常微分方程(三)

4.高阶一维线性方程

  • n n n阶线性微分方程的一般形式是 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = f ( t ) \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=f(t) dtndnx+a1(t)dtn1dn1x++an(t)x=f(t),其中 a k ( t ) a_k(t) ak(t) [ α , β ] [\alpha,\beta] [α,β]是连续函数,且 t t t也定义在 [ α , β ] [\alpha,\beta] [α,β]上,当 f ( t ) ≡ 0 f(t)\equiv0 f(t)0时,是齐次方程,否则是非齐次方程
  • x 1 = x , x 2 = d x d t , ⋯   , x n = d n − 1 x d t n − 1 x_1=x,x_2=\frac{dx}{dt},\cdots,x_n=\frac{d^{n-1}x}{dt^{n-1}} x1=x,x2=dtdx,,xn=dtn1dn1x
    • x 1 ′ = x ′ = x 2 , x 2 ′ = x 3 , ⋯   , x n − 1 ′ = x n , x n ′ = − a 1 ( t ) x n − ⋯ − a n ( t ) x 1 + f ( t ) x_1'=x'=x_2,x_2'=x_3,\cdots,x_{n-1}'=x_{n},x_n'=-a_1(t)x_n-\cdots-a_n(t)x_1+f(t) x1=x=x2,x2=x3,,xn1=xn,xn=a1(t)xnan(t)x1+f(t)
    • 显然,每个导数都是 x 1 , x 2 , ⋯   , x n x_1,x_2,\cdots,x_n x1,x2,,xn的线性组合,因此 [ x 1 ′ ( t ) x 2 ′ ( t ) . . . x n ′ ( t ) ] = A ( t ) [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) ] + [ 0 0 . . . f ( t ) ] \begin{bmatrix}x_1'(t)\\x_2'(t)\\...\\x_n'(t)\end{bmatrix}=A(t)\begin{bmatrix}x_1(t)\\x_2(t)\\...\\x_n(t)\end{bmatrix}+\begin{bmatrix}0\\0\\...\\f(t)\end{bmatrix} x1(t)x2(t)...xn(t)=A(t)x1(t)x2(t)...xn(t)+00...f(t),其中 A ( t ) = [ 0 1 0 . . . 0 0 0 1 . . . 0 . . . 0 0 0 . . . 1 − a n ( t ) − a n − 1 ( t ) − a n − 2 ( t ) . . . − a 1 ( t ) ] A(t)=\begin{bmatrix}0&1&0&...&0\\0&0&1&...&0\\...\\0&0&0&...&1\\-a_n(t)&-a_{n-1}(t)&-a_{n-2}(t)&...&-a_1(t)\end{bmatrix} A(t)=00...0an(t)100an1(t)010an2(t)............001a1(t)
    • 以此高阶一维方程转化为了一阶高维线性方程组 d y d t = A ( t ) y + g ( t ) \frac{dy}{dt}=A(t)y+g(t) dtdy=A(t)y+g(t),其中 y = [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) ] , g ( t ) = [ 0 0 . . . f ( t ) ] y=\begin{bmatrix}x_1(t)\\x_2(t)\\...\\x_n(t)\end{bmatrix},g(t)=\begin{bmatrix}0\\0\\...\\f(t)\end{bmatrix} y=x1(t)x2(t)...xn(t),g(t)=00...f(t)
  • 由于最开始令 x 1 = x x_1=x x1=x,因此 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = f ( t ) \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=f(t) dtndnx+a1(t)dtn1dn1x++an(t)x=f(t)的解是 d y d t = A ( t ) y + g ( t ) \frac{dy}{dt}=A(t)y+g(t) dtdy=A(t)y+g(t)的解的第一个元素 x 1 ( t ) x_1(t) x1(t)
  • 剩余的 [ x 2 ( t ) x 3 ( t ) . . . x n ( t ) ] = [ x ′ ( t ) x ′ ′ ( t ) . . . x ( n − 1 ) ( t ) ] = [ x 1 ′ ( t ) x 2 ′ ( t ) . . . x n − 1 ′ ( t ) ] \begin{bmatrix}x_2(t)\\x_3(t)\\...\\x_n(t)\end{bmatrix}=\begin{bmatrix}x'(t)\\x''(t)\\...\\x^{(n-1)}(t)\end{bmatrix}=\begin{bmatrix}x_1'(t)\\x_2'(t)\\...\\x_{n-1}'(t)\end{bmatrix} x2(t)x3(t)...xn(t)=x(t)x(t)...x(n1)(t)=x1(t)x2(t)...xn1(t),因此 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = f ( t ) \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=f(t) dtndnx+a1(t)dtn1dn1x++an(t)x=f(t)的解和 d y d t = A ( t ) y + g ( t ) \frac{dy}{dt}=A(t)y+g(t) dtdy=A(t)y+g(t)的解可以互相推出,即齐次高阶一维线性方程的解矩阵 [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) x 1 ′ ( t ) x 2 ′ ( t ) . . . ⋱ x 1 n − 1 ( t ) . . . . . . x n n − 1 ( t ) ] \begin{bmatrix}x_{1}(t)&x_{2}(t)&...&x_{n}(t)\\x_{1}'(t)&x_{2}'(t)&&\\...&&\ddots\\&\\x_{1}^{n-1}(t)&...&...&x_{n}^{n-1}(t)\end{bmatrix} x1(t)x1(t)...x1n1(t)x2(t)x2(t).........xn(t)xnn1(t)对应转化为齐次一阶高维线性方程的基解矩阵 X ( t ) = [ x 11 ( t ) x 12 ( t ) . . . x 1 n ( t ) x 21 ( t ) x 22 ( t ) . . . ⋱ x n 1 ( t ) . . . . . . x n n ( t ) ] X(t)=\begin{bmatrix}x_{11}(t)&x_{12}(t)&...&x_{1n}(t)\\x_{21}(t)&x_{22}(t)&&\\...&&\ddots\\&\\x_{n1}(t)&...&...&x_{nn}(t)\end{bmatrix} X(t)=x11(t)x21(t)...xn1(t)x12(t)x22(t).........x1n(t)xnn(t)
