The linear proton accelerator for the MYRRHA-ADS

Revue des Questions Scientifiques, 2013, 184 (1) : 99-114 The linear proton accelerator for the MYRRHA-ADS Dirk Vandeplassche, Luis Medeiros RomÃo SCK CEN/ANS/ADT 1. Introduction In order to obtain large yields of fast neutrons it is common practice to use a spallation mechanism by bombarding some high atomic number target material with a high energy and high intensity proton beam. This principle is used in powerful pulsed neutron sources like ISIS [1], SNS [2] and ESS [3]. It is also used as a continuous neutron generator, e.g. at SINQ [4]. The subcritical core of an Accelerator Driven System (ADS) requires a comparable neutron yield for the assembly to deliver a sizeable amount of fission power. Hence the same spallation mechanism is considered. In the case of the MYRRHA project [5], being an ADS delivering around 80 MWth, the neutron production has to be in the 10 17 n/s range. The geometry of its core is optimized for 600 MeV protons. The requested maximum beam intensity is then obtained from Monte-Carlo simulations using the MCNPX code [6] as 4 ma. The fundamental beam requirements for MYRRHA are summarized in table 1. particle beam energy beam current beam delivery mode beam MTBF P 600 MeV 4 ma Continuous Wave (CW) > 250 h Table 1 : Fundamental beam requirements for MYRRHA. MTBF stands for Mean Time Between Failures.

100 revue des questions scientifiques These energy and current characteristics may clearly be considered as safe in the context of present particle accelerator state-of-the-art. The specificity and difficulty of the Continuous Wave (CW) nature of the beam delivery are certainly acknowledged. In particular it is well understood that normal conducting accelerating structures (i.e. RadioFrequency (RF) cavities made from copper) may exhibit severe heating problems when operated in CW mode if not designed with great care for all the thermal issues. The really outstanding challenge, however, is the requirement regarding the Main Time Between Failures (MTBF). In the operational MYRRHA context, the beam is considered to fail if its delivery to the subcritical core is interrupted during a time period that lasts longer than 3 s. The MYRRHA cycle will span 3 months. During such a cycle it is requested to have not more than 10 beam failures, which translates into the quoted MTBF of 250 hours. It should be noted that shorter beam trips are tolerated at a virtually unlimited occurrence frequency. The accelerator for MYRRHA is being developed in a strongly collaborative mode, which is essentially based on the very extensive know-how that is present in the European accelerator community. The R&D activities have, to a large extent, been organized as subsequent European Framework Programme projects [7, 8, 9] 1. In order to support bilateral agreements for common R&D activities, Memories of Understanding (MoU) are concluded with several research institutes. Finally, partnerships based on a commercial agreement, with research institutes or selected industrial companies, allow to realize prototypes in a highly interactive mode. 2. The reliability goal In order to situate the reliability goal of the MYRRHA accelerator (as expressed by the allowed number of beam trips) with respect to actually achieved performances, fig. 1 shows a comparison between recorded SNSdata [10] and the MYRRHA request. In this graph one should distinguish 3 zones, according to the considered trip duration (t ): 1. t > 300 s. This is the zone which typically corresponds to the breakdown of some hardware component needing a local intervention. The 1. The present work is supported by the European Atomic Energy Community s Seventh Framework Programme under grant agreement nr. 269565 (MAX project).

the linear proton accelerator for the myrrha ads 101 graph makes clear that 1 to 2 orders of magnitude have to be gained on the beam trip frequency. The methods that are envisaged to achieve this are discussed below. 2. 3 s < t < 300 s. This interval is mostly characterized by events requiring operator intervention through the control system. The very significant reduction to be achieved here will require a generalization of automated procedures, so as to push the response times below 3 s. 3. t < 3 s. This zone is basically problemless, since the trip frequency is unspecified. Actually, the main mechanism of the reliability increase consists in limiting, whenever this possibility exists, the beam trip durations to this zone. Figure 1: Beam trip frequencies: recorded in SNS, accepted for MYRRHA The key concepts that have been identified in order to significantly increase the reliability of the accelerator for MYRRHA may be itemized as: make the design in such a way that all working points are kept well away from technological limits. use specifications that give useful margins with respect to the requirements (efficiency permitting). apply a strict control on the MTBF of as many components as possible. introduce, by design, the concept of fault tolerance.

