Geodesic Deviation and Tidal Action of Waves

A gravitational wave is a ripple in the metric, but a single freely falling mass cannot feel it: by the equivalence principle that mass follows a geodesic and rides the wave without any local sign of its passage. The physical, coordinate-independent content of the wave shows up only in the relative motion of two or more neighbouring free masses — the tidal stretch-and-squeeze that the wave imprints across a finite separation. That relative acceleration is exactly what the geodesic deviation equation computes, with the Riemann curvature tensor playing the role of the tidal-force operator. This article is where linearized gravitational-wave theory finally touches an instrument: the two preceding results — the wave equation for the perturbation $h_{\mu\nu}$ and the transverse-traceless (TT) gauge that isolates its two physical polarizations — are turned here into a measurable motion of matter. Two pictures of the same physics trip up newcomers, and reconciling them is half the work. In TT gauge the coordinates are pinned to the free masses themselves, so their coordinate positions never change while the metric — and therefore the proper distance between them — oscillates. In a local proper (Fermi normal) frame anchored to one mass, the coordinates are rigid and the second mass genuinely accelerates back and forth. Both are correct, linked by a gauge transformation, and both deliver the same proper-distance oscillation set by the single gauge-invariant curvature component $R_{i0j0} = -\tfrac12\ddot h^{TT}_{ij}$ . To leading order the tidal acceleration is $\ddot\xi^i = \tfrac12\ddot h^{TT}_{ij}\,\xi^j$ , an explicit, calculable function of the strain and its polarization. Applied to a ring of test masses it produces the quadrupolar breathing that is the real-space fingerprint of a spin-2 field — and the response a kilometre-scale interferometer is built to read.

Key takeaways

A single free mass detects nothing; a gravitational wave reveals itself only through the relative acceleration of separated free masses, because the gauge-invariant physical content is curvature, not the metric perturbation itself.
The geodesic deviation (Jacobi) equation $\frac{D^2\xi^\mu}{\mathrm{d}\tau^2}=-R^\mu{}_{\alpha\nu\beta}u^\alpha\xi^\nu u^\beta$ governs how a separation vector accelerates under the Riemann tensor.
For a weak wave the tidal field reduces to $\ddot\xi^i=-R^i{}_{0j0}\xi^j=\tfrac12\ddot h^{TT}_{ij}\xi^j$ — the acceleration is set by the second time derivative of the TT strain.
Integrated, this deforms separations as $\xi^i(t)\approx[\delta^i_j+\tfrac12 h^{TT}_{ij}(t)]\,\xi^j(0)$ : a plus wave stretches a ring along $x$ while squeezing along $y$ , the cross wave the same rotated by 45°.
The fractional length change is $\delta L/L=\tfrac12 h$ ; for GW150914 ( $h\sim10^{-21}$ , $L=4$ km) the differential arm change is $\Delta L=hL\sim4\times10^{-18}$ m — a thousandth of a proton radius.
Key idea: a gravitational wave is detected not by the motion of one mass but by the tidal Riemann curvature $R^i{}_{0j0}=-\tfrac12\ddot h^{TT}_{ij}$ acting between masses, producing a relative acceleration $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ — a gauge-invariant, quadrupolar stretching of spacetime.

Where this lives today

Geodesic deviation is no longer theory — it is the read-out of a running observatory network. Since GW150914 in 2015, the LIGO–Virgo–KAGRA detectors have logged roughly a hundred compact-binary mergers, each one a direct measurement of the tidal strain $\tfrac12 h^{TT}_{ij}$ acting between suspended mirrors, with peak strains around $10^{-21}$ and arm changes of $10^{-18}$ m tracked on the interferometer's dark fringe. Because the deviation pattern is purely quadrupolar, a network with differently oriented arms can decompose a signal into $h_+$ and $h_\times$ and test for the extra scalar "breathing" or vector modes that alternative theories of gravity predict — so far, consistent with pure spin-2 general relativity. At nanohertz frequencies, pulsar timing arrays (NANOGrav, EPTA, PPTA) turned the same physics inside out: a wave perturbs the proper distance along each Earth–pulsar line, and in 2023 these collaborations reported evidence for a stochastic background whose inter-pulsar correlations follow the Hellings–Downs curve — the angular signature of quadrupolar geodesic deviation projected onto the sky. Looking ahead, the space-based LISA mission (launch planned for the mid-2030s) will read the deviation across millions of kilometres in the millihertz band, and next-generation ground detectors (Einstein Telescope, Cosmic Explorer) aim for strains near $10^{-24}$ . Every one of these instruments is, at heart, a precision assay of $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ .

The physical picture

Picture a cloud of free dust grains adrift in deep space, each on its own geodesic. No grain feels a thing — drop an accelerometer on any one of them and it reads zero. But watch their relative positions and the wave becomes visible: the cloud breathes, elongating along one transverse axis while contracting along the perpendicular one, then swapping a half-period later. A gravitational wave is a pure tidal field. It has no monopole part (that would violate mass conservation) and no dipole part (that would violate momentum conservation); the leading, and lowest, moment it can carry is the quadrupole, and geodesic deviation renders that quadrupole as a stretch-and-squeeze in the plane transverse to propagation. The first figure freezes that pattern for the plus and cross polarizations.

Figure

A ring of freely falling test masses deformed by a wave travelling into the page. The plus wave stretches x while squeezing y; the cross wave does the same rotated by 45°. To linear order each mass moves as ξₓ = (1 + ½h₊)x₀ and ξ_y = (1 − ½h₊)y₀, then reverses each half-period — a purely quadrupolar, spin-2 deformation

Tidal forces are the gravitational analogue of the difference between two nearby field lines. A uniform field accelerates everything equally and is locally undetectable — that is the equivalence principle. Only the variation of the field across a finite separation, the curvature, produces relative acceleration. This is why the tidal acceleration in $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ carries a second time derivative and a factor $\tfrac12$ : a single derivative of $h$ can be set to zero by a coordinate choice along a geodesic (it is gauge), but two derivatives assemble the gauge-invariant Riemann tensor that no coordinate transformation can remove. That is the whole reason an interferometer works, and the second figure puts the real numbers on it.

Figure

A Michelson interferometer reading the tide. A plus wave lengthens the x-arm and shortens the y-arm by δL = ½hL each. For LIGO, L = 4 km and peak strain h ≈ 1×10⁻²¹ give δLₓ = +2×10⁻¹⁸ m, δL_y = −2×10⁻¹⁸ m, and a differential ΔL = hL = 4×10⁻¹⁸ m — about a thousandth of a proton radius, read off the dark fringe

The core equations

The starting point is the fully covariant law for how the separation between two nearby geodesics evolves. It is the exact statement that curvature — and only curvature — drives relative acceleration; everything specific to gravitational waves is obtained by evaluating its right-hand side for a weak TT perturbation.

\frac{D^2 \xi^\mu}{\mathrm{d}\tau^2} = -R^\mu{}_{\alpha\nu\beta}\,u^\alpha\,\xi^\nu\,u^\beta \qquad (1)

where: $\xi^\mu$ — separation (connecting) vector between two neighbouring geodesics; $u^\alpha=\mathrm{d}x^\alpha/\mathrm{d}\tau$ — four-velocity tangent to the fiducial geodesic; $R^\mu{}_{\alpha\nu\beta}$ — Riemann curvature tensor, the tidal-force operator; $D/\mathrm{d}\tau=u^\alpha\nabla_\alpha$ — covariant derivative along the geodesic; $\tau$ — proper time. Reading: the acceleration of the separation is linear in $\xi$ and linear in the curvature, so it vanishes identically in flat spacetime no matter how the masses are moving. For slowly moving masses the four-velocity is essentially $u^\alpha\approx(1,0,0,0)$ and the covariant derivative reduces to an ordinary time derivative. Only the "electric" components $R^i{}_{0j0}$ survive, and equation (1) collapses to a Newtonian-looking tidal equation.

\frac{\mathrm{d}^2\xi^i}{\mathrm{d}t^2} = -R^i{}_{0j0}\,\xi^j \qquad (2)

where: $R^i{}_{0j0}$ — the relativistic tidal tensor (electric part of Riemann); $\xi^i$ — spatial separation of the masses; $t\approx\tau$ — coordinate time, equal to proper time for slow masses to the order kept. Reading: a $3\times3$ tidal matrix acts on the separation; its eigenvalues give the stretch/squeeze rates along the principal axes. Now specialize the curvature to a weak wave. Computing the linearized Riemann tensor in TT gauge (Derivation B) leaves a single surviving term — the second time derivative of the strain.

R^i{}_{0j0} = -\tfrac12\,\partial_0\partial_0\,h^{TT}_{ij} = -\tfrac12\,\ddot h^{TT}_{ij} \qquad (3)

where: $h^{TT}_{ij}$ — transverse-traceless metric perturbation, the physical strain; $\ddot h^{TT}_{ij}$ — its second time derivative; the factor $\tfrac12$ comes directly from the linearized Riemann formula. Reading: the tidal field an interferometer sees is not $h$ , nor $\dot h$ , but $\ddot h$ — pure gauge-invariant curvature. Substituting (3) into (2) gives the master relation of the whole subject: the wave's strain acting as a tidal acceleration per unit separation.

