Journal of Petrology | Volume 43 | Number 10 | Pages 1787-1814 | 2002
© Oxford University Press 2002
The Substitution of Al and F in Titanite at High Pressure and Temperature: Experimental Constraints on Phase Relations and Solid Solution Properties
PETER TROPPER1,2,*,
CRAIG E. MANNING2 and
ERIC J. ESSENE1
1DEPARTMENT OF GEOLOGICAL SCIENCES, UNIVERSITY OF MICHIGAN, 2534 C. C. LITTLE BUILDING, ANN ARBOR, MI 48109-1063, USA
2DEPARTMENT OF EARTH AND SPACE SCIENCES, GEOLOGY BUILDING, UNIVERSITY OF CALIFORNIA AT LOS ANGELES, LOS ANGELES, CA 90095-1567, USA
Received
July 10, 2001;
Revised typescript accepted
March 26, 2002
 |
ABSTRACT
|
|---|
Experimental studies were carried out to evaluate phase relations
involving titanite–F–Al-titanite solid solution
in the system CaSiO
3–Al
2SiO
5–TiO
2–CaF
2. The
experiments were conducted at 900–1000°C and 1·1–4·0
GPa. The average F/Al ratio in titanite solid solution in the
experimental run products is 1·01 ± 0·06,
and X
Al ranges from 0·33 ± 0·02 to 0·91
± 0·05, consistent with the substitution [TiO
2+]
–1[AlF
2+]
1.
Analysis of the phase relations indicates that titanite solid
solutions coexisting with rutile are always low in X
Al, whereas
the maximum X
Al of titanite solid solution occurs with fluorite
and either anorthite or Al
2SiO
5. Reaction displacement experiments
were performed by adding fluorite to the assemblage anorthite
+ rutile = titanite + kyanite. The reaction shifts from 1·60
GPa to 1·15 ± 0·05 GPa at 900°C, from
1·79 GPa to 1·375 ± 0·025 GPa at
1000°C, and from 1·98 GPa to 1·575 ±
0·025 GPa at 1100°C. The data show that the activity
of CaTiSiO
4O is very close to the ideal molecular activity model
(X
Ti) at 1100°C, but shows a negative deviation at 1000°C
and 900°C. The results constrain
G°
f,298·15 of
CaAlSiO
4F to be -2595 ± 3 kJ/mol and S°
298·15 to be in the range of 105·2–109·6 J/mol
K, which in turn can be used to calculate petrogenetic grids
involving titanite solid solutions in the system CaTiSiO
4O–CaAlSiO
4F.
KEY WORDS: F–Al-titanite; thermodynamic data; TARK; fluorite; reaction displacement
|
INTRODUCTION
|
|---|
Geological background
Titanite is a common accessory mineral in mafic, pelitic and
granitic rocks from many geological environments (
Higgins & Ribbe, 1976;
Ribbe, 1982;
Enami et al., 1993). Titanite may
deviate significantly from its ideal composition by the substitution
Al and F for Ti and O (
Hollabaugh, 1980;
Franz & Spear, 1985;
Bernau et al., 1986;
Fehr, 1991;
Oberti et al., 1991;
Carswell et al., 1996). The Al + OH
Ti + O substitution leads
to the Al–OH end-member vuagnatite CaAlSiO
4(OH), which
has a different structure from titanite (
McNear et al., 1976)
and is typical of low-temperature geological environments (
Enami et al., 1993).
By contrast, the F–Al substitution is isostructural
and is common at high metamorphic temperatures (>500–600°C)
and pressures. For example, up to 55 mol % F–Al substitution
has been reported from high- and ultra-high-pressure metamorphic
rocks (
Franz & Spear, 1985;
Sobolev & Shatsky, 1991;
Carswell et al., 1996). The common association of F–Al-rich
titanites with high-pressure environments has led to the suggestion
that equilibria involving titanite solid solution may be useful
for constraining pressure, temperature and/or F
2 fugacity during
metamorphism (
Smith, 1977;
Franz & Spear, 1985;
Gibert et al., 1990;
Enami et al., 1993;
Carswell et al., 1996;
Markl & Piazolo, 1999).
Despite the importance of the F–Al content of titanite there are few experimental data available with which to evaluate the physical and chemical controls on titanite solid solutions. The F–Al substitution was examined by Smith (1981) and Troitzsch & Ellis (1999). Smith (1981) synthesized titanite with about 50 mol % F–Al titanite substitution at 1000–1200°C and 1·5–3·5 GPa, similar to the upper limit of substitution reported from various natural occurrences. He also found an increase in the F–Al substitution with falling temperature and increasing pressure. Troitzsch & Ellis (1999) synthesized end-member F–Al-titanite at 1100°C and high pressure, and demonstrated complete solid solution in the system CaTiSiO4O–CaAlSiO4F.
In high-pressure rocks, titanite and rutile also form the basis of a set of equilibria useful for the determination of pressure. Specific equilibria discussed by Manning & Bohlen (1991) include
Manning & Bohlen (1991) used these reactions to estimate pressures in high-pressure
rocks from Scotland and Austria and ultra-high-pressure rocks
from Kazakhstan. The successful application of reactions (1)
and (2) to thermobarometry illustrates the utility of titanite-bearing
equilibria in petrology. However, the accuracy of the results
depends on understanding the subsolidus properties of titanite
solid solutions and relevant phase equilibria. We focus here
on phase relations involving F–Al-titanite.
Preliminary analysis of phase relations
Examination of the compositional space relevant to F–Al-titanite solid solution provides guidance for the design and interpretation of the present experiments, and for comparison with previous work. Figure 1 illustrates that a variety of F-bearing phases must be considered. In the basal plane of the quaternary CaO–Al2O3–SiO2–CaF2 (Fig. 1a), anhydrous, high-temperature Ca–Al silicates are coplanar (zoisite is projected into this space for comparison with its fluorian analogue). Relevant F-bearing phases, which plot above the basal plane in Fig. 1a, include in order of increasing relative CaF2 content, hypothetical F-zoisite (Ca2Al3Si3O12F), F–Al-titanite (CaAlSiO4F), and the hypothetical F-grossular end-member (Ca3Al2
3F12). Also shown are the intermediate compositions on the grossular–F-grossular join, Ca3Al2Si2O8F4, which is collinear with anorthite and fluorite, and Ca3Al2Si1·51·5O6F6, which is coplanar with fluorite, quartz and kyanite.
