pith. sign in

arxiv: 2605.16493 · v1 · pith:SPDQBAVJnew · submitted 2026-05-15 · 🌌 astro-ph.SR · astro-ph.GA

Hot DQs, magnetic and metal-polluted white dwarfs: spectroscopic insights from a Gaia machine-learning-selected 500 pc sample

Enrique Miguel Garc\'ia Zamora , Santiago Torres Gil , Alberto Rebassa Mansergas , Aina Ferrer i Burjachs This is my paper

Pith reviewed 2026-05-19 21:23 UTC · model grok-4.3

classification 🌌 astro-ph.SR astro-ph.GA
keywords white dwarfsGaia spectramachine learningmagnetic white dwarfsDQ white dwarfsspectral classificationhelium-richmerger remnants
0
0 comments X

The pith

Machine-learning classifications from low-resolution Gaia spectra accurately identify white dwarf types, showing most massive DB candidates are magnetic white dwarfs or warm DQs instead of genuine helium-rich stars.

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

The paper tests machine-learning spectral classifications from Gaia data by obtaining medium-resolution spectra for 255 white dwarfs with the OSIRIS instrument on the Gran Telescopio Canarias. The automated labels match visual inspection more than 90 percent of the time for the main spectral types included in training sets. It further shows that objects labeled massive DB white dwarfs are mostly magnetic stars or warm carbon-dominated DQ stars, with only 5 of 112 observed confirmed as true DBs. Warm DQs appear along the Gaia Q branch and exhibit high tangential speeds consistent with a merger origin. The observations also identify dozens of magnetic white dwarfs and several rare subtypes including hot and warm DQs.

Core claim

By comparing Random Forest machine-learning classifications applied to Gaia spectra with visual spectral typing from medium-resolution OSIRIS observations, the study establishes that the algorithms achieve over 90 percent accuracy for spectral types in their training sets despite the low resolution. It determines that only 4.46 percent of the observed massive DB candidates are genuine helium-rich DB white dwarfs, while the remainder consist primarily of magnetic white dwarfs identified via Zeeman splitting and warm DQ white dwarfs. The properties of the warm DQs align with prior work supporting a merger-remnant origin for these carbon-rich objects.

What carries the argument

Cross-validation of Random Forest classifications on low-resolution Gaia spectra against visual inspection of medium-resolution OSIRIS spectra, with magnetic fields identified through Zeeman splitting.

Load-bearing premise

Visual inspection of the medium-resolution spectra provides an accurate and objective ground truth for spectral types without significant subjectivity or selection bias.

What would settle it

A larger follow-up spectroscopic survey of similar massive DB candidates that finds a substantially higher fraction confirmed as genuine DB white dwarfs would challenge the conclusion that most are misclassified magnetic or DQ objects.

Figures

Figures reproduced from arXiv: 2605.16493 by Aina Ferrer i Burjachs, Alberto Rebassa Mansergas, Enrique Miguel Garc\'ia Zamora, Santiago Torres Gil.

Figure 1
Figure 1. Figure 1: Gaia HR diagram showing the targets observed at the GTC during the 2023B, 2024B, 2025A and 2025B semesters, su￾perimposed on the Gaia 100 pc white dwarf sample of Jiménez￾Esteban et al. (2023). spectral classification. The GTC spectra were reduced using the pamela (Marsh 2014) and molly (Marsh 2019)1 packages. The Gaia HR diagram of the observed targets from each campaign, superimposed on the 100 pc Gaia w… view at source ↗
Figure 2
Figure 2. Figure 2: Observed versus predicted (García-Zamora et al. 2025) [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Colour-tangential velocity diagram for our warm DQs. It [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Hβ Zeeman splitting in DAH J1927+6938, where Hβ splits into seven components. We also performed a first-order estimate of the magnetic field strengths by fitting Gaussian profiles to the visible components of the Hα and Hβ lines. The line center was considered to lie at the mean of the Gaussian fits, and the spectral displacement was computed as the wavelength difference between the lateral and central pea… view at source ↗
read the original abstract

