Conditioning and High Power Operation of the Lower Hybrid Current Drive Launcher in JET

Transcription

1 JET P(97)19 Conditioning and High Power Operation of the Lower Hybrid Current Drive Launcher in JET A Ekedahl 1 J A Dobbing P Finburg P Froissard C Gormezano M Lennholm F Rimini P Schild F X Söldner. JET Joint Undertaking Abingdon Oxfordshire OX14 3EA UK. 1 Institute for Electromagnetic Field Theory and Plasma Physics Chalmers University of Technology S Göteborg Sweden. CEA Cadarache F 1318 Saint Paul lez Durance France. June 1997

2 This document is intended for publication in the open literature. It is made available on the understanding that it may not be further circulated and extracts may not be published prior to publication of the original without the consent of the Publications Officer JET Joint Undertaking Abingdon Oxon OX14 3EA UK. Enquiries about Copyright and reproduction should be addressed to the Publications Officer JET Joint Undertaking Abingdon Oxon OX14 3EA.

3 ABSTRACT The full lower hybrid current drive (LHCD) system on the Joint European Torus (JET) has been operational since the start of the pumped divertor phase of JET in By a campaign of vacuum conditioning in the beginning of operation the power transmission and conditioning of the full launcher was speeded up as compared with the prototype launcher used in 199/91. After vents of the tokamak the conditioning can be recovered by ~3 vacuum pulses over 3-4 days. Even after extensive conditioning and with efficient pumping the power handling of the full LHCD system on plasma is limited by the maximum electric field in the waveguides the limit being 4-5 kv/m. 1. INTRODUCTION Radio frequency (RF) techniques are widely used for heating and current drive in fusion research. Lower hybrid current drive (LHCD) which utilises waves in the frequency range 1-1 GHz has proven to be the most efficient of the various methods of non-inductive current drive in tokamaks (lower hybrid waves fast waves electron cyclotron waves neutral beam injection etc.). The main goal of the LHCD system on the Joint European Torus (JET) as well as in other tokamaks is to control the plasma current profile in order to suppress magnetohydrodynamic instabilities and to explore scenarios for steady state tokamak operation with improved confinement. For example minute long LH pulses at P LH =.4 MW and full non-inductive current drive at.8 MA for more than 1 minute have been achieved in Tore Supra [1]. In JET 13 s long LH pulses with up to 6 MW have been carried out and 3 MA full non-inductive current drive and suppression of sawteeth have been achieved with ~6 MW []. These scenarios imply operation of the LHCD systems at high power during long pulses which requires conditioning of the vacuum components in order that they can sustain the electric fields involved. This paper describes the conditioning and high power operation of the complete LHCD launcher installed in JET in 1994 and the results obtained. The results are also compared with those obtained with the prototype launcher which was used in the 199/91 experimental campaign in JET.. DESCRIPTION OF THE JET LHCD SYSTEM The LHCD launcher in JET [3 4] is composed of arrays of phased waveguides with internal power splitting so-called multijunctions. A similar design is used in high power LHCD 1

4 experiments in other tokamaks e. g. Tore Supra [5 6] and JT6-U [7]. The full LHCD system in JET (Fig. 1) which was brought into operation in early 1994 operates at 3.7 GHz and has a generator capability of 1 MW for s supplied by 4 klystrons. The power is transmitted through 4 copper waveguides 4 m long with dimensions 38.6 mm 77. mm to the torus hall Fig. 1: Layout of the transmission lines inside the torus hall launcher and cryopump of the JET LHCD system. where hybrid junctions split each transmission line in two. The 48 lines are then connected to the multijunctions while the fourth ports of the hybrid junctions connect to high power dummy loads in the torus hall where most of the reflected power is absorbed. The klystrons are protected by circulators which absorb any power reflected back to the klystrons. The transmission lines are filled with bar absolute SF 6 in order to prevent arcing. The launcher is surrounded by a vacuum tank connected to the torus. 48 double beryllium oxide windows separate the transmission lines from the torus vacuum (~1-7 mbar). The part under vacuum consists of 48 perforated waveguides for pumping. These are connected to 48 multijunctions made of copper coated stainless steel. A 1 l/s cryopump has been installed on top of the tank above the perforated waveguide section. The gas released from the waveguide walls during pulses is pumped through the perforations. The pressure is measured by a penning gauge located at the bottom of the tank below the cryopump.

