| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology | Volume 44 | Number 2 | Pages 355-385 | 2003
© Oxford University Press 2003
Glass-rich, Cordierite–Biotite Rhyodacite, Valle Ninahuisa, Puno, SE Peru: Petrological Evidence for Hybridization of ‘Lachlan S-type’ and Potassic Mafic Magmas
DEPARTMENT OF GEOLOGICAL SCIENCES AND GEOLOGICAL ENGINEERING, QUEEN'S UNIVERSITY, KINGSTON, ON., CANADA, K7L 3N6
Corresponding author. Present address: Canada-Nunavut Geoscience Office, PO Box 2319, Iqaluit, Nunavut, Canada X0A 0H0. E-mail: hsandema{at}nrcan.gc.ca
RECEIVED March 29, 2002; ACCEPTED August 29, 2002
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONGEOLOGICAL SETTINGPETROGRAPHY AND MINERAL...WHOLE-ROCK GEOCHEMISTRYDISCUSSIONCONCLUSIONS: IMPLICATIONS FOR...APPENDIX: ANALYTICAL METHODSREFERENCES |
|---|
The Revancha dyke, a 24·1 Ma,
5 m wide body of cordierite–biotite rhyodacite exposed in the Central Andean Inner Arc of southeastern Peru, is interpreted as a pristine analogue of a Lachlan-type, strongly peraluminous, S-type monzogranite. The preservation of both glass and microphenocrysts, as well as the hypabyssal setting, provide evidence of crystallization processes that are largely disguised in the more widespread plutonic or the largely pyroclastic extrusive examples of such rocks. The dyke contains up to 74% undevitrified glass, phenocrysts of plagioclase, biotite, Fe-cordierite and rare quartz and sanidine, and accessory sillimanite, apatite, zircon, ilmenite and monazite. Both plagioclase and biotite phenocrysts exhibit complex zonation and convincing evidence of magmatic dissolution, whereas groundmass microphenocrysts of these minerals commonly exhibit, respectively, more calcic and magnesian compositions than the phenocrysts. The magma was highly reduced, the sparse ilmenite approaching stoichiometric FeTiO3. Although comparable in most salient aspects with S-type granitoids of the Lachlan Fold Belt of SE Australia, the rhyodacite is enriched in K2O, Na2O, Al2O3, P2O5, U and Rb, but depleted in MgO, CaO, FeOT, TiO2, V and Co. The Revancha rhyodacite, like many peraluminous S-type suites, exhibits mineralogical, textural and chemical features strongly supporting shallow-crustal hybridization with a mafic melt, or melts, of mantle derivation, but the geochemical relationships suggest that the latter were probably markedly potassic rather than basaltic. The Revancha rhyodacite is a unit of the Picotani Group, a 22–26 Ma suite of diverse mafic and felsic rocks emplaced along the foreland boundary of the Central Andean orogen during the Aymará tectonic event. High heat-flow and relatively thin orogenic crust at this time promoted anatexis of clastic rocks at shallow depths. This geotectonic setting contrasts markedly with that of the subsequent, 7–17 Ma, Macusani Formation ash-flows, and other strongly peraluminous, ‘Himalayan-type’, two-mica rhyolites and granites of the Quenamari Group, emplaced during the major crustal thickening of the Quechuan orogeny. KEY WORDS: glassy rhyodacite; Lachlan S-type; sieve textures; K-rich mafic melt hybridization; petrogenesis
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGPETROGRAPHY AND MINERAL...WHOLE-ROCK GEOCHEMISTRYDISCUSSIONCONCLUSIONS: IMPLICATIONS FOR...APPENDIX: ANALYTICAL METHODSREFERENCES |
|---|
Chappell & White (1974) first proposed, on the basis of field studies of two restricted areas of the Palaeozoic Lachlan Fold Belt (LFB) of eastern Australia, that granitoid rocks globally fall into two broad categories, namely I-type, i.e. derived from (meta-)igneous protoliths, and S-type, ultimately of (meta-)sedimentary derivation. The former are generally hornblende bearing and hence metaluminous or, in more silicic members, weakly peraluminous [aluminium saturation indices (Zen, 1986)
1·1], whereas the latter were described as lacking hornblende but containing the Al-rich minerals muscovite (‘common’) and, less consistently, garnet and/or cordierite, in addition to abundant biotite. In a recent reappraisal of this critical concept, drawing on a quarter-century of continued research on the granitoid rocks of the Lachlan Fold Belt, Chappell & White (2001) maintained the fundamental validity of the two-fold classification, although noting that the more felsic members of the two suites converge towards similar minimum melt compositions. Although questioning the general significance of enclaves, they stated that hornblende-bearing and metasedimentary enclaves occur, respectively, only in I- and S-type suites, a distinction ‘supported by all subsequent studies’ (p. 495), a relationship considered to reflect the postulated ubiquitous retention of restite material (White & Chappell, 1977). They moreover argue for a close inherent relationship between muscovite-rich and cordierite-bearing peraluminous granites, interpreting the former as more strongly fractionated, and hence more hydrous equivalents of the latter. Finally, they propose that all markedly peraluminous granites are products of ‘low-temperature’ melts, emphasizing that the vast majority of such rocks in the Lachlan Fold Belt contain over 65 wt % SiO2. Significantly different petrogenetic models have, none the less, been proposed by other workers. Thus, Barbarín (1996), in a review of strongly peraluminous granitoid suites, argued that ‘two-mica’, i.e. muscovite-bearing granites (MPG) and rhyolites differ critically from cordierite–biotite-bearing rocks (CPG) in exhibiting positive rather than negative correlations of aluminium saturation index with SiO2 content and other differentiation indices. Sylvester (1998) similarly concluded that these two clans of peraluminous rocks, distinguished as Himalayan- and Lachlan-type, respectively, record fundamentally differing petrogenetic and geotectonic environments, the two-mica suites forming under higher pressures and lower temperatures in response to greater degrees of orogenic, post-collisional, crustal thickening. Patiño-Douce (1999), integrating the results of melting experiments with the major element compositions of diverse granitic rocks, has reiterated the case for distinguishing peraluminous, muscovite-rich leucogranites and rhyolites (‘PLGS’) from peraluminous S-type granites (‘PSGS’), which characteristically exhibit low-P, high-T mafic aluminous minerals, i.e. cordierite, spinel and/or Al-rich orthopyroxene. He emphasizes that among common granite types only the former are likely to represent pure anatectic melts, interpreting the generally less siliceous magmas, including PSGS, as incorporating residual (restite) and/or peritectic phases generated through the reaction of basaltic magmas with quartzo-feldspathic crustal protoliths. PSGS are thus considered as products of ‘hybrid magmas’ produced through shallow-crustal dehydration-melting of biotite-rich metapelites.
