ELM divertor heat loads in JET-ILW and full-W ASDEX Upgrade

1. ELM divertor heat loads in JET-ILW and full-W ASDEX Upgrade T.Eich, R.Scannel, B.Sieglin, G.Arnoux, S.Devaux, I.Balboa, A.Scarabosio, M.Leyland, S.Brezinsek, G.F.Matthews, S.Jachmich, H.Thomsen, A.Herrmann, P.DeMarne, M.Beurskens, W.Fundamenski, G.Huysmans PFMC Jülich Germany, 16.05.13

2. Outline • Combining type-I ELM heat load from various experimental campaigns in JET and ASDEX Upgrade (both C and W) • The story so far: Results from JET and AUG ‘carbon’ operation • Comparison of ‘W’ and ‘C’ ELM heat loads • A pedestal pressure based ELM divertor heat load scaling • Outlook & Summary and Conclusions • Not covered: Access to small ELM regimes

3. J. Linke ELMs: transient heat loads The transient heat flux factor has a simple relation to Energy (E), depositon Area (Adep) and characteristic time scale (tc): 0.5MJ/m2 heat flux factor Mitigation of transient events needs to reduce either the energy, increase the area or the characteristic time scale

4. Time scales: initial comparison of W and C • Comparison of ELM power fluxes by IR derived from CFC and W surfaces in JET-C gave fair agreement CFC #74380 Reference case • ELM outer divertor targetenergy ~ 0.35 of ELM lossenergy (same forWandC) W

5. 4401 ELMs, 25 discharges trise (μs) Measured ELM power load (JET) τ|| =Lc / cs (μs) Triniti Plasma Gun (normalized) ELM time scales in ‘Carbon’ (MW),(MW/m2) ITER assumes : ELM decay time : 500us ELM rise time: 200us This temporal shape was used for material studies (us) Temporal shape and time scales of ELM heat fluxes in JET and ITER are expected to be similar, since they scale with τII = Lc/cs

6. Relaxation of a Maxwell distribution ASDEX Upgrade • ELM energy release time into the SOL, τMHD << τII • ELM duration time x 2.4 ELM rise time W.Fundamenski, PPCF 2006 A.Kirk, PPCF 2006 T.Eich at al, JNM 2009

7. ELM time scales by AXUV studies • Neartarget ELM inducedradation (lowdensity) shows also fair agreementandis in linewithMaxwellianvelocity full-W ASDEX Upgrade operation

8. ELM heat loads in ripple experiments JET-C dBT=0.08% dBT=0.5% dBT=0.75% Natural ELMs 22Hz 30Hz 50Hz Vertical ‘kicks’ • For TF ripple studies, an increase of ELM frequencies is found (22Hz, 30Hz, 52Hz) • ELM peak heat fluxes are not reduced EELM(kJ)

9. Divertor peak heat flux vs EELM peak heat flux (MWm-2) B=2.2 T, 2 MA, (q95~3.6) δav=0.45, PNBI=10-14 MW JET: 2.5MA/2.5T ELM wetted area increases with ELM loss energy Wetted area (m2) ELM frequency (Hz) EELM(kJ)

10. ELM wetted area • Observedtrend: ELM wettedareaincreaseswith ELM losssize, resultseen in JET, DIII-D and ASDEX Upgrade JET λq,ELM = 20mm JET-ILW DIII-D λq,ELM = 5mm M.Jakubowski et al, Nucl.Fusion (2009) H.Thomsen et al, NF (2011) ITER: For minimum sized ELMs broadening (λq,ELM =5mm)

11. ELM ergodisation & filaments: complex deposition pattern • NoobviousdifferencesfoundbetweenWandCoperationw.r.t. ELM ergodisationorfilamentarysubstructure • However, detailedstudiesare in progress,aimingat • (quasi-) toroidal modenumbers • energydistributionbeweetn (radiallymoving) filamentsand parallel losses due toergodizationoffieldlines Δt = -215 µs Δt = -129 µs Δt = -43 µs Δt = 43 µs Δt = 129 µs Δt = 215 µs JOREK (ITER) JET-C 4.0MJ (cond.) 1.6MJ (conv.)

