American Mineralogist, Volume 80, pages 115-130, 1995

Ca-rich carbonate melts: A regular-solution model, with applications to carbonatite magma + vapor equilibria and carbonate lavas on Venus

ALLAN H. TRElMAN Lunar and Planetary Institute, 3600 Bay Area Boulevard, Houston, Texas 77058-1113, U.S.A.

ABSTRACT A thermochemical model of the activities of species in carbonate-rich melts would be useful in quantifying chemical equilibria between carbonatite magmas and vapors and in extrapolating liquidus equilibria to unexplored PTX. A regular-solution model of Ca-rich carbonate melts is developed here, using the fact that they are ionic liquids, and can be treated (to a first approximation) as interpenetrating regular solutions of cations and of anions. Thermochemical data on systems of alkali metal cations with carbonate and other anions are drawn from the literature; data on systems with alkaline earth (and other) cations and carbonate (and other) anions are derived here from liquidus phase equilibria. The model is validated in that all available data (at 1 kbar) are consistent with single values for the melting temperature and heat of fusion for calcite, and allliquidi are con- sistent with the liquids acting as regular solutions. At 1 kbar, the metastable congruent melting temperature of calcite (CaC03) is inferred to be 1596 K, with AH fu,(calcite)= 31.5 :t 1 kJ/moI. Regular solution interaction param- eters (W) for CaH and alkali metal cations are in the range -3 to -12 kJ/moP; Wfor CaH-BaH is approximately -11 kJ/moP; W for CaH-MgH is approximately -40 kJI moP, and W for CaH-LaH is approximately +85 kJ/moP. Solutions of carbonate and most anions (including OH-, F-, and SO~-) are nearly ideal, with Wbetween 0 (ideal) and -2.5 kJ/moP. The interaction of carbonate and phosphate ions is strongly nonideal, which is consistent with the suggestion of carbonate-phosphate liquid immiscibility. Interaction of carbonate and sulfide ions is also nonideal and suggestive of carbonate-sulfide liquid immiscibility. Solution of H20, for all but the most H20-rich compositions, can be mod- eled as a disproportionation to hydronium (H30+) and hydroxyl (OH-) ions with W for CaH-H30+ 33 kJ/moP. "" The regular-solution model of carbonate melts can be applied to problems of carbonatite magma + vapor equilibria and of extrapolating liquidus equilibria to unstudied systems. Calculations on one carbonatite (the Husereau dike, Oka complex, Quebec, Canada) show that the anion solution of its magma contained an OH- mole fraction of -0.07, although the vapor in equilibrium with the magma had P(H20) = 8.5 X P(C02). Fin carbonatite systems is calculated to be strongly partitioned into the magma (as F-) relative to coexisting vapor. In the Husereau carbonatite magma, the anion solution contained an F- mole fractionof -6 x 10-5. Calcite and anhydrite may be present on the surface of Venus, but they would not be molten at ambient surface temperature (660-760 K) because the minimum melt temper- ature (eutectic) for the calcite + anhydrite system is calculated to be 1250 K. The Venus atmosphere contains 5 ppb HF, which implies that the anion solution of a carbonate-rich magma in equilibrium with the atmosphere would contain a F- mole fraction of -7 x 10-3, or about 0.1 wtOJo.Although this proportion of F is much enriched compared with the atmosphere, it would have little effect on phase relations of the carbonatite.

INTRODUCfION (Treiman and Schedl, 1983; Dawson et aI., 1990; Keller and Krafft, 1990; Watson, 1991; Norton and Pinkerton, Carbonatites are igneous rocks that formed from car- 1992), their thermochemical properties have been stud- bonate-rich magmas. The petrogeneses of carbonatites are ied little (Bradley, 1962; Treiman, 1989). Thus, investi- imperfectly understood, in part because of uncertainties gations of carbonatites have not benefited from quanti- in the physical and chemical properties of their parent tative thermochemical models such as have been magmas. Although the physical and mass-transport prop- developed for silicate magmas (e.g., Ghiorso et aI., 1983; erties of carbonatite magmas are becoming appreciated Berman and Brown, 1987; Ghiorso, 1987). 0003--004X/95/0 102--0 115$02.00 115 ~ ------

116 TREIMAN: Ca-RICH CARBONATE MELTS

For instance, a thermochemical model of carbonate The Temkin model is consistent with ideal behavior in each melts would provide a quantitative link between the ion solution and also with regular behavior, in which there compositions of carbonatite magmas and the composi- is heat of mixing but no excess entropy of mixing (F0rland, tions of their associated volatile phases. Carbon dioxide 1955). Here, I use the simplest version of the Temkin mod- is obviously important in carbonatite magmas; H20 has el, in which all cations occupy identical quasi-lattice sites, played a prominent role in experimental studies of car- as do all anions. This model ignores complexation except bonatite genesis (Wyllie, 1989), and the potential impor- as reflected by regular solution behavior and ignores the tance of F has recently been reemphasized (Gittins et aI., expectation of differing sites in the liquid quasi-lattice. The 1990; Jago and Gittins, 1991). In addition, most plutonic Temkin model is only an approximation because local charge carbonatites are surrounded by volumetrically significant balance does not permit ions of different charges to inter- zones of metasomatized rock, i.e., fenite (McKie, 1966). change completely freely (e.g., Ca2+ vs. Na+) and because These zones bespeak large fluxes of volatiles associated common sense (and the Gibbs-Duhem relation) suggest that with carbonatite magmas. It has been possible to con- different ions affect their surrounding ions in different ways. strain the composition of the volatile phases through the For instance, one cannot expect ions of different sizes (e.g., compositions and phases of the solids with which they Ca2+ vs. Mg2+) and charges (e.g., Ca2+ vs. Na+) to maintain equilibrated (e.g., Rubie and Gunter, 1983; Treiman and identical distances and coordinations with surrounding ions. Essene, 1984; Kresten and Morogan, 1986; Andersen, With these caveats, the Temkin model is a good first 1986). However, it has been impossible to constrain com- approximation for the properties of many ionic liquids, positions of carbonatite magma from fluid compositions, including (as I show below) those of Ca- and carbonate- except in the most general terms. With a quantitative rich melts to the level of detail permitted by most avail- thermochemical model of carbonatite magmas, the con- able data. In addition, the regular solution model is fa- nection between fluid and magma compositions would miliar in the geological community and is the simplest be straightforward. formulation of real solutions (e.g., Ghiorso et aI., 1983; Similarly, a thermochemical model of carbonate melts Ghiorso, 1987; Berman and Brown, 1987; Helffrich and would permit extrapolation of known liquidus equilibria to Wood, 1989). More physically accurate models for ionic physical and chemical conditions that have not been studied salts (e.g., the reciprocal-salt or conformal-solution mod- experimentally. In this way, a thermochemical model would el: Blander and Topol, 1966; Kleppa, 1977, 1981) may provide a structure for understanding the results of experi- be better representations of reality, but they are not jus- ments already completed, a ready way of applying experi- tified by the quality and quantity of data available. mental results to complex natural systems, and an aid in designing new experimental programs. INTERPRETATIVE METHOD In this paper, I propose a thermochemical model of For some components in carbonate magmas, thermo- carbonate-rich magmas, based on regular-solution theory chemical data can be taken directly from the literature. and the observation that carbonatite magmas are ionic But for many major components, like Ca and Mg car- liquids (Treiman and Schedl, 1983). In the model, all bonates, and for magmas at high pressure, such data must available data (at 1 kbar) are consistent with a single tem- be gained indirectly. The most accessible sources of these perature of melting for calcite (metastable congruent data are liquidus phase equilibria (in effect, measure- melting), and a single value for the heat of fusion for ments of freezing point depressions), which can be ma- calcite. Similarly, the locations of all liquidus calcite-sat- nipulated to retrieve heats of fusion and activity-com- urated liquidus surfaces are consistent with the Ca-rich position relationships (Lewis and Randall, 1961). carbonate melts being regular solutions. Portions of this To simplify the interpretation ofliquidus surfaces, sol- model were presented by Treiman (1989), which is su- id and liquid phases both must be referred to the same perseded by this work. standard state. For simplicity and consistency with geo- logical applications, the standard state for a component TEMKIN MELT MODEL is taken as the chemically pure phase in its equilibrium Carbonate-rich melts are ionic melts or fused salts, liq- structure at the temperature of interest. Thus, pure solid uids in which the discrete entities (ions) are charged and phases below their melting temperatures have activities bound by electrostatic forces (Zarzycki, 1962; Sundheim, of unity; hypothetical pure liquids below their solidifi- 1964; Lumsden, 1966; Kleppa, 1977, 1981). Polymeriza- cation temperatures have activities exceeding unity. Ac- tion of anions (as in silicate liquids) is unimportant, and tivities of components in solutions (solid and liquid) are ionic complexes can be treated as distinct ionic species. referred to the same standard state. For solid phases be- Ionic liquids are amenable to relatively simple thermo- low their melting temperatures, this is a normal solvent chemical analysis because their cations and anions may be standard state: the ratio of activity to mole fraction for a treated, to a first approximation, as independent solutions. component (a/X) is unity for the pure component (X = This approximation is the quasi-lattice or Temkin (1945) I). For liquid solutions phases, this is also a solvent solid model. It is justifiable because enormous energy would be state, but with the pure solvent having nonunit activity needed to exchange, for instance, a cation surrounded by at subsolidification temperatures. The hypothetical pure anions for an anion surrounded by anions (Blander, 1964). liquid in its standard state must have the same structure TREIMAN: Ca-RICH CARBONATE MELTS 117 as the solution, which need not be the same liquid struc- (Andersen and Lindsley, 1981). In an ideal solution, all ture as in the pure system at its melting temperature. W = 0, and so 'Y = 1. In the Temkin model of ionic Freezing point depression is treated in detail in stan- liquids, the cations and anions are treated independently dard thermodynamics textbooks (e.g., Lewis and Rand- as regular solutions, each with its own X, n, W, and 'Y all, 1961). Consider the isobaric melting reaction A (sol- terms. Substituting Equations 3 and 5 into Equation 2 for id) ;=' A (liquid solution) in the system A-B at T < Tfu" the system A-B and rearranging into the format y = ax the congruent melting temperature of A (solid). One can + b yields approach this state in two steps: melting of pure A at T < Tfu" and isothermal solution of B into the melt and _ RT In(XA,melt) - RT In(aA"olid) solid until they are at equilibrium. For the first step, the T 1-- free energy of melting pure A at a temperature below T Tfu, < Tfu, is given as (1 - X A,melt)2 (A) (6) - - T = W AD + Illi fus . b.Gl~, = Illi ~u, I - y- (1) T ° ( ) 1-- fu' Tfu, where llli~u,(A) is the molar enthalpy change on melting (heat of fusion) of pure phase A at Tfu, (viz., Flood et a1., If the phase A is purely component A, its activity is unity, 1949; Lewis and Randall, 1961, p. 415). Equation I as- and Equation 6 simplifies to sumes that the heat of fusion is not a function of tem- _ RT In(XA,melt) (1 - XA,melt)2 IlliO ( ) ( ) perature, i.e., the effect of b.Cp,fu,(A)on melting A is rel- = WAB + fus A . 7 T T atively small, which is justifiable for carbonate melt 1-- 1-- systems at the present level of precision. Ignoring b.CP.ru, Tfu, Tfu, in carbonate, chloride, or nitrate systems causes < 1% For points on this A-saturated liquidus, a graph of - RT error in the Illi and < 10%error in regular solution fu' In(XA,melt)/(1- T/Tru,) vs. (I - XA,melt)2/(1- TIT fus)should parameters (Treiman, unpublished calculations). This yield a straight line of slope WABand intercept llli~u,(A) b.GTu,(A)is positive because a melt of pure A composition (e.g., Flood et a1., 1949). Typically, phase-equilibrium ex- is not stable relative to the solid at T < Tfu,' The second periments yield bracketed ranges in T and X within which step is isothermal solution of B into the melt and solid a liquidus must lie, and so this graph would consist of until they are at equilibrium, i.e., b.GTu,(A) + b.G;;'lution= brackets through which the straight line must pass. These O. Substituting this and the definition of free energy brackets can typically be satisfied by ranges of WABand changes with composition into Equation I yields a de- llli~u,(A). scription of the liquidus surface, These descriptions of the liquidus surface can be relat- ed to the Temkin model of ionic melts with a few defi- I - ~ Illi~u,(A) = -RT(ln aA,melt- In aA,rolid) (2) nitions of standard state for ion activities. As with solid Tfu, ( ) and liquid phases, the reference state for an ionic species where a is activity of a component in a phase, relative to is a crystalline solid containing that pure ionic species at the standard states given above. the temperature of interest. For example, systems in equi- The activity of A in the melt phase can be calculated librium with the pure solid phase Ln+Mn- have activities from the assumption of regular solution behavior by of ionic components Ln+ and Mn- of unity. If a solution phase contains both Ln+ and Mn- ions, means of the activity coefficient 'YA (3) (8)

