First Hubble constant from gravitational waves and galaxies: 67 km/s/Mpc, with wide error bars

A method that looks for where galaxy redshifts line up with gravitational-wave distances returns H0 = 67 (+18/−15) km/s/Mpc from two events, an interval that contains both the Planck and the SH0ES values.

Peer reviewed · Statistical analysis of two real events · Proof of concept

Astronomers have measured the Hubble constant, the universe’s present expansion rate, by a new route: matching the distances of gravitational-wave sources to the redshifts of galaxies in the same patch of sky. Isabela Santiago de Matos, Tessa Baker and colleagues report 67 (+18/−15) km/s/Mpc in The Astrophysical Journal, published online on 2 November 2026. The value sits close to both sides of the Hubble tension, because the uncertainty, about 25%, is far too large to choose between them.

Evidence at a glance

Is this a big deal?

It is a solid first step, not a verdict on the Hubble tension. The authors show a new kind of measurement works on real data, and the 5.9-sigma detection is the headline achievement. The H0 number is a demonstration: it agrees with everything and constrains little. The potential is in the scaling, covered below.

What scientists found

With GW190814 alone, the team detects the cross-correlation at 3.4 sigma (p-value 6.7×10^-4). Adding GW170817 raises it to 5.9 sigma (p-value 1.9×10^-9). These significances come from comparing the data with 6,000 or so simulations in which galaxies and wave sources are not related.

Converting the peak positions into a cosmic expansion rate gives H0 = 67 (+18/−15) km/s/Mpc at 68% for both events, and 53 (+31/−18) for GW190814 alone. The team also puts the first observational limit on “GW bias”, a number describing how strongly wave sources follow the cosmic web of matter: b_gw < 4.3 at 95% confidence. Only an upper bound is possible so far. The authors suggest two reasons: the bias trades off against H0, and the covariance is hard to model as the bias approaches zero.

How they found it

A gravitational wave from two merging objects reveals how far away they are, from the signal’s strength, but not their redshift. Galaxies give the opposite: redshift is easy, distance is hard. For small distances, redshift times the speed of light is roughly H0 times distance, so a matched pair of measurements gives H0.

The trick is how to match them. GW190814 was localised to 18.5 square degrees (90% confidence), about 90 full Moons of sky, and it came from some galaxy in there. The team divides the sky and distance into voxels, three-dimensional pixels, and counts galaxies in each. It does the same for the wave source’s probability map and measures the angular cross-correlation. Because wave sources live in galaxies, and galaxies cluster, the cross-correlation peaks where the galaxy redshift corresponds to the wave distance. For GW190814 that is near a distance of 255 megaparsecs (about 830 million light-years), at redshift of roughly 0.065. For GW170817 it is about 40 Mpc (130 million light-years).

The analogy is a shout heard in a dark city: loudness gives how far away, direction gives a rough neighbourhood, and the buildings in that neighbourhood at that distance are the candidates. The analogy breaks down because no single host is identified. The peak is a statistical excess spread over many galaxies.

The galaxy side is the K-band GLADE+ catalogue: roughly 280,000 galaxies after cuts, with the Milky Way’s plane masked out. The wave side is from GWTC-3.0, the LIGO-Virgo-KAGRA (LVK) catalogue. The team deliberately did not use GW170817’s known redshift, so that event, a neutron-star merger whose host galaxy is known, acts like an unlabelled one. They assumed flat priors of 20 to 120 km/s/Mpc for H0. The simulations are built at one fiducial cosmology from the cosmic microwave background (H0 = 67.36 km/s/Mpc) and then rescaled as H0 changes.

How this relates to the Hubble tension

The Hubble tension is a mismatch between two precise numbers. Planck’s measurements of the cosmic microwave background, interpreted with the standard cosmological model, give H0 = 67.4 ± 0.5 km/s/Mpc. The SH0ES team’s distance ladder, built on Cepheid stars and supernovae, gives 73.04 ± 1.04. The gap is about 5.6 km/s/Mpc, or roughly 4.9 times the two errors added in quadrature (SciQuest calculation).

The new interval, 52 to 85, contains both. The paper itself calls the result “not yet competitive” with the cosmic microwave background and standard-candle measurements, though compatible with other dark-siren results. Its pitch is different: gravitational waves bypass external distance calibration, so as an independent route it could eventually weigh in on the tension.

