Gauge Transformations in Linearized Gravity

Linearized gravity is what Einstein's theory becomes when you look at a nearly flat patch of spacetime and keep only the first whisper of curvature. Write the metric as $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$ with $|h_{\mu\nu}|\ll1$ , expand the field equations to first order, and the geometry dissolves into something familiar: a symmetric tensor field $h_{\mu\nu}$ on a fixed Minkowski background, obeying wave equations, radiating, carrying energy — a relativistic field theory on flat space. But a catch runs through the whole subject: not every $h_{\mu\nu}$ is physics. The full theory is invariant under arbitrary smooth relabellings of the points of spacetime (diffeomorphisms), and the shadow that invariance casts on the weak field is a gauge symmetry acting on $h_{\mu\nu}$ . Perform an infinitesimal coordinate change $x^\mu\to x^\mu+\xi^\mu(x)$ , with $\xi^\mu$ as small as $h$ itself, and the perturbation shifts by $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ . Two perturbations related this way are the same curved spacetime in two slightly different coordinates. This is not merely an analogy to the gauge freedom $A_\mu\to A_\mu-\partial_\mu\Lambda$ of electromagnetism — it is the same mathematical structure, with $\xi^\mu$ playing the role of $\Lambda$ . The consequences are practical: raw metric components mean nothing until you fix a gauge or build gauge-invariant combinations; the freedom collapses the field equations to a clean wave equation; and it explains — by a bookkeeping subtraction $10-4-4=2$ — why a gravitational wave carries exactly two polarizations. Get gauge wrong and you will, as Einstein himself once did, mistake a coordinate artifact for a physical wave.

Key takeaways

The weak-field metric splits as $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$ ; an infinitesimal coordinate change $x^\mu\to x^\mu+\xi^\mu$ shifts the perturbation by $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu=h_{\mu\nu}-2\partial_{(\mu}\xi_{\nu)}$ .
This is the linearized residue of full diffeomorphism invariance — a genuine redundancy, not a physical change — structurally identical to the gauge freedom of electromagnetism, with $\xi^\mu\leftrightarrow\Lambda$ .
The linearized Riemann tensor is invariant under any $\xi^\mu$ , so all curvature, tidal forces, and wave strain are gauge-independent; individual components of $h_{\mu\nu}$ are not observable.
Counting: $h_{\mu\nu}$ has 10 components; the 4 functions $\xi^\mu$ remove 4 and residual freedom removes 4 more, leaving 2 physical degrees of freedom — the graviton polarizations $h_+,h_\times$ .
Fixing the harmonic (Lorenz / de Donder) gauge $\partial^\mu\bar h_{\mu\nu}=0$ collapses the field equations to decoupled wave equations $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ ; residual freedom with $\Box\xi^\mu=0$ then reaches the transverse-traceless (TT) gauge.
Key idea: under $x^\mu\to x^\mu+\xi^\mu$ the perturbation transforms as $h_{\mu\nu}\to h_{\mu\nu}-2\partial_{(\mu}\xi_{\nu)}$ ; this is pure gauge — the linearized Riemann tensor and every curvature invariant are untouched, so physics lives only in gauge-invariant combinations.

The core insight

Under an infinitesimal coordinate change $x^\mu\to x^\mu+\xi^\mu$ the metric perturbation transforms as $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu=h_{\mu\nu}-2\partial_{(\mu}\xi_{\nu)}=h_{\mu\nu}-\mathcal{L}_\xi\eta_{\mu\nu}$ ; this is a pure gauge redundancy, the exact linear analogue of $A_\mu\to A_\mu-\partial_\mu\Lambda$ , and the linearized Riemann tensor is invariant under it.

