betaFeedback

WorksPhysics and Astronomy

High-transformer-ratio plasma-based afterburner for extending X-ray free-electron laser photon energies to ≳\gtrsim50 keV

Abstract

Bright coherent X-rays in the ≳\gtrsim50 keV range are highly desirable for applications in materials science, high-energy-density physics, and nuclear physics, but are difficult to obtain from present X-ray free-electron laser (XFEL) facilities. Plasma-based afterburners offer a compact route to increasing the electron-beam energy, yet high transformer ratio (TR) operation is vulnerable to hosing instability, which can severely degrade the witness beam quality. We propose using a transversely elliptical driver beam to excite an elliptical nonlinear plasma wake with an axially varying focusing gradient, thereby suppressing the witness beam hosing through betatron decoherence. Three-dimensional particle-in-cell simulations show that, for a 3 kA driver and a 2.35 kA witness beam with an rms length of 4 μ\mathrmμm and a beam separation of 94 μ\mathrmμm (TR ≈3\approx 3), this scheme can accelerate an 8 GeV witness beam to ∼30\sim30 GeV while nearly half of the electrons meet the XFEL requirements. Start-to-end simulations show that the resulting beam can drive ≳50\gtrsim50 keV XFEL radiation with peak powers of ∼10\sim10 GW and durations of ∼10\sim10 fs, even when quantum diffusion is included, offering a compact route to extend existing XFEL facilities to higher photon energies.

Introduction

X-ray free-electron lasers (XFELs) can deliver X-rays with unprecedented brightness[1, 2], ultrashort pulse duration, and exceptional transverse coherence, enabling applications across many disciplines, such as nonlinear X-ray physics[3, 4, 5], ultrafast science[6, 7, 8], and coherent diffractive imaging[9, 10, 11, 12]. Limited by facility footprint and construction cost, most XFELs employ high-quality electron beams with energies of ∼10\sim10 GeV to generate X-ray photons of ∼1\sim1 to ∼20\sim20 keV in magnetic undulators with a length of an order of 100 m[2, 13, 14, 15, 16, 17, 18, 19]. Extending the photon energy to the several 10s to 100 keV range is highly desirable[20] for advancing applications in fields such as deep-penetration materials science[21, 22, 23], physics under extreme conditions[24], and nuclear quantum optics[25]. For example, high-energy diffraction microscopy (HEDM) utilizes 50∼\sim100 keV X-rays to resolve grain-level three-dimensional structure and dynamics within bulk polycrystalline materials, which is beyond the reach of current XFELs[21, 22].

Shortening the undulator period λu\lambda_u or using a smaller undulator deflection parameter KK can increase the XFEL photon energy, but at the expense of a longer saturation length and reduced saturation power[1, 26]. In contrast, increasing the electron beam energy raises the photon energy without decreasing the saturation power, making it the most effective approach. Plasma-based afterburners (PBAs)[27, 28, 29, 30] with ∼10\sim10 GV/m acceleration gradient have been proposed as a means of boosting the electron beam energy at XFEL facilities by avoiding substantially increasing the facility footprint and cost[31, 32]. In a PBA, one electron beam from the radio-frequency (RF) accelerators is used as a driver to excite a nonlinear plasma wake[33, 34, 35] and accelerate another RF accelerator beam - the witness beam.

Although they have demonstrated 10-GeV-class energy gains in meter-long plasmas[28, 29, 36], it is challenging to preserve the beam quality to meet the stringent requirements posed by XFELs, i.e., ϵn≲γbλr4π\epsilon_n \lesssim\gamma_b \frac{\lambda_r}{4\pi} and Δγbγb≲ρ\frac{\Delta\gamma_b}{\gamma_b} \lesssim\rho, where λr\lambda_r is the radiation wavelength, ρ\rho is the FEL parameter characterizing FEL gain, ϵn,Δγbmc2,γbmc2\epsilon_n,\Delta\gamma_b mc^2,\gamma_b mc^2 are the normalized emittance, the slice energy spread and the average energy of the electron beam, respectively, mm is the electron mass and cc is the speed of light in vacuum[1, 26].

A recent work[32] studied the possible reasons of the beam quality degradation in a PBA, including axial chirp of the acceleration gradient[37, 38, 39] and ion motion[40, 41, 42], and showed that more than half of the witness beam can satisfy the subsequent XFEL requirements when utilizing the beam loading[43]. The transformer ratio (TR)[44, 45, 46, 47] which is defined as the ratio of the maximum accelerating field experienced by the witness beam to the maximum decelerating field acting on the driver reaches ∼1\sim1. To further improve the witness beam energy gain and the corresponding XFEL photon energy, a PBA operated with a large TR is needed which is susceptible to hosing instability (HI) seeded by misalignment between the driver and witness beams and/or the asymmetry of the beams[50, 51]. HI would wiggle the beams, amplify the transverse position and angle jitters, and degrade the beam quality rapidly, and it is thus necessary to suppress it. Several schemes have been proposed to suppress the HI through a decoherence of the electron betatron oscillations, e.g., applying a large energy chirp[52, 53], increasing the slice energy spread[52, 54, 55], utilizing the axial varying focusing gradient in the quasilinear regime[56] and a substantial ion motion[57, 58]. However, these schemes all degrade the beam quality and are therefore only applicable to suppressing the HI of the driver beam.

In this work, a transversely elliptical driver beam is employed to excite an elliptical nonlinear plasma wake with an axially varying focusing gradient, thereby mitigating the HI of the witness beam and maintaining its quality. We combine theoretical analysis with three-dimensional (3D) particle-in-cell (PIC) simulations to investigate the performance of the TR of the PBA and the usable fraction of the witness beam under different parameters of the driver and the witness beams. Simulations indicate that, in the axisymmetric wake generated by a round driver at TR ≈3\approx3, the HI seeded by the beam misalignment can cause significant particle loss and catastrophic quality degradation. In contrast, driver beams optimized in spot size and ellipticity can suppress HI almost completely with limited growth of the slice energy spread. Under these conditions the initial 8 GeV witness beam can be accelerated to ∼30\sim30 GeV while retaining its initial quality, enabling the production of ≳50\gtrsim50 keV x-rays with 10s GW peak power, even after accounting for quantum diffusion. The proposed PBA scheme can be integrated into contemporary XFEL facilities, substantially extending their capability for high-photon-energy x-ray generation and benefiting hard-x-ray users in nuclear physics, high-energy-density physics, and materials science.

High-quality PBA with a high TR

We consider contemporary XFEL facilities and set the energy of the electron beams from the RF accelerators to 8 GeV without loss of generality$^{2,13\text{--}19}$. As shown in Figure 1(a), two temporally separated electron beams propagate into a uniform plasma with a density of n0=7×1016 cm−3n_0 = 7 \times10^{16}\ \mathrm{cm}^{-3} where the driver transfers its energy to the witness beam through the nonlinear plasma wake. In the linear regime of plasma wakefield acceleration, the TR is limited to 2 for longitudinally symmetric drivers$^{44}$. Asymmetric drivers with slowly rising edge and rapidly falling edge can flatten the decelerating field and achieve a high TR which exceeds 2$^{45,46}$. In the 3D nonlinear blowout regime$^{37,39,59}$, a longitudinally symmetric beam driver with relatively long duration can also achieve a large TR since the profile of the acceleration field is governed by the motion of the boundary of the nonlinear wake rather than the falling edge of the driver$^{50,51,60,61}$.

Figure 1. Schematic and parameter scan of the two-bunch PBA. (a) Plasma electron density distribution together with the driver and witness beams. The plasma electron density is shown in blue, while the driver- and witness-beam density distributions are shown in red and green, respectively. The black curve shows the on-axis accelerating field EzE_z with the witness beam included, and the red curve shows the corresponding unloaded wake. (b–e) Dependence of the transformer ratio and usable fraction η\eta on the witness beam peak current IwI_w and beam separation Δ\Delta for four cases. White contours denote the transformer ratio, and color maps denote η\eta. For panels (b) and (c), the driver current is fixed at Id=2.0I_d = 2.0 kA, with scan step sizes of 0.05 kA in IwI_w and 0.4 μ\mum in Δ\Delta. For panels (d) and (e), the driver current is fixed at Id=3.0I_d = 3.0 kA, with scan step sizes of 0.1 kA in IwI_w and 0.6 μ\mum in Δ\Delta.

