Micro-structural and compositional variations of hydrothermal epidote-group minerals from a peralkaline granite, Corupá Pluton, Graciosa Province, South Brazil, and their petrological implications

Epidote-group minerals, together with albite, quartz, fluorite, Al-poor and Fe-rich phyllosilicates, zircon, and minor oxides and sulphides, are typical hydrothermal phases in peralkaline alkali-feldspar granites from the Corupá Pluton, Graciosa Province, South Brazil. The epidote-group minerals occur as single crystals and as aggregates filling in rock interstices and miarolitic cavities. They display complex recurrent zoning patterns with an internal zone of ferriallanite-(Ce), followed by allanite-(Ce), then epidote-ferriepidote, and an external zone with allanite-(Ce), with sharp limits, as shown in BSE and X-ray images. REE patterns show decreasing fractionation degrees of LREE over HREE from ferriallanite to epidote. The most external allanite is enriched in MREE. LA-ICP-MS data indicate that ferriallanite is enriched (>10-fold) in Ti, Sr and Ga, and depleted in Mg, Rb, Th and Zr relative to the host granite. Allanite has lower Ga and Mn and higher Zr, Nb and U contents as compared to ferriallanite, while epidote is enriched in Sr, U and depleted in Pb, Zr, Hf, Ti and Ga. The formation of these minerals is related to the variable concentrations of HFSE, Ca, Al, Fe and F in fluids remaining from magmatic crystallization, in an oxidizing environment, close to the HM buffer. L-MREE were in part released by the alteration of chevkinite, their main primary repository in the host rocks.


INTRODUCTION
Epidote-group minerals constitute complex solid solutions series crystallized in several geological environments, making part of a variety of paragenesis in magmatic and metamorphic rocks and pegmatites, as well as hydrothermal and metassomatic rocks (Deer et al. 1986, Gieré andSorensen 2004). The minerals in this group fit the general formula A (1) A (2) M (1) M (2) M (3) O(SiO 4 .Si 2 O 7 ) (OH,F), where the cations with octahedrycal and cubic or greater coordination fill in the M and A structural sites, respectively. The epidote-group mineral nomenclature, as approved by the IMA, was proposed by Armbruster et al. (2006), who recognized three main subgroups: clinozoisite, allanite, and dollaseite. Common epidote presents almost only Al 3+ in M (1) and M (2) , Fe 3+ in M (3) , and Ca in A (1) and A (2) ; in clinozoisite and piemontite, Al +3 and Mn +3 are essential cations in M (3) , respectively. Among the REE-rich minerals in this group, the most important is allanite, where REEs substitute for Ca in A (2) , while Fe 3+ substitutes for Fe 2+ in M (3) . In the dollaseite subgroup, Mg, Fe 2+ , and Mn 2+ are essential cations in M (1) and/ or M (3) . An almost complete solid solution series appears to exist between allanite and epidote endmembers. In ferriallanite (Kartashov et al. 2002) and ferriepidote, Fe 3+ is the dominant cation in M (1) . Substantial reviews on the mineralogy and petrology of this group can be found in Deer et al. (1986), Gieré and Sorensen (2004) and Franz and Liebscher (2004).
Primary epidote and allanite occur as accessory phases in many granitic rocks and register important petrological information. Magmatic epidote occurs in certain relatively Ca-rich metaluminous granitic rocks and may give important geobarometric and oxred informations (e.g. Zen and Hammarston 1984, Sial et al. 1999, Schmidt and Poli 2004. Allanite is a typical accessory mineral in metaluminous to moderately peraluminous granitic rocks and concentrates most of their LREE content. The chemical compositions of allanite correlate well with magma compositions and crystallization environments (Hoshino et al. 2006, Vlach and. Epidote-group minerals are also very common as post-magmatic or meteoritichydrothermal phases in granites and related rocks, where they crystallize from late, residual and/ or external fluids incorporated from the adjacent country rocks (e.g. Exley 1980, Gieré and Sorensen 2004, Jiang 2006. Relatively Ti-rich ferriallanite-(Ce) and allanite-(Ce) are typical primary minerals in metaluminous to moderately peraluminous biotite sieno-and monzogranites of the aluminous (subalkaline) petrographic association from the Graciosa Province of A-type granites and syenites in S-SE Brazil . They also occur in certain hybrid metaluminous syenites and granites from the coeval alkaline association. During the post-magmatic, hydrothermal stages, as well, allanite-(Y) and epidote were formed in slightly peraluminous granites, while ferriallanite-(Ce), allanite-(Ce) and epidote crystallized in peralkaline granites (cf. Vlach and Gualda 2007).
This work deals with micro-structural and compositional features of hydrothermal epidotegroup minerals occurring in a peralkaline alkali-feldspar granite from the Corupá Pluton, Graciosa Province, South Brazil. Optical, BSE and compositional X-ray images are integrated with quantitative chemical data for major-and traceelements obtained with EPMA and LA-ICP-MS in order to characterize their micro-structural and compositional variations and evaluate their role during post-magmatic evolution of peralkaline liquids.

