Clustering coefficient and periodic orbits in flow networks

TL;DR

This paper demonstrates that the clustering coefficient in flow networks can identify approximate locations of periodic trajectories and cyclic motions in fluid flows and dynamical systems, providing a spectral method to detect periodic orbits.

Contribution

It introduces a novel application of the clustering coefficient to flow networks for detecting periodic orbits and cyclic motions across various fluid and dynamical systems.

Findings

Clustering coefficient highlights approximate locations of periodic trajectories.

High mean clustering at specific  values indicates periodic orbits of period 3.

Method successfully applied to fluid flows, Lorenz system, and Mediterranean sea circulation.

Abstract

We show that the clustering coefficient, a standard measure in network theory, when applied to flow networks, i.e. graph representations of fluid flows in which links between nodes represent fluid transport between spatial regions, identifies approximate locations of periodic trajectories in the flow system. This is true for steady flows and for periodic ones in which the time interval $\tau$ used to construct the network is the period of the flow or a multiple of it. In other situations the clustering coefficient still identifies cyclic motion between regions of the fluid. Besides the fluid context, these ideas apply equally well to general dynamical systems. By varying the value of $\tau$ used to construct the network, a kind of spectroscopy can be performed so that the observation of high values of mean clustering at a value of $\tau$ reveals the presence of periodic orbits of period $3\tau$ which impact phase space significantly. These results are illustrated with examples of increasing complexity, namely a steady and a periodically perturbed model two-dimensional fluid flow, the three-dimensional Lorenz system, and the turbulent surface flow obtained from a numerical model of circulation in the Mediterranean sea.