pith. machine review for the scientific record. sign in

arxiv: 2605.05733 · v1 · submitted 2026-05-07 · ⚛️ physics.chem-ph · cond-mat.stat-mech

Recognition: unknown

Density diversity in training data governs thermodynamic transferability of machine learning interatomic potentials

Je-Yeon Jung, Minwoo Kim, Min Young Ha, Seungtae Kim, Won Bo Lee

Authors on Pith no claims yet

Pith reviewed 2026-05-08 04:16 UTC · model grok-4.3

classification ⚛️ physics.chem-ph cond-mat.stat-mech
keywords densitymlipsdiversitythermodynamictrainingacrossbehaviorcomputational
0
0 comments X

The pith

Diversifying density in training data produces machine learning interatomic potentials that transfer across thermodynamic states.

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

The paper establishes that including varied densities in training data for machine learning interatomic potentials yields better transferability across temperatures and phases than varying temperature alone. Foundation models trained on solids describe liquid densities accurately but fail in gases, while molecular models show the reverse pattern. Experiments with controlled training and distillation show density-diverse sets fix both gaps, because density alters local atomic coordination more than temperature does and therefore supplies greater structural variety. This approach allows reliable simulations of fluid processes under changing conditions without increasing the overall data budget.

Core claim

Diversifying the density of training configurations, rather than temperature, is the most effective strategy for building thermodynamically transferable MLIPs within a fixed computational budget. Foundation MLIPs trained on solid-state databases accurately describe liquid-like densities but fail at gas-like conditions, while molecular-database-trained models exhibit the opposite behavior. Controlled from-scratch training and distillation experiments confirm that density-diverse datasets resolve both failure modes, whereas temperature-diverse datasets cannot compensate for missing density regimes. Coordination number analysis reveals the physical origin of this behavior: local coordination is

What carries the argument

Density diversity in the training dataset, which increases the range of local atomic coordination environments more effectively than temperature diversity.

Where Pith is reading between the lines

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

  • The same density-first sampling principle may improve transferability in ML potentials trained for other state variables that strongly affect local structure, such as composition.
  • Models for inhomogeneous systems like interfaces could be built by deliberately sampling across density gradients rather than uniform temperature sweeps.
  • The validation framework based on density coverage offers a practical checklist for assessing existing foundation models before deployment in variable-pressure fluid simulations.

Load-bearing premise

The coordination number analysis and observed failure modes in the tested foundation and from-scratch models apply to other MLIP architectures and chemical systems.

What would settle it

Repeat the from-scratch training and distillation experiments on a different fluid system such as water and test whether density diversity still outperforms temperature diversity when evaluating transfer errors at densities outside the training range.

Figures

Figures reproduced from arXiv: 2605.05733 by Je-Yeon Jung, Minwoo Kim, Min Young Ha, Seungtae Kim, Won Bo Lee.

Figure 1
Figure 1. Figure 1: Pairwise interaction energy profiles for molecular dimers. (a) Ar – Ar, (b) view at source ↗
Figure 2
Figure 2. Figure 2: Radial distribution function (RDF) of CO view at source ↗
Figure 3
Figure 3. Figure 3: Density-dependent performance of MLIPs for CO view at source ↗
Figure 4
Figure 4. Figure 4: Effect of training-data diversity on the thermodynamic coverage of from-scratch view at source ↗
Figure 5
Figure 5. Figure 5: Thermodynamic coverage map of distilled MLIP with three different training view at source ↗
Figure 6
Figure 6. Figure 6: Physical origin of density-governed transferability. (a) Reference coordination view at source ↗
read the original abstract

Machine learning interatomic potentials (MLIPs) offer first-principles accuracy with reduced computational cost, but their transferability across different thermodynamic states remains questionable, particularly for fluid systems where molecules experience local environments far from crystalline equilibrium. Here, we demonstrate that diversifying the density of training configurations, rather than temperature, is the most effective strategy for building thermodynamically transferable MLIPs within a fixed computational budget. We first show that foundation MLIPs trained on solid-state databases accurately describe liquid-like densities but fail at gas-like conditions, while molecular-database-trained models exhibit the opposite behavior. Controlled from-scratch training and distillation experiments confirm that density-diverse datasets resolve both failure modes, whereas temperature-diverse datasets cannot compensate for missing density regimes. Coordination number analysis reveals the physical origin of this behavior: local coordination topology is more susceptible to density than temperature, leading to further structural diversity. These results establish density diversity as a design principle for thermodynamically transferable MLIPs and provide a validation framework for assessing the thermodynamic coverage of both foundation and from-scratch models, enabling reliable atomistic simulation of fluid-phase processes across diverse operating conditions.

