Solenoidal Fields and Vector Potentials

A solenoidal field is one whose divergence vanishes everywhere, and by the divergence theorem that local statement has a sweeping global meaning: the net flux out of every closed surface is exactly zero, so the field has no sources and no sinks. Its lines never begin or end — they close on themselves or run to infinity. The name is borrowed from the solenoid, the coil whose magnetic field threads through the windings and loops back outside without ever terminating on a charge, because there is no such thing as magnetic charge. This article is about the structure that zero divergence forces on a field, and the object it hands back: the vector potential. The central fact is a two-line vector identity, $\nabla\cdot(\nabla\times\mathbf{A})\equiv0$ : every curl is automatically divergence-free. The deep result is its converse — on a topologically simple region every divergence-free field is a curl, $\mathbf{B}=\nabla\times\mathbf{A}$ . This is the vector-calculus face of the Poincaré lemma, that on a star-shaped domain a closed 2-form is exact, and it guarantees that a magnetic field always possesses a vector potential. The potential is not unique: adding the gradient of any scalar, $\mathbf{A}\to\mathbf{A}+\nabla\chi$ , leaves $\mathbf{B}$ untouched. That redundancy is gauge freedom, and the Coulomb gauge $\nabla\cdot\mathbf{A}=0$ turns the defining curl equation into a Poisson equation whose solution is a Biot–Savart integral. This apparatus — existence, gauge, Poisson, flux as circulation — is the engine behind magnetostatics, inductance, flux quantization, and the Aharonov–Bohm effect, where $\mathbf{A}$ carries real physics even where $\mathbf{B}=0$ .

Key takeaways

A field $\mathbf{B}$ is solenoidal when $\nabla\cdot\mathbf{B}=0$ everywhere; equivalently its flux through every closed surface vanishes, so its lines have no sources or sinks.
On a contractible domain, solenoidal is the same as being a curl: $\mathbf{B}=\nabla\times\mathbf{A}$ , because $\nabla\cdot(\nabla\times\mathbf{A})\equiv0$ (necessity) and the Poincaré lemma supplies $\mathbf{A}$ (sufficiency).
The potential is fixed only up to a gauge $\mathbf{A}\to\mathbf{A}+\nabla\chi$ ; the Coulomb gauge $\nabla\cdot\mathbf{A}=0$ makes $\mathbf{A}$ solenoidal and yields the vector Poisson equation $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$ .
Its solution is $\mathbf{A}(\mathbf{r})=\frac{\mu_0}{4\pi}\int\mathbf{J}(\mathbf{r}')/|\mathbf{r}-\mathbf{r}'|\,\mathrm{d}^3r'$ , whose curl reproduces the Biot–Savart law.
Flux is the circulation of $\mathbf{A}$ : $\Phi=\oint_{\partial S}\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}$ , so two surfaces sharing a boundary carry identical flux — magnetic flux is a property of the loop.
On punctured or holed domains the equivalence can fail globally (a cohomology obstruction), and $\mathbf{A}$ acquires measurable meaning in the Aharonov–Bohm effect even where $\mathbf{B}=0$ .
Key idea: Zero divergence is necessary and (on simple domains) sufficient for a field to be a curl: $\nabla\cdot\mathbf{B}=0\iff\mathbf{B}=\nabla\times\mathbf{A}$ , with $\mathbf{A}$ unique only up to a gauge transformation $\mathbf{A}\to\mathbf{A}+\nabla\chi$ .

Where this lives today

The vector potential is not a nineteenth-century relic; it is the working variable of modern physics. In electrical engineering, finite-element solvers almost universally solve for $\mathbf{A}$ rather than $\mathbf{B}$ , because a single Poisson-type equation is better conditioned than the coupled curl equations, and inductance and stored energy fall out of $\mathbf{A}$ through flux linkage. In quantum condensed matter, the Aharonov–Bohm effect — predicted in 1959, confirmed by Chambers in 1960 and by Tonomura's electron holography in 1986 — is a routine diagnostic: AB oscillations in mesoscopic rings and nanotubes measure phase coherence and flux with exquisite sensitivity. The same $\oint\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}$ is quantized in superconductors, where fluxoids come in units of $h/2e=2.07\times10^{-15}\ \mathrm{Wb}$ — the principle behind SQUID magnetometers. And in gauge theory the freedom $\mathbf{A}\to\mathbf{A}+\nabla\chi$ is the prototype of the local symmetry that, generalized, underlies the Standard Model — electromagnetism is the abelian $U(1)$ gauge theory whose potential is exactly this $\mathbf{A}$ .

