Algorithm structure regularisation for x-ray spectroscopy with Diamond Lightsource
PI: Tanner, Jared
Department: Mathematical Institute
Diamond Light Source is the UK national synchrotron facility which enables in-situ imaging for a wide range of applications such as visualising the life cycle of virus. Improved imaging is achievable through both increasing the coherent flux or through algorithmic advances. The former has inherent limits due to thresholds above which a sample will be damaged by a given radiation dose. For instance, imaging the cycling of a battery, for example, the beam interaction with the electrolyte can stop the charge and discharge process; or measuring the chemical state of an anti-cancer drug in a cell, the beam can de-form the cell affecting information on location of the drug or decompose the drug under the beam, making measurements of the chemical state impossible. Algorithmic advances are needed to increase the complexity of objects that can be measured while benefitting from the growing available flux. Including application domain structure enables such improvement. For instance, a finite number of chemical states in a system can be modelled as a low-rank structure which has recently been exploited to reduce the needed flux by 15-20%, which alternatively can allow increased chemical state complexity that can be imaged at a fixed flux. This project will extend the algorithms to: include additional structure such as spatial connectivity, allow for robustness to arbitrary but few outliers to the model, and will explore yet further structure for the most compelling use-cases of spectral-ptycho-tomography. Recent advances in optimisation algorithms that don't rely as heavily on the knowledge of manifold structure will allow this further structure to be included. The algorithm advances will be trialed on specific applications, but is also widely transferable and its success is expected to encourage wider uptake at Diamond as well as other related facilities.