A physically based model for the water retention curve of clay pastes

Available approaches for modeling the soil water retention curve are based on parameters to be fitted using observed data. Due to the formal fitting to the data the soil water retention curves of nonswelling and swelling soils are frequently considered on equal grounds, and the data of both soil types are used as examples in validation. The objective of this work is to construct and validate a new physical model of the water retention curve (the drying branch) of a swelling clay paste (as the simplest case of swelling soils) in the broad area of water content and suction values where the effect of adsorption against capillarity is negligible. The approach is based on a unitary description of pore-dimension, pore-shape, and pore-water distributions, using the case of swell–shrink clay pastes as an example. The general expression of suction is factorized. The first factor depends on the characteristic dimensions of water-containing pores and, indirectly, on shrinkage–swelling. The second depends on the soil-porosity variation at shrinkage–swelling. The case of clay paste allows a number of specific simplifications that lead to a final expression for matric suction. The consideration as a whole is based on a recently suggested model connecting the microstructure, pore size distribution, and shrinkage curve of a clay paste. The four parameters of the constructed retention curve have a clear physical meaning and can be measured independently of the model to be developed. These parameters include the density of the solid phase, the volume fraction of the solid phase at maximum water content (at the liquid limit), the minimum relative volume of the clay paste, and the maximum external dimension of 3D pores of the clay matrix at maximum water content. The physical model explains the qualitative differences to be observed between the retention curves of the swelling clay soil pastes and soils with a rigid matrix. For model validation data on the properties and the water retention curve of two clay pastes are compared to a data analysis using the suggested approach. The data and model prediction are in satisfactory agreement.