Abstract
The dynamics of zero-range processes (ZRP) on complex networks are significantly influenced by the underlying topological structure. In the stationary state, a real-space complete condensation phase transition can occur, where particles accumulate on network hubs. This paper investigates the finite density effects of this transition in both stationary and dynamical ZRP on scale-free networks. Using the grand canonical ensemble method, we analytically predict scaling laws for the average occupation number relative to finite density. Furthermore, we explore relaxation dynamics through a hierarchical evolution hypothesis and scaling ansatz. Monte Carlo simulations on Barabási-Albert networks validate these predicted density scaling laws.
Introduction
The Zero Range Process (ZRP) is a cornerstone model in non-equilibrium statistical mechanics, providing insights into phenomena ranging from traffic flow to Bose-Einstein condensation. In a ZRP, particles hop between nodes at a rate determined solely by the occupancy of the departure node. On scale-free (SF) networks—characterized by a power-law degree distribution where a few "hubs" possess the majority of connections—particles tend to condense onto the largest hub when the hopping rate is sufficiently low.
While the existence of this transition is well-documented, previous studies often treated particle density as a fixed constant. However, in real-world systems, density is frequently a tunable or fluctuating parameter. This research addresses a critical gap: determining how total particle density ($\rho$) quantitatively dictates the threshold for condensation and the time required for the system to reach equilibrium.
Theoretical Framework and Methodology
The study employs a dual analytical and numerical approach:
- Steady-State Analysis: Utilizing the Grand Canonical Ensemble (GCE) formalism, the researchers introduced an "effective fugacity" and a "crossover degree" ($k_c$). This allowed the network to be partitioned into a "fluid-like" phase (low-degree nodes) and a "condensed" phase (high-degree nodes).
- Dynamical Analysis: To model system evolution, a hierarchical evolution hypothesis was applied. This assumes a bottom-up equilibration where smaller nodes reach steady state first, followed by larger nodes, until the hub finally captures the condensate. A scaling ansatz was used to predict the evolution of the average occupation number ($m_k$) and the Inverse Participation Ratio (IPR).
- Numerical Validation: Monte Carlo (MC) simulations were conducted on Barabási-Albert (BA) networks (with degree exponent $\gamma = 3$) to test the analytical predictions across varying densities and system sizes.
Steady-State Scaling Laws
A primary finding is that the "crossover degree" $k_c$—the boundary between fluid and condensing nodes—is density-dependent, scaling as $k_c \sim \rho^{-\delta}$. As density increases, the threshold for condensation effectively drops.
As shown in Figure 1, the average occupation number $m_k$ collapses onto a single universal curve when plotted against the scaled variable $k\rho^\delta$. This confirms that despite the complexity of the network, the system follows universal scaling functions governed by density.
Relaxation Dynamics
The relaxation process is inherently hierarchical. Small-degree nodes equilibrate almost instantly, while the hub acts as a temporary sink that grows in importance over time. Crucially, increasing particle density postpones the completion of the condensation transition. Higher density necessitates the redistribution of more particles, extending the time required for the hub to claim its share of the condensate.
Figure 3 demonstrates the success of the dynamical scaling ansatz. The relaxation time $t_R$ scales with system size $L$ as $t_R \sim L^\beta$, where $\beta = 1 - \delta$ is the dynamic exponent. The data collapse in the Inverse Participation Ratio (IPR) further validates the hierarchical nature of the transition.
Conclusion
This work provides the first comprehensive scaling laws for ZRP on scale-free networks with variable density. By identifying density as a primary driver of both the condensation threshold and relaxation speed, the authors offer a more nuanced understanding of non-equilibrium phase transitions. Future research may extend these findings to other hopping rules or more complex network topologies, such as highly clustered social or biological networks.
Summary updated