The Trace-Reversed Metric Perturbation

Linearized general relativity begins with a deceptively simple move: write the metric as $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$ with $|h_{\mu\nu}|\ll1$ and keep only terms linear in the perturbation. The trouble is that the linearized Einstein tensor, written directly in $h_{\mu\nu}$ , is a sprawling sum of six second-derivative terms — d'Alembertians, double divergences, and gradients of the trace all tangled together. Nothing about it says "wave". The trace-reversed metric perturbation is the single change of variables that untangles it. Define $\bar h_{\mu\nu}\equiv h_{\mu\nu}-\tfrac12\eta_{\mu\nu}h$ , subtracting a precise slice of the trace, and every explicit trace term is absorbed. What remains is one d'Alembertian plus three terms sharing a common factor: the divergence $\partial^\mu\bar h_{\mu\nu}$ . That surviving object is exactly what the residual coordinate freedom of linearized gravity can set to zero. Imposing the Lorenz gauge $\partial^\mu\bar h_{\mu\nu}=0$ kills the last three terms, and the field equations collapse to $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ — a decoupled, flat-space wave equation, one per component, sourced by the stress-energy tensor. This is the electromagnetism of gravity: $\bar h_{\mu\nu}$ is the analogue of the vector potential $A_\mu$ , the Lorenz gauge mirrors $\partial^\mu A_\mu=0$ , and the wave equation mirrors $\Box A_\mu=-\mu_0 J_\mu$ line for line. From it flow the Newtonian limit, the retarded potentials, the quadrupole formula, and the two gravitational-wave polarizations — the whole edifice of weak-field gravity resting on the sign of one $\tfrac12$ .

Key takeaways

The trace-reversed perturbation $\bar h_{\mu\nu}=h_{\mu\nu}-\tfrac12\eta_{\mu\nu}h$ is a linear repackaging of the weak-field metric that turns the linearized Einstein tensor into a single d'Alembertian.
Its name is literal: the trace flips sign, $\bar h=-h$ , so the operation is an involution — trace-reversing twice returns $h_{\mu\nu}$ .
In the Lorenz (harmonic) gauge $\partial^\mu\bar h_{\mu\nu}=0$ , the Einstein equations become the flat-space wave equation $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ , one decoupled equation per component.
The retarded solution $\bar h_{\mu\nu}=4G\!\int T_{\mu\nu}(t_{\rm ret})/|\mathbf{x}-\mathbf{x}'|\,d^3x'$ is the gravitational analogue of the electromagnetic retarded potential, and leads to the quadrupole radiation formula.
The static $00$ -component reproduces Poisson's equation $\nabla^2\Phi=4\pi G\rho$ with $h_{00}=-2\Phi$ ; the vacuum equation $\Box\bar h_{\mu\nu}=0$ carries the two gravitational-wave polarizations $h_+,h_\times$ .
Key idea: trace-reversing the perturbation sweeps every inconvenient trace and divergence in the linearized Einstein tensor into one gauge-removable term; in Lorenz gauge the field equations become a decoupled flat-space wave equation for each component of $\bar h_{\mu\nu}$ .

The core insight

Trace-reversing the perturbation moves all the inconvenient trace and divergence terms in the linearized Einstein tensor into a single gauge-removable divergence; in Lorenz gauge the field equations become a decoupled flat-space wave equation for each component of $\bar h_{\mu\nu}$ .

Where this lives today

The trace-reversed wave equation is the working engine of gravitational-wave astronomy. Every waveform template that LIGO, Virgo, and KAGRA match against their data is built on the post-Newtonian expansion of $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ , carried to high order in $v/c$ to track a chirping binary through thousands of cycles. Since GW150914 the field has catalogued dozens of mergers, including GW170817, whose electromagnetic counterpart was found because the same quadrupole formula pins down distance and sky position. The retarded integral (7) underlies numerical relativity's wave extraction, and the pulsar-timing arrays that reported a nanohertz background in 2023 analyze the same trace-reversed strain across the galaxy. Extended to include $T_{0i}$ , it yields the gravitomagnetic potentials whose frame-dragging Gravity Probe B measured in 2011. From the Solar System to black-hole mergers, the physics is one wave equation for one trace-reversed field.

