The Alboran basement at Site 976

ODP Site 976 is located in the west Alboran Basin on a buried basement high bounded by normal faults (Fig. 1), probably active in Early to Middle Miocene time (Comas et al., 1992; Watts et al., 1993: Comas & Soto, 1999). The crust in the area is probably between 15 and 20 km thick (Hatzfeld et al., 1978). The results of the earlier Deep Sea Drilling Project (DSDP) Site 121 drill hole nearby had previously suggested the presence of metamorphic continental basement (Kornprobst, 1973; Ryan et al., 1973). At Site 976 the metamorphic basement is covered by a 670 m thick, discontinuous sequence of sediments of Serravallian (Middle Miocene) to Pleistocene age (Shipboard Scientific Party, 1996). The basal contact of the sediments is sharp and characterized by the presence of a variety of basement pebbles, and appears to be an unconformity, although the presence of breccias down to 40 m below the contact and the significant difference in the depth of this contact between two adjacent holes indicates active faulting during early to middle Miocene time (Shipboard Scientific Party, 1996). The overlying sedimentary sequence therefore places a younger age limit of ~13 Ma on the exhumation of the metamorphic rocks.

The top 124 m of cored basement at Hole 976B (1108 m, water depth) consists of dark grey, quartz-biotite-sillimanite-plagioclase schist, with visible porphyroblasts of garnet and andalusite. This metapelite sequence has interlayers of calcite and dolomite marble and associated calc-silicate rocks as reaction bands that developed along the metapelite-metacarbonate contacts. Metacarbonate and associated rocks are particularly abundant in the upper part of the cored section (Fig. 2). This metapelite sequence overlies banded pelitic gneiss with coarse-grained K-feldspar, cordierite, and andalusite. Much of the gneiss is migmatitic, with centimetre-thick veins and segregations of weakly foliated or unfoliated felsic material (leucosome with a peraluminous granite composition) containing large crystals of cordierite, as well as biotite, andalusite, and sillimanite. The most common textural type for these migmatites is stromatic migmatites [terminology of Mehnert, (1968) and McLellan, (1983)]. In some instances, the granitic leucosome also occurs as veins cutting across the foliation and in pull-aparts, which locally gives the gneiss a brecciated structure (agmatitic texture). Leucosome geometries largely following pre-existing planes of anisotropy, the low volume of leucosome veins (<20%) with respect to the whole-rock volume, and the common occurrence of mafic selvedges (sillimanite-rich melanosomes) associated with them suggest that the low degree of melting was controlled by a strong, compositional foliation (see Brown et al., 1995; Vigneresse et al., 1996). Major and trace element contents of migmatite and pelite gneiss are similar to high-grade schist composition trends (Spadea & Prosser, 1999) suggesting a local source and low mobility of melts during migmatite formation. Therefore, these leucosomes represent metatexite migmatites [following Brown, (1973) and Sawyer, (1996)].

Figure 2. Schematic lithological columns of the metamorphic basement cored in ODP Hole 976B and Hole 976E [for more details, see Shipboard Scientific Party, (1996)]. Recovered intervals are shown in black in the recovery column. [triangle], location of samples discussed in the text.

Both rock units are cut by peraluminous, leucogranitic dykes, as well as numerous zones of brittle fault gouge and breccia. The high-grade schist and the gneiss show distinctive differences in their petrological and structural evolution, and at Hole 976B are probably separated by a significant tectonic boundary, represented by a thick band (up to 20 m thick) of fault rocks (consolidated clay breccia, fault gouges, and cataclastic rocks) (Comas & Soto, 1999).

The lithological association and metamorphic characteristics at the site broadly resemble those within parts of the Alpujarride Complex of the adjacent Betic Cordillera (and its counterparts the Sebtides in the Rif). A sequence formed by biotite + K-feldspar + sillimanite-rich pelitic schist, granular cordierite-bearing gneiss, and marble intercalations, commonly cut by leucogranite dykes, characterizes those Alpujarride units underlying the Ronda ultramafic massif some 80 km north of the drill-site (Fig. 1). These rocks reach high-amphibolite to granulite grade conditions with evidence of migmatization and partial melting under relatively low-pressure conditions (<6 kbar) (Loomis, 1972; Westerhof, 1977; Lundeen, 1978; Sánchez-Gómez et al., 1995; Tubía et al., 1997). The Alpujarride Complex as a whole comprises a stack of units with different metamorphic evolutions and with metamorphic grade generally decreasing upward, which have been grouped in terms of their metamorphic and lithologic characteristics [for a classification of the Alpujarride units, see Aldaya et al., (1979), Tubía et al., (1992) and Azañón et al., (1994)]. The inferred stratigraphic sequence in all the Alpujarride units consists of (from bottom to top): graphite-rich schists and metapsammites of presumed pre-Permian age (up to 2000 m), fine-grained schists (or phyllites) of probable Permian-Triassic age (up to 600 m), and carbonate rocks (locally marbles; up to 800 m) of Middle and Late Triassic age (Kozur et al., 1974; Delgado et al., 1981). Compared with this sequence, the basement section on Site 976 resembles the lower metapelite sequence of the Alpujarride units, of presumable Palaeozoic age.

Ductile fabrics

Two systematically identifiable sets of ductile fabrics and structures can be distinguished in the high-grade schist, referred to as D₁ and D₂. The earlier fabric, S₁, is characterized by compositional layering, transposed quartz veins, 0·5-2 mm scale quartz-rich and biotite-rich laminae and oriented biotite. Garnet, staurolite, and early plagioclase porphyroblasts have grown over S₁ and include it as graphite trails.

S₁ has been intensively affected by 1-50 mm scale D₂ folds. These are commonly tight and asymmetrical but may also be symmetrical. The compositional layering and quartz-biotite laminae that define S₁ are considerably thickened in the hinges of these folds relative to the limbs. Biotite is recrystallized in the fold hinges, and tends to lie parallel to the axial surface, as do elongated mats of fibrolite, the two together defining a weak axial planar fabric, S₂. In much of the high-grade schist D₂ has been so strong that the main foliation is in fact a composite fabric formed by the transposition and modification of S₁. D₂ was followed by static growth of second-stage garnet, and subsequently by andalusite, plagioclase, and fine-grained K-feldspar intergrown with quartz.

We have observed no clear transition between the fabrics in the high-grade schist and that in the gneiss, so that the age of the gneissic foliation in relation to the deformational history of the schist is uncertain. On the basis of the textural relationships discussed below, we believe that it is a composite foliation, which may have formed initially at about the same time as D₂ in the high-grade schist, but was subsequently heavily modified by compaction or further deformation during partial melting.

Andalusite and sillimanite

Sillimanite occurs as elongated fibrolitic mats, and occasionally as larger prismatic porphyroblasts. Both textural types of sillimanite define the S₂ planar fabric. Fibrolite in the matrix forms aggregates with sub-microscopic intergrowths of K-feldspar and plagioclase, and these aggregates are commonly intergrown with biotite (Fig. 5b and f). Andalusite forms large randomly oriented porphyroblasts (up to 2-3 cm long) that grew post-kinematically with respect to S₂. It commonly contains oriented inclusions of ilmenite, biotite, and fibrolite parallel to S₂, and corroded staurolite porphyroblasts partly transformed to ilmenite and minor spinel (Fig. 5f). A remarkable textural feature of andalusite is that it commonly shows irregular outlines interfingering with plagioclase and K-feldspar, and locally occurs as rounded or irregular blebs included in these minerals (Fig. 5g). Andalusite is also locally altered to randomly oriented white mica.

