arxiv: 2405.04967 · v2 · submitted 2024-05-08 · ❄️ cond-mat.mtrl-sci

Recognition: 2 theorem links

· Lean Theorem

MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures

Han Yang , Chenxi Hu , Yichi Zhou , Xixian Liu , Yu Shi , Jielan Li , Guanzhi Li , Zekun Chen

show 14 more authors

Shuizhou Chen Claudio Zeni Matthew Horton Robert Pinsler Andrew Fowler Daniel Z\"ugner Tian Xie Jake Smith Lixin Sun Qian Wang Lingyu Kong Chang Liu Hongxia Hao Ziheng Lu

Authors on Pith no claims yet

Pith reviewed 2026-05-17 00:06 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci

keywords deep learningmachine learning force fieldatomistic simulationmaterials propertiesGibbs free energythermodynamic propertieshigh-pressurehigh-temperature

0 comments

The pith

MatterSim predicts Gibbs free energies of inorganic solids at near first-principles accuracy across wide temperatures and pressures.

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

The paper presents MatterSim as a deep learning atomistic model trained on large-scale first-principles computations. It establishes that this model can efficiently perform simulations and predict properties such as lattice dynamics, mechanical behaviors, and thermodynamic quantities like Gibbs free energies for inorganic solids. The model achieves near first-principles accuracy and a resolution of 15 meV per atom for temperatures up to 1000 K when compared to experiments. A reader would care if true because it could drastically reduce the computational cost of exploring material behaviors under realistic conditions and enable prediction of phase diagrams. The model supports fine-tuning with additional data for specific needs, cutting data requirements by up to 97 percent.

Core claim

MatterSim acts as a machine learning force field that not only predicts ground-state material structures and energetics but also simulates their behavior under realistic temperatures and pressures. This leads to accurate computation of lattice dynamics, mechanical and thermodynamic properties comparable to first-principles methods. In particular, it predicts Gibbs free energies for a wide range of inorganic solids with near-first-principles accuracy and reaches 15 meV/atom resolution for temperatures up to 1000 K against experimental data. This capability allows for the prediction of experimental phase diagrams at minimal computational cost. Additionally, the model serves as a platform for a

What carries the argument

The deep learning neural network trained actively on extensive first-principles data, serving as a versatile machine learning force field that generalizes across elements, temperatures from 0 to 5000 K, and pressures up to 1000 GPa.

If this is right

Computes lattice dynamics and thermodynamic properties at first-principles accuracy but with much higher efficiency.
Enables prediction of material phase diagrams at low computational cost.
Supports customization through fine-tuning with domain-specific data, reducing required data by up to 97%.
Facilitates simulations of materials under extreme conditions up to 5000 K and 1000 GPa.
Provides a base for continuous learning and integration into materials design workflows.

Where Pith is reading between the lines

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

If correct, this approach could extend to predicting other complex properties not directly trained on, such as electronic or optical behaviors through transfer learning.
High data efficiency suggests potential for rapid adaptation to new levels of theory or specific material classes with minimal additional computations.
Could accelerate the discovery of materials for applications in energy, electronics, or structural uses by allowing broader exploration of the design space.

Load-bearing premise

The first-principles training data covers enough of the chemical space and conditions to allow the model to generalize accurately without large errors on unseen materials or extreme conditions.

What would settle it

Running MatterSim on a new composition or at a temperature or pressure outside the training range and comparing its predicted Gibbs free energy or other properties directly to experimental measurements or high-accuracy first-principles results; large discrepancies beyond 15 meV/atom would falsify the generalization claim.

read the original abstract

Accurate and fast prediction of materials properties is central to the digital transformation of materials design. However, the vast design space and diverse operating conditions pose significant challenges for accurately modeling arbitrary material candidates and forecasting their properties. We present MatterSim, a deep learning model actively learned from large-scale first-principles computations, for efficient atomistic simulations at first-principles level and accurate prediction of broad material properties across the periodic table, spanning temperatures from 0 to 5000 K and pressures up to 1000 GPa. Out-of-the-box, the model serves as a machine learning force field, and shows remarkable capabilities not only in predicting ground-state material structures and energetics, but also in simulating their behavior under realistic temperatures and pressures, signifying an up to ten-fold enhancement in precision compared to the prior best-in-class. This enables MatterSim to compute materials' lattice dynamics, mechanical and thermodynamic properties, and beyond, to an accuracy comparable with first-principles methods. Specifically, MatterSim predicts Gibbs free energies for a wide range of inorganic solids with near-first-principles accuracy and achieves a 15 meV/atom resolution for temperatures up to 1000K compared with experiments. This opens an opportunity to predict experimental phase diagrams of materials at minimal computational cost. Moreover, MatterSim also serves as a platform for continuous learning and customization by integrating domain-specific data. The model can be fine-tuned for atomistic simulations at a desired level of theory or for direct structure-to-property predictions, achieving high data efficiency with a reduction in data requirements by up to 97%.

