From Kuramoto to Crawford: exploring the onset of synchronization in populations of coupled oscillators
Summary (4 min read)
1. Introduction
- At first glance, the papers look technical, maybe even a bit intimidating.
- Technical, yes, but a technical tour de force.
2. Background
- The Kuramoto model was originally motivated by the phenomenon of collective synchronization, in which an enormous system of oscillators spontaneously locks to a common frequency, despite the inevitable differences in the natural frequencies of the individual oscillators [10–13].
- Collective synchronization was first studied mathematically by Wiener [27,28], who recognized its ubiquity in the natural world, and who speculated that it was involved in the generation of alpha rhythms in the brain.
- A more fruitful approach was pioneered by Winfree [10] in his first paper, just before he entered graduate school.
- Using numerical simulations and analytical approximations, Winfree discovered that such oscillator populations could exhibit the temporal analog of a phase transition.
- Kuramoto himself began working on collective synchronization in 1975.
3.1. Governing equations
- Kuramoto [5] put Winfree’s intuition about phase models on a firmer foundation.
- Even though the reduction to a phase model represents a tremendous simplification, these equations are still far too difficult to analyze in general, since the interaction functions could have arbitrarily many Fourier harmonics and the connection topology is unspecified — the oscillators could be connected in a chain, a ring, a cubic lattice, a random graph, or any other topology.
- Like Winfree, Kuramoto recognized that the mean-field case should be the most tractable.
- The frequenciesωi are distributed according to some probability densityg(ω).
- For simplicity, Kuramoto assumed thatg(ω) is unimodal and symmetric about its mean frequency , i.e.,g( +ω) = g( −ω) for allω, like a Gaussian distribution.
3.2. Order parameter
- The complex order parameter [5] r eiψ = 1 N N∑ j=1 eiθj (3.2) is a macroscopic quantity that can be interpreted as the collective rhythm produced by the whole population.
- The radiusr(t) measures the phase coherence, andψ(t) is the average phase (Fig. 1).
- If all the oscillators move in a single tight clump, the authors haver ≈ 1 and the population acts like a giant oscillator.
- Each oscillator appears to be uncoupled from all the others, although of course they are interacting, but only through the mean-field quantitiesr andψ .
- Moreover, the effective strength of the coupling is proportional to the coherence.
3.3. Simulations
- For concreteness, suppose the authors fixg(ω) to be a Gaussian or some other density with infinite tails, and vary the couplingK.
- Thenr(t) decays to a tiny jitter of size O(N−1/2), as expected for any random scatter ofN points on a circle (Fig. 2).
- But whenK exceedsKc, this incoherent statebecomes unstable andr(t) grows exponentially, reflecting the nucleation of a small cluster of oscillators that are mutually synchronized, thereby generating a collective oscillation.
- The numerics further suggest thatr∞ depends only onK, and not on the initial condition.
4. Kuramoto’s analysis
- In his earliest work, Kuramoto analyzed his model without the benefit of simulations — he guessed the correct long-term behavior of the solutions in the limitN→ ∞, using symmetry considerations and marvelous intuition.
- In contrast, the oscillators with |ωi >Kr are “drifting” — they run around the circle in a nonuniform manner, accelerating near some phases and hesitating at others, with the inherently fastest oscillators continually lapping the locked oscillators, and the slowest ones being lapped by them.
- The authors evaluate the locked contribution first.
- Thus, Kc = 2 πg(0) , which is Kuramoto’s exact formula for the critical coupling at the onset of collective synchronization.
5.1. Finite-N fluctuations
- In the last of her three Bowen lectures at Berkeley in 1986, Kopell pointed out that Kuramoto’s argument contained a few intuitive leaps that were far from obvious — in fact, they began to seem paradoxical the more one thought about them — and she wondered whether one could prove some theorems that would put the analysis on firmer footing.
- In particular, she wanted to redo the analysis rigorously for large but finiteN, and then prove a convergence result asN→ ∞. Whereas Kuramoto’s approach had relied on the assumption thatr was strictly constant, Kopell emphasized that nothing like that could be strictly true for any finiteN.
- For largeN, for most initial conditions, and for most realizations of theωi , the coherencer(t) approaches the Kuramoto valuer∞(K) and stays within O(N−1/2) of it for a large fraction of the time.
- Around the same time, Daido [30–33], and Kuramoto and Nishikawa [8,9] began exploring the finite-N fluctuations using computer simulations and physical arguments.
5.2. Stability
- The other major issue left unresolved by Kuramoto’s analysis concerns the stability of the steady solutions.
- Kuramoto was well aware of the stability problem; he writes [5] (p. 74): “One may expect that negativeµ (i.e., weaker coupling) makes the zero solution stable, and positiveµ ( . ., stronger coupling) unstable.
- Surprisingly enough, this seemingly obvious fact seems difficult to prove.
6. Stability theories of Kuramoto and Nishikawa
- Kuramoto and Nishikawa [8,9] were the first to tackle the stability problem.
