⚛️ general relativity

The impact of precession and higher-order multipoles for gravitational wave cosmological inference

This paper demonstrates that for inferring the Hubble constant via the mass spectrum method using current and near-future gravitational-wave data, employing simpler models that omit spin precession and higher-order multipoles yields comparable results to more complex models while reducing computational costs by a factor of six.

Original authors: Charlie Hoy, Konstantin Leyde

Published 2026-07-08

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Charlie Hoy, Konstantin Leyde

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 is a giant, expanding balloon. Astronomers want to know exactly how fast this balloon is inflating (a rate called the Hubble constant, or $H_0$ ). One of the newest ways to measure this is by listening to the "chirps" of colliding black holes, which act like cosmic lighthouses.

However, to measure the distance to these lighthouses, scientists need a very specific "rulebook" (a mathematical model) to translate the sound of the collision into a distance. The problem is, these rulebooks are getting incredibly complex. They are trying to include every tiny detail of how black holes spin and wobble, which makes the calculations take forever on supercomputers.

The Big Question:
Does adding all these fancy, complicated details to the rulebook actually help us measure the expansion of the universe more accurately? Or is it just extra work for no real gain?

The Paper's Answer:
The authors of this paper say: It's mostly extra work.

Here is the breakdown using simple analogies:

1. The "Fancy Map" vs. The "Simple Map"

Imagine you are trying to guess how far away a friend is by listening to their voice.

The Simple Model (IMRPhenomXAS): This is like using a basic map that assumes your friend is walking in a straight line. It's fast to calculate.
The Fancy Model (IMRPhenomXPHM): This is like a high-tech map that accounts for your friend spinning around, wobbling, and shouting in different directions (spin precession and higher-order multipoles). It's much slower to compute.

The paper tested both maps using real data from black hole collisions and also created a "worst-case scenario" simulation where the black holes were spinning wildly and had very uneven masses (the conditions where the fancy map should matter most).

2. The "Worst-Case" Test

To be sure, the scientists didn't just look at average black holes. They simulated a population of black holes that were designed to be difficult:

They were spinning fast and out of alignment (like a top wobbling).
They had very different sizes (one huge, one small).

In this chaotic scenario, the "Fancy Map" did give slightly better details about the specific distance and angle of a single black hole collision. It broke some "blind spots" in the data.

However, when they used these results to calculate the expansion rate of the universe ( $H_0$ ), the difference between the Simple Map and the Fancy Map vanished. The final answer for the universe's expansion rate was almost identical, regardless of which map they used.

3. The "Noise" Factor

Why didn't the fancy details help? The paper explains that the "noise" in the data (statistical uncertainty) is currently so loud that it drowns out the subtle improvements the fancy models offer.

Analogy: Imagine trying to hear a whisper in a rock concert. Even if you use a super-sensitive, expensive microphone (the Fancy Model) to hear the whisper, the result is the same as using a cheap microphone (the Simple Model) because the music (statistical noise) is just too loud. The extra clarity doesn't change the final message you hear.

4. The Cost-Benefit

The authors found a massive difference in cost:

The Simple Model was six times faster to run than the Fancy Model.
Since the final result for the universe's expansion rate was the same, the paper argues that for the next few years, we should stick to the simpler, faster models.

Summary

The paper concludes that while including complex physics (like spinning and wobbling black holes) makes our understanding of individual black hole collisions more precise, it does not significantly improve our measurement of the universe's expansion rate right now.

Because the universe is currently "noisy" with statistical uncertainty, using the simpler, cheaper models is just as effective for cosmology and saves a tremendous amount of computing power. This is crucial because, in the near future, we expect to detect thousands of these events, and we won't have the computing power to run the most complex models on all of them.

One Caveat: The authors note that this logic applies to the "Mass Spectrum Method" (using the population of black holes). If we ever detect a black hole collision that also has a visible light flash (an electromagnetic counterpart), then the fancy models would become important again. But for the current method of just listening to the waves, simple is best.

Technical Summary: The Impact of Precession and Higher-Order Multipoles for Gravitational Wave Cosmological Inference

Problem Statement
Gravitational wave (GW) astronomy offers a pathway to independently measure the Hubble constant ( $H_0$ ) using the "mass spectrum" (or spectral siren) method. This technique infers $H_0$ by combining measured detector-frame masses with the assumed source-frame mass distribution of binary black hole (BBH) populations, bypassing the need for electromagnetic counterparts. However, this inference relies on theoretical waveform models to estimate source properties (such as luminosity distance and component masses). Current models are incomplete descriptions of General Relativity; specifically, many analyses utilize simplified models that exclude spin-induced orbital precession and higher-order multipole moments. While including these physical effects breaks degeneracies (particularly between distance and inclination) and improves single-event parameter estimation, it significantly increases computational cost. The central question addressed is whether the exclusion of these complex physical phenomena introduces systematic biases in the hierarchical inference of $H_0$ that outweigh the benefits of using computationally cheaper, simpler models.

