Curl of B

The magnetostatics chapter has been assembling a differential portrait of $\vec B$ , and one panel remains. The Helmholtz theorem of the vector-calculus chapter set the terms: to know a field is to know its divergence and its curl. The divergence came last article — zero, everywhere, no monopoles. This article computes the curl, and the answer is the second field equation of magnetostatics and the future backbone of Ampère's law:

\nabla\times\vec B = \mu_0\,\vec J.

Read it as a locality statement, because that is its content: the swirl of the magnetic field at a point is set entirely by the current at that point. Empty space beside a roaring wire has enormous $\vec B$ but exactly zero curl — the circulation is all inherited from elsewhere; a paddle wheel dropped in the field there would feel torque-free flow. Where current density lives, and only there, the field's local rotation switches on, in strict proportion. Electrostatics had charge sourcing the divergence of $\vec E$ ; magnetostatics has current sourcing the curl of $\vec B$ . Charge makes fields spread; current makes fields wrap. The derivation is this chapter's graduation exercise: it takes the curl of the Biot–Savart integral and must spend, by name, nearly everything the track has banked — the delta-function identity $\nabla\cdot(\hat{\boldsymbol{\mathscr{r}}}/\mathscr{r}^2) = 4\pi\delta^3$ from the vector-calculus chapter, and the steady-current condition $\nabla\cdot\vec J = 0$ from the Continuity Equation, which enters mid-proof as a load-bearing wall rather than a footnote. The payoff cascade is immediate: Stokes' theorem converts the local law into the integral Ampère's law whose applications own the next article; a Poisson equation for the vector potential comes into view; and the law's built-in refusal to tolerate unsteady currents marks the exact doorstep where Maxwell will later force his displacement current through.

Key takeaways

The second field equation of magnetostatics: $\nabla\times\vec B = \mu_0\vec J$ — the local circulation of $\vec B$ is proportional to the current density at that very point.
It is a theorem, derived by taking the curl of the Biot–Savart law; the proof consumes the E0 delta identity $\nabla\cdot(\hat{\boldsymbol{\mathscr{r}}}/\mathscr{r}^2) = 4\pi\delta^3$ and the steady-current condition $\nabla\cdot\vec J = 0$ .
Away from currents, $\nabla\times\vec B = 0$ even where $\vec B$ is strong — the wire's exterior field is curl-free; all the swirl is concentrated on the current.
Stokes' theorem upgrades it to the integral Ampère's law $\oint\vec B\cdot\mathrm{d}\vec\ell = \mu_0I_{\rm enc}$ — stated here, spent on symmetric geometries by the next article.
Self-consistency check with teeth: the divergence of a curl vanishes, so the law demands $\nabla\cdot\vec J = 0$ — feed it a charging capacitor and it breaks, which is exactly where Maxwell's displacement current will enter (E6).
With both derivatives known ( $\nabla\cdot\vec B = 0$ , $\nabla\times\vec B = \mu_0\vec J$ ), Helmholtz says the magnetostatic field is completely determined — and writing $\vec B = \nabla\times\vec A$ turns the curl law into a Poisson equation $\nabla^2\vec A = -\mu_0\vec J$ (built properly in Magnetic Vector Potential).
KEY IDEA: current is the sole source of the curl of $\vec B$ — the field circulates around currents like water around a stirring rod, with circulation density $\mu_0\vec J$ , and exactly zero swirl anywhere current is absent.

The core insight

In magnetostatics, current is the sole source of the curl of $\vec B$ : $\nabla\times\vec B = \mu_0\vec J$ . The magnetic field circulates around currents the way water swirls around a stirring rod, with circulation density proportional to the local current density.

