Comparison of two regularization methods for soft x-ray tomography at Tore Supra

Abstract

Soft x-ray (SXR) emission in the range 0.1–20 keV is widely used to obtain valuable information on tokamak plasma physics, such as particle transport, magnetic configuration or magnetohydrodynamic activity. In particular, 2D tomography is the usual plasma diagnostic to access the local SXR emissivity. The tomographic inversion is traditionally performed from line-integrated measurements of two or more cameras viewing the plasma in a poloidal cross-section, like at Tore Supra (TS). Unfortunately, due to the limited number of measured projections and presence of noise, the tomographic reconstruction of SXR emissivity is a mathematical ill-posed problem. Thus, obtaining reliable results of the tomographic inversion is a very challenging task. In order to perform the reconstruction, inversion algorithms implemented in present tokamaks use a priori information as additional constraints imposed on the plasma SXR emissivity. Among several potential inversion methods, some of them have been identified as well suited to tokamak plasmas. The purpose of this work is to compare two promising inversion methods, i.e. the minimum fisher information method already used at TS and planned for WEST configuration, and the alternative 2nd order Phillips–Tikhonov regularization with smoothness constraints imposed on the second derivative norm. Respective accuracy of both reconstruction methods as well as overall robustness and computational time are studied, using several synthetic SXR emissivity profiles. Finally, a real case is studied through tomographic reconstruction from TS SXR database.