Astrophysical Sources of Gravitational Waves

Gravitational waves are ripples in the curvature of spacetime, propagating at the speed of light, sourced by accelerating mass-energy. In the weak-field, slow-motion regime a single expression — the quadrupole formula — governs both the radiated field and the luminosity through the second time derivative of the source's mass quadrupole moment. Two conservation laws close the loudest channels: energy conservation forbids monopole radiation, momentum conservation forbids mass-dipole radiation. Gravity radiates only from the quadrupole and higher multipoles, two orders up from electromagnetism, where the dipole dominates. That one structural fact organizes the entire catalogue: an efficient emitter must be massive, compact, fast, and non-spherical. The catalogue splits into four families. Compact binary coalescences — black holes and neutron stars spiralling together — are the loudest: as the orbit tightens, amplitude and frequency climb into the characteristic chirp, whose leading waveform is fixed by one number, the chirp mass $\mathcal{M}$ . Supernovae contribute short bursts; spinning deformed neutron stars emit nearly monochromatic continuous waves; and beneath every resolvable signal lies a stochastic background, the incoherent sum of countless unresolved sources, astrophysical and possibly cosmological. Which family you can hear is set by frequency — ground interferometers cover $10$ – $10^{4}\,\mathrm{Hz}$ , LISA the millihertz, pulsar timing arrays the nanohertz. GW150914 in 2015 and the nanohertz-background evidence of 2023 have opened this spectrum across nine decades.

Key takeaways

Astrophysical gravitational waves arise wherever a mass distribution has a time-varying mass quadrupole; monopole and dipole radiation are forbidden by conservation of energy and momentum, so the quadrupole leads.
The sources fall into four classes — coalescing compact binaries (transient chirps), supernova bursts, continuous waves from spinning deformed neutron stars, and stochastic backgrounds — each dominant in a distinct band.
Strains at Earth are minuscule: a neutron-star merger at $\sim100\,\mathrm{Mpc}$ gives peak strain $h\sim10^{-22}$ , a length change of $\sim10^{-19}\,\mathrm{m}$ across a kilometre-scale detector.
Compact-binary inspirals are the cleanest sources: the chirp mass $\mathcal{M}=(m_1m_2)^{3/5}/(m_1+m_2)^{1/5}$ alone sets the leading amplitude and the frequency sweep $\dot f$ , so measuring $f$ and $\dot f$ reads off $\mathcal{M}$ and the distance directly.
Detection band selects the population: ground interferometers ( $10$ – $10^{4}\,\mathrm{Hz}$ ) catch stellar-mass mergers, LISA ( $10^{-4}$ – $0.1\,\mathrm{Hz}$ ) massive black-hole and white-dwarf binaries, pulsar timing arrays ( $10^{-9}$ – $10^{-7}\,\mathrm{Hz}$ ) supermassive binaries.
GW150914 (2015) — black holes of $\sim36$ and $\sim29\,M_\odot$ merging at $\sim410\,\mathrm{Mpc}$ , radiating $\sim3\,M_\odot c^2$ — proved stellar-mass binary black holes exist and merge within a Hubble time.
Key idea: All astrophysical gravitational radiation flows from one root — a time-varying mass quadrupole. The monopole and dipole channels are closed by conservation of energy and momentum, so only compact, massive, fast, asymmetric systems radiate detectably, and the chirp mass alone fixes the leading inspiral waveform.

The core insight

All astrophysical gravitational radiation flows from one root: a time-varying mass quadrupole. Monopole and dipole channels are closed by conservation of energy and momentum, so only compact, massive, fast, asymmetric systems radiate detectably — and the chirp mass alone fixes the leading inspiral waveform.

