What the paper is about: The authors use three‑dimensional computer simulations to study the L–H transition in tokamaks. The L–H transition is the sudden improvement of confinement at the plasma edge that fusion researchers use to reduce heat losses. The simulations show that electromagnetic turbulence in the edge can spontaneously produce a sheared flow that cuts off turbulent transport. They find that this transition happens at lower heating power when the toroidal magnetic field points in the so‑called “favourable” direction, and they trace the asymmetry to finite collisionality breaking time‑reversal symmetry.

What the researchers did: The team ran flux‑driven two‑fluid simulations with the GBS code in a diverted tokamak geometry. The code solves a reduced set of fluid equations (drift‑reduced Braginskii) for the plasma edge. Typical parameters at the separatrix were R/ρs = 500, a/R ≃ 0.3, mi/me = 800, Z = 1 and an electron beta β = 10−4, with a collisionality parameter ν = 5×10−2 cs/R. Starting from a steady L‑mode state, they raised the core heating power by about 40% and watched the edge. In the favourable toroidal‑field case the energy confinement time τE rose by a factor of about two on a short time scale (~5 R/cs) and then settled at roughly 1.5 times its initial value. The same heating did not trigger the transition in the unfavourable case unless the input power was raised further.

How it works at a high level: In the simulations the edge turbulence is driven by resistive ballooning and electron drift‑wave modes. Electromagnetic effects (finite β, the ratio of plasma pressure to magnetic pressure) are essential: when β is set to zero the transition does not occur. Two linked mechanisms operate. First, increasing β changes the linear response of electrons and tends to stabilize drift waves. Second, finite‑β drift waves nonlinearly generate a mean E×B shear flow (the flow produced by the electric field across the magnetic field). That shear reduces turbulence and so reduces transport, producing an edge pressure pedestal and a barrier to heat loss. The simulations also show that removing the zonal (flux‑surface averaged) E×B component immediately restores high transport, confirming that the mean flow is the key player.