Fabrication and RF test of the 500 MHz-RFQ linear accelerator for a transportable neutron source RANS-III

Abstract

At RIKEN, a transportable accelerator-driven compact neutron source (RANS-III) is under development for an on-site nondestructive inspection of the degradation of old concrete and reinforcing steel. RANS-III consists of an ion source, a low-energy beam transport, a radio frequency quadrupole linear accelerator (RFQ linac), a radio frequency (RF) system, a high-energy beam transport, a target station and a neutron measurement system. Because the inner diameter of the RFQ linac is inversely proportional to the resonance frequency, the resonance frequency of the RANS-III RFQ linac in this study was chosen to be 500 MHz, which is 2.5 times that of the RANS-II RFQ linac. Therefore, the inner diameter and weight of the RANS-III RFQ linac were reduced to approximately half and one third, respectively, of those of the RANS-II RFQ linac. The RANS-III RFQ linac was designed to accelerate a proton beam with a 10 mA peak current and 100 μA average beam current from 30 keV to 2.49 MeV (Journal of Disaster Research12(3) (2017) 585–592). Based on the evaluations, an RFQ linac for RANS-III was fabricated, and the RF characteristics of the cavity, such as the resonant frequency and electric-field distribution, were measured using a low-power test and tuned using fixed tuners. In addition, RF couplers and RF systems were constructed to inject RF power into the RANS-III RFQ linac, and RF input tests were performed.

Keywords

Transportable compact neutron source RFQ linac proton accelerator

1. Introduction

Neutron beams are used as a material analysis tool for structural inspection and metallographic analysis because of their high penetrating power and ability to identify light elements. Neutron beams are usually produced by research reactors or large accelerators. On the other hand, compact accelerator-driven neutron sources, which can be installed and used in a single laboratory, have been developed in recent years.

At RIKEN, the RIKEN accelerator-driven compact neutron source (RANS) [7], which consists of a 7 MeV proton linear accelerator (linac) and a target station with a beryllium target, was installed in 2013. Measurement techniques have been developed for the inspection of infrastructures such as bridges. For example, a neutron scattering imaging technique was developed to detect water and voids in concrete, as well as to measure the salt content in reinforced concrete using prompt γ-ray analysis [4,10]. In addition, RANS-II, which consists of a 200 MHz radio frequency quadrupole (RFQ) accelerator and a target station with a lithium target, has been developed as a prototype of a portable small neutron source, and neutrons have been generated successfully through the ⁷Li(p, n)⁷Be reaction [5].

Fig. 1.

RIKEN transportable compact neutron source (RANS-III).

Table 1

Cavity parameters of the 500 MHz RFQ linac for RANS-III

Resonant frequency [MHz]	500
Input beam energy [keV]	30
Output beam energy [MeV]	2.49
Peak beam current [mA]	≦10.2
Input emittance (RMS, nom) [ $π mm \cdot mrad$ ]	0.1
Twiss parameter α of the matched beam at the RFQ entrance	0.61
Twiss parameter β of the matched beam at the RFQ entrance vane [mm/mrad]	3.0
Max duty cycle [%]	3.0
Average beam current [μA]	≦100
Power loss [kW, 100% Q]	183.3
Unloaded Q	8730
Cavity length [mm]	2370
Cavity outer diameter [mm]	∼300
Cavity weight [t]	0.7

In addition to RANS and RANS-II, RANS-III – a transportable accelerator-driven compact neutron source (see Fig. 1) – is being developed. Because RANS-III is to be installed in an automobile, its footprint and weight are limited. Therefore, a lithium target and an RFQ linac were adopted to design the RANS-III with the same neutron intensity as RANS-II, but with more compactness. The resonant frequency of the RFQ linac is inversely proportional to the inner diameter of the cavity. Based on this property, an RFQ linac with a resonant frequency of 500 MHz (2.5 times that of RANS-II) was designed for RANS-III. Thus, the inner diameter and weight of the RANS-III RFQ linac were reduced to approximately half and one third, respectively, of those of the RANS-II [3]. Design parameters of the 500 MHz RFQ linac for RANS-III are listed in Table 1.

