| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology Volume 42 Number 9 Pages 1595-1620 2001
© Oxford University Press 2001
The Distribution of H2O between Cordierite and Granitic Melt: H2O Incorporation in Cordierite and its Application to High-grade Metamorphism and Crustal Anatexis
DEPARTMENT OF GEOLOGY AND GEOPHYSICS, UNIVERSITY OF EDINBURGH, WEST MAINS ROAD, EDINBURGH EH9 3JW, UK
Received June 19, 2000; Revised typescript accepted February 22, 2001
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONExperiments defining the distribution of H2O [Dw = wt % H2O(melt)/wt % H2O(crd)]) between granitic melt and coexisting cordierite over a range of melt H2O contents from saturated (i.e. coexisting cordierite + melt + vapour) to highly undersaturated (cordierite + melt) have been conducted at 3–7 kbar and 800–1000°C. H2O contents in cordierites and granitic melts were determined using secondary ion mass spectrometry (SIMS). For H2O vapour-saturated conditions Dw ranges from 4·3 to 7 and increases with rising temperature. When the system is volatile undersaturated Dw decreases to minimum values of 2·6–5·0 at moderate to low cordierite H2O contents (0·6–1·1 wt %). At very low aH2O, cordierite contains less than 0·2–0·3 wt % H2O and Dw increases sharply. The Dw results are consistent with melt H2O solubility models in which aH2O is proportional to Xw2 (where Xw is the mole fraction of H2O in eight-oxygen unit melt) at Xw
0·5 and 0·25kw{exp[(6·52 - (2667/T)) x Xw]} at Xw > 0·5, coupled with cordierite hydration models in which aH2O is proportional to n/(1 - n), where n is the number of molecules of H2O per 18-oxygen anhydrous cordierite formula unit (n < 1). Combination of our 800–1000°C cordierite H2O saturation results with previous cordierite hydration data leads to the following geohydrometer relation, applicable for temperatures in the range 500–1000°C:
 |
KEY WORDS: cordierite; granite; melting; metamorphism; SIMS
|
INTRODUCTION |
|---|
Fluids and melts interacting with hot rocks during metamorphism and melting at mid- and deep-crustal levels are intrinsic to the evolution of metamorphic belts and the transfer of heat and matter in the crust. As a consequence, it is essential that a thorough understanding of the fluid regime of high-grade metamorphism is developed and constrained using a broad variety of approaches and techniques. The timing and extents of any fluid infiltration, whether by carbonic (CO2-rich), hydrous or saline fluids (e.g. Newton et al., 1980
; Santosh et al., 1990; Shmulovich & Graham, 1996; Buick et al., 1998), and prevalence of melt-present but fluid-absent conditions during the course of high-grade metamorphism (e.g. Waters & Whales, 1984; Waters, 1988; Barbey et al., 1990; Stevens & Clemens, 1993) are important factors influencing the behaviour of the deep crust during major tectonic events.
The high temperatures (>800°C) attained in many granulite terrains (Harley, 1989) ensure that some melt has had to be present in equilibrium with solid phases in most metapelitic rocks initially containing biotite, which has been shown experimentally to undergo extensive dehydration melting in the 800–900°C interval for a range of bulk-rock compositions (e.g. Le Breton & Thompson, 1988; Vielzeuf & Holloway, 1988; Puziewicz & Johannes, 1990; Patiño Douce & Johnston, 1991). This general perception is supported by studies of several high-grade terrains in which biotite dehydration-melting reactions have been deduced on the basis of textural (e.g. Vernon & Collins, 1988), petrological (e.g. Waters & Whales, 1984; Bhattacharya & Sen, 1986; Young et al., 1989) and geochemical (e.g. Waters, 1988; Barbey et al., 1990; Holtz & Barbey, 1991) observations from discordant leucosomes and segregations in pelitic gneisses. To understand and quantify the role and influence of melting in granulite generation and probe further into the links between anatexis as manifested in migmatites and felsic magmatism in orogenic belts (Clemens & Vielzeuf, 1987; White & Chappell, 1988; Stevens & Clemens, 1993), it is important to monitor any variations in volatile composition and abundance and deduce accurately the fluid contents of the melts that have migrated within or moved out of the region of high-temperature metamorphism. The melts themselves have crystallized, escaped, fractionated or been affected by post-crystallization metamorphism and so a proxy monitor phase of fluid content is required. This proxy would complement the indirect approaches, such as stable isotopic signatures and mineral equilibria calculations, that are commonly employed to evaluate the extents of fluid interaction and activities of fluid species.
The mineral cordierite, (Mg,Fe)2Al4Si5O18.(nH2O,mCO2), potentially provides such a direct mineralogical monitor. Cordierite is able to incorporate both the major volatile species in crustal metamorphism, CO2 and H2O, within the channels in its structure and hence can yield information on the relative proportions and contents of those species in any coexisting fluid or melt (Goldman et al., 1977; Armbruster & Bloss, 1980, 1982; Johannes & Schreyer, 1981; Kurepin, 1984; Schreyer, 1985; Le Breton, 1989; Vry et al., 1990). Furthermore, cordierite is a common phase in a variety of metapelitic assemblages and may be produced over a wide pressure (P)–temperature (T) range (3–11 kbar and 600–950°C) through a variety of processes involving rather different fluid regimes, including subsolidus reactions, melting, and metasomatic fluid infiltration. Cordierite-bearing S-type granitoids and upper amphibolite to granulite facies migmatites furthermore indicate that cordierite can occur in equilibrium with peraluminous felsic melts, produced either as a peritectic phase during melting or as a phenocrystic phase during the course of melt crystallization (e.g. Vry et al., 1990; Fitzsimons, 1994, 1996; Harley, 1994; Stevens et al., 1995; Carrington & Harley, 1996).
In fluid-present situations, the absolute abundance of channel volatiles and XCO2 [CO2/(CO2 + H2O)] in cordierite is a function of pressure, temperature and XCO2 in the coexisting fluid (Johannes & Schreyer, 1981; Kurepin, 1984). Hence knowledge of the volatile composition in cordierite provides constraints on the metamorphic fluid composition (e.g. Armbruster & Bloss, 1982; Armbruster et al., 1982; Schreyer, 1985; Vry et al., 1990). However, in cases where a free volatile phase is not present but temperatures are high (>750–800°C: Le Breton & Thompson, 1988; Vielzeuf & Holloway, 1988; Patiño Douce & Johnston, 1991), the absolute abundance of these channel volatiles depends instead upon partitioning of H2O and CO2 between the cordierite and a coexisting melt phase of broadly granitic composition (Harley, 1994; Stevens et al., 1995; Carrington & Harley, 1996). Measurements of the total volatile contents and compositions of cordierite formed at known P–T therefore have the potential to discriminate between fluid-present conditions and situations in which melt is present but a free H2O–CO2 fluid phase absent and provide quantitative fluid activity estimates to support textural and phase equilibrium approaches. The key prerequisites for this are the availability of (1) reliable and internally consistent experimental data defining the H2O, and H2O–CO2, saturation surfaces for cordierite and (2) related experimental data on the volatile contents of cordierite coexisting with fluid-undersaturated melt (Fig. 1; see Carrington & Harley, 1996) that define the partitioning in the undersaturated system and how this varies with P–T. Of course, application to metamorphic belts also requires spatial data on volatile distribution or zoning in natural cordierite so that the significance of any H2O or CO2 leakage from or uptake into the channels in cordierite during post-formation re-equilibration can be assessed (Lepezin & Melenevsky, 1977; Zimmermann, 1981; Jochum et al., 1983; Schreyer, 1985; Mirwald et al., 1986).
|
This paper presents the results of a polybaric–polythermal experimental study of the partitioning of H2O between cordierite and coexisting melt that ranges from vapour-saturated to highly undersaturated conditions (Fig. 1), and hence contributes to objectives (1) and (2) in the specific case of the CO2-free system. Thermodynamic modelling of the experimentally determined H2O contents of cordierite and melt allows calculation of water activities from cordierite-bearing assemblages and is used to improve present calibrations of the P–T dependence of saturation H2O contents (isohydrons) in Fe–Mg cordierites. The experimental calibrations are then used to constrain the water activities relevant to high-grade melting and metamorphism both for examples of key mineral reactions constrained from independent experimental data and grids, and for selected granulite and migmatite terrains for which cordierite volatile content data are available.
|
BACKGROUND AND EXPERIMENTAL RATIONALE |
|---|
H2O solubility in felsic and peraluminous melts
The maximum solubilities of H2O in eutectic melts within the Qz–Ab, Qz–Or, haplogranitic and simplified peraluminous granite systems are now well defined from several studies (e.g. Johannes & Holtz, 1990
; Holtz et al., 1992a, 1992b; Pichavant et al., 1992). Maximum H2O contents increase strongly with pressure but are only weakly dependent on temperature, and indicate that solubility is also to some extent affected by melt Ab–Or ratio and Al2O3 content.
At specified P–T conditions greater than the wet solidus hydrous melts can in principle have H2O contents anywhere between a maximum that is attained when any additional H2O forms a free fluid phase (L + V equilibrium: aH2O = 1), and a minimum H2O content at which the granitic melt crystallizes quartz and alkali-feldspar (L + Crystals equilibrium; Holtz & Johannes, 1994; Johannes & Holtz, 1996); only on the wet and dry solidi are the maximum and minimum H2O contents identical. This is a key feature of the melt system relevant to the H2O-undersaturated experiments carried out in this study, which spans both the H2O-saturated and -undersaturated portions of compositional space defined in the melt–cordierite–H2O system (Fig. 1). Melt H2O contents at the minimum boundary are difficult to measure experimentally as melt–crystal mixtures are produced, but are instead calculated using models for the relationships between P, T, aH2O and weight percent H2O in melts (e.g. Burnham & Nekvasil, 1986; Burnham, 1994; Moore et al., 1995, 1998; Papale, 1997). The H2O contents of melts formed in dehydration-melting reactions will also lie at a minimum value for the P–T conditions of melting, but hitherto have only been experimentally constrained using modal mass-balance calculations on run products (Vielzeuf & Holloway, 1988; Patiño Douce & Johnston, 1991; Carrington & Harley, 1995). As a consequence, the Burnham & Nekvasil (1986) approach has been widely used in predicting weight percent H2O in melts formed through dehydration-melting reactions and for estimation of the fertility of metapelitic rocks as sources of granitoid magmas (Clemens & Vielzeuf, 1987; White & Chappell, 1988; Vielzeuf & Clemens, 1992). The mass-balance relations used to predict melt percentages at a given temperature will be affected by the significant and variable H2O content of cordierite in those cases where it is produced as a peritectic phase coexisting with melt (Stevens et al., 1995; Carrington & Harley, 1996).
