Chapter § MODEL ORE-METAL PARTITIONING FROM MELTS INTO VAPOR AND VAPOR/BRINE MIXTURES P.A. Candela and P.M. Piccoli Laboratory for Mineral Deposits Research, University of Maryland at College Park, College Park MD 20742-4211 INTRODUCTION Experimental and computational studies of ore-metal-bearing systems can aid in the development of strategies that are required in the exploration for earth resources. Theoretical studies can suggest factors that maximize the probability of ore formation, and can also suggest specific tools for evaluating — magmatic- hydrothermal ore potential. In this paper, we prosent an update of the previous work performed by Candela & Holland (1986) and Candela (19892); our carlier papers suggested that porphyry copper ore-formation is favored by (1) high initial C1/H,0 ratios in the magma, and, (2) “early” saturation of a melt with respect to a volatile phase (because of the commonly compatible behavior of Cu). Further, Candela & Holland (1986) demonstrated that porphyry copper deposits can form given the calculated efficiencies of removal, and that reasonable initial magmatic copper concentrations in the melt are sufficient. However, those earlier models considered only one supercritical fluid phase, or a one-phase metastable fluid In this communication, earlier contributions are updated to include consideration of melt-vapor-brine equilibria Computational models of magmatic devola- tilization, building on those earlier efforts, have been continually under development in our laboratory. Our algorithms predict the concen- trations of ore metals, isotopes, and other elements in melts, vapors, brines, and crystals during volatile exsolution and magmatic crystallization. Rather than perform simple mass- balance calculations for each element individually, our algorithm MVP (Magmatic Volatile Phases), a second-boiling model (see Bowen 1928), considers the simultaneous equilibria (among melt, crystals, and volatile phases ineluding brine) that must be accounted for to model the efficiencies with which ore metals are removed from magmas into aqueous ore fluids, ‘These model calculations can be used to construct hypotheses that can be tested in the field. Further, many metals of economic im- portance are chloride-complexed in aqueous fluids at the temperatures that prevail in felsic magm systems, and melts may be saturated with respect to both vapor and brine phases, Therefore, we have considered explicitly melt-vapor-brine equilibria; in this treatment, attention is paid to ns in the composition of these tent with the phase rule. ‘Although few data exist on multi-component vapor/brine equilibria at magmatic temperatures, wwe present a model that estimates the effects of vapor-brine separation on the partitioning of elements such as copper.” Preliminary data are available on the partitioning of copper among the phases melt-vapor-brine (Williams e¢ al. 1995). The timing of brine saturation relative to crystallization progress, and the concentrations of selected elements in the brine, vapor, and devolatilizing melt, can be calculated by using our algorithm. Before proceeding to a discussion of metal partitioning, we must first consider the behavior of chlorine in magma/volatile systems. CHLORIDE PARTITIONING Many studies (Tables 1, 2, 3) have demon- strated the strong affect of chlorine on the partitioning of ore metals between melts and volatile phase(s). Since the pioneering efforts of 101Candela & Piecoth ‘Table 1. Metaluminous to slightly peraluminous (low Cl supereritical fluid/melt partition coefficients: alk and F) equilibrium vapor/melt, brine/melt, and ,alkaline-carth, and 2° transition metals Flement WV Mel Composition v/l _b/l afl TEC) P(Rb) Reference rr a ee T_T highsiicariyone DF wT Willams (1995) Li 1 haplogranite 1.03031 8002 Bos (1990) Li 1 topazthyolite 10s 8500S Webster e al. (1989) Li 1 topazshyolite 40 50 Webster etal (1989) Li 1 macusanite 10 ee London etal (1988) Be 2 topazrhyolite 100 3502 Webster eal. (1989) Mg 2. high-silica rhyolite 0.1680.04 770-880 14-24 Holland (1972) KL high-silica rhyolite 0.7440.06 770-880 1-24 Gammon et al. (1969) Holland (1972) Ca 2. highsslica rhyolite 03840.09 770-880 14-24 Holland (1972) Rb 1 QwAbea 0.774005 302 Carron & Legache (1980) Rb 1 QiaOrae 0.6640.06 30 2 Carron & Lepache (1980) Rb I haplogranite 0.613023 7502 Bos (1990) St 2 Qube 0.2840.07 si Carron & Lepache (1980) St 2 QuOrep 0,500.13 0d Carron & Legache (1980) Cs 1 QuAbey 1,080.08 a0 Catron & Legache (1980) CS 1 QuOrao 0.6730.04 800 2 Carron & Legache (1980) Ba 2 QuAby 0.2980.05 800 2 Carron & Legaehe (1980) Ba 2 QuOrw 0.4340.09 800 2 Catron & Legache (1980) Mn 2 highesilica rhyolite 65216 770-880 1.4-24 Holland (1972) u 1 highesliea rhyolite 2085.6 750 14 Candela & Holland (1984) CuI highesilica rhyolite 215273 21573 @iss73) 8001 Williams (1993) CuI highesilica rhyolite 1126 1186 (iss) 8500.5 Williams (1995) Zn___2. high-silica rhyolite 924 8002 Holland (1972) ‘General nates about equilibrium partitioning Tables: uncertain ‘where listed) are 16. Supercritical fluid is defined here as any fluid that occurs at « pressure above the maximum defined by the critical curve for a given temperature (see Bodnar et al, 1985) Nis the ligation numb« Holland (1972) and Kiline & Burnham (1972), researchers interested in the partitioning of metals between melts and volatile phases have recognized the importance of understanding the distribution of chloride in magma-volatile systems (cf. Shinohara ef al. 1989; Webster er al 1989; Webster 1992a,b; Mettich & Rutherford 1992). Of course, only components that can vary independently in a phase can have a real distribution coefficient; clearly, this is not the case with chloride, because Cl is not an independently variable component of the phases in question. Simple Nemnst-type partition coefficients are only crude approximations of the behavior of Cl, but we will use those formulations here for convenience. Our suggested thermodynamic formulation for chloride partitioning in melt + ‘vapor systems is presented in Appendix 2. 102 See the text for additional details Experimental Studies on Chlorine Partitioning between the Melt and Vapor + Brine at Pressures below 2 Kilobars Many studies, since the work of Holland (1972) and Kiline & Burnham (1972), have measured the distribution of chloride between coexisting melt and volatile phases. However, many chloride partitioning studies have been performed at pressures in excess of 2 kb, These data are not germane to. most magmati hydrothermal ore systems, or they have been performed in the two-phase region with only a few experiments aimed at the estimation of two- phase (vapor/melt or brine/melt) partition coefficients, Reconnaissance experiments performed by our laboratory at 1 kb and 800 °C in the high-silica rhyolite — aqueous vapor system (ie., concentration of Cl in the aqueous vaporOre-metal Partitioning ‘Table 2, Metaluminous to slightly peraluminous (low Cl and F) equilibrium vapor/melt, brine/melt, and supercritical fluid/melt partition coefficeients: rare earth, and high field strength transition elements lement_N Mel Composiion a eT TRG) PRB) —Reterence Ta 3 haha ryote Tye. 80) 7a 1995 Ce 3 Spruce Pine pegmatite 2.561089 8001.25 Flynn & Burham (1978) Ce 3. highs nyoite 4815177 8001 Read (1995) Né 3 highslica rhyolite SI3HL89 800 Rewd (1995) Sm 3. high-sliea rhyolite 446H1.64 800 1——_Read (1995) Ev 3 Aighsilla hyolite 5201192 800 1_Reed (1995) Gd 3 higheslica rhyolite Ao4si7i 8001 Reed (1993), T 3. highslfa ehyoite A715 8001 Reed (1995) Ho 3. hiphsilin hyoite 3788139 8001 Reed (1995) Yb 3. Spruce Pine pegmatite 1128039 800 1.25—_Flynn & Burnham (1978) Yb 3. higheilica yo, BU6ELI6 = 800 Reed (1995), 3 ks ROLE 8001 Reed (1995), say) 03 750-2 Keppler ae Wyllie (1991) age 2 07 8501 Nekrasov eral. (1980) Mir 2 quariealbite 8 8501 Nekrasove al (1980) BP aptograite 0.00.38 750 2 Keppler & Wyilie (1991) Th 2, topazthyolite 300 8502 Webstore af (1989) W 1 Qreabuor, 39813 8001 Manning & Henderson ‘General notes about equilibrium partiioning Tables: unceraintes (where listed) are To. Supercritical fluid is defined here as any fluid that occurs ata pressure above the maximum defined by the critical curve for a given temperatute (see Bodnar etal. 1985). Nis the ligation number, “At high CL:W the ligation for W may be 4 of 6 (Note: some experiments by Keppler & Wyllie (1991) suggest W partitioning is independent of Cl—see Table 3). Data for U were obtained from experiments which used H,OHTICt (only) as the starting solution. Data for W were obtained feom experiments in which the only chloride added was in the form NaCl, The ligation number for Th has been estimated from the 925-950 °C experiments of Webster etal. (1989), ‘Table 3. Metaluminous to slightly peraluminous (low Cl and F) nernstian vapor/melt, brine/melt, and ‘supereritical luid/melt partition coefficients lemeat Melt Dit pet Te) ki) Retrenee ¥ ‘0 __aplogranite HoIs 750-80 Pichavant 1981) B 0 macusanite 23 775° 2 London etal. (1988) F 0 topaz rhyolite 0183003 800 =~ Webster (1990) F 0 haplogranite 08:02 750° 2 Keppler & Wylie (1991) a 0 topaz rhyol 3.81? 725.797 1.0 Webster & Holloway (1988) a 0 high-silica rhyolite 16.683.9 810 1.2 Shinohara ea, (1989) a 0 high-silica rhyolite 3.7 Bs Y Mettich & Rutherford (1992) a 0 high-silica rhyolite 1.79204 800 1 Williams (1995) Mo, 0 highsslica rhyolite 25816 750 14 Candela & Holland (1984) Mo! 0 haplogranite 35 7802 Keppler & Wyllie (1991) Mp" 0 aplogranite 35128 750-2 Keppler & Wyllie (1991) wt 0 haplogranite 35209 750-2 Keppler & Wyllie (1991) we 0 haplogranite 072802 750-2 Keppler & Wyllie (1991) w 0 QayyAbs Otay 0.748025 800 i Manning & Henderson (1984) Th 0 haplogranite 01200137502 Keppler & Wylie (1991) ‘This aie contang Tafomation for elements which exh Nesian-ijpebahaviorUncenaitcs (where Tite) ar To the ligation number a 400 ppm Cin the vapor phos; and designate runs in which he starting solutions comand Ha ‘only, and #,0 + (XCl or HF) stating solutions, respectively, "Th has also been included in Tables 1 and 2 (Seto for etal). 