C. Wayne Burnham

Water and magmas; a mixing model☆

Abstract A model for the mixing of H2O and silicate melts has been derived from the experimentally determined effects of H2O on the viscosity (fluidity), volumes, electrical conductivities, and especially the thermodynamic properties of hydrous aluminosilicate melts. It involves primarily the reaction of H2O with those O−2 ions of the melt that are shared (bridging) between adjacent (Al, Si)O4 tetrahedra to produce OH− ions. However, in those melts that contain trivalent ions in tetrahedral coordination, such as the Al3+ ion in feldspathic melts, the model further involves exchange of a proton from H2O with a non-tetrahedrally coordinated cation that must be present to balance the net charge on the AlO4 group. This cation exchange reaction, which goes essentially to completion, results in dissociation of the H2O and is limited only by the availability of H2O and the number of exchangeable cations per mole of aluminosilicate. In the system NaAlSi3O8-H2O, upon which this thermodynamic model is based, there is 1 mole of exchangeable cations (Na+) per mole (GFW) of NaAlSi3O8, consequently ion exchange occurs for H2O contents up to a 1:1 mole ratio (Xmw = mole fraction H2O = 0.5). For mole fractions of H2O greater than 0.5, no further exchange can occur and the reaction with additional bridging oxygens of the melt produces 2 moles of associated OH− ions per mole of H2O dissolved. These reactions lead to a linear dependence of the thermodynamic activity of H2O (amw) on the square of its mole fraction (Xmw) for values of Xmw, up to 0.5 and an exponential dependence on Xmw at higher H2O contents. Thus, for values of X m w ↬ 0.5, a m w = k(X m w ) 2 , where k is a Henrys law constant for the dissociated solute. Extension of the thermodynamic model for NaAlSi3O8-H2O to predict H2O solubilities and other behavior of compositionally more complex aluminosilicate melts (magmas) requires placing these melts on an equimolal basis with NaAlSi3O8. This is readily accomplished using chemical analyses of quenched glasses by normalizing to the stoichiometric requirements of NaAlSi3O8, first in terms of equal numbers of exchangeable cations for mole fractions of H2O up to 0.5 and secondly in terms of 8 moles of oxygen for higher H2O contents. Chemical analyses of three igneous-rock glasses, ranging in composition from tholeiitic basalt to lithium-rich pegmatite, were thus recast and the experimental H2O solubilities were computed on this equimolal basis. The resulting equimolal solubilities are all the same, within experimental error, as the solubility of H2O in NaAlSi3O8 melt calculated from the thermodynamic relations. The equivalence of equimolal solubilities implies that the Henrys law constant (k), which is a function of temperature and pressure, is independent of aluminosilicate composition over a wide range. Moreover, as a consequence of the Gibbs-Duhem relation and the properties of exact differentials, it is clear that the silicate components of the melt, properly defined, mix ideally. Thus, a relatively simple mixing model for H2O in silicate melts has led to a quantitative thermodynamic model for magmas that has far-reaching consequences in igneous petrogenesis.

An experimental determination of rare earth partition coefficients between a chloride containing vapor phase and silicate melts

The partitioning behavior of cerium, europium, gadolinium and ytterbium between an aqueous “vapor” phase and water saturated silicate melt have been experimentally examined using a new experimental approach employing radioactive tracers and a double-capsule technique. Equilibrium was established by reversing the partition coefficient1 and by betatrack autoradiography. Aqueous solution compositions were varied by adding different amounts of chloride and in some cases fluoride or carbon dioxide. The H2O contents of the Spruce Pine pegmatite melts were varied by conducting experiments at 4.0 kb, 800°C and at 1.25 kb, 800°C. A jadeite-nepheline composition (75 wt% Jadeite) also was employed at 4.0 kb, 800°C. The chloride experiments (Spruce Pine 4 kb, 800°C) show a linear relationship between the cube of the chloride molality and the partition coefficients of the trivalent rare earths. Europium, under the experimental fO2 conditions (quartz-fayalite-magnetite buffer), varied linearly as the fifth power of the chloride molality. At the chloride molalities examined ( KPGd>KPYb. The smaller ytterbium ion was consistently concentrated in the melt phase relative to the other rare earths in all experiments on the Spruce Pine composition. Experiments on the jadeite-nepheline composition showed no relative fractionation and a positive europium anomaly. The 1.25 kb experiment on the Spruce Pine composition showed a negative europium anomaly in plots of KpRE vs. REE. The overall rare earth partitioning at a constant chloride molality (mCl = .914) was such that KPSP(1.25 kb) > KPSP(4.0 kb) > KPJd-Ne(4.0 kb), where SP = Spruce Pine, Jd-Ne = jadeitenepheli Using the model of Burnnam (1975), It is suggested that the trivalent rare earth partitioning is related to the cube of the melt octahedral site concentration; a property which 1n hydrous melts 1s dependent on melt composition and hydroxyl molality. Excellent agreement was found for the Spruce Pine melt, whereas the jadeite-nepheline melt gave apparent hydroxyl molalities which were too high for the measured partition coefficient. Additional octahedral sites are proposed for this unusual composition perhaps due to some aluminum in 6-fold coordination. The apparent compositional variation of europium partitioning at a constant oxygen fugacity is believed to be related to both the octahedral melt site concentration for trlvalent europium and an 8-coordinated site concentration for divalent europium. Any parameter which affects the numbers of these sites (PH2O, melt composition) will affect the rare earth partitioning. The observed dependency of the partition coefficient on the structural state of the melt could be as significant as its dependency on crystalline structural constraints. Furthermore, since PH2O can drastically effect the melt structural state, its effects could be reflected in melt/crystal partition coefficients.

