The proposed VHEE radiotherapy structure comprises a power source section and an accelerator section. The power source section occupies a larger area and incurs higher costs. Calculations indicate that using only a 50 MW klystron would require approximately 4 m to accelerate electrons to 100 MeV, while employing multiple klystrons would significantly increase both the spacial requirements and costs. To address this, we have opted to use a pulse compressor capable of achieving a flat-top power gain of three times [30]. This approach increases the gradient of the acceleration structure, allowing the VHEE radiotherapy system to remain compact while utilizing only one klystron.

Figure 2 illustrates the condition and results of the high-power test for the two-stage pulse compressor when it operates at the first-stage. The X-band amplitude modulation (AM) cavity chain features a novel design, where resonant cavities are connected at both ends of the dual-polarization mode coupler. This design reduces both transmission loss and the length of the AM cavity chain to half of its original size. The length of the first-stage compression part of the X-band pulse compression system is only 0.4 m. In high-power tests, a flat-top power gain of three times was achieved, with the amplitude and phase of the flat-top demonstrating good stability, making multi-bunch acceleration feasible. Thus, the first-stage compression section is well-suited for use in the compact VHEE radiotherapy structure.

For the VHEE radiotherapy facility, we developed a prototype of an X-band Backward Traveling Wave (BTW) accelerating structure intended to bunch a DC electron beam and accelerate it to approximately 8 MeV [31]. After optimizing the cavity shape, field distribution, and beam dynamics, the BTW structure was simulated and designed using CST Studio Suite [32]. The simulation results demonstrated a field distribution that closely matches that of the constructed prototype. Additionally, the time-domain circuit model was utilized to analyze the transient beam parameters of the buncher during its unsteady state.

The BTW structure was fabricated at Tsinghua University and underwent high-power test at the Tsinghua X-band High-power Test Stand, as illustrated in Figure 3. The structure was powered by 5 MW microwave pulses from the pulse compressor, operating in one-stage mode (with the second compressor detuned), achieving an average gradient of 45 MV/m. The electron beam was successfully accelerated to 8 MeV, with a pulse current of 108 mA. A comparison between the designed and measured values of the BTW structure is presented in Table 3, where the measured capture rate was calculated by dividing the target current by the gun current. Although some beam parameters did not fully meet the design specifications, the test preliminarily verified the high gradient and large current capabilities of the BTW structure. Future work will focus on addressing fabrication errors in the output coupler to ensure the designed capture rate, and optimizing the magnetic coupling holes between adjacent cells with smooth rounding to improve high-gradient performance.

A constant-impedance traveling-wave structure, consisting of 72 cells operating in the 2 π / 3 mode, named TTX-XC72 [33], was designed and fabricated for the Very Compact Inverse Compton Scattering Gamma-ray Source (VIGAS) program at Tsinghua University [34]. Although XC72 could reach almost 80 MV/m,its breakdown rate (BDR) is about 1 0 − 3 /pulse m, which is slightly higher than we required, mostly because the field in the first cell is too high. Therefore, we decided to switch to the constant gradient (CG) approach. Recently, the CG prototype XT72 was developed [35]. This structure is also suitable for use in the main acceleration section of the compact VHEE radiotherapy facility.

The XT72 structure comprises 70 accelerating cavities along with input and output couplers. The fabricated cells are shown in Figure 4. When an 80 MW pulse with top drop from the pulse compressor is introduced into the structure, it achieves an average acceleration gradient of 80 MV/m. With a total length of 0.63 m, the structure provides an energy gain of 50 MeV for electrons. The XT72 prototype underwent high-power test, demonstrating better performance compared to the XC72 structure, with an achieved gradient of 81.0 MV/m versus 78.7 MV/m, and a lower breakdown rate (BDR) of 1.5 × 1 0 − 4 . Additionally, XT72 exhibited a lower maximum surface electric field and reduced temperature rise from pulse heating. Table 4 shows the parameters of XT72.

The optimization results of the BTW’s geometric parameters are shown in Figure 7, where the first eight cells are bunching cells and the rest are accelerating cells.

