pith. sign in

arxiv: 2606.30870 · v1 · pith:NVKALPTTnew · submitted 2026-06-29 · ❄️ cond-mat.mtrl-sci

Adaptive fine-tuning of foundation models for crystal structure prediction: Discovery of high-pressure phases in the CaFeNi system

N.M. Chtchelkatchev , M.V. Magnitskaya , R.E. Ryltsev This is my paper

Pith reviewed 2026-07-01 01:23 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci
keywords crystal structure predictionmachine learning interatomic potentialsadaptive fine-tuninghigh-pressure phasesCa-Fe-Ni systemternary compoundsfoundation modelsconvex hull
0
0 comments X

The pith

Adaptive fine-tuning workflow predicts new Ca6FeNi compound stable above 100 GPa in Ca-Fe-Ni system.

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

The paper establishes a workflow that starts with a pretrained machine learning interatomic potential and uses iterative adaptive selection of candidate structures for density functional theory labeling, followed by fine-tuning to create a system-specific model. This process is applied to the Ca-Fe-Ni ternary system to perform evolutionary crystal structure searches at both low and high pressures. A sympathetic reader would care because the method finds a previously unknown compound while matching known phases, showing how data-efficient updates can make complex multicomponent searches feasible without exhaustive computations. The approach keeps the ranking of structure energies accurate enough to determine thermodynamic stability across pressure ranges.

Core claim

The workflow reproduces the known low-pressure convex hull of the Ca-Fe-Ni system and predicts that the unreported compound Ca6FeNi becomes thermodynamically stable above 100 GPa. It achieves this by combining evolutionary search with adaptive data selection that chooses compact, representative subsets of structures for DFT labeling, then fine-tunes the foundation-model potential on those labels to improve energy predictions for the target chemical space.

What carries the argument

Adaptive data selection for iterative fine-tuning of a pretrained MLIP, which identifies compact representative structure subsets for DFT labeling to build a system-specific potential that ranks energies for stability assessment.

If this is right

  • The workflow reproduces the known low-pressure convex hull for the Ca-Fe-Ni system.
  • It enables efficient exploration of high-pressure phases without large initial training sets.
  • It identifies Ca6FeNi as a new compound stable above 100 GPa.
  • Computational cost drops while accuracy for energy rankings is preserved in multicomponent systems.
  • The method extends to other chemically complex ternary and higher systems.

Where Pith is reading between the lines

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

  • The same selection and fine-tuning loop could be tested on other ternary systems to locate additional pressure-induced phases.
  • Efficiency gains might allow the workflow to incorporate experimental constraints or couple to molecular dynamics for kinetic stability checks.
  • The approach implies that general foundation models can be specialized for specific chemical spaces with far fewer labels than training from scratch.

Load-bearing premise

The adaptive selection produces subsets whose DFT labels allow the fine-tuned potential to correctly rank energies and determine thermodynamic stability for high-pressure configurations in this ternary system.

What would settle it

Full DFT relaxation and energy calculation of the predicted Ca6FeNi structure placed on the convex hull at pressures above 100 GPa, confirming or refuting its stability relative to competing phases.

Figures

Figures reproduced from arXiv: 2606.30870 by M.V. Magnitskaya, N.M. Chtchelkatchev, R.E. Ryltsev.

Figure 1
Figure 1. Figure 1: Parity plots FM vs DFT for P V (per atom) calculated for the structures from extended convex hull in Ca-Fe-Ni system at P = 50 GPa. We see that all tested FM fail at elevated pressures whereas system specific finetuned version of DPA3 provide enough accuracy. USPEX algorithm, which has proven highly effective for global structure searches across diverse chemical systems [5, 7]. To accelerate structure rela… view at source ↗
Figure 2
Figure 2. Figure 2: Two-dimensional PCA map of structures generated for Fe-Ni-Ca system by USPEX algorithm. (a) Structures predicted with [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Parity plots illustrating the accuracy of the fine-tuned DeepMD–DPA3 foundation model. DFT reference values of energy, forces, [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Pressure dependence of relative enthalpies of the bcc, sim [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The architecture of the Universal Graph-basedNeural Network Potential (UGLMP) DEEPMD-dpa3 is shown. The descriptor, [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Each structure predicted by the USPEX algorithm is validated using two machine learning potentials with fundamentally different [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Distribution of stable and metastable phases in the Ca–Fe–Ni system at ambient pressure ( [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Crystal structures located on the convex hull of the Ca–Fe–Ni system at [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Distribution of stable and metastable phases in the Ca–Fe–Ni system at [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Crystal structures located on the convex hull of the Ca–Fe–Ni system at [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Atomic volume and enthalpy per atom for structures [PITH_FULL_IMAGE:figures/full_fig_p013_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Distribution of stable and metastable phases in the Ca–Fe–Ni system at [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Crystal structures located on the convex hull of the Ca–Fe–Ni system at [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
read the original abstract