Where this lives today

Gravitational-wave astronomy is now a running catalogue. The LIGO-Virgo-KAGRA network has logged more than a hundred compact-binary mergers, mapping the mass distribution of black holes and neutron stars and finding "mass-gap" objects that stress formation models. GW170817 — the 2017 binary neutron-star merger seen at once in gravitational waves, a short gamma-ray burst, and an optical kilonova — founded multimessenger astronomy: it confirmed that mergers forge heavy r-process elements, pinned the speed of gravity to $c$ within one part in $10^{15}$ , and gave a standard-siren $H_0$ . In 2023 the pulsar-timing-array collaborations reported a nanohertz background with the tell-tale Hellings-Downs correlation, most plausibly the hum of supermassive black-hole binaries. Ahead, LISA (2030s) will open the millihertz band to massive black-hole mergers, extreme-mass-ratio inspirals, and Galactic white-dwarf binaries, while Einstein Telescope and Cosmic Explorer target stellar-mass mergers to the edge of the observable universe. Every one is the same time-varying quadrupole.

The physical picture

Picture a pair of masses whirling around each other as a rotating dumbbell. Its mass distribution is squashed out of spherical symmetry, and that bulge sweeps around the sky, shaking spacetime. Because the dumbbell looks identical after half a turn, the quadrupole pattern repeats twice per orbit, and the wave comes out at twice the orbital frequency. A perfectly spherical pulsation, a star that merely breathes, has no quadrupole and radiates nothing: only the changing shape couples to the wave field.

Figure

A rotating binary is a mass quadrupole: its non-spherical shape repeats every half turn, so it radiates gravitational waves at twice the orbital rate, f_GW = 2 f_orb. A neutron-star merger at about 100 Mpc reaches Earth as a peak strain of only h ≈ 2×10⁻²².

Gravity is the weakest interaction, so even a cataclysm produces a strain of order $10^{-21}$ . The energy budget is nonetheless staggering: GW150914 radiated about $3\,M_\odot c^2$ in a fraction of a second, briefly outshining the entire luminous universe — yet by the time the wave reached Earth it moved a four-kilometre detector arm by less than one-thousandth the width of a proton. Frequency is destiny: a binary's wave frequency rises as its orbit tightens, so heavier systems ring lower and lighter ones higher, and choosing a detector's arm length and noise floor is choosing which class of the cosmos you can hear, as the map below sets out.

Figure

Frequency is destiny. Pulsar timing arrays listen at 10⁻⁹–10⁻⁷ Hz to supermassive black-hole binaries, LISA at 10⁻⁴–0.1 Hz to massive black-hole and white-dwarf binaries, and ground interferometers at 10–10⁴ Hz to stellar-mass neutron-star and black-hole mergers — each band a different population.

The core equations

The starting point is the radiation field. Solving the linearized Einstein equation with a matter source and keeping the leading far-zone, slow-motion term gives the transverse-traceless strain as the second time derivative of the mass quadrupole, retarded and falling off as $1/r$ .

\bar{h}_{ij}^{\mathrm{TT}}(t,\mathbf{x}) = \frac{2G}{c^4 r}\,\ddot{I}_{ij}^{\mathrm{TT}}\!\left(t - \frac{r}{c}\right) \qquad (1)

where: $\bar{h}_{ij}^{\mathrm{TT}}$ — transverse-traceless metric perturbation, the physical strain; $I_{ij}$ — mass quadrupole moment of the source; $r$ — source-to-observer distance; $t-r/c$ — retarded time; $G,\,c$ — Newton's constant and the speed of light. Reading it: the wave is two time derivatives of the source's shape, so a static or uniformly moving mass radiates nothing, and the prefactor $G/c^4 \approx 8\times10^{-45}$ makes the effect faint. Squaring the field and averaging its energy flux over a few wavelengths (the Isaacson procedure) gives the total radiated power — the quadrupole luminosity, which ranks every source.

L_{\mathrm{GW}} = \frac{1}{5}\frac{G}{c^5}\left\langle \dddot{Q}_{ij}\,\dddot{Q}^{ij} \right\rangle \qquad (2)

where: $L_{\mathrm{GW}}$ — total gravitational-wave luminosity; $Q_{ij}$ — trace-free (reduced) mass quadrupole moment; $\langle\cdot\rangle$ — average over several periods. Reading it: the natural luminosity scale is $c^5/G \approx 3.6\times10^{52}\,\mathrm{W}$ ; comparable-mass mergers reach $\sim10^{-3}$ of it at peak — already brighter than all the stars in the sky combined. For a two-body system the entire mass dependence of the leading waveform collapses into one combination, the chirp mass — no other mass parameter appears at Newtonian order.