103Candela & Piccoli <0.32 molal), yield vapor/melt _ partition coefficients of approximately 3-5 (Williams, Candela & Piccoli, unpublished). These numbers are equivalent to those obtained by Webster (1992a) under similar conditions. Metrich & Rutherford (1992) also collected data on the chloride concentration of rhyolite at 1 kb, but at 845 °C. These experiments were performed in the three-phase field (melt-vapor-brine). Given the model NaCI-H,0 system from Bodnar ef al. (1985), we can estimate the vapor composition in the experiments of Metrich & Rutherford (1992), and a vapor/brine partition coefficient can be calculated; this analysis yields D¢j' = 3.7. All of these determinations are in agreement within error. The data from Shinohara (1987) yield partition coefficients that are higher under similar conditions. Webster (1992) suggested that this is due to a systematic error in the concentration of Clin the melt reported by Shinohara. Further, the Na/(NatK) ratios in Shinohara's melts are higher than in the other studies, and this may be a cause of some diserepaney. We tentatively conclude that the empirical DU at pressures less than ] kb is on the order of 5, and we use that value in our calculations. The study by Shinohara er al. (1989) showed convincingly that the Cl vapor/melt partition coefficient decreases with decreasing pressure. Therefore, even lower partition coefficients are to be expected at pressures less than I kb. It is instructive to consider the limitations that the phase rule places on considerations of melt ~ vapor + brine systems. Consider the simple system Ab-H;O-NaCl. As other components are added to create a more realistic magma ~ volatile system, restrictions of constant composition or phase saturation can be added as well. Ata given temperature and pressure, the phase assemblage ‘melt-vapor-brine is invariant. Therefore, when the melt is co-saturated with brine plus vapor, the ‘composition of the phases in the system is fixed. ‘The Cl concentration in the melt can vary only as a function of temperature and pressure. Of course, this simple ternary system is not of direct geological interest. However, when more com- ponents are added, restrictions are also added ‘The restrictions are of two types: 1) addition of a 104 new phase (resulting in a Gibbs-Duhem equation for that phase); or 2) restrictions on the composition of the system (e.g, if only small excursions in K/Na or AV(K+Na) occur in a system, then the chemical potentials of other components are limited in their possible variation). For example, the system Ab-H,0- NaCLOr-An-Q-FeO-MgO is in a state of isobaric~isothermal invariance when the phase assemblage is melt + vapor + brine + qz + af + pl + mt + bt, a very likely assemblage at vapor + brine saturation (restriction of type 1, above), However, a critic might note that the presence of HCI in the volatile phase is not accounted for in this 8-component system. As pointed out in Candela (1990), the component HNa accounts both for HCI in the volatile phases and for variation in the ASI of the melt, Given that the ASI of many melts cluster around unity, and do not vary greatly during the exsolution of vapor + brine, the variation in the partition coefficient for Cl due to the variance generated by this extra dogree of freedom will be small. Therefore, phase-rule constraints limit the variability in the Cl concentration in the melt saturated with vapor and brine at a given temperature and pressure. Similar conclusions have been reached by Shinohara (1994), and the interested reader is directed to that paper for further discussion. As we will demonstrate later in this paper, variations in temperature and pressure during vapor + brine saturation are rather restricted, and variations in the Cl concentration of the associated melt phase are apt to be small, These factors mitigate against large excursions in the Cl concentration of a melt when the melt is saturated with respect to both brine and vapor. Similar arguments apply when a ‘melt or aqueous fluid is saturated with respect to two feldspars (cf, Lagache & Weisbrod 1977), ot two (Na, K)-Cl solid solutions (of: Sterner et al. 1992). For a system of a given major-element composition, the Cl concentration of the coexisting melt, vapor, and brine, will be a function of temperature and pressure only, and will not allow the free variation of an element, as in, eo Rayleigh fractionation. Rayleigh fractionation requires a system with a variance greater than zero. Rayleigh fractionation requires a system with a variance greater than zero,‘THE PARTITIONING OF ORE- AND OTHER ELEMENTS BETWEEN MELTS AND MAGMATIC AQUEOUS FLUIDS Abundant experimental data exist conceming the partitioning of ore metals between melts and volatile phases. Here, we provide an overview of the data pertaining to vapor/melt, brine/melt, aqueous mixture/melt, and supercritical fluid/melt Partitioning in silicic systems. This review is not ‘exhaustive, but covers studies which have been performed with granitic melt compositions, and with ASI greater than but close to unity, We were interested mainly in data that provide the most complete experimental run-product analyses, that were performed at low pressures (0.5 to 2 kb), and. at temperatures near the granite solidus, Further, there are many complexities in these experiments which cannot be discussed here, and the readers are referred to the original sources for more detailed information; this treatment is meant only as an overview. General Notes about the Data in the Tables Formulation of the data ‘The elements presented in Tables 1, 2, and 3 have been grouped according to their behavior in the magmatic volatile phase(s) as: 1) Cl complexed elements (of the form Ki, where and j are the element and phase of interest, respectively); and 2) those elements exhibiting ‘Nernst-type behavior (i.e., elements which do not exhibit Cl-complexing, and have _ partition coefficients of the form D/"= C/ /C} ). The data are presented as vapor/melt (Ki; D?”"), brine/melt (Kise; D#"), [aqueous mixture of brine and vaporj/melt (K?%'; D?*”), or superer Aluid/melt (KI; Df") coefficients Many of the original papers report data for Cl- complexed elements in the form of Cl-dependent Nernst-type partition coefficients; this is not a desirable form for many elements (Candela 1990), Several workers have adopted a more thermodynamically rigorous formulation for Cl- complexed elements, i.e, exchange constants, KlNe (Holland 1972; Gammon et al, 1969; Ore-metal Partitioning Candela & Holland 1986; Bos 1990; Williams 1995), but there are a significant amount of good- quality experimental data reported in other forms, Data presented in the form of Nernst-type partition coefficients (for Cl-complexed elements) in original sources (e.g. Carron & Legache 1980; Flynn & Burnham 1978; Keppler & Wyllie 1991) have been recast in the form of exchange constants, formulated as: i [td op : ees] ay These exchange constants represent equilibria such as: 1 1 Mnsen) MH hucr Hey, Mvagy, ‘where is the element of interest, jis vapor, brine, or supercritical fluid, and 7 is the ligation number or number of Cl ions per cation in the presumed aqueous complex at pressure and temperature (for a thorough discussion of this procedure, readers are referred to Candela 1986a). An exchange constant, for example Kno, represents, simply, the partitioning of e.g., Zn, between vapor and ‘melt, relative to Na: 2Whect ho Mc? Minos, We consider equilibria involving ions to be anathema because: 1) there is no commonly accepted standard state for ions in melts; and 2) the volatile phases are mostly associated molecular fluids, We therefore express conditions for chemical equilibrium in terms of independently variable components of the phases, To transform data originally presented as Dj"'= Cj /C}_ into exchange constants requires that the experimentalist provide either Na con- centrations (melt and magmatic volatile phase), or at least analyses of CI” in the products of the aqueous run. When data were not available in the 10sCandela & Piccoli cited work, C4, /Cl,, was approximated by 0.46 + 0.04mer. This functional dependence is, consistent with the work of Holland (1972), Shinohara ef al. (1989) and Urabe (1985); the relation is purely empirical and is not thermo- dynamically rigorous, but seems to be valid where a haplogranitic melt was used in the experiments. The use of this empirical parameter may not be preferred, but its use is the only means, in some cases, by which simple partition ratios of Cl complexed metals can be approximated. ‘The data of Bodnar er al. (1985) on the system NaCI-H,O were used to determine if experiments were performed in the one- or two-volatile phase regions. All experiments at 2 kb are classed as supercritical regardless of temperature (over the temperature range of interest here). Several sets of data have no reported uncertainty, either because of a lack of uncertainty reported in the original sources, or because some of the results represent individual data points (see element- specific discussions below). Hydrogen (D) ‘The most important chloride leading to alteration associated with porphyry deposits is HCI [chloride of hydrogen (1); see Sverjensky et al. 1991]. H(), ie. Hz, takes part in oxidation— reduction equilibria and is NOT the topic of this discussion: note that Hz is a component that is distinct from HCI and is unrelated to these equilibria. Unlike fluoride, where HF comprises ‘most of the fluoride species in supercritical fluids (© 99% at T>550 °C, and 1 kb, for the assemblage wollastonite-fluorite-quartz; Zhu & Sverjensky 1991), much of the chloride at magmatic temperatures can be present as species other than HCl. ‘The H(1) concentration in the magmatic volatile phase in equilibrium with a rhyolitie melt (ASI ~ 1) at 1 km depth has about 20% of its Cl bound to hydrogen; at 6 km approximately 5% of Cl is as HCI (but with considerable scatter of the data; Shinohara er al, 1984). Williams (1995) calculated model exchange constants between brine and melt, and vapor and melt for the equilibrium: 106 aul » 1 hacrWhac Bc THnaasyo, where Hae = hy H{C #420, © and represents the excess Al activity in the melt. ‘An exchange constant for this expression can be represented by ae Chey © Kine” i Na” ase" CaCl where Ch, = (NadISi;0o)" In these experiments, the concentrations of NaCl and HCI were calculated from the concentrations of Na and H*, respectively, in the quenched product (homogenized from the vapor and brine at pressure and temperature). Model concentrations of NaCI and HCI in the vapor and brine phases at pressure and temperature were estimated given the total Cl concentration in the charge, and the composition of coexisting vapor and brine in the system NaCl-H,O. The ASI of the melt was calculated from the electron- microprobe analysis of the glass. ‘At a pressure of 1 kb (800 °C), Williams (1995) reported a value for K'zvia and Kins of 24 and 2, respectively. These results suggest HCUNaCI is higher in the vapor relative to brine by an order of magnitude. Lithium Lithium, present in significant quantities in highly evolved systems, is yet to be the emphasis of a comprehensive experimental study of melt- volatile phase partitioning, Data are presented from three studies with evolved melt compositions: macusanite (which is enriched in LisO in additional to F and B,O3; London er al. 1988), topaz rhyolite (Webster ef ai. 1989), and a haplogranite melt (Bos 1990). Based on the study by Bos (1990), and the oxidation state of Li, we adopt a ligation number for Li equal to unity[although some data from Webster et al. (1989) at 900-950 °C end 2 kb may suggest values as high as 3], which implies that Li is present in magmatic volatile phases as LiCl In a study of the partitioning of several elements between haplogranite melt and super critical fluids, performed at 800 °C and 2 kb, Bos (1990) reported the results from 15 experiments involving Li partitioning. His results, represented by the expression Dif” = 0.47mer (in HF-free experiments), imply that the Li-Na exchange constant is close to unity. Bos reported that Li partitioning into a