Starting materials

Natural mineral powders of Fe-rich biotite, spessartine garnet, petalite, quartz, and intermediate montebrasite-amblygonite were added to mineral mixtures (HGS4, HGS5, SPC), to green rhyolite obsidian from Macusani, Peru (e.g. London et al., 1988), and, in a few experiments, to sol-gel glasses (HGS1, HGS1F1, HGS1F2, HGS3) of haplogranite composition (Table 1). The haplogranite mineral mixture HGS4 was composed of albite, orthoclase, and quartz in a proportion to match the 200 MPa H₂O minimum Ab₃₈Or₂₈Qtz₃₄ (Tuttle & Bowen, 1958). Glasses HGS1 and HGS3 are both comparatively sodic but differ in their ASI values (Table 1); HGS1F1 and HGS1F2 have the same composition as HGS1 with the addition of 1 and 2 wt % F. Dehydrated synthetic gibbsite was added to HGS4 to raise the ASI of the bulk composition and produce HGS5. Mixture SPC contained quartz, albite, orthoclase, and muscovite. The analyses listed in Table 1 are for hydrous glasses produced by melting the bulk compositions at 200 MPa H₂O; HGS5 contained minor quantities (< ~0·1 vol. %) of corundum with glass. One set of experiments employed a 1:1 weight mixture of biotite + reagent Mg₃(PO₄)₂·8H₂O with an equal weight fraction of HGS5. A second set ofexperiments utilized spessartine garnet + reagent MnHPO₄·XH₂O in a 1:1 weight mixture added to an equal weight fraction of HGS5. For the petalite-montebrasite equilibrium, a milled mixture of petalite (± spodumene), montebrasite, and quartz in equal weight proportions of the three constituents was added to an equal weight fraction of the haplogranite mineral mixture HGS4 or macusanite glass. The low solidus of the macusanite obsidian (London et al., 1989) allowed experimentation to temperatures as low as 525°C. In the experiments with LiAl silicates and phosphates, F was added to the system via the natural amblygonite-montebrasite starting phase, as a component of HGS1F1 and HGS1F2, or by the Macusani obsidian (Table 1). In total, the F content of hydrous melt was varied from 0·6 to 1·8 wt % F in experiments that contained amblygonite-montebrasite.

Table 1. Compositions of starting materials

Experimental methods

In all runs, the mineral or mineral-glass mixtures were sealed with a slight excess of H₂O needed to saturate melt at P and T into 20 mm * 3 mm Au capsules, with the powder mix confined to a central 5 mm * 3 mm portion of capsule. Experiments were pressurized cold, then heated to the run temperature. Forward-direction experiments were taken directly to run T. Reverse-direction experiments were preconditioned at 850°C, quenched in air to room T, then raised to final run T as for the forward-direction experiments. The purpose of the reverse-direction experiments was to use high T to oversaturate the melt in the components of interest relative to their concentrations in melt at the final run T. In this way, the P content of melt via a particular reaction at a specific final T was approached from higher (reverse-direction) and lower (forward-direction) P in melt at the same final T. Experiments were considered to have attained equilibrium if the P contents of glasses produced by these two different approaches to the final state were equivalent. Experiments were 2-4 weeks long. Capsules were run subhorizontally in water-pressurized cold-seal René-41[reg] and NIMONIC 105[reg] vessels at 200 MPa. Pressure was measured with a factory-calibrated Heise bourdon tube gauge, with fluctuations of <5 MPa over the course of experiments. Temperature was monitored by internal chromel-alumel thermocouples, with an estimated total error ±5°C. Oxygen fugacity was not controlled, but the intrinsic oxygen fugacity of these vessels plus their Ni-metal filler rods and hydrocarbon rust inhibitor is slightly below the nickel-nickel oxide (NNO) buffer (e.g. Huebner, 1971). Runs were quenched isobarically in air (5-10°C/s). Capsules were reweighed to test for leaks; all experiments gained weight because Ni (systematic with run time and temperature) from the reaction vessel dissolved into the precious metal tubing; Ni did not contaminate the silicate charge.

