Sei sulla pagina 1di 32

Mineralogy and Petrology (2000) 68: 225256

Calciocarbonatite and magnesiocarbonatite rocks and magmas represented in the system CaO-MgO-CO2-H2O at 0.2 GPa
W.-J. Lee, M. F. Fanelli, N. Cava, and P. J. Wyllie
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA, USA With 11 Figures Received July 20, 1998; revised version accepted August 18, 1999 Summary The low-pressure eutectic for the coprecipitation of calcite, portlandite, and periclase/ brucite (with H2O-rich vapor) has served as a model for the existence and crystallization of carbonatite magmas. Attempts to determine conditions for the appearance of dolomite at this eutectic have been unsuccessful. We have discovered a second low-temperature eutectic for more magnesian liquids which excludes portlandite and includes dolomite (all results are vapor-saturated). Addition of Ca(OH)2-Mg(OH)2 to CaCO3-MgCO3 at 0.2 GPa depresses the liquidus to temperatures below the crest of the calcite-dolomite solvus; the vapor-saturated liquidus surface falls steeply, and the eld boundary for liquids coexisting with calcite and periclase reaches a peritectic at 880  C, where a narrow eld for liquidus dolomite begins, extending down to the eutectic at 659  C for the coprecipitation of calcite, dolomite and periclase (brucite should replace periclase at slightly higher pressures). The calcite liquidus is very large. The eld boundary for coexistence of calcite and dolomite extends approximately in the direction from CaMg(CO3)2 towards Mg(OH)2. The results illustrate conditions for the formation of mineral-specic cumulates from variable magma compositions. Hydrous (or sodic) carbonate-rich liquids with compositions from CaCO3 to CaMg(CO3)2 will precipitate calcite-carbonatites rst, followed by calcite-dolomite-carbonatites, with the prospect of precipitating dolomite-carbonatite alone through a limited temperature interval, and with periclase joining the assemblage in the closing stages. Periclase in the Fe-free system may represent the ubiquitous occurrence of magnetite in natural carbonatites. The restricted range for the precipitation of dolomite-carbonatites adds credibility to the evidence for primary magnesiocarbonatite (near-dolomite composition) magmas. Magnesiocarbonatite magmas can precipitate much calcite-carbonatite rock.

226 Zusammenfassung

W.-J. Lee et al.

Calciokarbonatitische und magnesiokarbonatitische Gesteine und Magmen im System CaO-MgO-CO2-H2O bei 0.2 GPa Das Niedrigdruck-Eutektikum der gemeinsamen Ausscheidung von Calcit, Portlandit und Periklas/Brucit (mit H2O-reicher Fluidphase) diente als Modell um die Existenz und Kristallisation karbonatitischer Magmen zu erklren. Versuche die Bedingungen a des Auftretens von Dolomit an diesem Eutektikum zu bestimmen blieben bisher ergebnislos. Wir entdeckten ein zweites Niedrigtemperatur-Eutektikum fr magneu siumreichere Schmelzen, das Portlandit ausschliet, aber Dolomit inkludiert (alle Ergebnisse bei Fluidsttigung). Die Zugabe von Ca(OH)2-Mg(OH)2 zu CaCO3-MgCO3 a bei 0.2 GPa senkt den Liquidus auf Temperaturen unter die Solvus-Schwelle von CalcitDolomit. Die uidgesttigte Liquidusche verluft steil und die Grenzche von a a a a Schmelze, die mit Calcit und Periklas koexistiert erreicht ein Peritektikum bei 880  C. Dort offnet sich ein schmales Feld fr Liquidus-Dolomit, das bis zum Eutektikum bei u 659  C reicht, an dem Calcit, Dolomit und Periklas (Brucit sollte Periklas bei geringfgig hoheren Drucken ersetzen) gemeinsam ausgeschieden werden. Der Calcitu Liquidus ist sehr gro. Die Linie an der Calcit und Dolomit koexistieren erstreckt sich ungefhr von CaMg(CO3)2 zu Mg(OH)2. Die Ergebnisse zeigen die Bildungsbedinguna gen fr die Bildung mineralspezischer Kumulate aus unterschiedlichen Magmenzuu sammensetzungen. Aus wssrigen (oder Na-reichen) karbonatreichen Schmelzen mit a Zusammensetzungen zwischen CaCO3 und CaMg(CO3)2 werden sich zuerst Calcit karbonatite und dann Calcit-Dolomitkarbonatite ausscheiden, mit der Moglichkeit Dolomitkarbonatite ber ein sehr eingeschrnktes Temperaturintervall zu bilden und u a mit Periklas, der zu dieser Vergesellschaftung im Endstadium hinzukommt. Periklas im Fe-freien System konnte das weitverbreitete Analog zu Magnetit in natrlichen u Karbonatiten sein. Der enge Bereich fr die Ausscheidung von Dolomitkarbonatiten u untermauert die Existenz primrer magnesiokarbonatitischer Magmen (nahe der a Zusammensetzung von Dolomit). Magnesiokarbonatitische Magmen konnen daher entsprechende Mengen an calcitkarbonatitischen Gesteinen ausscheiden.

Introduction This paper examines the relationships among dolomitic and calcitic melts (or magmas) and mineral precipitates (or cumulate rocks) in the system CaO-MgOCO2-H2O at 0.2 GPa. The experiments were made by M. Fanelli and N. Cava at the University of Chicago in 19761980, and a phase diagram was published by Fanelli et al. (1986) and Wyllie (1989, Fig. 20.6). We present here full details of the experiments with interpretations and applications based on recent discoveries and ideas in carbonatite petrogenesis. Carbonatite classication The nomenclature for carbonatite rocks, magmas, and synthetic carbonate-rich melts is rich and varied, with the employment of old rock names, of rock names based on mineralogy or on chemistry, and with magmas and experimental liquids dened on the basis of chemistry, or by the corresponding rock names, or by the nearest carbonate compositions. We outline some denitions in order to avoid

Calciocarbonatite and magnesiocarbonatite rocks and magmas

227

ambiguity in our comparison of experimental liquids and mineral assemblages with magmas and rocks. Detailed reviews of carbonatite mineralogy, composition and occurrences are contained in books by Heinrich (1966), Tuttle and Gittins (1966), Le Bas (1977), and Bell (1989). The dominance of intrusive calcite-carbonatites (sovites and alvikites) and less abundant dolomite-carbonatites (rauhaugite and beforesite) is established. Some carbonatites contain both calcite and dolomite, and dolomite may show extensive solid solution toward ankerite (Woolley and Kempe, 1989). Because of difculties in applying the IUGS mineralogical classication (Streckeisen, 1980), Woolley (1982) devised a system based on rock chemistry, which Woolley and Kempe (1989) later merged with the IUGS mineralogical system. In summary, (1) carbonatites are dened as having b 507 modal carbonates, (2) the old terms such as sovites and rauhaugite should be dropped, (3) names should be based on mineralogy if determined (e.g. apatite-pyrochloredolomite carbonatite), (4) names according to chemical analyses are calciocarbonatite, magnesiocarbonatite, and ferrocarbonatite, with boundaries: for C/ (C M F) b 0.8, calciocarbonatite; for C/(C M F) ` 0.8, magnesiocarbonatite if M b F, and ferrocarbonatite if F ` M (where C CaO, M MgO, F FeO Fe2O3 MnO, in weight). Calciocarbonatites form a distinct group, magnesiocarbonatites include most rauhaugites and beforsites (dolomite carbonatites), and ferrocarbonatites occupy a wider range, (5) natrocarbonatites are composed essentially of K-Na-Ca-carbonates. Gold's (1966) average carbonatite has SiO2 5.827. Woolley and Kempe (1989) determined average compositions for carbonatites with ` 107 SiO2. Features of the compositions (wt7) of average calciocarbonatite and magnesiocarbonatite are, respectively: CaO 49.12 and 30.12; MgO 1.80 and 15.06; SiO2 2.72 and 3.63; H2O 0.76 and 1.20. Magnesiocarbonatite is dominated by dolomite with variable ankerite in solid solution. When citing petrological accounts, we will use the authors' terms. For rocks, these may be the old names rejected by Woolley and Kempe (1989), and in some examples we may add a modern term in parentheses. Current preference for rocks is based on mineralogy, i.e. calcite-carbonatites and dolomite-carbonatites. For magmas and experimental melts, we will use the chemical denitions for calcioand magnesiocarbonatites, supported on occasion with the terms ``calcitic'' or ``dolomitic'', which experimentalists commonly use to describe carbonate-rich melts. The prex ``magnesio-'' is potentially confusing because it implies MgCO3 compositions, whereas liquids with dolomitic compositions are relevant in the range of experimental or magmatic interest. Sequence of crystallization in carbonatites From his detailed mapping of the Alno carbonatite complex, von Eckerman (1948, p. 148) concluded that sovite (calciocarbonatite) magma formed at an upper level (estimated pressure 90 MPa), and that beforsite (magnesiocarbonatite) magma formed at greater depth (estimated pressure 240 MPa). This sequence could be caused by pressure difference, or it could be time related, with the magmatic column solidifying downward (von Eckerman, 1966). Bailey (1960, personal

