目录
1. 矩阵序列
1.1 敛散性
定义: 将矩阵序列 A ( k ) = [ a i j ( k ) ] m × n A^{(k)}=\left[a_{i j}^{(k)}\right]_{m \times n} A(k)=[aij(k)]m×n, 记作 { A ( k ) } \left\{A^{(k)}\right\} {A(k)} 当 lim k → ∞ a i j ( k ) = a i j ( ∀ i , j ) \lim _{k \rightarrow \infty} a_{i j}^{(k)}=a_{i j}(\forall i, j) limk→∞aij(k)=aij(∀i,j) 时, 称矩阵序列 { A ( k ) } \left\{A^{(k)}\right\} {A(k)} 收敛于矩阵 A = ( a i j ) \boldsymbol{A}=\left(a_{i j}\right) A=(aij) 。记作
lim k → ∞ A ( k ) = A , 或者 A ( k ) → A ( k → ∞ ) \lim _{k \rightarrow \infty} A^{(k)}=A \text {, 或者 } A^{(k)} \rightarrow A(k \rightarrow \infty) k→∞limA(k)=A, 或者 A(k)→A(k→∞)若数列 ( a i j ( k ) ) \left(a_{i j}^{(k)}\right) (aij(k)) 之一发散, 称 { A ( k ) } \left\{A^{(k)}\right\} {A(k)} 发散
性质:(和、乘积、逆的收敛性一致):
(1) 若 lim k → ∞ A ( k ) = A m × n , lim k → ∞ B ( k ) = B m × n \lim _{k \rightarrow \infty} A^{(k)}=A_{m \times n}, \lim _{k \rightarrow \infty} B^{(k)}=B_{m \times n} limk→∞A(k)=Am×n,limk→∞B(k)=Bm×n 则
lim k → ∞ ( a A ( k ) + b B ( k ) ) = a A + b B , ∀ a , b \lim _{k \rightarrow \infty}\left(a A^{(k)}+b B^{(k)}\right)=a A+b B, \quad \forall a, b k→∞lim(aA(k)+bB(k))=aA+bB,∀a,b(2) 若 lim k → ∞ A ( k ) = A m × n , lim k → ∞ B ( k ) = B n × l \lim _{k \rightarrow \infty} A^{(k)}=A_{m \times n}, \lim _{k \rightarrow \infty} B^{(k)}=B_{n \times l} limk→∞A(k)=Am×n,limk→∞B(k)=Bn×l 则
lim k → ∞ ( A ( k ) B ( k ) ) = A B \lim _{k \rightarrow \infty}\left(A^{(k)} B^{(k)}\right)=A B k→∞lim(A(k)B(k))=AB(3) 若 A ( k ) A^{(k)} A(k) 与 A A A 是可逆矩阵, 且 lim k → ∞ A ( k ) = A \lim _{k \rightarrow \infty} A^{(k)}=A limk→∞A(k)=A, 则
lim k → ∞ ( A ( k ) ) − 1 = A − 1 \lim _{k \rightarrow \infty}\left(A^{(k)}\right)^{-1}=A^{-1} k→∞lim(A(k))−1=A−1
1.2 定理1(阐述矩阵序列收敛和矩阵范数之间的关系)
定理1: 设
A
(
k
)
,
A
∈
C
m
×
n
A^{(k)}, A \in C^{m \times n}
A(k),A∈Cm×n, 则
(1)
lim
k
→
∞
A
(
k
)
=
0
⟺
∀
∥
∙
∥
,
lim
k
→
∞
∥
A
(
k
)
∥
=
0
\lim _{k \rightarrow \infty} A^{(k)}=0 \iff \forall\|\bullet\|, \lim _{k \rightarrow \infty}\left\|A^{(k)}\right\|=0
limk→∞A(k)=0⟺∀∥∙∥,limk→∞∥∥A(k)∥∥=0
(2)
lim
k
→
∞
A
(
k
)
=
A
⟺
∀
∥
∙
∥
,
lim
k
→
∞
∥
A
(
k
)
−
A
∥
=
0
\lim _{k \rightarrow \infty} A^{(k)}=A \iff \forall\|\bullet\|, \lim _{k \rightarrow \infty}\left\|A^{(k)}-A\right\|=0
limk→∞A(k)=A⟺∀∥∙∥,limk→∞∥∥A(k)−A∥∥=0
1.3 定理2(阐述收敛矩阵和谱半径之间的关系)
谱半径:对
∀
A
∈
C
n
×
n
\forall A \in C^{n \times n}
∀A∈Cn×n, 谱半径为
ρ
(
A
)
=
max
i
∣
λ
i
∣
\rho(A)=\max _{i}\left|\lambda_{i}\right|
ρ(A)=maxi∣λi∣
定义: 若
A
n
×
n
A_{n \times n}
An×n 满足
lim
k
→
∞
A
k
=
0
n
×
n
\lim _{k \rightarrow \infty} A^{k}=0_{n \times n}
limk→∞Ak=0n×n, 称
A
A
A 为收玫矩阵
定理 2:
A
A
A 为收敛矩阵
⟺
ρ
(
A
)
<
1
\iff \rho(A)<1
⟺ρ(A)<1
1.4 定理3(阐述收敛矩阵和矩阵范数之间的关系)
定理 3: 若矩阵范数
∥
⋅
∥
M
\|\cdot\|_{M}
∥⋅∥M 使
∥
A
∥
M
<
1
\|A\|_{M}<1
∥A∥M<1, 则
A
k
→
0
\boldsymbol{A}^{k} \rightarrow \mathbf{0}
Ak→0
证明:
ρ
(
A
)
≤
∥
A
∥
M
<
1
⇒
A
k
→
0
\quad \rho(A) \leq\|A\|_{M}<1 \Rightarrow A^{k} \rightarrow 0
ρ(A)≤∥A∥M<1⇒Ak→0
2. 矩阵级数
定义: 设有矩阵序列 { A ( k ) } \left\{A^{(k)}\right\} {A(k)}, 其中 A ( k ) = [ a i j ( k ) ] ∈ C m × n A^{(k)}=\left[a_{i j}^{(k)}\right] \in C^{m \times n} A(k)=[aij(k)]∈Cm×n 称 A ( 0 ) + A ( 1 ) + ⋯ + A ( k ) + ⋯ A^{(0)}+A^{(1)}+\cdots+A^{(k)}+\cdots A(0)+A(1)+⋯+A(k)+⋯ 为矩阵级数。记为 ∑ k = 0 ∞ A ( k ) \sum_{k=0}^{\infty} A^{(k)} ∑k=0∞A(k) 部分和 S ( N ) = ∑ k = 0 N A ( k ) S^{(N)}=\sum_{k=0}^{N} A^{(k)} S(N)=∑k=0NA(k) 构成矩阵序列 { S ( N ) } \left\{S^{(N)}\right\} {S(N)}