For a multicomponent solution of species A, B, and C, If a solid phase contains only a single cation or anion the heat of mixing is given as species, the activity of that ion species is unity, and the activity of the phase is equal to the activity of the other Illimix = (nA + nB+ nd(XAXBWAB + XAXCWAC ion species. Activities of individual ion species in melt + XBXCWBC+ XAXBXCWABC) (4) solutions can be determined by their mole fractions and regular solution interaction parameters (Eqs. 3 and 5). where n is the number of moles of a species present, and The activity of a component in a melt solution is refer- W is the interaction parameter for that pair (or triplet) of enced, as before, to the pure component in its equilibrium species (Lewis and Randall, 1961). The ternary interac- phase at that T. Thus, in a melt in equilibrium with solid, tion parameter is assumed here to be zero, although this pure Ln+Mn-, the activity product aL"+'aM"- is unity. value is not required by theory (Helffrich and Wood, Other thermochemical quantities are calculated with 1989). The activity coefficient for species A is then 'Y standard methods. The molar entropy of melting is cal-

RT In 'YA = (X~ + XBXd WAB+ (X~ + XBXC)WAC culated from the heat of melting as (5) (9)

-- ---

llS TREIMAN: Ca-RICH CARBONATE MELTS

TABLE1. Melting of alkali and alkaline earth carbonates: Molar TABLE2. Melting of alkali and alkaline earth carbonates: Molar properties at 1 bar properties at 1 kbar

T... ilH~. ilS::" il~p.tu. ilV... T... ilH::.. ilS::" il~p.... ilV... Compound K kJ/mol J/(mol'K) J/(mol'K) cm3/mol Compound K kJ/mol J/(mol' K) J/(mol'K) cm3/mol

Na,C03 1131 29.7 26.3 -8.5 4.7* Na,C03 1145* 30.1 26.3 -8.0 4.7 K,C03 1171 27.6 23.6 -1.1 7.5** K,C03 1200* 28.2 23.5 -1.1 4.8 Li,C03 996 44.8 44.8 -7.6 2.6 Li,C03 1003** 45.0 45.0 -7.6 2.6 CaC03 t 1583 30.5 :t 1 19:t 2.5 CaC03 1596t 31.5 :t 1 19.7 :t 0.7 2.5 :t 0.1 CaC03:j: 1463 38.5 :t 4 26 :t 3 MgC03:j: 1750 32 :t 25? 18:t 15? 0.7 :t 0.6? Note: enthalpy and entropy from Janz et al. (1979), heat capacities Note: heats and entropies of alkali carbonates extrapolated from 1-bar from Selman and Maru (1981), and volumes from Klement and Cohen values; 7;", measured; volumes from high-pressure phase equilibria. Data (1975) and Janz et al. (1979). for alkaline earth carbonates as derived in text. * Volume from Klement and Cohen (1975). Janz et al. (1979) gave 7.5 * Koster van Groos and Wyllie(1966). cm3/mol. ** Klement and Cohen (1975). ** Volume from Klement and Cohen (1975). Janz et al. (1979) gave 10.3 t Extrapolated from Irving and Wyllie (1975); see text. cm3/mol. :j: Ttu, extrapolated from Irving and Wyllie (1975). Other values estimated t Appropriate for Ca-rich melts, and those containing significant K vs. from liquidus surface of Ragone et al. (1966) without consideration of Na. Recalculated from F0rland (1955), using Ttu,extrapolated from high possible experimental errors. See text. pressure. These data are consistent with high-pressure determinations; see text. :j:Appropriate for lower temperature, Na-rich melts. Recalculated from F0rland (1955) and Flood et al. (1949). fraction, rather than from density measurement (used by Janz et aI., 1979). because congruent melting of a pure phase is isothermal. CALCIUM CARBONATE Volume changes on melting can be derived from direct Calcite is the most abundant mineral in most carbon- measurement or from polybaric equilibria by the Clau- atites, intrusive and extrusive (Bailey, 1993), and so cal- sius-Clapeyron equation: cium carbonate is likely to be among the most important components in carbonate magmas. The melting proper- ~ -Vrus= ~Sfus (10) (dP/dT)rus' ties of CaC03 at I bar and I kbar must be inferred in- directly because calcite does not melt congruently at these Extrapolation of heats of melting over temperatures and pressures; pure calcite decarbonates below 40 bars (Ba- pressures follow from the partial derivatives of enthalpy: ker, 1962) and melts incongruently to liquid + vapor between 90 and -7000 bars (Irving and Wyllie, 1975; atlll rus - = ~ c p.rus and Huang and Wyllie, 1976). However, melting and solution aT properties of CaC03 can be measured directly for melts where ~C p.rusis the difference in heat capacities between that are not too rich in CaC03 component, given a tem- molten and solid phases (Tables I and 2). Note that perature for its (metastable) congruent melting. ~C p.rusis negative for the alkali carbonates (Tables I and All high-pressure liquidus surfaces and most I-bar li- 2), as it is for many ionic salts (Robie et al., 1979; DeKock, quidus surfaces are consistent with a single value for the 1986). A negative ~C P.fussuggests premelting structural temperature of melting (Trus) and heat of fusion changes in the solid and does not violate the second law tlll~us(calcite); unless specifically noted, all discussion here of thermodynamics. refers to numerical values consistent with l-kbar liquidus equilibria. However, some I-bar liquidi are consistent ALKALI CARBONATES with a separate set of Trusand tlll~us(calcite) (Table I); it Alkali carbonates are inferred to be important constit- is possible that Ca-rich carbonate melt might occur in uents of some carbonate magmas (LeBas, 1981; Dawson two distinct structures at I bar (viz., F0rland, 1955). For et aI., 1987; Gittins, 1989), although few carbonatites the most part, melting properties derived for high pres- contain alkali carbonate minerals. There is an extensive sure are appropriate for geological applications. literature on melting and mixing properties of the alkali carbonates, from which much of Table I is drawn directly Melting temperature or calculated. The congruent melting temperature for pure CaC03 in For volume changes on fusion, ~Vru" for the alkali the calcite structure, Trus(calcite), can be estimated by ex- carbonates (Table I), the data of Klement and Cohen trapolating the high-pressure congruent melting curve to (1975) are preferred over those of Janz et al. (1979), which lower pressures (Irving and Wyllie, 1975; Huang and are larger than those of Klement and Cohen by -20%. Wyllie, 1976). Between 10 and 20 kbar, the congruent The discrepancy in ~ V rus lies in the volumes of the solid melting curve for CaC03 has a slope of -12.5 K/kbar, phases, as both groups report comparable melt volumes. implying a congruent Trus(calcite) of 1583 K at I bar and Klement and Cohen's (1975) data are preferred, because 1596 K at I kbar (Tables I and 2). This l-kbar Tfus(calcite) their solid volumes are from high-temperature X-ray dif- is just above the experimentally determined bracket for TRElMAN: Ca-RICH CARBONATE MELTS 119

TABLE3. Expressions for ordinate Q values in Figs. 1 and 2 i!J 40 i!J - Q1 = -RTln(Xe''',~,;onm''''Xeo!',";onmelt) / (1- L) "0 !IJ !IJ Q2 -RTln(Xe,"~t;on~t) E = (1- L) -- 30 / ~- I> ,. Q3 = -RTln(XoHao~om"') (1- ~) ,.... . ~,. / 0 20 f I> Q4 = -RTln(XFao;onmelt) / (1- L) . )] OS = - [R T In(Xea"~t;onm~t)- R T In(aeaeo,,~ - ~) 0 0.05 0.1 0.15 0.2 0.25 0.3 / (1 _ [RTln(Xe"',~_~t'Xeo!-,oo;on~t) + Weo!- oH-(1 - XOW"o;onmelt)'] (1- X )2/( 1- Q6 = Ca2+,cation melt ~fuJ (1 - TlTtu,)

Fig. I. Regular-solution interpretation of calcite-saturated li- _ [R T In(Xca2+,cationmelt'Xco~-,anionmeh)+ Wc~- oH-(1 - Xc~-,aniOnmelt)2] Q7 = (1 quidus in CaC03-BaSO.-CaF2 (500 bars C02' Kuellmer et aI., - T/T.,,) 1966), following Eq. 12. See Table 3 for definition of Q I. Open and filled symbols represent liquid-only and liquid + calcite experiments, respectively. Lines from each data point represent a conservative estimate of errors (:t5 Kin T; :to.2 wt% in X), imately zero (derived below, viz., Table 4), so Equation showing only the error bar halves that contribute to uncertainty 8 reduces to in slope and intercept. The liquidus surfaces, linear with slope Wand ordinate intercept of Ml~u,(calcite), should lie between _ RT In(XcaH,cation melt' XcOj-,anion melt) these groups of experiments. Within error, all data are consistent T with ~H~u,(calcite) = 31.5 :t I kJ/mol and WCa'+_Ba2+= -11 :t 1-- 9 kJ/moP (thin solid lines). Tcu,

