Geodesic Deviation and Tidal Forces

Drop two ball bearings side by side from a tall tower. In the frame of either bearing the other floats weightless — the equivalence principle has abolished gravity locally. Yet watch long enough and the two do not stay put relative to one another: because both fall toward the centre of the Earth, their paths quietly converge, and stacked vertically the lower one would pull ahead. Gravity, banished from each worldline, reappears in the gap between worldlines. That irreducible relative behaviour is the tidal field, and the geodesic deviation equation states exactly what governs it: not the gradient of a force, but the curvature of spacetime. The setting is a one-parameter family of freely-falling paths — a congruence — with an affine parameter $\tau$ along each geodesic and a label $s$ picking out neighbours. The tangent $u^\mu=\partial x^\mu/\partial\tau$ obeys the geodesic equation; the connecting vector $\xi^\mu=\partial x^\mu/\partial s$ measures the separation. Because both come from the same coordinate net their bracket vanishes, $[u,\xi]=0$ , and differentiating the separation twice along the flow — invoking the Riemann tensor exactly where covariant derivatives refuse to commute — gives the cleanest physical reading of curvature there is. A single geodesic is locally straight; only a family reveals the geometry, converging where curvature is positive and spreading where it is negative. This one equation is the hinge of general relativity. Its Newtonian limit fixes curvature as the second derivative of the potential and pins the coupling constant in Einstein's equations; its trace becomes the Raychaudhuri focusing equation behind the singularity theorems; and its trace-free part is what a laser interferometer registers when a gravitational wave stretches one arm while squeezing the other. Tidal force, spaghettification, detector strain, and the inevitability of singularities are one geometric fact seen from different distances.

Key takeaways

Geodesic deviation gives the relative acceleration of two neighbouring free-fallers: $\frac{D^2 \xi^\mu}{D\tau^2} = -R^\mu{}_{\nu\rho\sigma}\,u^\nu \xi^\rho u^\sigma$ — pure curvature, no force.
Flat spacetime gives zero relative acceleration, so geodesic deviation is the operational signature of curvature; you cannot transform it away by any choice of coordinates.
Newtonian tidal forces $\partial_i\partial_j\Phi$ are the weak-field, slow-motion limit, identifying $R^i{}_{0j0}\approx\partial_i\partial_j\Phi$ and fixing Einstein's constant.
The covariant derivative $D/D\tau$ is essential: ordinary derivatives of vectors at different points are non-tensorial and mix genuine tides with coordinate artefacts.
A small ball of dust shears into a trace-free ellipsoid in vacuum (volume preserved) and contracts when matter is present — Raychaudhuri focusing, which seeds singularity theorems; near a point mass the tidal eigenvalues are $(+2,-1,-1)\,GM/r^3$ .
Key idea: The relative acceleration of two neighbouring freely-falling particles equals the Riemann curvature tensor contracted twice with their common four-velocity and once with their separation vector — tidal force IS curvature.

Where this lives today

The geodesic deviation equation is the working model behind an entire branch of observational physics. Every gravitational-wave detection — from the first LIGO event in 2015 through the merger catalogue now assembled by LIGO, Virgo, and KAGRA — is literally a measurement of $\ddot\xi^i=-R^i{}_{0j0}\xi^j$ : freely-suspended mirrors trace geodesics, and the differential arm-length change reads out the passing curvature to a part in $10^{21}$ , a principle the space-based LISA will extend to million-kilometre arms. In astrophysics, tidal disruption events — stars shredded near a supermassive black hole — are the $1/r^3$ tidal tensor caught in the act, their light curves probing black-hole masses. The same focusing term underpins the singularity theorems and trapped-surface analysis, tools sharp enough that Penrose's use of them earned the 2020 Nobel Prize. And in cosmology, geodesic deviation is the language of gravitational lensing: the shear and convergence that distort galaxy shapes in weak-lensing surveys are the optical scalars of a null congruence mapping the dark-matter tidal field across the sky.

The physical picture

Release a small spherical cloud of motionless test particles high above the Earth. Each falls along its own geodesic toward the planet's centre, and because those geodesics all aim at the same point, particles displaced sideways drift toward each other as the field lines converge, while particles stacked vertically separate, since the lower one sits in a stronger field. The sphere deforms — squeezed horizontally, stretched vertically — into a prolate ellipsoid, and to leading order in vacuum the squeeze and stretch balance exactly, so the volume is preserved while the shape distorts. That ellipsoid is the geodesic deviation equation made visible, and the trace-free tidal tensor is its generator.

