Inspiral Coefficient and Consistency Tests

Inspiral tests ask whether the frequency-dependent phase measured from a compact-binary signal has the mutually constrained structure predicted by general relativity (GR). The central observable is not a free post-Newtonian coefficient by itself, but its posterior after masses, spins, waveform uncertainty, calibration, detector noise, and event selection have been propagated. A split-signal comparison asks a related question—whether disjoint frequency regions predict a common remnant—but it is trustworthy only when the split, priors, and waveform domains cannot manufacture disagreement.

Helpful background. Weak-Field Gravity supplies the perturbative meaning of the inspiral expansion; Probability and Statistical Inference supplies normalized likelihoods and prior measures; Gravitational-Wave Data Analysis and Inference supplies calibrated strain and source-parameter estimation; Tests of Gravity supplies the theory-to-measurement logic; and Black-Hole Perturbation Theory supplies the remnant spectrum used by late-signal comparisons.

Inspiral phase as an overdetermined prediction

For a quasi-circular binary, use $\tilde h(f)=\int h(t)e^{+2\pi ift}dt$, with inverse kernel $e^{-2\pi ift}$. For positive frequency, write the detector-frame signal as

$$\tilde h(f)=A(f;\vartheta)\exp\!\left[i\Psi(f;\vartheta)\right], \qquad v=(\pi GM_z f)^{1/3},$$

where $f$ is the dominant gravitational-wave frequency, $M_z=(1+z)(m_1+m_2)$ is the redshifted total mass, and $\vartheta$ contains masses, spins, orientation, distance, and coalescence variables. The symmetric mass ratio is $\eta=m_1m_2/(m_1+m_2)^2$, between zero and $1/4$. The dimensionless $v$ is approximately the orbital speed: the wave frequency is twice the orbital frequency for the dominant quadrupole. A useful bottom-up deformation is

$$\Psi(f)=\Psi_{\rm GR}(f;\vartheta) +\frac{3}{128\eta v^5} \sum_k \varphi_k(\vartheta)\,\delta\hat\varphi_k\,v^k,$$

where the displayed sum suppresses the separate logarithmic terms in a full post-Newtonian expansion. This fractional form applies where the corresponding GR coefficient $\varphi_k$ is nonzero; coefficients absent in GR require an absolute additive deformation instead. GR is the joint point $\delta\hat\varphi_k=0$. A term $v^k$ is at $k/2$ post-Newtonian order relative to the leading phase. Thus a $k=-2$ correction has a stronger low-frequency dependence than the leading quadrupole phase and can represent dipole radiation in suitable theories. Its coefficient still depends on the bodies’ charges and cannot be translated into a universal coupling without that map.

The usefulness of phase comes from counting many cycles. At leading quadrupole order, energy balance gives $\dot f\propto\mathcal M_z^{5/3}f^{11/3}$, where $\mathcal M_z=\eta^{3/5}M_z$ is the redshifted chirp mass. Consequently a phase correction proportional to the leading $f^{-5/3}$ term can be largely absorbed by changing $\mathcal M_z$. Higher powers, spin dependence, and independent parts of the signal break such degeneracies. This is why a fitted coefficient becomes a test only after masses and spins are fitted jointly. Yunes, Yagi, and Pretorius 2016, §II develops the mapping between waveform deformations and specific physical mechanisms.

Only one or a controlled subset of deviations should be varied unless the data resolve their covariance. A one-at-a-time bound is conditional on fixing the other deviations to zero; it is not a simultaneous bound on all coefficients. A complete alternative theory can change several powers together, introduce new harmonics, or invalidate the adiabatic and quasi-circular assumptions.

An internal split test instead divides the likelihood at a declared frequency $f_\mathrm{split}$. Inspiral data infer a final mass and spin $(M_f,\chi_f)_\mathrm{low}$ through a GR evolution model; the higher-frequency data infer $(M_f,\chi_f)_\mathrm{high}$. Fractional differences such as

$$\Delta_M=\frac{2(M_f^\mathrm{low}-M_f^\mathrm{high})} {M_f^\mathrm{low}+M_f^\mathrm{high}}$$

has null value zero. Repeated GR signals should produce the stated coverage under the full analysis; every individual credible region need not contain zero. The test requires a common parameter measure, compatible waveform domains, and adequate signal-to-noise ratio on both sides. A split fixed before looking at the data can be held fixed; a data-dependent split requires the choice rule in the calibration, or an explicitly defined marginalization. A mismatch of induced priors can move the apparent GR quantile even for a GR injection.

From released strain to a bounded conclusion

For detectors $I$, a Gaussian-noise likelihood has the schematic normalized form

$$p(d\mid\vartheta,\delta,C,\mathcal M) =\frac{ \exp[-\tfrac12\sum_I(d_I-h_I\mid d_I-h_I)_I] }{Z_n(C)},$$

For stationary, mutually independent detector noises with one-sided power spectral densities $S_{n,I}(f)$, the inner product is

$$(a\mid b)_I=4\operatorname{Re}\int_{f_{\min}}^{f_{\max}} \frac{\tilde a_I^*(f)\tilde b_I(f)}{S_{n,I}(f)}\,df.$$

It weights a discrepancy most strongly where the detector is quiet. Correlated detector noise requires the inverse spectral matrix instead of the detector sum. If the spectrum is inferred, its determinant in $Z_n$ cannot be discarded as a constant.

where $C$ denotes calibration and noise-spectrum nuisance variables, $\mathcal M$ the waveform family, and $Z_n$ contains the noise normalization. The posterior multiplies this likelihood by explicitly declared source and deviation priors. The event selection must be stated because inspiral duration, band signal-to-noise ratio, precession, higher modes, and data quality determine which systems enter a test; catalog-wide language is not justified by a selected subset.

