LCLS Proves 100–400 Attosecond Hard X-Ray Pulses With Copper Fluorescence

Researchers at SLAC's Linac Coherent Light Source (LCLS) have shown that the facility can deliver hard X-ray pulses lasting 100 to 400 attoseconds. The pulses had photon energies of about 9 keV. The achievement is mostly a measurement result. According to the team's paper in Nature Photonics, it is an experimental confirmation of attosecond pulse generation in the hard X-ray regime. The accelerator physics was largely in place already. What was missing was a ruler short enough to check it.

01The problem was proof, not production

LCLS operators have been shaping electron bunches to emit very short X-ray flashes for years. SLAC reported the first attosecond hard X-ray laser pulses at LCLS back in 2017. Agostino Marinelli, a SLAC professor and study author, put the difficulty plainly. Over several years the team learned to create these pulses, but measuring pulses this short is very hard. He described the new work as the first diagnostic that proves the pulses are only a few hundred attoseconds long.

Conventional pulse-length diagnostics have trouble at this scale and photon energy. Laser-based streaking methods have been used for soft X-rays. Hard X-rays at around 9 keV are harder to characterise, because few materials respond to them in an easily readable way. The team instead used a nonlinear process that scales steeply with pulse intensity.

02How the pulses are made

The preprint describes the production side. The pump pulses are emitted from a current spike in the electron bunch. The bunch is highly compressed, with a peak current of about 10 kA. The preprint reports 9.05 keV photon energy. A related LCLS paper from the same community says this kind of spike comes from photocathode laser shaping. In that technique the drive laser's temporal profile is tailored so that the compressed electron bunch develops a short, high-current feature. Only the part of the bunch inside that spike lases efficiently, so the X-ray pulse is far shorter than the bunch.

Physics sets the floor. As an earlier LCLS paper explains, the shortest pulse an XFEL can produce is limited by its amplification bandwidth, which is roughly the same size as its extraction efficiency. At hard X-ray energies the LCLS bandwidth can support pulses shorter than 1 fs. Pulses of a few hundred attoseconds therefore sit well within what the machine allows in principle.

03Using copper fluorescence as a stopwatch

Lead author Ichiro Inoue of the University of Tokyo proposed using amplified spontaneous emission (ASE) to measure the duration of LCLS pulses indirectly. In ASE, intensely focused X-ray pulses excite a specific element, which then fluoresces in the forward direction. According to the Nature Photonics abstract, the analysis rests on the yield of this collective X-ray fluorescence, which is sensitive to how short the exciting pulse is.

The setup, from the SLAC release and the preprint, works as follows:

• Beam characterisation. Part of the beam goes to a detector that measures its energy and spectral shape. In the preprint this is a transmissive intensity monitor, which reads X-ray scattering from a 10-µm-thick Kapton film. A thin bent silicon (220) crystal spectrometer sits alongside it. Both measure each shot without destroying it.
• Nanofocus. Advanced Kirkpatrick-Baez total-reflection mirrors focus the pulses to a 150 nm FWHM spot.
• Target and readout. The focused beam hits a 20-µm copper foil. The ASE is measured with a flat Si (220) analyser crystal and a two-dimensional detector.

The logic is that for a given pulse energy and spectrum, a shorter pulse has higher peak intensity and so produces more ASE. Measuring pulse energy and spectrum shot by shot, and comparing them with the fluorescence yield, lets the team infer duration. The researchers report shot-to-shot fluctuations in duration, ranging from 100 as to 400 as. This is a model-dependent inference, not a direct time-domain trace. The paper's theory section covers how ASE yield maps to duration. This article does not assess that part.

04What the numbers say about pulse quality

The paper reports an inferred time–bandwidth product of about 2 fs·eV for 100-as pulses, indicating that a fraction of the pulses is nearly transform-limited. A transform-limited Gaussian pulse sits near 1.8 fs·eV (a standard figure, not from the sources). A value of about 2 means the pulses have little excess spectral width for their length. The phrase "a fraction of the pulses" matters. The preprint excerpt also notes that approximately 6.5% of pulses showed a single spike, though the excerpt is truncated. So this is a statistical picture of a fluctuating source. It is not a guarantee of 100 as on every shot. Users who need clean single-spike pulses would have to select or tag them shot by shot using the same kind of diagnostics.

05What the timing unlocks

The SLAC release contrasts the two X-ray regimes. A 2019 Marinelli-led team generated soft attosecond X-rays below 5 keV, which suit studying electron movement in chemical bonds. Hard X-rays above 5 keV penetrate deeper and can image with enough resolution to locate electrons at atomic scale. Together these give a probe that is fast and penetrating at once.

Dense plasmas. Sam Vinko of the University of Oxford studies the hot, dense matter found in stars and inertial fusion experiments. According to the release, until now X-rays were either too weak in energy to penetrate these plasmas or too slow to capture their short-lived core-excited states. This spring at LCLS, he used pulses that were energetic enough to get through and fast enough to freeze the plasma before it evolved. The release calls this a first at any XFEL. His comment on the usual alternative was blunt: experimenters normally rely on modelling and assumptions about what is happening.

Damage avoidance. Inoue says the pulses can capture electronic states in their pristine state, essentially outrunning the damage that X-rays themselves cause. This only works if the pulse is shorter than the damage processes, which is why a verified duration matters.

Imaging. Tais Gorkhover of the University of Hamburg used attosecond soft X-rays at LCLS for diffraction imaging. She expects attosecond hard X-rays to sharpen nanoscale absorption and scattering images further, and to help separate nuclear from electronic dynamics.

06Implications for instrument design and open questions

Nanofocusing was central to this experiment. A 2025 Optica memorandum co-authored by Inoue reports focused attosecond hard pulses with intensity exceeding 10^19 W/cm². That level is what makes ASE a usable diagnostic. For instrument designers, this has several consequences:

1. Pulse-duration diagnostics should be built into end stations as routine instruments, not one-off experiments.
2. Shot-to-shot metrology of energy, spectrum and duration has to run at the repetition rate and have enough throughput for selection.
3. Optics for attosecond pulses must preserve both tight focus and very large bandwidth. The 2026 follow-on preprint reports bandwidths exceeding 30 eV, a factor of two beyond previous demonstrations. Related preprints cover a hard X-ray attosecond split-delay line for pump-probe work.

Several things remain unproven. The diagnostic uses a specific class of target (3d transition metals) and needs very high intensity, so it is not a general-purpose, non-invasive monitor. The duration inference depends on a theory of ASE yield. The spread from 100 to 400 as shows the source is not yet deterministic. The plasma result Vinko describes is reported through the SLAC release, and this article did not locate a peer-reviewed paper for it. The Nature Photonics result gives other facilities and users a basis for comparison. It also sets a concrete target: stable, selectable, verifiable attosecond hard X-ray pulses delivered to experiments on demand.