To avoid the additional complexity associated with beam shaping, we consider tri-Gaussian distribution for both driver and witness beams in the subsequent discussions. We choose a driver with a root-mean-square (rms) length of σz,d=32 μm\sigma_{z,d} = 32\ \mu\mathrm{m} (kpσz,d=1.6k_p\sigma_{z,d} = 1.6) and consider four representative configurations of driver peak current and witness bunch length: Case I (Id=2I_d = 2 kA, σz,w=4 μm\sigma_{z,w} = 4\ \mu\mathrm{m}), Case II (Id=2I_d = 2 kA, σz,w=6 μm\sigma_{z,w} = 6\ \mu\mathrm{m}), Case III (Id=3I_d = 3 kA, σz,w=4 μm\sigma_{z,w} = 4\ \mu\mathrm{m}), and Case IV (Id=3I_d = 3 kA, σz,w=6 μm\sigma_{z,w} = 6\ \mu\mathrm{m}), where kp=ωpck_p = \frac{\omega_p}{c}, ωp=n0e2mϵ0\omega_p = \sqrt{\frac{n_0e^2}{m\epsilon_0}} is the plasma frequency, ee is the electron charge and ϵ0\epsilon_0 is the vacuum permittivity. For each case, we scan the peak current of the witness (IwI_w) and the beam separation (Δ\Delta) to investigate the TR and the preservation of witness beam quality. The beams are assumed to be transversely matched to the ion column force of the nonlinear plasma wake, i.e., their Twiss parameters satisfy β=2γbkp−1≈3.6\beta= \sqrt{2\gamma_b}k_p^{-1} \approx3.6 mm and α=0\alpha= 0$^{62}$. Achieving this millimeter-scale β\beta using magnetic focusing elements is challenging; however, this can be addressed by employing a properly tailored plasma density upramp to further focus the beam to the required β\beta$^{63,64}$. The initial emittance of the witness beam is set to 0.4 μ\mum, representative of state-of-the-art beam quality, while a larger value of 1.2 μ\mum is assigned to the driver. Their matched spot sizes are σr,d=0.52 μm\sigma_{r,d} = 0.52\ \mu\mathrm{m} and σr,w=0.30 μm\sigma_{r,w} = 0.30\ \mu\mathrm{m}. The plasma length is chosen such that the driver energy is almost fully depleted.

We use 3D quasi-static PIC code QuickPIC to model the PBA stage$^{65,66}$. The simulation setup can be found in Appendix A. The tightly focused driver beam with a peak density of nd≈2.5×1019 cm−3≫n0n_d \approx2.5 \times10^{19}\ \mathrm{cm}^{-3} \gg n_0 (for Id=2I_d = 2 kA) excites a fully blownout plasma wake and an ion column void of plasma electrons is present [Figure 1(a)]. The transverse wakefield in the ion column is linear as Wr=−12mωp2erW_r = -\frac{1}{2}\frac{m\omega_p^2}{e}r, and the acceleration gradient EzE_z is uniform in the transverse plane but varies longitudinally[38, 39]. The linear focusing force with a constant gradient (Wr/rW_r/r) can preserve the beam emittance[62, 67]. However, the axial variation of EzE_z increases the energy spread within a finite-thickness longitudinal slice of the witness beam. In FELs, electrons within one cooperation length, lc≈λr4π3ρl_c \approx\frac{\lambda_r}{4\pi\sqrt{3}\rho}, emit radiation cooperatively[1]. Consequently, the relative energy spread of a slice with thickness lcl_c must satisfy σγbγb<ρ\frac{\sigma_{\gamma_b}}{\gamma_b}<\rho to maintain FEL gain, where σ\sigma denotes the rms value. By assuming a non-evolving EzE_z profile (which is valid for most beam-driven plasma wakefield accelerators), and noting that the cooperation length lc≪λpl_c \ll\lambda_p, such that EzE_z varies approximately linearly over one cooperation length, the growth of the relative slice energy spread of the witness beam can be written as[32]

δγb≡σγbγb=(σγb0mc2)2+(elc23dEzdξL)2γb0mc2+eEzL(1)\delta_{\gamma_b} \equiv\frac{\sigma_{\gamma_b}}{\gamma_b} = \frac{\sqrt{\left(\sigma_{\gamma_{b0}}mc^2\right)^2+\left(e\frac{l_c}{2\sqrt{3}}\frac{\mathrm{d}E_z}{\mathrm{d}\xi}L\right)^2}}{\gamma_{b0}mc^2+eE_zL} \tag*{(1)}

where γb0mc2\gamma_{b0}mc^2 and σγb0mc2\sigma_{\gamma_{b0}}mc^2 are the initial average energy and rms energy spread of the witness beam, ee is the electron charge, LL is the acceleration distance, and a flattop slice with a total thickness of lcl_c is assumed. Because the cooperation length depends on both the electron beam and undulator parameters, its precise value is not uniquely fixed. Guided by the discussion in Sec. IV, we adopt lc≈60l_c \approx60 nm and ρ≈1.6×10−4\rho\approx1.6 \times10^{-4} as a representative value for the following calculations. Without the witness beam, EzE_z falls approximately linearly with dEzdξ∼\frac{\mathrm{d}E_z}{\mathrm{d}\xi} \sim GeV/m/µm in the acceleration region [red line in Figure 1(a)]. This would significantly increase the slice energy spread after a meter-scale plasma. A witness beam with a properly chosen delay and peak current can absorb the wake energy and flatten the EzE_z profile to reduce ∣dEzdξ∣\left|\frac{\mathrm{d}E_z}{\mathrm{d}\xi}\right| and the corresponding δW\delta_W after acceleration.

We introduce η\eta to quantify the fraction of electrons that satisfy the energy spread requirement after acceleration and perform one-step QuickPIC simulations to study the TR and η\eta using Eq. (1). As shown in Figure 1(b)-(Figure 1e), the transformer ratio exhibits a clear dependence on the witness beam current and beam separation: lower current or larger separation results in a higher TR. The achieved TR ranges from ∼2\sim2 to ∼4\sim4 across the four cases, which is higher than that obtained with a driver of σz,d=16\sigma_{z,d}=16 µm, for which TR ≈1\approx1[32]. TR can be further improved by using a shorter witness beam and/or a longer driver beam. These results are consistent with the beam loading analysis in the nonlinear plasma wake[43]. A larger Δ\Delta positions the witness beam closer to the wake tail, where EzE_z is stronger; meanwhile, a smaller IwI_w or a shorter σz,w\sigma_{z,w} causes less modification of the unloaded high acceleration gradient.

We divide the beam loading into three regimes[32]: the unloaded regime where dEzdξ>0\frac{\mathrm{d}E_z}{\mathrm{d}\xi}>0 over the witness beam, the well-loaded regime where dEzdξ≈0\frac{\mathrm{d}E_z}{\mathrm{d}\xi}\approx0 and the overloaded regime with dEzdξ>0\frac{\mathrm{d}E_z}{\mathrm{d}\xi}>0 close to the middle of the beam and away from the beam center. When IwI_w is low (≲1.0\lesssim1.0 kA for Case I and II, and ≲1.7\lesssim1.7 kA for Case III and IV), the witness beam carries insufficient charge to significantly modify the unloaded EzE_z profile, leaving dEzdξ\frac{\mathrm{d}E_z}{\mathrm{d}\xi} large and resulting in a slice energy spread that exceeds the Pierce parameter ρ\rho for most electrons [the deep blue regions in Figure 1(b)-(Figure 1e)]. As IwI_w increases, dEzdξ\frac{\mathrm{d}E_z}{\mathrm{d}\xi} decreases and η\eta increases correspondingly. When approaching the well-loaded regime, η\eta reaches a relatively high level and remains largely insensitive to small variations in IwI_w and Δ\Delta. As can be seen from Figure 1(b)-(Figure 1e), the current achieving the maximum η\eta decreases as Δ\Delta increases[32, 43]. A further increase of IwI_w results in the overloaded regime [upper right corners in Figure 1(b)-(Figure 1e)] and η\eta decreases.

Compared with Case I and II, the high-η\eta region in Case III and Case IV corresponds to a substantially larger IwI_w, since the stronger driver excites a larger plasma bubble with a steeper field gradient at the tail, requiring a higher IwI_w to reach the well-loaded regime. Since the saturation power of FELs is Psat∝Iw4/3P_{\mathrm{sat}}\propto I_w^{4/3}, a large IwI_w is favorable for FEL performance[1, 26]. Point O in Case III achieves TR ≈3.0\approx3.0 and η≈64%\eta\approx64\% at a peak current of Iw=2.2I_w=2.2 kA, is well separated from the low-η\eta region, and is thus chosen for the subsequent full PBA simulation with QuickPIC. Throughout the PBA stage, the plasma wake structure remains nearly stationary, and the witness beam is continuously accelerated with an acceleration gradient of 15.9 GV/m until the driver beam is depleted at a plasma length of L=1.30L=1.30 m. The center energy of the witness beam WcW_c reaches 28.9 GeV at the exit of the PBA stage, which is 3.6 times the initial energy. The current profile and the normalized emittance of the witness beam are preserved during the acceleration. Its relative energy spread closely matches the prediction obtained from Eq. (1) using the one-step simulation results, and the usable electron fraction η≈68%\eta\approx68\% is also in good agreement with the prediction of the one-step simulation.