At the evidence cutoff 2026-08-19, the peer-reviewed method baseline is the GWTC-3 analysis: its event and pipeline selections are tabulated, and its residual, inspiral–merger–ringdown, parameterized-inspiral, propagation, polarization, and ringdown tests are separated rather than pooled into one verdict LIGO–Virgo–KAGRA 2023, Table I and §§II–VIII. The current GWTC-5 test manuscript is a submitted draft, not a peer-reviewed replacement. It analyzes 168 test-selected signals from a catalog of 390 detections and retains explicit per-pipeline cuts LIGO–Virgo–KAGRA 2026, submitted manuscript P2500781-v17, Table 1 and §§2–3.

The reproducible strain and posterior baseline is the official GWTC-5/GWOSC stable parameter-estimation release 9. Its 16 kHz products match the source channels, whereas its 4 kHz products are decimated with a documented filter approximation; frequencies below 10 Hz are not calibrated and must not enter the likelihood GWOSC 2026, official GWTC-5.0 documentation, “Parameter Estimation” and “Notes on Strain Data”. P2600410-v1 separately supplies the released products for reproducing figures from the tests manuscript LIGO–Virgo–KAGRA 2026, official tests data release P2600410-v1. A rerun therefore pins the event list, file identifiers, calibration version, strain segment, data-quality flags, power spectral density, waveform and code versions, priors, sampler settings, and convergence diagnostics.

Results and limitations through 2026-08-19

Evidence or result	Interpretation	Limitation or scope
Peer-reviewed GWTC-3 consistency and coefficient analyses	Tested several complementary GR predictions with declared pipeline selections	Finite selected events, waveform families, and O3 calibration; not a test of every alternative theory
Submitted GWTC-5 coefficient analysis	A larger selected set is consistent with the tested inspiral structures under its stated priors	Submitted status; the 168 tested signals are not the full 390-event catalog
GR and deformed-signal injections	Can measure coverage, false deviations, and recovery of a controlled phase term	Valid only over the injected source, noise, calibration, and waveform population

The conservative position is that present posteriors have not established a departure from the tested GR waveform structure. A competing interpretation of an outlying coefficient must first survive waveform-family changes, calibration and noise marginalization, prior remapping, precession and eccentricity injections, and trials accounting. In particular, the submitted GWTC-5 study omits the traditional inspiral–merger–ringdown consistency result after demonstrated spurious biases, so that test must not be advertised as a current GWTC-5 result. This is negative evidence about that implementation, not evidence against the logical idea of cross-band consistency.

An adversarial independent check injects numerical-relativity and analytic GR signals across source parameters and noise realizations, then repeats the inference with different waveform families and split frequencies. It must recover calibrated coverage at zero deviation and recover a deliberately injected phase deformation with the expected sign and scaling. Failure of either control leaves the physical interpretation unresolved and model-dependent.

What could change the conclusion. The interpretation could change when a peer-reviewed GWTC-5 tests paper, superseding calibration or posterior release, corrected split-test construction, or injection campaign changes coverage, event selection, or the mapping from phase coordinates to complete theories.

Exercises

Suppose two split-signal analyses infer independent Gaussian remnant-mass estimates $M_L=70\pm4$ and $M_H=66\pm5$ in the same units. Estimate the significance of their difference, and explain why this number alone is not a GR test.

Solution

The difference is $M_L-M_H=4$, while independence gives a standard deviation $\sqrt{4^2+5^2}=\sqrt{41}\simeq6.4$. Therefore the nominal discrepancy is only about $0.62$ standard deviations. This comparison is not yet a GR test because the estimates may have different induced priors, shared calibration uncertainty, waveform bias, and a data-dependent split. A valid conclusion follows only after those covariances and the selection rule are included and GR injections demonstrate calibrated coverage.

References

• Gravitational Wave Open Science Center (2026), “GWTC-5.0 Data Release Documentation,” stable release 9 parameter-estimation, strain, and data-quality products, official release, DOI.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration, “Data Release for GWTC-5.0: Tests of General Relativity,” version P2600410-v1 (2026), official release.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration, “Tests of General Relativity with GWTC-3,” Physical Review D 107, 022005 (2023), DOI.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration, “Tests of General Relativity with GWTC-5.0,” submitted manuscript P2500781-v17 (2026), official DCC record.
• Nicolás Yunes, Kent Yagi, and Frans Pretorius, “Theoretical Physics Implications of the Binary Black-Hole Mergers GW150914 and GW151226,” Physical Review D 94, 084002 (2016), DOI.