Pub Date : 2026-01-15DOI: 10.1038/s41524-025-01916-8
Tiago F. T. Cerqueira, Haichen Wang, Silvana Botti, Miguel A. L. Marques
We present a novel approach to generate a fingerprint for crystalline materials that balances efficiency for machine processing and human interpretability, allowing its application in both machine learning inference and understanding of structure-property relationships. Our proposed material encoding has two components: one representing the crystal structure and the other characterizing the chemical composition, which we call Pettifor embedding. For the latter, we construct a non-orthogonal space where each axis represents a chemical element and where the angle between the axes quantifies a measure of the similarity between them. The chemical composition is then defined by the point on the unit sphere in this non-orthogonal space. We show that the Pettifor embeddings systematically outperform other commonly used elemental embeddings in compositional machine learning models. Using the Pettifor embeddings to define a distance metric and applying dimension reduction techniques, we construct a two-dimensional global map of the space of thermodynamically stable crystalline compounds. Despite their simplicity, such maps succeed in providing a physical separation of material classes according to basic physical properties.
{"title":"A non-orthogonal representation for materials based on chemical similarity","authors":"Tiago F. T. Cerqueira, Haichen Wang, Silvana Botti, Miguel A. L. Marques","doi":"10.1038/s41524-025-01916-8","DOIUrl":"https://doi.org/10.1038/s41524-025-01916-8","url":null,"abstract":"We present a novel approach to generate a fingerprint for crystalline materials that balances efficiency for machine processing and human interpretability, allowing its application in both machine learning inference and understanding of structure-property relationships. Our proposed material encoding has two components: one representing the crystal structure and the other characterizing the chemical composition, which we call Pettifor embedding. For the latter, we construct a non-orthogonal space where each axis represents a chemical element and where the angle between the axes quantifies a measure of the similarity between them. The chemical composition is then defined by the point on the unit sphere in this non-orthogonal space. We show that the Pettifor embeddings systematically outperform other commonly used elemental embeddings in compositional machine learning models. Using the Pettifor embeddings to define a distance metric and applying dimension reduction techniques, we construct a two-dimensional global map of the space of thermodynamically stable crystalline compounds. Despite their simplicity, such maps succeed in providing a physical separation of material classes according to basic physical properties.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"45 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968816","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-15DOI: 10.1038/s41524-026-01959-5
Shengtao Jiang, Arthur C. Campello, Wei He, Jiajia Wen, Daniel M. Pajerowski, Young S. Lee, Hong-Chen Jiang
Spin-1/2 kagome antiferromagnets are leading candidates for realizing quantum spin liquid (QSL) ground states. While QSL ground states are predicted for the pure Heisenberg model, understanding the robustness of the QSL to additional interactions that may be present in real materials is a forefront question in the field. Here we employ large-scale density-matrix renormalization group simulations to investigate the effects of next-nearest neighbor exchange couplings J2 and Dzyaloshinskii-Moriya interactions D, which are relevant to understanding the prototypical kagome materials herbertsmithite and Zn-barlowite. By utilizing clusters as large as XC12 and extrapolating the results to the thermodynamic limit, we precisely delineate the scope of the QSL phase, which remains robust across an expanded parameter range of J2 and D. Direct comparison of the simulated static and dynamic spin structure factors with inelastic neutron scattering reveals the parameter space of the Hamiltonians for herbertsmithite and Zn-barlowite, and, importantly, provides compelling evidence that both materials exist within the QSL phase. These results establish a powerful convergence of theory and experiment in this most elusive state of matter.
