Aharonov-Bohm effect and holonomy of U(1) connections

Intuition Beginner

A magnetic field can steer a moving charge. The usual rule is that the field has to be present where the charge actually travels. The Aharonov-Bohm effect breaks that rule. It says a charged particle can feel a magnetic field even when the field is zero everywhere along its path.

The trick is to confine the field inside a long thin coil, a solenoid, and let it leak nothing outside. Outside the coil there is no magnetic field at all. Yet if an electron splits, passes on both sides of the coil, and rejoins, the interference pattern shifts. The amount of shift depends only on how much field is trapped inside, not on where the electron went.

What carries the message is the potential, not the field. The potential survives outside the coil even though the field vanishes there. The electron reads the potential by accumulating a phase, like a clock hand turning slightly as it travels.

Visual Beginner

Picture a coil pointing out of the page with field trapped inside it. The shaded region around the coil has no field. Two paths leave a common start, pass on opposite sides of the coil, and meet again. Each path carries a clock-hand phase that turns as the electron moves along it.

Because the two paths together circle the coil once, the difference between their clock hands equals the trapped flux. Loop around the coil twice and you pick up twice the phase. The field hides inside; the loop still counts it.

Worked example Beginner

Suppose a solenoid traps a flux of $6$ units, in the convention where the accumulated phase equals the enclosed flux. An electron beam splits and the two halves pass on opposite sides of the coil. Together the two paths enclose the coil once.

The phase along the upper path minus the phase along the lower path equals the enclosed flux, so the phase difference is $6$ units. That difference is what shifts the interference fringes. Double the trapped flux to $12$ units and the phase difference becomes $12$, shifting the fringes twice as far. Reverse the current so the flux is $-6$ and the fringes shift the opposite way.

What this tells us: the observable shift tracks the trapped flux directly. No field ever touches the electron, yet the trapped amount controls the pattern. The potential outside the coil is the messenger, and the loop around the coil is how the message is read.

Check your understanding Beginner

Formal definition Intermediate+

Let $M$ be a smooth manifold carrying a principal $U(1)$-bundle with a connection whose local potential is $A\in {\Omega }^1(M)$ and whose curvature is the field strength $F=dA$ 03.07.29. Parallel transport of the connection along a path turns the internal phase, and for a closed path the net turn is the holonomy 03.05.11.

Definition (holonomy of a $U(1)$ connection). For a piecewise-smooth loop $\gamma :[0,1]\to M$ based at $x$, the holonomy of the connection around $\gamma$ is $$ \mathrm{hol}(\gamma) ;=; \exp!\Big(i \oint_\gamma A\Big) ;\in; U(1). $$ Because $U(1)$ is abelian, this scalar is independent of the basepoint within a path-component, and it is the parallel-transport phase the connection assigns to one circuit of $\gamma$.

The connection is flat when $F=dA=0$. The sign convention here takes the connection one-form to be $iA$ valued in $\mathfrak{u}(1)\cong i\mathbb{R}$, so that the holonomy is $\exp (i{\oint }_{\gamma }A)$ with a plus sign; the opposite convention flips the sign of every holonomy and leaves all statements below intact.

The Wilson loop of $\gamma$ is the gauge-invariant observable $$ W(\gamma) ;=; \oint_\gamma A \pmod{2\pi}, $$ equivalently $W(\gamma )=\mathrm{hol}(\gamma )$ read as an element of $U(1)$. Under a gauge transformation $A\mapsto A+d\lambda$ with $\lambda \in C^{\infty }(M)$, the integral changes by ${\oint }_{\gamma }d\lambda$, which is an integer multiple of $2\pi$ when $\lambda$ is a genuine $U(1)$-valued phase; hence $W(\gamma )$ and $\mathrm{hol}(\gamma )$ are gauge-invariant even though $A$ is not.