The content of `impurity' cations (M = Fe<sup>3+</sup>, Ti, Cr, Mg, and Mn; XMAlSiO5 is the sum of the `impurity' cations) confined to the octahedral site in coexisting prismatic sillimanite and andalusite is significantly lower in the corundum-bearing high-grade schist than in the other metapelites and gneiss (staurolite-bearing schist: average XMAlSiO5 = 0·013; corundum-bearing schist: average XMAlSiO5 = 0·007; gneiss: average XMAISiO5 = 0·048). This probably indicates a larger P-T area of coexistence of the two polymorphs in the gneissic rocks (Kerrick & Speer, 1988), which may help explain some of the textural observations.

Garnet

We have identified two distinct families of garnet porphyroblasts (250 µm-5 mm in diameter) by considering their textural relationships with S₂ [see also Shipboard Scientific Party, (1996)]. Garnet I usually has inclusion trails of graphite that define an internal foliation (S₁) oblique to the external S₂ foliation. This internal foliation is straight in the central part of the crystals, suggesting interkinematic growth between D₁ and D₂ (Fig. 5d). The porphyroblasts have inclusions of rutile, ilmenite, biotite, and early plagioclase (near the rims), and some are partly corroded to andalusite, plagioclase, K-feldspar, and biotite (Fig. 6), or to fine-grained aggregates of chlorite, muscovite, Na-rich plagioclase, and quartz.

Figure 6. Garnet porphyroblast partially replaced by syn-D₂ fibrolite, biotite, and K-feldspar, and post-D₂ andalusite, biotite, and K-feldspar [mineral abbreviations after Kretz, (1983); and Fib, fibrolite]. In (b), the late growth of muscovite in the presence of andalusite should be noted. The line marks the location of the profile given in Fig. 7g. Back-scattered electron images. Conditions for image acquisition: (a) 1024 * 1024 pixels and (b) 512 * 512 pixels. Sample 76B (161-976B-76R-1, 3-7 cm; staurolite-bearing high-grade schist).

Many garnet I porphyroblasts have an inclusion-poor outer zone with idiomorphic outlines (garnet II). This outermost rim usually has fibrolite inclusions that are continuous with the matrix. In some grains the boundary between the two zones is irregular and shows evidence of resorption (Fig. 5d). Garnet II grains without garnet I cores also occur in the matrix as large (up to 1 cm long), idiomorphic porphyroblasts with abundant inclusions parallel to the external S₂ foliation. Garnet II includes fibrolite, biotite, plagioclase and K-feldspar (Fig. 5e), tourmaline, and locally staurolite, chlorite, and corundum.

Garnet from corundum-bearing schist is characterized by increasing contents in almandine and pyrope and decreasing contents in spessartine and Fe/(Fe + Mg) from core to rims (e.g. in Fig. 7h: Alm_0·48-0·65, Pyr_0·05-0·08, Sps_0·12-0·15). Grossular always decreases towards the rims (Grs_0·20-0·35). The two garnet types distinguished in the staurolite-bearing high-grade schist have a distinctive zoning pattern (Fig. 7a-d). Garnet I porphyroblasts have a normal growth zoning pattern indicating prograde growth: a strongly modified bell-shaped spessartine profile, and increasing almandine and pyrope, from core to rim. There is also an overall gradual decrease in Fe/(Mg + Fe) from core (average 0·97) to rim (average 0·92). The inner cores of some of the larger garnet I porphyroblasts are depleted in the spessartine and grossular components, and the Fe/(Fe + Mg) ratio, and show a coupled increase in almandine and pyrope (Figs 7b and 8). This type of inner core could be the remains of an older relict garnet developed under prograde conditions in the presence of chlorite, which preferentially partitioned Mn (e.g. Spear, 1993). Incomplete or truncated zoning patterns similar to those observed in garnet I are present in grains included in andalusite porphyroblasts (Fig. 7g). Some of the garnet I rims show evidence of resorption, with a discontinuity in the Ca, Fe, and Mg zoning patterns (Figs 8 and 9).

Figure 7. Representative compositional profiles across garnet porphyroblasts in high-grade schist (HGS) [staurolite-bearing schist, (a-e) and (g); biotite-rich schist, (f); corundum-bearing schist, (h)]; and gneiss (i, j). All profiles are from rim to rim. Garnet profile in (e) corresponds to a garnet II porphyroblast adjacent to a garnet I grain [profile in (d)] (see location in Fig. 18b). The rim of this garnet II in contact with the matrix (grey area) has a remarkable increase in grossular, spessartine, and Fe/(Fe + Mg). Two-dimensional image maps of garnet profiles (a), (b), and (f) are shown in Figs 9, 8, and 10, respectively.

Figure 8. Back-scattered electron image (a) and X-ray composition maps (b-e) of a garnet I porphyroblast from sample 76B (161-976B-76R-1, 3-7 cm; staurolite-bearing high-grade schist). The line marks the location of the profile given in Fig. 7b. Arrows point to the irregular, sharp boundary between the garnet I core and the garnet II overgrowth, which correlates also with discontinuities in composition maps. Conditions for image acquisition: 1024 * 1024 pixels, beam size 1 µm, 3 µm/pixel, Faraday cup current 25 nA, dwell time 30 ms/pixel.

Figure 9. Back-scattered electron image (a) and X-ray composition maps (b-e) of a garnet I porphyroblast from sample 28 (161-976B-76R-1, 45-49 cm; staurolite-bearing high-grade schist). The line marks the location of the profile given in Fig. 7a. Arrows point to the irregular boundary between the garnet I core and the garnet II overgrowth, which correlates also with discontinuities in composition maps. Conditions for image acquisition: 512 * 512 pixels, beam size 1 µm, 5 µm/pixel, Faraday cup current 25 nA, dwell time 100 ms/pixel.

Garnet II rims and grains show fairly constant composition (Fig. 10), similar to the rims of garnet I (Figs 7f and 8). Flat zoning patterns in this garnet may be a result of rapid growth (Holm & Selverstone, 1990), rather than diffusive homogenization of originally zoned garnets (e.g. Tracy et al., 1976; Yardley, 1977), because flat zoning is also observed in larger (>1·5 mm) porphyroblasts [intragranular diffusion is usually enhanced in garnets with radii <1 mm; Florence & Spear, (1991); and Spear, (1991)] and because, when it occurs, it affects only the outermost rims of garnet I (<100 µm, e.g. Fig. 7a). Some small garnet II porphyroblasts show sharp increases in spessartine, Fe/(Fe + Mg) ratio, and grossular at the rims (Fig. 7e), but these increases are rare in the outermost rims of large garnet II porphyroblasts. This contrasting behaviour suggests that garnet grain size has a large effect on volume diffusion processes (Florence & Spear, 1991; Spear, 1991). There is a sharp inflection of the Fe/(Fe + Mg) and grossular contents from garnet I rims to garnet II overgrowths, suggesting that these two garnets have grown by different reactions or were separated by an inflection in the P-T path (e.g. Rumble & Finnerty, 1974; Karabinos, 1984; Spear et al., 1984; Whitney & Ghent, 1993).

Figure 10. Back-scattered electron image (a) and X-ray composition maps (b-e) of a garnet II porphyroblast from sample 61 (161-976E-14R-1, 98-102 cm; biotite-rich high-grade schist). The line marks the location of the profile given in Fig. 7f. Conditions for image acquisition: 1024 * 1024 pixels, beam size 1 µm, 3 µm/pixel, Faraday cup current 25 nA, dwell time 30 ms/pixel.

