Machine learning the microscopic form of nematic order in twisted double-bilayer graphene

This work demonstrates the use of convolutional neural networks (CNNs) to learn effective theoretical models from scanning tunneling microscopy (STM) data on correlated moiré superlattices, specifically focusing on electronic nematic order in twisted double-bilayer graphene (TDBG). The large moiré unit cells and tunability of these systems allow for the collection of high-dimensional datasets. The CNNs learn the microscopic nematic order parameter by incorporating correlations between the local density of states (LDOS) at different energies, distinguishing it from heterostrain. This methodology proves powerful for investigating microscopic details of correlated phenomena in moiré systems.

João Augusto Sobral, Stefan Obernauer, Simon Turkel, Abhay N. Pasupathy, Mathias S. Scheurer