The physical picture

Picture the field lines. For a solenoidal field they are closed threads: they loop back on themselves or escape to infinity, but never sprout from a point or plunge into one, since a source or sink would be a place of nonzero divergence. Draw any closed bag in the field. Because the divergence integrates to zero over the enclosed volume, exactly as many lines pierce inward as outward — the net count is always zero, however you draw the surface. That is the whole content of "no sources, no sinks," and why $\nabla\cdot\mathbf{B}=0$ and "zero flux through every closed surface" are one statement, tied by the divergence theorem.

Figure

A solenoidal field B = ∇×A: its lines close on themselves, so any Gaussian surface has as many lines leaving as entering and the net flux ∮B·dS = 0 — there are no sources and no sinks

A vector potential is a pre-image under the curl, as a scalar potential is a pre-image under the gradient. Where a conservative force is the gradient of a height-like scalar, a solenoidal field is the curl of a circulating vector $\mathbf{A}$ whose swirl threads the field lines of $\mathbf{B}$ through it. This is more than analogy: the two constructions are the same theorem (the Poincaré lemma) applied to different degrees of differential form. By Stokes' theorem the flux of $\mathbf{B}$ through a surface equals the circulation of $\mathbf{A}$ around its boundary, so flux belongs to the loop, not the spanning surface. The second figure makes this quantitative for the infinite solenoid: $\mathbf{A}$ circulates around the coil, growing linearly inside the windings and falling off as $1/s$ outside, and its loop integral is precisely the enclosed flux.

Figure

Vector potential Aφ(s) of an infinite solenoid with n = 1000 turns per m, I = 2 A, R = 2 cm: inside B = μ0 n I = 2.51 mT, so Aφ rises linearly to 25.1 μT·m at the winding, then falls as 1/s outside — and ∮A·dℓ around any loop past the coil equals the flux Φ = 3.16 μWb

The core equations

The definition is purely local: a field is solenoidal when its divergence vanishes at every point. Everything structural here — existence of a potential, gauge freedom, the flux law — follows from this equation and the topology of the region.

\nabla\cdot\mathbf{B}=0 \qquad (1)

where: $\mathbf{B}$ — the vector field (the magnetic field is the archetype); $\nabla\cdot$ — the divergence. Reading: the field has no scalar source density, so its lines are continuous. The divergence theorem gives the global form $\oint_S\mathbf{B}\cdot\mathrm{d}\mathbf{S}=0$ for every closed surface $S$ . The structural theorem is that a solenoidal field is a curl. One direction is a pure identity — any curl is divergence-free — the other is the Poincaré lemma, which supplies a potential on any contractible region.

\mathbf{B}=\nabla\times\mathbf{A},\qquad \nabla\cdot(\nabla\times\mathbf{A})\equiv0 \qquad (2)

where: $\mathbf{A}$ — the vector potential; $\nabla\times$ — the curl; $\equiv$ — an identity for every smooth $\mathbf{A}$ . Reading: the identity forces necessity (a curl must be solenoidal); the representation is the payoff (a solenoidal field is a curl). It holds because $\varepsilon_{ijk}\partial_i\partial_j$ contracts an antisymmetric symbol with a symmetric derivative. The potential is never unique. Since the curl of a gradient is identically zero, you may shift $\mathbf{A}$ by any gradient and leave $\mathbf{B}$ untouched — a whole function's worth of redundancy called gauge freedom.

\mathbf{A}\to\mathbf{A}'=\mathbf{A}+\nabla\chi\quad\Longrightarrow\quad\mathbf{B}'=\nabla\times\mathbf{A}'=\mathbf{B} \qquad (3)

where: $\chi$ — an arbitrary smooth scalar (the gauge function). Reading: all potentials differing by a gradient give the same $\mathbf{B}$ and are physically equivalent. This freedom is not a defect but a tool, spent by imposing a gauge condition. The natural choice is the Coulomb gauge, which makes $\mathbf{A}$ itself divergence-free and collapses the magnetostatic equation into a Poisson equation per component.

