Physical confirmation of the RR Lyrae substructure candidates from the Gaia hierarchical clustering forest and validation of the chemo-kinematic clustering metric (Cook et al. 2022)

A hierarchical clustering forest applied to 91,234 Gaia DR2 RR Lyrae variables, validated by correlated Gaia EDR3 proper motions, reported sixteen candidate Milky Way substructures smaller than 1 kpc — including stream-like extensions near globular clusters with known tidal tails, a mid-halo grouping (substructure 28) whose association with the nearby globular clusters IC 1257, NGC 6402 and NGC 6366 is dubious, and a grouping at about 55 kpc in the neighbourhood of the Large Magellanic Cloud (substructure 31) that may be a previously unknown Magellanic satellite. The candidates lack radial velocities and metallicities, so their physical status is undetermined, and the proposed chemo-kinematic clustering metric combining orbital energy, angular momentum and iron abundance was never applied or tested. The open problem is to determine the physical reality, membership, orbital coherence and progenitor association of these candidates, and to validate (or justifiably refine) the chemo-kinematic clustering methodology — including its conservative proper-motion validation and its limited computational scalability — on large, homogeneous RR Lyrae datasets. As of August 2026, no identified follow-up work has confirmed or refuted any candidate or tested the metric, although adjacent, independently developed methods now demonstrate chemo-kinematic clustering of RR Lyrae stars without spectroscopy.

Background

RR Lyrae variables are radially pulsating, metal-poor, core-helium-burning stars on the horizontal branch of old stellar populations; because their brightness is tied to their pulsation period (a period–luminosity relation that is strictest in the infrared), they serve as standard candles — objects of known intrinsic brightness from which distances are derived — and are among the best tracers of the Milky Way's stellar halo and its debris from accreted satellites. Cook et al. (2022) took the RRab (fundamental-mode RR Lyrae) catalogue of the Gaia mission's second data release, derived a distance for each of the 91,234 retained stars, and applied agglomerative hierarchical clustering with average linkage — a procedure that repeatedly merges the two closest clusters, where the distance between clusters is the average of all pairwise star-to-star distances — to the three-dimensional positions, computed separately within sky regions defined by the HEALPix pixelisation of the sphere (a standard scheme that partitions the sky into equal-area pixels). Because clustering in position alone produces many chance groupings, each candidate was validated using proper motions (angular sky-plane motion) from Gaia EDR3, the Gaia mission's early third data release: the statistic $\tilde{X}_{\theta}$ measures how tightly the member proper-motion vectors align with the group's median proper motion, and the adopted threshold $\tilde{X}_{\theta,\min}=0.98$ requires at least half of the stars to deviate from that median direction by at most 11.48 degrees — a deliberately conservative cut. This produced sixteen substructures of physical radius under 1 kpc. Five lie close to Milky Way globular clusters (dense, gravitationally bound old-star clusters) that have known tidal tails — streams of stars stripped by the Galactic tidal field — or proposed links to ancient merger events. Substructure 28 lies in the mid-halo along a sightline near three globular clusters whose distances are inconsistent with it by several kiloparsecs, making it an independent-structure candidate. Substructure 31, at a distance of about 55 kpc and projected near the Large Magellanic Cloud (LMC), the Milky Way's most massive satellite galaxy, was flagged as possibly a previously unknown satellite of the LMC — of particular interest because several lines of evidence suggest the Magellanic system was formerly a triplet of galaxies before the LMC accreted one member. The authors emphasise that confirming any such candidate requires spectroscopy (radial velocities and chemical abundances), which Gaia does not provide for these faint stars. Looking ahead, they proposed a 'chemo-kinematic' clustering metric in the space of classical integrals of motion — quantities conserved along stellar orbits that keep their values long after a stellar stream has dispersed in configuration space: the total orbital energy $E_{\mathrm{tot}}$, the vertical angular-momentum component $L_z$, and the magnitude $L_{\perp}$ of the angular-momentum component perpendicular to that axis — supplemented by the iron abundance $[\mathrm{Fe}/\mathrm{H}]$ (a logarithmic metallicity scale relative to the Sun). Their Equation (21) defines a squared distance $D_{\mathrm{modified}}$ as a weighted sum of squared differences in these four quantities, with scaling factors $\beta_i$ fixing units and relative weights; the authors state that determining whether this is a suitable metric requires further investigation, that no suitable Galactic-scale chemo-kinematic RR Lyrae dataset yet exists, and that infrared photometry from the Nancy Grace Roman Space Telescope combined with spectroscopic surveys could enable the test. They further note that average-linkage clustering costs $O(n^2 \log_2 n)$ in the number of stars, making density-based hierarchical clustering or sampling schemes necessary for substantially larger future catalogues, and that the conservative proper-motion threshold may reject genuine, strongly curved or phase-mixed (positionally dispersed but orbitally coherent) stream components.

