A sample problem for an exam is as follows:

Consider the wave equation $U_{tt} = 4U_{xx}, 0 < x < 1$ with $U(0,t)= U(l,t)= 0$ and $U(x,0)= x(1-x)$, $U_t(x,0)= \pi$.

Find $U(1/4,1/4)$ and $U(1/2,1/2)$ using the reflection principle.

We went over the solution in lecture but I lost my notes and I'm unsure how to proceed. If someone could provide a solution with some explanation I would greatly appreciate it.

Note: $U_x$ and $U_t$ denote the partial derivative with respect to $x$ or $t$ respectively.


The reflection method consists in extending the problem to $\mathbb R\times [0,\infty)$ in such a way that the given boundary condition $u(0,t)=u(1,t)=0$ becomes a consequence of some symmetry of initial conditions $u(\cdot, 0)=g$ and $u_t(\cdot, 0)=h$. In order to make sure that there is no motion at $x=0$, we need $g(-x)=-g(x)$ and $h(-x)=-h(x)$. Similarly, to fix the point at $x=1$ we need $g(2-x)=-g(x)$ and $h(2-x)=-h(x)$. Thus, both $g$ and $h$ are $2$-periodic functions. Their period from $-1$ to $1$ is shown below: red is the initial position $g$, blue is the velocity $h$.

Initial data

Now it remains to apply d'Alembert's formula with propagation speed $c=2$. $$u(x,t) = \frac12 (g(x-2t) + g(x+2t)) + \frac{1}{4} \int_{x-2t}^{x+2t} h(s)\,ds$$ The piecewise nature of $g$ and $h$ would make it awkward to write out these integrals in general, but it's not hard to evaluate them at particular points $(x,t)$, which is what you asked to do.


Here is a reference where the method is explained in detailes. Starting on page $19$.

  • $\begingroup$ That reference isn't helpful, because while the spatial domain of the problem stated by rmh52 is $x \in [0,\ell]$, the spatial domain in the reference is $x \in [0,\infty)$. $\endgroup$
    – Ron Gordon
    Dec 31 '12 at 16:27
  • $\begingroup$ @rlgordonma The reference is at least somewhat helpful, because the method can be adapted to bounded intervals. I posted an answer. $\endgroup$
    – user53153
    Dec 31 '12 at 18:12
  • $\begingroup$ Perhaps. But it just seems that the problem is more naturally solved by Fourier decomposition, which is more or less equivalent to your formulation. $\endgroup$
    – Ron Gordon
    Dec 31 '12 at 18:48
  • $\begingroup$ @rlgordonma If you wanted a general formula for $U(x,t)$, the Fourier series is a natural tool indeed. Here we just want a numeric value of $U(1/4,1/4)$, which according to my answer is $\frac{1}{2}(g(-1/4)+g(3/4))+\frac14 \int_{-1/4}^{3/4} h = 0 +\frac14 (\pi/2)=\pi/8$. Compare this effort with expanding $x(1-x)$ into a Fourier series, and then trying to sum the infinite series back to find what $U(1/4,1/4)$ actually is. $\endgroup$
    – user53153
    Dec 31 '12 at 19:04
  • $\begingroup$ @PavelM: Thanks for replying to his comment. $\endgroup$ Dec 31 '12 at 22:16

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.