Where this lives today

The curl law is the equation magnetostatic solvers actually discretize: finite-element magnetics codes assemble $\nabla\times(\nu\nabla\times\vec A) = \vec J$ (the Poisson form (4) with material reluctivity) over millions of mesh cells to design every motor, transformer and MRI magnet — the differential law as industrial infrastructure. Run in reverse it is a measurement principle: NASA's MMS quartet flies a literal curlometer, differencing $\vec B$ across a tetrahedron of spacecraft to map the currents of magnetic-reconnection events; solar physicists infer coronal currents from vector magnetograms the same way; and tokamak control rooms reconstruct plasma current profiles from edge magnetics in real time between plasma shots. Structurally, this equation is where magnetostatics closes. With divergence and curl both delivered, Helmholtz's theorem certifies the field of any steady current as completely determined — the chapter's remaining articles (Ampère applications, the vector potential, boundary conditions, multipoles) are harvest, not new sowing. And the law's one honest failure is the most productive crack in physics: its structural refusal of unsteady currents is exactly where the displacement current enters, turning $\nabla\times\vec B = \mu_0\vec J$ into the Ampère–Maxwell law, whose consequence — electromagnetic waves at speed $1/\sqrt{\mu_0\varepsilon_0}$ — is the reason the electrodynamics chapters exist.

Origins · 1 minJames Clerk Maxwell and how this was discovered

James Clerk Maxwell

1831–1879 · Edinburgh, Aberdeen, Cambridge

Maxwell is the reason this article's equation has its differential form — and its famous correction. In "On Faraday's Lines of Force" (1855–56) he translated Faraday's intuitive field-line pictures into exact local mathematics, expressing the relations between currents and magnetic circulation that Ampère had left in integral, action-at-a-distance dress. The 1861–62 vortex-model papers then exposed the steady-current law's breaking point and supplied the repair — the displacement current — from which electromagnetic waves and the electromagnetic theory of light followed in 1865. The Treatise (1873) delivered the full architecture that Heaviside would compress into today's four equations, this article's among them. Einstein kept Maxwell's portrait on his wall and called the field-equation revolution "the most profound and the most fruitful that physics has experienced since the time of Newton"; the curl of $\vec B$ , steady form here and corrected form ahead, is that revolution's signature line.

History

1820

— Ørsted, Biot, Savart and Ampère establish that currents make circulating magnetic fields — the phenomenon awaiting a local law.

1826

— Ampère's Mémoire encodes current–field relations in integral form; the differential version awaits field thinking.

1855

— Maxwell's On Faraday's Lines of Force begins translating field lines into local differential equations.

1861

— Maxwell's vortex model diagnoses the steady law's inconsistency and inserts the displacement current.

1873

— The Treatise on Electricity and Magnetism codifies the field equations; curl B = μ0 J stands as the magnetostatic limit.

1884

— Heaviside recasts Maxwell's system into the four vector equations taught today.

2015

— NASA's MMS tetrahedron flies a working curlometer, measuring space-plasma currents from field differences — the differential law as instrument.

The physical picture

Bring back E0's curl-meter: an imaginary paddle wheel, dropped into the field, spinning if the field circulates locally around its axle. Around a current-carrying wire the field lines are circles — surely maximal swirl? Drop the paddle wheel beside the wire and it refuses to turn. The field is stronger on the paddle's near side but the lines also curve and diverge in exactly the compensating way; the local rotation is zero. The $1/s$ profile of the wire's field is precisely the curl-free way to circulate — all of the circulation is concentrated on the axis, where the current is. Slide the paddle wheel into the wire itself, into the current-carrying metal, and it spins.

Figure

The paddle-wheel audit: swirl lives only where the current is

This is the resolution of an apparent paradox worth savoring. Circulation — the line integral $\oint\vec B\cdot\mathrm{d}\vec\ell$ around a big loop — is plainly nonzero around a wire. Curl is circulation per unit area, taken in the shrinking limit; and around the wire all circulation is inherited from the threading current, none generated locally in the empty space. The two statements — global circulation nonzero, local curl zero — coexist because the curl is singularly concentrated where $\vec J$ lives. The mathematics below makes that concentration literal: a Dirac delta at the source, smoothed into $\mu_0\vec J(\vec r)$ when the current is distributed. Set beside the last article, the full duality of statics clicks into place. Charge sources the divergence of $\vec E$ ; current sources the curl of $\vec B$ ; the other derivative of each field vanishes ( $\nabla\times\vec E = 0$ , $\nabla\cdot\vec B = 0$ ). Four local equations, two sources, complete symmetry of structure with total asymmetry of content — and by the Helmholtz theorem, four derivatives are exactly what it takes to pin both fields down.

