2020年10月5日 星期一

國立臺灣大學一百零九學年度轉學生入學考試試題詳解

※注意:請於試卷上「非選擇題作答區」標明大題及小題題號,並依序作答。

  1. Let $\left\{e_1,e_2,e_3\right\}$ be the standard basis for $\mathbb{R}^3$. Suppose that a linear transformation $T:\mathbb{R}^3\mapsto\mathbb{R}^3$ is defined by $T\left(x,y,z\right)=\left(2x+y,2y+z,2z\right)$.
    1. Write down the matrix of $T$ relative to the standard basis. (5 points.)
    2. Write down the matrix of $T$ relative to the ordered basis $\left\{e_3,e_2,e_1\right\}$. (5 points.)
    3. Find a matrix $P$ such that

      $\displaystyle P^{-1}\begin{pmatrix}a&1&0\\0&a&1\\0&0&a\end{pmatrix}P=\begin{pmatrix}a&0&0\\1&a&0\\0&1&a\end{pmatrix}$

      for all real numbers $a$. (5 points.)
    4. Prove that for any given $n\times n$ matrix $A$, there is a matrix $Q$ such that

      $Q^{-1}AQ=A^t$.

      (5 points.)
    5. Let

      $A=\begin{pmatrix}1&-1&0\\1&2&1\\0&1&1\end{pmatrix}$.

      Find a matrix $Q$ such that $Q^{-1}AQ=A^t$. (5 points.)
  2. 訣竅前兩小題按照矩陣表示法的定義求解;第三小題可由 Jordan塊矩陣的轉置相似概念求解;第四小題使用伴矩陣來求證(佈於實數中)或使用Jordan型(佈於複數中);第五題則執行前兩小題所描述的過程。
    解法
    1. 直接計算可知

      $\begin{aligned}&T\left(e_1\right)=\left(2,0,0\right)=2e_1+0e_2+0e_3,\\&T\left(e_2\right)=\left(1,2,0\right)=1e_1+2e_2+0e_3,\\&T\left(e_3\right)=\left(0,1,2\right)=0e_1+1e_2+2e_3.\end{aligned}$

      故相對於標準基底的矩陣表示法為

      $\begin{bmatrix}2&1&0\\0&2&1\\0&0&2\end{bmatrix}$

    2. 直接演算可知

      $\begin{aligned}&T\left(e_3\right)=\left(0,1,2\right)=2e_3+1e_2+0e_3,\\&T\left(e_2\right)=\left(1,2,0\right)=0e_3+2e_2+1e_1,\\&T\left(e_1\right)=\left(2,0,0\right)=0e_3+0e_2+2e_1.\end{aligned}$

      故相對於 $\left\{e_3,e_2,e_1\right\}$ 的矩陣表示法為

      $\begin{bmatrix}2&0&0\\1&2&0\\0&1&2\end{bmatrix}$

    3. 取 $P=\begin{bmatrix}0&0&1\\0&1&0\\1&0&0\end{bmatrix}$,那麼 $P^{-1}=P$ 且容易檢驗知

      $P^{-1}\begin{bmatrix}a&1&0\\0&a&1\\0&0&a\end{bmatrix}P=\begin{bmatrix}0&0&1\\0&1&0\\1&0&0\end{bmatrix}\begin{bmatrix}a&1&0\\0&a&1\\0&0&a\end{bmatrix}\begin{bmatrix}0&0&1\\0&1&0\\1&0&0\end{bmatrix}=\begin{bmatrix}0&0&a\\0&a&1\\a&1&0\end{bmatrix}\begin{bmatrix}0&0&1\\0&1&0\\1&0&0\end{bmatrix}=\begin{bmatrix}a&0&0\\1&a&0\\0&1&a\end{bmatrix}$.

