Deuterium-tritium plasmas in novel regimes in the Tokamak Fusion Test Reactor

of and and ﬁrst The interactions of energetic ions with ion Bernstein produced by mode conversion from fast in mixed-species studied as a possible mechanism the of the to


I. INTRODUCTION
Since the Tokamak Fusion Test Reactor ͑TFTR͒ 1 began its deuterium-tritium ͑D-T͒ phase of operation in December 1993, more than 1.2 GJ of D-T fusion energy has been produced. Over this period, 841 plasmas containing high concentrations of tritium have been made for a wide variety of experiments. About 90 g of tritium has been processed. TFTR has achieved high availability for experiments while maintaining a record of safe operation and compliance with the regulatory requirements of a nuclear facility. The tokamak, heating systems, and power supplies have all been operated at, or beyond, their original design specifications.
During the first year of D-T operation, experiments concentrated on achieving the maximum fusion power in order to validate the extrapolability of experience in deuterium plasmas to D-T plasmas and to study alpha-particle physics in the most reactor relevant conditions. 1 During that period, it became apparent that the fusion performance of TFTR was being limited by plasma stability and that the development of alternate modes of operation could extend its ability to explore reactor relevant phenomena in D-T plasmas. For the last 18 months, therefore, considerable effort has been devoted to developing new operational regimes which offer the possibility of increased plasma stability while preserving the good confinement and extremely high fusion reactivity of existing TFTR regimes.
In February 1995, it was discovered that plasmas in TFTR with reversed magnetic shear (‫ץ‬q/‫ץ‬r Ͻ 0) in the central region could undergo a spontaneous transition during neutral beam heating to a state of enhanced confinement, the so-called enhanced reverse shear ͑ERS͒ regime, 2 which appeared to be associated with the formation of a localized transport barrier in the interior of the plasma. A similar regime was also discovered in the DIII-D tokamak 3 at about the same time and has since been studied in several tokamaks, including JT-60U 4 and the Joint European Torus ͑JET͒. 5 Although it is produced by a different heating method, namely neutral beam injection, the ERS regime has strong similarities to two other regimes of improved confinement involving modification of the q profile, namely the pellet enhanced performance mode in JET 6 and that occurring in Tore-Supra with lower-hybrid current drive. 7 Since reversed magnetic shear also offers the prospect of improved stability to certain pressure-driven magnetohydrodynamic ͑MHD͒ modes, the ERS regime seemed particularly attractive for further exploration in TFTR. Experiments with this regime in the 1996 run are described in Sec. II.
A second line of investigation grew out of previous experiments to improve plasma stability by creating more highly peaked current profiles through current rampdown. 8 This technique, which increases the internal inductance parameter, l i , of the plasma, and produces what is called the high-l i regime, had already achieved high normalized-␤ and significant fusion power, but was limited operationally in its extrapolability to higher performance. An innovative method has now been developed to produce high-l i plasmas at much higher plasma current. Experiments utilizing this technique will be described in Sec. III.
In Sec. IV we discuss the D-T reactivity achieved in these different operational regimes and compare the achieved reactivity with that of extrapolations based on experience in deuterium plasmas. In Sec. V we present recent results in alpha-particle physics while in Sec. VI we describe the experiments with heating by waves in the ion-cyclotron range of frequencies ͑ICRF͒ in various plasma and wave coupling regimes.

