Liquid Immiscibility in the Join NaAlSiO 4 – NaAlSi 3 O 8 – CaCO 3 at 1 GPa : Implications for Crustal Carbonatites

The synthetic system Na2O–CaO–Al2O3–SiO2–CO2 has been parent (at crustal pressures) contain a maximum of 80 wt % CaCO3. There are two relevant paths for a silicate liquid which widely used as a model to show possible relationships among alkalic silicate magmas, calciocarbonatites, and natrocarbonatites. The exsolves carbonate-rich liquid (along with silicate mineral precipitates): (1) the assemblage is joined by calcite, or (2) the determined immiscibility between silicateand carbonate-rich liquids has been strongly advocated to explain the formation of natural assemblage persists without carbonate precipitation until all silicate liquid is used up. The phase diagrams indicate that high-temperature carbonatite magmas. Phase fields intersected at 1·0 GPa by the composition joins NaAlSiO3O8–CaCO3 (Ab–CC, published) and immiscible carbonate-rich liquids must be physically separated from parent silicate liquid before they can precipitate carbonate-rich NaAlSiO4(Ne)90Ab10–CC (new), along with measured immiscible liquid compositions, provide pseudoternary phase relationships for mineral assemblages. Path (1) then corresponds to the silicate–calcite liquidus field boundary, and a stage is reached where the carbonatethe composition triangles Ab–CC–Na2CO3(NC) and Ne90Ab10– CC–NC. Interpolation between these, and extrapolation within rich liquids will precipitate large amounts of calcite and fractionate toward alkali carbonates (not necessarily matching natrocarbonatite the CO2-saturated tetrahedron Al2O3–SiO2–CaO–Na2O, provides pseudoquaternary phase relationships defining the volume for the compositions). In path (2) the high-temperature immiscible carbonate liquid precipitates only silicates through a temperature interval until miscibility gap and the surface for the silicate–carbonate liquidus field boundary. The miscibility gap extends between 10 and 70 wt it reaches the silicate–carbonate liquidus field boundary, where it may precipitate calcite or nyerereite or gregoryite. Sövites are readily % CaCO3 on the triangle Ne–Ab–CC at 1·0 GPa; it does not explained as cumulates, with residual alkali-rich melts causing extend to the Na2O-free side of the tetrahedron. The liquidus fenitization. We can see no way in phase diagrams for vapor loss minerals in equilibrium with both silicateand carbonate-rich to remove alkalis and change immiscible natrocarbonatite liquids to consolute liquids are nepheline, plagioclase, melilite, and wolCaCO3-rich liquids; adjustments to vapor loss would be made not lastonite; with increasing Si/Al the liquidus for calcite reaches the by change in liquid composition but by precipitation of calcite and miscibility gap. We use these phase relationships to: (1) illustrate silicate minerals. The processes illustrated in this model system are possible paths of crystallization of initial CO2-bearing silicate applicable to a wide range of magmatic conditions, and they haplomagmas, (2) place limits on the compositions of immiscible complement and facilitate interpretation of phase relationships in carbonatite magmas which can be derived from silicate parent the single paths represented by each whole-rock phase equilibrium magmas, and (3) illustrate paths of crystallization of carbonatite study. magmas. Cooling silicate–CO2 liquids may reach the miscibility gap, or the silicate–calcite liquidus field boundary, or terminate at a eutectic precipitating silicates and giving off CO2. Silicate–CO2 liquids can exsolve liquids ranging from CaCO3-rich to alkalic carbonate compositions. There is no basis in phase relationships for the occurrence of calciocarbonatite magmas with ~99 wt % CaCO3;


RECEIVED NOVEMBER 8, 1996; REVISED TYPESCRIPT ACCEPTED APRIL 11, 1997
The synthetic system Na 2 O-CaO-Al 2 O 3 -SiO 2 -CO 2 has been parent (at crustal pressures) contain a maximum of 80 wt % CaCO 3 .There are two relevant paths for a silicate liquid which widely used as a model to show possible relationships among alkalic silicate magmas, calciocarbonatites, and natrocarbonatites.The exsolves carbonate-rich liquid (along with silicate mineral precipitates): (1) the assemblage is joined by calcite, or (2) the determined immiscibility between silicate-and carbonate-rich liquids has been strongly advocated to explain the formation of natural assemblage persists without carbonate precipitation until all silicate liquid is used up.The phase diagrams indicate that high-temperature carbonatite magmas.Phase fields intersected at 1•0 GPa by the composition joins NaAlSiO 3 O 8 -CaCO 3 (Ab-CC, published) and immiscible carbonate-rich liquids must be physically separated from parent silicate liquid before they can precipitate carbonate-rich NaAlSiO 4 (Ne) 90 Ab 10 -CC (new), along with measured immiscible liquid compositions, provide pseudoternary phase relationships for mineral assemblages.Path (1) then corresponds to the silicate-calcite liquidus field boundary, and a stage is reached where the carbonate-the composition triangles Ab-CC-Na 2 CO 3 (NC) and Ne 90 Ab 10 -CC-NC.Interpolation between these, and extrapolation within rich liquids will precipitate large amounts of calcite and fractionate toward alkali carbonates (not necessarily matching natrocarbonatite the CO 2 -saturated tetrahedron Al 2 O 3 -SiO 2 -CaO-Na 2 O, provides pseudoquaternary phase relationships defining the volume for the compositions).In path (2) the high-temperature immiscible carbonate liquid precipitates only silicates through a temperature interval until miscibility gap and the surface for the silicate-carbonate liquidus field boundary.The miscibility gap extends between 10 and 70 wt it reaches the silicate-carbonate liquidus field boundary, where it may precipitate calcite or nyerereite or gregoryite.So ¨vites are readily % CaCO 3 on the triangle Ne-Ab-CC at 1•0 GPa; it does not explained as cumulates, with residual alkali-rich melts causing extend to the Na 2 O-free side of the tetrahedron.The liquidus fenitization.We can see no way in phase diagrams for vapor loss minerals in equilibrium with both silicate-and carbonate-rich to remove alkalis and change immiscible natrocarbonatite liquids to consolute liquids are nepheline, plagioclase, melilite, and wol-CaCO 3 -rich liquids; adjustments to vapor loss would be made not lastonite; with increasing Si/Al the liquidus for calcite reaches the by change in liquid composition but by precipitation of calcite and miscibility gap.We use these phase relationships to: (1) illustrate silicate minerals.The processes illustrated in this model system are possible paths of crystallization of initial CO 2 -bearing silicate applicable to a wide range of magmatic conditions, and they haplomagmas, (2) place limits on the compositions of immiscible complement and facilitate interpretation of phase relationships in carbonatite magmas which can be derived from silicate parent the single paths represented by each whole-rock phase equilibrium magmas, and (3) illustrate paths of crystallization of carbonatite study.magmas.Cooling silicate-CO 2 liquids may reach the miscibility gap, or the silicate-calcite liquidus field boundary, or terminate at a eutectic precipitating silicates and giving off CO 2 .Silicate-CO 2 liquids can exsolve liquids ranging from CaCO 3 -rich to alkalic carbonate compositions.There is no basis in phase relationships for the occurrence of calciocarbonatite magmas with ~99 wt % CaCO 3 ; KEY WORDS: calciocarbonatite; join NaAlSiO 4 -NaAlSi 3 O 8 -CaCO 3 ; liquid immiscibility; natrocarbonatite carbonate liquids derived by immiscibility from a silicate-CO 2

INTRODUCTION
The simplest synthetic system representing extrusive natrocarbonatites, intrusive calcic carbonatites, and associated alkalic igneous rocks is the five-component system Na 2 O-CaO-Al 2 O 3 -SiO 2 -CO 2 .However, this system requires three dimensions for complete graphical representation with excess CO 2 (Fig. 1).The utility of phase relationships in such model systems as a guide to possible processes in natural systems, and for interpretation of phase equilibrium studies of more complex rock systems, was recently discussed and justified by Lee & Wyllie (1996).The topology of liquidus surfaces and field boundaries elucidates possible processes, although the precise details vary, as the relative positions of field boundaries vary with pressure (Lee & Wyllie, 1996, 1997), and with composition [e.g. with peralkalinity (Kjarsgaard & Peterson, 1991) and with added MgO (Lee & Wyllie, 1997)].
interpreted their experiments and presented a similar miscibility gap. Lee & Wyllie (1994) showed that with H 2 O added to the join Ne 90 Ab 10 -CC, the liquidus temperature was  Hamilton (1988) that a mis-CaO-Na 2 O. Figure 1 shows the compositions of relevant cibility gap in the dry system straddled their composition silicate and carbonate minerals, and the bold triangle join.shows the composition joins Ab-CC and Ne-CC.The Koster van Groos & Wyllie (1966Wyllie ( , 1968Wyllie ( , 1973) explored the phase relationships on the liquidus surface around reader should note the different ratios of Al 2 O 3 /SiO 2 in dehydrated at 800°C for 4 h.Starting mixtures of these components were ground with ethanol in an agate mortar for 1 h.All starting mixtures (~5 mg for each run) were held for >1 h in a vacuum drying oven at 110°C before being loaded into platinum capsules ~4 mm long.

