Thresholds for epidemic spreading in networks

We study the threshold of epidemic models in quenched networks with degree distribution given by a power-law. For the susceptible-infected-susceptible (SIS) model the activity threshold lambda_c vanishes in the large size limit on any network whose maximum degree k_max diverges with the system size, at odds with heterogeneous mean-field (HMF) theory. The vanishing of the threshold has not to do with the scale-free nature of the connectivity pattern and is instead originated by the largest hub in the system being active for any spreading rate lambda>1/sqrt{k_max} and playing the role of a self-sustained source that spreads the infection to the rest of the system. The susceptible-infected-removed (SIR) model displays instead agreement with HMF theory and a finite threshold for scale-rich networks. We conjecture that on quenched scale-rich networks the threshold of generic epidemic models is vanishing or finite depending on the presence or absence of a steady state.

💡 Research Summary

The paper investigates the epidemic threshold of two canonical contagion models—susceptible‑infected‑susceptible (SIS) and susceptible‑infected‑removed (SIR)—on quenched (static) networks whose degree distribution follows a power‑law. Classical heterogeneous mean‑field (HMF) theory predicts that the epidemic threshold λc depends only on the moments of the degree distribution and, for scale‑free networks with diverging second moment, λc should vanish. The authors show that this prediction holds only for models without a steady state (e.g., SIR). For SIS, which possesses a stationary infected state, the threshold disappears in the thermodynamic limit whenever the maximum degree kmax grows with system size N. The key mechanism is the emergence of a “self‑sustained hub”: a node with degree kmax becomes permanently active for any infection rate λ larger than 1/√kmax. Because this hub can continuously reinfect its neighbors, it acts as an autonomous source that keeps the infection alive throughout the network, regardless of the rest of the topology. Consequently, λc→0 as N→∞ even if the degree distribution is not strictly scale‑free, provided that kmax diverges.

In contrast, the SIR model lacks a stationary infected phase; infected nodes become removed and cannot be reinfected. The hub can trigger a large outbreak, but after it recovers the infection cannot be regenerated, and the epidemic dies out. Therefore the SIR threshold matches the HMF prediction and remains finite for “scale‑rich” networks (those with bounded kmax).

The authors support their analytical arguments with extensive numerical simulations on synthetic networks with various degree exponents γ and different scaling laws for kmax (e.g., kmax∝Nα). For SIS, they observe that the prevalence stays non‑zero for any λ>1/√kmax, confirming the vanishing of λc. For SIR, the prevalence drops to zero at a finite λc that agrees with the HMF formula λc=⟨k⟩/⟨k²⟩.

From these results they conjecture a general principle: on quenched, scale‑rich networks the epidemic threshold of any model is either zero or finite depending solely on whether the model admits a steady infected state. This insight refines the traditional view that the degree distribution alone determines epidemic resilience. It also has practical implications. In SIS‑like processes (computer viruses, persistent rumors) targeting the largest hub—through immunization, quarantine, or reducing its effective transmission rate—can restore a finite threshold and suppress endemic infection. In SIR‑like processes (acute diseases, one‑shot marketing campaigns) hub removal is insufficient; control strategies must instead reduce overall connectivity or lower the transmission probability below the HMF‑predicted λc.

The paper concludes by outlining future directions: extending the analysis to annealed or temporally evolving networks, considering multiple interacting contagions, and studying how adaptive rewiring of hubs influences the threshold. Such extensions would bridge the gap between the idealized static framework presented here and the complex, time‑varying structures encountered in real‑world epidemic and information‑spreading scenarios.