DETECTIFz galaxy groups in the REFINE survey - I. Group detection and quenched fraction evolution at z < 2.5

We use a large K-selected sample of 299 961 galaxies from the REFINE survey, consisting of a combination of data from three of the deepest near-infrared surveys, UKIDSS UDS, COSMOS/UltraVISTA, and CFHTLS-D1/VIDEO, that were homogeneously reduced to obtain photometric redshifts and stellar masses. We detect 2588 candidate galaxy groups up to z = 3.15 at S/N > 1.5. We build a very pure ($\gt 90{{\rm \,per\ cent}}$) sample of 448 candidate groups up to z = 2.5 and study some of their properties. Cluster detection is done using the DElaunay TEssellation ClusTer IdentiFication with photo-z (DETECTIFz) algorithm that we describe. This new group finder algorithm uses the joint probability distribution functions of redshift and stellar-mass of galaxies to detect groups as stellar-mass overdensities in overlapping redshift slices, where density is traced using Monte Carlo realization of the Delaunay Tessellation Field Estimator. We compute the algorithm selection function using mock galaxy catalogues taken from cosmological N-body simulation lightcones. Based on these simulations, we reach a completeness of $\sim \! 80{{\,\rm per\ cent}}$ for clusters (M₂₀₀ > 10¹⁴M_⊙) at a purity of $\sim \! 90{{\, \rm per\ cent}}$ at z < 2.5. Using our 403 most massive candidate groups, we constrain the redshift evolution of the group galaxy quenched fraction at 0.12 ≤ z < 2.32, for galaxies with 10.25 < log M_⋆/M_⊙ < 11 in 0.5 × R₂₀₀. We find that the quenched fraction in group cores is higher than in the field in the full redshift range considered, the difference growing with decreasing redshift. This indicates either more efficient quenching mechanisms in group cores at lower redshift or pre-processing by cosmic filaments.