Geodetic Precession (de Sitter Effect)

A gyroscope in free fall is the purest inertial-guidance device physics allows: cut every torque, and its spin axis should point at the same star forever. In flat spacetime it does. Willem de Sitter showed in 1916 that in curved spacetime it does not — the spin axis slowly swings, orbit after orbit, in the same sense as the motion. This is geodetic precession, the de Sitter–Fokker effect, one of the cleanest predictions in general relativity because it is pure geometry: a torque-free gyroscope parallel-transports its spin along its worldline, and parallel transport around a closed loop through curved geometry does not return a vector to itself. The leftover rotation — the holonomy of the connection — is the precession. What makes the effect physically rich rather than merely geometrical is that the coefficient is a sum of three comparable pieces that look unrelated until the metric ties them together: special relativity already forces an accelerated gyroscope to wobble (Thomas precession); gravitational time dilation tilts simultaneity around the orbit; and the curved space around a mass makes "straight ahead" itself bend. In Newtonian gravity all three vanish. In general relativity they combine into the factor three-halves, one part of which is spatial curvature (the PPN parameter γ) — the same γ that bends starlight and delays radar echoes. The prediction sat untested for decades until Gravity Probe B flew four fused-quartz gyroscopes in a polar orbit and read their drift with SQUID magnetometers, confirming the de Sitter rate to sub-percent precision — and the same physics, amplified a millionfold by strong fields, is watched directly in binary pulsars whose beams precess in and out of our line of sight.

Key takeaways

Geodetic precession is the rotation of a freely-falling gyroscope's spin relative to the distant stars, caused by parallel-transporting that spin through the curved spacetime around a mass as the gyroscope orbits.
For a circular orbit of radius $r$ about a non-rotating mass $M$ , the spin precesses prograde in the orbital plane at $\Omega_{\mathrm{geo}}=\tfrac{3}{2}\,\dfrac{GM}{c^2 r}\,\omega_{\mathrm{orb}}$ , with $\omega_{\mathrm{orb}}=\sqrt{GM/r^3}$ .
The coefficient $\tfrac{3}{2}$ is a sum of three pieces: Thomas precession $(-\tfrac12)$ , gravitational time dilation $(+1)$ , and spatial curvature $(+\gamma)$ ; in PPN language it reads $\gamma+\tfrac12$ , so gyroscope precession directly probes the curvature parameter $\gamma$ .
In low Earth orbit the rate is about $6.6$ arcsec/yr; Gravity Probe B measured $6601.8\pm18.3$ mas/yr against a prediction of $6606.1$ mas/yr, confirming general relativity at roughly the $0.3\%$ level.
Geodetic precession (sourced by mass $M$ , needs the gyroscope to move) is distinct from frame dragging (sourced by spin $\mathbf J$ , acts on a static gyroscope).
Strong fields amplify the effect enormously: the spin of PSR B1913+16 precesses at about $1.2^\circ$ /yr, seen as a secular change of its pulse profile.
Key idea: A freely-falling gyroscope parallel-transports its spin; the holonomy around one closed orbit rotates the spin by $\tfrac32(GM/c^2 r)$ radians in the prograde sense — a sum of Thomas precession, gravitational time dilation, and spatial curvature.

The core insight

A freely-falling gyroscope parallel-transports its spin; the holonomy of parallel transport around a closed orbit through curved spacetime rotates the spin by $\tfrac{3}{2}(GM/c^2 r)$ radians per orbit in the prograde sense, a sum of Thomas precession, gravitational time dilation, and spatial curvature.

Where this lives today

The geodetic effect is now a working tool of precision gravity. Gravity Probe B remains its landmark confirmation, but the same physics is measured in two other arenas. Lunar laser ranging watches the Earth–Moon system precess as de Sitter imagined it: the Moon's orbital angular momentum drifts by about $19.2$ mas/yr as the pair orbits the Sun, and millimetre-precision ranging to the Apollo retroreflectors confirms this to better than one percent — an independent solar-system test with entirely different systematics. Binary pulsars provide the strong-field frontier: in the double pulsar PSR J0737−3039 and in Hulse–Taylor's PSR B1913+16 the spin axis precesses at roughly a degree per year, and in the double pulsar this has literally swung one component's radio beam out of our line of sight, a disappearance timed to match general relativity. And because the coefficient is $\tfrac12(1+2\gamma)$ , every one of these measurements constrains the PPN parameter γ — independent of the light-bending and Shapiro-delay tests that pin $\gamma-1$ below $10^{-5}$ , and so a valuable probe of alternative theories of gravity.

