An Experimental Study of Fe – Al Solubility in the System Corundum – Hematite up to 40 kbar and 1300

The mutual solubility in the system corundum–hematite [a-(Al, Fe)2O3] was investigated experimentally using both synthetic and natural materials. Mixtures of g-Al2O3 and a-Fe2O3 (weight ratios of 8:2 and 10:1) were used as starting materials for synthesis experiments in air at 800–1300 C with run times of 7–34 days. Experiments at 8–40 kbar and 490–1100 C were performed in a piston-cylinder apparatus (run times of 0 8–7 4 days) using a natural diasporite consisting of 60–70 vol. % diaspore and 20–30 vol. % Ti-hematite. During the diasporite–corundite transformation, the FeTiO3 component (12–18 mol %) of Ti-hematite only slightly increased, implying that oxygen fugacity was maintained at high values. Run products were studied by electron microprobe and X-ray diffraction (Rietveld) techniques. An essentially linear volume of mixing exists in the solid solution with a slight positive deviation at the hematite side. Up to 1000 C, corundum contains <4 mol % Fe2O3 and hematite <10 mol % Al2O3; at 1200 C these amounts increase to 9 3 and 17 0 mol %, respectively. At 1300 C hematite was no longer stable and a-(Al0 88Fe0 12 3þ O3) coexists with the orthorhombic phase Fe0 53 3þ Al0 47O3. The present results agree with corundum (solvus) compositions obtained in previous studies but indicate a larger solubility of Al in hematite. The miscibility gap in the solution can be modelled with an asymmetric Margules equation with interaction parameters (2s uncertainties): WCor 1⁄4 89 10 kJ/mol; WCor 1⁄4 23 8 J/(Kmol); WHem 1⁄4 91 14 kJ/mol; WHem 1⁄4 14 10 J/(Kmol). Application of the corundum– hematite solution as a solvus geothermometer is limited because of the scarcity of suitable rock compositions.


INTRODUCTION
Corundum (a-Al 2 O 3 ) is the dominant mineral in bauxites and Al-rich laterites metamorphosed at medium to high metamorphic conditions (Feenstra, 1985;Yalc ¸in, 1987).It occurs, typically in moderate amounts, in other highgrade metamorphic rocks such as pelitic granulites, Al-rich marbles, and high-pressure mafic and ultramafic rocks (e.g.Enami & Zang, 1988;Morishita et al., 2001Morishita et al., , 2004;;L o opez S a anchez-Vizcaı ´no & Soto, 2002;Ouzegane et al., 2003;Zhang et al., 2004;and references therein).Hematite (a-Fe 2 O 3 ) is isostructural with corundum (space group R-3c); it is common in soils and oxidized metamorphic rocks, where it rarely persists up to high metamorphic grades.During metamorphism of Al-rich laterites and bauxites the oxidized nature of the protoliths is commonly preserved.This may result in the coexistence of the rhombohedral oxides corundum and Al-(Ti)-hematite, thus allowing the study of their mutual solubility relations in natural systems.
Both hematite and corundum are important compounds for the production of industrial materials and ceramics.Several recent studies have focused on the synthesis, including micron-to nanosized particles, of Al-substituted hematite and Fe-substituted corundum using a variety of precursors and methods (e.g.Bouree et al., 1996;Polli et al., 1996;Van San et al., 2001;Wells et al., 2001;Da Costa et al., 2002;Majzlan et al., 2002;Cornell & Schwertmann, 2003).Although 'dry' experiments in the Al-Fe-O system typically suffer from sluggish reaction at temperatures below 1000 C, the system was already investigated in the early days of experimental petrology at 1 bar and up to temperatures of 1725 C (Muan & Gee, 1956;Atlas & Sumida, 1958;Muan, 1958;Turnock & Eugster, 1962).In general, these studies revealed that there is only restricted mutual solubility between corundum and hematite with a wide miscibility gap.The corundum-hematite solvus does not close (in air and 1 bar) because of reduction of Al-hematite to Fe-Al spinel at $1390 C and the formation of intermediate orthorhombic FeAlO 3 at $1300 C (Muan & Gee, 1956).This restricts experimental investigation of the mutual corundum-hematite solubilities to these temperatures.Recently, Majzlan et al. (2002) determined the mixing enthalpies of various Al-hematite and Fe-corundum compositions by high-temperature calorimetry.
