Hannay's Angle

Every visitor to the Panthéon in Paris watches the same slow miracle: Foucault's pendulum swings all day, and its swing plane creeps around the hall, clocking Earth's rotation. Ask why and you will usually hear "the Coriolis force" — true, and computed in Coriolis Force. But there is a second, stranger answer. After exactly one day, the support of the pendulum has been carried around a closed loop in space and returned to its starting state. The pendulum's amplitude comes back too — the action is adiabatically invariant. Its phase does not. The swing plane fails to return by an angle that has nothing to do with how fast the loop was traversed and everything to do with the loop's geometry: the solid angle it encloses on the sphere of directions. That failure-to-return is Hannay's angle, discovered by John Hannay in 1985 — startlingly late for a theorem of classical mechanics, and only after Michael Berry had found its quantum twin. When an integrable system is carried adiabatically around a closed circuit in parameter space, the actions are protected by the adiabatic theorem, but the conjugate angles pick up a shift beyond the dynamical $\int\omega\,\mathrm{d}t$ : a holonomy, the exact analogue of a vector coming back rotated after parallel transport around a curved surface. This article builds the connection-and-curvature machinery on top of Action-Angle Variables and the Adiabatic Theorem for Slowly Varying Hamiltonians, derives the solid-angle law for the rotated oscillator, and lands the correspondence $\theta_{\rm Hannay} = -\partial\gamma_{\rm Berry}/\partial n$ — the cleanest bridge in this chapter between classical mechanics and the geometric-phase revolution that runs through modern quantum physics.

Key takeaways

Carry an integrable system slowly around a closed loop $C$ in parameter space: the actions $I_j$ return (adiabatic invariants), but each angle variable acquires an extra shift beyond the dynamical phase $\int\omega_j\,\mathrm{d}t$ .
That extra shift — Hannay's angle — is purely geometric: it depends on the loop traced in parameter space, not on the traversal time or rate.
It is a holonomy: the loop integral of a connection one-form (the angle-averaged response $\partial_I\langle p\,\partial_\lambda q\rangle$ ), equal by Stokes' theorem to the flux of a curvature two-form through any surface spanning $C$ .
Flagship example: a pendulum or bead-on-hoop whose axis is carried around a cone picks up $\Delta\theta = -2\pi(1-\cos\theta_c)$ — minus the enclosed solid angle. The Foucault pendulum's daily deficit is exactly this law at work.
The quantum connection is exact in the semiclassical limit: $\theta_{\rm Hannay}(I) = -\partial\gamma_{\rm Berry}/\partial n$ at $I = (n+\tfrac12)\hbar$ — Berry's phase is the quantum shadow of Hannay's angle, and vice versa.
No integrability, no angle variables, no Hannay angle: for chaotic systems the geometric residue of slow driving reappears as geometric magnetism and deterministic friction (Berry–Robbins).
Key idea: adiabatic transport around a loop in parameter space conserves the action but leaves the angle with a path-dependent geometric anholonomy — Hannay's angle — the loop integral of the angle-averaged one-form, i.e. the flux of the Hannay curvature.

The core insight

Adiabatic transport around a loop in parameter space conserves the action but leaves the angle variable with a path-dependent geometric anholonomy — Hannay's angle — equal to the loop integral of the angle-averaged one-form $\langle\partial_\lambda\theta\rangle\,\mathrm{d}\lambda$ , i.e. the flux of the Hannay curvature.