  • 满足初始条件 x ( t 0 ) = x 0 , x ′ ( t 0 ) = x 1 , ⋯   , x n − 1 ( t 0 ) = x n − 1 x(t_0)=x_0,x'(t_0)=x_1,\cdots,x^{n-1}(t_0)=x_{n-1} x(t0)=x0,x(t0)=x1,,xn1(t0)=xn1的方程 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = f ( t ) \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=f(t) dtndnx+a1(t)dtn1dn1x++an(t)x=f(t)的解是存在的且唯一的,可以通过转化为 d y d t = A ( t ) y + g ( t ) \frac{dy}{dt}=A(t)y+g(t) dtdy=A(t)y+g(t)来证明
  • 从解的角度上 d y d t = A ( t ) y + g ( t ) \frac{dy}{dt}=A(t)y+g(t) dtdy=A(t)y+g(t) d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = f ( t ) \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=f(t) dtndnx+a1(t)dtn1dn1x++an(t)x=f(t)是“等价”的
  • 定义 W ( t ) = d e t ( [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) x 1 ′ ( t ) x 2 ′ ( t ) . . . ⋱ x 1 n − 1 ( t ) . . . . . . x n n − 1 ( t ) ] ) W(t)=det(\begin{bmatrix}x_{1}(t)&x_{2}(t)&...&x_{n}(t)\\x_{1}'(t)&x_{2}'(t)&&\\...&&\ddots\\&\\x_{1}^{n-1}(t)&...&...&x_{n}^{n-1}(t)\end{bmatrix}) W(t)=det(x1(t)x1(t)...x1n1(t)x2(t)x2(t).........xn(t)xnn1(t))是解组 { x k ( t ) , k = 1 , 2 , ⋯   , n } \{x_k(t),k=1,2,\cdots,n\} {xk(t),k=1,2,,n} W r o n s k i Wronski Wronski行列式,其中的矩阵正是 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = 0 \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=0 dtndnx+a1(t)dtn1dn1x++an(t)x=0的解矩阵
  • x 1 ( t ) , x 2 ( t ) , ⋯   , x n ( t ) x_1(t),x_2(t),\cdots,x_n(t) x1(t),x2(t),,xn(t)线性无关    ⟺    \iff x 1 ( k ) ( t ) , x 2 ( k ) ( t ) , ⋯   , x n ( k ) ( t ) x_1^{(k)}(t),x_2^{(k)}(t),\cdots,x_n^{(k)}(t) x1(k)(t),x2(k)(t),,xn(k)(t)线性无关,因为求导之后的式子等价
  • L i o u v i l l e Liouville Liouville定理: n n n阶齐次线性微分方程 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = 0 \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=0 dtndnx+a1(t)dtn1dn1x++an(t)x=0的解组 { x k ( t ) , k = 1 , 2 , ⋯   , n } \{x_k(t),k=1,2,\cdots,n\} {xk(t),k=1,2,,n}线性无关的充要条件是它的 W r o n s k i Wronski Wronski行列式 W ( t ) W(t) W(t)在区间 [ α , β ] [\alpha,\beta] [α,β]上恒不为零,这等价于 W ( t ) W(t) W(t)在区间 [ α , β ] [\alpha,\beta] [α,β]上的某点 t 0 t_0 t0处不为零。方程的任一解组 { x k ( t ) , k = 1 , 2 , ⋯   , n } \{x_k(t),k=1,2,\cdots,n\} {xk(t),k=1,2,,n} W r o n s k i Wronski Wronski行列式满足 L i o u v i l l e Liouville Liouville公式, W ( t ) = W ( t 0 ) e − ∫ t 0 t a 1 ( τ ) d τ W(t)=W(t_0)e^{-\int_{t_0}^ta_1(\tau)d\tau} W(t)=W(t0)et0ta1(τ)dτ
    • 这表明 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = 0 \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=0 dtndnx+a1(t)dtn1dn1x++an(t)x=0解组对应的 W ( t ) W(t) W(t)只有两种情况:在任一点 t t t都不为零;在任一点 t t t都为零
