Experimental Determination of the K2CO3 Fusion Curve to 3 GPa and Constraints on the Liquid Equation of State

An equation of state (P-V-T relation) for carbonate liquids is of considerable geological interest despite the fact that carbonate magmas are volumetrically sparse compared to other igneous rocks. They are efficient agents of mantle metasomatism because of their high mobility and high concentrations of incompatible trace elements. They are also frequently associated with alkaline silicate magmas and may play a role in their genesis. Information on the density of carbonate liquids to high pressure is needed for an accurate assessment of their buoyancy in the deep mantle, as well as their thermodynamic stability to high pressure. Although the one-bar density and compressibility of alkaline carbonate liquids are well known (Zhu et al., 1991; Liu and Lange, 2003), the pressure dependence of the liquid compressibility (K?¡ê? = dK/dP) is not well constrained. For example, Dobson et al. (1996) report a density for K2CO3 liquid at 4 GPa and 1500\??C, obtained by the falling sphere method, of 3.10 g/cm3. Liu and Lange (2003) used the 3rd-order Birch-Murnaghan equation of state to show that this liquid density value is consistent with a liquid K?¡ê? value of 2.5, which in turn will lead to a strongly negative dT/dP slope in the K2CO3 fusion curve at pressures > 1.0 GPa. Thus the implications of this 4 GPa liquid density measurement are that alkaline carbonate liquids are stable to remarkably low temperatures at high pressure and that they are expected to become more dense than silicate melts (and possibly mantle minerals) at deep mantle depth, which has profound implications for the migration of carbonate melts in the mantle. Owing to the significance of these inferences, it is important to independently verify the density of high-pressure carbonate liquids. In this study, we provide this test by experimentally determining the fusion curve of K2CO3 to 3 GPa and comparing our results with the calculated melting reaction. The requisite thermodynamic data needed to calculate the fusion curve are available from the literature, with the only unknown being the pressure dependence of the liquid compressibility, namely K?¡ê?. At pressures <= 0.5 GPa, the effect of not including a liquid K?¡ê? value is negligible, and the calculated fusion curve is in excellent agreement with that determined experimentally to 0.45 GPa in an internally heated pressure vessel by differential thermal analysis (Klement and Cohen, 1975). In order to constrain the value of liquid K?¡ê? from fusion curve analysis, the melting reaction must be determined to higher pressure. In our experiments to 3 GPa in a piston-cylinder apparatus, crystalline K2CO3 powder is preheated to 400 C to eliminate the water and tightly packed in a platinum capsule whose top is marked carefully. A small Pt ball is put at the top of the capsule, covered by a thin layer of K2CO3 powder to prevent it from sticking to the top of the capsule. The Pt ball falls to the bottom when the carbonate becomes liquid and remains at the top of the capsule when the carbonate does not melt. Our results show that K2CO3 remains crystalline up to 1300??C at 3 GPa, which indicates an exceptionally high liquid K?¡ê? value that is >= 18 (based on use of the 3rd-order Birch-Murnaghan EOS). These data indicate that alkaline carbonate liquids are not stable at exceptionally low temperatures at high pressure and that they remain strongly buoyant relative to silicate melts in the deep mantle. These results also support the hypothesis that highly compressible liquids at one bar have correspondingly high K?¡ê? values (Lange, 2002).