• 【回复:四两寻找哀鸣之问】为什么一直辛苦劳作的农民没有富起来?专家解释的4点原因很靠谱 2020-01-18
  • 月薪过万白领辞职回乡养鸡 亏数十万后回城找工作 2019-12-25
  • 85后女青年刘阿娟带领果农发展绿色无公害果品 2019-12-25
  • 习近平应约同俄罗斯总统普京通电话 2019-12-08
  • 安徽九旬教师守候留守儿童:“鞠躬尽瘁,死而不已” 2019-11-27
  • 新华时评:成果不易,更需诚意与信任呵护 2019-11-25
  • 绍兴印染布质量比对成果发布 部分指标国际领先 2019-11-25
  • 当前金融市场流动性合理稳定 2019-10-27
  • 阳泉市春播生产平稳有序 2019-10-27
  • 外卖平台生鲜电商角逐 世界杯引爆外卖夜宵大战 2019-08-23
  • 回复@笑傲江湖V:七千二百多吧。 2019-08-13
  • 四大名著剧组首次同台忆往事 经典影视剧如何铸就? 2019-07-08
  • 老妈见14岁女儿和前闺蜜干架 竟递刀助威:快捅她胸部! 2019-07-08
  • 司法责任制改革:追责保护相统一 给司法人员“定心丸” 2019-06-25
  • 正确区分党徽和党员徽章 2019-06-25
  • 首 页
    滚动信息 更多 >>
    npj Computational Materials获得第一个SCI影响因... (2018-09-07)
    英文刊《npj Computational Materials(计算材料学... (2017-05-15)
    开放的数据库 (2015-12-17)
    上海硅酸盐所与Nature出版集团联合创办的中国首本n... (2015-11-27)
    Nature Publishing Group and Shanghai Institute ... (2015-09-24)
    快捷服务
    最新文章 研究综述
    过刊浏览 作者须知
    期刊编辑 审稿须知
    相关链接
    · 在线投稿
    会议信息
    友情链接
      新时时彩走势
      无机材料学报
      OQMD数据库
    mm.jpg
    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...
    【查看详细】
    近期文章 更多 >>

    Bio-inspired self-folding strategy to break the trade-off between strength and ductility in carbon-nanoarchitected materials (生物启发策略实现碳纳米结构宏观构筑材料强度延性的协同提升) 
    Xiangzheng JiaZe Liu & Enlai Gao
    npj Computational Materials 6:13(2020)
    doi:s41524-020-0279-8
    Published online:05 February 2020

    Abstract| Full Text | PDF OPEN

    摘要:石墨烯具有优异的力学,电学和热学性质,已成为构筑高性能多功能宏观碳材料的最具竞争力的候选构筑基元之一。然而,目前碳纳米结构宏观构筑材料出现了常见的材料强度-延性悖论,并且其若干关键力学性能,例如拉伸强度,仍远低于碳纳米结构单元。得益于折叠结构的构象及其之间的转变,蚕丝蛋白具有优异的力学特性。受此启发,本工作提出了一种自折叠碳纳米结构的拓扑构筑策略,获取兼具高强度高延展性的碳纳米结构宏观构筑材料。这种组装材料的优异力学性能主要归因于自折叠界面间的剪切,滑移以及折叠结构展开的贡献。分子模拟表明折叠几何结构和界面相互作用是调节材料强度与延性的有效参量,同时界面剪切,滑动以及折叠结构的展开可以大幅提高材料变形破坏所耗散的能量。最后,结合原子尺度变形分析与连续介质力学模型,本文讨论了该材料的变形与破坏机制。本工为制备高强高延宏观碳材料的提供了理论基础。 

    Abstract:Graphene possesses extraordinary mechanical, electronic, and thermal properties, thus making it one of the most promising building blocks for constructing macroscopic high performance and multifunctional materials. However, the common material strength–ductility paradox also appears in the carbon-nanoarchitected materials and some of the key mechanical performance, for example, the tensile strength of graphene-based materials, are still far lower than that of graphene. Inspired by the exceptional mechanical performance of silk protein benefiting from the conformations of folded structures as well as their transitions, this work proposed a topological strategy to yield graphene-based materials with ultrahigh ductility while maintaining decent tensile strength by self-folding graphene sheets. This drastically improved mechanical performance of graphene-based materials is attributed to the exploitation of shearing, sliding, and unfolding deformation at the self-folded interface. Molecular dynamics simulations show that both modulating self-folded length and engineering interface interaction can effectively control the strength, ductility, and the ductile failure of van der Waals interfaces among the self-folded structures, where interfacial shearing, sliding, and unfolding open channels to dissipate mechanical energy. Based on the insights into the atomic-scale deformation by molecular dynamics simulations, the underlying mechanism of deformation and failure of these materials is finally discussed with a continuum mechanics-based model. Our findings bring perceptive insights into the microstructure design of strong-yet-ductile materials for load-bearing engineering applications..

    Editorial Summary

    Bio-inspired Strategy: Constructing carbon-nanoarchitected materials having high strength and high ductility生物启发组装策略:构建高强度高延性碳材料

    受生物材料启发,本研究提出并验证了通过折叠碳纳米结构以实现其宏观构筑材料高强度和高延性协同提升的策略。来自武汉大学土建学院的高恩来研究团队(//enlaigao.whu.edu.cn/),报道了一种基于生物启发的拓扑折叠组装策略,可有效地调控碳纳米结构宏观构筑材料的强度和延性。首先,他们开展了自折叠微结构与宏观材料强度-延性的关联研究,构建了自折叠结构的计算模型;然后开展多尺度模拟与分析,获得了自折叠结构特征参数与宏观材料强度-延性的定量关系,揭示载荷在自折叠结构中的传递机制,进而建立嵌入自折叠结构特征参数的力学模型;最后对模型参数开展分析,提出了面向材料强度-延性需求的微结构设计,以及最大限度利用低维碳结构优异力学特性的组装策略。该工作为碳基构筑宏观材料的力学性能发展提供了理论基础,有助于推动新兴高强轻质碳材料的发展。

    In this work, a bio-inspired strategy of self-folding graphene into macroscopic materials is proposed and detailedly investigated.A team led by Enlai Gao from Wuhan University, reported a topological assembly strategy that can effectively tunethe strength and ductility of carbon-nanoarchitected materials. First, they construct a numerical model of the bio-inspired self-folding structures. Second, theyperformMD simulations to investigate the quantitative relationship between the self-folding structures and their mechanical properties, and then clarify the mechanism of load transfer in these structures. Then, a mechanical model embedding characteristic parameters of self-folding structures is constructed. Finally, sensitivity analysis of mechanical model responses is assessed to find the critical parameters for strengthening and toughening design of these materials. The outcomes of this work provide solid theoretical principles for the performance prediction and optimal design of nanocarbon-based oriented materials, and then advance efficient development of these emerging high-strength, lightweight materials.