HI suppression using transversely asymmetric drivers

The results above demonstrate that, under ideal driver–witness alignment and with axisymmetric beam distributions, a round driver can provide the required energy gain while preserving the witness-beam quality needed for subsequent XFEL operation. In practice, however, neither perfect alignment nor exact beam axisymmetry can be guaranteed. In the high-TR configuration studied here, finite initial driver–witness misalignment or asymmetric beam distributions can seed HI and rapidly degrade the witness-beam quality[48, 49]. The HI of the driver can be mitigated by an initially large energy chirp or slice energy spread, as well as by effects that develop during PBA, including energy chirp, slice-energy-spread growth, and the nonlinear, axially varying focusing wakefield at the front of the driver[52, 53, 54, 55, 56]. These mechanisms, however, are unsuitable for the witness beam, whose projected energy spread, slice energy spread, and emittance must be preserved. Here, we utilize the axially varying focusing gradient of an elliptical nonlinear plasma wake driven by an elliptical driver[53, 68] to suppress the HI of the witness beam via Balakin–Novokhatsky–Smirnov (BNS) damping[69]. We focus on HI seeded by a misalignment between the driver and the witness, and the same approach can be directly extended to suppress HI seeded by witness beam asymmetry.

We introduce an initial transverse offset δx=σr,w\delta_x=\sigma_{r,w} between the witness and driver at point O discussed in Sec. II, and use the slice at ξ=ξw+σz,w\xi=\xi_w+\sigma_{z,w} to illustrate the development of HI. Because the witness beam is located near the rear of the nonlinear wake, where the ion channel radius is rb∼O(0.1)kp−1r_b\sim O(0.1)k_p^{-1}, it couples strongly to the wake boundary, leading to rapid growth of the HI[49]. As shown by the blue line in Figure 2a, the oscillation amplitude of slice centroid ∣xb∣|x_b| grows rapidly and reaches the boundary of the ion column at s≈24s\approx24 cm. Electrons then begin to be lost, causing the centroid oscillation amplitude to decrease. The HI also drives growth of the projected emittance in the oscillation direction: εn,x\varepsilon_{n,x} increases from its initial value of 0.4 μ\mum to 2.5 μ\mum at s≈24s\approx24 cm, whereas εn,y\varepsilon_{n,y} remains nearly unchanged. For comparison, the black solid line in Figure 2a shows the centroid evolution predicted by the Huang model[49], using the channel radius and boundary pseudo-potential extracted from simulations and accounting for witness acceleration. The witness is located near the wake tail, where the wake radius varies rapidly along the longitudinal direction, thereby violating the short-pulse approximation adopted in the model. Consequently, the model underestimates the growth of the centroid oscillation amplitude.

Figure 2. Witness-beam HI seeded by driver–witness misalignment. (a) Evolution of the slice centroid at ξ=ξw+σz,w\xi=\xi_w+\sigma_{z,w} (blue solid line) and projected emittances within ξw±σz,w\xi_w\pm\sigma_{z,w} (red dashed and green dash-dotted lines) for an initial offset δx=σr,w\delta_x=\sigma_{r,w}. The black solid line shows the centroid evolution calculated using the Huang model[49]. (b) Dependence of the propagation distance at which the ξw+σz,w\xi_w+\sigma_{z,w} slice reaches the ion column boundary on the initial offset (blue solid line), together with the projected emittances evaluated at that distance (red dashed and green dash-dotted lines). Immobile ions are used here.

We then scan the initial offset and record the propagation distance at which the selected slice reaches the ion column boundary, shown by the blue solid line in Figure 2b. Even for an offset of δx=0.1σr,w\delta_x=0.1\sigma_{r,w}, the selected witness slice reaches the ion column boundary near the end of the plasma. For all values of δx\delta_x considered, εn,x\varepsilon_{n,x} grows to approximately 2.5 μ\mum, while εn,y\varepsilon_{n,y} remains close to its initial value. These simulation results demonstrate that the HI can develop rapidly even from a small initial offset, causing electron loss and emittance growth that would preclude the use of the accelerated beam for XFEL applications.

The space charge field of an electron beam driver with an elliptical transverse profile is also elliptical.

Take a Gaussian distribution nd=nd0exp⁡(−x2/α+αy22σr,d2)n_d=n_{d0}\exp\left(-\frac{x^2/\alpha+\alpha y^2}{2\sigma_{r,d}^2}\right) as an example,

Ex≈end0σr,d2xϵ0(x2+y2)[1+σr,d2(α−1/α)(x2−3y2)(x2+y2)2],Ey≈end0σr,d2yϵ0(x2+y2)[1+σr,d2(α−1/α)(3x2−y2)(x2+y2)2]E_x\approx\frac{en_{d0}\sigma_{r,d}^2x}{\epsilon_0(x^2+y^2)}\left[1+\frac{\sigma_{r,d}^2(\alpha-1/\alpha)(x^2-3y^2)}{(x^2+y^2)^2}\right],\quad E_y\approx \frac{en_{d0}\sigma_{r,d}^2y}{\epsilon_0(x^2+y^2)}\left[1+\frac{\sigma_{r,d}^2(\alpha-1/\alpha)(3x^2-y^2)}{(x^2+y^2)^2}\right]

when x,y≫σr,dx,y\gg\sigma_{r,d}, where α\alpha is aspect ratio of the driver. The second terms in the square brackets represent the nonaxisymmetric field components induced by the elliptical profile, which scale linearly with σr,d2\sigma_{r,d}^2 and α=1/α\alpha=1/\alpha. Such a space charge field structures a blowout plasma wake with an elliptical boundary[53, 68]. Consequently, the gradient of the linear focusing wakefield varies along the axial direction, rather than remaining constant at dWxdx=dWydy=−12mωp2\frac{dW_x}{dx}=\frac{dW_y}{dy}=-\frac{1}{2}m\omega_p^2 as in a round wake. Figure 3a shows the profiles of Gx=−dWxdxG_x=-\frac{dW_x}{dx} and Gy=−dWydyG_y=-\frac{dW_y}{dy} in the witness beam region under different driver spot sizes and aspect ratios. In these scans, the driver emittances in both transverse directions are adjusted to maintain transverse beam matching, while the driver current is kept fixed. A larger σr,d\sigma_{r,d} or a higher α\alpha leads to a more significant variation of Gx,yG_{x,y} along ξ\xi. This axially varying Gx,yG_{x,y} changes the Betatron oscillation frequencies of different witness slices and drives them out of the resonance, thereby providing BNS damping of the HI[69].

Figure 3. HI suppression by using transversely asymmetric drivers. (a) Axial variation of the focusing gradients for different driver configurations. (b) Relative transverse nonuniformity of the accelerating field for different driver configurations. The inset shows the transverse distribution of EzE_z at ξ=ξw\xi=\xi_w for σr,d=1.41 μm\sigma_{r,d}=1.41\ \mu\mathrm{m} and α=2\alpha=2. (c) Slice centroid evolution at ξ=ξw+σz,w\xi=\xi_w+\sigma_{z,w} during the PBA stage for an initial offset δx=σr,w\delta_x=\sigma_{r,w}. The full centroid oscillation is shown for σr,d=1.70 μm\sigma_{r,d}=1.70\ \mu\mathrm{m} and α=2\alpha=2 (blue solid curve), whereas only the low-frequency envelopes are shown for the other driver configurations. The black solid curve shows the centroid envelope for an initially tilted witness when (σr,d,α)=(1.70 μm,2)(\sigma_{r,d},\alpha)=(1.70\ \mu\mathrm{m},2). The inset shows the maximum centroid amplification Amax⁡A_{\max} versus the relative rms focusing-gradient variation D(G)D(G) for different driver spot sizes, with α=2\alpha=2; the point labels indicate σr,d\sigma_{r,d}.

