Chimera States

What you're seeing

Two crowds of identical clocks. Every oscillator is just a phase θ running around the circle, and — crucially — every oscillator has the same natural frequency and obeys the same coupling rule. The two groups are defined symmetrically: there is no built-in difference between A and B, no "leader" group and no "follower" group. The natural expectation, then, is that the two crowds should do the same thing — either both synchronize or both stay in disarray.

Instead the population splits. One group condenses into near-perfect synchrony: all its dots pile onto a single point on the ring and its arrow stretches to the rim (r ≈ 1). The other group refuses to lock: its dots stay smeared around the circle, its arrow short and restless, its order parameter fluctuating somewhere well below one. And the split persists. Neither the rules nor the groups single out which crowd plays which role — the system simply falls into the asymmetry on its own. A chimera state is exactly this: coexisting coherence and incoherence among oscillators that have every right to behave identically.

The rule

Each oscillator nudges its phase toward the others through the sine of their phase difference — the same ingredient as in the Atlas's Kuramoto model. Two things make a chimera possible. First, the oscillators sit in two groups, and the coupling is a little stronger within a group (strength μ) than across to the other group (strength ν), with μ > ν. Second, the coupling carries a small phase lag β, just under π/2. Writing the phase of oscillator i in group σ as θ^σ_i (with all natural frequencies equal and set to zero, i.e. working in the rotating frame):

dθ^σ_i/dt = − Σ_σ' (K_σσ'/N) Σ_j sin(θ^σ_i − θ^σ'_j + β),   K_AA = K_BB = μ,  K_AB = K_BA = ν.

As in Kuramoto, each group's all-to-all sum collapses into a single mean-field pull through its complex order parameter r_σ e^iψ_σ = (1/N) Σ_j e^iθσ_j, so an oscillator only has to react to two averages — its own group's and the other's:

dθ^σ_i/dt = −[ μ·r_σ·sin(θ^σ_i − ψ_σ + β) + ν·r_σ'·sin(θ^σ_i − ψ_σ' + β) ]

That is the entire model. There is no asymmetry anywhere in it — and yet, for the right β and the right coupling gap μ − ν, the steady behaviour is asymmetric. The simulation integrates these equations with a fourth-order Runge–Kutta step (forward Euler drifts on this system); the controls expose β, the asymmetry A = μ − ν (with μ + ν held at 1), and the number of oscillators per group.

Why it matters

The chimera state was a genuine surprise. For symmetric, identically-coupled oscillators, the textbook outcomes are full synchrony or full incoherence; a stable mixture of the two seemed to have no right to exist. Kuramoto and Battogtokh found it anyway in 2002, in a ring of oscillators with a nonlocal coupling kernel, and Abrams and Strogatz named it in 2004 after the Chimera of Greek myth — a single creature stitched together from a lion, a goat, and a serpent's tail, here a single state stitched together from order and disorder. The two-population version simulated here, due to Abrams, Mirollo, Strogatz and Wiley (2008), is the cleanest case: simple enough that the chimera's existence and stability can be worked out analytically.

What makes it conceptually important is that it is a form of spontaneous symmetry breaking. The equations are perfectly symmetric under swapping A and B, so for every chimera with A synchronized there is a mirror-image chimera with B synchronized — and the system simply falls into one of them, decided by where it happens to start. Hit randomize and it is close to a coin flip which group wins the coherent role. This is the same logic by which a pencil balanced on its tip must fall some way even though no direction is preferred. Contrast the plain Kuramoto model, where global coupling drives the whole crowd to a single shared fate; here the structured coupling lets one shared fate split into two.

In the wild

Chimeras began as a theoretical curiosity and were later produced deliberately in the lab. They have been realized in coupled-oscillator experiments of several kinds:

• Mechanical metronomes. Martens and colleagues (2013) coupled two swinging platforms of metronomes — strong coupling within each platform, weak coupling between — and saw one platform synchronize while the other stayed incoherent: a mechanical realization of exactly this two-population model.
• Chemical oscillators. Networks of Belousov–Zhabotinsky reaction droplets and electrochemical oscillators have been arranged so that part of the population locks while part drifts.
• Optical and electronic arrays. Coupled lasers, optical light-modulator feedback loops, and electronic circuit arrays have all hosted chimera-like states.
• A suggested link to brains. Some animals — dolphins and certain birds — show unihemispheric sleep: one brain hemisphere falls into slow-wave, synchronized sleep while the other stays awake and active. The coexistence of a synchronized and a desynchronized region has made chimeras a tempting analogy, and they have also been invoked in discussions of pathological synchrony in epilepsy.

Honesty matters here. In a finite system like this one, chimeras can be long-lived transients rather than truly eternal states — run small enough or long enough and the incoherent group may eventually collapse into sync. The two-population model is an idealization (all-to-all coupling within and between groups, a single phase lag, identical oscillators). And the brain and unihemispheric-sleep connections are suggestive analogies, not established mechanism: real neural tissue is not a population of identical phase oscillators, and no one has shown that a literal chimera state implements unihemispheric sleep. What the model genuinely establishes is the surprising fact itself — that symmetry and identical coupling do not force identical behaviour.

Try this

• Just watch the two corner readouts: |r_A| and |r_B| split apart and stay split. One climbs to ≈ 1 and pins there; the other settles into a restless value around 0.6–0.8 and never catches up. That persistent gap is the chimera.
• Slide the phase lag β toward π/2 for the sharpest chimera — the coherent group locks hardest and the incoherent group stays most scattered.
• Now slide β well away from π/2 (or push the asymmetry up): the split closes and both groups fall into the same behaviour — typically both lock. The chimera lives only in a window of these parameters.
• Hit randomize a few times and watch which group wins the coherent role. It's essentially a coin flip — the clearest sign that the rules don't prefer either group; the initial conditions break the tie.

References

1. Kuramoto, Y. & Battogtokh, D. (2002). "Coexistence of coherence and incoherence in nonlocally coupled phase oscillators." Nonlinear Phenomena in Complex Systems 5(4), 380–385.
2. Abrams, D. M. & Strogatz, S. H. (2004). "Chimera states for coupled oscillators." Physical Review Letters 93(17), 174102.
3. Abrams, D. M., Mirollo, R., Strogatz, S. H. & Wiley, D. A. (2008). "Solvable model for chimera states of coupled oscillators." Physical Review Letters 101(8), 084103. (The two-population model simulated here.)
4. Panaggio, M. J. & Abrams, D. M. (2015). "Chimera states: coexistence of coherence and incoherence in networks of coupled oscillators." Nonlinearity 28(3), R67–R87. (Review.)
5. Martens, E. A., Thutupalli, S., Fourrière, A. & Hallatschek, O. (2013). "Chimera states in mechanical oscillator networks." Proceedings of the National Academy of Sciences 110(26), 10563–10567.