主要内容

高等代数教学辅导

8.4 酉变换和酉矩阵、正交变换和正交矩阵

建设中!

子节 8.4.1 主要知识点

定义 8.4.1.

\(V\)\(n\)维酉(欧氏)空间,\(\varphi\)\(V\)的线性变换。如果\(\varphi\)保持内积,即对任意\(\alpha,\beta\in V\), 总成立
\begin{equation*} \left(\varphi(\alpha),\varphi(\beta)\right)=(\alpha,\beta), \end{equation*}
则称\(\varphi\)酉变换正交变换)。
  • \(\varphi,\psi\)是酉变换,则
    1. \(\varphi\)可逆,且\(\varphi^{-1}\)也是酉变换;
    2. \(\varphi\psi\)为酉变换。
  • \(\varphi,\psi\)是正交变换,则
    1. \(\varphi\)可逆,且\(\varphi^{-1}\)也是正交变换;
    2. \(\varphi\psi\)为正交变换。
\(n\)维欧氏空间\(V\)中的线性变换\(\varphi\)在标准正交基\(\xi_1,\xi_2,\ldots,\xi_n\)下的矩阵是正交矩阵\(A\)
  1. 如果\(|A|=1\),则称\(\varphi \)第一类正交变换旋转);
  2. 如果\(|A|=-1\),则称\(\varphi\)第二类正交变换

8.4.5.

\(\eta\)\(n\)维欧氏空间\(V\)中一单位向量,定义
\begin{equation*} \varphi(\alpha)=\alpha-2(\eta,\alpha)\eta. \end{equation*}
证明:
  1. \(\varphi\)是正交变换(称为镜面反射);
  2. 存在\(V\)的一个标准正交基,使得\(\varphi\)在这个标准正交基下的矩阵为
    \begin{equation*} \begin{pmatrix} -1 & 0\\ 0 & E_{n-1} \end{pmatrix} \end{equation*}
  3. \(\varphi\)是第二类正交变换。
  • \(|\lambda|=1\),故可设\(a=\cos\theta,b=-\sin\theta\)\(a^2+b^2=1\)\(A \alpha=a \alpha-b \beta=(\alpha,\beta)\begin{pmatrix} \cos\theta\\ \sin\theta \end{pmatrix} \text{,}\) \(A\beta=b \alpha+a \beta=(\alpha,\beta)\begin{pmatrix} -\sin\theta\\\cos\theta \end{pmatrix} \)
    \begin{equation*} A(\alpha,\beta)=(\alpha,\beta)\begin{pmatrix} \cos\theta & -\sin\theta\\ \sin\theta & \cos\theta \end{pmatrix} \end{equation*}

练习 8.4.2 练习

1.

\(\varphi\)是欧氏空间\(V\)上的变换,且对任意\(\alpha,\beta\in V\),都有
\begin{equation*} \left(\varphi(\alpha),\varphi(\beta)\right)=\left(\alpha,\beta\right), \end{equation*}
证明:\(\varphi\)是线性变换,因而是正交变换。
解答.
对任意\(\alpha,\beta\in V, a,b\in\mathbb{R}\),根据已知条件,有
\begin{equation*} \begin{array}{cl} &\left(\varphi(a\alpha+b \beta)-a\varphi(\alpha)-b\varphi(\beta),\varphi(a\alpha+b \beta)-a\varphi(\alpha)-b\varphi(\beta)\right)\\ =&\left(\varphi(a\alpha+b \beta),\varphi(a\alpha+b \beta)\right)-2a\left(\varphi(\alpha),\varphi(a\alpha+b \beta)\right)-2b\left(\varphi(\beta),\varphi(a\alpha+b \beta)\right)\\ &+a^2\left(\varphi(\alpha),\varphi(\alpha)\right)+2ab\left(\varphi(\alpha),\varphi(\beta)\right)+b^2\left(\varphi(\beta),\varphi(\beta)\right)\\ =&\left(a\alpha+b \beta,a\alpha+b \beta\right)-2a\left(\alpha,a\alpha+b \beta\right)-2b\left(\beta,a\alpha+b \beta\right)+a^2\left(\alpha,\alpha\right)\\ &+2ab\left(\alpha,\beta\right)+b^2\left(\beta,\beta\right)\\ =&0 \end{array} \end{equation*}
\(\varphi(a\alpha+b \beta)-a\varphi(\alpha)-b\varphi(\beta)=0\),即\(\varphi(a\alpha+b \beta)=a\varphi(\alpha)+b\varphi(\beta)\)。故\(\varphi\)是线性变换,因而是正交变换。

2.

