IMPACT OF DATABASE CHARACTERISTIC ON TRACE ELEMENT PARTITIONING MODELS FOR CLINOPYROXENE

Quantitative models of trace element (TE) partitioning between minerals and melt can be useful in simulating how specific processes drive changes in the TE signature of an evolving magmatic system. The models designed to predict the behavior of TEs are most commonly built off regression analysis of the dependence of the partition coefficient (^Mineral-MeltD_i) between mineral and melt on pressure, temperature, composition, and substitution mechanisms by which TEs enter into the crystal lattice structure of the minerals. Most predictive models for TE behavior are calibrated on experimental determinations. Such data is available in public databases (https://www.earthchem.org/communities/experimental-petrology/). Predictive models based on regression analysis are dependent on the input data. However, the characteristics of the clinopyroxene/melt TE partitioning dataset is known to be heterogeneous in terms of number of experiments and number of elements analyzed for different TEs. The goal of this investigation is to evaluate the influence of dataset characteristics on the calculated dependencies with a focus on rare earth elements (REEs) and high field strength elements (HFSEs) on ^Cpx-MeltD_i. Clinopyroxene was selected because of the large number of experiments in the literature (>980) and it accommodates a number of TEs at measurable natural levels, but accommodate them differently. We have updated the traceDs database with new experimental data and used that newly compiled data to examine the correlation of D_REE and D_HFSE with reciprocal temperature and composition. Our results show that the calculated temperature dependence for D_REE and D_HFSE are not correlated to ionic radius or valence. However, they are strongly correlated with both D_Ti and Al^IV to a similar degree – in spite of differences in valence and presumed mechanism for charge balance. We also found that R² for those linear regressions were negatively correlated to the number of experiments which may be attributed to our experimental design as well as analytical ability to measure specific elements.