§3 高阶微分方程
C1 线性微分方程的一般理论
1) n n n阶线性微分方程: y ( n ) + a 1 ( x ) y ( n − 1 ) + ⋯ + a n ( x ) y = f ( x ) y^{(n)}+a_1(x)y^{(n-1)}+\cdots+a_n(x)y = f(x) y(n)+a1(x)y(n−1)+⋯+an(x)y=f(x)
2)解的存在性定理: a i ( x ) , f ( x ) ∈ C [ a , b ] ⟹ ∀ t 0 ∈ [ a , b ] , ∀ f ( i ) ( t 0 ) a_i(x),f(x) \in C[a,b]\implies\forall t_0\in[a,b],\forall f^{(i)}(t_0) ai(x),f(x)∈C[a,b]⟹∀t0∈[a,b],∀f(i)(t0),存在唯一解
3)叠加原理: f 1 ( x ) , f 2 ( x ) , … , f n ( x ) f_1(x),f_2(x),\dots,f_n(x) f1(x),f2(x),…,fn(x)是齐次线性微分方程的解,则线性组合 ∑ i = 1 n c i f i ( x ) \sum_{i=1}^nc_if_i(x) ∑i=1ncifi(x)也是解
4)函数的相关性:
-
朗斯基行列式
:
W [ y 1 , y 2 , … , y n ] = ∣ y 1 y 2 ⋯ y n y 1 ( 1 ) y 2 ( 1 ) ⋯ y n ( 1 ) ⋮ ⋮ ⋱ ⋮ y 1 ( n − 1 ) y 2 ( n − 1 ) ⋯ y n ( n − 1 ) ∣ W[y_1,y_2,\dots,y_n] = \begin{vmatrix} y_1&y_2&\cdots&y_n\\ y_1^{(1)}&y_2^{(1)}&\cdots&y_n^{(1)}\\ \vdots&\vdots&\ddots&\vdots\\ y_1^{(n-1)}&y_2^{(n-1)}&\cdots&y_n^{(n-1)} \end{vmatrix} W[y1,y2,…,yn]=∣∣∣∣∣∣∣∣∣y1y1(1)⋮y1(n−1)y2y2(1)⋮y2(n−1)⋯⋯⋱⋯ynyn(1)⋮yn(n−1)∣∣∣∣∣∣∣∣∣ -
n n n个函数线性相关 ⟹ W ≡ 0 \implies W\equiv 0 ⟹W≡0
-
齐次线性微分方程的 n n n个解线性无关 ⟹ W ≠ 0 \implies W \ne 0 ⟹W=0
因此齐次线性微分方程的解的朗斯基行列式恒为0或恒不为0
-
齐次线性微分方程一定存在 n n n个线性无关的解
证明:取初值条件W = E
5)通解结构定理:若齐次线性微分方程组无关解为 y 1 , … , y n y_1,\dots,y_n y1,…,yn,则通解为 ∀ c i , ∑ i = 1 n c i y i \forall c_i,\sum_{i=1}^nc_iy_i ∀ci,∑i=1nciyi
证明:
- 由解的存在性定理,一定有n个无关解
- 由解叠加原理,该解是齐次线性微分方程的解
- 由于 J ( ϕ , … , ϕ ( n − 1 ) ; c 1 , … , c n ) = W ≠ 0 J(\phi,\dots,\phi^{(n-1)};c_1,\dots,c_n) = W\neq0 J(ϕ,…,ϕ(n−1);c1,…,cn)=W=0,故系数独立
- 该组解称为基本解组,构成一个 n n n维线性空间。 W ( x 0 ) = 1 W(x_0) = 1 W(x0)=1时,称标准基本解组
6)非齐次线性微分方程组的解 = 齐次通解+非齐次特解
- y 1 , y 2 y_1,y_2 y1,y2是非线性解,则 y 1 − y 2 y_1-y_2 y1−y2是齐次解
C2 直接解法
