使用深度学习的原子间潜力探索碳的相变和结构.
Kai Chen1, Riyi Yang1, Zhefeng Wang1
1Shanghai Ultra-Precision Optical Manufacturing Engineering Center, Department of Optical Science and Engineering, Fudan University, Shanghai 200433, China. songyouwang@fudan.edu.cn.
碳的新机器学习潜力使得大规模系统的高效,准确的模拟成为可能. 这一突破有助于研究碳相变,并在极端条件下发现新的碳结构.
科学领域:
- 材料科学 材料科学 材料科学
- 计算化学计算化学
- 凝聚物质物理学 凝聚物质物理学
背景情况:
- 在大规模系统中准确模拟相变对于理解材料行为至关重要.
- 像*ab initio*分子动力学 (AIMD) 这样的传统方法是准确的,但在计算上昂贵.
- 经验潜能提供速度,但缺乏相位过渡研究所需的精度.
研究的目的:
- 开发一种结合效率和精度的计算方法,用于研究大型碳系统中的相位过渡.
- 从C60和石墨烯前体中研究无形和多晶钻石的形成机制.
- 用先进的计算工具发现新的碳结构.
主要方法:
- 使用深度神经网络开发碳的机器学习潜力 (MLP).
- 在高压,高温 (HPHT) 条件下的大型系统的模拟.
- 使用结构搜索软件 (AIRSS) 来生成初始结构,然后通过MLP进行优化.
主要成果:
- 开发的MLP表现出强大的可扩展性,并使碳相过渡的有效研究成为可能.
- 成功阐明了无形和多晶钻石的形成机制.
- 确定了新的碳结构集群,MLP预测与高斯近似潜力 (GAP) 保持一致,但具有更高的计算效率.
结论:
- 新型的MLP为碳材料研究提供了强大而有效的工具,特别是对于相位过渡研究.
- 在极端条件下了解碳材料的行为方面取得了重大进展.
- 开辟了新的途径来探索碳异构体及其特性,并提高了计算效率.
更多相关视频
11:29Novel 3D/VR Interactive Environment for MD Simulations, Visualization and Analysis
Published on: December 18, 2014
09:17Structure-Based Simulation and Sampling of Transcription Factor Protein Movements along DNA from Atomic-Scale Stepping to Coarse-Grained Diffusion
Published on: March 1, 2022
相关概念视频
Electronic Structure of Atoms
An atom comprises protons and neutrons, which are contained inside the dense, central core called the nucleus, with electrons present around the nucleus. Taking into account the wave–particle duality of electrons and the uncertainty in position around the nucleus, quantum mechanics provides a more accurate model for the atomic structure. It describes atomic orbitals as the regions around the nucleus where electrons of discrete energy exist, characterized by four quantum...
Predicting Molecular Geometry
π Electron Effects on Chemical Shift: Overview
¹³C NMR: ¹H–¹³C Decoupling
A broadband decoupling technique is used to simplify these complex, sometimes overlapping, signals. Broadband decoupling relies on a...
Double Resonance Techniques: Overview
Spin decoupling is usually achieved by...
Covalent Bonding and Lewis Structures