PETROGRAPHY
The Corupá Pluton (Fig. 1) is one of the most typical occurrences of the alkaline association of A-type syenites and granites from the Graciosa Province, S-SE Brazil (Y. Garin, unpublished data, Garin et al. 2003). This province was emplaced in the late Neoproterozoic (ca. 580 Ma) during extensional tectonic regimes Vlach 2007, Vlach et al. 2011). The pluton presents a crescent shape and crops out over 80 km 2 in the southernmost part of the province. It is partially sided by vulcanosedimentary and sedimentary sequences from the Guaratubinha Basin and the Corupá Graben, respectively, and intrudes Archaean granulites from the Luis Alves Micro-plate. The pluton is made up mainly of syenites; granites occur as minor bodies concentrated in its internal zones, probably formed through the crystallization of residual liquids fractionated from more primitive syenitic magmas (Y. Garin, unpublished data). Coeval gabbro-diorites occur as sin-plutonic minor bodies, while hybrid monzodiorites and monzosyenites appear as rounded enclaves with variable dimensions within some medium-grained syenite varieties (Garin et al. 2003).
The syenitic rocks comprise green to pink colored, medium-to coarse-grained, alkali-feldspar syenites and quartz syenites (Y. Garin, unpublished data, Vlach and Gualda 2007). Structures are massive to slightly oriented by magmatic flow, and textures are inequigranular hypidiomorphic, with well-developed K-feldspar tabular crystals. They are metaluminous rocks showing mainly mesophertitic alkali feldspar, with variable amounts of mafic minerals (20 < M < 7) and quartz ( < 10 vol. %). The more mafic syenites contain fayalite, diopside-hedenbergite and calcic amphibole as primary phases. Hedenbergite, calcic and sodic-calcic amphiboles are observed in more evolved varieties. Medium-to coarse-grained alkali-feldspar melasyenites (M up to 65) occur as cumulatic rocks interlayered with the coarse grained syenites in stratified units with restricted distribution. Accessory minerals include mainly ilmenite, chevkinite, zircon and rare apatite; Ti-magnetite occur in some of the most primitive (quartz-absent) syenites, while late titanite is occasional in some evolved syenites. Most of the evolved syenites contain variable amounts of Al-poor, Fe-rich biotite and chlorite, stilpnomelane, actinolitic amphibole, albite and quartz as postmagmatic, hydrothermal minerals, filling in interstices among the main phases.

178
SILVIO R.F. VLACH granular hypidiomorphic. They are leucocratic to hololeucocratic rocks (6 < M < 4), with a peralkaline signature (1.03 < A/NK < 1.06) made of mesophertitic alkali feldspar, quartz and interstitial sodic amphibole (riebeckite), sometimes with sodiccalcic cores, as the main mafic phase. Graphic intergrowths of alkali-feldspar and quartz are common features. Ilmenite, chevkinite, zircon, late titanite and some sphalerite are the main primary accessories. The post-magmatic mineral assemblage will be discussed in the following sections.

