Luminosity-Dependent Assembly Bias of Central Galaxies from Weak Lensing and Clustering

TL;DR

This work provides observational evidence for luminosity-dependent halo assembly bias by leveraging the DESI Legacy Imaging Surveys and spectroscopic BCGs to perform lensing and clustering analyses with a mass-dependent halo-bias model. By separating mass effects from assembly history, the authors find that brighter central galaxies reside in halos that form earlier and exhibit weaker large-scale clustering at fixed mass, with a relative-bias signal $r_{b0}=0.89\pm0.06$ at $\sim3.2\sigma$. The results are independently supported by large hydrodynamical simulations MTNG and FLAMINGO, which reproduce the same qualitative HAB trends, strengthening the link between galaxy luminosity and halo formation history. The findings have important implications for modeling the galaxy–halo connection and for cosmological inferences that neglect halo assembly bias, motivating future work with larger spectroscopic samples and improved simulations.

Abstract

Assembly bias, which is the variation in halo clustering at fixed mass driven by formation history, has long been predicted by numerical simulations but remains difficult to confirm observationally. Previous studies have reported evidence for halo assembly bias by dividing samples according to galaxy stellar mass using various methods. In this work, we present observational measurements of halo assembly bias based on the luminosity of spectroscopically confirmed brightest cluster galaxies (BCGs). Using cluster catalogs and shear measurements from the DESI Legacy Imaging Surveys, we employ a mass-dependent halo-bias model to disentangle halo bias from its underlying mass dependence in galaxy-galaxy lensing and clustering measurements. We confirm that brighter BCGs reside in more concentrated halos yet are less strongly clustered on large scales, with a relative bias ratio deviating from unity at the $\sim 3σ$ level, providing clear evidence of assembly bias. The results are further supported by the FLAMINGO and MillenniumTNG hydrodynamical simulations, in which we directly detect assembly bias signals consistent with the observations, thereby strengthening the connection between galaxy luminosity and halo formation history.

Figures (6)

• Figure 1: Spectroscopic completeness of the BCG sample as a function of redshift $z$ and absolute $z$-band magnitude $M_z$. The color scale indicates the fraction of BCGs with spectroscopic redshifts, $N_{\mathrm{spec}}/N_{\mathrm{total}}$. The red solid and dashed lines mark the threshold at $-22.45$ and $-22.7$ respectively for two subsamples.
• Figure 2: Left panels: ESD profiles for the low-$L$ (blue) and high-$L$ (red) BCGs subsamples. The solid black lines show the best-fit models, while the colored dashed lines represent different model components. Right panels: Posterior distributions of five free parameters and their 68% and 95% confidence contours in the ESD model for the two subsamples. The upper-right inset displays the posterior distribution of the ratio of normalized halo bias, $A_b^{{\rm high}-L}/A_b^{{\rm low}-L}$.
• Figure 3: Upper panel: Projected correlation functions $w_p(R)$ measured with $\Pi_{\max} = 50{\rm Mpc}/h$ for the low-$L$ (blue) and high-$L$ (red) BCG subsamples. Middle panel: The relative bias ratio $r_b$ (defined in Eq.\ref{['eq:rb']}) between the two subsamples. A value of unity indicates no assembly bias. The error bars include the propagated uncertainties of halo masses measured via galaxy–galaxy lensing. Purple solid circles show the results using $\Pi_{\max} = 50 \, {\rm Mpc}/h$, while gray open circles represent the results with $\Pi_{\max} = 100 \, {\rm Mpc}/h$. The $R$ positions of the grey points are slightly shifted for visual clarity. Lower panel: The ratio of $r_b(R)$ obtained with $\Pi_{\max} = 100\,{\rm Mpc}/h$ to that with $\Pi_{\max} = 50\,{\rm Mpc}/h$. The consistency with one demonstrates the robustness of our measurements against projection effects.
• Figure 4: The upper panel shows the projected correlation functions of the high-$L$ and low-$L$ BCG subsamples for MTNG and FLAMINGO hydrodynamical simulations; the lower panel shows the corresponding relative bias ratio $r_b(R)$ between the two subsamples. The shaded regions represent the corresponding observational results shown in Figure \ref{['fig:wp']}.
• Figure 5: Left: Constant-fit results of the relative bias ratio, $r_{b0}$, obtained using different halo bias models. The horizontal axis labels the models in the order of tinker2010large, Cole1989, Jing1998, Sheth2001, Seljak2004, Pillepich2010, Bhattacharya2011, and Comparat2017. Right: Halo bias as a function of halo mass in different models.