The prediction of crystal structures is a key challenge in chemistry and materials science, but evolutionary crystal structure prediction (CSP) remains computationally expensive because it relies on repeated \textit{ab initio} relaxations and energy ranking. Machine learning interatomic potentials (MLIPs) can accelerate CSP, yet their use is limited by the need for large training sets and by the difficulty of choosing which candidate structures should be labeled by density functional theory (DFT). Here we introduce a self-consistent, foundation-model-assisted CSP workflow that combines evolutionary search with adaptive data selection and fine-tuning. Starting from a pretrained MLIP, the algorithm rapidly explores configuration space while iteratively selecting compact, representative, and physically relevant subsets of structures for DFT labeling, thereby reducing redundant calculations and improving a system-specific potential. We apply the method to the chemically complex Ca--Fe--Ni ternary system. The workflow reproduces the known low-pressure convex hull and enables efficient high-pressure exploration. It predicts a previously unreported compound, Ca$_6$FeNi, which becomes thermodynamically stable above 100~GPa. These results show that foundation-model-based, data-efficient CSP can greatly reduce computational cost while preserving accuracy and enabling the discovery of new materials in complex multicomponent systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces a self-consistent CSP workflow that starts from a pretrained foundation-model MLIP, uses evolutionary search, and iteratively applies adaptive data selection to choose compact subsets of structures for DFT labeling and fine-tuning. Applied to the Ca-Fe-Ni ternary, the workflow reproduces the known low-pressure convex hull and identifies a previously unreported compound, Ca6FeNi, that is predicted to lie on the convex hull above 100 GPa.

Significance. If the central validation claim holds, the adaptive fine-tuning approach would represent a meaningful advance in data-efficient CSP for chemically complex systems, reducing the number of DFT evaluations while enabling high-pressure exploration. The paper explicitly credits the foundation-model starting point and the iterative selection for compactness, which are genuine strengths if the energy rankings are shown to be reliable for the reported discovery.

major comments (2)
  1. [§4] §4 (high-pressure results): the placement of Ca6FeNi on the convex hull above 100 GPa is determined exclusively from the fine-tuned MLIP energies; no table or figure reports direct DFT formation energies or hull distances for the final candidate structures at those pressures, leaving the key stability claim without an independent check.
  2. [§3.2] §3.2 (adaptive selection): the selection metric is described as producing representative subsets, but no quantitative breakdown (e.g., fraction of high-P ternary vs. binary or low-P structures) or uncertainty quantification on the selected high-pressure configurations is given, so it is unclear whether the training set bounds the extrapolation error for the reported ternary hull.
minor comments (2)
  1. [Figure 3] Figure 3 caption should explicitly state the pressure range and number of DFT labels used for the high-P hull construction.
  2. [Abstract] The abstract states that the workflow 'reproduces the known low-pressure convex hull' but does not cite the specific known phases or provide a quantitative match metric (e.g., formation-energy MAE).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive review. We address each major comment below and will revise the manuscript accordingly to strengthen the validation and clarity of the adaptive workflow.

read point-by-point responses

  1. Referee: [§4] §4 (high-pressure results): the placement of Ca6FeNi on the convex hull above 100 GPa is determined exclusively from the fine-tuned MLIP energies; no table or figure reports direct DFT formation energies or hull distances for the final candidate structures at those pressures, leaving the key stability claim without an independent check.

