Interpreting NIR Spectra

Unlike the relatively straightforward infrared spectra, which arise mainly from fundamental molecular vibrations and rotations, and where molecular components can be readily ascribed to the 'fingerprint' of peaks, the visual interpretation of NIR spectra is virtually impossible and speculative at best.

There are two approaches to interpreting spectra (Bonanno et al., 1992, p. 19):

1. Try and relate observed bands and peaks with known absorbing functional groups or chemical compounds;

2. Take a chemometric approach: ignore the question as to what causes the peak, and select the absorbing wavelength on an empirical basis to give the best correlation with traditional chemical analyses.

The difficulty with the first approach can be seen in the case of chloroform (CHCl3), which has only one absorbing CH group, yet possesses no less than 62 possible combination bands (Kaye, 1954). The presence of many C-H bonds in different molecular locations in organic macromolecules, such as the ligno-celluloses, will lead to vast numbers of absorption frequencies. Most of the absorbing frequencies derive from overtones and combinations of fundamental vibrations involving hydrogenic stretching modes (Osborne and Fearn, 1986, p. 29). There is also extensive overlapping and perturbation of the NIR absorption bands. It must be noted, however, that only combination bands arising from two different vibrational modes of the same functional group and having the same symmetry are allowable (Bonanno et al., 1992, p. 22). The CH stretching absorber is also present in proteins, oils and starch as well as cellulose, complicating the spectral interpretation. A similar inter-pretational problem exists with OH absorbers. These are present in simple sugars, but many of the same bands also appear in starch and cellulose (Murray and Williams, 1987). Williams (1991) has stated that 'At any wavelength area between 750 and 2500 nm, there is a multiplicity of absorbers, all of which may contribute to the spectrum of a commodity. For example, in the 2100 nm area, nearly 20 absorbers, including 2nd and 3rd overtones, can be identified, and the assignment of the wavelength to any particular absorber becomes rather specious.' The second, empirical approach, however, lacks a sound physico-chemical basis, but can be made to work under the right conditions. Irrespective of the compounds causing the overlapping spectral bands, it is the shape, that is the rate of change in slope with respect to wavelength, that conveys compositional information (Deaville and Baker, 1993). Different types of organic composite substances thus possess a 'fingerprint', and the relative position and magnitude of the peaks can be interpreted to yield information on the composition and relative amounts of substances present.

There is a warning from Shenk et al. that 'simply running data through the latest mathematical algorithm will result in nothing interpretable and is only pseudoscience.' Nevertheless, Givens (1993) claims that in most cases, these equations have been shown to provide a better prediction of forage digestibility, for example, than laboratory procedures.

Was this article helpful?

0 0

Post a comment