Solubility of Au in Cl- and S-bearing hydrous silicate melts

Available online at www.sciencedirect.com

Geochimica et Cosmochimica Acta 74 (2010) 2396–2411 www.elsevier.com/locate/gca

Solubility of Au in Cl- and S-bearing hydrous silicate melts R.E. Botcharnikov a,*, R.L. Linnen b, F. Holtz a b

a Institut fu¨r Mineralogie, Leibniz Universita¨t Hannover, Callinstr. 3, D-30167 Hannover, Germany Department of Earth and Environmental Sciences, University of Waterloo, Waterloo, Ont., Canada N2L 3G1

Received 2 July 2009; accepted in revised form 16 September 2009; available online 22 September 2009

Abstract The solubility of Au in Cl- and S-bearing hydrous rhyodacitic and andesitic melts has been experimentally investigated at 1050 °C, 200 MPa and log fO2 close to the Ni/NiO solid oxygen buffer (NNO). The concentrations of Au in the experimental glasses have been determined using Laser Ablation ICP-MS (LA) with special efforts to avoid incorporation of Au micronuggets in the analysis. It is concluded that metal micronuggets are an experimental artefact and produced by Au partitioning into the fluids during heating with consequent precipitation on fluid dissolu#132;tion in the melting glass powder. Hence, the micronuggets do not represent quench phases and must be excluded from the analysis. The micro-analytical data obtained by LA show that Au concentrations vary from 0.2 to 2.5 ppm by weight, generally consistent with the literature data for other melt compositions. The measured Au concentrations increase with increasing amounts of Cl and S dissolved in the silicate melt and show a correlation with the apparent activities of Cl and S in the system. The apparent activities of Cl and S are defined by the simplified linear relationship between volatile concentrations in the melt and activity of volatiles. The maximum activity (a* = 1) is assumed to be reached at the saturation of the systems in respect of Cl-rich brine or FeS liquid for Cl and S, respectively. The dependence of Au solubility on the concentrations/activities of Cl and S at the fixed redox conditions shows that Au may form not only oxide- but also Cl- and S-bearing complexes in silicate melts. Furthermore, it indicates that exsolution of S and Cl from the melt by degassing/segregation/crystallization processes may lead to mobilization and extraction of Au into the fluid, liquid and/or mineral phase(s). Ó 2010 Published by Elsevier Ltd.

1. INTRODUCTION The porphyry-type deposits are important sources of ore metals and in particular of Au, Cu and Mo. The ore deposits are associated temporally and spatially with magmatic intrusions (e.g., Hunt, 1991; Sillitoe, 1979, 2000, 2008), and it is generally accepted that their formation is related to the precipitation of metals from the magmatichydrothermal fluid phase(s) exsolved from a subjacent magmatic reservoir (e.g., Bodnar, 1995; Halter et al., 2002; Harris et al., 2003, 2004; Hattori and Keith, 2001; Hedenquist et al., 1993, 1994; Hedenquist and Lowenstern, *

Corresponding author. Tel.: +49 511 762 19183; fax: +49 511 762 3045. E-mail address: [email protected] (R.E. Botcharnikov). 0016-7037/$ - see front matter Ó 2010 Published by Elsevier Ltd. doi:10.1016/j.gca.2009.09.021

1994; Holland, 1972; Lowenstern, 1993; Shinohara and Hedenquist, 1997; Taran et al., 2000; Ulrich et al., 1999). Moreover, recent studies (e.g., Audetat et al., 1998; Hedenquist et al., 1993; Heinrich et al., 2004) indicate that the composition of the ore (i.e., metal ratios) is directly related with the composition of the primary magmatic fluid (e.g., Halter et al., 2002, 2005; Ulrich et al., 1999, 2002), although the ore distribution and ore-grade in host rocks are mainly controlled by the processes occurring at the deposition site. A number of studies have emphasized the crucial role of mafic magmas and related fluids within intermediate to felsic plutons in the mobilization, extraction and transport of ore metals (Audetat and Pettke, 2006; Halter et al., 2005; Hattori and Keith, 2001; Keith et al., 1997; Maughan et al., 2002), reporting that felsic rocks hosting the mineralization may not always represent the primary source of metals.

Au solubility in volatile-bearing magmas

Fluid inclusions associated with ore mineralization are characteristically composed of low-salinity aqueous vapor and brines and often contain sulfide or sulfate mineral phases (e.g., Harris et al., 2003; Heinrich et al., 1999; Ulrich et al., 1999, 2002), indicating that Cl and S exert a significant control on the metal mobilization and transport by the magmatic/hydrothermal fluid(s). However, at given temperature, pressure and redox conditions, the partitioning of Cl and S between magmatic melt and exsolving fluid phase(s) is strongly dependent on the composition of silicate melt and on the concentrations of other dissolved volatiles. Hence, it can be suggested that the bulk composition of the magmatic source may have a significant influence on the activity and mobility of ore metals. To better understand magmatic processes in volatile-bearing (H2O, Cl, S) magmas, leading to the mobilization and transport of Au to the deposition site, quantitative estimates on the solubility of Au in felsic to mafic silicate melts are necessary. This can be assessed from experimental simulations of magmatic conditions, e.g., by performing solubility/partitioning experiments. In this study we present experimental data on the solubility of Au in rhyodacitic and andesitic melts at 200 MPa. The results are compared with the available literature data, mainly obtained for basaltic and rhyolitic melts, to constrain possible effects of melt composition and activity of volatiles on the Au capacity of natural melts. 2. STATE OF THE ART IN EXPERIMENTAL STUDIES ON Au SOLUBILITY So far, the experimental approaches were mostly focused on the behavior of Au in dry mafic magmas to place constraints on the processes, occurring during the formation of the core and mantle of the Earth and in hydrous rhyolitic and haplogranitic magmas to study the transport of metals in magmatic–hydrothermal systems assumed to be related to porphyry ore deposits. The solubility of Au in dry basaltic melts equilibrated with AuPd alloys or molten Au (coexisting with Ir- or Os-bearing crystals) was investigated at atmospheric pressure and at T > 1250 °C in a wide range of redox conditions (about 12 logarithmic units in fO2), demonstrating a positive dependence of gold solubility on fO2 (Borisov and Palme, 1996; Brenan et al., 2005). Both studies reported Au concentrations in the melt from <1 up to 50–60 ppm, and suggested that Au1+ is the major stable species in volatile-free basaltic melts at geologically relevant conditions. The study of Jana and Walker (1997) reported a positive effect of S on the solubility of Au in mafic silicate melts at 1500–2200 °C and 1–5 GPa. The Au concentrations in basaltic liquids were measured with electron microprobe to be about 30–240 ppm which is two orders of magnitude higher than in S-free runs. However, the experimental products contained metal blebs that might have contaminated the Au analyses (see also discussion of Borisov and Palme (1996). The experiments of Bezmen et al. (1994) performed at 1200–1300 °C and at 100–400 MPa fluid pressure show that graphite-saturated volatile-bearing (H–O–C–S-fluids) basaltic melts coexisting with Au-bearing PGE alloys and

2397

sulphide melts may contain up to 12 ppm Au. The bulk Au concentrations were determined by Instrumental Neutron Activation Analysis (INAA) which can be dramatically influenced by the presence of Au micronuggets in the silicate glasses. Moreover, since Au was only a trace element in the experimental metal alloys (the Au activity in the experiments was very low), these experiments do not represent reliable Au solubility data for volatile-bearing basalts. Several experimental studies were focused on the quantitative evaluation of Au partitioning between Au metal, hydrous silicic melts, minerals (±magnetite ± pyrrhotite) and coexisting fluid(s) in S-free (Frank et al., 2002; Simon et al., 2003, 2005, 2007, 2008) and S-bearing magmatic assemblages (Jugo et al., 1999; Simon et al., 2007, 2008; Bell et al., 2009). The experimental results showed that rhyolitic/ haplogranitic melts can dissolve up to 10 ppm of Au at temperatures of 800–850 °C and pressures of 100–150 MPa in both S-free and S-bearing systems. It must be noted that the common occurrence of Au micronuggets in experimental glasses makes the accurate analysis of Au concentrations difficult. The application of different analytical methods like INAA, Secondary Ion Mass Spectrometry (SIMS) and Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA) showed that the contribution of micronuggets to the measurement can be considerably minimized, using LA technique. The LA data of Simon et al. (2003, 2005, 2007, 2008) and Bell et al. (2009) indicate that the concentration of Au dissolved in volatile-bearing haplogranitic melts varies in a range from 0.02 to 0.65 ppm if micronuggets are excluded from the analysis. For rhyolitic melts, no systematic dependence of Au solubility on the concentrations of H2O, S and Cl was observed in studies based on the bulk determination of Au as well as in the studies based on results obtained from LA. However, it must be noted that the solubilities of S and Cl in haplogranitic and rhyolitic liquids are low at magmatic conditions (e.g., Carroll and Webster, 1994), probably limiting the ability of these volatiles to affect significantly the solubility of Au. On the other hand, the concentrations of Au in silicate melts are extremely low and even small variation in the molar concentration of S and Cl is expected to influence the behaviour of Au. The absence of clearly observed correlations indicates either that Au does not complex with both S and Cl or that detection of such correlation is overwhelmed by the analytical and experimental difficulties. 3. EXPERIMENTS AND ANALYTICS 3.1. Experimental strategy Synthetic glass powders of rhyodacitic and andesitic compositions were charged into Au capsules together with H2O and elemental S or with a HCl-solution. The capsules were run in an internally heated pressure vessel (IHPV) at 1050 °C, 200 MPa and under controlled redox conditions close to the Ni/NiO solid oxygen buffer (NNO) for 5 days. At these conditions, the metallic Au of the capsule walls was equilibrated with fluid-saturated silicate melts. After the experiment, the bubble- and crystal-free glasses were analyzed for major elements, Cl, S and H2O as well as for