Where this lives today

Gauge transformations in linearized gravity are the working language of gravitational-wave astronomy. Every LIGO–Virgo–KAGRA detection is analyzed in TT gauge, where $h_+$ and $h_\times$ are the templates matched against the data; matched filtering assumes the eight gauge components have been stripped away and only the invariant strain remains, and numerical-relativity simulations extract the radiated $h^{\mathrm{TT}}_{ij}$ so it can be compared gauge-invariantly across codes. The same separation of physical from pure-gauge modes drives cosmology: the Bardeen potentials are gauge-invariant combinations of metric and matter perturbations, so cosmic-microwave-background predictions do not depend on how one slices spacetime. At the frontier, the massless spin-2 invariance $\delta h_{\mu\nu}=-2\partial_{(\mu}\xi_{\nu)}$ is a constraint any quantum theory of gravity must respect, and it forbids a graviton mass without a delicate reorganization. Pulsar timing arrays now hunt a nanohertz background with the same TT-gauge strain, extending Einstein's 1916 linearization across nine decades of frequency.

The physical picture

Hold two pictures side by side. In the first, $h_{\mu\nu}$ is a tensor field on a rigid flat background, and a gauge transformation is just a smooth relabelling of the background points. Sliding every point by a tiny displacement $\xi^\mu(x)$ repaints the field's components — as rotating your axes changes the components of an ordinary vector — though no geometry has moved. That is why $\delta h_{\mu\nu}=-2\partial_{(\mu}\xi_{\nu)}$ is exactly minus the Lie derivative of the background, $-\mathcal{L}_\xi\eta_{\mu\nu}$ . Infinitely many $h_{\mu\nu}$ therefore describe the same spacetime; they form a "gauge orbit," and only the direction transverse to it — the gauge-invariant part — carries physics.

In the second picture we ask what a real detector feels. Because physics lives only in the gauge-invariant part, the transverse-traceless (TT) gauge throws away eight components and leaves the two — $h_+$ and $h_\times$ — that stretch and squeeze proper distances between freely falling masses. A wave sweeping through the Earth distorts a ring of test masses into a pulsing ellipse, and that distortion is a real, measured, staggeringly small number.

Figure

The two polarizations h₊ and h× stretch and squeeze a ring of free test masses in orthogonal patterns; for GW150914 the peak strain was h ≈ 1.0×10⁻²¹, so across a LIGO arm of length L = 4 km the length change is ΔL = hL ≈ 4×10⁻¹⁸ m — a few thousandths of a proton diameter

The core equations

Start from the split that defines the weak field. Every equation is an expansion in $h_{\mu\nu}$ to first order.

g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu},\qquad |h_{\mu\nu}|\ll1 \qquad (1)

where: $\eta_{\mu\nu}$ — Minkowski metric, signature $(-,+,+,+)$ ; $h_{\mu\nu}$ — symmetric perturbation, first order. Indices are raised and lowered with $\eta$ . The central object is the gauge transformation — what an infinitesimal coordinate change $x^\mu\to x^\mu+\xi^\mu$ does to the perturbation. Derivation A derives it from the tensor transformation law.

h_{\mu\nu}\to h'_{\mu\nu}=h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu=h_{\mu\nu}-2\partial_{(\mu}\xi_{\nu)} \qquad (2)

where: $\xi^\mu(x)$ — infinitesimal gauge vector, of order $h$ ; $\partial_{(\mu}\xi_{\nu)}\equiv\tfrac12(\partial_\mu\xi_\nu+\partial_\nu\xi_\mu)$ ; the variation equals $-\mathcal{L}_\xi\eta_{\mu\nu}$ . Gauge freedom never touches observables because the linearized Riemann tensor — which encodes tidal forces and wave strain — is invariant under (2).

R^{(1)}_{\mu\nu\rho\sigma}=\tfrac12\big(\partial_\nu\partial_\rho h_{\mu\sigma}+\partial_\mu\partial_\sigma h_{\nu\rho}-\partial_\mu\partial_\rho h_{\nu\sigma}-\partial_\nu\partial_\sigma h_{\mu\rho}\big) \qquad (3)

where: $R^{(1)}_{\mu\nu\rho\sigma}$ — Riemann tensor to first order in $h$ , antisymmetric in $[\mu\nu]$ and $[\rho\sigma]$ . Derivation B shows $\delta R^{(1)}_{\mu\nu\rho\sigma}=0$ under (2). To write the field equations compactly it pays to trace-reverse the perturbation — the change of variable that makes harmonic gauge look clean.

