美文网首页高等代数
高等代数理论基础79:若尔当标准形的几何理论(2)

高等代数理论基础79:若尔当标准形的几何理论(2)

作者: 溺于恐 | 来源:发表于2019-04-30 15:46 被阅读6次

    若尔当标准形的几何理论(2)

    定义:设\mathscr{A}\C上n维空间V上的一个线性变换,W是一个\mathscr{A}-不变子空间,若有\eta\in W,使W=P[\mathscr{A}]\eta,则称WV的一个\mathscr{A}-循环子空间

    注:定义对任一数域P有效

    引理:V,\mathscr{A}

    W=P[\mathscr{A}]\eta,\eta的最小多项式为p(\lambda),则\dim(W)=\partial(p(\lambda))

    证明:

    \partial(p(\lambda))=k,\forall w\in W

    f(\lambda)\in \C[\lambda],使w=f(\mathscr{A})\eta

    作带余除法

    f(\lambda)=q(\lambda)p(\lambda)+(l_0+l_1\lambda+\cdots+l_{k-1}\lambda^{k-1})

    w=f(\mathscr{A})\eta=q(\mathscr{A})p(\mathscr{A})\eta+(l_0\mathscr{E}+l_1\mathscr{A}+\cdots+l_{k-1}\mathscr{A}^{k-1})\eta

    =l_0\eta+l_1\mathscr{A}\eta+\cdots+l_{k-1}\mathscr{A}^{k-1}\eta

    \eta,\mathscr{A}\eta,\cdots,\mathscr{A}^{k-1}\eta的线性组合

    若有l_0\eta+l_1\mathscr{A}\eta+\cdots+l_{k-1}\mathscr{A}^{k-1}\eta=0

    (l_0\mathscr{E}+l_1\mathscr{A}+\cdots+l_{k-1}\mathscr{A}^{k-1})\eta=0

    p(\lambda)\eta的最小多项式,且为k

    l_0=\cdots=l_{k-1}=0

    \eta,\mathscr{A}\eta,\cdots,\mathscr{A}^{k-1}\eta线性无关,故为W的基

    \dim(W)=k=\partial(p(\lambda))\qquad\mathcal{Q.E.D}

    引理:V,\mathscr{A}

    若由\eta_1,\eta_2,\cdots,\eta_s\in V,使得

    1.V=P[\mathscr{A}]\eta_1+P[\mathscr{A}]\eta_2+\cdots+P[\mathscr{A}]\eta_s​

    2.设每个\eta_i\mathscr{A}的最小多项式为p_i(\lambda),且\sum\limits_{i=1}^s\partial(p_i(\lambda))=\dim(V)

    V=P[\mathscr{A}]\eta_1+P[\mathscr{A}]\eta_2+\cdots+P[\mathscr{A}]\eta_s​为直和

    证明:

    \dim(P[\mathscr{A}]\eta_i)=\partial(p_i(\lambda))

    \dim(V)=\sum\limits_{i=1}^s\partial(p_i(\lambda))=\sum\limits_{i=1}^s\dim(p[\mathscr{A}]\eta_i)

    \therefore V=P[\mathscr{A}]\eta_1+P[\mathscr{A}]\eta_2+\cdots+P[\mathscr{A}]\eta_s为直和\qquad\mathcal{Q.E.D}

    定理:V一定是一些\mathscr{A}-循环子空间的直和

    证明:

    \varepsilon_1,\varepsilon_2,\cdots,\varepsilon_nV的一组基,设\mathscr{A}\begin{pmatrix}\varepsilon_1\\\varepsilon_2\\\vdots\\\varepsilon_n\end{pmatrix}=A\begin{pmatrix}\varepsilon_1\\\varepsilon_2\\\vdots\\\varepsilon_n\end{pmatrix}

    (\mathscr{A}E-\mathscr{E}A)\begin{pmatrix}\varepsilon_1\\\varepsilon_2\\\vdots\\\varepsilon_n\end{pmatrix}=O

    \lambda E-A​,有可逆的P(\lambda),Q(\lambda)​,使得

    P(\lambda)(\lambda E-A)Q(\lambda)=\begin{pmatrix}h_1(\lambda)\\&h_2(\lambda)\\& &\ddots\\& & &h_n(\lambda)\end{pmatrix}​是对角形

    h_1(\lambda),\cdots,h_n(\lambda)的首项系数为1

    O=P(\mathscr{A})(\mathscr{A}E-\mathscr{E}A)\begin{pmatrix}\varepsilon_1\\\varepsilon_2\\\vdots\\\varepsilon_n\end{pmatrix}​

