Automatic Bulk Composition Analysis of Lunar Basalts: Novel Big-Data Algorithm for Energy-Dispersive X‑ray Spectroscopy

The bulk composition of lunar basaltic meteorites and
clasts provides
crucial information for understanding their petrogenesis and thus
lunar thermal evolution. Meanwhile, the basalt type of Chang’E-5
based on the bulk TiO2 contents remains debatable. Modal
recombination based on mineral volume fraction, densities, and average
compositions is currently the most popular method to determine the
bulk composition of lunar samples. Yet, the latter two parameters
can be biased markedly by ubiquitous compositional variations in pyroxene,
olivine, and plagioclase. To rectify these issues and provide more
accurate classifications, this study devises a novel big-data algorithm
that analyzes maps of energy-dispersive X-ray spectroscopy (EDS) data
of lunar basalts. The algorithm starts by labeling each point through
a newly devised mineral classifier, then uses the mean of all points
per mineral to represent average composition, and finally recalculates
the true density per mineral to replace standard density. The accuracy
of this mineral classifier is demonstrated by tests on a database
of lunar minerals. The accuracy and precision of EDS mapping were
verified by test analysis on certified reference minerals. Measurements
on a lunar meteorite sample with a known composition, NWA 4734, are
comparable to those measured using inductively coupled plasma optical
emission spectrometry and confirm the reliability of the bulk composition
algorithm. To demonstrate its utility for comprehensive understanding
of petrographic features, the high-efficiency algorithm was applied
to Chang’E-5 basalts. The results reveal that these basalts
are characterized by low-Ti and low-Mg features, thus distinct from
previous Apollo and Luna samples.