THE SYSTEM Na
Experiments were undertaken using 1•27-cm diameter piston-cylinder apparatus, with calcium fluoride as a pressure medium.Temperature was controlled and monitored by a W 95 Rh 5 -W 74 Rh 26 thermocouple with no correction for the effects of pressure on the e.m.f.Run duration varied from 1 to 12 h.Experiments were terminated by turning off the electrical power while the system remained near isobaric, and the quenching rate was ~100°C/s for the first 600°C.Pressure accuracy is about ±0•05 GPa, and temperature accuracy is estimated to be ±10°C.No pressure correction was made for the effects of friction.
Run products were mounted in a brass holder with indicate the liquidus fields for albite, wollastonite, and calcite.Liquidus phologies observed under the petrographic microscope piercing points p and q along Ab-CC were determined by Lee & and a Camscan scanning electron microscope fitted with Wyllie (1996), and the eutectic e between Wo and CC by Huang & an energy dispersive X-ray system (EDS).Phases were Wyllie (1974).Curves e-p-I 5 and I 5 ′-o indicate the compositions of liquids which coprecipitate silicate and carbonate minerals with vapor.analyzed by EDS using a beam current of 0•1 nA as Curve m-I 5 -k-I 5 ′-n delineates the limit of the miscibility gap where measured on brass.
silicate-and carbonate-rich liquids coexist with a mineral and vapor.k is the critical point where Ls=Lc.I 5 and I 5 ′ are pseudoternary isobaric invariant points for the assemblage Wo+CC+Ls+Lc+V.
the miscibility gap, but most subsequent studies presented EXPERIMENTAL RESULTS the immiscible liquids corresponding to an isothermal On the basis of the runs listed in Table 1, the phase miscibility gap. Lee & Wyllie (1992a, b, c) emphasized fields intersected by the composition join Ne 90 Ab 10 -CC that the existence of a miscibility gap did not mean that at 1•0 GPa are shown in Fig. 3.The interpretations of all carbonatites were immiscible fractions.The liquidus phases and textures are illustrated by the examples in paths followed by silicate-CO 2 magmas could (1) reach Fig. 4. The phases encountered include nepheline (Ne), a miscibility gap, (2) reach a silicate-carbonate liquidus anorthite (An), scapolite (Scap), melilite (Mel), calcite field boundary and coprecipitate silicates and carbonates, (CC), silicate-rich liquid (Ls), carbonate-rich liquid (Lc), or (3) terminate at a eutectic precipitating silicates with and vapor (V).As minerals were precipitated during the evolution of CO 2 vapor.The results in Fig. 2 illustrate quench, detailed study was required to interpret some of the tentative positions of the silicate-carbonate (e-p-I 5 ) the phase assemblages.Representative compositions of and albite-wollastonite (through q) liquidus field boundimmiscible silicate-carbonate liquid pairs are listed in aries.The new results in this contribution permit a more Table 2 and illustrated in Fig. 5.The positions of dashed complete construction of silicate-silicate field boundaries lines are uncertain, but they provide an internally conin the liquidus volume between Ab-Ne-CC and sistent interpretation of the available data.The ex-(Al 2 O 3 +SiO 2 ) in Fig. 1, providing additional constraints periments were aimed at determination of the liquidus on the conditions leading to each of the three processes relationships, the piercing points on either side of the outlined above.miscibility gap (P, Q), and the points separating different liquidus minerals (R, S, T).These points were determined partly on the basis of runs plotted in Fig. 3, and partly

EXPERIMENTAL METHODS
on tie-lines connecting measured liquid compositions (S, T, Fig. 5), with additional constraints required to com-The starting materials were: (1) primary standard grade plete the phase diagrams in Figs 6, 7 and 8. P and S are CaCO 3 powders (Alfa Product), dried at 110°C for at closely determined, and the possible error in Q and R least 1 day; (2) synthetic glass of 90 wt % NaAlSiO 4 and 10 wt % NaAlSi 3 O 8 (prepared by Koster van Groos), is discussed below.to a glass (Fig. 4a), but at lower temperatures the glass in Fig. 3 have been interpreted in terms of the phase contained quench nepheline (Fig. 4b).Carbonate-rich boundaries and phase assemblages as shown.The presliquids always formed calcite dendrites (~several m; Fig. ence of vapor is evident by the appearance of pore space 4b-d), with nyerereite and thin lamellae of silicate-rich on the top portion of a capsule, and bubbles trapped in material (Fig. 4f ).quenched liquids; however, complex quench textures in The upper-temperature stability limits for calcite and some lower-temperature experiments (e.g.148 and 118 nepheline are fairly well defined (Fig. 3, continuous lines).at 1150°C) made positive identification of vapor difficult.
The melting temperature of calcite at 1 GPa is 1480°C Liquid was produced in all experiments, and quenched to a variety of textures depending on composition and (Irving & Wyllie, 1975).Neither the fusion temperature   The bold lines (Fig. 3, solid and dashed) bounding the miscibility gap are based on the runs and on analyses of coexisting quenched liquids.The slope of the silicaterich limb above 1400°C is arbitrarily drawn, roughly extended from the section below 1400°C to point P. The carbonate-rich limb above Q is sketched with a negative slope symmetric to the silicate-rich counterpart, reflecting that the size of the hyper-liquidus miscibility gap decreases with increasing temperature (e.g.Freestone & Hamilton, 1980;Lee & Wyllie, 1996).There is no evidence for immiscible liquids in runs at 1200°C and below, indicating that the boundary of the miscibility gap closes with decreasing temperature, as indicated by the bold dashed line swinging towards the Ne-rich direction at polite, and anorthite are close to their ideal compositions (Geh, Scap, and An, respectively, in Fig. 1).Melilite commonly formed large prismatic crystals (lightest color of nepheline nor the melting relationships of the join NaAlSiO 4 -NaAlSi 3 O 8 have been determined ex-crystals, Fig. 4d and e) with sizes up to 100×30 m 2 , and contained a few percent of Na 2 O, representing solid perimentally at 1 GPa.The congruent melting temperature of Ne 90 Ab 10 in Fig. 3 is estimated to be ~1620°C, solution toward soda melilite.Run 121 was the only experiment containing scapolite, commonly with in-using published 1 atm data (Greig & Barth, 1938) extrapolated by 110°C corresponding to the pressure in-clusions of carbonate-rich melt, occurring as rectangular crystals up to 30 m in the layer of quenched carbonate-crease [for pure NaAlSi 3 O 8 as determined by Boyd & England (1963); see also Bell & Roseboom (1965)].
There are also some minute crystals of nepheline (several m across) near the large nepheline, not discernible in the picture.Nepheline contains variable amounts of CaO which increase with changing bulk composition towards CaCO 3 , from <1 wt % to ~6 wt % CaO.Nepheline compositions project onto the line between Ne and An in Fig. 1, suggesting solid solutions dominantly with anorthite.Bowen (1912) reported that at 1 atm nepheline could dissolve as much as 35 wt % anorthite, with CaO equivalent to 7 wt %.
The calcite liquidus also drops steeply to piercing point R for the calcite-melilite field boundary (near 1325°C, 85 wt % CaCO 3 ).Calcite in this join always formed a distinct layer at the lower portion of a charge, as shown in Fig. 4d, and was commonly intergrown with nepheline, melilite or anorthite.The calcite crystals exhibited a hexagonal habit, in contrast with the rounded calcite crystals reported in many silicate-carbonate systems (see Huang et al., 1980;Lee et al., 1994).The calcite-precipitating fields are first intersected by the melilite liquidus at R, then joined by scapolite, anorthite, and finally nepheline at 1200°C and below.Contrary to the results in the join Ab-CC (Lee & Wyllie, 1992a, b, 1996), the calcite liquidus here does not intersect the two-liquid field.
The melilite liquidus is intersected in a narrow compositional range between the miscibility gap and the calcite liquidus, Q to R, and completion of the phase diagram in Figs 6, 7 and 8 shows that the melilite-out boundary (Fig. 3) extends between the calcite liquidus 149).The fields 2, 4, and 7 in Fig. 3 indicate a small temperature interval where scapolite may be stable (assumed to cover a compositional range similar to that of (up to 30 m in length) within carbonate-rich liquids, or the melilite-out boundary).Below this interval the phase intergrown with nepheline, calcite or melilite (Fig. 4c assemblages are joined by anorthite through a wide and d).
compositional range.The scapolite-anorthite tem-The composition join intersects a series of piercing perature relationship is consistent with the results of points for liquidus field boundaries: P between the li- Goldsmith & Newton (1978), and we assume the field quidus surfaces for nepheline and silicate-rich immiscible boundary between the two stability fields to be a carliquid (Ls); Q between carbonate-rich immiscible liquid bonation-decarbonation relation.It should be noted that (Lc) and melilite; and R between melilite and calcite.between the Lc limb of the miscibility gap (dashed Points S and T within the miscibility gap are closely bold curve below Q) and the calcite field boundaries defined, as discussed below in connection with Figs 6 (continuous curve below R) are fields where carbonateand 7a.
rich liquids precipitate assemblages of silicate minerals, The nepheline liquidus drops steeply as CaCO 3 is joined by calcite at lower temperatures.added, down to the piercing point P (1335°C and 14 wt % CC), where it is joined by immiscible carbonate-

