Divergence of B and Absence of Magnetic Monopoles

Around the year 1269, the crusader-engineer Petrus Peregrinus mapped the field of a spherical lodestone with a compass needle, found the lines converging on two special points he named poles — and then discovered the fact that has survived every experiment since: break the magnet, and each fragment grows a fresh pair. North and south refuse to be separated. Seven and a half centuries of ever-more-heroic attempts — ocean water, moon rock, cosmic-ray detectors, the LHC's dedicated monopole trap — have found exactly zero isolated magnetic charges. Electromagnetism enshrines that empty-handedness as a field equation, the tersest of Maxwell's four: $\nabla\cdot\vec B = 0$ . This article earns the equation rather than postulating it. The magnetostatic field is manufactured by the Biot–Savart law, and a few lines of vector calculus show that everything so manufactured is automatically divergence-free — the cross-product geometry of the previous articles hard-wires the absence of sources into every field a current can make. From there the consequences cascade: zero net flux through every closed surface; field lines that never start or stop; a flux through a loop that depends only on the loop and not on the surface you drape across it (the fact Faraday's law and every henry of inductance will lean on); and the guarantee that $\vec B$ can always be written as the curl of a vector potential — the doorway to the gauge structure of modern physics. And because the strongest statements deserve the strongest scrutiny, the article takes the monopole question seriously: Dirac's 1931 argument that even one magnetic charge anywhere would explain why electric charge comes in integer lumps — a result so beautiful that the searches never stopped, and the reason $\nabla\cdot\vec B = 0$ is best read not as dogma but as a standing experimental scoreboard: monopoles 0, closed loops everything.

Key takeaways

The magnetic Gauss law $\nabla\cdot\vec B = 0$ holds everywhere: magnetic field lines have no sources or sinks — no magnetic charge exists (so far as any experiment has ever seen).
It is not an independent postulate of magnetostatics: the Biot–Savart law implies it — the divergence of every field built from $\vec J\times\hat{\boldsymbol{\mathscr{r}}}/\mathscr{r}^2$ vanishes identically, by pure vector-calculus structure.
Integral form: $\oint_S\vec B\cdot\mathrm{d}\vec a = 0$ through every closed surface — whatever flux enters must leave.
Corollary with a future: the flux through an open surface depends only on its boundary curve — the well-definedness that Faraday's law, inductance, and flux quantization all inherit.
Structural payoff: a divergence-free field is always a curl — $\vec B = \nabla\times\vec A$ exists globally, launching the vector potential two articles ahead.
Dirac's counterfactual: a single monopole of charge $q_m$ anywhere would force $q_eq_m = 2\pi n\hbar/\mu_0$ — electric charge quantization explained by one particle nobody has found. Searches (Cabrera, MoEDAL, IceCube) continue.
KEY IDEA: $\nabla\cdot\vec B = 0$ everywhere — magnetic field lines close on themselves, no magnetic charge sources them, and every magnetic field is the curl of a vector potential.

The core insight

$\nabla\cdot\vec B = 0$ everywhere — magnetic field lines are closed loops, there is no magnetic charge, and $\vec B$ is always the curl of a vector potential.

Where this lives today

The law is load-bearing infrastructure across physics and engineering. Every computational-electromagnetics code enforces or monitors $\nabla\cdot\vec B = 0$ (divergence cleaning is a named subroutine in magnetohydrodynamics simulation, where numerical drift can create spurious monopoles that wreck plasma dynamics); every flux-based device spec — transformer, inductor, SQUID, flux-gate compass — quotes flux-through-a-loop on the strength of the boundary theorem. In superconductors, flux through a ring is not only well defined but quantized in units $h/2e = 2.07\times10^{-15}$ Wb, the operating currency of SQUID magnetometers reading femtotesla brain fields. The monopole hunt remains one of experimental physics' great open stakeouts. Cabrera's 1982 Valentine's Day SQUID event — a single, perfect one-Dirac-charge flux jump, never repeated — is the field's most tantalizing ghost story. Modern searches ring the LHC (MoEDAL's trapping detectors), watch for monopole tracks in polar ice (IceCube), and comb ancient minerals; all null so far, pushing monopole fluxes below one per square kilometre per century. Meanwhile condensed matter delivered a twist: "magnetic monopole" quasiparticles in spin ice — emergent endpoints of flipped-spin strings that hop like free magnetic charges within the material. They obey a Coulomb law and even a Dirac-string picture, but they are collective excitations, not fundamental particles: $\nabla\cdot\vec B = 0$ holds exactly even there, with the "monopoles" bookkeeping $\vec H$ , not $\vec B$ . The distinction is precisely this article's — and grand-unified theories still predict the real thing, mass $\sim10^{16}$ GeV, which is one reason inflationary cosmology was invented: to explain where they all went.