2.1 敛散性
定义:
若 lim N → ∞ S ( N ) = S \lim _{N \rightarrow \infty} S^{(N)}=S limN→∞S(N)=S, 称 ∑ A ( k ) \sum A^{(k)} ∑A(k) 收敛于 S S S, 记做 ∑ A ( k ) = S \sum A^{(k)}=S ∑A(k)=S
若 { S ( N ) } \left\{S^{(N)}\right\} {S(N)} 发散, 称 ∑ A ( k ) \sum A^{(k)} ∑A(k) 发散
性质:
- ∑ A ( k ) = S ⟺ ∑ k = 0 ∞ a i j ( k ) = s i j ( all i , j ) \sum A^{(k)}=S \iff \sum_{k=0}^{\infty} a_{i j}^{(k)}=s_{i j}\left(\begin{array}{ll}\text { all } & i, j)\end{array}\right. ∑A(k)=S⟺∑k=0∞aij(k)=sij( all i,j)
- 若 ∑ ∣ a i j ( k ) ∣ \sum\left|a_{i j}^{(k)}\right| ∑∣∣∣aij(k)∣∣∣ 收敛 ( a l l i , j ) (a l l \quad i, j) (alli,j), 称 ∑ A ( k ) \sum A^{(k)} ∑A(k) 绝对收敛。
(1) ∑ A ( k ) \sum A^{(k)} ∑A(k) 绝对收敛 ⇒ ∑ A ( k ) \Rightarrow \sum A^{(k)} ⇒∑A(k) 收敛
(2) 若 ∑ A ( k ) \sum A^{(k)} ∑A(k) 绝对收敛于 S S S, 对 ∑ A ( k ) \sum A^{(k)} ∑A(k) 任意重组重排得 ∑ B ( k ) \sum B^{(k)} ∑B(k), 则 ∑ B ( k ) \sum B^{(k)} ∑B(k) 绝对收敛于 S S S 。
- ∑ A ( k ) = S \sum A^{(k)}=S ∑A(k)=S 绝对收敛 ⇔ ∀ ∥ ⋅ ∥ , ∑ ∥ A ( k ) ∥ \Leftrightarrow \forall\|\cdot\|, \sum\left\|A^{(k)}\right\| ⇔∀∥⋅∥,∑∥∥A(k)∥∥ 收敛
- ∑ A ( k ) \sum A^{(k)} ∑A(k) 收敛于 S ⇒ ∑ k = 0 ∞ P A ( k ) Q S \Rightarrow \sum_{k=0}^{\infty} P A^{(k)} Q S⇒∑k=0∞PA(k)Q 收敛于 P S Q P S Q PSQ
∑ A ( k ) \sum A^{(k)} ∑A(k) 绝对收敛 ⇒ ∑ k = 0 ∞ P A ( k ) Q \Rightarrow \sum_{k=0}^{\infty} P A^{(k)} Q ⇒∑k=0∞PA(k)Q 绝对收敛- 若 ∑ k = 1 ∞ A ( k ) \sum_{k=1}^{\infty} A^{(k)} ∑k=1∞A(k) 绝对收敛于 S m x x , ∑ k = 1 ∞ B ( k ) S_{m x x}, \sum_{k=1}^{\infty} B^{(k)} Smxx,∑k=1∞B(k) 绝对收玫于 T n x l T_{n x l} Tnxl
则Cauchy积 A ( 1 ) B ( 1 ) + [ A ( 1 ) B ( 2 ) + A ( 2 ) B ( 1 ) ] + [ A ( 1 ) B ( 3 ) + A ( 2 ) B ( 2 ) + A ( 3 ) B ( 1 ) ] A^{(1)} B^{(1)}+\left[A^{(1)} B^{(2)}+A^{(2)} B^{(1)}\right]+\left[A^{(1)} B^{(3)}+A^{(2)} B^{(2)}+A^{(3)} B^{(1)}\right] A(1)B(1)+[A(1)B(2)+A(2)B(1)]+[A(1)B(3)+A(2)B(2)+A(3)B(1)] + ⋯ + [ A ( 1 ) B ( k ) + A ( 2 ) B ( k − 1 ) + ⋯ + A ( k ) B ( 1 ) ] + ⋯ +\cdots+\left[A^{(1)} B^{(k)}+A^{(2)} B^{(k-1)}+\cdots+A^{(k)} B^{(1)}\right]+\cdots +⋯+[A(1)B(k)+A(2)B(k−1)+⋯+A(k)B(1)]+⋯
绝对收敛于 S T S T ST, 记作 ∑ A ( k ) ⋅ ∑ B ( k ) = S T \sum A^{(k)} \cdot \sum B^{(k)}=S T ∑A(k)⋅∑B(k)=ST
2.2 Neumann级数
定义: A n × n , ∑ k = 0 ∞ A k , ( A 0 = I ) A_{n \times n}, \sum_{k=0}^{\infty} A^{k},\left(A^{0}=I\right) An×n,∑k=0∞Ak,(A0=I)
性质: A n × n A_{n \times n} An×n
- ∑ A k \sum A^{k} ∑Ak收敛 ⇔ A k → 0 ∑ A k \Leftrightarrow A^{k} \rightarrow 0 \\ \sum A^{k} ⇔Ak→0∑Ak收敛时, 其和为 ( I − A ) − 1 (I-A)^{-1} (I−A)−1
- ∥ A ∥ < 1 ⇒ ∥ ( I − A ) − 1 − ∑ k = 0 N A k ∥ ≤ ∥ A ∥ N + 1 1 − ∥ A ∥ , N = 0 , 1 , 2 \|A\|<1 \Rightarrow\left\|(I-A)^{-1}-\sum_{k=0}^{N} A^{k}\right\| \leq \frac{\|A\|^{N+1}}{1-\|A\|}, N=0,1,2 ∥A∥<1⇒∥∥∥(I−A)−1−∑k=0NAk∥∥∥≤1−∥A∥∥A∥N+1,N=0,1,2
2.3 幂级数
幂级数: 对函数
f
(
z
)
=
∑
c
k
z
k
,
(
∣
z
∣
<
r
)
f(z)=\sum c_{k} z^{k},(|z|<r)
f(z)=∑ckzk,(∣z∣<r) 方阵
A
n
×
n
A_{n \times n}
An×n, 构造矩阵幕级数
f
(
A
)
=
∑
c
k
A
k
f(A)=\sum c_{k} A^{k}
f(A)=∑ckAk
谱半径:对
∀
A
∈
C
n
×
n
\forall A \in C^{n \times n}
∀A∈Cn×n, 谱半径为
ρ
(
A
)
=
max
i
∣
λ
i
∣
\rho(A)=\max_{i}\left|\lambda_{i}\right|
ρ(A)=maxi∣λi∣
定理 6 :
(1)
ρ
(
A
)
<
r
⇒
∑
c
k
A
k
\rho(A)<r \Rightarrow \sum c_{k} A^{k}
ρ(A)<r⇒∑ckAk 绝对收敛
(2)
ρ
(
A
)
>
r
⇒
∑
c
k
A
k
\rho(A)>r \Rightarrow \sum c_{k} A^{k}
ρ(A)>r⇒∑ckAk 发散
2.4 代入规则
性质(代入规则): 若 f ( z ) = g ( z ) f(z)=g(z) f(z)=g(z), 则 f ( A ) = g ( A ) f(A)=g(A) f(A)=g(A).
