1. 1 ELM triggering by deuterium pellets G. Kocsis 1) Acknowledgement: A. Alonso 2), L.R. Baylor 3), S. Kálvin 1), K. Lackner 5), P.T. Lang 5), A. Alonso 2), L.R. Baylor 3), G. Huysmans 4), S. Kálvin 1), K. Lackner 5), P.T. Lang 5), J. Neuhauser 5), M. Maraschek 5), B. Pegourie 6), G. Pokol 7), T. Szepesi 1), 1) KFKI RMKI, EURATOM Association, P.O.Box 49, H-1525 Budapest-114, Hungary 2) Laboratorio Nacional de Fusion, Euratom-CIEMAT, 28040 Madrid, Spain 3) Oak Ridge National Laboratory, Oak Ridge, TN, USA 4) Association Euratom-CEA Cadarache 5) MPI für Plasmaphysik, EURATOM Association, Boltzmannstrasse 2., D-85748 Garching, Germany 6) CEA, IRFM, F-13108 Saint-Paul-lez-Durance, France 7) BME NTI, EURATOM Association, P.O.Box 91, H-1521 Budapest, Hungary

2. 2 Outline Introduction: ELM problem, pellet pacemaking Summary and outlook Poloidal scan of pellet triggering capabilityRadial localisation of pellets at ELM trigger Pellet caused MHD perturbation in type-I, type-III and OH discharges Pellet caused plasma cooling in type-I ELMy H-modePellet plasma interaction:Nonlinear MHD modelling of pellet ELM triggering

3. 3 Introduction

4. 4 ELMs can cause critically high power load Type-I ELMs can cause critically high transient power load on plasma facing components. Full performance ITER plasma W pedestal can be as much as 100 MJ leading to ~20 MJ ELM loss if ν* ped is the controlling parameter. Loarte et al., Nuclear Fusion, ITER Physics Basis,Chapter 4 «Edge Localised Modes» (ELMs) are MHD instabilities destabilised by the pressure gradient in the H-mode edge pedestal: Losses up to 10% plasma energy in several 100 micro-seconds Filaments evolving during ELM on MAST. Kirk, et al. PRL 2004.

5. 5 A. Kallenbach, Ringberg seminar,2006 ELM pacing is mandatory in presence of medium-Z radiators ELMs are not only harmful, but also remove impurities from the plasma preventing impurity accumulation and radiation collapse. In presence of medium-Z radiators the ELM pacing is mandatory. Failure of two consecutive ELMs (pellets) are already problematic in scenarios with medium-Z radiators

6. 6 Solving the ELM problem: mitigation by frequency enhancement It was recognized that ELM energy loss scales inversely with ELM frequency ►►► envisaged solution for the ELM problem: Mitigate ELMs below the damage threshold by enhancing their frequency by pellet injection Observation: ΔW ELM ~ P HEAT f ELM for many scenarios with natural ELMs Approach: Split large ELMs into many small ones A. Herrmann, PPCF 44 (2002) 883

7. 7 Pellet ELM pacemaking Injection of frequent small and shallow penetrating cryogenic pellets has been found a promising technique to mitigate this effect. The technique works but the underlying physical processes of the ELM triggering are not well understood. Therefore our aim is to study where and how a pellet triggers an ELM to be able to make predictions for future machines and to optimise the pellet pacing tool.

8. 8 Poloidal scan of pellet triggering capability

9. 9 P.T. Lang et al., NF 47 (2007) 754 Pellet size: 4mm cube velocity: 150-300m/s Fueling pellets injected from all locations on JET trigger ELMs V L H Fueling pellets injected from all locations (V,H,L) on JET trigger prompt ELMs detected by Mirnov coils and on divertor Hα radiation

10. 10 Fueling pellets injected from all locations on DIII-D trigger ELMs Pellets injected from the 5 different injection locations on DIII-D are observed to trigger immediate ELMs. L. Baylor et al., POP 7 (2000) 1878 Pellet size: 2.7mm diam Pellet velocity:150-800m/s

11. 11 Pellets injected from all locations on ASDEX Upgrade trigger ELMs HFS: pellet size: 1.4-2.1mm velocity: 240-1000m/s LFS: pellet size: 1-2mm velocity: 100-200m/s Prompt pellet ELM trigger for HFS injection. ELM event released 20–50μs after ablation onset with a minor part of the pellet mass ablated. P.T. Lang et al., NF 47 (2004) 665

