pith. sign in

arxiv: 2605.23801 · v1 · pith:GTZMOYHAnew · submitted 2026-05-22 · 🌌 astro-ph.GA

Unsupervised Chemo-Dynamical Dissection of the Inner Galactic Halo: Discovery of Five Accreted Substructures with SDSS-V and Gaia

Furkan Akbaba , Olcay Plevne This is my paper

Pith reviewed 2026-05-25 03:15 UTC · model grok-4.3

classification 🌌 astro-ph.GA
keywords Galactic haloaccreted substructureschemo-dynamical analysisunsupervised clusteringMilky Way assemblyex-situ starsSDSS-VGaia
0
0 comments X

The pith

Unsupervised clustering on 12-dimensional chemo-dynamical data from SDSS-V and Gaia identifies five new tightly bound accreted substructures in the inner Galactic halo.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The authors apply UMAP dimensionality reduction and HDBSCAN clustering to a chemically selected sample of 2185 ex-situ stars without any kinematic pre-selection. Their blind search recovers nine groupings that match seven previously known substructures while also isolating five new candidates labeled FO1 through FO5. Four of the new groups appear as robust overdensities in both chemistry and dynamics, and one shows extreme nitrogen enhancement consistent with debris from a massive globular cluster. The work demonstrates that chemical dimensions break dynamical degeneracies that pure orbit-based methods cannot resolve in the crowded inner halo.

Core claim

A purely data-driven 12-dimensional chemo-dynamical analysis with UMAP and HDBSCAN on SDSS-V Milky Way Mapper DR19 and Gaia DR3 data recovers known substructures including Gaia-Enceladus/Sausage, the Helmi Streams, and Sequoia, and reports five new tightly bound candidate substructures FO1–FO5 with total energy ≤ −1.8 × 10^5 km² s^{-2}. Four candidates are confirmed as robust chemo-dynamical overdensities; FO2 exhibits [N/Fe] = +0.83 ± 0.16 suggestive of tidal debris from a disrupted massive globular cluster. High-dimensional chemical information differentiates structures that share similar orbits but distinct chemistry.

What carries the argument

UMAP+HDBSCAN unsupervised clustering pipeline applied to a chemically selected ex-situ sample in 12-dimensional chemo-dynamical feature space.

If this is right

  • High-dimensional chemical information resolves dynamical degeneracies between structures that share similar orbits.
  • The method recovers known substructures without kinematic pre-selection, confirming that the feature space preserves assembly history signals.
  • FO2's nitrogen enhancement points to tidal debris from a disrupted massive globular cluster.
  • The inner halo retains a record of multiple early accretion events in tightly bound orbits.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Extending the same blind clustering to larger spectroscopic samples could map additional low-energy substructures below current detection thresholds.
  • The separation of FO5 from Shiva and FO3 from the Helmi Streams illustrates that chemistry can split populations previously treated as single dynamical entities.
  • Nitrogen-rich candidates like FO2 may trace a distinct channel of globular-cluster formation inside accreted dwarf galaxies.

Load-bearing premise

The clusters found by the algorithm correspond to physically distinct accreted stellar populations rather than artifacts of the chosen hyperparameters or leftover contamination in the chemical selection.

What would settle it

Follow-up high-resolution spectroscopy that measures consistent chemical abundance patterns and orbital parameters for the FO1–FO5 candidate members distinct from the surrounding halo field stars.

Figures

Figures reproduced from arXiv: 2605.23801 by Furkan Akbaba, Olcay Plevne.

