46th Lunar and Planetary Science Conference (2015) 2737.pdf OLIVINE - MELT EQUILIBRIA IN LUNAR ULTRAMAFIC MAGMAS: INSIGHTS INTO MELT THERMODYNAMIC PROPERTIES Stephanie M. Brown1 and Timothy L. Grove1 , 1 Department of Earth, Atmospheric, and Planetary Science, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139 ([email protected], [email protected]) Experiments on lunar high-Ti melts revealed unusual variability of Fe-Mg partitioning between olivine Fe−Mg ). [1], and more recently [2], have and melt (KD shown that is it not clear what causes the non-ideal Fe−Mg behavior that leads to the observed KD variability. The lunar high-Ti experimental dataset, supplemented by new experiments on the intermediate-Ti ultramafic glasses, isolates the influence of silica activity and suggests that neither silica activity nor NBO/T (# of non-bridging oxygens per # of tetrahedrally coordiFe−Mg . Rather, nated cations) can sufficiently predict KD Fe−Mg KD variation is due to the complex solution behavior of network modifying Fe-Mg melt components. The relation between melt components, variable Fe−Mg f O2 , and the olivine-melt KD The observed deFe−Mg crease at low f O2 of the olivine KD in high-Ti lunar ultramafic glass experiments means that either the olivine is more forsteritic or the silicate melt has higher (FeO/MgO) [2]. This can be expressed as a reduction in the activity coefficient ratio of the Fe-Mg melt comFeO melt ponents, ( γγMgO ) . Because olivine is a nearly pure Fe-Mg solid solution with well known thermodynamic Fe−Mg properties (i.e. there is a wide range in KD for a given olivine composition), the culprit must be the melt. Thus there must be a change in the speciation of olivine forming silicate melt components under lowFe−Mg f O2 and high TiO2 conditions, resulting in low KD Fe−Mg γFeO melt and low ( γMgO ) corrected to values. When KD 1 atm is compared to the FeO melt ( γγMgO ) corrected to 1 atm FeO melt ratio decreases with decreasing f O2 . ) [3], the ( γγMgO γFeO melt The ( γMgO ) ratio is directly dependent on melt spe- ciation; if either component is bonding preferentially with other components in the melt, its activity will be reduced because the formation of such complexes stabilizes these components in the melt phase, reducing their availability for olivine formation. Thermodynamically, this occurs when there is a negative non-ideal free energy excess (-T∆Sexcess mix ) to form the melt complexes. The valence state of titanium is dependent on the f O2 conditions of the experiments: Ti4+ is likely being reduced to Ti3+ in the low-f O2 Fe capsule experiments [2]. Combining this with the above observation that there are f O2 -dependent changes in melt specia- tion strongly suggests that Fe and Ti3+ are complexing more efficiently together at low f O2 to reduce the Fe activity in the melt. An enticing solution is that Fe2+ , at the expense of Mg, is more efficiently complexing with Ti3+ than it did with Ti4+ in the melt, effectively increasing the amount of magnesium available for olivine. This type of reaction also predicts that the melt would become more olivine normative, which is observed experimentally by the expanding of the olivine primary phase volume (i.e., deepening of the multiple saturation point): the silicate-forming melt becomes more olivine normative as the Ti/(Fe+Mg) ratio of (Fe,Mg)-Ti melt components increases, thereby increasing the (FeO+MgO)/SiO2 ratio of the remaining liquid [2]. A pair of melt component reactions, similar to the one proposed by [2], can explain all the observed behavior. Reaction 1 is an oxidation-reduction reaction that describes the reduction of Ti4+ to Ti3+ by dissociation of a preexisting ilmenite (Fe,Mg)Ti4+ O3 melt component. The titanium in the ilmenite-like melt component is donated to the armalcolite-like melt component also present in the melt, forming a new more Ti-rich armalcolite melt component and releasing (Fe,Mg)O into the melt, causing the melt to become more olivine normative. In the limit of low f O2 , where x = 1, the