View larger version (13K):
[in this window]
[in a new window]
|
Fig. 1. (a) CaO–CaF2–Al2O3–SiO2 quaternary. Perovskite-type phases (CaTiO3 and intermediate CaTixSi1-xO3; e.g. Leinenweber et al., 1997) are omitted. The shaded area marks the triangle wollastonite–kyanite–fluorite and contains end-member F–Al-titanite (FAT), F-bearing zoisite (Fzo) and an intermediate F-grossular–grossular solid solution, Fgrs/3; •, minerals on basal CaO–Al2O3–SiO2 plane; gray circles, F-bearing minerals in wollastonite–Al2SiO5–fluorite plane;  , F-grossular (Fgr), which projects outside the quaternary volume. Tie lines link relevant solid solutions. (b) Wollastonite–rutile–kyanite–fluorite quaternary illustrating the phase relations among anorthite–kyanite–titanite–rutile–F–Al-titanite (FAT)–fluorite; •, minerals on or projecting onto wollastonite–Al2SiO5–fluorite plane; gray circles, Ti-bearing compositions; , F-grossular (Fgr), which projects outside the quaternary volume. Tie lines link relevant solid solutions. All mineral abbreviations according to Kretz (1983) except: Cats, Ca-Tschermaks; Coe, coesite; FAT, F–Al-titanite; Fgrs and Fgrs/3, F-grossular and F-grossular–grossular solid solution; Fzo, F-zoisite; Lar, larnite; Lm, lime; Rk, rankinite.
|
|
Mapping the coordinate system in Fig. 1a to CaSiO3–Al2SiO5–CaF2 allows addition of TiO2 (Fig. 1b) and illustrates the positions of titanite–F–Al-titanite solid solution. Anhydrous F-zoisite or F-grossular have not been reported from nature or experiments. Fluorian zoisite was identified by Troitzsch & Ellis (1999) in their F–Al-titanite synthesis experiments, but they determined that it contained significant hydroxyl. Fluorian grossular–andradite solid solutions occur naturally (e.g. Valley et al., 1983; Manning & Bird, 1990), but stoichimetric considerations imply significant OH as well. The extent to which hydroxyl substitution is required to stabilize these F-bearing phases is unknown; nevertheless, the possibility of F substitution along these trajectories must be considered, especially in anhydrous experimental systems at high T and P.
The mechanisms of F substitution differ in zoisite, grossular and titanite. In zoisite, a simple [OH-]-1[F-]1 exchange operates; in F-grossular, F substitution is coupled to Si vacancies via [SiO44-]-1[F44-]1; and the exchange vector for the titanite solid solution is [TiO2+]-1[AlF2+]1. Thus, in addition to P and T, the chemical potentials of the thermodynamic components F2, SiO2, H2O, Al2O3 and TiO2 must all be considered in evaluating phase equilibria involving F–Al-titanite solid solution.
Because of the many variables that might control phase assemblages and compositions, it is useful to identify that appropriate portion of compositional space that is most geologically relevant. Figure 1a shows that phases on the fluorite–wollastonite–kyanite plane are silica-rich and are, therefore, most likely to be associated with quartz in natural systems. Thus, we focused our efforts on the compositional region encompassed by the phases wollastonite–kyanite–rutile–titanite–fluorite; that is, the region depicted in Fig. 1b. Additional relevant phases within the bounding compositions are F-zoisite, F–Al-titanite, and F-grossular. Although the grossular–F-grossular solid solution intersects the anorthite–fluorite join at Ca3Al2Si2O8F4, we anticipate that this substitution is minimized when the investigated bulk compositions are at or near quartz saturation.
Ignoring F-zoisite, which is unknown in nature, there are two reactions involving F–Al-titanite (Fig. 1b):
Consideration of the geometric relations in Fig. 1b indicates that is there is only one other possible reaction among these phases,
which
was experimentally investigated by
Manning & Bohlen (1991).
We use the acronyms AFT, TAFT and TARK for reactions (3), (4)
and (5), respectively.
Relative stabilities in P–T space can be predicted using estimated properties of Al–F-titanite. The volume (V°298·15) of synthetic Al–F-titanite was determined by Troitzsch & Ellis (1999) to be 344·11 Å3 or 51·805 cm3/mol (Troitzsch & Ellis, 1999). This is 
7% smaller than the reported volume of end-member titanite CaTiSiO4O, which varies from 55·74 cm3/mol (Robie & Hemingway, 1995) to 55·64 cm3/mol (Xirouchakis & Lindsley, 1998). The volume is also slightly lower than the 52·5 cm3/mol that was predicted for pure CaAlSiO4F by Oberti et al. (1991) by extrapolation of volume data from natural Al–F-bearing titanites.
The entropy (S°298·15) is estimated here by additive techniques (Robinson & Haas, 1983). Two estimates are made. The first, based on a reaction among orthosilicates,
yields
S°
298·15 for Al–F-titanite of 105·2 J/mol K. The second
estimate used the reaction among oxides and fluorides, where
the data for the oxides were taken from
Holland (1989) and those
for fluorite were taken from
Robie & Hemigway (1995),
and gives
S°
298·15 = 109·6 J/mol K after applying a volume correction. These
entropy estimates of 105–110 J/mol K are significantly
lower than the entropy of titanite, which has been inferred
to range from 129·2 J/mol K (
Robie & Hemingway, 1995)
to 131·2 J/mol K (
Holland & Powell, 1998), but similar
to the value of 107·8 J/mol K estimated by
Xirouchakis & Lindsley (1998).
Their
S°
298·15 for titanite
is too low to be consistent with phase equilibrium constraints
of
Manning & Bohlen (1991).
The slopes of reactions (3)–(5) were calculated using our estimated ranges of S°298·15 and V°298·15 for F–Al-titanite, along with data for the other minerals from Robie & Hemingway (1995). The calculations yield dP/dT for reaction (3) ranging from 26·9 to 28·9 bar/°C. The Clapeyron slope of the TAFT reaction (4) is much steeper and ranges from 55·5 to 123·5 bar/°C. The estimate for reaction (5) is 17·2–19·2 bar/°C, similar to the Clapeyron slopes of 20·5–23·5 bar/°C obtained by Manning & Bohlen (1991). Our calculations therefore suggest that the AFT equilibrium (3) has a slightly steeper slope than equilibrium (5) suggesting that the two reactions intersect, with reaction (3) stable at higher pressures at high temperatures.
