🔭 astrophysics

Galaxy assembly bias in cosmological hydrodynamical simulations: a comparison between SIMBA and IllustrisTNG

This study compares galaxy assembly bias in SIMBA and IllustrisTNG simulations, revealing that while both show increasing bias from high to intermediate redshifts driven primarily by cosmic overdensity rather than tidal anisotropy, their low-redshift behaviors diverge significantly, with TNG maintaining strong bias while SIMBA's bias diminishes, prompting a new parameterization for density-dependent halo occupation distributions.

Original authors: Hong-Gang Yang, Marcos Pellejero, John Peacock, Romeel Davé

Published 2026-03-11

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Hong-Gang Yang, Marcos Pellejero, John Peacock, Romeel Davé

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

The Big Picture: Are Neighbors Important?

Imagine you are trying to predict how many people live in a specific house. The simplest rule you could use is: "The bigger the house, the more people live in it." In the world of astronomy, this is the standard rule. Astronomers usually assume that the number of stars in a galaxy depends almost entirely on the size of the invisible "dark matter" bubble (called a halo) that holds it together.

But this paper asks a tricky question: Is that rule enough?

What if the neighborhood matters too? What if two houses of the exact same size have different numbers of people because one is in a quiet, empty suburb and the other is in a bustling, crowded city center?

This phenomenon is called Galaxy Assembly Bias (GAB). It's the idea that a galaxy's "personality" (how many stars it has, how it clusters with others) isn't just about its own size, but also about where it lives in the cosmic neighborhood.

The Experiment: Two Different Chefs

To figure out if this "neighborhood effect" is real and how strong it is, the authors cooked up a comparison between two massive, state-of-the-art computer simulations of the universe: SIMBA and IllustrisTNG.

Think of these simulations as two different chefs trying to bake the perfect "Universe Cake."

The Ingredients: Both chefs use the same basic recipe (the laws of physics and the Big Bang).
The Secret Sauce: They use different methods to handle the "baryonic physics" (the messy stuff like stars exploding, black holes eating gas, and galaxies blowing winds). SIMBA uses one set of rules for these explosions; TNG uses a slightly different set.

The authors wanted to see: Does the "neighborhood effect" (GAB) taste the same in both cakes?

The Method: The "Shuffle" Game

To measure GAB, the scientists played a game of "Cosmus Shuffle."

The Original Setup: They looked at the galaxies as they naturally formed in the simulation.
The Shuffle: They took all the galaxies and randomly swapped them around. Crucially, they only swapped galaxies into halos (dark matter bubbles) of the same size.
- Analogy: Imagine you have a room full of people of different heights. You tell everyone to swap seats, but you can only sit in a chair that fits your height. You don't care who sits where, just that the chair size matches the person.
The Comparison: After shuffling, they measured how clumpy the galaxies were.
- If the "neighborhood" didn't matter, the shuffled galaxies would look exactly the same as the original ones.
- If the "neighborhood" did matter, the shuffled galaxies would look different (less clumpy), because you broke the special connection between a galaxy and its specific neighborhood.

The Results: A Tale of Two Universes

Here is where the story gets interesting. The two simulations agreed at first, but then went in opposite directions.

1. The Early Universe (High Redshift):
At the beginning of time (around 13 billion years ago), both simulations showed almost zero neighborhood effect. Galaxies were just forming, and their size was the only thing that mattered.

2. The Middle Era (Around 10 billion years ago):
Both simulations started to show a small neighborhood effect (about 5%). Galaxies in crowded areas started to look slightly different than those in empty areas.

3. The Divergence (The Present Day):
This is the big surprise. As time moved forward to the present day:

In SIMBA: The neighborhood effect disappeared. It went back to zero. It's as if the SIMBA universe decided that "location, location, location" doesn't matter anymore.
In IllustrisTNG: The neighborhood effect grew stronger, reaching about 10%. In this universe, being in a crowded neighborhood makes a huge difference to how galaxies behave.

Why the difference?
The authors suspect it comes down to Black Holes. In the SIMBA universe, the black holes are like aggressive, steady jets that blow gas away very efficiently, clearing out the neighborhood. In TNG, the black holes are a bit more chaotic. This difference in how they "clean up" their surroundings changes how galaxies form in crowded vs. empty areas.

The Culprit: Density vs. Tides

The authors also tried to figure out what exactly about the neighborhood was causing this. They looked at two factors:

Overdensity (The Crowd): How many other galaxies are nearby?
Tidal Anisotropy (The Stretch): Is the space being stretched or squeezed by gravity?

