Shattered Pellet Injection as the Primary Disruption Mitigation Technique for ITER

Transcription

1 1 EX/9-2 Shattered Pellet Injection as the Primary Disruption Mitigation Technique for ITER D. Shiraki 1, N. Commaux 1, L.R. Baylor 1, N.W. Eidietis 2, E.M. Hollmann 3, V.A. Izzo 3, C.J. Lasnier 4, R.A. Moyer 3, T.C. Jernigan 1, S.K. Combs 1 and S.J. Meitner 1 1 Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA 2 General Atomics, P.O. Box 85608, San Diego, CA 92186, USA 3 University of California-San Diego, La Jolla, CA 92093, USA 4 Lawrence Livermore National Laboratory, Livermore, CA 94551, USA contact of main author: shirakid@fusion.gat.com Abstract. The shattered pellet injection (SPI) technique has demonstrated disruption thermal quench (TQ) and current quench (CQ) control that scale to meet ITER disruption mitigation requirements. Direct comparisons of SPI with the massive gas injection (MGI) technique in comparable conditions shows improved mitigation results with the SPI approach. Faster particle delivery and enhanced core deposition allows SPI to achieve 20% lower divertor heat loads than equivalent MGI shutdowns of DIII-D plasmas, while the particle delivery timescales are expected to scale to ITER more favorably in the case of SPI. Techniques have also been developed to further tune the SPI process by injecting pellets composed of mixtures of neon and deuterium, allowing control of TQ and CQ properties by varying the composition of the pellet. Mitigation metrics are observed to saturate within scaled injection quantities anticipated for ITER, providing a possible technique for tuning disruption properties to remain within allowable ITER limits. The SPI technique has also been applied to mitigation of post-disruption runaway electron beams. Initial results demonstrate that neon shattered pellets can dissipate such runaway beams after their formation, but high- and low-z impurity species are found to have competing effects, indicating that careful selection of the pellet species is important. 1. Introduction Major disruptions in ITER, if unmitigated, have the potential to generate high heat loads during the thermal quench (TQ) and large mechanical stresses during the current quench (CQ), as well as multi-ma beams of relativistic runaway electrons [1]. Damage from such unmitigated disruption loads may require frequent repairs and costly downtime, necessitating effective mitigation techniques to reduce their severity [2]. Massive high-z impurity injection allows the mitigation of disruption loads, reducing peak divertor heat flux during the TQ by dissipating a large fraction of the plasma thermal energy through radiation, while reducing halo current forces by increasing the plasma resistivity and accelerating the plasma current decay during the CQ. Additionally, the large density rise due to assimilation of the injected impurities enhances the collisional drag on seed runaway electrons, allowing the suppression of large runaway beams if the density is high enough [3]. In the shattered pellet injection (SPI) approach [4,5], the impurity material is injected in the form of a solid cryogenic pellet which is shattered just prior to entering the plasma. The shattering of the pellet protects in-vessel components from direct impact by a large solid pellet, and simultaneously improves assimilation of the particles in the plasma by generating a high surface area spray of pellet fragments. The resulting shattered pellet material arrives at the plasma surface in a very brief interval. This is in contrast to the previously studied massive gas injection (MGI) technique, which injects the impurities as a high pressure gas pulse, and whose delivery is limited by gas flow rates through the delivery tube [6]. In this paper, we describe results from studies of the SPI technique on the DIII-D tokamak, which have led to its selection as the primary injection scheme for the ITER disruption

