Commissioning experiences with the spoke-based CW superconducting proton linac

Fang Yan· Hui-Ping Geng · Cai Meng · Ya-Liang Zhao · Hua Shi ·Hua-Fu Ouyang· Tong-Min Huang· Fei-Si He· Jian-Li Wang ·Rui Ge· Yan-Feng Sui · Qiang Ye · Rong Liu · Qun-Yao Wang ·Sheng-Chang Wang· Jian-She Cao· Wei-Min Pan

Abstract The 10-MeV Accelerator-Driven Subcritical(ADS) system Injector-I test stand at the Institute of High Energy Physics (IHEP) is a testing facility dedicated to demonstrating the feasibility of the spoke-based superconducting (SC) linear accelerator (linac) for the ADS project in China. The injector adopted a four-vane copper structure radio frequency quadrupole(RFQ)with an output energy of 3.2 MeV and an SC section accommodating 14 βg=0.12 single spoke cavities,14 SC solenoids,and 14 cold beam position monitors (BPMs). A 10-MeV pulsed beam with a beam current of 10 mA and a 2-mA continuous wave (CW) beam were successfully shooting through. The commissioning results confirmed the feasibility of using a 325-MHz spoke-type cavity for accelerating the proton beam in the low β and medium β sections. This paper describes the results achieved, the difficulties encountered,and the experiences obtained during commissioning.

Keywords Superconducting proton linac · Spoke cavity ·High intensity · CW commissioning

1 Introduction

The Accelerator-Driven Subcritical (ADS) project was launched in China in 2011 with the intention of developing the concept and design of a 1-GeV high-intensity superconducting (SC) linac with the aim of building a demonstration facility for an accelerator-driven subcritical system(ADS) in multiple phases. The driver linac operates in continuous wave (CW) mode and eventually delivers 10 ～20 MW of beam power[1,2].The linac includes two major sections: the injector and main linac sections. The injectors accelerate the proton beams up to 10 MeV, and the main linac boosts the energy from 10 MeV to 1 GeV.

To develop key technologies for the design,fabrication,and operation of a high-intensity CW linac and demonstrate the design schemes, two different injectors applying two different frequencies were proposed [3, 4], fabricated,and commissioned in two different institutes independently. The Institute of High Energy Physics (IHEP) was responsible for Scheme I(Injector-I)based on a 325-MHz(the same frequency as the main linac) room-temperature radio frequency quadrupole (RFQ) and SC spoke cavities of the same RF frequency.The Institute of Modern Physics(IMP)was responsible for Scheme II(Injector II)based on 162.5-MHz room-temperature RFQ and SC half-wave resonator (HWR) technology of the same frequency [5].

The general layout of the ADS Injector-I linac is shown in Fig. 1. Injector-I is composed of an electron cyclotron resonance (ECR) ion source, a low-energy beam transport(LEBT) line, a 4-vane type copper structure RFQ, a medium-energy beam transport(MEBT)line,an SC section,an energy divergence analysis (EDA) system, and a beam dump line.

Fig. 1 (Color online) The general layout of the ADS Injector-I linac

Table 1 ADS Injector-I test facility specifications

The specifications of the Injector-I are listed in Table 1.It was designed to accelerate the proton beam up to 10 MeV based on a frequency of 325 MHz. Although the injector was designed to be operated in CW mode with an average beam current of 10 mA, considering the CW operation difficulty for the high intensity proton linac,CW operation was not an essential goal for the first stage of injector commissioning. However, we still pursue the operation of CW proton beam with an energy of 10 MeV and as high average beam current as possible.

This contribution presents the challenges faced and experiences obtained during the commissioning of a spokebased SC linac based on the ADS Injector-I developed in IHEP. Section 1 briefly introduces the background of the project. Section 2 then summarizes the design philosophy of using spoke cavities for a 10-MeV CW injector. The commissioning details under pulsed and CW modes are presented in Sects. 3 and 4,respectively.The final section provides a brief summary.

2 Design philosophy

CW operation is a critical objective for a high-intensity proton linac used in ADS applications. Although there are no high-intensity SC CW machines operating at the design stage of ADS injector, several proposed CW proton/deuteron linac projects such as Project-X [6, 7], EFIT or MYRRHA [8], IFMIF [9], and EURISOL [10] are good models for the physical and technical designs of the accelerator.The ADS driver linac must be designed with a very high beam power (MW-scale), and the beam losses must be kept as low as possible along the linac with a normal acceptance of 1 W/m for all high-power proton accelerators. It must also be designed with extremely high reliability with limited beam trips at a time scale of seconds(as listed in Table 2),which surpasses those of the existing proton linacs. All of these requirements induce extremely high restrictions on the linac design and commissioning.

