A stoichio-kinetic model for a DPPH∙ -ferulic acid reaction.

Estimates of the activity of antioxidants in the literature often appear inconsistent. In the specific case of the DPPH∙ test, the diversity of measurements may arise from variations in the protocols followed. This paper proposes an unbiased method which models the reduction mechanism. This method is applied to the reduction of DPPH∙ by ferulic acid. A scheme with eight coupled reactions is proposed and has been validated on different solvents and using a wide range of DPPH ̇, ferulic acid, and 5,5'-diferulic acid concentrations, and verified using data from the literature on ferulic acid activity. This modeling approach permits a correction to the bias of the 8th reaction (spontaneous reduction of DPPH ̇), because of its sensitivity to solvent, which in most cases is not taken into account. The best experimental strategy to determine the Efficient Concentration of ferulic acid to reduce 20% (EC20) and 50% (EC50) of DPPH∙ is then detailed in terms of initial DPPH∙ concentrations and the duration of the experiment.

Modeling the formation of some polycyclic aromatic hydrocarbons during the roasting of Arabica coffee samples.

Roasting is a critical process in coffee production, as it enables the development of flavor and aroma. At the same time, roasting may lead to the formation of nondesirable compounds, such as polycyclic aromatic hydrocarbons (PAHs). In this study, Arabica green coffee beans from Cuba were roasted under controlled conditions to monitor PAH formation during the roasting process. Roasting was performed in a pilot-spouted bed roaster, with the inlet air temperature varying from 180 to 260 degrees C, for roasting conditions ranging from 5 to 20 min. Several PAHs were determined in both roasted coffee samples and green coffee samples. Different models were tested, with more or less assumptions on the chemical phenomena, with a view to predict the system global behavior. Two kinds of models were used and compared: kinetic models (based on Arrhenius law) and statistical models (neural networks). The numbers of parameters to adjust differed for the tested models, varying from three to nine for the kinetic models and from five to 13 for the neural networks. Interesting results are presented, with satisfactory correlations between experimental and predicted concentrations for some PAHs, such as pyrene, benz[a]anthracene, chrysene, and anthracene.