1)若方程 y ( n ) + a 1 ( x ) y ( n − 1 ) + ⋯ + a n ( x ) y = 0 y^{(n)}+a_1(x)y^{(n-1)}+\cdots+a_n(x)y = 0 y(n)+a1(x)y(n−1)+⋯+an(x)y=0中 a i ( x ) a_i(x) ai(x)均为实值函数,有复值解 y = ϕ + i ψ y = \phi+i\psi y=ϕ+iψ,则 y ˉ = ϕ − i ψ , R e y = ϕ , I m y = ψ \bar{y}=\phi-i\psi,\mathrm{Re}y = \phi,\mathrm{Im}y=\psi yˉ=ϕ−iψ,Rey=ϕ,Imy=ψ都是方程的解
2)若方程 y ( n ) + a 1 ( x ) y ( n − 1 ) + ⋯ + a n ( x ) y = u ( x ) + i v ( x ) y^{(n)}+a_1(x)y^{(n-1)}+\cdots+a_n(x)y = u(x)+iv(x) y(n)+a1(x)y(n−1)+⋯+an(x)y=u(x)+iv(x)中 a i ( x ) , u ( x ) , v ( x ) a_i(x),u(x),v(x) ai(x),u(x),v(x)均为实值函数,有复值解 y = U ( x ) + i V ( x ) y=U(x)+iV(x) y=U(x)+iV(x),则 U ( x ) U(x) U(x)和 V ( x ) V(x) V(x)分别是 y ( n ) + a 1 ( x ) y ( n − 1 ) + ⋯ + a n ( x ) y = u ( x ) y^{(n)}+a_1(x)y^{(n-1)}+\cdots+a_n(x)y = u(x) y(n)+a1(x)y(n−1)+⋯+an(x)y=u(x)和 y ( n ) + a 1 ( x ) y ( n − 1 ) + ⋯ + a n ( x ) y = v ( x ) y^{(n)}+a_1(x)y^{(n-1)}+\cdots+a_n(x)y = v(x) y(n)+a1(x)y(n−1)+⋯+an(x)y=v(x)的解
3)齐次常系数线性微分方程
-
特征方程
: y ( n ) + a 1 y ( n − 1 ) + ⋯ + a n y = 0 y^{(n)} + a_1y^{(n-1)}+\cdots+a_ny=0 y(n)+a1y(n−1)+⋯+any=0的特征方程为 λ n + a 1 λ n − 1 + ⋯ + a n = 0 \lambda^n+a_1\lambda^{n-1}+\cdots+a_n=0 λn+a1λn−1+⋯+an=0 -
特征根均为单实根: y = ∑ i = 1 n c i e λ i x y = \sum_{i=1}^nc_ie^{\lambda_ix} y=∑i=1ncieλix
朗斯基行列式为:
W = ∣ e λ 1 x e λ 2 x ⋯ e λ n x λ 1 e λ 1 x λ 2 e λ 2 x ⋯ λ n e λ n x ⋮ ⋮ ⋮ ⋮ λ 1 n − 1 e λ 1 x λ 2 n − 1 e λ 2 x ⋯ λ n n − 1 e λ n x ∣ = e ∑ i = 1 n λ i ∣ λ 1 λ 2 ⋯ λ n λ 1 2 λ 2 2 ⋯ λ n 2 ⋮ ⋮ ⋮ ⋮ λ 1 n − 1 λ 2 n − 1 ⋯ λ n n − 1 ∣ = e ∑ i = 1 n λ i ∏ 1 ≤ i < j ≤ n λ i − λ j \begin{aligned} W &= \begin{vmatrix} e^{\lambda_1x}&e^{\lambda_2x}&\cdots&e^{\lambda_nx}\\ \lambda_1e^{\lambda_1x}&\lambda_2e^{\lambda_2x}&\cdots&\lambda_ne^{\lambda_nx}\\ \vdots&\vdots&\vdots&\vdots\\ \lambda_1^{n-1}e^{\lambda_1x}&\lambda_2^{n-1}e^{\lambda_2x}&\cdots&\lambda_n^{n-1}e^{\lambda_nx} \end{vmatrix} = e^{\sum_{i=1}^n\lambda_i} \begin{vmatrix} \lambda_1&\lambda_2&\cdots&\lambda_n\\ \lambda_1^2&\lambda_2^2&\cdots&\lambda_n^2\\ \vdots&\vdots&\vdots&\vdots\\ \lambda_1^{n-1}&\lambda_2^{n-1}&\cdots&\lambda_n^{n-1} \end{vmatrix}\\ &=e^{\sum_{i=1}^n\lambda_i} \prod\limits_{1\le i\lt j\le n}\lambda_i-\lambda_j \end{aligned} W=∣∣∣∣∣∣∣∣∣eλ1xλ1eλ1x⋮λ1n−1eλ1xeλ2xλ2eλ2x⋮λ2n−1eλ2x⋯⋯⋮⋯eλnxλneλnx⋮λnn−1eλnx∣∣∣∣∣∣∣∣∣=e∑i=1nλi∣∣∣∣∣∣∣∣∣λ1λ12⋮λ1n−1λ2λ22⋮λ2n−1⋯⋯⋮⋯λnλn2⋮λnn−1∣∣∣∣∣∣∣∣∣=e∑i=1nλi1≤i<j≤n∏λi−λj
故 e λ i x e^{\lambda_ix} eλix线性无关 -
存在复根: k k k重复根 λ = α + i β \lambda = \alpha+i\beta λ=α+iβ的共轭复数也是 k k k重复根,因此得到 2 k 2k 2k个实值解:
e α t cos β x , x e α t cos β x , … , x k − 1 e α t cos β x e^{\alpha t}\cos \beta x ,xe^{\alpha t}\cos \beta x,\dots,x^{k-1}e^{\alpha t}\cos \beta x eαtcosβx,xeαtcosβx,…,xk−1eαtcosβx
e α t sin β x , x e α t sin β x , … , x k − 1 e α t sin β x e^{\alpha t}\sin \beta x ,xe^{\alpha t}\sin \beta x,\dots,x^{k-1}e^{\alpha t}\sin \beta x eαtsinβx,xeαtsinβx,…,xk−1eαtsinβx
证明:
-
先假设 λ = 0 \lambda = 0 λ=0
则 λ \lambda λ是 k k k重根意味着特征方程有因子 λ k \lambda^k λk,则 a n , … , a n − k + 1 = 0 a_n,\dots,a_{n-k+1}=0 an,…,an−k+1=0
故原微分方程化为 y ( n ) + a 1 y ( n − 1 ) + ⋯ + a n − k y ( k ) = 0 y^{(n)}+a_1y^{(n-1)}+\cdots+a_{n-k}y^{(k)}=0 y(n)+a1y(n−1)+⋯+an−ky(k)=0
显然有解 1 , x , x 2 , … , x n − 1 1,x,x^2,\dots,x^{n-1} 1,x,x2,…,xn−1
-
再假设 λ ≠ 0 \lambda\ne0 λ=0
作换元 y = z e λ x y = ze^{\lambda x} y=zeλx代入原微分方程( x ( m ) = e λ x ∑ i = 0 m C m i λ i y ( i ) x^{(m)}=e^{\lambda x}\sum_{i=0}^mC_m^i\lambda^iy^{(i)} x(m)=eλx∑i=0mCmiλiy(i)),
得新方程 z ( n ) + b 1 z ( n − 1 ) + ⋯ + b n z = 0 z^{(n)}+b_1z^{(n-1)}+\cdots+b_nz=0 z(n)+b1z(n−1)+⋯+bnz=0,有 k k k重特征根 0 0 0
-
4)非齐次常系数线性微分方程:
- 比较系数法:
- f ( x ) = P ( x ) e λ x , λ ∈ R , P ( x ) f(x) = P(x)e^{\lambda x},\lambda \in R,P(x) f(x)=P(x)eλx,λ∈R,P(x)为实系数多项式,则解为 x = t k Q ( m ) e λ x x = t^kQ(m)e^{\lambda x} x=tkQ(m)eλx, k k k为 λ \lambda λ的重数,若非特征根则重数为0, Q ( x ) Q(x) Q(x)也是次数不高于 P ( x ) P(x) P(x)的多项式,通过待定系数法,将 x x x代入方程后比较两端系数确定