2.1 Why 325 MHz?

As mentioned in the Introduction, Injector-I was designed based on a frequency of 325 MHz. Compared to 162.5 MHz used by Injector-II,the benefit of choosing 325 MHz for room-temperature RFQ is the acceleration effectiveness and smaller cavity size. In addition, the beam space charge effect with a 325-MHz frequency is also weaker compared with the half frequency because they both have the same average beam current (under the condition in which the beam size is the same to maintain the same acceptance for the downstream linac). The space charge effect is crucial for ADS applications with megawatt beam power because non-relativistic proton beams with stronger space charge effects are more sensitive to the mismatch and have a higher possibility of inducing parameter/structure and coupling resonances, leading to halo growth and particle losses. However, if the space charge effect is kept the same for both frequency choices,the beam size out of the injector must be increased for the half-frequency case.A larger longitudinal beam size means more cavities and more cost to maintain the same acceptance for the downstream linac. Another benefit we considered for choosing 325 MHz,was that no frequency jump will occur for the transition between the injector and main linac.Although the frequency jump can be realized through LLRF control [11], the transition part can easily be a source of halo development because a matched beam is difficult to achieve with strong space charge effect at the low-energy part where the periodic lattice is discontinued,and the situation will be worse if a frequency jump occurs.

There are also challenges for 325-MHz scheme. While operating in CW mode, the biggest issue of a copper structure RFQ is the power dissipation on the cavity surface. For the copper structure RFQ, the power density is closely related to the frequency (f). The power dissipation is proportional to f3/2, whereas the aperture is inverselyproportional to f. For the same cavity length, the power density will be 5.7-times larger for the 325 MHz RFQ than for the half frequency. This is a significant challenge for both the 325-MHz RFQ and power coupler during CW operation.

Table 2 Beam strips & availability requirement for ADS-based industrial power generation [1]

Although there are disadvantages to the choice of 325-MHz, we still selected the higher-frequency scheme.On the one hand, we obtained some experience based on the construction of the previous RFQ (or ‘‘973’’ RFQ: 3.5 MeV,352 MHz,[14]7%duty factor)built at IHEP several years ago. On the other hand, the successful operation of the 325-MHz injector can verify the feasibility of using a spoke-type cavity,not only for the low-energy accelerating part but also for the medium-energy part of the main linac,which is more important for the future China initiative Accelerator Driven System (CiADS) [2] constructions.More details on the Injector-I design can be found in[3, 15] and the references therein.

2.2 Why spoke cavity

The main design philosophy for the ADS driver linac is to use as many SC structures as possible. The major difficulty of a CW operation using a room-temperature structure is the large heat deposit on the cavity.If it cannot be effectively removed, the deposited heat will deform the cavity and even destroy the cavity components,leading to a degraded operating gradient and lower cavity efficiency.By contrast, the SC cavity is favored because the cavity loss is very small.This makes a CW operation feasible for high-intensity proton linacs. Meanwhile, the SC cavity is also favored for the independently phased feature, providing the possibility for SC cavities to realize local compensation [16] when the cavity fails during operation.This is very important for ADS accelerators to achieve the extremely strict reliability, as shown in Table 2. In addition,the SC cavity is also favored for much larger apertures than the room-temperature acceleration structure, which leads to largely reduced beam losses. This is a significant benefit for fulfilling the 1 W/m beam loss acceptance for a high-power CW linac.

Considering all of the advantages listed above, it is commonly understood that the SC proton linac is the best choice as an ADS driver accelerator[16,17],and the lower the input energy of the SC section, the easier it is for the realization of a CW operation. With well-approved SC radio frequency (RF) technology in recent decades, particularly the success of high-beta elliptical cavities at SNS[18, 19], the widely used low-beta QWR cavities [20, 21],the testing results of HWR cavities [22, 23], and most importantly, the test results of spoke cavities for medium and even low betas [24-26], it was thought that a proton linac with SC structures excluding the RFQs is possible.However,although the newly developed spoke-type cavity fills in the gap of the medium-energy part using the SC accelerating structure, and the test results also seem promising, no single cavity has yet been verified by the beam, let alone by a high-intensity CW proton beam.Therefore, a beta equal to 0.12 spoke-type cavity was chosen for Injector-I to verify the key technologies of the design, fabrication, and operation of the spoke cavities.

3 Beam commissioning at pulsed mode

Commissioning in pulsed mode is an essential step used to verify the hardware performance and physical design.The injector was commissioned according to the fabrication and installation phases. The first stage was for the ion source, LEBT, and RFQ test stand. After the beam performance of the RFQ was well understood, a test cryomodule(TCM)accommodating two single spoke cavities,two solenoids, and two beam position monitors (BPMs)were installed. The successful commissioning of the TCM verified the design and fabrication of the spoke cavity system. The tuning experiences of the independently phased resonators were obtained during this process. After TCM was dismantled, two new cryomodules (with seven spoke cavities in each)were installed and commissioned in succession. The beam parameters along the linac were carefully measured with a 10-mA pulsed beam to achieve proper optic settings for CW commissioning.