5 In each multijunction the incoming waveguide is first split vertically into two superposed lines in an H-plane junction and then further divided horizontally into four narrow waveguides by E- plane junctions in both levels. At the front end of the multijunction there are eight 9 mm 7 mm waveguides (Fig. ). Phase shifters are incorporated in the waveguides in such a way as to Fig. : Side top and front views of a multijunction of the LHCD launcher. give a π/ π 3π/ phase distribution of the forward power at the grill mouth. The multijunctions have silicon carbide coated graphite loads at the fourth port of the hybrid junctions which absorb part of the reflected power from the grill mouth and reduce the circulating power in the multijunctions. The multijunctions are mounted in six rows and eight columns and each klystron feeds two multijunctions placed one above the other. At the grill mouth there are twelve toroidal rows with 3 narrow waveguides in each row. The launcher is not actively cooled only inertially. The grill radiates slow waves in the direction of the electron current with parallel refractive 3

6 index N // >1 and electric field nearly parallel to the plasma magnetic field. The peak in the N // - spectrum can be varied between 1.4 and.3 by adjusting the phase of the incident wave between klystrons feeding adjacent multijunctions in a horizontal row. Normally the standard π/-phasing between waveguides is used which gives a spectrum of N // peaked at 1.84 with full width between zero points of N // =.46. The front of the grill is surrounded by a frame of graphite tiles which protects the grill from excessive heat deposition by plasma contact and impact of fast particles. The frame protrudes approximately mm in front of the waveguide openings. The protrusion must be kept small in order to avoid interaction of the slow wave with the protection frame while still protecting the grill from the heat load. Both the front of the grill and the protection frame have been designed to match the shape of the X-point plasmas in the divertor phase of JET. The front of the LHCD launcher with is protection frame is shown in the photograph in Fig. 3. Fig. 3: View of the LHCD launcher inside the JET tokamak. The divertor and ion cyclotron resonance heating (ICRH) antennas can also be seen. The coupling of the LH waves can be controlled during plasma pulses by a feedback loop on the reflection coefficient [8]. The radial position of the launcher is adjusted in real time to maintain 4

7 the desired value of the power reflection coefficient usually 3%. The launcher which weighs approximately 15 tonnes is moved by hydraulic actuators [9]. The hydraulic system controls the launcher movement to a total stroke of 1 mm and a maximum allowed speed of 5 mm/s. Before the complete launcher was brought into operation a prototype launcher with one third of the available power was used in JET in the experiments in 199/91 before the pumped divertor was installed. Some results of these experiments are summarised in the next section. 3. OPERATIONAL EXPERIENCE WITH THE L LAUNCHER The first LHCD experiments in JET started in 199 with a prototype launcher (L) [1]. It consisted of eight copper coated stainless steel multijunctions of the type that was later used in the full-sized launcher in JET and eight copper zirconium (CuZr) multijunctions of the type used in Tore Supra [11]. All multijunctions had an anti-multipactor coating of carbon. Eight klystrons with power capability of 6 kw powered the 16 multijunctions. The grill mouth was composed of eight toroidal rows with 16 narrow waveguides (9 mm) per row. After the first campaign of five months 3 kw per klystron could be transmitted to the plasma. The power handling of both types of multijunctions was essentially the same. The conditioning status was kept after a seven months long shutdown and when the second phase of operation began in June 1991 the previously achieved power level (3 kw per klystron) could be reached rapidly. Two air leaks at 35 C occurred during the second campaign. After these incidents the conditioning of the CuZr multijunctions could not be recovered and the power level at which clean pulses without breakdowns could be achieved remained below 3 kw per klystron. The stainless steel multijunctions on the other hand could recover their conditioning and continued the slow but steady increase in power transmission. 6kW klystron power was reached at the end of operation with L with 17 pulses on plasma and a total of 14 months operation. The power transmission capability would usually degrade after a full day of pulsing. This could be attributed to the increase in the waveguide temperature [1] (the launcher is inertially cooled only) and the associated outgassing from the waveguide surfaces which increases the risk of multipactor breakdowns [13]. Further in some plasma pulses a sudden decrease in reflection coefficient followed by an arc could be observed. This indicated a formation of a plasma in front of the grill mouth. A similar situation might occur during vacuum pulses if the outgassing from the waveguide surfaces is high and the gas is ionised by the electric field. Following these results the automatic grill conditioning system for the complete launcher (L1) was designed. It was also found that the carbon coating flaked off and it was therefore decided not to apply carbon coating inside the multijunctions on L1. 5

