Gianluca Martelloni — Università di Firenze #
Molecular dynamics simulation of landslide with two infiltration time scales due to micro and macro pore soil structure #
In this work we integrate existing soil infiltration modeling with particle based methods in order to 
simulate landslides triggered by rainfall. In literature, usually, the infiltration models are based on 
continuum schemes (e.g. Eulerian approach) by means of which it is possible to define the field of the 
pore pressure within a soil. Differently, the particle based method implements a Lagrangian scheme which 
allows to follow the trajectory of the particles and their dynamical properties. In order to simulate the 
triggering mechanism, we test the classical and fractional Richards equations adapted to the molecular 
dynamics approach using the failure criterion of Mohr-Coulomb. In our scheme the local positive pure 
pressures are simply interpreted as a perturbation of the rest state of each grain, i.e., the pore 
pressure function can be interpreted as a time-space dependent scalar field acting on the particles. To 
initialize the system we generate, using a molecular dynamics based algorithm, a mechanically stable 
sphere packing simulating a consolidate soil. In this way we obtain the input structure of our 
"fictitious" soil to model landslides, considering the infiltration processes caused by rainfall. 
Moreover, in our scheme, the particles are porous and therefore we take into consideration the micro pore 
structure at intra-particle level, while the macro pore structure is due to inter-particle interstices. In 
this way we have two different infiltration time scales, as observed experimentally. The inter-particle 
interactions are modeled through a force which, in the absence of suitable experimental data and due to 
the arbitrariness of the grain dimension, is derived from a Lennard-Jones like potential. For the 
prediction of the particle positions, after and during a rainfall, we use a standard molecular dynamics 
approach. We analyze the sensitivity of the models by varying some parameters (hydraulic conductivity, 
cohesion, slope and friction angle, soil depth, variation of random properties, fractional order of the 
generalized infiltration model, etc.) and considering both regular and random configuration of the 
particles. The outcome of the simulations is quite satisfactory and therefore, we can claim that this is a 
promising new method to simulate landslides triggered by rainfall.