\(\varphi\)是正交变换,\(U\)\(\varphi\)-子空间,证明:\(U^\perp\)也是\(\varphi\)-子空间。
解答.
证法一:由\(\varphi\)是正交变换可知,\(\varphi^{-1}\)也是正交变换。故\(\forall\alpha\in U^\perp ,\beta\in U\)
\begin{equation*} (\varphi(\alpha),\beta)=(\alpha,\varphi^{-1}(\beta)), \end{equation*}
因为\(U\)\(\varphi\)-不变子空间且\(\varphi\)可逆,所以\(U\)也是\(\varphi^{-1}\)-不变子空间,于是\(\varphi^{-1}(\beta)\in U\),则
\begin{equation*} (\varphi(\alpha),\beta)=(\alpha,\varphi^{-1}(\beta))=0, \end{equation*}
\(\varphi(\alpha)\in U^\perp\)。因此\(U^\perp\)也是\(\varphi\)-不变子空间。
证法二:设\(\xi_1,\cdots ,\xi_r\)\(U\)的一个标准正交基,将其扩充为\(V\)的一个标准正交基\(\xi_1,\cdots ,\xi_r,\xi_{r+1},\cdots ,\xi_n\),则
\begin{equation*} U^\perp =\langle \xi_{r+1},\cdots ,\xi_n\rangle. \end{equation*}
\(\varphi\)是正交变换,所以\(\varphi(\xi_1),\cdots ,\varphi(\xi_r),\varphi(\xi_{r+1}),\cdots ,\varphi(\xi_n)\)也是\(V\)的一个标准正交基。故\(\varphi(\xi_1),\cdots ,\varphi(\xi_r)\)线性无关。又\(U\)\(\varphi\)-不变子空间,所以\(\varphi(\xi_1),\cdots ,\varphi(\xi_r)\in U\)。因此\(\varphi(\xi_1),\cdots ,\varphi(\xi_r)\)\(U\)的一个基。
对任意\(r+1\leq i \leq n, 1\leq j\leq r\),由
\begin{equation*} (\varphi(\xi_i),\varphi(\xi_j))=(\xi_i,\xi_j)=0 \end{equation*}
\(\varphi(\xi_i)\in U^\perp\)。因此\(U^\perp\)也是\(\varphi\)-不变子空间。
证法三:设\(\xi_1,\cdots ,\xi_r\)\(U\)的一个标准正交基,将其扩充为\(V\)的一个标准正交基\(\xi_1,\cdots ,\xi_r,\xi_{r+1},\cdots ,\xi_n\),则
\begin{equation*} U^\perp =\langle \xi_{r+1},\cdots ,\xi_n\rangle. \end{equation*}
因为\(U\)\(\varphi\)-不变子空间,所以
\begin{equation*} \varphi(\xi_1,\xi_2,\cdots ,\xi_n)=(\xi_1,\xi_2,\cdots ,\xi_n)\begin{pmatrix} A&B\\0&C \end{pmatrix}. \end{equation*}
又因为\(\varphi\)是正交变换,故\(\begin{pmatrix} A&B\\0&C \end{pmatrix}\)是正交矩阵,则\(\begin{pmatrix} A&B\\0&C \end{pmatrix}^{-1}=\begin{pmatrix} A&B\\0&C \end{pmatrix}^T\),即
\begin{equation*} \begin{pmatrix} A^{-1}&-A^{-1}BC^{-1}\\0&C^{-1} \end{pmatrix}=\begin{pmatrix} A^T&0\\B^T&C^T \end{pmatrix}, \end{equation*}
\(B=0\)。从而\(U^\perp\)也是\(\varphi\)-不变子空间。

3.

