节 4.6 不变子空间
建设中!
子节 4.6.1
- \(\phi\)与\(\phi|_U\)的相同点是在\(U\)上对应法则一样;不同点是\(\phi\)是\(V\)的线性变换;而\(\phi|_U\)是\(U\)的线性变换。
- 定义中\(U\)是\(\phi\)的不变子空间这个条件不可少,否则无法导出\(U\)上的线性变换。
设\(\phi\in\mathcal{L}(V)\),\(U\)是\(\phi\)的不变子空间。设\(\varepsilon_1,\varepsilon_2,\ldots,\varepsilon_r\)是\(U\)的一个基,将其扩为\(V\) 的一个基\(\varepsilon_1,\varepsilon_2,\ldots,\varepsilon_r,\varepsilon_{r+1},\ldots,\varepsilon_n\),则\(\phi\)在此基下的矩阵是
\begin{equation}
\begin{pmatrix}
a_{1,1} & \cdots & a_{1,r} & a_{1,r+1} & \cdots & a_{1,n}\\
\vdots &\ddots & \vdots & \vdots & \ddots & \vdots\\
a_{r,1} & \cdots & a_{r,r} & a_{r,r+1} & \cdots & a_{r,n}\\
0 & \cdots & 0 & a_{r+1,r+1} & \cdots & a_{r+1,n}\\
\vdots &\ddots & \vdots & \vdots & \ddots & \vdots\\
0 & \cdots & 0 & a_{n,r+1} & \cdots & a_{n,n}\\
\end{pmatrix}.\tag{4.7}
\end{equation}
反之,若\(\phi\)在基\(\varepsilon_1,\varepsilon_2,\ldots,\varepsilon_r,\varepsilon_{r+1},\ldots,\varepsilon_n\)下的矩阵为(4.7),则\(\langle \varepsilon_1,\varepsilon_2,\ldots,\varepsilon_r\rangle\)是一个\(\phi\)-子空间。
设\(\phi\)是\(n\)维线性空间\(V\)的线性变换,\(V=V_1\oplus V_2\),\(V_1\)与\(V_2\)是\(\phi\)-子空间,\(\varepsilon_1,\varepsilon_2,\ldots,\varepsilon_r\)是\(V_1\)的基,\(\varepsilon_{r+1},\varepsilon_{r+2},\ldots,\varepsilon_n\)是\(V_2\)的基,则在基\(\varepsilon_1,\varepsilon_2,\ldots,\varepsilon_r,\varepsilon_{r+1},\ldots,\varepsilon_n\)下的矩阵是
\begin{equation}
\begin{pmatrix}
A_1 & 0\\
0 & A_2
\end{pmatrix}\tag{4.8}
\end{equation}
其中\(A_1\)是\(r\)阶方阵,\(A_2\)是\(n-r\)阶方阵。
反之,若\(\phi\)在基\(\varepsilon_1,\varepsilon_2,\ldots,\varepsilon_r,\varepsilon_{r+1},\ldots,\varepsilon_n\)下的矩阵是(4.8),令
\begin{equation*}
V_1=\langle \varepsilon_1,\varepsilon_2,\ldots,\varepsilon_r\rangle,\ \ V_2=\langle \varepsilon_{r+1},\varepsilon_{r+2},\ldots,\varepsilon_n\rangle ,
\end{equation*}
则\(V_1\)、\(V_2\)都是\(\phi\)-子空间,且
\begin{equation*}
V = V_1\oplus V_2.
\end{equation*}
定理 4.6.2.
设\(\phi\in\mathcal{L} (V)\),则\(\phi\)在\(V\)的某个基下矩阵是块对角矩阵\(\Leftrightarrow\)\(V\)可分解为一些\(\phi\)的不变子空间的直和。
练习 4.6.2 练习
1.
设\(\varphi\)是数域\(\mathbb{F}\)上\(n\)维线性空间\(V\)上的线性变换,取定\(\lambda\in\mathbb{F}\),记
\begin{equation*}
V_ \lambda=\{\alpha\in V\ |\ \varphi (\alpha)=\lambda \alpha\},
\end{equation*}
证明:\(V_ \lambda\)是\(\varphi\)-不变子空间。
解答.
因
\begin{equation*}
\varphi(0_V)=0_V=\lambda 0_V,
\end{equation*}
所以\(0_V\in V_\lambda\),\(V_\lambda\)是\(V\)的非空子集。对任意\(\alpha,\beta\in V_\lambda,a,b\in\mathbb{F}\),有
\begin{equation*}
\varphi (\alpha)=\lambda \alpha,\varphi(\beta)=\lambda \beta,
\end{equation*}
则
\begin{equation*}
\varphi (a \alpha+b \beta)=a\varphi(\alpha)+b\varphi(\beta)=a(\lambda \alpha)+b(\lambda \beta)=\lambda(a \alpha+b \beta)\mbox{。}
\end{equation*}
因此\(V_\lambda\)是\(V\)的子空间。又
\begin{equation*}
\varphi (\alpha)=\lambda \alpha\in V_\lambda,
\end{equation*}
故\(V_ \lambda\)是\(\varphi\)-不变子空间。
2.