supercritical gas is enhanced with increasing melt A/NK [A/NK = molar AlOy(NazO + K;0); in the range 0.9 to 1.1], and very limited data suggest D{,”’ increases with HF in supercritical fluids, The uncertainty in Kite reported here is due to the variation in reported supercritical fluid/melt partitioning data only (Bos 1990). London ef al. (1988) and Webster ef al. (1989) reported Li partitioning data for high-F systems (some with > 9000 ppm F in the melt) ‘The Di! of 0.4 reported in London er al. (1988) has been used to calculate the K in Table 1 given the 1:1 water:melt ratio in the experiment, and 4450 ppm Cl in the starting glass, At2 kb, the data of London ef al. (1988) suggest Kis = 10 (at 775 °C) in a macusanite melt (ASI of starting alass = Lf), whereas those of Webster er a (1989) suggest a significantly higher KY/he of 40 (at 850 °C; Run "15) in a topaz rhyolite melt (ASI of starting glass = 1.0). An additional run (Webster ef al, 1989; Run *1) at 0.5 kb (850 °C) within the two-phase field suggests significantly enhanced partitioning of Li into the aqueous mixture (K7P'M, = 105). Webster et al. (1989) also recognized a temperature dependence on Li partitioning: with increasing temperature, at 0.5 kb, the ‘distribution coefficient” decreases. ‘The opposite trend was found at 2 kb (Fig. 1a of Webster ef al. 1989). No uncertainty is reported here for either of the high-F sets of experiments. There is considerable range in the Li par- titioning data presented here, up to two orders of magnitude (which suggests these results should be Ore-metal Partitioning viewed as tentative), and it seems that additional work is needed to clear up the source of this variability. The limited data of Bos (1990) on Li partitioning suggest that some of the variation presented here may be due simply to increased HF in the aqueous phases and the associated melt, and we tentatively recommend his data be used alone when modelling low-fluorine metaluminous mag- matic systems Magnesium, potassium, calcium, manganese, zine Experiments by Gammon ef al. (1969) and Holland (1972) at 770-880 °C and 1.4-2.4 kb were mainly outside the two-phase region in the NaCl- HO system. Most of the values listed in Tables 1, 2, and 3 for Mg, K, Ca, Mn and Zn are dis- cussed in Candela (1986a) and will not be discussed in detail here. The results of Bos (1990) suggest a value of KEN, = 0.63, which is consistent with the results presented by Gammon et al. (1969) and Holland (1972): Kj, = 0.74. Urabe (1987) studied the effect of pressure (over the pressure range of 1.6-5 kb), and aluminosity (Urabe 1985; at 3.7 kb) on the par- titioning of Zn between magmatic volatile phases and high-silica melts, Urabe's results, based on 10 experiments at 1.6 kb and 800 °C, suggest a value of D3," approaching 11, but with considerable scatter in the data. As Urabe did not analyze the final solutions for Cl or major elements, it i difficult to determine the value for K%‘4,, but these may approach values as high as 50 (assuming that all chloride added to the experimental charge was present in the final solution), or approximately 5 times the value determined by Holland (1972). Urabe (1985) reported that some of this variation may be due to melt composition effects (specifically ASI). Strontium, rubidium, cesium, barium Carron & Legache (1980) performed experi- ‘ments to determine the partitioning behavior of Sr, Rb, Cs, and Ba between magmatic volatile phases and Q-Ab, or Q-Or melts, They reported results as Nemst partition coefficients, orl @.+e)|{erGuee)} 107Candela & Piccoli where # is the element of interest (for an element with a ligation of 1). Some assumptions need to be made to extract exchange constants (see the discussion in Candela 1986a). All reported values of Kir SI for these elements. Bos (1990) performed thirteen experiments for which Rb was measured in run-product solutions and glasses. These results were reported as Dif{! = 0.28mcr' which suggests Khe = 0.61 (40.23), which is in agreement with the results of Carron & Legache (1980). Copper Copper is one of the few elements for which there are both vapor/melt and brine/melt partitioning data (Kw and Ken), and data at supercritical conditions (Ki). Results of Cu partitioning experiments are imperative to our understanding of the genesis of porphyry copper deposits, wherein fluid inclusions containing both low-salinity vapors and saline brines have been identified Candela & Holland (1984) found that Khe = 20 45.6 at 14 kb and 750 °C in a high-silica rhyolite melt, and suggested that Cu has a ligation of 1 in magmatic volatile phases. Keppler & Wyllie (1991), who performed experiments at 2.0 kb and 750°C, obtained qualitatively similar results (but higher) to those of Candela & Holland (1984), but their results probably differed because of the nature of the aqueous starting material (pure HCI-H,0 solution versus mixed chloride-H,0 solutions, respectively) used in the experimental runs. Williams (1995) performed experiments at 800 °C at 1 kb, and 850 °C and 0.5 kb, in the system high-silica rhyolite melt-vapor + brine, Williams (1995) found that Cu/Na in the vapor and brine are equivalent (within error) when the brine composition is calculated, This suggests Wh hit wt that Keaav, Kbvne, and Keth are equivalent in magnitude, and are approximately 215 and I at 1.0 kb (at 800 °C) and 0.5 kb (at 850 °C), respectively. If most of this variation is ascribed to the effect of pressure, then the results from both sets of experiments suggest that Cu. partitioning into 108 magmatic volatile phases reaches a maximum near 1 kb, and is lower at both higher and lower pressures, This is consistent with the observation that many porphyry copper deposits are associated with plutons emplaced at 3-4 km depth, Rare-earth elements Partitioning of select REE between volatile phase(s) and melt was studied by Flynn & Burnham (1978) using Harding Pegmatite as starting material, ‘The original data are reported as Die: the Na concentrations in the run-products were not reported. Candela (1986a) estimated K igen. using several assumptions (for details, see Candela 1986a). The majority of the runs of Flynn & Burnham (1978) were within the two- phase field (mer = 0.45-0.91) although several were in the vapor-only field; however, both sets are considered together in this review. Although four REEs (Ce, Eu, Gd, Yb) were studied in Flynn & Burnham, only La and Yb have been presented in this paper The equilibria of Reed (1995) are presented as exchange constants with respect to Na. Reed performed experiments at 1 kb and 800 °C for the majority of REE. The values in, Table 2 represent an individual experiment (run “107) — estimated uncertainty is based on counting statistics only, ‘The 1-kb experiment of Reed was within the two- phase field (vapor + brine at magmatic tempera- tures and pressures). The Ki for the REE at | kb range from a maximum of approximately 5 for the LREE to 3 for the HREE (Reed 1993) Limited runs at higher pressure (Ce at 1.25 kb; Flynn & Burnham 1978) suggest Kin for the REE decreases with increasing pressure at a fixed temperature Tin Partitioning experiments involving Snare hampered by alloying problems in Pt capsules, Keppler & Wyllie (1991; 750 °C and 2 kb) studied the partitioning of Sn by performing experiments wherein the starting materials consisted of silicate glass and H,0 + HCI HP solutions. Nekrasov ef al, (1980; 850 °C and 1 kb) used H,0 and H,0- HCI as starting solutions. ‘The melts used in theexperiments by Nekrasov ef al. (1980) were in the binary system quartz-albite. In both sets of experiments, only the starting chloride concen- trations were reported, and we have assumed for purposes here that all of the chloride is present in the solution of the run-product. ‘The Sn equilibria presented in this review are from a very limited data set, and the data should be viewed as qualitative. Keppler & Wyllie (1991) showed that par- titioning of Sn in magmatic volatile phases is enhanced in HCI-bearing fluids relative to HF- bearing fluids. Results of these experiments have been modelled assuming that the ligation of Sn is 2, but this has yet to be determined conclusively. Based on the five experiments of Keppler & Wyllie (1991) which used H,0 and H,0-HCI (0.5- 2 molal HCl) aqueous starting solutions, we calculate a value for Kin = 0.3 for Sn-Na exchange between supercritical fluid and haplogranitic melt (at NNO). Nekrasov ef a/. (1980) reported the results of ‘en experiments on Sn partitioning at the W-WO, (Sn”* dominant?) oxygen fugacity buffer, and six experiments at the NNO (Sn‘* dominant?) oxygen fugacity buffer. These experiments were performed at 850 °C and 1 kb, using H,0-only, and 0.1 and 0.5 M HCI aqueous starting solutions (these HCI concentrations are lower than those of all the runs of Keppler & Wyllie 1991), Based on the study by Eugster & Wilson (1985), regarding the hydrothermal solubility of cassiterite, we consider the dominant aqueous species for Sn to be SnCl, at magmatic temperatures and pressures, and under moderately oxidizing conditions. The results of Nekrasov ef af. have been recast as Kil, m8 and Kee, = 0.7 assuming a ligation number of 2 in both cases). No uncertainty was reported by Nekrasov ef al. (1980); a coefficient of variation of up to 25% has been estimated from their data According to the results of Nekrasov er a (1980), the valence of Sn plays an important role in the partitioning of Sn between volatiles and melt phases (possibly up to an order of magnitude). Because of the uncertainty in assessing the correct ligation of Sn, it may not be Ore-metal Partitioning an prudent to suggest that Ks," y, is greater than 1. Furthermore, Sn“ and Sn?" may have variable ligation numbers as a function of chloride in the aqueous fluids. Uranium The most complete set of U partitioning data is from Keppler & Wyllie (1991). These experi- ‘ments were performed using HO + HCI HF CO; aqueous starting solutions. In the HO-HCI experiments, their results show clearly that U partitions into the volatile phase as a cube of the chloride concentration with Kis = 0.40 + 0.38, Furthermore, as the HF or CO; concentration in the fluid increases (in experiments with HF and CO; starting solutions), the U concentration in the fluid incteases. Tt is not clear how U partitions in mixed fluids (HCL-HF, HCI-CO;, HP-CO;), ‘There seems to be an fO; dependence on U parti- tioning into magmatic volatile phases (Keppler & Wyllie 1991). Keppler & Wyllie (1991) found that U has a ligation number of 3 in HCI- dominated fluids, and 2 in HF-dominated fluids Thorium Two sets of experiments concerning Th partitioning have been included here. Keppler & Wyllie (1991) performed experiments between hhaplogranite melts and HO +! HCl + HF starting solutions, No significant Th partitioning in either H,0-only, or H,0-HC! starting-solution experi- ments oceurred (D%,'< 0.01). Thorium has been included as exhibiting Nemst-type partitioning behavior; the value reported here is derived from experiments in 1-4 M HF solutions, In H,0-HF experiments (6 data points), enhanced partitioning ‘was observed (but with considerable scatter). A value of DY” = 0.12 4 0.13 has been adopted here from the HF-only experiments. Webster ef al. (1989) suggested slightly different partitioning behavior for Th relative to the study of Keppler & Wyllie (1991). According to the results of Webster er al. (1989), D%," varies linearly with (mcr)°, but this is somewhat tenuous. This discrepancy may be due to a ‘mixed 109Candela & Piccoli fluid’ effect: the experiments of Webster ef al. (1989), which used topaz rhyolite (>1 wt.