High Fusion Performance from Deuterium-Tritium Plasmas in JET

Transcription

1 JET (98)36 M Keilhacker et al High Fusion erformance from Deuterium-Tritium lasmas in JET

2 This document is intended for publication in e open literature. It is made available on e understanding at it may not be furer circulated and extracts may not be published prior to publication of e original, wiout e consent of e ublications Officer, JET Joint Undertaking, Abingdon, Oxon, OX14 3EA, UK. Enquiries about Copyright and reproduction should be addressed to e ublications Officer, JET Joint Undertaking, Abingdon, Oxon, OX14 3EA.

3 JET (98)36 High Fusion erformance from Deuterium-Tritium lasmas in JET M Keilhacker, A Gibson, C Gormezano, Lomas, R Thomas, M L Watkins, Andrew, B Balet, D Borba, C D Challis, I Coffey, G A Cottrell, H L De Esch, N Deliyanakis, A Fasoli, C W Gowers, H Y Guo, G T A Huysmans, T T C Jones, W Kerner, R T W König, M J Loughlin, A Maas, F Marcus, M F F Nave, F G Rimini, G Sadler, S Sharapov, G Sips, Smeulders, F Söldner, A Taroni, B J D Tubbing, M G von Hellermann, D J Ward, and e JET Team. JET Joint Undertaking, Abingdon, Oxfordshire, OX14 3EA, See Appendix I reprint of a aper to be submitted for publication in Nuclear Fusion October 1998

4 ABSTRACT High fusion power experiments using D-T mixtures in ELM-free H-mode and optimised shear regimes in JET are reported. A fusion power of 16.1 MW has been produced in an ELM-free H-mode at 4.2 MA / 3.6 T. The transient value of e fusion amplification factor was.95±.17, dia consistent wi e high value of n DT () τ E Ti()= m -3 s kev±2%, and maintained for about half an energy confinement time until excessive edge pressure gradients resulted in discharge termination by MHD instabilities. The ratio of D-D and D-T fusion powers (from separate but oerwise similar discharges) showed e expected factor of 21, validating D-D projections of D-T performance for similar pressure profiles and good plasma mixture control which was achieved by loading e vessel walls wi e appropriate D-T mix. Magnetic fluctuation spectra showed no evidence of Alfvénic instabilities driven by alpha particles, in agreement wi eoretical model calculations. Alpha particle heating has been unambiguously observed, its effect being separated successfully from possible isotope effects on energy confinement by varying e tritium concentration in oerwise similar discharges. The scan showed at ere was no or at most a very weak isotope effect on e energy confinement time. The highest electron temperature was clearly correlated wi e maximum alpha particle heating power and e optimum D-T mixture; e maximum increase was 1.3±.23 kev wi 1.3 MW of alpha particle heating power, consistent wi classical expectations for alpha particle confinement and heating. In e optimised shear regime, clear internal transport barriers were established for e first time in D-T, wi a power similar to at required in D-D. The ion ermal conductivity in e plasma core approached neoclassical levels. Real time power control maintained e plasma core close to limits set by pressure gradient driven MHD instabilities, allowing 8.2 MW of D-T dia fusion power wi n DT () τ E Ti () 1 21 m -3 s kev, even ough full optimisation was not possible wiin e imposed neutron budget. In addition, quasi steady-state discharges wi simultaneous internal and edge transport barriers have been produced wi high confinement and a fusion power of up to 7 MW; ese double barrier discharges show a great potential for steadystate operation. 1. INTRODUCTION Significant controlled fusion power was first produced during e reliminary Tritium Experiment (TE) in e Joint European Torus (JET) in 1991 [1], when a hot ion H-mode plasma containing 11% tritium in deuterium produced 2 MJ of fusion energy wi a peak fusion power DT =1.7 MW and a fusion power gain Q in = DT / in =.12, in being e total input power to e torus. Tritium usage and neutron production were deliberately kept low during e TE in order to limit vessel activation so at a pumped divertor [2] could be installed ree mons later during a manned intervention. This was e first of a series of divertors installed in JET between 1991 and 1996 to test e effect of increasing geometric closure on divertor behaviour and plasma 1

5 performance wi deuterium plasmas closest in scale and geometry to at foreseen for e International Thermonuclear Experiment Reactor (ITER) [3]. Meanwhile, e Tokamak Fusion Test Reactor (TFTR) in e USA operated in deuterium-tritium (D-T) mixtures and, using 5% of D and 5% of T, produced 1.7 MW of fusion power and a fusion gain Q in =.27 in e supershot regime [4]. This limiter tokamak also explored improved confinement in e enhanced reversed shear (ERS) regime, but e performance improvements achieved in deuterium [5] could not be translated into D-T [4, 6]. Wi e closure of TFTR in April 1997, JET is now e only fusion experiment able to operate wi D-T fuel mixtures. Furermore, and going beyond TFTR, JET has an effective divertor and can operate in a wider range of plasma regimes. In addition, JET has e first industrial scale plant for e closed cycle supply and processing of tritium (e Active Gas Handling System, AGHS [7]), togeer wi a proven remote handling capability. These have allowed e JET programme for e period to e end of 1999 to include bo a broad-based series of D-T experiments (DTE1) to address crucial issues of D-T physics and technology for ITER, and e development and exploration of divertor concepts for ITER. During DTE1, e JET torus was pumped continuously rough e AGHS and was supplied wi D-T by e gas introduction and neutral beam (NB) systems. The tritium was stored in uranium beds and re-processed in e AGHS to a purity of 99.4% by gas chromatography. In contrast to e TE, when e total amount of tritium available was.2 g, ere was no equivalent restriction for DTE1. The site inventory of 2 g of tritium was re-processed eight times by e AGHS, making e equivalent of 99.3 g of tritium available for DTE1. The fusion energy produced was 675 MJ, equivalent to e limit set at is stage in e JET programme on e total neutron production of less an 2.5x1 2 neutrons (compared to e limit of about 1.5x1 18 neutrons imposed for e TE) so at e subsequent activation of e JET vessel would not prevent manned in-vessel intervention for more an a year after e end of DTE1. The JET programme continued after DTE1 wi an ITER Urgent hysics hase and e remote handling replacement of e relatively closed Mark IIA divertor [8] used during DTE1 by an even more closed gas box divertor (Mark IIGB). The physics objectives of DTE1 can be divided into two areas. The first, discussed in a companion paper [9], was e characterisation of e baseline ITER mode of operation, e steady-state ELMy H-mode, in D-T. Of particular importance were assessments of e hydrogen isotope effects on e H-mode reshold power, e energy confinement and e edge operational space, of various heating schemes at e ion cyclotron resonance frequency (ICRF) in D-T plasmas, and of high fusion power and Q in steady-state. The aims of e second area, which is e subject of is paper, were to produce maximum fusion power and Q, to observe alpha particle heating and to study e stability of Alfvén Eigenmodes. These aims are inter-related since high fusion power corresponds to high alpha particle pressure, ereby maximising e effect of e alpha particles for heating and destabilising of Alfvén Eigenmodes. 2