Crystallization and Fractionation Trends in the System Andesite-H2O-CO2-O2 at Pressures to 10 Kb

Phase relations of a Mount Hood andesite, which has the composition of an average orogenic andesite, have been determined as a function of O 2 fugacity at 1 atm and of H 2 O fugacity to pressures of 10 kb, at O 2 fugacities of the quartz-fayalite-magnetite (QFM) buffer. All runs contained either a H 2 O or H 2 O–CO 2 fluid phase; melts in runs with a H 2 O–CO 2 fluid phase were H 2 O undersaturated. The H 2 O contents of the melts and H 2 O fugacities were calculated from NaAlSi 3 O 8 –H 2 O thermo-dynamic data on the assumption of ideal mixing in the system H 2 O–CO 2 . One-atmosphere runs show that melting relations of silicates are little affected by f o 2 but that both ilmenite- and magnetite-out temperatures are raised by higher f o 2 . Ilmenite precipitates at higher temperature than magnetite. In these runs and in all runs at high pressure with H 2 O and H 2 O–CO 2 fluid phases, oxides were not stable at temperatures of the silicate liquidus. Oxides might be stable on the silicate liquidus if f o 2 rose two or more log units above the Ni–NiO (NNO) buffer. However, calculations indicate that in natural magmas, those processes which might change f o 2 —crystal-liquid equilibria or exchange of H 2 , or H 2 and H 2 O with the wall rocks—cannot raise f o 2 by that magnitude. Because differentiation of basalt melts to andesite must involve iron-rich oxide phase subtraction, such fractionation models appear unreasonable. For the Mount Hood andesite composition, plagioclase is the liquidus phase under H 2 O–saturated conditions to 5 kb and under H 2 O–undersaturated conditions at 10 kb when the H 2 O content of the melt is less than 4.7 wt percent. For higher H 2 O contents, either orthopyroxene or, at H 2 O saturation at pressure greater than 8 kb, amphibole assumes the liquidus. In all cases, clinopyroxene crystallizes at lower temperature than orthopyroxene. Melting curves in the H 2 O–under-saturated region may be contoured either as percent H 2 O in melt or as P e H 2 O ; in either case, the topology of the various silicate melting curves is different from the case of H 2 O–saturated melting. Therefore, melting relations determined at H 2 O–saturated conditions cannot be used successfully to predict melting relations in the H 2 O–undersaturated region. Amphibole melting relations were studied isobarically at 5 kb as a function of temperature and fluid-phase composition. Amphibole has a maximum stability temperature of 940 ± 15°C for fluid compositions of 100 to 44 mole percent H 2 O; for fluids containing more CO 2 than 56 percent (or, equivalently, less than 4.4 wt percent H 2 O in melt), the melting temperature is lower. The same relations would be seen if CO 2 were not present and the melt were H 2 O undersaturated. These rather low melting temperatures, relative to other silicate phases, preclude andesite generation by basalt fractionation involving amphibole at pressures less than 10 kb.

