(315a) Multiscale CFD Model Parallelization with Application to PECVD of Thin Films | AIChE

(315a) Multiscale CFD Model Parallelization with Application to PECVD of Thin Films


Crose, M. - Presenter, University of California, Los Angeles
Tran, A., University of California, Los Angeles
Ding, Y., University of California, Los Angeles
Christofides, P., University of California, Los Angeles
Accurate and efficient modeling of plasma-enhanced chemical vapor deposition (PECVD) has remained the goal of many research institutions due to far reaching applications to the manufacturing of silicon thin films for use in the photovoltaic and microelectronics industries [1],[2]. In the effort to improve product quality and to cut down on manufacturing costs due to the difficulties associated with in situ measurements and the fabrication of reaction chamber components, three-dimensional (3D) modeling which captures the physical geometry of common PECVD reactors has become invaluable [3]. Recently, Crose et al. proposed a novel multiscale computational fluid dynamics (CFD) model, which combined a macroscopic CFD domain with a microscopic surface domain. While this model was successful in capturing the behavior of the PECVD reactor with regard to the non-uniform deposition of a-Si:H films, and in allowing for improved reactor geometries to be explored, the computational costs of the model were non-trivial. Specifically, single batch simulations required days to complete rendering batch-to-batch studies (e.g., run-to-run control strategies) infeasible.

Motivated by the above considerations, an improved parallel computation strategy is applied which allows for the discretization of both the macroscopic CFD volume and the microscopic kMC algorithm across 64 cores of UCLA’s Hoffman2 computational cluster. The outlined multiscale model is applied to the deposition of 300 nm thick a-Si:H films in two distinct PECVD reactor geometries. The results suggest a significant improvement to the uniformity of the thin-film product (~4% change in the film thickness uniformity) and the reduction in computational costs suggest future implementation of run-to-run control strategies similar to those demonstrated by Crose et al. [4] for two-dimensional systems.

[1] Collins, D., Strojwas, A., White, D., “A CFD Model for the PECVD of Silicon Nitride,” IEEE Trans. Semicond. Manuf., 7, 176–183, 1994.

[2] Da Silva, A. and Morimoto, N., “Gas flow simulation in a PECVD reactor,” In Proceedings of the 2002 International Conference on Computational Nanoscience and Nanotechnology, San Juan, Puerto Rico, 22–25, 2002.

[3] Crose, M., W. Zhang, A. Tran and P. D. Christofides, "Multiscale Three-Dimensional CFD Modeling of PECVD of Amorphous Silicon Thin Films,'' Comp. & Chem. Eng., 113, 184-195, 2018.

[4] Crose, M., JSI Kwon, A. Tran and P. D. Christofides, "Multiscale Modeling and Run-to-Run Control of PECVD of Thin Film Solar Cells,'' Renewable Energy, 100, 129-140, 2017.