Pith. sign in

REVIEW 3 major objections 5 minor 37 references

Actively-trained magnetic Moment Tensor Potentials for mechanical, dynamical, and thermal properties of paramagnetic CrN

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read An automatically trained magnetic Moment Tensor Potential reproduces elastic, dynamical, and thermal properties of paramagnetic B1-CrN against DFT and experiment.

desk verdict A solid methods paper that genuinely automates training of magnetic moment tensor potentials with cDFT and active learning, with honest validation on CrN; the main caveat is the unquantified infinite-temperature assumption behind the paramagnetic ensemble. read the letter →

arxiv 2412.20214 v2 pith:NOBGLDZG submitted 2024-12-28 cond-mat.mtrl-sci physics.atom-ph

Alexey S. Kotykhov , Max Hodapp , Christian Tantardini , Konstantin Kravtsov , Ivan Kruglov , Alexander V. Shapeev , Ivan S. Novikov This is my paper · ORCID
classification cond-mat.mtrl-sciphysics.atom-ph
keywords magneticmomenttensorpotentialactivelearningconstraineddensityfunctionaltheoryparamagneticCrNelasticconstantsphononspectralatticethermalexpansionspecificheatcapacity
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper establishes a fully automated protocol for fitting a machine-learned interatomic potential that includes collinear magnetic moments as explicit degrees of freedom. The key ingredients are an active-learning algorithm that selects the most extrapolative atomic configurations on the fly and constrained density functional theory, which supplies energies, forces, stresses, and magnetic forces for those configurations even when the magnetic moments are far from equilibrium. Tested on B1-CrN, the resulting potential reproduces elastic constants, phonon spectra, lattice thermal expansion, and specific heat capacity in the paramagnetic state, matching DFT and experimental values. The authors argue that this removes the manual trial-and-error that has previously dominated the construction of magnetic machine-learning potential training sets.

What carries the argument

The central object is the magnetic Moment Tensor Potential (mMTP), whose energy is a sum of per-atom contributions built from moment tensor descriptors that depend on interatomic distances, atomic types, and collinear magnetic moments, and whose total energy is explicitly invariant under inversion of all magnetic moments. Fitting uses energies, forces, stresses, and magnetic forces, with active learning based on the D-optimality criterion (maxvol) to select configurations that most increase the linear independence of the fitting matrix. Constrained DFT provides the crucial training labels for non-equilibrium magnetic states by fixing each atom's magnetic moment as an external parameter during self-consistency, which lets the potential learn the energy landscape of excited magnetic configurations.

What would settle it

Compute the elastic constants, phonon spectra, and thermal expansion of B1-CrN using a paramagnetic ensemble that includes finite-temperature short-range spin correlations (for example, from disordered-local-moment molecular dynamics or Monte Carlo spin flips at realistic temperatures) and compare against the paper's equal-weight 50-configuration ensemble; differences exceeding the stated fitting errors would falsify the central ensemble premise.

Watch

Extended reading notes

Core claim

The paper's central claim is that an actively trained magnetic Moment Tensor Potential can describe B1-CrN in its paramagnetic state without hand-crafted training sets. The paramagnetic state is represented as an equally weighted average over 50 randomly disordered collinear magnetic configurations, and the fitted potential reproduces the mechanical, dynamical, and thermal properties of this ensemble: elastic constants agree with DFT in the ferromagnetic, antiferromagnetic, and paramagnetic states; phonon spectra agree with DFT and with Raman/infrared measurements once non-analytic LO-TO corrections are included; and quasi-harmonic thermal expansion and heat capacity match experiment. The final training set contains 2423 configurations, roughly 96% of them with non-equilibrium magnetic moments computed with constrained DFT, and prediction errors are 1.7 meV/atom in energy, 108 meV/Å in forces, 0.4 GPa in stresses, and 64 meV/μB in magnetic forces.

Load-bearing premise

The paramagnetic state is modeled as an equally weighted average over 50 random collinear magnetic configurations, which is the infinite-temperature limit; if 50 configurations are not representative, or if the actual finite-temperature paramagnetic state contains spin correlations this ensemble cannot capture, the predicted properties could be systematically off.

