Mp. Seah et Mt. Brown, VALIDATION AND ACCURACY OF SOFTWARE FOR PEAK SYNTHESIS IN XPS, Journal of electron spectroscopy and related phenomena, 95(1), 1998, pp. 71-93
The development of a range of VAMAS format data files containing synth
etic I:est spectra and their use in evaluating the peak fitting routin
es of three commercial XPS data analysis systems A, B and C, is report
ed. The synthetic spectra allow tests for: (1) acceptance of VAMAS for
mats with kinetic energy or binding energy scales that may have positi
ve or negative abscissa increments; (2) Gaussian-Lorentzian sum or pro
duct function peaks; (3) effectiveness of Shirley background subtracti
on; (4) effectiveness of Tougaard background subtraction; (5) peak syn
thesis with partly resolved peaks with no background; and (6) peak syn
thesis with a Shirley background. The results show that ail the data s
ystems studied accept VAMAS format files but not all would accept both
of the options of direction of scan. In addition, each data system ha
d programming faults, some of which were later rectified. Synthetic sp
ectra approximating the carbon Is peaks for poly(methyl methacrylate),
poly(vinyl acetate), poly(vinyl chloride) and poly(isobutylene) allow
an analysis of accuracy of the fits in these tests. One full set of s
pectra is constructed with Gauss-Lorentz product function peaks and a
second full set with Gauss-Lorentz sum function peaks since each softw
are was designed for fitting only one of these options. The fitting of
the individual peaks is only good for software using the same peak sh
apes as those of the data. The fit quality can deteriorate markedly fo
r fitting to a peak shape different to that used in the peak fitting a
lgorithm (ie fitting sum function peaks with product function software
or vice versa). For unresolved peaks additional information is requir
ed to provide meaningful intensities. The three software systems studi
ed all have different facilities but it is clear that the quality of c
onvergence to the correct result gives a general order of preference i
n the fitting algorithm of B, A and C. Other aspects can favour softwa
re A which has more facilities. (C) 1998 Elsevier Science B.V. All rig
hts reserved.