\mathcal{M} = \frac{(m_1 m_2)^{3/5}}{(m_1+m_2)^{1/5}} \qquad (3)

where: $\mathcal{M}$ — chirp mass, the sole leading-order mass parameter of an inspiral; $m_1,\,m_2$ — component masses. Reading it: for two equal masses $m$ it equals $2^{-1/5}m \approx 0.87\,m$ , and it is what an inspiral measures — not the individual masses. Balancing radiated power against the shrinking orbital energy then gives how fast the frequency climbs — this equation is the chirp.

\dot{f}_{\mathrm{GW}} = \frac{96}{5}\pi^{8/3}\left(\frac{G\mathcal{M}}{c^3}\right)^{5/3} f_{\mathrm{GW}}^{11/3} \qquad (4)

where: $f_{\mathrm{GW}}$ — gravitational-wave frequency ( $=2f_{\mathrm{orb}}$ ); $\dot{f}_{\mathrm{GW}}$ — its rate of increase; $\mathcal{M}$ — chirp mass, extractable from $f$ and $\dot f$ alone. Reading it: the steep $f^{11/3}$ dependence makes the frequency run away near merger — the audible rush from tens to hundreds of hertz in a fraction of a second — and its slope weighs $\mathcal{M}$ . The observable amplitude of an inspiral follows by combining the quadrupole field with circular-orbit kinematics, carrying distance and chirp mass together.

h \sim \frac{4}{r}\left(\frac{G\mathcal{M}}{c^2}\right)^{5/3}\left(\frac{\pi f_{\mathrm{GW}}}{c}\right)^{2/3} \qquad (5)

where: $h$ — strain amplitude at the detector; $r$ — luminosity distance; other symbols as above. Reading it: because $\mathcal{M}$ is read from the phasing $(f,\dot f)$ while the amplitude scales as $1/r$ , an inspiral is a self-calibrating "standard siren" yielding its own distance. Two further observables, tabulated in the Limiting cases below, govern the diffuse and continuous classes: the stochastic background's fractional energy density $\Omega_{\mathrm{GW}}(f)=\rho_c^{-1}\,\mathrm{d}\rho_{\mathrm{GW}}/\mathrm{d}\ln f$ (with $\rho_c=3H_0^2 c^2/8\pi G$ ), and the near-monochromatic tone $h_0=(16\pi^2 G/c^4)\,\varepsilon I_{zz}f_{\mathrm{rot}}^2/r$ radiated at twice spin frequency by a neutron star of ellipticity $\varepsilon=(I_{xx}-I_{yy})/I_{zz}$ . Reading it: expected ellipticities are $\varepsilon\lesssim10^{-6}$ , so continuous-wave strains are $\sim10^{-25}$ or smaller — recoverable only by coherently integrating a known-frequency signal for months.

Derivation APower radiated by a circular binary and the inspiral lifetime

The goal is to turn the quadrupole luminosity into two testable numbers for the loudest source class: how brightly a circular Keplerian binary radiates, and how long it lives before the orbit shrinks to nothing. We build the reduced quadrupole of two Newtonian point masses, differentiate three times for the luminosity, set the radiated power equal to the loss of orbital energy, and integrate for the separation — arriving at the Peters merger-time formula, whose steep dependence on separation explains why the observable population is the tight, near-merger systems.

Step 1. Reduce to one effective particle: with reduced mass $\mu = m_1 m_2/(m_1+m_2)$ at separation $a$ , the trace-free quadrupole is $Q_{ij} = \mu\left(x_i x_j - \tfrac13\delta_{ij}r^2\right)$ . Why: only the relative motion carries a changing quadrupole — the centre of mass moves uniformly and cannot radiate.

Step 2. Put the particle on a circular orbit, $x=a\cos\Omega t$ , $y=a\sin\Omega t$ , and differentiate the quadrupole three times. Forming $\langle\dddot Q_{ij}\dddot Q^{ij}\rangle$ gives a constant proportional to $\mu^2 a^4\Omega^6$ . Why: circular motion makes the quadrupole rotate rigidly, so its third derivative has constant magnitude and the luminosity is steady, not pulsed.

