Skip to content

Lasing of a cavity-based X-ray source

FieldValue
JournalNature
Published2026-01-28
AuthorsPatrick Rauer, Immo Bahns, Bertram Friedrich, Sara Casalbuoni, Massimiliano Di Felice
DOI10.1038/s41586-025-10025-x
Citations5

A 132.8 m diamond Bragg cavity produced multi-pass X-ray lasing at 6.952 keV. Maximum outcoupled pulse energies reached 4 to 13 μJ, and the cavity synchronized to 2.23 MHz electron bunches at the European XFEL. The result establishes cavity-based X-ray free-electron laser operation in an accelerator environment, with thermal load limiting stable output.

The experiment used the last four SASE1 undulator segments, which provided 20 m of active length. The electron energy was 14 GeV, and each bunch carried 250 pC.

The cavity placed two retro-reflector assemblies 66.41 m apart. Each assembly combined one 100 μm-thick diamond crystal with two perpendicular, 70 mm Kirkpatrick-Baez mirrors.

The diamonds had ⟨100⟩ surface cuts and served as (400) Bragg reflectors in near-backscattering geometry. This reflection set the nominal photon energy at 6.9517 keV.

The mirrors carried boron carbide coatings and had a meridional curvature radius of about 30 km. Their projected 200 μm by 200 μm apertures required submicroradian and submicrometre alignment precision.

Within each assembly, the crystal and mirror surfaces were perpendicular. The returning beam passed 2.67 mm from the incident beam inside the same vacuum pipe.

A 5 m magnetic chicane coupled the electron beam into the cavity, producing a 4 mm offset at the crystal position. Its R56 of -22 μm did not deteriorate the beam.

The bunches were tuned to reduce energy spread and emittance. At the undulator, peak current was about 6 kA. The rms bunch length was 6.5 μm, and the central chirp was 4 ± 0.5 MeV per micrometre.

The trajectory through the four undulator cells was aligned within ±10 μm. The measured SASE gain length was about 5 m, and the last four cells had an integrated gain of 48.

The seeded gain was lower because the first two gain lengths entered the FEL lethargy regime. Differences in pulse duration and transverse overlap also reduced seeded gain.

Cavity length was scanned by moving the downstream crystal-mirror assembly in 1 μm steps over a 6 mm range. Resonance produced a persistent saturated photodiode signal across the bunch train.

Alignment used four-degree-of-freedom stages with nanometre-scale positioning and nanoradian-scale angular precision. The final adjustment maximized the cold-cavity ring-down signal.

The gain-based scan gave a seeding range of σz = (0.7 ± 0.06) μm around resonance. The diode-based extended scan gave (1.15 ± 0.05) μm because early saturation broadened its response.

The resonance scan averaged the ten strongest trains at each 1 μm delay setting. Near resonance, the spectrometer signal increased by orders of magnitude. Off resonance, the stable signal was three orders of magnitude weaker.

The authors treated the cavity-length dependence as conclusive evidence of oscillator lasing. Successful operation required femtosecond synchronization and micrometre-scale transverse overlap between bunches and recirculating photons.

Ring-down analysis estimated per-round-trip cavity reflectivity at 67.3(4)% with seeding and 75% for the cold cavity. Both estimates were corrected for diagnostic grating losses.

The theoretical combined peak reflectivity was 96.8%. A simulation gave 85% after finite mirror size and measured imperfections were included.

The shortfall was attributed to crystal strain, wavefront distortion, aperture cut-off losses and mechanical misalignment. Mechanical loss was considered small because the cavity had been aligned by maximizing ring-down.

Simulations reproduced the main pulse evolution but remained more symmetric spectrally and predicted higher reflectivity. The authors cite uncertain electron phase space, unknown static crystal strain and component alignment as sources of disagreement.

The single-shot spectrometer resolved the spectrum’s evolution across the bunch train:

Spectral referenceBandwidthContext
First-bunch spectrum(21 ± 6) eV FWHMUnsaturated single-pass SASE, measured
Seeded spectrum(0.24 ± 0.06) eV FWHM65th and final spectrometer bunch, measured
Typical hard X-ray SASE≥0.1%Prior-work bandwidth reference

The HIREX single-shot spectrometer used 0.08 eV bins and a 61 meV Gaussian resolution kernel. No dedicated outcoupler was installed. Photons transmitted through the downstream crystal were monitored instead.

Because the crystal reflected the spectrum’s central part, only its wings were transmitted. This outcoupling geometry produced an asymmetric double-hump seeded spectrum.

Reference to a calibrated X-ray gas monitor set the pulse-energy scale. The monitor’s average photon flux carried 9% uncertainty.