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

where: $\bar h_{\mu\nu}$ — trace-reversed perturbation; $h$ — the trace. Reversing twice returns the original. In these variables the linearized Einstein equations — valid in any gauge — read as one wave operator plus three divergence terms, which a gauge choice can kill.

\Box\bar h_{\mu\nu}-\partial_\rho\partial_\nu\bar h^\rho{}_\mu-\partial_\rho\partial_\mu\bar h^\rho{}_\nu+\eta_{\mu\nu}\partial_\rho\partial_\sigma\bar h^{\rho\sigma}=-16\pi G\,T_{\mu\nu} \qquad (5)

where: $\Box=\eta^{\rho\sigma}\partial_\rho\partial_\sigma$ — flat d'Alembertian; $T_{\mu\nu}$ — stress-energy at zeroth order in $h$ ; $G$ — Newton's constant; $c=1$ . Every term after $\Box\bar h_{\mu\nu}$ carries a divergence $\partial^\rho\bar h_{\rho\mu}$ . Imposing the harmonic (de Donder) gauge sets that divergence to zero — four conditions, one per free index $\nu$ . Derivation C proves it is always reachable.

\partial^\mu\bar h_{\mu\nu}=0 \qquad (6)

where: $\partial^\mu\bar h_{\mu\nu}$ — divergence of $\bar h$ . "Harmonic" reflects that the coordinates satisfy $\Box x^\mu=0$ ; "de Donder" and "Lorenz gauge" are the gravitational and electromagnetic names. With (6) imposed, the last three terms of (5) vanish and each component of $\bar h_{\mu\nu}$ obeys the same sourced wave equation — predicting gravitational radiation at $c$ .

\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu} \qquad (7)

where: the operator acts component-by-component, so the ten equations decouple and solve by retarded integrals. In vacuum, $\Box\bar h_{\mu\nu}=0$ . The residual freedom left after (6) reaches the transverse-traceless gauge, where a vacuum wave shows only its two physical polarizations.

h^{\mathrm{TT}}_{0\mu}=0,\qquad \partial^j h^{\mathrm{TT}}_{ij}=0,\qquad \delta^{ij}h^{\mathrm{TT}}_{ij}=0 \qquad (8)

where: $h^{\mathrm{TT}}_{ij}$ — the purely spatial, transverse, traceless perturbation. In TT gauge $\bar h^{\mathrm{TT}}_{ij}=h^{\mathrm{TT}}_{ij}$ , and the survivors are exactly $h_+$ and $h_\times$ .

Derivation AThe gauge law from an infinitesimal change of coordinates

The goal is to derive equation (2) from first principles — to show that $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ is not an ad-hoc rule but the unavoidable consequence of writing the same metric in shifted coordinates $x^\mu\to x'^\mu=x^\mu+\xi^\mu(x)$ , kept to first order, and to identify the variation with a Lie derivative that makes the "only a relabelling" slogan precise.

Step 1. Apply the tensor transformation law to the full metric: $g'_{\mu\nu}(x')=\dfrac{\partial x^\alpha}{\partial x'^\mu}\dfrac{\partial x^\beta}{\partial x'^\nu}\,g_{\alpha\beta}(x)$ , exact and saying nothing yet about smallness. Why: the metric is a rank-2 tensor, so its components in a new chart are fixed entirely by the Jacobian of the coordinate change.

Step 2. Invert the coordinate change. From $x'^\mu=x^\mu+\xi^\mu$ we get $x^\mu=x'^\mu-\xi^\mu+O(\xi^2)$ , so the Jacobian is $\dfrac{\partial x^\alpha}{\partial x'^\mu}=\delta^\alpha_\mu-\partial_\mu\xi^\alpha+O(\xi^2)$ . Why: to first order the shift and its inverse differ only in the sign of $\xi$ , and the derivative supplies the $\partial\xi$ pieces that become the gauge term.