The nature of multicomponent aluminosilicate melts

Abstract The parental silicate melts of the most abundant igneous rocks in the earths crust may be regarded as multicomponent solutions of as many as 15 chemically discrete, electrostatically neutral complexes or species. These species arecorrelated with thermodynamic components and, by adoption of a “quasi-crystalline” model, are identified, both in terms of chemical composition (stoichiometry) and their basic structural components, with the crystalline phases that form on the liquidus under equilibrium conditions. Among the approximately 10 end-member components (species) in any one of these igneous-rock melts, the three aluminosilicate (feldspar-like) components, NaAlSi3O8 (ab), CaAl2Si2O8 (an), and KAlSi3O8 (or), together comprise more than 50 mole % of the solution, hence they may be regarded as a multicomponent solvent in which the several other non-aluminosilicate species are dissolved. The three components of this multicomponent solvent, moreover, are found from the solution behavior of H2O to mix essentially ideally with each other. Similarly, the non-aluminosilicate component, Si4O8 (qz), is found to mix ideally with the aluminosilicates of the solvent. The thermodynamic behavior of the end-member components as a function of liquidus temperature, pressure, and composition (including H2O content), interpreted in accordance with the quasi-crystalline model, lead to the recognition of two basic types of homogeneous melt reactions that result in formation of other species from the end-member components. These end-member components—especially the feldspar-like aluminosilicates (a)—may undergo partial dissociation, or the aluminosilicates may interact with the non-aluminosilicate components. In either type, the reaction may involve no change in the coordination number of the Al atoms (IVAlO45− in the feldspar-like species), or it may involve a change to higher coordination numbers (VAlO57− and VIAlO69−) for at least part of the Al atoms. This tendency toward variable coordination of Al is regarded as the principal factor in determining the nature and extent of speciation reactions in aluminosilicate-rich melts, especially at high pressures where higher coordination numbers are favored. The basis for these interpretations is the difference between the mole fraction of end-member component i (Xiam), as obtained from published experimental liquidus data, and the activity of i (aiam), calculated from internally consistent sets of Gibbs free energies and volumes of melting for crystalline material of pure i composition. These differences for the end-member aluminosilicate components, Xaam − aaam, are either zero (no speciation reaction involving a) or greater than zero (speciation reaction consumes a) in solutions containing such chemically and structurally dissimilar species as Si4O8 (qz) and Mg4Si2O8 (fo), a forsterite-like species. The thermodynamic consequences of attributing the differences to speciation reactions are to return the aluminosilicate components to Raoultian behavior in melts where aluminosilicates crystallize on the liquidus. By the same procedures, Si4O8 and Mg4Si2O8 also become Raoultian in behavior when mixed with the aluminosilicates, but more experimental data are needed to evaluate the mixing properties of other geochemically important melt components. The relationships between Xim and aiam, when expressed in simple empirical equations of the form Xiam − aiam = Xiam ƒ(P, Xjam,Xwm), contain the information required to calculate liquidus equilibrium relations in hydrous (Xwm) and anhydrous (Xi,j,…am) melts of geochemical interest over a wide range of conditions. When interpreted in terms of the quasi-crystalline model, they also provide new insights into the nature of multicomponent aluminosilicate melts and into numerous magmatic phenomena.