Figure

A small ball of free-fallers above the Earth deforms into an ellipsoid: stretched radially by +2GM/r³ = +3.1×10⁻⁶ s⁻² per metre and squeezed transversely by −GM/r³ = −1.5×10⁻⁶ s⁻² per metre. The three eigenvalues +2, −1, −1 in units of GM/r³ sum to zero, so the volume is preserved — pure tidal shear

Now zoom out to the geometry that causes it. Two travellers set off due north from adjacent points on the equator, each along a great circle — a geodesic on the sphere. They begin exactly parallel, yet by the pole their paths have crossed: nothing pushed them together, the surface's positive curvature focused the geodesics. The separation $\xi(s)$ obeys $\mathrm{d}^2\xi/\mathrm{d}s^2=-\xi/a^2$ on a sphere of radius $a$ , whose solution $\xi(s)=\xi_0\cos(s/a)$ vanishes after a quarter circumference. This is the deviation equation in its simplest form; the spacetime version merely replaces the Gaussian curvature $1/a^2$ with the full Riemann tensor contracted along the four-velocity.

Figure

Two geodesics that start parallel on a positively curved surface are focused together: the connecting vector obeys ξ″ = −ξ/a² on a sphere of radius a, with solution ξ(s) = ξ₀ cos(s/a), so an initial gap ξ₀ closes to zero after a quarter circumference s = πa/2. Curvature, not a force, drives the convergence

Why insist on the covariant derivative rather than an ordinary one? Comparing the velocities of two particles at different points means comparing vectors in different tangent spaces — meaningless until one is parallel-transported to the other's location. The connection performs that transport, and the leftover, transport-independent piece is the genuine relative acceleration. In flat space transport is path-independent and the leftover vanishes; that it does not vanish in curved space, and is proportional to the Riemann tensor, is the whole content of the equation.

The core equations

The central statement is that the second covariant derivative of the separation is a linear map on $\xi$ built entirely from curvature. Everything below either defines the ingredients or contracts and specialises this one line.

\frac{D^2 \xi^\mu}{D\tau^2} = -R^\mu{}_{\nu\rho\sigma}\, u^\nu\, \xi^\rho\, u^\sigma \qquad (1)

where: $\xi^\mu=\partial x^\mu/\partial s$ — the deviation (connecting) vector between neighbouring geodesics; $u^\mu=\partial x^\mu/\partial\tau$ — the tangent four-velocity, obeying $u^\nu\nabla_\nu u^\mu=0$ ; $R^\mu{}_{\nu\rho\sigma}$ — the Riemann curvature tensor; $D/D\tau\equiv u^\nu\nabla_\nu$ — the covariant derivative along the worldline; $\tau$ — the affine parameter (proper time for timelike geodesics). Reading it: the right-hand side is a linear operator on $\xi$ , so neighbours accelerate apart or together in proportion to their separation and the local curvature — and not at all when $R=0$ . The operator $D/D\tau$ is the covariant directional derivative — it corrects the ordinary rate of change by the connection so the result transforms as a tensor, making "relative acceleration" a coordinate-independent statement.

\frac{D V^\mu}{D\tau} \equiv u^\nu \nabla_\nu V^\mu = \frac{\mathrm{d}V^\mu}{\mathrm{d}\tau} + \Gamma^\mu{}_{\nu\rho}\,u^\nu V^\rho \qquad (2)

where: $\Gamma^\mu{}_{\nu\rho}$ — the Christoffel symbols of the Levi-Civita connection; $\mathrm{d}/\mathrm{d}\tau$ — the ordinary derivative along the curve. Reading it: the $\Gamma$ term is the parallel-transport correction; drop it and you are back to non-tensorial coordinate differences a change of chart could invent or destroy. Because the tangent and the connecting vector come from a single coordinate net on the two-surface swept out by the congruence, their Lie bracket vanishes, and for a torsion-free connection this becomes an equality of covariant cross-derivatives.

u^\nu \nabla_\nu \xi^\mu = \xi^\nu \nabla_\nu u^\mu \qquad (3)

where: $[u,\xi]^\mu=u^\nu\partial_\nu\xi^\mu-\xi^\nu\partial_\nu u^\mu$ — the Lie bracket, zero because mixed partials commute. Reading it: the change of separation along the flow equals the change of the flow along the separation — the identity that lets the second derivative in $(1)$ collapse to a single curvature term. Contracting the Riemann tensor twice with the observer's four-velocity isolates the tidal tensor, whose eigenvalues are the principal stretches and squeezes felt in the rest frame.

