【矩阵论】2. 矩阵分解——正规谱分解——正规阵

矩阵论
1. 准备知识——复数域上的矩阵与换位公式)
1. 准备知识——复数域上的内积域正交阵
1. 准备知识——相似对角化与合同&正定阵
2. 矩阵分解—— SVD准备知识——奇异值
2. 矩阵分解——SVD
2. 矩阵分解——QR分解
2. 矩阵分解——乔利斯分解&平方根公式
2. 矩阵分解——正规谱分解——正规阵
2. 矩阵分解——正规分解
2. 矩阵分解——单阵及特征值特征向量一些求法


6.1 正规阵

6.1.1 正规阵定义

若方阵A满足 A H A = A A H A^HA=AA^H AHA=AAH ,则A为正规阵

正规条件: A H A = A A H A^HA=AA^H AHA=AAH

6.1.2 正规阵特点

  1. 正规阵必为方阵
  2. A 正规    ⟺    A H 正规 A正规\iff A^H正规 A正规AH正规 A 不正规    ⟺    A H 不正规 A不正规\iff A^H不正规 A不正规AH不正规

6.1.3 常见正规阵

a. 对角阵

对角阵 A = ( a 1 ⋱ a n ) A=\left(\begin{matrix}a_1&&\\&\ddots&\\&&a_n\end{matrix}\right) A= a1an 必正规

三角正规阵必对角

若三角阵 B = ( b 1 b 12 ⋯ b 1 n b 2 ⋯ b 2 n ⋱ b n ) B=\left(\begin{matrix}b_{1}&b_{12}&\cdots&b_{1n}\\&b_{2}&\cdots&b_{2n}\\&&\ddots&\\&&&b_n\end{matrix}\right) B= b1b12b2b1nb2nbn 正规,则 B = ( b 1 b 2 ⋱ b n ) B=\left(\begin{matrix}b_1&&&&\\&b_2&&\\&&\ddots&\\&&&b_n\end{matrix}\right) B= b1b2bn 为对角形

