.png){kind=icon}
.png){kind=icon}
Return
Flow condensation and nonlinear fragility in spatially embedded directed networks
.png){kind=icon}
PRE
AI
N
S
DOI：10.1016/j.physa.2026.131773.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Directed transport on spatially embedded weighted networks is constrained simultaneously by geometry, topology, and stochastic accessibility. In such systems, nonequilibrium probability current need not remain broadly distributed: it can condense onto a sparse backbone that sustains efficient target access while rendering the network exceptionally vulnerable to localized perturbations. We study this phenomenon in a Drosophila optic-lobe connectome through three coupled observables: an energy-like wiring-synapse cost Etotal(η) , a capped mean first-passage latency Lglobal(c) to an absorbing target set, and a reachability fraction C measuring the support of finite access. Relative to progressively constrained microcanonical maximum-entropy ensembles N0/N1/N2 , the empirical graph occupies an extreme sector of the joint (Etotal,Lglobal,C) space that is not reproduced by any null ensemble. A target-conditioned stationary diffusion field reveals strong current condensation: the top 1% of directed edges carries approximately 39% of the stationary flux. Removing that small subset triggers an approximately 84% collapse of finite-horizon target-hitting probability in the audited transport graph, exposing a nonlinear fragility that is absent from randomized controls and from a minimal generative rewiring model. We interpret the connectome as a single empirical realization, not as proof of universality: it demonstrates that efficient access and acute structural collapse can coexist in a dense spatial network under well-defined transport observables.
Journal
P
IF:
3.1
Papers:
1.3K
Citations:
3.6W
Organization
Cited Papers