Editorial extensions

If this is right

  • The same protocol can build magnetic machine-learning potentials for other materials without manually curated configurations, replacing trial-and-error training-set construction.
  • One fitted potential simultaneously describes ferromagnetic, antiferromagnetic, and paramagnetic states of B1-CrN, so it can be used for simulations crossing magnetic orderings.
  • Phonon spectra in the paramagnetic state require non-analytic term corrections to capture LO-TO splitting, and using experimental dielectric and Born charge inputs brings the mMTP spectrum close to Raman and infrared data.
  • Because anharmonic effects are negligible in CrN, the quasi-harmonic approximation suffices for thermal properties; heat capacity matches experiment above 400 K, while the 280–400 K discrepancy is attributed to the absence of Ortho-CrN data in the training set.
  • Active learning with constrained DFT is essential in practice: about 96% of the selected training configurations have non-equilibrium magnetic moments that ordinary DFT cannot provide.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Extension: the equally weighted 50-configuration ensemble represents the infinite-temperature paramagnetic limit; predictions near the Néel temperature could shift if real finite-temperature paramagnetism carries short-range spin correlations that this ensemble misses.
  • Extension: the protocol's reliance on constrained DFT means its transferability to other correlated magnets will depend on the choice of constraint spheres and Hubbard U, which would need separate validation for each new material.
  • Extension: the heat-capacity mismatch near the phase transition suggests that a B1-only training set cannot capture structural phase changes; adding Ortho-CrN data in the same active-learning loop would be a direct test of whether the protocol extends to phase boundaries.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes an automated active-learning protocol for fitting magnetic Moment Tensor Potentials (mMTPs) in which training configurations are selected during magnetic-moment equilibration, relaxation, and molecular dynamics, and are subsequently labeled by constrained DFT (cDFT) energies, forces, stresses, and magnetic forces. The method is applied to B1-CrN, and the paramagnetic state is represented by an equal-weight average over 50 randomly disordered collinear magnetic configurations. The authors report elastic constants, phonon spectra, thermal expansion, and specific heat capacity for B1-CrN, and claim that the automatically trained mMTP reproduces mechanical, dynamical, and thermal properties of the paramagnetic state with respect to DFT and experiments. The training set is small (2423 configurations) and is made openly available.

Significance. If the central claim holds, the paper is a useful methodological contribution: it replaces manual training-set construction for magnetic MLIPs with an active-learning loop that exploits cDFT to label non-equilibrium magnetic configurations, and it demonstrates the approach on a material with a complex magnetic order-disorder transition. The open dataset and the explicit fitting to magnetic forces are strengths, and the comparison against independent experimental thermal data gives the paper a falsifiable component. The main significance hinges, however, on whether the equal-weight random-spin ensemble used for the paramagnetic state is physically representative at the temperatures of interest, and on whether the reported agreement with DFT is as quantitative as the text suggests.

major comments (3)
  1. [Section III.B] The paramagnetic state is defined as an equal-weight average over 50 random collinear spin configurations, which the authors state corresponds to the infinite-temperature limit. The convergence test in Fig. 5 shows that adding more configurations changes the averaged energy and force by less than the mMTP fitting errors, but this only controls statistical sampling error; it does not address whether the infinite-temperature measure is representative of the short-range-correlated paramagnetic state just above the Néel temperature (280 K). Since every paramagnetic prediction in Table II and Figs. 8-10 is generated from this ensemble, a systematic bias in the ensemble would enter all of the central results. I ask the authors to provide a concrete test of ensemble representativeness, for example a comparison with Boltzmann-weighted spin-space averaging as used in Ref. [18], or a sensitivity study over ensembles with different degrees of short-range order, and to state clearly how the reported properties change under that test.
  2. [Table II] The text states that the mMTP and DFT elastic constants "coincide very well," but the table shows substantial disagreements for shear constants. For the AFM state, C44 is 158 GPa from mMTP versus 122 GPa (DFT[*]) and 124 GPa (DFT[18]), a deviation of about 28%. For the FM state, C44 is 188 GPa versus 164 and 162 GPa, a deviation of about 15%. Even the paramagnetic C44 deviates by about 10% (160 vs 145 GPa). The claim of very good agreement should be either quantified with error metrics or tempered, and the authors should discuss what these deviations imply for the intended downstream use of the potential in mechanical and dynamical simulations.
  3. [Section III.E, Fig. 10] The claim that the fitted mMTP reproduces thermal properties is supported by specific heat capacity agreement only above about 400 K. The discrepancy between 280 and 400 K is acknowledged and attributed to the absence of Ortho-CrN data, but it is not quantified, and the attribution is not tested. I request a numerical comparison (e.g., average absolute or relative error in the 280-400 K window and above 400 K), and either a demonstration that the discrepancy is dominated by the missing Ortho-CrN phase (for instance, a calculation of the Ortho-CrN/AFM contribution to the heat capacity) or an explicit statement that the thermal-property claim in the abstract and conclusion is limited to the B1 paramagnetic phase above 400 K.
minor comments (5)
  1. [Section II.A] The heading contains a typo: "Magnetic Moment T ensor Potential" should read "Magnetic Moment Tensor Potential."
  2. [Figure 9 caption] The caption reads "experimentally [18]"; it should read "experimental data from [18]" or "experimental values from [18]."
  3. [Section III.A] The sentence describing perturbations of magnetic moments is ambiguous: "we repeated the previous step for the magnetic moments by applying perturbations, again, two times, by at most 15% to each equilibrium magnetic moment." Please specify whether each of the two repetitions uses the same 15% bound and whether the perturbed moments are independent of the atomic displacements.
  4. [Section III.A] The term "non-equilibrium magnetic moments" is used frequently; a brief definition or a reference to the cDFT description would help readers understand that these are magnetic moments that are not the self-consistent ground-state moments for the given atomic configuration.
  5. [Table II] The labels DFT[*] and DFT[18] are clear enough, but the caption could state explicitly which magnetic ordering and which computational settings (U value, k-point mesh, cutoff) correspond to DFT[*] so that the comparison is reproducible without going back to Section II.D.