Meanwhile, the Panofsky–Wenzel theorem ensures that an axially varying focusing field must be accompanied by a transversely non-uniform accelerating gradient[70], i.e., ∂W∥∂ξ=∇⊥Ez\frac{\partial W_\parallel}{\partial\xi}=\nabla_\perp E_z. The inset in Figure 3b shows the EzE_z distribution in the x−yx-y plane, where a clear transverse nonuniformity is observed. To quantify the transverse nonuniformity of EzE_z, we define δEz(ξ)≡∫dxdynw(Ez−Eˉz)2∫dxdynw\delta E_z(\xi)\equiv\sqrt{\frac{\int dxdy n_w(E_z-\bar E_z)^2}{\int dxdy n_w}}, where Eˉz(ξ)≡∫dxdynwEz∫dxdynw\bar E_z(\xi)\equiv\frac{\int dxdy n_wE_z}{\int dxdy n_w} is the average acceleration gradient experienced by the witness beam. Figure 3b shows the relative nonuniformity, averaged over the witness beam current, for different drivers. The trend is consistent with Figure 3a, i.e., a larger σr,d\sigma_{r,d} or a higher α\alpha causes a larger ∣⟨δEzEˉz⟩ξ∣\left|\left\langle\frac{\delta E_z}{\bar E_z}\right\rangle_\xi\right|, where ⟨f⟩ξ≡∫dξIwf∫dξIw\langle f\rangle_\xi\equiv\frac{\int d\xi I_wf}{\int d\xi I_w}. Because the nonuniformity of EzE_z in an elliptical wake scales as x2−y2x^2-y^2, employing a tightly focused witness beam is beneficial for achieving a small ∣⟨δEzEz⟩ξ∣\left|\left\langle\frac{\delta E_z}{E_z}\right\rangle_\xi\right|.

With an elliptical driver, the centroid evolution at the witness beam slice ξ=ξw+σz,w\xi=\xi_w+\sigma_{z,w} for an initial offset of δx=σr,w\delta_x=\sigma_{r,w} is shown in Figure 3(c). When σr,d=1.70 μm\sigma_{r,d}=1.70\ \mu\mathrm{m} and α=2\alpha=2, the slice centroid oscillation (blue solid line) is suppressed effectively with a maximum xb≈1.5δxx_b\approx1.5\delta_x. In addition to the high-frequency betatron oscillation, the centroid exhibits a slow oscillation with a decaying amplitude[52, 56]. Numerical calculations of the coupled HI equations[49] reproduce this beat-wave behavior and show that it originates from resonance detuning induced by the axially varying focusing gradient. The beat-wave period can be estimated from the difference in betatron wavenumber between the beam head and a slice located at ξ=ξw+σz,w\xi=\xi_w+\sigma_{z,w}. This difference is Δkβ=G(ξ)γb(mωp2/e)kp−G0γb(mωp2/e)kp≈0.34kp2γb\Delta k_\beta=\sqrt{\frac{G(\xi)}{\gamma_b(m\omega_p^2/e)}}k_p-\sqrt{\frac{G_0}{\gamma_b(m\omega_p^2/e)}}k_p\approx0.34\frac{k_p}{\sqrt{2\gamma_b}}, where G0=0.5mωp2eG_0=0.5\frac{m\omega_p^2}{e} and G(ξ)=0.9mωp2eG(\xi)=0.9\frac{m\omega_p^2}{e} is used to reflect the increase of the focusing gradient during driver propagation (as discussed below). The betatron phase difference between these two slices therefore accumulates to 2π2\pi over a propagation distance 2πΔkβ≈182γbkp−1\frac{2\pi}{\Delta k_\beta}\approx18\sqrt{2\gamma_b}k_p^{-1}. This estimate is in reasonable agreement with the simulation result, shown by the blue line in Figure 3(c).

The envelope of the centroid oscillation for other drivers is also shown in Figure 3(c). A smaller σr,d\sigma_{r,d} results in less effective suppression of HI (red dashed line and green dash-dotted line), and vice versa (purple dotted line). This trend is consistent with the gradient profiles shown in Figure 3(a). To quantify the dependence of HI suppression on the longitudinal variation of the focusing gradient, we define D(G)=⟨[G−⟨G⟩ξ]2⟩ξ/⟨G⟩ξD(G)=\sqrt{\langle[G-\langle G\rangle_\xi]^2\rangle_\xi/\langle G\rangle_\xi}, where G(ξ)=Gx(ξ)G(\xi)=G_x(\xi) and the current-weighted average is taken over ξw−2σz,w≤ξ≤ξw+σz,w\xi_w-2\sigma_{z,w}\leq\xi\leq\xi_w+\sigma_{z,w}. The corresponding centroid amplification is characterized by Amax⁡=max⁡0≤s≤L∣xb(s)/δx∣A_{\max}=\max_{0\leq s\leq L}|x_b(s)/\delta_x| at ξ=ξw+σz,w\xi=\xi_w+\sigma_{z,w}. As shown in the inset of Figure 3(c), increasing D(G)D(G) from approximately 0.014 to 0.045 reduces Amax⁡A_{\max} from 4.94 to 1.06, consistent with stronger HI suppression through betatron-frequency detuning. As the head of the driver expands during propagation, it provides an increasingly nonaxisymmetric field component, and the axial variation of Gx,yG_{x,y} (as well as the transverse nonuniformity of EzE_z) increases correspondingly. Simulations show that for (σr,d,α)=(1.70 μm,4)(\sigma_{r,d},\alpha)=(1.70\ \mu\mathrm{m},4), GyG_y at ξw+σz,w\xi_w+\sigma_{z,w} becomes negative after a propagation distance of s≈1 cms\approx1\ \mathrm{cm}, causing substantial electron loss. This imposes a constraint on the transverse profile of the driver.

For subsequent XFEL operation, we need to examine the witness beam at the PBA exit in terms of its offset amplification, slice emittance and slice energy spread growth. The envelope of ∣xb/δx∣\left|x_b/\delta_x\right| at the plasma exit (s=1.3 ms=1.3\ \mathrm{m}) is approximately 3.3, 1.0, 0.3, and 0.2 for (σr,d,α)=(1.13 μm,2)(\sigma_{r,d},\alpha)=(1.13\ \mu\mathrm{m},2), (1.41 μm,2)(1.41\ \mu\mathrm{m},2), (1.70 μm,2)(1.70\ \mu\mathrm{m},2), and (1.98 μm,2)(1.98\ \mu\mathrm{m},2), respectively. Thus, the position offset of the witness beam is not amplified for σr,d≳1.41 μm\sigma_{r,d}\gtrsim1.41\ \mu\mathrm{m} and α=2\alpha=2. Since each slice of the witness beam is not transversely matched to the local focusing gradient, a finite slice energy spread can cause a growth of the slice emittance[63]. However, since the relative slice energy spread is very small, ∼10−4\sim10^{-4}, the resulting slice emittance growth is negligible over the considered 1.3-meter-long. plasma[63]. The slice energy spread grows due to the transverse non-uniformity of EzE_z in the elliptical wake. Accordingly, the usable fraction η\eta decreases to 62% when (σr,d,α)=(1.13 μm,2)(\sigma_{r,d},\alpha)=(1.13\ \mu\mathrm{m},2), and decreases further with increasing σr,d\sigma_{r,d}, i.e., η≈50%\eta\approx50\%, 41% and 30% for (1.41 μm,2)(1.41\ \mu\mathrm{m},2), (1.70 μm,2)(1.70\ \mu\mathrm{m},2), and (1.98 μm,2)(1.98\ \mu\mathrm{m},2), respectively. Considering the trade-off between the slice emittance and energy spread growth of the witness beam, and the HI suppression, we select a driver with σr,d=1.70 μm\sigma_{r,d}=1.70\ \mu\mathrm{m} and α=2\alpha=2 [blue solid line in Figure 3(c)]. Note that the peak current of the witness beam is slightly increased from 2.20 to 2.35 kA to compensate the change of on-axis EzE_z caused by the elliptical driver, increasing the usable fraction η\eta to approximately 50%.

To test whether HI suppression remains effective for a different type of initial seed, we also simulate an initially tilted witness when (σr,d,α)=(1.70 μm,2)(\sigma_{r,d},\alpha)=(1.70\ \mu\mathrm{m},2). Its initial centroid follows xb(0,ξ)=−σr,w[ξ−(ξw−3σz,w)]/(4σz,w)x_b(0,\xi)=-\sigma_{r,w}[\xi-(\xi_w-3\sigma_{z,w})]/(4\sigma_{z,w}), giving an offset magnitude of σr,w\sigma_{r,w} at ξ=ξw+σz,w\xi=\xi_w+\sigma_{z,w}. The corresponding centroid envelope, shown by the black solid curve in Figure 3(c), exhibits negligible amplification. Thus, HI suppression also persists for an initially tilted witness beam. Mobile lithium ions are included in the simulations, and the associated ion motion is found to be weak, with a negligible effect on the HI. Additional simulations further show that, although hydrogen ions can effectively suppress the HI, this suppression is accompanied by substantial emittance growth[42].