Step 3. Substitute and linearize. Insert the Jacobians and $g_{\alpha\beta}=\eta_{\alpha\beta}+h_{\alpha\beta}$ , drop products of two small quantities, and use $\partial_\mu\eta_{\alpha\beta}=0$ ; the cross-terms collapse to $g'_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ . Why: the constant background survives untouched, so the entire effect of the coordinate change is dumped into the perturbation — gauge acts on $h$ , not on $\eta$ .

Step 4. Read off the transformation and recognize it. Since $g'_{\mu\nu}=\eta_{\mu\nu}+h'_{\mu\nu}$ , matching gives $h'_{\mu\nu}=h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ , which is (2). The subtracted piece is $\partial_\mu\xi_\nu+\partial_\nu\xi_\mu=\mathcal{L}_\xi\eta_{\mu\nu}$ . Why: identifying $\delta h_{\mu\nu}=-\mathcal{L}_\xi\eta_{\mu\nu}$ shows the transformation is the flat background dragged along the flow $\xi$ — the exact linear image of a diffeomorphism.

Go deeper · Derivation B · 1 minWhy curvature does not care about gauge

Derivation BWhy curvature does not care about gauge

The goal is to prove that the linearized Riemann tensor (3) is exactly invariant under (2), so that $\delta R^{(1)}_{\mu\nu\rho\sigma}=0$ for every $\xi^\mu$ . This is the most important fact in the subject: no measurement built from curvature — tidal acceleration, geodesic deviation, interferometer strain — can depend on the gauge, so the raw components of $h_{\mu\nu}$ are not observable. The strategy is direct substitution followed by a cancellation that rides on the commuting of partial derivatives.

Step 1. Feed the variation into each slot. Replace every $h$ in (3) by $\delta h_{\alpha\beta}=-\partial_\alpha\xi_\beta-\partial_\beta\xi_\alpha$ ; the first term gives $\partial_\nu\partial_\rho\,\delta h_{\mu\sigma}=-\partial_\nu\partial_\rho\partial_\mu\xi_\sigma-\partial_\nu\partial_\rho\partial_\sigma\xi_\mu$ , and similarly for the other three. Why: the linearized Riemann tensor is linear in $h$ , so its variation is the same formula with $h\to\delta h$ — no product rule, no second-order debris.

Step 2. Lay out all eight third-derivative terms. Each of the four terms in (3) splits into two, giving eight of the schematic form $\pm\,\partial\partial\partial\xi$ , each with three of $\{\mu,\nu,\rho,\sigma\}$ on the derivatives and one on $\xi$ . Why: the cancellation is not a slogan but an explicit pairing, best seen written out.

Step 3. Pair the terms using commuting derivatives. The two terms carrying $\xi_\sigma$ are $-\tfrac12\partial_\nu\partial_\rho\partial_\mu\xi_\sigma$ and $+\tfrac12\partial_\nu\partial_\rho\partial_\mu\xi_\sigma$ ; since partial derivatives commute the ordering is irrelevant and they cancel. Why: the antisymmetries $[\mu\nu]$ and $[\rho\sigma]$ of Riemann generate every $\partial\partial\partial\xi$ term a second time with opposite sign, and equality of mixed partials makes the two the same object.

Step 4. Sum the pairs. Every term finds a partner of opposite sign, so the total vanishes: $\delta R^{(1)}_{\mu\nu\rho\sigma}=0$ for arbitrary $\xi^\mu$ . Why: gauge changes shuffle coordinates, and coordinates are exactly what the Riemann tensor is built to be independent of.

Go deeper · Derivation C · 1 minHarmonic gauge and the wave equation

Derivation CHarmonic gauge and the wave equation

The goal is threefold: to show the harmonic gauge $\partial^\mu\bar h_{\mu\nu}=0$ can always be reached, to use it to collapse the messy field equation (5) into the clean wave equation (7), and to expose the consistency condition it demands of the source. This is where gauge freedom becomes a tool, turning Einstein's weak-field equations into the same wave equation that governs light — and where the link between gauge symmetry and energy-momentum conservation shows itself.

