Status Report on the University of Washington Superconducting Booster Accelerator Project

Transcription

1 Status Report on the University of Washington Superconducting Booster Accelerator Project Derek W. Storm. D.T. Corcoran, M.A. Howe, Q.-X. Lin, and D.P. Rosenzweig Nuclear Physics Laboratory University of Washington Seattle, WA U.S.A. I. Introduction The University of Washington Superconducting Booster was proposed in 1982 and funded at the very end of The design goals were to be able to accelerate particles from protons to middle weight ions using our 9-MV FN tandem as an injector. We wanted to be able to double the proton energy and increase the heavy ion energy by a larger factor, so that the resulting beams would be at least equal in energy to those that would be obtained from an 18-MV tandem. (An earlier proposal to put an 18-MV tandem at our laboratory was very attractive, but in the final analysis appeared too expensive.) We found that the LINAC could be added to the existing tandem for something like half the cost of the new tandem and its associated tower. One special advantage for us was that the LINAC could be put into existing building space. Because we planned to accelerate lighter ions than other laboratories, we conceptualized the accelerator around "low-beta" resonators of P near 0. l and high-beta resonators of some value from 0.16 to 0.2. At the time, Ben-Zvi and Brennan had recently built a successful quarter-wave P = 0.09 prototype using lead plated copper, and the ANL P = niobium split loop resonators were in production. Shepard was progressing with development of a niobium split-loop P = 0.16 resonator, but the committment to build those resonators had not yet been made at Argonne. The wide transit-time factor curve of the quarter-wave resonator was attractive because of our desires to accelerate a wide mass range. It was also clear that the problems in phase locking resulting from mechanical vibrations were substantially less with the quarter-wave resonators than with split loops. We found that if we built the entire linac of P = 0.1 two-gap resonators, we would be able to achieve our energy goals. Thus

2 if we were unable to build a high-beta quarter-wave resonator, we would have a fallback. On the other hand, if we chose split-loop resonators, we would have serious problems if P = 0.1 was the highest available. The operating fields available from lead or niobium resonators were not very different, although lead resonators required more power than similar niobium ones. For the reasons presented above, we chose to base our linac on lead-plated copper quarter-wave resonators; we do not feel that we sacrificed performance in doing so, and to the contrary we gained a number of advantages. So far we have been happy with that choice. Although niobium is a better superconductor than lead, it is useful to compare the energy gain per unit charge per watt actually obtained with lead or with niobium resonators. Then one should balance the difference between the cost of niobium and lead plated copper resonators resonators against the cost of refrigeration. If we assume 3 MV/m average accelerating fields (appropriate for our recent operating experience as well as that quoted for Argonne at a similar point in their operating history), and a refrigerator capital cost of $1.3k/W (our 480 W refrigerator cost $623k in 1984), we get the following: UW P= 0.10 lead-plated copper quarter-wave ANL P=0.105 niobium split loop acceleration length 18 cm 35 cm cost $12 k $40 k cost/mv $22 k/mv $38 k/mv power 8 W 4 W power/mv 15 W/MV 4 W/MV cooling cost/mv $19 k/mv $ 5 k/mv combined cost/mv $4 1 k/mv $43 k/mv I have not included refrigerator operating costs, which would favor the niobium resonators. On the other hand, refrigerator capital costs will scale with power more slowly than linearly. In the following text we will describe the accelerator itself briefly and describe the resonators and rf control system in more detail. We will conclude with a report on our operating experience to date, which includes accelerating an oxygen beam and transporting it to a scattering chamber where we were able to use elastic scattering from a gold foil to study some of its properties.

3 11. Description of the LINAC The booster accelerator consists of 24,6 = 0.10 resonators in six cryostats and 12,6 = 0.21 resonators in another six similar cryostats. Between each cryostat there is a quadrupole doublet. There is a P= 0.10 buncher and a = 0.21 debuncher each in its own cryostat. Beam transport, diagnostics, cryogenics, and so on have been discussed previ~usly.~ A plan view of the accelerator is shown in Fig. 1. f UNAC BOOSTER AND BEAM HANDLING EQUPMENT I0 20 WFEn =ME L...,... e _ Y! Fig. 1. A plan view of the booster accelerator. The beam is bent through 180 degrees between the low-beta and high-beta halves of the accelerator, in order to utilize the space that was available in the Tandem vault. The resonators operate at 150 MHz (actually ). With an average accelerating field of 3.0 MV/m in each resonator, we would double the energy of the tandem proton beam and increase energies by about a factor of four for tandem beams of ions up to about mass 50. A graph describing the ion energies is shown in Fig. 2.