E_{\mu\sigma}\equiv R_{\mu\nu\sigma\rho}\,u^\nu u^\rho,\qquad E_{\mu\sigma}=E_{\sigma\mu},\qquad E_{\mu\sigma}u^\sigma=0 \qquad (4)

where: $E_{\mu\sigma}$ — the tidal (electric Weyl) tensor measured by the observer with four-velocity $u$ . Reading it: symmetric (Riemann's pair-exchange symmetry) and purely spatial (antisymmetry in the last pair), so tides act only in the slice orthogonal to the worldline — where Newtonian $\partial_i\partial_j\Phi$ lives. The deviation equation is then $D^2\xi^\mu/D\tau^2=-E^\mu{}_\rho\,\xi^\rho$ . The trace of the tidal tensor is a single Ricci contraction — the doorway from geometry to matter, since the Einstein equations tie it to the local energy density.

\mathrm{tr}\,E = E^\mu{}_\mu = R_{\mu\nu}u^\mu u^\nu,\qquad \text{(vacuum: } R_{\mu\nu}=0 \Rightarrow \mathrm{tr}\,E=0) \qquad (5)

where: $R_{\mu\nu}=R^\lambda{}_{\mu\lambda\nu}$ — the Ricci tensor. Reading it: the trace controls whether a dust ball's volume changes — zero in vacuum (pure shear), but with matter $R_{\mu\nu}u^\mu u^\nu>0$ and the ball contracts, which is gravitational focusing. Tracing the deviation of an entire congruence yields the Raychaudhuri equation, the evolution law for the expansion scalar that underlies the singularity theorems.

\frac{\mathrm{d}\theta}{\mathrm{d}\tau} = -\tfrac13\theta^2 - \sigma_{\mu\nu}\sigma^{\mu\nu} + \omega_{\mu\nu}\omega^{\mu\nu} - R_{\mu\nu}u^\mu u^\nu \qquad (6)

where: $\theta=\nabla_\mu u^\mu$ — the expansion (fractional rate of volume change); $\sigma_{\mu\nu}$ — the shear; $\omega_{\mu\nu}$ — the vorticity. Reading it: for irrotational geodesics every term but vorticity is non-positive when $R_{\mu\nu}u^\mu u^\nu\ge0$ , so $\theta\to-\infty$ in finite affine time once convergence begins — caustics form inevitably. Finally, the weak-field, slow-motion limit identifies the relevant curvature component with the second derivatives of the Newtonian potential, closing the circle to elementary tidal forces.

g_{00}=-(1+2\Phi),\qquad R^i{}_{0j0}\approx \partial_i\partial_j\Phi \qquad (7)

where: $\Phi$ — the Newtonian gravitational potential, $|\Phi|\ll1$ in units $c=1$ . Reading it: the four-velocity is nearly $u^\mu\approx(1,0,0,0)$ and $\tau\approx t$ , so the deviation equation collapses to $\ddot\xi^i=-\partial_i\partial_j\Phi\,\xi^j$ , the textbook tidal tensor, with trace $\nabla^2\Phi=4\pi G\rho$ recovering the Poisson equation.

Derivation AFull derivation of the geodesic deviation equation

The goal is to descend from a smooth one-parameter family of geodesics to the curvature-driven relative acceleration $D^2\xi^\mu/D\tau^2=-R^\mu{}_{\nu\rho\sigma}u^\nu\xi^\rho u^\sigma$ , using nothing beyond the geodesic equation, the commutation of the two coordinate-derived vector fields, and the definition of the Riemann tensor as the obstruction to commuting covariant derivatives. Let $x^\mu(\tau,s)$ be a congruence whose fixed- $s$ curves are affinely parametrised geodesics, with $u^\mu=\partial x^\mu/\partial\tau$ and $\xi^\mu=\partial x^\mu/\partial s$ . The whole content is that after two covariant derivatives every non-curvature term cancels, leaving exactly one Riemann contraction.

Step 1. Impose the geodesic condition on each member, $u^\nu\nabla_\nu u^\mu=0$ ; the tangent is parallel-transported along itself. Why: this is the only dynamical input — it makes the leftover acceleration purely geometric, not a mix of geometry and applied force.