Origins · 1 minPaul Dirac and how this was discovered

Paul Dirac

1902–1984 · Bristol, Cambridge, Tallahassee

Dirac built quantum mechanics' deepest structures with an engineer's economy — the transformation theory, the relativistic electron equation that predicted antimatter, quantum field theory's foundations — guided by a conviction that beautiful mathematics is physics' best compass. His 1931 monopole paper is the conviction at full strength: analyzing how wavefunction phases wind around a hypothetical magnetic charge's vector-potential "string," he showed consistency demands $q_eq_m = 2\pi n\hbar/\mu_0$ — so a single monopole anywhere would explain electric charge quantization, a fact otherwise unexplained. "One would be surprised if Nature had made no use of it," he wrote; Nature, so far, declines to confirm. The same paper's mathematics — phases, strings, potentials-over-fields — grew into gauge theory's fiber-bundle language, and the monopole became the theoretical laboratory (from 't Hooft–Polyakov solitons to dualities) where much of modern field theory was invented. The scoreboard stands: the particle, zero sightings; the idea, foundational.

History

1269

— Petrus Peregrinus maps a lodestone's poles and finds they cannot be separated — the law's first data point.

1600

— Gilbert's De Magnete declares Earth itself a magnet; poles always paired, never alone.

1820

— Ampère reads Ørsted's discovery as doctrine: all magnetism is circulating current — no magnetic charge required.

1873

— Maxwell writes div B = 0 among his equations: the eloquent zero.

1931

— Dirac shows one monopole would quantize electric charge — the hunt acquires its trophy value.

1982

— Cabrera's SQUID records a single perfect Dirac-charge event on Valentine's Day; it never repeats.

2009

— Spin-ice experiments realize emergent monopole quasiparticles — Coulombic, mobile, and not fundamental: div B still exactly zero.

The physical picture

Put a balloon — any closed surface, any size, anywhere — into an electric field, and Gauss's law reads its flux like a meter: net lines out = charge inside. Put the same balloon into a magnetic field and the meter always reads zero. Around a wire, the circular field lines pierce the balloon in and out in matched pairs. Around a bar magnet, lines that exit near the north face curve around and re-enter near the south. Slide the balloon inside the magnet and the lines run straight through it, in one face and out the other. There is no placement, in any magnetic field ever measured, that traps a net outflow — because there is nothing inside for lines to start from.

Figure

The balloon test: electric fields have sources, magnetic fields only throughways

The bar-magnet experiment makes the same point with a hacksaw. Cut a magnet hoping to separate its poles, and each half instantly presents a complete north-south pair — because the "poles" were never point sources at all, only the places where internal field lines (which run straight through the magnet's body, south face to north face inside) exit and re-enter. The lines are closed loops threading magnet and exterior alike; the pole faces are doors, not springs. Ampère's old hypothesis, vindicated by atomic physics, completes the picture: magnet = swarm of microscopic current loops, and every current loop's field, by the previous articles' geometry, wraps and closes. One more consequence deserves its physical statement before the math: because no lines end anywhere, the flux through a loop of wire is a property of the loop alone. Drape a flat disk across the loop or a billowing parachute — same flux, because the closed surface formed by the two together traps zero. When Faraday's law arrives (chapter E6) and says EMFs track "the change of flux through the circuit," it is this theorem that makes the phrase meaningful.

From Biot–Savart to the law

Magnetostatics does not need to assume the absence of magnetic charge; it inherits it from the machinery that builds the fields. The claim to prove:

\nabla\cdot\vec B = 0 \qquad (1)

where: the divergence is taken at any field point, for the $\vec B$ of any steady current distribution via the Biot–Savart law. Zero on the right — compare Gauss's law $\nabla\cdot\vec E = \rho/\varepsilon_0$ , where the right side is the whole point. The pair of divergences is the cleanest statement in physics of "electric charge exists; magnetic charge does not."