    Ultralow lattice thermal conductivity of chalcogenide perovskite CaZrSe3 contributes to high thermoelectric figure of merit (硫族化物钙钛矿CaZrSe3的超低晶格热导率有助于高热电性能) 
    Eric Osei-AgyemangChallen Enninful Adu & Ganesh Balasubramanian
    npj Computational Materials 5:116(2019)
    doi:s41524-019-0253-5
    Published online:04 December 2019

    Abstract| Full Text | PDF OPEN

    摘要:一种新兴的硫属钙钛矿CaZrSe3因具有显著的光学和电学性能,在能量转换应用方面拥有广阔的前景。但为能在潜在的热电等领域得到应用,了解其热性能极为重要,而之前还没有系统的相关报道。本研究使用密度泛函理论和玻尔兹曼输运计算,检测和解释了CaZrSe3中的晶格热输运机理。我们发现,CaZrSe3平均弛豫时间非常短,这说明,声子-声子散射得以增强,从而湮灭了声子模态,降低了热导率。另外,Grüneisen参数预测的结果显示,钙钛矿晶体的强非谐性和声模的低声子数密度,共同导致了晶格热导率限于1.17 W m-1 K-1。CaZrSe3块体样品(N→∞)中的声子平均自由程为138.1 nm,将纳米结构的CaZrSe3样品制成约10 nm会使热导率降低至0.23 W m-1 K-1。我们还发现,与n型掺杂相比,p型掺杂导致热电优值(ZT)的预测值更高,空穴浓度在1016–1017 cm-3、温度在600-700 K范围内时,ZT值约为0.95-1。 

    Abstract:An emerging chalcogenide perovskite, CaZrSe3, holds promise for energy conversion applications given its notable optical and electrical properties. However, knowledge of its thermal properties is extremely important, e.g. for potential thermoelectric applications, and has not been previously reported in detail. In this work, we examine and explain the lattice thermal transport mechanisms in CaZrSe3 using density functional theory and Boltzmann transport calculations. We find the mean relaxation time to be extremely short corroborating an enhanced phonon–phonon scattering that annihilates phonon modes, and lowers thermal conductivity. In addition, strong anharmonicity in the perovskite crystal represented by the Grüneisen parameter predictions, and low phonon number density for the acoustic modes, results in the lattice thermal conductivity to be limited to 1.17Wm-1K-1. The average phonon mean free path in the bulk CaZrSe3 sample (N→∞) is 138.1nm and nanostructuring CaZrSe3 sample to ~10nm diminishes the thermal conductivity to 0.23Wm-1K-1. We also find that p-type doping yields higher predictions of thermoelectric figure of merit than n-type doping, and values of ZT ~0.95–1 are found for hole concentrations in the range 1016–1017cm-3 and temperature between 600 and 700K.

    Editorial Summary

    Chalcogenide perovskite: high thermoelectric figure of merit硫族化物钙钛矿:高热电性能

    该研究采用详细的第一性原理计算探索了CaZrSe3中基本的晶格输运机理,然后预测了晶格热导率(kL)和热电优值ZT值。美国Lehigh大学机械工程与力学系的Ganesh Balasubramanian领导的课题组,求解了线性化的Boltzmann输运方程,并计算了声子群速度、声子寿命和平均Grüneisen参数。计算出的平均Grüneisen参数为3.75,表明CaZrSe3中存在强烈的非谐性,这有助于降低kL。约450 K的高德拜温度表明,在较低温度下只有很少的声子模处于活动状态,从而导致较低的晶格热导率。在声子谱的声区内,尽管群速度很高,但声子数密度相对较低。高Grüneisen参数、高德拜温度和声区中降低的声子数密度,表明晶体中存在明显的非谐性,所有这些都导致了CaZrSe3的超低导热率。计算得出在湮灭之前的声子传播热力学极限为138.1 nm。以10 nm的尺度对CaZrSe3作纳米结构化,kL可以进一步降低80%,达到0.23 W m-1 K-1,从而潜在地增大ZT。他们计算了不同温度下不同载流子浓度的ZT值,发现与n型掺杂相比,p型掺杂实现了最大的ZT值。而且,较低的载流子浓度产生相对较高的ZT值。对于p型掺杂,在1015 cm-3的载流子浓度及550 K600 K的温度下,计算出的ZT值为1.004;而对于n型掺杂,在550 K、载流子浓度为1015 cm-3时,所实现的最高ZT值为0.984。由于预测p型掺杂比n型掺杂具有更高的ZT值,在设计人员的CaZrSe3热电器件中,在较低的载流子浓度下、采用p型掺杂而不是n型掺杂,有望实现最佳性能。

    The fundamental lattice transport mechanisms in CaZrSe3 is revealed by employing detailed first principles calculations, and the lattice thermal conductivity (kL) and the thermoelectric figure of merit ZT are subsequently predicted. A group led by Ganesh Balasubramanian from the Department of Mechanical Engineering and Mechanics, Lehigh University, USA, solved the linearized Boltzmann transport equations and calculated the phonon group velocities, phonon lifetimes and average Grüneisen parameter. The calculated average Grüneisen parameter of 3.75 shows strong anharmonicity in CaZrSe3 that contributes to low kL. A high Debye temperature of ~450K indicates that only few phonon modes are active at lower temperatures leading to lower lattice thermal conductivity. The phonon number density is relatively low, albeit with high group velocities, in the acoustic region of the phonon spectra. Considerable anharmonicity in the crystal indicated by the high Grüneisen parameter, a high Debye temperature, a reduced phonon number density in the acoustic region, all contribute towards the ultralow thermal conductivity of CaZrSe3. The calculated thermodynamic limit at which phonons travel before annihilation is 138.1nm. By nanostructuring CaZrSe3 at a scale of ~10nm, kL can be further reduced by ~80% to attain 0.23Wm-1K-1, to potentially enhance ZT. They computed ZT for different carrier concentrations across different temperatures and found that the highest ZT values are achieved for p-type doping compared to n-type doping. Also, lower carrier concentrations yield relatively higher ZT values. For p-type doping, a ZT value of 1.004 is calculated at carrier concentration of 1015cm-3 and temperatures of 550 and 600K. However, for n-type doping the highest ZT value of 0.984 is achieved at 550K at a carrier concentration of 1015cm-3 Since higher ZT is predicted for p-type doping compared to n-type doping, in designer CaZrSe3 thermoelectrics, optimum performance will be achieved at lower carrier concentrations and for p-type doping rather than n-type.

    Topological electronic states in HfRuP family superconductors (HfRuP超导家族中的拓扑电子态) 
    Yuting Qian, Simin Nie, Changjiang Yi, Lingyuan Kong, Chen Fang, Tian Qian, Hong Ding, Youguo Shi,Zhijun Wang, Hongming Weng & Zhong Fang
    npj Computational Materials 5:121(2019)
    doi:s41524-019-0260-6
    Published online:10 December 2019

    Abstract| Full Text | PDF OPEN

    摘要:基于第一性原理计算和实验测量,我们报道了六方相的过渡金属磷化物TT'XT = Zr,Hf; T'= Ru; X = P,As)具有非平庸拓扑性质。这类材料同时也是广为人知的具有较高超导转变温度的非中心对称超导体。我们发现, 在超导转变之前,HfRuP属于外尔半金属相,具有12对第二类外尔点,而ZrRuAs、ZrRuPHfRuAs则属于具有平庸的Fu-Kane Z2指数,但非平庸的镜面陈数的拓扑晶体绝缘体相。我们制备了具有这两种不同拓扑态的高质量单晶样品,并通过实验证实了它们的超导性。ZrRuAs的大范围的能带结构被ARPES实验观测到,同时与理论计算结果符合得相当好。结合内禀的超导性,正常态的非平庸拓扑特性可能在体和表面中产生非常规的超导性。我们的发现将激发研究人员在这些化合物中寻找可能的拓扑超导的实验研究。 

    Abstract:Based on the first-principles calculations and experimental measurements, we report that the hexagonal phase of ternary transition metal pnictides TT’X (T=Zr, Hf; T’=Ru; X=P, As), which are well-known noncentrosymmetric superconductors with relatively high transition temperatures, host nontrivial bulk topology. Before the superconducting phase transition, we find that HfRuP belongs to a Weyl semimetal phase with 12 pairs of type-II Weyl points, while ZrRuAs, ZrRuP and HfRuAs belong to a topological crystalline insulating phase with trivial Fu-Kane Z2 indices but nontrivial mirror Chern numbers. High-quality single crystal samples of the noncentrosymmetric superconductors with these two different topological states have been obtained and the superconductivity is verified experimentally. The wide-range band structures of ZrRuAs have been identified by ARPES and reproduced by theoretical calculations. Combined with intrinsic superconductivity, the nontrivial topology of the normal state may generate unconventional superconductivity in both bulk and surfaces. Our findings could largely inspire the experimental searching for possible topological superconductivity in these compounds.

