|
 |
快捷服务 |
|
|
|
|
|
|
|
相关链接 |
|
|
|
|
|
|
会议信息 |
|
|
|
|
|
|
友情链接 |
|
|
|
|
|
|
|
|
| 期刊介绍 |
 |
|
《npj 计算材料学》是在线出版、完全开放获取的国际学术期刊。发表结合计算模拟与设计的材料学一流的研究成果。本刊由中国科学院上海硅酸盐研究所与英国自然出版集团（Nature Publishing Group，NPG）以伙伴关系合作出版。
主编为陈龙庆博士，美国宾州大学材料科学与工程系、工程科学与力学系、数学系的杰出教授。共同主编为陈立东研究员，中国科学院上海硅酸盐研究所研究员高性能陶瓷与超微结构国家重点实验室主任。
办刊目的与报道范围
《npj 计算材料学》是在线出版、完全开放获取的国际学术... 【查看详细】 |
|
|
|
|
|
|
Physics guided deep learning for generative design of crystal materials with symmetry constraints
Yong Zhao, Edirisuriya M. Dilanga Siriwardane, Zhenyao Wu, Nihang Fu, Mohammed Al-Fahdi, Ming Hu & Jianjun Hu
npj Computational Materials 9:38 (2022)
doi.org/10.1038/s41524-023-00987-9
Abstract: Discovering new materials is a challenging task in materials science crucial to the progress of human society. Conventional approaches based on experiments and simulations are labor-intensive or costly with success heavily depending on experts’ heuristic knowledge. Here, we propose a deep learning based Physics Guided Crystal Generative Model (PGCGM) for efficient crystal material design with high structural diversity and symmetry. Our model increases the generation validity by more than 700% compared to FTCP, one of the latest structure generators and by more than 45% compared to our previous CubicGAN model. Density Functional Theory (DFT) calculations are used to validate the generated structures with 1869 materials out of 2000 are successfully optimized and deposited into the Carolina Materials Database www.carolinamatdb.org, of which 39.6% have negative formation energy and 5.3% have energy-above-hull less than 0.25 eV/atom, indicating their thermodynamic stability and potential synthesizability.
摘要: 在材料科学中，发现新材料是一个具有挑战性的任务，对人类社会的进步至关重要。传统的基于实验和模拟的方法需要耗费大量的人力和成本，成功率很大程度上依赖于专家的启发式知识。在这里，我们提出了一种基于物理信息的深度学习晶体生成模型（PGCGM），用于高结构多样性和对称性的高效晶体材料设计。与目前最新的晶体结构生成器FTCP相比，我们的模型将生成有效性提高了700%以上，与我们先前的CubicGAN模型相比提高了45%以上。密度泛函理论（DFT）计算被用来验证生成的结构，其中1869种材料成功地被优化并存储在Carolina Materials数据库www.carolinamatdb.org中，其中39.6%具有负的形成能量，5.3%的相对稳定能（e_above_hull)在0.25 eV/atom以下，表明它们具有热力学稳定性和潜在的可合成性。
Editorial Summary
Material has their own way of handling things: Physics guided AI for generative materials design
The discovery of new materials plays a critical role in promoting disruptive innovation in technological fields such as national defense, aerospace, and semiconductor technology. However, the traditional trial-and-error based approach to new material research is often like finding a needle in a haystack, with long cycles and slow output. Utilizing known material databases and AI algorithms to accelerate the discovery of synthesizable, but structurally stable new materials has become a significant issue in the field of materials science. This study proposes a generative design model for crystalline materials with symmetry constraints using a deep learning algorithm based on physical information. This method utilizes a generative adversarial network (GAN) and proposes two loss functions based on atomic pair distance and structural symmetry as constraints, in addition to the standard discriminator loss function, to efficiently generate crystal structures. Prof. Jianjun Hu from the Department of Computer Science and Engineering and Prof. Ming Hu from the Department of Mechanical Engineering at the University of South Carolina in the United States, along with their team, utilized known crystal structure from material datasets such as Material Project, ICSD, and OQMD to train and validate a generative model called PGCGM. Then, they generated a large number of new crystal structures that passed verification through DFT first-principles calculations. Compared to previous work, significant improvement in efficiency and accuracy were achieved. The crystal generation model proposed in this study combines space group affine transformation and effective data self-augmentation methods. And the loss functions are based on atomic pair and crystal structure characteristics, and physical constraints are used to restrict the training of the generative model, thereby preventing the generation of invalid solutions and making the generated crystal structures more likely to satisfy physical constraints. The idea of using deep learning to incorporate constraints for more effective search, turning waste (constraints) into wonder (benefits), has the potential to be applied in other fields such as engineering design, molecular design, protein design, and more.
编辑概述
江湖事江湖了：用原子间固有的游戏规则来指导人工智能材料设计
新材料研发对推动国防科技、航天科技、半导体科技等技术领域的颠覆性创新具有关键作用。然而，基于试错式的传统新材料研究方法有如大海捞针，往往周期过长且产出缓慢。利用已知材料数据库和人工智能算法加速发现可合成的结构稳定的新材料已成为材料科学领域的重大问题。但材料科学领域的突破亟需一种生成式设计方法，以便利用基于物理信息的深度学习算法设计具有对称性约束的晶体材料。
来自美国的南卡罗莱纳大学的计算机科学与工程系的胡建军教授和机械工程系的胡明教授团队，利用Materials Project、ICSD和OQMD等材料数据集中的已知晶体结构训练及验证生成模型PGCGM，并大批量生成了通过DFT第一性原理计算检验的新晶体结构。他们利用生成对抗神经网络（GAN)，在标准鉴别器损失函数的基础上，提出以原子对距离和结构的对称性作为约束条件的两个损失函数，来高效地生成晶体结构。相比之前的算法，新方法在效率和准确率方面得到了极大提升。作者提出的晶体生成模型，结合了空间群仿射变换和有效的数据自增强方法，其损失函数基于原子对特性和晶体结构特性，通过物理约束来限制生成模型的训练，使得生成模型避免生成无效解，从而使得生成的晶体结构更容易满足物理约束。这一将深度学习融入约束以作更有效搜索的、化腐朽（约束）为神奇（有利）的思路在其他工程设计、分子设计、蛋白质设计等领域，都有较大的应用潜力。
|
|
|
|
Strong quartic anharmonicity, ultralow thermal conductivity, high band degeneracy and good thermoelectric performance in Na2TlSb
Tongcai Yue, Yinchang Zhao, Jun Ni, Sheng Meng & Zhenhong Dai
npj Computational Materials 9:17 (2022)
doi.org/10.1038/s41524-023-00970-4
Abstract: We employ first-principles calculations combined with self-consistent phonon theory and Boltzmann transport equations to investigate the thermal transport and thermoelectric properties of full-Heusler compound Na2TlSb. Our findings exhibit that the strong quartic anharmonicity and temperature dependence of the Tl atom with rattling behavior plays an important role in the lattice stability of Na2TlSb. We find that soft Tl-Sb bonding and resonant bonding in the pseudocage composed of the Na and Sb atoms interaction is responsible for ultralow κL. Meanwhile, the multi-valley band structure increases the band degeneracy, results in a high power factor in p-type Na2TlSb. The coexistence of ultralow κL and high power factor presents that Na2TlSb is a potential candidate for thermoelectric applications. Moreover, these findings help to understand the origin of ultralow κL of full-Heusler compounds with strong quartic anharmonicity, leading to the rational design of full-Heusler compounds with high thermoelectric performance.