T incongruent melting [1573-1593 K; Wyllie and Tuttle, 1-- 1960 (temperatures corrected by - 32 K per Gittins and Tcu, Tuttle, 1964); Irving and Wyllie, 1975], and is consistent + Ml~us(calcite). (12) within error with much of the liquidus equilibrium data for 1 bar (F0rland, 1955). As discussed below, some li- The data of Kuellmer et a1. (1966) are recast by this equa- quidi at 1 bar are consistent with Tcu,(calcite) = 1463 K. tion in Figure 1. Ifthe Temkin regular solution model is valid, and if the anion interaction parameters are effec- Heat of melting tively zero, the liquidus surface separating the liquid-only High pressure. The locations of calcite-saturated liqui- data points (open symbols) and liquid + calcite points di at high pressure (500- 1000 bars) imply that (filled symbols) should be representable as a straight line Ml~u,(calcite) = 31.5 :t 1 kllmo1. This value is con- with slope of Wca2+_Ba2+and a Y-axis intercept of Ml~us(ca1- strained most closely by the location of the calcite-satu- cite). The liquidus can in fact be represented as a straight rated liquidus in the system CaC03-CaF2-BaSO. (Kuell- line consistent with the error bars of all points, suggesting mer et aI., 1966), and is consistent with all available Ml~us(calcite) = 31.05 :t 0.25 kllmo1. However, these determinations of calcite-saturated liquidi at high pres- tight error limits are dictated by a single liquid-only point sure, in which calcite is a pure phase. (the open symbol that extends below the lines). The con- The experimental location of the calcite-saturated li- servative approach taken here is to assume that point is quidus in the system BaSO.-CaC03-CaF2 (Fig. 1, Table in error and to estimate Ml ~us(calcite)= 31.5 :t 1 kllmol 3; Kuellmer et aI., 1966) is the most restrictive available and WeaH-Ba'+= -11 :t 9 kllmoF from the remaining constraint on Ml~u,(calcite). Calcite grown in this system points (Fig. 1). The error limits correspond to a temper- is effectively pure; it does not accept significant SO~- of ature uncertainty of :t 5 K and a compositional uncer- F- in solid solution, and Kuellmer et a1. (1966) reported tainty of :to.2 wt% in the most abundant component. no indication of solid solution with Ba. To retrieve This value for Ml~us(ca1cite) is consistent with all avail- Ml~u,(calcite) from this system, one substitutes Temkin able high-pressure determinations of liquidi saturated in melt activity models (Eq. 5) for individual anion and cat- pure calcite. Figure 2a-2f show many sets of calcite-sat- ion solutions into Equation 8 and then substitutes the urated liquidi recast following Equation 7. Within uncer- resultant melt activity model into Equation 7, the de- tainty, all these liquidi are consistent with Ml~us(calcite) scription of the liquidus surface. The anion interaction = 31.5 :t 1 kllmo1. This value is essentially identical to parameters WCOj-_F-,Wsoa-F-, and Wcoj--soa- are approx- the only independent estimate of Ml~us(calcite) at high -----

120 TREIMAN: Ca-RICH CARBONATE MELTS

a. b. 35 35 ..--.. o -o E E 30 -""") -""") ~ ~ --- - 25 C\J C\J o Q 20 20

15 o 0.2 0.4 0.6 0.8 1 1.2 1.4 0.2 0.4 0.6 0.8 1 (1- Xea",cationmeu)2/(1- %fuS) (1- 1- Xca,..catiOnmeltY /( %;us) 38 c, 36 34 ..--.. 34 ..--.. <5 o 33 -E 32 E ~ 30 -""") 32 --- ~--- 28 r--- 31 26 J

24 II 30 220 0.5 1 1.5 0.2 0.4 0.6 0.8 1 1.2 1.4 (1- Xow, aJ1iOnmej /( 1- %J (1- 1- 35 XF-,anionmeltY/( %fuJ e, 30 f. -(5 25 E ..--.. -- <5 ~ 20 E - 15 - LO ---~ Q .. 10 o<.c 50 ~1 ~2 ~3 ~4 O~ 0.6

(1- 1- XC"',catiOnm,"Y /( If.,) 0.02 0.04 0.06 0.08 0.1 0.12 160 g, (1- Xc"'.c.tionm,uY /( l-Iffus)

120 -o E pressure,29 kJ/mol (Bradley, 1962). Bradley assumed that -~ 80 l-kbar melts in CaC03-Ca(OH)2 were ideal solutions and --- performed an analysis of the freezing point depression I'-- similar to that here. Results here on anion solutions are Q 40 consistent with near-ideal mixing of carbonate and hy- droxide anions (Fig. 2c, Table 4). o One bar. Most of the limited experimental data on the o 0.5 1,5 2 2.5 3 melting properties of calcite at 1 bar are consistent with results from high pressure: Tcu,(calcite) = 1583 K and Ml~u,(calcite) = 31.5 ::!::I kJ/mo1 [Table I; Flood et aI., TREIMAN: Ca-RICH CARBONATE MELTS 121

Fig. 2. Regular-solution interpretation of calcite-saturated li- Wea" -K+= -14.5 :t 2.5 kJ/moP. (c) CaC03-Ca(OH), (I-kbar quidi from other high-pressure experiments. Open and filled CO,; Wyllie and Tuttle, 1960; published temperatures corrected symbols represent liquid-only and liquid + calcite experiments, by - 32 K according to Gittins and Tuttle, 1964), following Eq. respectively. See Table 3 for expressions of ordinate 'Q' values. 7. Wcoj -ow = 2.3 :t 2.3 kJ/moP. (d) CaC03-CaF, (Gittins and Heavy lines from each data point represent a conservative esti- Tuttle, 1964; Kuellmer et aI., 1966), following Eq. 7. The data mate of errors (:t5 Kin T; :to.2 wt% in X), showing only the sets are not consistent within error limits; ignoring the two dis- error-bar halves that contribute to uncertainty in slope and in- crepant points (ordinate values of 30.5 and 32) permits tercept. The liquidus surfaces, linear with slope Wand ordinate Weoj-_F- = 0 :t 2 kJ/moP. (e) CaC03-MgC03 (IO-kbar CO,; intercept of Ml?u,(calcite), should lie between these groups of Byrnes and Wyllie, 1981), following Eq. 6. Activities of CaC03 experiments. Thin solid lines encompass range of liquidus sur- in calcite calculated ITom Anovitz and Essene (1987). Wea2+.Mg"= faces permitted by these data (with error bars), and Ml?u,(calcite) -40 :t 30 kJ/moP. (f) CaC03-Ca(OH),-La(OH)3 (I-kbar CO,; = 31.5 :t I kJ/mol (Fig. I); the given range of W values is for Jones and Wyllie, 1986), following Eq. 13. Wea" _La"= 85 :t 50 these solid lines. (a) CaC03-Na,C03 (I-kbar CO,; Cooper et aI., kJ/moP. (g) CaC03-H,O (I-kbar CO,; Wyllie and Tuttle, 1960, 1975), following Eq. 7. Wea2+_Na+= -6 :t 2 kJ/moP. (b) Caco3- temperatures corrected by -32 K according to Gittins and Tut- K,C03 (I-kbar CO,; Cooper et aI., 1975), following Eq. 7. tle, 1964), following Eq. 15. WeaH_H,o+= 33 :t 2 kJ/moP.

1949; Ferland, 1955; the location of the calcite-saturated consistent with the earlier liquidus experiments of Flood liquidus in CaC03-Na2C03 by Poletaev et aI., 1975, spans et ai. (1949). They determined the CaO-saturated liqui- too small a composition range to constrain .:lH?u,(calcite)]. dus surfaces in CaC03-Na2C03, CaC03-K2C03, and However, the liquidus position in the CaC03-Na2C03 CaC03-Li2C03 under 1 bar C02' from 1244 to 1378 K. system at temperatures below approximately 1170 K is Activities of calcite in the melt solutions were calculated consistent with Tru,(calcite) = 1463 K and .:lH?u,(calcite) from the pressure of C02 in equilibrium with calcite and ;::::38 kJ/mol (Table 1; Flood et aI., 1949; Ferland, 1955). lime; uncertainties were given only as error bars on graphs To account for these discrepancies, Ferland (1955) sug- and cannot be readily evaluated. Flood et ai. (1949) took gested that lower temperature melts in CaC03-Na2C03 the melting temperature for calcite to be 1613 K, used a do not have the same structure as higher temperature molecular mole fraction model for melt activities, and melts and melts in other systems (notably K-bearing). calculated (from Eq. 7) that .:lH?u,(calcite) = 14.2 kJ/moi. There is no evidence that the low-temperature structure Recalculating their data for CaC03-Na2C03 for melting persists to higher temperature or pressure in the Ca-rich temperatures of 1583 or 1463 K and with a Temkin melt systems examined here. model (e.g., ionic fractions) yields .:lH?u,(calcite) ;::::39 kJ/ Extrapolations of Tru,(calcite) and .:lH?ua(calcite) from mol, consistent with Ferland's (1955) data on Ca-poor high pressure are consistent with most of the I-bar liqui- compositions in CaC03. The determinations for CaC03- dus experiments of Ferland (1955), who calculated both K2C03 and CaC03-Li2C03 of Flood et ai. (1949) are more values from the compositions of melts saturated with Cao scattered and are consistent with either pair of Tfus(calcite) (lime) in the systems CaC03-Na2C03, CaC03-K2C03, and and .:lH&,,(calcite). CaC03-NaKC03 as functions of C02 pressure between To explain the discrepancies in .:lH?u,(calcite) and 1203 and 1273 K. Compositions were measured by weight T£o,(calcite), Ferland (1955) suggested that the more Na- loss (C02 loss); a (calcite) was calculated from measured rich and lower-temperature melts in CaC03-Na2C03 have C02 pressure and the known pressure of C02 in equilib- a different structure from those at higher Ca contents and rium with calcite and CaO; uncertainties were not given temperatures. On the basis of the .:lH ru,(calcite) values, and cannot be evaluated. The linear correlation of .:lG£0, the Ca-rich melt structure is present in all systems at high (calcite) and T for the systems CaC03-K2C03 and Cac03- pressure. Another speculative explanation is that the sol- NaKC03implied Tfus(calcite);::::1523 K and .:lH&,,(calcite) id in the CaC03-Na2C03 experiments was not actually ;:::: 35 kJ/moi. The original data are consistent with CaO but a mixed oxide phase in CaO-Na20. I am aware, Tfus(calcite)= 1583 K (extrapolated above from high-pres- however, of no reports of mixed Na-Ca oxide phases. sure equilibria),which yields.:lH?ua(calcite) ;::::30.5 kJ/mol, consistent with high-pressure phase equilibria. Liquidus ex- Melt volume periments at high Ca contents and higher temperatures in The volume change on melting calcite at high pressure the system CaC03-Na2C03, are also consistent with the high- may be calculated from Equation 9. The entropy of fu- pressure values, although data are limited. sion, .:lSru,(calcite), is calculated from .:lH?u,(calcite) and However, at lower temperatures and lower Ca con- Tru,(calcite), as in Table 2. The slope of the polybaric tents, the CaO-saturated liquidus in CaC03-Na2C03 is congruent melting for calcite curve is 80 K/bar (Irving not consistent with Tru,(calcite) and .:lH?u,(calcite) from and Wyllie, 1973, 1975), yielding .:lVru,(calcite) = 2.5 ::!: high-pressure phase equilibria. Rather, Ferland (1955) 0.1 cm3/mol at high pressure. This value is comparable found that these liquidus determinations suggested with .:lV rusfor Li2C03, but significantly smaller than those Tru,(calcite) = 1463 K and .:lH?u,(calcite) ;::::37.5 kJ/moi. for K2C03 and Na2C03 (Table 2). This higher .:lH?u,(calcite) and lower Tfus(calcite)are also The molar volume of CaC03 melt could now be esti- 122 TRElMAN: Ca-RICH CARBONATE MELTS

