Regensburg 2025 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

HL: Fachverband Halbleiterphysik

HL 50: 2D Materials: Stacking and Heterostructures (joint session O/HL)

HL 50.9: Talk

Thursday, March 20, 2025, 17:00–17:15, H6

Machine-Learning the Electronic Structure of Twisted Bilayer Graphene — •Lenz Fiedler¹, Agnieszka Kuc¹, Florian Arnold², and Attila Cangi¹ — ¹Helmholtz-Zentrum Dresden Rossendorf, Dresden, Deutschland — ²Technische Universität Dresden, Dresden, Deutschland

Twistronics, i.e., the study of twodimensional materials in which individual layers are twisted w.r.t.õne another, has the potential to significantly propel technological progress. Twisted bilayer materials, e.g., graphene, may exhibit a significant change in electronic structure and electrical properties based on twist angle. Their computational treatment with density functional theory (DFT) proves difficult, as small twist angles affect the periodicity of the cell and can only be simulated with large unit cells. In this talk, the recently introduced Materials Learning Algorithms (MALA) - a framework for accelerating DFT calculations based on machine learning - is applied to twisted bilayer graphene. Bilayer graphene serves as a proxy for the larger field of twistronics itself. It is shown how the electronic structure, including electronic density of states and electronic charge density, can be predicted from a small number of twist angles for a range of twisted bilayer graphene structures. Since the MALA framework uses the local density of states to encode the electronic structure on a numerical grid, predictions can be made on much larger length scales than with standard DFT calculations. This work demonstrates how machine learning can be used to computationally model twisted bilayer structures where standard first-principles methods are not viable.

Keywords: Density Functional Theory; Machine Learning; Neural Networks