Gravitational-Wave Luminosity and Energy Flux

Gravitational-wave luminosity is the rate at which a source bleeds energy into the ripples it launches, $L_{\rm GW}=\mathrm dE/\mathrm dt$ , and computing it forces a conceptual crisis first. The equivalence principle forbids a local, gauge-invariant energy density for the gravitational field: at any point one can transform to a freely-falling frame where the metric is flat and its first derivatives vanish, so "the energy here" is coordinate-dependent nonsense. The resolution, owned by aq-gr-isaacson-stress-energy, is to stop asking pointwise and average the second-order Einstein tensor over a region large compared with a wavelength but small compared with the background curvature radius. That short-wave average kills the gauge-dependent total-derivative terms and leaves an effective stress-energy $t_{\mu\nu}$ that curves the background and conserves energy like any other source. Everything else is bookkeeping on that idea. Far from the source the field is locally plane and can be put in transverse-traceless gauge (aq-gr-transverse-traceless-gauge), leaving only two polarizations; the quadrupole formula (aq-gr-quadrupole-formula) supplies the amplitude $h\sim(2G/c^4 r)\ddot Q$ ; and the flux $\langle t^{0r}\rangle\sim\dot h^2$ integrated over a distant sphere gives Einstein's result $L_{\rm GW}=(G/5c^5)\langle\dddot Q_{ij}\dddot Q_{ij}\rangle$ . The prefactor $G/c^5$ is so minute that even violent motions radiate a whisper of the natural scale $c^5/G\approx3.6\times10^{52}\,\mathrm W$ — yet that whisper drains the Hulse–Taylor pulsar's orbit at exactly the predicted rate, the first proof that gravity radiates.

Key takeaways

Gravitational-wave luminosity is a source's energy-loss rate, $L_{\rm GW}=\mathrm dE/\mathrm dt$ , got by integrating the Isaacson energy flux over a large sphere, the $r^2$ of the area cancelling the $1/r^2$ of the flux.
Because gravitational energy is not localizable, the flux $\langle t^{0i}\rangle=-\tfrac{c^4}{32\pi G}\langle\dot h^{\rm TT}_{jk}\,\partial_i h^{\rm TT}_{jk}\rangle$ is meaningful only after averaging over several wavelengths — a gauge-invariant average, not a pointwise density.
Feeding in the quadrupole formula gives the workhorse result $L_{\rm GW}=\tfrac{G}{5c^5}\langle\dddot Q_{ij}\dddot Q_{ij}\rangle$ , with $Q_{ij}$ the reduced (trace-free) mass quadrupole; the third derivative appears because flux is quadratic in $\dot h\propto\ddot Q$ .
Radiation starts at the quadrupole, not the monopole or dipole: mass-energy and momentum conservation kill the two lowest moments, which is why gravity radiates so weakly.
The coupling $G/c^5\approx2.76\times10^{-53}\,\mathrm{W^{-1}}$ inverts to the Dyson scale $c^5/G\approx3.6\times10^{52}\,\mathrm W$ ; the Earth–Sun system musters only about 200 W, while a black-hole merger reaches $\sim10^{-3}$ of the Dyson bound.
For a circular binary $L_{\rm GW}=\tfrac{32}{5}\tfrac{G}{c^5}\mu^2a^4\Omega^6$ , which drains the orbit and produces the inspiral chirp measured in PSR B1913+16 and heard by LIGO.
Key idea: The luminosity is flux times area — take the Isaacson stress-energy of the TT wave field, integrate its energy flux over a distant sphere, and substitute the quadrupole amplitude to obtain $L_{\rm GW}=(G/5c^5)\langle\dddot Q_{ij}\dddot Q_{ij}\rangle$ .

The core insight

The luminosity is flux times area: take the Isaacson stress-energy of the TT wave field, integrate its energy flux over a distant sphere, and substitute the quadrupole expression for the wave amplitude to obtain $L_{\rm GW}=(G/5c^5)\langle\dddot Q_{ij}\dddot Q_{ij}\rangle$ .