MATERIALS AND METHODS
Several samples of the Corupá peralkaline granites were studied in polished thin sections and the features described in this paper are common to most of them. Sample CO-2 ( Fig. 1) was selected for a detailed analytical work because it is very fresh, the described micro-structural features are better developed and an almost complete whole-rock and mineral geochemical data set is available for. Images of hydrothermal epidote-group minerals and quantitative analysis were obtained at the Optical, EPMA and LA-ICP-MS laboratories at Departamento de Mineralogia e Geotectônica, Instituto de Geociências, Universidade de São Paulo.
The EPMA laboratory has a JEOL-JXA8600S equipment with 5 wavelength and 1 dispersive spectrometers automated with the Thermo-Noran Voyager 4.3 system. Qualitative and quantitative analyses were made on C-coated polished thin sections, 30 and 80 µm thick. BSE-Compo imaging and combined WDS and EDS X-ray dot mapping were used to monitor micro-structural and compositional variations. Dot maps were obtained in selected areas, under 20 kV and 100 nA for the column acceleration voltage and beam current, with a resolution of 512 vs. 512 pixels and a dwell time of 0.1 s. Mapped elements were: Si Kα and Fe Kα(EDS), Al Kα, Ce Lα, CaKα, Ti Kα and SrLα(WDS). Si and Fe regions of interest were selected with the minimum possible width.
Quantitative WDS point analyses were made under 20 kV, 50 nA and 1-2 µm for the column acceleration voltage, beam current and diameter, respectively, following the procedures described by Vlach and Gualda (2007). Matrix effects were corrected with the PROZA procedure (e.g. Bastin and Heijligers 1990). Estimated precision was better than 2% for major, between 5 and 10% for minor, and worse for trace elements.
Over 60 point analyses were obtained for the epidote-group minerals. The Fe 2+ / Fe 3+ partition was calculated with the MinCal program (G.A.R. Gualda and S.R.F. Vlach, unpublished data) based on 8 total cations and 25 negative charges, according to Droop (1987). The studied minerals show significant chemical variations in the epidoteallanite-ferriepidote-ferriallanite system. They are very pristine and clear, as the radiation dose was not high enough to cause significant damage, given the low Th ( < 50 ppm) and U (< 10 ppm) contents. So, Fe 2+ / Fe 3+ were also computed following other methods discussed in the literature and results were compared. Structural formulae were calculated according to the suggestions of Dollase (1971), Ercit (2002), Franz and Liebscher (2004) and Armbruster et al. (2006).
LA-ICP-MS analyses were made both in spot (allanite) and raster (epidote) modes in 80-μm thick polished thin sections using the Elan 6100DRC ICP-MS equipped with a New Wave UP-213 LA system, with a 213 nm Y(Nd)AlG laser. The energy density and the beam diameter at the sample surface were close to 8.6 J/cm 2 and 65 (spot) or 40 (raster) µm, respectively. The used analytical protocol was similar to that presented by Gao et al. (2002), and 44 elements were quantified with a total acquisition time of 120s, equally divided for sample and blank measurements. Signal integration and dwell times were 8.3 and 1.66 ms, respectively. Daily oxide formation ratio was controlled by ThO formation EPIDOTE-GROUP MINERALS IN PERALKALINE GRANITE, GRACIOSA PROVINCE ratio and maintained below 1 %. Data acquisition, drift corrections, as well as data treatment and conversions to concentrations were done with the Glitter software (van Achterbergh et al. 2007). Synthetic glasses SRM-612 (NIST) and BHVO-2G were used as standards. Si and Ca contents, previously measured with EPMA, were used as internal standards for normalization purposes.
The spatial resolution of the LA-ICP-MS technique, about 4 to 5 orders of magnitude lower than EPMA, turns difficult to solve the thinnest compositional variations in the analyzed crystals and aggregates. The presented spot and raster analyses were obtained in the most homogeneous volumes found after high resolution optical and BSE imaging.