Circularity Check

1 steps flagged · score 3.0 of 10

Elastic-constant validation partially reduces to shear-displaced training data; thermal and vibrational tests remain externally anchored.

  1. fitted input called prediction [Section III.A (training-set construction) and Section III.C/Table II (elastic constants)]
    "Afterwards, we constructed 32 configurations by applying shear displacements to the equilibrated configurations of ±1% along the xx and yz directions for the ferromagnetic and five randomly disordered collinear magnetic states, and along the xx, xy, xz, zz directions for the configurations in the AFM state. These configurations were calculated with DFT and added to the training set."

    Elastic constants are the linear stress response to strain, i.e., second derivatives of energy with respect to strain. The training set explicitly includes DFT energies and stresses for ±1% shear-displaced configurations of FM, AFM, and randomly disordered collinear magnetic states, and the fitting functional (5) contains a stress term with weight ws. Since the PM state is defined as an average over the same kind of random collinear magnetic states, the PM elastic constants in Table II are substantially a re-reading of fitted input data rather than an extrapolative prediction. The agreement with DFT[18] therefore checks consistency between two DFT-based disorder averages, while the genuinely external tests are the experimental thermal expansion and heat capacity.

full rationale

The central derivation is not circular: the mMTP is a surrogate fitted to DFT/cDFT energies, forces, stresses, and magnetic forces, and the reported thermal properties are compared with experimental values that were not used as fitting labels. The PM ensemble is an equal-weight average over random collinear magnetic states by construction, and using it both for training and evaluation is a modeling assumption, not a logical tautology. The main circularity is localized to the elastic-constant validation: shear-displaced configurations with DFT stresses were deliberately added to the training set, so the Table II elastic constants are largely a consistency check on fitted data rather than an independent prediction. The LO-TO phonon comparison is also softened by the use of experimental dielectric tensor and Born effective charges from [16] in the non-analytic correction, but this is a calibration detail that does not affect the short-range force constants or the thermal properties. Overall, the protocol claim and the thermal predictions retain independent support, so the circularity is partial and property-specific.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The central claims rest on a chain of model and approximation choices: the DFT functional (LDA+U with U=3 eV), the cDFT constraint method, the mMTP functional form, the active learning criterion, the infinite-temperature spin averaging, and the QHA. None of these are introduced in this paper, but they are all load-bearing. No new physical entities are postulated. The main source of non-independence is the choice of U from the experimental lattice parameter, which anchors the thermal expansion prediction rather than leaving it fully free.