{"title":"Quantifying the phase diagram and Hamiltonian of S = 1/2 kagome antiferromagnets: bridging theory and experiment","authors":"Shengtao Jiang, Arthur C. Campello, Wei He, Jiajia Wen, Daniel M. Pajerowski, Young S. Lee, Hong-Chen Jiang","doi":"10.1038/s41524-026-01959-5","DOIUrl":"https://doi.org/10.1038/s41524-026-01959-5","url":null,"abstract":"Spin-1/2 kagome antiferromagnets are leading candidates for realizing quantum spin liquid (QSL) ground states. While QSL ground states are predicted for the pure Heisenberg model, understanding the robustness of the QSL to additional interactions that may be present in real materials is a forefront question in the field. Here we employ large-scale density-matrix renormalization group simulations to investigate the effects of next-nearest neighbor exchange couplings J2 and Dzyaloshinskii-Moriya interactions D, which are relevant to understanding the prototypical kagome materials herbertsmithite and Zn-barlowite. By utilizing clusters as large as XC12 and extrapolating the results to the thermodynamic limit, we precisely delineate the scope of the QSL phase, which remains robust across an expanded parameter range of J2 and D. Direct comparison of the simulated static and dynamic spin structure factors with inelastic neutron scattering reveals the parameter space of the Hamiltonians for herbertsmithite and Zn-barlowite, and, importantly, provides compelling evidence that both materials exist within the QSL phase. These results establish a powerful convergence of theory and experiment in this most elusive state of matter.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"266 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968814","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-15DOI: 10.1038/s41524-025-01938-2
Rogério Almeida Gouvêa, Pierre-Paul De Breuck, Tatiane Pretto, Gian-Marco Rignanese, Marcos José Leite Santos
This study introduces MatterVial, an innovative hybrid framework for feature-based machine learning in materials science. MatterVial expands the feature space by integrating latent representations from a diverse suite of pretrained graph-neural network (GNN) models—including structure-based (MEGNet), composition-based (ROOST), and equivariant (ORB) graph networks—with computationally efficient, GNN-approximated descriptors and novel features from symbolic regression. Our approach combines the chemical transparency of traditional feature-based models with the predictive power of deep learning architectures. When augmenting the feature-based model MODNet on Matbench tasks, this method yields significant error reductions and elevates its performance to be competitive with, and in several cases superior to, state-of-the-art end-to-end GNNs, with accuracy increases exceeding 40% for multiple tasks. An integrated interpretability module, employing surrogate models and symbolic regression, decodes the latent GNN-derived descriptors into explicit, physically meaningful formulas. This unified framework advances materials informatics by providing a high-performance, transparent tool that aligns with the principles of explainable AI, paving the way for more targeted and autonomous materials discovery.
{"title":"Combining feature-based approaches with graph neural networks and symbolic regression for synergistic performance and interpretability","authors":"Rogério Almeida Gouvêa, Pierre-Paul De Breuck, Tatiane Pretto, Gian-Marco Rignanese, Marcos José Leite Santos","doi":"10.1038/s41524-025-01938-2","DOIUrl":"https://doi.org/10.1038/s41524-025-01938-2","url":null,"abstract":"This study introduces MatterVial, an innovative hybrid framework for feature-based machine learning in materials science. MatterVial expands the feature space by integrating latent representations from a diverse suite of pretrained graph-neural network (GNN) models—including structure-based (MEGNet), composition-based (ROOST), and equivariant (ORB) graph networks—with computationally efficient, GNN-approximated descriptors and novel features from symbolic regression. Our approach combines the chemical transparency of traditional feature-based models with the predictive power of deep learning architectures. When augmenting the feature-based model MODNet on Matbench tasks, this method yields significant error reductions and elevates its performance to be competitive with, and in several cases superior to, state-of-the-art end-to-end GNNs, with accuracy increases exceeding 40% for multiple tasks. An integrated interpretability module, employing surrogate models and symbolic regression, decodes the latent GNN-derived descriptors into explicit, physically meaningful formulas. This unified framework advances materials informatics by providing a high-performance, transparent tool that aligns with the principles of explainable AI, paving the way for more targeted and autonomous materials discovery.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"17 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968817","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-14DOI: 10.1038/s41524-025-01889-8
Siyu Chen, Yao Wei, Bartomeu Monserrat, Jan M. Tomczak, Samuel Poncé
Rare-earth superhydrides have attracted considerable attention because of their high critical superconducting temperature under extreme pressures. They are known to have localized valence electrons, implying strong electronic correlations. However, such many-body effects are rarely included in first-principles studies of rare-earth superhydrides because of the complexity of their high-pressure phases. In this work, we use a combined density functional theory and dynamical mean-field theory approach to study both electrons and phonons in the prototypical rare-earth superhydride CeH9, shedding light on the impact of electronic correlations on its critical temperature for phonon-mediated superconductivity. Our findings indicate that electronic correlations result in a larger electronic density at the Fermi level, a bigger superconducting gap, and softer vibrational modes associated with hydrogen atoms. Together, the inclusion of these correlation signatures increases the Migdal-Eliashberg superconducting critical temperature from 47 K to 96 K, close to the measured 95 K. Our results reconcile experimental observations and theoretical predictions for CeH9 and herald a path towards the quantitative modeling of phonon-mediated superconductivity for interacting electron systems.