Figure 1
Figure 1. Figure 1: Distribution of the four PCA+KMeans populations in the [Al/Fe]–[Mg/Mn] chemical diagnostic plane. Each panel corresponds to one of the four recovered PCA+KMeans clusters. The dashed curve indicates the em￾pirical accreted/in-situ chemical boundary proposed by S. Alinder et al. (2025). Cluster 2 (purple) lies predominantly within the chemically accreted region above the boundary, confirming its ex-situ natu… view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of the four PCA+KMeans popula￾tions in the [Mg/Fe]–[Fe/H] plane. The colour coding corre￾sponds to the four chemically distinct populations identified through the PCA+KMeans analysis. Solid line is taken from J. T. Mackereth et al. (2019) and dashed line is taken from F. Akbaba et al. (2026). sis, all chemical dimensions were standardised to zero mean and unit variance using a StandardScaler. … view at source ↗
Figure 4
Figure 4. Figure 4: Three pair-wise projections of the three-dimensional UMAP embedding constructed from the 12-dimensional chemo-dynamical feature space of the ex-situ sample (N = 2185). The adopted feature space consists of six chemical dimensions ([Fe/H], [Mg/Fe], [Al/Fe], [Mn/Fe], [Ni/Fe], [Si/Fe]) and six dynamical dimensions (Etot, Lz, e, Zmax, JR, and Jz). Points are colour-coded according to their HDBSCAN cluster assi… view at source ↗
Figure 5
Figure 5. Figure 5: Combined chemo-dynamical χ 2 similarity matrix between the HDBSCAN groups identified in this work and the literature substructures compiled by D. Horta et al. (2023). Rows correspond to the recovered groups and columns to the reference structures. The combined metric represents the average of the chemical and dynamical χ 2 distances. Lower values indicate stronger similarity. Groups below the dashed line c… view at source ↗
Figure 6
Figure 6. Figure 6: Multi-dimensional overview of the recovered literature substructures. From left to right: [Mg/Mn]–[Al/Fe] plane, Etot–Lz plane, eccentricity–Rapo plane, √ JR–Lz action space, action diamond (Lz/Jtot versus (Jz − JR)/Jtot), and the VR–VZ velocity plane. Grey points indicate the full ex-situ sample, while coloured symbols represent the recovered literature structures. eccentricities (e = 0.48 ± 0.15) and rel… view at source ↗
Figure 7
Figure 7. Figure 7: Multi-dimensional overview of the five newly identified candidate structures FO1–FO5. Columns correspond to indi￾vidual FO groups, while rows show different chemo-dynamical diagnostic spaces: [Mg/Mn]–[Al/Fe], Etot–Lz, eccentricity–Rapo, √ JR–Lz, action diamond, and VR–VZ . Grey points indicate the full ex-situ population for reference [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Comparison between FO5 and the Shiva and Shakti structures proposed by K. Malhan & H.-W. Rix (2024). Left: distribution in the Etot–Lz plane showing the Shiva (green polygon) and Shakti (orange polygon) selection regions. Dark purple symbols indicate FO5 members located inside the Shiva region (N = 11), while light purple symbols correspond to FO5 stars outside the Shiva selection (N = 17). Centre: [Mg/Mn]… view at source ↗
read the original abstract

The inner Galactic halo is a complex graveyard of the Milky Way's earliest accretion events, where severe orbital phase-mixing challenges traditional dynamical stream-finding techniques. We present a purely data-driven, 12-dimensional chemo-dynamical analysis of the inner halo using \textsl{SDSS-V Milky Way Mapper} (DR19) and \textsl{Gaia} DR3. Utilizing an unsupervised machine learning framework based on UMAP and HDBSCAN, we perform a blind search for clustered populations within a chemically selected \textit{ex-situ} sample of 2,185 stars without kinematic pre-selection. Our pipeline recovers nine kinematic groupings corresponding to seven known substructures (including \textsl{Gaia}-Enceladus/Sausage, the Helmi Streams, and Sequoia), validating the robustness of the high-dimensional feature space. We also report five new tightly bound candidate substructures, designated FO1--FO5 ($E_{\rm tot} \leq -1.8 \times 10^5~\mathrm{km^2~s^{-2}}$). Four candidates (FO1, FO3, FO4, FO5) are confirmed as robust chemo-dynamical overdensities, while FO2 exhibits a striking nitrogen enhancement ($[\mathrm{N/Fe}] = +0.83 \pm 0.16$) suggestive of tidal debris from a disrupted massive globular cluster. Finally, we demonstrate that high-dimensional chemical information is critical for resolving dynamical degeneracies in the crowded inner halo, differentiating structures sharing similar orbits but distinct chemistry (e.g., FO5 and Shiva), and the reverse (e.g., FO3 and the Helmi Streams). These findings confirm that the deepest regions of the Galactic potential preserve a rich record of the Galaxy's assembly history.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper applies an unsupervised UMAP+HDBSCAN pipeline in a 12D chemo-dynamical space to a chemically selected ex-situ sample of 2185 stars from SDSS-V DR19 and Gaia DR3. It recovers nine groupings corresponding to seven known inner-halo substructures (Gaia-Enceladus/Sausage, Helmi Streams, Sequoia, etc.) and reports five new tightly bound candidates (FO1–FO5 with E_tot ≤ −1.8 × 10^5 km² s^{-2}), claiming four are robust chemo-dynamical overdensities and that FO2 shows extreme nitrogen enhancement ([N/Fe] = +0.83 ± 0.16). The work emphasizes that high-dimensional chemistry resolves orbital degeneracies.

