The Curl of the Electrostatic Field

Electrostatics stands on exactly two field equations. Gauss's law, $\nabla\cdot\mathbf{E}=\rho/\varepsilon_0$ , you have just seen put to work; it fixes how field lines begin and end — on charge. This article delivers the second, quieter equation: $\nabla\times\mathbf{E}=\mathbf{0}$ . It fixes what field lines never do — close on themselves. Where the divergence encodes the inverse-square strength of Coulomb's law, the vanishing curl encodes its other half: the force is central, pointing along the line joining the charges. Two derivative statements, and between them the whole of Coulomb's law is spent. The result sounds austere but it is the most consequential sentence in the chapter, because of what it licenses. A field with zero curl has zero circulation around every loop: carrying a charge around any closed path costs nothing, the electrostatic force is conservative, and — by the machinery built in Scalar Potentials — a single scalar function $V$ exists with $\mathbf{E}=-\nabla V$ . Three field components collapse into one function. Every volt on every battery, every equipotential map, every boundary-value trick of the next two chapters cashes this one check. The proof is honest and short: the field of one point charge is radial with a $1/r^2$ magnitude, so its line integral depends only on the endpoint radii — no path memory, no circulation — and superposition hands the property to every static field for free. We also flag, honestly, the escape clause: let a magnetic field change in time and Faraday's law switches the curl back on. Statics is the regime where the potential's existence is guaranteed; enjoy it while it lasts.

Key takeaways

The electrostatic field of any static charge distribution is irrotational: $\nabla\times\mathbf{E}=\mathbf{0}$ everywhere.
Equivalently, $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0$ around every closed loop — the field is conservative, and round trips cost zero work.
The root cause is Coulomb's law being central and inverse-square: a point charge's field is $-\nabla(kq/r)$ , a pure gradient, and curls of gradients vanish identically; superposition extends this to all static fields.
Zero curl is exactly the license to define the electrostatic potential: $\mathbf{E}=-\nabla V$ , the subject of The Electrostatic Potential next.
Together, $\nabla\cdot\mathbf{E}=\rho/\varepsilon_0$ and $\nabla\times\mathbf{E}=\mathbf{0}$ are the complete differential laws of electrostatics — by Helmholtz, they determine the field.
Strictly a statics result: Faraday's law $\nabla\times\mathbf{E}=-\partial\mathbf{B}/\partial t$ revives the curl the moment magnetic fields change in time.
Key idea: because every point-charge field is the gradient of a scalar and curls of gradients vanish identically, superposition forces $\nabla\times\mathbf{E}=\mathbf{0}$ for any electrostatic field — equivalently $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0$ — which is exactly the condition for a potential $\mathbf{E}=-\nabla V$ to exist.

Where this lives today

Every circuit you have ever analyzed leans on this article. Kirchhoff's voltage law — the sum of voltage drops around any closed loop is zero — is $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0$ wearing schematic-diagram clothes, and it holds to the extent that the circuit's fields are quasi-static. Power engineers, chip designers, and everyone running SPICE simulations are betting on the curl of $\mathbf{E}$ being negligible at their frequencies — a bet that starts to lose exactly where this article's caveat says it must: RF engineers at gigahertz frequencies watch "the voltage between two points" turn path-dependent, ground loops in audio equipment hum at 50 Hz because a stray changing flux threads the loop, and probing a fast switching converter with a sloppy oscilloscope ground lead measures the loop's Faraday EMF as faithfully as the signal. The law also earns its keep as a validity check in computation. Electrostatic field solvers and the interpolation schemes of particle-in-cell plasma codes are audited by computing $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}$ around mesh loops: any nonzero residual is numerical error masquerading as physics, and "curl-free to machine precision" is a standard regression test in 2026 solver suites. Meanwhile the modern metrology of voltage itself — Josephson-junction standards defining the volt — only makes sense because, in the static limit, potential differences are path-independent quantities worth standardizing.

