Paper TF-ThP13
Structure Characterization of PECVD a-SiCN:H Thin Films: Toward Machine Learning Algorithms for Modeling of Complex Disordered Solids

Thursday, October 24, 2019, 6:30 pm, Room Union Station B

Plasma-enhanced chemical vapor deposition (PECVD) is a routinely employed process in thin-film technologies. Despite its array of advantages and affordability, it suffers from the lack of systematic principles to define growth conditions for an intended output. A deeper understanding of plasma processes is necessary for rational design and strategic synthesis of robust materials spanning a broad spectrum of applications. The properties of these materials are highly dependent on structure; and the structure varies as a function of growth conditions. Interpreting or predicting the effects of PECVD process variables such as temperature, pressure, flow rate and plasma power density on structural features of thin-films is a formidable task. The traditional ‘cook and quench’ molecular dynamics approach is incapable of replicating the relatively longer time scales and non-thermodynamic nature of the actual experiment. An alternative approach entails advanced machine learning algorithms applied not to reproduce, but rather to map the process-structure-property correlations. However, this requires training data in the form of empirically determined chemical models obtained under known process conditions. Here, PECVD grown amorphous hydrogenated SiCN thin films obtained from structurally different molecular precursors are studied to compile such a data set due to their stability, scope for precursor synthesis, and compatibility with various characterization techniques: FT-IR, solid-state NMR, fluctuation electron microscopy (FEM), as well as X-ray and neutron diffraction. We present the effects of process parameters on a-SiCN:H thin films, extensive structure and property characterization, and propose chemical structure models.