Significance. If the new candidates survive rigorous stability tests, the result would add concrete building blocks to the inner-halo accretion inventory and strengthen the case that chemical dimensions are required to break dynamical degeneracies in phase-mixed regions. The recovery of multiple known structures already provides a useful internal validation of the 12D feature space.

major comments (3)
  1. [Methods] Methods section: No quantitative stability analysis (hyperparameter grid search, bootstrap resampling, or mock-data injection) is reported for the UMAP (n_neighbors, min_dist) and HDBSCAN (min_cluster_size, min_samples) choices that produced FO1–FO5. Because the central claim rests on these five new overdensities being physically distinct rather than clustering artifacts, the absence of such tests is load-bearing.
  2. [Results] Results (FO1–FO5 subsection): The statement that FO1, FO3, FO4, FO5 are “confirmed as robust” is given without the explicit robustness metrics, survival fractions across hyperparameter variations, or contamination-injection tests that would substantiate the claim.
  3. [Sample selection] Chemical-selection paragraph: The ex-situ cut and its estimated residual in-situ contamination fraction (especially for metal-poor stars) are not quantified with error budgets or selection-function modeling, leaving open the possibility that contamination contributes to the reported overdensities.
minor comments (2)
  1. [Figure 3] Figure captions and text should explicitly state the adopted UMAP random seed and the precise HDBSCAN parameters used for the final clustering run.
  2. [Abstract] The energy unit in the abstract and §4 is written km² s^{-2}; consistency with standard Galactic-dynamics notation (km² s^{-2}) should be checked throughout.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have helped us improve the methodological transparency and quantitative support for our claims. We address each major comment point by point below.

read point-by-point responses

  1. Referee: [Methods] Methods section: No quantitative stability analysis (hyperparameter grid search, bootstrap resampling, or mock-data injection) is reported for the UMAP (n_neighbors, min_dist) and HDBSCAN (min_cluster_size, min_samples) choices that produced FO1–FO5. Because the central claim rests on these five new overdensities being physically distinct rather than clustering artifacts, the absence of such tests is load-bearing.

    Authors: We agree that the absence of reported quantitative stability tests represents a genuine gap, as the identification of FO1–FO5 is central to the paper. In the revised manuscript we have added a new Methods subsection that presents a hyperparameter grid search over UMAP and HDBSCAN parameters, bootstrap resampling of the stellar catalog, and mock-data injection tests. These analyses are used to evaluate cluster stability and will be shown in new figures and tables. revision: yes

  2. Referee: [Results] Results (FO1–FO5 subsection): The statement that FO1, FO3, FO4, FO5 are “confirmed as robust” is given without the explicit robustness metrics, survival fractions across hyperparameter variations, or contamination-injection tests that would substantiate the claim.

    Authors: We accept that the original wording lacked supporting quantitative metrics. The revised Results section now explicitly references the survival fractions, hyperparameter stability ranges, and mock-injection outcomes from the new Methods analyses, replacing the unqualified statement of confirmation with data-driven language. revision: yes

  3. Referee: [Sample selection] Chemical-selection paragraph: The ex-situ cut and its estimated residual in-situ contamination fraction (especially for metal-poor stars) are not quantified with error budgets or selection-function modeling, leaving open the possibility that contamination contributes to the reported overdensities.