\(\varphi\)是酉变换,证明:\(\varphi\)的属于不同特征值的特征向量必正交。
解答.
\(\lambda, \mu\)\(\varphi\)的两个不同的特征值,\(\alpha ,\beta\)分别是\(\varphi\)的属于\(\lambda, \mu\)的特征向量,则
\begin{equation*} \varphi(\alpha)=\lambda \alpha,\varphi(\beta)=\mu \beta\mbox{,且}\left|\lambda\right|=\left|\mu\right|=1\mbox{。} \end{equation*}
\((\varphi(\alpha),\varphi(\beta))=(\alpha,\beta)\)
\begin{equation*} \lambda\overline{\mu}(\alpha,\beta)=(\alpha,\beta). \end{equation*}
注意到\(\left|\mu\right|=1\)\(\lambda\neq \mu\),所以\(\lambda\overline{\mu}\neq 1\)。因此\((\alpha,\beta)=0\)

4.

\(\varphi\)\(n\)维欧氏空间\(V\)的正交变换。若\(\varphi\)\(1\)作为一个特征值,且属于特征值\(1\)的特征子空间\(V_1\)的维数为\(n-1\),证明:\(\varphi\)是镜面反射。
解答.
\(\lambda_1,\lambda_2,\cdots ,\lambda_n\)\(\varphi\)的所有复特征值,由\(\varphi\)是正交变换,得
\begin{equation*} \lambda_1\lambda_2\cdots \lambda_n=\pm 1\mbox{且}\left|\lambda_i\right|=1,\ (i=1,2,\cdots ,n) \end{equation*}
注意到\(\dim V_1=n-1\),所以\(1\)的代数重数\(\geq n-1\)。不妨设\(\lambda_2=\cdots =\lambda_n=1\),则\(\lambda_1=1\)\(-1\)。若\(\lambda_1=1\),则\(\varphi\)在某组标准正交基下的矩阵为\(E_n\),即\(\varphi={\rm id}_V\),这与\(\dim V_1=n-1\)相矛盾。因此\(-1\)\(\varphi\)的一个特征值。
\(\xi_1\)\(\varphi\)的属于\(-1\)的单位特征向量,\(\xi_2,\cdots ,\xi_n\)是特征子空间\(V_1\)的一个标准正交基,则\(\xi_1,\xi_2,\cdots ,\xi_n\)\(V\)的一个标准正交基,且
\begin{equation*} \varphi (\xi_1,\xi_2,\cdots ,\xi_n)=(\xi_1,\xi_2,\cdots ,\xi_n)\begin{pmatrix} -1&0\\0&E_{n-1} \end{pmatrix}. \end{equation*}
因此,\(\varphi\)是镜面反射。

5.

\(\xi,\eta\)\(n\)维欧氏空间\(V\)中两个不同的单位向量,证明:存在一个镜面反射\(\varphi\),使得\(\varphi(\xi)=\eta\)
解答.
\(\alpha=\frac{\xi- \eta}{\|\xi-\eta\|}\),则\(\alpha\)是单位向量。定义\(V\)上变换\(\varphi\)如下
\begin{equation*} \varphi (\beta)=\beta-2(\alpha,\beta)\alpha, \end{equation*}
\(\varphi\)是镜面反射,且
\begin{equation*} \begin{array}{ccl}\varphi(\xi)&=&\xi-2(\frac{\xi- \eta}{\|\xi-\eta\|},\xi)\frac{\xi- \eta}{\|\xi-\eta\|}\\ &=&\xi-2\frac{(\xi-\eta,\xi)}{(\xi-\eta,\xi-\eta)}(\xi-\eta)\\ &=&\xi-2\frac{1-(\xi,\eta)}{2-2(\xi,\eta)}(\xi-\eta)\\ &=&\eta . \end{array} \end{equation*}

6.