Plagioclase and K-feldspar

Plagioclase appears in the matrix as large ovoid porphyroblasts (up to 5-7 mm long), showing two main growth stages. The first, plagioclase I, is represented by cores densely crowded with included graphite, defining a straight internal foliation. This texture indicates inter-kinematic growth between D₁ and D₂, but locally plagioclase I has grown over D₂ microfolds (early syn-D₂ growth). Plagioclase II forms inclusion-free rims, as well as new clear porphyroblasts associated with K-feldspar and biotite (Fig. 5b). These may contain inclusions of fibrolite parallel to S₂, indicating growth at a late stage (post-D₂). In some samples plagioclase II contains corroded relics of staurolite and andalusite.

K-feldspar occurs in several textural settings: finely intergrown within fibrolite mats (Fig. 5f); interstitially with granular quartz in veins (cutting across the S₂ foliation); as thin rims surrounding zoned plagioclase porphyroblasts; and as large irregular poikiloblasts. The last commonly contain abundant rounded inclusions of quartz and biotite, and may be intergrown with plagioclase. K-feldspar occasionally occurs as intergrowths with plagioclase within garnet II porphyroblasts (Fig. 5e).

The compositional range of plagioclase shows remarkable variations, both between rock types and within single samples, depending on textural position (Fig. 11a). Plagioclase in high-grade schist has a low to medium Ab content (staurolite-bearing schist X_Ab = 0·2-0·6; corundum-bearing schist X_Ab = 0·05-0·45). Plagioclase porphyroblasts in high-grade schist have variable non-symmetrical zoning patterns, with oscillatory (Fig. 12b and c) and normal zoning (i.e. more albite rich toward rims; Fig. 12a) as the most common types. Cores of plagioclase I, crowded with graphite and ilmenite inclusions, are characterized by low values of Fe³⁺ (Fig. 12a-d). Some plagioclase II porphyroblasts have thin rims of K-feldspar (total range X_Ab = 0·05-0·12) containing inclusions of fibrolite and biotite. The composition of plagioclase inclusions in garnet spans from intermediate Ab content (X_Ab = 0·35-0·65) in garnet II porphyroblasts, similar to the composition range of matrix porphyroblasts, to X_Ab = 0·20-0·23 in the outer rim of garnet I grains. The highest Ab contents are found in plagioclase rims and plagioclase included with K-feldspar in fibrolite mats (X_Ab >0·4).

Figure 11. Compositional variation of single analysis of plagioclase (a) and K-feldspar (b) porphyroblasts in all the population of samples analysed.

Figure 12. Representative compositional profiles across plagioclase porphyroblasts in high-grade schist (HGS) [corundum-bearing schist, (a, c, d): staurolite-bearing schist, (b)] and plagioclase phenocrysts in granite (e and f). All profiles are from rim to rim. Cores of plagioclase I, crowded with graphitic inclusions, are marked in grey, and are usually characterized by lower Fe³⁺ ratio values. [For location of profile (e), see Fig. 16b.]

K-feldspar porphyroblasts in high-grade schist are not zoned, and have low albite contents (X_Ab = 0·02-0·15) (Fig. 11b).

Oscillatory zoning in plagioclase may be the result of discontinuous growth, controlled by the local availability of reactant phases (e.g. garnet) and strain partitioning during deformation (Menard & Spear, 1996). The smooth variations in composition indicate that the zoning pattern was not controlled by any of the plagioclase compositional gaps (Smith, 1983). In the absence of any other recognizable synchronous Ca-bearing mineral phase, the normal and oscillatory zoning of plagioclase in the high-grade schist (i.e. growth of sodic plagioclase II rims after D₂) cannot be correlated with zoning of garnet I (i.e. growth of garnet I before D₂), which exhibits decreasing grossular contents from core to rim. The normal zoning in plagioclase could, however, be explained as a result of breakdown of the paragonite component in the white mica. The textural evidence suggests that white mica breakdown occurred during D₂, however, whereas the growth of sodic plagioclase rims and porphyroblasts (plagioclase II) is post D₂. This is one line of evidence suggesting that the breakdown of white mica produced a small amount of melt, which took up the sodium, and that plagioclase II grew during crystallization of this melt phase. The oscillatory zoning may therefore result from short-term variations in a(H2O) during crystallization, as is common in magmas crystallizing at low P (e.g. Scaillet et al., 1995).

Biotite and muscovite

Dark brown biotite porphyroblasts occur in very different textural positions: biotite is locally oriented parallel to S₁; it is commonly well oriented parallel to S₂, particularly in the hinge areas of D₂ folds; it occurs as randomly oriented aggregates, associated with plagioclase II and andalusite prophyroblasts; it occurs as inclusions in plagioclase II and garnet II porphyroblasts, and as reaction products of garnet I and staurolite. Biotite porphyroblasts defining S₂ are commonly closely associated or intergrown with fibrolite. In the corundum-bearing schist, large porphyroblasts of biotite with pyrite inclusions partly transformed to pyrrhotite are common. All of these textural relationships indicate that biotite was stable during almost the entire metamorphic evolution of these rocks. Mimetic alteration of biotite to chlorite has occurred at a very late stage.

Representative biotite analyses are illustrated in Fig. 13. We have rejected all the chemical analyses in high-grade schist with total oxide amounts <94·5 wt %, and selected those analyses with a sum of [xii] cations >1·6 ions p.f.u. This is to avoid possible chlorite intergrowths at small scale, mass deficiencies, and analytical uncertainties in the determination of alkali elements in micas (García-Casco et al., 1993).

Figure 13. Biotite composition per sample, expressed in terms of Al^[vi], Fe/(Fe + Mg), and Ti contents. Mica analyses were normalized to 20 oxygens, 4 OH^-, and Fe^total= Fe²⁺. Chemical analyses with total oxide amounts <94·5 wt % in the high-grade schist and <93·5 wt % in the gneiss and granite, and with a sum of [xii] cations <1·6 ions p.f.u., have been rejected for this study.

Biotite in the high-grade schist is in general Fe rich and exceptionally aluminous (Al^[vi] ranges between 0·75 and 1·25 ions p.f.u.). Matrix biotite is homogeneous in individual samples, but varies with rock type, suggesting that the main control is bulk-rock composition. In staurolite-bearing schist, for example, Fe/(Fe + Mg) is ~0·8 (range 0·67-0·85) and Ti varies from 0·3 to 0·5 ions p.f.u., whereas in corundum-bearing schist Fe/(Fe + Mg) is ~0·7 (range 0·5-0·8) and Ti ~0·2 ions p.f.u (Fig. 13). Biotite included in garnet porphyroblasts (important for thermobarometric calculations) shows similar Fe/(Fe + Mg) and Al^[vi] to matrix biotite, but shows bell-shaped and linear or crescent-shaped patterns of zoning in Fe, Al^[vi], and Ti (Fig. 14). The bell-shaped zoning pattern suggests diffusional interchange with the garnet host, and both zoning patterns contradict the common assumption of infinitely fast diffusion in biotite, which produces rapid homogenization of the crystals in this context (e.g. Spear & Florence, 1992; Spear & Parrish, 1996), and might suggest a rapid cooling history for these rocks.

Figure 14. Compositional profiles of biotite included in a large garnet II porphyroblast (for location of these profiles, see Fig. 18b). Sample 28 (161-976B-76R-1, 45-49 cm). Both profiles illustrate the two extreme zoning patterns observed in biotite in this context: (a) a bell-shaped zoning pattern; (b) a linear, crescent-shaped zoning pattern.

