Machine-Learned Dark Matter Subhalo Candidates in the 4FGL-DR2: Search for the Perturber of the GD-1 Stream

TL;DR

This study employs supervised machine learning to identify potential dark matter subhalo candidates among unassociated gamma-ray sources, providing new insights into dark matter substructure and its possible influence on stellar streams.

Contribution

It introduces a machine learning approach to find dark matter subhalo candidates in gamma-ray data and investigates their potential role in perturbing the GD-1 stellar stream.

Findings

Identified 73 dark matter subhalo candidates from unassociated gamma-ray sources.

Found 17 candidates with X-ray sources, 52 without, suggesting diverse properties.

Provided a new inventory to explore dark matter substructure in the Galactic halo.

Abstract

The detection of dark matter subhalos without a stellar component in the Galactic halo remains a challenge. We use supervised machine learning to identify high-latitude gamma-ray sources with dark matter-like spectra among unassociated gamma-ray sources in the 4FGL-DR2. Out of 843 4FGL-DR2 unassociated sources at $|b| \geq 10\mathrm{^\circ}$, we select 73 dark matter subhalo candidates. Of the 69 covered by the Neil Gehrels Swift Observatory (Swift), 17 show at least one X-ray source within the 95% LAT error ellipse and 52 where we identify no new sources. This latest inventory of dark subhalos candidates allows us to investigate the possible dark matter substructure responsible for the perturbation in the GD-1 stellar stream. In particular, we examine the possibility that the alleged GD-1 dark subhalo may appear as a 4FGL-DR2 gamma-ray source from dark matter annihilation into Standard Model particles.