Scope of the study
The relative shifts of the three equilibria (3)–(5) as a consequence of titanite solid solution are unknown. Quantitative analysis of these phase relations, and related solution energetics and thermodynamic properties, require experimental studies of appropriate equilibria and bulk compositions. The first part of this study involves the assessment of conditions of F–Al-titanite stability and examination of reaction (4) to explore its possible utility for application to metamorphic phase equilibria. The second part of this study focuses on the thermodynamics of mixing in CaTiSiO4O–CaAlSiO4F solid solutions and the derivation of thermodynamic properties of CaAlSio4F.
|
METHODS
|
|---|
Starting materials
Natural and synthetic minerals were used as starting materials.
Natural kyanite from Brazil (0·15 wt % Fe
2O
3,
Bohlen et al., 1991)
was boiled in concentrated HF for
3·5 h
to dissolve minor adhering sheet silicate grains and then fired
at 950°C overnight. We used synthetic titanite from
Manning & Bohlen (1991).
Anorthite was synthesized from high-grade
(99·99% pure) SiO
2, Al
2O
3 and CaCO
3. The mixture was
decarbonated at 1000°C overnight and then melted in a Pt
crucible at 1600°C for 30 min. The quenched glass was ground
and then crystallized at 1400°C over a period of 2 days,
with repeated grinding. Rutile was synthesized by firing TiO
2 at 1000°C over a period of 3 days, with repeated grinding.
Synthetic fluorite (Fisher Co.) and natural fluorite from the
microprobe standard collection of the Electron Microprobe Analysis
Laboratory, University of Michigan, were used. Anorthite, titanite
and kyanite were analyzed by electron microprobe (EMP) analyses
and X-ray powder diffraction to verify their purity. X-ray powder
diffraction confirmed the absence of additional phases in the
rutile and natural and synthetic fluorite.
Run procedures
All experiments were conducted in an end-loaded piston–cylinder apparatus at UCLA similar to that described by Boyd & England (1960) with 25·4 mm (1 inch) diameter furnace assemblies and pistons for runs below 2·0 GPa; at higher pressures, 12·7 mm (1/2 inch) furnace assemblies and pistons were used. The furnace assembly is primarily made of graphite, NaCl and MgO (Bohlen, 1984; Manning & Boettcher, 1994). The furnace assembly for high-temperature runs contained a Pyrex sleeve between the graphite furnace and the salt cell to prevent the penetration of salt melts into the graphite furnace (Boettcher et al., 1981; Manning & Boettcher, 1994).
Stoichiometric mixtures of the starting materials were sealed in a welded Ag80Pd20 capsule with 2 mm diameter. The capsule was placed horizontally in the graphite furnace and packed in boron nitride (BN) to reduce temperature gradients. A piece of Pt foil was placed on top of the capsule to prevent puncture by the thermocouple. Temperature was measured with a type S (Pt/Pt90Rh10) thermocouple, with a precision estimated to be ±3°C, and pressures was monitored using a Heise gauge having a precision of about ±0·01 GPa.
The piston-in method (Johannes et al., 1971) was used for every experiment. First, a pressure of about two-thirds of that to be maintained during the run was applied to the furnace assembly. The temperature was then raised to the desired temperature of the run over a period of 10 min and finally the pressure was increased to the desired value. In the high-temperature runs, pressure was increased to about 0·3–0·5 GPa and then the temperature was raised to 700°C to ensure ductile behavior of the Pyrex sleeve. Afterwards, the pressure was increased to 0·2 GPa below the desired value and then the temperature was raised to the final value.
Analytical techniques
X-ray powder diffraction was conducted using a conventional Norelco–Philips vertical diffractometer with variable slit geometry. All run products were analyzed by scanning electron microscopy (SEM) and EMP analysis. Electron microprobe analyses were obtained with the Cameca CAMEBAX EMP at the Electron Microbeam Analyses Laboratory (EMAL) of the University of Michigan. Additional EMP analyses were obtained with the Cameca CAMEBAX EMP at UCLA. Analyses were obtained at 15 kV and 10 nA beam current with a point beam. Natural and synthetic mineral standards included Topaz Mt. topaz (F), synthetic rutile (Ti) and natural tanzanite (Ca, Al, Si). Fluorine was analyzed on the multilayer crystal OV60 (6 nm d-spacing). The counting times were 50 s for F and 30 s for all other elements. Raw data were reduced with a PAP-type correction. The titanites of runs 265, 248 and 249 were analyzed with a JEOL 6310 SEM equipped with a LINK ISIS energy-dispersive system and a MICROSPEC wavelength-dispersive system at the Institute of Mineralogy and Petrology, University of Graz. The analytical conditions were 15 kV and 5 nA sample current. The elements Si, Al, Ca and Ti were analyzed with the energy-dispersive system and the F content was analyzed with the wavelength-dispersive system, because of the lower detection limit of 0·05–0·1 wt % of the wavelength-dispersive system.
|
RESULTS
|
|---|
Titanite solid solutions were produced with a range of starting
assemblages at 900–1100°C and 1·1–4·0
GPa (Table
1). We prepared starting mixtures with 12 molar ratios
of fluorite ± anorthite, rutile, kyanite, titanite and
quartz (series A–H; Table
1). The run duration was between
23 and 118 h for most experiments, depending on the temperature
as shown in Table
1. It is likely that such long durations are
conservative, inasmuch as two short runs (run 248, 8 h; run
245, 6 h) yielded sufficient reaction progress to determine
the reaction direction.
Equilibrium constraints
Owing to the high temperatures of all experiments (900–1100°C), reaction rates were rapid and the growth of F–Al-rich titanite could easily be observed in the run products. Using the starting assemblage titanite + kyanite + rutile + fluorite, we evaluated equilibration in our experiments in two ways. First, different bulk compositions with varying rutile and fluorite content (series F and G), or varying titanite, kyanite and fluorite content (series D1–D4) were run side by side at the same P, T and time. The compositions of newly grown titanitess in experiments with mix F (run 226) and mix G (run 227) were identical and yield XAl = 0·51 as shown in Table 2a and b. The results from the experiments with mixtures D1–D4 also result in the same XAl of 0·87–0·90 in titanitess (Table 2a). In a second test of equilibrium, experiments at 900 and 1000°C and 2·0 GPa (series F) and experiments at 1000°C and 1·35 GPa and 1100°C and 1·55 GPa (series H) were run for different durations to compare the compositions (XAl) of the newly grown titanitess as a function of time. In these runs, experiments with mix F (kyanite + rutile + titanite + fluorite) at 900°C yielded statistically indistinguishable XAl of 0·54 ± 0·04 and 0·55 ± 0·03 regardless of whether the run duration was 48 or 118 h (Fig. 2a). Similarly, titanitess from the experiments at 1000°C with the same assemblage have XAl of 0·52 ± 0·01 and 0·51 ± 0·00 obtained at 24 and 114 h run (Fig. 2b). Two experiments with mixture H were also run with different run durations. The experiment at 1000°C and 1·35 GPa was run for 43 h (run 249) and 75 h (run 269). Both runs yielded similar XAl of 0·48 ± 0·05 and 0·51 ± 0·03. At 1100°C, the experiment at 1·55 GPa was also conducted with different run times. Run 248 and run 266 yielded a similar XAl of 0·50 ± 0·04 and 0·49 ± 0·01 whether the run duration was 8 or 66 h (Fig. 2c).