摘要: 我们采用第一性原理计算，结合自洽声子理论和玻尔兹曼输运方程，研究了全赫斯勒化合物Na2TlSb的热输运和热电性质。研究结果表明，具有rattling行为的Tl原子的强四次非谐性和温度依赖性对Na2TlSb的晶格稳定性起着重要的作用。我们发现，软Tl-Sb键和在由Na和Sb原子之间的组成的赝笼中的共振键共同作用导致了超低晶格热导率。同时，多谷能带结构增加了带简并度，导致p型Na2TlSb的高功率因子。超低晶格热导率和高功率因数的共存表明Na2TlSb是一个潜在的热电应用候选材料。此外，这些发现有助于理解具有强四次非谐性的全赫斯勒化合物的超低晶格热导率的起源，从而合理设计具有高热电性能的全赫斯勒化合物。
Editorial Summary
Full-heusler compound:good thermoelectric performance
Thermoelectric materials can convert thermal energy into electrical energy, so they are widely applied to energy harvesting, thermoelectric cooling and thermopower generators. However, it is a challenge to realize a substantial improvement in the thermoelectric properties of materials due to the Seebeck coefficient S, the electrical conductivity σ and the thermal conductivity κ are mutually restricted, that is, κ and σ are positively correlated, while σ and S are negatively correlated. Reducing lattice thermal conductivity or finding materials with inherently low lattice thermal conductivity is an important way to obtain materials with high thermoelectric properties. This study investigates the effect of higher-order phonon anharmonicity in Na2TlSb on the lattice transport properties, and then predicts the lattice thermal conductivity and thermoelectric figure of merit. A team led by Prof. Zhenhong Dai from the Department of Physics, Yantai University, conducted an in-depth study on the thermoelectric transport mechanism of the full-Heusler compound Na2TlSb using first-principle calculations. In this study, we found that the fourth-order anharmonicity plays an important role in the lattice stability and phonon scattering processes of Na2TlSb. We find that soft Tl-Sb bonding and resonant bonding in the pseudocage composed of the Na and Sb atoms interaction is responsible for ultralow κL. Meanwhile, the multi-valley band structure increases the band degeneracy, results in a high power factor in p-type Na2TlSb. The coexistence of ultralow κL and high power factor presents that Na2TlSb is a potential candidate for thermoelectric applications. Moreover, these findings help to understand the origin of ultralow κL of full-Heusler compounds with strong quartic anharmonicity, leading to the rational design of full-Heusler compounds with high thermoelectric performance.
编辑概述
全赫斯勒化合物：高热电性能
热电材料可以将热能转换为电能，在能量采集、热电制冷和热力发电机等领域有着广阔的应用前景。但实现材料的热电性能大幅度提高仍然是一个挑战，由于塞贝克系数S 、电导率σ 和热导率κ 之间相互制约，即κ和σ是正相关的，σ和S是负相关的。而降低晶格热导率或者寻找本征低晶格热导率的材料是获得高热电性能材料的一个重要的方法。该研究探索了Na2TlSb中高阶声子非谐性对晶格输运性质的影响，然后预测了晶格热导率和热电品质因数。来自烟台大学物理与电子信息学院的戴振宏教授团队联合清华大学倪军教授团队和中科院物理所孟胜研究员团队，采用第一性原理计算对全赫斯勒化合物Na2TlSb的热电输运机制进行了深入的研究。该研究不仅发现四阶非谐性在Na2TlSb的晶格稳定性和声子散射过程中的重要的作用，还发现软Tl-Sb键和在由Na和Sb原子之间的组成的赝笼中的共振键结合导致了超低晶格热导率。此外，由于多谷电子能带结构增加了能带简并度，导致Na2TlSb高的功率因子。低的晶格热导率和高功率因子的共存表明Na2TlSb具有广阔的热电应用前景。该研究还有助于理解具有强四次非谐性的全赫斯勒化合物的超低κL的起源，从而合理设计具有高热电性能的全赫斯勒化合物。
|
|
|
|
Predicting lattice thermal conductivity via machine learning: a mini review
Yufeng Luo, Mengke Li, Hongmei Yuan, Huijun Liu and Ying Fang
npj Computational Materials 9: 04 (2022)
doi.org/10.1038/s41524-023-00964-2
Abstract: Over the past few decades, molecular dynamics simulations and first-principles calculations have become two major approaches to predict the lattice thermal conductivity (κL), which are however limited by insufficient accuracy and high computational cost, respectively. To overcome such inherent disadvantages, machine learning (ML) has been successfully used to accurately predict κL in a high-throughput style. In this review, we give some introductions of recent ML works on the direct and indirect prediction of κL, where the derivations and applications of data-driven models are discussed in details. A brief summary of current works and future perspectives are given in the end.
摘要: 在过去的几十年里，分子动力学模拟和第一性原理计算已经成为预测晶格热导率的两种主要方法，但是它们或者计算精度有限，或者计算成本太高。为了克服这些不足之处，机器学习被成功应用于高通量精确预测体系的晶格热导率。在这篇综述文章中，我们介绍了近年来机器学习在直接和间接预测晶格热导率方面的若干重要进展，详细讨论了数据驱动的机器学习模型的导出及其应用。文章最后，我们对目前的工作进行了简单的总结并对未来的发展和挑战进行了思考。
Editorial Summary
Accurately and readily predicting lattice thermal conductivity: direct and indirect machine learning approaches
The lattice thermal conductivity (κL) is a key design parameter for various technological applications. For example, heat sinks in the electronic devices require higher κL to dissipate the excessive thermal energy, while reducing κL is an effective approach to improve the efficiency of thermoelectric (TE) conversion. It is thus quite necessary to discover/design particular systems with desired κL. On the theoretical side, the most reliable approach for predicting κL is the solution of phonon Boltzmann transport equation (BTE) within the framework of density functional theory (DFT). However, the required calculations of the interatomic force constants (IFCs) are time-consuming, especially for those with large unit cell and low symmetry. As an alternative, the classic molecular dynamics (MD) simulations can be utilized to predict the κL of systems with complex crystal structure. Nevertheless, the accuracy of MD significantly depends on the choice of interatomic potentials, which also limits its wide application. In a word, there remains some challenges or difficulties to accurately predict the κL, especially in a high-throughput way. As an important technique of artificial intelligence, machine learning (ML) can efficiently determine the underlying connectivity among enormous data at extremely low cost. During the past few decades, many efforts have been devoted to evaluate the κL of various systems, both theoretically and experimentally. Based on these available data, ML can establish a mapping between the κL and the input features (such as the atomic mass, the phonon frequency, and the volume of unit cell). Compared with first-principles calculations and MD simulations, the data-driven ML models enable high-throughput evaluation of κL, which exhibit strong predictive power for systems both inside and beyond the training set. In addition to such direct prediction of κL, ML has been successfully used to build accurate interatomic potentials for MD simulations. Generally speaking, the machine learning potential (MLP) employs regression algorithm to determine the ab-initio potential energy surface (PES), and the atomic configurations are usually adopted as input features. On the other hand, as the derivative of total energy with respect to the atomic displacement, IFCs can be obtained from the Taylor expansion of the PES. The MLPs determine the accurate PES and thus can derive the IFCs at almost negligible computational cost, which enable accelerated solution of phonon BTE for the evaluation of κL. Collectively speaking, ML can overcome the inherent disadvantages of MD simulations and first-principles calculations to accurately and readily predict κL. As invited by Prof. Lidong Chen (Co-Editor-in-Chief of npj Computational Materials), a team led by Prof. Liu from Wuhan University wrote a review article about recent works on the prediction of lattice thermal conductivity via machine learning. In the section of “Direct prediction”, they give a detailed introduction of the dataset construction, the feature selection, and the training algorithms, which are then combined to obtain the high-throughput models for predicting κL. In the section of “Indirect approach”, they focus on the construction of machine learning potential, and highlight their first-principles level accuracy as well as advantages over general approaches in predicting lattice thermal conductivity. The review is concluded with a summary of current works and future perspectives.
