Pith. sign in

REVIEW 3 major objections 6 minor 64 references

A Machine Learning Framework for Modeling Ensemble Properties of Atomically Disordered Materials

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A GNN-plus-Monte-Carlo pipeline computes ensemble-averaged electrical and optical conductivity of disordered MXenes.

desk verdict A solid ML-MC framework for ensemble spectral properties, but the headline conductivity peak may be an artifact of the harmonic-mean averaging. read the letter →

arxiv 2506.15652 v1 pith:MU7IU6MF submitted 2025-06-18 cond-mat.mtrl-sci

Zhenyao Fang , Ting-Wei Hsu , Qimin Yan This is my paper · ORCID
classification cond-mat.mtrl-sci
keywords graphneuralnetworksMonteCarlosimulationconfigurationaldisorderMXenesurfaceterminationselectricalconductivityopticalensembleaveragingpersistenthomology
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper introduces a computational pipeline that combines graph neural networks (GNNs) with Monte Carlo simulations to compute thermodynamic and ensemble-averaged functional properties of atomically disordered materials, where direct density-functional sampling is too expensive. Applying it to the MXene monolayer Ti3C2O2−xFx with mixed F/O terminations and vacancies, the authors claim that surface-termination disorder changes the temperature dependence of electrical conductivity, producing a peak near the order-disorder phase-transition temperature. They explain the peak as a competition between F groups acting as scattering centers in the ordered phase and as electron donors that raise the Fermi level and restore metallicity in the disordered phase. By contrast, they find that optical conductivity stays nearly unchanged across the transition and is controlled mainly by the global F:O composition. The payoff would be a practical route to disorder-aware predictions for materials such as high-entropy alloys and disordered magnets.

What carries the argument

The central machinery is an equivariant graph neural network that takes a crystal structure as a graph and simultaneously predicts total energy, optical conductivity spectrum, and electrical conductivity spectrum, trained on thousands of configurations whose spectra are precomputed from maximally localized Wannier Hamiltonians. This surrogate is embedded in Metropolis Monte Carlo simulations, where the ensemble average of conductivities is taken as the harmonic mean across sampled configurations, treating the stack of configurational supercells as conductors in series. For partially terminated structures, the graph is augmented with virtual nodes for known vacancy sites and with persistent homology features that encode vacancy positions, since standard GNN message passing dilutes defect information.

What would settle it

Recompute the ensemble-averaged electrical conductivity with configuration-dependent scattering times obtained, for example, from first-principles electron-phonon or disorder scattering, or measure temperature-dependent conductivity of MXene films with controlled F:O ratios; if the peak near the order-disorder transition disappears in either test, the fixed-tau approximation is responsible for the central effect.

Watch

Extended reading notes

Core claim

The authors claim that the ensemble average of spectral transport properties, not just energies, can be computed for disordered materials by training an equivariant GNN on density-functional/Wannier data and running it inside Metropolis Monte Carlo at each temperature. On Ti3C2O2−xFx, this produces the paper's central finding: the configurational heat capacity has a peak marking an order-disorder transition at a temperature that rises with F and vacancy content, and the ensemble-averaged electrical conductivity shows a peak near that transition for F-rich stoichiometries, while staying monotonic for low F content. The microscopic story is that in the ordered phase isolated F groups act as scattering defects, while above the transition the average translational symmetry is restored and F doping raises the Fermi level, switching the transport from semimetal-like to metal-like. The same calculation shows that optical conductivity is essentially unaffected by local disorder and is governed by chemical composition, with the 1.5 eV peak intensity decreasing as F content increases. The paper frames these results as evidence that the framework captures disorder-driven physics that direct first-principles configurational sampling cannot reach.

Load-bearing premise

All electrical-conductivity spectra are computed with one fixed electron scattering time, tau = 10 fs, for every configuration and every temperature; if the real scattering time changes with the local termination environment or temperature, the predicted conductivity peak near the phase transition could be a numerical artifact rather than a physical transport signature.

Editorial extensions

If this is right

  • For F-rich Ti3C2O2−xFx, the ensemble electrical conductivity develops a maximum near the order-disorder transition, whereas low-F compositions show monotonic metallic-like decline.
  • Optical conductivity is insensitive to configurational disorder and temperature; the 1.5 eV peak amplitude tracks the F fraction, making it a composition probe.
  • Adding termination vacancies raises the order-disorder transition temperature and suppresses electrical conductivity by disrupting the periodic potential.
  • Grand-canonical simulations show that even small changes in F content with temperature can move the transition temperature and qualitatively alter transport behavior.
  • The same GNN-in-MC workflow is proposed as a general route for disorder-driven phenomena in high-entropy alloys and disordered magnetic compounds.