6 In D-D plasmas on JET, two modes of operation have consistently out-performed, at least transiently, e ELMy H-mode by a factor of about four in fusion performance. These are e hot ion ELM-free H-mode, which had already been tested in e TE and was considered to be e established route to high performance, and e optimised shear mode, which had been developed during e year leading up to DTE1. Their translation from D-D to D-T is of great interest for understanding ese regimes as well as producing high performance. There is also a long term interest in developing e latter regime towards steady state. These high performance discharges predominantly use high energy NB injection to heat and fuel e plasmas. During DTE1, e JET NB system was operated wi tritium in all eight beam sources of one of e two injectors. More an 1 MW of tritium NB power was delivered at voltages of 155 kv, for pulse durations 5 s. The oer injector operated exclusively in deuterium at 78 kv. ICRF was used to supplement NB heating. The experimental set-up, e heating systems and e D-T related diagnostics are discussed in more detail in e companion paper [9]. The paper is structured as follows. Section 2 considers e technical issues related to tritium plasma concentrations, tritium retention in e vessel wall and clean-up. The questions of fusion power performance (including e stability of Alfvén Eigenmodes) and e detection of alpha particle heating in e ELM-free H-mode are en addressed in Sections 3 and 4, respectively. Section 5 considers fusion power performance and physics in e optimised shear regime. The summary, conclusions and outlook follow in Section TRITIUM CONCENTRATIONS, VESSEL RETENTION AND CLEAN-U During DTE1 over 2 pulses had plasma tritium concentrations greater an 4%. Control of e isotopic mixture in e plasma was achieved by preceding e experiment wi a few ohmic or ICRF heated pulses to changeover e isotopic composition of e walls. Wi 1% tritium fuelling, plasma tritium concentrations greater an 9% were readily obtained (Fig. 1a). During e campaign, it was possible to achieve nearly complete tritium recovery (>98%) from e NB injector which operated in tritium. However, about 3% of e tritium input to e torus was retained in e torus (Fig. 1b). After e tritium experiments were completed, e experimental campaign continued wi about 2 mons of operation in deuterium and hydrogen, after which e torus tritium inventory had fallen to 17% of e torus input (6 g). The plasma tritium concentrations had fallen to.1% by is time, so it was clear at e tritium inventory in e torus was not contributing significantly to e isotopic mixture. The torus tritium inventory at e end of DTE1 was over 3 times larger an had been expected from e tritium retention results of e TE [1]. For e purposes of using e TE results to extrapolate to DTE1, a simple analytical fit to e TE data was used to represent e behaviour of retained tritium after each pulse. While is TE-based prediction closely followed e tritium concentrations, e difference in actual and predicted tritium inventories grew over 3

7 Sub-divertor T alpha e course of e campaign. This suggests at 1. Residual gas analyser a) Tritium concentrations ere was a sink for tritium during DTE1 which TE-based prediction was not present at e time of e TE. The high level of tritium retention during.8.6 DTE1 has been related to carbon films, satu- rated wi deuterium and tritium and located.4 in e divertor. While co-deposited films have.2 always been present in e JET vessel, e films wi e Mark IIA divertor were unprecedentedly ick and hydrogen rich. The films were found in cold regions of e divertor, shadowed from direct contact wi e plasma. While e majority of e JET vessel was heated to 32 C, ese cold regions are actively cooled b) Tritium inventory to 5 C (ey act as a heat shield to protect e in-vessel divertor coils), allowing e formation Cumulative Torus input 1 Torus inventory of stable films wi more an 4% TE-based prediction hydrogen concentrations in carbon [11]. In ulse Number contrast, at e time of e TE (i.e. before e FIG. 1. (a) Tritium concentrations and (b) tritium inventory for e second period of DTE1 from September installation of a divertor), e whole vessel interior was maintained at 3 C. The films which formed on water-cooled metal surfaces were found to flake off, probably on venting, to November 1997 (ulses No to 4323). Note at ere had already been a first period of D-T operation in May/June The tritium concentrations shown were measured by sampling neutrals in e sub-divertor leading to considerable debris in e bottom of volume during pulses, and sampling gas by residual gas e divertor. Even ough e hardware is similar analysis between pulses. The TE-based prediction is on e inner and outer legs of e divertor, based on exhaust measurements during e TE. e flakes were found only on e inner leg. The processes leading to is asymmetry and e high level of carbon erosion needed to form e films are still under investigation. Analysis of tiles and flakes removed from e vessel after e experimental campaign is underway and expected to identify e exact location of e tritium inventory still inside e vessel. This analysis will also determine e physical and chemical characteristics of e films. T / (T+D) Tritium inventory (g) JG98.38/8c 3. THE ELM-FREE H-MODE The ELM-free H-mode has been developed in JET over a wide range of plasma currents ( MA) and toroidal magnetic fields ( T) wi ree different divertor configurations. It has been e established route to high fusion performance in deuterium and was successfully used for e TE [1]. High values of e fusion triple product, n DT ()τ E T i (), were 4

