Estimation of effective sample size when analysing powders with diffuse reflectance near-infrared spectrometry

Diffuse reflectance near-infrared (NIR) spectrometry was used to analyse powder samples where the samples were presented to the incident light in layers of controlled powder thickness (0.5–4.0 mm). Measurements were carried out both with a laboratory NIR instrument using a standard sample cup and process instruments equipped with fibre-optic probes. The examined powders were polystyrene microspheres, microcrystalline cellulose, film-coated pellets and non-pareil pellets of different sieve fractions. It was found that the change of the sample absorbance signal (log 1/R) with increasing powder thickness can be described by an exponential function. A method is presented for estimation of the amount of sample contributing to the measured reflected light. This amount is here defined as the effective sample size. The method is based on a comparison of a fitted exponential function with the repeatability of the NIR measurements. For the analysed powders the sample size, expressed as the sample weight per unit area, was estimated for each measured wavelength in the visible-NIR region 400–2500 nm. The results show that the sample size depends strongly on the wavelength in the 400–1400 nm region and that above 1600 nm the depth penetrated by the radiation, yielding useful NIR information, is less than 0.75 mm for all examined powders. The estimated sample weight per unit area also depends on the particle size of the powder. Coarse particles showed larger sample sizes than finer particles. Implications of the results for on-line monitoring of blend homogeneity are briefly discussed.