编辑概述
又快又准地预测晶格热导率：直接和间接的机器学习方法
晶格热导率在各种技术应用中发挥着关键作用，例如电子器件的散热需要较高晶格热导率的材料，而热电材料要求尽可能低的晶格热导率。因此，发现或设计具有特定晶格热导率的体系至关重要。从理论上来说，预测晶格热导率最精确的方法是在密度泛函理论（DFT）框架下求解声子的玻尔兹曼输运方程（BTE），但这需要花费大量时间计算原子间相互作用力常数（IFCs），特别是对于具有较大原胞和较低对称性的体系。作为一种替代方法，分子动力学（MD）模拟适用于预测复杂结构的晶格热导率。然而，MD的准确性很大程度上取决于所选取的原子间势函数，这无疑限制了该方法的广泛应用。总而言之，通过DFT计算和MD模拟来获得晶格热导率目前还存在一些问题或挑战，特别是它们难以用于高通量预测。作为人工智能的一项重要技术，机器学习（ML）能够以极低的成本快速挖掘海量数据内部的特征和规律。基于实验和理论上已知的晶格热导率数据，ML可以建立起晶格热导率和特征参量（比如原子质量、声子频率、晶胞体积等）之间的映射关系。与DFT和MD方法相比，数据驱动的机器学习模型可以对训练集内外各种体系的晶格热导率进行精确的高通量预测。除了这种直接预测方法，ML还可以用来构建MD模拟所需的原子间势函数从而间接预测晶格热导率。一般来说，机器学习势（MLP）运用回归算法确定具有第一性原理精度的势能面，通常选取原子构型作为特征参量；此外，由于IFCs可以通过对势能面进行泰勒展开获得，MLP也可以用于导出IFCs以加速求解声子BTE，同样实现对晶格热导率的间接预测。总的来说，ML可以克服DFT计算和MD模拟的不足，从而快速准确地预测晶格热导率。最近，受到npj Computational Materials主编陈立东教授的邀请，武汉大学物理科学与技术学院的刘惠军教授团队撰文综述了近期利用机器学习预测晶格热导率的若干重要进展。在“直接预测”一节中，他们详细介绍了建立高通量模型预测晶格热导率所需要的数据集构建、特征参数选择、机器学习算法。在“间接预测”一节中，作者关注机器学习势的构造过程，强调了其第一性原理级别的精度以及相较于传统方法预测晶格热导率的优势。文章最后，作者对目前的工作进行了简单的总结并对未来的发展和挑战进行了思考。
|
|
|
|
Machine-learning atomic simulation for heterogeneous catalysis
Dongxiao Chen, Cheng Shang & Zhi-Pan Liu
npj Computational Materials 9:02(2022)
doi.org/10.1038/s41524-022-00959-5
Abstract: Heterogeneous catalysis is at the heart of chemistry. New theoretical methods based on machine learning (ML) techniques that emerged in recent years provide a new avenue to disclose the structures and reaction in complex catalytic systems. Here we review briefly the history of atomic simulations in catalysis and then focus on the recent trend shifting toward ML potential calculations. The advanced methods developed by our group are outlined to illustrate how complex structures and reaction networks can be resolved using the ML potential in combination with efficient global optimization methods. The future of atomic simulation in catalysis is outlooked.
摘要: 化学无序材料具有广泛的应用，而其结构或构型的确定是最重要和最具挑战性的问题之一。由于众所周知的指数墙问题(N-体体系中可能的结构数量呈指数增长)，传统方法对于大体系是非常低效或棘手的。在这里，我们介绍了一种通过主动学习结合第一性原理计算来预测化学无序材料的热力学稳定结构的有效方法。我们的“辣搜”方法可以有效地压缩采样空间，极大地降低计算成本。我们研究了三种不同的、典型的有限尺寸体系，包括阴离子无序体系BaSc(OxF1？x)3 (x=0.667)，具有较大尺寸的阳离子无序体系Ca1？xMnxCO3 (x=0.25)和具有较大采样空间的缺陷无序体系ε-FeCx (x=0.5)。常用的枚举方法需要分别显式计算2664、1033和10496个构型，而“辣搜”方法只需要分别显式计算约15、20和10个构型。除了有限尺寸体系之外，我们的“辣搜”方法也适用于准无限尺寸体系，以增强材料设计能力。
编辑概述
搭乘“特快列车”的异相催化原子模拟
异相催化是化学的核心问题之一。她不仅为整个人类社会带来了巨大收益（如合成氨制氮肥），还因自身的复杂性持续推动着化学前沿的发展。越来越多的证据表明，在催化过程中，真正的活性位点往往是原位形成的，而限于当今实验表征手段难以应对催化中的高温高压条件，使得基于密度泛函（DFT）的理论计算与模拟成为了解原子级别催化过程的有力手段。当前，随着人工智能浪潮对传统化学的冲击，基于机器学习势函数的方法在计算速度上远超DFT，这促使新时代背景下异相催化原子模拟研究的更新换代。基于上述背景，复旦大学化学系刘智攀教授团队从结构和反应两个角度，综述了近年来机器学习在异相催化领域的技术突破。现从中各选一例进行介绍。从结构角度，由于机器学习势函数的出现，人们可以将全局优化算法应用于复杂势能面全局最小值（即给定初始构型下最稳定结构）的搜索，且这一过程甚至可与DFT单步优化耗时拥有近乎相同的数量级。对于原子数可变（巨正则系综）条件下发生的复杂催化反应，其优势结构便可通过给定多种初始构型，采用并行全局优化计算的方法，对若干拥有不同原子数的稳定结构进行热力学稳定性的最终筛选与判别。在此思想上进一步优化，图1给出了此类方法在银表面氧化物上的应用实例。可以看到，在不依赖于任何实验先验下，其可给出不同原子数银氧表面结构在给定催化条件下的势能图，从中可清晰看到稳定结构对应的分布区域，以及前几名稳定结构的构型及其热力学稳定性上的差异，这无疑对实际催化过程（高温高压）下催化剂表面结构的认知具有直接的指导意义。从反应角度，机器学习势函数的出现使得过渡态的搜索十分廉价，从而为大规模反应路径的自动探索提供了可能。例如，首先可通过对初始反应物进行随机势能面搜索，采样可能的产物以构成反应对，然后进行大规模并行的过渡态搜索，得到若干包含过渡态信息的化学反应数据库。通过将能垒最低的产物重置为反应物并不断重复这一过程，便可向后延续整个反应网络的探索，完善反应数据库。如果能对反应数据库进行机器学习，则可实现对给定反应物进行产物和能垒的共同预测，逐级、全自动地构建反应网络。图2的示例实现了这一设想，其通过学习铜表面低碳化合物的反应数据库，在无实验先验的条件下给出了甘油分解的反应路径，提出的新反应通道解释了实验观测到的1,2-丙二醇高选择性之谜。
|
|
|
|
In-plane ferroelectric tunnel junctions based on 2D α-In2Se3/semiconductor heterostructures
Zifang Liu, Pengfei Hou, Lizhong Sun, Evgeny Y. Tsymbal, Jie Jiang & Qiong Yang
npj Computational Materials 9:12 (2022)
doi.org/10.1038/s41524-022-00953-x
