Gravitational Wave Extraction and Waveform Modeling

A binary black hole merger is the most violent event general relativity permits, yet what a detector on Earth records is a whisper: a dimensionless strain $h(t)$ of order $10^{-21}$ , a length change smaller than a proton across a four-kilometre arm. Between the two lies a chain of inference. A numerical-relativity code evolves the full nonlinear Einstein equations inside a finite box and outputs metric and curvature on a grid; far from the holes that data must become the two transverse-traceless polarizations $h_+$ and $h_\times$ a distant observer would measure — the problem of wave extraction. Two formalisms do the job: the Newman-Penrose Weyl scalar $\Psi_4$ , which by the peeling theorem falls off as $1/r$ and encodes purely outgoing radiation, and the gauge-invariant Regge-Wheeler-Zerilli master functions, which treat the far field as ripples on a Schwarzschild background split by parity. But extraction is only half the title. A single simulation costs weeks of supercomputer time and covers one point in a seven-dimensional space of mass ratio and two spins — far too slow for a matched-filter search. So the community distills numerical waveforms into fast template families — post-Newtonian inspiral, the effective-one-body resummation, phenomenological frequency-domain fits, and data-driven surrogates — each calibrated to reproduce the numerical truth to a fraction of a radian over $10^3$ to $10^5$ cycles. This article synthesizes both halves: how the asymptotic radiation is read off a strong-field simulation, and how it is compressed into the banks that let LIGO-Virgo-KAGRA hear a distant merger.

Key takeaways

Wave extraction turns a strong-field numerical-relativity spacetime into the strain a distant detector sees, through two channels: the Weyl scalar $\Psi_4$ (peeling, purely outgoing) and the Regge-Wheeler-Zerilli master functions (perturbed Schwarzschild, split by parity).
At large radius $\Psi_4=\ddot h_+ - i\,\ddot h_\times$ , so the complex strain $h=h_+ - i\,h_\times$ follows from two time integrations — done in the frequency domain to kill low-frequency drift.
The field expands in spin-weight $-2$ harmonics ${}_{-2}Y_{\ell m}$ ; the $(2,2)$ mode dominates while higher modes carry mass ratio, spin, precession and inclination.
Finite-radius spheres are gauge-biased; Cauchy-characteristic extraction or $1/r$ extrapolation pushes the answer to null infinity where news and strain are unambiguous.
Template families — PN, effective-one-body calibrated to numerical relativity, phenomenological, and surrogate — reproduce waveforms to mismatches below $10^{-3}$ and are the banks matched filtering runs on.
Key idea: The asymptotic outgoing radiation is a single complex field — $\Psi_4$ at large $r$ or the RWZ master functions — whose spin-weighted multipoles, extrapolated to null infinity, become the strain $h_+ - i\,h_\times$ that PN, EOB, phenom and surrogate templates must reproduce to fractions of a radian.

The core insight

The asymptotic outgoing radiation is encoded in one complex field — either $\Psi_4$ at large $r$ or the Regge-Wheeler-Zerilli master functions — whose spin-weight $-2$ multipoles, after extrapolation to future null infinity, become the strain $h_+ - i\,h_\times$ that template families (PN, EOB, phenom, surrogate) must reproduce to a fraction of a radian of phase.

Where this lives today

Extraction and modeling are the daily currency of gravitational-wave astronomy. High-accuracy catalogues such as SXS now run Cauchy-characteristic extraction as standard, delivering strain directly at null infinity — including the nonlinear memory effect that finite-radius methods miss. Every LIGO-Virgo-KAGRA parameter-estimation run leans on families that did not exist twenty years ago: SEOBNR effective-one-body models, the IMRPhenom family, and NR surrogates such as NRSur7dq4 that interpolate the full precessing parameter space to sub-percent accuracy, feeding searches over banks of order $10^5$ templates. The frontier is ringdown spectroscopy — resolving the $(2,2,1)$ overtone against the $(2,2,0)$ to test the no-hair theorem — and, ahead, LISA and the Einstein Telescope demanding waveforms accurate over $10^5$ cycles with memory, eccentricity and matter effects.

