Whisky, an analytical tale: the study of Scotch whisky and other spirits by advanced spectroscopic techniques

Abstract

The analysis of Scotch Whisky and other spirits is of great importance to the quality control, product protection and product development. The spirit industry uses various UKAS accredited analytical methods, such as those for chromatographic instru mentation that can be time consuming and require the use of multiple instruments.

In this thesis we explore Nuclear Magnetic Resonance (NMR) spectroscopy as a possible method for providing multiple characteristics of spirits in few experiments and explore their accuracy. We also analyse composition of ethanol wood extracts and compare them with compounds found in a concentrated whisky sample.

Quantitative NMR (qNMR) is a powerful analytic technique that is currently underutilised by the spirit industry. We have developed a method for quantification of 13 congeners applicable to high ethanol samples. The method uses a custom build library of individual compounds within the proprietary software program Chenomx. We were able to be confidently quantified these compounds using a single 1D ¹H NMR spectrum in 86 whisky sample (500 µl, 15 minutes per samples) within 6.4 % (standard deviation ± 5.0 % ) of their nominal concentration. We have conducted a chemometric analysis using NMR determined concentration of congeners (targeted analysis) achieving separation between single malt Scotch Whisky and Blended Scotch. Similar separation was realised through non-targeted analysis of 1D ¹H NMR spectra of the studied samples. Models were built using un-supervised (Principal Component Analysis) and supervised (Partial Least Squares - Discriminant Analysis) methods, which can be used as a method for classification of whisky samples and the identification of counterfeits.

NMR spectroscopy was also used to study ethanol/water mixtures aiming to develop a methodology for the determination of the ethanol strength. Several NMR parameters, such as ¹³C and ¹H chemical shift differences between the CH₂ and CH₃ resonances of ethanol (∆ δ¹³C and ∆ δ¹H), one-bond and long-range proton-carbon coupling constants of ethanol were examined as a function of ethanol concentration. Regions of ethanol concentrations that produced a step change in these parameters agreed approximately with the literature data obtained by other methods. Several parameters were taken forward and tested on real samples of spirit. It was found that samples containing high levels of sugar directly influence these NMR parameters and further investigation will be needed to develop this method for use in spirits with (low) carbohydrate content. Due to the magnitude of the changes, the ∆ δ¹³C parameter (2.1 Hz/ 1 % ABV) is most promising for further exploration. Using an external standard, a quantitative NMR method utilising the absolute integral of the CH₂ and CH₃ peaks obtained in a 1 scan ¹H decoupled ¹³C spectrum of neat spirits was developed yielding accuracy of 1 % over the full range of analysed samples. The use of ¹H NMR spectroscopy using benchtop instruments was also explored and the ethanol strength was determined with an accuracy of ±X% using integrals of ¹H resonances of ethanol.

Using an external standard, qNMR was explored for the determination of the site specific isotopic composition of ethanol. Both δ ¹³C and D/H ratios were investigated as these parameters are known to depend on the biosynthetic pathways of plant carbohydrates that are fermented to produce ethanol. A qNMR method using neat samples was developed that can determine site-specific D/H ratios to about 1 % accuracy, which is sufficient to link the origin of ethanol to C₃ or C₄/CAM biosynthetic pathways of plants. Chemometric analysis of the data was performed on range of samples containing different botanical origins, yielding a clear separation between C₃ and C₄/ CAM derived spirits. This method failed for δ ¹³C due to low accuracy of the integral intensities obtained from the ¹³C spectra.

Wood extracts of American oak wood showings were partially separated by Solid Phase Extraction and concentrated fractions were studied using high resolution 2D NMR methods on 800 MHz NMR. Eighteen compounds were identified, including eleven carbohydrates, three cyclitols, 2-phenolethanol and a larger compound (8-(4- Hydroxy-3,5-dimethoxyphenyl)-6,7-bis(hydroxymethyl)-1,3-dimethoxy-5,6,7, 8-tetrahydro-2-naphthalenol) believed to be a degradation of lignin. In addition, two ethyl glycosides were identified. These however could be an artefact of samples preparption as their existence in whisky samples was not reported before.