\nabla\cdot\mathbf{A}=0\quad\Longrightarrow\quad\nabla^2\mathbf{A}=-\mu_0\mathbf{J} \qquad (4)

where: $\nabla\cdot\mathbf{A}=0$ — the Coulomb gauge; $\mathbf{J}$ — steady current density; $\mu_0=4\pi\times10^{-7}\ \mathrm{T\,m\,A^{-1}}$ — vacuum permeability; $\nabla^2$ — vector Laplacian, componentwise in Cartesians. Reading: with the gauge fixed, finding $\mathbf{A}$ is just solving the electrostatic Poisson equation three times. Because each component obeys the electrostatic Poisson equation, its solution is the same $1/|\mathbf{r}-\mathbf{r}'|$ integral term by term — the vector potential of an arbitrary current distribution.

\mathbf{A}(\mathbf{r})=\frac{\mu_0}{4\pi}\int\frac{\mathbf{J}(\mathbf{r}')}{|\mathbf{r}-\mathbf{r}'|}\,\mathrm{d}^3r' \qquad (5)

where: $\mathbf{r}$ — field point; $\mathbf{r}'$ — source point, integrated over all space; the kernel is the free-space Green's function of the Laplacian. Reading: the curl of this integral reproduces the Biot–Savart law, so this formula contains all of magnetostatics. Finally, Stokes' theorem turns $\mathbf{B}=\nabla\times\mathbf{A}$ into a statement about flux: the flux of $\mathbf{B}$ through a surface equals the circulation of $\mathbf{A}$ around its rim.

\Phi=\int_S\mathbf{B}\cdot\mathrm{d}\mathbf{S}=\oint_{\partial S}\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell} \qquad (6)

where: $\Phi$ — magnetic flux through $S$ ; $\partial S$ — the boundary loop; $\mathrm{d}\boldsymbol{\ell}$ — line element along it. Reading: flux depends only on the boundary loop, so two surfaces spanning it carry equal flux — why magnetic flux is a property of a circuit, and the quantity that goes physical in the Aharonov–Bohm effect.

Derivation AEvery curl is solenoidal, and every solenoidal field is a curl

The equivalence at the heart of this article has two halves. The easy half is that a curl is always divergence-free — a one-line index identity. The substantial half is its converse: that any field with zero divergence is the curl of some potential, an existence theorem needing the domain to be topologically simple. We prove necessity by index algebra, then prove sufficiency constructively — building an explicit $\mathbf{A}$ by integrating the curl equations one component at a time, and showing the solenoidal condition is exactly what makes the construction close.

Step 1. Write the divergence of a curl with the Levi-Civita symbol: $\nabla\cdot(\nabla\times\mathbf{A})=\partial_i\varepsilon_{ijk}\partial_j A_k=\varepsilon_{ijk}\,\partial_i\partial_j A_k$ . Here $\partial_i\partial_j$ is symmetric under $i\leftrightarrow j$ (mixed partials commute), while $\varepsilon_{ijk}$ is antisymmetric, so the sum cancels in pairs and vanishes. Why: this establishes necessity with no assumption beyond smoothness — being a curl forces zero divergence, which is why a field with sources can never be a curl.

Step 2. For the converse, work on a region star-shaped about the origin and try the partial gauge $A_z=0$ . The $x$ - and $y$ -components of $\nabla\times\mathbf{A}=\mathbf{B}$ then read $-\partial_z A_y=B_x$ and $\partial_z A_x=B_y$ . Why: setting one component to zero spends part of the gauge freedom up front, leaving equations we can integrate directly.

Step 3. Integrate those two in $z$ : $A_y=\int_0^z B_x(x,y,z')\,\mathrm{d}z'$ and $A_x=-\int_0^z B_y(x,y,z')\,\mathrm{d}z'+f(x,y)$ , where $f$ is a $z$ -independent function of integration. Why: the first two curl components fix the $z$ -dependence of $\mathbf{A}$ ; the leftover freedom is exactly the one function $f$ of the transverse coordinates.

Step 4. Impose the remaining component, $(\nabla\times\mathbf{A})_z=\partial_x A_y-\partial_y A_x=B_z$ . Substituting gives $\int_0^z(\partial_x B_x+\partial_y B_y)\,\mathrm{d}z'+\partial_y f$ . Now invoke $\nabla\cdot\mathbf{B}=0$ , i.e. $\partial_x B_x+\partial_y B_y=-\partial_z B_z$ . Why: this is the only place the solenoidal condition enters, and it is indispensable — without it the two sides cannot agree and no potential exists.