设一元函数 f ( z ) f(z) f(z) 能展开为 z z z 的幕级数
f ( z ) = ∑ k = 0 ∞ c k z k ( ∣ z ∣ < r , r > 0 ) f(z)=\sum_{k=0}^{\infty} c_{k} z^{k} \quad(|z|<r, r>0) f(z)=k=0∑∞ckzk(∣z∣<r,r>0)其中 r > 0 r>0 r>0 表示该幂级数的收敛半径。当 n \mathrm{n} n 阶矩阵 A \mathrm{A} A 的 谱半径 ρ ( A ) < r \rho(A)<r ρ(A)<r 时, 把收玫的矩阵幂级数 ∑ k = 0 ∞ c k A k \sum_{k=0}^{\infty} c_{k} A^{k} ∑k=0∞ckAk 的和 为 f ( A ) f(\mathrm{~A}) f( A), 即 f ( A ) = ∑ k = 0 ∞ c k A k f(A)=\sum_{k=0}^{\infty} c_{k} A^{k} f(A)=∑k=0∞ckAk
2.5 级数展开式
e
z
=
1
+
1
1
!
z
+
⋯
+
1
k
!
z
k
+
⋯
(
r
=
+
∞
)
e^{z}=1+\frac{1}{1 !} z+\cdots+\frac{1}{k !} z^{k}+\cdots \quad(r=+\infty)
ez=1+1!1z+⋯+k!1zk+⋯(r=+∞)
e
A
=
I
+
1
1
!
A
+
⋯
+
1
k
!
A
k
+
⋯
(
∀
A
n
×
n
)
e^{A}=I+\frac{1}{1 !} A+\cdots+\frac{1}{k !} A^{k}+\cdots \quad\left(\forall A_{n \times n}\right)
eA=I+1!1A+⋯+k!1Ak+⋯(∀An×n)
sin
z
=
z
−
1
3
!
z
3
+
⋯
+
(
−
1
)
k
1
(
2
k
+
1
)
!
z
(
2
k
+
1
)
+
⋯
(
r
=
+
∞
)
\sin z=z-\frac{1}{3 !} z^{3}+\cdots+(-1)^{k} \frac{1}{(2 k+1) !} z^{(2 k+1)}+\cdots \quad(r=+\infty)
sinz=z−3!1z3+⋯+(−1)k(2k+1)!1z(2k+1)+⋯(r=+∞)
sin
A
=
A
−
1
3
!
A
3
+
⋯
+
(
−
1
)
k
1
(
2
k
+
1
)
!
A
(
2
k
+
1
)
+
⋯
(
∀
A
n
×
n
)
\sin A=A-\frac{1}{3 !} A^{3}+\cdots+(-1)^{k} \frac{1}{(2 k+1) !} A^{(2 k+1)}+\cdots \quad\left(\forall A_{n \times n}\right)
sinA=A−3!1A3+⋯+(−1)k(2k+1)!1A(2k+1)+⋯(∀An×n)
1
1
−
z
=
∑
k
=
0
∞
z
k
(
∣
z
∣
<
1
)
\frac{1}{1-z}=\sum_{k=0}^{\infty} z^{k} \quad(|z|<1)
1−z1=∑k=0∞zk(∣z∣<1)
1
I
−
A
=
∑
k
=
0
∞
A
k
(
ρ
(
A
)
<
1
)
\frac{1}{I-A}=\sum_{k=0}^{\infty} A^{k} \quad(\rho(A)<1)
I−A1=∑k=0∞Ak(ρ(A)<1)
2.6 定理7
一般来说,
e
A
+
B
≠
e
A
e
B
≠
e
B
e
A
e^{A+B} \neq e^{A} e^{B} \neq e^{B} e^{A}
eA+B=eAeB=eBeA
但
A
n
×
n
,
B
n
×
n
,
A
B
=
B
A
⇒
e
A
+
B
=
e
A
e
B
=
e
B
e
A
A_{n \times n}, B_{n \times n}, A B=B A \Rightarrow e^{A+B}=e^{A} e^{B}=e^{B} e^{A}
An×n,Bn×n,AB=BA⇒eA+B=eAeB=eBeA
3. 矩阵函数(以矩阵为自变量且取值为矩阵的一类函数)
3.1 矩阵函数值的求法——待定系数法
例6: A = [ 2 0 0 1 1 1 1 − 1 3 ] , A=\left[\begin{array}{ccc}2 & 0 & 0 \\ 1 & 1 & 1 \\ 1 & -1 & 3\end{array}\right], \quad A=⎣⎡21101−1013⎦⎤, 求 e A , e t A ( t ∈ R ) e^{A}, e^{t A} \quad(t \in R) eA,etA(t∈R)
第一步:
φ ( λ ) = det ( λ I − A ) = ( λ − 2 ) 3 \varphi(\lambda)=\operatorname{det}(\lambda I-A)=(\lambda-2)^{3} φ(λ)=det(λI−A)=(λ−2)3
( A − 2 I ) = [ 0 0 0 1 − 1 1 1 − 1 1 ] , ( A − 2 I ) 2 = O (A-2 I)=\left[\begin{array}{ccc}0 & 0 & 0 \\ 1 & -1 & 1 \\ 1 & -1 & 1\end{array}\right],(A-2 I)^{2}=O (A−2I)=⎣⎡0110−1−1011⎦⎤,(A−2I)2=O
取 ψ ( λ ) = m ( λ ) = ( λ − 2 ) 2 , ψ ( λ ) 为 首 一 多 项 式 且 可 被 det ( λ I − A ) 整 除 且 是 J o r d a n 中 的 最 高 次 \psi(\lambda)=m(\lambda)=(\lambda-2)^{2},\psi(\lambda)为首一多项式且可被\operatorname{det}(\lambda I-A)整除且是Jordan中的最高次 ψ(λ)=m(λ)=(λ−2)2,ψ(λ)为首一多项式且可被det(λI−A)整除且是Jordan中的最高次
第二步:
(1) f ( λ ) = e λ = ψ ( λ ) g ( λ ) + ( a + b λ ) , 其 中 f ( λ i ) = ∑ c k λ i k 绝 对 收 敛 f(\lambda)=e^{\lambda}=\psi(\lambda) g(\lambda)+(a+b \lambda),其中f\left(\lambda_{i}\right)=\sum c_{k} \lambda_{i}^{k}绝对收敛 f(λ)=eλ=ψ(λ)g(λ)+(a+bλ),其中f(λi)=∑ckλik绝对收敛