    4. 【方法一】 假若 $A$ 佈於複數體中,那麼每個矩陣相似一 Jordan 型式,即存在可逆矩陣 $P$ 與 Jordan 型矩陣 $J_A$ 使 $P^{-1}AP=J_A$,其中

      $J_A=\begin{bmatrix}J_1&0&\cdots&0&0\\0&J_2&\cdots&0&0\\\vdots&\vdots&\ddots&\vdots&\vdots\\0&0&\cdots&J_{k-1}&0\\0&0&\cdots&0&J_k\end{bmatrix}$

      此處 $J_A$ 為 $d_i$ 階方陣,而存在特徵值 $c_i$ 使 $J_i-c_i$ 為冪零矩陣。再者定義 $K_m$ 為 $m$ 階方陣如下

      $K_m=\begin{bmatrix}0&0&\cdots&0&1\\0&0&\cdots&1&0\\\vdots&\vdots&\ddots&\vdots&\vdots\\0&1&\cdots&0&0\\1&0&\cdots&0&0\end{bmatrix}$

      則容易知道 $K_{d_i}^{-1}J_iK_{d_i}=\left(J_i\right)^t$ 以及 $K_{d_i}^{-1}=K_{d_i}$。由此能知

      $\left(J_A\right)^t=\begin{bmatrix}\left(J_1\right)^t&\cdots&0\\\vdots&\ddots&\vdots\\0&\cdots&\left(J_k\right)^t\end{bmatrix}=\begin{bmatrix}K_{d_1}&\cdots&0\\\vdots&\ddots&\vdots\\0&\cdots&K_{d_k}\end{bmatrix}\begin{bmatrix}J_1&\cdots&0\\\vdots&\ddots&\vdots\\0&\cdots&J_k\end{bmatrix}\begin{bmatrix}K_{d_1}&\cdots&0\\\vdots&\ddots&\vdots\\0&\cdots&K_{d_k}\end{bmatrix}$

      因此

      $A\sim J_A\sim\left(J_A\right)^t\sim A^t$.


      【方法二】 假若矩陣僅佈於實數體或有理數體(或其他抽象的體)時。我們先觀察到這樣的事實:假若存在矩陣相似於 $A$ 且該矩陣 $B$ 相似於自身的轉置,則該矩陣 $A$ 也相似於自身的轉置。這是因為由條件中的兩個相似可知有兩個可逆矩陣 $P$ 與 $Q$ 滿足 $B=PAP^{-1}$ 與 $B^t=QBQ^{-1}$,如此可看出

      $A^t=\left(P^tQP\right)A\left(P^tQP\right)^{-1}$

      又因每個矩陣都與自己的有理型式(rational form)相似,故我們僅需對於有理型式的矩陣證明即可。再者,我們容易看出這樣的證明只需應用在每一個塊狀即可,即我們可不失一般性的假設 $A$ 是某個多項式 $f$ 的伴矩陣(即該多項式同時為矩陣 $A$ 的特徵多項式與極小多項式)。又容易知道伴矩陣的轉置有相同的特徵多項式與極小多項式,故兩者具有相同的有理型式,故 $A^t$ 與 $A$ 相似,證明完畢。
    5. 【方法一】先計算給定矩陣的特徵多項式為

      $\text{ch}_A\left(x\right)=\det\left(xI-A\right)=\begin{vmatrix}x-1&1&0\\-1&x-2&-1\\0&-1&x-1\end{vmatrix}=\left(x-1\right)^2\left(x-2\right)$

      故特徵值為 $1$ 與 $2$。分別計算特徵向量如下
      • 當 $x=1$ 時有 $A-I=\begin{bmatrix}0&-1&0\\1&1&1\\0&1&0\end{bmatrix}$,故取 $v_1=\begin{bmatrix}1\\0\\-1\end{bmatrix}$ 滿足 $Av_1=v_1$。又進一步可取 $v_2=\begin{bmatrix}1\\-1\\0\end{bmatrix}$ 能滿足 $\left(A-I\right)v_2=v_1$。
      • 當 $x=2$ 時有 $A-2I=\begin{bmatrix}-1&-1&0\\1&0&1\\0&1&-1\end{bmatrix}$,故取 $v_3=\begin{bmatrix}-1\\1\\1\end{bmatrix}$ 能滿足 $Av_3=2v_3$。
      據此取 $P=\begin{bmatrix}v_1&v_2&v_3\end{bmatrix}=\begin{bmatrix}1&1&-1\\0&-1&1\\-1&0&1\end{bmatrix}$ 有

      $P^{-1}AP=J:=\begin{bmatrix}1&1&0\\0&1&0\\0&0&2\end{bmatrix}$.