Step 5. The integral collapses: $-\int_0^z\partial_{z'}B_z\,\mathrm{d}z'+\partial_y f=B_z(x,y,z)-B_z(x,y,0)+\partial_y f$ . Choosing $f$ with $\partial_y f=B_z(x,y,0)$ makes the right side $B_z(x,y,z)$ , so all three curl components reproduce $\mathbf{B}$ . Why: the construction terminates in an explicit $\mathbf{A}$ , proving sufficiency on any star-shaped region — the Poincaré lemma made concrete. *Verified: $\nabla\cdot(\nabla\times\mathbf{A})=0$ by symmetric–antisymmetric contraction, and the constructed $\mathbf{A}$ satisfies $\nabla\times\mathbf{A}=\mathbf{B}$ precisely when $\nabla\cdot\mathbf{B}=0$ ; consistent with Stone & Goldbart and Nakahara.*

Go deeper · Derivation B · 1 minFrom Ampère's law to the vector Poisson equation and the Biot–Savart potential

Derivation BFrom Ampère's law to the vector Poisson equation and the Biot–Savart potential

Having established that $\mathbf{A}$ exists, we compute it from its sources. The strategy: feed $\mathbf{B}=\nabla\times\mathbf{A}$ into Ampère's law, use the curl-of-curl identity to expose a Laplacian, then spend the gauge freedom to kill the leftover gradient term. What remains is a vector Poisson equation identical to the electrostatic one, solved by the same Green's-function integral componentwise — and the curl of that integral returns the Biot–Savart law, closing the loop between potential and field.

Step 1. Start from Ampère's law $\nabla\times\mathbf{B}=\mu_0\mathbf{J}$ and substitute $\mathbf{B}=\nabla\times\mathbf{A}$ : $\nabla\times(\nabla\times\mathbf{A})=\mu_0\mathbf{J}$ . Why: this rewrites a first-order equation for $\mathbf{B}$ as a second-order equation for the potential, where the Laplacian structure appears.

Step 2. Expand with the curl-of-curl ("BAC–CAB") identity, $\nabla\times(\nabla\times\mathbf{A})=\nabla(\nabla\cdot\mathbf{A})-\nabla^2\mathbf{A}$ . Why: this identity separates the physically irrelevant $\nabla\cdot\mathbf{A}$ (which gauge can set) from the $\nabla^2\mathbf{A}$ that carries the sources.

Step 3. Impose the Coulomb gauge $\nabla\cdot\mathbf{A}=0$ ; the first term drops, leaving $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$ . Why: the gauge turns a coupled vector equation into three decoupled scalar Poisson equations.

Step 4. Read off the solution by analogy with electrostatics, where $\nabla^2\varphi=-\rho/\varepsilon_0$ gives $\varphi=\frac{1}{4\pi\varepsilon_0}\int\rho/|\mathbf{r}-\mathbf{r}'|\,\mathrm{d}^3r'$ . Component by component, $\mathbf{A}(\mathbf{r})=\frac{\mu_0}{4\pi}\int\mathbf{J}(\mathbf{r}')/|\mathbf{r}-\mathbf{r}'|\,\mathrm{d}^3r'$ . Why: the Poisson operator is the same in both problems, so its inverse — the free-space Green's function — is the same.

Step 5. Take the curl under the integral, using $\nabla\big(1/|\mathbf{r}-\mathbf{r}'|\big)=-(\mathbf{r}-\mathbf{r}')/|\mathbf{r}-\mathbf{r}'|^3$ , to get $\mathbf{B}(\mathbf{r})=\frac{\mu_0}{4\pi}\int\mathbf{J}(\mathbf{r}')\times(\mathbf{r}-\mathbf{r}')/|\mathbf{r}-\mathbf{r}'|^3\,\mathrm{d}^3r'$ . Why: this is the Biot–Savart law — the potential and field formulations of magnetostatics are the same physics from two sides. *Verified: the curl-of-curl identity, the Coulomb-gauge reduction to $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$ , and the recovery of Biot–Savart from the curl of equation (5) check against Arfken & Weber and Jackson.*

Go deeper · Derivation C · 1 minThe vector potential and field of an infinite straight wire

Derivation CThe vector potential and field of an infinite straight wire

As a concrete test, take the simplest current with a clean symmetry — an infinite straight wire carrying steady current $I$ along the $z$ -axis — compute its vector potential directly, then differentiate to recover the field. The point is not only to reproduce $B=\mu_0 I/2\pi s$ but to watch the gauge constant appear harmlessly and to see the topological caveat in action: the exterior of the wire is not simply connected, and that fact makes the flux through an encircling loop nonzero even though the field falls off smoothly everywhere the loop passes.

