Pith. sign in

REVIEW 3 major objections 5 minor 40 references

The Augmented Potential Method: Multiscale Modeling Toward a Spectral Defect Genome

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

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

Pith's one-line read This paper claims that replacing the quantum-mechanical core of hybrid defect simulations with a universal machine-learned potential gives near-DFT accuracy for grain-boundary segregation energies, enabling a 1,036-alloy spectral database…

desk verdict A genuinely useful integration—ML foundation potential as the QM/MM core to build a 1,036-alloy segregation database—with a narrow validation footprint that leaves the flagship Fe and database-wide accuracy claims as extrapolation. read the letter →

arxiv 2502.08014 v2 pith:CGSYFDUI submitted 2025-02-11 cond-mat.mtrl-sci

Nutth Tuchinda , Changle Li , Christopher A. Schuh This is my paper · ORCID
classification cond-mat.mtrl-sci
keywords augmentedpotentialmethodgrainboundarysegregationmachinelearninginteratomicmultiscalesimulationenergyspectradefectgenomesolute-soluteinteractionsbccironalloys
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 proposes the augmented potential method, a multiscale scheme that uses a universal machine-learning interatomic potential in a small core and buffer around a defect site and a fast classical potential everywhere else. The authors claim this yields grain-boundary segregation energies with accuracy close to density functional theory, while being fast enough to compute segregation spectra for 1,036 binary alloys across 14 solvent metals and 75 solutes. If correct, the resulting spectral database is roughly five times larger than previous compilations and extends to cases previously out of reach, including bcc Fe alloys where magnetism makes DFT-scale multiscale spectra impractical, and solute-solute interaction spectra that multiply the cost 50 to 100 times. A sympathetic reader would care because it offers a practical route from high-throughput interatomic-potential screening to near-quantum defect thermodynamics, and a step toward a systematic defect genome.

What carries the argument

The central mechanism is the augmented potential method: a hybrid multiscale setup in which the core and buffer around each sampled grain-boundary site are evaluated with a universal equivariant machine-learning potential, and the surrounding region by a classical embedded-atom potential, with the buffer size converged on force matching. The central identity is the dilute-limit segregation energy $\Delta E_i^{\mathrm{seg}} = E_i^{\mathrm{GB}} - E^{\mathrm{bulk}}$, the energy change when a solute replaces a solvent atom at a grain-boundary site referenced to the same substitution in the bulk. Site selection uses SOAP descriptors, principal-component reduction, and k-means clustering to choose representative grain-boundary sites, and the resulting spectra are fit to skew-normal distributions.

What would settle it

Re-run APM on the Fe-based systems or on a non-Al grain boundary type and compare the predicted segregation energies site-by-site against converged spin-polarized DFT calculations; an average error above the roughly 10 kJ/mol level of the Al validation would falsify the claim that the database carries quantum-like accuracy.

Watch

Extended reading notes

Core claim

The central claim is that a universal machine-learned potential can take the place of a quantum-mechanical core in a hybrid multiscale calculation without losing quantum-level accuracy for defect energetics. The method defines a core containing the substituted solute site, a buffer zone, and an outer molecular-mechanics region; the core and buffer energies and forces are evaluated by a pretrained universal potential rather than by DFT. Against DFT reference data on the Sigma5(210) grain boundary in Al, the method reproduces segregation energies for nine solutes within an error range similar to QM/MM, over a span exceeding 100 kJ/mol. The authors then apply it to produce grain-boundary segregation spectra for 1,036 binary pairs, bcc Fe-based spectra, and a Cu(Au) solute-solute interaction spectrum, arguing these applications are computationally impractical for quantum-accurate multiscale models.

Load-bearing premise

The method's database inherits its accuracy from the universal machine-learning potentials at low-symmetry grain-boundary sites, a property validated on only one aluminum boundary with nine solutes and not directly checked for the bcc-Fe flagship application.

Editorial extensions

If this is right

  • The 1,036-spectrum database stands as a coarse-grained surrogate for DFT-level segregation thermodynamics across 14 solvents and 75 solutes.
  • bcc Fe-based segregation spectra, including magnetic effects, become accessible; they were previously out of reach for quantum-accurate multiscale methods.
  • Solute-solute interaction spectra can be generated for non-dilute concentrations, with interaction magnitudes up to roughly 100 kJ/mol that are comparable to the segregation energies themselves.
  • The same augmented-potential construction transfers to triple junctions, dislocations, vacancy clusters, and, with modification, interstitial elements and chemically complex alloys.

