High‐resolution peak analysis in TOF SIMS data

• Published in: Surface and interface analysis Vol. 53; no. 1; pp. 53 - 67
• Main Authors: Gelb, Lev D., Shahrokh Esfahani, Negar, Walker, Amy V.
• Format: Journal Article
• Language: English
• Published: Bognor Regis Wiley Subscription Services, Inc 01-01-2021
• Subjects: Amorphous materials; Bias; Data analysis; Mathematical models; Metallic glasses; Multilayers; Secondary ion mass spectrometry; Shape functions
• ISSN: 0142-2421; 1096-9918
• DOI: 10.1002/sia.6872

High mass resolution time‐of‐flight secondary ion mass spectrometry (TOF SIMS) can provide a wealth of chemical information about a sample, but the analysis of such data is complicated by detector dead‐time effects that lead to systematic shifts in peak shapes, positions, and intensities. We introduce a new maximum‐likelihood analysis that incorporates the detector behavior in the likelihood function, such that a parametric spectrum model can be fit directly to as‐measured data. In numerical testing, this approach is shown to be the most precise and lowest‐bias option when compared with both weighted and unweighted least‐squares fitting of data corrected for dead‐time effects. Unweighted least‐squares analysis is the next best, while weighted least‐squares suffers from significant bias when the number of pulses used is small. We also provide best‐case estimates of the achievable precision in fitting TOF SIMS peak positions and intensities and investigate the biases introduced by ignoring background intensity and by fitting to just the intense part of a peak. We apply the maximum‐likelihood method to fit two experimental data sets: a positive‐ion spectrum from a multilayer MoS2 sample and a positive‐ion spectrum from a TiZrNi bulk metallic glass sample. The precision of extracted isotope masses and relative abundances obtained is close to the best‐case predictions from the numerical simulations despite the use of inexact peak shape functions and other approximations. Implications for instrument calibration, incorporation of prior information about the sample, and extension of this approach to the analysis of imaging data are also discussed.