% F) as starting material, also contained significant Cl (micy = 0.23-0.26 at 500 bars). ‘Three experiments at 0.5 kb, and 850-950 °C suggest KM! ~ 300. The partitioning of Th may also be dependent upon ASI, and temperature; these effects cannot be ruled out at this time, Clearly, more experimentation on Th is necessary. Tungsten Two sets of experiments (Manning & Henderson 1984; Keppler & Wyllie 1991) have been considered here for the partitioning of W between magmatic volatile phases and silicic melts. The relationship between the data in both sets of experiments is not readily apparent. Tungsten has often been thought to exhibit Nernstian-type behavior, and much of the evidence presented here supports that; however, there is some suggestion that chloride may play @ role in some instances, Neither source presents complete analyses of run-product solutions or ‘melts, so additional interpretation is difficult. Manning & Henderson (1984) performed W partitioning experiments by using a Qz-Ab-Or melt (at | kb and 800 °C) and aqueous starting solutions of Hj0 + NaF + NaCl (several other additional starting solutions were used, but will not be discussed here). All of their experiments which utilized H,O-only as a starting solution, contained W-oxide crystals and have been ignored. Results of three experiments with F added as NaF (0.5-1 M) suggest DY” = 0.74 0.15, with no apparent dependency on F. Seven experiments in which chloride was added to the runs as NaCl (0.5-3.4 M) show an obvious relationship between W partitioning and chloride, The authors suggested that W seems to form complexes with a ligation of 1 at low Cl:W in the fluid, and possibly a fourth- or six-order depen- dence at high mics this is represented in the table mit of equilibrium partitioning data as K’#™! = 3.9 4 1.3 assuming a ligation of 1 which was found at low Cl and W concentrations. All of the NaCl runs were within the two-phase field. 110 Keppler & Wyllie (1991) also performed W partitioning experiments, using haplogranite melt and starting solutions of H,0 + HCI + HF. The runs designated W" in Table 3 represent experiments with starting-solution mixtures of 05-2 MHCI, or 4 MHF. Results of the H,0-only experiments (designated W"' in Table 3) suggest that DY = 3.54 0.9. At 800 °C and 2 kb, data from Keppler & Wyllie (excluding runs which produced W-oxide crystals) are of the same order of magnitude as those of Manning & Henderson (1984) (D4 = 0,72 £0.19). ‘The experiments of Keppler & Wyllie (1991) suggest no dependence on mer. This discrepancy could be due to the fact, that Keppler & Wyllie added Cl to experiments as HCI, as opposed to the NaCI solutions in Manning, & Henderson (1984). Within the variation of the run-products, no discemable difference in partitioning is present in experiments with HCl- and HF-bearing starting solutions. Boron ‘Two sets of B experiments have been included here, both of which yield similar results Pichavant (1981) studied the partitioning of B between haplogranite melt and magmatic volatile phase at 750-800 °C and 1 kb, In that inves- tigation, B partitioning was not studied as a function of the Cl or F content of the fluid. The experiments of Pichavant suggest D@™! ~ 3. ‘These results are similar to those described by London et af. (1988) using macusanite melts (with notably much higher melt aluminosities), London et al, determined a value of DY/! = 2,3 at 775 °C. and 2 kb. Fluorine There is an abundance of F partitioning data in the literature, the majority of which is in high-F systems. The F partitioning data from Webster & Holloway (1990) discussed here represent runs with <10,000 ppm F in the melt (7 data points at 2 kb); these data can be used to derive a temperature-dependent Dz’, At 800 °C the cal- culated value of Dy” is 0.18 + 0.03. Webster &Holloway did note that at very high F concentrations in the melt, Dy’ > 1. Data from Keppler & Wyllie (1981) in 2 haplogranite-HE system (at 750 °C and 2 kb), are based on 3 data points with 1-4 M HF starting solutions, Their results suggest a value for D;” of approximately 0.8. ‘The data presented here suggest that F does not partition strongly into magmatic volatile phases. Molybdenum Two series of experiments on Mo have been included in Table 3 as Nemst-type partition coefficients. Candela & Holland (1984) deter- ‘mined that Mo exhibited Nemnstian-type behavior. ‘They performed experiments at 750 °C and 1.4 kb, and found Dim = 2.5 4 1.6. Limited experiments by Keppler & Wyllie (1991) with melts of composition similar to those of Candela & Holland suggest similar values (DY! = 1.8 + 08 using 1-2 M HE or 1-2 M HCI starting solutions). Two experiments in which water was the only starting solution yielded slightly higher partition coefficients (D7! = 5.5). Summary of Experimental Data Examination of these data suggests that there is a paucity of vapor/melt and brine/melt par- titioning data, expecially at pressures of I kb and below, and that experiments and experimental design need to be designed to retrieve exchange constants, where applicable. Run-products need to be analyzed completely so that phase relations and thermodynamically valid exchange constants can be formulated successfully. There is a wide range of values for several of the elements presented. Many of the experiments of Webster ef al. (1989) have elevated partition coefficients with respect to high-field-strength elements, and Li, ‘Th, and Be, which suggests that FCI mixed- halide complexes may play an important role in the partitioning of these elements. Experiments which use H,0 + HCI as opposed to H,0 + NaCl, etc, in starting solutions commonly gave significantly different results, probably because of anomalous partitioning of Cl, and activity- coefficient effects in the melt phase. ee Ore-metal Partitioning PARTITIONING OF ORE METALS IN NATURAL SYSTEMS Preamble to the Quantitative Model In the early stages of the solidification of a magma, crystallization occurs in the absence of a water-rich vapor phase, During this stage, ore metals are distributed among the melt and a number of crystalline phases. Much of the crystallization that occurs before significant water is lost from the magma to the volatile phase is, most likely, patt of the protracted crystallization and fractionation history that occurs before final plutonic emplacement. These early crystalline products may not reside at the level of exposed pluton, and ore metals sequestered in these phases are lost from the pluton-ore system. On the other hand, some portion of the crystallization that ‘occurs prior to significant water loss from the magma may occur at the level of final plutonic emplacement. Ore metals sequestered in these crystalline products may or may not be available for ore formation. Some proportion of the ore metals sequestered at the level of final plutonic emplacement may be stripped from their host minerals above ot below the solidus. However, the quantitative modelling in this paper does not consider whether sequestering occurs at or below the level of final plutonic emplacement, and the ‘model expressed herein considers metals that are Partitioned into. erystalline phases to be lost permanently from the ore-forming system. ‘The next stage in the evolution of a pluton ore system involves the crystallization of the magma in the presence of a water-rich volatile phase, or volatile-phase mixture (brine plus vapor). In the model presented in this paper, the evolution of the volatiles is fractional, following a Rayleigh law. Further, as in the models of Candela & Holland (1986) and Candela (1989), efficiencies of ore-metal removal from the magma into the volatile phase(s) are calculated. When the calculated efficiency of removal for a given element is multiplied by the ratio of the initial ore- ‘metal concentration in the melt to the initial water concentration in the melt, an “averaged concentration” of the ore metal in the magmatic volatile phase(s) is obtained (as discussed by Candela 1989b). As a pluton crystallizes from the utCandela & Piccoli outside inward because of heat loss to the country rocks (e.g., Brandeis & Jaupart 1987; Candela 1991; Averkin & Candela 1994), volatile exsolution occurs over a range of crystallization progress. In simple terms, this means that an “averaged fluid” gains egress from the pluton, and NOT the “instantaneous aliquots” modelled by Candela (1989) or Cline & Bodnar (1991) (or for example, those pictured in Fig, 2). Averaged concentrations of ore metals and other elements in the magmatic volatile phases may more closely approximate the fluids that are evolved from magmas, and are probably the best quantity to compare to fluid-inclusion compositions. Future modifications of this chemical model will include, explicitly, considerations of the physics of volatile exsolution, Candela (1994) pointed out that the evolution of the volatile phases from a crystallization interval may approximate a sequence of “batch vaporization” followed by “quasi-Rayleigh fractionation” (intermediate between pure batch and pure fractional crystallization). An exception to the production of an “averaged fluid” may occur when a water~ saturated melt is intruded to a high level in the earth's crust. Crystallization of the shallowly- emplaced magma may occur under non- equilibrium conditions due to the reduction in confining pressure below the vapor pressure of the melt, and the concomitant increase in the solidus temperature. In this case, the melt undercools, and crystallizes irreversibly. With crystallization only loosely tied to the variation in temperature with distance from the contact, a well-mixed magma may evolve volatiles from many points within the chamber, and not just from the crystallization interval. A given composition fluid or fluid mixture will be evolved from numerous sites in the pluton, and the composition of the volatile phase(s) will change progressively with time as the magma crystallizes and exsolves volatiles irreversibly. Quenched caps on plutons, or irregular zones of variably textured granite, characterized by dendritic quartz. crystals and/or micrographic textures, or zones of interconnected mmiarolitic cavities (Candela & Blevin 1993) may represent magma that has crystallized and devolatilized in this way. 2 The Quantitative Model: A Modification of Candela (19894) We have written an algorithm, MVP (Magmatic Volatile Phases), which can be used to calculate the composition of the magmatic volatile phase(s), vapor or vapor + brine, that are iso- barically expelled from granitic (s./.) magmas. This phenomenon of crystallization-driven