Parallel transport interpretation of Foucault precession

Look closely at Foucault's celebrated result and something is missing. The precession rate $\Omega\sin\lambda$ contains no pendulum length, no mass, no amplitude, no $g$ — every dynamical property of the apparatus has cancelled out. When a physical answer sheds all its dynamics, that is nature's way of confessing that the answer was never really about dynamics. The Foucault precession is, at bottom, a fact of geometry: the geometry of carrying a direction around a closed loop on a curved surface. Here is the reframing. The swing plane defines a direction — an arrow drawn on the local horizontal ground. As the Earth rotates, the pendulum's location is carried around a circle of latitude, and the ideal pendulum, feeling no torque about the local vertical, transports its arrow as faithfully as the curved Earth allows: never twisting it, only tilting it as the ground itself tilts. That operation has a name — parallel transport — and a famous property: around a closed loop on a curved surface, an arrow does not come home pointing where it started. It returns rotated by the holonomy angle, equal to the solid angle the loop encloses. For a latitude circle that solid angle is $2\pi(1-\sin\lambda)$ , and unpacking it recovers Foucault's sine law with almost no equations of motion at all. This geometric residue — an angle that depends only on the path, not on how fast the path is traced — was the first physically observed example of what is now called anholonomy or geometric phase. The same structure reappears in the Hannay angle of classical adiabatic systems and in Berry's phase in quantum mechanics: Foucault's pendulum, hanging in the Panthéon, was doing 1980s physics in 1851.

Key takeaways

The Foucault swing plane is a parallel-transported direction: the ideal pendulum exerts no torque about the local vertical, so its swing direction is carried around the latitude circle as straight as the sphere permits.
Parallel transport around a closed loop on a curved surface returns a vector rotated by the holonomy angle — equal to the solid angle the loop encloses (integrated Gaussian curvature).
For the latitude circle, the enclosed polar cap has solid angle $2\pi(1-\sin\lambda)$ : after one sidereal day the swing plane fails to close on its stellar orientation by exactly this angle, while turning $2\pi\sin\lambda$ relative to the ground.
The cone-unrolling construction makes this concrete: the tangent cone along the latitude circle unrolls into a flat sector of angle $2\pi\sin\lambda$ , and the missing wedge is the holonomy.
The result is rate-independent — a day or a year around the loop gives the same angle. That path-only dependence is the hallmark of a geometric phase.
The same anholonomy structure underlies gyroscope transport drift, the classical Hannay angle, and Berry's phase in quantum mechanics.
Key idea: the Foucault precession is the holonomy of parallel transporting the swing direction around a latitude circle on the sphere — the net rotation equals minus the enclosed solid angle $2\pi(1-\sin\lambda)$ , a geometric phase independent of pendulum dynamics.

The core insight

The Foucault precession is the holonomy of parallel transporting the swing direction (a tangent vector) around a latitude circle on the sphere; the net rotation equals minus the enclosed solid angle $2\pi(1-\sin\lambda)$ , a geometric phase independent of pendulum dynamics.

Where this lives today

Anholonomy graduated from curiosity to toolkit. In quantum processors, geometric (holonomic) gates steer a qubit's state around a closed loop in parameter space so that the acquired phase depends only on the enclosed solid angle — inheriting the Foucault property that the answer is immune to timing noise along the path, a robustness actively exploited in trapped-ion and superconducting platforms. In optics, the polarization of light guided along a coiled fibre rotates by the solid angle the propagation direction sweeps on its sphere — Berry's law confirmed by Tomita and Chiao in 1986, and now a routine consideration in fibre-optic gyroscope and quantum-communication design. The same transport arithmetic runs inertial navigation: strap-down algorithms integrate exactly the frame rotations this article computes, holonomy terms included, at kilohertz rates inside every airliner. The grandest verification is general-relativistic: a gyroscope orbiting a mass parallel-transports its spin through curved spacetime, and the per-orbit holonomy — the geodetic precession — was measured by Gravity Probe B at about $6.6$ arcseconds per year over Earth, confirming the curved-spacetime transport rule to $0.3\%$ . From a pendulum missing $88.9°$ a day in Paris to a satellite gyro missing $6.6''$ a year above it, the physics is one sentence: carried around a loop in a curved geometry, a direction comes home rotated by the curvature it enclosed.