{"title":"Impact of electronic correlations on the superconductivity of high-pressure CeH9","authors":"Siyu Chen, Yao Wei, Bartomeu Monserrat, Jan M. Tomczak, Samuel Poncé","doi":"10.1038/s41524-025-01889-8","DOIUrl":"https://doi.org/10.1038/s41524-025-01889-8","url":null,"abstract":"Rare-earth superhydrides have attracted considerable attention because of their high critical superconducting temperature under extreme pressures. They are known to have localized valence electrons, implying strong electronic correlations. However, such many-body effects are rarely included in first-principles studies of rare-earth superhydrides because of the complexity of their high-pressure phases. In this work, we use a combined density functional theory and dynamical mean-field theory approach to study both electrons and phonons in the prototypical rare-earth superhydride CeH9, shedding light on the impact of electronic correlations on its critical temperature for phonon-mediated superconductivity. Our findings indicate that electronic correlations result in a larger electronic density at the Fermi level, a bigger superconducting gap, and softer vibrational modes associated with hydrogen atoms. Together, the inclusion of these correlation signatures increases the Migdal-Eliashberg superconducting critical temperature from 47 K to 96 K, close to the measured 95 K. Our results reconcile experimental observations and theoretical predictions for CeH9 and herald a path towards the quantitative modeling of phonon-mediated superconductivity for interacting electron systems.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"21 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968818","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-14DOI: 10.1038/s41524-025-01932-8
Lilac Macmillan, Eduardo Costa Girão, Vincent Meunier
We introduce the embedding tensor Φ, a rank-3, purely integer tensor that elevates faces and edges to the same footing as vertices to provide a unique, coordinate-free representation of any three-connected two-dimensional carbon lattice. Defined by the incidence of vertices, edges, and polygonal faces, Φ obeys simple summation rules derived from Euler characteristics. Casting Φ into a flag graph enables exact, tolerance-free identification of wallpaper symmetries. Building on this algebraic framework, we develop an iterative add-dimer search that generates all structures with NF faces from all crystals with NF − 1 faces, while automatically discarding duplicates via tensor isomorphism and spotting non-primitive cells through symmetry checks. Exploiting symmetry keeps the combinatorial growth of dimer insertions tractable even for NF > 5, transforming an otherwise exponential search into a practically feasible approach for high-throughput exploration. Once candidate topologies are enumerated, approximate real-space coordinates and lattice vectors can be reconstructed analytically from Φ and sparse crossing matrices, providing initial geometries for electronic or vibrational calculations. The method delivers an end-to-end pipeline from exhaustive, symmetry-aware enumeration to metadata tagging and coordinate generation, while requiring only integer arithmetic.
