Post-Newtonian Waveform Modeling

A ground-based interferometer does not photograph a black-hole binary; it counts wave crests. Over the minutes a low-mass inspiral spends in the sensitive band, tens of thousands of gravitational-wave cycles sweep past, and the detector's entire sensitivity comes from correlating that measured phase against a theoretical template. Miss the rate of phase accumulation by even a fraction of a radian summed over the band and the matched filter loses lock — the signal-to-noise collapses and the inferred masses skew. Post-Newtonian (PN) theory is the discipline built to supply that phase to the demanded precision. It takes the leading quadrupole formula, which is only the first term of a series in $v/c$ , and computes the systematic relativistic corrections that track the waveform through the long, slow, well-separated regime where the two bodies still orbit at a modest fraction of light speed. The organizing idea is an expansion in the single small parameter $\epsilon \sim (v/c)^2 \sim GM/(rc^2)$ — velocity-squared and gravitational potential are the same order for a bound orbit, by the virial theorem. One expands the metric, the equations of motion, the conserved orbital energy $E(v)$ , and the radiated flux $\mathcal{F}(v)$ each as a power series in $\epsilon$ , labels the terms by PN order (0PN, 1PN, 1.5PN, …), and then closes the system with energy balance: the orbit loses exactly the energy the waves carry off, $\mathrm{d}E/\mathrm{d}t = -\mathcal{F}$ . Integrating that balance yields the chirp — the upward sweep of frequency — and hence the phase. The half-integer orders are where general relativity shows its nonlinear teeth: radiation reaction enters at 2.5PN, and at 1.5PN the waves scatter off the curvature the binary itself creates, a hereditary "tail" that depends on the source's entire past.

Key takeaways

Post-Newtonian theory expands the dynamics and radiation of a slow, well-separated compact binary in powers of $\epsilon \sim v^2/c^2 \sim GM/(rc^2)$ , systematically extending the leading quadrupole formula to the accuracy detectors demand.
A correction of relative order $(v/c)^{2n}$ beyond leading is labelled $n$ PN; the conservative dynamics are known to 4PN and the gravitational-wave phase to 3.5–4.5PN, with spin and tidal terms to comparable order.
The observable inspiral phase follows from energy balance $\mathrm{d}E/\mathrm{d}t = -\mathcal{F}$ : given the PN series for orbital energy $E(x)$ and flux $\mathcal{F}(x)$ , one integrates the chirp $\mathrm{d}x/\mathrm{d}t$ and accumulates the phase.
Half-integer orders are physical, not bookkeeping: 2.5PN is the leading radiation-reaction force, and 1.5PN is the first gravitational-wave tail — radiation backscattering off the binary's own spacetime curvature, a hereditary effect non-local in time.
The leading phase depends on the masses only through the chirp mass $\mathcal{M} = (m_1 m_2)^{3/5}(m_1+m_2)^{-1/5}$ , which is therefore the most precisely measured parameter of a low-mass inspiral.
PN waveforms govern the low-frequency, high-cycle-count inspiral; near merger ( $v \to c$ ) they are resummed by the effective-one-body method or matched to numerical relativity.
Key idea: expand everything — metric, equations of motion, energy $E(v)$ , and flux $\mathcal{F}(v)$ — in powers of $\epsilon \sim (v/c)^2$ , then impose energy balance $\mathrm{d}E/\mathrm{d}t = -\mathcal{F}$ to integrate the slow inspiral and pin the gravitational-wave phase to fractional-radian accuracy over tens of thousands of cycles.

The core insight

Treat $v/c$ as small and expand everything — metric, equations of motion, conserved energy $E(v)$ , and radiated flux $\mathcal{F}(v)$ — in powers of $\epsilon \sim (v/c)^2$ ; then impose energy balance $\mathrm{d}E/\mathrm{d}t = -\mathcal{F}$ to integrate the slow inspiral and accumulate the gravitational-wave phase to fractional-radian accuracy over tens of thousands of cycles.

Where this lives today

Post-Newtonian phasing is the analytic backbone of gravitational-wave astronomy as it is actually practised. Every matched-filter search runs the observed data against banks of templates whose low-frequency phase is the SPA series of equation (7), carried to 3.5PN in the non-spinning part and further for spins and tides; the fitted coefficients are literally how the chirp mass, mass ratio, and spins are read out of an event. For the binary neutron star GW170817 the chirp mass was pinned to roughly four significant figures — a precision that flows directly from the sixteen-thousand-cycle lever arm of Example 1. The PN coefficients are also the arena for testing general relativity itself: parameterized frameworks such as TIGER and the flexible-theory-agnostic (FTA) analysis let each phase coefficient float away from its GR value and bound the deviation, with the 1.5PN tail coefficient a prime target because it probes nonlinear, hereditary structure. And where PN theory finally runs out near merger, it does not vanish — it seeds the effective-one-body Hamiltonian and anchors the low-frequency end of the calibrated IMRPhenom and SEOBNR waveform families that dominate LIGO–Virgo–KAGRA analysis, and the same series is now being extended to the millihertz sources LISA will observe.

