Probing the Milky Way halo with RR Lyrae stars from Gaia Data Release 3

¹ INAF – Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, via Piero Gobetti 93/3, Bologna 40129, Italy
² Dipartimento di Fisica e Astronomia-Università di Bologna, via Piero Gobetti 93/2, Bologna 40129, Italy
³ Computer Science Department, Lucerne University of Applied Sciences and Arts, Luzern 6002, Switzerland
⁴ TOELT LLC, Machine Learning Research and Development Department, Winterthur 8406, Zurich, Switzerland

^★ Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Received: 15 January 2026
Accepted: 19 February 2026

Abstract

Context. The Milky Way (MW) stellar halo, containing debris from past accretion events, serves as a fossil record of hierarchical mass assembly. Due to their distinct properties, RR Lyrae stars (RRLs) serve as excellent tracers for identifying and characterising the halo’s sub-structures. Thanks to the advent of Gaia Data Release 3 (DR3), which includes high-precision positions, parallaxes, proper motions, and radial velocities, along with the identification and characterisation of thousands of RRLs, it has become possible to study the distribution, kinematics, and metallicity of RRLs in the various dynamical sub-structures with unprecedented detail.

Aims. Our primary goal is to identify and characterise the dynamical sub-structures of the MW halo using RRLs from Gaia DR3. Methods. We analysed a sample of 4933 RRLs, for which we calculated the integrals of motion and orbital parameters. We applied the domain-informed novelty detection CLustering in Multiphase Boundaries (CLiMB) framework to identify RRL membership in the MW sub-structures. We then used newly calibrated photometric metallicities available in the literature to study the metallicity distributions of RRLs in different sub-structures.

Results. We analysed the metallicity distributions of RRLs in major accreted system remnants as a snapshot of their chemical evolutionary status during early epochs. We calculated the weighted mean metallicity ([Fe∕H]) and the corresponding standard deviation for Gaia Sausage/Enceladus ([Fe∕H] = −1.57 ± 0.25 dex), Sequoia ([Fe/H] = −1.64 ± 0.26 dex), and the Helmi streams ([Fe/H] = −1.66 ± 0.19 dex). The metallicity distribution of RRLs in Thamnos was found to be bimodal, with the metal-poor peak likely representing the genuine (i.e. true) accreted Thamnos population ([Fe/H] = −1.94 ± 0.20 dex), in agreement with recent works based on spectroscopic abundances. Our analysis shows that the sub-structures ED-1 and L-RL3 are highly contaminated by thick-disc stars. However, the metal-poor tails in their metallicity distributions might be signatures of remnants from small accreted systems. We also identified some over-densities of RRLs in correspondence with the recently reported sub-structures Shiva and Shakti, which we suggest are of an in situ origin. Finally, we applied the RRL-based mass-metallicity relation of galaxies to test the nature of the identified dynamical sub-structures.