Preamble

Development of a compact X-band dual-mode medical linac based on a pulse compressor Haipeng Zhi, Jiaru Shi, 1, 2, Hao Zha, Jian Gao, Qiang Gao, and Huaibi Chen 1 Department of Engineering Physics, Tsinghua University, Beijing, 100084, China Key Laboratory of Particle and Radiation Imaging (Tsinghua University), Ministry of Education, Beijing 100084, China In this study, a compact X-band dual-mode electron linear accelerator (linac) was developed for medical radiotherapy, which is capable of generating both 100 mA

6 MeV

beams for X-ray radiotherapy (low-energy mode) and 10 mA

5 MeV

beams for electron radiotherapy (high-energy mode). The dual-mode linac leveraged a 3 MW multi-beam X-band klystron as the power source, with a total length of . A novel spherical SLAC Energy Doubler (SLED pulse compressor) was specially designed for the high-energy mode to achieve flat-top output, with a power gain factor of 2.2. Mode switching is realized through dynamic tuning of the pulse compressor’s resonant frequency. Furthermore, a standing-wave (SW) bi-periodic accelerating structure operating in mode was developed, which enables effective bunching and acceleration of electron beams under both modes. A comprehensive experimental platform has been established for the linac, and high- power RF tests have been conducted. The experimental results validate the feasibility of the dual-mode linac, demonstrating its potential for versatile radiotherapy applications with optimized spatial compactness.

Keywords

Dual-mode linac, SLED pulse compressor, Standing-wave accelerating structure, High-power RF experiment.

INTRODUCTION

Cancer remains a pressing global challenge. According to the 2024 statistical report by the International Agency for Research on Cancer (IARC), 2022 witnessed over 20 mil- lion new cancer cases and 9.5 million cancer-related deaths globally [ ], highlighting the urgent demand for advanced therapeutic modalities. Radiotherapy, employed as primary treatment or in combination with chemotherapy, surgery, or other modalities for over of cancer patients, contributes to nearly of curative outcomes [ ], underscoring its pivotal role in oncology. Innovations in radiotherapy tech- nologies, including improved dose precision [ ] and multi- mode treatment strategies [ ], not only hold profound sci-

entific significance but also yield substantial economic ben- 14

efits, thus making critical contributions to the progress of global health.

6 MeV

low-energy medical electron linac, designed for X-ray radiotherapy, features a compact configuration and low treatment costs, enabling its widespread adoption in pub- lic cancer care [ However, for superficial tumors amenable to electron therapy, such as breast, skin, and thy- roid cancers [ ], low-energy medical linacs are con- strained by their RF power sources, failing to produce elec- tron beams with higher energy.

Furthermore, most prior medical linacs engineered to produce higher-energy electron beams (10–20 ) have relied on higher-power RF sources, resulting in elevated costs, complex system configurations, and stringent maintenance requirements [ ]. Thus, there is an urgent demand for medical electron linacs that combine the compactness and cost-effectiveness of low-energy sys- tems with the capability of accelerating higher-energy elec- tron beams, especially for the advancement of cancer treat- ment capabilities in less developed regions.

Since its inception in the 1970s, the pulse compressor has been widely adopted in diverse accelerators [ compressing pulse width to enhance pulse power, it over- comes the power limitations of RF sources, enabling higher- energy electron acceleration. Most pulse compressors feature

a compact and mechanically simple structure. For example, 39

the SLED pulse compressor comprises only several passive components, such as a resonant cavity and a dual-mode RF polarizer, resulting in low manufacturing costs and high op- erational reliability. More importantly, leveraging the pulse compressing effect can drastically relax the performance re- quirements for the RF sources, including reduced klystron

output power, lower modulator voltage and diminished ther- 46

mal load on the cooling system, thereby significantly cutting 47

project funding, equipment costs and long-term operational complexity. Additionally, switching between the low-energy and high-energy modes of the linac can be easily achieved

through dynamic tuning of the pulse compressor’s resonant 51

frequency. In summary, compared to developing entirely new higher-power RF sources (including klystrons and mag- netrons), the adoption of pulse compressor provides a more cost-effective and engineering-feasible pathway for medical linacs to achieve high-energy electron acceleration.

Meanwhile, X-band high-gradient technology has progres-

sively matured in recent years, with significant advancements 58

achieved in accelerating structure design, manufacturing, and

tuning, as well as in research on beam break-up instability 60

]. Compared with conventional S-band and C-band ], X-band accelerating structures exhibit higher shunt impedance, which facilitates the compact designs of medical linacs with lower spatial requirements. While substantial re- search has been devoted to the development of X-band medi- cal linacs, existing efforts are predominantly restricted to the design of the accelerating structure themselves [ ], and no studies have been reported on integrating pulse compres- sors with X-band accelerating structures for the development of X-band dual-mode medical electron linacs.

Based on the above analysis, we have developed a com- pact X-band dual-mode medical electron linac integrated with a SLED pulse compressor, enabling seamless switching be- tween X-ray and electron therapy. The linac employs a

3 MW

marks the moment when the energy storage reaches a steady state. (b) Input and output fields of the pulse compressor. (c) Complete vacuum model of the pulse compressor. polarizer was developed for the pulse compressor [ Its RF characteristics emulate a coupler, with two or- thogonal modes in cylindrical port replace the tradi- tional two ports of the coupler, reducing the number of SLED cavities from two to one. Additionally, the butterfly- shaped structure eliminates the symmetric cylindrical waveg- uide, achieving vertical dimension reduction compared to conventional designs. The coupler optimization followed the CERN-proposed function [ ], which has been op-

timized to − 95 dB , ensuring minimal power leakage. 187

Fabrication of the spherical resonant cavity inevitably in- troduces deviations in sphericity, leading to resonant fre-

quency detuning between the two orthogonal TE 114 modes, 190

which significantly deviates the S-parameters of the two po- 191

larization modes. Therefore, independent measurements and

separate tuning of the two polarization modes were critical to 193

ensuring operational consistency. A single-mode RF polar- izer dedicated to independently measuring the S-parameters of the two polarization modes has been developed. By cou- pling the single-mode RF polarizer to the spherical resonant cavity at orthogonal orientations, the S-parameters of the two

The vacuum model and electric field magnitude of the com- 200

plete pulse compressor are presented in Fig. (c), with all simulated RF parameters matched the design specifications.

