pith. sign in

arxiv: 2403.10769 · v3 · submitted 2024-03-16 · ❄️ cond-mat.mtrl-sci · cond-mat.mes-hall

Smooth Overlap of Spin Orientations: Machine Learning Exchange Fields for Ab-initio Spin Dynamics

Yuqiang Gao , Menno Bokdam , Paul J. Kelly This is my paper

Pith reviewed 2026-05-24 03:45 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci cond-mat.mes-hall
keywords machine learning potentialspin dynamicsnoncollinear magnetismexchange fieldsbcc FeGaussian approximation potentialab initio methods
0
0 comments X

The pith

A two-body machine learning model reproduces noncollinear magnetic energies in bcc Fe within 1 meV per spin.

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

This work extends the Gaussian Approximation Potential to include noncollinear spin moments, creating a model for the magnetic potential energy surface. The model relies on spin coordinates and orientations under an adiabatic approximation to achieve efficiency for spin dynamics. A two-body implementation, using rotational symmetries of magnetic interactions, is tested on bcc Fe and matches constrained DFT results closely. Further multi-body terms are proposed to improve generality.

Core claim

The paper establishes that the Smooth Overlap of Spin Orientations model, implemented in two-body form for exchange, predicts total energies and local fields for arbitrary noncollinear spin arrangements in bcc iron with errors under 1 meV per spin relative to explicit constrained noncollinear density functional theory calculations.

What carries the argument

The Smooth Overlap of Spin Orientations kernel, which encodes spin orientations to capture exchange interactions while preserving rotational symmetries.

If this is right

  • The model supports efficient ab initio spin dynamics by avoiding repeated electronic structure calculations.
  • Incorporating three-body terms will allow description of more complex magnetic interactions.
  • The approach extends to diverse materials beyond bcc Fe.
  • Machine learning can now handle spin dynamics at scales previously limited by DFT cost.

Where Pith is reading between the lines

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

  • Such models could enable simulations of magnetic textures or dynamics in nanostructures where direct DFT is prohibitive.
  • Integration with structural ML potentials might allow coupled magneto-structural simulations.
  • Validation on materials with stronger anisotropy or frustration would test broader applicability.

Load-bearing premise

Spin directions and magnitudes adjust adiabatically and the energy depends only on the spin configuration without needing full electronic details at each step.

What would settle it

Performing constrained DFT calculations for a noncollinear spin structure in bcc Fe and finding energy differences exceeding 1 meV per spin from the model's predictions would disprove the accuracy.

Figures

Figures reproduced from arXiv: 2403.10769 by Menno Bokdam, Paul J. Kelly, Yuqiang Gao.

Figure 1
Figure 1. Figure 1: FIG. 1. (a) Total energy per atom as a function of the lat [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Two spins [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. (a) Total energy and (b) transverse exchange field [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Comparison of the total energies predicted by [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. (Top row) Comparison of total DFT energies calculated for 25 noncollinear test spin configurations described in [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Local magnetic moment distributions in the training [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Comparison of total DFT energies calculated for [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Transverse exchange field [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
read the original abstract

We add the magnetic degrees of freedom to the widely used Gaussian Approximation Potential of machine learning (ML) and present a model that describes the potential energy surface of a crystal based on the atomic coordinates as well as their noncollinear magnetic moments. Assuming an adiabatic approximation for the spin directions and magnitudes, the ML model depends solely on spin coordinates and orientation, resulting in computational efffciency and enabling ab initio spin dynamics. Leveraging rotational symmetries of magnetic interactions, the ML model can incorporate various magnetic interactions, expanding into two-body, three-body terms, etc., following the spirit of cluster expansion. For simplicity, we implement the ML model with a two-body form for the exchange interaction. Comparing total energies and local fields predicted by the model for noncollinear spin arrangements with explicit results of constrained noncollinear density functional calculations for bcc Fe yields excellent results, within 1 meV/spin for the total energy. Further optimization, including three-body and other terms, is expected to encompass diverse magnetic interactions and enhance the model's accuracy. This will extend the model's applicability to a wide range of materials and facilitate the machine learning ab initio spin dynamics.

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

1 major / 1 minor

Summary. The manuscript extends the Gaussian Approximation Potential framework to noncollinear magnetic moments via a Smooth Overlap of Spin Orientations (SOSO) model. Assuming an adiabatic approximation, the model depends only on spin coordinates and orientations; a two-body truncation of the symmetry-adapted expansion is implemented and compared to constrained noncollinear DFT on bcc Fe, with reported agreement within 1 meV/spin for total energies and local fields.