The physical picture

Carry an arrow around a closed loop on the curved surface of the Earth, always keeping it "as straight as possible" — parallel transport. Walk down a meridian to the equator, a quarter-turn along it, and back up to the pole, and the arrow returns rotated ninety degrees, though you never turned it locally; the deficit equals the enclosed curvature times area. Geodetic precession is exactly this applied to a gyroscope's spin carried around its orbit through curved spacetime: the connection has non-trivial holonomy, and the spin comes back rotated. Because a torque-free gyroscope in free fall obeys parallel transport, this geometric rotation is directly measurable.

Figure

A freely-falling gyroscope parallel-transports its spin around one orbit and returns rotated by Δφ = 3π GM/c²r in the prograde sense — for a 7027 km orbit about Earth this accumulates to about 6.6 arcsec/yr

Three effects of comparable size set the coefficient. Thomas precession is purely special-relativistic: a gyroscope whose velocity keeps turning wobbles because successive Lorentz boosts do not commute — it appears even without gravity, in a rotating lab. Layered on top are the warping of simultaneity by gravitational time dilation and the genuine curvature of the space around the mass, the "rubber sheet" whose bending is measured by the PPN parameter γ. In Newtonian gravity every one of these is zero, so a torque-free gyroscope keeps a fixed direction exactly. The surviving three-halves is a clean relativistic fingerprint — and because one part is the spatial-curvature term γ, measuring the precession measures the same curvature that bends starlight.

Figure

The three-halves coefficient dissected: Thomas precession contributes −1/2, gravitational time dilation +1, and spatial curvature +γ (equal to +1 in general relativity), summing to Ω_geo = (γ + 1/2) GM/c²r ω = 3/2 GM/c²r ω

The core equations

A torque-free gyroscope does not obey ordinary parallel transport in general — it obeys Fermi–Walker transport, which subtracts the rotation an accelerated observer would see, keeping the spin orthogonal to the four-velocity and non-rotating relative to a local inertial frame.

\frac{\mathrm{D}S^\mu}{\mathrm{d}\tau}=\frac{1}{c^2}\,u^\mu\big(a_\nu S^\nu\big),\qquad u_\mu S^\mu=0 \qquad (1)

where: $S^\mu$ — gyroscope spin four-vector; $u^\mu$ — four-velocity, $u^\mu u_\mu=-c^2$ ; $a^\nu=\mathrm{D}u^\nu/\mathrm{d}\tau$ — four-acceleration; $\mathrm{D}/\mathrm{d}\tau$ — covariant derivative along the worldline. Reading: the spin stays perpendicular to the motion; only acceleration drives the extra term. For a gyroscope in free fall the four-acceleration vanishes, $a^\nu=0$ , and Fermi–Walker transport collapses to pure parallel transport — the equation we integrate around the orbit, where the Christoffel symbols mix the spatial spin components as the four-velocity sweeps the loop.

\frac{\mathrm{D}S^\mu}{\mathrm{d}\tau}=\frac{\mathrm{d}S^\mu}{\mathrm{d}\tau}+\Gamma^{\mu}{}_{\alpha\beta}\,u^\alpha S^\beta=0 \qquad (2)

where: $\Gamma^{\mu}{}_{\alpha\beta}$ — Christoffel symbols of the metric. Reading: each spin component changes only through the connection acting on the orbital velocity — no external torque enters. Integrating (2) around one circular geodesic in Schwarzschild spacetime gives the exact de Sitter result — compact, and valid all the way down to the innermost stable orbit, not just in the weak field.