Persisting controversy regarding the genesis of true S-type granitic rocks, and particularly the extent to which their mineralogy and chemistry are influenced by restite retention (see Wyborn et al., 1981; Clemens & Wall, 1984), in part reflects the predominance of coarse-grained phaneritic intrusive facies, in which melt compositions are not preserved and crystallization history is disguised. Moreover, most volcanic representatives are either altered or pyroclastic, and the compositions of microphenocrysts, which in many volcanic rocks record the conditions of late-stage magmatic consolidation, have not been recorded in S-type suites. The salient objective of the present contribution is the documentation of a young Lachlan-type, or PSGS, rock emplaced in a transitional, hypabyssal setting, thereby preserving both unaltered phenocrysts and microphenocrysts, as well as undevitrified glass. The subject of discussion is a member of a well-documented suite in a clearly defined Central Andean geodynamic environment which was also the site of emplacement of muscovite-bearing, Himalayan-type, granites and rhyolites, including the well-known Macusani rhyolites. Moreover, a nearby, closely allied and coeval cordierite–biotite granite pluton hosts the richest giant hard-rock Sn deposit, San Rafael, and clarification of the petrogenesis and crystallization history of the S-type magmas of the region has clear implications for metallogenesis and ore-genesis. Comparisons are made herein with the classic S-type associations of eastern Australia, and with the few similarly fresh cordierite-phyric suites that have been described, namely, the largely ignimbritic Canberra (Wyborn et al., 1981) and Violet Town (Clemens & Wall, 1984) volcanics of the LFB, a dacite from the Upper Miocene Cerro del Hoyazo volcanic centre, SE Spain (Zeck, 1970, 1992), and the Upper Miocene cordierite rhyolites of the Morococala Meseta, Bolivia (Morgan et al., 1998).
|
GEOLOGICAL SETTING |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGPETROGRAPHY AND MINERAL...WHOLE-ROCK GEOCHEMISTRYDISCUSSIONCONCLUSIONS: IMPLICATIONS FOR...APPENDIX: ANALYTICAL METHODSREFERENCES |
|---|
The uppermost Oligocene, rhyodacitic Revancha dyke crops out in the upper Ninahuisa Valley (Fig. 1), at 14° 4'4''S latitude, 70° 35'3''W longitude in Puno Department of southeastern Peru. The valley is eroded into the SE quadrant of the Meseta de Quenamari, a high tableland underlain largely by the flat-lying Middle to Upper Miocene Macusani Formation (Fig. 1). The meseta and a chain of 4500–5200 m a.s.l. (above sea level) glaciated peaks rimming its southern margin were assigned by Laubacher (1978) to the Precordillera de Carabaya, a southwestern outlier of the main Cordillera de Carabaya, itself a segment of the Cordillera Oriental. The area lies within the Inner Arc magmatic domain of Clark et al. (1983, 1984, 1990), a petrographic subprovince characterized since the Permian by compositionally diverse magmatism with both crustal and mantle sources, focused along the craton–orogen interface (Kontak et al., 1984; Kontak, 1985). The Inner Arc is contrasted to the more homogeneous Main Arc of the Central Andes, which underlies the entire Cordillera Occidental and much of the Altiplano at this latitude.
|
The Cenozoic volcanic and sedimentary rocks of the Cordillera de Carabaya region were originally grouped together by Laubacher et al. (1988) as the Cayconi Formation, but were assigned to the Crucero Supergroup by Sandeman (1995) and Sandeman et al. (1995, 1997a) to express their petrological and petrogenetic diversity. On the basis of stratigraphic and geochronological relationships, this subaerial succession was subdivided into an older Picotani Group, 22–26 Ma in age and which includes the Revancha dyke, and a younger, 6·5–17 Ma, Quenamari Group, each associated with a cogenetic intrusive suite.
The study area is underlain by a thick sequence of clastic sedimentary strata of the Ordovician Sandia Group and Carboniferous Ambo Group, as well as basalts and trachyandesites of, respectively, the Permian Mitu Group and Jurassic AllincÄ&!aacute;pac Group (Kontak et al., 1990a). These are unconformably overlain by silicic pyroclastic rocks and high-K calc-alkaline basaltic flows of the Picotani Group (Sandeman, 1995; Sandeman et al., 1997a), which record the mid-Cenozoic revival of Andean magmatism along the foreland margin of the orogen, coincident with the Aymará (Pehuenchean) tectonic event (Sandeman et al., 1995). Tectonic and geomorphological relationships suggest that the Picotani Group was emplaced into continental crust of normal, or even below-normal thickness, whereas the ensuing Quenamari Group magmatic activity clearly accompanied major crustal thickening throughout the Middle and early Late Miocene (Kontak et al., 1990b; Sandeman et al., 1995).
Commingling and mixing of Lachlan S-type granitic magmas with absarokitic basaltic melts have been documented for Picotani Group suites elsewhere in the Precordillera de Carabaya by Kontak et al. (1986) and Kontak & Clark (1997, 2000), and Carlier et al. (1992, 1993, 1997), Sandeman & Clark (1993) and Sandeman et al. (1997a; unpublished data, 1995) have described similar relationships between S-type and minette and/or lamproite magmas in the Picotani Group and Intrusive Suite. Commingling between mantle-derived potassic and Lachlan-type, PSGS crustal melts therefore appears to have been wide-spread in the SE Peruvian Inner Arc domain in the late Oligocene–Early Miocene.
