Edited by Ralph J. Pasquinelli

The PET RFQ project is nearing the end of the first phase of what was envisioned to be a three phase project. This document is a full report on the accomplishments and difficulties encountered during this phase. As such it is important in the assessment of whether or not to proceed to the succeeding phases and, if so, is how to proceed.

This report contains documentation from the efforts of building the RFQ accelerator for production of PET isotopes. It is meant as a summary of design, construction, performance and improvements. Each chapter contains references to more detailed resources when appropriate. The contributors listed below are part of a staff approaching one hundred individuals that have worked on this project for over two years. Their efforts and dedication have created a functional ³He⁺⁺ accelerator.

Contributors to this report include:

Ralph Pasquinelli, Roy Rubinstein, Chuck Schmidt, Ding Sun, Kris Anderson, Bob Webber, Dave Peterson, Brian Chase, Joe Dey, Frank Bieniosek, Del Larson, Nancy Grossman, Dan Holslin, Howie Pfeffer, Duane Voy, Elliott McCrory, Bruce Hoffman, Mike Shea, Richard DeHaas, Phil Young.

The goal of this project was to accelerate ³He⁺⁺ for PET isotope production. Utilization of ³He does not require enriched target material and was believed to reduce neutron radiation flux. The starting point was an existing Strategic Defense Initiative Organization (SDIO) funded 8 MeV RFQ accelerator that had reliability and performance problems. This report describes the redesign/rebuild of almost all elements of that accelerator, including raising the energy to 10.5 MeV. The project design goals are:

• Peak Beam Current=12 milliamperes (mA electrical , 6 mA particle current)
• Average Beam Current = 300 microamperes (µA) (150 µA minimum)
• Duty Cycle= 2.5%
• Beam Uniformity on Target= +/- 25%
• Beam Reliability/Availability= 50% of scheduled time
• Minimal Radiation Shielding Outside the Target Vault, i.e. not a radiation area
• Extensive Radiochemistry Targetry Studies of ¹¹C,¹³N,¹⁵O,¹⁸F
• Technology Transfer for Potential Clinical Use

To date the accelerator has achieved 10.5 MeV ³He⁺⁺ beams with the following typical characteristics:

• Peak Beam Current= 5 mA (electrical, 2.5 mA particle current)
• Average Beam Current = 100 µA
• Duty Cycle= 2.0%
• Beam Uniformity on Target= +/- 50%
• Beam Reliability/Availability= 25-30%
• Entire Accelerator Foot Print is a Radiation Area

The collaboration believes that the reliability of the accelerator will improve, but that the minimum design average beam current of 150 µA will not be demonstrated before the beginning of radiochemistry experiments. The achieved performance is adequate to begin the radiochemistry program in January 1998. The cost to date is $8 million out of the original $10 million budgeted. It is unlikely that the technology transfer goals will be pursued in view of the remaining funds.

This introduction will try to serve as a narrative describing how the project was conceived, the initial goals, what actual performance was achieved, some of the short comings and the lessons learned. More specific technical details are described in subsequent chapters of this report.

The PET RFQ project (or "PET project") is a collaboration between Fermilab (FNAL, Batavia, Illinois), Science Applications International Corporation (SAIC, San Deigo, California), Biomedical Research Foundation (BRF, Shreveport, Louisiana), and the University of Washington (UW, Seattle, Washington) to create a linear accelerator for the production of Positron Emission Tomography (PET) isotopes utilizing Radio Frequency Quadrupoles (RFQs) . ³He is the particle to be accelerated to 10.5 MeV for the production of the four main radio isotopes used in PET, ¹⁸F,¹¹C,¹³N,¹⁵O. Most of the accelerator development has taken place at Lab G on the Fermilab campus. Originally, the project had three phases. The first phase was to design, construct and commission the ³He linear RFQ accelerator. The second phase was to carry out radiochemistry experiments. The final third phase was to investigate the possibility of technology transfer to commercialize the application of RFQ accelerators for PET isotope production. A number of difficulties and setbacks were uncovered during the design and commissioning of the accelerator complex that have made the possibility of phase three improbable with the current funding profile.

This report will concentrate on phase one of the project. Subsequent chapters of this report will detail the major technical problems encountered and their solutions. A second volume under separate cover will concentrate on the radiochemistry aspects of the project to date.

