Infiltration effects on a two-dimensional molecular dynamics model of landslides
In this paper we propose a two-dimensional (2D) computational model, based on a molecular dynamics (MD) approach, for deep landslides triggered by rainfall. Our model is based on interacting particles or grains and describes the behavior of a fictitious granular material along a slope consisting of a vertical section, i.e. with a wide thickness. The triggering of the landslide is caused by the passing of two conditions: a threshold speed and a condition on the static friction of the particles, the latter based on the Mohr-Coulomb failure criterion (Coulomb 1776; Mohr 1914). The inter-particle interactions are through a potential that, in the absence of suitable experimental data and due to the arbitrariness of the grain dimension is modeled by means of a potential similar to the Lennard-Jones one (Lennard-Jones 1924), i.e., with an attractive and a repulsive part. For the updating of the particle positions we use a MD method which results to be very suitable to simulate this type of systems (Herrmann and Luding 1998). In order to take into account the increasing of the pore pressure due to the rainfall, a filtration model is considered. Finally we also introduce in the model the viscosity as a term in the dynamic equations of motion. The outcome of simulations, from the point of view of statistical and dynamic characterization, is quite satisfactory relative to real landslides behavior and we can claim that this types of modeling can represent a new method to simulate landslides triggered by rainfall.
💡 Research Summary
The paper presents a novel two‑dimensional (2‑D) computational framework for simulating deep landslides triggered by rainfall, using a molecular‑dynamics (MD) approach. The authors model the slope as a collection of interacting particles (grains) that represent a fictitious granular material with a wide vertical thickness. Inter‑particle forces are described by a Lennard‑Jones‑like potential, which provides a short‑range repulsive component and a longer‑range attractive component. This choice compensates for the lack of precise experimental data on grain‑scale interactions and allows the model to capture both collision and cohesion effects that are essential for reproducing shear failure and flow.
Particle motion follows Newton’s second law and is integrated with the Verlet algorithm, a standard technique in MD that ensures energy‑conserving time stepping. To emulate the viscous dissipation observed in real soils, a linear damping term (viscosity) is added to the equations of motion. This term reduces kinetic energy, suppresses unrealistic oscillations, and reproduces the quasi‑elastic‑plastic behavior of saturated earth masses.
Triggering of the landslide is governed by two simultaneous criteria. First, a particle must exceed a prescribed threshold velocity (v_thr), representing the dynamic aspect of failure. Second, the static friction coefficient μ, which controls the Coulomb shear resistance, is reduced according to the Mohr‑Coulomb failure criterion as pore‑water pressure builds up. The pore pressure evolution is modeled by a simple one‑dimensional infiltration (Darcy) equation, which raises the effective stress σ′ = σ – p over time. When both the velocity condition and the reduced friction condition are satisfied, the particle loses its static equilibrium and begins to slide, initiating a shear band that propagates downslope.
The simulation setup consists of thousands of particles arranged on a 2‑D grid that represents a vertical cross‑section of a slope. Initially, particles are at rest under gravity and inter‑particle forces, forming a stable configuration. Rainfall is introduced by activating the infiltration model; pore pressure rises uniformly, gradually lowering μ. Once the critical velocity is reached, a rapid transition from static to dynamic behavior occurs.
Results show a clear sequence of events that mirrors observed landslide dynamics. (1) Shear failure nucleates at the slope surface and expands downward, forming a distinct failure plane. (2) Particle speed distributions evolve from near‑Gaussian (pre‑failure) to a Rayleigh‑like shape (post‑failure), indicating the emergence of high‑speed flows. (3) Total potential energy drops sharply while kinetic energy spikes, but the viscous term dissipates a portion of this energy as heat, reproducing the energy balance of real landslides. (4) The geometry of the failure zone exhibits fractal characteristics; its fractal dimension varies with infiltration intensity and friction reduction rate.
Statistical analyses quantify the propagation speed of the failure front, the distribution of particle accelerations, and the rate of energy loss. The front speed scales linearly with rainfall intensity and infiltration coefficient, confirming that the model captures the coupling between hydraulic loading and mechanical response. Sensitivity tests reveal that larger reductions in μ cause earlier failure and a broader spread of the sliding mass.
Comparison with field observations—such as measured run‑out distances, velocity time histories, and post‑event deposit patterns—demonstrates good agreement, suggesting that the MD framework can faithfully reproduce the essential physics of rain‑induced landslides.
The authors argue that the particle‑based nature of the model offers distinct advantages over continuum approaches. By adjusting grain‑scale parameters (size, stiffness, friction, cohesion) the model can represent a wide range of soil types and conditions without requiring extensive calibration data. This flexibility makes the method suitable for scenario analysis, risk assessment, and the design of mitigation strategies in data‑scarce environments.
In summary, the study introduces a robust, physics‑based MD model that integrates hydraulic infiltration, Mohr‑Coulomb failure, velocity thresholds, and viscous damping to simulate rainfall‑triggered landslides. The simulation outcomes align well with empirical data, establishing the approach as a promising tool for advancing our understanding of landslide initiation and for supporting engineering practice in slope stability assessment.