TABLE4. Mixing of ions in molten salts: Regular solution pa- in_Na at low pressure and relatively lower temperature, rameters Mf~us(calcite) - 38.5 kJ/mol, and Tfus(calcite) = 1463 K. W Counter On the basis of molar entropies, one may speculate that Ions (kJ/moJ2) ion Reference the former melts are structurally comparable to molten

CO~--OH- -2.3 :t 2.3 Ca2+ 1 K2C03 and the latter are structurally comparable with -0 Na+ 2 molten Na2C03. CO~--F- 0:t2 Ca2+ 1 K+ --2 3 MAGNESIUM CARBONATE CO~--CI- -1.7.0 Na+ 3.4 CO~--Br- -1.7 Na+ 4 The common presence of magnesian calcite, dolomite, Na+ CO~--Sry,- 0 5 and magnesian silicates in carbonatites shows that mag- O:t 1.5 Ca'+ 1 CO~--PO~- >65? Ca2+ 1 nesium carbonate is an important component in carbon- Cry,- _0'- -0 Na+ 2 atite petrogenesis. Unfortunately, data on the melting and CO~--O~- -0 Na+ 2 F--OH- +4.3 Ca2t 6 thermophysical properties of magnesium carbonate are F--SO~- -0 Na+ 7 either absent or uncertain. Dolomite, CaMg(C03)2, is the Ca'+ -Na + -6 :t 2 CO~- 1 most common Mg-bearing carbonate in carbonatites, but --10 CO~ 8 Ca'+-K+ -14.5 :t 2.5 CO§- 1 it decarbonates at low pressure and melts incongruently --24 CO~- 8 at high pressure. In addition, there appear to be no avail- Ca2+ + -Li -2.5 CO~- 4 able liquidus equilibria that can be used as above to de- Ca'+-H3O+ 36 :t 2 CO~--OH- 1 Ca2+-Mg'+ -40 :t 20 CO~- 1 rive its melting properties. Data on Mg in carbonate melts Ca2+-Ba'+ -11 :t 9 CO~--SO~- 1 must now come from the limited studies available in- Ca2t -La3+ +85 :t 50 CO§--OH- 1 volving MgC03, magnesite. Mg2+-K+ -20 :t 30? CO§- 1 Na+-K+ -5.6 CO~- 9 Magnesite melts incongruently at pressures below 25 Na+-Li+ -11.2 CO§- 9 kbar, and the hypothetical congruent melting tempera-

Note: 1 ~ this work; 2 ~ Selman and Maru (1981); 3 ~ from phase ture must be extrapolated from there to the range of in- diagrams in Levin et al. (1964. 1969); 4 = Lumsden (1966); 5 = Flood et terest. Using the high-pressure liquidus determinations of al. (1952); 6 = average value from Tacker and Stormer (1993); 7 = average Irving and Wyllie (1975), the I kbar congruent melting value from Kleppa and Julsrud (1980); 8 = F0rland (1955) and recalculation of Flood et al. (1949); 9 = average value from Andersen and Kleppa (1976). temperature for MgC03 may be extrapolated as 1753 K. The Mf~us(magnesite) is very poorly determined by the single available liquidus location, in MgC03-K2C03 (Ra- mated from this if the molar volume of solid CaC03 were gone et aI., 1966), for which magnesite is not a solid so- known. At its l-kbar melting temperature, CaC03 would lution. Taken at face value, the brackets of Ragone et a1. be in the CaC03 (V) polymorph (Carlson, 1983); unfor- (1966) on the magnesite-saturated liquidus surface only tunately, molar volumes have been measured only to 1148 restrict Mf~ulmagnesite) to a value of 32 :t 25 kJ/mol K and 1 bar, where CaC03 (V) is the stable polymorph and (Table 2), with a corresponding entropy of fusion of (Mirwald, 1979). Recklessly extrapolating molar volume 18:t 15 J/(mol'K) (Table 2) and a WMg2+.K+of -20 :t 30 and compressibility data for CaC03 (IV) (Mirwald, 1979) kJ/moP. However, including reasonable experimental un- to 1583 K and 1 kbar, I estimate a molar volume for certainties in the analysis (5 K and 0.2 mol% MgC03) only solid CaC03 of 39 cm3/mo1. This value leads to a molar restricts Mf~ulmagnesite) to >7 kJ/mol and WMgH.K+to volume for liquid CaC03 of 41.5 cm3/mol and a density < 10 kJ/moP. Oearly, much work remains. of 2.4 gm/cm3. This density is comparable with the 2.2 gm/cm3 inferred for a Ca-rich carbonatite magma (Nes- THERMOCHEMISTRY OF SOLUTION bitt and Kelly, 1977). In the regular-solution model, there is a heat effect in the formation of a solution, but no entropy effect beyond Structures of CaC03-rich melts that of random mixing of constituents (Eq. 4; viz., Lewis For modeling Ca-rich carbonate melt systems, it seems and Randall, 1961). The heat effect is described by a sin- reasonable to accept Wfus(calcite) and Trus(calcite) from gle interaction parameter, W, for each possible pair (or the l-kbar experiments (Table 2) and as extrapolated to multiplet) of species in a solution. For a Temkin ionic other pressures (Table 1). There are no obvious problems solution, there are independent Ws for the cation and with experiments or interpretation to explain the differ- anion solutions. The regular solution parameters Ware ences between the inferences from high-pressure phase obtained simultaneously with estimates of !::Jlfus>and so equilibria and the results of Flood et a1. (1949) and some have already appeared above in discussions of Figures I, results of F0rland (1955). It is quite reasonable to infer, 2, and 3. as did F0rland (1955), that carbonate melts at low pres- sure can adopt multiple structures. The structure ob- Anion mixing tained at high pressure (1 kbar) seems to be retained at I It is likely that the anion solution of carbonatite mag- bar for compositions rich in Ca and those containing sig- mas is dominated by the carbonate anion, but other an- nificant K; for these melts, W~us(calcite) = 31.5 :t I kJ/ ions may play an important or essential role in carbonate mol and Tfus(calcite) = 1583 K. For melts relatively rich petrogenesis. For instance, the presence of OH anions TREIMAN: Ca-RICH CARBONATE MELTS 123 permits carbonate-rich magmas to melt at geologically 1800 reasonable temperatures (Wyllie and Tuttle, 1960; Wyl- lie, 1989), and fluoride anions are inferred to have an equally large effect on liquidus phase relations (Gittins et aI., 1990; Jago and Gittins, 1991). g The anion solutions (Temkin model) in carbonate-rich I- CC+L melts are nearly ideal for all non polymerizing anions (Ta- ble 4): ". . . mixed anion-common cation fused salts of- ten are very nearly ideal solutions" (Kleppa and Julsrud, 1200 AH + CC 1980). Data are available from Ca-rich and some other binary systems on the interaction of carbonate anions o 0.2 0.4 0.6 0.8 1 with fluoride and other halide, hydroxide, oxide, perox- ide, and sulfate ions (Table 4). W values constrained here X(CaC03) include those for CO~--OH- (Fig. 2c) and CO~--F- (Fig. Fig. 3. Liquidus phase diagrams for CaCO,-CaSO.: phases 2d). Values of W in Table 4 for all anion solutions except are CC, calcite (CaCO,); AH, anhydrite (CaSO.), not including carbonate-orthophosphate are near zero, confirming their polymorphic transitions; and L, liquid. Solid lines are as pre- nearly ideal behavior. dicted by the regular-solution model (Eq. 7) and ideal mixing of Mixing of sulfate and carbonate in ionic melts appears CO~- and SO~- (Table 4) for a total pressure of -100 bars. Open essentially ideal. Flood et al. (1952) studied melts in the squares and dotted lines are experimental determinations of the system Na2C03-Na2SO.-C02 for electrochemical appli- liquidus and solidus by differential thermal analysis (Fuerstenau cations, and found that the melts behaved as ideal solu- et aI., 1981). Predicted and experimental positions of the calcite- saturated liquidus agree within error. The experimentally deter- tions. The ideality of carbonate-sulfate mixing extends to mined position of the anhydrite-saturated liquidus is not con- Ca-rich compositions, as the calcite-saturated liquidus sistent with the predicted liquidus and does not extrapolate to surfaceinCaC03-CaSO.at 1 bar (Fuerstenau et aI., 1981) the known melting temperature of anhydrite. Since Fuerstenau is consistent with Wcoj- -so~-= 0 kJ/moP, if t>.H?u,(calcite) et al. (1981) did not characterize their experiment products, it is = 31.5 kJ/mol (Fig. 3). possible that their liquidus surface represents growth of a mixed Mixing properties of carbonate and sulfide anions would anion solid that melts incongruently. be very useful in understanding redox states of natural systems and in some carbonatite-hosted ore deposits, but no definitive data are available. The liquidus location in CaC03-Ca(OH)2-CaS at 1 kbar by Helz and Wyllie (1979) cite) = 31.5 kJ/mol (Table 2), Wcoj-.ow = -2.3 kJ/moP cannot be interpreted uniquely here because WS2-.ow is (Table 4), and W PO~ - -ow ~ - 27 kJ/moP (calculated from not known. Their data on the binary join Ca(OH)2-CaS Table 4 of Tacker and Stormer, 1933). However, Biggar and estimates of the melting temperature and heat of fu- (1969) suggested that these experiments may be faulty sion for CaS suggest that WS2--ow is approximately + 15 and that the liquidus may lie at even higher temperatures, kJ/moP; this value is obviously suspect. The liquidus lo- in which case Wcoj- -PO~-would be even larger. Such a cation in CaC03-CaSO.-CaS determined at low pressure large positive Wcoj--PO~-implies that some carbonate-rich by Fuerstenau et al. (1981) cannot be interpreted unam- and phosphate-rich melts might be immiscible. In nature, biguously here because WS2--so~-is not known. However, phosphate-rich segregations (called phoscorite or cama- if WS2--so~-is taken as zero, WS2--coj- must be near + 100 forite) are common in some carbonatites (e.g., Eriksson, kJ/mol. 1989), and it has been suggested they form by carbonate- The behavior of phosphate may be far from ideal. The phosphate liquid immiscibility (Lapin, 1976). Continued mixing of carbonate and orthophosphate (P01-) anions experimentation will be required to understand the ther- was suggested to be nearly ideal by analogy with the mix- mochemistry of phosphate-carbonate mixing. ing of carbonate and sulfate anions (Table 4; Treiman, There are few quantitative data on the mixing behavior 1989). However, Tacker and Stormer (1993) have shown of carbonate and silicate anions in carbonate melts. The that solutions of molten calcium orthophosphate and other mixing of carbonate with orthosilicate (SiO:-) may be calcium salts (hydroxide, chloride, and fluoride) are not close to ideal (Treiman, 1989). However, the limited data ideal, with regular-solution interaction parameters more available (in the system CaO-Si02-C02-H20: Wyllie and negative than -20 kJ/moP. They suggested further that Haas, 1965) are difficult to interpret because the propor- carbonate-orthophosphate mixing might also be non- tions and speciations of H20 in the melts are not known. ideal. The data of Biggar (1969) appear to be the only Mixing of carbonate anions with more polymerized alu- liquidus determinations relevant to the mixing of alkaline minosilicate anions is far from ideal, as shown by the earth carbonates and phosphates. His experiments yield- immiscibility of carbonate and silicate melts. This liquid ed only a single bracket on the calcite-saturated liquidus immiscibility covers a wide range of synthetic and natural in a system containing orthophosphate. Taken at face val- compositions (e.g., Koster van Groos and Wyllie, 1966; ue, this liquidus bracket suggests Wcoj--po~- :::::+65 kJ/ Treiman and Essene, 1985; Kjarsgaard and Hamilton, moP, assuming a symmetrical regular solution, t>.H?u,(cal- 1989). Even in compositions without immiscibility, car- 124 TREIMAN: Ca-RICH CARBONATE MELTS bonate anions tend to form clusters that exclude poly- 1986). As calcite accepts little OH- or LaH in solid so- merized silicate anions (Mysen and Virgo, 1980). The lution, the calcite-saturated liquidus may be modeled as degree of this nonideality is a function of the composition of the silicate melt (Mysen and Virgo, 1980; Fine and Stolper, 1985) and remains to be characterized fully. - {[RT In(Xca2+,cat;on melt' X coj- ,an;onmelt)