4 E/A with 3.0 MV/m or Tandems I I I I I I l l I I I I I I l l ' - X 3.0 M/m LINE W Tandem X M Tandem - X X Oko utri~p-1 X + X + X + 0 X X m - 0 X P d t He C 0 Si Ca Fe Zr - I I I I I I I I l'& Fig. 2. Energy per unit mass for ions from the 9-MV tandem alone and from our LINAC injected by the tandem, assuming 3-MV/m fields in the LINAC and assuming most probable charge states following stripping in the tandem and between the tandem and LINAC. Energies for an 18-MV tandem with two strippers is also shown Construction of Resonators The high-beta resonators are double the diameter of the lowbeta ones. This is convenient, since we can put either two highbeta or four low-beta resonators in a cryostat. Drawings of the resonators are presented in Fig. 3. The low-beta resonators are similar to a design by Ben-Zvi and Brennan, except that we increased the inner radius of the outer conductor from 8 to 9 cm. According to some calculations by ~ e n - ~ v(done i ~ after the design was fixed), this change in radius should significantly reduce multipactoring. The design of the high-beta resonator was

5 discussed in some detail in a paper published in the proceedings of the workshop previous to this The main modification we made to the construction scheme developed by Ben-Zvi and Brennan was to make the inner conductor and shorting plate out of a single OHFC forging. This simplified construction by removing an e-beam weld step. Fig. 3. Low- and high-beta resonators. There have been three problems with the construction of these resonators. First, it is difficult to make the inside electron beam weld between the shorting plate and the outer conductor. The weld must be made with the material to be welded far from the electron gun and in a region where it is difficult

6 for the welder to observe the process. After some practice the welder was able to obtain a reasonable success rate with this weld. On site inspection of the weld using a dye penetrant was useful in detecting cracks or pores in the weld. These were then repaired immediately by rewelding. Second, we braze the ground drift tubes in a vacuum furnace, using a copper-silver-nickel braze (Nicusil-3, made by GTE- Wesgo). This was done at first with the resonator standing on its bottom, but later we did it with the resonator upside down, standing on its shorting plate. We depend on capillary action to fill the part of the joint that is up in the vacuum furnace. Sometimes there is a void. We found we could patch this void using tin-lead eutectic solder applied by hand after heating the resonator to the solder melting point with heat tape. When the brazing was done with the resonator upside down, the void was easy to fill by hand. Third, in the early production of the high-beta resonators, we found the inner e-beam weld to be cracked after brazing. We concluded that this was the result of a rapid cool down which thermally stressed this thin weld. After introducing a controlled cool down step, the cracking did not occur. Towards the end of our production runs we had a very high success rate (provided we do not consider the solder patching of the braze joints as an indication of a failure). Viking Forge, in Verdi, Nevada, supplied the OFHC forgings. We machined the outer conductors in our own shop, and Northwest Machine Products of Auburn, Washington made the inner conductors using their NC equipment. Radcliffe Engineering, in Pasadena California did the e-beam welding, and we did the vacuum brazing at Pacific Metalurgical in Kent Washington. The completed copper structure was polished using a tumble polishing technique. Tumbling took about a week. Before the final tumbling step the length of the outer conductor of the resonator was trimmed in order to achieve the frequency desired for 295 deg K. The final tumbling step, which was with a dry "walnut shell" material, was done just before plating. We built 29 low-beta and 13 high-beta resonators between January 1985 and April The resonators were lead plated and polished using a modified version5 of the technique developed at Cal Te~h.~ We did not try the "modernu7 thin plating technique during our production runs, since the more tedious plate and polish technique had been proven successful. Also, we had enough

