Monitoramento por espectroscopia dos compostos fenólicos e furaldeídos gerados no processamento de biomassa lignocelulósica

Abstract

The major process challenge of the second-generation ethanol (2G) production is related with characterization of hydrolyzate from lignocellulosic biomass, which often contains high quantities of phenolic compounds and furan derivatives. These components of hydrolyzate are responsible for inhibit and deactivate enzymes during hydrolysis in addition to negatively influence the fermentation step. The phenolic compounds and furaldehydes quantification could help to highlight the bioprocesses limitations. As a result, it could allow the process improvement that may be characterized by more productive, robust and tolerant to these compounds. Concerning about this objective, rapid, efficient, and low-cost technologies for monitoring the phenolic compounds and furan derivatives are essential for better control of the pretreatment, hydrolysis and fermentation steps during 2G ethanol production. For achieving that goal it was verified the viability of monitoring phenolic compounds and furaldheydes by the use of chemometric techniques. The Ultraviolet Visible and Near Infrared spectral regions were analyzed in association with Partial Least Squares (PLS) regression for monitoring the inhibitors from pretreatment hydrolyzate of sugarcane bagasse. Hydrolyzate samples from liquid hot water pretreatment of biomass plus synthetic samples were evaluated on distinct calibration and test trials. The negative effect in both hydrolysis and fermentation process were considered for monitoring components from hydrolyzate and synthetic mixtures. Then, the concentration of vanillin, hydroxymethylfurfural, furfural, as well as ferulic, gallic and p-coumaric acids were analyzed. It was found that the most accurate PLS model could be used to monitor phenolic compounds and furaldehydes during the liquid hot water pretreatment of lignocellulosic material from three different operating conditions. The best predicting concentrations provided satisfactory accuracy for each analyte by presenting PLS-UV-Vis models with potential for process monitoring (standard deviation of prediction for cross-validation leave-one-out (RMSECV) around 3.0 to 9.0% and residual predictive deviation (RPD) was from 2.0 up to 5.0).