102 revue des questions scientifiques Fault tolerance means that, by design, the function of a faulty element can be taken up by one or several other operational elements while preserving the nominal characteristics of the proton beam. From the allowable beam trip spectrum it follows that the sum of the corresponding reconfiguration time and the restart time must be kept below 3 s. It is obvious that this last item is of a much more fundamental and farreaching nature than the others. In fact, it has to be a permanent issue and concern during the entire design process, both in terms of global coherence of the accelerator and of a vast majority of its subsystems. The implementation of fault tolerance implies 1. redundancy. Common sense considers that redundancy is a key issue for reliability. Already during the FP5 PDS-XADS program this statement was confirmed through a reliability modeling study of the accelerator [11]. It is useful to note that, from a machine architecture point of view, redundancy is closely related to modularity. Modules may be arranged in a parallel or in a serial manner, and redundancy schemes based on them may be one-to-one or distributed. Economics and flexibility arguments clearly favour the distributed schemes whenever they may be realized. 2. high performance diagnostics. Reconfiguration upon a fault must be triggered by proper fault detection. Therefore the diagnostic system needs to feature an even higher level of reliability. Also, self-diagnostics are a very important capability in order to avoid the propagation of wrong information. Finally, the diagnostic system should acquire predictive power through advanced statistical analysis tools, so that fault conditions may be anticipated and cured before their actual occurrence. Note that the very high quality of the diagnostic system is also vital for the protection of the machine in the presence of a 2.4 MW beam. 3. powerful and fast controls. The 3 s limit is especially demanding with regard to the low level RF controls, which are the actual drivers of the reconfiguration and of the beam startup. However, at high level it is foreseen to run fast on-line simulation tools that will support defining the updated configuration. Furthermore, the diagnostic system heavily relies on the performance of the controls.

the linear proton accelerator for the myrrha ads 103 4. repairability. The mission time of the accelerator corresponds to the duration of 1 MYRRHA cycle, i.e. 3 months. It is foreseeable that several large scale entities, which consist of many subcomponents (like e.g. the RF system as a whole) will have MTBF-values that are significantly shorter than the mission time. So, in order to keep up the fault tolerance capability all along the mission time it is compulsory to repair failing components during it. It is a question of proper design to ensure this possibility. 3. Description of the linac The fundamental choices regarding the accelerator for MYRRHA have been made by considering the present state-of-the-art, the available experience, ongoing R&D programs and, of course, the findings of the preceding section. A first globally coherent description of this accelerator has been compiled at the end of the FP6 EUROTRANS project [12]. It is being completed and updated during the ongoing FP7 MAX project. A concise description may be found in [13]. In the underlying paper the principal characteristics will be highlighted. s# E in [MeV] cavity type #gaps/ cav. f RF [MHz] β #cav./ CM #CM length [m] E out [MeV] 1 17 spoke 2 352 0.35 2 24 69 81 2 81 elliptical 5 704 0.47 2 17 64 184 3 184 elliptical 5 704 0.65 4 15 101 600 Table 2: Structure of the superconducting linac for MYRRHA, consisting of 3 sections. CM stands for cryomodule. b = v/c and indicates the geometrical value that is used throughout the section. The accelerator for MYRRHA is a SuperConducting LINear ACcelerator (SC Linac), which corresponds to the fact that it mainly consists of a sequence of superconducting accelerating RF cavities, laid out in a linear geometry. In order to reach 600 MeV, this linac has an approximate length of 230 m. This value is relatively high with respect what is presently achievable, and reflects the need for ample margins driven by the reliability requirement. The fault tolerance imposes that all 142 cavities are individually controlled