The latest Gaia data release provides low-resolution spectra for approximately 100 000 white dwarfs. Though useful for pre-classification, they lack the resolution required for accurate spectral type and parameter determination, motivating spectroscopic follow-up campaigns. In this work, we assess the reliability of machine-learning spectral classifications derived from Gaia spectra through comparison with medium-resolution spectroscopy, determine the nature of objects classified as "massive helium-rich (DB)" by automated methods, and characterise the properties of warm and hot DQ (carbon-dominated) white dwarfs, magnetic and metal-polluted objects. To do this, we observed 255 white dwarfs with the Gran Telescopio Canarias equipped with the OSIRIS instrument (R ~ 1000). Spectral types were assigned through visual inspection and compared with machine-learning classifications applied to Gaia spectra. Magnetic objects were identified via Zeeman splitting, and magnetic field strengths were estimated. We find machine-learning classifications are highly accurate (> 90% for spectral types in their training sets), despite the low resolution of Gaia spectra. We show "massive DBs" to be mostly magnetic white dwarfs and warm DQs, with only 5 of 112 observed (4.46%) confirmed as genuine DBs. Warm DQs are found along the Gaia Q branch and exhibit unusually high tangential speeds. We provide spectral classifications for 255 white dwarfs, demonstrate that Random Forest algorithms reliably classify low-resolution Gaia spectra into main spectral types, determine the nature of "massive DBs", and identify a large population of magnetic white dwarfs and carbon-rich objects. Several rare subtypes are identified, including 1 DAQ, 1 DQZA, 4 hot, 29 warm DQ stars, and 63 magnetic white dwarfs. The properties of warm DQs are consistent with previous studies, supporting their proposed origin as merger remnants.

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

2 major / 2 minor

Summary. The manuscript reports medium-resolution (R~1000) OSIRIS spectroscopy of 255 Gaia-selected white dwarfs within 500 pc. Visual spectral types assigned by inspection are compared to machine-learning classifications derived from low-resolution Gaia spectra, yielding a claimed accuracy >90% for spectral types in the training sets. The authors reclassify 112 objects previously labeled 'massive DBs', finding only 5 (4.46%) to be genuine DBs while the majority are magnetic white dwarfs or warm DQs; they also characterize warm DQs along the Gaia Q branch (high tangential velocities), identify 63 magnetic objects, 29 warm DQs, 4 hot DQs, and several rare subtypes (1 DAQ, 1 DQZA), and argue that the warm-DQ properties support a merger-remnant origin.

Significance. If the visual classifications are robust, the work validates Random Forest methods for pre-classifying the ~100 000 Gaia white-dwarf spectra and clarifies the composition of the 'massive DB' population, with direct implications for white-dwarf merger rates and the origin of the Q branch. The addition of 63 newly identified magnetic white dwarfs and a statistically useful warm-DQ sample strengthens the observational basis for evolutionary models of carbon-rich and magnetic objects.

major comments (2)
  1. [Methods (spectral typing) and Results (ML comparison and DB reclassification)] The headline accuracy claim (>90%) and the reclassification result (only 5/112 genuine DBs) rest entirely on visual inspection of the R~1000 OSIRIS spectra serving as ground truth. No inter-rater reliability metric (Cohen's kappa or equivalent), no multiple independent classifiers, and no cross-check against a subset of higher-resolution (R>5000) spectra are reported. Small systematic shifts in line-strength or Zeeman-feature judgments would directly change both the agreement fraction and the DB confirmation rate. This is load-bearing for the central claims.
  2. [Sample selection and observational details] The 255-object sample and the 112-object 'massive DB' subsample lack explicit documentation of the parent Gaia ML selection criteria, any post-observation cuts, magnitude or color limits, and error bars on derived parameters (e.g., tangential velocity, magnetic field strength). Without these, it is impossible to assess selection biases or the robustness of the reported fractions and the warm-DQ kinematic properties.
minor comments (2)
  1. [Results] A summary table listing all 255 objects with Gaia ML label, OSIRIS visual type, and key features (magnetic field estimate, C2 strength, metals) would greatly improve readability and allow readers to reproduce the reclassification statistics.
  2. [Abstract and Results] The abstract states 'machine-learning classifications are highly accurate (>90% for spectral types in their training sets)'; the precise definition of 'training sets' and whether the quoted figure includes or excludes the rare subtypes should be clarified in the text.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive report. We address the major comments below and have revised the manuscript accordingly to improve clarity and robustness.