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

3 major / 2 minor

Summary. The paper claims that for machine learning interatomic potentials (MLIPs) applied to fluid systems, diversifying the density of training configurations is more effective than diversifying temperature for achieving thermodynamic transferability within a fixed computational budget. This is supported by showing that foundation models trained on solid-state databases succeed at liquid-like densities but fail at gas-like conditions (and vice versa for molecular-database models), with controlled from-scratch training and distillation experiments demonstrating that density-diverse datasets resolve both failure modes while temperature-diverse ones do not. Coordination-number analysis is presented as the physical explanation, since local coordination topology varies more strongly with density than with temperature.

Significance. If the central result holds, the work supplies a concrete, testable design rule for constructing thermodynamically transferable MLIPs and a validation framework based on density coverage. This is practically useful for fluid-phase simulations in chemistry and materials science, where operating conditions often span wide density ranges. The controlled experiments and coordination-number interpretation add interpretability and falsifiability that many MLIP transferability studies lack.

major comments (3)
  1. [Results and Discussion sections] The generalization claim that density diversity is the governing factor (abstract and final paragraph) rests on experiments performed only for the specific foundation models (solid-state vs. molecular databases) and from-scratch/distillation setups described. No additional architectures (different message-passing depths, cutoff schemes, or equivariant layers) or chemically dissimilar systems are tested, so it remains unclear whether coordination number remains the dominant transferable descriptor outside the examined cases.
  2. [Experimental results on foundation models and controlled trainings] Quantitative support for the failure modes and their resolution is insufficiently detailed. The abstract and results describe qualitative success/failure but do not report dataset sizes, model architectures, error bars, or specific error metrics (e.g., energy/force RMSE or structural deviation thresholds) that would allow assessment of effect size and statistical significance.
  3. [Coordination number analysis subsection] The coordination-number analysis is offered as the physical origin, yet the manuscript does not compare it against other structural descriptors (e.g., radial distribution functions, angular distributions, or ring statistics) to establish that density-induced changes in coordination are the primary driver rather than one of several correlated factors.
minor comments (2)
  1. [Figures] Figure captions and axis labels should explicitly state the number of configurations, temperature/density ranges, and error metrics used in each panel to improve reproducibility.
  2. [Abstract and Conclusions] The term 'parameter-free' or similar phrasing for the design principle should be avoided or qualified, since the conclusion is drawn from empirical comparisons rather than a derivation independent of the chosen models and systems.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback and for highlighting the practical value of our results. Below we respond point by point to the major comments, providing clarifications from the manuscript and indicating the revisions we will implement.

read point-by-point responses

  1. Referee: [Results and Discussion sections] The generalization claim that density diversity is the governing factor (abstract and final paragraph) rests on experiments performed only for the specific foundation models (solid-state vs. molecular databases) and from-scratch/distillation setups described. No additional architectures (different message-passing depths, cutoff schemes, or equivariant layers) or chemically dissimilar systems are tested, so it remains unclear whether coordination number remains the dominant transferable descriptor outside the examined cases.

    Authors: We agree that the experiments are confined to the models and chemical systems described and do not demonstrate universality. The controlled from-scratch and distillation protocols were deliberately chosen to isolate density versus temperature effects while holding other variables fixed. The coordination-number analysis supplies a physically interpretable mechanism, but we do not claim it is the sole descriptor in all cases. In the revised manuscript we will moderate the language in the abstract and concluding paragraph, add an explicit Limitations subsection that states the current scope, and outline the need for validation on additional architectures and dissimilar chemistries. revision: yes

  2. Referee: [Experimental results on foundation models and controlled trainings] Quantitative support for the failure modes and their resolution is insufficiently detailed. The abstract and results describe qualitative success/failure but do not report dataset sizes, model architectures, error bars, or specific error metrics (e.g., energy/force RMSE or structural deviation thresholds) that would allow assessment of effect size and statistical significance.

