Computational Materials Science最新文献_第7页

Linear-response TDDFT and supercell core-hole calculations of electron energy-loss spectra in polymorphic HfO2 多态HfO2中电子能量损失谱的线性响应TDDFT和超级单体核空穴计算

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-03-10 Epub Date: 2026-02-25 DOI: 10.1016/j.commatsci.2026.114601

Jiachen Fan, Shang-Peng Gao

Due to its outstanding electronic and dielectric properties, hafnium dioxide (HfO₂) has emerged as a promising material across diverse fields, from high-κ dielectrics to next-generation non-volatile memories and optical devices. A comprehensive understanding of the dielectric response and electronic excitation characteristics of HfO₂ is essential for both fundamental studies and device applications. Electron energy-loss spectroscopy (EELS), a key technique for probing dielectric behavior and electronic structure, plays a crucial role in characterizing polymorphic HfO₂. In this study, the low-loss and core-loss EELS spectra of cubic (c), tetragonal (t), monoclinic (m) and orthorhombic-III (oIII) HfO₂ are systematically investigated using first-principles calculations. In the low-loss region, anisotropic EELS spectra are obtained via time-dependent density functional theory (TDDFT) with random phase approximation (RPA) and adiabatic local density approximation (ALDA), including local-field effects (LFEs), and the influence of finite momentum transfer on the energy and intensity evolution of characteristic excitations is thoroughly examined. In the core-loss region, anisotropic O K-edge energy-loss near-edge structures (ELNES) are calculated using a core-excited pseudopotential approach incorporating core-hole effects. The spectral features are analyzed in conjunction with the projected density of states (PDOS) to elucidate their electronic origins, and the roles of local chemical coordination on the ELNES are further assessed. This work offers rigorous theoretical insight into the electronic excitation properties of polymorphic HfO₂. The findings provide a deeper understanding of electronic excitation behavior and dielectric response in different HfO₂ polymorphs, thereby advancing the interpretation of their EELS spectra and supporting the optimization of HfO₂-based electronic devices.

由于其出色的电子和介电性能，二氧化铪（HfO2）已成为从高κ介电材料到下一代非易失性存储器和光学器件等各个领域的有前途的材料。全面了解HfO2的介电响应和电子激发特性对于基础研究和器件应用都是必不可少的。电子能量损失谱（EELS）是探测HfO2介电行为和电子结构的关键技术，在表征多晶态HfO2中起着至关重要的作用。在本研究中，采用第一性原理计算系统地研究了立方(c)，四方(t)，单斜(m)和正交- iii (oIII) HfO2的低损耗和核心损耗EELS谱。在低损耗区，采用随机相位近似（RPA）和绝热局部密度近似（ALDA），利用随时间密度泛函理论（TDDFT）获得了EELS各向异性谱，包括局域场效应（LFEs），并深入研究了有限动量传递对特征激发能量和强度演化的影响。在核损耗区，利用核激发伪势方法计算了各向异性O - k边能量损失近边结构（ELNES）。结合预测态密度（PDOS）分析了光谱特征，阐明了它们的电子来源，并进一步评估了局部化学配位对ELNES的作用。这项工作为多晶HfO2的电子激发特性提供了严格的理论见解。研究结果对不同HfO2多晶的电子激发行为和介电响应有了更深入的了解，从而促进了对其EELS谱的解释，并为HfO2基电子器件的优化提供了支持。

{"title":"Linear-response TDDFT and supercell core-hole calculations of electron energy-loss spectra in polymorphic HfO2","authors":"Jiachen Fan, Shang-Peng Gao","doi":"10.1016/j.commatsci.2026.114601","DOIUrl":"10.1016/j.commatsci.2026.114601","url":null,"abstract":"<div><div>Due to its outstanding electronic and dielectric properties, hafnium dioxide (HfO<sub>2</sub>) has emerged as a promising material across diverse fields, from high-κ dielectrics to next-generation non-volatile memories and optical devices. A comprehensive understanding of the dielectric response and electronic excitation characteristics of HfO<sub>2</sub> is essential for both fundamental studies and device applications. Electron energy-loss spectroscopy (EELS), a key technique for probing dielectric behavior and electronic structure, plays a crucial role in characterizing polymorphic HfO<sub>2</sub>. In this study, the low-loss and core-loss EELS spectra of cubic (c), tetragonal (t), monoclinic (m) and orthorhombic-III (oIII) HfO<sub>2</sub> are systematically investigated using first-principles calculations. In the low-loss region, anisotropic EELS spectra are obtained via time-dependent density functional theory (TDDFT) with random phase approximation (RPA) and adiabatic local density approximation (ALDA), including local-field effects (LFEs), and the influence of finite momentum transfer on the energy and intensity evolution of characteristic excitations is thoroughly examined. In the core-loss region, anisotropic O K-edge energy-loss near-edge structures (ELNES) are calculated using a core-excited pseudopotential approach incorporating core-hole effects. The spectral features are analyzed in conjunction with the projected density of states (PDOS) to elucidate their electronic origins, and the roles of local chemical coordination on the ELNES are further assessed. This work offers rigorous theoretical insight into the electronic excitation properties of polymorphic HfO<sub>2</sub>. The findings provide a deeper understanding of electronic excitation behavior and dielectric response in different HfO<sub>2</sub> polymorphs, thereby advancing the interpretation of their EELS spectra and supporting the optimization of HfO<sub>2</sub>-based electronic devices.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"267 ","pages":"Article 114601"},"PeriodicalIF":3.3,"publicationDate":"2026-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403726","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Improved capabilities of the TurboGAP code for radiation induced cascade simulations: An illustration with silicon 改进的TurboGAP代码的能力，辐射诱导级联模拟：一个插图与硅