8 observed and transient Q DT 1 (as defined in [1]) was suggested on e basis of TRANS extrapolations from pure deuterium plasmas to 5:5 D-T mixtures [12]. In e TE e regime was ought to be limited by e power handling of e divertor target tiles rough e occurrence of a carbon impurity bloom [1]. Following e installation of e Mark I pumped divertor [2], which had sufficient power handling capability to avoid blooms, it was recognised at e performance limitations were of MHD origin [13] and at highly shaped plasmas, high plasma current and low recycling were necessary to maximise e ELM-free period [14]. However, outer modes [13], identified as external kinks associated wi high edge bootstrap currents [15], could still degrade performance, but could be ameliorated by ramping down e plasma current The D-D Experience and Extrapolation to D-T In order to permit a comparison between e fusion performance in D-D and D-T it is first necessary to discuss how deuterium discharges were expected to translate to D-T. Despite e changes to JET and its divertor since e TE, e neutron yield in pure deuterium, Y DD, has continued to scale strongly wi stored diamagnetic energy, W dia, wi roughly 2 constant reactivity, Y DD / W dia. This scaling covers e dependence of e neutron yield on e squares of e ermal and fast particle energies and eir product as expected from ermonuclear, beam-beam and beam-plasma reactions, and variations in e reactivity will arise from variations in profiles, T i, T i /T e, Z eff and beam energy. It is clear erefore at ese parameters have not varied significantly in is regime. D-T simulations using TRANS have been made for a number of such pulses, assuming transport coefficients deduced from e D-D pulses, i.e. neglecting isotopic effects, but including alpha particle heating. These simulations, which will be described in more detail in [16], also show a quadratic dependence between e predicted D-T fusion powers and diamagnetic stored energies (Fig. 2). It is important to point out at ese simulations suggest a fusion power multiplication factor of 21, consistent wi e profile weighted ratio of e fusion cross sections [17]. Figure 2 contains simulated fusion power versus simulated W dia and 21 times e measured D-D fusion power, DD, versus measured W dia. The two curves correspond to a quadratic dependence of 21x DD wi W dia, e first an upper bound of all e D-D data and e second a typical curve rough e combined NB and ICRF heating data (alough e bulk of is data is omitted for clarity). A large part of e scatter in predictions between e two curves indicate systematic variations in reactivity due to variations in T i / T e, Z eff and fast ion energy content. In particular, low power NB-only and high power combined heating tend to cluster about e lower curve, whereas high power NB-only data tend towards e upper curve. The TRANS simulations introduce additional scatter as a result of uncertainties in e kinetic measurements. Nevereless, ere is a clear strong improvement of expected performance wi increasing stored energy which provides a useful benchmark for e D-T measurements to be presented later. 5

9 eq DT (MW) x DD TRANS NB 5 NBRF W W W DIA (MJ) 435 eq FIG. 2. Equivalent D-T fusion power DT from D-T simulations based on D-D discharges using TRANS (see text) against simulated stored diamagnetic energy (circled symbols). The plain symbols are 21 times e measured D-D fusion power at e measured diamagnetic eq energy. The two curves are bo DT 2 Wdia; e upper curve delineates e upper bound of 21 DD for all ELM-free data in e period , and e lower curve delineates e typical trend of 21 DD for combined heating data in D-D for e Mark II divertor only. JG98.38/9c 3.2. Mixture Control in D-T The TE experiments wi 11% tritium injected by NB suggested [12] at e plasma mix reflected e NB mix, indicating at eier recycling was negligible (because e particle confinement was so good) or e injected species was reflected at e walls (raer an deuterium being desorbed). However, an experiment wi e Mark II divertor wi selective hydrogen loading of plasma facing surfaces suggested at e dilution of core fuelled deuterium discharges could be significant even under conditions of low recycling [18]. To clarify e relative contributions of NB fuelling and recycling to e plasma mix in e ELM-free H-mode regime, e NB mix was varied from pure deuterium to pure tritium whilst e walls were still predominantly deuterium and e D-T neutron rate was measured. Such a variation in NB mix was possible wi 1 MW of injected NB power by selecting appropriate beams from each of e two NB injectors. To interpret e results of is experiment it is only necessary to know e relative contributions to e neutron rate from beam-plasma reactions due to deuterium on tritium and tritium on deuterium, and from ermonuclear reactions. These were estimated from TRANS extrapolations from a pure deuterium discharge to D-T. The TRANS results were en scaled for different NB mixes by a simple model which assumed at e recycling contribution of deuterium was proportional to e total NB particle flux (deuterium and tritium), wi e constant of proportionality, f, to be determined experimentally. If recycling could be neglected (f=), e peak fusion yield would occur wi a NB mix of 45% tritium (weighted towards deuterium rich by e tritium on deuterium beam-plasma reactions). However, e maximum fusion yield was found experimentally wi 1% tritium NB injection, and e experimental dependence of fusion power wi NB mix indicated f=2, i.e at recycling dominated e plasma mix. TRANS analysis of is experiment qualitatively confirms e result of e simple model [19]. This experiment clearly demonstrated at it would be necessary to replace deuterium gas fuelling by mixed D-T fuelling and to load e walls wi an appropriate D-T mix for e high power experiments. 6