2 2 EX/9-2 mitigation system (DMS) [7]. In particular, comparisons of SPI with the MGI approach, methods of optimizing the SPI process through variation of the pellet composition, and application of SPI to mitigation of runaway electron beams are described. 2. Improved Mitigation over MGI MGI has been well studied on a number of tokamaks, demonstrating the ability to mitigate disruption TQ and CQ loads [1,2,8]. Thus dedicated experiments have focused on making detailed comparisons of SPI with the more studied MGI approach. On DIII-D, SPI and MGI injectors are located on the same port, so that by injecting identical impurity quantities in equivalent plasma targets, direct comparisons of the two techniques can be made [9]. Compared with equivalent MGI, SPI shows significant improvements in particle delivery rates. Because the impurities travel as a solid pellet until just prior to entering the plasma, they are not limited by fluid conductance as in the case of MGI. This results in a more instantaneous delivery of the injected material. Typical density evolutions following each injection type are shown in Fig. 1. Despite identical injection quantities (53 Pa-m 3 ), the fraction of injected particles observed in the plasma for SPI is a factor of two higher compared to MGI. This peak density is also reached much earlier in the disruption, during the TQ and early part of the CQ. This is advantageous for both thermal mitigation, when rapid assimilation of the radiating impurity is necessary during the TQ, and for collisional suppression of runaway electron seeds during the early CQ. FIG. 1. Comparison of line integrated densities during the CQ for equivalent neon SPI and MGI injections. FIG. 2. Histograms of peak ratios of line integrated densities during equivalent MGI and SPI. In addition to the faster delivery rate and higher total assimilation, SPI results in better core deposition than MGI, due to the existence of solid fragments which are able to penetrate beyond the plasma edge. This radial penetration of solid fragments is verified by fast framing visible camera images. The resulting core deposition is also characterized by comparing line integrated density signals from two interferometer chords: one which travels through the core of the plasma (CENTRAL), and a second which only samples the plasma edge (EDGE). A histogram of the peak ratios of these two signals is shown in Fig. 2. It is observed that SPI

3 3 EX/9-2 achieves a more centrally peaked deposition, due to the ballistic transport of pellet fragments into the plasma core, in contrast to MGI pulses which are stopped at the plasma edge and instead rely on MHD mixing for radial transport [10]. As a result of the faster particle delivery and higher global and core deposition, SPI is able to better mitigate conducted heat loads during the TQ. The improved delivery of radiating impurities allows more efficient dissipation of the plasma thermal energy through line radiation, which distributes the heat loads more isotropically than does conduction to the strike points. Fig. 3 shows that for a range of plasma thermal energies, SPI reduces peak divertor heat loads by an additional ~20% over equivalent MGI shutdowns. A similar reduction of peak heat loads is also observed for the outer strike point. While significant differences in mitigation results are already observed in DIII-D, the effectiveness of the two techniques are FIG. 3. Peak heat fluxes measured at the inner expected to scale differently towards ITER. strike point during equivalent MGI and SPI. Because the pellet remains solid until being shattered just outside the plasma, the particle delivery by SPI remains very rapid even in larger devices (after accounting for the required time-of-flight from the injector to plasma surface). In contrast, computational fluid dynamics calculations of MGI flow rates in ITER show that gas pulses become significantly spread out in time when accounting for realistic distances between the gas valves and the plasma, slowing injection timescales [7]. 3. Control of Disruption Properties ITER will require disruption properties to remain within certain allowable limits, even when mitigation techniques are applied [2]. For example, TQ radiation fractions must be high enough to prevent melting and erosion of plasma facing components, while simultaneously keeping the resulting CQ timescales above a lower bound set by induced eddy currents in the blanket modules. In addition to the electromagnetic loads, very fast CQs are undesirable due to the enhanced drive for runaway electron growth by the avalanche mechanism. Thus, mitigation techniques must allow control of TQ and CQ properties, in order to ensure that these disruption characteristics remain within allowable ranges. DIII-D experiments have demonstrated a technique to control disruption properties based on the injection of mixed species shattered pellets [11]. These pellets are frozen as a homogeneous mixture of neon and deuterium (due to the proximity of the triple points and sublimation curves of the two species), allowing the pellet composition to be varied freely using only a simple pipe-gun injector design. Although pellet sizes are determined by the injector barrel size, this technique allows pellet compositions to vary continuously, ranging from pure deuterium to pure neon, and including intermediate mixtures of arbitrary ratios. By varying the quantity of neon in a shattered pellet, the resulting TQ and CQ properties during the SPI shutdown can be controlled.