Characterization and chemical analysis

Chemical analyses were obtained by wavelength-dispersive spectrometry (WDS) on a Cameca SX-50 electron microprobe using crystalline standards with TAP, PET, LIF, and layered composite diffraction device (2d = 65 Å). The two-beam condition (2 nA and 20 nA regulated beam current) analysis procedure developed by Morgan & London, (1996) was used for electron microprobe analysis (EMPA) of glasses. Analyses of minerals were conducted mostly at 20 kV, 20 nA regulated beam, and 3 µm spot size. The PAP correction procedure was used (Pouchou & Pichoir, 1985). The detection limit for phosphorus was 0·04 wt % P₂O₅ (with a 20 nA beam) at 3[sgr] above mean background. Amblygonite-montebrasite is quickly damaged (evidenced by a substantial increase in the count rates for F with time) by a focused electron beam, and where possible beam diameters of 10-20 µm were used.

Reverse-direction experiments typically produce coarser-grained crystalline silicates more suitable for EMPA than forward-direction experiments, and the inheritance of relict starting material-problematic in forward-direction experiments-is minimized in reversed runs. Most of the phosphates crystallized by reverse-direction methods in this study, however, formed dendritic or vermicular skeletal crystals (albeit huge, covering areas of ~10⁵ µm²) that were commonly too narrow for accurate EMPA (Fig. 1e and f). Analyses of even the coarsest-grained FeMgMn phosphate minerals contained a melt component. For these phases, we used the K content of the phosphate minerals (and other components in their proportions within the surrounding glass) to correct the EMPA by subtraction of a melt component. The result yields very good fits to the phosphate mineral stoichiometries. Amblygonite-montebrasite that crystallized in reverse-direction experiments formed dendritic crystals that were mostly too fine for EMPA. Forward-direction experiments at low T produced new overgrowths (e.g. Fig. 1e) but grains were small. Because the requisite EMPA beam size is large and F = OH exchange produces a negligible change of mass (and hence back-scattered electron signal intensity), we cannot rule out the possibility that the microprobe analyses contain a component of the starting montebrasite from Tanco. The partitioning of F between amblygonite-montebrasite and melt (discussed below) suggests that this is indeed the case.

Figure 1. Back-scattered electron images of new experimental run producs. Run labels correspond to Tables 2- 7. Scales are for full-frame. Phase abbreviations are glass (gl), K-feldspar (Kfs), quartz (Qz), biotite (Bt), cordierite (Crd), spessartine (Sps), petalite (Pet), montebrasite (Mbr), sarcopside (Src), and eosphorite (Eos).

NEW EXPERIMENTAL RESULTS

Biotite-phosphate equilibria

An equilibrium among biotite (intermediate annite-phlogopite-siderophyllite solid solution), sarcopside, and cordierite (Fig. 1b) was attained with 1·30 wt % P₂O₅ (700°C) to 1·57 wt % P₂O₅ (850°C) in peraluminous melt (Tables 2 and 3) derived from the starting composition HGS5. The observed reaction relationship in this range of temperature is

3K(Fe,Mg)3AlSi3O10(OH)2 + 6Al2SiO5(m) + 9SiO2(m) + P2O5(m) =  (Fe,Mg)3(PO4)2 + 3(Mg,Fe)2Al4Si5O18 + 3KAlSi3O8 + 3H2O.

(3)

K-rich alkali feldspar (Fig. 1a) appeared as a crystalline phase in near-solidus experiments (e.g. experiment HGMgP-8, Table 2) or as an added component of melt at higher T. Starting biotite reacted with silica and excess alumina of HGS5 to form cordierite.

Table 2. Experimental run series HGMgP¹, glass compositions; HGS5 and SPC bulk compositions

Table 3. Experimental run series HGMgP, crystalline phase compositions; HGS5 bulk compositions

One reconnaissance experiment with peraluminous SPC mix plus biotite and triplite (TRP-6, Tables 1 and 2) produced minor triplite, (Mn,Fe,Mg)₂PO₄(F,OH), Mn-apatite, and various other crystalline Ca(Mn)-phosphates. We abandoned further work with the natural triplite because its Mn and Ca contents displaced the bulk composition too far away from that of reaction (3). However, the high Ca and F content of the starting triplite promoted saturation of melt in apatite and sharply reduced the solubility of P in melt.