The physical picture

Picture the merger as a source buried at the centre of nested spheres. Close in, the spacetime is fully nonlinear and the gauge is whatever the evolution code chose; there is no clean notion of "the wave". Move outward and the field organizes itself: by peeling the Weyl scalar $\Psi_4$ decays as $1/r$ while non-radiative pieces fall faster, so at large radius $\Psi_4$ is pure outgoing radiation. A detector measures relative accelerations of free masses — a curvature component — so it is no accident that the extracted quantity is a piece of the Weyl tensor. The first figure fixes the geometry.

Figure

Wave extraction geometry: the fully nonlinear merger sits at the centre, the field is read on nested coordinate spheres, and by peeling Ψ₄ = Ψ₄⁰/r + O(1/r²) — extrapolating in 1/r, or evolving a null foliation with Cauchy-characteristic extraction, delivers the strain at future null infinity ℐ⁺

The Regge-Wheeler-Zerilli picture looks at the same far field differently: it treats the radiation as small perturbations of a fixed Schwarzschild background, split by parity into two scalar wave equations with effective potentials, and reads the late-time signal as the ringing of those potentials. This is the spectroscopic view — the post-merger ringdown is literally the quasinormal-mode spectrum of those operators, a decaying sum of complex frequencies fixed by the remnant's mass and spin alone. The whole coalescence plotted against time is the iconic chirp: a slow inspiral rising in frequency and amplitude, a violent merger, and a damped ringdown, on which the second figure puts real numbers.

Figure

The inspiral-merger-ringdown strain h(t): frequency and amplitude sweep up, peak at merger, then decay as the remnant rings down in its fundamental quasinormal mode — for a 60 M⊙ Schwarzschild remnant the (2,2,0) mode has f ≈ 201 Hz, damping time τ ≈ 3.3 ms, and quality factor Q ≈ 2.1

The core equations

Wave extraction begins with a curvature component numerical relativity computes unambiguously in the wave zone. The Newman-Penrose scalar $\Psi_4$ projects the Weyl tensor onto a null tetrad aligned with the outgoing radiation, and of the five complex Weyl scalars it is the one that, by peeling, carries the outgoing transverse-traceless wave at order $1/r$ .

\Psi_4 = C_{\alpha\beta\gamma\delta}\, n^\alpha \bar m^\beta n^\gamma \bar m^\delta \qquad (1)

where: $C_{\alpha\beta\gamma\delta}$ — Weyl conformal curvature tensor; $n^\alpha$ — outgoing-aligned null tetrad leg (sign convention varies between codes); $m^\alpha,\ \bar m^\alpha$ — complex null legs spanning the extraction 2-sphere, bar is complex conjugation; $\Psi_4$ — complex scalar encoding the two outgoing polarizations. The reading: with a good tetrad far from the source, $\Psi_4$ is the wave, no gauge or near-zone contamination surviving at leading order. The whole reason $\Psi_4$ is useful: in the wave zone it equals the second time derivative of the strain a detector feels — the bridge from curvature to observable, and why two integrations recover $h$ .

\Psi_4 = \ddot h_+ - i\,\ddot h_\times \equiv \ddot{\bar h}, \qquad h \equiv h_+ - i\,h_\times \qquad (2)

where: $h_+,\ h_\times$ — the two transverse-traceless polarizations in an orthonormal sphere frame; $h$ — complex strain combining them; dots are derivatives in retarded time. The reading: $\Psi_4$ is "the acceleration of the wave", so the strain is its double antiderivative — exact at null infinity, approximate at finite radius. Neither $\Psi_4$ nor $h$ is a single number on the sphere; each is a field over angles. Because gravitational radiation has spin weight $-2$ , the natural basis is the spin-weight $-2$ spherical harmonics, one complex time series per multipole.

