High fusion performance from deuterium-tritium plasmas in JET

High fusion performance from deuterium-tritium plasmas in JET

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 128.143.23.241

This content was downloaded on 25/09/2013 at 06:16

Please note that terms and conditions apply.

1999 Nucl. Fusion 39 209

(http://iopscience.iop.org/0029-5515/39/2/306)

View the table of contents for this issue, or go to the journal homepage for more

Home Search Collections Journals About Contact us My IOPscience

iopscience.iop.org/page/terms

http://iopscience.iop.org/0029-5515/39/2

http://iopscience.iop.org/0029-5515

http://iopscience.iop.org/

http://iopscience.iop.org/search

http://iopscience.iop.org/collections

http://iopscience.iop.org/journals

http://iopscience.iop.org/page/aboutioppublishing

http://iopscience.iop.org/contact

http://iopscience.iop.org/myiopscience

High fusion performance from

deuterium–tritium plasmas in JET

M. Keilhacker, A. Gibson, C. Gormezano, P.J. Lomas, P.R. Thomas,M.L. Watkins, P. Andrew, B. Balet, D. Borba, C.D. Challis, I. Coffey,G.A. Cottrell, H.P.L. De Esch, N. Deliyanakis, A. Fasoli, C.W. Gowers,H.Y. Guo, G.T.A. Huysmans, T.T.C. Jones, W. Kerner, R.W.T. Konig,M.J. Loughlin, A. Maas, F.B. Marcus, M.F.F. Nave, F.G. Rimini, G.J. Sadler,S.E. Sharapov, G. Sips, P. Smeulders, F.X. Soldner, A. Taroni,B.J.D. Tubbing,M.G. von Hellermann, D.J. Ward, JET Team∗

JET Joint Undertaking, Abingdon,Oxfordshire, United Kingdom

Abstract. High fusion power experiments using DT mixtures in ELM-free H mode and optimized

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 the fusion amplification factor was 0.95 ± 0.17,

consistent with the high value of nDT (0)τdiaE Ti(0) = 8.7×1020± 20% m−3 s keV, and was maintained

for about half an energy confinement time until excessive edge pressure gradients resulted in discharge

termination by MHD instabilities. The ratio of DD to DT fusion powers (from separate but otherwise

similar discharges) showed the expected factor of 210, validating DD projections of DT performance

for similar pressure profiles and good plasma mixture control, which was achieved by loading the

vessel walls with the appropriate DT mix. Magnetic fluctuation spectra showed no evidence of Alfvenic

instabilities driven by alpha particles, in agreement with theoretical model calculations. Alpha particle

heating has been unambiguously observed, its effect being separated successfully from possible isotope

effects on energy confinement by varying the tritium concentration in otherwise similar discharges. The

scan showed that there was no, or at most a very weak, isotope effect on the energy confinement time.

The highest electron temperature was clearly correlated with the maximum alpha particle heating

power and the optimum DT mixture; the maximum increase was 1.3 ± 0.23 keV with 1.3 MW of

alpha particle heating power, consistent with classical expectations for alpha particle confinement

and heating. In the optimized shear regime, clear internal transport barriers were established for the

first time in DT, with a power similar to that required in DD. The ion thermal conductivity in the

plasma core approached neoclassical levels. Real time power control maintained the plasma core close

to limits set by pressure gradient driven MHD instabilities, allowing 8.2 MW of DT fusion power

with nDT (0)τdiaE Ti(0) ≈ 1021 m−3 s keV, even though full optimization was not possible within the

imposed neutron budget. In addition, quasi-steady-state discharges with simultaneous internal and

edge transport barriers have been produced with high confinement and a fusion power of up to 7 MW;

these double barrier discharges show a great potential for steady state operation. c©1999, Euratom

1. Introduction

Significant controlled fusion power was firstproduced during the Preliminary Tritium Exper-iment (PTE) in JET in 1991 [1], when a hot ionH mode plasma containing 11% tritium in deu-terium produced 2 MJ of fusion energy with a peakfusion power PDT = 1.7 MW and a fusion powergain Qin = PDT /Pin = 0.12, Pin being the totalinput power to the torus. Tritium usage and neutron

∗ See Appendix.

production were deliberately kept low during thePTE in order to limit vessel activation so that apumped divertor [2] could be installed three monthslater during a manned intervention. This was thefirst of a series of divertors installed in JET between1992 and 1998 to test the effect of increasing geomet-ric closure on divertor behaviour and plasma perfor-mance with deuterium plasmas closest in scale andgeometry to those foreseen for ITER [3].

Meanwhile, TFTR in the USA operated in DTmixtures and, using 50% of deuterium and 50% oftritium, produced 10.7 MW of fusion power and afusion gain Qin = 0.27 in the supershot regime [4].

Nuclear Fusion, Vol. 39, No. 2 1999, IAEA, Vienna 209

M. Keilhacker et al.

This limiter tokamak also explored improved confine-ment in the enhanced reversed shear (ERS) regime,but the performance improvements achieved indeuterium [5] could not be translated into DT [4, 6].

With the closure of TFTR in April 1997, JET isnow the only fusion experiment able to operate withDT fuel mixtures. Furthermore, and going beyondTFTR, JET has an effective divertor and can oper-ate in a wider range of plasma regimes. In addi-tion, JET has the first industrial scale plant for theclosed cycle supply and processing of tritium (theActive Gas Handling System (AGHS) [7]), togetherwith a proven remote handling capability. These haveallowed the JET programme for the period to the endof 1999 to include both a broadly based series of DTexperiments (DTE1) to address crucial issues of DTphysics and technology for ITER, and the develop-ment and exploration of divertor concepts for ITER.

During DTE1, the JET torus was pumped contin-uously through the AGHS and was supplied with DTby the gas introduction and neutral beam (NB) sys-tems. The tritium was stored in uranium beds andreprocessed in the AGHS to a purity of 99.4% by gaschromatography. In contrast to the PTE, when thetotal amount of tritium available was 0.2 g, there wasno equivalent restriction for DTE1. The site inven-tory of 20 g of tritium was reprocessed eight times bythe AGHS, making the equivalent of 99.3 g of tritiumavailable for DTE1.

The total fusion energy produced during DTE1was 675 MJ, equivalent to the limit set on the totalneutron production of less than 2.5× 1020 neutrons(compared with the limit of about 1.5×1018 neutronsimposed for the PTE), so that the subsequent acti-vation of the JET vessel would not prevent mannedin-vessel intervention for more than a year after theend of DTE1. The JET programme continued afterDTE1 with an ITER Urgent Physics Phase and theremote handling replacement of the relatively closedMark IIAP divertor [8] used during DTE1 by an evenmore closed ‘gas box’ divertor (Mark IIGB).

The physics objectives of DTE1 can be dividedinto two areas. The first, discussed in a companionarticle [9], was the characterization of the baselineITER mode of operation, the steady state ELMyH mode, in DT. Of particular importance wereassessments of the hydrogen isotope effects on theH mode threshold power, the energy confinementand the edge operational space, of various heatingschemes at the ICRF in DT plasmas, and of highfusion power and Q in steady state.

The aims of the second area, which are the sub-ject of this article, were to produce maximum fusionpower and Q, to observe alpha particle heating andto study the stability of Alfven eigenmodes. Theseaims are interrelated since high fusion power corre-sponds to high alpha particle pressure, thereby max-imizing the effect of the alpha particles for heatingand destabilizing of the Alfven eigenmodes.

In DD plasmas on JET, two modes of opera-tion have consistently outperformed, at least tran-siently, the ELMy H mode by a factor of about 4in fusion performance. These are the hot ion ELM-free H mode, which had already been tested in thePTE and was considered to be the established routeto high performance, and the optimized shear mode,which had been developed during the year leadingup to DTE1. Their translation from DD to DT is ofgreat interest for understanding these regimes as wellas producing high performance. There is also a longterm interest in developing the latter regime towardssteady state.

These high performance discharges predominantlyuse high energy NB injection to heat and fuel theplasmas. During DTE1, the JET NB system wasoperated with tritium in all eight beam sources ofone of the two injectors. More than 10 MW of tri-tium NB power was delivered at voltages of≈155 kV,for pulse durations ≤5 s. The other injector oper-ated exclusively with deuterium at ≈78 kV. ICRFwas used to supplement NB heating.

The experimental set-up, the heating systems andthe DT related diagnostics are discussed in moredetail in the companion article [9].

The present article is structured as follows. Sec-tion 2 considers the technical issues related to tritiumplasma concentrations, tritium retention in the ves-sel wall and tritium cleanup. The questions of fusionpower performance (including the stability of Alfveneigenmodes) and the detection of alpha particle heat-ing in the ELM-free H mode are then addressed inSections 3 and 4, respectively. Section 5 considers thefusion power performance and physics in the opti-mized shear regime. The summary, conclusions andoutlook follow in Section 6.

2. Tritium concentrations,vessel retention and cleanup

During DTE1, over 200 pulses had plasma tri-tium concentrations greater than 40%. Control ofthe isotopic mixture in the plasma was achieved by

210 Nuclear Fusion, Vol. 39, No. 2 (1999)

Article: High fusion performance from DT plasmas in JET

40

30

20

10

042400 42600 42800 43000 43200 43400

Trit

ium

inve

ntor

y (g

)

Pulse Number

JG98

.703

/1c

PTE-based prediction

b) Tritium inventory

1.0

0.8

0.6

0.4

0.2

0

Sub divertor T alphaResidual gas analyserPTE-based prediction

T /

(T+

D)

a) Tritium concentrations

Cumulative Torusinput

Torus inventory

Figure 1. (a) Tritium concentrations and (b) tritium

inventory for the second period of DTE1 from September

to November 1997 (discharges 42661–43023). Note that

there had already been a first period of DT operation in

May and June 1997. The tritium concentrations shown

were measured by sampling neutrals in the subdivertor

volume during discharges, and the sampling gas was mea-

sured by residual gas analysis between discharges. The

PTE based prediction is based on exhaust measurements

during the PTE.

preceding the experiment with a few ohmic or ICRFheated pulses to change over the isotopic compositionof the walls. With 100% tritium fuelling, plasma tri-tium concentrations greater than 90% were readilyobtained (Fig. 1(a)).

During the campaign, it was possible to achievenearly complete tritium recovery (>98%) from theNB injector which operated in tritium. However,about 30% of the tritium input to the torus wasretained in the torus (Fig. 1(b)). After the tritiumexperiments were completed, the experimental cam-paign continued with about two months of operationin deuterium and hydrogen, after which the torus tri-tium inventory had fallen to 17% of the torus input(6 g). The plasma tritium concentrations had fallen

to ≈0.1% by this time, so it was clear that the tri-tium inventory in the torus was not contributingsignificantly to this.

The torus tritium inventory at the end of DTE1was over 3 times larger than that which had beenexpected from the tritium retention results of thePTE [10]. For the purposes of using the PTE resultsto extrapolate to DTE1, a simple analytical fit tothe PTE data was used to represent the behaviourof retained tritium after each pulse. While this PTEbased prediction closely followed the actual tritiumconcentrations, the difference between the actual andthe predicted tritium inventory grew over the courseof the campaign. This suggests that there was a sinkfor tritium during DTE1 which was not present atthe time of the PTE.

The high level of tritium retention during DTE1has been related to carbon films, saturated withdeuterium and tritium and located in the divertor.While co-deposited films have always been present inthe JET vessel, the films with the Mark IIA diver-tor were unprecedentedly thick and hydrogen rich.The films were found in cold regions of the diver-tor, shadowed from direct contact with the plasma.While the majority of the JET vessel was heatedto 320◦C, these cold regions are actively cooled to≈50◦C (they act as a heat shield to protect the in-vessel divertor coils), allowing the formation of sta-ble films with more than 40% hydrogen concentra-tions in carbon [11]. In contrast, at the time of thePTE (i.e. before the installation of a divertor), thewhole vessel interior was maintained at 300◦C. Thefilms which formed on water cooled metal surfaceswere found to flake off, probably on venting, leadingto considerable debris in the bottom of the divertor.Even though the hardware is similar on the inner andouter legs of the divertor, the flakes were found onlyon the inner leg. The processes leading to this asym-metry and the high level of carbon erosion needed toform the films are still under investigation. Analysisof tiles and flakes removed from the vessel after theexperimental campaign is under way and is expectedto identify the exact location of the tritium inventorystill inside the vessel. This analysis will also deter-mine the physical and chemical characteristics of thefilms.

