AVS 45th International Symposium
    Applied Surface Science Division Thursday Sessions
       Session AS-ThM

Paper AS-ThM7
Trajectory Projection Factor Analysis

Thursday, November 5, 1998, 10:20 am, Room 307

Session: Gaede-Langmuir Award Address and Quantitative Surface Analysis
Presenter: D.J. Pocker, IBM SSD
Authors: D.J. Pocker, IBM SSD
B.R. York, IBM SSD
Correspondent: Click to Email
Linear least squares (LLS) analyses of arrays of Auger or ESCA spectra (as from depth profiles, line scans or maps) run into problems when the data does not contain regions with pure spectra for one or more of the compounds. In addition, target factor analysis (TFA) encounters problems when there is not prior knowledege of the profiles for each compound. We present here an eigenanalysis and coordinate transformation method which straightforwardly extracts spectra for the pure compounds along with their associated profiles from arrays with overlapping spectra of up to 4 compounds. In the method, the profiles for the mathematically distinct eigenspectra are plotted against each other, yielding a TRAJECTORY through the orthogonal eigenspace. PROJECTION of this trajectory onto/into a well chosen plane or volume then allows a simple transformation of coordinates to define a non-orthogonal, chemically interesting space. From this, the pure compound spectra and profiles are easily computed. (The limit of 4 is from the graphical nature of the analysis. The concept extends to an arbitrary number of compounds. Also, for favorable 4 compound cases, overlaps as bad as "never fewer than 3" can be treated.)