arxiv: 2605.07927 · v1 · submitted 2026-05-08 · ❄️ cond-mat.mtrl-sci

Recognition: no theorem link

MatterSim-MT: A multi-task foundation model for in silico materials characterization

Andrew Fowler, Chang Liu, Chenxi Hu, Claudio Zeni, Daniel Z\"ugner, Deniz Gunceler, Fabian Thiemann, Frank No\'e, Guanzhi Li, Han Yang, Hongxia Hao, Jielan Li, Junfu Tan, Lingyu Kong, Lixin Sun, Matthew Horton, Qian Wang, Robert Pinsler, Shuizhou Chen, Tian Xie, Xixian Liu, Yeqi Bai, Yicheng Chen, Yichi Zhou, Yu Shi, Yu Zhu, Zekun Chen, Ziheng Lu

Authors on Pith no claims yet

Pith reviewed 2026-05-11 03:06 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci

keywords multi-task foundation modelmaterials simulationphonon splittingferroelectric hysteresisredox transitionfirst-principles datain silico characterization

0 comments

The pith

A multi-task foundation model pretrained on millions of atomic structures can predict complex material behaviors such as pressure-dependent phonon splitting and ferroelectric hysteresis directly from its architecture.

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

The paper introduces MatterSim-MT as a single model that handles structure prediction, dynamics, thermodynamics and additional properties after broad pretraining on first-principles data. It is trained on over 35 million labeled structures spanning 89 elements under extreme temperatures and pressures, then fine-tuned on quantities including Bader charges, magnetic moments, Born effective charges and dielectric matrices. This setup lets the model perform simulations of phenomena that potential energy surfaces alone cannot capture, with reported agreement to experiment on silicon carbide phonon splitting under pressure, hysteresis loops in barium titanate, and redox shifts in lithium-rich battery materials. Accurate and scalable in silico characterization would shorten the time needed to evaluate candidate materials for real-world use. The authors also show the model improves with more data or parameters and adapts to new systems through active learning.

Core claim

MatterSim-MT, a multi-task foundation model pretrained on over 35 million first-principles-labeled structures covering 89 elements, temperatures up to 5000 K and pressures up to 1000 GPa, and fine-tuned on Bader charges, magnetic moments, Born effective charges and dielectric matrices, serves as a foundation model for material structure, dynamics and thermodynamics while also enabling complex simulations including pressure-dependent LO-TO phonon splitting in SiC, electric hysteresis in ferroelectric BaTiO3, and the cationic-to-anionic redox transition during delithiation of a Li-rich cathode material, all with close agreement to experiment.

What carries the argument

The multi-task architecture that jointly predicts structural, dynamical, thermodynamic and property-specific outputs such as charges and dielectric responses after broad pretraining.

If this is right

The model scales with additional data and parameters while maintaining performance.
It can be efficiently fine-tuned to higher levels of theory beyond the initial training.
New systems can be incorporated through active learning without full retraining.
It supports simulations of phenomena beyond those describable by potential energy surfaces alone.

Where Pith is reading between the lines

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

A single model handling multiple properties at once could reduce the computational cost of screening large libraries of candidate materials for specific applications.
Integration with experimental data streams might create closed-loop workflows that iteratively improve predictions for real devices.
The approach may lower the barrier for simulating coupled electro-mechanical or electro-chemical processes that currently require separate specialized tools.

Load-bearing premise

Broad pretraining on millions of structures plus fine-tuning on the listed properties is sufficient to capture and generalize to the demonstrated complex phenomena without task-specific retraining or hidden fitting that affects the reported experimental agreement.

What would settle it

Direct experimental measurements of LO-TO phonon frequencies in SiC under varying pressures that deviate substantially from the model's predictions, or electric polarization loops in BaTiO3 that fail to match the simulated hysteresis shapes.