设\(V\)是4维线性空间,\(V\)上线性变换\(\varphi\)在基\(\xi_1,\xi_2,\xi_3,\xi_4\)下的矩阵为
\begin{equation*}
\begin{pmatrix}
1&0&2&-1\\
0&1&4&-2\\
2&-1&0&1\\
2&-1&-1&2
\end{pmatrix},
\end{equation*}
- 证明:\(U=\langle \xi_1+2 \xi_2, \xi_2+\xi_3+2 \xi_4\rangle\)是\(\varphi\)-不变子空间;
- 求\(\varphi|_U\)在基\(\xi_1+2 \xi_2, \xi_2+\xi_3+2 \xi_4\)下的矩阵。
解答.
- 由已知条件知\(\left\{\begin{array}{l} \varphi (\xi_1)=\xi_1+2 \xi_3+2 \xi_4,\\ \varphi (\xi_2)=\xi_2- \xi_3- \xi_4,\\ \varphi (\xi_3)=2\xi_1+4 \xi_2- \xi_4,\\ \varphi (\xi_4)=-\xi_1-2 \xi_2+\xi_3+2 \xi_4,\\ \end{array}\right.\) 则\begin{equation*} \begin{array}{c}\varphi (\xi_1+2 \xi_2)=\varphi (\xi_1)+2 \varphi(\xi_2)=\xi_1+2 \xi_2\in U,\\ \varphi (\xi_2+ \xi_3+2 \xi_4)=\varphi (\xi_2)+\varphi (\xi_3)+2 \varphi(\xi_4)=\xi_2+\xi_3+2 \xi_4\in U,\end{array} \end{equation*}故\(U\)是\(\varphi\)-子空间。
- 因为\begin{equation*} \begin{array}{c} \varphi|_U(\xi_1+2 \xi_2)=\varphi(\xi_1+2 \xi_2)=\xi_1+2 \xi_2,\\ \varphi|_U(\xi_2+ \xi_3+2 \xi_4)=\varphi(\xi_2+\xi_3+2 \xi_4)=\xi_2+\xi_3+2 \xi_4, \end{array} \end{equation*}即\begin{equation*} \varphi(\xi_1+2 \xi_2,\xi_2+\xi_3+2 \xi_4)=(\xi_1+2 \xi_2,\xi_2+\xi_3+2 \xi_4)\begin{pmatrix} 1&0\\0&1 \end{pmatrix}, \end{equation*}所以\(\varphi|_U\)在基\(\xi_1+2 \xi_2, \xi_2+\xi_3+2 \xi_4\)下的矩阵为\(\begin{pmatrix} 1&0\\0&1 \end{pmatrix}\)。
3.
设\(\varphi ,\psi\)是\(n\)维线性空间\(V\)上线性变换,
- 若\(\varphi\psi=\psi\varphi\),证明:\({\rm Ker}\varphi\)与\({\rm Im}\varphi\)都是\(\psi\)-不变子空间;
- 若\(\varphi^2=\varphi\),证明:\({\rm Ker}\varphi\)与\({\rm Im}\varphi\)都是\(\psi\)-不变子空间的充分必要条件是\(\varphi\psi=\psi\varphi\)。
解答.
