In classical network theory, global connectivity emerges through a smooth, gradual process. As links are added to a random network, small clusters slowly merge until a single "giant component" smoothly comes to dominate the system. However, many real-world networks—and modern theoretical models—exhibit a far more abrupt phenomenon known as explosive percolation. One moment the network is completely fragmented; the next, a massive, macroscopic fraction of it snaps into a single connected web.
This sudden behavior raises a fundamental question in statistical physics: Is this sharp transition a true mathematical discontinuity (a first-order phase transition), or is it a very steep continuous transition (a second-order transition) whose true nature is masked by the finite size of our observations?
Weak versus Strong Discontinuities
To determine whether a sudden jump is a permanent feature of a system or an artifact of scale, researchers analyze how the transition behaves as the network grows toward the thermodynamic limit (an infinite number of nodes, $N \to \infty$). This scaling analysis distinguishes between two classes of abrupt transitions:
- Weak Discontinuities: Jan Nagler et al. and Jiang-Hai Qian et al. demonstrated that in many local competition models, the perceived jump is technically a finite-size effect. As the system size approaches infinity, the relative size of the connectivity jump vanishes, meaning the transition is mathematically continuous. However, because the jump size decays extremely slowly as the network grows, the transition remains practically discontinuous for any real-world, finite network. In these systems, even a single link addition can trigger a massive, disruptive shift in global connectivity.
- Strong Discontinuities: Conversely, true first-order transitions do exist at infinite scales. Raissa M. D’Souza demonstrated that by generalizing network growth rules, one can trigger a strongly discontinuous transition where the largest component jumps by a constant fraction of the total network size, regardless of how large the network is. Under these conditions, multiple giant components can burst into existence simultaneously and remain stable.
Jump Size
▲
│ ┌─────────────────────────┐
│ │ Strong Discontinuity │ (Persists at infinite scale)
│ └─────────────────────────┘
│
│ ┌─────────────────────────┐
│ │ Weak Discontinuity │ (Vanishes at N ➔ ∞, but very slowly)
│ └─────────────────────────┘
└────────────────────────────────► System Size (N)
Microscopic Drivers: Suppression and Higher-Order Interactions
Why do these sudden jumps happen? The core structural mechanism is the suppression of large clusters. If a network's growth rules systematically prevent medium-sized clusters from merging early on, the network accumulates a massive "powder keg" of isolated, similarly sized components. When a newly added link finally bridges these components, they merge in a rapid, explosive cascade.
Jiang-Hai Qian et al. explored this using a "best-of-$m$" node selection rule, proving that the transition is only strongly discontinuous if the selection pool $m$ is infinite; for any finite $m$, the transition eventually smooths out into a weak discontinuity as $N \to \infty$.
Beyond structural connectivity, these abrupt jumps also occur in dynamical processes like epidemics and information propagation due to cooperative effects and higher-order interactions:
- Cooperative Spreading: Kihong Chung et al. showed that when multiple exposures are required for an agent to become infected or adopt a behavior, the transition from a localized outbreak to a global epidemic shifts from a smooth, continuous process to a sudden, discontinuous jump.
- Simplicial Complexes: Traditional models look at pairwise interactions (edges). Wenjie Li et al. showed that when group interactions are modeled using simplicial complexes, the higher-order dynamics inherently cause epidemic growth to become discontinuous. A population can shift instantly from a healthy state to a widespread outbreak with no gradual warning phase.
The Bimodal Nature of Information Cascades
This physics of discontinuity directly explains the behavior of information cascades, viral marketing, and rumor propagation in social networks. When a system operates near a discontinuous phase transition, it exhibits a starkly bimodal distribution of outcomes.
Sameet Sreenivasan et al. analyzed this in feed-based networks where users have limited attention. In the supercritical regime, a message rarely achieves moderate success; it either disappears almost immediately or explodes to reach a massive, macroscopic fraction of the entire network.
This bimodal behavior implies the existence of a strict tipping point, or critical percolation threshold:
- Emergency Broadcasts: Charles Koll et al. found that an initial official broadcast must reach a specific minimum percentage of a population (typically 1% to 5%) to trigger a self-sustaining peer-to-peer relay. If the initial reach falls below this threshold, the informal network fails to bridge the gap, leaving large portions of the population unreached.
- Network Coupling: Dajun Qian et al. demonstrated that the presence of tightly knit physical clusters significantly lowers the transmission threshold required to trigger a global information cascade across multi-layered networks.
- Competitive Dynamics: When competing pieces of information spread simultaneously, Li Cao et al. showed that minor adjustments in network parameters or infectivity can cause a discontinuous shift, causing one piece of information to abruptly vanish while the other achieves total prevalence. Similar threshold dynamics govern viral marketing seeding (Hamed Amini et al.) and the sudden onset of social unrest (José Cadena et al.; Adrien Guille et al.).
Why Discontinuity Scaling Matters
Distinguishing whether a network property scales continuously or discontinuously dictates our ability to predict, model, and control complex systems.
Traditional frameworks for network scaling often assume that properties change smoothly as nodes or edges are added. However, as Desmond Higham et al. noted, discrete topological features like global connectivity can shift abruptly, violating the continuity assumptions required for predictable scaling laws. Furthermore, Xueqi Cheng et al. discovered that the largest connectivity jump does not always happen at the initial threshold; it can occur later in the supercritical regime due to the unstable coexistence and sudden merging of multiple giant components.
Similarly, Mikhail Tamm et al. showed that networks can undergo sharp transitions where they suddenly lock into structural "islands of stability," acting as traps that resist gradual change.
By using scaling analysis to track how the interval between successive connectivity transitions changes as a system approaches the thermodynamic limit, researchers can determine whether an observed sudden shift is an unmanageable, permanent feature of an infinite system or a scale-dependent phenomenon that can be mitigated through strategic structural interventions.
Key Foundational Work
Impact of single links in competitive percolation (2011) — contributed 4 concept elements to this direction.
Explosive Percolation with Multiple Giant Components (2011) — contributed 3 concept elements to this direction.
Marketing in a Random Network (2008) — contributed 1 concept elements to this direction.