read the original abstract

Accurate property characterization is a major bottleneck in materials design. While first-principles methods and task-specific machine-learning models have driven important progress, they remain fundamentally limited in scalability and generalizability across the vast space of structures and properties relevant to real-world materials design. We present MatterSim-MT, a multi-task foundation model for in silico materials simulation and property characterization. The model is pretrained on over 35 million first-principles-labeled structures covering 89 elements, temperatures up to 5000 K and pressures up to 1000 GPa, and is fine-tuned on various properties including Bader charges, magnetic moments, Born effective charges, and dielectric matrices. Out of the box, MatterSim-MT not only serves as a foundation model for predicting material structure, dynamics and thermodynamics, its multi-task architecture also enables a wide range of complex simulations that cannot be captured by potential energy surfaces alone. For example, we demonstrate pressure-dependent LO-TO phonon splitting in SiC with close agreement with experiment, electric hysteresis in ferroelectric BaTiO3, and the cationic-to-anionic redox transition during delithiation of a Li-rich cathode material. Finally, we show that MatterSim-MT scales well with more data and parameters, can be efficiently fine-tuned to higher levels of theory, and can be efficiently extended to new systems via active learning. Overall, we believe this approach provides a scalable route to accurate in silico materials characterization.

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.

Desk Editor's Note private letter to a colleague

MatterSim-MT shows a large-scale pretrain plus multi-task fine-tune setup that can run a few complex simulations from the listed properties, but the abstract leaves the quantitative support and exact inference steps unclear.

read the letter

The core claim is that pretraining on 35 million structures across many elements and conditions, then fine-tuning on Bader charges, magnetic moments, Born charges, and dielectric matrices, lets the model handle structure, dynamics, thermodynamics, and also things like pressure-dependent LO-TO splitting in SiC, ferroelectric hysteresis in BaTiO3, and redox shifts in a Li-rich cathode, all with close experimental match and without extra per-task retraining. That combination and the three concrete demos look new relative to the cited prior work. The scaling behavior with more data and parameters, plus the active-learning extension note, is a practical plus that people building materials models will notice right away. The paper also states it can be fine-tuned to higher levels of theory, which is useful if it holds up in the full results. The soft spot is the lack of visible error bars, baseline comparisons, data-split details, or a transparent recipe for how the three simulation examples were actually run from the fine-tuned model. The stress-test point lands: it is not obvious from the abstract whether those LO-TO, hysteresis, and redox results follow directly from the listed fine-tuning objectives or whether they needed additional MD protocols, post-processing choices, or system-specific adjustments that could affect the reported agreement. Without those steps shown, the central claim that the multi-task architecture enables the phenomena out of the box rests on thin evidence so far. This paper is aimed at groups already working on foundation models for materials or who need scalable property predictors for energy and catalysis applications. A reader who wants to see how far a single model can stretch across response properties will get value from the scale and the demo ideas, even if they have to wait for the methods section to judge reproducibility. It deserves a serious referee because the pretraining scale and the attempt to link multi-task outputs to complex simulations are worth checking in detail, even if the current write-up needs more quantitative grounding and clearer inference protocols before it can be taken as strong support for the headline claim.

Referee Report

3 major / 1 minor

Summary. The paper introduces MatterSim-MT, a multi-task foundation model pretrained on over 35 million first-principles structures spanning 89 elements, temperatures to 5000 K and pressures to 1000 GPa. It is fine-tuned on properties including Bader charges, magnetic moments, Born effective charges and dielectric matrices. The central claim is that the resulting model serves as a foundation for structure, dynamics and thermodynamics predictions while also enabling complex simulations (pressure-dependent LO-TO splitting in SiC, ferroelectric hysteresis in BaTiO3, cationic-to-anionic redox crossover in a Li-rich cathode) with close experimental agreement, all without task-specific retraining, and that the approach scales with data and parameters.

Significance. If the reported agreements and zero-shot capabilities are substantiated, the work would offer a meaningful step toward scalable, general-purpose in silico materials characterization that extends beyond conventional interatomic potentials. The scale of the pretraining corpus and the explicit multi-task fine-tuning on response properties constitute concrete strengths that could support broader adoption if accompanied by rigorous validation.

major comments (3)