In 1988, the Strategic Defense Initiative Organization (SDIO) of the US Department of Defense requested proposals for new accelerator technology, under development for military purposes, to be applied to medical research.¹ Science Applications International Corporation (SAIC) was awarded funding in 1989 to develop an 8 MeV ³He⁺⁺ RFQ accelerator for the production of PET isotopes. There were several advantages noted for this technique. The proposed RFQs were small and lightweight, and required little neutron shielding (compared to the more commonly used proton/deuteron cyclotrons). In addition, the targets for PET isotope production using a ³He⁺⁺ beam do not require enriched isotopes.

Dabiri et al² described in 1990 the proposed SAIC ³He⁺⁺ accelerator. It included a ³He⁺⁺ ion source (then currently under development at SAIC) and two 425 MHz RFQs for accelerating the particles to 8 MeV. The projected peak electrical current was to be 15 mA, with 166 microsecond pulses at 120 Hz for a 2% duty factor. The RFQ transmission efficiency was predicted to be 97%. In October 1990, SAIC announced³ the achievement of a 0.5 MeV proton beam in their RFQ prototype.

At the Targetry 91 workshop, Hagan et al⁴ updated the design parameters. The ion source was now to produce singly charged ³He⁺, followed by a 212 MHz RFQ accelerating to 1 MeV and then a charge stripper; this was followed by a two tank 425 MHz RFQ accelerating to 8 MeV. The output was to be 360 Hz, 55 microsecond pulses, giving 2% duty factor; pulse current was to be 15 mA (electrical), which corresponds to an average electrical current of 300 µAmps . The original ³He⁺⁺ ion source was felt to be too developmental, and there were also significant changes to the earlier RF system design. At the end of October 1991, a 9 mA peak pulse current had been achieved⁴ of 1 MeV ³He⁺.

In July 1992⁵, tests were carried out on the 8 MeV 3He⁺⁺ performance. A beam of 2.5 mA peak, 120 Hz, 33 microsecond pulses was achieved, giving an average electrical current of ~10 µAmps.

A letter in March 1993⁶ described the then state of the accelerator, following termination earlier that year of the PET RFQ program at SAIC. It noted that there were still significant technological challenges remaining in the RF system, and that several other accelerator systems would have to be improved in order to achieve the design current. A very rough estimate of the costs to bring the machine to design specifications was of order $1M.

A 1993 report⁷ summarized the results achieved before project termination. A current of 4 mA electrical of 8 MeV ³He⁺⁺ was obtained, at a 1.6% duty factor (these numbers correspond to about 65 µAmps electrical average current). Some reasons for the low current were given, including a charge doubling efficiency of only 20%, but the system was very unreliable.

The project has been technically reviewed three times over the last two years, twice by DOE and once by an independent review team appointed by the Fermilab Director. Each review committee consisted of experts on RFQs, accelerator technology, and radiochemistry. Suggestions from these committees were carefully scrutinized and, when possible, implemented in the redesign of the accelerator. Very little of the original SDIO funded accelerator remains with the current design. These details will be addressed in the following report.

The original intention was to complete machine design and commissioning at Fermilab followed by dismantling the accelerator complex and shipping to BRF in Shreveport, Louisiana. A facility in the physical plant at BRF has been completed should this option still be part of the collaboration’s plans. The decision whether or not to ship the accelerator will be made during Spring 1998. The radiochemistry program could also be carried out at Fermilab should the collaboration decide to keep the accelerator at its current location in Lab G. If the Fermilab option is chosen, some infrastructure modifications to lab G will be necessary to allow more efficient operations for the radiochemists.

Current status of the accelerator commissioning has the machine operating at Lab G under the stewardship of SAIC personnel. Chapter 16 will contain the details of operations at Lab G during a trial period with minimal FNAL intervention. The intention of this period was to simulate day to day operations of the accelerator in a remote location, an important factor to consider should shipping the accelerator to BRF be chosen.

There was an extensive period of litigation between government agencies as to when transfer of the existing accelerator hardware could be made between DOD and DOE. During this period, the accelerator had been moved to BRF from San Deigo for the dedication of the new Biomedical Research Institute (BRI). The project was officially resumed in September 1995, with a $10 million funding profile. At this time, there was no detailed design report available. The original machine concept described in Reference 2 was not the machine that was designed and built under the SDIO contract. What was built under the SDIO contract was a three RFQ system where the first RFQ operated at 212.5 MHz and was an integral part of what became a 1 MeV doubly charged ³He source.

