数学 旧 .com 迁移

偏微分方程:波动方程 / 高维初值问题

从旧 .com 全量搬运的历史内容,来源路径:/math/课程/偏微分方程/chapters/波动方程/高维初值问题/

迁移来源

初值问题 (三维情形)-球面平均法 (Kirchhoff) #

question
{utta2(ux1x1+ux2x2+ux3x3)=f(x,t),R3×(0,),u(x,0)=φ(x),xR3ut(x,0)=ψ(x),xR3.\begin{cases} u_{tt}-a^2\left(u_{x_1x_1}+u_{x_2x_2}+u_{x_3x_3}\right)=f(x,t),&\mathbb{R}^3\times (0,\infty),\\ u(x,0)=\varphi(x), & x\in\mathbb{R}^3\\ u_t(x,0)=\psi(x), & x\in\mathbb{R}^3. \end{cases}

由于高维的 \hyperref[Duhamel]{Duhamel 原理}也成立, 故我们只考虑 f=φ=0f=\varphi=0 的情形. 采用球面平均法.

考虑函数 h(x)h(x) 在以 xx 为心, 以 rr 为半径的球面上的平均值

I(x,r;h)=14πy=1h(x+ry)dSyI(x,r;h)=\frac{1}{4\pi}\iint_{|y|=1}h(x+ry)\text{d} S_y

下面的推导几个等式.

  • (1)
zRh(x+z)dz=0Rr2dry=1h(x+ry)dSy=0R4πr2I(x,r;h)dr.\begin{aligned} \iiint_{|z|\leqslant R} h(x+z)\text{d} z\\ =\int_0^R r^2\text{d} r\iint_{|y|=1}h(x+ry)\text{d} S_y\\ =\int_0^R 4\pi r^2 I(x,r;h)\text{d} r. \end{aligned}

该推导只用了变量替换.

  • (2)
Δ0R4πr2I(x,r;h)dr=zRΔxh(x+z)dz=zRΔzh(x+z)dz=z=Rh(x+z)nzdSz=z=Ri=13h(x+z)ziziRdSz=y=1R2i=13yih(x+Ry)xidSy=4πR2RI(x,R;h).\begin{aligned} \Delta\int_0^R 4\pi r^2 I(x,r;h)\text{d} r\\ =\iiint_{|z|\leqslant R}\Delta_x h(x+z)\text{d} z=\iiint_{|z|\leqslant R}\Delta_z h(x+z)\text{d} z\\ =\iint_{|z|=R}\frac{\partial h(x+z)}{\partial n_z}\text{d} S_z=\iint_{|z|=R}\sum\limits_{i=1}^3\frac{\partial h(x+z)}{\partial z_i}\frac{z_i}{R}\text{d} S_z\\ =\iint_{|y|=1}R^2\sum\limits_{i=1}^3 y_i\frac{\partial h(x+Ry)}{\partial x_i}\text{d} S_y\\ =4\pi R^2 \frac{\partial}{\partial R}I(x,R;h). \end{aligned}

其中 Δx\Delta_x 表示对 xx 的各个分量求偏导.

第二行: 设 u=x+zu=x+z, 则可由链式法则知 h(x+z)x=h(u)u1\frac{\partial h(x+z)}{\partial x}=\frac{\partial h(u)}{\partial u}\cdot 1. 从而 Δxh(x+z)=Δzh(x+z)\Delta_x h(x+z)=\Delta_z h(x+z).

第三行: 利用 \hyperref[Green]{Green 公式} 取 v=1v=1. 并用法向导数定义.

第四行: 换元, 其中 R2R^2R2dSy=dSzR^2\text{d} S_y=\text{d} S_z 给出.

  • (3)
RΔ0R4πr2I(x,r;h)dr=4πR2ΔI(x,R;h)\frac{\partial}{\partial R}\Delta\int_0^R 4\pi r^2 I(x,r;h)\text{d} r=4\pi R^2\Delta I(x,R;h)

其中将偏导与 Δ\Delta 换序后利用微积分基本定理.