[Results on complex simulations] Demonstration sections (LO-TO splitting, hysteresis, redox examples): the manuscript states 'close agreement with experiment' for the three complex phenomena but supplies no quantitative error metrics, error bars, baseline comparisons to existing models or methods, or details on data splits and independent validation sets. These omissions make it impossible to evaluate whether the multi-task architecture genuinely delivers the claimed accuracy.
[Model architecture and fine-tuning] Methods and inference description: it remains unclear whether the reported LO-TO, hysteresis and redox results follow directly from the listed fine-tuning objectives (Bader charges, Born charges, dielectric matrices) in a zero-shot manner or require additional, unstated MD protocols, property post-processing, system-specific adjustments or hidden data choices. This directly affects the central claim that the multi-task model enables such phenomena 'out of the box'.
[Methods] Validation and reproducibility: no information is provided on training/validation/test splits for the 35 M structures or the fine-tuning datasets, nor on how the model was tested for generalization to the specific demonstration systems. This is load-bearing for assessing scalability and reliability claims.

minor comments (1)

[Abstract] Abstract and introduction: the phrase 'out of the box' is used without a precise definition of the inference protocol; a short clarifying sentence would improve readability.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive review. The comments highlight important areas for strengthening the quantitative validation, methodological transparency, and reproducibility of our work. We address each major comment below and will revise the manuscript accordingly to provide the requested details and clarifications.

read point-by-point responses

Referee: Demonstration sections (LO-TO splitting, hysteresis, redox examples): the manuscript states 'close agreement with experiment' for the three complex phenomena but supplies no quantitative error metrics, error bars, baseline comparisons to existing models or methods, or details on data splits and independent validation sets. These omissions make it impossible to evaluate whether the multi-task architecture genuinely delivers the claimed accuracy.

Authors: We agree that additional quantitative metrics are necessary to substantiate the claims. In the revised manuscript, we will add explicit error metrics (e.g., MAE and RMSE) comparing simulated LO-TO frequencies, hysteresis loop areas, and redox potentials against experimental values, along with error bars obtained from multiple independent MD runs with different random seeds. We will also include baseline comparisons to direct DFT calculations and, where feasible, to existing ML potentials such as MACE or NequIP on the same systems. Regarding data splits, we will explicitly confirm that the demonstration systems (SiC, BaTiO3, and the Li-rich cathode) were held out from both pretraining and fine-tuning datasets, with the pretraining split performed by elemental composition and thermodynamic conditions to prevent leakage. revision: yes
Referee: Methods and inference description: it remains unclear whether the reported LO-TO, hysteresis and redox results follow directly from the listed fine-tuning objectives (Bader charges, Born charges, dielectric matrices) in a zero-shot manner or require additional, unstated MD protocols, property post-processing, system-specific adjustments or hidden data choices. This directly affects the central claim that the multi-task model enables such phenomena 'out of the box'.

Authors: The simulations are performed in a zero-shot manner using the fine-tuned model outputs. For pressure-dependent LO-TO splitting in SiC, we extract Born effective charges and dielectric matrices directly from the model and apply standard lattice dynamics post-processing (no additional training). For ferroelectric hysteresis in BaTiO3, the model serves as the interatomic potential in MD, with polarization computed from the fine-tuned Born charges and dielectric response. For the redox crossover, Bader charges and magnetic moments are used to track oxidation states during MD trajectories. We will expand the Methods section with a new subsection detailing these exact inference workflows, including any standard post-processing (e.g., Fourier analysis for phonons) and confirming that no system-specific retraining, hidden data, or non-standard adjustments were applied. This preserves the 'out of the box' multi-task claim. revision: yes
Referee: Validation and reproducibility: no information is provided on training/validation/test splits for the 35 M structures or the fine-tuning datasets, nor on how the model was tested for generalization to the specific demonstration systems. This is load-bearing for assessing scalability and reliability claims.

Authors: We will add a dedicated 'Data Splits and Generalization' subsection in the Methods. The 35 million pretraining structures were split 80/10/10 (train/validation/test) with stratification by elemental composition, temperature, and pressure ranges. Fine-tuning datasets for Bader charges, magnetic moments, Born charges, and dielectric matrices follow the same protocol, with sources and sizes now specified. We will explicitly state that the three demonstration systems were excluded from all training and fine-tuning data and report model performance on the held-out test sets to support the generalization claims. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper presents a standard multi-task ML foundation model pretrained on external first-principles calculations (35M structures) and fine-tuned on auxiliary properties (Bader charges, Born effective charges, dielectric matrices). The reported demonstrations of LO-TO splitting, ferroelectric hysteresis, and redox transitions are applications of the trained model to new simulation protocols, not reductions of the claimed outputs to the training inputs by definition or statistical forcing. No self-definitional equations, fitted parameters renamed as predictions, or load-bearing self-citations that collapse the central claims are present. The derivation chain remains self-contained against external DFT data and experimental benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the representativeness of the 35M first-principles structures and the sufficiency of the chosen fine-tuning tasks; no new physical entities are introduced.

