Assessing the feasibility of laser induced breakdown spectroscopy for detecting nitrogen in martian surface sediments

Abstract

Despite the detection of fixed nitrogen in meteorites and directly on Mars' surface, the abundance and distribution of nitrogen sequestered in the martian crust remains unknown. Given that nitrogen is a bioessential element that is required for the synthesis of amino acids, nucleic acids, and other organic molecules vital for life, this gap in knowledge is one of the most important challenges in constraining martian habitability. Laser-induced breakdown spectroscopy (LIBS) has the capability to detect N in natural rock samples and is available as a stand-off survey instrument on multiple currently active Mars rovers, creating an immediate opportunity to map the stratigraphic distribution of N within diverse depositional settings. However, little has been published regarding the detection of N with LIBS.

To lay a foundation for N detection on Mars using LIBS, we synthesized a comprehensive suite of samples with variable amounts of nitrogen (as nitrate or ammonium) in either a Mars regolith simulant or a clay matrix. We present baseline spectra of N emission in Mars-relevant matrices and identify spectral interferences. Our results indicate that 17 diagnostic N emission lines are reliably detectable from mineral-bound N against a basaltic background, but only four lines exhibit sufficient sensitivity to be detected across a range of N concentrations and within all tested matrices. To elucidate optimized strategies for quantification, we present an iterative series of PLS models. We find that prediction accuracy is improved by restricting the compositional range of the training set, normalizing the data, subtracting baseline continuum emission, and simultaneously modeling the emission behaviour of multiple diagnostic N lines at once. We observe that the prediction uncertainty increases (worsens) from 8.4% to 29.9% if models are used to predict N in samples with a dissimilar matrix than those used during training, suggesting poor generalizability outside the training range. Consequently, future work should focus on developing a larger, more diverse training set that encompasses the range of N concentrations and phases expected to be encountered on Mars, which may be used to train generalizable models. Overall, this work demonstrates that LIBS is a promising tool for determining the abundance of N sequestered in martian surface materials and lays a foundation for future development.