A non-example worth flagging: a flat connection need not have identity holonomy on every loop. Flatness forces $F=0$ pointwise, which kills the holonomy of every contractible loop, but on a non-simply-connected $M$ a loop that cannot be contracted may still carry $\mathrm{hol}(\gamma )\ne 1$. The Aharonov-Bohm configuration is exactly this case ^{[Sternberg Ch. 18 §3]}.

Key theorem with proof Intermediate+

Theorem (flat $U(1)$ holonomy is a character of ${\pi }_1$). Let $A$ be a flat connection ($dA=0$) on a connected manifold $M$. Then the holonomy $\mathrm{hol}:\gamma \mapsto \exp (i{\oint }_{\gamma }A)$ depends only on the homotopy class of $\gamma$ and defines a group homomorphism $$ \chi_A : \pi_1(M, x) \longrightarrow U(1), \qquad \chi_A([\gamma]) = \exp!\Big(i\oint_\gamma A\Big). $$

Proof. First, homotopy invariance. Let ${\gamma }_0$ and ${\gamma }_1$ be loops based at $x$ that are homotopic relative to the basepoint, and let $H:[0,1]\times [0,1]\to M$ be a homotopy with $H(\cdot ,0)={\gamma }_0$, $H(\cdot ,1)={\gamma }_1$, and $H(0,s)=H(1,s)=x$. The image $\Sigma =H([0,1{]}^2)$ is a singular surface whose oriented boundary is ${\gamma }_0$ followed by ${\gamma }_1$ run backwards. Pull $A$ back by $H$ and apply Stokes' theorem on the square: $$ \oint_{\gamma_0} A - \oint_{\gamma_1} A ;=; \int_{\partial [0,1]^2} H^* A ;=; \int_{[0,1]^2} H^(dA) ;=; \int_{[0,1]^2} H^ F ;=; 0, $$ where the side contributions cancel because $H$ is constant on the two vertical edges. Flatness gives $F=0$, so ${\oint }_{{\gamma }_0}A={\oint }_{{\gamma }_1}A$ and the holonomies agree. Hence $\mathrm{hol}$ descends to ${\pi }_1(M,x)$.

Second, the homomorphism property. For loops $\alpha ,\beta$ based at $x$, the concatenation $\alpha \cdot \beta$ traverses $\alpha$ then $\beta$, so $$ \oint_{\alpha \cdot \beta} A = \oint_\alpha A + \oint_\beta A, $$ because line integrals add along concatenated paths. Exponentiating, $$ \chi_A([\alpha][\beta]) = \exp!\Big(i\oint_\alpha A + i\oint_\beta A\Big) = \chi_A([\alpha]),\chi_A([\beta]). $$ The constant loop has $\oint A=0$ and holonomy $1$, and the reversed loop ${\gamma }^{-1}$ gives ${\oint }_{{\gamma }^{-1}}A=-{\oint }_{\gamma }A$, so ${\chi }_A([\gamma {]}^{-1})={\chi }_A([\gamma ]{)}^{-1}$. Thus ${\chi }_A$ is a homomorphism into the abelian group $U(1)$. Since $U(1)$ is abelian, ${\chi }_A$ factors through the abelianization ${\pi }_1(M,x{)}^{\mathrm{ab}}\cong H_1(M;\mathbb{Z})$, which is the statement that flat holonomy is a homology invariant 03.04.21. $\square$

Bridge. This theorem builds toward the topological reading of the Aharonov-Bohm phase, where the holonomy is identified with the period of the closed-not-exact connection one-form against a homology class 03.04.21, and it appears again in the prequantum line bundle 05.11.01, whose flat sections are governed by exactly such characters; it connects to the Ambrose-Singer holonomy theorem 03.05.17, the curved counterpart in which the holonomy algebra is generated by the curvature that here is set to zero, and it underlies the Wilson-loop variables of lattice gauge theory, where holonomies around plaquettes are the fundamental degrees of freedom.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none — Mathlib has the fundamental group, de Rham cohomology, and integration of forms, but not the parallel-transport / holonomy map of a principal connection, so the central statement cannot yet be stated intrinsically.

