www.elsevier.com/locate/nimb

Nuclear Instruments and Methods in Physics Research B 244 (2006) 467–472

NIMBBeam Interactions

withMaterials &Atoms

Proof examination of high efficiency IH-linac

Kazuo Yamamoto a,b,*, Toshiyuki Hattori a, Masahiro Okamura b, Noriyosu Hayashizaki a

a Research Laboratory for Nuclear Reactors, Tokyo Institute of Technology (T.I.Tech.), 2-12-1 O-okayama Meguro-ku, Tokyo 152-8550, Japanb The Institute of Physical and Chemical Research (RIKEN), 2-1 Wako-shi, Saitama 351-0198, Japan

Received 13 January 2005; received in revised form 16 August 2005Available online 24 January 2006

Abstract

An IH-linac which has features of high acceleration efficiency, a high energy gain and a low energy injection has been studied. Aneffect of End Ridge Tuner which characterizes electric-field distribution was investigated by a three-dimensional simulator. Also ahalf-scale cold model was constructed to confirm the distribution. To verify the design strategy to fulfill the features mentioned above,a test IH-linac has been manufactured. A shunt impedance of the linac was deduced by measuring energies of emitted X-rays from thecavity. These works proved that the IH structure can accommodate very low energy beam without pre-accelerator like RFQs and canaccelerate up to several-tens times of injection energy in a short single-cavity.� 2005 Elsevier B.V. All rights reserved.

PACS: 29.17.+w; 29.27.�a

Keywords: Heavy ion accelerator; IH-linac; Alternating phase focusing structure; Compact linac; 2 MeV proton; High effective shunt impedance

1. Introduction

Tumor therapy with carbon ion beams out of a synchro-tron is a topical accelerator application and a miniaturiza-tion of the accelerator complex is one of the main tasks formedical accelerator designers. However there is a limitationto minimize the synchrotron due to the desired particle’srigidity, therefore it will be important to develop a compactand simple injector linac. If the chain of the injector linacscan be replaced by a single short linac, the construction andoperation costs can be reduced. The simple robust struc-ture suites for medical facilities.

IH-linacs which have the substantial advantages ofsmall transverse dimensions and high shunt impedancesespecially at a low velocity region can be a good candidateto realize such a simple and compact injector. However this

0168-583X/$ - see front matter � 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.nimb.2005.08.154

* Corresponding author. Address: Research Laboratory for NuclearReactors, Tokyo Institute of Technology (T.I.Tech.), 2-12-1 O-okayamaMeguro-ku, Tokyo 152-8550, Japan. Tel./fax: +81 357343055.

E-mail address: k_yamam@nirs.go.jp (K. Yamamoto).

type of linacs has been commonly studied as a post-accel-erator due to mainly two difficulties in adopting a veryshort cell length at injection region.

The first thing to be solved is inducing transverse focus-ing forces with the short cell length. Typically the focusingelements like quadrupole magnets are installed in the drifttubes. In this case the cell length need to be at least severalcentimeters and the injecting beam should be pre-acceler-ated to have adequate cell length. To shorten the cell lengthto accept low energy particles from the ion source, otherschemes like alternate phase focusing (APF) should beapplied.

The other issue is in RF characteristics with the shortcell length at the entrance area of the IH-linac. To acceptthe low energy beam, the unit cell length varies precipi-tously in the single cavity. At the entrance area, many smallacceleration gaps concentrate and the voltage distributionis enhanced steeply. The large capacitance due to smallgaps tends to make un-balanced field distribution. On theother hand to keep maximum acceleration efficiency underan actual discharge limit, the field strength distribution has

0 20 40 60 80 100 1200.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0.0030

0.0035

0.0040

Em

od [

nor.

]

z cm

38cm 28cm 18cm

Fig. 1. Electric-field distributions and the ERT lengths. This figure showsthat the distribution can be optimized by varying the ERT length between18 and 28 cm. The field strength of the both edge of the cavity can bebalanced.

0 20 40 60 80 100 1200.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0.0030

Em

od [

nor.

]

z cm

26cm 24cm 22cm

Fig. 2. Fine tuning of the electric-field distribution. The ERT was changedby 1 cm step, 24 cm was considered to be an optimize point.

468 K. Yamamoto et al. / Nucl. Instr. and Meth. in Phys. Res. B 244 (2006) 467–472

to be constant. This area of research has been developed atTokyo University [1–3] and Tokyo Institute of Technology[4,5]; an energy gain achieve to around 10 times from 10 to161 keV by proton and from 240 keV/u to 2.4 MeV/u byheavy ions, respectively.

To realize the single-injector linac: (1) high energy gain,(2) high acceleration efficiency, (3) low energy injection and(4) enough transmission are needed to be established. Inthis paper, we focused on first three items (1)–(3), in the listand the performance of the built test linac is reported.

