🔭 astrophysics

Bounding the Effect of HOD Assumptions on Small-Scale Clustering Constraints

This study demonstrates that the constraining power of small-scale galaxy clustering on cosmological parameters is heavily dependent on the amount of prior information assumed about the galaxy-halo connection, as relaxing HOD assumptions significantly reduces the ability to exclude incorrect cosmologies.

Original authors: Nick Magnelli, Zachery Brown, Lado Samushia

Published 2026-06-11

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Nick Magnelli, Zachery Brown, Lado Samushia

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, three-dimensional puzzle. Astronomers want to solve this puzzle to understand the rules of the cosmos—how much dark matter there is, how fast the universe is expanding, and what dark energy is doing. To do this, they look at how galaxies are clustered together, like looking at how people are grouped at a massive concert.

This paper is about a specific challenge: How much can we learn from the "small talk" of galaxies (their clustering on small scales) if we don't perfectly understand the rules of how galaxies live inside their invisible "homes" (dark matter halos)?

Here is the breakdown of the paper's story, using simple analogies:

1. The Problem: The "House" vs. The "Family"

Think of a dark matter halo as a house, and a galaxy as a family living inside it.

The Cosmologists want to know the rules of the neighborhood (the universe's expansion, dark energy, etc.).
The Galaxy-Halo Connection (HOD) is the rulebook for how many families live in each house, how big the families are, and how they arrange themselves.

The problem is that we don't know the exact rulebook. We have to guess. If we guess wrong about the families, we might think the neighborhood rules are different than they actually are.

2. The Experiment: The "Floor" and the "Ceiling"

The authors wanted to test how much their guesses about the families (the HOD) mess up their understanding of the neighborhood (the cosmology). They used a supercomputer simulation called AbacusSummit to create 81 different versions of the universe, each with slightly different rules.

They tested two extreme scenarios, like setting a "floor" and a "ceiling" for how much they could learn:

The "Ceiling" (The Optimist's Dream):
Imagine you are given the exact rulebook for the families. You know exactly how many people live in every house. In this scenario, you can look at the galaxy clustering and say, "Aha! This specific neighborhood rule is definitely wrong!"
- Result: When they assumed they knew everything about the families, they could rule out 81% of the fake universes they tested. They were very strict.
The "Floor" (The Conservative Realist):
Imagine you have no idea about the families. You only know they exist. You are allowed to try any reasonable arrangement of families to make the data fit. You ask, "Is there any way to arrange the families so that this fake universe looks like our real data?"
- Result: When they allowed the family arrangements to change freely to fit the data, they could only rule out 25% of the fake universes. The "noise" of not knowing the families drowned out the signal of the universe's rules.

3. The Big Discovery

The paper found a massive gap between the "Ceiling" and the "Floor."

Many fake universes that looked totally wrong when you knew the family rules (Ceiling) looked perfectly fine when you were allowed to change the family rules to fit them (Floor).
The Metaphor: It's like trying to identify a song by listening to it through a wall. If you know exactly what the wall is made of (Ceiling), you can tell the song is different. If you don't know what the wall is and can assume it's made of anything (Floor), you can't tell the difference.

The Conclusion: The ability to learn about the universe from small-scale galaxy clustering depends heavily on how well we understand the relationship between galaxies and their dark matter homes. If we don't have strong prior knowledge of that relationship, we lose most of our ability to constrain the universe's rules.

4. The "Shape-Shifter" Problem

The authors also found something weird about the "best fit" family arrangements.

When they tried to find the perfect family arrangement that matched the data, they didn't find just one answer. They found thousands of completely different family arrangements that all produced the exact same galaxy pattern.
The Metaphor: It's like trying to guess the recipe for a cake by tasting the frosting. You might find that 1,000 different combinations of flour, sugar, and eggs could all result in the exact same taste of frosting. You can't say, "This is the only recipe."
Implication: Even if we find a "best fit" model, we can't trust the specific numbers we get for the galaxy families. They are "degenerate," meaning many different answers look the same.

5. What They Didn't Do

They did not apply this to real-world medical data or clinical treatments.
They did not claim to have solved the mystery of dark energy yet.
They did not say we should stop looking at small scales; they just said we need better "family rulebooks" (HOD priors) to make sense of it.

Summary

This paper is a warning label for cosmologists. It says: "Small-scale galaxy clustering contains a treasure trove of information about the universe, but that treasure is locked behind a door. The key is our understanding of how galaxies live in dark matter. If we don't have a good key (strong prior knowledge), we can't get in, and the information remains useless."

They built a tool (called HODmin) to help find the best possible "key" within a wide range of guesses, but they proved that without a good starting guess, the lock remains too hard to pick.

Technical Summary: Bounding the Effect of HOD Assumptions on Small-Scale Clustering Constraints

Problem Statement
While small-scale galaxy clustering (scales $r \in [5, 80]$ Mpc/h) contains substantial cosmological information beyond the standard Baryon Acoustic Oscillation (BAO) and Redshift-Space Distortion (RSD) regimes, extracting this information robustly is complicated by the necessity of modeling the galaxy–halo connection. Unlike large-scale analyses where perturbative theory suffices, small-scale modeling requires non-perturbative approaches that introduce nuisance parameters, specifically the Halo Occupation Distribution (HOD). The extent to which cosmological constraints on small scales depend on the assumed priors for these HOD parameters remains unclear. This paper investigates how much constraining power is lost when HOD parameters are treated conservatively (profiled over broad bounds) versus when they are assumed to be known exactly.