Step 3. Insert this into equation (2) and eliminate $\Omega$ with Kepler's third law $\Omega^2 = G(m_1+m_2)/a^3$ , giving $L_{\mathrm{GW}} = \tfrac{32}{5}(G^4/c^5)(m_1 m_2)^2(m_1+m_2)/a^5$ . Why: the $a^{-5}$ scaling is the engine of the story — halving the separation raises the power thirty-twofold, so radiation runs away as the orbit tightens.

Step 4. Impose energy balance. With $E=-Gm_1m_2/2a$ , so $\dot E = (Gm_1m_2/2a^2)\dot a$ , setting $\dot E = -L_{\mathrm{GW}}$ turns the luminosity into an equation for the shrinking separation. Why: the binary has no other energy reservoir, so every joule radiated comes out of the orbit, forcing $a$ to decay.

Step 5. The result is separable, $a^3\,\mathrm{d}a/\mathrm{d}t = -\tfrac{64}{5}(G^3/c^5)\,m_1 m_2(m_1+m_2)$ . Integrating from an initial separation $a_0$ down to $a=0$ gives the coalescence time $t_{\mathrm{merge}} = \dfrac{5}{256}\dfrac{c^5 a_0^4}{G^3 m_1 m_2(m_1+m_2)}$ . Why: the $a_0^4$ dependence is decisive — a binary twice as wide takes sixteen times as long, so wide binaries live longer than the universe while tight ones merge within its band.

t_{\mathrm{merge}} = \frac{5}{256}\,\frac{c^5 a_0^4}{G^3\, m_1 m_2 (m_1+m_2)}

With the eccentricity enhancement factor included, this formula reproduces the Hulse-Taylor pulsar's measured orbital decay $\dot P_b \approx -2.4\times10^{-12}\,\mathrm{s/s}$ to better than a percent.

Go deeper · Derivation B · 1 minWhy gravity has no monopole and no dipole radiation

Derivation BWhy gravity has no monopole and no dipole radiation

The goal is to prove, straight from conservation of the source's stress-energy, that the two lowest multipoles are radiatively silent — the structural reason gravitational waves are so faint. We compute the mass monopole and dipole, take their time derivatives using the continuity equation $\partial_\mu T^{\mu\nu}=0$ , and show that the derivatives radiation would need are killed by energy and momentum conservation, leaving the quadrupole leading.

Step 1. The monopole is the total mass-energy, $M = c^{-2}\int T^{00}\,\mathrm{d}^3x$ . Energy conservation gives $\dot M = 0$ , so the monopole part of the far field is static — no radiation. Why: a pulsating but spherical source (a "breathing" star) changes only its monopole, which cannot vary, so it is gravitationally silent.

Step 2. The mass dipole is $d^i = c^{-2}\int T^{00}x^i\,\mathrm{d}^3x = M x^i_{\mathrm{cm}}$ , the centre-of-mass position times the total mass. Why: it is the gravitational analogue of the electric dipole, so naively one expects it to radiate.

Step 3. Its first derivative, using continuity $\partial_0 T^{00} = -\partial_j T^{0j}$ and integrating by parts, is $\dot d^i = c^{-1}\int T^{0i}\,\mathrm{d}^3x = P^i$ , the total momentum. Why: the "current" of mass-energy is momentum, so the dipole simply tracks the uniformly drifting centre of mass.

Step 4. The second derivative is $\ddot d^i = \dot P^i = 0$ , because momentum is conserved for an isolated system. Since radiation depends on second time derivatives, the dipole channel is closed. Why: unlike electromagnetism, where positive and negative charges let a dipole oscillate, gravitational "charge" is one-signed mass, and momentum conservation forbids the oscillation outright.

Step 5. The same manipulation on the second moment $\int T^{00}x^i x^j$ does not vanish at second order: $\ddot I_{ij}\neq0$ generically. The quadrupole survives as the leading radiative multipole — one order up from electromagnetism. Why: this single fact makes gravitational radiation a feeble, quadrupolar effect, which is why detectors as extreme as LIGO are needed.

