A new technique that uses a chemical sensor coupled with multivariate analysis is evaluated and applied to discriminate soft drinks from different vendors. The headspace volatiles of different cola samples were analyzed using a chemical sensor that uses quadrupole mass-spectrometer technology. Analysis times per sample averaged between 3 to 4 minutes.

Soft drink samples were used to train the chemical sensor with acceptable mass spectra patterns. Small variations in the mass spectra profiles of the samples were detected using principal component analysis (PCA). In this application, PCA searches for correlations among all m/z abundances simultaneously and extracts linear combinations of highly correlated m/z abundances (principal components, or PCs) that describe the greatest amount of sample variability.

Mathematical proximity of the soft drinks’ mass spectra projections into the 3 dimension PC plot translates into chemical similarity since over 90% of the total variation was captured within the first 3 PCs. In this case, samples close to each other in the principal component plot resemble each other chemically. PCs of soft drinks from the same vendors clustered together in the three dimensional PC plot validating that this chemical sensor along with multivariate analysis is capable of differentiating mass spectra patterns for the soft drink samples.