H2O incorporation in cordierite
The maximum volatile contents of cordierite coexisting with H2O have been experimentally determined for pure Mg-cordierite (Mirwald & Schreyer, 1977; Mirwald et al., 1979; Johannes & Schreyer, 1981), an Mg-rich natural cordierite (Armbruster & Bloss, 1982), and more recently for Fe-cordierite (Boberski & Schreyer, 1990; Mukhopadhyay & Holdaway, 1994). Maximum H2O occupancies of the channels in cordierite are constrained by structure and have been variously modelled (Newton & Wood, 1979; Lonker, 1981; Martignole & Sisi, 1981; Bhattacharya & Sen, 1985; Carey, 1995) to yield isohydrons of weight percent H2O (or of molecules of H2O per 18-oxygen anhydrous formula unit) in cordierite as a function of P–T. Although the models differ in the calculated positions, slopes and curvatures of the isohydrons it is evident that the maximum H2O content of cordierite increases strongly with pressure and decreases with temperature. Indeed, the recent models of Carey (1995), Skippen & Gunter (1996) and Holland & Powell (1998) each reproduce most of the available H2O-system data to within ±0·15 wt % H2O. These models also demonstrate that the saturation H2O contents in the T range 500–800°C do not vary with parameters such as XMg of the cordierite. However, until now, no cordierite H2O data for T >800°C have been incorporated into the saturation models and, in addition, not all of the available lower-T data have been included in the different analyses. Therefore, in this study we have modelled the saturation H2O contents of Fe–Mg cordierites over the T range 500–1000°C, using previous literature data and our new experiments, to derive an improved calibration applicable to all metamorphic and igneous cordierites.
|
EXPERIMENTAL AND ANALYTICAL METHODS |
|---|
Experimental design and procedures
Experiments aimed at defining the distribution of water between cordierite and melt, Dw [= wt % H2O(melt)/wt % H2O(crd)], have been performed at 800, 900 and 1000°C and 3, 5 and 7 kbar in an internally heated gas apparatus at the University of Edinburgh. The experiments have been conducted using the same methods as employed by Carrington & Harley (1996)
. The pressure and temperature uncertainties associated with the apparatus are ±0·2 kbar and ±5°C, respectively (Carrington & Harley, 1995). Up to seven charges could be run in one operation at a particular P–T condition and for run times varying between 250 and 400 h. In most experiments natural Antarctic cordierite (BB3: Brattstrand Bluffs migmatite leucosome; XMg = 0·64; sieve fraction 125–250 mm) has been used (Fig. 2a; see Table 2, below), evacuated of all volatiles before the experiment. A more magnesian cordierite (AMNH: American Museum of Natural History specimen, gem fields, Zimbabwe; XMg = 0·85), also evacuated, has been used in 7 kbar runs and at equivalent conditions to BB3 to evaluate any effects of cordierite XMg on H2O contents. Evacuation of volatiles is accomplished by prolonged stepped heating under high vacuum at temperatures of up to 1200°C, with the sample held for 1 h at each 100°C temperature step below 1200°C, for 2 h at 1200°C, and heated for a minimum of 10 h in total. Secondary ion mass spectrometry (SIMS) analysis of cordierite grain fragments treated using this stepped-heating procedure shows that they are evacuated (i.e. volatile-free) with apparent H2O and CO2 contents at or below the SIMS background levels.
|
|
|
|
The granitic melt has been added in the form of an anhydrous gel (XMg = 0·25) synthesized in the K2O–Na2O–FeO–MgO–Al2O3–SiO2 system. The compositions of the starting gel and cordierite are given in Tables 1 and 2, respectively. Pre-determined quantities of H2O have been added to each of the cordierite–gel charges using a microsyringe so that Dw could be measured for a wide range of H2O contents in the experimental phases. Reversal experiments have been conducted in a similar way, but with the natural BB3 cordierite, which had first been evacuated, re-saturated with pure H2O at 900°C and 5·0 kbar. This procedure produces a starting cordierite H2O content of 1·7 wt %, as measured for sampled grain fragments using SIMS. Additional experiments have been carried out using a geometry in which a gel layer is sandwiched between a layer of evacuated-BB3 cordierite and a layer of pre-saturated BB3 cordierite. A series of such ‘sandwich’ experiments run for durations between 2 h and 2 weeks (Fig. 2b) have been used as part of a time-study to assess the approach to equilibrium in our experiments as a whole.
|
Following the experiments the charges were punctured and then impregnated with epoxy under vacuum to ensure that the run products retained their spatial and textural integrity. Each charge was then mounted so that the length of the capsule, and hence both cordierite and melt layers, could be exposed by grinding. Most charges contain well-defined cordierite-rich and melt-rich layers, retaining their original geometry (Fig. 2a). A few charges show relative movement of the Crd–L boundary, probably through crystal slumping and convective motion in the melt. Each experiment has been examined using reflected light microscopy and scanning electron microscopy [SEM; secondary electron imagery (SEI) and backscattered electron imagery (BEI)] to monitor the textural relations and features.
Analysis of run products
The H2O contents of the experimental cordierite and melt have been analysed using SIMS (Tables 3 and 4), and oxides other than H2O by energy-dispersive SEM and by wavelength-dispersive electron microprobe (EMP) (Tables 1 and 2). An Oxford Instruments scanning electron microscope at the University of Manchester was used, with beam conditions of 1·5 nA and 15 kV, an energy-dispersive spectrometry (EDS)/ZAF-4 analytical correction procedure, and supercooling to -190°C during analysis to minimize loss of alkali counts on the experimental glasses. Analyses obtained using this technique are comparable in terms of the ratios of non-alkali components (i.e. FMAS) with those obtained using EMP on the same samples. The analytical procedure and conditions for the SIMS and EMP analysis are as detailed by Carrington & Harley (1996).
SIMS calibration curves have been obtained in each analytical session using H2O-bearing glasses of appropriate granitic composition and differing but well-known H2O contents (2·1, 3·9 and 5·6 wt %) and natural cordierite standards with 1·56 ± 0·08 wt % and 0·80 ± 0·06 wt % H2O. An O- primary beam at 8 nA current was used in all SIMS sessions. Positive secondary ions were measured (H+, 30Si) with the ion counts determined at an energy offset of 75 V. Each analysis involved 30 cycles of H and Si counts following 5 min burn-in. The mean of the isotope ratios of the last 10 cycles is taken as the final analysis value, and has a statistical counting precision of better than 1% relative. Analyses are expressed as isotopic ratios of 1H/30Si and converted to weight percent H2O by comparison with standard cordierites or melts. A similar routine has been used in H2O–CO2 analysis of natural cordierites and of experimental run products except that SIMS is set up to measure negative secondary ions (H-, 12C, 28Si). In this case the analyses, expressed as isotopic ratios of 1H/28Si and 12C/28Si, are converted to weight percent H2O and CO2 respectively by comparison with standard cordierites or melts. The standard errors on analytical populations for individual standard melt and cordierite grains are typically ±0·2 and ±0·08 wt % respectively for each analytical session, which yield calibration curves with 2
uncertainties of 0·15 wt % and 0·06 wt % for melt and cordierite H2O contents of 3 wt % and 1 wt % respectively. Following SIMS analysis charges were routinely examined under BEI to confirm the positions and textural contexts of each analytical spot (e.g. Fig. 2b–e) and hence establish, for example, whether the melt (glass) analyses had incorporated any fine crystals and cordierite analyses any melt or spinel.
Each experiment has been assessed to determine whether it was conducted at water-saturated or water-undersaturated conditions. The release of fluid on puncturing the capsule and the extensive presence of vesicles in the experimental glasses are interpreted as a sign that the capsule had been water saturated. As described below, the melt water contents in most of these charges matched closely the saturation water contents reported by Holtz & Johannes (1994) and Johannes & Holtz (1996) for melts in closely comparable chemical systems.
|
EXPERIMENTAL RESULTS |
|---|
Approach to equilibrium: time-study
A time-study was conducted to assess the convergence towards equilibrium in the experiments. Two separate experimental charges, one with evacuated BB3 cordierite and the other containing pre-saturated BB3 cordierite, were run for 10 days to provide an experimental reversal bracket at an intermediate melt H2O content of 4–5 wt % (experiments C4-1 and C4-C). These were complemented by four sandwich-style experiments (C12-1 to C12-4) in which evacuated cordierite and re-saturated cordierite layers were separated by a melt layer (Fig. 2b). The sandwich experiment charges each had H2O added to them individually by microsyringe and so did not have identical initial total H2O contents. These four experimental charges were then run for different time intervals (Table 3).
The results of the time-study are plotted in Fig. 3. To evaluate the experiments, we have first calculated the value of Dw [= H2O(melt)/H2O(cordierite)] using the analysed H2O contents of the melt and the initially evacuated cordierite at one end of the capsule. This is then compared with the Dw calculated using the analysed H2O contents of the melt and the initially saturated cordierite at the other end of the capsule. Assessing the experiments in this way largely removes spurious effects of variations in analysed phase H2O contents, which are due to variations in the amounts of starting materials (including H2O fluid) added to each experimental charge. The two calculated values of Dw converge to within error after only 24 h and continue to be comparable for up to 240 h run duration, the shortest run duration used for the principal equilibrium experiments of this study. The increase in the Dw calculated using the pre-evacuated cordierite with increasing run duration is opposite to that which would be expected for identical experimental charges with identical total H2O contents. This trend, and that of overall increase in the converged Dw obtained from 24 h to 240 h, results from the differing total H2O content of each experiment. As is demonstrated by the full experimental dataset, Dw at a selected P–T condition is generally observed to increase with increasing H2O contents in the phases (see below). The convergent time-study results are consistent with this, and it is coincidental that the time-study experiments with the lowest total H2O contents (i.e. bulk weight percent H2O in the experimental charge) happen to be those with the shorter run durations. We are confident that the value of Dw reached at 240 h is close to equilibrium, as six previous and comparable experiments (Carrington & Harley, 1996) with run durations from 280 to 340 h show an average Dw of 4·4 (± 0·6), indistinguishable from the results of this time-study.
