The Massive and Distant Clusters of WISE Survey 2: Second Data Release

TL;DR

The paper presents the second data release of the MaDCoWS survey, expanding coverage to 6498 deg^2, cataloging over 133,000 galaxy cluster candidates up to redshift 2, and validating the catalog's accuracy and completeness using neural networks and mock data.

Contribution

This work provides a significantly expanded galaxy cluster catalog with improved detection methods and validation techniques, including CNN-based spurious detection removal and mock light cone analysis.

Findings

Catalog contains 133,036 cluster candidates at 0.1 ≤ z ≤ 2

Purity of the catalog exceeds 90% for high S/N detections

Recovery rate of known clusters is 59-91% in overlapping areas

Abstract

We present the second data release of the Massive and Distant Clusters of WISE Survey 2 (MaDCoWS2). We expand from the equatorial first data release to most of the Dark Energy Camera Legacy Survey area, covering a total area of 6498 deg^2. The catalog consists of 133,036 S/N $\geq5$ galaxy cluster candidates at $0.1\leq z \leq2$, including 6790 candidates at z > 1.5. We train a convolutional neural network (CNN) to identify spurious detections, and include CNN-based cluster probabilities in the final catalog. We also compare the MaDCoWS2 sample with literature catalogs in the same area. The larger sample provides robust results that are consistent with our first data release. At S/N $\geq5$, we rediscover 59-91% of clusters in existing catalogs that lie in the unmasked area of MC2. The median positional offsets are under 250 kpc, and the standard deviation of the redshifts is 0.031(1+z). We fit a redshift-dependent power law to the relation between MaDCoWS2 S/N and observables from existing catalogs. Over the redshift ranges where the surveys overlap with MaDCoWS2, the lowest scatter is found between S/N and observables from optical/infrared surveys. We also assess the performance of our method using a mock light cone measuring purity and completeness as a function of cluster mass. The purity is above 90%, and we estimate the 50% completeness threshold at a virial mass of log(M/M$_\odot$)$\approx14.3$. The completeness estimate is uncertain due to the small number of massive halos in the light cone, but consistent with the recovery fraction found by comparing to other cluster catalogs.