free parameters (1)

Model weights and architecture hyperparameters
Billions of parameters fitted during pretraining on 35 million structures and subsequent fine-tuning on property tasks.

axioms (1)

domain assumption First-principles calculations supply sufficiently accurate labels for the pretraining and fine-tuning tasks
The entire pipeline treats these calculations as ground truth without independent verification of their accuracy for the demonstrated properties.

pith-pipeline@v0.9.0 · 5663 in / 1418 out tokens · 59895 ms · 2026-05-11T03:06:20.929728+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

65 extracted references · 65 canonical work pages · 1 internal anchor

[1]

T. Chen, S. Luo, D. He, S. Zheng, T.Y. Liu, L. Wang, Geomformer: A general architecture for geometric molecular representation learning (2023)

work page 2023
[2]

H. Yang, C. Hu, Y. Zhou, X. Liu, Y. Shi, J. Li, G. Li, Z. Chen, S. Chen, C. Zeni, et al., Mattersim: A deep learning atomistic model across elements, temperatures and pressures. arXiv preprint arXiv:2405.04967 (2024)

work page arXiv 2024
[3]

Vaswani, Attention is all you need

A. Vaswani, Attention is all you need. Advances in Neural Information Processing Systems (2017)

work page 2017
[4]

Xie, J.C

T. Xie, J.C. Grossman, Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties. Physical review letters120(14), 145301 (2018)

work page 2018
[5]

Zhong, H

Y. Zhong, H. Yu, X. Gong, H. Xiang, Edge-based tensor prediction via graph neural networks. arXiv preprint arXiv:2201.05770 (2022) 28 Fig. S14: MAX force error against first-principles results in the test set from the finetuned model and the model trained from scratch

work page arXiv 2022
[6]

Decoupled Weight Decay Regularization

I. Loshchilov, F. Hutter, Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101 (2017)

work page internal anchor Pith review Pith/arXiv arXiv 2017
[7]

Tan, J.C

A.R. Tan, J.C. Dietschreit, R. Gomez-Bombarelli, Enhanced sampling of robust molecular datasets with uncertainty-based collective variables. arXiv preprint arXiv:2402.03753 (2024)

work page arXiv 2024
[8]

Krajewski, J.W

A.M. Krajewski, J.W. Siegel, J. Xu, Z.K. Liu, Extensible structure-informed prediction of for- mation energy with improved accuracy and usability employing neural networks. Computational Materials Science208, 111254 (2022)

work page 2022
[9]

Thomas-Mitchell, G

A. Thomas-Mitchell, G. Hawe, P.L. Popelier, Calibration of uncertainty in the active learning of machine learning force fields. Machine Learning: Science and Technology4(4), 045034 (2023)

work page 2023
[10]

Thaler, G

S. Thaler, G. Doehner, J. Zavadlav, Scalable bayesian uncertainty quantification for neural net- work potentials: promise and pitfalls. Journal of Chemical Theory and Computation19(14), 4520–4532 (2023)

work page 2023
[11]

Ong, W.D

S.P. Ong, W.D. Richards, A. Jain, G. Hautier, M. Kocher, S. Cholia, D. Gunter, V.L. Chevrier, K.A. Persson, G. Ceder, Python materials genomics (pymatgen): A robust, open-source python library for materials analysis. Computational Materials Science68, 314–319 (2013) 29

work page 2013
[12]

Kresse, J

G. Kresse, J. Furthmüller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Computational materials science6(1), 15–50 (1996)

work page 1996
[13]

Kresse, J

G. Kresse, J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Physical review B54(16), 11169 (1996)

work page 1996
[14]

Blöchl, Projector augmented-wave method

P.E. Blöchl, Projector augmented-wave method. Physical review B50(24), 17953 (1994)

work page 1994
[15]

