Towards accurate and efficient process simulations based on atomistic and neural network approaches

Leveraging a powerful new methodology based on machine learning techniques coupled with atomistic simulation approaches, a previously undiscovered reaction sequence is predicted for Si epitaxial growth using $\mathrm{S}\mathrm{i}_{2}\mathrm{H}_{6}$ precursor. The complex surface reaction mechanism revealed by metadynamics simulations involves process controllable intermediate $\mathrm{S}\mathrm{i}\mathrm{H}_{2}$ formation that explains a long standing unresolved experimentally reported growth plateau appearance during $\mathrm{S}\mathrm{i}_{2}\mathrm{H}_{6}$ adsorption in the presence of H 2 carrier gas. Due to its accuracy and transferable flexible design, the proposed general workflow model can be applied to any deposition and etching process condition optimization to enhance thin film quality and to provide yield improvement pathways for next generation electronic device materials.