In a laser-plasma accelerator1, the interaction of a high-intensity laser pulse with a plasma creates a trailing density modulation, the plasma wave, which supports electric fields several orders of magnitude larger than those provided by modern RF accelerator cavities. Correctly controlled, the plasma wave can trap electrons from the plasma background and then accelerate a well-confined phase-space volume, resulting in a highly relativistic, high-brightness electron beam from only a centimetre-scale plasma.
The field has seen rapid progress in recent years. A series of landmark experiments demonstrated advanced concepts to generate and characterize high-brightness beams10,11,12,13,14,15, new laser guiding concepts have extended the interaction length of laser and plasma to result in electron beams of GeV and higher energies5,6,7,8,9, first steps towards continuous operation have been made16 and, recently, the long anticipated first gain from a plasma-driven free-electron laser was reported17.
However, the reproducibility and stability of today’s laser-plasma accelerators are still less developed than those of modern RF machines. This can be linked to the micrometre-scale size of the plasma cavity, which leads to extreme accelerating fields and inherently short femtosecond electron bunch durations, but also makes it very challenging to precisely control the injection and acceleration process. Also, a new plasma cavity is created with every laser shot. Because the plasma cavity is essentially formed by the radiation pressure of the laser, even subtle variations of the drive pulse can result in a modified plasma cavity, thus changing the acceleration fields and dynamics.
The resulting percent-level energy spread and energy jitter typically associated with laser-plasma electron beams are particularly damaging and still effectively prevent laser-plasma accelerators from becoming a viable alternative accelerator technology. For example, free-electron lasers require permille-level energy spread beams18, whereas injectors for synchrotron light sources have a tight one-percent-level energy acceptance19,20. More generally, transporting large energy spread beams from the accelerator to an application can result in adverse chromatic effects that quickly degrade the beam quality available at the interaction point.
To address these challenges, tailored drive laser systems21 and the deployment of active stabilization techniques22,23,24 are expected to improve the performance of future laser-plasma accelerators, but the implementation of these concepts remains challenging.
A more fundamental approach is to improve the spectral properties of laser-plasma electron beams by exploiting their inherent short bunch duration and high peak currents.
For example, a decompression technique was proposed25,26, which stretches the bunch longitudinally to introduce an energy–position correlation (chirp) and thereby locally reduces the energy spread at the expense of peak current. This technique has enabled the demonstration of a seeded free-electron laser from initially large energy spread laser-plasma electron beams27, although scalability of the concept beyond a proof-of-principle experiment remains unclear.
Other techniques are based on passive structures to remove the energy chirp of a decompressed beam. In these dechirpers, the interaction of the electron bunch with a corrugated pipe28, a dielectric structure29 or a plasma30,31,32,33 drives an electric field that effectively removes the correlated energy spread. However, because a passive dechirper is driven by the electron bunch itself, any small variation in bunch length, charge or current profile also affects the dechirping result, and as they also do not correct the beam energy jitter, stability concerns remain.
More recently, it has been proposed to add an accelerating, that is, active, structure after decompression34,35,36,37 to greatly reduce both the correlated energy spread and energy jitter38,39, thereby addressing shortcomings of previous concepts and providing laser-plasma-generated electron beams of unprecedented quality and reproducibility.
In the following, we experimentally demonstrate, for the first time to our knowledge, active energy compression of a laser-plasma-generated electron beam. We improve the beam spectral properties by more than an order of magnitude and demonstrate performance previously only associated with modern RF accelerators.
Our energy compression scheme is illustrated in Fig. 1. A laser-plasma accelerator provides several-micrometres long, kiloampere peak current electron beams (Fig. 1a) of several-percent energy spread and energy jitter (Methods).
a, The laser-plasma accelerator provides a several-femtosecond duration electron bunch that has a several-percent energy spread and an energy deviation from the reference energy Eref. b, In a subsequent magnetic chicane, energy-dependent path-length differences result in a longitudinal energy chirp, which effectively stretches the bunch from micrometre to millimetre length. c,d, The RF field of a dechirper cavity then compensates the energy chirp by accelerating low-energy electrons while decelerating high-energy electrons, resulting in a narrowband energy-stabilized beam (d).
The laser-plasma accelerator is followed by a magnetic chicane. Here the first dipole introduces an energy-dependent deflection angle. The electron trajectories are then parallelized by a second dipole of inverse field. A third and fourth dipole close the symmetry and bring the beam back on the design axis. The chicane thereby introduces energy-dependent path-length differences that effectively stretch the bunch longitudinally (Fig. 1b) and induces an energy–position correlation (energy chirp).
After the chicane, the beam goes through an accelerating RF cavity, in which the positive gradient of the accelerating field cancels out the previously induced energy chirp (Fig. 1c). Through this mechanism, the set-up also removes the energy jitter: the electron spectrum is compressed to the electron energy that overlaps the zero crossing of the RF field (Fig. 1d).
Ideally, the beam energy spread is reduced proportionally to the bunch stretching, which can be more than two orders of magnitude. In practice, however, both the energy chirp introduced by the chicane and the sinusoidal RF field have small but non-negligible nonlinear terms that limit the energy compression. Yet, even including those nonlinear contributions, an energy spread reduction by more than an order of magnitude is readily possible (Methods). The scheme is thus ideally suited for short, high-current electron beams, as provided by a plasma accelerator, and applications that require only moderate peak current.
We have demonstrated this concept experimentally at the LUX laser-plasma accelerator. The drive laser provides 2.2-J, 35-fs (full width at half maximum (FWHM)) pulses on target at 1-Hz repetition rate. Through the interaction with a 5-mm-long plasma source, the set-up provides electron beams with an energy of 257 megaelectronvolts (MeV) at 41 pC (13 pC rms) of charge and a typical energy spread of 1.8% and energy jitter of 3.5%. From simulations, we estimate an initial bunch length of about 2 µm (rms) which corresponds to a peak current of 2.5 kA (Methods).