At a given temperature, the mg-number [molar 100 * Mg/(Mg + Fe)] of phases varied as mg-number_sarcopside > mg-number_cordierite mg-number_biotite mg-number_melt (Table 3). This set of experiments did not contain Mn above the microprobe detection level. The sum of mafic components (MgO + FeO) in glasses was low and increased from 1·09 wt % oxides at 650°C to 2·01 wt % oxides at 850°C. These values are equivalent to the mafic content of nominally P-free melts of similar composition (e.g. Icenhower & London, 1995). The P content of melt (1·30-1·57 wt % P₂O₅) at equilibrium (3) mirrored the small increase in mafic components. These P values, however, are nearly twice those of apatite-saturated experiments (e.g. Wolf & London, 1994).

Garnet-phosphate equilibria

An apparent equilibrium was attained between spessartine and sarcopside at 0·96 wt % P₂O₅ (600°C) to 2·40 wt % P₂O₅ (850°C) in peraluminous melts derived from HGS5 and HGS3 (Tables 4 and 5). The reaction relationship in this temperature range is expressed as

Mn3Al2Si3O12 + P2O5(m) = Mn3(PO4)2 + Al2SiO5(m) + 2SiO2(m).

(4a)

Garnet exhibits sharply euhedral outlines (Fig. 1c), which we infer to represent new growth. There was no discernible difference, however, in the composition of starting vs euhedral garnet; the phosphorus content of both is below EMPA detection. The mafic component of melt, almost entirely from MnO, increased from ~1·25 to 3·25 wt % between 600 and 850°C. Icenhower et al., (1994) and Icenhower & London, (1995) also observed that Mn is more soluble than Fe + Mg at comparable P, T, and melt compositions. The P₂O₅ content of melt tracked the concentration of MnO over the same T interval. Eosphorite appeared in addition to sarcopside mostly in forward-direction experiments to T <= 750°C (Fig. 1d and Table 4). We infer that eosphorite is a stable phase in the peraluminous system that contains garnet, sarcopside, and melt at T <= 750°C but generally failed to nucleate in melts cooled from higher temperatures. Hence, an overall model reaction for this system is

4Mn3Al2Si3O12 + 5P2O5(m) + 9H2O(m) =  2Mn3(PO4)2 + 6MnAlPO4(OH)·H2O + Al2SiO5(m) + 11SiO2(m).

(4b)

Table 4. Experimental run series HGMnP¹, glass compositions; HGS3 and HGS5 bulk compositions

Table 5. Experimental run series HGMnP, crystalline phase compositions; HGS3 and HGS5 bulk compositions

Lithium aluminosilicate-phosphate equilibria

The equilibrium attained among petalite (the analogous reaction with spodumene is obvious), quartz, and montebrasite (Fig. 1e) is represented by the reaction (London & Burt, 1982a)

LiAlSi4O10 + PO2(OH,F)(m) = LiAlPO4(OH,F) + 4SiO2.

(5)

Above 700°C, the LiAl-silicate virgilite took the place of petalite + quartz [also see London, (1984)]. A new and presumably metastable LiAl-silicate with feldspar stoichiometry (LiAl silicate, Tables 6 and 7) occurred with quartz at 675 and 700°C. Experiments below 675°C grew petalite (petalite, Table 7) plus quartz. The new petalite that formed in reverse-direction experiments between 575 and 635°C is noteworthy for its texture: whereas other phases had grain sizes of ~10-20 µm, the few petalite crystals formed oikocrysts of ~500 µm that contain inclusions of montebrasite and quartz (Fig. 1f).

Table 6. Experimental run series HPLi and MPLi, glass compositions

Table 7. Experimental run series HPLi and MPLi, crystalline phase compositions

The solubility of amblygonite-montebrasite with LiAl-silicates ± quartz increased rapidly with T (Fig. 2a) from 1·4 wt % P₂O₅ at 525°C to 7·2 wt % P₂O₅ in melt at 700°C (Table 6). Amblygonite-montebrasite was unstable with >9 wt % P₂O₅ in melt at 750°C. Table 6 lists four forward-direction experiments, HPLi-15 to HPLi-18, in which only the starting F content of the system varied. The results show an inverse variation of P with F (Fig. 2b), consistent with the solubility product [aLi][aP][aF]. The data, however, fit a third-order regression perfectly (R = 1·00). We offer two interpretations of these preliminary data: (1) amblygonite is more stable than montebrasite ([Delta]G_f^Amb < [Delta]G_f^Mbr) at the conditions of the experiments (H₂O-saturated), and (2) above ~1·5 wt % F in melt, F forms melt species that do not affect [aLi][aP][aF].

