Gauge Freedom in Linearized Gravity

Einstein's field equations are invariant under any smooth relabelling of spacetime points — the diffeomorphism group is their built-in symmetry. When we work in the weak field, writing $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$ and keeping only first order in $h$ , that infinite symmetry does not vanish; it descends to a sharp, tractable statement about the perturbation field. An infinitesimal coordinate shift $x^\mu\to x^\mu+\xi^\mu$ , with $\xi^\mu$ as small as $h$ itself, leaves the frozen Minkowski background $\eta_{\mu\nu}$ untouched but reshuffles the components of $h_{\mu\nu}$ . This reshuffling — $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ — is the gauge freedom of linearized gravity, and it is nothing but the linearized Lie derivative of the background along $\xi$ . The structure is an exact copy of electromagnetism, with the dictionary shifted up by one tensor rank. There the potential $A_\mu$ carries gauge freedom $A_\mu\to A_\mu+\partial_\mu\lambda$ and only the field strength $F_{\mu\nu}$ is physical; here the potential is the rank-two field $h_{\mu\nu}$ , the gauge parameter is a whole four-vector $\xi^\mu$ instead of a single scalar $\lambda$ , and the physical object is the linearized Riemann tensor. Because there are four gauge functions rather than one, gravity can delete four spurious components of $h_{\mu\nu}$ — and a residual freedom deletes four more. The bookkeeping is the payoff. A symmetric $h_{\mu\nu}$ has ten components; four gauge functions fix the Lorenz gauge, and four residual harmonic functions carry a vacuum wave to the transverse-traceless gauge, leaving $10-4-4=2$ . Those two survivors are the plus and cross polarizations of a gravitational wave — the two helicity states of the massless spin-2 graviton. Everything downstream, from LIGO's strain to the quadrupole formula, is the disciplined use of this one freedom.

Key takeaways

In linearized gravity the metric splits as $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$ with $|h_{\mu\nu}|\ll1$ , and an infinitesimal coordinate change $x^\mu\to x^\mu+\xi^\mu$ shifts the perturbation by $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ while the geometry stays physically identical.
This is the linearized face of diffeomorphism invariance, and it is the exact gravitational analogue of the electromagnetic gauge transformation $A_\mu\to A_\mu+\partial_\mu\lambda$ — with a four-vector $\xi^\mu$ in place of one scalar $\lambda$ .
The linearized Riemann tensor is gauge invariant, so every genuinely physical effect — curvature, tidal acceleration, the strain a detector reads — is untouched by the choice of $\xi^\mu$ .
Four functions $\xi^\mu$ impose the Lorenz gauge $\partial^\mu\bar h_{\mu\nu}=0$ , collapsing the field equations to ten decoupled wave equations $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ ; residual gauge with $\Box\xi^\mu=0$ then reduces a vacuum wave to the transverse-traceless gauge.
Counting: $10$ metric components $-4$ gauge $-4$ residual/constraint $=2$ physical polarizations, matching the two helicity- $\pm2$ states of the graviton.
Gauge artifacts can perfectly mimic a wave; only gauge-invariant quantities carry physics — which is why LIGO measures a Riemann-driven proper-length change of order $2\times10^{-18}\ \mathrm{m}$ , not a coordinate.
Key idea: Under $x^\mu\to x^\mu+\xi^\mu$ the perturbation transforms as $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ ; the linearized Riemann tensor is invariant, so the four functions $\xi^\mu$ are pure gauge and reduce $h_{\mu\nu}$ from ten to two physical components.

Where this lives today

Gauge freedom is not a historical curiosity; it is the daily working language of gravitational-wave astronomy. Every LIGO–Virgo–KAGRA template is written in the transverse-traceless gauge, precisely because that gauge cleanly isolates the two physical polarizations $h_+$ and $h_\times$ from the mountain of coordinate degrees of freedom, and because in it the freely suspended mirrors sit at fixed coordinates while the proper light-travel distance oscillates. One of the sharpest ways to test general relativity against the data is a polarization test: general relativity permits exactly two tensor modes, but a generic metric theory could support up to six (adding scalar "breathing" and vector modes). A detection of any non-tensor polarization would mean the graviton is not purely spin-2 — that gauge freedom removes fewer than eight components — and current multi-detector networks already bound such extra modes to be small. At the low-frequency extreme, pulsar-timing arrays such as NANOGrav search the same TT structure imprinted on pulsar arrival times, using the gauge-invariant Hellings–Downs correlation as the signature of a real nanohertz background. A live frontier is gravitational-wave memory — a permanent, gauge-invariant offset in the separation of free masses left behind after a wave passes. In every case the discipline is Einstein's hard-won lesson: trust the curvature, not the coordinates.