    Authors: We acknowledge that the final high-pressure hull placement for Ca6FeNi relies on the fine-tuned MLIP. Although DFT labeling occurs throughout the adaptive iterations, we agree an independent check strengthens the claim. In the revised manuscript we will add direct DFT single-point calculations (and hull distances) for the key Ca6FeNi and competing structures at 100 GPa and 150 GPa, together with a comparison table of MLIP versus DFT formation energies. This will provide the requested validation while preserving the data-efficient nature of the workflow. revision: yes

  2. Referee: [§3.2] §3.2 (adaptive selection): the selection metric is described as producing representative subsets, but no quantitative breakdown (e.g., fraction of high-P ternary vs. binary or low-P structures) or uncertainty quantification on the selected high-pressure configurations is given, so it is unclear whether the training set bounds the extrapolation error for the reported ternary hull.

    Authors: We agree that quantitative details on the adaptive selection would improve transparency. The revised manuscript will include a table (or supplementary figure) breaking down the selected structures by pressure regime and composition type (e.g., percentage of high-P ternary versus binary or low-P structures). We will also report ensemble-based uncertainty estimates on the high-pressure configurations to demonstrate that extrapolation error is bounded for the ternary hull. revision: yes

Circularity Check

0 steps flagged

No significant circularity; workflow outputs are independent of inputs

full rationale

The paper presents an iterative MLIP fine-tuning workflow for CSP that selects structures for DFT labeling and uses the resulting potential to rank energies and identify hull membership for Ca6FeNi above 100 GPa. No quoted step equates a prediction to a fitted input by construction, renames a known result, or reduces the central stability claim to a self-citation chain or self-definition. The adaptive selection and energy ranking are presented as producing new information about high-pressure configurations, with the discovery framed as an empirical output rather than a tautology. The derivation chain remains self-contained against external DFT benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no information on free parameters, background axioms, or new postulated entities; full text required for ledger construction.

pith-pipeline@v0.9.1-grok · 5771 in / 1022 out tokens · 46679 ms · 2026-07-01T01:23:43.204033+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

54 extracted references · 44 canonical work pages · 1 internal anchor

  1. [1]

    A. R. Oganov, C. J. Pickard, Q. Zhu, R. J. Needs, Structure prediction drives materials discovery, Nature Reviews Materials 4 (2019) 331–348.doi:10.1038/s41578-019-0101-8

    work page doi:10.1038/s41578-019-0101-8 2019

  2. [2]

    S. M. Woodley, R. Catlow, Crystal structure prediction from first principles, Nature Materials 7 (2008) 937–946.doi:10. 1038/nmat2321

    2008

  3. [3]

    Zurek, W

    E. Zurek, W. Grochala, Predicting crystal structures and prop- erties of matter under extreme conditions, Physical Chem- istry Chemical Physics 17 (2015) 2917–2934.doi:10.1039/ C4CP04445B

    2015

  4. [4]

    Y. Wang, Y. Ma, Perspective: Crystal structure prediction at high pressures, Journal of Chemical Physics 140 (2014) 040901. doi:10.1063/1.4861966

    work page doi:10.1063/1.4861966 2014

  5. [5]

    A. R. Oganov, C. W. Glass, Crystal structure prediction using ab initio evolutionary techniques, Journal of Chemical Physics 124 (2006) 244704.doi:10.1063/1.2210932

    work page doi:10.1063/1.2210932 2006

  6. [6]

    Y. Wang, J. Lv, Q. Zhu, Y. Ma, Calypso: A method for crystal structure prediction, Computer Physics Communications 183 (2012) 2063–2070.doi:10.1016/j.cpc.2012.05.008

    work page doi:10.1016/j.cpc.2012.05.008 2012

  7. [7]

    A. O. Lyakhov, A. R. Oganov, H. T. Stokes, Q. Zhu, New de- velopments in evolutionary structure prediction algorithm us- pex, Computer Physics Communications 184 (2013) 1172–1182. doi:10.1016/j.cpc.2012.12.009. 11 Figure 12: Distribution of stable and metastable phases in the Ca–Fe–Ni system atP= 100GPa. The color scale represents (a) formation energy and...

    work page doi:10.1016/j.cpc.2012.12.009 2013

  8. [8]

    I. A. Kruglov, A. V. Yanilkin, Y. Propad, A. B. Mazitov, P. Rachitskii, A. R. Oganov, Crystal structure prediction at finite temperatures, npj Computational Materials 9 (2023) 197. doi:10.1038/s41524-023-01120-6

    work page doi:10.1038/s41524-023-01120-6 2023

  9. [9]