Where this lives today

Geometric phases went from curiosity to organizing principle in a single generation, and Hannay's angle is their classical anchor. In condensed matter, the Berry curvature of Bloch bands — the quantum object whose $\partial/\partial n$ shadow is the Hannay curvature — controls the anomalous Hall effect, orbital magnetization, and the topological invariants of insulators and semimetals; when experimentalists reconstruct Berry curvature from wave-packet dynamics in optical lattices, the wave-packet equations they use are precisely classical transport with a Hannay-type connection term. Measured directly, the classical angle appears in swung and rotated mechanical resonators, in the polarization transport of light along coiled optical fibers (Tomita–Chiao, 1986 — a solid-angle law identical to the rotated-oscillator result), and in every Foucault pendulum on Earth. The idea also escaped Hamiltonian mechanics. Shapere and Wilczek recast locomotion at low Reynolds number as holonomy: a swimming microorganism cycles through shapes (a loop in shape space) and translates by the connection's holonomy — "swimming as a gauge theory", with the falling cat's self-righting as the rotational cousin. Robotics uses the same geometric-control language for snake robots and satellite reorientation. And where integrability fails and no angle variable exists, the geometric residue of slow driving survives in mutated form: Berry and Robbins showed the first-order reaction of a chaotic fast system back on its slow drivers contains geometric magnetism — a Lorentz-like force built from a curvature two-form — plus deterministic friction, the classical skeleton of dissipation. The chapter's arc — action frozen, angle remembering geometry — turns out to be the seed of half of modern topological physics.

The physical picture

Start with the picture that makes "geometric" literal. Stand at the north pole holding an arrow flat against the ground, walk south to the equator, walk a quarter of the way around it, and walk back to the pole — never once twisting the arrow relative to your path. The arrow comes back rotated by 90°: the solid angle of the spherical triangle you enclosed. Nothing rotated it; the curvature of the sphere did. This failure of transported directions to return — anholonomy — is the entire conceptual content of Hannay's angle. The transported object is the oscillator's phase; the "surface" is parameter space; the curvature is a property of how the family of invariant tori twists as the parameters change.

Figure

Parallel transport anholonomy: an arrow carried around a closed loop on a sphere returns rotated by the enclosed solid angle

Why does the geometry land in the angle and nowhere else? Because the adiabatic theorem has already spoken for the action. The action is the area of the torus cross-section, frozen to exponential accuracy by the averaging mechanism of the Adiabatic Theorem for Slowly Varying Hamiltonians; after a closed loop it returns to its initial value, full stop. The angle enjoys no such protection — it is a phase, and phases are exactly the objects that can store holonomy. So slow cyclic driving deposits all of its geometric memory into the one variable free to receive it. The split of the total angle advance is the operational heart. After one loop of duration $\tau$ , the angle has advanced by an enormous dynamical amount — the frequency integrated over the loop, growing without bound as $\tau\to\infty$ — plus a finite residue that stops depending on $\tau$ altogether once the motion is adiabatic. Slow the loop down by a factor of ten: the dynamical phase grows tenfold; the residue does not move. That $\tau$ -independent residue is Hannay's angle, and its indifference to timing is what "geometric" means operationally.

Figure

The angle advance splits into a rate-dependent dynamical part and a loop-dependent geometric part that survives the adiabatic limit

The connection and its curvature

The bookkeeping starts by splitting the total advance of each angle after one closed circuit of duration $T$ . The instantaneous frequencies contribute the obvious part; everything else is the object of study:

\theta_j(T)=\int_0^T\omega_j\big(I,\lambda(t)\big)\,\mathrm{d}t\;+\;\Delta\theta_j^{\mathrm{Hannay}} \qquad (1)

where: $\omega_j = \partial H/\partial I_j$ — the instantaneous frequency at the frozen parameter value, exactly as computed in Frequencies from Action-Angle Variables; $I$ — the action vector, unchanged around the loop; $\Delta\theta_j^{\mathrm{Hannay}}$ — the geometric remainder. The first term depends on the schedule $\lambda(t)$ ; the second, we will see, only on the oriented curve $C$ that the schedule traces. The remainder has the structure of a line integral of a connection — a rule for transporting the angle as the parameters move. Averaging the parameter-driven part of $\dot\theta$ over the fast motion (the same averaging that protects the action) yields

\Delta\theta_j^{\mathrm{Hannay}}=\oint_C \mathcal{A}_j^{(\mu)}(I,\lambda)\,\mathrm{d}\lambda_\mu,\qquad \mathcal{A}_j^{(\mu)}=-\frac{\partial}{\partial I_j}\Big\langle\frac{\partial G}{\partial\lambda_\mu}\Big\rangle_\theta \qquad (2)

where: $\mathcal{A}_j^{(\mu)}$ — the Hannay connection, a one-form on parameter space; $G(\theta,I;\lambda)$ — the generating function of the map to instantaneous action-angle variables; $\langle\cdot\rangle_\theta$ — the average over the torus angles. Read the structure: the connection is an angle-averaged, action-differentiated response of the system to parameter changes. It inherits a gauge freedom — the choice of where $\theta = 0$ sits on each torus — under which $\mathcal{A}$ shifts by a total $\lambda$ -derivative, leaving the loop integral untouched. Whenever a loop integral of a gauge-dependent one-form is secretly gauge-invariant, Stokes' theorem is telling you there is a curvature underneath:

\Delta\theta_j^{\mathrm{Hannay}}=\iint_S \mathcal{F}_j^{(\mu\nu)}\,\mathrm{d}\lambda_\mu\wedge\mathrm{d}\lambda_\nu,\qquad \mathcal{F}_j^{(\mu\nu)}=\partial_{\lambda_\mu}\mathcal{A}_j^{(\nu)}-\partial_{\lambda_\nu}\mathcal{A}_j^{(\mu)} \qquad (3)

where: $\mathcal{F}_j$ — the Hannay curvature two-form, gauge-invariant and locally measurable; $S$ — any surface spanning $C$ . This is the classical twin of the Berry curvature, and it makes two facts obvious at a glance: a contractible loop enclosing zero flux gives zero angle, and deforming the loop without sweeping curvature changes nothing. The quantum correspondence closes the circle of this chapter. Semiclassically, the $n$ -th eigenstate lives on the torus with $I = (n+\tfrac12)\hbar$ (the quantization rule of Bohr-Sommerfeld Quantization), and Berry's phase for that state relates to Hannay's angle by a derivative with respect to the quantum number:

\theta_{\mathrm{Hannay}}(I)=-\frac{\partial\gamma_{\mathrm{Berry}}(n)}{\partial n}\bigg|_{I=(n+1/2)\hbar} \qquad (4)

where: $\gamma_{\mathrm{Berry}}(n)$ — the geometric phase of the $n$ -th adiabatic eigenstate. The logic of the correspondence is worth internalizing: quantum phases are $S/\hbar$ -like objects, classical angles are their $\partial/\partial I$ derivatives — exactly the relation between the phase of a WKB wavefunction and the angle variable, developed in WKB and the Eikonal Connection. One more structural equation pins down where the geometry enters the dynamics. The moving transformation to instantaneous action-angle variables shifts the Hamiltonian by the time derivative of its generating function — the identical mechanism that opened the proof of the adiabatic theorem:

K(\theta,I,t)=H(I;\lambda)+\dot\lambda_\mu\,\frac{\partial G}{\partial\lambda_\mu} \qquad (5)

where: $K$ — the Hamiltonian in the co-moving action-angle frame; the second term — the sole carrier of both the action's (vanishing, after averaging) drive and the angle's (surviving) geometric drive. The action and the angle read the same small term through different operators: $-\partial/\partial\theta$ annihilates its average, $+\partial/\partial I$ does not. That asymmetry is the whole story.

Derivation AThe master formula: Hannay's angle as an angle-averaged loop integral

We derive equation (2): the geometric part of the angle advance is the closed-loop integral of the angle-averaged, action-differentiated response $\partial_I\langle p\,\partial_\lambda q\rangle$ . The strategy is the standard adiabatic one — transform to instantaneous action-angle variables, isolate the term injected by the motion of the transformation, and average over the fast angle — but now applied to the angle equation rather than the action equation. Watch the contrast with the adiabatic theorem: the drive term does not disappear on averaging, because $\partial/\partial I$ of an angle-average need not vanish.

Step 1. At each instant solve the time-independent problem at frozen $\lambda$ : a generating function $W(q,I;\lambda)$ produces $\theta = \partial W/\partial I$ , $p = \partial W/\partial q$ , with frozen Hamiltonian $H(I;\lambda)$ . Letting $\lambda$ move makes the transformation time-dependent, and by equation (5) the co-moving Hamiltonian is $K = H(I;\lambda) + \dot\lambda_\mu\,\partial W/\partial\lambda_\mu$ .