\(A\)\(n\)阶复正规矩阵,证明:\(A\)是Hermite矩阵的充分必要条件是\(A\)的特征值全是实数。
解答.
充分性:因为\(A\)\(n\)阶复正规矩阵,所以存在酉矩阵\(U\),使得
\begin{equation*} A=\overline{U}^{T}\begin{pmatrix} \lambda_1&&&\\ &\lambda_2&&\\ &&\ddots&\\ &&&\lambda_n \end{pmatrix}U, \end{equation*}
其中\(\lambda_1,\lambda_2,\cdots ,\lambda_n\)\(A\)的全部特征值。根据已知条件,\(\lambda_1,\lambda_2,\cdots ,\lambda_n\in\mathbb{R}\),则
\begin{equation*} \overline{A}^T=\overline{U}^{T}\begin{pmatrix} \overline{\lambda_1}&&&\\ &\overline{\lambda_2}&&\\ &&\ddots&\\ &&&\overline{\lambda_n} \end{pmatrix}U=\overline{U}^{T}\begin{pmatrix} \lambda_1&&&\\ &\lambda_2&&\\ &&\ddots&\\ &&&\lambda_n \end{pmatrix}U=A. \end{equation*}
因此\(A\)是Hermite矩阵。
必要性:设\(\lambda\in\mathbb{C}\)\(A\)的特征根,则存在\(0\neq X=(x_1,x_2,\cdots,x_n)^T\in\mathbb{C}^n\)使得
\begin{equation*} AX=\lambda X. \end{equation*}
于是,\(\overline{X}^TAX=\lambda\overline{X}^TX\)。另一方面,由\(A\) 是Hermite阵知\(A=\overline{A}^T\),则
\begin{equation*} \overline{X}^TAX=\overline{X}^T\overline{A}^TX=\overline{AX}^TX=\overline{\lambda}\overline{X}^TX. \end{equation*}
从而\(\lambda\overline{X}^TX=\overline{\lambda}\overline{X}^TX\)。由\(X\neq 0\)可知,\(\overline{X}^TX=\left|x_1\right|^2+\left|x_2\right|^2+\cdots +\left|x_n\right|^2>0\),故\(\overline{\lambda}=\lambda\),即\(\lambda\)是实数。

7.

\(A\)\(n\)阶复正规矩阵,证明:\(A\)是酉矩阵的充分必要条件是\(A\)的特征值全是模为\(1\)的复数。
解答.
充分性:因为\(A\)\(n\)阶复正规矩阵,所以存在酉矩阵\(U\),使得
\begin{equation*} A=\overline{U}^{T}\begin{pmatrix} \lambda_1&&&\\ &\lambda_2&&\\ &&\ddots&\\ &&&\lambda_n \end{pmatrix}U, \end{equation*}
其中\(\lambda_1,\lambda_2,\cdots ,\lambda_n\)\(A\)的全部特征值。由于\(|\lambda_i|=1\),即\(\overline{\lambda_i}\lambda_i=1\),故
\begin{equation*} \overline{A}^{T}A=\overline{U}^{T}\begin{pmatrix} \overline{\lambda_1}&&&\\ &\overline{\lambda_2}&&\\ &&\ddots&\\ &&&\overline{\lambda_n} \end{pmatrix}\begin{pmatrix} \lambda_1&&&\\ &\lambda_2&&\\ &&\ddots&\\ &&&\lambda_n \end{pmatrix}U=E_n. \end{equation*}
因此\(A\)是酉矩阵。
必要性:设\(\lambda\in\mathbb{C}\)\(A\)的特征根,则存在\(0\neq X=(x_1,x_2,\cdots,x_n)^T\in\mathbb{C}^n\),使得
\begin{equation*} AX=\lambda X. \end{equation*}
因为\(A\)是酉矩阵,即\(\overline{A}^TA=E\),所以
\begin{equation*} \overline{X}^TX=\overline{X}^T\overline{A}^TAX=\overline{(AX)}^T(AX)=|\lambda|^2\overline{X}^TX. \end{equation*}
\(X\neq 0\)可知,\(\overline{X}^TX=\left|x_1\right|^2+\left|x_2\right|^2+\cdots +\left|x_n\right|^2>0\),故\(|\lambda|=1\)

8.

\(A\)\(n\)阶复正规矩阵,证明:\(A\)是幂零矩阵的充分必要条件是\(A=0\)
解答.
充分性:显然成立。
必要性:因为\(A\)\(n\)阶复正规矩阵,所以存在酉矩阵\(U\),使得
\begin{equation*} A=\overline{U}^{T}\begin{pmatrix} \lambda_1&&&\\ &\lambda_2&&\\ &&\ddots&\\ &&&\lambda_n \end{pmatrix}U, \end{equation*}
其中\(\lambda_1,\lambda_2,\cdots ,\lambda_n\)\(A\)的全部特征值。由\(A\)是幂零矩阵知
\begin{equation*} \lambda_1=\lambda_2=\cdots=\lambda_n=0, \end{equation*}
\(A=0\)

9.