Muscovite in the high-grade schist is always retrogressive (e.g. replacing andalusite), and has low Tiand Si contents (usually <0·05 and 3·05-3·3 ions p.f.u, respectively), consistent with crystallization at relatively low P and T (Guidotti, 1984; Massone & Schreyer, 1987; Guidotti et al., 1988).

Staurolite

This mineral is present in two different textural positionsthat can coexist in the same rock sample. Small staurolite crystals (<200 µm) are preserved as corroded inclusions in garnet II, andalusite, and plagioclase II porphyroblasts. Inside andalusite, staurolite is locally transformed into an Fe-rich spinel (hercynite) (Fig. 5f). In some samples, it occurs as unaltered crystals in the matrix with an internal foliation defined by straight inclusions of graphite and ilmenite (Fig. 5c). Both textural positions suggest that staurolite grew inter-kinematically between D₁ and D₂.

Staurolite shows little chemical variation in Fe (Fig. 15a), but unaltered porphyroblasts in the matrix are zoned with increasing Fe/(Fe + Mg) toward the rims (0·925-0·946) (Fig. 16b). Mg and Zn contents are always very low (between 0·2-0·4 and 0·05-0·17 ions p.f.u., respectively). Relict inclusions of staurolite in garnet II and andalusite have lower Fe/(Fe + Mg) than matrix porphyroblasts (average 0·93 and 0·88, respectively), and may be derived from the cores of the original grains.

Figure 15. (a) Selected scatter bivariant diagram of staurolite composition, showing the best linear regression that fits the compositional trends. Open symbols represent non-transformed staurolite porphyroblasts in the matrix and filled symbols staurolite relics in andalusite porphyroblasts. (b) Core-to-rim compositional profile of a relict staurolite (sample 22; 161-976B-74X-1, 30-34 cm). Staurolite analyses were normalized to 46·5 oxygens (= 3·0 H⁺ ions p.f.u.), and assuming that Li = 0·2 ions p.f.u. and Fe³⁺ = 3·5% of total Fe, following the normalization procedures of Holdaway et al., (1986, , 1991) and Hawthorne et al., (1993).

Figure 16. Selected textural relationships in the gneiss and granite rocks [mineral abbreviations after Kretz, (1983); and Fib, fibrolite]. (a) Principal foliation defined by intensely oriented harmonious fibrolite, with augen magnetite and porphyroblasts of quartz and K-feldspar. Sample 51 (161-976B-95R-2, 24-28 cm; pelitic domain, migmatite gneiss). Photomicrograph, plane-polarized light. (b) Zoned plagioclase phenocryst surrounded by K-feldspar. Dark areas in the K-feldspar are small patches of Ab-rich plagioclase. Both feldspars are replaced by a fine intergrowth of white mica and sericite, with needle-like quartz (in black). The line marks the location of the profile given in Fig. 12e. Back-scattered electron image. Conditions for image acquisition: 512 * 512 pixels. Sample 43 (161-976B-82R-1, 93-98 cm; granite). (c) Andalusite with inclusions of fibrolite and biotite parallel to the external foliation. The andalusite porphyroblast shows apparently resorbed outlines with respect to surrounding plagioclase. Sample 161-976B-93R-1, 52-57 cm (pelitic gneiss). Photomicrograph, plane-polarized light. (d) Andalusite with augen texture. Opaque oxide inclusions define an included fabric at high angle to external foliation defined by biotite and fibrolite. Sample 161-976B-93R-1, 52-57 cm (pelitic gneiss). Photomicrograph, plane-polarized light. (e) Andalusite with inclusion trails of opaque oxide defining a folded internal fabric. External foliation is defined by oriented aggregates of fibrolite, clearly oblique to the internal fabric inside the andalusite. Sample 161-976B-96R-1, 30-34 cm (pelitic gneiss). Photomicrograph, plane-polarized light. (f) Idiomorphic andalusite phenocrysts included in a K-feldspar + quartz matrix. Sample 161-976B-96R-1, 46-48 cm (pelitic gneiss). Photomicrograph, plane-polarized light.

Corundum

Corundum forms elliptical porphyroblasts up to several millimetres in diameter in quartz-poor assemblages. The crystals have inclusion trails that are generally concordant with S₂, but showing some evidence of continued flattening of S₂ around the crystals. They are probably synkinematic with S₂. In some rocks there is a distinct halo of K-feldspar around the corundum, suggesting that it formed at the expense of muscovite (see also Shipboard Scientific Party, 1996).

Ti-Fe oxides

Ilmenite is the most abundant Ti-Fe oxide in all of the different types of high-grade schist. It is present as a matrix component parallel to S₂, or as inclusions in garnets I and II, plagioclase, andalusite, corundum, staurolite, and K-feldspar porphyroblasts. Inside andalusite, worm-like crystals of ilmenite with a radial pattern are common. Rutile is only present as needle-shaped crystals preserved in large ilmenite crystals, suggesting that rutile was transformed into ilmenite before D₂. Another Fe-rich oxide present as a minor component is spinel (hercynite), which has significant contents in Mg and Mn, and negligible Zn contents (average composition Fe_88·5 Mg_10·0 Mn_1·4 Zn_0·1). Hercynite commonly occurs as small grains in poikiloblastic andalusite surrounding relict inclusions of staurolite (Fig. 5f).

Reaction history

In the high-grade schist, we can identify an early (post-S₁, pre-S₂) mineral assemblage that includes garnet I + staurolite + biotite + plagioclase + Ti-Fe oxide + quartz (Assemblage 1). Given the bulk composition of the schist, muscovite was almost certainly present, but is not now preserved. In the KFMASH system Assemblage 1 corresponds to a divariant assemblage Grt + St + Bt + Ms [Als] (3a), and it is therefore unlikely that an Al-silicate was present in equilibrium with this assemblage (Fig. 4a).

The normal zoning pattern of garnet I suggests prograde metamorphism with rising temperature and possibly decreasing pressure, but the timing of this garnet growth is not clear from textural evidence, although thermobarometric calculations (see below) suggest that their rims grew near the Ky-Sil transition. These observations, together with the decrease in the Fe/(Fe + Mg) ratio in garnet I towards the rim, suggest that the growth of this mineral may have occurred by the divariant reaction [Als] (3a) in the Ky stability field (see Fig. 3c). Inner garnet cores characterized by a decreasing Fe/(Fe + Mg) ratio might be relics of previous garnet which may have grown in greenschist facies conditions from chlorite or chloritoid. Subsequent resorption of garnet I, before the growth of garnet II rims, may have occurred either as a result of the prograde univariant reaction

Grt + Chl + Ms = St + Bt + Qtz + H2O

(14)

or during decompression by the net transfer reaction [Als] (3a).