  • (4)
Δ(rI(x,r;h))=2r2(rI(x,r;h))\Delta (rI(x,r;h))=\frac{\partial ^2}{\partial r^2}(rI(x,r;h))

结合 (2),(3) 可得到.

我们取 hhf=φ=0f=\varphi=0 时的解 u(x,t)u(x,t), 并定义

M(x,r,t)=rI(x,r;u)M(x,r,t)=rI(x,r;u)

则有

a22Mr2(4)=a2ΔM=a2r4πy=1Δu(x+ry,t)dSyuttΔu=0=r4πy=1utt(x+ry,t)dSy=Mtt.\begin{aligned} a^2\frac{\partial^2 M}{\partial r^2}\overset{=}{(4)}a^2\Delta M=\frac{a^2r}{4\pi}\iint_{|y|=1}\Delta u(x+ry,t) \text{d} S_y\\ \overset{=}{u_{tt}-\Delta u=0}\frac{r}{4\pi}\iint_{|y|=1}u_{tt}(x+ry,t)\text{d} S_y=M_{tt}. \end{aligned}

M(x,r,0)=rI(x,r,0)=0M(x,0,0)=0Mt(x,r,0)=rI(x,r;ψ)\begin{aligned} M(x,r,0)=rI(x,r,0)=0\\ M(x,0,0)=0\\ M_t(x,r,0)=rI(x,r;\psi) \end{aligned}

故对每一个 xx, M(x,r,t)M(x,r,t) 是半无界问题

{Mtta2Mrr=0,r>0,t>0M(x,0,t)=0,M(x,r,0)=0,Mt(x,r,0)=rI(x,r;ψ).\begin{cases} M_{tt}-a^2 M_{rr}=0, & r>0,t>0\\ M(x,0,t)=0, &\\ M(x,r,0)=0, &\\ M_t(x,r,0)=rI(x,r;\psi). & \end{cases}

解得 0rat0\leqslant r\leqslant at 时有

M(x,r,t)=12aatrat+rrI(x,R;ψ)dRM(x,r,t)=\frac{1}{2a}\int_{at-r}^{at+r}rI(x,R;\psi)\text{d} R

从而

u(x,t)=limr01rM(x,r,t)=limr012aratrat+rRI(x,R;ψ)dR=1a[rI(x,r;ψ)]r=at=tI(x,at;ψ)=t4πy=1ψ(x+aty)dSy.\begin{aligned} u(x,t)=\lim\limits_{r\to 0}\frac{1}{r}M(x,r,t)\\ =\lim\limits_{r\to 0}\frac{1}{2ar}\int_{at-r}^{at+r}RI(x,R;\psi)\text{d} R\\ =\frac{1}{a}[rI(x,r;\psi)]\big|_{r=at}=tI(x,at;\psi)\\ =\frac{t}{4\pi}\iint_{|y|=1}\psi(x+aty)\text{d} S_y. \end{aligned}

Sr(x)={x+y:y=at}S_{r}(x)=\{x+y:|y|=at\}.

u2=14πa2tSat(x)ψ(y)dSu_2=\frac{1}{4\pi a^2t}\iint_{S_{at}(x)}\psi(y)\text{d} S

再由 \hyperref[Duhamel]{Duhamel 原理}得到

u1(x,t)=t[14πa2tSat(x)φ(y)dS]u_1(x,t)=\frac{\partial}{\partial t}\left[\frac{1}{4\pi a^2t}\iint_{S_{at}(x)}\varphi(y)\text{d} S\right] u3(x,t)=0t[14πa2(tτ)Sa(tτ)(x)f(y,τ)dS]dτu_3(x,t)=\int_0^t\left[\frac{1}{4\pi a^2(t-\tau)}\iint_{S_{a(t-\tau)}(x)}f(y,\tau)\text{d} S\right]\text{d} \tau

综上 u(x,t)=u1+u2+u3u(x,t)=u_1+u_2+u_3.

讨论

评论

正在加载评论...