The physical picture

A symmetric perturbation $h_{\mu\nu}$ carries two kinds of information: a "trace part" like a single scalar spread over the diagonal, and "trace-free" shear-like parts that squeeze the geometry one way and stretch it another. The linearized Einstein tensor mixes these, and the source $T_{\mu\nu}$ has its own trace. Trace reversal lines the two sides up, moving the scalar clutter out of the differential operator and into the definition of the field. The first figure shows this on the perturbation everyone knows — the Newtonian metric, where $h_{\mu\nu}$ is diagonal with every entry $-2\Phi$ — which collapses onto a single component.

Figure

Trace reversal acting on the Newtonian perturbation: the full metric perturbation h_{μν} = diag(−2Φ, −2Φ, −2Φ, −2Φ) with trace h = −4Φ collapses under h̄_{μν} = h_{μν} − ½ η_{μν} h onto a single component h̄_{μν} = diag(−4Φ, 0, 0, 0), and its trace flips sign to h̄ = +4Φ = −h

The deeper payoff is an exact analogy with electromagnetism. There the potential $A_\mu$ obeys $\Box A_\mu=-\mu_0 J_\mu$ once $\partial^\mu A_\mu=0$ is imposed, and retarded solutions give the field of any moving charge and, far away, radiation. Trace reversal makes gravity wear the same clothes: $\bar h_{\mu\nu}$ is the "gravitational potential", $\partial^\mu\bar h_{\mu\nu}=0$ its gauge, and $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ its sourced wave equation, with $T_{\mu\nu}$ as $J_\mu$ . The second figure lays the two theories side by side and follows the gravitational one to its famous consequence: a wave whose strain, reaching Earth from a black-hole merger, is a number LIGO actually measured.

Figure

The gravitational wave equation mirrors electromagnetism line for line: □ h̄_{μν} = −16πG T_{μν} beside □ A_μ = −μ₀ J_μ, both solved by retarded potentials; radiated outward, the trace-reversed field gives the strain LIGO recorded for GW150914, h ≈ 1×10⁻²¹ at a distance of 410 Mpc

The core equations

The definition subtracts a weighted slice of the trace from the perturbation. The factor $\tfrac12$ is not a convention but the unique number that makes the construction work in four dimensions, as the next equation shows.

\bar h_{\mu\nu}\equiv h_{\mu\nu}-\tfrac12\,\eta_{\mu\nu}\,h \qquad (1)

where: $h_{\mu\nu}$ — metric perturbation, $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$ with $|h_{\mu\nu}|\ll1$ ; $\eta_{\mu\nu}$ — Minkowski metric, signature $(-,+,+,+)$ ; $h$ — the trace $h=\eta^{\mu\nu}h_{\mu\nu}$ ; $\bar h_{\mu\nu}$ — the trace-reversed perturbation. Read it as: keep $h_{\mu\nu}$ , bleed a trace piece off the diagonal. Contracting with $\eta^{\mu\nu}$ shows why the name is literal: the metric contracts with itself to $4$ in four dimensions, and $4\cdot\tfrac12-1=1$ flips the trace's sign.

\bar h\equiv\eta^{\mu\nu}\bar h_{\mu\nu}=h-\tfrac12(4)h=-h \qquad (2)

where: $\bar h$ — trace of the reversed perturbation; $4=\delta^\mu_\mu=\eta^{\mu\nu}\eta_{\mu\nu}$ counts the dimensions. Because the trace merely changes sign, the operation is its own inverse. Since trace-reversing flips the trace, applying the recipe again recovers the original — an inverse that lets us solve for $\bar h_{\mu\nu}$ and read off the physical metric $h_{\mu\nu}$ .

h_{\mu\nu}=\bar h_{\mu\nu}-\tfrac12\,\eta_{\mu\nu}\,\bar h \qquad (3)

where: $\bar h=-h$ . Identical in shape to (1) — the mark of an involution. The payoff appears when the linearized Einstein tensor is written in the new variable: every explicit factor of the trace $h$ vanishes, leaving a d'Alembertian plus three terms that each contain the same divergence $\partial^\alpha\bar h_{\cdot\alpha}$ .

