(484f) Operation and Control of Spatial Atomic Layer Etching Process

With the growing dependence on semiconductors to sustain the next technological revolution, the semiconductor manufacturing industry has struggled to meet rising consumer demand [1]. Semiconductor materials, which have been consistently shrinking in size and agree with the trends postulated by Moore's Law, are becoming more difficult to fabricate due to their stricter conformance requirements, specifically, their need for precise measurements in the sub-5-nm dimensions. In response to this obstacle, the emergence of self-limiting processes including Atomic Layer Deposition (ALD) and Atomic Layer Etching (ALE) can deliver product conformance but are inefficient processes that fail to reduce the manufacturing time due to their need for long purging cycles. Spatial atomic layer deposition (SALD) and spatial atomic layer etching (SALE) are refined processes that reduce the length of purging times, but their lack of experimentation and characterization has made it difficult to integrate these developing processes into industries.

Prior research has been conducted to identify optimal operating conditions using machine learning [2] and reactor design [3] using a multiscale modeling approach that integrates microscopic modeling simulated by the kinetic Monte Carlo (kMC) algorithm and computational fluid dynamics (CFD) in the macroscopic scale. However, these optimal ranges require rigorous control systems to ensure that these operating conditions are regulated and the effects of disturbances are minimized. Prior research was conducted using a batch control algorithm known as run-to-run (R2R) control using exponentially weighted moving average (EWMA) regression modeling for ALE processes [4]. In order to ensure the efficacy of the process environment for SALE reactors, this research aims to construct a model predictive control (MPC)-based run-to-run (R2R) control system for SALE of aluminum oxide thin films. Specifically, we develop a control algorithm to regulate the speed of the wafer as it goes through the spatial ALE process to ensure complete etching by the end of the batch. The controller uses measurements of wafer etching completion at the end of the batch to determine the speed needed for an incoming wafer to achieve complete etching in the presence of process variability without using an unnecessarily slow speed that is economically suboptimal as it is currently done in industry. The control system implementation is demonstrated through application to a multiscale computational fluid dynamics model of the ALE process and its performance and robustness are evaluated via a series of industrially-relevant operational scenarios.

[1] J. Voas, N. Kshetri, J.F. DeFranco. Scarcity and global insecurity: the semiconductor shortage. IT Professional, 23, 78-82, 2021.

[2] S. Yun, M. Tom, J. Lou, G. Orkoulas, P.D. Christofides. Microscopic and data-driven modeling and operation of thermal atomic layer etching aluminum oxide thin films. Chemical Engineering Research and Design 177, 96-107.
[3] Yun, S., M. Tom, F. Ou, G. Orkoulas, P. D. Christofides. Multiscale computational fluid dynamics modeling of thermal atomic layer etching: application to chamber configuration design. Computers & Chemical Engineering, 161, 107757, 2022.
[4] Yun, S., M. Tom, F. Ou, G. Orkoulas, P. D. Christofides. Multivariable run-to-run control of thermal atomic layer etching of aluminum oxide thin films. Chemical Engineering Research & Design, 182, 1-12, 2022.