Compositions of immiscible liquids
rich liquid.Figure 4a illustrates unambiguous primary nepheline (outlined) from run 132, which forms large The compositions of seven pairs of immiscible liquids for the temperature interval 1250-1350°C are listed in Table crystal prisms up to 50×20 m 2 ; however, nepheline The open circles labeled 20, 30, 40, 60 correspond to below the values indicated on the starting join silicate-CaCO 3 , but liquids near Q and toward the silicate the projected starting mixtures from Fig. 3 with 20 wt % CaCO 3 , etc.It should be noted that the mixture '30' components become substantially lower in carbonate contents compared with the silicate-CaCO 3 join, being contains ~20 wt % CaO, and mixture '60' contains ~45 wt % CaO in the projection.The compositions of phase located between the join and the CO 2 -free silicate-CaO projection.boundaries and piercing points in Fig. 3 show liquid compositions in terms of wt % CaCO 3 , but the projected Runs 117, 119, 129, and 130 (Table 1, Fig. 3) yielded silicate glass, and runs 123, 127, and 145 (Table 1, analyses in Figs 5 and 6 show instead wt % CaO, consistently lower than the wt % CaCO 3 .The actual Fig. 3) contained silicate glass with quench nepheline.
Individual analyses for the quenched silicate-and car-carbonate compositions (wt %) remain unknown without measurements of the CO 2 content of pre-quench liquids.bonate-rich liquids from the first four runs are plotted in Fig. 5a, and from the other three in Fig. 5b.Averages The presence of vapor in all runs shows that the liquids in Fig. 3 Lee et al. (1994) determined the CO 2 solubilities in 6 and connected by tie-lines, and grouped in terms of temperature.parts of the system CaO-SiO 2 -CO 2 .The position of the CO 2 -saturated liquidus provides an indication of the Figure 5a shows three analyses for each liquid in four runs.The silicate analyses cluster closely, but the actual liquid composition.The geometrical relationship carbonate analyses show a wider spread.The quench bracketed at 14 wt % CaCO 3 .The calcite-out phase products of carbonate-rich liquids show large heteroboundary is closely defined, so the position of R depends geneity in calcite-nyerereite distribution (e.g.Fig. 4b, on the temperature of the melilite liquidus, which occurs 4f ).Areas up to 100 m 2 were analyzed by the EDS above 1300°C (run 120).The piercing point R must lie method, but the areas suitable for the rastering mode at a cotectic temperature minimum between the liquidus were still too small to compensate for the heterogeneity; fields for calcite and melilite (Fig. 8), requiring a temthe variation in analyses trends generally subparallel to perature maximum on the melilite liquidus between R the CaO-Na 2 O side of the projection.A line connecting and Q/T.We have assumed a maximum temperature each pair of average coexisting liquid compositions would of ~1350°C for the melilite liquidus.This provides a pass very close to the corresponding starting mixture, as cross-section with the calcite liquidus dropping through shown in Fig. 6.This is true even for run 119, which more than 150°C to reach the calcite-silicate field boundcontains some melilite and anorthite.
ary at R, which is consistent with several other calcite-Figure 5b shows similar analyses from the three runs silicate joins at several pressures (e.g.Maaloe & Wyllie, with quench nepheline in the glass.For the silicate liquid 1975; Huang et al., 1980;Lee & Wyllie, 1996).Given in each run, three types of analyses are plotted: the the steep liquidus surface for calcite, we conclude that open diamonds near Ne give compositions for quench the point R is close to 85 wt % CaCO 3 , with possible nepheline crystals; the open triangles are for the mineralerror of ±3 wt % CaCO 3 .The composition of Q in free glass; the filled diamonds are for larger areas (>1000 Fig. 3 is bracketed only between 60 wt % and 80 wt % m 2 ) including both glass and quench nepheline.The CaCO 3 , but it is also constrained because the 1300°C latter analyses represent the original liquid composition, isotherm in Fig. 6 must pass very close to Q.We place lying between the open diamonds and the squares as the position of Q at 72±8 wt % CaCO 3 , as indicated they should.The analyses of carbonate-rich liquids in for its projected position Q in Fig. 6.Despite the un-Fig.5b have a similar spread, and similar relationship certainties in Q and R, their relative positions in the phase to the silicate liquids as in Fig. 5a, although the range diagram cannot be changed, and hence the corresponding for run 127 is somewhat wider.Only one rastered analysis petrological conclusions remain unaffected by the errors.is given for run 145, covering ~1000 m 2 area.P, Q and R are plotted at 8 wt %, 59 wt % and 76 The average analyses for the seven pairs of coexisting wt % CaO, respectively, in the CO 2 -free projection in liquids, connected by tie-lines and identified in terms of Fig. 6.As discussed above, the actual liquids are situated temperatures, are plotted together in Fig. 6.A noteworthy between the Fig. 3  The experimental data obtained from the join Ab-1300°C for mixture '60'.The rotation of the tie-line for CaCO 3 (Lee & Wyllie, 1996) were presented in the 117 compared with the others could conceivably be a Hamilton projection (Fig. 1b) as pseudoternary phase temperature effect between 117 (1300°C) and 119 relationships (Fig. 2).Similarly, the new near-liquidus (1250°C) enhanced as the immiscible liquids approach experimental data obtained from the join Ne 90 Ab 10closure.
CaCO 3 can be treated as pseudoternary in projection, Small differences are distinguishable as a function of because the immiscible liquid compositions (molecular temperature.The short arcs connect analyses of quenched Al/Si of silicate-rich liquids ~0•95; Table 2) are very liquids at the three temperatures 1350°C, 1300°C and close to the triangular slice CaO-Na 2 O-Ne 90 Ab 10 (Al/ 1250°C, and thus represent portions of isotherms on the Si=0•9) extending through the excess-CO 2 tetrahedron liquidus surface of the miscibility gap, which becomes (Fig. 1).The data in Figs 3 and 6 are first combined to larger with decreasing temperature.The data points define the miscibility gap and liquidus field boundaries overlap, but the isotherms must be separate, as drawn.
around it, along with the primary minerals along the These partial isotherms, along with other definitive points field boundaries (Fig. 7).Sequences of crystallization from Fig. 3, permit construction of part of the liquidus illustrated in Fig. 3 are then combined with published surfaces and field boundaries intersected by the comor inferred data on bounding systems to begin mapping position join Ne 90 Ab 10 -CC.
out the field boundaries separating primary phase fields Additional liquid compositions are given by the points for the CO 2 -saturated silicate and carbonate liquidus P, Q and R from Fig. 3.These represent liquids for piercing points on liquidus field boundaries.P is closely surfaces (Fig. 8).isotherms on the miscibility gap.They also define the The pseudoternary system through limits for the liquidus minerals nepheline, anorthite, Ne 90 Ab 10 -CaCO 3 -Na 2 CO 3 melilite, anorthite, and nepheline.It should be noted The experimental data on phase fields (Fig. 3) and liquid that the primary mineral on the miscibility gap field compositions (Figs 5 and 6) permit construction of parts boundary for carbonate-rich liquids from I 1 ′ to near of the vapor-saturated liquidus surface for the miscibility Na 2 O is nepheline, not a carbonate [see Fig.There must be a critical point where Ls=Lc on respectively (Figs 3 and 6).The bold continuous line the miscibility gap field boundary between I 2 and I 2 ′, through P and Q is our estimate of the position of the associated with the liquidus for melilite.The distribution field boundary enclosing the miscibility gap, similar to of isotherms in Fig. 7 indicates that this is also a temthe topology in Fig. 2, with justification following.The perature maximum, which is consistent with the tentative silicate side of the field boundary is extended to terminate dashed lines in Fig. 3.We have estimated K near Q, at arbitrarily on the Ne-Na 2 O side at M, based on previous temperature near 1330°C.results in the Ca-free system from Koster van Groos & Figure 8 shows the completed pseudoternary phase Wyllie (1966Wyllie ( , 1973) ) and Kjarsgaard & Hamilton (1988, diagram, analogous with Fig. 2, showing the three major 1989) (see Lee & Wyllie, 1996; Figs 2 and 8), and the features, the miscibility gap (line-shaded two-liquid field), carbonate side is extended to show a narrow field for the silicate liquidus surface (gray), and the carbonate carbonate-rich liquids extending from CaO to Na 2 O (Lee liquidus surface (dotted).The miscibility gap is defined by & Wyllie, 1996; Figs 2 and 8).
the liquidus field boundaries and pseudoternary isobaric The isotherms for the miscibility gap liquidus given in invariant points transferred from Fig. 7: M-Fig.6 are more or less concentric with the limiting field I 1 -I 2 -K-I 2 ′-I 1 ′-N.The region near the CaO-Na 2 O side boundary, terminating on it, or remaining within it, and of the projection (I 1 ′ and I 2 ′) is distorted compared with fitted with the temperatures of P and Q, as well as with Fig. 7, to show some details for the carbonate-rich liquids.the temperatures of the additional tie-lines shown passing The primary liquidus minerals nepheline, anorthite, and through S and T. Nepheline and melilite are the primary melilite, on the silicate-rich field boundary M-K, must minerals on the silicate liquidus surface between P and be repeated along the carbonate-rich field boundary Ne, and between Q and R, respectively, and calcite K-N.Therefore, no primary carbonate mineral can be (R-CaO) is the primary mineral on the carbonate liquidus precipitated from the carbonate-rich liquids, Lc, which surface (Figs 3 and 7).
remain in equilibrium with the silicate-rich liquids, Ls. Figure 3 shows anorthite on the liquidus between Crystallization directions along parts of the boundaries nepheline and melilite (S-T).The anorthite liquidus is are indicated by arrows.therefore adjacent to the field boundary between the two On the silicate liquidus surface, several curves are pairs of pseudoternary isobaric invariant points I 1 -I 2 , and sketched for the field boundaries between Ne and An, I 1 ′-I 2 ′ (open squares) in Fig. 7.The tie-lines I 1 -I 1 ′ and and between An and Mel.For the silicate-rich liquids, I 2 -I 2 ′ must pass through the points S and T (Fig. 3), as the curves are largely schematic, starting at somewhat shown in Fig. 7.The corresponding phase assemblages arbitrary eutectics in the joins Ne-An and An-Mel, and are I 1 -S-I 1 ′ (1230°C): Ne+An+Ls+Lc+V; and I 2terminating at the better-constrained points I 1 -I 1 ′ and T-I 2 ′ (1270°C): An+Mel+Ls+Lc+V.The precise po-I 2 -I 2 ′.For the carbonate-rich liquids (see the enlarged sitions of these six points are uncertain, but the constraints inset diagram), two field boundaries are defined between are fairly close.The error bars in terms of composition I 1 ′ and I 3 , and I 2 ′ and I 4 , respectively, where I 3 and I 4 and temperature are within our normal brackets of are pseudoternary isobaric invariant points for 10 wt % CaCO 3 and 50°C.The tie-lines must be Ne+An+CC+Lc+V and An+Mel+CC+Lc+V.approximately parallel to the other tie-lines (Fig. 6).Given Points I 3 and I 4 are plotted close to the measured comthese constraints, there is little variation possible for the positions of carbonate-rich liquids in runs 148 and 121 positions of I 1 , I 1 ′, I 2 , I 2 ′, S and T. The positions of S with similar phase assemblages (Table 1, Fig. 3).The and T in Fig. 3 are drawn to correspond to the points arrangement of field boundaries and arrows in Fig. 8 constructed in Fig. 7.
(inset) is consistent with the sequences of crystallization The estimated invariant points I 1 , I 2 , I 2 ′ and I 1 ′ provide observed for liquids with compositions between Q and R in Fig. 3, neglecting the complication of possible temperature constraints for the terminations of liquidus primary scapolite (phase fields 2, 4 and 7 in Fig. 3).
The compositional differences between the two experimental joins have been noted with respect to Fig. 1a, There is probably a thermal divide on the melilite liquidus and the triangular slices drawn in Fig. 10a show the surface extending from the temperature maximum at difference clearly in terms of Al 2 O 3 /SiO 2 .The CO 2 -free critical point K.
projected phase relationships intersected by the tetra-The silicate-carbonate liquidus field boundary starts hedron may be treated as pseudoquaternary.The from a postulated eutectic E between Mel and CC on Ne 90 Ab 10 -CC-NC plane has Al/Si=0•9, and the Abthe (Al 2 O 3 +SiO 2 )-CaO baseline (compare e in Fig. 2 1a isobaric invariant points for assemblages of two liquids, which include Ne 90 Ab 10 (Fig. 8) and Ab (Fig. 2) intersect three minerals, and vapor, such as Plag+ the same major liquidus fields (the silicate-carbonate Wo+Mel+Ls+Lc+V at I 7 (open circle).The critical miscibility gap, silicate and carbonate liquidus surfaces), curve through K-k is the locus of critical liquids where but with some significant differences.The key features Ls=Lc, dividing the miscibility gap liquidus surface into of Figs 2 and 8 are compared in Fig. 9, which shows the consolute liquids Ls occupying a large area to the left, lower parts of the diagrams with vertical exaggeration of and Lc to the right concentrated near the CaO corner of two.In Figs 8 and 9a, the silicate-carbonate liquidus the tetrahedron and the axis CaO-Na 2 O, with maximum field boundary does not intersect the miscibility gap field 75-80 wt % CaO (Figs 2 and 8).This corresponds to boundary, and the miscibility gap is therefore surrounded no more than ~80 wt % dissolved CaCO 3 (assuming all by the liquidus surface for primary silicates.In Figs 2 CaO and Na 2 O assigned to carbonate; see Lee & Wyllie,and 9b, the silicate-carbonate liquidus field boundary 1996).It should be noted in Fig. 8 that the liquidus does intersect the miscibility gap field boundary, bringing field boundary for coprecipitation of melilite and calcite the carbonate liquidus surface into contact with the without liquid immiscibility can yield a liquid with ~85 miscibility gap between the isobaric pseudoinvariant wt % CaO (near I 4 ), or at most ~86 wt % CaCO 3 .The points I 5 and I 5 ′.
consolute liquids Ls and Lc do not correspond to liquidus The liquidus surfaces for silicates involve different surfaces with primary minerals silicates and carbonates, minerals and field boundaries, as expected given silicate respectively.Liquidus fields for both silicates and carend-members which are nepheline-normative (Figs 8 and bonates occur on both sides of the critical curve, Ls= 9a), or albite-normative (Figs 2 and 9b).In Fig. 9b, the Lc.This feature was emphasized by Lee & Wyllie (1996) albite-wollastonite field boundary has been extended in connection with the position of k and the silicate and through the determined point q (Fig. 2) to an estimated carbonate liquidus surfaces in Fig. 2. invariant point I 6 on the miscibility gap field boundary, The field boundaries sketched in Fig. 10b connect the which requires another tie-line I 6 -I 6 ′ for the pseudodetermined points, and locate the liquidus fields for invariant phase assemblage Ab+Wo+Ls+Lc+V.The immiscible liquids coexisting with nepheline, plagioclase, slope of the tie-line is comparable with those defined by melilite, wollastonite or calcite.The area 'Plag' indicates the alkali-rich two-liquid pairs in Kjarsgaard & Hamilton the field of two liquids with plagioclase changing composition continuously from anorthite to albite as the bulk (1989).changes in the direction of increasing Al/Si (from the Ab join toward the Ne join), the length of the intersection Figure 11 summarizes the main pseudoquaternary phase relationships based on the two pseudoternary dia-(along the field boundary for CC+Ls+Lc+V, Fig. 9b) is reduced as I 5 and I 5 ′ migrate toward the critical point grams, with possible cooling directions indicated on some of the field boundaries.The light-shaded surface for the k.For the composition with slightly higher Al/Si than that where I 5 and I 5 ′ become coincident on the critical miscibility gap is reproduced from Fig. 10b, and a sketch of the silicate-carbonate liquidus field boundary surface curve, the silicate-carbonate liquidus surface becomes separated from the miscibility gap, and carbonates do (dark-shaded) is based on Figs 2, 8, and 9.These two surfaces meet in the wollastonite-calcite field boundary not coexist with the immiscible liquids, which is the situation in Fig. 9a. on the miscibility gap surface (Fig. 10b).The silicate-Fig.11.Compositional tetrahedron showing the three major liquidus features, the miscibility gap (in Fig. 10b), silicate liquidus volume, and carbonate liquidus volumes.Possible cooling directions for some field boundaries are indicated by arrows.F is the eutectic e for the join Wo-CC in Fig. 2, and G is a hypothetical eutectic between Al 2 O 3 and CaCO 3 , located at ~30 wt % CaO.The carbonate liquidus volume near the CaO corner is for calcite, and its range becomes more limited at higher alkali contents, involving sodic carbonate minerals (e.g.nyerereite, Na 2 CO 3 ).
pseudoternary system in Fig. 2, and in Fig. 9b we extend the Ab-Wo liquidus field boundary to eutectic I 6 .Here we compare the 1•0 GPa conclusions (from Figs 2 and