    Authors: Dataset sizes, model architectures, training protocols, and error metrics (including energy/force RMSE with standard deviations from repeated runs) are reported in the Methods section and Supplementary Information. To improve accessibility we will extract the key quantitative values and statistical details into the main Results text, add error bars to the relevant figures, and include explicit numerical thresholds for the success/failure criteria used in the transferability tests. revision: yes

  3. Referee: [Coordination number analysis subsection] The coordination-number analysis is offered as the physical origin, yet the manuscript does not compare it against other structural descriptors (e.g., radial distribution functions, angular distributions, or ring statistics) to establish that density-induced changes in coordination are the primary driver rather than one of several correlated factors.

    Authors: Coordination number was highlighted because it exhibits the largest variation across the density range examined and correlates directly with the observed transferability failures. We recognize that other descriptors are correlated with density. In the revised manuscript we will add a short comparative analysis that includes radial distribution functions and angular distributions, demonstrating that coordination number provides the strongest discrimination between the density regimes where transferability succeeds or fails. revision: yes

Circularity Check

0 steps flagged

No circularity; results rest on comparative experiments and coordination analysis

full rationale

The paper advances its central claim through controlled from-scratch training, distillation, and foundation-model evaluation experiments that directly compare density-diverse versus temperature-diverse datasets, with coordination-number statistics offered as the physical explanation for observed transferability differences. No equations, fitted parameters renamed as predictions, or self-citation chains are invoked to derive the result; the findings are presented as empirical outcomes that can be replicated or falsified on the described systems and architectures. The derivation chain is therefore self-contained and does not reduce to its inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract supplies no explicit free parameters, axioms, or invented entities; the central variable is the sampling of density regimes during training data generation.

axioms (1)
  • domain assumption Local atomic coordination topology is the dominant physical feature controlling thermodynamic transferability of interatomic potentials.
    Invoked to explain why density matters more than temperature.

pith-pipeline@v0.9.0 · 5509 in / 1156 out tokens · 27221 ms · 2026-05-08T04:16:44.879368+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

67 extracted references · 7 canonical work pages

  1. [1]

    Simeski, M

    F. Simeski, M. Ihme, Supercritical fluids behave as complex networks, Nature Communications 14 (1) (2023) 1996

    2023

  2. [2]

    T. J. Yoon, M. Y. Ha, W. B. Lee, Y.-W. Lee, Molecular dynamics sim- ulation on the local density distribution and solvation structure of su- percritical co2 around naphthalene, The Journal of Supercritical Fluids 130 (2017) 364–372

    2017

  3. [3]

    J. Li, Y. Lang, B. Li, H. Zhang, J. Zhang, S. S. Rahman, Molecular mechanism of supercritical co2 enhancing shale oil production by ex- traction characteristics, Langmuir 41 (17) (2025) 11147–11160

    2025

  4. [4]

    Y. Han, B. Zhang, J. Jiang, L. Cen, J. Li, L. Zhao, Z. Xi, Understand- ing the enhancement of entrainers on the supercritical carbon dioxide extraction of phospholipids: A molecular dynamics simulation, Journal of Molecular Liquids 411 (2024) 125647

    2024

  5. [5]

    Muralidharan, M

    A. Muralidharan, M. I. Chaudhari, L. R. Pratt, S. B. Rempe, Molecular dynamics of lithium ion transport in a model solid electrolyte interphase, Scientific reports 8 (1) (2018) 10736

    2018

  6. [6]

    C. Park, S. Kim, S. Kim, M. Lee, S. Kim, J. Cho, A. Park, S. Kwon, M. Kim, S. Rho, et al., Lithium solvation structure and dynamics in an ionic liquid electrolyte: A deep learning-assisted analysis on polar- izable molecular dynamics simulations, Chemical Engineering Journal 508 (2025) 160662. 31

    2025

  7. [7]

    Kubisiak, A

    P. Kubisiak, A. Eilmes, Molecular dynamics simulations of ionic liquid based electrolytes for na-ion batteries: Effects of force field, The Journal of Physical Chemistry B 121 (42) (2017) 9957–9968

    2017

  8. [8]

    J. J. Winetrout, K. Kanhaiya, J. Kemppainen, P. J. in ‘t Veld, G. Sachdeva, R. Pandey, B. Damirchi, A. van Duin, G. M. Odegard, H. Heinz, Implementing reactivity in molecular dynamics simulations with harmonic force fields, Nature communications 15 (1) (2024) 7945

    2024

  9. [9]