证明:严格三角阵不是正规阵
设 B = ( b 1 b 12 b 13 b 2 b 23 b 3 ) 为正规阵, B H = ( b 1 ‾ b 12 ‾ b 2 ‾ b 13 ‾ b 23 ‾ b 3 ‾ ) B B H = ( b 1 b 12 b 13 b 2 b 23 b 3 ) ( b 1 ‾ b 12 ‾ b 2 ‾ b 13 ‾ b 23 ‾ b 3 ‾ ) = ( ∣ b 1 ∣ 2 + ∣ b 12 ∣ 2 + ∣ b 13 ∣ 2 ∣ b 23 ∣ 2 + ∣ b 2 ∣ 2 ∣ b 3 ∣ 2 ) B H B = ( b 1 ‾ b 12 ‾ b 2 ‾ b 13 ‾ b 23 ‾ b 3 ‾ ) ( b 1 b 12 b 13 b 2 b 23 b 3 ) = ( ∣ b 1 ∣ 2 ∣ b 12 ∣ 2 + ∣ b 2 ∣ 2 ∣ b 13 ∣ 2 + ∣ b 23 ∣ 2 + ∣ b 3 ∣ 2 ) , 若 B 满足正规阵,则 B H B = B B H , 即 ⇒ { ∣ b 1 ∣ 2 = ∣ b 1 ∣ 2 + ∣ b 12 ∣ 2 + ∣ b 13 ∣ 2 ∣ b 12 ∣ 2 + ∣ b 2 ∣ 2 = ∣ b 23 ∣ 2 + ∣ b 2 ∣ 2 ∣ b 3 ∣ 2 = ∣ b 13 ∣ 2 + ∣ b 23 ∣ 2 + ∣ b 3 ∣ 2 ⇒ { ∣ b 12 ∣ 2 = 0 ∣ b 13 ∣ 2 = 0 ∣ b 23 ∣ 2 = 0 ⇒ b 12 = b 13 = b 23 = 0 ,即 B = ( b 1 b 2 b 3 ) 为对角形 \begin{aligned} &设B=\left( \begin{matrix} b_1&b_{12}&b_{13}\\ &b_2&b_{23}\\ &&b_{3} \end{matrix} \right)为正规阵,B^H=\left( \begin{matrix} \overline{b_1}&&\\ \overline{b_{12}}&\overline{b_2}&\\ \overline{b_{13}}&\overline{b_{23}}&\overline{b_3} \end{matrix} \right)\\ &BB^H=\left( \begin{matrix} b_1&b_{12}&b_{13}\\ &b_2&b_{23}\\ &&b_{3} \end{matrix} \right)\left( \begin{matrix} \overline{b_1}&&\\ \overline{b_{12}}&\overline{b_2}&\\ \overline{b_{13}}&\overline{b_{23}}&\overline{b_3} \end{matrix} \right)\\ &=\left( \begin{matrix} \vert b_1 \vert^2+\vert b_{12} \vert^2+\vert b_{13} \vert^2 &&\\ &\vert b_{23} \vert^2+\vert b_2 \vert^2&\\ &&\vert b_3 \vert^2 \end{matrix} \right)\\ &B^HB=\left( \begin{matrix} \overline{b_1}&&\\ \overline{b_{12}}&\overline{b_2}&\\ \overline{b_{13}}&\overline{b_{23}}&\overline{b_3} \end{matrix} \right)\left( \begin{matrix} b_1&b_{12}&b_{13}\\ &b_2&b_{23}\\ &&b_{3} \end{matrix} \right)\\ &=\left( \begin{matrix} \vert b_1\vert^2 &&\\ &\vert b_{12}\vert^2+\vert b_2\vert^2&\\ &&\vert b_{13}\vert^2+\vert b_{23}\vert^2+\vert b_{3}\vert^2 \end{matrix} \right),\\ &若B满足正规阵,则B^HB=BB^H,即\\ &\Rightarrow \left\{ \begin{aligned} \vert b_1\vert^2 = \vert b_1 \vert^2+\vert b_{12} \vert^2+\vert b_{13} \vert^2\\ \vert b_{12}\vert^2+\vert b_2\vert^2=\vert b_{23} \vert^2+\vert b_2 \vert^2\\ \vert b_3 \vert^2=\vert b_{13}\vert^2+\vert b_{23}\vert^2+\vert b_{3}\vert^2 \end{aligned} \right.\Rightarrow \left\{ \begin{aligned} \vert b_{12} \vert^2=0\\ \vert b_{13} \vert^2=0\\ \vert b_{23} \vert^2=0 \end{aligned} \right.\\ &\Rightarrow b_{12}=b_{13}=b_{23}=0,即B=\left( \begin{matrix} b_{1}&&\\ &b_{2}&\\ &&b_{3} \end{matrix} \right)为对角形 \end{aligned} B= b1b12b2b13b23b3 为正规阵,BH= b1b12b13b2b23b3 BBH= b1b12b2b13b23b3 b1b12b13b2b23b3 = b12+b122+b132b232+b22b32 BHB= b1b12b13b2b23b3 b1b12b2b13b23b3 = b12b122+b22b132+b232+b32 ,B满足正规阵,则BHB=BBH, b12=b12+b122+b132b122+b22=b232+b22b32=b132+b232+b32 b122=0b132=0b232=0b12=b13=b23=0,即B= b1b2b3 为对角形

若分块阵 A = ( B C 0 D ) A=\left(\begin{matrix} B&C\\0&D \end{matrix}\right) A=(B0CD) 正规,则C=0,且B,D都正规, A = ( B 0 0 D ) A=\left(\begin{matrix} B&0\\0&D \end{matrix}\right) A=(B00D)
A H A = ( B H 0 C H D H ) ( B C 0 D ) = ( B H B B H C C H B C H C + D H D ) A A H = ( B C 0 D ) ( B H 0 C H D H ) = ( B B H + C C H C D H D C H C H C + D D H ) 由于 t r ( A H A ) = t r ( A A H ) , ∴ t r ( C C H ) = 0 , 利用迹公式可写 t r ( C C H ) = ∑ ∣ c i , j ∣ 2 = 0 ,其中 C = ( c i , j ) , C 为零阵 且 B B H = B H B , D D H = D H D \begin{aligned} &A^HA=\left( \begin{matrix} B^H&0\\ C^H&D^H \end{matrix} \right)\left( \begin{matrix} B&C\\ 0&D \end{matrix} \right)=\left( \begin{matrix} B^HB&B^HC\\ C^HB&C^HC+D^HD \end{matrix} \right)\\ &AA^H=\left( \begin{matrix} B&C\\ 0&D \end{matrix} \right)\left( \begin{matrix} B^H&0\\ C^H&D^H \end{matrix} \right)=\left( \begin{matrix} BB^H+CC^H&CD^H\\ DC^H&C^HC+DD^H \end{matrix} \right)\\ &由于tr(A^HA)=tr(AA^H),\therefore tr(CC^H)=0,\\ &利用迹公式可写tr(CC^H)=\sum \vert c_{i,j} \vert^2=0,其中C=(c_{i,j}),C为零阵\\ &且BB^H=B^HB,DD^H=D^HD \end{aligned} AHA=(BHCH0DH)(B0CD)=(BHBCHBBHCCHC+DHD)AAH=(B0CD)(BHCH0DH)=(BBH+CCHDCHCDHCHC+DDH)由于tr(AHA)=tr(AAH),tr(CCH)=0,利用迹公式可写tr(CCH)=ci,j2=0,其中C=(ci,j)C为零阵BBH=BHB,DDH=DHD
由证明过程可见,严格三角阵为非正规阵

