The Divergence (Gauss) Theorem

Every finite-volume simulation on the planet is the divergence theorem running in a loop. Computational fluid dynamics codes, weather models, and magnetohydrodynamics solvers store one field value per grid cell and update it by summing the fluxes through the cell's faces — Step 3 of the proof, executed trillions of times per second on a GPU cluster. The scheme conserves mass, charge, and energy to machine precision precisely because the interior-face cancellation is built into the arithmetic: whatever leaves one cell enters its neighbour, bit for bit. When plasma physicists battle the numerical sin of $\nabla\cdot\mathbf{B}\neq0$ in fusion and astrophysics codes, the fix ("constrained transport") is again the theorem, enforced cell by cell. The identity is just as alive in theory. Poynting's theorem — the energy audit of the electromagnetic field you will meet in E7 — is the divergence theorem applied to $\mathbf{E}\times\mathbf{B}$; the Maxwell stress tensor turns force-on-charges calculations into surface integrals; and in gravitational physics the ADM mass of a spacetime is read off from a flux integral at infinity, the general-relativistic descendant of $\oint\mathbf{g}\cdot\mathrm{d}\mathbf{a}=-4\pi G M_{\mathrm{enc}}$.

Imagine the vector field as the velocity of water laced with hidden faucets and drains. The divergence at a point is the net rate at which water is created there per unit volume — positive for a faucet, negative for a drain. Now seal off any region of the flow with an imaginary closed surface and do the audit: every litre per second manufactured inside has nowhere to go but out through the skin, and every litre swallowed by a drain must be supplied through the skin. Total interior production equals net outward flow. That sentence is the divergence theorem; the mathematics only makes it precise. The mechanism behind it is a cancellation you can draw. Chop the volume into tiny cubes. Each interior face is shared by two adjacent cubes, and the outward normal of one is the inward normal of the other: whatever flux leaves cube A across that face enters cube B, so in the grand sum those two contributions kill each other exactly. Stack up millions of cells and the interior becomes a perfectly silent ledger — only the faces on the outer boundary have no partner to cancel against, and they alone survive.

The cancellation buys you something remarkable for free: the shape of the surface does not matter. Wrap a source in a neat sphere or in a lumpy potato-shaped surface — as long as the same faucets are inside, the same net flux comes out, because the volume between the two surfaces contains no sources and therefore contributes nothing. This surface-independence is what will make Gauss's law a practical tool rather than a curiosity: you get to choose the surface that makes the integral trivial.

Two consequences of this picture are worth carrying with you. First, the theorem is bookkeeping, not physics: it holds for any smooth vector field whatsoever — electric, magnetic, fluid, or purely abstract. Physics enters only when a law tells you what the divergence equals, as in $\nabla\cdot\mathbf{E}=\rho/\varepsilon_0$. Second, it belongs to a family. All three integral theorems of this chapter share one skeleton — $\int_{\text{region}}(\text{derivative of field}) = (\text{field data on the boundary})$ — with the fundamental theorem for gradients working on curves, the Stokes theorem on surfaces, and this one on volumes. In the language of differential forms they are literally one theorem, $\int_M \mathrm{d}\omega = \int_{\partial M}\omega$.

Everything in this article expands one line. Suppose $\mathbf{F}$ is continuously differentiable throughout a compact region $V$, bounded by the closed, piecewise-smooth surface $S=\partial V$ with outward-pointing area elements. Then

where: $\mathrm{d}\tau$ — volume element; $\mathrm{d}\mathbf{a}=\hat{\mathbf{n}}\,\mathrm{d}a$ — outward area element on $S$; $\nabla\cdot\mathbf{F}$ — the divergence, the local flux density. Read it as a conservation audit: the left side counts sources point by point, the right side catches everything at the border. The outward orientation is part of the theorem — flip the normals and you flip the sign. The deepest way to see why (1) must be true is to recall the coordinate-free definition of divergence, which you met when the operator was first built:

where: the limit shrinks the volume onto a point. Equation (2) is the divergence theorem in miniature — flux per unit volume at a point. The theorem proper just integrates this local statement back up over a finite region, with the interior cancellations doing the assembly work. For computations you will usually deploy the Cartesian recipe $\nabla\cdot\mathbf{F}=\partial_x F_x+\partial_y F_y+\partial_z F_z$, and the proof below shows the two definitions cooperating.