    Q. Mao, M. Feng, X. Z. Jiang, Y. Ren, K. H. Luo, A. C. van Duin, Classical and reactive molecular dynamics: Principles and applications in combustion and energy systems, Progress in Energy and Combustion Science 97 (2023) 101084

    2023

  10. [10]

    J. Choi, B. Jun, Y. Jung, Reliable li-ion conductivity with efficient data generation and uncertainty estimation toward large-scale screen- ing, Chemical Engineering Journal 516 (2025) 163847

    2025

  11. [11]

    Zhang, Z

    P. Zhang, Z. Liu, X. Wang, R. Mi, Y. Pang, F. Pan, J. Li, Z. Wang, W. Wang, K. H. Luo, Unraveling the early-stage hydration mechanism of tricalcium silicate via machine learning potentials, Chemical Engi- neering Journal (2026) 176856

    2026

  12. [12]

    O. T. Unke, S. Chmiela, H. E. Sauceda, M. Gastegger, I. Poltavsky, K.T.Schutt, A.Tkatchenko, K.-R.Muller, Machinelearningforcefields, Chemical reviews 121 (16) (2021) 10142–10186

    2021

  13. [13]

    M. Yang, L. Bonati, D. Polino, M. Parrinello, Using metadynamics to 32 build neural network potentials for reactive events: the case of urea decomposition in water, Catalysis Today 387 (2022) 143–149

    2022

  14. [14]

    N. V. Avula, M. L. Klein, S. Balasubramanian, Understanding the anomalous diffusion of water in aqueous electrolytes using machine learned potentials, The Journal of Physical Chemistry Letters 14 (42) (2023) 9500–9507

    2023

  15. [15]

    M. Kim, S. Kim, C. Hyeon, J. W. Yu, S. Q. Choi, W. B. Lee, Anomalous water penetration in al3+ dissolution, The Journal of Physical Chem- istry Letters 15 (43) (2024) 10903–10908

    2024

  16. [16]

    J. H. Ryu, J. W. Yu, T. J. Yoon, W. B. Lee, Understanding the dielectric relaxation of liquid water using neural network potential and classical pairwise potential, Journal of Molecular Liquids 397 (2024) 124054

    2024

  17. [17]

    J. H. Ryu, S. Kim, M. Kim, J. W. Yu, T. J. Yoon, W. B. Lee, Explor- ing the reaction network of acetic acid in supercritical water via ma- chine learning interatomic potential, Journal of Chemical Information and Modeling 65 (16) (2025) 8614–8623

    2025

  18. [18]

    J. W. Yu, S. Kim, J. H. Ryu, W. B. Lee, T. J. Yoon, Spatiotemporal characterization of water diffusion anomalies in saline solutions using machine learning force field, Science Advances 10 (50) (2024) eadp9662

    2024

  19. [19]

    J. Zeng, D. Zhang, A. Peng, X. Zhang, S. He, Y. Wang, X. Liu, H. Bi, Y. Li, C. Cai, et al., Deepmd-kit v3: a multiple-backend framework for machine learning potentials, Journal of chemical theory and computa- tion 21 (9) (2025) 4375–4385. 33

    2025

  20. [20]

    Molinari, T

    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) (2022) 2453

    2022

  21. [21]

    Batatia, D

    I. Batatia, D. P. Kovacs, G. N. C. Simm, C. Ortner, G. Csanyi, MACE: Higher order equivariant message passing neural networks for fast and accurate force fields, in: A. H. Oh, A. Agarwal, D. Belgrave, K. Cho (Eds.), Advances in Neural Information Processing Systems, 2022. URLhttps://openreview.net/forum?id=YPpSngE-ZU

    2022

  22. [22]

    Y. Park, J. Kim, S. Hwang, S. Han, Scalable parallel algorithm for graph neural network interatomic potentials in molecular dynamics sim- ulations, J. Chem. Theory Comput. 20 (11) (2024) 4857–4868.doi: 10.1021/acs.jctc.4c00190

    work page doi:10.1021/acs.jctc.4c00190 2024

  23. [23]

    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) (2023) 1031–1041

    2023

  24. [24]

    I.Batatia, P.Benner, Y.Chiang, A.M.Elena, D.P.Kovács, J.Riebesell, X. R. Advincula, M. Asta, M. Avaylon, W. J. Baldwin, et al., A founda- tion model for atomistic materials chemistry, The Journal of chemical physics 163 (18) (2025)

    2025

  25. [25]