The Revancha dyke, a minor unit of the Picotani Intrusive Suite (Sandeman et al., 1997a), attains a maximum exposed width of 5 m. It cuts rhyodacitic ash-flows and monzogranitic intrusive rocks, with which it shows close mineralogical and chemical similarities, on the southern slope of the valley of the Ninahuisa River, 1.6 km WNW of Hacienda Ninahuisa. The dyke strikes 120°, dips 67° SW, and has been traced for 
400 m. To the NW, it is covered by fluvio-glacial sediments. Chilled margins, commonly flow-banded (Fig. 2a), enclose a porphyritic interior (Fig. 2b), but centimetre-scale variations in grain-size and crystallinity are locally observed. The rock has a striking, but misleading, mafic appearance in outcrop, with dark mauve cordierite phenocrysts scattered through a matrix of blackish glass. The dyke is unweathered and unaffected by hydrothermal activity.
|
|
PETROGRAPHY AND MINERAL CHEMISTRY |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGPETROGRAPHY AND MINERAL...WHOLE-ROCK GEOCHEMISTRYDISCUSSIONCONCLUSIONS: IMPLICATIONS FOR...APPENDIX: ANALYTICAL METHODSREFERENCES |
|---|
Both microscopic and outcrop-scale textural relationships indicate that, although the flow-banded, crystal-poor facies is characteristic of the margins of the dyke, in general this intruded more crystal-rich, coarser-grained rhyodacite; thus, flow-bands in the latter are truncated by the boundaries of the former. The two facies exhibit significantly different textural features and mineral compositions, and are hence distinguished in the ensuing discussion as, respectively, the glass-rich (GR) and glass-poor (GP) facies. Sanidine and quartz phenocrysts occur only in the GP facies, in which cordierite and biotite are also coarser, attaining diameters of 10 and 5 mm, respectively. Modal compositions of the two facies, based on at least 2000 points/section for 22 sections, are given in Table 1.
|
Plagioclase
Plagioclase is the dominant phenocryst phase in both GP and GR facies of the dyke, comprising 13–18 modal % of the rock. It exhibits a wide variety of grain-sizes, morphologies and textures. Phenocrysts, generally 0·5–4 mm in diameter, occur as resorbed, subhedral grains or grain fragments. The majority incorporate zones with sieve textures (Tsuchiyama & Takahashi, 1983), which may be classified as either coarse, interconnected, ‘honeycomb’ networks of glass tubules or ‘dusty’ associations of very small, typically unconnected glass inclusions. Elongate, prismatic microphenocrysts (microlites), generally
0.5 mm in length, are common throughout both facies of the dyke. Approximately 80% of the plagioclase phenocrysts in the GR facies are rounded to subhedral and exhibit extensive honeycomb sieve textures. Clear, but locally mineral inclusion-rich, cores (Fig. 3a) are enclosed by honeycombed mantles, which are in turn overgrown by clear, strongly zoned rims. The core-zones commonly display albite or carlsbad twinning that persists through the surrounding sieved zones (Fig. 3a). The honeycomb texture characteristically comprises concentric arrays of glass tubules, and is confined to specific internal surfaces such as twin planes, oscillatory zones whose composition appears to have been more sodic than that in equilibrium with the melt, and zones overgrowing truncated zoning (evidence of earlier thermal resorption). The outer zones of the grains exhibit marked gradational extinction (Fig. 3b), but Nomarski differential interference contrast (NDIC) images of fluoroboric acid-etched polished surfaces reveal very fine oscillatory laminations, which are typically truncated by irregular resorption surfaces at the grain boundaries. A few honeycomb-textured crystals show a complete spectrum of these features. Plagioclase phenocrysts with zones of fine ‘dusty’ sieve texture are rare in the GR facies of the dyke. Microlitic plagioclase grains in this facies are, like the phenocrysts, rounded, and similarly widely exhibit honeycomb textures, evidence of temperature rise at a very late, near-surface stage of emplacement. It should be emphasized that, like most plagioclase microphenocrysts in a wide range of volcanic rocks (e.g. Pearce & Clark, 1989), those in the dyke exhibit fine-scale oscillatory zoning in NDIC images. Such zoning is truncated by the honeycomb tubules, and these textures do not represent skeletal products of rapid growth.
|
Plagioclase phenocrysts in the GP facies of the dyke conform to three distinct morphotypes. Most abundant are subhedral grains that incorporate core zones with patchy extinction (Fig. 3c), and resorbed and weakly sieved margins. Carlsbad and albite twinning is largely obscured in the patchy textured cores, but well developed elsewhere. Glomerocrysts are abundant and comprise 3–5 grains, each with a patchy textured core zone. The second morphotype comprises subhedral and resorbed grains exhibiting extensive areas of fine, ‘dusty’ sieve texture (Fig. 3d), and the third, which may be xenocrystal, comprises subhedral, weakly resorbed grains that exhibit mineral inclusion-rich cores and oscillatorily zoned overgrowths (Fig. 3e). The matrix of the GP facies is characterized by numerous flow-aligned, albite-twinned, subhedral to euhedral plagioclase microphenocrysts that lack sieve texture but commonly exhibit minor resorption.
Plagioclase compositions (Table 2), including all phenocrysts (Figs 4a and b, 5a and b) and microphenocryst textural variants (Figs 4c and 5c) in both facies of the dyke, vary widely. Phenocrysts from the GP facies (An20–45: Figs 4a and 5a) are, however, overall less variable in composition than those of the GR facies (An20–67: Figs 4b and 5b). The most calcic plagioclase in the GR facies occurs at the inner boundaries of the oscillatorily zoned mantles surrounding areas of intense honeycomb sieve texture. These mantles commonly display, near the outer boundaries of the sieved zones, a strong increase in An (up to 12 mol %) relative to the cores (Fig. 3b). Such ‘calcic spikes’ are, however, succeeded outwards by abrupt and continuous decreases in An towards the rims of the grains. The most calcic analyses overlap in composition with a field defined by 60 phenocryst and microlitic plagioclases from uppermost Oligocene basaltic rocks of the region (MVP, Fig. 4, H. A. Sandeman, unpublished data, 1995). Patchy textured (Figs 3c and 5a) and fine dusty sieve-textured grains (Figs 3d and 5a) are typically more sodic, ranging from An22 to An44. Analysis of dusty sieve-textured grains also reveals high apparent K concentrations, attaining Or31, but this probably reflects incorporation of potassic glass (Fig. 3d). Phenocrysts with apparently undisturbed normal oscillatory zoning range overall from An20 to An59 (Fig. 5a and b), but compositions more calcic than An26 typically predominate. Many of these oscillatorily zoned grains (see Fig. 3e) exhibit a progressive rimward decrease in An content. Similarly simple normal zonation has been reported for plagioclase in comparable volcanic rocks, the Hawkins, Goobarragundra and Laidlaw Suites of the LFB, by Wyborn et al. (1981), who indeed interpreted the plagioclase, biotite and quartz phenocrysts of the Hawkins Suite as restitic phases. Clemens & Wall (1984), however, recorded plagioclase grains in the Violet Town volcanics, also in the LFB, with up to two internal sieved and Ca-rich zones indicative of partial resorption.
