Machine-learning Kohn-Sham potential from dynamics in time-dependent Kohn-Sham systems

Abstract The construction of a better exchange-correlation potential in time-dependent density functional theory (TDDFT) can improve the accuracy TDDFT calculations and provide more accurate predictions properties many-electron systems. Here, we propose machine learning method to develop energy Kohn-Sham system is proposed. based on dynamics does not require any data exact for training model. We demonstrate results our with 1D harmonic oscillator example two-electron example. show that machine-learned matches absence memory effect. Our still capture presence effects. developed this article provides insight into making approximations system.