- f ( x ) = ( A ( x ) cos β x + B ( x ) sin β x ) e α x , α , β f(x)=(A(x)\cos\beta x+B(x)\sin \beta x)e^{\alpha x},\alpha,\beta f(x)=(A(x)cosβx+B(x)sinβx)eαx,α,β为常数, A ( x ) , B ( x ) A(x),B(x) A(x),B(x)为实系数多项式。此时有解 f ( x ) = ( P ( x ) cos β x + Q ( x ) sin β x ) e α x , P ( x ) , Q ( x ) f(x)=(P(x)\cos\beta x+Q(x)\sin \beta x)e^{\alpha x},P(x),Q(x) f(x)=(P(x)cosβx+Q(x)sinβx)eαx,P(x),Q(x)是待定系数多项式,次数不高于 A ( x ) , B ( x ) A(x),B(x) A(x),B(x)
- 拉普拉斯变换法:
拉普拉斯变换
:复变函数 F ( s ) = ∫ 0 + ∞ f ( x ) e − s x d x F(s) = \int_0^{+\infin}f(x)e^{-sx}\mathrm{d}x F(s)=∫0+∞f(x)e−sxdx,定义于复平面 R e s > σ \mathrm{Re}\ s\gt\sigma Re s>σ。称为 f ( x ) f(x) f(x)的拉普拉斯变换,其中 f ( x ) f(x) f(x)在 x ≥ 0 x\ge 0 x≥0上有定义,且 ∣ f ( x ) ∣ < M e σ x , M , σ ∈ R + |f(x)|\lt Me^{\sigma x},M,\sigma \in R^+ ∣f(x)∣<Meσx,M,σ∈R+。 f ( x ) f(x) f(x)为原函数,而 F ( s ) F(s) F(s)为像函数- 解法:对方程两端实施拉普拉斯变换,解出像函数,再反变换求得原函数
- 非特殊形式,使用常数变易法
5)常数变易法:已知基本解组求非齐次微分方程通解
解得齐次解 y = ∑ i = 1 n c i y i y = \sum_{i=1}^nc_iy_i y=∑i=1nciyi,设 y = ∑ i = 1 n c i ( x ) y i y = \sum_{i=1}^nc_i(x)y_i y=∑i=1nci(x)yi
代入 y ( n ) + a 1 ( x ) y ( n − 1 ) + ⋯ + a n ( x ) y = f ( x ) y^{(n)}+a_1(x)y^{(n-1)}+\cdots+a_n(x)y = f(x) y(n)+a1(x)y(n−1)+⋯+an(x)y=f(x),可得关于 c i ( x ) c_i(x) ci(x)的一个方程式
为解得 n n n个 c i ( x ) c_i(x) ci(x),需要 n n n个方程式
y ( 1 ) = ∑ i = 1 n ( c i ( 1 ) ( x ) y i + c i ( x ) y i ( 1 ) ) y^{(1)} = \sum_{i=1}^n (c_i^{(1)}(x)y_i+c_i(x)y_i^{(1)}) y(1)=∑i=1n(ci(1)(x)yi+ci(x)yi(1)),令 ∑ i = 1 n c i ( 1 ) ( x ) y i = 0 \sum_{i=1}^nc_i^{(1)}(x)y_i = 0 ∑i=1nci(1)(x)yi=0,得 y ( 1 ) = ∑ i = 1 n c i ( x ) y i ( 1 ) ( x ) y^{(1)} = \sum_{i=1}^n c_i(x)y_i^{(1)}(x) y(1)=∑i=1nci(x)yi(1)(x)