Figure 2. Plots of (a) the concentration of P₂O₅ in melt vs temperature as determined by the equilibrium (5), with data from Tables 6 and 7; (b) the concentration of P₂O₅ in melt vs F in melt; and (c) the concentration of F in amblygonite-montebrasite vs F in melt.

Variations in F content of melt from 0·6 to 1·8 wt % F did not change the crystalline phase assemblage. It should be noted that petalite, as opposed to F-rich mica such as lepidolite, grew in runs MPLi-3 and MPLi-12 with ~1·5 wt % F in peraluminous melt (London, 1982). Applying a linear regression to the data in Tables 6 and 7, the apparent partition coefficient for F between amblygonite-montebrasite solid solutions and melt, D_F^Amb/melt, is 2·79 for experiments between 600 and 700°C (Fig. 2c), with a fair Pearson correlation (0·92) but an unrealistic y-intercept value (2·13 wt % F in montebrasite at zero F in melt). Either F = OH mixing in amblygonite-montebrasite solid solution is far less ideal than has been proposed (see Loh & Wise, 1976), or this high intercept value stems from incomplete equilibration of the starting material (i.e. inclusion of relict montebrasite in the EMPA of experimental products). We propose that the slope of C_F^Amb vs C_F^melt (= D_F^Amb/melt) should be ~4-5, as derived from the amblygonite-montebrasite at intermediate F content.

Apatite and the phosphorus content of initial melts

Magmas derived from the incipient partial melting of typical metapelite assemblages will, if equilibrium is attained during melting, be strongly peraluminous (ASI = 1·3-1·4) regardless of the activity of H₂O in melt (e.g. Holtz et al., 1992; Icenhower & London, 1995; Patiño Douce, 1996; Wolf & London, 1997). At such high ASI values, the melting of apatite can contribute up to 0·7 wt % P₂O₅ to melt at low temperatures [750°C in the experiments by Wolf & London, (1994)], and higher phosphorus contents with increasing T (e.g. Harrison & Watson, 1984).

Wolf & London, (1994) identified the local enhancement of apatite solubility caused by chemical diffusion between apatite grains and melt. This process should augment the dissolution of apatite. With high solubility in strongly peraluminous and typically Ca-poor liquids derived from such metapelites, it is likely that all apatite that is exposed to the melt will dissolve at the source, and hence that the P₂O₅ of initial melt will be limited by the abundance of apatite. In the unlikely case that P is available from sources other than apatite, monazite, and xenotime, then the limiting equilibria at source will be those such as the biotite-sarcopside reaction (3) above, because anatexis will probably not exhaust the mafic silicates. Pelitic rocks possess molar Ca/P ratios >1·67 (sources: Gromet et al., 1984; Haack et al., 1984; Coveney & Glascock, 1989; J. Ague, personal communication, 1995; Moss et al., 1996), the stoichiometry of apatite, and apatite is probably the sole reservoir of P. These same sources report <= 0·5 wt % P₂O₅ and mostly <0·2 wt % P₂O₅ in the pelite whole rocks. Thus, the likely scenario for metapelite-derived magmas is that they will be undersaturated with respect to apatite at the source, contain <0·7 wt % P₂O₅ at inception, and begin to fractionate plagioclase but not apatite with crystallization upon ascent.

Mafic silicate-phosphate equilibria

Primary mafic silicates, together with apatite as the only phosphate, are far more common in granites and pegmatites than are mafic phosphate minerals. This implies that the P contents of most granite-pegmatite melts do not reach the values of equilibria (3) and (4), but rather lie near the lower values for apatite saturation. For low-T, Ca-poor anatectic melts derived from metapelites, early saturation (e.g. Nekvasil, 1988) and crystallization of plagioclase can reduce the Ca/P ratio below that of apatite (e.g. London et al., 1989), which ensures that some P will be incorporated in other crystalline phosphates (in addition to the alkali feldspars). The presence of mafic phosphates but not mafic silicates with feldspars and quartz, conversely, suggests that the P content of the melt exceeded the buffer established by the relevant equilibria [e.g. (3) or (4) above]. Pegmatites in which most of the mafic components are boundin phosphates fall into Cerny's, (1991) beryl-columbite-phosphate pegmatite subtype.

