Principal Component Analysis in EDS

Spectral imaging is a very powerful microanalysis tool for chemical phase identification in a variety of samples. However, a spectral imaging (SI) data set contains vary large amounts of information and traditional approaches to analyzing these data sets rely on element identification and X-ray mapping. These methods have the potential to misidentify or completely overlook minor phases. To efficiently analyze these large amounts of data, automated methods are needed. To be useful for routine analysis, these methods must make no assumptions about the sample chemistry, be fairly robust, work with low volume or sparse data sets and deal with noisy data. The microanalysis software program, COMPASS, automatically performs data reduction and analysis on SI data sets and conforms to these requirements [1, 2].