f ′ ( λ ) = e λ = [ ψ ( λ ) g ( λ ) ] ′ + b f^{\prime}(\lambda)=e^{\lambda}=[\psi(\lambda) g(\lambda)]^{\prime}+b f′(λ)=eλ=[ψ(λ)g(λ)]′+b
f ( 2 ) = e 2 : ( a + 2 b ) = e 2 f ′ ( 2 ) = e 2 : b = e 2 } { a = − e 2 b = e 2 \left.\begin{array}{rl}f(2)=e^{2}: & (a+2 b)=e^{2} \\ f^{\prime}(2)=e^{2}: & b=e^{2}\end{array}\right\} \quad\left\{\begin{array}{l}a=-e^{2} \\ b=e^{2}\end{array}\right. f(2)=e2:f′(2)=e2:(a+2b)=e2b=e2}{a=−e2b=e2
e A = e 2 ( A − I ) = e 2 [ 1 0 0 1 0 1 1 − 1 2 ] e^{A}=e^{2}(A-I)=e^{2}\left[\begin{array}{ccc}1 & 0 & 0 \\ 1 & 0 & 1 \\ 1 & -1 & 2\end{array}\right] eA=e2(A−I)=e2⎣⎡11100−1012⎦⎤
3.2 数项级数求和法
例7: A = [ π 0 0 0 − π 0 0 0 1 0 ] A=\left[\begin{array}{cccc}\pi & 0 & 0 & 0 \\ & -\pi & 0 & 0 \\ & & 0 & 1 \\ & & & 0\end{array}\right] A=⎣⎢⎢⎡π0−π0000010⎦⎥⎥⎤, 求 sin A \sin A sinA
解: φ ( λ ) = ∣ λ I − A ∣ = λ 4 − π 2 λ 2 \varphi(\lambda)=|\lambda I-A|=\lambda^{4}-\pi^{2} \lambda^{2} φ(λ)=∣λI−A∣=λ4−π2λ2, 取 ψ ( λ ) = φ ( λ ) \psi(\lambda)=\varphi(\lambda) ψ(λ)=φ(λ) ψ ( A ) = 0 ⇒ A 4 = π 2 A 2 , A 5 = π 2 A 3 , A 7 = π 4 A 3 , ⋯ \psi(A)=0 \Rightarrow A^{4}=\pi^{2} A^{2}, A^{5}=\pi^{2} A^{3}, A^{7}=\pi^{4} A^{3}, \cdots ψ(A)=0⇒A4=π2A2,A5=π2A3,A7=π4A3,⋯ sin A = A − 1 3 ! A 3 + 1 5 ! A 5 − 1 7 ! A 7 + ⋯ \sin A=A-\frac{1}{3 !} A^{3}+\frac{1}{5 !} A^{5}-\frac{1}{7 !} A^{7}+\cdots sinA=A−3!1A3+5!1A5−7!1A7+⋯ = A + [ − 1 3 ! + π 2 5 ! − π 4 7 ! + ⋯ ] A 3 = A + 1 π 3 [ sin π − π ] A 3 = A − 1 π 2 A 3 = [ 0 0 0 0 0 0 0 0 1 0 ] ∵ A 3 = diag ( π 3 , − π 3 , 0 , 0 ) \begin{aligned}=& A+\left[-\frac{1}{3 !}+\frac{\pi^{2}}{5 !}-\frac{\pi^{4}}{7 !}+\cdots\right] A^{3} \\=& A+\frac{1}{\pi^{3}}[\sin \pi-\pi] A^{3}=A-\frac{1}{\pi^{2}} A^{3}=\left[\begin{array}{llll}0 & 0 & 0 & 0 \\ & 0 & 0 & 0 \\ & & 0 & 1 \\ & & & 0\end{array}\right] \\ & \because A^{3}=\operatorname{diag}\left(\pi^{3},-\pi^{3}, 0,0\right) \end{aligned} ==A+[−3!1+5!π2−7!π4+⋯]A3A+π31[sinπ−π]A3=A−π21A3=⎣⎢⎢⎡0000000010⎦⎥⎥⎤∵A3=diag(π3,−π3,0,0)
3.3 对角阵法
设
P
−
1
A
P
=
diag
(
λ
1
,
⋯
,
λ
n
)
=
Λ
P^{-1} A P=\operatorname{diag}\left(\lambda_{1}, \cdots, \lambda_{n}\right)=\Lambda
P−1AP=diag(λ1,⋯,λn)=Λ, 则
A
k
=
P
Λ
k
P
−
1
A^{k}=P \Lambda^{k} P^{-1}
Ak=PΛkP−1
f
(
A
)
=
∑
k
=
0
∞
c
k
A
k
=
P
⋅
diag
(
f
(
λ
1
)
,
⋯
,
f
(
λ
n
)
)
⋅
P
−
1
f(A)=\sum_{k=0}^{\infty} c_{k} A^{k}=P \cdot \operatorname{diag}\left(f\left(\lambda_{1}\right), \cdots, f\left(\lambda_{n}\right)\right) \cdot P^{-1}
f(A)=∑k=0∞ckAk=P⋅diag(f(λ1),⋯,f(λn))⋅P−1
3.4 Jordan标准形法
4. 矩阵函数利用Jordan标准形进行定义
设
P
−
1
A
P
=
J
=
diag
(
J
1
,
⋯
,
J
s
)
,
J
i
=
λ
i
I
+
I
(
1
)
P^{-1} A P=J=\operatorname{diag}\left(J_{1}, \cdots, J_{s}\right), J_{i}=\lambda_{i} I+I^{(1)}
P−1AP=J=diag(J1,⋯,Js),Ji=λiI+I(1)
如果
f
(
z
)
f(z)
f(z) 在
λ
i
\lambda_{i}
λi 处有
m
i
−
1
m_{i}-1
mi−1 阶导数, 令
f
(
J
i
)
=
∑
k
=
0
∞
c
k
J
i
k
=
f
(
λ
i
)
I
+
f
′
(
λ
i
)
1
!