3. The ELM-free H mode

The ELM-free H mode has been developed inJET over a wide range of plasma currents (1.7–4.2 MA) and toroidal magnetic fields (1.7–3.8 T)

Nuclear Fusion, Vol. 39, No. 2 (1999) 211


33643

0.083W2

0.055W2

40346 26087

40305

34236

34236

40365

15

10

5

5

PD

T (

MW

)

10 15WDIA (MJ)

210xPDD

TRANSP

NB NBRF

JG98

.703

/2c

eq

Figure 2. Equivalent DT fusion power P eqDT from DT

simulations based on DD discharges using the TRANSP

code (see text) against simulated stored diamagnetic

energy (circled symbols). The plain symbols are 210 times

the measured DD fusion power at the measured diamag-

netic energy. The two curves both have a P eqDT ∝ W 2

dia

dependence; the upper curve delineates the upper bound

of 210PDD for all ELM-free data in the period from 1991

to 1997, and the lower curve delineates the typical trend

of 210PDD for combined heating data in DD for the

Mark II divertor only.

with three different divertor configurations. It hasbeen the established route to high fusion per-formance in deuterium and was successfully usedfor the PTE [1]. High values of the fusion tripleproduct, nDT (0)τETi(0), were observed and tran-sient QDT ≈ 1 (as defined in Ref. [1]) was sug-gested on the basis of TRANSP extrapolationsfrom pure deuterium plasmas to 50:50 DT mix-tures [12]. In the PTE, the performance of thisregime was thought to be limited by the powerhandling of the divertor target tiles through theoccurrence of a carbon impurity ‘bloom’ [1]. Follow-ing the installation of the Mark I pumped diver-tor [2], which had sufficient power handling capa-bility to avoid blooms, it was recognized that theperformance limitations were of MHD origin [13]and that highly shaped plasmas, high plasma cur-rent and low recycling were necessary to maximizethe ELM-free period [14]. However, outer modes[13], identified as external kinks associated withhigh edge bootstrap currents [15], could still degrade

performance, but could be ameliorated by rampingdown the plasma current.

3.1. The DD experience andextrapolation to DT

In order to permit a comparison between thefusion performance in DD and DT it is first necessaryto discuss how deuterium discharges were expectedto translate to DT.

Despite the changes to JET and its divertor sincethe PTE, the neutron yield in pure deuterium YDD

has continued to scale strongly with stored diamag-netic energy Wdia with roughly constant reactiv-ity, YDD/W

2dia. This scaling covers the dependence

of the neutron yield on the squares of the ther-mal and fast particle energies and on their productas expected from thermonuclear, beam–beam andbeam–plasma reactions, and variations in the reac-tivity will arise from variations in profiles, Ti, Ti/Te,Zeff and beam energy. It is clear therefore that theseparameters have not varied significantly in thisregime.

Deuterium–tritium simulations using theTRANSP code have been made for a numberof such discharges, assuming transport coefficientsdeduced from the DD discharges, i.e. neglectingisotopic effects, but including alpha particle heating.These simulations, which will be described in moredetail in Ref. [16], also show a quadratic dependencebetween the predicted DT fusion powers and thediamagnetic stored energies (Fig. 2). It is importantto point out that these simulations suggest a fusionpower multiplication factor of 210, consistent withthe profile weighted ratio of the fusion cross-sections[17]. Figure 2 contains simulated fusion power versussimulated Wdia and 210 times the measured DDfusion power PDD versus measured Wdia. The twocurves correspond to a quadratic dependence of210PDD with Wdia, the first an upper bound of allthe DD data and the second a typical curve throughthe combined NB and ICRF heating data (althoughthe bulk of these data is omitted for clarity). Alarge part of the scatter in predictions betweenthe two curves indicate systematic variations inreactivity due to variations in Ti/Te, Zeff andfast ion energy content. In particular, low powerNB only and high power combined heating tendto cluster about the lower curve, whereas highpower NB only data tend towards the upper curve.The TRANSP simulations introduce additionalscatter as a result of uncertainties in the kinetic



Table 1. Maximum values for the eight high power ELM-free H mode pulses in DT mixtures

Discharge No. 42665 42675 42676 42677 42968 42974 42976 42979

Comment Deuterium Sawtooth NB + ωcH NB only NB + ωcH NB + ωcH Early

walls limited +2ωcT switch-off

Time (s) 13.14 13.15 13.34 13.3 13.25 13.37 13.4 13.0

Ip (MA) 3.7 3.65 3.64 3.62 3.99 3.98 3.95 4.03

BT (T) 3.4 3.4 3.4 3.4 3.6 3.6 3.6 3.6

PNB (MW) 18.3 18.7 18.8 21.3 18.7 22.3 22.3 20.3

PRF (MW) 2.7 3.6 3.6 0 5.4 3.0 3.1 3.1

Pin (MW) 21.3 23.2 22.6 21.6 25.1 25.6 25.7 23.8

PDT (MW) 8.7 10.0 12.9 12.3 12.4 15.8 16.1 10.7

Wdia (MJ) 12.3 13.2 15.3 13.4 15.8 16.9 17.0 13.4

nT /(nT + nD) 0.26 0.52 0.54 0.49 0.62 0.67 0.59 0.63

Qin = PDT/Pin 0.40 0.43 0.57 0.57 0.51 0.64 0.62 0.45

QPtot 0.7 0.70 0.77 0.95 0.95

Notes: Ip is the plasma current and BT is the toroidal field at R = 3 m. PNB , PRF and Pin are the power inputs to

the torus from NB, ICRF and total, including ohmic, heating. PDT is the total fusion power as measured by the total

neutron yield. Wdia is the diamagnetic stored plasma energy. nT /(nT + nD) is the fraction of tritium as measured by

the neutral particle analyser. The quoted QPtot values have an uncertainty of ±18%.

measurements. Nevertheless, there is a clear strongimprovement of expected performance with increas-ing stored energy, which provides a useful benchmarkfor the DT measurements to be presented later.

3.2. Mixture control in DT

The PTE experiments with 11% tritium injectedby NB suggested [12] that the plasma mix reflectedthe NB mix, indicating that either recycling wasnegligible (because the particle confinement was sogood) or the injected species was reflected at thewalls (rather than deuterium being desorbed). How-ever, an experiment with the Mark II divertor withselective hydrogen loading of plasma facing surfacessuggested that the dilution of core fuelled deuteriumdischarges could be significant even under condi-tions of low recycling [18]. To clarify the relativecontributions of NB fuelling and recycling to theplasma mix in the ELM-free H mode regime, theNB mix was varied from pure deuterium to pure tri-tium whilst the walls were still predominantly deu-terium and the DT neutron rate was measured. Sucha variation in NB mix was possible with 10 MW ofinjected NB power by selecting appropriate beamsfrom each of the two NB injectors. To interpret theresults of this experiment it is only necessary toknow the relative contributions to the neutron ratefrom beam–plasma reactions due to deuterium on tri-tium and tritium on deuterium, and from thermonu-clear reactions. These were estimated from TRANSP

extrapolations from a pure deuterium discharge toDT. The TRANSP results were then scaled for dif-ferent NB mixes by a simple model which assumedthat the recycling contribution of deuterium was pro-portional to the total NB particle flux (deuteriumand tritium), with the constant of proportionality, f ,to be determined experimentally. If recycling couldbe neglected (f = 0), the peak fusion yield wouldoccur with an NB mix of 45% tritium (weightedtowards being deuterium rich by the tritium on deu-terium beam–plasma reactions). However, the max-imum fusion yield was found experimentally with100% tritium NB injection, and the experimentaldependence of fusion power with NB mix indicatedf = 2, i.e. that recycling dominated the plasmamix. The TRANSP analysis of this experiment qual-itatively confirms the result of the simple model[19]. This experiment clearly demonstrated that itwould be necessary for the high power experiments toreplace deuterium gas fuelling by mixed DT fuellingand to load the walls with an appropriate DT mix.

3.3. High power discharges in DT

In all, eight high power ELM-free discharges wereproduced, as listed in Table 1. The first, which usedDT mixtures for both gas and NB fuelling whilstthe walls were still predominantly deuterium, con-firmed the conclusions of the mixture control experi-ment at higher NB particle fluxes. For the remainingexperiments, the walls were loaded using three to



PDT (MW)

Ip (MA)Dα + Tα

Ti, Te (keV)

PIN

PNB

WDIA

PRF

Zeff<ne>

11.5 12.0 12.5Time (s)

0

2

4

0

10

20

300

2

4

0

10

20

300

510

15

13.0 13.5 14.0

JG98

.703

/3c

<ne> 1019 m 3

ZeffBθ

Power (MW)Energy (MJ)

∼

∼Bθ

Dα + Tα

Ip

Ti

Te

Figure 3. Various time traces for the DT discharge with

the highest fusion power. Note that following the ELM

the density rises to ≈8× 1019 m−3 and remains off-scale.

five ohmic or ICRF heated discharges with the gasfuelling adjusted until the measured DT plasma mixwas close to 50:50. The target density and pre-programmed gas flows were adjusted in one or twoshort (<0.5 s) high power discharges. Finally, the fullduration discharge was selected.

The first series of experiments was performed withthe standard 3.8 MA/3.4 T scenario as developedin deuterium (including the same 0.4 MA/s currentramp-down to stabilize the outer modes) in order topermit a direct comparison between DD and DT forboth NB only and NB supplemented by 3 MW offundamental hydrogen minority ICRF heating. TheNB-only case produced 12.3 MW of fusion powerfor 21.3 MW of NB power, compared with a pro-jected 9.2 MW fusion power for 19.1 MW of NBpower from simulations based on the equivalent (butlower power) DD discharge. With combined heat-ing, 12.9 MW of fusion power was produced with22.4 MW of additional heating power, compared witha projected 12.2 MW for 22.0 MW input. The uncer-tainty in these measurements arising from the accu-racy of the neutron measurements is ±8%, so thisfusion power clearly exceeded the previous record [4].

The second series of experiments was carriedout at a higher toroidal field and plasma current

(4.2 MA/3.6 T) and took full advantage of thehigher NB power available in tritium. Mixed hydro-gen minority and second harmonic tritium ICRFheating was also tested [20], producing 12.4 MW offusion power for a total heating power of 25.1 MW,the additional RF power providing only electronheating. A new record DT fusion power, 16.1 MW,was obtained with a total heating power of 25.4 MWwhich included NB heating (22.3 MW) together withhydrogen minority ICRF heating (3.1 MW).

The maximum value of the ratio of Qin =PDT /Pin for these high power pulses is shown inTable 1 to vary from 0.40 to 0.64, reflecting differ-ences in fuel mix, Zeff , confinement and durationof the high performance phase. In comparison withthe highest fusion power discharge in TFTR withQin = 0.27 [4], the JET discharges produced higherfusion power for lower input power to the torus.

Some time traces for the best JET pulses areshown in Fig. 3. In common with the DD experience,stored energy, fusion power and electron density risemonotonically with time. The ion temperature levelsoff around 28 keV, significantly higher than the elec-tron temperature, which is about 14 keV. The dis-charge does not reach steady state conditions, but islimited by MHD activity: first an outer mode, thena giant ELM (as described previously for deuteriumplasmas). This MHD activity manifests itself in theintensity of the Balmer alpha emission, and, follow-ing detection of a giant ELM, the heating power isswitched off to limit the total neutron production.In DD, with the heating maintained, such dischargesevolved into a repetitive ELMy behaviour with lowerfusion performance. A more detailed comparison ofthe MHD behaviour in DD and DT plasmas will bepresented in Ref. [21].