证明:\(n\)维欧氏空间\(V\)中任意正交变换\(\varphi\)都可以表为一系列镜面反射的乘积。
解答.
证法一:设\(\xi_1,\xi_2,\cdots ,\xi_n\)\(V\)的一个标准正交基,\(\eta_i=\varphi(\xi_i),i=1,2,\cdots ,n\)。由\(\varphi\)是正交变换知,\(\eta_1,\eta_2,\cdots ,\eta_n\)也是\(V\)的一个标准正交基。
  1. \(\xi_i=\eta_i(i=1,2,\cdots ,n)\),令\(\varphi_1(\alpha)=\alpha-2(\xi_1,\alpha)\xi_1\),则
    \begin{equation*} \varphi_1(\xi_1)=-\xi_1,\varphi_1(\xi_j)=\xi_j(j=2,\cdots ,n), \end{equation*}
    \(\varphi_1\)是镜面反射且\(\varphi=\varphi_1\varphi_1\)
  2. \(\xi_1,\cdots ,\xi_n\)\(\eta_1,\cdots ,\eta_n\)不尽相同,不妨设\(\xi_1\neq \eta_1\),那么由上题结论,存在镜面反射\(\varphi_1\),使得
    \begin{equation*} \varphi_1(\xi_1)=\eta_1,\varphi_1(\xi_j)=\zeta_j(j=2,\cdots ,n), \end{equation*}
    \begin{equation*} \xi_1,\xi_2,\cdots,\xi_n\stackrel{\varphi_1}{\longrightarrow}\eta_1,\zeta_2,\cdots ,\zeta_n. \end{equation*}
    • \(\eta_j=\zeta_j(j=2,\cdots ,n)\),则\(\varphi=\varphi_1\),结论成立。
    • 否则,不妨假设\(\eta_2\neq\zeta_2\),令镜面反射
      \begin{equation*} \varphi_2(\alpha)=\alpha-2(\eta,\alpha)\eta,\ \mbox{其中}\eta=\frac{\eta_2-\zeta_2}{\|\eta_2-\zeta_2\|}, \end{equation*}
      \(\varphi_2(\eta_2)=\zeta_2\)。注意到\((\eta_2,\eta_1)=(\zeta_2,\eta_1)=0\),所以\(\varphi_2(\eta_1)=\eta_1\)。于是
      \begin{equation*} \xi_1,\xi_2,\cdots,\xi_n\stackrel{\varphi_1}{\longrightarrow}\eta_1,\zeta_2,\cdots ,\zeta_n\stackrel{\varphi_2}{\longrightarrow}\eta_1,\eta_2,\chi_3\cdots ,\chi_n. \end{equation*}
      依此类推,
      \begin{equation*} \xi_1,\xi_2,\cdots,\xi_n\stackrel{\varphi_1}{\longrightarrow}\eta_1,\zeta_2,\cdots ,\zeta_n\stackrel{\varphi_2}{\longrightarrow}\eta_1,\eta_2,\chi_3\cdots ,\chi_n \end{equation*}
      \begin{equation*} \stackrel{\varphi_3}{\longrightarrow}\cdots\stackrel{\varphi_s}{\longrightarrow}\eta_1,\eta_2,\cdots ,\eta_n. \end{equation*}
      \(\varphi=\varphi_s\varphi_{s-1}\cdots\varphi_1\),其中\(\varphi_i\)都是镜面反射\((i=1,2,\cdots,s)\)
证法二:对维数\(n\)用数学归纳法。
  1. \(n=1\)时,设\(\xi\)\(V\)的一个标准正交基,定义镜面反射
    \begin{equation*} \psi:V\rightarrow V,\ \alpha\mapsto \alpha-2(\xi,\alpha)\xi. \end{equation*}
    \(\varphi\)是正交变换,所以\(\varphi={\rm id}_V\)\(-{\rm id}_V\)
    1. \(\varphi=-{\rm id}_V\) 时,则\(\varphi=\psi\)是一个镜面反射。
    2. \(\varphi={\rm id}_V\)时,则\(\varphi=\psi\psi\)是两个镜面反射的乘积。
  2. 假设对于\(n-1\)维欧氏空间结论成立。对\(n\)维欧氏空间\(V\)上的任一正交变换\(\varphi\)
    1. \(\varphi={\rm id}_V\),则\(\varphi=\psi\psi\),其中\(\psi\)是任一镜面反射。
    2. \(\varphi\neq {\rm id}_V\),则存在\(V\)中单位向量\(\xi_1\),使得\(\varphi(\xi_1)\neq \xi_1\)。令\(\eta_1=\varphi(\xi_1)\),由\(\varphi\)是正交变换知,\(\|\eta_1\|=\|\varphi(\xi_1)\|=\|\xi_1\|=1\)。由上题结论知,存在镜面反射\(\varphi_1\),使得\(\varphi_1(\eta_1)=\xi_1\),则\(\varphi_1\varphi(\xi_1)=\xi_1\),即\(U=\langle \xi_1\rangle\)\(\varphi_1\varphi\)-不变子空间。由于\(\varphi_1\varphi\)仍是正交变换,所以由 练习 8.4.2.2 结论可知,\(U^\perp\)也是\(\varphi_1\varphi\)-不变子空间。故\(\left. (\varphi_1\varphi)\right|_{U^\perp}\)是正交变换,其中\(\dim U^\perp =n-1\)。由归纳假设,在\(U^\perp\)内存在单位向量\(\xi_2,\cdots,\xi_s\),它们分别决定\(U^\perp\)上的镜面反射\(\varphi_2,\cdots ,\varphi_s\),使得