2398

R.E. Botcharnikov et al. / Geochimica et Cosmochimica Acta 74 (2010) 2396–2411

Au in order to determine the effect of volatiles on Au solubility in silicate melts. 3.2. Experimental method 3.2.1. Starting compositions The starting materials are synthetic analogues of rhyodacitic and andesitic melts and are reported in Table 1. The same compositions have been already used in several experimental studies on the solubility of volatiles in silicate melts (e.g., Ohlhorst et al., 2001; Sato et al., 2005; Botcharnikov et al., 2004, 2006, 2007). The compositions were prepared from mixtures of oxides (SiO2, TiO2, Al2O3, Fe2O3, MnO, MgO) and carbonates (CaCO3, Na2CO3, K2CO3) ground in a rotary mortar. The mixed powders were melted for two hours in Pt crucibles at 1600 °C, 1 atm in air (log fO2 = 0.68). The melts were quenched to glass by placing the crucibles into a water bath. The glasses were crushed in an agate mortar and melted a second time for 0.5 h following the procedure above to improve homogeneity of the glass compositions. The homogeneity of silicate glasses was verified by electron microprobe (Table 1). The glasses were finally crushed in the mortar and sieved to the grain sizes of <100 lm and <200 lm. Two fractions were mixed together in a ratio 1: 1 to decrease the free volume in the experimental capsules and to minimize the incorporation of atmospheric nitrogen in experimental charges. 3.2.2. Experimental method The glass powders (about 50 mg) were placed together with ca. 10 wt.% of HCl-solution (containing about 2.5, 5 and 10 wt.% Cl) or with elemental S (about 0.25, 0.5, 1, and 2 wt.% bulk S) + 7–8 wt.% of H2O into Au capsules (1.5 mm long, 2.6 mm inner diameter). The initial amount of Cl in HCl solutions was checked using a Mettler DL25 Titrator. One capsule was charged with andesitic glass powder and 9 wt.% of pure H2O (sample A264) while another capsule was loaded with a rhyodacitic glass, ca. 1 wt.% of elemental S and about 5 wt.% of bulk H2O (sample SG38, Table 2). Capsules were welded shut and heated overnight in an oven at 110 °C to homogenize fluid distribution inside the capsule and check for leaks. Such amounts of added volatiles were sufficient to produce a free fluid phase in all experimental runs as verified by weight loss after opening the capsules (and also by strong H2Ssmell in S-bearing experiments). The capsules were run in an internally heated pressure vessel (IHPV) at 1050 °C (note that sample A264 was run 1025 °C), 200 MPa and redox conditions close to that buffered by NNO. The total pressure was measured by a pressure transducer calibrated against a strain gauge manometer with an uncertainty of about 1 MPa. The variations of pressure during the experiments were <5 MPa. Temperature was measured with four unsheathed S-type

(Pt–Pt90Rh10) thermocouples used to control the temperature gradient over a length of 30 mm inside the vessel. Temperature oscillations were less than 5 °C. The redox conditions were adjusted by loading a mixture of Ar and H2 gases in the IHPV. Hydrogen fugacity (fH2) in the experiments was controlled using a Shaw-membrane technique specially designed for the IHPV (Berndt et al., 2002). Within the sample capsule, the fH2 was fixed due to an in-diffusion of hydrogen through the capsule walls. This, in turn, controlled the fugacity of oxygen via the redox reactions inside the capsule. The initial amount of H2 charged in the vessel was chosen to achieve conditions close to NNO, assuming that the redox state in the capsules is mainly controlled by the equilibrium reaction of water formation (H2 + 1/2 O2 M H2O) and that the water activity in the system (aH2O) is close to unity. However, it must be noted that the presence of S in several experiments may have influenced the redox conditions in the capsules as well. Sample SG38 was run at a slightly higher fugacity of hydrogen, resulting in nominally more reduced conditions in the capsule (corresponding approximately to log fO2 = NNO0.6 for the system with aH2O = 1). The duration of most experimental runs was 120 ± 5 h (Table 2). Previous studies have proven that run duration of 2–5 days is enough to achieve equilibrium distribution of volatiles between fluids and silicic melts even at lower temperatures of 800–900 °C and similar pressures (e.g., Metrich and Rutherford, 1992; Signorelli and Carroll, 2000). This is in agreement with the determined diffusivities of volatiles in silicate liquids (e.g., Baker et al., 2005) and with the design of experimental charges, i.e., homogeneous distribution of fluids and rock powder at the beginning of the experiments. In other words, the small grain size of the rock particles (<200 lm) provides short diffusion paths for volatiles, dissolving in the melt during heating and equilibration. To our knowledge, the diffusivity data for Au in silicate melts are not available in the literature. However, the model estimations of Mungall (2002) show that Au diffusivity in mafic melts is about 10 11 to 10 10 m2/s at magmatic conditions, which is similar to the diffusion rates of volatiles. Thus, it is expected that the applied run duration was sufficiently long to attain equilibrium conditions in our experiments. All experimental runs were quenched isobarically by dropping the capsules into the cold zone of the IHPV. It should be emphasized that at the conditions of the experiments, the metallic Au of the capsule walls was equilibrated with fluid-saturated silicate melts, establishing the activity of Au in the system equal to unity. 3.3. Analytical techniques After the experiments, capsules were checked for leaks by weighing and were prepared for analytical investigation of the experimental products. Most products were com-

Table 1 Starting composition of synthetic andesite and rhyodacite (normalized to 100 wt.%). SiO2

TiO2

Al2O3

FeOtot

MnO

MgO

CaO

Na2O

K2O

Total

57.44 70.13

1.06 0.50

17.53 14.25

7.20 3.58

0.12 0.12

4.31 1.44

7.42 4.06

3.32 3.17

1.61 2.76

100.00 100.00

Au solubility in volatile-bearing magmas

3.3.1. Electron microprobe Glass fragments of each sample were mounted in epoxy for electron microprobe analysis. The analyses of the experimental glasses were performed with a Cameca SX100 electron microprobe equipped with 5 spectrometers. All data were obtained using 15 kV acceleration potential, 4–6 nA beam current, a defocused electron beam (10-lm diameter) and peak counting times for major elements of 5–10 s. In order to minimize the migration of the mobile elements (Na, K), alkalis were analyzed first, and no significant alkali loss was observed (see also Stelling et al., 2008). For the sulfur analyses, the exact wavelength of the SKa radiation was estimated in silicate glasses (Carroll and Rutherford, 1988), and peak position was compared with the standards (BaSO4 and ZnS). The SKa peak position in the glasses was close to that of sulfur bound as sulfide. Chlorine and sulfur were measured as the last elements with 30 nA beam current and the counting times of 100–200 s and of 60–100 s for S and Cl, respectively. Multiple measurements were made for each sample (7–30 analyses) to reduce possible analytical errors and to check for homogeneity (Table 2). 3.3.2. Determination of H2O content of the glass The concentrations of H2O in experimental glasses were determined by Karl–Fischer titration (KFT) and by NIR spectroscopy as reported in Table 3. The KFT method provides absolute mass of H2O in the analyzed sample that can be directly translated in H2O concentrations, knowing the weight of the sample. For the determination of H2O by NIR spectroscopy, doubly polished glass plates of 250 lm thickness were analyzed with a Bruker IFS88 spectrometer coupled with IR-Scope II microscope (MCT narrow range detector; tungsten lamp and CaF2 beamsplitter; spectral resolution of 4 cm 1). Typically 100 scans were used for analyses of background and glasses with a spot size of approximately 100  100 lm. Thickness of the glass plates was measured with a digital micrometer (Mitutoyo) with a precision ±2 lm. Densities of hydrous andesitic glasses were estimated from the equation proposed by Ohlhorst et al. (2001) for the same andesitic composition (q = 18.4
CH2Otot + 2661), where q is a density in g/L, CH2Otot is the total water content of the glass in wt.%). Densities of the H2O-bearing rhyodacitic glasses were estimated assuming an average value between the density of dacite (q = 11.8
CH2Otot + 2515) after Ohlhorst et al. (2001) and that of rhyolite (q = 14.7
CH2Otot + 2367)

after Tamic et al. (2001), containing the same amount of H2O as in the investigated samples. The concentrations of water dissolved in the glasses as molecular H2O (H2Omol) and as OH groups were determined from the heights of the baseline-corrected absorption bands at 5230 and 4520 cm 1, respectively. In the calculation of water concentration in andesitic glasses for H2Omol and OH peaks we used a tangential baseline correction and corresponding absorption coefficients of 0.86 and 0.68 L mol cm 1 as calibrated by Ohlhorst et al. (2001). For rhyodacitic glasses, the absorption coefficients were not specifically calibrated but Ohlhorst et al. (2001) proposed an empirical equation where coefficients for both peaks can be calculated as a function of glass composition (SiO2). The calculated values are 1.3 and 1.1 L mol cm 1 for H2Omol and OH peaks, respectively. It should be noted that a comparison of the data from NIR measurements with the data obtained by KFT for the same samples shows a significant systematic overestimation of water contents by NIR (Fig. 1). The reason for such overestimation is not clear and it is presumably related to the calibration of absorption coefficients. Since KFT provides absolute concentration values and does not require calibration, the KFT data can be used as reference values. Hence, to have a reliable data from NIR, the determined H2O concentrations have to be multiplied by a factor of 0.9, according to the KFT measurements (see Table 3). Since almost all samples (in several cases we did not have enough material for the KFT measurements) were analysed with KFT, we use the KFT data for the interpretation of our experimental results. Unfortunately, the range of water concentrations in our experiments is relatively narrow and we can not provide additional reliable calibration of the absorption coefficients for NIR. In the case of rhyodacite, the concentrations of H2O determined by NIR show a good agreement with the KFT data (except sample S45 with low H2O content) and no further correction is needed (see Fig. 1 and Table 3).