CRYSTALLIZATION PATHS OF
(1) reach the field boundary m-I 5 and exsolve im-CARBONATED SILICATE LIQUIDS IN miscible carbonate-rich liquid Lc (n-I 5 ′), then cool with MODEL SYSTEM composition changing toward eutectic I 6 while the car- Lee & Wyllie (1996) reviewed possible crystallization bonate-rich immiscible liquid similarly changes toward paths of carbonated silicate liquids using the phase re-I 6 ′; lationships intersected by Ab-CC at 2•5 and 1•0 GPa (2) reach the silicate-carbonate field boundary Wo-CC emphasizing the variations with composition (nephelineand coprecipitate wollastonite and calcite until the liquid or quartz-normative, Ca/Na) and pressure.The evolution reaches I 5 , where carbonate-rich liquid I 5 ′ is exsolved; of carbonated silicate liquids towards carbonatitic residua liquid I 5 will then change composition to the eutectic at is constrained not only by the liquid miscibility gap, but I 6 , exsolving Lc changing from I 5 ′ to I 6 ′; also by the silicate-carbonate liquidus field boundary.
(3) terminate at eutectics on the silicate-CO 2 liquidus with precipitation of silicate minerals and evolution of Lee & Wyllie (1996) considered these limits for the VOLUME 38 NUMBER 9 SEPTEMBER 1997 CO 2 .There are insufficient experimental data to evaluate Exsolved carbonate-rich liquids with compositions between N and K would not precipitate carbonates as long this possibility.
The most calcic carbonate-rich liquid that could be as they remained in equilibrium with the silicate-rich liquids.Crystallization paths of the liquids Lc when produced is I 5 ′ (70 wt % CaO in Fig. 2; maximum 73 wt % CaCO 3 ), but the silicate liquids trend towards I 6 , removed from the silicate liquids would pass down the steep liquidus surface with precipitation of combinations which produces a much more sodic carbonate liquid I 6 ′.There are no crystallization paths for silicate-CO 2 liquids of nepheline, anorthite and melilite until they reached the silicate-carbonate liquidus field boundary, I 4 -I 3 -O to produce carbonate-rich liquids between I 5 and I 5 ′.
The exsolved carbonate-rich liquids Lc along I 5 ′-I 6 ′-n in Figs 8 and 9a, where the silicate minerals would be joined by calcite.The liquid I 4 on the silicate-carbonate precipitate only silicates (Wo and Ab) as long as they remain in equilibrium with the silicate liquids (except at field boundary is richer in CaCO 3 than the immiscible liquids, containing ~85 wt % CaCO 3 .The crystallization I 5 ′, where wollastonite is joined by calcite).If Lc is removed from the coexisting Ls, then the liquid can cool path would then continue along I 3 -O toward residual liquids precipitating alkali carbonates such as nyerereite down the narrow, steep silicate liquidus surface I 5 ′-n-o until it reaches the silicate-carbonate field boundary (Cooper et al., 1975).I 5 ′-o, where a carbonate mineral is coprecipitated.These liquids will follow paths probably close to those determined in the carbonate system by Cooper et al. (1975) The pseudoquaternary system at lower pressures, with final liquids containing ap-Na 2 O-CaO-Al 2 O 3 -SiO 2 with excess CO 2 proximately equal proportions of CaCO 3 and Na 2 CO 3 .
The paths of silicate-CO 2 liquids in Fig. 9a and b discussed above assume ternary phase relationships.In fact, although the liquidus relationships are close to The pseudoternary system through ternary, liquid paths diverge from the triangular planes Ne 90 Ab 10 -CaCO 3 -Na 2 CO 3 when silicate minerals are precipitated.Therefore, crystallization paths are better described in the CO 2 -sat-The major difference between the pseudoternary systems based on Ab (Figs 2 and 9b) and Ne (Figs 8 and 9a) is urated, pseudoquaternary system Na 2 O-CaO-Al 2 O 3 -SiO 2 , as in Figs 10b and 11.Possible field bound-that the silicate-carbonate field boundary in the latter system does not intersect the miscibility gap field bound-aries for immiscible liquids on the shaded surface show the transition between the conditions described above ary, with the consequence that the silicate liquidus surface completely surrounds the miscibility gap, and all im-for the two pseudoternary systems (Fig. 10b).Volumes, surfaces and lines extend into the tetra-miscible consolute liquids, Ls and Lc, precipitate only silicate minerals.A second difference is that in the Ne-hedron from the areas, lines and points on the miscibility gap surface (Fig. 11).These phase relationships are based system the silicate-carbonate field boundary passes much closer to the CaO corner than in the Ab system, needed (but unknown) in order to trace in detail the paths of crystallization from original silicate-CO 2 melts.causing expansion of the silicate liquidus fields (melilite), and contraction of the calcite liquidus field (Fig. 9).
Original liquids in the silicate-CO 2 liquidus volume of Fig. 11 follow paths leading to similar residual products Let us consider the crystallization paths for silicate-CO 2 liquids with primary minerals Ne, An, or Mel.Liquid to those outlined above for the pseudoternary systems, with additional variations now apparent.Let us consider paths may: (1) reach the field boundary K-I 2 -I 1 -M, and exsolve silicate liquids with primary nepheline, plagioclase, wollastonite or melilite.immiscible carbonate-rich liquid Lc (K-I 2 ′-I 1 ′-N); the immiscible carbonate-rich liquids tend to be concentrated (1) These liquids may follow paths to the miscibility gap surface where carbonate-rich liquid is exsolved, and into the region of I 1 ′ (76 wt % CaO; maximum 77 wt % CaCO 3 ); they may migrate to the invariant point I 7 where the coexisting liquid Lc has composition I 7 ′, between I 2 ′ and (2) reach the field boundary for anorthite-melilite en route to the miscibility gap at I 2 , after which they may I 6 ′ in Figs 2, 8, and 9.
(2) There is no obvious path to the silicate-carbonate continue to I 1 , as in path (1) above; there is probably a thermal maximum across the melilite liquidus field from liquidus surface visible in Fig. 11.The path which appeared possible across the wollastonite liquidus in Figs 2 K to Mel, preventing the silicate-rich liquids from reaching the silicate-carbonate field boundary for melilite-and 9b appears to be precluded by the rising liquidus temperature from I 7 along the field boundary between calcite; (3) terminate at undetermined eutectics on the silicate-melilite and wollastonite.However, the possibility is real, as confirmed by the liquidus along the edge quartz-calcite CO 2 liquidus with precipitation of silicate minerals and evolution of CO 2 .
(Fig. 11; Lee et al., 1994). ( 3) The phase relationships associated with the field samples.We emphasize that although our model system compositions differ from those of natural rocks, the boundary extending from I 7 into the silicate liquidus volume provide conditions for liquids which solidify to topology of the main phase elements illustrates particular processes and insights about their controls.The phase yield silicate minerals and CO 2 without immiscible liquids or coprecipitation of calcite.
geometry and minerals vary as a function of pressure and composition (especially Mg/Ca).The formation of immiscible liquids Lc (except those on the field boundary between Wo and CC, Fig. 10b), We now consider the generation of haplocarbonatite magmas in the model system, selecting only the starting if they are separated from the host silicate liquid Ls, is followed by the precipitation of silicates only.The compositions corresponding to carbonated silicate sequence of crystallization is controlled by the armagmas (at temperatures on the liquidus, not above rangement of field boundaries associated with I 1 ′, I 2 ′, I 5 ′, it) which follow differentiation paths that intersect the I 6 ′ and I 7 ′ on the Lc surface, until the liquids reach the miscibility gap before they reach the silicate-carbonate silicate-carbonate liquidus surface to the right and rear liquidus boundary.They may follow two different types of Fig. 11.There, most of the liquids will coprecipitate of sequences of crystallization, outlined below in stages.calcite, but some original silicate liquids could yield Each stage involves a temperature interval: immiscible sodic liquids (e.g.I 6 ′) and coprecipitate ny-Stage 1: Ls+silicates.Ls is within the silicate liquidus erereite or gregoryite before calcite.volume (Fig. 11), represented in part by the silicate Of particular significance for the origin of carbonatite liquidus surfaces of Figs 2, 8 and 9.When Ls reaches magmas are the surfaces enclosing the liquidus volume the miscibility gap, the assemblage is joined by Lc. for the primary crystallization of calcite (and other car-Stage 2: Ls+Lc+silicates.Ls and Lc change compositions bonates).This volume is enclosed in Fig. 11 by the across the miscibility gap surface in Fig. 11, visualized sides of the tetrahedron extending from CaO, by the in part by the intersections of Ls in boundaries m-I 5 and silicate-carbonate liquidus surface, and by the lower M-K, and of Lc in boundaries n-I 5 ′ and N-K (Figs 2, portion of the miscibility gap surface with primary calcite 8 and 9).Crystallization of silicates and the exsolution (compare Fig. 10b).Silicate-CO 2 liquids may follow paths of Lc can be followed by either (a) the exhaustion of Ls to this surface by two routes: (1) precipitation of silicates, (or the fractional separation of Lc before Ls is used up), liquid immiscibility, continued precipitation of silicates, or (b) the precipitation of calcite, leading to Stage 3a or and finally coprecipitation of calcite and alkali carbonates 3b. with silicates; or (2) precipitation of silicates, and then Stage 3a: Lc+silicates.Liquid Lc will cool across the coprecipitation of calcite, which may be followed by silicate liquidus surface represented by the areas I 5 ′-n-o liquid immiscibility and paths corresponding to those in and K-I 2 ′-I 1 ′-N-O-I 3 -I 4 -E (Figs 2 and 8; not clearly (1).displayed in Fig. 11).; not clearly displayed in Fig. 11).This surface is of the phase fields.The phase diagrams also permit tests physically separated from the miscibility gap surface of the feasibility of hypotheses published during the except along the line of intersection of the two surfaces debates of the 1980s, which tended to be based on (between Wo and CC in Figs 10b and 11) corresponding petrography rather than phase diagrams, and on petrological imaginations as admirably fertile as some mantle to Stage 3b.Crystallization is dominated by calcite, except for original liquids with very high peralkalinity contained low but appreciable concentrations of Si, Ti, Al, Fe and Mg.No carbonate minerals were produced.(Kjarsgaard et al., 1995).In a natural system calcite would form cumulates (it sinks even in experimental capsules, The experimental results on mixtures of natural rocks with carbonates presented by Kjarsgaard & Peterson Fig. 4d; Wyllie & Tuttle, 1960).The residual liquid must be driven toward alkali carbonates (Cooper et al., 1975(Cooper et al., ), (1991)), Hamilton &Kjarsgaard (1993), andKjarsgaard et al. (1995) can be more fully illustrated in terms of paths but Kjarsgaard et al. (1995) pointed out that the liquids so derived do not match some definitive characteristics of crystallization in the framework of the comprehensive phase diagrams of Figs 2, 8 and 11, although the liquid of natural natrocarbonatites.There would be ample opportunity for the residual low-viscosity alkalic car-compositions and minerals differ in detail, and the pressures are different.The product of Kjarsgaard et al. (1995) bonatite magma to escape as a fenitizing fluid.