The physical picture

Picture the quadrupole formula as the first term of a Taylor series in $v/c$ . A binary at $v/c \approx 0.1$ has $\epsilon \approx 0.01$ , so each PN order is a percent-level correction to the one before — negligible in a single orbit, but the detector integrates the phase over $\sim 10^4$ cycles, and a one-percent error in the phasing rate compounds into many radians of mismatch. PN theory is precisely the craft of computing those small corrections to the precision the interferometer can resolve. The binary spirals inward because it bleeds energy and angular momentum into the waves: the orbit shrinks, the frequency climbs, and the waveform becomes an amplitude- and frequency-modulated sinusoid — the "chirp." The first figure shows that chirp for a canonical neutron-star binary and pins the real numbers to it.

Figure

The inspiral chirp of a 1.4 + 1.4 M⊙ binary neutron star. As energy radiates away the orbit shrinks and the gravitational-wave frequency sweeps upward from about 10 Hz to roughly 1.5 kHz at the innermost stable circular orbit, where v/c climbs from 0.08 to 0.41 — some 16000 wave cycles accumulate in the detector band over about 1000 s

The single most important conceptual novelty beyond the quadrupole picture is the tail. Waves emitted at one instant do not stream freely to infinity on flat spacetime; they scatter off the static curvature sourced by the binary's own total mass-energy, so the field a distant detector sees depends on the source's entire past. This makes the flux non-local in time and stamps the phase with characteristic $\pi$ -dependent (logarithmic and odd-power) terms that no flat-space radiation theory can produce. Because each PN order costs enormous analytic labour, the field pushes to high order only where it pays: the phase, being the measured quantity, has been driven furthest. The second figure is the payoff a data analyst actually cares about — the "phase budget," the number of gravitational-wave cycles each PN order contributes for a neutron-star binary, which is what dictates how many orders must be kept.

Figure

The post-Newtonian phase budget for a 1.4 + 1.4 M⊙ inspiral from 10 Hz to the ISCO. Each order contributes a measured number of accumulated gravitational-wave cycles — 0PN ≈ 16031, 1PN ≈ 439, the 1.5PN tail ≈ 211, 2PN ≈ 9.9, 2.5PN ≈ 12, 3PN ≈ 2.6 — bar length scales as log₁₀ of cycles, so terms out to 3.5PN are needed to hold the phase to under one cycle

Read the budget carefully: the 1.5PN tail is worth about 211 cycles — twenty times the entire 2PN instantaneous contribution. A purely nonlinear, hereditary effect is one of the loudest voices in the phasing, which is why it doubles as one of the cleanest tests of nonlinear gravity in the data.

The core equations

The starting point is the expansion parameter. For a bound orbit the kinetic energy $\tfrac12\mu v^2$ balances the potential $GM\mu/r$ , so orbital velocity-squared and gravitational potential are the same small quantity, and a single $\epsilon$ organizes the whole scheme.

\epsilon \;\sim\; \frac{v^2}{c^2} \;\sim\; \frac{GM}{r\,c^2} \qquad (1)

where: $v$ — characteristic orbital velocity; $M=m_1+m_2$ — total mass of the binary; $r$ — orbital separation; $G,\,c$ — Newton's constant and the speed of light. A term of relative order $\epsilon^n=(v/c)^{2n}$ beyond the leading behaviour is by convention called " $n$ PN." The engine that turns the two PN series into an observable is energy balance. In the adiabatic approximation the orbit is a sequence of circular orbits, each losing to the waves exactly the energy the flux carries to infinity.

\frac{\mathrm{d}E}{\mathrm{d}t} \;=\; -\,\mathcal{F} \qquad (2)

where: $E$ — conserved orbital binding energy as a function of orbital frequency; $\mathcal{F}$ — gravitational-wave luminosity at infinity; $t$ — coordinate time. Everything conservative lives in $E$ ; everything dissipative lives in $\mathcal{F}$ ; balance couples them. The conservative side is the PN orbital energy for a circular orbit, expressed in the dimensionless frequency variable $x$ . The leading term is the Newtonian $-\tfrac12\mu v^2$ ; the 1PN correction is the first relativistic bracket.

E(x) = -\frac{1}{2}\mu c^2 x\Big[1 - \tfrac{1}{12}(9+\nu)\,x + \dots\Big],\qquad x \equiv \Big(\frac{GM\omega}{c^3}\Big)^{2/3} \qquad (3)

where: $x$ — dimensionless PN frequency parameter, $x \sim (v/c)^2$ ; $\omega$ — orbital angular frequency; $\nu=\mu/M$ — symmetric mass ratio, $0<\nu\le \tfrac14$ ; $\mu=m_1m_2/M$ — reduced mass. Equal masses give $\nu=\tfrac14$ ; the test-mass limit is $\nu\to0$ . The dissipative side is the PN flux. Its leading $x^5$ scaling is the quadrupole formula; the 1PN bracket and, crucially, the $4\pi\,x^{3/2}$ term — the first tail — sit on top.

