An empirical model to predict the headspace concentration of volatile compounds above solutions containing sucrose

An empirical model was developed to describe and predict the change in gas–liquid partition behaviour of a wide range of volatile compounds in aqueous sucrose solutions. The static equilibrium headspace concentrations of 40 volatiles (from different chemical classes e.g. pyrazines, alcohols, esters and ketones and with different physical properties e.g. volatility and solubility), were measured above aqueous sucrose solutions [0–65% (w/v)]. As sugar concentration increased, the headspace concentration of some compounds increased, whilst others stayed the same or decreased. The changes in volatile headspace concentrations were correlated (using partial least squares regression) with a range of physicochemical descriptors which were calculated from the compound structure. The model was composed of seven physicochemical descriptors and had a regression coefficient, R2=0.74. An external test set was used to validate the model. The key descriptors in the model were (log P)2, LUMO energy and a connectivity index term.