再次求导得 y ( 2 ) = ∑ i = 1 n ( c i ( 1 ) ( x ) y i ( 1 ) + c i ( x ) y i ( 2 ) ) y^{(2)}=\sum_{i=1}^n(c_i^{(1)}(x)y_i^{(1)}+c_i(x)y_i^{(2)}) y(2)=∑i=1n(ci(1)(x)yi(1)+ci(x)yi(2)),令 ∑ i = 1 n c i ( 1 ) ( x ) y i ( 1 ) = 0 \sum_{i=1}^nc_i^{(1)}(x)y_i^{(1)}=0 ∑i=1nci(1)(x)yi(1)=0,得 y ( 2 ) = ∑ i = 1 n c i ( x ) y i ( 2 ) y^{(2)}=\sum_{i=1}^nc_i(x)y_i^{(2)} y(2)=∑i=1nci(x)yi(2)
同理,得到 y ( n − 1 ) = ∑ i = 1 n c i ( x ) y i ( n − 1 ) y^{(n-1)}=\sum_{i=1}^nc_i(x)y_i^{(n-1)} y(n−1)=∑i=1nci(x)yi(n−1), y ( n ) = ∑ i = 1 n ( c i ( 1 ) ( x ) y i ( n − 1 ) + c i ( x ) y i ( n ) ) y^{(n)}=\sum_{i=1}^n(c_i^{(1)}(x)y_i^{(n-1)}+c_i(x)y_i^{(n)}) y(n)=∑i=1n(ci(1)(x)yi(n−1)+ci(x)yi(n))
将 y , y ( 1 ) , … , y ( n ) y,y^{(1)},\dots,y^{(n)} y,y(1),…,y(n)代入 y ( n ) + a 1 ( x ) y ( n − 1 ) + ⋯ + a n ( x ) y = f ( x ) y^{(n)}+a_1(x)y^{(n-1)}+\cdots+a_n(x)y = f(x) y(n)+a1(x)y(n−1)+⋯+an(x)y=f(x),并注意到 y y y是齐次解,得 ∑ i = 1 n c i ( 1 ) ( x ) y i ( n − 1 ) = f ( x ) \sum_{i=1}^nc_i^{(1)}(x)y_i^{(n-1)}=f(x) ∑i=1nci(1)(x)yi(n−1)=f(x)
方程组:
{ ∑ i = 1 n c i ( 1 ) ( x ) y i = 0 ∑ i = 1 n c i ( 1 ) ( x ) y i ( 1 ) = 0 ⋯ ∑ i = 1 n c i ( 1 ) ( x ) y i ( n − 2 ) = 0 ∑ i = 1 n c i ( 1 ) ( x ) y i ( n − 1 ) = f ( x ) \begin{cases} \sum_{i=1}^n c_i^{(1)}(x)y_i = 0 \\ \sum_{i=1}^n c_i^{(1)}(x)y_i^{(1)} = 0 \\ \cdots \\ \sum_{i=1}^n c_i^{(1)}(x)y_i^{(n-2)} = 0 \\ \sum_{i=1}^n c_i^{(1)}(x)y_i^{(n-1)} = f(x) \end{cases} ⎩⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎧∑i=1nci(1)(x)yi=0∑i=1nci(1)(x)yi(1)=0⋯∑i=1nci(1)(x)yi(n−2)=0∑i=1nci(1)(x)yi(n−1)=f(x)
朗斯基行列式 W ≠ 0 W\ne 0 W=0,故可解得 c i ′ ( x ) c_i'(x) ci′(x),得 c i ( x ) = ∫ c i ′ ( x ) d x = ϕ i + γ i , γ i c_i(x) = \int c'_i(x)\mathrm{d}x=\phi_i+\gamma_i,\gamma_i ci(x)=∫ci′(x)dx=ϕi+γi,γi是任意常数代入得通解 y = ∑ i = 1 n ϕ i y i + ∑ i = 1 n γ i y i y = \sum_{i=1}^n \phi_iy_i +\sum_{i=1}^n \gamma_iy_i y=∑i=1nϕiyi+∑i=1nγiyi