The workhorse corollary: Gauss's law in two costumes

Why does an E&M course care so much about a calculus identity? Because Maxwell's first equation is a statement about a divergence, and experiments and problem-solving live on surfaces. Suppose you accept the local law $\nabla\cdot\mathbf{E}=\rho/\varepsilon_0$ — the field's flux density at each point is the charge density there. The divergence theorem instantly globalizes it.

The same two-line manoeuvre generates a family of results you will use all track long. Applied to $\mathbf{F}=\phi\nabla\psi$ together with the product rule $\nabla\cdot(\phi\nabla\psi)=\phi\nabla^2\psi+\nabla\phi\cdot\nabla\psi$, the theorem yields Green's first identity, and antisymmetrizing in $\phi\leftrightarrow\psi$ gives Green's second identity,

where: $\phi,\psi$ — any twice-differentiable scalar fields. This unassuming identity is the engine behind the uniqueness theorems of electrostatics and the whole method of Green's functions in E2. Applied instead to the current density with charge conservation $\nabla\cdot\mathbf{J}=-\partial\rho/\partial t$, the theorem gives the integral continuity statement

where: $\mathbf{J}$ — current density and $Q_{\mathrm{enc}}$ — the charge inside $S$: net current leaving any closed surface equals the rate at which the enclosed charge decreases. This is charge conservation in its global form, and Kirchhoff's current law is its steady-state shadow.

Which statement answers which question

1 · Auditing both sides over a unit cube (confidence builder)

Verify the divergence theorem for $\mathbf{F}=xy\,\hat{\mathbf{x}}+2yz\,\hat{\mathbf{y}}+3zx\,\hat{\mathbf{z}}$ over the unit cube $0\le x,y,z\le 1$.

Step 1. The volume side. $\nabla\cdot\mathbf{F}=\partial_x(xy)+\partial_y(2yz)+\partial_z(3zx)=y+2z+3x$. Integrate over the cube using $\int_0^1 x\,\mathrm{d}x=\tfrac12$ for each variable in turn: $\int_V(\nabla\cdot\mathbf{F})\,\mathrm{d}\tau = \tfrac12+2\cdot\tfrac12+3\cdot\tfrac12 = \boxed{3}$.

Step 2. The surface side, face by face. On $x=1$ ($\mathrm{d}\mathbf{a}=\mathrm{d}y\,\mathrm{d}z\,\hat{\mathbf{x}}$): $\mathbf{F}\cdot\mathrm{d}\mathbf{a}=y\,\mathrm{d}y\,\mathrm{d}z$, giving $\tfrac12$. On $x=0$ the integrand $xy$ vanishes: $0$.

Step 3. On $y=1$ ($\mathrm{d}\mathbf{a}=\mathrm{d}x\,\mathrm{d}z\,\hat{\mathbf{y}}$): $2z\,\mathrm{d}x\,\mathrm{d}z$ integrates to $1$. On $y=0$: the integrand $2yz$ vanishes, $0$. On $z=1$: $3x\,\mathrm{d}x\,\mathrm{d}y$ gives $\tfrac32$; on $z=0$: $0$.

Step 4. Total flux: $\tfrac12+1+\tfrac32 = \boxed{3}$ — equal to the volume integral, as the theorem promises. Notice the economy: the volume side was one integral; the surface side was six. That asymmetry is exactly why you will usually use the theorem to trade toward whichever side is easier.

