Characterizing and Predicting Spatial Nonuniformity in the Deep Reactive Ion Etching of Silicon

Abstract

We present a way of predicting spatial nonuniformity in deep reactive ion etching (DRIE). In addition to well-known feature size or aspect-ratio-dependent etch (ARDE) nonuniformity, etch rate can vary between regions of the wafer, such that supposedly identical microelectromechanical systems (MEMS) fabricated on that wafer may vary in performance or fail in use. Wafer- and die-scale uniformity may be improved by changing conditions in the etching chamber, but this usually compromises etching speed and may accentuate ARDE. An alternative approach is to lay out the patterns being etched so that they compete for reactants in a controlled way. We demonstrate a way of precharacterizing any DRIE tool plus associated “recipes” of operating parameters. Several test wafers are etched and measured, and simple data-fitting algorithms are run. The model constructed captures etch nonuniformity occurring over the diameter of a wafer plus variations caused by the localization of patterns within the wafer. Our technique models etch rate variation across experimental wafers with fitting errors of between 0.8 and 4.5% root–mean–square per wafer. Our model can then predict etch rate on a lateral grid for any etched pattern. We envisage this method being integrated into computer-aided MEMS design systems.