\Psi_4(t,r,\theta,\phi)=\sum_{\ell=2}^{\infty}\sum_{m=-\ell}^{\ell}\Psi_4^{\ell m}(t,r)\,{}_{-2}Y_{\ell m}(\theta,\phi) \qquad (3)

where: ${}_{-2}Y_{\ell m}$ — spin-weight $-2$ spherical harmonic; $\Psi_4^{\ell m}$ — complex modal coefficient, one time series per mode; $\ell\geq 2$ — the sum starts at the quadrupole because monopole and dipole do not radiate. The reading: the $(2,2)$ mode dominates a quasi-circular inspiral, while higher and precession-induced modes carry mass ratio and spin. The two extraction channels meet in the strain. Once extrapolated to null infinity, the physical strain at luminosity distance $R$ and inclination $(\theta,\phi)$ is the same spin-weighted sum, amplitudes related to $\Psi_4$ by two integrations.

h = -\frac{1}{R}\sum_{\ell m} H_{\ell m}(t)\,{}_{-2}Y_{\ell m}(\theta,\phi), \qquad \ddot H_{\ell m} = R\,\Psi_4^{\ell m} \qquad (4)

where: $R$ — luminosity (areal) distance to the source; $H_{\ell m}$ — strain mode amplitude; the second relation is equation (2) written mode by mode. The reading: extraction delivers $\Psi_4^{\ell m}$ ; two integrations and a $1/R$ scaling give the strain multipoles a template must match. The Regge-Wheeler-Zerilli channel replaces curvature projection with black-hole perturbation theory. Metric perturbations of a Schwarzschild background separate by parity into two gauge-invariant master functions, each obeying a Schrodinger-like equation in the tortoise coordinate with an effective potential.

\frac{\mathrm{d}^2\Psi^{(\mathrm{e/o})}_{\ell m}}{\mathrm{d}r_*^2}+\left[\omega^2-V^{(\mathrm{e/o})}_\ell(r)\right]\Psi^{(\mathrm{e/o})}_{\ell m}=0 \qquad (5)

where: $r_*$ — tortoise coordinate, $\mathrm{d}r_*=(1-2M/r)^{-1}\mathrm{d}r$ ; $V^{(\mathrm o)}_\ell$ — Regge-Wheeler (odd/axial) potential; $V^{(\mathrm e)}_\ell$ — Zerilli (even/polar) potential; $\omega$ — complex frequency. The reading: with outgoing conditions at infinity and ingoing at the horizon, the allowed $\omega$ are the quasinormal frequencies — the ringdown spectrum. The odd-parity potential is worth writing explicitly: its two terms have clean physical meaning, and its peak near the light ring $r\approx 3M$ is what traps and slowly leaks the ringing.

V^{(\mathrm o)}_\ell(r)=\left(1-\frac{2M}{r}\right)\left[\frac{\ell(\ell+1)}{r^2}-\frac{6M}{r^3}\right] \qquad (6)

where: $M$ — background (remnant) mass; $\ell$ — multipole index; the $\ell(\ell+1)/r^2$ term is the centrifugal barrier and $-6M/r^3$ the spin-2 coupling to background curvature. The reading: a scalar field would see $-2M/r^3$ and electromagnetism $0$ , so the $-6$ is the fingerprint of the graviton's spin. The luminosity carried to infinity is a Bondi flux built from the news — a sum over modes of the squared time integral of $\Psi_4^{\ell m}$ .

\frac{\mathrm{d}E}{\mathrm{d}t}=\lim_{r\to\infty}\frac{r^2}{16\pi}\sum_{\ell m}\left|\int_{-\infty}^{t}\Psi_4^{\ell m}\,\mathrm{d}t'\right|^2 \qquad (7)

where: $\mathrm{d}E/\mathrm{d}t$ — instantaneous gravitational-wave luminosity; $\int\Psi_4\,\mathrm{d}t'$ — the news function, equal to $\dot{\bar h}$ ; geometrized units $G=c=1$ . The reading: integrate once to the news, square and sum for the power — for GW150914 this peaked near $3.6\times10^{49}$ W, about $10^{-3}$ of $c^5/G$ . Modeling the inspiral analytically starts from the effective-one-body map, which resums the post-Newtonian series by turning two-body dynamics into geodesic-like motion of a test mass in a deformed metric with radial potential $A(u)$ , $u=GM/(rc^2)$ , PN-expanded then Pade-resummed. The physical two-body energy is then recovered from that effective one-body energy by a nonlinear mass-shell map.

