Fig. 2: Example studies on monolayer graphene and carbon nanotube (CNT).
From: A deep equivariant neural network approach for efficient hybrid density functional calculations
a Band structures of supercells of monolayer graphene computed by hybrid density functional theory (DFT-hybrid) and DeepH-hybrid. Representative test structures with the smallest, median and largest mean absolute error (MAE) in the prediction of hybrid-functional Hamiltonian are displayed. b Schematic workflow of the DeepH-hybrid method, which learns from training datasets of nearly flat structures and then generalizes to study curved nanotube structures. c Band structures and d real and imaginary parts of electric susceptibility χzz as a function of frequency ω of (49, 0) CNT computed by DFT-hybrid and DeepH-hybrid. The periodic direction of the nanotube is defined as the z-axis. Source data are provided as a Source Data file.