2. Basic design

A target species for acceleration was set to C4+ ion(q/A = 1/3) which is normally used in heavy ion tumortherapy facilities. Injection energy is 40 keV/u consideringextraction voltage from an ion source and realistic celllength (14 mm). Extraction energy is 2 MeV/u and anenergy gain is 50 times of the injection energy, which ismuch higher than usual amplification. A constant electricfield strength distribution was adopted to obtain highenergy gain in a single cavity. This means that appliedgap voltage is varied in ascending order from the injectionside to extraction side. For stable operation, the electric-field strength in the gaps was set not to exceed the twiceof the Kilpatrick limit [6]. The linac length was determinedto 1.5 m which is derived from an assumed achievable effec-tive acceleration voltage, 4 MeV/m. The number of cells is22, which was decided by the linac length divided by theaverage cell length. Each gap voltage and length wasdesigned as follows: (1) a total voltage to accelerate up todesigned-energy was calculated, (2) the first gap lengthwas given to be one-half of the first cell length, (3) the firstgap voltage was fixed to be the designed electric-fieldsstrength, (4) the sequential gap voltage distribution wasadjusted to maintain total design voltage and (5) all ofthe gap length was determined to be under twice the Kilpa-trick limit of the electric-field strength.

3. Research of IH-structure using a three-dimensional

simulator

Although model measurements have been commonlyused as a technique to investigate an IH-resonator toobtain designed electric-fields, we adopt a numericalthree-dimensional simulation in design stage. By changingthe cut shape of a plate which sustains the drift tubes,the electric-field distribution was controlled. We call thecut at the end of the plate End Ridge Tuner (ERT). TheERT manipulates the path of the axial magnetic flux andthe gap voltages near the ERT can be increased. The qual-ity factor and the shunt impedance of the resonant cavityremained practically unchanged. For accommodating thelow energy injection beam, due to condensation of thecapacitances, the injection side of the cavity intrinsicallyhas strong electric field strength. Therefore, the effect ofthe ERT was investigated carefully by OPERA-3d [7]. In

3D modeling, only one half of the cavity was simulated,since the longitudinal cut plane through the ridge makesmirror image. The memory size of the computer and thecomputation time restricted mesh size and fine mesh wasemployed only in the gap region. In the present conditions,meshes are divided to 1 mm step in the gap and 2 mm in thedrift tube region. Fig. 1 shows examples of the electric-fielddistributions with 18, 28 and 38 cm of the ERT at the exitof the linac, respectively. The total length of the ridge is116 cm. As can be seen in the figure, the effect of theERT is strong enough to control the field distributionalong the beam axis even in the cavity which accepts lowenergy injection.

Fine adjustment of the ERT was performed as shown inFig. 2. The electric-fields at the both ends are almost samewhen the ERT length is 24 cm, which corresponds to 21%of the total ridge length. The field strength at the center ofthe cavity was calculated as 83% of the strength at the bothend. This value is acceptable for the test linac. In case ofshorter IH cavity we can avoid the depression in the fielddistribution with the ERT tuning. If adopt a longer cavity,

Table 1Parameters of the linac

Particle C4+

Input energy 39 keV/uOutput energy 1.9 MeV/uFrequency 98.2 MHzNumber of cells 22Cavity length 1280 mmDiameter 560 mm

2000 4000 6000 80000.3

0.4

0.5

0.6

0.7

0.8

Ele

ctri

c-fi

eld

[arb

.]

z [arb.]

Fig. 4. Electric-field distribution of half-scale cold model with the ERTlength of 12 cm. It turned out that the measured electric-fields distributionat 12 cm of 1/2 model well agreed with the simulation of the 24 cm ofthree-dimensional model (1/1 model).

K. Yamamoto et al. / Nucl. Instr. and Meth. in Phys. Res. B 244 (2006) 467–472 469

we see serious depression at the center part and it should becared by other tuning schemes. The design parameters ofthe linac based on the simulation study are summarizedin Table 1.

4. Model measurements

A half-scale cold model was constructed to confirm theelectric-field distribution predicted by the simulation. Themodel was made of brass and 28 cm of inside diameterand 64 cm in length as shown in Fig. 3. The measurementwas carried out by a perturbation method; a small alumin-ium ball was placed onto the acceleration axis and the elec-tric-fields at this position were obtained by measuring thevariation in the frequency. The electric-fields distributionwas obtained by changing the ball position. The optimumelectric-fields distribution was obtained at an ERT lengthof 12 cm as shown in Fig. 4. The electric-fields strengthin center is lower (about 81%) than those in the bothend-area. These results have been predicted by the simula-tion. The numerical 3D simulation of the IH cavity couldachieve good enough accuracy. This also indicates thatwe can get more information on the RF characteristics ofthe cavity more easily with extendibility compared to themodel measurements. One can say that the 3D simulationbecame strong tool to optimize the cavity structure andcontributes the design of the high acceleration efficiencyaccelerator.