Figure 4 shows the measured DT fusion power as afunction of diamagnetic stored energy, which can becompared with the DT projections of Fig. 2. Neutralbeam heating only, combined heating (NBRF) andthe higher current data are distinguished by differentsymbols. The low power mixture control experimentstogether with the data from the alpha particle heat-ing experiment (Section 4) are also included, as arethe two 11% tritium pulses from the PTE. The lowpower experiments clearly find an optimum reactiv-ity and these data, taken with those at high power,are well described by PDT ∝ W 2

dia, indicating goodmixture optimization and high reactivity throughoutthe series, independent of power and plasma current.The high power NB-only data are a little above thecurve through the combined heating data, as tends



42677

42665

42676

15

10

5

5

PD

T (

MW

)

10 15

42968

42979

PTE

WDIA (MJ)

42675

42976

0.083W2

3.8 MA4.2 MA

NB NBRF

0.055W2

JG98

.703

/4c

42974

Figure 4. Measured DT fusion power for all ELM-

free plasmas in DT against measured diamagnetic energy.

The two points labelled PTE are from the Preliminary

Tritium Experiment. The curves through the data are

the same as those in Fig. 2.

0.4

0.3

0.2

Total pressure (MPa)

42976 DT

40305 DT simulation

42676 DT

40305 DD

ρ~ ra

Alpha pressure (MPa)42676

42976

40305sim

0.1

0 0.2 0.4 0.6 0.8 1.0

JG98

.380

/12c

Figure 5. Profiles of total pressure computed by

TRANSP for DD discharge 40305, the DT simulation

based on discharge 40305, DT discharge 42676 at the

same plasma current (3.8 MA) and heating power, and

the highest fusion power discharge, 42976 (at 4.2 MA).

The DT discharge at 3.8 MA is intermediate between the

DD case and the DT simulation.

10

15

5

0

2

3

1

0

6

4

2

00.8

0.4

0

Thermonuclear fraction

DT neutron rate

12.4 12.8 13.2

Time (s)

Pulse No: 42976 TRANSP run Y935

JG98

.703

/6c

P (

MW

)(1

018 s

-1)

Wdi

a (M

J)ƒ t

h

Measured

MeasureddWdia

dt

Measured

PIN

TRANSP

Dα + Tα (a.u.)

TRANSP

TRANSP

Figure 6. Validation of TRANSP analysis and kinetic

data for discharge 42976 by reference to direct measure-

ments of the diamagnetic energy Wdia , the derivative of

the diamagnetic energy dWdia/dt and the neutron rate.

The bottom panel shows the thermonuclear fraction.

to be the case in DD (Fig. 2). The improved fusionperformance at higher plasma current is a result ofthe higher stored energy; the 17 MJ attained is higherthan achieved previously in JET or in any other mag-netic plasma confinement experiment.

Comparing the NB only and NBRF data in Figs 2and 4, the ratio of the fusion power in DT to thatin DD is within ±10% of the expected ratio of 210at the same stored energy, in contrast to the resultsobtained on TFTR [4]. Figure 5 shows that the shapeof the total pressure profile in DT is similar to that inDD and, within experimental uncertainties, is closeto that projected for DT. In fact, the differencebetween the DD and DT profiles reflects the pop-ulation of energetic alpha particles produced by thefusion reactions. These observations apparently con-firm the assumption used in the extrapolations to DTthat the net isotope effect on energy confinement isnegligible.

3.4. TRANSP analysis of the DT discharges

Figure 6 shows that there is excellent agree-ment between directly measured quantities and those



Table 2. Parameters of interest for ELM-free H mode discharge 42976 with the highest DT fusion power, at the time

of maximum QPtot (13.13 s) and maximum value of each parameter, if significantly different

(The geometric parameters of this pulse are: minor radius a = 0.95 m, major radius R = 2.92 m, q95 = 3.47, elonga-

tion κ = 1.81, triangularity δ = 0.36. The divertor plasma is attached. The computed quantities are from TRANSP

run Y935.)

Parameter Values at 13.13 s Maximum values

Plasma current, Ip (MA) 4.02 4.18

Toroidal field, BT (T) 3.6

NB power, PNB (MW) 22.3

RF power, PRF (MW) 3.1

Total power input to torus, Pin (MW) 25.7

Absorbed external power, P thabs (MW) 21.9

Total thermal heating power, P thheat = P thabs + P totα abs (MW) 23.5

Loss power, P thloss = P thheat − dWth/dt (MW) 13.2 19 before ELM

Electron density, ne (1019 m−3) 4.1

Fuel density, nD + nT (1019 m−3) 3.3± 0.3

Mean Zeff 1.9± 0.2

Central ion temperature, Ti(0) (keV) 28± 2

Central electron temperature, Te(0) (keV) 14

Wth, Wdia (MJ) 12.0, 15.5 13.6, 17.0

dWth/dt, dWdia/dt (MW) 10.3, 10.2

τE thermal, diamagnetic (s) 0.90, 0.94

(nD + nT )Ti(0)τE thermal, diamagnetic (1020 m−3 keV s) 8.3, 8.7± 20%

(nD + nT )/ne: TRANSP 0.80

nT /(nD + nT ): NPA, TRANSP 0.59, 0.49

Neutron rate (1018 s−1) 5.0 5.7

DT fusion power, PDT (MW) 14.2 16.1

DT fusion energy (MJ) 13.8

Alpha particle source power, P sα, heating power, P totα abs (MW) 2.83, 1.62 3.15, 2.52

Qin = PDT/Pin 0.55 0.62

PDT/Pabs 0.60 0.74

QPth 0.69± 0.19

QPtot 0.95± 0.17

Thermonuclear fraction, fth 0.61± 0.05 0.63

P sα/Ptotloss, P

totα abs/P

thloss 0.17, 0.12

computed by TRANSP. In particular, the measureddiamagnetic energy is in excellent agreement withthat computed by TRANSP, Wdia = Wth + 3

2Wf⊥tot,

where Wth is the kinetic measurement of the thermalenergy content and W f

⊥tot is the computed total per-pendicular fast ion energy. It should be noted thatTRANSP (run Y935) computes the neutron yielddirectly from the kinetic measurements. The errorbar on this computation reflects the uncertaintiesin Ti(r) and Zeff (r) assuming that these are thedominant uncertainties. The agreement is satisfying,indicating not only the quality of the measurementsbut also that TRANSP contains the appropriatephysics.

The measured derivative of the diamagneticenergy has a resolution which can be as short as1 ms and resolves the effect of NB interruptionsand MHD events, as can be seen by comparingwith the traces Pin and Dα + Tα. With a TRANSPtime step of 10 ms and 2000 particles in the MonteCarlo NB slowing down calculation it is necessaryto smooth over 30–50 ms in order to minimize sta-tistical noise. With such smoothing there is rea-sonable agreement between measured and computeddWdia/dt over most of the pulse, even during fluc-tuations in Zeff (in magnitude comparable to theerror estimates) which contribute to fluctuationsin the kinetic measurement of dWth/dt. For this



30

20

10

01.0

0.5

0

1.0

0.5

012.0 12.5 13.0

Time (s)

13.5 14.0

Pulse No: 42976 TRANSP run Y935

JG98

.703

/7c

P (

MW

)τ

(s)

QtotP

QthP

Q

τEdia

τEth

Plosstot

Plossth

Pαabstot

Pheatth Pabs

thPin

Dα + Tα (a.u.)

in

Figure 7. Time dependent performance analysis for

discharge 42976. The top panel shows the global power

balance including the total power input to the torus

Pin = POH + PNB + PRF , the total power from external

sources absorbed by the thermal plasma, P thabs , the alpha

particle power transferred to the thermal plasma, P totα abs ,

and the total heating power to the thermal plasma,

P thheat = P th

abs + P totα abs . The power escaping from the ther-

mal plasma, P thloss , and the total loss power in all chan-

nels, P totloss = P th

loss + P rotloss + P fast

loss , are shown. The middle

panel comparesQin = PDT/Pin with the thermalQPth and

the total QPtot as defined in the text. The bottom panel

shows the diamagnetic and thermal energy confinement

times, τ diaE and τ th

E , and the Balmer α signal. Following

the ELM there is a transient increase in the total heating

power P thheat to about 44 MW, which is off-scale.

reason the derivative of the thermal energy iscomputed from dWth/dt = d(Wdia − 3

2Wf⊥tot)/dt

because the fast particle energy is almost constant. ATRANSP run (Y936) has been performed with theseZeff fluctuations removed, and this improves theagreement between measured and computed quan-tities and yet has a negligible (<5%) effect on thepower balance. It is therefore concluded that theTRANSP run (Y935) using only measured data pro-vides an adequate description of the behaviour of fastions and the thermal plasma.

Figure 6 also shows the thermonuclear fraction(the ratio of thermonuclear power to total fusionpower, fth = P thDT /PDT ) as computed by TRANSP,which reaches a value of about 0.63. In order to esti-mate the uncertainty in the thermonuclear fractionit is valid, given the small uncertainty in the mea-sured neutron rate, to constrain the computationto match the measured neutron rate exactly. Thisgives an uncertainty in fth of ±0.05 relative to thereference TRANSP run.

The measured and computed data for discharge42976, which produced the highest fusion power, aresummarized in Table 2 at a time (13.13 s) when theplasma is still quiescent. The first point to note isthe global power balance, which shows that of thetotal power input to the torus (Pin = 25.7 MW),the total power from external sources absorbed bythe thermal plasma is P thabs = 21.9 MW (includingdirect heating, fast particle heating and thermaliza-tion, and the heating by rotation friction). The dif-ference (3.6 MW) includes losses due to NB shine-through, fast particle charge exchange and rotation(but excluding frictional heating). An alpha parti-cle source power, P sα = 2.83 MW, is measured fromthe DT fusion neutron rate, and, taking into accountthe alpha particle slowing down, an alpha parti-cle heating power P totαabs = 1.62 MW is computedto be absorbed by the thermal plasma. The totalheating power transferred to the thermal plasmafrom all sources is therefore P thheat = 23.5 MW. Therate of change of plasma thermal energy dWth/dt islarge, and the thermal loss due to conduction, con-vection, radiation and thermal charge exchange isP thloss = P thheat − dWth/dt = 13.2 MW. The totalloss from the plasma including thermal, rotationaland fast ion loss channels (but excluding NB shine-through) is P totloss = 16.2 MW. A time dependentpower balance is shown in the top panel of Fig. 7,and this is used to compute the confinement time1

in the lower panel. The value of nDT (0)τETi(0) is8.3× 1020 or 8.7× 1020 m−3 s keV (±20%), depend-ing on whether the thermal or diamagnetic energyconfinement time (τ thE , τdiaE ) is used (see Table 2 forvalues of nDT (0), τ thE , τdiaE and Ti(0)).

1 The thermal energy confinement time is defined as τ thE =Wth/P

thloss, and the diamagnetic energy confinement time as

τdiaE = Wdia/(Pin+P sα−PSH−PBCX−dWdia/dt). Wth,Wdia,P thloss, Pin and P sα are defined in the text. In addition, PSH is

the NB shine-through power and PBCX is the net NB powerlost from NB fast ions by charge exchange on the backgroundneutrals.



The central fusion power density, 0.63 MW/m3,exceeds the central power density absorbed fromexternal sources, 0.59 MW/m3, and does so out to35% of the minor radius. The alpha particle sourcepower density is about 0.12 MW/m3 at the cen-tre, and the alpha particle heating power density(including orbit effects and slowing down) is about0.06 MW/m3, most of which is coupled to electrons.The central alpha particle heating of electrons iscomparable to each of the centrally absorbed powers(from external sources) to the electrons, equiparti-tion from ions to electrons and the change in elec-tron energy density, which are of comparable magni-tude. The alpha particle heating should therefore besignificant, and this is examined in Section 4.

3.5. DT fusion performance parameters

A value of nDT (0)τ thE Ti(0) ≈ 9× 1020 m−3 s keV(at the same Ti(0), pressure profile peaking and rel-ative ion and electron temperatures as discharge42976) would, for a pure thermonuclear plasma,correspond to Q = PDT /(Ploss − Pα) ≈ 1 andP sα/Ploss ≈ 16%. If the same conditions could bemaintained into steady state, Pin = Ploss − Pα, andtherefore Qin = PDT /Pin would also approach unity.