10 3.3. High ower ulses in D-T In all, eight high power ELM-free pulses were produced, as listed in Table I. The first, which used D-T mixtures for bo gas and NB fuelling whilst e walls were still predominantly deuterium, confirmed e conclusions of e mixture control experiment at higher NB particle fluxes. For e remaining experiments, e walls were loaded using 3-5 ohmic or ICRF heated pulses wi e gas fuelling adjusted until e measured D-T plasma mix was close to 5:5. The target density and pre-programmed gas flows were adjusted in 1 to 2 short (<.5 s) high power pulses. Finally, e full duration pulse was selected. Table I Maximum values for e eight high power ELM-free H-mode pulses in D-T ulse No Comment walls deuterium sawtoo limited NB+ω ch NB only NB+ω ch early NB+ω +2ω ch ct switch off Time (s) I (MA) B T (T) NB (MW) RF (MW) in (MW) DT (MW) Wdia (MJ) n T /(nt+n D ) Q in=dt / in Q ptot Notes: I p is e plasma current, B T is e toroidal field at R=3m. NB, RF and in are e power inputs to e torus from NB, ICRF and total, including ohmic. DT is e total fusion power as measured by e total neutron yield. W dia is e diamagnetic stored plasma energy. n T /(n T +n D ) is e fraction of tritium as measured by e neutral particle analyser. The quoted Q tot values have an uncertainty of ± 18%. The first series of experiments was performed wi e standard 3.8 MA / 3.4 T scenario as developed in deuterium (including e same.4 MA/s current ramp-down to stabilise outer modes) in order to permit a direct comparison between D-D and D-T for bo NB-only and NB supplemented by 3 MW of fundamental hydrogen minority ICRF heating. The NB-only case produced 12.3 MW of fusion power for 21.3 MW of NB power, compared wi a projected 9.2 MW fusion power for 19.1 MW of NB power from simulations based on e equivalent (but lower power) D-D discharge. Wi combined heating, 12.9 MW of fusion power was produced 7

11 wi 22.4 MW of additional heating power, compared wi a projected 12.2 MW for 22. MW input. The uncertainty in ese measurements arising from e accuracy of e neutron measurements is ±8%, so is fusion power clearly exceeded e previous record [4]. The second series of experiments was carried out at a higher toroidal field and plasma current (4.2 MA / 3.6 T) and took full advantage of e higher NB power available in tritium. Mixed hydrogen minority and second harmonic tritium ICRF heating was also tested [2], producing 12.4 MW of fusion power for a total heating power of 25.1 MW, e additional RF power providing only electron heating. A new record D-T fusion power, 16.1 MW, was obtained wi a total heating power of 25.4 MW which included NB heating (22.3 MW) togeer wi hydrogen minority ICRF heating (3.1 MW). The maximum value of e ratio of Q in = DT / in for ese high power pulses is shown in Table I to vary from.4 to.64, reflecting differences in fuel mix, Z eff, confinement and duration of e high performance phase. In comparison wi e highest fusion power discharge in TFTR wi Q in =.27 [4], e JET discharges produced higher fusion power for lower input power to e torus. Some time traces for e best JET pulse are shown in Fig. 3. In common wi e D-D experience, stored energy, fusion power and electron density rise monotonically wi time. The ion temperature levels off around 28 kev, significantly higher an e electron temperature which is about 14 kev. The discharge does not reach steady conditions, but is limited by MHD activity: first an outer mode, en a giant ELM (as described previously for deuterium plasmas) DT (MW) ower (MW) Energy (MJ) T i, T e (kev) I p (MA) D α + T α NB <n e > 1 19 m 3 Z eff <n e > B θ IN RF W DIA Z eff B θ Time (s) FIG. 3. Various time traces for e D-T pulse wi highest fusion power. Note at following e ELM e density rises to ~8x1 19 m -3 and remains off scale. JG98.632/2c DT (MW) MA 4.2 MA NB 5 NBRF TE.83W W DIA (MJ) W 2 FIG. 4. Measured D-T fusion power for all ELM-free plasmas in D-T against measured diamagnetic energy. The two points labelled TE are from e reliminary Tritium Experiment. The curves rough e data are e same as in Fig. 2. JG98.38/11c 8

12 This MHD activity manifests itself in e intensity of e Balmer alpha emission and, following detection of a giant ELM, e heating power is switched off to limit e total neutron production. In D-D, wi e heating maintained, such discharges evolved into a repetitive ELMy behaviour wi lower fusion performance. A more detailed comparison of e MHD behaviour in D-D and D-T plasmas will be presented in [21]. Figure 4 shows e measured D-T fusion power as a function of diamagnetic stored energy which can be compared wi e D-T projections of Fig. 2. Neutral beam heating only, combined heating (NBRF) and e higher current data are distinguished by different symbols. The low power mixture control experiments togeer wi e data from e alpha particle heating experiment (Section 4) are also included, as are e two 11% tritium pulses from e TE. The low power experiments clearly find an optimum reactivity and ese data, taken wi ose 2 at high power, are well described by DT W dia, indicating good mixture optimisation and high reactivity roughout e series, independent of power and plasma current. The high power NB only data is a little above e curve rough e combined heating data as tends to be e case in D-D (Fig. 2). The improved fusion performance at higher plasma current is a result of e higher stored energy; e 17 MJ attained is higher an achieved previously on JET or any oer magnetic plasma confinement experiment..4 Comparing e NB-only and NBRF data Total pressure (Ma) in Figs. 2 and 4, e ratio of e fusion power DT DT simulation in D-T to at in D-D is wiin ±1% of e DT expected ratio of 21 at e same stored energy, in contrast to e results obtained on TFTR DD [4]. Figure 5 shows at e shape of e total pressure profile in D-T is similar to at in D-D.1 Alpha pressure (Ma) and, wiin experimental uncertainties, is close to at projected for D-T. In fact, e difference between e D-D and D-T profiles reflects 435sim ρ~ a r e population of energetic alpha particles produced by e fusion reactions. On e face of FIG. 5. rofiles of total pressure computed by TRANS for D-D pulse No. 435, e D-T simulation based on it, ese observations confirm e assumption pulse No. 435, e D-T pulse No at e same plasma current (3.8 MA) and heating power, and e used in e extrapolations to D-T at e net highest fusion power pulse No (at 4.2 MA). The isotope effect on energy confinement is negligible. D-T pulse at 3.8 MA is intermediate between e D-D case and e D-T simulation TRANS Analysis of e D-T ulses JG98.38/12c Figure 6 shows at ere is excellent agreement between directly measured quantities and ose computed by TRANS. In particular, e measured diamagnetic energy is in excellent agreement wi at computed by TRANS, W dia = W + 3/2 W tot, where W is e f kinetic 9

