Experiments on liquid immiscibility in silicate melts with H₂O, P, S, F and Cl: implications for natural magmas

Abstract

Isobaric (200 MPa) experiments have been performed to investigate the effects of H₂O alone or in combination with P, S, F or Cl on liquid-phase separation in melts in the systems Fe₂SiO₄–Fe₃O₄–KAlSi₂O₆–SiO₂, Fe₃O₄–KAlSi₂O₆–SiO₂ and Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂ with or without plagioclase (An₅₀). Experiments were heated in a rapid-quench internally heated pressure vessel at 1,075, 1,150 or 1,200 °C for 2 h. Experimental fO₂ was maintained at QFM, NNO or MH oxygen buffers. H₂O alone or in combination with P, S or F increases the temperature and composition range of two-liquid fields at fO₂ = NNO and MH buffers. P, S, F and Cl partition preferentially into the Fe-rich immiscible liquid. Two-liquid partition coefficients for Fe, Si, P and S correlate well with the degree of polymerization of the SiO₂-rich liquid and plot on similar but distinct power-law curves compared with equivalent anhydrous or basaltic melts. The addition of 2 wt% S to the system Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂ stabilizes three immiscible melts with Fe-, FeS- and Si-rich compositions. H₂O-induced suppression of liquidus temperatures in the experimental systems, considered with the effects of pressure on the temperature and composition ranges of two-liquid fields in silicate melts, suggests that liquid-phase separation may be stable in some H₂O-rich silicate magmas at pressures in excess of 200 MPa.

Introduction

Liquid-phase separation is accepted as an important differentiation mechanism in diverse magmas. Immiscibility between felsic silicate-dominated and Fe-rich mafic silicate-dominated liquids (L ^f and L ^m) has been documented in layered mafic intrusions (McBirney 1975), anorthosite complexes (Darling and Florence 1995), mid-ocean ridge magma chambers (Dixon and Rutherford 1979), granitoids (Rajesh 2003; Johnson et al. 2002), lamprophyres (Philpotts 1967) and lunar and terrestrial volcanic rocks (Roedder and Weiblen 1971). Immiscibility also has been invoked as a primary genetic mechanism for some iron oxide-dominated base and precious metal mineral deposits, including Kiruna-type magnetite–apatite systems (Chen et al. 2010, Clark and Kontak 2004, Nyström and Henríquez 1994).

Experimental investigations of felsic silicate--mafic silicate liquid immiscibility to date have constrained the configuration of miscibility gaps in silicate melts as a function of temperature, parental melt composition and, to a lesser extent, pressure for the anhydrous systems Fe₂SiO₄–KAlSi₂O₆–SiO₂ with or without one or more of the following: Na, Ca, Mg, Ti or P (Roedder 1951, 1978; Watson 1976a, b, Visser and Koster van Groos 1979a, b; Naslund 1983; Bogaerts and Schmidt 2006), as well as for more chemically complex lunar and terrestrial basaltic liquids (Longhi 1990; Philpotts 1982; Hess et al. 1975; Rutherford et al. 1974). To develop further a predictive model of felsic silicate--mafic silicate immiscibility in magmas, we herein document experiments on the effects of the network-modifying component H₂O, alone and in combination with P, S, F or Cl, on selected compositions in the systems Fe₂SiO₄–Fe₃O₄–KAlSi₂O₆–SiO₂–Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂, Fe₃O₄–KAlSi₂O₆–SiO₂, Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂ and Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂–Ca_0.5Na_.5Al_1.5Si_2.5O₈ (An₅₀).

The stability and extent of two-liquid solvi in silicate melts are markedly sensitive to minor variations in the enthalpy of mixing term in the expression: ∆G _mix = ∆H _mix − T∆S _mix. Common constituents in natural melts such as OH, P, S, F and Cl have positive enthalpies of formation of anion--metal complexes which act to increase the value of the ∆G _mix term, thus favoring liquid unmixing (Ryerson and Hess 1978). Further, it has been demonstrated that OH, P, S, F and Cl influence liquid miscibility gap relations and the degree of polymerization in silicate melts, parameters that control the compositions and thermal ranges of two-liquid miscibility gaps (Botcharnikov 2008; Moore et al. 1998; Visser and Koster van Groos 1979a; Haughton et al. 1974; Dolejš and Baker 2007; Webster and De Vivo 2002).

The results reported herein show that H₂O, alone or in combination with P, S, F or Cl, acts to expand the composition range of experimental two-liquid miscibility gaps at temperatures that are geologically relevant (1,075–1,200 °C). The phase relationships we document, considered in conjunction with known chemical-thermal evolution trends in natural melts, provide a more complete framework for the discussion of the role of silicate-liquid immiscibility in petrogenesis and mass–flux in natural igneous systems.

