Compact Binary Coalescence Waveforms

Waveform modeling is the beating heart of an operating observatory. The LIGO–Virgo–KAGRA network matched-filters hundreds of thousands of templates in near-real time, issuing public alerts within minutes so telescopes can chase electromagnetic counterparts. Every event in the GWTC catalogs — now in the hundreds — is a full Bayesian inference against effective-one-body, IMRPhenom, and surrogate waveforms, and their accuracy is itself a limiting systematic: as detectors sharpen, waveform error rather than noise increasingly caps how well masses, spins, and tides are measured. Population inference over the catalog maps the mass spectrum (including a possible pair-instability gap), the spin distribution that separates isolated-binary from dynamical-cluster formation, and the merger-rate history. GW170817's tidal phase constrained the neutron-star radius and disfavored the stiffest equations of state, while its standard-siren distance gave a ladder-independent $H_0$. Ahead, LISA will track extreme-mass-ratio inspirals for years with self-force waveforms, and Einstein Telescope and Cosmic Explorer will demand models accurate to a part in $10^4$ — an active frontier.

Picture two masses orbiting their common center. General relativity says a time-varying mass quadrupole radiates energy as ripples in spacetime. As energy leaves, the orbit tightens and the bodies speed up, so both the wave's amplitude and its frequency climb together — a rising "chirp" you can literally hear when the strain is played as sound. Crucially the waveform is not a fit made after the fact; it is a first-principles prediction made before any data exist, which is exactly why a matched template can dig a signal out of noise a hundred times louder. The gravitational-wave frequency is twice the orbital frequency, because a mass quadrupole (a dumbbell) repeats twice per orbit. The first figure shows the resulting three-act signal.

The three phases are three languages for one story: post-Newtonian theory is a careful low-speed expansion, good while the orbit is wide; numerical relativity is brute-force integration of Einstein's equations when speeds reach a third of light and the fields overlap violently; and the ringdown is the settling tone of a struck bell, a black hole vibrating at frequencies fixed only by its mass and spin. Matched filtering ties them together — like listening for a known melody buried in static, sliding the expected tune across the recording to find the instant of maximum correlation. The second figure makes the chirp quantitative: the frequency sweep the Newtonian chirp law predicts for a GW150914-like binary.

The starting point is the leading-order strain from the mass-quadrupole formula: two point masses on a circular orbit radiate a wave whose two polarizations combine into a single complex amplitude that scales with reduced mass and orbital frequency and falls off inversely with distance.

where: $h_+, h_\times$ — plus and cross gravitational-wave polarizations; $d_L$ — luminosity distance; $\mu = m_1 m_2/M$ — reduced mass; $M = m_1+m_2$ — total mass; $\omega_{\rm orb}$ — orbital angular frequency; $\Phi(t)$ — orbital phase, and the GW phase is $2\Phi$. Read it as: amplitude grows with orbital frequency ($\propto \omega^{2/3}$); the phase advances at twice the orbital rate. The masses enter the leading strain and frequency sweep only through one combination, the chirp mass — the most precisely measured parameter of any inspiral. Derivation 1 shows why this grouping is forced on us.

where: $\mathcal{M}$ — chirp mass, sets the leading inspiral phase and amplitude; $\eta = m_1 m_2 / M^2$ — symmetric mass ratio, bounded $0 < \eta \le 1/4$. Read it as: two masses collapse into one measurable number; recovering $m_1$ and $m_2$ separately requires $\eta$, which hides in higher-order corrections. The engine is the quadrupole luminosity — the rate the orbit loses energy to gravitational waves — which for a circular orbit depends steeply on separation and orbital frequency.

where: $E$ — orbital energy; $a$ — orbital separation (circular orbit); $G, c$ — gravitational constant and speed of light. Read it as: the tiny prefactor $G/c^5$ is why gravitational radiation is negligible for planets but dominant for compact objects at small $a$ and high $\omega$. Balancing that luminosity against the orbit's binding energy gives the observable frequency sweep — the chirp rate that matched-filter searches ultimately track.

where: $f$ — gravitational-wave frequency $= \omega_{\rm orb}/\pi$; $\dot f$ — its time derivative. Read it as: frequency accelerates as $f^{11/3}$, so almost all the sweep happens in the final moments, and only $\mathcal{M}$ sets the rate at leading order. In the frequency domain the inspiral takes a compact closed form via the stationary-phase approximation — the workhorse "TaylorF2" template. The amplitude is a simple power law; all the parameter information lives in the phase, a post-Newtonian series.

where: $\tilde h(f)$ — Fourier transform of the strain; $\mathcal{A} \propto \mathcal{M}^{5/6}/d_L$ — amplitude; $\Psi(f)$ — frequency-domain phase; $t_c, \phi_c$ — coalescence time and phase; $\psi_k$ — PN phase coefficients depending on masses and spins. Read it as: the leading $(\pi\mathcal{M}f)^{-5/3}$ term is pure chirp mass; spins and tides enter as successive corrections. Detection is an inner product: a template is cross-correlated with the data weighted by the inverse noise spectrum, and the optimal signal-to-noise ratio is the template's norm in that same noise-weighted metric.

