Determine the Elemental Composition of Minerals From Complex Solid‐Solution Series by Raman Spectroscopy: Implications for Mars Exploration Missions

ABSTRACT Garnets are minerals offering valuable insights into geological and planetary processes, in both terrestrial and extraterrestrial contexts. Their presence in metamorphic rocks, as well as in meteorites, makes them key indicators of the conditions under which they formed, such as pressure, temperature, and chemical environment. This information is valuable for understanding the history and evolution of planetary bodies, including Earth and Mars. Developing automatic classification models for garnets using Raman spectroscopy has a potential application for planetary exploration, particularly for the Raman Laser Spectrometer (RLS) on the Rosalind Franklin rover of the upcoming ExoMars mission. In this work, we used a dataset combining spectra from ADAMM and RRUFF databases, ensuring a well‐distributed representation of the two main garnet groups: pyralspites and ugrandites. The spectra from ADAMM were obtained using the RLS SIM, a laboratory version of RLS, while RRUFF data helped increase the number of samples in the dataset. After standardizing all data, we designed a two‐step classification model: a top‐level model to classify into the two main groups and two specific models to classify the garnet type within each group. We tested multiple machine learning algorithms, including support vector machine (SVM), k‐nearest neighbors (KNN), artificial neural networks (ANN), decision tree, and naive Bayes. The top‐level classification reached 100% accuracy in testing, while the final combined model achieved 80.42% accuracy. These results demonstrate that machine learning and Raman spectroscopy can effectively classify garnets, providing a valuable tool for planetary missions and mineralogical studies.