II. REVERSED-SHEAR PLASMAS
Plasmas with reversed magnetic shear in the central region are produced in TFTR by applying a period of lowpower neutral beam injection ͑NBI͒ heating ͑typically Ͻ10 MW͒, to large cross-section plasmas while the toroidal current is being ramped up to its final level. 2 This heating, referred to as the NBI ''prelude,'' and the large plasma size combine to inhibit penetration of the induced current, thereby creating a hollow current profile and reversed magnetic shear. After the final current has been reached, a period of high-power NBI is applied to study the confinement and stability properties. The high-power phase may be followed by a second period of lower power NBI, known as the ''postlude'' phase, to sustain the period of ERS confinement. The q profile of a reversed-shear plasma may be characterized, at the most basic level, by the minimum q, q min , and by the normalized minor-radius, min (ϭr/a) of the surface of minimum q. Experiments in 1995 had developed a reliable startup for reversed-shear plasmas at a plasma current I p ϭ1.6 MA ͑major radius R p ϭ2.60 m; minor radius aϭ0.90 m; toroidal magnetic field B T ϭ4.6 T, q a Ϸ5.8͒. 2 These plasmas generally had 2 р q min р 3 and min ϭ 0.3-0.4 and in those plasmas that underwent ERS transitions, the region of improved confinement appeared to coincide with the region of shear reversal, i.e., р min . The 1.6 MA ERS plasmas exhibited a limiting Troyon-normalized-␤, ␤ N ͑ϭ10 8 ␤ T aB T /I p , where ␤ T ϭ2 0 ͗p͘/B T 2 and ͗p͘ is the volume-average plasma pressure͒, of about 2; this modest ␤-limit was attributed to the small volume of high-pressure plasma within the transport barrier. The 1996 reversed-shear experiments continued to use this reliable 1.6 MA plasma for studies of ERS transition physics and the formation of the transport barrier, 9 but a considerable effort was also devoted to exploring higher current scenarios with the goal of producing plasmas having 1Ͻq min Ͻ2 and larger min which theoretical studies 10 had suggested would have a substantially improved ␤-limit. A plasma current of 2.2 MA, corresponding to q a Ϸ4.3 with the other major parameters held fixed, was chosen for this development because it was expected to be compatible with producing a D-T fusion power approaching 10 MW at ␤ N slightly greater than 2.
To produce reversed shear at lower q a , it is necessary to avoid deleterious MHD instabilities, sometimes resulting in disruptions, during the current ramp phase, particularly when the edge q passes through integral values. In standard TFTR operation, the plasma is grown in minor cross-section during the current ramp to bring the q a to its final value as early as possible and then to maintain it constant; this procedure results in rapid current penetration and usually inhibits MHD activity during the current ramp. In order to avoid the MHD activity during the reversed-shear startup, it was found necessary to program brief reductions in the current ramp rate and the prelude NBI heating power as the troublesome integral q a values were approached. Disruptions were particularly a problem if, in addition to passing through an integral q a , the value of q min was simultaneously close to a rational value. As shown in Fig. 1, shear reversal was produced over a larger radius at the higher current. However, the desired reduction in q min could not be achieved simultaneously. Despite variations of the startup phase, including variations of the prelude NBI and the introduction of partial plasma growth to increase current penetration, within the accessible, reliable range of startup conditions at 2.2 MA, lower q min could only be achieved at the expense of reduced min . This apparent relationship between q min and min is illustrated in Fig. 2. Similar difficulty in achieving q min Ͻ2 has also been experienced by the JT-60U team developing reversed-shear plasmas in that device. 4 Once a reliable startup had been developed, it was found that ERS transitions resulting in an improvement in global confinement, such as those observed at 1.6 MA, did not occur spontaneously under similar conditions at high current, possibly because the threshold power for the transition had increased beyond the available NBI power. ͑The peak deuterium NBI power available for ERS studies has been limited to about 29 MW because a longer total NBI heating duration is required in this mode of operation to span the prelude and postlude phases.͒ However, the transient formation of regions of increased gradient in the temperature profiles, particularly of the electron temperature, as opposed to the density profile, was observed in some 2.2 MA reversed-shear plasmas. These events were found to be associated with q min crossing rational values, particularly q min ϭ 5 2 and 3; at higher rational values the temperature perturbation became progressively weaker. This phenomenon, while not producing profound changes in overall confinement, may shed light on the underlying mechanisms of confinement in these complex plasmas.