volatile exsolution in plutonic environments has been referred to as second boiling (e.g., Bowen 1928). Briefly, this algorithm models volatile saturation of a melt of a given bulk compo: and initial water concentration, at a specified temperature and pressure. Crystallization is modelled implicitly through the use of bulk crystal/melt partition coefficients for each element, Once the composition of the melt is calculated at volatile saturation (which is ‘equivalent to water saturation in this algorithm), the composition of successive aliquots of vapor, or vapor + brine are calculated, until the volatile- saturated melt has crystallized. Sufficient data do not exist yet for the composition of the brine to be calculated accurately; however, until such data are procured, we believe that the results presented here represent the most rigorous treatment possible at this time, and are illustrative of the effects of brine saturation, As more data are obtained from laboratory investigations, they can be added to this database. The end product of this modelling includes: 1) the elucidation of general trends in the calculated efficiencies with which metals are removed from crystallizing magmas into evolving aqueous/saline “ore fluids”; and 2) estimates of the concentrations of oremetals in the volatile phases. Appendix I and its accompanying flow chart provide a detailed description of the operation of the program. In the flow chart, numbers next to each heading relate to keyed comments in Appendix 1 ‘Modelling results We have modelled isothermal, isobaric volatile exsolution at $00 °C and 1 kb. Bulk Ds (bulk partition coefficients, or mineral weight- fraction weighted sums of the individual mineral/melt partition coefficients for any given element) are treated as constant throughoutj | | | i 1 | magma evolution for most elements, However, as discussed by Candela & Holland (1986) and Lynton et al. (1993), the bulk D for Cu should change from greater than one (which models “compatible” behavior), to a value much less than ‘one (which models incompatible behavior) at water saturation, because of: 1) the importance of small amounts of magmatic sulfide in the sequestering of Cu (Lynton ef ai, 1993), and 2) the fact that felsic melts seem to be saturated with a magmatic sulfide phase early in their history, but not at the later stages of crystallization (ce, sulfides are not uncommon as inclusions in early- formed minerals, but glass/sulfide contacts are rare in voleanic rocks). In this algorithm, saturation with a CO,-free vapor, or a CO,-free vapor + brine mixture can be modelled. Implicitly, the melt composition is modelled as "near minimum", and does not change in major-element composition with crystallization. ‘The algorithm routinely considers Cu, Mn, Zn, Ca, Mg, Yb, Ce, K, Na, Eu (exchange partitioning), and Mo, W, and B (Nemst-type behavior). It can easily be modified to include all of the elements in Tables 1, 2, and 3. Although a thermodynamic model for Cl partitioning is under development, we have not incorporated it into this ‘model at this time. An outline of the thermo- dynamic model for Cl is presented in Appendix 2. In the current version of the algorithm, a constant partition coefficient is used. There are sufficient data on melt/vapor partitioning to allow a first- order estimate of model efficiencies of removal of ore metals from the melt into the associated vapor. However, that is not true for the case of melt/brine equilibria. Based on data from our lab, we can estimate melt/brine equilibria for Cu and HCL, For other elements we assume that the vapor/melt exchange constant can be used as an estimate of the brine/melt exchange constant. That this is not unreasonable is shown by Williams ef al. (1995) for Cu, Nemst-type elements (e.g, Mo, F, and B) are modelled as partitioning into brine in Proportion to the ratio of water in the brine relative to the vapor; this approach is highly and should be taken only as a ation at this time, Therefore, in this tion, we emphasize the behavior of Cu in melt-vapor-brine systems; the consideration of Ore-metal Partitioning weeded so that a reasonable balance in the brine can be other elements estimate of Cl-c made. Our computations have a number of consequences for the evaluation of the behavior of chloride-complexed cations in magmatic hydro- thermal processes (Table 4 shows the initial concentrations and bulk crystal/melt partition coefficients for elements that were used in the computations). At any given temperature and pressure near the granite solidus, there is a unique weight ratio of Cl to HzO in the melt (at 800 °C and | kb, CVH;0 » 0.05) at which the melt is in equilibrium with both a vapor and a brine. When ‘a melt saturates with respect to a volatile phase at lower values of this ratio, at any given temperature and pressure, the system is vapor- saturated and the chloride concentration increases in the melt until brine + vapor saturation is reached. For melts that become vapor-saturated at low pressures, the usual course of evolution will be from a state of vapor saturation to one of vapor + brine saturation. As only melts that saturate Table 4, Bulk crystal/melt partition coefficients before/after volatile saturation (at NNO and MW), and average composition (in ppm) of Lachlan Fold Belt /-type granite used in the calculations as for Table 1. D before D after ‘Concen- Volatile Volatile tration Saturation: Saturation; (Ppm) NNO(MW) _NNO(MW) or 2Q) 0:2(0.2) 10 Mn 0.5(0.3) 0.5(0.5) 350 Zn 1) 1) 49 ca 1a) 10) 2500 Mg I(t) 1a) 8900 Yb 0.27027) 0.27(0.27) 83 Ce 1531.53) 1.53(1.53) 66 K 10) 10) 28700 Na 1a) 1) 23400 B 0.10.1) 0.10.1) 20 Mo — 0.2(0.4) 0.20.4) 1 Ww 0.50.4) 0.5(0.4) $ cal (0) (0) . Bu__1.612.48) 1.61.48) 12. “indicates that the concentration of Cl was varied in individual runs 113Candela & Piccoli with respect to vapor or vapor + brine are considered in this paper, we will limit our discussion to melts with Ci at or below the concentration on the vapor + brine cotectic at the stated pressure, At 1 kb and 800 °C, the critical concentration of Cl on the vapor-brine cotectic is approximately 2200 ppm given a vapor/melt partition coefficient for Cl equal to 5. At 500 bars, the critical Ci is just under 500 ppm. It certainly is possible, however, for a melt to become saturated with respect to brine only, ie., the melt saturates with respect to a hydrous molten salt mixture, but the vapor pressure of the volatiles has not yet reached the load pressure. According to the data summarized in Webster (1992a) and Metrich & Rutherford (1992), meta~ luminous melts saturated with NaCl are not very different from melts saturated with respect to a vapor that is near brine saturation, However, vapor-absent brine-saturated systems are not considered in this modelling effort. Table 5 shows an example of select results from a typical run of the MVP algorithm. In this example, a metaluminous oxidized [type granite melt, with approximately 2 wt.% H,0 and 530 ppm Cl (a reasonable Cl concentration for a Sierran granite; Piccoli & Candela 1994) is crystallized at I kb. Vapor saturation occurs at 50% crystallization, and brine saturation occurs at 80% crystallization, The example in Table 5 serves as a reasonable guide of the processes modelled by a typical MVP run. ‘The following paragraphs, and the accompanying Figures, ‘Table 5, Example of crystallization of hydrous felsic, low-COz magma with exsolution of volatiles Initial melt: magma = (100% melt as a limiting case) + major-clement composition » average Lachlan Fold Belt J4ype granite (see Table * C1 elt) = 530 ppm # 1,0 (melt) = 2.1 wt. % also, ASI=1.01, and 0,= NNO. + pressure ~ 1000 bars, temperature (at volatile saturation) = 800 °C ‘At 80% crystallization, the melt saturates with respect to VAPOR, and * Cl (melt) = 1060 ppm * Cl (vapor) = $120 ppm, (0.84 wt. % NaCl eq) @ Der=$ + H,0 (melt) =42 wl % [AU 75% erystallization (50% of water exsolved), the volatile phase is still “vapor only”, and i (melt) = 1830 ppm * i vapor) = 9160 ppm (1.3 wt %6 NaCl eq) + 11,0 (mel) = 42 w1.% ‘At 80% crystallization (60% of water exsolved), the melt saturates with respect to BRINE (system is now vapor + brine, Saturated). The varianee of the system drops by 1, and the conceatration of Cl in the melt is now eonstant, A constant ClH,O patio in the melt at vapor + brine saturation fixes the vapor/brine weight ratio in the exsolved product * Cl (melt) = 2182 ppm 10,900 ppm (1.8 wt. % NaCl eq) 68,000 ppm (60.8 wt, % NaCl ea.) + H,0 (mel = 42 wi% * brine(vapor + brine) = 0.11 by weight The concentrations of CI and water are constant in the melt, vapor and brine from 80 % 10 100% crystallization (in most nagmas, this interval of crystallization is nearly isothermal). However, other unbuffered concentrations (eg, Cu) can vary freely. In the above simulation, the concentration of Cu in the initial melt was 10 ppm. At brine saturation, the Cu concentration in the 20% melt remaining was 4 ppm (the effet of prolonged crystal fractionation before volatile exsolution, together with vapor-only exsolution prior to brine sat ation). The concentation of Cu in the vapor and brine at that point = 90 and $100 ppm, fespectively. The Cu concentration inal three phases plammets after brine saturation. The overall efficiency of removal of Cu is 14%, 14Efficiency (2%) [vapor + brine] Ore-metal Partitioning Cu Mn Zn Ca Mg Yb Ce K Na B Mo W Eu CiWater HBB 0.001 Init CWater BRI 0.08 Ir Figure 1. Efficiencies of removal of selected elements from the melt into the exsolving volatile phases forthe parameters given in Table 4, at two different CVE,O ratios. Caleulations were performed for a pressure equal to 1000 bars, an AST of 1.01, and an initial water concentration of 4 wt. illustrate how the efficiency of removal of Cu from a melt into exsolving volatile phases varies as a function of selected magmatic parameters. ‘The effect of increasing Cl/H2O of the starting melt (or concentration of Cl at a given initial chloride concentration in the melt, Ci) on the efficiency with which selected elements are removed into an evolving vapor or vapor + brine mixture is illustrated in Figure 1. Chloride has a great effect on elements such as Cu, Mn, Zn, Yb, and Ce. The effect is most dramatic for elements with ligation numbers greater than unity. For example, on the order of 10% or greater of the REE can be mobilized at reasonable values for CVH,O (e.g., 0.05 by weight) even when these elements are modelled as behaving compatibly ‘with respect to crystallizing phases. Copper, of course, pattitions strongly into the magmatic volatile phases for similar CI/H,O values. At a pressure of 1 kb (see Fig. 2a,b), the volatile phase/melt exchange constants for Cu are very high, and Cu is efficiently removed from the magma into the volatile phases. The concen- tration of Cu decreases as volatile exsolution progresses. However, at 500 bars (not shown), the volatile phase/melt exchange constants for Cu are lower; therefore, the concentration of Cu in successive aliquots of the brine and vapor increases as the volatile-saturated melt crys- tallizes. The study of Candela & Bouton (1990) showed that the behavior of Mo and W during crystallization depends upon the oxygen fugacity. In general, as fO, increases, Mo becomes more usCandela & Piecoli 0.017—— gee 0.05: ge 0024 0.001 Concentration of Guin Vapor (molest) 0.03 0.025: 0.02: ors: oor Concentration of Cures) 0.005: or 02 / 0.08 38 Bes centration of Cu in Brine (moles) 0.05 88 § Con 0304 05 06 d7 08 09 1 (Reaction Progress) Vapor Only ee aes on os 04 os 06 o7 da oa 4 Z (Faction Progress Vatiabi) [Vena Brine on: Concentration of W (moles) o.o001 o 116 on oz os 06 o7 ds o9 1 03_ Oa Z (Reaction Progress) por— Brine | Figure 2 Concentrations of Cu (a and b) and W (c) in the volatile phases as a function of devolatilization (reaction) progress under the specified conditions. Al calculations were performed at 800 °C, 1000 bars, for melts with an ASI= 1.01, and an oxygen fugacity of NNO. ‘The concentration of Cu (a and b) was calculated using Clwater of 0.05 and 0.025, respectively, and initial water concentrations (melt) of 4 wt.