Experimental demonstration of fresh bunch self-seeding in an X-ray free electron laser

X-ray free electron lasers (XFELs) are the brightest sources of coherent X-rays, generating transversely coherent femtosecond long pulses from nanometer to sub-angstrom wavelengths. Their peak brightness is presently nine orders of magnitude larger than that of storage ring based synchrotron radiation sources.¹ Existing XFELs have opened new frontiers in scientific research, fueling progress in biology, Atomic Molecular and Optical (AMO) science, materials science, condensed matter physics, and chemistry.² The standard mode of operation for current XFELs is Self Amplified Spontaneous Emission (SASE),^3,4 generating X-rays with a power transfer efficiency $P X - R a y / P Beam$ and a radiation bandwidth on the order of the FEL parameter ρ, typically around $10 − 3$⁠. Increasing the efficiency and improving the temporal coherence of XFELs have been achieved via self-seeding,^5–10 external seeding,^11,12 and tapering of the undulator magnetic field.¹³ 

A recent paper shows that the efficiency of XFELs can be increased to about 10%, generating multi-TW X-ray pulses few to 10 fs long.¹⁴ This can be achieved by optimized undulator tapering in combination with two features: a Fresh Bunch Self-Seeding (FBSS) method and an advanced superconducting helical undulator with strong transverse focusing. The result from Ref. 14 is based on theoretical analysis and simulations using typical XFEL electron beam parameters. For these parameters, the saturation power for SASE, self-seeding, and FBSS is 50 GW, 3.6 TW, and 6.3 TW, respectively.¹⁴ The great advantage of FBSS with respect to SASE and the doubling of peak power with respect to self-seeding is the motivation for our experiment to demonstrate the feasibility of the FBSS scheme in an XFEL. We note that the demonstration of the FBSS method reported here is limited in peak power by the use of existing hardware not optimized for this application. We remind the reader that alternative methods to reach TW-level pulses based on advanced compression schemes and superradiance have also recently been proposed.^15–18 

Pushing the state of the art of XFEL efficiency and peak power is important for many applications. In AMO science, for example, increasing the power enhances the scattering signal in imaging non-periodic objects, making the present very challenging experiments more accessible in the future and opening a new regime in non-linear atomic physics.¹⁹ In the field of quantum materials, increasing the resolution of off-resonant elastic and inelastic X-ray scattering experiments²⁰ is beneficial for probing and understanding the quantum states.²¹ Recent results in nonlinear Compton scattering also reveal the interest in increasing the peak power to bring the focused electric field closer to the Schwinger critical field.²² Finally, higher peak power is important for femtosecond protein crystallography and single particle imaging experiments using the “diffraction before destruction” principle.^23,24 These experiments require TW level, short pulses with a duration less than approximately 10 fs. In particular, this technique is attractive for obtaining diffraction patterns from small proteins which fail to form macroscopic crystals, such as the class of amyloid forming peptides underlying neurodegenerative diseases including Alzheimer's and Parkinson's.²⁵ 

With an eye towards these applications in this paper, we report the experimental demonstration of the FBSS method, paving the way to very high peak power, high brightness XFEL pulse generation. The FBSS technique uses either two different electron bunches or two single bunch slices, one to generate the seed signal and the other to amplify it in a tapered undulator to very high peak power. In conventional, single bunch, self-seeding operation, the beam energy spread must be kept small to allow amplification in the tapered undulator. As a consequence, the seed power and the output power are limited since the output power increases for a larger seed due to increased gain from undulator tapering. The FBSS scheme eliminates the trade-off between the seed power at the exit of the monochromator and the electron slice energy spread in the seeded undulator section.²⁶ Overcoming this trade-off is critical for high efficiency XFELs as it improves electron capture and reduces sensitivity to the sideband instability.^14,18,27 Furthermore, the scheme enables the control of the amplified seeded pulse duration^28–30 below the value set by the self-seeding monochromator, an important feature for experiments sensitive to radiation damage. The FBSS X-ray pulses obtained in our experiment are of high power (50 GW), narrow bandwidth (<8 × 10⁻⁵), and short duration (<10 fs). We measure and compare the FBSS performance with SASE and self-seeding performances at the same photon energy.

