Determining the number of constituent atoms in metallic nanoclusters (NCs) directly from imaging is key to understanding how atomicity governs their size-dependent properties. Scanning transmission electron microscopy (STEM), which captures real-space images of materials with tunable magnification down to the atomic scale, provides an invaluable means to probe such structures. However, despite these advantages, automated and accurate identification of NC atomicity remains challenging, requiring robust extraction of features such as projected shape and contrast distribution from imaging data. To address this challenge, we present a deep learning framework that classifies platinum NCs (Ptn; n = 19, 30, 41, 55, 70) using high-resolution aberration-corrected STEM images. A convolutional neural network extracts structural features that are separable in UMAP (Uniform Manifold Approximation and Projection) space, with class-specific focus visualized using Grad-CAM (Gradient-weighted Class Activation Mapping). The model achieves high accuracy, even for mixed-atomicity samples (n = 19, 41, 70) on a shared substrate. To address domain shift, we apply fine-tuning with high-confidence pseudo-labels, significantly recovering performance. A dual-channel model integrating Local Contrast Normalization (LCN) filtering achieves a coefficient of determination of R² = 0.94 ± 0.03, outperforming size-based classification. This framework automates atomic-scale classification from STEM images and advances autonomous workflows via real-time analysis and machine learning based decisions.
直接从成像中确定金属纳米团簇（nc）中组成原子的数量是理解原子性如何控制其尺寸依赖特性的关键。扫描透射电子显微镜（STEM）可以捕获材料的真实空间图像，其放大倍数可调至原子尺度，为探测此类结构提供了宝贵的手段。然而，尽管有这些优势，自动准确识别NC原子性仍然具有挑战性，需要从成像数据中提取投影形状和对比度分布等特征。为了应对这一挑战，我们提出了一个深度学习框架，该框架使用高分辨率像差校正的STEM图像对白金nc （Ptn; n = 19、30、41、55、70）进行分类。卷积神经网络提取UMAP（均匀流形逼近和投影）空间中可分离的结构特征，使用梯度加权类激活映射（Gradient-weighted Class Activation Mapping）将特定类的焦点可视化。即使对于共享衬底上的混合原子性样品（n = 19,41,70），该模型也具有很高的精度。为了解决领域转移问题，我们使用高置信度伪标签进行微调，显著恢复了性能。结合局部对比度归一化（LCN）滤波的双通道模型的决定系数R²= 0.94±0.03，优于基于尺寸的分类。该框架可以从STEM图像中自动进行原子尺度的分类，并通过实时分析和基于机器学习的决策推进自主工作流程。
{"title":"Interpretable deep learning for atomicity classification of platinum nanoclusters in STEM images","authors":"Keizo Tsukamoto, Naoyuki Hirata, Masahide Tona, Atsushi Nakajima","doi":"10.1038/s41524-026-02014-z","DOIUrl":"https://doi.org/10.1038/s41524-026-02014-z","url":null,"abstract":"Determining the number of constituent atoms in metallic nanoclusters (NCs) directly from imaging is key to understanding how atomicity governs their size-dependent properties. Scanning transmission electron microscopy (STEM), which captures real-space images of materials with tunable magnification down to the atomic scale, provides an invaluable means to probe such structures. However, despite these advantages, automated and accurate identification of NC atomicity remains challenging, requiring robust extraction of features such as projected shape and contrast distribution from imaging data. To address this challenge, we present a deep learning framework that classifies platinum NCs (Ptn; n = 19, 30, 41, 55, 70) using high-resolution aberration-corrected STEM images. A convolutional neural network extracts structural features that are separable in UMAP (Uniform Manifold Approximation and Projection) space, with class-specific focus visualized using Grad-CAM (Gradient-weighted Class Activation Mapping). The model achieves high accuracy, even for mixed-atomicity samples (n = 19, 41, 70) on a shared substrate. To address domain shift, we apply fine-tuning with high-confidence pseudo-labels, significantly recovering performance. A dual-channel model integrating Local Contrast Normalization (LCN) filtering achieves a coefficient of determination of R² = 0.94 ± 0.03, outperforming size-based classification. This framework automates atomic-scale classification from STEM images and advances autonomous workflows via real-time analysis and machine learning based decisions.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"13 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147320184","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 : 2026-02-26DOI: 10.1038/s41524-025-01887-w
