This is a Problem from Evans PDE 2nd Edition,Chapter 2 Problem 23. Let $S$ denote the square lying in $\Bbb R\times (0,\infty)$ with corners at the points $(0,1),(1,2),(0,3),(-1,2)$. Define $$f(x,t):=\left \{ \begin{array}{ll} \ -1, \text{ for } (x,t)\in S\cap \{ t>x+2\} \\ 1, \text{ for } (x,t)\in S\cap \{ t<x+2\} \\ \ 0, \text{ otherwise. } \end{array} \right.$$ Asume $u$ solves

$$\left \{ \begin{array}{ll} \ u_{ tt}-u_{ xx}=f, \text{ in } \Bbb R\times (0,\infty) \\ \ u=g,u_t=0,\text{ on } \Bbb R\times\{ t=0\}. \end{array} \right.$$ Describe the shape of $u$ for times $t>3$.


enter image description here

By Duhamel's principle, for any $(x, t)$, we have $$u(x,t) = \int^t_0 \int^{x + (t-s) }_{x-(t-s)} f(y,s)\, dy\, ds.$$ Refering to the picture, we see that for $t_0 \ge 3$, if $(x_0,t_0)$ is on the right hand side of the grey area, then the domain of dependence for $u(x_0,t_0)$ does not intersect the support of $f$ and thus $u(x_0,t_0) = 0$. Likewise, if $(x_0,t_0)$ is on the left hand side of the grey area then the domain of dependence of $u(x_0, t_0)$ contains the entire support of $f$ so since $f$ is balanced, the integral will be zero.

Because of this, the only ``interesting'' region is the grey region. As pictured, for $(x_0, t_0)$ in the grey region, $u(x_0, t_0)$ depends only on the length $d = d(x_0, t_0)$ of the orange line segment because it determines how much of the support of $f$ lies in the domain of dependence of $u(x_0,t_0)$ (the blue/red region is the intersection of the domain of dependence with the support of $f$). In this region we see that if $d = 0$, we should get $u(x_0,t_0) = 0$ and if $d = \sqrt 2$ then $u(x_0, t_0) = 0$. Further, as $d$ grows from $0$ to $\tfrac{\sqrt 2}2$ the value of $u$ is simply the amount of blue area. For $\tfrac{\sqrt 2}{2} \le d \le \sqrt 2$, $u$ is the blue area minus the red area. Calculating the area, we see $$u(x_0, t_0) = \left\{ \begin{matrix} d(x_0,t_0) \sqrt 2, & 0 \le d(x_0,t_0) \le \tfrac{\sqrt{2}}{2} \\ \tfrac{\sqrt 2}{2}\sqrt 2 - \sqrt 2(\tfrac{\sqrt 2}2 - (\sqrt 2 - d(x_0,t_0))), & \tfrac{\sqrt{2}}{2} \le d(x_0,t_0) \le \sqrt 2. \end{matrix} \right.$$ or simplifying a bit $$u(x_0, t_0) = \left\{ \begin{matrix} d(x_0,t_0) \sqrt 2, & 0 \le d(x_0,t_0) \le \tfrac{\sqrt{2}}{2} \\ 2 - d(x_0, t_0) \sqrt 2& \tfrac{\sqrt{2}}{2} \le d(x_0,t_0) \le \sqrt 2. \end{matrix} \right.$$ Pictorially, this $u$ looks like a triangular spike along the line between $t = x +1$ and $t = x+3$. Solving explicitly for $d(x_0, t_0)$ is not hard, but the algebra is somewhat ugly and the process is in no way edifying.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.