【矩阵论】5.复合矩阵

5. 复合矩阵

A A A m × n m\times n m×n 矩阵. 对 1 ⩽ i 1 < ⋯ < i k ⩽ m , 1 ⩽ j 1 < ⋯ < j l ⩽ n 1\leqslant i_1<\cdots<i_k\leqslant m,1\leqslant j_1<\cdots<j_l\leqslant n 1i1<<ikm,1j1<<jln ,
A [ i 1 , ⋯   , i k ∣ j 1 , ⋅ ⋅ ⋅ , j l ] A[i_1,\cdots,i_k|j_1,\cdotp\cdotp\cdotp,j_l] A[i1,,ikj1,⋅⋅⋅,jl] A A A 中由第 i 1 , ⋅ ⋅ ⋅ , i k i_1,\cdotp\cdotp\cdotp,i_k i1,⋅⋅⋅,ik k k k 行和第 j 1 , ⋅ ⋅ ⋅ , j l j_1,\cdotp\cdotp\cdotp,j_l j1,⋅⋅⋅,jl l l l
按自然顺序排列后组成的 k × l k\times l k×l 子矩阵.
A ( i 1 , ⋅ ⋅ ⋅ , i k ∣ j 1 , ⋅ ⋅ ⋅ , j l ) A(i_1,\cdotp\cdotp\cdotp,i_k|j_1,\cdotp\cdotp\cdotp,j_l) A(i1,⋅⋅⋅,ikj1,⋅⋅⋅,jl) 记从 A {A} A 中划去第 i 1 , ⋅ ⋅ ⋅ , i k i_1,\cdotp\cdotp\cdotp,i_k i1,⋅⋅⋅,ik k k k 行和第 j 1 , ⋯   , j l j_1,\cdots,j_l j1,,jl l l l 列后
余下的 ( m − k ) × ( n − l ) (m-k)\times(n-l) (mk)×(nl) 子矩阵.
A [ i 1 , ⋯   , i k ∣ j 1 A[i_1,\cdots,i_k|j_1 A[i1,,ikj1 ⋯   , j l ) \cdots,j_{l}) ,jl) 记从 k × n k\times n k×n 子矩阵 A [ i 1 , ⋯   , i k ∣ 1 , ⋯   , n ] A[i_1,\cdots,i_k|1,\cdots,n] A[i1,,ik∣1,,n] 中划去第 j 1 , ⋯   , j l j_1,\cdots,j_l j1,,jl 列后
余下的 k × ( n − l ) k\times(n-l) k×(nl) 子矩阵.
A ( i 1 , ⋅ ⋅ ⋅ , i k ∣ j 1 , ⋅ ⋅ ⋅ , j l ] A(i_1,\cdotp\cdotp\cdotp,i_k|j_1,\cdotp\cdotp\cdotp,j_l] A(i1,⋅⋅⋅,ikj1,⋅⋅⋅,jl] 记从 m × l m\times l m×l 子矩阵 A [ 1 , ⋅ ⋅ ⋅ , m ∣ j 1 , ⋯   , j i ] {A}[1,\cdotp\cdotp\cdotp,m|j_1,\cdots,j_i] A[1,⋅⋅⋅,mj1,,ji] 中划去第 i 1 , ⋯   , i k i_1,\cdots,i_k i1,,ik 行后
余下的 ( m − k ) × l (m-k)\times l (mk)×l 子矩阵.

5.1 复合矩阵的定义和性质

A A A m × n m\times n m×n 矩阵. 定义 Q k , m = { ( i 1 , … , i k ) ∣ 1 ≤ i , < ⋯ < i k ≤ n } Q_{k,m}=\{(i_{1},\ldots,i_{k})|1\leq i,<\cdots<i_{k}\leq n\} Qk,m={(i1,,ik)∣1i,<<ikn}
对于 1 ⩽ k ⩽ min ⁡ { m , n } 1\leqslant k\leqslant\min\{m,n\} 1kmin{m,n}, 定义 A A A k k k复合矩阵 C k ( A ) C_k(A) Ck(A) 如下:

C k ( A ) C_k(A) Ck(A) ( m k ) × ( n k ) \binom mk\times\binom nk (km)×(kn) 矩阵,它分别以 α ∈ Q k , m \alpha\in Q_{k,m} αQk,m β ∈ Q k . n {\beta}\in Q_{k.n} βQk.n 作为行标和列标,并均按字典式排序, C k ( A ) C_k({A}) Ck(A) ( α , β ) (\boldsymbol\alpha,\boldsymbol{\beta}) (α,β) 元是 det ⁡ A [ α ∣ β ] . \det\boldsymbol{A}[\boldsymbol{\alpha}|\boldsymbol{\beta}]. detA[αβ].

