Abstract
We consider random nondirected networks subject to dynamics conserving vertex degrees and study, analytically and numerically, equilibrium three-vertex motif distributions in the presence of an external field $h$ coupled to one of the motifs. For small $h$, the numerics is well described by the "chemical kinetics" for the concentrations of motifs based on the law of mass action. For larger $h$, a transition into some trapped motif state occurs in Erdős-Rényi networks. We explain the existence of the transition by employing the notion of the entropy of the motif distribution and describe it in terms of a phenomenological Landau-type theory with a nonzero cubic term. A localization transition should always occur if the entropy function is nonconvex. We conjecture that this phenomenon is the origin of the motifs' pattern formation in real evolutionary networks.
Introduction
In the study of complex networks, "motifs" are small, recurring subgraphs—such as triangles of three connected nodes—that serve as fundamental building blocks for network function. While specific motifs are frequently overrepresented in biological and social systems, the physical mechanisms that stabilize these distributions are not fully understood. This research investigates whether an underlying "entropic landscape" favors certain motif configurations, causing networks to "choose" specific structural states rather than drifting through all possible configurations.
Methodology
The researchers utilized Erdős-Rényi (ER) random networks as a baseline model, with sizes ranging from $N = 40$ to $320$ nodes and connection densities $p$ of $0.05$ and $0.35$. To simulate network evolution, they employed a "link permutation" method that swaps edges between nodes while keeping the individual node degrees constant.
An external field ($h$) was introduced to act as a mathematical bias, rewarding or punishing the formation of specific motifs. The team then applied Landau-type theory, a standard tool in statistical mechanics, to determine if the motif concentration (the order parameter $\phi$) changed smoothly or abruptly in response to this bias.
Key Insights: The Entropic Landscape
At low levels of external bias, motif formation follows the "law of mass action," similar to chemical reactions in a gas where components interact independently. However, the study revealed that the entropy of the motif distribution is non-convex. This non-convexity creates "islands of stability"—deep valleys in the structural energy landscape. Once a network falls into one of these valleys, it becomes "trapped" in a specific motif pattern, making that configuration highly stable even under small evolutionary pressures.
Results and Discussion
Numerical simulations demonstrated a first-order phase transition: as the bias $h$ increases, the network does not gradually add motifs but instead hits a critical point and "jumps" suddenly to a high-motif state.
This transition is characterized by hysteresis, meaning the network "remembers" its state; a pattern persists longer when the bias is decreased than it took to form initially. The researchers also found that the critical bias required to trigger this jump decreases as the network size increases, scaling approximately as $N^{-1/2}$. These behaviors were observed in both ER and scale-free networks, suggesting that these islands of stability are a fundamental property of complex system organization.
Conclusion
The existence of these entropic traps provides a physical explanation for why certain structural patterns persist in evolving networks. Future research will explore directed networks, where the increased variety of triads (16 types compared to 4 in undirected networks) likely creates an even more complex landscape of structural stability.
Summary updated