3.1 Ion source and LEBT

A schematic diagram of the ECR source and LEBT system is shown in Fig. 2. To have the flexibility of operating in both CW and pulsed modes, the front end needs to have the ability to supply an adjustable beam.The source was designed to provide a 35-keV proton beam(H+).The LEBT chopper was located at the entrance of the RFQ chopping the beam with a pulse width starting from 20 μs and a repetition frequency from 1 to 50 Hz.The rise and down times of the chopper were less than 20 ns. The first AC current transformer (ACCT) located between the chopping system and the RFQ was used for the beam current measurement. A maximum average current of 13 mA was delivered at the entrance of the RFQ.

The difficulty of the LEBT was to control the emittance growth and achieve a high transport efficiency,particularly for high-intensity beams. Thus, the LEBT was designed to be as short as possible with a total length of 1.67 m. An Alison detector installed in a moveable diagnostic bench was used for the LEBT emittance measurement. Twiss parameters close to the designed values at the entrance of the RFQ were achieved by adjusting the LEBT solenoid settings. Figure 3 shows the horizontal phase space of the beam 8.8-cm downstream of the LEBT exit.Table 3 shows the design and measurement of the beam current, Twiss parameters, and normalized root-mean-square (RMS)emittance. The measurement and design values agreed well, except for the position of the beam shift.

Fig.2 (Color online)Off-line layout of 2.45-GHz ECR proton source and LEBT [27]

3.2 RFQ accelerator

The 325-MHz RFQ bunches and accelerates the 35 keV beam from the ion source to 3.2 MeV. It is a 4-vane type copper structure RFQ with π mode stabilizers. The total length of the RFQ is 4.7 m. It is composed of two resonantly coupled physical segments, and each segment includes two technical modules connected with flanges. In total, four couplers are mounted on the RFQ and two couplers are mounted on one physical segment. A total of 64 tuners are plugged symmetrically on four sections for field flatness tuning and frequency adjustment. In total, 80 main cooling channels were designed for the RFQ vane and wall heat dissipation. There are also auxiliary cooling channels for the inner and outer conductors of the feedingforward power couplers(FPCs)and plug tuners.A detailed design can be found in [28]. The layout of the RFQ test stand is shown in Fig. 4.A movable diagnostic bench with two slits, one DC current transformer (DCCT), two fast current transformers(FCTs),four quadrupoles,four BPMs,and a simple beam dump line were mounted immediately after the RFQ for commissioning.

A pulsed proton beam with a peak current of 10 mA was used to verify the RFQ performance. To protect the power coupler,the transmission was measured for different cavity powers with a pulsed beam to determine the minimum power needed to obtain the top beam transmission of the RFQ. The power feeding in the RFQ cavity has to be sufficiently large to maintain a good beam transmission;otherwise, the cavity surface might be damaged by the beam lost inside the cavity, leading to a reduced cavity performance and service life. As shown in Fig. 5, the maximum transmission(at ～280 kW)achieved is 97%,in comparison with the design beam transmission of 98.7%,which fits well with the RFQGen [29] simulations. An output energy of 3.2 MeV was detected by two downstream FCTs using the time-of-flight (TOF) method with fine distance measurement with 0.05-mm alignment error. The distance between the two FCTs is 954.04 mm, and the expected time difference of the beam signals between two FCTs is 1.63 ns. The measurement result is 1.66 ns, as shown in Fig. 6, which shows that the RFQ performance meets the design specifications fairly well.

3.3 The MEBT and test cryomodule

The MEBT is used to match the RFQ and SC sections.As shown in Fig. 7 (and Fig. 4), it is composed of six quadruples, six steering magnets, and two bunchers. Beam diagnostic devices include six BPMs, two FCTs, one ACCT, and three wire scanners (WSs). Two FCTs with 1.67 m between them were used for the energy measurement. It should be noted that the distance between theFCTs must be sufficiently long to reduce the measurement errors.