new melt component becomes “anosovite” Ti4+ Ti32+ O5 : + + (Fe, Mg)Ti42 O5 + x(Fe, Mg)Ti4 O3 * ) + + 3 (Fe, Mg)1−x Ti42−x Ti2x O5 + 2x(Fe, Mg)O + xO2 . (1) Reaction 2 describes the Fe-Mg exchange in the Tirich modified armalcolite melt component, setting x = 0.5: + + Mg0.5 Ti41.5 Ti3 O5 + FeO * ) + + Fe0.5 Ti41.5 Ti3 O5 + MgO. (2) Reaction 2 proceeds to the right as the amount of titanium in the bulk composition increases causing preferential release of MgO into the melt, successfully preFe−Mg FeO melt dicting the decrease in ( γγMgO . In sum) and KD mary, this behavior suggests that the more Fe,Ti-rich modified armalcolite melt component is stable because it has a large excess non-ideal contribution to ∆Smix at low f O2 . 0.5 0.45 intermediate-Ti to high-Ti (yellow, orange, red, black) lunar ultramafic glass experiments 0.4 low-Ti (green, VLT) lunar ultramafic glass experiments 2737.pdf 0.45 Pressure corrected KDFe-Mg Predicted KDFe-Mg from Toplis 2005 46th Lunar 2 and Planetary Science Conference (2015) 0.4 0.35 0.35 0.3 0.25 0.25 0.2 0.15 0.3 0.2 0.2 0.25 0.3 0.35 0.4 0.45 0 10 20 30 40 50 60 70 FeO + MgO (wt%) 0.5 Measured KDFe-Mg Figure 1: The lunar experimental dataset cannot be Fe−Mg model predicted by the silica activity Fe-Mg KD of [3]. Data is from this work and [11, 6, 7, 12, 8, 9, 2] Figure 2: Fe-Mg olivine partitioning is not a quadratic function of FeO + MgO (wt%) or NBO/T. Grey dots and open circles are from [5], otherwise legend is the same as in Figure 1. Data is from this work and [11, 6, 7, 12, 8, 9, 2] Fe−Mg Predicting KD Many models exist for predictFe−Mg ing KD [4, 1, 5, 3, etc.]; however, these modFe−Mg els cannot account for the range in KD observed from lunar ultramafic experiments. For example, experiments on the low titanium lunar ultramafic glasses and mare basalts consistently record high olivine FeFe−Mg > 0.35 [1, 6, 7, 8, 9, 10] than compared Mg KD Fe−Mg to terrestrial rocks [3]. The cause of the KD variation must also be related to melt speciation. But unlike FeO melt the lower ( γγMgO ) of the intermediate - high TiO2 liquids, the low-titanium ultramafic lunar glasses exFeO melt hibit elevated ( γγMgO ) . Such high values typically correlate with terrestrial liquids that are 8-12 mol% more SiO2 rich [3] than the SiO2 content of the lowtitanium lunar ultramafic glasses. Therefore, the high Fe−Mg KD of these low SiO2 melts cannot be explained by a change in silica activity (Figure 1). Additionally, [5] suggested that the melt polymerization proxy FeO + MgO (wt%), or the equivalent Fe−Mg NBO/T, of an ultramafic liquid controlled KD . Fe−Mg They found that the KD reached a maximum at NBO/T = 2, or FeO + MgO ≈ 36 wt%. When we superimpose the lunar dataset on top of the [5] data (Figure 2), we find that the lunar data crosscuts this Fe−Mg correlation. The true cause of KD variation in ultramafic glasses is more complex than silica activity or NBO/T. As an alternative model, we are working on quantifying the role of network modifiers and network formFe−Mg ers for predicting KD as a function of temperature, pressure, f O2 , and composition. References [1] J. Longhi, et al. (1978) Geochimica et Cosmochimica Acta 42(10):1545 doi. [2] M. J. Krawczynski, et al. (2012) Geochimica et Cosmochimica Acta 79:1 doi. [3] M. J. Toplis (2005) Contributions to Mineralogy and Petrology 149(1):22 doi. [4] P. Roeder, et al. (1970) Contributions to Mineralogy and Petrology 29:275. [5] I. Kushiro, et al. (1998) Geophysical Research Letters 25(13):2337 doi. [6] H.-K. Chen, et al. (1982) Journal of Geophysical Research 87(S01):A171 doi. [7] H.-K. Chen, et al. (1983) Journal of Geophysical Research 88(S01):B335 doi. [8] L. Elkins, et al. (2000) Geochimica et Cosmochimica Acta 64(13):2339 doi. [9] L. T. Elkins-Tanton, et al. (2003) Meteoritics & Planetary Science 38(4):515 doi. [10] J. A. Barr, et al. (2013) Geochimica et Cosmochimica Acta 106:216 doi. [11] J. Delano (1980) Proc Lunar Planet Sci Conf 11:251. [12] T. Wagner, et al. (1997) Geochimica et Cosmochimica Acta 61(6):1315 doi.