XFELs with 50 keV photon energies

At the PBA exit, the witness beam is strongly focused with βx,y∼7 mm\beta_{x,y}\sim7\ \mathrm{mm}, making it unsuitable for direct injection into the undulator. To address this mismatch, a magnetic transport line is designed to increase the transverse β\beta-functions of the witness beam, match them to the undulator focusing lattice, and remove the energy-depleted driver beam[32]. The detailed design of this transport line is described in Appendix B.

Figure 4 shows the slice properties at the end of the transport line for a witness beam accelerated in an elliptical PBA with an initial driver–witness misalignment of σr,w\sigma_{r,w}. The current profile (black solid line) and the slice emittances in the two transverse planes (red and green solid lines) are well preserved over most of the witness bunch. The relative slice energy spread remains at the ∼10−4\sim10^{-4} level near the bunch center, as indicated by the blue dash-dotted line. Owing to the axially varying focusing gradient in the PBA, the witness beam is not transversely matched to the plasma wake. As a result, the Courant–Snyder parameters of the witness beam vary from slice to slice at the PBA exit. Consequently, the slice β\beta-functions exhibit axial oscillations after transport, as shown by the red and green dashed curves. Near the bunch center, where the slice energy spread is close to its minimum, both βx\beta_x and βy\beta_y are approximately 45 m.

According to the undulator resonance condition, λr=λu2γb2(1+K22)\lambda_r=\frac{\lambda_u}{2\gamma_b^2}\left(1+\frac{K^2}{2}\right), this ∼30 GeV\sim30\ \mathrm{GeV} and high-quality electron beam can drive FEL lasing at high photon energies in an undulator. However, quantum diffusion induced by continuous spontaneous-emission undulator radiation (SUR) reduces the electron average beam energy and increases its energy spread[1, 26]. While the energy loss can be compensated for by employing a tapered undulator[71], the growth of the energy spread ultimately limits the achievable photon energy.

The growth of the energy spread induced by quantum diffusion is described as[72] dσγb2dzu=715λcreγb4K2F(K2)\frac{\mathrm{d}\sigma_{\gamma_b}^2}{\mathrm{d}z_u}=\frac{7}{15}\lambda_c r_e\gamma_b^4 K^2 F\left(\frac{K}{\sqrt{2}}\right), where rer_e is the classical electron radius, λc=ℏmec\lambda_c=\frac{\hbar}{m_ec} is the reduced Compton wavelength, and F(x)=1.70x+11+1.88x+0.8x2F(x)=1.70x+\frac{1}{1+1.88x+0.8x^2} (for planar undulator) is a fit to simplify the dependence on various Bessel functions. As the FEL instability develops, quantum diffusion increases the beam energy spread; once the induced relative energy spread approaches the FEL Pierce parameter ρ\rho before saturation, the FEL performance is strongly degraded. In 1D FEL theory, ρ=(e2K2[JJ]2nb32ϵ0γb3mec2ku2)1/3\rho=\left(\frac{e^2K^2[J J]^2n_b}{32\epsilon_0\gamma_b^3m_ec^2k_u^2}\right)^{1/3}, where nbn_b is the peak beam density, and [JJ]∼1[J J]\sim1 is the coupling factor between the electrons and the radiation field in a planar undulator whose value depends on KK[1]. For a fixed target radiation wavelength, for which λu∝(1+K2/2)−1\lambda_u\propto(1+K^2/2)^{-1}, numerical calculations indicate that a larger λu\lambda_u and a smaller KK are usually favorable for FEL gain.

To assess the XFEL performance driven by the accelerated witness beam, we perform 3D FEL simulations with GENESIS for high-photon-energy XFEL operation, including quantum diffusion. For simplicity, we choose an undulator period of λu=3 cm\lambda_u=3\ \mathrm{cm} and vary the undulator deflection parameter KK to vary the radiation wavelength. A stepwise linear taper is applied from the undulator entrance, with KK decreasing by 1.32×10−41.32\times10^{-4}, 8.75×10−58.75\times10^{-5}, and 6.45×10−56.45\times10^{-5} per 1.98 m module for the initial K=2.83K=2.83, 2.40, and 2.12 cases, respectively, thereby keeping (1+K2/2)/γb2(1+K^2/2)/\gamma_b^2 approximately constant and maintaining the FEL resonance condition. The solid lines in Fig. 5(a) show the growth of the peak radiation power for three representative cases whereas the dashed lines indicate the quantum-diffusion-induced energy spread growth.

For λu=3 cm\lambda_u=3\ \mathrm{cm} and K=2.83K=2.83, the resonant photon energy is Er=54.3 keVE_r=54.3\ \mathrm{keV}. The 3D Pierce parameter estimated us-

Figure 4. The slice current II, relative energy spread δW\delta_W, normalized emittance ϵn\epsilon_n and β\beta-function of the witness bunch at the end of the beam transport line.

ing Ming Xie’s fitting formula[73] is ρ≃1.6×10−4\rho\simeq1.6 \times10^{-4}. Fitting the simulated power growth with P(zu)∝exp⁡(zu/Lg)P(z_{\mathrm{u}}) \propto\exp(z_{\mathrm{u}}/L_{\mathrm{g}}) over the interval 100 m≤zu≤140 m100\ \mathrm{m} \leq z_{\mathrm{u}} \leq140\ \mathrm{m} gives a power gain length LgL_{\mathrm{g}} corresponding to an effective Pierce parameter ρeff≡λu/(4π3Lg)≈0.54×10−4\rho_{\mathrm{eff}} \equiv\lambda_{\mathrm{u}}/(4\pi\sqrt{3}L_{\mathrm{g}}) \approx0.54 \times10^{-4}, approximately a factor of three smaller than the Ming Xie estimate. This reduction is mainly caused by the SUR-induced beam energy spread growth, as shown by the blue dashed line in Fig. 5(a). The radiation power saturates at approximately zu=300 mz_{\mathrm{u}} = 300\ \mathrm{m}, reaching a peak value of P≈19.6 GWP \approx19.6\ \mathrm{GW}.

As KK is decreased to 2.40 and 2.12, the corresponding photon energy increases to 69.7 and 83.5 keV and the Pierce parameter given by the Ming Xie’s formula decreases to 1.3×10−41.3 \times10^{-4} and 1.1×10−41.1 \times10^{-4}. As indicated by the dashed lines in Fig. 5(a), the energy spread growth is also slowed compared with the K=2.83K = 2.83 case. In both cases, the peak power saturates at around zu=350 mz_{\mathrm{u}} = 350\ \mathrm{m}, reaching P≈13.1P \approx13.1 and 7.4 GW, respectively.

The radiation spectra at saturation are shown in Fig. 5(b). They exhibit pronounced multi-spike structures, with relative rms bandwidths Δω/⟨ω⟩\Delta\omega/\langle\omega\rangle of 7.3×10−47.3 \times10^{-4}, 7.2×10−47.2 \times10^{-4}, and 7.5×10−47.5 \times10^{-4} for K=2.83K = 2.83, 2.40, and 2.12, respectively. These bandwidths are approximately 13.6, 13.9, and 20.3 times the corresponding effective Pierce parameters inferred from fits to the simulated exponential power growth. The larger full-pulse bandwidths arise from the residual longitudinal energy chirp acquired during the PBA stage: different slices have different resonant frequencies, introducing additional spectral broadening beyond the intrinsic SASE bandwidth[26, 32]. The temporal pulse profiles for these three cases are shown in Fig. 5(c). The radiation pulse develops around the beam center where the current is high and the energy spread is small. In all three cases, the pulse duration is approximately 10 fs. At the corresponding saturation distances, the pulse energies are estimated to be 33.0, 32.2, and 26.2 μ\muJ for photon energies of 54.3, 69.7, and 83.5 keV, respectively. These pulse energies correspond to approximately 3.8×1093.8 \times10^{9}, 2.9×1092.9 \times10^{9}, and 2.0×1092.0 \times10^{9} photons per pulse, respectively.