7 experience with that technique to know that if the plating looked good the resonator would perform well. Thus we required only that the plated resonator pass a visual inspection before we installed it in the accelerator. The main change we made was to use a reverse plating technique. We use a pulsed current source, operating with 50% duty factor on a 20 msec period. Then we would reverse this current, removing some of the lead that had been deposited. This was done with a 15 second period. We found that a smoother finish would result with the current reversed for 15% of the period. More reverse current would give a smoother surface, but would leave regions with small radii unplated. We found that 85% of the resonators plated this way had smooth unstained lead surfaces. The resonators that failed the visual examination were replated. IV. Rf Control System Each resonator is operated in a self excited loop with strong coupling. The quiescent power for the loop is set to about 50 W for the low-beta and 100 W for the high-beta resonators. Most of this power is reflected back from the resonator. We use a circulator to direct the reflected power to a 50-Ohm load. The amplitude of the field in the resonator is adjusted to roughly the desired value using the variable (inductive) coupler. An input attenuator in the loop is adjusted so that its output is a standard amplitude for the desired resonator field. The frequency is adjusted using a differential screw and lever mechanism to deform the bottom plate. This mechanism has a range of about 10 KHz for the low-beta and 30 KHz for the highbeta resonators, and this range requires lo4 steps of a motor mounted on the top of the cryostat. Tuning to within several Hz of the clock frequency is possible. Once the resonators are tuned they need to be retuned only if some large change in amplitude or coupling is made. The bandwidth of the resonators when overcoupled is typically tens of Hertz, although the critically coupled bandwidth is about 1 Hz. The "radiation pressure" (which really results from Coulomb forces between the center conductor and the tuning plate) shift is about 10 Hz between low fields and 3 MV/m. Once the resonator is running at about the right frequency and amplitude it is locked. Amplitude is monitored using a detector diode sampling the input attenuator output. Phase error

8 is developed using a double balanced mixer with the attenuator output compared with a clock signal. Adjustments in amplitude are accomplished by changing the attenuator, and changes in phase are accomplished by shifting the clock signal after it enters the control unit. Thus the signal being controlled remains at a constant amplitude. This control unit was developed at the Weizmann Institute and Stony Brook,' following a scheme of Dela~en.~ The output is developed using a "complex phasomodulator" built by Oelectron Co. This unit permits independent control of the real and the quadrature power in order to control amplitude and phase independently. The resonators are extremely stable in eigenfrequency. Normal laboratory sources of mechanical vibration do not influence the phase error signal. The vibrations induced by banging on a cryostat with a board do make a noticeable disturbance on the phase error signal, but such activity does not cause the resonators to go out of lock. During tests of the prototype high-beta resonators we detected mechanical vibrations of the tuning plate, which is essentially a diaphragm with a fundamental frequency of about 500 Hz. Once the plate was supported in the center by the tuner mechanism, the vibrations were not observed. We have observed a short term drift in the amplitude control when the resonator is first turned on. This drift appears to be due to the temperature coefficient of the detector and to heating of the control unit when the rf is turned on. When the resonator is turned off, we actually turn off the power to the on-board rf amplifiers. This effects the controller temperature. We are considering temperature stabilizing the control units. Long term amplitude drifts appear to be at the level of 1% or less. We have not studied them very carefully yet. Phase stability seems to be excellent, although it is difficult to study at the 0.1 deg level. V. Operating Experience Our resonator operating experience to date consists of a set of test measurements and conditioning runs with each resonator as well as three runs of several days with beam and a fourth several day run with all resonators running locked. The resonators do not show signs of ageing; that is we can reproduce Q vs field curves, and once multipactor conditioning has been successfully concluded, we do not see signs of multipactoring again, provided

9 the resonators are stored at liquid nitrogen temperature under reasonably good vacuum. Some resonators have been in their cryostats for 20 months. It is essential to perform helium conditioning or power conditioning in order to reach reasonable operating fields with our resonators. Before conditioning, we find that the Q factor falls rapidly at about 1 MV/m field with associated x-ray production. We condition the resonators first with a 200-watt amplifier and then with a simple grounded-grid amplifier which we can operate at about 1 kw. Helium is introduced into the vacuum to a pressure of about 2 X 1 ov5 torr. Conditioning is done with pulses that are critically coupled (after the initial transient) with a duty factor of 2 to 5% and a repetition rate of 2 to 3 Hz. Initially the pulsed field in the resonator shows rapid increase with time, and often we see numerous discharges. Usually the discharges stop in a few minutes, and after about 30 minutes the pulse field stops increasing. During conditioning we reach 7 or 8 MV/m in the best resonators. Then we find that field emission will set in between 3 and 4 MV/m. (We estimate peak surface fields to be about 4.8 times the average accelerating field.) If we don't get as high as 5 MV/m during conditioning, the resonator will not perform well, and we are presumably seeing losses due to normal places in the lead rather than due to field emission. 1n.Fig. 4 we present curves of Q vs field for the best high and low-beta resonator as well as for the median resonators. (The resonators were ranked by the fields we obtained during operation to determine the median.) VI. Recent Tests We have operated various subgroups of the resonators at various times, but recently we have had several tests during which nearly all resonators in the accelerator were locked and operated for several days. (One low-beta and one high-beta resonator have tuners which are stuck and one low-beta resonator has a short circuited rf monitor cable. The remaining 35 resonators were operated during the tests.) Excluding the buncher and rebuncher, we tuned the resonators to operate at about 6 to 8 Watts each for the low-beta and 14 to 16 Watts for the high-beta ones. This tuning is done with critical coupling; then we overcouple and amplitude lock at the same rf amplitude. Helium boil-off measurements confirm that the power used is consistent with the value expected from the original