Methodology
The authors utilize the AbacusSummit suite of high-resolution N-body simulations, analyzing 81 distinct $w_0wa$ CDM cosmologies. They generate mock luminous red galaxy (LRG) catalogs using a standard five-parameter HOD model (5pHOD) implemented via the AbacusHOD algorithm.

To address the computational cost of direct simulation evaluation during optimization, the authors construct interpolated 2-point correlation function (2PCF) multipoles (monopole and quadrupole) in HOD parameter space (the "5pIHOD" model). They introduce HODmin, a custom two-stage global optimization algorithm designed to navigate the non-smooth, multi-minima $\chi^2$ surfaces characteristic of simulation-based HOD modeling. HODmin employs adaptive pseudorandom domain compression (Stage 1) followed by local Nelder–Mead minimization (Stage 2) to find the global minimum $\chi^2$ for a given cosmology.

The study defines two limiting cases to bound the constraining power:

The "Floor" (Conservative): The HOD parameters are profiled over broad initial bounds (with modest filters on number density and satellite fraction). A cosmology is considered viable if any allowed HOD parametrization yields an acceptable fit to the data.
The "Ceiling" (Optimistic): The HOD parameters are assumed to be known exactly (fixed to the values used to generate the mock data vector). No profiling is performed; the $\chi^2$ is computed directly.

The analysis compares these cases against a fiducial mock data vector generated from the Planck $\Lambda$ CDM cosmology (c000) at phase ph000. The study validates the convergence of HODmin using self-consistent mocks where the expected minimum $\chi^2$ is near zero.

Key Contributions

HODmin Algorithm: Development and validation of a two-stage global optimization algorithm capable of efficiently minimizing $\chi^2$ in high-dimensional, non-smooth HOD parameter spaces derived from simulations.
Quantification of Prior Dependence: A systematic comparison of the "floor" and "ceiling" constraints across 81 cosmologies, explicitly isolating the impact of HOD assumptions on small-scale clustering constraints.
Degeneracy Analysis: Demonstration that even when a cosmology fits the data well, the best-fit HOD parameters are not uniquely determined, revealing substantial degeneracies in HOD space that prevent the extraction of a unique "summary" HOD model from 2PCF multipoles alone.

Results

Significant Discrepancy in Constraints: There is a wide gap between the floor and ceiling constraints. When comparing against the same Planck $\Lambda$ $Λ$ CDM mock data vector:
- In the ceiling case (known HOD), approximately 81% of the tested non-fiducial cosmologies are excluded at the $3\sigma$ level.
- In the floor case (profiled HOD), only 25% of the cosmologies are excluded at the $3\sigma$ level.
Cosmological Mismatch vs. HOD Flexibility: Many cosmologies that are strongly excluded in the ceiling case (by multiple orders of magnitude in $\chi^2$ ) become viable in the floor case. This indicates that the differences in 2PCF multipoles caused by varying cosmological parameters can be largely absorbed by adjusting the HOD parameters within broad bounds.
Sensitivity to Analysis Choices:
- Scale Cuts: Including smaller scales ( $r \in [5, 20]$ Mpc/h) strengthens constraints, even in the conservative floor case.
- Multipole Choice: Fitting both the monopole and quadrupole generally yields tighter constraints than fitting either alone, though the monopole contains the majority of unique cosmological information.
- Phase and HOD Realization: Constraints are sensitive to the specific phase realization of the mock and the specific HOD realization used to generate the data, though these effects are secondary to the floor-ceiling gap.
Parameter Degeneracy: In self-consistent tests, the authors find that 1,000 distinct best-fit HOD parameter sets can produce nearly indistinguishable 2PCF multipoles with $\chi^2_\nu \ll 1$ . This confirms that the 2PCF multipoles alone cannot uniquely identify the HOD model, even if the cosmology is known.

Significance and Claims
The paper concludes that the strength of cosmological constraints derived from small-scale galaxy clustering is heavily dependent on the amount of prior information assumed regarding the galaxy–halo connection. The authors state that while small-scale clustering contains substantial cosmological information, its robust extraction is currently limited by the inability to uniquely constrain HOD parameters.

The "floor–ceiling" bracket demonstrates that without informative galaxy–halo priors, the constraining power of small-scale clustering is significantly weakened, as the flexibility of the HOD model can mimic cosmological variations. Conversely, assuming the HOD is perfectly known yields unrealistically strong constraints. The authors assert that realistic analyses with informative but imperfect HOD priors will likely yield constraints falling between these two limits. Consequently, the paper argues that informative galaxy–halo priors are necessary to extract strong small-scale clustering constraints, and that current methods relying solely on profiling over broad HOD bounds may severely underestimate the true cosmological information content or fail to rule out viable alternative cosmologies.