Transverse-Traceless Gauge

A metric perturbation $h_{\mu\nu}$ in linearized gravity is a symmetric $4\times4$ array — ten independent numbers at every point. Yet a gravitational wave in empty space carries only two physical degrees of freedom, the same count Maxwell's theory assigns to the photon. Where did the other eight go? Four are pure constraint, fixed the moment one adopts the Lorenz (harmonic) gauge; four more are pure coordinate freedom, the residual liberty to relabel spacetime points without touching the geometry. The transverse-traceless gauge is the disciplined bookkeeping that spends every one of those eight and lays the surviving two bare on the table. It is the exact gravitational analogue of choosing radiation gauge in electromagnetism, where $A^0=0$ and $\nabla\!\cdot\mathbf{A}=0$ strip the four-potential down to its two transverse photon polarizations. The construction is a two-stage descent. Stage one imposes $\partial^\mu\bar h_{\mu\nu}=0$ , which decouples Einstein's equations into flat-space wave equations and cuts ten components to six. This does not fix the gauge completely: a residual family of coordinate shifts $\xi^\mu$ obeying $\Box\xi^\mu=0$ survives. Stage two spends that residual freedom, in vacuum, to zero the trace and the time components — and the field equations then force the spatial part to be divergence-free. What remains is $h^{\mathrm{TT}}_{ij}$ : purely spatial, transverse to the propagation direction, and traceless. For a wave along $\hat z$ it collapses to a single $2\times2$ symmetric traceless block — the plus and cross polarizations. Crucially, in TT coordinates freely-falling masses stay put; the wave warps the ruler between them, not the lattice. That is why $h^{\mathrm{TT}}_{ij}$ is exactly what a detector measures and what the quadrupole formula delivers.

Key takeaways

The transverse-traceless (TT) gauge is a vacuum specialization of the Lorenz gauge that spends the residual coordinate freedom to expose the two physical, propagating degrees of freedom of a gravitational wave.
Its three defining conditions are purely spatial $h^{\mathrm{TT}}_{0\mu}=0$ , transverse $\partial^j h^{\mathrm{TT}}_{ij}=0$ , and traceless $h^{i\,\mathrm{TT}}_{i}=0$ .
The degree-of-freedom ledger reads $10-4-4=2$ : ten components, minus four from the Lorenz condition, minus four spent by the residual harmonic freedom $\Box\xi^\mu=0$ .
Because the field is traceless, the trace-reversed and ordinary perturbations coincide, $\bar h_{\mu\nu}=h_{\mu\nu}$ , and the field equation collapses to the flat-space wave equation $\Box h^{\mathrm{TT}}_{ij}=0$ .
TT gauge is a vacuum statement; the projector $\Lambda_{ij,kl}$ that builds it is nonlocal, so a source region ( $T_{\mu\nu}\neq0$ ) cannot be put in TT gauge point-by-point.
The wave's physical effect is tidal: it stretches and squeezes a ring of freely-falling masses transverse to propagation, area-preserving to first order — the principle behind interferometric detection.
Key idea: spend the residual gauge freedom of the Lorenz gauge, in vacuum, to make the spatial perturbation transverse and traceless; the resulting $h^{\mathrm{TT}}_{ij}$ holds exactly the two gauge-invariant polarizations of a gravitational wave, and freely-falling particles stay at fixed TT coordinates while the proper distance between them oscillates.

The core insight

Spend the residual gauge freedom of the Lorenz gauge, in vacuum, to make the spatial perturbation transverse and traceless; the resulting $h^{\mathrm{TT}}_{ij}$ holds exactly the two gauge-invariant polarizations of a gravitational wave, and freely-falling particles stay at fixed TT coordinates while proper distances between them oscillate.

Where this lives today

The transverse-traceless amplitude is the working currency of gravitational-wave astronomy. When LIGO, Virgo, and KAGRA report a detection, the number they publish — the strain time series $h(t)$ — is literally the TT field projected onto each detector's arm geometry, and the matched-filter template banks that dig those signals out of the noise are libraries of predicted $h^{\mathrm{TT}}_{ij}(t)$ computed from post-Newtonian and numerical-relativity waveforms. The quadrupole formula $h^{\mathrm{TT}}_{ij}=(2G/r)\,\Lambda_{ij,kl}\,\ddot Q_{kl}(t-r)$ is where those predictions begin, with the very projector of equation (5) doing the gauge cleanup. The energy budget of a merger — how much of a black hole's mass is radiated, about $3M_\odot$ worth for GW150914 — is computed from the Isaacson stress-energy $\langle T^{\mathrm{GW}}_{\mu\nu}\rangle=(32\pi G)^{-1}\langle\partial_\mu h^{\mathrm{TT}}_{ij}\,\partial_\nu h^{ij\,\mathrm{TT}}\rangle$ , which is meaningful only for the gauge-invariant TT components. And the frontier instruments extend the same language across the spectrum: pulsar timing arrays (NANOGrav's 2023 evidence for a nanohertz background) and the planned space interferometer LISA both search for $h^{\mathrm{TT}}_{ij}$ , just at frequencies fifteen orders of magnitude apart. Every one of them is measuring the two numbers that survive the $10-4-4=2$ ledger.