\mathcal{F}(x) = \frac{32}{5}\frac{c^5}{G}\,\nu^2 x^5\Big[1 + \Big(-\tfrac{1247}{336}-\tfrac{35}{12}\nu\Big)x + 4\pi\,x^{3/2} + \dots\Big] \qquad (4)

where: $\tfrac{32}{5}\tfrac{c^5}{G}\nu^2 x^5$ — leading (quadrupole) flux; the $x$ term is 1PN; the $x^{3/2}$ term is 1.5PN, whose coefficient $4\pi$ is the leading hereditary tail. The prefactor $c^5/G \approx 3.6\times10^{52}\ \mathrm{W}$ is the natural luminosity scale of general relativity. Balance now fixes the chirp: dividing the flux by the energy's slope gives the secular rate at which the frequency variable climbs.

\frac{\mathrm{d}x}{\mathrm{d}t} = -\frac{\mathcal{F}(x)}{\mathrm{d}E/\mathrm{d}x} \qquad (5)

where: $\mathrm{d}x/\mathrm{d}t$ — secular rate of frequency increase (the chirp). Because $\mathcal{F}>0$ and $\mathrm{d}E/\mathrm{d}x<0$ , the right side is positive: the frequency always rises as the orbit decays. The phase is then just the time integral of the orbital frequency, and the gravitational-wave phase at the dominant harmonic is twice it.

\Phi(t) = \int \omega\,\mathrm{d}t,\qquad \omega = \frac{c^3}{GM}\,x^{3/2},\qquad \phi_{\rm GW}=2\Phi \qquad (6)

where: $\Phi$ — accumulated orbital phase; the factor of 2 is because the mass quadrupole of a binary rotates at twice the orbital rate, so the wave frequency is $2f_{\rm orb}$ . For matched filtering one wants the phase in the frequency domain, obtained by the stationary-phase approximation (SPA). This closed form, the actual template, exposes the PN coefficients as the numbers fitted to the data.

\Psi(f) = 2\pi f t_c - \phi_c - \frac{\pi}{4} + \frac{3}{128\,\nu\,v^5}\Big[1 + \Big(\tfrac{3715}{756}+\tfrac{55}{9}\nu\Big)v^2 - 16\pi\,v^3 + \dots\Big] \qquad (7)

where: $v=(\pi GM f/c^3)^{1/3}$ — PN velocity parameter at gravitational-wave frequency $f$ ; $t_c,\,\phi_c$ — coalescence time and phase (integration constants); the $v^2$ term is 1PN and the $-16\pi\,v^3$ term is the 1.5PN tail. Each PN coefficient is a separate testable number. Finally, the leading term of that phase depends on the two masses only through one combination — the chirp mass — which is therefore the tightest-measured parameter of a low-mass inspiral.

\mathcal{M} = (m_1 m_2)^{3/5}\,(m_1+m_2)^{-1/5} = \nu^{3/5}M \qquad (8)

where: $\mathcal{M}$ — chirp mass. For an equal-mass binary $\mathcal{M}=(\tfrac14)^{3/5}M \approx 0.435\,M$ ; the mass ratio and spins enter only at higher PN order and so are measured far less precisely.

Derivation AThe Newtonian chirp and the leading stationary-phase phase

The goal is to combine energy balance with the bare quadrupole flux and follow the algebra all the way to the leading frequency-domain phase coefficient $3/128$ , showing along the way how the two individual masses collapse into the single chirp-mass combination. We keep only the leading term of each PN series, treat the orbit as quasi-circular, integrate the chirp to find how frequency runs to the coalescence time, and finally pass to the frequency domain by stationary phase — the exact route a verifier would retrace to confirm the template's leading term.

Step 1. Take the leading pieces of equations (3) and (4): $E=-\tfrac12\mu c^2 x$ and $\mathcal{F}=\tfrac{32}{5}\tfrac{c^5}{G}\nu^2 x^5$ , with $\mu=\nu M$ . Why: at Newtonian order the binding energy is purely kinetic-plus-potential and the flux is purely quadrupolar, so no relativistic brackets survive and the derivation stays transparent.

Step 2. Form the chirp from equation (5). Since $\mathrm{d}E/\mathrm{d}x = -\tfrac12\mu c^2=-\tfrac12\nu M c^2$ , dividing gives $\mathrm{d}x/\mathrm{d}t = \tfrac{64}{5}\tfrac{c^3}{GM}\nu\,x^5$ . Why: this is the classic $x^5$ chirp law — the frequency's climb accelerates steeply as $x$ grows, which is why the waveform sweeps ever faster toward merger.