In particular, at an input power of

3 MW

, the maximum sur- face electric field occurring at the coupling aperture was only

19 MV m

− 1 . This low-field design significantly reduces the 205

risk of vacuum breakdown and improves the operational ro- bustness of the pulse compressor [ Lower-Power RF Experiment Cold test of the pulse compressor was performed using a Rohde & Schwarz Vector Network Analyzer (VNA). Before welding, the directivity of the dual-mode polarizer was first verified. Within the frequency range of 9.25-9.35 , both parameters remained below 30 dB , while exceeded 15 dB , demonstrating satisfactory per- formance of the dual-mode RF polarizer. Then, the single- mode RF polarizer was used to measure the frequency sep- aration of two orthogonal polarization modes in the spherical resonant cavity. The frequency separation between the two modes was finally tuned to less than 100 kHz

After welding, cold test and tuning of the complete pulse 220

compressor were conducted, as illustrated in Fig. Switching between the low-energy mode and high-energy mode is achieved by adjusting the insertion depth of the top

tuning pin to change the resonant frequency of the pulse com- 224

pressor. When the top tuning rod is fully retracted, the res- 225

onant frequency should exactly coincide with the operating frequency, thereby effectively compressing the klystron out- put power and enabling high-energy electron acceleration.

The primary objective of the cold test and tuning is to ver- 229

ify whether the characteristics of the pulse compressor in this state are consistent with the design expectations, including

resonant frequency f = 9 . 3 GHz , coupling factor β = 3 . 2 , 232

and quality factor Q 0 = 10 5 . 233

However, the tuning pin has a fragile structure and must 234

be connected to the resonant cavity via a flange to ensure air tightness. Premature installation will increase the risk of dam- age or vacuum leakage during subsequent transportation and

installation process, so the tuning pin was not installed during 238

the cold test. Instead, direct flange sealing is adopted. Con-

sidering that the diameter of the tuning pin hole at the top 240

of the resonant cavity is only 7 mm, corresponding to a cut- off frequency of

1 GHz

for the dominant mode, the

3 GHz

microwave undergoes rapid attenuation in the hole.

Thus, direct flange sealing yields the same electromagnetic

effects as when the tuning pin is fully retracted. 245

marks the moment when the energy storage reaches a steady state. (b) Input and output fields of the pulse compressor. (c) Complete vacuum model of the pulse compressor.

The target operating frequency of the pulse compressor is

3 GHz

, with an intended operating temperature of and an ion pump achieves high-vacuum environment within

the cavity ( < 10 − 6 Pa). Thus, the target frequency during cold 249

test need be calculated via Eq. (

f tun = f tar √ ϵ r + αf tar ( T tar − T tun ) (5) 251

where f tar = 9 .

3 GHz

is the operating frequency, ϵ r = 252

1 . 00059 is the relative atmospheric dielectric constant, α = 253

is the thermal expansion coefficient of cop-

per, T tar = 35 ◦ C is the target temperature and T tun is 255

the temperature during cold test, which was measured to

be 22 . 7 ◦ C . The target tuning frequency was calculated as 257

9 . 299 16 GHz and the tuning results are presented in Fig. 2 [FIGURE:2] 258

(b). At the resonant point, the S21 parameter is 81 dB corresponding to a coupling factor of 3.1, which is close to

the target coupling factor β = 3 . 2 . The loaded quality factor 261

of the pulse compressor is measured as using the half-power bandwidth method, corresponding to the intrinsic

quality factor Q 0 pc = 9 . 84 × 10 4 , which also agrees well with 264

the expected results. Furthermore, in the low-energy mode, simply inserting the

tuning pin to a sufficient depth induces a significant shift in 267

the resonant frequency of the pulse compressor. Under this condition, RF power is reflected at the coupling hole and ex- its from the dual-mode RF polarizer. The pulse compressor ceases pulse compression functionality, enabling low-energy mode operation. RF simulation demonstrated that a

diameter stainless-steel tuning pin inserted to 60 mm achieves 273

20 MHz

resonant frequency shift, fully meeting the opera-

tional requirements. The tuning pin has an actual maximum 275

insertion depth of 100 mm , providing greater tuning margin. 276

In summary, since low-energy mode only requires full inser-

tion of the tuning pin to detune the pulse compressor, elimi- 278

nating the need for calibration of insertion depth and resonant frequency offset, specialized cold test for the detuned state is unnecessary.

Then, leveraging the linear response characteristics of the pulse compressor, low-power RF experiment was conducted to evaluate its performance, with the experimental setup illus- trated in Fig. (a) and Fig. (b). As the power source for the low-power RF experiment, the low-level RF (LLRF) system includes an excitation source, a solid-state amplifier (SSA), a phase inverter, and a trigger source. The excitation source, serving as the primary generator of microwave signals, pro-

vides initial RF excitation for the system. The SSA is inte- 290

grated with the excitation source to amplify the microwave signal, serving as the klystron driving power. It was not ac- tivated during low-power RF experiment. The phase inverter performs dual functions: phase inversion and amplitude mod- ulation of RF signals. And the trigger signal source synchro-

nizes the entire LLRF system, ensuring precise timing coor- 296

dination among components for coherent operation. To keep consistency with high-power test conditions, the pulse com- pressor operated under low-vacuum conditions, maintained

by a mechanical pump, during the low-power RF experiment. 300

A Rohde Schwarz NRP-Z81 power meter was employed for measurement of the compressed pulse.