MICRO-STRUCTURAL FEATURES
In the studied samples, hydrothermal epidote-group minerals occur as isolated idiomorphic crystals and as aggregates of smaller prismatic to acicular grains, sometimes arranged in a radiated form. Most of them present submilimetric dimensions and fill in interstices between the primary minerals as well as miarolitic cavities in the granite. They constitute < 0.1 vol. % of the CO-2 sample and are mainly associated with clear albite (An < 2 mol. %), quartz, dark green Al-poor and Fe-rich biotite and chlorite, fluorite, zircon, and occasionally some magnetite, hematite and galena as minute opaque crystals.
The most typical micro-structures of the epidote-group minerals in the studied granite are exemplified in Fig. 2, which depicts plane polarized optical, BSE, as well as Ca, Al, Fe, Ce, Ti, and Sr X-ray images for a single crystal and/or a crystalline aggregate. In both cases, well marked complex oscillatory zoning patterns are observed. In general, four main compositional zones, with abrupt mutual contacts, may be recognized from core to rims: the most internal zone (I) displays higher light absorption with red to dark brown (Z) and dark red to brown (X) pleochroic colors, higher backscattered coefficients and relatively higher Fe, Ce, and Ti, and lower Al and Ca contents; the next zone (II) shows lighter colors, as is typical in common allanite, lower backscattered coefficients, and lower Fe, Ce and Ti, and higher Al and Ca contents. Along the fast growing direction of the crystals, which corresponds to the b crystallographic axis, zone I presents a typical swallow-tailed contact against zone II.
The third zone (III) is more transparent and clear with orange to yellow and green pleochroic colors, as in a typical epidote, with higher X-ray intensities for Ca and, to a lesser extent, Al, and lower for Ce and Ti. The last external zone (IV) is constituted by allanite with similar properties of zone II. Sr concentrates mainly in the epidote zone, particularly in epidote filling in interstices between crystal-forming aggregates, and correlates directly with Al and inversely with Fe. In general, zones I and II are more homogeneous, while thin zoning patterns identified by Ca, Al, Ce and/or Ti intensity changes appear in the epidote and the external allanite zones (III and IV, respectively) in both isolated crystals and crystal aggregates. A thin allanite zone with higher intensities of Ce and Ti, precipitated immediately over the epidote zone III, is a common feature in the internal part of the allanite zone IV. In the most external epidote and allanite zones, crystals in the aggregates lost their chemical individuality in the images.
In some cases (not shown) isolated crystals are devoid of zones III and IV, and some aggregates do not present zone I. Rare crystal aggregates appear to be constituted only by allanite with properties similar to the ones described for zone IV. In other cases, the allanite from zone I appears to be constituted of several acicular crystals immersed in allanite from zone II, which seem to diverge from a 180 SILVIO R.F. VLACH common growing substrate, as illustrated in Fig. 2.
The zones with the highest transparency, as epidote and allanite zones II, III, and IV, are always very pristine, with birefringence between 0.3 and 0.4, and do not show radioactive damage evidence.

CHEMICAL COMPOSITIONS
Representative WDS analyses and structural formulae for epidote-group minerals from sample CO-2 are presented in Table I. The whole data set is

