Gravitational-Wave Tidal Inference

Gravitational-wave tidal inference measures a small, mass-dependent correction to a binary inspiral, not an equation of state directly. The chain is: a material model determines each star’s Love response; the two responses enter a waveform phase; calibrated strain constrains masses, spins, and a binary tidal combination; and a common-EOS model maps those latent parameters back to pressure and radius. Every link introduces assumptions. This page fixes that chain, its statistical normalization, and the event-level reproduction boundary with an evidence cutoff of 2026-08-18.

Required background. Equivalence principles and gravitational measurement supplies the distinction between source parameters, detector response, and measured strain.

Helpful background. Stress–energy and conservation supplies the material source; the Einstein equation supplies the stellar equilibrium map; relativistic matter and compact objects develops stellar Love response; gravitational radiation and waveforms develops inspiral radiation and detector strain; and computational relativity supplies waveform and inference validation.

From stellar response to the binary tidal parameter

For a nonrotating star in general relativity, define compactness $C=GM/R$ and the dimensionless quadrupolar deformability

$$\Lambda=\frac{2}{3}\frac{k_2}{C^5},$$

where $R$ is the areal surface radius and $k_2$ is obtained by matching the regular static interior perturbation to the exterior tidal field. The theory, multipole, frequency regime, surface condition, normalization, and matching convention are part of the definition; Hinderer 2008, § 4.2, Eqs. (15)–(23), with the 2009 erratum gives the standard relativistic stellar construction. The erratum corrects Eqs. (20) and (23) and must accompany the original article.

At leading adiabatic order, a binary constrains mainly the mass-weighted combination

$$\widetilde\Lambda =\frac{16}{13} \frac{ (m_1+12m_2)m_1^4\Lambda_1 +(m_2+12m_1)m_2^4\Lambda_2 }{(m_1+m_2)^5}.$$

For a quasicircular inspiral, define $M=m_1+m_2$, $\eta=m_1m_2/M^2$ and $v=(\pi GMf)^{1/3}$. With $\widetilde h(f)=\int h(t)e^{2\pi ift}\,dt=\mathcal A(f)e^{i\Psi(f)}$ at positive frequency, the leading adiabatic quadrupolar contribution to the TaylorF2 phase is

$$\delta\Psi_{\rm tidal} =\frac{3}{128\eta v^5} \left(-\frac{39}{2}\widetilde\Lambda v^{10}\right) =-\frac{117}{256\eta}\widetilde\Lambda v^5.$$

The factor $v^{10}$ relative to the leading point-particle phase makes this a 5PN effect. This normalization and its next-order correction are given in Wade et al. 2014, Eqs. (4)–(5) and Appendix A.5. The masses and frequency must be in the same frame: observed $f$ pairs with detector-frame masses. Static response and adiabatic inspiral cease to suffice near resonances or merger.

At fixed masses and response, the magnitude of this leading phase term grows as $f^{5/3}$. Higher frequencies therefore amplify tidal information, while also increasing sensitivity to waveform error. The term is correlated with mass ratio, spins, spin-induced quadrupoles, and the point-particle baseline. A quoted interval on $\widetilde\Lambda$ is conditional on the waveform family and analysis band. Relating $\Lambda_1$ and $\Lambda_2$ through a common EOS can improve precision, but it also replaces two free responses with a restrictive physical prior. A phase transition that generates disconnected stable branches requires special care because mass-to-deformability may not be single-valued over the assumed family.

Calibrated strain and the event likelihood

For detectors $a$ with mutually independent, stationary Gaussian noise over the analyzed segments, inference uses the likelihood

$$p(d\mid\theta)\propto \exp\!\left[-\frac12\sum_a \langle d_a-h_a(\theta)\mid d_a-h_a(\theta)\rangle_a\right],$$

with

$$\langle x\mid y\rangle_a =4\,\mathrm{Re}\int_{f_{\min}}^{f_{\max}} \frac{\widetilde x^{*}(f)\widetilde y(f)}{S_{n,a}(f)}\,df.$$

The data product is calibrated strain, while $S_{n,a}$ is an estimated one-sided noise power spectral density and $h_a$ includes sky response, distance, orientation, masses, spins, tides, and calibration nuisances. Calibration amplitude and phase uncertainty can be sampled with response-spline parameters. Data selection must state detector network, science segments, data-quality vetoes, gates or glitch subtraction, sample rate, analysis duration, $f_{\min}$, $f_{\max}$, PSD estimation, and any marginalization over calibration or noise lines.

The primary public tidal benchmark remains the GWOSC GW170817 event release, using event-level HDF5 strain, data-quality information, and the release’s parameter-estimation products. GWOSC, GW170817 data release is the archive entry; Abbott et al. 2019, §§ II–IV and appendices documents the LIGO–Virgo likelihood choices and tidal analyses. At the evidence cutoff, GWTC-5.0, published 2026-05-26, is the current official cumulative catalog context. Its GWOSC GWTC-5.0 Parameter Estimation and Data Products sections list Stable Releases 6 and 9 and state that Event Portal parameter-estimation values and event links use the newer Stable Release 9 posterior records. Candidate, strain, and posterior products must therefore be pinned separately rather than described by one undifferentiated release number. A catalog row or population point estimate does not replace an event’s strain, calibration, and posterior metadata.