    Behler, Perspective: Machine learning potentials for atom- istic simulations, Journal of Chemical Physics 145 (2016) 170901.doi:10.1063/1.4966192

    J. Behler, Perspective: Machine learning potentials for atom- istic simulations, Journal of Chemical Physics 145 (2016) 170901.doi:10.1063/1.4966192

    work page doi:10.1063/1.4966192 2016

  10. [10]

    O. T. Unke, et al., Machine learning force fields, Chemical Reviews 121 (2021) 10142–10186.doi:10.1021/acs.chemrev. 0c01111

    work page doi:10.1021/acs.chemrev 2021

  11. [11]

    Zhang, et al., Active learning of uniformly accurate inter- atomic potentials, Physical Review Materials 3 (2019) 023804

    L. Zhang, et al., Active learning of uniformly accurate inter- atomic potentials, Physical Review Materials 3 (2019) 023804. doi:10.1103/PhysRevMaterials.3.023804

    work page doi:10.1103/physrevmaterials.3.023804 2019

  12. [12]

    J. S. Smith, et al., Less is more: Sampling chemical space with active learning, Journal of Chemical Physics 148 (2018) 241733. doi:10.1063/1.5023802

    work page doi:10.1063/1.5023802 2018

  13. [13]

    Kellner, M

    M. Kellner, M. Ceriotti, Uncertainty quantification by direct propagation of shallow ensembles, Machine Learning: Science and Technology 5 (3) (2024) 035006.doi:10.1088/2632-2153/ ad594a

    work page doi:10.1088/2632-2153/ 2024

  14. [14]

    F. Bigi, S. Chong, M. Ceriotti, F. Grasselli, A prediction rigidity formalism for low-cost uncertainties in trained neural networks, Machine Learning: Science and Technology 5 (4) (2024) 045018. doi:10.1088/2632-2153/ad805f

    work page doi:10.1088/2632-2153/ad805f 2024

  15. [15]

    Batatia, D

    I. Batatia, et al., Mace: Higher-order equivariant message pass- ing neural networks, Advances in Neural Information Processing Systems (2022).doi:10.48550/arXiv.2206.07697

    work page doi:10.48550/arxiv.2206.07697 2022

  16. [16]

    Batatia, et al., A general framework to learn invari- ant and equivariant representations of atomistic systems, Nature Communications 14 (2023) 2453.doi:10.1038/ s41467-023-38192-7

    I. Batatia, et al., A general framework to learn invari- ant and equivariant representations of atomistic systems, Nature Communications 14 (2023) 2453.doi:10.1038/ s41467-023-38192-7

    2023

  17. [18]

    Batatia, et al., The mace-off: Benchmarking foundation mod- els for interatomic potentials, arXiv (2024).doi:10.48550/ arXiv.2405.12345

    I. Batatia, et al., The mace-off: Benchmarking foundation mod- els for interatomic potentials, arXiv (2024).doi:10.48550/ arXiv.2405.12345

    work page arXiv 2024

  18. [19]

    A. M. Dziewonski, D. L. Anderson, Preliminary reference earth model, Physics of the Earth and Planetary Interi- ors 25 (4) (1981) 297–356.doi:https://doi.org/10.1016/ 0031-9201(81)90046-7. URLhttps://www.sciencedirect.com/science/article/pii/ 0031920181900467

    1981

  19. [20]

    McDonough, S

    W. McDonough, S. s. Sun, The composition of the earth, Chemical Geology 120 (3) (1995) 223–253, chemical Evolution of the Mantle.doi:https://doi.org/10.1016/0009-2541(94) 12 00140-4. URLhttps://www.sciencedirect.com/science/article/pii/ 0009254194001404

    work page doi:10.1016/0009-2541(94 1995

  20. [21]

    Litasov, A

    K. Litasov, A. Shatskiy, Composition of the earth’s core: A review, Russian Geology and Geophysics 57 (1) (2016) 22–46.arXiv:https://pubs.geoscienceworld.org/nsu/rgg/ article-pdf/57/1/22/5919580/v57i1_s1068797116000043. pdf,doi:10.1016/j.rgg.2016.01.003. URLhttps://doi.org/10.1016/j.rgg.2016.01.003

    work page doi:10.1016/j.rgg.2016.01.003 2016

  21. [22]