13 (1 18 s 1 ) (MW) W dia (MJ) ƒ ulse No: TRANS run Y935 TRANS D α + T α (a.u.) Measured JG98.661/3c measurement of e ermal energy content and f W tot is e computed total perpendicular fast ion energy. It should be noted at TRANS (run Y935) computes e neutron yield directly IN from e kinetic measurements. The error bar 2 TRANS 1 on is computation reflects e uncertainties dw dia Measured dt in T i (r) and Z eff (r) assuming at ese are e 6 DT neutron rate Measured dominant uncertainties. The agreement is satisfying, indicating not only e quality of e 4 TRANS 2 measurements but also at TRANS contains.8 Thermonuclear fraction e appropriate physics..4 The measured derivative of e diamagnetic energy has a resolution which can be as short as 1 ms and resolves e effect of NB Time (s) interrupts and MHD events as can be seen by FIG. 6. Validation of TRANS analysis and kinetic data for pulse No by reference to direct measurements comparing wi e traces in and D α + T α. of diamagnetic energy, W dia, derivative dw dia /dt and Wi a TRANS time step of 1 ms and 2 neutron rate. The bottom panel shows e ermonuclear particles in e Monte Carlo NB slowing down fraction. calculation it is necessary to smoo over 3-5 ms in order to minimise statistical noise. Wi such smooing ere is reasonable agreement between measured and computed dw dia /dt over most of e pulse, even during fluctuations in Z eff (in magnitude comparable to e error estimates) which contribute to fluctuations in e kinetic measurement of dw /dt. For is reason e derivative of e ermal energy is computed from dw /dt=d(w dia - 3/2 W tot )/dt because e fast particle energy is almost constant. f A TRANS run (Y936) has been performed wi ese Z eff fluctuations removed, and is improves e agreement between measured and computed quantities and yet has a negligible (<5%) effect on e power balance. It is erefore concluded at e TRANS run (Y935) using only measured data provides an adequate description of e behaviour of fast ions and e ermal plasma. Figure 6 also shows e ermonuclear fraction (e ratio of ermonuclear power to total fusion power, f = DT/DT ) as computed by TRANS, which reaches a value of about.63. In order to estimate e uncertainty in e ermonuclear fraction it is valid, given e small uncertainty in e measured neutron rate, to constrain e computation to match e measured neutron rate exactly. This gives an uncertainty in f of ±.5 relative to e reference TRANS run. The measured and computed data for pulse No which produced e highest fusion power, are summarised in Table II at a time (13.13 s) when e plasma is still quiescent. The first point to note is e global power balance, which shows at of e total power input to e torus ( in =25.7 MW), e total power from external sources absorbed by e ermal plasma is 1

14 arameter lasma Toroidal NB power, RF Total current, I field, B power, Absorbed NB RF p T power input to torus, t external power, Total ermal heating power, = + ot heat Loss abs t power, αabs t in h abs h = -dw loss heat Values at / dt 3. 2 Table II arameters of interest for ELM-free H-mode pulse No wi e highest D-T fusion power, at e time of maximum Q tot (13.13 s) and maximum value of each parameter, if significantly different. The geometric parameters of is pulse are: minor radius a =.95m, major radius R = 2.92m, q 95 = 3.47, elongation κ = 1.81, triangularity δ =.36. The divertor plasma is attached. Computed quantities are from TRANS run Y935. s Max. Values Units MA 3.6 T 22.3 MW 3.1 MW 25.7 MW 21.9 MW 23.5 MW 1 19 before ELM MW Electron density, ne 4.1 Fuel density, n D T 3.3±.3 Mean Zeff 1.9±.2 Central ion temperature, T ( ) i 28± 2 kev Central electron temperature, T ( ) e 14 kev W, Wdia dw n T / dt, dw di a / dt.3, m m , , 17. MJ 1 MW τ E ermal, diamagnetic.9,.94 s ( n D + n T ) T ( ) τ i E t hermal, dia 8.3, 8.7±2% 1 m kevs ( n D + n T )/ n, TRANS e.8 /( n ): NA, TRANS D T.59,.49 Neutron rate DT fusion power, DT 1 18 s MW DT fusion energy MJ Alpha heating, particle source power, tot αabs Q in = DT / in DT / abs Q Q tot Thermonuclear s α / tot loss, ot fraction, f s α 2.83, , 2.52 MW ±.19.95±.17 t / t h αabs loss.17,.12.61±