    =[P(\mathscr{A})(\mathscr{A}E-\mathscr{E}A)Q(\mathscr{A})]Q^{-1}(\mathscr{A})\begin{pmatrix}\varepsilon_1\\\varepsilon_2\\\vdots\\\varepsilon_n\end{pmatrix}​

    =\begin{pmatrix}h_1(\lambda)\\&h_2(\lambda)\\& &\ddots\\& & &h_n(\lambda)\end{pmatrix}\begin{pmatrix}\eta_1\\\eta_2\\\vdots\\\eta_n\end{pmatrix}​

    其中\begin{pmatrix}\eta_1\\\eta_2\\\vdots\\\eta_n\end{pmatrix}=Q^{-1}(\mathscr{A})\begin{pmatrix}\varepsilon_1\\\varepsilon_2\\\vdots\\\varepsilon_n\end{pmatrix}​

    h_i(\mathscr{A})\eta_i=0,h_i(\lambda)\eta_i的零化多项式,i=1,2,\cdots,n

    p_i(\lambda)\eta_i的最小多项式,则\partial(p_i(\lambda))\le \partial(h_i(\lambda))

    且有p_i(\lambda)|h_i(\lambda)

    \sum\limits_{i=1}^n\partial(h_i(\lambda))=\partial(|\lambda E-A|)=\dim(V)​

    \begin{pmatrix}\varepsilon_1\\\varepsilon_2\\\vdots\\\varepsilon_n\end{pmatrix}=Q(\mathscr{A})\begin{pmatrix}\eta_1\\\eta_2\\\vdots\\\eta_n\end{pmatrix}

    =\begin{pmatrix}\sum\limits_{j=1}^nq_{1j}(\mathscr{A})\eta_j\\\sum\limits_{j=1}^nq_{2j}(\mathscr{A})\eta_j\\\vdots\\\sum\limits_{j=1}^nq_{nj}(\mathscr{A})\eta_j\end{pmatrix}

    其中Q(\lambda)=(q_{ij}(\lambda))

    \varepsilon_i=\sum\limits_{j=1}^nq_{ij}(\mathscr{A})\eta_j\in P[\mathscr{A}]\eta_1+\cdots+P[\mathscr{A}]\eta_n

    易知V=P[\mathscr{A}]\eta_1+\cdots+P[\mathscr{A}]\eta_n

    \dim(V)\le \sum\limits_{i=1}^n\dim(P[\mathscr{A}]\eta_i)

    \dim(V)\le \sum\limits_{i=1}^m\dim(P[\mathscr{A}]\eta_i)=\sum\limits_{i=1}^n\partial(p_i(\lambda))

    \le\sum\limits_{i=1}^n\partial(h_i(\lambda))=\dim(V)

    上式成立,

    当所有等号成立即证

    \dim(V)=\sum\limits_{i=1}^n\dim(P[\mathscr{A}]\eta_i)=\sum\limits_{i=1}^n\partial(p_i(\lambda))​

    V=P[\mathscr{A}]\eta_1\oplus P[\mathscr{A}]\eta_2\oplus \cdots\oplus P[\mathscr{A}]\eta_n\qquad\mathcal{Q.E.D}

    注:

    1.可得p_i(\lambda)=h_i(\lambda),i=1,2,\cdots,n

    h_i(\lambda)\eta_i的最小多项式

    已证\sum\limits_{i=1}^n\partial(p_i(\lambda))=\sum\limits_{i=1}^n\partial(h_i(\lambda))

