Compositional dependent compressibility of dissolved

American Mineralogist, Volume 96, pages 1402–1409, 2011

Amorphous materials: Properties, structure, and durability†

Compositional dependent compressibility of dissolved water in silicate glasses Wim J. Malfait,1,* Carmen Sanchez-Valle,1,* Paola Ardia,1,2 Etienne Médard,3,4,5 and Philippe Lerch6 1 Institute for Geochemistry and Petrology, ETH Zurich, CH-8092 Zurich, Switzerland Department of Geology and Geophysics, University of Minnesota, Minneapolis 55455, U.S.A. 3 Laboratoire Magmas et Volcans, Clermont Université, Université Blaise Pascal, BP 10448, F-63000 Clermont-Ferrand, France 4 CNRS, UMR 6524, LMV, F-63038 Clermont-Ferrand, France 5 IRD, R 163, LMV, F-63038 Clermont-Ferrand, France 6 Swiss Light Source, Paul Scherrer Institute, CH-5232 Villigen-PSI, Switzerland 2

Abstract The sound velocities and elastic properties of a series of hydrous rhyolite, andesite, and basalt glasses have been determined by Brillouin scattering spectroscopy at ambient conditions to elucidate the effect of glass composition on the compressibility of dissolved water. Both the adiabatic bulk (KS) and shear modulus (µ) of the dry glasses decrease with increasing silica content (KS,basalt > KS,andesite > KS,rhyolite and µbasalt > µandesite > µrhyolite). For each composition, the shear modulus systematically decreases with increasing water content. Although the addition of up to 14 mol% water decreases the KS of andesite and basalt glasses by up to 6%, there is no discernable effect of water on the KS of the rhyolite glasses. The partial molar KS of dissolved water (KS) in rhyolite, andesite, and basalt glasses are 37 ± 5, 19 ± 7, and 40 ± 3 GPa, corresponding to partial molar isothermal compressibilities (βT) of 0.029 ± 0.005, 0.042 ± 0.004, and 0.026 ± 0.002 GPa−1, respectively. These results indicate that the compressibility of dissolved water strongly depends on the bulk composition of the glass; hence, the partial molar volume of water cannot be independent of the bulk composition at elevated pressure. If the compressibility of dissolved water also depends on composition in the analog melts at high temperature and pressure, these observations will have important consequences for magmatic processes such as magma mixing/unmixing and fractional crystallization. Keywords: Hydrous silicate melts, silicate glasses, sound velocities, compressibility, partial molar volume of water, Brillouin scattering

Introduction The density of hydrous magmas is a key parameter affecting the outcome and timescales of many magmatic processes, e.g., the crystal settling velocities in a magma chamber and the emplacement depth of intrusions. Direct density measurements on hydrous liquids at magmatic temperature and pressure conditions are, however, experimentally challenging. Nevertheless, some measurements have been made for hydrous compositions at simultaneous high temperature and pressure with various methods, including a customized internally heated pressure vessel (Burnham and Davis 1971), the sink-float method (Agee 2008; Matsukage et al. 2005; Sakamaki et al. 2006), X‑ray radiography (Sakamaki et al. 2009) and Brillouin scattering spectroscopy in a diamond-anvil cell (Tkachev et al. 2005). In the last decade, additional constraints on the volumetric and structural properties of volatile-bearing melts at high pressure-temperature conditions have been provided by ab initio calculations (Mookherjee et al. 2008). However, most of the available data correspond either to * E-mail: [email protected]; carmen.sanchez@erdw. ethz.ch † The Amorphous Materials special section papers are presented as a group at http://www.minsocam.org/msa/ammin/toc/ and on GSW at http://ammin.geoscienceworld.org/misc/virtual_special_list.dtl. 0003-004X/11/0809–1402$05.00/DOI: 10.2138/am.2011.3718

mafic and ultramafic melts at temperatures and pressures relevant to the mantle transition zone, or to more acidic compositions in a narrow low pressure-temperature range. Consequently, no experimental density data are available for hydrous melts with compositions and pressure-temperature conditions relevant for mid-oceanic ridges, subduction zones, and mid- to lower-crustal settings. Despite the fundamental dynamic and thermodynamic differences between the liquid and glassy state (Webb 1992), the physical properties of hydrous silicate glasses at ambient pressure and temperature conditions can provide insights into the properties of the analog melts. For instance, although there are noticeable differences between the thermal expansion of glasses and melts (Dingwell and Webb 1990; Webb 1992; Webb et al. 1992), the thermal expansion of hydrous silicate glasses up to the glass transition temperature (Tg) has been used to constrain the partial molar volume of water in silicate melts (Ochs and Lange 1999). Apart from non-ideal effects for Al2O3 (Ai and Lange 2008; Courtial and Dingwell 1999; Kress and Carmichael 1991), most of the experimentally derived density and compressibility data of anhydrous glasses and melts can be reproduced well by an ideal mixing model with respect to volume (Bottinga and Weil 1970; Nelson and Carmichael 1979; Rivers and Carmichael 1987). In this model, the molar volume V is given by V =