    Editorial Summary

    Superconductors: Topological states翁红明、王志俊研究员携手:本征超导体的拓扑电子态

    本研究基于第一性原理计算和实验测量发现具有较高超导转变温度的一类超导材料HfRuP家族,该家族具有非平庸的拓扑性质:HfRuP属于外尔半金属相,而ZrRuAs、ZrRuPHfRuAs则属于拓扑晶体绝缘体相。来自中国科学院物理研究所的翁红明(本刊副主编)、王志俊研究员领导的理论团队,通过第一性原理计算,发现了HfRuP超导家族材料具有拓扑非平庸电子结构。根据他们的理论发现,石友国,钱天、丁洪研究员领导的实验团队,通过单晶制备和物性测量,进一步确定了这类材料的超导性和拓扑性质。利用本征超导材料的拓扑电子结构性质,相关拓扑超导研究可以有效地克服异质结样品生长、样品掺杂等不利因素。目前本征超导材料FeSeTe的拓扑表面态在体态超导转变温度以下形成的二维拓扑超导能隙已经被ARPES实验观测到,相关STM实验也在紧张进行中。本研究又发掘出了一类新的本征超导材料,利用他们独特的电子结构,三维拓扑超导可通过超导电性与外尔半金属相结合的HfRuP来实现,拓扑镜面超导可在处于拓扑晶体绝缘体相的ZrRuAs、ZrRuPHfRuAs中寻找。与过去预言的需要压力/掺杂才能诱发/增强超导电性的拓扑超导材料相比,HfRuP家族可以得到很好的单晶样品,同时具有较高超导转变温度,为拓扑超导的研究提供了一个非常好的实验平台。

    Based on first-principles calculations and experimental measurements, this study reported that a class of superconducting materials HfRuP family with relatively high transition temperatures has nontrivial topological properties: HfRuP is a Weyl semimetal, ZrRuAs, ZrRuP, and HfRuAs belong to the topological crystalline insulating phase. A theoretical team led by Profs. Zhijun Wang and Hongming Wengfrom from the Institute of Physics, Chinese Academy of Sciences found the topological nontrivial properties of HfRuP family based on first-principles calculations. According to their theoretical findings, the experimental teams led by Profs. Youguo Shi, Tian Qian and Ding Hong further confirmed the superconductivity and topological properties of these materials by the growth of single crystals and physical property measurements. To study topological superconductivity in an intrinsic superconductor with nontrivial topology, it can overcome the disadvantages of the difficulties in the hetero-structure growth and doping experiments. So far, the superconducting gap in the surface of the FeSeTe superconductor has been  observed by the ARPES experiments, which has been proposed to have nontrivial spin-momentum-locking surface Dirac cone states on the 001 surface. Many further STM experiments are in progress. Taking advantage of these unique electronic structures in the HfRuP family, three-dimensional topological superconductor may be achieved by combining superconductivity and Weyl semimetal in HfRuP, and topological mirror superconductor in the topological crystalline insulating phase of ZrRuAs, ZrRuP, or HfRuAs. Unlike other proposed topological superconductor candidates that require pressure/doping to induce/enhance superconductivity, the single crystals of HfRuP family are intrinsic superconductors with relatively high transition temperatures, providing a good material platform for the study of topological superconductors.

    Tailored Morphology and Highly Enhanced Phonon Transport in Polymer Fibers: a Multiscale Computational Framework(聚合物纤维的构型调控和高度增强的声子输运:一个多尺度计算框架)
    Shangchao Lin, Zhuangli CaiYang WangLingling ZhaoChenxi Zhai
    npj Computational Materials 5:126(2019)
    doi:s41524-019-0264-2
    Published online:19 December 2019

    Abstract| Full Text | PDF OPEN

    摘要:尽管研究者们为提高聚合物纤维的热导率做出了巨大的努力,但纤维热拉伸过程参数和所产生的链取向度、结晶度以及声子输运性能之间的关系仍不明朗。通过大尺度分子动力学模拟和仔细训练的粗?;Τ?,我们系统地研究了块体聚乙烯样品的热拉伸参数。我们发现了一个适中的拉伸温度和应变速率的优化组合,可实现聚合物纤维最高程度的链取向度、结晶度和最终的热导率。这种组合可由粘弹性弛豫中的竞争效应合理地解释,并凝聚为无量纲的德博拉数——一个可预测热拉伸方案的指标,反映了应力局域化和分子链扩散的微妙平衡。拉伸变形后,无定形聚乙烯的热导率提高到了理论值极限(即其纯结晶相的热导率)的80%?;诎虢峋Ь酆衔锏拇?/span>-并联热导本质,本研究建立了一个等效介质理论模型,用于预测链取向度和结晶度对热导率的影响。热导率的提高主要归因于本征声子平均自由程和纵波群速度的增加。这项工作从根本上揭示了聚合物热拉伸工艺参数的影响,并为全有机电子设备中的声子输运和聚合物散热器效率的增强建立了完整的过程-结构-属性关系。 

    Abstract:Although tremendous efforts have been devoted to enhance thermal conductivity in polymer fibers, correlation between the thermal-drawing conditions and the resulting chain alignment, crystallinity, and phonon transport properties have remained obscure. Using a carefully trained coarse-grained force field, we systematically interrogate the thermal-drawing conditions of bulk polyethylene samples using large-scale molecular dynamics simulations. An optimal combination of moderate drawing temperature and strain rate is found to achieve highest degrees of chain alignment, crystallinity, and the resulting thermal conductivity. Such combination is rationalized by competing effects in viscoelastic relaxation and condensed to the Deborah number, a predictive metric for the thermal-drawing protocols, showing a delicate balance between stress localizations and chain diffusions. Upon tensile deformation, the thermal conductivity of amorphous polyethylene is enhanced to 80% of the theoretical limit, that is, its pure crystalline counterpart. An effective-medium-theory model, based on the serial-parallel heat conducting nature of semicrystalline polymers, is developed here to predict the impacts from both chain alignment and crystallinity on thermal conductivity. The enhancement in thermal conductivity is mainly attributed to the increases in the intrinsic phonon mean free path and the longitudinal group velocity. This work provides fundamental insights into the polymer thermal-drawing process and establishes a complete process–structure–property relationship for enhanced phonon transport in all-organic electronic devices and efficiency of polymeric heat dissipaters.

