I know that this question has already been asked multiple times but I´m still not getting on the mathematical details behind the answers... So I hope that this question doesn´t get closed.

First I will consider the case of discrete charges (I suppose that there are $M$ charges inside the volume and $N-M$ charges outside the volume; hence a total of $N$ charges):

$$\begin{align} \oint_{\partial\Omega}\vec{E}\cdot\vec{dA} & =\int_{\Omega}\nabla\cdot \vec{E} dV \\ & = {1\over 4\pi\epsilon_0}\sum_{k=1}^{N}q_k\int_{\Omega}\nabla\cdot({\vec{r}-\vec{r_k}\over |\vec{r}-\vec{r_k}|^3})dV \\ & = {1\over 4\pi\epsilon_0}\sum_{k=1}^{M}q_k\int_{\Omega}\nabla\cdot({\vec{r}-\vec{r_k}\over |\vec{r}-\vec{r_k}|^3})dV + {1\over 4\pi\epsilon_0}\sum_{k=M+1}^{N}q_k\int_{\Omega}\nabla\cdot({\vec{r}-\vec{r_k}\over |\vec{r}-\vec{r_k}|^3})dV \end{align}$$

Note that in the second term of the last equality is zero because we are just integrating over $\Omega$ and in this term the charges are outside $\Omega$ hence: $\int_{\Omega}\nabla\cdot({\vec{r}-\vec{r_k}\over |\vec{r}-\vec{r_k}|^3})dV=0$

Know we can compute the first term of the last equality using the fact that: $\nabla\cdot({\vec{r}-\vec{r_k}\over |\vec{r}-\vec{r_k}|^3})=4\pi\delta(\vec{r}-\vec{r_k})$

$${1\over 4\pi\epsilon_0}\sum_{k=1}^{M}q_k\int_{\Omega}\nabla\cdot({\vec{r}-\vec{r_k}\over |\vec{r}-\vec{r_k}|^3})dV={1\over \epsilon_0}\sum_{K=1}^{M}q_k\int_{\Omega}\delta(\vec{r}-\vec{r_k})dV={1\over \epsilon_0}\sum_{k=1}^{M}q_k$$

(In the last equality I´m not sure that I can use the fact that $\int_{\Omega}\delta(\vec{r}-\vec{r_k})dV=1$ because I´m not integrating over all the space)

Now if we want to derive the differential form: $$\int_{\Omega}\nabla\cdot \vec{E}dV={1\over \epsilon_0}\sum_{k=1}^{M}q_k={1\over \epsilon_0}\sum_{k=1}^{M}q_k\int_{\Omega}\delta(\vec{r}-\vec{r_k})d^3\vec{r}={1\over \epsilon_0}\int_{\Omega}\rho(\vec{r})d^3\vec{r}$$

And then we conclude that $$\nabla\cdot E={1\over \epsilon_0}\rho$$ (becuase the equality of integrals is valid for all region $\Omega$)

So if everthing is right then that is gauss law for discrete charges; Now what happens if a have a continuous charge that is inside $\Omega$? (for example a solid charged cone inside a sphere) this is the part where I´m having trouble:

The electric field of a continuous dsitribution is given by $$E(\vec r)={1\over 4\pi\epsilon_0}\int_{\Omega´}\rho(\vec s)({\vec{r}-\vec{s}\over |\vec{r}-\vec{s}|^3})d^3\vec{s}$$

where $\Omega´$ is the charged region in the space we have that $$\begin{align} \nabla\cdot{E(\vec{r})} & = {1\over 4\pi\epsilon_0}\int_{\Omega´}\rho(\vec s)\nabla \cdot({\vec{r}-\vec{s}\over |\vec{r}-\vec{s}|^3})d^3\vec{s} (\text{differentiation under the integral sign})\\ & = {1\over \epsilon_0}\int_{\Omega´}\rho(\vec{s})\delta(\vec{r}-\vec{s})d^3\vec{s} \\ & = {1\over \epsilon_0}\rho(\vec{r}) (\text{shifting property of delta function}) \end{align}$$

(I have a big issue in this part becuase the density $\rho$ is just defined in the charged solid region $\Omega´$ so it doesn´t make sense to talk about the density in all the space;also i´m not sure if the last equality is valid because I´m not integrating over all the space)

Know assuming that it makes sense to talk about $\rho$ in all the space then if we want to obtain the integral form we use the divergence theorem: $$\oint_{\partial \Omega}E\cdot \vec{dS}=\int_{\Omega}\nabla\cdot\vec{E}dV={1\over \epsilon_0}\int_{\Omega}\rho(\vec{r})dV={Q_{int}\over \epsilon_0}$$

I would really appreciate if you can help me with the "details" of the proof, also any comments or suggestion would be highly appreciated.


1 Answer 1


I'd have preferred to answer through a comment, but I'm not yet allowed to do so. Your proof is entirely correct. The $n$-dimensional delta distribution is in fact defined as follows:

$$ \int_{\Omega}\ d^{n}x\ \ \delta^{(n)}(\vec{x}-\vec{x}_{0})=\begin{cases}1&\vec{x}_{0}\in\Omega\\0&\vec{x}_{0}\notin\Omega\end{cases} $$ and it acts on functions as $$ \int_{\Omega}\ d^{n}x\ \ f(\vec{x})\ \delta^{(n)}(\vec{x}-\vec{x}_{0})=\begin{cases}f(\vec{x}_{0})&\vec{x}_{0}\in\Omega\\0&\vec{x}_{0}\notin\Omega\end{cases} $$ so it doesn't matter whether or not you are integrating over all space. Good job writing all in detail anyway.


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.