The Verdict:

The Crowd (Overdensity) is the King. Almost all of the "neighborhood effect" comes from simply being in a crowded area.
The Stretch (Tides) is a Jester. The stretching of space had almost no effect on the results.

The "Low-Mass" Secret

The paper also discovered that this neighborhood effect is mostly driven by small galaxies living in small dark matter bubbles.

Big Halos: If you have a massive dark matter bubble, it's almost guaranteed to be in a crowded area anyway. There's no variation, so no bias.
Small Halos: Small bubbles can be found in both empty deserts and crowded cities. The simulation showed that small bubbles in crowded cities are much more likely to host a galaxy than small bubbles in empty deserts.

Why Should You Care?

This paper is a warning for future astronomers. We are about to launch massive surveys (like the Euclid mission and DESI) that will map millions of galaxies to understand the universe's expansion and dark energy.

If we assume that "House Size = Number of People" and ignore the "Neighborhood Effect," our maps will be slightly wrong. The authors found that ignoring this bias could mess up our measurements by 10%.

The Takeaway:
The universe is a complex place. Just knowing how big a galaxy is, isn't enough. You also need to know if it's living in a "city" or a "desert." And, depending on how the universe's "black hole chefs" cook their gas, that neighborhood effect might be huge or non-existent. We need to understand these recipes perfectly to read the story of the universe correctly.

1. Problem Statement

The spatial distribution of galaxies is a primary observable for cosmology, tracing the underlying dark matter large-scale structure (LSS). Standard models, such as the Halo Occupation Distribution (HOD), assume that the number of galaxies in a dark matter halo depends solely on the halo's mass. However, Galaxy Assembly Bias (GAB) challenges this assumption, positing that galaxy clustering depends on secondary halo properties (e.g., formation time, concentration) and environmental factors (e.g., local overdensity, tidal fields) beyond just mass.

While GAB has been studied extensively in semi-analytical models and the IllustrisTNG simulation, its physical origin remains unclear. Different studies disagree on whether the driving factors are internal halo properties or environmental ones, and whether GAB affects central and satellite galaxies equally. Crucially, there has been no direct comparison of GAB evolution across different state-of-the-art hydrodynamical simulations with distinct baryonic physics prescriptions. This paper aims to:

Quantify the strength and redshift evolution of GAB in two major simulations: SIMBA and IllustrisTNG (TNG).
Determine the extent to which GAB is driven by smoothed matter overdensity ( $\delta_5$ ) versus tidal anisotropy ( $\alpha_5$ ).
Investigate the role of low-mass versus high-mass haloes in generating the GAB signal.
Provide a parameterization to incorporate density-dependent GAB into HOD models.

2. Methodology

The authors utilize the largest available runs of two cosmological hydrodynamical simulations:

SIMBA (m100n1024): Uses the GIZMO code with Meshless Finite Mass (MFM) hydrodynamics. It features specific implementations of galactic winds and AGN feedback (bipolar jets).
IllustrisTNG (TNG300-1): Uses the AREPO moving-mesh code. It features kinetic stellar feedback and multi-mode AGN feedback (thermal, kinetic, radiative).

Key Methodological Steps:

Galaxy Selection: Galaxies are selected by stellar mass to match number densities between the two simulations ( $N \approx 0.043 \, h^3 \text{Mpc}^{-3}$ at $z=0$ ).
Shuffling Technique: To isolate GAB, the authors employ a "shuffling" method. Galaxies are randomly reassigned to haloes of the same mass bin. This removes any dependence of galaxy occupancy on halo properties other than mass.
- Total GAB: Measured as the ratio of the projected correlation function $w_p(r_p)$ of the original sample to the shuffled sample: $[w_p(r_p)/w_p(r_p)_{\text{shuffled}} - 1]$ .
Double Shuffling: To isolate environmental drivers, galaxies are shuffled within bins of halo mass AND a specific environmental variable:
- Smoothed Overdensity ( $\delta_5$ ): Matter density smoothed over $5 \, h^{-1}\text{Mpc}$ .
- Tidal Anisotropy ( $\alpha_5$ ): A measure of the tidal field shape, calculated to be uncorrelated with $\delta_5$ .
HOD Analysis: The Halo Occupation Distribution is measured for haloes in the top and bottom quartiles of $\delta_5$ and $\alpha_5$ to see how environment modulates central and satellite occupation.
Mass-Dependent Shuffling: Shuffling is restricted to low-mass ( $M_h < 10^{11.5} M_\odot$ ) and high-mass haloes separately to determine the mass regime responsible for GAB.