The physical picture

Picture the ten components of $h_{\mu\nu}$ as a fixed budget. Four are spent enforcing the Lorenz condition $\partial^\mu\bar h_{\mu\nu}=0$ ; four more are spent by the residual harmonic freedom $\Box\xi^\mu=0$ , the leftover liberty to slide coordinates around without disturbing the curvature. What survives, $10-4-4=2$ , is the irreducible content of a massless spin-2 field. TT gauge is nothing more than the accounting that makes this subtraction visible, and its output is a single $2\times2$ traceless matrix carrying two numbers, $h_+$ and $h_\times$ . A gravitational wave has no monopole and no dipole — it is a purely tidal field, a quadrupolar stretch-and-squeeze. Transversality says the disturbance lives entirely in the plane perpendicular to the propagation direction $\hat n$ ; tracelessness says it preserves transverse area to first order, so whatever it stretches along $x$ it compresses along $y$ . That area-preserving shear is the visual fingerprint of the plus and cross patterns, and the first figure shows it as a ring of test masses breathing through one cycle.

Figure

A ring of freely-falling test masses under a passing wave along the line of sight: the plus polarization in amber stretches along x while squeezing along y then reverses each half cycle, the cross polarization in blue does the same along the 45 degree diagonals, and the fractional stretch obeys δL divided by L equals one half h₊ — transverse area is preserved to first order because the wave is traceless

In TT coordinates the masses themselves do not move: the geodesic equation gives $\mathrm{d}^2x^i/\mathrm{d}\tau^2=0$ for a particle that starts at rest, because the TT frame is comoving with free fall. All the physics hides in the metric — the proper distance between two fixed-coordinate masses oscillates even though their coordinate values are frozen. The coordinates are a free-falling scaffold; the wave warps the ruler laid across it. This is exactly why a coordinate description alone is treacherous in general relativity, and why the gauge-invariant TT amplitude, tied to real geodesic deviation, is the trustworthy observable. The second figure makes the counting concrete: it tracks the budget from ten components down to two and shows the traceless matrix that survives. The projection that extracts $h^{\mathrm{TT}}_{ij}$ from a generic $h_{ij}$ is built from the transverse projector $P_{ij}=\delta_{ij}-n_in_j$ , and it is intrinsically nonlocal in position space (it hides an inverse Laplacian). That nonlocality is the deep reason TT gauge is a clean idea only in vacuum, far from where the radiation was born.

Figure

The gauge budget of a metric perturbation: ten components minus four fixed by the Lorenz condition minus four spent by the residual harmonic freedom leaves exactly two physical degrees of freedom, the plus and cross amplitudes of the transverse-traceless matrix, so the ledger reads 10 minus 4 minus 4 equals 2

The core equations

Everything begins with the three algebraic conditions that define the gauge. They say the perturbation has no time components, no divergence on its spatial indices, and no trace. Two of the ten components can survive all three.

h^{\mathrm{TT}}_{0\mu}=0,\qquad \partial^{j}h^{\mathrm{TT}}_{ij}=0,\qquad \delta^{ij}h^{\mathrm{TT}}_{ij}=h^{i\,\mathrm{TT}}_{i}=0 \qquad (1)

where: $h^{\mathrm{TT}}_{\mu\nu}$ — the perturbation in transverse-traceless gauge; $\partial^j h^{\mathrm{TT}}_{ij}$ — spatial divergence, the transversality condition; $\delta^{ij}h^{\mathrm{TT}}_{ij}$ — the trace, taken with the flat spatial metric $\mathrm{diag}(1,1,1)$ . Read together: "purely spatial, divergence-free, and traceless." The entire gauge freedom of linearized gravity is a coordinate shift $x^\mu\to x^\mu+\xi^\mu$ . To first order this changes the perturbation without changing any curvature invariant, so it is genuine redundancy, not physics.

h_{\mu\nu}\;\longrightarrow\; h_{\mu\nu}-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu \qquad (2)

where: $\xi^\mu$ — an arbitrary infinitesimal coordinate shift; the combination $\partial_\mu\xi_\nu+\partial_\nu\xi_\mu$ is the Lie derivative of the flat metric along $\xi$ . Because the linearized Riemann tensor is invariant under (2), any two $h_{\mu\nu}$ differing by such a term describe the same spacetime. Fixing the Lorenz gauge does not exhaust this freedom. The shifts that preserve $\partial^\mu\bar h_{\mu\nu}=0$ are exactly the harmonic ones, four independent functions of spacetime.

