机器学习的原子间潜力的高效和准确的空间混合,用于材料科学.
Fraser Birks1, Matthew Nutter1,2, Thomas D Swinburne3
1Warwick Centre for Predictive Modelling, School of Engineering, University of Warwick, Coventry, UK.
概括
我们开发了ML-MIX,这是一种计算工具,通过混合机器学习的原子间潜力 (MLIPs) 来加速分子动力学模拟. 这种方法在不损害准确性的情况下实现了显著的加速度,使得大规模复杂系统的研究成为可能.
科学领域:
- 计算材料科学科学 计算材料科学
- 原子学模拟 原子学模拟
- 机器学习在物理学中的应用
背景情况:
- 机器学习的原子间潜能 (MLIP) 提供了高精度,但在计算上是密集的.
- 由于计算成本,现有的方法难以进行大规模模拟.
- 量子力学/分子力学 (QM/MM) 启发的方法可以弥合精度和效率的差距.
研究的目的:
- 推出ML-MIX,这是一个用于加速分子动力学模拟的新软件包.
- 为了使复杂的MLIP能够在实际计算预算内在更大的系统上使用.
- 为了证明空间混合原子间潜力的精度和效率的提高.
主要方法:
- 开发了ML-MIX,这是一个与CPU和GPU兼容的LAMMPS包.
- 实施了一个空间混合策略,用于不同复杂度的原子间潜能.
- 展示了潜在的蒸技术,从"昂贵"的"便宜"的潜力中产生"便宜"的潜力.
- 在Si,Fe和W-He系统中对点缺陷的验证性能.
主要成果:
- 在不损失精度的情况下,在 8000 个原子周围的系统中实现了高达 11 × 的加速度.
- 成功地应用ML-MIX来研究W中使用ACE潜力的螺杆位移性.
- 使用MACE/SNAP混合精确地复制了W的实验He反射系数,匹配了高达80 eV的观测结果.
- 展示了在大型,现实的系统上部署最先进的MLIP的能力.
结论:
- 通过高效组合MLIPs,ML-MIX显著加快了原子模拟.
- 该方法允许将先进的MLIP应用于大规模,科学相关的问题.
- ML-MIX弥合了材料模拟中的计算成本和准确性之间的差距,使得新的发现成为可能.
相关概念视频
Hybridization of Atomic Orbitals I
69.1K
The mathematical expression known as the wave function, ψ, contains information about each orbital and the wavelike properties of electrons in an isolated atom. When atoms are bound together in a molecule, the wave functions combine to produce new mathematical descriptions that have different shapes. This process of combining the wave functions for atomic orbitals is called hybridization and is mathematically accomplished by the linear combination of atomic orbitals. The new orbitals that...
69.1K
Intermolecular vs Intramolecular Forces
101.6K
Intermolecular forces (IMF) are electrostatic attractions arising from charge-charge interactions between molecules. The strength of the intermolecular force is influenced by the distance of separation between molecules. The forces significantly affect the interactions in solids and liquids, where the molecules are close together. In gases, IMFs become important only under high-pressure conditions (due to the proximity of gas molecules). Intermolecular forces dictate the physical properties of...
101.6K
The Thermodynamics of Mixing
59
Mixing is a fascinating phenomenon in thermodynamics, particularly when considering the Gibbs energy of a mixture at constant temperature and pressure. This energy, denoted as G, tends to decrease during spontaneous mixing processes, offering insights into the composition changes that occur.Imagine two ideal gases, initially separated in different containers, with amounts nA and nB, respectively, both at a temperature T and pressure p. The chemical potentials of these gases have their 'pure'...
59
Fermi Level Dynamics
923
The vacuum level denotes the energy threshold required for an electron to escape from a material surface. It is usually positioned above the conduction band of a semiconductor and acts as a benchmark for comparing electron energies within various materials.
Electron affinity in semiconductors refers to the energy gap between the minimum of its conduction band and the vacuum level and it is a critical parameter in determining how easily a semiconductor can accept additional electrons.
The work...
Electron affinity in semiconductors refers to the energy gap between the minimum of its conduction band and the vacuum level and it is a critical parameter in determining how easily a semiconductor can accept additional electrons.
The work...
923
Molecular Orbital Theory II
28.2K
Molecular Orbital Energy Diagrams
28.2K
Predicting Molecular Geometry
46.8K
VSEPR Theory for Determination of Electron Pair Geometries
46.8K