Fig. 3 (Color online) Beam phase space at the measured location (8.8-cm drift downstream of the LEBT exit):Simulation (left) and measurement (right) results

Table 3 Measurement results versus the design goal of the beam current, Twiss parameters, and normalized RMS emittance at the exit of the LEBT

Fig. 4 (Color online) Layout of RFQ test stand

Fig.5 (Color online)RFQ transmission scan at different input power

Two downstream BPMs were used for the phase scan to determine the buncher settings. For phase scanning, it is necessary to use two BPMs not far away to reduce the measurement error caused by the largely expanded phase spread of the beam. For energy measurements, it is recommended to measure with buncher cavities for the same reason. Figure 8 shows the phase-scanning results for bunchers I and II. The bunchers were scanned at different voltages,and the data were fitted using a cosine curve.The phase with the maximum energy gain corresponds to zero degrees, and -90°was accordingly determined. Because the change in particle velocity during phase scanning of the bunchers can be neglected, the scanning results can be fitted very well using the cosine function. It is worth mentioning that the phase information of the first BPM right after RFQ was used to check the RFQ status after recovery from RF quenches. This method was extended to all downstream BPMs to check the cavity repeatability from any RF quenches.

Fig. 6 (Color online) Beam energy measurement for RFQ using two FCTs

Fig. 7 (Color online) Layout of TCM

To verify the performance of the hardware systems,such as cavity, solenoid, and cryogenic systems, the TCM (as also shown in Fig. 7) housing two periods was installed and commissioned before the formal cryomodules. The commissioning of the TCM turned out to be necessary for the physical design and technical validations of the formal cryomodules. The accelerating gradient achieved for the TCM spoke cavity was 3.1 MV/m with a maximum peak gradient of 14 MV/m. This is far from the design specification of 6.08 MV/m because of the field emissions caused by contamination of the cavity, preventing the accelerating gradient from further increasing. After commissioning the TCM, the power coupler structure was changed to protect the ceramic windows. The diameter of the coupler port and the length of the vacuum part were reduced to help shielding the field emission(FE)electrons.Cleanliness during the assembly was also important for ensuring the cavity high-gradient operation. The power coupler was assembled into the cavity in a class-10 clean room for the formal module. All improved technologies and methods were used for the succeeding cryomodules:CM1 and CM2. More details can be found in [30]. The cavity operating gradient during the following commissioning stages was much higher.

Fig. 8 (Color online) Phase scanning curves for Buncher I (left) and Buncher II (right)

3.4 The SC sections

The SC section consists of two cryomodules. Each cryomodule has seven periods. The first and second cryomodules accelerate the proton beam to 5 and 10 MeV,respectively. The periodic lattice was broken at the interface of the two cryomodules because needed space must be maintained for the cold-to-warm transition. Although the simulated results are promising under an ideal situation, a break of the periodic lattice easily leads to mismatches during commissioning because of an unavoidable input beam mismatch and accumulated static and dynamic errors of the upstream elements.

Although a long single cryomodule is preferred from a beam dynamics perspective and for a higher cavity efficiency, two cryomodules were selected because they are much easier to install, align, and maintain. To keep the space between the cryomodules as short as possible,diagnostic devices and vacuum pumps were omitted. For the SC section of Injector-I, the space from the end of the last cold element of the upstream cryomodule to the entrance of the first cold element of the downstream cryomodule was minimized to 570 mm because two layers of cold shielding are necessary for the CW operation of the SC section at 2 K.

After the TCM was commissioned, it was replaced by CM1 and CM2. Each cryomodule includes seven βg=0.12 SC spoke cavities, seven SC solenoids, and seven cold BPMs.Based on TCM operational experience,the pressure stability of the liquid helium (LHe) was improved from±0.6 to ±0.05 mbar (measured during CM1 commissioning)with optimized control and operation parameters of the cryogenic system [31].

For the βg =0.12 spoke cavity,as shown in Table 4,the df/dp was large at approximately -130 Hz/mbar for the CM1 cavities and-83 Hz/mbar for the CM2 cavities after the improvements were made.This was found to have been caused mainly by the deformation of the beam pipe,which could be solved by adding stiffeners to the helium vessel and applying stronger tungsten inert gas (TIG) welding[31]. Limited improvements were achieved for the CM2 cavities because they were already undergoing fabrications after the TCM commissioning. For the βg = 0.21 cavities installed on the 25-MeV test stand [32] (the last cryomodule with six βg = 0.21 spoke cavities) based on Injector-II in the IMP, the design was further improved,i.e., df/dp was controlled down to approximately 10 Hz/mbar, as shown in Table 5, which makes it possible to operate under a 4-K system. The tuner design was also improved for this cryomodule by eliminating loose gears and joints,based on the operating experiences of CM1 and CM2 [31]. It is worth mentioning that the 25-MeV test linac (with six βg = 0.21 spoke cavities) has also been commissioned with a 2-mA CW beam having a maximum of 14 h without trip and the accumulated CW operation time of 110 h [33].