I
(
1
)
+
⋯
+
f
(
m
i
−
1
)
(
λ
i
)
(
m
i
−
1
)
!
I
(
m
i
−
1
)
f\left(J_{i}\right)=\sum_{k=0}^{\infty} c_{k} J_{i}^{k}=f\left(\lambda_{i}\right) I+\frac{f^{\prime}\left(\lambda_{i}\right)}{1 !} I^{(1)}+\cdots+\frac{f^{\left(m_{i}-1\right)}\left(\lambda_{i}\right)}{\left(m_{i}-1\right) !} I^{\left(m_{i}-1\right)}
f(Ji)=∑k=0∞ckJik=f(λi)I+1!f′(λi)I(1)+⋯+(mi−1)!f(mi−1)(λi)I(mi−1)
f
(
A
)
=
∑
k
=
0
∞
c
k
A
k
=
P
⋅
∑
k
=
0
∞
c
k
J
k
⋅
P
−
1
=
P
⋅
diag
(
f
(
J
1
)
,
⋯
,
f
(
J
s
)
)
⋅
P
−
1
f(A)=\sum_{k=0}^{\infty} c_{k} A^{k}=P \cdot \sum_{k=0}^{\infty} c_{k} J^{k} \cdot P^{-1}=P \cdot \operatorname{diag}\left(f\left(J_{1}\right), \cdots, f\left(J_{s}\right)\right) \cdot P^{-1}
f(A)=∑k=0∞ckAk=P⋅∑k=0∞ckJk⋅P−1=P⋅diag(f(J1),⋯,f(Js))⋅P−1
称
f
(
A
)
f(A)
f(A) 为对应于
f
(
z
)
f(z)
f(z) 的矩阵函数
例题:
A
=
[
2
1
2
1
2
1
2
]
,
f
(
z
)
=
1
z
A=\left[\begin{array}{llll}2 & 1 & & \\ & 2 & 1 & \\ & & 2 & 1 \\ & & & 2\end{array}\right], f(z)=\frac{1}{z}
A=⎣⎢⎢⎡2121212⎦⎥⎥⎤,f(z)=z1, 求
f
(
A
)
f(A)
f(A)
A
=
[
1
1
0
0
1
0
0
0
2
]
,
f
(
z
)
=
z
A=\left[\begin{array}{lll}1 & 1 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 2\end{array}\right], f(z)=\sqrt{z}
A=⎣⎡100110002⎦⎤,f(z)=z, 求
f
(
A
)
f(A)
f(A)
步骤:
先求矩阵的Jordan型,需要jordan矩阵和P(
P
−
1
A
P
=
J
=
diag
(
J
1
,
⋯
,
J
s
)
P^{-1} A P=J=\operatorname{diag}\left(J_{1}, \cdots, J_{s}\right)
P−1AP=J=diag(J1,⋯,Js))
再套公式求
f
(
J
i
)
=
f
(
λ
i
)
I
+
f
′
(
λ
i
)
1
!
I
(
1
)
+
⋯
+
f
(
m
i
−
1
)
(
λ
i
)
(
m
i
−
1
)
!
I
(
m
i
−
1
)
f\left(J_{i}\right)=f\left(\lambda_{i}\right) I+\frac{f^{\prime}\left(\lambda_{i}\right)}{1 !} I^{(1)}+\cdots+\frac{f^{\left(m_{i}-1\right)}\left(\lambda_{i}\right)}{\left(m_{i}-1\right) !} I^{\left(m_{i}-1\right)}
f(Ji)=f(λi)I+1!f′(λi)I(1)+⋯+(mi−1)!f(mi−1)(λi)I(mi−1)
最后
f
(
A
)
=
P
⋅
diag
(
f
(
J
1
)
,
⋯
,
f
(
J
s
)
)
⋅
P
−
1
f(A)=P \cdot \operatorname{diag}\left(f\left(J_{1}\right), \cdots, f\left(J_{s}\right)\right) \cdot P^{-1}
f(A)=P⋅diag(f(J1),⋯,f(Js))⋅P−1
5. 矩阵函数的性质
- f ( z ) = f 1 ( z ) + f 2 ( z ) ⇒ f ( A ) = f 1 ( A ) + f 2 ( A ) f(z)=f_{1}(z)+f_{2}(z) \Rightarrow f(A)=f_{1}(A)+f_{2}(A) f(z)=f1(z)+f2(z)⇒f(A)=f1(A)+f2(A)
- f ( z ) = f 1 ( z ) ∙ f 2 ( z ) ⇒ f ( A ) = f 1 ( A ) ∙ f 2 ( A ) = f 2 ( A ) ⋅ f 1 ( A ) f(z)=f_{1}(z) \bullet f_{2}(z)\Rightarrow f(A)=f_{1}(A) \bullet f_{2}(A)=f_{2}(A) \cdot f_{1}(A) f(z)=f1(z)∙f2(z)⇒f(A)=f1(A)∙f2(A)=f2(A)⋅f1(A)
6. 矩阵的微分和积分(对自变量)
微分定义:
如果矩阵
A
(
t
)
=
(
a
i
j
(
t
)
)
m
×
n
A(t)=\left(a_{i j}(t)\right)_{m \times n}
A(t)=(aij(t))m×n, 的每一个元素
a
i
j
(
t
)
a_{i j}(t)
aij(t) 是变量
t
t
t 的可微函数, 则
A
(
t
)
A(t)
A(t) 关于
t
t
t 的导数 (微商) 定义为
d
A
(
t
)
d
t
=
(
a
i
j
′
(
t
)
)
m
×
n
\frac{d A(t)}{d t}=\left(a_{i j}^{\prime}(t)\right)_{m \times n}
dtdA(t)=(aij′(t))m×n, 或者
A
′
(
t
)
=
(
a
i
j
′
(
t
)
)
m
×
n