Figure 8A shows the optimized energy distribution at the beam line exit. Within a macro-pulse, 24 nC of electron charge falls within the 100 ± 2 MeV range, while 11 nC lies outside. A significant proportion of the charge falls outside the specified energy range, which can be attributed not only to the tail generated during the hot cathode bunching process, but also the energy reduction caused by the unsteady state of the acceleration structure. Figure 8B, C illustrate the temporal variations in beam parameters within a macro pulse, highlighting a noticeable drop in capture rate at 100 ns due to field changes in the bunching section. Between 80 and 130 ns, before the main acceleration section reaches steady-state, captured electrons are only accelerated to lower energies. After 270 ns, as the input power decreases and residual electric field remains in the accelerating tube, captured electrons are again accelerated to a lower energies. Calculations show that electrons produced during these unsteady states contribute to 76% of the tail distribution.

Directly using the electron beam of Figure 8A in VHEE radiotherapy experiments would result in significant background, making it necessary to eliminate or suppress the tail. The most straightforward method is to place a dipole and a slit at the beam exit to filter the electrons by energy, as shown in Figure 9A. While effective in removing the tail, this approach requires a large magnet deflection radius which poses shielding challenges.

The second method is to position the dipole and slit at the exit of the bunching section. Since the beam energy is lower at this point, the required deflection radius is smaller, making shielding easier. However, calculations indicate that this method can only filter out 55% of the tail, as shown in Figure 9B. This limitation arises because, while some electrons’ energy falls within the specified range upon exiting the bunching section, their large phase deviation causes them to fall outside this range after passing through the acceleration section. Moreover, placing a dipole at the bunching section exit complicates both the microwave and beam systems.

The third method involves using a grid-controlled electron gun with precise beam control, activating the beam when the accelerating structure is nearly filled and deactivating it when input power starts to decline. Since that electrons produced during unsteady states contribute to 76% of the tail distribution, keeping the starting moment of electron emission close to the moment when the accelerator field reaches a steady state may greatly reduce the tail. Considering the beam loading effect, the activating moment is chosen to be when the accelerating structure is nearly filled instead of full filled. The specific activating moment needs to be compared and optimized. Despite this, the unsteady process caused by wakefield effects during beam injection still prevents complete tail removal, as illustrated in Figure 9C.

In summary, while directly screening electrons’ energy at the exit of beam line needs a larger dipole, it is more effective. Future work would be necessary to deal with shielding challenges.

Based on the optimization of the output charge, the Monte Carlo simulation software GEANT4 [37] was employed to preliminarily calculate the depth-dose distribution of the VHEE beam. To simplify the model, the electron beam was assumed to be parallel after expansion. The parallel beam traverses a 50 cm air gap and then interacts with a 40 cm water layer. The depth-dose distributions at the center of the beams, with cross-sections of 3 × 3   cm 2 and 10 × 10   cm 2 , for various energies, are displayed in Figure 10.

It is clear that higher energy levels allow electrons to penetrate deeper into the material. The designed output energy of the VHEE beam line in this study is 100 MeV. By comparing Figures 10A, B, it can be observed that a smaller beam cross-section results in a more rapid fall-off in the depth dose. This is due to increased scattering experienced by electron beams with smaller cross-sections. The dose at a given depth is approximately inversely proportional to the beam’s cross-section area.

For dose estimation, a dose-to-charge ratio of 20 Gy/ μ C • 10 × 10 c m 2 is used. This implies that a 1 μ C electron beam expanded to a 10 × 10 c m 2 area can deliver a 20 Gy dose at the target. This ratio is inversely proportional to the cross-sectional area of the expanded beam and can be applied to estimate the radiation field and dose rate.

Given that the beam line outputs 24 nC per macro pulse and operates at a repetition rate of 40 Hz, the estimated dose rate under a 10 × 10   cm 2 radiation field is: 20 Gy / μ C × 24   nC × 40   Hz = 19.2   Gy / s

When the radiation field is reduced to less than 6 × 6   cm 2 , the dose rate can exceed 40 Gy/s, which is ideal for FLASH radiotherapy research. Future system upgrades, such as increasing the repetition rate of the modulator, could further enhance the dose rate of the entire setup. In addition, it should be noted that this section’s dose calculation is only a rough approximation and could only preliminary confirm the beam line’s ability to deliver VHEE at UHDR. After the beam line is constructed and tested in the future, detailed dose simulation would be performed to compare the measured results.