Step 3. Integrate the chirp. Separating variables, $\int x^{-5}\,\mathrm{d}x \propto t$ , so $x^{-4}$ is linear in $(t_c-t)$ and $x(t)\propto (t_c-t)^{-1/4}$ , with the integration constant defining the coalescence time $t_c$ at which $x\to\infty$ . Why: the divergence at $t_c$ is the formal "merger" of the point-particle model, and identifying $t_c$ is what lets the whole waveform be written as a function of time-to-coalescence.

Step 4. Get the phase. Insert $\omega=(c^3/GM)x^{3/2}$ into $\Phi=\int\omega\,\mathrm{d}t$ ; using the chirp law the integral gives $\Phi \propto x^{-5/2}\propto (t_c-t)^{5/8}$ , a phase that scales as a single power of the frequency variable and carries the mass dependence as $\nu M^{5/3}=\mathcal{M}^{5/3}$ . Why: here the two masses fuse — only the combination $\mathcal{M}=\nu^{3/5}M$ appears, which is the algebraic origin of the chirp mass as the leading observable.

Step 5. Pass to the frequency domain. The Fourier transform of the slowly chirping sinusoid is dominated by the stationary point of its phase, giving $\Psi(f)=2\pi f t_c-\phi_c-\pi/4-\phi_{\rm GW}(f)$ ; substituting the Newtonian $\phi_{\rm GW}(f)$ yields $\Psi_{\rm 0PN}(f)=2\pi f t_c-\phi_c-\tfrac{\pi}{4}+\tfrac{3}{128}\big(\pi G\mathcal{M}f/c^3\big)^{-5/3}$ . Why: the numerical coefficient $3/128$ and the exponent $-5/3$ are exactly the leading template term LIGO fits, and the masses enter only through $\mathcal{M}$ — the calculation delivers both the number and the physical lesson at once. *Verified: the coefficient $3/128$ and exponent $-5/3$ reproduce Maggiore Vol. 1 Eq. (5.275) and Blanchet (2014); recompute by integrating $\mathrm{d}x/\mathrm{d}t=(64/5)(c^3/GM)\nu\,x^5$ and applying the SPA, and confirm the mass grouping $\mathcal{M}=\nu^{3/5}M$ .*

Go deeper · Derivation B · 2 minWhy gravitational-wave tails first appear at one-and-a-half post-Newtonian order

Derivation BWhy gravitational-wave tails first appear at one-and-a-half post-Newtonian order

The aim is to show heuristically that the leading hereditary "tail" enters the flux at relative order $x^{3/2}$ with coefficient $4\pi$ — a genuinely nonlinear, non-local-in-time effect absent from any flat-space radiation theory. We treat the outgoing quadrupole radiation as a wave trying to propagate not on flat space but on the weakly curved background produced by the binary's own total mass, track how backscattering builds a convolution over the source's past, and count the powers of the small parameter that fix its PN order.

Step 1. Recognize that the wave equation is not homogeneous once $M\ne0$ : the radiation obeys $\Box h \ne 0$ because it must climb out of the Schwarzschild-like potential $\sim GM/(rc^2)$ of the total mass. Why: on curved space part of an outgoing wave is reflected back in, so the field at infinity is not simply the retarded quadrupole — it carries an echo of the curvature.

Step 2. Write the leading correction as a hereditary integral, $h^{\rm tail}_{ij}(t) \propto \tfrac{GM}{c^3}\int_0^{\infty}\mathrm{d}\tau\,\dddot{Q}_{ij}(t-\tau)\,\ln\tau$ , a convolution of the quadrupole's history against a logarithmic kernel. Why: the backscattered piece depends on when the wave was emitted and how long it has been climbing the potential, so it is non-local in time — unlike the instantaneous multipoles that depend only on the source's present state.

Step 3. Count the PN order. The prefactor $GM/c^3$ times the source frequency $\omega$ builds the dimensionless combination $GM\omega/c^3=x^{3/2}$ , which multiplies the Newtonian flux; the logarithmic kernel integrates against a monochromatic quadrupole to yield a factor of $\pi$ , so the tail sits at $4\pi\,x^{3/2}$ relative to leading. Why: a single extra power of $GM\omega/c^3$ is exactly one-and-a-half PN orders, which is why the first tail is a half-integer effect and not a whole-order one.