H_{\mathrm{real}}=M c^2\sqrt{1+2\eta\left(\frac{H_{\mathrm{eff}}}{\mu c^2}-1\right)}, \qquad H_{\mathrm{eff}}=\mu c^2\sqrt{A(u)\bigl[1+p_\phi^2 u^2 + \dots\bigr]} \qquad (8)

where: $H_{\mathrm{real}}$ — physical Hamiltonian of the binary; $M=m_1+m_2$ — total mass; $\mu=m_1m_2/M$ — reduced mass; $\eta=\mu/M$ — symmetric mass ratio ( $0<\eta\leq1/4$ ); $H_{\mathrm{eff}}$ — effective one-body Hamiltonian built from $A(u)$ and the momenta $p_\phi,p_{r_*}$ . The reading: as $\eta\to0$ the map gives $H_{\mathrm{real}}\to H_{\mathrm{eff}}$ , so EOB reduces to a test particle orbiting Schwarzschild — the anchor that lets it survive merger where the bare PN series diverges. Surrogates abandon dynamics entirely and interpolate the numerical data directly, expanding any waveform in a small empirical basis with parameter-dependent coefficients.

h_{\mathrm{sur}}(t;\boldsymbol\lambda)=\sum_{i=1}^{n} B_i(t)\, A_i(\boldsymbol\lambda) \qquad (9)

where: $B_i(t)$ — reduced (empirical interpolation) basis from training waveforms; $A_i(\boldsymbol\lambda)$ — fitted coefficients of the physical parameters $\boldsymbol\lambda$ (masses, spins); $n$ — number of basis elements, typically tens to a hundred. The reading: with a good basis, evaluation takes milliseconds and mismatches against held-out waveforms stay below $10^{-3}$ inside the training region.

Derivation ARecovering the strain from the Weyl scalar by fixed-frequency integration

The goal is to turn the exact wave-zone relation $\Psi_4=\ddot{\bar h}$ into a usable recipe for the strain, and to see why integrating $\Psi_4$ twice in time fails so badly that every modern catalogue avoids it. The strategy is to pass to the Fourier domain, where two derivatives become multiplication by $-\omega^2$ , invert with a low-frequency cutoff that tames the divergent $1/\omega^2$ , and transform back — removing the spurious drifts of time-domain integration while leaving the physical signal untouched.

Step 1. Write the exact mode relation from equation (4), $\Psi_4^{\ell m}(t)=\ddot H_{\ell m}(t)/R$ , so recovering $H_{\ell m}$ is a double antidifferentiation. Why: each multipole is an independent scalar time series, so the vector problem decouples into identical scalar problems, one per mode.

Step 2. Fourier transform. Since a time derivative becomes multiplication by $i\omega$ , two derivatives give $\tilde\Psi_4(\omega)=-\omega^2\,\tilde h(\omega)$ , hence formally $\tilde h(\omega)=-\tilde\Psi_4(\omega)/\omega^2$ . Why: differentiation is diagonal in frequency, so the integration that is delicate in time is a single algebraic division in frequency.

Step 3. Regularize the division. As $\omega\to 0$ the factor $1/\omega^2$ blows up, amplifying low-frequency noise into a spurious linear-plus-quadratic ramp. Replace $\omega$ by $\mathrm{sgn}(\omega)\,\omega_0$ whenever $|\omega|<\omega_0$ , with $\omega_0$ just below the lowest physical mode frequency (for $(2,2)$ , a little under $2\Omega_{\mathrm{orb}}$ ). Why: no real signal power lives below $\omega_0$ , so flattening the divergent kernel there discards only noise and integration constants, not physics.

