Photoinduced reactive dynamics of complex systems

光诱导反应是许多化学和生化系统中的重要过程。要详细了解这些过程，通常需要对涉及许多非绝热跃迁的原子核进行严格的量子动力学模拟。本项目的目的是发展和应用数值方法，以促进大系统光致过程中原子核的详细量子动力学模拟。光致反应过程量子动力学模拟的核心步骤是构建描述电子态之间耦合的高维从头算势能面（PESs）。本项目主要采用直接动力学方法和基于神经网络的拟合，研究耦合PESs的过程驱动构建。结合非线性降维方法，采用从头算多重衍生法（AIMS）寻找和处理系统非绝热动力学中的相关区域。基于从头算数据，将使用基于现代神经网络的拟合和其他机器学习方法构建振动耦合PESs。多构型时变Hartree （MCTDH）计算将用于研究耦合PESs上的波包动力学，并计算基于轨迹的方法无法精确研究的复杂谱。有趣的应用包括π -> π *激发下乙烯的光诱导动力学和吡啶的开环反应。

Photoinduced reactions are important processes in many chemical and biochemical systems. The detailed understanding of these processes often requires a rigorous quantum dynamical simulation of the nuclei involving numerous nonadiabatic transitions. The aim of this project is the development and application of numerical methods to facilitate a detailed quantum dynamical simulation of the nuclei in photoinduced processes of large systems. A central step for the quantum dynamical simulation of photoinduced reactive processes is the construction of high-dimensional ab initio potential energy surfaces (PESs) describing the couplings between electronic states. The present project focuses on the process driven construction of coupled PESs employing direct dynamics methods and neural network based fitting. The ab initio multiple spawning (AIMS) approach combined with a nonlinear dimensionality reduction method will be used to find and process relevant regions in the nonadiabatic dynamics of the system. Based on the ab initio data, vibronically coupled PESs will be constructed using modern neural network based fitting and other machine learning approaches. Multi-configurational time-dependent Hartree (MCTDH) calculations will be used to study the wavepacket dynamics on the resulting coupled PESs and to compute complex spectra which can not be accurately studied using trajectory based methods.Interesting applications include photoinduced dynamics of ethylene upon pi -> pi * excitation and the ring-opening reaction of pyridine.