Discrimination of white wines by Raman spectroscopy coupled with chemometric methods

Abstract

France is the largest exporter of wine in the world. The export turnover is estimated at 8.7 billion euros in 2017 for 13 million hectoliters sold. This lucrative business pushes scammers to increase the value of some low-end wines by cheating on their appellations, quality or even their origins. These facts lead to losing 1.3 billion euros each year to the European Union’s wine and spirits companies. 

The control of wine quality is performed by analytical methods such as infrared, NMR or HPLC. Nevertheless, the presence of water and ethanol interferes with the determination of the other wine molecules. In addition, the complexity of the wine matrix and the chemical similarity between its main compounds complicate the extraction of information obtained by these analytical methods. Consequently, the need to develop more sensitive, fast and automated procedures remains a real need for investors and stakeholders in this area. Our study aims to evaluate the ability of Raman spectroscopy to discriminate wines depending on their origin and grape variety based on their spectral fingerprints. Wines from 8 grapes varieties have been studied: Chardonnay (Bourgogne), Riesling (Alsace), Gewurztraminer (Romania), Muscadet (Val de Loire), Sauvignon blanc (Bordeaux), Muscat (Pays d’Oc) and a blend with Semillon (Bergerac). The results showed that white wine has a rich spectral signature (excitation at 532 nm) which reflected its molecular composition. The application of statistical tests (Kruskal-Wallis) made it possible to classify 6 different groups thus confirming that the spectra of the analyzed wines are different. Principal component analysis and discriminant analysis showed a perfect discrimination between the different wines. The validation of the database with another wine that is not part of the model (Sauvignon blanc, Val de Loire) showed a very good discrimination between the different wines. Nevertheless, confusion was observed between the two Sauvignon because the model could not differentiate them despite their different origins. 

Raman spectroscopy allows the rapid identification of the grape variety. Nevertheless, a large number of samples must be analyzed in order to evaluate the industrial viability of this technique (variability between years, batches) and validate the approach on a large panel of wine belonging to grape varieties and different geographical areas.