The present study re-investigates the Al 2 O 3 -Fe 2 O 3 system at geologically relevant temperatures (up to 1300 C).In particular, we report data for the solubility of Fe in corundum, obtained by using synthetic oxide mixtures as well as natural Ti-hematite-containing diasporite as starting materials.Contrary to the previous studies, we included hydrous high-pressure experiments (up to 40 kbar) in piston-cylinder equipment.In the final part of this paper we will compare the derived corundum-hematite miscibility gap with data reported for natural rocks.

EXPERIMENTAL AND ANALYTICAL METHODS Experimental
Mixtures of g-Al 2 O 3 and a-Fe 2 O 3 were prepared with Al 2 O 3 :Fe 2 O 3 weight ratios of 8:2 and 10:1.We used g-Al 2 O 3 because this is more reactive than a-Al 2 O 3 .The oxide mixtures were pressed to 5-8 mm long circular tablets (Ø ¼ 5 mm) and heated in a three-zone horizontal furnace that was open to air.Experiments were carried out at temperatures ranging from 800 to 1300 C and lasted between 7 and 34 days (Table 1).Experimental temperature was monitored by two Pt-PtRh 87 thermocouples, between which the samples were located.Especially at the lower temperatures, run products were reground several times to promote reactivity.Runs were terminated by pulling the holder with samples and thermocouple into the cold part of the tube furnace; this resulted in cooling of the samples to <200 C within 30-40 s.

Electron microprobe analysis
Chemical compositions of corundum and hematite in all 30 experiments were determined by wavelengthdispersive electron microprobe analysis (EMPA) using a CAMECA SX-100 electron microprobe.The machine was operated at 15 kV with a beam current of 20 nA and spot size of 1-2 mm.Counting times were 30-60 s on peaks; backgrounds were counted for 15-30 s.The following synthetic minerals were used as standards: corundum (Al Ka); hematite (Fe Ka); eskolaite (Cr Ka); wollastonite (Si Ka, Ca Ka); rutile (Ti Ka), periclase (Mg Ka); pyrophanite (Mn Ka, Ti Ka).The spectrometer intensity data were corrected with the PAP program (Pouchou & Pichoir, 1985).

X-ray diffraction
Run products of all 1 bar experiments and six pistoncylinder experiments were investigated by X-ray powder diffraction (Table 1).The patterns were collected using a STOE STAPI P diffractometer in the 2Q range 5-125 for CuK a1 radiation.For quantitative phase analyses and for determination of the cell dimensions, Rietveld analyses were performed using the GSAS software package (Larson & Von Dreele, 1987).Initial structure models for corundum, hematite and FeAlO 3 used the structural parameters of Oetzel & Heger (1999), Blake et al. (1966) and Bouree et al. (1996), respectively.The lattice parameters of synthesized corundum and hematite were easily refined with the Rietveld method; c 2 ranged from 1Á0 to 1Á5 and the Durbin-Watson factor from 1Á8 to 1Á1.Although this indicates that the calculated standard deviations are reliable, uncertainties for lattice parameters listed in Table 1 are standard deviations multiplied by three.

RESULTS AND DISCUSSION
'Dry' vs hydrous synthesis of (Al,Fe) 2 O 3 phases As mentioned above, two types of synthesis experiments have been performed.The first type involved the synthesis of (Al,Fe) 2 O 3 phases from oxide mixtures in 1 bar   87 (4) 5 .  0332 (4) 13 .   763 (2) 301 .94 (5) Values in parentheses for compositional data represent one standard deviation of the mean in terms of the last digit(s) given; those for unit cell data three standard deviation of the mean in terms of the last digit(s) given.n.d., not determined; n.p., not present.
furnace experiments in air.Corundum in these experiments is generally xenoblastic; pores are ubiquitous within and between the grains (Fig. 1a and b), complicating EMPA.The grain size of corundum and hematite in the 'dry' 1 bar syntheses is commonly <20 mm; at T !1000 C local grain coarsening occurred, leading to porous corundum grains of up to 60 mm in diameter (Fig. 1a).The grain size of Al-hematite slightly increases with run temperature and reaches 30 mm at the highest temperatures.In contrast to corundum, the hematite grains are nearly devoid of pores.