Reported assemblages of biotite or garnet with mafic phosphate minerals appear to be rare, but this may be due in part to the nature of studies of the pegmatitic phosphates, which historically have focused more on mineralogy than on petrogenesis. If mafic silicates and phosphates do occur in the same granite or pegmatite body, they usually do not occur together, but rather are sequential to one another, i.e. the phosphates succeed the silicates with little or no overlap in their paragenesis. Biotite is rarely associated with the more typical (Li)Mn-phosphates. A few localities, however, possess assemblages for which equilibrium (3) may be applicable: biotite occurs in a primary association with triphylite in New Hampshire (Chapman, 1943) and with triplite in Colorado (Wolfe & Heinrich, 1947). Lithium-iron micas (Li-siderophyllite, approaching zinnwaldite) occur with combinations of childrenite [FeAlPO₄(OH)₂^.H₂O)]-eosphorite, triplite-zwieselite, or triphylite in evolved granites and pegmatites of the Krusne Hory, Czech Republic (Breiter et al., 1997), and the Erzgebirge region of southern Germany (Webster et al., 1997). Childrenite and triplite also occur with biotite and tourmaline in a marginal facies of the Alburquerque batholith, Badajoz, Spain, as discussed below. Assemblages of garnet + Mn-rich apatite are common in granites and pegmatites. As with biotite, however, there are few localities where primary mafic phosphate minerals such as sarcopside (e.g. Keller, 1988; Zhang, 1995) occur with garnet.

Lithium aluminosilicate-phosphate-quartz

In contrast to mafic silicate-phosphate assemblages, the assemblage of spodumene or petalite + amblygonite-montebrasite + quartz is common in Li-rich pegmatites and some granites. A few natural systems [e.g. the complex amblygonite subtypes (Cerny, 1991)] contain amblygonite-montebrasite in the absence of any LiAl-silicates (petalite, spodumene, or lepidolite) and probably exceeded the buffer capacity of reaction (5). In the Li-free haplogranite system, there is complete silicate-phosphate miscibility in melt and no P-saturating crystalline phase (e.g. lacroixite, ideally NaAlPO₄OH) with up to 14 wt % P₂O₅ in melt at T as low as 550°C (London et al., 1993b).

Experimental results expressed by reaction (5) are different from those of (3) and (4) in that the solubility of the phosphate phase increases sharply with increasing temperature, a characteristic not seen in the experiments with the mafic phosphates. This is not surprising, because the components of amblygonite-montebrasite represent an alkali charge-balanced melt-forming component similar to that of the haplogranitic melt. Reaction (5) has two important independent variables, T and P, and of these two, P is the better constrained, either through the utilization of the alkali feldspars [equation (1)] or the mafic silicate-phosphate equilibria (3) and (4). If the reactions (1), (3), and (4) buffer the P content of melt, then reaction (5) could be used crudely as a geothermometer. With ~1-2 wt % P₂O₅ in melt, the equilibrium (5) will occur at a temperature <~500-600°C. Such low temperatures are consistent with the majority of amblygonite-montebrasite occurrences in pegmatites.