228

W.-J. Lee et al.

communication) reported reversed sequences of this depth relationship. Le Bas (1977) concluded from extensive eld studies that dolomite is not a normal member of high-level subvolcanic carbonatite intrusions. According to Garson and Smith (1958), the successive emplacement of calcitic then ankeritic sovite, followed by sideritic carbonatites (ferrocarbonatites) is very common, a conclusion shared by Le Bas (1977). Heinrich (1966) reported the normal sequences of crystallization of calcite, dolomite, ankerite or siderite, which can also be deduced from textures in individual specimens (e.g. Zhabin and Cheripiivskaya, 1965). There are also examples of dolomite phenocrysts in a calcite matrix, as in the Meach Lake complex near Ottawa (Hogarth, 1966) and in Dicker Willem complex, southwest Namibia (Gittins, 1973; Cooper and Reid, 1991). In the Jacupiranga complex, Brazil, there are examples of calcite with interstitial dolomite, and of dolomite with interstitial calcite (J. Gaspar, 1983, personal communication). Carbonatite petrogenesis and experiments The eld occurrences of associated carbonatites and alkaline igneous rocks have generally been interpreted as indicating that small quantities of carbonatite magmas are derived by fractionation or liquid immiscibility from a parent alkaline silicate magma of mantle parentage (Heinrich, 1966; Tuttle and Gittins, 1966; Le Bas, 1977; Bell, 1989). The discovery that carbonated peridotites melt to yield dolomitic melts (Wyllie and Huang, 1975, 1976; Eggler, 1976, 1978; Wallace and Green, 1988) gave credence to the few eld-based proposals that some carbonatite magmas were primary from the mantle as initially suggested by von Eckerman (1948) and Holmes (1950). Isotope studies conrm that the parents of carbonatite rocks are mantlederived (e.g. chapters in Bell, 1989). Several petrologists have recently interpreted some carbonatites as having precipitated from primary calciocarbonatite magmas (Gittins, 1989; Bailey, 1993; Barker, 1996) or magnesiocarbonatite (dolomitic) magmas (Bailey, 1989; Harmer and Gittins, 1997). Wyllie (1978) and Lee and Wyllie (1997a) have maintained from experiments that primary magmas from lherzolite must be dolomitic, and not calcitic, and would probably be erupted as vesiculating or explosive magmas from a depth of $ 70 km. Dalton and Wood (1993) interpreted their experimental data in terms of metasomatism of lherzolite to wehrlite, which could permit an original dolomitic liquid to change composition progressively toward CaCO3 as it rose through metasomatic wehrlite from $ 70 km to whatever depth it escaped from wehrlite equilibration. They proposed that this process would yield the primary calciocarbonatite magmas deduced from Bailey's (1993) eld studies. Lee and Wyllie (2000) have conrmed and elucidated this process in the system CaO-MgOSiO2-CO2, noting that the primary magma was probably quite siliceous, and that formation of calcite-carbonatites would require fractionation in the crust. Many discussions of carbonatite magmas tend to relate them closely to the rocks of interest. There is no problem relating basaltic magmas and basaltic rocks, but the use of rock names for magmas in petrogenetic discussions on carbonatites can be misleading. Wyllie and Tuttle (1960) found that calcite sinks within a few minutes in low-temperature melts in the system CaO-CO2-H2O at crustal pressures,

Calciocarbonatite and magnesiocarbonatite rocks and magmas

229

and concluded that in magmatic carbonatite systems, almost complete separation of successive liquids from crystals should be expected. Rapid sinking of, and almost complete separation of calcite and quartz from liquid in higher-temperature CaO-SiO2-CO2 melting experiments was also illustrated by Lee et al. (1994, Fig. 4). Laboratory experiments thus support the view that most intrusive carbonatites are cumulates, and are not representative of liquid compositions. Le Bas (1977, p. 277) reached the same conclusion from eld studies: ``The chemical composition of a rock specimen of carbonatite is little indication of the chemical composition of carbonatite magma.'' There are several phase diagrams showing crystallization behavior of liquids with compositions corresponding to carbonatite compositions. The melting relationships for calcite and dolomite have been studied in the system CaCO3MgCO3 at 2.7 GPa and 1 GPa by Irving and Wyllie (1975) and Byrnes and Wyllie (1981). The solidus temperatures (greater than 13001100  C, 31 GPa) are too high for carbonatite magmas, according to eld evidence. Solid solutions between calcite and dolomite are precipitated from the high temperature carbonate melts, but at all pressures to 3 GPa, the solidus temperatures are higher than the crest of the calcite-dolomite solvus. The phase relationships need to be matched with magmatic temperatures and depths of formation and emplacement, and with the associated petrological processes. For calcite and dolomite to be coprecipitated from carbonate-rich melts, additional components are required to lower the solidus temperature well below the crest of the calcite-dolomite solvus (10001100  C, depending on pressure). H2O is a component capable of accomplishing this. Some petrologists appeal to uorine (Gittins, 1989) or alkali carbonates (Cooper et al., 1975) to produce similar temperature reductions. Synthetic calciocarbonatite magma paths at 0.1 GPa extend through several hundred degrees down to about 650  C, terminating at eutectics between calcite and portlandite in the system CaO-CO2-H2O (Wyllie and Tuttle, 1960). Wyllie (1965) attempted to deduce conditions for the precipitation of dolomite in the system CaOMgO-CO2-H2O, combining data from CaO-CO2-H2O and MgO-CO2-H2O (Walter et al., 1962), and concluded that dolomite should appear with calcite, portlandite and liquid at higher pressure. Wyllie and Boettcher (1969) and Boettcher et al. (1980) followed the solidus reactions in CaO-CO2-H2O and CaO-MgO-CO2-H2O to 4 GPa, but they did not nd dolomite with liquid. We have now discovered why: there is a second low-temperature eutectic involving calcite and dolomite, separated from the calcite-portlandite eutectic by a thermal barrier on the liquidus. This is associated with a large, relatively low-temperature liquidus surface with elds for primary calcite and dolomite, which elucidate the crystallization behavior of calciocarbonatite and magnesiocarbonatite (dolomitic) magmas, and the cumulate rocks they leave behind. These results show that liquids with compositions of magnesiocarbonatite can precipitate calcite-carbonatite rocks. Experimental design The framework of phase relationships is outlined in Figs. 1 and 2. For abbreviations in text, refer to legends of Figs. 1 and 2, as appropriate. Figure 1 shows the system CaCO3-MgCO3 at 0.2 GPa, using known solvus relationships,

230

W.-J. Lee et al.

Fig. 1. Partly schematic phase relationships for CaCO3-MgCO3 at 0.2 GPa, based on data from Goldsmith and Heard (1961), Irving and Wyllie (1975), and Byrnes and Wyllie (1981). CC calcite; Cc, Do, Mc solid solutions of calcite, dolomite and magnesite; Pe periclase; L liquid; V vapor

Fig. 2. Bounding ternary phase relationships of the system CaO-MgO-CO2-H2O at 0.2 GPa, deduced from published data (Wyllie and Tuttle, 1960, for CaO-CO2-H2O; Ellis and Wyllie, 1979, for MgO-H2O-CO2). Heavy lines give the composition of liquid saturated with vapor and a mineral. Cooling directions are indicated by arrows. CH portlandite. See Fig. 1 for other abbreviations

and estimated reactions involving liquid based on extrapolation from the data of Irving and Wyllie (1975) at 2.7 GPa and Byrnes and Wyllie (1981) at 1 GPa (the calcite fusion curve is known). Temperatures for the reactions below $ 1000  C are closely dened by the array of univariant reactions constructed by Irving and Wyllie (1975), but the compositions of the high-temperature calcite-dolomite solid solution and liquidus loops are not dened; there are no data on the slope of the calcite solid solution curve bounding Cc Pe V. However, our petrological