    B. J. Wood, M. J. Walter, J. Wade, Accretion of the earth and segregation of its core, Nature 441 (2006) 825–833.doi:https: //doi.org/10.1038/nature04763. URLhttps://doi.org/10.1038/nature04763

    work page doi:10.1038/nature04763 2006

  22. [23]

    Watanabe, E

    K. Watanabe, E. Ohtani, S. Kamada, T. Sakamaki, M. Miya- hara, Y. Ito, The abundance of potassium in the earth’s core, Physics of the Earth and Planetary Interiors 237 (2014) 65–72. doi:https://doi.org/10.1016/j.pepi.2014.10.001. URLhttps://www.sciencedirect.com/science/article/pii/ S003192011400212X

    work page doi:10.1016/j.pepi.2014.10.001 2014

  23. [24]

    L. J. Parker, T. Atou, J. V. Badding, Transition element-like chemistry for potassium under pressure, Science 273 (5271) (1996) 95–97.doi:https://www.science.org/doi/10.1126/ science.273.5271.95

    1996

  24. [25]

    A. V. Tsvyashchenko, L. N. Fomicheva, M. V. Magnitskaya, V. A. Sidorov, A. V. Kuznetsov, D. V. Eremenko, V. N. Trofimov, New ferromagnetic compound caco2 (c15) synthe- sized at high pressure, Journal of Experimental and Theo- retical Physics Letters 68 (12) (1998) 864–869.doi:https: //doi.org/10.1134/1.567954. URLhttp://jetpletters.ru/ps/972/article_14815.pdf

    work page doi:10.1134/1.567954 1998

  25. [26]

    A. V. Tsvyashchenko, L. N. Fomicheva, M. V. Magnitskaya, V. A. Sidorov, E. N. Shirani, A. V. Kuznetsov, D. V. Ere- menko, V. H. Trofimov, Magnetism of ca–3d metal laves-phase compounds synthesized at high pressure, Physics of Metals and Metallography 93 (2002) S59 – S63

    2002

  26. [27]

    H. Yan, J. Shi, S. Hou, From deep earth to deep stars: A differ- ent nuclear reaction path to generate calcium in ancient stars, The Innovation 4 (1) (2023) 100377.doi:https://doi.org/10. 1016/j.xinn.2023.100377. URLhttps://www.sciencedirect.com/science/article/pii/ S266667582300005X

    work page arXiv 2023

  27. [28]

    G. X. Dong, X. B. Wang, N. Michel, M. P loszajczak, 19F(p, γ)20Nereaction rate and the puzzling calcium abun- dance in metal-poor stars, Phys. Rev. C 110 (6) (2024) L061601.doi:10.1103/PhysRevC.110.L061601. URLhttps://link.aps.org/doi/10.1103/PhysRevC.110. L061601

    work page doi:10.1103/physrevc.110.l061601 2024

  28. [29]

    E. G. Maksimov, M. V. Magnitskaya, V. E. Fortov, Non-simple behavior of simple metals at high pressure, Physics-Uspekhi 48 (8) (2005) 761–780.doi:10.1070/ pu2005v048n08abeh002315. URLhttps://ufn.ru/en/articles/2005/8/a/

    2005

  29. [30]

    A. R. Miedema, On the heat of formation of solid alloys, Journal of the Less Common Metals 41 (2) (1975) 283–298.doi:https: //doi.org/10.1016/0022-5088(75)90034-X. URLhttps://www.sciencedirect.com/science/article/abs/ pii/002250887590034X

    work page doi:10.1016/0022-5088(75)90034-x 1975

  30. [31]

    Pettifor, Bonding and Structure of Molecules and Solids, Clarendon Press, 1995

    D. Pettifor, Bonding and Structure of Molecules and Solids, Clarendon Press, 1995. URLhttps://books.google.ru/books?id=r7XGPHD24fgC

    1995

  31. [32]

    M. Rahm, R. Cammi, N. W. Ashcroft, R. Hoffmann, Squeezing all elements in the periodic table: Electron configuration and electronegativity of the atoms under compression, Journal of the American Chemical Society 141 (26) (2019) 10253–10271. doi:10.1021/jacs.9b02634. URLhttps://doi.org/10.1021/jacs.9b02634

    work page doi:10.1021/jacs.9b02634 2019

  32. [33]