-- Pseudocode only: holonomy of a principal U(1) connection is not yet
-- available in Mathlib.
axiom Connection (M : Type*) [Manifold M] : Type*       -- local 1-form A
axiom holonomy {M : Type*} [Manifold M] :
    Connection M → Loop M → Circle                       -- exp(i ∮ A)

axiom flat {M : Type*} [Manifold M] : Connection M → Prop  -- dA = 0

-- For a flat connection, holonomy is a homotopy invariant and descends
-- to a character of the fundamental group.
axiom flat_holonomy_homotopy_invariant
    {M : Type*} [Manifold M] (A : Connection M) (h : flat A)
    (γ δ : Loop M) (H : Homotopic γ δ) :
    holonomy A γ = holonomy A δ

axiom flat_holonomy_character
    {M : Type*} [Manifold M] (A : Connection M) (h : flat A) (x : M) :
    FundamentalGroup M x →* Circle

The missing formalization work includes parallel transport along smooth paths, the loop holonomy as the resulting circle element, the Stokes-based homotopy invariance for flat connections, and the identification of the resulting character with a period class in $H^1(M;\mathbb{R})/H^1(M;\mathbb{Z})$.

Advanced results Master

The holonomy of a flat $U(1)$ connection is the geometric content of the period of a closed one-form. On a connected $M$ a flat connection has $dA=0$, so $[A]\in H_{\mathrm{dR}}^1(M;\mathbb{R})$ is a de Rham class, and the assignment $[\gamma ]\mapsto {\oint }_{\gamma }A$ is the pairing of that class with $H_1(M;\mathbb{Z})$ 03.04.21. Two flat connections are gauge-equivalent by a single-valued phase exactly when their periods differ by elements of $2\pi \mathbb{Z}$ on each integral homology cycle, so the moduli space of flat $U(1)$ connections is the torus $$ H^1(M;\mathbb{R}),/,H^1(M;\mathbb{Z}) ;=; \mathrm{Hom}\big(\pi_1(M), U(1)\big), $$ the character variety of the fundamental group. The Aharonov-Bohm phase is one coordinate on this torus, fixed by the enclosed flux modulo ${\Phi }_0$.

The solenoid realizes a specific point of this torus. With confined flux $\Phi$, the exterior region is ${\mathbb{R}}^2\setminus \{0\}$ times the axis, with ${\pi }_1\cong \mathbb{Z}$, and the flat potential $A=\frac{\Phi }{2\pi }d\theta$ has ${\oint }_{{\gamma }_0}A=\Phi$ on the generating loop. The holonomy $e^{i\Phi }$, restored to physical units $\exp (ie\Phi /\hslash )$, is the observed phase. The field strength vanishes on the entire accessible region, so the magnetic field is not the complete physical datum; the connection, recorded by its holonomy, is. This is the precise sense in which the Aharonov-Bohm effect shows the potential to carry information beyond $F$ ^{[Aharonov-Bohm 1959]}.

The nonabelian Wilson loop extends the holonomy to a structure group $G$. For a connection with potential $A$ valued in $\mathfrak{g}$, parallel transport around $\gamma$ is the path-ordered exponential $\mathrm{hol}(\gamma )=\mathcal{P}\exp \left({\oint }_{\gamma }A\right)\in G$, an element well-defined up to conjugation by the basepoint frame. The conjugation ambiguity is removed by the trace, and $W_R(\gamma )={\mathrm{tr}}_R\mathcal{P}\exp \left({\oint }_{\gamma }A\right)$ in a representation $R$ is the gauge-invariant Wilson loop. In the abelian case the path ordering is vacuous and the trace in the fundamental representation returns the scalar $e^{i{\oint }_{\gamma }A}$. Wu and Yang argued that this nonintegrable phase factor, rather than $A$ or $F$ separately, is the complete and minimal description of a gauge field ^{[Wu-Yang 1975]}.

