🔭 astrophysics

Web-Halo Model Peak-Background Split (WHM-PBS): halo bias as a distribution, not a number

The Web-Halo Model Peak-Background Split (WHM-PBS) proposes that large-scale halo bias is a skewed distribution inherited from the cosmic-web environment rather than a fixed number, providing a parameter-free analytic framework that successfully predicts bias relations, stochasticity trends, and assembly bias effects while improving constraints on synthetic data.

Original authors: Samuel Brieden, Alexander Tipp

Published 2026-08-25

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Samuel Brieden, Alexander Tipp

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 universe is not a smooth, uniform soup of matter. Instead, it is a vast, intricate network known as the cosmic web, where galaxies and dark matter clump together in a hierarchy of shapes: flat sheets, long filaments, and dense knots called haloes. These haloes are the gravitational cradles where galaxies form, and understanding how they cluster together is essential for cosmologists trying to map the history and structure of the cosmos. For decades, scientists have relied on a simplified rule to describe this clustering: they assumed that the tendency of a halo to clump, known as its "bias," was a single, fixed number determined solely by its mass. A heavy halo was thought to have one specific bias, and a light halo another, with no variation in between. This approach treated the universe as a place where mass alone dictated destiny, ignoring the complex environments in which these cosmic structures actually live.

A new study by Samuel Brieden and Alexander Tipp challenges this long-held view, proposing that a halo's bias is not a single number but a wide, skewed distribution. The researchers developed a theory called the Web–Halo Model Peak–Background Split, which treats the bias of a halo as something it inherits from its surroundings. In their picture, a halo does not form in isolation; it sits inside a filament, which itself is embedded within a larger sheet. Because a halo of a specific mass can end up in many different filaments of varying sizes, it inherits a range of different biases from these potential hosts. The authors used advanced mathematical tools to calculate the probability of a halo occupying any given host, revealing that the "bias" is actually a cloud of possibilities rather than a point. This shift in perspective transforms how we understand the relationship between matter and the structures it forms.

The team's work shows that while the average bias still follows the familiar trend of increasing with mass, the spread around that average is significant and physically meaningful. They found that this spread is not random noise but a direct consequence of the cosmic web's structure. By treating the bias as a distribution, the researchers could predict how haloes cluster in ways that match complex computer simulations of the universe. Specifically, they demonstrated that this distribution explains why haloes with the same mass can behave differently depending on their formation history and environment, a phenomenon known as assembly bias. Their model successfully predicted a specific reversal in the relationship between a halo's concentration and its clustering strength, a feature that had been observed in simulations but lacked a clear theoretical explanation until now.

Beyond explaining existing observations, the study provides a new toolkit for interpreting data from modern sky surveys. The researchers showed that their predicted distribution of biases can be used to set strict, physically motivated limits on the uncertainties in cosmological measurements. In current analyses, scientists often have to guess at these uncertainties, which can lead to errors in determining the expansion history of the universe. By using the distribution derived from the cosmic web itself, the authors demonstrated that these errors can be significantly reduced. Their approach effectively removes the "projection effects" that occur when too many unknown variables are allowed to vary freely, leading to more precise and reliable maps of the cosmos.

The study also tackled the issue of "stochasticity," or the random, shot-noise-like fluctuations in how haloes are distributed. The authors showed that the spread in inherited biases acts as a source of this randomness, creating a pattern of clustering that is slightly stronger than simple randomness would predict. When they combined this with the fact that haloes cannot overlap with one another, their model reproduced the exact trend of randomness seen in simulations, moving from a state of excess clustering to a state of suppressed clustering as haloes become more massive. This success suggests that the random variations in the cosmic web are not just background noise but a fundamental feature that carries information about the environment.

Ultimately, this work redefines the bias of a dark matter halo from a static property into a dynamic one, shaped by the hierarchy of the cosmic web. The researchers did not just find a new number; they found a new way of seeing the universe, where the environment is as important as the object itself. Their theory connects the large-scale structure of the cosmos to the small-scale details of halo formation, offering a unified explanation for several puzzling observations. By providing a framework that links the shape of the cosmic web to the behavior of galaxies, this study offers a clearer path forward for the next generation of cosmological experiments, ensuring that the maps we draw of the universe are as accurate as the data allows.

Technical Summary: Web–Halo Model Peak–Background Split (WHM-PBS)

Problem Statement
Standard analytic models of halo bias typically treat the bias parameters ( $b_1, b_2, b_s2, \dots$ ) as deterministic functions of halo mass, $b(M_h)$ . However, N-body simulations reveal that at fixed mass, haloes exhibit "assembly bias," clustering differently based on secondary properties like formation time, concentration, and environment. Consequently, the bias of a halo is not a single number but the mean of a distribution with significant intrinsic scatter. Current Effective Field Theory (EFT) analyses of large-scale structure often treat these bias coefficients as free nuisance parameters with broad priors, leading to projection effects that can bias cosmological inference, or fix them to potentially inaccurate values. There is a need for an analytic framework that predicts not only the mean bias relations but also the distribution (scatter) of these biases and their stochastic consequences from first principles.