G^{(1)}_{\mu\nu}=-\tfrac12\Big(\Box\bar h_{\mu\nu}+\eta_{\mu\nu}\,\partial^\alpha\partial^\beta\bar h_{\alpha\beta}-\partial_\nu\partial^\alpha\bar h_{\mu\alpha}-\partial_\mu\partial^\alpha\bar h_{\nu\alpha}\Big) \qquad (4)

where: $\Box=\eta^{\alpha\beta}\partial_\alpha\partial_\beta=-\partial_t^2+\nabla^2$ — the flat-space d'Alembertian; $\partial^\alpha\bar h_{\mu\alpha}$ — the divergence of $\bar h$ . The last three terms are the only obstruction to a clean wave equation. Linearized gravity keeps a gauge freedom: infinitesimal coordinate changes $x^\mu\to x^\mu+\xi^\mu$ shift $\bar h_{\mu\nu}$ without changing the physics — just enough to zero the troublesome divergence, the Lorenz (or harmonic, or de Donder) gauge.

\partial^\mu\bar h_{\mu\nu}=0 \qquad (5)

where: $\partial^\mu=\eta^{\mu\alpha}\partial_\alpha$ . This is the direct analogue of the electromagnetic $\partial^\mu A_\mu=0$ , always reachable, as Derivation 2 proves. Imposing (5) annihilates the last three terms, so $G^{(1)}_{\mu\nu}=-\tfrac12\Box\bar h_{\mu\nu}$ , and $G_{\mu\nu}=8\pi G\,T_{\mu\nu}$ becomes the master equation of weak-field gravity.

\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu} \qquad (6)

where: $T_{\mu\nu}$ — the stress-energy tensor (source); $G$ — Newton's constant; units $c=1$ . Each of the ten components obeys a flat-space wave equation, decoupled from the rest. Being linear, (6) is solved by the retarded integral of the source over the past light cone — the twin of the electromagnetic potential.

\bar h_{\mu\nu}(t,\mathbf{x})=4G\int\frac{T_{\mu\nu}\big(t-|\mathbf{x}-\mathbf{x}'|,\,\mathbf{x}'\big)}{|\mathbf{x}-\mathbf{x}'|}\,d^3x' \qquad (7)

where: $t-|\mathbf{x}-\mathbf{x}'|$ — the retarded time ( $c=1$ ); the coefficient $4G$ follows from the Green's function of $\Box$ . Signals propagate outward at the speed of light. Far from a slow source, the transverse-traceless part of the retarded solution is the radiation field, and only the second derivative of the mass quadrupole survives — gravity has no monopole or dipole radiation, so the quadrupole leads.

\bar h^{\,TT}_{ij}=\frac{2G}{r}\,\ddot Q^{\,TT}_{ij} \qquad (8)

where: $Q_{ij}$ — the mass quadrupole moment; $r$ — distance to the observer; $\ddot Q$ — its second time derivative at retarded time. This is the quadrupole formula, the basis of gravitational-wave astronomy.

Derivation AFrom the linearized Einstein tensor to a single wave equation

The goal is to show how the trace-reversed variable converts the unwieldy linearized Einstein tensor into one d'Alembertian, turning Einstein's equations into a flat-space wave equation. We start from the linearized Ricci tensor, assemble the Einstein tensor, substitute $h_{\mu\nu}=\bar h_{\mu\nu}-\tfrac12\eta_{\mu\nu}\bar h$ to watch every trace term cancel, and impose the Lorenz gauge to clear the residual divergences. That cancellation is the whole content of "trace reversal": the $\tfrac12$ is chosen precisely so the trace clutter disappears.

Step 1. Linearize the Ricci tensor. Expanding $R_{\mu\nu}=\partial_\alpha\Gamma^\alpha_{\mu\nu}-\partial_\nu\Gamma^\alpha_{\mu\alpha}+\Gamma\Gamma$ to first order (the $\Gamma\Gamma$ products are second order) gives $R^{(1)}_{\mu\nu}=\tfrac12(\partial^\alpha\partial_\mu h_{\nu\alpha}+\partial^\alpha\partial_\nu h_{\mu\alpha}-\Box h_{\mu\nu}-\partial_\mu\partial_\nu h)$ . Why: the connection is already first order in $h$ , so only the derivative-of-connection pieces of Riemann contribute.