Main features
The main chemical characteristics of the Corupá epidote-group minerals are well described by the compositional variations of the single zoned crystal shown in Fig. 2A, as depicted in Fig. 3. This figure nicely illustrates the main chemical variations obtained closely along the b crystallographic axis, expressed in oxide wt. % of major and minor elements, as well as Fe 3+ /(Fe 3+ +Fe 2+ ) ratios, and their correlations with zones I, II, III, and IV, as discussed in the following.
The chemical compositions here presented appear to cover almost all the range between the epidote and allanite end members. Fig. 4 shows the cationic proportions of (REE + Y) in the A site as plotted against the VI M 2+ cations in the M site. The compositional distributions along the epidoteallanite join in this diagram has been used as an evidence of extensive solid solution between these end members. Our results present a very good correlation (determination coefficient r 2 = 0.94) and show, in relation to the ideal allanite end-member, a shift towards a relative excess of the VIM2+ cations in M-sites, a feature shared by most available data in the literature, for which the Fe 3+ /Fe 2+ partition was estimated (e.g. Gieré and Sorensen 2004).
The cationic substitution schemes in the epidote-group minerals were revised by Gieré andSorensen (2004), Franz andLiebscher (2004) and Armbruster et al. (2006). The entry of the rare-earth elements in the (REE+Y)-bearing members involves exchange with Ca (and Sr),  Fig. 5), with an exceptional r 2 = 1, demonstrated that at least one of these exchange vectors plays a major role in the observed compositional variations. The obtained compositions are distributed along the entire epidote-allanite range (cf. Fig. 4), but a few ones plot in the intermediate zone of this diagram, with REE+ Y values between 0.3 and 0.6 a.p.f.u.
Compositional variations in the clinozoisiteepidote-allanite-ferriallanite are well depicted in the REE + Y + Th + Sr (A site occupancy excluding Ca) vs Al T diagram (Petríck et al. 1995), which also

184
SILVIO R.F. VLACH allows to estimate the Fe 3+ /(Fe 3+ +Fe2 + ) ratios, as discussed in the next section. An expanded version, to include the end-member ferriepidote, of this diagram is shown in Fig. 6. Most of the Corupá analyses plot in the quadrangular epidote-allaniteferriepidote-ferriallanite area; exceptions are some analyses with REE + Y + Th + Sr greater than 1 a.p.f.u. and others with deficient Al in the M (2) sites, or both. Such Al deficiency points to Fe occupancies in the M (2) sites up to 0.18 a.p.f.u. (see Table I).
towards the allanite-epidote join. Among the epidote compositions, those displaced towards the ferriepidote end-member are more representative of the crystal aggregates, while compositions that plot close to the epidote-ferriallanite join were obtained almost only in single crystals. The compositions of hydrothermal epidotegroup minerals from two other granite occurrences are plotted in this diagram for comparison purposes: allanite-(Y) from the Capivari Pluton of the Graciosa Province  and epidote and allanite from the Skye Igneous Complex, Scotland (Exley 1980). The compositions of the latter define a significant compositional range as well, parallel to the [REE 3+ Fe 2+ ][CaFe 3+ ] -1 exchange vector. As a whole, the Skye compositions are richer in Al T and their plots are displaced to the right of the diagram: compositions close to the ferriallanite and ferriepidote end-members are lacking, while some analyses yet contain a small contribution of the clinozoisite endmember. This correlates with the relatively more aluminous character of the host granites, in contrast with the marked peralkaline signature of the Corupá sample. The Y-richest analyses from Skye are  Fe 3+ /(Fe 3+ +Fe 2+ ) ratios It is difficult to estimate Fe 3+ /(Fe 3+ + Fe 2+ ) ratios (Fox) from EPMA analyses, specially for the REE, Th and U relatively rich end-members of the epidote-group minerals. This is due to the possible presence of vacancies in the A-sites, of trivalent Mn, site-occupancy uncertainties, as well as chemical exchanges launched by hydrothermal alteration (Wood andRicketts 2000, Poitrasson 2002) and structural damage caused by Th and U radioactive decays.
The best estimation methods, according to Ercit (2002), are based on 6 (M+T) or 8 (A+M+T) cations and 25 negative charges, the first being superior as it does not depend on A-site vacancies. However, the second method was indicated by the IMA, taking into account the data obtained by Bonazzi et al. (1996), which seems to indicate some Mn 2+ and, to a lesser extent, Fe 2+ filling in M (1) , M (3) and A (1) sites (cf. Armsbruster et al. 2006, Franz and Liescher 2004, Gieré and Sorensen 2004. It was nevertheless emphasized that this method does not work well when there are A-site vacancies and/or if the analyses are not complete. Fox ratios should also be estimated from (REE+Y+Th+Sr) and Al T relations in the subsystem clinozoisite-epidote-allanite-ferriallanite (cf. Fig.  6), as graphically depicted in Petríck et al. (1995) and algebraically wrote by Poitrasson (2002). The first authors showed that the obtained results were