Reading between the lines

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

  • A direct extrapolation of the paper's logic is that the database can seed mesoscale segregation isotherms without any new first-principles calculations; the paper demonstrates the spectral fits but not the downstream isotherm predictions.
  • The bivariate Gaussian form fitted to the Cu(Au) interaction spectrum suggests a low-dimensional parameterization of non-dilute segregation; the paper leaves deriving the implied concentration dependence to future work.
  • Applying the method to a second, structurally different grain boundary with DFT checks would test whether the Al Sigma5 validation transfers; the paper does not report such a check.
  • As universal potentials improve, the same hybrid architecture could in principle let the learned potential take over more of the cell; the paper keeps a classical region for speed, so the learned-classical boundary is a tunable knob.
Share X Bluesky LinkedIn Reddit HN

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 introduces the Augmented Potential Method (APM), a multiscale scheme in which a universal machine-learning interatomic potential (EquiformerV2 or MACE-omat) replaces the quantum-mechanical core of a QM/MM-style calculation, while a classical potential handles the outer MM region. The method is used to compute dilute-limit grain-boundary segregation energy spectra for 1,036 binary alloy pairs across 14 solvents and 75 solutes, to fit these spectra with skew-normal distributions, and to compute a solute-solute interaction spectrum for Cu(Au). The central validation is a comparison against DFT for nine solutes at the Al Sigma5[001](210) grain boundary (Fig. 2). The paper claims quantum-like accuracy, a roughly fivefold expansion of the available spectral database, and first-of-their-kind Fe-based and solute-solute interaction spectra that are argued to be beyond the reach of quantum-accurate multiscale methods.

Significance. If the transferability claim holds, the 1,036-spectrum database and the APM workflow would be a substantial practical contribution: the method would provide a low-cost route to defect thermodynamics for many alloy systems, including magnetic bcc Fe, and would open solute-solute interaction spectra to systematic computation. The paper has notable strengths: the DFT benchmark in Fig. 2 is external, no APM parameter was fit to reproduce that benchmark, the statistical summaries (skew-normal parameters) are computed from the simulated spectra rather than used as inputs, and the spectra and fitting parameters are made available in the supplemental material. The significance is nevertheless conditional: the single-solvent, single-boundary validation is narrow relative to the breadth of the claimed database, and the Fe results that are showcased as a flagship application have no direct reference validation. The proposed method is promising, but the evidence as presented supports a more cautious statement of accuracy than 'quantum-like' for all 1,036 systems.

