X-ray free-electron lasers (XFELs) offer extremely intense and brief femtosecond pulses, enabling the outrunning of radiation damage in biological samples. This capability allows the study of protein dynamics under near-physiological conditions at room temperature, avoiding the need for sample freezing as in conventional techniques. While flash X-ray imaging has advanced, resolution remains limited by the large dynamic range of diffracted intensities and the weakness of the diffraction signal. Averaging over a vast number of single-particle snapshots is crucial for high-resolution data, a process previously hindered by low hit probabilities and limited pulse repetition rates of older XFEL facilities. The European XFEL (EuXFEL), utilizing superconducting linear accelerator technology, overcomes this limitation with its high repetition rate, enabling an era of high-intensity, high-repetition-rate, and high-data-rate XFELs. However, this high repetition rate presents challenges, requiring advancements in sample injectors and X-ray detectors to ensure that the plasma created by each pulse does not interfere with subsequent measurements. This study aimed to demonstrate the feasibility of single-particle imaging at the EuXFEL's high intrabunch repetition rate using the Single Particles, Clusters, and Biomolecules & Serial Femtosecond Crystallography (SPB/SFX) instrument and the Adaptive Gain Integrated Pixel Detector (AGIPD).

Previous research established the principle of outrunning radiation damage with femtosecond X-ray pulses at the FLASH facility (Chapman et al., 2006). This technique, now routine in serial femtosecond crystallography, has been applied to image various biological samples, including living cells, organelles, and viruses (Bogan et al., 2008; van der Schot et al., 2015; Hantke et al., 2014; Munke et al., 2016; Reddy et al., 2017; Seibert et al., 2011; Ekeberg et al., 2015). Despite advancements, the resolution of 3D reconstructions remained limited, typically around a dozen voxels (Kurta et al., 2017; Rose et al., 2018; Lundholm et al., 2018). This limitation is attributed to factors such as the large dynamic range of diffracted intensities, the weak diffraction signal, and shot-to-shot variations in imaging conditions. High repetition rate serial crystallography at the EuXFEL has been demonstrated (Wiedorn et al., 2018; Grünbein et al., 2018; Yefanov et al., 2019), establishing a foundation for this study which extends that to single particle imaging.

The experiment utilized the SPB/SFX instrument at the EuXFEL, employing a 9.2 keV X-ray beam focused to a 15 × 15 µm² spot. Data were collected over five 12-hour shifts in December 2017 at 300 pulses per second, with an intrabunch repetition rate of 1.1 MHz. Iridium(III) chloride (IrCl3) and cesium iodide solutions, as well as Mimivirus and Melbourne virus samples, were aerosolized and injected into the beam using a gas dynamic virtual nozzle (GDVN). Diffraction patterns were recorded using the AGIPD detector. The experimental setup involved a three-slit collimation system to minimize background scattering. Data analysis included characterizing the instrument background, determining the particle size from diffraction patterns, and assessing the correlation between pulses within a train. Particle size was determined using a scattering model fitted to the diffraction patterns (using the spherical Bessel function of the first kind), considering factors like incident photon fluence and beam alignment. The detector’s performance was also characterized by analyzing its response to varying photon signals. The background was separated into instrument background and injection background. Hit/nonhit image classification was performed by counting lit pixels, significantly above the average, as recorded by the AGIPD. Size estimations of Mimivirus were refined using a continuous wavelet transform (CWT)-based procedure. Pulse independence was verified by analyzing the distribution of incident photon fluences and particle sizes for different pulses within trains.

The experiment yielded a large dataset of diffraction patterns: 11,255,800 frames were recorded, with 557,675 identified as hits. Analysis of IrCl3 scattering patterns provided estimates of particle sizes (80-800 nm diameter) and incident photon fluences (up to 2.8 × 10⁹ photons/µm²). A sensitivity limit was determined, indicating the smallest detectable particle size for the given setup. Background scattering was characterized, with a median background of about 4 × 10⁻⁴ photons per pixel. The variation in the position of diffraction pattern centers was low (order of magnitude lower than in similar LCLS measurements), indicating good instrument stability. Analysis of Mimivirus diffraction patterns showed a bimodal size distribution, with a peak around 500 nm corresponding to the virus particles and another at a smaller size potentially due to contaminants or aggregates. A filtering procedure, based on particle size estimation, significantly improved the percentage of single-hit patterns (from 39% to 88%). Analysis of pulses within trains showed no significant correlations between consecutive pulses, indicating that the pulses are independent, and thus that debris from one pulse does not interfere with the measurement of subsequent pulses. The hit probability was independent of the pulse's position within the train.

The findings demonstrate the feasibility of single-particle imaging at megahertz repetition rates at the EuXFEL. The low background noise and high stability of the instrument contribute to the successful acquisition of high-quality diffraction data. The maximum beam fluence obtained is consistent with expectations given the experimental conditions. The observed bimodal size distribution in Mimivirus analysis suggests a need to optimize the sample delivery method (e.g., using electrospray instead of GDVN) to minimize contamination and improve data quality. The independence of consecutive pulses within a train confirms the suitability of the EuXFEL's high repetition rate for single-particle imaging experiments. The high data throughput afforded by megahertz repetition rates, along with refinements in sample delivery and detector technology, has the potential to significantly advance the field by allowing for higher resolution imaging.

This study successfully demonstrated megahertz single-particle imaging at the EuXFEL, despite initial limitations in experimental parameters like focal spot size and wavelength. The instrument's stability and low background are promising for future experiments. Further improvements are expected with upgraded focusing optics and optimized sample delivery techniques (e.g., electrospray). The demonstration of independent pulses within a train paves the way for high-repetition-rate, high-data-rate single-particle imaging at XFELs, promising substantial advancements in biomolecular imaging and structural biology.

The experiment was conducted with a limited set of parameters (photon energy of 9.2 keV and a relatively large focal spot) due to limitations of the instrument at the time of the experiment. The use of GDVN resulted in a bimodal size distribution, suggesting the need for improved sample delivery techniques like electrospray for cleaner samples and reduced contamination issues. Higher resolution reconstruction of images from these single particle imaging is an area that needs improvement.