free parameters (6)
  • Hubbard U = 3 eV
    Adopted from refs 17,18 where it was chosen to match the experimental room-temperature lattice parameter and valence DOS of paramagnetic B1-CrN; affects all training data and therefore all predicted properties.
  • M_max for Cr and N = 3.2 µB (Cr), 0.1 µB (N)
    Define the domain of magnetic moments in the Chebyshev basis of the mMTP; chosen as maximum moments in the training set and affect the functional representation.
  • Fitting weights we, wf, ws, wt = 1, 0.01 Ų, 0.001, 0.1 µB²
    Weights in the objective function (5); chosen by hand without sensitivity analysis; influence the fit balance between energies, forces, stresses, and magnetic forces.
  • Active learning thresholds γlow and γup = γlow=2 (relaxation), 4 (MD); γup=10
    Control when configurations are preselected or terminate simulation; based on recommendations from refs 12 and 28, but the specific values affect the size and composition of the training set.
  • Cutoff and basis hyperparameters = Rcut=5.0 Å, Rmin=1.7 Å, level 16, Nφ=8, Nψ=2
    Model architecture choices for the mMTP; chosen empirically, affecting accuracy and computational cost.
  • Number of magnetic configurations for PM average = 50
    Selected from a convergence test in Fig. 5; the spread of averaged properties is not reported.
assumptions (7)
  • domain assumption LDA+U (U=3 eV) is an adequate electronic structure description for the magnetic states of CrN
    Invoked in Section II.D; all DFT/cDFT training data depends on this functional choice.
  • domain assumption The Gonze et al. cDFT method correctly imposes hard constraints on atomic magnetic moments and yields accurate energies, forces, stresses and magnetic forces for non-equilibrium spin states
    Used throughout Section II.C and II.D; the whole strategy of training on excited magnetic states rests on this.
  • domain assumption The mMTP functional form (Eqs. 1-4) can represent the potential energy of collinear magnetic systems and their magnetic forces
    Taken from prior work [7,19,21,24]; not re-derived here.
  • domain assumption The D-optimality (maxvol) active learning criterion identifies configurations whose addition most improves the mMTP
    Borrowed from non-magnetic MTP active learning [12,25].
  • domain assumption An equally weighted average over randomly disordered collinear spin configurations approximates the paramagnetic state
    Stated in Section III.B; corresponds to the infinite-temperature limit and underpins all paramagnetic-state predictions.
  • domain assumption The quasi-harmonic approximation is valid for the lattice thermal properties of B1-CrN
    Assumed in Section III.E based on an earlier estimate of small anharmonicity from ref 18.
  • domain assumption Experimental dielectric tensor (ε=22) and Born effective charge (Z=4.4) are appropriate for non-analytic LO-TO corrections
    Used in Section III.D when computing phonon spectra; these values are taken from experiment [16].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Actively-trained magnetic Moment Tensor Potentials for mechanical, dynamical, and thermal properties of paramagnetic CrN." pith.science (2026). https://pith.science/paper/NOBGLDZG

@misc{pith2026241220214,
  author       = {Pith},
  title        = {Pith review of: Actively-trained magnetic Moment Tensor Potentials for mechanical, dynamical, and thermal properties of paramagnetic CrN},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NOBGLDZG}},
  note         = {Machine review of arXiv:2412.20214}
}
read the original abstract

We present a protocol for automated fitting of magnetic Moment Tensor Potential explicitly including magnetic moments in its functional form. For the fitting of this potential we use energies, forces, stresses, and magnetic forces (negative derivatives of energies with respect to magnetic moments) of configurations selected with an active learning algorithm. These selected configurations are computed using constrained density functional theory, which enables calculating energies and their derivatives for both equilibrium and non-equilibrium (excited) magnetic states. We test our protocol on the system of B1-CrN and demonstrate that the automatically trained magnetic Moment Tensor Potential reproduces mechanical, dynamical, and thermal properties, of B1-CrN in the paramagnetic state with respect to density functional theory and experiments.

Figures

Figures reproduced from arXiv: 2412.20214 by the authors.

Figure 1
Figure 1. FIG. 1: Computation scheme of the per-atom energy contribution [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: Protocol of magnetic MTP active training. We start from an initial mMTP (the scheme of mMTP is also [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 4
Figure 4. FIG. 4: B1-CrN in the paramagnetic state is [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (6 more)
Figure 3
Figure 3. Figure 3: FIG. 3: B1-CrN in the FM and AFM states. Cr atoms [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]
Figure 5
Figure 5. Figure 5: FIG. 5: Dependence of the averaged forces and energies [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 8
Figure 8. Figure 8: FIG. 8: Phonon spectrum for B1-CrN in the [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 7
Figure 7. Figure 7: FIG. 7: Phonon spectrum for B1-CrN in the [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 9
Figure 9. Figure 9: FIG. 9: Linear thermal expansion coefficient for [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10: Specific heat capacity for B1-CrN in the [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 31 canonical work pages

