Full-scale ab initio simulations of laser-driven atomistic dynamics

Qiyu Zeng, Bo Chen, Shen Zhang, Dongdong Kang, Han Wang, Xiaoxiang Yu & Jiayu Dai

npj Computational Materials 9: 213 (2023)

https://doi.org/10.1038/s41524-023-01168-4

Published online: 25 November 2023

编辑概述

双温模型耦合的神经网络分子动力学：通往大尺度、第一性原理精度的原子尺度激光过程模拟

超快激光与物质相互作用，会产生由热电子与冷晶格构成的极端电子-离子非平衡状态，相应的微观动力学演化将受到激发态电子势能面的调制，同时伴随着电子-离子非绝热能量交换以及激光辐照样品的几何特征等多因素复杂耦合的影响，实验难以区分不同过程的贡献，传统第一性原理和分子动力学方法也面临着效率精度难以两全的挑战。

来自国防科技大学理学院的戴佳钰教授团队，提出了一种高效、准确的激发态分子动力学模拟框架——TTM-DPMD，构建显式依赖电子分布的神经网络激发态势能面，并与双温模型分子动力学深度耦合，可以内禀描述电子激发导致的非热效应和非绝热能量交换等，并将模拟规模拓展到百万原子量级。机器学习框架如图1所示。

FIG 1 Schematic diagram of workflow for efficient and accurate simulation of laser-driven atomistic dynamics.

    文章中以30 nm厚度的多晶钨薄膜为对象，实现了样品在厚度维度的全尺度模拟（包含75万原子），观察到：

1）即使在低激光能量密度下（激光能量密度为0.08 MJ/kg，对应电子温度约为5000 K），激光产生的热电子也足以诱导声子软化，显著改变电子衍射峰下降的动力学过程，从而获得与实验定量一致的结果，如图2所示。

  IG 2 capturing nonthermal effect with TTM-DPMD approach 2

FIG 2 Hot electron modifies the thermodynamic pathway and introduce significant inhomogeneity in thermodynamic profile and structural evolution

    在高激光能量密度下（激光能量密度为0.80 MJ/kg，对应电子温度约为11000 K），热电子会额外贡献高达10 GPa的非热应力，在薄膜表面驱动剧烈的单轴膨胀过程，使得系统的热力学、微观结构演化表现出强烈的非均匀性，与传统热过程截然不同，必须依赖于大尺度、激发态的双重描述，如图3所示。该研究为理解激光与物质相互作用过程中原子尺度动力学提供了可靠的研究手段，并指出在大尺度下，激发电子驱动的非热行为会产生截然不同的热力学路径和结构演化。

Editorial Summary

Neural network: Laser-excited potential energy surface

Ultrafast laser excitation can drive matter into extremely nnequilibrium states, in which the hot electron and cold lattice coexist. The subsequent atomistic dynamics is therefore a long-standing challenge because it is governed by the interplay between excited-electron-modulated potential energy surface, electron-ion coupling, and geometric characteristics of target sample.

Recently, a team led by Prof. Jiayu Dai from the College of Science, National University of Defense Technology, China, developed an efficient and accurate framework for simulating laser-excited atomistic dynamics. A constructed electron-temperature-dependent deep neural-network potential energy surface is deeply coupled with a hybrid atomistic-continuum approach, to inherently describe the laser-induced nonthermal behavior and non-adiabatic energy exchange, while extending the simulation size to millions of atoms.

The full-scale simulation of 30-nm-thick polycrystalline tungsten film (containing 752650 atoms) upon laser excitation is performed:

1. Even under moderate laser excitation (laser energy density of 0.08 MJ/kg to create initial electron temperature around 5000 K), hot electron can induce phonon softening to significantly modify the decay dynamics of Laue diffraction peak intensity, which enables consistent interpretation of experimental measurements.
2. Under severe laser excitation (energy density of 0.80 MJ/kg to increase electron temperature up to 10000 K), the hot electrons can contribute a non-negligible nonthermal stress of more than 10 GPa, leading to uniaxial expansion process near the free surface of the sample. Different with purely thermal process, the hot electron introduce significant inhomogeneity in thermodynamic state and microscopic structure is introduced, which requires both large scale simulation and laser-excited state description.

This work provides a path to understand the laser-driven atomistic dynamics, and demonstrate the vital role of hot electrons in determining the thermodynamic pathway and structural transformation dynamics.