|
Phase chemistry
Glasses (melts)
The melt gel composition (Table 1) has been selected to be close to the eutectic composition and in Fe–Mg equilibrium with BB3 cordierite (Table 2) at the central P–T conditions of the study, 900°C and 5·0 kbar (Carrington & Harley, 1996). The experimental glasses have been analysed to detect any significant deviations from eutectic compositions, which would result in reduced melt water contents at a given P–T condition (Table 2). Analyses have been obtained by analytical SEM using a liquid-nitrogen cold stage to minimize the alkali counting losses that are common in water- and alkali-rich glasses (SEM, Fig. 4), and with a Cameca CAMEBAX electron microprobe (EMP, Fig. 4) to cross-check the accuracy of SiO2, Al2O3, FeO and MgO as measured on the cold-stage scanning electron microscope. EMP analyses show variable to extensive alkali losses, leading to high normative quartz (Fig. 4), but project back into the cluster of analyses obtained by SEM. Hence the reported glass analyses, which show only minor variation in SiO2, Al2O3 and FeO centred around the starting composition, are reliable. The analysed glasses have consistently higher MgO and lower Na2O and K2O contents than the initial gel, but their compositional range overlaps that of eutectic compositions measured in independent experimental studies over a range of P–aH2O conditions (Johannes & Holtz, 1996, fig. 2.20). As the melts produced in our experiments appear to be close to eutectic compositions, any decrease in the maximum melt H2O content attainable at a given P–T condition arising from compositional variation is significantly less than the variations resulting from changes in pressure and aH2O.
|
In detail, normative quartz (Fig. 4) increases with increasing weight percent H2O to contents that are high compared with haplogranitic eutectic melts but consistent with independent experimental results for peraluminous melts character (Johannes & Holtz, 1996). Normative corundum in the experimental glasses increases from 2·2 to 6·2% in parallel with the increase in normative quartz. The higher levels of normative corundum are greater than that attained in haplogranite experiments (Johannes & Holtz, 1996), although comparable with those of natural leucogranites and rhyolites. The maximum melt H2O content at fluid saturation is not significantly dependent on normative corundum content, and for the total variation in normative corundum seen in our experimental study there is a variation of only 0·5 wt % in the maximum melt H2O content attainable at a specified P–T condition (Johannes & Holtz, 1996, fig. 4.8). The more quartz-normative of our experimental glasses are less orthoclase-normative than the previously determined eutectic compositions, but this deviation is unlikely to affect the maximum melt H2O contents greatly as the Or/Ab ratio has only a small effect on H2O solubility for the limited range of compositions (Or/Ab = 0·7–0·82) produced in this study (Holtz et al., 1992a, 1992b).
Cordierite
The cordierite crystals generally retain their original composition with respect to major oxides other than H2O. There are changes in the Fe/Mg ratio of the crystal rims and increases in the minor alkali content of the cordierites (Table 3).
The melt (XMg = 0·25) and cordierite (BB3: XMg = 0·64) compositions were originally chosen to be in Fe/Mg equilibrium at 900°C, 5·0 kbar and low aH2O (Carrington & Harley, 1995, 1996). At other conditions rims of higher, and occasionally lower, cordierite XMg are observed (Fig. 2d). The rims have a maximum thickness of 10 µm and their XMg increases with higher-P–T run conditions. In some runs grains of an evacuated cordierite with higher XMg (AMNH: XMg = 0·85) were used, to limit any reaction with melt and provide comparisons with lower-XMg cordierite. The maximum H2O content of cordierite attained at any specified P–T condition, in terms of moles per formula unit, has been demonstrated to be independent of its XMg (Boberski & Schreyer, 1990; Carrington & Harley 1996).
Cordierites show increased Na2O contents (to 0·32 wt %) that are essentially independent of both T and H2O content. Cordierite K2O contents increase with rising T (to 0·19 wt %) and decreasing H2O, and are in all cases higher than in the starting grains. The lack of correlation between T and Na2O in the cordierites in our experiments is inconsistent with the Na2O in cordierite geothermometer devised by Mirwald (1986), but the measured Na contents at 800–900°C and 5 kbar (Crd rims 0·04–0·07 Na cations p.f.u.) are comparable with Na contents reported at similar conditions by Knop et al. (1998). The relationships between cordierite alkali contents (both Na and K) and T–aH2O in our experiments have been described and analysed in detail elsewhere (Thompson et al., 2001) and will not be considered further here as they have no discernible bearing on the H2O data.
The potential for channel alkalis in cordierite to block the passage of water has been noted in previous diffusion studies (e.g. Johannes & Schreyer, 1981; Zimmermann, 1981). However, for two reasons we suggest that the increased alkali content of the cordierite in our experiments is more likely to prevent water loss on quenching than prevent water entry during the early stages of the run. First, the majority of the starting cordierite grains are initially anhydrous (as confirmed by SIMS analysis of evacuated grains) but end the experiments containing significant water. Second, the measured water contents of the phases in the water-saturated experiments are found to be very close to those found in previous studies conducted on cordierites of varied Fe/Mg ratio and synthesized both with and without alkalis.
In a small number of experiments, a few modal percent of new phases formed. Orthopyroxene was formed very rarely, as fine crystals in the melt. Biotite, feldspar and quartz were occasionally found in melt or on cordierite at the cordierite–melt interface in the lower-temperature runs (Fig. 2e). Spinel was found in the higher-P and -T experiments, usually formed on resorbed cordierite grain boundaries or along fractures in cordierite (Fig. 2b and f). At the highest-T, lowest-P H2O-rich run conditions (e.g. 1000°C, 3 kbar) cordierite showed extensive resorption and replacement by granular spinel, presumably as a result of incongruent dissolution in the H2O-rich melts via the reaction Crd = Spl + SiO2(L) (Fig. 2f). Cordierite–melt H2O contents measured from such charges have not been used in the data analysis and are not reported here.
Equilibrium H2O contents in cordierites and glasses (melts)
The H2O-content results of the experiments are presented in Table 4 and illustrated in three types of diagram to facilitate comparisons of experiments at single P–T conditions and to allow comparisons between polybaric and polythermal data. First, the averaged phase H2O wt % data are plotted simply as H2O wt % in melt against H2O wt % in coexisting cordierite for specified P–T conditions in Fig. 5. Second, the distribution coefficient Dw [= wt % H2O(melt)/wt % H2O(Crd)] calculated from such data is plotted against the H2O content of the product cordierite for given sets of P–T conditions in Fig. 6. Lastly, the normalized H2O content of melt is plotted against the normalized H2O of cordierite in Fig. 7. In the last type of diagram the measured H2O content of each phase is normalized against its maximum H2O content at fluid saturation at the specified P–T condition. This procedure effectively normalizes isobaric–isothermal data against unit water activity (aH2O) and hence allows comparison of all data with the effects of P and T on the maximum H2O contents attained at aH2O = 1 accounted for.
|
|
|
|
Melt–cordierite H2O content data at specified P–T conditions
H2O content data obtained at 5·0 kbar and 900°C (Fig. 5a) define a smooth and coherent convex trend. The melt H2O content varies from 10·0 to 2·1 wt %, which by comparison with the data of Holtz & Johannes (1994) spans the entire range from the maximum (aH2O = 1·0) to minimum possible H2O contents (aH2O << 1·0). Coexisting cordierite H2O contents vary from 1·70 to 
0·50 wt %. The curvature of this H2O distribution trend, previously described as a ‘dog-leg’ and ascribed to solubility mechanism effects in the melt (Carrington & Harley, 1996), is discussed further in a later section. The forward and reversal experiments are indistinguishable, consistent with a close approach to equilibrium.
The remainder of the data are presented in Fig. 5b and c. The 7·0 kbar, 900°C experiments (Fig. 5b) produced considerably higher maximum (saturation) melt and cordierite H2O contents than those at 5 kbar, consistent with published data for both phases (Holtz & Johannes, 1994; Mirwald et al., 1979). The 3·0 kbar, 900°C undersaturated experiments have slightly lower cordierite H2O contents for a given melt H2O content, when compared with 5·0 kbar data. Also included in Fig. 5b are data from experiments at 3·0 kbar and 800°C, which follow the same trend as that of the 5·0 kbar, 900°C data. All the reversal experiments shown in Fig. 5b are consistent with smooth trends defined by the forward experiments. The isobaric data (Fig. 5c) reproduce the range of melt and cordierite water contents measured at 5·0 kbar and 900°C. The 5·0 kbar, 800°C data show slightly higher cordierite H2O contents for a given melt H2O content, whereas three of the four 5·0 kbar, 1000°C data demonstrate lower cordierite H2O contents for a specified melt H2O. The one discrepant undersaturated experiment at this P–T condition (C16-2) is attributed to H2O loss from the quenched melt glass, probably during sample preparation for SIMS. The forward and reversal experiments run at 5·0 kbar, 800°C produce results within error of each other, although the cordierite water contents are systematically higher for the reversal experiments. A close, but not necessarily complete, approach to equilibrium is therefore inferred for these lower-temperature experiments.
Also depicted in Fig. 5 are reference distribution curves derived from modelling of the entire cordierite–melt dataset, as explained in a later section. These modelled distribution curves bracket all the experimental data but, it should be noted, are not curves fitted to individual datasets collected at specific P–T conditions. The curves, derived from the modelling described below, display a convexity that is consistent with the general curvature of the cordierite–melt H2O data obtained at each P–T condition.
Dw–cordierite H2O content relationships
The distribution of H2O between the experimental melt and cordierite pairs is illustrated in Fig. 6. The data obtained at 5·0 kbar and 900°C show that the distribution coefficient, Dw, is 5·9 (± 0·3) for H2O-saturated conditions and decreases smoothly to 4·2 (± 0·4), before possibly rising again to 5·0 (± 0·4) at the minimum cordierite H2O content attained. These data straddle and are consistent with a parabolic Dw–H2O(Crd) curve as depicted in Fig. 6a that has been produced from thermodynamic modelling of the entire dataset (continuous line) and calculated at 5·0 kbar and 900°C; the curve itself is not an explicit fit to the data presented in Fig. 5a.
The data from the experiments conducted at other P–T conditions are comparable in form with the results at 5·0 kbar and 900°C but demonstrate systematic P–T-dependent effects, in general being displaced consistently upwards (for higher T–lower P) or downwards (for lower T–higher P) with respect to the 5·0 kbar and 900°C data and its modelled trend. Isothermal data at 900°C (Fig. 6b) and 800°C (Fig. 6c) indicate that Dw increases with decreasing pressure for a given cordierite H2O content. The isobaric data at 5·0 kbar (Fig. 6d) and 3·0 kbar (Fig 6e) demonstrate that Dw increases with rising temperature for a given cordierite H2O content. The highly undersaturated experiment at 5·0 kbar and 1000°C (C5-4) that produced cordierite with a H2O content of only 0·36 wt % (Fig. 6d) appears to have an anomalously high Dw that is close to that otherwise attained only at saturation for this P–T condition. However, the thermodynamic modelling presented in the next section shows that at high temperatures where the minimum H2O contents of both the melt and cordierite are very low, high values of Dw are to be expected, as illustrated by the convex form of the modelled Dw–H2O function superimposed on the data for comparative purposes.
