Method and apparatus for endpoint detection using partial least squares

Abstract

An apparatus and method for detection of a feature etch completion within an etching reactor. The method includes determining a correlation matrix by recording first measured data regarding a first etch process over successive time intervals to form a first recorded data matrix (110), assembling a first endpoint signal matrix using target endpoint data for a specific etch process, performing a partial least squares analysis (130) on the recorded data matrix and the first endpoint signal matrix to refine the recorded data matrix, and computing a correlation matrix (180) based upon the refined recorded data matrix and the first endpoint signal matrix. The method further includes performing a second etch process to form a second recorded data matrix. The correlation matrix and the second recorded data matrix are analyzed to determine whether an endpoint of the second etch process has been achieved.

Description

[0001] Cross References to Related Applications [0002] This application claims priority to U.S. Application Serial No. 60 / 277,981, filed March 23,2001. The content of this application is hereby incorporated by reference. technical field [0003] The present invention relates generally to endpoint detection in semiconductor manufacturing processes. Background technique [0004] The present invention recognizes the problems associated with conventional fabrication of reaction chambers and methods of using those chambers, which are addressed by the use of the present invention. [0005] Typically, during semiconductor processing, material is removed or etched along fine lines or in vias or contacts patterned on a silicon substrate using a (dry) plasma etch process. A plasma etch process generally involves positioning a semiconductor substrate with an overlying patterned protective layer, such as a photoresist layer, in a processing chamber. Once the substrate is positioned...

AI Technical Summary

Problems solved by technology

However, a disadvantage of using PCA for multivariate analysis of light emission data includes the mathematical rigor and complexity inherent with such an analysis, and, more importantly, the lack of insight into etch process-related energy yields. The use of a set of physical criteria for the data including endpoint signals

Images