Step 4. Assemble and invert. The boxed relation $\tilde h_{\ell m}(\omega)=-\tilde\Psi_{4,\ell m}(\omega)/\max(|\omega|,\omega_0)^2$ is inverse-transformed to a clean $h_{\ell m}(t)$ . Why: the inverse FFT rebuilds the strain with the drift already excised — why fixed-frequency integration is the community default. *Verified: for a monochromatic test mode this reproduces the analytic $h=-\Psi_4/\omega^2$ exactly, and it is the method of Reisswig and Pollney (2011) used across modern waveform catalogues.*

Go deeper · Derivation B · 1 minThe gravitational-wave luminosity from the quadrupole formula

Derivation BThe gravitational-wave luminosity from the quadrupole formula

The goal is to derive the leading-order power radiated by a slowly moving source — the famous $G/5c^5$ luminosity — from the quadrupole strain and the effective stress-energy of a weak wave, and to see why this single prefactor sets the inspiral chirp rate every detection template is anchored to. The strategy is to feed the quadrupole strain into the Isaacson energy flux, cancel the strain's $1/r^2$ against the flux's $r^2$ , and do the angular integral of the transverse-traceless projector.

Step 1. Start from the Einstein quadrupole strain $h_{ij}^{TT}=(2G/c^4 r)\,\ddot Q_{ij}^{TT}(t-r/c)$ and the Isaacson flux $\mathrm{d}E/\mathrm{d}t=(c^3 r^2/32\pi G)\oint\langle \dot h_{ij}^{TT}\dot h^{ij}_{TT}\rangle\,\mathrm{d}\Omega$ . Why: the Isaacson flux is the gauge-invariant energy of a high-frequency wave, averaged over several wavelengths.

Step 2. Insert the strain. Then $\dot h_{ij}^{TT}=(2G/c^4 r)\,\dddot Q_{ij}^{TT}$ , and the flux's $r^2$ cancels the $1/r^2$ in $|\dot h|^2$ , so the answer is finite as $r\to\infty$ . Why: luminosity cannot depend on the arbitrary sphere radius, and the cancellation checks that falloff and flux normalization are consistent.

Step 3. Do the angular integral of the transverse-traceless projector, $\oint \dddot Q^{TT}_{ij}\dddot Q_{TT}^{ij}\,\mathrm{d}\Omega=\tfrac{4\pi}{5}\dddot Q_{ij}\dddot Q^{ij}$ , introducing the factor $1/5$ . Why: projecting the two physical polarizations and averaging over directions turns the tensor into one scalar rate; the $4\pi/5$ is pure sphere geometry.

Step 4. Collect constants to reach $\mathrm{d}E/\mathrm{d}t=(G/5c^5)\langle \dddot Q_{ij}\dddot Q^{ij}\rangle$ . Why: this prefactor fixes how fast an inspiral loses energy and sweeps up in frequency — exactly the chirp the leading post-Newtonian phase encodes. *Verified: applied to PSR B1913+16 the companion orbital-decay formula gives $\dot P_b=-2.40\times10^{-12}$ s/s, matching the observed Hulse-Taylor decay to about $0.2\%$ .*

Limiting cases

Different regimes call for different tools; the right column names what each is physically saying.

Case	Result	Meaning
Weak-field, slow-motion inspiral	$h_{ij}^{TT}=\dfrac{2G}{c^4 r}\,\ddot{Q}_{ij}^{TT}(t-r/c)$	Einstein quadrupole formula; the seed of post-Newtonian templates and the leading chirp.
Perturbed Schwarzschild (ringdown)	$\dfrac{\mathrm{d}^2\Psi}{\mathrm{d}r_*^2}+(\omega^2-V_\ell)\Psi=0$	RWZ master equation; complex $\omega$ are quasinormal modes that dominate the post-merger signal.
Fully nonlinear merger	$\Psi_4 = C_{\alpha\beta\gamma\delta}\,n^\alpha \bar m^\beta n^\gamma \bar m^\delta$	Newman-Penrose extraction from the full Weyl tensor; the only faithful capture of the strong-field peak.
Asymptotic limit / null infinity	$\Psi_4 \sim \dfrac{\Psi_4^0}{r}+O(r^{-2})$	Peeling: only the $1/r$ piece is radiation; CCE delivers it directly at $\mathscr{I}^+$ , free of finite-radius bias.
Test-mass limit $\eta\to 0$	$H_{\mathrm{real}}\to H_{\mathrm{eff}}$	EOB reduces exactly to a particle orbiting Schwarzschild — the anchor of the resummation.