Cation mixing + Wcoj--ow(l - Xcoj-,anionmelt)2] (1 - ~)} Within the accuracy limits of available data, cation / (I )2 mixing in Ca-rich carbonate melts is adequately de- - X a, 2+ ,cat;onmelt = Wa,2+.La3+ + M/\\'s(calcite) (13) scribed by the regular-solution mode1. The regular-solu- T tion parameters W in Table 4 are taken from the litera- 1-- ture and developed here (Figs. I, 2a-2b, 2e-2f, and 3). Tfus Regular-solution interaction parameters for CaH and following Equations 7 and 12. This equation is in the monovalent and divalent cations tend to be moderate format y = ax + b; when it is graphed in that manner, and negative, suggesting the association of unlike cations experimental brackets on the calcite-saturated liquidus in the melt. The only value for a trivalent cation, LaH, ought to permit it to be a straight line with a slope of is large and positive. WCa2+_LaHand an intercept of M/~us(calcite) = 31.5 :t 1 Alkali cations. The interaction parameters W for Ca- kllmo1. The limited data are consistent with regular-so- Na and Ca-K carbonate systems at high pressure (Table lution behavior and Wa,2+_La3+= +80 :t 50 kllmoP (Fig. 4) were derived in concert with M/~us(calcite) from data 2f, Table 4). of Cooper et a1. (1975) in Figure 2a and 2b. The Wa,2+.K+value for I bar are significantly higher (Table 4, Solution of H20 based on Hood et al., 1949, and F0rland, 1955); the source H20 is important, both as a flux to permit melting of of the discrepancy is unknown. A CaH-Li+ interaction carbonates at geologically reasonable conditions (Wyllie parameter is available in the literature (Lumsden, 1966). and Tuttle, 1960) and as a constituent of the vapors as- The WNa+.K+and WNa+.L;+of Table 4 are averages from sociated with carbonatites (e.g., Rankin, 1975; Nesbitt the data of Andersen and Kleppa (1976); their precise and Kelly, 1977; McKie, 1966; LeBas, 1977; Rubie and calorimetry showed that both values are slight functions Gunter, 1983). H20 is problematic within an ionic melt of composition across the respective joins. model as it is not an ionic liquid. Further, the speciation Alkaline earth cations. Very few data are available for of H20 in ionic solutions may be affected by cation com- the estimation of solution parameters for alkaline earths plexation, formation of mixed anions (like bicarbonate), (besides Ca) in carbonate melts. The value for Wa,2+_&2+ and changes in intrinsic variables (like acidity and fa,). was defined in Figure 1 during determination of In addition, H20 does not behave as OH does in carbon- M/~us(calcite). ate-rich ionic melts; the calcite-saturated liquidus in To estimate the Winteraction parameter for the CaH- CaC03-Ca(OH)2 is effectively straight and not inflected, MgH cation solution in carbonate melts, Equation 6 and whereas the calcite-saturated liquidus in CaC03-H20 is the activity-composition model of Anovitz and Essene strongly curved and inflected, concave up at high CaC03 (1987) can be applied to the liquidus phase equilibria of contents, and concave down at lower CaC03 contents Byrnes and Wyllie (1981). Figure 2e shows that liquidus, (Figs. 5 and 6, respectively, of Wyllie and Tuttle, 1960). with the range of permissible liquidus locations forced Even so, the solution of H20 in Ca-carbonate melts through M/~us(calcite)= 31.5 :t 1 kllmol on the vertical can be described with a simple regular-solution model, axis. The permissible liquidus locations correspond to on the basis of the liquidus surface in the system CaC03- W = -40 :t 30 kllmo1. This result must be used with H20 (Wyllie and Tuttle, 1960, temperatures corrected by caution because the liquidus of Byrnes and Wyllie (1981) - 32 K per Gittins and Tuttle, 1964). I will assume that was determined at 10 kbar; the activity-composition H20 ionizes completely into hydroxide and hydronium model has been extrapolated somewhat beyond its known ions on solution in a carbonate melt: range of applicability, and the validity of the activity 2H20 ;=' H30+ + OH- (14) model is in question (McSwiggen, 1993). As a further caution, a similar analysis performed on the 27-kbar li- and so affects both the cation and anion solutions direct- quidus in CaC03-MgC03 (Irving and Wyllie, 1975) is ly. The experiments of Wyllie and Tuttle (1960) were consistent with a regular solution W of approximately done with excess vapor in most cases; because the com- zero. Obviously, data in this system are too sparse for position of the vapor is unconstrained, I must assume firm conclusions, but changes in melt structures are pos- further that the mass of vapor was insignificant compared sible. with that of the solid (i.e., that the mass of H20 input to Other cations. Interpretable liquidus data for other cat- the charge effectively represents the mass of H20 in the ions in carbonate melts are limited to lanthanum in the liquid). With this speciation model, Equations 5 and 7 system CaC03 -Ca(OH)2 -La(OH)3 (Jones and Wyllie, can be combined and reduced to yield TREIMAN: Ca-RICH CARBONATE MELTS 125

vapor phase is fugitive, its composition is elusive. In rare - {[RT In(XeaH,cation melt' Xeoj-,anion melt) cases, its composition can be constrained by the mineral assemblage of the carbonatite (Treiman and Essene, 1984), + Wcoj-.ow(1 - XeOhnionmel.)2] (1- but not without dispute (Gittins et aI., 1990, 1992; Trei- / £)} man and Essene, 1992). A solution model for carbonate-rich melts provides a (1-XH ea ,cation melt)2 = WeaH.H30+ + .:lH?u,(calcite) (15) link between carbonatite magmas and vapors. From the T composition of a carbonatite magma (e.g., its halogen content), a solution model permits calculation of some compositional constraints on the vapor in equilibrium comparable with Equation 13 above. XeaH,cationmeltand with the magma. Alternatively, given the composition of Xco2 anionmelthave the same numerical values, because they a vapor phase (e.g., its H20 content), one can constrain both ~rise from disproportionation of H20. Constraints the composition of the magma. on the calcite-saturated liquidus in CaC03-H20 from F: Experimental calibration. Carbonatite magmas can Wyllie and Tuttle (1960) are recast in Figure 2g in this contain significant F. Most carbonatites contain fluora- form, with the term in brackets on the ordinate and the patite and some carry fluorite (CaF2); many other F-bear- term attached to Wea2+.H30+on the abscissa. Weoj-.ow is ing minerals may be present, e.g., pyrochlore, fluocerite from Table 4. For the model to be correct, the liquidus [(Ce,La)F3], bastnaesite [(Ce,La)(C03)F], and amphiboles surface must be representable as a straight line, with an (Hogarth, 1989). The natrocarbonatites of Oldoinyo ordinate intercept of .:lH ?u,(calcite) = 31.5 ::!::1 kllmol Lengai, Tanzania, also contain F-bearing nyerereite and a slope of Wca2+.H30+'Figure 2g shows that these cri- [(Na,K)2Ca(C03)2J, and gregoryite [(Na2,Ca)C03] (Peter- teria are met for the ten experiments at lower abscissa son, 1990). Many fenites associated with carbonatites values, which have <40 wt% H20 (equivalent to XH20 ::S contain fluorite, some in economic proportions. In ad- 0.9). To match these experiments, WCa2+.H30+= 36 ::!::2 dition, theoretical interest has recently focused on F be- kllmoF, which is high but not unreasonable compared cause of the experimental observation that F is similar to with other W values (Table 4). H20 as a flux to permit melting of carbonates at geolog- It remains unclear how closely this model cleaves to ically reasonable temperatures (Gittins et aI., 1990; Jago reality. Instead of a straight line in Figure 2g, one could and Gittins, 1991). also fit a smooth curve through all the liquidus brackets, The behavior of F in carbonatite magma-vapor sys- inconsistent with regular solution behavior. It is entirely tems can be modeled by the reaction possible that effects offormation of complexes, formation F2 + COj- 2F- + C02 + 1/202 (16) of mixed anions, or violation of other assumptions are "" vapor vapor vapor hidden in the correlation of Figure 2g. If accurate, how- melt melt ever, it suggests that dissolved H20 at high temperatures which describes the equilibrium distribution of F be- in ionizing environments behaves as the ionic compound tween carbonate melt and vapor. To calibrate the distri- hydronium hydroxide, H30+OH -. These hypotheses about bution of F, one must derive the standard Gibbs free H20 in carbonate magmas should be readily testable. energy of Reaction 16 in the temperature range of interest (ignoring the effects of pressure on melt components), ApPLICATIONS Beyond the value of a theoretical understanding of car- l1G~=_RTln feo2,vapo<'f32,vaPO', a~-,melt (17) f bonatite magmas, this regular-solution model can help ( F2,vapor acoJ- .melt) answer real geological questions involving carbonate melts. Here, two types of questions are considered: the where each a is the activity of a melt component and relationships between the compositions of carbonatite each f is the fugacity of a vapor species. If all the fugac- magmas and their vapors and the prediction of liquidus ities and activities can be determined at a given temper- equilibria. ature, the free energy of reaction is readily calculated. The Gibbs energy of Equation 17 can be calculated Carbonatite magma and vapor along the calcite + fluorite saturated liquidus in the sys- A fluid or vapor phase is commonly associated with tem CaF2-Ca(OH)2-CaC03' as drawn by Gittins and Tut- carbonatites, as shown by their fluid inclusions (e.g., Ran- tle (1964) from their l-kbar experimental results (their kin, 1975; Nesbitt and Kelly, 1977) and surrounding met- Fig. 4). The ratio of gas species fugacities in Equation 17 asomatic rocks (fenites: e.g., McKie, 1966; LeBas, 1977; is buffered by the presence of solid calcite and fluorite, Rubie and Gunter, 1983). An understanding of the vapor F2 + CaC03 CaF2 + C02 + 1/202, (18) phase is important for understanding the application of "" vapor solid vapor vapor experimental phase equilibria to natural carbonatites, the solid origins of carbonatite metasomatism, and the effects of Activities of melt species come directly from the analyzed volatile loss or gain on magma composition. Because the or inferred compositions of the melts, and activity coef- 126 TREIMAN: Ca-RICH CARBONATE MELTS

ficients come from the melt solution model. With Reaction 21 can be calculated for compositions along the WOH~_F~_eoj~and WF~_eoj~as zero (Table 4), activity co- calcite + portlandite [Ca(OH)2]-saturated liquidus in the efficients are given as system CaF2-Ca(OH)2-CaC03 (Gittins and Tuttle, 1964), the RT In 'Ycoj~= Wcoj~ same data set used for calibration of the carbonate + fluo- -ow (Xbw + XF~XOw) rine exchange above. The ratio of gas species fugacities in (19a) Equation 22 is buffered by the presence of solid calcite and portlandite and the equilibrium and H20 + CaC03 Ca(OH)2 + C02' (23) "" vapor solid solid vapor - Weoj~_ow(Xcoj~Xow) (19b) Activities of melt species are calculated from analyzed or from Equation 5. To obtain AG~ in Equation 17, the gas inferred melt compositions and regular-solution activity fugacity ratio is calculated using thermochemical data coefficients. With WOWF-eoj- and WF-_COj-as zero (Ta- from Chase et al. (1974) and Robie et al. (1979), and melt ble 4), activity coefficients are given as component activities are calculated from measured or in- RT In 'Yeoj- = Weoj-_OW(XbH~ + XF-XOW) ferred melt compositions (Gittins and Tuttle, 1964) and activity coefficients (Eq. 19), with W values from Table (24a) 4. Taking seven points from the liquidus of Gittins and and Tuttle (1964) from 1153-848 K yielded a regressed free energy equation of

