TH/P1-19

PlasmaDisruptionandVDEModelinginSupportofITERI.Bandyopadhyay1,a, A.K. Singh1, N. Eidietis2, D.A. Humphreys2, R. Granetz3,

G. Pautasso4, A. Isayama5, S. Miyamoto6 and S. C. Jardin7

1. ITER-India, Institute for Plasma Research, Bhat, Gandhinagar, India 2. General Atomics, PO Box 85608, San Diego, CA 92186-5608, USA 3. Massachusetts Institute of Technology, Cambridge, MA, 02139, USA 4. Max-Planck-Institut für Plasmaphysik, Association EURATOM, Garching, Germany 5. National Institute of Quantum and Radiological Science and Technology, Naka, Ibaraki 311- 0193, Japan

6. Research Organization for Information Science Technology, Tokai-mura, Japan 7. Princeton Plasma Physics Laboratory, Princeton, New Jersey, USA a.Email: indranil.bandyopadhyay@iter-india.org

Abstract: Accurate modeling of major disruption and vertical displacement events in ITER is necessary to determine the halo current amplitude during these events and hence the electromagnetic loads on the machine components. Predictive simulations for MD and VDE events in ITER have been carried out using DINA and TSC codes. However, in these simulations, the halo current amplitude depends critically on the choice of the halo parameters, namely the temperature and width of the halo region. Due to lack of credible experimental data of these two parameters and also no sound physics based model so far, these parameters are chosen rather ad-hoc in the simulations. For validating simulations with existing experiments, these parameters are chosen carefully for each experimental discharge so as to give a good match between the experiments and simulations. But for predictive simulations for ITER, this creates a problem as to what values to be chosen for these parameters. To resolve this issue, a concerted effort to validate the TSC model against a wider set of experiments in different machines is presently underway. We have selected a set of four shots in DIII-D and 5 shots CMOD, which are being simulated in TSC. The halo parameters are set carefully only for one experiment in each machine and for the rest of the shots, they are kept unchanged. Thus the difference between the experimental and simulated halo current amplitude in these discharges would give an indication of the possible error in predictive modeling. We have already modeled three DIII-D discharges and 1 CMOD discharges in which, we can reproduce the peak halo current amplitude within about 10-15% of their experimental value. 1. Introduction Major Disruptions (MDs) of the plasma current and Vertical Displacement Events(VDEs)areamajorconcerninanytokamak,astheyarenotonlyahindrancetosteadystate operations of the machine, but also subject the machine components to largeelectromagnetic forces. Therewill be no exception in ITER, but especially due to thelargeplasmacurrentof15MAinITER,thehalocurrentcanalsobehighandthatcrossedwiththeITER’slargetoroidalmagneticfield,cansubjectthemachinetounprecedentedJxBforces.Thus,theITERfirstwall,vacuumvesselandallin-vesselcomponentsmustbedesignedtowithstandtheselargeforces.TheprojectedhalocurrentamplitudeduringITERMDsandVDEsarepresentlybasedonDINAsimulations [1] andextrapolations from thedatabaseof theseevents inpresentdayexperiments.However,theproblemwiththisapproachistwo-fold:first,itwouldbeextremely difficult to design and build ITER or any reactor grade tokamak ifextrapolations fromtheoutliersofpresentexperimentaldatabaseareconsidered,e.g.,the non-axisymmetric VDEs in JET. Secondly, in themodeling predictions fromDINA,which were also benchmarked with TSC simulations, there are uncertainties in thepredictions due to themodel assumptions.While the TSC andDINA code predictionsmatch verywell with each other [2], especially when similarmodel assumptions areused, the most significant uncertainty in the model assumptions lie in the key halo