104 revue des questions scientifiques (implying 142 RF amplifiers, each with its own low level controller). Cavities are grouped in so-called cryomodules these are, from a practical point of view, the elementary building blocks of the modular superconducting linac. The accelerator is subdivided in 3 sections, according to the type of cavity and of cryomodule, as indicated in table 2. Focusing is obtained through warm quadrupole doublets that are installed between the cryomodules. These warm sections also house the beam diagnostics. This superconducting linac, starting at 17 MeV, is laid out in such a way that fault tolerance is achievable up to the cryomodule level: the function of any faulty cavity or even of any faulty entire cryomodule may be taken up by a retuning of the remaining elements. The retuning scheme typically favours the neighbouring elements, but a more global scheme may be envisaged as well. The serial redundancy mechanism is based on a fast sequence of 1. fault detection and localization 2. switching off the beam 3. detuning of the faulty cavit-y/ies + definition of the retuning scheme 4. application of the retuning scheme by reprogramming the concerned low level RF modules (and possibly of quadrupole magnet controllers) 5. beam re-injection with relaxed beam loss tolerances 6. settling of automatic feedback regulation loops 7. resetting operational beam loss levels Below 17 MeV the serial redundancy mechanism is not achievable, because quickly evolving beam parameters preclude the modularity. In order to preserve the fault tolerance capability of the linac, a parallel redundancy has to be implemented, implying the installation of 2 identical accelerator sections up to 17 MeV. Each of these 2 injectors mainly consists of these elements, starting at zero energy: 1. the ion source, of the Electron Cyclotron Resonance (ECR) type for optimal longevity, delivering a moderate 30 kev proton beam. 2. the Low Energy Beam Transport (LEBT) line, for low energy beam characterization and manipulation, and for appropriate beam matching into the next element. 3. the 4-rod RadioFrequency Quadrupole (RFQ), focusing, bunching and accelerating the beam up to 1.5 MeV.

the linear proton accelerator for the myrrha ads 105 4. 2 copper accelerating cavities of the so-called CH-type, bringing the beam to 3.5 MeV. 5. 4 superconducting CH-cavities, combined in a single cryomodule together with solenoids for focusing, up to 17 MeV. The injector is described in detail in [14]. Its latest layout is shown in fig. 2. Each of the injectors transports its beam through a Medium Energy Beam Transport (MEBT) line that ends in a common switching magnet. The polarity of this switching magnet selects one of the beams for delivery into the superconducting linac, whereas the other beam is directed into a dump. Figure 2: Schematic overview of the injector, 0 to 17 MeV. At the exit of the linac, at 600 MeV, starts the High Energy Beam Transport (HEBT) line. There are 3 possible end stations for the beam: the MYRRHA subcritical core. The beam is fed into the core from the top through an achromatic line, of which the geometry is defined by a first 45 vertical dipole magnet, directing the beam upwards. a second 45 vertical dipole magnet, bringing the beam horizontal again. a 90 vertical magnet right above the reactor. Focusing is obtained by 9 quadrupole magnets. The specificity and challenge of this beam line lies in the fact that the vertical part of it, past the 90 bend and into the reactor core, has no beam-optical elements despite of its length of around 25 m. The line is terminated by a window that is cooled by the liquid Pb-Bi being the coolant of the reactor and also the spallation target. In order to properly distribute the power deposit in this window, the beam impact area has to be widened by using 4 fast scanning magnets (redundancy) located next to the 90 bend.