Significance. A transferable ML representation of exchange fields could enable efficient spin dynamics at near-DFT accuracy. The symmetry-adapted cluster-expansion structure is a clear strength, and the explicit comparison to independent constrained-DFT data reduces circularity relative to purely fitted models.

major comments (1)
  1. [Abstract] Abstract: the central claim of 'excellent results, within 1 meV/spin for the total energy' is presented without any information on training-set size, validation protocol, error bars, or confirmation that the tested noncollinear configurations were excluded from the fit. Because the two-body model is constructed to reproduce the interactions it is trained on, this detail is load-bearing for whether the reported accuracy demonstrates generalization to unseen spin arrangements needed for spin dynamics.
minor comments (1)
  1. [Abstract] Abstract: 'efffciency' is a typographical error.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their positive summary and for identifying a clear way to strengthen the abstract. We agree that additional context on training, validation, and generalization is warranted and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim of 'excellent results, within 1 meV/spin for the total energy' is presented without any information on training-set size, validation protocol, error bars, or confirmation that the tested noncollinear configurations were excluded from the fit. Because the two-body model is constructed to reproduce the interactions it is trained on, this detail is load-bearing for whether the reported accuracy demonstrates generalization to unseen spin arrangements needed for spin dynamics.

    Authors: We agree that the abstract should be revised to include these details. The full manuscript describes a training set of spin configurations obtained from constrained noncollinear DFT and reports performance on additional noncollinear arrangements; we will extract the relevant numbers, protocol description, and explicit statement that the quoted 1 meV/spin figure includes held-out configurations, together with any available error bars or cross-validation statistics, and insert them into a revised abstract. This change will make the generalization claim explicit without altering the technical content of the work. revision: yes

Circularity Check

0 steps flagged

No significant circularity; validation rests on independent constrained DFT benchmarks

full rationale

The paper constructs an ML extension of the Gaussian Approximation Potential that incorporates spin orientations via a symmetry-adapted two-body expansion for exchange. Its central accuracy claim is supported by explicit numerical comparison of model outputs against separate constrained noncollinear DFT calculations on bcc Fe, reported as agreement within 1 meV/spin. This benchmark is external to the fitting procedure and does not reduce by construction to the training inputs or to any self-citation. No equations or statements in the provided text exhibit a fitted parameter being relabeled as a prediction, a self-definitional loop, or a uniqueness result imported from the authors' prior work. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the adiabatic spin approximation and the adequacy of a two-body truncation; no explicit free parameters or invented entities are named in the abstract.

free parameters (1)
  • ML model hyperparameters and cutoff radii
    Standard in GAP-style models; values are fitted to DFT data but not reported in abstract.
axioms (1)
  • domain assumption Adiabatic approximation for spin directions and magnitudes
    Explicitly stated in abstract as the basis for depending only on spin coordinates.

pith-pipeline@v0.9.0 · 5745 in / 1260 out tokens · 31822 ms · 2026-05-24T03:45:54.965099+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

105 extracted references · 105 canonical work pages

  1. [1]

    adiabatic spin approximation

    when the spin orientations vary. For this, we choose the collinear spin configuration with the same spatial co- ordinates {ri} ∆E({mi, ri}) = E({mi, ri}) − EFM({ri}), (36) where E and EFM are total energies for noncollinear and collinear ferromagnetic systems, respectively and the GAP-SOAP ML-FF framework can describe EFM({ri}) very well [65]. To test the...

    work page 2000

  2. [2]

    Variable mi To see how well the SOSO model works when the size of atomic moments depends on the local environment, we modified the Heisenberg Hamiltonian so that the moment of spin i depends linearly on the angles it makes with its nearest neighbours (nn) as mi = 2.23 − nnX j cos−1(ei · ej) nπ × 0.63, (42) in which n is the number of nearest neighbours of...

    work page

  3. [3]

    Variable atomic coordinates ri The noncollinear spin systems considered so far are for atoms (and spins) on fixed lattice sites. To evaluate the model’s performance for systems with thermally dis- ordered spin orientations and positions, we introduce a three-dimensional Gaussian distribution P (∆r) = 1 (2π)3/2σ3r exp − ∆2 r 2σ2r (44) of displacements ∆ r ...

