Pub Date : 2025-12-11DOI: 10.1038/s41524-025-01850-9
Shimanta Das, Noah Oyeniran, Joshua Walter, Aidan Gesch, Chongze Hu
The concurrent segregation of multiple solute elements at grain boundaries (GBs), also known as co-segregation, is a pervasive interfacial behavior that governs microstructural evolution and influences many properties of high-entropy alloys (HEAs). However, accurately predicting co-segregation behavior in HEAs is a challenging task due to the vast compositional space and complex interactions among multiple solute elements. In this paper, we developed a scalarization-based Bayesian optimization (SBO) framework integrated with high-throughput atomistic simulations to efficiently explore and optimize the large compositional space of CrMnFeCoNi HEAs for targeted co-segregation behavior and other desirable interfacial properties. Specifically, Thompson sampling is adopted to explore the input compositional space and identify HEA candidates representing two extremes: the strongest and weakest co-segregation of Cr and Mn at CrMnFeCoNi GBs. These SBO-predicted segregation extremes are subsequently validated by hybrid molecular dynamics/Monte Carlo simulations and first-principles calculations. Furthermore, electronic structure calculations demonstrate that the co-segregation of Cr and Mn can be ascribed to the hybridization of their d valence electrons promoted by the presence of Fe. While this SBO framework focuses on segregation behavior, it can be easily extended to optimize a wide range of interfacial properties in multicomponent systems. This study establishes a new paradigm for designing advanced HEAs through interfacial property optimization.
{"title":"Bayesian Optimization of Grain-Boundary Segregation in High-Entropy Alloys","authors":"Shimanta Das, Noah Oyeniran, Joshua Walter, Aidan Gesch, Chongze Hu","doi":"10.1038/s41524-025-01850-9","DOIUrl":"https://doi.org/10.1038/s41524-025-01850-9","url":null,"abstract":"The concurrent segregation of multiple solute elements at grain boundaries (GBs), also known as co-segregation, is a pervasive interfacial behavior that governs microstructural evolution and influences many properties of high-entropy alloys (HEAs). However, accurately predicting co-segregation behavior in HEAs is a challenging task due to the vast compositional space and complex interactions among multiple solute elements. In this paper, we developed a scalarization-based Bayesian optimization (SBO) framework integrated with high-throughput atomistic simulations to efficiently explore and optimize the large compositional space of CrMnFeCoNi HEAs for targeted co-segregation behavior and other desirable interfacial properties. Specifically, Thompson sampling is adopted to explore the input compositional space and identify HEA candidates representing two extremes: the strongest and weakest co-segregation of Cr and Mn at CrMnFeCoNi GBs. These SBO-predicted segregation extremes are subsequently validated by hybrid molecular dynamics/Monte Carlo simulations and first-principles calculations. Furthermore, electronic structure calculations demonstrate that the co-segregation of Cr and Mn can be ascribed to the hybridization of their d valence electrons promoted by the presence of Fe. While this SBO framework focuses on segregation behavior, it can be easily extended to optimize a wide range of interfacial properties in multicomponent systems. This study establishes a new paradigm for designing advanced HEAs through interfacial property optimization.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"15 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145718534","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 : 2025-12-11DOI: 10.1038/s41524-025-01872-3
Hendrik Kraß, Ju Huang, Seyed Mohamad Moosavi
Universal machine learning interatomic potentials (uMLIPs) have emerged as powerful tools for accelerating atomistic simulations, offering scalable and efficient modeling with accuracy close to quantum calculations. However, their reliability and effectiveness in practical, real-world applications remain an open question. Metal-organic frameworks (MOFs) and related nanoporous materials are highly porous crystals with critical relevance in carbon capture, energy storage, and catalysis applications. Modeling nanoporous materials presents distinct challenges for uMLIPs due to their diverse chemistry, structural complexity, including porosity and coordination bonds, and the absence from existing training datasets. Here, we introduce MOFSimBench, a benchmark for evaluating uMLIPs on key materials modeling tasks for nanoporous materials, including structural optimization, molecular dynamics (MD) stability, bulk property prediction, and host-guest interactions. Evaluating 20 models from various architectures, we find that top-performing uMLIPs consistently outperform classical force fields and fine-tuned machine learning potentials across all tasks, demonstrating their readiness for deployment in nanoporous materials modeling. Our analysis highlights that data quality plays a more critical role than model architecture in determining performance across all evaluated uMLIPs. We release our modular and extensible benchmarking framework at https://github.com/AI4ChemS/mofsim-bench, providing an open resource to guide the adoption for nanoporous materials modeling and further development of uMLIPs.