The Lorentz detuning factor for the spoke cavity(β=0.12) is approximately 10 Hz/(MV/m)2, as shown in Fig. 10, which is relatively high. So does for the tuning sensitivity, which is approximately 1.1 MHz/mm. These features make the cavity sensitive to external interfaces,such as microphonics and mechanical vibrations.Although the cavity is still controllable, it increases the difficulty of LLRF control.However,for the development of such small gap spoke cavities with a frequency of 325 MHz, it is difficult to increase the stiffness of the structure. But, for higher beta spoke cavities, the sensitivity can be improved much better, which will largely reduce the difficulty of LLRF control. For example, for βg = 0.21 cavities, the tuning sensitivity was reduced to 600 kHz/mm,and for βg= 0.5, the double spoke cavity developed in IHEP is 107 kHz/mm (simulation value) [34]. Of course, as for microphonics isolating or damping from the vibration source (if it can be traced) is always more effective in improving the noise level. During the commissioning of Injector-I, the vibration source was traced to the scroll pump from the cart inside the tunnel. After removal,the noise level is reduced by one order of magnitude [31].

Table 4 Measurement df/dp of 14 SC β=0.12 spoke cavities in Injector-I [31]

Table 5 Measurement df/dp of six SC β=0.21 spoke cavities on 25-MeV test linac [31]

Fig. 9 (Color online) Two cryomodules of Injector-I installed in the tunnel

A beam energy of 6.05 MeV was obtained at the exit of CM1 with a beam peak current of 10.6 mA, a pulse width of 1 ms,and a repetition frequency of 2 Hz.The maximum cavity operating gradient (Eacc) exceeded 7 MV/m. CM2 was installed immediately after CM1. Figure 9 shows the test stand with both cryomodules installed in the tunnel.A beam energy of 10.67 MeV was obtained at the exit of CM2 with a peak current of 10.6 mA.The cavity operating gradients are listed in Table 6.For the first few cavities and the cavities in the middle, the operating gradients were relatively lower. The low gradient of the first few cavities originates from contamination by an accidental vacuum leak of the warm section upstream.The last two cavities in CM1 and the first two cavities in CM2 were used for longitudinal matching.The operating gradients of the other cavities were all beyond the design specifications and well compensated for the lower gradient cavities.

The beam transmission through the cryomodule was approximately 100%, as shown in Fig. 11, and the current was measured using two ACCTs,i.e.,ACCT2(upstream of CM1, purple curve) and ACCT3 (downstream of CM2,yellow curve). The beam energy divergence at the exit of the cryomodules was measured by the EDA system,which includes two slits,one 90°magnet,and one Faraday cup,as shown in Fig. 1. The energy divergence was measured by scanning the fields of the upstream magnet after inserting two slits and obtaining signals from the Faraday cup (behind the second slit). Figure 12 shows the Faraday cup current signals versus the magnetic fields for a series of bunches. As shown in this figure, the energy divergences are almost the same for all bunches. The different magnet currents correspond to different beam energies. In Fig. 13,the blue, green, and red curves show the energy evolution for a 20 μs beam at the head,tail,and middle of the bunch.They did not show different energy divergences. The maximum energy difference for the beam with a 20 μs pulse width is 0.06 MeV. The RMS value for 10.67 MeV with a 10.6-mA peak current proton beam is 3.2%, which agrees with the simulation (RMS, 2.8%).

Fig.10 (Color online)Lorentz detuning factor of 14 spoke cavities in Injector-I [31]

Table 6 Operating gradients of 14 SC β=0.12 spoke cavities in Injector-I [31]

Fig. 11 (Color online) ACCT2(purple curve) and ACCT3(yellow curve) signals

Fig.12 (Color online)Current signals on faraday cup of EDA system versus magnet fields for a series of bunches at the exit of Injector-I

3.5 Transverse emittance measurement

Beam loss control is a crucial issue in the commissioning of a high-power proton linac.A mismatch is one of the major sources of beam loss, and thus the Twiss parameters and emittance measurement must be as accurate as possible. In our case, quadruple scan and double slit methods were used on the MEBT section and beam dump line, respectively.

As is well known, quadruple scan data with beam sizes crossing the waist could yield a much smaller calculation error;thus,the strengths of two upstream quadrupoles were scanned independently to obtain valid data for the horizontal and vertical emittance calculations,respectively.For an error reduction,two bunchers were turned off during the measurements.

For the quadruple scan, the first problem is the beam size fitting during data processing.A Gaussian curve fitting is typically used for calculating the RMS beam size.However, in our case, because the quadruple setting deviated significantly from the nominal design during the gradient scans,most of the measurement signals deviated fromthe standard Gaussian distribution, as shown in Fig. 14.The direct RMS formula, Gaussian fitting, and generalized Gaussian [35] fitting methods have been applied and compared for an RMS beam size calculation[36].Here,the generalized Gaussian fitting method was used to fit the beam sizes because it can better characterize the RMS size of the non-Gaussian distribution signal compared to the Gaussian fitting method (as shown in Fig. 14), and is less sensitive to the background signal compared to the direct RMS formula calculation method.