Step 2. Use the vanishing bracket: since $u$ and $\xi$ come from commuting partial derivatives, $[u,\xi]=0$ , and for a torsion-free connection $u^\nu\nabla_\nu\xi^\mu=\xi^\nu\nabla_\nu u^\mu$ . Why: it trades a derivative of $\xi$ along $u$ for one of $u$ along $\xi$ — the pivot of the calculation.

Step 3. Write the first covariant derivative of the separation, $\frac{D\xi^\mu}{D\tau}=u^\nu\nabla_\nu\xi^\mu=\xi^\nu\nabla_\nu u^\mu$ , using Step 2. Why: casting everything in terms of $\nabla u$ prepares the geodesic equation to cancel at second order.

Step 4. Differentiate again and apply the Leibniz rule,

\frac{D^2\xi^\mu}{D\tau^2}=u^\rho\nabla_\rho\big(\xi^\nu\nabla_\nu u^\mu\big)=(u^\rho\nabla_\rho\xi^\nu)\nabla_\nu u^\mu+u^\rho\xi^\nu\nabla_\rho\nabla_\nu u^\mu.

Why: these two pieces seed the cancellation — the first collides with a term hidden inside the second.

Step 5. Rewrite the first piece with Step 2 again, $(u^\rho\nabla_\rho\xi^\nu)\nabla_\nu u^\mu=(\xi^\rho\nabla_\rho u^\nu)\nabla_\nu u^\mu$ . Why: in this form it is ready to annihilate a matching term from Step 8.

Step 6. Reorder the double derivative in the second piece into a symmetric part plus the commutator,

u^\rho\xi^\nu\nabla_\rho\nabla_\nu u^\mu=u^\rho\xi^\nu\big(\nabla_\nu\nabla_\rho u^\mu+[\nabla_\rho,\nabla_\nu]u^\mu\big).

Why: the commutator is where the Riemann tensor enters; the symmetric remainder is where the geodesic equation strikes.

Step 7. Insert the definition of curvature, $[\nabla_\rho,\nabla_\nu]u^\mu=R^\mu{}_{\sigma\rho\nu}\,u^\sigma$ . Why: this single substitution is the origin of every $R$ in the answer — curvature as the failure of covariant derivatives to commute on $u$ .

Step 8. Attack the symmetric remainder with the geodesic equation,

u^\rho\xi^\nu\nabla_\nu\nabla_\rho u^\mu=\xi^\nu\nabla_\nu(u^\rho\nabla_\rho u^\mu)-(\xi^\nu\nabla_\nu u^\rho)\nabla_\rho u^\mu=-(\xi^\nu\nabla_\nu u^\rho)\nabla_\rho u^\mu.

The first term dies because $u^\rho\nabla_\rho u^\mu=0$ , and what remains exactly cancels Step 5. Why: the consistency miracle — all first-derivative-of- $u$ products vanish, leaving only curvature.

Step 9. Collect the survivor and fix the sign convention,

\frac{D^2\xi^\mu}{D\tau^2}=u^\rho\xi^\nu R^\mu{}_{\sigma\rho\nu}u^\sigma=-R^\mu{}_{\nu\rho\sigma}u^\nu\xi^\rho u^\sigma,

relabelling dummy indices and using the antisymmetry $R^\mu{}_{\sigma\rho\nu}=-R^\mu{}_{\sigma\nu\rho}$ . Why: that antisymmetry turns the raw expression into standard form and guarantees tides act orthogonal to $u$ . *The decisive check is the exact cancellation of the two non-curvature terms in Steps 5 and 8; this cancellation fails for any connection with torsion, confirming that the result is specific to the torsion-free Levi-Civita connection.*

Go deeper · Derivation B · 1 minNewtonian tidal tensor as a curvature limit

Derivation BNewtonian tidal tensor as a curvature limit

The goal is to show that the fully covariant deviation equation degenerates, in the weak static field of slow motion, into the elementary Newtonian tidal expression $\ddot\xi^i=-\partial_i\partial_j\Phi\,\xi^j$ , pinning the identification $R^i{}_{0j0}=\partial_i\partial_j\Phi$ that later fixes the coupling constant of the Einstein equations. Work in a nearly inertial frame with $g_{00}=-(1+2\Phi)$ , $g_{ij}=\delta_{ij}$ , and $u^\mu\approx(1,0,0,0)$ , $\tau\approx t$ ; every approximation below drops terms of second order in $\Phi$ or in the velocity.

