Figure 2. SFX data as a function of excitation fluence.

The DrBphP_darkstructure (green) is shown together with the observed difference electron density, contoured at 3.5 σ, at 1 ps collected using (a) 0.2 mJ/mm^₂, (b) 0.4 mJ/mm^₂, (c) 1.3 mJ/mm^₂, and (d) 1.7 mJ/mm^₂. All spot sizes were computed assuming Gaussian line shapes with the $(1/e^2)$ convention. The data shown in panel A-C were collected at SACLA in May 2019, whereas the data shown in panel D was collected in October 2018. The same experimental setup was used in both occasions. The laser energy of the experiment in 2018 can be found in the Materials and methods section. The energies for the experiment in 2019 were 16 µJ, 42 µJ, and 106 µJ (panels A-C, respectively). During the experiment in 2019, the femtosecond laser beam was misaligned by 50 µm distance from the interaction spot between X-rays and jet in the direction of flow. The laser intensities were corrected for this displacement assuming a Gaussian line shape. The excitation fluence is similar to previous femtosecond time-resolved SFX experiments (Nogly et al., 2018; Pande et al., 2016; Barends et al., 2015; Coquelle et al., 2018); however, we found high scattering in the grease/buffer mixture (Figure 1—figure supplement 1). Since the crystallographic signals were reduced when lowering the excitation fluence and disappeared completely when reaching 1/10 of the maximum value, we conclude that the excitation fluence that actually reaches the crystals in the grease matrix is much lower than the incoming photon fluence and that the photoexcitation is in the single-photon regime.