Abstract: Ferroelectric tunnel junctions (FTJs) have great potential for application in high-density non-volatile memories. Recently, α-In2Se3 was found to exhibit robust in-plane and out-of-plane ferroelectric polarizations at a monolayer thickness, which is ideal to serve as a ferroelectric component in miniaturized electronic devices. In this work, we design two-dimensional van der Waals heterostructures composed of an α-In2Se3 ferroelectric and a hexagonal IV-VI semiconductor and propose an in-plane FTJ based on these heterostructures. Our first-principles calculations show that the electronic band structure of the designed heterostructures can be switched between insulating and metallic states by ferroelectric polarization. We demonstrate that the in-plane FTJ exhibits two distinct transport regimes, tunneling and metallic, for OFF and ON states, respectively, leading to a giant tunneling electroresistance effect with the OFF/ON resistance ratio exceeding 1 ？ 104. Our results provide a promising approach for the high-density ferroelectric memory based on the 2D ferroelectric/semiconductor heterostructures.
摘要: 化学无序材料具有广泛的应用，而其结构或构型的确定是最重要和最具挑战性的问题之一。由于众所周知的指数墙问题(N-体体系中可能的结构数量呈指数增长)，传统方法对于大体系是非常低效或棘手的。在这里，我们介绍了一种通过主动学习结合第一性原理计算来预测化学无序材料的热力学稳定结构的有效方法。我们的“辣搜”方法可以有效地压缩采样空间，极大地降低计算成本。我们研究了三种不同的、典型的有限尺寸体系，包括阴离子无序体系BaSc(OxF1？x)3 (x=0.667)，具有较大尺寸的阳离子无序体系Ca1？xMnxCO3 (x=0.25)和具有较大采样空间的缺陷无序体系ε-FeCx (x=0.5)。常用的枚举方法需要分别显式计算2664、1033和10496个构型，而“辣搜”方法只需要分别显式计算约15、20和10个构型。除了有限尺寸体系之外，我们的“辣搜”方法也适用于准无限尺寸体系，以增强材料设计能力。
编辑概述
基于二维范德华异质结能带调控的平面铁电隧道结
基于电子器件小型化、非挥发、低功耗等发展需求，铁电隧道结被认为是最有前景的信息存储器之一。传统钙钛矿铁电薄膜材料在小尺寸下的性能退化是遏制铁电隧道结和相关铁电器件小型化的重要因素。CuInP2S6、In2Se3、HfO2等在原子尺度或数纳米下仍具有稳定铁电性的铁电薄膜的出现，为基于铁电材料的小型化电子器件的发展带来了契机。来自湘潭大学材料科学与工程学院的杨琼教授和姜杰副教授团队利用单层厚度的α-In2Se3铁电体和六方IV-VI族半导体(SnTe and PbSe)，设计了一类二维铁电/半导体范德华异质结。第一性原理研究表明，该异质结能够在α-In2Se3铁电极化翻转的调控下，实现从宽带隙绝缘态到金属态的转变。基于该类异质结显著的能带调控性质，研究者提出了一种新型的面内二维铁电隧道结。该铁电隧道结以α-In2Se3/IV-VI族半导体异质结为势垒区、以重电子掺杂的异质结作为虚拟电极。由于α-In2Se3面内、外协同的铁电极化，可以通过施加面内或面外的电场控制铁电隧道结整个势垒区的极化状态和电子能带结构，从而实现隧道结的开和关态。该铁电隧道结并非依靠传统的界面非对称性来调控电子隧穿势垒，因而也不依赖于电极的类型。研究表明，设计的平面铁电隧道结的开/关状态呈现金属导电和电子隧穿两种截然不同的导电机制，从而达到1 ？ 104以上的关/开电阻比，且其电致电阻效应随着势垒宽度的增加而显著增加。该类异质结也可以用于面外铁电隧道结。该研究结果为基于二维和超薄铁电材料的高密度存储器提供了有效方案。
|
|
|
|
Active learning to overcome exponential-wall problem for effective structure prediction of chemical-disordered materials
Xiaoze Yuan, Yuwei Zhou, Qing Peng, Yong Yang, Yongwang Li & Xiaodong Wen
npj Computational Materials 9:12 (2022)
doi.org/10.1038/s41524-023-00967-z
Abstract: Chemical-disordered materials have a wide range of applications whereas the determination of their structures or configurations is one of the most important and challenging problems. Traditional methods are extremely inefficient or intractable for large systems due to the notorious exponential-wall issue that the number of possible structures increase exponentially for N-body systems. Herein, we introduce an efficient approach to predict the thermodynamically stable structures of chemical-disordered materials via active-learning accompanied by first-principles calculations. Our method, named LAsou, can efficiently compress the sampling space and dramatically reduce the computational cost. Three distinct and typical finite-size systems are investigated, including the anion-disordered BaSc(OxF1？x)3 (x=0.667), the cation-disordered Ca1？xMnxCO3 (x=0.25) with larger size and the defect-disordered ε-FeCx (x=0.5) with larger space. The commonly used enumeration method requires to explicitly calculate 2664, 1033, and 10496 configurations, respectively, while the LAsou method just needs to explicitly calculate about 15, 20, and 10 configurations, respectively. Besides the finite-size system, our LAsou method is ready for quasi-infinite size systems empowering materials design.