major comments (3)
  1. [Fig. 2 and SM3] The transferability claim rests on one validation: nine solutes at the Al Sigma5[001](210) boundary, with all three unique sites, benchmarked against DFT from Ref. [71]. The paper then generalizes to 14 solvents, 75 solutes, and random polycrystals including magnetic bcc Fe, but no per-solvent validation is shown for any other solvent. The free-surface validation mentioned in the text is only cited (Ref. [72]) and not displayed. Because the database inherits any systematic error of the foundation model at low-symmetry sites, the claim of 'quantum-like accuracy' for the full 1,036-system database is currently under-supported. I would ask for at least a small set of additional DFT or literature validations covering a few chemically distinct solvents (e.g., one fcc, one bcc, one magnetic system) or, failing that, a clear statement that the accuracy is demonstrated only for Al and extrapolated elsewhere.
  2. [Fig. S1 and Eq. (S1)] The buffer-size convergence test in Fig. S1 is performed on bulk lattices, not on grain-boundary sites. Since the segregation energy in Eq. (S1) is the difference of two large total energies evaluated at a defect site and at a bulk site, a small systematic force or energy error that appears only at low-coordination, low-symmetry GB sites would not be detected by a bulk-lattice convergence test. The paper should either provide convergence or validation data at GB sites for the solvents used, or explicitly discuss how errors at GB sites are controlled. As written, the method has no internal checkpoint to detect a failure of the foundation model at the sites of interest.
  3. [Fig. 4 and Section 4] The Fe-based segregation spectra in Fig. 4 are presented as a 'state-of-the-art accuracy' demonstration, but no DFT or other quantum reference is provided for any Fe system. The difficulty of DFT for magnetic Fe is acknowledged, but this does not remove the need for some benchmark. Given that the entire Fe database is unvalidated, the phrase 'with state-of-the-art accuracy' (Section 4) is stronger than the evidence supports. I would ask the authors to either validate a subset of Fe segregation energies against available DFT data from the literature, or to rephrase the claim to indicate that the Fe spectra are predictions whose accuracy has not been directly established.
minor comments (5)
  1. [Abstract] The abstract states '1,050 binary alloy pairs' while the main text and Fig. 1 cite 1,036; the numbers should be reconciled.
  2. [Table S1] The header 'Butter Size' should be 'Buffer Size'.
  3. [Fig. 2 caption and text] The text says APM accuracy is 'of a similar magnitude to that achieved by a multiscale method with a quantum-mechanical calculation at the core (QM/MM)', but Fig. 2 does not show QM/MM results; the comparison is only APM vs DFT. Either add the QM/MM points or soften the wording.
  4. [Section 1 and SM1] The SOAP hyperparameters (cutoff, smearing width, n_max, l_max) are stated in the text, but the exact cutoff value is deferred to the supplemental material. Since this is a reproducibility-relevant detail, it would be helpful to state it explicitly in the main text.
  5. [Eq. (4)] The bivariate Gaussian in Eq. (4) is fitted to the (Delta_E_seg, omega_GB) data, but no goodness-of-fit or scatter plot is shown; a brief quantitative assessment (e.g., correlation coefficient) would help the reader judge the adequacy of the bivariate model.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: APM spectra are computed from external universal ML potentials and checked against independent DFT; skew-normal and covariance parameters are fitted outputs, not inputs.

full rationale

The paper's load-bearing quantities are segregation energies obtained from Eq. (S1) using a universal ML potential (EquiformerV2 or MACE) in the core and buffer, with a classical potential outside; the ML potentials are external models not fitted to any APM output or to the validation set. The skew-normal parameters of Eq. (2) and the bivariate-Gaussian parameters of Eq. (4) are statistical summaries fitted after the spectra and interaction coefficients are computed, so the tabulated database is not a fitted parameter renamed as a prediction. The validation in Fig. 2 compares APM against DFT data from Ref. [71]; although that reference shares authors with this paper, it is a separate first-principles calculation with no APM parameters in it, and it is therefore independent support rather than a circular input. Self-citations to prior Wagih–Schuh and Tuchinda–Schuh work supply the polycrystal sampling workflow, the SOAP/PCA site-selection approach, and the QM/MM-style buffer-convergence methodology, but those are methodological precedents and do not mathematically force the reported spectra. The genuine scientific risk — that ML-potential accuracy at low-symmetry grain-boundary sites is validated only for Al Σ5(210) and is unvalidated for magnetic bcc Fe and most solvent–solute pairs — is a transferability and correctness concern, not circularity, because the derivation does not assume the accuracy it claims. No step in the derivation chain reduces by construction to its own input, so the circularity score is 0.

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

The main calculation leans on pre-trained foundation models whose accuracy is inherited from their DFT training data, on sampling hyperparameters taken from prior literature, and on per-solvent buffer sizes chosen from bulk-lattice convergence tests. Six settings are effectively free parameters, dominated by the skew-normal fits that are the database output. No invented physical entities appear. The epistemic load concentrates in the unvalidated transferability of the ML potentials to non-Al solvents and to magnetic Fe.