Step 1. Compute the connection at first order, $\Gamma^i{}_{00}=-\tfrac12 g^{ij}\partial_j g_{00}=\partial_i\Phi$ ; only the spatial gradient survives for a static field. Why: this term already reproduces Newton's law $\ddot x^i=-\partial_i\Phi$ from the geodesic equation.

Step 2. Extract the curvature component, $R^i{}_{0j0}=\partial_j\Gamma^i{}_{00}-\partial_0\Gamma^i{}_{j0}+O(\Gamma^2)\approx\partial_i\partial_j\Phi$ ; time derivatives vanish (static) and quadratic-in- $\Gamma$ terms are second order in $\Phi$ . Why: this is the bridge — the curvature acting on the separation is literally the Hessian of the potential.

Step 3. Reduce the deviation equation with $u^\mu\approx\delta^\mu_0$ , so $D^2\xi^i/D\tau^2\approx-R^i{}_{0j0}\,\xi^j$ ; only $\nu=\sigma=0$ contributes and $\xi^j$ supplies the spatial index. Why: the four-index object collapses to a $3\times3$ matrix on the separation — the Newtonian tidal tensor's shape.

Step 4. Replace the covariant second derivative by the ordinary one, $D^2\xi^i/D\tau^2\approx\mathrm{d}^2\xi^i/\mathrm{d}t^2$ ; connection corrections are higher order. Why: only here may we speak of ordinary acceleration, while the covariant machinery silently keeps the result frame-independent.

Step 5. Assemble the identification, $\mathrm{d}^2\xi^i/\mathrm{d}t^2=-\partial_i\partial_j\Phi\,\xi^j$ , and take the trace: $\nabla^2\Phi=4\pi G\rho$ matches $R_{00}=4\pi G\rho$ via the Einstein equations. Why: this Newtonian limit plus a divergence-free field equation is one of the pillars selecting $G_{\mu\nu}=8\pi G\,T_{\mu\nu}$ . *The trace of the result reproduces the Poisson equation and hence fixes Einstein's constant $8\pi G$ , while the trace-free remainder is the genuine tidal shear.*

Limiting cases

The deviation equation spans the deepest geometry and the most elementary tidal physics; these regimes are the checks that keep it honest.

Case	Result	Meaning
Flat spacetime	$R^\mu{}_{\nu\rho\sigma}=0\;\Rightarrow\;\frac{D^2\xi^\mu}{D\tau^2}=0$	In Minkowski space neighbouring geodesics have no relative acceleration; any apparent acceleration is a coordinate artefact removed by passing to inertial Cartesian coordinates.
Newtonian weak field	$\frac{\mathrm{d}^2\xi^i}{\mathrm{d}t^2}=-\partial_i\partial_j\Phi\,\xi^j$	For slow motion and a weak static field $g_{00}=-(1+2\Phi)$ , the relevant curvature is $R^i{}_{0j0}\approx\partial_i\partial_j\Phi$ , reproducing the Newtonian tidal tensor exactly.
Point-mass tides	eigenvalues $(+2,-1,-1)\,GM/r^3$	Radial stretch is twice the transverse squeeze; trace zero. The $1/r^3$ scaling is why tides grow explosively on approach and dominate the physics near compact objects.
Vacuum / trace-free	$R_{\mu\nu}=0\;\Rightarrow\;\mathrm{tr}\,E=0$	The tidal tensor is traceless: a small dust ball is sheared into an ellipsoid at fixed volume to second order — pure shear, the hallmark of vacuum gravity including gravitational waves.
Matter present (Raychaudhuri)	$\frac{\mathrm{d}\theta}{\mathrm{d}\tau}\supset-R_{\mu\nu}u^\mu u^\nu=-4\pi G(\rho+3p)$	Contracting deviation gives the focusing term; positive energy density convergently focuses geodesics, the seed of the Penrose–Hawking singularity theorems.

Common trapConfusing the bending of one geodesic with the deviation of a family

Two traps compound. First, never use the ordinary second derivative $\mathrm{d}^2\xi^\mu/\mathrm{d}\tau^2$ as the relative acceleration — it is non-tensorial, and by choosing coordinates you can make it anything. Only the covariant $D^2\xi^\mu/D\tau^2$ is physical, and it equals a curvature contraction. Second, do not read curvature as the bending of a single geodesic: a single geodesic is locally straight by definition, since in its own local inertial frame the Christoffel symbols vanish at a point, so no measurement on one worldline detects gravity. Curvature lives only in the relationship between neighbouring geodesics — the second-order-in-separation tide that no change of coordinates can abolish. The equivalence principle removes the force from each worldline; it cannot remove the tide from the gap between them.

