Near-infrared reflection spectroscopy and partial least squares regression for determining the total carbon coverage of silica packings for liquid chromatography

Near-infrared reflection spectroscopy (NIRS) was used in combination with principal component analysis (PCA) and partial least squares (PLS) regression to determine the silica packing properties, whether endcapped (EC) or non-endcapped, and the amount of surface total carbon coverage (%). A preparation technique for the reproducible analysis of bulk materials is covered as well as spectral data pretreatments to enhance prediction accuracy of the PLS models. Especially derivatives and scatter correction methods turned out to be well suited. A standard error of prediction (SEP) of 0.57% C for the determination of the total carbon load of octadecylated (C₁₈) silica was found. In case of the C₁₈-EC silica a SEP of 0.51% carbon gives rise to a robust and accurate model. The presented method allows the simultaneous determination of several parameters, e.g. particle and pore size or surface modifications, from a single spectrum and is amenable to implementation for in-line or on-line analysis in the silica producing industry.