The physical picture

Picture the field lines of any collection of static charges: they stream out of positive charges, curve through space, and dive into negative ones — but they never form closed loops. That visual fact is the curl law. A nonzero curl would mean a little paddle wheel placed in the field gets torqued around; field lines circulating in closed rings would do exactly that. Electrostatic fields have sources and sinks but no whirlpools.

Figure

Allowed and forbidden field patterns — sources and sinks yes, closed circulation no

The energy version of the same statement is the one to internalize. Carry a test charge slowly around any closed path and tally the work the field does. If the tally came back positive, you could run the loop forever — a perpetual-motion machine powered by a stationary arrangement of charges. Conservation of energy demands that every round trip nets exactly zero, and that is precisely $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0$ . Zero curl is energy conservation, written locally. The deepest reason is geometric, and it lives one level down in Coulomb's law itself. The field of a single point charge points radially and its magnitude depends only on distance. Such a field is the slope of a hill — the hill of height $kq/r$ — and slopes of hills have no twist in them: walking any closed trail on a landscape returns you to your starting altitude. Centrality is doing the work here; the inverse square is almost incidental to this result (any central $f(r)\hat{\mathbf{r}}$ is curl-free — the $1/r^2$ is what Gauss's law spends). Superposition finishes the job: sums of twist-free fields are twist-free.

The two statements and the bridge between them

The law comes in a local and a global costume, and the Stokes theorem of E0 is the changing room. The local statement:

\nabla\times\mathbf{E}=\mathbf{0} \qquad (1)

where: $\mathbf{E}$ — the electrostatic field of any static charge distribution; the equation holds at every point of space, including inside charge distributions. It is one of the two field equations of electrostatics, the other being Gauss's law in differential form. The global statement:

\oint_{C}\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0 \qquad (2)

where: $C$ — any closed loop whatsoever; $\mathrm{d}\boldsymbol{\ell}$ — the directed line element along it. Physically: the work per unit charge done by the field around a round trip is zero. The two are equivalent by the Curl (Stokes) Theorem, $\oint_C\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=\int_S(\nabla\times\mathbf{E})\cdot\mathrm{d}\mathbf{a}$ : insert equation (1) and every circulation dies; conversely, if every circulation dies, shrinking loops around each point kills the curl pointwise. And the payoff statement, redeemed in full in The Electrostatic Potential:

\mathbf{E}=-\nabla V \qquad (3)

where: $V$ — a single-valued scalar field, existing because of equations (1)–(2) on the simply connected arena of electrostatics (the Scalar Potentials equivalence circle). One function will now carry all three field components.

Derivation AZero circulation from Coulomb's law done honestly

The goal is equation (2) — zero circulation — derived by actually computing the line integral of a point charge's field along an arbitrary path, not by waving at symmetry. The strategy: write $\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}$ in spherical coordinates centered on the charge, watch every angular piece drop out, and find that the integral remembers only the endpoint radii. Path independence and zero circulation then fall out together, and superposition donates the result to every electrostatic field. This is the calculation that makes the potential legal, so it is worth doing with all steps visible.

Step 1. Set up. A point charge $q$ sits at the origin; its field is $\mathbf{E}=\frac{kq}{r^2}\hat{\mathbf{r}}$ with $k=1/4\pi\varepsilon_0$ . A general path element in spherical coordinates is $\mathrm{d}\boldsymbol{\ell}=\mathrm{d}r\,\hat{\mathbf{r}}+r\,\mathrm{d}\theta\,\hat{\boldsymbol{\theta}}+r\sin\theta\,\mathrm{d}\phi\,\hat{\boldsymbol{\phi}}$ .

Step 2. Dot them. Only the radial pieces survive:

\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=\frac{kq}{r^2}\,\mathrm{d}r

Why the angular terms die: $\hat{\mathbf{r}}\cdot\hat{\boldsymbol{\theta}}=\hat{\mathbf{r}}\cdot\hat{\boldsymbol{\phi}}=0$ — the field is central, so sideways motion is always perpendicular to it and costs nothing. This is the precise point where "the force points along the line joining the charges" enters the physics.