The physical picture

Picture $h_{\mu\nu}$ as the numbers you get when you lay a coordinate grid over almost-flat spacetime and read off how far the metric departs from Minkowski. Wiggle the grid — relabel every point by $x^\mu\to x^\mu+\xi^\mu$ — and those numbers change, even though you have not bent spacetime one bit. The bending is curvature and it is real; the grid-wiggle is gauge and it is a fiction of your bookkeeping. The single unbreakable rule is that the physical curvature, the linearized Riemann tensor, comes out the same no matter how you wiggle. The first figure sets these two things side by side: the coordinate description on the left, which you can slosh around freely, and the invariant tidal deformation on the right, which you cannot.

Figure

Gauge versus physics — a coordinate wiggle ξ^μ changes the components of h_μν through h_μν → h_μν − ∂_μξ_ν − ∂_νξ_μ, yet the physical curvature R_μνρσ, read off as the tidal deformation of a ring of free-falling test masses, is left exactly unchanged

To decide whether something is real, you look at what two freely floating masses actually do. Their relative acceleration is set by the Riemann tensor through the geodesic-deviation equation, and that equation is gauge invariant — so no clever coordinate choice can erase a real wave, and no coordinate trick can conjure one. This is the whole working method of the field: compute a curvature, or the proper distance between test masses, never the bare components of $h_{\mu\nu}$ . The second figure turns the counting into a picture and attaches the real number that a detector cares about: ten components fall to two, and those two produce a proper-length change of order $10^{-18}\ \mathrm{m}$ across a LIGO arm.

Figure

The degree-of-freedom ladder 10 − 4 − 4 = 2 — the symmetric perturbation has 10 components, the four Lorenz-gauge functions ξ^μ remove 4, the four residual harmonic functions remove 4 more, leaving the 2 physical polarizations h₊ and h×; for a strain h ≈ 1×10⁻²¹ across a 4 km arm this is a proper-length change ΔL = ½hL ≈ 2×10⁻¹⁸ m

The core equations

Everything begins with the weak-field split: the true metric is Minkowski plus a small perturbation treated as a tensor field living on the flat background. This is the arena in which the whole discussion takes place, and the smallness of $h$ is what lets us drop every term of second order and higher.

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

where: $\eta_{\mu\nu}$ — Minkowski background, signature $(-,+,+,+)$ , held fixed; $h_{\mu\nu}$ — symmetric metric perturbation, ten independent components, treated as a field on flat space. Indices are raised and lowered with $\eta$ throughout, since corrections are higher order. The gauge freedom is generated by an infinitesimal change of coordinates. We shift every point by a smooth vector field $\xi^\mu$ that is as small as the perturbation itself, so that products $\xi\,h$ and $\xi\,\xi$ can be discarded to the order we keep.

x^\mu \to x'^\mu = x^\mu + \xi^\mu(x),\qquad |\partial_\nu\xi^\mu|\sim |h| \qquad (2)

where: $\xi^\mu$ — arbitrary smooth gauge vector, four free functions, first order in $h$ . It is the exact analogue of the electromagnetic gauge scalar $\lambda$ , promoted from one function to four. Feeding the coordinate change through the tensor transformation law and keeping first order gives the central result: the perturbation shifts by (minus) the symmetrized gradient of $\xi$ . This is the transformation that reshuffles $h_{\mu\nu}$ without touching the geometry.

h_{\mu\nu} \to h_{\mu\nu} - \partial_\mu \xi_\nu - \partial_\nu \xi_\mu \qquad (3)

where: $\xi_\mu=\eta_{\mu\nu}\xi^\nu$ — gauge covector; $\partial_\mu$ — flat-space partial derivative. The combination $\partial_\mu\xi_\nu+\partial_\nu\xi_\mu$ is exactly the Lie derivative $(\mathcal{L}_\xi\eta)_{\mu\nu}$ of the background along $\xi$ , which is why this is diffeomorphism invariance in miniature. It is often cleaner to work with the trace-reversed field $\bar h_{\mu\nu}$ , because that is what appears in the field equations. Its gauge law inherits an extra trace term from the $-\tfrac12\eta_{\mu\nu}h$ subtraction.