    work page 2000

  4. [4]

    first- principles

    core hours of computation for a 2 ×2×2 (respec- tively 3×3×3) supercell. To constrain the spin direction, λ is set to be 30 with the energy penalty converging to less than 0.5 × 10−4eV/atom. With these computational parameters, the atomic moments in the constrained non- collinear calculations are converged to better than ∼1%. The DFT energies of the nonco...

    work page

  5. [5]

    However, where Suzuki et al

    to the GAP-SOAP framework is similar in spirit to the present work. However, where Suzuki et al. con- struct effective magnetization densities by replacing the delta function of point multipoles with the product of a spatial Gaussian times the multipole moments, we re- place the delta-function orientation of the magnetic mo- ment with a Gaussian which all...

    work page

  6. [6]

    two-body

    was very recently applied to the description of iron including noncollinear magnetism [27]. Insofar as Rinaldi et al. included the magnitude of the atomic moments, their implementation is more general than what we have presented. A very extensive training set containing more than 70000 structures was used including different crystal structures of Fe with ...

    work page

  7. [7]

    Car and M

    R. Car and M. Parrinello, Unified approach for molecular dynamics and density-functional theory, Phys. Rev. Lett. 55, 2471 (1985)

    work page 1985

  8. [8]

    Behler and M

    J. Behler and M. Parrinello, Generalized neural-network representation of high-dimensional potential-energy sur- faces, Phys. Rev. Lett. 98, 146401 (2007)

    work page 2007

  9. [9]

    Hohenberg and W

    P. Hohenberg and W. Kohn, Inhomogeneous electron gas, Phys. Rev. 136, B864 (1964)

    work page 1964

  10. [10]

    Kohn and L

    W. Kohn and L. J. Sham, Self-consistent equations in- cluding exchange and correlation effects, Phys. Rev. 140, A1133 (1965)

    work page 1965

  11. [11]

    Kresse and D

    G. Kresse and D. Joubert, From ultrasoft pseudopoten- tials to the projector augmented-wave method, Phys. Rev. B 59, 1758 (1999)

    work page 1999

  12. [12]

    Torrent, F

    M. Torrent, F. Jollet, F. Bottin, G. Z´ erah, and X. Gonze, Implementation of the projector augmented- wave method in the ABINIT code: Application to the study of iron under pressure, Comp. Mat. Sci. 42, 337 (2008)

    work page 2008

  13. [13]

    Behler, Perspective: Machine learning potentials 21 for atomistic simulations, J

    J. Behler, Perspective: Machine learning potentials 21 for atomistic simulations, J. Chem. Phys. 145, 170901 (2016)

    work page 2016

  14. [14]

    A. P. Bart´ ok, M. C. Payne, R. Kondor, and G. Cs´ anyi, Gaussian Approximation Potentials: The Accuracy of Quantum Mechanics, without the Electrons, Phys. Rev. Lett. 104, 136403 (2010)

    work page 2010

  15. [15]

    Jinnouchi, F

    R. Jinnouchi, F. Karsai, and G. Kresse, On-the-fly ma- chine learning force field generation: Application to melt- ing points, Phys. Rev. B 100, 014105 (2019)

    work page 2019

  16. [16]

    A. P. Bart´ ok, R. Kondor, and G. Cs´ anyi, On representing chemical environments, Phys. Rev. B 87, 184115 (2013)

    work page 2013

  17. [17]

    Jinnouchi, J

    R. Jinnouchi, J. Lahnsteiner, F. Karsai, G. Kresse, and M. Bokdam, Phase Transitions of Hybrid Perovskites Simulated by Machine-Learning Force Fields Trained on the Fly with Bayesian Inference, Phys. Rev. Lett. 122, 225701 (2019)

    work page 2019

  18. [18]

    Skubic, J

    B. Skubic, J. Hellsvik, L. Nordstr¨ om, and O. Eriksson, A method for atomistic spin dynamics simulations: imple- mentation and examples, J. Phys.: Condens. Matter 20, 315203 (2008)

    work page 2008

  19. [19]

    Eriksson, A

    O. Eriksson, A. Bergman, L. Bergqvist, and J. Hellsvik, Atomistic Spin Dynamics: Foundations and Applications (Oxford University Press, Oxford, 2017)

    work page 2017

  20. [20]