    • 此处 L i o u v i l l e Liouville Liouville定理的证明可以通过 x ′ = A ( t ) x x'=A(t)x x=A(t)x L i o u v i l l e Liouville Liouville定理来推导记忆:此处的 A ( t ) = [ 0 1 0 . . . 0 0 0 1 . . . 0 . . . 0 0 0 . . . 1 − a n ( t ) − a n − 1 ( t ) − a n − 2 ( t ) . . . − a 1 ( t ) ] A(t)=\begin{bmatrix}0&1&0&...&0\\0&0&1&...&0\\...\\0&0&0&...&1\\-a_n(t)&-a_{n-1}(t)&-a_{n-2}(t)&...&-a_1(t)\end{bmatrix} A(t)=00...0an(t)100an1(t)010an2(t)............001a1(t),因此原公式中的 t r ( A ( t ) ) tr(A(t)) tr(A(t))显然就是此公式中的 − a 1 ( t ) -a_1(t) a1(t),对应的此处的 [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) x 1 ′ ( t ) x 2 ′ ( t ) . . . ⋱ x 1 n − 1 ( t ) . . . . . . x n n − 1 ( t ) ] \begin{bmatrix}x_{1}(t)&x_{2}(t)&...&x_{n}(t)\\x_{1}'(t)&x_{2}'(t)&&\\...&&\ddots\\&\\x_{1}^{n-1}(t)&...&...&x_{n}^{n-1}(t)\end{bmatrix} x1(t)x1(t)...x1n1(t)x2(t)x2(t).........xn(t)xnn1(t)中的 n n n个函数 x 1 ( t ) , ⋯   , x n ( t ) x_1(t),\cdots,x_n(t) x1(t),,xn(t)就是方程 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = 0 \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=0 dtndnx+a1(t)dtn1dn1x++an(t)x=0 n n n个解, W ( t 0 ) W(t_0) W(t0)就是原方程的 d e t ( X ( t 0 ) ) det(X(t_0)) det(X(t0))
  • 齐次线性微分方程 d n x d t n + a 1 ( t ) d n − 1 x d t n − 1 + ⋯ + a n ( t ) x = 0 \frac{d^{n}x}{dt^{n}}+a_1(t)\frac{d^{n-1}x}{dt^{n-1}}+\cdots+a_n(t)x=0 dtndnx+a1(t)dtn1dn1x++an(t)x=0在区间 [ α , β ] [\alpha,\beta] [α,β]上存在 n n n个线性无关的解 x 1 ( t ) , ⋯   , x n ( t ) x_1(t),\cdots,x_n(t) x1(t),,xn(t),且方程的通解是 x ( t ) = ∑ k = 1 n c k x k ( t ) x(t)=\sum_{k=1}^nc_kx_k(t) x(t)=k=1nckxk(t),其中, c 1 , ⋯   , c n c_1,\cdots,c_n c1,,cn是任意常数
  • 非齐次线性微分方程的通解是 x ( t ) = ∑ k = 1 n c k x k ( t ) + x ∗ ( t ) x(t)=\sum_{k=1}^nc_kx_k(t)+x^*(t) x(t)=k=1nckxk(t)+x(t),其中 c 1 , ⋯   , c n c_1,\cdots,c_n c1,,cn是任意常数,而 x ∗ = ∑ k = 1 n ∫ t 0 t x k ( t ) W k ( τ ) W ( τ ) f ( τ ) d τ x^*=\sum_{k=1}^n\int^{t}_{t_0}\frac{x_k(t)W_k(\tau)}{W(\tau)}f(\tau)d\tau x=k=1nt0tW(τ)xk(t)Wk(τ)f(τ)dτ是非齐次方程的一个特解, W ( t ) W(t) W(t)是解组的 W r o n s k i Wronski Wronski行列式, W k ( t ) W_k(t) Wk(t) W ( t ) W(t) W(t)中第 n n n行第 k k k列元素的代数余子式
    • 此处的非齐次高阶一维线性方程也可以对比非齐次一阶高维线性方程的解(由于最开始时令 x 1 = x x_1=x x1=x,即非齐次一阶高维线性方程的特解的第一行对应非齐次高阶一维线性方程的特解),非齐次一阶高维线性方程的特解是 X ( t ) ∫ t 0 t ( X ( τ ) − 1 f ( τ ) ) d τ X(t)\int^{t}_{t_0}(X(\tau)^{-1}f(\tau))d\tau X(t)t0t(X(τ)1f(τ))dτ,将非齐次高阶一维线性方程转化为对应的非齐次一阶高维线性方程 d y d t = A ( t ) y + g ( t ) \frac{dy}{dt}=A(t)y+g(t) dtdy=A(t)y+g(t)后,可以推出它的特解是 [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) x 1 ′ ( t ) x 2 ′ ( t ) . . . ⋱ x 1 n − 1 ( t ) . . . . . . x n n − 1 ( t ) ] ∫ t 0 t [ x 1 ( τ ) x 2 ( τ ) . . . x n ( τ ) x 1 ′ ( τ ) x 2 ′ ( τ ) . . . ⋱ x 1 n − 1 ( τ ) . . . . . . x n n − 1 ( τ ) ] − 1 [ 0 0 0 0 . . . f ( τ ) ] d τ \begin{bmatrix}x_{1}(t)&x_{2}(t)&...&x_{n}(t)\\x_{1}'(t)&x_{2}'(t)&&\\...&&\ddots\\&\\x_{1}^{n-1}(t)&...&...&x_{n}^{n-1}(t)\end{bmatrix}\int^{t}_{t_0}\begin{bmatrix}x_{1}(\tau)&x_{2}(\tau)&...&x_{n}(\tau)\\x_{1}'(\tau)&x_{2}'(\tau)&&\\...&&\ddots\\&\\x_{1}^{n-1}(\tau)&...&...&x_{n}^{n-1}(\tau)\end{bmatrix}^{-1}\begin{bmatrix}0\\0\\0\\0\\...\\f(\tau)\end{bmatrix}d\tau x1(t)x1(t)...x1n1(t)x2(t)x2(t).........xn(t)xnn1(t)t0tx1(τ)x1(τ)...x1n1(τ)x2(τ)x2(τ).........xn(τ)xnn1(τ)10000...f(τ)dτ,由于 g ( τ ) g(\tau) g(τ)的前 n − 1 n-1 n1个元素都是 0 0 0,因此逆矩阵只需要计算最后一列,而 A − 1 = 1 d e t ( A ) A ∗ A^{-1}=\frac{1}{det(A)}A^* A1=det(A)1A,因此根据伴随矩阵的定义和求法 [ x 1 ( τ ) x 2 ( τ ) . . . x n ( τ ) x 1 ′ ( τ ) x 2 ′ ( τ ) . . . ⋱ x 1 n − 1 ( τ ) . . . . . . x n n − 1 ( τ ) ] − 1 = 1 W ( τ ) [ ∗ ∗ . . . W 1 ( τ ) ∗ ∗ . . . ⋱ ∗ . . . . . . W n ( τ ) ] \begin{bmatrix}x_{1}(\tau)&x_{2}(\tau)&...