1402

MALFAIT ET AL.: COMPRESSIBILITY OF DISSOLVED WATER IN SILICATE GLASSES

ΣxiVi and the isothermal compressibility by βT = Σxi,VβiT, where xi, xi,V, Vi, and βiT are the mole fraction, volume fraction, partial molar volume, and partial molar isothermal compressibility of component i, respectively. The density data on hydrous glasses seem to indicate that the partial molar volume of water (VH2O) at atmospheric pressures is independent of the bulk composition of the glass (VH2O = 12 ± 0.5 cm3/mol) for a wide range of glass compositions (Ochs and Lange 1999; Richet et al. 2000; Richet and Polian 1998). In addition, the partial molar bulk modulus (KS) of dissolved water in andesite glasses, as determined by Brillouin scattering spectroscopy (18 ± 3 GPa), resembles that of Ice VII (23.9 ± 0.9 GPa) (Hemley et al. 1987; Richet and Polian 1998), suggesting that water behaves as a dense ice-like component in silicate glasses. These observations, combined with measurements of the thermal expansion of hydrous glasses (Ochs and Lange 1997, 1999), led to density models for hydrous melts that assume ideal mixing of the water component, i.e., a compositionally independent partial molar volume of the dissolved water component, in high-temperature high-pressure melts (Jing and Karato 2009; Ochs and Lange 1999). These models also assume that the partial molar compressibility of dissolved water is independent of melt composition and adopt a value for the compressibility of water derived from volumetric data on hydrous albite melts up to 0.85 GPa (Burnham and Davis 1971). However, since only hydrous melts with albite compositions have been investigated systematically at high-pressure and high-temperature conditions, a possible compositional dependence of the partial molar KS of dissolved water, hence the compressibility, cannot be excluded (Richet 2005; Richet et al. 2005). Indeed, a compositionally dependent partial molar volume of water at high pressure has been suggested based on the pressure dependence water solubility in silicate melts (Mysen and Acton 1999; Mysen and Wheeler 2000). To investigate whether the compressibility of dissolved water depends on glass composition, we have determined the acoustic velocities and elastic moduli of a series of hydrous rhyolite, andesite, and basalt glasses using Brillouin spectroscopy at ambient conditions. The velocities are used to determine the partial molar KS of dissolved water in the glasses and confirm that KS strongly depends on the silicate composition. The implications of this observation on the applicability of current density models for hydrous silicate melts will be discussed.

Experimental methods Sample synthesis and characterization Hydrous glasses with rhyolite, andesite, and basalt compositions have been studied (Tables 1, 2, and 3). The rhyolite and basalt glasses are part of a set of well-characterized samples recently used for viscosity measurements (Ardia et al. 2008) and phase relation investigations (Médard and Grove 2008), respectively; the hydrous andesite glasses were specifically synthesized for the present Brillouin scattering study. The compositional homogeneity, including water content, of all glasses has been confirmed by electron microprobe and infrared spectroscopy or secondary mass spectrometry. A detailed description of the synthesis conditions and characterization of the rhyolite glasses can be found in Ardia et al. (2008). The composition (Table 1) corresponds to an iron-free analog for rhyolite melts and lies close to the 0.5 GPa ternary minimum in the hydrous albite-orthoclase-quartz system (Holtz et al. 1992). The hydrous glasses have been synthesized in an internally heated pressure vessel at 0.18 GPa. The water contents were verified by Karl-Fischer titration (Behrens and Stuke 2003), infrared spectroscopy (Okumura and Nakashima 2005), and Raman spectroscopy (Zajacz et al. 2005). The dry glass (HGG0) was synthesized by dehydration of a hydrous glass at ambient pressure. To allow the comparison of the