During D₂, the stable assemblage in the high-grade schist was probably biotite + sillimanite + K-feldspar + plagioclase + ilmenite + quartz ± garnet (Assemblage 2). Staurolite became unstable when the conditions surpassed the univariant reaction (4). Muscovite breakdown occurred probably as a result of the dehydration melting reaction (6a) or its equivalent:

Ms + Pl + Qtz = Sil + Kfs + L

(6b)

(Thompson & Algor, 1977; Thompson & Tracy, 1979; Thompson, 1982). Reaction [Als] (3a), which probably occurred before (5b), also contributed to the consumption of muscovite. Textural evidence that suggests the presence of minor amounts of melts formed by this reaction are the occurrence of K-feldspar + plagioclase + quartz intergrowths included in the new garnet II porphyroblasts, and also K-feldspar on rims of plagioclase, and interstitial K-feldspar together with quartz in veins. Additionally, the apparent simultaneous crystallization of K-feldspar, plagioclase, and andalusite, to form Assemblage 3, is interpreted as resulting from the later crystallization of interstitial melts. The development of the assemblage corundum + biotite + K-feldspar + plagioclase in quartz-depleted rocks (corundum-bearing schist), represents the reaction

Ms = Crn + Kfs + H2O

(15)

(Huang et al., 1973; Huang & Wyllie, 1974).

Therefore, Assemblage 2 results from the change of the divariant assemblage Grt + Bt + Sil + Ms to form Grt + Bt + Sil + Kfs. Taken altogether, this sequence of assemblages indicates decompression, accompanied by constant or rising T.

Textural evidence suggests that garnet II grew after D₂, and therefore at lower P and/or higher T than the muscovite breakdown reaction. Nevertheless, barometric studies on inclusion suites inside these garnet porphyroblasts (see below) suggest that garnet grew under decreasing-P conditions (from the Ky-Sil transition to P ~3-4 kbar). Garnet growth at this stage probably involved staurolite breakdown (4), as suggested by the fact that some garnet II grains include corroded staurolite relics. Another garnet-forming reaction that could have been involved is the Crd-forming reaction (7), which is ruled out because of the absence of cordierite in the high-grade schist. The range of P estimations on this garnet, together with the flat-zoning pattern in the Fe/(Fe + Mg) ratio, can only be explained by a rising-T decompression P-T path (Fig. 3c). Along this path limited growth of garnet could have occurred by reaction (5b), but given the abundant evidence of disequilibrium in these rocks (in particular, the preservation of staurolite in the matrix), we suggest that garnet II grew also by overstepping of reaction (4) under low-P conditions.

The final assemblage we can establish in these rocks is biotite + andalusite + K-feldspar + plagioclase + ilmenite + quartz (Assemblage 3). The common occurrence of staurolite relics included in andalusite suggests that the breakdown of staurolite was considerably overstepped, and probably contributed to substantial growth of an Fe-rich biotite during this final stage. Retrogressive breakdown of garnet occurred during the cooling path by the divariant reaction [St] (5b), as is revealed by the occurrence of andalusite including garnet, K-feldspar, and fibrolite (Fig. 6).

Quantitative P-T determinations

We have applied several geothermometers and geobarometers to help constrain the P-T conditions achieved by the basement rocks. Table 2 shows selected P-T estimates for representative analyses per sample and in different textural positions, and illustrates the diversity of results obtained within a single thermometer or barometer depending on the type of calibration. The absence of suitable parageneses for thermobarometry in the gneiss together with the spurious and scattered results obtained with several geothermometers and geobarometers (e.g. garnet + cordierite geobarometer) give us additional data to suggest that many of the mineral assemblages in the gneissic rocks are not in equilibrium.

Table 2. Representative garnet, biotite, and plagioclase analyses with thermobarometric calculations

Among the different calibrations of the garnet + biotite (GARB) thermometer, the calibrations which include corrections for the biotite composition (in particular, for the Ti and Al^[vi] contents; e.g. Indares & Martignole, 1985; Patiño Douce et al., 1993; Holdaway et al., 1997) and for the non-ideality in both the garnet and biotite (e.g. Bhattacharya et al., 1992; Kleemann & Reinhardt, 1994; Holdaway et al., 1997) seem to be the most reasonable calibrations for the studied rocks, on the basis of biotite and garnet compositions (i.e. biotite with a significant Ti content and garnet relatively rich in Ca and Mn). The results obtained with the calibration of Holdaway et al., (1997) have a smaller range in temperature estimates in the high-grade schist (620-780°C), although scattered results are always obtained in the gneissic rocks (600-950°C). In general, the calibration of Holdaway et al., (1997) yields temperature estimates close to those obtained with the calibration of Kleeman & Reinhardt, (1994) (maximum differences of +15-20°C) and Perchuk & Lavrent'eva, (1983) (maximum differences of -20-40°C), whereas the calibration of Patiño Douce et al., (1993) always gives extremely high temperatures, with deviations of up to 175-180°C. Other calibrations of the GARB [such as those of Ferry & Spear, (1978) using the garnet-activity model of Berman, (1990), Hodges & Spear, (1982), and Indares & Martignole, (1985) with the garnet-activity model of Ganguly & Saxena, (1984)], produce scattered T estimates with deviations of up to 80-120°C over the values obtained with the calibration of Holdaway et al., (1997).

We have applied the GARB thermometer to garnet-biotite grains in contact, either in the matrix or as inclusions in large garnet II porphyroblasts (see Figs 18 and 19a). The temperatures obtained vary from 620 to 770°C [henceforth values with the calibration of Holdaway et al., (1997); always for an estimated pressure of 4 kbar]. The calculated temperatures obtained from single garnet II porphyroblasts crowded with biotite inclusions vary from 637 to 693°C (662 ± 14·5°C).

(a)

(b)

Figure 18. (a) Photomicrograph showing a large, elongated garnet II porphyroblast with inclusion trails of graphite, ilmenite, biotite, plagioclase, fibrolite and chlorite, parallel to the external foliation S₂. To the left of this grain there is a garnet porphyroblast showing two stages of growth, where a core crowded with opaque minerals and graphite (garnet I) is surrounded by an inclusion-poor overgrowth (garnet II). (b) Sketch of garnet showing values of Fe/(Fe + Mg) in single spot of garnet and biotite, and calculated garnet + biotite (GARB) temperatures using the indicated mineral pairs and the calibration of Holdaway et al., (1997) (for an estimated pressure of 4 kbar). Also shown are the mole fractions of grossular in single spots of garnet, anorthite in single analyses of plagioclase, and the calculated garnet + Al-silicate + quartz + plagioclase (GASP) pressures using the indicated mineral pairs and the calibration of Hodges & Crowley, (1985) (for an estimated temperature of 675°C). Garnet and biotite compositional profiles in Fig. 7 and Fig. 14, respectively, are indicated by thick white lines. Sample 28 (161-976B-76R-1, 45-49 cm; staurolite-bearing high-grade schist).