    J. Kim, J. Kim, J. Kim, J. Lee, Y. Park, Y. Kang, S. Han, Data- efficient multifidelity training for high-fidelity machine learning inter- 34 atomic potentials, J. Am. Chem. Soc. 147 (1) (2024) 1042–1054.doi: 10.1021/jacs.4c14455

    work page doi:10.1021/jacs.4c14455 2024

  26. [26]

    Kabylda, J

    A. Kabylda, J. T. Frank, S. Suárez-Dou, A. Khabibrakhmanov, L. Medrano Sandonas, O. T. Unke, S. Chmiela, K.-R. Muüller, A. Tkatchenko, Molecular simulations with a pretrained neural network and universal pairwise force fields, Journal of the American Chemical Society 147 (37) (2025) 33723–33734

    2025

  27. [27]

    Bowen, Materials Project Trajectory (MPtrj) Dataset

    B. Deng, Materials Project Trajectory (MPtrj) Dataset (7 2023). doi:10.6084/m9.figshare.23713842.v2. URLhttps://figshare.com/articles/dataset/Materials_ Project_Trjectory_MPtrj_Dataset/23713842

    work page doi:10.6084/m9.figshare.23713842.v2 2023

  28. [28]

    M. M. Ghahremanpour, P. J. Van Maaren, D. Van Der Spoel, The alexandria library, a quantum-chemical database of molecular properties for force field development, Scientific data 5 (1) (2018) 1–10

    2018

  29. [29]

    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. Zitnick, Z. W. Ulissi, Open materials 2024 (omat24) in- organic materials dataset and models, arXiv preprint arXiv:2410.12771 (2024)

    work page arXiv 2024

  30. [30]

    M. Y. Ha, T. J. Yoon, T. Tlusty, Y. Jho, W. B. Lee, Widom delta of supercritical gas–liquid coexistence, The journal of physical chemistry letters 9 (7) (2018) 1734–1738

    2018

  31. [31]

    A. Park, J. Ryu, W. B. Lee, Ionic liquid molecular dynamics simulation 35 with machine learning force fields: Dpmd and mace, arXiv preprint arXiv:2503.18249 (2025)

    work page arXiv 2025

  32. [32]

    Grunert, M

    M. Grunert, M. Großmann, J. Hänseroth, A. Flötotto, J. Oumard, J. L. Wolf, E. Runge, C. Dreßler, Modeling complex proton trans- port phenomena—exploring the limits of fine-tuning and transferability of foundational machine-learned force fields, The Journal of Physical Chemistry C 129 (21) (2025) 9662–9669

    2025

  33. [33]

    S. Ju, J. You, G. Kim, Y. Park, H. An, S. Han, Application of pre- trained universal machine-learning interatomic potential for physico- chemical simulation of liquid electrolytes in li-ion batteries, Digital Dis- covery 4 (6) (2025) 1544–1559

    2025

  34. [34]

    Z. A. Goodwin, M. B. Wenny, J. H. Yang, A. Cepellotti, J. Ding, K. Bystrom, B. R. Duschatko, A. Johansson, L. Sun, S. Batzner, et al., Transferability and accuracy of ionic liquid simulations with equivari- ant machine learning interatomic potentials, The Journal of Physical Chemistry Letters 15 (30) (2024) 7539–7547

    2024

  35. [35]

    W. G. Stark, C. van der Oord, I. Batatia, Y. Zhang, B. Jiang, G. Csányi, R. J. Maurer, Benchmarking of machine learning interatomic potentials for reactive hydrogen dynamics at metal surfaces, Machine Learning: Science and Technology 5 (3) (2024) 030501

    2024

  36. [36]

    O. T. Unke, M. Stöhr, S. Ganscha, T. Unterthiner, H. Maennel, S.Kashubin, D.Ahlin, M.Gastegger, L.MedranoSandonas, J.T.Berry- man, et al., Biomolecular dynamics with machine-learned quantum- 36 mechanical force fields trained on diverse chemical fragments, Science Advances 10 (14) (2024) eadn4397

    2024

  37. [37]

    J. Hoja, L. Medrano Sandonas, B. G. Ernst, A. Vazquez-Mayagoitia, R. A. DiStasio Jr, A. Tkatchenko, Qm7-x, a comprehensive dataset of quantum-mechanical properties spanning the chemical space of small organic molecules, Scientific data 8 (1) (2021) 43