Experimental method

Starting materials for the experiments included seven anhydrous base compositions prepared from SiO₂ (cristobalite), Al₂O₃, K₂Si₂O₅, FeO and Fe₂O₃. Each base mixture plots as a composition point on the 30 wt% FeO isopleth on the ternary join fayalite–leucite–silica (Fig. 1). Base compositions have an Al/K molar ratio of 1. To minimize the fO₂ gradient between melts and external solid buffers, Fe³⁺/Σ Fe values for the melts synthesized in this study were estimated with the method of Schuessler et al. (2008) at: fO₂ = quartz–fayalite–magnetite (QFM), nickel–nickel oxide (NNO) or magnetite-hematite (MH) buffers and T = 1,200 °C and P = 200 MPa. The 30 wt% FeO_total component of each base mixture comprises FeO and Fe₂O₃, and Fe₂P, FeS, FeCl₂ or FeF₂ in proportions that approximate the Fe³⁺/Σ Fe values calculated for the selected experimental conditions. Oxygen fugacity in the experimental capsules was controlled using the conventional double capsule, metal–metal oxide or metal oxide-silicate + water configuration (Chou and Cygan 1990).

Fig. 1

Backscattered electron images of run product, a liquid-phase separation and coalescence of liquid droplets (L ^m), composition: A-3 + H₂O + F, fO₂ = MH, P = 200 MPa, phase assemblage is the result of a thermal gradient, range ~ 950–1,230 °C (from left to right), b typical liquid-phase separation textures, composition: A-1 + H₂O + P, fO₂ = MH, 1,200 °C, P = 200 MPa, c two liquids plus magnetite, composition: A-3 + H₂O + F, fO₂ = MH, 1,075 °C, P = 200 MPa, d three-liquid-phase assemblage (L ^f, L ^m and L ^s) and three liquids plus pyrrhotite (Po) developed upon cooling, A-3 + H₂O + S, fO₂ = MH, 1,150 °C, P = 200 MPa. L ^f = Si-rich liquid, L ^m = Fe-rich liquid, L ^s = sulfide-rich liquid, Po = pyrrhotite, mt = magnetite

Experimental starting compositions containing either 1 wt% P, 2 wt% S, 6 wt% Cl or 6 wt% F (total wt. oxides) were prepared by the addition of Fe₂P, FeS, FeCl₂ or FeF₂ to the anhydrous base mixtures (Table 1). Fe salts were selected as the source of P, S, F and Cl in order to minimize the loss of volatile components during the welding of experimental capsules. Hydrous experiments incorporated 10 wt% H₂O (total wt. solids). Plagioclase-bearing experiments contain 1.3 wt% (total weight of solid oxides and halides) An₅₀, constituting 43 wt% of the feldspar component.

Table 1 Base compositions (wt%)

Experiments were carried out by loading the desired quantity of starting material, or starting material +H₂O, into a 2-mm-outside diameter, 1.25-cm-length, 0.1-mm-thick platinum capsule. Three to five experimental capsules were loaded into a 5-mm-outside diameter, 3-cm-length, 0.2-mm-thick platinum capsule containing H₂O and one of the selected metal–metal oxide or metal oxide--silicate buffers, QFM, NNO or MH. Both inner experimental capsules and outer buffer-bearing capsules were sealed by welding.

Experiments were carried out in Kanthal™ or platinum-wound furnaces placed in an internally heated pressure vessel under isobaric conditions (200 ± 10 MPa), isothermally at 1,075, 1,150 or 1,200 °C for two hours using argon as the pressure medium. The pressure vessel, similar in design to that described by Holloway (1971), was modified to allow the vessel to rotate from the horizontal run position to a vertical quench position. Rapid isobaric cooling of the experimental capsules was achieved as the vessel was rotated toward the vertical causing the capsule to drop from the hot spot to the unheated, water-cooled end of the pressure chamber (T < 250 °C). The quench rate is inferred to be 500 °C/s, similar to that reported by Holloway (1992) for a rapid-quench furnace with an equivalent thermal profile. The rotating furnace design used in this study provides a significant degree of control over the thermal characteristics of the critical heating zone. The temperature along with the length of the experimental capsules was measured using three inconel-sheathed, chrome-alumel thermocouples. Temperature differences between the distal thermocouples ranged from 1 to 16, ±2 °C. The argon medium pressure was measured using a Bourdon tube gauge, accurate to ±5 MPa.

Reversal experiments were performed to determine the time required to achieve chemical equilibrium in the experimental charges. Capsules (buffered at fO₂ = QFM, NNO and MH) containing experimental base compositions +10 wt% H₂O (total weight of solids in charge) were heated for 2 h at a temperature of 1,210 °C and then cooled to 1,075, 1,150 or 1,200 °C for one, two or 4 h, respectively, and subsequently quenched. The chemical compositions and textural characteristics observed in the experimental products produced in the reverse experiments are identical to those produced in forward experiments run at the same temperature, and it is concluded that equilibrium was obtained at run durations of less than 1 h. Solid-oxide buffer reactants were evaluated after cooling using X-ray powder diffraction analysis or microscopic phase identification.

Experimental samples were mounted in epoxy, polished and analyzed with a Cameca SX-100 electron microprobe at the University of Manitoba. Analytical conditions were set to an accelerating voltage of 15 kV, a 15 nA beam current and counting time of 20 s for all elements except P, S, F and Cl (30 s). The beam diameters were 5–10 μm for Si-rich and Fe-rich glasses and 2 um for quenched sulfide melt. Natural and synthetic oxides, silicates or sulfides were used as standards.