TH/P1-19

parameters, namely the temperature of the halo region (Thalo) and width of the haloregion(Whalo).Thesearetwokeyparameters,whichdeterminefinallytheamplitudeofthe halo current. Generally in both DINA and TSC simulations, due to a lack ofexperimentaldataonThaloandWhalo,asalsotheabsenceofasoundphysicsmodelmeansthattheirvaluesareusuallyadjustedad-hoctobestmatchexperimentaldata.To analyse in detail the effect of these two key halo parameters on the halo currentamplitude and the projected MD/VDE forces in ITER, a concerted effort is presentlyunderwayunderaworkinggroupWG-10oftheITPAMHDtopicalgroup,tobenchmarkthecodeswithawidevarietyofexperimentaldata.Inthepast,theTSCcodesimulationswere benchmarked with experiments in ASDEX-UG and NSTX, which was reportedearlier[3].ThesewerefollowedbyfurtherbenchmarkingoftheTSCcodeagainstfourdifferentVDEdischargesinASDEX-UG[4].Figure1givesanexampleofcomparisonof

theTSCsimulationresultswiththeexperimentaldatainASDEX-UG–thetoprowshowsthe evolution of plasma current (Ip), poloidal halo current (Ihalo) and plasma poloidalbetainsimulationandinexperiments.ObviouslythereisverygoodmatchbetweentheTSCsimulationsand theexperimentsvalues.But theproblem lies in thechoiceof theWhaloandThalousedinthesimulationsshowninthebottomrow.Itistobenotedthatineachofthesesimulations,WhaloandThaloareadjustedtogetthebestmatchbetweenthesimulated and experimental data. This poses a problem in predictive simulations forITERindecidingwhataretheidealvaluesofThaloandWhalotobechosen.Toresolvethis,theWG-10 ispresentlycarryingoutbenchmarkingsimulationswithTSCwithawidersetofexperimentaldata,mainlywithrepresentativedischargesintheDIII-DandCMODtokamaks.InDIII-Dwehaveselectedasetof4VDEdischarges,whileinCMODwehaveselected 3MD and 2 VDE discharges for these simulations, which best represent thevarietyofsuchscenarios in thesetwomachines. Inall thesedischargesthereareverygood measurements of the halo currents at poloidally different tile locations, which

Figure1:TSCsimulationsofdisruptionshots25000,24689and23297inASDEX-UG.Thetoprowgivestheevolutionofsimulated(solid lines)andexperimental(dashed lines) Ip, Ihaloandpoloidalbeta.Thelowerrowgivesthewidthandtemperatureofthehaloregionsetinthesimulations

TH/P1-19

should help in benchmarking the halowidthmodel used in TSC. The simulations arecarriedoutinTSCforthedisruptive/VDEphaseofthesedischarges.2.TSCSimulationsofthedisruption/VDEshotsThesimulationsarestartedwiththe initialplasmaequilibriumabout4-5millisecondsbeforetheonsetofthethermalquench(TQ)incaseofMDshotsandaftertheonsetofVDE for the VDE shots. The plasma equilibrium at that experimental time point iscarefully reproduced in TSC to have a very good match between the experimentalreconstructed and simulated equilibria including the shape, location of the magneticaxis, density and temperature profiles, poloidal beta and internal inductance. Themachinegeometry,includingthevacuumvesselandfirstwallcomponentsarecarefullyrepresentedinverygooddetailinTSC,withpossiblepoloidalhalocurrentpathsclearlydefined.Theplasmaevolution is then followed inTSCwithexperimentalcoil currentssetasinputstothesimulations.ThecurrentpeakingposttheTQissimulatedthroughahyper-resistivity parameter [5] introduced in the Ohm’s law. Beyond the currentpeaking the current quench (CQ) occurs naturally in the simulations with a ratedeterminedmainlybythepostTQplasmatemperature,theplasmaverticalgrowthrate,theevolvingpoloidalfieldconfigurationandthemachinegeometry.Thehalocurrentisswitched on in the simulations at the time when the plasma becomes limited. Thetransport equations are not solved in these simulations and the TQ is simulated byartificiallydroppingtheplasmapressuretoemulatetheexperimentaldataofthecentralelectrontemperature.In these simulations, thehalo currentparametersThaloandWhalo are set carefullyonlyfor one of the representative discharges in the machines. For the simulations of theotherexperimentaldischarges forthatmachine, theseparametersarekeptunchangedandonlytheinitialplasmaandcoilcurrentparametersareusedtomimicapredictivemodeling scenario. Thus we believe the difference between the experimental andsimulated halo current amplitudes for these discharges would give an idea of thepossiblepercentageerrorinthepredictivesimulations.3.ModelingofDIII-DdischargesForDIII-Dwehaveselectedfourdischargesforthesesimulations,namelyshotnumbers154143,154144,158207and144838,allType-IVDEsinDIII-D.Wehavecompletedthesimulations forthe first threeshots,whilesimulations forshot#144838arepresentlyongoing.Figure2belowgives theexperimentaldataofplasmaand totalpoloidalhalocurrents,plasmaverticalpositionandthecentralelectrontemperaturemeasuredinthesoftX-raychannel.WehaveadjustedtheThaloandWhaloparametersonlyfortheshot#154143,whichisanunmitigated forced downward VDE discharge triggered right from an H-mode at2000msec.TheplasmacolumngoesintoadownwardVDE,touchesthebottomwallanda fast thermalquenchoccursbetween2030.73-2030.81msec, thus in about80μs.Theshot#154144 is almost identical to#154143until the currentquench, but thereafterhasaslowercurrentquenchandlarger/longerhalopulsethan#154143,probablyduetodecreasedgassingfromthewall.Theshot#158207incomparisonhasamuchfasterVDE motion with shorter duration but almost equally high halo current. In the