In this study, a three-body RFQ accelerator – with two minor vanes and one major vane that were machined from copper blocks and assembled by bolting – was adopted. The resonant frequency and electric-field distribution of the RANS-III RFQ linac were measured using a vector network analyzer and were adjusted using fixed tuners.

An RF system was constructed to inject RF power into the RANS-III RFQ linac. The RF system consisted of four 75 kW semiconductor amplifiers, a 77D coaxial tube, and RF couplers connected to the cavity, allowing the system to feed 300 kW of RF power to the cavity. The RF coupler was reduced from 77D to 39D in the vacuum region so that the RF power output from the semiconductor amplifier could be fed into the RANS-III RFQ linac without discharged through a coupler port with the size of a 39D. The RF system was connected to the RANS-III RFQ linac, and RF conditioning was performed.

2. Fabrication and RF property measurement of the 500 MHz RFQ linac

Based on the design, a 500 MHz RFQ cavity for RANS-III was fabricated. As fabrication method, the 3-pieces RFQ fabrication method was adopted. In this method, the 3-pieces RFQ consists of one major vane and two minor vanes. Oxygen Free Copper blocks were machined to the shape of the minor vane and major vane by using a computer numerical control (CNC) milling machine. The RFQ cavity is assembled by putting the major vane between the two minor vanes and bolting them together (Fig. 2(a)) [8]. This structure is rigid and suitable for use as an RFQ linac for RANS-III.

Fig. 2.

Fabrication of the 500 MHz RFQ cavity.

The resonant frequency of the fabricated cavity was measured using a vector network analyzer. The frequencies of resonance modes are listed in Table 2. From the simulation results obtained using CST MICROWAVE STUDIO [1], the acceleration mode of the RFQ linac is Mode 5 (TE211 mode), and the resonance frequency is 502.09 MHz. The measured resonant frequency was 495.36 MHz and 6.73 MHz lower than the simulation value. This might be due to the increase in the capacitance of the cavity because the minor vanes were closer to the beam axis owing to assembly errors.

Next, the electric field strength distribution of the cavity was measured using the perturbation method, as shown in Fig. 3. Polytetrafluoroethylene ( $8 \times 8 \times 30 mm$ ) was used as the perturbator, and the perturbator was attached to a polyester thread with a diameter of 0.33 mm. The perturbator was placed between the RFQ electrodes in each quadrant, and the perturbator was moved inside the cavity. The electric-field strength distribution in the RFQ linac can be obtained by measuring the resonant frequency at each position of the perturbator with a vector network analyzer. The measurement position of the electric-field strength distribution in the third and fourth quadrants tended to shift owing to the weight of the perturbator. Therefore, the effect of misalignment due to the weight of the perturbator was evaluated by measuring the electric field strength distribution in the third and fourth quadrants with the cavity turned upside down. To reduce the misalignment of the perturbation, tension was harnessed to the thread by applying weight. Without the weight, the electric-field strength distribution in the third and fourth quadrants was at most 10% lower than the electric-field strength measured by inverting the acceleration cavity due to misalignment caused by the weight of the perturbator. By attaching a 4 kg weight to the thread, the misalignment due to the weight of the perturbator was compensated. Therefore, the electric-field strength in the third and fourth quadrants was as small as 3% difference from the electric-field strength measured by inverting the cavity.

Figure 4(a) shows the electric-field strength distribution of the RFQ cavity (without tuners and couplers) measured by the perturbation method. The dipole components are calculated $D1 = (Q1 - Q3) / 2$ and $D2 = (Q2 - Q4) / 2$ . The maximum difference in electric-field strength distributions in Q1 and Q2 with those in Q3 and Q4 is approximately 50%. It is considered that the position of the lower RFQ electrode is slightly closer to the beam axis direction than the design value because the measured resonance frequency is lower than the calculated value.