Step 1. Find how the divergence transforms. Under (2), $\delta\bar h_{\mu\nu}=-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu+\eta_{\mu\nu}\partial_\rho\xi^\rho$ ; taking its divergence and cancelling the trace piece gives $\partial^\mu\bar h'_{\mu\nu}=\partial^\mu\bar h_{\mu\nu}-\Box\xi_\nu$ . Why: the divergence is exactly what the gauge condition constrains, so its transformation law decides whether the condition is reachable.

Step 2. Solve for the required $\xi$ . Setting $\partial^\mu\bar h'_{\mu\nu}=0$ demands $\Box\xi_\nu=\partial^\mu\bar h_{\mu\nu}$ , an inhomogeneous wave equation that always admits a retarded-Green's-function solution — so a gauge vector reaching harmonic gauge always exists. Why: reachability is guaranteed by the solvability of the wave equation, which is why every weak field can be put in harmonic form.

Step 3. Collapse the field equations. With $\partial^\rho\bar h_{\rho\mu}=0$ , the three divergence terms in (5) vanish and only $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ survives — each component an independent sourced wave equation. Why: a single operator replaces the tangle of coupled second derivatives, and gravitational radiation falls out as its retarded solution.

Step 4. Read off the consistency condition. Take the divergence of (7): the left side gives $\partial^\mu\Box\bar h_{\mu\nu}=\Box(\partial^\mu\bar h_{\mu\nu})=0$ by the gauge condition, so the right side forces $\partial^\mu T_{\mu\nu}=0$ . Local energy-momentum conservation is not optional — it is required for the gauge-fixed equations to be consistent. Why: this is the linearized Bianchi identity, the field-theory shadow of the fact that gauge invariance and conservation of the source are two faces of one symmetry, just as $\partial^\mu J_\mu=0$ follows from Lorenz-gauge invariance in electromagnetism.

Limiting cases

Each row is the same theory viewed through a different gauge or source, showing how many physical fields survive.

Case	Result	Meaning
Vacuum, harmonic gauge	$\Box\bar h_{\mu\nu}=0,\ \ \partial^\mu\bar h_{\mu\nu}=0$	Residual freedom $\Box\xi^\mu=0$ removes 4 further components, reaching TT gauge: exactly 2 polarizations propagate at $c$ .
Static source (Newtonian limit)	$\nabla^2\bar h_{00}=-16\pi G\,\rho$	Time-independent gauge fixes the spatial coordinates; $\Phi=-\tfrac14\bar h_{00}$ obeys Poisson's equation $\nabla^2\Phi=4\pi G\rho$ .
Pure gauge	$h_{\mu\nu}=-2\partial_{(\mu}\xi_{\nu)}$	For any smooth $\xi^\mu$ the linearized Riemann tensor vanishes identically — flat spacetime in disguise, no physics at all.
Detector scale	$\ddot{\!\xi}^{\,i}=-R^{(1)}_{i0j0}\,\xi^j$	Geodesic deviation depends only on gauge-invariant $R^{(1)}$ ; the "motion" of a single mass is pure coordinate.

Common trapReading a physical effect off the raw components of h_μν

The most common and costly error in linearized gravity is to treat components of $h_{\mu\nu}$ as observable. They are not. A perfectly nonzero $h_{\mu\nu}=-2\partial_{(\mu}\xi_{\nu)}$ can describe flat, empty Minkowski space in wobbly coordinates — its linearized Riemann tensor vanishes everywhere, so there is no tidal force, no curvature, no wave. Conversely, a "gravitational effect" that appears in one gauge and disappears in TT gauge was never physical. The discipline: never claim a physical result from raw metric components until you have confirmed it survives in $R^{(1)}_{\mu\nu\rho\sigma}$ or another manifestly gauge-invariant quantity. This is exactly the mistake that led Einstein and Rosen in 1936 to nearly conclude gravitational waves do not exist — they had found a coordinate singularity, not a physical one.

