Dynamics of globally coupled oscillators: progress and perspectives

• Applications span electrochemistry, quantum electronics, bridge engineering, neuroscience, and social phenomena such as pedestrian synchrony.Source: §I · pp. 2–3
• Experiments and observations report collective oscillatory behavior in electrochemical and electronic oscillators, metronomes, Josephson junctions, laser arrays, yeast cells, and bacterial clocks.Source: §I · p. 3
• The globally coupled Winfree model exhibits a transition to a macroscopic synchronized state characterized by a non-zero mean field.Source: §II · p. 3
• Collective synchrony is a threshold phenomenon: it occurs when coupling is sufficiently strong or frequency-distribution inhomogeneity is sufficiently small.Source: §II · p. 3
• For the Kuramoto model, a critical coupling proportional to frequency-distribution width separates zero mean field from a non-zero collective mode.Source: §II · p. 4
• The Kuramoto collective mode can be interpreted as a second-order nonequilibrium phase transition, while a uniform frequency distribution produces a jump in the order parameter.Source: §II · p. 4

• The Kuramoto model and its phase-shift extension became paradigmatic models for analyzing large oscillator ensembles.Source: §III · pp. 4–5

A. Watanabe-Strogatz theory

• Watanabe–Strogatz theory describes identical oscillators under arbitrary common forcing using three global variables and N − 3 constants of motion.Source: A. Watanabe-Strogatz theory · p. 5
• The Watanabe–Strogatz equations completely describe the evolution of an ensemble of identical oscillators.Source: A. Watanabe-Strogatz theory · p. 6

• For a hierarchical population of M groups of identical units, the ensemble dynamics obey M coupled Watanabe–Strogatz equations.Source: A. Watanabe-Strogatz theory · p. 6
• A large population with a frequency distribution g(ω) is characterized by frequency-dependent Watanabe–Strogatz variables z(ω,t) and α(ω,t).Source: A. Watanabe-Strogatz theory · p. 6

Closed mean-field dynamics

• Forcing independent of the phase variable makes the phase equation irrelevant, leaving a closed equation for the order parameter that also appears in OA theory.The resulting closed equations are obtained for the order parameter variables.Source: B. From WS to Ott-Antonsen theory · p. 7

WS and OA reductions

• The OA manifold is a special WS solution and, for continuous frequency distributions, is argued to be the only attractor, although relaxation may be slow.The OA manifold corresponds to a uniform distribution of constants of motion.Source: B. From WS to Ott-Antonsen theory · p. 7

Lorentzian reduction

• For a Lorentzian frequency distribution, analyticity permits residue integration, giving Y = Z(i) and a reduced equation for the mean field.A stationary synchronous solution has amplitude R0 and mean-field frequency ν = (ε cos β −1) tan β.Source: B. From WS to Ott-Antonsen theory · pp. 7–8

WS and OA reductions

• The WS equations describe two interacting oscillator populations, while uniformly distributed constants of motion reduce the system from six to four dimensions on the OA manifold.The local WS variable coincides with the local Kuramoto mean field in the relevant setting.Source: B. From WS to Ott-Antonsen theory · pp. 6, 8

Chimera states

• The reduced two-population model admits chimera states in which one population synchronizes while the other remains partially synchronized; the full WS system also permits quasiperiodic chimeras.Theoretical predictions were confirmed in experiments with two groups of coupled metronomes.Source: B. From WS to Ott-Antonsen theory · p. 8

Nontrivial transitions for unimodal distributions

• Unimodal frequency distributions can produce first-order transitions and bistability, contrary to the earlier assumption of qualitative similarity with Lorentzian dynamics.These complex transition scenarios were demonstrated for some distributions.Source: C. Generalizations of the Kuramoto model · pp. 8–9

Complex coupling schemes

• Complex coupling schemes let the mean field drive nonlinear macroscopic variables, which in turn affect oscillator phases, as in pedestrian, electronic, and electrochemical systems.Examples include bridge motion driven by pedestrians and coupling through common current or voltage.Source: C. Generalizations of the Kuramoto model · p. 9