Experience at lower current had suggested a role for edge conditions in determining the threshold power for the ERS transition. In particular, the use of lithium pellet injection before the start of the high-power NBI phase ͑by 0.1-0.5 s͒ had been found to reduce the threshold power. Lithium pellet injection was also found to stimulate ERS transitions at 2.2 MA, but only when the pellet injection essentially coincided with the start of the high-power phase: a delay of as little as 0.15 s between the pellet and the start of the high-power phase would inhibit the effect. Since the effect of lithium injection on wall influxes is known 11 to persist for periods of the order of 1 s, the mechanism for stimulation of the ERS transition by the pellet must involve other effects on the plasma, perhaps on the heating and temperature profiles. Once stimulated by the pellet injection, the high-current ERS phase resembled that at lower current: the plasma developed a very well-confined core inside a region of very steep gradients, particularly in the density, as illustrated in Fig. 3. Comparing the 1.6 and 2.2 MA data in this figure, it can be seen that the location of the transport barrier has indeed expanded with the increase in min , as observed in similar regimes in other devices. 4,12 Analysis of the transport in the high-current ERS plasmas shows that, as at lower current, the ion thermal and the particle diffusivities are reduced but that the electron thermal diffusivity is only slightly affected. The suppression of the transport in the ERS regime is correlated with a reduction inside the transport barrier in the level of turbulent density fluctuations measured by a multi-channel microwave reflectometer. 13 The fluctuations become suppressed at the time of the ERS transition and reappear when the plasma reverts to L-mode ͑low-confinement͒ towards the end of the NBI pulse.
Whereas some 1.6 MA ERS plasmas had reached ␤ N Ϸ2.0 without disruption, the 2.2 MA ERS plasmas suffered frequent disruptions, not only during the high-power NBI phase when ␤ N was rising but also, as shown in Fig. 4, during the postlude phase when ␤ N was decreasing in time. The highest normalized-␤ achieved at 2.2 MA was 1.45. The cause of this low ␤-limit has been investigated experimentally and theoretically. 14 The disruption is believed to arise from the growth of an ideal infernal/kink mode with toroidal mode number nϭ1 that is driven primarily by the pressure gradient in the region of low magnetic shear around the surface of minimum q. During the ERS phase, extreme pressure gradients develop in this weak shear region and persist even when the NBI power is reduced in the postlude due to the low transport. The progressive reduction of the ␤-limit with time occurs because, as the q-profile evolves on a resistive time scale, q min approaches 2. Parametric studies of reversed-shear stability have shown that the ␤-limit decreases when q min is close to low-order rational values. 10 This emphasizes the importance of developing techniques for controlling both the current and pressure profiles if we are to take advantage of advanced operating modes, such as ERS, in the future. While the ␤-limit at 1.6 MA in TFTR, ␤ N Ϸ2.0, appears low compared to the highest value reported for this regime in DIII-D, 12 it is actually very similar to the highest values reached in high-performance reversed-shear modes in JT-60U 4 and in JET 5 at comparable magnetic field.
While considerable time during the last run was devoted to developing the ERS regime and studying the physics of the associated transport barrier, only a few such plasmas were attempted in D-T, all of these at 1.6 MA. The long overall NBI pulses required for the reversed-shear startup provides a practical constraint on the number of D-T shots that can be taken in this mode of operation in any one experiment. When D-T NBI was applied to 1.6 MA reversedshear plasmas, it was found that the threshold power for the ERS transition was considerably higher than for D-only NBI. Whereas with a well-conditioned limiter, about 16 MW of D-NBI ͑six NBI sources͒ was sufficient to induce an ERS transition, 27 MW was required in D-T ͑seven T-NBI and three D-NBI sources͒, at which power level the plasma would rapidly approach the ␤-limit following the ERS transition. The variation of the threshold power with isotopic mixture and also with magnetic field 9 provide clear tests for theories of ERS confinement. As a result of the difficulty of producing a suitable fuel mixture in the ERS plasmas, the maximum D-T fusion power produced by an ERS plasma has only reached 3.1 MW, although higher D-T power has been reached in both non-ERS reversed-shear plasmas and plasmas with weak positive shear having q 0 Ͼ1. The 2.2 MA reversed-shear plasmas suffered from an additional impediment to D-T operation: the lithium pellets injected at the start of the high-power NBI to produce ERS transitions compromised the fusion reactivity of these plasmas. The internal transport barrier of the ERS plasmas caused the injected lithium to be retained in the plasma core, significantly diluting the reacting species. The lithium density profile measured by charge-exchange recombination spectroscopy was also found to have a very steep gradient at the transport barrier during the ERS phase, similar to the electron density profile. In 2.2 MA deuterium ERS plasmas, the peak DD reactivity was between 40% and 80% of that expected on the basis of the plasma total stored energy, scaling from both ERS plasmas at 1.6 MA and supershots in similar conditions of plasma size, current, and magnetic field. This suggests that accumulation of helium ash may pose a problem for achieving sustained ignition in the ERS regime without active helium removal techniques.