\dot M = 0,\qquad \ddot d^{\,i} = \dot P^{\,i} = 0 \;\Rightarrow\; \text{leading radiation is quadrupolar}

Go deeper · Derivation C · 1 minThe chirp: reading the masses and distance off the frequency sweep

Derivation CThe chirp: reading the masses and distance off the frequency sweep

The goal is to show how the energy balance above becomes the central observable of gravitational-wave astronomy — the frequency evolution $\dot f(f)$ — and how measuring it turns a binary into both a mass scale and a distance marker, the "standard siren." We rewrite the balance through the gravitational-wave frequency, watch the mass combination collapse to the chirp mass, then combine phasing with amplitude to see the distance drop out.

Step 1. From energy balance $\dot E = -L_{\mathrm{GW}}$ with $E=-Gm_1m_2/2a$ and Kepler's law, re-express the shrinking orbit as a rising frequency using $f_{\mathrm{GW}} = \Omega/\pi = (1/\pi)\sqrt{G(m_1+m_2)/a^3}$ . Why: the detector measures frequency, not separation, so the physics must be rewritten in the observable variable.

Step 2. Substituting, every appearance of the masses reorganizes into the single combination $\mathcal{M}=(m_1m_2)^{3/5}/(m_1+m_2)^{1/5}$ , yielding $\dot f_{\mathrm{GW}} = \tfrac{96}{5}\pi^{8/3}(G\mathcal{M}/c^3)^{5/3} f_{\mathrm{GW}}^{11/3}$ . Why: this is the remarkable collapse — the two masses and total mass conspire into one number, so a Newtonian inspiral measures $\mathcal{M}$ and nothing else about the masses.

Step 3. Integrate the sweep for the time to coalescence from a frequency $f$ : $\tau = \tfrac{5}{256}(c^3/G\mathcal{M})^{5/3}(\pi f)^{-8/3}$ . Why: it converts the runaway into a stopwatch — how many seconds of signal remain once the source enters the band.

Step 4. Combine the phasing with the amplitude (5): $h\sim (4/r)(G\mathcal{M}/c^2)^{5/3}(\pi f/c)^{2/3}$ . Because $\mathcal{M}$ is already fixed by $\dot f$ , the only remaining unknown in the amplitude is $r$ . Why: the waveform therefore encodes its own luminosity distance with no cosmic distance ladder — the "standard siren" that measures $H_0$ directly.

\tau(f) = \frac{5}{256}\left(\frac{c^3}{G\mathcal{M}}\right)^{5/3}(\pi f)^{-8/3},\qquad h\,r = 4\left(\frac{G\mathcal{M}}{c^2}\right)^{5/3}\left(\frac{\pi f}{c}\right)^{2/3}

The sweep and time-to-merger formulas reproduce the $\sim0.2\,\mathrm{s}$ in-band duration of GW150914 and $\sim100\,\mathrm{s}$ of GW170817; the standard-siren logic underlies the GW170817 $H_0$ measurement.

Limiting cases

Each source class is one regime of the same quadrupole emission, in its own frequency band:

Case	Result	Meaning
Compact binary inspiral	$f_{\mathrm{GW}}=2f_{\mathrm{orb}},\ \ h\sim\dfrac{4}{r}\left(\dfrac{G\mathcal{M}}{c^2}\right)^{5/3}\left(\dfrac{\pi f_{\mathrm{GW}}}{c}\right)^{2/3}$	Deterministic chirp; amplitude and sweep fixed by $\mathcal{M}$ and $r$ . The cleanest, loudest source.
Rotating neutron star (continuous)	$h_0=\dfrac{16\pi^2 G}{c^4}\dfrac{\varepsilon I_{zz}f_{\mathrm{rot}}^2}{r}$	Near-monochromatic at $2f_{\mathrm{rot}}$ ; ellipticity $\varepsilon\lesssim10^{-6}$ makes it very weak, needing months of integration.
Burst (supernova)	$E_{\mathrm{GW}}\sim\dfrac{\pi^2c^3}{G}\,f^2\,\tau\,h^2 r^2$	Short, poorly-modelled transient from asymmetric collapse; efficiency uncertain, $\lesssim10^{-8}\,M_\odot c^2$ .
Stochastic background	$\Omega_{\mathrm{GW}}(f)=\dfrac{1}{\rho_c}\dfrac{\mathrm{d}\rho_{\mathrm{GW}}}{\mathrm{d}\ln f}$	Incoherent statistical signal; found by cross-correlating instruments (Hellings-Downs for pulsar arrays).