\Box\,\xi^\mu=0,\qquad \Box\equiv-\partial_t^2+\nabla^2 \qquad (3)

where: $\Box$ — the flat d'Alembertian; $\xi^\mu$ — the residual gauge vector, four free solutions of the wave equation. These four functions are precisely the budget spent in stage two to reach TT gauge. Once the perturbation is traceless, the trace-reversal operation becomes the identity, and the linearized vacuum field equation collapses to a bare wave equation for the spatial components.

\bar h_{\mu\nu}=h_{\mu\nu}-\tfrac12\eta_{\mu\nu}h \;\;\xrightarrow{\,h=0\,}\;\; \bar h^{\mathrm{TT}}_{\mu\nu}=h^{\mathrm{TT}}_{\mu\nu},\qquad \Box\,h^{\mathrm{TT}}_{ij}=0 \qquad (4)

where: $h=\eta^{\mu\nu}h_{\mu\nu}$ — the four-dimensional trace, zero in TT gauge; $\bar h_{\mu\nu}$ — the trace-reversed perturbation. With $h=0$ there is nothing to reverse, so $\bar h_{\mu\nu}$ and $h_{\mu\nu}$ are the same object and each satisfies the vacuum wave equation. To build a TT field from a generic spatial perturbation one needs the transverse projector and, from it, the transverse-traceless projector. The first kills any component along $\hat n$ ; the second additionally removes the trace.

P_{ij}(\hat n)=\delta_{ij}-n_in_j,\qquad \Lambda_{ij,kl}(\hat n)=P_{ik}P_{jl}-\tfrac12 P_{ij}P_{kl} \qquad (5)

where: $\hat n$ — unit propagation direction; $P_{ij}$ — projects a single index into the plane $\perp\hat n$ , with $P_{ij}n_j=0$ and $P_{ii}=2$ ; $\Lambda_{ij,kl}$ — projects a symmetric tensor onto its transverse-traceless part. $\Lambda$ is transverse, traceless, and idempotent — a genuine projection operator. Applying $\Lambda$ to any spatial $h_{kl}$ delivers its physical TT content; the result is a symmetric traceless transverse matrix that, for propagation along $\hat z$ , decomposes onto the two polarization basis tensors.

h^{\mathrm{TT}}_{ij}=\Lambda_{ij,kl}(\hat n)\,h_{kl}=h_+\,e^{+}_{ij}+h_\times\,e^{\times}_{ij} \qquad (6)

where: $h_{kl}$ — a generic spatial perturbation (e.g. from the quadrupole formula); $e^{+}_{ij}=\hat x_i\hat x_j-\hat y_i\hat y_j$ and $e^{\times}_{ij}=\hat x_i\hat y_j+\hat y_i\hat x_j$ for $\hat n=\hat z$ — the plus and cross basis tensors. The two coefficients $h_+,h_\times$ are the wave's physical amplitudes.

Derivation AReaching the transverse-traceless gauge by spending the residual freedom

The goal is to prove the central claim of degree-of-freedom counting: that starting from a Lorenz-gauge perturbation with six independent components in vacuum, the four residual harmonic functions $\xi^\mu$ obeying $\Box\xi^\mu=0$ are exactly enough to zero the trace and the three mixed time-space components, after which the Lorenz condition itself forces the remaining spatial part to be transverse — leaving precisely two physical numbers. We track the budget subtraction at every step so the final tally $10-4-4=2$ is not asserted but earned.

Step 1. Begin in Lorenz gauge, $\partial^\mu\bar h_{\mu\nu}=0$ . This is four linear conditions on the ten components of the symmetric $h_{\mu\nu}$ , leaving six independent functions, each obeying the vacuum wave equation $\Box\bar h_{\mu\nu}=0$ . Why: the Lorenz condition is what decoupled Einstein's equations into wave equations in the first place, so it is the natural launch point and its cost is a clean four.

Step 2. Note the leftover freedom. A further shift $\xi^\mu$ changes $\bar h_{\mu\nu}$ by $-\partial_\mu\xi_\nu-\partial_\nu\xi_\mu+\eta_{\mu\nu}\partial_\alpha\xi^\alpha$ , whose divergence is $-\Box\xi_\nu$ . The Lorenz condition therefore survives if and only if $\Box\xi_\nu=0$ — four independent harmonic functions remain to be spent. Why: this is the crucial fact that Lorenz gauge is not a complete fixing; without these four functions there would be no way to descend further.