{"title":"Graph embedding tensor: unifying topological description, symmetry detection, and structure generation for two-dimensional carbon allotropes","authors":"Lilac Macmillan, Eduardo Costa Girão, Vincent Meunier","doi":"10.1038/s41524-025-01932-8","DOIUrl":"https://doi.org/10.1038/s41524-025-01932-8","url":null,"abstract":"We introduce the embedding tensor Φ, a rank-3, purely integer tensor that elevates faces and edges to the same footing as vertices to provide a unique, coordinate-free representation of any three-connected two-dimensional carbon lattice. Defined by the incidence of vertices, edges, and polygonal faces, Φ obeys simple summation rules derived from Euler characteristics. Casting Φ into a flag graph enables exact, tolerance-free identification of wallpaper symmetries. Building on this algebraic framework, we develop an iterative add-dimer search that generates all structures with NF faces from all crystals with NF − 1 faces, while automatically discarding duplicates via tensor isomorphism and spotting non-primitive cells through symmetry checks. Exploiting symmetry keeps the combinatorial growth of dimer insertions tractable even for NF > 5, transforming an otherwise exponential search into a practically feasible approach for high-throughput exploration. Once candidate topologies are enumerated, approximate real-space coordinates and lattice vectors can be reconstructed analytically from Φ and sparse crossing matrices, providing initial geometries for electronic or vibrational calculations. The method delivers an end-to-end pipeline from exhaustive, symmetry-aware enumeration to metadata tagging and coordinate generation, while requiring only integer arithmetic.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"56 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968821","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
High-entropy layered oxides are promising sodium-ion battery (SIB) cathodes, yet the fundamental role of conformational entropy in stacking phase preference remains unclear. Here, we combine density functional theory (DFT), ab initio molecular dynamics (AIMD), and a fine-tuned CHGNet machine-learning interatomic potential (MLIP) to investigate representative high-entropy (Na0.8Ni0.2Fe0.2Co0.2Mn0.2Ti0.2O2) and low-entropy (Na0.8Mn0.6Co0.4O2) layered oxides in both O3 and P2 phases. A three-stage Monte Carlo sampling strategy was developed to explore transition-metal arrangements, Na/vacancy distributions, and representative low-energy conformations. The fine-tuned CHGNet achieved near-DFT accuracy while enabling large-scale sampling at orders of magnitude lower cost. Our analyses reveal that high-entropy oxides exhibit stronger Na–TMO2 interactions, broader O–TM bond length distributions, and smaller interlayer distance ratios compared with their low-entropy counterparts. These structural features favor O3-phase stabilization in cases where conventional ionic-potential descriptors are insufficient to clearly distinguish between O3- and P2-type layered oxides. Bond-length analyses further indicate that Jahn–Teller distortions in Mn are mitigated in high-entropy oxides, contributing to enhanced structural stability. This study establishes conformational entropy as a decisive factor, alongside Na ionic and cationic potentials, in governing stacking phase stability, and highlights the power of MLIP-accelerated modeling for exploring high-entropy materials and guiding the rational design of next-generation SIB cathodes.
{"title":"Probing entropic control of stacking phase preference in layered oxide cathodes for sodium-ion batteries via machine-learning potentials","authors":"Liang-Ting Wu, Zhong-Lun Li, Shih-Ying Yen, Payam Kaghazchi, Jyh-Chiang Jiang","doi":"10.1038/s41524-025-01954-2","DOIUrl":"https://doi.org/10.1038/s41524-025-01954-2","url":null,"abstract":"High-entropy layered oxides are promising sodium-ion battery (SIB) cathodes, yet the fundamental role of conformational entropy in stacking phase preference remains unclear. Here, we combine density functional theory (DFT), ab initio molecular dynamics (AIMD), and a fine-tuned CHGNet machine-learning interatomic potential (MLIP) to investigate representative high-entropy (Na0.8Ni0.2Fe0.2Co0.2Mn0.2Ti0.2O2) and low-entropy (Na0.8Mn0.6Co0.4O2) layered oxides in both O3 and P2 phases. A three-stage Monte Carlo sampling strategy was developed to explore transition-metal arrangements, Na/vacancy distributions, and representative low-energy conformations. The fine-tuned CHGNet achieved near-DFT accuracy while enabling large-scale sampling at orders of magnitude lower cost. Our analyses reveal that high-entropy oxides exhibit stronger Na–TMO2 interactions, broader O–TM bond length distributions, and smaller interlayer distance ratios compared with their low-entropy counterparts. These structural features favor O3-phase stabilization in cases where conventional ionic-potential descriptors are insufficient to clearly distinguish between O3- and P2-type layered oxides. Bond-length analyses further indicate that Jahn–Teller distortions in Mn are mitigated in high-entropy oxides, contributing to enhanced structural stability. This study establishes conformational entropy as a decisive factor, alongside Na ionic and cationic potentials, in governing stacking phase stability, and highlights the power of MLIP-accelerated modeling for exploring high-entropy materials and guiding the rational design of next-generation SIB cathodes.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"81 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968819","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-13DOI: 10.1038/s41524-025-01928-4