      又可以取 $R=\begin{bmatrix}v_2&v_1&v_3\end{bmatrix}=\begin{bmatrix}1&1&-1\\-1&0&1\\0&-1&1\end{bmatrix}$ 便有

      $R^{-1}AR=J^t=\begin{bmatrix}1&0&0\\1&1&0\\0&0&2\end{bmatrix}$

      同取轉置便有

      $P^{-1}AP=J=\left(R^{-1}AR\right)^t=R^tA^t\left(R^t\right)^{-1}$

      即得

      $\left(PR^t\right)^{-1}A(PR^t)=A^t$

      故取 $Q=PR^t$,即

      $Q=\begin{bmatrix}1&1&-1\\0&-1&1\\-1&0&1\end{bmatrix}\begin{bmatrix}1&-1&0\\1&0&-1\\-1&1&1\end{bmatrix}=\begin{bmatrix}3&-2&-2\\-2&1&2\\-2&2&1\end{bmatrix}$.


      【方法二】 承方法一可知計算給定矩陣的特徵多項式有

      $\text{ch}_A\left(x\right)=x^3-4x^2+5x-2$

      而特徵值為 $1$ 與 $2$,其中 $I-A=\begin{bmatrix}0&1&0\\-1&-1&-1\\0&-1&0\end{bmatrix}$,其秩數為 $2$,其核數為 $1$,從而極小多項式恰為特徵多項式。如此取 $v=\begin{bmatrix}1\\0\\0\end{bmatrix}$ 循環生成向量,即知 $\left\{v,Av,A^2v\right\}$ 為 $\mathbb{R}^3$ 中的基底。取 $P=\begin{bmatrix}v&Av&A^2v\end{bmatrix}=\begin{bmatrix}1&1&0\\0&1&3\\0&0&1\end{bmatrix}$,而得

      $P^{-1}AP=\begin{bmatrix}0&0&2\\1&0&-5\\0&1&4\end{bmatrix}$

      類似地,對於轉置矩陣 $A^t=\begin{bmatrix}1&1&0\\-1&2&1\\0&1&1\end{bmatrix}$ 可取 $u=\begin{bmatrix}1\\0\\0\end{bmatrix}$ 作為循環生成向量,即有 $\left\{u,Au,A^2u\right\}$ 作為 $\mathbb{R}^3$ 中的基底。取 $R=\begin{bmatrix}u&A^tu&(A^t)^2u\end{bmatrix}=\begin{bmatrix}1&1&0\\0&-1&-3\\0&0&-1\end{bmatrix}$,而得

      $R^{-1}A^tR=\begin{bmatrix}0&0&2\\1&0&-5\\0&1&4\end{bmatrix}$

      至此我們有 $P^{-1}AP=R^{-1}A^tR$,故 $(PR^{-1})^{-1}A(PR^{-1})=A^t$。因此取

      $Q=PR^{-1}=\begin{bmatrix}1&1&0\\0&1&3\\0&0&1\end{bmatrix}\begin{bmatrix}1&1&-3\\0&-1&3\\0&0&-1\end{bmatrix}=\begin{bmatrix}1&0&0\\0&-1&0\\0&0&-1\end{bmatrix}$.


  3. Let $A$ and $B$ be two square matrices over $\mathbb{C}$. Prove that the set of eigenvalues of $AB$ is the same as the set of eigenvalues of $BA$. (15 points.)
  4. 訣竅直接證明兩集合互相包含來證明,其中應特別留意特徵向量必須非零。
    解法設 $E_{AB}$ 與 $E_{BA}$ 分別為 $AB$ 與 $BA$ 的特徵值全體。若 $\lambda\in E_{AB}$,那麼存在一 $v\neq0$ 滿足 $ABv=\lambda v$,從而

    $BA\left(Bv\right)=B\left(ABv\right)=B\left(\lambda v\right)=\lambda Bv$.