Results

General

Compositions of the starting materials are given in Table 1, run conditions and phase relation data are summarized in Table 2, and electron microprobe analyses and mass concentration partition ratios (D _i  = concentration in the mafic liquid L ^m/concentration in the felsic liquid L ^f) are presented in Table 3.

Table 2 Experimental run conditions and phase assemblages

Table 3 Electron microprobe analyses of experimental run products and major element liquid partition coefficients, D_{Fe-liq/Si-liq} (wt% component in Fe-rich liquid/wt% component in silicate-rich liquid)

Immiscible liquids in experimental anhydrous silicate melts commonly occur as spheroidal or ellipsoidal droplets that range from submicron to over 500 μm in size (Fig. 1). Exsolving conjugate immiscible liquids nucleate and form droplets which grow and, to varying degree, coalesce (Fig. 1a). Each liquid commonly occurs as droplets or pools distributed within a larger volume of the other liquid (Bowen 1925; Philpotts 2008). Experimental melts produced in this study in systems with H₂O alone or H₂O + P, S or F exhibit a spherical geometry similar to that observed in anhydrous melts (Fig. 1b).

All of the experimental immiscible melt phase assemblages contain a vapor phase. The presumed equilibrium vapor bubbles are generally >10 μm in diameter and are defined by menisci that appear smooth under high-power magnification. These bubbles are distinguished from quench-induced vapor exsolution textures which form as small bubbles of <1 μm or as aggregate masses of bubbles with irregular geometries.

Macroscopic-scale liquid-phase aggregation and gravitational segregation has been reported in anhydrous immiscible silicate melts (Kyser et al. 1998). In the present study, the spatial orientation of experimental capsules was recorded prior to the heating of selected charges, and the gravitational settling (density segregation) of Fe-rich immiscible liquids was evident in the run products produced in each of the H₂O-bearing experimental systems.

Addition of H₂O

H₂O-induced liquidus suppression in both natural and synthetic silicate melts is well documented (Médard and Grove 2008; Gaetani et al. 1994). In the hydrous experimental systems considered here, the H₂O-induced thermal suppression of the liquidus surfaces exposes low-temperature, compositionally extensive, two- or three-liquid miscibility gaps, some of which lie partially or entirely below the liquidus surfaces in anhydrous systems of similar compositions (Fig. 2). In the system Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂–H₂O at fO₂ = MH, the composition range of the miscibility gap is increased relative to the range in equivalent anhydrous melts (Fig 3). The minimum temperature observed for the two-liquid field in the experimental melts with added H₂O is 1,150 °C, in anhydrous melts that are otherwise equivalent in composition, 1,375 °C (Naslund, 1983). At fO₂ = NNO, the addition of H₂O to the system Fe₃O₄–KAlSi₂O₆–SiO₂ displaces the range of the miscibility gap to higher SiO₂ compositions and presumably extends the two-liquid field outside of the experimental composition range. A single experiment performed at fO₂ = QFM, with base composition A-1 + H₂O (10 wt%), P = 200 MPa and T = 1,200 °C, produced two conjugate liquids. The A-1 base mixture used in the experiment is more SiO₂-rich (~5 wt%) than the most SiO₂-rich parental melt in equivalent anhydrous melts, suggesting that the miscibility gap either expands or shifts toward more SiO₂-rich compositions in hydrous melts at fO₂ = QFM.

Fig. 2

Phase relations in experimental melts (a–h). Composition: base mixtures A-1–A-7 systems Fe₃O₄–KAlSi₂O₆–SiO₂ and Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂ plus 10 wt% H₂O and either 1 wt% phosphorus, 2 wt% sulfur or 6 wt% fluorine. Conditions of experiments: T = 1,075, 1,150 or 1,200 °C, P = 200 MPa, duration—2 h, fO₂–NNO or MH buffers. Dashed lines represent the inferred region of phase-field boundaries

Fig. 3

Phase relations in experimental melts with 10 wt% H₂O and from equivalent anhydrous melts (Naslund 1983). a Phase assemblages of experimental melts with equivalent compositions run at oxygen fugacity conditions of MH buffer (this study), fixed at log fO₂ = −5 (Naslund). b Experimental melts with equivalent compositions run at oxygen fugacity conditions of NNO buffer (this study), fixed at log fO₂ = −9 (Naslund). L liquid, M magnetite, S silica mineral

Addition of H₂O and phosphorus

The miscibility gap in the system Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂–H₂O–P, fO₂ = MH (Fig. 2c), shows the same range of composition as that in the melts with H₂O alone, but the extent of the two-liquid plus magnetite field is significantly increased in the phosphorus-bearing system. In the system Fe₃O₄–KAlSi₂O₆–SiO₂–H₂O–P at fO₂ = NNO (Fig. 2d), the two-liquid field is shifted toward higher SiO₂ compositions but to a lesser degree than in melts with H₂O only. The temperatures of the magnetite saturation in the system at fO₂ = NNO are predictably lower than at fO₂ = MH, and consequently, the two-liquid field is thus extended to lower temperatures at more reducing conditions.