Figure

The four derivatives of statics: charge spreads E, current wraps B

The theorem and its price

The claim, with its small print:

\nabla\times\vec B = \mu_0\,\vec J \qquad (1)

where: $\vec J$ — the current density at the field point ( $\mathrm{A/m^2}$ ); $\mu_0 = 4\pi\times10^{-7}\ \mathrm{T\,m/A}$ . Locality is the message: the curl does not care about distant currents, however large. And the small print is enforceable: take the divergence of both sides — the left side dies identically (divergence of a curl), so the law can only hold where

\nabla\cdot\vec J = 0 \qquad (2)

where: this is precisely the steady-current contract signed in Continuity Equation. The law does not merely prefer steady currents; it is algebraically incapable of coexisting with any other kind. Keep that in view: it is both a hypothesis the derivation must use and the exact fault line along which Maxwell will later crack the equation open.

Derivation AThe curl of the Biot-Savart field — the chapter's debts called in

The goal is equation (1) from the volume Biot–Savart law — the companion calculation to last article's divergence, but harder and richer: this one produces a source. The strategy: push $\nabla\times$ under the integral, expand the curl-of-a-cross-product, and evaluate two surviving terms — one collapses onto a delta function (the E0 identity, giving the local $\mu_0\vec J$ ), the other must be shown to vanish, and its death certificate is signed by $\nabla\cdot\vec J = 0$ . Watch both banked results get spent; neither is decorative.

Step 1. Take the curl under the integral (field and source coordinates are independent):

\nabla\times\vec B = \frac{\mu_0}{4\pi}\int \nabla\times\left[\vec J(\vec r')\times\frac{\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}\right]\mathrm{d}\tau'.

Step 2. Expand with the curl-of-a-cross-product rule (E0's product rules), writing $\vec f = \hat{\boldsymbol{\mathscr{r}}}/\mathscr{r}^2$ . Of the four terms, the two containing field-coordinate derivatives of $\vec J(\vec r')$ vanish outright, leaving

\nabla\times(\vec J\times\vec f) = \vec J\,(\nabla\cdot\vec f) - (\vec J\cdot\nabla)\,\vec f.

Step 3. The first term delivers the law. The E0 delta identity — proved twice over in the vector-calculus chapter and recalled here, not re-derived — gives $\nabla\cdot\vec f = 4\pi\,\delta^3(\boldsymbol{\mathscr{r}})$ , and the delta collapses the integral onto the field point:

\frac{\mu_0}{4\pi}\int \vec J(\vec r')\,4\pi\,\delta^3(\vec r - \vec r')\,\mathrm{d}\tau' = \mu_0\,\vec J(\vec r).

Why this is the physical heart: the same singular concentration that made a point charge's divergence a delta now makes the field's swirl strictly local — the curl "sees" only the current at its own location, which is the locality announced in the overview.

Step 4. The second term must die, and here the current's steadiness earns its keep. Trade the field-coordinate derivative for a source-coordinate one ( $\vec f$ depends only on $\vec r - \vec r'$ , so $\nabla f_{\rm component} = -\nabla' f_{\rm component}$ ), then integrate by parts, moving $\nabla'$ onto $\vec J$ :

-\int (\vec J\cdot\nabla)\,\vec f\,\mathrm{d}\tau' = \int (\vec J\cdot\nabla')\,\vec f\,\mathrm{d}\tau' = -\int \vec f\,(\nabla'\cdot\vec J)\,\mathrm{d}\tau' + \text{surface term}.

The bulk piece vanishes because $\nabla'\cdot\vec J = 0$ — the steady-current condition, entering as a load-bearing wall. The surface term vanishes for any localized current distribution (push the boundary beyond the sources). Why the honesty matters: without steadiness this term survives, and the curl law fails — quantitatively, not cosmetically. This is the precise mathematical location of magnetostatics' boundary post: Ampère's law is not approximately true for unsteady currents; it is structurally incompatible with them.

Step 5. Collect:

\nabla\times\vec B = \mu_0\,\vec J(\vec r).

The divergence-consistency check closes cleanly: $\nabla\cdot(\mu_0\vec J) = 0$ ✓ for exactly the steady currents the derivation admits.

