Cross-correlating dark sirens across bands sharpens cosmology and gravitational-wave bias estimates

Gravitational-wave astronomy may be on the verge of its most powerful cosmological upgrade yet, according to a new theoretical study showing that combining space-based and ground-based gravitational-wave detectors into a single multi-band observing network could dramatically sharpen measurements of the universe’s expansion rate and reveal, for the first time at percent-level precision, how gravitational-wave sources cluster across cosmic history.

The study, published in The European Physical Journal C, presents the first Fisher-matrix forecast for so-called cross-correlation dark siren cosmology using multi-band gravitational-wave observations. A team of cosmologists at Northeastern University in China, led by Ji-Yu Song and including Ya-Nan Du, Yue-Yan Dong, Jing-Fei Zhang and Xin Zhang, cross-correlated simulated gravitational-wave events from a network comprising the proposed deci-hertz space observatory B-DECIGO and the third-generation ground-based detectors Einstein Telescope and two Cosmic Explorer facilities with the photometric galaxy catalog expected from China’s Space-station Survey Telescope, or CSST.

The motivation behind the work stems from two of the deepest tensions in modern cosmology. The Hubble tension, a discrepancy now approaching six standard deviations between the locally measured Hubble constant and the value inferred from the cosmic microwave background under the standard ΛCDM model, has resisted every attempt at explanation through systematic errors. Meanwhile, recent results from the Dark Energy Spectroscopic Instrument, based on baryon acoustic oscillation measurements of more than fourteen million galaxies and quasars, show an approximately three-sigma preference for evolving dynamical dark energy over the cosmological constant. Both puzzles underscore the urgent need for independent probes of cosmic expansion with entirely different sources of systematic uncertainty. Gravitational waves provide exactly such a probe.

The standard-siren technique exploits a fundamental property of gravitational-wave signals: the amplitude of the waveform from a compact binary coalescence is inversely proportional to the luminosity distance of the source. Unlike light from supernovae, which must be calibrated through a chain of distance indicators, gravitational-wave distances are absolute measurements requiring no external calibration ladder. When combined with the redshift of the host galaxy, they directly constrain the expansion history of the universe. The landmark event GW170817, a binary neutron star merger with an electromagnetic counterpart, demonstrated the technique’s power. But bright sirens with identified host galaxies are extraordinarily rare, expected to make up less than one percent of detections even in the era of third-generation ground-based detectors. The vast majority of events are dark sirens, invisible to telescopes, whose redshifts must be inferred statistically.

The cross-correlation method offers a way around this limitation. Instead of identifying individual host galaxies, researchers bin gravitational-wave events by luminosity distance and cross-correlate them with galaxies binned by redshift. The cross-correlation signal reaches its maximum only when the assumed distance-redshift relation matches the true cosmology, making the method sensitive to cosmological parameters while remaining largely immune to incompleteness in galaxy catalogs and to assumptions about the black hole population. Crucially, the same analysis simultaneously measures the gravitational-wave clustering bias, a quantity that describes how the spatial distribution of gravitational-wave sources relates to the underlying dark matter field.

Sky localization is the critical bottleneck, and this is where multi-band observation enters. B-DECIGO, a proposed space-borne interferometer operating in the 0.1 to 10 hertz deci-hertz band, would observe compact binary inspirals months or even years before their merger signals sweep into the hertz-band sensitivity range of ground-based detectors. By combining the long-baseline early inspiral tracking of the space detector with the high signal-to-noise merger observation from the ground network, sky localization improves by two to three orders of magnitude compared to single-band detection. The forecast quantifies this vividly: the multi-band configuration achieves a median 90-percent-credible sky localization area of roughly 0.02 square degrees, compared to about 16 square degrees for the ground-only network, despite both detecting a comparable sixty-four thousand binary black hole events per year.

The consequence for cosmology is substantial. In the ΛCDM model, after ten years of observation the multi-band network achieves a fractional precision on the dimensionless Hubble parameter of 0.35 percent, a 37 percent improvement over the ground-only configuration and an 86 percent improvement over B-DECIGO operating alone. Even with just one year of data, the multi-band network would already reach 0.70 percent precision. These constraints come on top of the galaxy survey’s own clustering information, which alone delivers 3.87 percent precision, demonstrating that the gravitational-wave cross-correlation contributes qualitatively new geometric information through the absolute distance scale rather than merely refining existing measurements.

