Addressing Artifacts in X-ray Photon Correlation Spectroscopy (XPCS) Data Analysis

X-ray Photon Correlation Spectroscopy (XPCS) data, like most experimental processes, has both elevated noise levels and the presence of various artifacts, which disrupt the extraction of dynamics of the sample under study. While extensive attention has been given to addressing noise in XPCS experiments, our focus is dedicated to tackling other types of non-random artifacts. These include detector and experimental setup bias, peak shape variations, and the presence of secondary peaks, such as that which could be caused by stray background light. These artifacts manifest in the dynamics as shoulders, inflated values, and unphysical upturns in calculated intensity-intensity autocorrelation functions.

Our work utilizes XPCS data from LESCO at BNL's CHX beamline as a case study, where we have developed a methodology to identify and eliminate these artifacts. Our approach emphasizes the importance of proper modeling for both the intensity-intensity autocorrelation function and the chosen region of interest on the detector. By using Gaussian modeling of peaks and prior knowledge of the phase-temperature relationship, accurate models can be developed to eliminate many of these artifacts. Furthermore, we delve into the discussion of optimizing this process using Machine Learning techniques, offering a promising path for future improvements.