Improvement of SPIDER tomographic diagnostic

The Neutral Beam Injection (NBI) consists in firing a high-energy beam of neutral particles inside the fusion plasma in a tokamak for fuel heating and fusion reaction triggering. One of such tokamak machines is ITER. For successful ITER operation, the beam must satisfy strict specifications, for example in terms of ionic current throughput (40 A at 1 MV), homogeneity (more than 90%) and divergence (less than 7 mrad). To reach these requirements, in Padua, at Consorzio RFX, two experiments are hosted: SPIDER and MITICA. SPIDER is the full- size prototype of ITER NBI negative ion source. This work focuses on one of its diagnostics, namely the visible tomography of the beam which will be installed in the future also on MITICA, ITER’s NBI full-scale prototype. Visible tomography is a non-invasive diagnostic which uses two-dimensional visible cameras collecting the light emitted from the ion beam-background interactions. The camera signal is an integrated measure of the two-dimensional beam emissivity along different Lines-Of-Signt (LOSs), which can be reconstructed using a suitable inversion algorithm. Therefore, the reconstructed emissivity can be used to characterise the beam divergence and homogeneity and, using suitable spectroscopic models, allows to estimate the beam current density. This work aims at improving the current SPIDER tomography by introducing two-dimensional LOSs in the reconstruction algorithm, to better account for the geometry of the diagnostic, and by further developing a model for the beam emission (by introducing new reactions and accounting for the effect of secondary electrons on the beam light), to better interpret the reconstructed emissivity in terms of negative ion current density. Testing on experimental data shows good agreement between the previous reconstruction results and the improved 2D-LOS one. Further testing on full-beam simulations shows that the algorithm performance is not affected by the beam features (e.g. beamlet width and uniformity) and the reconstruction error in ideal conditions (no light background and no signal noise) remains around 10% at 5-beamlet resolution or lower for lower resolutions (i.e. 10, 20, 40 or 80-beamlet resolution), showing the possibility for successful application of such upgrades with a sufficient level of detail using two-dimensional LOSs. Light background is shown to impact the most the reconstruction accuracy of the beam, up to 30% in the worst case of 5% background intensity (compared to the nominal beamlet luminosity). The beam emissivity as a function of the beam energy is assessed, showing a reduction as the beam energy increases. It also demonstrates that the single stripping dominates the beam emission at all energies, increasing as the beam energy does. At SPIDER’s nominal acceleration of 100 keV, single stripping processes account for 87.7% of the total emissivity, followed by excitation (5.7%) and secondary electrons (4.1%), also representing a possible cause of light background. Cameras are calibrated using a calibrated source and a Hα filter in order to obtain an equivalent Hα source whose emissivity is known. Setting the camera in front of the equivalent source allows to link the signal, collected by the completely illuminated pixels, to the emissivity, obtaining a calibration constant which is used to convert the integrated camera counts into radiant power integrals. The reconstructed emissivity allows, using the results from the beam model, to obtain, for the first time, the 2D pattern of the beam current density from its emissivity, which matches the same order of magnitude of the direct electrical measurements of the STRIKE calorimeter and the Beam Current Monitor.