Only the data obtained at 5·0 kbar, 900°C and 3·0 kbar, 800°C span the whole range of melt H2O contents from the maximum to minimum values given by Holtz & Johannes (1994). Our data at the other P–T conditions represent partial trends from saturation (or near-saturation) to lower, but not the lowest possible, H2O contents in coexisting melt and cordierite. This applies in particular to the 7·0 kbar, 900°C dataset (Fig. 6b), which lies parallel to the steeply decreasing part of the Dw trend seen in the 5·0 kbar, 900°C data and the modelled distribution curve. As will be seen in the modelling section, the shapes of the Dw–H2O (Crd) trends and range of values of Dw for all data except one H2O-undersaturated saturated Crd–L data point at 5 kbar, 1000°C (Fig. 6d) and a minimum-H2O data point at 3·0 kbar, 800°C (Fig. 6c) are consistent with simple thermodynamic formulations for the incorporation of H2O in both melt and cordierite.
Normalized melt–cordierite H2O relationships
Further interpretation of the experimental results from Figs 5 and 6 is complicated by the variations in the maximum H2O contents of the phases that arise from the different P–T conditions used in the experiments. To account for the P–T-dependent differences in the maximum H2O contents, and therefore compare the H2O-distribution behaviour over the whole P–T range of the experimental study, the H2O contents of both melt and cordierite are expressed as a fraction of the maximum H2O content in the phase at that P–T condition. At 5·0 kbar, 900°C; 5·0 kbar, 800°C; and 3 kbar, 800°C, the maximum fluid contents in cordierite and melt are taken from the experiments reported in this study. Maximum-H2O melt compositions obtained at 5·0 kbar, 1000°C and 7 kbar, 900°C, however, are considered to be unreliable because of H2O loss from the quenched glasses, as noted above, and hence in these cases the maximum melt H2O content at fluid saturation is taken from Holtz & Johannes (1994) and the cordierite from our experiments. As fluid-saturated conditions were not attained at 3 kbar, 900°C, this P–T dataset is not presented in Fig. 7. The normalized parameters used for presentation of the data in the polybaric–polythermal diagrams of Fig. 7 are
 |
 |
As in the previous diagrams the data obtained at 5·0 kbar, 900°C define a smooth and curved distribution trend (Fig. 7a). Data at 5·0 kbar, 800 and 1000°C (Fig. 7b) essentially follow the same normalized trend apart from the discrepant experiment C16-2, and hence indicate that temperature does not affect the shape and position of the normalized distribution curve. Similarly, isothermal data obtained at 800°C, 3·0 and 5·0 kbar show no effect of pressure on the shape of the normalized distribution curve.
The data obtained at 3·0 kbar, 900°C have not been normalized as fluid saturation was not achieved in experiment C8-1. This is borne out by comparison of our maximum melt H2O contents with those of Holtz & Johannes (1994), which are in complete agreement (within error) at 5·0 kbar, 900°C; 5·0 kbar, 800°C; and 3·0 kbar, 800°C. In contrast, the melt H2O content measured at 3·0 kbar, 900°C is 5·94 wt % (C8-1) compared with Holtz & Johannes’ saturation value of 7·4 wt %. Moreover, the measured Dw for this experiment (5·08) is consistent with our other undersaturated Crd–L data obtained at 900°C (Fig. 6b) and 3 kbar (Fig. 6e).
Figure 7 demonstrates that within the range 800–1000°C, 3·0–7·0 kbar the distribution of water between cordierite and granitic melts conforms fairly closely to a single normalized function. This function is represented by the dashed reference curve calculated from the modelling described below and added to each diagram in Fig. 7. The principal determinant of the relative magnitude of the distribution of water between melt and cordierite (Dw) is the system’s degree of undersaturation (aH2O) at the defined P–T condition.
In summary, the data presented in Fig. 5 consistently display the features observed and reported in detail for the 5 kbar, 900°C dataset (Carrington & Harley, 1996). Figure 6 shows that the general form of the Dw–H2O(Crd) curve at all P–T conditions is one of increasing Dw for cordierite H2O contents that are <0·75 wt % and >1·0 wt %. Dw values are in the range 5–6 near saturation but are lower in undersaturated conditions, with minima in the range 3·5–4·8 for cordierite with 0·8–1·0 wt % H2O. Figure 7 demonstrates that the entire Crd–L dataset obtained under polybaric–polythermal conditions follows a consistent pattern that can be described by a single normalized H2O distribution curve. This is accounted for by the thermodynamic formulations of the incorporation of H2O into cordierite and melt, as elucidated in the sections below.
|
THERMODYNAMIC MODELLING OF THE H2O SYSTEM |
|---|
The cordierite–melt experimental data can be treated using appropriate thermodynamic models to define the relationships between water activity (aH2O), Dw and the H2O contents of cordierite and melt. The thermodynamic models used are described first and then applied to the data to model the relationship between Dw and the H2O content of cordierite or melt illustrated in Fig. 5. The present results are then integrated with previous experimental data at lower T to develop a revised saturation isohydron set for cordierite in equilibrium with pure H2O in the P–T range 2–9 kbar and 500–1000°C.
Cordierite–hydrous cordierite
The energetics of hydration of cordierite have been considered by several workers (e.g. Newton & Wood, 1979; Lonker, 1981; Martignole & Sisi, 1981; Kurepin, 1984; Bhattyacharya & Sen, 1985). Largely influenced by the hydration experiments of Mirwald et al. (1979), some early models suggested incorporation of up to 1·2 moles of H2O per formula unit (p.f.u.) of cordierite, achieved under higher-P, lower-T conditions. However, more recent models (e.g. Mukhopadhyay & Holdaway, 1994; Carey, 1995; Skippen & Gunter, 1996) have generally agreed on a one-site, 1 mole mixing model for H2O incorporation, consistent with both calorimetric measurements (Carey & Navrotsky, 1992) and ab initio energy calculations (Winkler et al., 1994). Hence, the analysis of the thermodynamics of hydrous cordierite presented here follows the approach adopted by recent workers (Carey, 1995; Skippen & Gunter, 1996), such that the following relation governs H2O incorporation into cordierite:
 |
The reaction above then has the following equilibrium expression:
 |
 |
 |
Vs, and the volume of the vapour phase at the P and T of interest:
 |
Vs = 0, a situation considered and modelled by Carey (1995) and Skippen & Gunter (1996). In this paper we will re-examine whether Vs = 0 in the light of the full dataset compiled from our own and previous data, but for analysis of isobaric data this term will be implicit in the retrieved Hr and so will not be considered in the initial modelling. Noting that the integral term in equation (5) is related to the fugacity of H2O at the P–T of interest, f H2O, equating with RT ln fH2O, hence
 |
Before any polybaric–polythermal analysis, it is useful to note that at fixed P and T (and therefore also fixed f H2O at saturation with pure H2O) the left-hand side of equation (6) is a constant. Rearranging,
 |
 |
; Skippen & Gunter, 1996). The proportionality constant at a given P–T condition is denoted hereafter as zw, where w is for ‘water’. As is apparent from the derivation above, zw includes enthalpy and entropy terms that are obtainable from regression of our data and those of Mirwald et al. (1979), Boberski & Schreyer (1990), Carey (1995), Skippen & Gunter (1996) and others. Finally,
 |
 |
 |
We will first examine how well the isobaric–isothermal experimental data for cordierite can be fitted to the formulation described by (9) and (10), the first instance in which this has been possible by direct measurement. The variation of zw with pressure and temperature is then evaluated to derive a refined expression for the saturation value of n (nsat) over P–T space and a general relation for n as a function of aH2O at varied P–T conditions.
Incorporation of H2O into granitic or peraluminous melts
H2O can be incorporated into melts through the breaking of Si–O linkages producing –(OH) and (H)O– pairs and as molecular water [see review by Holloway & Blank (1994)]. The latter solubility mechanism, which becomes important at high H2O contents (e.g. >5 wt % H2O; McMillan, 1994), may be responsible for H2O increasingly preferring the melt over cordierite as a residence site, as seen in the Dw data of Fig. 6.
There have been several attempts at modelling the solubility of H2O in silicate melts (Stolper, 1982; Burnham & Nekvasil, 1986; McMillan & Holloway, 1987; Moore et al., 1995, 1998). Usually these models involve empirical fits to solubility data at saturation and for unit aH2O for a range of P–T conditions, coupled with compositional terms (e.g. Burnham, 1994; Holloway & Blank, 1994; Moore et al., 1995) to account for the effect of melt network modifiers on H2O contents. Alternatively, models have been developed that incorporate speciation data (Stolper, 1982) and allow iterative calculation of total H2O contents at given P–T knowing the speciation between molecular or free H2O and the H2O incorporated as a network-breaker (Holloway & Blank, 1994).