%. The concentration of W (c) was calculated tial water concentration of The partitioning of W between volatile phases and melt is modelled by using Nernst-type volatile phase/melt pattition coefficientsincompatible, and W becomes more compatible, The experiments were performed at 800 °C and 1 kb, and for a small selection of minerals; the general behavior, however, can be extended to more complex systems. The vapor/melt partition coefficients for Mo and W are not high enough to deplete the volatile-saturated melt in these constituents during volatile exsolution (see, for example, Fig. 2c). We have calculated. the efficiency with which Mo and W can be removed from a devolatilizing felsic magma, We used Duo = 0.2 and Dy = 0.5 at 0) = NNO (nickle-nickle oxide), and Dy = Dy = 0.4 at Oy = MW (magnetite-wilstite, which is close to #0 = GM). Figure 3. shows that Mo partitioning into the volatile phase is enhanced, relative to W, as the oxygen fugacity is increased from MW (characteristic of reduced granites such as the S- type granites of the Lachlan Fold Belt) to NNO, characteristic of the oxidized [type granites of the same area. The compatible elements decrease in concentration as crystallization proceeds, and incompatible elements increase in concentration during crystallization, Early development of Figure 3. Ratios of the efficiencies of removal of Mo and W as a function of the initial water concentration in the melt at two different oxygen fugacities. MW magnetite — wilstite (model S-lype oxygen fugacity); NNO= nickel — nickel oxide oxygen fugacity (model /- type oxygen fugacity). Calculations were performed for a pressure equal to 1000 bars. ‘The partitioning of Mo and W between volatile phases and melt is modelled by using Nernst-type volatile-phase/melt partition coefficients, Ore-metal Partitioning water-ich volatile phase is necessary for the efficient removal of compatible elements, such as Cu and some REE (Fig. 4). Removal of ore metals from the system is best demonstrated by examination of differentiation parameters such as the whole-rock MgO concentration (see Blevin & Chappell 1992, 1995). At a given load pressure, the initial water concentration will determine the timing of volatile saturation. Initial water concentrations can be related to the timing of saturation of a magma with respect to a water-rich phase whether or not the melt is COs-bearing. Other factors that increase Cu partitioning into “ore fluids” from ing magmas are: 1. Higher fO, in mt-saturated systems: this lowers the FeCly concentration (see number 2 below) in the magmatic volatile phases, and therefore makes more Cl available for complexation with Cu and other Cl- complexed elements. 2. Lower ASI in the melt: at ASI near unity, the HCl concentration and = FeCl, concentration (for mt-saturated systems at a given JO;) in the volatile phases are minimized in metaluminous to _per- aluminous magmas (see discussion in Candela 1989a); both HCI and Fe compete with Cu and other metals for Cl, and all Cl- complexed metals experience a decrease in the efficiency of removal asthe concentration of HCI and FeCl, increase (see Fig. 5), 3. Lower Na,O and K,0 in the melt (at a given melt ASD. As the total cation inventory of a melt increases, competition for Cl becomes acute, and overall efficiencies of removal decrease. Discussion ‘The probability of mineralization with respect to Cl-complexed ore metals increases with increasing CHO of the associated magma. As brine can form at relatively low magmatic CIO, the presence of brine + vapor-tich inclusions alone is not a good indicator of high CIM,0 of the initial melt. Model Cl u7Candela & Piccoli 100 90 80: ; 70: 60 50- 40 30 20! | 10 0- Efficiency (%) [vapor + brine] Cu Mn Zn Ca Mg Yb Ce K Na B Mo W Eu 04 Initial Water | Efficiency (%) [vapor + brine] 20: 10: Cu Mn Zn Ca Mg Yb Ce K Na B Mo W Eu Wy Asi=1.01 BB Asi-1.1 118 iconcentrations can be obtained from electron- probe microanalysis of apatite (Piccoli & Candela 1994), or from analyses of melt inclusions (Appendix 3) in comagmatie volcat rocks. The efficiency of removal of REE from a magma into exsolving volatile phases is a function of both the CVHO ratio of the magma and the extent to which the REE are taken up by crystallizing phases. For melts with low initial CVO ratios (0.01), brine saturation occurs relatively late if at all, and overall efficiencies of removal of chloride-complexed metals are low. *At the Cl concentrations that are found commonly in felsic magmas (see Appendix 3), saturation of magmas with respect to vapor + brine should not be uncommon at pressures below about 1.4 kb, The CV/H,O ratio in the melt is fixed at a given temperature and pressure when the system is saturated with respect to vapor and brine; further, the ratio of vapor to brine, and the composition of these phases, remain constant, The concentration of un- buffered trace elements (e.g., Cu in the absence of intermediate solid solution) may vary in the system, and may follow Rayleigh-type fractionation patterns, whereas Cl cannot (Table 5), Therefore, increases in magmatic Cl concentrations are not necessary consequences of crystal fractionation, even at low pressures. *Cu partitioning from magmas into Clbearing volatile phases is favored near | kb. Pressures of magmatic devolatitization can be estimated in a number of ways; however, most means of estimating pressure are fraught with difficulty, especially under epizonal conditions. ‘The chloride concentration in vapor-saturated brine in magmatic (quartz phenocryst) fluid inclusions, trapped near the solidus, may yield information on pressure. Pressure estimates, together with estimates of initial water concentrations, can yield information on the timing of saturation with respect to a water-rich Ore-metal Partitioning volatile phase relative to crystallization of early- crystallizing phases that can sequester ore ‘metals. At a given pressure of emplacement, the initial water concentration will determine the timing of volatile saturation. Initial water concentrations can be related to the timing of saturation of a magma with respect to a water-rich phase whether or not the melt is COy-bearing. Other factors, suggested by quantitative modelling, that may increase Cu partitioning into “ore fluids” are: higher fO, in mt-saturated systems because of both the destabilization of magmatic sulfides and the suppression of aqueous FeCly; lower ASI in the melt (for metaluminous to peraluminous compositions) because hydrogen competes with other cations for chloride, and also increases FeCl, in mt-saturated systems; and lower NaO and K,0 in the melt (at a given melt ASD), because of competition for chloride in the volatile phases. ACKNOWLEDGMENTS We thank M. Reed and T.J. Williams for providing partitioning data and for numerous discussions, We also thank H. Shinohara and J. Hemley for extensive reviews, and FH. Thompson for an insightful review and for inviting this contribution. We would like to acknowledge support from National Science Foundation grant EAR-9204671, and Department of Energy grant DEFGO7-901D13025 REFERENCES AVERKIN, Y.A. & CANDELA, P.A. (1994): Thermo- ‘and fluid dynamic effects of expansion in the ‘model granitic system Ab-Q-water during the crystallization and volatile separation. Trans, Am Geophys. Union 75, 363 (abstr) Figure 4 (p. 118, top). Efficiencies of removal of selected elements from the melt into the exsolving vapor or vapor + brine for the parameters given in Table 4, at two different initial water concentrations in the melt. Figure § (p. 118, bottom), Bificiencies of removal of selected elements from the melt into the exsolving vapor or vapor + brine for the parameters given in Table 4, at two different values for the Aluminum Saturation Index (ASI), 119Candela & Piccoli BACON, CR., NEWMAN, S. & STOLPER, E. (1992) Water, CO;, Cl, and F in melt inclusions in phenocrysts from three Holocene explosive eruptions, Crater Lake, Oregon. dm. Mineral. 77, 1021-1030, BLEVIN, P.L. & CHAPPELL, B.W. (1992): The role ‘of magma, oxidation states, and fractionation in determining the granite metallogeny of eastern ‘Australia, Trans, Royal Soc. Edinburgh, Earth Sci 83, 305-316. BLEVIN, P.L. & CHAPPELL, BW, (1995): Chemistry, origin and evolution of mineralized granites in the Lachlan Fold Belt, Australia: The metallogeny of I- and S-type granites. Econ, Geol. (accepted). BODNAR, RJ., BURNHAM, C.W. & STERNER, S.M. (1985): Synthetic fluid inclusions in natural quartz, Ill. Determination of phase equilibrium properties in the system H,O-NaCl to 1000°C and 1500 bars. Geochim. Cosmochim. Acta 49, 1861- 1873, BOS, A. (1990): Hydrothermal elemental distributions at high temperatures: An experimental study of the partitioning of major and trace elements between phlogopite, haplogranitic melt and vapor. Geol. Ultraieetina, Mededelingen van de Faculteit Aardwetenschappen der Rijksuniversiteit et Utrecht 69, 99 p. BOWEN, NL. (1928): The Evolution of Igneous Rocks Dover Publications, Inc., New York BRANDEIS, G. & JAUPART, C. (1987): Crystal sizes in intrusions of different dimensions: constraints fon the cooling regime and the crystallization kinetics. Jn Magmatic Processes: Physiochemical Principles (B.. Mysen, ed). The Geochemical Society, New York, 307-318, CANDELA, P.A. (1986a): Generalized mathematical ‘models for the fractional evolution of vapor from ‘magmas in terrestrial planetary crusts. Jn Advances in Physical Geochemistry 6: Chemistry and Physics of the Terrestrial Planets (S.K. Saxena, ed). Springer-Verlag, New York, 362-396. CANDELA, P.A. (19866): Toward a thermodynamic ‘model for the halogens in magmatic systems: an application to melt-vapor-apatite equilibria, Chem, Geol, $7, 289-301 120 CANDELA, P.A. (19899): Magmatic ore-forming fluids: "thermodynamic and mass transfer calculations of metal concentrations. In Ore Deposits Associated with Magmas (J.A. Whitney & Ad. Naldret, eds.) Reviews Econ. Geol. 4, 203- 22 CANDELA, P.A. (1989b): Calculation of magmatic fluid contributions to porphyry-type ore systems: predicting fluid inclusion chemistries. Geochem. J. 23, 295-308, CANDELA, P.A. (1990): Theoretical constraints on the chemistry of the magmatic aqueous phase. Jn Ore- bearing Granite Systems; Petrogenesis and Mineralizing Processes (H.J. Stein & J.L. Hannah, eds,), Geol. Soe, Am, Speclal Paper 246, 11-19. CANDELA, P.A. (1991): Physies of aqueous phase exsolution in plutonic environments. Am. Mineral. 