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-03-10 Epub Date: 2026-02-23 DOI: 10.1016/j.commatsci.2026.114560

Uttiyoarnab Saha , Ali Hamedani , Miguel A. Caro , Andrea E. Sand

TurboGAP is a software package designed for efficient molecular dynamics simulations using Gaussian approximation potential (GAP) machine-learning interatomic potentials (MLIP). In this work, we enhance the capabilities of TurboGAP for radiation damage simulations by implementing a two-temperature molecular dynamics model, based on electron density-dependent coupling of electronic and atomic subsystems. Additionally, we implement adaptive calculation of the timestep and grouping of atoms for cell-border cooling. Our implementation incorporates electronic stopping power either through a traditional friction-based model or a more realistic first-principles-derived model. By combining the computational efficiency of TurboGAP with the accuracy of GAP MLIP, we perform cascade simulations in silicon with primary knock-on atom (PKA) energies up to 10 keV. Our simulations scale to systems containing up to 1 million atoms. We study the generation and clustering of radiation-induced defects. We also calculate ion-beam mixing and compare our results with the experimental data, discussing how the GAP-MLIP along with the inclusion of a realistic electronic stopping model affects the prediction of experimental mixing values.

TurboGAP是一个使用高斯近似势（GAP）机器学习原子间势（MLIP）进行有效分子动力学模拟的软件包。在这项工作中，我们通过实现基于电子和原子子系统的电子密度依赖耦合的双温度分子动力学模型，增强了TurboGAP辐射损伤模拟的能力。此外，我们实现了自适应计算的时间步长和分组原子的细胞边界冷却。我们的实现通过传统的基于摩擦的模型或更现实的第一原理衍生模型集成了电子停止功率。通过将TurboGAP的计算效率与GAP MLIP的精度相结合，我们在硅中进行了一级撞击原子（PKA）能量高达10 keV的级联模拟。我们的模拟扩展到包含多达100万个原子的系统。我们研究了辐射缺陷的产生和聚类。我们还计算了离子束混合，并将结果与实验数据进行了比较，讨论了GAP-MLIP以及包含现实电子停止模型如何影响实验混合值的预测。

{"title":"Improved capabilities of the TurboGAP code for radiation induced cascade simulations: An illustration with silicon","authors":"Uttiyoarnab Saha , Ali Hamedani , Miguel A. Caro , Andrea E. Sand","doi":"10.1016/j.commatsci.2026.114560","DOIUrl":"10.1016/j.commatsci.2026.114560","url":null,"abstract":"<div><div>TurboGAP is a software package designed for efficient molecular dynamics simulations using Gaussian approximation potential (GAP) machine-learning interatomic potentials (MLIP). In this work, we enhance the capabilities of TurboGAP for radiation damage simulations by implementing a two-temperature molecular dynamics model, based on electron density-dependent coupling of electronic and atomic subsystems. Additionally, we implement adaptive calculation of the timestep and grouping of atoms for cell-border cooling. Our implementation incorporates electronic stopping power either through a traditional friction-based model or a more realistic first-principles-derived model. By combining the computational efficiency of TurboGAP with the accuracy of GAP MLIP, we perform cascade simulations in silicon with primary knock-on atom (PKA) energies up to 10 keV. Our simulations scale to systems containing up to 1 million atoms. We study the generation and clustering of radiation-induced defects. We also calculate ion-beam mixing and compare our results with the experimental data, discussing how the GAP-MLIP along with the inclusion of a realistic electronic stopping model affects the prediction of experimental mixing values.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"267 ","pages":"Article 114560"},"PeriodicalIF":3.3,"publicationDate":"2026-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403728","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Numerical efficiency of explicit time integrators for phase-field models 相场模型显式时间积分器的数值效率

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-03-10 Epub Date: 2026-02-20 DOI: 10.1016/j.commatsci.2026.114599

Marco Seiz , Tomohiro Takaki

Phase-field simulations are a practical but also expensive tool to calculate microstructural evolution. This work aims to compare explicit time integrators for a broad class of phase-field models involving coupling between the phase-field and concentration. Particular integrators are adapted to constraints on the phase-field as well as storage scheme implications. Reproducible benchmarks are defined with a focus on having exact sharp interface solutions, allowing for identification of dominant error terms. Speedups of 4 to 114 over the classic forward Euler integrator are achievable while still using a fully explicit scheme without appreciable accuracy loss. Application examples include final stage sintering with pores slowing down grain growth as they move and merge over time.

