Evaluating the serpentinization degree of Martian analogues through the RLS ExoMars simulator: comparison between univariate and multivariate semi-quantification methods

As part of the ESA ExoMars rover payload, the Raman Laser Spectrometer (RLS) is scheduled to deploy on Mars in 2021. Together with MicrOmega (NIR) and MOMA (GC-MS), the instrument will analyze Martian subsoil samples to determine their mineralogical composition and investigate the potential presence of biomarkers. Beside the challenges associated with the development of the first Raman spectrometer to be validated for planetary exploration (together with Mars2020/ Sherloc and Supercam systems), to optimize the scientific outcome of RLS spectra gathered on Mars has a crucial importance in the fulfillment of the mission aims. Thus, the RLS team is developing tailored chemometric tools that, taking into account technical specifications and the operational mode of the RLS system, could be used to semi-quantify the main phases composing Martian samples.

Considering that 1) the serpentinization of olivine-bearing rocks on Earth plays a key role in the proliferation of microorganisms and in the preservation of biomarkers, and 2) remote sensing systems (e.g. CRISM) detected vast serpentine-bearing deposits on Mars, the present work seek to provide the chemometric tools necessary to correctly define the serpentinization degree of Martian rock samples through the interpretation of RLS data.

To do so, olivine and serpentine certified materials were mixed at different concentration ratios and 39 spot of analysis por sample were analyzed by means of the RLS ExoMars Simulator. Data sets were then analyzed using uni-variate (intensity ratio between olivine and serpentine main peaks) and multi-variate (a combination of principal component analysis and artificial neural networks PCA-ANN) methods.

The two uni-variate and multi-variate semi-quantification models were finally applied to the study of serpentinized rocks sampled from the Leka Ophiolite Complex (LOC), being those part of the Planetary Terrestrial Analogue Library (PTAL) collection. RLS-based semi-quantification results were finally compared to those obtained from the use of a state-of-the-art laboratory X-ray diffractometer (XRD).

Our study suggest that the uni-variate method provide excellent results when the analyzed rocks are mainly composed of olivine and serpentine. However, the estimation reliability decreases when the mineralogical heterogeneity of the sample increases (Raman features of additional mineral phase may overlap the selected olivine and serpentine peaks). In these cases, the multi-variate method based on the combination of PCA and ANN helps to more accurate estimate the serpentinization degree of the terrestrial analogs.

In conclusion, the preliminary results summarized in this work indicates that the study of terrestrial analogs is of crucial importance to test and validate RLS-dedicated semi-quantification models. In a broader perspective, it also highlights the importance of developing multiple chemometric tools, since the effectiveness of each of them varies according to mineralogical complexity of the sample under study.