76, 1081-1091. CANDELA, P.A. (1994): Combined chemical and physical model for plutonic devolatilization: A non-Rayleigh fractionation algorithm. Geochim. Cosmochim, Acta 58, 2157-2167. CANDELA, P.A. & BLEVIN, P.L. (1993): Textural evidence for granitic volatile phase permeability: observations from the Lachlan and New England Fold Belts of easter Australia. Geol. Soc. Am. Abstr. Programs 25, 401. CANDELA, P.A. & BOUTON, S.L. (1990): The influence of oxygen fugacity on tungsten and molybdenum partitioning between silicate melts and ilmenites. Econ. Geol. 88, 633-640. CANDELA, P.A. & HOLLAND, HD. (1984): The partitioning of copper and molybdenum between silicate melts and aqueous fluids. Geochim. Cosmochim, Acta 48, 373-388 CANDELA, P.A. & HOLLAND, H.D. (1986): A mass transfer model for copper and molybdenum in ‘magmatic hydrothermal systems: ‘The origin of porphyry-type ore deposits. eon. Geol. 81, 1-19, CARRON, J.P. & LAGACHE, M. (1980): Etude expérimentale du fractionnement des éléments Rb, Cs, Sr, et Ba entre feldspaths alcatins, solutions hhydrothermals et liquides silieatés dans le systéme Q.Ab.Or-H,O a 2 kb, entre 700 et 800°C. Bull Minéral, 103, 571-578.CLINE, 18. & BODNAR, RJ. (1991): Can economic porphyry copper mineralization be generated by a typical calc-alkaline melt? J. Geophys. Research 96, 8113-8126. DUNBAR, N. & HERVIG, R.L. (1992a): Petrogenesis| ‘and volatile stratigraphy of the Bishop Tuff: Evidence from melt inclusion analysis. /. Geophys. Research 917, 15,129. DUNBAR, N. & HERVIG, RL. (1992b): Volatile and trace element composition of melt inclusions from the lower Bandelier Tuff: Implications for magma chamber processes and eruptive style. J. Geophys Research 97, 15151-15170. DUNBAR, N.W. & KYLE, PR. (1992); Volatile contents of obsidian clasts in tephra from the ‘aupo Volcanic Zone, New Zealand: Implications to eruptive style. J. Voleanol. Geotherm, Research 49, 127-145, BUGSTER, H.P. & WILSON, G.A. (1985): Transport ‘and deposition of ore-forming elements in hydrothermal systems associated with granites. High Heat Production Granites, Hydrothermal Circulation, and Ore Genesis (C. Hall, ed), Institute of Mining and Metallurgy, St. Austell, England, 87-98, FLYNN, R-T. & BURNHAM, CW. (1978): An experimental determination of rare earth partition coefficients between a chloride-containing vapor phase and silicate melts. Geochim. Cosmochim. Acta 42, 685-701. FRANTZ, .D., POPP, RK. & BOCTOR, NZ. (1981): Mineral-solution equilibria—V. Solubilities of rock-forming minerals in supercritical fluids. Geochim. Cosmochim, Acta 45, 69-11. GAMMON, I.B., BORCSIK, M. & HOLLAND, ILD. (1969): Potassium-sodium ratio in aqueous solutions and co-existing silicate melts, Seience 163, 179-181 HEDENQUIST, J.W. & LOWENSTERN, J.B. (1994) The role of magmas in the formation of hydrothermal ore deposits, Nature 370, 519-527, HERVIG, R.L, DUNBAR, N., WESTRICH, H.R, & KYLE, PR, (1989): Pre-eruptive water content of thyolitic magmas as determined by ion microprobe Ore-metal Partitioning analyses of melt inclusions in phenocrysts. J Volcan. Geotherm. Research 36, 293-302, HOLLAND, H.D. (1972): Granites, solutions and base ‘metal deposits. Eon, Geol, 67, 281-301 KEPLER, H. & WYLLIE, PJ. (1991): Partitioning of Cu, Sn, Mo, W, U, and Th between melt and aqueous fluid in’the systems haplogranite-H,0— HCI and haplogranite-H,0-HP. Contrib. Mineral. Petrology 109, 139-150, KILINC, LA, & BURNHAM, C.W. (1972); Par- titioning of chloride between a silicate melt and coexisting aqueous phase from 2 to 8 kilobars, Econ, Geol. 67, 231-235. LAGACHE, M. & WEISBROD, A. (1977): The system: two alkali feldspars — KCI-NaCI-H,O at moderate to high temperatures and low pressures. Contrib, Mineral. Petrology 62, 77-101 LONDON, D., HERVIG, RL. & MORGAN, G.B,, (1988): Melt-vapor solubilities and elemental partitioning in peraluminous granite-pegmatite systems: experimental results with Macusani glass at 200 MPa, Contrib. Mineral. Petrology 99, 360- 373, LOWENSTERN, J.B. (1994): Chlorine, fluid immiscibility, and degassing in _peralkaline ‘magmas from Pantelleia, Italy. dm. Mineral. 79, 353-369, LYNTON, 8.J,, CANDELA, P.A. & PICCOLI, P.M. (1993): Experimental study of the partitioning of copper between pyrrhotite and a_ high-silica thyolitic melt, Econ, Geol. 88, 901-915, MANNING, D.A.C. & HENDERSON, P. (1984): The behaviour of tungsten in granitic meltvapor systems. Contrib. Mineral, Petrology 86, 286-293, METRICH, N. & RUTHERFORD, MJ. (1992): B perimental study of chlorine behavior in hydrous silicate melts. Geochim, Cosmochim, Acta 86, 607- 616. NEKRASOV, L:Y., EPEL'BAUM, MB, & SOBOLEV, V.P, (1980): Partition of tin between melt and chloride fluid in the granite-SnO-SnO,-fuid system. Doklady Earth Sel. Section 282, 165-168. 121Candela & Piccoli PASTERIS, J.D, WOPENKA, B, WANG, A. & HARRIS, TN. (1994): A possible factor in the high SO, emissions of the Mt. Pinatubo eruption: non-synchronous saturation of the magma with fluid and anhydrite. Geol Soe. Am. Abstr Programs 26, 452. PERKINS, D., ESSENE, EJ. & WALL, VJ. (1987): ‘THERMO: A computer program for calculation of mixed-volatile equilibria. Am. Mineral. 72, 446- 447. PICCOLI, P.M. & CANDELA, P.A. (1994): Apatite in felsic rocks: a model for the estimation of initial halogen concentrations in the Bishop Tuff (Long Valley) and Tuolumne Intrusive Suite (Sierra Nevada Batholith) magmas, Am. J Sci. 294, 92- 135, PICHAVANT, M. (1981): An experimental study of the effect of boron on a water saturated haplogranite at 1 kb vapor pressure, Contrib Mineral. Petrology 76, 430-439. REED, MJ. (1995): Disiribution of Rare Earth Elements between Aqueous Fluids and Granitic ‘Melt, Ph.D. thesis, Univ. Maryland at College Park (anticipated. SHAW, H.R. & WONES, D.R. (1964) Fugacity coefficients for hydrogen gas between OC and 1000°C, for pressures to 3000 atm. Am. J. Sei, 262, 918-929, SHINOHARA, H. (1987): Partition of Chlorine Compounds in the System Silicate Melt and Hydrothermal Solutions. D.Se. thesis, Tokyo Institute of Technology, Tokoyo, Japan. SHINOHARA, H. (1994): Exsolution of immiscible ‘vapor and liquid phases from a crystallizing silicate tmelt: implications for chlorine and metal transport Geochim, Cosmochim. Acta 58, 5215-5221. SHINOHARA, H., IIYAMA, J.T. & MATSUO, S. (1984): Comportement du chlore dans les systémes ‘magma granitique-eau, C. R. Acad. Sei. Paris 298, 741-743, SHINOHARA, H., IIYAMA, 1.7. & MATSUO, S. (1989): Partition of chlorine compounds between silicate melts and hydrothermal solutions: 1 Partition of NaCI-KC. Geochim, Cosmochim. Acta 53, 2617-2630, 122 STERNER, $.M,, CHOU, I-MING, DOWNS, RT. & PITZER, K.S. (1992): Phase relations in the system NaCI-KCL-H,O: V. Thermodynamic-PTX analysis of sofid-liquid equilibria at high temperatures and pressures. Geochim. Cosmochim. Acta $6, 2295- 2309. SVERJENSKY, D.A., HEMLEY, J.J. & D'ANGELO, WM. (1991): Thermodynamic assessment of hydrothermal alkali feldspar-mica-aluminosilicate equilibria, Geochim. Cosmochim. Acta 55, 989- 1004, URABE, T. (1985): Aluminous granite as a source magma of hydrothermal ore deposits: an ex perimental study. con, Geol. 80, 148-157. URABE, T. (1987): The effect of pressure on the partitioning ratios of lead and zinc between vapor and thyolitic melts. Seon. Geol, 82, 1049-1052, WEBSTER, J.D. (1990) Partitioning of F between HO and CO, fluids and topaz rhyolite melt: Implications for —mineralizing magmatic hydrothermal fluids in F-tich granitic systems. Contrib, Mineral. Petrology 104, 424-438. WEBSTER, J.D. (1992a): Water solubility and chlorine partitioning in Clrich granitic systems: Effects of melt composition at 2 kbar and 800°C, Geochim Cosmochim. Acta 86, 659-678. WEBSTER, J.D. (1992b): Fluid-melt interactions involving Clich granites: Experimental study from 2 to 8 kbar. Geochim. Cosmochim. Acta 56, 679-687, WEBSTER J.D. & HOLLOWAY, JR. (1988): Experi ‘mental constraints on the partitioning of Cl between topaz rhyolite melt and HO and H,O + CO; fluids: New implications for granitic differentiation and ore deposition. Geochim Cosmochim. Acta 82, 2091-2108. WEBSTER, J.D, & HOLLOWAY, J.R. (1990): Par- titioning of F and Cl between magmatic hydrothermal fluids and highly evolved granitic magmas. In Ore-bearing Granitic Systems; Petrogenesis and Mineralizing Processes (H.I. Stein & LL, Hannah, eds,). Geol. Sac. Am. Special Paper 246, 21-34. WEBSTER, J.D., HOLLOWAY, JR. & HERVIG, RL (1989): Partitioning of lithophile trace elementsbetween H,0 and HO + CO; fluids and topaz rhyolite melt, Beon, Geol. 84, 116-134 WILLIAMS, TJ. (1995): Copper and HCI in Felsic ‘Magmatic Systems. Ph.D. thesis, University of Maryland at College Park, Maryland, WILLIAMS, T.J., CANDELA, P.A. & PICCOLI, PM. (1995): The partitioning of copper between silicate ‘melts and two-phase aqueous fluids: an experimental investigation at | kilobar, 800 °C and 0.5 Kilobar, 850 °C. Contrib. Mineral. Petrology (in revision), ZHU, C. & SVERJENSKY, D.A. (1991): Partitioning of F, Cl and OH between minerals and hydro- thermal fluids. Geochim, Cosmochim, Acta 58, 1837-1858, APPENDIX 1 Annotation to the MVP Flow Chart (Fig. Al-1) This appendix has been included to provide supplemental information to the flow chart for MVP calculations presented in Figure Al-1. ‘The ‘general’ mathematical approach used in these calculations is the same as that described in Candela (1986a, 1989a). Step 1: Pressures used should be consistent, broadly, with the partitioning data for individual elements. Oxygen fugacity affects the solubility of magnetite. At lower oxygen fugacities, the FeCl, concentration in the volatile phases is higher. Use of CV/H,O ratios outside the range of 0.01-0.05 may result in crashes of the program in some circumstances. Specifying pressure yields a “saturation water concentration” for the melt. Step 2: Only a limited range of pressures can be used; otherwise, vapor-brine equilibria cannot be properly modelled. The initial water concen- tration chosen in Step 1 cannot be greater than the saturation water concentration of the melt at the specified pressure; ie., the algorithm does not treat magmas that carry excess water to the site of emplacement as bubbles. Also, the initial CHO ratio in the melt cannot be greater than the value for this variable at vapor + brine saturation. Ore-metal Partitioning Step 3: Data are read in from disk and include: 1) initial melt composition (in this case, an ‘average’ /-type granite from the Lachlan Fold Belt; Table 4); 2) bulk partition coefficients (Table 4); and 3) equilibrium partition coefficients (Tables 1 and 2). Given the initial and saturation water concentrations of the melt specified in Step 1, and the bulk Ds (Table 4), the composition of the melt at the commencement of volatile exsolution is calculated. Step 4: The chloride concentration in the vapor is calculated by using its vapor/melt partition coefficient Step 5: The result from Step 4 is used to evaluate whether the system is brine-saturated. Step 6: If the system is brine-saturated, then CI/H,0 of the melt, vapor, and brine are fixed (see the discussion in the text). Step 7: Concentrations of elements that can be approximated as following Nemnst-type par- titioning behavior (Candela 1990) are calculated in the melt, and volatile phase(s). Elements which exhibit Nemnst-type behavior include, e.g., B, Mo, and possibly W (Table 3). Step 8: Estimation of Na in the vapor is critical as all exchange constants are formulated as exchanges with respect to Na. Sodium is calculated from a “charge-balance” (really a chloride-cation balance) in the volatite phase(s) by using initial estimates of the concentration of Cl-complexed elements (see Candela 1989a), Step 9: Refinement of the concentration, in the volatile and melt phases, of elements which complex with chloride (with the exception of Fe, H, and Na) including Mg, K, Ca, Rb, Sr, Mn, Cu, Zn, and REE (in addition to Li, Be, Cs, Ba, Th, U, and Sn). Step 10: Calculation of FeCly in the volatile phases is accomplished by considering the equilibrium: FeO, + 6HCI +H; @ 3FeCh +410 (7) from Frantz ef al, (1981; Equation 6). The appropriate equilibrium constant is: 123,Candela & Piccoli ‘i cos 3 tlement hich exe Nemsciype behavior No srr wih {Reape re ‘clement ich exit 1 Nanstayge behavior _ samara toe. ‘of Chompesedelemens! 