\Omega_{\mathrm{geo}}=\omega_{\mathrm{orb}}\left[\,1-\Big(1-\frac{3GM}{c^2 r}\Big)^{1/2}\,\right],\qquad \omega_{\mathrm{orb}}=\sqrt{\frac{GM}{r^3}} \qquad (3)

where: $\Omega_{\mathrm{geo}}$ — spin precession rate relative to the stars (per unit coordinate time); $\omega_{\mathrm{orb}}$ — orbital frequency, exactly Keplerian for Schwarzschild circular orbits; $r$ — Schwarzschild radial coordinate. Reading: the spin lags the orbit by the gap between $\omega_{\mathrm{orb}}$ and the transported frequency $\omega_{\mathrm{orb}}\sqrt{1-3GM/c^2 r}$ . Expanding the square root for $r\gg GM/c^2$ gives the weak-field form used throughout the solar system. As a vector it is manifestly prograde — parallel to the orbital angular momentum — and connects to the Newtonian potential $\Phi=-GM/r$ .

\boldsymbol\Omega_{\mathrm{geo}}=\frac{3}{2}\,\frac{GM}{c^2 r^3}\,(\mathbf r\times\mathbf v)=\frac{3}{2}\,\frac{\mathbf v\times\nabla\Phi}{c^2}\Big|_{\text{sense: prograde}} \qquad (4)

where: $\mathbf r$ — position from the mass; $\mathbf v$ — orbital velocity; $\nabla\Phi=GM\hat{\mathbf r}/r^2$ — Newtonian field gradient. Reading: the rate follows the orbital angular momentum direction $\mathbf r\times\mathbf v$ , so the spin precesses about the orbit normal, independent of its own magnitude. Specializing to a circular orbit collapses the vector into the headline scalar rate — the number compared to experiment.

\Omega_{\mathrm{geo}}=\frac{3}{2}\,\frac{GM}{c^2 r}\,\omega_{\mathrm{orb}}=\frac{3}{2}\,\frac{(GM)^{3/2}}{c^2\,r^{5/2}} \qquad (5)

where: all symbols as above. Reading: the rate falls off as $r^{-5/2}$ , minute for distant orbits, steep as the orbit tightens toward a compact body. Promoting the coefficient to a general metric theory replaces the pure number by a combination of the PPN parameter γ — turning a gyroscope into a test of the theory of gravity, not merely a demonstration of it.

\Omega_{\mathrm{geo}}=\Big(\gamma+\tfrac12\Big)\frac{GM}{c^2 r}\,\omega_{\mathrm{orb}}=\frac{1+2\gamma}{2}\,\frac{GM}{c^2 r}\,\omega_{\mathrm{orb}} \qquad (6)

where: $\gamma$ — PPN parameter measuring spatial curvature per unit mass; general relativity fixes $\gamma=1$ , recovering $\tfrac32$ . Reading: the $\tfrac12$ is Thomas-plus-time-dilation; the $\gamma$ is spatial curvature, the same γ measured by light bending and the Shapiro delay. Multiplying the rate by the orbital period gives the holonomy angle per revolution — a clean way to compare spin precession with orbit precession.

\Delta\phi_{\mathrm{orbit}}=\Omega_{\mathrm{geo}}\,T_{\mathrm{orb}}=3\pi\,\frac{GM}{c^2 r} \qquad (7)

where: $T_{\mathrm{orb}}=2\pi/\omega_{\mathrm{orb}}$ — orbital period; $\Delta\phi_{\mathrm{orbit}}$ — spin rotation per revolution. Reading: this is exactly half the perihelion advance $6\pi GM/c^2 a(1-e^2)$ — same geometry, but spin and orbit weight the metric components differently.

Derivation AGeodetic precession from parallel transport in Schwarzschild spacetime

The goal is to derive the exact de Sitter rate of equation (3) directly from the parallel-transport equation, with no weak-field shortcut. We integrate the spin around one circular equatorial orbit of the Schwarzschild metric, reduce the coupled component equations to a single harmonic oscillator in proper time, and compare its frequency to the orbital frequency — the mismatch is the precession. We work in the equatorial plane $\theta=\pi/2$ and follow MTW §40.7.

