High resolution Thomson scattering measurements with sub-centimetre radial resolution have been made during a sawtooth crash in a fusion plasma in MAST. As magnetic reconnection occurs, a growing magnetic island induces an increase in the temperature gradient at the island boundary layer, before a very rapid collapse of the core temperature. The increase in the local temperature gradient is sufficient to make the plasma core unstable to ideal magnetohydrodynamic instabilities, thought to be responsible for the rapidity of the collapse.PACS numbers: 52.55Fa, 52.35PyMagnetic reconnection is the phenomenon of the breaking and rejoining of magnetic field lines in a plasma. Examples of this process are solar flares in astrophysical plasmas [1,2] and the sawtooth instability in tokamak plasmas [3,4]. Whilst the sawtooth instability was first observed in 1974 [5], the process by which this periodic collapse of the core plasma temperature occurs is still only partially understood. Detailed diagnosis of the sawtooth crash has shown that the temperature profile is initially essentially axisymmetric, but is then deformed by a helical instability before a very rapid temperature collapse re-establishes an axisymmetric profile with a lower value at the magnetic axis [6,7].Tokamak plasmas are susceptible to sawtooth oscillations when the safety factor, q = rB φ /RB θ is less than unity [8], where r, R are the minor and major radii and B φ , B θ are the toroidal and poloidal magnetic fields. The helical perturbation which arises during the crash has an m = n = 1 structure, where m, n are the poloidal and toroidal periodicity of the wave. The first explanation of the periodic temperature collapses was proposed by Kadomtsev [3], who showed that in the nonlinear regime of the m = 1 mode, reconnection occurs at the separatrix on the characteristic Sweet-Parker timescale [1,2] 2 /ηc 2 is the resistive diffusion time, ρ is the mass density, η is the plasma resistivity and S = τ R /τ A is the Lundquist number. This timescale is up to two orders of magnitude too large to explain crash times in large modern-day tokamaks, where τ crash ∼ 20 − 100µs, whereas τ K = 2 − 10ms.The three principal observations which any theory must explain are (i) the rapidity of the temperature collapse, (ii) the sudden onset of the collapse and (iii) the incomplete relaxation of the current profile whereby q remains below unity whilst the temperature profile relaxes completely. The onset of the crash represents a theoretical challenge, since the incremental change in the safety factor which governs the stability of the m = n = 1 mode is unacceptably small to explain the rapid onset. Many alternative crash models have since been proposed, including resistive two-fluid MHD [9], collisionless kinetic effects [10,11], accelerated complete reconnection due to nonlinear collisionless effects [12], magnetic stochastization leading to enhanced perpendicular transport [13] and triggering of secondary instabilities [14][15][16]. Each of these models has had proponents...