Derivation ABiot-Savart implies div B = 0 — no monopoles by construction

The goal is equation (1) directly from the volume Biot–Savart law — showing the divergence-free property is manufactured into every magnetostatic field by the cross-product structure, not imposed as an extra law. The strategy: push $\nabla\cdot$ under the source integral (legal — it differentiates field coordinates, the integration is over source coordinates), expand the integrand with the product rule for $\nabla\cdot(\vec X\times\vec Y)$ , and watch both resulting terms die for structural reasons. Note what is not used: the closedness of circuits, or even $\nabla\cdot\vec J = 0$ — this proof is pure geometry, which is why even a lone element's field is divergence-free.

Step 1. Start from the volume law and move the divergence inside the integral:

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

Why this is legal: $\nabla$ differentiates with respect to the field point $\vec r$ , while the integral runs over source points $\vec r'$ — independent variables; the derivative slides through.

Step 2. Expand with the divergence-of-a-cross-product identity (E0's product rules):

\nabla\cdot\left(\vec J\times\frac{\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}\right) = \frac{\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}\cdot\left(\nabla\times\vec J\right) - \vec J\cdot\left(\nabla\times\frac{\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2}\right).

Step 3. Kill the first term. $\vec J(\vec r')$ depends only on source coordinates, so every field-coordinate derivative of it vanishes: $\nabla\times\vec J = 0$ .

Step 4. Kill the second term. The Coulomb-kernel identity from the vector-calculus chapter, $\hat{\boldsymbol{\mathscr{r}}}/\mathscr{r}^2 = -\nabla(1/\mathscr{r})$ , exposes the kernel as a gradient — and the curl of a gradient vanishes identically:

\nabla\times\frac{\hat{\boldsymbol{\mathscr{r}}}}{\mathscr{r}^2} = -\nabla\times\nabla\!\left(\frac{1}{\mathscr{r}}\right) = 0.

Why this is the heart: the same kernel, in Coulomb's law, sits alone and delivers a delta-function divergence — electric charge. In Biot–Savart it sits inside a cross product, where only its curl can contribute to the divergence — and as a pure gradient it has none. The 90° twist that makes magnetic fields circulate is exactly what deprives them of sources.

Step 5. Both terms vanish for every source element, so the integral vanishes:

\nabla\cdot\vec B = 0,

everywhere, for any steady $\vec J$ — no exceptions, no special geometries. The only ingredients are the two E0 identities $\nabla\times(\nabla f) = 0$ and the cross-product divergence rule, both recalled from the vector-calculus chapter rather than re-proved here. The contrast case is worth holding in mind: the kernel alone gives $\nabla\cdot(\hat{\boldsymbol{\mathscr{r}}}/\mathscr{r}^2) = 4\pi\delta^3(\boldsymbol{\mathscr{r}})$ — the E0 delta-function result that powers Gauss's law — highlighting that the difference between electric and magnetic divergence laws is structural, not accidental.

The integral form follows by the divergence theorem (E0): integrate (1) over any volume, convert, and

\oint_S \vec B\cdot\mathrm{d}\vec a = 0 \qquad (2)

where: $S$ — any closed surface whatsoever. This is "Gauss's law for magnetism," fourth of Maxwell's equations, with the eloquent zero where enclosed charge would sit. Its working corollary is used daily:

Figure

The pod argument: two surfaces on one rim must carry the same flux

Go deeper · Derivation B · 1 minFlux belongs to the boundary — the theorem Faraday's law will stand on

Derivation BFlux belongs to the boundary — the theorem Faraday's law will stand on

The goal is the statement that makes "the flux through a circuit" a legal phrase: for a fixed boundary curve, the magnetic flux $\Phi_B = \int_S\vec B\cdot\mathrm{d}\vec a$ is the same for every open surface $S$ spanning that curve. The strategy is a two-surface bookkeeping trick on equation (2) — one of the tidiest arguments in field theory, worth internalizing as a method: to compare two open surfaces, glue them into a closed one.

Step 1. Take two open surfaces $S_1$ and $S_2$ sharing the same boundary curve $\partial S$ — say, a flat disk across a wire loop and a dome bulging high above it — both oriented consistently with the loop (right-hand rule).