    2021

  38. [38]

    Medrano Sandonas, D

    L. Medrano Sandonas, D. Van Rompaey, A. Fallani, M. Hilfiker, D. Hahn, L. Perez-Benito, J. Verhoeven, G. Tresadern, J. Kurt Weg- ner, H. Ceulemans, et al., Dataset for quantum-mechanical exploration of conformers and solvent effects in large drug-like molecules, Scientific Data 11 (1) (2024) 742

    2024

  39. [39]

    Eastman, P

    P. Eastman, P. K. Behara, D. L. Dotson, R. Galvelis, J. E. Herr, J. T. Horton, Y.Mao, J.D.Chodera, B.P.Pritchard, Y.Wang, etal., Spice, a dataset of drug-like molecules and peptides for training machine learning potentials, Scientific Data 10 (1) (2023) 11

    2023

  40. [40]

    A. G. Donchev, A. G. Taube, E. Decolvenaere, C. Hargus, R. T. McGib- bon, K.-H. Law, B. A. Gregersen, J.-L. Li, K. Palmo, K. Siva, et al., Quantum chemical benchmark databases of gold-standard dimer inter- action energies, Scientific data 8 (1) (2021) 55

    2021

  41. [41]

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

    1996

  42. [42]

    V. I. Anisimov, J. Zaanen, O. K. Andersen, Band theory and mott 37 insulators: Hubbard u instead of stoner i, Physical Review B 44 (3) (1991) 943

    1991

  43. [43]

    Adamo, M

    C. Adamo, M. Cossi, V. Barone, An accurate density functional method for the study of magnetic properties: the pbe0 model, Journal of Molec- ular Structure: THEOCHEM 493 (1-3) (1999) 145–157

    1999

  44. [44]

    Tkatchenko, R

    A. Tkatchenko, R. A. DiStasio Jr, R. Car, M. Scheffler, Accurate and efficient method for many-body van der waals interactions, Physical re- view letters 108 (23) (2012) 236402

    2012

  45. [45]

    Frank, O

    T. Frank, O. Unke, K.-R. Müller, So3krates: Equivariant attention for interactions on arbitrary length-scales in molecular systems, Advances in Neural Information Processing Systems 35 (2022) 29400–29413

    2022

  46. [46]

    Grimme, J

    S. Grimme, J. Antony, S. Ehrlich, H. Krieg, A consistent and accurate ab initio parametrization of density functional dispersion correction (dft- d) for the 94 elements h-pu, The Journal of chemical physics 132 (15) (2010)

    2010

  47. [47]

    Grimme, S

    S. Grimme, S. Ehrlich, L. Goerigk, Effect of the damping function in dispersion corrected density functional theory, Journal of computational chemistry 32 (7) (2011) 1456–1465

    2011

  48. [48]

    A. Dunn, Q. Wang, A. Ganose, D. Dopp, A. Jain, Benchmarking mate- rials property prediction methods: the matbench test set and automat- miner reference algorithm, npj Computational Materials 6 (1) (2020) 138. 38

    2020

  49. [49]

    Cipriani, M

    P. Cipriani, M. Nardone, F. Ricci, Neutron diffraction measurements on co2 in both undercritical and supercritical states, Physica B: Condensed Matter 241 (1997) 940–946

    1997

  50. [50]

    J. G. Harris, K. H. Yung, Carbon dioxide’s liquid-vapor coexistence curve and critical properties as predicted by a simple molecular model, The Journal of Physical Chemistry 99 (31) (1995) 12021–12024

    1995

  51. [51]

    L. B. Skinner, C. Huang, D. Schlesinger, L. G. Pettersson, A. Nilsson, C. J. Benmore, Benchmark oxygen-oxygen pair-distribution function of ambient water from x-ray diffraction measurements with a wide q-range, The Journal of chemical physics 138 (7) (2013)

    2013

  52. [52]

    Riera, E

    M. Riera, E. P. Yeh, F. Paesani, Data-driven many-body models for molecular fluids: Co2/h2o mixtures as a case study, Journal of Chemical Theory and Computation 16 (4) (2020) 2246–2257

    2020

  53. [53]

    S.Yue, M.Riera, R.Ghosh, A.Z.Panagiotopoulos, F.Paesani, Transfer- ability of data-driven, many-body models for co2 simulations in the va- por and liquid phases, The Journal of Chemical Physics 156 (10) (2022)