\bar h_{\mu\nu}\to \bar h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu+\eta_{\mu\nu}\,\partial_\alpha\xi^\alpha \qquad (4)

where: $\bar h_{\mu\nu}=h_{\mu\nu}-\tfrac12\eta_{\mu\nu}h$ — trace-reversed perturbation; $h=\eta^{\mu\nu}h_{\mu\nu}$ — the trace. The added $\eta_{\mu\nu}\,\partial_\alpha\xi^\alpha$ is precisely what makes the divergence transform cleanly in equation (6). The object that stays put is the linearized Riemann tensor. Substituting the gauge shift (3) and using the symmetry of mixed partials, every $\xi$ -term cancels against a partner — the curvature is gauge invariant, exactly as the electromagnetic field strength $F_{\mu\nu}$ is invariant under $A_\mu\to A_\mu+\partial_\mu\lambda$ .

R^{(1)}{}_{\mu\nu\rho\sigma}=\tfrac12\big(\partial_\rho\partial_\nu h_{\mu\sigma}+\partial_\sigma\partial_\mu h_{\nu\rho}-\partial_\sigma\partial_\nu h_{\mu\rho}-\partial_\rho\partial_\mu h_{\nu\sigma}\big),\qquad \delta R^{(1)}{}_{\mu\nu\rho\sigma}=0 \qquad (5)

where: $R^{(1)}{}_{\mu\nu\rho\sigma}$ — linearized Riemann tensor, the physical curvature and the sole carrier of tidal effects; $\delta$ — the change under a gauge transformation. This single line is why $h_{\mu\nu}$ by itself is never observable. Spending the four functions $\xi^\mu$ to kill the divergence of $\bar h_{\mu\nu}$ defines the Lorenz (harmonic) gauge, and it always exists because the required $\xi$ solves an ordinary inhomogeneous wave equation. In that gauge the field equations become ten decoupled wave equations.

\Box\,\xi_\nu=\partial^\mu\bar h^{\text{(old)}}_{\mu\nu}\;\Longrightarrow\;\partial^\mu\bar h_{\mu\nu}=0,\qquad \Box\,\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu} \qquad (6)

where: $\Box=\eta^{\mu\nu}\partial_\mu\partial_\nu$ — flat-space d'Alembertian; $G$ — Newton's constant; $T_{\mu\nu}$ — stress-energy source; units $c=1$ . The left equation picks the gauge, the right one is the linearized Einstein equation it produces. The final accounting is a subtraction. Ten components, minus four gauge functions, minus four residual/constraint conditions, equals two — the physical count that the electromagnetic analogue reproduces as $4-1-1=2$ .

10 - 4 - 4 = 2 \qquad\Longleftrightarrow\qquad 4 - 1 - 1 = 2\ \text{(EM)} \qquad (7)

where: $10$ — components of symmetric $h_{\mu\nu}$ ; the first $4$ — gauge functions $\xi^\mu$ ; the second $4$ — residual harmonic conditions reaching the TT gauge; $2$ — physical polarizations, the helicity- $\pm2$ states of the graviton (compare the photon's helicity- $\pm1$ ).

Derivation AThe gauge law and why curvature does not care

The goal here is to earn equation (3) from first principles and then immediately cash it out as equation (5): first show that an infinitesimal coordinate change $x^\mu\to x^\mu+\xi^\mu$ forces $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ by pushing the metric through the ordinary rank-two tensor transformation law and keeping only first order, and then show by direct substitution that this same shift annihilates the linearized Riemann tensor, so the freedom is genuinely unphysical. Both halves use nothing but Taylor expansion and the commutativity of partial derivatives.

Step 1. Invert the coordinate change to the order we keep. From $x'^\mu=x^\mu+\xi^\mu$ the Jacobian is $\partial x^\alpha/\partial x'^\mu=\delta^\alpha_\mu-\partial_\mu\xi^\alpha+\mathcal{O}(\xi^2)$ . Why: the transformation law needs the inverse Jacobian, and to first order inverting a shift just flips the sign of its gradient.

Step 2. Transform the metric as a covariant rank-two tensor, $g'_{\mu\nu}=\dfrac{\partial x^\alpha}{\partial x'^\mu}\dfrac{\partial x^\beta}{\partial x'^\nu}\,g_{\alpha\beta}$ , and insert $g_{\alpha\beta}=\eta_{\alpha\beta}+h_{\alpha\beta}$ . Why: $g_{\mu\nu}$ is an honest tensor, so its components in the new chart are fixed entirely by this law — no physics has been assumed, only a relabelling.

Step 3. Expand and discard products of small quantities. Since $\eta$ is constant, $\eta_{\alpha\beta}\partial_\mu\xi^\alpha=\partial_\mu\xi_\beta$ , and the cross terms give $g'_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu+\mathcal{O}(2)$ . Reading off the new perturbation, $h'_{\mu\nu}=h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ . Why: the background is untouched because $\mathcal{L}_\xi\eta$ is a pure gradient, and only the perturbation absorbs the change — this is equation (3), equivalently $h'_{\mu\nu}=h_{\mu\nu}-(\mathcal{L}_\xi\eta)_{\mu\nu}$ .