Step 3. Integrate from $\mathbf{a}$ to $\mathbf{b}$ along any path:

\int_{\mathbf{a}}^{\mathbf{b}}\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=\int_{r_a}^{r_b}\frac{kq}{r^2}\,\mathrm{d}r=kq\left(\frac{1}{r_a}-\frac{1}{r_b}\right)

The wiggles of the path never appear — only the starting and ending distances from the charge. Why this is already path independence: two different paths between the same endpoints share $r_a$ and $r_b$ , hence share the integral.

Step 4. Close the loop: $\mathbf{b}=\mathbf{a}$ means $r_b=r_a$ , so $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=kq(1/r_a-1/r_a)=0$ for the point charge.

Step 5. Superpose. Any electrostatic field is a sum (or integral) of point-charge fields, and line integrals are linear: $\oint\mathbf{E}_{\mathrm{tot}}\cdot\mathrm{d}\boldsymbol{\ell}=\sum_i\oint\mathbf{E}_i\cdot\mathrm{d}\boldsymbol{\ell}=0$ . Every static distribution — dipoles, disks, thunderclouds — inherits zero circulation, term by term. Quick check: Step 3 evaluated between $r_a=0.10$ m and $r_b=0.25$ m with $q=5$ nC gives $44.95\times(10-4)=269.7$ V, the number reused in Worked example 1.

Figure

The round-trip audit — arcs are free, radial legs cancel, every loop nets zero

Go deeper · Derivation B · 1 minThe differential law and its Stokes consistency

Derivation BThe differential law and its Stokes consistency

The goal is the local statement (1), $\nabla\times\mathbf{E}=\mathbf{0}$ , obtained by exhibiting the point-charge field as a pure gradient and letting a vector identity do the killing; then a consistency pass through the Stokes theorem to confirm the global and local statements are one law, not two. The route is short, but each step leans on a specific E0 tool — this derivation is where the vector-calculus chapter pays rent.

Step 1. The gradient identity. For $\boldsymbol{\mathcal{R}}=\mathbf{r}-\mathbf{r}'$ (source at $\mathbf{r}'$ , field point at $\mathbf{r}$ , $\mathcal{R}=|\boldsymbol{\mathcal{R}}|$ ), a direct computation gives

\nabla\!\left(\frac{1}{\mathcal{R}}\right)=-\frac{\hat{\boldsymbol{\mathcal{R}}}}{\mathcal{R}^{2}}

Why: differentiate $\mathcal{R}=\sqrt{(x-x')^2+(y-y')^2+(z-z')^2}$ ; each component assembles into the unit vector over the square. This is the workhorse identity of the whole chapter — it will build the potential of every distribution.

Step 2. Recognize the field as a gradient. Coulomb's law says $\mathbf{E}=\frac{q}{4\pi\varepsilon_0}\frac{\hat{\boldsymbol{\mathcal{R}}}}{\mathcal{R}^2}$ , so by Step 1,

\mathbf{E}=-\nabla\!\left(\frac{q}{4\pi\varepsilon_0\,\mathcal{R}}\right)

The single-charge field is minus the slope of the scalar hill $kq/\mathcal{R}$ — the potential, arriving one article early.

Step 3. Kill the curl with the identity $\nabla\times(\nabla f)\equiv\mathbf{0}$ (equality of mixed partials — established in E0): $\nabla\times\mathbf{E}=-\nabla\times\nabla(kq/\mathcal{R})=\mathbf{0}$ , valid everywhere except the source point itself, where the field is singular.

Step 4. Superpose. The curl is a linear operator: $\nabla\times\sum_i\mathbf{E}_i=\sum_i\nabla\times\mathbf{E}_i=\mathbf{0}$ , and for continuous distributions the sum becomes an integral over source points with the curl acting on field coordinates — still zero under the integral sign. Why the interchange is legal: the curl differentiates with respect to $\mathbf{r}$ while the integral runs over $\mathbf{r}'$ ; away from singularities the derivatives pass through.

Step 5. Stokes consistency. Feed the result back through the Curl (Stokes) Theorem: $\oint_C\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=\int_S(\nabla\times\mathbf{E})\cdot\mathrm{d}\mathbf{a}=\int_S\mathbf{0}\cdot\mathrm{d}\mathbf{a}=0$ for every loop and every spanning surface — the honest line-integral computation of the previous derivation, recovered from the local law. One law, two costumes.