free parameters (6)
  • skew-normal fit parameters (alpha, mu, sigma) = Per-alloy values tabulated in the supplemental spreadsheet
    Fitted to each of the 1,036 computed spectra; these fitted parameters are the tabulated database output used for the segregation strength color scale.
  • SOAP hyperparameters = Cutoff in SM, width 0.5 A, nmax=10, lmax=5
    Chosen by hand following prior work (Refs 40-42); the descriptor construction determines what the sampling algorithm sees.
  • Number of principal components and k-means centroids = PCs 1-10, 100 centroids
    Follows the established spectral workflow (Ref 9); affects which sites are selected for the expensive core calculation.
  • Per-solvent core and buffer sizes = Table S1, e.g., Ni 4.26/6.5 A
    Set by force-convergence tests on bulk lattices (Fig. S1), not on GB sites; the buffer size controls accuracy of the hybrid coupling.
  • FIRE force tolerance = 0.02 eV/A (0.03 for V)
    Chosen below typical QMMM tolerances (Ref 30); relaxation quality depends on it.
  • MM potential scaling factors = Per Refs 30-31 scaling
    Empirical scaling to reduce mechanical mismatch between the ML and MM potentials at the interface.
assumptions (6)
  • domain assumption The skew-normal distribution captures general polycrystalline GB segregation spectra.
    Assumed in Eq. (2) and used to summarize all 1,036 spectra; justified by prior work (Refs 10, 23) rather than tested here.
  • domain assumption The 12-grain polycrystal model represents random polycrystal GB environments.
    Used to generate all sampling structures; cited to Refs 23-27, not re-validated in this work.
  • domain assumption 100 k-means centroids in the reduced SOAP space adequately represent the GB site distribution.
    Sampling step in Fig. 1d; inherited from Ref 9 without convergence testing here.
  • domain assumption Universal ML potentials are accurate at low-symmetry defect sites for all solvents and solutes in the database.
    The core of the method's accuracy claim; validated only on Al Sigma5(210) with 9 solutes and on cited free-surface checks (Ref 72).
  • standard math The difference-of-differences definition of segregation energy cancels constant energy offsets between the ML core and the MM region.
    Eq. (S1) defines the segregation energy as a double difference, so reference-energy offsets between potentials cancel; this is the mathematical basis of the hybrid coupling.
  • domain assumption DFT (PBE level) is the appropriate ground truth for segregation energies.
    Standard practice in the field; the validation and all accuracy claims are relative to DFT, not to experiment.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Augmented Potential Method: Multiscale Modeling Toward a Spectral Defect Genome." pith.science (2026). https://pith.science/paper/CGSYFDUI

@misc{pith2026250208014,
  author       = {Pith},
  title        = {Pith review of: The Augmented Potential Method: Multiscale Modeling Toward a Spectral Defect Genome},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CGSYFDUI}},
  note         = {Machine review of arXiv:2502.08014}
}
read the original abstract

The modeling of solute chemistry at low-symmetry defects in materials is historically challenging, due to the computation cost required to evaluate thermodynamic properties from first principles. Here, we offer a hybrid multiscale approach called the augmented potential method that connects the chemical flexibility and near-quantum accuracy of a universal machine learning potential at the site of the defect, with the computational speed of a long-range classical potential implemented away from the defect site in a buffer zone. The method allows us to rapidly compute distributions of grain boundary segregation energy for 1,050 binary alloy pairs (including Ag, Al, Au, Cr, Cu, Fe, Mo, Nb, Ni, Pd, Pt, Ta and V, W solvent), creating a database for polycrystalline grain boundary segregation. This database is ~5x larger than previously published spectral compilations, and yet has improved accuracy. The approach can also address problems far beyond the reach of any other method, such as handling bcc Fe-based alloys, or the complex solute-solute interactions in random polycrystals. The approach thus paves a pathway toward a complete defect genome in crystalline materials.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 30 canonical work pages

  1. [1]

    A. P. Thompson et al., LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales, Computer Physics Communications 271, 108171 (2022)

    work page 2022

  2. [2]

    W. M. Brown, A. Kohlmeyer, S. J. Plimpton, and A. N. Tharrington, Implementing molecular dynamics on hybrid high performance computers–Particle–particle particle-mesh, Computer Physics Communications 183, 449 (2012)

    work page 2012

  3. [3]

    W. M. Brown, P. Wang, S. J. Plimpton, and A. N. Tharrington, Implementing molecular dynamics on hybrid high performance computers–short range forces, Computer Physics Communications 182, 898 (2011)

    work page 2011

  4. [4]

    W. M. Brown and M. Yamada, Implementing molecular dynamics on hybrid high performance computers—three-body potentials, Computer Physics Communications 184, 2785 (2013)

    work page 2013

  5. [5]

