The final two elements in the process of wavenumber calibration of a spectrometer involve software processing. These are (a) the generation of a spectrum through the Fourier transform processing of the measured interferogram (for Fourier transform spectrometers and (b) the determination of a single wavenumber value for each calibration band or line in the spectrum. How these steps are carried out can significantly affect the final wavenumber values that are to be compared with the standard values. We have examined severalimportant elements of this software processing: the methods used for transmittance minima evaluation as well as the effects of data interval (zero-filling), apodization, and resolution.We have developed a pair of methods for minima evaluation that exhibit excellent wavenumber value reproducibility throngincorporation of elements of the centroid method at fixed fraction. These methods are called the 'extrapolated centroid' (EC) method an the 'enhanced computational bisecting' (EB) method. These two extrapolation algorithms provide wavenumber values that are directly comparable with those of more commonly used methods such as absolute minimum (or maximum), polynomial curve fit, zero derivativeand others. These EC and EB methods will be used for certification of several NIST wavelength Standard Reference Materials (SRMs).A comparison of a number of methods, including the EC and EB methods, is made using a large number of spectra from calibration measurements of NIST polystyrene wavenumber SRMS.Results of the comparison including mean wavenumber values ancorresponding variances will be presented. In addition, results of parameter studies including data interval, apodization and resolution will be discussed.
Handbook of Vibrational Spectroscopy
John Wiley & Sons, Ltd., Hoboken, NJ
band analysis, polystyrene, wavenumber standard