Step 2. Hamilton's equation for the angle in the new frame:

\dot\theta=\frac{\partial K}{\partial I}=\omega(I,\lambda)+\dot\lambda_\mu\,\frac{\partial}{\partial I}\!\left(\frac{\partial W}{\partial\lambda_\mu}\right)

Why both terms survive: the frequency is the frozen-dynamics part; the second term is first order in the slowness, and unlike its counterpart in the action equation ( $-\partial_\theta(\cdot)$ , a pure oscillation), it has a non-vanishing angle average.

Step 3. Average over the fast angle at fixed $\lambda$ . The oscillatory part of $\partial_I\partial_\lambda W$ integrates to zero over the slow loop (same spectral argument as the adiabatic theorem); the average survives:

\langle\dot\theta\rangle-\omega=\dot\lambda_\mu\,\frac{\partial}{\partial I}\Big\langle\frac{\partial W}{\partial\lambda_\mu}\Big\rangle_\theta

Step 4. Rewrite the average in phase-space terms. Since $p = \partial W/\partial q$ , the chain rule on the torus gives $\langle\partial_\lambda W\rangle_\theta = \langle p\,\partial_\lambda q\rangle_\theta$ up to a $\lambda$ -derivative of a single-valued function, which cannot contribute to a closed loop. Why this step matters: it converts a generating-function expression into an average of measurable quantities on the torus — this is the form in which the connection is actually computed in examples.

Step 5. Integrate over the loop, converting $\dot\lambda_\mu\,\mathrm{d}t \to \mathrm{d}\lambda_\mu$ :

\Delta\theta^{\mathrm{Hannay}}=\oint_C \frac{\partial}{\partial I}\Big\langle p\,\frac{\partial q}{\partial\lambda_\mu}\Big\rangle_\theta\,\mathrm{d}\lambda_\mu

The result depends on the oriented curve $C$ alone: reparametrizing time drops out with $\dot\lambda\,\mathrm{d}t$ , and gauge shifts of the angle origin change $\mathcal{A}$ by an exact form killed by the closed loop. Verified: matches Hannay (1985) Eq. (10) and Berry's classical-angle formula (J. Phys. A 18, 15, 1985); specializing to a rotated oscillator reproduces the solid-angle law derived next, and applying $-\partial/\partial n\,$ to the Berry phase of the corresponding WKB state recovers equation (4).

Go deeper · Derivation B · 2 minThe rotated oscillator: minus the enclosed solid angle

Derivation BThe rotated oscillator: minus the enclosed solid angle

The canonical one-degree-of-freedom example, and the one nature exhibits daily in the Foucault pendulum. An oscillator swings in a plane normal to an axis $\hat n(t)$ ; the axis is carried slowly around a closed cone on the unit sphere. We show the Hannay angle equals minus the solid angle the cone encloses. The strategy is transport bookkeeping: in the frame that follows the axis, the swing direction is parallel-transported, and the transport rule for a direction carried around a cone is classical differential geometry. Watch for the subtraction of the trivial full turn — the anholonomy is the deficit from naive expectations, not the raw rotation.

Step 1. Parametrize the axis $\hat n(t)=(\sin\theta_c\cos\phi,\,\sin\theta_c\sin\phi,\,\cos\theta_c)$ at fixed colatitude $\theta_c$ , with $\phi: 0 \to 2\pi$ slowly. The oscillation plane is normal to $\hat n$ ; the swing phase $\theta$ is measured from a reference direction in that plane.

Step 2. In the co-moving frame the equation for the swing direction acquires a rotation-rate term: the component of the frame's angular velocity along $\hat n$ is $\dot\phi\cos\theta_c$ , so

\dot\theta-\omega=-\dot\phi\,\cos\theta_c

Why only the axial component acts: rotations of the frame about axes in the oscillation plane tilt the plane but do not advance the phase within it; only the component about $\hat n$ rotates directions inside the plane. This is the same decomposition that isolates $\Omega\sin\lambda_0$ as the effective vertical rotation in the Foucault analysis of Coriolis Force.

