Abstract:
We have conducted Ar sorption experiments on two varietis of natural quartz minerals (Q112, Q113) and a synthetic one (QHE37). Runs were performed in an internally heated pressure vessel at 1300°C and pressures of up to 8000 bar for run times between 1 and 12 d, on grain sizes ranging from 11-20 to 60-80 μm. Run products were analysed by gas chromatography (GC), Knudsen cell mass spectroscopy (KMS), electron microprobe (EMP), and scanning electron microscopy (SEM). Release spectra of Ar desorption were monitored by KMS. For sample Q112 and QHE37 two release signals are observed (500-1000°C and 1200-1600°C). When two grain sizes of the same specimen (QHE37) are analysed, the high temperature peak does not vary whereas the low temperature peak is significantly increased with decreasing grain size, suggesting desorption of surface bonded Ar. Argon contents from the high temperature peak indicate an Ar sorption of 28 ppm (QHE37) and 48 ppm (Q112) at 4070 bar. Specimen Q113 does not exhibit low temperature release, nevertheless, its Ar content is higher in smaller grains (11-20 μm: 431 ppm, 60-80 μm: 128 ppm), while increasing the duration of the experiments from ~1 d to ~10 d does not change the Ar content. This apparently erratic behaviour suggests an extrinsic control for Ar sorption. EMP analysis of all samples at the μm scale reveals heterogeneous Ar distribution. A few enriched spots with Ar up to 4000 ppm are observed (e.g., Q113) compared to a background concentration below the detection limit of ~30 ppm. The average concentration measured by EMP is fairly similar to the high temperature step of bulk analytical methods (KMS or GC). We can conclude that bulk measurements of the sorption of Ar do not document equilibrium dissolution. Assuming the Ar diffusivity to be fast enough to permit saturation of at least the 11-20 μm grain fraction after ~10 d, at 8000 bar and 1300°C, an upper bound of Ar solubility can be given as 30 ppm. In contrast, bulk methods yield variable average Ar concentrations which depend on experimental conditions. This indicates that solubilities measured by bulk methods grossly overestimate the true solubility. A quartz/melt partition coefficient of less than 0.006 can be derived.