106 revue des questions scientifiques the full power beam dump. For the purpose of commissioning the accelerator in fully operational conditions, this 2.4 MW beam dump will be unavoidable. It represents a technical and a radiation protection challenge, but at the present stage it is only a concept. the ISOL@MYRRHA target station. The possibility exists to feed an Isotope Separator On Line target station in parallel with the ADS operation, by taking profit of the time holes (time intervals with zero beam) that need to be made in the ADS beam for subcriticality monitoring. By using a fast extraction setup (kicker + septum), these 200 µs holes repeated at a frequency of 250 Hz give the possibility to send 5% of the total beam intensity to such a target. There it is used for the production of radioactive ion beams that may serve several experiments in the domains of fundamental physics or of medical radioisotope production. A particularly strong point of ISOL@MYRRHA is the availability of very long beam time periods: 3 months of uninterrupted beam delivery allow to tackle statistically challenging problems. Besides all the equipment constituting the linac itself one should also consider the main ancillary systems. In the case of MYRRHA, these systems providing global services or implementing global functions all have to be chosen and designed in accordance with the plant s reliability goal. The issue is especially relevant for power converters in general. Modularity being a key issue for improving reliability through fault tolerance, this principle is also applicable to a large variety of transistor-based electronic devices, and in particular to DC power supplies and to power RF amplifiers. Some applications are not yet industrially mature. Nevertheless, they are sufficiently developed for R&D purposes and their future potential is widely acknowledged. fast digital low level controls. The generalized use of FPGA s now gives a flexible and affordable access to the development of these fast control electronics. In the domains of low level RF and of fast beam diagnostics they provide unprecedented opportunities. tailored industrial solutions. It is considered that the use of industrially available solutions may offer a suitable guarantee of reliability in such domains as cryogenics and slow beam diagnostics.

the linear proton accelerator for the myrrha ads 107 the unified high level control system. The well chosen architecture of this system is of utmost importance for the exploitation of the machine, which is estimated to involve around 100000 process variables. The control system should essentially be built along Big Physics proven lines, and it may have to be revisited for enhanced reliability. Today the development of an accelerator control system complying with such an approach is industrially available. Overview figures of the MYRRHA linac are shown in fig. 3. Figure 3: Overview of the MYRRHA linac and the high energy beam transport line into the reactor. The 3 drawings should be thought as a linear array. Top: injectors and spoke cavity section. Middle: 2 elliptical cavity sections. Bottom: High energy beam transport line with extraction to ISOL@MYRRHA (not shown), full power beam dump and access to the subcritical core in the reactor. see colour pictures on page 217

108 revue des questions scientifiques 4. Linac R&D program In view of all that precedes, putting a continuous emphasis on the reliability of the accelerator, it will not be surprising that the present R&D program around the MYRRHA linac is entirely focused on this topic. Actually, 4 lines of investigation have been defined around the subject of linac reliability. They are combined in a preliminary scheme that is based on 3 well identified activities, as presented in fig. 4. Figure 4: Scheme of the R&D activities that are foreseen in view of the linac for MYRRHA. 4.1 Yielding a vision on the implementation of a beam-mtbf > 250 hours Starting from a point where the fundamental feasibility of the fault tolerant scheme is well established on the basis of both beam dynamics simulations [15] and of experimental results [16], it is now time to perform the studies and develop the items that are effectively needed for a practical realization with a < 3 s switching time: a performant reliability model of the linac that will allow to identify MTBF requirements for individual components. This study will be based on a Reliability, Availability, Maintainability and Inspectability (RAMI) approach.

the linear proton accelerator for the myrrha ads 109 the error analysis of the linac in configurations exhibiting fault compensation. These studies will use start-to-end Monte Carlo simulations. fast digital Low Level RF (LLRF) controllers. In the case of the MYRRHA linac, the global set of LLRF controllers (one per RF amplifier, hence one per cavity) has an even more vital role than in a standard linac. Indeed, the vast majority of the fault recovery actions will be directed by the LLRF. Also in this case, modeling confirms the theoretical feasibility, given sufficiently fast digital feedback loops [17]. Possibly, adaptive and predicitve loop control may be usefully applied. Prototype controllers will be evaluated on a test cryomodule carrying a single b = 0.5 5-cell elliptical SC cavity (figs. 5 and 6). on-line beam simulators. A crucial step in the fault recovery procedure is the determination of the updated configuration. A most flexible manner of obtaining this is through fast linac simulation calculations, at all times based on the actual linac configuration. diagnostics with predictive capabilities and with self-diagnostics. The extremely difficult task of the diagnostic system will be to provide totally reliable machine protection while avoiding all false interlocks. Wrong pieces of information have to be identified and discarded, and the risk of upcoming interlocks has to be maximally pre-announced. Figure 6: CAD drawing of the operational elliptical test cryomodule. Figure 5: View of the elliptical test cavity being mounted in its cryomodule. see colour pictures on page 218

