| Several methods whose sum outlines a protocol for non-invasive wine fingerprinting is presented herein. First a low frequency, nu < 25 MHz, spectrometer capable of non-invasively and non-destructively screening the dielectric properties of full intact bottles of wine is described. A comprehensive review of dielectric properties and principal component analysis is also discussed. Then a separate nu < 30 MHz low frequency spectrometer capable of non-invasively and non-destructively screening the diamagnetic properties of full intact wine bottles is similarly described. Throughout, it will be demonstrated that wines inherent constituent variability resultant from its distinct enological and viticultural processes give rise to a bottles unique susceptibility fingerprint. The sensitivity of both these methods to such ionic and molecular constituents is established by analyzing standard solutions. In both methods principal component analysis is implemented as a means of dimensional reduction; a plot of whose first two principal components indicate wines of like origin and vintage empirically cluster in similar regions. Finally a bivariate correlated Student distribution is developed as a model to statistically quantify cluster attribution in the principle component basis. As a case study a larger, more statistically meaningful, 36 bottle estate wine library is screened for its diamagnetic properties. The library is constructed from a single bottle type obtained from four different sources; the estates onsite cellar, a private collector's garage and two distributers in Salinas and Monterey. Application of principal component analysis followed by analysis with the bivariate distribution permit the effect of wine transport and storage to be explored. The success of the application of these spectrometers coupled with the use of bivariate statistics suggests that the method may be used to non-invasively screen counterfeit wine. |