|
|
, Ma8915b; •, MAC175; 
, MAC137.
|
Plagioclase microphenocrysts in the dyke exhibit wide compositional ranges (Figs 4c and 5c) and, remarkably, many are as calcic as the most Ca-rich zones in the phenocrysts. This is particularly evident in the GP facies, in which microphenocrysts have compositions of An33–61, most being more calcic than the associated phenocrysts. Rounded and sieved, albite-twinned microphenocrysts in the GR facies exhibit a similar overall range from An25 to An61 (Figs 4c and 5c), but on average are less calcic than those in the GP facies. It is therefore apparent that both textural facies of the dyke record complex thermal histories, the late-crystallizing microphenocrysts revealing apparent temperature ranges similar to those over which the phenocrysts had formed. Moreover, the GP facies may have experienced an overall prograde evolution, its slightly more calcic microphenocrysts suffering less dissolution than those in the GR facies. The extreme variability in the composition of ‘contiguous’ microphenocrysts in both facies, however, is evidence for marked spatial variations in T.
|
Cathodoluminescence (CL) images (not shown) of plagioclase phenocrysts reveal that the oscillatory zoning is associated with variations in the intensity of a greenish yellow CL response at a comparable scale. CL emission spectra for several grains (A. N. Mariano, personal communication, 1997) are in conformity with excitation by Fe2+ and/or Mn2+. No crystals exhibit the reddish brown response characteristic of Fe3+ excitation, a feature shown by the corroded outer zones of phenocrysts in several volcanic units of the Picotani Group that record the mixing of reduced granitic and more oxidized minette magmas (Sandeman & Clark, 1993; Sandeman et al., 1997a). Indeed, the majority of the Revancha plagioclase phenocrysts are cut by networks of thin veinlets exhibiting intense yellow CL response; microphenocrysts are similarly affected. Electron microprobe X-ray mapping of plagioclase phenocrysts reveals that the selvages of the microfractures are enriched in Fe rather than Mn. The crosscutting features, most of which are not visible in transmitted light, may contain films of glass. We propose that the feldspar grains experienced late-stage thermal shock with the concentration of Fe2+-enriched plagioclase within the resultant fractures, i.e. local Fe metasomatism, or redistribution, occurred under reduced conditions.
Sanidine
In the GP facies of the dyke, large (2·5 cm), euhedral phenocrysts of translucent sanidine comprise up to 20 modal % of the rock, resulting in a porphyritic texture that closely resembles that in the adjacent phaneritic monzogranite. These crystals enclose numerous subhedral inclusions of oscillatorily zoned plagioclase, biotite and quartz, none exhibiting features suggestive of a restitic origin. This, along with its restriction to the coarsely crystalline facies, suggests that sanidine was paragenetically late.
Representative sanidine analyses are given in Table 3 and the entire dataset is plotted in Fig. 4d. The composition ranges from Or70Ab30 to Or75Ab25, with 1.0 wt % An and up to 0·39 wt % BaO. Sanidine in the glass-poor facies of the Revancha dyke is very similar in composition to alkali feldspars in unmetasomatized, but holocrystalline, Lachlan-type PSGS granites in SE Peru (Kontak, 1985; H. A. Sandeman, unpublished data, 1995), and generally comparable with those in the S-type granites and volcanic rocks of the Kosciusko and Melbourne Basement terranes of the Lachlan Fold Belt (Wyborn et al., 1981; Elburg, 1996a).
Quartz
Quartz occurs only in the GP facies of the dyke, in which it attains 5 modal %. It forms large (7 mm), rounded and embayed bipyramidal phenocrysts that commonly contain inclusions of apatite, biotite and plagioclase. It displays arcuate, anastamosing fractures, typically decorated by small melt inclusions, implying that fracturing occurred before solidification of the melt. Such features could result from heating during adiabatic magma ascent, but their close association with mafic enclaves elsewhere in the Picotani Group supports an origin through thermal shock during commingling with mafic magmas (Kontak et al., 1986; Kontak & Clark, 1997). Because quartz occurs only in the GP, it is inferred to have crystallized at a relatively late stage.
Biotite
Biotite forms 8–14 modal % of the rock, occurring both as ragged, stubby grains up to 5 mm in size and as elongate prismatic microphenocrysts (0·5 mm). The former exhibit pale to dark brown pleochroism and contain abundant inclusions of apatite, monazite and zircon, and sparse inclusions of ilmenite. Many have rounded, ragged and/or embayed forms (Fig. 6a), and enclose numerous blebs of glass and quartz. Those occurring in GR domains are typically more strongly resorbed than those in the GP facies, implying that the former may represent a hotter melt that intruded the latter. Biotite microphenocrysts in both facies, however, exhibit similar forms and, unlike the large phenocrysts, lack sieve textures.
|
Representative biotite analyses are reported in Table 4 and a more comprehensive dataset is plotted in Fig. 7, including both large, stubby sieve-textured grains and elongate prismatic microphenocrysts. Electron microprobe analysis was supplemented by titrametric determination of FeO on high-purity separates (250–354 µm) of micas from four specimens, permitting estimation of FeO:Fe2O3 ratios, at least on the multi-grain scale. The rocks were gently ground under inert liquids, and we consider it improbable that the micas were oxidized during either crushing or titration. The FeO content determined for a separate was assumed to apply to all microprobe analyses of phenocryst grains (>250 µm) in that specimen. Biotite phenocryst compositions are characterized overall by elevated TiO2 (3·2–4·5 wt %) and high MgO (10·4–13·7 wt %). There is extensive compositional overlap for phenocrysts in the GP and GR facies, but the former exhibit overall higher Fe(Fe + Mg) ratios (Fig. 7) paralleling the contrast in plagioclase phenocryst compositions. Biotite microphenocrysts from the dyke are characterized, on average, by more variable, and typically higher, TiO2 (3·17–5·79 wt %) and higher MgO (11·4–15·7 wt %) than the associated phenocrysts. Microphenocrysts in the GP facies exhibit a wider range of Fe(Fe + Mg) ratio than those in GR (Fig. 7), but there is probably no significant difference in the two populations. All microphenocrysts show wide variations in AlIV, but this feature is particularly evident in the GR facies.
