Accelerating accurate predictions of electronic response properties in extreme conditions

Electronic response properties of materials in extreme conditions play a crucial role in fusion energy research by facilitating diagnostics in experiments and parameterizing hydrodynamic simulations of fusion targets. While real-time time-dependent density functional theory (TDDFT) offers high-fidelity predictions for these properties, high computational costs preclude widespread use within the life cycle of experimental design and interpretation. More efficient models based on linear-response theory involve approximate electron-ion collision rates which do not always reproduce TDDFT response functions. We assess the prospects and limitations of Bayesian techniques for inferring accurate electron-ion collision rates from TDDFT predictions. We then investigate the extent to which the physically-motivated Bayesian framework allows interpolating between TDDFT calculations at different conditions. Throughout, we focus on predictions of the electron energy loss function (ELF), which is probed by inelastic x-ray scattering, relates to dynamic conductivity, and determines electronic stopping power. This work could enable efficient predictions of electronic response functions with first-principles accuracy.