Pittcon 2009 Day 1: FTIR Mixture Analysis Software Packages

The first day of Pittcon was exciting for me, for the wrong reasons. I have come down with a nasty cold which has made standing and talking, the most common activity at Pittcon, a little more interesting than I would like. I hope I have not inadvertently infected any of my colleagues.

I visited a number of FTIR instrument company booths on Monday. There is much new and interesting to talk about, but one thing that really excites me are new software programs that assist the FTIR user in mixture analysis. As I mention in my Fundamentals of FTIR and IR Spectral Interpretation I courses, mixture analysis is the biggest practical disadvantage of FTIR. The problem is that the more chemically complex a mixture becomes the more complex the infrared spectrum becomes, making it harder to figure out what peaks are from what molecules. In my training courses I teach attendees the 4 ways of tackling mixtures. These mixture analysis software packages may represent a fifth way.

Time for a disclaimer. Two vendors, the Bio-Rad Informatics division and Thermo-Nicolet are demonstrating mixture analysis software here at Pittcon. My discussion of their offerings is in no way an endorsement of their products. If I find other companies showing mixture analysis software I will talk about them in a later post.

These mixture analysis software packages work like library searching, which is a technique many FTIR users are familiar with. In this case the mixture spectrum is selected, one or more spectral libraries are selected, and then the mixture search is performed. The algorithms on both systems are trade secret, but my hunch is that they use some sort of chemometric modeling, perhaps principle components or partial least squares analysis. For the results to work out well you have to tell the software how many different chemical components you think there are in the sample. Knowing this information greatly improves the quality of the results. However, as those of us who work in the real world know, we don't always know the exact number of components in a sample. I also noticed that for the most part you are limited to mixtures with 2, 3, or 4 components. Not surprisingly, the calculation time increases for each added component.

My hunch then is if you tell the software you have a two component mixture it will take spectra from the selected libraries two at a time, calculate mixture spectra from them, and compare them to your sample spectrum. This comparison gives a number similar to the hit quality index (HQI) in a normal library search. The pair of library spectra that when added together give the best match to your sample spectrum should give the best HQI. You can visually compare the calculated and sample spectra. Other tools, including the spectral residual, which is the result of subtracting the calculated spectrum from the sample spectrum, are available to judge the quality of the results. I think certain FTIR users, particularly those unfamiliar with IR Spectral Interpretation, may find this type of software package useful.

I am sure many of you are wondering whether these mixture analysis programs can provide quantitative information i.e.the percentage of different compounds in a sample. There was much careful talk about this issue from which I was not able to draw a conclusion. This is something I would like to see an unbiased third party put to the test (hint hint).

A blurb about the Bio-Rad software is here: http://collateral.knowitall.com/collateral/95372-Mixture_Analysis_Datasheet.pdf#zoom=75%

A blurb about the Nicolet software is here: http://www.thermo.com/com/cda/product/detail/1,,10137344,00.html