Pellet induced high density phases during ELM suppression in ASDEX Upgrade

Pellet induced high density phases during ELM suppression in ASDEX Upgrade P. T. Lang 1), C. Angioni 1), R.M. Mc Dermott 1), R. Fischer 1), J.C. Fuchs 1), O. Kardaun 1), B. Kurzan 1), G. Kocsis 2), M. Maraschek 1), A. Mlynek 1), W. Suttrop 1), T. Szepesi 2), G. Tardini 1), K. Thomsen 3), H. Zohm 1), ASDEX Upgrade Team 1) MPI für Plasmaphysik, EURATOM Association., 85748 Garching, Germany 2) WIGNER RCP RMKI, EURATOM Association, POB 49, 1525 Budapest, Hungary 3) European Commission, DG for Research & Innovation, 1049 Brussels, Belgium E-mail contact of main author: peter.lang@aug.ipp.mpg.de Abstract. Magnetic perturbations (MP) with n = 2 have been found in ASDEX Upgrade to result in reproducible and robust ELM mitigation in a wide heating power and safety factor range. ELM mitigation is established for peripheral densities above a critical threshold. Pellets injected into mitigation phases do not trigger type-i ELM like events unlike when launched into unmitigated type-i ELMy plasmas. The absence of ELMs results in an improved pellet fuelling efficiency and persistent density build up, mostly eliminating the need for strong gas puff. No deleterious impact was found on MHD activity, plasma rotation or impurity transport. Notably, the pedestal density, temperature, pressure and rotation profiles remain virtually unchanged. Operation at line averaged densities from 0.75 up to 1.5 times n Gw (core densities of up to 1.6 times n Gw ) has been demonstrated, an upper density limit for the ELM-mitigated regime not been encountered so far. In the density regime near and above the Greenwald limit (up to about n/n Gw = 1.4) the confinement time at ASDEX Upgrade turns out to be prognosticated more truthfully by the multi-tokamak scaling from 2006 than by the IPB98(y,2) power law scaling. A prolonged train of pellets with repetition time t P produces an enhanced plasma particle inventory which exhibits a maximum just after the pellet arrives, and which drops with a decay time τ P towards a minimum just before the next pellet. For t P > τ P, an approximately linear relationship exists between the pellet flux and the time averaged plasma density. In parallel, a mild reversible plasma energy reduction takes place. This behaviour is already known from previous pellet fuelling studies performed under ELMy H-mode conditions; the achieved quasi steady-state operational boundaries can be well explained by additional pellet-driven convective losses. In the ELM suppressed regime, however, enhancing the pellet flux by reducing t P, below the initial value of τ P resulted in a strong sudden density enhancement. Furthermore, no further significant energy reduction takes place. The behaviour is caused by an increasing τ P, attributed mainly to changing transport properties, in addition to the reduced ELM losses. Usually, fuelling performance improvement can be gained solely by deeper pellet particle deposition, which results in a slightly enhanced pellet particle persistence time. 1. Introduction Studies for future fusion power plants based on the tokamak principle show that at high central pressure, the temperature in the core plasma would reach values in excess of the optimum for D-T fusion, if the device is operated below the empirical Greenwald density limit n Gw [1]. When using gas puff fuelling in present day tokamaks, density profiles become flat and are restricted to values below this limit. The latter is considered to be a limit of the plasma edge density and should not impose a restriction to the core density. Pellets made from frozen solid fuel, injected deep into the plasma in present devices, are the prime candidate to improve core fuelling. In reactor relevant H-mode