    Miltat, G

    J. Miltat, G. Albuquerque, and A. Thiaville, An Intro- duction to Micromagnetics in the Dynamic Regime, in Spin Dynamics in Confined Magnetic Structures I , Top- ics in Applied Physics, Vol. 83, edited by B. Hillebrands and K. Ounadjela (Springer, Berlin, Heidelberg, 2002) Chap. 1, pp. 1–34

    work page 2002

  21. [21]

    Tserkovnyak, A

    Y. Tserkovnyak, A. Brataas, G. E. W. Bauer, and B. I. Halperin, Nonlocal magnetization dynamics in ferromag- netic nanostructures, Rev. Mod. Phys. 77, 1375 (2005)

    work page 2005

  22. [22]

    A. P. Bart´ ok and G. Cs´ anyi, Gaussian approximation potentials: A brief tutorial introduction, International Journal of Quantum Chemistry 115, 1051 (2015)

    work page 2015

  23. [23]

    Behler, Four generations of high-dimensional neural network potentials, Chem

    J. Behler, Four generations of high-dimensional neural network potentials, Chem. Rev. 121, 10037 (2021)

    work page 2021

  24. [24]

    V. L. Deringer, A. P. Bart´ ok, N. Bernstein, D. M. Wilkins, M. Ceriotti, and G. Cs´ anyi, Gaussian process regression for materials and molecules, Chem. Rev. 121, 10073 (2021)

    work page 2021

  25. [25]

    A. P. Thompson, L. P. Swiler, C. R. Trott, S. Foiles, and G. J. Tucker, Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials, Journal of Computational Physics 285, 316 (2015)

    work page 2015

  26. [26]

    A. V. Shapeev, Moment tensor potentials: A class of sys- tematically improvable interatomic potentials, Multiscale Modeling & Simulation 14, 1153 (2016)

    work page 2016

  27. [27]

    Drautz, Atomic cluster expansion for accurate and transferable interatomic potentials, Phys

    R. Drautz, Atomic cluster expansion for accurate and transferable interatomic potentials, Phys. Rev. B 99, 014104 (2019)

    work page 2019

  28. [28]

    Drautz, Atomic cluster expansion of scalar, vectorial, and tensorial properties including magnetism and charge transfer, Phys

    R. Drautz, Atomic cluster expansion of scalar, vectorial, and tensorial properties including magnetism and charge transfer, Phys. Rev. B 102, 024104 (2020)

    work page 2020

  29. [29]

    Marco and B

    E. Marco and B. J¨ org, High-dimensional neural network potentials for magnetic systems using spin-dependent atom-centered symmetry functions, npj Computational Materials 7, 170 (2021)

    work page 2021

  30. [30]

    Domina, M

    M. Domina, M. Cobelli, and S. Sanvito, Spectral neigh- bor representation for vector fields: Machine learning po- tentials including spin, Phys. Rev. B 105, 214439 (2022)

    work page 2022

  31. [31]

    Novikov, B

    I. Novikov, B. Grabowski, F. K¨ ormann, and A. Shapeev, Magnetic Moment Tensor Potentials for collinear spin- polarized materials reproduce different magnetic states of bcc Fe, npj Computational Materials 8, 13 (2022); A. S. Kotykhov, K. Gubaev, V. Sotskov, C. Tantardini, M. Hodapp, A. V. Shapeeva, and I. S. Novikov, Fitting to magnetic forces improves th...

    work page 2022

  32. [32]

    Suzuki, T

    M.-T. Suzuki, T. Nomoto, E. V. Morooka, Y. Yanagi, and H. Kusunose, High-performance descriptor for magnetic materials: Accurate discrimination of magnetic struc- ture, Phys. Rev. B 108, 014403 (2023)

    work page 2023

  33. [33]

    Rinaldi, M

    M. Rinaldi, M. Mrovec, A. Bochkarev, Y. Lysogorskiy, and R. Drautz, Non-collinear magnetic atomic cluster expansion for iron, npj Computational Materials 10, 12 (2024)

    work page 2024

  34. [34]

    Kotochigova, Z

    S. Kotochigova, Z. H. Levine, E. L. Shirley, M. D. Stiles, and C. W. Clark, Local-density-functional cal- culations of the energy of atoms, Phys. Rev. A 55, 191 (1997), http://math.nist.gov/DFTdata/; Local- density-functional calculations of the energy of atoms, Phys. Rev. A 56, 5191 (1997)

    work page 1997

  35. [35]