Figure 19. Thermobarometric constraints on the P-T evolution of the Alboran basement. (a) Thermobarometric estimates in the high-grade schist (samples used for the plot: 22, 28, 32, 61, and 76B; see Table 2) with GARB thermometer (Holdaway et al., 1997; assuming that garnet contains 3% Fe³⁺ and biotite has 11·6% Fe³⁺), and GASP (Hodges & Crowley, 1985), GPBQ (Hoisch, 1990, , 1991), and GRAIL (Bohlen et al., 1983) barometers. Key reactions from the KFMASH grid of Spear & Cheney, (1989) are also shown for reference. Muscovite-out reactions [(6a) and (6b)] are from Huang et al., (1973), Huang & Wyllie, (1974), Thompson & Algor, (1977), and Thompson, (1982). The lightly shaded box represents the average temperatures and pressures (±1[sgr]) for the rims in contact of matrix porphyroblasts of garnet, biotite, and plagioclase. [Note the large dispersion (up to 775°C at 4 kbar) of the temperature estimates for the biotite porphyroblasts included in garnet II grains.] The proposed prograde and decompression P-T path for the high-grade schist have also been plotted for comparison. Dashed portion of the path shows approximate P-T conditions along the prograde trajectory (growth of garnet I). (b) Thermobarometric estimates in the high-grade schist (dotted lines) and granite + gneissic rocks (continuous lines) on plagioclase + K-feldspar intergrowths (selected results from samples: 22, 28, 43, 76, and SOT; see Table 3), performed with the ternary-feldspar calibration of the two-feldspar thermobarometer [calibration of Elkins & Grove, (1990)]. These triangular areas represent the P-T conditions achieved by these rocks during the cooling path. The lightly shaded area corresponds to one of the most accurate P-T estimates (sample SOT) obtained with this thermobarometer (centroid of this area is placed at 500°C and 3·2 kbar). Data sources for the different granite solidus have been summarized by Johannes & Holtz, (1996) [H₂O-saturated granite solidus is from Le Breton & Thompson, (1988); Al₂O₃-saturated granite solidus is from Holtz et al., (1992) and Joyce & Voigt, (1994)]. Dehydration melting reactions [(6a) and (6b)] are from Vielzeuf & Clemens, (1992). Long-dashed lines in the And-stability field represent the effects of the addition of F (4 wt %) and B₂O₃ (4 wt %) on the granite solidus [[solid square], experimentally derived T conditions (at 1 kbar) for B₂O₃-bearing (Pichavant, 1981, , 1987) and F-bearing (Manning, 1981) granite solidus]. The proposed cooling P-T path for the Alboran basement has also been plotted for comparison. Numbers in parentheses refer to reactions described in the text. Al-silicate triple point from Berman, (1988).

The wide span in the calculated temperature within single samples (some incompatible with the stability field of the observed mineral assemblage) gives further evidence of disequilibrium in these rocks. Additionally, the temperatures calculated with GARB should be used with circumspection because of the following contributing factors:

   (1) matrix biotites display variable Fe/(Fe + Mg) contents depending the textural position and reaction site.

   (2) Biotite in the matrix has probably undergone a late-stage homogenization towards more Fe-rich compositions (resulting in an overestimate of the temperature with GARB), by the breakdown of garnet at lower P to form andalusite + biotite by the net transfer reaction [St] (5b) (Fig. 3a and c).

   (3) Some of the biotite inclusions in garnet II porphyroblasts have also probably undergone this late-stage enrichment in Fe by diffusion along fractures in garnet (Whitney, 1996), which are particularly abundant in garnet II porphyroblasts (see Figs 8 and 18). The occurrence of Fe-rich chlorite inclusions in some of these porphyroblasts (Fig. 18) provides further evidence for retrogression, probably induced by diffusion along these fractures.

   (4) Dissolution of garnet exposed different rim compositions [in general with higher Fe/(Fe + Mg) ratios than the absolute rim] not in equilibrium with matrix biotites. This could explain some of the lower temperatures obtained with rims of garnet I porphyroblasts (Figs 18b and 19a).

All of these factors are related to the widespread disequilibrium relationships between garnet and biotite in these high-grade rocks, and suggest caution in the use of this geothermometer to estimate peak-temperature conditions (e.g. Ghent et al., 1982; Selverstone & Chamberlain, 1990; Spear & Florence, 1992; Spear & Parrish, 1996).

Application of the garnet + sillimanite + quartz + plagioclase (GASP) barometer (calibrations of Newton & Haselton, 1981; Hodges & Spear, 1982; Ganguly & Saxena, 1984; Hodges & Crowley, 1985; Koziol & Newton, 1988; Koziol, 1989) to plagioclase inclusions in garnet from the staurolite-bearing high-grade schist consistently gives pressures between 6 and 7 kbar in garnet I rims [X_Ab = 0·20-0·25 and assuming kyanite as the coexistingAl-silicate; henceforth quoted pressure estimates are those obtained with the calibration of Hodges & Crowley, (1985)], close to the kyanite-sillimanite transition, for an assumed temperature of 600°C (Table 2 and Fig. 19a). These estimates with GASP should be considered as maximum pressures because of the previously reported absence of an Al-silicate in equilibrium with garnet I (Assemblage 1). Using rim compositions for garnet I in contact with rims of matrix plagioclase (X_Ab = 0·45-0·60), and assuming that both minerals are in equilibrium with sillimanite, gives a wide range of lower pressures, 4-6·7 kbar at 675°C (Table 2, Fig. 19a). Plagioclase inclusions in garnet II porphyroblasts, which tends to be more Ab rich towards the garnet rim (plagioclase inclusions in the garnet II core: X_Ab = 0·33-0·45; inclusions near the garnet II rim: X_Ab = 0·48-0·38) also record decreasing pressures during their growth, from 7·3-6·2 kbar at the garnet core to 3·5-3 kbar at the garnet rim (at 675°C and coexisting with sillimanite; Fig. 18b). In view of the strong zonation of plagioclase in the matrix, and the evidence of late-stage re-equilibration of feldspars, the GASP results with rims of porphyroblasts should be treated with caution.

Additional P estimations have been obtained by using the garnet + plagioclase + biotite + quartz (GPBQ) barometer calibrated by Hoisch, (1990, , 1991), which gives pressures consistent with GASP in the range 8·6-7·6 kbar (for an assumed T of 675°C) for plagioclase included in garnet II porphyroblasts, and 5·3-4·3 kbar for rims in contact with matrix plagioclase (Table 2, Fig. 19a).

Pressure estimates for garnet I porphyroblasts have been performed by applying the garnet + rutile + Al-silicate + ilmenite + quartz (GRAIL) barometer (Bohlen et al., 1983) to rutile-ilmenite intergrowths occasionally preserved in these garnet porphyroblasts. Pressures calculated, assuming pure ilmenite, range from 10·5 to 8 kbar (at 500 and 600°C) for inclusions in garnet I core and rim, respectively (Fig. 19a). Considering the above-mentioned presumable absence of an Al-silicate in equilibrium with garnet I, the GRAIL estimates can be considered as tentative minimum pressures (see Bohlen et al., 1983, p. 1055).

Keeping in mind the need for caution in interpreting the thermobarometric results, we suggest that the results of the GASP and GBPQ barometers indicate a significant pressure drop during D₂, from 6-7 kbar to 3-4 kbar. Approximate temperature conditions at the end of D₂ are likely to have been between 650-700°C in the metapelites and 700-750°C in the gneiss. Approximate prograde P-T conditions in these rocks can only be estimated using the GRAIL barometer estimates, indicating a pressure of 10·5-8 kbar (for a presumed T of 500 and 600°C) for core and rim of garnet I porphyroblasts, respectively (see Fig. 19a).

P-T conditions along the cooling path

The decompression P-T path was followed by a cooling path (P < 3 kbar) into the andalusite stability field (T ~600°C). During this portion of the path, rapid crystallization of residual melts and a large overstepping of reactions occurred, allowing complex textural relationships involving andalusite to form. In the metapelites these include relict staurolite preserved inside large andalusite porphyroblasts, and intergrowths of andalusite and either plagioclase or K-feldspar. In the gneiss, andalusite is found either included in or intergrown with cordierite, plagioclase, and K-feldspar.