Ning Mao, Cheng Xu, Ting Bao, Nikolai Peshcherenko, Claudia Felser, Yang Zhang
While moiré phenomena have been extensively studied in low-carrier-density systems such as semiconductors, their implications for metallic systems with large Fermi surfaces remain largely unexplored. Using GPU-accelerated large-scale ab-initio quantum transport simulations, we investigate spin transport in two distinct platforms: twisted bilayer MoTe2 (semiconductor) and NbX2 (X = S, Se; metals). In twisted MoTe2, the spin Hall conductivity (SHC) evolves from (4frac{e}{4pi }) at 5.09° to (10frac{e}{4pi }) at 1.89°. Remarkably, in heavily doped metallic regimes where isolated Chern bands are absent, we observe a universal amplification of the SHC arising from Fermi surface reconstruction under a long-wavelength potential, with the peak SHC tripling from (6frac{e}{4pi }) at 5.09° to (17frac{e}{4pi }) at 3.89°. For moiré metals like twisted NbX2, we identify a record SHC of −5200 (ℏ/e)S/cm, surpassing all known bulk materials.
{"title":"Universal giant spin Hall effect in moiré metal","authors":"Ning Mao, Cheng Xu, Ting Bao, Nikolai Peshcherenko, Claudia Felser, Yang Zhang","doi":"10.1038/s41524-025-01887-w","DOIUrl":"https://doi.org/10.1038/s41524-025-01887-w","url":null,"abstract":"While moiré phenomena have been extensively studied in low-carrier-density systems such as semiconductors, their implications for metallic systems with large Fermi surfaces remain largely unexplored. Using GPU-accelerated large-scale ab-initio quantum transport simulations, we investigate spin transport in two distinct platforms: twisted bilayer MoTe2 (semiconductor) and NbX2 (X = S, Se; metals). In twisted MoTe2, the spin Hall conductivity (SHC) evolves from (4frac{e}{4pi }) at 5.09° to (10frac{e}{4pi }) at 1.89°. Remarkably, in heavily doped metallic regimes where isolated Chern bands are absent, we observe a universal amplification of the SHC arising from Fermi surface reconstruction under a long-wavelength potential, with the peak SHC tripling from (6frac{e}{4pi }) at 5.09° to (17frac{e}{4pi }) at 3.89°. For moiré metals like twisted NbX2, we identify a record SHC of −5200 (ℏ/e)S/cm, surpassing all known bulk materials.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"50 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147320185","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 : 2026-02-21DOI: 10.1038/s41524-026-02004-1
Jianpei Xing, Ying Zhao, Li Sun, Ming Xu, Xue Jiang, Jijun Zhao
{"title":"Coexistence of magnetism and ferroelectricity in the 2D inorganic molecular crystal SbI3•(S7N)3","authors":"Jianpei Xing, Ying Zhao, Li Sun, Ming Xu, Xue Jiang, Jijun Zhao","doi":"10.1038/s41524-026-02004-1","DOIUrl":"https://doi.org/10.1038/s41524-026-02004-1","url":null,"abstract":"","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"30 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146260843","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 : 2026-02-21DOI: 10.1038/s41524-026-02016-x
Tae Jung Moon, Seong Min Park, Han Gyu Yoon, Hee Young Kwon, Changyeon Won
Frustrated magnetic systems such as spin ice are key platforms for novel metamaterials. However, identifying their ground states in finite arrays is a formidable challenge, as boundary sensitivity and metastable states trap conventional optimization methods. We introduce a virtuous-cycle AI pipeline where a genetic algorithm explores the latent space of a variational autoencoder (VAE), with the best candidates progressively refining the VAE’s representation. Applied to Kagome spin ice, this method reveals how the boundary magnetism is determined: boundaries break the symmetry of the $$sqrt{3,}times sqrt{3,}$$3×3 magnetic superstructure while the bulk superstructure order in the interior maintains. Furthermore, it demonstrates that high geometric confinement induces a novel quasi-ferromagnetic phase, which breaks the interior superstructure order. Our work provides a predictive framework for designing frustrated materials and demonstrates a powerful AI approach for boundary-sensitive physical systems.