15 abs=21.9 MW (including direct heating, fast particle heating and ermalisation, and e heating by rotation friction). The difference (3.6 MW) includes losses due to NB shine-rough, fast particle charge exchange and rotation (but excluding frictional heating). An alpha particle source power, α=2.83 s MW is measured from e D-T fusion neutron rate, and, taking into account e alpha particle slowing down, an alpha particle heating power tot α abs =1.62 MW is computed to be absorbed by e ermal plasma. The total heating power transferred to e ermal plasma from all sources is erefore heat =23.5 MW. The rate of change of plasma ermal energy dw /dt is large, and e ermal loss due to conduction, convection, radiation and ermal charge exchange is loss= heat -dw /dt=13.2 MW. The total loss from e plasma including ermal, rotational and fast ion loss channels (but excluding NB shine-rough) is tot loss=16.2 MW. A time dependent power balance is shown in e top panel of Fig. 7, and is is used to compute e confinement time 1 in e lower panel. The value of n DT ()τ E T i () is 8.3 or 8.7x1 2 m -3 skev (±2%), depending on wheer e ermal or diamagnetic energy dia confinement time ( τ E, τ E ) is used (see Table II for values of n DT (), τ E, dia τe and T i ()). (MW) τ (sec) ulse No: TRANS run Y935 in tot loss Q tot Q heat loss Q in τ E abs τ E dia tot αabs D α + T α (a.u.) Time (s) FIG. 7. Time dependent performance analysis for pulse No The top panel shows e global power balance including e total power input to e torus in = OH + NB + RF, e total power, abs, from external sources absorbed by e ermal plasma, e alpha tot particle power transferred to e ermal plasma α abs and e total heating power to e ermal plasma heat = abs + tot α abs. The power escaping from e ermal plasma, loss, and e total loss power in all tot channels, loss= loss+ rot loss+ fast loss, are shown. The bottom panel shows e diamagnetic and ermal dia energy confinement times, τ E and τe, and e Balmer α signal. The middle panel compares e simple ratio, Q in = DT / in wi ermal, Q, and total, Qtot, as defined in e text. The central fusion power density,.63 MW/m 3, exceeds e central power density absorbed from external sources,.59 MW/m 3, and does so out to 35% of e minor radius. The alpha particle source power density is about.12 MW/m 3 at e centre and e alpha particle JG98.661/2c 1 The ermal energy confinement time is defined as τ E =W / loss, and e diamagnetic energy confinement time dia as τ E =Wdia /( in + α s - SH - BCX - dw dia /dt). W, W dia, loss, in and α s are defined in e text. In addition SH is e NB shinerough power and BCX is e net NB power lost from NB fast ions by charge exchange on e background neutrals. 12

16 heating power density (including orbit effects and slowing down) is about.6 MW/m 3 most of which is coupled to electrons. The central alpha particle heating of electrons is comparable to each of e centrally absorbed powers (from external sources) to e electrons, equipartition from ions to electrons and e change in electron energy density which are of comparable magnitude. The alpha particle heating should erefore be significant and is is examined in Section D-T Fusion erformance arameters A value of n DT () τ E Ti () 9x1 2 m -3 s kev (at e same T i (), pressure profile peaking and relative ion and electron temperatures as pulse No ) would, for a pure ermonuclear plasma, correspond to Q = DT /( loss - α ) 1 and α/ s loss 16%. If e same conditions could be maintained into steady-state, in = loss - α, and erefore Q in = DT / in would also approach unity. In reference [1] a definition of Q was chosen in such a way as to indicate, for a NB heated plasma, e likely stationary value of DT / in which would be obtained were it possible to maintain e same plasma parameters by reducing e input power by e amount of e derivative of stored energy. Following is approach would suggest at e steady-state equivalent of pulse No would correspond to DT /( in - SH - BCX ) in e range.8-.9 [22], dia approximately commensurate wi e observed values of n DT () τ E Ti (). However, is treatment neglects e finite slowing down time of e alpha particles, e RF driven fast particles, and e small but non-negligible effect of rotation in e power balance. In e following paragraphs two new parameters, Q and Qtot, are defined, which rigorously include all terms in e power balance and can be applied to bo transient and steady-state plasmas. Firstly, it is proposed to define a parameter Q to describe e ermonuclear performance of plasmas in terms of losses from e ermal plasma: Q = / ( loss - α abs ) which can be applied to bo transient and stationary cases. =f DT is e ermonuclear power, loss is e power escaping from e ermal plasma and α abs is e alpha particle heating power transferred to e ermal plasma from ermonuclear reactions. In fact, is definition differs from at used in [1] only in at it takes account of e alpha particle slowing down. Q is en a function of ( 2 τe ) x 2 loss and e reactivity / W, even for transient conditions, and encapsulates e underlying physics of confinement and ermal profiles. Of course Q does not necessarily remain constant as stationary conditions are approached unless 2 ( τ E ) x loss remains constant. Secondly, it is proposed to define a parameter Q tot which describes e total fusion performance in terms of all losses from e plasma: Q tot = DT / ( loss + rot loss fast + loss - tot α abs ) 13