Calciocarbonatite and magnesiocarbonatite rocks and magmas

231

Fig. 3. Estimated position of the vapor-saturated liquidus surface for liquids coexisting with minerals, and vapors near the line H2O-CO2 (modied from Wyllie, 1965, Fig. 2). The surface connects the vapor-saturated, ternary liquidus eld boundaries (heavy solid lines with arrows) as in Fig. 2. The heavy dashed quadrilateral Ca(OH)2-CaCO3-MgCO3Mg(OH)2 (CC-MC-MH-CH) include the compositions of starting mixtures (closed circles) for this study. Dashed line a-b indicates the intersection of the quadrilateral and vaporsaturated liquidus surface. The dashed lines connecting CaCO3, MgO and Vm form a triangle for a reaction determined in this study; the triangle and the vapor-saturated liquidus surface intersect at the dashed curve c-d

interest and new experiments with H2O are in the lower temperature range. Figure 1 is the basis for the vapor-saturated liquidus eld boundary in CaO-MgO-CO2, the heavy line in Fig. 2. Figure 2 also shows known and estimated liquidus eld boundaries at 0.2 GPa for the other three ternary systems bounding the tetrahedron CaO-MgO-CO2-H2O. The system CaO-CO2-H2O has been determined at 0.1 GPa and pressures up to 4 GPa (Wyllie and Tuttle, 1960; Wyllie and Boettcher, 1969). The geometry of the other two systems, and some of the liquidus temperatures, are readily estimated from previous experimental studies (e.g. Walter et al., 1962; Wyllie, 1965; Irving et al., 1977; Ellis and Wyllie, 1979). The heavy line in each system is the vapor-saturated liquidus eld boundary, giving the compositions of liquids coexisting with crystals and a vapor with composition near H2O-CO2. The dashed lines connect the four minerals used to make starting mixtures for the experiments, calcite (CC), magnesite (MC), brucite (MH) and portlandite (CH). Figure 3 shows the tetrahedron formed from the four ternary systems of Fig. 2. The four dashed lines frame the composition quadrilateral (CC-MC-MH-CH). The heavy lines are the ternary vapor-saturated liquidus eld boundaries, which are connected by the corresponding quaternary liquidus surface. The quadrilateral (CC-MC-MH-CH) and the liquidus surface intersect along the curved line a-b (estimated). The solid circles give the compositions of the starting mixtures, and these are not far separated from the estimated position of the vapor-saturated liquidus surface, with the separation increasing with Mg/Ca. The mixtures above

232

W.-J. Lee et al.

Fig. 4. Quadrilateral CC-MC-MH-CH to proper scale, showing the composition plane including the starting mixtures (see Fig. 3). Composition of each numbered mixture is given in Tables 1 and 2

the line a-b are in the excess vapor volume, and those below the line a-b are undersaturated in (CO2 H2O). The starting mixtures on the composition quadrilateral are shown more clearly in Fig. 4. The study was done in two stages. Fanelli (1978, unpublished M.S. Report) completed many runs on 47 mixtures in the triangle CC-MH-CH (Table 1). These results established the presence of a eutectic outside of that triangle, and Cava (1980, unpublished M.S. Report) dened its location with the additional mixtures in the triangle MC-MH-CH (Table 2). The results obtained for the two triangles are here combined in the quadrilateral (Figs. 3 and 4). Experimental methods Starting mixtures were prepared from (1) calcite (CaCO3Baker Analyzed Reagent; dried at 530  C for 48 h); (2) portlandite (Ca(OH)2Baker Analyzed Reagent; dried at 240  C for 3 h); (3) brucite (Mg(OH)2synthesized in a large volume Morey bomb using Baker Analyzed MgO; dried at 150  C for 5 h); (4) magnesite (MgCO3; prepared from basic magnesium carbonate in cold-seal pressure vessels with CO2 and open capsules; analyzed optically and by X-ray diffraction before use; dried at 150  C for 5 h). Components were mixed thoroughly in polyethylene vials in a ``Wig-l-Bug'' amalgamator, and the mixtures were stored at 110  C before using. Compositions of the starting mixtures are listed and numbered in Tables 1 and 2, and they are located by number in Fig. 4.

Calciocarbonatite and magnesiocarbonatite rocks and magmas

233

Table 1. Experimental results at 0.2 GPa in the join MgCO3-CaCO3-Ca(OH)2, by M. Fanelli

(continued)

234 Table 1 (continued)

W.-J. Lee et al.

(continued)

Calciocarbonatite and magnesiocarbonatite rocks and magmas Table 1 (continued)

235

The samples were sealed in gold capsules of 2 mm diameter and 1015 mm in length. Capsules were weighed before and after runs; capsules with leaks gain weight by water intake during experiments. All runs were held at 0.2 GPa in coldseal pressure vessels using distilled water as the pressure medium, and were then quenched to below 100  C using water (within 3045 sec) or air (34 min). Temperatures were measured internally by sheathed chromel-alumel thermocouples. Pressure is accurate to 57, and temperatures to 5  C (Piwinskii and Wyllie, 1968). Run durations were based on previous experience in the similar systems cited above. Time-series experiments conrmed that reaction rates with liquid present

236

W.-J. Lee et al.

Table 2. Experimental results at 0.2 GPa in the join Ca(OH)2-MgCO3-Mg(OH)2, by N. Cava

(continued)

Calciocarbonatite and magnesiocarbonatite rocks and magmas Table 2 (continued)

237

are fast. At the same P-T-X condition, runs from 3 h to 48 h produced the same phase assemblage with no change in crystal size or X-ray diffraction pattern (compare runs 41 and 5, 38 and 1, and 43 and 6; Table 1). Most experiments in the join CC-MC-CH were run for 320 h above 600  C, and up to 49.5 h below 600  C (Table 1). In contrast, the runs in the join MC-MH-CH were all done for 2 h, except for some reversal experiments (Table 2). Reversal experiments are denoted by R in Tables 1 and 2. Two runs on the join CC-MC-CH were held at 775  C for 3 h, and lowered rapidly to 675  C for additional 3 h. Each pair of forward and reversal experiments (compare runs 38 and 49R, and 24 and 51R) produced the same phase assemblage and nearly identical calcite compositions, conrming that equilibrium was achieved in short time in the presence of liquid. Mixes # 4852 on the join MC-MH-CH were used to reverse simultaneously both the reaction brucite periclase vapor, and the solidus. The runs were held for 2 h at 665  C, producing the assemblages Cc Pe L V or Cc Do L V, and then lowered to 655  C, 645  C and 640  C, and held for 2 or 168 h, as listed in Table 2. All liquids crystallized, reversing the solidus, but even after 168 h some metastable periclase remained. Subsolidus reactions are slower than the rapid hypersolidus reactions.

238

W.-J. Lee et al.

Identication of phases Capsules were examined under a binocular microscope to check for evidence of vapor present during a run: pits on the surface of the capsule and inated tops. Many capsules released gas or emitted water when punctured, indicating the presence of either a CO2-rich, or a H2O-rich vapor phase. The charges were removed from the capsules, and examined for evidence of crystal settling and vapor pits. One portion of the charge was then coarsely ground, and mounted in index oil for optical examination, while another was nely ground for X-ray diffraction analysis. Run products were scanned in the range 10 to 60 using ltered Cu K radiation. Some sample charges were polished and carbon-coated for additional textural examination using the scanning electron microscope (SEM), and for accurate determination of calcite compositions by electron microprobe analysis. The phases encountered were solid solutions of calcite (Cc), dolomite (Do), and magnesite (Mc), portlandite (CH), brucite (MH), periclase (Pe), liquid (L), and vapor (V). The carbonates were distinguished from other phases by their high birefringence and characteristic rhombohedral to rounded shapes. Distinction between carbonate phases was made by X-ray diffraction and on the basis of habit. Calcite formed euhedral rhombs in subsolidus runs, and larger, round crystals in the presence of liquid (e.g. Wyllie and Tuttle, 1960; Otto and Wyllie, 1993). Its size varied from 20 to 40 mm, but up to 80 mm in runs near the join CC-CH. Compositions were determined by X-ray powder diffraction, and electron microprobe. Calcite quenched from liquid formed dendrites of variable size and texture. Dolomite (a few mm in size) occurred as elongated prisms when an equilibrium phase, and dendritic when quenched from liquid. Magnesite occurred only in subsolidus runs as euhedral rhombohedra, as small as dolomite. The large size of calcite crystals in the presence of liquid provided the rst opportunity for reliable electron microprobe analyses of calcite solid solutions on the calcite-dolomite solvus. The unsatisfactory nature of analyses based on X-ray diffraction of ne-grained subsolidus carbonates was reviewed by Byrnes and Wyllie (1981). The calcite solvus was bracketed precisely by microprobe analyses of calcite solid solutions in forward and reversed runs between 825  C and 600  C. The curve was made available for the review of Goldsmith (1983), and detailed results are being prepared (Lee et al., manuscript in preparation). The analyses were made (197678) on the University of Chicago ARL-EMX electron microprobe tted with an energy dispersive X-ray system using Li drifted silicon detector, a beam current of 0.01 mA and a spot size of $ 5 mm. Two natural dolomites and a calcite were used as standards. Analyses were corrected for drift. Proles across hypersolidus calcites exhibit no zoning, so all analyses were taken at crystal centers to avoid contamination from surrounding phases. At each temperature, analyses of several minerals from forward and reversed experiments occupied a range of about 2 wt7 CaCO3, so we consider the average calcite compositions to be the reported values 17 CaCO3. The EMP measurements are consistent with the geometry of bracketed phase boundaries. Primary portlandite occurs as rounded platy crystals, 80120 mm across, with very low interference color and uniform extinction. Quench portlandite formed