Which statement do you reach for

you want	use	notes
test whether a given field could be electrostatic	compute $\nabla\times\mathbf{E}$	any nonzero component disqualifies it
justify defining a potential	$\nabla\times\mathbf{E}=\mathbf{0}$ + simply connected	the Scalar Potentials circle of equivalences
work done between two points	path independence, $\int_a^b\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}$	endpoints only; pick the laziest path
round-trip energy audit	$\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0$	Kirchhoff's voltage law in embryo
complete field equations of statics	(1) + Gauss's Law in Differential Form	Helmholtz: div + curl determine the field
time-varying regime	$\nabla\times\mathbf{E}=-\partial\mathbf{B}/\partial t$	Faraday's Law of Induction (E6) — the curl returns

Limiting cases

Case	Result	Meaning
single point charge	$\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0$ exactly	central + radial ⟹ only endpoint radii matter
any central field $f(r)\hat{\mathbf{r}}$	curl-free regardless of $f$	centrality alone kills the curl; $1/r^2$ is Gauss's business
arbitrary static distribution	$\nabla\times\mathbf{E}=\mathbf{0}$	superposition: linearity of curl over sources
inside charged matter	still $\nabla\times\mathbf{E}=\mathbf{0}$	the curl law holds where $\rho\neq0$ ; only the divergence sees charge
slowly varying fields (quasi-statics)	$\nabla\times\mathbf{E}\approx\mathbf{0}$	circuit theory's regime: voltages well-defined to excellent accuracy
time-varying $\mathbf{B}$	$\nabla\times\mathbf{E}=-\partial\mathbf{B}/\partial t\neq\mathbf{0}$	Faraday: circulation = EMF; single-valued $V$ alone no longer represents $\mathbf{E}$

Common trapZero curl is a law of nature for E — no, it is a law of statics

The equation $\nabla\times\mathbf{E}=\mathbf{0}$ has the typographic confidence of a fundamental law, and students promote it to one — then Faraday's law feels like a contradiction. It is not: the full law is $\nabla\times\mathbf{E}=-\partial\mathbf{B}/\partial t$ , and electrostatics is the special case where nothing moves and nothing varies, so the right side sleeps. Inside a transformer or a betatron, $\mathbf{E}$ circulates vigorously, loops of wire carry EMFs, and "the voltage between two points" becomes path-dependent — the potential alone can no longer represent the field. What survives universally is not zero curl but the bookkeeping built on it, upgraded in E9 to $\mathbf{E}=-\nabla V-\partial\mathbf{A}/\partial t$ . Within statics, however, the law is exact everywhere — including inside charge distributions, where many students wrongly suspect it fails. Charge density feeds the divergence, never the curl.

Worked examples

1 · Auditing a round trip near a point charge (confidence builder)

A point charge $q=5.0\ \mathrm{nC}$ sits at the origin. Carry a test charge $q_0=2.0\ \mathrm{nC}$ around the closed path: radially out from $A$ ( $r=0.10$ m) to $B$ ( $r=0.25$ m), along a circular arc at $r=0.25$ m to $C$ , radially in to $D$ ( $r=0.10$ m), then along the inner arc back to $A$ . Compute the work done by the field on each leg and confirm the round trip nets zero. ( $k=8.99\times10^9\ \mathrm{N\,m^2/C^2}$ .)

Step 1. The constant first: $kq = 8.99\times10^9\times5.0\times10^{-9}=44.95\ \mathrm{V\,m}$ .

Step 2. Leg $A\to B$ (radial, outward). From the honest line integral, $W_{AB}=q_0\,kq\left(\frac{1}{r_A}-\frac{1}{r_B}\right) = 2.0\times10^{-9}\times44.95\times(10-4) = 2.0\times10^{-9}\times269.7 = \boxed{+5.4\times10^{-7}\ \mathrm{J}}$ — the field pushes the positive test charge outward, doing positive work (539 nJ).