read point-by-point responses

  1. Referee: [Methods (spectral typing) and Results (ML comparison and DB reclassification)] The headline accuracy claim (>90%) and the reclassification result (only 5/112 genuine DBs) rest entirely on visual inspection of the R~1000 OSIRIS spectra serving as ground truth. No inter-rater reliability metric (Cohen's kappa or equivalent), no multiple independent classifiers, and no cross-check against a subset of higher-resolution (R>5000) spectra are reported. Small systematic shifts in line-strength or Zeeman-feature judgments would directly change both the agreement fraction and the DB confirmation rate. This is load-bearing for the central claims.

    Authors: We appreciate the referee's emphasis on the importance of validating our visual spectral classifications. The classifications were performed by the authors using standard visual inspection techniques for white dwarf spectra at this resolution, consistent with methods used in prior studies. Although we did not report inter-rater reliability metrics or employ multiple independent classifiers, nor did we cross-check with higher-resolution spectra, the excellent agreement with the machine learning classifications provides supporting evidence for the reliability. We will revise the manuscript to include a more detailed description of the spectral typing procedure in the Methods section and explicitly state the limitations regarding the lack of formal reliability metrics. This will strengthen the presentation without changing the core results. revision: partial

  2. Referee: [Sample selection and observational details] The 255-object sample and the 112-object 'massive DB' subsample lack explicit documentation of the parent Gaia ML selection criteria, any post-observation cuts, magnitude or color limits, and error bars on derived parameters (e.g., tangential velocity, magnetic field strength). Without these, it is impossible to assess selection biases or the robustness of the reported fractions and the warm-DQ kinematic properties.

    Authors: We agree that clearer documentation of the sample selection is necessary. The 255 targets were chosen from the Gaia ML-selected white dwarf catalog within 500 pc, with the specific criteria outlined in the paper's introduction. We will add explicit details on the parent sample selection, including any magnitude or color cuts, post-observation filters applied, and uncertainties or error bars on parameters such as tangential velocities and estimated magnetic field strengths. A new paragraph or subsection will be included in the revised Methods to facilitate assessment of selection effects and the robustness of our findings on warm DQs and other subpopulations. revision: yes

Circularity Check

0 steps flagged

No circularity: accuracy and reclassification rest on independent spectroscopic follow-up

full rationale

The paper's central results—ML classification accuracy >90% and the finding that only 5/112 'massive DBs' are genuine—are obtained by applying Random Forest models to Gaia spectra and then comparing those outputs against new medium-resolution OSIRIS spectra (R~1000) whose types were assigned by visual inspection. This comparison uses fresh observational data external to the Gaia inputs and to any prior training sets; no equation, fitted parameter, or self-citation is shown to reduce the reported agreement fraction or DB confirmation rate back to the original ML labels by construction. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard observational techniques and established white-dwarf classification schemes rather than new theoretical constructs or fitted parameters.

axioms (1)
  • domain assumption White dwarf spectral types can be reliably determined by visual inspection of medium-resolution optical spectra.
    This premise enables the direct comparison between machine-learning labels and the new observations.

pith-pipeline@v0.9.0 · 5895 in / 1376 out tokens · 62134 ms · 2026-05-19T21:23:29.290215+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages

  1. [1]