&x_{n}(\tau)\\x_{1}'(\tau)&x_{2}'(\tau)&&\\...&&\ddots\\&\\x_{1}^{n-1}(\tau)&...&...&x_{n}^{n-1}(\tau)\end{bmatrix}^{-1}=\frac{1}{W(\tau)}\begin{bmatrix}*&*&...&W_1(\tau)\\*&*&&\\...&&\ddots\\&\\*&...&...&W_n(\tau)\end{bmatrix} x1(τ)x1(τ)...x1n1(τ)x2(τ)x2(τ).........xn(τ)xnn1(τ)1=W(τ)1............W1(τ)Wn(τ),其中 W k ( τ ) W_k(\tau) Wk(τ) W ( τ ) W(\tau) W(τ)中第 n n n行第 k k k列元素的代数余子式,例如根据公式, W 1 ( t ) W_1(t) W1(t)的位置应该是 A n 1 A_{n1} An1,即划去第 n n n行第 1 1 1列后的代数余子式,因此特解 [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) x 1 ′ ( t ) x 2 ′ ( t ) . . . ⋱ x 1 n − 1 ( t ) . . . . . . x n n − 1 ( t ) ] ∫ t 0 t 1 W ( τ ) [ ∗ ∗ . . . W 1 ( τ ) ∗ ∗ . . . ⋱ ∗ . . . . . . W n ( τ ) ] [ 0 0 0 0 . . . f ( τ ) ] d τ = [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) x 1 ′ ( t ) x 2 ′ ( t ) . . . ⋱ x 1 n − 1 ( t ) . . . . . . x n n − 1 ( t ) ] ∫ t 0 t 1 W ( τ ) [ W 1 ( τ ) f ( τ ) W 2 ( τ ) f ( τ ) W 3 ( τ ) f ( τ ) W 4 ( τ ) f ( τ ) ⋯ W n ( τ ) f ( τ ) ] d τ = [ x 1 ( t ) x 2 ( t ) . . . x n ( t ) x 1 ′ ( t ) x 2 ′ ( t ) . . . ⋱ x 1 n − 1 ( t ) . . . . . . x n n − 1 ( t ) ] [ ∫ t 0 t 1 W ( τ ) W 1 ( τ ) f ( τ ) d τ ∫ t 0 t 1 W ( τ ) W 2 ( τ ) f ( τ ) d τ ∫ t 0 t 1 W ( τ ) W 3 ( τ ) f ( τ ) d τ ∫ t 0 t 1 W ( τ ) W 4 ( τ ) f ( τ ) d τ ⋯ ∫ t 0 t 1 W ( τ ) W n ( τ ) f ( τ ) d τ ] \begin{bmatrix}x_{1}(t)&x_{2}(t)&...&x_{n}(t)\\x_{1}'(t)&x_{2}'(t)&&\\...&&\ddots\\&\\x_{1}^{n-1}(t)&...&...&x_{n}^{n-1}(t)\end{bmatrix}\int^{t}_{t_0}\frac{1}{W(\tau)}\begin{bmatrix}*&*&...&W_1(\tau)\\*&*&&\\...&&\ddots\\&\\*&...&...&W_n(\tau)\end{bmatrix}\begin{bmatrix}0\\0\\0\\0\\...\\f(\tau)\end{bmatrix}d\tau=\begin{bmatrix}x_{1}(t)&x_{2}(t)&...&x_{n}(t)\\x_{1}'(t)&x_{2}'(t)&&\\...&&\ddots\\&\\x_{1}^{n-1}(t)&...&...&x_{n}^{n-1}(t)\end{bmatrix}\int^{t}_{t_0}\frac{1}{W(\tau)}\begin{bmatrix}W_1(\tau)f(\tau)\\W_2(\tau)f(\tau)\\W_3(\tau)f(\tau)\\W_4(\tau)f(\tau)\\\cdots\\W_n(\tau)f(\tau)\end{bmatrix}d\tau=\begin{bmatrix}x_{1}(t)&x_{2}(t)&...&x_{n}(t)\\x_{1}'(t)&x_{2}'(t)&&\\...&&\ddots\\&\\x_{1}^{n-1}(t)&...&...&x_{n}^{n-1}(t)\end{bmatrix}\begin{bmatrix}\int^{t}_{t_0}\frac{1}{W(\tau)}W_1(\tau)f(\tau)d\tau\\\int^{t}_{t_0}\frac{1}{W(\tau)}W_2(\tau)f(\tau)d\tau\\\int^{t}_{t_0}\frac{1}{W(\tau)}W_3(\tau)f(\tau)d\tau\\\int^{t}_{t_0}\frac{1}{W(\tau)}W_4(\tau)f(\tau)d\tau\\\cdots\\\int^{t}_{t_0}\frac{1}{W(\tau)}W_n(\tau)f(\tau)d\tau\end{bmatrix} x1(t)x1(t)...x1n1(t)x2(t)x2(t).........xn(t)xnn1(t)t0tW(τ)1............W1(τ)Wn(τ)0000...f(τ)dτ=x1(t)x1(t)...x1n1(t)x2(t)x2(t).........xn(t)xnn1(t)t0tW(τ)1W1(τ)f(τ)W2(τ)f(τ)W3(τ)f(τ)W4(τ)f(τ)Wn(τ)f(τ)dτ=x1(t)x1(t)...x1n1(t)x2(t)x2(t).........xn(t)xnn1(t)t0tW(τ)1W1(τ)f(τ)dτt0tW(τ)1W2(τ)f(τ)dτt0tW(τ)1W3(τ)f(τ)dτt0tW(τ)1W4(τ)f(τ)dτt0tW(τ)1Wn(τ)f(τ)dτ,因为只需要得到第一行的 x 1 ( t ) , ⋯   , x n ( t ) x_1(t),\cdots,x_n(t) x1(t),,xn(t),因此对应相乘后再相加,即得到 x ∗ = ∑ k = 1 n ∫ t 0 t x k ( t ) W k ( τ ) W ( τ ) f ( τ ) d τ x^*=\sum_{k=1}^n\int^{t}_{t_0}\frac{x_k(t)W_k(\tau)}{W(\tau)}f(\tau)d\tau x=k=1nt0tW(τ)xk(t)Wk(τ)f(τ)dτ是非齐次方程的一个特解