Step 2. Glue them. $S_1$ together with $S_2$ -reversed forms a closed surface (disk bottom + dome top = a closed pod). Apply (2) to the pod:

\Phi_1 - \Phi_2 = \oint_{\rm pod}\vec B\cdot\mathrm{d}\vec a = 0.

Why the minus sign: reversing $S_2$ 's orientation to make the pod's normals point consistently outward flips the sign of its flux — the bookkeeping that turns "flux through two surfaces" into "flux out of one closed surface."

Step 3. Conclude $\Phi_1 = \Phi_2$ : any two spanning surfaces carry identical flux, so

\Phi_B = \Phi_B(\partial S\ \text{only}) \qquad (3)

Why physics leans on this: Faraday's law will declare $\mathcal{E} = -\mathrm{d}\Phi_B/\mathrm{d}t$ for a circuit — meaningless unless $\Phi_B$ is the circuit's own property, independent of an arbitrary choice of spanning surface. Inductance ( $\Phi = LI$ ), SQUID magnetometry, and flux quantization in superconductors all quote flux-through-a-loop on the strength of (3). Worked example 2 runs the theorem with numbers: a hemisphere's curved surface carries exactly its equatorial disk's flux, no surface integral over the dome required. *Note that the argument fails exactly when equation (2) fails: if a monopole sat between the two surfaces, the closed-pod flux would no longer be zero, and flux through the loop would depend on which surface you chose — which is precisely how monopole detectors such as Cabrera's SQUID loop were designed to trigger.*

The last structural payoff: a field with no divergence is always somebody's curl. The Helmholtz machinery of the vector-calculus chapter guarantees that (1) is exactly the integrability condition for the existence of a vector potential,

\vec B = \nabla\times\vec A \qquad (4)

where: $\vec A$ — the magnetic vector potential, with $\nabla\cdot(\nabla\times\vec A) \equiv 0$ closing the consistency loop identically. Why $\vec B$ gets a vector potential while $\vec E$ enjoyed a scalar one is now transparent: $\vec E$ 's curl vanished (statics), so it was a gradient; $\vec B$ 's divergence vanishes, so it is a curl. The construction, gauge freedom included, is owned by Magnetic Vector Potential two articles ahead — but its birth certificate is signed here.

Go deeper · Derivation C · 2 minWhat one monopole would buy — Dirac's quantization argument

Derivation CWhat one monopole would buy — Dirac's quantization argument

The goal is the most celebrated "what if" in electromagnetism: if even one magnetic charge $q_m$ exists anywhere, quantum mechanics forces the product $q_eq_m$ to be quantized — and therefore explains why electric charge itself comes in integer multiples. The strategy here is the semiclassical angular-momentum route (Thomson's, polished by Saha and Wilson): compute the angular momentum stored in the fields of a static charge–monopole pair, notice it is independent of their separation, and quantize it. This is honest about its register: Dirac's own 1931 argument runs through vector-potential strings and wavefunction phases (a preview of gauge theory beyond this article); the field-momentum route gets the same answer with tools this chapter owns.

Step 1. Set up the pair: electric charge $q_e$ at the origin, hypothetical monopole $q_m$ (radial $\vec B = \frac{\mu_0}{4\pi}\frac{q_m}{r^2}\hat{\vec r}$ from it) a distance $d$ away. Crossed static $\vec E$ and $\vec B$ fields fill space, and electromagnetic fields carry momentum density $\propto \vec E\times\vec B$ (the conservation-laws chapter's business; we borrow only its existence).

Step 2. Compute the total field angular momentum about the pair's axis. The integral is a classic (Griffiths Prob. 8.19); every factor of distance cancels and

L_{\rm field} = \frac{\mu_0\,q_eq_m}{4\pi},

independent of the separation $d$ — the geometric miracle that makes the argument possible: the answer cannot depend on where the particles happen to be, so it is a property of the pair's very existence.