The resonant frequency and electric-field strength distribution were adjusted using the eight tuners (Fig. 5). First, the tuners were long enough to be inserted as far inside the inner wall of the RFQ linac as possible (35 mm from the cavity wall) to increase the resonant frequency. Figure 4(b) shows the electric-field strength distribution in the RFQ cavity when the tuners and RF couplers were installed. The electric field in the upper two quadrants (first and second quadrants) is weaker, whereas the electric field in the lower two quadrants (third and fourth quadrants) is stronger. The resonant frequency was 500.4 MHz. The electric-field in the quadrant where the tuner was inserted tended to be higher than the electric-field in the adjacent quadrants, whereas the electric-field in the opposite quadrant tended to be the lowest. Therefore, the fixed tuners were installed in all eight ports in the third and fourth quadrants.

Table 2

Resonance frequencies of each resonance mode

	Measurement value [MHz]	Simulation value [MHz]
Mode1 (TE111)	474.58	485.36
Mode2 (TE111)	481.08	485.52
Mode3 (TE112)	484.78	493.16
Mode4 (TE112)	489.32	493.59
Mode5 (TE211)	495.36	502.09
Mode6 (TE212)	499.82	506.08

Fig. 3.

Layout of the perturbation method.

Fig. 4.

Electric-field strength distribution of the RFQ cavity.

This result indicates that the variation in the electric-field strength distribution in each quadrant can be minimized by adjusting the length of each tuner. Therefore, dummy tuners were used (Fig. 6) to measure the electric-field strength distribution while changing the tuner length in steps of 5 mm to determine the tuner condition that minimizes the variation in the electric-field strength. Based on the measurement results of the electric-field strength distribution using dummy tuners, the optimum tuner inserted length from the inner wall of the RFQ cavity was obtained, as shown in Table 3, and the electric-field strength distribution was measured after the adjustment. The electric-field strength distribution after tuning is shown in Fig. 4(c)). The variation in the electric field in each quadrant is reduced to better than 20%. The resonant frequency of the acceleration cavity is 498.80 MHz, and the unloaded quality factor Q is 6364, which is approximately 73% of the design value.

Fig. 5.

Tuner insertion position.

Fig. 6.

Dummy tuners.

Table 3

Port conditions and the optimum tuner inserted length from the inner wall of the RFQ cavity

Position	Q1	Q2	Q3	Q4
A	TMP	Pick-up	Tuner (13 mm)	Tuner (8 mm)
B	Coupler	Coupler	Tuner (35 mm)	Tuner (35 mm)
C	Coupler	Coupler	Tuner (30 mm)	Tuner (35 mm)
D	Pick-up	TMP	Tuner (13 mm)	Tuner (8 mm)

3. High power test with the four set RF system

3.1. Configuration of the RF system for the 500 MHz RFQ linac

To accelerate proton beam in the 500 MHz RFQ linac for RANS-III, it is necessary to inject 250–300 kW of RF power into the RFQ cavity. However, because the coupler port of this cavity is small (diameter 46 mm), the voltage near the coupler port might become high, causing discharge. Therefore, four RF systems with 75 kW of RF power per system were used to provide 300 kW of RF power stably while reducing the voltage near the coupler port.

Fig. 7.

Schematic diagram of the high-frequency system.

The configuration of the RF system of the RFQ linac for RANS-III is shown in Fig. 7. Four sets of RF systems were connected in parallel in the RFQ cavity, each of these consisting of a semiconductor amplifier (AMP), 77D coaxial tube, directional coupler, and RF coupler. A semiconductor amplifier (CA500BW2-8085RP, R&K [2]) was used in the high-frequency amplifier. The dimensions of each semiconductor amplifier were $560 \times 1300 \times 1510 mm$ . Inside the semiconductor amplifier, 24 modules of the final amplifier were connected in parallel. The RF power output from each module was integrated in a radial combiner to amplify and output 500 MHz of RF power with up to 75 kW per unit. In the semiconductor amplifier, circulators and dummy loads were installed between the radial combiner and amplifier module to protect each amplifier module from the reflected waves on the load side.