b. H阵与斜H阵

Hermite阵与斜Hermite阵必正规
若 A 是 H e r m i t e 阵,则 A H = A , A H A = A A = A A H \begin{aligned} 若A是Hermite阵,则A^H=A,A^HA=AA=AA^H \end{aligned} AHermite阵,则AH=AAHA=AA=AAH

  • 实对称阵与反对称阵都是正规阵

c. U阵

U阵必正规(实正交阵)
A H A = I = A A H \begin{aligned} A^HA=I=AA^H \end{aligned} AHA=I=AAH

6.1.4 正规阵的构造方法

a. 倍数法则

若A正规,取倍数k,则kA为正规阵,

( 0 i i i 0 i i i i ) = i ( 0 1 1 1 0 1 1 1 1 ) , ( i i i 2 i ) = i ( 1 1 1 2 ) 都是正规阵 A = 1 2 ( i 1 1 i ) 为正规 U 阵,则 2 A = ( i 1 1 i ) \begin{aligned} \left( \begin{matrix} 0&i&i\\ i&0&i\\ i&i&i \end{matrix} \right)=i\left( \begin{matrix} 0&1&1\\ 1&0&1\\ 1&1&1 \end{matrix} \right),\left( \begin{matrix} i&i\\ i&2i \end{matrix} \right)=i\left( \begin{matrix} 1&1\\1&2 \end{matrix} \right)都是正规阵\\ A=\frac{1}{\sqrt{2}}\left( \begin{matrix} i&1\\ 1&i \end{matrix} \right)为正规U阵,则\sqrt{2}A=\left( \begin{matrix} i&1\\ 1&i \end{matrix} \right) \end{aligned} 0iii0iiii =i 011101111 ,(iii2i)=i(1112)都是正规阵A=2 1(i11i)为正规U阵,则2 A=(i11i)

b. 平移法则

若A正规,则 A ± c I A\pm cI A±cI 正规

若 A 是正规阵,则 ( A ± c I ) H ( A ± c I ) = ( A H ± c I ) ( A ± c I ) = A H A ± c A H ± c A + c 2 I = A A H ± c A ± c A H + c 2 I = ( A ± c I ) ( A H ± c I ) \begin{aligned} 若A是正规阵,则(A\pm cI)^H(A\pm cI)&=(A^H\pm cI)(A\pm cI)\\ &=A^HA\pm cA^H\pm cA+c^2I\\ &=AA^H\pm cA\pm cA^H+c^2I\\ &=(A\pm cI)(A^H\pm cI) \end{aligned} A是正规阵,则(A±cI)H(A±cI)=(AH±cI)(A±cI)=AHA±cAH±cA+c2I=AAH±cA±cAH+c2I=(A±cI)(AH±cI)

c. U相似

若A正规,则 Q H A Q Q^HAQ QHAQ 也正规,其中Q为U阵( Q H = Q − 1 Q^H=Q^{-1} QH=Q1),即正规阵的U相似阵一定正规

证明:
∵ A H A = A A H ,且存在 U 阵 Q ,使得 Q H A Q = B ,即证 B H B = B B H B H B = ( Q H A Q ) H ( Q H A Q ) = Q H A H Q Q H A Q = Q H A H A Q B B H = ( Q H A Q ) ( Q A Q H ) H = Q H A Q Q H A H Q = Q H A A H Q = A A H = A H A Q H A H A Q ∴ B H B = B B H , B 为正规阵 \begin{aligned} &\because A^HA=AA^H,且存在U阵Q,使得Q^HAQ=B,即证B^HB=BB^H\\ &B^HB=(Q^HAQ)^H(Q^HAQ)=Q^HA^HQQ^HAQ=Q^HA^HAQ\\ &BB^H=(Q^HAQ)(QAQ^H)^H=Q^HAQQ^HA^HQ=Q^HAA^HQ\overset{AA^H=A^HA}{=}Q^HA^HAQ\\ &\therefore B^HB=BB^H,B为正规阵 \end{aligned} AHA=AAH,且存在UQ,使得QHAQ=B,即证BHB=BBHBHB=(QHAQ)H(QHAQ)=QHAHQQHAQ=QHAHAQBBH=(QHAQ)(QAQH)H=QHAQQHAHQ=QHAAHQ=AAH=AHAQHAHAQBHB=BBH,B为正规阵

