QUANTIFYING AND VISUALIZING EARTH’S MINERAL CHEMISTRY THROUGH GEOLOGIC TIME

Earth’s mineralogy is governed not only by bulk composition, differentiation processes, and secondary geochemical processes, but also by the way its constituent elements parse themselves into mineral species (Hazen et al. 2015; Christy 2015). To gain insight into how elements associate in minerals, we used a large mineral database (mindat.org, as of 12/17/2015) to conduct a statistical analysis of the number of mineral species containing both elements X and Y for every possible X-Y pair (and X, Y and Z for every possible X-Y-Z triplet) of all 72 mineral-forming elements. The frequency with which an element pair or triplet occurred together in the nominal formula of a mineral was compared with the expected frequency if elements were distributed among mineral species randomly. This analysis was conducted for both a list of Hadean minerals that were likely present when Earth’s crust first formed (Hazen 2013), and for the full list of 5027 presently known minerals. We employ a new 3-dimensional data visualization technique, 3-D Klee diagrams, as well as force network diagrams to examine detailed patterns in the way elements combine to form Earth’s mineralogical diversity.

Results reveal that the strength of these correlations and anti-correlations ranged from ~1 to 10^-304 (as measured by their p-value in a χ² test for variable dependence for element pairs, and a Cochran-Mantel-Haenszel test for element triplets). Comparing element correlations between Hadean vs. Cenozoic minerals reveals that Si-O is the most important correlation on the early Earth, whereas O-H is the most important on present day Earth, providing valuable clues as to how post-Hadean processes such as plate tectonics and biology (Hazen et al. 2008) can dramatically affect a planet’s mineralogy. We also find that triplet correlations of the form X-O-H can be used to divide mineral-forming elements into hydrophilic vs. hydrophobic elements, and that force network diagrams indicate a grouping of elements into those associated with silicates vs. sulfides. Applying new statistical and visualization tools to large geological and biological databases promises to enable many more abductive, data-driven discoveries about how Earth’s geosphere and biosphere co-evolved.