Worked examples

1 · Tidal stretch at the Earth's surface (fully worked)

Two point masses hover just above the Earth's surface, one directly above the other, separated vertically by $\ell=1.8\ \mathrm{m}$ (a standing person's height). Using $GM_\oplus=3.986\times10^{14}\ \mathrm{m^3\,s^{-2}}$ and $R_\oplus=6.371\times10^{6}\ \mathrm{m}$ , find the radial tidal acceleration between them, the transverse squeeze, and confirm the tidal tensor is trace-free.

Step 1. The base tidal rate is $GM/r^3=(3.986\times10^{14})/(6.371\times10^{6})^3$ ; with $(6.371\times10^{6})^3=2.585\times10^{20}\ \mathrm{m^3}$ this gives $GM/r^3=\mathbf{1.54\times10^{-6}\ s^{-2}}$ , the natural tidal frequency-squared at the surface.

Step 2. The radial eigenvalue is $+2GM/r^3=+3.08\times10^{-6}\ \mathrm{s^{-2}}$ (stretch); times the separation, the relative acceleration between head- and foot-height masses is $\Delta a_\parallel=(3.08\times10^{-6})(1.8)=\mathbf{5.5\times10^{-6}\ m/s^2}$ — the differential $\Delta g$ across a $1.8\ \mathrm{m}$ body, about half a millionth of $g$ .

Step 3. The transverse eigenvalue is $-GM/r^3=-1.54\times10^{-6}\ \mathrm{s^{-2}}$ (squeeze), half the radial value and opposite in sign; a horizontal $1.8\ \mathrm{m}$ pair converges at $\Delta a_\perp=2.8\times10^{-6}\ \mathrm{m/s^2}$ .

Step 4. Check the trace: $(+2)+(-1)+(-1)=0$ in units of $GM/r^3$ , so $\mathrm{tr}\,(\partial_i\partial_j\Phi)=\nabla^2\Phi=0$ in the vacuum outside the Earth — the ball stretches vertically exactly as much as it squeezes horizontally, at fixed volume. Radial stretch +3.08×10⁻⁶ s⁻², transverse squeeze −1.54×10⁻⁶ s⁻² each, trace = 0; a standing person is pulled apart top-to-bottom at 5.5×10⁻⁶ m/s².

2 · Spaghettification: stellar versus supermassive black holes (fully worked)

Estimate the radial tidal gradient at the event horizon of (a) a $1\,M_\odot$ black hole and (b) a $10^{8}\,M_\odot$ black hole, and decide in each case whether a $1.8\ \mathrm{m}$ human survives the crossing. Use $G=6.674\times10^{-11}$ , $c=2.998\times10^{8}$ , $M_\odot=1.989\times10^{30}\ \mathrm{kg}$ .

Step 1. The radial tidal gradient $2GM/r^3$ , evaluated at the horizon $r=r_s=2GM/c^2$ , becomes $2GM/r_s^3=c^6/(4G^2M^2)$ — so the horizon tide decreases with mass as $M^{-2}$ , a counter-intuitive fact.

Step 2. For $M=1\,M_\odot$ : $r_s=\mathbf{2.95\ km}$ and the gradient is $c^6/(4G^2M^2)=\mathbf{1.03\times10^{10}\ s^{-2}}$ . Across $\ell=1.8\ \mathrm{m}$ that is $1.85\times10^{10}\ \mathrm{m/s^2}\approx\mathbf{1.9\times10^{9}\,g}$ — a human is torn into a stream of atoms well before the horizon.

Step 3. For $M=10^{8}\,M_\odot$ the gradient scales as $M^{-2}$ , giving $(1.03\times10^{10})/(10^{8})^2=\mathbf{1.03\times10^{-6}\ s^{-2}}$ ; across $1.8\ \mathrm{m}$ that is $\approx\mathbf{1.9\times10^{-7}\,g}$ , gentler than Earth's surface tide, with the horizon out at $r_s\approx2\ \mathrm{AU}$ .

Step 4. So the stellar-mass horizon is lethal while an observer crosses the supermassive horizon feeling essentially nothing, to be shredded much later inside. Stellar-mass horizon: ≈1.9×10⁹ g (fatal); 10⁸ M_⊙ horizon: ≈1.9×10⁻⁷ g (imperceptible). Bigger black holes have gentler horizons because tidal gradient ∝ M⁻².