The recent and comprehensive model of Papale (1997), which incorporates experimental data to 15 kbar, demonstrates that H2O solubility in granitic melts is not adequately represented by the Burnham model at high pressures. Although the P–T-dependent parameter kw defined from our high-P experiments is lower than would be predicted using the Burnham expressions for rhyolite glass (Burnham, 1994), and hence our saturation H2O contents are higher at pressures of 3–7 kbar, we have used the general formulations of Burnham in the absence of any better high-P model. The critical point of departure for the present analysis, therefore, is that in the granitic liquids at pressures between 3 and 7 kbar we assume that the activity of water can be approximated by the following expressions:
 |
 |
Modelled H2O distribution (Dw) and its variation with pressure, temperature and aH2O
In the case of Crd–melt equilibrium, for isobaric–isothermal experimental data obtained from saturated to highly undersaturated conditions aH2O defined from cordierite must equal aH2O defined from melt. Hence
 |
 |
Equations (13a) and (13b) can be combined with the relationships between n and weight percent H2O in cordierite or Xw and weight percent H2O in melt to determine the variation of Dw as a function of weight percent H2O in either phase for specified values of kw/zw. The theoretical form of Dw so derived is a combination of an inverse and linear function in H2O, with a calculated minimum region (± 0·2 units) occurring between 0·7 and 0·9 wt % H2O in cordierite (between 2·5 and 4·5 wt % H2O in the melt) and the magnitude of Dw then increasing gradually to saturation (Fig. 9). These two features are consistent with the experimental Dw data. The hydration models embodied in equations (13a) and (13b) also predict a rapid asymptotic increase in Dw with decreasing H2O below 0·2–0·4 wt % in cordierite (<1·2–1·7 wt % H2O in melt). These marked increases in Dw at low H2O contents are not readily observed in the experiments, partly because the crystallization of quartz and feldspar in near-solidus H2O-poor melts hinders the precise measurement of Dw at very low aH2O. However, the principal limitation is that these effects are predicted to occur at H2O contents generally less than the minimum possible in a granitic melt at the P–T conditions of most of the experiments: the rapid rise in Dw is largely restricted to an inaccessible region of extreme undersaturation (and low aH2O) where melting is not possible unless the temperatures are very high. This behaviour is expected to occur at temperatures near the dry granite liquidus (e.g. at 1050°C). Consistent with this prediction, an increase in Dw at low H2O contents is most evident in the 5 kbar, 1000°C data (Fig. 6d).
|
The analysis of isobaric–isothermal datasets can be generalized to all P–T conditions through normalization of the variables n/(1 – n) and Xw against their values at saturation at the studied P–T conditions [i.e. nsat/(1 – nsat) and Xw,sat]. This procedure makes the fewest assumptions about the values of kw and zw. The bulk of the experimental dataset (all except two points from 36) conforms to a straight-line relation between the melt and cordierite aH2O indicators and supports their applicability to the modelling of Dw relationships over a wide P–T range. Because the modelled Dw curves form a family with the position of any one curve characterized simply by kw/zw, they depend only on the relation between nsat/(1 – nsat) in cordierite and the functions in Xw,sat for the melt at a specified P–T condition and so can be derived from saturation isopleths (isohydrons) for melt and cordierite.
On the basis of existing saturation H2O isopleths for melts (Holtz & Johannes, 1994; Johannes & Holtz, 1996) and those modelled here for cordierite (see below), kw/zw varies between 3·0 and 7·0 across the P–T range 3–7 kbar and 700–1000°C, decreasing with rising T at constant P, and increasing with increasing P at fixed T. For kw/zw values between 3·0 and 7·0, modelled Dw lies in the range 2·5–4·5 for those highly undersaturated conditions where cordierite is in equilibrium with a melt that has the minimum possible H2O content for the specified P–T (Fig. 9). Maximum Dw values, which generally occur at H2O saturation, are in the range 4·5–6·5 for this degree of variation in kw/zw (Fig. 9). The Dw curves generated from the modelling can be used to estimate the H2O contents of melts coexisting with cordierites of known composition. For example, cordierites containing 1 wt % H2O could have equilibrated with melts containing 2·5–5 wt % H2O, depending on P and T, with the higher H2O content being in melts formed at high T and low P (i.e. low kw/zw).
|
CORDIERITE H2O SATURATION AND ISOHYDRONS TO 1000°C |
|---|
The experiments reported here provide six new, directly measured, data points on the saturation H2O contents of cordierite at 800–1000°C. Furthermore, these are supported by and consistent with several other data points obtained at the same P–T conditions. For example, the saturation H2O content of cordierite at 5 kbar and 900°C (1·69 ± 0·05 wt % for XMg = 0·67) is based on two saturation experiments but is also constrained to be in this range by the linear fit to all the other, H2O-undersaturated, data obtained at this P–T condition.
The new saturation data also access a higher T field than hitherto available for modelling the incorporation of H2O into cordierite, resulting in a significantly wider 1/T interval for fitting experimental hydration data. Accordingly, we have taken the hydration data from previous studies of Mg, Fe and Mg–Fe cordierite (Schreyer & Yoder, 1964; Mirwald et al., 1979; Boberski & Schreyer, 1990; Mukhopadhyay & Holdaway, 1994; Carey, 1995; Skippen & Gunter, 1996), added our own higher-T data, and fitted the entire dataset via a ln Keq vs 1/T plot, where ln Keq is evaluated from
 |
The ln Keq vs 1/T plot for all cordierite hydration data is presented in Fig. 10. This does not discriminate subjectively between data or classify data on the basis of measurement technique or run pressure. Although there is significant scatter in ln Keq at temperatures of 800 and 600°C, this is not consistently related to pressure or other intrinsic variables, and linear regression of all the data yields
 |
|
, saturation data obtained in this study at 800, 900 and 1000°C; 
, data from all other sources as outlined in the text. This plot presents all 155 data points, but the best-fit line (continuous line) is regressed through 98 data points that exclude the far outliers where duplicate data are available, several points that have large analytical uncertainties, and points obtained at extreme P conditions (
9 kbar).
Our preferred fit to the data, however, utilizes 98 points from the 155 available at 40 distinct P–T conditions. The data excluded from this fit are outliers obtained at P–T conditions where two or more other results are mutually compatible and cases in which the relative errors on H2O measurement are large (>20% relative). This fit, which includes all our data along with all those of Boberski & Schreyer (1990), Mukhopadhyay & Holdaway (1994) and Carey (1995), is as follows:
 |
Hr = -34·95 kJ/mol and entropy change Sr = -97·69 J/mol/K for the T range 500–1000°C.
To assess whether the assumption of Vs = 0 is reasonable we tested correlations of ln Keq against P for isothermal datasets obtained in previous experimental studies as well as those presented here. Whereas data at 800°C and 700°C show moderate correlations of ln Keq with P and yield similar d(ln Keq)/dP slopes of 0·88 J/kbar, any P dependence of data at 600°C, 750°C, 900°C and 1000°C is within the errors of measurement. In agreement with Carey (1995), we conclude that the extraction of a Vs term is not warranted, based on the existing hydration data and given the uncertainties in these data.
Recent heat capacity data for cordierite indicate that Cp of hydrous cordierite (H-Crd) is higher than that of anhydrous cordierite (A-Crd) by 8·3 J/mol/K (Carey & Navrotsky, 1992). Holland & Powell (1998) have accordingly fitted cordierite hydration data over the T range 500–800°C (i.e. excluding the data presented here) with the Cp of reaction included, to derive estimates for H298 and S298. Their retrieved values of -41·8 kJ/mol and -109·0 J/K/mol respectively are within error of our estimates of H298 (-41·97 kJ/mol) and S298 (-109·11 J/K/mol) obtained when the Cp from Carey & Navrotsky (1992) is included in the regression of the 500–1000°C data. In this case the best-fit result is
 |
Cp term.
In the light of the errors inherent in the experimental data we prefer the simpler fit given in equation (16) for direct calculation of saturation isohydrons of H2O in cordierite over the T range 500–1000°C. The isohydrons thus produced, expressed in terms of molecules of H2O per 18-oxygen anhydrous cordierite formula unit (i.e. nsat), are presented in Fig. 11 and compared with those presented by Skippen & Gunter (1996) and calculated using Holland & Powell (1998) in Fig. 12. The geohydrometer of Skippen & Gunter (1996) yields higher H2O contents in cordierite than predicted from our isohydrons at all P–T conditions, with the greatest differences occurring at lower pressures and for lower temperatures (2–4 kbar, 600–700°C; Fig. 12a). The two isohydron sets are within error (±0·015 p.f.u.) at high temperatures. Isohydrons calculated from Holland & Powell (1998) are lower than the best-fit isohydrons produced here (Fig. 12b), but only by 0·005–0·01 p.f.u. over most of the P–T range considered. Such differences equate to differences in H2O of only ±0·02–0·04 wt % and are within the errors of measurement of H2O contents in cordierite using SIMS and other methods (usually > ±0·05 wt %). In summary, the isohydron set presented in Fig. 11 defines the P–T dependence of the hydration of cordierite in the presence of pure H2O and is consistent with the best available experimental data as modelled in previous studies as well as with our new data at 800–1000°C. Equation (16), or the isohydrons of Fig. 11, can be used in concert with measured contents of H2O in natural cordierites and independent estimates of the P–T conditions of equilibration to calculate aH2O via equations (9), (10) or (11). Examples of such applications are presented below.
|
|
|
APPLICATIONS TO MELTING AND HIGH-TEMPERATURE FLUID PROCESSES |
|---|
Dw for H2O-saturated conditions and the wet granite solidus
Superimposition of the saturation isohydrons for cordierite on the isopleths for maximum H2O contents in peraluminous granitic melts allows calculation of kw/zw over the P–T field of melt stability and also direct determination of the Dw at saturation [Dw(sat)]. Contours of Dw(sat) are presented along with cordierite and melt saturation isohydrons and isopleths in Fig. 13a. Over the P–T field 1–8 kbar and 650–900°C, Dw(sat) varies in the range 4·1–6·2. The slightly steeper positive slopes of the cordierite isohydrons compared with melt saturation H2O isopleths cause Dw(sat) to increase with rising temperature and pressure over most of the P–T range of mid- and deep-crustal melting, but the Dw(sat) contours switch from these negative dP/dT slopes to positive dP/dT at pressures <2 kbar. Dw(sat) only varies between 4·2 and 5·0 along the water-saturated solidus in the NKFMASH or simplified granite system, for pressures in the range 1–6 kbar. The experimental data indicate that an average Dw of 4·3 ± 0·2 is appropriate for cordierite crystallized in the majority of wet- or near-saturated granites that are emplaced at shallow crustal pressures and exhibit andalusite–sillimanite type contact metamorphism.
|
Dw for H2O-undersaturated conditions in the realm of dehydration-melting
In the KFMASH system at temperatures higher than 800°C the majority of reactions involving cordierite will be dehydration-melting reactions (e.g. Carrington & Harley, 1995
), and in such cases the cordierite will be constrained by the highly H2O-undersaturated melts produced through dehydration melting to have H2O-undersaturated compositions (Stevens et al., 1995; Carrington & Harley, 1996). In such circumstances, cordierite H2O contents will be buffered through the exchange equilibrium
 |
; Johannes & Holtz, 1996). Hence, the dehydration-melting KFMASH grids of Carrington & Harley (1995), contoured for XMg isopleths by Harley (1998a), can also be contoured for the H2O content of cordierite in melt-bearing assemblages provided that the melt minimum-H2O contours are independently estimated.