Three of the issues debated in the 1980s were: (1) Are at the lowest temperature (700°C) corresponds to a liquid Lc on the surface passing through field boundaries I 5 ′-n silicate-and carbonate-rich liquids which do not mix at temperatures above their liquidi really conjugate liquids? and K-N (Figs 2, 8 and 9), about midway between CaO and Na 2 O, and still in equilibrium with the silicate (2) Are immiscible carbonate-rich liquids in experimental studies 'superheated' compared with natural carbonatite liquid, Ls, and silicate minerals.They recognized that the immiscible haplonatrocarbonatite magma would not magmas?(3) Are natrocarbonatite magmas derived from parental Ca-and Mg-carbonate magmas, or vice versa?precipitate carbonates without further fractionation of silicates, concluding (p.184): 'One can envisage efficient Some aspects of the debates are clarified and resolved unambiguously by the phase diagrams.The positions of fractionation of these ferromagnesian solids concurrent with the separation of exsolved carbonate liquids from those engaged in these debates have shifted through time, and the following quotations from recent papers will their silicate liquid host.'According to the phase diagrams (Figs 2 and 8), these two processes are not concurrent-suffice as reviews.Barker (1996a) referred to 'The carefully reasoned debate between M. J. Le Bas (1987, the exsolved carbonate liquids Lc must first be physically separated from the silicate host magma Ls before they 1989) and J. Gittins (1989;Twyman & Gittins, 1987)' as 'recommended reading', illustrating the complex prob-become capable of fractionating down the silicate liquidus to reach the silicate-carbonate liquidus surface (passing lems in carbonatite petrogenesis.He cited recent petrographic observations favoring liquid immiscibility in lavas through I 5 ′-o and I 3 -O, Figs 2, 8, and 9).
Kjarsgaard & Peterson (1991; using a pseudobinary (Dawson et al., 1992;Macdonald et al., 1993;Church & Jones, 1995), and recalled the arguments that these may diagram) and Hamilton & Kjarsgaard (1993;using a Hamilton projection) reported a sequence of phase as-not be 'true immiscibility', but a reluctance of two liquids at different temperatures to mix.
semblages (neglecting oxides) obtained from two sets of experiments using Shombole lavas±added CaCO 3 , The immiscibility issue, (1), and the 'superheat' issue, (2), were dealt with by Kjarsgaard et al. (1995, pp. which correspond as follows to the stages outlined above: Ls; Stage 2, Ls+Lc+silicates; and Stage 3b, 178-179), and they correctly concluded that this is not a problem at all.Twyman & Gittins (1987) had suggested Ls+Lc+silicates+calcite.The liquid composition paths plotted by Hamilton & Kjarsgaard (1993) are reasonably that the occurrence of immiscibility in the Freestone & Hamilton (1980) experiments proved nothing and was consistent with the miscibility gap field boundaries in Figs 2 and 8 (considering the differences in bulk com-an artifact of melting rocks with sharply divergent liquidi (natrocarbonatite ~500°C; silicate ~1000°C), i.e. the position and pressure), diverging from K with decreasing temperature from 1025°C to 900°C.The 900°C silicate liquids are immiscible but do not represent a conjugate pair because they did not exist as a homogeneous, high-liquid is greatly enriched in (Na 2 O+SiO 2 )/CaO.This liquid composition is constrained by the phase assemblage temperature liquid.Both Le Bas (1981Bas ( , 1987Bas ( , 1989) ) and Twyman & Gittins (1987) noted that 'the carbonate (Stage 3b) to lie on the curve of intersection of the miscibility gap surface and the silicate-carbonate liquidus liquids . . .were exsolved at temperatures 350-550°C hotter ('superheated') than the liquidus temperature of (Figs 10b and 11).This indicates that for the natural rock compositions used in the experiments, the primary natrocarbonatite'.Kjarsgaard et al. (1995) reported experiments using a natural wollastonite nephelinite lava calcite field on the miscibility gap surface extends very close to the side (SiO 2 -Na 2 O-Al 2 O 3 ) of the pseudo-and synthetic natrocarbonatite.They used a pseudobinary system to illustrate the sequence of phase as-quaternary system in Fig. 11; the calcite liquidus volume is similarly greatly expanded to low CaO contents com-semblages produced experimentally, and demonstrated the coexistence of silicate and alkalic carbonate liquids pared with that depicted in Fig. 11.
With respect to the third issue, according to Barker as conjugate pairs at 750°C and 700°C with melanite garnet, clinopyroxene and other silicate and oxide min-(1996a), 'Le Bas (1987Bas ( , 1989) ) championed natrocarbonatite liquid as parent to the more Ca-and Mg-erals, which confirmed that the carbonate liquid was not 'superheated'.The carbonate liquid compositions rich carbonate liquids, whereas Gittins (1989; Twyman trace elements and Sr isotope ratios that some of the above, carbonatite magmas are formed when (1) the calcite in the dyke (both in the groundmass and in certain silicate magma reaches the miscibility gap, or (2) the amygdales) had precipitated from residual kimberlite silicate magma differentiation path bypasses thermal liquid, whereas that in other amygdales was of secondary maxima and reaches the silicate-calcite liquidus boundorigin.They also showed that the primary calcites had ary under conditions where this boundary does not high Sr and Ba and low 87 Sr/ 86 Sr ratios, indicating that intersect the miscibility gap; this permits the liquid to these shared chemical characteristics with carbonate in continue differentiating toward CaCO 3 (as in Figs 8 and carbonatites. 11; contrast Fig. 2).The system Na 2 O-CaO-Al 2 O 3 -SiO 2 -CO 2 is not rel-What carbonatite magma compositions can be derived from evant for most mantle processes and products, but it alkalic parent magmas?In general, the higher the Na/Ca includes compositions representing some evolved nein the silicate magma, the more sodic the immiscible phelinites and phonolites, calciocarbonatites, and nacarbonate-rich magma (Fig. 9), but the Al/Si ratio is also trocarbonatites.The different types of differentiation influential (compare tie-lines I 6 -I 6 ′ and I 1 -I 1 ′ in Fig. 9).paths of carbonated alkalic silicate magmas depend on Kjarsgaard & Peterson (1991) emphasized the importance the original liquid composition, the distribution of liquidus of peralkalinity on the compositions of immiscible carfield boundaries, and the distribution of thermal maxima bonate-rich liquids.The immiscible carbonate-rich lion the liquidus.The petrological problems which can be quids separated from many starting silicate compositions addressed are presented as a series of questions, some of tend to be concentrated in the region I 1 ′, I 2 ′, I 5 ′, and I 7 ′ which were considered by Lee & Wyllie (1996) using Figs (Figs 8, 9, 10 and 11).These are calciocarbonatite 2 and 9b as guides.We can now evaluate them with magmas, with somewhat less than 80 wt % CaCO 3 but another dimension, using Figs 9, 10 and 11.
with significant contents of alkalis (up to 20 wt % or Can calcite spherules or ocelli in mantle xenoliths represent more Na 2 CO 3 ).For some high-alkali evolved (low-MgO) magma compositions?Calcite spherules in mantle xenoliths silicate magmas, there is a prospect of generating more have been interpreted as representing immiscible carsodic immiscible carbonatite magmas (I 6 -I 6 ′ in Fig. 9b), bonate liquids (e.g.Pyle & Haggerty, 1994;Seifert & which could produce natrocarbonatites with compositions Thomas, 1995;Kogarko et al., 1995a).The size of the considerably richer in Na 2 O than nyerereite (Nye, Fig. calcite liquidus volume illustrated in Figs 9, 10 and 11 1).Kjarsgaard et al. (1995) demonstrated that immiscible shows that no silicate-derived carbonatitic magma can natrocarbonatite liquids could be produced on both sides have composition with >85 wt % CaCO 3 , and such a of the nyerereite-fairchildite thermal divide.Therefore, liquid cannot solidify completely to a 99 wt % calcite natrocarbonatites corresponding to those at Oldoinyo spherule.Lee et al. (1994) and Lee & Wyllie (1996, 1997) Lengai which precipitate gregoryite can be produced by maintained that the reported calcite spherules probably liquid immiscibility from an alkali silicate parent.represented rounded crystalline calcite grown from a What crystallization paths are followed by derivative carbonatite silicate-carbonate magma.This conclusion was supmagmas?Some interesting questions about fractionation ported with phase diagrams, and with previous deof carbonatite magmas cannot be evaluated here, because scriptions of rounded calcite crystals grown our system contains neither Mg nor Fe; we are concerned experimentally in melts of varied composition.Rounded with variations in amount of carbonates, Ca/Na, and carbonate crystals are also known in lavas, e.g.gregoryite Al/Si.Carbonatite magmas precipitate calcite only when (with oscillatory zoning) in natrocarbonatite flows at they reach the silicate-calcite liquidus boundary, either Oldoinyo Lengai (Cooper et al., 1975;Church & Jones, by direct crystallization without immiscibility, or via the 1995).