Methodology
The authors develop the Web–Halo Model Peak–Background Split (WHM-PBS), an analytic theory extending the excursion-set formalism. The core methodology involves:

Cosmic-Web Excursion Set: The model utilizes the moving-barrier formalism (Shen et al.) to describe the hierarchical collapse of matter. Instead of a single constant barrier (spherical collapse), it employs morphology-dependent moving barriers for sheets, filaments, and haloes. A random walk in the smoothed density field first crosses the sheet barrier, then the filament barrier, and finally the halo barrier, establishing a nested hierarchy: a halo sits inside a filament, which sits inside a sheet.
Conditional Mass Functions: Using the exact Zhang–Hui (ZH) integral equation (rather than truncated Taylor series approximations), the authors compute the conditional mass functions $P(M_{host}|M_h)$ . This quantifies the probability distribution of host environments (filament or sheet mass) for a halo of a given mass $M_h$ .
Environment-Averaged Bias: The central theoretical innovation is the definition of halo bias as the environment-averaged response. The bias of a halo is the bias of its host environment, averaged over the conditional mass function:
$b(M_h) = \langle b(M_{host}) \rangle_{M_{host}|M_h}$
Because the host mass $M_{host}$ is a random variable at fixed $M_h$ , the resulting bias is a distribution, not a number.
Bias and Stochasticity Derivation:
- Density Biases ( $b_1, b_2, b_3$ ): Derived as the response of the abundance to a long-wavelength density background. The model shows these relations are "tight" (universal) when expressed in terms of $b_1$ , with scatter arising only from the aggregation of samples along the host curve.
- Tidal Bias ( $b_{s2}$ ): Derived as the response to a long-wavelength tidal shear. Unlike density biases, the tidal response depends on the specific halo-host pair (due to the barrier slope), resulting in a "broken" relation with genuine scatter at fixed $b_1$ .
- Stochasticity: The scatter in the bias distribution is interpreted as a random field, sourcing a super-Poisson "clustering" contribution to the halo power spectrum. This is combined with a sub-Poisson "exclusion" term to model the full stochasticity matrix.
- Assembly Bias: The correlation between the host environment and halo internal structure (e.g., concentration) is modeled as population mixing. The model predicts the sign inversion of the bias–concentration correlation near the characteristic mass $m^*$ .

Key Contributions and Results

Bias as a Distribution: The paper demonstrates that halo bias is intrinsically a skewed distribution. The mean of this distribution reproduces standard large-scale bias predictions (though with a known $\sim18\%$ amplitude offset relative to Tinker et al. calibrations due to the barrier choice), while the width ( $\sigma \approx 0.6-1.0$ ) represents the physical scatter of environments.
Universal Relations and Priors:
- Density Sector: The relations $b_2(b_1)$ and $b_3(b_1)$ are found to be tight and universal. The model provides a "selection prior band" derived from the aggregation rule, which encompasses N-body measurements.
- Tidal Sector: The relation $b_{s2}(b_1)$ is genuinely broad. The model calibrates a prior band for this sector by combining the selection uncertainty with an assembly-bias term, constrained by simulation measurements of concentration/spin splits (Lazeyras et al.), limiting the shift to $\lambda_S \approx 0.35 \sigma_{\text{band}}$ .
Stochasticity Prediction: The model predicts the full stochasticity matrix (diagonal and off-diagonal) without free parameters. It successfully reproduces the transition from super-Poisson (clustering-dominated) to sub-Poisson (exclusion-dominated) noise observed in simulations (Baldauf et al.), attributing the super-Poisson component directly to the host bias scatter.
Assembly Bias Inversion: The model analytically predicts the inversion of the bias–concentration correlation near $M_h \simeq 1.7 \times 10^{13} h^{-1} M_\odot$ ( $b_1 \simeq 1.5$ ) as a consequence of the median host mass ratio crossing a turnaround threshold ( $M_f/M_h \approx 1.76$ ), without tuning parameters to assembly bias.
EFT Inference Demonstration: In a synthetic real-space EFT inference (DESI-like), the authors show that using WHM-PBS priors mitigates projection effects (biases in cosmological parameters like $\Omega_m$ and $h$ ) that arise when nuisance parameters are left free with broad priors. Conversely, fixing nuisance parameters to values $1\sigma$ away from the truth introduces significant biases, whereas the WHM-PBS bands absorb such errors.

Significance and Claims
The authors claim that WHM-PBS provides the first analytic, first-principles prediction of halo bias relations, their scatter, and the associated stochasticity. The significance lies in shifting the paradigm from treating bias as a set of fixed numbers to treating it as a distribution derived from the cosmic web hierarchy.

Physical Priors: The derived distributions serve as physically motivated priors for EFT analyses, tightening constraints and reducing systematic errors compared to broad priors or fixed values.
Unified Framework: The model unifies the treatment of bias, stochasticity, and assembly bias as different manifestations of the same underlying host-environment distribution.
Robustness: While the absolute amplitude of the linear bias is over-predicted (a known limitation of the specific barrier calibration), the relations between bias parameters ( $b_N(b_1)$ ) and the scatter are robust and cosmology-universal, making them suitable for constraining cosmology where $b_1$ is marginalized.

The paper concludes that the width of the bias distribution is as physically predictive as the mean, offering a new tool for interpreting large-scale structure surveys. The code is publicly available to reproduce these results.

Technical Summary: Web–Halo Model Peak–Background Split (WHM-PBS)

More like this