The physical picture

Try the experiment with a spear instead of a pendulum. Stand somewhere on a globe, point the spear along the ground in some direction, and start walking a closed loop — always carrying the spear as "straight ahead as possible": never deliberately turning it left or right, letting it tilt only because the ground under your feet tilts. On a flat parking lot this procedure is boringly reliable: walk any loop and the spear comes home pointing exactly as it left. On the curved Earth it fails. Walk a loop and the spear returns rotated — not because you twisted it, not because any force acted on it, but because the patchwork of tilted tangent planes you marched through does not assemble into a flat floor. The failure angle is called the holonomy of the loop, and it is a property of the loop alone.

Figure

Carry a direction around a closed loop on the sphere and it comes home rotated

The Foucault pendulum runs this spear experiment involuntarily and to perfection. Its "spear" is the swing direction — the line in the horizontal plane along which the bob oscillates. Its "walk" is supplied by the rotating Earth, which carries the pendulum's location once around its circle of latitude every sidereal day. And its "carry it straight" discipline is enforced by dynamics: an ideal spherical pendulum has no torque about the local vertical, so nothing ever twists the swing plane within the horizontal — the plane changes orientation only in the minimal, unavoidable way forced by the local vertical itself swinging around the Earth's axis. That is precisely the definition of parallel transport. The pendulum is a direction being carried around a loop as straight as the sphere permits. Once you see this, the two headline mysteries of the Foucault effect dissolve at a glance. Why does the precession not care about length, mass, amplitude, or $g$ ? Because the answer is the holonomy of the loop, a property of the path — the dynamics only had to be good enough to enforce faithful transport, and any ideal pendulum is. Why the sine of the latitude? Because $2\pi(1-\sin\lambda)$ is the solid angle of the polar cap the latitude circle encloses — the sine arrives through the geometry of the sphere, not through any force balance. The dynamical derivation gets the same answer by patiently averaging the Coriolis force over swings; the geometry explains why that answer was destined to be so clean. There is a beautiful two-sided consistency here, and it is worth stating carefully because each frame sees a different "rotation." Relative to the ground, the swing plane turns $2\pi\sin\lambda$ per sidereal day — that is what the museum visitor sees. Relative to the stars, the plane fails to close by $2\pi(1-\sin\lambda)$ — that is the holonomy. The two add up to the ground frame's own $2\pi$ daily turn, exactly as they must. Only at the poles (cap shrinks to a point, holonomy zero) is the plane truly star-fixed; only at the equator (great circle, a geodesic) does the transported arrow come home unrotated — the $360°$ "deficit" there is a full turn, which is no rotation at all.

The mathematics of a carried direction

The transport rule comes first, because everything else is bookkeeping on top of it. Let $\hat{\mathbf{u}}$ be the unit vector along the swing direction, tangent to the sphere at the pendulum's current location, and let $\hat{\mathbf{z}}$ be the local vertical there. As the location moves, $\hat{\mathbf{z}}$ changes, and $\hat{\mathbf{u}}$ must stay horizontal. Parallel transport is the rule that does this with zero twist:

\frac{\mathrm{d}\hat{\mathbf{u}}}{\mathrm{d}t} = -\left(\hat{\mathbf{u}}\cdot\frac{\mathrm{d}\hat{\mathbf{z}}}{\mathrm{d}t}\right)\hat{\mathbf{z}} \qquad (1)

where: the right side is the smallest correction that keeps $\hat{\mathbf{u}}$ perpendicular to the tilting vertical — a pure projection, with no component that rotates $\hat{\mathbf{u}}$ within the horizontal plane. Any rotation about $\hat{\mathbf{z}}$ would be a twist, and the rule forbids twists. This innocent-looking equation is the Levi-Civita connection of the sphere in embedded disguise; we will use it only in this concrete vector form. The headline number is the failure of transport to close around a full latitude circle:

\Delta\phi_{\mathrm{inertial}} = -\,2\pi(1-\sin\lambda) \qquad (2)

where: $\Delta\phi_{\mathrm{inertial}}$ — the angle by which the transported direction misses its original stellar orientation after one loop; the sign says the miss is opposite to the sense of travel (a clockwise lag, for eastward transport in the north). Relative to the ground the same day of transport shows up as