When the analysis extends to the more flexible w0waCDM framework, which allows the dark energy equation of state to evolve with time, the multi-band advantage on cosmological parameters moderates considerably. The additional degeneracies between the Hubble parameter and the dark energy parameters weaken all constraints, and the multi-band improvement shrinks to roughly 4 percent for the Hubble parameter. The authors attribute this to the physics of the measurement: distance-scale information is probed primarily at large angular scales, where both ground-only and multi-band configurations retain adequate signal, so the dramatic localization advantage matters less for dark energy parameters. Improvements to the dark energy equation of state parameters themselves reach about 14 to 15 percent.

The most striking result of the study concerns the gravitational-wave clustering bias, where the multi-band localization advantage proves decisive. Because the bias measurement requires high angular multipole modes, which are severely damped by poor sky localization in the ground-only configuration, the contrast between networks is dramatic. At redshifts between one and two, where the gravitational-wave event density peaks, the multi-band network constrains the clustering bias to roughly 3 percent precision per redshift bin, compared with 8 to 60 percent for the ground-only configuration and 20 to 33 percent for B-DECIGO alone. The authors emphasize that this level of precision, resolved across fifteen independent redshift bins, opens an entirely new window onto the astrophysics of compact binary mergers.

The reason astrophysicists care deeply about the clustering bias is that different formation channels for black hole binaries predict distinct clustering signatures. Binaries formed through isolated stellar evolution trace their host galaxies and should show a clustering bias that increases with redshift. Binaries assembled dynamically in dense stellar environments such as globular clusters, or assisted by the disks of active galactic nuclei, would cluster differently. Most provocatively, a population of primordial black holes formed in the early universe would track the dark matter distribution itself, with a clustering bias near unity and roughly independent of redshift. Percent-level, redshift-resolved bias measurements of the kind forecast for the multi-band network could distinguish among these scenarios, potentially answering whether some fraction of the dark matter consists of primordial black holes.

The technical machinery behind the forecast is careful. The team computed tomographic angular power spectra using the public code pyccl under the Limber approximation, retaining density, lensing, and redshift-space or luminosity-distance-space distortion contributions while discarding subdominant Doppler and gravitational potential terms. Gravitational-wave events were binned in luminosity-distance space, and because the mapping from distance to redshift depends on the cosmological parameters, perturbing the cosmology shifts the effective redshift range of each gravitational-wave kernel and changes its overlap with the galaxy bins. This sensitivity is the physical engine of the cosmological constraint. The authors treated galaxy and gravitational-wave clustering biases as free parameters in each of fifteen bins, a deliberately conservative choice that avoids assuming a functional form for bias evolution, and marginalized over all thirty bias parameters alongside the cosmological parameters in a joint Fisher matrix analysis.

The authors also examined the robustness of their assumptions. Re-weighting the simulated black hole population to the latest catalog-based distribution and updating the merger rate tightened the Hubble parameter constraints by only a few percent, leaving the best-constrained bias bin at the 3 percent level, so the main conclusions are insensitive to population modeling. They acknowledge remaining limitations, notably the omission of photometric redshift calibration biases, which could make the forecasts somewhat optimistic, and the inherent restriction of the Fisher formalism to Gaussian posteriors.

Future extensions could push the constraints further. Combining the cross-correlation with cosmic shear measurements in a full three-by-two-point analysis, adding cosmic microwave background priors to break parameter degeneracies, incorporating spectroscopic galaxy surveys to reduce photo-z scatter, and merging the technique with spectral siren methods that extract cosmological information from gravitational-wave mass distributions would all sharpen the picture further. Independent calibration of the clustering bias from simulations or multi-messenger observations could convert it from a nuisance parameter into a genuine astrophysical observable.

What emerges is a coherent vision of the next decade of gravitational-wave cosmology. Ground-based detectors supply the event numbers; space-based deci-hertz observatories supply the angular precision; and galaxy surveys supply the three-dimensional map against which the gravitational-wave sky is measured. The study argues that none of these components alone can unlock the full potential of dark siren cosmology, but together they could deliver sub-percent constraints on the Hubble constant and, perhaps more importantly, the first precision atlas of where in the universe black hole binaries choose to form.