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 scales a single GNN force field across the periodic table and wide T/P ranges with claimed 15 meV/atom Gibbs accuracy, but the generalization rests on undocumented training coverage.

read the letter

The core advance is a single model trained via active learning on first-principles data that targets ground-state structures, energetics, and finite-temperature thermodynamics for inorganic solids from 0-5000 K and 0-1000 GPa. It reports usable accuracy on lattice dynamics, mechanical properties, and Gibbs free energies at roughly 15 meV/atom versus experiment up to 1000 K, plus faster run times than direct DFT. The active-learning loop and the fine-tuning path that cuts data needs by up to 97% are practical additions for users who want to adapt the model to specific chemistries or levels of theory. Those pieces give the work real engineering value if the numbers hold outside the training distribution. The main weakness is the missing documentation on coverage. The abstract supplies no histograms or counts for how temperatures, pressures, and element combinations were sampled, no explicit train/test splits, and no error bars on the headline metrics. Without those, it is hard to tell whether the experimental agreement comes from dense interpolation on well-sampled regions or from genuine extrapolation. The circularity risk is also present: many comparisons are to the same first-principles data used in training. This paper is aimed at computational materials groups that run high-throughput screening or need quick finite-temperature estimates for phase diagrams. A reader already working with graph neural network potentials will find the scale and the reported benchmarks worth examining. I would send it to peer review. The scope is large enough and the practical claims are sharp enough that referees should see the full methods and validation figures, even if the paper needs clearer evidence on data coverage before publication.

Referee Report

2 major / 2 minor

Summary. MatterSim is a deep learning atomistic model actively learned from large-scale first-principles computations. It functions as a machine-learning force field for efficient simulations across the periodic table, with claimed applicability from 0 to 5000 K and pressures up to 1000 GPa. The model is reported to deliver up to ten-fold precision gains over prior best-in-class approaches for ground-state structures, energetics, lattice dynamics, mechanical properties, and thermodynamic quantities. A central result is the prediction of Gibbs free energies for inorganic solids at near-first-principles accuracy, achieving 15 meV/atom resolution versus experiment for temperatures up to 1000 K, thereby enabling low-cost computation of experimental phase diagrams. The work also positions the model as a platform for continuous learning and fine-tuning with reduced data requirements (up to 97 %).

Significance. If the generalization and accuracy claims are substantiated, MatterSim would represent a meaningful advance in universal machine-learning potentials for materials science. It could accelerate high-throughput exploration of thermodynamic stability and phase behavior under realistic operating conditions without repeated DFT calculations. The active-learning strategy and support for domain-specific fine-tuning are positive features that align with practical needs in the field. Credit is due for targeting a broad temperature-pressure range and for framing the model as extensible rather than a one-off fit.

major comments (2)

[Dataset construction] Dataset construction section: the manuscript provides no quantitative description of the training distribution, such as the number of unique compositions or elements covered, temperature/pressure histograms, or the fraction of data at high-T/high-P regimes. This information is load-bearing for the central generalization claim that the 15 meV/atom Gibbs free energy accuracy extends across unseen inorganic solids up to 1000 K and 1000 GPa.
[Thermodynamic properties results] Thermodynamic properties results: the reported 15 meV/atom resolution for Gibbs free energies versus experiment is given without error bars on the model predictions, without explicit train/test split details, and without confirmation that the experimental comparison set is disjoint from the actively learned first-principles data. These omissions prevent assessment of whether the accuracy reflects true out-of-distribution performance or interpolation on well-sampled subsets.

minor comments (2)

[Abstract] Abstract: the phrase 'up to ten-fold enhancement in precision' should specify the baseline model, the exact metric (e.g., force or energy RMSE), and the conditions under which the factor is measured.
[Figures] Figure captions throughout: parity plots and error distributions would benefit from explicit statements of the number of structures or conditions included and whether any data points were excluded as outliers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and valuable feedback on our manuscript. We address each of the major comments below and have updated the manuscript to incorporate the suggested improvements where appropriate.

read point-by-point responses

Referee: [Dataset construction] Dataset construction section: the manuscript provides no quantitative description of the training distribution, such as the number of unique compositions or elements covered, temperature/pressure histograms, or the fraction of data at high-T/high-P regimes. This information is load-bearing for the central generalization claim that the 15 meV/atom Gibbs free energy accuracy extends across unseen inorganic solids up to 1000 K and 1000 GPa.

