Choosing a Gaussian Surface

The problem

Gauss's law is always true. It is only useful when you can pull \(E\) out of the integral, and that needs the right surface. Here are the three cases where it works, one where it does not, and the reasoning that tells them apart.

The requirement, stated precisely

\[\oint\mathbf{E}\cdot d\mathbf{A} = \frac{Q_{\rm enc}}{\varepsilon_0}\]

To get \(E\) out you need a surface on which, at every point, either

\(\mathbf{E}\) is constant in magnitude and parallel to \(d\mathbf{A}\) (so the integrand is \(E\,dA\)), or
\(\mathbf{E}\) is perpendicular to \(d\mathbf{A}\) (so that part contributes nothing).

Only the symmetry of the charge distribution can guarantee that. So the real question is never "what surface should I draw?" but "what symmetry does the charge have?"

Case 1: spherical symmetry

A solid insulating sphere, radius \(R = 0.1\) m, total charge \(Q = 5\) nC, uniformly distributed.

The charge looks the same from every direction, so \(\mathbf{E}\) must be radial with magnitude depending only on \(r\). Surface: a concentric sphere.

Outside (\(r > R\)): \(Q_{\rm enc} = Q\), so \(E\cdot4\pi r^2 = Q/\varepsilon_0\):

\[E = \frac{Q}{4\pi\varepsilon_0 r^2} \qquad\Rightarrow\qquad E(0.2\,\text{m}) = \frac{(8.99\times10^9)(5\times10^{-9})}{0.04} = 1.12\times10^{3}\ \text{N/C}\]

Identical to a point charge at the centre — the sphere's extent is invisible from outside.

Inside (\(r < R\)): only the charge within \(r\) counts, and it scales with volume: \(Q_{\rm enc} = Q(r/R)^3\). Then

\[E = \frac{Qr}{4\pi\varepsilon_0R^3} \qquad\Rightarrow\qquad E(0.05\,\text{m}) = \frac{(8.99\times10^9)(5\times10^{-9})(0.05)}{10^{-3}} = 2.25\times10^{3}\ \text{N/C}\]

Note the shape: linear rise inside, peak at the surface (\(4.49\times10^3\) N/C at \(r = R\)), \(1/r^2\) decay outside. And note that \(E\) at \(r = R/2\) exceeds \(E\) at \(r = 2R\) — a common surprise.

Case 2: cylindrical symmetry

An infinite line with charge per length \(\lambda\).

Symmetry: same along the line, same around it. \(\mathbf{E}\) is radial. Surface: a coaxial cylinder of radius \(r\), length \(L\).

The two end caps have \(\mathbf{E}\perp d\mathbf{A}\), contributing zero — this is the second escape clause earning its keep. The curved side gives \(E\cdot2\pi rL = \lambda L/\varepsilon_0\):

\[E = \frac{\lambda}{2\pi\varepsilon_0 r}\]

\(L\) cancels, as it must — the answer cannot depend on how much of the line you chose to enclose. That cancellation is a good check on any Gauss's-law calculation.

Case 3: planar symmetry

An infinite sheet with charge per area \(\sigma\).

\(\mathbf{E}\) points away from the sheet, same everywhere. Surface: a pillbox straddling it.

The sides contribute nothing (\(\mathbf{E}\perp d\mathbf{A}\)); the two faces each give \(EA\):

\[2EA = \frac{\sigma A}{\varepsilon_0} \qquad\Rightarrow\qquad E = \frac{\sigma}{2\varepsilon_0}\]

Independent of distance. On a conductor's surface, though, the field is zero on the inside, so only one face contributes and \(E = \sigma/\varepsilon_0\) — twice as large. Mixing these two up is the single most common error in this topic.

Case 4: where it fails

A uniformly charged cube.

Gauss's law still holds: the flux through any enclosing surface is \(Q/\varepsilon_0\), exactly. But there is no surface on which \(E\) is constant. A sphere around the cube has stronger field near the corners; a cube-shaped surface has \(\mathbf{E}\) non-perpendicular over most of its area.

You cannot factor \(E\) out, so you learn only the total flux — which you already knew. The field of a charged cube has no elementary closed form and must be integrated numerically.

This is the honest lesson: Gauss's law does not compute fields. It converts a symmetry you already possess into an answer. No symmetry, no answer.

The decision procedure

What symmetry does the charge have? (Not the surface — the charge.)
That symmetry dictates the direction of \(\mathbf{E}\) and what it can depend on.
Draw the surface matching that symmetry.
Check that unwanted parts contribute zero, and that irrelevant dimensions cancel.
If no symmetry survives step 1, use Coulomb's law and integrate.

Gauss's law · Electric flux · Coulomb's law · Conductors & insulators · E&M field explorer · Ampère's law — the identical logic for magnetism