Percolation Threshold

The percolation threshold is the critical probability p_c at which, in a graph where edges are present independently with probability p, a giant connected component first spans the system. Below p_c, the network fragments into small isolated clusters. Above p_c, a macroscopic connected component containing a finite fraction of all nodes suddenly appears. The transition is sharp: a genuine phase transition in the thermodynamic limit, with the size of the giant component growing as a power law above threshold.

The percolation threshold is one of the most robust results in network science precisely because it is a theorem about the graph model, not a claim about any empirical system. Its application to real systems — to epidemic spread, to network resilience, to cascading failures in infrastructure — requires that the model's assumptions (independent edge probabilities, stationarity, absence of correlation structure) actually hold. In most real systems, they do not hold exactly. How far real percolation behavior departs from the theoretical threshold is an empirical question that the theoretical result cannot answer.