To constrain the P-T conditions experienced by the rocks during the cooling history, we have used the two-feldspar geothermometer, based on the common interpretation that this thermometer could yield estimates of solidus temperatures for granitic systems (Anderson, 1996). We have used the binary-feldspar calibrations of Stormer, (1975), Powell & Powell, (1977), Whitney & Stormer, (1977), and Haselton et al., (1983), together with the ternary-feldspar calibrations of Ghiorso, (1984), Green & Usdansky, (1986), Fuhrman & Lindsley, (1988), and Elkins & Grove, (1990). Mole fraction information of coexisting feldspars and the solutions to the different calibrations are presented in Table 3. There is a large variation in the calculated temperatures between the different calibrations, with maximum differences up to 140°C (values always for an estimated P of 3 kbar), although the majority of the temperatures are between 450 and 550°C. By comparing the results obtained with a specific calibration (e.g. Fuhrman & Lindsley, 1988), the calculated temperatures tend to fall from the granite and gneissic rocks (570-470°C) to the metapelites (530-300°C), and within single rock types with the sequence of crystallization deduced by textural relationships. For example, two-feldspar pairs observed as inclusions in garnet II porphyroblasts have higher temperatures of crystallization than those formed in the outermost rims of zoned plagioclase porphyroblasts in the matrix (420°C and 300°C, respectively). Or, in the gneissic rocks, temperatures obtained in feldspars from the pelitic domains are systematically higher than in leucosome domains (550°C and 480-500°C, respectively).

Table 3. Selected plagioclase and K-feldspar analyses with two-feldspar thermometric calculations

The calibrations of the ternary-feldspar thermometer could be also used to estimate the pressure conditions and the degree of equilibrium during crystallization (e.g. Green & Usdansky, 1986; Fuhrman & Lindsley, 1988). The graphical solutions of the three end-member calibrations (T_Ab, T_An, and T_Or) of the ternary-feldspar system have been represented in a P-T diagram (Fig. 19b) for the most representative mineral pair per sample. The intersections define a broad P-T area placed at 450-550°C and 2-4 kbar, which represents the P-T conditions during the subsolidus recrystallization of the feldspars. These P-T conditions are clearly below any of the granite solidi, either the H₂O-saturated (Le Breton & Thompson, 1988) or the Al₂O₃-saturated granite solidus (Joyce & Voigt, 1994; Johannes & Holtz, 1996) (see Fig. 19b). The absence of a unique intersection between the three end-member solutions in all the selected pairs represented in Fig. 19b suggests also the existence of a variable degree of disequilibrium among the two feldspars in these rocks.

Modelling of the P-T path in the KNaCaFMASH system

Table 4 lists the mineral compositions for Assemblage 2 biotite + sillimanite + K-feldspar + plagioclase ± garnet II + quartz, which is assumed to represent equilibrium conditions immediately after D₂ in the high-grade schist. From the thermobarometric studies we have assigned P-T conditions of 675°C and 4 kbar. To overcome the observed disequilibrium relationships, we have back-calculated the modal abundance of the phases under these P-T conditions by using bulk-rock chemistry data. The results are also included in the Table 4, and constitute the starting P-T-X-M point for the modelling. We have computed in the KNaCaFMASH system the isopleths of mineral composition and mineral abundance(expressed as molar proportions of the minerals, M) for the following trivariant fields (from lower- to higher-P conditions):

Crd-Kfs-Grt-Bt-Pl-Qtz-H2O:
Grt + Kfs + H2O + [Als] Bt + Crd + Pl + Qtz	(8b)
Als-Kfs-Grt-Bt-Pl-Qtz-H2O:
Grt + Kfs + H2O + [St] Bt + Als + Pl + Qtz	(5c)
Ms-Als-Grt-Bt-Pl-Qtz:
Grt + Ms [St, H2O] Bt + Als + Pl + Qtz	(5d)
and St-Ms-Grt-Bt-Pl-Qtz-H2O:
Grt + Ms + H2O + [Als] St + Bt + Pl + Qtz	(3b)

which correspond, in the KFMASH system, to the divariant fields [Als] (8a), [St] (5b), [St, H₂O] (5a), and [Als] (3a), respectively. The boundary reactions of these trivariant fields are now divariant in the new system (for the calculation details, see Table 4), and correspond to the previously reported univariant reactions: Crd-in (7), Ms-out (6b), and St-out (4). As we reported for the KFMASH system, all the calculations have been done with the Gibbs method, using the programs provided by Spear et al., (1991). Because the assemblage at the starting point of the modelling is not in full equilibrium, however, the detailed results of the modelling can only be used to give a qualitative indication of the P-T path.

Table 4. P-T-X-M values for the peak metamorphic assemblage in the high-grade schist of the Alboran basement, and P-T-X-M values calculated with the Gibbs method in the KNaCaFMASH system

The diagrams in Fig. 20 show the P-T evolution which best reconciles the reaction history in the high-grade schist with the P-T-X-M phase relations in the model KFNaCaMASH system. The depicted P-T evolution has also been constructed following the more reliable results of the thermobarometric calculations. This path has three domains well constrained by textural, chemical, and thermobarometric analyses:

   (1) Prograde growth of garnet I porphyroblasts from the chlorite-in to the staurolite-in stability fields. This portion of the P-T path occurs from 10-10·5 kbar to 6-8 kbar, near the Ky-Sil transition, as is revealed by GRAIL and GASP results. Along this path, T probably rose from 500°C to 600°C to account for the observed garnet I zoning patterns.

   (2) Growth of garnet II porphyroblasts along a path with decreasing P and rising T. This path occurs, as is documented by GARB, GASP, and GPBQ results, from the lower-P portion of the staurolite-in stability field (6-7 kbar) to low-P conditions (4 kbar at 650-700°C), once the muscovite dehydration reactions [(6a), (6b), and (15)] have been surpassed.

   (3) In the gneiss, we have evidence of higher-T conditions during the decompression (T > 700°C) which resulted in a long-lived history of melting reactions, the possible participation of some biotite-breakdown reactions [e.g. (11) and (12)], and the crossing of the cordierite-forming reaction (7). From the present study, we can rule out the occurrence of this reaction in the high-grade schist.

   (4) The decompression was followed by near-isobaric cooling (P < 3 kbar), producing rapid crystallization of residual melts in the andalusite stability field (T < 600°C), and a late growth of muscovite in coexistence with andalusite.

The prograde portion of the P-T path is shown tentatively as it is only supported by the zoning pattern and GRAIL pressure estimates on garnet I. Growth of inner zones of garnet I could have occurred from chlorite andbiotite on a trajectory from 500°C, 10 kbar to 560°C,8·5 kbar. These conditions satisfy GRAIL data and yieldlarge amounts of garnet zoned down to an Fe/(Fe + Mg)ratio of ~0·92. After the intersection with the staurolite-in reaction (14) and within the staurolite stability field (i.e. from 560°C, 8 kbar to 600-620°C, 6-7 kbar), garnet could have been stable without significant resorption (Fig. 20h) in the presence of biotite, staurolite, and plagioclase progressively enriched in Ca (Fig. 20b). In conclusion, by correlating GRAIL pressure estimates with the compositional zoning observed in garnet I porphyroblasts, we suggest that this garnet grew from at least 500°C at 10-10·5 kbar to 600°C at 6-7 kbar.