    P. L. Williams, Y. Mishin, and J. C. Hamilton, An embedded-atom potential for the Cu–Ag system, Modelling and Simulation in Materials Science and Engineering 14, 817 (2006)

    work page 2006

  6. [6]

    M. I. Mendelev, M. J. Kramer, C. A. Becker, and M. Asta, Analysis of semi-empirical interatomic potentials appropriate for simulation of crystalline and liquid Al and Cu, Philosophical Magazine 88, 1723 (2008)

    work page 2008

  7. [7]

    P. A. T. Olsson, Transverse resonant properties of strained gold nanowires, Journal of Applied Physics 108, 034318 (2010)

    work page 2010

  8. [8]

    S. M. Eich, D. Beinke, and G. Schmitz, Embedded-atom potential for an accurate thermodynamic description of the iron–chromium system, Computational Materials Science 104, 185 (2015)

    work page 2015

Show all 40 references
  1. [9]

    M. I. Mendelev and A. H. King, The interactions of self-interstitials with twin boundaries, Philosophical Magazine 93, 1268 (2013)

    2013

  2. [10]

    Y. Chen, X. Liao, N. Gao, W. Hu, F. Gao, and H. Deng, Interatomic potentials of W–V and W–Mo binary systems for point defects studies, Journal of Nuclear Materials 531, 152020 (2020)

    2020

  3. [11]

    M. R. Fellinger, H. Park, and J. W. Wilkins, Force-matched embedded-atom method potential for niobium, Phys. Rev. B 81, 144119 (2010)

    2010

  4. [12]

    M. I. Mendelev, M. J. Kramer, S. G. Hao, K. M. Ho, and C. Z. Wang, Development of interatomic potentials appropriate for simulation of liquid and glass properties of NiZr2 alloy, Philosophical Magazine 92, 4454 (2012)

    2012

  5. [13]

    S. M. Foiles, M. I. Baskes, and M. S. Daw, Embedded-atom-method functions for the fcc metals Cu, Ag, Au, Ni, Pd, Pt, and their alloys, Phys. Rev. B 33, 7983 (1986)

    1986

  6. [14]

    Hirata, L

    A. Hirata, L. J. Kang, T. Fujita, B. Klumov, K. Matsue, M. Kotani, A. R. Yavari, and M. W. Chen, Geometric frustration of icosahedron in metallic glasses, Science 341, 376 (2013). 9

    2013

  7. [15]

    Y. Chen, J. Fang, L. Liu, W. Hu, N. Gao, F. Gao, and H. Deng, Development of the interatomic potentials for W-Ta system, Computational Materials Science 163, 91 (2019)

    2019

  8. [16]

    M. I. Mendelev, S. Han, W. Son, G. J. Ackland, and D. J. Srolovitz, Simulation of the interaction between Fe impurities and point defects in V, Phys. Rev. B 76, 214105 (2007)

    2007

  9. [17]

    D. R. Mason, D. Nguyen-Manh, and C. S. Becquart, An empirical potential for simulating vacancy clusters in tungsten, Journal of Physics: Condensed Matter 29, 505501 (2017)

    2017

  10. [18]

    Chanussot, A

    L. Chanussot, A. Das, S. Goyal, T. Lavril, M. Shuaibi, M. Riviere, K. Tran, J. Heras-Domingo, C. Ho, and W. Hu, Open catalyst 2020 (OC20) dataset and community challenges, Acs Catalysis 11, 6059 (2021)

    2021

  11. [19]

    Barroso-Luque, M

    L. Barroso-Luque, M. Shuaibi, X. Fu, B. M. Wood, M. Dzamba, M. Gao, A. Rizvi, C. L. Zitnick, and Z. W. Ulissi, Open materials 2024 (omat24) inorganic materials dataset and models, arXiv Preprint arXiv:2410.12771 (2024)

    2024 arXiv

  12. [20]

    Y.-L. Liao, B. Wood, A. Das, and T. Smidt, Equiformerv2: Improved equivariant transformer for scaling to higher-degree representations, arXiv Preprint arXiv:2306.12059 (2023)

    2023 arXiv

  13. [21]

    B. Deng, P. Zhong, K. Jun, J. Riebesell, K. Han, C. J. Bartel, and G. Ceder, CHGNet: Pretrained Universal Neural Network Potential for Charge-Informed Atomistic Modeling, arXiv:2302.14231

    arXiv

  14. [22]