A^{\prime}(t)=\left(a_{i j}^{\prime}(t)\right)_{m \times n}
A′(t)=(aij′(t))m×n
微分性质:
定理8: 设
A
(
t
)
,
B
(
t
)
\boldsymbol{A}(\boldsymbol{t}), \boldsymbol{B}(\boldsymbol{t})
A(t),B(t) 可导, 则有
(1)
d
d
t
[
A
(
t
)
+
B
(
t
)
]
=
d
d
t
A
(
t
)
+
d
d
t
B
(
t
)
\frac{d}{d t}[A(t)+B(t)]=\frac{d}{d t} A(t)+\frac{d}{d t} B(t)
dtd[A(t)+B(t)]=dtdA(t)+dtdB(t)
(2)
A
m
×
n
,
f
(
t
)
A_{m \times n}, f(t)
Am×n,f(t) 可导
d
d
t
[
f
(
t
)
A
(
t
)
]
=
f
′
(
t
)
A
(
t
)
+
f
(
t
)
A
′
(
t
)
\frac{d}{d t}[f(t) A(t)]=f^{\prime}(t) A(t)+f(t) A^{\prime}(t)
dtd[f(t)A(t)]=f′(t)A(t)+f(t)A′(t)
(3)
A
m
×
n
,
B
n
×
l
:
d
d
t
[
A
(
t
)
B
(
t
)
]
=
A
′
(
t
)
B
(
t
)
+
A
(
t
)
B
′
(
t
)
A_{m \times n}, B_{n \times l}: \frac{d}{d t}[A(t) B(t)]=A^{\prime}(t) B(t)+A(t) B^{\prime}(t)
Am×n,Bn×l:dtd[A(t)B(t)]=A′(t)B(t)+A(t)B′(t)
定理9: 设
A
n
×
n
A_{n \times n}
An×n 为常数矩阵, 则有
(1)
d
d
t
e
t
A
=
A
e
t
A
=
e
t
A
A
\frac{d}{d t} e^{t A}=A e^{t A}=e^{t A} A
dtdetA=AetA=etAA
(2)
d
d
t
cos
(
t
A
)
=
−
A
⋅
sin
(
t
A
)
=
−
sin
(
t
A
)
⋅
A
\frac{d}{d t} \cos (t A)=-A \cdot \sin (t A)=-\sin (t A) \cdot A
dtdcos(tA)=−A⋅sin(tA)=−sin(tA)⋅A
(3)
d
d
t
sin
(
t
A
)
=
A
⋅
cos
(
t
A
)
=
cos
(
t
A
)
⋅
A
\frac{d}{d t} \sin (t A)=A \cdot \cos (t A)=\cos (t A) \cdot A
dtdsin(tA)=A⋅cos(tA)=cos(tA)⋅A
积分定义:
如果矩阵
A
(
t
)
=
(
a
i
j
(
t
)
)
m
×
n
A(t)=\left(a_{i j}(t)\right)_{m \times n}
A(t)=(aij(t))m×n 的每一个元素
a
i
j
(
t
)
a_{i j}(t)
aij(t) 在
[
t
0
,
t
]
\left[t_{0}, t\right]
[t0,t] 上可积, 称
A
(
t
)
A(t)
A(t) 可积, 记为
∫
t
0
t
A
(
τ
)
d
τ
=
(
∫
t
0
t
a
i
j
(
τ
)
d
τ
)
m
×
n
\int_{t_{0}}^{t} A(\tau) d \tau=\left(\int_{t_{0}}^{t} a_{i j}(\tau) d \tau\right)_{m \times n}
∫t0tA(τ)dτ=(∫t0taij(τ)dτ)m×n
积分性质:
(1)
∫
t
0
t
[
A
(
τ
)
+
B
(
τ
)
]
d
τ
=
∫
t
0
t
A
(
τ
)
d
τ
+
∫
t
0
t
B
(
τ
)
d
τ
\int_{t_{0}}^{t}[A(\tau)+B(\tau)] d \tau=\int_{t_{0}}^{t} A(\tau) d \tau+\int_{t_{0}}^{t} B(\tau) d \tau
∫t0t[A(τ)+B(τ)]dτ=∫t0tA(τ)dτ+∫t0tB(τ)dτ
(2)
A
A
A 为常数矩阵:
∫
t
0
t
[
A
∙
B
(
τ
)
]
d
τ
=
A
⋅
[
∫
t
0
t
B
(
τ
)
d
τ
]
\int_{t_{0}}^{t}[A \bullet B(\tau)] d \tau=A \cdot\left[\int_{t_{0}}^{t} B(\tau) d \tau\right]
∫t0t[A∙B(τ)]dτ=A⋅[∫t0tB(τ)dτ]
B
B
B 为常数矩阵:
∫
t
0
t
[
A
(
τ
)
⋅
B
]
d
τ
=
[
∫
t
0
t
A
(
τ
)
d
τ
]
⋅
B
\int_{t_{0}}^{t}[A(\tau) \cdot B] d \tau=\left[\int_{t_{0}}^{t} A(\tau) d \tau\right] \cdot B
∫t0t[A(τ)⋅B]dτ=[∫t0tA(τ)dτ]⋅B
(3) 设
a
i
j
(
t
)
∈
C
[
t
0
,
t
1
]
a_{i j}(t) \in C\left[t_{0}, t_{1}\right]
aij(t)∈C[t0,t1] 则:
d
d
t
∫
t
0
t
A
(
τ
)
d
τ
=
A
(
t
)
\frac{d}{d t} \int_{t_{0}}^{t} A(\tau) d \tau=A(t)
dtd∫t0tA(τ)dτ=A(t)
(4) 设
a
i
j
′
(
t
)
∈
C
[
t
0
,
t
1
]
a_{i j}^{\prime}(t) \in C\left[t_{0}, t_{1}\right]
aij′(t)∈C[t0,t1], 则:
∫
t
0
t
1
A
′
(
τ
)
d
τ
=
A
(
t
1
)
−
A
(
t
0
)
\int_{t_{0}}^{t_{1}} A^{\prime}(\tau) d \tau=A\left(t_{1}\right)-A\left(t_{0}\right)
∫t0t1A′(τ)dτ=A(t1)−A(t0)