Fig. 1 .
Compositional tetrahedron Al 2 O 3 -SiO 2 -CaO-Na 2 O, projected from CO 2 , showing silicate and carbonate phases relevant to the system examples of calciocarbonatites which were not dominated Na 2 O-CaO-Al 2 O 3 -SiO 2 -CO 2 .Χ, compositions of these phases.Ab, by alkalic rocks, suggesting that some of them represented albite; Ne, nepheline; An, anorthite; Gr, grossular; Wo, wollastonite; primary magmas from the mantle.Scap, scapolite; Lar, larnite; Geh, gehlenite; Can, cancranite; CC, calcite; Nye, nyerereite; NC, sodium carbonate.Β, projection of the It is well established experimentally that the comphases on the basal triangle SiO 2 -CaO-Na 2 O from Al 2 O 3 .The bold position of magma from carbonated peridotite does not triangle indicates the studied join Ne-Ab-CC.(b) Compositional tri- correspond to calciocarbonatite, but to a calcic dolomitic angle (Al 2 O 3 +SiO 2 )-CaO-Na 2 O (+CO 2 ; Hamilton projection), show- ing the same silicate and carbonate phases (Χ).Our starting mixtures magma with silica and alkali content depending on the lie slightly below line Ne-CC, with 10 wt % Ab in the bulk silicate 2 O-CaO-Al 2 O 3 -SiO 2 -CO 2 lowered sufficiently for the silicate-carbonate field bound-The phase relationships can be presented in the tetary to pass underneath the high-temperature miscibility rahedron Al 2 O 3 -SiO 2 -CaO-Na 2 O with excess CO 2 (Fig. gap.This confirmed and explained the result of Wat-1a), and in the Hamilton projection (Fig. 1b) used by kinson & Wyllie (1971) for the join Ne-CC-H 2 O, given Freestone & Hamilton (1980): (Al 2 O 3 +SiO 2 )the result of Kjarsgaard &