Step 3. Arc $B\to C$ at fixed $r=0.25$ m: the field is radial, the motion is tangential, $\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0$ at every point. $W_{BC}=\boxed{0}$ — no sweat, no work.

Step 4. Leg $C\to D$ (radial, inward): same endpoints radii as $A\to B$ but reversed, $W_{CD}=q_0\,kq\left(\frac{1}{0.25}-\frac{1}{0.10}\right)=-5.4\times10^{-7}\ \mathrm{J}$ . The inner arc $D\to A$ : zero again.

Step 5. Total: $W = 5.4\times10^{-7}+0-5.4\times10^{-7}+0=\boxed{0}$ ✓. The arcs were free and the radial legs canceled — the loop integral vanished leg by leg, exactly as the derivation promised. Any wilder loop just interleaves more free arcs and more canceling radial excursions.

2 · Could this field exist? The curl as a lie detector (exam level)

Two candidate fields are proposed, with $K=10$ in SI units and coordinates in metres. (a) $\mathbf{F}_1=K\,(xy\,\hat{\mathbf{x}}+2yz\,\hat{\mathbf{y}}+3xz\,\hat{\mathbf{z}})$ . (b) $\mathbf{F}_2=K\,(y^2\,\hat{\mathbf{x}}+(2xy+z^2)\,\hat{\mathbf{y}}+2yz\,\hat{\mathbf{z}})$ . Which could be an electrostatic field? For the impostor, compute the circulation around the unit square in the $xy$ -plane ( $z=0$ , corners $(0,0)$ , $(1,0)$ , $(1,1)$ , $(0,1)$ ) and interpret it.

Step 1. Curl of $\mathbf{F}_1$ : the components are $(\partial_y F_z-\partial_z F_y,\ \partial_z F_x-\partial_x F_z,\ \partial_x F_y-\partial_y F_x) = K(0-2y,\ 0-3z,\ 0-x) = -K(2y,\,3z,\,x)$ . Not identically zero — $\mathbf{F}_1$ cannot be electrostatic, full stop; no arrangement of static charges produces it.

Step 2. Curl of $\mathbf{F}_2$ : $\big(\partial_y(2yz)-\partial_z(2xy+z^2),\ \partial_z(y^2)-\partial_x(2yz),\ \partial_x(2xy+z^2)-\partial_y(y^2)\big) = K(2z-2z,\ 0-0,\ 2y-2y)=\mathbf{0}$ ✓. $\mathbf{F}_2$ passes and indeed $\mathbf{F}_2=-\nabla V$ with $V=-K(xy^2+yz^2)$ — check: $-\partial_xV=Ky^2$ , $-\partial_yV=K(2xy+z^2)$ , $-\partial_zV=2Kyz$ ✓.

Step 3. Circulation of the impostor around the unit square at $z=0$ . Bottom ( $y=0$ , $x:0\to1$ ): $F_{1x}=Kxy=0$ , contributes $0$ . Right side ( $x=1$ , $y:0\to1$ ): $F_{1y}=2Kyz=0$ at $z=0$ , contributes $0$ . Top ( $y=1$ , $x:1\to0$ ): $\int_1^0 Kx\,\mathrm{d}x = -K/2$ . Left side ( $x=0$ , $y:1\to0$ ): $F_{1y}=0$ , contributes $0$ .

Step 4. Total circulation: $\oint\mathbf{F}_1\cdot\mathrm{d}\boldsymbol{\ell} = -K/2 = \boxed{-5.0\ \mathrm{V}}$ per lap. Cross-check via Stokes: the enclosed curl component is $(\nabla\times\mathbf{F}_1)_z=-Kx$ , and $\int_0^1\!\!\int_0^1(-Kx)\,\mathrm{d}x\,\mathrm{d}y = -K/2$ ✓.

Step 5. Interpret: the counterclockwise circulation is $-5.0$ V, so a $1$ C charge driven clockwise around the loop would gain $5$ J per lap — for free, forever. That perpetual-motion verdict is why no static charge arrangement can build $\mathbf{F}_1$ . (Fields with exactly this character do exist — around a changing magnetic flux, where Faraday's law supplies the energy honestly through whatever drives $\mathbf{B}$ .)