scenarios, conventional outboard pellet injection showed low fuelling efficiency as a result of the B drift and ELM triggering. While the first hindrance was many years ago turned into an advantage through the use of inboard pellet injection, the second obstacle was overcome only recently at ASDEX Upgrade (AUG) by external magnetic perturbations (MP) suppressing strong type-i ELMs [2]. Whereas for fuelling size pellets injected into a type-i ELM phase an initial ELM is already triggered while ablation is still on going, the same pellets injected under almost identical plasma conditions but during ELM suppressed phases do not trigger ELMs. Large ELMs remain suppressed even in the presence of strongest pellet perturbations. Pellet fuelling has already proven its ability to access plasma densities somewhat beyond the Greenwald density but at the expense of some confinement degradation [3]. Showing enhanced fuelling efficiency

during suppression phases now high density operation far beyond the Greenwald density with good confinement has been achieved. Such a scenario could provide a high density, high confinement operation regime for ITER and DEMO [4]. 2. Experimental boundary conditions ASDEX Upgrade is a divertor tokamak with all plasma facing components completely covered with tungsten (W). In-vessel saddle coils that can produce non-axisymmetric magnetic perturbations were installed; the initial half of B-coils (shown in red in figure 1) prior to the 2010 campaign, the augmenting B-coils prior to the 2011 campaign. The projected final configuration is shown in the left part of figure 1, composed of 16 B-coils (each 8 upper and lower ones, referred to as Bu- and Bl-coils, respectively) and 8 A-coils. The present coil configuration consists of two rows of 8 coils evenly spaced toroidally and mounted to the passive stabilization loop on the low field side. The coils have five turns each and create a mainly radial field with toroidal mode numbers up to n = 4 [5]. Experiments presented here were performed during the 2010 campaign restricting the coil configuration yet up to n = 2. To perform a first set of pellet fuelling studies the decommissioned high speed inboard launching system based on a centrifuge accelerator and a looping transfer system was revitalized. A sketch of the system is displayed in the right part of figure 1. During the 2010 campaign the system was capable of delivering pellets with a nominal particle content ranging from 1.5 3.7 x 10 20 D in the velocity range 240 560 m/s from the magnetic high field side of the torus with repetition rates of up to 47 Hz. Within a given pellet train launched into a discharge both pellet speed and size were fixed. However repetition rates can be changed to a fixed fraction of the centrifuge revolution frequency. Prior and during the 2011 campaign, the pellet systems was subject to a deep modernization aiming for more flexibility and further enhanced fuelling performance. The pellet observation system was also refurbished and upgraded to include two ultra-fast CMOS cameras and is now capable of fast individual pellet tracking up to 1 Mframe/s. FIG. 1: Left: Configuration of in-vessel saddle coils in ASDEX Upgrade, projected ultimately to consist of 8 A-coils and 2x8 B-coils (upper and lower) installed at the low field side. Currently installed are all B-coils, results reported here were achieved during the campaign 2011 with the initial set of coils only (2x4 coils, allowing yet for n=2 magnetic field perturbations) shown in red. Right: Centrifuge launcher and pellet guiding system capable of high-speed inboard launch and in vessel configuration of ASDEX Upgrade with divertor IId as operated in the 2011 campaign. Flux surfaces from a typical magnetic configuration are also shown in red.

