Run-to-Run Control of Thermal Atomic Layer Deposition

Abstract

This work focuses on the development of a computational fluid dynamic (CFD) model of a batch atomic layer deposition (ALD) process and an associated run-to-run control scheme. Specifically, a cylindrical furnace reactor is analyzed for a SiO 2 thin-film ALD using BTBAS and ozone as precursors. First, a high-fidelity 2D axisymmetric multiscale CFD model for an industrial-scale furnace ALD system is developed in ANSYS Fluent to characterize the gas-phase development and the surface deposition, which is based on the previously constructed database using the kinetic Monte-Carlo (kMC) algorithm. After the validation of the multiscale CFD model, it is utilized to investigate a wide range of operating conditions, from which a regression model is developed to describe the input-output relationship between the inlet feed flow rate and the half-cycle time. Next, a run-to-run (r2r) control scheme is formulated, which uses the post-batch feedback information to adjust the operating conditions using the regression relationship and an exponentially weighted moving average (EWMA) algorithm. Finally, the multiscale CFD model and the r2r controller are integrated to generate a closed-loop system via a message-passing interface (MPI) and a data synchronization scheme to evaluate the performance of the r2r controller.