Where this lives today

The quadrupole luminosity is the daily working formula of gravitational-wave astronomy. Every compact-binary waveform template used by LIGO, Virgo, and KAGRA rests on the energy-balance equation $\dot E_{\rm orb}=-L_{\rm GW}$ : at leading order it fixes $\dot f\propto f^{11/3}\mathcal M^{5/3}$ , so the chirp mass is read straight from how fast the signal sweeps up, with masses and spins from higher post-Newtonian corrections. Pulsar timing arrays (NANOGrav, EPTA, PPTA) hunt the stochastic hum of supermassive-black-hole binaries whose collective luminosity sets the nanohertz background reported in 2023. Near merger the slow-motion expansion breaks down and numerical relativity takes over — yet the quadrupole result remains the benchmark against which every numerical waveform is normalized, and the future detector LISA will track $L_{\rm GW}$ through millions of orbits, turning the Einstein rate into a precision test of general relativity itself.

The physical picture

Picture the source at the center of an enormous imaginary sphere and the wave field as a stream of ripples flowing outward across it. Each ripple carries energy proportional to the square of its slope — as a stretched string carries power $\propto(\partial_t y)(\partial_x y)$ — and the luminosity is the total energy crossing the sphere per unit time. Two things make this rigorous. The flux must be averaged over several wavelengths, the gravitational analogue of replacing $\sin^2$ by $\tfrac12$ for a light wave, because before averaging the local density is gauge and not even sign-definite. And it must fall as $1/r^2$ so that, times the area $4\pi r^2$ , it gives a luminosity independent of $r$ — energy conservation in the radiation zone made visible.

Figure

Energy crossing a distant sphere: the quadrupole source radiates a flux that falls as ⟨t⁰ʳ⟩ ∝ 1/r², so the total power L = ∮ ⟨t⁰ʳ⟩ r² dΩ is independent of the sphere radius r. The natural scale is the Dyson luminosity c⁵/G ≈ 3.6×10⁵² W, which no ordinary source approaches

Why the third time-derivative of $Q$ ? The amplitude measures the acceleration of the mass distribution, $h\sim\ddot Q/r$ ; the flux is quadratic in the field's velocity, $\dot h^2\sim(\dddot Q)^2/r^2$ ; and the power is $r^2$ times that flux, so the $r$ 's cancel and $L$ depends on $(\dddot Q)^2$ . The contrast with electromagnetism is structural: conservation of mass-energy freezes the monopole and conservation of momentum freezes the mass dipole ( $\dot{\mathbf P}=0$ ), so radiation begins two multipole orders higher than the electric dipole — the deep reason gravitational radiation is so feeble. The clinching demonstration is a real system: the orbital clock of a binary pulsar, ticking down at precisely the rate this formula predicts.

Figure

The Hulse–Taylor binary PSR B1913+16: the cumulative shift of periastron time accumulates as a parabola because the orbital period itself is shrinking at dP_b/dt = −2.40×10⁻¹² s/s. The dots are timing data over three decades; the curve is the parameter-free general-relativity prediction from the quadrupole luminosity — they agree to about 0.2%

The core equations

Start from the definition of a luminosity: the total power crossing a sphere is the radial energy flux integrated over solid angle, with the area element $r^2\,\mathrm d\Omega$ turning flux into power. The angle brackets are the Isaacson short-wave average, without which the integrand has no gauge-invariant meaning.

L_{\rm GW}=\frac{\mathrm dE}{\mathrm dt}=\oint_{S}\langle t^{0i}\rangle\,n_i\,r^2\,\mathrm d\Omega \qquad (1)

where: $L_{\rm GW}$ — total power radiated in gravitational waves; $t^{0i}$ — Isaacson energy flux, the effective stress-energy of the waves; $n_i$ — outward unit normal to the sphere $S$ of radius $r$ ; $\langle\cdot\rangle$ — average over several wavelengths or periods; $\mathrm d\Omega$ — solid-angle element. Read it as "flux times area, summed over all directions", the average making it gauge-invariant. The flux comes from the Isaacson stress-energy tensor, which repairs non-localizability by averaging the second-order Einstein tensor and moving it to the source side of the field equations.

t_{\mu\nu}=\frac{c^4}{32\pi G}\,\langle\partial_\mu h^{\rm TT}_{ij}\,\partial_\nu h^{\rm TT}_{ij}\rangle \qquad (2)

where: $h^{\rm TT}_{ij}$ — transverse-traceless metric perturbation; $\partial_\mu$ — partial derivative, repeated indices $i,j$ summed; $c^4/32\pi G$ — normalization fixing the energy scale. It is the gravitational twin of the electromagnetic $T_{\mu\nu}\sim E^2+B^2$ : energy quadratic in the field's derivatives. The time–space component is the energy flux — the analogue of the Poynting vector — depending for an outgoing wave only on $\dot h$ .