    Schmidt, T

    J. Schmidt, T. F. T. Cerqueira, A. H. Romero, A. Loew, F. Jäger, H.-C. Wang, S. Botti, and M. A. L. Marques, Improving machine-learning models in materials science through large datasets, Materials Today Physics 48, 101560 (2024)

    2024

  15. [23]

    Passaro and C

    S. Passaro and C. L. Zitnick, Reducing SO (3) Convolutions to SO (2) for Efficient Equivariant GNNs, in International Conference on Machine Learning (PMLR, 2023), pp. 27420–27438

    2023

  16. [24]

    Batatia, D

    I. Batatia, D. P. Kovacs, G. Simm, C. Ortner, and G. Csányi, MACE: Higher order equivariant message passing neural networks for fast and accurate force fields, Advances in Neural Information Processing Systems 35, 11423 (2022)

    2022

  17. [25]

    Yuan Chiang, mace-universal, (n.d.)

  18. [26]

    Batatia, S

    I. Batatia, S. Batzner, D. P. Kovács, A. Musaelian, G. N. C. Simm, R. Drautz, C. Ortner, B. Kozinsky, and G. Csányi, The design space of E(3)-equivariant atom-centred interatomic potentials, Nature Machine Intelligence 7, 56 (2025)

    2025

  19. [27]

    Batatia, P

    I. Batatia, P. Benner, Y. Chiang, A. M. Elena, D. P. Kovács, J. Riebesell, X. R. Advincula, M. Asta, M. Avaylon, and W. J. Baldwin, A foundation model for atomistic materials chemistry, arXiv Preprint arXiv:2401.00096 (2023)

    2023 arXiv

  20. [28]

    Bitzek, P

    E. Bitzek, P. Koskinen, F. Gähler, M. Moseler, and P. Gumbsch, Structural Relaxation Made Simple, Phys. Rev. Lett. 97, 170201 (2006)

    2006

  21. [29]

    Guénolé, W

    J. Guénolé, W. G. Nöhring, A. Vaid, F. Houllé, Z. Xie, A. Prakash, and E. Bitzek, Assessment and optimization of the fast inertial relaxation engine (fire) for energy minimization in atomistic simulations and its implementation in lammps, Computational Materials Science 175, 1...

    2020

  22. [30]

    Wagih and C

    M. Wagih and C. A. Schuh, Learning Grain-Boundary Segregation: From First Principles to Polycrystals, Phys. Rev. Lett. 129, 046102 (2022)

    2022

  23. [31]

    T. D. Swinburne and J. R. Kermode, Computing energy barriers for rare events from hybrid quantum/classical simulations through the virtual work principle, Phys. Rev. B 96, 144102 (2017)

    2017

  24. [32]

    Tuchinda, G

    N. Tuchinda, G. B. Olson, and C. A. Schuh, Grain Boundary Segregation and Embrittlement of Aluminum Binary Alloys from First Principles, Acta Materialia 121058 (2025)

    2025

  25. [33]

    Zhang, O

    S. Zhang, O. Y. Kontsevoi, A. J. Freeman, and G. B. Olson, First principles investigation of zinc- induced embrittlement in an aluminum grain boundary, Acta Materialia 59, 6155 (2011)

    2011

  26. [34]

    Hadian, B

    R. Hadian, B. Grabowski, and J. Neugebauer, GB code: A grain boundary generation code, The Journal of Open Source Software 3, (2018)

    2018

  27. [35]

    Kresse and J

    G. Kresse and J. Hafner, Ab initio molecular dynamics for liquid metals, Phys. Rev. B 47, 558 (1993)

    1993

  28. [36]

    Kresse and J

    G. Kresse and J. Furthmüller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Computational Materials Science 6, 15 (1996). 10

    1996

  29. [37]

    Kresse and J

    G. Kresse and J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Phys. Rev. B 54, 11169 (1996)

    1996

  30. [38]

    Kresse and D

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

    1999

  31. [39]

    P. E. Blöchl, Projector augmented-wave method, Phys. Rev. B 50, 17953 (1994)

    1994

  32. [40]

    J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple, Phys. Rev. Lett. 77, 3865 (1996)

    1996

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.