Reading between the lines

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

  • The single-tau approximation likely understates disorder scattering in the ordered phase; a config-dependent tau could strengthen or weaken the predicted peak, so the quantitative shape of the peak should be read as a model prediction rather than a direct observable.
  • The harmonic-mean ensemble average is physically motivated for series-connected configurational slabs, but arithmetic or effective-medium averages would give different magnitudes; comparing them would test how much the peak depends on the averaging rule.
  • The framework could be extended to predict scattering times or spectral broadening directly from the GNN, turning the fixed-tau assumption into a learned quantity and making the conductivity prediction falsifiable per configuration.
  • Since the optical 1.5 eV peak is tied to Fermi-level filling, optical measurements on samples with controlled F content offer a quick experimental check of the composition-governance claim, independent of transport.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. This paper introduces a computational framework that combines graph neural networks (GNNs) with Monte Carlo (MC) simulations to predict ensemble-averaged spectral properties of atomically disordered materials. The authors train an equivariant GNN to predict total energy, optical conductivity, and electrical conductivity spectra for 3,000 configurations of surface-termination-disordered MXene monolayers (Ti3C2O2−x−yFx), using DFT plus Wannier-interpolated Kubo-Greenwood/Boltzmann transport calculations as ground truth. The trained GNN is then used as the energy and property evaluator inside Metropolis MC simulations (canonical and semi-grand canonical ensembles) to compute heat capacities and ensemble-averaged conductivities as functions of temperature. The central physical claim is that the ensemble-averaged electrical conductivity exhibits a peak near the order-disorder transition temperature for F-rich stoichiometries, attributed to a competition between scattering and electron-filling effects, whereas the optical conductivity is robust to local disorder and reflects only the global chemical composition.

Significance. The framework addresses a real computational bottleneck: direct DFT sampling of configurational disorder for transport properties is prohibitive, and the GNN appears to achieve high accuracy on the reduced targets (R² of 0.99 for energy, 0.89–0.87 for optical, 0.96–0.98 for electrical conductivity; MAPE below 4% for the fully terminated set). The MC protocol is standard, and the use of persistent homology features for vacancy identification is a useful technical contribution. If the averaging prescription and transport approximations are validated, the framework could enable statistically meaningful ensemble predictions that are inaccessible to direct DFT. However, the novelty lies mainly in the combination of existing methods; the physical conclusions about the conductivity peak are conditional on the modeling choices discussed below.

major comments (3)
  1. [Computational Methods, Eq. (5)] The harmonic-mean prescription for the ensemble-averaged conductivity is not physically justified for a two-dimensional monolayer, and this choice is load-bearing for the central claim. The Metropolis MC samples alternative configurations of the same 2D sheet; these are not a stack of conductors in series. For an in-plane electric field, the configurational average of the conductivity should be an arithmetic mean (or an effective-medium average appropriate to the 2D geometry), not a harmonic mean, which is dominated by the least-conductive configurations. Because the ordered and disordered phases have very different conductivity distributions, the harmonic mean can produce a peak near the transition temperature even if the per-configuration conductivities are monotonic in temperature. The authors should test the robustness of the peak to the averaging rule (e.g., arithmetic mean or Bruggeman effective-medium approximation) and provide a more concrete justification for the series picture.
  2. [Computational Methods, Eqs. (2)-(3)] The electrical conductivity of each configuration is computed with a constant relaxation time tau = 10 fs, independent of configuration, temperature, and energy. As a result, the temperature dependence of sigma_ele within a single configuration is controlled only by the Fermi-Dirac derivative and the configuration-dependent transport distribution function; no microscopic disorder-scattering mechanism is included. The 'scattering effects' invoked in the interpretation of the conductivity peak therefore enter only through the configurational averaging in Eq. (5). Since Eq. (5) is itself in question (see previous comment), the attribution of the peak to a competition between scattering and electron filling is not supported by the calculations as presented. The authors should either demonstrate that the peak and its position are robust to tau within a reasonable range (e.g., 1-100 fs), adopt an energy- and/or configuration-dependent relaxation-time model, or soften the claim to refer to the configurational average of the transport distribution function rather than to scattering.
  3. [MC Simulation] The Monte Carlo averages are reported without error bars or convergence diagnostics. The paper states that each temperature uses 2e5 steps with a 20% burn-in, but no autocorrelation analysis, block averaging, or multiple independent runs are provided. The conductivity peak in Fig. 3(b) is a relatively small feature, and it is essential to demonstrate that it is not a Monte Carlo fluctuation. Please report error bars on the heat capacity and averaged conductivities (e.g., from jackknife over blocks or independent runs) and the autocorrelation times of the energy and conductivity time series.
minor comments (6)
  1. [Figure 3 caption] The caption of Figure 3 states 'under grand canonical ensemble' for results that are canonical; the grand canonical results appear in Figure 4. Please correct the captions to avoid confusion.
  2. [References] References [22] and [39] are duplicates of the same Graph Attention Networks paper; please consolidate.
  3. [Introduction] The phrase 'the recent development of machine learning techniques' is nearly repeated in the first two paragraphs; please streamline the wording.
  4. [Computational Methods, First-Principles Calculations] The sentence 'we employed the DFT+U method, with a U value of 3 eV is added to the 3d orbitals of Ti atoms' is grammatically incorrect; please revise.
  5. [Figure 2] The parity plots in Figure 2(b-d) report R² and MAPE for the reduced (mean) quantities; please also report the mean absolute error on the full spectra, or at least for the temperatures/photon energies shown in Figure 2(e-f), so the reader can judge spectral accuracy.
  6. [MC Simulation] The text refers to the 'grand canonical ensemble' while stating that the total number of surface termination groups remains fixed; this is a semi-grand canonical ensemble. Please define the ensemble more precisely or use standard terminology.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the conductivity peak is an emergent GNN/MC result, and the paper's self-citations are not load-bearing.