Common trapExpecting dipole radiation, and trusting the quadrupole formula through merger

Two traps recur. First, the electromagnetic analogy tempts one to expect dipole gravitational radiation — but the gravitational "charge" is mass, its dipole moment $\sum_a m_a\mathbf{x}_a = M\mathbf{x}_{\mathrm{cm}}$ has a first derivative equal to total momentum and a second derivative that vanishes by momentum conservation. There is no dipole channel; the quadrupole is leading, which is exactly why the effect is so weak. Second, the quadrupole formula is only the leading post-Newtonian (weak-field, slow-motion) term. It works beautifully through the long inspiral but breaks down when the signal is strongest — the final orbits and merger, where the bodies move at a sizeable fraction of $c$ and the fields are strong. Those waveforms require full numerical relativity, and quoting a precise chirp mass from data that includes the merger using only equation (4) is a common, serious error.

Worked examples

1 · GW150914 — chirp mass, strain, and energy (fully worked)

The first detection came from black holes of $m_1\approx36\,M_\odot$ and $m_2\approx29\,M_\odot$ at luminosity distance $r\approx410\,\mathrm{Mpc}$ . Find the chirp mass, estimate the peak strain near $f_{\mathrm{GW}}=150\,\mathrm{Hz}$ , and check the radiated-energy scale.

Step 1. Chirp mass from equation (3): $\mathcal{M}=(36\cdot29)^{3/5}/(65)^{1/5}\,M_\odot$ . Numerically $(1044)^{0.6}=64.7$ and $(65)^{0.2}=2.30$ , so $\mathcal{M}=64.7/2.30\approx\mathbf{28.1\,M_\odot}$ — the single number the inspiral phasing measures.

Step 2. Convert to length units: $G\mathcal{M}/c^2 \approx 4.15\times10^{4}\,\mathrm{m}$ (about $41\,\mathrm{km}$ ), and $r=410\,\mathrm{Mpc}=1.27\times10^{25}\,\mathrm{m}$ .

Step 3. Strain from equation (5) at $f_{\mathrm{GW}}=150\,\mathrm{Hz}$ : with $(G\mathcal{M}/c^2)^{5/3}\approx5.0\times10^{7}$ and $(\pi f/c)^{2/3}\approx1.36\times10^{-4}$ (SI), $h\approx (4/1.27\times10^{25})\cdot5.0\times10^{7}\cdot1.36\times10^{-4}\approx\mathbf{2\times10^{-21}}$ , matching the observed peak strain of order $10^{-21}$ .

Step 4. Energy: the merger radiated about $3\,M_\odot c^2$ out of a $\sim65\,M_\odot$ system — a fraction $3/65\approx\mathbf{4.6\%}$ of the total rest mass, an efficiency no chemical or nuclear process approaches. The peak luminosity $\sim3.6\times10^{49}\,\mathrm{W}$ is about $10^{-3}$ of the natural scale $c^5/G\approx3.6\times10^{52}\,\mathrm{W}$ , briefly greater than the combined light of every star in the observable universe.

2 · How long does a binary sing in the band?

Contrast the two flagship events using the time-to-coalescence $\tau = \tfrac{5}{256}(c^3/G\mathcal{M})^{5/3}(\pi f)^{-8/3}$ from Derivation C.

Step 1. GW150914 ( $\mathcal{M}=28.1\,M_\odot$ ) enters the LIGO band at $f_{\mathrm{GW}}\approx35\,\mathrm{Hz}$ . With $G\mathcal{M}/c^3 = 4.15\times10^4/2.998\times10^8\approx1.38\times10^{-4}\,\mathrm{s}$ , evaluating gives $\tau\approx\mathbf{0.19\,s}$ — the signal is a fifth of a second, a true "chirp."