摘要: 化学无序材料具有广泛的应用，而其结构或构型的确定是最重要和最具挑战性的问题之一。由于众所周知的指数墙问题(N-体体系中可能的结构数量呈指数增长)，传统方法对于大体系是非常低效或棘手的。在这里，我们介绍了一种通过主动学习结合第一性原理计算来预测化学无序材料的热力学稳定结构的有效方法。我们的“辣搜”方法可以有效地压缩采样空间，极大地降低计算成本。我们研究了三种不同的、典型的有限尺寸体系，包括阴离子无序体系BaSc(OxF1？x)3 (x=0.667)，具有较大尺寸的阳离子无序体系Ca1？xMnxCO3 (x=0.25)和具有较大采样空间的缺陷无序体系ε-FeCx (x=0.5)。常用的枚举方法需要分别显式计算2664、1033和10496个构型，而“辣搜”方法只需要分别显式计算约15、20和10个构型。除了有限尺寸体系之外，我们的“辣搜”方法也适用于准无限尺寸体系，以增强材料设计能力。
Editorial Summary
LAsou: highly efficient structure prediction for chemical-disordered materials
To determine the microstructure of materials is the premise of understanding and modifying materials. Here we focus on studying the structure of chemical-disordered materials. Here, the term ‘chemical-disordered materials’ stems from the semi-ordered materials whose lattice is periodic (thus crystal) but the occupying atom species are non-periodic in space. From the point view of chemical compositions, the chemical-disordered materials can be classified into anionic, cationic, and defected counterparts, which can be simply considered as the anions, cations and defects occupy the non-periodic sites. Chemical-disordered materials are widely used in many areas including semiconductors, high-temperature superconductors, Li-ion batteries, metal alloys, ceramics, and heterogeneous catalysts due to their special properties and performances. It is very important to determine the structure of disordered materials for further study of the properties of disordered materials. The exact structure of chemical-disordered materials has not been solved because of the uncertainty of atomic occupancy at some lattice sites. Structural prediction of chemical-disordered materials is challenging and becomes more complex with increasing system size. Previous studies have used crystal symmetry, cluster expansion, and machine learning potential to accelerate structural prediction of chemical-disordered materials, but still need a large number of expensive first-principles calculations. Xiao-dong Wen and Yu-Wei Zhou from Institute of Coal Chemistry, CAS/Synfuels China Technology Co. Ltd, and Qing Peng from Institute of Mechanics, CAS developed the LAsou method for highly efficient structure prediction of chemical-disordered materials. Based on the active learning strategy, LAsou method can quickly predict the thermodynamically stable structure of chemical-disordered materials with only a few first-principles calculations. Tests on three different typical finite-size systems, which included the anion-disordered BaSc(OxF1？x)3 (x=0.667), the cation-disordered Ca1？xMnxCO3 (x=0.25) and the defect-disordered ε-FeCx (x=0.5) show that the LAsou method can quickly find thermodynamically stable structures with very little first-principles calculations compared to traditional enumeration methods. Besides the finite-size system, LAsou method will be helpful for a wide range of applications for the larger, more complex, quasi-infinite size systems and the new materials that occurs in nanoparticles, catalysts, solid solutions, high-entropy alloys and high-entropy oxides, and so on.
编辑概述
辣搜：化学无序材料的高效结构预测
确定材料的微结构是认识和改性材料的前提，尤其是确定化学无序材料的结构。“化学无序材料”源于“半无序材料”，是指周期性的晶格被不同元素原子随机占据的材料。从化学成分的角度来看，化学无序材料可分为阴离子、阳离子和缺陷对应体材料，可以简单地认为是阴离子、阳离子和缺陷占据了非周期位点。化学无序材料由于其独特的性质，广泛应用半导体、高温超导体、金属合金、陶瓷和沸石催化剂等领域。研究无序材料的结构对于理解无序材料性质和指导实验具有非常重要的价值。因为部分晶格位点的原子占据不确定性，化学无序材料的确切结构一直难以解决，其结构预测更是极具挑战，随着体系尺寸增加预测会变得愈加复杂。目前可利用晶体对称性除重、集团展开和机器学习势等技术，来加速化学无序材料的结构预测，但仍面临大量昂贵的第一性原理计算。来自中科院山西煤化所/中科合成油技术股份有限公司的温晓东研究员、周余伟博士和中科院力学所的彭庆研究员组成的联合团队，创建了“辣搜”(Large space sampling and Active labeling for searching)方法，用来作化学无序材料的高效结构预测。“辣搜”方法基于主动学习策略，仅用少量第一性原理计算就可快速预测化学无序材料的热力学稳定结构。三种不同的、典型的有限尺寸体系测试表明，阴离子无序的BaSc(OxF1？x)3、阳离子无序的Ca1？xMnxCO3和缺陷无序的ε-FeCx与传统枚举法相比，“辣搜”方法在仅需非常少的第一性原理计算均可快速找到热力学稳定的结构。除了有限尺寸体系外，“辣搜”方法还有望广泛应用于更大、更复杂、准无限尺寸体系，以及纳米颗粒、催化剂、固溶体、高熵合金和高熵氧化物等新材料。
|
|
|
|
A Generative Deep Learning Framework for Inverse Design of Compositionally Complex Bulk Metallic Glasses
Ziqing Zhou, Yinghui Shang, Xiaodi Liu & Yong Yang
npj Computational Materials 9:15 (2022)
doi.org/10.1038/s41524-023-00968-y
Abstract: The design of bulk metallic glasses (BMGs) via machine learning (ML) has been a topic of active research recently. However, the prior ML models were mostly built upon supervised learning algorithms with human inputs to navigate the high dimensional compositional space, which becomes inefficient with the increasing compositional complexity in BMGs. Here, we develop a generative deep-learning framework to directly generate compositionally complex BMGs, such as high entropy BMGs. Our framework is built on the unsupervised Generative Adversarial Network (GAN) algorithm for data generation and the supervised Boosted Trees algorithm for data evaluation. We studied systematically the confounding effect of various data descriptors and the literature data on the effectiveness of our framework both numerically and experimentally. Most importantly, we demonstrate that our generative deep learning framework is capable of producing composition-property mappings, therefore paving the way for the inverse design of BMGs.
摘要: 通过机器学习设计大块金属玻璃一直是近年来材料科学中活跃的研究课题。然而，之前的机器学习模型主要基于监督学习的算法并通过人工输入来寻找块体金属玻璃的可能成分。然而，随着金属玻璃成分复杂性的增加，这种方法变得效率低下。为了克服这个问题，我们提出一个基于生成式深度学习的框架来直接生成成分复杂的块体金属玻璃(例如高熵金属玻璃)。我们的深度学习框架通过无监督型的生成对抗网络 (generative adversarial network) 算法来产生金属玻璃的成分，并利用监督型的提升树(boosted trees)算法来进行数据评估。我们进一步通过数值模型和大量实验系统地研究了各种数据描述符以及已有的文献数据对此框架有效性的影响。我们证明了这种生成式深度学习框架能够产生属于金属玻璃的成分属性关系，从而为复杂成分金属玻璃的逆向设计铺平道路。
Editorial Summary
Inverse Design of Metallic Glasses via Artificial Intelligence: From Properties to Compositions
In materials science, it has been an issue of scientific importance to find alloy systems that can form bulk glasses with target properties, which also has broad application prospects and commercial value. However, traditional approaches to obtain bulk metallic glasses with target properties through trial and error are time-consuming and costly, especially for metallic glasses with complex compositions, such as high-entropy metallic glasses. Professor Yang Yong's team from the Department of Mechanical Engineering, City University of Hong Kong, recently proposed a deep learning framework that combines the unsupervised algorithm of generative adversarial networks and the supervised algorithm of boosted trees that could enable the reverse design of compositionally complex bulk metallic glass. Through this deep learning framework, people can generate the composition-property relationships of bulk metallic glasses from existing experimental data, thereby facilitating the design of bulk metallic glasses from properties to composition (reverse design). At present, this deep learning framework has been applied to generate metallic glass compositions with target glass transition temperature, crystallization temperature and liquidus temperature, which have been experimentally verified. With this deep learning framework, it is possible to generate octanary alloys that can form bulk metallic glasses. The outcome of the research may greatly accelerate the research and development of bulk metallic glasses with unique properties.
