Modeling of Plasma Etch Systems Using Ordinary Least Squares, Recurrent Neural Network, and Projection to Latent Structure Models

Abstract

In microelectronics manufacturing, control strategies for plasma etch systems have been limited to traditional statistical process control and recipe control techniques. The lack of in situ real-time measurements of process performance and appropriate models has hindered the introduction of feedback control systems. This paper focuses on empirical model building for advanced process control using two real-time diagnostic sensors for measurement of the reactor state. Laser interferometry for measurement of etch rate and voltage and current probes for measurement of effective radio-frequency power and sheath voltage, coupled with data acquisition hardware and software, provided the foundation for steadystate and dynamic model development of the plasma etch process. Several linear and nonlinear steady-state techniques including ordinary least squares, neural networks, and projection to latent structures were used in empirical model building. Both linear regression and recurrent neural network model structures provided a satisfactory fit of the data for the operating space investigated. Projection to latent structures techniques indicated that the most relevant variables were power, pressure, and chamber impedance. The addition of the impedance measurement significantly improved the predictive capability of the model. ) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.83.63.180 Downloaded on 2015-06-05 to IP ) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.83.63.180 Downloaded on 2015-06-05 to IP

Resolving critical dimension drift over time in plasma etching through virtual metrology based wafer-to-wafer control

“…[24][25][26][27] Since data generated in plasma etch equipment has collinearity, measurement noise and huge amount of sensor variables, PLSR works well in plasma etching. 13,[28][29][30] PLSR can avoid collinearity problems among variables in such a way to utilize orthogonal latent variables. The orthogonal latent variables are result of re-arrangement in such a way to have maximum correlation between input and output variables.…”

Feedback control of plasma etching reactors for improved etching uniformity

“…The improved understanding of plasma properties has led to the development of accurate mathematical models for various plasma etching processes which can be used to compute reactor con"gurations that reduce the etching nonuniformity (e.g. Park & Economou, 1990;Park & Economou, 1991;Bushman, Edgar, & Trachtenberg, 1997;Venkatesan, Trachtenberg, & Edgar, 1987). Such studies have revealed that the use of a showerhead arrangement to introduce the precursor gas into the reaction chamber, signi"cantly reduces the radial etching nonuniformity.…”

A study in dynamic neural control of semiconductor fabrication processes

“…We offer a software implementation based on the latter approach using a cascade correlation neural network for the critical prediction algorithm. While other such neural-network approaches to quality metric prediction for semiconductor tools have been reported [3], [14]- [16], we offer a predictive model that accounts for both continuously modulated variables as well as time-dependent maintenance variables, such as parts cleaning, and replacement. This mixed model was developed to serve as a real time aid for the tool operator and engineer, and incorporates all the actions that are considered for maintenance of process control on a daily basis.…”