Machine-Learning Electron Dynamics with Moment Propagation Theory: Application to Optical Absorption Spectrum...

Area of Science:

• Quantum Chemistry
• Computational Physics
• Materials Science

Background:

• Simulating quantum electron dynamics is computationally intensive.
• Existing methods often struggle with complex molecular and condensed matter systems.
• Machine learning offers potential for accelerating these simulations.

Purpose of the Study:

• To apply Moment Propagation Theory (MPT) with machine learning for efficient electron quantum dynamics simulations.
• To train the MPT equation of motion using real-time time-dependent density functional theory (RT-TDDFT) data.
• To demonstrate the method's efficacy in calculating optical absorption spectra.

Main Methods:

• Utilized a novel theoretical formulation: Moment Propagation Theory (MPT).
• Employed real-time time-dependent density functional theory (RT-TDDFT) for first-principles data generation.
• Trained second-order time derivatives of moments using machine learning with maximally localized Wannier functions (MLWFs).

Main Results:

• Achieved efficient electron dynamics simulations via machine-learned MPT.
• Successfully computed optical absorption spectra for diverse systems.
• Demonstrated applicability to isolated molecules (water, benzene, ethene) and condensed matter (liquid water, silicon).

Conclusions:

• The ML-trained MPT provides an efficient and accurate method for quantum electron dynamics.
• The approach is versatile, applicable to both molecular and condensed matter systems.
• Explored the utility of electron nearsightedness for further optimization.