The curved case is governed by Ambrose-Singer 03.05.17: the Lie algebra of the holonomy group is spanned by the curvature evaluated on pairs of vectors, transported back to the basepoint. Setting $F=0$ collapses the holonomy group to the image of ${\pi }_1$, recovering the flat statement above as the degenerate boundary of the Ambrose-Singer mechanism ^{[Ambrose-Singer 1953]}. Flux quantisation in a superconductor is the integrality that arises when the configuration space changes: in a superconducting ring the order parameter is single-valued, forcing the holonomy of the electromagnetic connection acting on the charge-$2e$ condensate to be $1$, so the enclosed flux is locked to integer multiples of $h/2e$. The same holonomy that is a continuous Aharonov-Bohm phase on a non-simply-connected vacuum becomes a quantised invariant once a single-valued condensate is present.

Synthesis. The Aharonov-Bohm effect is the holonomy $\mathrm{hol}(\gamma )=\exp (i{\oint }_{\gamma }A)$ of a flat $U(1)$ connection on a non-simply-connected region, and the flat condition $F=dA=0$ forces this holonomy to factor through ${\pi }_1(M)\to U(1)$ as the period pairing of a closed-not-exact one-form against $H_1(M;\mathbb{Z})$; this builds toward the character-variety description of flat connections as the torus $H^1(M;\mathbb{R})/H^1(M;\mathbb{Z})$ on which the enclosed flux is a coordinate, appears again in the prequantum line bundle 05.11.01 whose flat sections realize exactly these characters, connects to the Ambrose-Singer holonomy theorem 03.05.17 as the $F=0$ degeneration of the curvature-generated holonomy algebra, and underlies the gauge-invariant Wilson loop $W(\gamma )={\oint }_{\gamma }A\mathrm{mod}2\pi$ that the nonabelian path-ordered exponential generalizes and that lattice gauge theory takes as its fundamental variable. The magnetic field where the particle travels is not the complete physical datum, because two configurations with the same vanishing $F$ on the accessible region but different enclosed flux give measurably different interference patterns; the holonomy is the complete gauge-invariant content, and flux quantisation in a superconductor is the same invariant forced to be integral by a single-valued condensate.

Full proof set Master

Proposition (period independence of the capping surface). Let $A$ be a $U(1)$ connection on $M$, flat on an open region $U\subset M$, and let $\gamma \subset U$ be a loop bounding two surfaces $S$ and $S^{\prime }$ in $M$ (not necessarily inside $U$) with $\partial S=\partial S^{\prime }=\gamma$. Then ${\int }_{S}F={\int }_{S^{\prime }}F$ modulo $2\pi$ times the integral charges enclosed, and in particular $\mathrm{hol}(\gamma )=\exp \left(i{\int }_{S}F\right)$ is independent of the chosen capping surface as an element of $U(1)$.

Proof. The two surfaces fit together into a cycle: orient $S$ by $\gamma$ and $S^{\prime }$ oppositely, so that $S\cup (-S^{\prime })$ is a closed singular surface $\Sigma$ with $\partial \Sigma =0$. By Stokes applied to a chain bounding $\Sigma$, or directly because $\Sigma$ is a $2$-cycle, $$ \int_S F - \int_{S'} F ;=; \int_\Sigma F. $$ If $F$ is exact on a neighbourhood of $\Sigma$, the right side is $0$ and the periods agree exactly. In general ${\int }_{\Sigma }F$ is the integral of the curvature over a closed surface, which by the integrality of the first Chern class is $2\pi$ times an integer (the net charge enclosed by $\Sigma$) 03.07.29. Exponentiating, $\exp (i{\int }_{S}F)=\exp (i{\int }_{S^{\prime }}F)$ in $U(1)$, since the discrepancy is $e^{2\pi i\mathbb{Z}}=1$. Hence the holonomy is independent of the capping surface. $\square$

Proposition (the Aharonov-Bohm holonomy of the solenoid). Let the exterior region of an idealized infinite solenoid carrying confined flux $\Phi$ be $M=({\mathbb{R}}^2\setminus \{0\})\times \mathbb{R}$, with flat connection $A=\frac{\Phi }{2\pi }d\theta$. Then for the loop ${\gamma }_0$ winding once around the axis, $\mathrm{hol}({\gamma }_0)=e^{i\Phi }$, and $A$ is gauge-equivalent to the product connection $A=0$ if and only if $\Phi \in 2\pi \mathbb{Z}$.