Addition of H₂O and sulfur

Immiscibility between silicate and sulfide melts is well documented (e.g., Naldrett 2005), but no experimental investigations of the effect of sulfur on liquid-phase separation in silicate melts have been documented. Laroque and Stimac (2000) reported the conversion of Fe-sulfide melts to Fe-oxide melts in basaltic, intermediate and granitic systems, providing chemical and textural evidence that immiscible FeO-rich liquids are widely produced from sulfide liquids as a consequence of sulfur loss during magma devolatilization. The addition of H₂O + 2 wt% S to the system Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂–H₂O at fO₂ = MH stabilizes a three-liquid field (Fig. 2e), a Fe-rich mafic silicate liquid, a Fe-poor felsic silicate liquid and a Fe-sulfide liquid (FeS₆₃–FeS₇₄). The three-liquid field range extends from base composition A-1 to A-6, delimited, in the low-silica portion of the experimental field, by a region in which two liquids plus magnetite are stable (Fig. 2e). The three liquids are preserved in experimental glasses as spheroidal droplets, each distributed within the other two, or as crystalline Fe-sulfide that nucleated and grew rapidly in the sulfide liquid upon quench. The sulfide solid phase shows dendritic growth habit and is best developed on the boundary surface between magnetite crystals in some of the quenched mafic conjugate liquids (Fig. 1c).

Experiments in the system Fe₃O₄–KAlSi₂O₆–SiO₂–H₂O–S at fO₂ = NNO did not reach sulfide liquid saturation, but a compositionally extensive, two-liquid field stable below 1,075 °C is observed in the intermediate- to high-silica portion of the experimental composition range (Fig. 2f).

At fO₂ = NNO, the mafic liquids in the two-liquid portion of the system accommodate over 10 wt% sulfur. At fO₂ = MH, sulfur is strongly partitioned into the Fe–S–O melt, which accommodates up to 23 wt% S and 60–65 wt% FeO_total. The phase relations in the high-silica portion of the system Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂–H₂O–S at fO₂ = MH, i.e., between the three-liquid field and the assemblage one liquid + magnetite + silica in temperature range of temperature range 1,075–1,150 °C, are undetermined. The exsolution of a sulfide liquid in the S-bearing melts is notable, in part, because sulfate is commonly considered the predominant sulfur specie in silicate melts under oxygen fugacity conditions above NNO, e.g., Wallace and Carmichael 1994. The assumption of equilibrium fO₂ conditions is supported by the partitioning trends of S in the two-melt fields, as well as composition data from reversal experiments. The simplest explanation for the presence of sulfide is that the speciation of sulfur is controlled by the amount of Fe²⁺ in the melts, which in turn is probably elevated because of the lack of mono- and divalent cations required to charge-balance any network-forming Fe³⁺ in the melts.

The addition of H₂O and fluorine

The addition of fluorine to the system Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂–H₂O at fO₂ = MH increases both the composition range and the silica mineral saturation surface temperatures in the experimental regime (Fig. 2g) in agreement with the general principles of the acid–base equilibrium. A similar effect of the addition of F has been observed in the haplogranitic system, e.g., Manning (1981).

The expansion of the silica mineral stability region limits the extent of the two-liquid field and shifts the composition range of the miscibility gap to more intermediate-silica compositions. The two-liquid field in the system Fe₂O₃–KAlSi₂O₆–SiO₂–H₂O–F at fO₂ = NNO is limited to high-silica compositions (~66 wt% SiO₂).

The addition of H₂O and chlorine

No two-liquid field was observed in this study in the system Fe₃O₄–KAlSi₂O₆–SiO₂–H₂O–Cl, fO₂ = NNO. The addition of H₂O + Cl to the Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂–H₂O fO₂ = MH system increases the temperatures of the silica mineral saturation surface to the extent that two liquids were observed only in base composition A-1 + H₂O + Cl at 1,200 °C (Table 2). Melts produced from starting mixtures of mafic to intermediate composition become significantly enriched with iron (40–64 wt% FeO_total) and H₂O with decreasing temperature and the progressive crystallization of silica minerals.

The addition of H₂O plus P, S, F or Cl to the system Fe₃O₄–KAlSi₂O₆–SiO₂–Ca_0.5Na_.5Al_1.5Si_2.5O₈

Experiments were performed in the system Fe₃O₄–KAlSi₂O₆–SiO₂–Ca_0.5Na_.5Al_1.5Si_2.5O₈ (An₅₀) with 10 wt% H₂O, 1 wt% P, 2 wt% S and 6 wt% F or Cl at 1,200 °C, fO₂ = NNO and P = 200 MPa. Stable two-liquid assemblages were observed in the melts with H₂O alone and with H₂O and P, S or F (Table 2). No liquid-phase separation was observed in the melts with H₂O plus Cl.