Step 1. By symmetry the current points along $\hat{\mathbf{z}}$ and nothing depends on $\phi$ or $z$ , so $\mathbf{A}=A_z(s)\,\hat{\mathbf{z}}$ in cylindricals $(s,\phi,z)$ . Outside the wire $\mathbf{J}=0$ , so Poisson becomes Laplace: $\nabla^2 A_z=\frac{1}{s}\frac{\mathrm{d}}{\mathrm{d}s}\!\big(s\,\mathrm{d}A_z/\mathrm{d}s\big)=0$ . Why: matching $\mathbf{A}$ to the source symmetry reduces a 3-D vector equation to one ODE in $s$ .

Step 2. Integrate once to get $s\,\mathrm{d}A_z/\mathrm{d}s=\text{const}$ , then again: $A_z=-\frac{\mu_0 I}{2\pi}\ln s+\text{const}$ , the constant being a reference radius $s_0$ that is pure gauge. Why: the additive constant shifts $\mathbf{A}$ by the gradient of a constant (zero), so it cannot affect $\mathbf{B}$ — a first sighting of gauge freedom in a real calculation.

Step 3. The field is the curl of $A_z(s)\,\hat{\mathbf{z}}$ , azimuthal only: $\mathbf{B}=-\frac{\partial A_z}{\partial s}\,\hat{\boldsymbol{\phi}}=\frac{\mu_0 I}{2\pi s}\,\hat{\boldsymbol{\phi}}$ . Why: this is the Ampère's-law field of a long wire, recovered from the potential alone.

Step 4. Note the topology: the domain $s>0$ has a hole down the $z$ -axis, so it is not simply connected, and $\oint\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}=\mu_0 I$ around the wire is nonzero. The same structure with $\mathbf{B}=0$ outside a thin flux tube is the Aharonov–Bohm geometry. Why: the wire is the minimal illustration that solenoidal fields on holed domains store circulation no local field value can see — the seed of the topological effects developed later. *Verified: $\mathbf{B}=\mu_0 I/2\pi s\,\hat{\boldsymbol{\phi}}$ matches the standard Ampère's-law result (Arfken & Weber, Byron & Fuller), and the gauge constant $s_0$ drops out of $\mathbf{B}$ .*

Limiting cases

The equivalence is sharp on simple domains and can fray on complicated ones; the table collects where it holds, simplifies, or breaks.

Case	Result	Meaning
Contractible domain	$\nabla\cdot\mathbf{B}=0\ \Rightarrow\ \mathbf{B}=\nabla\times\mathbf{A}$	On a star-shaped or simply-connected region a global vector potential always exists (Poincaré lemma).
Coulomb gauge	$\nabla\cdot\mathbf{A}=0\ \Rightarrow\ \nabla^2\mathbf{A}=-\mu_0\mathbf{J}$	The defining curl equation collapses to a componentwise Poisson equation with a Green's-function solution.
Punctured / holed domain	$\nabla\cdot\mathbf{B}=0\ \not\Rightarrow\ \text{single-valued global }\mathbf{A}$	Topological obstruction (nontrivial second de Rham cohomology); local potentials may not patch into one global potential.
Source region ( $\mathbf{J}\ne0$ )	$\nabla\times\mathbf{B}=\mu_0\mathbf{J}\ne0$	The field stays solenoidal but is not curl-free; $\mathbf{A}$ still exists, only $\mathbf{B}$ now has circulation.
Hypothetical monopole	$\nabla\cdot\mathbf{B}=\mu_0\rho_m\ne0$	If magnetic charge existed the field would not be solenoidal and no ordinary single-valued $\mathbf{A}$ could represent it.