full rationale

The central claim—that ensemble-averaged electrical conductivity peaks near the order-disorder transition—is not imposed by construction. Per-configuration conductivities are computed from Wannier-interpolated band structures via Boltzmann transport with a stated constant τ=10 fs, and the ensemble average is taken as a harmonic mean over Metropolis-sampled configurations (Eqs. 2, 3, and 5). The peak emerges from the temperature-dependent configurational weights and the distinct per-configuration σ(T) of O-rich versus F-rich terminations; it is not a fitted parameter and is not a restatement of the input data. The harmonic-mean ansatz and fixed τ are modeling assumptions external to the target result; even if physically debatable, they are not circular because the predicted peak is not equivalent to those assumptions by definition. The paper does cite the authors' prior Wannier workflow (ref 36) and persistent-homology GNN (ref 25), but these are independent tools used to generate or featurize the training data, not to impose the ensemble-transport result. No uniqueness theorem is imported, and no known result is merely renamed. The GNN is benchmarked against DFT on held-out test configurations with R² values of 0.99, 0.89, and 0.96, providing external grounding for the learned surrogates. Overall, the derivation chain is self-contained at the level of the claimed prediction, and the circularity burden is low.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claims rest on DFT+U as ground truth, a fixed 10 fs relaxation time, and the harmonic-mean ensemble convention. The GNN and Monte Carlo are standard tools, and no new physical entities are introduced.

free parameters (3)
  • Hubbard U on Ti 3d orbitals = 3 eV
    Chosen as DFT+U correction; affects band structure and hence both conductivity spectra and order-disorder energetics.
  • Constant relaxation time tau = 10 fs
    Used in Boltzmann transport for all electrical conductivity values; taken from prior literature, not fitted to this system, and not allowed to vary across configurations or temperature.
  • Gaussian broadening eta for Delta function = 0.05 eV
    Used in the transport distribution function; broadens spectral features and affects absolute conductivity values.
assumptions (6)
  • domain assumption DFT+U with PBE functional and D3 correction provides accurate ground-state energies and band structures for Ti3C2T2 configurations.
    All labels for GNN training and Monte Carlo energies come from this approximate functional.
  • domain assumption Wannier-interpolated tight-binding Hamiltonians reproduce the DFT electronic structure near the Fermi level within a few meV.
    Invoked via Ref. 36; conductivity spectra are computed from these Hamiltonians.
  • domain assumption The 5x5x1 supercell is large enough for the sampled configurational ensemble to represent disorder effects.
    All configurations are generated in a 5x5x1 supercell with 175 atoms; finite-size effects are not discussed.
  • ad hoc to paper Harmonic mean of single-configuration conductivities gives the correct ensemble average for a stack of supercells in series.
    Defined in Eq. (5); this averaging choice is physically motivated by series resistors but is a modeling convention, not directly validated.
  • domain assumption Metropolis Monte Carlo runs of 2e5 steps per temperature with 20% burn-in are converged.
    No convergence test or error bar is reported for heat capacity or averaged conductivities.
  • domain assumption GNN prediction errors do not materially change thermodynamic averages and transition temperatures.
    The GNN surrogate replaces DFT for energies and spectra inside the Monte Carlo loop; error propagation is not analyzed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Machine Learning Framework for Modeling Ensemble Properties of Atomically Disordered Materials." pith.science (2026). https://pith.science/paper/MU7IU6MF

