Artificial intelligence for artificial materials: moiré atoms

Semiconductor moiré materials have emerged as a pivotal experimental platform for realizing novel strongly-correlated states of matter, providing new opportunities for designing artificial quantum materials. Despite its importance, understanding the many strongly-correlated electron physics in these systems is challenging due to the limited accuracy and scalability of the conventional methods. In this work, we propose a new approach of integrating multi-scale modeling with artificial intelligence to investigate the artificial quantum materials. We study the regime that the system can be modeled as an array of ``moiré atoms”, each of which consists of several strongly-correlated electrons confined to a moiré superlattice potential minimum. We develop a scalable method with a state-of-the-art 2D fermionic neural network to accurately simulate the physics. We find that strong Coulomb interactions in combination with the anisotropic moiré potential lead to striking ``Wigner molecule", which charge density distributions can be observed with scanning tunneling microscopy.