相场模拟是一种实用但昂贵的计算微观结构演变的工具。这项工作旨在比较涉及相场和浓度之间耦合的广泛的相场模型的显式时间积分器。特定的积分器适应于相场的约束以及存储方案的含义。定义可重复的基准时，重点是要有精确的接口解决方案，以便识别主要的错误项。4到114的加速超过经典的前向欧拉积分器是可以实现的，同时仍然使用一个完全显式的方案，没有明显的精度损失。应用实例包括烧结的最后阶段，随着时间的推移，孔隙的移动和合并会减缓晶粒的生长。

引用次数: 0

Beyond prediction: Assessing stability in feature selection methods for materials science applications 超越预测：评估材料科学应用中特征选择方法的稳定性

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-03-10 Epub Date: 2026-02-25 DOI: 10.1016/j.commatsci.2026.114609

Yoshiyasu Takefuji

This study examines the reliability of feature selection methods in materials science, where machine learning applications have surged despite widespread misapplications stemming from limited understanding of interpretability constraints. We compare supervised models (XGBoost, Random Forest), unsupervised techniques (Feature Agglomeration, HVGS), and statistical methods (Spearman's correlation) through a novel stability testing framework using a public materials dataset. Our results reveal that despite high predictive accuracy (R² > 0.95), supervised models produce unstable feature rankings when the highest-ranked feature is removed—a critical flaw when identifying structure-property relationships. Common misapplications include over-reliance on black-box models for scientific interpretation, insufficient cross-validation procedures, and failure to test feature importance stability. In contrast, unsupervised methods and Spearman's correlation demonstrate perfect ranking stability while maintaining competitive performance. This highlights a fundamental distinction between prediction accuracy and feature importance reliability. We recommend that materials researchers supplement supervised learning with model-agnostic approaches to avoid misinterpretation of material-property relationships and ensure scientifically robust conclusions about causal mechanisms in materials development.

本研究考察了材料科学中特征选择方法的可靠性，尽管由于对可解释性约束的理解有限而导致广泛的误用，但机器学习应用仍在激增。我们通过使用公共材料数据集的新型稳定性测试框架，比较了有监督模型（XGBoost, Random Forest）、无监督技术（Feature Agglomeration， HVGS）和统计方法（Spearman’s correlation）。我们的研究结果表明，尽管预测精度很高（R2 > 0.95），但当移除排名最高的特征时，监督模型会产生不稳定的特征排名——这是识别结构-属性关系时的一个关键缺陷。常见的错误应用包括过度依赖黑盒模型进行科学解释，交叉验证过程不足，以及未能测试特征重要性的稳定性。相比之下，无监督方法和斯皮尔曼相关性在保持竞争绩效的同时表现出完美的排名稳定性。这突出了预测准确性和特征重要性可靠性之间的根本区别。我们建议材料研究人员用模型不可知的方法来补充监督学习，以避免对材料-属性关系的误解，并确保对材料开发中的因果机制得出科学可靠的结论。

{"title":"Beyond prediction: Assessing stability in feature selection methods for materials science applications","authors":"Yoshiyasu Takefuji","doi":"10.1016/j.commatsci.2026.114609","DOIUrl":"10.1016/j.commatsci.2026.114609","url":null,"abstract":"<div><div>This study examines the reliability of feature selection methods in materials science, where machine learning applications have surged despite widespread misapplications stemming from limited understanding of interpretability constraints. We compare supervised models (XGBoost, Random Forest), unsupervised techniques (Feature Agglomeration, HVGS), and statistical methods (Spearman's correlation) through a novel stability testing framework using a public materials dataset. Our results reveal that despite high predictive accuracy (R<sup>2</sup> > 0.95), supervised models produce unstable feature rankings when the highest-ranked feature is removed—a critical flaw when identifying structure-property relationships. Common misapplications include over-reliance on black-box models for scientific interpretation, insufficient cross-validation procedures, and failure to test feature importance stability. In contrast, unsupervised methods and Spearman's correlation demonstrate perfect ranking stability while maintaining competitive performance. This highlights a fundamental distinction between prediction accuracy and feature importance reliability. We recommend that materials researchers supplement supervised learning with model-agnostic approaches to avoid misinterpretation of material-property relationships and ensure scientifically robust conclusions about causal mechanisms in materials development.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"267 ","pages":"Article 114609"},"PeriodicalIF":3.3,"publicationDate":"2026-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403733","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Expanding the search space of high entropy oxides and predicting synthesizability using machine learning interatomic potentials 扩展高熵氧化物的搜索空间并利用机器学习原子间势预测可合成性

