Espectroscopia ótica para discriminação de misturas de café arábica e robusta

Abstract

Coffee is one of the most important food products in the world economy. Brazil is the largest producer and exporter of the grain and the second largest consumer of coffee in the world. Most commercially available coffee is arabica coffee (Coffea arabica L.) and robusta/conilon coffee (Coffea canephora L.) type species as well their blends. The arabica coffee has higher commercial value and its general sensorial characteristics are more favorable when compared to robusta coffee one. This work aimed to differentiate coffee species after roasting by optical spectroscopy. Thus, blends with different ratios of roasted and ground arabica/robusta coffees in three grades of roasting (light, medium and dark) were studied using the Fourier-Transform Infrared Spectroscopy (FTIR), fluorescence and time resolved fluorescence (TCSPC) spectroscopy techniques. Notably similar in their chemical composition, two multivariate methods were used to analyze the sample spectra: principal component analysis (PCA) and partial least squares regression (PLS). From the spectra analyzed with PCA, it was possible to differentiate the coffee species and to identify compounds that are good discriminants among them, such as caffeine and lipids, present in larger quantities in the robusta and arabica coffee, respectively. In addition, by associating the FTIR spectra to the PLS, a model capable of quantitatively predict the proportion of each specie in the studied samples was developed, with a determination coefficient higher than 0.99. Time resolved fluorescence and fluorescence analyzes were performed using excitation/emission wavelengths at 280/480 nm, 310/470 nm and 400/605 nm. The emission spectra obtained by the technique discriminate the species and indicate the presence of fluorophores in only one of the coffees analyzed. As for the time-correlated single photon count (TCSPC), the mean intensity of the emission lifetime was calculated, in which the set 310/470 nm showed a difference between species of more than 20%. The results obtained in this work demonstrate the fact that the techniques employed are promising since they are able to quickly distinguish the species or even quantify the percentage of the blends.