The first course of action after unpacking the boxes shipped from Shreveport, was to evaluate the hardware. Several tests were the preliminary goal, 1. To understand the performance of the RFQs, 2. to assemble the singly charged ³He ion source, and 3. to experiment with the SAIC gas jet charge doubler. Four to six months was necessary to assemble the test stand to carry out these investigations. By spring of 1996, work on all three objectives was well underway. Several very important discoveries were made during this period.

The x-ray emissions from the 212 RFQ greatly exceeded the minimal level required for continuous occupancy. Numerous lead blankets were wrapped around the RFQ to reduce the radiation level to allow continuous occupancy. The RF system used to drive the RFQ was one of the surplus FNAL Linac stations that had recently been decommissioned. Even though the RF station was adequate for the test, it clearly lacked controls and interfaces necessary for integration into a reliable accelerator (all the old Linac interfaces, interlocks and controls were not part the surplus hardware). Because these tests were only to 1 MeV, there were none of the problems of phase lock and stability that would be discovered as subsequent RFQs were put on line. The testing continued very smoothly through the summer of 1996.

An additional test was carried out on one of the 425 MHz RFQs once surplus hardware was procured from the canceled GTA project at Los Alamos. It was fortunate for the PET project that the GTA utilized the same 4616 tetrode in one of their RF drivers. The PET project purchased three 4616 amplifier cavities and four solid state amplifier drivers. These early tests of the 425 MHz RFQ showed some sparking of the RFQ vanes when driven with 73 kWatts, the estimated correct value for achieving the gradient to accelerate ³He⁺⁺ from 1 to 5 MeV. There was no real spark protection in the test setup. Later inspection of the RFQ showed discoloration and loss of copper plating on the vane tips in the middle of the RFQ structure. (In a subsequent inspection of these vanes by Gerry McMichael of Argonne National Lab, he disclosed that the damage to the vanes was not detrimental to their future use.)

The tests on the ion source showed a number of design flaws that needed attention. Initial testing indicated excessive usage of ³He gas. At $115 per gaseous liter, this would become a serious operational expense if not remedied. The source was totally disassembled and revealed a number of "holes" in the source chambers that passed the He gas through to the vacuum pump unnecessarily. Once plugged, gas consumption was immediately reduced. The arc modulator power supply was not robust or reliable. A redesign of a pulse forming network power supply similar to units used at Fermilab was pursued.

The gas stripper/charge doubler was initially based on a static gas cell. Striping efficiency was low and could be improved by a gas jet, a totally new design. The valve used for the SDIO project was piezo electric and very expensive. It had a limited lifetime estimated at 1-100 million cycles. With the estimated accelerator design duty cycle of 2.5% running at 360 Hz, this translates to 0.77 to 77 hours of operation, clearly an unacceptable replacement interval. A new design based on an automobile fuel injector was adopted. The gas load necessary to strip ³He⁺ to ³He⁺⁺ was such that a new vacuum system with ample pumping was necessary. In addition, these first tests experimented with a number of different gases to test stripping efficiency, including helium, nitrogen, argon, carbon dioxide and xenon.

The first nine months of effort were spent trying to understand exactly what the characteristics of the inherited project were. It was at this same time that the first formal DOE review was conducted on May 29-31, 1996. The project was closely scrutinized by the review team and the following action items were disclosed: 1. Perform an ion source to target particle tracing with space charge effects included, i.e. an end to end simulation. 2. Set up a detailed project management scheme to aid tracking of project progress. 3. Develop a specific set of performance standards that would be agreed upon by all collaborating institutions. Meet these minimum requirements before shipping the accelerator back to BRF. 4. Hold a second review of the project in six months.

Action items 1 and 3 were absolutely critical to the technical success of the project. It was indeed true that no full simulation had been carried out by the SDIO project. By the same token, we were discovering the performance of the existing hardware and it became evident that some reverse engineering would be necessary. The end to end simulation was part of this process. Upon rebirth of the project, it was believed that the lack of a reliable RF system was a major cause for poor operations. At this stage, this simple assessment was not true as many subsystems required attention. In the period between DOE reviews, extensive work was done on the simulations and redesign of hardware support systems continued in parallel. The performance goals for the project were optimistic as are most goals of accelerator design. A minimum acceptable performance was drafted and agreed upon by all collaborators on August 30, 1996. This document has been included as an appendix.

