1. Investigación

Variation in methylxanthines (theobromine and caffeine) and flavanols (catechin and epicatechin) was studied in a large set of cocoa powders (covering different origins, processing parameters and alkalization levels). The content of these compounds was established by high-performance liquid chromatography (HPLC), whose results showed that the alkalisation process lowered the content of all analytes, whose loss was more evident in flavanols. Therefore, the determination of these analytes in a huge set of samples allowed not only better knowledge of the concentration variability in natural commercial cocoas from different origins, but also the understanding of the effect that industrial alkalisation has on these contents. The feasibility of reflectance near-infrared spectroscopy (NIRS) combined with partial least square (PLS) to non-destructively predict these contents, was also evaluated. All the analytes were generally well predicted, with better predictions for methylxanthines (R2 P 0.882 for both analytes; RMSEP 0.020-0.061%, bias -0.027-0.006) than for flavanols (R2 P 0.818-0.863; RMSEP 6.63-15.87%, bias 1.942-3.056). Thus NIRS could be an alternative fast reliable method for the routine assessment of these analytes in the cocoa industry.

Cocoa shell must be removed from the cocoa bean before or after the roasting process. In the case of a low efficient peeling process or the intentional addition of cocoa shell to cocoa products (i.e. cocoa powders) to increase the economic benefit, quality of the final product could be unpleasantly affected. In this scenario, the Codex Alimentarius on cocoa and chocolate has established that cocoa cake must not contain more than 5% of cocoa shell and germ (based on fat-free dry matter). Traditional analysis of cocoa shell is very laborious. Thus, the aim of this work is to develop a methodology based on near infrared (NIR) spectroscopy and multivariate analysis for the fast detection of cocoa shell in cocoa powders. For this aim, binary mixtures of cocoa powder and cocoa shell containing increasing proportions of cocoa shell (up to ca. 40% w/w based on fat-free dried matter) have been prepared. After acquiring NIR spectra (1100-2500 nm) of pure samples (cocoa powder and cocoa shell) and mixtures, qualitative and quantitative analysis were done. The qualitative analysis was performed by using principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA), finding that the model was able to correctly classify all samples containing less than 5% of cocoa shell. The quantitative analysis was performed by using a partial least squares (PLS) regression. The best PLS model was the one constructed using extended multiple signal correction plus orthogonal signal correction pre-treatment using the 6 main wavelengths selected according to the Variable Importance in Projection (VIP) scores. Determination coefficient of prediction and root mean square error of prediction values of 0.967 and 2.43, respectively, confirmed the goodness of the model. According to these results it is possible to conclude that NIR technology in combination with multivariate analysis is a good and fast tool to determine if a cocoa powder contains a cocoa shell content out of Codex Alimentarius specifications.

Cocoa powder is a global product of great value that can be adulterated with low-cost raw materials such as carob flour without changing the characteristics of color, aroma and flavor of the product. The use of rapid methods, as a NIR technology combined with multivariate analysis, is of interest for this detection. In this work, 216 adulterated samples prepared by blending commercial cocoa powders with different alkalization levels (n = 12) with commercial carob flour (n = 6) in different proportions (0-60% of adulteration) were analyzed. The diffuse reflectance spectra of the samples were acquired from 1100 to 2500 nm using a Foss NIR spectrophotometer. A qualitative and quantitative analysis was done. For the qualitative analysis, a principal component analysis (PCA) and a partial least squares discriminant analysis (PLS-DA) was performed. The coefficient of determination (R2) of the model PLS-DA was 0.969 and the coefficient of determination of the validation (R2CV), based on a full cross validation was 0.901 indicating good calibration with good predictability. These results indicate that it is possible to distinguish between pure cocoa powders from the adulterated samples. For the quantitative analysis a partial least squares (PLS) regression analysis was performed. The most robust model of PLS prediction was obtained with 1 factors (LV) at coefficient of determination (R2) of 0.980 and a root mean square error of prediction (RMSEp) of 3.237 % for the external validation set. These data lead to the conclusion that NIR technology combined with multivariate analysis allows the identification and determination of the amount of natural cocoa powder present in a mixture adulterated with carob flour.