\frac{\mathrm{d}^2\xi^i}{\mathrm{d}t^2} = \tfrac12\,\ddot h^{TT}_{ij}\,\xi^j \qquad (4)

where: $\xi^j$ — separation, held at its unperturbed value $\xi^j(0)$ on the right since corrections are second order in $h$ ; $\tfrac12\ddot h^{TT}_{ij}$ — tidal acceleration per unit length. Reading: each free mass is driven by an effective force per unit mass $\tfrac12\ddot h^{TT}_{ij}\xi^j$ , quadrupolar and transverse. Because $h$ is tiny, the right side is nearly constant in $\xi$ , so equation (4) integrates twice to a clean deformation of the separation vector — the form used to draw the deforming ring.

\xi^i(t) \approx \xi^j(0)\left[\delta^i_j + \tfrac12\, h^{TT}_{ij}(t)\right] \qquad (5)

where: $\xi^j(0)$ — initial separation before the wave arrives; $\delta^i_j$ — undisturbed separation (Kronecker delta); $\tfrac12 h^{TT}_{ij}$ — fractional strain matrix. Reading: the separation is stretched or squeezed by half the strain — the origin of the observable $\delta L/L=\tfrac12 h$ . Projected onto one interferometer arm this is the number the instrument reports; the trace-free, quadrupolar structure of $h^{TT}_{ij}$ is what makes the two perpendicular arms respond oppositely. Alongside it, the low-velocity, static limit of (2) recovers ordinary Newtonian tides, anchoring the relativistic result to familiar physics.

\frac{\delta L}{L} = \tfrac12\,h_+ \qquad\text{and}\qquad R^i{}_{0j0}\ \xrightarrow{\ \text{static}\ }\ \partial_i\partial_j\Phi \qquad (6)

where: $\delta L/L$ — fractional change in a detector arm's proper length; $h_+$ — plus-polarization strain amplitude along that arm; $\Phi$ — Newtonian potential, whose tidal tensor $\partial_i\partial_j\Phi$ has trace $\nabla^2\Phi=4\pi G\rho$ . Reading: the wave contributes a traceless tidal tensor (no local source, $\rho=0$ ), which is exactly why the deformation stretches one axis by as much as it squeezes the other.

Derivation AThe geodesic deviation equation from a family of geodesics

The goal is to derive the covariant law $\frac{D^2\xi^\mu}{\mathrm{d}\tau^2}=-R^\mu{}_{\alpha\nu\beta}u^\alpha\xi^\nu u^\beta$ from first principles, showing that the Riemann tensor is not inserted by hand but forced out by taking two covariant derivatives of a separation vector between neighbouring free-fall trajectories. The strategy is to set up a smooth two-parameter family of geodesics, exploit the fact that the tangent and connecting vectors come from commuting coordinate flows, and let the non-commutativity of covariant differentiation produce the curvature. This is the rigorous backbone behind the tidal picture and it is completely independent of any weak-field assumption.

Step 1. Consider a one-parameter family of geodesics $x^\mu(\tau,s)$ , where $\tau$ labels proper time along each curve and $s$ labels which curve. Define the tangent $u^\mu=\partial x^\mu/\partial\tau$ and the connecting (separation) vector $\xi^\mu=\partial x^\mu/\partial s$ . Because these are coordinate basis vectors of the $(\tau,s)$ surface, their Lie bracket vanishes, which for a torsion-free connection gives $u^\nu\nabla_\nu\xi^\mu=\xi^\nu\nabla_\nu u^\mu$ . Why: equal mixed partials $\partial_\tau\partial_s=\partial_s\partial_\tau$ become the statement that $u$ and $\xi$ "commute", the identity that lets us swap the order of differentiation later.

Step 2. Compute the second covariant derivative of $\xi$ along the fiducial geodesic: $\frac{D^2\xi^\mu}{\mathrm{d}\tau^2}=u^\alpha\nabla_\alpha(u^\beta\nabla_\beta\xi^\mu)$ . Use Step 1 to rewrite the inner factor as $\xi^\beta\nabla_\beta u^\mu$ , giving $u^\alpha\nabla_\alpha(\xi^\beta\nabla_\beta u^\mu)$ . Why: we are asking how fast the separation's rate of change itself changes — an acceleration of separation — and the commuting-flows identity converts $\tau$ -derivatives of $\xi$ into $s$ -derivatives of $u$ , where the geodesic condition can act.

Step 3. Expand with the Leibniz rule and commute the two covariant derivatives. The commutator of covariant derivatives acting on a vector is the curvature, $[\nabla_\alpha,\nabla_\beta]u^\mu=R^\mu{}_{\nu\alpha\beta}u^\nu$ , while the geodesic equation $u^\alpha\nabla_\alpha u^\mu=0$ annihilates the term in which both derivatives fall on $u$ along its own direction. Why: this is the decisive step — the physical effect survives precisely because covariant derivatives do not commute, and the leftover piece is the Riemann tensor, not a coordinate artifact.