|
|
Figure 7 highlights the elevated Mg contents of the Revancha biotites in comparison with those from a range of peraluminous (Lachlan- and Himalayan-type), and even I-type rocks. There is, in contrast, considerable overlap in the AlIV contents. The Revancha phenocryst biotites overlap with the AlIV and Mg-rich sector of the field delimited for micas from other cordierite–biotite monzogranitic or rhyodacitic rocks of the Picotani Group and Intrusive Suite (Kontak, 1985; Kontak et al., 1986; Kontak & Clark, 1997), particularly those from units known to have undergone commingling with mafic potassic magmas (Sandeman & Clark, 1993). Close similarities are apparent with the biotite from the commingled Minastira pluton (Kontak & Clark, 1997). The biotites predictably exhibit higher mg-numbers than those in the Himalayan-type, two-mica suites of the Cordillera de Carabaya and are significantly more magnesian than those in the cordierite rhyolites of the Morococala Field, Bolivia (Morgan et al., 1998). They are, moreover, more magnesian than biotites in the S-type plutonic and volcanic rocks of the LFB. Microphenocryst biotites exhibit highly variable AlIV and even more magnesian compositions than the stubby, sieved phenocrysts, and represent exceptional compositions for biotite in a silicic rock.
Analysis of biotite phenocryst separates indicates that these are much richer in Fe3+ than would be expected from the reduced nature of the bulk-rock, as indicated by the absence of magnetite and the scarcity of ilmenite. The unweathered and unaltered condition of the dyke implies that little post-crystallization alteration has occurred, and the rocks do not contain xenocrysts, such as of Cr-rich phlogopite, derived from mixing with minette magmas, a common feature in other units of the Picotani Group (Sandeman & Clark, 1993). The analysed biotites are therefore interpreted as true phenocrysts. The bulk compositions plot adjacent to the hematite–magnetite buffer in the Fe2+–Fe3+–Mg2+ diagram of Wones & Eugster (1965). Apparent intra-sample variation is evident, but this is ascribed to the assignment of a single FeO content to the several microprobe analyses determined for each hand-specimen: thus, the variation in Fe2O3 is estimated by difference, and is probably an exaggeration of the real range. However, an inter-sample variation in the total iron (FeT) and Mg contents is evident, with a decrease in Mg from sample MAC-2 to sample MAC-175, respectively GR (74% glass) and GP (45%). That the biotites crystallized, or recrystallized, under variable oxidation conditions, ranging from the Ni–NiO to the Hem.–Mag. buffers, is shown by the variable Fe2+:Fe3+ ratios (Table 4), and it is inferred that those in the GR facies record the lower fO2 conditions. The apparent trend in biotite composition with crystallization is comparable with that documented for the Ploumanac'h granitoids by Barrière & Cotton (1979), who attributed it to early fractionation of ilmenite with subsequent late- or post-magmatic oxidation of Fe2+. We therefore suggest that the magmatic fO2 at the time of initial biotite crystallization is revealed by the least oxidized grains, i.e. it slightly exceeded the Ni–NiO buffer.
It should be noted that comparable, or even higher, Fe3+:Fe2+ ratios were recorded by Kontak (1985) for biotites from the highly reduced Macusani Formation rhyolites: the Fe3+ enrichment, equivalent to fO2 values well above the Ni–NiO buffer, was tentatively ascribed to late-stage oxidation associated with volatile concentration before eruption. These chemical data were reported, but not discussed, by Pichavant et al. (1988b, table 6). It should also be noted that Kontak (1985) recorded a wide range of Fe2+:Fe3+ ratios in biotite in several holocrystalline cordierite–biotite monzogranites of the Picotani Intrusive Suite, ranging from the QFM (quartz–fayalite–magnetite) buffer to well above the Ni–NiO buffer, and inferred that the silicic magmas in several of the hypabyssal centres experienced significant late-magmatic oxidation.
|
Cordierite
Cordierite, with an intense mauve colour, forms up to 4 modal % of the rock as euhedral, pseudo-hexagonal prisms, typically
4 mm in diameter but attaining 10 mm in GP specimens (Fig. 6c). The phenocrysts commonly exhibit sector-twinning and incipient alteration to pinite assemblages along grain boundaries and fractures, and contain rare sillimanite inclusions. Because of its abundance and large grain-size in the GP domains, and the absence of feldspar inclusions, cordierite is inferred to be paragenetically early (see Morgan et al., 1998). Cordierite phenocrysts are normally zoned, the cores and rims exhibiting mg-numbers [100 x molecular Mg/(Mg + FeT)] of 71 and 
55, respectively, with a mean value of 60. Representative compositions are presented in Table 5. The cordierite compositions are comparable with those from other plutonic and volcanic settings (Deer et al., 1986), including the broadly similar PSGS granites of the Cornubian batholith (Stimac et al., 1995), the Violet Town, Hawkins and Goobarragandra Volcanics of the LFB (Wyborn et al., 1981; Clemens & Wall, 1984) and the PSGS volcanic rocks of the Morococala Field, Bolivia (Morgan et al., 1998). Cordierite from both SE Australian and SE Peruvian PSGS monzogranites contains slightly less SiO2 than that reported by Montel et al. (1986), but is significantly lower in Al2O3 and higher in MgO and FeO than cordierite-like phases from the Macusani Formation (Pichavant et al., 1988b).
|
Accessory phases
Sillimanite occurs as sparse, isolated, prismatic microphenocrysts in the glass matrix, but more commonly forms acicular crystals included in sanidine, plagioclase and, rarely, biotite and cordierite. Ilmenite, the only Fe–Ti mineral, is rare, occurring both as small subhedral microphenocrysts in the glass matrix and, more widely, as inclusions in other phases, in particular biotite. Monazite and zircon occur as tiny dispersed prismatic grains but more commonly form inclusions in other phases. Apatite occurs as subhedral microphenocrysts (Fig. 6b) and as inclusions in biotite, and commonly hosts numerous micro-inclusions of sillimanite, monazite, zircon and ilmenite.
Representative electron microprobe analyses of selected accessory phases are presented in Table 6. Apatites in the Revancha dyke are fluorapatites, with compositions similar to those from the Cornubian batholith (Stimac et al., 1995); it should be noted that F is mobile under the analytical conditions employed, and the data should be considered semi-quantitative. Cathodoluminescence emission spectra of several Revancha apatites (A. N. Mariano, personal communication, 1997) record enrichment in heavy rare earth elements (HREE).
