Automated detection of merging galaxies at z = 0.25 − 1.0 in the CLAUDS+HSC survey using random forests

Abstract

Using a sample of galaxies (M⋆ ≥ 10^10.5M_⊙) covering an effective area of ∼ 20 deg² in the CLAUDS+HSC survey, we apply a Random Forest Classifier to automatically identify merger candidates in deep r-band images. We identify a largely pure, ∼ 90% complete sample of mergers which we use to derive the evolution in the merger fraction from 0.25 ≤ z_phot ≤ 1.0. We parameterize the merger fraction evolution with a power law of the form f_m = f₀(1+z)^m . Simulating the effects of increasing redshift on the detectability of mergers, we correct our merger fractions for incompleteness to obtain a local merger fraction of f₀ = 1.0%±0.2% and power-law index of m = 2.3±0.4, which is inconsistent with the mild or non-evolving merger scenario (m < 1.5) with 96.6% confidence. Finally, we estimate 0.3 merging events to occur per massive galaxy since z = 1.