Are any scrips available to analyze FTIR spectra?

I was wondering if there are any scrips available to analyze FTIR spectrum (.csv). In particular I wanted to find the area under a functional group peak (or peak region) to more accurately compare relative concentration of functional groups. Maybe the best way to do this would be some Fourier series approximations of the data and than integrate?

Any help is greatly appreciated.

Answer 1

[Walter Roberson]

I have not worked with FTIR, but what you write reminds me of a program we developed about 20-ish years ago.

We had MRS data in which the data was noise plus the sum of contributions from a number of chemicals, with each contribution being a Gaussian. For any given peak, the position identified the species, and the amplitude and phase could be used to calculate the area (concentration.)

We varied two parameters per peak: the position and the amplitude. Given a list of those pairs, we could calculate the phase (and, I seem to recall, one other parameter, possibly half-width.) Then generate the Gaussians, sum them, subtract, sum of squares to get the error estimate.

The method we used to alter the position / amplitude pairs was Levison-Marquardt.

The method did require a guess about the number of peaks. Sometimes guessing too many peaks resulted in a Gaussian with large phase being non-trivially matched, but much more usual was for an extra peak to match a noise artifact. We did find, though, that every once in a while, an extra peak would end up spread very broadly, in a manner similar to subtracting near constant offsets, different ones for the left and right side of the "position" of the "peak".

We did find that the LM tended to get stuck, so we added a test for being stuck, and generated new random positions in that case.