7. 函数对矩阵的导数(包括向量)
定义: 设
X
=
(
ξ
i
j
)
m
×
n
,
m
n
X=\left(\xi_{i j}\right)_{m \times n}, m n
X=(ξij)m×n,mn 元函数
f
(
X
)
=
f
(
ξ
11
,
ξ
12
,
⋯
,
ξ
1
n
,
⋯
,
ξ
m
×
n
)
f(X)=f\left(\xi_{11}, \xi_{12}, \cdots, \xi_{1 n}, \cdots, \xi_{m \times n}\right)
f(X)=f(ξ11,ξ12,⋯,ξ1n,⋯,ξm×n)
定义
f
(
X
)
f(X)
f(X) 对矩阵
X
X
X 的导数为
d
f
d
X
=
(
∂
f
∂
ξ
i
j
)
m
×
n
=
[
∂
f
∂
ξ
11
⋯
∂
f
∂
ξ
1
n
⋮
⋮
∂
f
∂
ξ
m
1
⋯
∂
f
∂
ξ
m
n
]
\frac{d f}{d X}=\left(\frac{\partial f}{\partial \xi_{i j}}\right)_{m \times n}=\left[\begin{array}{ccc} \frac{\partial f}{\partial \xi_{11}} & \cdots & \frac{\partial f}{\partial \xi_{1 n}} \\ \vdots & & \vdots \\ \frac{\partial f}{\partial \xi_{m 1}} & \cdots & \frac{\partial f}{\partial \xi_{m n}} \end{array}\right]
dXdf=(∂ξij∂f)m×n=⎣⎢⎢⎡∂ξ11∂f⋮∂ξm1∂f⋯⋯∂ξ1n∂f⋮∂ξmn∂f⎦⎥⎥⎤
8. 函数矩阵对矩阵的导数
定义:
设
X
=
(
ξ
i
j
)
m
×
n
,
f
k
l
(
X
)
=
f
k
l
(
ξ
11
,
ξ
12
,
⋯
,
ξ
1
n
,
⋯
,
ξ
m
×
n
)
X=\left(\xi_{i j}\right)_{m \times n}, f_{k l}(X)=f_{k l}\left(\xi_{11}, \xi_{12}, \cdots, \xi_{1 n}, \cdots, \xi_{m \times n}\right)
X=(ξij)m×n,fkl(X)=fkl(ξ11,ξ12,⋯,ξ1n,⋯,ξm×n)
F
=
[
f
11
⋯
f
1
s
⋮
⋮
f
r
1
⋯
f
r
s
]
,
∂
F
∂
ξ
i
j
=
[
∂
f
11
∂
ξ
i
j
⋯
∂
f
1
s
∂
ξ
i
j
⋮
⋮
∂
f
r
1
∂
ξ
i
j
⋯
∂
f
r
s
∂
ξ
i
j
]
\boldsymbol{F}=\left[\begin{array}{ccc}f_{11} & \cdots & f_{1 s} \\ \vdots & & \vdots \\ f_{r 1} & \cdots & f_{r s}\end{array}\right], \quad \frac{\partial F}{\partial \xi_{i j}}=\left[\begin{array}{ccc}\frac{\partial f_{11}}{\partial \xi_{i j}} & \cdots & \frac{\partial f_{1 s}}{\partial \xi_{i j}} \\ \vdots & & \vdots \\ \frac{\partial f_{r 1}}{\partial \xi_{i j}} & \cdots & \frac{\partial f_{r s}}{\partial \xi_{i j}}\end{array}\right]
F=⎣⎢⎡f11⋮fr1⋯⋯f1s⋮frs⎦⎥⎤,∂ξij∂F=⎣⎢⎢⎡∂ξij∂f11⋮∂ξij∂fr1⋯⋯∂ξij∂f1s⋮∂ξij∂frs⎦⎥⎥⎤
则:
d
F
d
X
=
(
1
d
X
)
⊗
d
F
=
[
∂
F
∂
ξ
11
⋯
∂
F
∂
ξ
1
n
⋮
⋮
∂
F
∂
ξ
m
1
⋯
∂
F
∂
ξ
m
n
]
\frac{d F}{d X}=\left(\frac{1}{d X}\right) \otimes d F=\left[\begin{array}{ccc} \frac{\partial F}{\partial \xi_{11}} & \cdots & \frac{\partial F}{\partial \xi_{1 n}} \\ \vdots & & \vdots \\ \frac{\partial F}{\partial \xi_{m 1}} & \cdots & \frac{\partial F}{\partial \xi_{m n}} \end{array}\right]
dXdF=(dX1)⊗dF=⎣⎢⎡∂ξ11∂F⋮∂ξm1∂F⋯⋯∂ξ1n∂F⋮∂ξmn∂F⎦⎥⎤
9. 转换定理
设
X
m
×
n
=
[
x
i
j
]
,
Y
p
×
q
=
[
y
i
j
]
X_{m \times n}=\left[x_{i j}\right], Y_{p \times q}=\left[y_{i j}\right]
Xm×n=[xij],Yp×q=[yij] 矩阵
A
p
×
m
,
B
n
×
q
,
C
p
×
n
,
D
m
×
q
A_{p \times m}, B_{n \times q}, C_{p \times n}, D_{m \times q}
Ap×m,Bn×q,Cp×n,Dm×q 是
X
X
X 的函数, 则下列两条等价
(1)
∂
Y
∂
x
i
j
=
A
E
i
j
B
+
C
E
i
j
T
D
,
i
=
1
,
…
,
m
;
j
=
1
,
…
,
n
;
E
i
j
∈
R
m
×
n
\frac{\partial Y}{\partial x_{i j}}=A E_{i j} B+C E_{i j}^{T} D, \quad i=1, \ldots, m ; j=1, \ldots, n ; E_{i j} \in R^{m \times n}
∂xij∂Y=AEijB+CEijTD,i=1,…,m;j=1,…,n;Eij∈Rm×n
(2)
d
y
i
j
d
X
=
