For decades, cosmologists modeled the large-scale distribution of galaxies using a featureless power law. It was a simple, elegant description, but it lacked physical insight. When large-scale galaxy surveys matured, they revealed distinct inflections and departures from this power law. Galaxies, it turned out, do not trace matter in a simple, linear fashion.
To understand why galaxies cluster the way they do, we need a physical bridge between the visible universe and the invisible scaffolding of dark matter. The Halo Occupation Distribution (HOD) framework provides this bridge. Instead of simulating computationally expensive gas physics and star formation across billions of light-years, HOD uses a probabilistic approach: it models the probability $P(N|M)$ that a dark matter structure (a "halo") of a given mass $M$ hosts $N$ galaxies of a specific type.
The Core Intuition: One-Halo vs. Two-Halo Scales
To see why HOD works, consider the galaxy two-point correlation function, which measures the excess probability of finding a pair of galaxies at a given separation distance. As shown by Zehavi et al. (2004), the observed correlation function departs significantly from a pure power law. HOD explains this departure by decomposing galaxy pairs into two distinct physical regimes:
- The One-Halo Term (Small Scales): At short distances, pairs of galaxies reside within the same dark matter halo. The clustering signal here is dictated by how galaxies are spatially distributed inside individual halos.
- The Two-Halo Term (Large Scales): At large distances, pairs of galaxies reside in different halos. The clustering signal here is driven entirely by how the host dark matter halos themselves cluster across the cosmic web.
The inflection point where the power law breaks is simply the transition zone between these two regimes. By modeling how galaxies occupy halos, HOD naturally unifies these small-scale and large-scale behaviors into a single physical framework.
The Central-Satellite Split
The core mathematical mechanism of HOD relies on a crucial distinction: separating a halo's galaxy population into a single central galaxy at the core and a surrounding population of satellite galaxies in the periphery ($N = N_{\text{cen}} + N_{\text{sat}}$).
This decomposition is not just a labeling convenience; it reflects fundamentally different physical processes:
- Centrals follow sub-Poissonian statistics. A dark matter halo must reach a minimum mass threshold before it can host even a single luminous galaxy. Once it crosses this threshold, the probability of hosting a central galaxy jumps rapidly from 0 to 1, acting like a step function. You cannot have two central galaxies in a single isolated halo.
- Satellites follow a Poisson distribution. Once a central galaxy is established, any additional mass gained by the halo allows it to attract or retain satellite galaxies. The number of satellites scales roughly as a power law with halo mass. Lin et al. (2009) demonstrated that treating satellite counts as independent, Poisson-distributed events drastically simplifies the statistical modeling of massive structures like Luminous Red Galaxies (LRGs) in clusters.
This mass-dependent occupancy explains why galaxy clustering varies so strongly with luminosity and color. As Zehavi et al. (2005) and Zehavi et al. (2011) demonstrated, highly luminous, red galaxies cluster far more strongly than faint, blue ones. HOD provides the intuition: brighter galaxies require more massive host halos to form. Because massive halos are rarer and naturally form within the densest nodes of the cosmic web, the galaxies residing within them inherit this heightened clustering. Ishikawa et al. (2021) reinforced this by showing that a halo's mass, rather than a galaxy's intrinsic stellar mass, is the primary driver governing galaxy quenching and evolution.
From Simulations to Cosmological Inference
To extract cosmological parameters from universe-spanning surveys, theorists must match observational data with N-body dark matter simulations. However, these simulations only output dark matter particles and halos, not visible galaxies.
HOD acts as the translation layer. By populating simulated dark matter halos with mock galaxies according to HOD parameters, researchers can generate synthetic universes and tune them until they match real-world observations.
[N-Body Simulation] ---> [Dark Matter Halos (Mass M)]
|
v <--- Apply HOD: P(N|M)
[Observed Survey Data] <--- [Mock Galaxy Catalog]
This technique has become essential for modern precision cosmology:
- Emulators: Wibking et al. (2018) and Wibking et al. (2019) developed simulation emulators that simultaneously vary cosmological parameters (like the amplitude of matter clustering, $\sigma_8$) and HOD parameters. This allows them to exploit deeply non-linear scales and galaxy-galaxy lensing data to break degeneracies between galaxy bias and underlying cosmology.
- Forward Modeling: Instead of relying on analytical approximations, Nadler et al. (2020) used a forward-modeling HOD approach to map faint Milky Way satellite galaxies to dark matter subhalos, successfully constraining galaxy formation efficiency and accounting for the gravitational disruptions caused by massive neighbors like the Large Magellanic Cloud.
- Isolating Cosmological Signals: As surveys push to higher redshifts, integrated modeling frameworks—such as those discussed by Guo et al. (2013)—become vital. For instance, Hadzhiyska et al. (2024) showed how adapting HOD models for specific survey targets allows researchers to disentangle galaxy assembly effects from pristine cosmological signals like primordial non-Gaussianity.
Beyond Mass: The Next Generation of HOD
While standard HOD assumes that a halo's mass alone dictates its galaxy content, real-world astrophysics is more nuanced. Modern iterations of the framework extend the model to incorporate secondary physical properties, a direction often called generalized HOD.
One major refinement is assembly bias—the reality that two halos of the identical mass can have different clustering behaviors based on their formation history or local environment. To capture these effects, Yuan et al. (2018) introduced flexible frameworks like GRAND-HOD, which incorporate satellite velocity bias and assembly bias to resolve tensions in higher-order statistics like the three-point correlation function.
Another critical concept is central-satellite conformity. Standard models assume satellite occupancy is independent of the central galaxy's state. However, Rocher et al. (2023) demonstrated that for specific populations like Emission Line Galaxies (ELGs), the presence of an active central galaxy physically alters the likelihood and spatial distribution of surrounding satellites.
To make these complex, multi-parameter models computationally feasible over millions of simulated structures, researchers have shifted from analytical profile-fitting to particle-based methods. Frameworks like AbacusHOD, developed by Yuan et al. (2021), directly associate mock galaxies with individual dark matter particles within the simulation. By bypassing rigid analytical profiles, these modern implementations—alongside dynamic models incorporating environmental quenching and satellite motions like those by Favole et al. (2026)—allow the HOD framework to maintain its intuitive simplicity while capturing the full, messy reality of galaxy assembly.
Key Foundational Work
On Departures from a Power Law in the Galaxy Correlation Function (2004) — contributed 4 concept elements to this direction.
The Luminosity and Color Dependence of the Galaxy Correlation Function (2005) — contributed 1 concept elements to this direction.