Surface destabilization and laboratory-induced non-stoichiometry in San Carlos olivine

Abstract

Annealing experiments on natural olivine (Mg_1-x Fe _x )₂SiO₄ (with x≈0.11) crystals (San Carlos, Arizona, spinel-lherzolite context) have been performed between T=1,100° C and 1,500° C for oxygen partial pressures pO₂=10⁻³ to 10⁻¹³ bar and times of 1 to 140 h in CO/CO₂ or H₂/H₂O gas mixtures. Even specimens annealed within the T-pO₂ theoretical stability field (TSF) calculated for stoichiometric olivine (Nitsan 1974) show systematic alterations developed within the first few microns of the surface of the crystals. Pyroxene crystals or melt form on the original olivine surface even at T=1,100° C, with preference of pyroxene when T<1,350° C and SiO₂-rich glass if T>1,350° C. This glass (rhyolite-like) can concentrate calcium from the starting olivine, and aluminum when Cr-Al spinels are present as inclusions. These observations are in contradiction with the TSF. They are obviously due to the presence of platinum used as a container of our samples, even if the contact between olivine and platinum is very weak. Rapid surficial diffusion of iron toward platinum (or via the gas phase) induces a Fe-depleted surface. According to the TSF, this more forsteritic surface should have a wider pO₂ range of stability. This is not the case, just because this situation is largely out of equilibrium. This iron loss induces a departure from cationic stoichiometry: (Mg, Fe)_2(1−δ), SiO₄ with δ small and positive. We extended the model that Nakamura and Schmalzried (1983) (N.S.) developed for fayalite (x=1) to our natural olivine composition, under the assumption that the majority defects are magnesium vacancies, Fe³⁺ occupying octahedral and tetrahedral sites, and the more complex neutral defect corresponding to Coulombic attraction between neighboring Fe³⁺ ions. We have recalculated the olivine stability field in pO₂ vs. δ space at T=1,300° C using this model for x≈0.1 (its extreme limit of validity) and conclude that olivine is stable only in a very narrow range in pO₂ which depends on δ. The calculation shows also that when olivine has nearly cationic stoichiometry (δ=0) as we believe for our starting material, the pO₂ range of stability is narrower than indicated by the TSF. In particular, it explains why Fe precipitates from the olivine (δ=0) (in absence of any other precipitation of SiO₂-rich phases) between 10⁻¹¹ and 10⁻¹³ bar at 1,300° C; this was not predicted by the TSF. Magnetite or iron precipitates also coexist with SiO₂-rich exsolutions or pyroxene when pO₂ is close to the upper or lower boundaries of the TSF, respectively. The N.S. model may have important implications for the interpretation of the existence of partial melting and/or the low-viscosity/low velocity zone in the upper mantle.