Accelerating atomic fine structure determination with graph reinforcement learning

Journal article

Abstract Atomic data determined by analysis of observed atomic spectra are essential for plasma diagnostics. For each low-ionisation open d- and f-subshell atomic species, around 10 3 fine structure energy levels can be determined through years of analysis of 10 4 observable spectral lines. We propose a partial automation of this task by casting the analysis procedure as a Markov decision process and solving it by graph reinforcement learning using reward functions partly learned on historical human decisions. In our evaluations on existing spectral line lists and theoretical calculations for Co II, Nd II and Nd III, hundreds of energy levels were identified and determined in hours, agreeing with published values in 95% of cases for Co II and 54–87% for Nd II and Nd III. As the current efficiency in atomic fine structure determination struggles to meet growing atomic data demands, our artificial intelligence approach sets the stage for closing this gap.

Authors: Ding, M; Darvariu, V-A; Ryabtsev, AN; Hawes, N; Pickering, JC

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Nature Research
• Journal: Communications Physics
• Publication date: 2026-03-19
• Acceptance date: 2026-03-02
• DOI: 10.1038/s42005-026-02582-y
• EISSN: 2399-3650
• ISSN: 2399-3650
• Language: English
• Keywords: Automation; Observable; Graph; Task (project management); Spectral line; Markov decision process; Markov chain; Reinforcement learning