A case study: the Alburquerque granite at Tres Arroyos, Badajoz, Spain

The Alburquerque batholith, a zoned, strongly peraluminous granite and its related pegmatite group, near the Tres Arroyos Sn prospect, Badajoz province, Spain (Gallego Garrido, 1992; London et al., 1993a), provides excellent control for a study of a magma's P budget across a narrow marginal interval that exhibits extreme chemical fractionation (Table 8). The Alburquerque batholith is a large composite pluton that spans the border from near Nisa, Portugal, to Alburquerque, Spain (Fig. 3). Its most fractionated portions occur along its SE margin in the vicinity of the Sn mining prospect at Tres Arroyos, ~10 km WNW of Alburquerque. There, the batholith contains at least five distinct granite-pegmatite facies, which, based on field relations and major- and trace-element chemistry, appear to have evolved by differentiation from one common magma without multiple injection or post-emplacement metamorphism. The rock types (Table 8) include the main central granite facies (CGF), which grades WSW to a narrow marginal granite facies (MGF), which in turn is transitional into a classic layered aplite-pegmatite border facies (LAP). The MGF is <1 km wide as exposed, and the LAP is discontinuous and rarely more than 200 m wide as exposed. Thin leucocratic dikes (LED), mostly <1 m thick, intruded into host phyllites ~100 m beyond the LAP contact with the phyllites. The LED dikes appear to have been a magma feeder system to the closely associated, zoned Li pegmatites (LIP), which are of the complex lepidolite subtype (Cerny, 1991). The LIP dikes are not exposed beyond 300 m from the LAP-phyllite contact. The CGF is coarse grained with subhedral K-feldspar phenocrysts in a hypidiomorphic granular groundmass. The MGF is fine grained with uniform hypidiomorphic granular texture. Aplitic portions of the LAP are texturally similar to the MGF but with intercalated pods of very coarse-grained miarolitic pegmatite. The LED is fine grained and texturally serriate to porphyritic. The LIP is one of several zoned bodies in which very coarse-grained pegmatitic texture is absent, but unidirectional solidification textures inward from the margins and graphic intergrowths of plagioclase, micas, and quartz are ubiquitous. The inward sequence of mineral assemblages in the LIP appears to be: (1) albite-quartz-topaz, (2) quartz-K-feldspar-albite, (3) albite-lepidolite-quartz, (4) massive quartz, and (5) lepidolite-topaz replacement along the boundary between (2) and (3). The bulk compositions of zoned pegmatites are notoriously difficult to ascertain, and although great effort was expended to obtain large representative samples of the LIP (trench samples taken perpendicular to the layering) in a large quarry, the footwall portion of this dike is below water level and was not sampled. The most uncertainty in any of the compositions lies with these pegmatite bodies. However, because the LED is a fine-grained rock that appears to have fed magma to the LIP, it may represent the LIP as well. All of the granitic and pegmatitic facies are strongly peraluminous and whole-rock (WR) P₂O₅ is high, so that the feldspars are useful monitors of the phosphorus content of melt.

Table 8. Compositions of granitic and pegmatitic facies of the Alburquerque batholith near Tres Arroyos, Badajoz, Spain

Figure 3. Location and regional geologic map around the Alburquerque batholith, Spain.

Major elements

The transition from CGF to LIP produces a marked decrease in the K* and silica content. This variation is consistent with the accumulation of P, F, and B in residual melt via crystal fractionation [e.g. fig. 2 of London et al., (1993b)]. The ASI values of fresh whole rocks also increase from CGF to LIP and approach the value of 1·35, which is the ASI of minimum melts derived from muscovite-bearing protoliths (e.g. Icenhower & London, 1995). Both trends imply that each successive granite facies at Tres Arroyos represents a (mostly) liquid differentiate that has separated from a cognate crystalline fraction (see Bea et al., 1994).

Silicate-phosphate assemblages

The mafic silicate mineral assemblage changes from biotite + cordierite in the CGF to biotite + tourmalinein the MGF and tourmaline only in the LAP. The LED and LIP both lack mafic minerals other than NbTaSn-oxides. Other aluminous phases (together with muscovite) in these assemblages include andalusite in the CGF and MGF, topaz in the LAP, LED and LIP, and lepidolite in the LED and LIP. Among the phosphates, coarse-grained apatite occurs mostly in the CGF and decreases sharply in abundance into the MGF. Childrenite appears in the CGF and is prevalent in the MGF and LAP (Table 9) along with eosphorite, sarcopside, and triplite. Although the textures of most of the mafic and other phosphate minerals (e.g. goyazite, Fig. 4b) connote a secondary or subsolidus origin (Fig. 4e and f), triplite and sarcopside occur as minute euhedral crystals in coarse-grained quartz (Fig. 4g and h). Some fraction of these mafic phosphate minerals, therefore, appears to have crystallized above the solidus of the MGF (with biotite and tourmaline) and the LAP (with tourmaline).Accessory mineral assemblages change from the LAP to the LED and LIP. In the LED and LIP, amblygonite-montebrasite is the principal and only primary-appearing phosphate mineral. Fine-grained secondary apatite and goyazite occur mostly along cleavages or fractures within or between quartz and feldspar, though very fine-grained apatite in the finer portions of the LED may be part of the (quenched?) silicate groundmass.