During the low-power RF experiment, the measured tem- perature of the pulse compressor was . According to Eq. ( ), the output frequency of the excitation source was set

Experimental results. to 9.30018 GHz to match the resonant frequency of the pulse compressor. The experimental results are shown in Fig. with the input pulse consisting of a flat-top signal and 500 ns inverted compensation signal. With an input pulse power of , an output pulse power of was suc- cessfully measured, achieving the predetermined power gain target of However, the phase inverter demonstrated suboptimal per- formance, with the modulation of the inverted signal exhibit- ing abrupt transitions.

The imperfect performance of the phase inverter directly impacted the output waveforms of the pulse compressor, causing the output waveforms to exhibit ripples. Additionally, at the theoretical phase inversion time , a spike with a pulse width of approximately 100 ns was observed. A potential cause lies in the non-ideal transient response of the phase inverter [ ]. Specifically, IQ modula- tion is adopted to realize phase inversion and amplitude mod- ulation of the RF signal in LLRF. A minor timing mismatch may exist between I/Q channels, or there may be overshoot on the edges of control signals. These two situations may cause the vector synthesis of phase and amplitude to deviate from the ideal condition at the moment of switching, thereby gen- erating the aforementioned spike. Because of the spike, the measured usable pulse width was only 400 ns , narrower than the designed 500 ns , but still satisfied the high-energy mode operational requirements of the linac.

To improve the output waveform of pulse compressor, the fundamental approach is to adopt a vector modulator with superior dynamic performance, thereby enhancing the tran- sient response speed of RF signal modulation. In addition, the adoption of predistortion compensation can improve the flatness of the output waveform. Nevertheless, from the per- spective of principle feasibility verification, the current exper- imental results already meet the requirements for high-power RF experiment.

X-BAND SW ACCELERATING STRUCTURE

In this section, based on a grid-controlled thermionic cath- 342

ode DC gun, we developed a X-band

3 GHz

SW acceler- ating structure operating in mode, which is capable of accelerating both 100 mA

6 MeV

10 mA

5 MeV

electron beams. The pulse current is adjusted by changing the voltage of the DC gun. The capture ratio and acceleration efficiency of the structure were systematically investigated in different operating modes to optimize its performance.

RF Design and Analysis Structural compactness represents the pivotal advantage of the dual-mode medical linac. The accelerator adopted a X-

3 GHz

mode bi-periodic SW accelerating struc- ture, which maximizes the frequency separation between ad- jacent resonant modes while improving the acceleration effi- ciency [ ]. In addition, nose-cone structures were incorpo- rated to improve the longitudinal shunt impedance, and mag- netic coupling holes were integrated into the cavity chains to enable efficient power transmission. These designs collec- tively contribute to improving the compactness of the accel- erating structure.

The bunching section, consisting of four bunching cavities, was designed through iterative calculations of field distribu-

tion and beam dynamics. To maintain structural compact- ness, the accelerating structure employs a single-feed SW de- sign. However, this introduces an inherent limitation: un- der the two operating modes with different input powers, the bunching section maintains consistent normalized field dis- tributions while the amplitude scales with the

lationship, leading to significant disparities in bunching effi- 370

ciency between the two modes. To mitigate this discrepancy, we adjusted the gun voltage and emission current, aiming to identify a appropriate bunching section field distribution that achieves acceptable capture ratios for both modes. Concur- rently, the iris radius was increased from to reduce beam transport losses. Distinct from conventional low-energy electron linacs, the high-energy mode represents the key innovation of the new linac, so the bunching section prioritized the optimization for

5 MeV

electron beams, achieving electron capture ratios of in low-energy mode in high-energy mode.

The DC gun parameters were ultimately determined as follows: gun voltage and 500 mA emission pulse current for the low-energy mode; and 10 kV gun voltage and 30 mA emission pulse current for the high-energy mode.

The comprehensive vacuum model of the accelerating structure was constructed in Ansys HFSS, with an overall length of , comprising 4 bunching cavities and 18 stan- dard cavities, as shown in Fig. (a). The simulation results, including the S11 curve, smith plot and normalized on-axis electric field distribution are presented in Figs. (d), respectively. The longitudinal effective shunt impedance of the standard cavities is

115 MΩm

Combining theoretical analysis and engineering experi- 395

ence, the coupling factor of the accelerating structure was finally determined to be 1.1.