Perdew, K

J.P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made simple. Physical review letters77(18), 3865 (1996)

work page 1996
[16]

Chen, S.P

C. Chen, S.P. Ong, A universal graph deep learning interatomic potential for the periodic table. Nature Computational Science2(11), 718–728 (2022)

work page 2022
[17]

B. Deng, P. Zhong, K. Jun, J. Riebesell, K. Han, C.J. Bartel, G. Ceder, Chgnet as a pretrained universal neural network potential for charge-informed atomistic modelling. Nature Machine Intelligence5(9), 1031–1041 (2023)

work page 2023
[18]

Schmidt, N

J. Schmidt, N. Hoffmann, H.C. Wang, P. Borlido, P.J. Carriço, T.F. Cerqueira, S. Botti, M.A. Marques, Machine-learning-assisted determination of the global zero-temperature phase diagram of materials. Advanced Materials35(22), 2210788 (2023)

work page 2023
[19]

Barroso-Luque et al., Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models

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

work page arXiv 2024
[20]

Curtarolo, G.L

S. Curtarolo, G.L. Hart, M.B. Nardelli, N. Mingo, S. Sanvito, O. Levy, The high-throughput highway to computational materials design. Nature materials12(3), 191–201 (2013)

work page 2013
[21]

A. Togo. Atsushi togo. URL https://doi.org/10.48505/nims.4197

work page doi:10.48505/nims.4197
[22]

Giannessi, S

F. Giannessi, S. Di Cataldo, S. Saha, L. Boeri, A database of high-pressure crystal structures from hydrogen to lanthanum. Scientific Data11(1), 766 (2024)

work page 2024
[23]

Elena, Dávid P

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

work page arXiv 2023
[24]

Y. Park, J. Kim, S. Hwang, S. Han, Scalable parallel algorithm for graph neural network interatomic potentials in molecular dynamics simulations. Journal of Chemical Theory and 30 Computation (2024)

work page 2024
[25]

Neumann, J

M. Neumann, J. Gin, B. Rhodes, S. Bennett, Z. Li, H. Choubisa, A. Hussey, J. Godwin, Orb: A fast, scalable neural network potential. arXiv preprint arXiv:2410.22570 (2024)

work page arXiv 2024
[26]

A. Togo, L. Chaput, T. Tadano, I. Tanaka, Implementation strategies in phonopy and phono3py. J. Phys. Condens. Matter35(35), 353001 (2023). https://doi.org/10.1088/1361-648X/acd831

work page doi:10.1088/1361-648x/acd831 2023
[27]

First-principles Phonon Calculations with Phonopy and Phono3py

A. Togo, First-principles phonon calculations with phonopy and phono3py. J. Phys. Soc. Jpn. 92(1), 012001 (2023). https://doi.org/10.7566/JPSJ.92.012001

work page doi:10.7566/jpsj.92.012001 2023
[28]

Perdew, A

J.P. Perdew, A. Ruzsinszky, G.I. Csonka, O.A. Vydrov, G.E. Scuseria, L.A. Constantin, X. Zhou, K. Burke, Restoring the density-gradient expansion for exchange in solids and surfaces. Physical review letters100(13), 136406 (2008)

work page 2008
[29]

A. Togo. Private communication (2024)

work page 2024
[30]

A. Togo, L. Chaput, T. Tadano, I. Tanaka, Implementation strategies in phonopy and phono3py. Journal of Physics: Condensed Matter35(35), 353001 (2023)

work page 2023
[31]

Kruglov, D.V

I.A. Kruglov, D.V. Semenok, H. Song, R. Szczęśniak, I.A. Wrona, R. Akashi, M.M. Davari Esfa- hani,D.Duan,T.Cui,A.G.Kvashnin,etal.,Superconductivityoflah10andlah16polyhydrides. Physical Review B101(2), 024508 (2020)

work page 2020
[32]

Drozdov, P

A. Drozdov, P. Kong, V. Minkov, S. Besedin, M. Kuzovnikov, S. Mozaffari, L. Balicas, F. Bal- akirev, D. Graf, V. Prakapenka, et al., Superconductivity at 250 k in lanthanum hydride under high pressures. Nature569(7757), 528–531 (2019)

work page 2019
[33]

