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

Abstract

We extend the DeePMD neural network architecture to predict electronic structure properties necessary to perform non-adiabatic dynamics simulations. While learning the excited state energies and forces follows a straightforward extension of the DeePMD approach for ground-state energies and forces, how to learn the map between the non-adiabatic coupling vectors (NACV) and the local chemical environment descriptors of DeePMD is less trivial. Most implementations of machine-learning-based non-adiabatic dynamics inherently approximate the NACVs, with an underlying assumption that the energy-difference-scaled NACVs are conservative fields. We overcome this approximation, implementing the method recently introduced by Richardson [J. Chem. Phys. 158 011102 (2023)], which learns the symmetric dyad of the energy-difference-scaled NACV. The efficiency and accuracy of our neural network architecture is demonstrated through the example of the methaniminium cation CH${}_2$NH${}_2^{+}$.

Presentation of the 3 challenges in learning NACVs was clarified, with references added. A supplementary material section was added to clarify some points made in main text about methods' properties, together with more tests and comparisons of the different methods. Main results of these comparisons are summarized in main text with new figures. Ancillary files with code and data were added

References in corpus (9)

• DeePMD-kit v2: A software package for Deep Potential models
• Machine learning for electronically excited states of molecules
• Combining SchNet and SHARC: The SchNarc machine learning approach for excited-state dynamics
• Raman Spectrum and Polarizability of Liquid Water from Deep Neural Networks
• Excited state, non-adiabatic dynamics of large photoswitchable molecules using a chemically transferable machine learning potential
• Physically inspired deep learning of molecular excitations and photoemission spectra
• Machine learning and excited-state molecular dynamics
• A Tungsten Deep Neural-Network Potential for Simulating Mechanical Property Degradation Under Fusion Service Environment
• MLatom software ecosystem for surface hopping dynamics in Python with quantum mechanical and machine learning methods

Cited by in corpus (4)

• Roadmap for Molecular Benchmarks in Nonadiabatic Dynamics
• A Descriptor Is All You Need: Accurate Machine Learning of Nonadiabatic Coupling Vectors
• Machine-Learned Potentials for Solvation Modeling
• Machine learning potential as a guide for eutectic in ultra-refractory multicomponent ceramics