For high-energy mode with 10 mA

5 MeV

beams, the beam power is merely

14 MW

, much lower than the field-building power of

80 MW

calculated by Eq. ( ), and the corresponding theo- retically optimal coupling factor is approximately 1.03. How- ever, local temperature rise at the coupling hole during op- eration causes deformation and slight decrease in . Thus,

the accelerating structure was designed with β = 1 . 10 to en- 404

sure optimal power feeding for high-energy mode. In con- trast, for the low-energy mode with 100 mA

6 MeV

beams, the beam power and field-building power are

95 MW

, respectively, corresponding to a matched coupling

factor of 1.63. Although β = 1 . 10 induces beam load mis- 409

match and power reflection during low-energy mode opera- tion, the power loss is acceptable. Power reflection coefficient in the low-energy mode is , calculated via Eq. (

P ref /P inc = [ β − β 0 β + β 0 ] 2 (6) 413

where β = 1 . 10 and β 0 = 1 . 63 . 414

In summary, the determination of coupling factor priori- tizes two key considerations. First, the high-energy mode re- RF design of the SW accelerating structures.(a) Vacuum models of the SW accelerating structure. (b) Parameters. (c) Smith plot. (d) Normalized on-axis electric field. quires

94 MW

input power. Accounting for trans- mission loss, the total input power reaches

22 MW

, nearing the power limit of the pulse compressor (

6 MW

). There- fore, power coupling matching of the high-energy mode is

Kinetic energy and pulse current of the high-energy mode. (c) Transverse phase space of the low-energy mode. (d)Transverse phase space of the high-energy mode. (e) Transverse emittance. (f) Energy spectrum the core design objective.

Second, the low-energy mode only requires an input power of

55 MW

. Even account- ing for reflected power from beam load mismatch transmission loss, the total input power remains

02 MW

, which can be fully satisfied by the

3 MW

multi- beam klystron. Thus, the low-energy mode imposes relatively

lenient requirements on coupling matching. 427

Beam dynamics simulation was conducted using ASTRA.

The kinetic energy and pulse beam current of the two oper- ating modes are shown in Figs. (a) and (b). According to our final design, at an output RF power of

0 MW

the klystron, electrons gain

0 MeV

energy with the 100 mA pulse current in low-energy mode operation. For the high- energy mode, the RF power transmission proceed as follows:

The klystron initially outputs 2 .

8 MW

of RF power. After 435

compression by the pulse compressor (with a power gain of 2.2) and accounting for of transmission loss,

9 MW

of RF power is expected to be fed into the accelerating struc- ture and electrons gain

5 MeV

energy with 10 mA pulse current. The above results are basically consistent with the

expected parameters during the pulse compressor design.

The transverse phase space distributions at the exit of the accelerating structure are shown in Figs. (c) and (d). And the transverse emittance curves are shown in Figs. (e). In the high-energy mode, lower pulsed current and higher longi- tudinal velocity collectively suppress the divergence of elec- tron bunches caused by space charge effect, leading to supe- rior transverse emittance compared to the low-energy mode, characterized by more concentrated position distribution and smaller angular divergence.

Energy spectrum for both modes are shown in Fig. (f). In the high-energy mode, the rms energy of 95% electrons in the bunch is

53 MeV

. And the relative energy spread is defined as the ratio of rms energy spread to its rms energy.

The low-energy mode yields an rms energy of

11 MeV

an relative energy spread of 8 . 3% . The significant discrep- 456

ancy of relative energy spread between the two modes pri- marily arises from the longitudinal bunching dynamics, par- ticularly the bunching section, which were optimized specifi- cally for the high-energy mode. In electron therapy, the dose is deposited directly by the beam, and a larger energy spread directly flattens the dose fall-off at the end of its range, com- promising its core advantage of sparing healthy tissues be- hind superficial tumors.

The energy spread design priori- tizes the high-energy mode, primarily based on the follow- ing considerations. In contrast, for X-ray therapy, the dose distribution results from polyenergetic photons generated by

bremsstrahlung, making it less sensitive to the initial electron 468

energy spread. Meanwhile, an integrated design of the bunch- ing and accelerating sections was adopted for structural com- pactness. This integration leads to a reduced field amplitude in the bunching section under low-energy mode, resulting in suboptimal bunching efficiency and a wider phase distribu- tion, thereby inducing a larger relative energy spread.

To further improve the capture ratio and relative energy spread of the low-energy mode, consideration can be given to a scheme where RF power is fed separately to the bunching section and the accelerating section. In this scheme, only the RF power fed to the accelerating section is adjusted across different operating modes to achieve distinct energy gains.

Although this approach compromises structural compactness, it maintains a constant bunching field across different modes, thereby preserving the bunching efficiency.

Cold Test After fabrication of the accelerating structure, the resonant frequency, quality factor and coupling coefficient of each cell were examined using plunger method. Notably, due to the

thin thickness, traditional tuning mechanisms such as tun- 488

ing holes or tuning pins could not be incorporated during 489

the manufacturing process. Thus, tuning of each cell was 490

achieved by reprocessing unsatisfactory copper disks with high-precision machine tools. Following two iterative repro- cessing cycles, all cavities were tuned into an appropriate range around the operating frequency of shown in Fig. (a). Deviation of the inter-cavity coupling coefficient from the design values was within , a level of precision that eliminated the need for further adjustments.

Subsequently, the assembled copper disks and waveguide

coupler were integrated into a cohesive unit through brazing, 499

utilizing a silver-copper alloy braze. This brazing material was selected for its excellent thermal conductivity, high me-

chanical strength, and compatibility with copper, ensuring ro- 502

bust bonding and minimal RF loss at the interfaces. The com- 503

plete accelerating structure is shown in Fig. (b). Then, cold test of the complete accelerating structure was conducted us- ing the VNA. The parameter, sampled at

3 GHz

, ex-

hibited a value of − 23 . 7 dB , corresponding to β acc = 1 . 14 . 507

was measured to be 3641 using the half-power band-

width method. From this, τ acc = 125 ns , which is close to the 509

expected value of . And the on-axial electric field dis- tribution of the accelerating structure was measured by bead- pull method showing good agreement with RF simulation re- sults, as illustrated in Fig.