Calciocarbonatite and magnesiocarbonatite rocks and magmas

239

transparent plates and blades many times larger than primary portlandite. These crystals are angular, had nger-like intergrowths of liquid, and usually exhibited irregular extinction or twinning. Subsolidus brucite formed irregular-shaped akes, up to 10 mm across. It did not appear as an equilibrium phase coexisting with liquid in any of our runs above 650  C, which is consistent with the dissociation reaction (Irving et al., 1977). X-ray diffraction conrmed that brucite formed from liquid during quench, but it was too ne-grained for detection by microscope. Periclase formed angular and equant grains when primary (1030 mm), and rounded and irregular when quenched from a liquid. In the near-solidus experiments, periclase crystals are generally small (3 to 5 mm), which made the distinction between primary and quench periclase difcult. Liquid in this system quenched to complex crystal intergrowths composed of the above minerals, with size and texture strongly dependent on bulk composition, temperature, and the primary minerals present. In many examples, the quenched products from near-solidus runs looked very similar to subsolidus assemblages observed with the optical microscope. Subtle textural differences became apparent under the SEM. Experimental results Experimental results are listed in Tables 1 and 2. Detection of trace amounts of vapor was difcult. The estimated liquidus surface in Fig. 3 indicates that vapor should be present in most of the melting experiments, those above line a-b, even when not reported in the tables. The results are presented in isothermal sections at 575  C, 675  C, 775  C and 825  C, as shown in Figs. 58. Additional runs were done at 500  C, and between 575  C and 675  C to bracket the temperatures of near-solidus reactions. The location of the phase boundaries at each temperature is constrained by the assemblages indicated in the tables, and internal consistency with results in other isothermal sections. Vapor-saturated liquidus eld boundaries intersected by the quadrilateral CC-MC-MH-CH are constructed in Fig. 9 from the corners of vapor-saturated 4-phase triangles in Figs. 68, and these are projected from the coexisting vapor compositions back to the vapor-saturated liquidus surface in the tetrahedron CaO-MgO-CO2-H2O, Fig. 10. Subsolidus relationships The subsolidus phase elds intersected at 575 C shown in Fig. 5a are consistent with results at 500  C (Table 1). The narrow widths for the join Cc-MH and DoMH in Fig. 5a were inferred from the limited solid solutions of calcite and dolomite, as illustrated in Fig. 1. The phase assemblages Cc MH and Do MH divide the quadrilateral into three regions for the elds Cc CH MH, Cc Do MH, and Do Mc MH. These ve phase elds are transferred onto the quadrilateral in the tetrahedron Fig. 5b, and they dene the quadrilateral subsolidus phase volumes for each phase assemblage, coexisting with periclase or with vapor. The subsolidus vapor phases coexisting with CC CH and MC MH both have high H2O/CO2 (Wyllie and Tuttle, 1960; Walter et al., 1962), as indicated

240

W.-J. Lee et al.

Fig. 5. a Isothermal phase elds intersected by the quadrilateral CC-MC-MH-CH at 575  C and 0.2 GPa, based on the experiments from mixtures indicated by closed circles (see results tabulated in Table 1). b Subsolidus phase volumes in the quaternary system. MH brucite. See Figs. 1 and 2 for other abbreviations

by CC CH Vm in Fig. 5b. Note that dolomite and magnesite cannot coexist with portlandite at these temperatures. Liquid is present at 675  C (Fig. 6), but three subsolidus phase elds remain. The eld Do Mc MH at 575  C has become Do Mc Pe V, and the two narrow elds Cc MH and Do MH are replaced by cross-cutting triangular

Calciocarbonatite and magnesiocarbonatite rocks and magmas

241

Fig. 6. Phase elds intersected by CCMC-MH-CH at 675  C and 0.2 GPa (Tables 1 and 2)

elds, Cc-Pe-V and Do-Pe-V, with H2O-rich vapor near Vm in Fig. 3. These changes are caused by the quadrilateral reaction (Walter et al., 1962): MgOH2 MgCO3 MgO V MH Mc Pe 1

which occurs between 655 and 660  C; many runs in Table 2 show the replacement of MH with Pe between temperatures of 655  C and 660  C, along with the appearance of liquid. The reaction was conrmed by the appearance of brucite below 655  C in the reversal experiments (from 665  C) in Table 2. Melting reactions The isothermal section at 675  C (Fig. 6) intersects several 4-phase tetrahedra, which appear on the quadrilateral as triangles giving compositions of 2 minerals and projected liquid composition, coexisting with vapors near Vm (Fig. 5b). We will refer to these as 4-phase triangles. There are two melting regions, involving the subsolidus composition triangles CC-MH-CH and Cc-Do-MH (Fig. 5). The former is associated with the known eutectic on CC-CH, and the other is related to a eutectic within Cc-Do-Pe-V. Figures 5a and 6 show that both solidus temperatures are between 575 and 675  C (as is reaction 1). The runs in Table 2 discussed above place the solidus reaction involving CcDo-MH-V between 655 and 660  C, indistinguishable from the brucite dissocia-

242

W.-J. Lee et al.

tion. We tentatively locate the dissociation at 656 C (reaction 1), and the solidus for CaCO3 CaMgCO3 2 MgO V L 2 Cc Do Pe at 659  C. Notice that with the quaternary phase relationships, when vapor is generated, the liquid compositions diverge from the quadrilateral plane CC-MCMH-CH, but the construction of Fig. 2 suggests that the divergence is fairly small for most compositions used. Additional runs in the join CC-MC-CH were conducted between 585  C and 600  C (not listed in Table 1) to locate the solidus reaction associated with the triangle CC-MH-CH. Samples remained subsolidus at 585  C, and showed trace melts at 600  C. We adopted a temperature of 600  C for the eutectic reaction, about 50  C lower than the dissociation reaction (1), indicating that brucite, not periclase, is part of the solidus reaction: CaCO3 CaOH2 MgOH2 V L Cc CH MH 3

Vapor-saturated liquidus eld boundaries The 4-phase triangles located by experimental brackets in Figs. 68 become larger with decreasing temperature, as the areas for L V become smaller. The positions of the triangles are closely dened by the runs, the compositions of minerals, and the overall phase relationships (continuity of each boundary as a function of temperature). The compositions of CH, Pe and MC remain essentially constant, and calcite in the area CC-MH-CH exhibits only minor solid solution. The compositions of coexisting Cc and Do are situated on the solvus limbs (Fig. 1), and the calcite compositions were measured by electron microprobe (Goldsmith, 1983; Lee et al., manuscript in preparation). The corners of the triangles are the projected compositions of liquids on the vapor-saturated liquidus eld boundaries. Curves connecting these corners thus dene the projected liquidus eld boundaries, and these are extended to the sides of the quadrilateral, to points corresponding to the bounding ternary reactions. The positions of the projected eld boundaries so determined are shown in Fig. 9. In the area CC-Do-MH, the three 4-phase triangles approaching each other with decreasing temperature represent the eld boundaries for L(Do, Pe, V), L(Cc, Do, V) and L(Cc, Pe, V), and they meet at eutectic E, 659  C, for reaction (2). From the sides of the area CC-MH-CH, we have intersected only two of the three 4-phase triangles, L(CH, Pe, V) and L(Cc, Pe, V). The third boundary, L(Cc, CH, V) exists only between 655  C and 600  C. The arrangement of eld boundaries near F satises the experimental data in Figs. 68, and our adopted value of 600  C and 650  C, respectively, for the solidus reaction F (3) and the dissociation of brucite (1). Connecting the two low-temperature regions around E and F (Fig. 9) is the eld boundary for coprecipitation of periclase and calcite, L(Cc, Pe, V), which passes over the thermal maximum m (with estimated temperature 900  C) where the two

Calciocarbonatite and magnesiocarbonatite rocks and magmas

243

Fig. 7. Phase elds intersected by CCMC-MH-CH at 775  C and 0.2 GPa (Tables 1 and 2)

Fig. 8. Phase elds intersected by CCMC-MH-CH at 825  C and 0.2 GPa (Tables 1 and 2)

244

W.-J. Lee et al.