12. 12 Radial localisation of pellets at ELM trigger

13. 13 Radial localisation of HFS pellets at ELM trigger Assumption: to trigger an ELM pellet has to reach a certain magnetic surface (l seed ) independently of the pellet mass and velocity dt ELM ONSET = (l seed – l separatrix ) / V P + t 0 Perturbation spreads: finally an instability starts to grow which develops into an ELM. ELM (type-I) ELM- delay B31-14 500  s separatrix crossing time PELLET Ablation monitor Pick-up coil + + Pellet velocity scan: t 0, l seed can be determined Every pellet injected into type-I ELMy H-mode plasma of ASDEX Upgrade triggered an ELM This ELM was observed only if the pellet entered into the confined plasma crossing the separatrix: dt ELM ONSET = t ELM ONSET – t pellet @ separatrix >0 dt ELM ONSET depends on pellet velocities l sep l ped l seed instability growth rate pert 1 pert 2 Pellet path (l) ELM onset time

14. 14 Radial localisation of HFS pellets at ELM trigger t 0 = 50 ± 7 μs l seed = 2.7 ± 0.4 cm dt ELM ONSET = (l seed – l separatrix ) / V P + t 0 G. Kocsis et al., NF 47 (2007) 1166 Consistently with peeling-ballooning model of the ELM predicting instability onset localized to the pedestal steep gradient region Only 5-15% of the expected pellet mass is ablated until the position of the seed perturbation Still question: smaller pellets still reaching the same location of pedestal gradient region can trigger an ELM or not (according perturbation reduces with pellet size) pedestal

15. 15 0.00 0.02 0.04 0.06 0.08 0.10 1.5 1.6 1.7 1.8 1.9 -40 -20 0 20 40 60 80 2772.22772.42772.62772.8 2773.0 -50 0 50 100 150 200 Pellet Enters Plasma ELM Triggered ELM End n e L(10 14 m -2 ) dB r /dt (a.u.) Inner wall dB r /dt (a.u.) Outer shelf Divertor D  HFS Pellet Induced ELM Details from DIII-D Pellet D  The ELM is triggered 0.05 ms after the pellet enters the plasma or 147 m/s* 0.05 ms = 7.4 mm penetration depth. DIII-D 129195 1.8mm pellet injected from inner wall is ~10% density perturbation No MHD precursor to pellet induced ELM Pellet D  represents ablation of pellet in the plasma with assumed constant radial speed. Time (ms) L.R. Baylor et al, Fuelling WS, Crete, 2008

16. 16 Non-RMP HFS RMP HFS Non-RMP LFS RMP LFS 0 5 10 15 20 25 30 0.60.70.80.911.1 P e (kPa)  Electron Pressure Pedestal DIII-D 129195 2.772s Pre-pellet and ELM Pellet Location when ELM Triggered Localisation of pellet when ELM is triggered in DIII-D L.R. Baylor et al, Fuelling WS, Crete, 2008 Data from DIII-D indicates that the pellet triggers an ELM before the pellet reaches half way up the pedestal (% Ped is % of Te pedestal height). Note: Pellets must penetrate deeper to trigger an ELM when RMP is applied.

17. 17 Radial location of pellets at trigger on JET P.T. Lang et al., NF 47 (2007) 754 For all H track pellets an ELM onset delay of 200 ± 30 μs is derived, related to position of s = 30 ± 4.5mm along the pellet trajectory or r = 23 ± 3.4 mm radilly on the outer mid- plane, respectively. Until the ELM trigger only 1% (or less) of the initial pellet mass is ablated. Pellet size: 4mm cube velocity: 150-300m/s Intrinsic ELM Triggered ELM

18. 18 Is the ELM triggered at the location of the pellet? Is the perturbation localised polidally/toroidally?