III. HIGH-l i REGIME EXPERIMENTS
Plasmas in the high-l i regime, with the current profile modified by ramping down the total current before or during the NBI heating pulse, have previously been shown to have improved stability, as measured by increases in the normalized-␤ sustainable without disruption. 8 However, in terms of the absolute-␤, ␤ T , such plasmas did not exceed the level that could have been achieved at the maximum plasma current before the current rampdown. A new technique has been developed in TFTR to produce current profiles similar to those generated by rampdown but at higher plasma current. 15,16 The technique, which is illustrated in Fig. 5, involves producing a high-current Ohmically heated plasma with reduced minor cross-section, and therefore, very low edge q, typically 2.3. This low-q plasma is then expanded in cross-section immediately before NBI heating to produce a plasma with a low current density in the outer region and, consequently, increased internal inductance. The major task in developing this regime was to produce the low-q plasma in a way that was compatible with achieving good limiter conditioning, i.e., low edge influxes of both hydrogen isotopes and carbon, during the NBI phase. This was accomplished by starting the plasma on the outboard limiter, with an aspect ratio of 7 initially, using gas puffing to control MHD activity while q a was reduced to 2.3, and then, after the current had been raised at constant q a , making a transition to the main inboard limiter where lithium pellet conditioning could be applied to control the edge influxes. Once optimized, the low-q a phase of these discharges was remarkably reproducible and devoid of MHD activity, although, following the final expansion, locked modes occasionally developed. These modes, which increased the edge influxes during the NBI and degraded confinement, were controlled by a brief period of co-tangential NBI which induced rapid plasma rotation before the main NBI pulse.
This new high-l i startup was developed for plasma currents up to I p ϭ 2.3 MA ͑Rϭ2.52 m, aϭ0.87 m, B T ϭ5.5 T͒. The product I p •I t •l i , where I t is the threading current of the toroidal field coil, which is a measure of the expected maximum plasma energy content, reached 208 MA 2 in these plasmas, exceeding the maximum produced with normal supershot startup techniques at higher plasma current. Compared to standard plasmas with the same global parameters, the sawtooth inversion radius was expanded as a result of the increased current density in the inner region of the plasma. The central q was measured to be in the range 0.75-0.80 ͑Ϯ0.04͒.
These plasmas were extensively studied using both D-only and D-T NBI. With extensive lithium conditioning applied to the limiter, the high-l i plasmas exhibited confinement properties very similar to supershots. The lithium conditioning was performed by the standard TFTR technique of pellet injection 11 and, once, by a new technique of evaporative coating in situ. This utilized a small oven inserted on a probe into the vacuum chamber between shots which deposited the equivalent of about 50 standard lithium pellets. This technique was successful in enhancing the confinement on the subsequent five or six shots with high-power NBI. Ultimately, however, the major limitation on D-T fusion performance of the high-l i plasmas during the last run period was the power handling capability of the limiter. At high NBI power in D-T, the influx of hydrogen isotopes and lithium from the edge increased dramatically during the pulse, degrading confinement to the point where it was not possible to reach the ␤-limit at the highest plasma current. Preliminary experiments were conducted at the end of the last run investigating the use of a radiating boundary, induced by puffing into the plasma small amounts of either argon or krypton, to reduce the peak power flux to the limiter. While the initial results were encouraging, i.e., the radiated power fraction could be increased significantly without affecting global confinement adversely, there was not time to develop this technique for use specifically with the high-l i D-T plasmas.
In order to test the ␤-limit in the high-l i regime, it was necessary to reduce both the plasma current and the toroidal field. In a plasma with I p ϭ2.0 MA, B T ϭ4.74 T, which achieved a transient confinement time of 0.24 s, a fusion power of 8.7 MW was reached before the plasma disrupted at a normalized-␤ of 2.35. The evolution of this shot during NBI is compared with that of a high-current supershot producing a similar fusion power in Fig. 6. This shot at reduced current and field was the only high-l i plasma which reached The internal inductance calculated from magnetic diagnostic data is extrapolated through the NBI pulse when the plasma pressure becomes anisotropic. The increase in the normalized ␤-limit is proportional to the increase in l i . the ␤-limit and the only one which disrupted during the NB heating phase. 16 If the power handling capability of the TFTR limiter can be improved through the use of a radiating boundary, or other means, the high-current versions of the high-l i plasmas already developed should be capable of producing D-T fusion power considerably above 10 MW.