Step 4. Collecting the surviving term and using the antisymmetry $R^\mu{}_{\nu\alpha\beta}=-R^\mu{}_{\nu\beta\alpha}$ to fix the index order yields $\frac{D^2\xi^\mu}{\mathrm{d}\tau^2}=-R^\mu{}_{\alpha\nu\beta}u^\alpha\xi^\nu u^\beta$ , equation (1). Why: the result is exact and geometric — two masses in free fall accelerate toward or away from each other at a rate set entirely by the curvature they sit in, with no reference to any force.

Go deeper · Derivation B · 2 minThe tidal field of a wave from the linearized Riemann tensor

Derivation BThe tidal field of a wave from the linearized Riemann tensor

The aim is to show that for a transverse-traceless gravitational wave the tidal tensor $R^i{}_{0j0}$ collapses to the strikingly simple $-\tfrac12\ddot h^{TT}_{ij}$ , and hence that the master deviation law is $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ . The route is to write the linearized Riemann tensor in terms of second derivatives of the perturbation, substitute the defining conditions of TT gauge, and watch three of the four terms vanish. This computation shows why the surviving factor of one-half and the appearance of the second time derivative are the load-bearing facts of gravitational-wave detection.

Step 1. To first order in $h_{\mu\nu}$ the Riemann tensor is $R_{\mu\nu\alpha\beta}=\tfrac12(\partial_\nu\partial_\alpha h_{\mu\beta}+\partial_\mu\partial_\beta h_{\nu\alpha}-\partial_\mu\partial_\alpha h_{\nu\beta}-\partial_\nu\partial_\beta h_{\mu\alpha})$ , which is manifestly gauge-invariant at this order. Setting the indices to $(\mu,\nu,\alpha,\beta)=(i,0,j,0)$ gives $R_{i0j0}=\tfrac12(\partial_0\partial_j h_{i0}+\partial_i\partial_0 h_{0j}-\partial_i\partial_j h_{00}-\partial_0\partial_0 h_{ij})$ . Why: the electric component $R_{i0j0}$ is the piece that appears in the slow-motion deviation equation, so it is the only combination of curvature we need for a laboratory detector.

Step 2. Impose the transverse-traceless gauge conditions $h^{TT}_{0\mu}=0$ (no time-time or time-space components) and $h^{TT}_{ij}$ purely spatial, transverse, and traceless. The first three terms each carry a factor $h_{0\mu}$ or $h_{00}$ and therefore vanish identically, leaving only the last term. Why: TT gauge is chosen precisely to strip $h$ down to its two physical, radiative degrees of freedom, and here that choice does the algebra for us — the tidal field can depend only on the physical strain.

Step 3. The survivor is $R_{i0j0}=-\tfrac12\,\partial_0\partial_0 h^{TT}_{ij}=-\tfrac12\ddot h^{TT}_{ij}$ ; raising the spatial index with the flat metric $\eta$ leaves the numerical value unchanged to this order, so $R^i{}_{0j0}=-\tfrac12\ddot h^{TT}_{ij}$ . Why: for a wave propagating along $z$ the spatial derivatives that might have entered $R_{i0j0}$ do not appear in this component, so only the time derivative survives — the tidal field oscillates at the wave frequency, in step with $\ddot h$ .

Step 4. Insert this into the slow-motion deviation equation $\ddot\xi^i=-R^i{}_{0j0}\xi^j$ to obtain the master formula $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ , and integrate twice (holding $\xi^j\approx\xi^j(0)$ , since corrections are $\mathcal{O}(h^2)$ ) to get $\xi^i(t)\approx[\delta^i_j+\tfrac12 h^{TT}_{ij}(t)]\xi^j(0)$ . Why: the two integrations turn an acceleration into the directly observable fractional deformation $\tfrac12 h$ , closing the gap between the abstract strain and a moving mirror.

Go deeper · Derivation C · 2 minFrozen coordinates, moving masses: proper distance and gauge invariance

Derivation CFrozen coordinates, moving masses: proper distance and gauge invariance

The objective is to dissolve the most persistent confusion in the subject — that in TT gauge "the masses do not move, so nothing happens" — by computing the physically measurable proper distance directly, and then showing that the alternative Fermi-normal frame, in which the masses visibly accelerate, gives the identical answer. The strategy is to place two free masses at fixed TT coordinates, integrate the oscillating spatial metric to get their proper separation, and then perform the gauge transformation to the local proper frame and confirm the same relative acceleration emerges. The lesson is that only the gauge-invariant Riemann tensor decides what is real, and both bookkeeping schemes agree on it.

