pith. sign in

arxiv: 2606.13548 · v1 · pith:46VQV64Ynew · submitted 2026-06-11 · ❄️ cond-mat.mtrl-sci · physics.data-an· stat.ML

Symmetry-electronic fingerprints reveal competing magnetic phases in two-dimensional materials

Addis Fuhr , Zachary R. Fox , David Parker , Ayana Ghosh This is my paper

Pith reviewed 2026-06-27 05:59 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci physics.data-anstat.ML
keywords symmetry-electronic fingerprinttwo-dimensional magnetsmagnetic orderingmachine learningcompeting phasesStoner ferromagnetismsuperexchangemagnetic frustration
0
0 comments X

The pith

Symmetry-electronic fingerprints allow machine learning models to classify magnetic ordering in 2D materials and use uncertainty to identify competing ferromagnetic and antiferromagnetic phases.

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

The paper introduces the symmetry-electronic fingerprint representation to encode crystallographic symmetry operations, Wyckoff-site geometry, and site-resolved electronic structure for magnetism predictions in two-dimensional materials. Combined with random forest ensemble learning, the approach classifies magnetic ordering, regresses moments and anisotropy energies, and distinguishes itinerant Stoner ferromagnetism from localized superexchange. Elevated model uncertainty is interpreted as a signal of materials where these mechanisms compete. First-principles calculations on Co- and Ni-based halides and oxides confirm that high-uncertainty regions correspond to near-degenerate FM and AFM phases with frustration and suppressed anisotropy. This turns model uncertainty into a diagnostic for materials where small perturbations can drive magnetic phase transitions.

Core claim

The symmetry-electronic fingerprint (SEF) encodes crystallographic symmetry operations, Wyckoff-site geometry, together with site-resolved electronic structure. When paired with ensemble random forest learning, SEF-trained models accurately classify magnetic ordering while regressing moments alongside anisotropy energies while simultaneously resolving the distinct regimes of itinerant Stoner ferromagnetism from localized superexchange. Regions of elevated model uncertainty identify materials where these mechanisms compete, corresponding to genuine near-degenerate FM and AFM phases with magnetic frustration, suppressed anisotropy, and emergent non-collinear ordering as validated by first-prin

What carries the argument

The symmetry-electronic fingerprint (SEF), a representation that encodes crystallographic symmetry operations, Wyckoff-site geometry, and site-resolved electronic structure to capture exchange physics for machine learning models of magnetism.

If this is right

  • Models classify magnetic ordering accurately in two-dimensional materials.
  • Models regress magnetic moments and anisotropy energies from the SEF representation.
  • Models resolve distinct regimes of itinerant Stoner ferromagnetism from localized superexchange.
  • Elevated uncertainty identifies materials with near-degenerate FM and AFM phases.
  • These phases exhibit magnetic frustration, suppressed anisotropy, and potential non-collinear ordering.

Where Pith is reading between the lines

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

  • The SEF approach could extend to databases of 2D materials beyond the Co- and Ni-based compounds used for validation.
  • High-uncertainty materials may respond to external perturbations like strain or doping to switch between collinear and non-collinear states.
  • The uncertainty-as-diagnostic strategy might apply to other competing order problems such as charge density waves or superconductivity.
  • Integration with additional electronic structure features could refine predictions of anisotropy in frustrated systems.

Load-bearing premise

Elevated model uncertainty in the SEF-random forest models reliably signals genuine physical competition between magnetic phases rather than data sparsity or model limitations.

What would settle it

First-principles calculations on a high-uncertainty material that find a large energy separation between ferromagnetic and antiferromagnetic states instead of near-degeneracy would falsify the claim that uncertainty indicates competing phases.

Figures

Figures reproduced from arXiv: 2606.13548 by Addis Fuhr, Ayana Ghosh, David Parker, Zachary R. Fox.