Table 9. Composition of phosphates from Tres Arroyos

Figure 4. Back-scattered electron and X-ray images of samples from the Alburquerque batholith at Tres Arroyos, Badajoz, Spain: (a) fine-grained apatite (Ap) is associated with sericite (Ms) as an alteration product in plagioclase (Pl) from the marginal granite facies (MGF); (b) secondary skeletal apatite (Ap) and radial goyazite (Goy) associated with aluminosilicate (Als) at a grain boundary with plagioclase (Pl) from the MGF; (c) potassium K[alpha] X-ray image of small K-feldspar grain (Kfs) plus included plagioclase (Pl), and muscovite (Ms) from the layered aplite of the LAP; (d) a P K[alpha] X-ray image of the same grain shows a heterogeneous zonation of P in the Kfs, more variability in the Pl, and a labyrinth of secondary phosphates along fractures and cleavage; (e) eosphorite (Eos) filling fractures and interstices between plagioclase (Pl) and quartz (Qz) from the LAP; (f) a typical interstitial vein filling of secondary apatite (Ap), wyllieite (Wyl), and sarcopside (Src) or vivianite (Viv) in the LAP; (g) euhedral grain of triplite (Trp) as an inclusion in quartz from the LAP; (h) minute euhedral crystal aggregate of childrenite-eosphorite, sarcopside, or triplite with biotite (Bt) as an inclusion in quartz from the MGF.

Chemistry and history of the feldspars

The alkali feldspars (Afs) mostly show normal fractionation trends in major-element chemistry from the CGF to the LIP. The An content of plagioclase (Pl) drops sharply from An_07-09 in the CGF to An_01-02 in the MGF and the other remaining facies (Table 10). The Or content of K-feldspar (Kfs) increases more gradually from the CGF (Or₈₇) to the LIP (Or₉₈). Samples from most facies contain multiple textural generations of Afs; e.g. porphyritic vs groundmass phases, Pl exsolved from Kfs, and, in many samples, a late interstitial generation of Kfs.

Table 10. Compositions and phosphorus contents of feldspars from the Alburquerque granite-pegmatite complex at Tres Arroyos, Badajoz, Spain

The P₂O₅ contents of the coarse-grained, primary-appearing Afs (Table 10) generally increase from the CGF (average 0·43 wt % P₂O₅: Table 11) to the LAP (average 1·21 wt % P₂O₅: Table 11) then fall in the LED and the LIP, where the P₂O₅ content of coarse-grained plagioclase is similar to that of the CGF. The distribution of P is heterogeneous within individual crystals; the 1[sgr] deviations (Table 10) are mostly 20-50% of the mean value. X-ray images reveal that the heterogeneity of P in the Afs is patchy [also see London, (1992)] and lacks any obvious relation to the likely pattern of growth zoning [compare fig. 7 of London, (1992)]. The full range of values (Min/Max, Table 10) is substantially larger. From this data set, the single highest value (individual point analysis) from Kfs in the LAP is 2·60 wt % P₂O₅. This is the highest value yet reported from any natural feldspar.

Table 11. Calculated P₂O₅ in melt and actual P₂O₅ in whole rock

One of the more interesting and generally useful features of the feldspar chemistry is revealed by the empirical distribution coefficient, D_P^Kfs/Ab, between the Kfs and Pl phases (Table 10). Because these natural feldspars are close to their Ab and Or end members, the distribution of P between Kfs-Pl pairs in equilibrium should be ~1·2 [equation (2)]. Deviations from this value reflect a lack of equilibrium between the Kfs-Pl pairs. Several specific illustrations are given below.