These results show that newly grown titanite compositions are reproducible and that run durations were sufficient. We take these conditions to reflect equilibrium.
Role of buffering assemblages in F–Al-titanite growth
We explored the role of starting assemblages in F–Al-titanite growth by attempting to grow titanite solid solutions at 1000°C and 2·0 GPa from a range of materials and stoichiometries. The bulk compositions were in the subvolume to the right of the fluorite–anorthite–titanite plane in Fig. 1b.
Examples of run products and typical textures are shown in Fig. 3a–d. The bulk compositions are compared in Fig. 4. Starting mixture A (runs 246 and 228) did not yield F–Al-titanite; however, F-bearing zoisite and quartz nucleated and grew (Fig. 3a). The F–Al-titanite solid solutions grew in all other starting mixtures. The microprobe analyses of titanites were usually performed in the rims of the newly grown titanites. If compositional zoning was visible, the newly grown core was also analysed. In experiments of shorter duration, there is a sharp transition within titanite grains between newly grown F–Al-titanite rims and titanite seeds in the core (Fig. 3b), whereas longer experiments produced a gradual variation in composition from core to rim (Fig. 3c). All run product assemblages included fluorite. Anorthite was partially consumed in all experiments in which F–Al-titanite solid solutions grew (series B–H); in some instances (e.g. run 231) only minor anorthite remained. Most of the experiments had yields of newly grown titanite solid solution. Starting CaTiSiO4O was nearly completely consumed (Fig. 3d). Minor corundum grew in experiment run 226. Tiny crystals of newly grown rutile often occur as oriented inclusions in the cores of newly grown F–Al-rich titanite, mostly together with small quartz inclusions (Fig. 3d). In the experiments on the TAFT reaction (4) with mix F, newly grown titanitess overgrew kyanite. In the experiments at 1·1 GPa and 1·3 GPa at 1000°C and 1100°C, kyanite is rimmed by anorthite, indicating overstepping of the TARK equilibrium (5). Quartz is present in the run products of all experiments.
The experiment with starting mixture B (run 247) yielded <5 µm wide rims of F–Al-titanite that overgrew starting titanite seeds (Fig. 3b). The experimental series with starting mixtures C and D yielded similar results. Mix C1 (run 231) gave high yields of F–Al-titanite, with crystals larger than 10 µm in diameter and almost no anorthite left (Fig. 3c). The F–Al-rich titanites from this experiment occasionally show zoning with Ti-rich cores and Ti-poor rims. Relict titanite from the starting mixture and quartz crystals also occur in the run products. This kind of zoning was also observed in other experiments, where the Ti-rich cores contain rutile inclusions. Mixtures D1–D4 (runs 233, 79, 80 and 81) and E (run 232) yield substantial volumes of F–Al-rich titanite. Rutile-bearing starting mixtures (F, G and H) yielded newly grown F–Al-titanite and rutile, whereas only small patches of end-member titanite from the starting assemblage remain (Fig. 3d).
The bulk compositions of the starting assemblages of series A–H are linked by tie lines in Fig. 4 to the compositions of titanite that grew in them; titanite compositions are shown alone in Fig. 5 in terms of the extent of F–Al exchange (see Table 1 for average XAl in titanite, and Table 2a–c for representative titanite analyses). For mixture H, we used the titanite composition of experiment 230 at 1000°C and 2·0 GPa. The XAl of titanitess varies in the experiments from 0·51 ± 0·02 (mixes F, G and H) to 0·91 ± 0·05 (mix B), indicating a wide and continuous range of F–Al-titanite substitution. The errors are 1
in the mean. The pure F–Al end-member CaAlSiO4F was never synthesized in the experiments: the titanites always contain some Ti, required by the presence of Ti in all starting mixtures. The F/Al ratio analyzed in titanites from all experiments is 1·01 ± 0·06, indicating a coupled substitution of F with Al. Linking these data to the bulk composition of the buffering assemblages of the mixtures may be depicted in a kyanite–wollastonite–rutile triangle (Fig. 4). These data indicate that the starting composition is not relevant and the final titanite composition depends on P, T and the final buffering assemblage.
The P–T dependence of the F–Al substitution in titanite
Another set of experiments explored the dependence of F–Al substitution in titanite on P and T. This requires an appropriate mineral assemblage to simultaneously buffer the compositions of the F–Al and Ti–O components of titanite solid solutions. If the titanite composition is a binary solid solution, as required by the Gibbs–Duhem equation, this conditions is met by reactions (3)–(5). In this study, we explored the P–T dependence with the TAFT reaction (4). Starting mixtures F and H, which contained kyanite, rutile, fluorite, titanite ± anorthite (Table 1), were utilized in these experiments. At high P and T, reaction (4) is driven to the right as F–Al-titanite and rutile grow at the expense of kyanite, titanite and fluorite, until equilibrium titanite compositions are generated. Representative compositions are given in Table 2b. In contrast to the reactions involving anorthite + fluorite, these experiments never yielded F-bearing zoisite.
