📝 SummarXiv

Abstract

We use a combination of self-supervised machine learning and visual classification to identify tidal features in a sample of 34,331 galaxies with stellar masses $\log_{10}(M_{*}/\rm{M}_{\odot})\geq9.5$ and redshift $z\leq0.4$, drawn from the Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) optical imaging survey. We assemble the largest sample of 1646 galaxies with confirmed tidal features, finding a tidal feature fraction $f=0.06^{+0.05}_{-0.01}$. We analyse how the incidences of tidal features and the various classes of tidal features vary with host galaxy stellar mass, photometric redshift, and colour, as well as halo mass. We find an increasing relationship between tidal feature fraction and host galaxy stellar mass, and a decreasing relationship with redshift. We find more tidal features occurring in group environments with $12.0<\log_{10}(M_{200}/\rm{M}_{\odot})<14.0$ than in the field or in denser, cluster environments. We also find that the central galaxies of the most massive ($\log_{10}(M_{200}/\rm{M}_{\odot})>14.1$) groups and clusters exhibit higher rates of tidal features than the satellite galaxies. We find good agreement between the trends we observe and the results obtained from purely visual or other automated methods, confirming the validity of our methodology and that using machine learning can drastically reduce the workload of visual classifiers, having needed to visually classify less than 30 per cent of our sample. Such methods will be instrumental in classifying the millions of suitable galaxies to be observed by large upcoming imaging surveys such as the Vera C. Rubin Observatory's Legacy Survey of Space and Time.