{"title":"Boundary sensitivity in finite-sized artificial spin ice explored via AI-assisted genetic algorithms","authors":"Tae Jung Moon, Seong Min Park, Han Gyu Yoon, Hee Young Kwon, Changyeon Won","doi":"10.1038/s41524-026-02016-x","DOIUrl":"https://doi.org/10.1038/s41524-026-02016-x","url":null,"abstract":"Frustrated magnetic systems such as spin ice are key platforms for novel metamaterials. However, identifying their ground states in finite arrays is a formidable challenge, as boundary sensitivity and metastable states trap conventional optimization methods. We introduce a virtuous-cycle AI pipeline where a genetic algorithm explores the latent space of a variational autoencoder (VAE), with the best candidates progressively refining the VAE’s representation. Applied to Kagome spin ice, this method reveals how the boundary magnetism is determined: boundaries break the symmetry of the <jats:inline-formula> <jats:alternatives> <jats:tex-math>$$sqrt{3,}times sqrt{3,}$$</jats:tex-math> <mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"> <mml:mrow> <mml:msqrt> <mml:mrow> <mml:mn>3</mml:mn> <mml:mspace/> </mml:mrow> </mml:msqrt> <mml:mo>×</mml:mo> <mml:msqrt> <mml:mrow> <mml:mn>3</mml:mn> <mml:mspace/> </mml:mrow> </mml:msqrt> </mml:mrow> </mml:math> </jats:alternatives> </jats:inline-formula> magnetic superstructure while the bulk superstructure order in the interior maintains. Furthermore, it demonstrates that high geometric confinement induces a novel quasi-ferromagnetic phase, which breaks the interior superstructure order. Our work provides a predictive framework for designing frustrated materials and demonstrates a powerful AI approach for boundary-sensitive physical systems.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"10 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146260842","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 : 2026-02-20DOI: 10.1038/s41524-026-01994-2
Rongfeng Guan, Ang Liu, Yang Song
{"title":"Deep learning generative model for conditional crystal structure prediction of sodium amide","authors":"Rongfeng Guan, Ang Liu, Yang Song","doi":"10.1038/s41524-026-01994-2","DOIUrl":"https://doi.org/10.1038/s41524-026-01994-2","url":null,"abstract":"","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"322 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223036","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 : 2026-02-19DOI: 10.1038/s41524-026-02008-x
Yuhao Luo, Tianyi Wang, Zhihao Huang, Yanqing Su, Shuozhi Xu, Peter K. Liaw, Xiang-Guo Li
{"title":"Dislocation-induced ordering as a source of strengthening in refractory multi-principal element alloys","authors":"Yuhao Luo, Tianyi Wang, Zhihao Huang, Yanqing Su, Shuozhi Xu, Peter K. Liaw, Xiang-Guo Li","doi":"10.1038/s41524-026-02008-x","DOIUrl":"https://doi.org/10.1038/s41524-026-02008-x","url":null,"abstract":"","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"1 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223028","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 : 2026-02-19DOI: 10.1038/s41524-026-02013-0
Enda Xiao, Terumasa Tadano
We present a machine learning-accelerated high-throughput (HTP) workflow for the discovery of functional materials. As a test case, quaternary and all- d Heusler compounds were screened for stable compounds with large magnetocrystalline anisotropy energy ( Eaniso ). Structure optimization and evaluation of formation energy and energy above the convex hull were performed using the eSEN-30M-OAM interatomic potential, while local magnetic moments, phonon stability, magnetic stability, and Eaniso were predicted by eSEN models trained on our DxMag Heusler database. A frozen transfer learning strategy was employed to improve accuracy. Candidate compounds identified by the ML-HTP workflow were validated with density functional theory, confirming high predictive precision. We also benchmark the performance of different uMLIPs, discuss the fidelity of local magnetic moment prediction, and demonstrate generalization to unseen elements via transfer learning from a universal interatomic potential.