7 1:1

6

H2O NIR, wt%

posed of quenched crystal- and bubble-free silicate glasses and fluids while two samples (S110 and SG38) contained also FeS-bearing phase. The rounded shape of FeS blebs indicates that this phase was a sulphide liquid during the experiment. Several methods were used to analyze the composition of the glasses: (i) electron microprobe for the concentrations of major oxides as well as for Cl and S; (ii) Karl–Fischer titration (KFT) and Fournier-transform near-Infrared (NIR) spectroscopy for the determination of H2O contents; (iii) colorimetric wet-chemical method for the analysis of the redox state of Fe in the quenched glasses; and (iv) Laser Ablation ICP-MS (LA) for the measurements of Au concentrations.

2399

5

4

Rhyodacite Andesite

3 3

4

5

6

7

H2O KFT, wt%

Fig. 1. Comparison between H2O concentrations of run-product glasses determined by KFT and NIR spectroscopy. Note systematic overestimation of H2O contents by NIR for andesitic glasses.

R.E. Botcharnikov et al. / Geochimica et Cosmochimica Acta 74 (2010) 2396–2411

3.3.3. Wet-chemical colorimetric determination of ferrous Fe in glasses The analytical approach for the determination of ferrous Fe in experimental silicate glasses is based on the analytical technique of Wilson (1960) modified by Schuessler et al. (2008) to minimize the use of toxic materials. The analytical method was optimized to measure the Fe2+/RFe ratio of milligram-sized samples. The accuracy and precision were tested with international reference materials and with standards analyzed by other methods. The replicate measurements for different rocks and minerals containing between 1 and 8 wt.% ferrous Fe showed that the precision of Fe2+/RFe determination using this technique is within ±0.03 (2r). Based on the calibration, it was concluded that this method can be reliably applied for silicate glasses with Fe2+/RFe ratios in the range from 0.4 to 0.9 (Schuessler et al., 2008). Three to eleven milligrams of glasses from our samples were measured, applying this technique. The 2r errors of the Fe2+/RFe determination include reproducibility and analytical uncertainty of the method and vary from ±0.03 to 0.11 (Table 2).

10 9

#S45 Sm = 190 ppmw

(a) micronuggets

8 7 6

Au (ppm)

2400

5 4 average values

3 2

area 2 area 1

1 0 60000

80000

100000

120000

140000

160000

180000

time (ms) s 10.0 9.0 8.0

(b)

#Cl85 Clm = 0.48 wt%

Au (ppm)

7.0

3.3.4. Au analysis by Laser-Ablation ICP-MS (LA) Analyses of Au (197Au) in the glasses were performed using a laser ablation technique (coupled with an MCICP-MS spectrometer; LA-ICP-MS) in the Great Lakes Institute for Environmental Research, University of Windsor, Canada. The glass samples were ablated using a Quantronix Integra-C Fs laser system at the fundamental wavelength of 785 nm. It is a regenerative and multi-pass Titanium-doped sapphire (Ti:sapphire) laser ablation system based on the Chirped Pulse Amplification (CPA) technique. The LA system operating conditions were a 50 Hz repetition rate, 0.36 mJ/ pulse energy, pin hole diameters of 2.0, 3.0 or 3.5 mm and a 10 objective lens which resulted in a 65 lm beam width in NIST610. The ablation material was transported from the ablation cell to the ICP-MS in argon as carrier gas and analyzed using a Thermo Electron X series II quadrupole ICP-MS. The data acquisition protocol was time-resolved analysis, the scanning mode was peak-jumping mode, dwell times were 10 ms for 29Si, 43Ca, 44Ca, and 57Fe; 20 ms for 33S and 34 S; and 50 ms for 197Au. 33S, 43Ca, 44Ca, and 57Fe were analysed at standard resolution and 29Si, 34S, and 197Au at high resolution. The ICP-MS analyses were calibrated using a NIST610 glass standard and Ca concentration from electron microprobe analysis was used as an internal standard. Data reduction was performed off-line using commercial software and in-house written program based on Longrich et al. (1996). Fig. 2 shows the typical time-resolved spectra for glasses containing Au. The absolute concentration of Au in the glass was calculated based on the known concentration of Ca measured by electron microprobe and LA in the same sample and in the reference NIST610 glass. All investigated samples show evidence for the presence of micronuggets of Au metal (sharp spikes in the spectra in Fig. 2). The relatively high time resolution of the LA analyses allowed identification of micronuggets that can be relatively well filtered out in order to define the “flat” regions of the spectra and,

6.0 5.0 4.0 3.0 2.0

0.0 60000

area 2

area 1

1.0 80000

100000

120000

140000

160000

180000

time (ms) s

Fig. 2. Representative LA spectra for the determination of Au concentration in quenched glasses for S- (a) and Cl-bearing (b) systems (see Tables 1–3 for other details). Large peaks are correspondent to the ablation of Au micronuggets. Outlined areas show the filter-out technique applied to minimize the influence of micronuggets. Thick black lines represent average Au concentrations within the outlined areas. Typically two to three areas were analysed to check for reproducibility and analytical uncertainty within each area and between the areas on the sample. Runproducts numbers shown are samples S45 and Cl85; S and Cl concentrations of glasses are reported in figures.

hence, to determine the reliable Au concentration in the glass (as illustrated in Fig. 2). The analytical errors were calculated from the reproducibility of data within and between the selected areas. It must be noted that nuggets did not show any systematic spatial distribution in the analysed samples. The amount of micronuggets is not directly related with the composition of the silicate melt or with the amount of dissolved volatiles but it seems that the Cl-bearing silicate melts contain higher proportion of micronuggets than the S-bearing melts, presumably indicating higher solubility of Au in Cl-bearing fluids (see discussion on the formation of micronuggets below, Section 5.1). In addition, the bulk concentrations of Au in the samples were determined by standard ICP-MS analyses (coop-

Au solubility in volatile-bearing magmas

eration with R. Schoenberg at Hannover). The HNO3 solutions with dissolved glass pieces (preliminary digested in HF) from the samples were analysed for 197Au. The analysis showed that the concentrations of Au in starting glasses do not exceed 0.17 ppm. The presence of Au in starting materials presumably indicates initial contamination during glass synthesis due to use of Pt crucibles or due to impurities in the chemical reagents. The bulk concentration of Au in the experimental samples varied in the range from 3 to 26 ppm Au and showed no systematics as a function of melt or system composition. The obtained range of concentrations is up to one order of magnitude higher than the concentrations determined by LA technique, emphasising the dramatic contribution of micronuggets to the bulk Au concentration in the experimental charge. 4. RESULTS 4.1. Major-element composition of glasses The glass compositions for the samples Cl84–Cl86 have been already reported by Botcharnikov et al. (2007) while the composition of glass from the run A264 was shown by Botcharnikov et al. (2008). Here, we present them to have a consistent experimental dataset and to provide some additional information (e.g., Fe2+/RFe ratios). In general, the measured compositions of experimental glasses are similar to the starting materials as listed in Table 2. It must be noted that the concentrations of Fe are slightly lower in samples S110 and SG38 because of segregation of FeSbearing phase/sulphide liquid (FeSm). The presence of FeS phases, typically occurring as rounded blebs, was detected by back-scattered electron (BSE) images and the composition was identified by electron microprobe analysis. The S110 sample contains only few but large FeS blebs with a size up to 30–40 lm while the typical size of FeS blebs in SG38 was less than 10 lm as was detected using BSE images. It was difficult to estimate the modal abundance of sulphide blebs from the BSE images because blebs were heterogeneously distributed in the samples (very often located close to the capsule wall). However, the mass-balance analysis shows that the Fe-loss from the silicate melt into sulphide phase was of about 5–10 and about 40 relative % in S110 and SG38 samples, respectively. No considerable changes for the other major elements have been observed, indicating that no significant partitioning of major elements from the melt into the fluid occurred. 4.2. Concentrations of volatiles (Cl, S and H2O) in glasses The concentrations of Cl, S and H2O in the glasses are reported in Tables 2 and 3. The concentrations of Cl vary in the range from ca. 0.2 to 1.0 wt.% Cl in both andesitic and rhyodacitic glasses. Considering the analysed Cl concentrations in glasses as well as the total amounts of Cl added in both systems, only a single-fluid phase (low-density vapour) should be present at the studied temperature (T) and pressure (P), according to the experimental data on phase relations in water and salt-bearing systems (see review of Webster and Mandeville, 2007). The distribution of

2401

Cl between andesitic melt and coexisting single-fluid phase (hereafter vapour) is discussed in detail by Botcharnikov et al. (2007). The information about the partitioning of Cl and S between phases in rhyodacitic systems at 850 °C, 200 MPa can be found in Botcharnikov et al. (2004). Here, we emphasize that the Cl concentration in silicate melt increases non-linearly with increasing Cl content in the coexisting vapour at T–P conditions of this study due to strong non-ideality of mixing in Cl-bearing fluids (see Fig. 4 in Botcharnikov et al. (2007) and the review of Webster and Mandeville (2007) for details). As a consequence, an increase in bulk Cl content of the system up to 1.1 wt.% and, hence, Cl concentration in the melt from 0 to 1 wt.% has only minor effect on the concentrations of dissolved H2O (Fig. 3a). The water content in andesitic melts increases slightly but systematically with increasing Cl concentration which is generally consistent with observations for basaltic melts reported by Stelling et al. (2008). The variations in H2O contents of rhyodacitic melts do not show any systematics as a function of Cl. The concentrations of dissolved S in glasses vary between 20 and 410 ppm by weight (Table 2) but the concentration of S at FeSm saturation is about 100 ppm higher in andesite than in rhyodacite (410 vs. 330 ppm, respectively). The observed difference can be attributed to the compositional dependence of S solubility in silicate melts (see e.g., Liu et al. 2007). The concentration of S continuously increases with increasing bulk S content in the system as illustrated in Fig. 4. Remarkable is that S content of the glass strongly depends on the bulk proportions of S and H2O in the system (see Table 2). In Fig. 4, sample SG38 has significantly higher S concentration at the same amount of bulk S but at two times lower H2O content, indicating that proportions of S and H2O in the free fluid have an influence on volatile distribution. In other words, fluid composition exerts indeed a firm control on volatile activities in the system. Furthermore, the concentrations of dissolved H2O strongly decrease with increasing bulk amount of S. Both andesitic and rhyodacitic systems illustrate almost linear relationships between melt concentrations of H2O and S (Fig. 3b). Such relationships differ strongly from the observations made for the Cl-bearing systems (Fig. 3a). A simple mass-balance analysis for the distribution of volatiles between fluid and melt phases was performed based on initial masses of added volatiles and on measured concentrations of volatiles in the glasses. The results of calculations indicate that the concentrations of both Cl and S in the melts increase with increasing Cl/H2O and S/H2O in the coexisting fluids (mass ratios) as shown in Fig. 5. It must be noted that the S/H2O ratio of the fluids (Fig. 5a) is about one order of magnitude higher than the Cl/H2O ratio (Fig. 5b). However, the analysis of these diagrams is difficult because both Cl and S can be present as various complexes (e.g., NaCl, CaCl2, HCl or hydrated Cl-bearing species; S2, H2S, SO2, etc.) in the fluid phase and the occurrences and abundances of such complexes are not known. Moreover, S is partly consumed by the segregated FeSm and S can also react with the capsule walls. Thus, the calculated proportions or mass ratios provide maximum estimates for the composition of the fluids, especially in