    G., Córsico, A

    Althaus, L. G., Córsico, A. H., Isern, J., & García-Berro, E. 2010, A&A Rev., 18, 471 Bédard, A. 2024, Ap&SS, 369, 43 Bédard, A., Bergeron, P., & Brassard, P. 2022a, ApJ, 930, 8 Bédard, A., Brassard, P., Bergeron, P., & Blouin, S. 2022b, ApJ, 927, 128

    work page 2010

  2. [2]

    2019, ApJ, 876, 67

    Bergeron, P., Dufour, P., Fontaine, G., et al. 2019, ApJ, 876, 67

    work page 2019

  3. [3]

    2011, ApJ, 737, 28

    Bergeron, P., Wesemael, F., Dufour, P., et al. 2011, ApJ, 737, 28

    work page 2011

  4. [4]

    2023, MNRAS, 523, 3363

    Blouin, S., Bédard, A., & Tremblay, P.-E. 2023, MNRAS, 523, 3363

    work page 2023

  5. [5]

    Blouin, S., Dufour, P., Thibeault, C., & Allard, N. F. 2019, ApJ, 878, 63

    work page 2019

  6. [6]

    A., Torres, S., & Rebassa-Mansergas, A

    Camisassa, M., Baiko, D. A., Torres, S., & Rebassa-Mansergas, A. 2024, A&A, 683, A101

    work page 2024

  7. [7]

    2023, A&A, 674, A213

    Camisassa, M., Torres, S., Hollands, M., et al. 2023, A&A, 674, A213

    work page 2023

  8. [8]

    E., Althaus, L

    Camisassa, M. E., Althaus, L. G., Córsico, A. H., et al. 2019, A&A, 625, A87

    work page 2019

  9. [9]

    M., Weiler, M., Jordi, C., et al

    Carrasco, J. M., Weiler, M., Jordi, C., et al. 2021, A&A, 652, A86

    work page 2021

  10. [10]

    M., et al

    Cepa, J., Bongiovanni, A., Pérez García, A. M., et al. 2013, in Highlights of Spanish Astrophysics VII, ed. J. C. Guirado, L. M. Lara, V . Quilis, & J. Gor- gas, 868–873 Córsico, A. H., Althaus, L. G., Gil-Pons, P., & Torres, S. 2021, A&A, 646, A60

    work page 2013

  11. [11]

    2019, ApJ, 885, 74

    Coutu, S., Dufour, P., Bergeron, P., et al. 2019, ApJ, 885, 74

    work page 2019

  12. [12]

    2020, Monthly Notices of the Royal Astronomical Society, 492, 3540 De Angeli, F., Weiler, M., Montegriffo, P., et al

    Cukanovaite, E. 2020, Monthly Notices of the Royal Astronomical Society, 492, 3540 De Angeli, F., Weiler, M., Montegriffo, P., et al. 2023, A&A, 674, A2 de Jong, R. S., Agertz, O., Berbel, A. A., et al. 2019, The Messenger, 175, 3

    work page 2020

  13. [13]

    2017, in Astronomical Society of the Pacific Conference Series, V ol

    Dufour, P., Blouin, S., Coutu, S., et al. 2017, in Astronomical Society of the Pacific Conference Series, V ol. 509, 20th European White Dwarf Workshop, ed. P. E. Tremblay, B. Gaensicke, & T. Marsh, 3

    work page 2017

  14. [14]

    D., & Behara, N

    Dufour, P., Fontaine, G., Liebert, J., Schmidt, G. D., & Behara, N. 2008, ApJ, 683, 978

    work page 2008

  15. [15]

    2007, Nature, 450, 522

    Dufour, P., Liebert, J., Fontaine, G., & Behara, N. 2007, Nature, 450, 522

    work page 2007

  16. [16]

    2013, in Astronomi- cal Society of the Pacific Conference Series, V ol

    Dufour, P., V ornanen, T., Bergeron, P., & Fontaine, A., B. 2013, in Astronomi- cal Society of the Pacific Conference Series, V ol. 469, 18th European White Dwarf Workshop., ed. J. Krzesi´nski, G. Stachowski, P. Moskalik, & K. Bajan, 167

    work page 2013

  17. [17]