A number of MEBT designs were considered. The one chosen was the idea of Delbert Larson, an SAIC consultant. The new MEBT would consist of an isochronous 540 degree magnetic lattice to preserve bunch length. An added benefit of the design was that the accelerator now folded back on itself essentially shortening the overall length and easing the siting problem. (As it turned out, the project’s future home was prudently relocated to the BRI physical plant because the unknown nature of the shielding requirements could have made location on the second floor impossible.) There were serious concerns over the space charge effects of the beam in the MEBT. Extensive modeling was conducted, but understanding any nonlinear space charge effects would have to wait until actual beam tests.

Once adopted, the design for the MEBT was pursued vigorously. The lattice required the design and construction of two large 270 degree bending dipoles and seven quadrupoles of various strengths. All magnets were designed and fabricated at the Fermilab Technical Support Division. This effort continued through November 1996. There were a few technical problems that required some rework, but for the most part the magnets were produced on schedule and have performed remarkably well.

A great deal of activity took place between the DOE reviews of the project. The following is a paraphrasing of findings of the second review committee and subsequent actions by the PET design team.

The original ion source operated well in preliminary testing. The optics for the ion source were redesigned and required a new vacuum vessel. From this point on in the project, the reliability of the ion source filament became a source of concern. The components for the ion source contain some organic plastic insulators and the entire source was not stored in an inert environment. It is believed that some form of contamination is the culprit for filament failures. The filament continues to be problematic, but new and improved designs and procedures are being followed to improve performance and reliability. Emittance measurements were performed with equipment borrowed from the FNAL Linac department. Two recommendations were presented. Relocating the steering magnet and use of consistent units of emittance. Both recommendations were or are being acted upon.

The transmission efficiency of the 212 MHz RFQ was typically 65% in actual testing at lab G. Further simulations with the RFQ code PARMTEQ supported this transmission efficiency. In the period between DOE reviews, the 212 MHz RFQ sprang a leak between the vacuum and water cooling passages of the vanes. The leak was caused by the corrosive action of water on the aluminum alloy (7075-T6) from which the vanes are fabricated. This problem was temporarily repaired with an impregnation technique to stop the leak (more details in chapter 12) and allowed continued testing. Three recommendations were presented. 1. Additional probes to measure fields in the RFQs should be added. A total of 12 such probes were added to all of the RFQs. 2. Continued modeling of the RFQ should utilize higher order term analysis software. This should be completed before making new replacement vanes. It was decided to take this opportunity to not only improve modeling, but also redesign the RFQ in an attempt to improve transmission efficiency. 3. X-ray production endpoints should be conducted to truly understand the intervane voltage. These tests were conducted on the 212 MHz RFQ with conclusions that show agreement between the theoretical fields and the extinction level of the x-rays produced.

The charge stripper testing with the fuel injector was successful in prototype testing to 60 Hz repetition rates. No measurable emittance dilution was observed at stripping efficiencies approaching 83%. The final stripper design was able to utilize a single fuel injector operating at the full duty cycle of 2.5% at 360 Hz. Additional pumping was added to the final stripper, a total of 4000 liters per second. Final stripping efficiencies at full duty cycle are approximately 70%. Argon gas showed the highest stripping efficiency, but its high molecular weight and low thermal conductivity caused excessive heating of the turbo molecular pumps at full duty cycle. Nitrogen gas is now utilized daily with only a few percent degradation in stripping efficiency, but vastly improved vacuum performance. Operating at the higher duty cycle reduced stripping efficiency because the pumps are not able to fully recover vacuum between 360 Hz cycles.

The MEBT received the highest level of scrutiny. The review committee was convinced that the MEBT performance as designed was questionable at best. The recommendations were to look into developing contingency plans in case the MEBT was a failure. Other suggestions were: addition of an isolation vacuum valve to the MEBT diagnostic box, use TRACE3-D to evaluate magnet settings, add as many diagnostics as possible, try to do additional modeling utilizing the code PARMILLA and WARP3D. At this stage in the project, the PET team was committed to the MEBT design. All the magnets in the lattice were close to completion, those that were complete met the specifications. More modeling did take place as suggested. The diagnostic box was jammed full of diagnostics including segmented Faraday apertures, a current toroid, and beam profile monitors for use at low duty cycles during commissioning. As it turned out, it was mechanically very difficult to include a vacuum isolation valve in the MEBT, hence that recommendation was not followed. Some back up plans were discussed, but if the MEBT did not work, there would be a big delay in the project. It was a bit of a gamble, but a victory for the project. The MEBT has worked very close to design performance and has been very reliable.