    Editorial Summary

    Multiscale Computational Framework: Tailored Morphology and Enhanced Phonon Transport in Polymer Fibers上海交大揭示:聚合物纤维的构型调控和声子输运增强

    本研究通过大尺度分子动力学模拟和仔细训练的粗?;Τ?,完整地建立了聚合物纤维的加工-结构-导热性能的复杂非线性关系。来自上海交通大学的林尚超副教授领导的团队,系统性地研究了纤维热拉伸条件参数对块体聚乙烯链取向度、结晶度和热导率的影响。研究发现一个适中的拉伸温度和应变速率的优化组合可实现最高程度的链取向度、结晶度和最终的热导率。这种组合可由粘弹性弛豫中的竞争效应合理地解释,凝聚为无量纲的德博拉数,反映了应力局域化和分子链扩散效应之间的巧妙平衡。他们建立了一个串-并联复合热导等效介质模型,以预测链取向度和结晶度对热导率的影响。热导率的提高主要归因于本征声子平均自由程和纵波群速度的增加。这项工作从根本上揭示了聚合物热拉伸工艺参数的影响,并为全有机电子设备中的声子输运和聚合物散热器效率的增强建立了完整的加工-结构-属性关系。

    Using large-scale molecular dynamics simulations and a carefully trained coarse-grained force field, this study establishes a complete process–structure–property relationship for heat conduction in polymer fibers. A team led by Shangchao Lin, an Associate Professor from Shanghai Jiao Tong University, systematically interrogated the thermal-drawing conditions of bulk polyethylene samples. An optimal combination of moderate drawing temperature and strain rate is found to achieve highest degrees of chain alignment, crystallinity, and the resulting thermal conductivity. Such combination is rationalized by competing effects in viscoelastic relaxation and condensed to the Deborah number. An effective-medium-theory model is developed here to predict the impacts from both chain alignment and crystallinity on thermal conductivity. The enhancement in thermal conductivity is mainly attributed to the increases in the intrinsic phonon mean free path and the longitudinal group velocity. This work provides fundamental insights into the polymer thermal-drawing process and establishes a complete process–structure–property relationship for enhanced phonon transport in all-organic electronic devices and efficiency of polymeric heat dissipaters.

    Reliable and explainable machine-learning methods for accelerated material discovery(可靠、可解释的机器学习方法加速材料发现)
    Bhavya KailkhuraBrian GallagherSookyung KimAnna Hiszpanski & T. Yong-Jin Han
    npj Computational Materials 5:108(2019)
    doi:s41524-019-0248-2
    Published online:14 November 2019

    Abstract| Full Text | PDF OPEN

    摘要:尽管机器学习(ML)在商业应用中表现出色,但将其应用于材料科学研究却仍存在一些非同寻常的挑战。在这种情况下,本研究有两个重点。首先,当ML从代表性不足或失衡的材料数据中学习时,我们确定了现有ML技术的常见陷阱。具体而言,我们的研究展示,在数据不平衡的情况下,评估ML模型质量的标准方法会崩溃,并产生误导性的结论。此外,我们发现模型自身的置信度得分并不可信,模型自省方法(使用更简单的模型)也无济于事,因为它们会引起预测性能下降(可靠性与可解释性之间的权衡)。其次,为了克服这些挑战,我们提出了一个通用的可解释且可靠的机器学习框架。具体而言,我们提出了一种通用管道,该管道采用较简单模型组合来可靠地预测材料属性。我们还提出了一种转移学习技术,通过利用不同材料特性之间的相关性,来克服因模型简单而导致的性能损失?;固岢隽艘恢中碌钠拦乐副旰鸵桓鲂湃纹婪?,以更好地量化预测的置信度。为提高可解释性,我们在框架中添加了基本原理生成器组件,该组件提供了模型水平和决策水平的解释。最后,我们证明了我们的技术在两类应用中的多功能性:1)预测晶体化合物的特性;2)确定潜在的稳定太阳能电池材料。我们还指出了ML在材料科学中成功应用尚须解决的一些悬而未决的问题。 

    Abstract:Despite ML’s impressive performance in commercial applications, several unique challenges exist when applying ML in materials science applications. In such a context, the contributions of this work are twofold. First, we identify common pitfalls of existing ML techniques when learning from underrepresented/imbalanced material data. Specifically, we show that with imbalanced data, standard methods for assessing quality of ML models break down and lead to misleading conclusions. Furthermore, we find that the model’s own confidence score cannot be trusted and model introspection methods (using simpler models) do not help as they result in loss of predictive performance (reliability-explainability trade-off). Second, to overcome these challenges, we propose a general-purpose explainable and reliable machine-learning framework. Specifically, we propose a generic pipeline that employs an ensemble of simpler models to reliably predict material properties. We also propose a transfer learning technique and show that the performance loss due to models’ simplicity can be overcome by exploiting correlations among different material properties. A new evaluation metric and a trust score to better quantify the confidence in the predictions are also proposed. To improve the interpretability, we add a rationale generator component to our framework which provides both model-level and decision-level explanations. Finally, we demonstrate the versatility of our technique on two applications: 1) predicting properties of crystalline compounds and 2) identifying potentially stable solar cell materials. We also point to some outstanding issues yet to be resolved for a successful application of ML in material science.

    Editorial Summary

    Reliable and explainable machine-learning methods for accelerated material discovery可靠、可解释的机器学习方法:加速材料发现

    尽管机器学习(ML)在商业应用中表现出色,但将其应用于材料科学研究却仍存在一些非同寻常的挑战。在这应用材料信息学过程中,可靠且可解释的机器学习解决方案的构建面临挑战,本研究为应对这一挑战迈出了第一步。本研究的主要贡献是双重的。首先,以代表性不充分和分布失衡的数据作机器学习训练的同时,我们在现有的材料信息学管道中找出了一些训练、测试和不确定性量化步骤中的缺陷。我们的发现引起了人们对现有材料信息学管道可靠性的严重关注。其次,为了克服这些挑战,我们提出了一种通用的、可解释的、可靠的机器学习方法,用于从代表性不足和分布失衡的数据中进行可靠的学习。我们提出了以下解决方案:1)学习架构偏向于训练过程,以实现不平衡域的目标;2)采用抽样方法来操纵训练数据的分布,从而允许使用标准的ML模型;3)可靠的评估指标和不确定性量化方法,以更好地捕获应用程序偏差。与其他针对每个属性而训练独立回归模型的现有方法相反,我们为提高可解释性,采用了一种简单且计算便宜的分区方案。该方案首先根据材料的属性值,将数据划分为若干材料亚类,然后为每组训练单独的更简单的回归模型。通过为每个材料亚类提供特色重要性,这一点可以实现。注意,我们的方法动机(和操作)与Ward等所用的类似概念不同。和朱等与以前的方法相反,我们进行分区的目的是增强“可解释性”,这在以前的方法中,是执行计算量大的穷举搜索法来查找人工组,以提高预测的准确性。在我们的案例中,我们增强可解释性的划分方案实际上损害了我们的预测性能(或准确性)。为了弥补这种性能损失,我们通过利用不同材料属性之间的相关性,来利用转移学习,以提高回归性能。我们的研究表明,所提出的转移学习技术可以克服由于模型简单而导致的性能损失。为了进一步提高ML系统的可解释性,我们在框架中添加了基本原理生成器组件?;驹砩善鞯哪勘晔撬氐模?/span>1)提供与单个预测相对应的解释;2)提供与回归模型相对应的解释。对于单个预测,基本原理生成器提供了有关原型(或相似但已知的化合物)的解释。这有助于材料科学家使用他/她的领域知识来验证类似的已知化合物或原型是否满足所施加的要求或约束。另一方面,对于回归模型,基本原理生成器提供了有关整个材料亚类的全局说明。最后,我们提出了一种新的评估指标和置信度评分,以更好地量化置信度,并建立对ML预测的信任度。通过将其用于两类应用,他们证明了该技术的适用性:1)预测晶体化合物的五种不同的物理特性,以及2)确定潜在稳定的太阳能电池材料。