In Fig. 13b are plotted isopleths of the minimum H2O contents of granitic melts for the P–T conditions of dehydration-melting in KFMASH. These melt (L) H2O isopleths are modified from Holtz & Johannes (1994) and Johannes & Holtz (1996) for peraluminous melts in KFMASH and are consistent with minimum H2O content data reported by Carrington & Harley (1996) and presented in this work. The diagram is also contoured with isohydrons (in wt %) of H2O in cordierite coexisting with such melts in the assemblages cordierite + garnet + sillimanite + quartz + K-feldspar + melt (higher-T) and cordierite + biotite + sillimanite + quartz + K-feldspar + melt (lower-T). These isohydrons have been obtained by combining the melt minimum-H2O isopleths with the P–T-dependent Dw functions for melt–cordierite partitioning developed here and correcting the weight percent H2O in cordierite for the XMg of the cordierite present in the assemblage at the relevant P–T condition. The isohydrons can also be positioned by calculating aH2O from the melt isopleths and then solving for n(Crd) at that aH2O and the given P–T using expressions (9), (11) and (16) given above.
The cordierite H2O-isopleths in this situation (i.e. in univariant and divariant dehydration-melting) have much steeper positive dP/dT slopes than isohydrons in the saturated cases (34–75 bar/°C compared with 2·5–10 bar/°C) because they are governed by the relation (18) and because the melt minimum-H2O isopleths are strongly temperature, rather than pressure, sensitive. In dehydration-melting the cordierite isohydrons are slightly shallower than the melt isopleths (dP/dT = 84 bar/°C for melts; Fig. 13b), whereas in the case of H2O-saturated melting (Fig. 13a) the isohydrons are steeper than the saturation H2O isopleths in the melt. This reflects a marked difference in the P–T behaviour of Dw when melts with the lowest H2O contents attainable are compared with the saturated case.
As illustrated in Fig. 13c, Dw values relevant to dehydration melting [hereafter denoted Dw(dm)] in the T range 750–900°C vary between 2·6 and 5·0, with the lowest values calculated at high-P, lower-T conditions where Dw(sat) is high (e.g. 8 kbar, 750°C: Dw(sat) = 6·0 compared with Dw(dm) = 2·6). Dw(dm) contours have shallow to steep positive dP/dT slopes and always increase with rising temperature. As a consequence, very high Dw(dm) values of 6–8 are attained only at high-T, low-P conditions where the H2O contents in each phase are the range in which the Dw curves display asymptotic behaviour (Fig. 9).
Of particular interest are the variations in cordierite H2O contents, aH2O and Dw(dm) calculated along the dehydration-melting or vapour-absent melting reaction
 |
), aH2O increases as pressure decreases along the reaction, from 0·16 at 9 kbar to >0·31 at 4 kbar when calculated using the modelled cordierite H2O contents coupled with equations (9), (10) and (16). The biotite dehydration-melting reaction emanating to higher pressures than the invariant point shows a calculated increase in aH2O with pressure (from 0·16 to 0·20), consistent with Schreinemakers’ rules, as its metastable extension, rather than stable portion, intersects the wet granite solidus at low pressure.
P–T stability of cordierite with fixed H2O content
Figure 13d uses the data from Fig. 13a and b and the models formulated in equations (9) and (16) to illustrate the stability region for a cordierite with a specific weight percent H2O content. In this example the number of molecules of H2O per cordierite formula unit is 0·60; for an XMg of 0·65 this equates to 1·75 wt % H2O in the cordierite. This diagram is contoured for aH2O, ranging from the minimum possible at high P–T conditions where the cordierite is in equilibrium with a H2O-undersaturated dehydration-melt (from Fig. 13b) to unit aH2O, where the cordierite (and melt) would be H2O saturated and in equilibrium with pure water (Fig. 13a). Hence, the shaded field represents the P–T range in which a cordierite with 1·75 wt % H2O is stable and can, in principle, in the absence of any fluid-related perturbations and exchange or mineral reactions retain this H2O content. In the absence of a fluid or melt phase with which to exchange or equilibrate, and provided porosity is trivial, there is no necessity for the cordierite, once formed, to either lose or gain H2O as conditions change within this P–T field. For example, a cordierite with 1·75 wt % H2O formed at 7 kbar and 900°C could be subjected to near-isothermal decompression through 2 kbar under fluid- and melt-absent conditions before becoming ‘oversaturated’ and likely to lose H2O from its channels.
As is emphasized from inspection of Fig. 13d, there is no unique line in P–T space that defines the stability of a cordierite with a specific H2O content. Even in the absence of CO2, at high temperatures where melt is present cordierite with a specific H2O content can be formed over a rather large P–T field depending on the imposed aH2O. In the example shown (Fig. 13d) there is a 5 kbar difference, at 850°C, between the pressure at which the cordierite with 1·75 wt % H2O would form coexisting with aqueous fluid and the pressure at which it would form in equilibrium with a melt containing 3·5 wt % H2O.
Migmatites and metamorphism: aH2O estimates and melt H2O contents
The cordierite hydration calibration developed here, and the Dw relationships defined from our experiments and modelling, can, in principle, be used to calculate the aH2O of melting and metamorphism in high-grade metamorphic terrains characterized by cordierite migmatites and applied to infer the H2O contents of the melts formerly present during the metamorphism. Conversely, where the P–T–aH2O conditions of melting and metamorphism are well known from independent equilibria, cordierite H2O contents predicted from our modelling can be compared with measured contents to evaluate the role of H2O addition into or leakage from cordierite subsequent to melting and metamorphism. Table 5 presents a summary of cordierite data from selected migmatite terrains and cordierite-bearing granites and pegmatites. The cordierite volatile content data in most of the cited cases have been obtained using SIMS analysis with negative secondary ions, as described in a previous section and by Carrington & Harley (1996).
|
Cordierite occurs as a euhedral to subhedral phase in many low- and medium-pressure migmatites found in both regional and contact metamorphism (e.g. Waters & Whales, 1984
; Stevens & Clemens, 1993; Carrington & Harley, 1995). The common occurrence of cordierite as single grains and grain clusters in leucosome veins and patches, as inclusion-free grains along mesosome–leucosome contacts (e.g. Harley, 1994; Fitzsimons, 1996), and as inclusion-free rinds overgrowing sillimanite- or biotite-bearing cordierite cores (e.g. Buick et al., 1998) is consistent with its generation through dehydration-melting reactions of the form
 |
The H2O contents of cordierites from several examples of such migmatite occurrences, some of which will be described in more detail elsewhere, are presented in Table 5. The data are coupled with independent P–T estimates to allow calculation of apparent aH2O, Dw, and hence the nominal H2O content (in wt %) of coexisting granitic melt, using equations (9), (10) and (16) as given above. Cordierites from several lower-P migmatitic granulite areas (<5 kbar, Table 5) yield aH2O estimates of between 0·2 and 0·6 and melt H2O contents between 2·5 and 4 wt % at temperatures of 675–800°C. The higher aH2O conditions tend to be calculated in cases where extensive segregated leucosomes and granitic sheets occur (e.g. Mt Stafford, Argentina, Kimberly region of West Australia), comparable with aH2O estimates from cordierites in many leucogranites and well-segregated pegmatites (Table 5). For example, cordierite leucogranite from Beatenally quarry, in the Proterozoic Northampton Block of Western Australia, contains 1·3 wt % H2O and implies 4·6 ± 0·4% H2O in the coexisting granite liquid at an aH2O of 0·44 ± 0·04, whereas a restitic cordierite–spinel–sillimanite gneiss xenolith in granite from the same locality suggests a much lower aH2O (0·12 ± 0·03) and potential melt H2O content (1·9 ± 0·4 wt %) if it equilibrated at the same P–T conditions. In this example, and in other cases where cordierite remains in contact with segregated melt, it is considered likely that the higher H2O contents and aH2O attained in the granites and pegmatites reflect the addition or uptake of H2O into cordierite as the melt increases in its H2O content during crystallization following segregation and ponding.
The well-developed migmatites and leucogranites described by Smith (1996) from the Cretaceous Fosdick Complex of Antarctica contain cordierites whose H2O contents indicate aH2O (0·36 ± 0·05) and melt H2O contents (3·9 ± 0·5 wt %) that are consistent with biotite dehydration-melting at the inferred temperature conditions of the metamorphic event and hence argue against significant H2O leakage or addition into the cordierites subsequent to the melting event. The same conclusions can be drawn from various other examples distinguished in Table 5. For example, at Mount Stafford in the Arunta Complex (Vernon et al., 1990; Greenfield et al., 1998) cordierite in granitic leucosomes contains 1 wt % H2O (Table 5). Using our hydration model, at 2·5 kbar and 650°C we obtain aH2O of 0·4 from these cordierites, and calculate 4·2 wt % H2O in the coexisting melts, reasonably consistent with phase equilibria constraints (Greenfield et al., 1998).
Cordierite is often considered to be susceptible to post-formation leakage from the relatively ‘open’ channels (e.g. Aines & Rossman, 1984; Mirwald et al., 1986) and it is possible that in many cases the peak volatile contents are not retained. In accord with this, some of the migmatite examples in Table 5 may be interpreted to have experienced H2O loss as their preserved H2O contents yield ‘melt’ H2O contents that are too low to be consistent with dehydration-melting at the independently inferred P–T conditions. Sample 9017, a cordierite–quartz–feldspar patch leucosome from the Reynolds Range, Arunta Complex, central Australia (Buick et al., 1998), yields an aH2O of only 0·16 and melt H2O contents (2·2 wt %) that are too low for the specified peak temperatures of 775°C. Likewise, several cordierites from garnet-bearing migmatites from Prydz Bay, East Antarctica, imply apparent melt H2O contents of only 1·6–2·0 wt %. These are substantially less than the 3–3·5 wt % minima required for granitic melts at 5·5–6·5 kbar and 825–850°C, the inferred conditions of melting in this case (Fitzsimons, 1996). Cordierite in a leucosome from Madagascar, supplied by C. Nicollet, displays the same type of H2O deficiency in comparison with melt H2O requirements and again indicates that volatile loss from cordierite in leucosomes has to be considered when evaluating the measured H2O data. In all these cited examples the alternative explanation of the low phase H2O contents and aH2O would be that the peak temperature estimates used for the rock suites are lower than those at which cordierite and melt equilibrated at or near the metamorphic peak, a possibility that requires further evaluation using independent thermometry and assemblage constraints in each case.