Authors: We agree with the referee that a quantitative description of the training distribution is important to support the generalization claims. In the revised manuscript, we have expanded the Dataset construction section to include the number of unique compositions and elements covered, as well as histograms showing the distribution of temperatures and pressures in the training data. We have also specified the fraction of data in the high-temperature (above 1000 K) and high-pressure (above 100 GPa) regimes. These additions provide the necessary context for evaluating the model's performance across the claimed range. revision: yes
Referee: [Thermodynamic properties results] Thermodynamic properties results: the reported 15 meV/atom resolution for Gibbs free energies versus experiment is given without error bars on the model predictions, without explicit train/test split details, and without confirmation that the experimental comparison set is disjoint from the actively learned first-principles data. These omissions prevent assessment of whether the accuracy reflects true out-of-distribution performance or interpolation on well-sampled subsets.

Authors: We appreciate this comment and have revised the Thermodynamic properties results section accordingly. We have added error bars to the reported Gibbs free energy predictions to indicate the uncertainty in the model outputs. Additionally, we have provided explicit details on the train/test splits used in the active learning process and confirmed that the set of materials used for experimental comparison is disjoint from the first-principles data employed in training the model. This clarification demonstrates that the 15 meV/atom accuracy is achieved on out-of-distribution examples, supporting the generalization capability. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on held-out tests and external experimental benchmarks

full rationale

The paper trains MatterSim on first-principles data via active learning and reports accuracy on Gibbs free energies versus independent experiments (15 meV/atom up to 1000 K). This comparison uses external data outside the training distribution. No equations or steps reduce by construction to the inputs (e.g., no fitted parameters renamed as predictions, no self-definitional loops, no load-bearing self-citations). The derivation chain is self-contained against external benchmarks, consistent with standard ML force-field validation.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The model rests on the assumption that a graph neural network trained on DFT data can extrapolate to unseen compositions and thermodynamic conditions; the ledger therefore records the neural-network weights as free parameters and the representativeness of the DFT training distribution as the key domain assumption.

free parameters (1)

neural network weights
All parameters of the deep learning model are fitted to the first-principles dataset; their number is large and not enumerated in the abstract.

axioms (1)

domain assumption DFT calculations provide sufficiently accurate reference data for the target properties across the claimed range of elements, temperatures, and pressures.
The entire training pipeline and accuracy claims rest on this premise; it is invoked implicitly throughout the abstract.

pith-pipeline@v0.9.0 · 5662 in / 1485 out tokens · 30171 ms · 2026-05-17T00:06:16.986880+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

MatterSim employs an active learning approach... deep graph neural networks, uncertainty-aware sampling... first-principles supervisor... M3GNet and Graphormer backbones
IndisputableMonolith/Foundation/BlackBodyRadiationDeep.lean blackBodyRadiationDeepCert unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

predicts Gibbs free energies... 15 meV/atom resolution for temperatures up to 1000 K

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 20 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Lang2MLIP: End-to-End Language-to-Machine Learning Interatomic Potential Development with Autonomous Agentic Workflows
cs.LG 2026-05 unverdicted novelty 7.0

Lang2MLIP is an LLM multi-agent framework that automates end-to-end development of machine learning interatomic potentials from natural language input for heterogeneous materials systems.
Micro-environment of the Eu interstitial in $\beta$-SiAlON:Eu$^{2+}$ green phosphor
cond-mat.mtrl-sci 2026-05 conditional novelty 7.0

First-principles ΔSCF and Huang-Rhys calculations validate the Eu-N₉ coordination model in β-SiAlON:Eu²⁺ at low z and explain the red-shift of emission via zero-phonon line trends, modest Huang-Rhys increases, and con...
Inverse Materials Design via Joint Generation of Crystal Structures and Local Electronic Descriptors
cond-mat.mtrl-sci 2026-05 unverdicted novelty 7.0

A joint diffusion framework for crystal structures and local electronic descriptors improves inverse materials design success rates and structural quality over structure-only models under band-gap and formation-energy...
Fast and Accurate Prediction of Lattice Thermal Conductivity via Machine Learning Surrogates
cond-mat.mtrl-sci 2026-05 conditional novelty 6.0

Machine learning models, especially certain deep neural networks, can predict lattice thermal conductivity with useful accuracy across different generalization tests while being orders of magnitude faster than first-p...
Intervention-Based Time Series Causal Discovery via Simulator-Generated Interventional Distributions
cs.LG 2026-05 unverdicted novelty 6.0

SVAR-FM uses simulator clamping to produce interventional distributions and flow matching to identify time series causal structures, with an error bound that predicts sign reversal of causal effects below a simulator ...
CrystalREPA: Transferring Physical Priors from Universal MLIPs to Crystal Generative Models
cond-mat.mtrl-sci 2026-05 unverdicted novelty 6.0