In Ref. [1] a definition of Q was chosen in sucha way as to indicate, for an NB heated plasma,the likely stationary value of PDT /Pin which wouldbe obtained were it possible to maintain the sameplasma parameters by reducing the input power bythe amount of the derivative of the stored energy. Fol-lowing this approach would suggest that the steadystate equivalent of discharge 42976 would corre-spond to PDT /(Pin − PSH − PBCX ) in the range0.8–0.9 [22], approximately commensurate with theobserved values of nDT (0)τdiaE Ti(0). However, thistreatment neglects the finite slowing down time ofthe alpha particles, the RF driven fast particles andthe small but non-negligible effect of rotation in thepower balance. In the following paragraphs two newparameters, QPth and QPtot, are defined, which rig-orously include all terms in the power balance andcan be applied to both transient and steady stateplasmas.

Firstly, it is proposed to define a parameter QPth todescribe the thermonuclear performance of plasmasin terms of losses from the thermal plasma,

QPth = Pth/(P thloss − P thα abs)

which can be applied to both transient and sta-tionary cases. Pth = fthPDT is the thermonuclear

power, P thloss is the power escaping from the thermalplasma and P thα abs is the alpha particle heating powertransferred to the thermal plasma from thermonu-clear reactions. In fact, this definition differs fromthat used in Ref. [1] only in that it takes accountof the alpha particle slowing down. QPth is then afunction of (τ thE )2P thloss and the reactivity Pth/W

2th,

even for transient conditions, and encapsulates theunderlying physics of confinement and thermal pro-files. Of course, QPth does not necessarily remain con-stant as stationary conditions are approached unless(τ thE )2P thloss remains constant.

Secondly, it is proposed to define a parameter QPtotwhich describes the total fusion performance in termsof all losses from the plasma,

QPtot = PDT /(P thloss + P rotloss + P fastloss − P totαabs)

and deals consistently with the more complex sit-uation where the power lost from the plasma nowalso includes rotation, P rotloss, and NB/RF fast parti-cle losses, P fastloss (such as charge exchange and orbitlosses). P totα abs is the total heating by alpha particlesfrom both thermonuclear and non-thermal reactions,taking account of the slowing down of the alpha par-ticles. The losses in this expression can be computedfrom the TRANSP power balance2, provided thatthe derivatives can be computed accurately, as wasdemonstrated in Section 3.4.

2 The denominator in this equation for QPtot can be

computed from the power balance

P thloss + P rotloss + P fastloss − Ptotα abs = POH + P sNB + P sRF

− d(Wth +W fRF +W f

NB +W rot)/dt.

The ohmic power POH is computed by TRANSP and is verysmall. The NB power source P sNB in the plasma (i.e. NBinput minus shine-through) is well known and the ICRF source

power P sRF in the plasma due to fast particles and direct heat-ing is computed. The derivative includes the energy stored inthe thermal plasma Wth, in the ICRF and NB fast particles(W f

RF and W fNB , respectively) and the computed rotational

energy W rot. It can be computed from the temporal deriva-

tive of

Wdia −3

2W f⊥tot +W f

RF +W fNB +W rot.

The value of this derivative is dominated by Wdia with thefast ion corrections being small (≈1 MW) and is relativelyindependent of the TRANSP assumptions, as shown by the

good agreement in Fig. 6. The normal levels of accuracy ofthe input power, neutron yield and diamagnetic stored energyare ±10, ±8 and ±5%. If it is assumed that the errors areGaussian and that these errors dominate the computation,then the absolute uncertainty in QPtot is ±18%. If it is assumed

that the uncertainties in kinetic data introduce a scatter of±10% in the TRANSP computation of QPth, then the totaluncertainty in QPth is likely to be ±(10+18) = ±28%.



A fairly comprehensive picture of the time depen-dent performance can be obtained by comparingQin,QPtot and QPth, as shown in Fig. 7 (middle panel).Whereas Qin increases monotonically up to the ELM(to a maximum of 0.62), both QPtot and QPth riserapidly over the first 0.8 s and remain approximatelyconstant for 0.4–0.5 s indicating that, though fusionand loss powers are changing rapidly with time,they are doing so in proportion. The peak valuesQPtot = 0.95±0.17 and QPth = 0.69±0.19 are broadlyin line with expectations based on the correspondingvalues of nDT (0)τETi(0).

Note that there is a degradation in these perfor-mance indicators with the MHD burst from around13.2 s (before the ELM) which is also seen on theconfinement time (lowest panel in Fig. 7). However,Qin continues to increase up to the time of the ELM,suggesting that the MHD activity increases the losseswith little effect on the reactivity of the plasma. Atthe ELM there is a significant rise in density and fallin electron temperature. The consequent change inslowing down transiently increases the alpha parti-cle heating power to ≈8 MW and the total thermalheating power to ≈44 MW (off-scale). The drop inQPtot accurately reflects the reduction in performanceduring this phase whilst the high value of Qin clearlyhas no meaning.

The performance parameters QPtot and QPth permita more detailed comparison of the eight dischargesshown in Table 1. In particular, these parameterscan be used to identify the effect of a non-optimalDT mix, which affects QPth strongly, to compare NB-only heating with combined NB and RF heating andto determine the effect of plasma current. It is foundthat QPtot is the same for NB-only and NB supple-mented by hydrogen minority ICRF heating, that itdegrades weakly with increasing additional heatingpower, but improves strongly with plasma current(from ≈0.7 at 3.64 MA to ≈0.95 at 4.0 MA), as willbe described in Ref. [16].

In this section, a new formulation for Q has beenproposed which rigorously takes into account all theterms in the power balance for a plasma heated byNB injection, RF waves and alpha particles, andwhich can be applied equally well to transient andsteady state plasmas. The three parametersQPth,QPtotand Qin provide a complete description of the perfor-mance, which is in line with the usual nDT (0)τETi(0)approach. The transient values of these parametersare in accord with the expectations based on deu-terium performance and as such confirm that sig-nificant progress has been made towards a tokamak

reactor. From the analysis presented it is self-evidentthat to achieve PDT /Pin ≥ 1 in such transient plas-mas would require a significant improvement in con-finement and/or reactivity. Conversely, only a rela-tively modest improvement in confinement and/orreactivity would be required to reach such a mile-stone were it possible to overcome the MHD limitswhich render these plasmas transient.

3.6. Stability of Alfven eigenmodes

Investigation of Alfven eigenmodes (AEs) drivenby energetic particles in tokamaks is motivated bythe potential for such instabilities to eject energeticfusion product alpha particles and fast particles pro-duced by additional heating from the core of a fusionreactor, possibly leading to a significant reduction ofalpha particle heating and to first wall damage. Thestudy of AEs in JET during DTE1 has demonstratedthat AEs can be destabilized by the energetic ionsproduced by auxiliary heating [23], but no evidenceof alpha particle driven AEs has been observed inthe high performance discharges [24]. Similar workhas been reported from TFTR [25] and JT-60 [26].

In the high performance DT hot ion ELM-freeH mode experiments on JET with fusion powers upto 16.1 MW and central βα(0) ≤ 0.7% (discharge42976), no observable AE activity was found on theexternal magnetic measurements up to 500 kHz. Fig-ure 8 shows spectrograms of the magnetic fluctua-tions in a DD reference discharge and the highestperformance DT discharge. In the DD reference dis-charge, at a toroidal magnetic field of 3.4 T, AEsare excited by 4.5 MW of hydrogen minority ICRFheating (but not by 3 MW). In the DT discharge, at3.8 T, there is no evidence of AE excitation eitherby the 3 MW of ICRF heating or by the 3.2 MWof alpha particle source power. The absence of alphaparticle driven AEs is in agreement with CASTOR-K stability calculations [24], where the least dampedAEs (n ≈ 5) are found to be stable due to the radialextent of the eigenmodes at high β and the additionaldamping provided by the high energy (≈160 kV)tritium NB injection.

4. Alpha particle heating

One of the original objectives of JET was thestudy of alpha particle production, confinement andconsequent plasma heating. A wide range of experi-ments, reviewed in Ref. [27], has demonstrated thatsingle fast ions in the 100 keV–MeV range are



500

400

300

200

100

0

Pulse No: 40308

12.6 12.8 13 .0Time (s)

13.2 13.4

Freq

uenc

y (k

Hz)

JG98

.703

/8a

TAEs

500

400

300

200

100

0

Pulse No: 42976

12.6 12.8 13 .0Time (s)

13.2 13.4

Freq

uenc

y (k

Hz)

JG98

.703

/8b

(a)

(b)

Figure 8. Spectrograms representing the amplitude of

the magnetic fluctuations measured by external Mirnov

coils in (a) a DD reference discharge (No. 40308)

and (b) the DT highest fusion performance discharge

(No. 42976).

trapped and slow down classically in quiescent plas-mas. However, fast ion losses can be caused by saw-tooth crashes, toroidal field ripple or AEs. Theseeffects are not expected to be important in the alphaparticle heating experiment described here becausethe measurements were made just before the collapse

of sawteeth which had periods (1–1.5 s) longer thanthe alpha particle slowing down time in the plasmacentre, the toroidal field ripple in JET is low andthe alpha particle pressure was not high enough toexcite AEs (Section 3.6). Thus, alpha particles in thisexperiment were expected to behave classically and,this being the case, were expected to transfer mostof their energy to the plasma electrons.

The first observation of measurable alpha parti-cle heating of the electrons in a magnetically con-fined plasma was made in TFTR under conditionswith a fusion gain Qin = 0.27 [28]. The hot ionELM-free H mode regime in JET is well suited tothe measurement of alpha particle heating since ithas a high current (3.8–4.2 MA) giving small alphaparticle orbits, Qin is about 0.6 and the alpha par-ticle heating of electrons in the plasma core is com-parable to, or greater than, each of the NB, ohmicor equipartition heating sources. The alpha particleheating is therefore important in the electron powerbalance and should be observable by increases inthe electron temperature. The larger value of Qin inJET should permit a clearer measurement of alphaparticle heating than was possible in TFTR.

4.1. Design of the alpha particleheating experiment

In order to maximize the possibility of observ-ing alpha particle heating, the discharge duration,and especially the ELM-free period, the sawtooth-free period and the electron energy confinement time,should be longer than the alpha particle slowingdown time of typically 1 s. Furthermore, the electrontemperature should be sufficiently high and the elec-tron density sufficiently low for ion–electron equipar-tition and direct electron heating by NB heating tobe low. In the JET experiments, the plasma had acentral electron temperature greater than 10 keV, acentral electron density less than 4 × 1019 m−3 anda total power to the electrons from these sources lessthan 30% of the total input power. ICRF heating wasnot used since it predominantly heats electrons andmight be confused with the alpha particle heating.

Since experiments in other plasma regimes onother tokamaks have indicated the possibility of aplasma isotope effect on energy confinement, theJET alpha particle heating experiment [29] wasdesigned to separate clearly any rise in electron tem-perature due to alpha particle heating from such anisotope effect on confinement. This was accomplished



by scanning the plasma mixture from pure deuteriumto nearly pure tritium. The DT fusion power, andhence the alpha particle heating, would be maxi-mized at intermediate concentrations of deuteriumand tritium, whilst the isotopic effects could be iso-lated by comparing pure deuterium and nearly puretritium plasmas.

In order to avoid temporal or spatial variationsin the DT mixture, the scan was performed usingmatched NB injection, gas fuelling and recyclingmixtures. The last required the walls to be loadedto the required mixture using tokamak dischargesbetween the main discharges in the scan. In this way,discharge 43011 with a nearly pure tritium plasma(nT /(nT +nD) ≈ 92%) was obtained. The configura-tion of the NB system for DTE1 allowed a five pointscan at a heating power of 10.4 MW, with the particlesource varying between 6 × 1020 atoms/s in tritiumand 8×1020 atoms/s in deuterium. The prototype forthe experiment was the deuterium discharge 40365,which was a low power version of the 3.8 MA/3.4 Thigh performance discharges. With 10.4 MW of NBpower, a hot ion H mode with an ELM-free period ofover 2 s was obtained. This had a DD neutron yieldwhich projected to 4–5 MW of fusion power with a50:50 DT mixture. The ability to detect alpha par-ticle heating at the 1 MW level was confirmed in atest experiment which used hydrogen minority ICRFheating as a substitute for alpha particle heating.