Step 2. GW170817 is a binary neutron star, $\mathcal{M}\approx1.188\,M_\odot$ , entering at $f_{\mathrm{GW}}\approx24\,\mathrm{Hz}$ . Because $\tau\propto\mathcal{M}^{-5/3}$ , the $\sim24\times$ smaller chirp mass lengthens the signal enormously: $\tau\approx\mathbf{100\,s}$ in band — long enough to localize on the sky and alert telescopes.

Step 3. Where does the chirp stop? A rough cutoff is the innermost stable circular orbit, $f_{\mathrm{ISCO}}\approx c^3/(6^{3/2}\pi G M_{\mathrm{tot}})$ ; for GW150914's $M_{\mathrm{tot}}\approx65\,M_\odot$ the point-mass estimate gives $\sim70\,\mathrm{Hz}$ , yet the true merger rang up to $\sim250\,\mathrm{Hz}$ — for comparable masses ISCO only orients you, and the real waveform needs numerical relativity. Heavy systems chirp briefly at low frequency; light neutron stars chirp for minutes at high frequency. (By contrast a deformed pulsar — $\varepsilon\sim10^{-6}$ , $I_{zz}\sim10^{38}\,\mathrm{kg\,m^2}$ , $f_{\mathrm{rot}}=100\,\mathrm{Hz}$ at $1\,\mathrm{kpc}$ — gives only $h_0\approx4\times10^{-26}$ , recoverable solely by months of coherent integration.)

Physics of the portrait

The theory of astrophysical gravitational-wave sources is where a fifty-year argument about whether the waves were even real gave way to precision science — and one figure sits at that hinge.

Origins · 1 minKip Thorne and how this was discovered

Kip Thorne

1940– · theory of astrophysical gravitational-wave sources and their detection

Kip Thorne turned gravitational waves from a theorist's curiosity into an observational program. When he began, it was not even settled that the waves carried energy — Einstein himself had wavered, and the point was only resolved at the 1957 Chapel Hill conference through Bondi's and Feynman's "sticky-bead" argument that a passing wave does real work. Thorne took the next step: if the waves are real, which systems make detectable ones? Through the 1970s and 1980s he and his students built the modern source catalogue — the quadrupole estimates for compact binaries, supernovae, and spinning neutron stars — and codified the multipole formalism for the radiation of any source. He championed the standard-siren idea and argued, against deep skepticism, that laser interferometers could reach the $10^{-21}$ strains the sources demanded. That conviction became LIGO, co-founded with Rainer Weiss and Ronald Drever; GW150914 in 2015 vindicated the enterprise, and Thorne shared the 2017 Nobel Prize with Weiss and Barry Barish. The sources in this article are largely the ones he first taught the field to expect.

History

1916

— Einstein predicts gravitational waves from the linearized field equations

1918

— Einstein derives the quadrupole formula for the radiated power

1963

— Peters and Mathews compute the waveform and energy loss of a Keplerian binary

1974

— Hulse and Taylor discover the binary pulsar PSR B1913+16, indirect proof of emission

2015

— LIGO records GW150914, the first direct detection of a compact-binary merger

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), Ch. 36 — generation of gravitational waves · Wald, General Relativity (1984), §4.4 — energy and gravitational radiation · Carroll, Spacetime and Geometry (2004), Ch. 7 — gravitational waves and the binary pulsar · Hartle, Gravity (2003), Ch. 16 & 23 — emission and detection · Schutz, A First Course in General Relativity (2009), Ch. 9 — gravitational radiation · Maggiore, Gravitational Waves Vol. 1 (2008), Ch. 3–4 — generation and sources. Key numerical anchors: $\mathcal{M}(36,29\,M_\odot)=28.1\,M_\odot$ , $\mathcal{M}(\text{GW170817})\approx1.188\,M_\odot$ , $\tau(35\,\mathrm{Hz})\approx0.19\,\mathrm{s}$ , and $\tau(24\,\mathrm{Hz})\approx100\,\mathrm{s}$ . Event parameters are from Abbott et al., PRL 116, 061102 (2016) for GW150914 and PRL 119, 161101 (2017) for GW170817; the nanohertz background evidence from the 2023 pulsar-timing-array papers.