Step 4. Read the physics. This $4\pi\,x^{3/2}$ term is the lowest-order purely nonlinear contribution to the radiation — gravity acting as a source for its own propagation — and, converted into the SPA phase, it becomes the celebrated $-16\pi\,v^3$ coefficient of equation (7). Why: detecting that coefficient in the inspiral phasing is a direct measurement of how the gravitational field scatters off itself, one of the sharpest available probes of the theory's nonlinearity. *Verified: the $4\pi$ tail coefficient at 1.5PN is the standard Blanchet & Damour (1992) result, reproduced in Blanchet, Living Reviews (2014), Eq. (314); its SPA image is the $-16\pi\,v^3$ phase term of Maggiore Vol. 1, Sec. 5.6.*

Go deeper · Derivation C · 2 minCounting cycles and why high post-Newtonian orders are unavoidable

Derivation CCounting cycles and why high post-Newtonian orders are unavoidable

The purpose here is to derive the closed-form leading-order count of gravitational-wave cycles a binary spends between a starting frequency and coalescence, and to use it to explain why sub-radian phase accuracy forces the field to compute so many PN orders. We integrate the Newtonian chirp once more, but now bookkeep the accumulated phase directly as a number of cycles, then estimate the size of the successive PN corrections to that number using the smallness of $v/c$ in band.

Step 1. Start from the Newtonian chirp of the previous derivation and change variables from time to frequency, writing the number of accumulated cycles as $N=\phi_{\rm GW}/2\pi=\int (f/\dot f)\,\mathrm{d}\ln f$ . Why: what a detector integrates is cycles, not seconds, so recasting the chirp as cycles-per-log-frequency is the physically relevant bookkeeping.

Step 2. Carry out the integral to leading order. Using $\dot f\propto \mathcal{M}^{5/3}f^{11/3}$ , the count from a low frequency $f$ up to coalescence is $N=\tfrac{1}{32\pi}\big(\pi G\mathcal{M}f/c^3\big)^{-5/3}$ . Why: this compact formula shows the cycle count is governed entirely by the chirp mass and the entry frequency, and it diverges as $f\to0$ — the lower the detector reaches, the more cycles it banks.

Step 3. Put in numbers for a neutron-star binary. With $\mathcal{M}=1.219\,M_\odot$ and $f=10\ \mathrm{Hz}$ , the formula gives $N\approx 1.6\times10^4$ cycles. Why: this is the concrete origin of the "sixteen thousand cycles" figure — the enormous lever arm that makes gravitational-wave phasing so sensitive and so demanding.

Step 4. Estimate the correction sizes. Each PN order multiplies the integrand by a factor of order $(v/c)^{2n}$ , and integrated over the band the 1PN piece is worth a few hundred cycles, the 1.5PN tail a couple hundred, and so on down the ladder as in the phase-budget figure. Why: to keep the total phase error below one cycle out of sixteen thousand, one must retain every term whose contribution exceeds unity — which for a neutron-star binary runs out to 3.5PN, settling how far the series must be pushed. *Verified: $N=\tfrac{1}{32\pi}(\pi G\mathcal{M}f/c^3)^{-5/3}=1.6\times10^4$ for $\mathcal{M}=1.219\,M_\odot$ , $f=10\ \mathrm{Hz}$ ; the per-order cycle contributions 16031 / 439 / 211 / 9.9 match the accumulated-phase expansion in Blanchet, Living Reviews (2014).*

Limiting cases

PN modeling is controlled by the smallness of $v/c$ ; its accuracy and its breakdown are governed by where the binary sits in its evolution.

Case	Result	Meaning
Slow inspiral (PN valid)	$v/c \lesssim 0.3,\quad \epsilon=v^2/c^2 \ll 1$	The well-separated regime: the series converges usefully and supplies nearly all the detectable cycles and the parameter estimation.
Newtonian / quadrupole limit	$\mathcal{F} \to \tfrac{32}{5}\tfrac{G}{c^5}\mu^2 r^4\omega^6$	Dropping every PN bracket recovers the leading quadrupole flux for a circular orbit of separation $r$ and angular frequency $\omega$ .
Test-mass limit	$\nu=\mu/M \to 0$	For $m_1\gg m_2$ the PN series must reproduce the exact black-hole perturbation (Teukolsky) result — a stringent check on every coefficient.
Near-merger breakdown	$v/c \to \mathcal{O}(1),\quad r\to r_{\rm ISCO}$	At the innermost stable circular orbit $v/c\approx0.41$ ; the parameter is no longer small, and the series must be resummed (EOB) or replaced by numerical relativity.

Common trapThe nPN label counts powers of v/c, not the order a term first appears in the metric

Two traps recur. First, " $n$ PN" means relative order $(v/c)^{2n}$ beyond the leading behaviour of the quantity in question — and the energy and flux series begin at different absolute orders, so one must always say which quantity's leading term the count is measured from. Quoting "the 3PN flux" and "the 3PN metric" as if they were the same expansion depth is a category error. Second, the half-integer orders are not bookkeeping filler inserted to make the series look symmetric: they are physical, dissipative, time-asymmetric effects. Radiation reaction — the local back-force that actually shrinks the orbit — first appears at 2.5PN, and the leading gravitational-wave tail appears at 1.5PN. Treating them as optional "small odd terms" discards precisely the pieces that make the orbit decay and that test the nonlinearity of gravity.