4.2. Measurement of alpha particle heating

Discharge 42847, with a tritium concentration of60%, produced the highest fusion power (6.7 MW)in the scan at an NB input power of 10.4 MW, cor-responding to an absorbed power of about 9.4 MW.This is compared with the discharge (No. 43011) witha tritium concentration of 92% in Fig. 9. Just afterthe end of the transition ELMs, both discharges hada sawtooth before 13 s and then remained sawtooth-and ELM-free for at least 1.4 s. This allowed thecentral ion temperature to reach 16–17 keV and wasresponsible for the better than anticipated fusionpower in discharge 42847. Figure 9 shows that thealpha particle heating power in discharge 43011 ismuch lower than that in discharge 42847. Thus, theaim to have a clear peak in the alpha particle heatingpower, with DT mixture, was achieved.

The effects of alpha particle heating are visibleas differences in plasma thermal energy and in thecentral ion and electron temperatures. It should benoted that the line averaged electron density is about

JG98

.703

/9c

Pulse No: 42847 Pulse No: 43011

10

5

02

1

0

10

0

20

10

5

0

12 13

Time (s)

14 15

Pab

s,N

BI (

MW

)P

α (M

W)

Wth

(MJ)

Ti,e

(0)

keV

) Ti

Te

Figure 9. Comparison of the absorbed NB heating

power, alpha particle heating power, plasma energy, and

ion and electron temperatures for discharges in the scan

with the highest fusion power (discharge 42847, solid

curves) and the highest tritium concentration (discharge

43011, dashed curves). The error bars shown on the

ion temperature traces are the statistical errors for sin-

gle time slices. The systematic, calibration error on the

electron temperature is 5%. The maximum central elec-

tron temperatures are 10.5 keV for discharge 43011 and

12.3 keV for discharge 42847.

10% higher in discharge 43011 than in discharge42847 at 14.2 s. However, the central densities areidentical at 4 × 1019 m−3. Interestingly, the edgetemperatures adjust so that the edge pressure is thesame in both discharges, and all of the differencein energy content comes from within the inner halfminor radius. Whilst some of the temperature differ-ence might, in principle, be due to the density varia-tion, regression analysis on the full data set showedthat the central electron temperature did not dependon the line averaged density.

The peak electron temperature achieved in eachdischarge in the scan is plotted versus time and DTmixture in Fig. 10. Between 12.0 s and about 13.5 salpha particle heating is small and the electron tem-perature is nearly independent of the DT mixture.Later, as the fusion power becomes significant andthe alpha particles have started to thermalize, theeffect of alpha particle heating becomes evident. It



12 13 14 150

0.2

0.4

0.6

0.8

1.0

# 40365

# 42870

# 42856# 42847

# 42840

# 43011

4.2-7 keV

7-9 keV

9-10 keV

10-11 keV

11-11.5 keV

11.5-12 keV

>12 keV

Giantsawteeth

Time (s)

JG98.703/10c

n T /(

n T +

nD)

Figure 10. Contour plot of the central electron temper-

ature, for all the discharges in the alpha particle heating

scan, against DT mixture on the vertical axis and time

on the horizontal axis. Giant sawteeth are marked by

squares.

can be seen that the maximum electron temperatureincreases with tritium concentration, up to 12 keV,when the concentration is 60% (discharge 42847),and then falls. The electron temperatures of nearlypure deuterium and tritium discharges are similar,making the identification of alpha particle heatingmore straightforward. The collapses of giant saw-teeth are marked with squares in Fig. 10. It canbe seen that the sawtooth period increases with tri-tium concentration. This is ascribed to an increasein the NB fast ion pressure within the q = 1 surface[30]. In the light of this variation in sawtooth period,measurements in the saturated period before thesawtooth crash have been used in the following.

The central electron temperature is shown inFig. 11 to be proportional to the total heating power,including absorbed NB, ICRF (where appropriate),ohmic and alpha particle heating power. The vari-ation in heating power in the DT discharges issubstantially due to the alpha particle heating (0–1.3 MW), although there is a variation in the nomi-nal NB power, which can be as much as 10% at lowtritium concentrations, as indicated by the horizon-tal bars. Also included in the data set are the testdischarges with hydrogen minority ICRF emulatingalpha particle heating. It can be seen from Fig. 11that alpha particle heating and the ICRF heating are

JG98

.380

/17c

10

11

12

13

99.6 10.6

POH + PNBI + PICRH + Pα (MW)

11.6 12.6

Te

(0)

(keV

)

D-T Pulses

ICRH in D-D

Figure 11. Central electron temperature, just before

sawtooth crashes, plotted against total absorbed heating

power, including alpha particle heating power. The bars

indicate deviations from the nominal value of the NB

heating power of 10.5 MW. Bars to the right indicate a

shortfall of power.

equally effective. This is reassuring because hydrogenminority ICRF heating is known from modulationexperiments [31] to have high efficiency and a highlypeaked deposition profile.

A regression fit to the data of Fig. 11 gives

Te(0) = (0.21± 0.99) + (0.99± 0.09)P thheat

where P thheat is the total power absorbed by the ther-mal plasma, including alpha particle heating. A fitwhich allows the contribution from the alpha parti-cle heating to be separated from variations in the NBheating power gives

Te(0) = (0.07± 1.04) + (1.0± 0.1)(P thheat − P thα abs)

+ (0.99± 0.13)P thαabs.

The result is a maximum change in Te(0) of 1.3 ±0.23 keV with 1.3 MW of alpha particle source power.If the electron density or the tritium fraction isincluded in the regression fit, their weights are zerowithin errors. This fit can be used to display theeffect of alpha particle heating directly. Figure 12shows a plot of ∆Te(0) = Te(0) − 0.07 − (P thheat −P thαabs) versus P thα abs. This shows that once the effectof NB power variations is removed, the alpha particleheating is consistent over the scan and that the stan-dard error on the temperature change with 1.3 MW



JG98

.380

/18c

0

0.5

1.0

1.5

–0.5–0.5 0

P (MW)

0.5 1.0 1.5

∆Te

(0)

(keV

)

D-T Pulses

ICRH in D-D

thα abs

Figure 12. Data of Fig. 11 re-plotted to show the change

in electron temperature, identified by the regression fit as

being due to alpha particle heating, versus alpha particle

heating power.

alpha particle power is a good reflection of the accu-racy of the data.

4.3. TRANSP analysis of the power balance

Figure 13(a) shows the electron power balancewithin the inner half minor radius for discharge42847. The stacked curves represent sums of con-tributions from each term. The rate of change ofenergy content and the convection loss are showninverted for clarity. The conduction loss is the differ-ence between the upper and lower envelopes. It canbe seen that the alpha particle heating is the sin-gle largest contributor to the electrons in the plasmacore at 14 s, and constitutes nearly one third of thetotal electron source power. Furthermore, the rate ofchange of energy content is the largest sink, taking1 MW out of 1.8 MW at 14 s. In contrast, in dis-charge 43011 with a tritium concentration of 92%,the rate of change of energy content is only 0.5 MWout of 1.4 MW, as shown in Fig. 13(b). The alphaparticle heating is a minor part of the power bal-ance and its absence is reflected in the rate of changeof energy content. Apparently, the conduction lossis more or less the same, which seems reasonablebecause the plasma conditions are comparable. Note

12.0 12.5 13.0 13.5 14.0 14.5 2.0

1.5

1.0

0.5

0.0

0.5

1.0

1.5

2.0

P (

MW

)

Time (s)

Pulse No: 42847a)

Pbe

Pie

POH

Pαe

dW /dt

Pconve

JG98

.703

/13a

c

e

12.0 12.5 13.0 13.5 14.0 14.5 2.0

1.5

1.0

0.5

0.0

0.5

1.0

1.5

2.0

P (

MW

)

Time (s)

Pulse No: 43011b)

Pbe

Pie

POHPαe

Pconve

JG98

.703

/13b

dW /dte

Figure 13. Electron power balance within r/a = 0.5

for the discharges in the alpha particle heating scan with

(a) the highest fusion power (discharge 42847) and (b)

the highest tritium concentration (discharge 43011). The

rate of change of energy content and the convective loss

are shown inverted for clarity.

that the glitches in the traces, observable in bothdischarges before 13 s, are due to sawteeth.

It is important to note that the correspondencebetween the difference in rate of change of electronenergy content and the total electron input power,within r/a = 0.5 in the two pulses shown in Fig. 13,is a certain indication that the alpha particle heating



profile is similar to that computed from the alphaparticle source and classical slowing down.

TRANSP analysis has also been used to com-pute the thermal energy confinement time. Withinerrors, the pure deuterium and nearly pure tritiumdischarges have the same confinement time of 1.2 s.The discharges with significant alpha particle heatinghave slightly higher confinement times (1.3–1.4 s),but the errors are larger. This increase in confine-ment time is, at least partly, due to the alpha parti-cle heating source being more peaked than either theNB or ohmic heating. This same analysis shows thatthe plasma energy content is increased by 1 MJ byalpha particle heating, and that it would be higherif the ELM-free period were longer.

4.4. Unambiguous evidencefor alpha particle heating

The data presented above constitute unambigu-ous evidence for alpha particle heating because theexperiment was designed to eliminate other sourcesof temperature change, such as any plasma isotopeeffect. There were clear peaks in temperature andplasma energy content as the mixture was scannedfrom pure deuterium to nearly pure tritium. Thesepeaks correlate well with the alpha particle heat-ing power, as the regression analysis shows. Thelocal electron power balances are consistent with theclassical alpha particle power deposition profiles.

It should be stressed that the alpha particle heat-ing data have been taken from MHD quiescentplasmas. Some other discharges, from both the alphaparticle heating experiment and the preparationexperiments for the high fusion power studies, couldhave been used to extend the data set but showedsigns of MHD activity. This MHD activity certainlyreduced the alpha particle heating, but there is noway to identify whether this is due to a loss of alphaparticles or a loss of plasma confinement. A futureDT experiment should address this issue by prepar-ing plasmas, in both DD and DT, with the chosentype of MHD activity, such as sawteeth, ELMs orexternal kink instabilities.

5. The optimized shear regime

The second high performance regime studied dur-ing DTE1 is the optimized shear scenario which wasdeveloped in JET to produce plasmas with energyconfinement and fusion performance higher than thepresent ELMy H modes. Improved confinement of

the central plasma was first obtained at JET by deeppellet fuelling of non-sawtoothing plasmas and sub-sequent heating of the core [32]. The analysis of thesedischarges showed that the magnetic shear in theplasma centre was low and slightly negative and thatthe transport was reduced [33]. With negative mag-netic shear, plasmas are predicted to remain stableat higher plasma pressure, and toroidal drift insta-bilities such as trapped particle and ion tempera-ture instabilities are predicted to be stabilized [34].In several experiments including JET, current profilecontrol has allowed the production of regimes wherethe heat transport is reduced significantly, close toneoclassical levels, within an internal transport bar-rier (ITB) [34–37]. This is in line with predictivemodelling [38] of such discharges, which takes intoaccount the effect of low magnetic shear togetherwith stabilization by rotational shear. However, scal-ing laws for plasmas with ITBs have not yet beendetermined. In particular, the plasma isotope depen-dence of the threshold power for triggering an ITBhas not been firmly established, although it has beenreported from TFTR [4, 6] that the power needed forthe formation of an ITB was much higher in DT thanin DD.

