The Aharonov-Bohm Effect via Path Integrals

Classically, a charged particle only knows about the electric and magnetic fields: the Lorentz force $\mathbf{F}=q(\mathbf{E}+\mathbf{v}\times\mathbf{B})$ vanishes identically in any region where $\mathbf{E}=\mathbf{B}=0$ , so such a region is dynamically inert. Thread a magnetic flux through a thin, impenetrable solenoid and let an electron travel only through the field-free space outside it, and Newton says nothing should happen. Quantum mechanics disagrees, decisively. Switch the flux on and the electron's interference fringes slide sideways — a real, measured shift — even though the particle never touched a field. This is the Aharonov–Bohm effect, and it is the sharpest evidence we have that the electromagnetic potentials, not just the fields, are the physical objects quantum theory is built on. The path integral makes the effect almost inevitable rather than paradoxical. Feynman's prescription weights every trajectory by $e^{iS/\hbar}$ , and the action of a charged particle carries the minimal-coupling term $q\int\mathbf{A}\cdot\mathrm{d}\mathbf{l}$ — a line integral of a one-form. Because $\mathrm{d}t$ cancels, this term is purely geometric: it depends on the route through the potential, not on the speed or timing. It therefore factors out of each path's amplitude as a phase, and when you compare two routes that pass on opposite sides of the flux, their phase difference is fixed entirely by the flux they bracket, $\Delta\varphi=q\Phi/\hbar$ , by Stokes' theorem. The structure is genuinely topological. With the flux line excluded, configuration space is multiply connected, and paths sort into homotopy classes labelled by how many times they wind around the flux. The amplitude is a sum over classes weighted by $e^{in\,q\Phi/\hbar}$ , so every observable is periodic in $\Phi$ with period the flux quantum $\Phi_0=h/q$ . You already hold The Feynman Path Integral Formulation and the geometric-phase machinery of Geometric (Berry) Phase; here that loop logic lands its cleanest instance — flux through the loop, no field on the path — and points straight at the Wilson loops of gauge theory.

Key takeaways

A charged particle picks up the phase $\exp\!\big(\tfrac{iq}{\hbar}\int_\Gamma\mathbf{A}\cdot\mathrm{d}\mathbf{l}\big)$ along a path; in the path integral this is the only place the vector potential enters.
For two paths bracketing a confined flux $\Phi$ the relative phase is $\Delta\varphi=\tfrac{q}{\hbar}\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l}=\tfrac{q\Phi}{\hbar}$ — even though $\mathbf{B}=0$ everywhere the particle goes.
The open-path integral is gauge-dependent; only the closed-loop integral (the enclosed flux) is gauge invariant and observable.
The effect is topological: it depends on the winding number around the excluded flux line, classified by $\pi_1=\mathbb{Z}$ , so observables are periodic in $\Phi$ with period $\Phi_0=h/q$ .
The single-electron flux quantum is $h/e\approx4.136\times10^{-15}\,\mathrm{Wb}$ ; in superconductors the Cooper-pair charge $2e$ halves it to $\Phi_0=h/2e\approx2.068\times10^{-15}\,\mathrm{Wb}$ .
Doubted for decades, the effect was settled by Tonomura's 1986 electron holography with flux fully sealed inside a superconductor.
Key idea: in the path integral the vector potential enters only through the gauge-invariant holonomy $\exp\!\big(\tfrac{iq}{\hbar}\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l}\big)=\exp(iq\Phi/\hbar)$ , an observable, topological phase that survives even where the fields vanish.

The core insight

In the path integral the vector potential enters only through the gauge-invariant phase factor $\exp\!\big(\tfrac{iq}{\hbar}\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l}\big)=\exp(iq\Phi/\hbar)$ accumulated around a closed loop; this holonomy is observable through interference even where the fields vanish, and it is periodic in the enclosed flux with period $\Phi_0=h/q$ .

Where this lives today

The Aharonov–Bohm phase is now a working tool across mesoscopic physics and quantum technology. In micron-scale metal and semiconductor rings, the electrical conductance oscillates periodically with applied flux at period $h/e$ as electron waves interfere between the two arms — Aharonov–Bohm conductance oscillations are a standard probe of phase coherence in nanostructures, and their $h/2e$ -periodic cousins (Al'tshuler–Aronov–Spivak oscillations) diagnose weak localization. The very same single-valuedness that quantizes the AB phase quantizes the flux threading a superconducting ring in units of $\Phi_0=h/2e$ , and the SQUID magnetometer turns that quantization into the most sensitive magnetic-field detector ever built, resolving small fractions of a flux quantum. Topologically, the effect is the abelian shadow of the Wilson loop; lattice gauge theory computes the non-abelian generalization, and the same holonomy logic drives geometric and topological quantum computing, where information is stored in phases that only global paths can change.