3. Synergy of MP ELM mitigation and pellet fuelling Magnetic perturbations with n = 2 have been found in AUG to result in reproducible and robust ELM mitigation in a wide heating power and safety factor range. ELM mitigation is established for peripheral densities above a critical threshold; so far it is not possible to distinguish whether this requirement corresponds to an edge collisionality threshold or a critical fraction of the Greenwald density limit. Once the mitigation regime is properly established, type-i ELMs disappear completely and are replaced by small ELM-like events [2]. ELM mitigation persists in a high density, high collisionality regime even with the strongest applied pellet perturbations. FIG. 2: Slow fuelling size pellets (largest achievable perturbation magnitude) at the tail-end of a MP generated ELM mitigated phase. Neither during phases with full or ramped down coil current nor in the following phase with sustained ELM mitigation a pellet triggers an ELM. Instead, they show a high fuelling efficiency and a smooth density relaxation after each pellet induced density jump. The compatibility of fuelling size pellet injection with magnetic perturbation induced ELM mitigation is shown in figure 2. A pellet train covers the end of the coil current steady state phase, the coil current ramp down (100 ms), and also the phase immediately afterwards showing sustained ELM mitigation. After termination of the coil current a slow gradual increase of the plasma stored energy sets in, lasting several hundred ms. Finally, the level of type-i activity is approached and type-i ELMs reappear. As can be seen by the absence of large ELM-induced divertor current spikes (3rd trace from top), neither in the ELM mitigated phase with activated coils nor in the following sustained phase do pellets trigger strong type-i ELMs. The example displays a pellet sequence consisting of 8 pellets (the second pellet is very small) where the 7 large pellets impose the strongest local perturbation attainable with

the available launching system. The pellets exhibit remarkably good fuelling behaviour. Since no initial ELM is triggered, high fuelling efficiency (estimated 70 100% of the arriving pellet mass) is achieved by the pellet particle deposition followed by a smooth density evolution. Little impact is observed on confinement which quickly recovers after every pellet. As an initial fuelling application, pellets were employed to assist access to the ELM mitigated regime while operating with freshly boronized walls. Due to the high wall pumping under such conditions the requested peripheral densities are hard to establish. The massive gas bleeding required would result in unwanted gas loading of the wall. Pellets are able to establishing easily the required conditions. Under standard operational conditions with unboronized walls a gas puff still suitable with respect on its impact on the confinement is sufficient to reach the peripheral line density. In a freshly boronized vessel causes strong wall pumping often full suppression of type-i ELMs without confinement degradation cannot be established by pure gas puffing. Pellet injection can be helpful to access the ELM-mitigation regime by raising the peripheral density beyond the required threshold level, mostly eliminating the need for strong gas puffing. For appropriate initial plasma parameters the sudden pellet initiated density step is able to kick the edge into the ELM mitigation regime even with boronized walls. To optimize edge fuelling, large pellets at the lowest achievable speed of 240 m/s were employed at a modest repetition rate. For such low repetition rate, during the early part of the pellet train the induced density enhancement usually drops back below the critical density level and type-i ELMs re-occur until the arrival of the next pellet. Persistent mitigation is achieved once the frequency of the pellets is sufficiently high to keep the edge density above the threshold level. In order to achieve an overall optimized performance, the gas puffing required reaching ELM mitigation can be largely replaced by the more efficient pellet fuelling. However, experiments aimed at determining the minimum amount of gas puffing required to maintain MP ELM mitigation showed that the gas puff cannot be entirely replaced by pellets; with too strong a reduction causes type I ELM activity to reappear. Using the pellets dedicated fuelling experiment were performed to achieve higher densities in order to expand the operational space of the MP ELM-mitigated regime beyond the limit encountered with pure gas puffing and to explore if there is an upper limit in density. This has been done for the high current, I p = 1.0 MA, MP ELM mitigation scenario with strong auxiliary heating (mix of neutral beam and central electron and ion cyclotron resonance heating) using mid-sized (nominal