🔭 astrophysics

Dark siren cross-correlations and the sensitivity of $H_0$ to methodological choices

This paper demonstrates that systematic biases in the gravitational wave cross-correlation method for measuring the Hubble constant ( $H_0$ ) can be effectively mitigated through appropriate modeling choices and large samples, while also showing that catalogue incompleteness can be directly incorporated into theoretical predictions without explicit modeling of missing populations.

Original authors: Madeline L. Cross-Parkin, Cullan Howlett, Leonardo Giani, Chris Blake, Tamara M. Davis

Published 2026-05-11

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Madeline L. Cross-Parkin, Cullan Howlett, Leonardo Giani, Chris Blake, Tamara M. Davis

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, invisible ocean. To understand how fast this ocean is expanding, scientists need to measure the distance between islands (galaxies) and the waves crashing against them (gravitational waves).

This paper is like a "quality control manual" for a new, high-tech way of measuring that expansion rate, known as the Hubble Constant ( $H_0$ ). The authors are testing a specific method called "Dark Siren Cross-Correlation."

Here is the breakdown of their work using simple analogies:

1. The Two Main Characters

The Galaxies (The Map): Think of galaxies as lighthouses scattered across the sky. We know where they are, but we don't know exactly how far away they are unless we assume a specific map of the universe.
The Gravitational Waves (The Ruler): When black holes crash, they send out ripples in space-time. These ripples act like a "standard ruler" that tells us the true distance to the crash. However, we often can't see the specific lighthouse (galaxy) that caused the crash. These are called "Dark Sirens."

2. The New Method: The "Crowd Match"

Instead of trying to find the exact lighthouse for every single wave crash (which is hard because the waves are fuzzy), this method looks at the crowd.

Imagine you are at a concert. You can't see the stage clearly (the gravitational wave location is blurry), but you can hear the music. You also have a map of where the audience is sitting (the galaxy catalog).

The Old Way: Try to guess exactly which seat the musician is sitting in.
The New Way (Cross-Correlation): Look at the pattern of the audience. If the musician is in a specific section, the crowd density in that section will be higher. By matching the "blurry" wave pattern with the "sharp" galaxy pattern, you can figure out how far away the musician is, even if you can't see them clearly.

3. What the Paper Actually Tested

The authors didn't just use this method; they stress-tested it to see if the "ruler" breaks under different conditions. They asked: "Does our math change the answer if we tweak how we do the calculation?"

They tested four main things:

A. The "Summary vs. The Whole Story" (Compression)

The Test: Do we need to look at every single detail of the crowd pattern (all the data), or is a single "summary score" enough?
The Finding: Using a single summary score (like an average) loses a tiny bit of precision (about 40-50% more uncertainty), but it saves a massive amount of computer time. It's like taking a photo of a crowd vs. counting every single person. For a quick estimate, the photo is fine.

B. The "Guessing Game" (Covariance)

The Test: When calculating errors, do we use a simple mathematical formula (a guess) or do we run thousands of fake universe simulations (a reality check)?
The Finding: If you have a small number of events, the simple formula guesses too optimistically and underestimates the error. The "fake universe" simulations are much more reliable. It's like predicting the weather: a simple rule of thumb works for a sunny day, but you need a supercomputer simulation to predict a hurricane.

C. The "Distorted Lens" (Systematic Biases)

The Test: What happens if our measurements are slightly off?
- Distance Errors: If our ruler is slightly wobbly, do we get the wrong answer?
- Bias: Do we assume galaxies are clustered differently than they really are?
The Finding:
- Wobbly Rulers: If the distance measurements are very fuzzy, the math naturally shifts the results. However, if you build this "fuzziness" directly into your computer model (forward modeling), the final answer stays correct.
- Wrong Assumptions: If you assume the crowd is static when it's actually moving, you get a biased answer. But, if you let the math adjust for this movement, you still get the right expansion rate.

D. The "Missing People" (Incompleteness)

The Test: What if our galaxy map is incomplete? What if we can only see the bright lighthouses and miss the dim ones?
The Finding: This is the biggest win of the new method. In the old way, missing lighthouses ruined the measurement. In this new "Crowd Match" method, you can simply tell the computer, "Hey, we only see the bright ones." The math adjusts automatically. You don't need to guess where the missing people are; you just use the pattern of the people you do see.

4. The "Bucket" Size (Binning)

To make the math work, scientists group data into "buckets" (bins) of distance and redshift.

The Test: Should the buckets be tiny (high detail) or big (low detail)?
The Finding:
- Galaxies: We have so many of them that tiny buckets are great.
- Gravitational Waves: We have very few of them. If you make the buckets too small, they become empty, and the "noise" (static) drowns out the signal.
- The Sweet Spot: You need buckets that are wide enough to catch enough waves to be statistically safe, but narrow enough to keep the distance information sharp.

The Bottom Line

The paper concludes that this "Dark Siren Cross-Correlation" method is robust and powerful.

Even with imperfect data, missing galaxies, or fuzzy distance measurements, the method can still accurately measure the expansion of the universe—provided you build a computer model that accounts for these imperfections from the start. It's a promising new tool for solving the mystery of why the universe is expanding at the rate it is.

Technical Summary: Dark Siren Cross-Correlations and the Sensitivity of $H_0$ to Methodological Choices

Problem Statement
Gravitational wave (GW) sources serve as "standard sirens," providing absolute distance measurements that offer a promising independent avenue for constraining the Hubble constant ( $H_0$ ) and resolving the current tension in cosmological parameters. While "bright sirens" (with electromagnetic counterparts) and "dark sirens" (statistically associating GW localizations with galaxy catalogues) have advanced significantly, dark siren methodologies face challenges related to incomplete galaxy catalogues, systematic biases, and the computational complexity of modeling large-scale structure (LSS).