d. 多项式正规

若 A 正规,则 f ( A ) = λ 0 I + λ 1 A + λ 2 A 2 + ⋯ + λ n A K f(A)=\lambda_0I+\lambda_1A+\lambda_2A^2+\cdots+\lambda_nA^K f(A)=λ0I+λ1A+λ2A2++λnAK 正规

6.1.5 正规阵与其H阵的特征向量相同

若A正规,则 A H A^H AH 与 A 有相同的向量

若 A 正规,且 A X = λ X ,则 A H X = λ ‾ X \begin{aligned} 若A正规,且AX=\lambda X,则A^HX=\overline{\lambda}X \end{aligned} A正规,且AX=λX,则AHX=λX

证明
只需证 ( A H − λ ‾ I ) X = 0 , 即 ( A − λ I ) H X = 0 , 由 ( A − λ I ) X = 0 ∣ A − λ I ∣ 2 = 0 ⇒ ( ( A − λ I ) X ) H ( A − λ I ) X = 0 ⇒ X H ( A − λ I ) H ( A − λ I ) X = 0 由于 A − λ I 正规, ( A − λ I ) H ( A − λ I ) = ( A − λ I ) ( A − λ I ) H 即有 X H ( ( A − λ I ) H ) H ( A − λ I ) H X = 0 ⇒ ( ( A − λ I ) H X ) H ( A − λ I ) H X = ∣ ( A − λ I ) H X ∣ 2 = 0 ⇒ ( A − λ I ) H X = 0 ⇒ ( A H − λ ‾ I ) X = 0 ⇒ A H X = λ ‾ X 故结论得证,若 A 正规,则 A X = λ X    ⟺    A H X = λ ‾ X 其中,若 λ ( A ) = { λ 1 , ⋯   , λ n } , 则 λ ( A H ) = { λ 1 ‾ , ⋯   , λ n ‾ } \begin{aligned} &只需证 (A^H-\overline{\lambda}I)X=0,即(A-\lambda I)^HX=0,由(A-\lambda I)X=0\\ &\vert A-\lambda I \vert^2=0\Rightarrow ((A-\lambda I)X)^H(A-\lambda I)X=0\Rightarrow X^H(A-\lambda I)^H(A-\lambda I)X=0\\ &由于A-\lambda I 正规,(A-\lambda I)^H(A-\lambda I)=(A-\lambda I)(A-\lambda I)^H\\ &即有 X^H((A-\lambda I)^H)^H(A-\lambda I)^HX=0\\ &\Rightarrow ((A-\lambda I)^HX)^H(A-\lambda I)^HX=\vert (A-\lambda I)^HX\vert^2=0\\ &\Rightarrow (A-\lambda I)^HX=0\Rightarrow (A^H-\overline{\lambda}I)X=0\Rightarrow A^HX=\overline{\lambda}X\\ &故结论得证,若A正规,则AX=\lambda X\iff A^HX=\overline{\lambda}X\\ &其中,若\lambda(A)=\{\lambda_1,\cdots,\lambda_n\},则\lambda(A^H)=\{\overline{\lambda_1},\cdots,\overline{\lambda_n}\} \end{aligned} 只需证(AHλI)X=0,(AλI)HX=0,(AλI)X=0AλI2=0((AλI)X)H(AλI)X=0XH(AλI)H(AλI)X=0由于AλI正规,(AλI)H(AλI)=(AλI)(AλI)H即有XH((AλI)H)H(AλI)HX=0((AλI)HX)H(AλI)HX=(AλI)HX2=0(AλI)HX=0(AHλI)X=0AHX=λX故结论得证,若A正规,则AX=λXAHX=λX其中,若λ(A)={λ1,,λn},λ(AH)={λ1,,λn}

你可能感兴趣的:(数学,#,矩阵论,矩阵,线性代数)