IV. SCALING OF DT REACTIVITY AND MODELING FROM D PLASMAS
An important issue for the design of future fusion experiments is the extrapolability of data obtained from experiments with deuterium plasmas to eventual operation with D-T plasmas. There are two types of effects to consider here: effects due to changes in the energy dependence of the reaction cross-sections and effects on the plasma itself resulting from the change in plasma composition, the so-called isotopic effects. While the first type might seem straightforward to calculate, the result can be changed significantly in practice by a combination of subtle changes of the second type, particularly because a plasma is usually subject to multiple constraints simultaneously. For example, in TFTR, a change from D to T NBI is accompanied by a change in beam acceleration voltage, total beam power, beam species mix, power and particle deposition profiles, ripple loss, beam thermalization time, ion and electron heating fractions, and also a change in the underlying confinement of the thermalized plasma. 17 The expected fusion reactivity enhancement in D-T plasmas over otherwise identical deuterium counterparts can be estimated from the ratio of the velocityweighted fusion cross-sections for DT and DD reactions. For fixed fuel density and temperatures the fusion power ratio, P D-T /P D-D , of purely thermal reactions reaches an idealized maximum of ϳ225 for T i ϳ12 keV, but the ratio falls to 150 at T i ϭ30 keV. In plasmas with a significant population of nonthermal fuel ions from neutral beam injection, the beamtarget reactivity enhancement also drops for T i above 15 keV. Furthermore, in D-T plasmas, the ion temperature is generally higher than in comparable D plasmas, which is a manifestation of the favorable isotopic effect but which actually penalizes the D-T reactivity. As a result, the measured ratio of fusion power in TFTR supershots is ϳ115 if plasmas with the same stored energy are compared. This ratio would be appropriate if the D plasma were at the ␤-limit, for example. When comparing plasmas with the same heating power, the isotope effect on confinement raises the DT fusion power and the fusion power ratio is then ϳ140. Furthermore, higher neutral beam power can be achieved with D-T operation due to the higher neutralization efficiency of tritium. As a result of this increase in power, the highest DT fusion power is actually 165 times the highest DD fusion power achieved in TFTR. However, it must be noted that to achieve this power ratio, the plasma energy increased from 5.6 MJ in the D plasma to 7.0 MJ in the D-T plasma. This complex behavior is illustrated in Fig. 7. This figure makes use of the fact that in TFTR supershots, in which the plasma energy is dominated by the ion component, a very constrained relationship is observed between the plasma energy and the fusion power output in both D and D-T plasmas. 18 The data in Fig. 7 emphasizes the importance of improving stability limits to achieving high fusion performance and demonstrates that the extrapolation of the highest performance D-only results, which are often limited by stability or power handling, to D-T plasmas is not a simple matter of idealized species substitution.

V. ALPHA-PARTICLE PHYSICS
Alpha-particle physics continues to be a major focus of the TFTR D-T program. Recent experiments in this area include the study of toroidal Alfvén eigenmodes ͑TAEs͒ driven by the alpha particles in plasmas with reduced magnetic shear in the central region, and measurements of the effects of sawteeth on the spatial and energy distributions of confined alpha particles.
