We formulate theoretical approaches to understand the mechanisms leading to unstable deformations in saturated and unsaturated soils. We focus on both localized and diffuse failure modes, addressing both brittle and plastic geomaterials. For this purpose, we derive failure indices via the mathematical concepts of controllability and bifurcation. These methods are combined with constitutive models for fluid-infiltrated porous media, with the main purpose to predict and explain the onset of failure mechanisms nder laboratory conditions and/or at engineering scale.

We use geomechanical models and stability criteria to explain catastrophic slope failures under field conditions. Our analyses focus on the onset of rapid flowslides, i.e. slope failures involving unexpected solid-to-fluid transitions. In particular, we address both underwater and subaerial slope failures, starting from a mechanistic explanation of the unstable processes that promote the mobilization of shallow soil veneers. This type of research aims to distinguish whether a natural deposit is likely to undergo a soils slip of limited mobility or a chaotic runaway failure. The ultimate goal is to quantify landslide susceptibility and landscape dynamics across scales.

We use multi-scale methods to capture the effects of particle fragmentation in unsaturated granular assemblies. In particular, we focus on the interplay between particle phenomena and solid-fluid interactions in crushable media subjected to high pressures. This research is conducted by combining continuum thermodynamics, particle-scale models and discrete element analyses of unsaturated granular media.

We use thermodynamic models to capture the macroscopic implications of evolving microstructural attributes. For this purpose, we describe the effects of chemical reactions at particle/pore scale, by modeling the evolution of particle geometry and pore size characteristics during dissolution and precipitation, as well as their implications in terms of mechanical properties at the continuum level.