Discussion

A high transformer ratio requires a long driver beam and a short witness beam, with the driver carrying substantially more charge than the witness. Overall, only approximately 5% of the charge delivered by the RF accelerator contributes to FEL lasing in the undulator and generates x-rays with high photon energies. In this sense, the PBA acts as an effective photon energy up-conversion stage: it sacrifices the number of radiating electrons in exchange for a substantial increase in photon energy. Without the afterburner, all electrons from the RF accelerator radiate in the undulator, but at relatively low photon energies. With the afterburner, only ∼5%\sim5\% of the electrons radiate, yet the emitted photons reach much higher energies. A crude estimate suggests that this trade-off is equivalent to converting approximately 10310^{3} photons at several keV into one photon above 50 keV.

In conclusion, we use 3D PIC simulations to demonstrate that 8 GeV electron beams from RF accelerators can be boosted to ∼30\sim30 GeV in a meter-long PBA with a transformer ratio of ≈3\approx3. The beam quality can be preserved well even in the presence of alignment errors, with nearly half of the electrons satisfying the requirements of subsequent high-photon-energy XFELs. The misalignment between the driver and witness beams can trigger the HI and we employ an elliptical beam driver to drive an elliptical plasma wake which can suppress the HI of the witness beam effectively and maintain the beam quality. The parameter study presented in Sec. III shows that the driver spot size needs to be chosen properly to achieve both effective HI suppression and minimal growth of slice energy spread. This poses challenges for the electron beams delivered by the RF accelerators, i.e., the spot sizes of the driver are different along the two transverse directions, and the average spot size is approximately 6 times that of the witness beam. Our simulations show that 10 GW-class x-rays with ≥50\geq50 keV photon energy can be produced in undulators hundreds of meters long after accounting for quantum diffusion.

The TR of approximately 3 considered here represents an operating point compatible with the beam parameters and XFEL application targeted in this study. The maximum TR at which HI can still be effectively suppressed depends on the competition between HI growth and longitudinal betatron-frequency detuning, subject to the beam-quality constraints associated with the specific driver–witness configuration. Although longitudinally shaped beams can provide higher TRs[61], the effectiveness of the proposed suppression mechanism in such configurations requires separate assessment. For the high-photon-energy XFEL application considered here, maximizing the TR is not the primary design objective. A TR of approximately 3 already provides sufficient witness energy for the demonstrated XFEL operation above 50 keV, provided that the required beam quality is preserved. Increasing the TR and witness energy does not necessarily improve the usable XFEL output: for fixed undulator parameters and propagation length, quantum-diffusion-induced relative energy spread increases with electron energy, while the Pierce parameter decreases when the beam current and transverse beam size are held fixed. Quantum diffusion already noticeably affects FEL performance in the cases studied here. Improving beam brightness offers a promising route by increasing the Pierce parameter and shortening the saturation length, thereby reducing the accumulated quantum-diffusion-induced energy spread[20].

Acknowledgments

This work was supported by the Fundamental and Interdisciplinary Disciplines Breakthrough Plan of the Ministry of Education of China-JYB2025XDXM204, the National Grand Instrument Project No. SQ2019YFF01014400, National Natural Science Foundation of China No. 12375147, Beijing outstanding young scientist project, and the Fundamental Research Funds for the Central Universities, Peking University. The simulations were supported by the High-performance Computing Platform of Peking University.

Author contributions

Conflict of Interest

The authors have no conflicts to disclose.

Author Contributions

Songtang Han: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (equal); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). Letian Liu: Formal analysis (equal); Investigation (equal); Methodology (equal); Software (supporting); Writing – review & editing (supporting). Zhiyan Yang: Investigation (equal); Software (supporting); Writing – review & editing (supporting). Zhuo Pan: Investigation (equal); Methodology (equal); Writing – review & editing (supporting). Xueqing Yan: Funding acquisition (equal); Supervision (equal); Writing – review & editing (equal). Xinlu Xu: Conceptualization (lead); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Resources (lead); Supervision (lead); Writing – original draft (equal); Writing – review & editing (equal).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Setup of the simulations

All simulations of the PBA stage are performed with the quasi-static 3D PIC code QuickPIC[65, 66]. A moving window propagating at the speed of light in vacuum cc is adopted, with a box size of 160 μm × 160 μm × 260 μm, and is discretized into 1024 × 1024 × 1024 cells in the xx, yy, and ξ\xi directions, respectively. A convergence test on the grid resolution is performed to verify that the chosen grid size is appropriate. Both the driver and witness beams are represented by approximately 4.3×1064.3 \times10^{6} macro particles. The plasma electrons and ions are initialized with 4 macro particles per cell. In the quasi-static algorithm, as the driver-witness bunch pair passes through each plasma slice, the plasma response is solved and the resulting electromagnetic fields are then used to advance the positions and momenta of the beam particles. In the simulations presented here, the bunch pair is advanced with a longitudinal step of 2 mm.

The beam transport is simulated using the particle tracking code ELEGANT[74] as detailed in Appendix B. The FEL process is simulated with the 3D time-dependent FEL code GENESIS[75]. The six-dimensional phase-space distribution of the transported witness beam is imported into GENESIS at the undulator entrance. To preserve the transverse beam size along the undulator, a FODO focusing lattice is included in the GENESIS simulations. Guided by the optimum β\beta-functions obtained from the Ming Xie fitting formula, the focusing and defocusing quadrupoles in the FODO lattice are chosen to have a length of 4 cm and a gradient of 106 T/m, with each quadrupole pair separated by a 1.98 m-long undulator module. The GENESIS simulations include both the average energy loss and the energy spread growth induced by SUR, with a stepwise linear undulator taper applied to compensate for the average energy loss, as described in Section IV.

Beam transport from the PBA to the undulator

A magnetic transport line is designed using the particle tracking code ELEGANT[74]. As shown in Figure 6, it consists of two quadrupole triplets (QT1 and QT2) and a symmetric C-chicane[32]. QT1, in a D-F-D configuration, is placed 18 cm downstream of the plasma exit, where ‘D/F’ indicates defocusing/focusing in the xx-direction. It consists of two defocusing quadrupoles (length 17.1 cm, gradient −748.6-748.6 T/m) and one focusing quadrupole (length 34.2 cm, gradient 606.9 T/m), separated by 41 cm drifts. The energy-depleted driver diverges more rapidly than the witness beam in QT1. A rectangular collimator of length 50 cm with an aperture of 0.15 mm ×\times 0.6 mm is placed 60 cm downstream of QT1 to remove the majority of the driver electrons. The subsequent symmetric C-chicane consists of four dipoles, each with a length of 10 cm and a magnetic field of 2.9 T, corresponding to a bending angle of 3 mrad for 29.3 GeV electrons. The drift lengths between consecutive dipoles are 1.00 m, 0.50 m, and 1.00 m. A beam dump is placed at the center of the chicane to remove the residual driver, which is horizontally separated from the witness beam by the dispersion introduced in the first half of the chicane. QT2, in an F-D-F configuration, is placed 89 cm downstream of the chicane exit. It consists of two focusing quadrupoles (length 13.8 cm, gradient 440.4 T/m) and one defocusing quadrupole (length 23.2 cm, gradient −373.4-373.4 T/m), separated by 1.34 m drifts. QT2 matches the witness beam to the target Twiss parameters βx=48\beta_x = 48 m, βy=44\beta_y = 44 m, αx=1.04\alpha_x = 1.04, and αy=−0.96\alpha_y = -0.96 at the undulator entrance. These parameters are optimized using the Ming Xie’s fitting formula[73] to maximize the 3D Pierce parameter ρ\rho and are approximately matched with the FODO along the undulator. Additional tracking simulations show that the transport largely preserves the entrance slice emittances under energy and emittance jitters, with moderate changes in the β\beta-functions.

Schematic of the beam transport line from the plasma exit to the undulator entrance

Figure 6. Schematic of the beam transport line from the plasma exit to the undulator entrance (not to scale). After the PBA stage, the accelerated witness beam and the energy-depleted driver beam are captured by the first quadrupole triplet, QT1. A rectangular collimator removes most of the driver electrons, while the residual driver is further separated from the witness beam in a symmetric C-chicane and dumped at the chicane center. The second quadrupole triplet, QT2, matches the witness beam to the target Twiss parameters at the undulator entrance for subsequent high-photon-energy FEL.

