Experimental determination of argon solubility in silicate melts: An assessment of the effects of liquid composition and temperature

The argon solubility of 38 liquids in the system Na2O-CaO-MgO-Al2O3-SiO2 (NCMAS) has been determined at 1873 K and 1 bar, the argon concentration of presaturated glasses being measured using a static mass spectrometer. For compositions in the subsystem diopside (CaMgSi2O6), nepheline (NaAlSiO4), albite (NaAlSi3O8), anorthite (CaAl2Si2O8), argon solubility is generally a linear function of the relative proportion of each end member, solubility being lowest in diopside melt (1.53 10−5 cm3 STP · g−1 · bar−1) and highest in albite melt (2.88 10−4 cm3 STP · g−1 · bar−1). For the tectosilicate joins studied (SiO2-Na2Al2O4, SiO2-CaAl2O4, SiO2-MgAl2O4) solubility decreases with decreasing silica content in all cases, being highest for Na-bearing liquids and lowest for Mg-bearing liquids at constant molar silica content. Where comparison is possible our results are in good agreement with data from the literature. When our data are considered in isolation we find that argon solubility shows an excellent correlation with calculated ionic porosity. The covariation of argon solubility and liquid density is also reasonable, that with molar volume less convincing and that with polymerization state (as defined by the ratio of the number of nonbridging oxygens and tetrahedral network forming cations; NBO/T) nonexistent. However, when our data are combined with those from the literature no well constrained correlation between argon solubility and ionic porosity is apparent. Based upon this observation and consideration of the temperature dependence of noble gas solubility it is concluded that ionic porosity is not a universally applicable parameter which may be used to predict noble gas solubility as a function of composition, temperature and pressure. Two new models for calculating argon solubility are proposed, both employing the notion of partial molar argon solubilities. The first uses oxide components, for which partial molar argon solubility is directly proportional to partial molar ionic porosity calculated at 1873 K, irrespective of the temperature of experimental equilibration. The second model, which offers the best fit to the available data, employs tetrahedral units rather than oxides as the proposed melt components. This latter model successfully accounts for reported argon solubilities in simple Al-free systems, in simple Al-bearing systems and in natural liquids. This is interpreted to infer that argon is incorporated in large sites in the liquid structure (such as the space within rings of n-tetrahedra) although further work is required to understand the quantitative links between melt structure and noble gas solubility.