The HEBT comments included lack of instrumentation in the beam line, unclear how 10.5 MeV is verified, use of a dog leg to reduce neutron flux in the reverse direction. The committee was impressed with progress on the target window development. Recommendations included characterizing the HEBT magnets and look into an insertable profile monitor. The magnets were indeed sent to the magnet test facility for characterization. They were also carefully studied for cooling efficiency since the HEBT magnets were designed to focus 8 MeV beam not 10.5 MeV. As a result, they would need to be run at higher currents. The magnets passed the thermal testing at design current for 10.5 MeV. Because of the high beam energy in the HEBT, a profile monitor of the wire type would not survive for very long. Instead a set of movable collimator jaws was designed and built that could handle the full beam power if necessary. A software application was written to do collimator scans to test for beam uniformity on the target window. This program has worked well and has been verified against carbon foil irradiations. (A carbon foil is irradiated, cut into strips, then very carefully counting for residual activity to measure beam uniformity.)

Recommendations for the RF systems was to improve the arc protection protocol and add RF field probes to all RFQs. The PET team has indeed incorporated a systematic means of controlling RF fields in the RFQs after arcing occurs. Because each of the RFQs needed to be rebuilt due to water corrosion of the vanes, the probes were easily added to all RFQs.

Action items from the second review: 1. Continue modeling of RFQs. 2. Appoint single lead subsystem mangers. 3. The Fermilab Director should appoint an independent advisory committee to review work on Accelerator construction and plans for future radiochemistry experiments. 4. Conduct another DOE review if the accelerator has not met project milestones by spring of 1997. Modeling continued and was completed. Single subsystem managers were appointed, an advisory committee was chartered, and there was no need for the next review as 10.5 MeV beam was accelerated in April of 97.

By the time the advisory committee convened, the accelerator complex was very close to completion. Beam had already been accelerated to 10.5 MeV. A number of useful comments on operations were presented and appreciated. A large part of the review concentrated on the targetry program and safety issues related to relocation of the accelerator to Shreveport. As the targetry and Louisiana safety regulations are not part of this report, nothing further will be mentioned here.

All of 1997 was devoted to final construction, assembly and commissioning of the accelerator complex at Lab G. Even though a number of serious problems had been addressed and corrective action was taken, there were still more unforeseen problems ahead.

One of the more serious delays was caused by the problems encountered with machining and plating of the RFQs. The water corrosion problem meant that all four RFQs would have to be built by the project. (Only the 425 MHz RFQ C was in the original plan.) There were several iterations with machine shops to find a company that could precisely machine the vane contours. The major headache was finding a plating house that could successfully apply the copper plate to the aluminum alloy. Each of the RFQ vanes and the RFQ C housing experienced plating problems and were returned to the vendor. A new plating vendor was finally found that could complete the job successfully. By April of 1997, all major components were in place and 10.5 MeV beam was first accelerated.

As commissioning ensued, one problem after another surfaced. The filaments in the ion source were failing after one to two weeks of operation. Previous work with the ion source showed filament lifetimes exceeding two months. This performance was not repeatable. The RFQs were slow to condition, the RF power systems had several infant mortality problems. As duty cycle was increased to the full 2.5 %, the RFQ coupling loops began failing due to overheating. The design of the coupling loop was such that the vacuum sealing o-ring was located in the region of highest electric field. As a result, they overheated and caused vacuum failure. The loops were redesigned to relocate the o-ring to the interior of the center conductor. This fix was effective as no new loop failures occurred. The water systems for the project were specified internally, but procured on the outside. They experienced cooling compressor failure, valve control failure, flow transducer failure, and improper control loop parameters. All have been fixed and the system has been performing reliably since. As is often the case with Murphy phenomena, all of the failures occurred separately, few in parallel, with the end result of extended down time necessary to fix the problems. These types of difficulty plagued the project for the last half of 1997 and made commissioning and operation of the accelerator a harrowing endeavor.