\Delta\phi_{\mathrm{ground}} = +\,2\pi\sin\lambda \qquad \text{per sidereal day} \quad\Longrightarrow\quad \dot\phi = \Omega\sin\lambda \qquad (3)

where: dividing by $T_{\mathrm{sid}}$ recovers the Foucault rate — the geometric and dynamical routes agree number for number: in Paris ( $\sin\lambda = 0.753$ ) the plane turns $271.1°$ per day against the floor while lagging $88.9°$ against the stars, and $271.1° + 88.9° = 360°$ , the ground's own daily turn. The deficit angle has a geometric identity: it is the solid angle of the polar cap bounded by the latitude circle,

\Omega_{\mathrm{cap}} = \int_0^{2\pi}\!\!\int_{\lambda}^{\pi/2} \cos\lambda'\,\mathrm{d}\lambda'\,\mathrm{d}\varphi = 2\pi(1-\sin\lambda) \qquad (4)

where: $\lambda'$ runs over latitude, $\varphi$ over longitude, and $\cos\lambda'\,\mathrm{d}\lambda'\,\mathrm{d}\varphi$ is the area element of the unit sphere. At the pole the cap closes to nothing; at the equator it opens to the full hemisphere, $2\pi$ steradians. Holonomy equals enclosed solid angle — this is the sphere's case of the general law holonomy = integrated Gaussian curvature, and both derivations below land on it from opposite directions.

Which statement answers which question

you want	use	notes
the transport rule itself	eq. (1)	tilt without twist — pure projection
net rotation after one loop, vs stars	$-2\pi(1-\sin\lambda)$ , eq. (2)	path property; rate-independent
precession rate vs ground	$\dot\phi = \Omega\sin\lambda$ , eq. (3)	the dynamical law, recovered
the geometry behind the angle	$\Omega_{\mathrm{cap}} = 2\pi(1-\sin\lambda)$ , eq. (4)	holonomy = enclosed solid angle
a hands-on construction	cone unrolling, Derivation A	scissors-and-paper honest
the general theorem	Gauss–Bonnet, Derivation B	works for any loop, any surface

Derivation AThe cone construction — unroll the latitude circle flat

We are deriving the holonomy deficit (2) with nothing but scissors-and-paper geometry. The strategy exploits a lucky fact: parallel transport along a curve depends only on the surface's tangent planes along that curve, so any two surfaces tangent to each other along the whole curve transport vectors identically. The sphere is awkward (curved everywhere), but the cone tangent to the sphere along the latitude circle is developable — it unrolls flat without stretching — and on flat paper, parallel transport is just "keep the arrow pointing the same way." The step to watch is the sector angle: the unrolled cone is not a full disc, and the missing wedge is the entire effect.

Step 1. Build the tangent cone. In a meridian cross-section through the sphere of radius $R$ , the point $P$ at latitude $\lambda$ sits at $(R\cos\lambda,\ R\sin\lambda)$ . The tangent line at $P$ (perpendicular to the radius $OP$ ) meets the spin axis at an apex $A$ above the pole. The slant distance is

\ell_s = \lvert AP\rvert = \frac{R\cos\lambda}{\sin\lambda}

Why a cone at all: rotating this tangent line about the axis sweeps a cone touching the sphere along the entire latitude circle — so along that circle, sphere and cone share every tangent plane, and transport on one is transport on the other.

Step 2. Unroll the cone onto a flat sheet. It becomes a circular sector of radius $\ell_s$ . The sector's arc must equal the latitude circle's true circumference $2\pi R\cos\lambda$ , so the sector's opening angle is

\Theta = \frac{2\pi R\cos\lambda}{\ell_s} = 2\pi\sin\lambda

where: $\Theta$ — the flat angle the full trip around the Earth occupies on paper. It is less than $2\pi$ : the unrolled cone is a pac-man, and the missing wedge has angle $2\pi(1-\sin\lambda)$ . Why the arc is preserved: unrolling a developable surface bends without stretching — lengths along the surface are exact.