\langle t^{0i}\rangle=-\frac{c^4}{32\pi G}\,\langle\dot h^{\rm TT}_{jk}\,\partial_i h^{\rm TT}_{jk}\rangle \qquad (3)

where: $\dot h^{\rm TT}_{jk}$ — coordinate-time derivative of the TT amplitude, $\partial_t h^{\rm TT}_{jk}$ ; $\partial_i$ — spatial gradient, radial far from the source; $\langle t^{0i}\rangle$ — the physical energy flux (power per unit area), i.e. $c$ times the raw tensor component, the minus sign coming from raising the time index. For an outgoing wave $\partial_r h\to-\tfrac1c\dot h$ , so $\langle t^{0r}\rangle=\tfrac{c^3}{32\pi G}\langle\dot h^2\rangle>0$ — the square of the field's velocity. Inserting the quadrupole formula for $h^{\rm TT}$ and integrating over the sphere collapses everything to Einstein's 1918 luminosity, the central formula of the subject.

L_{\rm GW}=\frac{G}{5c^5}\,\langle\dddot Q_{ij}\,\dddot Q_{ij}\rangle \qquad (4)

where: $Q_{ij}$ — reduced (trace-free) mass quadrupole moment; $\dddot Q_{ij}$ — its third time derivative; $G/5c^5$ — coupling combining the flux normalization with the angular average, the $1/5$ being the fingerprint of the transverse projection averaged over all directions. The moment that appears is the trace-free part of the second moment — only the shape of the source, not its size, radiates.

Q_{ij}=\int\rho(\mathbf x)\left(x_i x_j-\tfrac13\delta_{ij}\,r^2\right)\mathrm d^3x \qquad (5)

where: $\rho$ — mass-energy density; $x_i$ — Cartesian position relative to the center of mass; $r^2=x_k x_k$ . Subtracting the trace $\tfrac13\delta_{ij}r^2$ is not cosmetic: the trace part is a monopole in disguise and does not radiate, and the coefficients $1/5$ and $32/5$ assume it removed. Evaluating $Q_{ij}$ for two point masses on a circular orbit and differentiating three times yields the Peters–Mathews binary luminosity that governs every compact-binary inspiral.

L_{\rm GW}=\frac{32}{5}\frac{G}{c^5}\,\mu^2 a^4\Omega^6=\frac{32}{5}\frac{G^4}{c^5}\frac{m_1^2 m_2^2(m_1+m_2)}{a^5} \qquad (6)

where: $\mu=m_1 m_2/(m_1+m_2)$ — reduced mass; $a$ — orbital separation; $\Omega$ — orbital angular frequency, Kepler's law $\Omega^2=GM/a^3$ giving the right-hand form. The steep $a^{-5}$ is why inspiral is glacial when the stars are far apart and catastrophic at merger.

Derivation AQuadrupole luminosity from the Isaacson flux

The goal is to turn the abstract flux integral of equation (1) into Einstein's concrete number $L_{\rm GW}=(G/5c^5)\langle\dddot Q_{ij}\dddot Q_{ij}\rangle$ : start from the asymptotic wave field a quadrupole produces, square its time derivative to build the Isaacson flux, and integrate over a distant sphere. The whole content sits in one place — the transverse-traceless projection of the quadrupole, averaged over all directions, produces the pure number $2/5$ that fixes the coupling.

Step 1. The quadrupole formula (from aq-gr-quadrupole-formula) gives the far field in TT gauge as $h^{\rm TT}_{ij}=\tfrac{2G}{c^4 r}\,\Lambda_{ij,kl}(\mathbf n)\,\ddot Q_{kl}(t-r/c)$ , with the TT projector $\Lambda_{ij,kl}=P_{ik}P_{jl}-\tfrac12 P_{ij}P_{kl}$ built from $P_{ij}=\delta_{ij}-n_i n_j$ . Why: only the part of the quadrupole transverse to the line of sight $\mathbf n$ radiates, and $\Lambda$ extracts it while removing the trace.

Step 2. Differentiate in time, using $\partial_r\to-\tfrac1c\partial_t$ on the retarded argument, to get $\dot h^{\rm TT}_{ij}=\tfrac{2G}{c^4 r}\,\Lambda_{ij,kl}\,\dddot Q_{kl}$ and a radial flux $\langle t^{0r}\rangle=\tfrac{c^3}{32\pi G}\langle\dot h^{\rm TT}_{ij}\dot h^{\rm TT}_{ij}\rangle=\tfrac{G}{8\pi c^5 r^2}\,\Lambda_{ij,kl}\langle\dddot Q_{ij}\dddot Q_{kl}\rangle$ . Why: squaring promotes $\ddot Q$ to $\dddot Q$ and, because $\Lambda$ is idempotent ( $\Lambda\Lambda=\Lambda$ ), only a single projector survives.