Additional experiments

Additional experiments were performed using base compositions A-1 or A-3 with or without H₂O, at 1,200 °C, P = 200 MPa and fO₂ = QFM, NNO or MH (Table 2). Two-liquid assemblages were observed in melts with compositions A-1 + H₂O (QFM), A-1 + H₂O + F (QFM), A-1 + P (QFM), A-1 + F (QFM) and A-1 + Cl (MH). Two-liquid- plus solid-phase assemblages were produced in melts A-3 + H₂O + P (QFM), A-1 + Cl (QFM), A-1 + F (NNO) and A-1 + S (MH).

Discussion

Effects of H₂O, P, S, F and Cl on the T–X range of two-liquid fields

The temperature and compositional ranges of two-liquid fields in silicate melts are dependent on the position and configuration of the miscibility gap relative to the saturation surfaces of liquidus minerals. The phase relations given in Fig. 2 reflect the combined and, in some instances, competing effects of H₂O and P, S or F on the liquidus surface configuration in each of the experimental systems. Water has the most pronounced effects on the phase relations in the experimental compositions. Thus, the addition of H₂O dramatically suppresses liquidus surface temperatures, expands the stability field of magnetite and decreases that of silica minerals, thereby increasing the T–X range of the miscibility gap (Fig. 3). In anhydrous melts in the systems Fe₂SiO₄–KAlSi₂O₆–SiO₂ and Fe₂SiO₄–KAlSi₃O₈–SiO₂, with or without CaO, TiO₂ and MgO, the addition of phosphorus expands the two-melt field, decreasing the extent of the fayalite field and increasing that of silica minerals (Bogaerts and Schmidt 2006). It is likely that P and S (at concentrations below sulfide saturation) have a similar effect on the H₂O-rich melts in the simplified systems considered here, but H₂O-induced suppression of the silicate–mineral liquidus surface limits the stability of silica minerals to high SiO₂ compositions at low temperatures in the S-bearing melts and eliminates it altogether in the P-bearing melts. In contrast, F increases the silica mineral stability field to the extent that the two-liquid field is truncated by the silica–mineral saturation surface.

Major element partitioning and melt structure relations in volatile-rich melts

Major element partitioning trends in the volatile-rich experimental melts are similar to those reported in equivalent anhydrous melts (Watson 1976b; Naslund 1983; Bogaerts and Schmidt 2006); Fe, Ca and P preferentially partition into the mafic melt and Si, Al, K and Na into the felsic. No data have been recorded regarding the partitioning of S and F between anhydrous immiscible silicate melts, but in the H₂O-bearing melts produced in this study, S partitions strongly into the mafic melt (Fig. 4b). In contrast, F partitions nearly equally into the mafic and silicate liquids (D _F = 1 ± 0.6). The absolute values of major element partition coefficients between coexisting experimental melts are generally higher than in equivalent anhydrous melts.

Fig. 4

Partition coefficients and nbo/t^f between conjugate experimental melts (L ^m, FeO-rich melt; L ^f, SiO₂-rich melt). a variation in the partition coefficients between melts for Fe as a function of nbo/t ^f, b variation in the partition coefficients between melts for P, S and F as a function of nbo/t ^f, literature data for P in the system Fe₂SiO₄–KAlSi₂O₆–SiO₂ are from Visser and Koster van Groos (1979b, c) and Freestone and Powell (1983), data for the basaltic system (tholeiitic and lunar basalts) are from Rutherford et al. (1974), Ryerson and Hess (1980), Dixon and Rutherford (1979), Ryerson and Hess (1980), Philpotts and Doyle (1983) and Longhi (1990), modified from Bogaerts and Schmidt (2006). MME partition data from non-equilibrium, coexisting microgranular enclaves in granitic melts (Bogaerts and Schmidt 2006). c Variation in the partition coefficients between melts for Si and K as a function of nbo/t ^f in experimental melts with H₂O and H₂O plus P, S or F (this study), and in immiscible basaltic melts (Bogaerts and Schmidt 2006). d Fe partition coefficients between melts as a function of nbo/t ^f for coexisting melt inclusions in rocks from El Laco Fe-oxide deposit, Chile; Antauta hypabyssal complex of the Picotani Group, Peru; and the Dongargarh Supergroup, India (Naslund et al. 2002; Clark and Kontak 2004; Sensarma et al. 2000)

The relationship between major element partition ratios and melt structure in immiscible silicate melts has been used to assess the role of liquid-phase separation in the evolution of coexisting silicate magmas of differing composition. Specifically, major element mass partition ratios (D _i ) plotted as a function of the melt polymerization parameter nbo/t ^f (nbo number of non-bridging oxygens; t tetrahedrally coordinated network-forming cations; f = the felsic member of conjugate immiscible liquids pairs) define power-law relationships that are distinct from those for coexisting melt pairs that have not undergone liquid-phase separation. Bogaerts and Schmidt (2006) demonstrated that power-law curves (for D _i as a function of nbo/t) for the elements Fe, Ti, P, Si and K for immiscible melts of basaltic composition can be applied as a means to assess rocks formed from coexisting magmas over a wide range of compositions and petrogenetic conditions.