HIGH-POWER RF EXPERIMENT The pulse compressor has been validated via low-power RF experiment and cold test of the SW accelerating structure has been conducted. By integrating the SW accelerating struc- ture as the load terminal, configuring a high-voltage power for the DC gun, and expanding synchronous trigger settings, we successfully established a high-power RF experimental plat- form for the X-band dual-mode linac. This section details the construction and configuration of the high-power RF exper- imental platform, along with the sampling, processing, and analysis of experimental data.

High-Power RF Experiment Setup The photograph of the X-band dual-mode experimental platform is shown in Fig. (a). Benefiting from the compact X-band pulse compressor and SW accelerating structure, the main body of the experimental platform has an overall length of only , advancing the way for the linac toward com- mercialization.

For the RF power section, a

3 MW

X-band

3 GHz

multi- beam klystron served as the primary power source. A phase inverter capable of amplitude modulation was interposed be- tween the LLRF excitation source and the solid-state ampli- fier (SSA), enabling phase inversion and amplitude modula- tion of the RF power. The klystron modulator had a maximum high-voltage pulse width of , imposing a strict constraint on the RF power pulse width. To protect the klystron from reflected power, a four-port circulator connected with water loads was installed between the klystron and pulse compres- The resonant frequency of the pulse compressor was

tuned via a mechanical tuning pin positioned axially above 543

its cavity. When the pulse compressor was detuned from the operating frequency, it functioned only as a lossy waveg- uide with measurable power attenuation of approximately

Cold test of the X-band SW accelerating structure. (a) Single-cavity frequency tuning process. (b) Photograph of the accelerating structure. (c) Normalized on-axis field distribution measured by bead-pull method. . The pulse compressor maintains a constant cav- ity temperature of through four internal water cooling channels, each configured with a flow rate of

10 L

. No- tably, we specifically reinforced the cooling structure near the

coupling holes with high surface fields to prevent significant 551

changes in the coupling coefficient due to thermal deforma- tion, thereby ensuring stable operation of the pulse compres- sion system.

In the accelerating section, a directional coupler was in- stalled between the pulse compressor and the SW accelerat- ing structure to enable real-time sampling of incident and re- flected RF power, thereby detecting vacuum breakdown in the

structure. The grid-controlled thermionic cathode DC gun, 559

serving as the electron source, allowed precise adjustment of emission current by regulating the gun voltage, accommodat- ing the beam current demands for both operation modes. The DC gun had a maximum pulse width of , enabling full coverage of RF power pulse. Electron beams were expected to be bunched and accelerated to

5 MeV

the SW structure in the low-energy mode and high-energy -diameter internal water channels on copper disks with a flow rate of

3 L

Titanium window or tungsten target was installed at the exit 570

of the SW accelerating structure to extract electron beams or generate X-rays through bremsstrahlung. The linac was

sealed using RF windows, and a 4 kV titanium pump main- 573

tained high-vacuum environment with air pressure below in the accelerator, preventing arcing and ensuring beam stability. Corresponding detectors were installed down- stream for beam measurements, and their operational princi- ples were detailed in the measurement results section.

The pulse compressor and SW accelerating structure were conditioned in advance to reduce the breakdown rate (BDR) ]. Breakdown events were identified by abnormal re- flected wave signals and transient increases in vacuum pres-

sure. During the conditioning process, RF power injection 583

into the accelerator was gradually increased to avoid any irre- versible damage to the structure. After the accelerating struc- ture was fully conditioned with RF macro pulses under each of the two operation modes, the BDR was calculated to be less than /pulse for both.

Low-Energy Mode Measurement Results When the pulse compressor is detuned, RF power is cou- pled into the accelerator without pulse compression, thereby enabling low-energy mode operation under a heavy pulse cur- rent of 100 mA , as depicted in Fig. (b). Measurements in the low-energy mode specifically includes the beam energy, pulse current and X-ray dose rate generated by bremsstrahlung.

(c) Pulse current of the low-energy mode. (PC stands for the pulse compressor.) (d) Schematic of the high-power RF experiment in the high-energy mode. (e) Pulse current of the high-energy mode. (PI stands for the phase inverter.) An integrated current loop was employed for pulse cur- rent measurement.

The loop sensitively detects magnetic field variations induced by electron beams and convert such changes into electrical signals that are routed to an oscillo- scope via a BNC interface for observation.

As shown in 116 mA with a pulse width of , which is shorter than pulse width of the high-voltage modulator. This dis-

crepancy arises from two technical considerations. First, to 604

avoid voltage fluctuations at pulse edges, the pulse width of klystron drive signal (output from the SSA) was set to approx- imately , ensuring operation within the flat-top region of the modulator pulse and thus stable microwave power out-

put. Second, during the initial part of the microwave pulse, 609

the accelerating field has not yet reached steady state, leading to poor electron bunching efficiency. Most electrons in this phase are lost within the accelerating structure, further reduc- ing the effective beam pulse width to the observed The electron energy in the low-energy mode was measured via current attenuation method, where 0.1 steel sheets

terminated aluminum collector were used to char- acterize the beam attenuation. By measuring current signals with different number of steel sheets and fitting the data to GB/T 25306-2010, the electron energy was determined to be , consistent with the designed value.

A water-cooled tungsten target was mounted at the accelerator exit, facilitating X-ray generation through bremsstrahlung. The X-ray dose rate was measured using

a Farmer-type ionization chamber positioned at 1 m from 624

the tungsten target. At a repetition frequency of

150 Hz

and duty factor of , the measured dose rate was . The experimental results of the low-energy mode are listed in table Parameter Value Repetition Rate

150 Hz

Pulse Width Duty Factor High-Energy Mode Measurement Results When the pulse compressor is tuned to the

3 GHz

oper- ating frequency, the RF power undergoes pulse compression and amplitude modulation, generating a 400 ns flat-top pulse with the gain factor of 2.2, as depicted in Fig. (d). This en- ables the linac to operate in high-energy mode with the pulse current of 10 mA Pulse current measurements in the high-energy mode reused the same integrated current loop as in the low-energy mode experiment, as shown in Fig.