Pressure dependence
Figure 6a–c shows schematically the predicted shapes of reactions (3)–(5) in a P–XAl diagram. Reaction (3), AFT, will increase in pressure with increasing Al substitution in titanite, whereas reaction (5), TARK, will decrease in pressure with increasing Al substitution in titanite. Reaction (4), TAFT, also has a positive slope, which changes for intermediate values of XAl. The pressure dependence of titanite composition in the TAFT equilibrium was investigated at a range of pressures at 900°C, 1000°C and 1100°C (Table 1). New F–Al-titanite and rutile grew in all runs, as demonstrated by F–Al-titanite rims on titanite, and rutile inclusions with or without quartz or coesite in F–Al-titanite (Fig. 3d). Rutile clearly grew during these experiments, as its modal abundance strongly increases. In contrast to rutile from the starting material, the newly grown rutiles are extremely small and form oriented inclusions in cores of newly formed F–Al-titanites. In two experiments (46 and 256) with mixture F at 1·1 GPa at 1000°C and 1·3 GPa at 1100°C, anorthite grew in addition to F–Al-titanite and rutile. This is similar to experiments with mixture H below 1·2 GPa at 900°C, 1·4 GPa at 1000°C and 1·6 GPa at 1100°C (262, 238, 239, 249, 242, 243 and 248), where kyanite was nearly completely consumed and is rimmed by anorthite.
View larger version (18K):
[in this window]
[in a new window]
|
Fig. 6. Schematic P–XAl and T–XAl diagrams to illustrate the compositional changes of titanite solid solution along reactions (3)–(5).
|
|
The change in XAl of newly grown titanitess with pressure is given in Table 2b and displayed in Fig. 7a and b, illustrating the results from mixtures F (Fig. 7a and b) and H (Fig. 7b). Values of the XAl in mixture F increase with increasing pressure at 900°C, 1000°C and 1100°C, with values ranging from 0·33 ± 0·05 to 0·57 ± 0·01 (Fig. 7a, Table 2). The change in composition with pressure is greater when titanite coexists with anorthite below the TARK equilibrium than with kyanite above it (Fig. 7a). At any given pressure, titanite composition buffered by the TAFT equilibrium is the same at 1000 and 1100°C. The results with mixture H are similar (Fig. 7b). Values of XAl increase with increasing pressure at 900°C from 0·42 to 0·48, at 1000°C from 0·41 to 0·54, and at 1100°C from 0·47 to 0·54 (Fig. 7b, Table 2c).
Temperature dependence
The starting assemblage kyanite + rutile + fluorite ± anorthite + titanite (series F and H) was also run at temperatures of 900°C, 1000°C and 1100°C. Mixture F was run at 2·0 GPa to investigate the temperature dependence of the F–Al substitution in titanite (Table 2b). Along with the starting assemblage, only newly grown F–Al-titanite and quartz were present in the run products; no anorthite was observed. Figure 6d–f shows schematically the predicted shapes of reactions (3)–(5) in a T–XAl diagram. Reaction (3), AFT, will decrease in temperature with increasing Al substitution in titanite, whereas reaction (5), TARK, will increase in temperature with increasing Al substitution in titanite. Reaction (4), TAFT, also has a positive slope, which changes for intermediate values of XAl. Figure 7c shows that at 2·0 GPa XAl decreases from 0·55 ± 0·03 to 0·51 ± 0·01 as temperature rises from 900°C to 1000°C. At 1100°C, XAl is the same as at 1000°C.
Mixture H was run at 1·4, 1·6, 1·8 and 2·0 GPa at 1000°C and 1100°C. In the experiments at 1·6–2·0 GPa, no change in XAl was found (Fig. 7b). Only in the experiments at 1·4 GPa did XAl increase with falling temperatures, from 0·48 to 0·53; this effect is probably due to the change in the assemblage from anorthite at 1100°C to kyanite at 1000°C.
Displacement of the TARK equilibrium through addition of CaF2
The addition of fluorite to the assemblage anorthite + rutile + titanite + kyanite shifts the TARK equilibrium to lower pressures from the experimentally determined reversals (Manning & Bohlen, 1991). In the experiments at 900°C, only minor reaction occurred at 1·2 GPa. In contrast, the reaction direction could easily be determined in the experiment at 1·1 GPa where anorthite overgrew kyanite. Small (<3 µm) dendritic crystals of a Ca–Al-rich phase grew in one run (run 245), possibly indicating a small degree of melting. In many experiments where Al–F-titanite overgrew starting titanite, no sharp grain boundary existed between the two minerals (Fig. 3e and f). Figure 3e shows that some Al–F-titanites display a slight zoning with brighter Ti-enriched cores, containing numerous inclusions of rutile, and darker Ti-poor rims. In the experiments at 1·1 and 1·2 GPa (runs 251 and 262) at 900°C, 1·30, 1·35 and 1·40 GPa (runs 239, 249, 269, and 234) at 1000°C, and 1·55 and 1·60 GPa (runs 248, 266 and 240) at 1100°C, the isobaric or isothermal invariant assemblage Al–F-titanite + anorthite + kyanite + rutile + fluorite remained. In the experiments at 900°C and 1·1 GPa (run 262), 1000°C and 1·30 GPa (run 239), and 1100°C and 1·55 GPa (runs 248 and 266), anorthite overgrows kyanite, indicating the overstepping of the TARK reaction (5), as shown in Fig. 3g. In the experiments above 1·35 GPa at 1000°C and 1·55 GPa at 1100°C, the modal amount of anorthite strongly decreases with the formation of Al–F-titanite and kyanite (Fig. 3f). Below 1·3 GPa at 1000°C and 1·5 GPa at 1100°C, kyanite disappeared (Fig. 3e). Rutile is stable throughout the range of investigation, and tiny, newly formed rutile needles are always found as inclusions in Al–F-titanite. Two experiments with mixture H (runs 243 and 266) also showed small amounts of corundum. Although all experiments contain small amounts of quartz, it occurs only as inclusions in F–Al-titanite, and thus is never found in contact with corundum.
The chemical data from the experiments with assemblage H are shown in Table 2c. The (Al + F) content in titanite does not exceed 0·63. The total variation in XAl is between 0·38 and 0·54 at 900°C, 0·37 and 0·59 at 1000°C, and between 0·42 and 0·63 at 1100°C. Overall, the data show a slight decrease in XAl when titanitess coexists with the assemblage anorthite + rutile + fluorite. Some titanites contain slightly more titanian cores (XAl of 0·4). Kyanite and anorthite were also analyzed (Table 3). Kyanite contains small amounts of F (0·003–0·033 a.p.f.u.) and Ca (0·011–0·019 a.p.f.u.), and its mole fraction of Al [Al/(Al + Ti)] is therefore very close to unity (0·995–0·998). Anorthite shows a slight Al deficiency (1·96 ± 0·02 a.p.f.u.) and also contains small amounts of F (0·06 ± 0·03 a.p.f.u.) and Ti (0·01 ± 0·01 a.p.f.u.). Owing to the small grain size of the newly formed rutile needles, no reliable analyses could be obtained, although previous experimental studies suggest minor Al (0·02 a.p.f.u.) in rutile (Manning & Bohlen, 1991).