Worked examples

1 · The strain of GW150914 and why it is gauge-invariant (fully worked)

A gravitational wave from a binary black-hole merger passes through a LIGO detector with orthogonal arms of length $L=4$ km. At peak the dimensionless strain is $h\approx1.0\times10^{-21}$ . Find the change in arm length, confirm it is the TT strain and not a coordinate artifact, and estimate the wave's reduced wavelength at the $\sim100$ Hz band.

Step 1. In TT gauge the proper distance between two freely falling masses separated by coordinate distance $L$ along a polarization axis is $\ell\approx L(1+\tfrac12 h^{\mathrm{TT}})$ , so $\delta\ell/L=\tfrac12 h^{\mathrm{TT}}$ . Detectors quote the strain $h\equiv\Delta L/L$ as the differential effect across the two arms, restoring the factor of two: $\Delta L=hL$ . The length change is $\Delta L=hL=(1.0\times10^{-21})(4\times10^3\ \mathrm{m})=4\times10^{-18}\ \mathrm{m}$ .

Step 2. Put that in perspective. A proton has diameter $\approx1.7\times10^{-15}$ m, so $\Delta L/d_{\text{proton}}\approx(4\times10^{-18})/(1.7\times10^{-15})\approx2.4\times10^{-3}$ . The mirrors move by a few thousandths of a proton diameter — the measurement that earned the 2017 Nobel Prize.

Step 3. Confirm gauge-invariance. A detector reads the relative proper acceleration of its test masses, governed by $\ddot{\xi}^{\,i}=-R^{(1)}_{i0j0}\xi^j$ . Because $R^{(1)}_{\mu\nu\rho\sigma}$ is gauge-invariant (Derivation B), the same displacement is predicted in any gauge — TT gauge merely makes $h_+,h_\times$ manifest. A different gauge gives differently-looking $h_{\mu\nu}$ but the identical measurable $\Delta L$ .

Step 4. Estimate the wavelength. With $f\approx100$ Hz near merger, $\lambda=c/f=(3.0\times10^8)/100=3.0\times10^6$ m $=3000$ km, vastly larger than the 4 km arms — which is why the long-wavelength TT-gauge description as a nearly uniform strain across the detector is valid. $\lambda\approx3000$ km $\gg L$ .

2 · The Newtonian limit of the Sun (fully worked)

Show that harmonic gauge reproduces Newtonian gravity, and check the weak-field assumption $|h_{\mu\nu}|\ll1$ at the surface of the Sun ( $M=1.99\times10^{30}$ kg, $R=6.96\times10^8$ m).

Step 1. For a static, slow source only $T_{00}=\rho$ matters and $\Box\to\nabla^2$ , so (7) gives $\nabla^2\bar h_{00}=-16\pi G\rho$ . Define $\Phi\equiv-\tfrac14\bar h_{00}$ ; then $\nabla^2\Phi=-\tfrac14(-16\pi G\rho)=4\pi G\rho$ . This is exactly Poisson's equation $\nabla^2\Phi=4\pi G\rho$ — Newtonian gravity recovered from a gauge-fixed linearized Einstein equation.

Step 2. Undo the trace reversal. With $\bar h_{00}=-4\Phi$ , the relation $h_{\mu\nu}=\bar h_{\mu\nu}-\tfrac12\eta_{\mu\nu}\bar h$ gives $h_{00}=-2\Phi$ and $h_{ij}=-2\Phi\,\delta_{ij}$ , so $\mathrm{d}s^2=-(1+2\Phi)\,\mathrm{d}t^2+(1-2\Phi)\,\delta_{ij}\mathrm{d}x^i\mathrm{d}x^j$ . The single potential $\Phi$ sets both the time-time and space-space parts of the metric.

Step 3. Check weak-field validity at the Sun's surface via $|h_{00}|=2GM/(Rc^2)=2(6.674\times10^{-11})(1.99\times10^{30})/\big[(6.96\times10^8)(2.998\times10^8)^2\big]$ . This evaluates to $|h_{00}|\approx4.2\times10^{-6}$ , and $\Phi/c^2\approx2.1\times10^{-6}$ .