5.复值解和级数解法

  • 方程在 R n R^n Rn内无解,而在 C n C^n Cn内可能有解,扩大考虑范围可能得到更多解

  • 对于复值矩阵函数 A ( t ) = A R ( t ) + i A I ( t ) A(t)=A_R(t)+iA_I(t) A(t)=AR(t)+iAI(t),定义求导与积分

    • d A d t = d A R ( t ) d t + i d A I ( t ) d t \frac{dA}{dt}=\frac{dA_R(t)}{dt}+i\frac{dA_I(t)}{dt} dtdA=dtdAR(t)+idtdAI(t)
    • ∫ α β A ( t ) d t = ∫ α β A R ( t ) + i ∫ α β A I ( t ) \int^{\beta}_{\alpha}A(t)dt=\int^{\beta}_{\alpha}A_R(t)+i\int^{\beta}_{\alpha}A_I(t) αβA(t)dt=αβAR(t)+iαβAI(t)
  • 欧拉公式: λ = a + i b , e λ t = e a t ( c o s b t + i s i n b t ) \lambda=a+ib,e^{\lambda t}=e^{a t}(cosbt+isinbt) λ=a+ib,eλt=eat(cosbt+isinbt)

  • z ( t ) = x ( t ) + i y ( t ) z(t)=x(t)+iy(t) z(t)=x(t)+iy(t),复值函数 z ( t ) z(t) z(t)与其共轭 z ( t ) ˉ \bar{z(t)} z(t)ˉ线性无关    ⟺    \iff 它们的实部 x ( t ) x(t) x(t)和虚部 y ( t ) y(t) y(t)线性无关

