Studying galaxy cluster morphological metrics with Mock-X

TL;DR

This study uses the Mock-X framework to quantify how nine relaxation metrics applied to synthetic X-ray images of clusters from IllustrisTNG, BAHAMAS, and MACSIS distribute, correlate, and define relaxed samples across redshift and mass. By comparing to observational samples and analyzing redshift evolution and model dependence, the paper shows that metric distributions are broadly log-normal and that thresholds vary in effectiveness with redshift and simulation specifics. The results reveal systematic selection effects when constructing relaxed-cluster subsets from different metrics, with strong correlations among metrics but limited cross-metric consistency, especially at higher redshift or for certain subgrid physics. These findings underscore the need for multidimensional or calibrated relaxation criteria and provide a framework for interpreting cluster morphologies in simulations and observations, informing cosmological mass calibration strategies.

Abstract

Dynamically relaxed galaxy clusters have long played a role in galaxy cluster studies because it is thought their properties can be reconstructed more precisely and with less systematics. As relaxed clusters are desirable, there exist a plethora of criteria for classifying a galaxy cluster as relaxed. In this work, we examine $9$ commonly used observational and theoretical morphological metrics extracted from $54,000$ Mock-X synthetic X-ray images of galaxy clusters taken from the IllustrisTNG, BAHAMAS and MACSIS simulation suites. We find that the simulated criteria distributions are in reasonable agreement with the observed distributions. Many criteria distributions evolve as a function of redshift, cluster mass, numerical resolution and subgrid physics, limiting the effectiveness of a single relaxation threshold value. All criteria are positively correlated with each other, however, the strength of the correlation is sensitive to redshift, mass and numerical choices. Driven by the intrinsic scatter inherent to all morphological metrics and the arbitrary nature of relaxation threshold values, we find the consistency of relaxed subsets defined by the different metrics to be relatively poor. Therefore, the use of relaxed cluster subsets introduces significant selection effects that are non-trivial to resolve.

Figures (8)

• Figure 1: Four exemplar smoothed surface brightness maps computed from synthetic X-ray images of simulated clusters that pass all (top left), the theoretical (top right), the observational (bottom left) and none (bottom right) of the relaxation criteria examined in this work. The dashed blue line denotes $r_{\mathrm{500,crit}}$ of the cluster. Observational metrics focus on strong central emission and a smooth azimuthal distribution of emission, but theoretical parameters appear more sensitive to structure within the ICM.
• Figure 2: Log-normal distribution and CDF fit of the centre of mass displacement morphological metric, $X_{\mathrm{off}}$, for the TNG300-L1 sample at $z = 0.1$. The solid blue and dashed red lines show the sample distribution and best fit, respectively. The shaded region denotes the $1\sigma$ uncertainty computed via $10,000$ boostrap resamples. The dash-dot black line illustrates the literature threshold Neto2007.
• Figure 3: Comparison between simulated (solid blue) and observed (dashed red) morphological criteria distributions. The observed distributions are extracted from Mantz2015, Lovisari2017, and Nurgaliev2017. The simulated samples are cut to ensure the median mass/temperature and redshift are well matched to the observed sample. The blue (red) shaded region denotes the fraction of the distribution that would be classified as relaxed. The black dash-dot line denotes the threshold value taken from the literature. We note that the SPA criteria axes are inverted such that relaxed clusters always appear on the left of the panel.
• Figure 4: Redshift evolution of the morphological criteria examined in this work. Cluster samples from TNG300 level 1 (orange diamond), level 2 (red down triangle), level 3 (purple upward triangle), BAHAMAS (blue circle) and MACSIS (green square) are shown. The error bars show $68$ and $95$ percent of the sample. The black dash-dot shows the literature threshold value. For clarity, we have introduced small redshift offsets between the points.
• Figure 5: Correlation coefficient matrices for the $9$ morphological metrics at $z=0.1$ (top row) and $z=1.0$ (bottom row). Results for the MACSIS (left), BAHAMAS (middle) and TNG300 level 1 (right) samples are presented. Pearson correlation coefficients (Spearman's rank correlation coefficients) are shown in the lower left (upper right) triangle area. We note that we invert the values of the "positive" SPA criteria to ensure that if both criteria predict a more relaxed cluster the correlation is positive.