{"title":"MOFSimBench: evaluating universal machine learning interatomic potentials in metal-organic framework molecular modeling","authors":"Hendrik Kraß, Ju Huang, Seyed Mohamad Moosavi","doi":"10.1038/s41524-025-01872-3","DOIUrl":"https://doi.org/10.1038/s41524-025-01872-3","url":null,"abstract":"Universal machine learning interatomic potentials (uMLIPs) have emerged as powerful tools for accelerating atomistic simulations, offering scalable and efficient modeling with accuracy close to quantum calculations. However, their reliability and effectiveness in practical, real-world applications remain an open question. Metal-organic frameworks (MOFs) and related nanoporous materials are highly porous crystals with critical relevance in carbon capture, energy storage, and catalysis applications. Modeling nanoporous materials presents distinct challenges for uMLIPs due to their diverse chemistry, structural complexity, including porosity and coordination bonds, and the absence from existing training datasets. Here, we introduce MOFSimBench, a benchmark for evaluating uMLIPs on key materials modeling tasks for nanoporous materials, including structural optimization, molecular dynamics (MD) stability, bulk property prediction, and host-guest interactions. Evaluating 20 models from various architectures, we find that top-performing uMLIPs consistently outperform classical force fields and fine-tuned machine learning potentials across all tasks, demonstrating their readiness for deployment in nanoporous materials modeling. Our analysis highlights that data quality plays a more critical role than model architecture in determining performance across all evaluated uMLIPs. We release our modular and extensible benchmarking framework at https://github.com/AI4ChemS/mofsim-bench, providing an open resource to guide the adoption for nanoporous materials modeling and further development of uMLIPs.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"146 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145718536","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 : 2025-12-11DOI: 10.1038/s41524-025-01907-9
Yan-Xin Guo, Hai-Le Yan, Dong Wang, Ming-Hui Cai, Na Xiao, Nan Jia, Liang Zuo
The influence of 3d transition elements and Al on the intrinsic mechanical properties of CoNiCr was investigated by a combined first-principles calculation and chemical bonding study. All alloying elements tend to reduce elastic moduli, hardness, and ideal tensile strength, with Ti, V, Cu, and Al causing the most pronounced weakening effects while Mn and Fe exerting minor influences. These changes stem from variations in electron band filling and orbital hybridization. Specifically, elements with valence electron concentration (VEC) that are significantly different from CoNiCr destabilize bonding by either depleting bonding states (Ti and V), reducing d-d hybridization and overpopulating antibonding states (Cu), or altering orbital hybridization from d-d to p-d (Al). In contrast, Mn and Fe with comparable VEC preserve the bonding strength. Regarding deformation mechanism, all dopants tend to increase stacking fault energy (γisf). Most alloys exhibit the co-activation of slip and stacking fault, while Cu alloying favors twinning alongside dislocation slip. Notably, the competition between stacking fault and twinning is governed by 1/2γisf. VEC is identified as a critical parameter influencing γisf, with alloys possessing high VEC typically showing larger γisf. These findings establish a theoretical basis for designing high-performance fcc multicomponent alloys through composition-controlled chemical bonding engineering.