    \partial(p_i(\lambda))\le \partial(h_i(\lambda)),i=1,2,\cdots,n

    可得\partial(p_i(\lambda))=\partial(h_i(\lambda)),i=1,2,\cdots,n

    p_i(\lambda)|h_i(\lambda)都为首一多项式

    h_i(\lambda)=p_i(\lambda)

    h_i(\lambda)\eta_i的最小多项式

    2.若某h_i(\lambda)=1,则h_i(\mathscr{A})\eta_i=\mathscr{E}\eta_i=0

    \eta_i=0

    \eta_1,\cdots,\eta_n中去掉\eta_i=0

    将剩下的\eta_i重新编号,仍记作\eta_1,\eta_2,\cdots,\eta_s

    V=P[\mathscr{A}]\eta_1\oplus \cdots\oplus P[\mathscr{A}]\eta_s

    且各\eta_i的最小多项式次数\ge 1

    引理:设V=P[\mathscr{A}]\eta,\eta的最小多项式为h(\lambda)=(\lambda-\mu_1)^{l_1}(\lambda-\mu_2)^{l_2}\cdots(\lambda-\mu_t)^{l_t},\mu_i互不相同,则有\xi_1,\xi_2,\cdots,\xi_t\in V,使V=P[\mathscr{A}]\xi_1\oplus P[\mathscr{A}]\xi_2\oplus \cdots \oplus P[\mathscr{A}]\xi_t,且\xi_i对于\mathscr{A}的最小多项式是(\lambda-\mu_i)^{l_i}

    证明:

    m_i(\lambda)={h(\lambda)\over (\lambda-\mu_i)^{l_i}}

    \xi_i=m_i(\mathscr{A})\eta,1\le i\lt t

    易知\xi_i的最小多项式为(\lambda-\mu_i)^{l_i}

    m_1(\lambda),\cdots,m_t(\lambda)互素

    u_1(\lambda),u_2(\lambda),\cdots,u_t(\lambda)使得

    u_1(\lambda)m_1(\lambda)+u_2(\lambda)m_2(\lambda)+\cdots+u_t(\lambda)m_t(\lambda)=1

    \eta=u_1(\mathscr{A})m_1(\mathscr{A})\eta+u_2(\mathscr{A})m_2(\mathscr{A})\eta+\cdots+u_t(\mathscr{A})m_t(\mathscr{A})\eta

    =u_1(\mathscr{A})\xi_1+u_2(\mathscr{A})\xi_2+\cdots+u_t(\mathscr{A})\xi_t\in P[\mathscr{A}]\xi_1+\cdots+P[\mathscr{A}]\xi_t

    V=P[\mathscr{A}]\eta=P[\mathscr{A}]\xi_1+P[\mathscr{A}]\xi_2+\cdots+P[\mathscr{A}]\xi_t

    \partial(h(t))=\dim(P[\mathscr{A}]\eta)

    \sum\limits_{i=1}^t\dim(P[\mathscr{A}]\xi_i)=\sum\limits_{i=1}^tl_i

    =\partial(h(t))=\dim(P[\mathscr{A}]\eta)=\dim(V)

    V=P[\mathscr{A}]\xi_1\oplus P[\mathscr{A}]\xi_2\oplus P[\mathscr{A}]\xi_t\qquad\mathcal{Q.E.D}

    定理:V,\mathscr{A}​,则有\alpha_1,\alpha_2,\cdots,\alpha_t\in V​,使V=P[\mathscr{A}]\alpha_1\oplus \cdots\oplus P[\mathscr{A}]\alpha_s​,且\alpha_i​\mathscr{A}​的最小多项式为(\lambda-\lambda_i)^{k_i},k_i\ge 1​

    证明:

    \eta_1,\eta_2,\cdots,\eta_s\in V,使V=P[\mathscr{A}]\eta_1\oplus\cdots\oplus P[\mathscr{A}]\eta_s

    可将每个P[\mathscr{A}]\eta_i继续分解,直到满足要求

    故最后有\alpha_1,\alpha_2,\cdots,\alpha_t\in V,使V有分解式

    V=P[\mathscr{A}]\alpha_1\oplus \cdots\oplus P[\mathscr{A}]\alpha_s

    \alpha_i​\mathscr{A}​的最小多项式为(\lambda-\lambda_i)^{k_i},k_i\ge 1\qquad\mathcal{Q.E.D}​

    定理:V,\mathscr{A},则V中有基,使\mathscr{A}在该组基下的矩阵为若尔当标准形,且除去各若尔当块的排列顺序外,若尔当标准形由\mathscr{A}唯一确定

    相关文章

      网友评论

        本文标题:高等代数理论基础79:若尔当标准形的几何理论(2)

        本文链接:https://www.haomeiwen.com/subject/lixowqtx.html