The second type of experiments involved the hydrous synthesis of corundum and coexisting Ti-Al-hematite using a natural diasporite as starting material.These piston-cylinder runs have been performed in the context of a larger experimental study investigating phase relations and metasomatic interactions in a reacting diasporite-dolomite rock system (Feenstra & Wunder, 2002, and in preparation).Corundum in the hydrous syntheses forms sub-idioblastic crystals displaying a platy hexagonal habit (Fig. 1c and d).Even at the lowest synthesis temperatures (e.g.run AFD07: 8 kbar, 490 C; see Fig. 1c) the grain diameter is 10-30 mm with a crystal length of up to 50 mm.The corundum grains increase to 50-80 mm in size in the high-temperature runs.Ti-Al hematite grains are xenoblastic and 10-40 mm in diameter.Optical studies and EMPA indicate that the hematite occasionally contains lamellae of rutile, whereas lamellae or blebs of ilmenite are generally lacking.Individual hematite grains are homogeneous but slight differences in the ilmenite (FeTiO 3 ) component of hematite exist between grains in different parts of the experimental charge in some runs (e.g. in AFD02 and 05).In such cases the hematite in the external part of the charge (close to the gold capsule) has the highest FeTiO 3 component.
Inspection of Table 1 shows that all hydrothermally grown Fe-corundum is homogeneous in composition within the analytical uncertainty of the EMPA method, whereas 'dry' 1 bar synthesis at 800 C probably failed to produce equilibrium compositions during run times of 28 days.Comparison of runs Cor1b and Cor7, both at 1 bar and 1000 C, demonstrates that increasing the run duration of 7 towards 21 days (with several regrindings) results in a significantly higher Fe 3þ /(Fe 3þ þ Al) for Fe-corundum (Table 1).For synthesis temperatures <1000 C, Al-hematite produced by the 'dry' method is in part inhomogeneous (see Table 1).In the hydrous experiments Al saturation in hematite appears to be lacking in most runs up to 800 C; at higher temperatures the Fe 3þ /(Fe 3þ þ Al) ratio of hematite is uniform with compositional uncertainties inherent to EMPA.Regarding the hydrous syntheses it should be noted that, contrary to corundum, which formed from diaspore, Tihematite was already present in the diasporite from Samos.The initial Ti-hematite contained 12-18 mol % FeTiO 3 component in solution whereas its Al 2 O 3 content was <0Á10 wt %.After the experiments the FeTiO 3 component of Ti-hematite ranged from 12 to 32 mol % with 20 mol % FeTiO 3 in 15 of 17 runs (Table 1).Because the Ti-hematite composition did not change considerably, a strong potential to drive recrystallization and Al saturation in Ti-hematite did not exist during the experiments.This could explain why in the hydrothermal runs Ti-hematite failed to become saturated with respect to Al at temperatures 800 C, whereas corundum probably did with respect to ferric iron.
Comparison of the overall results of the 'dry' and hydrothermal synthesis demonstrates that reactivity is greatly improved in the latter case.This is obvious from better chemical homogeneity (Table 1) and coarser grain size, particularly of corundum, in the hydrothermal runs.The difference in reaction kinetics may result not only from the presence of water as a catalytic reaction medium but also from the positive influence of high pressure on mineral growth.

