XPS spectromicroscopy: exploiting the relationship between images and spectra

Abstract

The application of spectroscopic processing techniques to multi-spectral XPS data sets has enabled the acquisition of quantitative surface chemical state images. Such data sets are necessarily large, incorporating many spectra, so prohibiting interactive processing. Instead multivariate analytical techniques are used to reduce the dimensionality of the data, and also to increase the signal/noise, there bye speeding acquisition. These techniques may also be used to classify regions in images according to different chemistry, that is changes in photoelectron intensity, changes in binding energy and changes in the inelastic background. Spectra from classified regions may then be summed to aid visualisation, obviating the need for multivariate curve resolution with its attendant uncertainties. Further the inelastic background of transmission corrected spectra from classified regions may be modelled to provide spatially resolved in-depth information. Such classification also aids curve fitting, since curve fit models can be applied to regions of similar chemistry. IntroductionX-ray photoelectron spectroscopy is a mature surface analytical technique, with a large installed instrument base, providing relatively easily quantified chemical state information, which is reflected by the number of publications in scientific journals. However, despite the fact that the first commercially available XPS instrument capable of imaging was announced in 1990, [1] relatively few publications, by comparison, have concerned imaging XPS. This is a consequence of instrument manufacturers implementing the acquisition of single energy images as a guide to selected area analysis, and as a result, losing the very features that made XPS spectroscopy popular, namely, quantification and the acquisition of chemical state information. Single energy images, and even peak minus background images, are incapable of accounting for the background below photoelectron peaks, or for changes in the peak shape due to chemical shifts. Further, they cannot resolve overlapping photoelectron peaks. Indeed, even as a guide to selected area analysis, single energy images perform poorly, since they can only truly represent changes occurring across an image if there are only two components which change. Yet the acquisition of quantifiable chemical state images is clearly preferable to analyses at a number of discrete points.More recently, however, a number of publications have described the acquisition of spectrum image datasets, [2] where each pixel in an image contains a spectrum. Such datasets may be acquired by scanning an X-ray probe over the surface while acquiring a spectrum at each pixel, or by acquiring a series of energy-filtered images incremented in energy, known as parallel imaging. The latter mode is preferred since it is quicker and minimises sample damage due to X-ray exposure. Two different types of instruments are capable of parallel acquisition. One employs a cathode lens, [3] the other a combination magnetic/electrostatic lens. Using laborator...

High‐resolution XPS spectromicroscopy

“…24 of Ref. [1]), seem limited in practice to low magnification levels (with typical field of view in the 200-800 µm range) [17,18,19] allowing reasonable acquisition times. For parallel imaging, the reason is twofold: on the one hand, the resolution limit set by the aberrations of the entrance lens of the imaging spectrometers is in the order of 3 µm: this yields practical routine resolutions at least one order of magnitude higher with reasonable acquisition times; on the other hand, despite the high transmission of the available microscopes the counting statistics might still be insufficient for higher magnification, routine sub-10 µm XPS imaging.…”

Rapid multivariate analysis of 3D ToF‐SIMS data: graphical processor units (GPUs) and low‐discrepancy subsampling for large‐scale principal component analysis

“…PCA has therefore been an important method of analysing spectra and images in surface analysis for at least the last 25 years. There exist excellent examples [2][3][4][5][6][7][8] and analytical reviews [9][10][11] of its use in the literature, applied to a range of problems.…”

Multivariate analysis of extremely large ToFSIMS imaging datasets by a rapid PCA method

“…PCA has therefore been an important method of analysing spectra and images in surface analysis for at least the last 25 years. There exist excellent examples [2][3][4][5][6][7][8] and analytical reviews [9][10][11][12][13] of its use in the literature, applied to a range of problems. At the core of PCA software is Singular Value Decomposition (SVD), a matrix algebra method for decomposing spectra into orthogonal (i.e.…”