Raman spectroscopy for monitoring carotenoids in processed Bunchosia glandulifera pulps.

Raman spectroscopy is a rapid and non-destructive analytical technique that has found a growing interest in the characterization and quantification of microconstituents in foods. In this work, Raman spectroscopy was used to evaluate the carotenoids content of processed Bunchosia glandulifera, a native fruit from the Brazilian Atlantic Forest. The B. glandulifera pulps were dried in hot air at 65 and 85 °C, resulting in carotenoids degradation of about 75 and 80%, respectively. The degradation profile of carotenoids over time was analyzed by Raman spectroscopy, with and without the use of an internal standard (TiO2). The reproducibility of the analyses was evaluated by PCA of spectral data. PLS regression was applied for modelling the total carotenoids in B. glandulifera. Resulting models show the Raman data correlate with carotenoids content on samples resulting in a satisfactory coefficient of determination for all sets of samples.

Development of a quantitative approach using Raman spectroscopy for carotenoids determination in processed sweet potato.

The orange-fleshed sweet potato is a vegetable-rich in carotenoids. The thermic treatment for sweet potato processing can decrease the content of these constituents in the foods, lowering their bioactive properties. Raman spectroscopy has been growing as a fast tool to food analysis, especially for detection of low concentrations of carotenoids and to the monitoring of its degradation profile over time. Therefore, in this work were evaluated two methods of drying, hot air and microwaving with rotary drum, combined with quantitative Raman spectroscopy. The results showed carotenoids degradation around 50% for both types of drying processes studied. PCA plot proved the potential of reproducibility of analyses for microwave drying samples. For samples heated with hot air, the best linear correlation achieved was R2 = 0.90 and by microwave was R2 = 0.88. Also, partial least squares (PLS) regression models were constructed obtaining a satisfactory coefficient of determination.