Step 1. In TT gauge a mass initially at rest stays at fixed coordinates, because the relevant Christoffel symbol $\Gamma^i_{00}=\tfrac12(2\partial_0 h^{TT}_{0i}-\partial_i h^{TT}_{00})$ vanishes when $h^{TT}_{0\mu}=0$ ; so the geodesic equation gives $\mathrm{d}^2x^i/\mathrm{d}\tau^2=0$ . Two masses sit at $x=0$ and $x=L$ for all time. Why: this is the true content of "coordinates are frozen" — it is a statement about the coordinate labels, which are glued to the masses by construction, not about physical distance.

Step 2. The spatial metric, however, oscillates: with a plus wave $h^{TT}_{xx}=h_+\cos\omega t$ , one has $g_{xx}=1+h_+\cos\omega t$ . The proper distance between the masses is $s=\int_0^L\sqrt{g_{xx}}\,\mathrm{d}x=\int_0^L\sqrt{1+h_+\cos\omega t}\,\mathrm{d}x$ , which to linear order is $s\approx L\,[1+\tfrac12 h_+\cos\omega t]$ . Why: proper distance — light travel time between the masses — is what a ruler or a laser beam measures, and it genuinely breathes with amplitude $\tfrac12 h_+ L$ even though the coordinate distance $L$ never changes.

Step 3. Now change to a local proper (Fermi normal) frame about one mass via the infinitesimal transformation $x^i_{\rm proper}=x^i_{TT}+\tfrac12 h^{TT}_{ij}(t)\,x^j_{TT}$ , which makes the metric flat at the origin. The second mass now sits at $\xi^i_{\rm proper}(t)=[\delta^i_j+\tfrac12 h^{TT}_{ij}(t)]\xi^j(0)$ , and differentiating twice gives $\ddot\xi^i_{\rm proper}=\tfrac12\ddot h^{TT}_{ij}\xi^j$ . Why: in this frame the coordinates are rigid, so the whole effect shows up as a genuine acceleration of the second mass — the "masses really move" picture — and it matches the master formula exactly.

Step 4. Both frames share the same gauge-invariant curvature $R^i{}_{0j0}=-\tfrac12\ddot h^{TT}_{ij}$ , so both predict the same relative acceleration and the same proper-distance oscillation $s\approx L(1+\tfrac12 h_+\cos\omega t)$ . The Fermi construction is valid within a region small compared with a wavelength, where higher curvature corrections are negligible. Why: agreement on the Riemann tensor is the guarantee that the physics is coordinate-independent — the two descriptions are the same phenomenon told in two bookkeepings, and neither is "more real".

Limiting cases

The tidal-action formula specializes cleanly into the regimes one actually uses:

Case	Result	Meaning
Plus polarization, ring of masses	$\ddot\xi^x=\tfrac12\ddot h_+\,\xi^x,\ \ddot\xi^y=-\tfrac12\ddot h_+\,\xi^y$	A circle deforms into an ellipse oscillating between horizontal and vertical — the classic "+" pattern.
Cross polarization	$\ddot\xi^x=\tfrac12\ddot h_\times\,\xi^y,\ \ddot\xi^y=\tfrac12\ddot h_\times\,\xi^x$	The same ellipse rotated by 45° — diagonal stretch and squeeze.
Static / DC limit	$\ddot h^{TT}_{ij}\to0\ \Rightarrow\ \ddot\xi^i\to0$	A constant, non-oscillating $h$ produces no relative acceleration; only time variation of the wave drives masses.
Newtonian tidal limit	$R^i{}_{0j0}\to\partial_i\partial_j\Phi$	A slow, weak, static source reduces the deviation equation to the Newtonian tidal tensor with $\nabla^2\Phi=4\pi G\rho$ .
Along the propagation axis ( $z$ )	$\ddot\xi^z=0$	The wave is transverse: masses separated along the direction of travel feel no tidal force.

Common trapIn TT gauge the coordinates do not change, so a gravitational wave has no physical effect

This is the single most common error, and it is exactly backwards. In TT gauge the coordinates are defined to ride along with the free masses, so their coordinate positions are frozen by construction — but that is a statement about labels, not about physics. The spatial metric $g_{ij}=\delta_{ij}+h^{TT}_{ij}$ oscillates, so the proper distance between the masses, $s\approx L(1+\tfrac12 h)$ , genuinely changes; a laser beam bouncing between them takes a measurably different round-trip time. In the complementary Fermi-normal frame the coordinates are rigid and the masses visibly accelerate — same physics, different bookkeeping. What settles the matter is the gauge-invariant curvature $R^i{}_{0j0}=-\tfrac12\ddot h^{TT}_{ij}$ , which is nonzero in every frame. Confusing "coordinates are static" with "nothing happens" would predict that LIGO cannot work; it does.

Worked examples

1 · The smallest length LIGO must measure (fully worked)

A plus-polarized wave with peak strain $h_+=1.0\times10^{-21}$ and frequency $f=100$ Hz ( $\omega=2\pi f\approx628\ \mathrm{s}^{-1}$ ) passes a detector whose arms have proper length $L=4.0$ km, one along $x$ , one along $y$ . Find the peak change in each arm, the differential signal, and the peak tidal acceleration of a mirror.