{"title":"Chemical bonding dictates alloying effect on inherent mechanical strength and plastic deformation mechanism in CoNiCr multicomponent alloy","authors":"Yan-Xin Guo, Hai-Le Yan, Dong Wang, Ming-Hui Cai, Na Xiao, Nan Jia, Liang Zuo","doi":"10.1038/s41524-025-01907-9","DOIUrl":"https://doi.org/10.1038/s41524-025-01907-9","url":null,"abstract":"The influence of 3d transition elements and Al on the intrinsic mechanical properties of CoNiCr was investigated by a combined first-principles calculation and chemical bonding study. All alloying elements tend to reduce elastic moduli, hardness, and ideal tensile strength, with Ti, V, Cu, and Al causing the most pronounced weakening effects while Mn and Fe exerting minor influences. These changes stem from variations in electron band filling and orbital hybridization. Specifically, elements with valence electron concentration (VEC) that are significantly different from CoNiCr destabilize bonding by either depleting bonding states (Ti and V), reducing d-d hybridization and overpopulating antibonding states (Cu), or altering orbital hybridization from d-d to p-d (Al). In contrast, Mn and Fe with comparable VEC preserve the bonding strength. Regarding deformation mechanism, all dopants tend to increase stacking fault energy (γisf). Most alloys exhibit the co-activation of slip and stacking fault, while Cu alloying favors twinning alongside dislocation slip. Notably, the competition between stacking fault and twinning is governed by 1/2γisf. VEC is identified as a critical parameter influencing γisf, with alloys possessing high VEC typically showing larger γisf. These findings establish a theoretical basis for designing high-performance fcc multicomponent alloys through composition-controlled chemical bonding engineering.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"111 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145718537","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 : 2025-12-10DOI: 10.1038/s41524-025-01901-1
Kai Yang, Daniel Schwalbe-Koda
Generative models show great promise for the inverse design of molecules and inorganic crystals, but remain largely ineffective within more complex structures such as amorphous materials. Here, we present a diffusion model that reliably generates amorphous structures up to 3 orders of magnitude times faster than conventional simulations across processing conditions, compositions, and data sources. Generated structures recovered the short- and medium-range order, sampling diversity, and macroscopic properties of silica glass, as validated by simulations and an information-theoretical strategy. Conditional generation allowed sampling large structures at low cooling rates of 10−2 K/ps to uncover a ductile-to-brittle transition and mesoporous silica structures. Extension to metallic glassy systems accurately reproduced local structures and properties from both computational and experimental datasets, demonstrating how synthetic data can be generated from characterization results. Our methods provide a roadmap for the design and simulation of amorphous materials previously inaccessible to computational methods.
{"title":"A generative diffusion model for amorphous materials","authors":"Kai Yang, Daniel Schwalbe-Koda","doi":"10.1038/s41524-025-01901-1","DOIUrl":"https://doi.org/10.1038/s41524-025-01901-1","url":null,"abstract":"Generative models show great promise for the inverse design of molecules and inorganic crystals, but remain largely ineffective within more complex structures such as amorphous materials. Here, we present a diffusion model that reliably generates amorphous structures up to 3 orders of magnitude times faster than conventional simulations across processing conditions, compositions, and data sources. Generated structures recovered the short- and medium-range order, sampling diversity, and macroscopic properties of silica glass, as validated by simulations and an information-theoretical strategy. Conditional generation allowed sampling large structures at low cooling rates of 10−2 K/ps to uncover a ductile-to-brittle transition and mesoporous silica structures. Extension to metallic glassy systems accurately reproduced local structures and properties from both computational and experimental datasets, demonstrating how synthetic data can be generated from characterization results. Our methods provide a roadmap for the design and simulation of amorphous materials previously inaccessible to computational methods.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"31 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145711580","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 : 2025-12-10DOI: 10.1038/s41524-025-01871-4
Mohammed Al-Fahdi, Riccardo Rurali, Jianjun Hu, Christopher Wolverton, Ming Hu
Designing materials with targeted lattice thermal conductivity (LTC) demands electronic-level insight into chemical bonding. We introduce two bonding descriptors, namely normalized negative integrated COHP (-ICOHP) and normalized integrated COBI, that correlate strongly with LTC and rattling (mean-squared displacement), surpassing empirical rules and the unnormalized −ICOHP across >4500 inorganic crystals by first-principles. We train a crystal attention graph neural network (CATGNN) to predict these descriptors and screen ~200,000 database structures for extreme LTCs. From 367 (533) candidates with low (high) normalized -ICOHP and normalized ICOBI, first-principles validation identifies 106 dynamically stable compounds with LTC < 5 W m−1 K−1 (68% <2 W m−1 K−1) and 13 stable compounds with LTC > 100 W m−1 K−1. The descriptors’ low cost and clear physical meaning provide a rapid, reliable route to high-throughput discovery and inverse design of crystalline materials with ultralow or ultrahigh LTC for applications in thermal insulation, thermoelectrics, and electronics cooling.
