|
|
快捷服务 |
|
|
|
|
|
|
相关链接 |
|
|
|
|
|
|
会议信息 |
|
|
|
|
|
|
友情链接 |
|
|
|
|
|
|
近期文章 |
|
|
Predicting electronic structures at any length scale with machine learning |
发布时间:2023-09-19 |
Predicting electronic structures at any length scale with machine learning
Lenz Fiedler, Normand A. Modine, Steve Schmerler, Dayton J. Vogel, Gabriel A. Popoola, Aidan P. Thompson, Sivasankaran Rajamanickam & Attila Cangi
npj Computational Materials 9: 115(2023)
doi.org/10.1038/s41524-023-01070-z
Abstract: The properties of electrons in matter are of fundamental importance. They give rise to virtually all material properties and determine the physics at play in objects ranging from semiconductor devices to the interior of giant gas planets. Modeling and simulation of such diverse applications rely primarily on density functional theory (DFT), which has become the principal method for predicting the electronic structure of matter. While DFT calculations have proven to be very useful, their computational scaling limits them to small systems. We have developed a machine learning framework for predicting the electronic structure on any length scale. It shows up to three orders of magnitude speedup on systems where DFT is tractable and, more importantly, enables predictions on scales where DFT calculations are infeasible. Our work demonstrates how machine learning circumvents a long-standing computational bottleneck and advances materials science to frontiers intractable with any current solutions.
摘要: 物质中电子的性质非常重要。它们产生了几乎所有的物质性能,并决定了从半导体器件到巨大气体行星内部的物体的物理原理。这些不同应用的建模和模拟主要依赖于密度泛函理论(DFT),这已经成为预测物质电子结构的主要方法。虽然DFT计算已经被证明非常有用,但它们的计算尺度限制在小体系上。我们开发了一个机器学习框架,用于预测任何长度尺度上的电子结构。它在DFT适用的体系上显示了三个数量级的加速,更重要的是,它可以在DFT计算不可行的尺度上进行预测。我们的工作展示了机器学习如何绕过一个长期存在的计算瓶颈,并将材料科学推进到任何当前解决方案都难以解决的前沿。
Editorial Summary
Machine Learning: The Versatile Decoder of Electronic Structures
The electronic structure governs the reactivity of molecules and the properties of materials, making the accurate prediction and understanding of electronic structure critically important. Density Functional Theory (DFT) has emerged as the primary method for predicting material electronic structures; however, it has limitations when dealing with large-scale systems. In recent years, machine learning (ML) methods have been applied to learn electronic structures, but maintaining the precision of DFT for larger systems remains challenging. To overcome the limitation, This study has proposed a machine learning framework capable of efficiently and accurately predicting electronic structures across various scales. A team led by Prof. Attila Cangi from the Center for Advanced Systems Understanding and the Helmholtz-Zentrum Dresden-Rossendorf, Germany, has developed a machine learning approach with computational costs that scale linearly with system size, enabling efficient and accurate predictions of electronic structures at any scale. Unlike existing ML methods, this approach provides direct access to electronic structures and is not constrained by specific observable parameters. The study demonstrates that this method, applicable to systems where DFT is feasible, significantly accelerates computation speeds by up to three orders of magnitude, all while maintaining high computational accuracy. Most importantly, it enables electronic structure predictions at scales where traditional DFT calculations are unfeasible. This research successfully overcomes the computational bottleneck of DFT calculations for large-scale systems, offering crucial support for accelerating the discovery and optimization of new materials.
机器学习:电子结构的全能解码者
电子结构决定了分子的反应性以及材料的性能,因此准确预测和理解电子结构至关重要。密度泛函理论(DFT)已经成为预测物质电子结构的主要方法,但它在处理大尺寸体系时存在限制。近年来,机器学习(ML)方法已被用于学习电子结构,但对于较大尺寸的物质,保持DFT的精度仍然充满挑战。为了克服该限制,本研究提出了一种机器学习框架,可以高效且准确预测各种尺度上的电子结构。为了克服该限制,来自德国高级系统理解中心和亥姆霍兹德累斯顿罗森多夫研究中心的Attila Cangi研究团队,开发了一种机器学习方法,其计算成本与体系尺寸呈线性比例,可以高效且准确预测任何尺度的电子结构。与现有ML方法不同,该方法提供了对电子结构的直接访问,并不受特定可观测量的限制。研究表明,在DFT适用的体系上,该方法在几乎不损失计算精度的情况下,计算速度加快了三个数量级。更重要的是,它可以在DFT计算难以应对的尺度上进行电子结构的预测。这项研究成功克服了DFT计算在处理大尺寸体系时的计算瓶颈,为加速新材料的发现和性能优化提供了重要支持。
|
|
【打印本页】【关闭本页】 |
|
|
|