2402

Time, h

H2O ini, wt.%a

Cl ini, wt.%b

S ini, c, wt.%

n*

SiO2

46 120 120 120 120 120 120 120

8.97 8.94 8.77 8.51 7.30 7.47 7.45 9.13

— 0.26 0.52 1.07 — — — —

— — — — 0.23 0.48 0.96 1.88

7 30 20 25 20 20 16 10

54.21 53.60 53.58 53.36 53.81 55.08 53.89 56.52

(39) (26) (20) (24) (20) (18) (16) (28)

0.98 0.96 0.97 0.95 0.98 1.00 0.99 1.00

(4) (3) (4) (4) (2) (1) (1) (3)

16.69 16.18 16.09 16.02 16.07 16.55 16.15 17.17

Rhyodacite Cl81 120 Cl82 120 Cl83 120 S43 120 S44 120 S45 120 120 SG38e

8.81 8.69 8.43 7.24 7.27 7.32 4.74

0.25 0.51 1.06 — — — —

— — — 0.22 0.48 0.96 1.00

20 25 25 16 13 11 10

65.58 65.42 65.18 66.03 66.21 67.37 67.26

(30) (28) (25) (20) (14) (21) (49)

0.46 0.47 0.45 0.48 0.49 0.49 0.47

(2) (2) (3) (1) (1) (1) (3)

13.32 13.33 13.29 13.49 13.48 13.80 13.27

Sample Andesite A264d Cl84 Cl85 Cl86 S46 S47 S48 S110e

*

TiO2

FeOtot

MnO

MgO

CaO

(20) (13) (12) (15) (7) (8) (13) (23)

6.63 6.95 7.06 6.90 6.95 6.82 6.83 6.16

(32) (20) (32) (27) (15) (13) (7) (33)

0.15 0.10 0.11 0.09 0.10 0.11 0.12 0.09

(12) (3) (4) (5) (3) (2) (3) (9)

4.05 4.00 3.94 3.96 4.03 4.00 4.02 4.22

(13) (8) (8) (7) (7) (6) (5) (17)

6.91 7.00 7.00 6.93 7.00 7.21 7.01 7.46

(11) (10) (18) (10) (11) (10) (15)

3.34 3.35 3.27 3.21 3.21 2.93 1.88

(21) (17) (18) (11) (12) (11) (12)

0.11 0.11 0.11 0.13 0.12 0.11 0.01

(3) (3) (4) (3) (3) (2) (7)

1.35 1.34 1.35 1.37 1.38 1.41 1.32

(6) (5) (6) (4) (4) (3) (5)

3.80 3.74 3.76 3.81 3.84 3.89 3.73

Al2O3

Na2O

K2O

(23) (16) (14) (12) (9) (9) (9) (23)

2.82 3.03 3.00 2.97 3.00 3.14 3.01 3.17

(22) (16) (12) (14) (10) (10) (9) (27)

1.50 1.55 1.54 1.54 1.51 1.60 1.52 1.61

(9) (11) (10) (7) (6) (6) (15)

2.97 2.93 2.88 2.86 2.95 3.12 3.00

(10) (13) (14) (6) (13) (12) (17)

2.58 2.58 2.55 2.33 2.40 2.60 2.44

n, number of electron microprobe analyses of glasses for major elements. Initial amounts of added H2O in the starting charge (in Cl-bearing series it was calculated from the concentration of H2O in HCl solution). b Initial amount of added Cl in the starting charge (calculated from the concentration of Cl in HCl solution). c Initial amount of added S in the starting charge (volatiles + glass). d Experiment was conducted at 1025 °C as reported by Botcharnikov et al. (2008). e Samples S110 and SG38 contain sulfide-melt phase (regarded as FeSm). a

Cl

S

Total

Fe2+/ RFe

(6) (5) (8) (5) (5) (6) (3) (5)

— 0.24 (1) 0.48 (1) 0.97 (1) — — — —

— — — — 0.004 0.009 0.022 0.041

(2) (3) (3) (4)

93.95 93.61 93.78 93.70 93.47 95.54 93.56 97.44

0.84 0.78 0.85 0.73 0.83 0.84 0.86 0.88

(3) (8) (3) (11) (4) (3) (3) (3)

(8) (8) (8) (6) (7) (8) (9)

0.23 (1) 0.46 (1) 0.90 (3) — — — —

— — — 0.002 0.004 0.019 0.033

(1) (1) (3) (4)

93.74 93.74 93.75 93.71 94.08 95.75 93.41

0.80 0.80 0.81 0.90 0.81 0.90 0.82

(6) (6) (6) (3) (5) (3) (6)

R.E. Botcharnikov et al. / Geochimica et Cosmochimica Acta 74 (2010) 2396–2411

Table 2 The run conditions and glass compositions (in wt.%) from the 200 MPa and 1050 °C experiments (1r error).

Table 3 Concentrations of volatiles and Au in the experimental glasses (1r error), calculated fluid ratios and apparent activities of Cl, S and H2O. H2O KFT, wt.%

H2Omol NIR wt.%a

OH NIR wt.%b

H2O total NIR wt.%c

H2O melt corrected wt.%d

Cl/H2O fluid wt.%e

S/H2O fluid wt.%e

aCl*

Andesite A264c Cl84 Cl85 Cl86 S46 S47 S48 S110

n.a. 5.75 5.86 5.97 5.75 5.69 3.92 n.a.

3.97 4.03 4.10 4.20 3.77 2.16 3.79 1.54

(3) (3) (2) (2) (3) (3) (2) (2)

2.51 2.52 2.53 2.56 2.45 2.21 2.52 2.00

(3) (4) (2) (2) (3) (2) (2) (3)

6.47 6.55 6.63 6.76 6.22 6.32 4.37 3.54

(4) (5) (3) (3) (4) (4) (3) (3)

5.82 (4) — — — — — — 3.19 (3)

— 0.008 0.018 0.047 — — — —

— — — — 0.071 0.120 0.118 0.124

— 0.10 0.20 0.41 — — — —

— — — — 0.10 0.23 0.56 1.00

1.00 0.99 1.01 1.03 0.99 0.98 0.67 0.55

0.32 0.29 0.28 0.71 0.45 0.64 0.77 2.47

(17) (16) (13) (28) (20) (22) (26) (42)

Rhyodacite Cl81 Cl82 Cl83 S43 S44 S45 SG38

6.02 5.72 5.90 5.55 5.53 3.95 n.a.

3.81 3.90 3.92 3.67 3.38 1.88 0.32

(2) (2) (2) (2) (2) (5) (6)

1.96 1.99 2.00 1.96 1.95 1.70 0.96

(3) (2) (3) (3) (2) (6) (5)

5.78 5.89 5.92 5.63 5.33 3.58 1.28

(4) (3) (4) (4) (3) (12) (11)

— — — — — — 1.28 (11)

0.011 0.023 0.074 — — — —

— — — 0.062 0.123 0.124 0.125

0.16 0.33 0.64 — — — —

— — — 0.08 0.13 0.63 1.00

1.00 0.92 0.98 0.94 0.89 0.60 0.21

0.33 0.48 0.86 0.24 0.43 0.83 1.18

(15) (18) (34) (12) (19) (26) (47)

Sample

(10) (7) (4) (10) (10) (9)

aS*

f

aH2O*

f

Au, ppmwg

Au solubility in volatile-bearing magmas

(10) (9) (10) (9) (14) (7)

f

n.a., not analysed. a Concentration of molecular H2O in quenched glasses as determined by NIR spectroscopy. b Concentration of OH in quenched glasses as determined by NIR spectroscopy. c Bulk water concentration determined by summing H2Omol and OH concentration from NIR measurements. d Corrected concentration for andesite samples according to the KFT analyses (H2O total by NIR multiplied by 0.9); KFT data or corrected H2O values were used in the paper (except sample SG38 for which no KFT analysis was available). No correction for unextracted water was done (see Botcharnikov et al. 2007). e Mass ratios determined by simple mass balance (see text). f Apparent activities of volatiles (see text for details). g The 1r errors of Au concentration include the accuracy based on the LA analysis of the NIST610 reference glass, precision obtained from the variations of the LA signal and reproducibility of the nugget-filtered LA spectra (see Fig. 2).

