Chemometrics at SIMS XVII
Sep 17, 2009
The the 17th International Conference on Secondary Ion Mass Spectrometry, SIMS XVII, is being held this week in Toronto, CA. I am here with our Willem Windig. I’ve had an interest in SIMS for about a dozen years now, after being introduced to it by Anna Belu. Anna invited me to do a workshop on chemometrics in SIMS, gave me some data sets to work with, and off I went. SIMS was especially interesting to me because it provides very rich data sets, and is also often done over surfaces, producing what we refer to as multivariate images.
The use of chemometric techniques in SIMS has expanded considerably since I first got involved. The primary tool is still Principal Components Analysis (PCA), though there are certainly applications of other methods, including Maximal Autocorrelation Factors (MAF). However, I still see lots of examples where authors present 10, 20 or even 30 (!) images at different masses that are obviously highly correlated. These images could be replaced with one or two scores images and corresponding loadings plots: there would be less to look at, the images would be clearer due to the noise reduction aspects of PCA, and more might be learned about the chemistry from the loadings. So we still have a way to go before we reach everybody!
Over the years I’ve learned quite a bit about the non-idealities in SIMS. It certainly isn’t like like NIR spectra that follow Beer’s Law and have very high signal-to-noise! But I’ve gotten an additional appreciation for some of the issues from some of the talks here. There are plenty of “problems” such as detector dead time, low ion count statistics, effects due to topography, etc., which must be considered while doing an analysis.
I’ve been known to say that the secret to EVRI is data preprocessing, i.e. what you do the data before it hits the analysis method, e.g. PCA or PLS. That is certainly true with SIMS, and it’s nice to see that this is well appreciated among this group. The challenge for us as software developers will be to integrate these “correction” methods into our packages in a way that makes using them convenient.
I think that SIMS is ripe for more applications of Multivariate Curve Resolution (MCR), and in fact we focused quite a bit on that in our course here, Chemometrics in SIMS. With proper preprocessing, many applications of curve resolution should be possible. The more chemically meaningful results should be a hit with this audience, I would think.
I hope to see more MCR by the time of the SIMS XVIII, which will be held in Riva del Garda, ITALY, September 18-23, 2011. I plan to keep working in this area, if only so I can go to that! Hope to see you there.