17 and deals consistently wi e more complex situation where e power lost from e plasma rot now also includes rotation, loss, and NB/RF fast particle losses, fast loss (such as charge-exchange and orbit losses). α abs is e total heating by alpha particles from bo ermonuclear tot and non-ermal reactions taking account of e slowing down of e alpha particles. The losses in is expression can be computed from e TRANS power balance, 2 provided at e derivatives can be computed accurately as was demonstrated in Section 3.4. A fairly comprehensive picture of e time dependent performance can be obtained by comparing Q in, Q tot and Q, as shown in Fig. 7 (second panel). Whereas Qin increases monotonically up to e ELM (to a maximum of.62), bo Q tot and Q rise rapidly over e first.8 s and remain approximately constant for.4-.5 s indicating at, ough fusion and loss powers are changing rapidly wi time, ey are doing so in proportion. The peak values Q tot =.95±.17 and Q =.69±.19 are broadly in line wi expectations based on e corresponding values of n DT ()τ E T i (). Note at ere is a degradation in ese performance indicators wi e MHD burst from around 13.2 s (before e ELM) which is also seen on e confinement time (lowest panel in Fig. 7). However, Q in continues to increase up to e time of e ELM, suggesting at e MHD activity increases e losses wi little effect on e reactivity of e plasma. At e ELM ere is a significant rise in density and fall in electron temperature. The consequent change in slowing down transiently increases e alpha particle heating power to 8 MW and e total ermal heating power to 44 MW (off scale). The drop in Q tot accurately reflects e reduction in performance during is phase whilst e high value of Q in clearly has no meaning. The performance parameters Q tot and Q permit a more detailed comparison of e eight pulses shown in Table I. In particular, ese parameters can be used to identify e effect of 2 The denominator in is equation for Qtot loss rot + loss+ fast loss - tot s α abs = OH + NB can be computed from e power balance: s + RF f - d(w + W RF f + W NB + W rot )/dt s The ohmic power OH is computed by TRANS and is very small. The NB power source NB in e plasma (i.e. s NB input minus shine-rough) is well known and e ICRF source power RF in e plasma due to fast particles and direct heating is computed. The derivative includes e energy stored in e ermal plasma W, in e ICRF f f and NB fast particles (W RF and W NB ) and e computed rotational energy W rot. It can be computed from e temporal derivative of f W dia - 3/2W tot + W RF f f + W NB + W rot. The value of is derivative is dominated by W dia wi e fast ion corrections being small ( 1 MW) and is relatively independent of TRANS assumptions, as shown by e good agreement in Fig. 6. The normal level of accuracy of e input power, neutron yield and diamagnetic stored energy are ±1%, ±8% and ±5%. If it is assumed at e errors are gaussian and at ese errors dominate e computation, en e absolute uncertainty in Q tot is ±18%. If it is assumed at e uncertainties in kinetic data introduce a scatter of ±1% in e TRANS computation of Q, en e total uncertainty in Q is likely to be ±(1+18) = ±28%. 14

18 a non-optimal D-T mix, which affects Q strongly, to compare NB-only wi combined NBRF and to determine e effect of plasma current. It is found at Q tot is e same for NB-only and NB supplemented by hydrogen minority ICRF heating, at it degrades weakly wi increasing additional heating power, but improves strongly wi plasma current (from.7 at 3.64 MA to.95 at 4. MA), as will be described in [16]. In is section a new formulation for Q has been proposed which rigorously takes into account all terms in e power balance for a plasma heated by NB injection, RF waves and alpha particles, and can be applied equally well to transient and steady-state plasmas. The ree parameters Q, Qtot and Qin provide a complete description of e performance, which is in line wi e usual n DT ()τ E T i () approach. The transient values of ese parameters are in accord wi e expectations based on deuterium performance and as such confirm at significant progress has been made towards a tokamak reactor. From e analysis presented it is self evident at to achieve DT / in 1 in such transient plasmas would require a significant improvement in confinement and/or reactivity. Conversely, only a relatively modest improvement in confinement and/or reactivity would be required to reach such a milestone were it possible to overcome e MHD limits which render ese plasmas transient Stability of Alfvén Eigenmodes Investigation of Alfvén Eigenmodes (AEs) driven by energetic particles in tokamaks is motivated by e potential for such instabilities to eject energetic fusion product alpha particles and fast particles produced by additional heating from e core of a fusion reactor, possibly leading to significant reduction of alpha particle heating and to first wall damage. The study of AEs in JET during DTE1 has demonstrated at AEs can be destabilised by e energetic ions produced by auxiliary heating [23], but no evidence of alpha particle driven AEs has been observed in e high performance discharges [24]. Similar work has been reported from TFTR [25] and JT-6 [26]. In e high performance D-T hot ion ELM-free H-mode experiments on JET wi fusion powers up to 16.1 MW and central β α ().7% (pulse No ), no observable AE activity was found on e external magnetic measurements up to 5 khz. Figure 8 shows spectrograms of e magnetic fluctuations in a D-D reference discharge and e highest performance D-T discharge. In e D-D reference discharge, at a toroidal magnetic field of 3.4 T, AE are excited by 4.5 MW of hydrogen minority ICRF heating (but not by 3 MW). In e D-T discharge, at 3.8 T, ere is no evidence of AE excitation eier by e 3 MW of ICRF heating or by e 3.2 MW of alpha particle source power. The absence of alpha particle driven AEs is in agreement wi CASTOR-K stability calculations [24], where e least damped AEs (n 5) are found to be stable due to e radial extent of e eigenmodes at high β and e additional damping provided by e high energy ( 16 kv) tritium NB injection. 15

19 ulse No: TAEs JG98.578/1c ulse No: JG98.578/2c 4 4 Frequency (khz) 3 2 Frequency (khz) Time (s) Time (s) FIG. 8. Spectrograms representing e amplitude of e magnetic fluctuations measured by external Mirnov coils in (a) a D-D reference discharge (pulse No. 438) and (b) e D-T highest fusion performance discharge (pulse No ). 4. ALHA ARTICLE HEATING One of e original objectives of JET was e study of alpha particle production, confinement and consequent plasma heating. A wide range of experiments, reviewed in [27], has demonstrated at single fast ions in e 1keV-MeV ranges are trapped and slow-down classically in quiescent plasmas. However, fast ion losses can be caused by sawtoo crashes, toroidal field (TF) ripple or Alfvén eigenmodes (AEs). These effects are not expected to be important in e alpha particle heating experiment described here because e measurements were made just before e collapse of sawtee which had periods (1-1.5 s) longer an e alpha particle slowing-down time in e plasma centre, e toroidal field ripple in JET is low, and e alpha particle pressure was not high enough to excite AEs (see Section 3.6 above). Thus, alpha particles in is experiment were expected to behave classically and, is being e case, were expected to transfer most of eir energy to e plasma electrons. The first observation of measurable alpha particle heating of e electrons in a magnetically confined plasma was made in TFTR under conditions wi a fusion gain Q in =.27 [28]. The hot ion ELM-free H-mode regime in JET is well-suited to e measurement of alpha particle heating since it has high current ( MA) giving small alpha particle orbits, Q in is about.6, and e alpha particle heating of electrons in e plasma core is comparable to, or greater, an each of e NB, ohmic or equipartition heating sources. The alpha particle heating is erefore important in e electron power balance and should be observable by increases in e electron temperature. The larger value of Q in in JET should permit a clearer measurement of alpha particle heating an was possible in TFTR. 16