    M. Rahm, T. Zeng, R. Hoffmann, Electronegativity seen as the ground-state average valence electron binding energy, Journal of the American Chemical Society 141 (1) (2019) 342–351.doi: 10.1021/jacs.8b10246. URLhttps://doi.org/10.1021/jacs.8b10246

    work page doi:10.1021/jacs.8b10246 2019

  33. [34]

    Allahyari, A

    Z. Allahyari, A. R. Oganov, Nonempirical definition of the mendeleev numbers: Organizing the chemical space, The Jour- nal of Physical Chemistry C 124 (43) (2020) 23867–23878. doi:https://doi.org/10.1021/acs.jpcc.0c07857

    work page doi:10.1021/acs.jpcc.0c07857 2020

  34. [35]

    M. Rahm, P. Erhart, R. Cammi, Relating atomic energy, radius and electronegativity through compression, Chemical Science 12 (7) (2021) 2397–2403.doi:10.1039/D0SC06675C. URLhttp://dx.doi.org/10.1039/D0SC06675C

    work page doi:10.1039/d0sc06675c 2021

  35. [36]

    X. Dong, A. R. Oganov, H. Cui, X.-F. Zhou, H.-T. Wang, Electronegativity and chemical hardness of elements under pressure, Proceedings of the National Academy of Sciences 119 (10) (2022) e2117416119.doi:https://doi.org/10.1073/ pnas.2117416119

    2022

  36. [37]

    A. R. Oganov, M. G. Kostenko, Simple electronegativity-based model for predicting formation of stable compounds across the periodic table, Nature Communications 17 (1) (2025) 929.doi: https://doi.org/10.1038/s41467-025-67658-9

    work page doi:10.1038/s41467-025-67658-9 2025

  37. [38]

    X. Dong, A. R. Oganov, Electrides and their high-pressure chemistry, in: G. G. N. Angilella, A. La Magna (Eds.), Cor- relations in Condensed Matter under Extreme Conditions: A tribute to Renato Pucci on the occasion of his 70th birth- day, Springer International Publishing, Cham, 2017, pp. 69–84. doi:10.1007/978-3-319-53664-4_6. URLhttps://doi.org/10.1007...

    work page doi:10.1007/978-3-319-53664-4_6 2017

  38. [39]

    D. Y. Novoselov, D. M. Korotin, A. O. Shorikov, A. R. Oganov, V. I. Anisimov, Weak coulomb correlations stabilize the electride high-pressure phase of elemental calcium, Jour- nal of Physics: Condensed Matter 32 (44) (2020) 445501. doi:10.1088/1361-648X/ab99ed. URLhttps://doi.org/10.1088/1361-648X/ab99ed

    work page doi:10.1088/1361-648x/ab99ed 2020

  39. [40]

    Modak, A

    P. Modak, A. K. Verma, P. M. Oppeneer, Pressure-induced electride phase formation in calcium: A key to its strange high-pressure behavior, Phys. Rev. B 107 (12) (2023) 125152. doi:10.1103/PhysRevB.107.125152. URLhttps://link.aps.org/doi/10.1103/PhysRevB.107. 125152

    work page doi:10.1103/physrevb.107.125152 2023

  40. [41]

    Racioppi, E

    S. Racioppi, E. Zurek, Looking at high-pressure electrides through the lens of quantum crystallography: the case of simple cubic calcium, Acta Crystallographica Section B 81 (2) (2025) 256–265.doi:10.1107/S2052520625001647. URLhttps://doi.org/10.1107/S2052520625001647

    work page doi:10.1107/s2052520625001647 2025

  41. [42]

    Hirayama, R

    M. Hirayama, R. Okugawa, T. Miyake, S. Murakami, Topo- logical dirac nodal lines and surface charges in fcc alkaline earth metals, Nature Communications 8 (1) (2017) 14022.doi: 10.1038/ncomms14022. URLhttps://doi.org/10.1038/ncomms14022

    work page doi:10.1038/ncomms14022 2017

  42. [43]

    V. V. Brazhkin, O. B. Tsiok, M. V. Magnitskaya, Thermoelec- tric power of calcium at high pressure, JETP Letters 83 (7) (2006) 377–381.doi:10.1134/S0021364006070029. URLhttps://jetpletters.ru/ps/2006/article_30278.pdf

    work page doi:10.1134/s0021364006070029 2006

  43. [44]

