线性代数 05.02 方阵的特征值与特征向量

$\color{blue}{\S 第五章 第二节 方阵的特征值与特征向量}$

$\color{blue}{一、特征值与特征向量的概念}$

$定义6.设 A 为 n 阶方阵,如果数 \lambda 和 n 维非零列向量 x ,\\ 使关系式\\ Ax = \lambda x \qquad (1) \\ 成立,那么,称数 \lambda 为方阵 A 的特征值,\\ 非零向量x称为 A 的对应于特征值 \lambda 的特征向量.$
$(1)也可以写成(A - \lambda E)x = 0 \qquad (2)$
$这是n个未知数n个方程的齐次线性方程组, \\ 它有非零解的充分必要条件是系数行列式 \\ |A - \lambda E| = 0 \qquad (3)$
$由此,我们可由(3)求A的特征值,由(2)求A的特征向量.$

$\color{blue}{1.特征多项式}$

$f(\lambda) = |\lambda E - A| = \begin{vmatrix} \lambda -a_{11} & -a_{12} & \cdots & -a_{1n} \\ -a_{21} & \lambda -a_{22} & \cdots & -a_{2n} \\ \cdots & \cdots & \cdots & \cdots \\ -a_{n1} & -a_{n2} & \cdots & \lambda -a_{nn} \\ \end{vmatrix} \\ = \lambda^n - (a_{11} + a_{22} + \cdots + a_{nn})\lambda^{n - 1} - \cdots - (-1)^n|A|$

$\color{blue}{2.特征方程}$

$|A - \lambda E| = 0$
$即 \lambda^n - (a_{11} + a_{22} + \cdots + a_{nn})\lambda^{n - 1} + \cdots + (-1)^n|A| = 0$

$\color{blue}{3.A的迹}$

$A的迹 = a_{11} + a_{22} + \cdots + a_{nn}$

$\color{blue}{二、特征值与特征向量的性质}$

$设\lambda_1, \lambda_2, \cdots, \lambda_n是A的n个特征值,由高次方程的韦达定理. 不难证明: \\ 性质1.\lambda_1 + \lambda_2 + \cdots + \lambda_n = a_{11} + a_{22} + \cdots + a_{nn} \\ 即方阵A的{\color{blue}{特征值的和}}等于{\color{blue}{A的迹}}$

$一元二次方程的韦达定理: \\ x^2 + px + q = 0 \\ x_1 + x_2 = -p \\ x_1x_2 = q$

$性质2.\lambda_1 \lambda_2 \cdots \lambda_n = |A|$

$例1.证明方阵A可逆的充分必要条件是零不是A的特征值.$
$证.设\lambda_1, \lambda_2, \cdots, \lambda_n,是A的n个特征值.\\ 必要性:因为A可逆,所以|A| \neq 0, 由性质2 \\ \lambda_1 \lambda_2 \cdots \lambda_n \neq 0 \\ 故 \lambda_1, \lambda_2, \cdots, \lambda_n 不为零,从而零不是A的特征值. \\ 充分性:由于 \lambda_1, \lambda_2, \cdots, \lambda_n均不为零,\\ 从而 \lambda_1 \lambda_2 \cdots \lambda_n \neq 0 \\ 由性质2, |A| = \lambda_1 \lambda_2 \cdots \lambda_n \neq 0,故A可逆.$
$同理可证:A不可逆的充要条件是A有零特征值.$

$\color{blue}{三、特征值与特征向量的求法}$

$1.由|A - \lambda E| = 0,求A的n个特征值.$
$2.由Ax = \lambda x,求抽象矩阵的特征值.$
$3.由(A - \lambda E)x = 0,求A的特征向量.$