设计具有目标晶格热导率（LTC）的材料需要对化学键的电子级洞察力。我们引入了两个键描述符，即归一化负积分COHP （-ICOHP）和归一化积分COBI，它们与LTC和咔嗒（均方位移）密切相关，超越了经验规则和非归一化-ICOHP在>4500无机晶体中的第一性原理。我们训练晶体注意图神经网络（CATGNN）来预测这些描述符，并为极端ltc筛选约200,000个数据库结构。从367（533）个具有低（高）归一化-ICOHP和归一化ICOBI的候选化合物中，第一性原理验证确定了106个LTC为100 W m−1 K−1的动态稳定化合物。描述符的低成本和明确的物理含义为在绝热、热电和电子冷却应用中具有超低或超高LTC的晶体材料的高通量发现和逆向设计提供了快速、可靠的途径。
{"title":"Accelerated discovery of extreme lattice thermal conductivity by crystal graph attention networks and chemical bonding","authors":"Mohammed Al-Fahdi, Riccardo Rurali, Jianjun Hu, Christopher Wolverton, Ming Hu","doi":"10.1038/s41524-025-01871-4","DOIUrl":"https://doi.org/10.1038/s41524-025-01871-4","url":null,"abstract":"Designing materials with targeted lattice thermal conductivity (LTC) demands electronic-level insight into chemical bonding. We introduce two bonding descriptors, namely normalized negative integrated COHP (-ICOHP) and normalized integrated COBI, that correlate strongly with LTC and rattling (mean-squared displacement), surpassing empirical rules and the unnormalized −ICOHP across >4500 inorganic crystals by first-principles. We train a crystal attention graph neural network (CATGNN) to predict these descriptors and screen ~200,000 database structures for extreme LTCs. From 367 (533) candidates with low (high) normalized -ICOHP and normalized ICOBI, first-principles validation identifies 106 dynamically stable compounds with LTC < 5 W m−1 K−1 (68% <2 W m−1 K−1) and 13 stable compounds with LTC > 100 W m−1 K−1. The descriptors’ low cost and clear physical meaning provide a rapid, reliable route to high-throughput discovery and inverse design of crystalline materials with ultralow or ultrahigh LTC for applications in thermal insulation, thermoelectrics, and electronics cooling.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"144 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145711581","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}
Autonomous material exploration systems that integrate robotics, material simulations, and machine learning have advanced rapidly in recent years. Although their number continues to grow, these systems currently operate in isolation, limiting the overall efficiency of autonomous material discovery. In analogy to how human researchers advance materials science by sharing knowledge and collaborating, autonomous systems can also benefit from networking and knowledge exchange. Here, we propose a framework in which multiple autonomous material exploration systems form a network via transfer learning, selectively utilizing relevant knowledge from other systems in real time. We demonstrate this approach using three distinct autonomous systems and show that such networking significantly enhances the efficiency of material discovery. Our results suggest that the proposed framework can enable the development of large-scale autonomous material exploration networks, ultimately accelerating progress in material development.
{"title":"Networking autonomous material exploration systems through transfer learning","authors":"Naoki Yoshida, Yutaro Iwabuchi, Yasuhiko Igarashi, Yuma Iwasaki","doi":"10.1038/s41524-025-01851-8","DOIUrl":"https://doi.org/10.1038/s41524-025-01851-8","url":null,"abstract":"Autonomous material exploration systems that integrate robotics, material simulations, and machine learning have advanced rapidly in recent years. Although their number continues to grow, these systems currently operate in isolation, limiting the overall efficiency of autonomous material discovery. In analogy to how human researchers advance materials science by sharing knowledge and collaborating, autonomous systems can also benefit from networking and knowledge exchange. Here, we propose a framework in which multiple autonomous material exploration systems form a network via transfer learning, selectively utilizing relevant knowledge from other systems in real time. We demonstrate this approach using three distinct autonomous systems and show that such networking significantly enhances the efficiency of material discovery. Our results suggest that the proposed framework can enable the development of large-scale autonomous material exploration networks, ultimately accelerating progress in material development.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"13 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705142","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}
Probing the ideal limit of interfacial thermal conductance (ITC) in two-dimensional (2D) heterointerfaces is of paramount importance for assessing heat dissipation in 2D-based nanoelectronics. Using graphene/hexagonal boron nitride (Gr/h-BN), a structurally isomorphous heterostructure with minimal mass contrast, as a prototype, we develop an accurate yet highly efficient machine-learned potential (MLP) model, which drives nonequilibrium molecular dynamics (NEMD) simulations on a realistically large system with over 300,000 atoms, enabling us to report the ideal limit range of ITC for 2D heterostructures at room temperature. We further unveil an intriguing stacking-sequence-dependent ITC hierarchy in the Gr/h-BN heterostructure, which can be connected to moiré patterns and is likely universal in van der Waals layered materials. The underlying atomic-level mechanisms can be succinctly summarized as energy-favorable stacking sequences facilitating out-of-plane phonon energy transmission. This work demonstrates that MLP-driven MD simulations can serve as a new paradigm for probing and understanding thermal transport mechanisms in 2D heterostructures and other layered materials.