CrystalREPA closes the representation gap between crystal generators and universal MLIPs via contrastive alignment, yielding more stable and valid generated crystals while revealing that MLIP teacher quality is better...
Compact SO(3) Equivariant Atomistic Foundation Models via Structural Pruning
cs.LG 2026-05 unverdicted novelty 6.0

Structural pruning of SO(3) equivariant atomistic models from large checkpoints yields 1.5-4x fewer parameters and 2.5-4x less pre-training compute than small models trained from scratch, while outperforming them on m...
MatterSim-MT: A multi-task foundation model for in silico materials characterization
cond-mat.mtrl-sci 2026-05 unverdicted novelty 6.0

MatterSim-MT is a foundation model pretrained on over 35 million first-principles structures that predicts material structure, dynamics, and thermodynamics while enabling multi-task simulations of phonon splitting, fe...
Generative structure search for efficient and diverse discovery of molecular and crystal structures
cs.AI 2026-04 unverdicted novelty 6.0

GSS unifies diffusion generation and random structure search into a single sampling process using learned scores and physical forces, recovering diverse metastable structures at over tenfold lower cost than pure RSS w...
Agentic Fusion of Large Atomic and Language Models to Accelerate Superconductor Discovery
cs.LG 2026-04 unverdicted novelty 6.0

An agentic framework fusing large atomic and language models rediscovers 66 known superconductors and guides experimental verification of four new ones with transition temperatures from 2.5 K to 6.5 K.
Iterative learning scheme for crystal structure prediction with anharmonic lattice dynamics
cond-mat.mtrl-sci 2025-12 unverdicted novelty 6.0

An iterative scheme using foundation models and SSCHA enables efficient crystal structure prediction with anharmonic effects, shown to match DFT benchmarks on the H3S system from 50 to 200 GPa.
An experimentally validated end-to-end framework for operando modeling of intrinsically complex metallosilicates
cond-mat.mtrl-sci 2025-12 conditional novelty 6.0

An end-to-end framework combining domain separation, lightweight ML potentials, and de novo in silico synthesis enables quantitative atomistic modeling of mesoporous metallosilicates that matches experimental densitie...
Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models
cond-mat.mtrl-sci 2024-10 accept novelty 6.0

OMat24 provides over 110 million DFT calculations and EquiformerV2 models that reach state-of-the-art performance on material stability and formation energy prediction.
GEWUM: General Exploration Workflow for the Utopia of Materials: A Unified Platform for Automated Structure Generation, Selection, and Validation
cond-mat.mtrl-sci 2026-04 unverdicted novelty 5.0

GEWUM is a unified open-source platform that combines selective random structure search with universal machine learning interatomic potentials to automate materials structure prediction, selection, and validation.
Finetuning-Free Diffusion Model with Adaptive Constraint Guidance for Inorganic Crystal Structure Generation
cond-mat.mtrl-sci 2026-04 unverdicted novelty 5.0

A finetuning-free diffusion model uses adaptive constraint guidance to generate thermodynamically plausible inorganic crystal structures that meet targeted geometric and chemical constraints.
Comparing the latent features of universal machine-learning interatomic potentials
physics.chem-ph 2025-12 unverdicted novelty 5.0

Different uMLIPs encode chemical space in distinct ways, with high cross-model feature reconstruction errors, and fine-tuning preserves strong pre-training bias in the latent features.
Assessing foundational atomistic models for iron alloys under Earth's core conditions
physics.geo-ph 2026-05 unverdicted novelty 4.0

Foundational atomistic models reproduce some structural and dynamical properties of iron alloys under core conditions but none consistently match first-principles benchmarks due to missing explicit treatment of therma...
Quasiparticle Dynamics in the 4d-4f Ising-like Double Perovskite Ba2DyRuO6 studied using Neutron Scattering and Machine-Learning Framework
cond-mat.str-el 2025-12 unverdicted novelty 4.0

Ba2DyRuO6 undergoes simultaneous Ru5+ and Dy3+ antiferromagnetic ordering at 47 K into a collinear Ising state with magnons below 10 meV and CEF levels at 46.5 and 71.8 meV.
Learning Structure, Energy, and Dynamics: A Survey of Artificial Intelligence for Protein Dynamics
q-bio.BM 2026-04 unverdicted novelty 2.0

A review summarizing AI techniques for protein conformation generation, trajectory modeling, Boltzmann generators, machine learning potentials, and related challenges in scalability and physical consistency.
Inverse Design of Inorganic Compounds with Generative AI
physics.chem-ph 2026-04 unverdicted novelty 2.0

A review of generative AI for inverse design of inorganic compounds, analyzing adaptations for their complexity in composition, geometry, symmetry, and electronic structure, with discussion of future benchmarks and sy...