1403

results on glasses quenched from identical pressure conditions, an additional dry sample (HGG0b) was prepared at elevated pressure: pieces of the nearly anhydrous glass HGG0 were welded into Pt capsules, held for 30 min at 1000 °C and 0.18 GPa in an externally heated TZM pressure vessel and quenched to below Tg with quench rates on the order of 120 K/min. With the exception of a lower Al content, the composition of the andesite glasses used in this study are close to those previously investigated by Brillouin scattering spectroscopy (Richet and Polian 1998). The glasses were synthesized in welded Pt capsules in an end-loaded piston-cylinder apparatus using a talc-Pyrex-MgO assembly and a graphite heater. The glasses were prepared from decarbonated mixtures of oxides, hydroxides, and carbonates and water was added as AlOOH. The melts were held for 4 h at 0.7 GPa and 1673 K and quenched to glasses by turning off the power of the furnace (estimated quench rate of 100–200 K/s). The composition of the run products, analyzed by electron microprobe, agrees with the target composition within analytical uncertainty (Table 2). The water content was analyzed by infrared absorption spectroscopy on the same double-side polished sections that were subsequently used for the Brillouin measurements. The OH− and H2Omol content, determined from the height of the 4500 and 5200 cm–1 bands (Malfait 2009; Scholze 1966) and the average of the molar absorption coefficients obtained from the literature (Mandeville et al. 2002), agree with the target composition within 0.2 wt% (Table 2). The analytical uncertainties are most likely larger than the errors introduced during the sample synthesis. Thus, throughout this study, we use the nominal water content values (Table 1). Based on the Fourier transform infrared (FTIR) spectra, the CO2/carbonate content was smaller than 80 ppm for all samples. The basalt glasses were synthesized using powders from a high-alumina basaltic rock from the Giant Crater Lava Field, Medicine Lake Volcano, California [sample 82-72f (Donnelly-Nolan et al. 1991)]. The density of the andesite and basalt glasses was determined using the sink/ float method in a diiodomethane-acetone mixture. In this method, the density of the glass equals the density of the mixture for which the sample neither floats nor sinks, which can be calculated from the acetone/diiodomethane ratio. The reproducibility for repeated measurements is better than 0.1% and the densities of an α-quartz and

Table 1. Nominal compositions (wt%) and synthesis conditions Rhyolite Andesite Basalt SiO2 78.14 64.53 47.73 TiO2 – 0.57 0.59 Al2O3 11.79 17.58 18.51 FeO – – 8.21 MnO – – 0.15 MgO – 3.33 10.51 CaO – 9.39 12.01 Na2O 5.49 3.64 2.16 K2O 4.59 0.96 0.07 P2O5 – – 0.06 H2O 0–5.24 0–3.01 0–4.40 T (K) 1273–1873 1673 1290–1486 P (GPa) 0–0.18 0.70 0–0.20

Table 2. Composition of the hydrous andesite glasses* Sample SiO2 TiO2 Al2O3 MgO CaO Na2O K2O H2O† Total (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) Nominal composition Mix dry2 64.53 0.57 17.58 3.33 9.39 3.64 0.96 0.00 100 Mix 1/3 63.88 0.56 17.41 3.30 9.30 3.60 0.95 1.00 100 Mix 2/3 63.23 0.56 17.23 3.26 9.20 3.57 0.94 2.01 100 Mix wet 62.58 0.55 17.06 3.23 9.11 3.53 0.93 3.01 100 Analyzed composition Mix dry2 63.7(2) 0.54(2) 17.4(1) 3.28(3) 9.60(7) 3.44(4) 0.98(1) 0.04 98.98 Mix 1/3 63.2(3) 0.55(1) 17.0(2) 3.26(3) 9.57(9) 3.36(5) 0.96(1) 1.06 98.95 Mix 2/3 62.5(2) 0.54(1) 16.8(2) 3.22(4) 9.4(1) 3.30(4) 0.95(1) 2.18 98.91 Mix wet‡ 62.2(2) 0.54(1) 16.4(1) 3.16(4) 9.2(1) 3.23(3) 0.94(2) 2.87 98.61 * The electron microprobe analyses (EPMA) were carried out on a Jeol JX8200 with peak counting times of 60 s, an acceleration voltage of 15 kV, a beam current of 10 nA, and a spot size of 20 µm; the count rates for Na and K were stable over time with these conditions. Reported values are averages of 10 point analysis. † The infrared absorption spectra were collected with a Bruker Hyperion 3000 microscope connected to a Vertex 70 interferometer. Spectra were collected on 5 points for each section. The uncertainty (10% relative) is mostly due to the uncertainty on the molar absorption coefficient. ‡ This sample contains traces of small bubbles (