-
对任意\(\alpha\in{\rm Ker}\varphi\), 有\(\varphi(\alpha)=0\)。因为\(\varphi\psi =\psi\varphi\),所以\begin{equation*} \varphi(\psi(\alpha))=\psi(\varphi (\alpha))=\psi (0)=0, \end{equation*}即\(\psi (\alpha)\in{\rm Ker}\varphi\)。因此,\({\rm Ker}\varphi\)是\(\psi\)-子空间。对任意\(\beta\in{\rm Im}\varphi\),存在\(\alpha\in V\),使得\(\beta=\varphi(\alpha)\)。因为\(\varphi\psi=\psi\varphi\),所以\begin{equation*} \psi(\beta)=\psi\varphi(\alpha)=\varphi(\psi(\alpha))\in{\rm Im}\varphi, \end{equation*}故\({\rm Im}\varphi\)是\(\psi\)-子空间。
-
由\((1)\),充分性成立。下证必要性。对任意\(\alpha\in V\),有\(\varphi^2(\alpha)-\varphi(\alpha)=0\),即\(\varphi(\alpha)-\alpha\in {\rm Ker}\varphi\)。因为\({\rm Ker}\varphi\)是\(\psi\)-不变子空间,所以\(\psi(\varphi(\alpha)-\alpha)\in{\rm Ker}\varphi\),即\(\varphi\left(\psi(\varphi(\alpha)-\alpha)\right)=0\),故\begin{equation} \varphi\psi(\alpha)=\varphi\psi\varphi(\alpha).\tag{4.9} \end{equation}注意到\({\rm Im}\varphi\)是\(\psi\)-不变子空间,且\(\varphi(\alpha)\in{\rm Im}\varphi\),所以\(\psi \left(\varphi(\alpha)\right)\in {\rm Im}\varphi\),即存在\(\beta\in V\)使得\(\psi \left(\varphi(\alpha)\right)=\varphi(\beta)\)。于是,\begin{equation} \varphi\psi\varphi(\alpha)=\varphi\left(\varphi(\beta)\right)=\varphi^2(\beta)=\varphi(\beta)=\psi \left(\varphi(\alpha)\right).\tag{4.10} \end{equation}
4.
设\(\varphi:\mathbb{F}^2\rightarrow\mathbb{F}^2,\ (a,b)^T\mapsto (b,a)^T\),试求所有非平凡的\(\varphi\)-不变子空间。
解答.
设\(U\)是非平凡的\(\varphi\)-不变子空间,则\(\dim U=1\)。设\((a,b)^T\)是\(U\)的基,则\(\varphi((a,b)^T)=(b,a)^T\in U\),即存在\(k\in\mathbb{F}\)使得\((b,a)^T=k(a,b)^T\),故\(b=\pm a\)。因此\(U=\langle (1,1)\rangle\)或\(\langle (1,-1)\rangle\)。
5.
设\(\varphi\)是\(n\)维线性空间\(V\)上的线性变换,\(\varphi\)在基\(\xi_1,\xi_2,\cdots ,\xi_n\)下的矩阵是
\begin{equation*}
\begin{pmatrix}
a&0&0&\cdots&0&0\\
1&a&0&\cdots&0&0\\
0&1&a&\cdots&0&0\\
\vdots&\vdots&\vdots&&\vdots&\vdots\\
0&0&0&\cdots&a&0\\
0&0&0&\cdots&1&a
\end{pmatrix},
\end{equation*}
证明:
- 设\(U\)是\(\varphi\)-子空间,且\(\xi_1\in U\),则\(U=V\);
- 对于任意非零\(\varphi\)-子空间\(U\),总有\(\xi_n\in U\);
- \(V\)不能分解为两个非平凡的\(\varphi\)-子空间的直和;
- 求\(\varphi\)的所有不变子空间。
解答.
- 依题意,\begin{equation*} \varphi(\xi_1)=a\xi_1+\xi_2,\varphi(\xi_2)=a\xi_2+\xi_3,\cdots ,\varphi(\xi_{n-1})=a\xi_{n-1}+\xi_n,\varphi(\xi_n)=a\xi_n. \end{equation*}因\(U\)是\(V\)的\(\varphi\)-子空间,且\(\xi_1\in U\),所以\(\xi_2=\varphi(\xi_1)-a\xi_1\in U\)。于是,\quad \(\xi_3=\varphi(\xi_2)-a\xi_2\in U\)。依此类推,得\(\xi_1,\xi_2,\cdots ,\xi_n\in U\)。故\(U=V\)。