Step 3. Do the angular integral with the isotropic averages $\int n_i n_j\,\tfrac{\mathrm d\Omega}{4\pi}=\tfrac13\delta_{ij}$ and $\int n_i n_j n_k n_l\,\tfrac{\mathrm d\Omega}{4\pi}=\tfrac1{15}(\delta_{ij}\delta_{kl}+\delta_{ik}\delta_{jl}+\delta_{il}\delta_{jk})$ . Why: the projector $\Lambda(\mathbf n)$ is a polynomial in the components of $\mathbf n$ , so integrating it over the sphere reduces to these two tensor averages.

Step 4. Contract against $\dddot Q_{ij}\dddot Q_{kl}$ and use tracelessness $Q_{ii}=0$ to drop every term with a $\delta$ that contracts onto a trace; the residue is $\oint\Lambda_{ij,kl}\tfrac{\mathrm d\Omega}{4\pi}\,\dddot Q_{ij}\dddot Q_{kl}=\tfrac25\,\dddot Q_{ij}\dddot Q_{ij}$ . Why: tracelessness turns a messy sum of Kronecker deltas into the single clean coefficient $2/5$ .

Step 5. Assemble, $L_{\rm GW}=r^2\oint\langle t^{0r}\rangle\,\mathrm d\Omega=\tfrac{G}{8\pi c^5}\cdot4\pi\cdot\tfrac25\langle\dddot Q_{ij}\dddot Q_{ij}\rangle=\tfrac{G}{5c^5}\langle\dddot Q_{ij}\dddot Q_{ij}\rangle$ . Why: the $r^2$ from the area cancels the $1/r^2$ in the flux, so $L$ is independent of where the sphere sits — no energy is lost between the source and infinity.

Go deeper · Derivation B · 1 minThe luminosity of a circular binary

Derivation BThe luminosity of a circular binary

The goal is to evaluate the general formula (4) for the most important source in the sky — two masses on a circular orbit — and land on the Peters–Mathews power $L_{\rm GW}=\tfrac{32}{5}\tfrac{G}{c^5}\mu^2 a^4\Omega^6$ . The route is direct: write the quadrupole of the relative orbit, notice it oscillates at twice the orbital frequency, differentiate three times, and average the square. Watch the factor of two in frequency and the pure number $32$ .

Step 1. Only the relative motion radiates, so use $Q_{ij}=\mu(x_i x_j-\tfrac13\delta_{ij}r^2)$ with $\mathbf x=a(\cos\Omega t,\sin\Omega t,0)$ . Expanding, $Q_{xx}=\tfrac{\mu a^2}{2}\cos2\Omega t+\text{const}$ , $Q_{yy}=-\tfrac{\mu a^2}{2}\cos2\Omega t+\text{const}$ , $Q_{xy}=\tfrac{\mu a^2}{2}\sin2\Omega t$ . Why: squaring $\cos\Omega t$ and $\sin\Omega t$ produces $\cos2\Omega t$ and $\sin2\Omega t$ — the mass distribution returns to the same shape twice per orbit, so it radiates at $2\Omega$ , not $\Omega$ .

Step 2. Differentiate three times: $\dddot Q_{xx}=4\mu a^2\Omega^3\sin2\Omega t$ , $\dddot Q_{yy}=-4\mu a^2\Omega^3\sin2\Omega t$ , $\dddot Q_{xy}=-4\mu a^2\Omega^3\cos2\Omega t$ . Why: each derivative brings down $2\Omega$ , so three of them carry $(2\Omega)^3=8\Omega^3$ times the amplitude $\tfrac{\mu a^2}{2}$ , giving the $4\mu a^2\Omega^3$ prefactor.

Step 3. Form $\langle\dddot Q_{ij}\dddot Q_{ij}\rangle=\langle\dddot Q_{xx}^2+\dddot Q_{yy}^2+2\dddot Q_{xy}^2\rangle=(4\mu a^2\Omega^3)^2(\tfrac12+\tfrac12+2\cdot\tfrac12)=32\,\mu^2 a^4\Omega^6$ . Why: the off-diagonal $Q_{xy}=Q_{yx}$ is counted twice, and each $\langle\sin^2\rangle=\langle\cos^2\rangle=\tfrac12$ , so the bracket collapses to the integer 32.