    以下分兩種情形考慮
    • 若 $\lambda\neq0$,那麼 $\lambda v\neq0$,從而 $Bv\neq0$,故 $\lambda$ 為 $BA$ 的特徵值且特徵向量為 $Bv$。
    • 若 $\lambda=0$,那麼直接看出 $\det\left(AB\right)=0$,如此 $\det\left(BA\right)=\det\left(A\right)\det\left(B\right)=\det\left(AB\right)=0$,從而 $\lambda=0$ 也為 $BA$ 的特徵值。
    綜上兩種情形恆知 $\lambda\in E_{BA}$,即 $E_{AB}\subseteq E_{BA}$。反之也有 $E_{BA}\subseteq E_{AB}$,故知 $E_{AB}=E_{BA}$,證明完畢。

  5. Let $\theta$ be a real number that is not an integer multiple of $\pi$ (so that $\sin\theta\neq0$). For a positive integer $n$, let $A_n=\left(a_{ij}\right)$ be the $n\times n$ matrix defined by

    $a_{ij}=\begin{cases}0,&\text{if}~\left|i-j\right|>1,\\1,&\text{if}~\left|i-j\right|=1,\\2\cos\theta,&\text{if}~i=j.\end{cases}$

    Prove that $\displaystyle\det A_n=\frac{\sin\left(n+1\right)\theta}{\sin\theta}$. (15 points.)
  6. 訣竅透過數學歸納法求解。
    解法設 $c_n=\det A_n$,那麼容易知道 $\displaystyle c_1=2\cos\theta=\frac{2\sin\theta\cos\theta}{\sin\theta}=\frac{\sin2\theta}{\sin\theta}$ 以及

    $\displaystyle c_2=\begin{bmatrix}2\cos\theta&1\\1&2\cos\theta\end{bmatrix}=4\cos^2\theta-1=\frac{4\sin\theta\left(1-\sin^2\theta\right)-\sin\theta}{\sin\theta}=\frac{3\sin\theta-4\sin^3\theta}{\sin\theta}=\frac{\sin3\theta}{\sin\theta}$.

    再者,對於第一列進行降階後針對第二項再對第一行降階可得

    $c_{n+1}=2c_n\cos\theta-c_{n-1}$

    考慮線性遞迴的輔助方程式 $\lambda^2-2\lambda\cos\theta+1=0$,可解得 $\lambda=\cos\theta\pm i\sin\theta=e^{\pm i\theta}$。如此遞迴關係式的通解為

    $c_n=A\cos\left(n\theta\right)+B\sin\left(n\theta\right)$

    其中 $A$ 與 $B$ 為常數。現利用 $c_1$ 與 $c_2$ 的結果來決定 $A$ 與 $B$ 如下

    $\left\{\begin{aligned}&A\cos\theta+B\sin\theta=2\cos\theta,\\&A\cos\left(2\theta\right)+B\sin\left(2\theta\right)=4\cos^2\theta-1.\end{aligned}\right.$

    第一式乘以 $2\cos\theta$ 與第二式相減可知

    $\left[2\cos^2\theta-\cos\left(2\theta\right)\right]A=1$

    故 $A=1$,而 $B=\cot\theta$。因此有

    $\displaystyle c_n=\cos\left(n\theta\right)+\cot\theta\sin\left(n\theta\right)=\frac{\sin\theta\cos\left(n\theta\right)+\cos\theta\sin\left(n\theta\right)}{\sin\theta}=\frac{\sin\left(n+1\right)\theta}{\sin\theta}$.


  7. Let

    $A=\begin{pmatrix}3&-2&-2\\-2&0&1\\-2&1&0\end{pmatrix}$.