    Dunlap, B. H. & Clemens, J. C. 2015, in Astronomical Society of the Pacific Conference Series, V ol. 493, 19th European Workshop on White Dwarfs, ed. P. Dufour, P. Bergeron, & G. Fontaine, 547

    work page 2015

  18. [18]

    A., Redfield, S., Dufour, P., & Hambly, N

    Farihi, J., Barstow, M. A., Redfield, S., Dufour, P., & Hambly, N. C. 2010, MN- RAS, 404, 2123

    work page 2010

  19. [19]

    & Dufour, P

    Fortier, A. & Dufour, P. 2015, in Astronomical Society of the Pacific Conference

    work page 2015

  20. [20]

    493, 19th European Workshop on White Dwarfs, ed

    Series, V ol. 493, 19th European Workshop on White Dwarfs, ed. P. Dufour, P. Bergeron, & G. Fontaine, 41 Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al. 2016, A&A, 595, A1 Gänsicke, B. T., Koester, D., Girven, J., Marsh, T. R., & Steeghs, D. 2010, Sci- ence, 327, 188 García-Zamora, E. M., Torres, S., & Rebassa-Mansergas, A. 2023, A&A, 679, A1...

    work page 2016

  21. [21]

    & Bergeron, P

    Genest-Beaulieu, C. & Bergeron, P. 2019, ApJ, 871, 169 Gentile Fusillo, N. P., Gänsicke, B. T., Farihi, J., et al. 2017, MNRAS, 468, 971

    work page 2019

  22. [22]

    Hall, P. B. & Maxwell, A. J. 2008, ApJ, 678, 1292

    work page 2008

  23. [23]

    A., Stopkowicz, S., Kitsaras, M.-P., et al

    Hollands, M. A., Stopkowicz, S., Kitsaras, M.-P., et al. 2023, MNRAS, 520, 3560

    work page 2023

  24. [24]

    A., Tremblay, P.-E., Gänsicke, B

    Hollands, M. A., Tremblay, P.-E., Gänsicke, B. T., et al. 2020, Nature Astronomy, 4, 663

    work page 2020

  25. [25]

    A., Tremblay, P

    Hollands, M. A., Tremblay, P. E., Gänsicke, B. T., Gentile-Fusillo, N. P., & Too- nen, S. 2018, MNRAS, 480, 3942

    work page 2018

  26. [26]

    2024, ApJ, 974, 12 Jiménez-Esteban, F

    Jewett, G., Kilic, M., Bergeron, P., et al. 2024, ApJ, 974, 12 Jiménez-Esteban, F. M., Torres, S., Rebassa-Mansergas, A., et al. 2023, MN- RAS, 518, 5106

    work page 2024

  27. [27]

    C., Dalton, G

    Jin, S., Trager, S. C., Dalton, G. B., et al. 2024, MNRAS, 530, 2688

    work page 2024

  28. [28]

    Jura, M. & Xu, S. 2012, AJ, 143, 6

    work page 2012

  29. [29]

    2023, MNRAS, 520, 6299

    Kawka, A., Ferrario, L., & Vennes, S. 2023, MNRAS, 520, 6299

    work page 2023

  30. [30]

    O., Pelisoli, I., Koester, D., et al

    Kepler, S. O., Pelisoli, I., Koester, D., et al. 2015, MNRAS, 446, 4078

    work page 2015

  31. [31]

    2024, ApJ, 965, 159

    Kilic, M., Bergeron, P., Blouin, S., et al. 2024, ApJ, 965, 159

    work page 2024

  32. [32]

    R., et al

    Kilic, M., Bergeron, P., Brown, W. R., et al. 2025, ApJ, 990, 62

    work page 2025

  33. [33]

    2026, arXiv e-prints, arXiv:2602.07246

    Kilic, M., Bergeron, P., Moss, A., et al. 2026, arXiv e-prints, arXiv:2602.07246

    work page arXiv 2026

  34. [34]