8 4. DESCRIPTION OF CONDITIONING METHOD FOR L1 Before installation on JET the complete launcher (L1) was installed in a vacuum tank baked up to 35 o C for 6 months and tested to 1 kw per multijunction for s. Once installed on JET a large part of the initial phase of operation was aimed at conditioning of the multijunctions by pulsing them in vacuum in repetitive pulses. The purpose was to increase the temperature of the multijunctions above the normal operating temperature and to force outgassing from the waveguide surfaces before the plasma operation started thus avoiding breakdowns in the multijunctions due to multipactoring during the experiments. When power is launched from the multijunctions into vacuum it is equivalent to placing a short circuit in front of the open ended waveguide and total reflection will occur. For the multijunctions at JET the reflection coefficient measured back in the transmission line is approximately 8%. From the temperature increase of the vacuum loads it is estimated that the power absorbed in the loads account for 1-3% of the power reflected back in the transmission lines. When a plasma exists in the vacuum vessel it acts as a load with a loading resistance dependent on the electron density in front of the waveguide opening. It is necessary to minimise the reflection coefficient and avoid large voltage standing wave ratio (VSWR) in the waveguides during plasma operation. On the other hand the large electric field produced when pulsing in vacuum gives an efficient method of conditioning the waveguides. Another advantage is that vacuum pulses can be carried out at high rate approximately two pulses per minute between plasma pulses or during commissioning periods. If one should rely on plasma pulses as the only method of conditioning the rate would be reduced to two pulses per hour. The expected power loss in the copper coated stainless steel multijunctions is calculated as follows. For power P incident in a waveguide the lost power is P L = P ( 1 e αd )= P αd where α is the attenuation constant and d the length of the waveguide. The energy P L τgives an increase in the waveguide temperature which equals T = P Lτ ρadc P τα ρac where τ is the pulse length A the cross sectional area of the waveguide ρ the density and C the specific heat of the waveguide material. For stainless steel at 5 C ρ = 79 kg/m 3 and C = 54 J/(kg K). The attenuation constant α for the TE 1 -mode is given by [14] 6

9 α = 1 + a c f a 1cα. b bf c 1 bf where f = 3.7 GHz c = m/s and a and b are the width and height of the waveguide respectively. The thickness of the copper coating in the multijunctions is 15 mm and the skin depth for copper at 3.7 GHz is 1.1 mm. The stainless steel below the copper surface causes a decrease in the electrical conductivity by approximately % of that for pure copper so that σ (Ωm) -1 at 5 C. When calculating the power in the multijunction the reflection coefficient is taken into account as follows. If R is the power reflection coefficient measured back from the multijunction in the transmission line then the power in the two-waveguide section of the multijunction is P o (1 + R). In the eight-waveguide section the wave is reflected twice due to the phase shifters which means that the reflection coefficient at the grill mouth is R and the total power is P o (1+ R). The calculated attenuation constants and the resulting temperature increase for the two-waveguide and eight-waveguide sections of the multijunctions (see Fig. ) are given in Table 1 and the estimated temperature increase of the waveguides for pulses with P = 1 kw and τ = 3. s both with R = 1% (in vacuum) and R = 3% (on plasma) are given in Table. TABLE 1 Section a b (mm mm) α (mm -1 ) A (mm ) T ( C) -waveguide τp (1+R) 8-waveguide P τ(1+ R) TABLE T ( C) R = 1. (vacuum) R =.3 (plasma) -waveguide.4. 8-waveguide.3.8 7

