Identifying Lyα emitter candidates with Random Forest: Learning from galaxies in the CANDELS survey

¹ INAF – Osservatorio Astronomico di Roma, via Frascati 33, 00078 Monteporzio Catone, Italy
e-mail: lorenzo.napolitano@inaf.it
² Dipartimento di Fisica, Università di Roma Sapienza, Città Universitaria di Roma – Sapienza, Piazzale Aldo Moro 2, 00185 Roma, Italy
³ Dipartimento di Fisica, Università di Roma Tor Vergata, Via della Ricerca Scientifica 1, 00133 Roma, Italy
⁴ The Cosmic Dawn Centre (DAWN), Niels Bohr Institute, University of Copenhagen, Lyngbyvej 2, 2100 Copenhagen, Denmark
⁵ Dipartimento di Fisica e Astronomia, Università di Padova, Vicolo dell’Osservatorio 3, 35122 Padova, Italy
⁶ INAF – Osservatorio Astronomico di Padova, Vicolo dell’Osservatorio 5, 35122 Padova, Italy
⁷ Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD, 21218 USA
⁸ INAF – Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, Via P. Gobetti 93/3, 40129 Bologna, Italy
⁹ Institute for Advanced Research, Nagoya University, Nagoya, 464-8601 Japan
¹⁰ Department of Physics, Graduate School of Science, Nagoya University, Nagoya, 464-8602 Japan

Received: 26 May 2023
Accepted: 20 July 2023

Abstract

The physical processes that make a galaxy a Lyman alpha emitter have been extensively studied over the past 25 yr. However, the correlations between physical and morphological properties of galaxies and the strength of the Lyα emission line are still highly debated. Here, we investigate the correlations between the rest-frame Lyα equivalent width and stellar mass, star formation rate, dust reddening, metallicity, age, half-light semi-major axis, Sérsic index, and projected axis ratio in a sample of 1578 galaxies in the redshift range of 2 ≤ z ≤ 7.9 from the GOODS-S, UDS, and COSMOS fields. From the large sample of Lyα emitters (LAEs) in the dataset, we find that LAEs are typically common main sequence (MS) star-forming galaxies that show a stellar mass ≤10⁹ M_⊙, star formation rate ≤ 10^0.5 M_⊙ yr⁻¹, E(B − V)≤0.2, and half-light semi-major axis ≤1 kpc. Building on these findings, we have developed a new method based on a random forest (RF) machine learning (ML) classifier to select galaxies with the highest probability of being Lyα emitters. When applied to a population in the redshift range z ∈ [2.5, 4.5], our classifier holds a (80 ± 2)% accuracy and (73 ± 4)% precision. At higher redshifts (z ∈ [4.5, 6]), we obtained an accuracy of 73% and precision of 80%. These results highlight the possibility of overcoming the current limitations in assembling large samples of LAEs by making informed predictions that can be used for planning future large-scale spectroscopic surveys.