To test whether the power-law relationships for major element partition data and nbo/t ^f in anhydrous silicate melts (described above) can be applied to the assessment of melts with H₂O, P, S or F, D _i values for several of the elements are plotted as a function of the nbo/t ^f values of the volatile-rich melts produced in this study (Fig. 4). For purposes of comparison, nbo/t ^f values are calculated using the method of Bogaerts and Schmidt (2006), in which t = Si + Al + P, and all Fe is treated as a network modifier, an assumption, justified in part, because the Al/K = 1 molar ratio in the experimental melts severely limits the number of cations available to charge-balance any network-forming Fe³⁺. Water is not included in the nbo/t ^f calculation scheme because its concentration in both natural and synthetic melts is difficult to quantify and the method would be impractical if the power-law relationship between partitioning and the nbo/t ^f parameter were strongly dependent upon the H₂O content of the melt.

Fe, Si and K partitioning

D_Fe and nbo/t data in melts with H₂O only, with P or S, overlap (Fig. 4a), yielding power-law curves which are indistinguishable from each other. Power-law equations fitting the data are as follows: for melts with H₂O only, D _Fe = 1.29 (nbo/t ^f)^−0.84 (r ² = 0.82), for melts with H₂O + P, D _Fe = 1.22 (nbo/t ^f)^−0.95 (r ² = 0.82) and for melts with S, D _Fe = 1.62 (nbo/t ^f)^−0.82 (r ² = 0.93). The curve for D _Fe as a function of nbo/t ^f in F-bearing melts differs from the curves calculated for melts with H₂O, S or P reflecting both higher D _Fe values and a shift in the miscibility gap toward more polymerized compositions (Fig. 5). The power-law equation fitting the data for F-bearing melts is D_Fe = 1.75 (nbo/t ^f)^−0.99 (r ² = 0.74).

Fig. 5

Width of the miscibility gap expressed as a function of temperature plotted against nbo/t of conjugate liquid pairs in experimental melts with H₂O, and H₂O with P, S or F. Data for the basaltic system are from Ryerson and Hess (1980), for the systems Fe₂SiO₄–KAlSi₂O₆–SiO₂–P, Visser and Koster van Groos (1979b, c), modified from Bogaerts and Schmidt (2006), and for the system Fe₂SiO₄–KAlSi₂O₆–SiO₂, Bogaerts and Schmidt (2006)

The addition of plagioclase (An₅₀) to melts with H₂O, or without H₂O plus P or F, does not induce significant changes in Fe partitioning, but D_Fe values in melts with plagioclase and S are greater than D _Fe values for the plagioclase-bearing melts with H₂O, or H₂O plus P or F. Temperature and oxygen fugacity have a minimal effect on the relationship between Fe partitioning and nbo/t in both the volatile-rich experimental melts and similar anhydrous immiscible melts (Bogaerts and Schmidt 2006).

Partitioning data for Si and K in melts with H₂O, P, S or F show significant overlap and are more dispersed than those for Fe, but data for D _Si as a function of nbo/t ^f in the experimental melts considered in toto yield a power-law curve that is clearly distinct from that defined by Si in coexisting melts that have not undergone liquid-phase separation (Fig. 4c).

Partitioning of P, S and F

Power-law curves for D _S (fO₂ = NNO or MH) and D _P (fO₂ = NNO) in the experimental melts are similar to those for D _Fe and plot above the curve for D_P in immiscible basaltic liquids (Fig. 4b). Partitioning coefficient values (D _P) at fO₂ = MH are greater than D _P at fO₂ = NNO, but increased polymerization of the felsic melt shifts the D _P–nbo/t ^f field toward more polymerized compositions that plot closer to the power-law curve for P in immiscible basaltic melts than the data from melts at fO₂ = NNO. Power-law equations fitting the data for the hydrous melts are D _P (NNO) = 1.07 (nbo/t ^f)^−1.79 (r ² = 0.79) and D _S (NNO, MH) = 4.15(nbo/t ^f)⁻¹⁰⁸ (r ² = 0.72). Partitioning coefficients for melts with F (D _F = 1 ± 0.6), however, do not yield a power-law curve. The addition of plagioclase (An₅₀) to melts in the system Fe₃O₄–KAlSi₂O₆–SiO₂ increases the quantities of P, S and F that can be accommodated in the felsic melt.

Application of partitioning–polymerization relationships in experimental melts to the assessment of coexisting, volatile-rich natural magmas

Bogaerts and Schmidt (2006) demonstrated that power-law curves for D _i as a function of nbo/t ^f for the elements Fe, Si and Ti in immiscible melts in the basaltic system can be used to discriminate between coexisting melts generated by liquid-phase separation and those resulting from other processes. Major element partitioning data from rocks formed from coexisting immiscible magmas plot either proximal to or as extensions of the power-law curves derived for immiscible melts in the basaltic system. Major element partitioning data from rocks formed from coexisting magmas that were generated by other processes are distinct from those of the immiscible basaltic system. The method has been applied to rocks generated by both intrusive and extrusive coexisting magmas over a wide range of P–T–X–fO₂ conditions. A comparison of power-law relationships established for D _i –nbo/t ^f relationships for the elements Fe, Si and P in the melts produced in this study with those calculated for the same elements in the basaltic system shows that the power-law equations calculated for the basaltic system can be applied to volatile-rich magmas (Fig. 5) over a wide fO₂ interval, e.g., QFM–MH. Further, the method is applicable even if the H₂O and Fe³⁺ contents of the magmas are not considered in calculating the polymerization parameter nbo/t. The finding is non-trivial because quantification of the volatile constituents and Fe³⁺ in igneous rocks and magmas is often problematic.