例,当 m = n = 3 , k = 2 m=n=3,k=2 m=n=3,k=2 时, C 2 ( A ) C_2(A) C2(A) 3 × 3 3×3 3×3 矩阵:
C 2 ( A ) = ( det ⁡ A [ 1 , 2 ∣ 1 , 2 ] det ⁡ A [ 1 , 2 ∣ 1 , 3 ] det ⁡ A [ 1 , 2 ∣ 2 , 3 ] det ⁡ A [ 1 , 3 ∣ 1 , 2 ] det ⁡ A [ 1 , 3 ∣ 1 , 3 ] det ⁡ A [ 1 , 3 ∣ 2 , 3 ] det ⁡ A [ 2 , 3 ∣ 1 , 2 ] det ⁡ A [ 2 , 3 ∣ 1 , 3 ] det ⁡ A [ 2 , 3 ∣ 2 , 3 ] ) . C_2(A)=\begin{pmatrix}\det A[1,2\mid1,2]&\det A[1,2\mid1,3]&\det A[1,2\mid2,3]\\\det A[1,3\mid1,2]&\det A[1,3\mid1,3]&\det A[1,3\mid2,3]\\\det A[2,3\mid1,2]&\det A[2,3\mid1,3]&\det A[2,3\mid2,3]\end{pmatrix}. C2(A)= detA[1,21,2]detA[1,31,2]detA[2,31,2]detA[1,21,3]detA[1,31,3]detA[2,31,3]detA[1,22,3]detA[1,32,3]detA[2,32,3] .

特别地,总有 C 1 ( A ) = A C_1({A})={A} C1(A)=A ; 当 m = n m=n m=n , C n ( A ) = ( det ⁡ A ) ,{C}_n({A})=(\det{A}) ,Cn(A)=(detA)(作为 1 阶方阵).

性质: ( 1 ) ( C k ( A ) ) T = C k ( A T ) ( 2 ) C k ( I n ) = I ( n k ) ( 3 ) D 是对角阵 ⇒ C k ( D ) 也是对角阵 且  C k ( D ) 的对角元集合 = D 中对角元集合中所有k个乘积 ( 4 ) 上述 ( 3 ) 中结论对下三角 ( 上三角 ) 成立。 \begin{aligned}&(1)(C_k(A))^T=C_k(A^T)\quad(2)C_k(I_n)=I_{\binom nk}\\&(3)D\text{是对角阵}\Rightarrow C_k(D)\text{也是对角阵}\\&\text{且}~C_k(D)\text{的对角元集合}=D\text{中对角元集合中所有k个乘积}\\&(4)上述(3)中结论对下三角(上三角)成立。\end{aligned} (1)(Ck(A))T=Ck(AT)(2)Ck(In)=I(kn)(3)D是对角阵Ck(D)也是对角阵 Ck(D)的对角元集合=D中对角元集合中所有k个乘积(4)上述(3)中结论对下三角(上三角)成立。

定理 5.1.1 A A A B B B 分别是 m × p m\times p m×p p × n p\times n p×n 矩阵, 1 ⩽ k ⩽ min ⁡ { m , p , n } 1\leqslant k\leqslant\min\{ m,p,n\} 1kmin{m,p,n}, 则有 C k ( A B ) = C k ( A ) C k ( B ) ; C_{k}\left ( AB\right ) = C_{k}\left ( A\right ) C_{k}\left ( B\right ) ; Ck(AB)=Ck(A)Ck(B);

证:左 ( α , β ) (\alpha,\beta) (α,β) = det ⁡ ( A B [ α ∣ β ] ) =\det(AB[\alpha|\beta]) =det(AB[αβ])
( α , β ) (\alpha,\beta) (α,β) = ∑ V ∈ Q k , p C k ( A ) ( α , v ) C k ( B ) ( v , β ) Cauchy- Binet → ∑ v ∈ Q k , p det ⁡ A [ α ∣ v ] det ⁡ B ( v , β ) =\sum_{V\in Q_{k,p}}C_{k}(A)_{(\alpha,v)}C_{k}(B)_{(v,\beta)}\underrightarrow{\text{Cauchy- Binet}}\sum_{v\in Q_{k,p}}\det A_{[\alpha|v]}\det B_{(v,\beta)} =VQk,pCk(A)(α,v)Ck(B)(v,β) Cauchy- BinetvQk,pdetA[αv]detB(v,β)