  • A ( t ) = A R ( t ) + i A I ( t ) A(t)=A_R(t)+iA_I(t) A(t)=AR(t)+iAI(t)是复值连续矩阵, f ( t ) = f R ( t ) + i f I ( t ) f(t)=f_R(t)+if_I(t) f(t)=fR(t)+ifI(t)是复制连续列向量,若 z ( t ) = x ( t ) + i y ( t ) z(t)=x(t)+iy(t) z(t)=x(t)+iy(t)是线性方程组 d z d t = A ( t ) z + f ( t ) \frac{dz}{dt}=A(t)z+f(t) dtdz=A(t)z+f(t)的复值解,则实连续函数 x = x ( t ) , y = y ( t ) x=x(t),y=y(t) x=x(t),y=y(t) d d t [ x y ] = [ A R ( t ) − A I ( t ) A I ( t ) A R ( t ) ] [ x y ] + [ f R ( t ) f I ( t ) ] \frac{d}{dt}\begin{bmatrix}x\\y\end{bmatrix}=\begin{bmatrix}A_R(t)&-A_I(t)\\A_I(t)&A_R(t)\end{bmatrix}\begin{bmatrix}x\\y\end{bmatrix}+\begin{bmatrix}f_R(t)\\f_I(t)\end{bmatrix} dtd[xy]=[AR(t)AI(t)AI(t)AR(t)][xy]+[fR(t)fI(t)]的解

    • 注意 x ( t ) x(t) x(t) y ( t ) y(t) y(t)都是多维列向量函数,而不是一般函数
    • 复值向量函数 z ( t ) = x ( t ) + i y ( t ) z(t)=x(t)+iy(t) z(t)=x(t)+iy(t)是实系数矩阵(即 A ( t ) = A R ( t ) A(t)=A_R(t) A(t)=AR(t))齐次线性方程组 d z d t = A ( t ) z \frac{dz}{dt}=A(t)z dtdz=A(t)z的解    ⟺    \iff 实部 x ( t ) x(t) x(t)和虚部 y ( t ) y(t) y(t)都是方程 d z d t = A ( t ) z \frac{dz}{dt}=A(t)z dtdz=A(t)z的解, z ( t ) ˉ \bar{z(t)} z(t)ˉ也是 d z d t = A ( t ) z \frac{dz}{dt}=A(t)z dtdz=A(t)z的解,可以理解为齐次线性方程组的叠加原理
  • C a u c h y Cauchy Cauchy定理:如果函数 f ( x , t ) f(x,t) f(x,t)在矩阵区域 R = { ( t , x ) ∈ R 2 : ∣ t − t 0 ∣ < α , ∣ x − x 0 ∣ < β } R=\{(t,x)\in\R^2:|t-t_0|<\alpha,|x-x_0|<\beta\} R={(t,x)R2:tt0<α,xx0<β}上可以展开成 f ( t , x ) = ∑ k , l = 0 ∞ γ k l ( t − t 0 ) k ( x − x 0 ) l f(t,x)=\sum^{\infty}_{k,l=0}\gamma_{kl}(t-t_0)^k(x-x_0)^l f(t,x)=k,l=0γkl(tt0)k(xx0)l(此时 f ( t , x ) f(t,x) f(t,x)可解析,即多元函数可展开为收敛的幂级数),则初值问题 d x d t = f ( t , x ) , x ( t 0 ) = x 0 \frac{dx}{dt}=f(t,x),x(t_0)=x_0 dtdx=f(t,x),x(t0)=x0 t 0 t_0 t0附近存在唯一的解析解(解析解指解函数可以展开成幂级数)

    • 证明存在唯一性,即已知有解析解,证明它唯一:

      • 将解 x ( t ) x(t) x(t) t 0 t_0 t0处泰勒展开, x ( t ) = x ( t 0 ) + x ′ ( t 0 ) ( t − t 0 ) + ⋯ + x ( n ) ( t 0 ) n ! ( t − t 0 ) n = x 0 + ∑ j = 1 ∞ c j ( t − t 0 ) j x(t)=x(t_0)+x'(t_0)(t-t_0)+\cdots+\frac{x^{(n)}(t_0)}{n!}(t-t_0)^n=x_0+\sum^{\infty}_{j=1}c_j(t-t_0)^j x(t)=x(t0)+x(t0)(tt0)++n!x(n)(t0)(tt0)n=x0+j=1cj(tt0)j是解析解所泰勒展开的幂级数,要证明解的唯一性只需证明 c 1 , c 2 , ⋯   , c n c_1,c_2,\cdots,c_{n} c1,c2,,cn是惟一的
      • 因为 x ( t ) x(t) x(t)是方程 d x d t = f ( t , x ) \frac{dx}{dt}=f(t,x) dtdx=f(t,x)的解即满足 d x d t = f ( t , x ( t ) ) \frac{dx}{dt}=f(t,x(t)) dtdx=f(t,x(t)) c 1 = x ′ ( t 0 ) = f ( t 0 , x ( t 0 ) ) = f ( t 0 , x 0 ) c_1=x'(t_0)=f(t_0,x(t_0))=f(t_0,x_0) c1=x(t0)=f(t0,x(t0))=f(t0,x0),因为 f ( t , x ) f(t,x) f(t,x)是可解析函数,根据多元函数的泰勒展开 f ( t , x ) f(t,x) f(t,x)在点 ( t 0 , x 0 ) (t_0,x_0) (t0,x0)处的泰勒展开为 f ( t , x ) = f ( t 0 , x 0 ) + ( t − t 0 ) f t ′ ( t 0 , x 0 ) + ( x − x 0 ) f x ′ ( t 0 , x 0 ) + 1 2 ! ( t − t 0 ) 2 f t t ′ ′ ( t 0 , x 0 ) + 1 2 ! ( t − t 0 ) ( x − x 0 ) f x t ′ ′ ( t 0 , x 0 ) + 1 2 ! ( t − t 0 ) ( x − x 0 ) f t x ′ ′ ( t 0 , x 0 ) + 1 2 ! ( x − x 0 ) 2 f x x ′ ′ ( t 0 , x 0 ) + O n f(t,x)=f(t_0,x_0)+(t-t_0)f'_t(t_0,x_0)+(x-x_0)f'_x(t_0,x_0)+\frac{1}{2!}(t-t_0)^2f''_{tt}(t_0,x_0)+\frac{1}{2!}(t-t_0)(x-x_0)f''_{xt}(t_0,x_0)+\frac{1}{2!}(t-t_0)(x-x_0)f''_{tx}(t_0,x_0)+\frac{1}{2!}(x-x_0)^2f''_{xx}(t_0,x_0)+O^n f(t,x)=f(t0,x0)+(tt0)ft(t0,x0)+(xx0)fx(t0,x0)+2!1(tt0)2ftt(t0,x0)+2!1(tt0)(xx0)fxt(t0,x0)+2!1(tt0)(xx0)ftx(t0,x0)+2!1(xx0)2fxx(t0,x0)+On,因此 c 1 = f ( t 0 , x 0 ) + 0 = f ( t 0 , x 0 ) c_1=f(t_0,x_0)+0=f(t_0,x_0) c1=f(t0,x0)+0=f(t0,x0),即 f ( t , x ) = ∑ k , l = 0 ∞ γ k l ( t − t 0 ) k ( x − x 0 ) l f(t,x)=\sum^{\infty}_{k,l=0}\gamma_{kl}(t-t_0)^k(x-x_0)^l f(t,x)=k,l=0γkl(tt0)k(xx0)l k = l = 0 k=l=0 k=l=0的情况(此时只有常数项),而 c 2 = 1 2 ! x ′ ′ ( t 0 ) = 1 2 ! [ f t ′ ( t 0 , x ( t 0 ) ) + f x ′ ( t 0 , x ( t 0 ) ) f ( t 0 , x 0 ) ] c_2=\frac{1}{2!}x''(t_0)=\frac{1}{2!}[f'_t(t_0,x(t_0))+f'_x(t_0,x(t_0))f(t_0,x_0)] c2=2!1x(t0)=2!1[ft(t0,x(t0))+fx(t0,x(t0))f(t0,x0)] x ′ ′ ( t ) x''(t) x(t)相当于 x ′ ( t ) x'(t) x(t) t t t求导,即 f ( t , x ( t ) ) f(t,x(t)) f(t,x(t)) t t t求导,根据链式法则, ∂ f ∂ t = ∂ f ∂ t + ∂ f ∂ x d x d t \frac{\partial f}{\partial t}=\frac{\partial f}{\partial t}+\frac{\partial f}{\partial x}\frac{dx}{dt} tf=tf+xfdtdx),对比式子 f ( t , x ) = ∑ k , l = 0 ∞ γ k l ( t − t 0 ) k ( x − x 0 ) l f(t,x)=\sum^{\infty}_{k,l=0}\gamma_{kl}(t-t_0)^k(x-x_0)^l f(t,x)=k,l=0γkl(tt0)k(xx0)l f ( t , x ) = f ( t 0 , x 0 ) + ( t − t 0 ) f t ′ ( t 0 , x 0 ) + ( x − x 0 ) f x ′ ( t 0 , x 0 ) + 1 2 ! ( t − t 0 ) 2 f t t ′ ′ ( t 0 , x 0 ) + 1 2 ! ( t − t 0 ) ( x − x 0 ) f x t ′ ′ ( t 0 , x 0 ) + 1 2 ! ( t − t 0 ) ( x − x 0 ) f t x ′ ′ ( t 0 , x 0 ) + 1 2 ! ( x − x 0 ) 2 f x x ′ ′ ( t 0 , x 0 ) + O n f(t,x)=f(t_0,x_0)+(t-t_0)f'_t(t_0,x_0)+(x-x_0)f'_x(t_0,x_0)+\frac{1}{2!}(t-t_0)^2f''_{tt}(t_0,x_0)+\frac{1}{2!}(t-t_0)(x-x_0)f''_{xt}(t_0,x_0)+\frac{1}{2!}(t-t_0)(x-x_0)f''_{tx}(t_0,x_0)+\frac{1}{2!}(x-x_0)^2f''_{xx}(t_0,x_0)+O^n f(t,x)=f(t0,x0)+(tt0)ft(t0,x0)+(xx0)fx(t0,x0)+2!1(tt0)2ftt(t0,x0)+2!1(tt0)(xx0)fxt(t0,x0)+2!1(tt0)(xx0)ftx(t0,x0)+2!1(xx0)2fxx(t0,x0)+On f ′ ( t 0 , x 0 ) f'(t_0,x_0) f(t0,x0)对应的 γ \gamma γ t − t 0 t-t_0 tt0 1 1 1,而 x − x 0 x-x_0 xx0 0 0 0的时候,即 f ′ ( t 0 , x 0 ) = γ 10 f'(t_0,x_0)=\gamma_{10} f(t0,x0)=γ10,以此类推, f x ′ ( t 0 , x 0 ) = γ 01 f'_x(t_0,x_0)=\gamma_{01} fx(t0,x0)=γ01,因此 c 2 = 1 2 ! ( γ 10 + γ 01 γ 00 ) c_2=\frac{1}{2!}(\gamma_{10}+\gamma_{01}\gamma_{00}) c2=2!1(γ10+γ01γ00),归纳推得 c n = P ( γ 00 , γ 00 , γ 01 , ⋯   , γ n − 1 , 0 ) c_n=P(\gamma_{00},\gamma_{00},\gamma_{01},\cdots,\gamma_{n-1,0}) cn=P(γ00,γ00,γ01,,γn1,0),其中 P P P是一个正系数多项式,因此每个 c k c_k ck都是惟一确定的公式计算得到的, Q . E . D . Q.E.D. Q.E.D.
    • 具体证明 C a u c h y Cauchy Cauchy定理方法