Despite careful scrutiny, the early D-T plasmas in TFTR, which were predominantly in the supershot regime, showed no signs of any TAE instability attributable to the presence of the energetic fusion alpha particles, despite reaching central ␤ ␣ up to about 0.3%. Recently, the more comprehensive TAE theory, which was developing in parallel with and driven by these experiments, suggested that by modifying the q profile in the core of the plasma, it might be possible to destabilize the TAE in TFTR. 19,20 This would occur if the gap structure in the Alfvén continuum were more closely aligned to the region of the highest spatial gradient in the alpha-particle pressure. Thus, a search for TAE instability was recently undertaken in plasmas with increased central q, q 0 ϭ1.1-2.5 and reduced magnetic shear in the central region. As predicted by the theory, transient modes in the Alfvén frequency range, 150-250 kHz, with toroidal mode number n ϭ 2, 3, 4, were observed in D-T plasmas 0.1-0.3 s following the end of the NBI heating pulse. 21,22 Over this timescale following NBI, the alpha-particle population remains sufficiently energetic to drive the TAE, but the NBinjected ions, which damp the instability, have become ther- FIG. 7. In NBI-heated supershots and high-l i plasmas, there is a close relationship between the volume-average plasma energy density (͗W͘ ϰ ␤B 2 ) and the volume-average fusion power density for both DD (ϰ͗W͘ 1.8 ) and DT ( ϰ ͗W͘ 1.7 ) reactions. The data is for plasmas with I p у2 MA and volumes 31-46 m 3 . The D-T plasmas are restricted to those with nearly optimal D-T mixture. The arrowed lines indicate the reactivity ratio achievable under various constraints: ͑a͒ for constant ␤; ͑b͒ for constant NBI power, taking advantage of the isotope effect on global confinement; ͑c͒ at maximum supershot performance, taking advantage of the higher NBI power available with tritium NBI. malized. The TAE has been observed both on signals from Mirnov coils and on a microwave reflectometer signal from the region r/aϭ0.3-0.4 which coincides with the maximum "␤ ␣ . Typical results for a plasma with q 0 ϭ1.1-1.3 are shown in Fig. 8. The mode rises in frequency as the density at the mode location decays following the NBI pulse. For these plasmas, the TAE was observed when the peak fusion power exceeded 2.5 MW, corresponding to ␤ ␣ (0)Ͼ0.03% at the onset of the mode. These alpha-driven TAEs have not yet been sufficiently strong to cause detectable losses of the alpha particles.
Measurements have been made of the effects of the naturally occurring sawtooth oscillations on the spatial and energy distributions of the confined alpha particles in D-T plasmas. 23 Passing ͑nontrapped͒ alpha particles in the energy range 0.15-0.6 MeV are detected by charge-exchange recombination radiation spectrometry ͑Alpha-CHERS͒, 24 while trapped alphas in the energy range 0.5-3.8 MeV are detected as escaping neutral helium atoms following double charge-exchange of alphas with neutrals in a pellet ablation cloud ͑PCX͒. 25 A comparison of the radial profiles of the two classes of alpha particles before and after sawtooth crash is shown in Fig. 9. Calculations of the distributions following the crash using a magnetic reconnection model are also shown. 23 For the trapped particles, satisfactory agreement with the data can be obtained by including not only the magnetic effects but also that of the helical electric field induced by the reconnection. The substantial redistribution of alphas produced by the sawtooth may pose a problem for reactors designed to operate in regimes where large, albeit infrequent, sawteeth are expected.

VI. RF HEATING EXPERIMENTS IN D-T PLASMAS
Heating of D-T plasmas by waves in the ion-cyclotron range of frequencies ͑ICRF͒ is proposed for the International Thermonuclear Experimental Reactor ͑ITER͒ 26 as a means of reaching ignition. TFTR has been in a unique position to study the physics of various schemes for coupling ICRF power to D-T plasmas. Effective heating was previously reported using the second-harmonic tritium resonance, not only in D-T supershot plasmas, 27,28 where the presence of beam-injected tritons ensured good RF absorption, but also in Ohmically heated, gas-fueled target plasmas ͓n e (0) Ϸ5ϫ10 19 m Ϫ3 , T e (0) Ϸ 3 keV initially͔ prototypical of the startup phase of ITER. 29 The ICRF heating using the fundamental hydrogenminority