Evaluación de los resultados de la propuesta

The linear instabilities discussed in this chapter can only grow so much before they saturate. The amplitude at which they do saturate requires including the full non-

linear physics to determine3. The exact physical process of nonlinear saturation is currently an open question, with several players already known[27]. Coupling be- tween different scales via the convective derivative in the fluid conservation equations leads to energy being transferred from the linear microinstabilities that exist at a given scale to other modes, which may even be linearly stable[36]. These modes can then go on to generate further modes at different scales again, leading to a “cascade” of energy into broad range of scales, including down to the smallest scales where the fluctuations dissipate and turn into heat. This eventually leads to saturation when the nonlinear transfer of energy to the damping scales balances the release of free energy from the driving gradients. Broad fluctuation spectra are one of the characteristic features of turbulence, and this has been observed in linear machines, where it is possible to drive and observe single, coherent drift waves. For example, using parallel-current driven instabilities, researchers using linear devices[37] were able to observe the change from a narrow band of fluctuation frequencies to a broad spectrum by increasing the drive for the mode. This forms part of the experimental evidence for drift wave turbulence in tokamaks. For a more complete review, see [25, 27]. It is also possible for secondary instabilities to develop, generated by the locally increased gradients formed by the movement of plasma parcels around a drift wave[32].

Turbulence often has a threshold gradient value, which, when exceeded, leads to a drastic increase in the transport of equilibrium quantities. This increased transport then acts to reduce the gradients back down to the critical value. For this reason, profiles in tokamaks are often described as “stiff” - they become insensitive to the heating. Any increase in the heating only leads to a correspondingly increased transport. Equilibrium profiles, then, should be marginally stable - that is, they exist near the critical gradient.

Equations (3.11) and (3.12) lead to two different diffusivities. The Bohm diffusivity[25] is given by

DB =

cTe

eB, (3.13)

and the gyro-Bohm diffusivity is

DGB =ρ∗DB. (3.14)

The Bohm diffusivity is independent of the system size, and arises due to mesoscale

3

The difference between linear and nonlinear simulations generally means whether or not the particles are influenced by the perturbed electrostatic field, with nonlinear simulations including

structures, whereas the gyro-Bohm diffusivity scales with the system size and is due to turbulence at the gyroradius scale. There are a variety of mechanisms predicted from theory, or see in simulations and experiments that can explain the reason for these two different diffusivities. In certain conditions, turbulence can lead to non-diffusive, non-local transport. Avalanches are intermittent events with two fronts - a positive fluctuation of heat or particles propagates down the corresponding gradient, while a negative perturbation propagates up the gradient. Streamers are radially elongated structures, larger than the transport scale length of the driving microinstability, Both these processes can lead to increased cross-field transport as their sizes scale with the system size, but they can also both be regulated by sheared flows (see sections 3.4.1 and 3.5). Turbulence can also spread into linearly stable regions via nonlinear processes, leading to increased transport in those regions.

3.4.1 Zonal Flows

One of the most important physical processes for determining the saturation level is zonal flows[36, 38], which are axisymmetric structures, with m, n = 0, and are well-localised radially (i.e. they have largekr). They arise from the divergence of

the Reynolds stress[36], ∇hδvxδvyi, where δvx,y are fluctuating velocities in differ-

ent directions, and h i is a time-average. The Reynolds stress merely rearranges momentum, and does not act as a source or sink[34]. Alternatively, one can think of the generation of zonal flows as being due to a nonlinear transfer of energy in the Fourier domain, with large k structures driving energy in small k (specifically

k= 0).

Their symmetric structure has two important consequences. The first is that they are not subject to Landau damping[38]. Secondly, due to their flows being perpendicular to the temperature and density gradients, they cannot access the free energy locked up in the gradients[36]. Their structure also means that they can suppress turbulence due to their shear flows (see section 3.5).

Zonal flows are linearly stable, therefore they have to be driven solely by nonlinear turbulent processes[27]. The effect of this is that they are a form of self- regulation for the turbulence, as they both regulate the level of turbulent transport and die away without the background turbulence. At low collisionality, therefore, they can be a significant yet harmless store of free energy.

3.4.2 Turbulence diagnostics

The diagnosis of turbulence in tokamak plasmas is a complicated and tricky practice, but there are several different techniques available to experimentalists (a review of the techniques discussed here, as well as several others may be found in [27]). One of the earliest methods, Langmuir probes, used on the first plasma devices is still in everyday use on state-of-the-art machines. These measureδn, δφ, and δTe, and the

correlations between the fluctuating quantities by physically probing the plasma with a biased electrode. The particle, heat and momentum fluxes can all be de- duced from these measurements. However, as Langmuir probes require a physical intrusion into the plasma, they are only suitable for use in low-temperature, low- density regions of the tokamak (i.e. outside the separatrix), lest they be vapourised. Short-wavelength lasers may be used to measureδn, as the light will be scattered via Thomson scattering due to the change in refractive index caused by the density perturbations. By measuring at different scattered angles, the wavenumber spec- trum may be obtained, as the scattered angle is determined by the wavelength of the fluctuations. Microwaves can also be used, in a technique called reflectometry, whereby microwave radiation is bounced off the cutoff layer in the plasma. This gives the radial location of a particular density. There are various different sub- classes of reflectometry, that can be used to get a wide variety of information - from

ω, kspectra of the fluctuating density to 2Dimaging of the turbulence and poloidal velocity measurements.

The above diagnostics are all active methods - passive methods are also pos- sible. For example, beam emission spectroscopy (BES) measures the light emitted from collisional processes due to the injection of neutral heating beams[39]. This will be expanded upon in section 6.5, where we discuss how to directly compare simulation and experiment.