10 measurements. Under these conditions we find the average lowbeta resonator produces an average accelerating field of 2.83 MV/m, while for the high-beta resonators the value is 2.53 MV/m. Q vs (E*) ~ o ~ - ~ ~ ~ ~ ~ ~ I I I, I I I ~ ~ I I (4-1) Tob #7 low p + l (1-1) h# M - 1 (1-1) l.b 07 a Q 108 S // I& 0 0 t I O ~ ; ~ ~ " 1 " " " " " 2 " " " ' 3 " ' ~ ,, I, 1 I, I I I I I I I I I./ U 2 + l1 (la-l) Aag m h I 0 1 ( 0-1) Tob 07 Fig. 4. Q vs Average accelerating field for the best and the median low- and high-beta resonators in the accelerator. The best high-beta resonator is serial number 11, while the median one is no. 1. The best low-beta unit is no. 23, while the median one is also no. 1.

11 The distribution of fields is fairly symmetric, with a standard deviation of 0.34 for each type of resonator. The field values are determined for each resonator by measuring the energy increase it gives the beam. For the low-beta resonators, 2.83 MV/m at 8.0 W corresponds to a Q of 5.9 X 1 o', while for the high-beta ones 2.53 MV/m at 15 W corresponds to a Q of 1.O X 10'. Except for one high-beta resonator, these are the units that were originally installed in the accelerator as it was assembled. As mentioned above, we installed plated resonators which had a reasonably smooth lead surface, and which showed no copper spots or severe staining. No testing other than a visual inspection was performed on the individual resonators before installing them in the accelerator. Resonators were stored under vacuum for a day or more and reinspected before installation. If there were cracked welds or voids in the braze joints, the plating might trap electrolite in these spaces but cover them. During the vacuum storage this electrolyte would dissolve the covering lead and make visible deposits on the remaining lead surf ace. Our most recent test involved the entire accelerator, using the resonators as described above. We were able to put an oxygen beam into one of our scattering chambers. By observing elastic scattering from a gold target, we determined that the energy resolution was better than 1% and the time resolution was better than 1 /3 nsec. The beam was stable. We obtained 30% transmission from the entrance of the linac to the target. The resonators stayed locked at the fields mentioned above during the tests. Those figures, although not yet our design goals, represent a major step toward achieving a working accelerator. We expect that by replating the poorer performing resonators we should be able to achieve our design fields of 3.0 MV/m (Note that this involves improvement of the resonator performance by 0.5 or 1.4 standard deviation, respectively, for the low and high-beta units.) We are looking forward to beginning some nuclear physics experiments as well as to fine tuning the accelerator.

12 References I. Ben-Zvi and J.M. Brennan, Nucl. Instrum. Methods 212, 73 ( ; J.M. Brennan, B. Kurup, I. Ben-Zvi, and J.S. Sok010wski~ Nucl. Instrum. Methods A242, 23 (1985). J.F. Amsbaugh, & d, Rev. Sci. Instrum. 57, 761 (1986); D.W. Storm, & &, IEEE Trans. Nucl. Sci. NS-32, 3262 (1985). I. Ben-Zvi and J.S. Sokolowski, Rev. Sci. Instrum. 57, 776 (1986); I. Ben-Zvi, (private communication). D.W. Storm, & &J., Proceedings of the Second Workshop on rf Superconductivity, Geneva, (1984) p D.W. Storm, J.M.Brennan, and I. Ben-Zvi, IEEE Trans. Nucl. Sci. NS-32, 3607 (1985). W.W. Burt, Adv. Cryog. Eng. Mater. 29, 159 (1983); G.J. Dick, J.R. Delayen, and H.C. Yen, IEEE Trans. Nucl. Sci I (1977). J.R. Delayen, Rev. Sci. Instrum. 57, 766 (1986). I. Ben-Zvi, et al, Nucl. Instrum. Methods A245 1 (1986). J.R. Delayen, G.J. Dick, and J.E. Mercereau, IEEE Trans. on Nucl. Sci. NS-24, 1759 (1977).