Fig. 9. Vapor-saturated liquidus eld boundaries intersected by the quadrilateral CH-CC-MC-MH, dened by the loci of corners of phase triangles (or phase volumes in tetrahedron, as in Fig. 5b) in the isothermal sections at 675  C, 775  C and 825  C (Figs. 68). The temperatures for the eutectics E and F, and peritectic P, and the corresponding reactions, are discussed in the text. Point m (estimated $ 900  C) indicates a thermal maximum on the liquidus eld boundary between primary periclase and calcite, which is also on the dashed line corresponding to c-d in Fig. 3

4-phase triangles Cc Pe L V coalesce to the dashed line Cc-Pe-V. This line is the line of intersection between the quadrilateral and the triangle Cc-Pe-Vm in Fig. 3, which contains the reaction: CaCO3 MgO Vm Lm Cc Pe 4

Returning to eutectic E in Fig. 9, the eld boundary L(Cc, Do, V) is well dened by experiments. Figures 68 show the 4-phase triangle Do Pe L V only as a narrow triangle; the join Do-MH thus constrains the position of the eld boundary L(Do, Pe, V) quite closely. These two eld boundaries enclose the projected vapor-saturated liquidus surface for primary dolomite. Figure 1 shows that the maximum temperature for the coexistence of calcite and dolomite at this pressure is about 880  C, where dolomite dissociates yielding the assemblage Cc Pe V. Therefore, the two eld boundaries must meet at 880  C at peritectic P (well dened by extrapolation from Figs. 6, 7, and 8) corresponding to the reaction in Fig. 1, with liquid included: CaMgCO3 2 V CaCO3 MgO L Do Cc Pe 5

The eld boundary L(Cc, Pe, V) rises steeply from P to the high-temperature reaction where magnesian calcite melts incongruently, as shown in Figs. 1 and 2, by the reaction (at $ 1900  C): CaY MgCO3 MgO L VCO2 Cc Pe 6

Calciocarbonatite and magnesiocarbonatite rocks and magmas

245

The liquidus prole from Fig. 1, including the temperature minimum, is shown in Fig. 9 by the arrows on the join CaCO3-MgCO3. The two liquidus eld boundaries L(Cc, Do, V) and L(Do, Pe, V) between P and E change from absorption reactions which consume Cc and Pe, respectively, at higher temperatures near P, to subtraction reactions for the coprecipitation of Do and Cc, and Do and Pe, respectively, at lower temperatures down to E. The vapor-saturated liquidus surface The phase elds and liquidus eld boundaries intersected by the quadrilateral in Fig. 9 have been projected onto the vapor-saturated liquidus surface in Fig. 10. Each point in Fig. 10 is determined by projection of the corresponding point in Fig. 9 along a line from the coexisting vapor phase composition. Vapor phase compositions change from near H2O (Vm, Fig. 2) at low temperatures, to increasing CO2 /H2O as temperature increases from E and F toward the carbonate side. Values have been determined for CaO-CO2-H2O at 0.1 GPa (Wyllie and Tuttle, 1960), and we assume on this basis that vapors retain high H2O/CO2 up to 900  C. Although the projection is only semi-quantitative, the arrangement of the boundaries and reaction points remain well-constrained by Figs. 69. The narrow dolomite eld appears much more elongated in the 3-dimensional perspective view of Fig. 10 than in Fig. 9. The estimated compositions (wt7) of the eutectic E at 659  C is 357 CC 217 MC 447 MH (less some H2O and CO2 as vapor), and that for the peritectic P at 880  C is 487 CC, 477 MC, and 57 MH (less some CO2 and H2O as vapor). Figure 10a shows the overall geometry in the tetrahedron, and Fig. 10b shows an enlarged version of the vapor-saturated liquidus surface with estimated isotherms drawn at 500  C and 100  C intervals above and below 1000  C, respectively. Note the steep liquidus surface for periclase, dropping down from the MgO and the carbonate side (see Fig. 1) to the low temperature eutectics E and F. The isotherms show the temperature maximum on the periclase surface, and a similar kind of surface on the Ca(OH)2 side of the maximum down to the eutectic F, involving calcite, portlandite and brucite, but no dolomite. The topography of the liquidus surface on the carbonate side of the liquidus maximum is remarkable. The very deep, basin-like liquidus for calcite and dolomite slices into the periclase liquidus, generating on three sides the steepest parts of the surface for primary periclase. The dashed curve on the carbonate liquidus (Fig. 10b) represents a thermal through extending from the liquidus-solidus minimum of CaCO3-MgCO3 in Fig. 1. It is sketched to disappear towards the low temperature region. Petrological applications Problems involving the petrogenesis of carbonatites include: (1) the processes and conditions leading to the generation of carbonatite magmas, (2) the chemical compositions of primary and derivative carbonatite magmas, and (3) the crystallization paths of carbonatite magmas, and their relationships to carbonatite rocks.

246

W.-J. Lee et al.

Fig. 10. a Field boundaries on the vapor-saturated liquidus surface, projected through estimated vapor phase compositions from their intersections on the quadrilateral CC-MC-MH-CH. Temperatures for the key reactions are given in Fig. 9. b Sketch of isotherms on the vapor-saturated liquidus surface of a. Temperatures in brackets are rough estimates from literature

There is broad agreement that carbonatite magmas are derived either as primary melts from mantle peridotite, or from mantle-derived alkalic silicate parent magmas by a process of fractional crystallization or liquid immiscibility (e.g. Bell, 1989). Carbonatite magmas therefore follow silicate-carbonate liquidus eld boundaries. The magmas must be dominated by carbonates, but they all precipitate some silicate and accessory minerals. Our experiments show the effect of H2O on the temperatures of melts from CaCO3-MgCO3, representing calciocarbonatite and

Calciocarbonatite and magnesiocarbonatite rocks and magmas

247

magnesiocarbonatite magmas. The effects of alkalis are shown by experiments of Cooper et al. (1975) and Harmer and Gittins (1997). Several phase diagrams suggest a clustering of immiscible carbonate-rich liquids at compositions corresponding to nepheline-sovite (Lee and Wyllie, 1996, 1997b, 1998). There is evidence for immiscible Mg-bearing carbonate-rich liquids in some melilititecarbonate compositions (Kjarsgaard and Hamilton, 1989). Lee and Wyllie (1994, 1997b, 1998) have emphasized that carbonatite magmas must follow carbonatesilicate liquidus eld boundaries, and that magmas cannot enter the carbonate liquidus elds (forbidden zones). Therefore, although the system CaO-MgO-CO2H2O shows the compositions and temperatures of carbonate-rich liquids, more precise modeling of carbonatite magmas requires data involving silicate minerals. Reasons were outlined in the ``Introduction'' for our conclusion that most plutonic carbonatite rocks are crystal cumulates. Crystallization paths of carbonaterich liquids in CaO-MgO-CO2-H2O model those of carbonatite magmas, and the precipitated mineral assemblages represent carbonatite rocks. The data of Franz and Wyllie (1967) and Otto and Wyllie (1993) with silicate components added indicate the additional silicate minerals to be expected from these Mg-bearing, low-alkali carbonatite magmas. Lee and Wyllie (1996, 1997b, 1998) have illustrated crystallization paths for low-Mg, Ca-Na carbonatite magmas at pressures of 1 GPa and above. Effect of volatile components on the precipitation of calcite and dolomite Melting temperatures in the system CaCO3-MgCO3 are relatively high (e.g. Fig. 1). This system strongly inuences phase relationships on the carbonate-silicate liquidus boundary which controls the compositions of primary carbonatite melts from carbonated mantle peridotite (Wyllie and Huang, 1976; Wyllie and Lee, 1998). At 1 GPa ($ 35 km depth), carbonate liquids may precipitate continuous solid solutions between calcite and dolomite at temperatures above 1080  C (Byrnes and Wyllie, 1981). With decreasing pressure this range becomes more restricted, concomitant with expansion of the periclase eld. At 0.2 GPa the range of solid solutions is from calcite to about $ 30 wt7 MgCO3 at temperatures between about 1100  C and 1400  C (Fig. 1). Addition of H2O, or Ca(OH)2, lowers the liquidus temperature of calcite (Wyllie and Tuttle, 1960). Figures 9 and 10 show that at 0.2 GPa, the solution of Ca(OH)2 generates liquids which precipitate only calcite, corresponding to calcite-carbonatites, between 1360  C and 655  C. Addition of MgO to these liquids generates low-temperature elds for brucite and periclase (near F, Fig. 10), but no dolomite (Wyllie, 1965; Boettcher et al., 1980). Therefore, although magmas related to the eld boundary from m to eutectic F (Fig. 9) could explain calcite-carbonatites, they provide no evidence for the formation of dolomite-carbonatites. No dolomite is precipitated from liquids on the join CaCO3-MgCO3 at 0.2 GPa (Fig. 1), but with addition of Mg(OH)2 dolomite appears as the liquidus temperatures are lowered (Fig. 10). The minimum anhydrous liquidus temperature of 1100  C is lowered to 659  C at E, with a liquidus eld for dolomite appearing below 880  C at P. Therefore, dolomite-carbonatites could be precipitated from magmas related to the area P-E. Figure 11a shows the phase elds intersected by

248

W.-J. Lee et al.

