LINEARIZED GRAVITY

In June 1916, seven months after he published the field equations, sent a paper to the Prussian Academy with a striking claim: his theory of gravity predicted waves. Just as Maxwell's equations had yielded electromagnetic waves traveling at the speed of light, the field equations — properly approximated — yielded gravitational waves traveling at the same speed. Einstein worked out the radiated power to lowest order and found it depended on the third time-derivative of a mass distribution's quadrupole moment.

It was a remarkable result, and Einstein half-distrusted it from the start. The waves were absurdly weak. The coupling constant $8\pi G/c^4$ in the field equations is roughly $2 \times 10^{-43}$ in SI units, so any laboratory source radiates a vanishing trickle of gravitational energy. Worse, the mathematics was treacherous. Because general relativity lets you change coordinates freely, a "wave" might be nothing more than a wobble in the coordinate grid — a mathematical artifact with no physical content. Distinguishing real ripples in spacetime from coordinate noise would occupy theorists for forty years.

The doubt nearly became official. In 1936 Einstein, then at Princeton with his assistant Nathan Rosen, submitted a paper to Physical Review titled "Do Gravitational Waves Exist?" — and his answer was no. The manuscript argued that exact plane gravitational waves contained singularities and could not be physical. The journal sent it to an anonymous referee (the cosmologist Howard Percy Robertson), who found an error: the singularities were coordinate artifacts, removable by a change of variables. Einstein, unused to peer review, withdrew the paper in anger. But Robertson was right. After Leopold Infeld relayed the correction, Einstein reversed himself, and the rewritten 1937 paper — now titled "On Gravitational Waves" — concluded that cylindrical gravitational waves do exist. The episode is the cleanest illustration of the central difficulty: in general relativity, telling physics from bookkeeping is genuinely hard. This topic builds the machinery that settles it.

The trick is to abandon the full nonlinear theory and study its small-amplitude limit. Suppose spacetime is nearly flat — the geometry of an empty room, perturbed by a faint ripple passing through. Then the Metric tensor $g_{\mu\nu}$ can be written as the flat Minkowski metric $\eta_{\mu\nu} = \mathrm{diag}(-1, 1, 1, 1)$ plus a small correction:

$g_{\mu\nu} = \eta_{\mu\nu} + h_{\mu\nu}, \qquad |h_{\mu\nu}| \ll 1$

In words: write the curved metric as the flat background $\eta$ plus a tiny perturbation $h$, and agree to keep only terms linear in $h$, throwing away everything quadratic or higher. The perturbation $h_{\mu\nu}$ is what we will call the gravitational field — it lives on top of flat spacetime, like a sound wave on top of still air.

Substituting this into the Einstein field equations and discarding every term with two or more powers of $h$ turns the formidable nonlinear $G_{\mu\nu} = (8\pi G/c^4)\, T_{\mu\nu}$ into a linear equation. It is convenient to package $h_{\mu\nu}$ together with its trace into the "trace-reversed" perturbation $\bar h_{\mu\nu} = h_{\mu\nu} - \tfrac{1}{2}\eta_{\mu\nu} h$, where $h = \eta^{\mu\nu} h_{\mu\nu}$. In terms of $\bar h$, the linearized field equations collapse to something a sophomore would recognize:

$\Box\, \bar h_{\mu\nu} = -\frac{16\pi G}{c^4}\, T_{\mu\nu}$

The symbol $\Box = -\tfrac{1}{c^2}\partial_t^2 + \nabla^2$ is the d'Alembertian — the flat-spacetime wave operator. This single line says: the gravitational field obeys a wave equation, sourced by matter, with the same operator that governs sound, light, and a vibrating string. In vacuum ($T_{\mu\nu} = 0$) it reduces to $\Box\, \bar h_{\mu\nu} = 0$, whose solutions are waves traveling at exactly the speed of light.

The picture above is the entire conceptual content of linearized gravity in one image. Flat spacetime is the stage; the wave is a small, transverse undulation of distances that sweeps across it. Nothing about the stage moves — it is the relationships between points that ripple.

Two features of the solution are worth stating precisely. First, the wave is transverse: for a wave propagating along the $z$-axis, the perturbation lives entirely in the $x$–$y$ plane perpendicular to the direction of travel, exactly as light's electric and magnetic fields are perpendicular to its motion. Second, the wave is massless and therefore non-dispersive: every frequency travels at $c$, so a sharp pulse stays sharp as it propagates. This is the dispersion relation built into $\Box \bar h_{\mu\nu} = 0$,

$\omega = c\,|\mathbf{k}|$

which simply says the angular frequency $\omega$ equals the speed of light times the wavenumber $|\mathbf{k}|$ — the signature of any disturbance that moves at $c$ without spreading. A gravitational wave from a billion light-years away arrives with its waveform intact, which is precisely why detectors can match it against a precomputed template.