12. Tolerable ELM size in ITER • AELM = 2 π Rdiv * λELM * fx = 0.90m2 (λELM=5mm, fx=6.5, Rdiv,inner=4.4m) • No Radiation, 100% in deposited in divertor, In/Out Asymmetry 2:1 (favouring the inner) • Result: Etol = 0.5MJ/m2 * 1.5 * 0.90m2 = 0.7MJ As AELM increases with ELM loss, how scales effectively the ELM energy density?

13. Comparing W- and C- ELMs NB: Experimental execution of discharges of JET-ILW and JET-C (slightly) different for dedicated ELM heat load studies • Shaping, Triangularity, strike lines on target: identical • Larger NBI heating power required for similar pedestal ne, Te • Bt / Ip scan executed first at identical PNBI for C/W and with much increased PNBI for W (up to 26MW) • H98y about 1 for best discharges in JET-ILW PNBI>20MW PNBI<13MW

14. W versus C: time scales (1) • Type-I ELMs in JET-ILW do not follow the simple scaling found for JET-C • Interpretation: ELM energy release time (ELM MHD time?) is larger than parallel transport time JET-ILW JET-C • τMHD << τII ? • τMHD >= τII

15. W versus C: time scales (2) • Also the temporal shape of the ELM power fluxes approaches the shape observed for type-I ELMs in JET-C when τIIis shortest in DB • Reminder: τII ≈ Te-0.5 • Same holds true for type-I ELMs full-W ASDEX Upgrade operation with good confinement (e.g. with N2 seeding) JET-ILW JET-C

16. example plots A couple of examples showing the increase of the ELM power deposition length, W-ELMs appear to be stretched for low Te full-W ASDEX Upgrade JET-ILW (2.0 & 2.4MA) / 2.5T PNI= 9MW, Te,ped = 480 eV PNI=10MW, Te,ped = 600 eV PNI=22MW, Te,ped = 1100eV Plot needed JPN 82630 JPN 83438 JPN 82644

17. H98y vs time scales • In summary we find, for good confinement conditions (at high pedestal temperature) there is almost no difference between W-ELMs and C-Elms w.r.t time scales • Such ‘good’ conditions are achieved in JET-ILW at higher heating power or with e.g. N2 seeding • (In line with ASDEX Upgrade N2 seeded experiments (t.b.c.))

18. A note on ‘long’ ELMs • ‘Longish’ time scales for pedestal collapse are observed e.g. by Thomson-Scattering or ECE measurements confirming previous assumption of τMHD >= τII • Such conditions were rarely observed in JET-C, but e.g. in Helium discharges • For conditions with ‘good’ confinement and high pedestal temperatures, short time scales are recovered (for AUG, PC: A.Burckhardt) Private communication, JET: L.Frassinetti, D.Dodt & AUG: A.Burckhardt

19. Duration of T e,ped drop: 0.3-3ms ILW (no seeding) ILW with N2 seeding JET-C at rpol=0.9 at rpol=0.9 at rpol=0.9 82537 82817 79501 Dt≈1ms Dt≈2-3ms Dt≈0.3-0.5ms Dt≈10ms • Behaviour relatively similar for each ELM • Time interval to reach the minimum: Dt≈1ms • Two different behaviours • Time interval to reach the • minimum: Dtfast≈2-3ms • Dtslow≈10ms • Behaviour relatively similar for each ELM • Time interval to reach the • minimum: Dt≈0.3-0.5ms Courtesy of L.Frassinetti