110 revue des questions scientifiques 4.2 Addressing critical issues through prototyping A number of linac components have been identified as critical items; it is considered that each of them should be the object of an in depth prototyping activity, obviously with strong reliability focus. It must be stressed that prototyping is very effective in the context of a highly modular concept, as is the linac for MYRRHA. Of each type of cryomodule present in the concept (CH, spoke, elliptical b = 0.47, elliptical b = 0.65) a prototype has to be built, and ample time should be foreseen for testing and feedback to the design. Today, this line of activity is (partly) covered by the European FP7 MAX program. In particular the design of a spoke cryomodule for MYRRHA is ongoing. However, the scenario for its construction and experimental exploitation is not defined yet. Figure 7: 2 inside views of the short test 4-rod RFQ cavity. This cavity will be used for thermal behaviour experiments at double linear power density w.r.t. nominal. see colour pictures on page 219 The critical nature of a CW RFQ is beyond doubt. As a first step, a short RFQ section (40 cm) has been built for the evaluation of its thermal behaviour. This activity is led by IAP 2 in the MAX framework. Fig. 7 shows this section; the experimental program is starting. Subsequently the full size RFQ prototype will be realized. This has the significant advantage that is may conveniently be operated and tested with beam by addition of an ion source/ LEBT. Besides for the RFQ itself, such an initial injector then becomes a perfect test platform for a number of other critical issues: 2. Institut für Angewandte Physik, Frankfurt, Germany

the linear proton accelerator for the myrrha ads 111 application of high power Solid State RF amplification non-interceptive diagnostic devices for high intensity CW beams robust controls, at the 3 layers of a typical Big Physics Control System low energy beam collimation, cleaning and chopping methods space charge compensation behaviour in the LEBT full experimental characterization of the low energy beam, allowing to insert optimal initial conditions in beam simulations It is planned to execute this experimental program (under the name RFQ@UCL) in close collaboration with and located at the Cyclotron Center at Louvain-la-Neuve, Belgium. 3 On the other hand it is based on strong bilateral collaboration agreements with the CNRS/IN2P3 laboratories IPNO 4 and LPSC 5 and with the aforementioned laboratory IAP. Finally, the RFQ@ UCL program is obviously also heavily interacting with and relying on the European FP7 MAX program. An overview of the foreseen setup is shown in fig. 8. Figure 8: Overview of the phase 1 experimental setup of the RFQ@UCL program. The round structures represent the 160 kw Solid State RF amplifier that will be installed on top of the bunker. 3. Centre de Ressources du Cyclotron (CRC), Chemin du Cyclotron 2, BE1348 Louvainla-Neuve 4. Institut de Physique Nucléaire, Orsay, France 5. Laboratoire de Physique Subatomique et de Cosmologie, Grenoble, France

112 revue des questions scientifiques At a later stage it will be useful to extend the RFQ@UCL activity towards a higher beam energy, by adding CH-type accelerating cavities to the existing setup. It will then also acquire a clear educational role. Some other items, components of the present design of the MYRRHA linac, have been identified as critical and therefore requiring special experimental and prototyping attention. However, at present the corresponding activities have not been defined, for instance concerning the fast switching magnet permitting to adequately commute from one injector to the other. the beam visualization device to be placed at the very end of the HEBT, just before the beam window, where a VIMOS-like system [18] is foreseen. 4.3 Investigation of future oriented solutions A significant task of the R&D program is to look ahead and to explore promising technological solutions that are expected to have become mature by the time the linac will have to be built. Today it is possible to envisage reliability enhancing developments like Solid State power RF amplifiers. modular DC power supplies, possibly even with hot swap capability. telecom industry based low level control architectures. Other emerging techniques may considerably increase the energy efficiency of the superconducting machine: high Q SC cavities made from ingot Nb. Some of these innovating techniques will be evaluated by effectively applying them in specific R&D programs see e.g. section 4.2. Others will be the object of systematic monitoring of their use in other projects. 4.4 Initiating collaborations with industry It is felt that several industrial companies are likely to play an important role in the realization of the accelerator. Such an industrial involvement may occur at different levels:

the linear proton accelerator for the myrrha ads 113 1. manufacturing of single items or prototypes 2. series manufacturing of recurring accelerator components (e.g., magnets, RF cavities, cryomodules) 3. design, building, installation and commissioning of overall support infrastructures (e.g., cooling, cryogenics) 4. definition and implementation of the control system 5. installation and alignment 6. global coordination and integration of the manufacturing and installation activities At the level of the present R&D program, industrial involvement is limited to points 1, 3 and 4 of the above list. Considering section 4.2, industry is or will be strongly involved for the following: ECR ion source RFQ manufacturing magnets and power converters for the LEBT EPICS-based control system for RFQ@UCL cooling system for RFQ@UCL SC spoke cavity manufacturing spoke cryomodule prototyping It should be clear that several of these activities, while being launched in the framework of an R&D program, are themselves to be considered as investigation objects. The evaluation of the individual industrial partners on one hand, of the industry involving methodology on the other hand, are both part of the R&D program as a whole. References [1] http://www.isis.stfc.ac.uk/about-isis/ how-isis-works---in-depth4371.html [2] http://neutrons.ornl.gov/apgroup/papers/sc LinacCDR.pdf [3] S. Peggs (ed.), ESS Conceptual Design Report, ESS report ESS-2012-1 (2012) [4] http://www.psi.ch/asq/sinq-facility [5] H. Aït Abderrahim, AccApp 11, Knoxville, April 2011, edited by American Nuclear Society (2012) 1

114 revue des questions scientifiques [6] http://mcnpx.lanl.gov/ [7] B. Carluec, P. D hondt, Utilisation and Reliability of High Power Proton Accelerators, Daejeon, Korea, May 2004, edited by OECD (2005) 113 [8] http://cordis.europa.eu/projects/rcn/85226 en.html [9] http://ipnweb.in2p3.fr/max/index.php/maxproject [10] J. Galambos, T. Koseki and M. Seidel, Hadron Beams 2008, Nashville, Tennessee, August 2008, CPL04 (2008) 489 [11] L. Burgazzi, P. Pierini, Reliability Engineering and System Safety 92 (2007) 449 [12] J.L. Biarrotte, A.C. Mueller, H. Klein, P. Pierini and D. Vandeplassche, INAC 10, Tsukuba, September 2010, TUP020 (2010) 440 [13] D. Vandeplassche, J.L. Biarrotte, H. Klein and H. Podlech, IPAC 11, San Sebastian, September 2011, WEPS090 (2011) 2718 [14] H. Podlech, M. Busch, F.D. Dziuba, H. Klein, D. Mder, U. Ratzinger, A. Schempp, R. Tiede, C. Zhang and M. Amberg, IPAC 11, San Sebastian, September 2011, WEPS039 (2011) 2574 [15] J.L. Biarrotte and D. Uriot, Phys. Rev. ST Accel. and Beams 11 (2008) 072803 [16] J. Galambos, S. Hendersen, A. Shishlo and Y. Zhang, HPPA5, Mol, Belgium, ay 2007, edited by Nuclear Energy Agency (2008) 161 [17] F. Bouly, J.L. Biarrotte and C. Joly, LINAC 10, Tsukuba, September 2010, MOP082 (2011) 244 [18] K. Thomson, DIPAC 09, Basel, May 2009, TUPB18 (2009) 206