- 因为\(U\)是\(V\)的非零\(\varphi\)-子空间,所以存在\(\alpha\in U\),满足\(\alpha\neq 0\)。由\(0\neq\alpha\in V\),可设\(\alpha=a_1\xi_1+a_2\xi_2+\cdots +a_n\xi_n\),这里\(a_1,a_2,\cdots ,a_n\)不全为零。假设\(a_i\)是\(a_1,a_2,\cdots ,a_n\)中第一个不为零的数,此时\begin{equation*} \alpha=a_i\xi_i+a_{i+1}\xi_{i+1}+\cdots +a_n\xi_n\in U. \end{equation*}根据\(\varphi(\alpha)-a \alpha\in U\)得\begin{equation*} a_i \xi_{i+1}+a_{i+1}+\xi_{i+2}+\cdots +a_{n-1}\xi_n\in U. \end{equation*}记 \(\beta=a_i \xi_{i+1}+a_{i+1}+\xi_{i+2}+\cdots +a_{n-1}\xi_n\),根据\(\varphi(\beta)-a \beta\in U\)得\begin{equation*} a_i\xi_{i+2}+a_{i+1}\xi_{i+3}+\cdots +a_{n-2}\xi_n\in U. \end{equation*}依此类推,我们有\(a_i\xi_n\in U\)。因为\(a_i\neq 0\),所以\(\xi_n=\frac{1}{a_i}(a_i \xi_n)\in U\)。
- 由\((2)\),任意两个非平凡\(\varphi\)-子空间\(V_1,V_2\)必含\(\xi_n\),故\(V_1\bigcap V_2\neq 0\)。从而\(V\)不能分解成两个非平凡\(\varphi\)-不变子空间的直和。
- 依题意知,子空间\(0,U_i=\langle \xi_i,\xi_{i+1},\cdots ,\xi_n\rangle\)是\(\varphi\)-不变子空间。下证\(V\)的\(\varphi\)-不变子空间有且只有以上这些。设\(U\neq 0\)是\(V\)的\(\varphi\)-不变子空间,记\begin{equation*} i_0=\min\{i\ |\ a_i\neq 0\mbox{且}a_i \xi_i+a_{i+1} \xi_{i+1}+\cdots +a_n \xi_n\in U\}, \end{equation*}则\(U\subseteq\langle \xi_{i_0},\xi_{i_0+1},\cdots ,\xi_n\rangle\)。我们断言,\(U=\langle \xi_{i_0},\xi_{i_0+1},\cdots ,\xi_n\rangle\)。事实上,由\(i_0\)的定义知:存在\(a_{i_0},a_{i_0+1},\cdots ,a_n\in\mathbb{F}\)使得\(a_{i_0}\xi_{i_0}+a_{i_0+1}\xi_{i_0+1}+\cdots +a_n \xi_n\in U\)。由\((2)\)知\(\xi_n\in U\),所以\begin{equation*} (a_{i_0}\xi_{i_0}+a_{i_0+1}\xi_{i_0+1}+\cdots +a_n \xi_n)-a_n \xi_n\in U. \end{equation*}记\(\beta_1\triangleq a_{i_0}\xi_{i_0}+a_{i_0+1}\xi_{i_0+1}+\cdots +a_{n-1} \xi_{n-1}\),则\(\beta_1\in U\)。于是,\begin{equation*} \varphi(\beta_1)-a \beta_1-a_{n-1}\xi_n\in U, \end{equation*}即\begin{equation*} \gamma_1\triangleq a_{i_0}\xi_{i_0+1}+a_{i_0+1}\xi_{i_0+2}+\cdots +a_{n-2} \xi_{n-1}\in U, \end{equation*}再根据\(\varphi(\gamma_1)-a \gamma_1-a_{n-2}\xi_n\in U\)得\begin{equation*} \gamma_2\triangleq a_{i_0}\xi_{i_0+2}+a_{i_0+1}\xi_{i_0+3}+\cdots +a_{n-3} \xi_{n-1}\in U, \end{equation*}依此类推,有\(a_{i_0}\xi_{n-1}\in U\)。由\(a_{i_0}\neq 0\)可知\(\xi_{n-1}\in U\)。于是,\begin{equation*} \beta_2\triangleq\beta_1-a_{n-1}\xi_{n-1}= a_{i_0} \xi_{i_0}+a_{i_0+1}\xi_{i_0+1}+\cdots +a_{n-2}\xi_{n-2}\in U. \end{equation*}同理,我们有\begin{equation*} \varphi(\beta_2)-a\beta_2-a_{n-2}\xi_{n-1}\in U, \end{equation*}即\begin{equation*} \eta_1\triangleq a_{i_0}\xi_{i_0+1}+a_{i_0+1}\xi_{i_0+2}+\cdots +a_{n-3}\xi_{n-2}\in U. \end{equation*}再由\(\varphi(\eta_1)-a\eta_1-a_{n-3}\xi_{n-1}\in U\),得\begin{equation*} \eta_2\triangleq a_{i_0}\xi_{i_0+2}+a_{i_0+1}\xi_{i_0+3}+\cdots +a_{n-4}\xi_{n-3}\in U. \end{equation*}依此类推,\(a_{i_0}\xi_{n-2}\in U\)。由\(a_{i_0}\neq 0\)可知\(\xi_{n-2}\in U\)。于是,\begin{equation*} \beta_3\triangleq\beta_2-a_{n-2}\xi_{n-2}= a_{i_0} \xi_{i_0}+a_{i_0+1}\xi_{i_0+1}+\cdots +a_{n-3}\xi_{n-3}\in U. \end{equation*}重复上述步骤,最后有\(a_{i_0}\xi_{i_0}\in U\)。由\(a_{i_0}\neq 0\)得\(\xi_{i_0}\in U\)。因此,\begin{equation*} U=\langle \xi_{i_0},\xi_{i_0+1},\cdots ,\xi_n\rangle . \end{equation*}