Simulating the ghost: quantum dynamics of the solvated electron.

Area of Science:

• Physical Chemistry
• Computational Chemistry
• Materials Science

Background:

• The hydrated electron, a highly reactive species, presents significant challenges for experimental and theoretical studies due to its transient nature and the need for advanced electronic structure theories.
• Conventional empirical force fields struggle to model the solvated electron because it lacks a classical atomistic structure.

Purpose of the Study:

• To develop a novel machine-learning model capable of accurately describing the hydrated electron's effect on surrounding water structure without explicitly including the electron.
• To enable accurate quantum statistical and dynamical simulations of the solvated electron's properties.

Main Methods:

• A flexible machine-learning model was trained on high-level correlated wave function data.
• The model implicitly accounts for the electron's influence on the water environment.
• Quantum statistical and dynamical simulations were performed using the trained model.

Main Results:

• The machine-learning potential accurately reproduces the stable cavity structure of the hydrated electron.
• It correctly captures the localization dynamics following electron injection in water.
• The model achieves accuracy comparable to state-of-the-art correlated wave function methods at a lower computational cost.

Conclusions:

• The developed machine-learning model offers a computationally efficient and accurate approach for studying the solvated electron.
• It facilitates precise determination of the solvated electron's structure, diffusion mechanisms, and vibrational spectroscopy.
• This work overcomes limitations of traditional methods for modeling highly reactive species in solution.