Step 1. Take the Schwarzschild line element $\mathrm{d}s^2=-(1-2GM/c^2 r)c^2\mathrm{d}t^2+(1-2GM/c^2 r)^{-1}\mathrm{d}r^2+r^2\mathrm{d}\phi^2$ and a circular geodesic with $u^\mu=(u^t,0,0,u^\phi)$ . The coordinate orbital frequency is $\Omega\equiv\mathrm{d}\phi/\mathrm{d}t=u^\phi/u^t=\sqrt{GM/r^3}$ , exactly the Keplerian value. Why: the circular-geodesic condition $\Gamma^r{}_{tt}(u^t)^2+\Gamma^r{}_{\phi\phi}(u^\phi)^2=0$ fixes the frequency, and remarkably the relativistic answer coincides with the Newtonian one in these coordinates.

Step 2. Write the transport equations $\mathrm{d}S^\mu/\mathrm{d}\tau=-\Gamma^\mu{}_{\alpha\beta}u^\alpha S^\beta$ . Only the radial and azimuthal components couple dynamically: $S^t$ is pinned by orthogonality $u_\mu S^\mu=0$ and $S^\theta$ decouples, leaving $\dot S^r=-\Gamma^r{}_{tt}u^t S^t-\Gamma^r{}_{\phi\phi}u^\phi S^\phi$ and $\dot S^\phi=-\Gamma^\phi{}_{r\phi}u^\phi S^r$ . Why: the spin lives in the radial–tangential two-plane, exactly where the precession shows up; the time component is not an independent degree of freedom.

Step 3. Differentiate the $S^r$ equation and substitute to eliminate $S^\phi$ and $S^t$ , giving a single harmonic oscillator $\mathrm{d}^2 S^r/\mathrm{d}\tau^2=-\bar\Omega^2 S^r$ with $\bar\Omega^2=(u^t)^2\Omega^2\,(1-3GM/c^2 r)$ . Why: the crucial factor $1-3GM/c^2 r$ is where the geometry leaves its fingerprint — the same combination that governs stability at the innermost stable circular orbit.

Step 4. The spin oscillates at proper-time frequency $\bar\Omega$ , i.e. coordinate frequency $\Omega\sqrt{1-3GM/c^2 r}$ , while the orbit advances at $\Omega$ . The spin therefore precesses forward relative to the stars at $\Omega_{\mathrm{geo}}=\Omega-\Omega\sqrt{1-3GM/c^2 r}=\omega_{\mathrm{orb}}\big[1-\sqrt{1-3GM/c^2 r}\big]$ . Why: the transported spin returns to its original direction after a proper angle $2\pi\sqrt{1-3GM/c^2 r}$ , short of the full $2\pi$ swept by the orbit; the shortfall per unit time is the precession.

Step 5. Expand for $r\gg GM/c^2$ : $1-\sqrt{1-3GM/c^2 r}\approx \tfrac{3GM}{2c^2 r}$ , giving $\Omega_{\mathrm{geo}}\approx\tfrac32(GM/c^2 r)\,\omega_{\mathrm{orb}}$ . Why: this recovers de Sitter's 1916 weak-field rate and shows the exact formula is not an approximation but the parent expression — the three-halves is merely its leading term. *Verified: the exact form reduces to the de Sitter rate for large $r$ (Hartle §15.4, Rindler); at the ISCO $r=6GM/c^2$ it gives $\Omega_{\mathrm{geo}}=0.293\,\omega_{\mathrm{orb}}$ , i.e. $105^\circ$ of spin precession per orbit.*

Go deeper · Derivation B · 2 minWhy the coefficient is three-halves: Thomas, time dilation, and space curvature

Derivation BWhy the coefficient is three-halves: Thomas, time dilation, and space curvature

The goal is to show physically why the pure number is $\tfrac32$ rather than the bare Thomas factor $\tfrac12$ , by computing the precession from the parametrized-post-Newtonian metric and separating it into three distinct contributions. We identify the special-relativistic Thomas term from the orbital acceleration, then add the two gravitational terms — one from the time-time part of the metric, one from the space-space part scaled by γ — to expose the combination $\gamma+\tfrac12$ . This is the decomposition that makes gyroscope precession a probe of γ.