AG~ = -412.0 (:t2.8) - 0.091 (:to.003)T(kJ/mol) (20) + Weoj-_OH-(Xl:oj~ + XCOj-XF-) (24b) with r2 = 0.995, where the uncertainties are the 20-stan- from Equation 5. To obtain AG~ in Equation 22, the gas dard errors of regression. If the regular-solution model fugacity ratio is calculated using thermochemical data from here is accurate, this calibration should be valid for all Chase et al. (1974) and Robie et al. (1979), and melt com- Ca-rich carbonate-rich melts, whether in synthetic or ponent activities are calculated from measured melt com- complex natural systems. The free energies of reaction positions (Gittins and Tuttle, 1964). Activity coefficients are negative and imply that Reaction 16 favors strongly (Eqs. 24a and 24b) using W values are from Table 4. The the production of fluoride ions and carbon dioxide gas. calcite + portlandite liquidus spans only a small tempera- Thus, the vast proportion ofF in Ca-rich carbonatite melt- ture interval, 926-848 K (Gittins and Tuttle, 1964). Taking vapor systems is in the melt, and the vapor contains little compositions at the end points and one intermediate point F (Treiman and Essene, 1992). yields a regressed free energy equation of Hydroxide: Experimental calibration. Carbonatite AG~ = 79.2 (:t3.l) magmas may contain significant H20, as suggested by - 0.038 (:to.003)T(kJ/mol) (25) Wyllie and Tuttle (1960) to explain how carbonate-rich with r2 = 0.995, where the uncertainties are only the 20- melts might form at reasonable geological temperatures. standard errors of regression. This calibration is probably Since that time, the importance of H20 as a flux in car- more uncertain than the regression errors would indicate, bonatite magmas has gained wide acceptance, although as potential experimental errors have not been included, some experiments suggest that F may be equally effective and the temperature range of calibration is small. as a flux (Gittins et aI., 1990; Jago and Gittins, 1991). However, the positive free energy for Reaction 21 im- The exchange of H20 and OH between carbonatite plies that H20 is preferentially partitioned into the vapor magma and vapor systems can be modeled, in the same phase and that a H20-rich vapor would coexist with a manner as F above, by the reaction relatively H20-poor carbonatite magma. If the regular solution model here is accurate, this calibration should H20 + COj- 20H- + C02' (21) "" be valid for all Ca-rich carbonate-rich melts, whether in vapor melt melt vapor synthetic or complex natural systems. Unlike the case for F, there is no dependence here on An example from nature. Within the Oka carbonatite 10,. Calibration of this equilibrium requires derivation complex, Quebec, Canada, is the Husereau carbonatite of its standard Gibbs free energy in the temperature range dike. That unique dike contains such a restrictive mineral of interest (if one ignores the effects of pressure on melt assemblage that Treiman and Essene (1984) were able to components), calculate the composition of its F-poor vapor phase and infer by difference that the vapor was mostly H20. Gittins _RTln /c02.vaPO,. abH-,melt (22) (1989) and Gittins et al. (1990) criticized their conclu- AG~ = f ( H20,vapor acoj- ,melt) sion, asserting a much more prominent role for F. One where each a is the activity of a melt component and each source of disagreement was the abundance of F in a car- I is the fugacity of a vapor species. The Gibbs energies of bonatite magma compared with its vapor (Treiman and TREIMAN: Ca-RICH CARBONATE MELTS 127

Essene, 1992; Gittins et aI., 1992). Application of the is correct, the only way for the melt to have had more F regular-solution model for carbonate-rich magmas can is for it to have been present in complexes, like molecular resolve this issue. HF, SiF~-, etc. There is no evidence for or against the To calculate the concentrations of fluoride and OH in presence of such fluoride complexes in the Husereau dike. the Husereau carbonatite magma, one can apply Reac- tions 16 and 21 using the calibration in Equations 20 and Carbonate magma on the surface of Venus 25. Treiman and Essene (1984) inferred that the Huser- Carbonate-rich magmas may not be unique to the Earth, eau carbonatite magma equilibrated with calcite, apatite, but may be of broader planetary importance, possibly oxides, and other phases at 913 K and 1 kbar of total occurring on Mars (Longhi, 1991) and Venus. The tem- pressure. The gas phase was calculated to contain 110 perature and pressure of the Venus surface could be con- bars COz; by difference, HzO was inferred to account for ducive to carbonate volcanism (Sill, 1984), and Magellan 882 bars of the total vapor. radar images oflong sinuous channels on the Venus sur- To calculate the abundance ofOH in the Husereau dike face (presumably H20-free) have given that idea new life magma, one can use the calculated composition of the (Baker et aI., 1992; Komatsu et aI., 1992; Kargel et aI., Husereau gas phase in Reaction 21. Applying the cali- 1993). The possibility of carbonate magmas on Venus bration of Equation 25 yields must be considered in the context of its surface condi- tions: a global mean temperature of 740 K and a mean a6w,melt = 2.8 x 10-3. (26) atmospheric pressure of 95 bars (Seiff, 1983); the atmo- acoj- ,melt sphere consists of96.5% COz, 3.5% Nz, and -150 ppm If one assumes the magma contained no abundant ion SOz (Fegley et aI., 1992). The Venera and VEGA chem- species besides carbonate and hydroxide, the regular-so- ical analyses of the Venus surface are consistent with lution activity model (Eq. 5, Table 4), allows Equation tholeiitic or alkaline basalt, with two analyses suggesting 26 to be satisfied by Xow ::::0.07 and Xcoj- ::::0.93. This felsic or peralkaline compositions (Barsukov, 1992). ratio is consistent with the absence of portlandite or other Melting of calcite + anhydrite. Near-surface rocks on hydroxides as magmatic phases (viz., Gittins and Tuttle, Venus are inferred to contain both calcite (CaC03) and 1964) and affirms that carbonatite magmas were much anhydrite (CaS04) as weathering products of igneous poorer in HzO than their equilibrium vapors (here 882 minerals (Fegley and Prinn, 1989; Fegley et aI., 1992); bars HzO and 110 bars COz). There is no mineralogic alkali carbonates and sulfates are not predicted to be sig- evidence for this abundance of magmatic OH in the Hu- nificant in the weathering assemblage. Might calcite + sereau dike, and one may speculate that it was expelled anhydrite be molten under Venus surface conditions, or during solidification as a HzO-rich vapor phase. Some of might calcite + anhydrite melt be formed under reason- this HzO-rich vapor might have been consumed in the able geological conditions (e.g., volcanic or impact heat- pervasive conversion of magmatic peric1ase in the dike ing)? On Venus, is it also possible that carbonate-sulfate to brucite (Treiman and Essene, 1984), but most of it magmas might be generated by liquid immiscibility from must have left the dike completely. volatile-rich basaltic magma (Kargel et aI., 1993)? An- Fugacities of F-bearing gas species were calculated as- swering these questions requires knowing the liquidus suming F-OH exchange equilibria between apatite in the surface in the system CaC03-CaS04, which can be cal- dike rock and vapor. From the inferred f 0, of the QFM culated using the regular solution model developed here. buffer (10-186 bars), Treiman and Essene calculated that Melting relations in the system CaC03-CaS04 can be the gas fugacity ofF2 was 10-43.9bars. From the temper- modeled with the thermochemical parameters derived and ature, gas fugacities, and Equation 20, one can now cal- compiled here and compared with the experimentally de- culate the ratio of specie activities in the Husereau car- termined phase equilibria of Fuerstenau et ai. (1981). To bonatite melt as calculate the liquidus surface, one requires heats of melt- ing, temperatures of melting, solution models for calcite a~-,melt and anhydrite, and a solution model for their melts. Melt- = 4.8 X 10-9. (27) acoj- ,melt ing data for calcite are calculated for 100 bars from data in Tables 1 and 2. The melt solution of COj- and SO~- If the activity of carbonate in the melt was approximately is essentially ideal (Table 4). It is assumed that calcite unity, the activity of fluoride was 7 x 10-5. The anion and anhydrite are pure phases, that the Ml?u,(anhydrite) fraction of fluoride in the melt can be calculated knowing is 28 kJ/mol, for a I-bar melting temperature of 1723 K the anion fraction of carbonate in the melt, 0.93, as cal- (Robie et aI., 1979), and that this Ml?u,(anhydrite) is rel- culated above. This value gives a fluoride activity of 6.7 evant to 100 bars pressure. At 1468 K, anhydrite inverts x 10-5 (Eq. 27), a fluoride activity coefficient of 1.06 (Eq. to a high-temperature phase, which melts at 1737 :t 4 K 19b), and so a fluoride anion fraction of 6.3 x 10-5. (Rowe et aI., 1965). The lower temperature chosen for Clearly, this is not much fluoride in the melt, and it is this calculation is estimated for metastable melting ofthe consistent with the lack ofF-bearing minerals other than lower temperature anhydrite polymorph, which would be apatite in the dike. If the thermochemical analysis so far stable at the Venus surface. 128 TREIMAN: Ca-RICH CARBONATE MELTS