Immiscibility in volatile-rich magmas

H₂O, P, S, F and Cl are common constituents in most magmatic systems, and understanding their effects on silicate liquid-phase separation constitutes a critical step in the assessment of the role of silicate immiscibility in petrogenesis. Although the melt compositions employed herein are simplified, the documented effects of H₂O, P, S, F and Cl in the experimental systems provide a basis for understanding the potential influence of these constituents in natural magmatic systems.

The effect of H₂O alone or H₂O in combination with P or S is to increase the T–X range of miscibility gaps in silicate melts (Fig. 5) by suppressing the saturation temperatures of liquidus minerals. F and Cl increase the activity of SiO₂ in the melt, thereby expanding the T–X stability fields of SiO₂ minerals. The addition of H₂O and F or Cl to the experimental mixtures increases the SiO₂ mineral saturation surface, reducing the compositional extent of the miscibility gap in the F-bearing system and eliminating it altogether in the melts with Cl. Predictably, the stability fields of SiO₂ minerals and magnetite expand with increasing oxygen fugacity. The phase relations demonstrated here have general implications for the genesis of magnetite deposits of the Kiruna and El Laco type. The origin of these deposits has long been controversial. Sillitoe (2003) and Hitzman et al. (1992) among others propose that these deposits are hydrothermal lithologies, whereas Chen et al. (2010), Naslund et al. (2002) and others invoke liquid-phase separation of immiscible volatile, Fe- and Si-rich magmas as a petrogenetic mechanism. Rocks associated with these deposits are typically enriched with phosphorus and sulfur and have mineralogy and morphology characteristic of H₂O enrichment (Nystroem and Henrique 1994; Naslund et al. 2002).

Melt inclusions in dacite from the El Laco Fe-oxide deposit (Chile), comingled absarokitic basalt and peraluminous monzogranite from the Antauta hypabyssal complex of the Picotani Group (Peru) and Fe-rich spheres in andesitic rock from the Dongargarh Supergroup (India) have been interpreted as having formed from the unmixing of volatile-rich silicate magma (Naslund et al. 2002; Clark and Kontak 2004; Sensarma et al. 2000). Iron partition coefficient (D_Fe) and felsic-melt polymerization values (expressed as nbo/t ^f) calculated for the coexisting melts in these rocks are distinct from values calculated for coexisting rocks formed by other processes, and they plot above the power-law curves derived for the conjugate liquids produced in this study (Fig. 4a, d).

The expanded T–X range of the silicate-liquid miscibility gap and degree of Fe enrichment of the mafic conjugate melts (up to 72 wt% FeO_total) produced by the addition of H₂O, P and S to melts in this study, although by no means conclusive, support an immiscible petrogenetic hypothesis for some Fe-oxide deposits. In addition, the reduction in melt viscosity produced by the presence of H₂O favors the efficient separation of conjugate liquids by density, an important component of the immiscible petrogenetic model for the Kiruna ore deposit type.

Implications for the pressure stability of volatile-rich immiscible magmas

The effects of H₂O on the T–X configuration of the liquidus surface relative to the miscibility gap in the experimental melt systems have important implications for the pressure stability of two-liquid fields in H₂O-rich natural magmas. To clarify the effects of pressure on liquid-phase separation in H₂O-rich silicate melts, it is useful to consider: (a) the effects of pressure on the mixing parameters in the melts, specifically, on the T–X range of the two-liquid field, independent of the liquidus surface configuration; (b) the differences in the effects of pressure on the T–X configuration of the liquidus mineral saturation surface between anhydrous and H₂O-rich melts; and (c) the combined effects of (a) and (b) on the T–X configuration of two-melt field relative to the liquidus surface and the upper temperature limit of the miscibility gap.

In anhydrous silicate melts, increasing pressure either has little effect or expands the T–X extent of two-melt field miscibility gaps (Hudon et al. 2004; Visser and Koster van Groos 1979c; Watson and Naslund 1978), but increasing pressure commonly elevates liquidus surface temperatures above the miscibility gap, with the net effect that two-liquid fields are rarely intersected during the liquid line of descent of most anhydrous or H₂O-poor magmas.