Step 3. Integrate over the loop:

\Delta\theta=-\int_0^{2\pi}\cos\theta_c\,\mathrm{d}\phi=-2\pi\cos\theta_c

Step 4. Subtract the reference turn. The co-moving reference direction itself returns having made one full turn of $2\pi$ about the axis as $\phi$ wound around (the frame is dragged around the cone); the physical anholonomy — the mismatch between transported swing direction and restored reference — is

\Delta\theta_{\mathrm{Hannay}}=2\pi-2\pi\cos\theta_c \;(\mathrm{mod}\ 2\pi,\ \text{orientation})=-2\pi(1-\cos\theta_c)=-\,\Omega(C)

with $\Omega(C) = 2\pi(1-\cos\theta_c)$ the solid angle enclosed by the axis loop. The sign says the swing lags the transported frame. Why the answer is manifestly geometric: the rate $\dot\phi$ canceled between Steps 3 and 4 — fast or slow (within adiabaticity), one loop gives one solid angle. The curvature two-form here is the solid-angle form on the sphere, and equation (3) integrates it over the enclosed cap. Verified: with $\theta_c = 90° - \lambda_0$ this yields the Foucault deficit $2\pi(1-\sin\lambda_0)$ per sidereal day, equivalent to the textbook precession rate $\Omega_\oplus\sin\lambda_0$ ; agrees with Hannay (1985) and Berry (1985), and with the rotating-frame result in Coriolis Force.

Limiting cases

Case	Result	Meaning
Strict adiabatic limit	$\tau/T_{\mathrm{orbit}}\to\infty$	The dynamical phase grows without bound; the Hannay angle converges to a fixed, timing-independent value — its defining property.
Contractible / retraced loop	$\mathrm{area}(C)\to0$	Zero enclosed curvature flux ⇒ zero Hannay angle; a path retraced backward cancels exactly.
Semiclassical correspondence	$\theta_{\mathrm{H}}=-\partial\gamma_{\mathrm{B}}/\partial n$ at $I=(n+\tfrac12)\hbar$	The classical angle is the leading WKB content of the quantum geometric phase — the two formalisms see one geometry.
Rotated oscillator / Foucault	$\Delta\theta=-2\pi(1-\cos\theta_c)$	The angle deficit equals minus the solid angle enclosed by the axis loop — rate-independent, latitude-dependent.
Non-integrable dynamics	no angle variables	Hannay's construction needs tori; for chaotic systems the geometric residue of slow driving becomes geometric magnetism (Berry–Robbins), not an angle.

Common trapHannay's angle is not a small correction to the dynamical phase — it is a different kind of object

The dynamical phase $\int\omega\,\mathrm{d}t$ diverges as the loop slows; Hannay's angle converges. Comparing their sizes is a category error students make on every first pass: one scales with time, the other with enclosed geometry. Operationally, always subtract the instantaneous frequency integrated along the actual schedule — not the initial frequency times the elapsed time — or the leftover will be contaminated by ordinary frequency drift. And remember the actions are only conserved to adiabatic accuracy: the Hannay angle is the $O(1)$ , rate-independent part of the residue, cleanly separable from the $O(\varepsilon)$ non-geometric corrections only in the slow limit.

Worked examples

Example 1 — The Panthéon pendulum as a Hannay meter. Foucault's pendulum in Paris hangs at latitude $\lambda_0 = 48.85°\mathrm{N}$ . Compute the precession rate, the time for a full turn of the swing plane, and the geometric angle deficit after one sidereal day ( $T_\oplus = 23.93$ h).

Step 1. The axis (local vertical) is carried around the circle of latitude once per sidereal day — a cone of colatitude $\theta_c = 90° - 48.85° = 41.15°$ , so $\cos\theta_c = \sin\lambda_0 = 0.7529$ .

Step 2. Precession rate: the swing plane turns at $\Omega_\oplus\sin\lambda_0 = 360° \times 0.7529 = 271.1°$ per sidereal day, i.e. $271.1/23.93 = 11.3°$ per hour, clockwise seen from above.

Step 3. Full turn of the swing plane takes $360/271.1 \times 23.93\ \mathrm{h} = \boldsymbol{31.8\ \mathrm{h}}$ — the familiar "pendulum day", longer than 24 h everywhere south of the pole.

