Dresden 2026 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
TT: Fachverband Tiefe Temperaturen
TT 48: 2D Materials: Electronic structure, excitations, etc. II (joint session O/HL/TT)
TT 48.8: Vortrag
Mittwoch, 11. März 2026, 12:15–12:30, TRE/MATH
Long-Range Interactions in Twisted Bilayer Materials with Machine Learning for the Electronic Density — •Zekun Lou1, Alan Lewis2, and Mariana Rossi1 — 1MPI for the Structure and Dynamics of Matter, Hamburg, Germany — 2Department of Chemistry, University of York, York, U.K.
Moiré superlattices in twisted bilayer (TB) 2D materials exhibit extraordinary quantum phenomena, but first-principles understanding remains limited by computational costs. While most machine learning (ML) methods for density functional theory (DFT) acceleration are based on the locality assumption, we demonstrate that accurate moiré electronic structure prediction requires long-range encoding due to charge rearrangement, orbital hybridisation, and moiré potential modulation. Using long-range representations [1] for electronic-density prediction [2,3], we achieve low-energy band-structure predictions with <15 meV errors across twisted bilayer graphene (TBG), hBN, and transition-metal dichalcogenides (TMDCs), while ∼100 times faster than DFT. Descriptor requirements are material-dependent: homoatomic systems (e.g., TBG) are well-described by local descriptors, while hBN and TMDCs require long-range encoding. We summarise the physical implications of these findings that marry machine learning and the fundamental physics that governs the electronic density of twisted bilayer materials.
[1] A. Grisafi, M. Ceriotti, JCP 151, 204105 (2019)
[2] A. Lewis, A. Grisafi, M. Ceriotti, M. Rossi, JCTC 17, 7203 (2021)
[3] A. Grisafi, A. Lewis, M. Rossi, M. Ceriotti, JCTC 19, 4451 (2023)
Keywords: machine learning; density functional theory; twisted bilayer materials; moiré superlattice; electronic structure