Step 3. Spend one function on the trace. Choose $\xi^\mu$ so that $h=\eta^{\mu\nu}h_{\mu\nu}=0$ . Since the trace obeys its own wave equation and $\xi^\mu$ can shift it by an arbitrary harmonic function, one component of the freedom sets $h\equiv0$ for all time. With $h=0$ the trace-reversed field equals the field, $\bar h_{\mu\nu}=h_{\mu\nu}$ . Why: killing the trace is what makes the whole scheme self-consistent — it is the difference between "TT" and merely "Lorenz."

Step 4. Spend three functions on the time-space components. Use the three remaining harmonic functions $\xi^i$ to set $h_{0i}=0$ . The residual freedom is now completely exhausted: four functions have imposed one trace condition and three $h_{0i}$ conditions. Why: these are the components that a boosted observer could confuse with genuine radiation, so removing them is what makes the surviving field frame-honest.

Step 5. Let the field equations finish the job. With $h=0$ and $h_{0i}=0$ , the $\nu=0$ Lorenz condition reads $\partial^\mu h_{\mu 0}=\partial^0 h_{00}=0$ , so $h_{00}$ is static; a radiative, oscillating solution has no static piece, hence $h_{00}=0$ and the whole time sector vanishes. The $\nu=j$ Lorenz condition then reduces to $\partial^i h_{ij}=0$ — transversality, imposed for free. Why: this shows transversality is not an extra hand-chosen gauge condition but a consequence of the constraints already in force.

Step 6. Count the survivors. The spatial symmetric $h_{ij}$ has six components; tracelessness removes one and the three transversality conditions remove three, leaving $6-1-3=2$ . Equivalently, from the top: $10-4$ (Lorenz) $-4$ (residual) $=2$ . Why: both routes must and do agree on two — the plus and cross polarizations of a massless spin-2 field.

Go deeper · Derivation B · 2 minThe plus and cross structure of a plane wave

Derivation BThe plus and cross structure of a plane wave

The goal is to show that the three TT conditions, applied to a single monochromatic plane wave travelling along $\hat z$ , force the amplitude matrix into a two-parameter symmetric traceless block confined to the $xy$ -plane — the plus and cross polarizations — and to see that the wave is necessarily null and luminal. We take a plane-wave ansatz, impose the wave equation, then transversality and tracelessness in turn, and read off what is left.

Step 1. Write $h^{\mathrm{TT}}_{ij}=A_{ij}\,\mathrm{Re}\,e^{ik_\alpha x^\alpha}$ with wavevector $k^\alpha=(\omega,0,0,k_z)$ for propagation along $+z$ , and $A_{ij}$ a constant symmetric matrix. Why: a plane wave is the natural mode of the flat-space wave equation, and any wave packet is a superposition of these, so solving one solves all.

Step 2. Impose $\Box h^{\mathrm{TT}}_{ij}=0$ . Acting on the exponential gives $-k_\alpha k^\alpha=0$ , i.e. $\omega^2=k_z^2$ : the wave is null and its crests move at the speed of light. Set $k_z=\omega$ . Why: this is the statement that gravitational waves travel at $c$ , a prediction later confirmed to one part in $10^{15}$ by the near-simultaneous arrival of GW170817 and its gamma-ray burst.

Step 3. Impose transversality $\partial^j h^{\mathrm{TT}}_{ij}=0$ , which becomes $k^j A_{ij}=0$ . With $k^j=(0,0,\omega)$ this forces $A_{i3}=0$ for every $i$ : the entire third row and, by symmetry, third column vanish. Only the $2\times2$ block in the $xy$ -plane survives. Why: transversality is what confines the disturbance to the plane perpendicular to the line of sight — the "transverse" in the name.

Step 4. Impose tracelessness $\delta^{ij}A_{ij}=A_{11}+A_{22}=0$ (the $A_{33}$ entry is already zero), so $A_{22}=-A_{11}$ . Define $A_{11}\equiv h_+$ . Symmetry leaves $A_{12}=A_{21}\equiv h_\times$ as the only other free entry. Why: tracelessness is what ties the $xx$ stretch to an equal $yy$ squeeze, encoding the area preservation seen in the ring figure.