CONTACT METAMORPHISM OF MAGNESIAN LIMESTONES AT CRESTMORE, CALIFORNIA

The contact-metamorphic rocks at Crestmore, California, occur between magnesian marbles and a plutonic mass of quartz diorite (Bonsall tonalite) and between the same marbles and a relatively small pipelike hypabyssal mass of quartz monzonite porphyry. The marbles are preserved as two crudely lenticular bodies about 400 and 500 feet thick, respectively, that occur as screens in the quartz dioritic intrusive rocks of the southern California batholith. Both bodies, which are very similar petrographically, are composed of alternating layers of predazzite and coarsely crystalline calcite marbles and are nearly free of silica, alumina, iron, and the alkalies. The quartz diorite that engulfed the marbles is contaminated locally very near the contacts where minor monzonitic and gabbroic variants are present. Its exomorphic effects on the carbonate rocks consist mainly of: (1) the formation of metasomatic silicate contact rocks that generally are less than 1 foot thick and are composed of diopside, wollastonite, and grossularite; and (2) the conversion of nearly all the magnesium-bearing carbonate beds to periclase marbles, which subsequently were altered to predazzites. In contrast, the younger quartz monzonite porphyry, which was injected into the upper or Sky Blue marble unit, is nearly all contaminated as a result of reactive assimilation of marble. Moreover, its exomorphic silicate aureole is as much as 50 feet thick and contains the numerous complex mineral assemblages for which Crestmore is famous. Several lines of evidence indicate that much of this aureole was formed prior to the final consolidation of the porphyry intrusive mass. The thicker parts of the silicate contact aureole that surrounds the quartz monzonite porphyry exhibit the following well-defined zonal distribution of mineral assemblages, as traced outward from the intrusive body: (1) a garnet zone that is composed of grossularite and lesser amounts of wollastonite and diopside; (2) a zone characterized by only one mineral, idocrase; and (3) a zone in which monticellite is the most abundant mineral, but in which there are various amounts of clinohumite, cuspidine, ellestadite, forsterite, melilite, merwinite, perovskite, spinel, spurrite, tilleyite, and xanthophyllite. This zonation of mineral assemblages reflects a corresponding zonation in the bulk chemical composition of the rocks, as shown by changes in the ratio of metasomatic to indigenous constituents (Si + Al + Fe/Ca + Mg) from 0.66 in the monticellite zone, through 1.15 in the idocrase zone, to 1.62 in the garnet zone. There also is a strong tendency toward a mineralogical zonation within the monticellite zone, such that clinohumite, forsterite, spurrite, and spinel are concentrated in the silica-poor and calcite-rich outer part, and merwinite, cuspidine, and melilite in the more silica-rich inner part. Hence, a sequential occurrence of mineral assemblages and a change in bulk chemical composition can be traced inward toward the intrusive mass from un-metasomatized marbles. Moreover, textural features reveal a corresponding paragenetic sequence in which the more highly metasomatized assemblages appear to have formed at the expense of those that were less metasomatized. Available evidence indicates that: (1) the contact-metamorphic mineral assemblages at Crestmore and their zonation are largely the compositionally controlled products of silica, alumina, and iron metasomatism of relatively pure magnesian limestones; (2) temperatures of 625°C. or higher were reached prior to the introduction of silica into the present monticellite-zone rocks; and (3) the so-called “high-temperature” assemblages, such as monticellite, spurrite, and melilite, formed directly from the magnesian marbles without the intervention of “lower-temperature” steps that involve diopside, wollastonite, and grossularite. Therefore, it is proposed that contact metamorphism at Crestmore should be viewed as progressive metasomatism with consequent decarbonation at elevated temperatures rather than as progressive decarbonation attendant simply upon rising temperature.

Solution of H2O in diopside melts: a thermodynamic model

Structural similarities between dry diopside melt and superhydrous albite melt (Xw>0.5) — both lack three-dimensional silicate units — suggest that thermodynamic relations may be similar. A model based on that assumption successfully predicts diopside melting relations and H2O solubilities. For the model, the three partial differential equations describing solution of H2O in albite melt for Xw>0.5 have been integrated for diopside melt from Xw=0 to Xw at least as large as 0.76, with two exceptions: an alternative partial differential equation for Henrian solution of H2O in dilute melts was applied for Xw<0.20, and an alternative differential equation for the pressure dependence of aw at pressures below 2 kbar was developed. The latter alternative equation yields relatively small ¯Vws at low pressures rather than the large ¯Vws calculated from the equation from the albite system. Available experimental solubility data are not precise enough to offer a choice between the small-¯Vw and large-¯Vw equations. Integration of all the partial differential equations was constrained solely by the P and T of a single experimentally-determined point on the H2O-saturated solidus.Solubilities calculated by a Henrian-analogue solution model (adi=Xdi2) from the experimental H2O saturated solidus lie outside experimental solubility constraints for dilute melts. On the other hand, a Henrian model (adi=Xdi) successfully predicts solubilities in dilute melts. The formulation of the Henrian model and magnitudes of model molar entropies of solution are consistent with the hypothesis that H2O dissolves in diopside melt as an essentially undissociated species with little ordering on melt structural sites. That species could in turn be consistently, if not uniquely, interpreted to be molecular H2O or a hydroxylation (OH−) complex formed from nonbridging oxygens.

Calculated melt and restite compositions of some Australian granites

The thermodynamic relations embodied in the Quasicrystalline Model of Burnham and Nekvasil (1986), as recently extended by the author, have been used to quantitatively assess the feldspar-quartz liquidus relations in two I-type (Jindabyne and Moruya) and two S-type (Bullenbalong and Dalgety) suites of Australian granites, using analytical data provided by B. W. Chappell and co-workers. Among the more notable results obtained from these calculations at a constant pressure of 5.0 kbar and X w m =0.30 (≃2.8 wt% H 2 O), for purposes of comparison, are that: (1) felsic melts of remarkably uniform, but distinctive composition can be extracted from each suite, leaving solid residues in amounts up to 65 mol%