    & Kepler, S

    Koester, D. & Kepler, S. O. 2019, A&A, 628, A102

    work page 2019

  35. [35]

    1982, A&A, 116, 147

    Koester, D., Weidemann, V ., & Zeidler, E.-M. 1982, A&A, 116, 147

    work page 1982

  36. [36]

    E., Raichoor, A., et al

    Levi, M., Allen, L. E., Raichoor, A., et al. 2019, in Bulletin of the American Astronomical Society, V ol. 51, 57

    work page 2019

  37. [37]

    C., Dahn, C

    Liebert, J., Harris, H. C., Dahn, C. C., et al. 2003, AJ, 126, 2521

    work page 2003

  38. [38]

    2019, molly: 1D astronomical spectra analyzer, Astrophysics Source Code Library, record ascl:1907.012

    Marsh, T. 2019, molly: 1D astronomical spectra analyzer, Astrophysics Source Code Library, record ascl:1907.012

    work page 2019

  39. [39]

    Marsh, T. R. 2014, PAMELA: Optimal extraction code for long-slit CCD spec- troscopy, Astrophysics Source Code Library, record ascl:1406.002 O’Brien, M. W., Tremblay, P. E., Gentile Fusillo, N. P., et al. 2023, MNRAS, 518, 3055 Ould Rouis, L. B., Hermes, J. J., Guidry, J. A., et al. 2026, arXiv e-prints, arXiv:2602.02670

    work page arXiv 2014

  40. [40]

    1986, ApJ, 307, 242 Pérez-Couto, X., Pallas-Quintela, L., Manteiga, M., Villaver, E., & Dafonte, C

    Pelletier, C., Fontaine, G., Wesemael, F., Michaud, G., & Wegner, G. 1986, ApJ, 307, 242 Pérez-Couto, X., Pallas-Quintela, L., Manteiga, M., Villaver, E., & Dafonte, C. 2024, ApJ, 977, 31

    work page 1986

  41. [41]

    N., Liebert, J., & Schmidt, G

    Reid, I. N., Liebert, J., & Schmidt, G. D. 2001, ApJ, 550, L61

    work page 2001

  42. [42]

    M., Greenstein, J

    Sion, E. M., Greenstein, J. L., Landstreet, J. D., et al. 1983, ApJ, 269, 253

    work page 1983

  43. [43]

    & Bues, I

    Unglaub, K. & Bues, I. 2000, A&A, 359, 1042

    work page 2000

  44. [44]

    A., Jordan, S., et al

    Vincent, O., Barstow, M. A., Jordan, S., et al. 2024, A&A, 682, A5

    work page 2024

  45. [45]

    2023, MNRAS, 521, 760

    Vincent, O., Bergeron, P., & Dufour, P. 2023, MNRAS, 521, 760

    work page 2023

  46. [46]

    1995, Cataclysmic variable stars, V ol

    Warner, B. 1995, Cataclysmic variable stars, V ol. 28

    work page 1995

  47. [47]

    & Phinney, E

    Wegg, C. & Phinney, E. S. 2012, MNRAS, 426, 427

    work page 2012

  48. [48]

    1983, ApJS, 53, 523

    Williams, G. 1983, ApJS, 53, 523

    work page 1983

  49. [49]

    A., Winget, D

    Williams, K. A., Winget, D. E., Montgomery, M. H., et al. 2013, ApJ, 769, 123

    work page 2013

  50. [50]

    Williams, R. E. 1980, ApJ, 235, 939

    work page 1980

  51. [51]

    M., & Jura, M

    Zuckerman, B., Koester, D., Melis, C., Hansen, B. M., & Jura, M. 2007, ApJ, 671, 872 Article number, page 12 of 17 García-Zamora et al.: Hot DQs, magnetic and metal-polluted white dwarfs in aGaiasample Appendix A:Gaiacolour-magnitude diagrams for the visually identified spectral types and subtypes ¡0:5 0 0:5 1:0 1:5 BP ¡ RP 8 10 12 14 16 MG DAs DABs DAHs ...

    work page 2007