Аннотация:Thanks to its versatility, laser-induced breakdown spectroscopy (LIBS) can be used to analyse unique specimens, such as Mars surface, ocean floor, molten metal and radioactive waste. In many cases like these, there are no appropriate certified reference materials. Boltzmann plot is a well-known approach for calibration-free (CF) LIBS; it requires optically thin plasma under local thermodynamic equilibrium. It is frequently impossible to strictly fulfil these requirements; extreme conditions like open space or deep sea can further complicate preliminary checking of these conditions. Stationary laser-induced plasma can be modelled, synthesising a spectrum for a given elemental composition and other conditions [1]. Using multiparametric optimization, it’s possible to fit the model to the experimental data, with relative elemental concentrations and plasma conditions serving as variables. A gradient-free optimisation method has been chosen for reasons related to implementation. The model was tested on steel samples and aluminium alloys. Various shapes of the loss function have been considered, making it possible to account for analytical lines of very different intensity and highlight different detail of the spectra. The stability of the solution has been verified by repeated runs with random initial parameters. It has been shown that the homogeneous plasma model is enough to predict several elements with lines in narrow spectral regions, while a multi-zone model with different temperatures and electron densities provides the best results for the full spectra, although in the latter case the accuracy is lower. Finally several Curiosity ChemCam spectra were fitted by our algorithm to predict the composition of the studied object. The reported study was funded by Grant of the President of the Russian Federation (No. MK-5513.2021.6). References [1] S. M. Zaytsev, A. M. Popov, T. A. Labutin, Spectrochim Acta, Part A. 158 (2019) 105632.