Synchronization of heterogeneous Kuramoto oscillators with graphs of diameter two

Andrey Gushchin, Enrique Mallada, Ao Tang

Abstract

In this article, we investigate the synchronization of Kuramoto oscillators with heterogeneous natural frequencies on networks with a diameter of two. By assuming uniform coupling strengths, we derive an analytic condition that guarantees the existence of a Positively Invariant Set (PIS), which we demonstrate is sufficient for frequency synchronization. Our findings show that for diameter-two graphs, these conditions are significantly less conservative than existing general bounds for arbitrary topologies. Furthermore, we address non-uniform coupling through an optimization framework designed to minimize the total coupling strength required for synchronization.

Introduction: Bridging the Gap in Network Topology

Synchronization—the process where independent oscillators lock into a common frequency—is a fundamental phenomenon observed in biological systems like pacemaker cells and engineered systems like power grids. The Kuramoto model serves as the primary mathematical framework for studying these dynamics. However, existing research often focuses on two extremes: complete graphs (where every node is connected) or arbitrary topologies. While complete graphs are well-understood, they are physically unrealistic; conversely, general topology bounds are often excessively conservative, requiring coupling strengths far beyond what is necessary in practice.

This research focuses on graphs of diameter two, where any two nodes are either direct neighbors or share at least one common neighbor. This topology is a common feature in many real-world "small-world" networks. The challenge lies in heterogeneity, where each oscillator possesses a unique natural frequency, naturally resisting the consensus required for synchronization.

A simple graph of diameter two, where nodes 2 and 3 are connected via node 1.

Methodology: Stability and Optimization

The study employs a combination of control theory and graph theory to derive tighter bounds:

  • Positively Invariant Sets (PIS): Rather than seeking a single equilibrium, the authors identify a region (a PIS) where phases are trapped. They prove that if a system's phases start within this set, frequency synchronization is guaranteed.
  • Lyapunov Stability: A potential function approach, combined with LaSalle’s Invariance Theorem, is used to demonstrate that trajectories eventually settle into a synchronized state.
  • The Role of Common Neighbors: The researchers introduce a metric, $P_{ij}$, which accounts for both direct edges and shared neighbors. In diameter-two graphs, these shared neighbors act as mediators, pulling distant nodes toward synchronization more effectively than in sparse, high-diameter networks.
  • Linear Programming: For non-uniform coupling, the authors formulate an optimization problem to find the minimum total coupling strength needed to maintain synchronization, strategically allocating "energy" to the most critical links.

Key Results and Comparative Performance

The primary contribution is a new analytic bound for uniform coupling $K$: $$K \ge \frac{n \cdot |\bar{\omega}_i - \bar{\omega}_j|}{P_{ij} \cdot \sin D}$$ This formula directly links the required coupling strength to the frequency spread and the specific connectivity of the diameter-two graph.

Simulation showing the phases of three oscillators converging toward a common frequency over time.

In numerical simulations with 5 to 10 oscillators, this new condition proved remarkably robust. While existing general-purpose conditions from literature were satisfied by less than 20% of random diameter-two samples, the authors' specific condition was satisfied by 100% of the same samples. This demonstrates that narrowing the topological scope allows for significantly more efficient synchronization requirements—sometimes by an order of magnitude.

Discussion and Future Directions

While the diameter-two constraint provides powerful analytic gains, it remains a specific topological requirement. Many real-world networks have low diameters but may not strictly meet the diameter-two criteria. Additionally, the current results assume initial phases are relatively close (within a range of $\pi$).

Future work may extend these principles to larger diameters by identifying clusters of oscillators with similar frequencies or by exploring different optimization objectives, such as minimizing the maximum individual coupling strength rather than the total sum. This research paves the way for more realistic and efficient control of synchronized networked systems.

Summary updated

Abstract

In this article we study synchronization of Kuramoto oscillators with heterogeneous frequencies, and where underlying topology is a graph of diameter two. When the coupling strengths between every two connected oscillators are the same, we find an analytic condition that guarantees an existence of a Positively Invariant Set (PIS) and demonstrate that existence of a PIS suffices for frequency synchronization. For graphs of diameter two, this synchronization condition is significantly better than existing general conditions for an arbitrary topology. If the coupling strengths can be different for different pairs of connected oscillators, we formulate an optimization problem that finds sufficient for synchronization coupling strengths such that their sum is minimal.
BibTeX
@misc{gushchin2015synchronization,
  author = {Andrey Gushchin and Enrique Mallada and Ao Tang},
  title = {{Synchronization of heterogeneous Kuramoto oscillators with graphs of diameter two}},
  year = {2015},
  eprint = {1502.06137},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1502.06137},
}
Figure