Two orthopyroxene-bearing migmatitic samples provide good evidence for low aH2O conditions related to high-T melting. Sample 9052, a cordierite–orthopyroxene segregation in a garnet-bearing semipelitic gneiss from the Strangway Range of the Arunta Complex, suggests melting at aH2O of 0·12 to yield a melt with 2 wt % H2O, at 900°C. Ultrahigh-temperature melting and migmatization at Mather Peninsula in the Rauer Islands (Harley, 1998b) produced orthopyroxene + sillimanite + garnet and locally cordierite at P–T conditions of 8–9 kbar and 920–950°C. The aH2O (0·1) and melt H2O content (2 wt %) estimated from cordierite H2O data in this example are consistent with values expected from independent experimental data at these P–T conditions (Carrington & Watt, 1995).
In many of the cases listed in Table 5 and considered above CO2 is an additional important volatile in cordierite, ranging from >0·7 wt % (e.g. Prydz Bay, Reynolds Range) to more typical contents in the range 0·1–0·4 wt %. It is beyond the scope of this contribution to analyse and review the impact of CO2 on the Dw relationships developed from the H2O-only study. However, on the basis of experiments in the H2O–CO2 system (Thompson et al., 2001) the effect of CO2 on Dw at a given cordierite H2O content can be demonstrated to be negligible.
|
CONCLUSIONS |
|---|
Experimental study of the distribution of H2O between cordierite and coexisting granitic melt demonstrates that this distribution, Dw, is a predictable function of P, T and aH2O. Dw generally lies in the range 2·5–6·0 and at any selected P–T condition will vary depending on cordierite H2O content, with a minimum region corresponding to low to moderate aH2O. The H2O-undersaturated experiments prove that cordierite can have a range of equilibrium H2O contents at any specified P–T condition in the regime of melting, up to maxima defined by the H2O-saturated, fluid-present, system.
The isothermal–isobaric H2O content data for melt and cordierite are mutually consistent with simple solution models and, when coupled with the measured H2O data for cordierite in the saturated system, allow refinement of estimates for the thermodynamic parameters for cordierite hydration to 1000°C. The isohydron set so derived is similar to those recently produced by Skippen & Gunter (1996) over a smaller temperature range and within error of that which would be produced using the internally consistent dataset of Holland & Powell (1998).
The Dw–P–T–aH2O relations derived in this study allow calculation of the equilibrium H2O contents of cordierite coexisting with melts with given H2O at specified P and T, and vice versa, estimation of aH2O and cordierite–melt H2O contents along selected melting equilibria, calculation of H2O isohydrons for cordierite produced with melts formed through di- and multivariant dehydration-melting processes, and independent estimation of aH2O in cordierite-bearing migmatites, granites and granulite terrains. Such calculations provide a framework for examining the effect of H2O in cordierite upon melt production in high-grade metamorphism. They also demonstrate that whereas some cordierite migmatites formed under aH2O conditions consistent with biotite dehydration-melting, the H2O contents of other high-grade cordierites may reflect post-melting processes including H2O uptake during melt crystallization and H2O leakage following melt removal.
|
ACKNOWLEDGEMENTS |
|---|
Our thanks go to the following people: John Craven and Richard Hinton for assistance with SIMS analysis; Simon Burgess and Peter Hill for assistance with EMP analysis; Tim Hopkins and David Plant for assistance with SEM analysis; Volker Schenk, Dave Mattey, and Tony Fallick for analysis of cordierite standards; Francois Holtz for the provision of SIMS melt standards; Dave Mattey and Ian Fitzsimons for cordierite evacuation; and Pauline Thompson for discussions. Our thanks are due also to Dave Pattison, Gary Stevens and an anonymous reviewer for perceptive and detailed reviews that, it is hoped, have improved the clarity of this paper. This research was funded by the UK Natural Environment Research Council through grant GR3/09099 to S.L.H.
|
FOOTNOTES |
|---|
*Corresponding author. Telephone: 0044-131-6508547. Fax: 0044-131-6683184. E-mail: sharley{at}glg.ed.ac.uk
|
REFERENCES |
|---|
Aines, R. D. & Rossman, G. R. (1984). The high-temperature behavior of water and carbon dioxide in cordierite and beryl. American Mineralogist 69, 319–327.[Abstract]
Armbruster, Th. & Bloss, F. D. (1980). Channel CO2 in cordierites. Nature 286, 140–141.
Armbruster, Th. & Bloss, F. D. (1982). Orientation and effects of channel H2O and CO2 in cordierite. American Mineralogist 67, 284–291.[Abstract]
Armbruster, Th., Schreyer, W. & Hoefs, J. (1982). Very high CO2 cordierite from Norwegian Lapland: mineralogy, petrology, and carbon isotopes. Contributions to Mineralogy and Petrology 81, 262–267.[Web of Science]
Barbey, P., Macaudiere, J. & Nzenti, J. P. (1990). High-pressure dehydration melting of metapelites: evidence from the migmatites of Yaoundé (Cameroon). Journal of Petrology 31, 401–427.
Bhattacharya, A. & Sen, S. K. (1985). Energetics of hydration of cordierite and water barometry in cordierite-granulites. Contributions to Mineralogy and Petrology 89, 370–378.
Bhattacharya, A. & Sen, S. K. (1986). Granulite metamorphism, fluid buffering, and dehydration melting in the Madras charnockites and metapelites. Journal of Petrology 27, 1119–1141.
Boberski, C. & Schreyer, W. (1990). Synthesis and water contents of Fe2+-cordierites. European Journal of Mineralogy 2, 565–584.[Web of Science]
Buick, I. S., Cartwright, I. & Harley, S. L. (1998). The retrograde P–T–t path for low-pressure granulites from the Reynolds Range, central Australia: petrological constraints and implications for low-P/high-T metamorphism. Journal of Metamorphic Geology 16, 511–529.
Burnham, C. W. (1994). Development of the Burnham model for prediction of H2O solubility in magmas. In: Carroll, M. R. & Holloway, J. R. (eds) Volatiles in Magmas. Mineralogical Society of America, Reviews in Mineralogy 30, 123–129.
Burnham, C. W. & Nekvasil, H. (1986). Equilibrium properties of granitic pegmatite magmas. American Mineralogist 71, 239–263.[Abstract]
Carey, J. W. (1995). A thermodynamic formulation for hydrous cordierite. Contributions to Mineralogy and Petrology 119, 155–165.
Carey, J. W. & Navrotsky, A. (1992). The molar enthalpy of dehydration of cordierite. American Mineralogist 77, 930–936.[Abstract]
Carrington, D. P. & Harley, S. L. (1995). Partial melting and phase relations in high-grade metapelites: an experimental petrogenetic grid in the KFMASH system. Contributions to Mineralogy and Petrology 120, 270–291.
Carrington, D. P. & Harley, S. L. (1996). Cordierite as a monitor of fluid and melt water contents in the lower crust: an experimental calibration. Geology 24, 647–650.
Carrington, D. P. & Watt, G. R. (1995). A geochemical and experimental study of the role of K-feldspar during water-undersaturated melting of metapelites. Chemical Geology 122, 59–76.
Clemens, J. D. & Vielzeuf, D. (1987). Constraints on melting and magma production in the crust. Earth and Planetary Science Letters 86, 287–306.[Web of Science]
Fitzsimons, I. C. W. (1994). Cordierite migmatites from East Antarctica: geochemical constraints on volatile distribution during crustal anatexis. Mineralogical Magazine 58a, 274–275.
Fitzsimons, I. C. W. (1996). Metapelitic migmatites from Brattstrand Bluffs, east Antarctica—metamorphism, melting and exhumation of the mid-crust. Journal of Petrology 37, 395–414.
Goldman, D. S., Rossman, G. R. & Dollase, W. A. (1977). Channel constituents in cordierite. American Mineralogist 62, 1144–1157.[Abstract]
Greenfield, J. E., Clarke, G. L. & White, R. W. (1998). A sequence of partial melting reactions at Mt Stafford, central Australia. Journal of Metamorphic Geology 16, 363–378.[Web of Science]
Harley, S. L. (1989). The origins of granulites: a metamorphic perspective. Geological Magazine 126, 215–247.[Abstract]
Harley, S. L. (1994). Cordierite as a sensor of fluid and melt distribution in crustal metamorphism. Mineralogical Magazine 58a, 374–375.
Harley, S. L. (1998a). On the occurrence and characterization of ultrahigh-temperature (UHT) crustal metamorphism. In: Treloar, P. J. & O’Brien, P. (eds) What Controls Metamorphism and Metamorphic Reactions? Geological Society, London, Special Publications 138, 75–101.
Harley, S. L. (1998b). Ultrahigh temperature granulite metamorphism (1050°C, 12 kbar) and decompression in garnet (Mg70)–orthopyroxene–sillimanite gneisses from the Rauer Group, East Antarctica. Journal of Metamorphic Geology 16, 541–562.
Holland, T. J. B. & Powell, R. (1998). An internally consistent thermodynamic data set for phases of petrological interest. Journal of Metamorphic Geology 16, 309–343.[Web of Science]
Holloway, J. R. & Blank, J. G. (1994). Application of experimental results to C–O–H species in natural melts. In: Carroll, M. R. & Holloway, J. R. (eds) Volatiles in Magmas. Mineralogical Society of America, Reviews in Mineralogy 30, 187–230.
Holtz, F. & Barbey, P. (1991). Genesis of peraluminous granites II. Mineralogy and chemistry of the Tourem Complex (North Portugal). Sequential melting vs restite unmixing. Journal of Petrology 32, 959–978.
Holtz, F. & Johannes, W. (1994). Maximum and minimum water contents of granitic melts: implications for chemical and physical properties of ascending magmas. Lithos 32, 149–159.
Holtz, F., Behrens, H., Dingwell, D. B. & Taylor, R. B. (1992a). Water solubility in aluminosilicate melts of haplogranite composition at 2 kbar. Chemical Geology 96, 289–302.
Holtz, F., Johannes, W. & Pichavant, M. (1992b). Peraluminous granites: the effect of alumina on melt composition and coexisting mineral. Philosophical Transactions of the Royal Society of Edinburgh 83, 409–416.
Jochum, C., Mirwald, P. W., Maresch, W. V. & Schreyer, W. (1983). The kinetics of H2O exchange between cordierite and fluid during retrogression. Fortschritte der Mineralogie 61, 103–105.
Johannes, W. & Holtz, F. (1990). Formation and composition of H2O-undersaturated granitic melts. In: Ashworth, J. R. & Brown, M. (eds) High Temperature Metamorphism and Crustal Anatexis. London: Unwin Hyman, pp. 87–104.
Johannes, W. & Holtz, F. (1996). Petrogenesis and Experimental Petrology of Granitic Rocks. Berlin: Springer.