Reference graph

Works this paper leans on

158 extracted references · 158 canonical work pages · cited by 20 Pith papers · 2 internal anchors

[1]

Fiori, F

G. Fiori, F. Bonaccorso, G. Iannaccone, T. Palacios, D. Neumaier, A. Seabaugh, S.K. Banerjee, L. Colombo, Electronics based on two-dimensional materials. Nature nanotechnology 9(10), 768–779 (2014)

work page 2014
[2]

T. Li, G. Galli, Electronic properties of mos2 nanoparticles. The Journal of Physical Chemistry C 111(44), 16192–16196 (2007)

work page 2007
[3]

Mizushima, P

K. Mizushima, P. Jones, P. Wiseman, J.B. Goodenough, Lixcoo2 (0¡ x¡-1): A new cathode material for batteries of high energy density. Materials Research Bulletin 15(6), 783–789 (1980)

work page 1980
[4]

Ceder, Y.M

G. Ceder, Y.M. Chiang, D. Sadoway, M. Aydinol, Y.I. Jang, B. Huang, Identification of cathode materials for lithium batteries guided by first-principles calculations. Nature 392(6677), 694– 696 (1998)

work page 1998
[5]

Tibbitt, C.B

M.W. Tibbitt, C.B. Rodell, J.A. Burdick, K.S. Anseth, Progress in material design for biomed- ical applications. Proceedings of the National Academy of Sciences 112(47), 14444–14451 (2015)

work page 2015
[6]

X. Li, J. Xie, C. Jiang, J. Yu, P. Zhang, Review on design and evaluation of environmental photocatalysts. Frontiers of Environmental Science & Engineering 12, 1–32 (2018)

work page 2018
[7]

X. Hu, G. Li, J.C. Yu, Design, fabrication, and modification of nanostructured semiconductor materials for environmental and energy applications. Langmuir 26(5), 3031–3039 (2010)

work page 2010
[8]

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 materials 12(3), 191–201 (2013)

work page 2013
[9]

Choudhary, B

K. Choudhary, B. DeCost, C. Chen, A. Jain, F. Tavazza, R. Cohn, C.W. Park, A. Choudhary, A. Agrawal, S.J. Billinge, et al., Recent advances and applications of deep learning methods in materials science. npj Computational Materials 8(1), 59 (2022)

work page 2022
[10]

Xie, J.C

T. Xie, J.C. Grossman, Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties. Physical review letters 120(14), 145301 (2018)

work page 2018
[11]

Merchant, S

A. Merchant, S. Batzner, S.S. Schoenholz, M. Aykol, G. Cheon, E.D. Cubuk, Scaling deep learning for materials discovery. Nature pp. 1–6 (2023) 72

work page 2023
[12]

Chen, D.T

C. Chen, D.T. Nguyen, S.J. Lee, N.A. Baker, A.S. Karakoti, L. Lauw, C. Owen, K.T. Mueller, B.A. Bilodeau, V. Murugesan, et al., Accelerating computational materials discovery with artificial intelligence and cloud high-performance computing: from large-scale screening to experimental validation. arXiv preprint arXiv:2401.04070 (2024)

work page arXiv 2024
[13]

Lindsey, L.E

R.K. Lindsey, L.E. Fried, N. Goldman, Chimes: A force matched potential with explicit three- body interactions for molten carbon. Journal of chemical theory and computation 13(12), 6222–6229 (2017)

work page 2017
[14]

Sch¨ utt, P.J

K. Sch¨ utt, P.J. Kindermans, H.E. Sauceda Felix, S. Chmiela, A. Tkatchenko, K.R. M¨ uller, Schnet: A continuous-filter convolutional neural network for modeling quantum interactions. Advances in neural information processing systems 30 (2017)

work page 2017
[15]

Musaelian, S

A. Musaelian, S. Batzner, A. Johansson, L. Sun, C.J. Owen, M. Kornbluth, B. Kozin- sky, Learning local equivariant representations for large-scale atomistic dynamics. Nature Communications 14(1), 579 (2023)

work page 2023
[16]

Batzner, A

S. Batzner, A. Musaelian, L. Sun, M. Geiger, J.P. Mailoa, M. Kornbluth, N. Molinari, T.E. Smidt, B. Kozinsky, E (3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. Nature communications 13(1), 2453 (2022)

work page 2022
[17]

