I'm trying to determine the electrostatic potential caused by a specified charge density function: $$ \rho_c(\vec{r}) =\begin{cases} 1 & \vec{r} \in V\\ 0 & \text{otherwise} \end{cases} $$

The solution using Green's function is then: $$ \phi(\vec{r}) = \iiint_{V} \frac{\rho_c(\vec{r}')}{4 \pi \epsilon_0 \|\vec{r}-\vec{r}'\|} d \vec{r}' $$

I can understand how this integral can be easily evaluated so long as $\vec{r} \not\in V$ since there is no singularity in the integral, however I don't know how to resolve the integral if $\vec{r} \in V$, since this appears to cause the integral to become infinite/undefined.

Suppose for a concrete example that $V$ is an axis-aligned rectangular prism with opposing corners at $(x_0,y_0,z_0)$ and $(x_1,y_1,z_1)$. How do I find $\phi(\vec{r})$ inside of this box, and how do I generalize that procedure to any arbitrary $V$?

  • $\begingroup$ The singularity is integrable... $\endgroup$ Aug 31 at 11:46
  • $\begingroup$ I'd imagine it is since I can see one alternative solution to be to use the Green's function method to produce boundary conditions outside of $V$ for a Poisson PDE solver, however I don't know what mathematical tools I need to use to evaluate this integral "directly". Suppose I took the 1D case with $\rho_c(x) = 1$ for $x \in [-1, 1]$. I think this produces the integral $\phi(0) = \int_{-1}^1 \frac{1}{4\pi \epsilon_0 |x|} dx$, which doesn't converge, but using the "Poisson with BC's" approach gives $\phi(0) = \frac{6 \pi + \log(3)}{4 \pi \epsilon_0}$. $\endgroup$ Aug 31 at 12:38
  • $\begingroup$ In 1D there is no Green function, but in 2D there is and it is proportional to $\ln \sqrt{x^2+ y^2}$ which is locally integrable as well in $dxdy$ $\endgroup$ Aug 31 at 13:17

Here's a very simple example: what is the improper integral $\int_0^1\frac{dt}{\sqrt{t}}$? It is $\frac{1}{2}$ even though $\frac{1}{\sqrt{t}}\to \infty$ as $t\to 0^+$. The reason is that although the function is unbounded, it is not "bad enough" for the purposes off integration. This is what @Valter Moretti means by "the singularity is integrable". It goes off to $\infty$ slowly enough that it can be integrated over to yield a finite result. On the other hand, $\int_0^1\frac{dt}{t}=\infty$.

So, for integration purposes, mere unboundedness of the function alone is not enough to deduce anything about the finiteness of the integral. In $3$ dimensions, \begin{align} \int_{\|\mathbf{r}'\|\leq 1}\frac{dV'}{\|\mathbf{r}'\|}=\int_0^1\frac{1}{r'}4\pi r'^2\,dr'=4\pi\int_0^1r'\,dr' \end{align} is certainly finite. And in general, in $\Bbb{R}^n$, \begin{align} \int_{\|\xi\|\leq 1}\frac{1}{\|\xi\|^p}\,d^n\xi&=\int_0^1\frac{1}{r^p}A_{n-1}r^{n-1}\,dr=A_{n-1}\int_0^1\frac{dr}{r^{p+1-n}} \end{align} so this is finite if and only if $p+1-n<1$, if and only if $p<n$ (here $A_{n-1}=\frac{2\pi^{n/2}}{\Gamma(n/2)}$ is the surface area of the unit sphere $S^{n-1}\subset\Bbb{R}^n$).

  • $\begingroup$ hmm, so the 1D case is still unbounded? Or am I miss-understanding the second criteria since I can still directly solve the 1D Poisson's equation for $\rho_c = 1$ for $x \in [-1,1]$? $\endgroup$ Aug 31 at 12:48
  • $\begingroup$ @helloworld922 in the 1-D case it is not integrable as I mentioned: $\int_0^1\frac{dt}{t}=\infty$. In 3-dimensions there's no issues. I even gave the condition in $n$-dimensions for when such inverse-powers of distance yield finite integrals. (though for $1D$ solving Poisson's equation is trivial, you just integrate $u''(x)=\rho(x)$ twice (I'm ignoring the various constants)) $\endgroup$
    – peek-a-boo
    Aug 31 at 12:49
  • $\begingroup$ I suppose a follow-up question is why can I solve Poisson's equation "directly" by applying appropriate BC's at $x=\pm 2$ and find $\phi(0)$? $\endgroup$ Aug 31 at 12:51
  • $\begingroup$ what do you mean? $\endgroup$
    – peek-a-boo
    Aug 31 at 12:52
  • $\begingroup$ $\iint \frac{d^2 \phi(x)}{dx^2} dx dx = \phi(x)$, $\iint h(1-x) h(1+x) dx dx = \frac{1}{2} (2 c_0 x + (x - 1)^2 h(x-1) - (x+1)^2 h(x+1))+c_1$, then I just need to apply appropriate boundary conditions to find the constants of integration? $\endgroup$ Aug 31 at 13:03

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.