Orthorhombic FeAlO 3 phase
At 1300 C and 1 bar, we synthesized orthorhombic FeAlO 3 with a molar Fe 3þ /(Al þ Fe 3þ ) ratio of 0Á532 AE 0Á012 coexisting with Fe-corundum with molar X Fe of 0Á115 AE 0Á006 (run Cor6b; Table 1 and Fig. 1b).Our synthesis temperature of 1300 C is somewhat lower than the 1318 C reported by Muan & Gee (1956) and Muan (1958) as a minimum temperature for producing FeAlO 3 from Al 2 O 3 -Fe 2 O 3 mixtures.Using bulk compositions with molar Fe 3þ /(Al þ Fe 3þ ) between 0Á30 and 0Á50, Polli et al. (1996) synthesized orthorhombic FeAlO 3 by pyrolysis of Al-and Fe 3þ -nitrates at temperatures as low as 700 C. The poorly crystalline orthorhombic phase, coexisting with g-(Al,Fe)O 3 solid solution, probably formed metastably.During subsequent annealing at 900-1000 C the orthorhombic phase disappeared and the run products recrystallized to stable a-(Al,Fe)O 3 solid solutions.Polli et al. (1996) succeeded in synthesizing stable orthorhombic FeAlO 3 at 1350 C, which is in agreement with the phase diagram of Muan & Gee (1956).Majzlan et al. (2002) used a similar synthesis temperature to produce FeAlO 3 for their calorimetric studies.On the basis of thermodynamic analysis they argued that FeAlO 3 is unstable at 298 K with respect to corundum and hematite, and becomes entropy-stabilized at high temperatures.
Volume-composition relations for (Al,Fe) 2 O 3 Unit cell parameters of corundum and hematite determined by XRD are plotted versus chemical composition determined by electron microprobe in Fig. 2. Fe-corundum shows a linear relationship between composition and unit cell parameters in agreement with the findings of Majzlan et al. (2002).The volumes of our high-P Fe-corundum, grown hydrothermally at 650-950 C, tend to fall somewhat below the ideal line of mixing.Considering the compositional uncertainty (Table 1) this tendency is hardly significant.Hydrothermal Fe-corundum grown at 35 kbar and 1100 C plots again in line with 'dry' Fe-corundum synthesized from oxide mixtures at 1 bar and similar temperature.
Al-hematite shows a slight positive deviation from an ideal volume of mixing, particularly expressed by the a parameter (Fig. 2a).The volume-composition (V-X) relationships of our Al-hematites synthesized at 1 bar and temperatures ranging from 800 to 1200 C (Table 1) are in good agreement with the data of Majzlan et al. (2002) for high-T, Al-hematite (annealed at 1350 C and 1 bar).Volumes of low-T, Al-hematite prepared by Majzlan et al. (2002) at 702 C (and 1 bar) by thermal decomposition of Al-goethite and those of the most aluminous hematites from the studies of Van San et al. (2001) and Wells et al. (2001) display larger positive deviations from ideal mixing than our Al-hematite (Fig. 2).In the last two studies Al-hematite was synthesized from lepidocrocite (at 1 bar, 652 C) and Al-ferrihydrite gel (at 1 bar, 700 C), respectively.
Investigations of Al-substituted hematite produced at relatively low temperatures indicate that the unit cell edge a increases with decreasing synthesis temperature, whereas c remains rather constant [for a summary, see Cornell & Schwertmann (2003)].This relationship is attributed to incorporation of structural OH À in the hematite lattice at the lower temperatures.The hydrogen may have been inherited from the precursor phase.The loss on ignition of such hematites can be considered as a measure of initial OH À contents.After heating to 1000 C their a parameters converge to a systematic relationship with Fe 3þ /(Al þ Fe 3þ ) as shown by our 1 bar data and the high-T data of Majzlan et al. (2002).This relationship is obtained irrespective of synthesis method (Stanjek & Schwertmann, 1992;Cornell & Schwertmann, 2003).
Because V-X relationships in the hematite-ilmenite solid solution are non-linear and complex (e.g.Harrison & Redfern, 2001), it seems unjustified to project the ternary volume data on the hematite-corundum binary.Therefore, we do not show lattice parameters for our high-P hematite in Fig. 2.
Al 2 O 3 -Fe 2 O 3 system Figure 3 depicts the chemical composition of corundum and hematite as a function of synthesis condition (see Table 1).We have not plotted the high-T, 1 bar experiments with starting Al 2 O 3 :Fe 2 O 3 weight ratio of 10:1, in which corundum could not achieve Fe saturation and Al-hematite occurs in trace amounts (runs Cor4a and 2a; see Table 1).Likewise, Al-(Ti)-hematite compostions (runs with T 800 C) are not shown when it is suspected that the hematite is not saturated with Al [in Table 1; Fe 3þ /(Fe 3þ þ Al) !0Á984].Corundum, on the other hand, is chemically homogeneous in all hydrothermal experiments and its Fe 3þ /(Fe 3þ þ Al) increases systematically with run temperature.Hence corundum compositions are all depicted in Fig. 3.