Worked examples

1 · Chirp mass and cycle count of a neutron-star binary (fully worked)

Two $1.4\,M_\odot$ neutron stars inspiral through a ground-based detector. Find the chirp mass, the number of gravitational-wave cycles accumulated from a 10 Hz starting frequency to coalescence, and the corresponding time in band.

Step 1. The symmetric mass ratio of an equal-mass binary is $\nu=m_1m_2/M^2=(1.4)(1.4)/(2.8)^2=\tfrac14$ , and the total mass is $M=2.8\,M_\odot$ . The chirp mass follows from equation (8): $\mathcal{M}=\nu^{3/5}M=(0.25)^{0.6}\times2.8\,M_\odot$ . Since $(0.25)^{0.6}=0.4353$ , this gives $\mathcal{M}=\mathbf{1.219\,M_\odot}$ — the value measured for GW170817 to four significant figures.

Step 2. Convert to seconds via $G\mathcal{M}/c^3$ . With $\mathcal{M}=1.219\,M_\odot=2.42\times10^{30}\ \mathrm{kg}$ , one finds $G\mathcal{M}/c^3=6.0\times10^{-6}\ \mathrm{s}$ , so at $f=10\ \mathrm{Hz}$ the velocity parameter is $v=(\pi G\mathcal{M}f/c^3)^{1/3}$ -scale — small, confirming the PN expansion is deep in its domain.

Step 3. Apply the cycle formula from Derivation 3: $N=\tfrac{1}{32\pi}(\pi G\mathcal{M}f/c^3)^{-5/3}$ . Plugging in $G\mathcal{M}f/c^3=6.0\times10^{-5}$ gives $N=\mathbf{1.6\times10^{4}\ \text{cycles}}$ . Sixteen thousand cycles of phase are exactly the lever arm that makes the neutron-star chirp mass the best-measured quantity in gravitational-wave astronomy.

Step 4. The time in band follows from integrating the chirp: $\tau=\tfrac{5}{256}(c^3/G\mathcal{M})^{5/3}(\pi f)^{-8/3}$ . Evaluating gives $\tau\approx\mathbf{1.0\times10^{3}\ s}$ from 10 Hz — about seventeen minutes of signal, shrinking to roughly $54\ \mathrm{s}$ if the detector only reaches 30 Hz. So $\mathcal{M}=1.219\,M_\odot$ , $N\approx16000$ cycles, $\tau\approx1000$ s.

2 · The tail is a two-hundred-cycle effect

Using the phase-budget figure, show quantitatively why the 1.5PN tail cannot be neglected, and contrast it with the 2PN instantaneous term.

Step 1. Read the accumulated cycle contributions for the same $1.4+1.4\,M_\odot$ binary from 10 Hz to the ISCO: 0PN contributes $\approx16031$ cycles, the 1PN correction $\approx+439$ , the 1.5PN tail $\approx-211$ , and the 2PN term only $\approx+9.9$ . These are differences of the accumulated-phase series evaluated at the two band edges.

Step 2. Compare magnitudes. The 1.5PN tail removes about 211 cycles of phase — more than twenty times the entire 2PN instantaneous contribution of $\sim10$ cycles, even though 1.5PN is a "lower" order than 2PN. A matched filter that omitted the tail would drift by $\sim211$ cycles, i.e. $\sim1300\ \mathrm{rad}$ , and lose the signal completely.

Step 3. Draw the lesson. The odd-power, hereditary tail is not a small correction sitting neatly below the instantaneous terms; its coefficient $4\pi$ (flux) / $-16\pi$ (phase) is numerically large, so a genuinely nonlinear effect dominates a nominally higher instantaneous order. The tail is worth $\sim$ 211 cycles versus $\sim$ 10 for 2PN — nonlinearity is loud in the phasing.

3 · A stellar-mass black-hole binary spends far less time in band

Repeat the estimate for a GW150914-like binary, $m_1=36\,M_\odot$ , $m_2=29\,M_\odot$ , entering at 30 Hz, to see why heavy binaries are "merger-dominated."

Step 1. Chirp mass: $\mathcal{M}=(36\cdot29)^{3/5}/(65)^{1/5}\,M_\odot=\mathbf{28.1\,M_\odot}$ , about twenty-three times the neutron-star value.

Step 2. Because $N\propto\mathcal{M}^{-5/3}$ , scaling from Example 1 (and adjusting the entry frequency to 30 Hz) gives only $N\approx\mathbf{14}$ gravitational-wave cycles from 30 Hz to merger, lasting $\tau\approx\mathbf{0.3\ s}$ . The ISCO frequency $f_{\rm ISCO}=c^3/(6^{3/2}\pi GM)\approx\mathbf{68\ \mathrm{Hz}}$ sits barely above the entry frequency.