      \begin{equation*} \left. (\varphi_1\varphi)\right|_{U^\perp}=\varphi_2\cdots\varphi_s. \end{equation*}
现将\(\varphi_i(i=2,\cdots, s)\)定义范围扩大到\(V\)上,即补充定义\(\varphi_i(\xi_1)=\xi_1\)
\(\forall \alpha\in V\),存在唯一的\(\alpha_1=a \xi_1\in U, \alpha_2\in U^\perp\),使得\(\alpha=\alpha_1+\alpha_2\),注意到
\begin{equation*} \varphi_i(\alpha_1)=a\varphi_i(\xi_1)=a \xi_1=\alpha_1, \end{equation*}
所以
\begin{equation*} \begin{array}{ccl} \varphi_i(\alpha)&=&\varphi_i(\alpha_1)+\varphi_i(\alpha_2)=\alpha_1+\alpha_2-2(\xi_i,\alpha_2)\xi_i\\ &=&\alpha-2(\xi_i,\alpha_1+\alpha_2)\xi_i=\alpha-2(\xi_i,\alpha)\xi_i,\end{array} \end{equation*}
\(\varphi_i\)\(\xi_i\)\(V\)上确定的镜面反射。又
\begin{equation*} \begin{array}{ccl} \varphi_2\cdots\varphi_s(\alpha)&=&\varphi_2\cdots\varphi_s(\alpha_1)+\varphi_2\cdots\varphi_s(\alpha_2)=\alpha_1+\varphi_2\cdots\varphi_s(\alpha_2)\\&=&\varphi_1\varphi(\alpha_1)+\varphi_1\varphi(\alpha_2)=\varphi_1\varphi(\alpha),\end{array} \end{equation*}
所以\(\varphi_1\varphi=\varphi_2\cdots \varphi_s\)。注意到\(\varphi_1^2={\rm id}_V\),上式两边同时左乘\(\varphi_1\),得
\begin{equation*} \varphi=\varphi_1\varphi_2\cdots\varphi_s \end{equation*}
是一系列镜面反射的乘积。
证法三:因 \(\varphi\)\(n\)维欧氏空间 \(V\) 中正交变换,所以存在 \(V\)的一组标准正交基 \(\xi_1,\ldots ,\xi_n\),使得
\begin{equation*} \varphi(\xi_1,\ldots ,\xi_n)=(\xi_1,\ldots ,\xi_n)\begin{pmatrix} -E_s&&&&\\ &E_r&&&\\ &&R_1&&\\ &&&\ddots&\\ &&&&R_l\\ \end{pmatrix}, \end{equation*}
其中 \(R_j=\begin{pmatrix} \cos\theta_j&-\sin\theta_j\\ \sin\theta_j&\cos\theta_j \end{pmatrix},j=1,\ldots ,l\)。对任意 \(1\leq j\leq l\),记
\begin{equation*} M_j=\begin{pmatrix} \cos\theta_j&\sin\theta_j\\ \sin\theta_j&-\cos\theta_j \end{pmatrix},N_j=\begin{pmatrix} 1&0\\ 0&-1 \end{pmatrix}. \end{equation*}
定义 \(V\)上线性变换 \(\varphi_i,\psi_j,\sigma_j(i=1,\ldots ,s,j=1,\ldots ,l)\)满足
\begin{equation*} \varphi_i(\xi_1,\ldots ,\xi_n)=(\xi_1,\ldots ,\xi_n)E(i(-1)), \end{equation*}
\begin{equation*} \psi_j(\xi_1,\ldots ,\xi_n)=(\xi_1,\ldots ,\xi_n)\begin{pmatrix} E_{r+s}&&&&&\\ &E_2&&&&\\ &&\ddots&&&\\ &&&M_j&&\\ &&&&\ddots&\\ &&&&&E_2\\ \end{pmatrix}, \end{equation*}
\begin{equation*} \sigma_j(\xi_1,\ldots ,\xi_n)=(\xi_1,\ldots ,\xi_n)\begin{pmatrix} E_{r+s}&&&&&\\ &E_2&&&&\\ &&\ddots&&&\\ &&&N_j&&\\ &&&&\ddots&\\ &&&&&E_2\\ \end{pmatrix}, \end{equation*}
\begin{equation*} \varphi=\varphi_1\cdots\varphi_s\psi_1\sigma_1\cdots\psi_l\sigma_l. \end{equation*}
根据 练习 8.4.2.4\(\varphi_i,\psi_j,\sigma_j(i=1,\ldots ,s,j=1,\ldots ,l)\)都是镜面反射。因此\(\varphi=\varphi_1\cdots\varphi_s\psi_1\sigma_1\cdots\psi_l\sigma_l\)是一系列镜面反射的乘积。