Fig. 13 (Color online) Beam energy divergence at the exit of Injector-I

Fig.14 (Color online)Fitting results for three typical beam signal distributions with generalized Gaussian(ggGaussian)and standard Gaussian distributions

After obtaining the beam sizes, to include the space charge effect, the multi-objective genetic algorithm(MOGA) [37, 38] was used to fit the beam parameters by calling the multi-particle tracking code TraceWin [39]. As shown in Fig. 15, the MOGA optimization and measurement results agreed well with each other. The MOGAoptimized beam parameters are listed in Table 7. We can see that, for the measurement results, the horizontal emittance is approximately 20%smaller than that of the vertical plane, whereas roughly equivalent emittances were expected for the two planes. This is possibly due to the unideal beam injecting the RFQ.

The double slit method was used for the transverse beam parameter measurement out of the 5-MeV test stand. Two issues are critical with this method. One is the number of measurement grids, and the other is the threshold of themeasured signals. Although measurements using more grids are favored owing to their more accurate results,they take more time to conduct. Simulations show that the Twiss parameters converge when the number of grids increases to up to 64 times 64 [40]. Therefore, 64 was defined as the number of slices for the measurement. The second is determination of the background threshold.In our case, a new method is proposed to determine the background threshold [40]: As shown in Fig. 16, the second derivative of the emittance (ε)and Twiss parameter(α and β) is obtained as a function of the threshold, and when the value converges to zero,it is considered to be the threshold of the measurement data. Figure 17 (upper figure) shows the measurement phase space distributions at the exit of the 5-MeV test stand using the double-slit method, and the simulated phase space distributions at the same position(lower figure) are also presented for comparison. During the simulation,the input parameters(at the RFQ exit)were updated with the measurement data using the quadruple scan method, as previously mentioned. Both the simulated and measured Twiss parameters and emittances are presented in Table 8 for comparison.

Fig. 15 MOGA optimization results with space charge effects included and a comparison with the measurement results

Table 7 Beam parameters at RFQ exit using MOGA optimization method with space charge included

Fig. 16 (Color online) Emittance, beta function, alpha (left), and second derivatives (right) as functions of the signal threshold

4 Beam commissioning in CW mode

We accumulated significant experiences and collected a large amount of data during the commissioning of Injector-I in pulsed mode. However, CW commissioning is more difficult. As mentioned previously, the most difficult parts were the RFQ and spoke cavities. Although the RFQ and SC cavity could operate stably with full power without a beam or with a small duty factor (e.g.,＜1%) beam, they tripped more frequently once the beam duty factor approached the CW.

Many efforts have been made to keep the machine stable in CW mode. For the RFQ, cooling is critical.Several accidents during commissioning have led to the development of RFQ components.For the spoke cavity,the challenge has been the frequency stability and anti-interference capability for the smallest (βg = 0.12) spoke-type cavity used around in the world. According to the TCM operating experiences, the cavity, coupler structures, and cryogenic system were improved to enhance the stability of the system. In addition, we carefully matched the optics,used a metering probe along the SC section for field emission surveys, attached temperature probes around the small aperture beam pipe to align the beam getting through the step field magnets (SFMs), and monitored the vacuum status at the dump line.The stable CW operation time went from dozen of seconds up to several minutes, and finally more than 20 min with an output energy of 10 MeV and an average beam current of ～2 mA. Details are presented in this section.

4.1 RFQ

Several problems occurred during RF conditioning toward a CW, leading to an improvement of several RFQ components and assemblies.A vacuum leakage occurred at the welding port, leading to an upgrade to simplified cooling water channels at the end plates and the coupling plate [41]. An arc accompanied by a very low vacuum pressure at the coupler position turned out to be a burned RF contact spring [42]. We learned from this that the fingers of the loose contact spring could easily cause a discharge under high-power conditioning. Therefore, the RF contact spring was canceled in the new couplers using a copper vacuum flange gasket instead [41, 42]. However, it was retained for the plug tuners of the RFQ. Although a CW operation has been achieved for the RFQ with an average beam current of approximately 2 mA,the spring is still considered to be a weak point affecting the stability operation of the RFQ,which is not an advisable choice for high-power CW machines. Later, the coupler ceramic window cracked because the inner and outer conductors were shortened by the condensate water. The problem was solved by reducing the humidity of the tunnel and increasing the temperature of the cooling water for the coupler [42]. After fixing each of these problems, 90%transmission was achieved for the RFQ with a 90% beam duty factor after shooting in an 11-mA proton beam.

