Element Partitioning between Immiscible Carbonatite and Silicate Melts for Dry and H2O-bearing Systems at 1-3 GPa

Martin, Lukas H. J. ; Schmidt, Max W. ; Mattsson, Hannes B. ; Guenther, Detlef

In: Journal of Petrology, 2013, vol. 54, no. 11, p. 2301-2338

Ajouter à la liste personnelle
    Summary
    Carbonatite and silicate rocks occurring within a single magmatic complex may originate through liquid immiscibility. We thus experimentally determined carbonatite/silicate melt partition coefficients (Dcarbonate melt/silicate melt, hereafter D) for 45 elements to understand their systematics as a function of melt composition and to provide a tool for identifying the possible conjugate nature of silicate and carbonatite magmas. Static and, when necessary, centrifuging piston cylinder experiments were performed at 1-3 GPa, 1150-1260°C such that two well-separated melts resulted. Bulk compositions had Na K, Na ∼ K, and Na K; for the latter we also varied bulk H2O (0-4 wt %) and SiO2 contents. Oxygen fugacities were between iron-wüstite and slightly below hematite-magnetite and were not found to exert significant control on partitioning. Under dry conditions alkali and alkaline earth elements partition into the carbonatite melt, as did Mo and P (DMo >8, DP= 1·6-3·3). High field strength elements (HFSE) prefer the silicate melt, most strongly Hf (DHf = 0·04). The REE have partition coefficients around unity with DLa/Lu = 1·6-2·3. Transition metals have D < 1 except for Cu and V (DCu ∼ 1·3, DV = 0·95-2). The small variability of the partition coefficients in all dry experiments can be explained by a comparable width of the miscibility gap, which appears to be flat-topped in our dry bulk compositions. For all carbonatite and silicate melts, Nb/Ta and Zr/Hf fractionate by factors of 1·3-3·0, in most cases much more strongly than in silicate-oxide systems. With the exception of the alkalis, partition coefficients for the H2O-bearing systems are similar to those for the anhydrous ones, but are shifted in favour of the carbonatite melt by up to an order of magnitude. An increase of bulk silica and thus SiO2 in the silicate melt (from 35 to 69 wt %) has a similar effect. Two types of trace element partitioning with changing melt composition can be observed. The magnitude of the partition coefficients increases for the alkalis and alkaline earths with the width of the miscibility gap, whereas partition coefficients for the REE shift by almost two orders of magnitude from partitioning into the silicate melt (DLa = 0·47) to strongly partitioning into the carbonatite melt (DLa = 38), whereas DLa/DLu varies by only a factor of three. The partitioning behavior can be rationalized as a function of ionic potential (Z/r). Alkali and alkaline earth elements follow a trend, the slope of which depends on the K/Na ratio and H2O content. Contrasting the sodic and potassic systems, alkalis have a positive correlation in D vs Z/r space in the potassic case and Cs to K partition into the silicate melt in the presence of H2O. For the divalent third row transition metals on the one hand and for the tri- and tetravalent REE and HFSE on the other, two trends of negative correlation of D vs Z/r can be defined. Nevertheless, the highest ionic strength network-modifying cations (V, Nb, Ta, Ti and Mo) do not follow any trend; understanding their behavior would require knowledge of their bonding environment in the carbonatite melt. Strong partitioning of REE into the carbonatite melt (DREE = 5·8-38·0) occurs only in H2O-rich compositions for which carbonatites unmix from evolved alkaline melts with the conjugate silicate melt being siliceous. We thus speculate that upon hydrous carbonatite crystallization, the consequent saturation in fluids may lead to hydrothermal systems concentrating REE in secondary deposits