Force rearrangements and statistics of hyperstatic granular force networks

TL;DR

This paper investigates the statistical properties of hyperstatic force networks in granular media, revealing how force rearrangements and boundary effects influence macroscopic stress fluctuations through a maximum entropy framework.

Contribution

It introduces a detailed analysis of force rearrangements in hyperstatic granular networks, linking them to floppy modes and developing a maximum entropy model to predict microscopic statistics.

Findings

Rearrangements govern stress fluctuations and scale with distance to isostaticity.

Boundary conditions significantly alter force distribution statistics.

Maximum entropy approach accurately captures microscopic force statistics.

Abstract

The heterogeneous force networks in static granular media --- formed from contact forces between grains and spanning from boundary to boundary in the packing --- are distinguished from other network structures in that they must satisfy constraints of mechanical equilibrium on every vertex/grain. Here we study the statistics of ensembles of hyperstatic frictionless force networks, which are composed of more forces than can be determined uniquely from force balance. Hyperstatic force networks possess degrees of freedom that rearrange one balanced network into another. We construct these rearrangements, count them, identify their elementary building blocks, and show that in two dimensions they are related via duality to so-called floppy modes, which play an important role in many other aspects of granular physics. We demonstrate that the number of rearrangements governs the macroscopic statistical properties of the ensemble, in particular the macroscopic fluctuations of stress, which scale with distance to the isostatic point. We then show that a maximum entropy postulate allows one to quantitatively capture many features of the microscopic statistics. Boundaries are shown to influence the statistics strongly: the probability distribution of large forces can have a qualitatively different form on the boundary and in the bulk. Finally, we consider the role of spatial correlations and dimension. All predictions are tested against highly accurate numerical simulations of the ensemble, performed using umbrella sampling.