Invited Paper AS-MoM8
Quantitative Surface Chemical Microscopy

Monday, October 15, 2007, 10:20 am, Room 610

X-ray photoelectron spectroscopy is widely regarded as a mature technique, with a large instrument base, not restricted to specialist laboratories. One of the reasons for this is the apparent ease of quantification compared with other surface analytical techniques, and its ability to provide chemical state information. Yet comparison of quantified data between different instruments is still problematic. A procedure will be described for determining the instrument response function for selected modes of operation to allow quantification using theoretically determined sensitivity factors, enabling transfer of results between instruments. XPS is much less frequently used in imaging mode, which was originally developed as a guide for small area analysis. However, the consequence of the acquisition of single energy images is that the aspects that make XPS spectroscopy appealing, ie ease of quantification and provision of chemical state information, are not available, since quantification requires peak area measurement after a suitable background subtraction and chemical state information is often dependent on resolution of overlapping photoelectron peaks. These limitations can be overcome by acquiring a spectrum at each pixel in an image, known as spectromicroscopy, so that the accepted processing procedures used in spectroscopy may be applied to the spectrum image data set. The acquisition of spectrum image data sets which may consists of 1000 images each containing 256 by 256 pixels presents further challenges for the analyst charged with the interpretation of thousands of spectra with low signal/noise. The use of multivariate statistical analysis to reduce the dimensionality of the data and to improve signal/noise will be demonstrated. Procedures will then be described to characterise the instrument performance in imaging mode, and to apply a modified quantification procedure to obtain atomic concentration images. Further, it will be shown that by maintaining the relationship between images and spectra so that pixels may be classified by chemistry, leads to improved curve fitting, and provides an alternative to multivariate curve resolution in visualizing physically meaningful spectra. Finally the ability to obtain spatially resolved nanostructural information will be discussed.