    • 推论:如果方程 d 2 x d t 2 = a ( t ) d x d t + b ( t ) x = 0 \frac{d^2x}{dt^2}=a(t)\frac{dx}{dt}+b(t)x=0 dt2d2x=a(t)dtdx+b(t)x=0的系数 a ( t ) a(t) a(t) b ( t ) b(t) b(t)在区间 ∣ t − t 0 ∣ < ρ |t-t_0|<\rho tt0<ρ内都可以展开成 ( t − t 0 ) (t-t_0) (tt0)的收敛的幂级数,则方程 d 2 x d t 2 + a ( t ) d x d t + b ( t ) x = 0 \frac{d^2x}{dt^2}+a(t)\frac{dx}{dt}+b(t)x=0 dt2d2x+a(t)dtdx+b(t)x=0在区间 ∣ t − t 0 ∣ < ρ |t-t_0|<\rho tt0<ρ上存在收敛的幂级数解 x ( t ) = ∑ j = 0 ∞ c j ( t − t 0 ) j x(t)=\sum^{\infty}_{j=0}c_j(t-t_0)^j x(t)=j=0cj(tt0)j

      • d 2 x d t 2 + a ( t ) d x d t + b ( t ) x = 0 \frac{d^2x}{dt^2}+a(t)\frac{dx}{dt}+b(t)x=0 dt2d2x+a(t)dtdx+b(t)x=0可以写成 [ x x ′ ] ′ = [ 0 1 − b ( t ) − a ( t ) ] [ x x ′ ] \begin{bmatrix}x\\x'\end{bmatrix}'=\begin{bmatrix}0&1\\-b(t)&-a(t)\end{bmatrix}\begin{bmatrix}x\\x'\end{bmatrix} [xx]=[0b(t)1a(t)][xx],而利用 C a u c h y Cauchy Cauchy定理可以证明高维的一阶的方程如果满足条件,仍然会有幂级数解
      • 为了计算其解 x ( t ) x(t) x(t),先把 a ( t ) a(t) a(t) b ( t ) b(t) b(t)展开成幂级数 a ( t ) = ∑ j = 0 ∞ a j ( t − t 0 ) j a(t)=\sum_{j=0}^{\infty}a_j(t-t_0)^j a(t)=j=0aj(tt0)j b ( t ) = ∑ j = 0 ∞ b j ( t − t 0 ) j b(t)=\sum_{j=0}^{\infty}b_j(t-t_0)^j b(t)=j=0bj(tt0)j,然后代入 d 2 x d t 2 + a ( t ) d x d t + b ( t ) x = 0 \frac{d^2x}{dt^2}+a(t)\frac{dx}{dt}+b(t)x=0 dt2d2x+a(t)dtdx+b(t)x=0,利用待定系数法,对应同幂次的系数,即可确定幂级数的系数 c 1 , ⋯   , c n c_1,\cdots,c_n c1,,cn
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

_森罗万象

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付:¥1
获取中
扫码支付

您的余额不足,请更换扫码支付或充值

打赏作者

实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值