The physical picture

Picture a long solenoid seen end-on, its magnetic field trapped entirely inside. Outside, $\mathbf{B}=0$ — but the vector potential $\mathbf{A}$ does not vanish; it circulates around the solenoid like the velocity field of a bathtub vortex, falling off as $1/r$ but never turning off. Classically that circulation is invisible: no field, no force. Quantum mechanically, the phase of the wavefunction is exquisitely sensitive to it. Send an electron beam past both sides of the solenoid and recombine it, and the two partial waves have run through opposite senses of the circulation. Their interference fringes carry the difference — and that difference is fixed by the flux the two routes enclose between them.

Figure

The setup: field trapped inside, potential circulating outside, two routes bracket the flux

In the path integral this is transparent. Every trajectory carries a phase, and the vector potential adds exactly $\tfrac{q}{\hbar}\int\mathbf{A}\cdot\mathrm{d}\mathbf{l}$ to it — nothing else. The free dynamics of the two routes is unchanged; only the topological relationship of each route to the flux line differs. Compare the routes and the free parts largely cancel, leaving the difference of the two line integrals, which Stokes' theorem turns into the enclosed flux. The particle's local experience is field-free the whole way; the global winding is what it records. The deepest reading is topological. Excise the impenetrable flux line and the accessible space is multiply connected: a loop encircling the flux cannot be shrunk to a point. Paths fall into homotopy classes labelled by winding number $n$ , and the amplitude is a sum over classes weighted by $e^{in\theta}$ with $\theta=q\Phi/\hbar$ . Because the label $n$ is an integer that no continuous deformation can change, the effect is robust — and because $e^{in(\theta+2\pi)}=e^{in\theta}$ , every observable repeats each time the flux advances by one quantum $h/q$ .

Figure

Punchline: fringes slide rigidly with Φ, periodic with period Φ₀ = h/q — with B = 0 on the paths

The core equations

The vector potential enters the dynamics through minimal coupling. The Hamiltonian of a charge $q$ in a potential $(\phi,\mathbf{A})$ replaces the kinetic momentum with $\mathbf{p}-q\mathbf{A}$ , and this single substitution is the entire origin of the effect.

H=\frac{1}{2m}\left(\mathbf{p}-q\mathbf{A}\right)^2+q\phi \qquad (1)

where: $\mathbf{p}=-i\hbar\nabla$ — canonical momentum; $q$ — charge (an electron has $q=-e$ ); $\mathbf{A}$ — vector potential, $\mathbf{B}=\nabla\times\mathbf{A}$ ; $\phi$ — scalar potential; $m$ — mass. Reading it physically: it is not $\mathbf{B}$ but $\mathbf{A}$ that appears in the operator, so the wavefunction responds to $\mathbf{A}$ directly — the field only enters through its curl, which can vanish where $\mathbf{A}$ does not. The corresponding classical action carries a velocity-coupling term. Because $\dot{\mathbf{x}}\,\mathrm{d}t=\mathrm{d}\mathbf{l}$ , that term is a geometric line integral along the trajectory — the key structural fact.

S[\mathbf{x}(t)]=\int\!\Big(\tfrac12 m\dot{\mathbf{x}}^2+q\,\dot{\mathbf{x}}\cdot\mathbf{A}-q\phi\Big)\mathrm{d}t=S_0+q\!\int_\Gamma\mathbf{A}\cdot\mathrm{d}\mathbf{l} \qquad (2)

where: $S_0$ — the free-plus-scalar action; $\int_\Gamma\mathbf{A}\cdot\mathrm{d}\mathbf{l}$ — the magnetic line integral, independent of speed and timing. Reading it physically: the magnetic coupling is purely geometric. Reparametrize the path in time however you like and this term is unchanged; only the shape of the route through $\mathbf{A}$ matters. Feeding this into The Feynman Path Integral Formulation's sum over paths, the vector potential factors out of each amplitude as a pure phase.