Alberto Carta, Anwesha Panda, Claude Ederer
Calculations combining density functional theory (DFT) and dynamical mean-field theory (DMFT) for transition metal (TM) oxides and similar compounds usually focus on improving the description of the TM d states. Here, we emphasize the importance of also accounting for corrections of the ligand p states. We demonstrate that focusing exclusively on improving the description of the TM d states results in difficulties to obtain the correct insulating behavior for a variety of materials, and requires to use values for the local interaction parameters that are inconsistent with values obtained using, e.g., the constrained random phase approximation (cRPA). We demonstrate that, to a large part, these inconsistencies arise from using local/semi-local DFT as starting point for computing interaction parameters, and we show that applying a simple empirical correction to the low energy states not included in the correlated subspace results in improved values for the interaction parameters that then allow to obtain the correct insulating behavior. Moreover, we show that applying an approximate but realistic Hartree-Fock-like correction specifically to the O p orbitals, when they are explicitly included in the DMFT subspace, significantly improves the quantitative accuracy of the DFT+DMFT description for prototypical Mott insulators, including LaTiO3, LaVO3, and the perovskite rare-earth nickelates (RNiO3).
{"title":"Importance of ligand on-site interactions for the description of Mott-insulators in DFT+DMFT","authors":"Alberto Carta, Anwesha Panda, Claude Ederer","doi":"10.1038/s41524-025-01928-4","DOIUrl":"https://doi.org/10.1038/s41524-025-01928-4","url":null,"abstract":"Calculations combining density functional theory (DFT) and dynamical mean-field theory (DMFT) for transition metal (TM) oxides and similar compounds usually focus on improving the description of the TM d states. Here, we emphasize the importance of also accounting for corrections of the ligand p states. We demonstrate that focusing exclusively on improving the description of the TM d states results in difficulties to obtain the correct insulating behavior for a variety of materials, and requires to use values for the local interaction parameters that are inconsistent with values obtained using, e.g., the constrained random phase approximation (cRPA). We demonstrate that, to a large part, these inconsistencies arise from using local/semi-local DFT as starting point for computing interaction parameters, and we show that applying a simple empirical correction to the low energy states not included in the correlated subspace results in improved values for the interaction parameters that then allow to obtain the correct insulating behavior. Moreover, we show that applying an approximate but realistic Hartree-Fock-like correction specifically to the O p orbitals, when they are explicitly included in the DMFT subspace, significantly improves the quantitative accuracy of the DFT+DMFT description for prototypical Mott insulators, including LaTiO3, LaVO3, and the perovskite rare-earth nickelates (RNiO3).","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"21 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145956359","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-13DOI: 10.1038/s41524-025-01939-1
William Thornley, Sam Sullivan-Allsop, Rongsheng Cai, Nick Clark, Roman Gorbachev, Sarah J. Haigh
The Noise2Void technique is demonstrated for successful denoising of atomic resolution scanning transmission electron microscopy (STEM) images. The technique is applied to denoising atomic resolution images and videos of gold adatoms on a graphene surface within a graphene liquid-cell, with the denoised experimental data qualitatively demonstrating improved visibility of both the Au adatoms and the graphene lattice. The denoising performance is quantified by comparison to similar simulated data and the approach is found to significantly outperform both total variation and simple Gaussian blurring. Compared to other denoising methods, the Noise2Void technique has the combined advantages that it requires no manual intervention during training or denoising, no prior knowledge of the sample and is compatible with real-time data acquisition rates of at least 45 frames per second.