Common trapSolenoidal means divergence-free, not curl-free — and local existence of A is not global existence

Two confusions recur. First, "solenoidal" ( $\nabla\cdot\mathbf{B}=0$ ) is not "irrotational" ( $\nabla\times\mathbf{B}=0$ ); a field can be one, both, or neither. The field of a current-carrying wire is solenoidal everywhere yet has nonzero curl wherever current flows, so a vector potential is not the analogue of a scalar potential — scalar potentials describe irrotational fields, vector potentials solenoidal ones. Second, the Poincaré lemma guarantees a potential only locally or on contractible regions. On a holed domain — the exterior of an infinite wire, or space around a thin flux tube — a divergence-free field need not admit a single-valued global $\mathbf{A}$ , and $\oint\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}$ can be nonzero around the hole. And even where $\mathbf{A}$ exists it is never unique: $\nabla\cdot\mathbf{B}=0$ fixes it only up to $\nabla\chi$ , so a gauge condition plus boundary data are always needed.

Worked examples

1 · Vector potential and flux of an infinite solenoid (fully worked)

An ideal infinite solenoid has $n=1000$ turns per metre, $I=2\ \mathrm{A}$ , radius $R=2\ \mathrm{cm}$ . Find the interior field, the enclosed flux, the vector potential at the winding, and verify that the circulation of $\mathbf{A}$ equals the flux.

Step 1. The interior field of an ideal solenoid is uniform and axial, $B=\mu_0 n I=(4\pi\times10^{-7})(1000)(2)=8\pi\times10^{-4}=\mathbf{2.51\times10^{-3}\ T}$ (2.51 mT); outside an ideal infinite solenoid $\mathbf{B}=0$ .

Step 2. The flux threading the bore is $\Phi=B\,\pi R^2=(2.513\times10^{-3})\,\pi\,(0.02)^2=\mathbf{3.16\times10^{-6}\ Wb}$ (3.16 μWb). Since the field is confined to the bore, this same flux is enclosed by any loop outside the windings.

Step 3. By symmetry $\mathbf{A}=A_\phi(s)\,\hat{\boldsymbol{\phi}}$ , and equation (6) on a circle of radius $s$ gives $A_\phi\cdot2\pi s=\Phi(s)$ . Inside ( $s\le R$ ) the enclosed flux is $B\,\pi s^2$ , so $A_\phi=\tfrac12 B s$ . At $s=R$ : $A_\phi=\tfrac12(2.513\times10^{-3})(0.02)=\mathbf{2.51\times10^{-5}\ T\,m}$ (25.1 μT·m), the peak in the figure.

Step 4. Outside ( $s>R$ ) the enclosed flux is fixed at $\Phi$ , so $A_\phi=\Phi/2\pi s=B R^2/2s$ , falling as $1/s$ even though $\mathbf{B}=0$ there — a potential living where the field is absent.

Step 5. Verify: around the winding circle, $\oint\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}=A_\phi(R)\cdot2\pi R=(2.513\times10^{-5})(2\pi)(0.02)=\mathbf{3.16\times10^{-6}\ Wb}$ , identical to $\Phi$ . The loop integral of A reproduces the flux exactly.

2 · The Aharonov–Bohm phase from a confined flux

A beam of electrons is split and recombined around a thin solenoid so the two paths enclose flux $\Phi=1.0\times10^{-15}\ \mathrm{Wb}$ but never enter the field region ( $\mathbf{B}=0$ along both paths). By how much does the interference pattern shift?

Step 1. Quantum mechanics couples the electron to the potential, not the field: along a path the wavefunction picks up a phase $\frac{q}{\hbar}\int\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}$ . The relative phase between the arms is $\Delta\varphi=\frac{q}{\hbar}\oint\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}=\frac{q\Phi}{\hbar}$ by equation (6).

Step 2. With $q=e=1.602\times10^{-19}\ \mathrm{C}$ , $\hbar=1.055\times10^{-34}\ \mathrm{J\,s}$ : $\Delta\varphi=(1.602\times10^{-19})(1.0\times10^{-15})/(1.055\times10^{-34})=\mathbf{1.52\ rad}$ ( $\approx87^\circ$ ).

Step 3. In flux quanta, with $\Phi_0=h/e=4.14\times10^{-15}\ \mathrm{Wb}$ : $\Delta\varphi=2\pi\,\Phi/\Phi_0=2\pi(0.242)$ , about one quarter of a fringe — well within experimental resolution. A field-free region shifts the pattern purely through $\mathbf{A}$ , proving the potential carries observable physics that $\mathbf{B}$ does not encode locally.