Before activat- ing the phase inverter, the RF power from klystron failed to undergo effective pulse compression, leading to reduction in

field magnitude and phase mismatch in the accelerating struc- 643

ture. This resulted in ineffective electron capture and accel- eration, with no noticeable current observed, as indicated by to produce electron beams without inverse signal excitation,

aligning with theoretical predictions. 648

Following phase inverter activation, an obvious pulse cur- rent signal was recorded, as indicated by the red curve. Based on the low-power RF experiment in Section II.B, the pulse compressor can actually output a pulse width of 500 ns , but 400 ns meets the expected power gain. Thus, the emis-

sion duration of the electron gun is set to 400 ns . The rising

edge during the initial stage of the pulse current characterizes 655

the process by which the SW electromagnetic field gradually stabilizes within the accelerator and the electron capture ratio increases to the stable value. Consequently, the latter 200 ns of the pulse current reached the expected value, with an aver- age current of . Additionally, non-zero baseline and crosstalk were noted, though they did not affect the identifi- cation of the pulse current. The non-zero baseline is likely caused by imperfect grounding, while crosstalk may arise from electromagnetic interference in the experimental plat- form or high-voltage line coupling, which requires resolution through shielding optimization or equipment calibration.

The electron energy spectrum was measured using a mag- netic analyzer, whose physical image and schematic diagram are shown in Fig. (a) and Fig. (b), respectively. The mea- surement setup was mainly composed of three parts: the beam transport device, the magnetic analyzer, and the imaging sys- tem. The vacuum pump was connected to the magnetic an- alyzer via a tee joint and a bellows. To improve the energy spectrum resolution, a slit was installed at the entrance of the magnetic analyzer for geometric collimation. By adjusting the excitation current, the deflection magnetic field strength can be precisely regulated, thereby changing the deflection radius of the electron beam. The electron energy formula con- sidering relativistic effects can be expressed as: 1 + (

E = [ �

where is the deflection magnetic field, and is the deflec- tion radius.

The imaging system utilized optical imaging method.

When the electrons bombarded the YAG screen, fluorescent material was excited, generating visible light reflected by a mirror and then captured by a CCD camera. By inte- grating the pre-calibrated correlation between excitation cur- rent and magnetic field, the electron energy spectrum was accurately derived through analyzing the spatial distribution and intensity variation of the optical signals. The measured energy spectrum peak is at

9 MeV

, with a FWHM of

43 MeV

, as shown in Fig. (c). This result slightly ex- ceeds the high-energy mode design target of 13.5 MeV, which is speculated to originate from the actual power attenuation in the high-energy experiment being less than the expected value . The actual power attenuation calculated from the measurement results is -0.73 dB.

Notably, there is a discrepancy in the shape of energy spectrum between the experiment (Fig. (c)) and simulation (Fig. (e)), with an isosceles triangle in the experiment versus a right triangle in the simulation. The discrepancy primarily arising from the inconsistency between the ideal conditions adopted in beam dynamics simulations and the actual experi- mental conditions. In the simulation, the center of the Gaus- sian bunch is locked at the optimal acceleration phase: the bunch center, with the highest charge density, attains the max- imum energy gain, while the head and tail, with lower charge density, experience lower energy gains, resulting in a right triangular with the hypotenuse on the low-energy side. How- ever, in the experiment, the limited phase-locking precision prevents the center of the bunch from being accurately aligned with the optimal acceleration phase, leading to slightly higher energy gains for electrons in the head or tail of the bunch. Ad- ditionally, a relatively wide collimating slit before the mag- netic analyzer may cause some electrons with large transverse velocities to be collected and misidentified as high-energy electrons. Combined with the effects of factors such as the

initial energy spread of the bunch, non-ideal charge density 718

distribution, and higher-order modes in the cavity, the energy spectrum ultimately exhibits an isosceles triangle shape. The experimental results of high-energy mode are listed in table Parameter Value Repetition Rate

20 Hz

Pulse Width 200 ns Duty Factor mission, corresponding to a power retention ratio, a cor- is induced, yielding an energy of

15 .

5 MeV

under lossless conditions. Further minimizing the 728

power attenuation of RF components would enhance electron energy in the high-energy mode. icated to radiotherapy applications was first developed in the

Accelerator Laboratory of Tsinghua University. The accel- 734

source, coupled with a SLED pulse compressor, enabling dual-mode operation: low-energy mode delivering 100 mA electron radiotherapy. With lower treatment costs and spatial

requirements, the linac can significantly facilitate the popu- 741

larization of cancer therapy. Design, fabrication and cold test of the pulse compres- sor and accelerating structure have been conducted, along- side the establishment of the high power RF experimental platform.

Low-energy mode experiments generated electron beams with 116 mA pulse current and

05 MeV

ergy at

150 Hz

repetition rate, demonstrating a dose rate of at 1 m from the tungsten target. High-energy mode experiments yielded electron beams featuring 200 ns pulse width, pulse current,

9 MeV

central energy

43 MeV

FWHM. The results validated the feasibility of the linac, demonstrating its potential for versatile radio- therapy applications with optimized spatial compactness.

The linac is currently undergoing system integration de- sign, and preparations are underway for subsequent biologi- cal experiments in collaboration with hospitals. This compact

X-band dual-mode linac provides valuable references for the development of future compact medical linacs.