Methodology
The authors employ a hierarchical Bayesian framework to compare cosmological inferences derived from different waveform models. The study utilizes two primary datasets:

Re-analysis of GWTC-3: The authors re-analyze 42 BBH mergers from the third Gravitational Wave Transient Catalog (GWTC-3). They utilize posterior samples generated with four distinct Phenom waveform models:
- XAS: Aligned spins, dominant quadrupole only.
- XP: Misaligned spins (precession), dominant quadrupole.
- XHM: Aligned spins, higher-order multipoles (up to $\ell < 5$ ).
- XPHM: Misaligned spins, higher-order multipoles (the most complete model).
Simulated Populations: To test the limits of current conclusions, the authors simulate a "worst-case" population of 80 BBHs with high spin magnitudes (mean $a_i=0.7$ ) and preferentially asymmetric mass ratios ( $\beta_q = -1.1$ ), conditions expected to maximize the impact of precession and higher-order modes. They also simulate an astrophysically motivated low-spin population. All signals are injected into idealized Gaussian noise representing the Advanced LIGO and Virgo network sensitivities.

For both datasets, single-event parameter estimation is performed using the bilby package with the Dynesty sampler. These single-event posteriors are then fed into the icarogw pipeline to perform hierarchical inference on the hyperparameters, including $H_0$ , the mass distribution power-law indices, and the redshift evolution. The statistical agreement between posteriors derived from different models is quantified using the Jensen-Shannon Divergence (JSD).

Key Results

GWTC-3 Re-analysis: The inferred posterior distribution for $H_0$ shows negligible dependence on the inclusion of spin-precession or higher-order multipoles. The JSD between the most complex model (XPHM) and simpler models (XP, XHM) is approximately $4 \times 10^{-3}$ bits and $5 \times 10^{-4}$ bits, respectively. These values are well below the threshold ( $5 \times 10^{-2}$ bits) for statistical significance and are comparable to variations caused by random seeds. While minor differences appear in population hyperparameters (e.g., the maximum black hole mass $m_{max}$ and mass ratio power-law $\beta_q$ ), the 90% credible intervals for $H_0$ remain consistent across all models.
Simulated High-Spin Population: Even when analyzing a population specifically engineered to exhibit strong precession and higher-order multipole effects (where single-event analyses show significant differences in distance and mass ratio posteriors), the hierarchical inference of $H_0$ $H_{0}$ remains largely agnostic to the waveform model. All models recover the fiducial $H_0$ $H_{0}$ within the $1\sigma$ $1 σ$ credible interval.
- Note on Mass Ratio: The authors observe that neglecting precession biases the inferred mass ratio distribution ( $\beta_q$ ) toward more symmetric binaries, consistent with single-event degeneracies. However, this bias in the mass distribution does not propagate into a significant bias in the inferred $H_0$ .
- Note on Higher-Order Modes: Neglecting higher-order multipoles has a negligible impact on both $H_0$ and the mass distribution in this population, provided the number of events is not excessively large.
Computational Efficiency: The computational cost of analyzing the 80 simulated events with XAS (aligned, quadrupole only) was approximately 30 CPU days, compared to 180 CPU days for XPHM. This represents a six-fold reduction in computational cost for comparable $H_0$ constraints.
Waveform Systematics: A limited comparison between the Phenom (XPHM) and SEOBNR (SEOBNRv5PHM) families for six high-precession signals showed remarkable agreement in marginalized posteriors for mass and distance, suggesting that waveform family choice is not currently a limiting factor for $H_0$ inference.

Significance and Claims
The paper concludes that for current and near-future GW event numbers (up to $\sim$ 1000 events), there is no significant advantage in using computationally expensive models that include spin-precession and higher-order multipoles for the specific purpose of inferring $H_0$ via the mass spectrum method. The statistical uncertainties inherent in individual GW analyses dominate over the systematic errors introduced by missing physics in the waveform models.

The authors argue that while complex models are necessary for breaking degeneracies in single-event parameter estimation (yielding tighter constraints on distance and inclination), the "exact single-event posterior shape" is not critical when aggregating data to estimate the population-level $H_0$ . Consequently, using simpler, less accurate models (such as XAS) allows for comparable cosmological inferences at a fraction of the computational cost. This finding is particularly relevant as next-generation detectors are expected to detect thousands of events, where the computational burden of full-physics models may become prohibitive. The paper notes, however, that this conclusion applies specifically to the mass spectrum method; if direct electromagnetic counterparts are available (the "bright siren" method), accurate spin-precession models remain important for precise distance estimation.

1. The "Fancy Map" vs. The "Simple Map"

2. The "Worst-Case" Test

3. The "Noise" Factor

4. The Cost-Benefit

Summary

More like this