186
SILVIO R.F. VLACH in well agreement with independent Mössbauer determinations. This method may be applied to overcome minor problems related to analytical errors on Si and Fe and incomplete analyses, as well as to some late chemical transformations not affecting the involved elements. There are cases, however, in which minerals in this subsystem contain substantial Ti amounts, a feature typical of allanite and ferriallanite from A-type granites and related rocks, and the results will be overestimated . A graphical comparison of the Fox ratios estimated with these methods for the obtained data is presented in Fig. 7. The ratios computed on the basis of 6 (M+T) and 8 (A+M+T) cations are equivalent within the whole compositional range and, thus, incompatible with the presence of significant Fe 2+ or Mn 2+ contents, as well as vacancies, in the A site. The correlation between them and the ratios computed according Petríck et al. (1995) is always poor, as the epidote-ferriepidote compositions, related by the [Fe 3+ ][Al] -1 exchange vector, are outside the considered subsystem and our allanite-ferriallanite compositions present variable but significant Ti contents, incorporated through the coupled [Ti 4+ +Fe 2+ ][Al 3+ ] -2 exchange vector .
The available data plotted in Fig. 8 were used to devise a correction for the Ti effect on the Fox estimates obtained by the method of Petríck et al. (1995), which may be written as follows: where Fox C is the corrected and Fox P is the Petríck et al. ratio, respectively.

REE AND TRACE ELEMENTS PATTERNS
The REE + Y contents in the hydrothermal Corupá ferriallanite-(Ce) amount to up 1.05 a.p.f.u. (26.6 wt. %), while the equivalent contents in epidote-ferriepidote

188
SILVIO R.F. VLACH the epidote-group minerals, as well as for the host granite, are presented in Figs. 9A and 9B. The results obtained through both methods are in good agreement in the range of the L-MREE, which is measurable with EPMA. As expected, the epidote and ferriepidote EPMA-derived patterns are more irregular and subjected to higher analytical uncertainties due to the relatively low REE and Y contents. The patterns for the ferriallanite-rich compositions show the highest fractionation degrees, with La N /Nd N , Sm N /Y N (all data), and Er N /Lu N (LA-ICP-MS data) ratios varying between 3-4, 40-80 and 2-6; typical allanite has relatively lower fractionated patterns, with ratios between 2-3, 20-60 and 1-2, respectively. Epidote and ferriepidote compositions present the lowest fractionation degrees, with La N /Nd N and La N /Y N ratios close to unity. Negative Eu anomalies are well developed in all cases, with 0.02 < Eu/Eu* < 0.05 [Eu* = (Sm N *Gd N ) 0.5 ], the lower values appearing only in ferriallanite. It should be noted that allanite compositions from zone IV, external to the epidote zone, as well as most compositions observed in aggregates made up only of allanite, are significantly richer in MREE (see Fig. 9A), with La N /Nd N and Sm N /Y N ratios between 2-3 and 20-40, respectively.