5.1. Current rise scenario

The method used for optimizing the currentprofile in JET experiments has been described inRefs [37, 39]. The plasma current is ramped fairlyrapidly, at 0.4 MA/s, so that current diffusion fromthe plasma edge to the plasma centre takes longerthan the current ramp-up. A faster ramp-up resultsin strong MHD activity, leading either to anoma-lous current diffusion or even to disruptions. Lowpower lower hybrid current drive, at the level of1–2 MW, is applied promptly after the formation ofthe plasma to ensure a broad current profile in theearly phase of the discharge and to provide strongelectron heating in order to arrest current diffusion.The plasma, which is in a single null X point configu-ration from the first second of the discharge, is thenfurther preheated with 1–2 MW ICRF heating forseveral seconds before the main heating phase latein the current ramp. The ICRF frequency is suchthat heating (via minority species and/or at a har-monic resonance of the main ions) takes place closeto the plasma centre. This combined preheat pro-vides high central electron temperatures (typically6–8 keV) and further arrests the diffusion of the cur-rent. The preheating phase ends with the increase of



0 0.1 0.2 0.3 0.4 0.5 0.6r/a

42746D-T

42940D-T

42426

42682D-T

No ITB

ITBsuccessfullyformed

D-D

2.4

2.3

2.2

2.1

2.0

1.9

1.8

1.7

1.6

q

2.7 < Ip < 3.1MA 3.45 < Bt < 3.85T

JG98

.380

/19c

Figure 14. Profiles of q from magnetic reconstruction

(EFIT) at the start of the high power phase for discharges

with and without production of an ITB. Discharge 42682

in DT is an exact repeat of DD discharge 42426. Dis-

charges 42940 and 42746 use 3He puffs and slightly mod-

ified plasma current and ICRF heating waveforms. A typ-

ical error bar around q = 2 is given. The closer to the

plasma centre, the larger are the error bars.

the ICRF heating power up to 6 MW and the startof NB injection at powers up to 10 MW. The mainheating phase starts approximately 0.4 s later, com-bining NB and ICRF heating of typically 25 MW.It has been observed that a substantial part of theICRF heating is damped, via the second harmonicdeuterium resonance, on the NB ions or on the ther-mal ions when the ion pressure is sufficiently high[40, 41].

5.2. Performance in DD plasmasand modifications for DT plasmas

The highest fusion performance in DD plasmashas been obtained when an H mode transition wasdelayed for as long as possible. This was achievedby using a low target density, a magnetic config-uration which allowed maximum pumping by thedivertor cryopump, a low triangularity and bymaintaining the current ramp-up throughout themain heating phase. When an ITB is established,the resulting good core confinement maintains theplasma loss power below the level required to trigger

an H mode, thus preserving a low pressure gradientat the edge, as in the L mode. The highest fusionyield in JET DD plasmas has been achieved in thisway [40–43].

Initial experiments with tritium gas fuelling anddeuterium NB injection showed that the H modethreshold power was significantly lower in DT plas-mas than in DD plasmas [44, 45], and this wasexpected to be the main problem when translatingthe high performance obtained in DD to DT plas-mas. In spite of this concern, it was found that ITBscan be produced in JET in DT plasmas with similaradditional heating power levels and with similar cur-rent profiles to those in DD plasmas [46]. However,this required some developments of the current risescenario, since the current profiles at the start of themain heating phase in DD and DT discharges werefound to be different, as shown in Fig. 14. In DD plas-mas, the evolution of the radial location of the footof the steep ion temperature gradient is very similarto the evolution of the radius of the q = 2 magneticsurface as inferred using magnetic equilibrium recon-struction calculations [39–41]. Furthermore, the for-mation of an ITB first occurs when a q = 2 magneticsurface appears at a small plasma radius.

In DT plasmas, the calculated central q value wasgenerally higher than in DD plasmas. The differencein current profiles between DD and DT plasmas hasnot been fully understood. As shown in Fig. 14, sim-ilar q profiles can, however, be achieved in DT plas-mas by using some 3He gas in the plasma breakdownphase, modifying slightly the plasma current ramprate and increasing the ICRF heating power duringthe preheat phase. For the formation of the ITB asystematic analysis has also indicated a strong linkbetween the total energy input during the preheatingphase and the time of the L–H transition. When thepreheating energy from both NB and ICRF heatingexceeds 3 MJ, an H mode is triggered early. By usingmodified power waveforms, where the NB preheatenergy is much reduced, and establishing a currentprofile at the onset of the high power phase such thata q = 2 magnetic surface is present, ITBs have beenproduced in DT plasmas under similar conditions tothose in DD plasmas.

5.3. Internal transport barriersin DT plasmas

The time histories of various signals for two simi-lar DT discharges are compared with a DD dischargein Fig. 15. All waveforms are very similar, except for



Table 3. Parameters of interest for the high fusion power DT optimized shear discharges 42746

(at 6.82 s) and 42940 (at 6.25 s)

ValuesParameter

Discharge 42746 Discharge 42940

Ip (MA) 3.2 3.3

BT (T) 3.45 3.85

q95 3.9 3.8

PNB (MW) 18 20

PRF (MW) 2.0 3.0

POH (MW) 0.2 0.3

Pα (MW) 1.6 1.5

PSH (MW) 2.6 2.9

PCX (MW) 0.7 0.5

τdiaE (s) 1.1 0.8

ne(0) (1019 m−3) 3.9 4.4

nD(0) + nT (0) (1019 m−3) 3.2 3.4

Ti(0) (keV) 33 38

Te(0) (keV) 13 12.6

Wdia (MJ) 11 10

dWdia/dt (MW) 9 8.4

[nD(0) + nT (0)]τdiaE Ti(0) (1020 m−3 s keV) 11± 20% 10± 20%

[nD(0) + nT (0)]/ne(0) 0.82 0.8

nT (0)/[nD(0) + nT (0)]: TRANSP 0.34 0.34

nT /(nD + nT ) at edge: recycling light 0.14 0.2

Neutron rate (1018 s−1) 2.9± 10% 2.6± 10%

Fusion power, PDT (MW) 8.2± 10% 7.3± 10%

the plasma current in the high power phase, whichis ramped down for DT discharge 42750, and thefrequency of the ICRF heating, which is adjustedfor hydrogen minority in DT discharge 42746 andDD discharge 40847 and for 3He minority in DT dis-charge 42750. In DT, a sequence of events similar tothose in DD has been achieved, namely the trigger-ing of an ITB with an L mode edge followed by anELM-free H mode.

However, the time duration of the L mode edgephase is shorter in DT than in DD, consistent witha lower H mode threshold power in DT. It is to benoted that the ramp-down in plasma current resultsin shortening the L mode edge phase but increasesthe ELM-free H mode edge phase. The central iontemperature rises to 35 keV (compared with 28 keVin DD) and the fusion power reaches 8.2 MW fordischarge 42746 as shown in Table 3 and Fig. 15.The initial plasma tritium concentration, nT /(nD +nT + nH), which has been kept low to reduce thechance of an H mode transition, has been varied from17 to 30% with no effect on the formation of the

ITB. Figure 15 shows that the full potential of theoptimized shear discharges in DT has not yet beenrealized.

It has been shown [40, 41] that the density buildupwithin the ITB can be attributed mainly to NBfuelling during the L mode edge phase. Recyclingfrom the walls appears to be negligible during thatphase. Therefore, the NB system has been configuredto maximize the injection of tritium. The average tri-tium concentration in the plasma, measured by neu-tral particle analysis (NPA), is shown in Fig. 16 fortypical optimized shear discharges in DT. When theinitial tritium content is low, the tritium concentra-tion increases steadily but does not reach the levelof the NB mixture during the L mode phase. Whenthe ELM-free H mode is triggered, the tritium con-centration rises substantially due to the influx of tri-tium from the walls. Data consistency analyses usingTRANSP code simulations have indicated a tritiumconcentration of 29% at 6.8 s for discharge 42746 andof 34% at 6.2 s for discharge 42940, in reasonableagreement with the NPA data. It is clear from these



Figure 15. Time history of typical signals for three sim-

ilar discharges at BT = 3.43 T. Ip increases from 2.4 MA

at t = 5 s up to 3.25 MA at t = 7.2 s for discharges 42746

and 40847. For discharge 42750, Ip is ramped down from

2.9 MA at t = 6.2 s.

5.0 5.5 6.0 6.5 7.0

0.1

0.2

0.3

0.4

0.5

20%

17%

30%

H mode

H mode

42940

42750

42746

Time (s)

JG98

.703

/16c

n T/(

n T+

n D)

Initialconcentration

Figure 16. Time evolution of tritium concentration for

typical optimized shear DT discharges from neutral par-

ticle analyser measurements. The initial concentration is

estimated from visible spectroscopy measurements.

data that the DT mixture has not yet been opti-mized for fusion power production: the discharges

3.53.0 4.0R (m)

0

0.5

1.0

1.5

Pulse No: 42746 (t = 6.6s)Pulse No: 40847 (t = 6.3s)

2.0

2.5

P (

105

Pa)

Pkin

Pi

Pe

JG97

.582

/4c

Figure 17. Comparison of pressure profiles for two simi-

lar discharges in DD and DT one second after the start of

the main heating phase. The ion pressure is from charge

exchange spectroscopy measurements and the electron

pressure from lidar measurements.

are a compromise between obtaining the optimummixture and postponing the H mode transition.

It is to be noted that the central densities in DTplasmas are lower than those in DD plasmas. Thisis mainly due to a lower fuelling rate as a conse-quence of the need to maximize the tritium to deu-terium fuelling ratio. Tritium being fuelled by thehigh energy NB injector (151 keV) and deuteriumby the low energy NB injector (76 keV) led to higheraverage injected energies and lower fuelling. In orderto maintain the same total heating power, whilstmaximizing the tritium to deuterium fuelling ratio,the power injected by the lower energy deuteriuminjector (typically ≈ 76 kV) was reduced for theseDT experiments. For similar power DD and DT NBinjection, the tritium fuelling concentration was upto 45% but the overall fuelling within the ITB waslower by at least 15%, with a comparable reduction inplasma density. As a result, the density is lower andthe ion temperature is higher in DT plasmas. Theresulting central pressures at comparable times dur-ing the L mode edge phase nevertheless reach simi-lar values, as shown in Fig. 17. It can also be seenin Fig. 17 that the ITB is slightly narrower in DTthan in DD, which may be due to differences in thecurrent profiles.



5.0 5.5 6.0 6.5 7.0Time (s)

20

10

0630

10

05

4020

0

321

0

8

4

JG98

.380

/21cPfusion

Pulse No: 42746 Pulse No: 42940(M

W)

(keV

)(1

019 m

–3 )

WD

IA (M

J)(M

W)

P (

MW

)

neo

PICRH

PNBI

Ti (3.1m)

Figure 18. Time history of typical signals for discharges

42940 and 42746 in DT at BT = 3.85 T and 3.45 T,

respectively. Ip is increased from 2.6 MA at t = 4.8 s up

to 3.3 MA at t = 6.18 s.

0.2 0.3 0.4 0.5 0.6 0.7

0.01

0.1

1

10

χ i (

m2 /

s)

r/a

Pulse No:42940 BT = 3.85T

t = 0.1s

t = 0.6s

t = 0.9s

Neoclassical values at t = 0.9s

JG98

.703

/19c

Figure 19. Ion thermal diffusivities from a TRANSP

analysis for discharge 42940 in DT as compared with neo-

classical values. The time 0.1 s is shortly before the for-

mation of an ITB (at 0.3 s), while the time 0.9 s is just

before termination of the pulse.

It is to be noted that the central impurity con-centration in these discharges remains low, withZeff between 1.4 and 1.8, and there is no indica-tion of impurity accumulation within the ITB. The

concentration of the hydrogenic species ((nD +nT +nH)/ne) remains above 80%. These low values of Zeff

are attributed to the clean starting conditions and tothe continuous fuelling within the ITB.

While the early part of the main heating phaseleading to the triggering of an ITB has been reason-ably well optimized in DT, the limited neutron bud-get has prevented full optimization of the high powerphase. Such an optimization uses real time control ofboth the ICRF and the NB heating power to max-imize the core pressure, whilst avoiding disruptionsdue to excessive plasma pressure gradients. As dis-cussed in Ref. [47], the core pressure can be increasedwhen the volume of the ITB expands. In DD plasmas,this optimization has been achieved by controllingthe power in real time with feedback on the neutronyield. The ICRF power is normally stepped downwhen the neutron yield reaches a pre-set value; sub-sequently the NB power is controlled by a pre-setneutron yield waveform which usually increases lin-early with time. In DT plasmas, where the neutronyield depends on the DT mix, an effective substi-tute for this signal has been used, consisting of thesquare of the diamagnetic energy multiplied by theratio between the central and the edge line aver-aged electron density from the infrared interferom-eter (representing the pressure peaking), which canbe computed in real-time.