10 The grill conditioning system that was designed for L1 is specified as follows. The six klystron modules are pulsed in sequence in vacuum in repetitive cycles. The pulse length can be varied but is usually kept at 3. s. The protection system detects arcs at the vacuum windows with optical arc detectors VSWR at the klystron 1.8 and unbalanced reflection in the two waveguides of the same klystron. The protection system also detects a decrease in reflection coefficient which is an indication of a plasma formation inside or in front of a waveguide. When any of these events are detected the generator power is switched off for a pre-set period usually 1 ms then re-applied. Depending on the number of trips at each generator the klystron power is either increased or decreased by 5 kw for the following cycle. The klystrons are pulsed up to kw in this mode i.e. 1 kw per multijunction. 5. RESULTS OF OPERATION WITH THE L1 LAUNCHER 5.1 Vacuum conditioning After installation of the L1 launcher on JET the first three months of operation was mainly devoted to vacuum conditoning as described in the previous section. During the first weeks before the cryopump had become operational significant outgassing was measured on the mass spectrometer in the torus during vacuum conditioning. The pumping speed in the launcher was only 3 l/s arising from the conductance between the torus and the launcher. Fig. 4 shows.8 P torus = 1 x1 6 mbar T waveguides o C.6 Without cryopump Pressure (1 6 mbar).4. Conditioning pulses Time (Minutes) JG97.3/3c Fig. 4: Gas release seen on the mass spectrometer in the torus during the second day of vacuum conditioning after the start of operation in April

11 the outgassing of mass (H ) 18 (H O) 8 (CO N ) and 44 (CO ) during the conditioning pulses in the beginning of the campaign. Two months later when the LHCD cryopump was in use the increase in partial pressures had reduced significantly see Fig. 5. The increase in launcher pressure during the vacuum pulses had reduced by two orders of magnitude while the pumping speed had increased by a factor of ~3 i.e. from 3 l/s to 1 l/s. Improvement in power transmission both in vacuum and on plasma was observed during this period. After approximately 5 pulses on plasma 5 kw klystron power was reached in one klystron module and 4 kw on several other klystrons. For comparison maximum 3 kw klystron power was reached on L after approximately the same length of operation and similar number of pulses on plasma. This showed that the baking of the launcher by vacuum pulsing did speed up the conditioning of the waveguides. The 1994/95 experimental campaign was interrupted by several vents of the vacuum vessel. After these vents the conditioning could be recovered rapidly by RF conditioning in vacuum. By pulsing the klystrons repetitively in vacuum for a total number of ~3 pulses on each klystron over a period of three to four days the temperature of the waveguides increased by 5-8 C measured at the two-waveguide section of the multijunctions. Fig. 6 shows the pressure increase in the launcher during vacuum pulses at 7 kw / 5 s per klystron after the third vent before and after conditioning had been carried out..5.4 P torus = 3 x1 7 mbar T waveguides 5 o C With cryopump LHe Pressure (1 6 mbar) JG97.3/4c Time (Minutes) Fig. 5: Gas release in the torus during vacuum conditioning with cryopump two months after the start of operation. 9

12 1..8 With cryopump LHe JG97.3/1c Before conditioning After conditioning Launcher pressure (1 6 mbar).6.4. LH Time (s) Fig. 6: Increase in neutral pressure in the launcher during vacuum pulses after a vent of the tokamak before and after conditioning. Although vacuum conditioning is efficient for outgassing of the waveguides the electric field distribution in the multijunctions gives a local conditioning. On plasma the phase of the reflected wave and the standing wave pattern varies which means that the parts of the waveguides not conditioned during vacuum pulses will be conditioned during plasma pulses. It was therefore found important to complete the vacuum conditioning after a torus vent with approximately ten high power pulses (P LH > 4 MW) on plasma before satisfactory conditioning and reliable power transmission to the plasma was reached. dt/dt ( o C/hour) 16 Row 1 Row 1 Row Days after start of operation dp/dt (mbar/hour) JG97.3/c No cryo LN only LHe Days after start of operation JG97.3/1c Fig. 7: (a Left) Average rate of temperature increase in the three double rows of multijunctions during conditioning. (b Right) )Rate of pressure increase in the launcher. The dashed areas indicate periods of torus vents. The conditioning was recovered rapidly after the vents. 1