The composition of the ilmenite has been examined in detail. No evidence was found for intragrain zonation. The Fe3+/(Fe2+ + Fe3+) ratios range narrowly from 0·000 to 0·094 for all analysed grains (Table 6), and the data form a tight cluster near stoichiometric FeTiO3, possibly with a restricted extension towards TiO2. The elevated MgO contents (1·38–2·35 wt %), confirmed by both WDS and EDS techniques, are interpreted to record crystallization from a relatively primitive, high-T magma. Similar MgO contents, i.e. 3·13 wt %, have been determined by WDS analysis for ilmenites from the coeval San Rafael cordierite–biotite monzogranite (A. H. Clark, unpublished data, 1997). The close approach to stoichiometry implies strongly reducing conditions, a feature in apparent conflict with the variable compositions of the biotite phenocrysts. However, it should be noted that the comparably reduced Macusani Formation rhyolitic ash-flows contain ilmenite with a markedly more variable composition, embracing not only a wide range of Mn:Fe ratios, but also much more extensive solid solution of Fe2O3 and apparent alteration to ‘pseudorutile’ (Pichavant et al., 1988b). Both the Macusani and Morococala ilmenites are significantly less magnesian, with 0·35 and 0·41 ± 0·16 wt % MgO, respectively, than those from Revancha, a contrast paralleling that shown by the associated biotites. Moreover, the ilmenite from the Morococala quartz latite, a rock similar to the Revancha rhyodacite in SiO2 content, is markedly less magnesian (0·67 ± 0·25% MgO: Morgan et al., 1998), although the Cr contents of these rocks are comparable.
Possible xenocryst material
A painstaking search of the 22 sections cut from the dyke revealed only four polycrystalline micro-inclusions that may represent restite, xenocryst bodies or, less probably, material incorporated from a cogenetic mafic melt. In addition, a single grain of orthopyroxene (Fig. 6c) was observed. The micro-inclusions comprise: (1) aggregates of 2 mm diameter garnet + plagioclase + biotite + cordierite + orthopyroxene + quartz (Fig. 6d); (2) 1 mm aggregates of plagioclase + biotite + spinel + fibrolitic sillimanite (Fig. 6e); (3) 1·5 mm plagioclase + biotite aggregates; (4) 2·5 mm aggregates of cordierite + fibrolitic sillimanite + spinel. Orthopyroxene, spinel and garnet have been observed only in the above settings, but spinel and garnet do not occur together. Plagioclase grains in three of the micro-inclusions are mainly intergrown in a granoblastic texture, show albite twinning and, except in the first-mentioned inclusion, do not exhibit sieve textures (see below). They are clearly distinct from both phenocryst and microphenocryst plagioclase grains in the rocks. Spinel grains are colourless to pale green, anhedral, and generally intimately intergrown with sheaves of fibrolitic sillimanite. Garnet is pale red to pink, and anhedral, with embayed outlines apparently resulting from corrosion. In the inclusions, cordierite forms large, anhedral to subhedral grains, intergrown with the replacing anhedral orthopyroxene; it is in part sector-zoned and commonly is weakly pinitized along grain margins and fractures.
The compositions of silicate minerals in these sparse aggregates, including the single orthopyroxene grain from the glass-rich facies of the dyke and the four micro-inclusions, are presented in Table 7. The orthopyroxene yielded compositions of En59·8Fs39·6Wo0·6 to En58·2Fs41·2Wo0·6, with no discernible zonation in BSE images. Plagioclase from inclusion 3 ranges from An55 to An71 (2% Or), and is thus significantly more calcic than the majority of the plagioclase phenocrysts and microphenocrysts, but comparable in composition with plagioclase from high-K mafic volcanic rocks of the region (H. A. Sandeman, unpublished data, 1995). However, plagioclase intergrown with garnet in inclusion 1 displays a composition, An24–42, similar to that of the average plagioclase in the rock (An20–45). Garnet from micro-inclusion 1 is almandine–pyrope (Alm76–77Py17–20) with minor grossular (Grs2–4) and spessartine (Sps0·5–4·3), and exhibits a slight enrichment in Mn from core to rim. Fibrolitic sillimanites from inclusions 2 and 4 approach Al2SiO5, with only 0·15–0·16 wt % FeO and 0·05 wt % of other cations. Spinels from inclusions 2 and 4 are hercynitic with up to 10·5 wt % ZnO (Table 7). Biotite intergrown with spinel in inclusion 2 is more Fe-rich (17·2–17·4 wt % FeOT) than the microphenocrysts of the host dyke, but similar in composition to the phenocrysts therein. Biotite grains occurring adjacent to garnet in inclusion 1 have compositions comparable with those of the phenocryst and microphenocryst grains, whereas those coexisting with plagioclase in inclusion 3 exhibit Fe/(Fe + Mg) values similar to the biotite microphenocrysts, but with lower calculated AlIV (Fig. 7). Cordierite in inclusions 1 and 4 is comparable in composition with the phenocrysts in the dyke.
|
Groundmass glass
Averaged, replicate analyses of groundmass glass from several samples of the dyke are presented in Table 8. The electron microprobe analyses were obtained using ‘macusanite’ glass sample JV-1 (Pichavant et al., 1987) as a secondary standard, thereby obviating problems of Na migration (see Appendix), and permitting direct comparison with strongly peraluminous obsidians from the Macusani Formation. Macusanite has a composition comparable with rare-element pegmatites and has been interpreted as a product of extreme fractionation of a two-mica peraluminous magma (Pichavant et al., 1987, 1988a, 1988b). Glass from Revancha has much lower Al2O3 contents (12·5–13·2 vs 15·8–16·4 wt %), but significantly higher K2O:Na2O ratios (1·74–2·65 vs 0·75–0·86) than macusanite, whereas the remainder of the major element concentrations are comparable (Table 8). High K2O:Na2O ratios similar to those of the Revancha glass have been reported for finely crystalline matrix assemblages, inferred to represent devitrified glass, in the Ministira granite, another unit of the Picotani Intrusive Suite (Kontak & Clark, 1997). Groundmass glasses from the cordierite rhyolites of the Morococala Field are higher in Al2O3 and SiO2, but are characterized by lower abundances of CaO, MgO and, in particular, lower K2O/Na2O ratios than the Revancha dyke.
|
Although the Revancha glasses probably represent residual rather than original melt compositions, they provide, to our knowledge, the first direct evidence for melt compositions in a relatively unfractionated S-type system and, as such, directly complement the data for the Morococala cordierite rhyolites (Morgan et al., 1998).