2403

2404

R.E. Botcharnikov et al. / Geochimica et Cosmochimica Acta 74 (2010) 2396–2411 1.2

450 Andesite S

400

Rhyodacite Cl

1

S in melt, ppm

Cl in melt, wt%

Rhyodacite S

350

Andesite Cl

0.8 0.6 0.4

FeSm

300

FeSm

250 200 150 100

0.2

50

(a)

(a)

0

0 0

1

2

3

4

5

6

0

7

0.1

0.2

0.3

S / H2O fluid, mass

450 400

Rhyodacite S

350

(b)

Andesite S

0.8

Cl in melt, wt.%

FeSm

300

S in melt, ppm

1

FeSm

250 200 150

0.6

0.4 Andesite Cl

100

0.2

50

Rhyodacite Cl

(b) 0

0 0

1

2

3

4

5

6

7

0

0.02

0.04

0.06

0.08

0.1

Cl / H2O fluid, mass

H2O in melt, wt%

Fig. 3. Relationships between dissolved concentrations of H2O and S (a) as well as between ones of H2O and Cl (b) in silicate melts. Note significant differences between behaviour of volatiles in these two systems.

Fig. 5. The dependences of S and Cl concentrations in the melt on the mass proportions of S and H2O (a) and Cl and H2O (b) in the fluids. Arrows in (a) show possible shift in S/H2O ratio to lower values due to partial loss of S from melt and fluid into FeSm phases in samples SG38 and S110 (these phases were not considered in mass-balance calculations).

450 FeSm saturation in Andesite

400

S in melt, ppm

350

FeSm saturation in Rhyodacite

300 250

based on the measured concentrations of volatiles in glasses and will be focused on the assumed apparent activities of volatiles (see below). 4.3. The redox state of the glasses based on Fe2+/RFe ratios

4.7 wt.% bulk H2O

200 150

Andesite

100

Rhyodacite

50 0 0

0.5

1

1.5

2

2.5

S added, bulk wt%

Fig. 4. Dependence of S concentration in the melt on the bulk amount of S charged in the system. All samples have about 7–9 wt.% bulk H2O while sample SG38 has only 5 wt.% bulk H2O, resulting in higher S concentration in the melt S at a given bulk S content.

S-bearing systems. Furthermore, the simplified mass balance can not be directly used to constrain activities of volatile components and our following discussion will be

The measured Fe2+/RFe ratios of the experimental glasses are reported in Table 2. The determined values vary from 0.73 to 0.85 in Cl-bearing glasses and from 0.81 to 0.90 in S-bearing glasses (Fig. 6). Taking the analytical and experimental uncertainties into account, the differences are within error and an average value is close to 0.80–0.85, even for the samples S110 and SG38 where FeSm is stable. Despite the fact that the nominal redox conditions were slightly more reduced in SG38 run, the measured proportion of Fe2+ is similar to the other experiments (Fig. 6). The determined Fe2+/RFe values correspond to relatively reduced conditions close to the NNO oxygen buffer, and it is in agreement with experimental observations of Fe2+/ RFe vs. log fO2 relationships for other hydrous melt compositions (e.g., Botcharnikov et al., 2005; Schuessler et al., 2008). This implies that, although the concentrations of

Au solubility in volatile-bearing magmas 500 450

2.8

Rhyodacite S

2.4

300

FeSm

250 200 150 100 50

Rhyodacite S Andesite S

FeSm

350

Au in melt, ppm

S in melt, ppm

400

Andesite S

2405

FeSm

2.0 1.6

FeSm 1.2 0.8

(a) 0.4

0

(a)

0.0

1.2

0

0.01

0.02

0.03

0.04

0.05

S in melt, wt%

1.6

0.8

Rhyodacite Cl

Andesite Cl

0.6

1.4

0.4 0.2

(b) 0

Andesite Cl

Rhyodacite Cl

Au in melt, ppm

Cl in melt, wt%

1

0.5

0.6

0.7

0.8

0.9

1

Fe2+ / Fe total

Fig. 6. The Fe2+/Fe vs. S (a) and Cl (b) concentrations in the melt. Note the almost constant Fe redox ratio, indicating similar redox conditions in the experiments.

Cl, S and H2O in the fluid phase may vary in a wide range, the multicomponent fluid composition has no significant influence on the prevailing redox conditions in the capsules. It may also indicate that the speciation and fugacities of fluid components and consequently the concentrations in the silicate melt are mainly controlled by externally adjusted fugacity of hydrogen. 4.4. The solubility of Au in volatile-bearing melts The concentrations of Au in experimental glasses vary in the range from 0.2 to 2.5 ppm by weight as reported in Table 3 (the uncertainties are 1 sigma). The relationships between measured Au concentrations and volatile content of the glasses are plotted in Fig. 7. The solubility of Au in Cl- and S-free andesitic melt is about 0.3 ppm and it systematically increases with increasing concentrations of both S and Cl in the melt at given redox conditions (Fig. 7a and b). The effect of melt composition on Au solubility can not be resolved on the basis of the obtained data since both andesitic and rhyodacitic melts show similar Au contents within the uncertainty at a given S or Cl concentration. The error on Au determination increases with the Au concentration due to increasing amount of Au micronuggets and background noise in the LA spectra of the glass samples. Despite the large errors for Au, dissolved S and Cl show different effects on the Au metal/silicate melt parti-

1.2 1 0.8 0.6 0.4 0.2

(b)

0 0

0.2

0.4

0.6

0.8

1

1.2

Cl in melt, wt%

Fig. 7. Dependence of Au solubility in the rhyodacitic and andesitic melts on the concentrations of dissolved S (a) and Cl (b).

tioning. Although the solubility of Au increases in both cases, small amounts of dissolved S (ca. 100 ppm) produce an increase in Au concentrations by a factor of 2–3, while the influence of Cl is detectable only at Cl content >6000 ppm in the melt. 5. DISCUSSION 5.1. Formation of Au micronuggets in solubility experiments The formation of metal nuggets in the experimental products, in particular in silicate glasses, is a usual problem affecting the quality of metal-solubility data. It was proposed that Au metal nuggets are formed during quench when silicate melt becomes oversaturated with respect to metal with decreasing temperature (e.g., Frank et al., 2002; Simon et al., 2007). Thus, it was assumed that the Au micronuggets represent a quench phase and must be included in the analysis of the bulk concentration of dissolved Au at the experimental conditions (Frank et al., 2002; Simon et al., 2007). Moreover, it was concluded that the Au concentration data from bulk-analytical techniques such as INAA are comparable with ones obtained by the micro-analytical methods like SIMS and LA, providing reasonable estimates for Au solubility (Simon et al., 2007).

2406

R.E. Botcharnikov et al. / Geochimica et Cosmochimica Acta 74 (2010) 2396–2411

However, direct experimental evidence for the nucleation and growth of Au nuggets with decreasing temperature is not available up to date. It must be noted that the LA and bulk ICP-MS data for our samples show a dramatic difference in measured Au concentrations and absence of any systematics as a function of any parameter, indicating that bulk methods are not appropriate for the analysis of experimental glasses. Alternatively, gold nuggets can be an experimental artefact in volatile-bearing systems. Usually, the starting assemblage placed into the capsules is composed of glass powder and fluid source(s). At the onset of heating, the fluid phase will be quickly distributed between the grains of solid glass powder. According to the experimental studies on Au solubility in magmatic and hydrothermal fluids (e.g., Zotov et al., 1991; Archibald et al., 2001; Frank et al., 2002; Stefansson and Seward, 2003a,b, 2004; Hanley et al., 2005; Simon et al., 2005, 2007; Zezin et al., 2007; Pokrovski et al., 2009a,b), considerable amounts of Au can be dissolved in Cl- and S-bearing fluids already in early stages of the experiment, i.e., at relatively low temperatures during heating. Furthermore, the solubility of Au in the fluid phase increases with increasing temperature. In other words, experimental fluids may contain significant amounts of Au before the beginning of melting of the glass powder at high pressures and temperatures of the experiment. When silicate powder starts to melt, the Au-bearing fluids dissolve progressively in the melt. Taking into account that the solubility of Au in fluids is significantly higher than that in silicate melts (e.g., Frank et al., 2002; Simon et al., 2005, 2007), the excess Au, which was initially dissolved in the fluid, is expected to form small aggregates or micronuggets after partial dissolution of the fluid phase in the melt. This explanation for the formation of micronuggets is also supported by higher proportion of micronuggets in Cl-bearing glasses when compared with the S-bearing ones (Fig. 2) which can be attributed to higher solubility of Au in Cl-bearing fluids. Moreover, Frank et al. (2002) showed that different initial porosity of the starting silicate glass powder may have an effect on the amount of fluid bubbles and Au micronuggets in the quenched haplogranitic glasses, supporting our interpretation of micronugget formation in fluid-bearing experimental systems. Hence, the application of bulk-analytical methods for the determination of Au concentrations in experimental glasses is expected to lead to an overestimation of Au solubility. Although LA can be used for quantitative analysis, the presence of Au micronuggets results in large uncertainties. Thus, alternative or modified experimental approaches are required for accurate determination of Au partitioning in magmatic assemblages. For instance, instead of using dry glasses and volatiles as starting materials, volatile-bearing silicate melts should be synthesized in Au-free containers, avoiding contamination by micronuggets. The obtained quenched glasses can be used for further experiments. 5.2. Behaviour of H2O, S and Cl in rhyodacitic and andesitic systems In order to discuss the effect of S and Cl on the partitioning of Au between metal and silicate melt phases, it is useful