Figure 20. (a-h) P-T diagrams showing composition (expressed as mole fractions) and abundance (expressed as molar proportions of the phases) isopleth contours corresponding to the assemblages St + Ms + Grt + Bt + Pl + Qtz + H₂O [Als] (3b), Ms + Als + Grt + Bt + Pl + Qtz [St, H₂O] (5d), Kfs + Als + Grt + Bt + Pl + Qtz + H₂O [St] (5c), and Crd + Kfs + Grt + Bt + Pl + Qtz + H₂O [Als] (8b), in the KNaCaFMASH system. The isopleths for the Crd-bearing assemblage have been omitted in most of the diagrams [with the exception of (e) and (g)] for simplicity and because of the lack of cordierite in the high-grade schist. Contours corresponding to the assemblage Chl + Ms + Grt + Bt + Pl + H₂O are shown only for reference in (d) and (h) for the molar abundance and Fe/(Fe + Mg) composition of garnet. (i) P-T diagram showing the isopleths of M_Grt (expressed as wt % of garnet) for the reactions (4) St + Ms + Qtz + Pl = Grt + Bt + Als + H₂O and (7) Bt + Als + Qtz + Pl = Grt + Crd + Kfs + H₂O. The proposed P-T path for the high-grade schist has also been plotted in the diagrams for discussion of predicted P-T-X-M phase relations (see text). Dashed portion of the path shows approximate P-T conditions along the prograde trajectory of the high-grade schist based on zoning patterns of garnet I porphyroblasts. Starting P-T-X-M conditions at 675°C and 4 kbar are shown by a white circle. Calculated intersections with divariant reaction (7), (6b), and (4) are marked with white squares, and their P-T-X-M conditions are listed in Table 4. Approximate composition of chlorite at the intersection with the divariant staurolite-in reaction (14) has been calculated from the invariant point [Crd, Als, Tlc] of Spear & Cheney, (1989) [at 589°C and 9·5 kbar, X_Fe in chlorite 0·337, chlorite abundance 13·5 wt %]. Calculations have been done using the Gibbs method of Spear et al., (1982) and Spear, (1988, , 1993), and the thermodynamic database of Berman, (1988) modified by Spear & Cheney, (1989). Numbers in parentheses refer to reactions described in the text. Al-silicate triple point from Berman, (1988).

The decompression path occurred under rising-T conditions from 6-7 kbar at ~600°C to 3-4 kbar at 650-700°C. This portion of the path is supported by the following phase relations and data:

   (1) Along this path, garnet II grows essentially by the breakdown of staurolite by (4). This univariant reaction could produce a large amount of garnet within a small P drop (Fig. 20i), and therefore it results in a quick consumption of staurolite and a rapid growth of garnet with flat zoning-patterns, as is observed in many of the garnet II porphyroblasts. In the light of these results, the preservation under low-P conditions (P ~ 3 kbar) of staurolite in the matrix and within andalusite porphyroblasts gives further evidence for a rapid decompression and cooling P-T path, with partial consumption of the phases by overstepped reactions [in this case by [Grt] (1) and (4)].

   (2) Garnet II rims are equilibrated at low-P conditions (P = 3-4 kbar), once the muscovite-breakdown reaction (6b) has been surpassed, as is documented by GASP results and the preservation of plagioclase + K-feldspar inclusions. The amount of melt formed by this reaction was probably low (<15%) and proportional to the amount of muscovite in the rock (Clemens & Vielzeuf, 1987). The melting of muscovite will result in a Ca enrichment of garnet (Spear & Kohn, 1996) when it grows by the reaction [St] (5b). Constant Fe/(Fe + Mg) ratio from garnet I rims to garnet II porphyroblasts requires rising temperatures along a decompression P-T path (Figs 3c and 20d).

   (3) All along the decompression path, biotite grew in large amounts (Fig. 20f), becoming progressively more rich in Fe during the decompression. At P > 6 kbar and during the decompression path, staurolite becomes progressively richer in Fe (Fig. 20e). Both predictions are supported by our observations, including the type of zoning in staurolite (e.g. Fig. 15b) and the mineral chemistry and high abundance of biotite.

   (4) The predictions by the model for the plagioclase composition (Fig. 20c) are not confirmed by our observations. The main reason for this discrepancy comes probably from the muscovite-breakdown reaction (6b), which generated melts relatively enriched in Na. The preferential partitioning of Na in the melts will result in a relative enrichment of Ca in the growing plagioclase, and a final growth of Ab-rich outer rims, similar to zoning patterns we have described (Fig. 12a and b). Consequently, the Ab-rich composition of plagioclase considered at the starting point (Table 4) generates An isopleths useless for the consideration of phase relations.

   (5) With respect to phase relations under the lower-P conditions of the P-T evolution of the gneissic rocks, we can support with the modelling a large growth of cordierite with a fairly constant X_Fe composition (0·30-0·37) (Fig. 20e and g). It is important to note that contours of the molar abundance of cordierite allow for an increasing and continuous growth of this mineral along the cooling path in the gneiss (from T > 700°C and P ~ 3 kbar; Fig. 20g). If we explore the phase relations of the cordierite-forming divariant reaction [equivalent to (7)] we obtain a marked positive slope of the isopleths of moles of garnet, which diverge with temperature (Fig. 20i). Therefore, we can deduce qualitatively that a large amount of garnet could be produced under low-T conditions (650-750°C) if this reaction is crossed (the same conclusions could be depicted for moles of cordierite). Consequently, some amount of garnet could be produced in the gneiss by this reaction, and we reinforce the assumption of relatively higher T during the decompression in the gneiss (T > 700°C) because reaction (7) results probably in the production of Grt + Crd under higher P than the reaction depicted by Spear & Cheney, (1989). In conclusion, the main assemblage in the gneissic rocks requires T > 700°C at P > 4 kbar. An upper temperature limit of 750-800°C is indicated by the absence of orthopyroxene (Fig. 21). Hence we suggest that the melting event in the gneissic rocks occurred probably in the range 700-750°C and 4-6 kbar (Soto et al., 1999).

Figure 21. P-T path for the high-grade schist of the Alboran basement, and approximate melting conditions in the gneiss (stippled area). The displayed cooling P-T path is constrained by the thermobarometric data and mineral reactions shown in Fig. 19b. Stability fields for the different assemblages discussed in the text are shown in shaded areas. These P-T fields [Grt I core, Grt I rim (Assemblage 1), Grt II which is equivalent to the D₂ deformation event (Assemblage 2), and And + Ms (Assemblage 3)] are constrained by thermobarometric estimates, phase relations consideration, and textural observations (see text for discussion and uncertainties). Selected P-T conditions for the `peak' metamorphic assemblage (Bt + Sil + Kfs + Pl + Grt + Qtz, at the end of D₂) in the high-grade schist are shown with a large grey circle. P-T grid for pelites is composed from Spear & Cheney, (1989) KFMASH grid; dehydration melting reactions from Huang et al., (1973), Huang & Wyllie, (1974), Vielzeuf & Clemens, (1992), and Aranovich & Newton, (1998); granite solidi from Johannes & Holtz, (1996). Numbers in parentheses refer to reactions described in the text. Al-silicate triple point from Berman, (1988).

Finally, during the cooling P-T path, Assemblage 3 was formed in the high-grade schist including andalusite + K-feldspar + plagioclase + quartz. We interpret the textural relations between the andalusite and the feldspars as indicating crystallization of small amounts of granitic melt in the andalusite stability field. Assuming the availability of water during crystallization, the lack of muscovite suggests that conditions still lay above the muscovite breakdown curve. These considerations suggest P-T conditions of T ~600°C and P < 2 kbar for this assemblage. Cooling continued through 450-550°C as indicated by the results obtained with the two-feldspar thermobarometer. Very similar textures involving andalusite also developed in the gneissic rocks, with the difference that a much larger proportion of melt was involved, and cordierite also crystallized together with andalusite and the granitic assemblage. The P-T conditions for crystallization of this late assemblage must have been similar to those in the high-grade schist.