In the low-level radio frequency (LLRF) system, the RF, timing, and sampling/hold signals that are output from a signal generator and a pulse generator are branched into four, and the branched RF signals are input into each semiconductor amplifier after adjusting their frequency, amplitude, and phase.

The LLRF detects changes in resonance frequency due to thermal deformation of the RFQ cavity from the pickup and directional coupler, and adjusts the operating frequency of the RF signal to the resonance frequency of the cavity.

3.2. Development of the RF coupler

To feed the RF power from the semiconductor amplifier into the 500 MHz RFQ cavity efficiently, it is necessary to install an RF coupler between the coaxial tube and the RFQ cavity to match the impedance. The shape of the RF coupler depends on the resonant mode, shape of the cavity, and mutual inductance of the coupler. In the RFQ linac for RANS-III, the coupler port diameter is 46 mm owing to the downsizing of the RFQ cavity, and a reduction from 77D to 39D is necessary at the coupler.

A cross-sectional view of the RF coupler is shown in Fig. 8. It consists of a 77D coaxial tube, RF window, 77D–39D reducer, 39D coaxial tube, and loop electrode from the RF amplifier side to the cavity side. The RF window is an alumina disk with a chromium nitride coating on its surface to suppress the secondary electron emission. On the vacuum side from the RF window, the size of the coaxial tube is reduced from 77D to the equivalent of 39D, which can be connected to the RFQ cavity. Alumina disks are inserted into a 39D coaxial tube to support the inner conductor. The alumina disk has an outer diameter of 40.8 mm, an inner diameter of 6.5 mm, and four holes each of 6 mm diameter for vacuum evacuation. To prevent the falling of inner conductor parts when moving in a vehicle, the outer conductor has a 2 mm step where the alumina disk is installed, and the inner conductor parts are supported by the outer conductor and alumina disk. The loop electrode is a 5-mm-thick and 10-mm-wide copper wire, bent into a loop, connected to the outer conductor by brazing, and connected to the inner conductor by bolting. The length of the loop electrode is 62 mm, width is 31.3 mm, and cross-sectional area in the loop is 1560 mm².

Fig. 8.

Cross-sectional view of the RF coupler.

The dimensions of the RF coupler were designed using the 3D electromagnetic field simulation software CST MICROWAVE STUDIO. The dimensions a–c in Fig. 8 are adjusted so that the S₁₁ parameter between the RF coupler inlet and loop electrode was less than $- 40 dB$ and the total length of the reducer part was as short as possible. As a result of the design, the distance from the RF window to the beginning of the loop electrode was 92 mm, and the S₁₁ parameter was $- 53 dB$ . The high-frequency couplers fabricated based on this design are shown in Fig. 9.

Fig. 9.

Fabricated RF couplers.

Four couplers were installed in the RFQ cavity, and the coupling factors of each coupler was adjusted. Owing to the layout of the tuner, vacuum pump, etc., two of the four couplers were installed in the first quadrant and others were installed in the second quadrant. When the forward power from each semiconductor amplifier is $P_{f}$ , reflected power is $P_{r}$ , voltage reflection coefficient is Γ, and total coupling factor of each port is $β_{sum}$ , the relationship between these parameters is given in Equations (1) and (2). To reduce the reflected power, the coupling factor of each coupler must be adjusted so that the total coupling factor is about 1 [6,9]. $\begin{array}{l} (1) & P_{r} = Γ^{2} P_{f} \\ (2) & Γ = (β_{sum} - 1) / (β_{sum} + 1) \end{array}$

First, the unloaded Q ( $Q_{0_i^{'}}$ ) and loaded Q ( $Q_{L_i^{'}}$ ) of each coupler when all couplers are installed are obtained from the measurement results of S₁₁. However, the other couplers are terminated to 50 Ω, and $Q_{0_i^{'}}$ is the value when the cavity and the three other couplers are installed. From the measurement results, the external Q ( $Q_{ext_i}$ ) of each coupler and the total external Q ( $Q_{ext}$ ) are obtained from Equations (3) and (4). $\begin{array}{l} (3) & Q_{ext_i} = {(\frac{1}{Q_{L_i^{'}}} - \frac{1}{Q_{0_i^{'}}})}^{- 1} \\ (4) & Q_{ext} = {(\frac{1}{Q_{ext_1}} + \frac{1}{Q_{ext_2}} + \frac{1}{Q_{ext_3}} + \frac{1}{Q_{ext_4}})}^{- 1} \end{array}$