The median gain per round trip was 0.4 in the linear-gain regime. Only outcoupled energy was measured, so the number of stored photons was expected to be higher but remained unquantified.

The diagnostic grating coupled 2.8% into first-order diffraction for the photodiode. It was retracted during integrated-energy measurement to minimize cavity loss.

The ring-up and ring-down traces were averaged over 1,000 pulse trains. Detector saturation limited quantitative use of the earliest ring-up pulses.

The pulse energy reached a maximum, decayed and oscillated within each train. The median over 2,700 trains smoothed much of this oscillation. The long period across several bunches could not be explained by shot-to-shot beam quality fluctuations.

The authors attributed this behavior to thermal feedback in the diamond crystals. First, nonuniform heating created lattice heat bumps and reduced overlap between reflected photons and the electron beam.

Second, temperature drifts of a few kelvin shifted the two crystals’ lattice constants. Their reflection curves then stopped overlapping, reducing the photon field and heat load. The lower load allowed the sequence to repeat, producing pulse-energy oscillation.

A measured upstream-downstream temperature difference of 2 K was included in the simulation. The simulation reproduced the main experimental behavior.

Cooling the diamonds to cryogenic temperatures is expected to reduce the thermal effect and raise photon output. The authors caution that increased thermal load may cause instabilities in the low-temperature regime.

The proof-of-concept output remained below the peak intensities of state-of-the-art SASE and self-seeded XFELs. Self-seeding improves spectral purity and spectral flux, but it does not fully address stability and coherence limitations.

The narrow seeded spectrum agrees with the central CBXFEL prediction. The authors present fully coherent X-ray pulses as a potential next step, not a measured user-beam result.

These performance figures remain simulated projections. They cover a ten- to hundred-fold pulse-energy increase and an increase of up to three orders of magnitude in spectral flux.

The projection exceeds the cited capabilities of self-seeding and SASE. Expected improvements depend on electron-beam and cavity optimization, cryogenic crystal cooling and advanced outcoupling.

The FEL process is expected to saturate around 1 mJ, above the present output. Further experiments will test these optimized configurations.

Engineering implications for X-ray sources

Section titled “Engineering implications for X-ray sources”

Diamond Bragg optics are validated for X-ray resonators, while the experiment identifies thermal feedback as a design limit. Recirculation around a few undulator segments may reduce the need for very long undulator lines.

Timing precision, angular alignment and thermal control therefore belong in source design rather than later commissioning. Diamond selection must account for crystal cut, Bragg orientation, strain and thermal deformation.

Future cavities also need a dedicated outcoupler to avoid using crystal transmission as the output channel. Further experiments will test optimized beam and cavity settings, cryogenic cooling and advanced outcoupling.

The authors identify millielectronvolt-resolved inelastic and nuclear resonance spectroscopy as prospective applications. They also list hard X-ray ultrafast pump-probe resonant inelastic scattering, nonlinear wave mixing and quantum X-ray optics.

These applications remain proposed uses rather than measurements in the present experiment. Thermal stability and controllable outcoupling are the next engineering requirements for moving beyond the proof of concept.

The cavity needs two 100 μm-thick single-crystal diamond Bragg reflectors with ⟨100⟩ surface cuts. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal diamond plates and wafers up to 125 mm, with thickness from 0.1 to 500 micrometers.

The specified 100 μm thickness falls within that range, while cavity-specific integration can be reviewed through engineering support before ordering optics.

View Original Abstract

Abstract The invention of the laser transformed optics by providing intense, coherent light in the visible region, but extending this concept to X-rays has been hindered by a lack of suitable gain media and mirrors. Current hard X-ray free-electron laser (XFEL) facilities 1-5 overcome this by amplifying shot noise from a high-peak-current electron bunch via self-amplified spontaneous emission 6 in a single pass through long undulators, delivering very high brightness but with a noisy, multi-spiked temporal and spectral profile. Cavity-based XFELs (CBXFELs) 7-9 were proposed to close this gap by recirculating spectrally filtered X-ray pulses in a Bragg-reflecting cavity synchronized to a high-repetition-rate electron beam. Here we show lasing with multi-pass gain at 6.952 keV in a 132.8-m round-trip diamond-based Bragg cavity 10 at the European XFEL, matched to the 2.23-MHz bunch spacing of the superconducting accelerator 5 . Under stringent length and angular stability requirements, a ring-up in the cavity across successive bunches was observed, producing spectrally pure, microjoule-level pulses. This establishes the feasibility of CBXFELs in an accelerator environment and validates diamond Bragg optics for X-ray resonators. The demonstrated spectral purity opens a path to next-generation X-ray science, which demands highly coherent, stable sources.