Step 1. Take the isotropic weak-field metric $\mathrm{d}s^2=-(1+2\Phi/c^2)c^2\mathrm{d}t^2+(1-2\gamma\Phi/c^2)\delta_{ij}\mathrm{d}x^i\mathrm{d}x^j$ with $\Phi=-GM/r<0$ . The time-time part encodes gravitational time dilation; the spatial part encodes curvature, scaled by γ. Why: keeping γ explicit lets us track exactly which physical effect each term carries, instead of collapsing everything into the number $\tfrac32$ prematurely.

Step 2. Even in flat spacetime, an accelerated gyroscope precesses by the Thomas rate $\boldsymbol\Omega_T=\tfrac{1}{2c^2}\,\mathbf a\times\mathbf v$ . For an orbit the relevant acceleration is the centripetal one, $\mathbf a=-\nabla\Phi$ , which makes the Thomas contribution retrograde: in units of $(GM/c^2 r^3)(\mathbf r\times\mathbf v)$ it contributes a coefficient $-\tfrac12$ . Why: this is the purely kinematic, non-commuting-boosts effect; it would be present for any turning gyroscope, gravitational or not, and it subtracts from the total.

Step 3. The gyroscope is actually in free fall, so its proper acceleration is zero; the effect replacing the inertial "a" comes from the metric gradients. The time-time part $g_{00}=-(1+2\Phi/c^2)$ contributes a prograde $+1$ (the time-dilation / geodesic piece), and the spatial part $g_{ij}=(1-2\gamma\Phi/c^2)\delta_{ij}$ contributes a prograde $+\gamma$ (dragging the spin through curved space). Why: these are the two genuinely gravitational contributions — one from how clocks run, one from how rulers bend — and only the second carries γ, which is why the precession probes spatial curvature specifically.

Step 4. Sum the three: $-\tfrac12+1+\gamma=\gamma+\tfrac12$ , equal to $\tfrac32$ when $\gamma=1$ . Equivalently, Thomas $(-\tfrac12)$ partially cancels time dilation $(+1)$ to leave a net $+\tfrac12$ , on top of which sits the full spatial-curvature $\gamma$ . Why: this is the whole story of the three-halves — not a Thomas factor, not pure curvature, but the specific tug-of-war the metric enforces, matching Weinberg §9.7 and Carroll §7.4. *Verified: the split $-\tfrac12$ (Thomas) $+1$ (time dilation) $+\gamma$ (curvature) $=\gamma+\tfrac12$ reproduces the geodesic-transport result for $\gamma=1$ ; the net non-curvature part $+\tfrac12$ is a common textbook grouping.*

Limiting cases

How the geodetic rate behaves in physically distinct regimes:

Case	Result	Meaning
Weak field, circular orbit	$\Omega_{\mathrm{geo}}=\tfrac{3}{2}\dfrac{GM}{c^2 r}\omega_{\mathrm{orb}}=\tfrac32\dfrac{(GM)^{3/2}}{c^2 r^{5/2}}$	Standard de Sitter rate; scales as $r^{-5/2}$ , steep for tight orbits.
PPN / general metric theory	$\Omega_{\mathrm{geo}}=\big(\tfrac12+\gamma\big)\dfrac{GM}{c^2 r}\omega_{\mathrm{orb}}$	The $\tfrac12$ is Thomas plus time dilation, the γ is spatial curvature; GR sets $\gamma=1$ .
Newtonian limit	$\displaystyle\lim_{c\to\infty}\Omega_{\mathrm{geo}}=0$	No precession: an $O(c^{-2})$ effect, gone when spacetime is flat.
Strong field (near ISCO)	$\Omega_{\mathrm{geo}}=\omega_{\mathrm{orb}}\big[1-\sqrt{1-3GM/c^2 r}\big]$	Exact form; at $r=6GM/c^2$ the spin turns $105^\circ$ per orbit.
Photon sphere ( $r\to 3GM/c^2$ )	$\Omega_{\mathrm{geo}}\to\omega_{\mathrm{orb}}$	The transported frequency vanishes; the spin makes a full turn per orbit.