K(\mathbf{x}_f,\mathbf{x}_i)=\int\mathcal{D}[\mathbf{x}]\;e^{iS_0/\hbar}\,\exp\!\Big(\frac{iq}{\hbar}\int_\Gamma\mathbf{A}\cdot\mathrm{d}\mathbf{l}\Big) \qquad (3)

where: $K$ — amplitude from $\mathbf{x}_i$ to $\mathbf{x}_f$ ; $\mathcal{D}[\mathbf{x}]$ — the path measure. Reading it physically: the AB phase is not an approximation or a perturbation — it is an exact multiplicative factor on every path. This is why the path integral makes the effect look inevitable: there is simply nowhere else for $\mathbf{A}$ to hide. Comparing two routes on opposite sides of the flux, the difference of their phases collapses by Stokes' theorem to the enclosed flux. This is the master result.

\Delta\varphi=\frac{q}{\hbar}\Big(\int_{\Gamma_1}-\int_{\Gamma_2}\Big)\mathbf{A}\cdot\mathrm{d}\mathbf{l}=\frac{q}{\hbar}\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l}=\frac{q}{\hbar}\int_S(\nabla\times\mathbf{A})\cdot\mathrm{d}\mathbf{S}=\frac{q\Phi}{\hbar} \qquad (4)

where: $\Gamma_1,\Gamma_2$ — the two routes; $\Phi=\int_S\mathbf{B}\cdot\mathrm{d}\mathbf{S}$ — flux through the surface they bound. Reading it physically: the observable phase difference sees only $\Phi$ . Two paths in the same homotopy class differ by a contractible loop enclosing zero flux, so the class-integral is path-independent — the effect is topological, not geometric in the fine sense. Two consequences pin the physics. Gauge freedom shifts the open-path integral but not the loop, and the loop's $2\pi$ -periodicity sets the flux quantum.

\mathbf{A}\to\mathbf{A}+\nabla\Lambda\;\Rightarrow\;\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l}\;\text{unchanged};\qquad \Phi_0^{(e)}=\frac{h}{e}\approx4.136\times10^{-15}\,\mathrm{Wb},\quad \Phi_0=\frac{h}{2e}\approx2.068\times10^{-15}\,\mathrm{Wb} \qquad (5)

where: $\Lambda(\mathbf{x})$ — a single-valued gauge function; $h=6.626\times10^{-34}\,\mathrm{J\,s}$ ; $e=1.602\times10^{-19}\,\mathrm{C}$ . Reading it physically: the closed-loop holonomy $U(\Gamma)=\exp(\tfrac{iq}{\hbar}\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l})$ is the gauge-invariant observable — the abelian Wilson loop — and observables repeat every time $\Phi$ advances by $h/q$ .

Derivation AThe interference shift is the enclosed flux by Stokes' theorem

Our goal is equation (4): to show that in a two-route (double-slit + solenoid) geometry the fringe pattern shifts rigidly by $\Delta\varphi=q\Phi/\hbar$ , set by the enclosed flux and nothing else, even though $\mathbf{B}=0$ along both routes. The strategy is to attach the factorized AB phase from equation (3) to each route's free amplitude, subtract, and convert the resulting closed-loop line integral into a surface integral of $\mathbf{B}$ with Stokes' theorem. The subtlety to watch is that this is legitimate only because the accessible region is field-free, which makes the phase constant within each homotopy class.

Step 1. Write the amplitude on each route as its free amplitude times its AB phase: $\psi_j=\psi_j^{(0)}\exp\!\big(\tfrac{iq}{\hbar}\int_{\Gamma_j}\mathbf{A}\cdot\mathrm{d}\mathbf{l}\big)$ for $j=1,2$ . Why: equation (3) shows the vector potential multiplies each path's amplitude by exactly this factor; the free physics lives entirely in $\psi_j^{(0)}$ .

Step 2. Within one route family the line integral is the same for every member. Two paths in class $j$ differ by a contractible loop enclosing zero flux, so $\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l}=\int_S\mathbf{B}\cdot\mathrm{d}\mathbf{S}=0$ over that loop. Why: $\mathbf{B}=0$ throughout the accessible region, so any loop that does not encircle the flux line encloses no flux; this is what lets us collapse each family to a single phase.