Fig. 3. Photograph of constructed half-scale cold model with the ERTlength of 12 cm.

5. Manufacture

Based on the simulation and the low power RF fieldmeasurements of the cold model, an IH-linac was con-structed to demonstrate the efficient structure particularlyin the low energy range. The cavity was separated intothree parts for easy fabrication and easy modification ofthe drift tubes. The middle plate was machined from singlestainless plate within ±0.05 mm error.

The drift tubes and the supporting stems of the linac aremade of copper. The edge of the drift tubes was cut off toform a round shape to reduce the excess concentration ofthe field strength. Based on 2D electro-static simulation, amaximum field strength on the surface of the drift tubeswas estimated as 30 MV/m which corresponds to 2.6 timesof the Kilpatrick’s discharge limit at 98.2 MHz (11.27MV/m). The high field spots were found in injection sideof the cavity. In case of the small size of the drift tubes, highbravery factor, up to 3.5 times, has been reported in [8].High power operation is expected after the aging. Afterplating Cu on the middle plate, the drift tubes were fixedon the ridge with cap screws. Cooling water channels weredrilled in the middle plate and the copper made stems wereextended into the channel to diffuse the heat. The alignmentof the drift tubes on transverse plane was checked by insert-ing a stainless steel rod through the drift tube bores. Forlongitudinal alignment, the gap lengths were adjusted byusing shims. Due to machining error and thickness of theplating, the longitudinal position errors of the drift tubeswere piled up. We decided to adjust the gap length of cellNo. 18, which was longest gap length, to absorb all thelongitudinal position errors. Consequently, almost all thegap lengths increased to about +0.5% and only the 18thgap length was reduced to �1.8% from the design value.Although the synchronous phase became shifted graduallyand slightly from the design value up to a few degrees, itwas calculated that the influence can be negligible.

470 K. Yamamoto et al. / Nucl. Instr. and Meth. in Phys. Res. B 244 (2006) 467–472

After the alignments and the vacuum test the top andbottom vessels were assembled. The resonant frequencyand quality factor of the linac were measured to be98.2 MHz and approximately 10,800, respectively.

6. Acceleration test bench

An acceleration of a proton beam was carried out in thetest bench as shown in Fig. 5. A PIG ion source with anextraction voltage of 15 kV was utilized to extract protonsand additional high voltage was applied to a column whichaccelerates the beams up to injection energy. In a low energybeam transport line, there is bending magnet to select theproton and focusing lenses before and after the magnet.The accelerated particles after the IH-linac were analyzedby a bending magnet which was placed after a triplet-elec-tric-quadrupole lens. A metal mesh was attached in frontof the Faraday cup to reduce the noise from the cavity.

The required RF power was fed to the linac with a singleloop coupler which was made of 5 mm wide and 1 mmthick copper strip. The coupling strength between theamplifier and the resonator is adjustable by rotating theloop from outside of the vacuum.

7. Results of acceleration test

Fig. 6 shows the energy spectrum of accelerated protons.The figure clearly shows that the protons were accelerated

Fig. 5. Layout of the e

up to the design energy. During this measurement, thefocusing-lens strength was adjusted so that the acceler-ated-ion current would be maximum at the Faraday cupafter the bending magnet. Since the slit in front of the Far-aday cup was fully opened, it is difficult to discuss aboutthe energy spread from the signal widths in the energy spec-trum. A peak appeared at around 1 MeV seems partiallyaccelerated beam which was not captured by designed lon-gitudinal bucket. Some particles could be transmittedtrough the linac with some energy, because the injectionbeam was DC beam and some amount of the beam couldnot be fit into the design longitudinal acceptance. This phe-nomena was proved by our simulation (Fig. 7). We plan toinstall a buncher before the linac which will provide goodefficiency to capture the injected beam and will improvethe transmission of the beam.

Fig. 8 shows accelerated energy and ion current as afunction of RF power. The gap voltage, obtained fromthe measured Q-value, corresponds to an RF power of24 kW. To match the DC beam from the PIG ion source,it is recommended to employ a buncher cavity before thelinac.

A ratio of extraction currents divided by injection cur-rents was 6.1 · 10�2%. Considering the RF and the ionsource operation, a transmission efficiency was estimatedas 9.5%. Since no focusing elements are employed, thetransverse acceptance is small and we think the obtainedtransmission is reasonable value. Fig. 9 shows the varying

xperimental setup.

Fig. 6. Energy spectrum of the accelerated beam. The accelerated proton is shown at energy of 1.9 MeV.

Fig. 7. Energy diagram of acceleration scheme. Zero emittance intransverse plane, on the axis beam, was assumed. The phase angle ofthe injected particle was changed in every 5�, 0–360�, represented by theeach trajectory.

