Speaker
Description
The memristive properties of transition metal dichalcogenides, such as MoS₂, have attracted significant attention and have recently been linked to the dynamics of sulfur vacancies [1, 2]. However, the inherently slow kinetics of sulfur vacancy migration in MoS₂ renders direct ab initio simulations computationally infeasible.
To overcome this limitation, we train both graph neural network machine-learning interatomic potentials (MLIPs) and Gaussian approximation potentials for this system.
Beyond systematically evaluating model architectures and training strategies, we emphasize the importance of validating physically relevant observables—specifically diffusion barrier heights—rather than relying solely on conventional metrics such as energy and force errors on randomly sampled test sets prior to molecular dynamics (MD) simulations.
Nanosecond-scale MD simulations applying the resulting MLIP reveal key mechanisms of cooperative vacancy transport providing a coherent explanation of irradiation-induced vacancy patterns, especially the formation of line defects spanning tens of nanometers [3].
[1] D.Li, B.Wu, X.Zhu, J.Wang, B.Ryu, W.D.Lu, W.Lu, X.Liang, “MoS₂ Memristors Exhibiting Variable Switching Characteristics toward Biorealistic Synaptic Emulation”, ACS Nano 12 (2018): pp. 9240–9252. https://doi.org/10.1021/acsnano.8b03977.
[2] B.Spetzler, D.Abdel, F.Schwierz, M.Ziegler, P.Farrell, “The Role of Vacancy Dynamics in Two-Dimensional Memristive Devices”, Adv. Electron. Mater. 10 (2023): 2300635. https://doi.org/10.1002/aelm.202300635
[3] A.Flötotto, B.Spetzler, R.von Stackelberg, M.Ziegler, E.Runge, C.Dreßler, “Large-Scale Cooperative Sulfur Vacancy Dynamics in Two-Dimensional MoS₂ From Machine Learning Interatomic Potentials”, Small 20 (2026): e10679. https://doi.org/10.1002/smll.202510679