Applications of Kinetic Monte Carlo Simulations and Machine Learning to model Atomic Layer Deposition (ALD) of Metal Oxides.

Metal-oxides such as ZnO or Al₂O₃ synthesized through Atomic Layer Deposition (ALD) have been of great research interests as the candidate materials for ultra-thin tunnel barrier layers is. In this study, we have applied a 3D on-lattice Kinetic Monte Carlo (kMC) code developed by Timo Weckman’s group to simulate the growth mechanisms of the tunnel barrier layer and to evaluate the role of various experimentally relevant factors of the ALD processes. We have systematically studied the effect of parameters such as the chamber pressure and temperature, pulse/purge times, as well as the coverage of wetting layer on the substrate. The database generated from the kMC simulations was subsequently used as descriptors in the subsequent analyses via Machine Learning algorithms. The results of a combined approach of kMC and ML were then compared to the experimental results. The support from NSF (EPMD Division) Award No. 1809284 and the computational support from NERSC are gratefully acknowledged.