编辑概述
人工智能逆向设计金属玻璃：从性能到成分
以数据为中心的科学已被确定为科学研究的第四个范式。这种范式引入两种新颖的科学研究方法：一方面是创建大型的、相互关联的科学数据库；另一方面是利用人工智能算法研究科学数据，来探索很难通过人类观察得到的模式和趋势。在过去几年中，材料科学在这两个方面都取得了进展。NOMAD存储库和存档在2014年底被构建，它是第一个针对计算材料科学数据的可查找、可访问、可互操作和可重复使用的存储设施。它存储了50多种不同原子尺度代码的输入和输出文件，具有超过1亿个总能量计算。NOMAD通过元数据模式对数据进行转换、标准化和特征化，使数据便于进行AI分析。在本工作中，为了更方便地利用NOMAD，来自柏林洪堡大学的Luigi Sbailò等人，推出了NOMAD人工智能工具包。这个工具包主要有三方面应用：1. 通过最先进的AI工具提供一个API和库来访问和分析NOMAD存档数据。2. 提供一套浅显易懂的教程，从AI技术的实践入门到掌握为止。3. 维护一个社区驱动的、不断增长的计算笔记集合。通过提供带注释的数据和分析脚本，来自世界各地的学生和学者都能够追溯原始研究人员所遵循的所有步骤，以达到发表级的结果。NOMAD人工智能工具包的主要特点在于将存储在NOMAD存档中的数据与其人工智能分析连接在同一基础设施中。此外，用户在同一环境中拥有所有可用的人工智能工具以及访问NOMAD数据的权限，无需安装任何内容。该工具包的发展将有助于提高数据驱动的材料科学论文的可重复性，并降低新人进入该领域的学习壁垒。如何找到能形成块体玻璃的合金体系是一个长期以来人们关注的科学问题。在此之上，如何设计具有目标性能的块体金属玻璃更是材料科学中的一个热点问题，并且具有广阔的应用前景和商业价值。然而，传统方法通过不断试错来获得具有目标性能的块体金属玻璃不仅耗时而且成本高昂，尤其对于具有复杂成分的金属玻璃（例如高熵金属玻璃）更是如此。来自香港城市大学机械工程系的杨勇教授团队，最近提出了一种结合无监督型(unsupervised)的生成式对抗神经网络和监督型的（supervised）提升树的深度学习框架，可用于复杂成分块体金属玻璃的逆向设计。通过该深度学习框架，人们可从已有的实验数据中生成块体金属玻璃成分性能关系，从而完成块体金属玻璃从性能到成分的设计（逆向设计）。目前此深度学习框架已被应用于生成具有目标玻璃转变温度，晶化温度以及液相线温度的金属玻璃成分，并且获得了实验验证。而且通过此深度学习模型，可以生成可形成块体金属玻璃的八元合金。这极大地加快了具有独特性能的块状金属玻璃的研究和开发。
|
|
|
|
The NOMAD Artificial-Intelligence Toolkit: turning materials-science data into knowledge and understanding
Luigi Sbailò, ádám Fekete, Luca M. Ghiringhelli & Matthias Scheffler
npj Computational Materials 8:250 (2022)
doi.org/10.1038/s41524-022-00935-z
Abstract: We present the Novel-Materials-Discovery (NOMAD) Artificial-Intelligence (AI) Toolkit, a web-browser-based infrastructure for the interactive AI-based analysis of materials-science findable, accessible, interoperable, and reusable (FAIR) data. The AI Toolkit readily operates on the FAIR data stored in the central server of the NOMAD Archive, the largest database of materials-science data worldwide, as well as locally stored, users’ owned data. The NOMAD Oasis, a local, stand-alone server can be also used to run the AI Toolkit. By using Jupyter notebooks that run in a web-browser, the NOMAD data can be queried and accessed; data mining, machine learning, and other AI techniques can be then applied to analyze them. This infrastructure brings the concept of reproducibility in materials science to the next level, by allowing researchers to share not only the data contributing to their scientific publications, but also all the developed methods and analytics tools. Besides reproducing published results, users of the NOMAD AI toolkit can modify the Jupyter notebooks toward their own research work.
摘要: 我们提出了Novel-Materials-Discovery（NOMAD）人工智能（AI）工具包，这是一个基于Web浏览器的基础设施，用于可查找、可访问、可互操作和可重复使用（FAIR）数据的基于AI的交互式材料分析。AI工具包可以轻松操作存储在NOMAD Archive的中央服务器中存储的FAIR数据，该数据库是全球最大的材料科学数据库，同时也可以操作本地存储的用户数据。本地独立服务器NOMAD Oasis也可以用于运行AI工具包。通过在Web浏览器中运行的Jupyter笔记本，可以查询和访问NOMAD数据；然后可以应用数据挖掘、机器学习和其他AI技术进行分析。该工具包将材料科学中的可重复性概念提升到了一个新的水平，使研究人员不仅可以分享其科学出版物中所涉及的数据，还可以分享所有开发的方法和分析工具。除了重现已发表的结果外，NOMAD AI工具包的用户还可以修改Jupyter笔记本以适用于自己的研究工作。
Editorial Summary
NOMAD Artificial-Intelligence Toolkit
Data-centric science has been identified as the 4th paradigm of scientific research. It is observed that the novelty introduced by this paradigm is twofold. First, the creation of large, interconnected databases of scientific data; On the other hand, it is to use artificial intelligence algorithms to study scientific data in order to explore patterns and trends that are difficult to observe through human observation. In the past few years, material science has made progress in both areas. The NOMAD Repository & Archive was constructed in late 2014, which is the first Findable, Accessible, Interoperable, and Reusable (FAIR) storage facility for computational materials science data. It stores input and output files of more than 50 different atomistic codes, with over 100 million total energy calculations. The NOMAD converted, normalized, and characterized data through metadata schemas, making it ready for AI analysis. To facilitate the use of these databases, in this work Luigi Sbailò and colleagues from Humboldt University in Berlin have presented the NOMAD AI Toolkit. This toolkit has three main applications:1. Providing an API and libraries for accessing and analyzing the NOMAD Archive data via state-of-the-art (and beyond) AI tools. 2. Providing a set of shallow-learning-curve tutorials from the hands-on introduction to the mastering of AI techniques. 3. Maintaining a community-driven, growing collection of computational notebooks, each dedicated to an AI-based materials-science publication. By providing both the annotated data and the scripts for their analysis, students and scholars worldwide are enable to retrace all the steps that the original researchers followed to reach publication-level results. The main specificity of the NOMAD AI toolkit is in connecting within the same infrastructure the data, as stored in the NOMAD Archive, to their AI analysis. Moreover, users have in the same environment all available AI tools as well as access to the NOMAD data, without the need to install anything. This will allow for enhanced reproducibility of data-driven materials science papers and dampen the learning curve for newcomers to the field.