Although oxygen fugacities were not exactly similar in the 1 bar (in air) and hydrothermal experiments, we do not observe a systematic dependence of the Fe 3þ / (Fe 3þ þ Al) of corundum on experimental approach used.Also, our results using a 'sliding' Ti-hematite þ rutile internal buffer are in excellent agreement with the hydrothermal results of Turnock & Eugster (1962), who used the magnetite-hematite buffer (Fig. 3).Furthermore, Muan & Gee (1956), who studied the Al 2 O 3 -Fe 2 O 3 system both in air and at 1 bar oxygen pressure, found only minor topological differences between the two f O 2 conditions.Thus it seems that oxygen fugacity in all these studies was high enough to allow 'maximum' solubility of ferric iron in corundum for the prevailing P-T conditions.Consequently, we treated our experimental data independently of oxygen fugacity in the thermodynamic data analysis.
Assuming that compositions of coexisting Fecorundum and Al-hematite in the synthesis experiments  Thompson, 1967;Chatterjee, 1991;Anderson & Grerar, 1993).The PW v i term was neglected because the composition-volume relations for synthesized a-(Al,Fe 3þ ) 2 O 3 indicate a negligible excess volume (Fig. 2).Furthermore, the chemical compositions of coexisting Fe-corundum and Al-hematite do not vary systematically (at constant temperature) with pressure for the 0Á001-40 kbar range studied.The experimental data, weighted according to their compositional uncertainty (Table 1), were fitted with the program Mathematica (Wolfram Research, Inc., 2003) using pairs of coexisting Fe-corundum and Al-hematite.An asymmetric Margules model , where X Hem ¼ Fe 3þ /(Fe 3þ þ Al), excellently reproduces the experimental data (Fig. 3) with the following interaction parameters (2s uncertainties): Hem ¼ 91 AE 14 kJ/mol; W S Hem ¼ 14 AE 10 J/K mol.Symmetric or T-independent models (only W i H ) inadequately fitted the data.
Our results for Fe solubility in corundum concur with previous experimental studies (Muan & Gee, 1956;Atlas & Sumida, 1958;Turnock & Eugster, 1962).The first two studies of the Al 2 O 3 -Fe 2 O 3 join were performed at 1 bar in air and temperatures !1000 C. In their extensive study of Fe-Al oxide phase relationships below 1000 C, Turnock & Eugster (1962) hydrothermally studied the Al 2 O 3 -Fe 2 O 3 join at 2 and 4 kbar using the hematite-magnetite buffer.Their results for the corundum limb of the solvus are in excellent agreement with our low-T, hydrothermal experiments.For the hematite limb they determined, like Atlas & Sumida (1958), a higher Al solubility in hematite than we obtained (Fig. 3).On the other hand, Muan & Gee (1956) obtained for the range 1000-1350 C a lower Al solubility in hematite than indicated by our study.It should be kept in mind, however, that these 40-50-year-old experimental studies preceded routine electron microprobe analysis.Phase compositions were normally determined by XRD measurement using a single reflection.Using the (124) reflection of a-(Al,Fe 3þ ) 2 O 3 , Turnock & Eugster (1962) Majzlan et al. (2002, MNE2002).Majzlan et al. (2002) determined the enthalpies of mixing of various Fe-corundum and Al-hematite solid solutions by high-temperature oxide-melt calorimetry (Fig. 4).Excluding the Al-hematites synthesized by thermal decomposition of Al-goethite at 700 C, which contained considerable amounts of H 2 O, they fitted the calorimetric results with a symmetric Margules equation with W H ¼ 116 AE 10 kJ/mol and W S ¼ 32 AE 4 J/K mol.Compared with our results the resulting miscibility gap agrees well at the corundum limb but is clearly wider at the hematite limb of the solvus (Fig. 3).With regard to the experimental study of Muan & Gee (1956), Majzlan et al. (2002) noted that the deduced W parameters tend to overestimate the crest of the solvus, the exact position of which is unknown because of instability of corundum þ hematite at such conditions.Fitting the phase diagram of Muan & Gee (1956) with an asymmetric T-independent Margules model, Majzlan et al. (2002) 4 compares our mixing model for the corundum-hematite solid solution with that of Majzlan et al. (2002) and shows that within experimental uncertainties the calorimetric data also agree with our model.Majzlan et al. (2002) argued that the positive excess entropy required to model the miscibility gap may relate to increasing ordering for compositions approaching that of the orthorhombic FeAlO 3 phase.For dilute compositions close to the end-members (at low T ) ordering is probably of little importance.