定理 5.1.2 A A A n n n 阶(复)方阵, S p e c A = { λ 1 , ⋯   , λ n } \mathrm{Spec}\boldsymbol{A}=\left\{\lambda_1,\cdots,\lambda_n\right\} SpecA={λ1,,λn},
S p e c   C k ( A ) = { λ i 1 ⋯ λ i k ∣ 1 ⩽ i 1 < ⋯ < i k ⩽ n } . \mathrm{Spec}~\boldsymbol{C}_k(\boldsymbol{A})=\{\lambda_{i_1}\cdots\lambda_{i_k}\mid1\leqslant i_1<\cdots<i_k\leqslant n\}. Spec Ck(A)={λi1λik1i1<<ikn}.

A A A 必相似于上三角方阵 T T T , 即有可逆方阵 S S S, 使 S A S − 1   =   T   =   ( λ 1 ∗ ⋱ 0 λ n ) . SAS^{-1}\:=\:T\:=\:\begin{pmatrix}\lambda_1&&&*\\&\ddots&&\\\mathbf{0}&&&\lambda_n\end{pmatrix}. SAS1=T= λ10λn .
因此 , C k ( T ) = C k ( S ) C k ( A ) ( C k ( S ) ) − 1 . ,\boldsymbol{C}_k(T)=\boldsymbol{C}_k(\boldsymbol{S})\boldsymbol{C}_k(\boldsymbol{A})(\boldsymbol{C}_k(\boldsymbol{S}))^{-1}. ,Ck(T)=Ck(S)Ck(A)(Ck(S))1. 但注意到上三角方阵 T T T k k k 级复合矩 C k ( T ) C_k(T) Ck(T) 也一定是上三角方阵(因为若 α , β ∈ Q k . n \alpha,\beta\in Q_{k.n} α,βQk.n, 且按字典式排序 , α ,\alpha ,α β \beta β 之后,则不难验证 det ⁡ T [ α ∣ β ] = 0 \det T[\alpha|\beta]=0 detT[αβ]=0),而且这个上三角方阵 C k ( T ) C_k(T) Ck(T) 的主对角线上元素的集合正是 { { λ i 1 ⋯ λ i k ∣ 1 ⩽ i 1 < ⋯ < i k ⩽ n } . \{\left\{\lambda_{i_1}\cdots\lambda_{i_k}\mid1\leqslant i_1<\cdots<i_k\leqslant n\right\}. {{λi1λik1i1<<ikn}.

5.2 加性复合矩阵

定义 5.2.1 A A A n n n 阶方阵 , t ,t ,t 是不定元. 记 C k ( I n   +   t A ) C_{k}\left ( I_{n}\:+ \:t\mathbf{A} \right ) Ck(In+tA) = I ( n k ) \boldsymbol{I}_{{\binom nk}} I(kn) + t Δ k ( A ) t\Delta _{k}\left ( \boldsymbol{A}\right ) tΔk(A) + t 2 R t^{2}\boldsymbol{R} t2R ,
其中 Δ k ( A ) ∈ M ( n k ) ( C ) , R ∈ M ( n k ) ( C [ t ] ) . \Delta_k(\boldsymbol{A})\in M_{\binom nk}(\boldsymbol{C}),\boldsymbol{R}\in M_{\binom nk}(\mathbf{C}[t]). Δk(A)M(kn)(C),RM(kn)(C[t]). ( n k ) \binom nk (kn) 阶方阵 Δ k ( A ) \Delta_k(\boldsymbol{A}) Δk(A) A \boldsymbol{A} A k k k加性复合矩阵.

易知 k = 1 k=1 k=1 时, Δ 1 ( A ) = A ; \Delta_1(\boldsymbol{A})=\boldsymbol{A}; Δ1(A)=A; k = n k=n k=n , Δ n ( A ) = t r A . ,\Delta_n(\boldsymbol{A})=\mathrm{tr}\boldsymbol{A}. ,Δn(A)=trA.

