Oxford Centre for High Energy Density Science (OxCHEDS)

Abstract:

<jats:title>Abstract</jats:title> <jats:p>Machine learning (ML) density functional approximations (DFAs) have seen a lot of interest in recent years, often being touted as the replacement for well-established non-empirical DFAs, which still dominate the field. Although highly accurate, ML-DFAs typically rely on large amounts of data, are computationally expensive, and fail to generalize beyond their training domain. In this work we show that differentiable DFT with Kohn-Sham (KS) regularization can be used to accurately capture the behaviour of known local density approximations (LDA) from small sets of synthetic data without using localized density information. At the same time our analysis shows a strong dependence of the learning on both the amount and type of data as well as on model initialization. By enabling accurate learning from sparse energy data, this approach paves the way towards the development of custom ML-DFAs trained directly on limited experimental or high-level quantum chemistry datasets.</jats:p>

Abstract:

A robust and flexible architecture capable of providing real-time analysis on diagnostic data is of crucial importance to physics experiments. In this paper, we present such an online framework, used in June 2025 as part of the HRMT-68 experiment, performed at the HiRadMat facility at CERN, using the Super Proton Synchrotron (SPS) beam line. HRMT-68 was a fixed-target laboratory astrophysics experiment aiming to identify plasma instabilities generated by a relativistic electron-positron beam during traversal of an argon plasma. This framework was essential for experimental data acquisition and analysis, and can be adapted for a broad range of similar-scale experiments with a variety of experimental diagnostics, even those without a standard direct network communication interface. The developed framework’s customizable design enabled us to rapidly observe and extract emergent features from a diverse range of diagnostic data. Simultaneously, its modularity allowed for a quick introduction of new diagnostic devices and the modification of our analysis as features of interest were identified. As a result, we were able to effectively diagnose equipment malfunction, and infer the beam’s response to varying bunch duration, beam intensity, and the plasma state without resorting to offline analysis, at which time adjustment or improvement would have been impossible. We present the features of this agile framework, whose codebase we have made publicly available so that it may be adapted for future experiments with minimal modification.