Origins · 1 mingeorge gabriel stokes and how this was discovered

george gabriel stokes

1819–1903 · Pembroke College, Cambridge

Stokes held the Lucasian chair — Newton's chair — for over half a century, and ran much of Victorian physics from it as the long-serving secretary of the Royal Society. Fluid dynamics (the Navier–Stokes equations), the explanation of fluorescence, the first correct account of what makes the sky's polarization — his fingerprints are everywhere. The theorem bearing his name he did not, in fact, prove: it reached him in an 1850 letter from William Thomson (Kelvin), and Stokes set it as a Smith's Prize examination question at Cambridge in 1854 — a young Maxwell sat that exam. The name stuck to the examiner, not the author, and Maxwell went on to make the circulation-to-curl bridge a load-bearing wall of electromagnetic theory. Stokes' own electrostatics sensibility survives in every use of the theorem here: convert what you cannot see (a local twist) into what you can measure (work around a loop), and let the geometry arbitrate.

History

1785

— Coulomb's torsion-balance experiments establish the central, inverse-square force between charges — both properties this article spends.

1828

— Green's Essay introduces the potential function for electricity, presupposing exactly the conservative property proved here.

1831

— Faraday's induction experiments reveal circulating electric fields around changing flux — the discovery that fences this article's law into statics.

1850

— Kelvin communicates the circulation–curl theorem to Stokes in a letter; Stokes sets it as a Smith's Prize question in 1854, and the name sticks.

1865

— Maxwell's field equations enshrine both regimes: zero curl for statics, minus the flux rate in general.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §2.3.1–2.3.2 · Purcell & Morin, Electricity and Magnetism, 3rd ed., §2.1 · Jackson, Classical Electrodynamics, 3rd ed., §1.5 · Zangwill, Modern Electrodynamics, §3.3 · Feynman Lectures, Vol. II, Ch. 4 Key numerical checks: $\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=(kq/r^2)\mathrm{d}r$ follows directly from the spherical line element; $kq=44.95$ V·m for $q=5$ nC and the leg work $q_0\,kq(10-4)=5.39\times10^{-7}$ J for $q_0=2$ nC; both curls in Worked example 2 are recomputable componentwise and the square circulation comes out $-K/2=-5.0$ V two ways — direct leg integration and Stokes with $(\nabla\times\mathbf{F}_1)_z=-Kx$ ; and $V=-K(xy^2+yz^2)$ reproduces $\mathbf{F}_2$ by gradient.

Builds on: Coulomb's Law · The Electric Field · Superposition Principle for Electric Fields · Continuous Charge Distributions Leads to: The Electrostatic Potential · Gauss's Law in Differential Form · Poisson's Equation · Potential of a Localized Charge Distribution

Is the curl law independent information, or does Gauss's law already contain it?

Genuinely independent — the two equations spend different parts of Coulomb's law. Gauss's law encodes the inverse-square magnitude: change the exponent and flux through a sphere would depend on its radius. The curl law encodes the direction — centrality: the field of each charge points along the line to it, making it a gradient of a function of distance alone. You can invent fields that satisfy one and violate the other (a swirling divergence-free field satisfies Gauss with $\rho=0$ but fails the curl test). By the Helmholtz theorem, specifying both the divergence and the curl (with decay at infinity) pins the field uniquely — which is precisely why electrostatics needs exactly these two equations, and why neither is redundant.

Does the curl of E vanish inside charged matter, or only in vacuum?

Everywhere, in statics — this is the trap in the last quiz question. Look at where charge enters the derivations: $Q$ and $\rho$ feed the divergence equation; the curl computation never consults the local charge density. Each point charge's field is a gradient whether you stand near it or far, and superposition sums gradients into a gradient. So inside a uniformly charged ball, inside a battery's electrolyte, inside a thundercloud — wherever things are static — $\nabla\times\mathbf{E}=\mathbf{0}$ holds and a potential exists. What does change inside matter is how hard the divergence works. The frequent confusion comes from conflating "the field is complicated here" with "the field circulates here"; complexity lives in the sources, circulation would need time dependence.