Figure 1
Figure 1. Figure 1: FIG. 1. Key steps in predicting and understanding magnetic properties of two-dimensional materials using the symmetry [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Symmetry (a) and electronic (b) fingerprints. The major symmetry operations are shown for the symmetry fingerprint [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Distribution of magnetic states in the C2DB database (a). The classification task is limited to the FM and NM [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. (a) Relative total energies of ferromagnetic (FM) and antiferromagnetic (AFM1–AFM3) configurations referenced to [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. (a-c) Luttinger–Tisza exchange-energy landscapes for NiCl [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Performance of regression models trained with [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. SOM map of feature space divided into regions based on magnetic strength ( [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
read the original abstract

Two-dimensional magnets offer compelling platforms for spintronics and quantum technologies, yet predicting their magnetic ground states, moments, and anisotropy remains challenging. This limitation primarily arises because existing machine-learning representations encode chemical environments without capturing the symmetry or exchange physics that govern magnetism. In this work, we introduce the symmetry-electronic fingerprint (SEF), a physically interpretable representation that encodes crystallographic symmetry operations, Wyckoff-site geometry, together with site-resolved electronic structure. Combined with ensemble learning with random forests, the SEF accurately classifies magnetic ordering while regressing moments alongside anisotropy energies while simultaneously resolving the distinct regimes of itinerant Stoner ferromagnetism from localized superexchange. What sets the SEF-trained models apart is that regions of elevated model uncertainty are not a failure but a diagnostic, identifying materials where these mechanisms compete. First-principles calculations on Co- and Ni-based halides and oxides confirm that these regions correspond to genuine near-degenerate FM and AFM phases with magnetic frustration, suppressed anisotropy, and emergent non-collinear ordering. By encoding symmetry together with exchange physics directly into the representation unlike conventional descriptors, the SEF transforms model uncertainty into a compass pointing toward two-dimensional materials where small perturbations drive transitions between collinear, frustrated, or non-collinear magnetic phases.

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 / 0 minor

Summary. The manuscript introduces the symmetry-electronic fingerprint (SEF), a representation that encodes crystallographic symmetry operations, Wyckoff-site geometry, and site-resolved electronic structure for two-dimensional materials. Combined with random-forest ensemble learning, the SEF is claimed to classify magnetic ordering, regress magnetic moments and anisotropy energies, distinguish itinerant Stoner ferromagnetism from localized superexchange, and treat elevated model uncertainty as a diagnostic for materials exhibiting competing magnetic phases; first-principles calculations on Co- and Ni-based halides and oxides are presented as confirmation that high-uncertainty regions correspond to near-degenerate FM/AFM states with frustration and suppressed anisotropy.

Significance. If the quantitative performance and uncertainty interpretation hold under rigorous validation, the SEF would provide a physically motivated descriptor that turns ensemble variance into a pointer toward 2D materials with tunable or frustrated magnetism. The explicit incorporation of symmetry and exchange physics into the representation is a conceptual strength relative to purely chemical-environment descriptors.

major comments (2)
  1. [Abstract] Abstract: the central claims of accurate classification, regression, mechanism resolution, and diagnostic use of uncertainty are asserted without any reported quantitative metrics (accuracy, R², MAE, error bars), dataset size, train/test split protocol, or cross-validation details, rendering it impossible to assess whether the SEF-RF models actually outperform conventional descriptors or whether the uncertainty signal is load-bearing.
  2. [Validation] Validation (implied results section): first-principles checks are restricted to Co- and Ni-based halides/oxides; no quantitative correlation (e.g., uncertainty vs. DFT FM-AFM energy difference), no ablation on training-set density, and no test compounds outside this chemistry are provided, so the claim that elevated uncertainty specifically flags physical competition rather than data sparsity or extrapolation remains untested.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. The comments highlight important areas for clarification and strengthening of the presentation. We respond to each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claims of accurate classification, regression, mechanism resolution, and diagnostic use of uncertainty are asserted without any reported quantitative metrics (accuracy, R², MAE, error bars), dataset size, train/test split protocol, or cross-validation details, rendering it impossible to assess whether the SEF-RF models actually outperform conventional descriptors or whether the uncertainty signal is load-bearing.