Step 3. Quantize. Quantum mechanics allows angular momentum only in half-integer units of $\hbar$ ; demanding $L_{\rm field} = n\hbar/2$ :

q_e\,q_m = \frac{2\pi n\hbar}{\mu_0}, \qquad n \in \mathbb{Z} \qquad (5)

Step 4. Read it both ways. Given the electron's charge, the minimal monopole is enormous in coupling: $q_m^{(1)} = 2\pi\hbar/\mu_0 e = h/\mu_0 e \approx 3.3\times10^{-9}\ \mathrm{A\,m}$ — equivalently $\mu_0q_m = h/e = 4.14\times10^{-15}\ \mathrm{Wb}$ , one full flux quantum's worth of flux fanning out of the particle, making monopoles gloriously conspicuous to a flux detector (twice the superconducting flux quantum $h/2e$ , whose Cooper-pair factor of 2 is a different story). Read the other way — Dirac's way — one monopole anywhere in the universe forces every electric charge to be an integer multiple of a fundamental unit, because (5) must hold for each charge separately. Charge quantization, observed to exquisite accuracy and otherwise unexplained in classical electromagnetism, would have a reason. The field-angular-momentum result is the classic "Thomson dipole" calculation; in SI conventions Dirac's condition is often written $eg = 2\pi n\hbar$ with $g$ absorbing $\mu_0$ , which is equivalent to (5). The minimal charge evaluates to $h/(\mu_0 e) = 3.29\times10^{-9}$ A·m. A note on register: the half-integer quantization of field angular momentum used here is the semiclassical shortcut; Dirac's own string/phase argument and the modern fiber-bundle formulation place (5) on rigorous footing and yield the same condition.

Limiting cases

Case	Result	Meaning
any closed surface	$\oint\vec B\cdot\mathrm{d}\vec a = 0$	in = out, no exceptions measured, ever
two surfaces, one boundary	$\Phi_1 = \Phi_2$	"flux through a circuit" is well defined
inside a bar magnet	lines run straight through	poles are doors, not sources; cutting makes pairs
candidate field with $\nabla\cdot\vec B \neq 0$	not a magnetic field	the legality test of worked example 1
contrast: $\nabla\cdot\vec E = \rho/\varepsilon_0$	electric lines end on charge	the asymmetry is empirical, not mathematical
if one monopole existed	$q_eq_m = 2\pi n\hbar/\mu_0$	charge quantization explained (Dirac)
structural corollary	$\vec B = \nabla\times\vec A$ always	vector potential exists; gauge story begins

Common trapField lines close — but div B = 0 is the law, not the slogan

The slogan "magnetic field lines form closed loops" is a good picture and a slightly false theorem: divergence-free fields can have lines that spiral forever on a torus without closing (a real subtlety in tokamak fields), or run to infinity. What is exactly true is (1) and (2): no line ever begins or ends, and no closed surface traps net flux. Two more traps ride with this topic. First, "poles" language: the north face of a magnet is where internal lines exit — it is not a reservoir of magnetic charge, and treating pole strength as literal charge works only as a far-field bookkeeping fiction (the matter chapter's $\vec H$ -formalism makes it respectable). Second, the divergence theorem's reach: equation (2) holds for closed surfaces; for open surfaces flux is generally nonzero — that is the whole point of flux as a useful quantity — and it is only the boundary-dependence theorem (3) that lets you swap one open surface for another.

Worked examples

1 · Is this field legal? (confidence builder)

A simulation hands you a candidate magnetostatic field in a region: $\vec B = k\,(x\,\hat{\vec x} + y\,\hat{\vec y})$ with $k = 0.020\ \mathrm{T/m}$ . A colleague proposes instead $\vec B' = k\,(x\,\hat{\vec x} - y\,\hat{\vec y})$ . Which, if either, could be a real magnetic field?

Step 1. Apply the legality meter — equation (1). For $\vec B$ : $\nabla\cdot\vec B = k\,\partial_x x + k\,\partial_y y = 2k = 0.040\ \mathrm{T/m} \neq 0$ . Illegal: this field pours net flux out of every box — it would need magnetic charge filling the region ( $\rho_m = 2k/\mu_0$ in monopole units), and no current arrangement whatsoever can produce it.

Step 2. For $\vec B'$ : $\nabla\cdot\vec B' = k - k = 0$ ✓ everywhere. Admissible — and recognizable: it is the field near the center of a quadrupole magnet, the beam-focusing workhorse (the anti-Helmholtz midplane of the previous article's catalog).

Step 3. Audit a box to feel the theorem: a unit cube at the origin in field $\vec B$ has outflow $2k \times V = 0.040\ \mathrm{Wb}$ per cubic metre of enclosed "nothing" — the meter reading that instantly falsifies the simulation. Divergence is the first sanity check every computed or measured field map must pass; magnetics software runs precisely this test as a solver diagnostic.