- They proposed two different theories, both based on plausible physical reasoning, but neither of which ultimately turned out to be correct.
- Nevertheless, it is interesting to look back at their pioneering ideas, partly because they came tantalizingly close to the truth, and partly to remind us how subtle the stability problem appeared at the time.
6.1. First theory
- In their first approach, Kuramoto and Nishikawa [8] tried to derive an evolution equation forr(t) in closed form, a dynamical extension of the earlier self-consistency equation (4.5).
- The hope was that this might be possible close to the bifurcation, wherer(t) would be expected to evolve extremely slowly compared to the relaxation time of the individual oscillators.
- To push this strategy through, Kuramoto and Nishikawa [8] made several approximations whose validity was uncertain.
- Note the peculiar extra factor ofr on the right-hand side as compared to the usual normal form near a pitchfork bifurcation.
6.2. Second theory
- Kuramoto and Nishikawa soon realized that something was wrong.
- They now believed that the drifting oscillators arenotnegligible throughout the whole evolution of r(t) — rather, these oscillators play a decisive dynamical role in the earliest stages, thanks to their rapid response to fluctuations inr(t), though in the long run they still do not affect the steady value ofr.
- The oscillators are initially distributed according to the stationary densityρ(θ ,ω) found in Section 4, whereh0 plays the role ofr in the earlier formulas.
- Remember, the integral equation (6.2) was not derived in any systematic way from the governing equation (3.1).
7. Continuum limit of the Kuramoto model
- It was against this confusing backdrop that Mirollo and I began thinking about the stability problem.
- At the time, it was unclear how to formulate the problem mathematically.
- Sakaguchi [35] did not present a stability analysis of his model.
8. Stability of the incoherent state
- The linear stability problem for the incoherent state of Sakaguchi’s model was solved in [34].
- (8.2) Here c.c. denotes complex conjugate, andη⊥ contains the second and higher harmonics ofη.
- In summary, the linearization about the incoherent state of the Kuramoto model has a purely imaginary continuous spectrum forK<Kc, and the discrete spectrum is empty.
- AsK increases, a real eigenvalueλ emerges from the continuous spectrum and moves into the right half plane forK > Kc (Fig. 4).
9.1. The long-sought integral equation
- Fortunately it was now possible to derive such an equation systematically, as follows [37].
- More generally, Matthews, Mirollo, and I found that forK<Kc, the asymptotic behavior ofR(t) depends crucially on whetherg(ω) is supported on a finite interval [−γ ,γ ] or the whole real line (these are the only possibilities, by their hypotheses thatg is even and nowhere increasing forω > 0).
- The decay is caused by a pole in the left half plane — pole not of the integrand but of itsanalytic continuation(as required for the validity of the usual contour manipulations).
12. Epilog
- The last time I saw Crawford was in spring 1998, at the Pattern Formation meeting at the Institute for Mathematics and its Applications.
- It was his first conference after many bouts of chemotherapy, and although he was a little weak, he was all smiles and his manner was as gracious as ever.
- The authors enjoyed some fun times together that week, especially during a dinner with Mirollo.
- Over pizza and a few beers, the three of us discussed the linear stability problem for the entire branch of partially synchronized states in the Kuramoto model.
- With Crawford on their team, I bet the authors could have done it.
Did you find this useful? Give us your feedback
Citations
9,715 citations
7,665 citations
Cites background from "From Kuramoto to Crawford: explorin..."
...Twenty-five years later, the Kuramoto model continues to surprise us (see ref...
[...]
3,683 citations
3,424 citations
2,953 citations
Cites background from "From Kuramoto to Crawford: explorin..."
...One of these approaches, as already stated, considers a system made up of a huge population of weakly-coupled, nearly identical, interacting limit-cycle oscillators, where each oscillator exerts a phase dependent influence on the others and changes its rhythm according to a sensitivity function [27, 28]....
[...]
...Demanding some conditions for the stationary distribution of drifting oscillators with frequency ωi and phases θi [27], a self-consistent equation for r can be derived as...
[...]
References
6,101 citations
"From Kuramoto to Crawford: explorin..." refers background or methods in this paper
...In the 1990s, Crawford wrote a series of papers about the Kuramoto model of coupled oscillators [1–3]....
[...]
...Keywords:Kuramoto model; Coupled oscillators; Kinetic theory; Plasma physics...
[...]
...Technical, yes, but a technical tour de force....
[...]
4,525 citations
3,424 citations
3,077 citations
"From Kuramoto to Crawford: explorin..." refers background in this paper
...If the coupling strength exceeds a certain threshold, the system exhibits a phase transition: some of the oscillators spontaneously synchronize, while others remain incoherent....
[...]
2,491 citations
"From Kuramoto to Crawford: explorin..." refers methods in this paper
...Keywords:Kuramoto model; Coupled oscillators; Kinetic theory; Plasma physics...
[...]