Proof. On $M$ the angle one-form $d\theta$ is smooth and closed, $d(d\theta )=0$, so $dA=\frac{\Phi }{2\pi }d(d\theta )=0$ and the connection is flat 03.04.21. The angle form has period ${\oint }_{{\gamma }_0}d\theta =2\pi$ on a single circuit of the axis, hence $$ \oint_{\gamma_0} A = \frac{\Phi}{2\pi}\oint_{\gamma_0} d\theta = \frac{\Phi}{2\pi}\cdot 2\pi = \Phi, $$ so $\mathrm{hol}({\gamma }_0)=\exp (i{\oint }_{{\gamma }_0}A)=e^{i\Phi }$. For the gauge statement, suppose $A=d\lambda$ for a single-valued phase $e^{i\lambda }$. Then $\Phi ={\oint }_{{\gamma }_0}d\lambda \in 2\pi \mathbb{Z}$ by single-valuedness. Conversely, if $\Phi =2\pi k$ with $k\in \mathbb{Z}$, then $A=kd\theta =d(k\theta )$ and $e^{ik\theta }$ is single-valued on $M$, so the gauge transformation by $e^{-ik\theta }$ sends $A$ to $0$. Thus gauge-equivalence to $A=0$ holds precisely when $\Phi$ is an integer multiple of $2\pi$, which in physical units is the flux quantum ${\Phi }_0=2\pi \hslash /e$. $\square$

Proposition (fringe displacement in the two-slit experiment). In the double-slit Aharonov-Bohm geometry, the interference intensity as a function of transverse position $y$ is $I(y)\propto 1+\cos (ky+e\Phi /\hslash )$, so the fringe pattern is rigidly displaced by an amount proportional to the enclosed flux $\Phi$, with full-period displacement at each increment $\Delta \Phi ={\Phi }_0=2\pi \hslash /e$.

Proof. Let ${\psi }_1$ and ${\psi }_2$ be the amplitudes for the upper and lower paths, with relative dynamical phase $ky$ from the path-length difference at transverse position $y$. The connection contributes to each path the factor $\exp \left(\frac{ie}{\hslash }{\int }_{\text{path}}A\right)$. Their difference is the holonomy of the loop formed by the two paths, which encircles the solenoid once: $\frac{e}{\hslash }({\int }_{1}A-{\int }_{2}A)=\frac{e}{\hslash }\oint A=\frac{e\Phi }{\hslash }$ by the previous proposition. The total amplitude is ${\psi }_1+{\psi }_2\propto 1+\exp (i(ky+e\Phi /\hslash ))$, and $$ I(y) = |\psi_1 + \psi_2|^2 \propto 1 + \cos!\Big(ky + \frac{e\Phi}{\hbar}\Big). $$ The flux enters only as an additive phase $e\Phi /\hslash$ inside the cosine, so increasing $\Phi$ slides the entire pattern in $y$ without changing its shape. A shift of one full period requires $e\Phi /\hslash$ to increase by $2\pi$, i.e. $\Delta \Phi =2\pi \hslash /e={\Phi }_0$. $\square$

Connections Master

• Electromagnetism as a U(1) Yang-Mills theory 03.07.29 — the holonomy developed here is the parallel-transport phase of the connection whose curvature is the field strength $F=dA$; this unit takes the holonomy of the abelian connection introduced there and isolates the flat case where $F$ vanishes but the holonomy does not.

• Closed and exact forms; the Poincaré lemma 03.04.21 — the Aharonov-Bohm phase is the period ${\oint }_{\gamma }A$ of a closed-not-exact one-form, so the effect is nonzero exactly because $A$ represents a nonzero class in $H^1$ of the non-simply-connected exterior region.