13 The conditioning effect both at the beginning of operation of the L1 launcher and after the vents of the vacuum vessel is demonstrated in Fig. 7. Fig. 7(a) shows the average rate of temperature increase of the three double rows of waveguides during vacuum conditioning. This can be used as an indication of the increase in power which could be transmitted in vacuum. Fig. 7(b) shows the rate of pressure increase in the launcher tank during conditioning averaged over approximately 3 minutes. This demonstrates the evolution in gas release from the waveguide walls as the conditioning of the multijunctions improve. The rate of pressure increase dropped by two orders of magnitude during the first three months while the pumping speed increased by a factor of ~3 i.e. from 3 l/s (no cryopump) to 1 l/s (cryopump at liquid helium temperature). 5. Plasma operation High power LH current drive and profile control experiments have been carried out in the divertor phase on JET which started in Full current drive up to 3 MA and suppression of sawteeth have been obtained in plasmas with line average electron density in the range n e = m -3 []. High power pulses with P LH 5 MW and 1 s duration have been delivered to the plasma and the maximum coupled energy has reached 68 MJ (Fig. 8). The neutral gas pressure in the launcher is kept well below mbar during the long pulses by use of the cryopump. According to Fig. 8 the outgassing rate in the second pulse is.5 mbar l/s and.1 mbarl/s/mw. The maximum coupled LH power has reached 7.3 MW using 8. MW of generator power. High power (P LH > 6 MW) can only be coupled when the reflection coefficient is low ( 3%) and the feedback control on the reflection coefficient is therefore used routinely in the experiments (Fig. 9) [8]. (MW) (1 6 mbar) (MW) (1 6 mbar) Pulse No: 353 P LH Neutral pressure Pulse No: 3539 P LH Neutral pressure Time (s) P NBI P ICRH P ICRH JG97.3/5c Fig. 8: Maximum coupled LH energy during plasma pulses and evolution in neutral gas pressure. 11

14 (MW) (1 6 mbar) (%) Pulse No: P Generator P Rad Neutral pressure Reflection coefficient <R> P Coupled (cm) 4 LCFS Launcher Time (s) JG97.3/7c Fig. 9: Maximum coupled LH power. The position control system adjusts the launcher position to maintain the requested reflection coefficient (.3%). Pulse No: 349 Before conditioning Pulse No: After conditioning (cm) (%) (1 6 mbar) (MW) P LH Neutral pressure (Cryo on) <R> LCFS Launcher P LH Neutral pressure (Cryo off) <R> LCFS Launcher Time (s) Time (s) JG97.3/6c Fig. 1: LH pulses in similar plasma configurations before conditioning (#349) and after conditioning (#34414). 1

15 An example of the improvement in power transmission on plasma by vacuum conditioning is shown in Fig. 1. It shows two LH pulses on plasma at similar power level and in similar plasma configurations; the first pulse (#349) shortly after a torus closure before any conditioning had been carried out and the second pulse (#34414) five days later after the waveguides had been conditioned in vacuum. In #349 the applied power could not be sustained without trips while in the #34414 a clean pulse with only a small increase in neutral pressure was obtained. Fig. 11 Temperature ( o C) Temperature ( o C) Pulse No: 349 Pulse No: 349 Lower Upper Lower A3 Waveguides Pulse No: A3 Vacuum loads Pulse No: JG97.3/11c Upper Time (Hours) Fig. 11: Evolution in temperature of two multijunctions measured at the two-waveguide section and two vacuum loads on L1 during one week of conditioning and plasma operation. The pulses shown in Fig. 1 are indicated in the temperature trends. shows the evolution in temperature of two multijunctions measured at the two-waveguide section and two vacuum loads during vacuum conditioning and plasma operation that week. Plasma operation was carried out in the morning of the first day and during the last two days of the week. The vacuum conditioning can be distinguished by the large rate of change in temperature. The plasma pulses shown in Fig. 1 are indicated in the temperature trends. Further by increasing the temperature of the waveguides by RF conditioning and force outgassing from the waveguide surfaces before an experiment and the leave them to cool down again the degradation in power transmission with increasing temperature as observed on L could be avoided. After a quite rapid increase in power transmitted to plasmas during the first three months of operation a saturation seemed to occur. More than 5 kw has not been reached in any klystron 13