@misc{pith2026250615652,
  author       = {Pith},
  title        = {Pith review of: A Machine Learning Framework for Modeling Ensemble Properties of Atomically Disordered Materials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MU7IU6MF}},
  note         = {Machine review of arXiv:2506.15652}
}
abstract

Disorder, though naturally present in experimental samples and strongly influencing a wide range of material phenomena, remains underexplored in first-principles studies due to the computational cost of sampling the large supercell and configurational space. The recent development of machine learning techniques, particularly graph neural networks (GNNs), has enabled the efficient and accurate predictions of complex material properties, offering promising tools for studying disordered systems. In this work, we introduce a computational framework that integrates GNNs with Monte Carlo simulations for efficient calculations of thermodynamic properties and ensemble-averaged functional properties of disordered materials. Using the surface-termination-disordered MXene monolayer \ch{Ti3C2T}$_{2-x}$ as a representative system, we investigate the effect of surface termination disorder involving \ch{-F}, \ch{-O}, and termination vacancies on the electrical and optical conductivity spectra. We find that surface termination disorder affects the temperature dependence of electrical conductivity, inducing a peak close to the order-disorder phase transition temperature that reflects the competition between scattering and electron filling effects of the surface termination groups across the phase transition. In contrast, optical conductivity remains robust to local disorder across a wide temperature range and is governed primarily by the global chemical composition of surface terminations. These results demonstrate the utility of our machine-learning-assisted framework for statistically modeling disorder effects and ensemble properties in complex materials, opening new avenues for future studies of disorder-driven phenomena in systems such as high-entropy alloys and disordered magnetic compounds.

Figures

Figures reproduced from arXiv: 2506.15652 by the authors.

Figure 4
Figure 4. Therefore, under the gas-phase annealing conditions, surface termination vacancy [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 1
Figure 1. A schematic plot of the computational framework to integrate GNN and MC [PITH_FULL_IMAGE:figures/full_fig_p027_1.png] view at source ↗
Figure 2
Figure 2. (a) Examples of configurations in the fully terminated MXene Ti [PITH_FULL_IMAGE:figures/full_fig_p028_2.png] view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: The thermodynamic property and the ensemble averaged electrical and optical [PITH_FULL_IMAGE:figures/full_fig_p029_3.png]
Figure 4
Figure 4. Figure 4: The thermodynamic property and the ensemble averaged electrical and optical [PITH_FULL_IMAGE:figures/full_fig_p030_4.png]
Figure 5
Figure 5. Figure 5: The thermodynamic property and the ensemble averaged electrical and optical [PITH_FULL_IMAGE:figures/full_fig_p031_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 60 canonical work pages

  1. [1]

    J.; Jain, J

    Pu, S.; Sreejith, G. J.; Jain, J. K. Anderson Localization in the Fractional Quantum Hall Effect. Phys. Rev. Lett. 2022, 128, 116801

    work page 2022

  2. [2]

    Cutler, M.; Mott, N. F. Observation of Anderson Localization in an Electron Gas. Phys. Rev. 1969, 181, 1336--1340

    work page 1969

  3. [3]

    Anderson, P. W. Absence of Diffusion in Certain Random Lattices. Phys. Rev. 1958, 109, 1492--1505

    work page 1958

  4. [4]

    Phonon transport in disordered alloys: A Multiple-scattering approach

    Alam, A. Phonon transport in disordered alloys: A Multiple-scattering approach. Phys. Rev. B 2021, 104, 104202

    work page 2021

  5. [5]

    Dynamic disorder phonon scattering mediated by Cu atomic hopping and diffusion in Cu3SbSe3

    Wang, C.; Wu, Y.; Pei, Y.; Chen, Y. Dynamic disorder phonon scattering mediated by Cu atomic hopping and diffusion in Cu3SbSe3. npj Comput. Mater. 2020, 6, 155

    work page 2020

  6. [6]

    Thermal conductivity modeling on highly disordered crystalline Y _ 1-x Nb _x O _ 1.5+x : Beyond the phonon scenario