Effects of noise

• Noise makes synchronization threshold-like even for identical oscillators, with critical coupling proportional to noise intensity, whereas common noise tends to synchronize identical units.Independent noise is represented by a nonlinear Fokker–Planck equation in the thermodynamic limit.Source: C. Generalizations of the Kuramoto model · p. 10

Finite-size fluctuations

• Finite populations with heterogeneous frequencies fluctuate around the synchronization transition; small uniform-frequency populations can be chaotic beforehand, while large near-critical systems can be quasiperiodic.The maximal Lyapunov exponent decreases as λ ∼ N^-1, and regular dynamics are observed above the transition.Source: C. Generalizations of the Kuramoto model · p. 10

Kuramoto model on networks

• Synchronization transitions depend on network topology, with studies spanning small-world, lattice, modular, and hub-centered networks, including power-grid applications.The network generalization extends the original globally coupled formulation.Source: C. Generalizations of the Kuramoto model · p. 10

External forcing and collective phase resetting

• Periodic forcing can entrain the mean-field frequency, while pulse forcing shifts the collective phase and defines a collective phase-response curve.The macroscopic order parameter obeys an equation for a self-sustained oscillator.Source: C. Generalizations of the Kuramoto model · pp. 10–11

General coupling analysis

• For general coupling, Fourier expansion introduces generalized order parameters whose values must generally be determined self-consistently.The paper states that a complete analysis remains missing.Source: §IV · p. 11

Clustering

• Even identical oscillators can form several clusters, with each cluster consisting of fully synchronized units.This is a clustering regime beyond the direct single-coherence picture.Source: §IV · p. 11

Heteroclinic cycles

• Identical oscillators with first- and second-harmonic coupling can exhibit clustering and switching between cluster states through a heteroclinic cycle.The heteroclinic-cycle mechanism is well understood for small networks.Source: §IV · p. 11

Multi-branch entrainment

• Two harmonic components can create two stable entrainment branches, allowing the same mean field to entrain oscillators at two microscopic phases.This produces many microscopic states and a complex structure of macroscopic regimes.Source: §IV · p. 11

• Phase reductions are appropriate when coupling weakly perturbs trajectories from the autonomous limit cycle; strong coupling generally requires full dynamical models that are difficult to treat analytically.Source: §V · p. 12
• Numerical studies find collective synchrony across periodic, noisy, and chaotic oscillators, including spiking Hindmarsh-Rose neurons.Experiments with chaotic electrochemical oscillators confirm theoretical predictions.Source: §V · p. 12
• Stuart-Landau ensembles exhibit oscillation quenching when strong coupling adds effective damping to individual elements.Source: §V · p. 12
• They can also display collective chaos even when the constituent oscillators are initially periodic.Source: §V · p. 12
• Non-identical chaotic Rössler oscillators phase-synchronize and produce a nearly periodic mean field while remaining individually chaotic.Amplitude fluctuations are largely averaged out, with residual mean-field fluctuations attributed presumably to finite size.Source: §V · p. 12
• Inertial rotators can depart sharply from Kuramoto dynamics, with synchronization transitions exhibiting hysteresis analogous to a first-order phase transition.Source: §V · p. 12

A. Partial synchrony

• Partial synchrony describes states between fully synchronous and fully asynchronous behavior, including clustered and non-clustered arrangements.Source: A. Partial synchrony · p. 13
• In the Hindmarsh-Rose ensemble, synchrony loses stability at a critical coupling through a Hopf-like bifurcation, then returns for very large coupling.The bifurcation occurs when two complex multipliers cross the unit circle and create a new frequency.Source: A. Partial synchrony · p. 14
• Beyond synchrony breaking, oscillators form a slowly interchanging stripe with matched average frequency and slowly modulated phase shifts, producing possibly weak chaos.The characteristic interchange time is tens of periods.Source: A. Partial synchrony · p. 14
• With nonlinear Josephson-junction coupling, the synchronous state destabilizes through a real multiplier, leading to partial synchrony whose order parameter decreases smoothly with coupling.Source: A. Partial synchrony · p. 14
• In that Josephson-junction regime, the mean field oscillates faster than individual junctions, and their frequency difference grows with ε − εc.Source: A. Partial synchrony · p. 14
• A general theory of partially synchronous states appearing after synchrony breaking is still missing and requires further investigation.Source: A. Partial synchrony · p. 14

