Discrepancy Between Different Melting Indicators in Mantle Peridotites Caused by Garnet Field Melting

We present here new results from spinel lherzolites from one ultramafic massif (Ozren massif, Dinaride Ophiolite Belt) and one ultraslow-spreading mid-ocean ridge (Gakkel Ridge, Arctic Ocean). In terms of their modal and mineral composition, the investigated peridotites are comparable to the most fertile peridotites from different tectonic settings (high modal cpx; spinel Cr# 10-14; Al2O3 wt% in opx 5-6% (core), in cpx 6-7% (core)). In contrast, their cpx LREE content is unexpectedly low (CeCN 0.02-0.1), considering their high HREE cpx content (YbCN 8-9) and the overall fertility of these lherzolites. In fact, cpx LREE concentrations are an order of magnitude lower than predicted by even a pure fractional melting model in the spinel field. One possible explanation would be the change of the major element composition by later post-melting processes. Therefore it is important to evaluate factors that could modify spinel compositions. All pyroxenes in the investigated peridotites are zoned (although cpx to a lesser extent), with high Al2O3(6- 7%wt) and Cr2O3(0.9-1%wt) in the cores and low Al2O3(3-4%wt) and Cr2O3(0.4-0.5%wt) at their rims, but at fairly constant Cr#. In contrast, spinel is unzoned. The positive correlation between Cr and Al in pyroxenes can not be explained by melting, but rather indicates cooling-induced subsolidus Cr and Al exchange to the existing spinel, formation of new spinel and subsequent spinel equilibration. Since diffusion of Al is several orders of magnitude faster in spinels than in pyroxenes, spinel can equilibrate while pyroxenes remain zoned. In this process spinel Cr# is notably changed but is still controlled by the pyroxene composition preserving the very good correlation between Cr# in spinels and pyroxenes at global scale. Therefore, even if spinel composition is changed during cooling it reflects melting in and is still vital for our understanding of mantle processes. However, cooling can not explain the discrepancy between Cr# in mantle minerals and cpx-LREE content. The LREE-depleted cpx could be best explained by several degrees of melting in the garnet stability field (2- 3%) followed by several degrees of melting in the spinel stability field (2-3%), especially taking into account that the cpx MREE content is also affected and shifted towards lower concentrations than predicted by spinel field melting alone. Hellebrand et al. 2001 state that there is a good correlation between spinel Cr# and cpx HREE concentrations, and a poor one between spinel Cr# and cpx LREE. Since LREE are easily influenced by late processes such as refertilization, they conclude that cpx HREE and spinel Cr# reflect melting processes while LREE do not. However, the relative fractionation of LREE and MREE from the HREE is very sensitive to melting in the presence of garnet. At the same time, alumina as well as HREE behaves compatible and both will be left nearly unchanged by melting in the garnet field thus not disturbing their relative correlation. We conclude that spinel Cr# and cpx HREE fail to account for garnet field melting, but reflect exclusively spinel field melting. As a result the total degree of melting would be underestimated using the method by Hellebrand et al. 2001 alone. Hellebrand et al. (2001):Nature 410,677-680