The REE whole rock pattern of the host peralkaline granite (Fig. 9B) is also characterized by a higher fractionation degree in the LREE side. Normalized La N /Nd N , Sm N /Ho N and Er N /Lu N ratios are 2.4, 1.5, and 1.0, respectively, with Eu/ Eu* = 0.04. Its LREE pattern is controlled mainly by primary chevkinite and the hydrothermal ferriallanite and allanite discussed in this work. Albeit the higher L-MREE chevkinite contents, when compared to ferrriallanite, their patterns are fractionated to a similar degree . The M-HREE whole rock contents depends on allanite (particularly from zone IV), primary and hydrothermal zircon, as well as late titanite patterns, a theme to be explored in a next contribution.
Other trace element patterns are emphasized in the multi-element diagram depicted in Fig.  10. Relatively to the whole rock abundances, ferriallanite is enriched ( > 10-fold) in Ti, Sr, and Ga (up to 90 times) and depleted in Mg, Rb, Th and Zr (by a factor of 0.002-0.02). Ga and Mn contents in allanite are lower than in ferriallanite, while Zr, Ta, Nb and U are higher. Relative to ferriallanite and allanite, epidote is enriched in U, Sr (up to 2000-fold) and depleted in Nb, Pb, Zr, Hf, Ti, Zn and Ga. As it is well known, epidote concentrates significant amounts of Sr (e.g. Frei et al. 2004). The measured Pb contents (50 -430 ppm) in the Corupá epidote-group minerals are in great excess in relation to the expected radiogenic Pb generated by  decays. This puts some constraints on allanite U-Pb dating. Sn, not represented in Fig. 10, occurs in significant amounts, between 90 and 375 ppm in all these minerals (Table II).
suggests a relatively fast grow. This ferriallanite was overgrown by common allanite. The external part of the epidote-ferriepidote zone and the most external zone IV constituted by allanite display a somewhat convolute-like zoning pattern. These zones and all internal parts of the epidoteferriepidote zone III are overprinted by thin regular and recurrent compositional zones. The external zone IV is more homogeneous and continuous. Relatively Sr-rich epidote appears filling in interstices among common epidote crystals. This complex pattern suggests contrasted crystallization conditions or rates.
The compositional variation range observed in the hydrothermal epidote-group minerals covers a significant range along the allanite-ferriallanite and epidote-ferriepidote joints. A compositional gap may exist, however, between the more typical compositions of allanite-ferriallanite and epidoteferriepidote in the studied case. In fact, despite the high spatial resolution of the EPMA, some results could represent mixed analyses, a combination of very thin contrasted compositional domains. It must be emphasized that both BSE and compositional images show abrupt and sharp contacts between the allanite and epidote zones. This is intriguing, as literature data point to a compositional continuous between these end-members (e.g. Deer et al. 1986, Gieré andSorensen 2004), as also indicated by Fig. 3.
The REE patterns of allanite from the internal zones I and II show a higher fractionation of LREE over HREE, while allanite from the external zone IV has a somewhat less fractionated pattern and/ or contains higher MREE abundances. Epidote and ferriepidote present lower fractionation degrees and much lower LREE, and similar or somewhat higher HREE amounts as compared with allanite.
These compositional variations and the recurrence of allanite in the zoning patterns should be a response to local fluctuations in the concentrations of Ca, Al, Fe 3+ and REE, among other minor elements, due to their preferences for crystallizing