Cochran, R

W. Cochran, R. Cowley, Dielectric constants and lattice vibrations. Journal of Physics and Chemistry of Solids23(5), 447–450 (1962)

work page 1962
[34]

Giannozzi, S

P. Giannozzi, S. De Gironcoli, P. Pavone, S. Baroni, Ab initio calculation of phonon dispersions in semiconductors. Physical Review B43(9), 7231 (1991)

work page 1991
[35]

Baroni, S

S. Baroni, S. De Gironcoli, A. Dal Corso, P. Giannozzi, Phonons and related crystal properties from density-functional perturbation theory. Reviews of modern Physics73(2), 515 (2001)

work page 2001
[36]

Karch, F

K. Karch, F. Bechstedt, P. Pavone, D. Strauch, Pressure-dependent dynamical and dielectric properties of cubic sic. Journal of Physics: Condensed Matter8(17), 2945 (1996) 31

work page 1996
[37]

Olego, M

D. Olego, M. Cardona, Pressure dependence of raman phonons of ge and 3 c-sic. Physical Review B25(2), 1151 (1982)

work page 1982
[38]

J. Wang, L. Zhang, Q. Zeng, G.L. Vignoles, L. Cheng, Surface relaxation and oxygen adsorption behavior of different sic polytypes: a first-principles study. Journal of Physics: Condensed Matter 22(26), 265003 (2010)

work page 2010
[39]

Physical Review B89(19), 195112 (2014)

J.H.Skone,M.Govoni,G.Galli,Self-consistenthybridfunctionalforcondensedsystems. Physical Review B89(19), 195112 (2014)

work page 2014
[40]

Madelung, Physics of group iv elements and iii-v compounds

O. Madelung, Physics of group iv elements and iii-v compounds. Landolt-Bornstein, Numerical Data and Functional Relationships in Science and Technology, New Series17(1982)

work page 1982
[41]

Olego, M

D. Olego, M. Cardona, P. Vogl, Pressure dependence of the optical phonons and transverse effective charge in 3 c-sic. Physical Review B25(6), 3878 (1982)

work page 1982
[42]

Yu,Fundamentalsof semiconductors (Springer, 2005)

P. Yu,Fundamentalsof semiconductors (Springer, 2005)

work page 2005
[43]

McColl, S.W

K. McColl, S.W. Coles, P. Zarabadi-Poor, B.J. Morgan, M.S. Islam, Phase segregation and nanoconfined fluid o2 in a lithium-rich oxide cathode. Nature Materials pp. 1–8 (2024)

work page 2024
[44]

McWilliams, D.K

R.S. McWilliams, D.K. Spaulding, J.H. Eggert, P.M. Celliers, D.G. Hicks, R.F. Smith, G.W. Collins, R. Jeanloz, Phase transformations and metallization of magnesium oxide at high pressure and temperature. Science338(6112), 1330–1333 (2012)

work page 2012
[45]

Dubrovinskaia, S

N. Dubrovinskaia, S. Petitgirard, S. Chariton, R. Tucoulou, J. Garrevoet, K. Glazyrin, H.P. Liermann, V.B. Prakapenka, L. Dubrovinsky, B1-b2 phase transition in mgo at ultra-high static pressure. arXiv preprint arXiv:1904.00476 (2019)

work page arXiv 1904
[46]

Physical Review B81(5), 054110 (2010)

A.B.Belonoshko,S.Arapan,R.Martonak,A.Rosengren,Mgophasediagramfromfirstprinciples in a wide pressure-temperature range. Physical Review B81(5), 054110 (2010)

work page 2010
[47]

Bouchet, F

J. Bouchet, F. Bottin, V. Recoules, F. Remus, G. Morard, R. Bolis, A. Benuzzi-Mounaix, Ab initio calculations of the b1-b2 phase transition in mgo. Physical Review B99(9), 094113 (2019)

work page 2019
[48]

Cheng, E.A

B. Cheng, E.A. Engel, J. Behler, C. Dellago, M. Ceriotti, Ab initio thermodynamics of liquid and solid water. Proceedings of the National Academy of Sciences116(4), 1110–1115 (2019)

work page 2019
[49]