C. Chen, W. Ye, Y. Zuo, C. Zheng, S.P. Ong, Graph networks as a universal machine learning framework for molecules and crystals. Chemistry of Materials 31(9), 3564–3572 (2019)

work page 2019
[18]

Choudhary, B

K. Choudhary, B. DeCost, Atomistic line graph neural network for improved materials property predictions. npj Computational Materials 7(1), 185 (2021)

work page 2021
[19]

Chen, S.P

C. Chen, S.P. Ong, A universal graph deep learning interatomic potential for the periodic table. Nature Computational Science 2(11), 718–728 (2022)

work page 2022
[20]

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 Intelligence 5(9), 1031–1041 (2023)

work page 2023
[21]

A foundation model for atomistic materials chemistry

I. Batatia, P. Benner, Y. Chiang, A.M. Elena, D.P. Kov´ acs, 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) 73

work page arXiv 2023
[22]

Zhang, X

D. Zhang, X. Liu, X. Zhang, C. Zhang, C. Cai, H. Bi, Y. Du, X. Qin, J. Huang, B. Li, et al., Dpa-2: Towards a universal large atomic model for molecular and material simulation. arXiv preprint arXiv:2312.15492 (2023)

work page arXiv 2023
[23]

Shoghi, A

N. Shoghi, A. Kolluru, J.R. Kitchin, Z.W. Ulissi, C.L. Zitnick, B.M. Wood, From molecules to materials: Pre-training large generalizable models for atomic property prediction. arXiv preprint arXiv:2310.16802 (2023)

work page arXiv 2023
[24]

Kresse, J

G. Kresse, J. Furthm¨ uller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Physical review B 54(16), 11169 (1996)

work page 1996
[25]

A. Dunn, Q. Wang, A. Ganose, D. Dopp, A. Jain, Benchmarking materials property prediction methods: the matbench test set and automatminer reference algorithm. npj Computational Materials 6(1), 138 (2020)

work page 2020
[26]

Kresse, J

G. Kresse, J. Furthm¨ uller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Computational materials science 6(1), 15–50 (1996)

work page 1996
[27]

Kohn, L.J

W. Kohn, L.J. Sham, Self-consistent equations including exchange and correlation effects. Physical review 140(4A), A1133 (1965)

work page 1965
[28]

Hohenberg, W

P. Hohenberg, W. Kohn, Inhomogeneous electron gas. Physical review 136(3B), B864 (1964)

work page 1964
[29]

Perdew, K

J.P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made simple. Physical review letters 77(18), 3865 (1996)

work page 1996
[30]

Anisimov, J

V.I. Anisimov, J. Zaanen, O.K. Andersen, Band theory and mott insulators: Hubbard u instead of stoner i. Physical Review B 44(3), 943 (1991)

work page 1991
[31]

Jain, S.P

A. Jain, S.P. Ong, G. Hautier, W. Chen, W.D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, et al., Commentary: The materials project: A materials genome approach to accelerating materials innovation. APL materials 1(1) (2013)

work page 2013
[32]

J.E. Saal, S. Kirklin, M. Aykol, B. Meredig, C. Wolverton, Materials design and discovery with high-throughput density functional theory: the open quantum materials database (oqmd). Jom 65, 1501–1509 (2013)

work page 2013
[33]

Kirklin, J.E

S. Kirklin, J.E. Saal, B. Meredig, A. Thompson, J.W. Doak, M. Aykol, S. R¨ uhl, C. Wolverton, The open quantum materials database (oqmd): assessing the accuracy of dft formation energies. 74 npj Computational Materials 1(1), 1–15 (2015)

work page 2015
[34]

Schmidt, N

J. Schmidt, N. Hoffmann, H.C. Wang, P. Borlido, P.J. Carri¸ co, T.F. Cerqueira, S. Botti, M.A. Marques, Machine-learning-assisted determination of the global zero-temperature phase diagram of materials. Advanced Materials 35(22), 2210788 (2023)

work page 2023
[35]

C. Ying, T. Cai, S. Luo, S. Zheng, G. Ke, D. He, Y. Shen, T.Y. Liu, Do transformers really perform badly for graph representation? Advances in neural information processing systems 34, 28877–28888 (2021)

work page 2021
[36]

Y. Shi, S. Zheng, G. Ke, Y. Shen, J. You, J. He, S. Luo, C. Liu, D. He, T.Y. Liu, Benchmarking graphormer on large-scale molecular modeling datasets. arXiv preprint arXiv:2203.04810 (2022)

work page arXiv 2022
[37]

Riebesell, H

J. Riebesell, H. Yang, R. Goodall, S.G. Baird. Pymatviz: visualization toolkit for materials informatics (2022). https://doi.org/10.5281/zenodo.7486816. URL https://github.com/janosh/ pymatviz. 10.5281/zenodo.7486816 - https://github.com/janosh/pymatviz

work page doi:10.5281/zenodo.7486816 2022
[38]