    Authors: We agree that a quantified contamination estimate with error budget and selection-function modeling was missing. The revised sample-selection section now includes these elements, derived from a control-sample comparison and modeled selection function, together with a brief discussion of how residual contamination could affect the reported overdensities. revision: yes

Circularity Check

0 steps flagged

No significant circularity; analysis is data-driven clustering on public observations.

full rationale

The paper applies UMAP+HDBSCAN to a 12D chemo-dynamical feature space drawn from SDSS-V and Gaia observations of 2185 stars pre-selected by chemistry as ex-situ. It recovers known substructures and reports new candidates based on observed overdensities in the data. No equations, fitted parameters, or derivations are present that reduce the reported clusters to inputs by construction. No self-citation chains or uniqueness theorems are invoked as load-bearing for the central claim. The pipeline is self-contained against external benchmarks (recovery of Gaia-Enceladus, Helmi Streams, etc.) and does not rely on any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim depends on the chemical pre-selection successfully isolating ex-situ stars and on the clustering algorithm returning physically meaningful groups; both steps introduce modeling choices whose impact is not quantified in the abstract.

free parameters (1)
  • UMAP and HDBSCAN hyperparameters
    Embedding dimension, n_neighbors, min_cluster_size and related tuning parameters control which overdensities are reported as substructures.
axioms (1)
  • domain assumption Chemically selected sample consists of ex-situ stars with negligible in-situ contamination
    The pipeline begins with a chemically selected ex-situ sample of 2185 stars without kinematic pre-selection.

pith-pipeline@v0.9.0 · 5874 in / 1371 out tokens · 34032 ms · 2026-05-25T03:15:59.325745+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

58 extracted references · 58 canonical work pages · 8 internal anchors

  1. [1]

    2026, arXiv e-prints, arXiv:2603.19230, doi: 10.48550/arXiv.2603.19230

    Akbaba, F., Horta, D., & Plevne, O. 2026, arXiv e-prints, arXiv:2603.19230, doi: 10.48550/arXiv.2603.19230

    work page doi:10.48550/arxiv.2603.19230 2026

  2. [2]

    2025, arXiv e-prints, arXiv:2511.10092, doi: 10.48550/arXiv.2511.10092

    Alinder, S., Bensby, T., & McMillan, P. 2025, arXiv e-prints, arXiv:2511.10092, doi: 10.48550/arXiv.2511.10092

    work page doi:10.48550/arxiv.2511.10092 2025

  3. [3]

    2015, in American Astronomical Society Meeting Abstracts, Vol

    Allende-Prieto, C., & Apogee Team. 2015, in American Astronomical Society Meeting Abstracts, Vol. 225, American Astronomical Society Meeting Abstracts #225, 422.07 Allende Prieto, C., Beers, T. C., Wilhelm, R., et al. 2006, ApJ, 636, 804, doi: 10.1086/498131

    work page doi:10.1086/498131 2015

  4. [4]

    Laporte, C. F. P., & Deg, N. 2022, ApJ, 937, 12, doi: 10.3847/1538-4357/ac8b0d Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33, doi: 10.1051/0004-6361/201322068 Astropy Collaboration, Price-Whelan, A. M., Sip˝ ocz, B. M., et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4f Astropy Collaboration, Price-Whelan, A. ...

    work page doi:10.3847/1538-4357/ac8b0d 2022

  5. [5]

    Estimating distances from parallaxes. V: Geometric and photogeometric distances to 1.47 billion stars in Gaia Early Data Release 3

    Demleitner, M., & Andrae, R. 2021, AJ, 161, 147, doi: 10.3847/1538-3881/abd806 Barb´ a, R. H., Minniti, D., Geisler, D., et al. 2019, ApJL, 870, L24, doi: 10.3847/2041-8213/aaf811

    work page internal anchor Pith review doi:10.3847/1538-3881/abd806 2021

  6. [6]