Common trapGeodetic precession is not frame dragging, and the three-halves is not a Thomas factor

Two traps hide here. First, do not conflate geodetic precession with frame dragging (Lense–Thirring). Geodetic precession is sourced by the mass $M$ and requires the gyroscope to move — proportional to $\mathbf v$ , lying in the orbital plane. Frame dragging is sourced by the body's spin $\mathbf J$ and acts even on a static gyroscope, about a different axis; for Gravity Probe B the geodetic drift was about $180$ times larger, and the two were made orthogonal on the sky so they could be separated. Second, the coefficient $\tfrac32$ is not the Thomas factor $\tfrac12$ in disguise: Thomas contributes only $-\tfrac12$ , and the full result needs the gravitational time-dilation term $(+1)$ and spatial-curvature term $(+\gamma)$ . Calling geodetic precession "just Thomas precession in a gravitational field" is wrong by a factor of three and misses the curvature physics.

Worked examples

1 · Gravity Probe B in low Earth orbit (fully worked)

Four gyroscopes fly a near-polar orbit of semi-major axis $r=7027\ \mathrm{km}$ about Earth ( $GM_\oplus=3.986\times10^{14}\ \mathrm{m^3/s^2}$ ). Predict the geodetic drift rate in milliarcseconds per year and compare with the measured value.

Step 1. Compactness: $\dfrac{GM_\oplus}{c^2 r}=\dfrac{3.986\times10^{14}}{(8.988\times10^{16})(7.027\times10^{6})}=\mathbf{6.31\times10^{-10}}$ — of order $10^{-9}$ , confirming a weak-field effect where the leading three-halves form is ample.

Step 2. Orbital frequency: $\omega_{\mathrm{orb}}=\sqrt{GM_\oplus/r^3}=\mathbf{1.072\times10^{-3}\ rad/s}$ , a period $T=2\pi/\omega_{\mathrm{orb}}\approx\mathbf{97.7\ min}$ — consistent with Gravity Probe B's actual $97.65$ -min polar orbit.

Step 3. Rate: $\Omega_{\mathrm{geo}}=\tfrac32\,(6.31\times10^{-10})(1.072\times10^{-3})=\mathbf{1.015\times10^{-12}\ rad/s}$ , or $3.20\times10^{-5}$ rad over one year ( $3.156\times10^7\ \mathrm{s}$ ).

Step 4. Convert with $206\,265$ arcsec/rad: $\Omega_{\mathrm{geo}}=\mathbf{6.60\ arcsec/yr}=\mathbf{6604\ mas/yr}$ . Gravity Probe B's full prediction was $6606.1$ mas/yr and it measured $\mathbf{6601.8\pm18.3\ mas/yr}$ — agreement at the $0.3\%$ level, alongside the far smaller frame-dragging signal $37.2\pm7.2$ mas/yr. The de Sitter effect is confirmed: predicted $\approx6606$ mas/yr, measured $6602\pm18$ mas/yr.

2 · A relativistic binary pulsar (order-of-magnitude)

PSR B1913+16, the Hulse–Taylor binary, has orbital period $P_b=0.323\ \mathrm{day}$ , eccentricity $e=0.62$ , a companion mass $\approx1.39\,M_\odot$ , and relative semi-major axis $a\approx1.95\times10^9\ \mathrm{m}$ . Estimate the pulsar's geodetic spin precession rate.

Step 1. Field compactness at the orbit: $\dfrac{GM_2}{c^2 a}\approx\mathbf{1.05\times10^{-6}}$ — about a thousand times stronger than Gravity Probe B's orbit, which is why the effect is so dramatic here.

Step 2. Orbital frequency: $\omega_{\mathrm{orb}}=2\pi/P_b=2\pi/(2.79\times10^4\ \mathrm{s})=\mathbf{2.25\times10^{-4}\ rad/s}$ .

Step 3. Weak-field estimate with the eccentricity enhancement $1/(1-e^2)$ : $\Omega_{\mathrm{geo}}\approx\tfrac32(1.05\times10^{-6})(2.25\times10^{-4})/(1-0.62^2)\approx\mathbf{1.0^\circ/yr}$ . This is within a factor of order unity of the accepted $\approx1.2^\circ$ /yr — the residual is the binary mass-ratio factor of the exact Barker–O'Connell formula, omitted by our single-mass estimate. Strong fields turn a solar-system arcsecond-per-year into a degree-per-year, seen as a slow change in the pulsar's beam.