4 4 EX/9-2 FIG. 4. TQ radiation fraction as a function of the neon quantity in variable composition mixed species shattered pellets. The orange band indicates the scaled ITER injection quantities. Fig. 4 shows measured TQ radiation fractions as a function of the neon quantity in the pellet. It is observed that radiation fractions quickly saturate as a function of the injected neon quantity, due to the efficiency of neon as a radiating species. At the highest injected quantities (~50 Pa-m 3, corresponding to pure neon for the pellet sizes used here), radiation fractions approach 90%, which is the target value for the ITER DMS at high thermal energies [2]. These injection quantities can be scaled to ITER, by assuming that required impurity quantities scale proportionally with the plasma thermal energy W th which must be dissipated (although other dependencies such as the thermal energy fraction W th /W mag, not considered here, may be possible FIG. 5. TQ radiation fractions and normalized CQ durations, shown as a function of the neon quantity in variable composition mixed species pellets (shown logarithmically). The scaled ITER targets for these quantities are shown in blue. [12]). The range of injection quantities anticipated for ITER (up to 10 kpa-m 3 [7]), scaled to DIII-D based on this assumed W th scaling, is highlighted in Fig. 4. Radiation fractions are observed to approach saturation within this range, indicating that TQ properties can be widely varied within the anticipated injection quantities. The resulting CQ durations following mixed species SPI are shown in Fig. 5, with the neon quantity now shown logarithmically (and the corresponding radiation fractions shown again for comparison). By varying the pellet mixture through the full range between pure deuterium and pure neon, CQ durations can be varied by over a factor of two in otherwise identical plasma conditions. The poloidal halo current amplitudes during these CQs and the resulting vacuum vessel displacements are also observed to be well controlled by the pellet composition, varying by a factor of two over the full range of Ne/D 2 mixtures [11]. Extrapolating to ITER (using the plasma cross-section normalization for CQ timescales), these results show a potential operational space in the SPI Ne/D 2 mixture where both TQ and CQ targets (shown in Fig. 5) may potentially be met. However, this allowable operating range may be rather narrow, implying that similar characterizations during the early operation of ITER (when stored energies and plasma currents are relatively low) are likely to be necessary.

5 5 EX/ Dissipation of Runaway Electron Beams The impact of a large runaway electron beam on in-vessel components can cause severe localized damage, and avoidance or mitigation of such events remains a critical challenge. The injection of large particle quantities may collisionally suppress runaway electrons and prevent the formation of large beams, although the required densities may be large [3]. Previously, the SPI technique has demonstrated the highest achieved CQ densities on DIII-D approaching ~20% of the Rosenbluth density [4]. However, the complexities of the runaway electron generation and damping mechanisms leave large uncertainties in the densities required for full suppression on ITER. Therefore, techniques for dissipating a fully formed runaway electron plateau remain important, in the event that collisional suppression cannot be reliably achieved. In DIII-D experiments, runaway electrons are reliably formed through the injection of small 2 Pa-m 3 argon killer pellets. During the subsequent runaway plateau phase, the residual argon in the background plasma from this initial injection dominates the dissipation of runaways [13], but feedback control is used to maintain the total runaway current at 0.25 MA. Injection of neon and deuterium shattered pellets (of various mixing ratios) into this runaway beam has been tested, and compared with equivalent MGI. Since the runaway plateau phase is relatively long-lived in this scenario, the differences in particle delivery timescales between the two injection techniques are less critical for this application. However, the subsequent interaction with the FIG. 6. Neon SPI dissipation of a 0.25 MA runaway electron (RE) beam. The loop voltage during the runaway plateau varies as the feedback control attempts to maintain the target runaway current, indicating changes in the total dissipation. runaways is found to depend on the choice of injected species, with neon and deuterium having nearly opposite effects on the runaway beam. Neon SPI results in a significant dissipation of the runaway plateau current. This is shown in Fig. 6, following the injection of a shattered pellet with 170 Pa-m 3 of neon and 7 Pa-m 3 of deuterium (which is used to form the outer part of the pellet due to its lower shear strength compared to neon, assisting with the pellet breakaway). The SPI occurs once the target runaway current is matched, with the rapid assimilation of the shattered pellet enhancing the runaway current dissipation, raising the externally measured loop voltage which is applied by the plasma current feedback system as it attempts to maintain the target current. The initial current decay rate following SPI is similar to that due to the same quantity of pure neon MGI, which appears to be dominated by collisions with impurity ions [14]. These SPI dissipation results are in contrast to similar experiments on JET where high-z (including argon, krypton,