to evaluate the activities of S and Cl in the charges. Considering that we performed fluid-present experiments, the effect of S and Cl on solubilities of Cl-H2O and S-H2O fluids in the melts should be also taken into account. The observed small increase in H2O content of the andesitic melts with increasing amount of Cl can be probably attributed to the changes in the activity coefficient of H2O in the melt due to complexing with Cl as discussed by Stelling et al. (2008). Moreover, the absence of any significant negative effect of Cl on the dissolved H2O is a consequence of the strong non-ideal mixing in Cl–H2O-bearing fluids, resulting in enhanced activity coefficients of fluid components. Similar Cl and H2O concentration values obtained for rhyodacitic melt show that Cl and H2O partitioning is mainly controlled by the properties of the fluid phase and to a smaller extent by the compositional change of the melt from andesite to rhyodacite. The maximum amounts of Cl that can be dissolved in andesitic and rhyodacitic melts at the given experimental conditions are about 2.4 and 1.4 wt.% Cl, respectively, as calculated by the model of Webster and De Vivo (2002). The experimental data for andesites show consistent values of about 2.2–2.5 wt.% Cl (Webster et al., 1999; Webster and De Vivo, 2002; Botcharnikov et al., 2007). Moreover, the data indicate that the maximum Cl content of the melt can be reached if a silicate melt coexists with two immiscible fluids, i.e., with a low-density vapour and a Cl-rich brine. Since there are no direct experimental determinations for the relationships between Cl content of the silicate melt and Cl activity in the system, a term of apparent activity can be proposed, assuming that (1) Cl concentration in the melt is a linear function of Cl activity for the equilibrium silicate melt-vapour until the system is not saturated with respect of a brine phase and (2) since brine coexists with a vapour phase the apparent activity value for Cl (aCl*) is close to unity. Note that both assumptions are valid only for the T–P conditions at which fluid immiscibility is expected as Cl content of the system increases. The data on Cl partitioning between andesitic melts and coexisting fluids at the P–T conditions of our experiments indicate that fluids can reach subcritical state (Botcharnikov et al., 2007). Thus, using this approach, the Cl concentration values in the melts obtained in this study can be recalculated in terms of apparent activity of Cl. Assuming that the maximum Cl concentrations (2.4 and 1.4 wt.% for andesite and rhyodacite, respectively) correspond to the conditions at which aCl* = 1, the apparent activity values can be calculated by a simple equation: aCl* = Clm /Clmmax, where Clm is the Cl concentration in the melt from an individual experiment while Clmmax is the concentration in the melt at the saturation with Cl-rich brine. The calculated aCl* values vary in the range from 0.10 to 0.41 and from 0.16 to 0.64 for andesitic and rhyodacitic systems, respectively as reported in Table 3. In the case of S- and H2O-bearing systems, the relationship between the concentrations of dissolved S and H2O in the melt shows a negative trend for both andesite and rhyodacite, in contrast to the relationships between dissolved Cl and H2O. The observed decrease in water concentration with increasing S can be attributed to the increasing

Au solubility in volatile-bearing magmas

proportion of S in the system, leading to a reaction between volatile components which produced different S-, H-, Obearing species in the fluid. This relationship between H2O and S is very similar to the solubility behaviour of H2O and CO2-bearing fluids, as shown in Fig. 8 for rhyolitic and andesitic melts. This observation indicates that in a first approximation the behaviour of H2O- and S-bearing fluids is close to ideal, as for the H2O- and CO2-bearing fluids. Since the exact composition of the S- and H2O-bearing fluid phase is not known, we follow a similar approach as it was done for Cl- and H2O-bearing systems for defining an apparent activity of S. Sulphur is dissolved mainly as a S2 species in silicate melts at redox conditions close to NNO (e.g., Carroll and Rutherford, 1988) and the predominant species are most probably FeS complexes. Hence, we assume that the apparent S activity (aS*) represents the activity of FeS complexes and is equal to unity in FeSmsaturated silicate melts. Assuming in a first approximation that S content of the melt is a direct manifestation of S activity in the system, i.e., that S solubility is linearly dependent on S activity, the apparent activity values can be calculated by a simple equation: aS* = Sm/SFeSm , where Sm is the actual concentration of S in the melt and SFeSm is S content at the saturation of the melt with FeSm phase. The calculated apparent activities of S are reported in Table 3. The apparent activity of water (aH2O*) was calculated, assuming that water activity (aH2O) in the experiments without added Cl and S is equal to unity and assuming a linear relationship between aH2O in the system and water concentration in the melt. Strictly, this is not the case, as illustrated by the experiments on solubility of H2O-and CO2-bearing fluids. However, the application of the model of Burnham (1979) results in an underestimation of the H2O solubility values for both andesite and rhyodacite, resulting in unrealistically high aH2O (>1). Hence, considering that the values of aS* and aCl* are based on the assumptions of linear activity–composition relationships, we preferred to use this simple linear activity–composition

2500 Rhyodacite

H2O-CO2 in Andesite at 200 MPa, 1100-1200°C

Andesite

S or CO 2 in melt, ppm

2000

Andesite; B2006 Rhyolite; T2001

1500

1000 H2O-CO2 in Rhyolite at 200 MPa, 1100°C

500

FeSm FeSm

0 0

1

2

3

4

5

6

7

H2O in melt, wt%

Fig. 8. A comparison between solubilities of S- and H2O-bearing fluids from this study with the data on the solubility of H2O- and CO2-bearing fluids in rhyolite (Tamic et al., 2001; T2001) and andesite (Botcharnikov et al., 2006; B2006).

2407

relationship for water also in the further discussion. The concentration of water in Cl- and S-free andesite was measured to be 5.8 wt.% while the H2O concentration in rhyodacite (6.0 wt.%) was estimated by the extrapolation of the data from S-bearing system (see Fig. 3). The apparent activity was calculated as aH2O* = H2Om /H2Omsat, where H2Om is the concentration of H2O in the experimental melts while H2Omsat is the water content in water-saturated melts. The calculated apparent H2O activities (aH2O*) vary in the range from 0.95 to 1.03 for Cl-bearing and from 0.2 to 1 in S-bearing systems (Table 3). 5.3. The effect of melt composition, Cl and S on the solubility of Au The available literature data for Au solubility in silicate melts are compiled in Fig. 9. Note that the experimental data for different melt compositions are obtained at different temperatures and pressures, i.e., felsic melts have been investigated at 800–900 °C and 100–150 MPa, rhyodacitic and andesitic systems (this work) have been studied at 1025 and 1050 °C and 200 MPa, while basaltic compositions have been run at 1260 to 1480 °C and atmospheric pressure. The Au solubility values for the basaltic melt compositions were extracted from the experimental datasets at the redox conditions of our experiments (NNO) after Borisov and Palme (1996) and after Brenan et al. (2005). The concentrations of Au in dry basalts are comparable at 1260 and 1300 °C and they are only slightly higher than that in water-bearing Cl- and S-free andesite. The difference increases with increasing temperature (>1300 °C), though the effect of melt composition on Au solubility can not be excluded. The experimental data for volatile-bearing haplogranitic melt compositions show a scattering of the Au concentrations in both Cl- and S-bearing systems over almost three orders of magnitude. Although the data were obtained in the same temperature range (800–850 °C), no systematic dependence of gold solubility in the melt on the volatile concentrations can be resolved from the available datasets. We suggest that the observed scattering of the data is mainly due to the (bulk) analytical method assuming that Au micronuggets were present in the glasses as discussed above. Nevertheless, based on the experimental data obtained in this study in which the effect of micronuggets is minimized, it can be concluded that the effect of melt composition on the metal/melt partitioning of Au is not very significant. This seems to be valid at least for the volatilebearing rhyolites to andesites and for the dry basalts at the studied conditions. It must be noted that the experiments of Borisov and Palme (1996) and Brenan et al. (2005) for volatile-free basaltic liquids illustrated that Au solubility is strongly dependent on the redox conditions, leading the authors to the conclusion that Au is dissolved mainly as an oxide complex. The same effect could be expected for the Au concentration in the melt in S- and Clbearing systems but there is no available dataset allowing us to discuss the effect of fO2 in this case. The experimental data obtained in this study indicate that both Cl and S increase the Au content of the melt up to a factor of 4–5 or even higher, considering sample

R.E. Botcharnikov et al. / Geochimica et Cosmochimica Acta 74 (2010) 2396–2411 10

Rhyodacite Andesite S08 S07 J99 Be09

Au in melt, ppm

1

(a)

2.8 2.4

Au in melt, ppm

2408

FeSm

2.0 1.6 1.2 FeSm

0.8

0.1

0.4 0.0

(a)

0

0.2

0.01 0

0.02

0.04

0.06

0.08

0.1

0.12

1

0.01

0

0.2

0.4

0.6

0.8

Andesite S07 F02 B96 Br05 1

Andesite-S

Rhydacite-Cl

Andesite-Cl

FeSm

2 Au (ppmw) = 0.9327(a* Cl,S) + 0.2518

1.6

R2 = 0.9038

0.8

0

1.2

Cl in melt, wt%

Fig. 9. A compilation of available experimental data on the solubility of Au in silicate melts. The data are after Simon et al. (2005, S05; 2007, S07; 2008, S08; haplogranite at T = 800 and 1050 °C, P = 110–145 MPa); Jugo et al. (1999, J99; haplogranite at T = 850 °C, P = 100 MPa); Frank et al. (2002, F02; haplogranite at 800 °C, P = 100 MPa); Bell et al. (2009, Be09; haplogranite at T = 800 °C, P = 150 MPa); Borisov and Palme (1996, B96; basalt at T = 1300 °C, P = 0.1 MPa), and Brenan et al. (2005, Br05; basalt at T = 1260–1350 °C, P = 0.1 MPa). Note that Au concentrations are shown in a logarithmic scale.

