Understanding and Correcting the Factors Which Affect Diffuse Transmittance Spectra

transmitting samples are characterised by poor reproducibility, poor linearity, high noise and high sensitivity to sample measurement geometry. Typical calibration procedures for such spectra involve pretreatment with multiplicative scatter correction or standard normal variant correction, first or second derivative, and partial least squares regression. Artificial neural networks are also used for calibration when large numbers of samples are available. These procedures are used because they provide a satisfactory answer, and not because they directly address the sources of the problems. This paper presents a proposed explanation of the scatter effects and a procedure to correct for the effects to obtain a simple calibration for diffusely transmitting samples. The basic statement relating concentration, pathlength, absorptivity and transmittance of clear solutions is given by Beer’s Law.