    In this paper, we take some first steps in addressing the challenge of building reliable and explainable ML solutions for Materials Informatics applications. The main contributions of the paper are twofold. First, we identify some shortcoming with training, testing, and uncertainty quantification steps in existing Materials Informatics pipelines while learning from underrepresented and distributionally skewed (or imbalanced) data. Our finding raises serious concerns regarding the reliability of existing Materials Informatics pipelines. Second, to overcome these challenges, we propose a general-purpose explainable and reliable machine-learning method for enabling reliable learning from underrepresented and distributionally skewed data. We propose the following solutions: (1) learning architecture to bias the training process to achieve the goals of imbalanced domains; (2) sampling approaches to manipulate the training data distribution so as to allow the use of standard ML models; and (3) reliable evaluation metrics and uncertainty quantification methods to better capture the application bias. To improve the explainability, as opposed to other existing approaches, which train an independent regression model per property, we employ a simple and computationally cheap partitioning scheme. This scheme first partitions the data into subclasses of materials based on their property values and train separate simpler regression models for each group. Note that our approach differs in its motivation (and operation) from similar concept utilized by Ward et al. and Zhu et al. Our motivation behind partitioning is to enhance the “explainability”, as opposed to the previous approaches, where a computationally expensive exhaustive search was performed to find artificial groups to enhance the accuracy of predictions. In our case, our explainability enhancing partitioning scheme in fact hurts our predictive performance (or accuracy). To compensate for this performance loss, we utilize transfer learning by exploiting correlation among different material properties to improve the regression performance. We show that the proposed transfer learning technique can overcome the performance loss due to simplicity of the models. To further improve the interpretability of the ML system, we add a rationale generator component to our framework. The goal of the rationale generator is twofold: (1) provide explanations corresponding to an individual prediction and (2) provide explanations corresponding to the regression model. For individual prediction, the rationale generator provides explanations in terms of prototypes (or similar but known compounds). This helps a material scientist to use his/her domain knowledge to verify if similar known compounds or prototypes satisfy the requirements or constraints imposed. On the other hand, for regression models, the rationale generator provides global explanations regarding the whole material sub-classes. This is achieved by providing feature importance for every material sub-class. Finally, they propose a new evaluation metric and a trust score to better quantify confidence and establish trust in the ML predictions. We demonstrate the applicability of our technique by using it for two applications: (1) predicting five physically distinct properties of crystalline compounds, and (2) identifying potentially stable solar cell materials.

    Photorealistic modelling of metals from first principles根据第一原理对金属进行真实感建模
    Gianluca PrandiniGian-Marco Rignanese & Nicola Marzari
    npj Computational Materials 5:129(2019)
    doi:s41524-019-0266-0
    Published online:20 December 2019

    Abstract| Full Text | PDF OPEN

    摘要:自古以来,金属的颜色就吸引着人类的注意力,有色金属(尤其是金化合物)被当作象征权力美学的工具和物品。本研究开发了一个综合框架以获得金属的反射率和颜色,结果显示,可基于标准近似的直接第一原理技术,预测金属光学特性和颜色的趋势。我们将其用于预测几种金属元素和不同类型的金属间化合物(主要考虑基于贵金属的二元合金,包括金属间化合物、固溶体和异质合金)的反射率和颜色。通过与实验数据和已知有色金属的逼真渲染作广泛比较,我们验证了该数值方法。 

    Abstract:The colours of metals have attracted the attention of humanity since ancient times, and coloured metals, in particular gold compounds, have been employed for tools and objects symbolizing the aesthetics of power. In this work, we develop a comprehensive framework to obtain the reflectivity and colour of metals, and show that the trends in optical properties and the colours can be predicted by straightforward first-principles techniques based on standard approximations.We apply this to predict reflectivity and colour of several elemental metals and of different types of metallic compounds (intermetallics, solid solutions and heterogeneous alloys), considering mainly binary alloys based on noble metals.We validate the numerical approach through an extensive comparison with experimental data and the photorealistic rendering of known coloured metals.

    Editorial Summary

    Photorealistic modelling of metals from first principles贵金属七彩模拟筛?。何笥汛蛟旄寺氖资?/span>

    该研究报道了一种理论-近似方法,可预测金属的实际光学性质,并通过对材料的计算筛选来探索组分空间,从而帮助寻找具有特定光学特性的新材料,进而帮助寻找定制的颜色。来自瑞士国家材料计算设计与发现中心(MARVEL)的Nicola Marzari团队,首先建立了一种通用的计算方法,该方法可用于金属颜色的实感模拟,即,通过第一性原理技术估算金属晶体的反射率和颜色。然后,通过对元素金属的系统研究以及与实验数据的广泛比较证明,他们所采用的理论和数值近似方法能够再现反射率曲线的正确趋势,并能展现整个元素周期表中金属元素光学性质的主要差异。最后,作者通过研究不同类型的金属间化合物(有序金属间化合物、无序固溶体和异质合金),而对金属合金也作了类似的研究。特别地,通过与实验结果的比较发现,若使用适当的方法模拟不同类型的化合物,则i)已知有色金属间化合物的模拟颜色具有定性且通常与实验定量一致,ii)可再现贵金属基二元合金的主要颜色规律。该研究有助于激发未来通过第一原理技术实现不同类型材料颜色逼真模拟的研究。如,对元素金属和二元合金模拟方法的系统验证,可看作是实感模拟的必要前期步骤,以便后续对具有更多组成元素(如三元、四元等)、更复杂金属合金的实感模拟,这与该技术的应用(如,超合金和高熵合金)更为相关。

    A theoretical and approximationmethodis reported to predict trends in real metallic systems and to help the search for novel materials with specific optical properties, and therefore also colours, by exploring the composition space through the computational screening of materials. A team led byNicola Marzari from the Theory and Simulation of Materials (THEOS) and National Centre for Computational Design and Discovery of Novel Materials (MARVEL), Switzerland, firstly established a general computational approach that can be used for the photorealistic simulation of metals, showing how the reflectivity and colour of metallic crystals can be estimated by means of first-principles techniques.They then demonstrated through a systematic study on elemental metals and extensive comparisons with experimental data that the theoretical and numerical approximations adopted are able to reproduce the correct behaviour of the reflectivity curves and to capture the main differences in optical properties across the periodic table. Finally, they performed a similar study on metal alloys by considering different types of compounds, that is, ordered intermetallics, disordered solid solutions and heterogeneous alloys. In particular, the authors showed through a comparison with experimental results that, if the appropriate methods are used for the simulation of the different types of compounds, 1) the simulated colours of known coloured intermetallics are in qualitative and most often in quantitative agreement with experiments and that 2) one can reproduce the main colour trends in noble-metal-based binary alloys.Moreover, their work may help stimulate future studies aiming to achieve the photorealistic simulation of different types of materials by means of first-principles techniques.For example, the systematic validation of the approach performed on elemental metals and binary alloys can be seen as a necessary preliminary step for the photorealistic simulation of more complex metallic alloys having a larger number of constituent elements, such as ternaries, quaternaries, and so on, which are more relevant for technological applications (e.g. superalloys and high-entropy alloys).