The effect of pressure on two-melt fields in H₂O-rich melts is likely to differ significantly from that in anhydrous systems because the suppression of liquidus temperatures produced by H₂O is nearly independent of pressure up to 2 GPa (Médard and Grove 2008; Almeev et al. 2007; Gaetani and Grove 1998). The extent of stable immiscibility in H₂O-rich melts at pressures commensurate with mid- to deep crust and upper mantle environments is therefore dependent on composition and two competing effects: the increase in anhydrous melt liquidus temperatures as a function of increasing pressure and the magnitude of H₂O-induced liquidus suppression in H₂O-rich melts. For example, in basaltic melts at 2 GPa (fO₂ = NNO), 5 wt% H₂O lowers the olivine liquidus temperature by 137 °C, whereas the dry olivine liquidus temperature is increased as a function of pressure over a 2 GPa range by 130 °C. The net result is that at 2 GPa, the olivine liquidus temperature in H₂O-rich basaltic melts is slightly lower than that of the anhydrous melt at 0.101 MPa, 1,241 °C and 1,248 °C (Médard and Grove 2008). The H₂O-rich experimental melts produced in this study differ from the basaltic system in that they have fewer network- modifying species, and magnetite rather than forsterite is the primary liquidus mineral. Nonetheless, the magnitude of the H₂O-induced liquidus suppression in the experimental melts, although fO₂ dependent, is similar to that of the basaltic system, e.g., ∆T liquidus of the experimental melts, ~120 °C at fO₂ = NNO, ~200° at fO₂ = MH and basaltic melt ~137 °C at fO₂ = NNO.

Considering that liquid miscibility gaps (sensu stricto) in silicate liquids are either independent of or enhanced by increasing pressure and that H₂O-induced suppression of liquidus temperatures is nearly independent of pressure over a wide range of compositions at pressures up to 2 GPa, we infer that the T–X configuration of two-liquid fields in H₂O-rich silicate melts at 2 GPa will be similar to that of compositionally equivalent anhydrous melts at low pressures. In addition, although the upper critical temperatures of the miscibility gaps in the experimental systems considered here have not been established, they must be above the experimental upper thermal range (>1,210 °C), in excess of the upper critical temperature of similar anhydrous melts (Fig. 5). The relative increase in upper critical temperature in the hydrous melts is likely to extend further the area of the two-liquid field above the liquidus in H₂O-rich silicate melts.

Water-rich magmas are typically generated in primary melts in subduction zones by volatile-flux melting or in highly evolved magma systems that are enriched with volatile components during the crystal fractionalization process. Magmas in supra-subduction zone environments that contain sufficient H₂O to lower liquidus temperatures enough to stabilize two-liquid fields at depth (≥5 wt%) have been described in experimental and theoretical models of mantle wedge melting (e.g., Grove et al. 2006; Ulmer 2001). The studies delineate an extensive polythermal region within the mantle wedge, where primary, H₂O-rich melts (5–28 wt% H₂O) are generated. The T–X H₂O field that we infer to be permissive of silicate-liquid immiscibility lies within the T–X region where volatile-enriched melts are generated (Fig. 6), suggesting that silicate magma unmixing may be stable in liquids in deep arc environments. The process, though conjectural, describes an interesting mass-flux mechanism which, in conjunction with the contemporaneous mechanical segregation of conjugate melts, could concentrate and transport small but cumulatively significant quantities of Fe and volatile elements in supra-subduction zone environments.

Fig. 6

Inferred T–X H₂O range of two-liquid silicate melts in volatile-flux melting environment (temperature distribution and flux melting processes in the mantle wedge from Grove et al. 2006)

Conclusions

The addition of H₂O alone or in combination with small amounts of P, S or F (1, 2 and 6 wt% oxide totals, respectively) to melts in the systems Fe₂SiO₄–Fe₃O₄–KAlSi₂O₆–SiO₂, Fe₃O₄–KAlSi₂O₆–SiO₂ and Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂ expands the T–X range of the two-liquid miscibility gap. P and S partition strongly into the mafic melt, whereas F is nearly equally partitioned between the conjugate melts. Liquid-phase separation in melts with H₂O + Cl is restricted to a narrow composition range as the result of the Cl-induced increase in the stability of silica minerals. The addition of 2 wt% S to the system Fe₃O₄–Fe₂O₃–KAlSi₂O₆–SiO₂ stabilizes three immiscible melts with Fe-rich mafic silicate, Fe-poor felsic silicate and FeS compositions.

Power-law curves calculated for D _i –nbo/t ^f relationships for the elements Fe, Si and P in the melts produced in this study are similar to, but distinct from, those calculated for the same elements in immiscible basaltic melts. The results show that the power-law equations calculated for the basaltic system by Bogaerts and Schmidt (2006) can be applied to assess coexisting volatile-rich magmas over a wide range of P–T–X–fO₂ conditions. Further, the method is applicable even if the H₂O and Fe³⁺ contents of the magmas are not considered in calculating the polymerization parameter nbo/t ^f.

Water-induced suppression of liquidus temperatures in the experimental systems, considered with the effects of pressure on the temperature and composition ranges of two-liquid fields in silicate melts, suggests that liquid-phase separation may occur in some H₂O-rich silicate magmas at pressures up to 2GPa. The expanded T–X range of the silicate-liquid miscibility gap and degree of Fe enrichment of the mafic conjugate melts (up to 72 wt% FeO_total) produced by the addition of H₂O, P and S to melts in this study add support to the hypothesis that silicate magma unmixing is involved in the genesis of some Fe-oxide deposits.