2 · The hemisphere and the free lunch (exam level)

A uniform field $B = 0.50\ \mathrm{T}$ points along $\hat{\vec z}$ . A hemispherical dome of radius $R = 10\ \mathrm{cm}$ sits with its equatorial rim in the $xy$ -plane. Find the flux through the curved dome surface — by theorem, then by honest integration.

Step 1. By the boundary theorem (3): the dome and the flat equatorial disk share the boundary circle, so their fluxes are equal. The disk is trivial:

\Phi_{\rm dome} = \Phi_{\rm disk} = B\,\pi R^2 = (0.50)\,\pi(0.10)^2 = \boxed{1.57\times10^{-2}\ \mathrm{Wb}}.

Step 2. Honest integration as the cross-check. On the dome, $\vec B\cdot\mathrm{d}\vec a = B\cos\theta\,(R^2\sin\theta\,\mathrm{d}\theta\,\mathrm{d}\phi)$ with $\theta$ the polar angle:

\Phi = BR^2\int_0^{2\pi}\!\mathrm{d}\phi\int_0^{\pi/2}\cos\theta\sin\theta\,\mathrm{d}\theta = BR^2\,(2\pi)\left[\tfrac{\sin^2\theta}{2}\right]_0^{\pi/2} = B\pi R^2\ ✓.

Step 3. The lesson generalizes with force: for any bulging, crumpled, or exotic surface spanning that same rim — in any field, uniform or not — the flux equals the disk's. When Faraday's law (E6) asks for "the flux through this loop of wire," you may always compute it on the most convenient spanning surface; equation (2) pays for the free lunch.

Step 4. Closed-surface closure: dome plus disk form a closed surface, and their fluxes, with consistent outward orientation, cancel to zero exactly — the balloon test passing in the lab frame.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §5.3.1–5.3.2 (and Prob. 8.19 for the charge–monopole angular momentum) · Jackson, Classical Electrodynamics, 3rd ed., §5.3 and §6.12–6.13 · Purcell & Morin, Electricity and Magnetism, 3rd ed., Ch. 6 · Zangwill, Modern Electrodynamics, §10.2. The divergence derivation uses the E0 identities $\nabla(1/\mathscr{r}) = -\hat{\boldsymbol{\mathscr{r}}}/\mathscr{r}^2$ , the cross-product divergence rule, and $\nabla\times\nabla f = 0$ , all from the vector-calculus chapter. The boundary-flux theorem is the standard two-surface argument. The Dirac condition in SI form is $q_eq_m = 2\pi n\hbar/\mu_0$ , obtained via the semiclassical field-angular-momentum route. Key numbers to keep in hand: hemisphere flux $B\pi R^2 = 1.57\times10^{-2}$ Wb (at 0.5 T, $R = 10$ cm); legality test $\nabla\cdot\vec B = 2k \neq 0$ vs $0$ ; minimal Dirac charge $h/\mu_0 e = 3.29\times10^{-9}$ A·m, equivalent flux $h/e = 4.14\times10^{-15}$ Wb (distinct from the superconducting flux quantum $h/2e$ ). Experimental status: Cabrera (1982) single candidate event; MoEDAL and IceCube searches remain null.

Builds on: Biot-Savart Law · Magnetic Field of Steady Currents · Electric Current and Current Density

Leads to: Magnetic Vector Potential · Curl of B · Ampère's Law · Gauge Freedom and Coulomb Gauge

Is div B = 0 a law of nature or a mathematical identity?

Within magnetostatics, an identity — Derivation A shows every Biot–Savart field is divergence-free by construction, no physics input beyond the law itself. But the deeper claim, " $\nabla\cdot\vec B = 0$ exactly, always, for all fields however produced," is an experimental law that could fail tomorrow: one confirmed monopole and Maxwell's fourth equation acquires a right-hand side ( $\nabla\cdot\vec B = \mu_0\rho_m$ ), with a corresponding magnetic-current term joining Faraday's law. The mathematics is even ready — the extended "duality-symmetric" Maxwell equations are textbook exercises. Physics keeps the zero because seven centuries of searches keep returning it, not because the formalism requires it.

If magnetic field lines are closed, where exactly do a bar magnet's field lines go inside it?