Problem Statement

Determine the physical status of the candidate RR Lyrae substructures reported by the Gaia hierarchical clustering forest, and establish a validated chemo-kinematic clustering methodology for RR Lyrae samples of this kind. The common objective is to complete the chemo-kinematic stage of the clustering-forest survey — converting its three-dimensionally identified, proper-motion-validated candidate groupings into physically characterized (or refuted) substructures. Concretely, using the available astrometric, photometric and published spectroscopic data for stars in the candidate fields, together with simulations where the required observations do not yet exist: (1) establish for each of the sixteen reported groupings — in particular the mid-halo substructure 28 and the roughly 55 kpc grouping (substructure 31) projected near the Large Magellanic Cloud — whether it is a genuinely coherent stellar substructure and, if so, determine its membership, orbital coherence, and progenitor association (a globular cluster, a dwarf galaxy, the Magellanic system, or Galactic merger debris), or else demonstrate that it is a chance superposition, a known object, or otherwise not a distinct substructure, so far as the available data permit and stating what observation would decisively settle any candidate that they cannot decide; (2) test and validate — or justifiably refine — the proposed chemo-kinematic distance metric combining total energy, the angular-momentum components $L_z$ and $L_{\perp}$, and $[\mathrm{Fe}/\mathrm{H}]$ with explicit scaling factors $\beta_i$, on a large, homogeneous RR Lyrae dataset, quantifying substructure recovery against known structures and against simulations; and (3) determine whether the existing machinery blocks objectives (1) and (2) — that is, whether the conservative median-proper-motion validation (threshold $\tilde{X}_{\theta,\min}=0.98$) and the $O(n^2\log_2 n)$ clustering implementation remain adequate for these datasets — and, where they do not, propose validation statistics and algorithms that do not reject genuine strongly curved or phase-mixed stream components and that scale to the catalogue sizes involved. These three components are each required and jointly serve the single objective, the task being resolved when the physical status of the reported groupings has been determined as far as the available data and simulations permit (objective 1), the chemo-kinematic metric has been validated or justifiably refined (objective 2), and the adequacy of the existing validation statistic and clustering implementation has been decided (objective 3): the metric is the proposed tool for objective (1) once spectroscopic data exist, and the validation and scalability refinements remove identified blockers to applying it to the candidates and to future, larger RR Lyrae catalogues.

References

1. Cook, B. T., Woods, D. F., Ruprecht, J. D., Varey, J., Mastandrea, R., de Soto, K., Harburg, J. F., Rebbapragada, U., Mahabal, A. A. 2022, 'Tracing Milky Way substructure with an RR Lyrae hierarchical clustering forest', Monthly Notices of the Royal Astronomical Society 513, 2509–2521, DOI 10.1093/mnras/stac1007, https://arxiv.org/abs/2204.05868
2. Sun, S., Wang, F., Zhang, H., Xue, X.-X., Huang, Y., Zhang, R., Rix, H.-W., Li, X., Liu, G., Zhang, L., Yang, C., Zhang, S. 2025, 'Identifying Galactic Substructures in 5D Space Using All-sky RR Lyrae Stars in Gaia DR3', The Astrophysical Journal 979, 213, DOI 10.3847/1538-4357/ad9de0, https://arxiv.org/abs/2411.13122
3. Cristi-Cambiaso, N., Navarrete, C., Catelan, M., Zoccali, M., Quezada, C. 2026, 'A clustering-based search for substructures in the Galactic plane and bulge using RR Lyrae stars as tracers', Astronomy & Astrophysics 706, A298, DOI 10.1051/0004-6361/202556821, https://arxiv.org/abs/2512.23846
4. Kunder, A., Prudil, Z., Skaggs, C., Reggiani, H., Nataf, D. M., Hughes, J., Covey, K. R., Devine, K. 2024, 'The Galactic Bulge Exploration. III. Calcium Triplet Metallicities for RR Lyrae Stars', The Astronomical Journal 168, 139, DOI 10.3847/1538-3881/ad6262, https://arxiv.org/abs/2407.01515
5. Medina, G. E., Hansen, C. J., Muñoz, R. R., Grebel, E. K., Vivas, A. K., Carlin, J. L., Martínez-Vázquez, C. E. 2023, 'RR Lyrae stars as probes of the outer Galactic halo: chemical and kinematic analysis of a pilot sample', Monthly Notices of the Royal Astronomical Society 519, 5689–5722, DOI 10.1093/mnras/stac3800, https://arxiv.org/abs/2309.03271
6. Feng, Y., Guhathakurta, P., Peng, E. W., Cunningham, E. C., Côté, P., Ferrarese, L., Gwyn, S. D. J. 2026, 'Kinematics of Distant Milky Way Halo RR Lyrae Stars out to 160 kpc', The Astrophysical Journal 998, 157, DOI 10.3847/1538-4357/ae313b, https://arxiv.org/abs/2512.09795
7. Apfel, M., McKinnon, K., Rockosi, C. M., Guhathakurta, P., Johnston, K. V. 2025, 'Constraining the Milky Way Halo Accretion History with Simulated Stellar Halos: Designing the HALO7D-X Survey', The Astrophysical Journal 988, 225, DOI 10.3847/1538-4357/ade43d, https://arxiv.org/abs/2507.05239