10h imate be covet FeCl and HCl in vapor tom imsolubliy and met ASI, seme is anti ‘of Cleonpieet cent | ae Glee cfcercleestremoval Figure A1-1 Flow char ofthe procedure wed teal the composition of the magmatic volatile phase(). | statement of equilibrium, using the equation log K nen) Gino) (8) = 4+ BIT(R), where A and B are constants at 1.0 , kb (A = -25.218, B = 30361.8) (Frantz et al. Gin.) Cnr} CFa0% 1981). The activity of hydrogen (in Equation 8) is calculated from the fugacities of oxygen and "water, and the fugacity of pure hydrogen at the P fOp. Values for log K were calculated for this nd T of interest. The f2/;0 and flly for the pure | v mtg, ftom 0.1 to 1.0 as a function We varied a'tt (from 0.1 to 1.0 asa function of 124 agases have been calculated using Perkins ef al (1987) and tabulated values from Shaw & Wones (1964) [where fia, = P(bars) - Ty, respectively. ‘The mycyhas been calculated from: mcr = (Kiya Hase {&) o Values Ky, and K%4, are from Williams (1995: Table 1). The term Hase is defined as Hasc= (Ch, +, +2-C!,) (AST) (10) The molality of FeCl, in vapor or brine can therefore be calculated using (ay 1) na) {a Pro) Mec, A parallel calculation for mpacr, ean be performed for the brine. The equilibrium constant for mt solubility as originally formulated by Frantz. ef al (1981) was expressed in terms of activities of all phase components. In the equilibrium constant for mt solubility, the mole fraction of water is used as an estimate of the activity of water; this is justified given that the activity of the solvent approaches its mole fraction as the mole fraction approaches | (the mole fraction of water in brine- saturated vapor ~ 0.98). Step 11: Calculations of concentrations are iterated until values from integration (7) and iteration (+1) agree within a user-specified tolerance. The volatile exsolution (reaction) progress variable in this algorithm is the proportion of water-saturated melt crystallized, and is given by Z. Step 12: The efficiency of removal of metals form the melt is calculated from the integral of the concentration function over the full extent of Ore-metal Partitioning volatile exsolution, and the initial concentration of water and the metals in the melt. APPENDIX 2 The Thermodynamics of Chloride Partitioning between Melt and Vapor: An Introduction The thermodynamics of chloride partitioning are rather subtle, and have not been dealt with in terms of thermodynamically valid (ie., definable, and independently variable) components of the system. A full development cannot be given here, but the basie principles can be outlined. Consider ‘the simplest system that contains the essential chloride equilibria: NaAISijOs-NaCI-H,0-KNa.,- HINa, (SiO, could be added, but would not contribute to these equilibria; therefore, it has been left out so as not to cloud the discussion), The pertinent phases are the volatile phase, with the components HO, NaCl, KNa., and HNa, , and the melt with the components NaAISi,Os - HO-NaCl - [HNa, = 1/2 (HO + Al,O3 + 6Si0, + NaAISi;0,)]. There are four independent con- ditions of chemical equilibrium in this melt-vapor system: NaCl’ + NaCt (this study) (2) K-Na exchange (KNa.,) (Gammon et al, 1969)(13) HO" = HO! (water solubility); and (14) HNa!, = HNa!, (as) (Williams, Candela, and Piccoli unpublished data) Note that there is no one equilibrium for “chloride Partitioning”. This points out that equilibria should not be considered independently, Rather, the formulation of partitioning equilibria is dependent upon the choice of phase components, and the other equilibria that form the set of independent relations at equilibrium. Hence, the equilibrium NaCl" = NaCl’ , by itself, does not account for “chloride partitioning” However, the above set of equations (plus associated activity 125Candela & Piccoli coefficients that remain to be determined) should be sufficient to account for CI partitioning in the simple system defined here (appropriate equilibria can be added when considering more complex systems). The equilibrium constant for Equation 12 can be represented by ie Gyucr __ Mngt nac/ 55S eee Ne 16) ahaa [BE] Xue ‘This suggests that the thermodynamic “partition coefficient” for chloride, in light of the above equilibria, is best represented as Mme _ KM y] + (Na#] 555 (07) where X( is an expression for the mole fraction of Cl. {This could be developed as expressing the mixing of Cl on an anion matrix in the melt (Candela 1986b) that accounts for mixing of quasi-chemical species such as Cl, OH, Hz0, 0", and F; O* represents nonbridging oxygens that are associated with network-modifying cations, including aluminum in melts with ASI>I [ASI = ‘Aluminum Saturation Index: molar AV/(Nat K:(0.5)Ca)] in the melt, and cations such as FeO, CaO, MgO, efc.}. Nat = molar Na/(Na + K) in the melt, ([ly) = appropriate ratio of activity coefficients from Equation 16, When combined with equilibria for relations 13 and 15 above, we cean express the empirical ratio Cl/Cl(m) (the partition coefficient for Cl) as Dé} = K + (ly): 55.5 N |(Na#) + (K#)+ Kine(AS-D] 8) iN in the melt where Kit is the molar K/(Na#K) rai 1N is a factor that converts Xty—>m¢y (Candela 1986b), and is equal to the number of moles per 126 kilogram of Cl + F + OH + H,0 +O" N increases with ASI, AVSi (if ASI > 1), FeO, CaO, and MgO. Therefore, Equation 18 predicts that De, increases with increasing ASI, AV/Si, or alkaline-earth or transition-metal concentrations in the melt, as long as Kyys is low, as it presumably is at pressures greater than 2 kb, For simplicity, we have related these equilibria to vapor/melt partition coefficients only, even though equilibria involving brine are desired also, Chemical equilibria such as those presented above for melt/vapor equilibria are necessary and sufficient conditions for chemical equilibrium. APPENDIX 3 What are Reasonable Estimates of Melt (Magma) Cl and Hy0 in Silicic Magmas? Although several magmatic parameters have been identified as critical in the formation of various types of magmatic-hydrothermal deposits (e.g, pressure, types of crystallizing phases in an ore-related magma, and oxygen fugacity), the concentrations of Cl and H;O seem to play a dominant role. It has been shown that magmatic C1 and water concentrations (initial and saturation) are imperative in the removal of some elements into magmatic-hydrothermal fluids (Candela 1989, and the calculations in this manuscript. Rarely do paleomagmatic halogen signatures remain preserved in ore-beating systems—direct evidence is usually obscured. Nevertheless, several types of data can be used to obtain information about the chloride content of silicic and plutonic systems, Recently, fluid inclusions, ‘mineral chemistry, measurement of high- temperature volcanic fumaroles, and melt inclusions have been used for this purpose. Melt inclusions are commonly interpreted as recording major- and trace-element signatures of melts prior to eruption (e.g., Lowenster 1994). Assuming that melt inclusions retain magmatic halogen concentrations upon entrapment, they can be used as indicators of Cl/water ratio in silicic systems of interest here. We have included this Appendix to illustrate0.25 02: os clue. %) on Ore-metal Partitioning Water (wt. %) [2 Berta > tao Obsidian Dome ® Crater Lake Figure A3-1. Chlorine and water concentrations in melt inclusions from the Bandelier Tuff (Dunbar etal. 19926), Bishop Tul (Dunbar et al, 19922); Obsidian Dome (Hervig ef al. 1989), and Crater Lake (Bacon ef al. 1992), the range of Cl and water concentrations in some silicic systems, Although the absolute Cl con- centrations of melts may vary significantly (given metaluminous to slightly peraluminous thyolitic compositions), the CV/H,0 ratios are compara- tively constant. The limited data set presented here can be summarized as follows: in the Bandelier Tuff (melt inclusions: N= 96; ASI = 0.98; SiO, 76.8; Cl 1990 + 440 ppm; HO 3.2 + 1.2 wt.%; CVH,O = 0.072 + 0.033) [Dunbar & Hervig 1992b], Bishop Tuff (V= 134; ASI = 1.02; SiOy 77.2; Cl 720 + 160 ppm; H,0 3.0 + 15 wt.%; CIMH;0 = 0.03 + 0.02) [Dunbar & Hervig 1992a), Crater Lake (V = 14; ASI = 1.03; SiO, 73.0; Cl 1880 140 ppm; H,0 3.7 * 1.1 wt%s CVH,O = 0.06 + 0.04) [Bacon et al, 1992}, Obsidian Dome (W = 9; ASI = 1.07; SiO2 74.6; Cl 820 + 390 ppm; HO 4.02 + 1.42 wt.%; CVH,0 = 0.024 + 0.016) [Hervig et al. 1989]. Overall, the results of 257 melt-inclusion analyses with CH,O of 0.047 + 0.033 (16) and range from 0.006 to 0.250). ‘These studies suggest that melt CV/H,O values of 0.1 and 0.01 are quite common in felsic magmatic systems (Fig. A3-1). Note that CVH,O ratios of trapped melt inclusions are a function of: 1) pressure during trapping, which affects both water solubility in icic melts, and chlorine partitioning into the magmatic vapor and brine relative to the silicate melt; 2) melt composition; 3) the phase that inclusions are hosted within; and 4) timing of volatile saturation in a magma, but more likely time of entrapment (e.g, nlon-synchronous saturation of volatiles; ie., differing timing of saturation of HO, CO, and SO as recorded in melt inclusions (Pasteris e al. 1994)), CVO in melt inclusions commonly significantly from CV/H,O in obsidian and matrix. glass. For example, obsidian from Taupo, New Zealand, has CHO = 0.25 (N= 19; Dunbar & Kyle 1992) whereas in associated melt inclusions, CVO = 0.04 (Hervig ef al, 1989), Note that the absolute Cl concentrations in the obsidian and melt inclusions are similar (~1500 ppm), but the water concentration varies significantly (obsidian = 0.78 wi.%: MI ~4.5 wt.%). 127