Step 3. Form the detected intensity from the coherent sum: $P=|\psi_1+\psi_2|^2=|\psi_1^{(0)}|^2+|\psi_2^{(0)}|^2+2|\psi_1^{(0)}\psi_2^{(0)}|\cos\!\big(\delta_0+\Delta\varphi\big)$ , with $\Delta\varphi=\tfrac{q}{\hbar}\big(\int_{\Gamma_1}-\int_{\Gamma_2}\big)\mathbf{A}\cdot\mathrm{d}\mathbf{l}$ . Why: only the relative phase survives in the interference cross-term; $\delta_0$ is the geometric phase difference already present at zero flux.

Step 4. Close the difference into a loop and apply Stokes: $\int_{\Gamma_1}-\int_{\Gamma_2}=\oint_{\Gamma_1-\Gamma_2}\mathbf{A}\cdot\mathrm{d}\mathbf{l}=\int_S(\nabla\times\mathbf{A})\cdot\mathrm{d}\mathbf{S}=\Phi$ , so $\Delta\varphi=q\Phi/\hbar$ and $P$ shifts rigidly with $\Phi$ ; one quantum $\Phi=h/q$ restores the original pattern. Why: running $\Gamma_2$ backwards joins the two routes into a closed loop encircling the solenoid, whose only enclosed flux is the trapped $\Phi$ ; the fringe cosine's argument therefore advances linearly with $\Phi$ . Verified: `verify_aq-qm-aharonov-bohm-path-integral.py` confirms $P|_{\Phi=0}$ is the ordinary two-slit pattern, that $P$ is periodic in $\Phi$ with period $h/q$ , and that $\Delta\varphi(\Phi=\tfrac12 h/e)=\pi$ (bright↔dark), matching the Stokes reduction.

Go deeper · Derivation B · 2 minTopological origin: the sum over winding classes and the flux period

Derivation BTopological origin: the sum over winding classes and the flux period

Our goal is to expose the AB phase as a one-dimensional unitary representation of the fundamental group of the field-free region, $\pi_1=\mathbb{Z}$ , and to derive the exact flux period $\Phi_0=h/q$ from it. The strategy is to split the path integral on the multiply-connected space into homotopy classes labelled by winding number $n$ , attach the winding-proportional phase to each class, and then demand that the physics be single-valued in the phase $\theta=q\Phi/\hbar$ . Watch that the class label $n$ is a discrete topological invariant, not a continuous geometric quantity — that is exactly what makes the result robust.

Step 1. Classify paths by homotopy: with the flux line removed, $\pi_1(\mathbb{R}^3\setminus\text{line})\cong\mathbb{Z}$ , so every path from $\mathbf{x}_i$ to $\mathbf{x}_f$ carries an integer winding number $n$ counting signed encirclements of the flux. Why: a loop around the excluded line cannot be contracted; the number of times a path winds is a deformation invariant, the generator of the fundamental group.

Step 2. Split the kernel into class sums: $K=\sum_{n\in\mathbb{Z}}e^{in\theta}K_n$ with $K_n=\int_{\Gamma\in\text{class }n}\mathcal{D}[\mathbf{x}]\,e^{iS_0/\hbar}$ and $\theta=q\Phi/\hbar$ . Why: a path winding $n$ times encircles the flux $n$ times, so by equation (4) its AB phase is $n\theta$ (up to a common reference); the magnetic phase is constant on each class and pulls out of $K_n$ .

Step 3. Recognize the weights as a representation of $\pi_1$ : the map $n\mapsto e^{in\theta}$ is a homomorphism from $\mathbb{Z}$ into the unit circle $U(1)$ , i.e. a one-dimensional unitary representation of the fundamental group. Why: consistency of the path integral on a multiply-connected space (Laidlaw–DeWitt / Schulman) requires the class weights to form a unitary $1$ -D representation of $\pi_1$ ; for $\pi_1=\mathbb{Z}$ these are exactly the phases $e^{in\theta}$ .

Step 4. Impose single-valuedness: since $e^{in(\theta+2\pi)}=e^{in\theta}$ for all integer $n$ , physics is $2\pi$ -periodic in $\theta$ , hence periodic in flux with period $\Phi=2\pi\hbar/q=h/q$ . This is the flux quantum $\Phi_0=h/q$ . Why: shifting $\theta$ by $2\pi$ leaves every class weight untouched, so no observable can distinguish $\Phi$ from $\Phi+h/q$ ; the period is fixed by the charge alone, giving $h/e$ for electrons and $h/2e$ for Cooper pairs. Verified: the verifier confirms $e^{in\theta}$ is unchanged under $\theta\to\theta+2\pi$ for $n=-3\ldots3$ , that the implied period is $h/q$ , and reproduces $\Phi_0=h/e$ and $h/2e$ numerically.