• Horizontal lift and parallel transport 03.05.11 — the holonomy $\exp (i{\oint }_{\gamma }A)$ is the value of parallel transport around a loop, and the flatness condition is the statement that horizontal lifts of homotopic paths have the same endpoint.

• Ambrose-Singer holonomy theorem 03.05.17 — the curved generalization in which the holonomy Lie algebra is generated by the curvature; the flat $U(1)$ case is the degeneration where curvature vanishes and the holonomy reduces to a character of ${\pi }_1$.

• Prequantum line bundle and the integrality condition 05.11.01 — flat $U(1)$ connections and their holonomy characters control the flat sections and the Bohr-Sommerfeld data of geometric quantization, where the same period torus reappears.

Historical & philosophical context Master

Ehrenberg and Siday derived the interference phase shift from the vector potential in 1949 in the context of electron optics, observing that the potential and not the field controls the fringe positions, though the result drew little attention at the time ^{[Ehrenberg-Siday 1949]}. Aharonov and Bohm rediscovered and sharpened the prediction in 1959, framing it as a question about the physical significance of the electromagnetic potentials in quantum theory and proposing the solenoid interference experiment that now bears their names ^{[Aharonov-Bohm 1959]}. Wu and Yang gave the geometric reformulation in 1975, identifying the gauge-invariant content of a gauge field with the nonintegrable phase factor, that is the holonomy, and showing that neither $A$ alone nor $F$ alone is the complete description ^{[Wu-Yang 1975]}. Ambrose and Singer had already proved in 1953 the theorem relating curvature to holonomy that underlies the curved generalization ^{[Ambrose-Singer 1953]}. The experimental confirmation by Tonomura and collaborators in the 1980s, using superconducting toroidal magnets to confine the flux completely, settled the residual objections about field leakage.

Bibliography Master

@article{EhrenbergSiday1949,
  author  = {Ehrenberg, W. and Siday, R. E.},
  title   = {The Refractive Index in Electron Optics and the Principles of Dynamics},
  journal = {Proceedings of the Physical Society B},
  volume  = {62},
  pages   = {8--21},
  year    = {1949}
}

@article{AharonovBohm1959,
  author  = {Aharonov, Y. and Bohm, D.},
  title   = {Significance of Electromagnetic Potentials in the Quantum Theory},
  journal = {Physical Review},
  volume  = {115},
  pages   = {485--491},
  year    = {1959}
}

@article{AmbroseSinger1953,
  author  = {Ambrose, W. and Singer, I. M.},
  title   = {A Theorem on Holonomy},
  journal = {Transactions of the American Mathematical Society},
  volume  = {75},
  pages   = {428--443},
  year    = {1953}
}

@article{WuYang1975,
  author  = {Wu, Tai Tsun and Yang, Chen Ning},
  title   = {Concept of Nonintegrable Phase Factors and Global Formulation of Gauge Fields},
  journal = {Physical Review D},
  volume  = {12},
  pages   = {3845--3857},
  year    = {1975}
}

@article{Dirac1931,
  author  = {Dirac, P. A. M.},
  title   = {Quantised Singularities in the Electromagnetic Field},
  journal = {Proceedings of the Royal Society of London A},
  volume  = {133},
  pages   = {60--72},
  year    = {1931}
}

@article{Tonomura1986,
  author  = {Tonomura, A. and Osakabe, N. and Matsuda, T. and Kawasaki, T. and Endo, J. and Yano, S. and Yamada, H.},
  title   = {Evidence for Aharonov-Bohm Effect with Magnetic Field Completely Shielded from Electron Wave},
  journal = {Physical Review Letters},
  volume  = {56},
  pages   = {792--795},
  year    = {1986}
}

@book{Sternberg2012,
  author    = {Sternberg, Shlomo},
  title     = {Curvature in Mathematics and Physics},
  publisher = {Dover Publications},
  year      = {2012}
}