References

  1. [1]C. Pellegrini, A. Marinelli, and S. Reiche, “The physics of X-ray free-electron lasers,” Rev. Mod. Phys. 88, 015006 (2016).DOI
  2. [2]N. Huang, H. Deng, B. Liu, D. Wang, and Z. Zhao, “Features and futures of X-ray free-electron lasers,” The Innovation 2, 100097 (2021).DOI
  3. [3]L. Young, E. P. Kanter, B. Kraessig, Y. Li, A. M. March, et al., “Femtosecond electronic response of atoms to ultra-intense X-rays,” Nature 466, 56–61 (2010).DOI
  4. [4]M. Fuchs, M. Trigo, J. Chen, S. Ghimire, S. Schwartz, et al., “Anomalous nonlinear X-ray Compton scattering,” Nat. Phys. 11, 964–970 (2015).
  5. [5]C. Weininger, M. Purvis, D. Ryan, R. A. London, J. D. Bozek, et al., “Stimulated electronic X-Ray Raman scattering,” Phys. Rev. Lett. 111, 233902 (2013).
  6. [6]H. Yong, N. Zotev, J. M. Ruddock, B. Stankus, M. Simmermacher, et al., “Observation of the molecular response to light upon photoexcitation,” Nat. Commun. 11, 2157 (2020).DOI
  7. [7]Z.-H. Loh, G. Doumy, C. Arnold, L. Kjellsson, S. Southworth, et al., “Observation of the fastest chemical processes in the radiolysis of water,” Science 367, 179–182 (2020).
  8. [8]K. Pande, D. H. Hutchison, G. Groenhof, A. Aquila, J. S. Robinson, et al., “Femtosecond structural dynamics drives the trans/cis isomerization in photoactive yellow protein,” Science 352, 725–729 (2016).
  9. [9]H. N. Chapman, A. Barty, M. J. Bogan, S. Boutet, M. Frank, et al., “Femtosecond diffractive imaging with a soft-X-ray free-electron laser,” Nat. Phys. 2, 839–843 (2006).DOI
  10. [10]H. N. Chapman, P. Fromme, A. Barty, T. A. White, R. A. Kirian, et al., “Femtosecond X-ray protein nanocrystallography,” Nature 470, 73–77 (2011).
  11. [11]T. Ekeberg, M. Svenda, C. Abergel, F. R. N. C. Maia, V. Seltzer, et al., “Three-dimensional reconstruction of the giant mimivirus particle with an X-ray free-electron laser,” Phys. Rev. Lett. 114, 098102 (2015).pubmed.ncbi.nlm.nih.gov/25793853
  12. [12]S. Boutet, L. Lomb, G. J. Williams, T. R. Barends, A. Aquila, et al., “High-resolution protein structure determination by serial femtosecond crystallography,” Science 337, 362–364 (2012).DOI
  13. [13]P. Emma, R. Akre, J. Arthur, R. Bionta, C. Bostedt, et al., “First lasing operation of an ångstrom-wavelength free-electron laser,” Nat. Photonics 4, 641–647 (2010).DOI
  14. [14]E. Schneidmiller and M. Yurkov, “Baseline parameters of the European XFEL,” in Proceedings of the 38th International Free Electron Laser Conference (FEL2017) (2017) pp. 109–112.
  15. [15]M. Yabashi, H. Tanaka, and T. Ishikawa, “Overview of the SACLA facility,” J. Synchrotron Rad. 22, 477–484 (2015).DOI
  16. [16]C. J. Milne, T. Schietinger, M. Aiba, A. Alarcon, J. Alex, et al., “SwissFEL: the Swiss X-ray free electron laser,” Appl. Sci. 7, 720 (2017).DOI
  17. [17]J. Stöhr, “Linac coherent light source II (LCLS-II) conceptual design report,” Tech. Rep. (SLAC National Accelerator Laboratory, 2011).DOI
  18. [18]I. S. Ko, H.-S. Kang, H. Heo, C. Kim, G. Kim, et al., “Construction and commissioning of PAL-XFEL facility,” Appl. Sci. 7, 479 (2017).DOI
  19. [19]Z. Y. Zhu, Z. T. Zhao, D. Wang, Z. Liu, R. X. Li, et al., “SCLF: An 8-GeV CW SCRF linac-based X-ray FEL facility in Shanghai,” in Proceedings of the 38th International Free Electron Laser Conference (FEL2017) (2017) pp. 182–184.
  20. [20]W. H. Tan, R. Robles, J. Hernandez, E. A. Nanni, and A. Dhar, “Nanosecond radio-frequency pulse driven photon for very hard X-ray free-electron laser,” (2025), arXiv:2511.07592 [physics.acc-ph].arxiv.org/abs/2511.07592
  21. [21]U. Lienert, S. F. Li, C. M. Hefferan, J. Lind, R. M. Suter, et al., “High-energy diffraction microscopy at the advanced photon source,” JOM 63, 70–77 (2011).DOI
  22. [22]J. V. Bernier, R. M. Suter, A. D. Rollett, and J. D. Almer, “High-energy X-ray diffraction microscopy in materials science,” Annu. Rev. Mater. Res. 50, 395–436 (2020).DOI
  23. [23]H. Mimura, G. Yamaguchi, K. Tamura, H. Yumoto, T. Koyam, et al., “High-speed imaging of metal drilling using 100-keV X-rays at SPring-8,” in Advances in X-Ray/EUV Optics and Components XVII, Proceedings of SPIE, Vol. 12240 (SPIE, 2022) p. 1224009.DOI
  24. [24]U. Zastrau, K. Appel, C. Baehtz, O. Baehр, L. Batchelor, et al., “The high energy density scientific instrument at the European XFEL,” J. Synchrotron Rad. 28, 1393–1416 (2021).
  25. [25]B. W. Adams, C. Buth, S. M. Cavaletto, J. Evers, Z. Harman, et al., “X-ray quantum optics,” J. Mod. Opt. 60, 2–21 (2013).DOI
  26. [26]Z. Huang and K.-J. Kim, “Review of X-ray free-electron laser theory,” Phys. Rev. ST Accel. Beams 10, 034801 (2007).DOI
  27. [27]S. Lee, T. Katsouleas, P. Muggli, W. B. Mori, C. Joshi, et al., “Energy doubler for a linear collider,” Phys. Rev. ST Accel. Beams 5, 011001 (2002).DOI
  28. [28]I. Blumenfeld, C. E. Clayton, F.-J. Decker, M. J. Hogan, C. Huang, et al., “Energy doubling of 42 GeV electrons in a metre-scale plasma wakefield accelerator,” Nature 445, 741–744 (2007).pubmed.ncbi.nlm.nih.gov/17301787
  29. [29]M. Litos, E. Adli, W. An, C. Clarke, C. Clayton, et al., “High-efficiency acceleration of an electron beam in a plasma wakefield accelerator,” Nature 515, 92–95 (2014).DOI
  30. [30]R. D’Arcy, J. Chappell, J. Beinortaite, S. Diederichs, G. Boyle, et al., “Recovery time of a plasma-wakefield accelerator,” Nature 603, 58–62 (2022).
  31. [31]S. Schroder, J. Osterhoff, S. Wesch, and C. B. Schroeder, “Application of a plasma-based energy booster in operational free-electron lasers,” J. Phys.: Conf. Ser. 3124, 012018 (2025).arxiv.org/abs/2407.15583
  32. [32]L. Liu, Q. Ma, Y. Xia, Z. Wang, Y. Chen, et al., “Improvement of photon energy at X-ray free-electron lasers using plasma-based afterburner,” Matter Radiat. Extremes 10, 047202 (2025).DOI
  33. [33]T. Tajima and J. M. Dawson, “Laser electron accelerator,” Phys. Rev. Lett. 43, 267–270 (1979).DOI
  34. [34]P. Chen, J. Dawson, R. W. Huff, and T. Katsouleas, “Acceleration of electrons by the interaction of a bunched electron beam with a plasma,” Phys. Rev. Lett. 54, 693 (1985).
  35. [35]E. Esarey, C. B. Schroeder, and W. P. Leemans, “Physics of laser-driven plasma-based electron accelerators,” Rev. Mod. Phys. 81, 1229–1285 (2009).DOI
  36. [36]M. Litos, E. Adli, J. M. Allen, W. An, C. I. Clarke, et al., “9 GeV energy gain in a beam-driven plasma wakefield accelerator,” Plasma Phys. Control. Fusion 58, 034017 (2016).DOI
  37. [37]J. B. Rosenzweig, B. Breizman, T. Katsouleas, and J. J. Su, “Acceleration and focusing of electrons in two-dimensional nonlinear plasma wake fields,” Phys. Rev. A 44, R6186–R6192 (1991).
  38. [38]I. Kostyukov, A. Pukhov, and S. Kiselev, “Phenomenological theory of laser-plasma interaction in “bubble” regime,” Phys. Plasmas 11, 5256–5264 (2004).DOI