20 4.1. Design of e Alpha article Heating Experiment In order to maximise e possibility of observing alpha particle heating, e discharge duration, and especially e ELM-free period, e sawtoo-free period and e electron energy confinement time, should be longer an e alpha particle slowing down time of typically 1 s. Furermore, e electron temperature should be sufficiently high and e electron density sufficiently low for ion-electron equipartition and direct electron heating by NB heating to be low. In e JET experiments, e plasma had a central electron temperature greater an 1 kev, a central electron density less an 4x1 19 m -3 and a total power to e electrons from ese sources less an 3% of e total input power. ICRF heating was not used since it predominantly heats electrons and might be confused wi e alpha particle heating. Since experiments in oer plasma regimes on oer tokamaks have indicated e possibility of a plasma isotope effect on energy confinement, e JET alpha particle heating experiment [29] was designed to separate clearly any rise in electron temperature due to alpha particle heating from such an isotope effect on confinement. This was accomplished by scanning e plasma mixture from pure deuterium to nearly pure tritium. The D-T fusion power, and hence e alpha particle heating, would be maximised at intermediate concentrations of deuterium and tritium, whilst e isotopic effects could be isolated by comparing pure deuterium and nearly pure tritium plasmas. In order to avoid temporal or spatial variations in e D-T mixture, e scan was performed using matched NB injection, gas fuelling and recycling mixtures. The last required e walls to be loaded to e required mixture using tokamak pulses between e main pulses in e scan. In is way, pulse No wi a nearly pure tritium plasma (n T /(n T +n D ) 92%) was obtained. The configuration of e NB system for DTE1 allowed a 5 point scan at a heating power of 1.4 MW, wi e particle source varying between 6x1 2 atoms/s in tritium and 8x1 2 atoms/s in deuterium. The prototype for e experiment was e deuterium pulse No. 4365, which was a low power version of e 3.8 MA / 3.4 T high performance pulses. Wi 1.4 MW of NB power, a hot ion H-mode wi an ELM-free period of over 2 s was obtained. This had a D-D neutron yield which projected to 4-5 MW of fusion power wi a 5:5 D-T mixture. The ability to detect alpha particle heating at e 1 MW level was confirmed in a test experiment which used hydrogen minority ICRF heating as a substitute for alpha particle heating Measurement of Alpha article Heating ulse No , wi a tritium concentration of 6%, produced e highest fusion power (6.7 MW) in e scan at an NB input power of 1.4 MW, corresponding to an absorbed power of about 9.4 MW. This is compared wi e pulse (No. 4311) wi a tritium concentration of 92% in Fig. 9. Just after e end of e transition ELMs, bo pulses had a sawtoo before 13 s and en remained sawtoo- and ELM-free for at least 1.4 s. This allowed e central ion 17

21 temperature to reach kev and was responsible ulse No: ulse No: 4311 for e better an anticipated fusion power in pulse No Figure 9 shows at e alpha particle heating power in pulse No is much lower an at in pulse No Thus, e aim to have a clear peak in e alpha particle heating power, wi D-T mixture, was achieved. The effects of alpha particle heating are visible as differences in plasma ermal energy, T i 1 and central ion and electron temperatures. It T e should be noted at e line averaged electron density is about 1% larger in pulse 4311 an in pulse at 14.2 seconds. However, e central densities are identical at 4x1 19 m -3. Interestingly, Time (s) FIG. 9. A comparison of e absorbed NB heating power, alpha particle heating power, plasma energy, and ion e edge temperatures adjust so at and electron temperatures for pulses in e scan wi e highest fusion power (pulse No solid) and e edge pressure is e same in bo pulses e highest tritium content (pulse No dashed). and all of e difference in energy content comes from wiin e inner half minor radius. Whilst some of e temperature difference The error bars shown on e ion temperature traces are e statistical errors for single time-slices. The systematic, calibration error on e electron temperature is 5%. The maximum central electron temperatures are might, in principle, be due to e density variation, regression analysis on e full data set No kev for pulse No and 12.3 kev for pulse showed at e central electron temperature did not depend on e line averaged density. The peak electron temperature achieved in each discharge in e scan is plotted versus time and D-T mixture in Fig. 1. Between 12. s and about 13.5 s alpha particle heating is small and e electron temperature is nearly independent of e D-T mixture. Later, as e fusion power becomes significant and e alpha particles have started to ermalise, e effect of alpha particle heating becomes evident. It can be seen at e maximum electron temperature increases wi tritium concentration, up to 12 kev, when e concentration is 6% (pulse No ), and en falls. The electron temperatures of nearly pure deuterium and tritium discharges are similar, making e identification of alpha particle heating more straightforward. The collapses of giant sawtee are marked wi squares in Fig. 1. It may be seen at e sawtoo period increases wi tritium concentration. This is ascribed to an increase in e NB fast ion pressure wiin e q=1 surface [3]. In light of is variation in sawtoo period, measurements in e saturated period before e sawtoo crash have been used in e following. The central electron temperature is shown in Fig. 11 to be proportional to e total heating power, including absorbed NB, ICRF (where appropriate), ohmic and alpha particle heating abs,nbi (MW) α (MW) W (MJ) T i,e () kev) JG98.38/15c 18