    1. Find the maximum of $X^tAX$ among all $X\in\mathbb{R}^3$ subject to $X^tX=1$. Give an example of $X$ that attains the maximum. (8 ponts.)
    2. Find the minimum of $\text{tr}\left(Y^tAY\right)$ among all $3\times2$ matrices $Y$ over $\mathbb{R}$ subject to $Y^tY=I_2$. Give an example of $Y$ that attains the minimum. (7 points.)
  8. 訣竅經由對稱正定的概念可知二次型的最大值為最大的特徵值,且達到極值的方向為該特徵值的特徵向量;又注意到跡為特徵值之和,故透過 min-max 原理盡可能取較小的特徵值即可。
    解法
    1. 先計算特徵多項式如下

      $\begin{aligned}\text{ch}_A\left(x\right)&=\det\left(A-xI\right)=\begin{vmatrix}3-x&-2&-2\\-2&-x&1\\-2&1&-x\end{vmatrix}\\&=x^2\left(3-x\right)+4+4+4x+\left(x-3\right)+4x=-x^3+3x^2+9x+5\end{aligned}$

      那麼解 $\text{ch}_A\left(x\right)=0$ 可得 $x=-1$(二重根)與 $x=5$,故 $X^tAX$ 的最大值為 $5$。
      再者,計算 $A-5I=\begin{bmatrix}-2&-2&-2\\-2&-5&1\\-2&1&-5\end{bmatrix}$ 可取 $\displaystyle u=\frac1{\sqrt6}\begin{bmatrix}2\\-1\\-1\end{bmatrix}$ 滿足 $Av=5v$。據此可發現取 $X=u$ 能使 $X^tAX=5$。
      更精確地說明如下,計算 $x=1$ 時的特徵向量有 $\displaystyle v=\frac1{\sqrt5}\begin{bmatrix}1\\2\\0\end{bmatrix}$ 與 $\displaystyle w=\frac1{\sqrt{30}}\begin{bmatrix}2\\-1\\5\end{bmatrix}$。取 $P=\begin{bmatrix}u&v&w\end{bmatrix}$,那麼 $P^tAP=D$ 或寫為 $A=PDP^t$,其中

      $D=\begin{bmatrix}5&0&0\\0&-1&0\\0&0&-1\end{bmatrix}$

      記 $X'=P^tX=\left(x_1',x_2',x_3'\right)$,那麼

      $X^tAX=X^tPDP^tX=X'^tDX'=5x_1'^2-x_2'^2-x_3'^2$

      容易注意到

      $x_1'^2+x_2'^2+x_3'^2=X'^tX'=X^tPP^tX=X^tX=1$.

      $X^tAX=6x_1'^2-1\leq6-1=5$

      而等號成立為 $x_1'=1$、$x_2'=x_3'=0$,即 $X=PX'=u$ 能產生最大值。
    2. 承前一小題地過程並由 min-max 原理的概念可知最小值為 $-2$,取

      $\displaystyle Y=\begin{bmatrix}v&w\end{bmatrix}=\frac1{\sqrt{30}}\begin{bmatrix}\sqrt6&2\\2\sqrt6&-1\\0&5\end{bmatrix}$

      即可滿足條件。

  9. Let $V=\left\{a_0+a_1x+a_2x^2~:~a_j\in\mathbb{R}\right\}$ be the vector space of all polynomials of degree $2$ or less over $\mathbb{R}$. Define three linear functionals on $V$ by

    $\displaystyle f_1\left(p\right)=\int_0^1p\left(x\right)dx$, $\displaystyle f_2\left(p\right)=\int_0^2p\left(x\right)dx$, $\displaystyle f_3\left(p\right)=\int_{-1}^0p\left(x\right)dx$.

    Find a basis $\mathcal{B}$ for $V$ such that $\left\{f_1,f_2,f_3\right\}$ is its dual basis. (15 points.)
  10. 訣竅按照對偶基底的定義寫出聯立方程並求解即可。
    解法設所求的基底為 $\mathcal{B}=\left\{p_1,p_2,p_3\right\}$,其中 $p_1\left(x\right)=a_0+a_1x+a_2x^2$、$p_2\left(x\right)=b_0+b_1x+b_2x^2$ 以及 $p_3\left(x\right)=c_0+c_1x+c_2x^2$。如此有方程式

