2020/08/11 by Claire‐Alice Hébert, Claire-Alice Hébert, Hébert, Claire-Alice +9 · 1 citation
Engineering · Mathematics · #Algorithm #Applications (stat.AP) #Calibration #Computer science #FOS: Computer and information sciences #Gaussian #Gaussian process #Laser-induced breakdown spectroscopy #Laser-induced spectroscopy and plasma #Latin hypercube sampling #Machine learning #Mars Exploration Program #Martian #Martian surface #Materials science #Mathematics #Matrix (chemical analysis) #Monte Carlo method #Physics #Spectroscopy #Statistics #stat.AP
paper · pdf · doi:10.48550/arxiv.2008.04982
published in arXiv (Cornell University) (Cornell University) · 10 pages, 5 figures, special issue
arxiv created 2020/08/11 · openalex publication_date 2020/08/11 · arxiv updated 2020/08/13 · openalex created_date 2022/07/26 · openalex updated_date 2026/08/05
The Mars Curiosity rover carries an instrument, ChemCam, designed to measure\nthe composition of surface rocks and soil using laser-induced breakdown\nspectroscopy (LIBS). The measured spectra from this instrument must be analyzed\nto identify the component elements in the target sample, as well as their\nrelative proportions. This process, which we call disaggregation, is\ncomplicated by so-called matrix effects, which describe nonlinear changes in\nthe relative heights of emission lines as an unknown function of composition\ndue to atomic interactions within the LIBS plasma. In this work we explore the\nuse of the plasma physics code ATOMIC, developed at Los Alamos National\nLaboratory, for the disaggregation task. ATOMIC has recently been used to model\nLIBS spectra and can robustly reproduce matrix effects from first principles.\nThe ability of ATOMIC to predict LIBS spectra presents an exciting opportunity\nto perform disaggregation in a manner not yet tried in the LIBS community,\nnamely via Bayesian model calibration. However, using it directly to solve our\ninverse problem is computationally intractable due to the large parameter space\nand the computation time required to produce a single output. Therefore we also\nexplore the use of emulators as a fast solution for this analysis. We discuss a\nproof of concept Gaussian process emulator for disaggregating two-element\ncompounds of sodium and copper. The training and test datasets were simulated\nwith ATOMIC using a Latin hypercube design. After testing the performance of\nthe emulator, we successfully recover the composition of 25 test spectra with\nBayesian model calibration.\n