C 1 ( I + t A ) = I + t A , C n ( I + t A ) = det ⁡ ( I + t A ) C_1(I+tA)=I+tA,C_n(I+tA)=\det(I+tA) C1(I+tA)=I+tA,Cn(I+tA)=det(I+tA) ( f ( t ) + 1 t 多项式 g ( t ) + 1 t 多项式 h ( t ) + 1 ) \begin{pmatrix}f(t)+1&&t多项式\\&g(t)+1&\\t多项式& &h(t)+1\end{pmatrix} f(t)+1t多项式g(t)+1t多项式h(t)+1 t t t 的一次项仅在 n n n 个对角元乘积 ( t a 11 + 1 ) ⋯ ( t a m + 1 ) (ta_{11}+1)\cdots(ta_{m}+1) (ta11+1)(tam+1) 中, 即 t r ( A ) tr(A) tr(A)

性质: 若 A A A 是下三角,则 Δ k ( A ) \Delta_k(A) Δk(A) 也是下三角,且 Δ k ( A ) \Delta_k(A) Δk(A) 的对有元集合 = A =A =A 的对角元集合中所有 k k k 个元之和.

A = ( 1 2 3 1 2 3 ) , k = 2 A=\begin{pmatrix}1&2&3\\&1&2\\&&3\end{pmatrix},k=2 A= 121323 ,k=2
C 2 ( I 2 + t A ) = ( ( t + 1 ) 2 2 t ( t + 1 ) t 2 − 3 t 0 ( t + 1 ) ( t + 1 ) 2 t ( 3 t + 1 ) 0 0 ( t + 1 ) ( 3 t + 1 ) ) = I ( 3 2 ) + t ( 2 2 − 3 0 4 2 0 0 4 ) + t 2 ( 1 2 1 0 3 6 0 0 3 ) \begin{aligned}C_2(I_2+tA)&=\begin{pmatrix}(t+1)^2&2t(t+1)&t^2-3t\\0&(t+1)(t+1)&2t(3t+1)\\0&0&(t+1)(3t+1)\end{pmatrix}\\&=I_{\binom 32}+t\begin{pmatrix}2&2&-3\\0&4&2\\0&0&4\end{pmatrix}+t^2\begin{pmatrix}1&2&1\\0&3&6\\0&0&3\end{pmatrix}\end{aligned} C2(I2+tA)= (t+1)2002t(t+1)(t+1)(t+1)0t23t2t(3t+1)(t+1)(3t+1) =I(23)+t 200240324 +t2 100230163

定理 5.2.2 A A A B B B 都是 n n n 阶方阵,则

  1. Δ k ( A + B ) = Δ k ( A ) + Δ k ( B ) ; \Delta_{k}(A+B)=\Delta_{k}(A)+\Delta_{k}(B); Δk(A+B)=Δk(A)+Δk(B);
  2. 对可逆方阵 S , Δ k ( S A S − 1 ) = C k ( S ) Δ k ( A ) C k ( S ) − 1 S,\Delta_k(SAS^{-1})=C_k(S)\Delta_k(A)C_k(S)^{-1} S,Δk(SAS1)=Ck(S)Δk(A)Ck(S)1, 从而由 A A A B B B 相似,
    可得 Δ k ( A ) \Delta_k(A) Δk(A) Δ k ( B ) \Delta_k(B) Δk(B) 相似;
  3. 若记 S p e c A = { λ 1 , ⋯   , λ n } Spec\boldsymbol{A}=\{\lambda_1,\cdots,\lambda_n\} SpecA={λ1,,λn},则 S p e c   Δ k ( A ) = { λ i 1 + ⋯ + λ i k ∣ 1 ⩽ i 1 < ⋯ < i k ⩽ n } . Spec~\Delta_k(\boldsymbol{A})=\{\lambda_{i_1}+\cdots+\lambda_{i_k}\mid1\leqslant i_1<\cdots<i_k\leqslant n\}. Spec Δk(A)={λi1++λik1i1<<ikn}.