F. Bray, M. Laversanne, H. Sung, et al ., Global cancer statis- tics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 74(3) , 229-263 (2024).

R. Atun, D. Jaffray, M. Barton, et al ., Expanding global ac- cess to radiotherapy. Lancet. Oncol. 16(10) , 1153-1186 (2015).

S. C. Huang, H. Zhang, K. Bai, et al ., Monte Carlo study of the neutron ambient dose equivalent at the heavy ion medical machine in Wuwei. Nucl. Sci. Tech. 33(9) , 119 (2022). 10.1007/s41365-022-01093-z

[4] W. Wang, X. X. Yuan, X. H. Cai, et al ., A beam range monitor 771

based on scintillator and multi-pixel photon counter arrays for heavy ions therapy. Nucl. Sci. Tech. 33(10) , 123 (2022). 10.1007/s41365-022-01113-y H. H. Xiao, L. L. Liu, W. Y. Li, et al ., TLD calibration and absorbed dose measurement in a radiation-induced liver injury model under a linear accelerator. Nucl. Sci. Tech. 34(4) (2023).

Y. N. Gu, W. J. Zhao, X. G. Cao, et al ., Feasibility study of the photonuclear reaction cross section of medical radioisotopes using a laser Compton scattering gamma source. Nucl. Sci.

Tech. 35(9) , 155 (2024).

[7] J. Chen, Z. Hu, J. Tong, et al ., Study of BNCT neutronics op- 783

timization for out-of-beam dosimetry based on radiobiological figures of merit. Nucl. Instrum. Meth. B. , 1–9 (2021).

Y. Shi, M. Z. Zhang, L. H. Ou-Yang, et al ., Design of a rapid- cycling synchrotron for flash proton therapy. Nucl. Sci. Tech. 34(10) , 145 (2023).

M. Z. Zhang, D. M. Li, L. R. Shen, et al ., SAPT: a synchrotron- based proton therapy facility in Shanghai. Nucl. Sci. Tech. 34(10) , 148 (2023).

Z. P. Qiao, Y. C. Hu, Q. X. Jiang, et al ., Coin-structured tunable beam shaping assembly design for accelerator-based boron neutron capture therapy for tumors at different depths and sizes. Nucl. Sci. Tech. 34(12) 186 (2023). 10.1007/s41365-023-01325-w D. Ha, S. Lee, M. Ghergherehchi, et al 6MV X- band linear accelerator for stereotactic body radiation ther- apy. Nucl. Eng. Technol. 56(11) , 4502-4511 (2024).

D. Ha, S. H. Lee, M. Ghergherehchi, et al ., High preci-

sion tuning of RF cavity for 6MeV SKKU X-band medical 803

LINAC. J. Korean Phys. Soc. 85(7) , 591-599 (2024). 10.1007/s40042-024-01151-2 M. Uesaka, T. Natsui, K. Lee, et al ., 950keV, 3.95MeV and 6MeV X-band linacs for nondestructive evaluation and medicine. Nucl. Instrum. Meth. A. 657(1) , 82-87 (2011).

S. H. Lee, S. W. Shin, J. Lee, et al ., X-band Linac for a 6MeV dual-head radiation therapy gantry. Nucl. Instrum. Meth. A. , 40-45 (2017).

S. V. Kutsaev, R. Agustsson, A. Arodzero, et al ., Com- pact X-Band electron linac for radiotherapy and security ap- plications. Radiat. Phys. Chem. , 109494 (2021).

B. E. Bjärngard, G. T. Y. Chen, R. W. Piontek, et al ., Analysis of dose distributions in whole body superficial electron therapy.

Int. J. Radiat. Oncol. Biol. Phys. , 319-324 (1977). 10.1016/0360-3016(77)90090-6 I. J. Das, K. R. Kase, J. F. Copeland, et al ., Electron beam modifications for the treatment of superficial malignancies. Int.

J. Radiat. Oncol. Biol. Phys. 21(6) , 1627-1634 (1991). 10.1016/0360-3016(91)90342-2 D. Krim, B. Whelan, M. Harkness, et al ., Monte Carlo sim- ulation of a novel medical linac concept for highly conformal x-ray FLASH cancer radiotherapy. Sci. Rep. , 17604 (2025).

[19] M. V. Zheltonozhskaya, P. D. Remizov, M. V. Lenivkin, 829

et al ., Study of Light Nuclei Activation in Human Blood During High-Energy Linear Accelerator (Linac) Radiation

Therapy. Moscow Univ. Phys. 78 , 557–563 (2023). 832

10.3103/S0027134923040239 J. M. Schippers, M. Seidel, et al ., Operational and de- sign aspects of accelerators for medical applications, Phys.

Rev. Accel. Beams. , 034801 (2015). RevSTAB.18.034801 Z. D. Farkas, H. A. Hoag, G. A. Loew, et al ., SLED: A method of doubling SLAC’s energy, in Proceedings of 9th International Conference on the High-energy Accelerators, Stanford, Cali-

fornia, 1974. 841

I. Syrachev, V. Vogel, H. Mizuno, et al ., The results of RF high power tests of X-band open cavity RF pulse compression system VPM (JLC), in Proceedings of the 1994 International LINAC Conference, Tsukuba, Japan, 1994.

S. G. Tantawi, C. D. Nantista, V. A. Dolgashev, et al ., High- power multimode X-band rf pulse compression system for future linear colliders. Phys. Rev. Accel. Beams. , 042002 (2005).

J. W. Wang, S. G. Tantawi, C. Xu, et al ., Development for a supercompact X-band pulse compression system and its appli- cation at SLAC. Phys. Rev. Accel. Beams. , 110401 (2017).

