Magnetic spectroscopy signatures in hard-x-ray photoemission of photocatalytic materials for green hydrogen production

The idealist’s dream of an abundant and clean energy fuel, has long motivated photocatalytic materials research for the production of green hydrogen. While the underlying mechanism for photocatalysis has yet to be definitively understood, maximized light harvesting properties, optimized charge transfer, long lifetimes of intermediate species for redox reactions at the solid/water interface and, in a variety of materials systems, the presence of delocalized spin-states, have been inferred to improve photocatalytic efficiency. In this work we explore the predictive strategies of ensemble machine learning models for inverse modeling of experimental data and density functional theory calculations for inverse materials design. Spectroscopic signatures, in x-ray absorption, photoemission and x-ray emission, have long informed the underlying nature of electronic and magnetic structure. To directly reveal the geometric, electronic and magnetic structure of multifunctional photocatalytic candidate material systems, a formalism for the variation in the photoemission angular distribution in the magnetic linear dichroism geometry, when the excitation photon energy traverses through the crystallographic Bragg condition, is presented. Large synthetic datasets have been generated and served as training datasets for a deep learning model which is tested against the ground truth - experimental data.