Limiting cases

case	result	meaning
zero flux, $\Phi\to0$	$\Delta\varphi\to0$	ordinary two-slit pattern; no shift
integer quantum, $\Phi=n\,h/e$	$\Delta\varphi=2\pi n$ , $e^{i\Delta\varphi}=1$	pattern returns to its zero-flux position (periodicity)
half quantum, $\Phi=h/2e$ (electron)	$\Delta\varphi=\pi$	maximal shift: bright fringes move onto former dark fringes
Cooper pairs, $q=2e$	period $\Phi_0=h/2e\approx2.07\times10^{-15}$ Wb	superconducting flux quantum, half the single-electron period
classical limit, $\hbar\to0$	$\Delta\varphi\to\infty$	phase spins infinitely fast; fringes wash out, no observable effect — classical insensitivity restored

Common trapThe vector potential along one path is not observable — only the closed-loop flux is

It is tempting to say "the electron measures $\int_\Gamma\mathbf{A}\cdot\mathrm{d}\mathbf{l}$ along its path." It does not, and cannot: under a gauge transformation $\mathbf{A}\to\mathbf{A}+\nabla\Lambda$ that open-path integral changes by $\Lambda(\mathbf{x}_f)-\Lambda(\mathbf{x}_i)$ , a pure wavefunction phase you can redefine away. Only the closed-loop integral $\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l}=\Phi$ is gauge invariant, because $\oint\nabla\Lambda\cdot\mathrm{d}\mathbf{l}=0$ for single-valued $\Lambda$ . The observable is the enclosed flux, i.e. the relative phase between two routes — never one route's line integral. A related trap: the single-electron flux quantum is $h/e$ ; the famous $h/2e$ is a superconducting value, halved because the carriers are charge- $2e$ Cooper pairs, not electrons.

Worked examples

Example 1 — one flux quantum restores the pattern. An electron interferometer encircles a solenoid. By how much flux must you change $\Phi$ to slide the fringes back to their starting position, and what is that flux numerically?

Step 1. The pattern returns when $\Delta\varphi=q\Phi/\hbar$ advances by $2\pi$ , i.e. when $\Phi$ increases by one flux quantum $\Phi_0^{(e)}=h/|q|=h/e$ .

Step 2. Insert constants: $h=6.62607\times10^{-34}\,\mathrm{J\,s}$ , $e=1.602177\times10^{-19}\,\mathrm{C}$ , so $\Phi_0^{(e)}=h/e=6.62607\times10^{-34}/1.602177\times10^{-19}$ .

Step 3. Evaluate: $\Phi_0^{(e)}=\mathbf{4.136\times10^{-15}\ Wb}$ (equivalently $4.136\,\mathrm{mWb}$ per $10^{12}$ , or $4.136\times10^{-15}$ T·m $^2$ ). At exactly this flux $\Delta\varphi=2\pi$ and $e^{i\Delta\varphi}=1$ , so the interference pattern is identical to the zero-flux case. One tesla threaded through a $1\,\mu\mathrm{m}^2$ core is $10^{-12}$ Wb $\approx242$ quanta — the fringes cycle 242 times.

Example 2 — the fringe shift for a real geometry. A $100\,\mathrm{eV}$ electron beam ( $\lambda_{\mathrm{dB}}\approx0.123\,\mathrm{nm}$ ) passes a solenoid; the two arms are separated by $d=1\,\mu\mathrm{m}$ and interfere on a screen $L=0.5\,\mathrm{m}$ away, with trapped flux $\Phi=\tfrac12\Phi_0^{(e)}=2.07\times10^{-15}\,\mathrm{Wb}$ . Find the lateral fringe shift $\Delta x$ .

Step 1. The zero-flux fringe spacing is $w=\lambda L/d$ . With $\lambda=1.23\times10^{-10}\,\mathrm{m}$ , $L=0.5\,\mathrm{m}$ , $d=10^{-6}\,\mathrm{m}$ : $w=(1.23\times10^{-10}\cdot0.5)/10^{-6}=6.15\times10^{-5}\,\mathrm{m}=61.5\,\mu\mathrm{m}$ .