A number of attempts to start preliminary radiochemistry experiments began in late May 1997. The first few attempts were focused on operational procedures and shakedown of the hardware. By the end of the year, most of the low beam current tests had been performed and the need for sustained beam currents approaching 100 uAmps was desirable. By December 1997 most subsystems were operating reliably. There was one major component that was still outstanding, the fast valve trigger. A fast valve (one able of closing in 7 msec) is located in the HEBT. Its primary purpose is to close rapidly in the event a target window shatters. Earlier experiments showed that the fast valve trigger sensor supplied with the valve was so sensitive as to trip from outgassing of the target window with moderate beam current levels. A new sensor was on back order for many months. A backup sensor was employed, but not as fast as one would like. The bottom line was that a window broke in December, the fast valve stopped most of the debris, but operation of the accelerator is not like horse shoes, close does not count. The material from the window contaminated the RFQ C and returning to full gradient became impossible. The RFQ was disassembled and clearly had signs of serious arcing damage. The debris deposited in the RFQ was also now radioactive, making for a contaminate waste dilemma. The RFQ was cleaned, reassembled, and put back under vacuum before the Christmas holiday. In early January, the RFQ was reconditioned and beam commissioning resumed. By this time, the new fast valve sensor was delivered, inspected, installed and tested before the next chemistry experimentation period.

With most of the problems of reliability solved, one main headache remained, keeping the RFQs, all four of them, tuned on resonance and at full gradient. For some as yet unexplained reason, the conditioning of the RFQs has remained a mystery. Several weeks are required to recondition and recover from maintenance periods that bring the vanes up to atmosphere. Even when conditioning has been reached, the acceleration of beam exacerbates the spark rate in the 425 MHz RFQ chain. The beam losses between RFQs approach 20% and seem to increase the spark rate. The detuning of the RFQs approaches several kilohertz per second after a spark. The thermal tuning loops are unable to respond quickly and the RFQ frequencies start to wander. When there are four nomadic RFQs, it can take 20-30 minutes to settle down. This proves to be a very frustrating environment to perform chemistry experiments. A solution to this problem is now in progress. The original SDIO accelerator utilized mechanical tuning paddles for resonance control. By themselves, they were not capable of maintaining resonance control and were abandoned early in the project. A thermal control loop was adopted, but lacked the response time needed to recover quickly from sparks. The hybrid system now being implemented uses both thermal and mechanical tuning. It is expected that resonance control recovery will now be seconds not minutes.

Success at Last

In early February 1998, the first serious radiochemistry experiments were successfully carried out. With numerous failures and fixes in the past, the accelerator was able to deliver sustained beam currents of 50-75 uAmps to the target. The radiochemists were able to obtain twice as much useful data as in all previous attempts put together. While the resonance tuning problems continued, they were not so severe as to stop the experiments. A March 1998 run was very successful, with the new tuning paddles providing prompt recovery from arcs. Plans are to install tuning paddles for the remaining 212 and 425 C RFQs.

As is evident from this introduction, the project was not without difficulties. There were a number of designs that needed re-engineering. The project clearly exceeded the initial time estimate for completion of the accelerator and the cost estimate. As progress on the accelerator continued, many times it felt like opening Pandora’s Box. Although this project utilized some of the hardware from the SDIO program and surplus RF stations from the FNAL Linac, the design of this accelerator has provided a number of novel hardware contributions and important changes to software programs utilized in designing accelerators. The MEBT design may be a useful mechanism for matching RFQs to subsequent accelerating stages. High quality 270 degree bending magnets were designed and implemented. The simple but effective use of a fuel injector for a gas stripper provides a robust and reliable gas jet target. The extensive modeling of the accelerator revealed a number of fundamental errors in codes that have been used for years in the field of accelerator design. All of these contributions and more will be described in subsequent chapters.

It was initially believed (SDIO) that the beam loss between RFQs could be tolerated because of optimistic goals for accelerator components. In reality, no part of the accelerator achieved the original design parameters. The operating parameters that have been achieved are listed in the table below. If the use of a ³He beam for producing PET isotopes proves to be superior to proton cyclotrons, a better means of accelerating the beam would be to use the same 1 MeV source and stripper, but to follow the MEBT with a short drift tube Linac. The drift tube section would have a larger aperture, lower losses, and higher reliability over the RFQ structures used in this project.