Step 4. The Hannay deficit after one sidereal day: $\Delta\theta_{\rm Hannay} = -2\pi(1-\sin\lambda_0) = -2\pi(0.2471) = -1.553\ \mathrm{rad} = \boldsymbol{-88.9°}$ . Check the bookkeeping: raw rotation $271.1°$ plus deficit $88.9°$ equals exactly $360°$ — the swing plane's failure to complete the full turn is the enclosed solid angle $\Omega = 2\pi(1-\sin\lambda_0)$ . At the pole ( $\lambda_0 = 90°$ ): deficit zero, full turn per day. At the equator: deficit $2\pi$ , no precession at all.

Example 2 — Separating geometric from dynamical, by experiment design. A bead oscillates on a hoop with period $T_{\rm osc} = 2.00$ s. The hoop's axis is carried once around a cone of half-angle $60.0°$ , first in $\tau_1 = 600$ s, then in $\tau_2 = 1200$ s. Predict the accumulated phases and show how the two runs isolate the Hannay angle.

Step 1. Dynamical phase, run 1: $\theta_{\rm dyn} = \omega\tau_1 = (2\pi/2.00)(600) = 1885.0\ \mathrm{rad}$ — exactly $600/2.00 = 300.0$ complete cycles, since the frequency is unchanged by the (pure-rotation) parameter loop.

Step 2. Hannay angle: $\Delta\theta_{\rm H} = -2\pi(1-\cos 60°) = -2\pi(0.500) = -\pi\ \mathrm{rad} = \boldsymbol{-180°}$ — the bead arrives half a cycle out of phase with a stationary twin oscillator.

Step 3. Run 2 at half speed: dynamical phase doubles to $3770.0$ rad ( $600.0$ cycles); the geometric shift stays $-\pi$ exactly. Subtracting the (known) integer cycle counts, both runs land the same residual half-cycle offset.

Step 4. The experimental logic: plot residual phase against loop time $\tau$ — a constant. Any rate-dependent contamination (frequency drift, $O(\varepsilon)$ corrections with $\varepsilon = T_{\rm osc}/\tau \le 3.3\times10^{-3}$ ) appears as $\tau$ -dependence and extrapolates away as $\tau\to\infty$ . The plateau at $-180°$ is the Hannay angle, measured exactly as equation (1) instructs: total advance minus $\int\omega\,\mathrm{d}t$ .

Historical arc

Origins · 1 minMichael Berry and how this was discovered

Michael Berry

1941– · Bristol

Berry's 1984 paper on quantal phase factors — written after a question about degeneracies in molecular spectra refused to go away — uncovered a quantity quantum mechanics had carried unnoticed for sixty years: a phase depending only on the circuit traversed in parameter space. Within a year his Bristol colleague John Hannay asked the sharp classical question — what does an angle variable do under the same slow circuit? — and Berry immediately supplied the semiclassical bridge, $\theta_{\rm H} = -\partial\gamma_{\rm B}/\partial n$ , publishing back-to-back with Hannay in 1985. The pattern is quintessential Berry: find the asymptotic borderland between two theories (his phrase is "physics of the in-between") and extract structures both sides had overlooked — geometric phases, superoscillations, the Riemann–zeta/quantum-chaos dialogue. Knighted in 1996, he shared the 1998 Wolf Prize with Aharonov; the Berry curvature that organizes topological matter, and the classical angle that organizes this article, are the same geometry he taught physics to see.

History

1851

— Foucault hangs a 67 m pendulum in the Panthéon; its precessing swing plane is the first public display of what is now read as a geometric angle

1956

— Pancharatnam, studying polarized light, constructs the geometric phase for optical beams — a generation ahead of its recognition

1979

— Mead and Truhlar find the molecular geometric phase at conical intersections, quantum chemistry's independent discovery of the same structure

1984

— Berry isolates the quantum geometric phase for adiabatic cycles; the subject acquires its name

1985

— Hannay derives the classical angle anholonomy for integrable systems; Berry links the two by the semiclassical derivative relation

