Accelerating Electron Dynamics Simulations through Machine Learned Time Propagators
This work addresses the problem of slow electron dynamics simulations for researchers in computational chemistry and materials science, offering a potential incremental improvement for real-time modeling of laser-irradiated systems.
The paper tackles the computational cost of electron dynamics simulations in time-dependent density functional theory by introducing autoregressive neural operators as time-propagators, achieving superior accuracy and speed compared to traditional solvers, as demonstrated on one-dimensional diatomic molecules.
Time-dependent density functional theory (TDDFT) is a widely used method to investigate electron dynamics under various external perturbations such as laser fields. In this work, we present a novel approach to accelerate real time TDDFT based electron dynamics simulations using autoregressive neural operators as time-propagators for the electron density. By leveraging physics-informed constraints and high-resolution training data, our model achieves superior accuracy and computational speed compared to traditional numerical solvers. We demonstrate the effectiveness of our model on a class of one-dimensional diatomic molecules. This method has potential in enabling real-time, on-the-fly modeling of laser-irradiated molecules and materials with varying experimental parameters.