M. Horton. Add strict anions option to MaterialsProject2020Compatibility by mkhorton (2024). Accessed May 07, 2024

work page 2024
[39]

C. Zeni, R. Pinsler, D. Z¨ ugner, A. Fowler, M. Horton, X. Fu, S. Shysheya, J. Crabb´ e, L. Sun, J. Smith, et al., Mattergen: a generative model for inorganic materials design. arXiv preprint arXiv:2312.03687 (2023)

work page arXiv 2023
[40]

T. Xie, X. Fu, O.E. Ganea, R. Barzilay, T. Jaakkola, Crystal diffusion variational autoencoder for periodic material generation. arXiv preprint arXiv:2110.06197 (2021)

work page arXiv 2021
[41]

Pickard, R

C.J. Pickard, R. Needs, Ab initio random structure searching. Journal of Physics: Condensed Matter 23(5), 053201 (2011)

work page 2011
[42]

Schmidt, H.C

J. Schmidt, H.C. Wang, T.F. Cerqueira, S. Botti, M.A. Marques, A dataset of 175k stable and metastable materials calculated with the pbesol and scan functionals. Scientific Data 9(1), 64 (2022)

work page 2022
[43]

Schmidt, N

J. Schmidt, N. Hoffmann, H.C. Wang, P. Borlido, P.J. Carri¸ co, T.F. Cerqueira, S. Botti, M.A. Marques, Large-scale machine-learning-assisted exploration of the whole materials space. arXiv preprint arXiv:2210.00579 (2022)

work page arXiv 2022
[44]

Bergerhoff, I

G. Bergerhoff, I. Brown, F. Allen, et al., Crystallographic databases. International Union of Crystallography, Chester 360, 77–95 (1987) 75

work page 1987
[45]

Bloch, ¨Uber die quantenmechanik der elektronen in kristallgittern

F. Bloch, ¨Uber die quantenmechanik der elektronen in kristallgittern. Zeitschrift f¨ ur physik 52(7), 555–600 (1929)

work page 1929
[46]

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 Physics 73(2), 515 (2001)

work page 2001
[47]

Baroni, P

S. Baroni, P. Giannozzi, A. Testa, Green’s-function approach to linear response in solids. Physical review letters 58(18), 1861 (1987)

work page 1987
[48]

Giannozzi, S

P. Giannozzi, S. De Gironcoli, P. Pavone, S. Baroni, Ab initio calculation of phonon dispersions in semiconductors. Physical Review B 43(9), 7231 (1991)

work page 1991
[49]

Kresse, J

G. Kresse, J. Furthm¨ uller, J. Hafner, Ab initio force constant approach to phonon dispersion relations of diamond and graphite. Europhysics Letters 32(9), 729 (1995)

work page 1995
[50]

H. Yang, M. Govoni, A. Kundu, G. Galli, Combined first-principles calculations of electron– electron and electron–phonon self-energies in condensed systems. Journal of Chemical Theory and Computation 17(12), 7468–7476 (2021)

work page 2021
[51]

H. Yang, M. Govoni, A. Kundu, G. Galli, Computational protocol to evaluate electron–phonon interactions within density matrix perturbation theory. Journal of Chemical Theory and Computation 18(10), 6031–6042 (2022)

work page 2022
[52]

S. Fang, M. Geiger, J.G. Checkelsky, T. Smidt. Phonon predictions with e(3)-equivariant graph neural networks (2024)

work page 2024
[53]

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

work page doi:10.48505/nims.4197
[54]

Tolborg, J

K. Tolborg, J. Klarbring, A.M. Ganose, A. Walsh, Free energy predictions for crystal stability and synthesisability. Digital Discovery 1(5), 586–595 (2022)

work page 2022
[55]

Bartel, S.L

C.J. Bartel, S.L. Millican, A.M. Deml, J.R. Rumptz, W. Tumas, A.W. Weimer, S. Lany, V. Ste- vanovi´ c, C.B. Musgrave, A.M. Holder, Physical descriptor for the gibbs energy of inorganic crystalline solids and temperature-dependent materials chemistry. Nature communications9(1), 4168 (2018)

work page 2018
[56]

B1-B2 phase transition in MgO at ultra-high static pressure

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) 76

work page internal anchor Pith review Pith/arXiv arXiv 1904
[57]

Zhang, R

S. Zhang, R. Paul, S. Hu, M.A. Morales, Toward an accurate equation of state and b1-b2 phase boundary for magnesium oxide up to terapascal pressures and electron-volt temperatures. Physical Review B 107(22), 224109 (2023)

work page 2023
[58]