Step 5. Assemble the result. The amplitude matrix is $A_{ij}=\begin{pmatrix}h_+&h_\times&0\\h_\times&-h_+&0\\0&0&0\end{pmatrix}$ : exactly two independent amplitudes. The circularly polarized combinations $h_+\pm ih_\times$ are eigenstates of rotation about $\hat z$ with phase $e^{\mp2i\psi}$ — the factor of two marking a spin-2, helicity- $\pm2$ field. Why: this $2\times2$ traceless block is the endpoint of the whole construction and the object every detector, template bank, and quadrupole calculation ultimately speaks in.

Go deeper · Derivation C · 1 minWhy the Lambda operator really projects onto the transverse-traceless part

Derivation CWhy the Lambda operator really projects onto the transverse-traceless part

The goal is to verify that $\Lambda_{ij,kl}=P_{ik}P_{jl}-\tfrac12 P_{ij}P_{kl}$ , built from the transverse projector $P_{ij}=\delta_{ij}-n_in_j$ , deserves the name projector: that its output is transverse (annihilated by $n_i$ ), traceless (annihilated by $\delta_{ij}$ ), and that applying it twice does nothing new (idempotence). These three properties are what license the one-line recipe $h^{\mathrm{TT}}_{ij}=\Lambda_{ij,kl}h_{kl}$ for extracting the physical field from any source-derived $h_{kl}$ .

Step 1. Establish the building blocks. From $P_{ij}=\delta_{ij}-n_in_j$ with $|\hat n|=1$ one gets $n_iP_{ij}=n_j-n_j=0$ (transverse), $P_{ik}P_{kj}=P_{ij}$ (idempotent, since the cross terms cancel), and $P_{ii}=3-1=2$ (rank two, the dimension of the transverse plane). Why: every property of $\Lambda$ reduces to these three identities for $P$ , so they are worth stating cleanly up front.

Step 2. Show transversality. Contract $\Lambda_{ij,kl}$ with $n_i$ : the first term gives $n_iP_{ik}P_{jl}=0$ and the second gives $\tfrac12 n_iP_{ij}P_{kl}=0$ , both because $n_iP_{ik}=0$ . Hence $n_i\Lambda_{ij,kl}=0$ , and the output carries no index along $\hat n$ . Why: this is what guarantees $\partial^j h^{\mathrm{TT}}_{ij}=0$ in Fourier space, the transversality condition itself.

Step 3. Show tracelessness. Contract with $\delta_{ij}$ : the first term is $\delta_{ij}P_{ik}P_{jl}=P_{kl}$ (using idempotence), and the second is $\tfrac12\delta_{ij}P_{ij}P_{kl}=\tfrac12(2)P_{kl}=P_{kl}$ . They cancel: $\delta_{ij}\Lambda_{ij,kl}=P_{kl}-P_{kl}=0$ . Why: the factor of $\tfrac12$ in $\Lambda$ is tuned precisely so that $P_{ii}=2$ makes the trace cancel — the origin of the one-half in the definition.

Step 4. Show idempotence. A direct contraction $\Lambda_{ij,mn}\Lambda_{mn,kl}$ using $P^2=P$ , $P_{ii}=2$ , and the two results above returns $\Lambda_{ij,kl}$ unchanged. Since $\Lambda^2=\Lambda$ , it is a true projector; its trace as an operator equals two, confirming it selects exactly two components. Why: idempotence is what makes "apply the projector once and you are done" legitimate — a second application cannot remove anything more.

Limiting cases

Case	Result	Meaning
Vacuum requirement	$T_{\mu\nu}=0$	Full TT gauge (spatial, transverse, traceless) is attainable only where the stress-energy vanishes; inside matter only the Lorenz gauge survives, because the TT projector is nonlocal.
Plane wave along $\hat z$	$h^{\mathrm{TT}}_{ij}=\begin{pmatrix}h_+&h_\times&0\\h_\times&-h_+&0\\0&0&0\end{pmatrix}\cos[\omega(t-z)]$	Transversality kills the third row and column; tracelessness ties $xx$ to $-yy$ ; only $h_+,h_\times$ remain.
Static / Newtonian limit	$\omega\to0$	No time dependence means no propagating modes; the residual field is pure gauge — radiation requires a time-varying quadrupole.
Electromagnetic analogue	$A^0=0,\ \nabla\!\cdot\mathbf{A}=0$	TT gauge is the spin-2 counterpart of radiation gauge, isolating two transverse photon polarizations from the four-potential.
Along the source axis	$\Lambda$ removes longitudinal parts	Even a spherically symmetric pulsation ( $h_{ij}\propto\delta_{ij}$ ) projects to zero — there is no monopole gravitational radiation.