    Parametrically constrained geometry relaxations for high-throughput materials science (高通量材料科学的参数化约束几何弛豫)
    Maja-Olivia LenzThomas A. R. PurcellDavid HicksStefano CurtaroloMatthias Scheffler & Christian Carbogno
    npj Computational Materials 5:123(2019)
    doi:s41524-019-0254-4
    Published online:17 December 2019

    Abstract| Full Text | PDF OPEN

    摘要:利用对称性来减少参数空间,可大大提高电子结构计算的速度和质量。不幸的是,当晶体全局对称性被破坏时,许多传统的方法都会失败,即使这种扭曲只是轻微的扰动(如类似于Jahn-Teller的扭曲)。本研究提出了一种灵活、可推广的参数弛豫方案,并在全电子密度泛函理论程序FHI-aims中得到实现。这种方法利用参数约束在任何水平上保持对称。在证实了该方法对亚稳态结构的弛豫能力后,我们在13个晶格原型的359种材料的测试集上强调了它的适应性和性能。最后,我们展示了这些约束条件是如何将局部晶格畸变弛豫所需的步骤减少一个数量级的。这些约束条件的灵活性显著加快了面向多种应用的新材料的高通量搜索。 

    Abstract:Reducing parameter spaces via exploiting symmetries has greatly accelerated and increased the quality of electronic-structure calculations. Unfortunately, many of the traditional methods fail when the global crystal symmetry is broken, even when the distortion is only a slight perturbation (e.g., Jahn-Teller like distortions). Here we introduce a flexible and generalizable parametric relaxation scheme and implement it in the all-electron code FHI-aims. This approach utilizes parametric constraints to maintain symmetry at any level. After demonstrating the method’s ability to relax metastable structures, we highlight its adaptability and performance over a test set of 359 materials, across 13 lattice prototypes. Finally we show how these constraints can reduce the number of steps needed to relax local lattice distortions by an order of magnitude. The flexibility of these constraints enables a significant acceleration of high-throughput searches for novel materials for numerous applications.

    Editorial Summary

    Constrained geometry relaxations: parametrical and high-throughput德国马普学会提出:弛豫的参数化约束

      该研究提出了一种在一般对称约简空间中参数化结构弛豫的新方案。 德国马普学会弗里茨-哈伯研究所(Fritz-Haber-Institut)的Thomas A. R. Purcell领导的团队,在解释了该算法之后,在359种不同的材料上进行了测试。在所有情况下,新方法能够严格保持材料的对称性,并平均减少了50%结构收敛所需的步骤。他们还以氧化铋为例演示了如何利用约束条件来弛豫结构到亚稳相。最后,作者用MgO中已知的极化子畸变来展示了该方法的结构弛豫与局部对称性破坏。这一新方法将对计算材料的发现产生深远的影响。结构弛豫成本的降低不仅提高了高通量搜索的速度,而且同时可以探索材料中的亚稳相和动力学稳定的结构相。该方法还有望提高超胞结构的计算效率和基本结构研究的学习。最后,通过监测计算直接得到的力和考虑对称性的力之间的差异,可以从亚稳态或不稳定同质多形体中发现新的稳定相。虽然作者证明了所提出的算法适用于加速和改进标准固态物理计算,但其灵活性是它可应用于更广泛的问题,如过渡态搜索或界面弛豫等。类似地,该方法很容易推广到其他形式的坐标,并且在任何基于电子结构理论的程序包中直接实现,而且作者已经使用ASE将其实现。

    A new scheme for parametrically relaxing structures in a general, symmetry-reduced space is presented. A team led by Thomas A. R. Purcell from the Fritz-Haber-Institut der Max-Planck-Gesellschaft, Germany, firstly explained the algorithm and then tested it on 359 different materials across a broad range of material classes. In all cases, the new method was able to strictly preserve the symmetry of the materials, and on average reduced the number of steps needed to converge a material by 50%. They also demonstrated for the example of bismuth oxide how the constraints can be used to relax to metastable phases. Finally, they showcased the relaxation of structures with local symmetry breaking with known distortion patterns for polarons in MgO.  

    This new method will have a profound impact on computational materials discovery. Not only does the decreased cost of relaxing a material increase the velocity of high-throughput searches but it also allows for those searches to explore metastable and dynamically stabilized structures. The method also has the promise to improve the efficiency of supercell calculations and to study only physically relevant structures. By monitoring the difference between the full forces and symmetrized forces, new stable phases can potentially be discovered from metastable or unstable polymorphs. Although the authors showed that the proposed algorithm is applicable to accelerate and improve standard solid-state physics calculations, its flexibility allows it to be applied to a much wider range of problems, e.g., transition-state searches or interface relaxations. Similarly, it is easily generalizable to other forms of coordinates and straightforwardly implementable in any electronic-structure theory code, and has already been implemented as constraints with ASE by the authors

    .

    Identification of stable adsorption sites and diffusion paths on nanocluster surfaces: an automated scanning algorithm(识别纳米团簇表面上的稳定吸附位点和扩散路径:自动扫描算法
    Tibor SzilvasiBenjamin W. J. Chen & Manos Mavrikakis
    npj Computational Materials 5:101(2019)
    doi:s41524-019-0240-x
    Published online:25 October 2019

    Abstract| Full Text | PDF OPEN

    摘要:离散的三维(3D)纳米团簇表面的不同配位环境对其独特的催化性能起了重要作用。然而,识别这些团簇上的大量吸附位点和扩散路径是冗长和耗时的,尤其对于那些大型的、不对称的纳米团簇。本研究提出了一种简单、自动化的方法来构建近似的二位势能面,以便在最少的人工干预下处理好不同原子/原子团在3D纳米团簇的表面的吸附。这些势能面可以完全表征纳米簇表面上重要的吸附位点和扩散路径,其精确度与目前的计算方法相似,且计算成本相当。我们的方法可以处理复杂的纳米团簇,如合金纳米团簇,并解释团簇弛豫和吸附诱导的重构,为能量的精确计算奠定重要基础。此外,其高度并行化的特性非常适合当代超级计算机架构。我们以Au18Pt55两个团簇为例展示了我们的方法。对于Au18来说,原子氢在最稳定位点之间的扩散是通过非直观的路径发生的,这就强调了探索完整势能面的必要性。我们的方法能够对大型复杂纳米团簇上的吸附和扩散进行快速而客观的评估,这将有助于促进新材料的发现和催化剂的合理设计。 

    Abstract:The diverse coordination environments on the surfaces of discrete, three-dimensional (3D) nanoclusters contribute significantly to their unique catalytic properties. Identifying the numerous adsorption sites and diffusion paths on these clusters is however tedious and time-consuming, especially for large, asymmetric nanoclusters.Here, we present a simple, automated method for constructing approximate 2D potential energy surfaces for the adsorption of atomic species on the surfaces of 3D nanoclusters with minimal human intervention. These potential energy surfaces fully characterize the important adsorption sites and diffusion paths on the nanocluster surfaces with accuracies similar to current approaches and at comparable computational cost.Our method can treat complex nanoclusters, such as alloy nanoclusters, and accounts for cluster relaxation and adsorbate-induced reconstruction, important for obtaining accurate energetics.Moreover, its highly parallelizable nature is ideal for modern supercomputer architectures. We showcase our method using two clusters: Au18 and Pt55.For Au18, diffusion of atomic hydrogen between the most stable sites occurs via non-intuitive paths, underlining the necessity of exploring the complete potential energy surface.By enabling the rapid and unbiased assessment of adsorption and diffusion on large, complex nanoclusters, which are particularly difficult to handle manually, our method will help advance materials discovery and the rational design of catalysts.

