Paths to synchronization in the Kuramoto model with inertia

Synchronization is ubiquitous across natural and synthetic systems, yet most prior studies focus on the inertia-free Kuramoto model and do so at the macroscopic level<ref:2607.09171#pg2>. This study investigates the inertial Kuramoto model to analyze the kinetics of individual synchronized clusters that emerge in underdamped dynamics driven by interactions among multiple clusters with different frequencies<ref:2607.09171#pg2>.

The Role of Frequency Distribution Shape

The shape of the intrinsic frequency distribution dictates qualitatively different synchronized cluster organizations based on whether it is unimodal Gaussian or multimodal uniform<ref:2607.09171#pg3>. Specifically, a unimodal Gaussian distribution gives rise to a hierarchical organization for the clusters, whereas a multimodal uniform distribution results in a homogeneous organization<ref:2607.09171#pg3>. This contrast leads to qualitatively different behaviors of the order parameter R: it increases smoothly with coupling strength for the Gaussian distribution, while for the uniform distribution, it grows through discrete jumps that trace out the size of the Devil’s staircase (DS)<ref:2607.09171#pg3>.

Kinetics and Cluster Assembly

The route to synchronization also depends on the distribution type when resolving kinetics at the cluster level<ref:2607.09171#pg3>. For a Gaussian distribution, a single dominant cluster forms and gradually entrains the remaining oscillators, whereas with a uniform distribution, synchronization proceeds via successive cluster mergers initiated from peripheral seeds associated with the high-frequency periphery<ref:2607.09171#pg3>. The contrasting structures are seen in time-averaged angular velocity versus intrinsic frequency for a single realization, where the Gaussian case produces a hierarchical Devil’s staircase centered around one dominant cluster, while the uniform distribution yields clusters of comparable size spaced by uniform gaps<ref:2607.09171#pg6>.

Robustness and Stability

The shape of the distribution governs system stability by constraining assembly pathways, ensuring that entrainment-driven or merger-driven mechanisms remain robust in complementary coupling regimes<ref:2607.09171#pg3>. The Gaussian recovers the peripheral oscillators through individual re-entrainment at weak to intermediate coupling, while the uniform recovers them through cluster merging at strong coupling<ref:2607.09171#pg9>. This pathway in turn fixes stability, with entrainment dominating at weak coupling and merger dominating at strong coupling<ref:2607.09171#pg9>.

Microscopic Determinants of Synchronization

The breakdown of the single-cluster criterion is resolved by analyzing the initial energy of each oscillator in its candidate cluster’s center-of-mass frame<ref:2607.09171#pg26>. The separation between captured and uncaptured oscillators is robust inside the Melnikov radius at all coupling strengths, with the Gaussian case maintaining a dominant diagonal core while the uniform case fractures into multiple horizontal and vertical bands<ref:2607.09171#pg23>. This leads to a stochastic capture-and-escape process about the Melnikov radius, where capture is governed by two probabilities a and b<ref:2607.09171#pg24>.

Implications for Physical Systems

These findings suggest that modern power grids increasingly rely on merger-driven synchronization, whose robustness rests on whole groups of high-power peripheral buses rejoining together rather than on individual nodes re-entraining. The shape of the frequency distribution determines whether the system relies on central entrainment or peripheral merger, which has potential implications for grid operators needing to prioritize protection for these peripheral buses.

Consistency Across Parameters

The contrast between hierarchical and homogeneous organization, as well as the smooth versus stepwise growth of the order parameter, is a robust signature of the distribution shape rather than of any particular inertia or variance<ref:2607.09171#pg19>. This distinction holds throughout parameter space across inertia and variance variations<ref:2607.09171#pg31>. The precision and participation of the rational locking are universal, remaining fixed features independent of the distribution shape<ref:2607.09171#pg29>.

Summary and Discussion

Inertia first breaks the conventional mean-field description by allowing multiple clusters with distinct rotational frequencies to coexist and interact, turning synchronization into a multi-frequency synchronized cluster interaction-driven process<ref:2607.09171#pg4>. The shape of the distribution governs the dynamics throughout, fixing the cluster organization, assembly pathway, and ultimately the system’s robustness<ref:2607.09171#pg4>. The Gaussian organizes hierarchically around one dominant cluster with unequal gap ratios such as 2: 1 and grows that cluster by entraining oscillators outward from the center<ref:2607.09171#pg6>. In contrast, the uniform distribution forms clusters of comparable size with equal 1: 1 gaps, reflecting a homogeneous form of the Devil’s staircase<ref:2607.09171#pg6>. The pathway in turn fixes stability: the Gaussian recovers oscillator by oscillator through re-entrainment and is thus robust from weak to intermediate K, whereas the uniform recovers cluster by cluster through merger and is thus robust at strong K<ref:2607.09171#pg4>. The single structural feature—the central peak—thus runs through the entire chain, from cluster organization through the assembly pathway to system robustness<ref:2607.09171#pg4>.