  39. [39]W. Lu, C. Huang, M. Zhou, W. B. Mori, and T. Katsouleas, “Nonlinear theory for relativistic plasma wakefields in the blowout regime,” Phys. Rev. Lett. 96, 165002 (2006).DOI
  40. [40]J. B. Rosenzweig, A. M. Cook, A. Scott, M. C. Thompson, and R. B. Yoder, “Effects of ion motion in intense beam-driven plasma wakefield accelerators,” Phys. Rev. Lett. 95, 195002 (2005).
  41. [41]R. Golziahdez, T. Katsouleas, P. Muggli, C. Huang, and W. Mori, “Preservation of beam emittance in the presence of ion motion in future high-energy plasma-wakefield-based colliders,” Phys. Rev. Lett. 104, 155001 (2010).
  42. [42]W. An, W. Lu, C. Huang, X. Xu, M. J. Hogan, et al., “Ion motion induced emittance growth of matched electron beams in plasma wakefields,” Phys. Rev. Lett. 118, 244801 (2017).
  43. [43]M. Tzoufras, W. Lu, F. S. Tsung, C. Huang, W. B. Mori, et al., “Beam loading in the nonlinear regime of plasma-based acceleration,” Phys. Rev. Lett. 101, 145002 (2008).DOI
  44. [44]R. D. Ruth, A. Chao, P. B. Wilson, and P. Morton, “A plasma wake field accelerator,” Part. Accel. 17, 171 (1984).DOI
  45. [45]K. L. F. Bane, P. Chen, and P. B. Wilson, “On collinear wake field acceleration,” IEEE Trans. Nucl. Sci. 32, 3524–3526 (1985).DOI
  46. [46]P. Chen, J. J. Su, M. Dawson, K. L. F. Bane, and P. B. Wilson, “Energy transfer in the plasma wake-field accelerator,” Phys. Rev. Lett. 56, 1252–1255 (1986).pubmed.ncbi.nlm.nih.gov/10032613
  47. [47]J. B. Rosenzweig, “Nonlinear plasma dynamics in the plasma wake-field accelerator,” Phys. Rev. Lett. 58, 555–558 (1987).pubmed.ncbi.nlm.nih.gov/10034971
  48. [48]D. H. Whittum, W. M. Sharp, S. Xu, M. Lampe, and G. Joyce, “Electron-hose instability in the ion-focused regime,” Phys. Rev. Lett. 67, 991–994 (1991).pubmed.ncbi.nlm.nih.gov/10045042
  49. [49]C. Huang, W. Lu, M. Zhou, C. E. Clayton, C. Joshi, et al., “Hosing instability in the blow-out regime for plasma-wakefield acceleration,” Phys. Rev. Lett. 99, 255001 (2007).
  50. [50]G. Loisch, G. Asova, P. Boonpornprasert, R. Brinkmann, Y. Chen, et al., “Observation of high transformer ratio plasma wakefield acceleration,” Phys. Rev. Lett. 121, 064801 (2018).DOI
  51. [51]R. Roussel, G. Andonian, W. Lynn, K. Sanwalka, R. Robles, et al., “Single shot characterization of high transformer ratio wakefields in nonlinear plasma acceleration,” Phys. Rev. Lett. 124, 048802 (2020).
  52. [52]T. J. Mehrling, R. A. Fonseca, A. Martinez de la Ossa, and J. Vieira, “Mitigation of the hose instability in plasma-wakefield accelerators,” Phys. Rev. Lett. 118, 174801 (2017).
  53. [53]L. Hildebrand, W. An, X. L. Xu, F. Li, Y. Zhao, et al., “Mitigation techniques for witness beam hosing in plasma-based acceleration,” in Proceedings of the 2018 IEEE Advanced Accelerator Concepts Workshop (AAC 2018) (2018) pp. 1–5.
  54. [54]T. J. Mehrling, R. A. Fonseca, A. Martinez de la Ossa, and J. Vieira, “Mechanisms for the mitigation of the hose instability in plasma-wakefield accelerators,” Phys. Rev. Accel. Beams 22, 031302 (2019).
  55. [55]O. G. Finnerud, C. A. Lindstrøm, and E. Adli, “Parametric mapping of the efficiency–instability relation in plasma-wakefield accelerators,” Phys. Rev. Accel. Beams 28, 071301 (2025).arxiv.org/abs/2501.10602
  56. [56]R. Lehe, C. B. Schroeder, J.-L. Vay, E. Esarey, and W. P. Leemans, “Saturation of the hosing instability in quasilinear plasma accelerators,” Phys. Rev. Lett. 119, 244801 (2017).
  57. [57]T. J. Mehrling, C. Benedetti, C. B. Schroeder, E. Esarey, and W. P. Leemans, “Suppression of beam hosing in plasma accelerators with ion motion,” Phys. Rev. Lett. 121, 264802 (2018).
  58. [58]L. Hildebrand, Y. Zhao, W. An, F. Li, Q. Su, et al., “Beam realignment with emittance preservation in a plasma wakefield-accelerator stage,” Phys. Rev. Lett. 135, 195002 (2025).
  59. [61]Q. Su, J. Larson, T. N. Dalichaouch, F. Li, W. An, et al., “Optimization of transformer ratio and beam loading in a plasma wakefield accelerator with a structure-exploiting algorithm,” Phys. Plasmas 30, 053108 (2023).
  60. [62]X. L. Xu, J. F. Hua, F. Li, C. J. Zhang, L. X. Yan, et al., “Phase-space dynamics of ionization injection in plasma-based accelerators,” Phys. Rev. Lett. 112, 035003 (2014).
  61. [63]X. L. Xu, J. F. Hua, Y. P. Wu, C. J. Zhang, F. Li, et al., “Physics of phase space matching for staging plasma and traditional accelerator components using longitudinally tailored plasma profiles,” Phys. Rev. Lett. 116, 124801 (2016).
  62. [65]C. Huang, V. K. Decyk, C. Ren, M. Zhou, W. Lu, et al., “QUICKPIC: A highly efficient particle-in-cell code for modeling wakefield acceleration in plasmas,” J. Comput. Phys. 217, 658–679 (2006).
  63. [66]W. An, V. K. Decyk, W. B. Mori, and T. M. Antonsen, “An improved iteration loop for the three dimensional quasi-static particle-in-cell algorithm: QuickPIC,” J. Comput. Phys. 250, 165–177 (2013).
  64. [67]C. A. Lindstrøm, J. Beinortas, J. Björklund Svensson, L. Boulton, J. Chappell, et al., “Emittance preservation in a plasma-wakefield accelerator,” Nat. Commun. 15, 6097 (2024).DOI
  65. [68]P. Manwani, Y. Kang, J. Mann, B. Naranjo, G. Andonian, et al., “Analysis of the blowout plasma wakefields produced by drive beams with elliptical symmetry,” Phys. Rev. Lett. 135, 095001 (2025).
  66. [69]V. E. Balakin, A. V. Novokhatsky, and V. P. Smirnov, “VLEPP: transverse beam dynamics,” in Proceedings of the 12th International Conference on High Energy Accelerators (HEACC 1983) (1984) pp. 119–120.
  67. [70]W. K. H. Panofsky and W. A. Wenzel, “Some considerations concerning the transverse deflection of charged particles in radio-frequency fields,” Rev. Sci. Instrum. 27, 967–967 (1956).
  68. [71]N. Kroll, P. Morton, and M. Rosenbluth, “Free-electron lasers with variable parameter wigglers,” IEEE J. Quantum Electron. 17, 1436–1468 (1981).
  69. [72]E. Saldin, E. Schneidmiller, and M. Yurkov, “Calculation of energy diffusion in an electron beam due to quantum fluctuations of undulator radiation,” Nucl. Instrum. Methods Phys. Res., Sect. A 381, 545–547 (1996).
  70. [73]M. Xie, “Design optimization for an X-ray free electron laser driven by SLAC linac,” in Proceedings of the 1995 Particle Accelerator Conference (PAC 1995) (1995) pp. 183–185.
  71. [74]M. Borland, “Elegant: A flexible SDDS-compliant code for accelerator simulation,” Tech. Rep. (Argonne National Laboratory, 2000).DOI
  72. [75]S. Reiche, “GENESIS 1.3: a fully 3D time-dependent FEL simulation code,” Nucl. Instrum. Methods Phys. Res., Sect. A 429, 243–248 (1999).DOI

Paper details

vlnc:W_htbm7exr
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents