Greifswald 2024 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

P: Fachverband Plasmaphysik

P 3: Plasma Wall Interaction I

P 3.5: Talk

Monday, February 26, 2024, 15:15–15:30, ELP 6: HS 3

Data-integrated multiphysics simulations of reactive magnetron sputtering — •Tobias Gergs¹, Luca Vialetto^1,2, Christian Stüwe¹, and Jan Trieschmann¹ — ¹Theoretical Electrical Engineering, Kiel University, Kaiserstraße 2, 24143 Kiel, Germany — ²Department of Aeronautics and Astronautics, Stanford University, 496 Lomita Mall, Stanford, CA 94305, United States of America

Reactive magnetron sputtering is widely used in science and industry. However, the understanding of the physical kinetics remains incomplete, primarily because the intrinsic length and time scales of the plasma and the surface differ by orders of magnitude. Individual scientific disciplines have frequently concentrated on only one of these aspects in detail (i.e., plasma or surface), while the other aspect may have been considered in a simplified manner. In this work, established and novel methods are combined to adequately describe the coupled plasma and surface physics involved in the sputter deposition of silicon oxide in Ar/O₂ discharges. The dynamics of the plasma are described by 2d3v particle-in-cell simulations with a Monte Carlo transport scheme for charged particles, energetic neutrals, and sputtered atoms. The surface evolution is determined by rate equations for the surface coverage, which account for chemisorption, physisorption, diffusion of adatoms, and physical sputtering. The energy and angular distributions of sputtered particles are incorporated by an integrated machine learning model, which was trained with Monte Carlo simulation data. The influence of process parameters (e.g., admixtures of O₂) on phenomena such as target poisoning is emphasized.

Keywords: plasma-surface interaction; sputtering; machine learning; artificial neural network; silicon oxide