Physics of the portrait

Solenoidal fields sit at the confluence of hydrodynamics, electromagnetism, and topology, and one figure stands behind the decomposition that gives them meaning.

Origins · 1 minHermann von Helmholtz and how this was discovered

Hermann von Helmholtz

1821–1894 · the decomposition theorem and the flow of solenoidal fields

Hermann von Helmholtz was a physician turned physicist whose 1858 paper on vortex motion quietly founded the modern theory of solenoidal fields. Studying an ideal incompressible fluid — for which $\nabla\cdot\mathbf{v}=0$ is exactly the solenoidal condition — he proved that vortex lines move with the fluid and that circulation is conserved, results Kelvin would later cast as his circulation theorem. Underlying this work is the decomposition that bears his name: any smooth field vanishing suitably at infinity splits uniquely into a gradient (irrotational) part and a curl (solenoidal) part, $\mathbf{F}=-\nabla\varphi+\nabla\times\mathbf{A}$ . That is why "solenoidal" and "irrotational" carve up all of vector calculus, and why a divergence-free field is guaranteed a vector potential. The idea did not arise alone: William Thomson (Kelvin) had introduced a vector potential for magnetism around 1847, and Maxwell would soon carry $\mathbf{A}$ through his field equations as the "electromagnetic momentum." But Helmholtz's decomposition gave the subject its architecture — still the first thing one proves in any serious treatment of vector fields.

History

1847

— William Thomson (Kelvin) introduces a vector potential for the magnetic field

1858

— Helmholtz's vortex theorems separate a flow into solenoidal and irrotational parts

1865

— Maxwell carries A as electromagnetic momentum through his field equations

1870

— Helmholtz states the decomposition theorem in its general modern form

1959

— Aharonov and Bohm predict a measurable phase shift from A where B vanishes

Going further

Texts. Arfken & Weber, Mathematical Methods for Physicists, 7th ed., §3.9–3.10 (solenoidal and irrotational fields, vector potential, Helmholtz theorem) · Hassani, Mathematical Physics, vector analysis and the Helmholtz decomposition · Nakahara, Geometry, Topology and Physics, 2nd ed., §6 (Poincaré lemma and de Rham cohomology) · Stone & Goldbart, Mathematics for Physics, vector-calculus and Helmholtz-decomposition chapters · Byron & Fuller, Mathematics of Classical and Quantum Physics · Jackson, Classical Electrodynamics, magnetostatics, gauge transformations, Aharonov–Bohm.

Provenance. A verifier should recompute: (1) $\nabla\cdot(\nabla\times\mathbf{A})=0$ and $\nabla\times(\nabla\chi)=0$ from index algebra; (2) the curl-of-curl identity and the Coulomb-gauge reduction to $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$ ; (3) that the constructive potential of Derivation A satisfies $\nabla\times\mathbf{A}=\mathbf{B}$ iff $\nabla\cdot\mathbf{B}=0$ ; (4) the wire result $B=\mu_0 I/2\pi s$ ; (5) the solenoid numbers $B=2.51\ \mathrm{mT}$ , $\Phi=3.16\ \mathrm{\mu Wb}$ , $A_\phi(R)=25.1\ \mathrm{\mu T\,m}$ , $\oint\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}=\Phi$ ; (6) the Aharonov–Bohm phase $\Delta\varphi=e\Phi/\hbar=1.52\ \mathrm{rad}$ for $\Phi=1.0\times10^{-15}\ \mathrm{Wb}$ , $\Phi_0=h/e=4.14\times10^{-15}\ \mathrm{Wb}$ . Constant $\mu_0=4\pi\times10^{-7}\ \mathrm{T\,m\,A^{-1}}$ (SI, pre-2019 exact).

Prerequisites. Divergence of a Vector Field · Curl of a Vector Field · The Del Operator and Vector Differential Identities · The Divergence (Gauss) Theorem · The Laplacian Operator. Leads to. Stokes' Theorem · Surface Integrals and Flux.

Ask the tutor: Want to see why the Coulomb gauge is always reachable, or work the Aharonov–Bohm phase for a real superconducting flux quantum? Ask me to walk through it step by step.

Is "solenoidal" the same as "irrotational"?