Step 4. Substitute into (4): $L_{\rm GW}=\tfrac{G}{5c^5}\cdot32\mu^2 a^4\Omega^6=\tfrac{32}{5}\tfrac{G}{c^5}\mu^2 a^4\Omega^6$ , and eliminate $\Omega$ with $\Omega^2=GM/a^3$ to reach $L_{\rm GW}=\tfrac{32}{5}\tfrac{G^4}{c^5}m_1^2 m_2^2 M/a^5$ . Why: expressing the power through the separation $a$ rather than $\Omega$ makes the $a^{-5}$ scaling — and the runaway inspiral — explicit.

Go deeper · Derivation C · 1 minOrbital decay and the inspiral chirp

Derivation COrbital decay and the inspiral chirp

The goal is to convert the binary luminosity into an observable: the rate at which the orbit shrinks and the period decreases, the quantity measured in the Hulse–Taylor pulsar. The physics is a single balance statement — the radiated power must come out of the orbital energy — combined with the Newtonian relations between energy, separation, and period, yielding both the inspiral rate $\dot a$ and the period derivative $\dot P_b$ that carries the chirp mass.

Step 1. The Newtonian binding energy of a circular orbit is $E_{\rm orb}=-\tfrac{Gm_1 m_2}{2a}$ , and energy conservation demands $\dot E_{\rm orb}=-L_{\rm GW}$ . Why: the orbit is the only reservoir, so radiating forces $a$ to shrink, and since $E<0$ the system becomes more tightly bound as it loses energy.

Step 2. Differentiate, $\dot E_{\rm orb}=\tfrac{Gm_1 m_2}{2a^2}\dot a$ , set it equal to $-L_{\rm GW}$ , and solve: $\dot a=-\tfrac{64}{5}\tfrac{G^3 m_1 m_2 M}{c^5 a^3}$ . Why: $\dot a<0$ confirms inspiral, and the $a^{-3}$ makes the shrinkage accelerate without bound, giving a finite coalescence time $\tau=\tfrac{5}{256}\,c^5 a_0^4/(G^3 m_1 m_2 M)$ .

Step 3. Recast via Kepler's law as $\dot P_b=-\tfrac{192\pi}{5}\big(\tfrac{2\pi G\mathcal M}{c^3 P_b}\big)^{5/3}$ , with chirp mass $\mathcal M=(m_1 m_2)^{3/5}/M^{1/5}$ , times $f(e)=(1+\tfrac{73}{24}e^2+\tfrac{37}{96}e^4)/(1-e^2)^{7/2}$ for eccentricity. Why: the masses enter $\dot P_b$ only through $\mathcal M$ , which is why a detector reads the chirp mass straight off the waveform without resolving $m_1$ and $m_2$ .

Limiting cases

The quadrupole luminosity is the leading term of a slow-motion, weak-field expansion; its regime of validity and its most-used specializations are collected here.

Case	Result	Meaning
Slow-motion, weak-field	$v/c\ll1,\ \ L_{\rm GW}=\tfrac{G}{5c^5}\langle\dddot Q_{ij}\dddot Q_{ij}\rangle$	The leading term; corrections enter at $O(v^2/c^2)$ through the current quadrupole and mass octupole (the post-Newtonian series).
Circular binary	$L_{\rm GW}=\tfrac{32}{5}\tfrac{G}{c^5}\mu^2 a^4\Omega^6=\tfrac{32}{5}\tfrac{G^4}{c^5}\tfrac{m_1^2 m_2^2 M}{a^5}$	Uses $\Omega^2=GM/a^3$ ; the $a^{-5}$ scaling drives the inspiral, glacial when wide and explosive near merger.
Eccentric orbit	$L_{\rm GW}=\tfrac{32}{5}\tfrac{G^4 m_1^2 m_2^2 M}{c^5 a^5}\,f(e),\ \ f(e)=\tfrac{1+\frac{73}{24}e^2+\frac{37}{96}e^4}{(1-e^2)^{7/2}}$	Peters–Mathews factor; eccentricity sharply boosts emission at periastron ( $f\approx12$ for the Hulse–Taylor $e=0.617$ ).
Natural luminosity scale	$L_{\rm GW}\lesssim\tfrac{c^5}{G}\approx3.6\times10^{52}\,\mathrm W$	The Dyson luminosity, an absolute ceiling approached only by near-equal-mass black-hole mergers at coalescence.