  1. [18]

    L. Zhou, F. K¨ ormann, D. Holec, M. Bartosik, B. Grabowski, J. Neugebauer, and P. H. Mayrhofer, Structural stability and thermodynamics of crn magnetic phases from ab initio calculations and experiment, Phys- ical Review B 90, 184102 (2014)

    work page 2014

  2. [1]

    Heck, Magnetic materials and their applications (El- sevier, 2013)

    C. Heck, Magnetic materials and their applications (El- sevier, 2013)

    work page 2013

  3. [2]

    K. M. Krishnan, Fundamentals and applications of mag- netic materials (Oxford University Press, 2016)

    work page 2016

  4. [3]

    Crisan, P

    V. Crisan, P. Entel, H. Ebert, H. Akai, D. D. John- son, and J. Staunton, Magnetochemical origin for invar anomalies in iron-nickel alloys, Physical Review B 66, 014416 (2002)

    work page 2002

  5. [4]

    Smirnov, W

    A. Smirnov, W. Shelton, and D. Johnson, Importance of thermal disorder on the properties of alloys: Origin of paramagnetism and structural anomalies in bcc-based Fe1−xAlx, Physical Review B—Condensed Matter and Materials Physics 71, 064408 (2005)

    work page 2005

  6. [5]

    Ekholm, H

    M. Ekholm, H. Zapolsky, A. Ruban, I. Vernyhora, D. Ledue, and I. Abrikosov, Influence of the magnetic state on the chemical order-disorder transition temper- ature in fe-ni permalloy, Physical review letters 105, 167208 (2010)

    work page 2010

  7. [6]

    Eckhoff and J

    M. Eckhoff and J. Behler, High-dimensional neural network potentials for magnetic systems using spin- dependent atom-centered symmetry functions, npj Com- putational Materials 7, 170 (2021)

    work page 2021

  8. [7]

    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)

    work page 2022

Show all 37 references
  1. [8]

    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)

    2024

  2. [9]

    H. Yu, Y. Zhong, L. Hong, C. Xu, W. Ren, X. Gong, and H. Xiang, Spin-dependent graph neural network potential for magnetic materials, Physical Review B 109, 144426 (2024)

    2024

  3. [10]

    H. Yu, B. Liu, Y. Zhong, L. Hong, J. Ji, C. Xu, X. Gong, and H. Xiang, Physics-informed time-reversal equivariant neural network potential for magnetic materials, Physical Review B 110, 104427 (2024)

    2024

  4. [11]

    T. S. Kostiuchenko, A. V. Shapeev, and I. S. Novikov, Interatomic interaction models for magnetic materials: Recent advances, Chinese Physics Letters (2024)

    2024

  5. [12]

    E. V. Podryabinkin and A. V. Shapeev, Active learning of linearly parametrized interatomic potentials, Compu- tational Materials Science 140, 171 (2017)

    2017

  6. [13]

    Gonze, B

    X. Gonze, B. Seddon, J. A. Elliott, C. Tantardini, and A. V. Shapeev, Constrained density functional theory: A potential-based self-consistency approach, Journal of Chemical Theory and Computation 18, 6099 (2022)

    2022

  7. [14]

    Tantardini, D

    C. Tantardini, D. Fazylbekova, S. V. Levchenko, and I. S. Novikov, Quantum modelling of magnetism in strongly correlated materials: Evaluating constrained dft and the hubbard model for y114, Computational Materials Sci- ence 246, 113465 (2025)

    2025

  8. [15]

    M. W. Chase, Nist-janaf thermochemical tables 4th ed., J. of Physical and Chemical Reffernce Data , 1529 (1998)

    1998

  9. [16]

    Zhang and D

    X. Zhang and D. Gall, Crn electronic structure and vi- brational modes: An optical analysis, Physical Review B—Condensed Matter and Materials Physics 82, 045116 (2010)

    2010

  10. [17]

    Alling, T

    B. Alling, T. Marten, and I. Abrikosov, Effect of magnetic disorder and strong electron correlations on the thermo- dynamics of crn, Physical Review B—Condensed Matter and Materials Physics 82, 184430 (2010)

    2010

  11. [19]

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

    2016

  12. [20]

    Gubaev, E

    K. Gubaev, E. V. Podryabinkin, and A. V. Shapeev, Machine learning of molecular properties: Locality and active learning, The Journal of chemical physics 148, 241727 (2018)

