Network dismantling by physical damage

Abstract

We explore the robustness of complex networks against physical damage. We focus on spatially embedded network models and datasets where links are physical objects or physically transfer some quantity, which can be disrupted at any point along its trajectory. To simulate physical damage, we tile the networks with boxes of equal size and sequentially damage them. By introducing an intersection graph to keep track of the links passing through tiles, we systematically analyze the connectivity of the network and explore how the physical layout and the topology of the network jointly affect its percolation threshold. We show that random layouts make networks extremely vulnerable to physical damage, driven by the presence of very elongated links, and that higher-dimensional embeddings further increase their vulnerability. We compare this picture against targeted physical damages, showing that it accelerates network dismantling and yields non-trivial geometric patterns. Finally, we apply our framework to several empirical networks, from airline networks to vascular systems and the brain, showing qualitative agreement with the theoretical predictions.