Transmission Raman measurement directly through packed corn kernels to improve sample representation and accuracy of compositional analysis.

The potential of transmission Raman spectroscopy for direct analysis of packed granular samples, one of the most frequently encountered sample types in the field of non-destructive spectroscopic analysis, has been evaluated. For this purpose, transmission Raman spectra were collected by laser illumination through packed corn kernels to determine their protein concentration. Back-scattering Raman spectra of the same samples were also collected for comparison. Raman spectral features of the major kernel constituents were initially characterized, and Raman mapping over the whole kernel face was performed to investigate the inhomogeneous distribution of constituents in a kernel. Possible variations of transmission spectral features depending on the laser illumination on different locations of a kernel were investigated, since the orientation of kernels in the packing was essentially random. Rotation of kernel packing during spectral collection was helpful in improving the compositional representation of packed kernels. With partial least squares (PLS) regression, the protein concentrations were determined using both spectral collection methods and the resulting accuracies were compared. As a result, the transmission measurement provided a more accurate determination of protein concentration since it enabled deeper sampling across the packed kernels, leading to a better compositional representation of them. By contrast, in the back-scattering measurement, kernels on the top of the packing were mainly sampled for the spectral acquisition. Moreover, the back-scattering spectral feature, more weighted to constituents localized at the outer portion of a kernel, was short of representing the overall composition of a kernel.