A Coupled THM Model for Predicting Frost Heave in Unsaturated Fine-Grained Soil under Step-Cooling Conditions

An accurate simulation of frost heave in unsaturated fine-grained soils is critical for ensuring the long-term performance of infrastructure in cold regions. In response to this need, a coupled thermo-hydro-mechanical (THM) numerical model was developed in this study to characterize freeze-induced deformation of subgrade soils under unidirectional freezing. Specifically, the model incorporated void ratio and matric suction as key variables and integrated the van Genuchten (VG) model to represent unsaturated permeability and water retention behavior based on measured soil-water characteristic curves. Notably, unlike previous studies, this model introduced an innovative three-phase (liquid water–vapor–ice) coupling framework for moisture migration. In this framework, vapor transport was simulated using a pore-scale corrected diffusion coefficient equation that considered the blocking effect of ice crystals on vapor diffusion paths during freezing and was linked to matric suction and temperature gradients. This enables quantitative characterization of vapor contribution to total moisture flux, a rarely addressed aspect in existing models. Furthermore, to improve engineering relevance, a step-cooling scheme that could be more representative of in situ thermal conditions was adopted for laboratory validation. Experimental data from soil samples at the optimum water content (14%) were used to verify model accuracy. Subsequent simulations explored varying initial water contents (10%, 12%, 16%, and 18%) under open-system step cooling. The results indicated that water migrated from unfrozen to frozen zones, accumulating near the freezing front. A higher initial water content led to faster cooling, a more pronounced frost heave, and greater ice lens formation. Additionally, the void ratio in frozen regions was consistently higher than that in the unfrozen zones and shifted toward the warm boundary as the water content increased. Overall, the proposed framework offers a robust tool for predicting freeze-induced volumetric changes in unsaturated fine-grained soils and establishes a foundation for future multifield coupling models to enhance infrastructure resilience in seasonally frozen regions.