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-03-10 Epub Date: 2026-02-11 DOI: 10.1016/j.commatsci.2026.114581

Oliver A. Dicks , Solveig S. Aamlid , Alannah M. Hallas , Joerg Rottler

We propose an efficient computational methodology for predicting the synthesizability of high entropy oxides (HEOs) in a large space of possible candidate compounds. HEOs are a growing field with an enormous potential chemical composition space, and yet the discovery of new HEOs is slow and driven by experimental trial-and-error. In this work, we attempt to speed up this process by using a machine learned interatomic potential offering DFT-level accuracy. Our methodology starts by identifying a set of crystal structures and elements for screening, building a large random unit cell of each composition and structure, then relaxing this structure. The most promising candidates are distinguished based on the variance of the individual cation energies, which we introduce as our novel entropy descriptor, and the enthalpy of mixing, which is used as the enthalpy descriptor. The approach is applied to tetravalent HEOs, and its validity is confirmed by comparison to alternative descriptors and DFT calculations for a set of 7 elements. The search is then extended to a set of 14 elements and three crystal structures, where it successfully identifies the only known stable 4-component HEO in the

α

-PbO

_{2}

structure, as well as predicting several new 5-component candidate systems. This approach can straightforwardly be applied to new sets of elements and structures, allowing for the accelerated discovery of new HEOs.

我们提出了一种有效的计算方法来预测高熵氧化物（HEOs）在可能的候选化合物的大空间中的可合成性。heo是一个不断发展的领域，具有巨大的潜在化学成分空间，然而新的heo的发现是缓慢的，并且受到实验试错的驱动。在这项工作中，我们试图通过使用提供dft级别精度的机器学习原子间势来加速这一过程。我们的方法首先确定一组用于筛选的晶体结构和元素，然后构建每个组成和结构的大型随机单元格，然后放松该结构。最有希望的候选者是根据单个阳离子能量的方差来区分的，我们引入了新的熵描述符，以及混合焓，它被用作焓描述符。将该方法应用于四价heo，并通过与7个元素的替代描述符和DFT计算的比较证实了其有效性。然后将搜索扩展到14个元素和3种晶体结构，成功地确定了α-PbO2结构中唯一已知的稳定的4组分HEO，并预测了几个新的5组分候选体系。这种方法可以直接应用于新的元素和结构，从而加速发现新的heo。

{"title":"Expanding the search space of high entropy oxides and predicting synthesizability using machine learning interatomic potentials","authors":"Oliver A. Dicks , Solveig S. Aamlid , Alannah M. Hallas , Joerg Rottler","doi":"10.1016/j.commatsci.2026.114581","DOIUrl":"10.1016/j.commatsci.2026.114581","url":null,"abstract":"<div><div>We propose an efficient computational methodology for predicting the synthesizability of high entropy oxides (HEOs) in a large space of possible candidate compounds. HEOs are a growing field with an enormous potential chemical composition space, and yet the discovery of new HEOs is slow and driven by experimental trial-and-error. In this work, we attempt to speed up this process by using a machine learned interatomic potential offering DFT-level accuracy. Our methodology starts by identifying a set of crystal structures and elements for screening, building a large random unit cell of each composition and structure, then relaxing this structure. The most promising candidates are distinguished based on the variance of the individual cation energies, which we introduce as our novel entropy descriptor, and the enthalpy of mixing, which is used as the enthalpy descriptor. The approach is applied to tetravalent HEOs, and its validity is confirmed by comparison to alternative descriptors and DFT calculations for a set of 7 elements. The search is then extended to a set of 14 elements and three crystal structures, where it successfully identifies the only known stable 4-component HEO in the <span><math><mi>α</mi></math></span>-PbO<span><math><msub><mrow></mrow><mrow><mn>2</mn></mrow></msub></math></span> structure, as well as predicting several new 5-component candidate systems. This approach can straightforwardly be applied to new sets of elements and structures, allowing for the accelerated discovery of new HEOs.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"267 ","pages":"Article 114581"},"PeriodicalIF":3.3,"publicationDate":"2026-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403829","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A deep learning strategy to calibrate heteroatomic interactions in metal alloys 一种校准金属合金中杂原子相互作用的深度学习策略

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-03-10 Epub Date: 2026-02-14 DOI: 10.1016/j.commatsci.2026.114586

Luca Benzi , Diana Nelli , Pascal Andreazza , Riccardo Ferrando , Georg Daniel Förster

A machine-learning-assisted strategy is proposed to calibrate the heteronuclear parameters of the Tight-Binding Second-Moment Approximation (TB–SMA) potential using finite-temperature experimental data. The method involves the use of neural-network surrogate models trained on a large dataset of fictitious binary alloys, generated by randomly sampling TB–SMA parameter sets within physically meaningful intervals. Each surrogate model learns to predict thermodynamic observables — mixing enthalpy and lattice parameter — directly from the potential parameters. Once trained, the networks provide instantaneous predictions, eliminating the need for costly simulations during the optimization loop. The surrogate models are then embedded in a minimization scheme that adjusts the mixed interaction parameters to reproduce experimental thermodynamic data at selected compositions and at given temperatures. This workflow is applied to ten binary alloys formed by Cu, Ni, Pt, Pd, and Rh, obtaining parametrizations that accurately match experimental trends. The approach is general and well adapted to complex multi-element systems as high-entropy alloys. It can be extended to other potential forms and target properties.