The CW commissioning of Injector-I was started with a small average beam current (～2 mA), achieving a low beam loading effect because the LLRF feed forward control and frequency control were lacking.To generate such a low beam current, a small aperture, as shown in Fig. 18,was installed right after the first LEBT solenoid, and the wire scanner was replaced, as shown in Fig. 2. However,the measured maximum transmission of the RFQ was 90%,as shown in Fig. 19 (left), which was not as good as expected, the main reason we suspect is that the smallaperture was not at the center of the beam.It is known from the simulation that a 0.5-mm off-center displacement at the entrance of the RFQ will lead to a decrease in transmission from 98.7% to 92.7% (as shown in Fig. 19, right).

Table 8 Beam parameters at CM1 exit

Fig. 18 (Color online) Small aperture in the LEBT section

Fig. 19 (Color online) RFQ transmission scan @2mA (left) and RFQ transmission with transverse displacement (right)[43]

To achieve CW operation of the RFQ,one critical aspect was to use reflecting power interlock protection.Reflecting power was used to determine the arc at the initial stage.Once an arc signal was sent out, the feed in power was immediately cut. The power was then recovered after a short period (e.g., several hundred microseconds) to prevent the cavity from cooling down(without feed in power).The power stop interval was not fixed but was determined by the status of the arc. When a succeeding arc occurred,the stop period was slightly longer. In this way, the cavity and couplers were effectively protected, and the RFQ recovery efficiency remarkably increased.

Another critical aspect is the use of the self-excitation function. During commissioning, the operated cavity frequency of the RFQ was fixed by controlling the temperature difference of the cooling water between the cavity vane and wall. However, for CW operation with and without full power,there was a frequency difference of 111 kHz. During the power feeding process, the self-excitation function allows the changing of generator frequency for tracing the RFQ cavity frequency automatically, which greatly increases the efficiency of the field establishment once the cavity needs to be restarted.

4.2 SC cavity section

Fig. 20 (Color online) Power density distribution along the SC linac with 2 mA @2.5 ms

During the commissioning of the SC section,two issues repeatedly appeared.First,the solenoid and corrector at the back end of the linac are frequently tripped. We suspected that this might be caused by the beam loss owing to a longitudinal mismatch; for a transverse mismatch, the beam envelope should be oscillating along the linac such that the beam losses will not concentrate in a single location. However, for a longitudinal mismatch, the transverse beam envelope increases in size gradually with accumulated errors, and beam loss is very likely at the end of the linac. Therefore, the strengths of the 12th and 13th solenoids were increased to control the beam size behind,apparently improving the situation.

For a longitudinal mismatch, the beam loading effect is a major concern.The TraceWin code was used to study the beam loading effect along with Lorentz detuning during the transient process[44].The simulation starts with a nominal design without errors after setting up the design values for all optics. LLRF control was removed to better understand the mechanism. Thus, although the time of the beam loss was postponed during the simulation, the tendency should be the same. Figure 20 shows the beam power density distributions at 2.5 ms with an average beam current of 2 mA. The results show that the beam is mismatched longitudinally from the very beginning of the SC linac,some of which easily rotated out of the longitudinal bucket with optical errors accumulating along the linac, and is finally lost at the back-end of the linac after coupling to the transverse plane. Notably, the mismatch was quickly magnified after the beam passed through the matching part of two SC sections where the periodical structure was interrupted (although the situation could be improved by increasing the longitudinal acceptance of the injector, the cavity gradient has to be accordingly increased).The losses gradually increased along the linac, and the most serious loss occurred at the very end,as shown in Fig. 21,which is consistent with the experimental results. Later, the strengths of the 12th and 13th solenoids increased from 170 to 180 A,and the trip frequency of the solenoids decreased significantly. Meanwhile, as shown in Fig. 22, the beam loss during the simulation was also greatly improved after changing the settings of these two solenoids. This was treated as an explanation for the experimental phenomenon that occurred.

Another issue that frequently occurred during the commissioning of the SC cavity section was the deterioration ofthe vacuum status at the beam dump line once the beam duty factor approached the CW, and the interlock system was frequently triggered to stop the ion source.It was noted that once the beam size on the beam dump was reduced,the vacuum status worsened.This verifies that the phenomenon originated from the outgassing of the beam dump target when the high-intensity beam was shot. The problem was mitigated by slowly increasing the beam duty factor and maintaining the vacuum status in the beam dump line under the interlock threshold. However, it would be better if one more vacuum pump could be installed in this section and assist in quickly pumping the vacuum.