No — they are independent, in fact complementary, properties. Solenoidal means zero divergence ( $\nabla\cdot\mathbf{B}=0$ ); irrotational means zero curl ( $\nabla\times\mathbf{B}=0$ ). A field can be one without the other: the field around a wire is solenoidal but not irrotational, while a point charge's field is irrotational but not solenoidal. The Helmholtz decomposition writes any reasonable field as a purely solenoidal part plus a purely irrotational part, so the two notions exhaust the ways a field can behave. Solenoidal fields are curls of a vector potential; irrotational fields are gradients of a scalar potential.

If A is not unique, how can it be physical?

Only gauge-invariant quantities built from $\mathbf{A}$ are physical, and those are unique. The field $\mathbf{B}=\nabla\times\mathbf{A}$ is unchanged by $\mathbf{A}\to\mathbf{A}+\nabla\chi$ , and so is any loop integral $\oint\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}$ , since $\oint\nabla\chi\cdot\mathrm{d}\boldsymbol{\ell}=0$ . The Aharonov–Bohm phase depends on exactly this loop integral, which is why it is observable even though $\mathbf{A}$ is not. So the non-uniqueness of $\mathbf{A}$ is a redundancy of description, like choosing the zero of a scalar potential — the physics lives in the gauge-invariant combinations, and gauge freedom is a feature you exploit, not a defect.

Why is the Coulomb gauge the natural choice for solenoidal fields?

Because it makes the potential share the property of the field. Setting $\nabla\cdot\mathbf{A}=0$ renders $\mathbf{A}$ itself solenoidal and kills the $\nabla(\nabla\cdot\mathbf{A})$ term in the curl-of-curl identity, reducing the magnetostatic equation to $\nabla^2\mathbf{A}=-\mu_0\mathbf{J}$ — three decoupled copies of the electrostatic Poisson equation. You can always reach the gauge: given any $\mathbf{A}$ with $\nabla\cdot\mathbf{A}=g$ , solve $\nabla^2\chi=-g$ and the shifted $\mathbf{A}+\nabla\chi$ is divergence-free. Other gauges (Lorenz, axial) suit time-dependent or relativistic problems better.

When can a divergence-free field fail to have a vector potential?

Locally, never — the Poincaré lemma guarantees a potential in any small ball or contractible region. The failure is global and topological. On a domain with a hole or cavity, the local potentials on overlapping patches may refuse to agree into a single smooth global $\mathbf{A}$ ; the obstruction is the second de Rham cohomology of the region. The classic example is a thin flux tube: outside it $\mathbf{B}=0$ , yet $\oint\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}$ around it equals the enclosed flux and cannot be gauged away. So "divergence-free implies curl" is a theorem about contractible domains, and topology is what breaks it.

How does flux end up being a property of the loop, not the surface?

Through Stokes' theorem plus zero divergence. Stokes says $\int_S\mathbf{B}\cdot\mathrm{d}\mathbf{S}=\oint_{\partial S}\mathbf{A}\cdot\mathrm{d}\boldsymbol{\ell}$ , and the right side depends only on the loop $\partial S$ , not the surface you stretch across it. Equivalently: two surfaces sharing a loop bound a closed volume, and $\oint_{\text{closed}}\mathbf{B}\cdot\mathrm{d}\mathbf{S}=\int\nabla\cdot\mathbf{B}\,\mathrm{d}V=0$ since $\mathbf{B}$ is solenoidal, so their fluxes are equal. This is why you can speak of "the flux through a circuit" without naming a surface, and why flux linkage — the basis of inductance and Faraday's law — is well defined.

Does the absence of magnetic monopoles matter here, or is it just convention?

It is the physical reason the magnetic vector potential exists globally and exactly. Maxwell's $\nabla\cdot\mathbf{B}=0$ says there are no magnetic charges, and it is precisely this solenoidal condition that (on ordinary space) lets you write $\mathbf{B}=\nabla\times\mathbf{A}$ everywhere. If monopoles existed, $\nabla\cdot\mathbf{B}=\mu_0\rho_m\ne0$ , and no single-valued $\mathbf{A}$ could represent the field — one would need a singular "Dirac string," which is how Dirac's monopole is built and how it forces charge quantization. So the clean vector potential of magnetostatics is a direct gift of the fact that magnetic charge has never been found.