    Gunnarsson, Band model for magnetism of transition metals in the spin-density-functional formalism, J

    O. Gunnarsson, Band model for magnetism of transition metals in the spin-density-functional formalism, J. Phys. F: Met. Phys. 6, 587 (1976)

    work page 1976

  36. [36]

    U. K. Poulsen, J. Kollar, and O. K. Andersen, Magnetic and cohesive properties from canonical bands, J. Phys. F: Met. Phys. 6, L241 (1976)

    work page 1976

  37. [37]

    J. F. Janak, Uniform susceptibilities of metallic elements, Phys. Rev. B 16, 255 (1977)

    work page 1977

  38. [38]

    (), Because the LDA is known to underestimate the equi- librium lattice constant of the ferromagnetic 3 d metals by about 3%, we used the Generalized Gradient Approx- imation (GGA) of Perdew, Burke and Enzerhof [95] to describe the effects of exchange and correlation. The cal- culated results were obtained with a plane-wave basis and the projector augment...

    work page

  39. [39]

    K¨ ubler,Theory of Itinerant Electron Magnetism (Ox- ford University Press, Oxford, 2000)

    J. K¨ ubler,Theory of Itinerant Electron Magnetism (Ox- ford University Press, Oxford, 2000)

    work page 2000

  40. [40]

    Mohn, Magnetism in the Solid State: An Introduction , edited by M

    P. Mohn, Magnetism in the Solid State: An Introduction , edited by M. Cardona, P. Fulde, K. von Klitzing, and H.- J. Queisser, Springer Series in Solid-State Sciences, Vol. 134 (Springer, Berlin, 2003)

    work page 2003

  41. [41]

    S. V. Halilov, H. Eschrig, A. Y. Perlov, and P. M. Oppeneer, Adiabatic spin dynamics from spin-density- functional theory: Application to fe, co, and ni, Phys. Rev. B 58, 293 (1998)

    work page 1998

  42. [42]

    Dal Corso and S

    A. Dal Corso and S. de Gironcoli, Ab initio phonon dis- persions of fe and ni, Phys. Rev. B 62, 273 (2000)

    work page 2000

  43. [43]

    Wakabayashi, R

    N. Wakabayashi, R. H. Scherm, and H. G. Smith, Lattice dynamic of Ti, Co, Tc, and other hcp transition metals, Phys. Rev. B 25, 5122 (1982)

    work page 1982

  44. [44]

    P. Escudier, L’anisotropie de l’aimantation: un param` etre important de l’´ etude de l’anisotropie magn´ etocristalline (magnetization anisotropy - im- portant parameter for study of magnetocrystalline anisotropy) (1975) pp. 125–173

    work page 1975

  45. [45]

    D. M. Paige, B. Szpunar, and B. K. Tanner, The magne- tocrystalline anisotropy of cobalt, J. Magn. Magn. Mater. 44, 239 (1984). 22

    work page 1984

  46. [46]

    G. Chen, J. Zhu, A. Quesada, J. Li, A. T. N’Diaye, Y. Huo, T. P. Ma, Y. Chen, H. Y. Kwon, C. Won, Z. Q. Qiu, A. K. Schmid, and Y. Z. Wu, Novel chiral magnetic domain wall structure in Fe/Ni/Cu (001) films, Phys. Rev. Lett. 110, 177204 (2013); H. Yang, G. Chen, A. A. C. Cotta, A. T. N’Diaye, S. A. Nikolaev, E. A. Soares, W. A. A. Macedo, K. Liu, A. K. Schm...

    work page 2013

  47. [47]

    Z. S. Basinski, W. Hume-Rothery, and A. L. Sutton, The lattice expansion of iron, Proc. R. Soc. London, Ser. A 229, 459 (1955)

    work page 1955

  48. [48]

    Vincent and M

    F. Vincent and M. Figlarz, Quelques pr´ ecisions sur les param` etres cristallins et l’intensit´ e des raies Debye- Scherrer du cobalt cubique et du cobalt hexago- nal, Comptes Rendus Hebdomadaires des S´ eances de l’Acad´ emie des Sciences Serie C264, 1270 (1967)

    work page 1967

  49. [49]