    Deason, A. J. 2018, MNRAS, 478, 611, doi: 10.1093/mnras/sty982

    work page doi:10.1093/mnras/sty982 2018

  7. [7]

    2022, MNRAS, 514, 689, doi: 10.1093/mnras/stac1267

    Belokurov, V., & Kravtsov, A. 2022, MNRAS, 514, 689, doi: 10.1093/mnras/stac1267

    work page doi:10.1093/mnras/stac1267 2022

  8. [8]

    2019, MNRAS, 482, 1417, doi: 10.1093/mnras/sty2813

    Bennett, M., & Bovy, J. 2019, MNRAS, 482, 1417, doi: 10.1093/mnras/sty2813

    work page internal anchor Pith review doi:10.1093/mnras/sty2813 2019

  9. [9]

    galpy: A Python Library for Galactic Dynamics

    Bovy, J. 2015, ApJS, 216, 29, doi: 10.1088/0067-0049/216/2/29

    work page internal anchor Pith review doi:10.1088/0067-0049/216/2/29 2015

  10. [10]

    S., & Vaughan, A

    Bowen, I. S., & Vaughan, A. H., J. 1973, ApOpt, 12, 1430, doi: 10.1364/AO.12.001430

    work page doi:10.1364/ao.12.001430 1973

  11. [11]

    K., et al

    Buder, S., Lind, K., Ness, M. K., et al. 2022, MNRAS, 510, 2407, doi: 10.1093/mnras/stab3504

    work page doi:10.1093/mnras/stab3504 2022

  12. [12]

    Campello, R. J. G. B., Moulavi, D., & Sander, J. 2013, Advances in Knowledge Discovery and Data Mining, 160

    work page 2013

  13. [13]

    2019, ApJ, 883, 107, doi: 10.3847/1538-4357/ab38b8

    Conroy, C., Bonaca, A., Cargile, P., et al. 2019, ApJ, 883, 107, doi: 10.3847/1538-4357/ab38b8

    work page doi:10.3847/1538-4357/ab38b8 2019

  14. [14]

    V., Hasselquist, S., et al

    Cunha, K., Smith, V. V., Hasselquist, S., et al. 2017, ApJ, 844, 145, doi: 10.3847/1538-4357/aa7beb

    work page doi:10.3847/1538-4357/aa7beb 2017

  15. [15]

    2020, MNRAS, 493, 5195, doi: 10.1093/mnras/stz3537 De Silva, G

    Das, P., Hawkins, K., & Jofr´ e, P. 2020, MNRAS, 493, 5195, doi: 10.1093/mnras/stz3537 De Silva, G. M., Freeman, K. C., Bland-Hawthorn, J., et al. 2015, MNRAS, 449, 2604, doi: 10.1093/mnras/stv327

    work page doi:10.1093/mnras/stz3537 2020

  16. [16]

    2025, A&A, 700, A154, doi: 10.1051/0004-6361/202554252

    Dodd, E., Matsuno, T., Helmi, A., et al. 2025, A&A, 700, A154, doi: 10.1051/0004-6361/202554252

    work page doi:10.1051/0004-6361/202554252 2025

  17. [17]

    W., Rix, H.-W., et al

    Eilers, A.-C., Hogg, D. W., Rix, H.-W., et al. 2022, ApJ, 928, 23, doi: 10.3847/1538-4357/ac54ad

    work page doi:10.3847/1538-4357/ac54ad 2022

  18. [18]

    C., Horta, D., et al

    Fernandes, L., Mason, A. C., Horta, D., et al. 2023, MNRAS, 519, 3611, doi: 10.1093/mnras/stac3543

    work page doi:10.1093/mnras/stac3543 2023

  19. [19]

    2002, ARA&A, 40, 487, doi: 10.1146/annurev.astro.40.060401.093840 Gaia Collaboration, Prusti, T., de Bruijne, J

