What We Have Learned About the Existing Trace Element Partitioning data During the Population Phase of traceDs

Abstract:

The database traceDs is designed to provide a transparent and accessible resource of experimental partitioning data. It now includes ~ 90% of all the experimental trace element partitioning data (~4000 experiments) produced over the past 45 years, and is accessible through a web based interface (using the portal lepr.ofm-research.org).

We set a minimum standard for inclusion, with the threshold criteria being the inclusion of:

• Experimental conditions (temperature, pressure, device, container, time, etc.)
• Major element composition of the phases
• Trace element analyses of the phases

Data sources that did not report these minimum components were not included. The rationale for not including such data is that the degree of equilibration is unknown, and more important, no rigorous approach to modeling the behavior of trace elements is possible without knowledge of composition of the phases, and the temperature and pressure of formation/equilibration.

The data are stored using a schema derived from that of the Library of Experimental Phase Relations (LEPR), modified to account for additional metadata, and restructured to permit multiple analytical entries for various element/technique/standard combinations.

In the process of populating the database, we have learned a number of things about the existing published experimental partitioning data. Most important are:

• ~ 20% of the papers do not satisfy one or more of the threshold criteria.
• The standard format for presenting data is the average. This was developed as the standard during the time where there were space constraints for publication in spite of fact that all the information can now be published as electronic supplements. 
• The uncertainties that are published with the compositional data are often not adequately explained (e.g. 1 or 2 sigma, standard deviation of the average, etc.). 

We propose a new set of publication standards for experimental data that include the minimum criteria described above, the publication of all analyses with error based on peak count rates and background, plus information on the structural state of the mineral (e.g. orthopyroxene vs. pigeonite).