Step 4. Now insert $\delta h_{\mu\nu}=-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ into the linearized Riemann tensor of equation (5). Each of the four terms produces two third-derivative pieces of $\xi$ ; for example $\partial_\rho\partial_\nu\,\delta h_{\mu\sigma}$ contributes $-\partial_\rho\partial_\nu\partial_\mu\xi_\sigma-\partial_\rho\partial_\nu\partial_\sigma\xi_\mu$ . Why: writing every term out is the whole trick — the cancellation is visible only once the third derivatives are on the page.

Step 5. Collect the eight pieces. Because partial derivatives commute, $\partial_\rho\partial_\nu\partial_\mu\xi_\sigma$ from the first term is cancelled by $\partial_\rho\partial_\mu\partial_\nu\xi_\sigma$ from the fourth, and likewise for the remaining three pairs, giving $\delta R^{(1)}{}_{\mu\nu\rho\sigma}=0$ . Why: the antisymmetric pairing of the Riemann tensor is exactly engineered so that a symmetrized gradient drops out — the same structural reason $F_{\mu\nu}=\partial_\mu A_\nu-\partial_\nu A_\mu$ kills $\partial_\mu\lambda$ .

Go deeper · Derivation B · 1 minReaching the Lorenz gauge

Derivation BReaching the Lorenz gauge

The aim is to prove equation (6): that the four gauge functions $\xi^\mu$ can always be chosen to enforce $\partial^\mu\bar h_{\mu\nu}=0$ , and that once this holds the linearized Einstein equations collapse from a coupled mess into ten independent flat-space wave equations with the stress-energy as source. The engine is the trace-reversed gauge law (4); applying a divergence to it turns the gauge choice into a solvable wave equation for $\xi$ , and the existence of retarded solutions guarantees the gauge is reachable from any starting configuration.

Step 1. Take the divergence of the trace-reversed gauge law (4). The three $\xi$ -terms give $\partial^\mu\bar h'_{\mu\nu}=\partial^\mu\bar h_{\mu\nu}-\Box\xi_\nu-\partial_\nu(\partial^\mu\xi_\mu)+\partial_\nu(\partial_\alpha\xi^\alpha)$ . Why: differentiating the whole transformation is what exposes how the divergence — the quantity we want to zero — responds to a gauge choice.

Step 2. Notice the last two terms are identical and cancel, leaving the clean law $\partial^\mu\bar h'_{\mu\nu}=\partial^\mu\bar h_{\mu\nu}-\Box\xi_\nu$ . Why: this cancellation is the entire reason for the extra $\eta_{\mu\nu}\partial_\alpha\xi^\alpha$ term in (4); the trace-reversed field is engineered so its divergence transforms by a plain d'Alembertian of $\xi$ .

Step 3. Demand the new divergence vanish. That requires $\Box\xi_\nu=\partial^\mu\bar h^{\text{(old)}}_{\mu\nu}$ , an inhomogeneous wave equation with a fixed source, which always has a (retarded) solution. Why: solvability is what makes the Lorenz gauge universally attainable — no matter how awkward the initial $h_{\mu\nu}$ , a suitable $\xi$ exists to straighten it.

Step 4. Impose $\partial^\mu\bar h_{\mu\nu}=0$ in the linearized Einstein equation. The linearized Einstein tensor $G^{(1)}_{\mu\nu}=-\tfrac12\big(\Box\bar h_{\mu\nu}-\partial^\alpha\partial_\mu\bar h_{\alpha\nu}-\partial^\alpha\partial_\nu\bar h_{\alpha\mu}+\eta_{\mu\nu}\partial^\alpha\partial^\beta\bar h_{\alpha\beta}\big)$ loses its last three terms, leaving $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ . Why: the divergence-free condition kills exactly the terms that couple different components, decoupling the system into ten scalar wave equations — the doorway to retarded solutions and the quadrupole formula.

Go deeper · Derivation C · 1 minResidual freedom and the transverse-traceless gauge

Derivation CResidual freedom and the transverse-traceless gauge

The target is the transverse-traceless (TT) description of a source-free wave and the final degree-of-freedom count. Fixing the Lorenz gauge does not exhaust the freedom: any further $\xi^\mu$ obeying $\Box\xi^\mu=0$ preserves $\partial^\mu\bar h_{\mu\nu}=0$ , and these four harmonic functions are just enough to zero the time components and the trace of a vacuum wave. We take a plane-wave ansatz, read off the constraints the Lorenz gauge already imposes on the amplitude, then spend the residual freedom to strip the amplitude down to the two independent numbers that survive.