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. Science 338(6112), 1330–1333 (2012)

work page 2012
[59]

O.T. Unke, S. Chmiela, H.E. Sauceda, M. Gastegger, I. Poltavsky, K.T. Sch¨ utt, A. Tkatchenko, K.R. M¨ uller, Machine learning force fields. Chemical Reviews121(16), 10142–10186 (2021)

work page 2021
[60]

X. Fu, Z. Wu, W. Wang, T. Xie, S. Keten, R. Gomez-Bombarelli, T. Jaakkola, Forces are not enough: Benchmark and critical evaluation for machine learning force fields with molecular simulations. arXiv preprint arXiv:2210.07237 (2022)

work page arXiv 2022
[61]

Deringer, M.A

V.L. Deringer, M.A. Caro, G. Cs´ anyi, A general-purpose machine-learning force field for bulk and nanostructured phosphorus. Nature communications 11(1), 5461 (2020)

work page 2020
[62]

Zhou, S.R

Y. Zhou, S.R. Elliott, V.L. Deringer, Structure and bonding in amorphous red phosphorus. Angewandte Chemie International Edition 62(24), e202216658 (2023)

work page 2023
[63]

Skinner, C

L.B. Skinner, C. Benmore, J.C. Neuefeind, J.B. Parise, The structure of water around the compressibility minimum. The Journal of chemical physics 141(21) (2014)

work page 2014
[64]

W. Chen, F. Ambrosio, G. Miceli, A. Pasquarello, Ab initio electronic structure of liquid water. Physical review letters 117(18), 186401 (2016)

work page 2016
[65]

Soper, C

A. Soper, C. Benmore, Quantum differences between heavy and light water. Physical review letters 101(6), 065502 (2008)

work page 2008
[66]

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 physics 141(8) (2014)

work page 2014
[67]

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 Sciences 116(4), 1110–1115 (2019)

work page 2019
[68]

Monserrat, J.G

B. Monserrat, J.G. Brandenburg, E.A. Engel, B. Cheng, Liquid water contains the building blocks of diverse ice phases. Nature communications 11(1), 5757 (2020) 77

work page 2020
[69]

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 Sciences 114(41), 10846–10851 (2017)

work page 2017
[70]

R. Ruff, P. Reiser, J. St¨ uhmer, P. Friederich, Connectivity optimized nested graph networks for crystal structures. arXiv preprint arXiv:2302.14102 (2023)

work page arXiv 2023
[71]

De Breuck, M.L

P.P. De Breuck, M.L. Evans, G.M. Rignanese, Robust model benchmarking and bias-imbalance in data-driven materials science: a case study on modnet. Journal of Physics: Condensed Matter 33(40), 404002 (2021)

work page 2021
[72]

Chmiela, V

S. Chmiela, V. Vassilev-Galindo, O.T. Unke, A. Kabylda, H.E. Sauceda, A. Tkatchenko, K.R. M¨ uller, Accurate global machine learning force fields for molecules with hundreds of atoms. Science Advances 9(2), eadf0873 (2023)

work page 2023
[73]

T.W. Ko, J.A. Finkler, S. Goedecker, J. Behler, Accurate fourth-generation machine learning potentials by electrostatic embedding. Journal of Chemical Theory and Computation 19(12), 3567–3579 (2023)

work page 2023
[74]

PyTorch 2 : Faster machine learning through dynamic P ython bytecode transformation and graph compilation

J. Ansel, E. Yang, H. He, N. Gimelshein, A. Jain, M. Voznesensky, B. Bao, P. Bell, D. Berard, E. Burovski, G. Chauhan, A. Chourdia, W. Constable, A. Desmaison, Z. DeVito, E. Elli- son, W. Feng, J. Gong, M. Gschwind, B. Hirsh, S. Huang, K. Kalambarkar, L. Kirsch, M. Lazos, M. Lezcano, Y. Liang, J. Liang, Y. Lu, C. Luk, B. Maher, Y. Pan, C. Puhrsch, M. Reso...

work page doi:10.1145/3620665.3640366 2024
[75]

T.W. Ko, M. Nassar, S. Miret, E. Liu, J. Qi, S.P. Ong. Materials Graph Library (2021). https://doi.org/10.5281/zenodo.8025189

work page doi:10.5281/zenodo.8025189 2021
[76]

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
[77]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A.N. Gomez, L. Kaiser, I. Polosukhin, Attention is all you need. Advances in neural information processing systems 30 (2017) 78

work page 2017
[78]

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
[79]

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
[80]

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 Science 208, 111254 (2022)

work page 2022

Showing first 80 references.