    Editorial Summary

    Identification of stable adsorption sites and diffusion paths on nanocluster surfaces: an automated scanning algorithm纳米团簇表面的自动扫描:洞察秋毫、静观其变

      本研究介绍一种简单的方法(称为团簇表面自动扫描法,ACSS),用来生成3D团簇表面的势能面(PES),并通过探测Au18Pt55纳米团簇表面吸附氢原子的势能面,验证了该方法的准确性和高效性。来自美国威斯康星大学麦迪逊分?;в肷锕こ滔档?/span>Manos Mavrikakis教授领导的小组,提出了这一自动化、可高度并行化的方法,来生成3D纳米簇表面的近似2DPES。他们之所以选择用Au18Pt55纳米团簇为例,是因为在先前的实验结果发现,金和铂的小团簇在吸附表面H的各种催化过程中都是有活性的。他们首先通过与标准手动执行(MP)的计算对比,证明了用ACSS方法获得的稳定吸附位点和金属纳米团簇表面上任意两个位点的扩散路径;其次,他们对ACSS引入的误差做了量化,评估了ACSS中假设的有效性;最后,他们比较了ACSS法与MP方法的计算成本。以Au18Pt55表面吸附原子H和扩散的测试作为实例,ACSS法定性和定量地重现了传统MP计算得到的结果,同时降低了计算成本。重要的是,计算成本的收益会随团簇的尺寸增大而急剧增加,而在这种情况下,凭人类直觉来预测吸附位点和扩散路径往往非常不可靠,且容易出错。 

      ACSS方法消除了如探测稳定的吸附位点和扩散势垒等繁琐、重复的任务。此外,它们消除了用户的人为错误,降低了遗漏潜在的重要吸附位点或扩散路径的可能性。由于该方法的通用性,可以预见它将成为实现大型模型纳米团簇复杂表面自动探测的基础。这类纳米团簇不仅在催化领域有技术应用,而且在化学和材料科学领域也有应用,例如分析吸附在复合电池正极材料表面和晶界上的扩散。作者所建立的方法可极大地加速新材料的发现和催化剂的合理设计

      A simple way to generate PESs for 3D cluster surfaces, called Automated Cluster Surface Scanning (ACSS) method, is presented, and its accuracy and efficiency is demonstrated by probing the Potential energy surfaces (PESs) for adsorption of atomic H on Au18 and Pt55 nanoclusters. A group led by Prof. Manos Mavrikakis from the Department of Chemical & Biological Engineering, University of Wisconsin-Madison, USA, presented this automated, highly parallelizable approach—the ACSS method—to generate approximate 2D PESs of 3D nanocluster surfaces. They chose these clusters since previous experimental results showed that small gold and platinum clusters are active in various catalytic processes involving adsorbed surface H. They firstly demonstrated their method via comparisons with standard manually performed (MP) calculations that the ACSS methodology to identify the most important stable adsorption sites and diffusion paths connecting any two of these sites on a given metal nanocluster surface. Then, they quantified the errors introduced by ACSS and assess the validity of the assumptions in ACSS. Lastly, they compared the computational cost of the ACSS method with that of MP calculations. Using the test cases of atomic H adsorption and diffusion on Au18 and Pt55, the ACSS method qualitatively and quantitatively reproduced results obtained by traditional MP calculations at similar or reduced computational cost. Importantly, the realized gains in computational cost increase dramatically with cluster size, a regime exactly where human intuition and detailed accounting of sites and paths would be more likely to fail. 

    The ACSS method eliminates tedious, repetitive tasks, such as the probing of stable adsorption sites and diffusion barriers. Additionally, they remove user biases and reduce the likelihood of missed potentially important adsorption sites or diffusion paths. Due to the generalizability of the method, one can envision that it will serve as a foundation for enabling the automated exploration of complex surfaces of large model nanoclusters. Such nanoclusters have technological applications not only in catalysis, as showcased in this work, but also in chemistry and materials science, for example, for analyzing the diffusion of adsorbates on the surfaces and grain boundaries of complex battery cathode materials. The authors’ method will greatly accelerate the discovery of new materials and the rational design of catalysts.

    Coarse-graining auto-encoders for molecular dynamics (用于分子动力学模拟的粗粒度自动编码器)
    Wujie Wang & Rafael Gómez-Bombarelli
    npj Computational Materials 5:125(2019)
    doi:s41524-019-0261-5
    Published online:18 December 2019

    Abstract| Full Text | PDF OPEN

    摘要:分子动力学模拟为凝聚态材料的微观行为研究提供了理论知识,并作为一种预测工具,可进行新化合物的计算设计。但是,由于材料中热力学和动力学现象的时空尺度很大,原子模拟在计算上通常没有可行性。粗粒度方法可通过减小大型系统的尺寸、复制更多的时间步长、平均化快速运动来模拟大型系统。粗粒度涉及两个耦合学习的问题:定义从全原子表示到简化表示的映射,以及在粗粒度坐标上对哈密顿量进行参数化。我们提出了一个基于变分自动编码器的生成建??蚣?,以统一学习离散的粗粒度变量、解码回原子细节和参数化粗粒度力场的几种任务。该框架已在包括单分子和体相周期性模拟在内的多种模型系统上成功完成了测试。 

    Abstract:Molecular dynamics simulations provide theoretical insight into the microscopic behavior of condensed-phase materials and, as a predictive tool, enable computational design of new compounds. However, because of the large spatial and temporal scales of thermodynamic and kinetic phenomena in materials, atomistic simulations are often computationally infeasible. Coarse-graining methods allow larger systems to be simulated by reducing their dimensionality, propagating longer timesteps, and averaging out fast motions. Coarse-graining involves two coupled learning problems: defining the mapping from an all-atom representation to a reduced representation, and parameterizing a Hamiltonian over coarse-grained coordinates. We propose a generative modeling framework based on variational auto-encoders to unify the tasks of learning discrete coarse-grained variables, decoding back to atomistic detail, and parameterizing coarse-grained force fields. The framework is tested on a number of model systems including single molecules and bulk-phase periodic simulations.

    Editorial Summary

    Coarse-graining auto-encoders for molecular dynamics粗粒度自动编码器:克服分子动力学模拟的缺陷

    该研究提出了一种将粗粒度坐标作为潜在变量进行粗粒度分子动力学采样的方法。美国麻省理工学院材料科学与工程系的科研人员Rafael Gómez-BombarelliWujie Wang,通过力的正则化对潜在空间进行正则化处理,训练了编码映射(一种确定性解码)和粗粒度势函数,可以用来更长时间地模拟大型系统,从而加速分子动力学模拟。在统计学习理论和离散优化方法的推动下,他们提出了基于自动编码器的生成建??蚣?,该框架可1)在三维空间中学习离散的粗粒度变量,并通过几何反映射将其解码回原子细节;2)利用重构损失,从全原子数据中捕获显著的集合特征;3)利用半监督平均瞬时力极小化正则化粗粒度空间,得到平滑的粗粒度自由能??;4)通过变量法找到与作用在全原子训练数据上的瞬时平均力相匹配的高度复杂的粗粒度势函数。此外,他们的工作还允许使用统计学习作为跨越多尺度粗粒度模拟的基础。

    A method to treat the coarse-grained coordinates as latent variables which can be sampled with coarse-grained molecular dynamics is proposed. Two researchers, Rafael Gómez-Bombarelli and Wujie Wang from the Department of Materials Science and Engineering, Massachusets Institute of Technology, USA, by regularizing the latent space with force regularization, trained the encoding mapping, a deterministic decoding, and a coarse-grained potential that can be used to simulate larger systems for longer times and thus accelerate molecular dynamics simulations. Motivated by statistical learning theory and advances in discrete optimization, they proposed an auto-encoder-based generative modeling framework that (1) learns discrete coarse-grained variables in 3D space and decodes back to atomistic detail via geometric back-mapping; (2) uses a reconstruction loss to help capture salient collective features from all-atom data; (3) regularizes the coarse-grained space with a semi-supervised mean instantaneous force minimization to obtain a smooth coarse-grained free-energy landscape; and (4) variationally finds the highly complex coarse-grained potential that matches the instantaneous mean force acting on the all-atom training data. Furthermore, their work also enables the use of statistical learning as a basis to bridge across multi-scale coarse-grained simulations.