15 20 25 30 35 40

1.2

1.3

1.4

1.5

1.6

1.7

F.C

. cur

rent

[pA

]

RF power [kW]

Fig. 8. Current and energy of the accelerated proton beam as a functionof the RF power.

K. Yamamoto et al. / Nucl. Instr. and Meth. in Phys. Res. B 244 (2006) 467–472 471

acceptance in different RF phase angles. The injected beamemittance (nor. 0.6p mm mrad) are expressed by an enve-lope and only small fraction of the beam emittance canbe matched to the acceptance. For practical use, the trans-verse focusing scheme without quadrupole magnets in thedrift tubes will be an interesting and indispensable issuein the next step.

8. Measurements of RF cavity fields level

Energy spectroscopy measurements of X-rays from thecavity provided accelerating-gap voltages in the multi-cellaccelerating structures. An X-ray detector was used to

measure the energy of the emitted X-rays. The maximumenergy of the X-ray energy spectrum corresponds to thepeak-gap voltage in the RF structure. Sets of these dataprovide the scaling relation between cavity power and RFvoltage and this scaling relation provides the informationof the shunt impedance.

The X-ray detector was positioned at the behind of thecavity. The RF power was turned on to the desired level(= 24 kW) and the X-ray data spectrum was collected atthe fixed RF level until sufficient statistics were accumulatedto permit fitting to the energy endpoint. The peak-gap volt-age was estimated to be 175 kV as shown in Fig. 10. Thisvalue agrees well with the design within a few % error.

0 20 40 60 80 100 120 140 160 180 200

10

100

1000

10000

Cou

nt [

Log

_sca

le]

Energy [keV]

175keV

Fig. 10. Energy spectrum of the X-ray from the cavity. The maximumX-ray energy is shown as 175 keV.

-8 -6 -4 -2 0 2 4 6 8-50-40-30-20-10

01020304050

x', y

' [m

rad]

x, y [mm]

-5 -15 -25 -35

Fig. 9. The simulated phase acceptance. Since the assumed initial phaseangle of the particle was varied, the acceptance of the linac wastransformed in the simulation. The envelope corresponds to the emittanceof the real injected beam (nor. 0.6p mm mrad). Each acceptance only canaccept around 9% of the injected beam emittance.

472 K. Yamamoto et al. / Nucl. Instr. and Meth. in Phys. Res. B 244 (2006) 467–472

The shunt impedance was confirmed to over 223 MX/m.This means that the IH-linac which is specially tuned to

accept the low energy injection beam can demonstrate highshunt impedance as same as other IH-linac have.

9. Conclusion

The electric-field distribution in the IH-linac whichaccelerated C4+ up to 1.9 MeV/u from 39 keV/u was con-trolled very well by the end ridge tuner and high shuntimpedance of around 223 MX/m was performed. Theenergy of the accelerated beam achieved approximately50 times of injection energy within the single resonator. Itwas proved that the structure could accommodate verylow energy beam without pre-accelerator like the RFQs.We plan to study focusing schemes like finger focusing [9]or alternating phase focusing [10] those will improve thetransverse acceptance. The entire chain of the injectorlinacs for the medical synchrotrons might be replaced bya single-IH-linac.

References

[1] T. Hattori, S. Yamada, E. Tozyo, et al., in: Proc. 8th LinearAccelerator Meeting, Vol. 8, 1983, p. 59.

[2] S. Yamada, T. Hattori, T. Fujino, et al., in: Proc. of Int. IonEngineering Conguress ISIA 83 & IPAT, Vol. 83, 1983, p. 623.

[3] S. Yamada, T. Hattori, T. Fujino, T. Fukushima, T. Murakami, E.Tojyo, K. Yoshida, INS-NUMA-57, 1985.

[4] T. Hattori, K. Sato, H. Suzuki, et al., in: Proc. 10th LinearAccelerator Meeting, Vol. 10, 1985, p. 99.

[5] T. Hattori, K. Sato, H. Suzuki, et al., in: Proc. 1986 Int. Conf. ofLinear Accelerator, Stanford, 1986, p. 377.

[6] W.D. Kilpatrick, Rev. Sci. Instr. 28 (10) (1957).[7] OPERA-3D ver.10, copyright�, Vector Fields Limited, England.[8] J. Broere, H. Kugler, M. Vretener, U. Ratzinger, B. Krietenstein,

in: Proc. of 1998 Int. Linear Accelerator Conference, Chicago, 1998,p. 771.

[9] T. Ito, N. Hayashizaki, S. Matsui, K. Sasa, H. Schubert, E. Osvath,T. Hattori, Nucl. Instr. and Meth. B 161–163 (2000) 1164.

[10] T. Hattori, S. Matsui, N. Hayashizaki, et al., Nucl. Instr. and Meth.B 161–163 (2000) 1174.