19. 19 JET fast framing camera observation geometry 1 Inter limiter box 253467 Limiter box pellet box View: full poloidal cross section covering 90 degrees toidally and seeing the plasma facing components (limiters) on the outboard wall. The cross section of the LFS pellet injection falls into the middle of the view. Time resolution: 10-15μs LFS pellet injection Kocsis et al, EPS 2009

20. 20 1 Inter limiter box 253467 Limiter box pellet box LFS Pellet triggered ELM: ♠ field line elongated filament is growing out from the pellet cloud ♠ bright spots on limiter elements: interaction of this filament with the plasma facing components ♠ appearance of this field aligned structure coincides with the MHD ELM onset. ♠ after detachment from the pellet cloud - bright spots in limiter elements move upward - slowed down after 50-100μs: rotates toroidally/poloidally Filament observation during ELM trigger for LFS pellet injection Kocsis et al, EPS 2009 1234567 1234567 12345671234567 12345671234567

21. 21 Localisation of the seed perturbation for LFS pellets Plasma wall interaction footprint evolution is more regular No delay relative to magnetic ELM onset Pellet triggered type-I ELM (100-200kJ) Up to 80 μs delay relative to magnetic ELM onset Plasma wall interaction footprint evolution is irregular Natural type-I ELM (100-200kJ) The seed perturbation for the ELM triggering may be the pellet cloud The ELM is probably born at the location of the pellet ablation but this picture should be further refined/confirmed investigating HFS pellet ELM triggering

22. 22 Pellet – plasma interaction

23. 23 → neutral cloud formed on µs timescale → diamagnetic pellet cloud launches broadband Alfvén waves (v A ~5 ·10 6 m/s ) that communicated on the whole magnetic surface on a few 10µs -100μs → ionised cloud elongated along field lines expansion vel.: ion sound speed < 10 5 m/s localised in non-axisymetric filament < 10m high pressure Pellet-plasma interaction Geometry at ASDEX Upgrade: Type-I ELMy H-mode pedestal along the pellet trajectory ~ 6cm Pellet cloud radius perpendicular to the magnetic field ~ 1cm Pellet velocity: 240-1000m/s, pellet mass: 3-9(20) 10 19 deuterium Incoming incoming heat flux pellet spherical neutral cloud ionised cloud Pellet ablation in the first 50-100µs: → pellet cloud absorbs the incoming energy flux, cooling wave travels with electron thermal speed ~ 10 7 m/s → homogenised on the whole magnetic surface in a few 10µs-100μs → further complicated by grad B caused cloud drift to LFS

24. 24 Axisymmetric plasma cooling

25. 25 Axisymmetric plasma cooling I ELM onset v P =1000m/s Pellet trajectory v P =600m/s ELM onset Pellet trajectory v P =240m/s ELM onset Pellet trajectory Pellet caused local cooling is homogeneously distributed on the magnetic surface in a few 10µs-100μs (fast electron cooling wave travels with electron thermal speed ~10 7 m/s). The local cooling appears on fast ECE electron temperature measurement located toroidally 90 o from the location of the pellet injection in a few 10µs. 200µs Kocsis et al, EPS 2008

26. 26 Pellet caused cooling: ● appears almost immediately after the pellet reached the according magnetic surface ● causing remarkable temperature drop on a short timescale ● the cooling front moves together with the pellet for all pellet velocities. ● pellet plasma cooling lasts until the pellet is completely ablated and the plasma starts to recover but on a ms timescale. ● relative temperature drop is in the range of few 10% seems to depend on the pellet velocity. ● temperature decrease caused by the triggered ELM is slower than direct pellet one, therefore they can be discriminated. Axisymmetric plasma cooling II v P =600m/s ELM onset Pellet trajectory v P =240m/s ELM onset Pellet trajectory 200µs Kocsis et al, EPS 2008

27. 27 Pellet caused magnetic perturbation

28. 28 Pellet-driven MHD masked by type-I ELM footprint: in the early phase of the ablation (100- 200 μs) the ELM is triggered the pellet-driven perturbation is small, masked by the ELM after the ELM the pellet-driven mode is visible if the pellet lifetime is long enough → but only for long-life pellets (type-I) ELM pellet ELM-delay B31-14 500  s separatrix crossing tt distance from separatrix Therefore: the pellet driven magnetic perturbation was analyzed in different scenarios: OH, HD type- III and type-I H-mode Pick-up coil signals were analyzed: high frequency component: bandpower in 100-300 kHz called ‘envelope’: magnetic perturbation strength spectra and toroidal mode number spectra were studied Pellet caused magnetic perturbation in OH, HD type-III and type-I H-mode Pick-up coil Ablation monitor Szepesi, EPhF 2009