3 · Gravitational-wave strain at a detector

A gravitational wave of dimensionless strain amplitude $h\approx10^{-21}$ (the order of GW150914) passes through an interferometer with arm length $L=4\ \mathrm{km}$ . Estimate the change in arm length it induces.

Step 1. In the transverse-traceless gauge the deviation equation gives $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ , whose time-integrated effect is a fractional length change $\delta L/L=\tfrac12 h$ . The wave is a vacuum, trace-free tidal field, so it stretches one arm while squeezing the perpendicular one.

Step 2. Numerically, $\delta L=\tfrac12 h L=\tfrac12(10^{-21})(4\times10^{3})=\mathbf{2\times10^{-18}\ m}$ . That is about 1/400 of a proton radius (≈8.4×10⁻¹⁶ m) — the deviation equation, not any force on the mirrors, is what a kilometre-scale interferometer measures to detect a passing curvature ripple.

Physics of the portrait

Geodesic deviation is the moment differential geometry became a theory of gravity — sitting at the confluence of Riemann's curvature, Christoffel's connection, and the tensor calculus that made Einstein's theory expressible at all, with one figure building the bridge from abstract parallelism to physical tides.

Origins · 1 minTullio Levi-Civita and how this was discovered

Tullio Levi-Civita

1873–1941 · parallel transport and the geometry of curved spaces

Tullio Levi-Civita, a student of Gregorio Ricci-Curbastro at Padua, gave curvature a moving, physical meaning. In 1900 the pair published the absolute differential calculus — the tensor calculus Einstein would seize upon between 1912 and 1915 to write general relativity; the two corresponded, and Einstein said he admired Levi-Civita's work above almost any other. His decisive step came in 1917 with parallel transport: a rule for sliding a vector along a curve while keeping it "as parallel as possible", which turned the Christoffel symbols from a computational nuisance into a genuine geometric connection and made the covariant derivative intuitive. Curvature could now be defined operationally — as the failure of a vector to return to itself after transport around a closed loop — precisely the static content that the geodesic deviation equation recasts dynamically. In the 1920s Levi-Civita and, independently, J. L. Synge developed the deviazione geodetica itself, closing the loop between parallel transport and the measurable tidal field. Dismissed from his Rome chair in 1938 under the Fascist racial laws, he died isolated in 1941, but the Levi-Civita connection remains the default geometry of every general-relativity calculation.

History

1854

— Riemann's habilitation lecture defines the curvature of an n-dimensional manifold

1917

— Levi-Civita introduces parallel transport, giving the connection its geometric meaning

1926

— Levi-Civita and Synge cast the relative acceleration of neighbouring geodesics as the deviation equation

1955

— Raychaudhuri derives the focusing equation for a geodesic congruence

2015

— LIGO detects gravitational waves as an oscillating tidal deviation of freely-falling test masses

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §1.6, §11.3, §13.5 · Wald, General Relativity (1984), §3.3 and §9.2 · Carroll, Spacetime and Geometry (2004), §3.3 and §4.1 · Stone & Goldbart, Mathematics for Physics (2009), geometry of curved spaces · Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., tensor analysis · Hobson, Efstathiou & Lasenby, General Relativity (2006), Ch. 7. Key cross-checks: the deviation equation follows from commuting covariant derivatives with $[u,\xi]=0$ , with both non-curvature terms cancelling; the Newtonian limit gives $R^i{}_{0j0}=\partial_i\partial_j\Phi$ from $g_{00}=-(1+2\Phi)$ ; the point-mass tidal eigenvalues $(+2,-1,-1)\,GM/r^3$ follow from the Hessian of $\Phi=-GM/r$ and match the Schwarzschild orthonormal-frame components; and the anchor numbers are Earth $GM/r^3=1.54\times10^{-6}\ \mathrm{s^{-2}}$ , solar-mass horizon gradient $1.03\times10^{10}\ \mathrm{s^{-2}}$ at $r_s=2.95\ \mathrm{km}$ , and LIGO $\delta L=2\times10^{-18}\ \mathrm{m}$ for $h=10^{-21}$ , $L=4\ \mathrm{km}$ .

Prerequisites. Parallel Transport and Geodesics · The Covariant Derivative · Christoffel Symbols and Affine Connections · Covariant and Contravariant Tensors · The Metric Tensor. Leads to. Tensor Contraction and the Trace · Raising and Lowering Indices · Manifolds and Tangent Spaces.