From these data and Equation 7, one can map the li- fluorine-carbonate exchange between carbonate melts and quidus surface in CaC03-CaSO., ignoring the polymor- vapor in Reaction 16 can be applied here, assuming that phic transition in solid CaSO. (Fig. 3). The calcite-satu- its free energy calibration (Eq. 20) can be extrapolated to rated and anhydrite-saturated liquidus curves are 740 K from its minimum calibration temperature of848 calculated independently. The intersection of the curves K. With thermochemical data from Robie et aI. (1979) (the stable coexistence of calcite, anhydrite, and melt) at and the near-surface Venus atmospheric abundances of 1250 K is the eutectic point, the lowest temperature at CO = ~ 20 ppm and H20 = 20 ppm, the reactions which melt can exist in CaC03-CaSO. The eutectic tem- 2HF + C02'" F2 + H20 + CO (28a) perature is 335 K below the melting temperature of cal- cite alone and 473 K below the melting temperature of and anhydrite alone. Predicted melting relations in CaC03-CaSO. are in (28b) partial accord with the experimental results of Fuerst enau imply that fF2 = 1.1 X 10-.7 bars and f02 = 3.5 X 10-22 et aI. (1981), as shown in Figure 3. Fuerstenau et aI. in- bars, within the magnetite stability field in Fe-O (Fegley ferred phase relations by differential thermal analysis on et aI., 1992). From Reaction 16 and its calibrations in sealed charges. They had little control on pressure, which Equations 17 and 20, one can now calculate the activity was inferred to vary between 15 and 35 bars. Fuerstenau of fluoride in a carbonate-rich melt in equilibrium with et aI. did not analyze their experiment products, and so the Venus atmosphere. With the derivation of Equation their identification of solid phases is completely by infer- 21, fluoride and carbonate activities are related as ence. aF-,melt2 For the CaC03-rich calcite-saturated limb of the liqui- = 4.4 x 10-5. (29) dus (Fig. 3), there is excellent agreement between the pre- aCOJ- ,melt dictions here and the experiments of Fuerstenau et aI. This implies that fluoride activity in a carbonate melt is, (1981). However, prediction and experiment do not agree at most, 7 x 10-3. The fluoride activity coefficient is for the CaSO.-saturated limb of the liquidus. The exper- unity in the absence ofOH (Eq. 19b), implying the same imental determination of this limb, inferred to be anhy- value for the anion concentration of fluoride; this abun- drite + liquid (Fuerstenau et al., 1981), is not consistent dance corresponds to a F abundance of approximately with Trus(anhydrite)= 1723 K (viz., Robie et aI., 1979) 0.1 wt% (calculated for a melt containing only CaH, but rather with Trus::::::1550 K (Fig. 3). Using the latter COj-, and F-). Thus, F from the atmosphere would be Trusimplies (by means ofEq. 6) that the CaSO.-rich solid strongly concentrated in carbonate-rich melts on the Ve- ~ 28 kJ/mol, which is reason- has t!Jl rus per cation of nus surface. However, the proportion of F would be so able for an ionic compound. The source of these dis- low as to have little effect in stabilizing those melts (Jago crepancies is unclear; experimental error is possible, and Gittins, 1991). but a polymorphic transition in solid CaSO. is unlikely (Rowe et aI., 1965). To me, it seems likely that the Caveats liquidus phase detected by Fuerstenau et al. (1981) is Caution must be exercised in applying this regular solu- a previously unreported intermediate mixed-anion com- tion model of carbonate magmas, particularly in the preci- pound, for instance Ca3(SO.MC03), that melts incongru- sion of its implications. Many parameters in the model are ently to anhydrite + liquid. poorly known, including heats of melting for magnesium Applying either predicted or experimental liquidus sur- carbonate and W interaction parameters for all cations with faces to Venus, one can see that mixtures of calcite + MgH. The Temkin regular-solution model is only a first anhydrite will not melt at normal Venus surface temper- approximation to real behavior (Andersen and Kleppa, 1976) atures, 660-760 K (Seiff, 1983). To melt a mixture of and can undoubtedly be improved given additional high- calcite + anhydrite on the Venus surface would require quality data. Finally, there remain the possibilities of mul- temperatures at least 500 K above ambient, which could tiple melt structures in Ca-rich carbonate magmas and of easily be supplied by impact heating or basaltic volca- previously unreported liquidus phases in unexplored sys- nism (solidus temperatures near 1300-1400 K). Addi- tems. It is hoped that this model can serve as a starting tional components, like alkalis or halides, would decrease point for a deeper understanding of the thermochemical the melting temperature, but the availability of these behavior of carbonate-rich magmas. components is not clear. Similarly, the eutectic temper- ature calculated here does not invalidate the speculation ACKNOWLEDGMENTS that carbonate-sulfate magma could exsolve from basaltic This work was started ten years ago during my doctoral studies at the magmas (Kargel et aI., 1993). University of Michigan, under the tutelage of EoJ. Essene and W, Kelly, F in Venus carbonatites. The near-surface atmosphere and with the assistance ofL.M, Anovitz and J,W, Valley. More recently, I am grateful to B. Jago and J. Gittins for reviving my interest in carbon- of Venus contains approximately 5 ppb of HF (Fegley et atite thermochemistry and to B. Fegley, Jr. and J. Kargel for collabora- aI., 1992), and one may inquire whether this much F in tions on carbonate-sulfate magmas on Venus. The calculations here would the atmosphere would significantly affect the composi- have been impossible without the careful experiments reported over many tions of ionic melts on the Venus surface. The model for years by P.J. Wyllie, J. Gittins, and their colleagues. This paper has ben- TREIMAN: Ca-RICH CARBONATE MELTS 129 efited from careful and constructive critiques by E.1. Essene, C.T. Herz- mineralogy: Observations and theoretical constraints. Proceedings of berg, J.H. Jones, G. Ryder, B. Schuraytz, B.J. Wood, and an anonymous Lunar and Planetary Science, 22, 3-20. reviewer. Lunar and Planetary Institute contribution no. 840. Fine, G., and Stolper, E. (1985) The speciation of carbon dioxide in so- dium aluminosilicate glasses. Contributions to Mineralogy and Petrol- ogy, 91, 105-121. REFERENCES CITED flood, H., Ferland, T., and Roald, B. (1949) The equilibrium CaCO"m.", Andersen, B.K., and Kleppa, 0.1. (1976) Enthalpies of mixing in binary = CaCO,s, + CO,. The activity coefficients of calcium carbonate. Jour- liquid alkali carbonate solutions. Acta Chemica Scandinavica, A30, nal of the American Chemical Society, 71, 572-575. 751-758. flood, H., Ferland, T., and Motzfeld, K. (1952) On the oxygen electrode Andersen, 0.1., and Lindsley, D.H. (1981) A valid Margules formulation in molten salts. Acta Chemica Scandinavica, 6, 257-269. for an asymmetric ternary solution: Revision of the olivine-ilmenite Ferland, T. (1955) An investigation of the activity of calcium carbonate thermometer, with applications. Geochimica et Cosmochimica Acta, in mixtures off used salts. Journal of Physical Chemistry, 59,152-156. 45, 847-853. Fuerstenau, M.C., Shen, C.M., and Palmer, B.R (1981) Liquidus tem- Andersen, T. (1986) Magmatic fluids in the Fen carbonatite complex, S.E. peratures in the CaCO,-Ca(OH),-CaO and CaC03-CaSO,-CaS ternary Norway: Evidence of mid-crustal fractionation from solid and fluid systems: II. Industrial and Engineering Chemical Process Design and inclusions in apatite. Contributions to Mineralogy and Petrology, 93, Development, 20, 443-445. 491-503. Ghiorso, M.S. (1987) Modeling magmatic systems: Thermodynamic re- Anovitz, L.M., and Essene, E.1. (1987) Phase equilibria in the system lations. In Mineralogical Society of America Reviews in Mineralogy, CaC03-MgCO,-FeC03. Journal of Petrology, 28, 389-414. 17, 443-465. Bailey, O.K. (1993) Carbonate magmas. Journal of the Geological Society Ghiorso, M.S., Carmichael, I.S.E., Rivers, M.L., and Sack, R.O. (1983) of London, 150,637-651. The Gibbs free energy of mixing in natural silicate liquids: An expand- Bakcr, E.H. (1962) Calcium oxide-carbon dioxide system in the pressure ed regular solution approximation for the calculation of magmatic in- range 1-300 atmospheres. Chemical Society Journal, 1962,464-470. tensive variables. Contributions to Mineralogy and Petrology, 84, 107- Baker, R.V., Komatsu, G., Parker, T.J., Gulick, V.c., Kargel, J.S., and 145. Lewis, J.S. (1992) Channels and valleys on Venus: Preliminary analysis Gittins, J. (1989) The origin and evolution of carbonatite magmas. In K. of Magellan data. Journal of Geophysical Research, 97, 13421-\3444. Bell, Ed., Carbonatites: Genesis and evolution, p. 580-600. Unwin- Barsukov, V.L. (1992) Venusian igneous rocks. In V.L. Barsukov, AT. Hyman, London. Basilevsky, V.P. Volkov, and V.N. Zharkov, Eds., Venus geology, geo- Gittins, J., and Tuttle, O.F. (1964) The system CaF,-Ca(OH),-CaCO,. chemistry, and geophysics, p. 165-176. University of Arizona Press, American Journal of Science, 262, 66-75. Tucson. Gittins, J., Beckett, M.F., and Jago, B.c. (1990) Composition of the fluid Berman, RG., and Brown, T.H. (1987) Development of models for mul- phase accompanying carbonatite magma: A critical examination. ticomponent melts: Analysis of synthetic systems. In Mineralogical So- American Mineralogist, 75, 1106-1109. ciety of America Reviews in Mineralogy, 17,405-442. - (I 992) Composition of the fluid phase accompanying carbonatite Biggar, G.M. (1969) Phase relationships in the join Ca(OH),-CaCO,- magma: A critical examination-Reply. American Mineralogist, 77, Ca3(PO,),-H,0 at 1000 bars. Mineralogical Magazine, 37, 75-82. 666-667. Blander, M. (1964) Thermodynamic properties of molten salt solutions. Helffrich, G., and Wood, B. (1989) Subregular model for multicomponent In M. Blander, Ed., Molten salt chemistry, p. 127-237. Interscience, systems. American Mineralogist, 74, 1016-1022. New York. Helz, G.R., and Wyllie, P.J. (1979) Liquidus relationships in the system Blander, M., and Topol, L.E. (1966) The topology of phase diagrams of CaCO,-Ca(OH),-CaS and the solubility of sulfur in carbonatite mag- reciprocal molten salt systems. Inorganic Chemistry, 5, 1641-1645. mas. Geochimica et Cosmochimica Acta, 43, 259-265. Bradley, RS. (1962) Thermodynamic calculations on phase equilibria Hogarth, D.O. (1989) Pyrochlore, apatite, and amphibole: Distinctive involving fused salts: I. General theory and application to equlibria minerals in carbonatite. In K. Bell, Ed., Carbonatites: Genesis and involving calcium carbonate at high pressure. American Journal ofSci- evolution, p. 105-148. Unwin-Hyman, London. ence, 260, 374-382. Huang, W.-L., and Wyllie, P.J. (1976) Melting relationships in the system Byrnes, A.P., and Wyllie, P.J. (1981) Subsolidus and melting relations for CaO-CO, and MgO-CO, to 33 kilobars. Geochimica et Cosmochimica the join CaC03-MgC03 at 10 kbar. Geochimica et Cosmochimica Acta, Acta, 40, 129-\32. 45, 321-328. Irving, AJ., and Wyllie, P.1. (1973) Melting relationships in CaO-CO, Carlson, W.O. (1983) The polymorphs ofCaC03 and the aragonite-calcite and MgO-CO, to 36 kilobars with comments on CO, in the mantle. transformation. In Mineralogical Society of America Reviews in Min- Earth and Planetary Science Letters, 20, 220-225. eralogy, II, 191-225. - (1975) Solidus and melting relationships for calcite, magnesite, Chase, M.W., Curnutt, J.L., Hu, A.T., Prophet, H., Syverud, A.N., and and the join CaCO,-MgCO, to 36 kilobars. Geochimica et Cosmochim- Walker, L.c. (1974) JANAF thermochemical tables: 1974 Supplement. ica Acta, 39, 35-53. Journal of Physical and Chemical Reference Data, 3, 311-480. Jago, B.c., and Gittins, J. (1991) The role of fluorine in carbonatite mag- Cooper, AF., Gittins, J., and Tuttle, O.F. (1975) The system Na,C03- ma evolution. Nature, 349, 56-58. K,CO,-CaC03 at I kilobar and its significance in carbonatite petrogen- Janz, G.1., Allen, D.B., Bansal, N.P., Murphy, RM., and Tompkins, R.P.T. esis. American Journal of Science, 275, 534-560. (1979) Physical properties data compilations relevant to energy storage: Dawson, J.B., Garson, M.S., and Roberts, B. (1987) Altered former al- II. Molten salts: Data on single and multi-component systems. Nation- kalic carbonatite lavas from Oldoinyo Lengai, Tanzania: Inferences for al Standards Reference Data Series. NSRDS-NBS 61, Part II. U.S. calcite carbonatite lavas. Geology, 15, 756-768. Government Printing Office, Washington, DC. Dawson, J.B., Pinkerton, H., Norton, G.E., and Pyle, D.M. (1990) Phys- Jones, A.P., and Wyllie, P.J. (1986) Solubility ofrare earth elements in icochemical properties of alkali carbonatite lavas: Data from the 1988 carbonatite magmas, indicated by the liquidus surface in CaCO,- eruption ofOldoinyo Lengai, Tanzania. Geology, 18,260-263. Ca(OH),-La(OH)3' Applied Geochemistry, 1,95-102. DeKock, C.W. (1986) Thermodynamic properties of selected metal sul- Kargel, J.S., Komatsu, G., Baker, R.V., and Strom, R.G. (1993) The vol- fates and their hydrates. Information Circular 9081: U.S. Bureau of canology of Venera and VEGA landing sites and the geochemistry of Mines, 59 p. U.S. Government Printing Office, Washington, DC. Venus. Icarus, 103, 253-275. Eriksson, S.c. (1989) Phaloborwa: A saga of magmatism, metasomatism, Keller, J., and Krafft, M. (1990) Effusive natrocarbonatite activity of 01- and miscibility. In K. Bell, Ed., Carbonatites: Genesis and evolution, doinyo Lengai, June 1988. Bulletin of Volcanology, 52, 629-645. p. 221-254. Unwin-Hyman, Boston. Kjarsgaard, B.A, and Hamilton, D.J. (1989) The genesis of carbonatites Fegley, B., Jr., and Prinn, R.G. (1989) Estimation ofthe rate of volcanism by immiscibility. In K. Bell, Ed., Carbonatites: Genesis and evolution, on Venus from reaction rate measurements. Nature, 337, 55-58. p. 388-404. Unwin-Hyman, London. Fegley, B., Jr., Treiman, A.H., and Sharpton, V.L. (1992) Venus surface Klement, W., Jr., and Cohen, L.H. (1975) Solid-solid and solid-liquid 130 TREIMAN: Ca-RICH CARBONATE MELTS