TH/P1-19

simulations of these discharges it is of utmostimportance to have the DIII-D vessel, in-vesselcomponents and poloidal field coil system to bemodeled inTSCasaccuratelyaspossibleandalsotohavetheinitialTSCcalculatedplasmaequilibriaat the start of the simulations to have almost

identical to the reconstructedDIII-D equilibria.We have created forDIII-D a detailedmodelofthevacuumvesselandthein-vesselcomponentsandPFcoilsystemona2cmx2cmcomputationalgridwiththepoloidalhalocurrentpathinthefirstwallandvacuumvesseldefinedinreasonabledetail.WehavefirsttestedtheL/Rtimeofthevesselandpassivestructureagainst theengineeringparameter for them=1modeand it theTSCcalculatedvalueof4.6msecand6.2msec respectivelywithonlyvessel andvesselplus

Figure3: (Clockwise from top left) Evolutionof theplasmacurrent, verticalposition, centralelectron temperature (Soft X-raymeasurements) and the halo current in DIII-D forced VDEshots154143,154144and158207,whicharemodeledinTSC.

Figure 2: TSC simulated equilibria (toprow) and EFIT reconstructed equilibria(bottom row) for DIII-D shots 154143,154144and158207atthetimeofstartofthesimulations.

Table1:ComparisonofequilibriumparametersoftheDIII-Dshots154143,154144and158207inEFITreconstructionsagainstTSCcalculatedvaluesatthetimeofstart(showninparenthesisatthetop)ofthesimulations.

TH/P1-19

other in-vessel components are identical to the engineering value. The initial plasmaequilibriaatthestartofthesimulation, forthethreedischargessimulatedisshowninFigure 3 and the equilibrium parameters as calculated in TSC as against EFITreconstructedvaluesareshowninTable1.Inthesesimulations,theequilibriumcontrolsystems are switched off after calculations of the initial equilibrium and the plasmacolumnisallowedtodrift freely,withthepoloidal fieldduetothePFcoilcurrentssetidenticaltotheexperimentalvalues.Thehalocurrentmodelisswitchedonassoonastheplasmabecomeslimited.

Figure4:TSCsimulatedfluxsurfaceevolutionofthedisruptingphaseoftheVDEshot#154143.Theorangeregionoftheopenfluxsurfacesrepresentsthehaloregion.

Figure4 shows theequilibrium flux surfacesof thedisruptingplasma inVDEand thehaloregionasmodeledbyTSC.Itisinterestingtonotethatduringtheevolutionofthehalo region in thismodel, thehalo region canbe in contact simultaneouslywithbothupperaswellasbottom-innerpartofthefirstwall,asisseeninthefluxsurfaceplotattime2030msec.Suchahaloregionwouldallowhalocurrentstoflowfromtheplasmaboth to the lower-inner as also the upper part of the first wall and their supportstructurestothevacuumvessel,whichshouldshowinthepoloidaldistributionof the

halocurrentmeasurements.Wehavenotdonethedetailedinvestigationofthepoloidalhalocurrentdistribution,whichweintendtodoinfuture.Theevolutionoftheplasmaand halo currents and the plasma vertical position as modeled by TSC and in theexperiments,asalsotheassumedhalowidthevolutioninthemodel forshot#154143areshowninFigure5.Thehalocurrentmodelisswitchedonafter2020msecwhentheplasmabecomeslimitedanditswidthisraisedlinearlyfrom0to0.5(fractionofplasmaflux in the halo region) in 5msec and thereafter kept constant, while the halo

Figure 5: Simulated and experimental plasma current, halo current, plasma vertical position and the halowidthtemporalprofileusedinthesimulationforDII-Dshot#154143.

