Amphibole-, Clinopyroxene- and Plagioclase-Melt Partitioning of Trace and Economic Metals in Halogen-Bearing Rhyodacitic Melts. (17th July 2018)

• Record Type: Journal Article
• Authors: Iveson, Alexander A; Rowe, Michael C; Webster, James D; Neill, Owen K
• Is Part Of: Journal of petrology. Volume 59:Number 8(2018:Aug.)
• Page Start: 1579
• Page End: 1604
• Publication Date: 2018-07-17
• Subjects: experimental petrology -- partitioning -- rhyodacite -- trace elements; Petrology -- Periodicals; 552
• DOI: 10.1093/petrology/egy072
• Languages: English
• ISSNs: 0022-3530

Abstract

Abstract: The crystal-melt partitioning behaviour of a suite of trace and economically important elements—Li, Sc, V, Mn, Ni, Cu, Zn, Ga, Rb, Sr, Y, Zr, Nb, Mo, Sn, Ba, Hf, W and Pb—has been investigated experimentally under fluid-saturated conditions for rhyodacitic melts from a natural volcanic sample from Mt. Usu, Japan. Homogeneous starting glasses were doped with some of these trace elements, and aqueous solutions buffered the F and Cl concentrations in the melts and coexisting fluids. The partitioning of these trace elements between amphibole, plagioclase, clinopyroxene and melt was investigated at conditions of 810–860°C, 1·5–4·05 kbar and oxygen fugacity ( ƒ O2 ) ≈ NNO -0·5 to NNO +2 log units. The trace element contents of the product phases were measured by laser ablation mass spectrometry. Over the duration of the experiments, Cu was depleted from the melt by diffusive loss to the walls of the pure Au capsules; thus some capsules were pre-saturated with ∼4 wt % Cu through electroplating and annealing prior to the experiment. The results of 24 experiments show that all analysed trace elements appear to obey Henry's Law partitioning; the Nernst-type crystal-melt partition coefficients for these elements range from strongly incompatible (e.g. D Nb mineral / melt ≈ 0 for plagioclase) to strongly compatible (D Sc mineral / melt >15 for amphibole). Unlike the other trace elements, the average partitioning behaviours of Li and Cu are similar for all three phases, with average D Li cpx / melt (0·26 ± 0·04) ≈ D Li plag / melt (0·23 ± 0·02) ≈ D Li amph / melt (0·19 ± 0·06), and D Cu cpx / melt (0·11 ± 0·04) ≈ D Cu plag / melt (0·08 ± 0·01) = D Cu amph / melt (0·07 ± 0·02). The data show that fractionation of amphibole, clinopyroxene, and plagioclase should be an efficient mechanism for enrichment of Li, Cu, Mo, W and Pb in sulfide-poor rhyodacitic melts, with moderately to strongly incompatible D mineral/melt values for these elements in all three phases. Conversely, Mn, Zn and Sn will be depleted by extensive amphibole and clinopyroxene crystallisation, given average D mineral/melt values >2 for these metals. Of all the elements investigated, only Ga and Sr show compatible partitioning (D mineral/melt ≥ 1) in favour of plagioclase. Relative to major element substitution controls and variations in melt chemistry (in particular the molar ratio of Al2 O3 /(CaO+Na2 O+K2 O) and halogen contents), the narrow range in temperature, pressure, and ƒ O2 of these experiments appears not to affect trace element partitioning behaviour strongly. However, V partitioning between amphibole, clinopyroxene and melt shows a strong dependence on experimental ƒ O2, becoming more compatible in amphibole and clinopyroxene crystallised in more reduced experiments with concurrent lower crystal Mg/Mg+FeTOT (Mg#) values (D V amph / melt >26 where amphibole Mg# = 0·63, compared with D V amph / melt = 0·75 where amphibole Mg# = 0·82). Comparison of our partitioning data with relevant literature data for the same trace elements and T, P, ƒ O2 conditions suggests that subtle differences in melt composition, in particular melt F concentrations, may induce significant variations in D amph/melt values specifically. Therefore, where published trace element partition coefficients are applied to model natural fractional crystallisation processes, care must be taken to account for the combined controls of major element stoichiometry and/or melt composition on the particular partition coefficient.