    Taylor, Lattice parameters of binary nickel-cobalt al- loys, Journal of the Institute of Metals 77, 585 (1950)

    A. Taylor, Lattice parameters of binary nickel-cobalt al- loys, Journal of the Institute of Metals 77, 585 (1950)

    work page 1950

  50. [50]

    F. L. Durhuus, T. Skovhus, and T. Olsen, Plane wave implementation of the magnetic force theorem for mag- netic exchange constants: application to bulk Fe, Co and Ni, J. Phys.: Condens. Matter 35, 105802 (2023)

    work page 2023

  51. [51]

    R. E. Camley and K. L. Livesey, Consequences of the dzyaloshinskii-moriya interaction, Surface Science Re- ports 78, 100607 (2023)

    work page 2023

  52. [52]

    Heine, O

    M. Heine, O. Hellman, and D. Broido, Temperature- dependent renormalization of magnetic interactions by thermal, magnetic, and lattice disorder from first princi- ples, Phys. Rev. B 103, 184409 (2021)

    work page 2021

  53. [53]

    Pajda, J

    M. Pajda, J. Kudrnovsk` y, I. Turek, V. Drchal, and P. Bruno, Ab initio calculations of exchange interactions, spin-wave stiffness constants, and Curie temperatures of Fe, Co, and Ni, Phys. Rev. B 64, 174402 (2001)

    work page 2001

  54. [54]

    Born and R

    M. Born and R. J. Oppenheimer, Zur Quantentheorie der Molekeln (On the Quantum Theory of Molecules), Annalen der Physik 84, 457 (1927)

    work page 1927

  55. [55]

    V. P. Antropov, M. I. Katsnelson, M. van Schilfgaarde, and B. N. Harmon, Ab Initio spin dynamics in magnets, Phys. Rev. Lett. 75, 729 (1995); V. P. Antropov, M. I. Katsnelson, B. N. Harmon, M. van Schilfgaarde, and D. Kusnezov, Spin dynamics in magnets: Equation of motion and finite temperature effects, Phys. Rev. B 54, 1019 (1996)

    work page 1995

  56. [56]

    Kikuchi, A theory of cooperative phenomena, Phys

    R. Kikuchi, A theory of cooperative phenomena, Phys. Rev. 81, 988 (1951)

    work page 1951

  57. [57]

    Behler, Atom-centered symmetry functions for con- structing high-dimensional neural network potentials, J

    J. Behler, Atom-centered symmetry functions for con- structing high-dimensional neural network potentials, J. Chem. Phys. 134, 074106 (2011)

    work page 2011

  58. [58]

    Maresca, D

    F. Maresca, D. Dragoni, G. Cs´ anyi, N. Marzari, and W. A. Curtin, Screw dislocation structure and mobility in body centered cubic Fe predicted by a Gaussian Ap- proximation Potential, npj Computational Materials 4, 69 (2018)

    work page 2018

  59. [59]

    E. R. M. Davidson, T. Daff, G. Csanyi, and M. W. Finnis, Grand canonical approach to modeling hydrogen trap- ping at vacancies in α-Fe, Phys. Rev. Materials 4, 063804 (2020)

    work page 2020

  60. [60]

    Vandermause, S

    J. Vandermause, S. B. Torrisi, S. Batzner, Y. Xie, L. Sun, A. M. Kolpak, and B. Kozinsky, On-the-fly active learn- ing of interpretable bayesian force fields for atomistic rare events, npj Computational Materials 6, 20 (2020)

    work page 2020

  61. [61]

    Jansen, G

    T. Jansen, G. Brocks, and M. Bokdam, Phase transi- tions of LaMnO 3 and SrRuO3 from DFT+ U based ma- chine learning force fields simulations, Phys. Rev. B 108, 235122 (2023)

    work page 2023

  62. [62]

    Lahnsteiner and M

    J. Lahnsteiner and M. Bokdam, Anharmonic lattice dy- namics in large thermodynamic ensembles with machine- learning force fields: Cspbbr 3, a phonon liquid with cs rattlers, Physical Review B 105, 024302 (2022)

    work page 2022

  63. [63]

