Russell O. Colson

Partitioning in REE-saturating minerals: Theory, experiment, and modelling of whitlockite, apatite, and evolution of lunar residual magmas

We present compositions, including REEs determined by ion microprobe, of apatite and whitlockite in lunar rock assemblages rich in incompatible trace elements. Total concentrations of REE oxides in whitlockites range from 9–13 wt%, and those in apatites range from 0.15 to 1 wt%. Ratios of REE concentrations in whitlockite to those in coexisting apatite range from ~ 10 to 60. The distribution of Mg and Fe between apatite and whitlockite is correlated to that of coexisting mafic silicates: Magnesium is strongly preferred by whitlockite, and Fe is preferred by apatite. Incorporation of REEs in whitlockite is dominated by the coupled substitution of 2REE3+ in Ca(B) sites + vacancy in Ca(IIA) for 2Ca+2 in Ca(B) sites and (Ca2+,Na+) in Ca(IIA). Other substitutions account for only a small portion of the REEs in whitlockite over the observed concentration range; thus, REE concentrations become partially saturated as the primary substitution approaches its stoichiometric limit of two REEs per fifty-six oxygens, leading to reduced whitlockite/melt distribution coefficients e.g., decreasing from twenty-five to ten for Nd. The REE concentrations of lunar residual melts are not depleted by whitlockite crystallization in assemblages consisting mainly of other minerals in typical proportions. Distribution coefficients for the REEs in lunar apatite appear to be low and variable e.g., ~0.2–0.8 for Nd. Variations in the modal ratio of whitlockite to apatite, specifically the abundance of whitlockite, lead to a range of REE concentrations in the phosphates and variations in the magnitude of REE concentration ratios between whitlockite and apatite. If apatite crystallizes before whitlockite or in the absence of whitlockite, as textures in several samples indicate, then apatite zoned in REEs and apatite crystals of different REE concentrations may occur in a given sample, provided there is some amount of fractional crystallization and apatite does not later equilibrate. This may occur because, in the absence of whitlockite in the crystallizing assemblage, the REEs are highly incompatible relative to the crystalline assemblage, so REE concentrations in lunar residual melts increase strongly during small increments of late-stage crystallization. Once whitlockite begins to crystallize, bulk distribution coefficients for the REEs, although still < 1, are only mildly incompatible, so the change in REE concentrations of residual melts with further crystallization is small, consistent with the lack of REE zoning in whitlockite. The REE concentrations in lunar whitlockites are modelled as resulting mainly from equilibrium crystallization of the assemblages in which they occur; metasomatism or other secondary metamorphic processes are not indicated. Local melt-pocket equilibrium at advanced stages of crystallization may lead to variable REE concentrations and variable whitlockite /apatite concentration ratios within the same sample. Parent melts with extremely high REE concentrations are not required in order to crystallize REE-rich lunar whitlockite if modal proportions of whitlockite are low.

Reinterpretation of reduction potential measurements done by linear sweep voltammetry in silicate melts

Abstract We have determined experimentally the equilibrium concentrations of Ni between silicate melt and Pt as a function of oxygen fugacity. The results demonstrate that metallic species derived in linear sweep voltammetry experiments in silicate melts are diffusing into Pt electrodes and not into the melt, as was concluded by previous studies. This requires reinterpretation of previous linear sweep voltammetry results and recalculation and correction of reported reduction potentials. In this paper, we report these corrected reduction potentials. We also report activity coefficients for Ni in synthetic basalt and diopsidic melts and for Co in diopsidic melt.

Electrochemistry of cations in diopsidic melt - Determining diffusion rates and redox potentials from voltammetric curves

We have used Linear Sweep Voltammetry to measure properties of selected ions in diopsidic melt. In general, the redox reactions studied are complicated by such factors as slow reaction kinetics, monomerizations, and dimerizations, or presence of multi-step reductions that overlap on voltammetric curves. Criteria for recognizing these complications are gleaned from the literature and their application to reductions in high temperature silicate melts is discussed. Criteria that establish whether the reduction potentials obtained from the curves represent equilibrium conditions are also discussed. Diffusion coefficients were measured in diopsidic melt for cations of Eu, Mn, Cr and In. We observe that diffusion rates decrease as the size of the diffusing cation increases. Diffusional activation energies are similar to those reported for other silicate melt compositions. Enthalpies and entropies of reduction were determined for the cations V(V), Cr3+, Mn2+, Mn3+, Fe2+, Cu2+, Mo(VI), Sn(IV) and Eu3+. The free energies can be used to compute relative proportions of multivalent cations in various valence states as a function of temperature and fO2. Reduction potentials also provide a means for studying the structural state of cations in the melt. Free energies of reduction in melts of different compositions cannot be modeled solely by changes in oxide ion activity, but must include contributions from changes in the activities of the cations as well.

Diffusion and activity of NiO in CaO-Mg0-Al203-Si02 melts considering effects of αo2− and γNi2+

Variations in activity coefficients for oxides in silicate melts are complex functions of silicate melt composition. For example, the activity coefficient (y) for NiO shows a minimum when plotted against a parameter reflecting the degree of polymerization (or basicity or fraction of bridging oxygens) (e.g., Pretorius and Muan, 1992). In this paper, we propose that this complexity occurs in part because NiO is not an actual species in silicate melts. Variations in y can be better understood and predicted if the activity of NiO is treated as the product of activities of Ni2+ and 02− ions. Using voltammetric methods, we have measured independent activity coefficients for Ni2+ and O- for compositions between diopsidic and anorthitic melt and have found that variations in these values are more easily understood in terms of the melt composition and structure and permit qualitative variations in γNiOo (activity coefficient of NiO relative to the free energy of formation reported in Robie and Waldbaum, 1968) to be predicted in compositions other than those studied. We suggest that a similar consideration of ionic behavior might improve our understanding of activity coefficients for other oxides in silicate melts as well. In addition, we report diffusion rates for Ni and free energies for the reaction Ni2+ + 02- ⇌ Ni0 + 1202, for melt compositions along the compositional joins CaMgSi206-CaAl2Si208 and CaMgSi206-MgAl2Si 2O8.