where: $\rho$ — signal-to-noise ratio; $S_n(f)$ — one-sided detector noise power spectral density; $s$ — detector data; $\langle h|s\rangle$ — noise-weighted overlap that matched filtering maximizes over $\theta$. Read it as: sensitivity is weighted toward where the detector is quiet, so the noise curve $S_n(f)$ shapes what is detectable. After merger the remnant rings down as a sum of damped sinusoids — quasi-normal modes whose frequencies are fixed by the final mass and spin alone, the no-hair theorem made audible.

where: $M_f, a_f$ — final black-hole mass and dimensionless spin; $\Omega_{\ell m n}$ — dimensionless mode frequency from black-hole perturbation theory; $\tau_{\ell m n}$ — mode damping time; $\ell, m, n$ — angular and overtone indices, dominant mode $\ell=m=2, n=0$. Read it as: measuring two or more modes over-determines $(M_f, a_f)$, so consistency is a null test of general relativity. For neutron stars, finite size leaves a fingerprint in the high-frequency phase — the tidal deformability, entering formally at 5PN order and measurable only near merger.

where: $\tilde\Lambda$ — mass-weighted binary tidal deformability; $\Lambda_i = \tfrac{2}{3}k_2^{(i)}(c^2 R_i/G m_i)^5$ — dimensionless tidal deformability of body $i$; $k_2$ — tidal (apsidal) Love number; $R_i$ — neutron-star radius. Read it as: because $\Lambda \propto (R/m)^5$, the phase is exquisitely sensitive to the radius, and black holes have $k_2 = 0$ so $\tilde\Lambda = 0$.

The waveform's behavior in limiting regimes exposes which physics dominates where, and which parameters are measurable when.

1 · The chirp mass and mass ratio of GW150914 (confidence builder)

GW150914 was fit by two black holes of roughly $m_1 = 36\,M_\odot$ and $m_2 = 29\,M_\odot$ (source frame). Compute the chirp mass and the symmetric mass ratio, and say what each tells you.

Step 1. Total mass $M = 65\,M_\odot$ and product $m_1 m_2 = 1044\,M_\odot^2$ — the two combinations that build every mass observable. M = 65 M⊙.

Step 2. Chirp mass from equation (2): $\mathcal{M} = (1044)^{3/5}/(65)^{1/5}\,M_\odot$. With $(1044)^{0.6} = 64.0$ and $(65)^{0.2} = 2.29$, $\mathcal{M} = 64.0/2.29 = \mathbf{28.0\ M_\odot}$ (source frame) — the number the early inspiral pins down most tightly.

Step 3. Symmetric mass ratio $\eta = m_1 m_2/M^2 = 1044/4225 = \mathbf{0.247}$, just under the equal-mass ceiling $1/4$. So GW150914 was a nearly equal-mass merger — weak higher harmonics, dominant $\ell=m=2$ mode. The detector-frame chirp mass is larger, $\mathcal{M}^{\rm det} = (1+z)\mathcal{M} \approx 30.7\,M_\odot$ at $z\approx0.09$, the value the figures use. 𝓜 = 28.0 M⊙ (source), η = 0.247 ≈ 1/4.

2 · Time in band and cycle count: neutron stars versus black holes (fully worked)

Contrast a binary neutron star ($1.4+1.4\,M_\odot$) with GW150914 ($\mathcal{M}\approx28\,M_\odot$). For the BNS, find the chirp mass, the time to coalescence from 30 Hz, and the number of GW cycles between 30 and 1000 Hz. Then compare with the black-hole binary.

Step 1. BNS chirp mass: with $m_1 = m_2 = 1.4\,M_\odot$, $\mathcal{M} = \eta^{3/5}M = (1/4)^{3/5}\times2.8 = 0.435\times2.8 = \mathbf{1.22\ M_\odot}$. A tiny chirp mass — and every timescale below scales as $\mathcal{M}^{-5/3}$, so "tiny" means "long-lived."

Step 2. Time to coalescence from 30 Hz via $\tau = \tfrac{5}{256}(G\mathcal{M}/c^3)^{-5/3}(\pi f)^{-8/3}$. Insert $G\mathcal{M}/c^3 = 6.01\times10^{-6}\ \mathrm{s}$ and $f = 30\ \mathrm{Hz}$: $\tau = \mathbf{53.5\ s}$ — the better part of a minute sweeping up through the band.

Step 3. Cycle count from Derivation 2, $N_{\rm cyc} = \tfrac{1}{32}\pi^{-8/3}(G\mathcal{M}/c^3)^{-5/3}(f_{\rm low}^{-5/3} - f_{\rm high}^{-5/3})$ with $f_{\rm low}=30$, $f_{\rm high}=1000\ \mathrm{Hz}$. The low-frequency term dominates ($30^{-5/3} = 3.6\times10^{-3}$ vs $1000^{-5/3} = 1.0\times10^{-5}$), giving $N_{\rm cyc} \approx \mathbf{2.6\times10^{3}}$ cycles.

Step 4. For GW150914 ($\mathcal{M}\approx28\,M_\odot$) the chirp mass is $\sim23\times$ larger, so $\tau$ and $N_{\rm cyc}$ shrink by $23^{5/3}\approx180$: $N_{\rm cyc}(35\to250\ \mathrm{Hz})\approx\mathbf{10}$ cycles in $\approx0.16\ \mathrm{s}$. The BNS gives ~2600 cycles over ~50 s; the BBH gives ~10 cycles over ~0.2 s — hundreds of times more phase information for the lighter system, which is why BNS mergers are the equation-of-state and cosmology workhorses.

3 · The ringdown frequency of the GW150914 remnant

The merger left a remnant black hole of $M_f \approx 62\,M_\odot$ and dimensionless spin $a_f \approx 0.67$. Estimate the dominant ($\ell=m=2, n=0$) quasi-normal-mode frequency and damping time.

Step 1. The dimensionless mode frequency from perturbation-theory fits is $M_f\Omega_{220} \approx 1.5251 - 1.1568(1-a_f)^{0.1292}$. With $a_f = 0.67$, $(1-0.67)^{0.1292} = 0.865$, giving $M_f\Omega_{220} \approx 1.525 - 1.001 = 0.524$. A dimensionless frequency ≈ 0.53.

Step 2. Convert to physical frequency with equation (7): $f_{220} = \Omega_{220} c^3/(2\pi G M_f)$. The mass scale $c^3/(2\pi G M_f) = (2.998\times10^8)^3/[2\pi(6.674\times10^{-11})(62\times1.989\times10^{30})] = 5.2\times10^2\ \mathrm{Hz}$, so $f_{220} = 0.524\times520 \approx \mathbf{272\ Hz}$ — the horizontal red line in the second figure.

Step 3. The quality factor for this spin is $Q\approx3.2$, and $\tau_{220} = Q/(\pi f_{220}) = 3.2/(\pi\times272) \approx \mathbf{3.7\ ms}$. So the remnant rings for only a few cycles before it damps away — which is why ringdown tests of the no-hair theorem demand high signal-to-noise right at merger. f₂₂₀ ≈ 272 Hz, τ ≈ 3.7 ms.

Compact binary coalescence sits at the meeting point of a century of theory and a half-century of instrument building: a prediction of Einstein's 1916 field equations, made observable only in 2015. No single person owns it, but one figure organized the theoretical and experimental program that turned a waveform into an observatory.

Texts. Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §16–17 (general relativity, gravitational radiation, compact remnants) · Maggiore, Gravitational Waves, Vol. 1: Theory and Experiments (the definitive treatment of inspiral waveforms and matched filtering) · Ryden, Introduction to Cosmology (standard sirens and $H_0$) · Dodelson, Modern Cosmology (gravitational-wave sourcing) · Buonanno & Damour, Effective one-body approach to general relativistic two-body dynamics, Phys. Rev. D 59, 084006 (1999).

Provenance. A verifier should recompute the chirp-mass identity $\mu M^{2/3} = \mathcal{M}^{5/3}$; the coalescence time $\tau = \tfrac{5}{256}(G\mathcal{M}/c^3)^{-5/3}(\pi f)^{-8/3}$ by integrating $\dot f = \tfrac{96}{5}\pi^{8/3}(G\mathcal{M}/c^3)^{5/3}f^{11/3}$; and the cycle count $N_{\rm cyc}$. Worked numbers checked independently: GW150914 source-frame $\mathcal{M} = 28.0\,M_\odot$, $\eta = 0.247$ ($m_1=36$, $m_2=29\,M_\odot$); BNS $\mathcal{M} = 1.22\,M_\odot$, $\tau(30\ \mathrm{Hz}) = 53.5\ \mathrm{s}$, $N_{\rm cyc}(30\to1000\ \mathrm{Hz}) \approx 2.6\times10^3$; remnant $f_{220}\approx272\ \mathrm{Hz}$, $\tau_{220}\approx3.7\ \mathrm{ms}$ ($M_f=62\,M_\odot$, $a_f=0.67$). Event parameters are the published LIGO–Virgo values. Sources: Abbott et al., Phys. Rev. Lett. 116, 061102 (2016) and 119, 161101 (2017); Buonanno & Damour, Phys. Rev. D 59, 084006 (1999).

Builds on: Neutron Stars · Equation of State of Dense Matter · Tolman–Oppenheimer–Volkoff Equation Leads to: Neutron Star Structure and Interior · Magnetars.

Ask the tutor: Want to work through how a single chirp fixes the masses — or why lowering a detector's low-frequency wall multiplies the science? Ask and we will trace it step by step.