    2018

  13. [21]

    A. S. Kotykhov, K. Gubaev, M. Hodapp, C. Tantar- dini, A. V. Shapeev, and I. S. Novikov, Constrained dft- based magnetic machine-learning potentials for magnetic alloys: a case study of fe–al, Scientific Reports 13, 19728 (2023)

    2023

  14. [22]

    Landau and E

    L. Landau and E. Lifshitz, On the theory of the disper- sion of magnetic permeability in ferromagnetic bodies, Phys. Z. Sowjetunion 8, 101 (1935)

    1935

  15. [23]

    T. L. Gilbert, A lagrangian formulation of the gyromag- netic equation of the magnetization field, Phys. Rev.100, 1243 (1955)

    1955

  16. [24]

    A. S. Kotykhov, K. Gubaev, V. Sotskov, C. Tantardini, M. Hodapp, A. V. Shapeev, and I. S. Novikov, Fitting to magnetic forces improves the reliability of magnetic mo- ment tensor potentials, Computational Materials Science 245 (2024)

    2024

  17. [25]

    S. A. Goreinov, I. V. Oseledets, D. V. Savostyanov, E. E. 12 Tyrtyshnikov, and N. L. Zamarashkin, How to find a good submatrix, in Matrix Methods: Theory, Algorithms And Applications: Dedicated to the Memory of Gene Golub (World Scientific, 2010) pp. 247–256

    2010

  18. [26]

    Gonze, B

    X. Gonze, B. Amadon, G. Antonius, F. Arnardi, L. Baguet, J.-M. Beuken, J. Bieder, F. Bottin, J. Bouchet, E. Bousquet, et al. , The abinit project: Im- pact, environment and recent developments, Computer Physics Communications 248, 107042 (2020)

    2020

  19. [27]

    A. H. Romero, D. C. Allan, B. Amadon, G. Anto- nius, T. Applencourt, L. Baguet, J. Bieder, F. Bottin, J. Bouchet, E. Bousquet, et al. , Abinit: Overview and focus on selected capabilities, The Journal of chemical physics 152 (2020)

    2020

  20. [28]

    I. S. Novikov, K. Gubaev, E. V. Podryabinkin, and A. V. Shapeev, The mlip package: moment tensor potentials with mpi and active learning, Machine Learning: Science and Technology 2, 025002 (2020)

    2020

  21. [29]

    Corliss, N

    L. Corliss, N. Elliott, and J. Hastings, Antiferromagnetic structure of crn, Physical Review 117, 929 (1960)

    1960

  22. [30]

    I. A. Abrikosov, A. Ponomareva, P. Steneteg, S. Baran- nikova, and B. Alling, Recent progress in simulations of the paramagnetic state of magnetic materials, Cur- rent Opinion in Solid State and Materials Science 20, 85 (2016)

    2016

  23. [31]

    K¨ ormann, A

    F. K¨ ormann, A. Dick, B. Grabowski, T. Hickel, and J. Neugebauer, Atomic forces at finite magnetic temper- atures: Phonons in paramagnetic iron, Physical Review B—Condensed Matter and Materials Physics 85, 125104 (2012)

    2012

  24. [32]

    Born and K

    M. Born and K. Huang, Dynamical theory of crystal lat- tices (Oxford university press, 1996)

    1996

  25. [33]

    A. Togo, L. Chaput, T. Tadano, and I. Tanaka, Imple- mentation strategies in phonopy and phono3py, J. Phys. Condens. Matter 35, 353001 (2023)

    2023

  26. [34]

    Togo, First-principles phonon calculations with phonopy and phono3py, J

    A. Togo, First-principles phonon calculations with phonopy and phono3py, J. Phys. Soc. Jpn. 92, 012001 (2023)

    2023

  27. [35]

    Y. Wang, J. Wang, W. Wang, Z. Mei, S. Shang, L. Chen, and Z. Liu, A mixed-space approach to first-principles calculations of phonon frequencies for polar materials, Journal of Physics: Condensed Matter 22, 202201 (2010)

    2010

  28. [36]

    Browne, P

    J. Browne, P. Liddell, R. Street, and T. Mills, An investi- gation of the antiferromagnetic transition of crn, physica status solidi (a) 1, 715 (1970)

    1970

  29. [37]

    https://gitlab.com/ivannovikov/datasets_for_ magnetic_MTP/-/tree/main/CrN_paramagnetic_AL

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.