A
T
E
i
j
B
T
+
D
E
i
j
T
C
,
i
=
1
,
…
,
p
;
j
=
1
,
…
,
q
;
E
i
j
∈
R
p
×
q
\frac{d y_{i j}}{d X}=A^{T} E_{i j} B^{T}+D E_{i j}^{T} C, \quad i=1, \ldots, p ; j=1, \ldots, q ; E_{i j} \in R^{p \times q}
dXdyij=ATEijBT+DEijTC,i=1,…,p;j=1,…,q;Eij∈Rp×q
10. 矩阵分析的应用
例21:
设
A
=
[
1
1
0
1
0
2
]
A=\left[\begin{array}{lll}1 & 1 & 0 \\ & 1 & 0 \\ & & 2\end{array}\right]
A=⎣⎡111002⎦⎤ 求
x
′
(
t
)
=
A
∙
x
(
t
)
x^{\prime}(t)=A \bullet x(t)
x′(t)=A∙x(t) 的通解
解:
e
t
A
=
[
e
t
t
e
t
0
e
t
0
e
2
t
]
e^{t A}=\left[\begin{array}{ccc}e^{t} & t e^{t} & 0 \\ & e^{t} & 0 \\ & & e^{2 t}\end{array}\right]
etA=⎣⎡ettetet00e2t⎦⎤
x
(
t
)
=
e
t
A
c
=
[
c
1
e
t
+
c
2
t
e
t
c
2
e
t
c
3
e
2
t
]
x(t)=e^{t A} c=\left[\begin{array}{c} c_{1} e^{t}+c_{2} t e^{t} \\ c_{2} e^{t} \\ c_{3} e^{2 t} \end{array}\right]
x(t)=etAc=⎣⎡c1et+c2tetc2etc3e2t⎦⎤
例22:
矩阵函数
e
t
A
e^{t A}
etA 的列向量
x
1
(
t
)
,
⋯
,
x
n
(
t
)
x_{1}(t), \cdots, x_{n}(t)
x1(t),⋯,xn(t) 构成 齐次方程
x
′
(
t
)
=
A
∙
x
(
t
)
x^{\prime}(t)=A \bullet x(t)
x′(t)=A∙x(t) 的基础解系
解:
e
t
A
可逆
⇒
x
1
(
t
)
,
⋯
,
x
n
(
t
)
e^{t A \text { 可逆 }} \Rightarrow x_{1}(t), \cdots, x_{n}(t)
etA 可逆 ⇒x1(t),⋯,xn(t) 线性无关
取
c
=
e
j
⇒
x
j
(
t
)
=
e
t
A
c
c=e_{j} \Rightarrow x_{j}(t)=e^{t A} c
c=ej⇒xj(t)=etAc 是
x
′
(
t
)
=
A
⋅
x
(
t
)
x^{\prime}(t)=A \cdot x(t)
x′(t)=A⋅x(t) 的一个解
通解
x
(
t
)
=
e
t
A
c
=
c
1
∙
x
1
(
t
)
+
⋯
+
c
n
⋅
x
n
(
t
)
x(t)=e^{t A} c=c_{1} \bullet x_{1}(t)+\cdots+c_{n} \cdot x_{n}(t)
x(t)=etAc=c1∙x1(t)+⋯+cn⋅xn(t)
例23: 设
A
=
[
1
1
0
1
0
2
]
,
b
(
t
)
=
[
1
0
e
2
t
]
,
x
(
0
)
=
[
−
1
1
0
]
A=\left[\begin{array}{lll}1 & 1 & 0 \\ 1 & 0 \\ & 2\end{array}\right], b(t)=\left[\begin{array}{c}1 \\ 0 \\ e^{2 t}\end{array}\right], x(0)=\left[\begin{array}{c}-1 \\ 1 \\ 0\end{array}\right]
A=⎣⎡111020⎦⎤,b(t)=⎣⎡10e2t⎦⎤,x(0)=⎣⎡−110⎦⎤ 求
x
′
(
t
)
=
A
⋅
x
(
t
)
+
b
(
t
)
x^{\prime}(t)=A \cdot x(t)+b(t)
x′(t)=A⋅x(t)+b(t) 满足初始条件
x
(
0
)
x(0)
x(0) 的特解
e
t
A
=
[
e
t
t
e
t
0
e
t
0
e
2
t
]
e
−
τ
A
b
(
τ
)
=
[
e
−
τ
0
1
]
,
∫
0
t
e
−
τ
A
b
(
τ
)
d
τ
=
[
1
−
e
−
t
0
t
]
x
(
t
)
=
e
t
A
⋅
[
(
−
1
1
0
)
+
[
1
−
e
−
t
0
t
]
]
=
[
t
e
t
−
1
e
t
t
e
2
t
]
\begin{gathered} e^{t A}=\left[\begin{array}{ccc} e^{t} & t e^{t} & 0 \\ & e^{t} & 0 \\ & e^{2 t} \end{array}\right] \\ e^{-\tau A} b(\tau)=\left[\begin{array}{c} e^{-\tau} \\ 0 \\ 1 \end{array}\right], \quad \int_{0}^{t} e^{-\tau A} b(\tau) d \tau=\left[\begin{array}{c} 1-e^{-t} \\ 0 \\ t \end{array}\right] \\ x(t)=e^{t A} \cdot\left[\left(\begin{array}{c} -1 \\ 1 \\ 0 \end{array}\right)+\left[\begin{array}{c} 1-e^{-t} \\ 0 \\ t \end{array}\right]\right]=\left[\begin{array}{c} t e^{t}-1 \\ e^{t} \\ t e^{2 t} \end{array}\right] \end{gathered}
etA=⎣⎡ettetete2t00⎦⎤e−τAb(τ)=⎣⎡e−τ01⎦⎤,∫0te−τAb(τ)dτ=⎣⎡1−e−t0t⎦⎤x(t)=etA⋅⎣⎡⎝⎛−110⎠⎞+⎣⎡1−e−t0t⎦⎤⎦⎤=⎣⎡tet−1ette2t⎦⎤