Builds on: Quadrupole Formula · Gravitational Wave Equation · Transverse-Traceless Gauge · Isaacson Stress-Energy · Plus and Cross Polarizations. Leads to: Multipole Expansion of Gravitational Radiation · Isaacson Stress-Energy of Gravitational Waves.

Ask the tutor: Ask me to walk through the Peters merger-time integral, or to estimate the chirp mass and in-band lifetime of any binary you name.

Why does gravity radiate from the quadrupole while light radiates from the dipole?

Because the "charge" of gravity is mass, which has only one sign. In electromagnetism a positive and a negative charge oscillate against each other, giving a radiating time-varying dipole. For gravity the mass dipole is $M\mathbf{x}_{\mathrm{cm}}$ ; its first time derivative is the total momentum and its second the net force on an isolated system — which is zero. Since radiation depends on second derivatives of the moments, both the monopole (fixed by energy conservation) and the dipole (fixed by momentum conservation) are silent, and the lowest freely-varying moment is the quadrupole, the shape of the source. That is why gravitational radiation is intrinsically weak and only compact, fast, non-spherical systems are detectable.

What is the chirp mass, and why does it matter more than the individual masses?

The chirp mass $\mathcal{M}=(m_1m_2)^{3/5}/(m_1+m_2)^{1/5}$ is the one mass combination appearing in the Newtonian-order waveform — in both the amplitude and, crucially, the frequency sweep $\dot f$ . Because the phasing is measured very precisely, $\mathcal{M}$ is the best-determined parameter of any inspiral, often to a few percent even for a marginal detection. The individual masses enter only through weaker higher-order post-Newtonian corrections, so they are measured far less precisely and are partly degenerate with the spins — the chirp tells you $\mathcal{M}$ loudly and the mass ratio only faintly.

How can a gravitational wave carry energy if the metric perturbation can be gauged away locally?

Locally you can always choose coordinates in which the first derivatives of the metric vanish (the equivalence principle), so there is no gauge-invariant local energy density for the field. The resolution is to average over a region several wavelengths across: the Isaacson stress-energy tensor, built from the second-order short-wavelength-averaged curvature, is gauge-invariant and genuinely transports energy — exactly the flux in the quadrupole luminosity (2) that drives a binary's orbit to shrink. The "sticky-bead" argument makes it concrete: a passing wave does measurable work on freely suspended masses, so the energy is real even without a local density.

Why does the quadrupole formula break down at merger, when the signal is loudest?

The formula is the leading post-Newtonian term, valid when the source is slow ( $v\ll c$ ) and weakly self-gravitating — conditions that hold beautifully through the long inspiral. But in the final orbits the bodies reach a substantial fraction of $c$ and their fields become strong, so the neglected higher-order and nonlinear terms are no longer small, precisely where the amplitude peaks. Accurate merger waveforms therefore require full numerical relativity or calibrated effective-one-body models. The quadrupole formula still sets the scale and inspiral phasing, but cannot be trusted for the peak strain or ringdown.

What is a standard siren, and how does it measure the Hubble constant?

An inspiral is a "standard siren" — the gravitational analogue of a standard candle — because its waveform encodes its own luminosity distance. The sweep $\dot f$ fixes the chirp mass, which fixes the intrinsic amplitude; comparing that to the measured strain gives the distance $r$ directly, with no cosmic distance ladder. If the host galaxy's redshift is also known — as for GW170817, through its electromagnetic counterpart — pairing distance with redshift yields $H_0$ independently of Cepheids or supernovae. This ladder-free rung may, with many future sirens, help resolve the current tension between local and cosmological values of $H_0$ .

Why do different detectors target completely different frequencies?

Because a binary's gravitational-wave frequency is set by its orbital frequency, which scales inversely with size and mass. Stellar-mass objects merge when their orbits shrink to tens of kilometres, ringing at tens to thousands of hertz — the ground-interferometer band. Massive black holes ( $10^4$ – $10^7\,M_\odot$ ) are far larger, so their mergers ring at millihertz frequencies only LISA can reach. Supermassive black-hole binaries orbit over years, producing nanohertz waves seen only by watching pulsar clocks drift for decades. No single instrument spans this range, so the spectrum is carved among complementary detectors, each tuned to a distinct population.