提出了一种机器学习辅助策略，利用有限温度实验数据校准紧密结合第二矩近似（TB-SMA）势的异核参数。该方法使用神经网络代理模型，该模型是在虚拟二元合金的大型数据集上训练的，该数据集是在物理上有意义的间隔内随机抽样TB-SMA参数集生成的。每个代理模型学习预测热力学观测-混合焓和晶格参数-直接从势参数。一旦训练，网络提供即时预测，消除了优化循环期间昂贵的模拟的需要。然后将代理模型嵌入到最小化方案中，该方案调整混合相互作用参数，以在选定成分和给定温度下再现实验热力学数据。该工作流程应用于由Cu， Ni, Pt， Pd和Rh组成的十种二元合金，获得准确匹配实验趋势的参数化。该方法具有通用性，适用于复杂的多元素系统，如高熵合金。它可以扩展到其他潜在的形式和目标属性。

{"title":"A deep learning strategy to calibrate heteroatomic interactions in metal alloys","authors":"Luca Benzi , Diana Nelli , Pascal Andreazza , Riccardo Ferrando , Georg Daniel Förster","doi":"10.1016/j.commatsci.2026.114586","DOIUrl":"10.1016/j.commatsci.2026.114586","url":null,"abstract":"<div><div>A machine-learning-assisted strategy is proposed to calibrate the heteronuclear parameters of the Tight-Binding Second-Moment Approximation (TB–SMA) potential using finite-temperature experimental data. The method involves the use of neural-network surrogate models trained on a large dataset of fictitious binary alloys, generated by randomly sampling TB–SMA parameter sets within physically meaningful intervals. Each surrogate model learns to predict thermodynamic observables — mixing enthalpy and lattice parameter — directly from the potential parameters. Once trained, the networks provide instantaneous predictions, eliminating the need for costly simulations during the optimization loop. The surrogate models are then embedded in a minimization scheme that adjusts the mixed interaction parameters to reproduce experimental thermodynamic data at selected compositions and at given temperatures. This workflow is applied to ten binary alloys formed by Cu, Ni, Pt, Pd, and Rh, obtaining parametrizations that accurately match experimental trends. The approach is general and well adapted to complex multi-element systems as high-entropy alloys. It can be extended to other potential forms and target properties.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"267 ","pages":"Article 114586"},"PeriodicalIF":3.3,"publicationDate":"2026-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403408","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Sample-efficient active learning for materials informatics using integrated posterior variance 基于后验方差的材料信息学样本有效主动学习

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-03-10 Epub Date: 2026-02-09 DOI: 10.1016/j.commatsci.2026.114551

Ramsey Issa , Said Hamad , Ricardo Grau-Crespo , Emad Awad , Taylor D. Sparks

Developing accurate machine learning models with minimal data remains a central challenge in materials informatics. Efficient models can significantly reduce costly computational simulations and time-intensive experimentation by providing reliable predictions of material properties. In this work, we investigate the integrated posterior variance acquisition function within an active learning framework, comparing its performance against three established methods: random sampling, point-wise uncertainty sampling, and query-by-committee. We evaluate these methods across three diverse datasets: AutoAM, Thermoelectric, and NMR. Our results demonstrate that integrated posterior variance consistently outperforms conventional methods in selecting candidates that minimize prediction error with fewer labeled samples. We identify two key limitations: computational overhead that increases with dataset size and diminished effectiveness in high-dimensional feature spaces where distance metrics become less meaningful. Despite these constraints, our approach demonstrates how strategic experimental selection can substantially improve model performance across varying materials informatics domains while minimizing the number of required experiments, offering significant resource savings for materials discovery workflows.