Above 1·35 GPa at 1000°C and 1·6 GPa at 1100°C, anorthite rapidly disappears, as does kyanite below these pressures. Because our experiments were conducted at 900°C, 1000°C and 1000°C, the calculated positions of the TARK equilibrium from Holland & Powell (1998) at 1·60 GPa at 900°C, 1·79 GPa at 1000°C and 1·98 GPa at 1100°C were used to determine the extent of the shift. The locus of this curve shifted to 1·1–1·2 GPa at 900°C, 1·30–1·35 GPa at 1000°C and 1·55–1·6 GPa at 1100°C (Fig. 8). Taking the midpoints of the brackets, the displacement in pressure is 0·45 GPa at 900°C, 0·46 GPa at 1000°C and 0·405 GPa at 1100°C. Figure 8 shows equilibrium (5) with activities of CaTiSiO4O in titanite ranging from 1·0 to 0·3, assuming the other phases are stoichiometric, along with the experimental results of this study. The curves were calculated with the updated version of the program THERMOCALC v. 2.7 (T. J. B. Holland, personal communication, 1999) and the database of Holland & Powell (1998). The experiments show that the magnitude of displacement decreases with rising temperature, which requires that the activity of CaAlSiO4F in titanite solid solution buffered by anorthite + kyanite + rutile + fluorite increases with rising temperature.
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 8. The end-member curve of reaction (5) calculated with different titanite activities with the program THERMOCALC v. 2.7 and the database of Holland & Powell (1998). The numbers on the curves indicate the activity of titanite. {dtri}, growth of (Al + F) titanite + kyanite + rutile;  , (Al + F) titanite + anorthite + rutile growth.
|
|
|
DISCUSSION
|
|---|
Our results show that: (1) equilibrium compositions were obtained;
(2) it is possible to synthesize highly aluminous (
XAl >0·5)
F–Al-titanites at 1000°C and 2·0 GPa; (3) a
continuous solid solution exists in titanites in the compositional
range of our investigation (
XAl = 0·33–0·91);
(4) titanite seeds in the starting material are needed for the
nucleation of F–Al-titanite (
Troitzsch & Ellis, 1999).
In addition, compositions of the newly grown titanite solid
solutions show equality of F and Al concentrations in F–Al-titanite,
which indicates that OH substitution for F (vuagnatite-type
substitution), if it is present, is too minor to detect. Product
assemblages included titanite
ss + anorthite + fluorite, titanite
ss + anorthite + kyanite + fluorite, titanite
ss + anorthite + quartz
+ fluorite and titanite
ss + kyanite + rutile + fluorite. The
relationship between the growth of titanite solid solutions
and variable compositions in different starting assemblages
will be discussed below. In the presence of the assemblage titanite
ss + kyanite + rutile + fluorite,
XAl in titanite is fixed at constant
P and
T and varies only slightly with changes in
P and
T.
Constraints on P–XAl and T–XAl phase relations
The phase relations implied by our experiments may be interpreted in terms of isothermal P–XAl and isobaric T–XAl diagrams (Fig. 9a–d). In these diagrams, the positions of reactions (3), (4) and (5) are calculated with the activities obtained above and the compressibility and expansivity data from Table 4. These reactions intersect in an isothermal or isobaric invariant point. The titanite compositions in the presence of rutile, kyanite and fluorite define the position of TAFT. Experiments in which anorthite coexists with fluorite, titanite and either rutile or kyanite constrain the position of AFT. The TARK equilibrium shifts with XAl from the end-member position to the isobaric–isothermal invariant point (Fig. 9).
The position of TAFT is constrained by kyanite-bearing, anorthite-free product assemblages. The TAFT reaction shifts to intermediate compositions because of the capability of mutual solution of titanite and Al–F-titanite phase components. Above 1·2 GPa at 900°C, Al–F-titanite coexists with kyanite and XAl increases to 0·51–0·55 (Fig. 9a). This increase in XAl was also observed in the experiment series with mixtures F and H at 1000°C and 1100°C where XAl ranges from 0·33 to 0·48 in the presence of anorthite and 0·50 to 0·54 in the presence of kyanite (Fig. 9b and c). Thus, there is little change in XAl with pressure in equilibrium with rutile, kyanite and fluorite at the conditions investigated. TAFT therefore has a steep slope in P–XAl and T–XAl space (Fig. 9).
The location of TAFT shown in Fig. 9 is consistent with results of Smith (1981). At 1100°C and 2·0 GPa, Smith (1981) generated a Ca–Al silicate melt coexisting with rutile, fluorite and corundum. In addition, he reported that ‘several’ run products also contained ‘fluorite or quartz or kyanite’ in experiments at 1100–1200°C and 1·5–3·5 GPa. We interpret this to mean that fluorite, quartz and kyanite did not all coexist with rutile and titanite in any of his experiments. The absence of full TAFT assemblage, combined with the presence of melt and corundum in his experiments, probably results, at least in part, from the fluorite-poor, rutile–kyanite-rich bulk compositions he used. Regardless, the lower Al content of titanite reported by Smith (1981) is consistent with the phase relations depicted in Fig. 9b, in that titanite coexisting with rutile, in the absence of kyanite and fluorite, should be less aluminous than titanite buffered by the TAFT assemblage.
Anorthite-bearing, kyanite-free product assemblages constrain the P–XAl and T–XAl geometry and provide an upper pressure bound for the AFT reaction (Fig. 9c). In addition, our unreversed experiments on bulk compositions C–E also provide a maximum P limit on AFT at high XAl (Fig. 9b).
Figure 9a–d shows that the isothermal–isobaric invariant point of reactions (3)–(5) intersects at intermediate XAl of 0·5. The change in assemblage explains the large shift in titanite composition, owing to the shallower slope of this reaction. In addition, the different assemblages and F–Al-titanite compositions at 1·1–1·2 GPa and XAl 0·40–0·50 at 900°C, 1·30–1·35 GPa and XAl 0·45–0·50 at 1000°C, and 1·55–1·60 GPa and XAl 0·50 at 1100°C set a firm constraint on the position of the invariant point (Fig. 9a–d).