Figure 18 shows the time history of typical sig-nals for two DT discharges at 3.45 and 3.85 T.For discharge 42746, the L–H transition occurredearly, limiting the fusion power to 8.2 MW. Fordischarge 42940 the demand for the neutron yieldwas increased but the resulting pressure gradientswere too large and a disruption occurred at 6.3 s;the maximum fusion yield was 7.3 MW. The centralion temperature is very high, close to 40 keV, withion temperature gradients of 150 keV/m, toroidalrotation speeds of 750 km/s and plasma pressuregradients of 1 MPa/m [46]. This high tempera-ture can be explained by the combined action ofhigh energy NB injection and effective dampingof the ICRF waves not only on hydrogen minor-ity ions but also on deuterium ions (second har-monic) and tritium ions (third harmonic) [40, 41].TRANSP code analysis shows [46] that, within theITB, the ion thermal diffusivity decreases dramat-ically, approaching neoclassical [48] values in theplasma centre (Fig. 19), as in DD plasmas. It isto be noted that the ion heat transport is lower at3.85 T than at 3.45 T, in line with the higher iontemperature.



Figure 20. Time evolution of βN for typical DD and

DT discharges compared with the unstable region calcu-

lated for the n = 1 ideal pressure driven kink mode [47].

In general, power is controlled to avoid disruption (not

successful for discharge 42940).

An important aspect of these discharges is thestability of the plasma core. An MHD stabilityanalysis [47] indicates the need to avoid global n = 1ideal pressure driven kink modes. Domains of insta-bility are indicated in Fig. 20 for typical DT dis-charges and their counterparts in DD. The path ofβN versus peaking of the plasma pressure in DDdischarges indicates the route for optimizing the

Figure 21. Time history of discharge 42733 which devel-

ops an ITB with an ELMy H mode edge. BT = 3.4 T

and the Ip is ramped from 2.3 MA at t = 4.8 s to 3 MA

at t = 6.3 s. The power was stepped down to limit the

number of 14 MeV neutrons.

performance of such plasmas. The peaking of theplasma pressure is high when the ITB is first formedand βN increases quickly towards the stability limit.The ITB then starts to expand after its formationand, by controlling the injected power, it is possibleto increase further the plasma performance by main-taining βN close to its limiting value. The peakingdecreases substantially when the ELM-free H modeis formed, due to the large pressure gradient at theedge. The high performance phase of the dischargeis then terminated by a giant ELM. Discharge 42940in DT was not optimized, and better control of theplasma pressure would have allowed substantial gainsin fusion power.

5.4. Simultaneous internal andedge transport barriers

The highest fusion performance was normallyobtained by prolonging the phase during which theplasma edge was in L mode; the subsequent ELM-free H mode phase made these discharges transient.A significant number of discharges, however, devel-oped both an ITB and an ELMy H mode edge [43]as illustrated in Fig. 21, with a substantial fusion



yield being produced. In discharge 42733, an ITBis formed and the central ion temperature reaches24 keV, while the edge ion temperature is about3 keV, typical of an ELMy H mode plasma. Such dou-ble barrier discharges show great potential for steadystate operation.

In discharge 42733 the fusion power increases fromthe start of the main heating phase until it reaches6.8 MW, at which time the input power is reduced toeconomize on DT neutrons. This increase in fusionyield is due to a continuous buildup of central densitytogether with an increase of the tritium concentra-tion. The stored plasma energy reaches 8.8 MJ for atotal injected power of 18.4 MW and a correspondingconfinement enhancement factor H89 ≈ 2.3 relativeto the ITER89-P scaling [49]. In this discharge, asin similar DT and DD discharges, the positions ofthe q = 2 magnetic surface and the ITB change onlyslowly with time; this is a consequence of the genera-tion of an edge bootstrap current. At 6.3 s, the powerwas stepped down, and it is interesting to note thatthe subsequent collapse of the ITB triggers an ELM-free H mode.

This route, which could not be explored furtherduring DTE1 due to the imposed constraint on thenumber of DT neutrons produced, shows significantpromise for steady state high fusion yield DT plas-mas, but would require a technique for better controlof the plasma edge and/or current profile.

5.5. Stability of Alfven eigenmodes

In the optimized shear scenario, AEs are found tobe always unstable, due to the presence of a largehigh energy tail generated by the high ICRF heat-ing power typically required to produce the ITB inthis regime. The most favourable scenario for theobservation of alpha particle effects on AEs is createdby suddenly switching off the auxiliary heating [50].Alfven eigenmode activity is observed around 0.5 sinto the ‘afterglow’ phase of some of these discharges.Stability calculations performed by the CASTOR-Kmodel show that, after a few hundreds of millisec-onds into the afterglow, the drive from slowing downalpha particles can overcome the damping mecha-nisms for the AEs. However, the presence of a slow-ing down tail of ICRF generated ions provides anadditional drive which can be comparable with thealpha particle drive. A definitive identification of theadditional contribution of the alpha particles to theICRF drive is very difficult due to uncertainties inthe ICRF drive calculations.

6. Summary, conclusions and outlook

The main results of the article are now placed inthe wider context of present tokamak research, thefuture JET programme and the final goal of a fusionreactor.

The ELM-free H mode has been the establishedroute to high performance, setting world recordsin fusion performance, validating DD projectionsof DT performance, and demonstrating clear alphaparticle heating. The experiments not only led to16.1 MW of fusion power and transient values ofQPtot of 0.95 ± 0.17, consistent with the high valuesof nDT (0)τdiaE Ti(0) = 8.7× 1020 m−3 s keV, but alsodemonstrated the expected factor of 210 between DDand DT fusion power, confirming the quality of theJET kinetic data and the physics contained withinthe TRANSP code.

The high values of Q correspond to high values ofPα/Ploss ≈ 16%, so that the alpha particle heatingbecomes a measurable contribution to the power bal-ance. Indeed the alpha particle heating produced arise in electron temperature of the order of 1 keV permegawatt of alpha particle heating, clearly separatedfrom isotope effects, which were shown to be verysmall. This behaviour is consistent with the classicalalpha particle power deposition profiles.

The mixture control experiments enabled an opti-mum DT plasma mix to be achieved (even with anon-optimal NB mix) and played a crucial role inthe success of the ELM-free H mode programme inDT. The methodology developed is of general appli-cation and will be of value to future DT experimentson JET and elsewhere.

High performance ELM-free H mode plasmas werelimited by giant ELMs to βN ≈ 2.2, thus prevent-ing Qin from exceeding 0.64. This value of βN islower than the expected Troyon limit (βN ≈ 2.8)because the pressure profile is flatter than the opti-mum and the limit is set by a steep pressure gradientin the plasma edge. On the other hand, in regimeswith more peaked pressure profiles such as the TFTRsupershot regime or the optimized shear regimediscussed in Section 5, βN is limited to ≈1.9 byMHD instabilities associated with the steep pressuregradients in the plasma core.

The different plasma conditions obtained in thesupershot and ELM-free H mode regimes haveallowed complementary fusion studies to be pur-sued. In the TFTR supershot, higher input powersand peaked pressure profiles led to a higher centralfusion power density which has been exploited for the



investigation of MHD effects (including sawteeth) onalpha particles. In the JET ELM-free H mode, betterconfinement led to a higher fusion power and Q andclearly visible alpha particle heating (despite flatterpressure profiles), with the highest electron tem-perature showing a clear correlation with the max-imum alpha particle heating power and the opti-mum (≈40:60) DT mixture. This demonstration ofalpha particle heating, in conditions where there isno isotope effect on energy confinement and TAEsare stable (in agreement with predictions), reveals nounexpected effects which might prevent ignition in alarger device. A study of alpha particle heating in thepresence of alpha particle driven TAEs would requirea factor of 2 increase in 〈βα〉 above that obtained sofar in JET hot ion H mode discharges.

The optimized shear regime is less mature thaneither the ELM-free H mode or the supershot mode,but has the potential for combining the best fea-tures of both these regimes and to further the studyof alpha particle physics. Substantial progress wasmade during the scenario development which led tothe production of 8.2 MW of fusion power. First, andmost importantly, strong internal transport barrierswere shown to exist in DT. These barriers could beestablished with similar powers to those required inDD, in contrast to the edge H mode barrier whosethreshold is lower in DT. Second, the poor mix-ture control and reduced core fuelling by NB injec-tion (tritium concentration only 30%) limited thefusion performance, but was related to short termtechnical constraints rather than to a fundamen-tal deficiency. Future JET experiments would ben-efit from tritium injection from both NB injectors,and the methodology for optimizing the plasma mix-ture applied successfully to the ELM-free H modecould then also be adopted. Third, the evolution ofthe pressure profile peaking was successfully con-trolled very close to the maximum permitted bypressure gradient driven stability in DD, but couldnot be optimized in DT (βN ≈ 1.9). It is reason-able to conclude that, given sufficient experimentaltime and DT neutron budget, this regime shoulddeliver the fusion performance and Q commensu-rate with the high values of the fusion triple product(nDT (0)τETi(0) ≈ 1021 m−3 s keV) demonstrated.

The high performance regimes considered in thisarticle would be even more reactor relevant if theycould be maintained in steady state. To realizethe potential suggested by the high values of QPtotobtained in the ELM-free H mode would requireexercising control over the steep gradients in the

edge transport barrier which have led, so far, inex-orably towards instability. Such control would alsobe required for scenarios in which an ITB co-existedwith an ELM-free H mode edge, since high per-formance is then also terminated by a giant ELM.A pressure profile intermediate between that of theELM-free H mode and the optimized shear regimewith an ITB and an L mode edge would be required.Such pressure profiles are to be found in dischargesin which an ITB has been obtained simultaneouslywith an ELMy H mode edge, as already observedin DD on JET, JT-60U and DIII-D, and in DT onJET [43]. This scenario could offer an alternative forsteady state high fusion performance to the stan-dard ELMy H mode (the presently established sce-nario for ITER), but would require high βp (plasmapressure normalized to the pressure of the poloidalmagnetic field) for steady state operation and highβN together with high confinement for high perfor-mance. Calculations show that such discharges onJET can be MHD stable up to βN = 2.5 [47], andfurther optimization of the current profile can evenlead to stable βN values in excess of 3 [51]. Thiswould require not only edge control to maintain non-disruptive edge pressure gradients (e.g., using cur-rent ramps or impurity radiation at the edge), butalso careful current profile control in addition tothe pressure profile control already demonstrated toavoid pressure driven disruptions at high βN . Theachievement of such high performance conditions anda double transport barrier would allow the effectsof ELMs on alpha particle behaviour to be inves-tigated with the presently available additional heat-ing, whereas ELMy H mode plasmas without an ITBwould require a significant increase in heating powerfor such studies.

The DT results presented in this and the com-panion [9] article confirm a mature understand-ing of H mode plasmas and indicate significantprogress towards the achievement of a similar levelof understanding of plasmas with ITBs. The signifi-cant progress in understanding made over the yearsin tokamak fusion physics can now be exploited inJET and future devices to advance the realization ofa tokamak fusion reactor.