Step 1. Identify the residual freedom. From Step 2 of the previous derivation, a further gauge shift changes the divergence by $-\Box\xi_\nu$ , so staying in Lorenz gauge requires $\Box\xi_\nu=0$ — four harmonic functions still free. Why: gauge-fixing is rarely complete; the leftover harmonic $\xi$ is precisely what carries a wave from Lorenz to TT form.

Step 2. Insert a vacuum plane wave $\bar h_{\mu\nu}=\mathrm{Re}\!\left(C_{\mu\nu}\,e^{ik_\alpha x^\alpha}\right)$ into $\Box\bar h_{\mu\nu}=0$ , forcing $k_\alpha k^\alpha=0$ — the wave is null, moving at the speed of light. The Lorenz condition becomes $k^\mu C_{\mu\nu}=0$ , four algebraic constraints on the ten-component amplitude. Why: the null condition is the physics (gravitational waves are luminal), and the transversality $k^\mu C_{\mu\nu}=0$ is the first four of the eight components we will remove.

Step 3. Spend the residual freedom. Writing the harmonic gauge vector as $\xi_\mu=\mathrm{Re}\!\left(i\,\epsilon_\mu\,e^{ik_\alpha x^\alpha}\right)$ with constant $\epsilon_\mu$ , the amplitude shifts by $C_{\mu\nu}\to C_{\mu\nu}+k_\mu\epsilon_\nu+k_\nu\epsilon_\mu-\eta_{\mu\nu}k_\alpha\epsilon^\alpha$ . The four constants $\epsilon_\mu$ let us impose tracelessness $C^\mu{}_\mu=0$ and orthogonality $C_{\mu\nu}u^\nu=0$ to a chosen observer four-velocity $u^\nu$ , removing the time components $C_{0\mu}$ . Why: these are the remaining four conditions; together with the four from Step 2 they account for all eight non-physical components.

Step 4. Read off the survivors. For a wave along $z$ the amplitude collapses to $C_{\mu\nu}=\begin{pmatrix}0&0&0&0\\0&C_+&C_\times&0\\0&C_\times&-C_+&0\\0&0&0&0\end{pmatrix}$ — exactly two independent numbers, the plus and cross polarizations. The count is $10-4-4=2$ . Why: these two amplitudes are the physical content of a massless spin-2 field, and the TT gauge is reachable only in vacuum, where $\Box h_{\mu\nu}=0$ holds; inside a source one keeps Lorenz but cannot complete the projection.

Limiting cases

Case	Result	Meaning
Lorenz gauge fixing	$\partial^\mu \bar h_{\mu\nu}=0$	four functions $\xi^\mu$ solving $\Box\xi_\nu=\partial^\mu\bar h^{\text{old}}_{\mu\nu}$ ; always achievable, decouples the field equations
Residual freedom in Lorenz gauge	$\Box\,\xi^\mu=0$	transformations preserving $\partial^\mu\bar h_{\mu\nu}=0$ ; four harmonic functions used to reach the TT gauge
Transverse-traceless (vacuum) gauge	$h^{\text{TT}}_{0\mu}=0,\ h^{\text{TT}\,i}{}_i=0,\ \partial^j h^{\text{TT}}_{ij}=0$	reachable only where $T_{\mu\nu}=0$ ; leaves exactly the two polarizations $h_+,h_\times$
Newtonian / static limit	$h_{00}=-2\Phi,\ h_{ij}=-2\Phi\,\delta_{ij}$	gauge fixed to the longitudinal (Poisson) form; the physical content is the potential $\Phi$ , not the individual components
Electromagnetic analogue	$4-1-1=2$	one gauge scalar $\lambda$ , one residual, two helicity- $\pm1$ photon states — same logic, lower rank

Common trapReading physics off the components of h_μν

The single most common error in linearized gravity is to treat $h_{\mu\nu}$ — or worse, one component of it — as if it were observable. It is not. A gauge transformation can make individual components appear or vanish at will: you can write down a perfectly time-dependent, wave-looking $h_{\mu\nu}$ whose linearized Riemann tensor is identically zero (worked below), and you can hide a genuine wave behind an unlucky coordinate choice. Only gauge-invariant quantities carry physics: the linearized Riemann tensor, the relative acceleration of freely falling masses, the radiated luminosity, the strain a detector reads. A companion trap is to forget that the full TT gauge exists only in vacuum — where $T_{\mu\nu}=0$ . Inside matter you can still impose the Lorenz gauge, but the extra TT conditions rely on $\Box h_{\mu\nu}=0$ and fail wherever there is a source. When in doubt, compute a curvature; never a bare component.

