Potential Consequences of the Compositional Distribution of Trace Element Partitioning Experiments
Abstract
Our trace element partitioning study aims at creating a database for developing predictive models of trace element behaviour in natural geochemical systems. How this goal is pursued and what each investigator defines as a relevant database varies dramatically from study to study. Over the past several years, we have developed traceDs as part of the IEDA initiative. The purpose in developing traceDs was to provide a transparent and accessible resource of experimental partitioning data for the community. TraceDs now includes the experimental trace element partitioning data (>4900 experiments), including the experimental conditions, and the major and trace element composition of the phases. The existing database has specifically focused on following data characteristics:
Data quality - reflect reproducibility, precision, equilibration of phases, and completeness of analysis Data relevance - indicate how the data are relevant to a modeled system (appropriate T, P, X, etc.) Methodologies - for generating models Another important characteristic of the existing data that has yet to be fully examined is the data distribution in compositional space. Most models include regression parameters for composition as well as experimental conditions. This is because the mineral and melt component activities, the substitution of trace elements into specific sites, and the role of paired substitutions are influenced by composition. Existing data illustrates that the number of experimental determinations is different from element to element - even for a series of elements such as the REEs. For example, there are more determinations for the partitioning of the LREE between plagioclase and melt than there are for the HREE. In addition, there are more determinations for the REE in plagioclase than there are for the HFSE, and equally important, the composition of the population of determinations also vary with different elements. For instance, the population of liquids for the experiments where Zr partitioning was determined for pyroxene are different from those where Sc partitioning was determined. This means that any regression analysis must "see through" these differences in order to be applicable for a wide range of composition and to provide a set of modeled partition coefficients characterized by a systematic "smooth" pattern.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.V23I0163N
- Keywords:
-
- 0498 General or miscellaneous;
- BIOGEOSCIENCESDE: 1039 Alteration and weathering processes;
- GEOCHEMISTRYDE: 1065 Major and trace element geochemistry;
- GEOCHEMISTRYDE: 1914 Data mining;
- INFORMATICS