mass 2.9 x 10 20 D) pellets at 560 m/s. The pre-programmed pellet particle flux was adjusted by the pellet frequency. Successful fuelling replacement was achieved with a total gas puff reduction of about 5 times the applied pellet flux. Pellet injection shows advantageous fuelling behaviour as compared to gas puffing and allows for high density operation with less fuel consumption and less detrimental impact on the edge. Experiments were performed to determine the largest possible density enhancement with the least deleterious impact on confinement. This can be achieved by the deepest possible pellet penetration and deposition, i.e. maximised pellet mass and speed for any given set of plasma parameters. Maximum fuelling conditions at hand for this experiment were the largest available pellet size at v P = 560 m/s, the maximum repetition rate was 47 Hz. Applying this maximum pellet flux a line averaged density = 1.8 x 10 20 m -3 is achieved, corresponding to a Greenwald factor of 1.5. During this entire high density phase the ELMs remain mitigated. So far this is the highest density obtained with ELM mitigation. Progressing beyond the Greenwald limit by means of pellet fuelling in AUG is achieved by overcoming the flat density profiles typically observed in high density operation with gas puffing. Pellet fuelling experiments in ELMy H-mode plasmas (with less efficient fuelling) showed that a gradual pressure/energy loss can be avoided when approaching the edge density limit and thereby, high density operation can be combined with reasonable confinement [3]. As can be seen in

the density profiles shown in figure 3, the density increase during the pellet train takes place entirely inside the pedestal top while the pedestal itself remains, within diagnostic resolution, unchanged. For the case of maximum available pellet fuelling (red curve, #27119) in the plasma centre the density is almost doubled with respect to an identical reference discharge without pellets (black curve, #26987) reaching values of = 2.0 x 10 20 m -3 (corresponding to a Greenwald factor of 1.6). No deleterious impact is found on MHD activity, plasma rotation or impurity transport. Temperature and plasma rotation profiles, like the density, show no significant change at the pedestal top. The electron temperature (T e ) pedestal profiles as measured by the Thomson scattering diagnostic agree within the data resolution for all three phases. An enduring effect of the strong pellet fuelling is only observed inside the edge transport barrier. Altogether, reliable and reproducible operation at line averaged densities from 0.75 up to 1.5 times n Gw (core densities of up to 1.6 times n Gw ) has been demonstrated using pellets at AUG. An upper density limit for the ELM-mitigated regime has not been encountered so far; limitations were set solely by technical restrictions of the pellet launcher. FIG. 3: Radial density profiles in a phase with strong ELM suppression by magnetic perturbation for gas puff fuelling (black) and for pellet fuelling at different frequencies of 35 Hz (blue) to 47 Hz (red). The region of particle deposition by an ablating pellet is shown as well; Thomson scattering measurements (diamonds) suggest a related density gradient zone. FIG. 4: The strong density increase is caused by a pronounced enhancement of the pellet particle sustainment time. This is indicated by the decay time as calculated from line averaged density evolution. There is also some influence by the changing pellet penetration depth and by profile peaking (brown dots: estimated evolution of τ P ). 4. Advanced performance at high pellet injection frequency Due to its inherently transient character, pellet fuelling cannot establish truly steady state conditions. A train of pellets with repetition time t P produces an enhanced plasma particle inventory which exhibits a maximum just after the pellet arrives, and which drops with a decay time τ P towards a minimum just before the next pellet. For t P > τ P, an approximately linear relationship exists between the pellet flux and the time averaged plasma density. In parallel, a mild reversible plasma energy reduction takes place. This behaviour is already known from previous pellet fuelling studies performed under ELMy H-mode conditions; the