Common trapThe energy flux is a real local density you can measure point by point

It is not. Two related traps sink students. First, the integrand $t^{0i}$ is not a genuine local energy flux — it is gauge-dependent point by point, and the equivalence principle lets you transform it away at any single event. It acquires meaning only after the Isaacson average over a region several wavelengths across but small compared with the background curvature radius, which cancels the gauge-dependent total-derivative terms. Quoting "the energy density at this point" is a category error; only the averaged flux, and the luminosity built from it, are physical. Second, the moment in the formula is the reduced quadrupole $Q_{ij}=I_{ij}-\tfrac13\delta_{ij}I_{kk}$ , not the full $I_{ij}$ : using $I_{ij}$ smuggles in the non-radiating trace and gives the wrong coefficient — the celebrated $1/5$ and $32/5$ presuppose the trace-free moment.

Worked examples

1 · How feeble is gravitational radiation? (confidence builder)

Estimate the gravitational-wave luminosity of the Earth orbiting the Sun, and compare it with the natural scale $c^5/G$ .

Step 1. Since $m_\oplus\ll M_\odot$ , the reduced mass is essentially the Earth's, $\mu\approx m_\oplus=5.97\times10^{24}\,\mathrm{kg}$ , with $a=1.496\times10^{11}\,\mathrm m$ and $\Omega=2\pi/(1\,\text{yr})=1.99\times10^{-7}\,\mathrm{rad\,s^{-1}}$ . The orbit is nearly circular, so equation (6) applies directly.

Step 2. Assemble $L_{\rm GW}=\tfrac{32}{5}\tfrac{G}{c^5}\mu^2 a^4\Omega^6$ with $G/c^5=2.76\times10^{-53}\,\mathrm{W^{-1}}$ , $\mu^2=3.57\times10^{49}$ , $a^4=5.01\times10^{44}$ , $\Omega^6=6.2\times10^{-41}$ .

Step 3. The product is $L_{\rm GW}\approx 2.0\times10^{2}\,\mathrm W$ — the entire Earth–Sun system radiates gravitationally at the power of two incandescent bulbs. Against the Dyson scale this is $L_{\rm GW}/(c^5/G)\approx5\times10^{-51}$ , a fifty-order-of-magnitude shortfall: $G/c^5$ is so tiny the orbit will not decay appreciably in the age of the universe. Key result: about 200 W, or $5\times10^{-51}$ of the natural luminosity scale.

2 · The Hulse–Taylor orbital decay

PSR B1913+16 is a pair of neutron stars with $m_1=1.44\,M_\odot$ , $m_2=1.39\,M_\odot$ , orbital period $P_b=7.75\,\mathrm h=2.79\times10^4\,\mathrm s$ , and eccentricity $e=0.617$ . Predict the rate of change of the orbital period and compare with observation.

Step 1. Form the chirp mass $\mathcal M=(m_1 m_2)^{3/5}/M^{1/5}$ with $M=2.83\,M_\odot$ : $\mathcal M=1.23\,M_\odot=2.45\times10^{30}\,\mathrm{kg}$ , the only mass combination $\dot P_b$ needs.

Step 2. The dimensionless orbital factor is $\dfrac{2\pi G\mathcal M}{c^3 P_b}=1.37\times10^{-9}$ , and $(1.37\times10^{-9})^{5/3}=1.68\times10^{-15}$ . Times $-192\pi/5=-120.6$ this gives the circular part, $-2.03\times10^{-13}\,\mathrm{s/s}$ .

Step 3. The orbit is strongly eccentric, so apply $f(e)$ with $e^2=0.381$ : numerator $1+\tfrac{73}{24}(0.381)+\tfrac{37}{96}(0.145)=2.21$ , denominator $(1-0.381)^{7/2}=0.187$ , so $f=11.9$ . This nearly-twelvefold boost is emission concentrated at periastron, where the stars swing closest and fastest.

Step 4. Combine: $\dot P_b=-2.03\times10^{-13}\times11.9=\mathbf{-2.40\times10^{-12}\,s/s}$ . Three decades of pulsar timing measure exactly this, agreeing to about $0.2\%$ — the first quantitative proof, and the 1993 Nobel Prize, that gravitational waves carry energy at the Einstein rate. Key result: $\dot P_b\approx-2.4\times10^{-12}$ s/s, matching observation to 0.2%.