Fig. 2 .
Fig. 2. Miscibility gap (two-liquid field), silicate liquidus surface, and petropoxy, polished by Al 2 O 3 powders and cleaned by carbonate liquidus surface at 1 GPa presented in the Hamilton procompressed air, without using water, and then carbonjection, based on results intersected by the join Ab-CC at 1-2•5 GPa

Fig. 3 .
Fig. 3. Phase fields intersected by the join Ne 90 Ab 10 -CC at 1•0 GPa.Bold curves locate the fields for the existence of immiscible liquids Ls and Lc.Dashed curves indicate tentative determination of the phase field boundaries.Solidus is not determined.It should be noted that the dashed horizontal line near 1200°C indicates the appearance of an isobaric invariant assemblage: Ne+An+Mel+CC+Lc+V.Ne, nepheline; An, anorthite; Scap, scapolite; Mel, melilite; CC, calcite; Ls, silicate-rich liquid; Lc, carbonate-rich liquid.(See text for descriptions of the liquiduspiercing points P, Q, R, T and S.)

Fig. 6 .
Fig. 6.Average two-liquid compositions (Table 2) with corresponding tie-lines, including results at 1350°C (runs 130 and 129), 1300°C (117, 127 and 145), and 1250°C (119 and 123).Three pairs of two-liquid isotherms are drawn passing through the data, with those at lower temperature showing wider separation.The positions of piercing points P, Q (with error bar), and R along the join Ne 90 Ab 10 -CC are also indicated.