coupling scheme in D and D-T plasmas provides a unique means to examine the scaling of electron transport with plasma isotopic composition because the heating is essentially independent of the majority-ion composition. Furthermore, the regime resembles that of alpha-particle heating in plasmas with T i ϷT e considered prototypical of ignited plasmas in ITER. For neutral beam heating, the situation is complicated by differences in the beam composition, ionization, and thermalization processes for D and T and the fact that the auxiliary power flows to the electrons predominantly through coupling with the hot (T i ϾT e ) thermalized ions. An experiment was conducted to compare the confinement of nominally D-only ͑80% D, 1% T, 8% H, 2%-3% C͒ and D-T ͑ϳ40% D, ϳ40% T, 5% H, 2%-3% C͒ plasmas fueled by gas puffing. 30 The ICRF power up to 4.4 MW at 43 MHz was applied for 1.2 s. The H-minority heating profile was calculated to be similar and the total stored energy in the energetic minority-ion tail, determined from the pressure anisotropy measured by the magnetic diagnostics, was the same for the D and D-T plasmas. Calculations showed negligible absorption of the ICRF power by either the secondharmonic D or the third-harmonic T resonance in either case. The global confinement time of the D-T plasmas was consistently higher than their D-only counterparts, consistent with a scaling of confinement time with average isotopic mass, A, E ϰ A y where yϭ0.3-0.5. This is illustrated in Fig. 10. While this scaling is roughly consistent with both previous results from TFTR using NBI heating in both supershot and L-mode regimes 31  A scheme for electron heating and current drive utilizing mode conversion of the ICRF fast wave to an ion Bernstein wave ͑IBW͒ in a mixed-species plasma has previously been demonstrated in TFTR. 29,33,34 In 3 He-4 He plasmas with the composition controlled by gas puffing, 35 electron heating on axis to 11 keV was observed with 4 MW of coupled ICRF power at 43 MHz. Using the same rf frequency and field in 3 He-D plasmas with a slightly higher 3 He fraction, the heating was localized off axis, r/aϷ0.15, and a hollow electron temperature profile was generated which persisted for up to 0.3 s, about twice the global energy confinement of the plasma. Currents up to 0.12 MA driven by the modeconverted IBW ͑the MCCD scheme͒ have been inferred by comparing plasmas with co-and counter-directed phasing of the launched waves; the driven currents were in good agreement with theoretical predictions.
Prior to the 1996 TFTR experiments, the generators driving two of the ICRF launchers were modified to operate at 30 MHz for mode-conversion studies in D-T plasmas. An experiment was conducted in which the tritium fraction of the plasma (n T /n e ) was varied from about 15% to 55% to scan the D-T ion-ion hybrid resonance layer across the central region. The ICRF power coupled directly to the electrons by the IBW remained unexpectedly low, in the range 10%-30%, rather than the 80%-90% expected for tritium fractions above about 30%. An explanation for this discrepancy may be found in the use of lithium injection in preceding experiments, both for confinement enhancement and to promote ERS transitions. As a result of this extensive use of lithium, even after some effort had been made to clean the limiter by running discharges with high-power H-minority ICRF heating, a small amount of lithium continued to be introduced at the edge from the limiter, resulting in a lithium concentration, estimated spectroscopically to be about n Li /n e Ϸ0.5%, in the core of the target plasmas for the modeconversion experiments. The natural lithium used for conditioning consists mainly ͑92%͒ of 7 Li, which has a charge-tomass ratio ͑0.43͒ between those of deuterium and tritium, with the result that it becomes an efficient minority-ion ab-sorber of the fast waves, thereby blocking the modeconversion process. For future experiments, it is planned to use isotopically enriched 6 Li for the conditioning process in TFTR since its charge-to-mass ratio coincides with that of deuterium and the intrinsic carbon impurity. It should be noted that 9 Be, the only stable isotope of beryllium, also has a charge-to-mass ratio between tritium and deuterium, which could make its presence in the first-wall materials a threat to similar ICRF heating schemes for D-T plasmas in ITER.