$例2.求矩阵 A = \begin{pmatrix} -1 & 1 & 0 \\ -4 & 3 & 0 \\ 1 & 0 & 2 \\ \end{pmatrix}的特征值和特征向量.$
$解: \\ ①由|A - \lambda E | = 0,求A的全部特征值, \\ |A - \lambda E| = \begin{vmatrix} -1 - \lambda & 1 & 0 \\ -4 & 3 - \lambda & 0 \\ 1 & 0 & 2 - \lambda \\ \end{vmatrix} \\ =(-1)^{(3+3)}(2-\lambda)[(-1-\lambda)(3-\lambda) - (-4 \times 1)] \\ = (2 - \lambda)(1 - \lambda)^2 = 0 \\ 得A的特征值为:\lambda_1 = 2, \lambda_2 = \lambda_3 = 1 \\ ②由(A - \lambda E) x = 0,求A的特征向量. \\ 当\lambda = 2时, 解方程(A - 2E) x = 0 由 \\ A - 2E = \begin{pmatrix} -3 & 1 & 0 \\ -4 & 1 & 0 \\ 1 & 0 & 0 \\ \end{pmatrix} \sim \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 0 \\ \end{pmatrix} \\ 得基础解系p_1 = \begin{pmatrix} 0 \\ 0 \\ 1 \end{pmatrix} \\ 所以kp_1(k \neq 0)是对应于\lambda_1 = 2的全部特征向量. \\ 当\lambda_2 = \lambda_3 = 1时,解方程(A - E) x = 0由 \\ (A - E) = \begin{pmatrix} -2 & 1 & 0 \\ -4 & 2 & 0 \\ 1 & 0 & 1 \\ \end{pmatrix} \sim \begin{pmatrix} 1 & 0 & 1 \\ 0 & 1 & 2 \\ 0 & 0 & 0 \\ \end{pmatrix} \\ 得基础解系 p_2 = \begin{pmatrix} -1 \\ -2 \\ 1 \\ \end{pmatrix}, \\ 所以k_2p_2(k_2 \neq 0)是对应于\lambda_2 = \lambda_3 = 1的全部特征向量.$

$例3.求矩阵 A = \begin{pmatrix} -2 & 1 & 1 \\ 0 & 2 & 0 \\ -4 & 1 & 3 \\ \end{pmatrix}的特征值和特征向量.$
$解: ①由|A - \lambda E| = 0求A的全部特征值. \\ |A - \lambda E| = \begin{vmatrix} -2 - \lambda & 1 & 1 \\ 0 & 2 - \lambda & 0 \\ -4 & 1 & 3 - \lambda \\ \end{vmatrix} \\ = (-1)^{(2 + 2)}(2 - \lambda)[(-2 - \lambda)(3 - \lambda) - (-4 \times 1)] \\ = -(\lambda + 1)(\lambda - 2)^2 = 0 \\ 方程的特征值为 \lambda_1 = -1, \lambda_2 = \lambda_3 = 2 \\ ②由(A - \lambda E) x= 0,求特征向量.\\ 当\lambda = -1时,解方程(A + E) x = 0,由 \\ A + E = \begin{pmatrix} -1 & 1 & 1 \\ 0 & 3 & 0 \\ -4 & 1 & 4 \\ \end{pmatrix} \\ \sim \begin{pmatrix} 1 & 0 & -1 \\ 0 & 1 & 0 \\ 0 & 0 & 0 \\ \end{pmatrix} \\ 得基础解系 p_1 = \begin{pmatrix} 1 \\ 0 \\ 1 \end{pmatrix}, \\ 所以k_1p_1(k_1 \neq 0)是对应于\lambda_1 = -1时的全部特征向量. \\ 当\lambda_2 = \lambda_3 = 2时, 解方程(A - 2E) x= 0,由 \\ A - 2E = \begin{pmatrix} -4 & 1 & 1 \\ 0 & 0 & 0 \\ -4 & 1 & 1 \\ \end{pmatrix} \sim \begin{pmatrix} 1 & -\dfrac{1}{4} & -\dfrac{1}{4} \\ 0 & 0 & 0 \\ 0 & 0 & 0 \\ \end{pmatrix} \\ 得基础解系 p_2 = \begin{pmatrix}\dfrac{1}{4} \\ 1 \\ 0 \end{pmatrix}， p_3 = \begin{pmatrix}\dfrac{1}{4} \\ 0 \\ 1 \end{pmatrix}, \\ 所以k_2 p_2 + k_3 p_3 \\ = k_2\begin{pmatrix}\dfrac{1}{4} \\ 1 \\ 0 \end{pmatrix} + k_3\begin{pmatrix}\dfrac{1}{4} \\ 0 \\ 1 \end{pmatrix},(k_2 \neq 0, k_3 \neq 0) \\ 是\lambda_2 = \lambda_3 = 2时的所有特征向量.$

$例4.设\lambda是方阵A的特征值,则\lambda^2也是A^2的特征值.$
$证:因为\lambda是方阵A的特征值,设P \neq 0,\\ 使AP = \lambda P,于是 \\ A^2P = A(AP) = A(\lambda P) = \lambda (AP) \\ = \lambda (\lambda P) = \lambda ^2 P$

$例5.设3阶方阵A满足A^3 - 3A^2 + 2A = 0,求A的特征值.$
$解: 设\lambda是A的特征值,x是A的关于\lambda所对应的特征向量.\\ 则Ax = \lambda x,从而 \\ A^3x - 3A^2x + 2Ax = 0 \\ 由例2得\\ (\lambda^3 - 3\lambda^2 + 2\lambda) x = 0 \quad x \neq 0 \\ \lambda^3 - 3\lambda^2 + 2\lambda = 0 \\ \lambda(\lambda -1)(\lambda -2) = 0 \\ 故得A的特征值为:\lambda_1 = 0, \lambda_2 = 1, \lambda_3 = 2.$

