This paper uses the binary neutron star merger GW170817 and its electromagnetic counterpart to test whether Newton’s gravitational constant G has changed over time. The authors build a model that lets G vary slowly with time. They then fit that model to the observed gravitational waves and to independent electromagnetic measurements of the source location, distance, and orientation. They find no evidence that G is changing and place a tight limit on its fractional time derivative.

Why a changing G matters: the constancy of G is a core part of the strong equivalence principle, a fundamental assumption behind general relativity. If G varied in time, it would change how compact objects like neutron stars move and how gravitational waves travel through an expanding universe. The authors include both effects in a single waveform model so the test is self-consistent. They also account for how neutron stars respond to changes in G through object-dependent “sensitivities,” and they enforce mass limits that come from the Tolman–Oppenheimer–Volkoff (TOV) equations, which describe neutron star structure.

What the team did: they analyzed publicly available data from the two LIGO detectors (Hanford and Livingston) downloaded from the Gravitational Wave Open Science Center. The analysis used the PyCBC parameter-estimation framework with a low-frequency cutoff of 20 Hz and a reference frequency of 40 Hz. To check robustness, they ran the analysis with two different gravitational-wave waveform families: TaylorF2 and IMRPhenomXAS_NRTidalv2. They combined the gravitational-wave data with electromagnetic constraints from the short gamma-ray burst GRB 170817A and other follow-up observations to break parameter degeneracies that affect gravitational-wave–only fits.

What they found: the nonstandard parameter that measures the ratio of G at the source to G at the detector is consistent with one. Using the lookback time to GW170817, their results give a conservative three-sigma bound on the fractional time derivative of G of ȦG/G ∈ [−3.36×10^−9, 5.34×10^−10] per year. In their wording, this is the most stringent bound obtained so far from real gravitational-wave observations. The posterior distributions of the binary parameters remained consistent with general relativity across both waveform models, which supports the stability of the result against modeling choices.