Tracking Subsurface Active Weathering Processes in Serpentinite

Abstract

We conducted a novel study to capture the on‐going advancement of mineral weathering within a serpentinite formation by using an integrated approach of multi‐scale quantitative rock magnetic analyses and nano‐resolution geochemical imaging analyses. We studied a suite of rock samples from the Coast Range Ophiolite Microbial Observatory (CROMO) in California to conduct rock magnetic analyses enabling us to determine character of Fe‐bearing minerals and to predict locations of reaction boundaries among various stages of weathering. QEMSCAN® and other electron micro‐imagery analyses highlighted microstructural changes in amorphous minerals, and possible changes in porosity and coincides with the iron‐enrichment region. This iron enrichment indicates initiation of iron (‐oxides) nucleation, resulting in extremely fine gain magnetite formation. This is a newly documented mode of magnetite production in serpentinites and enhances the application of magnetite abundance as a proxy for the degree and extent of water‐rock interaction in mantle peridotite and serpentinite.

Interrogation of Ecotoxic Elements Distribution in Slag and Precipitated Calcite through a Machine Learning‐Based Approach Aided by Mass Spectrometry

“…Additional components increase the computational time, and may reduce the effectiveness of the clustering performance due to the curse of dimensionality as reported elsewhere. [28,40] The application of the HDBSCAN method on the dimensionally reduced data depicts that the studied samples exhibit a variety of complexity and that the provided workflow can analyze samples containing several phases that have characteristic compositions. Samples A and C containing 7 and 5 clusters respectively, including up to ≈20% outliers, while sample C shows 14 clusters with ≈35% outliers, as depicted in Figure 5.…”

“…Dimensionality Reduction and Clustering: The obtained EDS map data were exported as .raw and .rpl files from Aztec 6.0 software and were imported and separately processed in Jupyter Notebook using HyperSpy, SciPy, pandas, Matplotlib, Seaborn, and NumPy dependencies. [51][52][53][54][55][56] The current workflow builds on the work of Tominaga and coworkers [40] for processing of the EDS data and it is described as follows. The obtained BSE images were used to visualize the samples and to identify pore space such that it could be masked from subsequent dimensionality reduction and clustering analysis.…”

“…Previous work demonstrated that ≈700 counts per pixel was sufficient to facilitate data analysis. [ 40 ] However, there is no direct entry for the number of counts per pixel in SEM‐EDS control software. The counts per pixel can be increased through increasing the dwell time and/or the number of acquired frames.…”