Figure 9d shows the inferred phase relations at 2·0 GPa as a function of temperature and XAl. Our experiments at 900–1100°C imply that the titanite compositions in the TAFT assemblage shift to increasing XAl with falling temperature at constant pressure. The small change in the pressure of the invariant point between 900 and 1100°C leads us to suggest that it lies at >1300°C at 2·0 GPa, although it seems reasonable to expect that it will occur at similar XAl as at 900°C, 1000°C and 1100°C. As discussed above, the low-Al titanites coexisting with rutile of Smith (1981) are consistent with the absence of fluorite + kyanite + quartz in his run products.
The compositions of titanite solid solutions associated with different mineral assemblages can be inferred from Fig. 9a–d and can be summarized on P–XAl and T–XAl diagrams by considering coexistence of quartz/coesite with fluorite, aluminosilicate, or rutile (Fig. 10a–c). In the presence of quartz/coesite and fluorite (Fig. 10a), titanite solid solution is stable only in the stippled regions of the relevant P–XAl and T–XAl diagrams. The AFT equilibrium requires that as pressure decreases or temperature rises, the maximum Al in titanite decreases. In the wollastonite–rutile–kyanite ternary, phase relations are uncertain above the titanite–anorthite/F-titanite join because these Ca-rich bulk compositions are not yet fully investigated (Tropper, 1998). With increasing pressure at constant temperature (or falling temperature at constant pressure) the two-phase field (+ fluorite, + SiO2) for rutile + titanite of variable composition expands, and the XAl in titanite coexisting with rutile and anorthite increases to a maximum at the intersection, where the discontinuous TARK reaction limits stability of this assemblage to higher presssure (or lower temperature). Beyond the intersection, kyanite + rutile coexist with titanite of increasing XAl, and the two-phase field titanite + rutile continues to expand. The maximum XAl in titanite coexists with anorthite and fluorite and kyanite, until the stoichiometric AFT reaction (1) is crossed. At higher pressures or lower temperatures, anorthite and fluorite are no longer stable together because of end-member F–Al-titanite stability.
View larger version (37K):
[in this window]
[in a new window]
|
Fig. 10. Schematic P–XAl and T–XAl diagram with compositional ternaries. These diagrams are based on the P–XAl and T–XAl plots from Fig. 9. The labels of the three curves (AFT, TARK and TAFT) are the same as in Fig. 9 and have been omitted for clarity. The diagrams illustrate the range of Al + F substitution in titanitess coexisting with any of the assemblages fluorite + quartz/coesite, kyanite/sillimanite + quartz/coesite and rutile + quartz/coesite, as indicated by the stippled area.
|
|
In the presence of stable Al2SiO5 and SiO2 polymorphs, titanitess is limited to pressure above (or temperature below) the intersection (Fig. 10b). With increasing pressure (falling temperature), the first titanite to form is therefore intermediate, and the compositional range permitted with Al2SiO5 expands as pressure increases further. At a fixed P and T below the end-member TARK and AFT reactions, there are thus three three-phase fields (+ Al2SiO5 and SiO2) with invariant titanite compositions in the fluorite–rutile–anorthite ternary (anorthite + rutile + titanite, rutile + fluorite + titanite, and anorthite + fluorite + titanite) and three two-phase fields in which titanite composition is variable (titanite + anorthite, titanite + rutile, and titanite + fluorite). At higher pressure (or lower temperature), titanite of variable composition can coexist with anorthite or grossular (+ Al2SiO5 + SiO2) but not fluorite, or fluorite or rutile but not anorthite or grossular. The three-phase assemblage rutile + fluorite + titanite is stable to high pressure, but titanite composition is fixed at constant P and T.
Titanite solid solutions coexisting with rutile are always low in XAl, because they must lie to the left of the TAFT equilibrium (Fig. 10c). In this compositional region, the only important reaction encountered with increasing pressure (falling temperature) is the TARK equilibrium, along which titanite composition is fixed at any P and T. The maximum XAl of titanite solid solution in the presence of rutile and SiO2 occurs with fluorite and either anorthite or Al2SiO5.
Figure 11 shows the relative changes in the phase relations in terms of approximate P, T and titanite composition. The small P and T dependence of phase relations in this system implies that the AFT reaction lies at a similar P and T, and has a similar slope, as the TARK reaction, at least at >800°C. This means that the invariant point shifts little in terms of XAl. The main change in Fig. 11 is that the kyanite–sillimanite equilibrium crosses the invariant point between 1000°C and 1100°C.
View larger version (37K):
[in this window]
[in a new window]
|
Fig. 11. Schematic P–T–XAl diagram illustrating phase relations involving titanite solid solution at high temperature and pressure.
|
|
Growth of F-zoisite and quartz
Minor quantities of F-bearing zoisite and quartz grew in all anorthite-rich bulk compositions. F-zoisite was not observed in bulk compositions with high TiO2/CaF2 ratios (series E–H, Table 1). Zoisite, Ca2Al3Si3O12(OH), has the hypothetical fluorian analogue Ca2Al3Si3O12F which, like F–Al-titanite, is collinear with anorthite and fluorite (Fig. 3b) as can be seen from the reaction
However,
because F-zoisite probably contains both F and OH, it may be
stabilized by an additional component, such as H
2O. The presence
of H
2O could have an influence on the stability of FAT by leading
to a solid solution between FAT and the Al–OH end-member
CaAlSiO
4OH (vuagnatite,
McNear et al. 1976) although preliminary
experimental investigations by
Fehr (1991) showed that vuagnatite
is stable only at low temperatures (<290°C at 0·3
GPa, <600°C at 3·0 GPa). The occurrence of H
2O
in the capsule either through moisture in the air or fluid inclusions
in the natural fluorite, allows F
2-producing reactions such
as
Reactions
(8) and (9) are written with representative starting minerals
as reactants. Many reactions were proceding simultaneously in
these experiments, so these reactions are several of the numerous
possible pathways for the production of F–OH-zoisite.
Fluorine contents of analyzed zoisites range from 1·11 to 1·78 wt % or 0·27 to 0·42 F per formula unit (Table 5). The low analytical totals and F <1 imply the presence of additional OH or O. Troitzsch & Ellis (1999) also reported F-zoisite. They confirmed the presence of OH by IR spectroscopy and energy-dispersive analysis for oxygen. IR spectroscopy was not possible with the zoisite from our study because of small crystal sizes (<2–3 µm) and low yields, but we assume that the electric charge imbalance in our zoisites is also accommodated by OH. The growth of minor F- and OH-bearing zoisite implies that small amounts of H2O were incorporated into the charges although the experiments were nominally H2O-free.