Johannes, W. & Schreyer, W. (1981). Experimental introduction of CO2 and H2O into Mg-cordierite. American Journal of Science 281, 299–317.[Web of Science]
Knop, E., Scheilk, M. & Mirwald, P. W. (1998). Incorporation of sodium into magnesium cordierite below and above the solidus. Abstract Supplement 1, Terra Nova 10, 30–31.
Kurepin, V. A. (1984). H2O and CO2 contents of cordierite as an indicator of thermodynamical conditions of formation. Geokhimiya 8, 1125–1134.
Le Breton, N. (1989). Infrared investigations of CO2-bearing cordierites: some implications for the study of metapelitic granulites. Contributions to Mineralogy and Petrology 103, 387–396.
Le Breton, N. & Thompson, A. B. (1988). Fluid-absent (dehydration) melting of biotite in metapelites in the early stages of crustal anatexis. Contributions to Mineralogy and Petrology 99, 226–237.
Lepezin, G. G. & Melenevsky, V. N. (1977). On the problem of water diffusion in the cordierites. Lithos 10, 49–57.[Web of Science]
Lonker, S. W. (1981). The P–T–X relations of the cordierite–garnet–sillimanite–quartz equilibrium. American Journal of Science 281, 1056–1090.[Web of Science]
Martignole, J. & Sisi, J.-C. (1981). Cordierite–garnet–H2O equilibrium: a geological thermometer, barometer and water fugacity indicator. Contributions to Mineralogy and Petrology 77, 38–46.
McMillan, P. F. (1994). Water solubility and speciation models. In: Carroll, M. R. & Holloway, J. R. (eds) Volatiles in Magmas. Mineralogical Society of America, Reviews in Mineralogy 30, 131–156.
McMillan, P. F. & Holloway, J. R. (1987). Water solubility in aluminosilicate melts. Contributions to Mineralogy and Petrology 97, 320–332.
Mirwald, P. W. (1986). Ist Cordierite ein Geothermometer? Fortschritte der Mineralogie 64, 119.
Mirwald, P. W. & Schreyer, W. (1977). Die stabile und metastabile Abbaureaktion von Mg-cordierit in Talk, Disthen und Quartz und ihre Abhangigkeit vom Gleichgewichts-wassergehalt des Cordierits. Fortschritte der Mineralogie 55, 95–97.
Mirwald, P. W., Maresch, W. V. & Schreyer, W. (1979). Der Wassergehalt von Mg-cordierit zwischen 500° und 800°C sowie 0·5 und 11 kbar. Fortschritte der Mineralogie 57, 101–102.
Mirwald, P. W., Jochum, C. & Maresch, W. V. (1986). Rate studies on hydration and dehydration of synthetic Mg-cordierite. Materials Science Forum 7, 113–122.
Moore, G., Vennemann, T. & Carmichael, I. S. E. (1995). Solubility of water in magmas to 2 kbar. Geology 23, 1099–1102.
Moore, G., Vennemann, T. & Carmichael, I. S. E. (1998). An empirical model for the solubility of H2O in magmas to 3 kilobars. American Mineralogist 83, 36–42.[Abstract]
Mukhopadhyay, B. & Holdaway, M. J. (1994). Cordierite–garnet–sillimanite–quartz equilibrium I. New experimental calibration in the system FeO–Al2O3–SiO2–H2O and certain P–T–XH2O relations. Contributions to Mineralogy and Petrology 116, 462–472.
Newton, R. C. & Wood, B. J. (1979). Thermodynamics of water in cordierite and some petrologic consequences of cordierite as a hydrous phase. Contributions to Mineralogy and Petrology 68, 391–405.
Newton, R. C., Smith, J. V. & Windley, B. F. (1980). Carbonic metamorphism, granulites, and crustal growth. Nature 288, 45–49.
Papale, P. (1997). Thermodynamic modelling of the solubility of H2O and CO2 in silicate liquids. Contributions to Mineralogy and Petrology 126, 237–251.
Patiño Douce, A. E. & Johnston, A. D. (1991). Phase equilibria and melt productivity in the pelitic system: implications for the origin of peraluminous granitoids and aluminous granulites. Contributions to Mineralogy and Petrology 107, 202–218.[Web of Science]
Pichavant, M., Holtz, F. & McMillan, P. F. (1992). Phase relations and compositional dependence of H2O solubility in quartz–feldspar melts. Chemical Geology 96, 303–319.
Puziewicz, J. & Johannes, W. (1990). Experimental study of a biotite-bearing granitic system under water-saturated and water-undersaturated conditions. Contributions to Mineralogy and Petrology 104, 397–406.
Santosh, M., Harris, N. B. W., Jackson, D. H. & Mattey, D. P. (1990). Dehydration and incipient charnockite formation: a phase equilibria and fluid inclusion study from south India. Journal of Geology 98, 915–926.
Schreyer, W. (1985). Experimental studies on cation substitutions and fluid incorporation in cordierite. Bulletin de Minéralogie 108, 273–291.
Schreyer, W. & Yoder, H. S. (1964). The system Mg-cordierite–H2O and related rocks. Neues Jahrbuch für Mineralogie, Abhandlungen 101, 271–342.
Shmulovich, K. I. & Graham, C. M. (1996). Melting of albite and dehydration of brucite in H2O–NaCl fluids to 9 kbars and 700–900°C: implications for partial melting and water activities during high pressure metamorphism. Contributions to Mineralogy and Petrology 124, 370–382.
Skippen, G. B. & Gunter, A. E. (1996). The thermodynamic properties of H2O in magnesian and iron cordierite. Contributions to Mineralogy and Petrology 124, 82–89.
Smith, C. H. (1996). H2O–CO2 contents of cordierite in migmatites of the Fosdick Mountains, Marie Byrd Land. Terra Antarctica 3, 11–22.
Stevens, G. & Clemens, J. D. (1993). Fluid-absent melting and the roles of fluids in the lithosphere: a slanted summary? Chemical Geology 108, 1–17.[Web of Science]
Stevens, G., Clemens, J. D. & Droop, G. T. R. (1995). Hydrous cordierite in granulites and crustal magma production. Geology 23, 925–928.
Stolper, E. M. (1982). The speciation of water in silicate melts. Geochimica et Cosmochimica Acta 46, 2609–2620.
Thompson, P., Harley, S. L. & Carrington, D. P. (2001) Sodium and potassium in cordierite — a potential thermometer for melts? European Journal of Mineralogy, in press.
Vernon, R. H. & Collins, W. J. (1988). Igneous microstructures in migmatites. Geology 16, 1126–1129.
Vernon, R. H., Clarke, G. L. & Collins, W. J. (1990). Local, mid-crustal granulite-facies metamorphism and melting: an example in the Mount Stafford area, Central Australia. In: Ashworth, J. R. & Brown, M. (eds) High-temperature Metamorphism and Crustal Anatexis. London: Unwin Hyman, pp. 272–319.
Vielzeuf, D. & Clemens, J. D. (1992). The fluid-absent melting of phlogopite + quartz: experiments and models. American Mineralogist 77, 1206–1222.[Abstract]
Vielzeuf, D. & Holloway, J. R. (1988). Experimental determination of the fluid-absent melting relations in the pelitic system. Consequences for crustal differentiation. Contributions to Mineralogy and Petrology 98, 257–276.[Web of Science]
Vry, J. K., Brown, P. E. & Valley, J. W. (1990). Cordierite volatile content and the role of CO2 in high-grade metamorphism. American Mineralogist 75, 71–88.[Abstract]
Waters, D. J. (1988). Partial melting and the formation of granulite facies assemblages in Namaqualand, South Africa. Journal of Metamorphic Geology 6, 387–404.[Web of Science]
Waters, D. J. & Whales, D. (1984). Dehydration melting and the granulite transition in metapelites from southern Namaqualand, S. Africa. Contributions to Mineralogy and Petrology 88, 269–275.
White, A. J. R. & Chappell, B. W. (1988). Some supracrustal (S-type) granites of the Lachlan fold belt. Philosophical Transactions of the Royal Society of Edinburgh 79, 169–181.
Winkler, B., Milman, V. & Payne, M. C. (1994). Orientation, location, and total energy of hydration of channel H2O in cordierite investigated by ab-initio total energy calculations. American Mineralogist 79, 200–204.[Abstract]
Young, E. D., Anderson, J. L., Clarke, H. S. & Thomas, W. M. (1989). Petrology of biotite–cordierite–garnet gneiss of the McCullough Range, Nevada I. Evidence for Proterozoic low-pressure fluid-absent granulite-grade metamorphism in the southern Cordillera. Journal of Petrology 30, 39–60.
Zimmermann, J.-L. (1981). The liberation of H2O, CO2 and hydrocarbons from cordierites: kinetics, structural sites and petrogenetic implications. Bulletin de Minéralogie 104, 325–338.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. J. Rigby and G. T.R. Droop The cordierite fluid monitor: case studies for and against its potential application European Journal of Mineralogy, October 1, 2008; 20(5): 693 - 712. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Rigby, H. Mouri, and G. Brandl P-T conditions and the origin of quartzo-feldspathic veins in metasyenites from the Central Zone of the Limpopo Belt, South Africa South African Journal of Geology, September 1, 2008; 111(2-3): 313 - 332. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. E. Paukov, Y. A. Kovalevskaya, N.-S. Rahmoun, and C. A. Geiger Heat capacity of synthetic hydrous Mg-cordierite at low temperatures: Thermodynamic properties and the behavior of the H2O molecule in selected hydrous micro and nanoporous silicates American Mineralogist, February 1, 2007; 92(2-3): 388 - 396. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. KAINDL, P. TROPPER, and I. DEIBL A semi-quantitative technique for determination of CO2 in cordierite by Raman spectroscopy in thin sections European Journal of Mineralogy, June 1, 2006; 18(3): 331 - 335. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. JOHNSON and M. BROWN Quantitative Constraints on Metamorphism in the Variscides of Southern Brittany--a Complementary Pseudosection Approach J. Petrology, June 1, 2004; 45(6): 1237 - 1259. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Cesare, B. Cesare, G. Cruciani, and U. Russo Hydrogen deficiency in Ti-rich biotite from anatectic metapelites (El Joyazo, SE Spain): Crystal-chemical aspects and implications for high-temperature petrogenesis American Mineralogist, April 1, 2003; 88(4): 583 - 595. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. THOMPSON, S. L. HARLEY, and D. P. CARRINGTON Sodium and potassium in cordierite - a potential thermometer for melts? European Journal of Mineralogy, June 1, 2002; 14(3): 459 - 469. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||