Temperature-Selected Defect–Moiré Coupling Controls Nonradiative Charge Recombination in Twisted MoS2

Abstract

Moiré superlattices in twisted transition-metal dichalcogenides provide a versatile platform for engineering charge carrier dynamics, yet interactions between intrinsic defects and dynamically reconstructed moiré potentials under experimentally relevant conditions remain poorly understood. Here, we develop a machine-learning-accelerated nonadiabatic molecular dynamics (ML-NAMD) framework combining a DeePMD-based ML force field, an E(3)-equivariant Hamiltonian neural network, and NAMD, enabling nanosecond-scale structural sampling and electronically resolved carrier recombination dynamics in thousand-atom pristine and sulfur-vacancy-containing 3.48° twisted bilayer MoS2 supercells. We show that the sulfur vacancy is not a static recombination center; instead, its local moiré environment is thermally selected through coupling to dynamic lattice reconstruction. At 300 K, the vacancy-adjacent region evolves toward an RMoMo-like environment, an R-type Mo-on-Mo stacking with vertically aligned Mo atoms across the layers, whereas at 50 K it favors reconstructed commensurate domains or nearby domain wall configurations. This temperature-selected local stacking governs defect–moiré coupling and, consequently, carrier recombination. Sulfur vacancies induce femtosecond electron capture into defect trap states, but subsequent nonradiative recombination is strongly stacking dependent. In reconstructed regions, a deep-trap-mediated pathway accelerates recombination, whereas in the RMoMo region it is strongly suppressed, producing a 6.7-fold longer carrier lifetime. These findings establish a microscopic mechanism of temperature-selected defect–moiré coupling, in which local lattice reconstruction determines whether defect and moiré potentials cooperate to activate, or compete to suppress, trap-mediated nonradiative recombination. This mechanism suggests practical strategies for reducing nonradiative losses and improving the performance of moiré optoelectronic devices by controlling local reconstruction, temperature, twist angle, and defect passivation.