There are several possible explanations for the growth of hydrous phases during nominally anhydrous experimental conditions: (1) water contamination during welding; (2) presence of water in the natural minerals; (3) incomplete dehydration or dehydroxylation during drying treatment before experiments; (4) exchange of H2O, H2 or O2 across capsule walls during experiments. Although we used the standard method of drying capsules at 110°C before welding, the welding step itself could lead to reintroduction of some water into the charge. This is because welding of the Ag80Pd20 capsules in air requires rapid conduction of heat away from the weld, which is done by wrapping the capsule in a water-moistened tissue. Water and/or moist air could thus be trapped in the capsule as it is welded shut. Troitzsch & Ellis (1999) suggested that this was the source of water in F-zoisites in their run products.
In addition to contamination during welding, water could have been been introduced into some of our experiments via fluid inclusions in natural phases, particularly fluorite. Run numbers 18–120 used unfired natural fluorite. However, this cannot be the sole source of H2O because experiments using synthetic fluorite, fired at 950°C, also produced F-zoisite. Nonetheless, slightly higher modal abundances of F-zoisite were observed in runs with natural fluorite.
With the exception of the natural fluorite, each starting phase was fired at 950°C, and standard drying measures were applied to the capsules before welding. Nevertheless, it is possible that these treatments do not fully dehydroxylate surfaces of finely ground starting materials. After individual phases are fired at high temperature, water could be adsorbed from the atmosphere despite storage in evacuated bell jars. Drying of the weighed charge before welding may not completely dehydrate mineral surfaces. Given the small quantities of F-zoisite (<<1%), the charges contain less than 0·0005 mg of H2O. As zoisite contains relatively small quantities of H2O for a hydrous mineral, it will form in comparatively higher—that is, noticeable—modal abundances in the presence of only minor H2O contamination.
A final mechanism for forming hydrous phases is diffusion of H–O species across the capsule walls. Our experiments were not externally buffered with respect to the fugacities of these components. (One experiment, run 246, was conducted in a sealed gold capsule with a magnetite–hematite buffer to minimize hydrogen diffusion into the capsule; F-zoisite still grew.) Recent studies have demonstrated H2O loss from hydrous melting experiments in Au or AgPd alloys, implying that this molecule migrates through noble metals (Patiño-Douce & Beard, 1994; Truckenbrodt & Johannes, 1999).
All experiments contain quartz as a run product. This could be due to small amounts of quartz in the synthetic anorthite and titanite. Another possibility could be the formation of quartz through H2O-involving reactions. The presence of H2O either through moisture in the air or fluid inclusions in the natural fluorite allows F2-producing reactions such as reaction (7) to proceed. The initial fO2 of the capsules is high because of the minor Fe2O3 in the natural kyanite and a small amount of O2-bearing air included during welding. Equilibration at high fO2 could yield quartz by a reaction such as
Reaction (10) might occur
in the experiments with mixes F and G (titanite + kyanite +
rutile + fluorite) and could be responsible for the formation
of quartz and rutile inclusions in newly formed F–Al-titanites.
From the discussion above we conclude that the production of F-bearing zoisite and quartz is readily explained in light of the potential sources of H–O species in the experiments. As F-zoisite is evidently hydrous, whereas F–Al-titanite is not, we infer that zoisite is stabilized by the unanticipated presence of H–O species. However, the low modal proportion of this phase indicates that the concentration of H2O was low.
Activity–composition relations in titanite solid solutions
The activity of a species in a given phase can be obtained from the displacement of an end-member reaction curve by dilution of one or more participating phases (Wood & Fraser, 1977; Schmid et al., 1978; Wood, 1988). In this study, titanite was diluted with the (Al + F) titanite component, which shifted the position of reaction (5) to lower pressures by about 0·4 GPa at 1000°C and 1100°C. If the experiments are carried out at constant temperature, the activity of CaTiSiO4O in (Al + F) titanite can be obtained from the relations
where
and where
Vr is
the average volume change of reaction (5) between the conditions
of the end-member curve and our brackets. The activities of
Al
2SiO
5 in kyanite and CaAl
2Si
2O
8 in anorthite were approximated
with ideal ionic models, and are so close to ideal (0·994–0·998)
that within error they are unity. As no analytical data for
rutile were available, the activity of TiO
2 in rutile was also
assumed to be unity, consistent with the low Al contents (
XTi = 0·98) observed by
Manning & Bohlen (1991). Based
on these assumptions, equation (12) reduces to
Assuming the volume change is constant over this
pressure interval (0·450 GPa at 900°C, 0·465
GPa at 1000°C and 0·405 GPa at 1100°C) and that
the volume relations in the CaTiSiO
4O–CaAlSiO
4F solid
solution are linear, equations (11) and (12) can be combined
to give
where
PExp is the pressure
of the midpoint of the reversal of this study, involving the
solid solution in titanite, and
P0 is the pressure of the calculated
end-member curve from the dataset of
Holland & Powell (1998).
Solution of equation (14) yields the activity of CaTiSiO4O in (Al + F) titanite solid solutions of 0·42 ± 0·04, 0·48 ± 0·02 and 0·52 ± 0·02 at 900°C, 1000°C and 1100°C, respectively. The uncertainties are 1, arising from the width of the experimental brackets, but are probably underestimated because of the assumptions outlined above. These data indicate that the activity decreases with falling temperature. The activities from this temperature range are also similar to the activities from the mole fractions of Ti (XTi) of 0·49 ± 0·04 at 900°C, 0·49 ± 0·03 at 1000°C and 0·51 ± 0·01 at 1100°C; the activity coefficient of CaTiSiO4O in titanite,
TOP
ABSTRACT
INTRODUCTION
, composition of the starting titanite. The dashed line indicates the compositional trend. The error bars are one standard deviation (1
, the corresponding F–Al-titanite compositions. (b) Compositions in the system anorthite + fluorite + titanite + kyanite ± quartz; •, mixtures D1–D4, E, F, G and H; 
, mix C1 (anorthite + fluorite + anorthite); 
, mix C2 (anorthite + fluorite + anorthite); 
assemblage Al–F-titanite + kyanite + rutile + fluorite; 
assemblage Al–F-titanite + anorthite + fluorite ± rutile ± quartz. 