The integral upgrade is two lines and a theorem from E0. Integrate (1) over any open surface $S$ and apply Stokes' theorem to the left side:

\oint_{\partial S}\vec B\cdot\mathrm{d}\vec\ell = \mu_0\int_S\vec J\cdot\mathrm{d}\vec a = \mu_0\,I_{\rm enc} \qquad (3)

where: $I_{\rm enc}$ — the net current threading the loop $\partial S$ (through any spanning surface — well-defined precisely because $\nabla\cdot\vec J = 0$ , the same pod argument as last article's flux theorem, run on $\vec J$ ). Equation (3) is Ampère's law in integral form: circulation of $\vec B$ = $\mu_0$ × threaded current. Its exploitation — the symmetry ritual that cracks wires, solenoids, toroids and planes in one line each — is the whole business of the next article; here it is the corollary, minted and set aside.

Go deeper · Derivation B · 1 minCalibrating the curl-meter on the wire's field

Derivation BCalibrating the curl-meter on the wire's field

The goal is a hands-on verification that closes the paddle-wheel story with actual derivatives: take the known exterior field of a straight wire, $\vec B = (\mu_0I/2\pi s)\hat{\boldsymbol\phi}$ , and show its curl vanishes identically off the axis — then locate the missing circulation on the axis itself. The strategy is direct computation in cylindrical coordinates (E0's curl formula), plus a shrinking-loop argument for the singular axis. This is the standard sanity ritual for any proposed magnetostatic field, run here once in full.

Step 1. Write the cylindrical curl for a purely azimuthal field $\vec B = B_\phi(s)\,\hat{\boldsymbol\phi}$ . Only one component survives:

\nabla\times\vec B = \frac{1}{s}\,\frac{\partial}{\partial s}\left(s\,B_\phi\right)\hat{\vec z}.

Step 2. Feed it the wire's field, $B_\phi = \mu_0I/2\pi s$ :

\frac{1}{s}\,\frac{\partial}{\partial s}\left(s\cdot\frac{\mu_0 I}{2\pi s}\right) = \frac{1}{s}\,\frac{\partial}{\partial s}\left(\frac{\mu_0I}{2\pi}\right) = 0 \qquad (s \neq 0).

Why zero: $s\,B_\phi$ is constant — the $1/s$ profile is exactly the azimuthal field whose curl vanishes, the unique curl-free way to circulate. Any other radial power would leave residual swirl in empty space, demanding current there.

Step 3. Find the missing circulation. Around any circle of radius $s$ , $\oint\vec B\cdot\mathrm{d}\vec\ell = (\mu_0I/2\pi s)(2\pi s) = \mu_0I$ — independent of $s$ . Shrink the circle: the circulation does not fade, so the curl (circulation per area) diverges as the area vanishes — a two-dimensional delta function on the axis:

\nabla\times\vec B = \mu_0 I\,\delta^2(s)\,\hat{\vec z},

which is precisely $\mu_0\vec J$ for the idealized line current. The books balance: zero curl where there is no current, all of it concentrated where there is.

Step 4. The same audit run inside a thick wire (uniform $J$ , interior field $B_\phi = \mu_0Js/2$ from the next article's ritual): $\frac{1}{s}\partial_s(s\cdot\mu_0Js/2) = \frac{1}{s}\cdot\mu_0Js = \mu_0J$ ✓ — the curl-meter reads exactly the local current density at every interior point, delta functions nowhere needed when the source is honestly distributed. The shrinking-loop limit reproduces the line-current idealization consistently with the substitution rule — a line current is simply a delta-concentrated $\vec J$ .

One more structural door, opened and labeled. Since $\nabla\cdot\vec B = 0$ guarantees $\vec B = \nabla\times\vec A$ (last article), substitute into (1) and use the double-curl identity owned by E0's Del Operator article:

\nabla\times(\nabla\times\vec A) = \nabla(\nabla\cdot\vec A) - \nabla^2\vec A = \mu_0\vec J.

Choosing the gauge $\nabla\cdot\vec A = 0$ (legitimacy and freedom explained in Gauge Freedom and the Coulomb Gauge) reduces the curl law to

\nabla^2\vec A = -\mu_0\,\vec J \qquad (4)

where: each Cartesian component obeys a Poisson equation — the same equation electrostatics spent two chapters solving, now with $\mu_0J_i$ in the source seat. The entire boundary-value technology of E1–E2 transfers wholesale; that harvest is Magnetic Vector Potential's to reap, two articles ahead.