Why it matters to you

Nothing changes in daily life. A persistent mismatch between the early-universe and local values would mean either the standard cosmological model or some measurements need revisiting; background is in our explainer on the Big Bang theory. This result, with its wide interval, neither supports nor undermines either possibility.

What the evidence does not show

It does not show that the Hubble tension is real or resolved. It is not an independent check at useful precision either: with 25% uncertainty, almost any recent value fits.

Several assumptions matter. The simulations assume every galaxy is equally likely to host a merger, and GLADE+ is shallow, which the authors name as the main limitation of this work. Galaxy redshift errors are modelled as Gaussian, and peculiar velocities (galaxies’ own motions) were not included, though the authors expect the resulting shift to be smaller than the final uncertainty. Almost all of the signal comes from one event. The H0 and bias parameters also trade off against each other.

How this fits what we already knew

This is the third way to find redshifts for gravitational-wave sources. GW170817 is the only “bright siren”, with a visible counterpart and a known host. “Dark sirens” are events without a counterpart, normally analysed by checking candidate galaxies along the sky position and using assumptions about merger rates and black-hole masses. The most recent LVK constraint, cited in the paper, combines GWTC-5.0 with DES Y6 and gives 71.7 (+9.4/−7.5) km/s/Mpc.

Peak Sirens differs by using the three-dimensional clustering of galaxies and avoiding mass-distribution assumptions. That is the authors’ claim and has not been independently tested. The paper also notes that an earlier attempt with similar data returned a null result. The authors say that analysis collapsed the distance dimension, which removes the very correlation this method relies on.

Outside views and the authors’ framing

SciQuest has not interviewed an independent cosmologist and found no on-record outside comment. In the paper, the authors say that combining the new result with a spectral-siren analysis, which uses the black-hole mass distribution, gives 68 (+16/−8) km/s/Mpc. They call that a first approximation because it ignores correlations between the two analyses. They close by calling the work “the first stone of a castle of many opportunities.”

How other headlines framed it

SciQuest found no independent news outlet covering this paper by the time of writing, so there is no framing to compare. The result has been visible to specialists as an arXiv preprint since December 2025.

What happens next

The authors plan to repeat the analysis with deeper galaxy catalogues, naming Gaia and DESI Legacy as examples, and with more well-localised events from GWTC-3.0 and the later stages of the fourth LVK observing run. Their projection is that 100 events with sky maps like GW190814’s would give an H0 measurement near 4%. That is a forecast, not a result, and it assumes the same quality of localisation. They also plan tests of gravity theories in which waves travel differently from light. The related physics of how such events form is in our explainer on black-hole formation.

In one sentence

Gravitational-wave events and galaxy clustering have now yielded a first Hubble constant, 67 (+18/−15) km/s/Mpc, which proves the method works but cannot yet choose between the competing values.

Read next

• What is the Big Bang theory? Background on the expansion that H0 measures.
• What is the universe made of? A plain-language look at the standard-model inventory of matter and energy, which sets what H0 is compared against.
• NASA’s Roman Space Telescope launches to map the dark universe: a survey mission built to map galaxies.

Sources and disclosure

1. Santiago de Matos I., Dalang C., Baker T., Abramo R., Ferri J., Quartin M. “First Measurement of the Hubble Constant from Gravitational Wave–Galaxy Cross Correlations.” The Astrophysical Journal 1011, 175 (2026). doi:10.3847/1538-4357/ae9a7a. Preprint: arXiv:2512.15380 (v2, 14 August 2026, full text read).
2. Planck Collaboration. “Planck 2018 results. VI. Cosmological parameters.” Astronomy & Astrophysics 641, A6 (2020): arXiv:1807.06209.
3. Riess A.G. et al., SH0ES Cepheid and supernova measurement of H0, 2022: arXiv:2112.04510, DOI 10.3847/2041-8213/ac5c5b.

Disclosure: This article was drafted with AI assistance and checked against the sources above. The authors report funding from an ERC Starting Grant (SHADE), a Royal Society fellowship and Brazilian agencies; SciQuest has no connection to them. Calculations by SciQuest: uncertainty percentages from quoted intervals; gap 73.04 − 67.4 = 5.64, divided by √(0.5² + 1.04²) = 1.15 gives about 4.9; 255 Mpc × 3.26 = 830 million light-years; 18.5 deg² ÷ 0.21 deg² per full Moon ≈ 90. The graphic is a SciQuest original. Spotted an error? See our corrections policy.