Multiscale Modeling for MOCVD Synthesis and Characterization of Transition Metal Dichalcogenides

MOCVD enables to synthesize high-quality TMD layers from vaporized precursors by providing flexibility in the selection of the precursors and their flow rate. Herein, we develop a multiscale approach as a combination of ReaxFF, continuum fluid dynamics, phase-field (PF) and machine learning (ML) to model the gas-phase kinetic of the MOCVD growth and to connect further the MOCVD control parameters to the morphology, size, and distribution of the synthesized TMD materials. The results of the model that we developed, first, for MOCVD gas-phase kinetic of 2D-WSe2 correlate well with the experimental thickness measurements of 2D-WSe2 and show that the model is capable of simulating the experimentally observed trend [1]. We further extend this model to the combination of ReaxFF, ML, and PF, in progress. A systematic representative data set is generated based on the ReaxFF potential to train an ML-model describing the edge energies and edge-growth rates of 2D-WSe2 flakes as a function of key parameters, then to incorporate them into the PF simulations. The target of this work is to bridge spatial scales that range from 10−9 to 10-3 m in space and explore the optimal growth conditions resulting in high-quality TMD materials.
[1]Xuan et al. J. Cryst. Growth 2019, 527, 125247