The relationship between scrape-off layer filaments and density profiles in ASDEX Upgrade

This thesis reports on the experimental analysis of the turbulence statistics of the ASDEX Upgrade scrape-off layer (SOL), and the filaments that populate it, for different L and H-mode discharges using the gas puff imaging (GPI) diagnostic. A synthetic GPI diagnostic is constructed and applied to a toy model to account for the smearing of filaments caused by imperfect alignment of the lines-of-sight and the local magnetic field.

Radial increases in relative fluctuation amplitude, skewness, and kurtosis are seen due to the presence of filaments whose characteristics vary as they propagate radially. The probability distribution functions (PDFs) of filament amplitude for a given pixel measurement at the GPI detector follow Gamma distributions which become increasingly skewed and flattened with increasing radius. These are similar between most of the discharge types, lending further evidence towards the so-called ‘universality of the scrape-off layer’, but the L-mode discharge with nitrogen seeding has an increase in each metric, and the most skewed and flattened PDFs, in the far SOL compared to the other discharge types. This is though to be caused by changes to the filaments and is corroborated with independent measurement of an increase in the filament amplitudes in the 2D GPI data.

Using GPI, independent measurements of individual filament poloidal sizes and radial velocities are made, and reported with respect to the Myra filament model. Most filaments are measured with sizes above the fundamental blob size with velocities under the sheath-limited regime velocity scaling, except for SOL locations closest to the separatrix. Distribution functions of filament sizes and velocities are measured, as well as the interdependence between different filament attributes, which are used as inputs to the Garcia-Militello SOL model.

This model is based on a statistical framework that links filaments and their motion to SOL density profiles. In the single-filament model it is shown that profile decay lengths are directly proportional to the the radial filament velocity, vx, and parallel loss timescale, τ∥, the characteristic time at which filaments reduce in amplitude due to parallel losses. This model is expanded to show how radial changes of vx or τ∥ can generate profiles with either increasing or decreasing decay lengths. From simulations of this model, measurements of the higher-order statistical profiles, as well as single-point PDFs, are compared to the experimental statistical characterisation of the SOL, allowing us fully use the model to determine which change of variable may result in the change of the density profile.

For the first time, this model is extended to include the distributions of filament attributes. It is shown, both analytically and in simulations, how the amplitude distribution is required for the quadratic scaling of kurtosis with skewness, seen for Gamma distributed fluctuation signals, and how the radial velocity distributions, in addition to spatial changes in vx or τ∥, can lead to non-exponential decay profiles, as well as radial increases in higher-order statistical profiles. The model is first used in 1D and then extended to 2D, showing no change of the general results, and there is therefore no dependence of profile decay length on perpendicular filament size or velocity, even for large poloidal-to-radial velocity ratios.