)\otimes C=(A\otimes C)+(B\otimes C) (A+B)⊗C=(A⊗C)+(B⊗C)
- A ⊗ B = 0 A \otimes B=0 A⊗B=0当且仅当 A = 0 A=0 A=0或 B = 0 B=0 B=0
- ( A ⊗ B ) ( C ⊗ D ) = ( A C ) ⊗ ( B D ) (A\otimes B)(C\otimes D)=(AC)\otimes(BD) (A⊗B)(C⊗D)=(AC)⊗(BD)
设 A ∈ M m , B ∈ M n A\in M_{m},B\in M_{n} A∈Mm,B∈Mn
- 若 A , B A,B A,B对称,则 A ⊗ B A \otimes B A⊗B对称
- 若 A , B A,B A,B为 H e r m i t Hermit Hermit矩阵,则 A ⊗ B A\otimes B A⊗B是 H e r m i t Hermit Hermit矩阵
- 若 A , B A,B A,B可逆,则 A ⊗ B A\otimes B A⊗B也可逆,且 ( A ⊗ B ) − 1 = A − 1 ⊗ B − 1 (A\otimes B)^{-1}=A^{-1}\otimes B^{-1} (A⊗B)−1=A−1⊗B−1
- 若 A , B A,B A,B为正规矩阵,则 A ⊗ B A\otimes B A⊗B是正规矩阵
- 若 A , B A,B A,B为酉矩阵,则 A ⊗ B A\otimes B A⊗B是酉矩阵
- 若 λ ∈ σ ( A ) \lambda\in\sigma(A) λ∈σ(A), x x x是对应的特征向量, μ ∈ σ ( B ) \mu\in\sigma(B) μ∈σ(B), y y y是对应的特征向量,则 λ μ ∈ σ ( A ⊗ B ) \lambda\mu\in\sigma(A\otimes B) λμ∈σ(A⊗B), x ⊗ y x\otimes y x⊗y是对应的特征向量
- 若
σ
(
A
)
=
{
λ
1
,
⋯
,
λ
m
}
,
σ
(
B
)
=
{
μ
1
,
⋯
,
μ
n
}
\sigma(A)= \{ \lambda_1,\cdots,\lambda_{m}\},\sigma(B)= \{ \mu_1,\cdots,\mu_{n}\}
σ(A)={λ1,⋯,λm},σ(B)={μ1,⋯,μn},则
σ ( A ⊗ B ) = { λ i μ j ∣ i = 1 , ⋯ , m , j = 1 , ⋯ , n } \sigma(A\otimes B)=\{\lambda_{i}\mu_{j} | i=1,\cdots,m,j=1,\cdots,n\} σ(A⊗B)={λiμj∣i=1,⋯,m,j=1,⋯,n} - d e t ( A ⊗ B ) = ( d e t A ) n ( d e t B ) m det(A\otimes B)=(detA)^n(detB)^m det(A⊗B)=(detA)n(detB)m
- 若
s
v
(
A
)
=
{
s
1
,
⋯
,
s
m
}
,
s
v
(
B
)
=
{
t
1
,
⋯
,
t
n
}
sv(A)= \{ s_1,\cdots,s_{m}\},sv(B)= \{ t_1,\cdots,t_{n}\}
sv(A)={s1,⋯,sm},sv(B)={t1,⋯,tn},则
s v ( A ⊗ B ) = { s i t j ∣ i = 1 , ⋯ , m , j = 1 , ⋯ , n } sv(A\otimes B)=\{s_{i}t_{j} | i=1,\cdots,m,j=1,\cdots,n\} sv(A⊗B)={sitj∣i=1,⋯,m,j=1,⋯,n} - r a n k ( A ⊗ B ) = ( r a n k A ) ( r a n k B ) rank(A\otimes B)=(rankA)(rankB) rank(A⊗B)=(rankA)(rankB)
矩阵的拉直
矩阵的拉直的定义
设
A
=
(
a
1
,
a
2
,
⋯
,
a
n
)
∈
M
m
,
n
A=(a_1,a_2,\cdots,a_n)\in M_{m,n}
A=(a1,a2,⋯,an)∈Mm,n,则
v
e
c
(
A
)
=
(
a
1
a
2
⋮
a
n
)
.
vec(A)=\begin{pmatrix} a_{1}\\ a_{2}\\ \vdots \\ a_{n} \end{pmatrix}.
vec(A)=⎝⎜⎜⎜⎛a1a2⋮an⎠⎟⎟⎟⎞.
矩阵的拉直的性质
1、设
A
∈
M
m
,
n
,
B
∈
M
n
,
k
,
C
∈
M
k
,
t
,
A\in M_{m,n},B\in M_{n,k},C\in M_{k,t},
A∈Mm,n,B∈Mn,k,C∈Mk,t,则
v
e
c
(
A
B
C
)
=
(
C
T
⊗
A
)
v
e
c
B
.
vec(ABC)=(C^T\otimes A)vecB.
vec(ABC)=(CT⊗A)vecB.
2、存在一个只依赖于
m
,
n
m,n
m,n的
m
n
mn
mn阶置换矩阵
P
(
m
,
n
)
P(m,n)
P(m,n)使得
v
e
c
X
T
=
P
(
m
,
n
)
v
e
c
X
,
vecX^T=P(m,n)vecX,
vecXT=P(m,n)vecX,对任何
X
∈
M
m
,
n
X\in M_{m,n}
X∈Mm,n成立.