Peak Beam Current 12 mA (electrical ++) 5 mA (electrical ++)

Average Beam Current 300 uA (electrical ++) 100 uA (electrical++)

Ion source Current 25 mA peak 21 mA peak

Ion source emittance 0.76 pi mm.mr. (90%) 1 pi mm.mr. (90%)

Duty Cycle 2.5% 2.0%

Stripping efficiency 70% 68-70%

MEBT transmission efficiency 100% 70%

MEBT input emittance 0.8 pi mm.mr. (90%) 0.9 pi mm.mr. (90%)

MEBT output emittance ^a 0.9 pi mm.mr (90%) 0.8 pi mm.mr. (90%)

RFQ 212 transmission ^b 80% 80%

RFQ 425 A-B-C transmission 43% 34%

HEBT beam uniformity +/- 25 % +/-50 %

Personnel occupancy Not a radiation area Accelerator footprint is

radiation area

(a) Scraping is observed to lower emittance in the MEBT

(b) - Accurate measurements of emittance and Twiss parameters exist only for

the old 212 RFQ. The new 212 produces more beam (18 mA=75% ?) but much of

the additional current is lost in the MEBT/stripper. Losses may be caused

by scraping due to a different exit angle (Twiss parameters), or increased

output emittance. Unfortunately we did not have the opportunity to remeasure

output emittance and Twiss parameters for the new RFQ.

(c) Initial 'design value' numbers were very aggressive and perhaps unrealistic.

1. See the introduction in Reference 7.

2. A. E. Dabiri et al., Nuclear Technology, Vol. 92, p. 127, October 1992.

3. Announcement of Proton Beam in SAIC Radio Frequency Quadrupole Accelerator.

4. W. Hagan et al., Proceedings of the IVth International Workshop on Targetry and Target Chemistry. PSI Villigen, Switzerland; September 9–12, 1991; edited by R. Weinreich, p. 19.

5. Memo K. Krohn and J. Link to P. Young and W. Hagan, 15 July 1992, subject: Trip report for July 6–9, 1992.

6. P. Young to K. Krohn, March 29, 1993.

7. K. A. Krohn et al., Proceedings of the Fifth International Workshop on Targetry and Target Chemistry. Brookhaven National Laboratory, September 19 to 23, 1993; edited by R. Dahl, R. Ferrieri, R. Finn, D. Schlyer, p. 20.

Chapter 1. Pre 1995 Project History. A short description of the project before it was restarted with the current collaboration.

Chapter 2. Ion Optical Design of the RFQ PET Accelerator. The design of a doubly charged ion source was beyond the scope and budget for this project. A singly charged ion source original to the SDIO project has been modified and improved. Filament problems have plagued this source since the early stages of the project. When it appeared that the problems with filament burnout were overcome, they resurfaced in the last 3-6 months. A serious effort is now underway to understand the lack of reliability. Currently produced filaments typically last only 2-3 weeks. Filaments have been produced that lasted 2-3 months, an acceptable lifetime. This level of reliability must be re-established. The ion source does deliver the design beam current when the filament is operational.

The RFQs for the project are basically the same design as the original RFQs under the SDIO funded program. All RFQs needed to be rebuilt because of water corrosion problems. It was decided to take this opportunity to redesign the first RFQ in the chain to improve its capture efficiency. In addition a fourth RFQ was added to increase the energy of the beam to 10.5 MeV. All four RFQs were designed, built, tuned and tested within the current collaboration.

Chapter 3. RFQ Tuning Systems. The resonance tuning system employed in the SDIO program exhibited a number of reliability problems associated with mechanical paddle tuners. The complete system was redesigned utilizing thermal tuning via water cooling/heating.

Chapter 4. Low Level RF. The LLRF system is a completely new design based on the latest electronic techniques. Extensive digital signal processing is employed to give the system a high degree of flexibility.

Chapter 5. High Level RF. The High Level RF of the original SDIO program was a serious reliability problem. The use of planar array power amplifiers was a constant source of operational down time. At the outset of the current project, it was believed that the surplus FNAL Linac RF stations could be modified to provide the RF power. These surplus stations had a 25 year high reliability operational history. The stations were indeed modified and are now operating with minimal maintenance. Additional 425 MHz surplus hardware was purchased from Los Alamos National Lab.