Fig. 11. a Part of phase elds intersected by Do-MH at 0.2 GPa, based on results in Figs. 510. b Liquidus prole intersected by the join Do-NC (Na2CO3) at 0.1 GPa, after Harmer and Gittins (1997). Ny nyerereite, Na2Ca(CO3)2

the TX cross-section when Mg(OH)2 is added to dolomite. Three eld boundaries are intersected, as shown by the join CaMg(CO3)2-Mg(OH)2 in Fig. 9, giving piercing points separating the liquidus elds for periclase and calcite, calcite and dolomite, and dolomite and periclase. Thus, adding Mg(OH)2 to molten dolomite causes the precipitation of calcite through a temperature interval before dolomite is precipitated. Harmer and Gittins (1997) demonstrated that Na2CO3 (NC) has a similar effect. Figure 11b reproduces their experimental results, showing a portion of the liquidus phase relationships intersected by Do-NC at 0.1 GPa. The general arrangement of the liquidus surfaces intersected in Fig. 11b is remarkably similar to that in Fig. 11a, despite the different conditions of the two sets of experiments. The traverse from CaMg(CO3) to Ca0.5Mg0.5(OH)2 is accomplished by increasing the H2O/CO2 ratio. The phase elds intersected by this cross-section (see Fig. 9) would intersect the plateau-like liquidus for calcite (at temperature near 850  C), bounded on both sides by the steep periclase surface. Dolomite is precipitated only at temperatures below the liquidus. The large liquidus surface for calcite extending to eutectic E (Figs. 9 and 10) shows that hydrous dolomitic (magnesiocarbonatite) liquids precipitate calcite through some temperature interval before dolomite is precipitated. The phase elds for Cc L V determined in

Calciocarbonatite and magnesiocarbonatite rocks and magmas

249

Figs. 68 demonstrate that highly magnesian carbonate-rich liquids precipitate calcites exhibiting only limited solid solution of MgCO3; this suggests that magnesiocarbonatite magmas can precipitate calcite-carbonatites through a wide range of P-T-X conditions. Paths of crystallization, and cumulate rocks Consider the liquidus surface for calcite, which drops down steeply from CaCO3, divided into two relatively low temperature surfaces by the thermal ridge extending from CaCO3 towards thermal maximum m (Figs. 9 and 10). Most liquids with compositions on the MgCO3 side of the ridge precipitate calcite followed by dolomite, and all terminate at eutectic E; coexisting vapor phase compositions range from high-temperature CO2 to low-temperature H2O-rich Vm. Most liquids with compositions on the Ca(OH)2 side of the ridge precipitate calcite followed by periclase or brucite, and all terminate at eutectic F; coexisting vapor phase compositions range from high-temperature CO2 to a low-temperature vapor with composition extremely rich in H2O (Wyllie and Tuttle, 1960; Wyllie, 1965). Consider three paths of crystallization of a vapor-saturated liquid on the MgCO3 side of the ridge on the calcite liquidus. It is geometrically difcult to illustrate precise liquid paths on a curved surface in a tetrahedron, but Fig. 9 permits visualization with only minor distortion. All cooling liquids begin with the precipitation of calcite (slightly magnesian) following slightly curved paths directed away from CaCO3. (1) A limited range of liquid compositions would reach the eld boundary between m-E, then coprecipitate calcite and periclase, and solidify at E with addition of dolomite. (2) The adjacent zone of liquids would reach the calcite-dolomite eld boundary rising from E, with both minerals being coprecipitated until they were joined by periclase at E, where the liquid was used up. (3) The next zone of liquids would reach the calcite-dolomite eld boundary at higher temperatures, where it is a reaction boundary. Then, as temperature decreased, calcite would be resorbed as dolomite was precipitated. (3A) For some conditions, all calcite could be resorbed, and the liquid would then follow a path across the dolomite liquidus, precipitating only dolomite until it reached the dolomite-periclase eld boundary, and made its way down to the eutectic E where calcite would again be coprecipitated. (3B) For other conditions, resorption of calcite would cease before it disappeared, and then calcite and dolomite would be coprecipitated down to eutectic E, as in path (2). Consideration of the vapor-absent phase relationships introduces additional variety in paths of crystallization. Behind each liquidus area on the surface in Fig. 10a is a corresponding vapor-absent liquidus volume, the sides of which are shown in the ternary diagrams of Fig. 2. However, this should introduce no signicantly different sequences of mineral precipitation. Furthermore, the similarity between Figs. 11a and 11b with respect to the calcite and dolomite elds suggests that similar precipitates may be produced through a range of temperatures from

250

W.-J. Lee et al.

CaCO3-MgCO3 liquids, either by increasing H2O/CO2 or Na2CO3 content (Harmer and Gittins, 1997). In summary, hydrous (or sodic) carbonate-rich liquids with compositions from CaCO3 to CaMg(CO3)2 are likely to precipitate calcite rst (calcite-carbonatites), followed by calcite and dolomite (calcite-dolomite-carbonatites, with the prospect that at higher temperatures calcite is being resorbed), with the possibility of precipitating dolomite alone (dolomite-carbonatite) through a limited temperature interval, and with periclase joining the assemblage in the closing stages (periclasecalcite-dolomite-carbonatite). At somewhat higher pressures, brucite should replace periclase at the eutectic. There is only a very restricted range of liquid compositions which will precipitate dolomite rst. Bailey (1993) considered that the bimodality of magma types, or ``the compositional dichotomy between calcioand magnesiocarbonatites'' F F F ``represents a major challenge to carbonatite petrology''. The phase relationships in Figs. 9 and 10 indicate that a continuity of liquid compositions between calcio- and magnesiocarbonatites can provide a bimodality of cumulate rocks, due simply to depression of the liquidus to intersect the calcite-dolomite solvus; the liquidus depression can be caused by increasing concentrations of H2O or Na2CO3 in residual liquids, under equilibrium or fractional conditions. Periclase in phase diagrams may represent magnetite in carbonatites Periclase is rare in carbonatites. The assemblage calcite-dolomite-periclase has been reported from the Oka carbonatite (Treiman and Essene, 1984). Mariano and Roeder (1983) have described a carbonatite from Kerimasi with euhedral periclase, which is commonly overgrown on magnesioferrite, in a matrix of calcite and monticellite. Magnesioferrite-magnetite minerals (with composition near Fe3O4) are ubiquitous in carbonatites. Our earlier experiments in CaO-FeO-CO2-H2O and other Fe-bearing synthetic model systems, using cold-seal pressure vessels (1960s, unpublished), produced samples obscured by the precipitation of ne magnetite. These observations suggest that (Fe Mg)O in carbonatite magmas may mimic the role of MgO in the synthetic system, with partial oxidation to yield magnetite rather than wustite. Relationship of CaCO3-MgCO3-rich magmas to silicate parents The phase relationships at mantle pressures for the carbonate-silicate eld boundary adjacent to the liquidus elds for CaCO3-MgCO3 in model systems (CaO-MgO-Al2O3-SiO2-CO2) has been investigated. It is established that the nearsolidus liquid from carbonate-lherzolite at pressures extending at least between 2.8 and 7 GPa is a magnesiocarbonatite, with composition dominated by calcic dolomite, containing about 5 wt7 SiO2 (Wyllie and Huang, 1976; Eggler, 1978; Wyllie and Lee, 1998; Dalton and Presnall, 1998). Dalton and Wood's (1993) experiments indicating that magnesiocarbonatite liquid at pressures below 2.5 GPa would metasomatize lherzolite to wehrlite, with concomitant change of liquid to calciocarbonatite, have been conrmed in the model system by Lee and Wyllie (2000). The effect of H2O on the solidus of model dolomite-lherzolite has been

Calciocarbonatite and magnesiocarbonatite rocks and magmas

251

determined from 2 to 3 GPa (White and Wyllie, 1992), but many details remain to be studied. Bailey (1989, 1993) and Harmer and Gittins (1997) have concluded that some dolomite-carbonatites represent primary magmas from mantle peridotite, which is consistent with phase relationships for carbonated lherzolite, as outlined above. The restricted conditions for the formation of dolomite-carbonatite cumulates in Figs. 9 and 10 become irrelevant for evidence of primary magmas rising directly from the mantle, and may even be considered as circumstantial evidence for the formation of such liquids as primary magmas rather than as low pressure differentiates. The controlling phase relationships for the precipitation of carbonatite magmas represented in Figs. 910 are given by the eld boundaries on the carbonate-silicate liquidus generated when silicate components are added to CaO-MgO-CO2-H2O. Experiments on the addition of many silicate components to the join CaCO3Ca(OH)2 conrm that the carbonate-silicate liquidus in the region of F contains only about 57 dissolved silicate, with the silicate liquidus temperature rising steeply above the carbonate-silicate boundary (for reviews, see Wyllie, 1966, 1989). Franz and Wyllie (1967) added Mg2SiO4 at 0.1 GPa and encountered three reactions between 600  C and 650  C which added monticellite to the three reactions around the brucite eld near F. Reactions at higher temperatures introduced merwinite (at 755  C) and forsterite (at 895  C). Otto and Wyllie (1993) located the carbonate-silicate boundary at 0.2 GPa associated with the calcite liquidus on the MgCO3 side of the thermal maximum CaCO3-m (Figs. 9 and 10). They added SiO2 (in form of silica gel, with 4.37 H2O) to three mixtures with compositions close to a line between CaCO3-E. On each of the three composition joins, the calcite-silicate liquidus piercing point contains between 2.5 and 57 SiO2, and this is followed by an extensive liquidus for forsterite. No other silicate mineral is encountered until more than 187 SiO2 is added. Monticellite appears as a liquidus phase only at temperatures above 1000  C. The solidus reaction, with composition near that of E and temperature between 650  C and 660  C, was determined: calcite dolomite periclase forsterite vapor liquid Cc Do Pe Fo V L 7