TH/P1-19

temperature is kept constant at 6.5eV. It is evident that TSC can follow the plasmacurrentevolutiontoverygoodaccuracy,exceptforsomedifferenceinthefinaldyingIpevolution,duringwhich theexperimentaldatashowsasmoothercurrent termination,which TSC cannot follow. The vertical position is also followed in TSC to very goodaccuracy.Thedifferencebetween theexperimentalZpafterabout2031msec isdue tothefactthattheexperimentaldataconsidersthecurrentcenteroftheplasmaincluding

the expanding halo region,while TSC plots only thecenter of the closed fluxregion. The halo widthevolution temporal profileused for #154143 is keptidenticaltotheother2DIII-Dshotsmodeled–#154144and #158207. Figure 6shows the experimentaland simulated plasma andhalo current evolution inthese two shots. In these

simulationsalso,TSCcanfollowboththeIpquenchandthepeakhalocurrentamplitudeto very good agreement with the experiments and the peak halo current amplitudematchingtowithin10-15%oftheexperimentalvalueinthesimulations.Althoughthereis some difference in the durations of the halo current, the reasons forwhichwe arepresentlyanalyzing.4.ModelingofCMODdischarges

For CMOD, we have selected theshots 1020801010, 1020801025,1021010022, 1020903026 and1020904014.Ofthese,thefirstthreeareMDshots,whilethelasttwoareVDEs.Theshotswereselected fromthe database carefully with havingmaximum Ihalo/Ip fraction, initialplasma current before onset ofMD/VDE above 1MA. Of particularinterest are the shots 1021010022,which clearly shows signature ofrotating halo current and the VDEshot 1020904014, which shows apolarityreversalinthehalocurrent.Figure7 shows the evolutionof themeasuredplasmaandhalocurrents,vertical plasma centroid positionand the central electrontemperature in the MD shot1020801010onthe leftcolumnandthe VDE shot 1020903026, where

Figure 7: Evolution of plasma and halo currents, verticalpositionandcentralelectrontemperatureinCMODdisruptionshot 1020801010 (left column) and VDE shot 1020903026whicharemodeledinTSC.

Figure 6: Experimental (blue) and TSC simulated (red) Plasma andpoloidalhalocurrentevolutioninDIII-DVDEshots154144and158207.

TH/P1-19

wehaveplottedonlythepartofthedischargesmodeledinTSC.TheotherMDandVDEdischargesalsohavesimilarbehaviour for theseparameters. It is tobenoted that theMD discharges startwith the TQ, followed by the CQ and vertical drift of the plasmacolumn, while in the VDE discharges start with the vertical drift, while the plasmatemperature and current remains constant till the edge safety factor becomes low,finallyleadingtoTQduetoMHDeventsandsubsequentlytheCQ.ThisgeneralbehaviorisflowedinalltheMD/VDEdischarges.ThustheMDshotsimulationsarealsocarriedout by first calculating the plasma equilibriumabout 5msec prior to theTQ and then

emulating the TQ by artificiallydroppingplasmapressure to followtheexperimental central electrontemperature. In the VDE shotsimulations, the plasma equilibrium atthe start of the VDE is reproduced andtheplasmaisallowedtomovefreelybyswitching off the control systems andthe thermal quench emulated similarlybydroppingthepressure.Wehavecarriedoutsimulationsfortheshot#1020801010andwepresentherethe preliminary results, whilesimulationsfortheotherdischargesarein progress. For the CMOD simulationsalso, we have carefully calculated theinitial plasma equilibria for the shots atthestartofthesimulationandvalidateditwithEFITreconstructedequilibria.Figure8 shows the comparison of the EFITreconstructed equilibrium with the TSCcalculated equilibrium at 940.0msecwhen we start the simulation for thisshot,which showsaverygoodmatch. Infact we have calculated the initialequilibria for all the CMOD shots to bemodeled and like in theDIII-D cases, forall of these shots, the TSC calculatedplasma equilibrium parameters matchclosely with the EFIT reconstructedvalues.Theevolutionoftheexperimental

