Separation kinetics and phase behaviour of volatile flavour active compounds in aqueous food streams

Volatile organic compounds (VOC) present at low concentrations may have a tremendous effect on the flavour profile of e.g. fermented beverages such as wine and beer. Controlling the flavour profile requires a detailed understanding of the physiochemical interactions of these components with the food matrices in which they occur. This thesis tries to answer the question of whether it   is possible to separate any of the major volatile flavours in lager beers, especially isoamyl acetate. The proposed technique to do so had to meet a series of criteria. It had to be mild to protect biological ingredients such as proteins from being damaged, be able to handle high throughput (~100 to 500 m3/hr), satisfy food-grade conditions and must be compatible with current flavour stripping processes. To design such a process, more fundamental knowledge was needed,  especially regarding the thermodynamics of solutions and kinetics of separation. These two topics are therefore addressed throughout the thesis.

In chapter II we start with a critical review of the current knowledge on flavour-matrix interactions in aqueous systems. We found that the fundamental data needed to design flavour separation processes for beer have neither been investigated experimentally nor through predictive modelling. The main focus in most papers was on the sensory aspects of flavour retention and release and directly linked to food design, and not to flavour design. Along with providing the first results from our work, we discussed and challenged the experimental techniques and data interpretation methods that are currently common practice, and recommended techniques to avoid pitfalls.

Based on our observations in chapter 2, we quantified Henry’s Law Constant (HLC) in chapter   III for the three major flavour compounds in beer (ethyl acetate, isoamyl acetate and isoamyl alcohol) in a model system containing ethanol concentrations below 25 %(v/v). We used the   static headspace analysis method and showed that a major effect on flavour retention takes place  at ethanol concentrations between 15 and 20 %(v/v) and higher. We also modelled this   behaviour using Henry coefficients for aqueous binary, and ternary systems using the Wohl expansion for excess Gibbs free energy coupled with the one-parameter Margules equation. The first approach was not complex enough to cover the behaviour of the components, but based on the second model, we could. Wohl’s expansion parameter for ethanol-water can be interpreted as solvent-solvent interaction. Furthermore, we quantified HLCs based on Van ’t Hoff parameters between 30 to 60˚C.

In chapter IV, we used a stripping column with structural packing to observe the effect of beer  dry matter on flavour behaviour. We observed that the major components are carbohydrates and small proteins that in general enhance migration of esters from the aqueous body of the beer. However, there was a slight retention of isoamyl alcohol that was due to changes in mass transfer resistance. The effect of gas flow rate on the partition coefficient of the compounds was minor,  but it almost doubled the mass transfer coefficient of volatile flavour compounds. For isoamyl alcohol, the mass transfer resistance was found to be in both phases whereas for ester groups and ethanol the major resistance is in the gas phase which explains the difference in behaviour.

In chapter V we examined the possibility of using frictional diffusion (FricDiff) for separation of volatile compounds from a model gas mixture as it may exit a stripping column operating with carbon dioxide. We showed that a flat sheet FricDiff module can separate isoamyl acetate from  the gas feed and that the presence of water or ethanol on the sweep side creates additional   friction on flavour compounds and thus enhances separation but not enough to retain them completely. This may be further improved by adding flavours to the sweep gas, or alternatively,  the thickness of the membrane and the pore size would need to be adjusted, which in turn would theoretically also improve the selectivity of the process in absence of wall friction.

In the general discussion (chapter VI), we bring all the findings presented earlier together and wrap up with recommendations for process design.