3. Key Contributions

Cross-Simulation Comparison: First direct comparison of GAB evolution between SIMBA and TNG, highlighting how different baryonic physics (specifically AGN feedback) alters GAB trends.
Environmental Decomposition: Quantification of the relative contributions of local overdensity ( $\delta_5$ ) and tidal anisotropy ( $\alpha_5$ ) to the total GAB signal.
Low-Mass Dominance: Demonstration that GAB is almost exclusively driven by low-mass haloes, particularly those near the "turn-on" mass for galaxy formation.
HOD Parameterization: Development of a fitting formula that modifies standard HOD parameters (specifically the slopes of the central and satellite "turn-on") as a linear function of the environment rank, allowing for the generation of GAB-inclusive mock catalogs.

4. Key Results

A. Redshift Evolution of GAB

High Redshift ( $z > 2$ ): Both simulations show negligible GAB at $z=5$ , increasing to $\sim 5\%$ at $z=2$ .
Low Redshift Divergence ( $z < 2$ ):
- SIMBA: GAB declines after $z=2$ , reaching nearly 0% at $z=0$ .
- TNG: GAB continues to rise, reaching $\sim 10\%$ at $z=0$ .
Interpretation: The divergence coincides with the onset of substantial AGN feedback. SIMBA's bipolar jet feedback efficiently suppresses star formation in group-scale haloes and evacuates gas, reducing the environmental dependence of galaxy formation at low redshift. TNG's randomized jet direction and different feedback coupling lead to a stronger persistence of environmental dependence.

B. Environmental Drivers ( $\delta_5$ vs. $\alpha_5$ )

Overdensity ( $\delta_5$ ): Accounts for the majority of the GAB signal in both simulations.
Tidal Anisotropy ( $\alpha_5$ ):
- SIMBA: Contribution is negligible.
- TNG: Contributes a small but non-negligible fraction, though still minor compared to $\delta_5$ .
Conclusion: The primary driver of GAB is the local matter overdensity, not the tidal field shape.

C. HOD and Mass Regime

HOD Variation: In TNG, high- $\delta_5$ environments significantly increase the probability of hosting both central and satellite galaxies in low-mass haloes. In SIMBA, the effect is primarily on central galaxies.
Low-Mass Haloes: Shuffling restricted to low-mass haloes ( $M_h < 10^{11.5} M_\odot$ ) recovers nearly the entire GAB signal. High-mass haloes contribute negligibly.
Mechanism: Low-mass haloes in high-density regions are more likely to host galaxies. High-mass haloes are almost always in dense regions, leaving little scatter in their environment, hence no GAB signal.
Stellar Mass Threshold: Increasing the stellar mass threshold in TNG (to $10^{10} M_\odot$ ) slightly reduces the total GAB amplitude but preserves the trend that $\delta_5$ drives the signal via low-mass haloes.

D. HOD Parameterization

The authors propose a modified HOD model where the slopes ( $\mu$ and $\nu$ ) governing the "turn-on" of central and satellite populations depend linearly on the rank of the local overdensity ( $r_\delta$ ):
$\mu = \mu_{\text{tot}} - C_\delta (r_\delta - 0.5)$
This allows mock catalogs to reproduce the density-dependent clustering seen in hydrodynamical simulations.

5. Significance

Precision Cosmology: GAB can alter galaxy clustering amplitudes by up to 10%. Ignoring this effect in future surveys (e.g., DESI, Euclid) could bias cosmological parameter estimation.
Baryonic Physics Sensitivity: The stark difference in GAB evolution between SIMBA and TNG demonstrates that GAB is highly sensitive to the specific implementation of baryonic physics (particularly AGN feedback). This suggests that GAB could serve as a probe to constrain feedback models.
Modeling Improvement: The provided parameterization offers a practical tool for HOD-based mock generation, moving beyond the assumption that halo mass is the sole determinant of galaxy occupancy.
Observational Tension: The finding that tidal anisotropy plays a minor role in simulations contrasts with some recent observational claims, highlighting a need for further investigation into the empirical nature of assembly bias.

In summary, the paper establishes that galaxy assembly bias is a real, significant, and environment-driven phenomenon, primarily affecting low-mass haloes in high-density regions, with its magnitude and evolution heavily dependent on the specific baryonic physics models used in cosmological simulations.