From the calculated $Q_{ext}$ and the $Q_{L_i}$ of each coupler, the unloaded Q value ( $Q_{0_i}$ ) of the cavity can be obtained from the measurement of each coupler, and the coupling factor $β_{i}$ of each coupler can be obtained from the following equation. $\begin{array}{l} (5) & Q_{0_i} = {(\frac{1}{Q_{L_i^{'}}} - \frac{1}{Q_{ext}})}^{- 1} \\ (6) & β_{i} = \frac{Q_{0_i}}{Q_{ext}} \end{array}$

Using the above procedure, the couplers were adjusted. The coupling factor of each coupler was adjusted by rotating the coupler to change the cross-sectional area of the loop electrode through which the magnetic field passes. The coupling factor and Q-value are listed in Table 4. Because the measured unloaded Q-value was 73% of the design value, the nominal wall loss of the RFQ cavity was approximately 250 kW, and the beam loading was approximately 25 kW for acceleration of 10 mA proton beam. The total coupling factor was calculated to be 1.10 considering the beam loading. However, the measured $β_{sum}$ after the adjustment was 1.04. The coupling factors of each coupler were well balanced with variations within the range of $\pm 0.01$ .

Table 4

Coupling factor (β) and Q-value of each coupler and the cavity

	No. 1	No. 2	No. 3	No. 4
$Q_{0_i^{'}}$	24588	24336	25021	23890
$Q_{L_i}$	3079	3117	3157	3118
$Q_{ext_i}$	24588	24336	25021	23890
$Q_{ext}$	6113
$Q_{0_i}$	6204	6362	6528	6362
$Q_{0}$	6364 (73% of design value)
$β_{i}$	0.26	0.26	0.25	0.27
$β_{sum}$	1.04

3.3. RF conditioning

After impedance matching of the RF coupler, the RF conditioning of the RFQ linac was carried out by connecting a 77D coaxial tube and a semiconductor amplifier. First, the phase difference and operating frequency of each RF signal were adjusted using the LLRF system to minimize the amplitude of each reflected waveform with the same shape, while evaluating the reflected waveform of each semiconductor amplifier at 1 kW per unit with a pulse width of 100 μsec and repetition rate of 10 Hz.

Fig. 10.

Pickup voltage relative to RF power (maximum value normalized to 1).

Fig. 11.

Each RF signal with 300 kW RF power (the pulse width: 100 μsec, the repetition rate: 10 Hz).

After adjusting the phase difference and operating frequency of each RF signal, the RF power was increased at a pulse width of 300 μsec and repetition rate of 10 Hz, carefully monitoring the vacuum, outer wall temperature, and continuous discharge to degas gases such as H₂O, O₂, and N₂ on the inner wall of the cavity. Pickup voltage relative to RF power is shown in Figure 10. Maximum value is normalized to 1. When the RF power was more than 40 kW per amplifier (denoted as 40 kW/AMP), discharge occurred frequently in the acceleration cavity, so the RF power was maintained at 40 kW/AMP for 15 h. After conditioning at 40 kW RF power/AMP, the discharge frequency decreased sufficiently.

Thereafter, the RF power was increased from 40 to 75 kW/AMP with a pulse width of 100 μsec and a repetition rate of 10 Hz. Figure 11 shows the pickup voltage of the acceleration cavity and the reflected power waveform for each amplifier at a total RF input power of 300 kW. The vacuum level was $2.0 \times 10^{- 5} Pa$ at this point. The reflected power waveform was similar, and the reflected power was no longer small, which means that the impedance of each coupler matched.