Limiting cases

Case	Result	Meaning
empty space near currents	$\nabla\times\vec B = 0$ , $\vec B \neq 0$	strong field, zero swirl — circulation is inherited
inside a uniform current	$\nabla\times\vec B = \mu_0J\,\hat{\vec z}$	curl-meter reads the local $J$ , point by point
idealized line current	curl $= \mu_0I\,\delta^2(s)\hat{\vec z}$	all circulation concentrated on the axis
integrate over a surface	$\oint\vec B\cdot\mathrm{d}\vec\ell = \mu_0I_{\rm enc}$	Ampère's law — applications next article
take the divergence	$0 = \mu_0\nabla\cdot\vec J$	steady currents only — enforced, not optional
unsteady current	law fails structurally	Maxwell's $\mu_0\varepsilon_0\,\partial\vec E/\partial t$ repairs it (E6)
via $\vec B = \nabla\times\vec A$ , Coulomb gauge	$\nabla^2\vec A = -\mu_0\vec J$	Poisson again — E1/E2 machinery transfers
both derivatives known	Helmholtz: $\vec B$ unique	magnetostatics is now a closed theory

Common trapCurl-free is not field-free — and circulation is not curl

The two conflations this topic invites. First: outside the wire $\nabla\times\vec B = 0$ while $\vec B$ is at full strength — curl measures local swirl, not magnitude; a paddle wheel beside the wire does not turn. (Electrostatics had the same lesson transposed: $\nabla\cdot\vec E = 0$ in the empty space where $\vec E$ is strongest.) Second: nonzero circulation around a big loop does not mean nonzero curl along it — around a wire the circulation is $\mu_0I$ for every loop, yet the curl vanishes at every point of empty space the loop passes through; the circulation is entirely inherited from the singular concentration on the current. And the standing structural trap: equation (1) is a fair-weather law — apply it near a charging capacitor plate ( $\nabla\cdot\vec J \neq 0$ ) and different spanning surfaces give different $I_{\rm enc}$ , a genuine contradiction, not a subtlety. That failure is scheduled, and its repair — the displacement current — is the opening act of the electrodynamics chapter.

Worked examples

1 · The curl-meter inside a busbar (confidence builder)

A straight wire of radius $a = 1.0\ \mathrm{mm}$ carries $I = 10\ \mathrm{A}$ spread uniformly over its cross-section. Find $J$ , the interior field, and verify the curl law pointwise.

Step 1. Current density: $J = I/\pi a^2 = 10/\pi(10^{-3})^2 = \boxed{3.2\times10^{6}\ \mathrm{A/m^2}}$ , along $\hat{\vec z}$ .

Step 2. Interior field (from the integral law (3) on a circle of radius $s < a$ , or the next article's ritual): $B_\phi(s) = \mu_0Js/2$ . At $s = 0.5\ \mathrm{mm}$ : $B = (4\pi\times10^{-7})(3.2\times10^{6})(5\times10^{-4})/2 = \boxed{1.0\ \mathrm{mT}}$ , rising linearly to $2.0\ \mathrm{mT}$ at the surface.

Step 3. The pointwise audit — cylindrical curl of $B_\phi = \mu_0Js/2$ :

\left(\nabla\times\vec B\right)_z = \frac{1}{s}\frac{\partial}{\partial s}\left(s\cdot\frac{\mu_0Js}{2}\right) = \frac{1}{s}\cdot\mu_0Js = \mu_0J\ ✓

— the curl returns exactly $\mu_0$ times the local current density, at every interior point, no averaging or loops required. Differential Ampère is a pointwise meter, and this is what it reads.

Step 4. Cross the surface: outside, $B_\phi = \mu_0I/2\pi s$ and the curl drops dead to zero (Derivation B) — the meter tracks $\vec J$ through the boundary discontinuously, exactly as a local law must.

2 · Reading a hidden current distribution from a measured field (exam level)

A plasma-physics diagnostic maps an azimuthal field $\vec B = C\,s^2\,\hat{\boldsymbol\phi}$ (with $C = 0.10\ \mathrm{T/m^2}$ ) inside a column of radius $R = 2.0\ \mathrm{cm}$ . Infer the current density profile, the total current, and cross-check with the integral law.