Step 2. Build the Einstein tensor. Contract to the Ricci scalar $R^{(1)}=\partial^\alpha\partial^\beta h_{\alpha\beta}-\Box h$ , then form $G^{(1)}_{\mu\nu}=R^{(1)}_{\mu\nu}-\tfrac12\eta_{\mu\nu}R^{(1)}$ . Why: the field equations set the Einstein tensor to $8\pi G\,T_{\mu\nu}$ , and only it obeys the Bianchi identity guaranteeing conservation.

Step 3. Substitute the trace-reversed variable. Replace every $h_{\mu\nu}$ using $h_{\mu\nu}=\bar h_{\mu\nu}-\tfrac12\eta_{\mu\nu}\bar h$ ; all explicit-trace terms cancel, collapsing $G^{(1)}_{\mu\nu}$ to $-\tfrac12(\Box\bar h_{\mu\nu}+\eta_{\mu\nu}\partial^\alpha\partial^\beta\bar h_{\alpha\beta}-\partial_\nu\partial^\alpha\bar h_{\mu\alpha}-\partial_\mu\partial^\alpha\bar h_{\nu\alpha})$ . Why: this cancellation is the entire reason the variable was defined — the $\tfrac12$ was tuned so the trace terms annihilate rather than accumulate.

Step 4. Impose the Lorenz gauge. Each of the last three terms contains $\partial^\alpha\bar h_{\cdot\alpha}$ ; setting $\partial^\alpha\bar h_{\mu\alpha}=0$ kills all three at once, leaving $G^{(1)}_{\mu\nu}=-\tfrac12\Box\bar h_{\mu\nu}$ . Why: the divergence is what the residual coordinate freedom can freely adjust, so it is exactly what a gauge choice should eliminate.

Step 5. Read off the field equation. Setting $-\tfrac12\Box\bar h_{\mu\nu}=8\pi G\,T_{\mu\nu}$ gives $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ . Why: a decoupled inhomogeneous wave equation, one per component, whose retarded solution is immediate. *As a check: tracing both sides of the field equation gives $-\Box h=-16\pi G\,T$ , consistent with $R^{(1)}=-8\pi G\,T$ from $G^{(1)}=8\pi G\,T$ , confirming the cancellation and the factor $16\pi G$ .*

Go deeper · Derivation B · 1 minReaching the Lorenz gauge with residual coordinate freedom

Derivation BReaching the Lorenz gauge with residual coordinate freedom

The goal is to prove that the Lorenz condition $\partial^\mu\bar h_{\mu\nu}=0$ can always be imposed, so the previous derivation is never obstructed. The key is that an infinitesimal coordinate change acts on $\bar h_{\mu\nu}$ like a gauge transformation, and its effect on the divergence is governed by a wave equation for the generator $\xi^\mu$ that always has a solution. We track how $h_{\mu\nu}$ , $\bar h_{\mu\nu}$ , and its divergence transform, and solve for the $\xi^\mu$ that zeroes it.

Step 1. Transform the perturbation. Under $x^\mu\to x^\mu+\xi^\mu$ with $\xi^\mu$ small, the perturbation shifts by $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ , the linearized Lie derivative along $\xi$ . Why: this is the residual gauge freedom — coordinate changes of the same order as $h$ , leaving $\eta_{\mu\nu}$ intact but reshuffling $h_{\mu\nu}$ .

Step 2. Transform the trace-reversed variable. Trace-reversing the shifted $h_{\mu\nu}$ gives $\bar h_{\mu\nu}\to\bar h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu+\eta_{\mu\nu}\partial_\alpha\xi^\alpha$ . Why: because $\bar h$ carries a trace piece, its law differs from that of $h_{\mu\nu}$ by exactly the extra $\eta_{\mu\nu}\partial_\alpha\xi^\alpha$ term.

Step 3. Transform the divergence. Taking $\partial^\mu$ , the $\eta$ -term and one double-derivative term cancel, leaving $\partial^\mu\bar h_{\mu\nu}\to\partial^\mu\bar h_{\mu\nu}-\Box\xi_\nu$ . Why: the divergence we want to kill responds to a gauge change simply through $\Box\xi_\nu$ , a solvable operator.