20. TIME SCALES: DISTRIBUTIONS • CFC and ILW (not seeded) have clearly two different time scales • CFC and seeded ILW are comparable if the stored energy is similar • Seeded ILW are comparable to not seeded ILW if the stored energy is low • The ELM time scale seems to be related more to the stored energy than to the wall CFC Ip2.5MA only ILW (not seeded) Wth (MJ) ILW with N2 JET-C ILW w/o N2 ILW (seeded) with high Wth • Shots considered: Ip=2.5MA • and Pnet≈15-19MW • 5 CFC plasmas • 6 ILW plasmas (not seeded) • 11 ILW shots (with N2 seeding) • - 6 with Wth comparable to CFC • - 5 with low Wth ILW (seeded) with low Wth

21. Definition of ELM energyfluency • The ELM energy fluency is the peak of the time integrated heat flux profile (energy / area) εmax Typical numbers for large ELMs at JET: 150 kJ/m2 inter-ELM forreference (5ms)

22. W/C: ELM energy fluency A jump ahead: Attempt to scale or order the ELM energy fluency NB: Data are mapped to parallel field lines in order to compare the different divertor geometries JET-C :εtarget x 20 = εII JET-ILW:εtarget x 12 = εII Important conclusion: Though distributed on a longer time scale, deposited energy / area is the same (!)

23. W/C: ELM energy fluency EELM/Wplasma (%) Regression result for JET-C and JET-ILW ELM energy fluency (combined DB) Worth notifying: Very weak dependency on the relative ELM loss size (!) v*

24. Only JET-ILW data Result for JET-ILW only: Almost linear to the pedestal pressure B.Sieglin

25. Outlook • The ultimate goal of this study is to provide a Multi-Machine Scaling for ELM energy fluency and the power deposition ELM time scalesby combining JET, DIII-D and ASDEX Upgrade divertor ELM heat load data • The next step is the inclusion of ASDEX Upgrade data and to provide (i) major R scaling (ii) extrapolation to ITER and (iii) case to compare with ELM models and ELM modelling (JOREK) • For this endeavour we have run test pulses in GLADIS with JET lamellae, W-coated CFC target and AUG Div-III solid W target plates, in order to cross check JET and ASDEX Upgrade heat load data • Latter experiments in GLADIS are presented in the poster of Bernd Böswirth (Date, Poster ID)

26. Summary & Conclusions • Pedestal top pressure and temperature is reduced for the reference pulses with same Ip / Btor and heating power in JET-ILW • At identical pedestal top densities and temperatures, ELM heat load time scales in JET-ILW and full-W AUG w.r.t ‘carbon’ similar • ELM peak energy fluency (J/m2) for JET-C and JET-ILW at given pedestal top pressure is very similar • Simple regression reveals weak dependency of divertor peak energy fluency on relative ELM loss for JET data base • Latter explains the observed absence of a mitigation of divertor peak heat fluxes with increased ELM frequencies at constant pressure, e.g. by kicks, in ripple discharges, ELM pellet pacing or simple gas puffing • However, exceptions are e.g. B-coils in AUG or pellets in DIII-D

27. Back Up (GLADIS) • Results from GLADIS (B.Böswirth & B.Sieglin) α=333 kW/K/m2

28. EFCC & kicks for ELM mitigation kicks 6 • fELMs up ~3 (in this example) • ELM size reduced ΔWELM by a factor of ~2.5 EFCCs 5 ne,ped(1019 m-3) 4 1.4 1.2 1.0 Te,ped(keV) 5 4 Wthermal(MJ) DW~10% 3 • δav=0.45, 2.2T/2.0MA (q95=3.6) • Moderate reduction in Wth<10% • ne reduction (edge & core) ~ 30% : slightly higher for kicks (higher fELM) • Te,ped, (and Ti,ped) up by ~ 25% Details: 30 20 wrot(krad/sec) (pedestal) 10 0 80 ~15 Hz 40 45 Hz+/-0.4 0 80 40 40 Hz+/-15 0 14 16 18 20 22 Time (sec) Courtesy of E. de la Luna

29. εELM versus pped,e Assessing ELM mitigation techniques:

30. To Do, Improvements Suttrop: AUG B-coils Jachmich: EFCC Results Power versus time for AUG

31. ELM Heat Load • Good agreement of ILW data with free streaming approach • No dependence on relative ELM size