A schematic of the FBSS system used in the experiment is shown in Fig. 1. Selectively lasing with different slices of the electron beam is achieved with the fresh slice method.³² The electron bunch experiences a head-tail transverse kick by the wakefield generated in a structure, the dechirper,^33–35 set to an offset from the machine axis. Orbit correctors are used to steer the bunch on a tail-lasing orbit before the first undulator section. A saturated photon pulse is produced in the first section on the bunch tail. Subsequently, the photon pulse hits the diamond monochromator. A narrow bandwidth portion of it is diffracted and then discarded. The transmitted X-ray pulse presents a wide-bandwidth short pulse followed by a long narrow-bandwidth tail.³⁶ The chicane is used to delay the electron bunch head such that it is overlapped with the narrow-bandwidth tail of the photon pulse. The bunch orbit is switched to a head-lasing one around the chicane, and the monochromatic seed is amplified by the fresh-electrons on the bunch head in the second undulator section. The motion of the dechirper jaws can be adjusted in real time, enabling the control of the width of the lasing electron slice and consequently the X-ray pulse duration, in both the SASE and seeded undulator sections.³² 

The demonstration of FBSS was performed using the Linac Coherent Light Source (LCLS) at the SLAC National Accelerator Laboratory. The electron bunch parameters were 11.1 GeV energy, 4 kA current (core), normalized transverse emittance $ϵ x , y ∼$ 0.4 μm, and photon energy of 5.5 keV. The machine was set-up to collimate the charge from 250 pC at the cathode to 180 pC in the undulator.^37,38 We have chosen this photon energy because the LCLS power gain length at 5.5 keV is short enough to reach SASE saturation before the monochromator. Having a shorter gain-length allows us to reach higher seed power in the first undulator section, whose length is limited to 14 undulator segments. While it is possible to reach saturation earlier at longer wavelengths, the diamond crystal damage threshold limits the maximum tolerable seed power for safe operation.³⁹ The pulse duration was measured by imaging the electron beam longitudinal phase space using an X-band transverse deflecting cavity.⁴⁰ This important diagnostic allows us to determine the location and width of the lasing electron slice for a given dechirper setting (gap and offset) and a given bunch trajectory in the undulator. The radiation spectrum was measured with a hard X-ray bent crystal spectrometer.⁴¹ 

An optimal SASE configuration was first established with the dechirper and self-seeding monochromator extracted and all undulators inserted. The undulator segments allow small tuning of the undulator parameter $K = e B u / m e c k u$⁠, where B_u and k_u are the peak magnetic field and undulator wavenumber, in the range of 3.44–3.51. The restricted K tuning range limits the achievable FBSS power enhancement from tapering compared to a fully gap-tunable undulator.¹⁴ The average SASE pulse energy over 1470 shots was 2 mJ with a 40 fs pulse duration and a FWHM (±rms) spectral bandwidth of 9.69 ± 0.93 eV (see Fig. 2(a)). Using the optimal SASE configuration, we adjusted the vertical dechirper jaw and steered the tail on axis in the first undulator section. We adjust the matching in the first undulator section for the tail slices to account for quadrupole focusing from the dechirper wakefields.⁴² We applied a horizontal kick at the location of the self-seeding chicane to suppress all lasing downstream. As shown in Fig. 2(c), the FBSS performance was optimized by introducing a quadratic taper in the first undulator section to maximize the energy transfer from the electrons to the X-ray beam.^43,44 The average intensity before the monochromator was around 100 μJ in a short ∼5–10 fs slice on the tail of the electron beam (see Fig. 3(f)). We inserted the self-seeding monochromator and steered the core slices of the beam on-axis in the second undulator section. The chicane delay was set around 40 fs such that the core of the beam was overlapped with the first maximum of the narrow bandwidth seed wake pulse.³⁶ As illustrated in Fig. 3, the longitudinal phase space of the electron beam in the FBSS scheme shows two different slices with significant energy loss and large energy spread denoting strong lasing. At the tail of the bunch (t = 20 fs), the reconstructed X-ray power was 6 GW consistent with the measured pulse energy before the monochromator. In the core of the beam, the energy loss of the seeded electron slice is much larger, with some electrons losing ∼90 MeV of energy. The seeded X-ray power was 50 GW with a FWHM pulse duration of 9 fs. A fit of the integrated spectral intensity for the FBSS scheme (Fig. 2(c)) shows that 21% of the photons are in the SASE pre-pulse and 79% are in the seeded core defined by the bandwidth window $| Δ E seed |$ = 0.75 eV. The SASE radiation pre-pulse can be diverted on a timing tool and used for synchronization of the seed arrival time with experiments.