Physics of the portrait

The quadrupole formula made a prediction so small it took a century, a binary pulsar, and a four-kilometre interferometer to confirm — exactly.

Origins · 1 minquadrupole formula and how this was discovered

quadrupole formula

the Einstein luminosity L = (G/5c⁵)⟨Q⃛ᵢⱼ Q⃛ᵢⱼ⟩ · weak-field gravitational radiation

The quadrupole formula has the strangest biography of any equation in general relativity — its own author doubted it twice. Albert Einstein derived a version in 1916, spoiled it with a calculational error, and repaired it in a 1918 paper — which still carried a factor-of-two error in the radiated power, corrected by Eddington in 1922, after which the luminosity $L=(G/5c^5)\langle\dddot Q_{ij}\dddot Q_{ij}\rangle$ stood in the form that still bears his name. For decades no one was sure the waves were real rather than coordinate artifacts; Einstein himself, with Nathan Rosen in 1936, submitted a paper titled "Do Gravitational Waves Exist?" arguing they did not, before a referee caught the error. It was made rigorous piecemeal: Peters and Mathews (1963) gave the binary luminosity and its eccentricity factor, and Richard Isaacson (1968) settled the localizability problem with the short-wave average that makes $t_{\mu\nu}$ gauge-invariant. The decisive evidence was astronomical: Hulse and Taylor's 1974 discovery of PSR B1913+16 gave a clock precise enough to watch an orbit shrink, and by the 1980s its decay matched the prediction to a fraction of a percent. LIGO's 2015 detection of GW150914 closed the arc, measuring a merger whose peak luminosity — around $3.6\times10^{49}\,\mathrm W$ , momentarily outshining every star in the observable universe combined — is this same formula pushed to the edge where it hands off to numerical relativity.

History

1918

— Einstein derives the quadrupole luminosity L = (G/5c⁵)⟨Q⃛Q⃛⟩, still off by a factor of two until Eddington's 1922 correction

1963

— Peters and Mathews compute the binary luminosity and the eccentricity enhancement factor

1968

— Isaacson fixes the gauge-invariant wave stress-energy by short-wave averaging

1974

— Hulse and Taylor discover PSR B1913+16, the first binary-pulsar laboratory

2015

— LIGO's GW150914 directly measures a merger radiating about 3.6×10⁴⁹ W at peak

Going further

References. Misner, Thorne & Wheeler, Gravitation (1973), §35.7, §36.10 (quadrupole luminosity, Isaacson average) · Wald, General Relativity (1984), §4.4b · Carroll, Spacetime and Geometry (2004), §7.6 · Schutz, A First Course in General Relativity, 2nd ed. (2009), §9.3 · Hartle, Gravity (2003), §23.5 · Maggiore, Gravitational Waves Vol. 1 (2008), §3.3–3.4 · Peters & Mathews, Phys. Rev. 131, 435 (1963); Peters, Phys. Rev. 136, B1224 (1964). Key numerical values used throughout: $c^5/G=3.63\times10^{52}\,\mathrm W$ and $G/c^5=2.76\times10^{-53}\,\mathrm{W^{-1}}$ (from CODATA $G$ and $c$ ); Earth–Sun $L_{\rm GW}\approx196\,\mathrm W$ ; Hulse–Taylor $\dot P_b=-2.40\times10^{-12}\,\mathrm{s/s}$ with $m_1=1.44\,M_\odot$ , $m_2=1.39\,M_\odot$ , $P_b=7.75\,\mathrm h$ , $e=0.617$ , giving chirp mass $1.23\,M_\odot$ and eccentricity factor $11.9$ . The angular averages $\langle n_i n_j\rangle=\tfrac13\delta_{ij}$ and $\langle n_i n_j n_k n_l\rangle=\tfrac1{15}(\delta_{ij}\delta_{kl}+\delta_{ik}\delta_{jl}+\delta_{il}\delta_{jk})$ confirm the projector integral yields $2/5$ , and the binary average yields the integer $32$ .

Prerequisites. Quadrupole Formula · Isaacson Stress-Energy of Gravitational Waves · Transverse-Traceless Gauge · Linearized Einstein Equations. Leads to. Multipole Expansion of Gravitational Radiation · Plus and Cross Polarizations · Geodesic Deviation and Tidal Action of Waves.

