Nonadiabatic Derivative Couplings Calculated Using Information of Potential Energy Surfaces without Wavefunctions: Ab Initio and Machine Learning Implementations

Wen-Kai Chen,Sheng-Rui Wang,Xiang-Yang Liu,Wei-Hai Fang,Ganglong Cui

DOI: https://doi.org/10.3390/molecules28104222

IF: 4.6

2023-05-22

Molecules

Abstract:In this work, we implemented an approximate algorithm for calculating nonadiabatic coupling matrix elements (NACMEs) of a polyatomic system with ab initio methods and machine learning (ML) models. Utilizing this algorithm, one can calculate NACMEs using only the information of potential energy surfaces (PESs), i.e., energies, and gradients as well as Hessian matrix elements. We used a realistic system, namely CH2NH, to compare NACMEs calculated by this approximate PES-based algorithm and the accurate wavefunction-based algorithm. Our results show that this approximate PES-based algorithm can give very accurate results comparable to the wavefunction-based algorithm except at energetically degenerate points, i.e., conical intersections. We also tested a machine learning (ML)-trained model with this approximate PES-based algorithm, which also supplied similarly accurate NACMEs but more efficiently. The advantage of this PES-based algorithm is its significant potential to combine with electronic structure methods that do not implement wavefunction-based algorithms, low-scaling energy-based fragment methods, etc., and in particular efficient ML models, to compute NACMEs. The present work could encourage further research on nonadiabatic processes of large systems simulated by ab initio nonadiabatic dynamics simulation methods in which NACMEs are always required.

chemistry, multidisciplinary,biochemistry & molecular biology

What problem does this paper attempt to address?

This paper aims to solve several key problems in the calculation of non - adiabatic coupling matrix elements (NACMEs): 1. **High computational cost**: Calculating NACMEs using ab - initio methods is usually very computationally expensive, especially for polyatomic systems. Moreover, not all electronic structure programs or methods have implemented wave - function - based algorithms to calculate NACMEs. 2. **Difficulty in obtaining wave - function information**: Although some electronic structure methods or fragmentation methods can provide information on potential energy surfaces (PESs), they are difficult or unable to provide wave - function information, which limits their application in calculating NACMEs. To solve these problems, the author proposes an approximate algorithm for calculating NACMEs based on PESs information. This algorithm only requires information on energy, gradients, and Hessian matrix elements. Through this method, NACMEs can be efficiently calculated without wave - function information, thus broadening its application range in different electronic structure methods. Specifically, the author uses the CH₂NH molecule as an actual system to compare the NACMEs calculated by the PESs - based approximate algorithm and the wave - function - based exact algorithm. The research results show that, except at the strictly degenerate energy levels (i.e., conical intersection points), the PESs - based approximate algorithm can give results as accurate as those of the wave - function - based algorithm. In addition, the author also tests the method of calculating NACMEs by combining PESs information with a machine - learning (ML) model. The results show that this method is not only equally accurate but also more computationally efficient. This provides new possibilities for future non - adiabatic dynamics simulations in large systems.

Nonadiabatic Derivative Couplings Calculated Using Information of Potential Energy Surfaces without Wavefunctions: Ab Initio and Machine Learning Implementations

An Undecomposed Hybrid Algorithm for Nonlinear Coupled Constitutive Relations of Rarefied Gas Dynamics

Learning the Quantum Centroid Force Correction in Molecular Systems: A Localized Approach

Ab-Initio Molecular Dynamics Acceleration Scheme with an Adaptive Machine Learning Framework

Accelerated Quantum Mechanics/Molecular Mechanics Simulations Via Neural Networks Incorporated with Mechanical Embedding Scheme.

Accurate neural-network-based fitting of full-dimensional N2 –Ar and N2–CH4 two-body potential energy surfaces aimed at spectral simulations

Incorporating Electronic Information into Machine Learning Potential Energy Surfaces via Approaching the Ground-State Electronic Energy as a Function of Atom-Based Electronic Populations

Exciting DeePMD: Learning excited state energies, forces, and non-adiabatic couplings

Machine Learning Nonadiabatic Dynamics: Eliminating Phase Freedom of Nonadiabatic Couplings with the State-Intraction State-Averaged Spin-Restricted Ensemble-Referenced Kohn-Sham Approach

Non-adiabatic coupling matrix elements in a magnetic field: geometric gauge dependence and Berry phase

First-Principles Nonadiabatic Dynamics of Molecules at Metal Surfaces with Vibrationally Coupled Electron Transfer

Accurate interatomic force field for molecular dynamics simulation by hybridizing classical and machine learning potentials

DeePKS+ABACUS as a Bridge between Expensive Quantum Mechanical Models and Machine Learning Potentials

Deep Learning for Nonadiabatic Excited-State Dynamics.

Analytic evaluation of non-adiabatic couplings within the complex absorbing potential equation-of-motion coupled-cluster method

Combining Ab Initio and Machine Learning Method to Improve the Prediction of Diatomic Vibrational Energies

Nonadiabatic Excited-State Molecular Dynamics Methodologies: Comparison and Convergence

DeePKS + ABACUS as a Bridge between Expensive Quantum Mechanical Models and Machine Learning Potentials

On the representation of coupled adiabatic potential energy surfaces using quasi-diabatic Hamiltonians: a distributed origins expansion approach