Appendix

The JET Team

J.M. Adams1, P. Ageladarakis, B. Alper, H.Altmann, V. Amasov2, P. Andrew, Y. Andrew3,



S. Arshad, D. Bailey, N. Bainbridge, B. Balet, Y.Baranov4, P. Barker, R. Barnsley5, M. Baronian,D.V. Bartlett, A.C. Bell, L. Bertalot6, E. Bertolini,V. Bhatnagar, A.J. Bickley, H. Bindslev, K. Black-ler, D. Bond, T. Bonicelli, D. Borba7, M. Bran-don, P. Breger, H. Brelen, P. Brennan, W.J. Brew-erton, M.L. Browne, T. Budd, R.V. Budny8, A.Burt, T. Businaro, M. Buzio, C. Caldwell-Nichols,D. Campling, P. Card, C.D. Challis, A.V. Chankin,D. Chiron, J. Christiansen, P. Chuilon, D. Ciric,R. Claesen, H.E. Clarke, S. Clement, J.P. Coad,I. Coffey9, S. Conroy10, G. Conway11, S. Cooper,J.G. Cordey, G. Corrigan, G. Cottrell, M. Cox9,S.J. Cox, R. Cusack, N. Davies, S.J. Davies, J.J.Davis, M. de Benedetti, H. de Esch, J. de Haas, E.Deksnis, N. Deliyanakis, A. Dines, S.L. Dmitrenko,J. Dobbing, N. Dolgetta, S.E. Dorling, P.G. Doyle,H. Duquenoy, A.M. Edwards9, A.W. Edwards, J.Egedal, J. Ehrenberg, A. Ekedahl12, T. Elevant12,J. Ellis, M. Endler13, S.K. Erents9, G. Ericsson10,L.G. Eriksson, B. Esposito6, H. Falter, J.W. Far-thing, M. Fichtmuller, G.M. Fishpool, K. Fullard, M.Gadeberg, L. Galbiati, A. Gibson, R.D. Gill, D. God-den, A. Gondhalekar, D. Goodall9, C. Gormezano,C. Gowers, M. Groth14, K. Guenther, H. Guo, A.Haigh, B. Haist15, C.J. Hancock, P.J. Harbour,N.C. Hawkes9, N.P. Hawkes1, J.L. Hemmerich, T.Hender9, J. Hoekzema, L. Horton, A. Howman, M.Huart, T.P. Hughes, F. Hurd, G.T.A. Huysmans, C.Ingesson16, B. Ingram, M. Irving, J. Jacquinot, H.Jaeckel, J.F. Jaeger, O.N. Jarvis, M. Johnson, E.M.Jones, T.T.C. Jones, J.-F. Junger, F. Junique, C.Jupen, J. Kallne10, A. Kaye, B.E. Keen, M. Keil-hacker, W. Kerner, N.G. Kidd, S. Knipe, R. Konig,A. Korotkov, A. Krasilnikov2, J.G. Krom, P. Kup-schus, R. Lasser, J.R. Last, L. Lauro-Taroni, K.Lawson9, M. Lennholm, J. Lingertat, X. Litaudon17,T. Loarer, P.J. Lomas, M. Loughlin, C. Lowry, A.C.Maas16, B. Macklin, C.F. Maggi, M. Mantsinen18, V.Marchese, F. Marcus, J. Mart, D. Martin, G. Matth-hews, H. McBryan, G. McCracken, P.A. McCullen,A. Meigs, R. Middleton, P. Miele, F. Milani, J.Mills, R. Mohanti, R. Monk, P. Morgan, G. Mur-phy, M.F. Nave7, G. Newbert, P. Nielsen, P. Noll,W. Obert, D. O’Brien, M. O’Mullane, E. Oord, R.Ostrom, S. Papastergiou, V.V. Parail, R. Parkin-son, W. Parsons, B. Patel, A. Paynter, A. Peacock,R.J.H. Pearce, A. Pereverentsev, M.A. Pick, J. Plan-coulaine, O. Pogutse, R. Prentice, S. Puppin, G.Radford19, R. Reichle, V. Riccardo, F.G. Rimini, F.Rochard17, A. Rolfe, A.L. Roquemore8, R.T. Ross,A. Rossi, G. Sadler, G. Saibene, A. Santagiustina,

R. Sartori, R. Sauders, P. Schild, M. Schmid, V.Schmidt, B. Schokker16, B. Schunke, S.M. Scott, S.Sharapov, A. Sibley, M. Simon, R. Simonini, A.C.C.Sips, P. Smeulders, P. Smith, R. Smith, F. Soldner,J. Spence, E. Springmann, R. Stagg, M. Stamp, P.Stangeby20, D.F. Start, D. Stork, P.E. Stott, J.D.Strachan8, P. Stubberfield, D. Summers, L. Svens-son, P. Svensson, A. Tabasso3, M. Tabellini, J. Tait,A. Tanga, A. Taroni, C. Terella, P.R. Thomas, K.Thomsen, E. Traneus10, B. Tubbing, P. Twyman, A.Vadgama, P. van Belle, G.C. Vlases, M. von Heller-mann, T. Wade, R. Walton, D. Ward, M.L. Watkins,N. Watkins1, M.J. Watson, M. Wheatley, D. Wilson,T. Winkel, D. Young, I.D. Young, Q. Yu21, K.-D.Zastrow, W. Zwingmann.

Permanent affiliations

1. UKAEA, Harwell, Didcot, Oxfordshire, UK.2. TRINITI, Troitsk, Moscow, Russian Federation.3. Imperial College, University of London, UK.4. A.F. Ioffe Institute, St. Petersburg, Russian

Federation.5. University of Leicester, Leicester, UK.6. ENEA, CRE Frascati, Rome, Italy.7. IST, Centro de Fusao Nuclear, Lisbon, Portugal.8. Princeton Plasma Physics Laboratory, Prince-

ton University, Princeton, New Jersey, USA.9. UKAEA Culham Laboratory, Abingdon,

Oxfordshire, UK.10. Department of Neutron Research, Uppsala Uni-

versity, Sweden.11. University of Saskatchewan, Saskatoon, Canada.12. Royal Institute of Technology, Stockholm, Swe-

den.13. Max-Planck-Institut fur Plasmaphysik, Gar-

ching, Germany.14. University of Manchester Institute of Science

and Technology, Manchester, Lancashire, UK.15. FZ, Julich, Germany.16. FOM Instituut voor Plasmafysica, Nieuwegein,

Netherlands.17. CEA Cadarache, St. Paul-lez-Durance, France.18. Helsinki University of Technology, Espoo, Fin-

land.19. Institute of Theoretical Physics, University of

Oxford, Oxfordshire, UK.20. Institute for Aerospace Studies, University of

Toronto, Ontario, Canada.21. Institute of Plasma Physics, Hefei, China.



References

[1] JET Team, Nucl. Fusion 32 (1992) 187.

[2] JET Team (presented by D. Stork), in Plasma

Physics and Controlled Nuclear Fusion Research

1994 (Proc. 15th Int. Conf. Seville, 1994), Vol. 1,

IAEA, Vienna (1995) 51.

[3] Aymar, R., et al., in Fusion Energy 1996 (Proc. 16th

Int. Conf. Montreal, 1996), Vol. 1, IAEA, Vienna

(1997) 3.

[4] McGuire, K.M., et al., ibid., p. 19.

[5] Levington, F.M., et al., Phys. Rev. Lett. 75 (1995)

4417.

[6] Scott, S.D., et al., in Fusion Energy 1996 (Proc. 16th

Int. Conf. Montreal, 1996), Vol. 1, IAEA, Vienna

(1997) 573.

[7] Lasser, R.D., et al., in Fusion Engineering (Proc.

17th IEEE/NPSS Symp. San Diego, 1997), Vol. 1,

IEEE, Piscataway, NJ (1997) 292.

[8] JET Team (presented by G.C. Vlases), in Fusion

Energy 1996 (Proc. 16th Int. Conf. Montreal, 1996),

Vol. 1, IAEA, Vienna (1997) 371.

[9] Jacquinot, J., et al., Nucl. Fusion 39 (1999)

235.

[10] Andrew, P., et al., Nucl. Fusion 33 (1993) 1389.

[11] Peacock, A., et al., Proc. 13th Int. Conf. on Plasma

Surface Interactions, San Diego, 1998, J. Nucl.

Mater. (in press).

[12] Balet, B., et al., Nucl. Fusion 33 (1993) 1345.

[13] Nave, M.F.F., et al., Nucl. Fusion 37 (1997) 809.

[14] JET Team (presented by P.J. Lomas), in Plasma

Physics and Controlled Nuclear Fusion Research

1994 (Proc. 15th Int. Conf. Seville, 1994), Vol. 1,


[15] Huysmans, G.T.A., et al., Nucl. Fusion 38 (1998)

179.

[16] Lomas, P.J., et al., “Performance of the ELM-free

H mode in DD and DT: The realisation of significant

fusion power” (in preparation).

[17] Marcus, F.B., et al., Nucl. Fusion 33 (1993) 1325.

[18] JET Team (presented by P.J. Lomas), in Fusion

Energy 1996 (Proc. 16th Int. Conf. Montreal, 1996),

Vol. 1, IAEA, Vienna (1997) 239.

[19] Jones, T.T.C., et al., in Controlled Fusion and

Plasma Physics (Proc. 25th Eur. Conf. Prague,

1998), Vol. 22C, European Physical Society, Geneva

(1998) 369.

[20] Rimini, F.G., et al., “Combined heating experi-

ments in ELM-free H modes in JET” (in prepara-

tion).

[21] Nave, M.F.F., et al., “Discharge optimisation and

the control of edge stability” (in preparation).

[22] Gibson, A., JET Team, Phys. Plasmas 5 (1998)

1839.

[23] Fasoli, A., et al., Plasma Phys. Control. Fusion 39

(1997) B287.

[24] Sharapov, S.E., et al., “Stability of alpha particle

driven Alfven eigenmodes in high performance JET

DT plasmas”, submitted to Nucl. Fusion.

[25] Nazikian, R., Fu, G.Y., Phys. Rev. Lett. 78 (1997)

2976.

[26] Kramer, G.J., et al., Phys. Rev. Lett. 80 (1998)

2594.

[27] Heidbrink, W., Sadler, G.J., Nucl. Fusion 34 (1994)

535.

[28] Taylor, G., et al., Phys. Rev. Lett. 76 (1996) 2722.

[29] Thomas, P.R., et al., Phys. Rev. Lett. 80 (1998)

5548.

[30] Nave, M.F.F., et al., in Controlled Fusion and

Plasma Physics (Proc. 25th Eur. Conf. Prague,

1998), Vol. 22C, European Physical Society, Geneva

(1998) 365.

[31] Start, D.F.H., et al., in Controlled Fusion and

Plasma Heating (Proc. 17th Eur. Conf. Amsterdam,

1990), Vol. 14B, Part III, European Physical Soci-

ety, Geneva (1990) 1019.

[32] Jacquinot, J., et al., Plasma Phys. Control. Fusion

30 (1988) 1467.

[33] Hugon, M., et al., Nucl. Fusion 32 (1992) 33.

[34] Kessel, C., et al., Phys. Rev. Lett. 72 (1994) 1212.

[35] Strait, E.J., et al., Phys. Rev. Lett. 75 (1995)

4421.

[36] Ushigusa, K., JT-60 Team, in Fusion Energy 1996

(Proc. 16th Int. Conf. Montreal, 1996), Vol. 1,


[37] JET Team (presented by C. Gormezano), ibid.,

p. 487.

[38] Parail, V.V., et al., “Predictive modelling of JET

optimized shear discharges”, submitted to Nucl.

Fusion.

[39] Sips, A.C.C., et al., Plasma Phys. Control. Fusion

40 (1998) 1171.

[40] Cottrell, G.A., et al., in Controlled Fusion and

Plasma Physics (Proc. 24th Eur. Conf. Berchtes-

gaden, 1997), Vol. 21A, Part I, European Physical

Society, Geneva (1997) 81.

[41] Cottrell, G.A., et al., “Ion cyclotron heating of JET

DD and DT plasmas with optimized shear”, Nucl.

Fusion (in press).

[42] JET Team (presented by F.X. Soldner), Plasma

Phys. Control. Fusion 39 (1997) B353.

[43] Soldner, F.X., et al., “Approach to steady state high

performance in DD and DT with optimized shear on

JET”, submitted to Nucl. Fusion.

[44] JET Team (presented by M. Keilhacker), Plasma

Phys. Control. Fusion 39 (1997) B1.

[45] JET Team (presented by J. Jacquinot), in Con-

trolled Fusion and Plasma Physics (Proc. 24th Eur.

Conf. Berchtesgaden, 1997), Vol. 21A, Part IV,

European Physical Society, Geneva (1997) 1865.

[46] Gormezano, C., et al., Phys. Rev. Lett. 80 (1998)

5544.



[47] Huysmans, G.T.A., et al., in Controlled Fusion

and Plasma Physics (Proc. 24th Eur. Conf.

Berchtesgaden, 1997), Vol. 21A, Part I, European

Physical Society, Geneva (1997) 21.

[48] Hinton, F.L., Hazeltine, R.D., Rev. Mod. Phys. 48

(1976) 239.

[49] Schissel, D.P., et al., Nucl. Fusion 31 (1991) 73.

[50] Budny, R.V., et al., Nucl. Fusion 32 (1992) 429.

[51] Bondeson, A., et al., Nucl. Fusion 37 (1997) 1419.

(Manuscript received 29 June 1998Final manuscript accepted 20 November 1998)

E-mail address of M. Keilhacker: [email protected]

Subject classification: F0, Te; G0, Te; M0, Te


Documents

High fusion performance from deuterium-tritium plasmas in JET