If any central field is curl-free, why does the article make a fuss about Coulomb's law specifically?

Because superposition needs each individual source's field to be central from its own location — and nature obliged. A single hypothetical charge with field $f(r)\hat{\mathbf{r}}$ , any $f$ , is curl-free; that much is cheap. The physical content is that real electrostatic fields are built by superposing such central fields from many source points, each a gradient of a function of its own $\mathcal{R}$ , so the sum is a gradient too. Had the force between charges carried a sideways component (violating centrality), single-charge fields would already circulate and no potential would exist anywhere in electrostatics. The inverse square, meanwhile, is spent elsewhere: it is what makes the divergence come out as a delta function on the source — Gauss's law's inheritance.

How is a battery consistent with zero circulation — it drives charge around a loop, doesn't it?

Beautiful question, and the resolution is that a battery is not an electrostatic object. Around a circuit loop, the electrostatic field still satisfies $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=0$ faithfully. What pushes charge through the battery's interior against that field is chemistry — a non-electrostatic force per unit charge (often written $\mathbf{f}_s$ ) confined to the electrolyte. The EMF is $\oint\mathbf{f}_s\cdot\mathrm{d}\boldsymbol{\ell}$ , nonzero because $\mathbf{f}_s$ is not a gradient field. Outside the battery, the electrostatic field does positive work on the current; inside, chemistry hauls the charge back uphill. Energy per lap is supplied by consuming reactants, not by violating this article. The same accounting handles generators, where the loop-driving agent is Faraday's law instead of chemistry.

What breaks first when fields start changing in time slowly?

Quantitatively: the induced curl is $|\nabla\times\mathbf{E}|=|\partial\mathbf{B}/\partial t|$ , so the circulation around a loop of area $A$ is about $A\,\dot{B}$ . Compare that with the typical voltages in your problem and you have your own validity meter. A 50 Hz, 0.1 T stray field threading a 10 cm² loop gives $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}\sim 10^{-4}\times 0.1\times 2\pi\times 50\approx 3$ mV per loop — irrelevant to a 12 V power circuit, catastrophic to a microvolt sensor measurement. That is the quasi-static regime in one estimate: circuit theory's "voltage" remains an excellent fiction while $A\dot{B}$ is small against your signal, and RF/EMC engineering begins where it is not. The formal repair — potentials that survive time dependence — arrives with $\mathbf{E}=-\nabla V-\partial\mathbf{A}/\partial t$ in E9.

Why does zero curl guarantee the potential is single-valued?

Because the potential difference between two points is defined by a line integral, and zero curl (on a simply connected region) makes that integral path-independent — every route from $\mathbf{a}$ to $\mathbf{b}$ reports the same number, so "the potential at $\mathbf{b}$ relative to $\mathbf{a}$ " is one well-defined value rather than a path-dependent family. If circulation lurked around some loop, walking that loop would increment your "potential" by the circulation each lap, like a spiral staircase pretending to be a single floor — multivalued and useless for physics. The fine print "simply connected" matters in magnetostatics (loops around current-carrying wires), but electrostatic space, even dotted with point charges, leaves the argument intact: the full equivalence chain is walked carefully in Scalar Potentials.

Is zero curl an approximation or exact — and how well is it tested?

Within genuine statics it is exact, inherited from the exactness of centrality in Coulomb's law, and its tests piggyback on the most precise experiments in physics. Path-independence of potential differences underlies voltage metrology at the parts-per-billion level; any intrinsic circulation of static fields would wreck the agreement between Josephson-standard volts realized in different laboratory geometries. Photon-mass bounds constrain related deviations: a massive photon modifies the field structure, and satellite and laboratory limits put $m_\gamma < 10^{-54}$ kg territory. In practice, when an experiment does see $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}\neq0$ , the resolution is always a hidden $\partial\mathbf{B}/\partial t$ — a moving magnet, a switching current — never a failure of the static law. The theorist's version: zero curl is what "static" means.