Going further

References. Hannay, J. H., 'Angle variable holonomy in adiabatic excursion of an integrable Hamiltonian', J. Phys. A 18, 221 (1985) · Berry, M. V., 'Classical adiabatic angles and quantal adiabatic phase', J. Phys. A 18, 15 (1985) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 11-12 · Landau & Lifshitz, Mechanics, Vol. 1, §49-51 (adiabatic invariants) · Morin, Introduction to Classical Mechanics, Ch. on Lagrangian/adiabatic methods · Taylor, Classical Mechanics, Ch. 11 · Berry, M. V., 'Quantal phase factors accompanying adiabatic changes', Proc. R. Soc. A 392, 45 (1984)

Provenance. All equations cross-checked against Hannay (J. Phys. A 18, 221, 1985) and Berry (J. Phys. A 18, 15, 1985), with adiabatic-invariance facts from Landau-Lifshitz Mechanics §49-51 and action-angle formalism from Goldstein Ch. 10-11. A verifier should recompute: (1) the solid-angle law $-2\pi(1-\cos\theta_c)$ for the rotated oscillator; (2) the Foucault precession $\Omega_\oplus\sin\lambda_0$ per sidereal day; (3) the master one-form $\Delta\theta=\oint\partial_I\langle p\,\partial_\lambda q\rangle\,\mathrm{d}\lambda$ from the time-dependent canonical transformation; and (4) the Hannay-Berry differentiation relation. No non-standard numbers are asserted. The numeric examples (Paris pendulum: $271.1°$ /day, $31.8$ h pendulum day, deficit $-88.9°$ ; cone at $60°$ : Hannay $-\pi$ against dynamical $1885.0$ rad at $\tau=600$ s) were computed for this article and are hand-verifiable.

Builds on: Action-Angle Variables · Action Integrals and Invariant Tori · Frequencies from Action-Angle Variables · Liouville-Arnold Theorem · Generating Function to a Trivial Hamiltonian Leads to: Liouville Integrable Systems · Hamilton-Jacobi Equation · Bohr-Sommerfeld Quantization

Is the Foucault pendulum "really" explained by the Coriolis force or by geometry? Which explanation is the right one?

Both, exactly, and checking their agreement is a wonderful exercise. The rotating-frame calculation (done in Coriolis Force) treats the Coriolis term as a force and integrates the equations of motion: precession rate $\Omega_\oplus\sin\lambda_0$ . The Hannay calculation never mentions forces: it parallel-transports the swing direction as the local vertical is carried around the latitude circle and reads off the solid-angle deficit. Same number. The geometric view earns its keep in what it makes obvious: the answer cannot depend on the pendulum's length, amplitude, or frequency (none enter the connection), and the "pendulum day" is a property of the path on the globe, which is why a pendulum trucked slowly around any closed route accumulates the same deficit as one sitting at a latitude enclosing equal solid angle.

What exactly is the "gauge freedom" classically? There's no wavefunction to re-phase.

It is the freedom to choose where $\theta = 0$ sits on each torus, separately for every parameter value $\lambda$ . Any smooth re-origin $\theta \to \theta + \chi(\lambda)$ is legal bookkeeping; under it the connection shifts as $\mathcal{A} \to \mathcal{A} + \partial_\lambda\chi$ — precisely how a gauge potential transforms. Loop integrals kill the exact piece, so the Hannay angle is gauge-invariant, and the curvature $\mathcal{F}$ is gauge-invariant pointwise. This is not an analogy but the same mathematics as electromagnetic gauge freedom: the Hannay angle is a holonomy of a connection on a bundle of tori over parameter space, as the Berry phase is for a bundle of eigenstate rays.

Why does the general formula carry a derivative with respect to the action, when the hoop answer didn't seem to use it?

The master formula is $\Delta\theta = \oint \partial_I\langle p\,\partial_\lambda q\rangle\,\mathrm{d}\lambda$ , and generically the average $\langle p\,\partial_\lambda q\rangle$ depends on which torus you sit on, so the angle shift is action-dependent — different amplitudes acquire different shifts. Pure rotations are special: they displace every torus rigidly, the average is linear in $I$ , and $\partial_I$ produces an $I$ -independent answer — the solid-angle law, identical for all amplitudes. That degeneracy is why the Foucault deficit is universal. Deform the loop to include stiffness changes alongside rotations and the $I$ -dependence returns; correspondingly, on the quantum side, the Berry phase becomes $n$ -dependent beyond linear order.