Step 2. The AB phase shifts the pattern by a fraction $\Delta\varphi/2\pi=(q\Phi/\hbar)/2\pi=\Phi/\Phi_0^{(e)}$ of a full fringe. Here $\Phi/\Phi_0^{(e)}=\tfrac12$ , so the shift is half a fringe spacing.

Step 3. Therefore $\Delta x=(\Phi/\Phi_0^{(e)})\,w=\tfrac12\times61.5\,\mu\mathrm{m}=\mathbf{30.75\ \mu m}$ . A half-quantum of flux moves the whole fringe pattern sideways by 31 μm — a shift comfortably resolved in an electron microscope, and with $\mathbf{B}=0$ on both electron arms the entire displacement is the AB phase.

Origins · 1 minaharonov-bohm effect and how this was discovered

aharonov-bohm effect

predicted 1959, confirmed 1986 · Yakir Aharonov and David Bohm showed that electromagnetic potentials have observable quantum consequences where the fields vanish; the topological interference phase now anchors gauge theory

The effect was predicted by Yakir Aharonov and David Bohm in 1959 (Phys. Rev. 115, 485), building on a little-noticed 1949 analysis by Werner Ehrenberg and Raymond Siday. It provoked a decade-long controversy over whether a field-free vector potential could be "real," with sceptics arguing that unavoidable field leakage explained early experiments. The debate was settled by Akira Tonomura and collaborators in 1986 (Phys. Rev. Lett. 56, 792), who sealed a toroidal magnet inside superconducting niobium — the Meissner effect guaranteed zero external field — and measured the predicted phase shift by electron holography. The same holonomy structure, promoted from $U(1)$ to non-abelian groups, became the Wilson loop at the centre of modern gauge theory.

History

1949

— Ehrenberg and Siday note the potential-dependence of electron interference, an unheeded precursor

1959

— Aharonov and Bohm predict the effect and its topological, gauge-invariant character (Phys. Rev. 115, 485)

1960

— Chambers observes the fringe shift using an electron biprism and a magnetic whisker

1961

— Fowler, Marton and collaborators report further interferometric confirmations

1986

— Tonomura seals the flux inside a superconductor and confirms the effect by electron holography (Phys. Rev. Lett. 56, 792)

Going further

This article grafts the minimal-coupling action onto the sum over paths of The Feynman Path Integral Formulation, where the vector potential factors out as a pure phase, and it realizes the loop logic developed abstractly in Geometric (Berry) Phase — which reads the Aharonov–Bohm effect as a geometric phase (flux through the loop, no field on the path) and gives the spin-½ solid-angle result $\gamma=-\Omega/2$ . Here the loop's holonomy is a topological $U(1)$ phase set by the enclosed flux. The abelian holonomy $\exp(\tfrac{iq}{\hbar}\oint\mathbf{A}\cdot\mathrm{d}\mathbf{l})$ is the Wilson loop; the non-abelian gauge theory that follows promotes it to a path-ordered exponential and makes it the central gauge-invariant observable, while requiring the AB phase around a monopole's Dirac string to be single-valued yields the Dirac charge-quantization condition.

Provenance. Base article QUANT.RQM.AB.1102 (AllQFT "quantum-mechanics" track, Griffiths §10.2 / Sakurai §2.7 / Shankar Ch. 21). Upgraded to v3 canon: the Stokes reduction of the fringe shift, the winding-class decomposition and its $h/q$ period, the gauge-invariance of the loop, and the flux-quantum constants ( $h/e=4.136\times10^{-15}$ Wb, $h/2e=2.068\times10^{-15}$ Wb from CODATA) are all recomputed in `verify_aq-qm-aharonov-bohm-path-integral.py`. The base's physics was sound; the upgrade adds the topological derivation, the numeric worked geometry, and the history/scope framing (the effect's decade of controversy, settled by Tonomura 1986).

Prerequisites. The Feynman Path Integral Formulation · Minimal Coupling and Gauge Invariance · Schrödinger Equation in Electromagnetic Fields. Leads to. Geometric (Berry) Phase · Wilson loops and non-abelian gauge theory · Dirac monopole and charge quantization.

If B = 0 everywhere the electron goes, what physically causes the shift?