编辑概述
NOMAD人工智能工具包
以数据为中心的科学已被确定为科学研究的第四个范式。这种范式引入两种新颖的科学研究方法：一方面是创建大型的、相互关联的科学数据库；另一方面是利用人工智能算法研究科学数据，来探索很难通过人类观察得到的模式和趋势。在过去几年中，材料科学在这两个方面都取得了进展。NOMAD存储库和存档在2014年底被构建，它是第一个针对计算材料科学数据的可查找、可访问、可互操作和可重复使用的存储设施。它存储了50多种不同原子尺度代码的输入和输出文件，具有超过1亿个总能量计算。NOMAD通过元数据模式对数据进行转换、标准化和特征化，使数据便于进行AI分析。在本工作中，为了更方便地利用NOMAD，来自柏林洪堡大学的Luigi Sbailò等人，推出了NOMAD人工智能工具包。这个工具包主要有三方面应用：1. 通过最先进的AI工具提供一个API和库来访问和分析NOMAD存档数据。2. 提供一套浅显易懂的教程，从AI技术的实践入门到掌握为止。3. 维护一个社区驱动的、不断增长的计算笔记集合。通过提供带注释的数据和分析脚本，来自世界各地的学生和学者都能够追溯原始研究人员所遵循的所有步骤，以达到发表级的结果。NOMAD人工智能工具包的主要特点在于将存储在NOMAD存档中的数据与其人工智能分析连接在同一基础设施中。此外，用户在同一环境中拥有所有可用的人工智能工具以及访问NOMAD数据的权限，无需安装任何内容。该工具包的发展将有助于提高数据驱动的材料科学论文的可重复性，并降低新人进入该领域的学习壁垒。
|
|
|
|
Inverse design of metal–organic frameworks for C2H4/C2H6 separation
Musen Zhou & Jianzhong Wu
npj Computational Materials 8:256 (2022)
doi.org/10.1038/s41524-022-00946-w
Abstract: Efficient separation of C2H4/C2H6 mixtures is of paramount importance in the petrochemical industry. Nanoporous materials, especially metal-organic frameworks (MOFs), may serve the purpose owing to their tailorable structures and pore geometries. In this work, we propose a computational framework for high-throughput screening and inverse design of high-performance MOFs for adsorption and membrane processes. High-throughput screening of the computational-ready, experimental (CoRE 2019) MOF database leads to materials with exceptionally high ethane-selective adsorption selectivity (LUDLAZ: 7.68) and ethene-selective membrane selectivity (EBINUA02: 2167.3). Moreover, the inverse design enables the exploration of broader chemical space and identification of MOF structures with even higher membrane selectivity and permeability. In addition, a relative membrane performance score (rMPS) has been formulated to evaluate the overall membrane performance relative to the Robeson boundary. The computational framework offers guidelines for the design of MOFs and is generically applicable to materials discovery for gas storage and separation.
摘要: C2H4/C2H6混合物的高效分离对于石化工业至关重要。纳米多孔材料，特别是金属有机框架（MOFs），由于其可定制的结构和孔隙几何形状，有望实现这一目标。在本工作中，我们提出了一个计算框架，对吸附工艺和膜工艺中具有优异性能的MOF进行高通量筛选和逆向设计。对“计算就绪、实验性”MOF数据库CoRE 2019的高通量筛选结果表明，LUDLAZ对乙烷具有极高的吸附选择性，可达7.68；EBINUA02对乙烯具有极高的膜选择性，可达2167.3。此外，逆向设计能够探索更加广阔的化学空间、识别具有更高的膜选择性和渗透性的MOF结构。我们还制定了一个相对膜性能评分（rMPS），用于评估相对于Robeson边界的整体膜性能。该计算框架为MOF的设计提供了指导，并且适用于气体储存与分离材料的发现。
Editorial Summary
Inverse design of MOFs: Efficient separation of C2H4/C2H6
The efficiency of C2H4/C2H6 separation is important for the petrochemical industry because high-purity C2H4 is used as the primary feedstock for the synthesis of diverse chemical products, including plastics, polyesters, and rubber materials. Conventional processes for C2H4/C2H6 separation are mostly based on high-pressure cryogenic distillation, which requires extensive energy input while suffering from low separation efficiency. To reduce the energy cost and increase the selectivity, it is desirable to develop alternative approaches such as adsorption or permeation processes based on nanoporous materials. Metal–organic frameworks (MOFs) are ideal candidates for efficient separation of C2H4/C2H6 because they have good mechanical stability, large specific surface area, and tailorable pore structure and geometry. In particular, such materials show promising performance for separating molecules with similar size and interaction energy. Although computational methods (e.g., MD, grand canonical Monte Carlo simulation, and density functional theory) have been well established for accurate prediction of gas adsorption and diffusivity, the inverse design of nanoporous materials for separation processes remains a theoretical challenge from both computational and practical perspectives. In this work, Musen Zhou et al. from the Department of Chemical and Environmental Engineering of University of California, Riverside, proposed a computational framework for high-throughput screening and inverse design of high-performance MOFs for adsorption and membrane processes. High-throughput screening of the computational-ready, experimental (CoRE 2019) MOF database leads to materials with exceptionally high ethane-selective adsorption selectivity (LUDLAZ: 7.68) and ethene-selective membrane selectivity (EBINUA02: 2167.3). For the adsorption separation, the separation selectivity of ethene-selective MOF decreases with the increase of separation capacity, while the selectivity of ethane-selective MOF increases with the adsorption capacity. Compared with that in the adsorption process, the selectivity of the membrane process is less compromised by the increase of the separation capacity. Moreover, the inverse design enables the exploration of broader chemical space and identification of MOF structures with even higher membrane selectivity and permeability. In addition, a relative membrane performance score (rMPS) has been formulated to evaluate the overall membrane performance relative to the Robeson boundary. High rMPS favors MOFs with high permeability and intermediate membrane selectivity because high membrane selectivity requires a large energy barrier along the minimum energy path and leads to slow diffusion. The computational framework offers guidelines for the design of MOFs and is generically applicable to materials discovery for gas storage and separation.