Step 1. Differentiate — the curl law run in reverse (field measured, source inferred):

\mu_0 J_z(s) = \frac{1}{s}\frac{\partial}{\partial s}\left(s\cdot Cs^2\right) = \frac{3Cs^2}{s} = 3Cs \qquad\Longrightarrow\qquad J_z(s) = \frac{3Cs}{\mu_0}.

The current density grows linearly outward — a hollow-ish profile, densest at the edge: $J(R) = 3(0.10)(0.02)/\mu_0 = \boxed{4.8\times10^{3}\ \mathrm{A/m^2}}$ .

Step 2. Total current by integrating the inferred profile:

I = \int_0^R J_z\,2\pi s\,\mathrm{d}s = \frac{6\pi C}{\mu_0}\int_0^R s^2\,\mathrm{d}s = \frac{2\pi C R^3}{\mu_0} = \boxed{4.0\ \mathrm{A}}.

Step 3. Cross-check with the integral law at the boundary: $\oint\vec B\cdot\mathrm{d}\vec\ell = B(R)\cdot2\pi R = CR^2\cdot2\pi R = 2\pi CR^3$ , so $I_{\rm enc} = 2\pi CR^3/\mu_0 = 4.0\ \mathrm{A}$ ✓ — differential and integral bookkeeping agree to the digit.

Step 4. The professional moral: this inverse use — measure $\vec B$ , differentiate, read $\vec J$ — is a working diagnostic. Tokamak current-profile reconstruction and the four-spacecraft "curlometer" of the MMS magnetospheric mission compute space-plasma currents exactly this way: nature hands you the field; the curl law hands you the source.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §5.3.2–5.3.3 · Jackson, Classical Electrodynamics, 3rd ed., §5.3 · Purcell & Morin, Electricity and Magnetism, 3rd ed., Ch. 6 · Zangwill, Modern Electrodynamics, §10.3. The curl computation uses the distributional identity $\nabla\cdot(\hat{\boldsymbol{\mathfrak{r}}}/\mathfrak{r}^2)=4\pi\delta^3(\boldsymbol{\mathfrak{r}})$ and integration by parts with $\nabla'\cdot\mathbf{J}=0$ for localized steady currents; the Stokes upgrade to $\oint\mathbf{B}\cdot\mathrm{d}\boldsymbol{\ell}=\mu_0I_{\rm enc}$ and the Coulomb-gauge Poisson equation $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$ are standard. The value $\mu_0=4\pi\times10^{-7}$ T·m/A is used throughout. The two-term expansion confirms the delta term yields $\mu_0\mathbf{J}$ and the transport term dies by parts plus steadiness; the cylindrical curls check out (exterior $1/s$ field → 0; interior $\mu_0Js/2$ → $\mu_0J$ ). Worked numbers: busbar $J = 3.2\times10^{6}$ A/m², $B(0.5\,\mathrm{mm}) = 1.0$ mT, surface 2.0 mT; plasma column $J(R) = 4.8\times10^{3}$ A/m², $I = 2\pi CR^3/\mu_0 = 4.0$ A by both routes.

Builds on: Biot-Savart Law · Electric Current and Current Density · Continuity Equation · Divergence of B and Absence of Magnetic Monopoles

Leads to: Ampère's Law · Magnetic Vector Potential · Gauge Freedom and Coulomb Gauge · Magnetic Field of Steady Currents

If the curl is zero outside the wire, why does a compass beside the wire deflect?

Because the compass responds to $\vec B$ itself, not to its curl. The field beside the wire is strong and perfectly real — it is only the local rotation of that field that vanishes. Curl-free does not mean field-free (the warning's first trap): the exterior field is analogous to the velocity field of water draining around a plughole, which sweeps floating objects around in circles even though each fluid element undergoes no local rotation — a leaf carried by such flow orbits the center while keeping its orientation fixed. The compass needle is the leaf; the paddle wheel is the curl-meter; they measure different things.

How can the integral circulation be μ0 I when the curl vanishes at every point of my loop?