Step 3. Transport on the flat sector. A parallel-transported vector on paper simply keeps its direction fixed. Start at one straight edge of the sector with the vector at some angle to the radius; travel along the arc to the other edge; the vector still points the same way on paper — but the two edges of the sector are the same meridian on Earth, glued back together. Regluing rotates one edge onto the other by the missing wedge.

Step 4. Read off the holonomy: relative to the local landmarks (the meridian and the latitude circle), the vector returns rotated by the wedge angle

\Delta\phi = 2\pi - \Theta = 2\pi(1-\sin\lambda)

lagging behind the sense of travel — equation (2), and via the daily bookkeeping (3), the Foucault law. Why the limits are instant checks: at the pole the cone flattens into the tangent plane ( $\ell_s \to 0$ region degenerates, $\Theta \to 2\pi$ ), no wedge, no holonomy — the plane is star-fixed. At the equator the cone becomes a cylinder ( $\ell_s \to \infty$ ), $\Theta \to 0$ : the "deficit" is a full $2\pi$ , i.e. the transported vector returns exactly to itself — transport around a geodesic great circle is trivial, and the ground-relative precession $2\pi\sin 0 = 0$ vanishes. *Verified: the wedge $2\pi(1-\sin\lambda)$ equals the cap solid angle (4) computed independently by integration; construction follows von Bergmann & von Bergmann, Am. J. Phys. 75, 888 (2007).*

Figure

The tangent cone unrolled — the missing wedge is the holonomy

The cone gives the answer for a latitude circle; Gauss–Bonnet gives it for any loop, and shows what the angle really measures.

Go deeper · Derivation B · 2 minHolonomy equals enclosed curvature — the Gauss-Bonnet route

Derivation BHolonomy equals enclosed curvature — the Gauss-Bonnet route

We are re-deriving the deficit (2) from the Gauss–Bonnet theorem, to expose the general law behind the cone trick: the holonomy of any loop equals the total Gaussian curvature it encloses — for the unit sphere, the enclosed solid angle. The strategy is to compute the one nontrivial ingredient, the geodesic curvature of a latitude circle, and let the theorem do the rest. The point to watch is what geodesic curvature means: it is the rate at which the path itself turns within the surface — the steering a car on the sphere would need — and holonomy is precisely the mismatch between the path's total steering and the $2\pi$ that a flat-land loop would require.

Step 1. State the theorem for a smooth loop $\partial\Sigma$ bounding a disc-like region $\Sigma$ on a surface:

\oint_{\partial\Sigma}\kappa_g\,\mathrm{d}s \;+\; \iint_{\Sigma} K\,\mathrm{d}A \;=\; 2\pi \qquad (5)

where: $\kappa_g$ — geodesic curvature (in-surface steering rate of the path); $K$ — Gaussian curvature, $K = 1/R^2$ for the sphere. Why this is plausible: on a flat plane $K = 0$ and the theorem says any simple loop steers through a total of $2\pi$ — the familiar fact that a closed circuit turns you once around. Curvature enclosed inside the loop eats part of that $2\pi$ .

Step 2. Compute $\kappa_g$ for the latitude circle. The circle has radius $r = R\cos\lambda$ in space, but only the component of its curvature vector tangent to the sphere counts as steering. The in-space curvature $1/r$ points toward the circle's own center (on the axis); projecting onto the tangent plane leaves the fraction $\sin\lambda$ :

\kappa_g = \frac{\sin\lambda}{R\cos\lambda} = \frac{\tan\lambda}{R}

Why the projection: the component of curvature along the surface normal is the surface's business (it bends because the sphere bends); only the tangential remainder represents genuine steering within the surface. A great circle has its full curvature along the normal — $\kappa_g = 0$ , a geodesic, no steering at all.

Step 3. Integrate the steering around the loop of circumference $2\pi R\cos\lambda$ :

\oint\kappa_g\,\mathrm{d}s = \frac{\tan\lambda}{R}\times 2\pi R\cos\lambda = 2\pi\sin\lambda

where: the $R$ cancels — steering totals are scale-free, as any angle must be. Following the latitude circle requires $2\pi\sin\lambda$ of deliberate turning per lap: none at the equator (geodesic), a full $2\pi$ at a tiny circle hugging the pole (an ordinary flat-land circle).