10.

\(\alpha_1,\alpha_2,\cdots ,\alpha_m\)\(\beta_1,\beta_2,\cdots ,\beta_m\)\(n\)维欧氏空间\(V\)中的两个向量组,证明:存在\(V\)上的一个正交变换\(\varphi\),使得
\begin{equation*} \varphi(\alpha_i)=\beta_i,\ i=1,2,\cdots ,m \end{equation*}
的充分必要条件是
\begin{equation*} \left(\alpha_i,\alpha_j\right)=\left(\beta_i,\beta_j\right),\ i,j=1,2,\cdots ,m. \end{equation*}
解答.
必要性:因为\(\varphi\)是正交变换且\(\varphi(\alpha_i)=\beta_i\),所以
\begin{equation*} \left(\alpha_i,\alpha_j\right)=\left(\varphi(\beta_i),\varphi(\beta_j)\right)=\left(\beta_i,\beta_j\right),\ i,j=1,2,\cdots ,m. \end{equation*}
充分性:设\(\alpha_1,\cdots ,\alpha_r\)\(\alpha_1,\alpha_2,\cdots ,\alpha_m\)的一个极大无关组,则
\begin{equation*} \det G(\alpha_1,\cdots ,\alpha_r)\neq 0. \end{equation*}
\(\left(\alpha_i,\alpha_j\right)=\left(\beta_i,\beta_j\right)\)
\begin{equation*} \det G(\beta_1,\cdots ,\beta_r)=\det G(\alpha_1,\cdots ,\alpha_r)\neq 0, \end{equation*}
\(\beta_1,\cdots ,\beta_r\)线性无关。
\(\alpha_1,\cdots ,\alpha_r\)\(\alpha_1,\alpha_2,\cdots ,\alpha_m\)的一个极大无关组知,\(\forall 1\leq i\leq m\),存在\(a_{1i},\cdots ,a_{ri}\in\mathbb{R}\),使得\(\alpha_i=a_{1i}\alpha_1+\cdots +a_{ri}\alpha_r\),则
\begin{equation*} \left(\beta_i-\sum\limits_{j=1}^ra_{ji}\beta_j,\beta_i-\sum\limits_{j=1}^ra_{ji}\beta_j\right)=\left(\alpha_i-\sum\limits_{j=1}^ra_{ji}\alpha_j,\alpha_i-\sum\limits_{j=1}^ra_{ji}\alpha_j\right)=0, \end{equation*}
\(\beta_i=a_{1i}\beta_1+\cdots +a_{ri}\beta_r\),故\(\beta_1,\cdots,\beta_r\)是向量组\(\beta_1,\beta_2,\cdots ,\beta_m\)的一个极大无关组。令\(U_1=\langle \alpha_1,\cdots ,\alpha_m\rangle ,\ U_2=\langle \beta_1,\cdots ,\beta_m\rangle\),则\(\dim U_1=\dim U_2=r\)
定义\(U_1\)\(U_2\)的线性映射\(\varphi_1\),满足\(\varphi_1(\alpha_i)=\beta_i,\ i=1,2,\cdots ,r\),则\(\varphi_1\)是线性同构。由\(\left(\alpha_i,\alpha_j\right)=\left(\beta_i,\beta_j\right)\)知:\(\forall \alpha=\sum\limits_{k=1}^r a_k \alpha_k ,\beta=\sum\limits_{l=1}^r b_l \alpha_l\in U_1\),有
