Small-angle X-ray scattering (SAXS) is a versatile method to characterize the hierarchical structure of cellulosic materials under various conditions, including different moisture states. However, applying it to resolve the nanostructure of wood cell walls and other cellulosic materials requires complex data analysis, which aims to convert the scattering patterns into realistic real-space representations. Such analyses are typically based on oversimplified models and neglect a considerable part of the experimental scattering data. More automatized scattering data analysis methods are also needed for high-throughput synchrotron experiments.
In this collaborative project, we aim to create a workflow to turn SAXS data from wood materials into three-dimensional real-space representations. Our approach is to develop a numerical toolkit to generate model structures based on a set of structural input parameters and calculate the scattering patterns from the generated structures. We then use machine learning to invert the problem to obtain the set of parameters best describing a scattering pattern. The structural parameters consist of quantities like volume fraction of crystalline microfibrils, their size distribution, curvature and agglomeration tendency. Heterogeneities in the microscale structure and variations in orientation will also be considered. To ensure that the generated structures are realistic, physical constraints for instance based on electron tomography experiments are used to limit the explored parameter space. The electron tomography data can also be compared with experimental SAXS data via computed scattering patterns. 
Our target is to make the data analysis workflow publicly available, so that it can be used by synchrotron users to analyze their data measured from wood samples. It can also be further adapted to other cellulosic systems, including pulp fibers and cellulose nanomaterials. We expect that this automatized data analysis workflow will give views into the diverse structure of wood cell walls with unprecedented efficiency, based on SAXS data measured from them.