|
WHOLE-ROCK GEOCHEMISTRY |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGPETROGRAPHY AND MINERAL...WHOLE-ROCK GEOCHEMISTRYDISCUSSIONCONCLUSIONS: IMPLICATIONS FOR...APPENDIX: ANALYTICAL METHODSREFERENCES |
|---|
Major elements
The Revancha dyke is strongly peraluminous (sensu Sylvester, 1998), with whole-rock A/CNK ratios of 1·14–1·22 (Table 9) and normative corundum contents of 2·71–3·55 wt %. It is dacitic in terms of total alkalis vs SiO2 (LeBas et al., 1986) or monzogranitic in the QAP (quartz–alkali feldspar–plagioclase) granitoid classification plot (LeBas & Streckeisen, 1991), and falls within the field defined by other Lachlan-type PSGS granitoids of SE Peru (Kontak, 1985; Kontak et al., 1986; Sandeman, 1995; Kontak & Clark, 1997). Herein, we retain the term rhyodacite, which appropriately distinguishes these intermediate rocks from the locally abundant rhyolites and less common dacites of the Crucero Supergroup (Sandeman et al., 1997a).
|
Figure 8, a plot of normative corundum vs wt % SiO2, compares the Al enrichment of the Revancha rhyodacite with that of other peraluminous rock suites. The dyke samples occupy a tight data cluster, lying within the field for other Peruvian cordierite + biotite monzogranites but having low normative corundum values in comparison with the S-type granitoids and volcanic rocks of the Kosciusko and Melbourne Basement terranes of the Lachlan Fold Belt with moderate SiO2 contents (Hine et al., 1978; Wyborn et al., 1981; Clemens & Wall, 1984; Elburg, 1996a, 1996b; OZCHEM National Whole-rock Geochemical Database). Cordierite rhyolites of the Morococala Field of Bolivia have higher SiO2 but comparable normative corundum contents. Also shown are fields for the Himalayan-type, muscovite + biotite ± andalusite ash-flow tuffs and obsidian (‘macusanite’) of the Macusani Formation, SE Peru (Noble et al., 1984; Pichavant et al., 1988b; Sandeman, 1995). It is apparent that Lachlan-type PSGS rocks are comparable in normative corundum content with the more silicic, Himalayan-type, two-mica rhyolitic or leucogranitic suites. Matrix and quartz-enclosed glasses from the Morococala cordierite rhyolites are somewhat higher (2·7–3·9 and 2·8–3·6 wt %, respectively) in normative corundum than those from the dyke (2·4–3·1 wt %) and similarly overlap in composition with Himalayan-type PLGS rocks. Although glasses from the Revancha dyke exhibit higher SiO2 contents than the whole-rocks, their normative corundum contents and A/CNK ratios are, on average, slightly lower. If the whole-rock analyses are close to liquid compositions, this implies that melt peraluminosity decreased with fractionation, a feature considered by Barbarín (1996) to be characteristic of Lachlan-type PSGS suites. A similar relationship is exhibited by the Morococala cordierite rhyolites, whereas in the ‘Himalayan’ Macusani rhyolites, glass is markedly more aluminous than the corresponding whole-rocks.
|
Figure 9a and b compare major element data for the dyke and its residual glass with various experimental peraluminous glasses as well as with selected rock suites. The glasses synthesized by Patiño-Douce & Johnston (1991) are generally depleted in CaO relative to the Revancha rhyodacite, presumably reflecting the pelitic source composition used in their experiments, whereas those generated in the experiments of Green (1976) and Vielzeuf & Holloway (1988), using metagreywackes as starting materials, are more comparable with those of the Revancha dyke. The cordierite rhyolites of the Morococala Field and their glasses have somewhat more aluminous compositions (Morgan et al., 1998), whereas the Cerro del Hoyazo dacitic lava is strongly enriched in Fe and Mg relative to the Revancha dyke (Zeck, 1992).
|
The dyke is similar in major element composition to other, largely holocrystalline, cordierite–biotite monzogranites of SE Peru (Kontak, 1985; Kontak et al., 1986, 1987; Sandeman, 1995), overlapping with the more SiO2-poor examples of the suite. Like the other monzogranites of SE Peru, the dyke differs in composition from the S-type rocks of the Kosciusko and Melbourne Basement terranes of the Lachlan Fold Belt (Hine et al., 1978; Wyborn et al., 1981; Clemens & Wall, 1984; Elburg, 1996a, 1996b; OZCHEM National Whole-rock Geochemical Database), being generally enriched in K2O, Na2O, Al2O3 and P2O5, but depleted in MgO, CaO, Fe2O3T and TiO2 [Fig. 10; also shown for comparison are the average compositions of the S- and I-type granitoids of the LFB White & Chappell, (1983)]. Cordierite rhyolites from the Morococala Field of Bolivia have higher SiO2, significantly lower MgO, slightly lower CaO, Fe2O3T and TiO2, and comparable contents of Al2O3, Na2O, K2O and P2O5.
|
) and their groundmass glass (•) compared with analyses of PSGS monzogranites from southeastern Peru (
) and I-type (
) granites from that area (Trace elements
The Revancha rhyodacite is similar in its trace element contents to coeval cordierite–biotite monzogranites of SE Peru (Kontak, 1985; Kontak et al., 1986; Kontak & Clark, 1997; Sandeman, 1995), all being enriched in U, Rb, F and Pb, but depleted in V, Co, Zr, Y and Yb relative to the Australian rocks of similar mineralogy (Hine et al., 1978; Wyborn et al., 1981; Clemens & Wall, 1984; Elburg, 1996a, 1996b; OZCHEM National Whole-rock Geochemical Database: Fig. 10). In comparison with the cordierite rhyolites of Bolivia, the Revancha dyke has much higher Cr and Th, mildly higher Ba and Zr, comparable Rb and U, and lower Sr abundances. However, it is comparable in composition with the quartz latites documented by Morgan et al. (1998).