The vector potential $\mathbf{A}$ , which does not vanish outside the solenoid even though $\mathbf{B}=\nabla\times\mathbf{A}$ does. In quantum mechanics the dynamics is governed by $\mathbf{A}$ through minimal coupling, and the wavefunction accumulates $\tfrac{q}{\hbar}\int\mathbf{A}\cdot\mathrm{d}\mathbf{l}$ along its path. The two routes bracket the flux, and by Stokes' theorem their phase difference equals $q\Phi/\hbar$ . Nothing acts on the electron locally — no force, no field — yet the global topological relationship of its two routes to the trapped flux is physical and measurable. This is precisely why the effect forced physicists to accept the potentials, not just the fields, as carriers of physical content.

Is the Aharonov–Bohm effect action at a distance?

No. There is no signal or force reaching out from the field region. The effect is entirely local in the sense that the electron obeys the local Schrödinger equation with the local $\mathbf{A}$ everywhere along its path; it is non-local only in that the observable (the loop integral) is a property of a whole closed curve, not of any point. The modern resolution is that $\mathbf{A}$ (more precisely the $U(1)$ connection) is the genuinely local object, and the flux is its holonomy. Calling it action at a distance mistakes a topological, whole-loop quantity for a force propagating through empty space.

Why does the timing or speed of the electron not matter?

Because the magnetic contribution to the action, $q\int\dot{\mathbf{x}}\cdot\mathbf{A}\,\mathrm{d}t=q\int_\Gamma\mathbf{A}\cdot\mathrm{d}\mathbf{l}$ , is a line integral of a one-form: the $\mathrm{d}t$ cancels between $\dot{\mathbf{x}}$ and $\mathrm{d}t$ , leaving a purely geometric integral over the route. Reparametrize the trajectory in time however you like — speed it up, slow it down, pause — and this term is unchanged. This is the same reason the AB phase is topological: it responds to the shape and winding of the path through $\mathbf{A}$ , not to the kinematics. The free part of the action does depend on timing, but that cancels in the relative phase between the two routes.

How does this relate to Berry's geometric phase?

They are close cousins: both are loop holonomies of a connection, observable through interference. In Geometric (Berry) Phase the connection lives in parameter space and the phase is minus half the solid angle subtended, $\gamma=-\Omega/2$ for spin ½; here the connection is the electromagnetic $\mathbf{A}$ in real space and the phase is the enclosed flux, $q\Phi/\hbar$ . The Berry curvature there plays the role that $\mathbf{B}$ plays here — a curvature concentrated in a region the system does not enter. The AB effect is arguably the earliest and cleanest instance of a geometric/topological phase, predating Berry's general formulation by two decades.

Why was the effect controversial for so long?

Sceptics argued that no real experiment could perfectly confine the field, so the measured shifts might be ordinary responses to stray leakage rather than a genuine field-free potential effect. The objection was serious: early solenoids and whiskers did leak. The decisive answer came from Tonomura's 1986 experiment, which enclosed a toroidal magnet in superconducting niobium; the Meissner effect expels all field from the superconductor, guaranteeing $\mathbf{B}=0$ in the electron's region to extraordinary precision, and the observed phase still matched $q\Phi/\hbar$ . This closed the loophole and established the effect as physical beyond dispute.

Does the sum over winding classes double-count anything?

No, because the classes are disjoint by construction: every path belongs to exactly one homotopy class, labelled by its integer winding number around the flux line, and no continuous deformation moves a path between classes. The kernel $K=\sum_n e^{in\theta}K_n$ sums each class once, with the magnetic phase $e^{in\theta}$ pulled out because it is constant on the class. This is the standard Laidlaw–DeWitt/Schulman construction for path integrals on multiply-connected spaces; the same machinery classifies the statistics of identical particles (bosons vs. fermions in 3D, anyons in 2D), where the relevant group is the braid or permutation group instead of $\mathbb{Z}$ .

What sets the flux quantum, and why is it exactly h/q?

Single-valuedness of the physics under $\theta\to\theta+2\pi$ . The class weights $e^{in\theta}$ with integer $n$ are unchanged when $\theta=q\Phi/\hbar$ advances by $2\pi$ , so no observable can distinguish $\Phi$ from $\Phi+2\pi\hbar/q=\Phi+h/q$ . The period is therefore $\Phi_0=h/q$ , fixed entirely by the charge of the interfering particle: $h/e$ for a single electron, $h/2e$ for a Cooper pair. In a superconducting ring the same requirement — that the macroscopic pair wavefunction be single-valued around the ring — quantizes the trapped flux in these units, which is why measuring $h/2e$ was direct evidence for pairing.