Modeling and control of a semiconductor manufacturing process with an automata network: An example in plasma etch processing

Abstract

A neural network model of a plasma gate etch process is described and compared with a statistical model. The neural network model has a correlation of 0.68 while the statistical model has a correlation of 0.45. From our model we deduce that the flow rate of the etching gases and the induced d.c.-bias are the key factors driving the etching and thus the remaining oxide thickness at the end of the etch. An adaptive neural network controller for water-to-water plasma etch control is also described. It uses real time process signatures and historical data from a relational database for a computation of the overetch time for the current wafer etching within the reactor. For an MOS gate etch the standard deviation of the oxide thickness between the gate and the source (or drain) is in the range of 10 Å. This is comparable to open-loop control or timed etch where the operator selects the ideal overetch time. The controller has thus achieved better than human equivalence.