编辑概述
MOF逆向设计实现C2H4/C2H6高效分离
由于高纯度乙烯是塑料、聚酯纤维、橡胶等多种化学产品合成的主要原料， C2H4/C2H6的分离效率对于石化工业至关重要。传统的C2H4/C2H6分离工艺大多基于高压低温蒸馏，该方法不仅需要投入大量的能量，而且分离效率较低。为降低能量损耗并提高选择性，亟需开发出替代方案，如基于纳米多孔材料的吸附与渗透工艺。金属有机框架（MOF）具有良好的机械稳定性、较大的比表面积以及可定制的孔隙结构和几何形状，是高效分离C2H4/C2H6的理想候选材料。这类材料在分离具有相似尺寸和相互作用能的分子方面表现出优异的性能。虽然分子动力学、巨正则蒙特卡罗模拟和密度泛函理论等计算方法已经可以准确地预测气体吸附和扩散率，但用于分离过程纳米多孔材料的反设计仍然是一个理论挑战。在本工作中，来自美国加州大学河滨分校化学与环境工程系的Musen Zhou等人，提出了一个计算框架，可以对吸附工艺和膜工艺中具有优异性能的MOF进行高通量筛选和逆向设计。对“计算就绪、实验性”MOF数据库CoRE 2019的高通量筛选结果表明，LUDLAZ对乙烷具有极高的吸附选择性，可达7.68；EBINUA02对乙烯具有极高的膜选择性，可达2167.3。对于吸附分离而言，具有乙烯选择性的MOF材料的分离性能随着吸附容量的增加而降低，而具有乙烷选择性的MOF材料的分离性能随着吸附容量的增加而增加。与吸附工艺相比，吸附容量的增加对膜工艺的选择性影响较小。此外，逆向设计能够探索更加广阔的化学空间、识别具有更高的膜选择性和渗透性的MOF结构。研究者还制定了一个相对膜性能评分（rMPS），用于评估相对于Robeson边界的整体膜性能。渗透性高、膜选择性适中的MOF具有较高的rMPS，膜选择性过高会导致气体沿着最小能量路径需要克服较大的能垒，从而扩散缓慢。本工作提出的计算框架为MOF的设计提供了指导，并且适用于气体储存与分离材料的发现。
|
|
|
|
Forecasting of in situ electron energy loss spectroscopy
Nicholas R. Lewis, Yicheng Jin, Xiuyu Tang, Vidit Shah, Christina Doty, Bethany E. Matthews, Sarah Akers & Steven R. Spurgeon
npj Computational Materials 8:252 (2022)
doi.org/10.1038/s41524-022-00940-2
Abstract: Forecasting models are a central part of many control systems, where high-consequence decisions must be made on long latency control variables. These models are particularly relevant for emerging artificial intelligence (AI)-guided instrumentation, in which prescriptive knowledge is needed to guide autonomous decision-making. Here we describe the implementation of a long short-term memory model (LSTM) for forecasting in situ electron energy loss spectroscopy (EELS) data, one of the richest analytical probes of materials and chemical systems. We describe key considerations for data collection, preprocessing, training, validation, and benchmarking, showing how this approach can yield powerful predictive insight into order-disorder phase transitions. Finally, we comment on how such a model may integrate with emerging AI-guided instrumentation for powerful high-speed experimentation.
摘要: 预测模型是许多控制系统的一个中心部分，必须对长时间延控制变量做出高风险决策。预测模型与新兴人工智能（AI）引导的仪器尤其相关，都需要规定性知识来指导自主决策。在这里，我们描述了一个长短期记忆模型（LSTM），用于对材料和化学系统中最为丰富的分析探针—原位电子能量损失谱（EELS）进行数据预测。我们描述了数据收集、预处理、训练、验证和基准测试等关键因素，并展示了该方法对有序-无序相变强大的预测洞察力。最后，我们讨论了这一模型将如何与新兴人工智能引导的仪器进行集成，从而用于强大的高速实验。
Editorial Summary
Forecasting of in situ electron energy loss spectroscopy
Effective forecasting is essential for many disciplines and technologies, ranging from meteorology to the power grid and from stock trading to logistics. When performed correctly, forecasting can save time, reduce cost, and guide scientific discovery by helping direct decision-making. Forecasting models are a central part of many control systems, where high-consequence decisions must be made on long latency control variables. These models are particularly relevant for emerging artificial intelligence guided instrumentation, in which prescriptive knowledge is needed to guide autonomous decision-making. For many studies, both ex and in situ, we must make rapid decisions on high-latency control parameters using information from high-throughput, multimodal data streams. However, we currently lack the necessary low-level control, descriptive models, and forecasting (prescriptive) approaches to implement more powerful decision-making. In this work, Nicholas R. Lewis et al from the Department of Chemical Engineering, University of Washington, described the implementation of a long short-term memory model for forecasting in situ electron energy loss spectroscopy (EELS) data, one of the richest analytical probes of materials and chemical systems. They explored the crystalline-to-amorphous phase transition in the archetypal perovskite oxide SrTiO3, utilizing the electron beam itself to drive reduction and associated changes in core-loss EELS spectra. The authors systematically explored the key considerations for data preprocessing, model architecture, hyperparameter optimization, training, validation, and benchmarking. The model possesses good predictive power relative to ground truth experimental data and may serve as a basis for future model-predictive control approaches. Forecasting models of the kind shown in this work will find important usage in emerging self-driving microscope platforms. The ability to run these models in real-time and predict a future state of a chemical reaction will allow for optimization of experimental parameters, such as beam electron dose, sampling, and current, and realize richer, more accurate studies of crystal nucleation and growth, battery cycling, mechanical testing, and quantum behavior.
编辑概述
原位电子能量损失谱的预测
无论是从气象学到电网，还是从股票交易到物流，高效的预测对许多学科和技术都至关重要。如果预测正确，不仅可以节省时间、降低成本，而且能够通过帮助直接决策来指导科学发现。预测模型是许多控制系统的核心部分，在这些系统中必须对长时延控制变量做出高风险决策。预测模型与新兴人工智能引导的仪器尤其相关，二者都需要规定性知识来指导自主决策。对于许多研究，无论是离位还是原位，都必须使用高通量、多模态的数据流信息对高时延控制参数做出快速决策。然而，目前缺乏必要的低水平控制、描述性模型和预测（规定性）方法来实现更加强大的决策。在本工作中，来自美国华盛顿大学化学工程系的Nicholas R. Lewis等人，描述了一个长短期记忆模型，用于对材料和化学系统中最为丰富的分析探针—原位电子能量损失谱（EELS）进行数据预测。他们利用电子束自身驱动钙钛矿型氧化物SrTiO3还原，并从芯损失EELS谱的相应变化中探索了SrTiO3的晶态-非晶态相变。作者系统探索了数据预处理、模型架构、超参数优化、训练、验证和基准测试等关键因素。与真实的实验数据相比，该模型具有良好的预测能力，可能成为未来模型预测控制方法的基础。本工作提出的预测模型将在新兴自驱动显微平台中得到重要应用。实时运行这些模型并预测化学反应的未来状态将允许对电子束剂量、采样、电流等实验参数进行优化，从而实现对晶体成核和生长、电池循环、机械测试和量子行为等更加丰富和准确的研究。
|
|
|
|
|
|
沪公网安备 31010502006565号
