EDS Analysis for Petrology: A Probabilistic Framework With GPyEDS

Abstract Microtextural and chemical data from minerals in igneous rocks are critical for unpacking and developing understanding of processes in magmatic systems. Recent advancements leveraging unsupervised machine learning methods offer novel approaches for phase classification without requiring prior knowledge of phases or chemistry. The present work expands on these methods through a probabilistic framework to enable a high‐throughput approach, which can result in significant improvements in segmentation of noisy EDS data collected during SEM mapping of rock specimens. We demonstrate that linear matrix decomposition methods, such as principal component analysis (PCA), can be applied to examine the chemical zonation of mineral phases without prior knowledge of their chemistry. To calibrate phase chemistry, we introduce a Bayesian Markov Chain Monte Carlo (MCMC) approach that scales high spatial resolution energy dispersive spectroscopy (EDS) data to high precision and accuracy electron probe microanalysis (EPMA) profiles. We use this technique to derive high quality and reproducible chemical data sets across large spatial fields. The proposed method couples textural and chemical observations that allow petrologists to better interpret magmatic systems and understand crustal processes. Plain Language Summary The composition of crystals found in volcanic rocks contains crucial information about the processes that occur prior to large eruptions. These crystals can be examined by making chemical maps of small (cm‐scale) polished samples of the rocks and then probing them with an electron beam. X‐rays then released from the crystals have energy that is characteristic of the chemistry of the crystals. Technological advances allow for relatively rapid scanning of these sections, but the data are sometimes noisy and not straightforward to convert into sample maps that are useful for exploring sub‐volcanic processes. We have developed a new machine‐learning approach that automates the identification of different crystal types in the samples and provides a powerful method for converting noisy map data into images of crystals and their chemical zonation that can be used to understand the deep roots of volcanoes. Key Points New machine learning approach improves segmentation of chemical maps of thin sections Gaussian processes are added to previous autoencoder models to aid and improve clustering of mineral phases in compositional space A new method to quantify internal chemical zonation in minerals from noisy SEM map data and high‐precision/accuracy electron probe point data is also presented