Physics of the portrait

Geodetic precession is one of the earliest quantitative predictions of general relativity — de Sitter published it in 1916, within months of Einstein's field equations. It sat as a paper prediction for the better part of a century, and its confirmation braided together relativity, cryogenics, and precision gyroscopy.

Origins · 1 minWillem de Sitter and how this was discovered

Willem de Sitter

1872–1934 · general relativity and cosmology

Willem de Sitter was a Dutch astronomer at Leiden and one of the earliest, most influential interpreters of Einstein's general relativity. In 1916, only months after the field equations appeared, he computed the relativistic precession of the Earth–Moon system treated as a gyroscope orbiting the Sun — the effect that now bears his name. His three-part series in the Monthly Notices of the Royal Astronomical Society carried general relativity to the English-speaking world during the First World War: it was through de Sitter that Eddington first learned the theory he would test at the 1919 eclipse. De Sitter also found, in 1917, the empty-but-expanding cosmological solution (de Sitter space) that anticipated the accelerating universe. Adriaan Fokker sharpened the gyroscope calculation shortly after, which is why the effect is sometimes called the de Sitter–Fokker precession. The vindication of his 1916 prediction by Gravity Probe B came seventy-seven years after his death — a reminder that a clean theoretical result can outlast every instrument capable of testing it and still be waiting, correct, when technology finally catches up.

History

1916

— de Sitter computes the relativistic precession of the Earth–Moon gyroscope about the Sun

1918

— Fokker refines the calculation; the effect becomes the de Sitter–Fokker precession

1960

— Schiff and Pugh independently propose an orbiting-gyroscope experiment to test it

2004

— Gravity Probe B launches four cryogenic quartz gyroscopes into polar orbit

2011

— Gravity Probe B reports the geodetic drift confirmed to sub-percent precision

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §40.7 and Box 40.3 (gyroscope precession) · Wald, General Relativity (1984), §6.5 · Carroll, Spacetime and Geometry (2004), §7.4 and the appendix · Hartle, Gravity (2003), §14.3 and §15.4 · Schutz, A First Course in General Relativity, 2nd ed. (2009), ch. 11 · Weinberg, Gravitation and Cosmology (1972), §9.7 · Everitt et al., "Gravity Probe B: Final Results of a Space Experiment to Test General Relativity", Phys. Rev. Lett. 106, 221101 (2011).

Provenance. The exact Schwarzschild rate and its weak-field limit follow MTW §40.7 and Hartle §15.4; the Thomas / time-dilation / spatial-curvature decomposition follows Weinberg §9.7 and Carroll §7.4 (net "geodesic" part $+\tfrac12$ , curvature part $\gamma$ ). Numbers are standard published values: the Gravity Probe B prediction $6606.1$ mas/yr and measurement $6601.8\pm18.3$ mas/yr (frame dragging $37.2\pm7.2$ mas/yr) from Everitt et al. 2011; lunar de Sitter $\approx19.2$ mas/yr; Hulse–Taylor spin precession $\approx1.2^\circ$ /yr. The worked Gravity Probe B estimate ( $6604$ mas/yr from $r=7027$ km) and per-orbit angle $3\pi GM/c^2 r$ were re-derived here. A verifier should reproduce equation (3), Taylor-expand to $\tfrac32 GM/c^2 r\,\omega$ , confirm the PPN coefficient $\tfrac12(1+2\gamma)$ , and check the numbers. No values are fabricated.

Prerequisites. Weak-Field Limit and Linearized Gravity · Schwarzschild Geodesics and the Effective Potential · Newtonian Limit of General Relativity · PPN Formalism. Leads to. Frame Dragging and the Lense–Thirring Effect · Gravitomagnetism · PPN Parameters Beta and Gamma.

Ask the tutor: Ask me to parallel-transport a spin vector around one Schwarzschild orbit step by step, or to work out the geodetic drift for any orbit you name — from a Gravity Probe B replica to a pulsar orbiting a neutron star.