Ask the tutor: Ask me to work through the tidal tensor for a specific spacetime — Schwarzschild, a gravitational wave, or FRW cosmology — and watch the same deviation equation produce spaghettification, detector strain, or cosmic focusing.

Isn't gravity supposed to be abolished for a free-falling observer? Then what is a tidal force?

The equivalence principle abolishes gravity only locally — at a single point, to first order. In a small enough free-falling laboratory you feel nothing and a single test particle floats. But "small enough" is the key phrase: over a finite region, neighbouring free-fallers accelerate relative to one another, and that relative acceleration is the tidal force. It is second order in the separation, so it shrinks as you shrink the lab, which is why a point observer misses it. What you cannot do is transform it away everywhere at once, because it is set by curvature — derivatives of the connection that no coordinate choice can zero out over an extended region.

What is the difference between the Riemann tensor and the tidal tensor?

The Riemann tensor $R^\mu{}_{\nu\rho\sigma}$ is the full curvature — a four-index object carrying all the geometry at a point, valid for any observer. The tidal tensor $E_{\mu\sigma}=R_{\mu\nu\sigma\rho}u^\nu u^\rho$ is what a particular observer with four-velocity $u$ measures: the Riemann tensor contracted twice with that velocity, leaving a symmetric, purely spatial $3\times3$ matrix whose eigenvalues are the principal stretches and squeezes in the rest frame. Different observers slice the same Riemann tensor differently, so they disagree on the tidal tensor just as they disagree on which directions are "space". The Riemann tensor is invariant; the tidal tensor is its shadow in one rest frame.

Why does the volume of a dust ball stay fixed in vacuum but shrink when matter is present?

The rate of volume change is set by the trace of the tidal tensor, $R_{\mu\nu}u^\mu u^\nu$ — the Ricci contraction. In vacuum the Einstein equations force $R_{\mu\nu}=0$ , so the trace vanishes: the ball shears into an ellipsoid, stretching and squeezing by exactly compensating amounts at fixed volume to second order. With ordinary matter, $R_{\mu\nu}u^\mu u^\nu=4\pi G(\rho+3p)>0$ and the ball contracts. This convergence is gravitational focusing — the physical reason a cloud of dust collapses under its own gravity rather than merely distorting.

How is geodesic deviation related to the Raychaudhuri equation and singularity theorems?

Raychaudhuri's equation comes from tracing the deviation of an entire congruence rather than one connecting vector. It evolves the expansion scalar $\theta$ — the fractional rate of change of cross-sectional volume — with a self-focusing $-\tfrac13\theta^2$ term, shear, vorticity, and the Ricci focusing term $-R_{\mu\nu}u^\mu u^\nu$ . For irrotational geodesics obeying an energy condition, every contribution drives $\theta$ downward, so once convergence starts it runs away to $\theta\to-\infty$ in finite affine time — the geodesics cross at a caustic. Penrose and Hawking turned this guaranteed focusing into theorems proving singularities are generic in gravitational collapse and cosmology.

Why does the tidal tensor act only in the directions perpendicular to the observer's motion?

Because the Riemann tensor is antisymmetric in its last pair of indices. Forming $E_{\mu\sigma}u^\sigma=R_{\mu\nu\sigma\rho}u^\nu u^\sigma u^\rho$ contracts the antisymmetric pair $(\sigma\rho)$ with two copies of the same vector $u$ , which necessarily vanishes. So the tidal tensor has no component along the four-velocity: it lives entirely in the spatial slice orthogonal to the worldline. This matches the Newtonian picture, where $\partial_i\partial_j\Phi$ has only spatial indices and acts within the observer's instantaneous rest space — there is no "tidal force in the time direction".

Does the deviation equation also apply to light rays, not just massive particles?

Yes, with one adjustment. Photons follow null geodesics, for which proper time is unavailable, so you use an affine parameter $\lambda$ and the equation reads $D^2\xi^\mu/D\lambda^2=-R^\mu{}_{\nu\rho\sigma}k^\nu\xi^\rho k^\sigma$ with $k^\mu$ the null tangent. Tracing the deviation of a bundle of rays gives the optical scalar equations, whose expansion and shear are exactly the convergence and shear of gravitational lensing. When a distant galaxy's image is magnified and sheared by intervening mass, that is geodesic deviation of a null congruence — the same equation that stretches LIGO's arms distorts the lensed sky.