Step 4. Interpret. Since $4.2\times10^{-6}\ll1$ , linearization is superb at the Sun's surface — corrections of order $h^2\sim10^{-11}$ are why solar-system tests probe general relativity only at parts-per-million. The weak-field expansion is valid, with time-independent gauge freedom fixing the isotropic spatial coordinates above.

Physics of the portrait

The gauge structure of linearized gravity is inseparable from one man, who made both the discovery and the classic mistake — Einstein predicted gravitational waves, then twenty years later nearly convinced himself they were a fiction, tripped up by the gauge freedom he had helped uncover.

Origins · 1 minAlbert Einstein and how this was discovered

Albert Einstein

1879–1955 · weak-field gravity, gravitational waves, and the gauge that fooled him

Having completed general relativity in 1915, Albert Einstein turned at once to its weak-field limit. In 1916 he wrote the metric as a small perturbation of flat space, introduced what we now call harmonic coordinates, and found gravitational waves travelling at the speed of light; a 1918 paper corrected his own factor-of-two error and gave the quadrupole formula still used today. But the story that best illuminates gauge came later. In 1936 Einstein and Nathan Rosen submitted a paper to Physical Review titled "Are Gravitational Waves Real?" arguing — from an apparent singularity in a cylindrical-wave solution — that they were not. The referee (Howard Percy Robertson) recognized the singularity as a mere coordinate artifact, removable by a change of gauge; an indignant Einstein withdrew the paper but on reflection reached the same conclusion, and the 1937 version reversed the claim. Even Einstein could mistake a gauge effect for physics — only the invariance of curvature settles the matter. Théophile de Donder had meanwhile (1921) formalized the harmonic condition $\partial^\mu\bar h_{\mu\nu}=0$ , and in 1939 Fierz and Pauli recast the structure as the gauge theory of a massless spin-2 field. Vindication came in 2015: LIGO recorded a strain of $10^{-21}$ and heard two black holes merge.

History

1916

— Einstein linearizes his field equations and predicts gravitational waves

1921

— de Donder formalizes the harmonic coordinate condition ∂ᵘ h̄_μν = 0

1936

— Einstein and Rosen argue waves are unphysical; a referee catches the gauge error

1939

— Fierz and Pauli recast linearized gravity as the massless spin-2 gauge theory

2016

— LIGO announces GW150914, the first direct detection of gravitational waves

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §18.1–18.3 (linearized theory and gauge) · Wald, General Relativity, §4.4 (linearized equations, degrees of freedom) · Carroll, Spacetime and Geometry, §7.1 (gauge, harmonic and TT gauge) · Schutz, A First Course in General Relativity, §8.3 · Hartle, Gravity, Ch. 21 · Maggiore, Gravitational Waves Vol. 1, §1.2. Key self-consistency checks: the gauge law $\delta h_{\mu\nu}=-2\partial_{(\mu}\xi_{\nu)}$ follows from the coordinate transformation (Derivation A); substituting it into $R^{(1)}_{\mu\nu\rho\sigma}$ confirms invariance (Derivation B); the wave equation $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ requires $\partial^\mu T_{\mu\nu}=0$ for consistency (Derivation C); the degree-of-freedom count gives $10-4-4=2$ ; and the anchor numbers are $\Delta L=hL\approx4\times10^{-18}$ m and $|h_{00}|=2GM_\odot/(R_\odot c^2)\approx4.2\times10^{-6}$ . The factor $16\pi G$ assumes $c=1$ and signature $(-,+,+,+)$ .

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

Ask the tutor: Want to see the eight $\partial\partial\partial\xi$ terms cancel line by line, or walk through how the residual $\Box\xi^\mu=0$ freedom is spent to reach TT gauge for a plane wave?

Is a gauge transformation the same thing as a coordinate transformation?

