Phantom Traffic Jams: The Physics of Congestion Waves

A jam with no cause

A phantom traffic jam — also called a ghost jam or jamiton — is a region of dense, slow-moving or stopped traffic with no physical bottleneck: no accident, no on-ramp merge, no lane closure. It forms spontaneously once traffic density crosses a critical threshold and any small perturbation ripples through the stream. Japanese researchers demonstrated this cleanly in 2008: 22 cars were placed on a circular track and told to hold a constant, even spacing. Within minutes, small random variations in speed cascaded into a stop-and-go wave that circled the track indefinitely — and crucially, the wave moved backwards, opposite the direction of traffic flow, at roughly 15 km/h regardless of how fast the cars were actually driving.

Traffic as a compressible fluid

Traffic engineers describe flow with two quantities: flow q (vehicles per hour past a point) and density k (vehicles per kilometre), related by q = k·v. At low density drivers travel fast; as density rises, speed drops, producing a hump-shaped fundamental diagram that peaks at a critical density k_crit. Past that point the road is in the congested regime, and the Lighthill–Whitham–Richards (LWR) model treats it as a compressible fluid obeying a conservation law:

∂k/∂t + ∂q/∂x = 0,   q = k·V(k)
Greenshields:  V(k) = v_max · (1 − k/k_jam)
wave speed c = dq/dk = V + k·V'  → negative when k > k_crit

A negative wave speed means the disturbance itself propagates upstream — against the direction cars are moving. When you're stuck in a phantom jam, you're driving through a wave that is stationary or even drifting backwards relative to the road, not one being carried along by it.

The Nagel–Schreckenberg cellular automaton

The simplest microscopic traffic model is the Nagel–Schreckenberg (NaSch) model (1992). The road is a ring of cells; each cell holds at most one car with integer speed v ∈ {0, …, v_max}. Every step applies four rules in parallel across all cars: accelerate toward v_max, brake to whatever gap is open ahead, randomly lose one unit of speed with probability p (modelling human imperfection), then move forward by the new speed. That single randomisation step is enough: even with p ≈ 0.3 and moderate density, the model spontaneously produces stop-and-go waves that are statistically indistinguishable from real motorway data — a striking example of emergent behaviour from simple deterministic rules plus noise.

Jamitons — self-sustaining traffic waves

In 2009, MIT researchers (Flynn, Kasimov, Nave, Rosales, Seibold) proved analytically that traffic flow equations admit exact travelling-wave solutions they named jamitons, by analogy with solitons. A jamiton is a self-sustaining density pulse: vehicles enter the back, slow or stop, then re-accelerate out of the front, with the pulse holding its shape indefinitely whenever density is above k_crit and drivers have a finite reaction time τ > 0. Real human reaction time is 0.5–1.5 s; ordinary cruise control is 0.1–0.3 s; cooperative adaptive cruise control (vehicles communicating directly) can push τ below 0.05 s — a regime that completely suppresses jamiton formation. The single most effective thing an individual driver can do is counter-intuitive: leave a larger gap ahead, which absorbs small braking pulses before they amplify into a jamiton.

Frequently asked questions

Try it live

Everything above runs in your browser — open Traffic Jams (NaSch), place cars on a ring road, and adjust density and randomisation p to watch phantom jams emerge from nothing, then track the backward-propagating jam fronts in the live space-time diagram.