A predictive model based on the Krichevsky-Ilinskaya equation (KIE) combined with Peng Robinson EOS and the modified Huron-Vidal mixing rules has been proposed to estimate hydrocarbon solubility in both the vapour and liquid phases. For solubility in the liquid phase, the proposed model estimates the partial molar volume from an equation of state with Huron-Vidal mixing rules whereby a modified UNIFAC model is used to predict the asymmetric activity coefficient. The hydrocarbon data considered for modeling were non-polar liquid hydrocarbons such as n-alkanes and aromatic benzene compounds. For the hydrocarbons studied, the proposed model satisfactorily predicted the supercritical carbon dioxide solubility in the vapour phase. It was comparable to that correlated using equations of state with fitted interaction parameter applied to both phases. However, the liquid phase were less satisfactorily predicted with an AAD ranging from 5 to 24%. Partial molar volume at infinite dilution was found to be a critical input parameter as changing this parameter improved the error deviation between 4.1 and 10%. (C) 2001 Elsevier Science B.V. All rights reserved.