    Freeman, K., & Bland-Hawthorn, J. 2002, ARA&A, 40, 487, doi: 10.1146/annurev.astro.40.060401.093840 Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al. 2016, A&A, 595, A1, doi: 10.1051/0004-6361/201629272 Gaia Collaboration, Vallenari, A., Brown, A. G. A., et al. 2023, A&A, 674, A1, doi: 10.1051/0004-6361/202243940 Garc´ ıa P´ erez, A. E., Allend...

    work page doi:10.1146/annurev.astro.40.060401.093840 2002

  20. [20]

    E., Siegmund, W

    Gunn, J. E., Siegmund, W. A., Mannery, E. J., et al. 2006, AJ, 131, 2332, doi: 10.1086/500975

    work page doi:10.1086/500975 2006

  21. [21]

    2016, ApJ, 833, 81, doi: 10.3847/1538-4357/833/1/81

    Hasselquist, S., Shetrone, M., Cunha, K., et al. 2016, ApJ, 833, 81, doi: 10.3847/1538-4357/833/1/81

    work page doi:10.3847/1538-4357/833/1/81 2016

  22. [22]

    , keywords =

    Hasselquist, S., Hayes, C. R., Lian, J., et al. 2021, ApJ, 923, 172, doi: 10.3847/1538-4357/ac25f9

    work page doi:10.3847/1538-4357/ac25f9 2021

  23. [23]

    D., et al

    Haywood, M., Di Matteo, P., Lehnert, M. D., et al. 2018, ApJ, 863, 113, doi: 10.3847/1538-4357/aad235

    work page doi:10.3847/1538-4357/aad235 2018

  24. [24]

    2020, ARA&A, 58, 205, doi: 10.1146/annurev-astro-032620-021917

    Helmi, A. 2020, ARA&A, 58, 205, doi: 10.1146/annurev-astro-032620-021917

    work page doi:10.1146/annurev-astro-032620-021917 2020

  25. [25]

    H., et al

    Helmi, A., Babusiaux, C., Koppelman, H. H., et al. 2018, Nature, 563, 85, doi: 10.1038/s41586-018-0625-x

    work page doi:10.1038/s41586-018-0625-x 2018

  26. [26]

    P., Mackereth, J

    Horta, D., Schiavon, R. P., Mackereth, J. T., et al. 2021, MNRAS, 500, 1385, doi: 10.1093/mnras/staa2987

    work page doi:10.1093/mnras/staa2987 2021

  27. [27]

    P., Mackereth, J

    Horta, D., Schiavon, R. P., Mackereth, J. T., et al. 2023, MNRAS, 520, 5671, doi: 10.1093/mnras/stac3179

    work page doi:10.1093/mnras/stac3179 2023

  28. [28]

    Hunter, J. D. 2007, Computing in Science & Engineering, 9, 90, doi: 10.1109/MCSE.2007.55

    work page doi:10.1109/mcse.2007.55 2007

  29. [29]

    , year = 1994, month = jul, volume =

    Ibata, R. A., Gilmore, G., & Irwin, M. J. 1994, Nature, 370, 194, doi: 10.1038/370194a0

    work page doi:10.1038/370194a0 1994

  30. [30]

    P., Font, A

    Kisku, S., Schiavon, R. P., Font, A. S., et al. 2025, MNRAS, 542, 76, doi: 10.1093/mnras/staf1075

    work page doi:10.1093/mnras/staf1075 2025

  31. [31]

    SDSS-V: Pioneering Panoptic Spectroscopy

    Kollmeier, J. A., Zasowski, G., Rix, H.-W., et al. 2017, arXiv e-prints, arXiv:1711.03234, doi: 10.48550/arXiv.1711.03234

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1711.03234 2017

  32. [32]

    A., Rix, H.-W., Aerts, C., et al

    Kollmeier, J. A., Rix, H.-W., Aerts, C., et al. 2026, AJ, 171, 52, doi: 10.3847/1538-3881/ae0576

    work page doi:10.3847/1538-3881/ae0576 2026

  33. [33]