Worked examples

1 · The ringdown as black-hole spectroscopy (fully worked)

A binary black hole merges into a non-spinning remnant of mass $M=60\,M_\odot$ . Using the fundamental $\ell=m=2$ quasinormal mode of Schwarzschild, $M\omega_{220}=0.3737-0.0890\,i$ in geometric units, find the ringdown frequency, damping time, and quality factor a detector would see.

Step 1. Convert the remnant mass to a time. One solar mass in geometric units is $GM_\odot/c^3=4.925\times10^{-6}$ s, so $M=60\,M_\odot$ gives $\tau_M = 60\times4.925\times10^{-6}=\mathbf{2.955\times10^{-4}\ s}$ — the single scale, since the dimensionless eigenvalue $M\omega$ is the same for every Schwarzschild hole.

Step 2. Real frequency. $\omega_R=\mathrm{Re}(M\omega)/\tau_M=0.3737/2.955\times10^{-4}=1264\ \mathrm{rad/s}$ , so $f_{220}=\omega_R/2\pi=\mathbf{201\ Hz}$ — squarely in the LIGO band.

Step 3. Damping time. The imaginary part sets the decay: $\tau=\tau_M/|\mathrm{Im}(M\omega)|=2.955\times10^{-4}/0.0890=\mathbf{3.32\ ms}$ , so only a handful of cycles survive.

Step 4. Quality factor. $Q=\omega_R/(2|\omega_I|)=0.3737/(2\times0.0890)=\mathbf{2.10}$ — a heavily damped bell, which is why extracting overtones to test the no-hair theorem is so demanding. For a $60\,M_\odot$ Schwarzschild remnant: $f_{220}\approx201$ Hz, $\tau\approx3.32$ ms, $Q\approx2.10$ . A real remnant spins, shifting $f$ up and $\tau$ longer; GW150914's remnant ( $\sim62\,M_\odot$ , spin $\sim0.68$ ) rang near $250$ Hz.

2 · The Hulse-Taylor chirp rate grounds the quadrupole prefactor

The binary pulsar PSR B1913+16 has orbital period $P_b=7.75$ hr, eccentricity $e=0.617$ , and near-equal masses $m_1\approx m_2\approx1.4\,M_\odot$ . Show that the quadrupole luminosity predicts an orbital decay consistent with the measured $\dot P_b$ .

Step 1. The orbit-averaged decay from the quadrupole formula (Peters 1964) is $\dot P_b=-\dfrac{192\pi}{5}\Bigl(\dfrac{2\pi G}{c^3 P_b}\Bigr)^{5/3}\dfrac{m_1 m_2}{(m_1+m_2)^{1/3}}\,F(e)$ , with the eccentricity enhancement $F(e)=(1+\tfrac{73}{24}e^2+\tfrac{37}{96}e^4)(1-e^2)^{-7/2}$ . The $192\pi/5$ carries the same $1/5$ that appeared in the luminosity derivation.

Step 2. Evaluate the enhancement. With $e=0.617$ , $F(e)=\mathbf{11.85}$ — the eccentric orbit radiates almost twelve times faster than a circular one of the same period, the bursts near periastron dominating.

Step 3. Assemble the numbers ( $P_b=2.79\times10^4$ s, $m_1=1.44\,M_\odot$ , $m_2=1.39\,M_\odot$ ). The formula gives $\dot P_b=\mathbf{-2.40\times10^{-12}\ s/s}$ ; the measured intrinsic decay is $-2.398\times10^{-12}$ s/s, so observation over prediction is $\mathbf{0.997}$ — quadrupole radiation confirmed to a fraction of a percent, and the first evidence that gravitational waves carry energy. The same $G/5c^5$ prefactor that sets a template's chirp passed its first test in a radio pulsar.

Physics of the portrait

Wave extraction and waveform modeling are a seventy-year relay — from perturbation theorists who never imagined a detector to the numerical relativists and modelers who made a merger legible.

Origins · 1 mingravitational wave extraction and how this was discovered

gravitational wave extraction

the bridge from Ψ₄ at large r to the strain h at null infinity · black-hole perturbation theory, numerical relativity, and gravitational-wave astronomy

The idea that a black hole has a characteristic sound predates any hope of hearing it. In 1957 Tullio Regge and John Wheeler asked whether a Schwarzschild black hole is stable and, in answering, derived the odd-parity master equation that still bears their name; Frank Zerilli completed the even-parity half in 1970, and the two utterly different potentials proved to share an identical quasinormal spectrum (isospectrality). Meanwhile in 1962 Ezra Newman and Roger Penrose built the null-tetrad formalism whose scalar $\Psi_4$ became the workhorse of numerical extraction, and Bondi defined the news function at null infinity that makes radiated energy unambiguous. For decades these were tools without a source to point at. That changed when numerical relativity finally evolved a binary black hole through merger — Frans Pretorius in 2005, then the moving-puncture breakthroughs — producing the first strong-field waveforms to extract, while Alessandra Buonanno and Thibault Damour's 1999 effective-one-body resummation gave analytic dynamics a chance of surviving to merger. By 2015 the LIGO-Virgo collaboration had banks, pipelines and null-infinity waveforms ready, and GW150914's chirp matched a numerical-relativity-calibrated model to a fraction of a radian — the relay that began as a stability calculation ending as an observatory.

History

1957

— Regge and Wheeler derive the odd-parity master equation for a perturbed Schwarzschild black hole

1962

— Newman and Penrose introduce the null-tetrad Weyl scalar Ψ₄, and Bondi defines the news at null infinity

1970

— Zerilli completes the even-parity master function; the two parities prove isospectral

1999

— Buonanno and Damour formulate the effective-one-body resummation of the two-body dynamics

2005

— Pretorius achieves the first stable binary-black-hole merger simulation, giving numerical waveforms to extract

Going further

Texts. Misner, Thorne and Wheeler, Gravitation, ch. 35-36 (linearized waves, quadrupole formula) · Wald, General Relativity, ch. 4 and 7 (linearized gravity, radiation) · Carroll, Spacetime and Geometry, ch. 7 (gravitational radiation) · Schutz, A First Course in General Relativity, ch. 9 · Hartle, Gravity, ch. 16 · Regge and Wheeler, Phys. Rev. 108, 1063 (1957) · Zerilli, Phys. Rev. D 2, 2141 (1970) · Newman and Penrose, J. Math. Phys. 3, 566 (1962) · Buonanno and Damour, Phys. Rev. D 59, 084006 (1999).

Provenance. Equations cross-checked against MTW ch. 35-36 ( $G/5c^5$ prefactor), Wald ch. 4/7 and Carroll ch. 7 (linearized radiation, TT gauge), and the original Regge-Wheeler (1957) and Zerilli (1970) papers. A verifier should recompute: the $G/5c^5$ prefactor; the potential terms $\ell(\ell+1)/r^2-6M/r^3$ ; the tetrad relation $\Psi_4=\ddot h_+-i\ddot h_\times$ ; the Schwarzschild $(2,2,0)$ values $f\approx201$ Hz, $\tau\approx3.32$ ms, $Q\approx2.10$ at $60\,M_\odot$ (eigenvalue from Berti-Cardoso-Will); and the Hulse-Taylor decay $\dot P_b\approx-2.40\times10^{-12}$ s/s, ratio $\approx0.997$ . Peak GW150914 luminosity $\approx3.6\times10^{49}$ W $\approx10^{-3}c^5/G$ ; EOB follows Buonanno-Damour (1999).