2 · Gauss's law inside a charged ball (exam level)

A solid sphere of radius $R=0.10\ \mathrm{m}$ carries charge density $\rho(r)=kr$ with $k=2.0\ \mu\mathrm{C/m^4}$. Find the total charge, and the electric field at $r=0.05\ \mathrm{m}$ and at the surface.

Step 1. Total charge, integrating in spherical shells $\mathrm{d}\tau = 4\pi r^2\,\mathrm{d}r$: $Q=\int_0^R kr\cdot4\pi r^2\,\mathrm{d}r = \pi k R^4 = \pi\times2.0\times10^{-6}\times(0.10)^4 = \boxed{6.3\times10^{-10}\ \mathrm{C}\approx 0.63\ \mathrm{nC}}$.

Step 2. By symmetry $\mathbf{E}=E(r)\,\hat{\mathbf{r}}$, so on a concentric Gaussian sphere the flux is $E\cdot4\pi r^2$ and the integral form of Gauss's law — equation (1) applied to $\mathbf{E}$ — gives $E(r) = Q_{\mathrm{enc}}(r)/(4\pi\varepsilon_0 r^2)$ with $Q_{\mathrm{enc}}(r)=\pi k r^4$.

Step 3. Simplify: $E(r) = \dfrac{\pi k r^4}{4\pi\varepsilon_0 r^2} = \dfrac{k r^2}{4\varepsilon_0}$ inside the ball — the field grows like $r^2$ because enclosed charge accumulates faster ($r^4$) than the surface dilutes it ($r^2$).

Step 4. Numbers. At $r=0.05\ \mathrm{m}$: $E = \dfrac{2.0\times10^{-6}\times(0.05)^2}{4\times8.854\times10^{-12}} = \boxed{1.4\times10^{2}\ \mathrm{V/m}}$. At the surface: $E = \dfrac{kR^2}{4\varepsilon_0} = \dfrac{2.0\times10^{-6}\times0.01}{3.54\times10^{-11}} = \boxed{5.6\times10^{2}\ \mathrm{V/m}}$ — and a cross-check from outside, $E=Q/(4\pi\varepsilon_0R^2)=6.3\times10^{-10}/(1.11\times10^{-10}\times0.01)$, gives the same $5.6\times10^{2}\ \mathrm{V/m}$.

Step 5. Where the theorem worked: no differential equation was solved anywhere. The divergence theorem converted the local law into a flux statement, symmetry pulled $E$ outside the integral, and the rest was arithmetic.

References. Griffiths, Introduction to Electrodynamics, 4th ed., §1.3.4–1.3.5 · Jackson, Classical Electrodynamics, 3rd ed., §I.5 and front-cover formula sheet · Purcell & Morin, Electricity and Magnetism, 3rd ed., §2.9–2.11 · Zangwill, Modern Electrodynamics, §1.4 · Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., §3.5 All numbers in this article are exact analytic constants: the flux $4\pi$ and the factor of $3$ in $\nabla\cdot\mathbf{r}=3$ (and hence $\oint_S\mathbf{r}\cdot\mathrm{d}\mathbf{a}=3V$) involve no empirical values. The worked-example results — both sides equal $3$ for the unit-cube audit of $\mathbf{F}=xy\,\hat{\mathbf{x}}+2yz\,\hat{\mathbf{y}}+3zx\,\hat{\mathbf{z}}$; $Q=0.63$ nC, $E=1.4\times10^2$ V/m at $r=5$ cm, and $E=5.6\times10^2$ V/m at the surface for $\rho=kr$ — are arithmetic consequences of the boxed formulas.

Builds on: The Divergence of a Vector Field · Surface Integrals and Flux · Volume Integrals · The Del Operator (Nabla) · Scalar and Vector Fields Leads to: The Curl (Stokes) Theorem · The Fundamental Theorem for Gradients · Second Derivatives and the Laplacian