Effects of fractional wettability on capillary pressure–saturation–relative permeability relations of two-fluid systems

Capillary pressure (P_c)–saturation (S)–relative permeability (k_r) relationships must be quantified to accurately predict non-aqueous phase liquid (NAPL) distribution in the subsurface. Several experimental techniques are presented here for two-fluid P_c–S–k_r relationships for various saturation paths to better define the effect of fractional wettability on these relationships. During the primary drainage path of the P_c–S curves, the air–water system showed no distinct trend as a function of the fraction of sand treated by organosilane (S) to render it non-water wetting. In a NAPL–water system, however, a consistent decrease of capillary pressure with increase of the fraction of non-water wetting sands was observed. The much lower contact angle for air–water (a–w) system may result in the observed insensitivity of the a–w P_c–S curves to fractional wettability, at least for the PD pathway. For the main imbibition path of NAPL–water system, capillary pressure decreased as the fraction of the S component increased, requiring forced imbibition (negative capillary pressures) for a certain range of saturations. Systems with an increasing percentage of the S component also exhibited a higher water k_r and lower NAPL or air k_r at a given saturation for the primary drainage and main imbibition paths in both air–water and NAPL–water systems. The increase of water k_r with increase of the fraction of the S component can be explained by the ability of water to occupy larger and highly conductive pores in such a system. Experimental k_r–S data for the primary drainage path of NAPL–water system presented here were used to test the Bradford et al. Bradford SA, Abriola LM, Leij FJ. Wettability effects on two- and three-fluid relative permeabilities. J Contam Hydrol 1997;28:171–91] model and the modified Mualem model for estimating the k_r–S curves from measured P_c–S data as a function of fractional wettability. Both models predicted significantly less variation in the k_r–S curves than measured indicating that they did not adequately represent the system under investigation.