矩阵的方程
定理1: 矩阵方程 A X − X B = C ( 称 为 S y l v e s t e r 方 程 ) , A ∈ M m , B ∈ M n , C ∈ M m , n AX-XB=C(称为Sylvester方程), A\in M_m,B\in M_n,C\in M_{m,n} AX−XB=C(称为Sylvester方程),A∈Mm,B∈Mn,C∈Mm,n有唯一解当且仅当 A A A和 B B B没有公共特征值
定理2: 矩阵方程
A
X
−
X
B
=
C
(
称
为
S
y
l
v
e
s
t
e
r
方
程
)
,
A
∈
M
m
,
B
∈
M
n
,
C
∈
M
m
,
n
AX-XB=C(称为Sylvester方程), A\in M_m,B\in M_n,C\in M_{m,n}
AX−XB=C(称为Sylvester方程),A∈Mm,B∈Mn,C∈Mm,n有解当且仅当
(
A
0
0
B
)
\begin{pmatrix} A&0\\ 0&B \end{pmatrix}
(A00B)和
(
A
C
0
B
)
\begin{pmatrix} A&C\\ 0&B \end{pmatrix}
(A0CB)相似
例1、 设
A
∈
C
m
×
m
,
B
∈
C
n
×
n
,
X
(
t
)
∈
C
m
×
n
A\in \mathbf{C}^{m\times m},B\in \mathbf{C}^{n\times n},X(t)\in \mathbf{C}^{m\times n}
A∈Cm×m,B∈Cn×n,X(t)∈Cm×n,求下列微分方程初值问题的解:
{
d
X
(
t
)
d
t
=
A
X
(
t
)
+
X
(
t
)
B
X
(
0
)
=
X
0
\begin{cases}\dfrac{dX(t)}{dt}=AX(t)+X(t)B\\ X(0)=X_0 \end{cases}
⎩⎨⎧dtdX(t)=AX(t)+X(t)BX(0)=X0
引理: 设矩阵
A
∈
C
m
×
m
,
B
∈
C
n
×
n
A\in \mathbf{C}^{m\times m},B\in\mathbf{C}^{n\times n}
A∈Cm×m,B∈Cn×n,则
e
A
×
I
n
=
e
A
×
I
n
e^{A\times I_n}=e^A\times I_n
eA×In=eA×In,
e
I
m
×
B
=
I
m
×
B
e^{I_m\times B}=I_m\times B
eIm×B=Im×B.
p
r
o
o
f
:
proof:
proof:
e
A
×
I
n
=
∑
k
=
1
∞
1
k
!
(
A
⊗
I
)
k
=
∑
k
=
1
∞
1
k
!
(
A
k
⊗
I
k
)
=
(
∑
k
=
1
∞
1
k
!
A
k
)
⊗
I
=
e
A
×
I
n
e^{A\times I_n}=\sum\limits_{k=1}^{\infty}\frac{1}{k!}(A\otimes I)^k=\sum\limits_{k=1}^{\infty}\frac{1}{k!}(A^k\otimes I^k)=(\sum\limits_{k=1}^{\infty}\frac{1}{k!}A^k)\otimes I=e^A\times I_n
eA×In=k=1∑∞k!1(A⊗I)k=k=1∑∞k!1(Ak⊗Ik)=(k=1∑∞k!1Ak)⊗I=eA×In
同理可得:
e
I
m
×
B
=
I
m
×
B
e^{I_m\times B}=I_m\times B
eIm×B=Im×B
解、 对微分方程两边拉直,易得:
{
d
v
e
c
X
(
t
)
d
t
=
(
I
n
⊗
A
+
B
T
⊗
I
m
)
v
e
c
X
(
t
)
v
e
c
X
(
0
)
=
v
e
c
X
0
\begin{cases}\dfrac{dvecX(t)}{dt}=(I_n\otimes A+B^T\otimes I_m)vecX(t)\\ vecX(0)=vecX_0 \end{cases}
⎩⎨⎧dtdvecX(t)=(In⊗A+BT⊗Im)vecX(t)vecX(0)=vecX0
由引理可得:
v
e
c
X
(
t
)
=
e
t
(
I
n
⊗
A
+
B
T
⊗
I
m
)
v
e
c
X
0
=
(
e
B
T
t
⊗
e
A
t
)
v
e
c
X
0
=
v
e
c
(
e
A
t
X
0
(
e
B
T
t
)
T
)
=
v
e
c
(
e
A
t
X
0
e
B
t
)
vecX(t)=e^{t(I_n\otimes A+B^T\otimes I_m)}vecX_0=(e^{B^Tt}\otimes e^{At})vecX_0=vec(e^{At}X_0(e^{B^Tt})^T)=vec(e^{At}X_0e^{B^t})
vecX(t)=et(In⊗A+BT⊗Im)vecX0=(eBTt⊗eAt)vecX0=vec(eAtX0(eBTt)T)=vec(eAtX0eBt)
于是
X
(
t
)
=
e
A
t
X
0
e
B
t
X(t)=e^{At}X_0e^{B^t}
X(t)=eAtX0eBt为微分方程的解