4. Summary and future research

A 500 MHz RFQ linac system for the transportable accelerator-based neutron source RANS-III was developed. First, the RFQ cavity was fabricated based on the design evaluations, and the resonant frequency, the unloaded quality factor Q, and electric-field strength distribution were measured by the low-power test. Based on the measurement results, tuning was performed with the fixed tuners, which have a fixed length and do not move during accelerator operation, to obtain a resonant frequency and electric-field strength distribution close to the design values. The tuning results show that the resonant frequency was 498.8 MHz, the deviation of the relative electric field distribution in the TE211 mode was within 20%, and the unloaded quality factor was 73% of the design value.

To inject RF power into the cavity, RF couplers were developed, and a four-line RF system was constructed. After RF conditioning, it was possible to inject 100 μsec of RF power at 300 kW from the four-line RF systems into the cavity with a repetition frequency of 10 Hz.

In the future, we are planning to conduct further RF conditioning to achieve 300 kW RF power input with 3% duty cycle. When operated at 3% duty cycle, the power consumption of each semiconductor amplifier is about 5.2 kW and the wall loss of the RFQ cavity is approximately 7.5 kW (peak wall loss: 250 kW, duty cycle: 3%). Stable operation of the semiconductor amplifiers and RFQ cavity at 3% duty cycle requires the cooling system that maintains the RFQ cavity and the semiconductor amplifiers at a constant temperature. Therefore, we plan to construct a cooling system consisting of two air-cooled chillers (RKE3750-V-G2, ORION) with a cooling capacity of 12.2 kW each to cool two semiconductor amplifiers and an air-cooled chiller (RKE2200B1-V-G2, ORION) will be used to cool the RFQ cavity.

We plan to develop a permanent-magnet-type electron cyclotron resonance ion source and a double Einzel-lens-type low-energy beam transport for efficient acceleration of proton beams into the cavity. After the development of the ion source system, we intend to conduct a beam acceleration test of the RFQ linac with the four-feed RF system and the ion source system, as well as to evaluate whether the RFQ linac is suitable for the transportable neutron source.

References

http://www.cst.com.

https://rk-microwave.com/ca500bw2-8085rp/.

Ikeda

Otake

Kobayashi

and Hayashizaki

, Design of 500 MHz RFQ linear accelerator for a compact neutron source, RANS-III, Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms461 (2019), 186–190. doi:10.1016/j.nimb.2019.09.051.

Ikeda

Otake

and Mizuta

, Nondestructive measurement method to detect water/void inside slabs using compact neutron source by backscattered neutrons, Journal of Advanced Concrete Technology15(10) (2017), 603–609. doi:10.3151/jact.15.603.

Kobayashi

Ikeda

Otake

Ikeda

and Hayashizaki

, Completion of a new accelerator-driven compact neutron source prototype RANS-II for on-site use, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment994 (2021), 165091. doi:10.1016/j.nima.2021.165091.

Koubek

Grudiev

and Timmins

, RF measurements and tuning of the 750 MHz radio frequency quadrupole, Physical Review Accelerators and Beams20(8) (2017), 080102. doi:10.1103/PhysRevAccelBeams.20.080102.

Otake

Seki

Wakabayashi

Ikeda

Hashiguchi

Yoshimura

Sunaga

Taketani

Mizuta

Oshima

and Ishida

, Research and development of a non-destructive inspection technique with a compact neutron source, Journal of Disaster Research12(3) (2017), 585–592. doi:10.20965/jdr.2017.p0585.

Patent JP5317062B2, US8928216B2.

Takahiro

et al., Integration of 175-MHz LIPAc RF system and RFQ linac for beam commissioning, IEEE Transactions on Plasma Science48(6) (2020), 1489–1495. doi:10.1109/TPS.2020.2985020.

10.

Wakabayashi

Yoshimura

Mizuta

Ikeda

Hashiguchi

Kobayasi

Taketani

Yanagimachi

Goto

Sunaga

Ikeda

and Otake

A study on the non-destructive detection of salt in concrete using neutron-captured prompt-gamma rays at RANS, Plasma and Fusion Research13 (2018), 2404052.