transitions in K,CO" Na,CO" and Li,CO,: Investigations to >5 khar rium diagram of the system Caco, + Na,S04 .. CaS04 + Na,CO,. by differential thermal analysis; thermodynamics and structural cor- Russian Journal ofInorganic Chemistry, 20, 1742-1743. relations. Berichte der Bunsengesellschaft, 79, 327-334. Ragone, S.E., Datta, R.K., Roy, D.M., and Tuttle, O.F. (1966) The system KIeppa, 0.1. (1977) Thermodynamic properties of molten salt solutions. potassium carbonate-magnesium carbonate. Journal of Chemical Phys- In D.G. Fraser, Ed., Thermodynamics in geology, p. 279-299. Reidel, ics, 70, 2260-2261. Boston. Rankin, A.H. (1975) Fluid inclusion studies in apatites from carbonatites -(1981) Thermodynamics of simple molten salt mixtures. In R.c. of the Wasaki area of western Kenya. Lithos, 8, 123-136. Newton, A. Navrotsky, and RJ. Wood, Eds., Thermodynamics of min- Robie, R.A., Hemingway, RS., and Fisher, J.R. (1979) Thermodynamic erals and melts, p. 179-188. Springer, New York. properties of minerals and related substances at 298.15 K and I bar KIeppa, O.J., and Julsrud, S. (1980) Thermodynamics of charge-unsym- (10' pascals) pressure and at higher temperatures, corrected, 456 p. U.S. metrical anion mixtures: I. The liquid systems AF-A,S04' Acta Chem- Geological Survey Bulletin 1452. U.S. Government Printing Office, ica Scandinavica, A34, 655-665. Washington, DC. Komatsu, G., Kargel, J.S., and Baker, V.R. (1992) Canali-type channels Rowe, J.J., Morey, G.W., and Hansen, I.D. (1965) The binary system on Venus: Some genetic constraints. Geophysics Research Letters, 19, K,S04-CaS04' Journal ofInorganic and Nuclear Chemistry, 27,53-58. 11415-11418. Rubie, D.C., and Gunter, W.D. (1983) The role of speciation in alkaline Koster van Groos, A.F., and Wyllie, P.1. (1966) Liquid immiscibility in igneous fluids during fenite metasomatism. Contributions to Mineral- the system Na,o-AI,O,-SiO,-CO, at pressures to I kilobar. American ogy and Petrology, 82, 165-175. Journal of Science, 264, 234-255. Seiff, A (1983) Thermal structure of the atmosphere of Venus. In D.M. Kresten, P., and Morogan, V. (1986) Fenitization at the Fen complex, Hunten et aI., Eds., Venus, p. 215-279. University of Arizona Press, southern Norway. Lithos, 19,27-42. Tucson. Kuellmer, F.1., Visocky, A.P., and Tuttle, O.F. (1966) Preliminary survey Selman, J.R., and Maru, H.c. (1981) Physical chemistry and electrochem- of the system barite-calcite-fluorite at 500 bars. In O.F. Tuttle and J. istry of alkali carbonate melts, with special reference to molten-car- Gittins, Eds., Carbonatites, p. 353-364. Wiley, London. bonate fuel cells. In G. Mamantov and 1. Braunstein, Eds., Advances Lapin, AV. (1976) Geological examples of limited miscibility in ore- in molten salt chemistry, vol. 4, p. 159-389. Plenum, New York. silicate-carbonate melts. D0klady Akademie Nauk SSSR, 231, 163- Sill, G.T. (1984) Possible carbonatite volcanism on Venus (abs.). Bulletin 166. ofthe American Astronomical Society, 16,696. LeBas, M.1. (1977) Carbonatite-nephelinite volcanism, 347 p. Wiley, New Sundheim, RR., Ed. (1964) Fused salts, 435 p. McGraw-Hill, New York. York. Tacker, R.c., and Stormer, J.C. (1993) Thermodynamics of mixing of -(1981) Carbonatite magmas. Mineralogical Magazine, 44, 133- liquids in the system Ca,(P04),-CaCI,-CaF,-Ca(OH),. Geochimica et 140. Cosmochimica Acta, 57,4663-4676. Levin, E.M., Robbins, C.R., and McMurdie, H.F. (1964) Phase diagrams Temkin, M. (1945) Mixtures of fused salts as ionic solutions. Acta Phys- for ceramists, vol. I, 60 I p. American Ceramic Society, Columbus, icochimica USSR, 20, 411-420. Ohio. Treiman, A.H. (1989) Carbonatite magma: Properties and processes. In - (1969) Phase diagrams for ceramists, vol. 2, 625 p. American Ce- K. Bell, Ed., Carbonatites: Genesis and evolution, p. 89-104. Unwin- ramic Society, Columbus, Ohio. Hyman, London. Lewis, G.N., and Randall, M. (1961) Thermodynamics (2nd edition), Treiman, A.H., and Essene, E.1. (1984) A periclase-dolomite-calcite car- (revised by Pitzer, K.1., and Brewer, L.), 723 p. McGraw-Hill, New bonatite from the Oka complex, Quebec, and its calculated volatile York. composition. Contributions to Mineralogy and Petrology, 85,149-157. Longhi, J. (1991) Complex magmatic processes on Mars: Inferences from Treiman, A.H., and Essene, E.1. (1985) The Oka carbonatite complex, the SNC meteorites. Proceedings of Lunar and Planetary Science, 21, Quebec: Geology and evidence for silicate-carbonate liquid immisci- 695-709. bility. American Mineralogist, 70, 1101-1113. Lumsden, J. (1966) Thermodynamics of molten salt mixtures, 146 p. Treiman, A.H., and Essene, E.J. (1992) Composition of the fluid phase Academic Press, New York. accompanying carbonatite magmas: A critical examination-Discus- McKie, D. (1966) Fenitization. In O.F. Tuttle and J. Gittins, Eds., Car- sion. American Mineralogist, 77, 663-665. bonatites, p. 261-294. Wiley, London. Treiman, AH., and Schedl, A. (1983) Properties of carbonatite magma McSwiggen, P.L. (1993) Alternative solution model for the ternary car- and processes in carbonatite magma chambers. Journal of Geology, 91, bonate system CaCO,-MgCO,-FeCO,: II. Calibration of a combined 437-447. ordering model and mixing model. Physics and Chemistry of Minerals, Watson, E.R (1991) Diffusion in fluid-bearing and slightly-melted rocks: 20, 42-55. Experimental and numerical approaches illustrated by iron transport Mirwald, P.W. (1979) Determination of a high-temperature transition of in dunite. Contributions to Mineralogy and Petrology, 107,417-434. calcite at 80O"C and I bar CO, pressure. Neues Jahrbuch f1.irMineral- Wyllie, P.J. (1989) Origin of carbonatites: Evidence from phase equilib- ogie Monatshefte, 309-315. rium studies. In K. Bell, Ed., Carbonatites: Genesis and evolution, p. Mysen, RO., and Virgo, D. (1980) Solubility mechanisms of carbon di- 500-545. Unwin-Hyman, London. oxide in silicate melts: A Raman spectroscopic study. American Min- Wyllie, P.J., and Haas, J.L. (1965) The system CaO-SiO,-CO,-H,O: I. eralogist, 65, 885-899. Melting relationships with excess vapor at I kilobar pressure. Geo- Nesbitt, RE., and Kelly, W.e. (1977) Magmatic and hydrothermal inclu- chimica et Cosmochimica Acta, 29, 871-892. sions in carbonatites of the Magnet Cove complex, Arkansas. Contri- Wyllie, P.1., and Tuttle, O.F. (1960) The system CaO-CO,-H,O and the butions to Mineralogy and Petrology, 63, 271-294. origin of carbonatites. Journal of Petrology, I, 1-46. Norton, G.E., and Pinkerton, H. (1992) The physical properties of car- Zarzycki, J. (1962) High-temperature X-ray diffraction studies of fused bonatite lavas: Implications for planetary volcanism (abs.). Lunar and salts: Structures of molten alkali carbonates and sulfates. Discussions Planetary Science, 23, 1001-1002. of the Faraday Society, 32, 38-48. Peterson, T.D. (1990) Petrology and genesis of natrocarbonatite. Contri- butions to Mineralogy and Petrology, 105, 143-155. MANUSCRIPT RECEIVED DECEMBER 30, 1993 Poletaev, I.F., Lyudomirskaya, A.P., and Tsiklina, N.D. (1975) Equilib- MANUSCRIPT ACCEPTED JULY 27, 1994