    2022

  54. [54]

    J. Kim, J. You, Y. Park, Y. Lim, Y. Kang, J. Kim, H. Jeon, S. Ju, D. Hong, S. Y. Lee, S. Choi, Y. Kim, J. W. Lee, S. Han, Optimizing cross-domain transfer for universal machine learning interatomic poten- tials, arXiv (2025).doi:10.48550/arxiv.2510.11241

    work page doi:10.48550/arxiv.2510.11241 2025

  55. [55]

    T. J. Yoon, M. Y. Ha, E. A. Lazar, W. B. Lee, Y.-W. Lee, Topological characterization of rigid–nonrigid transition across the frenkel line, The Journal of Physical Chemistry Letters 9 (22) (2018) 6524–6528. 39

    2018

  56. [56]

    T. J. Yoon, M. Y. Ha, W. B. Lee, Y.-W. Lee, E. A. Lazar, Topological generalization of the rigid-nonrigid transition in soft-sphere and hard- sphere fluids, Physical Review E 99 (5) (2019) 052603

    2019

  57. [57]

    T. J. Yoon, M. Y. Ha, E. A. Lazar, W. B. Lee, Y.-W. Lee, Topological extension of the isomorph theory based on the shannon entropy, Physical Review E 100 (1) (2019) 012118

    2019

  58. [58]

    Behler, M

    J. Behler, M. Parrinello, Generalized neural-network representation of high-dimensional potential-energy surfaces, Physical review letters 98 (14) (2007) 146401

    2007

  59. [59]

    Hjorth Larsen, J

    A. Hjorth Larsen, J. Jørgen Mortensen, J. Blomqvist, I. E. Castelli, R. Christensen, M. Dułak, J. Friis, M. N. Groves, B. Hammer, C. Har- gus, et al., The atomic simulation environment—a python library for working with atoms, Journal of Physics: Condensed Matter 29 (27) (2017) 273002

    2017

  60. [60]

    M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, G. A. Petersson, H. Nakat- suji, X. Li, M. Caricato, A. V. Marenich, J. Bloino, B. G. Janesko, R. Gomperts, B. Mennucci, H. P. Hratchian, J. V. Ortiz, A. F. Izmaylov, J. L. Sonnenberg, D. Williams-Young, F. Ding, F. Lipparini, F. Egidi, J. Goin...

    2016

  61. [61]

    Weigend, R

    F. Weigend, R. Ahlrichs, Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for h to rn: Design and assessment of accuracy, Physical Chemistry Chemical Physics 7 (18) (2005) 3297–3305

    2005

  62. [62]

    G. J. Martyna, A. Hughes, M. E. Tuckerman, Molecular dynamics algo- rithms for path integrals at constant pressure, The Journal of chemical physics 110 (7) (1999) 3275–3290

    1999

  63. [63]

    M. E. Tuckerman, Statistical mechanics: theory and molecular simula- tion, Oxford university press, 2023

    2023

  64. [64]

    Takamoto, C

    S. Takamoto, C. Shinagawa, D. Motoki, K. Nakago, W. Li, I. Kurata, T. Watanabe, Y. Yayama, H. Iriguchi, Y. Asano, T. Onodera, T. Ishii, T. Kudo, H. Ono, R. Sawada, R. Ishitani, M. Ong, T. Yamaguchi, T. Kataoka, A. Hayashi, T. Ibuka, Pfp: Universal neural network po- tential for material discovery (2021).arXiv:2106.14583

    work page arXiv 2021

  65. [65]

    Martínez, R

    L. Martínez, R. Andrade, E. G. Birgin, J. M. Martínez, Packmol: A package for building initial configurations for molecular dynamics simu- lations, Journal of computational chemistry 30 (13) (2009) 2157–2164. 41

    2009

  66. [66]

    Riera, C

    M. Riera, C. Knight, E. F. Bull-Vulpe, X. Zhu, H. Agnew, D. G. Smith, A. C. Simmonett, F. Paesani, Mbx: A many-body energy and force cal- culatorfordata-drivenmany-bodysimulations, TheJournalofChemical Physics 159 (5) (2023)

    2023

  67. [67]

    Ceriotti, J

    M. Ceriotti, J. More, D. E. Manolopoulos, i-pi: A python interface for ab initio path integral molecular dynamics simulations, Computer Physics Communications 185 (3) (2014) 1019–1026. 42

    2014