Using the same Bragg reflection on the monochromator, we compared the performance of FBSS with regular self-seeding. The self-seeded X-ray spectrum, pulse duration, and statistical properties are shown in Figs. 2–4. The bandwidth of self-seeding was 0.44 eV comparable to the 0.4 eV in the FBSS scheme. In both cases, the relative electron slice energy spread $∼ 10 − 4$ increases the bandwidth beyond the value set by the monochromator (0.1 eV) and the Fourier limit (0.18 eV and 0.06 eV for FBSS and self-seeding respectively). The small relative increase in self-seeding compared to FBSS may be attributed to the additional nonlinearities in the electron beam phase space sampled by the longer seed pulse.⁴⁵ The X-ray pulse duration was now set by the length and duration of the monochromatic wake and spanned most of the electron bunch of ∼25–35 fs FWHM. The average (±rms) pulse intensity was thus larger for self-seeding at 573 ± 290 μJ compared to 330 ± 179 μJ for FBSS. The shorter pulse duration in the FBSS case, however, gives higher peak power, with around 50 GW compared to 30 GW for self-seeding as shown in Fig. 3. The ∼50% fluctuations in intensity are due to the intrinsic SASE fluctuations at the monochromator and electron beam energy jitter.⁹ We note that in regular self-seeding operation, the first 2 undulators are extracted to limit the electron bunch energy spread after the monochromator and optimize performance. From the spectral data in Fig. 2, the average bandwidth of self-seeding and FBSS is a factor of 22–24 narrower than SASE. We compare the relative brightness, defined as the number of photons per unit phase space volume, of FBSS with that of SASE and self-seeding.^1,46 We assume that the transverse phase space area is the same in all three cases, the bandwidth is the FWHM value for the average spectrum, and the number of photons is given by the average intensity scaled by the fraction of photons within the FWHM bandwidth window. The estimated brightness ratios are $B FBSS / B SASE =$ 12.5 and $B FBSS / B self - seeding =$ 2.4. We note that since the bandwidth is very similar in FBSS and self-seeding, the increase in brightness is due to the increase in power in the FBSS scheme.

In conclusion, in this paper, we report the experimental demonstration of short, high brightness, narrow bandwidth hard X-ray pulse generation using the FBSS scheme in an XFEL. The scheme consists in lasing with different slices of the electron beam before and after the self-seeding monochromator, in order to circumvent the trade-off between the seed power and energy spread, which limits the performance of self-seeded systems. Selective lasing is achieved by introducing transverse oscillations in bunch slices and steering different slices of the electron bunch on axis before and after the self-seeding monochromator. We describe how this method increases the peak power and brightness of XFEL pulses compared to SASE and self-seeding with measurements of the X-ray intensity, pulse duration, and spectrum at a photon energy of 5.5 keV. We estimate a brightness increase of around a factor of 12 for the FBSS scheme compared to SASE and a factor 2 compared to self-seeding at this energy. We note that the FBSS scheme is a convenient method for generating high intensity self seeded X-rays with a short pulse duration (<10 fs), not set by the temporal width of the monochromatic wake generated by the monochromator. Following this successful demonstration, this method can be used in future XFELs aiming to achieve TW peak powers, avoiding the parasitic time dependent effects which limit the efficiency of tapered self-seeded systems.

The authors would like to thank D. Ratner and the LCLS operation group for their support. This research was partly supported by the U.S. DOE Office of Science Grant Nos. DE-SC0009983 and DE-AC02-76SF00515.