Prerequisites. ADM 3+1 Decomposition of Spacetime · BSSN Formulation · Numerical Relativity and Binary Black Hole Mergers · Gauge Conditions in Numerical Relativity. Leads to. Black Hole Excision and Puncture Methods · Harmonic and Generalized Harmonic Gauge.

Ask the tutor: Ask me to walk through recovering a strain multipole from a raw $\Psi_4$ time series, or to compute the ringdown frequency and damping of any remnant mass and spin.

Why extract a curvature scalar Ψ₄ instead of just reading off the metric perturbation?

Because the metric on a finite-radius sphere is drenched in gauge. The evolution code chooses coordinates for numerical convenience, leaving imprints in $h_{\mu\nu}$ that look like waves but are not. The Weyl scalar $\Psi_4$ is a curvature component projected onto a physically chosen null tetrad, so at leading order in $1/r$ it is gauge-invariant and purely outgoing: it isolates the true radiative degrees of freedom from the coordinate noise. That is why $\Psi_4$ , not the raw metric, is the workhorse of extraction, the Regge-Wheeler-Zerilli master functions playing the same role in the perturbative channel.

What is future null infinity, and why do we extrapolate the waveform there?

Future null infinity, written $\mathscr{I}^+$ , is the idealized destination of outgoing light rays. It is the only place where "the gravitational wave" is unambiguous: the news and strain are defined there without gauge or near-zone contamination, and the asymptotic BMS symmetry makes energy and momentum flux well posed. A real simulation lives in a finite box, so we either extract on several spheres and extrapolate the modes in powers of $1/r$ , or evolve the Einstein equations along outgoing null cones (CCE) — both removing the offsets that plague any single finite-radius sphere.

Why does integrating Ψ₄ twice in time produce garbage, and what fixes it?

The relation $\Psi_4=\ddot{\bar h}$ says the strain is a double antiderivative, defined only up to constants of integration. In practice those constants — plus slow gauge transients and low-frequency noise — get amplified into a spurious drift growing linearly and quadratically in time, burying the low-amplitude ringdown tail. The cure is to integrate in the Fourier domain, where two derivatives become multiplication by $-\omega^2$ , and replace the divergent $1/\omega^2$ kernel below a cutoff $\omega_0$ with a flat value — discarding only the sub-signal frequencies where no real power lives.

What is the difference between EOB, phenomenological, and surrogate waveform models?

They sit on a ladder of decreasing first-principles content and increasing speed. Effective-one-body models resum the analytic post-Newtonian dynamics onto a deformed test-particle problem and attach a numerical-relativity-tuned merger and ringdown, carrying real Hamiltonian structure and a few calibration parameters. Phenomenological models fit closed-form frequency-domain amplitude and phase for speed. Surrogates make no dynamical assumptions: they interpolate the numerical data directly using a reduced basis and empirical time nodes, with the highest fidelity inside their training region but no guarantee outside it.

Why do the Regge-Wheeler and Zerilli potentials give the same ringdown frequencies?

The two potentials describe perturbations of opposite parity — Regge-Wheeler the axial (odd) modes, Zerilli the polar (even) — and look completely different as functions of radius, yet produce an identical quasinormal spectrum. This isospectrality is no coincidence: the operators are related by a Darboux transformation that preserves the spectrum while changing the potential's shape. Physically, a Schwarzschild black hole rings at the same complex frequencies whatever the parity of what perturbs it, so the ringdown frequencies are a clean function of mass alone (mass and spin for Kerr) — the foundation of the no-hair test.

What extra information do the higher modes beyond (2,2) carry?

The $(2,2)$ mode dominates a quasi-circular binary and encodes the overall chirp, but it is nearly degenerate in some parameters. Higher modes break those degeneracies: the relative amplitudes of $(2,1)$ , $(3,3)$ and $(4,4)$ grow with mass asymmetry, so they measure the mass ratio, and their phasing helps break the distance-inclination degeneracy a single mode cannot resolve. When spins are misaligned with the orbital angular momentum the plane precesses and pumps power between $m$ -modes, directly probing spin-orbit misalignment and hence how the binary formed — which is why precessing multi-mode surrogates are essential.