Rare earth elements
Chondrite-normalized (Sun & McDonough, 1989) rare-earth element (REE) contents of seven specimens of the dyke are shown in Fig. 11. All are enriched in the light-REE (LREE; LaN/LuN = 19·4–27·7) and exhibit minor negative Eu anomalies (Eu/Eu* 0·52–0·62: Eu* after Taylor & McLennan, 1985). The compositions fall within a field defined by 14 samples of monzogranite from the region (Kontak, 1985; Sandeman, 1995), although the dyke corresponds to the most REE-enriched holocrystalline granites and also exhibits a generally less pronounced negative Eu anomaly (Fig. 11). The cordierite rhyolites of Bolivia have REE abundances and patterns very similar to those of the Revancha rocks. In comparison, selected S-type rocks of the Kosciusko and Melbourne Basement terranes of the LFB (Hine et al., 1978; Elburg, 1996a; OZCHEM National Whole-rock Geochemical Database), with LREE abundances very similar to the Revancha dyke, exhibit higher HREE abundances, possibly recording the presence of an HREE-bearing phase, probably garnet, in the source of the Revancha magmas.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGPETROGRAPHY AND MINERAL...WHOLE-ROCK GEOCHEMISTRYDISCUSSIONCONCLUSIONS: IMPLICATIONS FOR...APPENDIX: ANALYTICAL METHODSREFERENCES |
|---|
Implications of field and geochronological data
The Revancha Dyke is a 24·1 ± 0·2 Ma body of glass-rich rhyodacite that intruded pyroclastic and porphyritic monzogranitic rocks of similar age and mineralogy, together making up part of the
20–26 Ma Picotani Group of the Crucero Supergroup in the Cordillera Oriental of SE Peru (Sandeman, 1995; Sandeman et al., 1997a). The dyke exhibits two distinct facies, glass rich (74 vol. % glass) and glass poor (51%), and textural evidence implies that the former intruded the latter. The occurrence of high-K mafic rocks of the same age as the Revancha dyke and the other S-type rocks of the region, in conjunction with widespread evidence for felsic–mafic magma commingling and mixing, implies that mantle-derived magmas were both heat sources for the production of peraluminous magmas in southeastern Peru, and mafic end-members involved in magmatic commingling. However, unambiguous macroscopic evidence of commingling with mafic magmas is not observed in the Revancha dyke itself.
Implications of mineral textures and composition
The preserved phenocryst mineral assemblage, plagioclase + biotite + cordierite ± quartz ± sanidine, and the corresponding microphenocryst assemblage, ilmenite (minor) + monazite + apatite + zircon ± sillimanite, are compatible with an origin through dehydration melting of a semipelitic, reduced crustal protolith (Clarke, 1981; Vielzeuf & Montel, 1994; Patiño-Douce & Beard, 1995). The compositions of sanidine, cordierite and ilmenite are comparable with those documented in Lachlan-type PSGS suites worldwide. However, the textural features and compositional variations of plagioclase and biotite in both facies of the dyke are complex, and provide evidence for magmatic resorption of these phenocrysts, which is inferred to have accompanied major temperature rises in the melt, probably during interaction of the PSGS magma with a mafic melt.
Abundant sieve textures in phenocryst plagioclase in both glass-rich and glass-poor facies of the dyke, and in microphenocryst plagioclase in the glass-rich facies, as well as the resorption of large, stubby biotite phenocrysts in both facies, are evidence for thermal disturbance of the magma. The predominance of honeycomb-textured plagioclase in the glass-rich facies, characterized by strongly zoned, unsieved overgrowths that, in turn, exhibit marginal resorption, indicates that the magma experienced at least two distinct thermal perturbations during crystallization. Honeycomb-textured plagioclase grains incorporate calcic spikes (12 An %) on the inner margins of the sieved zones, but the mantles of these grains exhibit a strong outward depletion in calcium, a relationship implying that the enclosing melt as a whole did not experience enrichment in CaO. Comparable honeycomb textures were not observed in the glass-poor facies, but a prograde early thermal history may be recorded by the mottled, patchy extinction in the cores of many of the large plagioclase grains (Fig. 3c). Such zones have commonly been interpreted as representing residual, restitic plagioclase, but these mottled zones are characterized by highly variable compositions, with calcic zones immediately adjacent to sodic ‘patches’, a feature not readily attributable to crystallization during either magmatic cooling, or residuum development at the site of anatexis. The Ca-rich patches may instead represent regions of calcic plagioclase originally surrounded by sodic glass resulting from partial melting of plagioclase crystals in a magma. This is in accord with the observation of Tsuchiyama & Takahashi (1983) that residual, resorbed plagioclase becomes increasingly more calcic with rising temperature. We therefore tentatively interpret the unusual extinction patterns as recording early honeycomb sieve-textured domains that have partially annealed, and infer that, like the glass-rich, the glass-poor facies experienced an early thermal perturbation.
Plagioclase microphenocrysts in both facies of the dyke not only exhibit large ranges in composition but are on average more calcic than most phenocrysts. The microphenocrysts of the glass-poor facies, in particular, are significantly more calcic than the associated phenocrysts, mimicking the compositional variations of the calcic mantles on honeycomb sieve-textured phenocrysts in the glass-rich facies. These relationships imply that the melt from which the microphenocrysts crystallized was hotter and possibly more calcium-rich than that from which the bulk of the phenocrysts precipitated at an earlier stage. Indeed, the latest textural feature preserved in the plagioclase phenocrysts in all specimens is a corrosion of their outer surfaces, implying that the melt was exposed to thermal perturbation before quenching.
The textural–compositional relationships of the plagioclase are parallelled by those of biotite. Thus, biotite phenocrysts in the glass-rich, and to a lesser extent the glass-poor facies of the dyke, exhibit ragged margins and commonly host numerous glass inclusions. Similar textural features have been described by Brearley (1987) for biotite undergoing dehydration breakdown in experimental charges. Thus, following phenocryst growth, the magma is inferred to have experienced a thermal pulse, probably at 850–900°C. Biotite microphenocrysts in both facies, however, exhibit generally euhedral, elongate prismatic forms and are enriched in MgO and TiO2 relative to the phenocrysts. These microphenocrysts exhibit large ranges in calculated AlIV contents, a feature not readily explained through crystallization from a homogeneous silicic PSGS melt. We suggest instead that the microphenocrysts crystallized from a heterogeneous, hybrid melt fraction, or fractions, characterized by variable Al2O3. In conjunction with the documented low Al2O3 and high K2O contents of the Revancha groundmass glass (Table 7), this model offers an explanation for the wide range in plagioclase and biotite microphenocryst compositions and the variable AlIV contents of the latter. The combined implication of these observations is that the PSGS magma represented by both facies of the dyke underwent late-stage mixing with hotter, more mafic, potassic to ultrapotassic magmas.
Physical evolution of the Revancha magmas
The evolution of the Revancha dyke may be deduced from the textures and corresponding c