Generation of 565 MW of $X$-band power using a metamaterial power extractor for structure-based wakefield acceleration

We report the generation of up to 565 MW, 2.7 ns (FWHM) pulses at 11.7 GHz from a metamaterial structure in test at the Argonne Wakefield Accelerator. The highest power was generated by a train of eight 65 MeV electron bunches spaced at 1.3 GHz with a total charge of 355 nC. The metamaterial structure consists of 100 copper unit cells each consisting of a wagon-wheel plate and a spacer plate with a total structure length of 0.2 m. The 565 MW pulse generates a wakefield with a peak on-axis gradient of $135\mathrm{MV}/\mathrm{m}$ that could be used to accelerate a trailing main bunch. An estimated surface electric field of over $1\mathrm{GV}/\mathrm{m}$ is generated on the metamaterial plates at the peak power level but no evidence of breakdown was observed during testing. Tests with single electron bunches and with trains of bunches of up to 100 nC produced output power levels in excellent agreement with simulations. At higher total bunch charge, offsets of the bunches from the axis resulted in beam interception and reduced output power. Simulations indicate that a perfectly aligned bunch train would generate more than 1 GW of power from the structure.

Figure 1

Schematic of two-beam structure-based wakefield acceleration. Rf power is generated from a drive beam (green) by the power extractor structure. This rf pulse is transported through a waveguide to a separate accelerator structure, where the main beam (blue) is accelerated.

Figure 2

Top: exploded view of two wagon wheel metamaterial unit cells. Each cell consists of a structure plate and a spacer plate, each 1 mm in thickness. A beam traversing the center of the cells will radiate backwards in reverse Cherenkov radiation. Bottom: dimensions of the wagon wheel structure plate.

Figure 3

Simulated dispersion curve for the wagon wheel metamaterial. The rf generated by the interaction with the beam will be at the frequency determined by the intersection of the dispersion curve and the beamline.

Figure 4

Comparison of ${\mathrm{S}}_{21}$ measurements showing the reduction in transmission loss in the Stage III experiment.

Figure 5

The asymmetric coupler in the Stage I/II experiments (top right) generated a vertically-asymmetric mode in the coupling cell (top left). The symmetric coupler implemented in the Stage III experiment (bottom right) eliminated this asymmetric mode (bottom left) and the associated transverse kick. The field plots show the normalized magnitude of $E_y$ and $E_x$, respectively.

Figure 6

The asymmetric mode in the Stage I/II coupling cell generated transverse electric field components in the fundamental mode. The magnitude of the deflecting field normalized to the accelerating gradient is shown in solid blue for the asymmetric design. The symmetry of the Stage III coupler eliminates this transverse field, as shown in dashed red.

Figure 7

The symmetric coupler design improves structure performance for a misaligned beam. For a set of 8 bunches centered 1 mm off-axis, the symmetric coupler (solid trace) produces 1.6x higher peak power and 1.4x more pulse energy than the asymmetric coupler design (dashed trace) by helping maintain linear bunch trajectory.

Figure 8

A CAD model of the Stage III structure, including symmetric high power output coupler and asymmetric low-power forward coupler.

Figure 9

$E_{\mathrm{pk}}/E_{\mathrm{acc}}$, the ratio of peak surface field to accelerating field, as a function of wagon wheel edge radius.

Figure 10

Assembled Stage III metamaterial power extractor at MIT prior to initial cold tests.

Figure 11

S-parameters of the assembled structure were measured with a VNA. ${\mathrm{S}}_{21}$ (top) and ${\mathrm{S}}_{11}$ (bottom) show excellent agreement with simulation.

Figure 12

Simulated dispersion compared to measured dispersion from beadpull cold test. The intersection point with the light line shows the experimental interaction frequency of 11.675 GHz.

Figure 13

Simulated (normalized) axial electric field compared to measured values from beadpull experiment at 11.675 GHz.

Figure 14

A simplified model of the AWA beamline setup during the Stage III experimental run. Bunch trains were monitored using YAG screens before and after the structure and charge was measured with ICTs. Dipole trim (not shown) and quadrupole magnets allow for steering. Figure modified from [46].

Figure 15

Vacuum chamber in which the power extractor was installed on the AWA beamline. Both the high power and low power rf output waveguides were sampled with a directional coupler and breakdown-detection diodes (not shown), with the power absorbed by rf loads at the end of the waveguide.

Figure 16

Schematic of breakdown detection system. A spare ${1.33}^{\prime \prime }$ port in the output waveguides is fitted with a taper, separated by a porous gasket to reject rf leakage. The collimator outside the window picks up light and focuses into a fiber optic cable. The light is detected by a fast diode detector and monitored on an oscilloscope.

Figure 17

Experimental rf pulse generated by a single 7 nC bunch, along with accompanying PIC simulation results.

Figure 18

FFT of the rf pulse in Fig. 17 from a single 7 nC bunch and accompanying PIC simulation results.

Figure 19

Rf Pulse of a eight bunch train with total charge of 12 nC, along with accompanying PIC simulation results.

Figure 20

FFT of the rf pulse in Fig. 19 from an eight bunch train with total charge of 12 nC bunch, along with accompanying PIC simulation results.

Figure 21

Plot of output power vs incident charge for a sample of 600 experimental shots from Run 1 and Run 2 compared to theory. The deviation from theory starts to appear above 100 nC of total charge.

Figure 22

YAG images taken before (upstream from) the power extractor show an bunch-to-bunch offset that is on the order of the bunch diameter. This is believed to be due to undamped transverse wakes in the linear accelerator. The total charge in this bunch train is 230 nC.

Figure 23

Experimental plot of typical 350 MW experimental trace in solid blue, from a 224 nC 8-bunch train. The red dashed simulation assumes an 224 nC eight-bunch train that centered on-axis. The yellow dot-dashed simulation assumes the same bunch train but offset by 0.05 mm per bunch and tilted by 0.025° per bunch.

Figure 24

Experimental trace showing 510 MW of generated output power at 11.7 GHz. The charge measured before entering the structure was 280 nC. The red simulation assumes the bunch train is offset by 0.25 mm per bunch and tilted by 0.12° per bunch.

Figure 25

Experimental trace showing 565 MW of generated output power at 11.7 GHz. The charge measured before entering the structure was 355 nC. The red simulation assumes the bunch train is offset by 0.3 mm per bunch and tilted by 0.21° per bunch.

Figure 26

Comparison of plate surface finish after one, two, three, and seven baths. Surface damage, circled, is clearly visible after a single 2% Citranox bath. While two baths still leave obvious surface defects, plates started to exhibit noticeable pitting after more than four baths (seven bath case is shown for clarity).

Figure 27

Visible microscope images of the edges of plates before (left) and after (right) treatment with Citranox bath procedure. Edges are visibly rounded by the process.

Figure 28

(Angled View) Scanning electron microscope images of the plate edge from before (top) and after (bottom) Citranox treatment.