H2O solubility in haplogranitic melts; compositional, pressure, and temperature dependence

Abstract

H₂O solubility has been determined in haplogranitic melts (system SiO₂-NaAlSi₃O₈-KAlSi₃O₈, Qz-Ab-Or) in the range 0.5–8 kbar and 800–1350 °C. Three types of starting materials were used: dry glass cylinders, prehydrated glass pieces, or dry glass blocks surrounded by glass powder. All the starting materials gave consistent results. The H₂O contents of the glasses were determined by Karl-Fischer titration. Dissolved H₂O was demonstrated to be distributed homogeneously throughout the isobarically quenched melts (glasses) using infrared spectroscopy. The compositional dependence of H₂O solubility was mainly determined at 0.5 kbar, 900 and 1000 °C; 1 kbar, 850 °C; and 4.8 kbar, 800 °C. Seventeen compositions containing 25, 35, or 45 wt% normative Qz and with various Or/(Or + Ab) ratios (0.86–0.09, Ab and Or expressed as normative weight percent) were investigated. At 0.5 kbar, H₂O solubility was little affected by the anhydrous composition. By contrast, molar H₂O solubility in aluminosilicate melts was significantly dependent upon anhydrous composition between 1 and 5 kbar. The highest solubility values were obtained for the most Ab-rich melts. This alkali effect has important implications for the physical and chemical properties of granitic melts. The effect of pressure (P) on H₂O solubility at P ≥ 3 kbar is greater than that reported in previous studies. Between 3 and 8 kbar at 800 °C, there is a (nearly linear) positive correlation between P and H₂O solubility. The effect of temperature (T) on H₂O solubility was investigated for a composition Qz₂₈Ab₃₈Or₃₄ (normative weight percent) in the P-T range 0.5–8 kbar and 800–1350 °C. Water solubility ranged from retrograde (with increasing T) at P ≤ 4 kbar through temperature independence at approximately 4.5 kbar to prograde at P = 5 kbar. Calculated H₂O solubilities using the model of Burnham and Nekvasil (1986) are slightly high at 0.5 kbar and significantly low at 5 kbar, compared with the experimental data. This implies that calculated H₂O activities for haplogranitic systems using the H₂O content of the melt may be overestimated at high pressure (P ≥ 5 kbar). Using the thermodynamic model of Silver and Stolper (1985) and assuming a proportion of molecular H₂O and OH groups close to that defined for albite melts by Silver and Stolper (1989), we found that the partial molar volume of H₂O in a melt with a composition Qz₂₈Ab₃₈Or₃₄ has to be close to 10–12 cm³/mol to obtain a good agreement between the calculated and the experimentally determined H₂O solubility curves in the pressure range 1–8 kbar at 900 °C.