achieved quasi steady-state operational boundaries can be explained well by additional pelletdriven convective losses. In the ELM suppressed regime a situation where t P < τ P can be easily obtained. In this case, a strong sudden density enhancement, as shown in figure 3, is observed. Whereas for a pellet rate of 35 Hz (blue curve, #27029) a modest increase of density takes place, increasing the pellet rate und flux by about 1/3 results in a drastic change. The decay time τ P can be obtained from density evolution of volume averaged density <n e > = N e /V with N e the particle inventory and V the volume of confined plasma. The temporal evolution after every single pellet injected at t can be described quite well by the relation <n e > (t) = < > + <n e > (t) with <n e > (t) =. Yielding a better temporal resolution, as measured by the DCN and CO 2 laser interferometers was taken for analysis. For the regime investigated the relation between and <n e > can to be described quite well by <n e > = 0.9 m with α a smooth density dependent fit parameter dropping only slightly below unity at densities significantly beyond the Greenwald density. An example of the temporal evolution of the pellet induced density surplus to the initial background density for the case of maximum available flux as measured by is displayed in figure 4 (first track from top); plotting ln( ) (2 nd track) it becomes already obvious sustainment times of pellet density enhancements increase strongly with density. Calculating directly from the evolution and correcting for profile peaking effects to obtain τ P = /α it turns out the initial sustainment time of about 35 ms increases by a factor of about 3 when approaching the high density regime. Notably, this initial decay time further increases to about 200 ms about 50-100 ms after the pellet injection. With the pellet deposited particles sustaining now much longer, the density enhancement is strongly boosted once the applied pellet flux is sufficient to enter this regime. Hence, thus once the regime is just approached, a mild increase of pellet rate and flux can result in a strong density enhancement. It is worth noting this effect is not primarily due to deeper pellet penetration. Without doubt, the cooler and denser plasma in the high density regime favours pellet penetration while in turn deeper penetration and hence particle deposition results in a longer particle sustainment. Albeit present here as well, the increasing pellet penetration accounts only for a minor contribution. Pellet penetration depths are determined for all pellets from the duration of pellet ablation and for some pellets from the fast framing camera images. Obtained values found in very good agreement are displayed in the lower track of figure 4. Whilst the pellet sequence enhances density, pellet penetration increases slightly as expected. However, some pellets later in sequence arriving accidentally with smaller mass do penetrate much less than the initial pellets but nevertheless still show a major fraction of the improvement in τ P. The observed behaviour is hence attributed mainly to changing transport properties. Investigations are underway to determine access conditions and understand the underlying physics of this possibly advanced performance regime in the ELM mitigated plasma phase. The advanced particle confinement is particularly attractive when one considers other options to achieve improved pellet particle persistence times which require significant engineering efforts to obtain higher pellet speed and/or mass. 5. Confinement at high density and scaling considerations The observed advanced performance allows to access to high densities far beyond Greenwald density with a remarkably low pellet particle flux. For the investigated scenarios, pellet fuelling efficiencies higher than gas puffing allowed to achieve this high density regime by just replacing a part of the required gas flux. As a consequence, significant confinement losses were not suffered with respect to initial phases or reference discharges. However, there is also not a confinement improvement anymore in the density regime above about n Gw = 0.85. Hence, the validity of the IPB98(y,2) energy confinement scaling, which contains a - dependence, has to be questioned for densities above 0.85 n Gw [4]. On the other hand, the

ITERH06-IP(y,dd) scaling, which predicts stagnating confinement near = n Gw describes the observations quite well. In effect, for inductive operation of ITER, optimal operational parameters to achieve a maximum Q value (fusion power divided by external heating power), do change notably compared to the values based on IPB98(y,2); this was reported recently in [1]. However, further refinement of the new scaling is still useful since some interactions, e.g. between / n Gw and the heating power per particle (P/N e ) as well as between aspect ratio and safety factor, have not yet been taken into account. The general rationale of looking at the quality of plasma confinement, expressed as an H- factor with respect to some empirical scaling, is to provide a benchmark for the performance of a particular device and scenario. Moreover, validation of an established scaling in a wide range of devices and scenarios is required for empirically predicting performance of future devices. Our recent results indicate that a closer look at this scaling of the confinement time near the Greenwald limit is still worthwhile. Therefore, we