Impact of phases distribution on mixing and reactions in unsaturated porous media

The impact of phases distribution on mixing and reaction is hardly assessable experimentally. We use a multiple point statistical method, which belongs to the family of machine learning algorithms, to generate simulations of phases distributions from data out of laboratory experiments. The simulations honour the saturation of the laboratory experiments, resemble the statistical distributions of several geometric descriptors and respect the physics imposed by capillary forces. The simulated phases distributions are used to compute solute transport. The breakthrough curves reveal that different phases distributions lead to broad ranges of early arrival times and long-term tailings as saturation decreases. For a given saturation, a similar long-term scaling of mixing area, interface length, and corresponding reactivity is observed regardless of phases distribution. However, phases distribution has a clear impact on the final values (before breakthrough) of area of mixing, interface length and mass of reaction product.