MICRO-STRUCTURAL AND COMPOSITIONAL VARIATIONS
The BSE and the compositional X-ray images demonstrate complex zoning patterns in hydrothermal single crystals and crystal aggregates of epidote-group minerals in the Corupá peralkaline granite. In several crystal aggregates, zone I ferriallanite is made of irradiating, thin acicular crystals with swallow-tail terminations, which Symbols as in Fig. 9. hydrothermal phases or for the remaining fluid and to chemical refreshment by recurrent fluid input. The higher contents of Ce and Ti in the internal allanite zone IV, as compared with the external zone II, indicate significant oversaturation of these elements in the fluid phase prior to the reprecipitation of allanite after epidote-ferriepidote. The REE contrasted patterns may also suggest some kind of structural control on the LREE and HREE A-site occupancies, allowing the preferential entry of HREE in the Sr-, Fe 3+ -and Al-rich end-members (cf. Bonazzi and Menchetti 1995).

THE HYDROTHERMAL ENVIRONMENT AND REE BEHAVIOR AND MOBILITY
Epidote-group minerals present a very large stability field (e.g. Franz andLiebscher 2004, Gieré andSorensen 2004), and the crystallization environment of the Corupá peralkaline granites is poorly constrained in terms of T, P and ƒ volatiles intensive parameters. Geological and petrographical evidences, as well as some geobarometric estimates available for related plutons within the province, indicates magmatic emplacement and crystallization in lower-crustal levels . Inferences based on experimental data (e.g. Scaillet and MacDonald 2001) suggest possible solidus temperatures for the granites about 670°C and, thus, an upper temperature limit for the hydrothermal environment. The occurrence of magnetite, hematite and galena in the hydrothermal assemblage points to relatively high ƒ O2 values, close to the magnetite-hematite (HM) buffer, under S 2saturation conditions.
The widespread distribution of hydrothermal allanite and epidote, as well as zircon and fluorite, in the studied peralkaline granites demonstrates the HFSE-and F-enriched nature of the postmagmatic fluids. A common feature related to the evolution of peralkaline liquids in the Graciosa Province and elsewhere is the abrupt change in their chemical signature as they cross from magmatic to post-magmatic conditions (Salvi et al. 2000, Graser and Markl 2008, Vilalva and Vlach 2010, a theoretical limit not easily recognized in real rocks. In general, magmatic evolution by crystal fractionation promotes an increase in alkalinity and accentuates the peralkaline character observed in mineral and/ or whole-rock compositions, while in the postmagmatic stage a sudden decrease in the alkalinity is observed, accompanied by the development of Ca-, Al-and F-bearing mineral assemblages, with relatively high REE contents. In the Corupá Pluton there is no clear evidence pointing to an external fluid input with the appropriate (HFSE-rich) composition; however a meteoritic contribution to the volatile constituents can not be disregarded.
The increase in the relative abundance of the HFSE in late-to post-magmatic fluids is a well-known phenomena in such environments, as these elements are relatively soluble in Ca-poor, F-bearing peralkaline liquids and fluids (e.g. Salvi and Williams-Jones 1996, Salvi et al. 2000. In these media, the Ca contents of the residual fluids seem to play a fundamental role (Salvi and Williams-Jones 1990, 1996, Boily and Williams-Jones 1994, and even low Ca contents lead to the crystallization of fluorite and a fast decrease of the F activity in the fluid phase. As a consequence, the HFSE solubility decreases and they are preferentially incorporated into the mineral structures. The main primary repository mineral of LREE in syenites and granites from the Corupá Pluton is chevkinite, a typical accessory phase with REE contents up to 49 wt. % (REE) 2 O 3 , which is very susceptible to post-magmatic alteration . Among the related chemical transformations, the authors emphasized a significant increase in Ti and a decrease in total Fe and LREE, indicating that these latter elements moved preferentially to the volatile phase. Importantly, REE EPIDOTE-GROUP MINERALS IN PERALKALINE GRANITE, GRACIOSA PROVINCE and specially LREE fractionation patterns observed for primary and altered chevkinite are very similar and so must be the pattern acquired by the fluids. It is worth to note also that REE-and F-rich hydrothermal phases are almost absent in the largely dominant metaluminous alkali-feldspar syenites and quartz syenites, with post-magmatic assemblages made up mainly of actinolitic amphibole, phyllosilicates, stilpnomelane, albite and quartz. This indicates that LREE released from chevkinite from syenites concentrates also in the latest fluids percolating granites. Primary chevkinite in syenites and granites is thus a suitable REE source, and its alteration should contribute to the LREE-enriched nature of the late fluids in the Corupá Pluton.

CONCLUSIONS
The hydrothermal crystallization of epidote-group minerals in peralkaline alkali-feldspar granites from the Corupá Pluton, South Brazil, indicates that the late fluids were relatively rich in HFSE, particularly the LREE. These higher LREE amounts were probably maintained in solution in Ca-poor and F-rich fluids, until the precipitation of fluorite in the evolved granites and consequent decrease in their solubility. The LREE, and also other HFSE, were concentrated mostly by magmatic fractionation in a peralkaline environment, as well as released through the hydrothermal alteration of early formed primary chevkinite. In this sense, these hydrothermal minerals are related mostly to an isochemical redistribution of the REE, specially LREE, during the post-magmatic crystallization stages of the Corupá Pluton.
The remarkable micro-structural and compositional zoning patterns observed in the epidote-group minerals point to contrasting growth conditions and significant compositional variations in the evolving hydrothermal fluid, under oxidizing conditions. The observed variations in the REE fractionation patterns indicates that fluid compositions evolved progressively towards higher MREE/LREE and/or HREE/LREE ratios.