Worked examples

1 · The strain LIGO measures is a curvature, not a coordinate (fully worked)

A gravitational wave of peak strain $h\approx1.0\times10^{-21}$ — the value at the peak of GW150914 — passes a LIGO detector whose arms are $L=4\ \mathrm{km}$ long, with freely suspended mirrors at the ends. Take the $+$ polarization aligned with the arms. How far does an arm actually change length, and why is that number gauge invariant?

Step 1. Write the physical relation. In the TT gauge the mirror coordinates do not move; the proper distance between them changes because the metric does. For two masses separated by $L$ along $x$ under a $+$ wave, $\delta L/L=\tfrac12\,h^{\text{TT}}_{xx}=\tfrac12 h$ . Equivalently, the gauge-invariant geodesic-deviation equation $\ddot\xi^i=\tfrac12\,\ddot h^{\text{TT}}_{ij}\,\xi^j$ gives the same relative acceleration — the two pictures agree because both are driven by the same Riemann component $R^{(1)}_{i0j0}=-\tfrac12\ddot h^{\text{TT}}_{ij}$ .

Step 2. Put in numbers. $\delta L=\tfrac12\,h\,L=\tfrac12\,(1.0\times10^{-21})(4\times10^{3}\ \mathrm{m})=\mathbf{2\times10^{-18}\ m}$ . For scale, a proton is about $1.7\times10^{-15}\ \mathrm{m}$ across, so this displacement is roughly $\delta L\approx1.2\times10^{-3}$ of a proton width — about one-thousandth of a proton.

Step 3. Argue the invariance. Compute the wave in any other gauge — add any $\xi^\mu$ you like — and the components of $h_{\mu\nu}$ change, but $R^{(1)}_{i0j0}$ does not (equation 5), so $\delta L$ does not. The mirrors' proper separation is a gauge-invariant observable; the coordinate you attach to them is not. This is why a $2\times10^{-18}\ \mathrm{m}$ number is meaningful at all — it is set by curvature, the one thing a change of coordinates cannot touch.

2 · A wave that is not there (pure gauge, zero curvature)

Start from flat spacetime, $h_{\mu\nu}=0$ , and apply the gauge vector $\xi_\mu=\big(C\cos[\omega(t-z)],\,0,\,0,\,0\big)$ propagating along $z$ . Show that the generated $h_{\mu\nu}$ oscillates like a wave yet carries no curvature.

Step 1. Generate the perturbation from the gauge law $h_{\mu\nu}=-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ . Only $\xi_0$ is nonzero, so $h_{00}=-2\partial_0\xi_0=2C\omega\sin[\omega(t-z)]$ and $h_{0z}=h_{z0}=-\partial_z\xi_0=-C\omega\sin[\omega(t-z)]$ , with every other component zero. These are manifestly oscillating functions of $t-z$ — to the untrained eye, a wave moving at the speed of light.

Step 2. Test it with the one honest question: is there curvature? Evaluate the relevant Riemann component $R^{(1)}_{0z0z}=\tfrac12\big(2\partial_0\partial_z h_{0z}-\partial_z^2 h_{00}-\partial_0^2 h_{zz}\big)$ . With $h_{zz}=0$ , the two surviving terms are $2\partial_0\partial_z h_{0z}=-2C\omega^3\sin[\omega(t-z)]$ and $-\partial_z^2 h_{00}=+2C\omega^3\sin[\omega(t-z)]$ .

Step 3. Add them. $R^{(1)}_{0z0z}=\tfrac12\big(-2C\omega^3\sin+2C\omega^3\sin\big)=\mathbf{0}$ , and the same cancellation kills every other component — as it must, since $h_{\mu\nu}=-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ is pure gauge. Two free masses placed in this "wave" feel zero relative acceleration: it is a coordinate ripple, not radiation. The lesson of Example 1 in reverse — an oscillating $h_{\mu\nu}$ with $R^{(1)}=0$ is a fiction of the grid.

3 · Counting the freedom against electromagnetism

Verify the physical polarization count for gravity and check that the electromagnetic case follows the same recipe.

Step 1. Gravity. The symmetric $h_{\mu\nu}$ has $\tfrac{4\times5}{2}=\mathbf{10}$ independent components. The Lorenz gauge $\partial^\mu\bar h_{\mu\nu}=0$ is four conditions, fixed by the four functions $\xi^\mu$ — subtract $4$ . The residual harmonic freedom $\Box\xi^\mu=0$ supplies four more usable conditions (traceless plus $C_{0\mu}=0$ ) — subtract $4$ . Result: $10-4-4=\mathbf{2}$ .