C3 间接解法
1)降阶:
-
F ( x , y ( k ) , … , y ( n ) ) = 0 F(x,y^{(k)},\dots,y^{(n)})=0 F(x,y(k),…,y(n))=0,使用 z = y ( k ) z = y^{(k)} z=y(k)换元
-
F ( y , … , y ( n ) ) = 0 F(y,\dots,y^{(n)})=0 F(y,…,y(n))=0,即不显含 x x x,令 p = y ′ p = y' p=y′,视 y y y为自变量得 y ′ = p , y ′ ′ = d p d x = d p d y p , y ′ ′ ′ = p ( d p d y ) 2 + p 2 d p d y , … y'=p,y''=\frac{\mathrm{d}p}{\mathrm{d} x}=\frac{\mathrm{d}p}{\mathrm{d}y}p,y'''=p(\frac{\mathrm{d}p}{\mathrm{d}y})^2+p^2\frac{\mathrm{d} p}{\mathrm{d} y},\dots y′=p,y′′=dxdp=dydpp,y′′′=p(dydp)2+p2dydp,…,代入原方程,可降一阶
-
若已知齐次微分方程 k k k个线性无关的解,可降低 k k k阶
假设 y 1 , … , y k y_1,\dots,y_k y1,…,yk是 k k k个线性无关的解,令 y = y k z y = y_kz y=ykz
则 y ( m ) = ∑ i = 0 m C m i y k ( i ) z ( m − i ) y^{(m)} = \sum_{i=0}^mC_m^iy_k^{(i)}z^{(m-i)} y(m)=∑i=0mCmiyk(i)z(m−i),代入原齐次方程,其中 z z z的系数为0
使用换元 p = z ′ = ( y y k ) ′ p = z' = (\frac{y}{y_k})' p=z′=(yky)′,可将其降低一阶,同时,原剩下 k − 1 k-1 k−1解对应新解 p i = ( y i y k ) ′ , i = 1 , 2 , … , k − 1 p_i = (\frac{y_i}{y_k})',i=1,2,\dots,k-1 pi=(ykyi)′,i=1,2,…,k−1
因此,每多知道一个无关解,就可以将方程降低一阶
2)二阶微分方程的幂级数解法:
-
解法:设 y = ∑ n = 0 + ∞ a n x n y = \sum_{n=0}^{+\infin}a_nx^n y=∑n=0+∞anxn,有 y ′ = ∑ n = 0 + ∞ ( n + 1 ) a n x n , y ′ ′ = ∑ n = 0 + ∞ ( n + 2 ) a n x n y' = \sum_{n=0}^{+\infin}(n+1)a_nx^n,y''=\sum_{n=0}^{+\infin}(n+2)a_nx^n y′=∑n=0+∞(n+1)anxn,y′′=∑n=0+∞(n+2)anxn,代入原方程,求出 a n a_n an
-
对方程 y ′ ′ + p ( x ) y ′ + q ( x ) y = 0 y''+p(x)y'+q(x)y=0 y′′+p(x)y′+q(x)y=0,若 p ( x ) , q ( x ) p(x),q(x) p(x),q(x)都可展为收敛幂级数,则存在幂级数解 y = ∑ n = 0 + ∞ a n x n y=\sum_{n=0}^{+\infin}a_nx^n y=∑n=0+∞anxn
-
对方程 y ′ ′ + p ( x ) y ′ + q ( x ) y = 0 y''+p(x)y'+q(x)y=0 y′′+p(x)y′+q(x)y=0,若 x p ( x ) , x 2 q ( x ) xp(x),x^2q(x) xp(x),x2q(x)都可展为收敛幂级数,则存在幂级数解 y = x α ∑ n = 0 + ∞ a n x n , α y=x^\alpha\sum_{n=0}^{+\infin}a_nx^n,\alpha y=xα∑n=0+∞anxn,α是待定常数