{"title":"Probing the ideal limit of interfacial thermal conductance in two-dimensional van der Waals heterostructures","authors":"Ting Liang, Ke Xu, Penghua Ying, Wenwu Jiang, Meng Han, Xin Wu, Wengen Ouyang, Yimin Yao, Xiaoliang Zeng, Zhenqiang Ye, Zheyong Fan, Jianbin Xu","doi":"10.1038/s41524-025-01885-y","DOIUrl":"https://doi.org/10.1038/s41524-025-01885-y","url":null,"abstract":"Probing the ideal limit of interfacial thermal conductance (ITC) in two-dimensional (2D) heterointerfaces is of paramount importance for assessing heat dissipation in 2D-based nanoelectronics. Using graphene/hexagonal boron nitride (Gr/h-BN), a structurally isomorphous heterostructure with minimal mass contrast, as a prototype, we develop an accurate yet highly efficient machine-learned potential (MLP) model, which drives nonequilibrium molecular dynamics (NEMD) simulations on a realistically large system with over 300,000 atoms, enabling us to report the ideal limit range of ITC for 2D heterostructures at room temperature. We further unveil an intriguing stacking-sequence-dependent ITC hierarchy in the Gr/h-BN heterostructure, which can be connected to moiré patterns and is likely universal in van der Waals layered materials. The underlying atomic-level mechanisms can be succinctly summarized as energy-favorable stacking sequences facilitating out-of-plane phonon energy transmission. This work demonstrates that MLP-driven MD simulations can serve as a new paradigm for probing and understanding thermal transport mechanisms in 2D heterostructures and other layered materials.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"20 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705143","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 : 2025-12-08DOI: 10.1038/s41524-025-01896-9
Harikrishna Sahu, Akhlak Mahmood, Labeeba B. Shafique, Rampi Ramprasad
Advances in machine learning have transformed materials discovery, yet challenges remain due to the lack of informatics-ready data and the complexity of numerical descriptors. Scientific knowledge is scattered across publications, making comprehensive data extraction difficult. This study presents a large language model (LLM)-driven framework to accelerate organic solar cell (OSC) materials discovery by extracting structured data from literature and predicting device performance using natural language embeddings. Trained on a curated dataset of 422 OSC devices, the fine-tuned LLM demonstrated strong predictive accuracy across key performance metrics: power conversion efficiency (PCE, R2: 0.87), short-circuit current (JSC, R2: 0.82), open-circuit voltage (VOC, R2: 0.89), and fill factor (FF, R2: 0.59). The models are then used to explore the space of 1.4 million combinations of materials, experimental variables and device architectures. The analysis provides data-driven design guidelines, identifying optimal donor-acceptor combinations and processing conditions that consistently yield higher device performance.