This is Stokes' theorem doing exactly its job with a singular integrand. The theorem equates the loop circulation to the surface integral of the curl over any spanning surface — and every spanning surface of a loop that encircles the wire must pierce the wire, picking up the delta-concentrated curl there. The empty-space points of the loop contribute nothing locally; the wire crossing contributes everything. Try to dodge it by choosing a surface avoiding the wire and topology defeats you: no such spanning surface exists for a linking loop. Circulation is a global quantity that Stokes localizes onto the current, which is also why $I_{\rm enc}$ is surface-independent exactly when $\nabla\cdot\vec J = 0$ .

Why doesn't the curl law care about distant currents when the field itself obviously does?

Precisely the division of labour between a field and its derivatives. $\vec B$ at a point is assembled by Biot–Savart from all currents everywhere — hugely nonlocal. But the derivation shows all the distant contributions arrive curl-free at your location (they are built from the gradient kernel away from its singularity); only the current at your point injects local swirl. Same structure as electrostatics: distant charges shape $\vec E$ , but $\nabla\cdot\vec E$ at your location reads only the charge density under your feet. Locality of the differential laws is what makes them field equations — statements a point can verify with its immediate neighborhood — and it is why they, not the integral solutions, generalize to electrodynamics unchanged in form.

Is curl B = μ0 J an approximation, a definition, or an exact law?

Within magnetostatics: an exact theorem, derived from Biot–Savart with no approximations. Within full electrodynamics: the steady-state limit of the exact Ampère–Maxwell law $\nabla\times\vec B = \mu_0\vec J + \mu_0\varepsilon_0\,\partial\vec E/\partial t$ . The correction term is not small print: it is what makes the system consistent with charge conservation for all currents (take the divergence of the corrected law and the continuity equation pops out identically) and what turns the equations into a wave machine. Practical validity criterion, same as Biot–Savart's: quasistatic — currents changing on timescale $T$ are handled correctly within distances $r \ll cT$ . At 50 Hz that is a thousand kilometres; inside every motor and transformer ever built, the magnetostatic curl law is effectively exact.

The derivation dropped a surface term at infinity. When is that actually justified?

For any localized current distribution — everything in a lab, where $\vec J$ vanishes beyond some radius, the parts-integration boundary term dies once the surface passes the sources. The textbook idealizations that reach infinity (the infinite wire, the infinite solenoid) technically flunk the hypothesis, but they are limits of localized circuits (a giant return loop closed at large distance), and their fields are the corresponding limits — every conclusion survives. Where care genuinely earns its keep: problems with currents entering or leaving the region through boundaries (electrodes, semi-infinite wires used in isolation), where "the" enclosed current can become bookkeeping-dependent. The safe ritual is always to complete the physical circuit, at least conceptually, before applying the machinery — the same closed-loop doctrine stamped on Biot–Savart itself.

Why is having BOTH div B and curl B such a big deal — what exactly does Helmholtz's theorem close?

Because a vector field in three dimensions has exactly the freedom that one divergence (a scalar's worth) and one curl (a divergence-free vector's worth) pin down, given decay at infinity — that is the Helmholtz theorem the E0 chapter proved. Before this article, $\vec B$ was known only through the integral recipe of Biot–Savart; now it is characterized axiomatically: the unique field with $\nabla\cdot\vec B = 0$ , $\nabla\times\vec B = \mu_0\vec J$ , and $\vec B \to 0$ far away. That flip — from recipe to axioms — is what lets the rest of the chapter run on differential methods (potentials, Poisson equations, boundary-value problems) with electrostatics' entire toolkit, and it is the form in which magnetostatics survives, curl term corrected, inside the full Maxwell equations.

In the inverse problem — measuring B and differentiating to get J — doesn't noise wreck the derivatives?

It genuinely tries to, and managing that is the working craft of curlometry. Differentiation amplifies high-frequency noise, so practical current-inference smooths or fits the field first (spline fits to magnetogram data, tetrahedron-scale linear gradients for the MMS curlometer — which also computes $\nabla\cdot\vec B$ from the same data as a built-in quality flag, since any nonzero reading is pure error by last article's law). Resolution is bounded by sensor separation: MMS infers only currents on scales larger than its ~10 km tetrahedron. The physics lesson survives the practicalities: the curl law is an instrument, and like all instruments it has a noise floor and a bandwidth — but what it measures, current from field geometry alone, nothing else can.