Step 4. Solve for the generator. Choosing $\xi_\nu$ with $\Box\xi_\nu=\partial^\mu\bar h_{\mu\nu}$ sends the new divergence to zero, and such a $\xi_\nu$ always exists because $\Box$ has a retarded Green's function for any right-hand side. Why: this is what makes the Lorenz gauge always attainable. A residual freedom remains — any further $\xi^\mu$ with $\Box\xi^\mu=0$ preserves the gauge, later spent to reach the transverse-traceless gauge in vacuum. *As a check: the residual $\Box\xi^\mu=0$ freedom removes four of the ten components of $\bar h_{\mu\nu}$ , leaving the two physical polarizations.*

Go deeper · Derivation C · 1 minThe Newtonian limit and the gravitational potential

Derivation CThe Newtonian limit and the gravitational potential

The goal is to show that the trace-reversed wave equation, in the static weak-field regime, reproduces Newtonian gravity exactly — Poisson's equation and the identification $h_{00}=-2\Phi$ . We specialize the source to slow, low-pressure matter so $T_{00}=\rho$ dominates, drop time derivatives, solve for $\bar h_{00}$ , and untrace-reverse to recover the metric component a slow test particle feels through the geodesic equation.

Step 1. Dominant source. For non-relativistic matter the rest-energy density swamps momentum flux and pressure, so $T_{00}\approx\rho$ while $T_{0i}\approx0$ and $T_{ij}\approx0$ ; only $\bar h_{00}$ is sourced. Why: the components of $T_{\mu\nu}$ scale as $\rho$ , $\rho v$ , $\rho v^2$ , and the slow-motion limit keeps only the first.

Step 2. Drop time derivatives. For a static field $\partial_t\to0$ , so $\Box\to\nabla^2$ and the equation becomes $\nabla^2\bar h_{00}=-16\pi G\rho$ . Why: a stationary mass gives a stationary potential; the wave operator degenerates to the Laplacian, as in electrostatics.

Step 3. Untrace-reverse. With only $\bar h_{00}$ non-zero, $\bar h=\eta^{00}\bar h_{00}=-\bar h_{00}$ , so $h_{00}=\bar h_{00}-\tfrac12\eta_{00}\bar h=\tfrac12\bar h_{00}$ , giving $\nabla^2h_{00}=-8\pi G\rho$ . Why: a slow test particle couples through the geodesic equation to $h_{00}$ , not $\bar h_{00}$ , so we convert back before comparing with Newton.

Step 4. Match to Newton. The geodesic equation reproduces $g_{00}=-(1+2\Phi)$ , i.e. $h_{00}=-2\Phi$ ; substituting into $\nabla^2h_{00}=-8\pi G\rho$ yields $\nabla^2\Phi=4\pi G\rho$ . Why: Poisson's equation with the correct constant, confirming that general relativity contains Newtonian gravity as its static weak-field limit. *As a check: the spatial part follows as $h_{ij}=-2\Phi\,\delta_{ij}$ , giving the standard weak-field line element $ds^2=-(1+2\Phi)dt^2+(1-2\Phi)\delta_{ij}dx^idx^j$ .*

Limiting cases

Case	Result	Meaning
Static Newtonian source	$\nabla^2\bar h_{00}=-16\pi G\rho$	Time derivatives drop; with $T_{00}=\rho$ this gives $h_{00}=-2\Phi$ and Poisson's equation $\nabla^2\Phi=4\pi G\rho$ , recovering Newtonian gravity.
Vacuum (radiation)	$\Box\bar h_{\mu\nu}=0$	With $T_{\mu\nu}=0$ the equation is the homogeneous wave equation; plane waves in transverse-traceless gauge carry the two polarizations $h_+,h_\times$ .
Trace identity	$\bar h=-h,\quad\bar{\bar h}_{\mu\nu}=h_{\mu\nu}$	Trace reversal is an involution; applying it twice restores the original perturbation exactly.
Pure-trace (conformal) mode	$h_{\mu\nu}=\tfrac14\eta_{\mu\nu}h\Rightarrow\bar h_{\mu\nu}=-\tfrac14\eta_{\mu\nu}h$	A pure-trace perturbation stays pure-trace under reversal, with the sign flipped — isolating the conformal degree of freedom.
$D$ spacetime dimensions	$\bar h_{\mu\nu}=h_{\mu\nu}-\tfrac1{D-2}\eta_{\mu\nu}h$	The coefficient generalizing $\tfrac12$ ; only in $D=4$ does the trace-reversal factor equal one-half, and gravity has no propagating waves for $D<4$ .

