So this question says to take $u(x,t) = v(x^2/t)$ to solve the 1d heat equation. That is, $$ u_t = u_{xx} $$ and it gives the general solution in the form $$ v(z) = c\int_{0}^z e^{-s/4}s^{-1/2} ds + d $$ (you can verify that this indeed is a solution to the heat equation). Anyway, the question involves choosing $c$ such that this solution is the fundamental solution, so that the initial condition $\Phi(x,0) = \delta(x)$ holds. I'm not sure how to go about doing this, can you perhaps point me in the right direction?

I think that $c$ solves $$ c = \frac{1}{\int_{-\infty}^\infty e^{-s^2/4t}\left(\frac{s^2}{t}\right)^{-1/2} ds } $$ Can someone verify this/explain why this?


The heat equation (also known as diffusion equation) conserves total mass, which by definition is the integral $M (t) = \int_{-\infty}^\infty u(x,t)\,dx$. (This can be proved by taking time derivative of $M$ and using the PDE.) Since $\delta$ is a unit mass, in order to satisfy $u(x,0)=\delta(x)$ we need $M$ to be $1$.

This is why we normalize the fundamental solution $$u(x,t)=\frac{c}{\sqrt{t}}e^{-x^2/(4t)}$$ so that its mass is $1$. Which leads to $$c^{-1} = \int_{-\infty}^{\infty} e^{-x^2/4}\,dt =\sqrt{4\pi}$$ Above, the mass is computed at time $t=1$ and equated to $1$. It could be computed in any other moment $t_0>0$; the result is the same.

  • $\begingroup$ Shouldn't the last integral, be an integral over space and not time? $\endgroup$ – Tucker May 29 '17 at 5:04

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.