Effect of rewiring for a sandpile model on a directed network

Abstract

Several studies have considered sandpile dynamics not over regular grids, but over networks. In this case, avalanches redistribute grains not between neighboring sites in a geometrical sense, but between connected sites, in a topological sense. However, depending on how nodes are connected, grains may never leave the system, preventing energy release. In this work, we study the simplest case, the BTW model in one and two dimensions, rewiring the nodes so that at every rewiring step, the energy release is always possible, and study avalanche statistics as a function of rewiring. In the 1D case, a transition is observed in the Gini coefficient of the load distribution per node at about 85% the number of possible rewirings, a transition which is not evident with other measures, such as the size distribution of avalanches or the mean distance between nodes in the network. In the 2D case, energy release follows a power law even when the grid is fully rewired, while the Gini coefficient, unlike the 1D case, decreases at a steady rate, with a smoother transition. The effect of network size N is studied, finding that there is a transition for the Gini coefficient at the thermodynamic limit N \to \infty for both the 1D and 2D cases, transition which is also observed in the betweenness centrality, but not in other topological measures. Finally, the dependence of the results with the load per rewiring iteration, and the avalanche threshold is studied.