Above a critical coupling K, independent oscillators spontaneously synchronise — a second-order phase transition from chaos to coherence.
In 1975, Japanese physicist Yoshiki Kuramoto simplified Winfree's (1967) model of biological oscillators into an exactly solvable mean-field system. The model describes N oscillators, each with a natural frequency ωᵢ drawn from a probability distribution g(ω), coupled through the phase differences of neighbours.
dθᵢ/dt = ωᵢ + (K/N)·Σⱼ sin(θⱼ − θᵢ)r·e^{iψ} = (1/N)·Σⱼ e^{iθⱼ}. When r≈0 phases are incoherent; r≈1 means full synchrony. ψ is the mean phase.
For a Lorentzian g(ω)=γ/π(ω²+γ²), the critical value is Kc=2γ. For Gaussian g, Kc=2/[π·g(0)].
For K>Kc, a bifurcation gives r≈√(8(K−Kc)/Kc·π·g(0)) — a continuous transition with β=1/2.
Using the order parameter, each oscillator couples to the collective mean field: dθᵢ/dt = ωᵢ + K·r·sin(ψ−θᵢ).
Kuramoto solved the N→∞ limit exactly using the self-consistency equation. In steady state with r constant, oscillators with |ωᵢ − Ω| ≤ Kr are locked to the mean field at frequency Ω; those outside this band remain drifting.
r = K·∫_{−Kr}^{Kr} cos(θ*)·g(Ω+Kr·sin θ*)·Kr·cos θ* dθ*1 = K·∫_{−π/2}^{π/2} cos²(θ)·g(Kr·sin θ) dθ,
with the trivial solution r=0 always present and a non-trivial r>0 branch appearing at K=Kc.
The density of locked oscillators grows as K exceeds Kc, contributing a coherent component to r. The Ott–Antonsen ansatz (2008) showed the dynamics reduce exactly to two real ODEs for r and ψ when g(ω) is a rational function, making the Kuramoto model analytically tractable far from equilibrium.
| Preset | N | K | ω distribution | Expected behaviour |
|---|---|---|---|---|
| 🌀 Subcritical | 80 | 0.5 | Gaussian σ=1.0 | r≈0; incoherent drifting phases |
| ⚡ Near Critical | 80 | 1.6 | Gaussian σ=1.0 | r fluctuates around Kc; partial locking |
| ✨ Synchronized | 80 | 4.0 | Gaussian σ=1.0 | r→0.8–0.95; majority phase-locked |
| 〰 Bimodal ω | 100 | 3.0 | Bimodal ±1.5 | Two clusters; harder to synchronise |
| 🔵 Large Network | 200 | 2.5 | Gaussian σ=1.0 | Large-N limit; sharper transition |
| 💫 Pulse Reset | 60 | 5.0 | Tight σ=0.2 | Strong coupling; oscillator flashing |
| Phenomenon | Condition | Description |
|---|---|---|
| Incoherence | K < Kc | Phase distribution remains uniform on [0,2π]; r≈0 |
| Partial synchrony | K ≈ Kc | A locked cluster coexists with drifting oscillators; r intermediate |
| Full synchrony | K ≫ Kc | Nearly all oscillators lock; r→1; coherent mean field dominates |
| Chimera states | Non-uniform coupling | Coexisting coherent and incoherent domains; found in 2D networks |
| Travelling waves | Bimodal g(ω) | Two clusters counter-rotate; beating oscillations in r(t) |
| Explosive sync | Correlated K−ω | Discontinuous (first-order) transition if large K assigned to high-ω oscillators |
| Domain | System | Role of synchronisation |
|---|---|---|
| Neuroscience | Gamma-band oscillations (40 Hz) | Cortical binding; attention; epileptic seizures when over-synchronised |
| Biology | Firefly flash synchrony | Males synchronise flashes to attract females; Southeast Asian species Pteroptyx malaccae |
| Cardiology | Sinoatrial node pacemaker cells | ~10 000 cells synchronise to drive heartbeat at consistent rate |
| Chronobiology | Suprachiasmatic nucleus (SCN) | ~20 000 neurons synchronise circadian rhythms; jet lag is de-synchronisation |
| Engineering | Power grid frequency (50/60 Hz) | Generators must maintain phase lock; loss of synchrony causes blackouts |
| Physics | Josephson junction arrays | Superconducting junctions synchronise to emit coherent microwave radiation |
| Model | Agents | Interaction | Order parameter |
|---|---|---|---|
| Kuramoto (1975) | Oscillators on ℝ | Phase difference sin(θⱼ−θᵢ) | r = |⟨e^{iθ}⟩| |
| Vicsek (1995) | Moving particles in 2D | Average velocity direction | φ = |⟨v/|v|⟩| |
| Boids Reynolds (1987) | 2D agents with positions | Sep/Ali/Coh forces | Polarisation P |
| Ising model (1925) | Lattice spins ±1 | Ferromagnetic coupling J | Magnetisation m |
| Level | Key concept |
|---|---|
| A-level / Pre-university | Simple harmonic oscillators; phase; coupled pendula |
| Undergraduate Physics | Nonlinear dynamics; mean-field theory; bifurcations; statistical mechanics analogy |
| Graduate / Research | Ott–Antonsen reduction; finite-size fluctuations; chimera states; network topology effects |
| Interdisciplinary | Systems biology; neuroscience; power systems; complex network science |
Finite-size fluctuations prevent exact r=1. In the N→∞ limit, r→1 only if all ωᵢ are identical (δ-distribution). For any distribution with finite spread, a fraction of drifting oscillators always exists at finite K, keeping r < 1. Increasing N sharpens the transition and reduces fluctuations, but the asymptotic r∞(K) < 1 for finite spread remains.
When oscillators initialise with uniformly random phases, they sample the circle uniformly — the Kuramoto model's incoherent state corresponds to a uniform distribution on S¹. As coupling increases and a cluster forms, the phases of locked oscillators converge toward ψ while drifting oscillators spread uniformly. The visual ring arrangement helps distinguish these two populations: the dense arc is the locked cluster.
A bimodal g(ω) = [δ(ω−ω₀) + δ(ω+ω₀)]/2 creates oscillators in two groups rotating in opposite directions. They form two competing clusters. For small |ω₀| the transition is still continuous; at a tricritical point (|ω₀| = √3/4·Kc) the transition becomes discontinuous — an explosive first-order synchronisation where r jumps from 0 to a finite value.
From Boids flocking to the Ising phase transition — browse 47+ interactive simulations covering physics, maths, and beyond.
All Simulations →