and TSC simulated plasma and halo currents between 940-950 msec are shown inFigure 9. While TSC is able to follow the Ip quench to good accuracy, there is somedifference.TheexperimentalIpquenchshowsclearlytwodifferentdecayrates–oneabitslowerquenchafterthecurrentpeakingtill946.5msec,thenafasterquenchbetween946.5-947.5msec. This could be due to a small change in the residual plasmatemperaturepostTQ,whichisdifficulttodeterminefromtheexperimentaldataduetohigh noise and very small temperature value post TQ. We have assumed a constantplasma temperature of 10eV in the simulations, which is kept constant post thermalquench.Wearestillinvestigatingifassumingdifferentplasmatemperatureduringthese

Figure8:EFITreconstructed(left)andTSCcalculatedplasmaequilibriumforCMODshot#1020801010at940.0msec,whichshowsverygoodmatch.

Figure 9: Plasma and halo current evolution in TSCsimulations and experiment for CMOD shot#1020801010

TH/P1-19

times can reproduce the dual slope in the Ip quench. We are able to reproduce thenatureofthehalocurrentevolutioninCMODforthisdischarge,whichshowsaninitialperiodoflowhalocurrent,finallyterminatingwitharapidpeakingto126.15kA.Inthesimulationsalsothereisaninitialperiodofabout2.5msecoflowhalocurrentandthenarapidpeakingtoabout114.6kA,althoughthetimeofthepeakingofthehalocurrentsdiffer in thesimulations– itappearsaboutamillisecond later than in theexperiment.ThiscouldbeduetothedifferenceinslopesintheIpquench,whichweareinvestigatingnow.Figure10showstheevolutionofthefluxsurfacesinthissimulationwiththehalocurrentmodelswitchedonfrom945msec.Duringthefastcurrentquenchphase,asthecoreplasma column shrinks and thehalo region expands, thewetted areaof the firstwallchanges,whichisalsodeterminedtoalargeextentbythemachinegeometry.Theeffect of this on the poloidally distributed halo current measurements should becarefullyanalysed.

5.SummaryBenchmarking simulationsofmajordisruptions andVDEs for anumberofDIII-D andCMODdischargesarebeingcarriedoutinTSCtohavemoreinsightintotheadhochalocurrentmodelsusedintheseandothersimilarsimulations.Webelievethattheresultsof these benchmarking simulations should provide a good estimate of the possibleerrorsinthepredictionofthehalocurrentamplitudesarisingfromuncertaintiesofthehalo parameters. We have so far carried out the benchmarking simulations in 3dischargesinDIII-DandoneinCMODandinallthesecaseswecanreproducethepeakhalocurrentamplitudetoabout10-15%oftheexperimentalvalueevenwiththead-hocmodel.We have also implemented new diagnostics in the TSC code to get the spaceresolved halo current distribution around the separatrix point. The simulated spatialdistributionof thehalocurrentwillbecomparedwiththeexperimentaldata inDIII-DandCMODwith themeasurements in theRogowski/partialRogowskicoils in the tilesand support structures. Thiswill give a very good idea of howwell the experimentalhalocurrentwidthsarereproducedinthesimulations.References: [1] M. Sugihara et al, Nucl. Fusion 47 (2007) 337. [2] S. Miyamoto et al, Nucl.Fusion54(2014)083002 [3] I. Bandyopadhyay et al, IAEA FEC 2010, IT/P-16. [4] S. Miyamoto et al, ITER Task Agreement C19TD28FJ final report, JADA report no. JADA-00914PRfinal [5] A.H. Boozer, 1986 Plasma Physics 35, 133.

Figure10:FluxsurfaceevolutioninTSCsimulationsofCMODshot#1020801010