Multivariate Analysis of SIMS Spectra

The information content from a single SIMS experiment is phenomenal. The technique is so rich in detail that it is difficult to know where, or how, to begin the interpretation stage. Each year that goes by sees an increase in one, or more, of mass resolution, sensitivity and speed of acquisition. The spectra become more detailed and the ability to collect lots of them becomes easier. In an attempt to stem the tide, most practitioners have chosen to implement statistical approaches to produce summaries of the data. However, the complexities of modern spectra require advanced multivariate statistics to extract the maximum amount of information. While univariate statistics is taught in most schools, the researcher is often more interested in developing instrumental and sample preparation abilities than developing their mathematical skills. This can leave a gap in knowledge that we will attempt to bridge in this text. There are any number of books available discussing multivariate statistical analysis, many with a chemometric bias. Here we concentrate on the aspects of multivariate analysis that the SIMS analyst will encounter and we seek to explain, in depth, the processes involved in performing, optimising and validating our procedures.