• The extended Stuart-Landau model uses mean-field amplitude R and coupling-dependent functions α(ε, R) and β(ε, R) to generalize Kuramoto-Sakaguchi dynamics.Source: B. Self-organized quasiperiodic dynamics · p. 15
• For α(ε, R) = (1 − εR^2)R, increasing coupling changes the fully synchronous state from stable attraction to effective repulsion and produces a self-organized bunch state.Source: B. Self-organized quasiperiodic dynamics · p. 15
• In the bunch state, oscillator phases spread around the circle while the bunch remains stationary in a rotating coordinate frame with frequency ω.Source: B. Self-organized quasiperiodic dynamics · p. 15
• For β(ε, R) = β0 + β1ε^2R^2, synchrony loses stability at β0 + β1ε^2R^2 = π/2, and the resulting state has distinct, generally irrationally related collective and individual frequencies.Source: B. Self-organized quasiperiodic dynamics · p. 15
• This two-frequency state is termed self-organized quasiperiodicity, or SOQ, and quantitative analyses compute both collective-mode and oscillator frequencies.SOQ states have been demonstrated experimentally in electronic circuits with global nonlinear coupling.Source: B. Self-organized quasiperiodic dynamics · pp. 15–16
• Complex collective states can also arise without a desynchronization transition when both fully synchronous and completely asynchronous states are unstable.Source: B. Self-organized quasiperiodic dynamics · p. 16

• Chimeralike states split a globally coupled population into coherent and incoherent groups, despite identical elements receiving the same force.Source: C. Chimeralike states in globally… · p. 16
• A modified Stuart-Landau construction uses two stable limit cycles with frequencies Ω2 > Ω1 and coupling mediated by a harmonic oscillator.Source: C. Chimeralike states in globally… · p. 17
• Choosing Ω1 < η < Ω2 makes the coupling synchronize large-amplitude oscillations while preventing synchronization of low-amplitude ones.Source: C. Chimeralike states in globally… · p. 17
• These delayed-feedback states occur when both synchronous and zero-mean-field asynchronous states are unstable, forcing the system between full coherence and incoherence.Source: C. Chimeralike states in globally… · p. 17
• Coupling can create bistability and thereby sustain a chimeralike state even when the autonomous delayed-feedback oscillators are monostable.Source: C. Chimeralike states in globally… · p. 17

Neuronal synchronization

• Synchronization has been observed in fully connected neuronal models during both periodic spiking and bursting.Source: §VII · p. 18

Network structure

• 0.5% connectivity produced synchronization properties practically indistinguishable from those of the fully connected network in randomly coupled map-based neurons.Source: §VII · p. 18
• Heterogeneous mean-field modeling preserves connection disorder as system size grows while describing microscopic and global features of neuronal synchrony.Source: §VII · p. 18
• For large coupling, diluted random spike-coupled networks can exhibit an exponentially long transient disordered state with negative largest Lyapunov exponent.Source: §VII · pp. 18–19

Stimulation and control

• Deep brain stimulation is framed as a desynchronization problem for pathological neuronal rhythms, with the goal of suppressing unwanted activity using minimal stimulation.Source: A. Control of collective synchrony · p. 19
• Feedback control can compensate the mean field, allowing oscillators to desynchronize; after suppression, vanishing stimulation tends toward zero.Source: A. Control of collective synchrony · p. 19

VIII. SURPRISES AND OUTLOOK

• Even the simplest sine-coupled phase-oscillator model exhibits partial integrability and an exact low-dimensional manifold.Source: §VIII · p. 20
• The same model supports nontrivial transition scenarios for unimodal frequency distributions, chimera states, and self-organized quasiperiodicity.Source: §VIII · p. 20
• Generalizations of the basic model display a broad range of dynamical phenomena that remains far from exhausted.Source: §VIII · p. 20
• Synchronization theory may extend beyond nonlinear dissipative systems to quantum objects, while improved experiments and analysis could reveal microscopic synchronization patterns.Source: §VIII · p. 20