revisit this topic here. The thermal H-mode confinement time scaling at present mostly used is the simple power law scaling IPB98(y,2) [6]. Following earlier database WG discussions, interactions were detected between the cylindrical q value q cyl and n e, and even between them and machine size, see [7]. This led to more careful prediction for large devices, including the implications of an uncertainty estimate [8]. For scalings that went beyond a simple power-law, the size dependence interaction has been separated from a log-quadratic dependence on n e /n Gw, as well as from a log-quadratic interaction between the shape factor F sh = q 95 / q cyl (which is some alternative to the triangularity δ) and n e /n Gw [9]. While referring to the original contribution [10], a simplified functional form of the previous IPP-02 scaling, re-fitted and supported by additional data, was analysed, using MKSA units, in [1] with a view towards ITER performance. The scaling from 2006 was based on an iteratively updated version of the multimachine database [11]. An extensive set of selection rules had been applied to obtain a standardized dataset (D into D discharges only), of data from fourteen tokamaks, including two low aspect-ratio, spherical devices (MAST and NSTX). The present version of the dataset, documented at EFDA, is DB4v5. The ITERH06-IP(y,dd) scaling, expressed in MKSA units, and corrected for M eff = 2.5 (instead of 2.0), i.e. with the intercept compatible to 290 MJ thermal energy content for ITER FEAT (see [10]), reads ( ) ( ) with M eff the effective atomic mass of the ions, A = R/a the aspect ratio, and where is the cross-section based, effective elongation. (For the other symbols the usual definitions apply, as specified e.g. in [1]. ) Pellet fuelling experiments performed at JET and AUG showed in ITER relevant scenarios a density regime beyond values accessible with gas puffing can be accessed but by no means with further growing energy confinement [12]. Ideally, the initial confinement can just be kept constant by avoiding pellet added convective losses as confinement evolution observed could be modeled taking into account the initial confinement and degradation by losses of thermalized pellet flux. A similar behavior was again found in this study. Access to the high density regime is possible but no confinement increase as predicted by the ITERH98P(y,2) scaling is found. As a consequence, the ratio H98 of achieved thermal energy confinement times to the ITERH98P(y,2) scaling predictions drops significantly for densities above 0.85 n Gw. This can be seen from the lower part of figure 5, displaying H98 versus the achieved fraction of the Greenwald density for discharges of the 1.0 MA n = 2 mitigation scenario. Data shown were taken from reference plasmas with gas fuelling yet without (open circles) and with MP ELM mitigation applying different heating power levels. Pellet fuelling to moderate and high densities (filled squares) is achieved by varying the pellet particle flux.

All data can be ordered by means of the H98 value, however obviously the predicted energy confinement time becomes more and more misleading with an increasing / n Gw ratio. On the other hand, the ITERH06-IP(y,dd) scaling, which predicts stagnating confinement near = n Gw describes the observations quite well. Here, the energy confinement time ratio H06 keeps the same level up to the highest densities established so far. To obtain H98 and H06, the standard approach (calculating the thermal energy confinement time excluding fast ion contributions and taking into account heating power losses) was applied also at AUG. FIG. 5: Ratio between the observed thermal energy confinement times and the predictions from ITERH98P(y,2) (lower) and ITERH06-IP(y,dd) (upper) scalings versus achieved fraction of the Greenwald density (1.0 MA scenario only). Reference data as obtained for type-i ELMing phases (open circles) and mitigated phases (filled circles) with gas puffing. References [1] JOHNER, J., Fusion Science and Technology, 59 (2010) 308. [2] SUTTROP, W., et al., Phys. Rev. Lett. 106, (2011) 225004. [3] LANG, P.T., et al., Nucl. Fusion 40, (2000) 245. [4] ZOHM, H., et al., this conference, FTP/3-3 [5] SUTTROP, W., et al., this conference, EX/3-4 [6] ITER EXPERT GROUPS, et al, Nucl. Fusion 39, (1999) 2175. [7] KARDAUN, O., Plasma Phys. Control. Fusion 41, (1999) 429. [8] KARDAUN, O., Nucl. Fusion 42, (2002) 841. [9] KARDAUN, O., Interval estimate of the global energy confinement time during ELMy H-mode in ITER FEAT, IPP-IR-2002/5-1.1. Report MPI für Plasmaphysik, Garching, 2002. [10] KARDAUN, O., for the ITPA global confinement database WG, Proc. 21th Conf. Plasma Phys. Control. Nucl. Fusion Res., Chengdu 2006, IT/P1-10, [11] MCDONALD, D., et al. Nucl. Fusion 47, (2007) 147. [12] LANG, P.T., et al., J. Nucl. Mater. 290 292, (2001) 374.