Step 4. A parallel-transported vector, by definition, ignores the path's steering — it only inherits the unavoidable rotation. The transported direction therefore returns rotated, relative to its start, by the difference between flat-land expectation and actual steering:

\Delta\phi = 2\pi - \oint\kappa_g\,\mathrm{d}s = 2\pi(1-\sin\lambda) = \iint_\Sigma K\,\mathrm{d}A = \Omega_{\mathrm{cap}}

Why the last equality: by (5) the shortfall is exactly the enclosed curvature integral, which on the unit sphere is the cap area — the solid angle (4). Holonomy is stored curvature; the loop merely collects it. *Verified: $\oint\kappa_g\,\mathrm{d}s + \Omega_{\mathrm{cap}} = 2\pi\sin\lambda + 2\pi(1-\sin\lambda) = 2\pi$ , closing Gauss–Bonnet exactly; the cone construction gives the identical wedge, and both reduce to zero holonomy at the pole and trivial $2\pi$ at the equator.*

Why the pendulum obeys the transport rule

The geometry is compelling, but it earns its keep only if the dynamics actually enforces equation (1) — otherwise the pendulum is a spear-carrier with no discipline. The honest argument has two parts. First, torque: gravity and the wire tension both act through the vertical plane containing the bob, so neither exerts any torque about the local vertical axis through the pivot. Nothing in the ideal pendulum's force inventory can rotate the swing direction within the horizontal plane. Second, time scales: the swing is enormously faster than the transport ( $\omega_0 \sim 0.4\ \mathrm{rad/s}$ against $\Omega \sim 7.3\times10^{-5}\ \mathrm{rad/s}$ — four orders of magnitude), so over any single oscillation the transport is effectively frozen and the swing direction is a well-defined, slowly evolving line. This is an adiabatic separation: the fast system (the swing) preserves its orientation because the slow change (the moving vertical) is too gradual to excite any twist, and the only evolution left to the swing direction is the minimal tilt that keeps it horizontal — which is precisely the projection rule (1). That is also the deep reason the Foucault angle previews modern physics: "fast system carried slowly around a closed circuit in parameter space, returning with a residual angle set by the enclosed geometry" is, word for word, the recipe for the Hannay angle in classical mechanics and Berry's phase in quantum mechanics. The pendulum's parameter space is the sphere of local verticals; the enclosed geometry is the solid angle of the cap.

Limiting cases

Case	Result	Meaning
pole, $\lambda = 90°$	holonomy $= 0$ ; plane star-fixed	cap closes to a point; ground-relative turn is the full $2\pi$ per day, all of it the floor's doing
equator, $\lambda = 0$	deficit $= 2\pi \equiv 0$ ; no precession	great circle is a geodesic: $\kappa_g = 0$ , transport is trivial, and $2\pi\sin 0 = 0$ against the ground
small circle near pole	$\oint\kappa_g\,\mathrm{d}s \to 2\pi$	tiny loop on an effectively flat cap — ordinary flat-land steering, no curvature enclosed
Paris, $\lambda = 48.85°$	ground $+271.1°$ /day, stars $-88.9°$ /day	the two rotations sum to $360°$ — the ground frame's own daily turn
flat rotating platform	precession at full $\Omega$ everywhere	no curvature, no sine law: the latitude dependence is the Earth's roundness speaking
any loop, any surface	$\Delta\phi = \iint K\,\mathrm{d}A$	the general law: holonomy is the curvature you enclose, latitude circles are just the special case

Common trapSomething must be pushing the swing plane around

Nothing pushes it. An ideal pendulum feels zero torque about the local vertical — gravity and tension both pass through the vertical plane of the bob — so no force ever rotates the swing direction within the horizontal. The precession is not a forced rotation but a bookkeeping mismatch: the local frame (north, east) twists as the Earth carries it around, while the faithfully transported swing direction refuses to twist along. Seen from the rotating frame, the Coriolis force "explains" the drift; seen geometrically, the drift is the holonomy of the path and no agent acts at all. Related trap: do not conclude from rate-independence that dynamics is irrelevant — the dynamics is what enforces the transport rule (no vertical torque, fast swing, adiabatic separation). Break it — drive the pivot, swing elliptically, use a short stiff pendulum — and the clean geometric angle is contaminated by ordinary dynamical precession, which is exactly the plague of real installations discussed in the Foucault pendulum article.