Common trapFixed TT coordinates mean the wave has no effect

Two traps live here. First, in TT gauge a freely-falling mass that starts at rest keeps constant spatial coordinates — but this does not mean nothing happens. The geodesic acceleration $\mathrm{d}^2x^i/\mathrm{d}\tau^2$ vanishes only because the TT frame is glued to free fall; the physical, measurable quantity is the proper distance between two such masses, and it oscillates as $\delta L/L=\tfrac12 h_+$ . The wave warps the ruler, not the lattice. Second, TT gauge is a vacuum concept. The projector $\Lambda$ is built from inverse Laplacians and is therefore nonlocal, so you cannot "put a source in TT gauge" point-by-point where $T_{\mu\nu}\neq0$ ; the TT field is well-defined only in the radiation zone far outside the source. Confusing "coordinates don't move" with "physics doesn't happen" is the single most common error in reading TT gauge.

Worked examples

1 · Projecting a messy perturbation onto its physical TT part (fully worked)

A calculation in some frame hands you a symmetric spatial perturbation (in units of $10^{-21}$ )

h_{ij}=\begin{pmatrix}2&1&3\\1&0&1\\3&1&4\end{pmatrix},

and you know the wave travels along $\hat n=\hat z$ . Extract the physical amplitudes $h_+$ and $h_\times$ .

Step 1. Write the projector ingredients. For $\hat n=(0,0,1)$ the transverse projector is $P_{ij}=\delta_{ij}-n_in_j=\mathrm{diag}(1,1,0)$ . The TT projection can be applied in the compact matrix form $h^{\mathrm{TT}}=P\,h\,P-\tfrac12 P\,\mathrm{tr}(Ph)$ , which is just $\Lambda_{ij,kl}h_{kl}$ written with matrix products.

Step 2. Sandwich with $P$ . Multiplying $P\,h\,P$ simply deletes every entry that carries a $3$ -index (the whole third row and column):

P\,h\,P=\begin{pmatrix}2&1&0\\1&0&0\\0&0&0\end{pmatrix}.

This enforces transversality — the longitudinal $h_{i3}$ pieces, including the large $h_{33}=4$ , are discarded as unphysical for a wave along $\hat z$ .

Step 3. Remove the residual trace. The transverse trace is $\mathrm{tr}(Ph)=h_{11}+h_{22}=2+0=2$ , so subtract $\tfrac12 P\,\mathrm{tr}(Ph)=\mathrm{diag}(1,1,0)$ :

h^{\mathrm{TT}}=\begin{pmatrix}2&1&0\\1&0&0\\0&0&0\end{pmatrix}-\begin{pmatrix}1&0&0\\0&1&0\\0&0&0\end{pmatrix}=\begin{pmatrix}1&1&0\\1&-1&0\\0&0&0\end{pmatrix}.

Step 4. Read off and check. The surviving matrix is transverse (third row and column zero) and traceless ( $1-1=0$ ), as it must be. Comparing with the standard block, $h_+=1$ and $h_\times=1$ (in units of $10^{-21}$ ). Of the six independent components handed in, four were gauge or longitudinal and only these two are physical — a concrete instance of $6\to2$ .

2 · The strain a detector actually measures

The first detection, GW150914, produced a peak strain of about $h\sim1\times10^{-21}$ at Earth. LIGO's arms are $L=4\ \mathrm{km}$ long. How far does one mirror move, and how does that compare to the size of a proton?

Step 1. Relate strain to displacement. In TT gauge the mirrors sit at fixed coordinates while the proper length of an arm changes by $\delta L=\tfrac12 h\,L$ (the factor $\tfrac12$ is the same one in $\delta L/L=\tfrac12 h_+$ from the ring picture).

Step 2. Put in numbers. $\delta L=\tfrac12\,(1\times10^{-21})(4\times10^3\ \mathrm{m})=\mathbf{2\times10^{-18}\ m}$ . The two arms move oppositely — one lengthens while the other shortens — so the differential signal is twice this, which is what the interferometer reads as a dark-fringe shift.

Step 3. Grasp the scale. A proton is about $1.7\times10^{-15}\ \mathrm{m}$ across, so $\delta L$ is roughly one part in 850 of a proton diameter. Detecting it at all is possible only because the light bounces $\sim300$ times in each Fabry–Pérot cavity (effectively lengthening the arm) and because photon shot noise is beaten down by high laser power and squeezed light. The physical observable is $\delta L=2\times10^{-18}$ m, set entirely by the gauge-invariant TT amplitude $h$ .

3 · No monopole radiation, in one line

Show that a purely "breathing" spatial perturbation $h_{ij}=\phi(t,r)\,\delta_{ij}$ carries no transverse-traceless content.

