Selecting Clusters and Protoclusters via Stellar Mass Density: II. Application to HSC-SSP Observations

TL;DR

This work identifies galaxy clusters and protoclusters across $0.1 \le z \le 2$ in the HSC-SSP Wide PDR3 using a stellar-mass-density, dominant-galaxy finder to locate (proto)BCGs and their members. It builds new photometric redshifts by fusing spectroscopic HSC-SSP data, COSMOS2020, and unWISE infrared information within a Bayesian neural-network framework, achieving performance comparable to leading photo-$z$ catalogs. The study produces 16{,}007 candidate structures over ~$850\,\mathrm{deg^2}$ with predicted purity around 90% and robust central-galaxy recovery, while revealing substantial complementarity with CAMIRA and WH21, and moderate cross-correlation with X-ray cluster catalogs. The resulting catalog, including high-redshift candidates and detailed BCG/member properties, offers a rich target list for the Prime Focus Spectrograph and advances our understanding of cluster assembly and galaxy evolution in dense environments.

Abstract

We present a selection of candidates of clusters and protoclusters of galaxies identified in the photometric data of the HSC-SSP Wide Public Data Release 3 (PDR3), spanning the redshift range $\rm 0.1 \leq z \leq 2$. The selection method, detailed in Vicentin et al. (2025), involves detecting massive galaxies located in high-density regions of matter, identified as potential central dominant galaxies, i.e., (proto)BCGs. Probabilistic criteria based on proximity to the candidate central galaxy and the expected stellar mass of member galaxies are applied to identify likely members of each structure. We produced updated photometric redshift estimates using deep learning methods trained on a dataset combining spectroscopic redshifts from the HSC-SSP Wide PDR3, high-accuracy photometric redshifts from the COSMOS2020 catalog, and mid-infrared data from the unWISE catalog for matched sources. Our method achieves a predicted purity of $\sim 90\%$ in detecting (proto)clusters, with $\gtrsim 65\%$ correctly identifying the (proto)BCG. A total of 16,007 candidate (proto)clusters were identified over an effective area of $\rm \sim 850 \ deg^{2}$ within the HSC-SSP Wide footprint. Comparisons with other existing catalogs reveal a good level of consistency, while also highlighting that different methods yield complementary discoveries. We further compare richness and halo masses from our optical catalog with those from recent X-ray cluster catalogs (eROSITA and MCXC-II), finding a moderate positive correlation and a scatter of $\rm \sim 0.4$ dex. This catalog provides a valuable new set of targets for the Prime Focus Spectrograph (PFS) instrument.

Figures (16)

• Figure 1: The blue histogram represents the redshift distribution of the original training sample, consisting of spectroscopic data and photometric redshifts from COSMOS2020. The dot-dashed histogram indicates the distribution of the test sample. The red histogram shows the distribution of the unbiased training sample (Sections \ref{['sec: unbias']} and \ref{['sec: unders']}). The yellow dashed histogram represents the photometric redshift distribution from COSMOS2020. All histograms are normalized so that the total area sums to unity.
• Figure 2: From left to right: Photometric redshifts ($\rm z_{phot}$), uncertainties $\sigma_{NMAD}$ (\ref{['eq: nmad']}), $Bias$ (\ref{['eq: bias']}), and outlier fraction $f_{out} ($\ref{['eq: f_out']}), respectively, as a function of spectroscopic or COSMOS2020 redshifts ($\rm z_{spec|COSMOS2020}$). Blue lines show the results obtained in this work, while purple lines the results obtained by nishizawa20. The dashed line in the second plot denotes the result obtained by wh21.
• Figure 3: Comoving number density of galaxy structure candidates as a function of cluster photometric redshift. The dotted lines represent the halo mass functions obtained by tinker08 for three different mass limits, as indicated in the figure. The solid red, dashed green, and dash-dotted black lines denote the results obtained by this work, CAMIRA PDR3, and WH21, respectively.
• Figure 4: Upper panel: Spectroscopic redshift of the dominant galaxies ($\rm z_{spec, BCG}$) as a function of the redshift of the structure they inhabit ($\rm z_{cl}$). Lower panel: Metrics as a function of $\rm z_{cl}$. The blue line represents the dispersion $\sigma_{z}$, while the purple line indicates the bias $\delta_{z}$.
• Figure 5: Fraction of cross-matched structures as a function of richness. Each plot shows the results in a given redshift interval, as indicated above each panel. Solid red and dashed blue denote the comparison between this work with CAMIRA and WH21, respectively. The fraction is calculated by applying cuts on the estimated richness from this work, defined as the ratio between the number of structures in our catalog with richness above a given threshold that have a match in the other catalogs (solid red and dashed blue lines for CAMIRA and WH21, respectively), and the total number of structures in our catalog above that same threshold. Therefore, a richness cut of zero corresponds to considering the entire sample.