29. 29 ELM pellet type-I type-III OH → no dependence on pellet mass → slight/no dependence on pellet speed → depends on pellet penetration  BUT only in one scenario  implies dependence on plasma parameters instead of penetration  ELM- and pellet-related MHD activity can be separated OH type-I type-III Magnetic perturbation strength Electron pressure [Pa] Magnetic perturbation strength [a.u.] Szepesi, EPhF 2009

30. 30 Spectral properties of the magnetic perturbation magnetic spectrumcoherent mode spectrum OH type-III type-I → in type-I the ELM makes the plasma prone to the n = -6 oscillation, and this is enhanced by the pellet (cooling of plasma edge, emergence of turbulence) → type-III case: more similar to OH → the pellet produces no new phenomena, only enhances what is already present Pellet-induced perturbation: → mode frequency: 100 - 300 kHz → tor. mode number: n = -6 (ion diamagnetic drift direction) TAE: the same parameters but: ELMs are different n = 3, 4 (Neuhauser, NF 2008) TAE oscillation Maraschek, PRL79 pel Washboard n = 3, 4 Szepesi, EPhF 2009 Onliving mode n=-6

31. 31 Nonlinear MHD simulation of pellet ELM triggering

32. 32 flux-aligned grid (reduced resolution) Non linear MHD code JOREK JOREK has been developed with the specific aim to simulate ELMs –domain with closed and open field lines –non-linear reduced MHD in toroidal geometry Density, temperature, electric potential (perp. flow), parallel velocity, poloidal flux Ideal wall conditions on walls Mach one, free outflow at divertor target Aim of the pellet related simulations is to answer the following questions: ELM occurs when plasma crosses linear MHD limit after which plasma returns to stable state How can a pellet trigger an ELM in the stable inter-ELM phase? Is pellet a trigger or a cause for an ELM? Huysmans, EPS 2009

33. 33 Pellets are approximated as an initial perturbation to the density profile: –poloidally and toroidally localised Amplitude typically 25 times central density Total amount of added particles 3-6% –constant pressure (i.e. low temperature) Pellet triggered MHD in H-mode pedestal: –Initial non-linear MHD simulations of pellets injected in H-mode pedestal show: Destabilisation of medium-n ballooning modes High pressure in pellet plasmoid drives MHD instability forming a single helical perturbation at the pellet position –suggests pellets can cause ELMs (instead of being a trigger) Non linear MHD code JOREK: modelling of pellet ELM triggering Huysmans, EPS 2009 x ~0.5μs

34. 34 Summary Millimeter sized fuelling pellets trigger ELMs independently of the poloidal injection angle (LFS, HFS) Pellets are located in the mid-pedestal at ELM trigger, and only a small fraction of their mass is ablated until the trigger, but it is still not known that smaller pellets just reaching the same radial location are still trigger ELMs For LFS pellets the high pressure pellet cloud formed around the pellet seems to be the seed perturbation but for HFS injection the situation is not yet clear The high pressure pellet cloud, the axisymmetric plasma cooling and the magnetic perturbation of the pellet cloud have been recognized as candidate perturbation for trigger mechanism, and were investigated in details Initial nonlinear MHD simulation for pellet ELM triggering shows destabilization of medium-n ballooning modes. High pressure in pellet plasmoid drives MHD instability forming a single helical perturbation at the pellet position for LFS injected pellets, which agrees with the observations on JET

35. 35 Outlook HFS-LFS asymmetry of the ELM triggering will be further investigated ►► on ASDEX Upgrade with LFS injection (measurement of the internal delay for LFS) ►► on JET with HFS injection (fast visible imaging of filaments) The investigation of pellet caused magnetic perturbation and plasma cooling is an ongoing project on JET, results will be published soon. The simulation should be run to clarify the determining mechanism of the pellet Elm triggering (HFS/LFS) Further investigations would be necessary with reduced pellet size or with small impurity pellets.

