Artifact Identification in X-ray Diffraction Data Using the Gradient Boosting Method

The in-situ synchrotron high-energy X-ray powder diffraction (XRD) technique is widely used to study the crystalline structures of materials in many settings, such as battery materials and extreme environments like diamond anvil cells or synthesis reactors. The technique involves analyzing XRD images captured under various conditions for identifying atomic structure based on its diffraction patterns, which provides information on the material's structure, such as defects, microstrain, and crystallite size. The XRD images are typically captured using area detectors and show diffraction patterns that can appear as powder diffraction rings or display other characteristics like textures, preferred orientations, and/or single crystal diffraction spots. The exclusion of single crystal diffraction spots allows a precise analysis of the powder diffraction rings of interest. In this paper, we investigate single crystal diffraction spot separation and identification in XRD images using gradient boosting techniques. The gradient boosting method was found to have high accuracy and reduced time compared to conventional methods when trained on small, diverse datasets. A more accurate study of the diffraction rings can be achieved by excluding artifacts from the XRD image integration.