16 which is less than for the prototype launcher (L). Fig. 1 shows a histogram of the output power from a representative klystron to L1. The graph shows the conditioning effect at the beginning of the campaign. Similar trend is seen in Fig. 7 with the exception that in Fig. 7(b) there is a tendency of improved conditioning with time measured over the whole period of 14 months. 6 Klystron D1 5 Vents Klyston power (kw) Pulse Number JG97.3/8c Fig. 1: Power generated from a representative klystron to the L1 launcher during plasma pulses. One difference between L and L1 seems to be the reflection coefficient linked to the electric field strength which the multijunctions could sustain without trips. On L the reflection coefficient for the bottom part of the launcher (with stainless steel multijunctions) was usually 5-1% but more than 5 kw klystron power was reached at this reflection coefficient. This corresponds to a maximum electric field strength reached of ~7 kv/m. On L1 a reflection coefficient of 3% was necessary in order to transmit more than 4 kw which corresponds to ~5 kv/m. Fig. 13 shows the klystron power for the same klystron as in Fig. 1 plotted against the reflection coefficient. The curve of a constant electric field strength of 53 kv/m is drawn in the graph. Analysis of the power transmission capability for all multijunctions on L1 shows that the electric field at which reliable power transmission without recurring trips can be obtained is approximately 4 kv/m. The electric field limit on L1 may possibly be linked to multipactoring in the multijunctions. As mentioned above the multijunctions on L1 are not carbon coated. Another difference between the complete launcher and the prototype is the number of waveguides in the toroidal direction at the grill mouth (3 compared with 16). According to Fuchs et al. [15] acceleration of edge electrons up to a few kev can take place in the near field of the waveguide array. The final energy of these initially rather cold electrons (~5 ev) increases with number of waveguides in the toroidal direction. This effect is thought to be responsible for hot 14

17 spots and impurity production reported in some experiments [16]. On some occasions in JET influx of metallic impurities have been observed when the launcher was approached very close to the plasma (< cm) in discharges with > m -3 and when the applied power was above 5 kw per klystron (~45 kv/m). Increase in radiation could be seen on spectroscopic measurements of radiation from iron and copper and on bolometry measurements of the radiated power. The bursts of radiation were correlated with breakdowns in the upper rows of multijunctions 6 Klystron D1 5 E = 53 kv/m Klyston power (kw) Reflection coefficient (%) Fig. 13: Klystron power versus reflection coefficient for a representative klystron indicating that arcs occurred in front of the grill mouth on the upper part of the launcher which in the divertor configuration in JET is the nearest to the plasma. In-vessel inspection of the grill mouth has revealed melting on the front end of the waveguides mainly at the midplane of the multijunctions which corresponds to the location of the maximum electric field. The top part of the launcher was more damaged than the bottom part which agrees with the impurity behaviour observed during plasma operation. In recent experiments this potential problem was avoided by keeping the launcher behind the poloidal limiters and providing good coupling conditions by injecting gas in the tokamak vessel from a gas feed near the launcher [17]. This method was first developed in ASDEX [18] and has also been explored in other tokamaks in order to improve the coupling of the LH waves. JG97.3/9c 6. SUMMARY The automatic grill conditioning technique employed for L1 did speed up the conditioning of the L1 launcher as compared with L. 5 kw was reached in one klystron module and 4 kw 15