Chapter 6. Charge Doubler. The 425 MHz RFQs are designed to accelerate doubly charged ³He⁺⁺. Calculations and analysis for removing the second electron showed the need for a gas jet charge stripper versus a stripping foil. The technique from the original SDIO project of utilizing a piezo valve for the gas jet proved unacceptable due to short lifetime and poor performance of the piezo valve. A new gas jet stripper utilizing an automobile fuel injector was designed, built, and tested. This device now routinely provides 70-80% charge stripping efficiency.

Chapter 7. Machine Operations and Performance. The chronology of events leading up to present day operations and current performance is presented.

Chapter 8. Diagnostics. A complete suite of diagnostic instrumentation was designed for the accelerator. Some of the devices include charge integrators, wire profile monitors, Faraday apertures and cups, and a beam intercepting longitudinal monitor. An emittance scanner was borrowed from the ion source group at FNAL to do emittance scans at 1 and 5 MeV. The equipment has worked extremely well, allowing for ease in commissioning all sections of the accelerator.

Chapter 9. Equipment Interlocks. Virtually all of the hardware associated with the accelerator is new to the current project. An extensive hardware protection interlock system was designed and built based on commercially available Programmable Logic Controllers (PLCs). This system has already proven valuable in protection of RF, vacuum, and water subsystems.

Chapter 10. Radiation Safety and Shielding. Radiation safety for this machine demanded a careful in-depth effort. A number of serious problems were encountered, analyzed and mitigated. Although one of the selling points of a linear particle beam was low radiation levels, the opposite was found to be true. The initial SDIO effort concentrated more on light weight design than on low radiation levels. As such the RFQ vanes, housings and vacuum vessels were designed from aluminum alloys. The x-ray radiation levels of the RFQs in the aluminum vacuum vessel exceeded 10 Rads/hr on contact. New Stainless steel chambers 1-5/16 inch thick were designed, installed and eliminated the x-ray hazards. The neutron production from 8 MeV ³He⁺⁺ on aluminum was difficult to estimate. Radiation levels near the RFQ junctions were found to exceed 100 mrem/hr due to particle beam losses associated with the gaps between adjacent RFQs. This required a radiation interlock system to be installed. The entire accelerator enclosure is now an interlocked radiation area. Additional shielding will be necessary for relocation of the machine to Shreveport. (This problem is currently being addressed.) The radiation levels in the target vault exceed the shielding capabilities of the current vault design. While this does not present a problem in the lab G location at Fermilab, a new or modified vault design will be necessary for BRF in Shreveport.

Chapter 11. Magnet Power Supply Systems. With the addition of the MEBT to the project came the need for nine additional powered magnets. Together with existing magnets for LEBT and HEBT, the system now has in excess of twenty power supplies. The MEBT has the additional requirement for stable current regulation that exceeds standard catalog power supply specifications. A complete remotely controlled and monitored power supply system has been designed, built, installed, and runs reliably.

Chapter 12. Vacuum and Mechanical Systems. The complete mechanical systems for the accelerator were redesigned from the ground up, literally. New stands for supporting the accelerator structure were designed to allow easy assembly and maintain ten mil alignment tolerances. A new vacuum system was carefully specified, analyzed, designed and installed. All mechanical designs have taken "mobility" into consideration so that the accelerator could be easily relocated to Shreveport. A great deal of mechanical design was required to create the functional Charge Doubler described above. Water corrosion problems in the RFQs required a temporary impregnation fix as well as coating new RFQs with Teflon for long term protection.

Chapter 13. Water Cooling Systems. The design of the accelerator calls for self contained cooling systems. The water systems for the project are based on commercially manufactured components. The systems include three types of cooling loops, Low Conductivity Water (LCW), four independent RFQ cooling loops, and a central heat exchanger loop. After some initial infant mortality of components, the water systems are functioning well. Some water temperature control problems are being addressed as the factory settings on the heat exchanger control loop were not optimized for our system.

Chapter 14. Control System Hardware. The control system hardware is the spinal chord of the accelerator. Clearly the design of this accelerator held many challenges. One system where challenges were not welcome was the control system. It was decided to adopt the field proven design of the FNAL Linac control system. The hardware did not require extensive redesign, hence reducing cost to the project. The control system has indeed provided reliable operational experience.

Chapter 15. Control Software. Although the control system hardware existed, a great deal of application code needed to be developed to support accelerator commissioning and operations. In addition, it was decided to create a user friendly graphic control environment to facilitate daily operations and diagnostics. Software was written utilizing commercial and in house applications.