$例6.若\lambda 是可逆矩阵A的特征,x 是A的关于 \lambda 所对应的特征向量,则\\ (1) \dfrac{1}{\lambda} 是 A^{-1} 的特征值. \\ (2) \dfrac{|A|}{\lambda}是A^{*}的特征值, x仍是它们的特征值向量.$
$证:(1)由已知,A x = \lambda x,两边左乘A^{-1} \\ x = \lambda A^{-1} x \\ 从而 A^{-1} x = \dfrac{1}{\lambda} x \\ 由定义知, \dfrac{1}{\lambda} 是 A^{-1} 的特征值. \\ (2) 因 Ax = \lambda x,两边左乘 A^ { * } \\ A^ { * }A x = \lambda A^ { * }x \\ 即 |A| x = \lambda A^{*} x \\ 从而 A^{*} x = \dfrac{|A|}{\lambda} x \\ 由定义知, \dfrac{|A|}{\lambda}是A^{*}的特征值, x仍是它们的特征值向量$

$\color{blue}{由上面各例类推,不难证明,若\lambda是A的特征值,\\ 则\lambda^k是A^k的特征值,\\ \varphi(\lambda)是\varphi(A)的特征值. \\ (其中\varphi(\lambda) = a_0 + a_1 \lambda + \cdots + a_m \lambda^m, \\ \varphi(A) = a_0 E + a_1 A + \cdots + a_m A^m) }$

$例7.设有4阶方阵A满足 |A + 3E | = 0, 且AA^T = 2E, \\ |A| < 0,求A^{*}的一个特征值.$
$解: 由|A + 3E | = 0, 即 |A - (-3) E| = 0,\\ 所以 \lambda = -3 是A的一个特征值.\\ 又 |AA^T| = |2E| = 2^4,得|A|^2 = 16, \\ 从而 |A| = \pm 4,而|A| < 0,所以 |A| = -4, \\ 故 A ^ { * }的一个特征值为:\dfrac{ |A| } { \lambda } = \dfrac{-4}{-3} = \dfrac{4}{3}.$

$\color{blue}{四、特征值与特征向量的有关定理}$

$定理2.设\lambda_1, \lambda_2, \cdots, \lambda_m是A的m个特征值p_1, p_2, \cdots, p_m\\ 依次与之对应的特征向量,若\lambda_1, \lambda_2, \cdots, \lambda_m各不相同,\\ 则p_1, p_2, \cdots, p_m线性无关.$
$证:设有常数x_1, x_2, \cdots, x_m使\\ x_1p_1 + x_2p_2 + \cdots + x_m p_m = 0 \qquad (1) \\ 则A(x_1p_1 + x_2p_2 + \cdots + x_m p_m) = 0,即 \\ \lambda_1 x_1 p_1 + \lambda_2 x_2 p_2 + \cdots + \lambda_m x_m p_m = 0 \qquad (2) \\ 在(2)两边左乘A,并Ap_j = \lambda_jp_j, (j = 1, 2, \cdots, m) 得\\ \lambda_1^2 x_1 p_1 + \lambda_2^2 x_2 p_2 + \cdots + \lambda_m^2 x_m p_m = 0 \qquad (3) \\ 依次做下去 \\ \lambda_1^{m -1} x_1 p_1 + \lambda_2^{m -1} x_2 p_2 + \cdots + \lambda_m^{m -1} x_m p_m = 0 \qquad (m) \\ 将上面m个式子联立成线性方程组,得向量方程组$
$\left \{ \begin{array}{l} x_1p_1 + x_2p_2 + \cdots + x_m p_m = 0 \\ \lambda_1 x_1 p_1 + \lambda_2 x_2 p_2 + \cdots + \lambda_m x_m p_m = 0 \\ \lambda_1^2 x_1 p_1 + \lambda_2^2 x_2 p_2 + \cdots + \lambda_m^2 x_m p_m = 0 \\ \cdots \qquad \cdots \qquad \cdots \qquad \cdots \qquad \cdots \\ \lambda_1^{m -1} x_1 p_1 + \lambda_2^{m -1} x_2 p_2 + \cdots + \lambda_m^{m -1} x_m p_m = 0 \end{array} \right.$
$由于系数行列式是范德蒙行列式,所以\\ D = \begin{vmatrix} 1 & 1 & \cdots 1 \\ \lambda_1 & \lambda_2 & \cdots & \lambda_m \\ \lambda_1^2 & \lambda_2^2 & \cdots & \lambda_m^2 \\ \cdots & \cdots & \cdots & \cdots \\ \lambda_1^{m - 1} & \lambda_2^{m - 1} & \cdots & \lambda_m^{m - 1} \\ \end{vmatrix} \\ = \prod_{m \geq i > j \geq 1}(\lambda_i - \lambda_j) \\ 又因为\lambda_1, \lambda_2, \cdots, \lambda_m各不相同,所以D \neq 0 \\ 由克莱默法则知,向量方程组仅有零解.即\\ x_jp_j = 0 (j = 1, 2, \cdots, m) \\ 而p_j \neq 0 (j = 1, 2, \cdots, m), \\ 而x_j = 0 (j = 1, 2, \cdots, m), 由(1)知 \\ p_1,p_2, \cdots, p_m线性无关.$