Ask the tutor: Ask me to walk through why the reduced quadrupole's third derivative — and nothing lower — sets the rate at which any source can bleed energy into gravitational waves.

Why does gravitational radiation start at the quadrupole when light starts at the dipole?

Because the "charges" of gravity are masses, which obey stricter conservation laws than electric charge. Electromagnetic radiation begins at dipole order because the dipole $\sum q_i\mathbf x_i$ can oscillate freely. In gravity the monopole is the total mass-energy, conserved, so its derivatives vanish; the dipole is $\sum m_i\mathbf x_i$ , whose first derivative is the total momentum and whose second derivative is $\dot{\mathbf P}=0$ , so it too is silent. The lowest moment free to change is the quadrupole. Radiation therefore begins two multipole orders higher than in electromagnetism — the deep reason gravitational waves are so weak.

What does it mean that gravitational energy is "not localizable"?

It means there is no tensor giving an unambiguous energy density of the gravitational field at a single point. The equivalence principle guarantees that at any event you can choose a freely-falling frame in which the metric is flat and its first derivatives vanish — so whatever "field energy" you thought was there transforms to zero. Any candidate local density is therefore coordinate-dependent, and the second-order Einstein tensor that would source it is not gauge-invariant point by point. Isaacson's resolution is to average over a region several wavelengths across: the average washes out the gauge-dependent oscillating terms and leaves a genuine, conserved, gauge-invariant stress-energy $t_{\mu\nu}$ . Energy is real; it is just not localizable to a point.

Why do the waves come out at twice the orbital frequency?

Because the source's quadrupole moment, not its position, is what radiates, and the quadrupole returns to the same value twice per orbit. Position is a dipole quantity that oscillates once per period, but the quadrupole depends on products like $x^2$ and $xy$ ; substituting $x=a\cos\Omega t$ gives $\cos^2\Omega t=\tfrac12(1+\cos2\Omega t)$ , an oscillation at $2\Omega$ . Physically, a dumbbell rotated by $180°$ looks identical to its start, so its far field completes a full cycle in half an orbit. The gravitational-wave frequency of a binary is thus twice its orbital frequency — a fact used routinely to infer orbital parameters from a signal.

Why must I use the reduced (trace-free) quadrupole and not the full second moment?

Because only the trace-free part of the mass distribution radiates. The full second moment $I_{ij}=\int\rho\,x_i x_j\,\mathrm d^3x$ contains a trace $I_{kk}=\int\rho\,r^2\,\mathrm d^3x$ that is a monopole-like quantity — a spherically symmetric breathing of the source — and a spherical mass distribution radiates nothing (Birkhoff's theorem). Subtracting $\tfrac13\delta_{ij}I_{kk}$ to form $Q_{ij}$ throws away exactly that non-radiating piece. Keeping the trace and plugging $I_{ij}$ into the formula double-counts a silent mode and gives the wrong number; the famous $1/5$ and $32/5$ hold only for the reduced moment.

If the Earth–Sun system radiates only 200 watts, how does LIGO ever detect anything?

By choosing utterly different sources. The luminosity scales as $\mu^2 a^4\Omega^6$ , and through Kepler's law $\Omega^6\propto M^3/a^9$ , so the power grows explosively as the separation shrinks and the masses become compact. Two thirty-solar-mass black holes a few hundred kilometres apart, orbiting hundreds of times per second, have an $\Omega$ larger than the Earth's by $\sim10^{10}$ , pushing the luminosity up more than fifty orders of magnitude to near the Dyson scale. That is why a black-hole merger briefly outshines the visible universe while the Earth–Sun system radiates two light bulbs' worth. LIGO detects the former, in the last fraction of a second before coalescence, where the quadrupole formula strains at its limits.

Where does the quadrupole formula break down, and what replaces it?

It breaks down when the source stops being slow and weak — precisely at the merger detectors most want to model. The formula is the leading term of an expansion in $v/c$ , with corrections at order $v^2/c^2$ from the current quadrupole, mass octupole, and non-linear "tail" terms where the outgoing wave scatters off the source's own curvature. For an inspiral these post-Newtonian corrections are computed to high order and are essential for templates, but as $v/c$ approaches $0.4$ – $0.5$ near merger the series stops converging usefully, and only numerical relativity — solving the full Einstein equations on a computer — is reliable. Even then the quadrupole luminosity remains the anchor: it fixes the leading behaviour every post-Newtonian order and every numerical waveform must reproduce in the slow-motion limit.