证 (1)将等式
C k ( ( I + t A ) ( I + t B ) ) = C k ( I + t A ) C k ( I + t B ) C_k((I+tA)(I+tB))=C_k(I+tA)C_k(I+tB) Ck((I+tA)(I+tB))=Ck(I+tA)Ck(I+tB)两边分别按 t t t 展开后,得
右边 = ( I + t Δ k ( A ) + t 2 R ) ( I + t Δ k ( B ) + t 2 S ) = I + t ( Δ k ( A ) + Δ k ( B ) ) + t 2 ( ⋯   ) , =(I+t\Delta_k(\boldsymbol{A})+t^2\boldsymbol{R})(I+t\Delta_k(\boldsymbol{B})+t^2\boldsymbol{S})=I+t(\Delta_{k}(A)+\Delta_{k}(B))+t^{2}(\cdots), =(I+tΔk(A)+t2R)(I+tΔk(B)+t2S)=I+t(Δk(A)+Δk(B))+t2(),
左边 = C k ( I + t ( A + B )   + t 2 A B ) =   I   +   t Δ k ( A + B ) + t 2 ( ⋯   ) . =C_{k}(I+t(A+B)\:+t^{2}AB)=\:I\:+\:t\Delta_{k}\left(A+B\right)+t^{2}\left(\cdots\right). =Ck(I+t(A+B)+t2AB)=I+tΔk(A+B)+t2().
比较两边 t t t 的系数即得(1).
(2)    C k ( S ( I + t A ) S − 1 ) = C k ( S ) C k ( I + t A ) ( C k ( S ) ) − 1 = I + t C k ( S ) Δ k ( A ) ( C k ( S ) ) − 1 + t 2 ( ⋯ ) . ~~\begin{aligned}C_{k}\left(\boldsymbol{S}(\boldsymbol{I}+t\boldsymbol{A})\boldsymbol{S}^{-1}\right)&=\boldsymbol{C}_{k}\left(\boldsymbol{S}\right)\boldsymbol{C}_{k}\left(\boldsymbol{I}+t\boldsymbol{A}\right)\left(\boldsymbol{C}_{k}\left(\boldsymbol{S}\right)\right)^{-1}\\&=\boldsymbol{I}+t\boldsymbol{C}_{k}\left(\boldsymbol{S}\right)\Delta_{k}\left(\boldsymbol{A}\right)\left(\boldsymbol{C}_{k}\left(\boldsymbol{S}\right)\right)^{-1}+t^{2}\left(\boldsymbol{\cdots}\right).\end{aligned}   Ck(S(I+tA)S1)=Ck(S)Ck(I+tA)(Ck(S))1=I+tCk(S)Δk(A)(Ck(S))1+t2().
(3) 和讨论 C k ( A ) C_{k}(A) Ck(A) 的谱一样,设可逆方阵 S S S 使 S A S − 1 = T = ( λ 1 ∗ ⋱ 0 λ n ) . SAS^{-1}=T=\begin{pmatrix}\lambda_1&&&*\\&\ddots&&\\\mathbf{0}&&&\lambda_n\end{pmatrix}. SAS1=T= λ10λn .
由 (2) 得 Δ k ( A ) \Delta_k(\boldsymbol{A}) Δk(A) Δ k ( T ) \Delta_k(T) Δk(T) 相似. 但 C k ( I + t T ) = I + t Δ k ( T ) + t 2 ( ⋯   ) C_k(I+t\boldsymbol{T})=\boldsymbol{I}+t\Delta_k(\boldsymbol{T})+t^2(\cdots) Ck(I+tT)=I+tΔk(T)+t2() 的左边是上三角方阵,其主对角线上元素的集合是 { ( 1 + t λ i 1 ) ⋯ ( 1 + t λ i k ) ∣ 1 ⩽ i 1 < ⋯ < i k ⩽ n } \{(1+t\lambda_{i_1})\cdots(1+t\lambda_{i_k})\mid1\leqslant i_1<\cdots<i_k\leqslant n\} {(1+tλi1)(1+tλik)1i1<<ikn}, 所以右边的 Δ k ( T ) \Delta_k(T) Δk(T) 也是上三角方阵,其主对角线上元素的集合正是 { λ i 1 + ⋯ \{\lambda_{i_1}+\cdots {λi1+ + λ i k ∣ 1 ⩽ i 1 < ⋅ ⋅ ⋅ < i k ⩽ n ∣ = S p e c   Δ k ( T ) = S p e c   Δ k ( A ) . +\lambda_{i_k}\mid1\leqslant i_1<\cdotp\cdotp\cdotp<i_k\leqslant n\mid=Spec~\Delta_k(T)=Spec~\Delta_k(A). +λik1i1<⋅⋅⋅<ikn∣=Spec Δk(T)=Spec Δk(A).

  • 14
    点赞
  • 9
    收藏
    觉得还不错? 一键收藏
  • 0
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值