    Authors: We agree that the abstract should include quantitative metrics to support the claims. In the revised manuscript we will update the abstract to report classification accuracy, regression R² and MAE values, dataset size, train/test split details, and cross-validation protocol. revision: yes

  2. Referee: [Validation] Validation (implied results section): first-principles checks are restricted to Co- and Ni-based halides/oxides; no quantitative correlation (e.g., uncertainty vs. DFT FM-AFM energy difference), no ablation on training-set density, and no test compounds outside this chemistry are provided, so the claim that elevated uncertainty specifically flags physical competition rather than data sparsity or extrapolation remains untested.

    Authors: The validation is focused on Co- and Ni-based systems because these permit direct first-principles confirmation of the physical meaning of high-uncertainty predictions. We will add a quantitative correlation between model uncertainty and DFT FM-AFM energy differences. We acknowledge the absence of ablation studies and external-chemistry tests as a limitation of the current scope and will expand the discussion section to address generalizability while retaining the core physical demonstration within the studied chemistry. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper defines the SEF representation from crystallographic symmetry, Wyckoff geometry, and site-resolved electronic structure, then trains random-forest ensembles on this descriptor against external first-principles data for classification of magnetic order, regression of moments and anisotropy, and use of ensemble variance as a diagnostic. No equation or workflow step reduces a claimed prediction or uniqueness result to a quantity defined in terms of the model's own fitted outputs; the validation on Co/Ni halides and oxides is presented as an independent first-principles check rather than a tautology. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Assessment based solely on abstract; full details on training data, feature construction, and model hyperparameters unavailable.

axioms (1)
  • domain assumption Machine learning models trained on symmetry-plus-electronic representations can distinguish itinerant Stoner ferromagnetism from localized superexchange mechanisms.
    Invoked when claiming the SEF resolves distinct regimes of magnetism.
invented entities (1)
  • Symmetry-electronic fingerprint (SEF) no independent evidence
    purpose: Physically interpretable input representation for ML models that encodes crystallographic symmetry, Wyckoff geometry, and site-resolved electronic structure.
    New representation introduced to address limitations of existing descriptors.

pith-pipeline@v0.9.1-grok · 5765 in / 1360 out tokens · 24246 ms · 2026-06-27T05:59:52.091262+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

33 extracted references · 1 linked inside Pith

  1. [27]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Computational materials science6, 15 (1996). 16

    1996

  2. [31]

    M.-P. Miao, N. Liu, W.-H. Zhang, J.-W. Zhou, D.-B. Wang, C. Wang, W. Ji, and Y.-S. Fu, Proceedings of the National Academy of Sciences122, e2422868122 (2025). ∗ fuhras@ornl.gov † ghosha@iitm.ac.in thebibliography31

    2025

  3. [32]

    Gong and X

    C. Gong and X. Zhang, Science363, eaav4450 (2019)

    2019

  4. [33]

    Gibertini, M

    M. Gibertini, M. Koperski, A. F. Morpurgo, and K. S. Novoselov, Nature nanotechnology14, 408 (2019)

    2019

  5. [34]

    Torelli, K

    D. Torelli, K. S. Thygesen, and T. Olsen, 2D Materials 6, 045018 (2019)

    2019

  6. [35]

    T. D. Rhone, W. Chen, S. Desai, S. B. Torrisi, D. T. Larson, A. Yacoby, and E. Kaxiras, Scientific Reports 10, 15795 (2020)

    2020

  7. [36]

    S.-O. Kaba, B. Groleau-Par´ e, M.-A. Gauthier, A.-M. Tremblay, S. Verret, and C. Gauvin-Ndiaye, Physical Review Materials7, 044407 (2023)

    2023

  8. [37]

    Ghosh, F

    A. Ghosh, F. Ronning, S. M. Nakhmanson, and J.-X. Zhu, Physical Review Materials4, 064414 (2020)

    2020

  9. [38]

    M. K. Horton, J. H. Montoya, M. Liu, and K. A. Persson, npj Computational Materials5, 64 (2019)

    2019

  10. [39]

    Y. Xie, G. A. Tritsaris, O. Gr˚ an¨ as, and T. D. Rhone, The Journal of Physical Chemistry Letters12, 12048 (2021)

    2021

  11. [40]