Straight through the body, from the south face to the north face — completing, inside the material, the loops seen outside. This surprises intuitions trained on "field goes from north to south": that is true outside only. Inside, $\vec B$ runs south-to-north face, and indeed into the region iron-filing pictures leave blank. (The auxiliary field $\vec H$ of the matter chapter genuinely does reverse inside a magnet — one of several reasons $\vec B$ and $\vec H$ must not be conflated, previewed here because the distinction is born exactly at this law: $\vec B$ is the divergence-free one.)

Doesn't the north pole of a magnet act exactly like a positive magnetic charge? Compass needles certainly think so.

As a far-field approximation, yes — and it is a legitimate calculational fiction. The field outside a long bar magnet is well modeled by fictitious "pole charges" $\pm q_m^{\rm eff}$ on the end faces (the matter chapter's magnetic-scalar-potential method formalizes this for $\vec H$ ). The fiction breaks exactly where Derivation A says it must: on closed surfaces and inside the material. A balloon around the "north charge" still reads zero net $\vec B$ -flux, because the flux that fans out of the pole face returns through the magnet's body — the internal throughway the pole model omits. Use poles for quick exterior estimates; trust only closed loops for the books.

What would actually happen in a detector if a monopole passed through — how would we know?

The cleanest signature is this article's integral law failing on schedule. Pass a monopole through a superconducting ring and the flux through the ring changes permanently by exactly $\mu_0q_m = nh/e$ — a quantized, latching jump that no dipole, however violent, can mimic (dipoles give transient wiggles netting zero). That is Cabrera's induction technique, and its 1982 single perfect jump remains unrepeated and unexplained. Complementary signatures: anomalously heavy ionization along a track (monopoles couple like a charge of $\sim68.5e$ ), and trapping in material followed by SQUID scanning — MoEDAL's method. The detectors are, in effect, machines for auditing $\oint\vec B\cdot\mathrm{d}\vec a = 0$ event by event.

Are the monopoles discovered in spin ice the real thing?

No — and the distinction is a beautiful exam on this article. Spin-ice "monopoles" are emergent quasiparticles: endpoints of strings of flipped atomic moments, which behave within the material like mobile magnetic charges obeying a Coulomb law. But they source $\vec H$ , not $\vec B$ : every atomic moment remains a closed-loop (dipole) source, $\nabla\cdot\vec B = 0$ holds exactly at every point of the sample, and no net $\vec B$ -flux ever crosses a closed surface around one. They cannot be extracted into vacuum, and passing one through Cabrera's ring would leave no latching jump. They are magnificent condensed-matter physics — Coulomb phases, Dirac strings made visible — and zero threat to Maxwell's zero.

Why does div B = 0 guarantee a vector potential exists — and why should I want one?

The vector-calculus chapter's converse theorems say: curl-free fields are gradients, divergence-free fields are curls (on well-behaved domains). Equation (1) is exactly the hypothesis of the second theorem, so some $\vec A$ with $\vec B = \nabla\times\vec A$ exists globally — and the explicit construction $\vec A = \frac{\mu_0}{4\pi}\int\vec J/\mathscr{r}\,\mathrm{d}\tau'$ arrives two articles ahead. Why want it: $\vec A$ turns three coupled field components into a simpler integral (no cross product in the integrand), converts flux into a line integral ( $\Phi = \oint\vec A\cdot\mathrm{d}\vec\ell$ , by Stokes — the boundary theorem (3) made manifest), and carries the gauge freedom that becomes the organizing principle of quantum electrodynamics. In quantum mechanics $\vec A$ is more physical than classical intuition suggests: the Aharonov–Bohm effect shifts electron interference in regions where $\vec B = 0$ but $\vec A$ is not.

How well tested is div B = 0, numerically speaking?

On the monopole side: searches in bulk matter (moon rocks, ocean sediments, meteorites) bound monopole densities below $\sim10^{-29}$ per nucleon; cosmic-flux bounds (Parker limit and its descendants, IceCube, ANTARES) sit below $\sim10^{-18}\ \mathrm{cm^{-2}s^{-1}sr^{-1}}$ for a wide mass range — roughly, less than one crossing per square kilometre per century. On the field-equation side: precision magnetometry and the consistency of flux-based metrology (the quantized $h/2e$ in superconducting rings, reproducible to parts in $10^{9}$ ) would betray any local violation. It is among the best-confirmed zeros in physics — which is precisely why a single confirmed exception would be revolutionary, and why the experiments politely continue.