This paper investigates the robustness of the angular cross-correlation method, which correlates GW events with galaxy catalogues via their shared large-scale structure, against various methodological choices. Specifically, the authors examine how covariance treatment, bias parametrization, distance/redshift binning, and catalogue incompleteness impact the inferred value of $H_0$ . The goal is to determine whether systematic biases can be effectively mitigated through appropriate modeling, thereby validating the cross-correlation approach for future precision cosmology.

Methodology
The authors employ a tomographic angular cross-correlation framework, analyzing the clustering of GW events and galaxies across redshift and luminosity distance bins.

Formalism: The method utilizes the angular cross-power spectrum ( $C_{\ell}$ ) between GW and galaxy tracers. To reduce computational cost and dimensionality, the authors introduce a compressed statistic, $S_{ij}$ , defined as a weighted sum of the cross-power spectrum over multipoles $\ell$ . This statistic effectively measures the line-of-sight cross-correlation between the two tracers.
Simulations: The analysis is based on 600 log-normal mock catalogues generated under a flat $\Lambda$ CDM cosmology ( $H_0 = 70$ km s $^{-1}$ Mpc $^{-1}$ , $\Omega_m = 0.25$ ). These mocks span the redshift range $0 \le z \le 0.5$ and incorporate realistic selection functions, including magnitude cuts and sky localization uncertainties.
Pipeline: Theoretical predictions are computed using CAMB under the Limber approximation, incorporating forward modeling of selection effects (Malmquist and Eddington biases) directly into the projection kernels. Parameter inference is performed via Markov Chain Monte Carlo (MCMC), fitting for $H_0$ and bias parameters ( $b_{gal}, b_{GW}$ ).
Covariance Estimation: The study compares analytical covariance matrices (derived from Gaussian assumptions) against empirical covariance matrices estimated from the 600 mock realisations, applying Hartlap and Percival corrections to account for finite sample sizes.

Key Contributions and Results

Compressed vs. Uncompressed Statistics:
The authors demonstrate that compressing the full angular power spectrum into the summary statistic $S_{ij}$ retains the majority of cosmological information. While using $S_{ij}$ results in a modest loss of constraining power (increasing $H_0$ uncertainty by a factor of $\sim 1.4$ for 200 events and $\sim 1.5$ for 1000 events compared to full $C_{\ell}$ ), it offers significant computational advantages and numerical stability by avoiding direct integration of Bessel functions.
Covariance Modeling:
A critical finding is the sensitivity of parameter constraints to the choice of covariance matrix.
- For large samples ( $\sim 10^3$ events), analytical and mock covariances yield consistent results if off-diagonal terms are included.
- For smaller samples (200 events), the analytical covariance underestimates uncertainties by a factor of $\sim 1.5$ compared to mock-derived covariances. Furthermore, including off-diagonal terms in the analytical covariance for small samples leads to numerical instability in the MCMC inversion, preventing convergence.
- The study concludes that mock-based covariance estimates are essential, particularly for smaller event counts, to avoid overly optimistic constraints.
Systematic Uncertainties:
- Malmquist and Eddington Biases: The paper confirms that luminosity distance uncertainties induce Eddington and Malmquist biases, which can systematically bias $H_0$ high if unmodeled. However, these effects are effectively mitigated by forward modeling the detection probability and selection thresholds within the theoretical kernel.
- Bias Parametrization: The study tests constant, linear, and quadratic parametrizations for galaxy and GW bias. While constant parametrization introduces a bias due to its inability to capture redshift evolution, linear and quadratic forms recover unbiased $H_0$ estimates. This suggests that precise knowledge of bias evolution is not strictly required for near-future observations, provided the parametrization is sufficiently flexible.
- Catalogue Incompleteness: A major advantage identified is the method's robustness to incomplete catalogues. Unlike standard dark siren approaches that require modeling the missing population, the cross-correlation method only requires the selection function to be consistently forward-modeled in the theoretical prediction. This allows the observed overdensity field to be used directly, yielding unbiased $H_0$ constraints even with radial selection effects (e.g., magnitude limits). However, combining catalogues with varying completeness across the sky can degrade constraints by reducing the power to constrain global bias parameters.
Impact of Binning:
The choice of redshift and luminosity distance binning involves a trade-off between signal resolution and shot noise.
- Finer redshift binning significantly improves $H_0$ constraints because galaxy shot noise remains low even with high resolution.
- Finer luminosity distance binning yields diminishing returns for GW events due to their sparsity; subdividing these bins increases shot noise, which can offset the benefits of higher resolution.
- The optimal strategy involves bin widths comparable to or larger than the typical GW distance uncertainty to absorb measurement errors without excessive signal dilution.

Significance and Claims
The paper claims that the angular cross-correlation method is a powerful and complementary tool for dark siren cosmology, capable of achieving unbiased $H_0$ constraints provided that methodological choices are carefully managed. The authors emphasize that:

Systematic biases arising from distance uncertainties, bias evolution, and catalogue incompleteness can be effectively mitigated through forward modeling rather than explicit reconstruction of missing populations.
The use of mock-derived covariance matrices is critical for reliable uncertainty estimation, particularly for smaller event samples where analytical approximations fail.
The compressed statistic $S_{ij}$ offers a computationally efficient alternative to full power spectrum analysis with minimal loss of information.

Ultimately, the work establishes that with appropriate modeling and a sufficiently large sample of precise GW events, the cross-correlation method can robustly address the Hubble tension, offering a distinct advantage over standard galaxy catalogue approaches by naturally incorporating selection effects into the theoretical framework.