The difficulty of obtaining efficient IBW mode conversion in D-T plasmas during the last experimental run prevented a direct test of the physics basis for the alphachanneling scheme, 36 i.e., the process of coupling part of the energy of fusion alpha particles to the fuel ions through a series of wave-particle interactions, rather than through collisional processes that tend to heat electrons rather than ions. However, experiments were conducted to characterize the interaction of energetic ions with the IBW produced by mode conversion in D-3 He plasmas using the 43 MHz generators at a toroidal field of 4.4-5.3 T. Some of these energetic ions diffuse onto unconfined orbits, are lost from the plasma, and are detected by an array of four energy and pitch-angle resolving detectors near the vacuum vessel wall at poloidal angles of 20°, 45°, 60°, and 90°below the outboard midplane. 37 With these detectors, it has been possible to verify two features of the IBW interaction essential for alpha-channeling. First, by comparing the lost-ion signals during co-parallel NBI for different spectra of the ICRF waves, nominally co-and counter-parallel, we have confirmed the reversal of the parallel wave vector of the IBW with respect to the fast-wave spectrum. Such a reversal is required for the channeling interaction. Second, the interactions of beam-injected deuterons with the IBW have been found to approach the collisionless limit, i.e., the waveparticle coupling is strong enough for channeling to occur at reasonable ICRF power levels, about 3 MW in TFTR. On the basis of these results, simulations have been performed which show that in a D-T reversed-shear plasma in TFTR, cooling of a significant portion the alpha-particle population, mediated by the IBW interaction, could be expected and that, furthermore, a characteristic signature of the process would be observable in the lost-alpha distribution. 38

VII. SUMMARY AND PLANS
In the past year, substantial progress has been made in developing two newly discovered advanced operational regimes in TFTR. The high-l i regime has already produced DT fusion power of 8.7 MW at lower current and toroidal magnetic field than supershots producing comparable power. This technique for increasing the stability of the plasma, utilizing expansion of an ultra-low-q Ohmic plasma prior to neutral beam heating, has already been extended to higher currents and awaits the application of new techniques for wall conditioning and for handling the power load to the limiter to achieve higher fusion power and, therefore, selfheating of the plasma by the fusion alpha particles. In the reversed-shear regime, progress has been made in developing plasmas at higher current. The internal transport barriers characteristic of the ERS plasmas have been produced at higher current in deuterium plasmas, but only by using lithium pellet injection to stimulate the transition; this has resulted in significant lithium contamination of the wellconfined plasma core which reduced the fusion reactivity significantly. Although both the radius of the surface of minimum q and the radius of the transport barrier have been increased, the ␤-limit of the high-current ERS plasmas has not increased significantly, apparently because the transport barrier and q profile evolve in a way which decreases stability through the ERS phase. This points out the necessity of developing tools to control transport barriers if we are to make use of them in advanced tokamak designs. In this regard, progress has been made in understanding the origin of the reduced transport in the TFTR ERS plasmas through the stabilization of microturbulence by sheared plasma flow, as discussed by Synakowski et al. 9 In view of the lithium dilution and stability issues encountered at high current, D-T ERS plasmas were only investigated at lower current. In these experiments, there were clear indications that the NBI power threshold for the ERS transition was higher in D-T than in D plasmas.
In weak-shear plasmas with q 0 Ͼ1, a TAE instability driven by the fusion alpha particles has been observed for the first time. The observation of this mode, which was predicted theoretically to occur in specific plasma conditions, provides strong confirmation for the validity of TAE theory which has been advanced significantly since the start of D-T operation on TFTR. The observed redistribution of alpha particles by sawteeth and its theoretical explanation provides important data for the design of ITER.
Following the 1996 experiments, the vacuum vessel of TFTR was opened, for the first time in three years of intensive D-T operation, to install new ICRF antennas and to upgrade some diagnostic capabilities, particularly the MSE system. With one of the new ICRF antennas, which has been installed in the IBW polarization (E RF ʈ B), it is intended to produce controllable transport barriers in TFTR similar to those achieved in Princeton Beta Experiment-Modified ͑PBX-M͒ 39 in the so-called ''CH'' mode. The other two new ICRF antennas will have four, rather than two, conductor straps which will improve the k-spectrum of the waves launched into the plasma. This will provide better control and localization of the driven current in the MCCD scheme. With these modifications, it is hoped to extend the performance of the ERS regime in particular, both by increasing the ␤-limit and by avoiding the contamination of the wellconfined plasma core that occurs as a result of the lithium presently injected to stimulate formation of the transport barrier. This would open the door to more extensive studies of D-T and alpha-particle physics in this regime.

ACKNOWLEDGMENTS
In undertaking these experiments, we have depended on the skill and hard work of the entire staff of the TFTR Project. We thank them for their dedication and their unstinting efforts. We thank Dr. R. C. Davidson for his support and encouragement.
This work is supported by U.S. Department of Energy Contract No. DE-AC02-76-CH03073.