    H., Helmi, A., Massari, D., Price-Whelan, A

    Koppelman, H. H., Helmi, A., Massari, D., Price-Whelan, A. M., & Starkenburg, T. K. 2019, A&A, 631, L9, doi: 10.1051/0004-6361/201936738 L¨ ovdal, S. S., Ruiz-Lara, T., Koppelman, H. H., et al. 2022, A&A, 665, A57, doi: 10.1051/0004-6361/202243060 19

    work page doi:10.1051/0004-6361/201936738 2019

  34. [34]

    Dynamical heating across the Milky Way disc using APOGEE andGaia,

    Mackereth, J. T., Bovy, J., Leung, H. W., et al. 2019, MNRAS, 489, 176, doi: 10.1093/mnras/stz1521

    work page doi:10.1093/mnras/stz1521 2019

  35. [35]

    R., Schiavon, R

    Majewski, S. R., Schiavon, R. P., Frinchaboy, P. M., et al. 2017, AJ, 154, 94, doi: 10.3847/1538-3881/aa784d

    work page doi:10.3847/1538-3881/aa784d 2017

  36. [36]

    2024, ApJ, 964, 104, doi: 10.3847/1538-4357/ad1885

    Malhan, K., & Rix, H.-W. 2024, ApJ, 964, 104, doi: 10.3847/1538-4357/ad1885

    work page doi:10.3847/1538-4357/ad1885 2024

  37. [37]

    2019, ApJL, 874, L35, doi: 10.3847/2041-8213/ab0ec0

    Matsuno, T., Aoki, W., & Suda, T. 2019, ApJL, 874, L35, doi: 10.3847/2041-8213/ab0ec0

    work page doi:10.3847/2041-8213/ab0ec0 2019

  38. [38]

    Accelerated Hierarchical Density Clustering

    McInnes, L., & Healy, J. 2017, arXiv e-prints, arXiv:1705.07321, doi: 10.48550/arXiv.1705.07321

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1705.07321 2017

  39. [39]

    UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

    McInnes, L., Healy, J., & Melville, J. 2018, arXiv e-prints, arXiv:1802.03426, doi: 10.48550/arXiv.1802.03426

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1802.03426 2018

  40. [40]

    McMillan, P. J. 2017, MNRAS, 465, 76, doi: 10.1093/mnras/stw2759 M´ esz´ aros, S., Jofr´ e, P., Johnson, J. A., et al. 2025, AJ, 170, 96, doi: 10.3847/1538-3881/ade4b9

    work page internal anchor Pith review doi:10.1093/mnras/stw2759 2017

  41. [41]

    2019, MNRAS, 488, 1235, doi: 10.1093/mnras/stz1770

    Belokurov, V. 2019, MNRAS, 488, 1235, doi: 10.1093/mnras/stz1770

    work page doi:10.1093/mnras/stz1770 2019

  42. [42]

    , keywords =

    Naidu, R. P., Conroy, C., Bonaca, A., et al. 2020, ApJ, 901, 48, doi: 10.3847/1538-4357/abaef4

    work page doi:10.3847/1538-4357/abaef4 2020

  43. [43]

    2020, Nature Astronomy, 4, 1078, doi: 10.1038/s41550-020-1131-2

    Necib, L., Ostdiek, B., Lisanti, M., et al. 2020, Nature Astronomy, 4, 1078, doi: 10.1038/s41550-020-1131-2

    work page doi:10.1038/s41550-020-1131-2 2020

  44. [44]

    W., Campante, T

    Neitzel, A. W., Campante, T. L., Bossini, D., & Miglio, A. 2025, A&A, 695, A243, doi: 10.1051/0004-6361/202451718

    work page doi:10.1051/0004-6361/202451718 2025

  45. [45]

    2011, Journal of Machine Learning Research, 12, 2825

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825

    work page 2011

  46. [46]

    2025, ApJ, 991, 207, doi: 10.3847/1538-4357/adfd5e

    Plevne, O., & Akbaba, F. 2025, ApJ, 991, 207, doi: 10.3847/1538-4357/adfd5e

    work page doi:10.3847/1538-4357/adfd5e 2025

  47. [47]

    2022, ApJ, 941, 45, doi: 10.3847/1538-4357/ac9e01 Sch¨ onrich, R., Binney, J., & Dehnen, W

