Microscopical differentiation of lignocellulosic fibers using reflectance near infrared spectroscopy and multivariate image analysis

Content and microscopical distribution of lignocellulosic components in wood could explain recalcitrance of samples for diverse biotechnological applications and suggest mechanisms to improve their quality for future processes in biofules and biomaterials production. In this work, lignocellulosic fibers obtained from Pinus radiata wood after solvent extraction pretreatment were evaluated at microscopic scale using near infrared spectroscopy (NIRS) and multivariate images analysis (MIA). Hyperspectral images were acquired with a Spotligth 400 FTIR-NIR & Imaging System (Perkin Elmer) in the range of 7800 – 4000 cm-1 with spectral resolution of 8cm-1 and 16 scans from areas of 200 x 200 mm using a pixel size of 6.25 mm in total reflectance mode. Acquired data organized in a hypercube were reshaped in a matrix containing pixel position and wavelengths and submitted to principal component analysis (PCA) and partial least squares (PLS) methods in order to evaluate the inter and intra-samples chemical variation, as well as their lignin content quantification, respectively. Analysis of sets of three and four samples with different content of lignin determined by the reference gravimetric and UV-spectroscopy method, were used for MIA. Results of spectra PCA show that scores of each sample have different dispersion and spectral homogeneity and are clustered into distinct groups among the samples. PCA of images showed differences in chemical distribution of lignocellulosic components among the samples, although identification of these components was difficult due to the near infrared bands nature and loadings obtained from derivative spectra. PLS models for lignin quantification in samples show standard errors of validation around 2.4 % in a range of 19.0 and 28.0 % of lignin in the samples. Although this values can be seen high, the average lignin content of all the pixels are very closed to the lignin content measured by the reference method, with relative errors between 1 and 5%.