$例8.设A为4阶方阵,其特征值\lambda_1,\lambda_2, \lambda_3, \lambda_4各不相同,\\ 对应的特征向量依次为\alpha_1, \alpha_2, \alpha_3, \alpha_4, \\ 令\beta = \alpha_1 + \alpha_2 + \alpha_3 + \alpha_4, \\ 证明: \beta, A\beta, A^2\beta, A^3 \beta线性无关.$
$证明:因为A\alpha_j = \lambda_j \alpha_j (j = 1, 2, 3, 4),所以 \\ A\beta = A(\alpha_1 + \alpha_2 + \alpha_3 + \alpha_4) \\ = \lambda_1\alpha_1 + \lambda_2 \alpha_2 + \lambda_3 \alpha_3 + \lambda_4 \alpha_4 \\ A^2\beta = A(\lambda_1\alpha_1 + \lambda_2 \alpha_2 + \lambda_3 \alpha_3 + \lambda_4 \alpha_4) \\ = \lambda_1^2\alpha_1 + \lambda_2^2 \alpha_2 + \lambda_3^2 \alpha_3 + \lambda_4^2 \alpha_4 \\ A^3\beta = A(\lambda_1^2\alpha_1 + \lambda_2^2 \alpha_2 + \lambda_3^2 \alpha_3 + \lambda_4^2 \alpha_4) \\ = \lambda_1^3\alpha_1 + \lambda_2^3 \alpha_2 + \lambda_3^3\alpha_3 + \lambda_4^3 \alpha_4 \\ 设常数k_1, k_2, k_3, k_4使 \\ k_1 \beta + k_2 A\beta + k_3A^2 \beta + k_4 A^3 \beta = 0 \\ 即 k_1(\alpha_1 + \alpha_2 + \alpha_3 + \alpha_4) \\ + k_2(\lambda_1\alpha_1 + \lambda_2 \alpha_2 + \lambda_3 \alpha_3 + \lambda_4 \alpha_4) \\ + k_3(\lambda_1^2\alpha_1 + \lambda_2^2 \alpha_2 + \lambda_3^2 \alpha_3 + \lambda_4^2 \alpha_4) \\ + k_4(\lambda_1^3\alpha_1 + \lambda_2^3 \alpha_2 + \lambda_3^3\alpha_3 + \lambda_4^3 \alpha_4) = 0 \\ 整理,得:\\ (k_1 + k_2\lambda_1 + k_3 \lambda_1^2 + k_4 \lambda_1^3)\alpha_1 \\ + (k_1 + k_2\lambda_2 + k_3 \lambda_2^2 + k_4 \lambda_2^3)\alpha_2 \\ + (k_1 + k_2\lambda_3 + k_3 \lambda_3^2 + k_4 \lambda_3^3)\alpha_3 \\ + (k_1 + k_2\lambda_4 + k_3 \lambda_4^2 + k_4 \lambda_4^3)\alpha_4 = 0$
$\left \{ \begin{array}{l} k_1 + k_2\lambda_1 + k_3 \lambda_1^2 + k_4 \lambda_1^3 = 0 \\ k_1 + k_2\lambda_2 + k_3 \lambda_2^2 + k_4 \lambda_2^3 = 0 \\ k_1 + k_2\lambda_3 + k_3 \lambda_3^2 + k_4 \lambda_3^3 = 0 \\ k_1 + k_2\lambda_4 + k_3 \lambda_4^2 + k_4 \lambda_4^3 = 0 \end{array} \right.$
$其系数行列式是4阶范德蒙行列式,且\lambda_1, \lambda_2, \lambda_3, \lambda_4各不相同,\\ 所以上面的方程组仅有零解 k_1 = k_2 = k_3 = k_4 = 0. \\ 即 \beta, A\beta, A^2\beta, A^3 \beta线性无关.$