    Rix, H.-W., Chandra, V., Andrae, R., et al. 2022, ApJ, 941, 45, doi: 10.3847/1538-4357/ac9e01 Sch¨ onrich, R., Binney, J., & Dehnen, W. 2010, MNRAS, 403, 1829, doi: 10.1111/j.1365-2966.2010.16253.x

    work page doi:10.3847/1538-4357/ac9e01 2022

  48. [48]

    H., Wheeler, A., et al

    Sit, T., Weinberg, D. H., Wheeler, A., et al. 2024, ApJ, 970, 180, doi: 10.3847/1538-4357/ad4ed2

    work page doi:10.3847/1538-4357/ad4ed2 2024

  49. [49]

    V., Bizyaev, D., Cunha, K., et al

    Smith, V. V., Bizyaev, D., Cunha, K., et al. 2021, AJ, 161, 254, doi: 10.3847/1538-3881/abefdc

    work page doi:10.3847/1538-3881/abefdc 2021

  50. [50]

    2020, AJ, 160, 82, doi: 10.3847/1538-3881/ab9ab9

    Steinmetz, M., Matijeviˇ c, G., Enke, H., et al. 2020, AJ, 160, 82, doi: 10.3847/1538-3881/ab9ab9

    work page doi:10.3847/1538-3881/ab9ab9 2020

  51. [51]

    2015, ApJ, 807, 104, doi: 10.1088/0004-637X/807/1/104 Van Rossum, G., & Drake Jr, F

    Ting, Y.-S., Conroy, C., & Goodman, A. 2015, ApJ, 807, 104, doi: 10.1088/0004-637X/807/1/104 Van Rossum, G., & Drake Jr, F. L. 1995, Python reference manual (Centrum voor Wiskunde en Informatica Amsterdam)

    work page doi:10.1088/0004-637x/807/1/104 2015

  52. [52]

    , keywords =

    Vasiliev, E., & Baumgardt, H. 2021, MNRAS, 505, 5978, doi: 10.1093/mnras/stab1475

    work page doi:10.1093/mnras/stab1475 2021

  53. [53]

    H., Holtzman, J

    Weinberg, D. H., Holtzman, J. A., Johnson, J. A., et al. 2022, ApJS, 260, 32, doi: 10.3847/1538-4365/ac6028

    work page doi:10.3847/1538-4365/ac6028 2022

  54. [54]

    C., Hearty, F

    Wilson, J. C., Hearty, F. R., Skrutskie, M. F., et al. 2019, PASP, 131, 055001, doi: 10.1088/1538-3873/ab0075

    work page doi:10.1088/1538-3873/ab0075 2019

  55. [55]

    SEGUE: A Spectroscopic Survey of 240,000 stars with g=14-20

    Yanny, B., Rockosi, C., Newberg, H. J., et al. 2009, AJ, 137, 4377, doi: 10.1088/0004-6256/137/5/4377

    work page internal anchor Pith review doi:10.1088/0004-6256/137/5/4377 2009

  56. [56]

    2018, ApJ, 863, 26, doi: 10.3847/1538-4357/aacd0d

    Yuan, Z., Chang, J., Banerjee, P., et al. 2018, ApJ, 863, 26, doi: 10.3847/1538-4357/aacd0d

    work page doi:10.3847/1538-4357/aacd0d 2018

  57. [57]

    C., & Huang, Y

    Yuan, Z., Chang, J., Beers, T. C., & Huang, Y. 2020, ApJL, 898, L37, doi: 10.3847/2041-8213/aba49f

    work page doi:10.3847/2041-8213/aba49f 2020

  58. [58]

    2012, Research in Astronomy and Astrophysics, 12, 723, doi: 10.1088/1674-4527/12/7/002

    Zhao, G., Zhao, Y.-H., Chu, Y.-Q., Jing, Y.-P., & Deng, L.-C. 2012, Research in Astronomy and Astrophysics, 12, 723, doi: 10.1088/1674-4527/12/7/002

    work page doi:10.1088/1674-4527/12/7/002 2012