    Attribute driven inverse materials design using deep learning Bayesian framework (使用深度学习贝叶斯框架驱动材料逆设计)
    Piyush M. TagadeShashishekar P. AdigaShanthi PandianMin Sik ParkKrishnan S. Hariharan & Subramanya Mayya Kolake
    npj Computational Materials 5:127(2019)
    doi:s41524-019-0263-3
    Published online:20 December 2019

    Abstract| 新时时彩走势 | PDF OPEN

    摘要:很大以部分计算材料学研究都集中于对材料性能的快速、准确、前瞻性的预测,比如,依据分子结构预测其电子性能。之前可用第一性原理计算预测,最近又可用机器学习预测,不过前者的计算量大、不适合高通量筛选。为任意给定的应用搜索合适的材料,遵循逆向路径——依据给出的属性,找到合适的材料。本研究提出了一种深度学习逆向预测框架,即使用“新型条件采样进行性能驱动的材料设计的结构学习(SLAMDUNCS),”以有效而准确地预测具有目标性能的分子。我们将此框架应用于三类有机分子的计算设计:用于薄膜晶体管的有机半导体、用于太阳能电池的小型有机受体、具有高氧化还原稳定性的电解质添加剂。我们的方法具有普适性,足以扩展到无机化合物,代表了基于深度学习的全自动材料探索向前迈出的重要一步。 

    Abstract:Much of computational materials science has focused on fast and accurate forward predictions of materials properties, for example, given a molecular structure predict its electronic properties. This is achieved with first principles calculations and more recently through machine learning approaches, since the former is computation-intensive and not practical for high-throughput screening. Searching for the right material for any given application, though follows an inverse path—the desired properties are given and the task is to find the right materials. Here we present a deep learning inverse prediction framework, Structure Learning for Attribute-driven Materials Design Using Novel Conditional Sampling (SLAMDUNCS), for efficient and accurate prediction of molecules exhibiting target properties. We apply this framework to the computational design of organic molecules for three applications, organic semiconductors for thin-film transistors, small organic acceptors for solar cells and electrolyte additives with high redox stability. Our method is general enough to be extended to inorganic compounds and represents an important step in deep learning based completely automated materials discovery.

    Editorial Summary

    Inverse materials design: deep learning Bayesian framework需要什么性能就做出什么材料?基于深度学习的材料逆设计

    该研究开发了一种基于深度学习的逆预测框架SLAMDUNCS,用于设计具有目标性能的材料。来自印度三星研发院三星技术研究所下一代研究实验室的Piyush M. TagadeShashishekar P. Adiga,提出了这一方法,即使用“新型条件采样进行属性驱动的材料设计的结构学习”(SLAMDUNCS),以有效而准确地预测具有目标性能的分子。特别是,他们将贝叶斯推论应用于三种代表性应用中,设计具有所需性能的有机分子:锂离子电池的高稳定性电解质、有机薄膜晶体管的n型和p型有机半导体、有机太阳能电池中使用的有机小受体分子。作者成功地在数据库外生成了具有目标性能的分子,其中的一小部分已通过第一性原理模拟作过验证。这主要基于他们的贝叶斯框架实现的,该框架可以获得一组分子结构,而不是由现有技术最优化方法预测的单个结构。此外,他们使用受限玻尔兹曼机编码基本化学规则,从而能够生成数据库非固有的分子结构,而无需使用任何试探性或组合性规则,进而能针对目标应用作完全自动化分子设计。尽管他们只是将SLAMDUNCS用于有机分子的设计,但能很容易地用于无机材料的设计,如半导体、压电和热电材料。他们的工作清楚地表明,SLAMDUNCS确实可以快速预测具有目标性能的分子/材料,也可以轻松地将该方法扩展到其他类型的材料和多种性能(包括热、电和光学性能)的预测。而且,SLAMDUNCS可用多元概率分布来简单量化似然函数,以便用于具有多个目标性能的分子逆预测。

    A deep learning based inverse prediction framework, SLAMDUNCS, for design of materials with target properties is developed. A team co-led by Piyush M. Tagade and Shashishekar P. Adiga from the Next Gen Research, Samsung Advanced Institute of Technology, Samsung R & D Institute, India, presented this method, Structure Learning for Attribute-driven Materials Design Using Novel Conditional Sampling (SLAMDUNCS), for efficient and accurate prediction of molecules exhibiting target properties. In particular, they applied Bayesian inference for design of organic molecules with desired properties in three representative applications: high stability electrolytes for lithium-ion batteries, n- and p-type organic semiconductors for organic thin film transistors, and small organic acceptor molecules for use in organic solar cells. They successfully generated molecules not in the database with target properties, a small subset of which were validated with first principles simulations. This is primarily enabled by use of their Bayesian framework that allowed to obtain a set of molecular structure as against single structures predicted by the state of the art optimisation methods. Moreover, their use of restricted Boltzmann machine to encode the fundamental chemical rules allowed to generate the molecular structures extrinsic to the database without using any heuristic or combinatorial rules, enabling fully automated design of molecules for target applications. While they had applied SLAMDUNCS for designing organic molecules, it could be readily adapted to design inorganic materials such as semiconductors, piezoelectric and thermoelectric materials. Their work clearly demonstrated that a fast prediction of molecules/materials with target properties is indeed possible, and it can be easily extended to other types of materials and a diverse set of properties, including thermal, electrical and optical performance. Moreover, SLAMDUNCS can be used for inverse prediction of molecules with multiple target properties simply by using multivariate probability distribution to quantify the likelihood function.

    版权所有 © 中国科学院上海硅酸盐研究所
    地址:上海市长宁区定西路1295号 邮政编码:200050
  • 【回复:四两寻找哀鸣之问】为什么一直辛苦劳作的农民没有富起来?专家解释的4点原因很靠谱 2020-01-18
  • 月薪过万白领辞职回乡养鸡 亏数十万后回城找工作 2019-12-25
  • 85后女青年刘阿娟带领果农发展绿色无公害果品 2019-12-25
  • 习近平应约同俄罗斯总统普京通电话 2019-12-08
  • 安徽九旬教师守候留守儿童:“鞠躬尽瘁,死而不已” 2019-11-27
  • 新华时评:成果不易,更需诚意与信任呵护 2019-11-25
  • 绍兴印染布质量比对成果发布 部分指标国际领先 2019-11-25
  • 当前金融市场流动性合理稳定 2019-10-27
  • 阳泉市春播生产平稳有序 2019-10-27
  • 外卖平台生鲜电商角逐 世界杯引爆外卖夜宵大战 2019-08-23
  • 回复@笑傲江湖V:七千二百多吧。 2019-08-13
  • 四大名著剧组首次同台忆往事 经典影视剧如何铸就? 2019-07-08
  • 老妈见14岁女儿和前闺蜜干架 竟递刀助威:快捅她胸部! 2019-07-08
  • 司法责任制改革:追责保护相统一 给司法人员“定心丸” 2019-06-25
  • 正确区分党徽和党员徽章 2019-06-25
  • 棋牌游戏注册送金币的 黑龙江p62开奖2018047 2017年双色球新的走势规律 宁夏十一选五 舟山星空棋牌游戏大厅 2013真人游戏 下载好友赣南麻将 两肖两码中特 宁夏十一选五开奖图 中彩网福彩3d走势图 腾讯游戏哈尔滨麻将 辽宁快乐12预测五号 大庆冠通棋牌手机版本 诈金花游戏破解版 广东时时彩 南粤福彩26选5开奖结果