S110 with 2.5 ppm Au. Since the Fe2+/RFe ratio remains almost constant in all experimental glasses (Fig. 6), we can assume that the redox conditions prevailing in the capsules were also almost identical. Hence, the observed increase in Au concentration can not be attributed to higher oxygen fugacity and should be a result of increasing concentration/ activity of Cl and S. Fig. 10 shows the relationships between apparent activities of H2O, Cl and S and the concentrations of dissolved Au in the melt. Although the solubility of Au increases with decreasing apparent activity of H2O in Sbearing systems, it shows no dependence on aH2O* in Clbearing systems (Fig. 10a), indicating that H2O activity does not have a significant influence on Au solubility in the melt. On the other hand, the Au concentration increases linearly with increasing aCl* and aS* (Fig. 10b), illustrating that Au partitioning between metal and melt phases can be strongly controlled by the activities of these two volatiles

FeSm

1.2

FeSm or Cl-brine

0.4

Rhyodacite S08 J99 S05 Be09

Rhyodacite-S

2.4

Au in melt, ppm

Au in melt, ppm

(b)

2.8

(b)

0.1

1

0.14

S in melt, wt% 10

0.4 0.6 0.8 apparent activity of H2O

0

0.2 0.4 0.6 0.8 apparent activity of Cl and S (a* Cl,S)

1

Fig. 10. The dependence of Au solubility on the apparent activity of H2O (a), S and Cl (b). The definition and method of calculation of the apparent activities are described in the text. The maximum apparent activities of S and Cl are assumed to be reached at the saturation of silicate melt with FeSm (sulphide liquid) or Cl-rich brine, respectively. The equation represents a linear regression for the concentrations of Au as a function of both S and Cl (note that S110 sample is not included in the data regression).

and not by the presence of H2O. It must be noted that one sample (S110) shows significantly higher Au concentrations compared with the other samples. It can be supposed that this discrepancy is attributed to the presence of tiny FeS blebs in the experimental glasses which were involved in the analysis. Since the partition coefficient of Au into sulphide phases is very high (e.g., Simon et al., 2007; Bell et al., 2009), the incorporation of such tiny FeS blebs can potentially contaminate the LA analysis. However, in the presence of FeS, the spikes of Fe are expected to be visible in LA spectra and this was not the case for both FeS-bearing samples S110 and SG38. Alternatively, it can be suggested that SG38 sample represents the more reduced conditions than the other samples. Since it is known that Au solubility has a positive dependence on fO2, the Au concentration in the glass of SG38 experiment might correspond to more reduced conditions and, hence it is lower than that expected at the NNO oxygen buffer. However, based on the overall dataset and on the general consistence of the experimental data for glass compositions, redox conditions and volatile contents we do not have any specific reasons to exclude one of the FeS-saturated samples.

Au solubility in volatile-bearing magmas

Interestingly, at least in the range aCl* and aS* = 0–0.6, the increase of Au solubility with increasing aCl* and aS* is very similar in both Cl- and S-bearing systems (Fig. 10b). Although the concentrations of Cl and S in the melt may differ by a factor of ten, the apparent volatile activities might be comparable due to different solubility behaviour of Cl and S. The well-developed correlation between Au and volatile activities indicates that Au may form complexes with both Cl and S in silicate melts. This finding means that not only oxide- but also Cl- and S-bearing Au complexes may control the mobility of Au in natural magmas. The formation of such complexes may lead to a decrease in Au activity coefficients and to a higher solubility of Au in the melt. This conclusion is valid at least for the experimental conditions of this study and the application of our data to other magmatic conditions and to other magma compositions requires further experimental efforts. Nevertheless, the obtained data clearly indicate that magma degassing/segregation/crystallization processes may lead to a significant mobilization of Au in natural magmas. 6. CONCLUSIONS The experiments on the partitioning of gold between Au metal and volatile-bearing rhyodacitic and andesitic melts show that Au may interact with volatiles in the melt and form some Cl- and S-bearing Au complexes. Increasing concentrations of dissolved Cl (up to 1 wt.%) and S (up to 0.04 wt.%) lead to an increase of the concentration of dissolved Au by a factor of 5 or even higher. A possible effect of melt composition can not be clearly observed on the basis of the obtained experimental dataset. The release of volatiles on degassing/segregation/crystallization processes in magmas may lead to a significant mobilization and probably extraction of Au from the melt into other coexisting/ formed magmatic phases. ACKNOWLEDGMENTS We thank O. Diedrich for the preparation of the samples for analysis; R. Schoenberg for the bulk ICP-MS analyses of experimental glasses; Z. Yang and B. Fryer from Great Lakes Institute for Environmental Research, University of Windsor, Canada for the help with LA-ICP-MS measurements; D. Kosanke and A. Husen for the help with determination of Fe2+/RFe ratios in the glasses. A. Simon and J. Webster are gratefully acknowledged for the constructive reviews. The financial support for this research was provided by the DFG (BO2941/1).

REFERENCES Archibald S. M., Migdisov A. A. and Williams-Jones A. E. (2001) The stability of Au–chloride complexes in water vapor at elevated temperatures and pressures. Geochim. Cosmochim. Acta 65, 4413–4423. Audetat A., Gunther D. and Heinrich C. A. (1998) Formation of a magmatic–hydrothermal ore deposit: Insights with LA-ICP-MS analysis of fluid inclusions. Science 279, 2091–2094. Audetat A. and Pettke T. (2006) Evolution of a porphyry–Cu mineralized magma system at Santa Rita, New Mexico (USA). J. Petrol. 47, 2021–2046.

2409

Bell A. S., Simon A. and Guillong M. (2009) Experimental constraints on Pt, Pd and Au partitioning and fractionation in silicate melt-sulfide-oxide-aqueous fluid systems at 800 °C, 150 MPa and variable sulfur fugacity. Geochim. Cosmochim. Acta 73, 5778–5792. Baker D. R., Freda C., Brooker R. A. and Scarlato P. (2005) Volatile diffusion in silicate melts and its effects on melt inclusions. Ann. Geophys. 48, 699–717. Berndt J., Liebske C., Holtz F., Freise M., Nowak M., Ziegenbein D., Hurkuck W. and Koepke J. (2002) A combined rapidquench and H2-membrane setup for internally heated pressure vessels: description and application for water solubility in basaltic melts. Am. Mineral. 87, 1717–1726. Bezmen N. I., Asif M., Brugmann G. E., Romanenko I. M. and Naldrett A. J. (1994) Distribution of Pd, Rh, Ru, Ir, Os, and Au between sulfide and silicate metals. Geochim. Cosmochim. Acta 58, 1251–1260. Bodnar, R.J., 1995. Fluid-inclusion evidence for a magmatic source for metals in porphyry copper deposits. In: J.F.H., T. (Ed.), Magmas, fluids and ore deposits, vol. 23. The Mineralogical association of Canada, Short Course, pp. 139-152. Borisov A. and Palme H. (1996) Experimental determination of the solubility of Au in silicate melts. Mineral. Petrol. 56, 297– 312. Botcharnikov R. E., Behrens H. and Holtz F. (2006) Solubility and speciation of C–O–H fluids in andesitic melt at T = 1100–1300 °C and P = 200 and 500 MPa. Chem. Geol. 229, 125–143. Botcharnikov R. E., Behrens H., Holtz F., Koepke J. and Sato H. (2004) Sulfur and chlorine solubility in Mt. Unzen rhyodacitic melt at 850 °C and 200 MPa. Chem. Geol. 213, 207–225. Botcharnikov R. E., Holtz F., Almeev R. R., Sato H. and Behrens H. (2008) Storage conditions and evolution of andesitic magma prior to the 1991–95 eruption of Unzen volcano: Constraints from natural samples and phase equilibria experiments. J. Volcanol. Geotherm. Res. 175, 168–180. Botcharnikov R. E., Holtz F. and Behrens H. (2007) The effect of CO2 on the solubility of H2O–Cl fluids in andesitic melt. Eur. J. Mineral. 19, 671–680. Botcharnikov R. E., Koepke J., Holtz F., McCammon C. and Wilke M. (2005) The effect of water activity on the oxidation and structural state of Fe in a ferro-basaltic melt. Geochim. Cosmochim. Acta 69, 5071–5085. Brenan J. M., McDonough W. F. and Ash R. (2005) An experimental study of the solubility and partitioning of iridium, osmium and gold between olivine and silicate melt. Earth Planet. Sci. Lett. 237, 855–872. Burnham C. W. (1979) The importance of volatile constituents. In The Evolution of the Igneous Rocks: Fiftieth Anniversary Perspectives (ed. J. H. S. Yoder). Princeton University Press, Princeton, NJ, pp. 439–482. Carroll M. R. and Rutherford M. J. (1988) Sulfur speciation in hydrous experimental glasses of varying oxidation state; results from measured wavelength shifts of sulfur X-rays. Am. Mineral. 73, 845–849. Carroll M. R. and Webster J. D. (1994) Solubilities of sulfur, noble gases, nitrogen, chlorine, and fluorine in magmas. In Volatiles in magmas, vol. 30 (eds. R. Carroll Michael and R. Holloway John). Mineralogical Society of America, Washington, DC, United States, pp. 231–279. Frank M. R., Candela P. A., Piccoli P. M. and Glascock M. D. (2002) Gold solubility, speciation, and partitioning as a function of HCl in the brine-silicate melt-metallic gold system at 800 °C and 100 MPa. Geochim. Cosmochim. Acta 66, 3719– 3732. Halter W. E., Heinrich C. A. and Pettke T. (2005) Magma evolution and the formation of porphyry Cu–Au ore fluids:

2410

R.E. Botcharnikov et al. / Geochimica et Cosmochimica Acta 74 (2010) 2396–2411

evidence from silicate and sulfide melt inclusions. Mineralium Deposita 39, 845–863. Halter W. E., Pettke T. and Heinrich C. A. (2002) The origin of Cu/Au ratios in porphyry-type ore deposits. Science 296, 1844– 1846. Hanley J. J., Pettke T., Mungall J. E. and Spooner E. T. C. (2005) The solubility of platinum and gold in NaCl brines at 1.5 kbar, 600–800 °C: a laser ablation ICP-MS pilot study of synthetic fluid inclusions. Geochim. Cosmochim. Acta 69, 2593–2611. Harris A. C., Kamenetsky V. S., White N. C. and Steele D. A. (2004) Volatile phase separation in silicic magmas at Bajo de la Alumbrera porphyry Cu–Au deposit, NW Argentina. Res. Geol. 54, 341–356. Harris A. C., Kamenetsky V. S., White N. C., van Achterbergh E. and Ryan C. G. (2003) Melt inclusions in veins: linking magmas and porphyry Cu deposits. Science 302, 2109–2111. Hattori K. I. and Keith J. (2001) Contribution of mafic melt to porphyry copper mineralization: evidence from Mount Pinatubo, Philippines, and Bingham Canyon, Utah, USA. Mineralium Deposita 36, 799–806. Hedenquist J. W., Aoki M. and Shinohara H. (1994) Flux of volatiles and ore-forming metals from the magmatic–hydrothermal system of Satsuma Iwojima volcano. Geology 22, 585– 588. Hedenquist J. W. and Lowenstern J. B. (1994) The role of magmas in the formation of hydrothermal ore deposits. Nature 370, 519–527. Hedenquist J. W., Simmons S. F., Giggenbach W. F. and Eldridge C. S. (1993) White-Island, New-Zealand, volcanic-hydrothermal system represents the geochemical environment of highsulfidation Cu and Au ore deposition. Geology 21, 731–734. Heinrich C. A., Driesner T., Stefansson A. and Seward T. M. (2004) Magmatic vapor contraction and the transport of gold from the porphyry environment to epithermal ore deposits. Geology 32, 761–764. Heinrich C. A., Gunther D., Audetat A., Ulrich T. and Frischknecht R. (1999) Metal fractionation between magmatic brine and vapor, determined by microanalysis of fluid inclusions. Geology 27, 755–758. Holland H. D. (1972) Granites, solutions, and base metal deposits. Economic Geology 67, 281–301. Hunt, J. P., 1991. Porphyry copper deposits: Historical perspectives of genetic concepts and case histories of famous discoveries, Economic Geology Monographs. Economic Geology Publishing Co., Lancaster 8, 192–206. Jana D. and Walker D. (1997) The influence of sulfur on partitioning of siderophile elements. Geochim. Cosmochim. Acta 61, 5255–5277. Jugo P. J., Candela P. A. and Piccoli P. M. (1999) Magmatic sulfides and Au: Cu ratios in porphyry deposits: an experimental study of copper and gold partitioning at 850 °C, 100 MPa in a haplogranitic melt pyrrhotite intermediate solid solution gold metal assemblage, at gas saturation. Lithos 46, 573–589. Keith J. D., Whitney J. A., Hattori K., Ballantyne G. H., Christiansen E. H., Barr D. L., Cannan T. M. and Hook C. J. (1997) The role of magmatic sulfides and mafic alkaline magmas in the Bingham and Tintic mining districts, Utah. J. Petrol. 38, 1679–1690. Liu Y., Samaha N.-T. and Baker D. R. (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim. Cosmochim. Acta 71, 1783–1799. Longrich H. P., Jackson S. E. and Gu¨nther D. (1996) Laser ablation-inductively coupled plasma-mass spectrometric transient signal data acquisition and analyte concentration calculation. J. Anal. Atom. Spectrom. 11, 899–904.

Lowenstern J. B. (1993) Evidence for a copper-bearing fluid in magma erupted at the Valley of ten thousand smokes Alaska. Contrib. Mineral. Petrol. 114, 409–421. Maughan D., Keith J., Christiansen E., Pulsipher T., Hattori K. and Evans N. (2002) Contributions from mafic alkaline magmas to the Bingham porphyry Cu–Au–Mo deposit, Utah, USA. Mineralium Deposita 37, 14–37. Metrich N. and Rutherford M. J. (1992) Experimental study of chlorine behavior in hydrous silicic melts. Geochim. Cosmochim. Acta 56, 607–616. Mungall J. E. (2002) Kinetic controls on the partitioning of trace elements between silicate and sulfide liquids. J. Petrol. 43, 749– 768. Ohlhorst S., Behrens H. and Holtz F. (2001) Compositional dependence of molar absorptivities of near-infrared OH- and H2O bands in rhyolitic to basaltic glasses. Chem. Geol. 174, 5– 20. Pokrovski G. S., Tagirov B. R., Schott J., Bazarkina E. F., Hazemann J.-L. and Proux O. (2009a) An in situ X-ray absorption spectroscopy study of gold-chloride complexing in hydrothermal fluids. Chem. Geol. 259, 17–29. Pokrovski G. S., Tagirov B. R., Schott J., Hazemann J.-L. and Proux O. (2009b) A new view on gold speciation in sulfurbearing hydrothermal fluids from in situ X-ray absorption spectroscopy and quantum-chemical modeling. Geochim. Cosmochim. Acta 73, 5406–5427. Sato H., Holtz F., Behrens H., Botcharnikov R. and Nakada S. (2005) Experimental petrology of the 1991–1995 Unzen dacite, Japan. Part II: Cl/OH partitioning between hornblende and melt and its implications for the origin of oscillatory zoning of hornblende phenocrysts. J. Petrol. 46, 339–354. Schuessler J. A., Botcharnikov R. E., Behrens H., Misiti V. and Freda C. (2008) Amorphous materials: properties, structure, and durability: oxidation state of iron in hydrous phonotephritic melts. Am. Mineral. 93, 1493–1504. Shinohara H. and Hedenquist J. W. (1997) Constraints on magma degassing beneath the far southeast porphyry Cu–Au deposit, Philippines. J. Petrol. 38, 1741–1752. Signorelli S. and Carroll M. R. (2000) Solubility and fluid-melt partitioning of Cl in hydrous phonolitic melts. Geochim. Cosmochim. Acta 64, 2851–2862. Sillitoe R. H. (1979) Some thoughts on gold-rich porphyry copperdeposits. Mineralium Deposita 14, 161–174. Sillitoe R. H. (2000) Gold-rich porphyry deposits: descriptive and genetic models and their role in exploration and discovery. Rev. Econ. Geol. 13, 315–345. Sillitoe R. H. (2008) Special paper: major gold deposits and belts of the north and south American Cordillera: distribution, tectonomagmatic settings, and metallogenic considerations. Econ. Geol. 103, 663–687. Simon A. C., Candela P. A., Piccoli P. M., Mengason M. and Englander L. (2008) The effect of crystal-melt partitioning on the budgets of Cu, Au, and Ag. Am. Mineral. 93, 1437– 1448. Simon A. C., Frank M. R., Pettke T., Candela P. A., Piccoli P. M. and Heinrich C. A. (2005) Gold partitioning in melt-vaporbrine systems. Geochim. Cosmochim. Acta 69, 3321–3335. Simon A. C., Pettke T., Candela P. A., Piccoli P. M. and Heinrich C. A. (2003) Experimental determination of Au solubility in rhyolite melt and magnetite: constraints on magmatic Au budgets. Am. Mineral. 88, 1644–1651. Simon A. C., Pettke T., Candela P. A., Piccoli P. M. and Heinrich C. A. (2007) The partitioning behavior of As and Au in S-free and S-bearing magmatic assemblages. Geochim. Cosmochim. Acta 71, 1764–1782.

Au solubility in volatile-bearing magmas Stefansson A. and Seward T. M. (2003a) The hydrolysis of gold(I) in aqueous solutions to 600 °C and 1500 bar. Geochim. Cosmochim. Acta 67, 1677–1688. Stefansson A. and Seward T. M. (2003b) Stability of chloridogold(I) complexes in aqueous solutions from 300 to 600 °C and from 500 to 1800 bar. Geochim. Cosmochim. Acta 67, 4559– 4576. Stefansson A. and Seward T. M. (2004) Gold(I) complexing in aqueous sulphide solutions to 500 °C at 500 bar. Geochim. Cosmochim. Acta 68, 4121–4143. Stelling J., Botcharnikov R. E., Beermann O. and Nowak M. (2008) Solubility of H2O- and chlorine-bearing fluids in basaltic melt of Mount Etna at T = 1050–1250 °C and P = 200 MPa. Chem. Geol. 256, 102–110. Tamic N., Behrens H. and Holtz F. (2001) The solubility of H2O and CO2 in rhyolitic melts in equilibrium with a mixed CO2– H2O fluid phase. Chem. Geol. 174, 333–347. Taran Y. A., Bernard A., Gavilanes J. C. and Africano F. (2000) Native gold in mineral precipitates from high-temperature volcanic gases of Colima volcano, Mexico. Appl. Geochem. 15, 337–346. Ulrich T., Guenther D. and Heinrich C. A. (1999) Gold concentrations of magmatic brines and the metal budget of porphyry copper deposits. Nature 399, 676–679. Ulrich T., Gunther D. and Heinrich C. A. (2002) Evolution of a porphyry Cu–Au deposit, based on LA-ICP-MS analysis of

2411

fluid inclusions: Bajo de la Alumbrera, Argentina (vol. 96, p. 1743, 2001). Econ. Geol. Bull. Soc. 97, 1889–1920. Webster J. D. and De Vivo B. (2002) Experimental and modeled solubilities of chlorine in aluminosilicate melts, consequences of magma evolution, and implications for exsolution of hydrous chloride melt at Mt. Somma–Vesuvius. Am. Mineral. 87, 1046– 1061. Webster J. D., Kinzler R. J. and Mathez E. A. (1999) Chloride and water solubility in basalt and andesite melts and implications for magmatic degassing. Geochim. Cosmochim. Acta 63, 729– 738. Webster J. D. and Mandeville C. W. (2007) Fluid immiscibility in volcanic environments. Rev. Mineral. Geochem. 65, 313–362. Wilson A. D. (1960) The micro-determination of ferrous iron in silicate minerals by a volumetric and colorimetric method. Analyst 85, 823–827. Zezin D. Y., Migdisov A. A. and Williams-Jones A. E. (2007) The solubility of gold in hydrogen sulfide gas: an experimental study. Geochim. Cosmochim. Acta 71, 3070–3081. Zotov A. V., Baranova N. N., Dar’yina T. G. and Bannykh L. M. (1991) The solubility of gold in aqueous chloride fluids at 350– 550 °C and 500–1500 atm: Thermodynamic parameters of AuCl2-(aq) up to 750 °C and 5000 atm. Geochem. Int. 28, 63–71. Associate editor: Edward M. Ripley