18 in several other klystrons after three months operation and approximately 5 pulses on plasma. On L 3 kw was reached after similar length of time and number of pulses on plasma. After venting the tokamak vessel the conditioning on L1 was recovered after ~3 vacuum pulses which lead to an increase in temperature by 5-8 C (measured at the two-waveguide section) plus approximately ten plasma pulses. By increasing the temperature of the waveguides above the normal operating temperature (-5 C) before an experiment and then leaving them to cool down again the degradation in power transmission with increasing temperature could be avoided. The LHCD cryopump is used during vacuum conditioning to pump the gas released during the LH pulses. During plasma operation the neutral gas pressure in the launcher is maintained well below mbar during long pulses with the cryopump. Although the conditioning showed promising results initially the end result after 14 months of operation with L1 was less successful than with L. An increase in transmitted power up to 4-5 kw klystron power was obtained during the first three months then a saturation was reached. On the stainless steel multijunctions on L 6 kw klystron power was reached at the end of the campaign after a total of 14 months operation. The electric field at which reliable power transmission can be achieved is lower on L1 than on L. A reliable electric field limit for the L1 launcher is 4-5 kv/m. With a reflection coefficient of 3% this corresponds to a power density at the grill mouth of approximately -3 MW/ m which is the range considered for ITER. The limit in electric field can possibly be linked to effects like multipactoring in the multijunctions or acceleration of electrons in the near field of the grill. The latter effect will limit the number of waveguides in the toroidal direction of a launcher (compare L1 and L in JET i.e. 3 and 16 waveguides respectively). In a large antenna separators would need to be inserted between sections of waveguides in order to limit the consecutive number of waveguides in a toroidal row. REFERENCES 1. Y. Peysson G. Rey R. Arslanbekov et al. Proc. 16th IAEA Fusion Energy Conference Montreal Canada 1996 (International Atomic Energy Agency Vienna) IAEA-CN-64/E- 4.. F. G. Rimini B. Alper Y. Baranov et al. Proc. 11th Topical Conf. on Radio Frequency Power in Plasmas Palm Springs CA USA 1995 (American Institute of Physics New York) p M. Pain H. Brinkschulte G. Bosia M. Brusati J. A. Dobbing et al. Proc. 13th Symposium on Fusion Engineering Knoxville Tennessee USA

19 4. F. X. Söldner M. Brusati A. Ekedahl P. Froissard C. Gormezano et al. Proc. IAEA Meeting on RF Launchers for Plasma Heating and Current Drive Naka Japan 1993 JAERI-Conf 94-1 p X. Litaudon G. Berger-By P. Bibet J. P. Bizarro et al. Nucl. Fusion (199). 6. P. Froissard P. Bibet G. Agarici S. Berio C. Deck et al. Proc. 19th Symposium on Fusion Technology Lisbon Portugal 1996 (Elsevier Science Amsterdam). 7. Y. Ikeda O. Naito M. Seki T. Kondoh S. Ide et al. Fusion Engineering and Design 4 87 (1994). 8. M. Lennholm Y. Baranov J. A. Dobbing A. Ekedahl P. Finburg et al. Proc. 16th Symposium on Fusion Engineering Urbana-Champaign Illinois USA C. I. Walker C. Gormezano A. Kaye M. Lennholm P. Paling R. Price and P. Schild Proc. 17th Symposium on Fusion Technology Rome Italy 199 (Elsevier Science Amsterdam) Vol. 1 p J. A. Dobbing G. Bosia H. Brinkschulte M. Brusati A. Ekedahl et al. Proc. 17th Symposium on Fusion Technology Rome Italy 199 (Elsevier Science Amsterdam) Vol. 1 p G. Rey R. Aymar G. Berger-By P. Bibet M. Goniche et al. Proc. 15th Symposium on Fusion Technology Utrecht The Netherlands 1988 (Elsevier Science Amsterdam) Vol. 1 p A. Ekedahl H. Brinkschulte M. Brusati J. A. Dobbing C. Gormezano et al. Proc. Europhysics Topical Conf. on Radio Frequency Heating and Current Drive of Fusion Devices Brussels Belgium 199 (European Physical Society Geneva) Vol. 16E p A. S. Kaye Plasma Phys. Control. Fusion 35 Suppl. A A71 (1993). 14. P. Lorrain and D. Corson Electromagnetic Fields and Waves nd Ed. (Freeman and Co. San Francisco 197) Chapt V. Fuchs M. Goniche Y. Demers P. Jacquet and J. Mailloux Physics of Plasmas 3 43 (1996). 16. M. Goniche J. Mailloux Y. Demers D. Guilhem J. H. Harris J. T. Hogan P. Jacquet et al. Proc. 3rd EPS Conference on Controlled Fusion and Plasma Physics Kiev Ukraine 1996 (European Physical Society Geneva) Vol. C Part II p A. Ekedahl Y. Baranov J. A. Dobbing B. Fischer M. Goniche et al. Proc. 1th Topical Conf. on Radio Frequency Power in Plasmas Savannah GA USA 1997 (American Institute of Physics New York). 18. F. Leuterer F. Söldner M. Brambilla M. Münich F. Monaco M. Zouhar et al. Plasma Phys. Control. Fusion (1991). 17