    Huang, M.; Liu, X.; Zhang, P.; Qian, X.; Feng, Y.; Li, Z.; Pan, W.; Wan, C. Thermal conductivity modeling on highly disordered crystalline Y _ 1-x Nb _x O _ 1.5+x : Beyond the phonon scenario . Appl. Phys. Lett. 2021, 118, 073901

    work page 2021

  7. [7]

    Atomic Disorder Enables Superior Catalytic Surface of Pt-Based Catalysts for Alkaline Hydrogen Evolution

    Xie, Y.; Cai, J.; Wu, Y.; Hao, X.; Bian, Z.; Niu, S.; Yin, X.; Pei, Z.; Sun, D.; Zhu, Z.; Lu, Z.; Niu, D.; Wang, G. Atomic Disorder Enables Superior Catalytic Surface of Pt-Based Catalysts for Alkaline Hydrogen Evolution. ACS Mater. Lett. 2021, 3, 1738--1745

    work page 2021

  8. [8]

    A.; Morgen, P.; Chamier, J.; Ravnsbæk, D

    Sharma, R.; Karlsen, M. A.; Morgen, P.; Chamier, J.; Ravnsbæk, D. B.; Andersen, S. M. Crystalline Disorder, Surface Chemistry, and Their Effects on the Oxygen Evolution Reaction (OER) Activity of Mass-Produced Nanostructured Iridium Oxides. ACS Appl. Energy Mater. 2021, 4, 2552--2562

    work page 2021

Show all 64 references
  1. [9]

    J.; Pan, J.; Tamboli, A

    Cordell, J. J.; Pan, J.; Tamboli, A. C.; Tucker, G. J.; Lany, S. Probing configurational disorder in ZnGeN _ 2 using cluster-based Monte Carlo. Phys. Rev. Mater. 2021, 5, 024604

    2021

  2. [10]

    J.; Dove, M

    Cliffe, M. J.; Dove, M. T.; Drabold, D. A.; Goodwin, A. L. Structure Determination of Disordered Materials from Diffraction Data. Phys. Rev. Lett. 2010, 104, 125501

    2010

  3. [11]

    M.; Yim, S.-Y.; Agarwal, R

    Leem, Y.-C.; Fang, Z.; Lee, Y.-K.; Kim, N.-Y.; Kakekhani, A.; Liu, W.; Cho, S.-P.; Kim, C.; Wang, Y.; Ji, Z.; Patra, A.; Kronik, L.; Rappe, A. M.; Yim, S.-Y.; Agarwal, R. Optically Triggered Emergent Mesostructures in Monolayer WS2. Nano Letters 2024, 24, 5436--5443

    2024

  4. [12]

    Large scale hybrid Monte Carlo simulations for structure and property prediction

    Prokhorenko, S.; Kalke, K.; Nahas, Y.; Bellaiche, L. Large scale hybrid Monte Carlo simulations for structure and property prediction. npj Comput. Mater. 2018, 4, 80

    2018

  5. [13]

    Wang, F.; Landau, D. P. Efficient, Multiple-Range Random Walk Algorithm to Calculate the Density of States. Phys. Rev. Lett. 2001, 86, 2050--2053

    2001

  6. [14]

    C.; Torbr\"ugge, S.; Landau, D

    Zhou, C.; Schulthess, T. C.; Torbr\"ugge, S.; Landau, D. P. Wang-Landau Algorithm for Continuous Models and Joint Density of States. Phys. Rev. Lett. 2006, 96, 120201

    2006

  7. [15]

    Cluster expansion method for multicomponent systems based on optimal selection of structures for density-functional theory calculations

    Seko, A.; Koyama, Y.; Tanaka, I. Cluster expansion method for multicomponent systems based on optimal selection of structures for density-functional theory calculations. Phys. Rev. B 2009, 80, 165122

    2009

  8. [16]

    Atomic cluster expansion for accurate and transferable interatomic potentials

    Drautz, R. Atomic cluster expansion for accurate and transferable interatomic potentials. Phys. Rev. B 2019, 99, 014104

    2019

  9. [17]

    Sanchez, J. M. Cluster expansions and the configurational energy of alloys. Phys. Rev. B 1993, 48, 14013--14015

    1993

  10. [18]

    Constructing and Evaluating Machine-Learned Interatomic Potentials for Li-Based Disordered Rocksalts

    Choyal, V.; Sagar, N.; Sai Gautam, G. Constructing and Evaluating Machine-Learned Interatomic Potentials for Li-Based Disordered Rocksalts. Journal of chemical theory and computation 2024, 20, 4844--4856