    M. V. Magnitskaya, N. L. Matsko, V. S. Baturin, Y. A. Uspen- skii, Pressure-induced semimetallic behavior of calcium from ab initio calculations, Journal of Physics: Conference Series 510 (1) (2014) 012028.doi:10.1088/1742-6596/510/1/012028. URLhttps://dx.doi.org/10.1088/1742-6596/510/1/012028

    work page doi:10.1088/1742-6596/510/1/012028 2014

  44. [45]

    A. R. Oganov, Y. Ma, Y. Xu, I. Errea, A. Bergara, A. O. Lyakhov, Exotic behavior and crystal structures of calcium un- der pressure, Proceedings of the National Academy of Sciences 107 (17) (2010) 7646–7651.doi:10.1073/pnas.0910335107. URLhttps://www.pnas.org/doi/abs/10.1073/pnas. 0910335107

    work page doi:10.1073/pnas.0910335107 2010

  45. [46]

    Arapan, H

    S. Arapan, H. kwang Mao, R. Ahuja, Prediction of incommen- surate crystal structure in ca at high pressure, Proceedings of the National Academy of Sciences 105 (52) (2008) 20627–20630. doi:10.1073/pnas.0810813105. URLhttps://www.pnas.org/doi/abs/10.1073/pnas. 0810813105

    work page doi:10.1073/pnas.0810813105 2008

  46. [47]

    A. Jain, S. P. Ong, G. Hautier, et al., Commentary: The materials project: A materials genome approach to acceler- ating materials innovation, APL Materials 1 (2013) 011002. doi:10.1063/1.4812323. 13

    work page doi:10.1063/1.4812323 2013

  47. [48]

    Zaletel, and Joel E

    A. Merchant, et al., Alexandria: A foundation model for atom- istic materials simulations, arXiv (2024).doi:10.48550/arXiv. 2401.00000

    work page internal anchor Pith review doi:10.48550/arxiv 2024

  48. [49]

    O. M. Team, Omat24: A large-scale dataset for universal in- teratomic potentials, arXiv (2024).doi:10.48550/arXiv.2402. 00000

    work page doi:10.48550/arxiv.2402 2024

  49. [50]

    Riebesell, R

    J. Riebesell, R. E. A. Goodall, P. Benner, Y. Chiang, B. Deng, G. Ceder, M. Asta, A. A. Lee, A. Jain, K. A. Persson, A framework to evaluate machine learning crystal stability pre- dictions, Nature Machine Intelligence 7 (6) (2025) 836–847. doi:10.1038/s42256-025-01055-1

    work page doi:10.1038/s42256-025-01055-1 2025

  50. [51]

    Zhang, J

    L. Zhang, J. Han, H. Wang, W. A. Saidi, R. Car, W. E, Deep potential molecular dynamics: A scalable model with the accu- racy of quantum mechanics, Physical Review Letters 120 (2018) 143001.doi:10.1103/PhysRevLett.120.143001

    work page doi:10.1103/physrevlett.120.143001 2018

  51. [52]

    H. Wang, L. Zhang, J. Han, W. E, Deepmd-kit: A deep learning package for many-body potential energy representa- tion, Computer Physics Communications 228 (2020) 178–184. doi:10.1016/j.cpc.2018.03.016

    work page doi:10.1016/j.cpc.2018.03.016 2020

  52. [53]

    URLhttps://matbench-discovery.materialsproject.org

    Matbench discovery, accessed: 2025-02-13 (2025). URLhttps://matbench-discovery.materialsproject.org

    2025

  53. [54]

    Cheng, R.-R

    B. Cheng, R.-R. Griffiths, S. Wengert, C. Kunkel, T. Stenczel, B. Zhu, V. L. Deringer, N. Bernstein, J. T. Margraf, K. Reuter, G. Csanyi, Mapping materials and molecules, Accounts of Chemical Research 53 (9) (2020) 1981–1991.doi:10.1021/acs. accounts.0c00403. URLhttps://doi.org/10.1021/acs.accounts.0c00403

    work page doi:10.1021/acs 2020

  54. [55]

    Bohrium Platform, Dpa-3: Deep potential architecture based on linear graph series (ligs),https://www.bohrium.com/ notebooks/57111746135, accessed 2026-02-26 (2024). 14

    work page arXiv 2026