The liquid composition is estimated to be in the range: 15207 MgCO3, 35407 CaCO3, 40457 Mg(OH)2, 567 Mg2SiO4. These results indicate that the paths of crystallization described for the carbonate-rich liquids in the SiO2-free system are very similar to those on the carbonate-silicate liquidus boundaries, with the addition of monticellite for liquids associated with F, and with forsterite for liquids on the MgCO3 side of the temperature maximum CaCO3-m, extending down to the region of E. The liquids terminating at E are better models for carbonatite magmas than those reaching F, because they do not involve portlandite (unknown as a primary mineral in carbonatites) and they do involve dolomite (present in carbonatites). Also, olivine is a widespread constituent of many plutonic carbonatites, whereas monticellite is less abundant (Heinrich, 1966).

252

W.-J. Lee et al.

Conclusions We conclude from the cited experiments in the system CaO-MgO-SiO2-CO2-H2O at 0.10.2 GPa that carbonatite magma crystallization produces the sequences of cumulates described above for liquids in the area CaCO3-m-P-E, together with a few percent of forsterite. Given the large liquidus eld for calcite (and the corresponding liquidus for calcite-forsterite), the fractional crystallization of both calciocarbonatite and magnesiocarbonatite magmas should yield forsterite-calcitecarbonatites rst, and these should be more abundant than the following precipitates of forsterite-calcite-dolomite-carbonatites. Forsterite-dolomite-carbonatites should be less abundant, and appear to require rather restricted conditions. Dolomite-carbonatites with few associated igneous rocks are prime candidates for precipitation from primary carbonatite magmas. According to Figs. 9 and 10, and the results of Otto and Wyllie (1993), we expect the late stages of (low-alkali) carbonatite magma evolution to be represented by the assemblage forsterite-calcite-dolomite-periclase (or magnetite) precipitated between 700600  C, with nal liquids containing MgO : CaO near 2 : 1 by weight (plus accessory minerals). The coexisting vapor composition has not been measured, but H2O/CO2 shouldbehigh, bycomparison withVm(Fig.10). TreimanandEssene (1984) reported a eutectic assemblage in a dike from Oka, with modal composition 52.37 calcite, 15.17 dolomite, 17.57 brucite (in part altered periclase), 3.87 periclase, and 0.67 forsterite (assemblage for reaction 7), together with 7.67 opaques, and 3.57 apatite. They calculated the vapor phase composition to be high in H2O/CO2. The occurrence of olivine in natural carbonatites, and the olivine-calcite liquidus eld boundaries in the model system are consistent with the derivation of carbonatite magmas from a mantle-derived, olivine-bearing silicate parent. Wyllie (1966) and Otto and Wyllie (1993) noted the occurrence of thermal barriers on the liquidus of many phase diagrams between high temperature silicate liquids and low temperature carbonate-rich magmas, but Watkinson and Wyllie (1971) and Lee and Wyllie (1994) located continuous liquidus paths from hydrous nepheline-normative liquids to CaCO3-rich liquids. The phase diagrams are thus consistent with the process of derivation of carbonatite magmas by fractional crystallization of primary olivine-nephelinites, with cumulus calciocarbonatites being the dominant product even for magmas with high Mg/Ca ratios. Other phase diagrams illustrate liquidus paths for olivine-nepheline-normative silicate liquids which intersect a liquid miscibility gap. Experiments with mixtures of magnesian nephelinite, Na2CO3 and CaMg(CO3)2 located an extensive liquidus eld for olivine adjacent the miscibility gap (Baker and Wyllie, 1990; Lee and Wyllie, 1997a), and a liquidus eld boundary between olivine and calcite for lowalkali compositions. For carbonatite magmas with signicant alkali contents, which is expected for magmas derived through liquid immiscibility, the carbonate-silicate eld boundaries would be associated with other silicates, including melilite, and they would not terminate near E. These liquids would continue to lower temperatures along liquidus paths leading to alkali carbonates (Cooper et al., 1975), as described and illustrated by Lee and Wyllie (1997b). The phase relationships discussed above establish the broad outline of the behavior of magmas reaching a silicate-calcite/dolomite eld boundary at low

Calciocarbonatite and magnesiocarbonatite rocks and magmas

253

crustal pressures. With continued differentiation and the concentration of minor elements, and decrease in pressure as magmas approach the surface, more complex relationships are likely to develop (e.g. Kjarsgaard and Peterson, 1991; Macdonald et al., 1993; Kjarsgaard et al., 1995).
Acknowledgments We thank J. A. Dalton and R. Luth for constructive reviews. This research was supported by US National Science Foundation Grant EAR-9218806.

References
Bailey DK (1989) Carbonate melt from the mantle in the volcanoes of south-east Zambia. Nature 338: 415418 Bailey DK (1993) Carbonate magmas. J Geol Soc Lond 150: 637651 Baker MB, Wyllie PJ (1990) Liquid immiscibility in a nephelinite-carbonate system at 25 kbars and implications for carbonatite origin. Nature 346: 168170 Barker DS (1996) Carbonatite volcanism. In: Mitchell RH (ed) Undersaturated alkaline rocks: mineralogy, petrogenesis, and economic potential. Mineralogical Association of Canada, Short Course 24, pp 4561 Bell K (ed) (1989) Carbonatites: genesis and evolution. Unwin Hyman, London, 618 pp Boettcher AL, Robertson JK, Wyllie PJ (1980) Studies in synthetic carbonatite systems: solidus relationships for CaO-MgO-CO2-H2O to 40 kilobars and CaO-MgO-SiO2-CO2H2O to 10 kilobars. J Geophys Res 85: 69376943 Brogger WC (1921) Die Eruptivgesteine des Kristianiagebietes. IV. Das Fengebiet in Telemark, Norwegen. Norsk Vidensk Selsk Skrift I, Math Naturw Klasse 9 Byrnes AP, Wyllie PJ (1981) Subsolidus and melting relations for the join CaCO3-MgCO3 at 10 kb. Geochim Cosmochim Acta 45: 321328 Cooper AF, Reid DL (1991) Textural evidence for calcite carbonatite magmas, Dicker Willem, southwest Namibia. Geology 19: 11931196 Cooper AF, Gittins J, Tuttle OF (1975) The system Na2CO3-K2CO3-CaCO3 at 1 kilobar and its signicance in carbonatite petrogenesis. Am J Sci 275: 534560 Dalton JA, Wood BJ (1993) The compositions of primary carbonate melts and their evolution through wallrock reaction in the mantle. Earth Planet Sci Lett 119: 511525 Dalton JA, Presnall DC (1998) Carbonatitic melts along the solidus of model lherzolite in the system CaO-MgO-Al2O3-SiO2-CO2 from 3 to 7 GPa. Contrib Mineral Petrol 131: 123135 von Eckermann H (1948) The alkaline district of Alno Island. Sveriges Geologiska Undersokning, Serie Ca 36: 176 von Eckermann H (1966) Progress of research on the Alno carbonatite. In: Tuttle OF, Gittins J (eds) Carbonatites. John Wiley, New York, pp 331 Eggler DH (1976) Does CO2 cause partial melting in the low-velocity layer of the mantlec Geology 4: 6972 Eggler DH (1978) The effect of CO2 upon partial melting of peridotite in the system Na2OCaO-Al2O3-MgO-SiO2-CO2 to 35 kb, with an analysis of melting in a peridotite-H2OCO2 system. Am J Sci 278: 305343 Ellis D, Wyllie PJ (1979) Carbonation, hydration, and melting relations in the system MgOH2O-CO2 at pressures up to 100 kilobars. Am Mineral 64: 3240 Fanelli MF, Cava N, Wyllie PJ (1986) Calcite and dolomite without portlandite at a new eutectic in CaO-MgO-CO2-H2O, with applications to carbonatites. Proceedings 13th General Meeting, International Mineralogical Association, Bulgaria, pp 313322

254

W.-J. Lee et al.