    2024

  11. [19]

    M.; Isayev, O

    Anstine, D. M.; Isayev, O. Machine Learning Interatomic Potentials and Long-Range Physics. The Journal of Physical Chemistry A 2023, 127, 2417--2431

    2023

  12. [20]

    S.; Armiento, R.; Alling, B

    Casillas-Trujillo, L.; Parackal, A. S.; Armiento, R.; Alling, B. Evaluating and improving the predictive accuracy of mixing enthalpies and volumes in disordered alloys from universal pretrained machine learning potentials. Phys. Rev. Mater. 2024, 8, 113803

    2024

  13. [21]

    Xie, T.; Grossman, J. C. Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties. Phys. Rev. Lett. 2018, 120, 145301

    2018

  14. [22]

    Graph Attention Networks

    Veličković, P.; Cucurull, G.; Casanova, A.; Romero, A.; Liò, P.; Bengio, Y. Graph Attention Networks. International Conference on Learning Representations. 2018

    2018

  15. [23]

    Fung, V.; Zhang, J.; Juarez, E.; Sumpter, B. G. Benchmarking graph neural networks for materials chemistry. npj Comput. Mater. 2021, 7, 84

    2021

  16. [24]

    Graph neural networks for materials science and chemistry

    Reiser, P.; Neubert, M.; Eberhard, A.; Torresi, L.; Zhou, C.; Shao, C.; Metni, H.; van Hoesel, C.; Schopmans, H.; Sommer, T.; Friederich, P. Graph neural networks for materials science and chemistry. Commun. Mater. 2022, 3, 93

    2022

  17. [25]

    Leveraging Persistent Homology Features for Accurate Defect Formation Energy Predictions via Graph Neural Networks

    Fang, Z.; Yan, Q. Leveraging Persistent Homology Features for Accurate Defect Formation Energy Predictions via Graph Neural Networks. Chemistry of materials 2025, 37, 1531--1540

    2025

  18. [26]

    Bai, J.; Du, Y.; Wang, Y.; Kong, S.; Gregoire, J.; Gomes, C. P. Xtal2DoS: Attention-based Crystal to Sequence Learning for Density of States Prediction. NeurIPS 2022 AI for Science: Progress and Promises. 2022

    2022

  19. [27]

    T.; Alatas, A.; Kong, J.; Li, M

    Chen, Z.; Andrejevic, N.; Smidt, T.; Ding, Z.; Xu, Q.; Chi, Y.; Nguyen, Q. T.; Alatas, A.; Kong, J.; Li, M. Direct Prediction of Phonon Density of States With Euclidean Neural Networks. Advanced science 2021, 8, e2004214--n/a

    2021

  20. [28]

    T.; Okabe, R.; Chotrattanapituk, A.; Li, M

    Hung, N. T.; Okabe, R.; Chotrattanapituk, A.; Li, M. Universal Ensemble‐Embedding Graph Neural Network for Direct Prediction of Optical Spectra from Crystal Structures. Advanced materials (Weinheim) 2024, 36, e2409175--n/a

    2024

  21. [29]

    M.; Charpagne, M.-A.; Latypov, M

    Hestroffer, J. M.; Charpagne, M.-A.; Latypov, M. I.; Beyerlein, I. J. Graph neural networks for efficient learning of mechanical properties of polycrystals. Computational materials science 2023, 217, 111894--

    2023

  22. [30]

    Deep-learning density functional theory Hamiltonian for efficient ab initio electronic-structure calculation

    Li, H.; Wang, Z.; Zou, N.; Ye, M.; Xu, R.; Gong, X.; Duan, W.; Xu, Y. Deep-learning density functional theory Hamiltonian for efficient ab initio electronic-structure calculation. Nature Computational Science 2022, 2, 367--377

    2022

  23. [31]

    Towards accurate prediction of configurational disorder properties in materials using graph neural networks

    Fang, Z.; Yan, Q. Towards accurate prediction of configurational disorder properties in materials using graph neural networks. npj computational materials 2024, 10, 91--7

    2024

  24. [32]

    A.; Pereyra, C.; Soroush, M.; Rappe, A

    Fang, H.; Thakur, A.; Zahmatkeshsaredorahi, A.; Fang, Z.; Rad, V.; Shamsabadi, A. A.; Pereyra, C.; Soroush, M.; Rappe, A. M.; Xu, X. G.; Anasori, B.; Fakhraai, Z. Stabilizing Ti3C2Tx MXene flakes in air by removing confined water. Proceedings of the National Academy of Science...