用最少的数据开发准确的机器学习模型仍然是材料信息学的核心挑战。有效的模型可以通过提供可靠的材料性能预测，显著减少昂贵的计算模拟和时间密集的实验。在这项工作中，我们研究了主动学习框架内的综合后验方差获取函数，并将其与三种既定方法（随机抽样、点不确定性抽样和按委员会查询）的性能进行了比较。我们在三个不同的数据集上评估这些方法：AutoAM， Thermoelectric和NMR。我们的结果表明，综合后验方差始终优于传统的选择方法，以减少标记样本的预测误差。我们确定了两个关键的限制：计算开销随着数据集大小的增加而增加，并且在距离度量变得没有意义的高维特征空间中有效性降低。尽管存在这些限制，我们的方法证明了策略性实验选择如何在不同材料信息学领域显著提高模型性能，同时最大限度地减少所需实验的数量，为材料发现工作流程节省大量资源。

{"title":"Sample-efficient active learning for materials informatics using integrated posterior variance","authors":"Ramsey Issa , Said Hamad , Ricardo Grau-Crespo , Emad Awad , Taylor D. Sparks","doi":"10.1016/j.commatsci.2026.114551","DOIUrl":"10.1016/j.commatsci.2026.114551","url":null,"abstract":"<div><div>Developing accurate machine learning models with minimal data remains a central challenge in materials informatics. Efficient models can significantly reduce costly computational simulations and time-intensive experimentation by providing reliable predictions of material properties. In this work, we investigate the integrated posterior variance acquisition function within an active learning framework, comparing its performance against three established methods: random sampling, point-wise uncertainty sampling, and query-by-committee. We evaluate these methods across three diverse datasets: AutoAM, Thermoelectric, and NMR. Our results demonstrate that integrated posterior variance consistently outperforms conventional methods in selecting candidates that minimize prediction error with fewer labeled samples. We identify two key limitations: computational overhead that increases with dataset size and diminished effectiveness in high-dimensional feature spaces where distance metrics become less meaningful. Despite these constraints, our approach demonstrates how strategic experimental selection can substantially improve model performance across varying materials informatics domains while minimizing the number of required experiments, offering significant resource savings for materials discovery workflows.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"267 ","pages":"Article 114551"},"PeriodicalIF":3.3,"publicationDate":"2026-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403833","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

On rapid solidification and multiscale modeling in metal additive manufacturing: A review 金属增材制造中的快速凝固和多尺度建模研究进展

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-03-10 Epub Date: 2026-02-17 DOI: 10.1016/j.commatsci.2026.114583

Chongfeng Zhang , Yi Song , Leiji Li , Xiaopeng Shen , Weijun Wang , Tianchi Zhu , Fei Xiao

Metal additive manufacturing (AM) offers unprecedented design flexibility and efficiency, yet its performance is critically governed by rapid solidification phenomena. In this paper, we offer an in-depth analysis regarding non-equilibrium effects. Specifically, the discussion centers on critical mechanisms including solute trapping, solute drag as well as interface dynamics, and their role in shaping microstructure evolution during rapid cooling. Specific attention will be given to dendritic, eutectic, peritectic solidification, and banded structures, which are characteristic of metal AM. In parallel, the review highlights the latest advances in multiscale modeling, spanning molecular dynamics, kinetic Monte Carlo, cellular automata, and phase-field approaches. By linking atomistic processes to mesoscopic pattern formation, this article will offer a comprehensive perspective that connects fundamental solidification science with predictive simulation tools. The paper closes by identifying critical obstacles and potential avenues for future research.

金属增材制造（AM）提供了前所未有的设计灵活性和效率，但其性能受到快速凝固现象的严重影响。在本文中，我们对非均衡效应进行了深入的分析。具体来说，讨论集中在关键机制，包括溶质捕获，溶质阻力和界面动力学，以及它们在快速冷却过程中形成微观结构演变中的作用。将特别注意枝晶、共晶、包晶凝固和带状组织，这是金属增材制造的特征。同时，回顾了多尺度建模的最新进展，包括分子动力学、动力学蒙特卡罗、元胞自动机和相场方法。通过将原子过程与介观图案形成联系起来，本文将提供一个综合的视角，将基础凝固科学与预测模拟工具联系起来。论文最后指出了未来研究的关键障碍和潜在途径。

{"title":"On rapid solidification and multiscale modeling in metal additive manufacturing: A review","authors":"Chongfeng Zhang , Yi Song , Leiji Li , Xiaopeng Shen , Weijun Wang , Tianchi Zhu , Fei Xiao","doi":"10.1016/j.commatsci.2026.114583","DOIUrl":"10.1016/j.commatsci.2026.114583","url":null,"abstract":"<div><div>Metal additive manufacturing (AM) offers unprecedented design flexibility and efficiency, yet its performance is critically governed by rapid solidification phenomena. In this paper, we offer an in-depth analysis regarding non-equilibrium effects. Specifically, the discussion centers on critical mechanisms including solute trapping, solute drag as well as interface dynamics, and their role in shaping microstructure evolution during rapid cooling. Specific attention will be given to dendritic, eutectic, peritectic solidification, and banded structures, which are characteristic of metal AM. In parallel, the review highlights the latest advances in multiscale modeling, spanning molecular dynamics, kinetic Monte Carlo, cellular automata, and phase-field approaches. By linking atomistic processes to mesoscopic pattern formation, this article will offer a comprehensive perspective that connects fundamental solidification science with predictive simulation tools. The paper closes by identifying critical obstacles and potential avenues for future research.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"267 ","pages":"Article 114583"},"PeriodicalIF":3.3,"publicationDate":"2026-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147403835","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A neuroevolution potential for predicting the lattice thermal conductivity of structurally disordered γ-Ga2O3 预测结构无序γ-Ga2O3晶格热导率的神经进化潜力