Worked examples

1 · Paris, both ledgers (confidence builder)

For the Panthéon pendulum at $\lambda = 48.85°$ N, compute the daily swing-plane rotation relative to the ground and relative to the stars, verify they are consistent, and check the stellar deficit against the cap solid angle.

Step 1. $\sin 48.85° = 0.753$ . Ground ledger, from (3): $\Delta\phi_{\mathrm{ground}} = 360°\times0.753 = \boxed{271.1°\ \text{per sidereal day}}$ , clockwise from above. Divided by $23.93\ \mathrm{h}$ this is $11.3°/\mathrm{h}$ — the rate every visitor's guidebook quotes.

Step 2. Stellar ledger, from (2): $\Delta\phi_{\mathrm{inertial}} = -360°\times(1-0.753) = \boxed{-88.9°\ \text{per day}}$ — after one sidereal day the plane is nearly a right angle away from its stellar orientation. The swing plane is emphatically not fixed among the stars.

Step 3. Consistency: the ground frame itself rotates $+360°$ per sidereal day about the local vertical. Ground-relative turn = stellar turn + frame turn: $-88.9° + 360° = 271.1°$ . ✓

Step 4. Solid-angle check, from (4): $\Omega_{\mathrm{cap}} = 2\pi(1-0.753) = 1.552\ \mathrm{sr}$ , and $1.552\ \mathrm{sr}\times(180/\pi)° /\mathrm{rad} = 88.9°$ . The stellar deficit is the cap, angle for steradian. ✓

2 · A gyroscope flown around the world (exam level)

A navigation-grade gyroscope, mounted torque-free in gimbals, is flown eastward around the Earth along the $60°$ N parallel, returning to its start. (a) By what angle has its spin axis rotated relative to local north, from the transport alone? (b) Repeat for a route along $30°$ N, and along the equator. (c) Why does the flight speed not matter, and what must be subtracted in a real measurement?

Step 1. The gimbals guarantee no torque about any axis: the gyro axis is parallel-transported along the flight path. From (2), the loop at $\lambda = 60°$ : bearing shift $= 360°(1-\sin 60°) = 360°\times(1-0.866) = \boxed{48.2°}$ , lagging the direction of travel (rotated clockwise relative to north, viewed from above, for an eastward circuit).

Step 2. At $\lambda = 30°$ : $360°\times(1-0.500) = \boxed{180°}$ — the gyro comes home pointing backwards relative to its departure bearing. A dramatic, purely geometric reversal.

Step 3. Along the equator: $360°\times(1-0) = 360° \equiv \boxed{0°}$ — the equator is a geodesic; transport around it is trivial and the axis returns exactly.

Step 4. Speed independence: the holonomy is a property of the path, not the schedule — halving the flight time changes nothing in (2), the signature of a geometric phase. What a real gyro also accumulates is the Earth-rotation contribution: during a flight of duration $t$ the planet's own spin parallel-transports and rotates local frames, adding the Foucault-type drift of order $\Omega\sin\lambda\, t$ ( $\approx 156°$ at $60°$ N for a 12-hour circuit). A real measurement logs both and subtracts the timed Earth-spin part; the leftover, schedule-independent $48.2°$ is the geometry of the flown loop. Inertial navigation systems perform exactly this bookkeeping continuously.

Origins · 1 minCarl Friedrich Gauss and how this was discovered

Carl Friedrich Gauss

1777–1855 · Göttingen Observatory

Gauss spent the 1820s doing physics with his boots on: directing the triangulation survey of the Kingdom of Hanover, hauling instruments up hills to measure a curved Earth with straight lines of sight. Out of that mud came the Disquisitiones generales circa superficies curvas (1827), where he proved the result he privately called his theorema egregium — the "remarkable theorem" that Gaussian curvature is intrinsic, measurable by a surveyor confined to the surface with no view of the embedding space. The angle deficit of triangles, the failure of transported directions to return — these are consequences of curvature a flatlander can detect, and Gauss showed how to compute them. The theorem now named Gauss–Bonnet, tying a loop's total steering to the curvature it encloses, is the engine of this article: the Foucault pendulum is a surveying instrument in Gauss's exact sense, reading the Earth's intrinsic curvature from inside. Gauss never saw Foucault's pendulum swing — he died four years after the Panthéon demonstration — but the angle it fails to close by is computed with his mathematics.