Step 3. Interpret. A heavy binary reaches $v/c\approx0.41$ within a handful of cycles, so PN theory governs only a sliver of the observed signal and the information lives in the merger and ringdown — which is why heavy-binary analysis leans on effective-one-body and numerical-relativity waveforms, while the neutron-star case is almost entirely PN-dominated. $\mathcal{M}=28.1\,M_\odot$ , $\sim$ 14 cycles, $\tau\approx0.3$ s: heavy = merger-dominated.

Physics of the portrait

Post-Newtonian waveform theory is the work of many hands over a century, but its modern high-order form — the multipolar post-Minkowskian machinery, the systematic treatment of tails, and the resummation that bridges to merger — bears the stamp of one figure above all.

Origins · 1 minThibault Damour and how this was discovered

Thibault Damour

born 1951 · the multipolar post-Newtonian formalism, gravitational-wave tails, and the effective-one-body method

Thibault Damour has shaped the analytic theory of the two-body problem in general relativity more than any living physicist. Trained in France and long based at the Institut des Hautes Études Scientifiques, he confronted in the 1980s a problem that had defeated the field for decades: computing the equations of motion and radiation of a compact binary to an order high enough to matter for the gravitational-wave detectors then being planned. With Luc Blanchet he built the multipolar post-Minkowskian formalism, a rigorous scheme for iterating the nonlinear Einstein equations and matching the near zone to the wave zone — and in 1992 the two established that gravitational-wave "tails," the backscattering of radiation off the source's own curvature, enter the flux at 1.5PN, the first purely hereditary effect. Damour's "effacement principle" explained why the internal structure of each body decouples from the orbital dynamics until very high order, justifying the point-particle description. In 1999, with Alessandra Buonanno, he introduced the effective-one-body method, which resums the divergent PN series into a form that follows the binary all the way through merger and ringdown; its descendants are among the template families LIGO and Virgo use today. When GW150914 arrived in 2015, the analytic scaffolding that turned a wobble in an interferometer into two black-hole masses had been built, in large part, in his notebooks.

History

1938

— Einstein, Infeld and Hoffmann derive the first post-Newtonian equations of motion for gravitating bodies

1965

— Chandrasekhar formulates the post-Newtonian hydrodynamics of self-gravitating fluids

1974

— Hulse and Taylor discover the binary pulsar PSR B1913+16, later confirming orbital decay at the quadrupole rate

1992

— Blanchet and Damour establish the 1.5PN gravitational-wave tail as the first hereditary effect

2015

— LIGO detects GW150914, matched by post-Newtonian, effective-one-body and numerical-relativity templates

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), Ch. 36 · Wald, General Relativity (1984), Ch. 4 · Carroll, Spacetime and Geometry (2004), Ch. 7 · Hartle, Gravity (2003), Ch. 23 · Schutz, A First Course in General Relativity, 2nd ed. (2009), Ch. 9 · Maggiore, Gravitational Waves Vol. 1 (2008), Ch. 5 · Blanchet, "Gravitational Radiation from Post-Newtonian Sources and Inspiralling Compact Binaries," Living Rev. Relativity 17, 2 (2014) · Buonanno & Damour, Phys. Rev. D 59, 084006 (1999).

Provenance. The PN energy and flux coefficients (1PN energy $-(9+\nu)/12$ ; flux $-1247/336-35\nu/12$ at 1PN; the $4\pi$ tail at 1.5PN) follow Blanchet, Living Rev. Relativity 17, 2 (2014). The SPA phase coefficient $3/128$ , the 1PN term $3715/756+55\nu/9$ , and the $-16\pi$ tail term match Maggiore, Gravitational Waves Vol. 1 (2008), Sec. 5.6. The cycle-budget numbers (16031 / 439 / 211 / 9.9 for a $1.4+1.4\,M_\odot$ binary, 10 Hz to ISCO) are computed here from the accumulated-phase expansion and agree with the standard tables. Quadrupole-formula background is from MTW Ch. 36 and Carroll Ch. 7. A verifier should recompute the leading SPA phase by integrating $\mathrm{d}x/\mathrm{d}t=(64/5)(c^3/GM)\nu\,x^5$ and applying stationary phase, check $\mathcal{M}=\nu^{3/5}M=1.219\,M_\odot$ for equal $1.4\,M_\odot$ masses, and confirm $N=\tfrac{1}{32\pi}(\pi G\mathcal{M}f/c^3)^{-5/3}\approx1.6\times10^4$ at 10 Hz.

Prerequisites. Quadrupole Formula · Multipole Expansion of Gravitational Radiation · Lorenz (Harmonic) Gauge · Linearized Einstein Equations · Isaacson Stress-Energy of Gravitational Waves. Leads to. Effective One-Body Formalism · Numerical Relativity Waveforms · Matched Filtering and Parameter Estimation.