{"title":"From Corpus to Innovation: Advancing Organic Solar Cell Design with Large Language Models","authors":"Harikrishna Sahu, Akhlak Mahmood, Labeeba B. Shafique, Rampi Ramprasad","doi":"10.1038/s41524-025-01896-9","DOIUrl":"https://doi.org/10.1038/s41524-025-01896-9","url":null,"abstract":"Advances in machine learning have transformed materials discovery, yet challenges remain due to the lack of informatics-ready data and the complexity of numerical descriptors. Scientific knowledge is scattered across publications, making comprehensive data extraction difficult. This study presents a large language model (LLM)-driven framework to accelerate organic solar cell (OSC) materials discovery by extracting structured data from literature and predicting device performance using natural language embeddings. Trained on a curated dataset of 422 OSC devices, the fine-tuned LLM demonstrated strong predictive accuracy across key performance metrics: power conversion efficiency (PCE, R2: 0.87), short-circuit current (JSC, R2: 0.82), open-circuit voltage (VOC, R2: 0.89), and fill factor (FF, R2: 0.59). The models are then used to explore the space of 1.4 million combinations of materials, experimental variables and device architectures. The analysis provides data-driven design guidelines, identifying optimal donor-acceptor combinations and processing conditions that consistently yield higher device performance.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"4 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145711517","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 : 2025-12-07DOI: 10.1038/s41524-025-01890-1
Flóra B. Németh, Andrea Hamza, Beatrix Tugyi, Maya El-Ali, Luca Szegletes, Ádám Madarász, Imre Pápai
An interactive web tool, PredPotS, has been developed for predicting one-electron standard reduction potentials of organic molecules in aqueous solutions. The predictions are generated using deep learning models trained and validated on a chemically diverse database comprising reduction potentials of approximately 8000 organic compounds. The reduction potentials of this database were computed using a composite computational protocol that combines the semiempirical quantum chemical method (GFN2-xTB) and a well-established DFT approach (M06-2X functional along with the SMD solvent model). While this computational approach is cost-effective, it is subject to certain limitations, which are nonetheless duly accounted for in the development of the database. The applied graph-based deep learning methods perform remarkably well in terms of the standard performance metrics. By entering or uploading the SMILES codes of the molecules, PredPotS provides fast and sensible predictions for one-electron standard reduction potentials for a diverse set of organic molecules also in the range compatible with the electrochemical stability of aqueous electrolytes. The PredPotS web tool is particularly well-suited for screening redox-active candidates for aqueous organic redox flow batteries, but it may also prove useful in a variety of other electrochemical applications.
{"title":"PredPotS: web tool for predicting one-electron standard reduction potentials for organic molecules in aqueous phase","authors":"Flóra B. Németh, Andrea Hamza, Beatrix Tugyi, Maya El-Ali, Luca Szegletes, Ádám Madarász, Imre Pápai","doi":"10.1038/s41524-025-01890-1","DOIUrl":"https://doi.org/10.1038/s41524-025-01890-1","url":null,"abstract":"An interactive web tool, PredPotS, has been developed for predicting one-electron standard reduction potentials of organic molecules in aqueous solutions. The predictions are generated using deep learning models trained and validated on a chemically diverse database comprising reduction potentials of approximately 8000 organic compounds. The reduction potentials of this database were computed using a composite computational protocol that combines the semiempirical quantum chemical method (GFN2-xTB) and a well-established DFT approach (M06-2X functional along with the SMD solvent model). While this computational approach is cost-effective, it is subject to certain limitations, which are nonetheless duly accounted for in the development of the database. The applied graph-based deep learning methods perform remarkably well in terms of the standard performance metrics. By entering or uploading the SMILES codes of the molecules, PredPotS provides fast and sensible predictions for one-electron standard reduction potentials for a diverse set of organic molecules also in the range compatible with the electrochemical stability of aqueous electrolytes. The PredPotS web tool is particularly well-suited for screening redox-active candidates for aqueous organic redox flow batteries, but it may also prove useful in a variety of other electrochemical applications.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"4 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145696935","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 : 2025-12-06DOI: 10.1038/s41524-025-01881-2
Pierre-Paul De Breuck, Hai-Chen Wang, Gian-Marco Rignanese, Silvana Botti, Miguel A. L. Marques
The rapid rise of generative artificial intelligence is reshaping materials discovery by offering new ways to propose crystal structures and, in some cases, even predict desired properties. This review provides a comprehensive survey of recent advancements in generative models specifically for inorganic crystalline materials. We outline architectures, representations, conditioning mechanisms, data sources, metrics, and applications, and organize existing models into a unified taxonomy.
{"title":"Generative AI for crystal structures: a review","authors":"Pierre-Paul De Breuck, Hai-Chen Wang, Gian-Marco Rignanese, Silvana Botti, Miguel A. L. Marques","doi":"10.1038/s41524-025-01881-2","DOIUrl":"https://doi.org/10.1038/s41524-025-01881-2","url":null,"abstract":"The rapid rise of generative artificial intelligence is reshaping materials discovery by offering new ways to propose crystal structures and, in some cases, even predict desired properties. This review provides a comprehensive survey of recent advancements in generative models specifically for inorganic crystalline materials. We outline architectures, representations, conditioning mechanisms, data sources, metrics, and applications, and organize existing models into a unified taxonomy.","PeriodicalId":19342,"journal":{"name":"npj Computational Materials","volume":"5 1","pages":""},"PeriodicalIF":9.7,"publicationDate":"2025-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145680698","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：