Bayesian analysis of optical emission spectroscopy measurements coupled to a collisional-radiative model: application to low-pressure misty plasmas

Abstract Optical emission spectroscopy (OES) is widely used for non-invasive plasma characterisation. OES data is often coupled with a collisional-radiative model (CRM) to get insight on the fundamental plasma properties. A CRM comes with many adjustable parameters, whose value must provide the best agreement between the measured and simulated spectra. In this work, a CRM is used to analyse time-resolved OES measurements of low-pressure ‘misty’ plasmas, in this case Ar plasmas in which liquid droplets are injected as aerosols in a pulsed fashion. It is shown that series of liquid pulses can lead to significant quenching of Ar(1 s ), even though a single injection does not. The search for the best-matching spectrum is framed in terms of probability distributions, taking full advantage of the principles of Bayesian inference. Some of the tools provided by Bayesian analysis are illustrated in the specific case of OES-CRM analysis of Ar 2 p -1 s transitions. Both experimental and modelling uncertainties are considered in a rigorous mathematical framework, and results are expressed as probability distributions instead of single values. The question of model selection is addressed through the estimation of the Bayes factor, applied as a quantitative filter for superfluous adjustable parameters.