Fig. 7 .
(a) Lower portion of the Hamilton projection (Fig. 6; vertical (R) and the defined point T (Figs 3 and 7a) within the exaggeration of two), showing the miscibility gap field boundary (bold miscibility gap (compare runs 117 and 119).The errors curve) through P, I 1 , I 2 , Q, K, I 2 ′ and I 1 ′, two-liquid isotherms at 1350, 1300 and 1250°C, and tie-lines I 1 -I 1 ′ and I 2 -I 2 ′ intersecting the studied associated with the location of points Q and R will be join at S and T (see text).Italic characters Ne, An and Mel surrounding considered in connection with Fig. 7. Melilite becomes the field boundary indicate the liquidus fields for nepheline, anorthite unstable below the fields 3 and 5 (compare runs 119 and and melilite.(b) Complete range of the field boundary to proper scale.

Fig. 8 .
Fig. 8. Completed liquidus field boundary diagram based on Fig. 3, and Figs 5-7, showing the miscibility gap, and the silicate and carbonate liquidus surfaces (compare Fig. 2).The key boundaries, cooling directions, and pseudoternary isobaric invariant points are also indicated (see text).The details near the CaO corner of the projection are sketched in the upper-right inset.It should be noted that the compositions of Ne, An, and Mel (Χ) are ideal values (e.g.Ne, An, and Geh, Fig. 1).
contain less CO 2 than indicated by the wt % of these analyses (2-3 measurements for each quenched liquid except one for Lc of run 145) are plotted in Fig. CaCO 3 in the starting mixtures.
join and the CO 2 -free projection.feature is how closely the tie-lines pass through the starting compositions.The tie-lines are subparallel at all temperatures within the narrow interval 1250-1350°C, with a regular arrangement of the two-phase tie-lines as MISCIBILITY GAP AND LIQUIDUS FIELD a function of changing bulk composition (including run BOUNDARIES IN THE SYSTEM 119 for composition '60' at 1250°C containing some Na 2 O-CaO-Al 2 O 3 -SiO 2 -CO 2 melilite and anorthite).The only exception is run 117 at 2 boundary gap, bounded by the field boundaries for two immiscible I 5 ′-n, and Lee & Wyllie (1996), boundaries G-N and liquids Ls+Lc coexisting with one mineral and vapor.g-n in fig.15].Carbonate is precipitated along with the Figure 7a with vertical exaggeration of two shows the silicate at lower temperatures, as indicated by the mineral data sources and construction methods, and Fig. 7b sequences in the right-hand side of Fig. 3 [see Fig. 2 shows the projected CO 2 -free results to scale.Points P boundary I 5 ′-o, and Lee & Wyllie (1996), boundaries (1335°C) and Q (1325°C) are on the miscibility gap G-O and g-o in fig.15].field boundary, coexisting with nepheline and melilite, ), CC-NC plane has Al/Si=0•33.Curves A and B on passes through the estimated piercing point R (Figs 3 and these two planes in Fig. 10a are the miscibility gap field 6), and then passes through the deduced pseudoperitectic boundaries from Figs 2, 8, and 9 along with the isobaric points I 4 and I 3 where melilite is successively replaced by pseudoinvariant points and the critical points, K and k, anorthite and nepheline.It then extends to O as discussed from Fig. 9.The immiscible liquid projections A and B, above (compare Fig. 2).The carbonate liquidus surface the field boundaries in Figs 2 and 8, then correspond is thus restricted to a field with primary calcite near closely to the curves of intersection of the miscibility gap CaO, which extends along a narrow area close to the side with the two planes.The possible range of the miscibility CaO-Na 2 O, with decreasing temperature.The primary gap is extrapolated down to the Al-free basal triangle liquidus carbonate minerals involved change from calcite SiO 2 -CaO-Na 2 O (Qz-CC-NC), and denoted by curve to alkali-rich carbonates (e.g.nyerereite, sodium car-C.It is, however, difficult to extrapolate further towards bonate), as indicated by the classic study of Cooper et al. the Al-rich region, and this part of the miscibility gap (1975) at lower pressures.remains unplotted.The three curves A, B and C from Fig. 10a provide the framework for the shaded surface of the miscibility gap in Fig. 10b (curves A and B are dashed lines).The The pseudoternary systems through pseudoternary isobaric invariant points on the triangular Ne 90 Ab 10 -CaCO 3 -Na 2 CO 3 and slices in Fig. 10a are located on the miscibility gap field Ab-CaCO 3 -Na 2 CO 3 boundaries traversing the shaded surface, as shown in The two pseudoternary isobaric phase diagrams for the Fig 10b.The field boundaries meet in pseudoquaternary triangular slices CaO-Na 2 O-silicate through Fig.

Fig. 9 .
Fig. 9. (a) and (b), comparison of the lower portions of phase diagrams in Figs 2 and 8 (vertical exaggeration of two).The Ab-Wo liquidus field boundary in (b) now extends to the miscibility gap at I 6 , with the corresponding immiscible carbonate-rich liquid I 6 ′ (exact position not shown) along the section I 5 ′-n.

Fig. 10 .
(a) CO 2 -saturated, compositional tetrahedron (Fig. 1a), show-9b) with the corresponding pseudoternary paths for the ing the miscibility gap field boundaries of Figs 8 and 2 in space (curves new results from the Ne 90 Ab 10 -CC join (Figs 8 and 9a), A and B near the planes Ne-CC-NC and Ab-CC-NC).The miscibility and then illustrate more clearly the effect of composition gap field boundary C on the base of the tetrahedron is extrapolated from A and B. K-k indicates the critical curve for Ls=Lc.(b) Same by combining the results in the pseudoquarternary system projection as in (a), showing the surface of the miscibility gap (shaded (Fig. 11).area) with five-phase field boundaries (two liquids, two minerals, and vapor).Italic characters Ne, Plag, Mel, Wo, and CC on the surface mark the fields for nepheline, plagioclase (Ab to An), melilite, wollastonite, and calcite, coexisting with immiscible liquids (with vapor).(See Fig.The pseudoternary system through 1a for the positions of An, Mel/Geh, and Wo.) Figure 11 provides a topological responding to a haplocarbonatite magma, changes comframework for tracing paths of crystallization leading to position along the silicate-carbonate field boundary haplocarbonatite magmas.This overall view of the phase surface, visualized in part by surface intersections shown relationships facilitates the interpretations of results from by the field boundaries I 5 ′-o and E-I 4 -I 3 -O (Figs 2 and individual rock studies, which sample only a small fraction 8 miscibility gap.Some liquids cool down the steep silicate What is the maximum percentage of CaCO 3 in immiscible liquidus (connecting the area between K-I 2 ′-I 1 ′-N and carbonatite magmas?We see no evidence for >80 wt % I 4 -I 3 -O in Fig. 8 with I 5 ′-n-o in Fig. 2) until they reach CaCO 3 in Hamilton projections of experimental results the boundary represented by the shaded surface behind (Figs 2 and 8; Hamilton & Kjarsgaard, 1993; Macdonald the sharp bend below G in Fig. 11.The carbonatite et al., 1993; Lee & Wyllie, 1997, with MgO added).magma compositions are concentrated in the range I 4 , Hamilton & Kjarsgaard (1993) reported immiscible li-I 3 to I 5 ′ (Figs 2, 8, and 9), but for peralkaline compositions quids reaching 90 wt % CaCO 3 when plotted in terms they may extend to significantly more sodic compositions of carbonate components only, but when the presence near I 6 ′ (Figs 2, 9, and 10).Differentiation along the of silicate components is taken into account, they plot the silicate-carbonate liquidus boundary (under conditions same point in a Hamilton projection which corresponds to where it does not intersect the miscibility gap, Figs 8 and a composition with no more than 75 wt % CaCO 3 , 9a) can yield melts corresponding to calciocarbonatites consistent with the other published miscibility gaps.(e.g.near I 4 , I 3 in Figs 8 and 9a).The results of Cooper What conditions and processes can lead to the formation of carbonatite magmas from alkalic parent magmas?As shown et al. (1975) in the model system CaCO 3 -Na 2 CO 3 -K 2 CO 3
The surfaces are steep, and the Original silicate-CO 2 liquids can generate carbonateamount of precipitation is small.Although Lc is carbonate rich liquids with compositions on the surface of the rich, it cannot precipitate carbonate minerals until it shaded carbonate volume, but it is impossible for liquids reaches the silicate-calcite field boundary, and Stage 4. to follow paths inside this volume.Therefore, this is a Stage 3b: Ls+Lc+silicates+calcite.This stage corforbidden volume for carbonatite magmas in this model responds to Ls reaching the limiting silicate-calcite field system at 1•0 GPa.boundary on the miscibility gap surface between Wo and CC in Figs 10b and 11, the locus of points corresponding to I 5 and I 5 ′ in Figs 2 and 9b; this is the curve of intersection