Copper and Li diffusion in plagioclase, pyroxenes, olivine and apatite, and consequences for the composition of melt inclusions

Silicate melt inclusions are an important tool to reconstruct original concentrations of volatiles and metals in silicate melts, as these constituents are partly lost during magma solidification. However, melt inclusions analyzed in co-precipitated minerals in volcanic rocks commonly show strongly divergent Cu contents, raising doubts about the validity of their composition. To understand the origin of this divergence a multiple approach was followed. First, the phenomenon was documented by recording petrographic details of five volcanic samples and analyzing 189 melt inclusions by laser-ablation ICP-MS. Second, diffusion experiments on gem-quality plagioclase, clinopyroxene, orthopyroxene, apatite and olivine were performed in a gas mixing furnace to constrain diffusion coefficients of Cu and to a lesser extent of Li. Third, re-equilibration experiments were performed on melt inclusions within plagioclase crystals to induce changes in their Cu and Li contents. The LA-ICP-MS analyses reveal that plagioclase-hosted and orthopyroxene-hosted melt inclusions in volcanic rocks commonly contain an order of magnitude more Cu than melt inclusions in co-precipitated clinopyroxene and olivine. Plagioclase-hosted melt inclusions in intrusive rocks, on the other hand, do not show this divergence. The diffusion experiments conducted on minerals find that Cu and Li diffusion in plagioclase is extremely fast (log D = −13.0 to −11.5 m2 s−1 at 1000 °C) and in both cases occurs via two separate diffusion mechanisms. Copper diffusion coefficients in apatite, clinopyroxene, orthopyroxene and olivine are 2–3 orders of magnitude lower but are still high compared to most other elements. Both the re-equilibration experiments on melt inclusions and quantitative modeling based on the measured diffusion coefficients demonstrate that at 1000 °C plagioclase-hosted melt inclusions can re-equilibrate their Cu and Li content with that of the surrounding magma within a few hours to a few weeks, whereas apatite-, clinopyroxene-, orthopyroxene- and olivine-hosted melt inclusions require tens of years to hundreds of years to do so. Abnormally high Cu contents of plagioclase- and orthopyroxene-hosted melt inclusions appear to result from postentrapmental Cu gain. In the case of orthopyroxene-hosted melt inclusions Cu seems to have been gained as a consequence of postentrapmental sidewall crystallization, which triggered the formation of sulfide globules that acted as a Cu sink. This process likely occurred at depth in magma chambers. Plagioclase-hosted melt inclusions, on the other hand, seem to have gained Cu in exchange for hydrogen that diffused out of the melt inclusions during or after volcanic eruption. Our results suggest that Cu (and also Li) concentrations measured in melt inclusions have to be generally treated with caution, and that particular prudence has to be used in the case of plagioclase and orthopyroxene hosts.