Jon Blundy á John Dalton Experimental comparison of trace element partitioning between clinopyroxene and melt in carbonate and silicate systems, and implications for mantle metasomatism Received: 15 July 1999 / Accepted: 18 February 2000 Abstract Experiments in the systems diopside-albite (Di-Ab) and diopside-albite-dolomite (Di-Ab-Dmt), doped with a wide range of trace elements, have been used to characterise the dierence between clinopyrox- ene-silicate melt and clinopyroxene-carbonate melt partitioning. Experiments in Di-Ab-Dmt yielded clino- pyroxene and olivine in equilibrium with CO 2 -saturated dolomitic carbonate melt at 3 GPa, 1375 °C. The experiments in Di-Ab were designed to bracket those conditions (3 GPa, 1640 °C and 0.8 GPa, 1375 °C), and so minimise the contribution of dierential temperature and pressure to partitioning. Partition coecients, de- termined by SIMS analysis of run products, dier markedly for some elements between Di-Ab and Di-Ab- Dmt systems. Notably, in the carbonate system clino- pyroxene-melt partition coecients for Si, Al, Ga, heavy REE, Ti and Zr are higher by factors of 5 to 200 than in the silicate system. Conversely, partition coecients for Nb, light REE, alkali metals and alkaline earths show much less fractionation (<3). The observed dierences compare quantitatively with experimental data on par- titioning between immiscible carbonate and silicate melts, indicating that changes in melt chemistry provide the dominant control on variation in partition coe- cients in this case. The importance of melt chemistry in controlling several aspects of element partitioning is discussed in light of the energetics of the partitioning process. The compositions of clinopyroxene and car- bonate melt in our experiments closely match those of near-solidus melts and crystals in CMAS-CO 2 at 3 GPa, suggesting that our partition coecients have direct relevance to melting of carbonated mantle lherzolite. Melts so produced will be characterised by elevated in- compatible trace element concentrations, due to the low degrees of melting involved, but marked depletions of Ti and Zr, and fractionated REE patterns. These are common features of natural carbonatites. The dierent behaviour of trace elements in carbonate and silicate systems will lead to contrasted styles of trace element metasomatism in the mantle. Introduction Carbonate melts are frequently invoked as agents of metasomatic mass transfer in the upper mantle (e.g. Green and Wallace 1988). Such melts can be produced by partial melting of carbonate-bearing mantle rocks at pressure >2 GPa, and temperatures of ³1000 °C (Wallace and Green 1988; Dalton and Wood 1993). The very low degree of melting involved in carbonatite gen- eration imparts very high incompatible trace element concentrations. Combined with their extreme mobility (Hunter and McKenzie 1989) this makes carbonate melts very ecient metasomatising agents. However, only rarely do metasomatised mantle rocks themselves preserve unambiguous evidence for the passage of a carbonate melt, such as carbonate crystals (e.g. Ionov et al. 1993, 1996; McInnes and Cameron 1994; Schiano et al. 1994; Yaxley et al. 1998). Typically, evidence for the passage of metasomatic carbonate melt in mantle rocks is cryptic, for example the presence of secondary clinopyroxene (cpx) reaction rims on primary orthopy- roxene (e.g. Yaxley et al. 1991; Hauri et al. 1993; Col- torti et al. 1998), and/or unusual bulk-rock trace element signatures (e.g. Yaxley et al. 1991; Dautria et al. 1992; Rudnick et al. 1993; Coltorti et al. 1998). In no case is the pristine metasomatic agent itself preserved and its composition must always be inferred through calculation (e.g. Hauri et al. 1993; Rudnick et al. 1993). Such calculations hinge critically on assumptions about Contrib Mineral Petrol (2000) 139: 356±371 Ó Springer-Verlag 2000 J. Blundy (&) á J. Dalton CETSEI, Department of Earth Sciences, University of Bristol, Wills Memorial Building, Bristol, BS8 1RJ, UK e-mail: Jon.Blundy@bristol.ac.uk J. Dalton Department of Geosciences, University of Texas at Dallas, Richardson, Texas 75083, USA Editorial responsibility: V. Trommsdor