We have used the word gauge twice now without earning it. Here is the idea. In general relativity you are free to relabel the points of spacetime — an infinitesimal coordinate change $x^\mu \to x^\mu + \xi^\mu$ shifts the perturbation by

$h_{\mu\nu} \to h_{\mu\nu} - \partial_\mu \xi_\nu - \partial_\nu \xi_\mu$

This says: a different choice of coordinates $\xi^\mu$ changes the numbers in $h_{\mu\nu}$ without changing any physical fact, exactly as adding $\partial_\mu \chi$ to the electromagnetic potential $A_\mu$ leaves the fields $\mathbf{E}$ and $\mathbf{B}$ untouched. The freedom to make this change is gauge freedom, and it is the source of all the confusion that nearly led Einstein to disown his own waves: some of the apparent motion in $h_{\mu\nu}$ is real tidal stretching, and some is just the coordinate grid sloshing around.

The cure is to spend the gauge freedom buying simplicity. Of the ten components of the symmetric $h_{\mu\nu}$, gauge transformations let us eliminate eight, leaving exactly two physical degrees of freedom. The standard choice that does this in vacuum is the Transverse-traceless gauge — the TT gauge — defined by three conditions: the perturbation is purely spatial ($h_{0\mu} = 0$), divergence-free or transverse ($\partial^i h_{ij} = 0$), and traceless ($h^i{}_i = 0$). In the TT gauge the coordinates ride along with freely-falling test masses, so every wobble that remains is genuine physics.

That the physics survives any coordinate choice is not merely reassuring — it is what makes a detection meaningful. When LIGO records a strain, it is measuring the gauge-invariant tidal effect, the relative change in distance between mirrors, which is exactly the quantity the TT gauge isolates. The sticky-bead argument of 1957 — due to Hermann Bondi and Richard Feynman, who pointed out that a passing wave would slide beads along a rod against friction and so deposit heat — finally convinced the community that gravitational waves carry real, extractable energy. They are not bookkeeping. They are physics.

Linearized gravity was not discovered by accident; it was built in the image of electromagnetism, step for step. The strategy is worth seeing laid out explicitly, because every move in the gravitational derivation has an electromagnetic ancestor that physicists already trusted.

The wave equation EQ.02 makes the parallel exact. Compare the electromagnetic wave equation in Lorenz gauge,

$\Box\, A^\mu = -\mu_0 J^\mu \qquad \longleftrightarrow \qquad \Box\, \bar h_{\mu\nu} = -\frac{16\pi G}{c^4}\, T_{\mu\nu}$

The two equations are the same equation with a richer index structure on the right: the photon's source is the four-current $J^\mu$ (one index), while the graviton's source is the stress-energy tensor $T_{\mu\nu}$ (two indices). That extra index is the whole story. Electromagnetism's source is a vector, so its lowest radiating multipole is the dipole, and its field carries spin-1. Gravity's source is a rank-2 tensor, so the dipole term is forbidden by momentum conservation and the lowest radiating multipole is the quadrupole; correspondingly the gravitational field carries spin-2. The spin sets the polarization geometry, and the geometry is what a ring of test masses reveals.

Where the analogy breaks is just as instructive. Electromagnetism is exactly linear: photons do not carry charge, so light waves pass through one another untouched. Gravity is not — the full field equations are nonlinear because gravitational energy itself gravitates, and a gravitational wave is, at second order, a source of further curvature. Linearized gravity is the regime where this self-interaction is negligible. It is a first chapter, not the whole book, but it is the chapter that turns Einstein's geometry into something a detector can hear.

We have reduced Einstein's nonlinear theory, in its weak-field limit, to a wave equation indistinguishable in form from Maxwell's — solutions traveling at $c$, transverse, massless, carrying two polarizations and real energy. That is the foundation. Everything in gravitational-wave astronomy is built on top of it.

The immediate next question is geometric: what does the wave actually do to matter? The two polarizations of the TT perturbation deform a ring of free test masses in two distinct patterns — the "plus" mode stretching and squeezing along the coordinate axes, the "cross" mode doing the same along the diagonals, the two related by a 45° rotation that is the unmistakable fingerprint of a spin-2 field. That is the subject of polarization modes, where the same quadrupole structure that forbids dipole radiation also dictates the shape of the deformation. From there the story becomes astrophysical: a binary inspiral is a quadrupole whose moment changes as two compact objects spiral together, radiating away orbital energy until they merge — the chirp that LIGO recorded on September 14, 2015.

Linearized gravity is also the bridge back to the full theory. The energy a wave carries, computed by averaging the second-order perturbation, is precisely what drains a binary's orbit and what the Hulse–Taylor pulsar confirmed to better than a fraction of a percent across three decades of timing. Einstein doubted in 1936 that the waves were real. The sticky bead, the binary pulsar, and finally a strain of $10^{-21}$ measured across four kilometers of vacuum proved otherwise. The faint ripple on flat spacetime turned out to be one of the most precisely tested predictions in all of physics.