The use of second-derivative spectra for pure-variable based self-modeling mixture analysis techniques

Abstract

Recently, the use of second-derivative spectra was introduced as an intermediate step in self-modeling mixture analysis. Since special problems arise using second-derivative spectra, a follow-up study was needed to evaluate the calculation of second-derivative spectra with respect to self-modeling mixture analysis techniques. A critical parameter is the size of the smoothing window, which should be large enough to reduce noise, but small enough to avoid distortion of the spectra. This paper will discuss this follow-up study and show methods to obtain the optimal smoothing, using a near-IR spectral data set of solvents, and an FTIR-microscopy data set of polymer laminate spectra. The results are obtained using the SIMPLISMA approach, but can also be applied to other pure-variable methods.