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-02-28 Epub Date: 2026-02-01 DOI: 10.1016/j.commatsci.2026.114555

Fangwei Yang , Haoran Sun , Xiaoxin Yang , Xu Li , Gang Yang

In recent years, the lattice thermal conductivity of γ-Ga₂O₃ with a defective spinel structure has attracted widespread attention from both industry and academia. However, due to its inherent structural disorder, accurately predicting its thermal conductivity using first-principles methods remains challenging. To overcome this challenge, this study developed a machine-learning interatomic potential applicable to multiple γ-Ga₂O₃ configurations, based on the neuroevolution potential framework combined with a multi-round active-learning strategy. Using this potential, the thermal conductivity of different γ-Ga₂O₃ configurations along various crystallographic directions was calculated. The results show that, within the same structure, the thermal conductivity along the [100] and [010] directions is essentially the same, while it is significantly lower along the [001] direction. Furthermore, the thermal conductivity of all configurations originates primarily from low-frequency phonons in the 0–6 THz range. The highly disordered structure intensifies phonon scattering and significantly reduces the group velocity, resulting in limited actual contribution of high-frequency phonons to thermal transport. Additionally, different configurations exhibit high similarity in phonon transport characteristics, resulting in relatively small differences in thermal conductivity among them.

近年来，具有缺陷尖晶石结构的γ-Ga2O3晶格导热性能受到了业界和学术界的广泛关注。然而，由于其固有的结构无序性，使用第一性原理方法准确预测其导热系数仍然具有挑战性。为了克服这一挑战，本研究基于神经进化势框架结合多轮主动学习策略，开发了一种适用于多种γ-Ga2O3构型的机器学习原子间势。利用该势，计算了不同构型γ-Ga2O3沿不同结晶方向的热导率。结果表明，在同一结构内，沿[100]和[010]方向的导热系数基本相同，而沿[001]方向的导热系数明显较低。此外，所有构型的热导率主要来源于0-6太赫兹范围内的低频声子。高度无序的结构加剧了声子散射，显著降低了群速度，导致高频声子对热输运的实际贡献有限。此外，不同构型的声子输运特性具有较高的相似性，导致它们之间的导热系数差异相对较小。

{"title":"A neuroevolution potential for predicting the lattice thermal conductivity of structurally disordered γ-Ga2O3","authors":"Fangwei Yang , Haoran Sun , Xiaoxin Yang , Xu Li , Gang Yang","doi":"10.1016/j.commatsci.2026.114555","DOIUrl":"10.1016/j.commatsci.2026.114555","url":null,"abstract":"<div><div>In recent years, the lattice thermal conductivity of γ-Ga<sub>2</sub>O<sub>3</sub> with a defective spinel structure has attracted widespread attention from both industry and academia. However, due to its inherent structural disorder, accurately predicting its thermal conductivity using first-principles methods remains challenging. To overcome this challenge, this study developed a machine-learning interatomic potential applicable to multiple γ-Ga<sub>2</sub>O<sub>3</sub> configurations, based on the neuroevolution potential framework combined with a multi-round active-learning strategy. Using this potential, the thermal conductivity of different γ-Ga<sub>2</sub>O<sub>3</sub> configurations along various crystallographic directions was calculated. The results show that, within the same structure, the thermal conductivity along the [100] and [010] directions is essentially the same, while it is significantly lower along the [001] direction. Furthermore, the thermal conductivity of all configurations originates primarily from low-frequency phonons in the 0–6 THz range. The highly disordered structure intensifies phonon scattering and significantly reduces the group velocity, resulting in limited actual contribution of high-frequency phonons to thermal transport. Additionally, different configurations exhibit high similarity in phonon transport characteristics, resulting in relatively small differences in thermal conductivity among them.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"266 ","pages":"Article 114555"},"PeriodicalIF":3.3,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146186284","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

ReaxFF parameter optimization for β-Ga₂O₃ MD simulations using Gaussian process Bayesian optimization 基于高斯过程贝叶斯优化的β-Ga₂O₃MD模拟ReaxFF参数优化

