Jackson derives Gauss's Law for electrostatics by transforming the surface integral of the electric field due to a single point charge over a closed surface into the integral of the solid angle, demonstrating that the integral depends only on the charge enclosed by the surface.

$$\mathbf{E}\cdot\mathbf{n}\, da = q\frac{cos\theta}{r^2}\,da$$

$$\mathbf{E}\cdot\mathbf{n}\, da = q\, d\Omega$$

And, apparently, it is "easy to see" that

$$\oint_S \mathbf{E}\cdot\mathbf{n}\,da=\begin{cases}\begin{align}&4\pi q & ~~~~~~~~~~~~~~~~~~~~&\text{if q lies inside S} \\&0 & ~~~~~~~~~~~~~~~~~~~~~& \text{if q lies outside S}\\\end{align}\end{cases}$$

Now, intuitively, this is pretty obvious, but I have no idea how to demonstrate that the integral of the solid angle of some closed surface at a point outside the surface is equal to zero. Even inside the surface, I wouldn't know how to show that the integral is equal to $4\pi$ for an implicit function $f(r,\theta,\phi)=c$ where I can't just use the spherical Jacobian transformation.

I'd like to be able to somehow generically parameterize a closed surface, or find a generic Jacobian for the surface area element, but it's just really not clear where to begin. I'd like to be able to show that

$$\frac{\hat{r}\cdot\hat{n}\,da}{r^2} = \nabla\times A$$

when the point P is outside the surface, but no approach is presenting itself.


  • 1
    $\begingroup$ Is this for a single point charge, multiple point charges, or a continuous charge distribution? Do we know that for a point charge $\nabla\cdot \mathbf E = 0$ away from the point? $\endgroup$ – Ted Shifrin Jul 21 '16 at 17:36
  • $\begingroup$ Edited. A single point charge. $\nabla\cdot E$ is obvious equal to zero when $\rho=0$, but he does not rely on that for this derivation. He shows that surface integral over any surface around the point charge is equal to $4\pi q$, and then uses the divergence theorem to show that $\nabla\cdot E = 4\pi\rho$. $\endgroup$ – JAustin Jul 21 '16 at 17:39
  • $\begingroup$ OK, so you're sure that he's not explicitly calculating the divergence for the inverse square force field? From what you've written, it is easy to see that the flux of $\mathbf E$ becomes the integral $\int q\,d\Omega$, and, if you understand the definition of solid angle, this is just the surface integral over the unit sphere of $q\,d\sigma$ (where $d\sigma$ is the surface area element of the sphere), so this gives you $4\pi q$. $\endgroup$ – Ted Shifrin Jul 21 '16 at 17:49
  • 1
    $\begingroup$ Correct, Jacob, and that's why I don't like this approach. I prefer to calculate the divergence explicitly for the case of the inverse-square force and then apply the divergence theorem to conclude that the flux agrees with the flux across a small sphere inside. $\endgroup$ – Ted Shifrin Jul 21 '16 at 18:18
  • 1
    $\begingroup$ By the way, you will never find a vector potential, because of the singularity. You would need divergence $0$ everywhere to get that. $\endgroup$ – Ted Shifrin Jul 21 '16 at 18:24

Jacob, you probably won't like this, but the easiest way I see to do the calculation rigorously is to use differential forms. (One reference that's somewhat accessible is my textbook Multivariable Mathematics ..., but you can find plenty of others.)

If $S$ is a closed (oriented) surface in $\Bbb R^3$ not containing the origin and $S^2$ is the unit sphere, then we consider the mapping $f\colon S\to S^2$ given by $f(\mathbf x)=\dfrac{\mathbf x}{\|\mathbf x\|}=\dfrac{\mathbf x}r$. Let $\omega = x\,dy\wedge dz + y\,dz\wedge dx + z\,dx\wedge dy$ be the area $2$-form on $S^2$. (We can also think of $\omega$ as the restriction to the unit sphere of the $2$-form $$\eta =\frac{x\,dy\wedge dz + y\,dz\wedge dx + z\,dx\wedge dy}{r^3}$$ on $\Bbb R^3-\{0\}$. This is the $2$-form corresponding to the electric field of a point charge with strength $1$ at the origin.) It follows from the change of variables theorem that $$\int_S f^*\omega = \int_{S^2}\omega = 4\pi.$$ (This takes care of the subtle cancellation issues you were worrying about when $f$ is not one-to-one. Because $S$ is a smooth surface, the projection map $f$ has degree $1$.)

The crucial calculation (which embodies the surface area statement in your question) is this: $f^*\omega = \eta$. To verify this, you'll need to know that $f^*(\mathbf x) = f(\mathbf x) = \dfrac{\mathbf x}r$, and $f^*(d\mathbf x) = df = \dfrac{d\mathbf x}r - \dfrac{\mathbf x}{r^2}dr$. So (take a deep breath), remembering that $\wedge$ is skew-symmetric: \begin{align*}f^*\omega &= \frac xr\left(\frac{dy}r-\frac y{r^2}dr\right)\wedge\left(\frac{dz}r-\frac z{r^2}dr\right)+\frac yr\left(\frac{dz}r-\frac z{r^2}dr\right)\wedge\left(\frac{dx}r-\frac x{r^2}dr\right)+\\& \hspace{1.5in}\frac zr\left(\frac{dx}r-\frac x{r^2}dr\right)\wedge\left(\frac{dy}r-\frac y{r^2}dr\right)\\ &= \eta - \frac1{r^4}\left(xz\,dy\wedge dr+xy\,dr\wedge dz + xy\,dz\wedge dr + yz\,dr\wedge dx + yz\,dx\wedge dr + xz\,dr\wedge dy\right) \\ &= \eta,\end{align*} as required.

For further motivation, in spherical coordinates $\eta = \sin\phi\,d\phi\wedge d\theta$ (using mathematicians' convention that $\phi$ is the angle from the positive $z$-axis). But you still ultimately need to use my remark above that the degree of $f$ is $1$.

  • $\begingroup$ Why "This takes care of the subtle cancellation issues you were worrying about when $f$ is not one-to-one. Because $S$ is a smooth surface, the projection map $f$ has degree 1"? Why can we use change of variables even if it is not one-to-one? $\endgroup$ – Marco Nervo Mar 1 '19 at 17:50
  • 1
    $\begingroup$ @MarcoAll-inNervo: Differential forms are more powerful than just plain integration of functions. The general fact is that if $f\colon M\to N$ is a smooth mapping between oriented, compact $n$-dimensional manifolds and $\omega$ is an $n$-form on $N$, then $$\int_M f^*\omega = \deg(f)\int_N \omega.$$ The integral keeps track of the regions where $f$ is orientation-preserving and -reversing. $\endgroup$ – Ted Shifrin Mar 1 '19 at 18:19
  • $\begingroup$ Ok, that's make sense to me. Why does $f$ have degree 1? $\endgroup$ – Marco Nervo Mar 1 '19 at 18:22
  • 1
    $\begingroup$ @MarcoAll-inNervo: When $S$ is a smooth closed surface with the origin in its interior, projection to a sphere centered at the origin always has degree $1$. This does not mean that that the mapping is necessarily one-to-one; it means that a generic point on the sphere has one preimage point, counting multiplicities (some points may definitely have multiple preimage points). I can see this by applying Stokes's Theorem to the region $R$ between a small sphere $\Sigma$ centered at the origin and the surface $S$, noting that $\int_S\eta - \int_\Sigma\eta = \int_R d\eta = 0$. $\endgroup$ – Ted Shifrin Mar 1 '19 at 18:33

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.