Step 1. From equation (5), each arm's fractional change is $\delta L/L=\tfrac12 h_+$ . The $x$ -arm lengthens and the $y$ -arm shortens by the same magnitude: $\delta L_x=+\tfrac12 h_+ L=+\tfrac12(10^{-21})(4.0\times10^{3}\ \mathrm{m})=+2.0\times10^{-18}\ \mathrm{m}$ , and $\delta L_y=-2.0\times10^{-18}\ \mathrm{m}$ . Each arm moves by 2×10⁻¹⁸ m — about one five-hundredth of a proton radius.

Step 2. A Michelson interferometer reads the difference of the two arm lengths, so the physical signal is $\Delta L=\delta L_x-\delta L_y=h_+ L=(10^{-21})(4.0\times10^{3}\ \mathrm{m})=\mathbf{4.0\times10^{-18}\ m}$ . This factor-of-two gain from differencing is the whole reason the geometry is an "L": the quadrupolar, traceless strain guarantees the arms move oppositely, so their difference doubles the signal while common-mode noise cancels.

Step 3. The peak tidal acceleration of a mirror at $\xi=L$ follows from the master formula (4). With $h^{TT}$ oscillating as $\cos\omega t$ , $\ddot h_+=-\omega^2 h_+\cos\omega t$ has peak magnitude $\omega^2 h_+$ , so $|\ddot\xi_x|_{\rm peak}=\tfrac12\omega^2 h_+ L=\tfrac12(628)^2(10^{-21})(4.0\times10^{3})\approx\tfrac12(3.94\times10^{5})(4.0\times10^{-18})\approx\mathbf{7.9\times10^{-13}\ m/s^2}$ . A mirror is pushed at under a picometre per second squared — utterly negligible as a force, yet coherent over thousands of cycles.

Step 4. Sanity check on the displacement: integrating a sinusoidal acceleration of peak $\tfrac12\omega^2 h_+ L$ gives a displacement amplitude $\tfrac12 h_+ L=2.0\times10^{-18}$ m, matching Step 1 — the two routes (integrated strain vs. integrated acceleration) agree, as they must. δL = 2×10⁻¹⁸ m, ΔL = 4×10⁻¹⁸ m, peak mirror acceleration ≈ 8×10⁻¹³ m/s².

2 · A ring of test masses under a passing wave

A circular ring of free test masses of radius $R_0=1.0$ m lies in the $xy$ -plane, transverse to a plus wave of peak strain $h_+=1.0\times10^{-21}$ . Find the semi-axes of the ellipse at peak strain and the shape at a quarter and half period.

Step 1. From equation (5) the mass initially at angle $\theta$ , position $(R_0\cos\theta,\,R_0\sin\theta)$ , moves to $x=(1+\tfrac12 h_+\cos\omega t)R_0\cos\theta$ and $y=(1-\tfrac12 h_+\cos\omega t)R_0\sin\theta$ . The ring stays an ellipse with semi-axes $a_x=R_0(1+\tfrac12 h_+\cos\omega t)$ and $a_y=R_0(1-\tfrac12 h_+\cos\omega t)$ .

Step 2. At $\omega t=0$ (peak): $a_x=1.0\,(1+5.0\times10^{-22})$ m and $a_y=1.0\,(1-5.0\times10^{-22})$ m. The ring is stretched along x and squeezed along y by ±5×10⁻²² m — half a zeptometre on a one-metre ring.

Step 3. At $\omega t=\pi/2$ : $\cos\omega t=0$ , so $a_x=a_y=R_0$ — the ring is momentarily a perfect circle again. At $\omega t=\pi$ : the deformation reverses, $a_x=R_0(1-\tfrac12 h_+)$ and $a_y=R_0(1+\tfrac12 h_+)$ — now stretched along $y$ , squeezed along $x$ .

Step 4. Note the area is preserved to first order: $a_x a_y=R_0^2(1-\tfrac14 h_+^2\cos^2\omega t)\approx R_0^2$ . The ring breathes between a horizontal and a vertical ellipse, area-preserving, returning to a circle twice per cycle — the traceless, quadrupolar signature of a spin-2 wave.

3 · Why a wave along the arm produces nothing

The same wave now travels along $x$ , along one of the arms, instead of along $z$ . What differential signal does the detector register?

Step 1. A wave propagating along $x$ has $h^{TT}_{ij}$ supported in the transverse $yz$ -plane; its $xx$ -component vanishes, $h^{TT}_{xx}=0$ . By the master formula the $x$ -arm feels $\ddot\xi^x=\tfrac12\ddot h^{TT}_{xj}\xi^j=0$ . The arm lying along the propagation direction does not respond at all.