Step 2. Electromagnetism. The potential $A_\mu$ has $\mathbf{4}$ components; the Lorenz gauge $\partial^\mu A_\mu=0$ removes one, and residual gauge with $\Box\lambda=0$ removes one more, leaving $4-1-1=\mathbf{2}$ .

Step 3. Read the pattern. Both are massless fields, so both keep exactly two transverse states — but the helicities differ: $\pm1$ for the spin-1 photon, $\pm2$ for the spin-2 graviton. The doubled gauge structure (a four-vector $\xi^\mu$ versus a scalar $\lambda$ ) is the fingerprint of the higher spin, and it is why gravitational radiation has no monopole or dipole mode.

Physics of the portrait

Gauge freedom in linearized gravity is where Einstein's own struggle with the meaning of coordinates comes to a head. He built general relativity on the principle that coordinates are arbitrary labels, then spent years working out — and once nearly denying — that the waves his weak-field equations predicted were physical rather than coordinate artifacts.

Origins · 1 minAlbert Einstein and how this was discovered

Albert Einstein

1879–1955 · linearized gravity and the reality of gravitational waves

Einstein reached the generally covariant field equations in November 1915 only after abandoning his own 1913 "hole argument," in which he had convinced himself that general covariance was impossible because two coordinate descriptions of the same matter distribution seemed to give different fields — the very ambiguity we now name gauge freedom and treat as a symmetry rather than a paradox. Within months, in his 1916 paper on the approximate integration of the field equations, he linearized about flat space, imposed the harmonic coordinate condition that is the direct ancestor of the Lorenz gauge, and found wave solutions travelling at the speed of light — predicting gravitational waves and, in 1918, deriving the quadrupole formula (after correcting a factor-of-two slip). Yet the coordinate confusion never fully released him: in 1936 he and Nathan Rosen submitted a paper arguing gravitational waves do not exist, having mistaken a coordinate singularity in cylindrical waves for a genuine one. The referee — now known to be Howard Percy Robertson — caught the gauge artifact; Einstein, initially furious, revised the paper to the opposite conclusion. The reality of the waves was settled only in 1957, at Chapel Hill, by the "sticky bead" argument that a passing wave does mechanical work on beads free to slide on a rod — a gauge-invariant effect no coordinate choice can remove. Einstein's arc, from the hole argument to the 1936 error, is the whole physics of this article lived out: the perennial danger of reading physics off $h_{\mu\nu}$ instead of off the curvature.

History

1916

— Einstein linearizes the field equations, imposes harmonic coordinates, and predicts gravitational waves

1918

— Einstein derives and corrects the quadrupole formula for radiated power

1936

— the Einstein–Rosen paper wrongly denies gravitational waves, misreading a gauge artifact as a singularity

1957

— the Chapel Hill sticky-bead argument shows the waves do physical work and are not pure gauge

2016

— LIGO's GW150914 directly detects the two transverse-traceless polarizations

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §18.1–18.4 and §35.4 · Wald, General Relativity, §4.4 · Carroll, Spacetime and Geometry, §7.1–7.2 · Hartle, Gravity, ch. 16 · Schutz, A First Course in General Relativity, ch. 8–9 · Maggiore, Gravitational Waves, Vol. 1, §1.2 and §2.2.

Provenance. The gauge law $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ , the invariance $\delta R^{(1)}{}_{\mu\nu\rho\sigma}=0$ , the Lorenz-gauge wave equation $\Box\bar h_{\mu\nu}=-16\pi G\,T_{\mu\nu}$ , and the TT amplitude matrix are cross-checked against Carroll §7.1–7.2, Wald §4.4, and MTW §18 and §35.4. A verifier should: (1) recover equation (3) from the rank-two tensor transformation law to first order in $\xi$ ; (2) confirm $\delta R^{(1)}{}_{\mu\nu\rho\sigma}=0$ by direct substitution using $\partial_a\partial_b=\partial_b\partial_a$ ; (3) reproduce $\Box\xi_\nu=\partial^\mu\bar h_{\mu\nu}$ and the decoupled field equation; (4) verify $10-4-4=2$ and the $+z$ amplitude matrix; (5) check the worked numbers $\delta L=\tfrac12 hL=2\times10^{-18}\ \mathrm{m}$ for $h=10^{-21}$ , $L=4\ \mathrm{km}$ , and $R^{(1)}_{0z0z}=0$ for the pure-gauge Example 2. The only empirical number quoted, the GW150914 peak strain $h\approx1.0\times10^{-21}$ , is the standard published value.

Prerequisites. Metric Perturbation and the Trace-Reversed Field · Linearized Einstein Equations · Diffeomorphisms and the Lie Derivative. Leads to. The Lorenz (Harmonic) Gauge · The Transverse-Traceless Gauge · Helicity and the Spin-2 Graviton · Plus and Cross Polarizations.