COMPARISON WITH NATURE
Corundum has been documented in a variety of metamorphic and magmatic rocks covering a wide field of P-T conditions ranging from the middle crust to those of the diamond stability field.Many of these occurrences are found in Fe-poor rocks commonly devoid of Fe-(Ti)oxides with ferric iron.Such corundum typically contains <0Á5 wt % Fe 2 O 3 and is of little significance for comparative purposes with the present study.(Majzlan et al., 2002).
Owing to their Al-Fe-rich and oxidized nature, the rare occurrences of metamorphosed bauxite and Al-rich laterite are favourable to provide information on maximum Fe solubility in corundum for the prevailing metamorphic conditions.The highest Fe 2 O 3 content (9Á17 wt %) in natural corundum was probably reported by Agrell & Langley (1958) from such a geological environment.Those workers described yellow corundum occurring in porcellanite at a contact of a dolerite intrusion at Tievebulliagh (Northern Ireland).The corundum is associated with hematite, pseudobrookite and ironmullite, indicating a very high oxygen fugacity during contact metamorphism of a lateritized basalt.Contact metamorphic temperatures were estimated to be between 900 and 1150 C. Assuming that the 9Á17 wt % Fe 2 O 3 (molar X Fe ¼ 0Á061) represents solvus composition, we obtain a temperature of $1050 C when applying our corundum-hematite solution model (Fig. 3).
Another example of metabauxitic Fe-corundum containing up to 2Á0-2Á9 wt % (1Á2-1Á9 mol %) Fe 2 O 3 was documented by Altenberger et al. (1992) and Adusumalli (1994) from the Odenwald complex (Germany).Here corundum þ sillimanite þ magnetite/ilmenite felses, interpreted to be metamorphosed Al-rich palaeosols, are locally present along the contact of gabbros.Using our solvus data, the Odenwald corundum yields a temperature of 650-800 C.This is lower than the 860-900 C deduced from Fe-Ti-oxide thermometry for the ( $362 Ma) Seeheim contact metamorphism (Altenberger et al., 1992) but falls in the range of estimated regional metamorphic conditions (Adusumalli, 1994) for the corundum occurrences in the Odenwald complex.
Corundum from the metamorphic bauxites of Naxos (Greece) contains 0Á5-0Á8 mol % Fe 2 O 3 in solid solution at the highest P-T conditions ( $6 kbar, $700 C) reached during regional metamorphism on the island (Feenstra, 1985;A. Feenstra, unpublished results, 1983A. Feenstra, unpublished results, -1985)).Such amounts of Fe 2 O 3 are lower than implied by the present experiments for a similar temperature (Fig. 3, Table 1).The difference may indicate that at the f O 2 conditions of coexisting magnetite þ (hemo)ilmenite in the Naxos metabauxites, there is considerably less Fe 3þ solubility in corundum than in the more oxidized experiments.Alternatively, it is possible that the Naxos corundum is not saturated with respect to Fe 3þ for the prevailing amphibolite-facies conditions.Corundum in the Naxos metabauxites formed during regional metamorphism by breakdown of diaspore (at $6 kbar and $420 C) and initially coexisted with Ti-hematite þ rutile (Feenstra, 1985;Urai & Feenstra, 2001).The corundum from near the corundum-in isograd is comparable in Fe 2 O 3 contents with that of the highest grade zones, and a correlation between Fe 2 O 3 in corundum and metamorphic temperature on Naxos is not found (Feenstra, 1985;A. Feenstra, unpublished results, 1983A. Feenstra, unpublished results, -1985)).It is thus possible that corundum preserved its initial Fe 3þ /(Fe 3þ þ Al) and during progressive metamorphism (rising T ) failed to adapt compositionally owing to its sluggishness to equilibrate.