Common trapConfusing the trace-reversed perturbation with the transverse-traceless part

The trace reversal $\bar h_{\mu\nu}=h_{\mu\nu}-\tfrac12\eta_{\mu\nu}h$ is a purely algebraic operation, defined for any $h_{\mu\nu}$ before any gauge and regardless of any source. The transverse-traceless (TT) part is something else: a further projection that only makes sense in vacuum, after the Lorenz gauge, using the leftover $\Box\xi^\mu=0$ freedom to strip $\bar h_{\mu\nu}$ to its two radiative components. In general $\bar h_{\mu\nu}$ is neither transverse nor traceless — indeed $\bar h_{00}=-4\Phi$ is the whole Newtonian field, as far from radiation as one can get. A subtler trap: at linear order all indices on $h_{\mu\nu}$ and $\bar h_{\mu\nu}$ are raised and lowered with $\eta$ , never with the full $g$ , which would introduce spurious second-order terms.

Worked examples

1 · Trace-reversing a perturbation by hand (confidence builder)

Take the Newtonian metric perturbation $h_{\mu\nu}=\mathrm{diag}(-2\Phi,-2\Phi,-2\Phi,-2\Phi)$ in signature $(-,+,+,+)$ . Compute its trace, its trace-reversed form $\bar h_{\mu\nu}$ , verify the trace flips sign, and confirm the operation is an involution.

Step 1. Compute the trace with the flat metric: $h=\eta^{\mu\nu}h_{\mu\nu}=-h_{00}+h_{11}+h_{22}+h_{33}=-(-2\Phi)+3(-2\Phi)=2\Phi-6\Phi=\mathbf{-4\Phi}$ . The single raised time index supplies the crucial minus sign.

Step 2. Apply $\bar h_{\mu\nu}=h_{\mu\nu}-\tfrac12\eta_{\mu\nu}h$ component by component. For the time-time slot, $\bar h_{00}=-2\Phi-\tfrac12(-1)(-4\Phi)=-2\Phi-2\Phi=\mathbf{-4\Phi}$ . For each spatial slot, $\bar h_{ii}=-2\Phi-\tfrac12(+1)(-4\Phi)=-2\Phi+2\Phi=\mathbf{0}$ . The perturbation has collapsed onto $\bar h_{\mu\nu}=\mathrm{diag}(-4\Phi,0,0,0)$ — a single non-zero component.

Step 3. Check the trace flip: $\bar h=\eta^{00}\bar h_{00}=(-1)(-4\Phi)=\mathbf{+4\Phi}=-h$ , exactly as equation (2) demands.

Step 4. Confirm the involution. Trace-reverse again: $\bar{\bar h}_{00}=\bar h_{00}-\tfrac12\eta_{00}\bar h=-4\Phi-\tfrac12(-1)(4\Phi)=-4\Phi+2\Phi=-2\Phi$ , and $\bar{\bar h}_{ii}=0-\tfrac12(+1)(4\Phi)=-2\Phi$ . This returns $\mathrm{diag}(-2\Phi,-2\Phi,-2\Phi,-2\Phi)=h_{\mu\nu}$ . The single surviving component $\bar h_{00}=-4\Phi$ then obeys $\nabla^2\bar h_{00}=-16\pi G\rho$ , which is just $\nabla^2\Phi=4\pi G\rho$ — Poisson's equation drops out.

2 · The strain of a gravitational wave from GW150914

Two black holes of about $35\,M_\odot$ each spiral together and merge $410\ \mathrm{Mpc}$ away, at the moment when their orbital speed reaches roughly half the speed of light. Estimate the gravitational-wave strain $h$ that reaches Earth using the quadrupole formula, restoring factors of $c$ .

Step 1. The quadrupole formula (8) with $c$ restored reads $h\sim\dfrac{2G}{c^4 r}\ddot Q$ . For a binary $Q\sim\mu d^2$ with reduced mass $\mu=\dfrac{m_1 m_2}{m_1+m_2}$ , so $\ddot Q\sim2\mu v^2$ with $v$ the orbital speed, giving $h\sim\dfrac{4G\mu v^2}{c^4 r}$ .