How the training signal is built

The transient capture probability Pts(∆ω) shows a central dip and splits into two lobes around and beyond ∆ω = 1 for the uniform distribution as K grows, confirming the shift from central to peripheral seeding<ref:2607.09171#pg9>. This migration of seeds is a robust signature of the distribution shape, not of any particular inertia or variance<ref:2607.09171#pg4>. The Gaussian case maintains a unimodal peak at ∆ω = 0 for all K, indicating central seeding<ref:2607.09171#pg9>.

Where the gains come from

The final link in the chain running from distribution shape through cluster organization and assembly pathway to system robustness is determined by whether the growth is entrainment-driven or merger-driven<ref:2607.09171#pg4>. The Gaussian recovers readily at weak to intermediate K, its perturbed curves tracking the unperturbed one closely as its periphery is reclaimed oscillator by oscillator through re-entrainment. Conversely, the uniform stays trapped in a low-R plateau there, a wide gap separating its perturbed curves from the unperturbed one, and recovers only at strong K where its displaced periphery re-forms clusters and rejoins by merger.

Pruning redundant reasoning tokens

The order parameter R grows not smoothly but through discrete jumps, its susceptibility ∆R spiking at each one as nearly identical initial conditions diverge into different final states<ref:2607.09171#pg5>. The shape’s role must therefore be understood at the mesoscopic level of the clusters, where it sets both structure and route<ref:2607.09171#pg4>. A Gaussian, with its dense central peak, organizes hierarchically around one dominant cluster with unequal gap ratios such as 2: 1 and grows that cluster by entraining oscillators outward from the center<ref:2607.09171#pg6>. A uniform distribution forms clusters of comparable size with equal 1: 1 gaps—a homogeneous DS—and assembles its giant cluster by merging peripheral seeds<ref:2607.09171#pg6>. The contrast in size structure, where the Gaussian carries many small clusters alongside its dominant one (ρL low, ρS high), while the uniform is almost all large clusters (ρL → 1, ρS → 0), is a robust signature of the distribution shape<ref:2607.09171#pg31>. The size structure splits the same way in every case, confirming that the large clusters form first and small ones only afterward<ref:2607.09171#pg31>.

The precision and participation of the rational locking are fixed features of the dynamics, independent of distribution shape<ref:2607.09171#pg29>. Both properties—the precision of the locking and its participation in it—underpin the staircase itself<ref:2607.09171#pg29>. This high participation is consistent with rational locking being energetically favorable, the commensurate ratios acting as the preferred configuration rather than a rare coincidence<ref:2607.09171#pg29>.

1. Adaptive Synchronization Strategy for Power Grids: AI systems can dynamically switch between entrainment-driven synchronization (for Gaussian frequency distributions) and merger-driven synchronization (for uniform frequency distributions) based on real-time grid conditions. This allows the system to utilize entrainment, working oscillator by oscillator at weak coupling and merger, working cluster by cluster at strong coupling for optimal stability.

2. Predictive Cluster Organization: AI can predict the resulting cluster organization—hierarchical (Gaussian) or homogeneous (uniform)—given a specific intrinsic frequency distribution of renewable energy sources. This prediction helps in anticipating the structure of the synchronization transition before it occurs, which is currently only observed after dynamics unfold.

3. Robustness Assessment Under Perturbation: AI can quantify system robustness against peripheral disruptions by predicting whether a synchronized state will recover via individual re-entrainment or cluster merging. This capability allows grid operators to prioritize protection for peripheral buses when the system is operating in the uniform distribution regime, where recovery relies on cluster mergers at strong coupling.

4. Optimal Frequency Allocation: For systems with a uniform frequency distribution, AI can determine the optimal coupling strength K that maximizes synchronization efficiency by analyzing discrete jumps in the order parameter R and corresponding susceptibility peaks log Delta R. This ensures the system operates at the most sensitive point of transition for rapid coherence gain.

Synchronization is ubiquitous across natural and synthetic systems, yet most prior studies focus on the inertia-free Kuramoto model and do so at the macroscopic level. In this study, we instead investigate the inertial Kuramoto model and analyze the kinetics of individual synchronized clusters that emerge in the underdamped dynamics, driven by the interactions among multiple synchronized clusters with different frequencies. Specifically, we explore two forms of intrinsic frequency distribution -- unimodal Gaussian and multimodal uniform -- and show that they give rise to qualitatively different synchronized clusters: a hierarchical organization for the Gaussian distribution and a homogeneous organization for the uniform distribution. This contrast leads to qualitatively different behaviors of the order parameter: for the Gaussian distribution, it increases smoothly with increasing coupling strength, while for the uniform distribution, it grows through a series of discrete jumps that trace out the size of the Devil's staircase (DS). By resolving the kinetics at the cluster level, we further find that the route to synchronization also depends on the distribution type: with a Gaussian distribution, a single dominant cluster forms and gradually entrains the remaining oscillators, whereas with a uniform distribution, synchronization proceeds via successive cluster mergers initiated from peripheral seeds associated with the high-frequency periphery. Taken together, these findings provide a new perspective on collective synchronization dynamics in inertial complex systems.