Trace element partitioning coefficients between silicate melts and orthopyroxene: Parameterizations of D variations

Nernst partition coefficient data between orthopyroxene and silicate melts show variability for most trace elements which cannot be neglected when designing petrogenetic trace element models. Orthopyroxene/LiquidD data were examined for variation against pressure, temperature; orthopyroxene (Opx) wollastonite content (Wo = 100 × Ca/Ca + Mg + Fe), mg# (Mg/Mg + Fetotal), and tetrahedral Al (Aliv) content; and wt.% melt SiO2, Al2O3, FeO, MgO, H2O and MgO# (MgO/MgO + FeOtotal). For most elements, Opx/LiquidD values correlate best against melt LnMgO (wt.%) content, Opx mg#, and melt SiO2, with most Opx/LiquidD values increasing as LnMgO and Opx mg# decrease and as SiO2 increases. Variations of Opx/LiquidD with other factors show more diffuse patterns. The effect of pressure is important for Opx/LiquidDNa and DCs. Log fO2 may be important in constraining Opx/LiquidDV. Multiple regression analysis of combinations of these input variables generally yields better solutions, although in most cases the simple regressions against LnMgO or Opx mg# capture most of the variance. Multi-element Opx/LiquidD profiles calculated from Opx/LiquidD regressions against melt MgO and SiO2 content, Opx mg#, and multiple regression analysis of melt and mineral compositional variables, are compared to results of the Lattice Strain Model; generating parameterizations of D0 (the strain compensated partition coefficient), E3+M (Young's Modulus), and r0 (the size of the M site), which can be used to calculate Opx/LiquidD values for the rare-earth elements (REE) and Y. These parameterizations of Opx/LiquidD variations yield results comparable to experimental or natural partitioning data, and provide better constrained solutions to trace element models applied to magma genesis and differentiation. Regressions were also performed to generate a number of orthopyroxene-saturation geothermometers, which are accurate to ca. 50°K. Variations of the Fe–Mg exchange coefficient were also parameterized.