    S. N. Pozdnyakov, M. J. Willatt, A. P. Bart´ ok, C. Ortner, G. Cs´ anyi, and M. Ceriotti, Incompleteness of Atomic Structure Representations, Phys. Rev. Lett. 125, 166001 (2020)

    work page 2020

  64. [64]

    spin” and “magnetic moment

    (), Equation (6) is conventionally referred to as a spin Hamiltonian with mi ≡ si a spin operator whose magni- tude is fixed. In the micromagnetics literature, it is com- mon to use the magnetisation M(r) and assuming that its magnitude is not variable, to characterise its orienta- tion using the unit vector bm; to avoid confusion, we will designate a uni...

    work page

  65. [65]

    R. M. White, Quantum theory of magnetism , 2nd ed., edited by M. Carona, P. Fulde, K. von Klitzing, and H.-J. Queisser, Springer Series in Solid-State Sciences, Vol. 32 (Springer, Berlin, 1983)

    work page 1983

  66. [66]

    Kaufmann and W

    K. Kaufmann and W. Baumeister, Single-centre expan- sion of Gaussian basis functions and the angular decom- position of their overlap integrals, J. Phys. B: Atomic, Molecular and Optical Physics 22, 1 (1989)

    work page 1989

  67. [67]

    The Gaussian basis functions need to be orthogonalised

    (), In the quip code from Cambridge University, the ra- dial parts are expanded in a set of equally spaced Gaus- sian functions while in the vasp ML package the radial basis functions are normalised spherical Bessel functions. The Gaussian basis functions need to be orthogonalised. The normalised spherical Bessel functions are mutually orthogonal

    work page

  68. [68]

    Abramowitz and I

    M. Abramowitz and I. A. Stegun, Handbook of Mathe- matical Functions (Dover, New York, 1965)

    work page 1965

  69. [69]

    E. P. Wigner, Group Theory and its application to the quantum mechanics of atomic spectra (Academic Press, 1959)

    work page 1959

  70. [70]

    (), Because we approach ML-EF from the DFT point of view, we switch from the conventional potential energy notation U of classical molecular dynamics to the total (electronic) energy E of DFT

    work page

  71. [71]

    Jana and M

    R. Jana and M. A. Caro, Searching for iron nanoparti- cles with a general-purpose gaussian approximation po- tential, Phys. Rev. B 107, 245421 (2023)

    work page 2023

  72. [72]

    (), Because the coefficients eC i nl in (35) used to describe the similarity kernel (34) are expressed in terms of the spherical harmonics Ylm(ei), the derivative of Ylm(ei) with respective to ei in (38) can be performed analyt- ically

    work page

  73. [73]

    Wang, P.-W

    H. Wang, P.-W. Ma, and C. H. Woo, Exchange interac- tion function for spin-lattice coupling in bcc iron, Phys. Rev. B 82, 144304 (2010)

    work page 2010

  74. [74]

    J. M. Ziman, Principles of the Theory of Solids , 2nd ed. (Cambridge University Press, Cambridge, 1972)

    work page 1972

  75. [75]

    N. W. Ashcroft and N. D. Mermin, Solid State 23 Physics (Holt-Saunders International Editions, Philadel- phia, 1976)

    work page 1976

  76. [76]

    but can extend prac- tically to about T > ΘD/3 in Pt, Cu and Au and to about T > ΘD/5 in Al” [99]

    (), Though the onset of the equipartition regime is strictly speaking at the Debye temperature ΘD, “in prac- tice the high- T limit is identified as the equipartition regime with linear-in-T resistivity ... but can extend prac- tically to about T > ΘD/3 in Pt, Cu and Au and to about T > ΘD/5 in Al” [99]. For typical transition-metal Debye temperatures, th...

    work page

  77. [77]

    Ma and S

    P.-W. Ma and S. L. Dudarev, Constrained density func- tional for noncollinear magnetism, Phys. Rev. B 91, 054420 (2015)

    work page 2015

  78. [78]

    (), We use the vasp option I CONSTRAINED M=1 to calculate ∂∆E/∂ ei

    work page

  79. [79]

    Y. Liu, A. A. Starikov, Z. Yuan, and P. J. Kelly, First- principles calculations of magnetization relaxation in pure Fe, Co, and Ni with frozen thermal lattice disor- der, Phys. Rev. B 84, 014412 (2011)

    work page 2011

  80. [80]

    Gibertini, M

    M. Gibertini, M. Koperski, A. F. Morpurgo, and K. S. Novoselov, Magnetic 2D materials and heterostructures, Nat. Nanotechnol. 14, 408 (2019)

    work page 2019

Showing first 80 references.