36. 36

37. 37 Backup transparencies

38. 38 Pellet cloud releases from pellet and expands along a flux tube. Density from the cloud expands along flux tube at the sound speed c s. Temperature ‘cold wave’ travels along the flux tube at the thermal speed. Heat is absorbed in the cloud resulting in a temperature deficit far from the cloud. Pressure decays and expands along the flux tube with a lower pressure far from the cloud. Strong local cross field pressure gradients result along the flux tube that form on  s time scales. What Causes an ELM Trigger from a Pellet? L nene cscs TeTe qeqe L PePe L   LRB Crete Jun2008

39. 39 long exposure Cam. exposures Ablation monitor signal time [s] Trajectory recons- truction separatrix t pellet @ separatrix Vide angle view photodiode Experimental set-up I short multiple exposures on time: 10µs off time: 90µs exposures Pellet localization (space, time): - fast CCD cameras + spatial calibration → def: penetration = distance from separatrix Injection geometry

40. 40 Processing of the pickup coil signals -eliminate LF component by moving box average - calculate the envelope of the remaining HF component (25  s box) - assume: envelope ~ MHD perturbation magnitude Experimental set-up II Toroidal pick-up coil set about 180 0, 5 coils Mirnov coil set 360 0, 30 coils Poloidal pick-up coil set, 7 coils about 60 0 ELM (type-I) ELM-delay B31-14 500  s separatrix crossing Calibrated Electron Cyclotron Emission (ECE) profiles are measured in second harmonic X-mode with a fast 60-channel heterodyne radiometer PELLET Ablation monitor Pick-up coil signal

41. 41 Experimental set-up III short-time Fourier transform (STFT): time shift and frequency shift invariant continuous analytical wavelet transform: time shift and scaling invariant Mode numbers as least squares fit to cross-phase as function of relative probe position : Processing of pick-up coil signals: spectral properties

42. 42 Radial localisation of pellets at ELM trigger Radial localisation of HFS pellets at ELM trigger Low pellet injection frequency : trigger events occur at different times in the ELM cycle that is we perform the analysis as a function of the time elapsed after the previous ELM Pellets can trigger ELMs at any time in the ELM cycle: plasma edge is not stable against HFS pellet induced seed perturbation on ASDEX Upgrade dt ELM_ONSET saturates with increasing dt elapsed dt elapsed > 8ms: dt ELM_ONSET ~ constant G. Kocsis et al., NF 47 (2007) 1166

43. 43 → axisymmetric decrease of the plasma pressure causing a pressure gradient increase in front of the pellet rough estimate: 10-30% for AUG → the incoming energy flux is absorbed and concentrated in the HFS localised helical non axisymmetric cloud causing a cloud pressure higher then that of the target plasma rough estimate: >10 P e Pellet-plasma interaction Pellet affects the plasma at a magnetic surface as long as t=Diam CLOUD /v P =20-100µs Estimation of the pellet cloud pressure and plasma pressure drop

44. 44 Injection scenarios Measurement of the ELM onset delay as a function of the pellet velocity → V P =240,600,880,1000 m/s HFS looping system: pellet erosion is velocity dependent → r P =0.71, 0.67, 0.58, 0.51 mm N P =9, 7, 5, 3 x 10 19 Characterization of the injection scenarios → Neutral Gas Shielding model calculation dN P /dt = - Const. r P 4/3 n e 1/3 T e 1.64 dN P /dl V P = - Const. r P 4/3 n e 1/3 T e 1.64 Designated pellet path Integrating this diff. equation along the designated pellet path the ablation rate (ablated particles /s) is calculated P.B. Parks et al., PoP 21 (1978) 1735

45. 45 240 m/s, r=0.71mm, N=9 ·10 19 600 m/s, r=0.67mm, N=7.4 880 m/s, r=0.58mm, N=5.e19 1000m/s,r=0.51mm, N=3.3e19 Injection scenarios Pellets penetrate deep into the plasma crossing the pedestal region completely In the pedestal region the ablation rate is nearly similar for all injection scenarios Particle deposition scales with 1/V P

46. 46 slowing down, n=4 Type-I ELMy H-mode el. diam. Drift, n>0 ion diam. Drift n<0 washboard-type precursor mode No precursor 15 pellet induced ELMs have been analysed for their basic poloidal/toroidal structure. The single, upward rotating structure represents a rather specific case. Typically, two or more such pronounced structures occur simultaneously. Despite the fixed pellet launch position, they appear initially at a more or less random toroidal position relative to the pellet, i.e. they are not directly growing out of the pellet plasmoid. Prompt triggered delayed triggered Neuhauser et al, NF, 48, 045005, 2008