Impact of projection-induced optical selection bias on the weak lensing mass calibration of galaxy clusters

TL;DR

The paper addresses a bias in weak-lensing mass calibration for optically selected galaxy clusters caused by projection effects that coherently boost richness $\lambda$ and the lensing signal $\Delta\Sigma(r_p)$. It uses two mock catalogs, MiniUchuu-based HOD and Cardinal Addgals with redMaPPer, to forward-model stacked lensing signals and perform mock mass calibrations. They find a mass calibration bias of about 20–50% on average, rising to 20–80% on large scales ($r_p \gtrsim 3$ Mpc), with Cardinal showing stronger and redshift–richness dependent biases. The authors argue this selection bias dominates current systematics and advocate simulation-based forward modeling and multiwavelength, multi-probe strategies to mitigate and self-calibrate.

Abstract

Weak gravitational lensing signals of optically identified clusters are impacted by a selection bias -- halo triaxiality and large-scale structure along the line of sight simultaneously boost the lensing signal and richness (the inferred number of galaxies associated with a cluster). As a result, a cluster sample selected by richness has a mean lensing signal higher than expected from its mean mass, and the inferred mass will be biased high. This selection bias is currently limiting the accuracy of cosmological parameters derived from optical clusters. In this paper, we quantify the bias in mass calibration due to this selection bias. Using two simulations, MiniUchuu and Cardinal, with different galaxy models and cluster finders, we find that the selection bias leads to an overestimation of lensing mass at a 20-50% level, with a larger bias 20-80% for large-scale lensing (>3 Mpc). Even with a conservative projection model, the impact of selection bias significantly outweighs the impact of other currently known cluster lensing systematics. We urge the cluster community to account for this bias in all future optical cluster cosmology analyses, and we discuss strategies for mitigating this bias.

Figures (3)

• Figure 1: Stacked lensing signal derived from the MiniUchuu-based mock (black) in four richness bins at $z=0.3$, with richness modeled by counts-in-cylinders ($\pm 30~h^{-1} \rm cMpc$ distance uncertainties). The red curves show the analytic predictions based on the mean mass and concentration of halos in each bin. The mean lensing signals of richness-selected clusters are biased high, and the bias is higher on large scales than on small scales.
• Figure 2: Analogous to Fig. \ref{['fig:cylinder_richness_lensing']} but for Cardinal, based on the Addgals galaxy model and the redMaPPer cluster finder. The lensing bias persists across all scales and is stronger than that shown in Fig. \ref{['fig:cylinder_richness_lensing']}.
• Figure 3: Fractional mass bias resulted from projection-induced selection bias, for the MiniUchuu mock (left) and Cardinal (right). We fit the analytic model to the mock lensing signals. The points and error bars correspond to the posterior means and 68% credible intervals. The top panels use radial range [0.2 Mpc, 3 Mpc), the middle panels use [3 Mpc, 30 Mpc), while the bottom panels combine both radial ranges. Cardinal tends to have larger mass biases than MiniUchuu. For Cardinal, the bias is higher for high-redshift and low-richness clusters.