Is geodetic precession the same thing as parallel transport, or a consequence of it?

It is a direct physical realization of it. A torque-free gyroscope in free fall parallel-transports its spin along its geodesic worldline — that is the content of the transport law once the four-acceleration is zero. Parallel transport around any closed loop in a curved space fails to return a vector to itself; the leftover rotation is the holonomy of the connection, fixed by the curvature enclosed by the loop. Geodetic precession is simply that holonomy made observable, with the "loop" being one orbit and the "vector" being the gyroscope's spin. This is why the effect is often called pure geometry: no forces act on the spin, and the rotation is entirely a property of how the spacetime is curved.

Why is the coefficient exactly three-halves and not two, or one?

Because it is a sum of three contributions with a specific sign structure. The special-relativistic Thomas precession contributes $-\tfrac12$ : an accelerated gyroscope wobbles retrograde because successive Lorentz boosts do not commute. Gravitational time dilation, encoded in the time-time part of the metric, contributes $+1$ . The curvature of space itself, encoded in the spatial part and measured by the PPN parameter γ, contributes $+\gamma$ , which equals $+1$ in general relativity. Adding them, $-\tfrac12+1+1=\tfrac32$ . Change the theory of gravity and γ changes, so the coefficient becomes $\gamma+\tfrac12$ — which is exactly why measuring gyroscope precession measures spatial curvature.

How can Gravity Probe B separate the geodetic drift from frame dragging if both are tiny?

By geometry and by size. The two effects point in orthogonal directions on the sky: the geodetic drift lies in the orbital plane, while the Lense–Thirring drift is perpendicular to it, in the plane of Earth's equator. The geodetic signal is also about $180$ times larger — roughly $6600$ versus $37$ mas/yr. So the experiment reads two nearly orthogonal drift components of the same gyroscopes, the large in-plane one geodetic and the small out-of-plane one frame dragging. The real difficulty was not separating them but modelling out unexpected electrostatic patch-effect torques on the rotors, which dominated the error budget and delayed the analysis by years.

Does the gyroscope's own spin rate or orientation affect the precession?

Not to leading post-Newtonian order. The rate $\Omega_{\mathrm{geo}}$ depends on the orbit — the mass $M$ , the radius $r$ , the orbital velocity — but not on how fast the gyroscope spins or which way its axis points within the orbital plane. The spin vector is simply carried by parallel transport, and that law does not care about the vector's length. This is what makes the effect a clean geometric measurement: any gyroscope in the same orbit precesses at the same rate. Spin magnitude matters only practically — a faster, more symmetric rotor is a better-behaved instrument with smaller residual torques.

Why is the spin holonomy per orbit exactly half of Mercury's perihelion advance?

Both come from the same Schwarzschild geometry — the radial and tangential parts of the metric distorted by the mass — but a precessing orbit and a transported spin sample that distortion with different weights. The perihelion advance is $6\pi GM/c^2 a(1-e^2)$ per orbit and the spin holonomy is $3\pi GM/c^2 r$ per orbit, exactly half in the circular limit. The factor of two reflects that the orbit's turning point responds to the full combination of radial and tangential metric coefficients while a parallel-transported vector responds to a particular half of it. It is a satisfying check that two distinct relativistic precessions are governed by one underlying geometry.

Could you ever see geodetic precession with an everyday gyroscope on Earth's surface?

In principle yes, but it is hopelessly small for a lab gyroscope sitting still, because the effect needs orbital motion — it scales with velocity through the field. A gyroscope carried around with Earth's rotation experiences a related precession, but at the surface the rate is milliarcseconds per year, swamped by mechanical torques, bearing friction, and vibration in any ordinary instrument. This is precisely why Gravity Probe B had to go to space: a drag-free polar orbit gives genuine free fall (proper acceleration essentially zero), the fused-quartz rotors were machined to near-perfect spheres to kill Newtonian torques, and the apparatus was cooled to liquid-helium temperatures so SQUID magnetometers could read the London moment of the superconducting rotors. The effect is real everywhere; measuring it just demands removing every larger effect first.