Quantifying optical scattering losses in pump-probe serial femtosecond crystallography experiments.

Ultrafast pump-probe serial femtosecond crystallography (SFX) experiments are typically performed with extremely high excitation laser fluence to maximize the occupancy of the light-induced state. This has been justified by assuming that a dominant fraction (up to 99%) of the pump laser light is scattered by the jetting medium, strongly reducing the pump laser fluence in the crystals which, therefore, do not absorb multiple photons and thus undergo the biologically relevant single-photon photoreaction. However, this notion is strongly contested in the field. To address the issue of pump laser scattering losses in high viscosity jets, we performed time-resolved SFX experiments on fatty acid photodecarboxylase crystals using high-viscosity extrusion (HVE) and both patterned and non-patterned sheet-on-sheet (SOS) fixed-target sample delivery systems on the Cristallina-MX instrument at SwissFEL. Tightly focused pump laser beams were used to mitigate light contamination (the undesired accidental illumination of neighboring crystals) in SOS chips, paving the way for their use in pump-probe investigations of lipidic-cubic-phase-grown membrane protein crystals. Determination of the light-induced reaction intermediate occupancies enabled a comparison of the influence of the delivery method on the pump laser fluence reaching the crystals. We show that optically transparent high-viscosity jets and chip crystal delivery methods result in photoproduct yields that approach 70-80% and 60-70%, respectively, of computationally derived yields that take into account excitation probabilities. These results demonstrate that clear viscous jets scatter pump laser light far less than has been claimed. Consequently, the use of excessive photoexcitation energy densities is neither needed nor justified in optically pumped time-resolved SFX experiments when using appropriately sized crystals.