\begin{equation*} \begin{array}{ccl} (\varphi_1(\alpha),\varphi_1(\beta))&=&\left(\sum\limits_{k=1}^r a_k \varphi_1(\alpha_k),\sum\limits_{l=1}^r b_l \varphi_1(\alpha_l)\right)=\sum\limits_{k=1}^r\sum\limits_{l=1}^r a_kb_l\left(\varphi_1(\alpha_k),\varphi_1(\alpha_l)\right)\\ &=&\sum\limits_{k=1}^r\sum\limits_{l=1}^r a_kb_l\left(\beta_k,\beta_l\right)=\sum\limits_{k=1}^r\sum\limits_{l=1}^r a_kb_l\left(\alpha_k,\alpha_l\right)\\ &=&\left(\sum\limits_{k=1}^r a_k \alpha_k,\sum\limits_{l=1}^r b_l \alpha_l\right)=(\alpha,\beta) \end{array} \end{equation*}
\(\varphi_1\)是欧氏空间\(U_1\)\(U_2\)的同构映射。
又因为\(\dim U_1^\bot=\dim U_2^\bot=n-r\),所以存在欧氏空间同构映射
\begin{equation*} \varphi_2:U_1^\bot\rightarrow U_2^\bot \end{equation*}
注意到\(V=U_1\oplus U_1^\bot\),所以可定义\(V\)上的变换\(\varphi\)如下
\begin{equation*} \varphi(\alpha+\beta)=\varphi_1(\alpha)+\varphi_2(\beta),\ \forall \alpha\in U_1,\ \beta\in U_1^\bot . \end{equation*}
不难验证,\(\varphi\)\(V\)上的线性变换,下证\(\varphi\)保内积。对任意\(X,Y\in V\),存在唯一的\(X_1,Y_1\in U_1, X_2,Y_2\in U_1^\bot\),使得\(X=X_1+X_2,\ Y=Y_1+Y_2\),则
\begin{equation*} \begin{array}{ccl} \left(\varphi(X),\varphi(Y)\right)&=&\left(\varphi_1(X_1)+\varphi_2(X_2),\varphi_1(Y_1)+\varphi_2(Y_2)\right)\\ &=&\left(\varphi_1(X_1),\varphi_1(Y_1)\right)+\left(\varphi_1(X_1),\varphi_2(Y_2)\right)+\left(\varphi_2(X_2),\varphi_1(Y_1)\right)\\ &&+\left(\varphi_2(X_2),\varphi_2(Y_2)\right)\\ &=&\left(\varphi_1(X_1),\varphi_1(Y_1)\right)+\left(\varphi_2(X_2),\varphi_2(Y_2)\right)\\ &=&(X_1,Y_1)+(X_2,Y_2)\\ &=&(X_1,Y_1)+(X_1,Y_2)+(X_2,Y_1)+(X_2,Y_2)\\ &=&(X,Y), \end{array} \end{equation*}
\(\varphi\)保内积,进而\(\varphi\)是正交变换,且
\begin{equation*} \varphi(\alpha_i)=\varphi_1(\alpha_i)=\varphi_1(a_{1i}\alpha_1+\cdots +a_{ri}\alpha_r)=a_{1i}\beta_1+\cdots +a_{ri}\beta_r=\beta_i,\ i=1,2,\cdots ,m\mbox{。} \end{equation*}