Step 2. The $y$ -arm can respond (it is transverse to this wave), so the differential signal is generally nonzero but reduced, and it depends on the wave's polarization angle about $x$ . The lesson is that a detector's response is a function of source direction and polarization — the antenna pattern — which is why a network of detectors with different orientations can triangulate a source and separate $h_+$ from $h_\times$ . A wave down the arm is invisible to that arm; sky localization and polarimetry both trace back to this directional tidal response.

Physics of the portrait

The tidal action of a gravitational wave is where general relativity stopped being a debate about coordinates and became an experimental science. For four decades after Einstein's 1916 prediction, physicists genuinely argued whether gravitational waves were real or a mathematical illusion of a bad coordinate choice — and the argument was settled by the realization that geodesic deviation, the tidal Riemann tensor acting between masses, is what a wave physically does.

Origins · 1 minHermann Bondi and how this was discovered

Hermann Bondi

1919–2005 · the sticky-bead argument and the physical reality of gravitational waves

Hermann Bondi, an Austrian-British mathematical physicist and cosmologist, is remembered here for helping end the "do gravitational waves exist?" controversy that had smouldered since 1916. Einstein himself had wavered — in 1936 he and Nathan Rosen submitted a paper arguing waves did not exist, before a referee (Howard Percy Robertson) caught the error. The decisive conceptual step came from Felix Pirani, who in 1956 showed that the physically meaningful, coordinate-independent effect of a wave is the geodesic deviation it drives through the Riemann tensor — reframing the whole question in terms of what happens to real matter. At the 1957 Chapel Hill conference (GR1), Richard Feynman sharpened this into the "sticky bead" argument: imagine beads threaded on a rigid rod, free to slide against friction. A passing wave drives the beads back and forth relative to the rod via exactly the tidal deviation $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ ; the friction generates heat, so the wave demonstrably does work and therefore carries energy. Bondi championed and published this argument, and with Pirani and Ivor Robinson constructed exact plane-wave solutions (1959) that carry genuine energy-momentum, cementing the physical reality of radiation. His broader legacy — the Bondi mass, the Bondi–Sachs formalism for radiation at null infinity — grew from taking that reality seriously. The line from Pirani's tidal insight through Bondi's sticky bead runs straight to LIGO: a detector is a sticky-bead experiment executed with laser light and hanging mirrors.

History

1916

— Einstein predicts gravitational waves from the linearized field equations

1956

— Pirani shows the Riemann tensor gives the gauge-invariant, observable effect of a wave

1957

— the Chapel Hill sticky-bead argument (Feynman, Bondi) shows the waves do real work

1959

— Bondi, Pirani and Robinson give exact plane-wave solutions carrying energy

2015

— LIGO records GW150914, measuring a geodesic-deviation strain of h ≈ 10⁻²¹

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §§1.6, 11.1–11.4, 35.4 (geodesic deviation and wave detection) · Wald, General Relativity (1984), §4.4 · Carroll, Spacetime and Geometry (2004), §3.3, §7.4–7.5 · Hartle, Gravity (2003), §16.2, §21.4 · Schutz, A First Course in General Relativity, 2nd ed. (2009), §9.5 · Maggiore, Gravitational Waves, Vol. 1 (2008), Ch. 1. The linearized Riemann tensor in TT gauge yields $R^i{}_{0j0}=-\tfrac12\ddot h^{TT}_{ij}$ and the master formula $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ ; the factor $\tfrac12$ and the vanishing of the $h_{0\mu}$ and $h_{00}$ terms follow directly from imposing the TT conditions, and the proper-distance integral gives $s\approx L(1+\tfrac12 h)$ . The numerical estimate $\Delta L\sim4\times10^{-18}$ m for $h\sim10^{-21}$ , $L=4$ km matches published LIGO design figures; the peak mirror acceleration $\tfrac12\omega^2 h L\approx8\times10^{-13}$ m/s² at 100 Hz follows directly from equation (4).

Prerequisites. Transverse-Traceless Gauge · Gravitational Wave Equation · Plus and Cross Polarizations · Linearized Einstein Equations. Leads to. Isaacson Stress-Energy of Gravitational Waves · Gravitational-Wave Luminosity and Energy Flux · Helicity and the Spin-2 Graviton · The Quadrupole Formula.

Ask the tutor: Ask me to walk through why the interferometer arms move oppositely, or to work the antenna pattern for a wave arriving from any sky direction.

If a freely falling mass feels nothing, how can any detector work?

Because a detector never relies on one mass — it compares two (or more) separated masses. The equivalence principle guarantees that a single free mass follows a geodesic and reads zero on any local accelerometer; the wave's effect is spacetime curvature, which by definition shows up only across a finite separation as relative acceleration. LIGO's mirrors are two such free masses hung far apart; the laser measures how their proper separation breathes. In sticky-bead language, one bead tells you nothing, but two beads on a rod slide relative to each other and heat it. The physical content of a gravitational wave is irreducibly a two-point (tidal) quantity, which is exactly what geodesic deviation computes.

