A future fusion reactor is expected to have all-metal plasma facing materials (PFM) to ensure low erosion rates, low tritium retention and stability against high neutron fluences. As a consequence, intrinsic radiation losses in the plasma edge and divertor are low in comparison to devices with carbon PFMs. To avoid localized overheating in the divertor, intrinsic low-Z and medium-Z impurities have to be inserted into the plasma to convert a major part of the power flux into radiation and to facilitate partial divertor detachment. For burning plasma conditions in ITER, which operates not far above the L-H threshold power, a high divertor radiation level will be mandatory to avoid thermal overload of divertor components. Moreover, in a prototype reactor, DEMO, a high main plasma radiation level will be required in addition for dissipation of the much higher alpha heating power. For divertor plasma conditions in present day tokamaks and in ITER, nitrogen appears most suitable regarding its radiative characteristics. If elevated main chamber radiation is desired as well, argon is the best candidate for simulataneous enhancement of core and divertor radiation, provided sufficient divertor compression can be obtained. The parameter P sep /R, the power flux through the separatrix normalized by the major radius, is suggested as a suitable scaling (for a given electron density) for the extrapolation of present day divertor conditions to larger devices. The scaling for main chamber radiation from small to large devices has a higher, more favourable dependence of about P rad,main /R 2 . Krypton provides the smallest fuel dilution for DEMO conditions, but has a more centrally peaked radiation profile compared to argon. For investigation of the different effects of main chamber and divertor radiation and for optimization of their distribution, a double radiative feedback system has been implemented in ASDEX Upgrade. About half the ITER/DEMO values of P sep /R have been achieved so far, and close to DEMO values of P rad,main /R 2 , albeit at lower P sep /R. Further increase of this parameter may be achieved by increase of the neutral pressure or improved divertor geometry.

Abstract. Detachment of high power discharges is obtained in ASDEX Upgrade by simultaneous feedback control of core radiation and divertor radiation or thermoelectric currents by the injection of radiating impurities. So far 2/3 of the ITER normalized heatflux P sep /R= 15 MW/m has been obtained in ASDEX Upgrade under partially detached conditions with a peak target heatflux well below 10 MW/m 2 . When the detachment is further pronounced towards lower peak heatflux at the target, substantial changes in ELM behaviour, density and radiation distribution occur. The time-averaged peak heat flux at both divertor targets can be reduced below 2 MW/m 2 , which offers an attractive DEMO divertor scenario with potential for simpler and cheaper technical solutions. Generally, pronounced detachment leads to a pedestal and core density rise by about 20-40 %, moderate ( 20 %) confinement degradation and a reduction of ELM size. For AUG conditions, some operational challenges occur, like the density cut-off limit for X-2 ECRH heating, which is used for central tungsten control. Partial detachment of high power discharges in ASDEX Upgrade2

Abstract. ELM triggering and pacing in an all-metal wall environment shows significant differences to a firstwall configuration containing carbon. Here we report on experiments performed at ASDEX Upgrade revisiting the issue with all plasma-facing surfaces now fully replaced by tungsten. This investigation was motivated by experimental findings indicating that ELM triggering becomes more intricate when the carbon is replaced by a metal wall. ELM pacing could no longer be achieved by magnetic triggering in ASDEX Upgrade under conditions that previously showed a positive response. Also, recent investigations at JET indicate that a lag time occurs in pellet ELM triggering when operating with the new ITER-like wall. The ASDEX Upgrade centrifugebased launching system was revitalized and upgraded for this study, now allowing detailed analysis of the ELM trigger response. The appearance of a lag time for pellet ELM triggering in an all-metal wall environment was confirmed. While different lag time durations were found for several type-I ELMy H-mode scenarios, the magnitude of the pellet perturbation was found to cause no difference. Reducing the auxiliary heating power for ELM triggering clearly makes the pellet tool less efficient for ELM control purposes; however, this affords a major benefit when applied for fuelling. Plasma operation with benign ELM behaviour at core densities far beyond the Greenwald limit was demonstrated, this being fully reversible and not affecting the energy confinement.