In sample TA22 of the CGF (Table 10), coarse-grained and primary-appearing Kfs and Pl have high and nearly equivalent P₂O₅ values and hence appear to have grown from melt in near-equilibrium. Plagioclase patches and stringers that appear to be exsolved from the Kfs, however, have similar P contents. Interstitial Kfs, which occurs texturally late in the paragenesis, contains lower P₂O₅ and is not in equilibrium with any of the other Afs phases. A similar relationship between Pl and interstitial Kfs exists in LIP samples TA28C and TA29, in which D_P^Kfs/Ab is much less than unity (Table 10). The P content of the interstitial Kfs in these facies identifies this Kfs as subsolidus in origin and not in equilibrium with any other Afs. The patchy and stringer Pl in Kfs, however, has a different origin. If this exsolved at subsolidus conditions, then it too should have low P. The high P content of the Pl in Kfs, however, is consistent with a primary intergrowth of these feldspars (e.g. Petersen & Lofgren, 1986; London et al., 1989). Using the P content as a guide, primary patchy or stringer intergrowths of Pl in Kfs can be distinguished from perthite of subsolidus origin, and this distinction is important for geothermometry based on integrated feldspar compositions.

An underlying assumption of the example above is that the P content of melt, and hence the Afs, increases until the eventual precipitation of phosphate minerals at near- or subsolidus conditions consumes all available P. Deviations of D_P^Kfs/Ab from unity, however, do not always signify that one or the other feldspar is hydrothermal in origin. In sample TA 21 of the CGF (Table 10), for example, coarse-grained Kfs and Pl appearto have crystallized close to equilibrium from melt (D_P^Kfs/Ab= 1·10). Finer-grained Pl has higher P₂O₅ values than either of the coarser-grained primary phases (D_P^Kfs/Ab = 0·62), which in this case probably reflects the continued crystallization of Pl from melt.

If D_P^Kfs/Ab is much greater than unity, the Pl phase has probably recrystallized in the subsolidus after most P has been precipitated. For example, the value of D_P^Kfs/Ab (3·75) between Kfs and stringer Pl in sample TA23G from the LIP identifies this Afs as a typical perthite formed by exsolution of Pl. In sample TA7 from the CGF-MGF margin (Table 10), Kfs and Pl occur asmutually separate and coarse-grained crystals, but D_P^Kfs/Ab= 5·50 and the P content of the Pl is low in comparisonwith other samples; from this we recognize that the Kfs is of igneous origin, but the Pl, though of similar grain size and primary appearance, is not, or at least is not comagmatic with this suite. In sample TA19 from the MGF (Table 10), the medium-grained Kfs and Pl appear to represent a comagmatic pair (D_P^Kfs/Ab = 0·99), but the coarse-grained Pl, which on the basis of texture is most likely to be magmatic, is evidently not (D_P^Kfs/Ab = 7·00). In this case, the two texturally distinct Pl phases could have been distinguished by their An contents-An₀₄ for the medium-grained Pl but An₀₀ for the coarse-grained phase.

Finally, there is a significant disparity between the P₂O₅ contents of Kfs and Pl in the layered aplitic portions of the LAP vs Kfs and Pl in the pegmatitic segregations (samples TA20 and TA23G, Table 10). The high P content of Afs in the layered aplite indicates that this body is magmatic in origin. Low P contents of the Afs in the miarolitic pegmatites, however, could be taken to indicate their crystallization from vapor only, as the partition coefficient for P between granitic melt and vapor is much less than unity (London et al., 1993b; Keppler, 1994). X-ray maps of these pegmatitic Afs, however, show them to be riddled with thin veinlets of mafic and aluminous (goyazite) phosphates. These feldspars probably once possessed higher, magmatic P contents similar to those of Afs in the aplite facies. Like the interstitial Kfs, the pegmatitic feldspars from the LAP probably reflect where aqueous vapor was concentrated at the transition from magmatic to hydrothermal conditions. If this is correct, then the other feldspars with high P contents preserve the igneous history of the rocks.

Table 10 shows other examples of comagmatic pairs, along with clear and not-so-clear evidence for exsolution of Pl from Kfs, and other cases where one or another of the texturally distinct Afs phases is not in equilibrium with another. The causes of the selective recrystallization of one feldspar phase, like the causes of low phosphorus contents in feldspars of apparently hydrothermal origin, are not readily evident. The important point, however, is that these different generations of feldspar can be readily identified.