6 6 EX/9-2 and xenon) MGI had no significant effects on the runaway beam [15], although this difference is not yet fully understood. In contrast to neon, deuterium injection with either technique results in a drop in the effective resistivity and background plasma electron density. This occurs simultaneously with a drop in spectral lines from argon, suggesting a reduction in the residual argon content of the background plasma which may be responsible for the reduced dissipation. Similar results have been seen for MGI with other low-z species such as helium [14]. For injections of species mixtures, deuterium is observed to have the dominant effect over high-z impurities, with a decrease in dissipation observed for relatively low deuterium quantities. For the shot shown in Fig. 6, even the small quantity of deuterium in the pellet (8% of the total injected atoms) is believed to reduce the overall dissipation (from that which would have resulted from pure neon), resulting in the residual runaway current observed afterwards. This is supported by similar results with MGI, where pure argon injection achieves complete dissipation of the current, but mixtures with 18% deuterium atoms result in a residual runaway current, similar to that observed in Fig. 6. Upgrades of the SPI system to allow injections of deuterium-free pellets (with a mechanical punch to assist the pellet breakaway) are planned in order to test this hypothesis. More detailed studies of the relative roles of high- and low-z impurities are planned, but these initial results indicate that careful selection of the injected species will be necessary in order to optimize injection schemes for runaway dissipation. 5. Summary and Discussion SPI has demonstrated effective mitigation of TQ and CQ loads that scale to meet ITER disruption mitigation requirements. Detailed comparisons of SPI with the MGI technique have shown that the SPI approach achieves improved TQ mitigation, with better protection of the divertor due to faster injection timescales, higher global assimilation, and improved core particle deposition. Particle delivery rates for SPI also extrapolate to ITER more favorably than for MGI, since it is not limited by gas conductance between the injector and the plasma surface. Further flexibility of the mitigation process is achieved through variation of the pellet composition using the mixed species pellet approach. Variation of the neon quantity in these pellets allows control of TQ and CQ properties within scaled ITER injection quantities, allowing a possible approach for tuning disruption characteristics to meet ITER targets. SPI has also demonstrated the dissipation of an existing post-disruption runaway electron beam with neon pellets, but efforts to determine optimum pellet mixtures are ongoing. SPI research on DIII-D continues to address important issues for the final design and operation of the ITER DMS. A second SPI system using a three-barrel ITER prototype design [7] is currently under installation on DIII-D. This system combined with the existing injector significantly extends the DMS capabilities, allowing ITER-relevant synchronized and/or successive multi-pellet injections, including toroidally separated independent injection capabilities. Future studies with this expanded system will investigate the cumulative nature of multiple shattered pellets, the impact of SPI trajectories on particle assimilation, 3D effects including potential radiation asymmetries, and continued optimization of injection schemes for runaway electron dissipation. In addition to the existing efforts, scaling and extrapolation of SPI results to ITER will require experimental characterization on other devices (such as JET), as well as the development and empirical testing of 3D extended MHD models for the SPI process, similar to those developed for MGI [16,17]. Improved understanding of the ablation and propagation of pellet shards in hotter and higher current plasmas, the mechanisms and timescales for toroidal/poloidal spreading of the shattered pellet material, and the impact of the resulting particle source

7 7 EX/9-2 distribution on radiation asymmetry, will allow more confident projections of SPI performance and optimization of ITER DMS operations. 6. Acknowledgment This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Fusion Energy Sciences, using the DIII-D National Fusion Facility, a DOE Office of Science user facility, under Awards DE-AC05-00OR22725, DE-FC02-04ER54698, DE-FG02-07ER54917 and DE-AC52-07NA DIII-D data shown in this paper can be obtained in digital format by following the links at References [1] HENDER, T.C., et al., Nucl. Fusion 47 (2007) S128 [2] LEHNEN, M., et al., J. Nucl. Mater. 463 (2015) 39 [3] ROSENBLUTH, M.N., PUTVINSKI, S.V., Nucl. Fusion 37 (1997) 1355 [4] COMMAUX, N., et al., Nucl. Fusion 50 (2010) [5] COMBS, S.K., et al., IEEE Transactions on Plasma Science 38 (2010) 400 [6] COMMAUX, N., et al., Nucl. Fusion 51 (2011) [7] BAYLOR, L.R., et al., Fusion Sci. Tech. 68 (2015) 211 [8] N.W. Eidietis, Nucl. Fusion 55 (2015) [9] COMMAUX, N., et al., Nucl. Fusion 56 (2016) [10] HOLLMANN, E.M., et al., Nucl. Fusion 48 (2008) [11] SHIRAKI, D., et al., Phys. Plasmas 23 (2016) [12] ROMANELLI, F., Nucl. Fusion 55 (2015) [13] HOLLMANN, E.M., et al., Nucl. Fusion 53 (2013) [14] HOLLMANN, E.M., et al., Phys. Plasmas 22 (2015) [15] REUX, C., et al., Nucl. Fusion 55 (2015) [16] IZZO, V.A., Phys. Plasmas 20 (2013) [17] IZZO, V.A., et al., Nucl. Fusion 55 (2015)