{"title":"Accurate screening of functional materials with machine-learning potential and transfer-learned regressions: Heusler alloy benchmark","authors":"Enda Xiao, Terumasa Tadano","doi":"10.1038/s41524-026-02013-0","DOIUrl":"https://doi.org/10.1038/s41524-026-02013-0","url":null,"abstract":"We present a machine learning-accelerated high-throughput (HTP) workflow for the discovery of functional materials. As a test case, quaternary and all- <jats:italic>d</jats:italic> Heusler compounds were screened for stable compounds with large magnetocrystalline anisotropy energy ( <jats:italic>E</jats:italic> <jats:sub>aniso</jats:sub> ). Structure optimization and evaluation of formation energy and energy above the convex hull were performed using the eSEN-30M-OAM interatomic potential, while local magnetic moments, phonon stability, magnetic stability, and <jats:italic>E</jats:italic> <jats:sub>aniso</jats:sub> were predicted by eSEN models trained on our DxMag Heusler database. A frozen transfer learning strategy was employed to improve accuracy. Candidate compounds identified by the ML-HTP workflow were validated with density functional theory, confirming high predictive precision. We also benchmark the performance of different uMLIPs, discuss the fidelity of local magnetic moment prediction, and demonstrate generalization to unseen elements via transfer learning from a universal interatomic potential.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"5 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223029","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 : 2026-02-19DOI: 10.1038/s41524-026-02002-3
Jonas Scheunert, Shamail Ahmed, Thomas Demuth, Andreas Beyer, Sebastian Wissel, Bai-Xiang Xu, Kerstin Volz
Polycrystalline materials have numerous applications due to their unique properties, which are often determined by the grain orientation relationships. Hence, quantitative characterization of grain as well as interface orientation is essential to optimize these materials, particularly energy materials. Using scanning transmission electron microscopy (TEM), materials can be analysed in an extremely fine grid of scan points via electron diffraction patterns at each scan point. By matching the diffraction patterns to a simulated database, the crystal orientation of a grain at each scan point can be determined. In this work, we train convolutional neural networks on dynamically simulated diffraction patterns of LiNiO 2 , an important cathode-active material for lithium-ion batteries, to predict the orientation of grains in terms of three Euler angles for the complete fundamental orientation region. Results demonstrate that these networks outperform the conventional pattern-matching algorithm with increased accuracy and efficiency. The former can be attributed to the fact that these models are trained by data incorporating dynamical effects. This work is an attempt to apply deep learning for the analysis of TEM data to determine the grain orientation and enlighten the great potential of machine learning to accelerate the analysis of electron microscopy data, toward a high-throughput characterization technique.
{"title":"Determining the grain orientations of battery materials from electron diffraction patterns using convolutional neural networks","authors":"Jonas Scheunert, Shamail Ahmed, Thomas Demuth, Andreas Beyer, Sebastian Wissel, Bai-Xiang Xu, Kerstin Volz","doi":"10.1038/s41524-026-02002-3","DOIUrl":"https://doi.org/10.1038/s41524-026-02002-3","url":null,"abstract":"Polycrystalline materials have numerous applications due to their unique properties, which are often determined by the grain orientation relationships. Hence, quantitative characterization of grain as well as interface orientation is essential to optimize these materials, particularly energy materials. Using scanning transmission electron microscopy (TEM), materials can be analysed in an extremely fine grid of scan points via electron diffraction patterns at each scan point. By matching the diffraction patterns to a simulated database, the crystal orientation of a grain at each scan point can be determined. In this work, we train convolutional neural networks on dynamically simulated diffraction patterns of LiNiO <jats:sub>2</jats:sub> , an important cathode-active material for lithium-ion batteries, to predict the orientation of grains in terms of three Euler angles for the complete fundamental orientation region. Results demonstrate that these networks outperform the conventional pattern-matching algorithm with increased accuracy and efficiency. The former can be attributed to the fact that these models are trained by data incorporating dynamical effects. This work is an attempt to apply deep learning for the analysis of TEM data to determine the grain orientation and enlighten the great potential of machine learning to accelerate the analysis of electron microscopy data, toward a high-throughput characterization technique.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"7 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2026-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223025","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：