Almost. A gauge transformation in linearized gravity is an infinitesimal coordinate transformation $x^\mu\to x^\mu+\xi^\mu$ , re-read as an active change of the field $h_{\mu\nu}$ on a fixed flat background rather than a passive relabelling; the two viewpoints give identical formulas. We call it "gauge" to stress the electromagnetic analogy and to emphasize that it is a redundancy of description — two $h_{\mu\nu}$ related this way are physically indistinguishable. Finite coordinate transformations in the full theory are diffeomorphisms; these gauge transformations are their linearized shadow.

Why exactly two degrees of freedom — where do the other eight go?

Start with the 10 components of the symmetric $h_{\mu\nu}$ . The four gauge functions $\xi^\mu$ impose the four harmonic conditions $\partial^\mu\bar h_{\mu\nu}=0$ , leaving six. But harmonic gauge does not use up all the freedom: any further $\xi^\mu$ with $\Box\xi^\mu=0$ preserves the condition, and this residual freedom removes four more (setting $h_{0\mu}=0$ and the trace to zero). Six minus four is two — the transverse-traceless components $h_+$ and $h_\times$ , the helicity- $\pm2$ graviton states. The count $10-4-4=2$ is the gravitational analogue of the photon's $4-2=2$ , with an extra subtraction because the graviton is spin-2.

Is the harmonic gauge the same as the Lorenz gauge in electromagnetism?

They are exact analogues, term for term. The Lorenz gauge $\partial^\mu A_\mu=0$ turns Maxwell's equations into $\Box A_\mu=-\mu_0 J_\mu$ with residual freedom $\Box\Lambda=0$ ; the harmonic (de Donder) gauge $\partial^\mu\bar h_{\mu\nu}=0$ turns Einstein's into $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ with residual freedom $\Box\xi^\mu=0$ . The dictionary is $A_\mu\leftrightarrow h_{\mu\nu}$ , $\Lambda\leftrightarrow\xi^\mu$ , $J_\mu\leftrightarrow T_{\mu\nu}$ , and the consistency conditions match too: $\partial^\mu J_\mu=0$ versus $\partial^\mu T_{\mu\nu}=0$ . The one difference is the extra index — gravity's source is a tensor, so there is one more layer of gauge fixing to reach the physical polarizations.

If gauge changes h_μν, how can gravitational-wave detectors measure anything?

Because detectors do not measure $h_{\mu\nu}$ — they measure the relative proper acceleration of freely falling test masses, governed by $\ddot{\xi}^{\,i}=-R^{(1)}_{i0j0}\xi^j$ . Since the linearized Riemann tensor is gauge-invariant, so is that acceleration and so is the interferometer fringe pattern. TT gauge is chosen not because it is "the real one" but because it is convenient: the mirrors stay at fixed coordinates and the whole physical effect sits in $h_+,h_\times$ . Any other gauge gives a differently-looking $h_{\mu\nu}$ but the identical measured strain — gauge invariance is what makes the measurement well-defined at all.

What actually went wrong in the 1936 Einstein–Rosen argument?

Einstein and Rosen studied cylindrical gravitational waves and found the metric became singular on an axis, reading this as evidence that no regular wave solution exists. The flaw was gauge: their coordinate system broke down on the axis, producing a coordinate singularity, not a curvature one — the Riemann tensor stayed finite, and a change to well-behaved coordinates removes the pathology. The referee (later revealed to be H. P. Robertson) spotted this, and after initial fury Einstein agreed, publishing a corrected paper in 1937. The lesson is the warning box: a feature living in the metric components but absent from the curvature invariants is an artifact of the chart, not a fact about spacetime.

Does the gauge freedom mean gravitational waves carry no energy?

No — but gauge is at the heart of the subtlety. A wave's energy is not localizable in a gauge-invariant, pointwise way, because you can always transform to a local freely-falling frame where the metric's first derivatives vanish. What is well-defined is the energy averaged over a region several wavelengths across, captured by the Isaacson stress-energy tensor built from the gauge-invariant $h^{\mathrm{TT}}_{ij}$ ; averaged, the radiated power reduces to Einstein's quadrupole formula. So waves do carry energy — it is why binary orbits decay — but gauge freedom forbids assigning that energy to a point, only to a region.