Ask the tutor: Ask me to walk through why the linearized Riemann tensor is gauge invariant term by term, or to build the pure-gauge "fake wave" of Example 2 and watch its curvature vanish.

Is gauge freedom the same thing as general covariance?

They are the linearized and full versions of one idea. General covariance says the full field equations hold in any coordinate system, so any smooth diffeomorphism maps a solution to a physically identical solution. When you fix a background — here Minkowski — and keep only first order in the perturbation, that infinite diffeomorphism symmetry shrinks to the specific transformation $h_{\mu\nu}\to h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ , generated by infinitesimal shifts $\xi^\mu$ . So gauge freedom is what general covariance becomes once you split $g=\eta+h$ : the same symmetry, now acting linearly on a tensor field over flat space, which is exactly why it looks and behaves like the gauge symmetry of a field theory.

Why is the linearized Riemann tensor singled out as physical?

Because it is gauge invariant and it drives the one thing you can actually measure. Substituting the gauge shift $\delta h_{\mu\nu}=-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ into the linearized Riemann tensor gives zero, term by term, because its antisymmetric structure kills any symmetrized gradient — the same reason the electromagnetic $F_{\mu\nu}$ ignores $\partial_\mu\lambda$ . And physically the Riemann tensor is what appears in the geodesic-deviation equation, so it sets the relative acceleration of freely falling masses — the tidal effect a detector reads out. A quantity that is both invariant under the symmetry and directly tied to an observable is the definition of "physical," and the bare components of $h_{\mu\nu}$ fail both tests.

How can a wave-looking h_μν carry no energy or curvature?

Take any gauge vector $\xi^\mu$ that oscillates, and build $h_{\mu\nu}=-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu$ . This is a pure gauge transformation of flat space, so by construction its linearized Riemann tensor vanishes identically, even though the components of $h_{\mu\nu}$ ripple along at the speed of light (Worked Example 2 does this explicitly). No curvature means no tidal force, no work done on free masses, and no radiated energy. The moral is that "looks like a wave" is a statement about coordinates, while "is a wave" is a statement about curvature. Only when the Riemann tensor is nonzero — genuinely, in every gauge — do you have radiation.

What is the difference between the Lorenz gauge and the TT gauge?

The Lorenz gauge $\partial^\mu\bar h_{\mu\nu}=0$ is four conditions that you can impose everywhere, in vacuum or inside matter; it decouples the field equations into ten wave equations and is the natural setting for retarded solutions and the quadrupole formula. The transverse-traceless gauge is a further specialization, using the residual harmonic freedom ( $\Box\xi^\mu=0$ ) to also set $h_{0\mu}=0$ and make the spatial part traceless and transverse. That extra step relies on $\Box h_{\mu\nu}=0$ , which holds only in vacuum, so the TT gauge is a radiation-zone luxury: it is the cleanest description of a free wave, displaying just $h_+$ and $h_\times$ , but it cannot be reached where a source sits.

Does gauge freedom mean gravitational waves are not real?

No — and this is precisely the confusion Einstein himself fell into in 1936. Gauge freedom means some apparent "waves" are coordinate artifacts, but it also gives you an infallible test: compute a gauge-invariant quantity. A real wave has nonzero linearized Riemann tensor in every gauge, does mechanical work on free masses (the Chapel Hill sticky-bead argument), and carries energy captured by the Isaacson stress-energy. GW150914 moved LIGO's mirrors by a curvature-driven, gauge-invariant $2\times10^{-18}\ \mathrm{m}$ . Gauge freedom does not threaten the reality of waves; it disciplines the bookkeeping so that only the genuinely physical part survives, and everything that survives that filter is as real as any tidal force.

Why exactly two polarizations, and why helicity ±2?

The two comes from the counting $10-4-4=2$ : ten components of a symmetric $h_{\mu\nu}$ , minus four killed by the Lorenz gauge, minus four more by the residual harmonic freedom. What remains for a wave along $z$ is a symmetric, traceless, transverse $2\times2$ block with just two independent numbers, $h_+$ and $h_\times$ . Their helicity is $\pm2$ because under a rotation by angle $\psi$ about the propagation axis the combination $h_+\pm i h_\times$ picks up a phase $e^{\mp 2i\psi}$ — twice the rotation angle, the signature of a spin-2 field. Contrast the photon, whose analogue rotates by $e^{\mp i\psi}$ : helicity $\pm1$ , spin-1. The doubled gauge structure of gravity is the direct fingerprint of the doubled spin.