    C. M. Acosta, E. Ogoshi, J. A. Souza, and G. M. Dalpian, ACS Applied Materials & Interfaces14, 9418 (2022)

    2022

  12. [41]

    Singh, P

    S. Singh, P. Joshi, A. Sharma, and A. Kashyap, Journal of Magnetism and Magnetic Materials , 173541 (2025)

    2025

  13. [42]

    Elrashidy, J

    A. Elrashidy, J. Della-Giustina, and J.-A. Yan, The Journal of Physical Chemistry C128, 6007 (2024)

    2024

  14. [43]

    Chotrattanapituk, R

    A. Chotrattanapituk, R. Okabe, E. Rha, M. Al-Hinai, E. Jiang, D. Pajerowski, Y. Cheng, J. J. Turner, and M. Li, arXiv preprint arXiv:2605.16230 (2026)

    Pith/arXiv arXiv 2026

  15. [44]

    Litvin, Foundations of Crystallography64, 419 (2008)

    D. Litvin, Foundations of Crystallography64, 419 (2008)

    2008

  16. [45]

    Bruno, Physical Review B39, 865 (1989)

    P. Bruno, Physical Review B39, 865 (1989)

    1989

  17. [46]

    Dzyaloshinsky, Journal of physics and chemistry of solids4, 241 (1958)

    I. Dzyaloshinsky, Journal of physics and chemistry of solids4, 241 (1958)

    1958

  18. [47]

    P. W. Anderson, Physical Review79, 350 (1950)

    1950

  19. [48]

    J. B. Goodenough, Physical Review100, 564 (1955)

    1955

  20. [49]

    Kanamori, Journal of Physics and Chemistry of Solids 10, 87 (1959)

    J. Kanamori, Journal of Physics and Chemistry of Solids 10, 87 (1959)

    1959

  21. [50]

    Oviedo, J

    F. Oviedo, J. L. Ferres, T. Buonassisi, and K. T. Butler, Accounts of Materials Research3, 597 (2022)

    2022

  22. [51]

    Zhong, B

    X. Zhong, B. Gallagher, S. Liu, B. Kailkhura, A. Hisz- panski, and T. Y.-J. Han, npj computational materials 8, 204 (2022)

    2022

  23. [52]

    Kohonen,Self-organizing maps, Vol

    T. Kohonen,Self-organizing maps, Vol. 30 (Springer Sci- ence & Business Media, 2012)

    2012

  24. [53]

    Haastrup, M

    S. Haastrup, M. Strange, M. Pandey, T. Deilmann, P. S. Schmidt, N. F. Hinsche, M. N. Gjerding, D. Torelli, P. M. Larsen, A. C. Riis-Jensen,et al., 2D Materials5, 042002 (2018)

    2018

  25. [54]

    A. Togo, K. Shinohara, and I. Tanaka, Science and Technology of Advanced Materials: Methods4, 2384822 (2024)

    2024

  26. [55]

    S. P. Ong, W. D. Richards, A. Jain, G. Hautier, M. Kocher, S. Cholia, D. Gunter, V. L. Chevrier, K. A. Persson, and G. Ceder, Computational Materials Science 68, 314 (2013)

    2013

  27. [56]

    D. E. Sands,Introduction to crystallography(Courier Corporation, 1993)

    1993

  28. [57]

    Breiman, Machine learning45, 5 (2001)

    L. Breiman, Machine learning45, 5 (2001)

    2001

  29. [58]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Computational materials science6, 15 (1996)

    1996

  30. [59]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Physical review B54, 11169 (1996)

    1996

  31. [60]

    K. S. Burch, D. Mandrus, and J.-G. Park, Nature563, 47 (2018)

    2018

  32. [61]

    Hou and R

    Y. Hou and R. Wu, Advanced Functional Materials35, e09453 (2025)

    2025

  33. [62]

    M.-P. Miao, N. Liu, W.-H. Zhang, J.-W. Zhou, D.-B. Wang, C. Wang, W. Ji, and Y.-S. Fu, Proceedings of the National Academy of Sciences122, e2422868122 (2025)

    2025