I'm trying to solve an equation whose solutions I know are plane waves but there are a few nuances.

First, the equation is of the form $$ \partial^2_t \psi + A(r)\partial^2_r \psi +B(r) \partial_r \psi + C(r)\psi =0 $$

Second, my boundary conditions are $ \psi(a)=\psi(b)=0 $

Third, my "initial" conditions are: $$ \psi(t,r)=f(r) $$ and $$\frac{\partial \psi}{\partial t}(t,r) =g(r)$$

And finally, I'm integrating the equation backwards in time. I have a solution at a certain time $t$ and I need the solution at a much earlier time.

Given these conditions, can I just convert the partial derivatives into finite differences and proceed with the usual method applied to wave equations? Will this yield the correct results?

I apologize if this is a trivial question but I just want to confirm that I'm on solid footing before I begin developing the tedious algorithm.

  • 2
    $\begingroup$ Did you perhaps change notation halfway through the question? Is the field to-be-solved-for $\psi$ or $u$? And is position $r$ or $x$? A bit murky. $\endgroup$ Mar 2, 2018 at 19:58
  • $\begingroup$ @rchilton1980 That was really sloppy of me. Thanks a lot for pointing that out. I've corrected it now $\endgroup$
    – pkg
    Mar 3, 2018 at 8:13

1 Answer 1


Yes. You can. The wave equation is time reversal.

Your PDE, namely $$\partial_{tt}u-\Delta u=0$$ is invariant under the transformation $t\to -t$. This means that if $u(x,t)$ is a solution, $u(x,-t)$ is also a solution.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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