Step 2. The reduced mass is $\mu=\dfrac{35\times35}{70}M_\odot=17.5\,M_\odot=3.5\times10^{31}\ \mathrm{kg}$ , and the useful combination is the gravitational length $\dfrac{G\mu}{c^2}\approx\mathbf{2.6\times10^4\ m}$ (about $26\ \mathrm{km}$ ).

Step 3. With $v^2=(0.5c)^2=0.25\,c^2$ cancelling two powers of $c$ , $h\sim\dfrac{G\mu}{c^2 r}$ , and the distance is $r=410\ \mathrm{Mpc}\approx1.27\times10^{25}\ \mathrm{m}$ .

Step 4. Divide: $h\sim\dfrac{2.6\times10^4}{1.27\times10^{25}}\approx\mathbf{1\times10^{-21}}$ . This matches the peak $\sim10^{-21}$ LIGO recorded on 14 September 2015 — a length change of $\sim10^{-18}\ \mathrm{m}$ across the $4\ \mathrm{km}$ arms, a thousandth of a proton's width. For scale, even at the Sun's surface $|h_{00}|=2GM_\odot/(R_\odot c^2)\approx4.2\times10^{-6}$ , only parts per million — which is why weak-field gravity is so accurate and its radiative corrections so hard to catch.

Physics of the portrait

The trace-reversed wave equation is the heart of the boldest prediction Einstein ever hedged on: that spacetime itself can ripple. He wrote it down soon after the field equations, doubted for decades whether the ripples were real, and never saw them confirmed.

Origins · 1 minAlbert Einstein and how this was discovered

Albert Einstein

1879–1955 · linearized general relativity and the prediction of gravitational waves

Within months of completing general relativity in November 1915, Einstein turned to its weak-field limit. In a 1916 paper he linearized his equations, introduced what we now call the trace-reversed perturbation and the harmonic gauge, and found the metric obeys a wave equation moving at light speed — gravitational waves. His first attempt had an error in the energy formula, corrected in 1918 to give the quadrupole formula: a system radiates only through the changing shape of its mass. Then came the doubt. In 1936 Einstein and Nathan Rosen submitted a paper titled "Do Gravitational Waves Exist?" arguing they did not; the referee Howard Percy Robertson caught the flaw — their apparent singularity was merely a bad coordinate choice for cylindrical waves. The waves' physicality was settled only at the 1957 Chapel Hill conference, through the "sticky bead" argument of Feynman and Bondi. His linearized theory then waited for the technology to catch up. It did on 14 September 2015, when LIGO detected the strain his 1916 wave equation had predicted, a century on.

History

1867

— Lorenz introduces the retarded-potential gauge condition later carried into gravity

1916

— Einstein linearizes his field equations and predicts gravitational waves

1918

— Einstein corrects the theory and derives the quadrupole radiation formula

1921

— De Donder formalizes the harmonic (Lorenz) gauge for general relativity

1957

— the Chapel Hill conference settles that gravitational waves carry energy

2015

— LIGO records GW150914, confirming the wave equation's prediction

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §18.1–18.2 · Wald, General Relativity (1984), §4.4 · Carroll, Spacetime and Geometry (2004), §7.1 · Schutz, A First Course in General Relativity, 2nd ed., §8.3 · Hartle, Gravity (2003), §21.1 The definition and trace identity follow by direct contraction ( $\eta^{\mu\nu}\eta_{\mu\nu}=4$ ); the involution on the Newtonian perturbation gives $\bar h_{00}=-4\Phi$ with the spatial diagonal emptied. The linearized Ricci and Einstein tensors, the trace cancellation, the Lorenz-gauge reduction, and $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ are all in units $c=1$ . The Newtonian recovery yields $h_{00}=-2\Phi$ and $\nabla^2\Phi=4\pi G\rho$ . The GW150914 strain ( $h\sim10^{-21}$ from $\mu=17.5\,M_\odot$ , $v\sim0.5c$ , $r=410\ \mathrm{Mpc}$ ) and the Solar-surface value ( $|h_{00}|\approx4.2\times10^{-6}$ ) agree with the published LIGO strain.

Prerequisites. The Weak-Field Approximation · The Einstein Tensor · The Einstein Field Equations · Local Conservation of Energy-Momentum. Leads to. The Newtonian Limit · The Stress-Energy Tensor.