IF 3.3 3区材料科学 Q2 MATERIALS SCIENCE, MULTIDISCIPLINARY

Computational Materials Science

Pub Date : 2026-02-28 Epub Date: 2026-02-07 DOI: 10.1016/j.commatsci.2026.114577

Stepan Savka, Andriy Serednytski, Dmytro Popovych

β-Gallium oxide (β-Ga₂O₃) is a promising wide-bandgap semiconductor for power electronics, requiring accurate molecular dynamics (MD) simulations to understand its atomic-scale behavior. This work presents the first automated optimization of ReaxFF parameters for β-Ga₂O₃ using Gaussian Process (GP) Bayesian optimization with a multi-objective framework incorporating pressure matching, force matching, and NVE stability testing. We optimized 22 critical ReaxFF parameters including bond energies, bond lengths, angle parameters, van der Waals interactions, and electronic properties. Reference data were obtained from MACE-MP-0, a universal machine learning potential trained on >150,000 DFT calculations. The multi-objective optimization achieved validated NVE ensemble stability at 0.1 fs timestep, with equilibrium pressure matching within 1.2% of MACE-MP-0 predictions (6.75 vs 6.67 GPa). The optimized parameters accurately reproduce experimental structural properties (lattice parameters within 0.3–2.6%, GaO bonds within 1%) and elastic constants within 2% of DFT values. Systematic timestep testing at 0.1, 0.25, and 0.5 fs confirmed that 0.1 fs is optimal for stable dynamics, characteristic of ReaxFF potentials with stiff bond terms. Parameter importance analysis revealed that van der Waals interactions and bond energies are most critical for accurate Ga₂O₃ modeling. The GP-Bayesian framework with multi-objective optimization successfully produced production-ready ReaxFF parameters for β-Ga₂O₃ MD simulations, demonstrating an efficient approach for developing reactive force fields with validated dynamic stability.

β-氧化镓（β-Ga₂O₃）是一种很有前途的用于电力电子的宽带隙半导体，需要精确的分子动力学（MD）模拟来理解其原子尺度的行为。这项工作首次使用高斯过程（GP）贝叶斯优化对β-Ga₂O₃的ReaxFF参数进行了自动优化，该优化具有多目标框架，包括压力匹配、力匹配和NVE稳定性测试。我们优化了ReaxFF的22个关键参数，包括键能、键长、角参数、范德华相互作用和电子性质。参考数据来自MACE-MP-0， MACE-MP-0是一种通用机器学习潜力，经过150,000次DFT计算训练。多目标优化在0.1 fs时间步长下实现了有效的NVE集成稳定性，平衡压力匹配在MACE-MP-0预测的1.2%以内（6.75 vs 6.67 GPa）。优化后的参数准确再现了实验结构性能（晶格参数在0.3-2.6%之间，GaO键在1%之间）和弹性常数在DFT值的2%以内。在0.1、0.25和0.5 fs的系统时间步长测试证实，0.1 fs是稳定动力学的最佳选择，具有刚性键项的ReaxFF电位的特征。参数重要性分析表明，范德华相互作用和键能对于精确的Ga₂O₃建模是最关键的。基于多目标优化的GP-Bayesian框架成功地为β-Ga₂O₃MD模拟生成了生产就绪的ReaxFF参数，展示了一种有效的方法来开发具有动态稳定性的反作用力场。

{"title":"ReaxFF parameter optimization for β-Ga₂O₃ MD simulations using Gaussian process Bayesian optimization","authors":"Stepan Savka, Andriy Serednytski, Dmytro Popovych","doi":"10.1016/j.commatsci.2026.114577","DOIUrl":"10.1016/j.commatsci.2026.114577","url":null,"abstract":"<div><div>β-Gallium oxide (β-Ga₂O₃) is a promising wide-bandgap semiconductor for power electronics, requiring accurate molecular dynamics (MD) simulations to understand its atomic-scale behavior. This work presents the first automated optimization of ReaxFF parameters for β-Ga₂O₃ using Gaussian Process (GP) Bayesian optimization with a multi-objective framework incorporating pressure matching, force matching, and NVE stability testing. We optimized 22 critical ReaxFF parameters including bond energies, bond lengths, angle parameters, van der Waals interactions, and electronic properties. Reference data were obtained from MACE-MP-0, a universal machine learning potential trained on >150,000 DFT calculations. The multi-objective optimization achieved validated NVE ensemble stability at 0.1 fs timestep, with equilibrium pressure matching within 1.2% of MACE-MP-0 predictions (6.75 vs 6.67 GPa). The optimized parameters accurately reproduce experimental structural properties (lattice parameters within 0.3–2.6%, Ga<img>O bonds within 1%) and elastic constants within 2% of DFT values. Systematic timestep testing at 0.1, 0.25, and 0.5 fs confirmed that 0.1 fs is optimal for stable dynamics, characteristic of ReaxFF potentials with stiff bond terms. Parameter importance analysis revealed that van der Waals interactions and bond energies are most critical for accurate Ga₂O₃ modeling. The GP-Bayesian framework with multi-objective optimization successfully produced production-ready ReaxFF parameters for β-Ga₂O₃ MD simulations, demonstrating an efficient approach for developing reactive force fields with validated dynamic stability.</div></div>","PeriodicalId":10650,"journal":{"name":"Computational Materials Science","volume":"266 ","pages":"Article 114577"},"PeriodicalIF":3.3,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146186227","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0