The minor and trace element chemistry of gem-quality corundum has been the subject of several detailed studies.Garland (2002) investigated alluvial sapphire deposits of western Montana and compared the sapphire chemistry with that of world-wide occurrences of corundum types of different metamorphic and magmatic origin.The study revealed that corundum associated with alkali basalts is highest in iron of all genetic types containing on average 7000 ppm Fe.For example, corundum from Shandong (China) and Australia (NSW regions and Queensland) contains up to 12 000-13 000 ppm Fe ( $1Á8 wt % Fe 2 O 3 ).When applying our solvus data, even such Fe concentrations give unrealistically low temperatures for basaltic melts, implying that physico-chemical conditions (e.g., f O 2 activity of Fe 3þ ) in the magmatic systems must differ considerably from those in our experiments imposing 'maximum' Fe solubility in corundum.The iron content of gem-quality corundum from metamorphic origin widely varies but is typically lower than that of the basaltic type.Hunstiger (1988) investigated the mineral chemistry and petrogenesis of a variety of ruby-type corundumcontaining metamorphic rocks including amphibolites, Al-rich gneisses and anatexites, and metacarbonate rocks.Whereas corundum in marbles is very poor in iron, the Fe 2 O 3 content of corundum in the other rock types may be up to 0Á8 wt %.Such concentrations are low compared with our experimental data considering the amphibolite-to granulite-facies conditions to which the rocks have been subjected.As discussed above, the difference probably reflects much lower activity of Fe 3þ and/or lack of chemical equilibrium in the natural rocks.

CONCLUSIONS
(1) The present experimental study confirms the results of classical (40-50 years old) investigations of the corundum-hematite system (at low P ), which demonstrated that there exists a wide miscibility gap that does not close.Up to pressures of 40 kbar there is a negligible effect of pressure on the width of the solvus expressed in an insignificant excess volume for the solid solution.Our results for 'maximum' Fe solubility in corundum as a function of temperature are in excellent agreement with those of the previous studies; however, we determined a larger Al solubility in hematite.
(2) The corundum-hematite solvus has limited potential as a geothermometer bcause of the scarcity of suitable rock types for application and the restricted Fe 3þ -Al exchange in the solution at T 1000 C that introduces large inaccuracy in T. The sluggish nature of corundum to re-equilibrate may also hamper application.Geothermometric application of Al solubility in hematite is even more difficult because of the extreme scarcity of hematite coexisting with corundum in metamorphic and magmatic rocks.In addition, the complex mineral chemistry of natural hematite complicates application.Nevertheless, we presented some examples of high-grade bauxitic or lateritic rocks for which realistic T estimates were obtained using our experimental calibration.

Table 1 :
Experimental conditions, chemical compositions and volume data of synthesized(Al,Fe 3þ indicated a compositional accuracy of AE3 wt % Al 2 O 3 .The accuracy of these older phase diagrams therefore should be judged in this context.Chemical compositions of Fe-corundum and Al-hematite as a function of synthesis temperature and pressure (see text).Fe-corundum in our run Cor6b (1 bar, 1300 C) coexists with orthorhombic Fe 1Á06 Al 0Á94 O 3 instead of Al-hematite.Results of experiments of Turnock & Eugster (1962, T&E1962) at 2 and 4 kbar and Atlas & Sumida (1958, A&S1958) at 1 bar are shown for comparison.The dotted curve shows the twoparameter Margules model fitted to our data (W H Cor ¼ 89 AE 10 kJ/mol; W S Cor ¼ 23 AE 8 J/(K mol); W H Hem ¼ 91 AE 14 kJ/mol; W S Hem ¼ 14 AE 10 J/(K mol); the curve with crosses depicts the symmetric solution model (W H ¼ 116 AE 10 kJ/mol; W S ¼ 32 AE 4 J/(K mol) of 3+ /(Fe 3+ + Al) (mole ratio) 3 Fig. 3.