Why does the tidal force depend on the second derivative of h, not on h itself?

Because relative acceleration is governed by the Riemann tensor, and in TT gauge $R^i{}_{0j0}=-\tfrac12\ddot h^{TT}_{ij}$ — two time derivatives of the strain. The deeper reason is gauge invariance: the value of $h$ at a point, and even its first derivative along a geodesic, can be changed or set to zero by a coordinate transformation, so neither can drive a coordinate-independent physical effect. It takes two derivatives to build the Riemann tensor, the lowest-order gauge-invariant object, and that is what physically accelerates the masses. So the observable tidal field oscillates as $\ddot h$ , in step with — not lagging — the strain, and it vanishes for any static, non-oscillating $h$ .

In TT gauge the coordinates don't move — is anything actually happening?

Yes, emphatically. "Coordinates don't move" is true but misleading: in TT gauge the coordinate grid is deliberately attached to the free masses, so their coordinate labels are constant by construction. What changes is the metric, and therefore the proper distance, $s\approx L(1+\tfrac12 h)$ — the quantity a ruler or a light beam actually measures. If you instead work in a local proper (Fermi) frame, the coordinates are rigid and the masses visibly accelerate. Both descriptions are correct and are related by a gauge transformation; they agree on the one thing that is coordinate-independent, the curvature $R^i{}_{0j0}=-\tfrac12\ddot h^{TT}_{ij}$ , which is nonzero. Mistaking frozen coordinates for "no effect" would wrongly predict that interferometers cannot detect anything.

Are the TT and Fermi-frame pictures really the same physics?

They are — that is the whole point of gauge invariance. In TT gauge the masses hold fixed coordinates while the metric oscillates, so the proper distance changes. In a Fermi normal frame anchored to one mass, the metric is flat at the origin and the second mass physically accelerates as $\ddot\xi^i=\tfrac12\ddot h^{TT}_{ij}\xi^j$ . A single infinitesimal coordinate transformation, $x^i_{\rm proper}=x^i_{TT}+\tfrac12 h^{TT}_{ij}x^j_{TT}$ , maps one into the other, and both yield the identical proper-distance oscillation and the identical Riemann tensor. Which picture is more convenient depends on the problem: TT gauge for propagation and radiation calculations, the Fermi frame for intuition about "forces" on a laboratory-scale detector, whose size is far smaller than a wavelength.

Why is the deformation quadrupolar, and what does that reveal about the wave?

A gravitational wave cannot carry a monopole moment (that would require the source's mass to oscillate, violating mass conservation) or a dipole moment (that would require the centre of mass or total momentum to oscillate, both forbidden by conservation laws). The lowest allowed radiating moment is the quadrupole, and the tidal pattern of geodesic deviation reflects exactly that: a traceless, area-preserving stretch-and-squeeze. Under a 180° rotation about the propagation axis the pattern maps onto itself, which is the real-space fingerprint of a helicity- $\pm2$ , spin-2 field. Detecting any monopole "breathing" mode or dipole mode would signal physics beyond general relativity — which is precisely what detector networks search for.

Does the ring of masses actually change size, or is it just the ruler changing?

It changes size in every operationally meaningful sense. If you measure the ring with a light beam — timing a photon's round trip — the travel time oscillates as $\tfrac12 h$ , because the proper distance does. A rigid material ruler is more subtle: its atoms are held by electromagnetic forces, so a stiff ruler partly resists the tide and reports a smaller effect, while the free masses follow geodesics and deform fully. This is why detectors use freely suspended mirrors, not a rigid frame: you want the test masses to track spacetime, not fight it. The unambiguous, gauge-invariant statement is that the proper distance between free masses oscillates, and a light-based measurement reads that oscillation directly.

How does this relativistic tidal effect connect to ordinary Newtonian tides?

Smoothly. In the slow-motion, weak-field, static limit the tidal tensor $R^i{}_{0j0}$ reduces to the Newtonian tidal tensor $\partial_i\partial_j\Phi$ , and the geodesic deviation equation becomes the familiar $\ddot\xi^i=-\partial_i\partial_j\Phi\,\xi^j$ that stretches a falling body along the radial direction toward a mass. The crucial difference is the trace: a Newtonian tide near matter has $\nabla^2\Phi=4\pi G\rho\neq0$ , but a gravitational wave in vacuum has $\rho=0$ , so its tidal tensor is traceless — one axis stretches by exactly as much as the perpendicular axis squeezes, preserving area. That tracelessness is the vacuum-radiation, spin-2 fingerprint that distinguishes a passing wave from the static tide of a nearby mass.