VISUALIZATION AND ANALYSIS OF METSILDB AND SULFSILDB : TWO NEW DATABASES FOR EXPERIMENTAL METAL-SILICATE AND SULFIDE-SILICATE EQUILIBRIA

A large number of experimental studies have investigated the distribution of elements between metal, silicate and sulfide liquids to better understand core formation processes and ore deposits. Here, we present two new databases, MetSilDB and SulfSilDB, which include 2880 literature data for experimental metal-silicate and sulfide-silicate equilibria, respectively, compiled from 144 peer-reviewed publications. We survey experimental conditions for each set of experiments, including ranges of pressure, temperature, oxygen fugacity, sulfur fugacity and chemical composition. Periodic tables showing the affinity of elements with metals (siderophile elements), sulfides (chalcophile) and silicates (lithophile) are constructed. These different types of visualization methods enable highlighting gaps in the experimental data. We also performed network analysis to investigate relationships between experiments and the elements that studies have been focused on. Specifically, these networks are bipartite with nodes being either experiments or elements which partition between metal and silicate or sulfide and silicate, as determined in each experiment. We found a bipolarity in the networks with the highly siderophile and highly lithophile elements located on each side of the network, which can be a source of bias in experimental setups. These results additionally show that the effects of light elements (Si, O, S, C) present in the metal on the elemental partitioning have mostly been investigated for lithophile elements, and needs to be addressed for other elements. Our study shows how machine learning algorithms and visualization techniques enable locating gaps and biases in the data, which will help better target future experimental studies.