    2024

  25. [33]

    Lim, K. R. G.; Shekhirev, M.; Wyatt, B. C.; Anasori, B.; Gogotsi, Y.; Seh, Z. W. Fundamentals of MXene synthesis. Nature Synthesis 2022, 1, 601--614

    2022

  26. [34]

    The world of two-dimensional carbides and nitrides (MXenes)

    VahidMohammadi, A.; Rosen, J.; Gogotsi, Y. The world of two-dimensional carbides and nitrides (MXenes). Science (American Association for the Advancement of Science) 2021, 372

    2021

  27. [35]

    Chemical Origin of Termination-Functionalized MXenes: Ti3C2 T 2 as a Case Study

    Hu, T.; Li, Z.; Hu, M.; Wang, J.; Hu, Q.; Li, Q.; Wang, X. Chemical Origin of Termination-Functionalized MXenes: Ti3C2 T 2 as a Case Study . The Journal of Physical Chemistry C 2017, 121, 19254--19261

    2017

  28. [36]

    Database of Tensorial Optical and Transport Properties of Materials From the Wannier Function Method

    Fang, Z.; Hsu, T.-W.; Yan, Q. Database of Tensorial Optical and Transport Properties of Materials From the Wannier Function Method. 2025; https://arxiv.org/abs/2504.00771

    2025 arXiv

  29. [37]

    A.; Yates, J

    Marzari, N.; Mostofi, A. A.; Yates, J. R.; Souza, I.; Vanderbilt, D. Maximally localized Wannier functions: Theory and applications. Rev. Mod. Phys. 2012, 84, 1419--1475

    2012

  30. [38]

    High-Throughput Screening and Automated Processing toward Novel Topological Insulators

    Zhang, Z.; Zhang, R.-W.; Li, X.; Koepernik, K.; Yao, Y.; Zhang, H. High-Throughput Screening and Automated Processing toward Novel Topological Insulators. The Journal of Physical Chemistry Letters 2018, 9, 6224--6231

    2018

  31. [39]

    How Attentive are Graph Attention Networks? International Conference on Learning Representations

    Brody, S.; Alon, U.; Yahav, E. How Attentive are Graph Attention Networks? International Conference on Learning Representations. 2022

    2022

  32. [40]

    Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification

    Shi, Y.; Huang, Z.; Feng, S.; Zhong, H.; Wang, W.; Sun, Y. Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification. Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence. 2021

    2021

  33. [41]

    P.; Kornbluth, M.; Molinari, N.; Smidt, T

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

    2022

  34. [42]

    B.; Anasori, B.; Hong, S

    Shahzad, F.; Alhabeb, M.; Hatter, C. B.; Anasori, B.; Hong, S. M.; Koo, C. M.; Gogotsi, Y. Electromagnetic interference shielding with 2D transition metal carbides (MXenes). Science (American Association for the Advancement of Science) 2016, 353, 1137--1140

    2016

  35. [43]

    Lyu, B.; Kim, M.; Jing, H.; Kang, J.; Qian, C.; Lee, S.; Cho, J. H. Large-Area MXene Electrode Array for Flexible Electronics. ACS nano 2019, 13, 11392--11400

    2019

  36. [44]

    Advanced materials (Weinheim) 2011, 23, 4248--4253

    Two-Dimensional Nanocrystals Produced by Exfoliation of Ti3AlC2. Advanced materials (Weinheim) 2011, 23, 4248--4253

    2011

  37. [45]

    The Rise of MXenes

    Gogotsi, Y.; Anasori, B. The Rise of MXenes. ACS nano 2019, 13, 8491--8494

    2019

  38. [46]

    Li-intercalation boosted oxygen vacancies enable efficient electrochemical nitrogen reduction on ultrathin TiO2 nanosheets

    Zhao, R.; Wang, G.; Mao, Y.; Bao, X.; Wang, Z.; Wang, P.; Liu, Y.; Zheng, Z.; Dai, Y.; Cheng, H.; Huang, B. Li-intercalation boosted oxygen vacancies enable efficient electrochemical nitrogen reduction on ultrathin TiO2 nanosheets . Chemical Engineering Journal 2022, 430, 133085

    2022

  39. [47]

    Cation-induced Ti3C2Tx MXene hydrogel for capacitive energy storage

    Zhang, Z.; Yao, Z.; Li, Y.; Lu, S.; Wu, X.; Jiang, Z. Cation-induced Ti3C2Tx MXene hydrogel for capacitive energy storage . Chemical Engineering Journal 2022, 433, 134488

