AVS 53rd International Symposium
    Applied Surface Science Thursday Sessions
       Session AS-ThP

Paper AS-ThP34
Preliminary Report of Evaluation of Automated Peak Detection Procedure in X-ray Photoelectron Spectra

Thursday, November 16, 2006, 5:30 pm, Room 3rd Floor Lobby

Session: Aspects of Applied Surface Science Poster Session
Presenter: M. Suzuki, ULVAC-PHI, Inc., Japan
Authors: M. Suzuki, ULVAC-PHI, Inc., Japan
S. Fukushima, NIMS, Japan
S. Tanuma, NIMS, Japan
Correspondent: Click to Email
X-ray photoelectron spectra should be analyzed to detect peaks prior to elemental/chemical assignment and the peak detection routine has been discussed in ISO/TC201/SC3 activity. It is needed to estimate its effectiveness of the software as well as the uncertainties of the results and it has been investigated in VAMAS activity. Prior to the world-wide interlaboratory study, the preliminary investigation was performed to make artificial spectral data and analyze their efficiency using a peak detection software routine that included the three algorithms with (1) second derivative method, (2) peak-to-background method, and (3) background estimation method. Test spectra were composed from the actually measured Au, Ag, and Cu spectra and the artificially noise-superposed spectra were also composed. The number of the basic spectra is three for the mixing ratio of (Au, Ag, Cu) as (1, 1, 1), (1, 0.1, 0.01), and (1, 0.01, 0.001). The noise-superposed spectra have been prepared for each basic spectrum, considering the noise level and the randomization for each noise level. These basic and noise-superposed spectra were analyzed for peak detection using the software built according to the algorithms above mentioned. The number of detected peaks and the efficiency that is the relative number against the detection number by eyes have been compared for the three basic spectra. The quality of (1, 1, 1) is shown to be different from the other two spectra of (1, 0.1, 0.01), and (1, 0.01, 0.001). For the spectrum of (1, 1, 1) the second derivative method gave a large number of detected peaks than the others. On the other hand, for the spectra of (1, 0.1, 0.01), and (1, 0.01, 0.001) the background method gave a greater number of peaks than the other methods. The number of detected peaks for the noise-superposed spectra drastically decreased from those for the basic test spectra. Detailed results and investigations will be presented and discussed at the conference site.