Ask the tutor: Ask me to walk through how integrating the Newtonian chirp $\mathrm{d}x/\mathrm{d}t\propto x^5$ collapses the two masses into a single chirp mass, or how the 1.5PN tail becomes the $-16\pi\,v^3$ term you can test in the data.

What exactly does "post-Newtonian order" count?

It counts powers of $(v/c)^2$ relative to the leading behaviour of whatever quantity you are expanding. The Newtonian result is 0PN; a correction suppressed by $(v/c)^2$ is 1PN, by $(v/c)^3$ is 1.5PN, and so on. The subtlety is that "leading" is quantity-specific: the flux starts at a different absolute order from the metric or the equations of motion, so "3PN flux" and "3PN metric" are not the same depth of calculation. Always anchor the count to the quantity's own leading term, and remember that half-integer orders carry the dissipative, radiation-reaction and tail physics.

Why is the gravitational-wave phase, rather than the amplitude, the thing that matters?

Because matched filtering extracts a signal by correlating the data with a template, and that correlation is destroyed by phase misalignment long before amplitude errors matter. A detector may see tens of thousands of cycles; if the template's phase drifts from the true phase by even a fraction of a radian per cycle, the accumulated mismatch washes the correlation to zero and the signal-to-noise vanishes. The amplitude sets how loud an event is, but the phase is where the physics — masses, spins, tidal deformability — is encoded and where PN theory concentrates its effort. This is why the field pushed the phase to 3.5PN while the amplitude is often kept only to lower order.

What makes a "tail" different from an ordinary multipole contribution?

An ordinary multipole term is instantaneous: the field at a given retarded time depends only on the source's state at that one earlier moment. A tail is hereditary — it depends on the source's entire past through a time integral, because the outgoing wave backscatters off the static curvature produced by the binary's own mass and only reaches the detector after wandering through that potential. Mathematically the tail is a convolution of the quadrupole's history against a logarithmic kernel. Physically it is the first sign that gravity is nonlinear: the field acts as a source for its own propagation. The leading tail enters the flux at 1.5PN with coefficient $4\pi$ .

How can the number of cycles be so precise if each correction is only a percent?

The precision comes from the length of the lever arm, not the size of any single correction. A neutron-star binary banks about sixteen thousand cycles between 10 Hz and merger, and the detector tracks the phase coherently across all of them. A one-percent error in the phasing rate, integrated over that many cycles, is not one percent of a cycle — it is of order a hundred cycles of drift, which annihilates the matched-filter correlation. So the requirement flips: you need each PN correction not because it is individually large but because the huge cycle count amplifies even tiny per-cycle errors into a total phase error that must stay below roughly one cycle.

Does the adiabatic (energy-balance) approximation ever fail?

Yes, near merger. Energy balance assumes the orbit evolves through a slow sequence of quasi-circular orbits, each shedding exactly the energy the waves carry off, so that the radiation-reaction timescale is long compared with an orbital period. That holds beautifully in the wide, slow inspiral, but as the binary approaches the innermost stable circular orbit the inspiral timescale shortens until it is comparable to the orbital period, the "sequence of circular orbits" picture collapses, and the plunge is no longer adiabatic. There the PN series must be resummed — the effective-one-body method builds a Hamiltonian that captures the plunge — or replaced outright by a numerical-relativity simulation.

Why does the chirp mass appear at leading order but the mass ratio only later?

Because the leading (Newtonian) phase, worked out in Derivation 1, depends on the two masses solely through the combination $\mathcal{M}=\nu^{3/5}M$ . The algebra of integrating the quadrupole chirp simply does not separate $m_1$ from $m_2$ at that order — only their chirp-mass combination survives. The symmetric mass ratio $\nu$ first enters as an independent handle at 1PN, through the bracket $\big(\tfrac{3715}{756}+\tfrac{55}{9}\nu\big)v^2$ . Since the leading term dominates the accumulated phase, $\mathcal{M}$ is tightly constrained while $\nu$ — and hence the split into individual masses — rides on weaker, higher-order corrections and is measured far less precisely.

How do post-Newtonian waveforms connect to the models LIGO actually uses?

They are the low-frequency foundation. The production template families are not raw PN series — near merger those series diverge — but PN results seed them. The effective-one-body approach resums the PN energy and flux into a Hamiltonian and radiation-reaction force that remain accurate through merger, then calibrates a few free functions against numerical-relativity simulations, yielding the SEOBNR family. The phenomenological IMRPhenom family stitches a PN-based inspiral phase directly onto numerical-relativity merger–ringdown fits in the frequency domain. In both, the many-cycle inspiral — where numerical relativity is far too expensive to simulate — is governed by PN theory, so the analytic coefficients derived here are quietly present in every parameter-estimation run.