{"title":"Noise2Void for denoising atomic resolution scanning transmission electron microscopy images","authors":"William Thornley, Sam Sullivan-Allsop, Rongsheng Cai, Nick Clark, Roman Gorbachev, Sarah J. Haigh","doi":"10.1038/s41524-025-01939-1","DOIUrl":"https://doi.org/10.1038/s41524-025-01939-1","url":null,"abstract":"The Noise2Void technique is demonstrated for successful denoising of atomic resolution scanning transmission electron microscopy (STEM) images. The technique is applied to denoising atomic resolution images and videos of gold adatoms on a graphene surface within a graphene liquid-cell, with the denoised experimental data qualitatively demonstrating improved visibility of both the Au adatoms and the graphene lattice. The denoising performance is quantified by comparison to similar simulated data and the approach is found to significantly outperform both total variation and simple Gaussian blurring. Compared to other denoising methods, the Noise2Void technique has the combined advantages that it requires no manual intervention during training or denoising, no prior knowledge of the sample and is compatible with real-time data acquisition rates of at least 45 frames per second.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"52 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145956349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-13DOI: 10.1038/s41524-025-01952-4
Bilvin Varughese, Troy D. Loeffler, Suvo Banik, Aditya Koneru, Sukriti Manna, Karthik Balasubramanian, Rohit Batra, Mathew J. Cherukara, Orcun Yildiz, Tom Peterka, Bobby G. Sumpter, Subramanian K.R.S. Sankaranarayanan
The development of next-generation molecular simulation models requires moving beyond predefined functional forms toward machine learning (ML) techniques that directly capture multiscale physics. Here, we demonstrate such an approach using symbolic regression (SR) with equation learner networks and a reinforcement learning search engine to derive interpretable equations for interatomic interactions. Training data were generated through nested ensemble sampling with density functional theory (DFT) energetics, spanning crystalline to highly disordered states. The optimization of the learner network employed continuous-action Monte Carlo Tree Search (MCTS) combined with gradient descent, enabling efficient exploration of function space. For copper as a representative transition metal, an unconstrained search produced models that outperformed fixed-form Sutton–Chen EAM potentials. The SR-derived models (SR1 and SR2) reproduced key material properties—lattice constants, cohesive energies, equations of state, elastic constants, phonon dispersion, defect formation energies, surface/bulk energetics, and phase transformation with significantly improved accuracy. Furthermore, stringent melting simulations using two-phase solid-amorphous interfaces confirmed that SR models accurately capture the interplay of vibrational entropy, cohesive energy, and structural dynamics, surpassing SC-EAM in both qualitative and quantitative predictions. This highlights the potential of SR to deliver fast, accurate, flexible, and physically meaningful potentials, advancing predictive modeling across scales.
{"title":"Physically interpretable interatomic potentials via symbolic regression and reinforcement learning","authors":"Bilvin Varughese, Troy D. Loeffler, Suvo Banik, Aditya Koneru, Sukriti Manna, Karthik Balasubramanian, Rohit Batra, Mathew J. Cherukara, Orcun Yildiz, Tom Peterka, Bobby G. Sumpter, Subramanian K.R.S. Sankaranarayanan","doi":"10.1038/s41524-025-01952-4","DOIUrl":"https://doi.org/10.1038/s41524-025-01952-4","url":null,"abstract":"The development of next-generation molecular simulation models requires moving beyond predefined functional forms toward machine learning (ML) techniques that directly capture multiscale physics. Here, we demonstrate such an approach using symbolic regression (SR) with equation learner networks and a reinforcement learning search engine to derive interpretable equations for interatomic interactions. Training data were generated through nested ensemble sampling with density functional theory (DFT) energetics, spanning crystalline to highly disordered states. The optimization of the learner network employed continuous-action Monte Carlo Tree Search (MCTS) combined with gradient descent, enabling efficient exploration of function space. For copper as a representative transition metal, an unconstrained search