    2022

  40. [48]

    Fang, H.; Fang, Z.; Thakur, A.; Rad, V.; S, N. C. B.; Michałowski, P.; Soroush, M.; Anasori, B.; Rappe, A. M.; Fakhraai, Z. Why Ti3C2Tx MXenes Are Conductive but Not Plasmonic in the Optical Domain. 2024

    2024

  41. [49]

    Caffrey, N. M. Effect of mixed surface terminations on the structural and electrochemical properties of two-dimensional Ti3C2T2 and V2CT2 MXenes multilayers. Nanoscale 2018, 10, 13520--13530

    2018

  42. [50]

    Chalcogen and halogen surface termination coverage in MXenes—structure, stability, and properties

    Dahlqvist, M.; Rosen, J. Chalcogen and halogen surface termination coverage in MXenes—structure, stability, and properties. NPJ 2D materials and applications 2024, 8, 65--13

    2024

  43. [51]

    Xu, B.; Fang, Z.; Sanchez-M\'artınez, M.-A.; Venderbos, J. W. F.; Ni, Z.; Qiu, T.; Manna, K.; Wang, K.; Paglione, J.; Bernhard, C.; Felser, C.; Mele, E. J.; Grushin, A. G.; Rappe, A. M.; Wu, L. Optical signatures of multifold fermions in the chiral topological semimetal CoSi. ...

    2020

  44. [52]

    Ab initio molecular dynamics for liquid metals

    Kresse, G.; Hafner, J. Ab initio molecular dynamics for liquid metals. Phys. Rev. B 1993, 47, 558--561

    1993

  45. [53]

    Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set

    Kresse, G.; Furthmüller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comput. Mater. Sci. 1996, 6, 15--50

    1996

  46. [54]

    From ultrasoft pseudopotentials to the projector augmented-wave method

    Kresse, G.; Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 1999, 59, 1758--1775

    1999

  47. [55]

    Bl\"ochl, P. E. Projector augmented-wave method. Phys. Rev. B 1994, 50, 17953--17979

    1994

  48. [56]

    P.; Burke, K.; Ernzerhof, M

    Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865--3868

    1996

  49. [57]

    I.; Zaanen, J.; Andersen, O

    Anisimov, V. I.; Zaanen, J.; Andersen, O. K. Band theory and Mott insulators: Hubbard U instead of Stoner I. Phys. Rev. B 1991, 44, 943--954

    1991

  50. [58]

    G., Stephan Ehrlich Effect of the Damping Function in Dispersion CorrectedDensity Functional Theory

    Stefan Grimme, L. G., Stephan Ehrlich Effect of the Damping Function in Dispersion CorrectedDensity Functional Theory. Journal of Computational Chemistry 2011, 32, 1456--1465

    2011

  51. [59]

    Animalu, A. O. E. Optical Conductivity of Simple Metals. Phys. Rev. 1967, 163, 557--562

    1967

  52. [60]

    BoltzWann: A code for the evaluation of thermoelectric and electronic transport properties with a maximally-localized Wannier functions basis

    Pizzi, G.; Volja, D.; Kozinsky, B.; Fornari, M.; Marzari, N. BoltzWann: A code for the evaluation of thermoelectric and electronic transport properties with a maximally-localized Wannier functions basis. Computer Physics Communications 2014, 185, 422--429

    2014

  53. [61]

    High-throughput prediction of the carrier relaxation time via data-driven descriptor

    Zhou, Z.; Cao, G.; Liu, J.; Liu, H. High-throughput prediction of the carrier relaxation time via data-driven descriptor. npj computational materials 2020, 6

    2020

  54. [62]

    Sernelius, B. E. Intraband relaxation time in highly excited semiconductors. Phys. Rev. B 1991, 43, 7136--7144

    1991

  55. [63]

    Optuna: A Next-Generation Hyperparameter Optimization Framework

    Akiba, T.; Sano, S.; Yanase, T.; Ohta, T.; Koyama, M. Optuna: A Next-Generation Hyperparameter Optimization Framework. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2019; p 2623–2631

    2019

  56. [64]

    Algorithms for Hyper-Parameter Optimization

    Bergstra, J.; Bardenet, R.; Bengio, Y.; K\' e gl, B. Algorithms for Hyper-Parameter Optimization. Advances in Neural Information Processing Systems. 2011 mcitethebibliography main.tex0000664000000000000000000013751015024572733011243 0ustar rootroot [journal=jacsat,manuscript=a...

    2011

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.