MURAL - Maynooth University Research Archive Library



    Unsupervised Feature Extraction Techniques for Plasma Semiconductor Etch Processes


    Ma, Beibei (2009) Unsupervised Feature Extraction Techniques for Plasma Semiconductor Etch Processes. PhD thesis, National University of Ireland Maynooth.

    [img] Download (17MB)


    Share your research

    Twitter Facebook LinkedIn GooglePlus Email more...



    Add this article to your Mendeley library


    Abstract

    As feature sizes on semiconductor chips continue to shrink plasma etching is becoming a more and more critical process in achieving low cost high-volume manufacturing. Due to the highly complex physics of plasma and chemical reactions between plasma species, control of plasma etch processes is one of the most di±cult challenges facing the integrated circuit industry. This is largely due to the di±culty with monitoring plasmas. Optical Emission Spectroscopy (OES) technology can be used to produce rich plasma chemical information in real time and is increasingly being considered in semiconductor manufacturing for process monitoring and control of plasma etch processes. However, OES data is complex and inherently highly redundant, necessitating the development of advanced algorithms for e®ective feature extraction. In this thesis, three new unsupervised feature extraction algorithms have been proposed for OES data analysis and the algorithm properties have been explored with the aid of both arti¯cial and industrial benchmark data sets. The ¯rst algorithm, AWSPCA (AdaptiveWeighting Sparse Principal Component Analysis), is developed for dimension reduction with respect to variations in the analysed variables. The algorithm gener- ates sparse principle components while retaining orthogonality and grouping correlated variables together. The second algorithm, MSC (Max Separation Clustering), is devel- oped for clustering variables with distinctive patterns and providing e®ective pattern representation by a small number of representative variables. The third algorithm, SLHC (Single Linkage Hierarchical Clustering), is developed to achieve a complete and detailed visualisation of the correlation between variables and across clusters in an OES data set. The developed algorithms open up opportunities for using OES data for accurate pro- cess control applications. For example, MSC enables the selection of relevant OES variables for better modeling and control of plasma etching processes. SLHC makes it possible to understand and interpret patterns in OES spectra and how they relate to the plasma chemistry. This in turns can help engineers to achieve an in-depth under- standing of underlying plasma processes.

    Item Type: Thesis (PhD)
    Keywords: Feature Extraction Techniques; Plasma Semiconductor Etch Processes;
    Academic Unit: Faculty of Science and Engineering > Electronic Engineering
    Item ID: 4079
    Depositing User: IR eTheses
    Date Deposited: 15 Jan 2013 11:19
    URI:

      Repository Staff Only(login required)

      View Item Item control page

      Downloads

      Downloads per month over past year