produced models that outperformed fixed-form Sutton–Chen EAM potentials. The SR-derived models (SR1 and SR2) reproduced key material properties—lattice constants, cohesive energies, equations of state, elastic constants, phonon dispersion, defect formation energies, surface/bulk energetics, and phase transformation with significantly improved accuracy. Furthermore, stringent melting simulations using two-phase solid-amorphous interfaces confirmed that SR models accurately capture the interplay of vibrational entropy, cohesive energy, and structural dynamics, surpassing SC-EAM in both qualitative and quantitative predictions. This highlights the potential of SR to deliver fast, accurate, flexible, and physically meaningful potentials, advancing predictive modeling across scales.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"52 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145956350","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this article, we present a first-principles field-effect transistors (FETs) contact study based on density functional theory and the non-equilibrium Green’s function method. We estimate device performance for three transition-metal-dichalcogenide (TMD) channel materials (WSe 2 , WS 2 , and MoS 2 ), including metal contacts (Ni) at source and drain for the first time. The results show that the variation in Rc has less impact on ION and IOFF at a given V DD than the variation in subthreshold swing ( SS ; with differences exceeding 30 mV/dec), suggesting SS may be more sensitive to the contacting material choice than previously realized at gate lengths below 15 nm. Among the channel and contact material combinations studied, Ni/WSe 2 FET leads to the best short-channel device performance. The quantum transport calculation shows the highest density of charge accumulation at the Ni/WSe 2 contact edge. Inspired by this first-principles study, we performed X-ray photoelectron spectroscopy and verified the bonding strength at the Ni/WSe 2 contact to be stronger than Ni/WS 2 and Ni/MoS 2 contacts. This supports the theoretical finding that the contact/channel materials need to be chosen to optimize SS and ION in short-channel TMD FETs.
{"title":"Investigating contact-limited scaling in sub-15-nm TMD FETs from first-principles","authors":"Kuan-Bo Lin, Hui-Ting Liu, Shin-Yuan Wang, Shu-Jui Chang, Chao-Cheng Kaun, Chenming Hu","doi":"10.1038/s41524-025-01947-1","DOIUrl":"https://doi.org/10.1038/s41524-025-01947-1","url":null,"abstract":"In this article, we present a first-principles field-effect transistors (FETs) contact study based on density functional theory and the non-equilibrium Green’s function method. We estimate device performance for three transition-metal-dichalcogenide (TMD) channel materials (WSe <jats:sub>2</jats:sub> , WS <jats:sub>2</jats:sub> , and MoS <jats:sub>2</jats:sub> ), including metal contacts (Ni) at source and drain for the first time. The results show that the variation in <jats:italic>R</jats:italic> <jats:sub> <jats:italic>c</jats:italic> </jats:sub> has less impact on <jats:italic>I</jats:italic> <jats:sub> <jats:italic>ON</jats:italic> </jats:sub> and <jats:italic>I</jats:italic> <jats:sub> <jats:italic>OFF</jats:italic> </jats:sub> at a given V <jats:sub>DD</jats:sub> than the variation in subthreshold swing ( <jats:italic>SS</jats:italic> ; with differences exceeding 30 mV/dec), suggesting <jats:italic>SS</jats:italic> may be more sensitive to the contacting material choice than previously realized at gate lengths below 15 nm. Among the channel and contact material combinations studied, Ni/WSe <jats:sub>2</jats:sub> FET leads to the best short-channel device performance. The quantum transport calculation shows the highest density of charge accumulation at the Ni/WSe <jats:sub>2</jats:sub> contact edge. Inspired by this first-principles study, we performed X-ray photoelectron spectroscopy and verified the bonding strength at the Ni/WSe <jats:sub>2</jats:sub> contact to be stronger than Ni/WS <jats:sub>2</jats:sub> and Ni/MoS <jats:sub>2</jats:sub> contacts. This supports the theoretical finding that the contact/channel materials need to be chosen to optimize <jats:italic>SS</jats:italic> and <jats:italic>I</jats:italic> <jats:sub> <jats:italic>ON</jats:italic> </jats:sub> in short-channel TMD FETs.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"47 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938273","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：