History

1827

— Gauss publishes the theorema egregium: curvature is intrinsic to a surface, detectable by measurements made entirely within it.

1851

— Foucault's pendulum turns under the Panthéon dome; the sine law is announced, its geometric meaning unsuspected.

1917

— Levi-Civita defines parallel transport, giving the "carry it as straight as possible" rule its exact mathematical form and connecting it to curvature.

1984

— Michael Berry identifies the geometric phase in quantum mechanics: adiabatic cycles return a system with a phase set by enclosed geometry, not elapsed time.

1985

— John Hannay works out the classical counterpart, the Hannay angle, with the Foucault pendulum as the founding example of a classical anholonomy.

2007

— von Bergmann and von Bergmann publish the elementary cone-unrolling derivation, making the holonomy computable with scissors and paper.

Going further

References. Morin, Introduction to Classical Mechanics, Ch. 10 (Coriolis force and the Foucault pendulum) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 4–5 (rotating frames) · Landau & Lifshitz, Mechanics, Vol. 1, §39 (motion in a non-inertial frame) · Taylor, Classical Mechanics, Ch. 9 (rotating reference frames) · Berry, 'Quantal phase factors accompanying adiabatic changes', Proc. R. Soc. Lond. A 392 (1984) · von Bergmann & von Bergmann, 'Foucault pendulum through basic geometry', Am. J. Phys. 75, 888 (2007)

Provenance. Precession rate $\dot\phi=\Omega\sin\lambda$ : verify by solving the linearized Coriolis equations (Morin Ch.10, Goldstein Ch.4, Taylor Ch.9). $\Omega=2\pi/T_{\mathrm{sid}}$ with $T_{\mathrm{sid}}\approx86164\,\mathrm{s}$ gives $\Omega\approx7.292\times10^{-5}\,\mathrm{rad/s}$ (standard astronomical value). Cap solid angle $2\pi(1-\sin\lambda)$ : recompute the surface integral. Holonomy = integrated Gaussian curvature: Gauss–Bonnet theorem (standard differential geometry; do Carmo). Parallel-transport derivation and geometric-phase identification follow von Bergmann & von Bergmann, Am. J. Phys. 75, 888 (2007) and Berry, Proc. R. Soc. A 392 (1984). Paris latitude $\approx48.85^\circ$ and full-circle time $\approx31.8$ h follow directly from the rate. Worked-example numbers on this page (Paris $271.1°$ /day ground, $-88.9°$ /day stellar, cap $1.552$ sr; gyroscope $48.2°$ at $60°$ N, $180°$ at $30°$ N; cone slant $\ell_s = R\cos\lambda/\sin\lambda$ , sector angle $2\pi\sin\lambda$ ; $\oint\kappa_g\,\mathrm{d}s = 2\pi\sin\lambda$ from $\kappa_g = \tan\lambda/R$ ) are illustrative arithmetic recomputed from the boxed formulas.

Builds on: Coriolis force · Angular velocity vector of a rotating frame · Time derivatives in rotating frames · Effective gravity on the rotating Earth. Leads to: Non-inertial reference frames · Centrifugal force.

How can the answer not depend on how fast the loop is traversed? Surely a faster Earth would drag the plane around faster

Two different quantities hide in "faster." The rate of precession against the ground does scale with $\Omega$ — spin the Earth faster and the museum visitor sees a faster creep. But the angle per completed loop, $2\pi(1-\sin\lambda)$ , is fixed by the path alone: doubling $\Omega$ halves the loop time and doubles the rate, leaving the per-lap angle untouched. That is what "geometric" means — the holonomy is the line integral of a connection along the path, with time nowhere in it. The same signature identifies Berry phases in quantum mechanics: change the schedule, keep the circuit, and the geometric part of the answer does not move.

