Femtosecond resolution pump-probe experiments are now routinely carried out at X-ray Free
Electron Lasers, enabled by the development of cross-correlation “time-tools” which correct the
picosecond-level jitter between the optical and X-ray pulses. These tools provide very accurate,
<10 fs, measurement of the relative arrival time, but do not provide a measure of the absolute
coincidence time in the interaction. Cross-correlation experiments using transient reflectivity in a
crystal are commonly used for this purpose, and to date no quantitative analysis of the accuracy or
stability of absolute coincidence time determination has been performed. We have performed a
quantitative analysis of coincidence timing at the SACLA facility through a cross-correlation of
100 6 10 fs, 400 nm optical pulses with 7 fs, 10.5 keV X-ray pulses via transient reflectivity in a
cerium-doped yttrium aluminum garnet crystal. We have modelled and fit the transient reflectivity,
which required a convolution with a 226 6 12 fs uncertainty that was believed to be dominated by
X-ray and laser intensity fluctuations, or assuming an extinction depth of 13.3 lm greater than the
literature value of 66.7 lm. Despite this, we are able to determine the absolute coincidence time to
an accuracy of 30 fs. We discuss the physical contributions to the uncertainty of coincidence time
determination, which may include an uncharacterised offset delay in the development of transient
reflectivity, including cascading Auger decays, secondary ionisation and cooling processes.
Additionally, we present measurements of the intrinsic short-term and long-term drifts between the
X-rays and the optical laser timing from time-tool analysis, which is dominated by a thermal
expansion of the 25 m optical path between tool and the interaction region, seen to be 60 fs over
a period of 5 h.