Franz GW, Wyllie PJ (1967) Experimental studies in the system CaO-MgO-SiO2-CO2H2O. In: Wyllie PJ (ed) Ultramac and related rocks. John Wiley, New York, pp 323326 Garson MS, Smith WC (1958) Chilwa Island Memoir no 1. Geol Surv Dept Nyasaland Gittins J (1973) The signicance of some porphyritic textures in carbonatites. Can Mineral 12: 226228 Gittins J (1989) The origin and evolution of carbonatite magmas. In: Bell K (ed) Carbonatites: genesis and evolution. Unwin Hyman, London, pp 580600 Gold DP (1966) The average and typical chemical composition of carbonatites. International Mineralogical Association, Papers, Fourth General Meeting, India, pp 8391 Goldsmith JR (1983) Phase relations of rhombohedral carbonates. In: Reeder RJ (ed) Reviews of mineralogy, vol 11. Carbonates: mineralogy and chemistry. Mineral Soc Am, Washington, DC, pp 4976 Goldsmith JR, Heard HC (1961) Subsolidus phase relationships in the system CaCO3MgCO3. J Geol 69: 4574 Harmer RE, Gittins J (1997) The origin of carbonatites: eld and experimental constraints. J Afr Earth Sci 25: 528 Heinrich EW (1966) The geology of carbonatites. Rand McNally, Chicago, 555 pp Hogarth DD (1966) Intrusive carbonate rock near Ottawa, Canada. Papers and Proceedings of the 4th General Meeting, International Mineralogical Association, India, pp 4553 Holmes A (1950) Petrogenesis of katungite and its associates. Am Mineral 35: 772792 Irving AJ, Wyllie PJ (1975) Subsolidus and melting relationships for calcite, magnesite, and the join CaCO3-MgCO3 to 36 kilobars. Geochim Cosmochim Acta 39: 3553 Irving AJ, Huang WL, Wyllie PJ (1977) Phase relations of portlandite Ca(OH)2, and brucite, Mg(OH)2, to 33 kilobars, Am J Sci 277: 313321 Kjarsgaard BA, Hamilton DL (1989) Carbonatite origin and diversity. Nature 338: 547548 Kjarsgaard B, Peterson T (1991) Nephelinite-carbonatite liquid immiscibility at Shombole Volcano, East Africa: petrographic and experimental evidence. Mineral Petrol 43: 293 314 Kjarsgaard BA, Hamilton DL, Peterson TD (1995) Peralkaline nephelinite/carbonatite liquid immiscibility: comparison of phase compositions in experiments and natural lavas from Oldoinyo Lengai. In: Bell K, Keller J (eds) Carbonatite volcanism: Oldoinyo Lengai and the petrogenesis of natrocarbonatites. Springer, Berlin Heidelberg New York Tokyo, pp 163190 (IAVCEI Proceedings in Volcanology 4) Le Bas MJ (1977) Carbonatite-nephelinite volcanism. John Wiley, London, 347 pp Lee WJ, Wyllie PJ (1994) Experimental data bearing on liquid immiscibility, crystal fractionation, and the origin of calciocarbonatites and natrocarbonatites. Int Geol Rev 36: 797819 Lee WJ, Wyllie PJ (1996) Liquid immiscibility in the join NaAlSi3O8-CaCO3 to 2.5 GPa and the origin of calciocarbonatite magmas. J Petrol 37: 11251152 Lee WJ, Wyllie PJ (1997a) Liquid immiscibility between nephelinite and carbonatite from 2.5 to 1.0 GPa compared with mantle melt compositions. Contrib Mineral Petrol 127: 116 Lee WJ, Wyllie PJ (1997b) Liquid immiscibility in the join NaAlSiO4-NaAlSi3O8-CaCO3 at 1.0 GPa: implications for crustal carbonatites. J Petrol 38: 11131135 Lee WJ, Wyllie PJ (1998) Petrogenesis of carbonatite magmas from mantle to crust, constrained by the system CaO-(MgO FeO)-(Na2O K2O)-(SiO2 Al2O3 TiO2)CO2. J Petrol 39: 495517

Calciocarbonatite and magnesiocarbonatite rocks and magmas

255

Lee WJ, Wyllie PJ (2000) The system CaO-MgO-SiO2-CO2 at 1 GPa, metasomatic wehrlites, and calciocarbonatite magmas. Contrib Mineral Petrol (in press) Lee WJ, Wyllie PJ, Rossman GR (1994) CO2-rich glass, round calcite crystals and no liquid immiscibility in the system CaO-SiO2-CO2 at 2.5 GPa. Am Mineral 79: 11351144 Macdonald R, Kjarsgaard BA, Skilling IP, Davies GR, Hamilton DL, Black S (1993) Liquid immiscibility between trachyte and carbonate in ash ow tuffs from Kenya. Contrib Mineral Petrol 114: 276287 Mariano AN, Roeder PL (1977) Kerimasi: a neglected carbonatite volcano. J Geol 91: 449 455 Otto JW, Wyllie PJ (1993) Relationships between silicate melts and carbonate-precipitating melts in CaO-MgO-SiO2-CO2-H2O at 2 kbar. Mineral Petrol 48: 343365 Piwinskii AJ, Wyllie PJ (1968) Experimental studies of igneous rock series: a zoned pluton in the Wallowa Batholith, Oregon. J Geol 76: 205234 Streckeisen AL (1980) Classication and nomenclature of volcanic rocks, lamprophyres, carbonatites and melilitic rocks. IUGS Subcommission on the Systematics of Igneous Rocks. Geol Rundsch 69: 194207 Treiman AH, Essene EJ (1984) A periclase-dolomite-calcite carbonatite from the Oka complex, Quebec, and its calculated volatile composition. Contrib Mineral Petrol 85: 149157 Tuttle OF, Gittins J (1966) Carbonatites. Interscience, New York, 591pp Wallace ME, Green DH (1988) An experimental determination of primary carbonatite magma composition. Nature 335: 343346 Walter LS, Wyllie PJ, Tuttle OF (1962) The system MgO-CO2-H2O at high pressures and temperatures. J Petrol 3: 4964 Watkinson DH, Wyllie PJ (1971) Experimental study of the join NaAlSiO4-CaCO3-H2O and the genesis of alkalic rock-carbonatite complexes. J Petrol 12: 357378 White BS, Wyllie PJ (1992) Solidus reactions in synthetic lherzolite-H2O-CO2 from 2030 kbar, with applications to melting and metasomatism. J Volcanol Geotherm Res 50: 117130 Woolley AR (1982) A discussion of carbonatite evolution and nomenclature, and the generation of sodic and potassic fenites. Mineral Mag 46: 1317 Woolley AR, Kempe DRC (1989) Carbonatites: nomenclature, average chemical compositions, and element distribution. In: Bell K (ed) Carbonatites: genesis and evolution. Unwin Hyman, London, pp 114 Wyllie PJ (1965) Melting relationships in the system CaO-MgO-CO2-H2O with petrological applications. J Petrol 6: 101123 Wyllie PJ (1966) Experimental studies of carbonatite problems: the origin and differentiation of carboibatite magmas. In: Tuttle OF, Gittins J (eds) Carbonatites. John Wiley, New York, pp 311352 Wyllie PJ (1978) Mantle uid compositions buffered in peridotite-CO2-H2O by carbonates, amphibole, and phlogopite. J Geol 86: 687713 Wyllie PJ (1989) Origin of carbonatites: evidence from phase equilibrium studies. In: Bell K (ed) Carbonatites: genesis and evolution. Unwin Hyman, London, pp 500545 Wyllie PJ, Boettcher AL (1969) Liquidus phase relationships in the system CaO-CO2-H2O to 40 kilobars pressure with petrological applications. Am J Sci 267A: 489508 Wyllie PJ, Huang WL (1975) Peridotite, kimberlite, and carbonatite explained in the system CaO-MgO-SiO2-CO2. Geology 3: 621624 Wyllie PJ, Huang WL (1976) Carbonation and melting reactions in the system CaO-MgOSiO2-CO2 at mantle pressures with geophysical and petrological applications. Contrib Mineral Petrol 54: 79107

256 W.-J. Lee et al.: Calciocarbonatite and magnesiocarbonatite rocks and magmas Wyllie PJ, Lee WJ (1998) Model system controls on conditions for formation of magnesiocarbonatite and calciocarbonatite magmas from the mantle. J Petrol 39: 1885 1893 Wyllie PJ, Tuttle OF (1960) The system CaO-CO2-H2O and the origin of carbonatites. J Petrol 1: 146 Zhabin AG, Cherepivskaya GY (1965) Carbonatite dikes as related to ultrabasic extrusive igneous activity. Dokl Acad Nauk SSR Earth Science Section 160: 135138 Authors' address: W.-J. Lee, M. F. Fanelli, and P. J. Wyllie, Division of geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, U.S.A., e-mail: wyllie@gps.caltech.edu

Potrebbero piacerti anche