BROWDIE: a New Machine Learning Model for Searching T&Y Dwarfs Using the UKIDSS J, H, K Band Survey

TL;DR

This paper introduces BROWDIE, a machine learning model that uses UKIDSS J, H, K band data to effectively identify T and Y brown dwarfs, which are faint and difficult to detect, thereby advancing celestial object classification.

Contribution

The study presents a novel ML model, BROWDIE, that utilizes simultaneous multi-band UKIDSS data to improve T and Y dwarf detection, requiring fewer observations than previous models.

Findings

Identified 118 T dwarfs and 14 Y dwarfs in UKIDSS DR11PLUS LAS L4 zone.

Demonstrated improved detection accuracy with multi-band ML approach.

Enhanced understanding of faint brown dwarf populations.

Abstract

We propose a new T, Y dwarf search model using machine learning (ML), called the "BROWn Dwarf Image Explorer (BROWDIE). Brown dwarfs (BD) are estimated to make up 25 percent of all celestial objects in the Galaxy, yet only a small number have been thoroughly studied. Homogeneous and complete samples of BDs are essential to advance the studies. However, due to their faintness, conducting spectral studies of BDs can be challenging. T\&Y brown dwarfs, a redder and fainter subclass of BDs, are even harder to detect. As a result, only a few T\&Y dwarfs have been extensively studied. Numerous attempts, including ML using various color band observations, have been made to identify BDs based on their colors. However, those models often require a large number of color observations, which can be a limitation. This study implemented an ML model by utilizing data from the J, H, and K photometry of the UKIRT Infrared Deep Sky Surveys (UKIDSS) simultaneously, effectively distinguishing celestial objects similar to BDs. The BROWDIE model was trained using UKIDSS data, and based on the model, 118 T dwarfs and 14 Y dwarfs were found in the UKIDSS DR11PLUS LAS L4 zone.