Priors, selections, and waveform systematics

The parameter prior must expose detector-frame component masses, spin magnitudes and tilts, luminosity distance, sky and orientation, coalescence parameters, calibration, and either independent $\Lambda_i$ or a common EOS. Source-frame conversion requires a redshift model. A low-spin prior informed by Galactic systems and a broad spin prior answer different conditional questions. Priors uniform in $\Lambda$, radius, spectral EOS coefficients, pressure knots, or sound speed do not induce the same measure on tidal phase.

For a single already-detected event, the detection selection function is not divided into the ordinary parameter-estimation likelihood. It enters when events are combined as a population or when detection-conditioned rates and EOS hyperparameters are inferred. Then

$$\alpha(\lambda)=\int d\theta\, p_{\rm det}(\theta)\,p(\theta\mid\lambda)$$

must use the same search threshold and population model as the hierarchical numerator. Binary-neutron-star selection depends strongly on masses and distance and more weakly on tidal effects; omitting it is an assumption that should be tested, not hidden.

Waveform systematic uncertainty is probed by changing the point-particle approximant, tidal resummation or calibration, spin-quadrupole treatment, merger termination, and upper frequency. Reanalysis with several numerical-relativity-informed tidal models in Narikawa et al. 2020, §§ II–IV shows how $f_{\max}$ and waveform choice are varied rather than absorbed into the statistical interval. Other required alternatives are PSD duration, glitch treatment, calibration response, and independent versus EOS-related deformabilities.

When posterior samples are recycled into an EOS analysis, the event interim prior must be divided out. A kernel density estimate can bias a thin, curved mass–tidal likelihood; direct likelihood evaluation or validated density estimation is preferable. Effective sample size after reweighting, sampler convergence, and injection recovery are part of the uncertainty budget.

Reproduction path and scope of the result

A bounded rerun downloads the exact GW170817 strain files and data-quality segments from GWOSC, verifies file identities, recreates conditioning and PSDs, and evaluates a declared waveform over a declared frequency interval. It records inference-software and waveform versions, calibration model, priors, sampler settings and seeds, convergence statistics, and posterior samples. Scientific validation then repeats the run with at least one independent tidal approximant, an alternative $f_{\max}$, a broader spin prior, and perturbed calibration; a zero-tide injection checks that the pipeline does not manufacture positive deformability.

This page does not claim a new strain analysis. Its evergreen conclusion is structural: the detector constrains a correlated likelihood in masses, spins, and tidal response. An EOS constraint is valid only after the waveform, event prior, population selection, stellar branch, causality/stability restrictions, and inter-event commonality assumption have been declared. The dataset version and evidence date travel with any numerical bound.

Check your understanding

Show that for an equal-mass binary with identical stars, $m_1=m_2=m$ and $\Lambda_1=\Lambda_2=\Lambda$, the binary parameter reduces to $\widetilde\Lambda=\Lambda$.

Solution

Each numerator term is $(m+12m)m^4\Lambda=13m^5\Lambda$. There are two terms, so

$$\widetilde\Lambda =\frac{16}{13} \frac{26m^5\Lambda}{(2m)^5} =\frac{16}{13}\frac{26}{32}\Lambda =\Lambda.$$

This check tests the normalization. It does not make the general unequal-mass parameter an arithmetic mean; its $m_i^4$ weighting preferentially emphasizes the deformability associated with the mass combination appearing in the phase.

References

• Abbott, B. P., et al. “Properties of the Binary Neutron Star Merger GW170817.” Physical Review X 9 (2019): 011001. DOI.
• GWOSC. “GW170817 Event Release” and “GWTC-5.0 Data Release Documentation.” Gravitational Wave Open Science Center, LIGO Laboratory (accessed 2026-08-18). GW170817; GWTC-5.0.
• Hinderer, Tanja. “Tidal Love Numbers of Neutron Stars.” Astrophysical Journal 677 (2008): 1216–1220; erratum 697 (2009): 964. DOI. Open PDF, including erratum.
• Narikawa, Tatsuya, et al. “Reanalysis of the Binary Neutron Star Mergers GW170817 and GW190425 Using Numerical-Relativity Calibrated Waveform Models.” Physical Review Research 2 (2020): 043039. DOI.
• Wade, Leslie, Jolien D. E. Creighton, Evan Ochsner, Benjamin D. Lackey, Benjamin F. Farr, Tyson B. Littenberg, and Vivien Raymond. “Systematic and Statistical Errors in a Bayesian Approach to the Estimation of the Neutron-Star Equation of State Using Advanced Gravitational Wave Detectors.” Physical Review D 89 (2014): 103012. DOI; Open PDF.