Ask the tutor: Ask me to trace-reverse a specific $h_{\mu\nu}$ of your choosing, walk through the cancellation of the trace terms in $G^{(1)}_{\mu\nu}$ , or estimate the strain of any binary you name.

Why is the factor exactly one-half and not something else?

Because four spacetime dimensions make it so. The subtraction $\bar h_{\mu\nu}=h_{\mu\nu}-c\,\eta_{\mu\nu}h$ produces a new trace $\bar h=h-c(\eta^{\mu\nu}\eta_{\mu\nu})h=h-4c\,h$ ; to flip the sign, $\bar h=-h$ , you need $1-4c=-1$ , so $c=\tfrac12$ . In general $D$ dimensions the same demand gives $c=1/(D-2)$ , one-half only when $D=4$ . The coefficient is not a free convention but a fixed consequence of dimensionality — the same algebra that makes the Einstein tensor $R_{\mu\nu}-\tfrac12 g_{\mu\nu}R$ the trace-reversed Ricci tensor.

What actually goes wrong if I do not trace-reverse?

Nothing is incorrect — the physics is the same — but the equation stays ugly. Written directly in $h_{\mu\nu}$ , the linearized Einstein tensor keeps explicit trace pieces such as $\partial_\mu\partial_\nu h$ and $\eta_{\mu\nu}\Box h$ that do not decouple. Even after a Lorenz-like gauge on $h_{\mu\nu}$ itself, those trace terms survive, so you cannot reduce the field equations to a single d'Alembertian: each component stays coupled to the trace of the whole tensor. Trace reversal is precisely the move that cancels them, which is why every textbook adopts $\bar h_{\mu\nu}$ first.

Is the Lorenz gauge unique?

No. After imposing $\partial^\mu\bar h_{\mu\nu}=0$ a residual freedom remains: any further coordinate change generated by a field $\xi^\mu$ with $\Box\xi^\mu=0$ preserves the Lorenz condition. This leftover has four functions' worth of choice, exactly enough to remove four of the ten components of $\bar h_{\mu\nu}$ in vacuum. Spending it defines the transverse-traceless gauge, stripping the field to its two physical polarizations $h_+$ and $h_\times$ . So the Lorenz gauge is a family, not a single choice — and that is what makes the gravitational-wave degrees of freedom come out to two.

How does the formalism guarantee energy-momentum conservation?

It comes out for free. Take the divergence of the field equation: $\partial^\mu(\Box\bar h_{\mu\nu})=\Box(\partial^\mu\bar h_{\mu\nu})$ since partials commute on flat space. The Lorenz condition sets $\partial^\mu\bar h_{\mu\nu}=0$ , so the left side vanishes; but the field equation gives $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ , so $\partial^\mu T_{\mu\nu}=0$ follows automatically. This is the linearized shadow of the contracted Bianchi identity $\partial^\mu G^{(1)}_{\mu\nu}=0$ : the geometry is built so its source must be conserved. Conservation is not an extra postulate but a consistency condition baked into the wave equation.

Why does gravity radiate through the quadrupole rather than the dipole?

Because the lower multipoles cannot change. Electromagnetic dipole radiation needs an oscillating charge dipole, but the gravitational "charge" is mass-energy: its total (the monopole) is conserved, and the mass dipole's rate of change is just the total momentum, also conserved for an isolated system. With the monopole and dipole channels frozen, the first moment that can radiate is the quadrupole. This is why the leading field is $\bar h^{TT}_{ij}\propto\ddot Q_{ij}$ , and why gravitational waves are so weak: they are sourced only by the changing shape of a mass, never its bulk motion.

Does the analogy with electromagnetism ever break down?

Yes, and instructively. At linear order the parallel is exact: $\bar h_{\mu\nu}$ mirrors $A_\mu$ , the Lorenz gauges match, the wave equations are twins. But gravity is nonlinear — the full Einstein tensor has $h^2$ and higher terms the linearization discards. Gravitational waves carry energy and therefore gravitate, feeding back into the metric; their effective stress-energy (the Isaacson tensor, built from gradients of $\bar h^{TT}_{\mu\nu}$ ) is a second-order effect with no electromagnetic analogue, since photons carry no charge. The linear theory is thus the first term of an expansion, superb in the weak field and corrected by post-Newtonian and numerical relativity when fields grow strong.