Step 1. Apply the projector. Since $\Lambda$ is traceless, $\Lambda_{ij,kl}\delta_{kl}=\delta_{ij}-\delta_{ij}$ -type cancellation: explicitly $\Lambda_{ij,kl}\delta_{kl}=P_{ik}P_{jk}-\tfrac12 P_{ij}P_{kk}=P_{ij}-\tfrac12 P_{ij}(2)=P_{ij}-P_{ij}=0$ .

Step 2. Conclude. $h^{\mathrm{TT}}_{ij}=\phi\,\Lambda_{ij,kl}\delta_{kl}=0$ . A spherically symmetric pulsation radiates nothing — the gravitational analogue of the statement that a radially breathing charge distribution emits no electromagnetic waves, and the deep reason gravitational radiation starts at quadrupole order.

Physics of the portrait

The transverse-traceless gauge is really the resolution of a forty-year argument about whether gravitational waves are real at all — an argument won by learning to separate genuine curvature from coordinate bookkeeping. No one framed that distinction more sharply than the man who first caught Einstein's "waves" carrying no energy.

Origins · 1 minArthur Eddington and how this was discovered

Arthur Eddington

1882–1944 · separating physical gravitational waves from coordinate artifacts

Arthur Stanley Eddington was the astrophysicist who made general relativity respectable in the English-speaking world — he led the 1919 eclipse expedition that measured the deflection of starlight and wrote the first major English treatise on Einstein's theory. But his sharpest contribution to the subject of this article came in 1922. Einstein's 1916–1918 papers had found wave solutions of the linearized field equations and had counted, loosely, three kinds of them. Eddington examined these modes in different coordinate systems and discovered that two of the three were spurious: they could be made to travel at any speed one liked, including infinite speed, by a mere change of coordinates. In a famous phrase he wrote that these waves "propagate at the speed of thought" — they were artifacts of the coordinate labels, not ripples in spacetime. Only one class of solution had a coordinate-independent, energy-carrying reality, and it moved at the speed of light. This was the first clear demonstration that the ten components of $h_{\mu\nu}$ are mostly gauge, and that the physical content must be extracted by fixing that freedom — exactly what the transverse-traceless construction does systematically. The debate Eddington opened ran until Pirani, Bondi, and Feynman settled it with the geodesic-deviation and "sticky bead" arguments of the 1950s, vindicating the two physical, luminal, transverse-traceless degrees of freedom that LIGO would finally measure in 2015.

History

1916

— Einstein finds wave-like solutions of the linearized field equations

1918

— Einstein derives the quadrupole formula fixing the amplitude radiated by a source

1922

— Eddington shows two of the three wave modes are coordinate artifacts that "travel at the speed of thought"

1957

— the Chapel Hill "sticky bead" argument establishes that transverse-traceless waves carry real energy

2015

— LIGO detects GW150914, measuring the transverse-traceless strain of a binary black-hole merger

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §35.4 and Box 35.1 (TT gauge and the degree-of-freedom count) · Wald, General Relativity, §4.4 · Carroll, Spacetime and Geometry, §7.1–7.2 · Schutz, A First Course in General Relativity, §9.1 · Hartle, Gravity, Ch. 16 · Maggiore, Gravitational Waves, Vol. 1, §1.2 (the $\Lambda$ projector). Key self-consistency checks: $\Box\xi^\mu=0$ preserves the Lorenz condition because the divergence of the shifted trace-reversed field picks up a $-\Box\xi_\nu$ term; the degree-of-freedom tally $10-4-4=2$ is confirmed by both the top-down route and the spatial count $6-1-3=2$ ; the plane-wave amplitude matrix follows from $k^jA_{ij}=0$ and $A_{ii}=0$ ; the transverse, traceless, and idempotent properties of $\Lambda_{ij,kl}$ all reduce to $P_{ii}=2$ ; the worked projection gives $h_+=1,\,h_\times=1$ for the sample $h_{ij}$ ; and the LIGO arm displacement is $\delta L=\tfrac12 hL=2\times10^{-18}$ m. The quadrupole field $h^{\mathrm{TT}}_{ij}=(2G/r)\Lambda\ddot Q$ and the Isaacson flux formula are standard results. Geometric units $c=1$ are used throughout except in the numerical LIGO example.

Prerequisites. Lorenz (Harmonic) Gauge · Gauge Freedom in Linearized Gravity · Metric Perturbation and Trace-Reversed Field · Linearized Einstein Equations. Leads to. Plus and Cross Polarizations · Geodesic Deviation and Gravitational Waves · Helicity and the Spin-2 Graviton · The Quadrupole Formula.

Ask the tutor: Ask me to project any spatial perturbation onto its plus and cross parts, walk the $10-4-4=2$ counting line by line, or connect the TT amplitude to what LIGO actually reads on its dark fringe.