    $\left\{\begin{aligned}&1=f_1\left(p_1\right)=a_0+\frac{a_1}2+\frac{a_2}3\\&0=f_2\left(p_1\right)=2a_0+\frac{4a_1}2+\frac{8a_2}3\\&0=f_3\left(p_1\right)=a_0-\frac{a_1}2+\frac{a_2}3\end{aligned}\right.\quad\left\{\begin{aligned}&0=f_1\left(p_2\right)=b_0+\frac{b_1}2+\frac{b_2}3\\&1=f_2\left(p_2\right)=2b_0+\frac{4b_1}2+\frac{8b_2}3\\&0=f_3\left(p_2\right)=b_0-\frac{b_1}2+\frac{b_2}3\end{aligned}\right.\quad\left\{\begin{aligned}&0=f_1\left(p_3\right)=c_0+\frac{c_1}2+\frac{c_2}3\\&0=f_2\left(p_3\right)=2c_0+\frac{4c_1}2+\frac{8c_2}3\\&1=f_3\left(p_3\right)=c_0-\frac{c_1}2+\frac{c_2}3\end{aligned}\right.$

    解聯立方程組可得

    $\left\{\begin{aligned}&a_0=1\\&a_1=1\\&a_2=-\frac32\end{aligned}\right.\qquad\left\{\begin{aligned}&b_0=-\frac16\\&b_1=0\\&b_2=\frac12\end{aligned}\right.\qquad\left\{\begin{aligned}&c_0=\frac13\\&c_1=-1\\&c_2=\frac12\end{aligned}\right.$

    故可取 $\displaystyle\mathcal{B}=\left\{1+x-\frac{3x^2}2,-\frac16+\frac{x^2}2,\frac13-x+\frac{x^2}2\right\}$ 使 $\left\{f_1,f_2,f_3\right\}$ 為 $V^*$ 的對偶基底。

  11. Find all possible Jordan forms for $7\times7$ real matrices having $x^2\left(x-1\right)^2$ as minimal polynomial. (15 points.)
  12. 訣竅對各種可能的情況進行仔細的分類。
    解法設 $J_n\left(\lambda\right)$ 表 $n$ 階基本上三角 Jordan 塊,其中對角線為 $\lambda$,即

    $J_n\left(\lambda\right)=\begin{bmatrix}\lambda&1&0&\cdots&0&0\\0&\lambda&1&\cdots&0&0\\0&0&\lambda&\cdots&0&0\\\vdots&\vdots&\vdots&\ddots&\vdots&\vdots\\0&0&0&\cdots&\lambda&1\\0&0&0&\cdots&0&\lambda\end{bmatrix}$

    那麼由極小多項式的形式可以知道對於 $\lambda=0$ 與 $\lambda=1$ 的塊的大小至多為二階方陣且至少有一塊為二階,從而依特徵值由大而小,且盡可能使大的特徵值有最多的塊進而減少,如此可排序如下:
    • $J_2\left(1\right)\oplus J_2\left(1\right)\oplus J_1\left(1\right)\oplus J_2\left(0\right)$
    • $J_2\left(1\right)\oplus J_1\left(1\right)\oplus J_1\left(1\right)\oplus J_1\left(1\right)\oplus J_2\left(0\right)$
    • $J_2\left(1\right)\oplus J_2\left(1\right)\oplus J_2\left(0\right)\oplus J_1\left(0\right)$
    • $J_2\left(1\right)\oplus J_1\left(1\right)\oplus J_1\left(1\right)\oplus J_2\left(0\right)\oplus J_1\left(0\right)$
    • $J_2\left(1\right)\oplus J_1\left(1\right)\oplus J_2\left(0\right)\oplus J_2\left(0\right)$
    • $J_2\left(1\right)\oplus J_1\left(1\right)\oplus J_2\left(0\right)\oplus J_1\left(0\right)\oplus J_1\left(0\right)$
    • $J_2\left(1\right)\oplus J_2\left(0\right)\oplus J_2\left(0\right)\oplus J_1\left(0\right)$
    • $J_2\left(1\right)\oplus J_2\left(0\right)\oplus J_1\left(0\right)\oplus J_1\left(0\right)\oplus J_1\left(0\right)$
    如上所列,共八種可能。

沒有留言:

張貼留言