Fig.21 (Color online)RMS phase spread evolution along the Injector-I SC section with 2-mA average beam current and the beam losses along the linac @2.5 ms

4.3 Beam performance in CW mode

Fig. 23 (Color online) CW commissioning results of Injector-I [45]

Fig.24 (Color online)Transmission during CW operation at 10 MeV,DCCT signal on LEBT DCCT at 1.206×2=2.412 mA(left figure),and DCCT at the exit of SC section at 2.128 mA (right figure)

Numerous efforts have been made to maintain the stability of the machine when shooting a CW beam. The stable operation time ranged from dozen of seconds up to several minutes, and finally, more than 20 min of stable beam time was obtained with an output energy of 10 MeV and an average beam current of ～2 mA using only LLRF feedback control (in which the feed-forward and frequency controls are absent). The maximum stable operation time reached was 23 min with an output energy of 10 MeV and an average CW proton beam current of 1.6 mA.Figure 23 shows the output energy at the exit of the RFQ/cryomodule and the beam current information. Although the RFQ could deliver a higher CW beam current, the SC section could not handle higher intensity beams; in addition,the beam loading effect was not compensated,and the frequency control loop was absent.The transmission of the CW beam from the LEBT to the end of the linac was 88.2%, as shown in Fig. 24, with an input beam current of 2.412 mA at the LEBT section and an output beam current at the exit of the SC section of 2.128 mA.Unfortunately,a few days later, the CW commissioning was stopped by a breakdown of the RFQ klystron, and we have since been unable to continue any further.

For the operating gradient during CW commissioning,all cavities fulfilled the design specifications (an accelerating gradient of no smaller than 6.08 MV/m)except those near room temperature,and the average peak gradient(Ep)was approximately 27 MV/m [31]. The first four cavities were contaminated by an accidental vacuum leakage at the MEBT section, which led to degraded operating gradients.If the first four cavities are removed, the average Ep is increased to ～30 MV/m. This is roughly half of the gradient reached compared with the vertical test, or even lower. As a good reference to this result, for higher beta spoke cavities, even if a conservative design gradient with a peak gradient of 27 MV/m is used, the accelerating gradient(Eacc)can still be much higher than small β spoke cavities with a smaller Ep/Eacc (for example,Ep/Eacc is 3.4 for a beta equal to a 0.5 spoke cavity,whereas Ep/Eacc is approximately 4.5, for beta equal to a spoke cavity of 0.12).This can serve as a reference for future linac designs with higher beta spoke cavities.

5 Summary and outlook

The China ADS injector-I testing facility was commissioned using pulsed and CW beams.A proton beam with an energy level of over 10 MeV and a peak beam current of more than 10 mA was obtained after accelerating through a room-temperature 3.2-MeV RFQ and 14 βg = 0.12 single spoke cavities, which fulfill the design expectations very well.A CW beam with an energy of 10 MeV was also shot through to check the RFQ, and particularly the spoke cavity performance, and an average beam current of approximately 2 mA was successfully obtained at the exit of the linac with LLRF feedback without feed-forward and frequency controls. The small gap gives the βg = 0.12 single spoke cavity a sensitive feature,making it difficult to sufficiently reinforce the structure. This leads to a higher tuning sensitivity and higher Lorentz detuning factor of the cavities, which makes them more sensitive to external interfaces and larger frequency detuning during the transient beam loading process.The situation is much better for βg =0.21 spoke cavities installed in the 25-MeV test stand,and the tuning sensitivity is reduced by almost half.For the βg = 0.5 double spoke cavity developed in the IHEP, the tuning sensitivity is even smaller, which is one order of magnitude smaller compared with the βg = 0.12 single spoke cavity.

Overall,for higher beta spoke cavities,the cavity would be less sensitive, which caused the difficulty of LLRF control. For higher beta spoke cavities, the operating gradients could be much higher because Ep/Eacc and Bp/Eacc are much smaller than the low β spoke cavities.

Preliminary experimental results obtained on the basis of spoke type cavities are encouraging. Applying a 325-MHz spoke-type cavity was proven to be a feasible approach to accelerate proton beams in medium and even low β sections. It was also proved that a proton linac with superconducting accelerating structures, excluding the RFQs, is possible, as is the realization of a high-intensity CW proton linac.

AcknowledgementsThe authors want to thank all the ADS teams in IHEP for their support,especially the beam dynamics group,for their comments,suggestions,and discussions.Special thanks are expressed to Professor Yuan Zhang in IHEP for organizing discussions and reviewing this paper, and also for Zheqiao Geng from PSI for useful discussions on LLRF control and beam loading issues.

Author ContributionsAll authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Fang Yan, Hui-Ping Geng, Cai Meng and Ya-Liang Zhao.The first draft of the manuscript was written by Fang Yan and all authors commented on previous versions of the manuscript.All authors read and approved the final manuscript.