Why is it called "transverse-traceless" and not just "radiation gauge"?

The two names describe the same idea in different theories. "Radiation gauge" (or Coulomb gauge) is the electromagnetic version: $A^0=0$ and $\nabla\!\cdot\mathbf{A}=0$ , leaving the two transverse photon polarizations. "Transverse-traceless" is the gravitational version, and it needs the extra word traceless because gravity's field is a rank-two symmetric tensor, not a vector. The vector case has nothing to take a trace of; the tensor case does, and killing that trace is what removes the would-be scalar mode and enforces the area-preserving, quadrupolar character of the wave. So the name literally lists the two conditions on the spatial part — divergence-free (transverse) and trace-free — that a vector gauge does not need spelled out.

If the coordinates don't move, how does an interferometer ever see anything?

This is the most common conceptual hurdle, and the resolution is that detectors measure proper distances, not coordinates. In TT gauge the suspended mirrors keep fixed coordinate values, but the metric component $h^{\mathrm{TT}}_{ij}$ that sets the proper distance between them oscillates, so the light-travel time along an arm changes by $\delta L/L=\tfrac12 h_+$ . An interferometer compares the phase of light returning from two perpendicular arms; when one arm's proper length grows and the other's shrinks, the recombined beams shift off their dark fringe. You can equally work in a "detector frame" where the mirrors visibly accelerate under a tidal force $\ddot\xi^i=\tfrac12\ddot h^{\mathrm{TT}}_{ij}\xi^j$ — the same physics, gauge-shifted. Both descriptions predict the identical, gauge-invariant fringe shift.

Why can't I put the source itself into transverse-traceless gauge?

Because the operation that makes a field TT is nonlocal. The projector $\Lambda_{ij,kl}(\hat n)$ is defined for a single propagation direction $\hat n$ , and to apply it to a general field you must first Fourier-decompose, project each mode, and transform back — an operation that in position space involves the inverse Laplacian $\nabla^{-2}$ , which reaches across all space. Far outside a source, in the radiation zone, the field is (locally) a superposition of outgoing plane waves with a common $\hat n\approx\hat r$ , so the projection is clean and $h^{\mathrm{TT}}_{ij}$ is well-defined. Inside or near the source, where $T_{\mu\nu}\neq0$ and many directions contribute, there is no single $\hat n$ and no local TT field. The most you can impose there is the Lorenz gauge, which is local.

Does tracelessness mean the volume of a mass ring is conserved?

To first order in $h$ , yes — and this is exactly the area-preserving stretch you see in the ring figure. Tracelessness, $\delta^{ij}h^{\mathrm{TT}}_{ij}=0$ , means the sum of the diagonal strains is zero, so a stretch along one transverse axis is compensated by an equal squeeze along the perpendicular one, leaving the enclosed transverse area unchanged to linear order. This is why a gravitational wave produces a shape distortion (an ellipse) rather than a size distortion (a breathing circle) — the breathing, monopole mode is precisely what tracelessness forbids. At second order in $h$ there are small volume effects, but the leading, measurable signature is pure area-preserving shear.

Where exactly did the "missing" eight components go?

They split into two groups of four. Four are removed by the Lorenz condition $\partial^\mu\bar h_{\mu\nu}=0$ ; these are genuine constraints that also happen to decouple the field equations. The other four are pure gauge — the residual coordinate freedom $\xi^\mu$ with $\Box\xi^\mu=0$ — and they are spent setting the trace to zero (one function) and the three mixed time-space components $h_{0i}$ to zero (three functions). The field equations then force $h_{00}=0$ and the transversality of the spatial part as consequences, not new choices. None of the eight carried physical, coordinate-independent information: they were constraints and labels. Only the two amplitudes $h_+,h_\times$ change the curvature that a detector, or a ring of masses, actually feels.

How does this connect to the graviton being spin-2?

Directly. The two surviving amplitudes combine into circular polarizations $h_+\pm ih_\times$ , and under a rotation by angle $\psi$ about the propagation axis these pick up phases $e^{\mp2i\psi}$ . The "2" in that exponent is the helicity, and a massless field with helicity $\pm2$ is what "spin-2" means for a quantum of radiation. Contrast the photon, whose transverse polarizations transform as $e^{\mp i\psi}$ (helicity $\pm1$ , spin-1). So the transverse-traceless counting — two physical states, transforming with a factor of two under rotation — is the classical shadow of the graviton's spin. The full quantum story (a massless spin-2 particle, and why that forces general relativity at low energies) is developed in the helicity and spin-2 graviton article; TT gauge is where its two polarization states first appear.