The angle equation's drive term looks just like the action's. Why does averaging kill one and not the other?

Both read the same small Hamiltonian term $\dot\lambda\,\partial_\lambda G$ , but through different derivatives. The action's drive is $-\partial_\theta(\dot\lambda\,\partial_\lambda G)$ : a $\theta$ -derivative of a periodic function, hence structurally zero-mean — that is the adiabatic theorem. The angle's drive is $+\partial_I(\dot\lambda\,\partial_\lambda G)$ : an $I$ -derivative, which has no reason to average to zero, and its mean is exactly the connection $\mathcal{A}\dot\lambda$ . One symbol's difference in Hamilton's equations decides that actions forget the journey and angles remember it. This asymmetry — conservation from one canonical partner, holonomy in the other — is the cleanest statement of why geometry hides in phases throughout physics.

What happens to Hannay's angle if the system is chaotic instead of integrable?

The construction needs invariant tori: no integrability, no global angle variable, no Hannay angle per se. But the geometric residue of slow driving does not vanish — it changes species. Berry and Robbins (1993) showed that the averaged reaction of a chaotic fast subsystem on its slow parameters contains, at first order in slowness, two universal terms: geometric magnetism — a velocity-dependent force $F_\mu = B_{\mu\nu}\dot\lambda_\nu$ with $B$ an antisymmetric curvature two-form built from correlation functions — and deterministic friction, a dissipative reaction rooted in the chaos. Integrable Hannay holonomy and chaotic geometric magnetism are the two faces of the same adiabatic reaction, which is a satisfying preview of how this chapter's integrable machinery degrades gracefully into the quantum-chaos world.

How is Hannay's angle actually measured, given that the dynamical phase is thousands of times larger?

By exploiting its rate-independence, exactly as in worked example 2. Run the same closed parameter loop at several slow speeds, count total accumulated phase against a reference oscillator, subtract the computed $\int\omega\,\mathrm{d}t$ for each run, and extrapolate the residual to zero speed: the plateau is geometric. Real implementations: rotated mechanical resonators and torsion oscillators; the polarization plane of light in a helically coiled fiber (Tomita–Chiao 1986), where each photon's arrival is one "loop" and the rotation matches the fiber's solid angle to a percent; and neutron and trapped-ion interferometry on the quantum side, where the same subtract-and-extrapolate logic isolates the Berry phase. The Foucault pendulum needs no subtraction at all — there the dynamical motion (the swing) and the geometric drift (the plane's precession) live in different variables, visible to the naked eye.

Does the Hannay angle accumulate if I go around the loop N times?

Yes — linearly. Holonomy composes: $N$ circuits enclose $N$ times the curvature flux, so $\Delta\theta = -N\Omega$ for the rotated oscillator, while the dynamical phase grows like the total elapsed time. This is how weak geometric effects are amplified experimentally (many fiber coils, many driving cycles), and it is also the sharp distinction from any effect of the loop's shape traversal rate, which would not scale cleanly with $N$ at fixed slowness. On the quantum side the same $N$ -fold winding multiplies the Berry phase — the spectroscopic sidebands of periodically driven systems shift by integer multiples of the geometric phase per cycle.

Is there a Hannay angle for the second and third adiabatic invariants of a magnetized particle, or only for simple oscillators?

Any integrable (or near-integrable) fast motion with a conserved action carries a conjugate angle that can pick up holonomy — the construction is degree-of-freedom-agnostic. A magnetized particle has three nested action-angle pairs (gyration, bounce, drift), and each angle can acquire geometric shifts when the field configuration is cycled slowly: the gyrophase picks up a solid-angle contribution as the local field direction wanders (this is measurable in plasma turbulence contexts), and slow deformations of the mirror geometry shift the bounce phase. Multi-frequency Hannay angles come with the usual caveat of this chapter: the frequencies must stay non-resonant along the loop, or the averaging that defines the connection fails at the resonant surfaces — the recurring theme picked up in Degeneracy and Commensurate Frequencies.