After the target, electrons are transported to the magnetic chicane, characterized by the chicane strength parameter R56 = 100 mm, which stretches 1% energy spread beams by a factor of about 1,000 to 1 mm length and induces an energy chirp of 1.0% per mm. The dechirper cavity (Methods) is a 5-m-long RF structure operated at 10-cm wavelength (S-band) and can change the beam energy by about 50 MeV.
After dechirping, electron beams are sent into a spectrometer and dispersed by a dipole magnet onto a scintillating screen to record the energy spectra with a resolution of order 0.07% (Methods). Not shown in Fig. 1, we have implemented several diagnostics throughout the set-up, including scintillating screens to measure the electron beam transverse profile and beam position monitors to non-invasively measure the transverse beam positions and charge (Methods).
First, we calculated the RF amplitude to remove the energy chirp, which is 45.4 MV (see Methods) for our chicane of R56 = 100 mm. We then scanned the RF phase to compress the electron beam energy (see Fig. 2).
a, Phase scan between electron bunch and RF field in steps of 15° (ref. 40). For better visibility, the spectral density is normalized for each step, averaging over 50 shots each. The shifted electron energy (red dots) follows the sinusoidal RF field (dashed black line). The energy jitter is denoted by red bars. b, At optimum compression, the energy spread is minimized. c–f, For illustration, we calculate the phase space of an initially chirped electron bunch (light blue) after interaction with the RF (dark blue) at distinct phases.
As we scan the phase, the median electron beam energy follows the sinusoidal RF field (red dots).
At 0°, the bunch is centred at the zero crossing of the RF electric field (Fig. 2a). Electrons at the head of the bunch are decelerated, whereas electrons at the back of the bunch are accelerated, effectively reducing the chirp and, thus, energy spread. The opposite effect happens at a phase of ±180°, at which the slope of the RF is inverted: electrons at the head of the bunch are now accelerated, whereas electrons at the back of the bunch are decelerated, effectively increasing the chirp and broadening the spectrum. Around ±90°, the bunch is collectively decelerated and accelerated, respectively, which shifts the energy spectrum.
Notably, we find the smallest energy spread not at 0° but at a slightly shifted phase of −23.6° (Fig. 2b), which can be understood as follows. The second-order dispersion of the chicane adds a small curvature to the linear energy chirp. Therefore, we need to operate the RF slightly below 0°, at which the small curvature of the sinusoidal RF field just compensates the nonlinear chirp (Methods). Operating instead at 0°, the RF field is almost linear and cannot compensate the curvature of the chirp, resulting in a larger energy spread. In general, owing to these nonlinearities, there is a unique pair of amplitude and phase that can be calculated analytically (Methods) and results in the smallest possible energy spread.
At the optimum set point with minimum energy spread, we recorded about 1,000 shots with the RF turned off and on, shown in Fig. 3.
a, Series of approximately 1,000 energy spectra on the electron spectrometer with RF off and on40. b, Average spectral density before (blue) and after (green) energy compression. The uncompressed spectrum is scaled by a factor of 10 for better visibility. The aperture of the beam pipe in the chicane defines a transmission window ranging from 227 to 285 MeV.
With the RF on (Fig. 3a), the energy jitter reduced by a factor of 72 from 3.5% to 0.048% and the energy spread reduced by a factor of 18 from 1.8% to 0.097%. Operating at a phase of −23.6°, the fully energy-compressed beam was shifted from a median energy of 257 MeV (RF off) to 275 MeV (RF on). The energy-compressed beams have a mean charge of 32 pC (12 pC rms). The peak spectral density reached as high as 70 pC per MeV. About 50% of all shots feature a sub-permille energy spread (Extended Data Fig. 1). Some shots were compressed to an energy spread as small as 0.068%, which is at the estimated resolution limit of our electron spectrometer. With compression, the charge provided inside a ±1% window of the median energy improved from 18.1% to 99.9%.
These results correspond to the best energy compression settings in the experiment, but we can further explore different capabilities of the set-up.
For example, stretching the electron bunch more lowers the energy chirp and thus reduces the required RF amplitude and thus power to operate the cavity. As the longer bunch then covers a larger phase of the RF, nonlinearities will reduce the energy compression performance. We tested this behaviour (R56 = 170 mm, amplitude 28 MV) and could still reduce the energy jitter and spread to 0.09% and 0.13%, respectively, while consuming a factor of three less RF power.
Furthermore, we can vary the RF phase to fine-tune the target energy within a several-percent range without notable loss of compression performance.
In summary, our set-up provides electron beams with a performance in energy jitter and spread previously only obtained from modern RF accelerators, opening up widespread deployment of laser-plasma accelerator technology.
An important application for such an energy-compressed laser-plasma electron beam is an injector for a future synchrotron storage ring. This application takes full advantage of the picosecond-level, several-ampere electron bunches of permille energy spread and jitter that our set-up already delivers today. Direct storage ring injection, as recently proposed20, typically requires GeV-level electron beams. Recent work has already demonstrated laser-plasma accelerators delivering up to 10 GeV beam energy5,6,7,8,9 using advanced laser guiding schemes to extend the interaction length of the drive laser and plasma. Furthermore, using X-band RF technology, the energy compression set-up could scale to higher beam energies without a marked increase in footprint20,39. With further development of high-efficiency, high-average-power laser drivers, a plasma-based injector could become a compact and energy-efficient alternative to RF technology20.
Other applications requiring higher beam currents could use the stronger dechirping gradients of X-band RF technology39 or plasmas38 to improve the beam spectral properties while maintaining peak current.
