Machine learning plasma-surface interface for coupling sputtering and gas-phase transport simulations

A unified model of thin film deposition by means of sputter deposition requires a description which bridges the intrinsic time and length scales of both the solid and the gas-phase. As a proof of concept, in this work such a description is established based on machine learning techniques using artificial neural networks. Initially, sputtered particle distributions are obtained from transport of ions in matter based simulations for Ar projectiles bombarding a Ti–Al composite for a number of representative incident particle energy distribution functions. Subsequently, a multilayer perceptron network is trained and verified with this set of incident/outgoing distributions. An error analysis is carried out for the obtained training results and their quality is compared and discussed for two sets of hyperparameters. Therewith, it is demonstrated that the trained network is able to predict the sputtered particle distributions for unknown, arbitrarily shaped incident ion energy distributions. It is consequently argued that the trained network may be readily used as a machine learning based model interface, which is sufficiently accurate also in scenarios which have not been previously trained.