P. Wang, J. R. Shi, H. Zha, et al ., Development of an S-band spherical pulse compressor. Nucl. Instrum. Meth. A. , 84- 91 (2018).

P. Wang, H. Zha, I. Syratchev, et al ., RF design of a pulse com- pressor with correction cavity chain for klystron-based com- pact linear collider. Phys. ReV. Accel. Beams. , 112001 (2017).

J. Y. Liu, J. R. Shi, H. Zha, et al ., Analytic RF design of a linear accelerator with a SLED-I type RF pulse compressor.

Nucl. Sci. Tech. 31(11) , 107 (2020). X. C. Lin, H. Zha, J. R. Shi, et al ., X-band two-stage rf pulse compression system with correction cavity chain. Phys. Rev.

Accel. Beams. : 120401 (2022). celBeams.25.120401 S. M. Hanna, Applications of X-band technology in medical accelerators, in Proceedings of the 1999 Particle Accelerator Conference, New York, USA, 1999.

M. M. Peng, J. R. Shi, H. Zha, et al ., Development and high- gradient test of a two-half accelerator structure. Nucl. Sci.

Tech. 32(6) , 60 (2021).

[31] J. R. Shi, A. Grudiev, W. Wuensch, et al ., Tuning of X-band 875

traveling-wave accelerating structures. Nucl. Instrum. Meth. A. , 14 (2013).

[32] X. C. Lin, H. Zha, J. R. Shi, et al ., Fabrication, tuning, and 878

high-gradient testing of an X-band traveling-wave accelerating structure for VIGAS. Nucl. Sci. Tech. , 102 (2022). 10.1007/s41365-022-01086-y X. C. Lin, H. Zha, J. R. Shi, et al ., A compact X-band backward traveling-wave accelerating structure. Nucl. Sci. Tech. 35(5) 40 (2024).

Q. Gao, H. Zha, J. R. Shi, et al ., Design and test of an X-band constant gradient structure. Phys. ReV. Accel.

Beams. , 090401 (2024). Beams.27.090401 J. Gao, H. Zha, J. R. Shi, et al ., Design, fabrication, and testing of an X-band 9-MeV standing-wave electron linear accelerator.

Nucl. Sci. Tech. 34(7) , 110 (2023). J. Gao, H. Zha, J. R. Shi, et al ., Study of wakefield-induced beam breakup instability on an X-ban accelerating struc- ture. IEEE Trans. Nucl. Sci. , 2542-2552 (2023). 10.1109/TNS.2023.3322220 L. Y. Zhou, H. Zha, J. R. Shi, et al ., A non-invasive diagnostic

method of cavity detuning based on a convolutional neural net- 898

work. Nucl. Sci. Tech. 33(7) , 94 (2022). 022-01069-z J. H. Shao, Y. C. Du, H. Zha et al., Development of a C-band 6 MeV standing-wave linear accelerator. Phys.

Rev. Accel. Beams. , 090102 (2013). RevSTAB.16.090102 X. C. Lin, H. Zha, J. R. Shi, et al ., Design, fabrication, and testing of low-group-velocity S-band traveling-wave acceler- ating structure. Nucl. Sci. Tech. 33(11) , 147 (2022). 10.1007/s41365-022-01124-9 X. C. Lin, H. Zha, J. R. Shi, et al ., Development of a seven- cell S-band standing-wave RF-deflecting cavity for Tsinghua Thomson scattering X-ray source. Nucl. Sci. Tech. 32(4) (2021).

S. Zhang, C. Meng, Z. S. Zhou, et al ., Design of 10 MeV electron linear accelerator for space environment simulation.

Nucl. Sci. Tech. 35(10) , 177 (2024).

[42] O. A. Ivanov, M. A. Lobaev, A. L. Vikharev, et al ., Ac- 917

tive microwave pulse compressor using an electron-beam trig- gered switch. Phys. Rev. Lett. , 115002 (2013). 10.1103/PhysRevLett.110.115002 T. P. Wangler, RF linear accelerators, 1st edn. (Wiley, Hoboken, 2008).

A. Grudiev., Design of compact high power RF components at X-band. CLIC Notes, , (2016).

S. A. Barengolts, V. G. Mesyats1, V. I. Oreshkin, et al ., Mech-

anism of vacuum breakdown in radio-frequency accelerating 926

structures. Phys. Rev. Accel. Beams. , 061004 (2018). 10.1103/PhysRevAccelBeams.21.061004 Z. F. Xiong, C. Cheng, J. Yu, et al ., Switching speed ef- fect of phase shift keying in SLED for generating high power microwaves. Chin. Phys. C. : 017006 (2016). 10.1088/1674-1137/40/1/017006

[47] A. Degiovanni, W. Wuensch, G.N. Jorge, Comparison of the 933

conditioning of high gradient accelerating structures. Phys. 934

Rev. Accel. Beams. , 032001 (2016). RevAccelBeams.19.032001 J. H. Shao, S. P. Antipov, S. V. Baryshev, et al ., Observa- tion of Field-Emission Dependence on Stored Energy. Phys.

Rev. Accel. Beams. , 264802 (2015). RevLett.115.264802 X. W. Wu, J. R. Shi, H. B. Chen, et al ., High-gradient break- down studies of an X-band Compact Linear Collider prototype structure. Phys. Rev. Accel. Beams. , 052001 (2017). 10.1103/PhysRevAccelBeams.20.052001 J. R. Shi, H. B. Chen, C. X. Tang, et al ., Fabrication and low- power testing of an L-band deflecting cavity for emittance- exchange at ANL, in Proceedings of 11th European Particle Accelerator Conference, Genova, Italy, 2008.