Plasma-Surface Interactions I

Multiscale models for plasma-surface interfaces (PSIs) often involve coarse-grained surface models for particle exchange as a boundary condition in plasma codes because the disparate scales challenge multiscale integration of atomistic surface simulations with plasma models. However, these approaches often neglect the underlying surface composition/morphology that is needed to capture the complexity of particle exchange in a PSI, such as implantation, reflection, and physical/chemical sputtering. This talk will describe the recent development of a surface-state-dependent model for particle exchange in a carbon-tungsten PSI through machine learning (ML) of atomistic simulations of independent ion bombardment events and progress applying these methods to capture surface state evolution and particle exchange during etching of chlorinated silicon by argon. Molecular dynamics (MD) simulations were produced by Sandia's massively parallel MD code, LAMMPS, to obtain statistics on incoming/outgoing particle distributions and changes to the surface state. ML models were trained to the data to extract a complete, but tractable set of surface parameters that influences particle exchange in the interface, track evolution of the surface parameters, and predict outgoing particle distributions as a function of surface state and incident particle energy/trajectory. Model accuracy will be evaluated through comparison of out-of-sample data sets and the challenges and next steps associated with capturing atomistic character within coarse-grained models will be discussed.

Funding acknowledgement

This work is supported by Sandia National Laboratories under the LDRD program. Sandia is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.