Machine-Learning Electron Dynamics with Moment Propagation Theory: Application to Optical Absorption Spectrum Computation using Real-Time TDDFT (2024)

Undergraduate: Nicholas Boyer

Faculty Advisor: Yosuke Kanai
Department: Chemistry

---

We present the application of our new method, moment propagation theory (MPT), to develop machine-learning techniques to simulate the quantum dynamics of electrons using first-principles theory. In particular, we use real-time time-dependent density functional theory (RT-TDDFT) in the gauge of so-called maximally localized Wannier functions (MLWFs) for computing the optical absorption spectrum. MPT provides both a concise representation of spatially-localized MLWFs using increasing orders of moments, the equation of motion for these moments can be integrated in time. Using this MPT framework, machine-learning techniques can be used to perform accurate and fast computations of the second order time derivatives of the moments needed to reproduce the electron dynamics. The application to computing optical absorption spectrum for various systems is demonstrated here as a proof-of-principles example of this approach. In addition to isolated molecules (water, benzene, and ethene), we study liquid water and explore how the principle of the nearsightedness of electrons can be employed in this context.