Inject methane and air into a thin circular gap between two plates. Under fuel-rich conditions, a flame forms at the edge and begins to rotate. It sustains itself without external heating, moving faster than the normal flame speed, and nobody asked it to spin.
The structure (arXiv:2603.22116) is bibrachial — a diffusion branch gliding along the cell edge and a premixed branch extending inward. The rotation emerges from a balance between local flame speed and gas velocity at the edge, stabilized by wall heat loss and flow expansion. Increase the flow rate and the single-headed flame splits into multiple heads, spacing themselves approximately equally around the circumference.
Note (2026-11-13; read against the paper's full text): the paper does not describe the change from one flame head to several as a bifurcation at a flow rate. It says the number of flame waves showed a certain level of hysteresis because it was affected by the instantaneous thermal boundary condition, that the boundary between single and multiple waves appeared fuzzy and was not recorded, and that above about 6.8 SLPM (at one condition) two or more heads are possible but depend on the time-histories of wall temperature and flow perturbations. The abstract's phrasing is that the rotation comes from a balance between local flame speed and unburned-gas velocity near the cell edge, with wall heat loss and flow expansion stabilising it, as suggested by simulations; the essay's picture of the flame moving sideways to stay ahead of a quenching zone is the essay's wording. The rotating flames were observed only under fuel-rich conditions.
The rotation frequency increases with flow rate for single-headed flames but follows a different pattern for multi-headed ones. A semi-empirical model predicts both frequency and shape from just two inputs: mass flow rate and surface temperature.
The phenomenon is self-organization from competing constraints. The flame wants to propagate outward (chemical energy). The walls want to quench it (heat loss). The fresh mixture wants to enter the gap (flow velocity). The compromise is rotation — the flame moves laterally to continuously encounter fresh mixture while staying ahead of the quenching zone.
The mode transitions are the striking part. One head to two heads isn't gradual — it's a bifurcation, a qualitative change in the flame's topology triggered by a quantitative change in flow rate. The flame discovers a new way to exist.
Micro-combustion applications care about this because small-scale flame stabilization is hard. The rotating mode is a solution the system finds for itself when the conventional solutions (stable anchored flame, uniform propagation) are excluded by the geometry.