Monserrat, J.G

B. Monserrat, J.G. Brandenburg, E.A. Engel, B. Cheng, Liquid water contains the building blocks of diverse ice phases. Nature communications11(1), 5757 (2020) 32

work page 2020
[50]

Parrinello, A

M. Parrinello, A. Rahman, Polymorphic transitions in single crystals: A new molecular dynamics method. Journal of Applied physics52(12), 7182–7190 (1981)

work page 1981
[51]

Nosé, A unified formulation of the constant temperature molecular dynamics methods

S. Nosé, A unified formulation of the constant temperature molecular dynamics methods. The Journal of chemical physics81(1), 511–519 (1984)

work page 1984
[52]

Martyna, D.J

G.J. Martyna, D.J. Tobias, M.L. Klein, Constant pressure molecular dynamics algorithms. The Journal of chemical physics101(5), 4177–4189 (1994)

work page 1994
[53]

Capinski, H

W. Capinski, H. Maris, E. Bauser, I. Silier, M. Asen-Palmer, T. Ruf, M. Cardona, E. Gmelin, Thermalconductivityofisotopicallyenrichedsi. Appliedphysicsletters71(15),2109–2111(1997)

work page 1997
[54]

Shinoda, M

W. Shinoda, M. Shiga, M. Mikami, Rapid estimation of elastic constants by molecular dynamics simulation under constant stress. Physical Review B69(13), 134103 (2004)

work page 2004
[55]

Praiteri

M. Praiteri. GPTA: Github repository for gpt assisted tasks. https://github.com/praiteri/GPTA (2023). Accessed: 2024-03-18

work page 2023
[56]

Soper, C

A. Soper, C. Benmore, Quantum differences between heavy and light water. Physical review letters101(6), 065502 (2008)

work page 2008
[57]

DiStasio, B

R.A. DiStasio, B. Santra, Z. Li, X. Wu, R. Car, The individual and collective effects of exact exchange and dispersion interactions on the ab initio structure of liquid water. The Journal of chemical physics141(8) (2014)

work page 2014
[58]

Gaiduk, J

A.P. Gaiduk, J. Gustafson, F. Gygi, G. Galli, First-principles simulations of liquid water using a dielectric-dependenthybridfunctional. Thejournalofphysicalchemistryletters9(11),3068–3073 (2018)

work page 2018
[59]

Zheng, M

L. Zheng, M. Chen, Z. Sun, H.Y. Ko, B. Santra, P. Dhuvad, X. Wu, Structural, electronic, and dynamical properties of liquid water by ab initio molecular dynamics based on scan functional within the canonical ensemble. The Journal of Chemical Physics148(16) (2018)

work page 2018
[60]

Einstein, Annalen der physik

A. Einstein, Annalen der physik. Nr10, 891ff (1905)

work page 1905
[61]

Maginn, R.A

E.J. Maginn, R.A. Messerly, D.J. Carlson, D.R. Roe, J.R. Elliot, Best practices for computing transport properties 1. self-diffusivity and viscosity from equilibrium molecular dynamics [article v1. 0]. Living Journal of Computational Molecular Science1(1), 6324–6324 (2019) 33

work page 2019
[62]

Chen, H.Y

M. Chen, H.Y. Ko, R.C. Remsing, M.F. Calegari Andrade, B. Santra, Z. Sun, A. Selloni, R. Car, M.L. Klein, J.P. Perdew, et al., Ab initio theory and modeling of water. Proceedings of the National Academy of Sciences114(41), 10846–10851 (2017)

work page 2017
[63]

Mills, Self-diffusion in normal and heavy water in the range 1-45

R. Mills, Self-diffusion in normal and heavy water in the range 1-45. deg. The Journal of Physical Chemistry77(5), 685–688 (1973)

work page 1973
[64]

Krynicki, C.D

K. Krynicki, C.D. Green, D.W. Sawyer, Pressure and temperature dependence of self-diffusion in water. Faraday Discussions of the Chemical Society66, 199–208 (1978)

work page 1978
[65]

Hardy, A

E.H. Hardy, A. Zygar, M.D. Zeidler, M. Holz, F.D. Sacher, Isotope effect on the translational and rotational motion in liquid water and ammonia. The Journal of Chemical Physics114(7), 3174–3181 (2001) 34

work page 2001