The deficit at the equator is 360 degrees, but the article says equatorial transport is trivial. Which is it?

Both — a rotation by $360°$ is the identity for a direction. The formula $2\pi(1-\sin\lambda)$ measures the mismatch against the local landmarks after regluing the cone's seam, and at $\lambda = 0$ it returns $2\pi$ : the transported vector comes home exactly aligned with its start. Consistently, the equator is a geodesic ( $\kappa_g = \tan 0/R = 0$ ): a path that needs no steering imposes no net rotation. And the ground-relative precession is $2\pi\sin 0 = 0$ — the pendulum on the equator never turns against the floor. The apparent paradox is just the difference between "rotated by a full turn" and "rotated by nothing," which for physical directions is no difference at all.

Is anything actually exerting a torque on the swing plane to make it precess?

No — and that is the whole point of the geometric view. Gravity pulls through the vertical; the wire tension pulls through the pivot in the bob's vertical plane; neither has a moment about the local vertical axis. The swing direction is never twisted by any force. What changes is the stage: the local frame of north and east twists as the Earth carries it around its latitude circle, while the faithfully transported swing direction declines to twist along. The rotating-frame observer books the mismatch as a Coriolis effect; the inertial geometer books it as holonomy. Same physics, two ledgers — the Coriolis force article runs the first ledger, this article the second.

What exactly is parallel about parallel transport if the vector visibly changes direction in 3D?

"Parallel" is judged from inside the surface, not from outside. The rule (1) changes the vector only by the projection needed to keep it tangent as the tangent plane tilts — it never rotates the vector within the surface. An inhabitant of the surface, comparing the vector to their own straight-ahead, sees no turning at all along the way; the rotation appears only when the loop closes and the start and end copies sit side by side. From the 3D outside, of course, the vector tilts continuously — but that extrinsic tilting is the sphere's business, not the vector's. Gauss's theorema egregium is precisely the statement that the interesting part (curvature, and hence holonomy) is intrinsic — measurable by the inhabitant with no external view.

Does a gyroscope carried around the Earth really behave like the Foucault pendulum?

Yes — a torque-free gyroscope is an even purer direction-carrier than the pendulum, with no swing dynamics to average over. Carried around a closed path, its axis returns rotated by the enclosed solid angle, exactly the pendulum's law; the worked example computes $48.2°$ for a lap of the $60°$ N parallel. This is not academic: inertial navigation systems must integrate these transport rotations continuously, and marine gyrocompasses are engineered around them. The relativistic upgrade replaces the curved Earth by curved spacetime: Gravity Probe B's orbiting gyroscopes precessed by $6.6''$ per year — the geodetic effect — confirming that parallel transport, now with spacetime curvature under the integral, still writes the answer.

If the Earth were flat but still rotating, would a Foucault pendulum precess?

It would — at the full rate $\Omega$ , everywhere on the platform, with no latitude dependence. On a rotating flat disc the local vertical is parallel to the rotation axis at every point, so the whole rotation projects onto the vertical and the swing plane lags the floor at $\Omega$ exactly, just as at Earth's pole. What the flat world loses is the sine law: the variation of the rate with position exists only because the Earth's surface curves, tilting the local vertical away from the spin axis by $90°-\lambda$ . So the pendulum's precession by itself proves rotation; its latitude dependence is the part that quietly measures the roundness — the enclosed-solid-angle geometry this article develops.

Is the Berry phase really the same mathematics, or just a loose analogy?

The same mathematics, made precise. In both problems a fast system is carried slowly around a closed circuit of external parameters, and returns with a residual angle (or phase) equal to a geometric quantity of the circuit — a solid angle in the simplest cases — independent of the traversal schedule. For the pendulum: fast swing, slow transport of the local vertical around a cone of directions, residue $2\pi(1-\sin\lambda)$ . For a quantum spin in a slowly reorienting magnetic field: fast Larmor precession, slow field loop, Berry phase $=$ (spin quantum number) $\times$ (enclosed solid angle). Hannay formalized the classical case in 1985 with the Foucault pendulum as the founding example. The dictionary is exact: adiabatic separation ↔ adiabatic theorem, transport rule ↔ connection, holonomy ↔ geometric phase.