Stable ( C , O , S ) isotopes and whole-rock geochemistry of carbonatites from Alto Paranaíba Igneous Province , SE Brazil

2Faculdade de Ciências e Tecnologia, Universidade Federal de Goiás – UFG, Goiânia (GO), Brazil. E-mail: jabrod@gmail.com, tcjbrod@gmail.com, elisa.barbosa@uol.com.br 3Centro Regional para o Desenvolvimento Tecnológico e Inovação, Universidade Federal de Goiás – UFG, Goiânia (GO), Brazil. E-mail: jabrod@gmail.com 4Instituto de Geociências, Campus Universitário Darcy Ribeiro, Universidade de Brasília – UnB, Brasília (DF), Brazil. E-mail: lucieth@gmail.com, rventura@unb.br 5CICTERRA, Facultad de Ciencias Exactas, Físicas y Naturales, Universidad Nacional de Córdoba, Córdoba, Argentina. E-mail: ipetrinovic@yahoo.com 6Laboratório de Estudos Geodinamicos e Ambientais, Campus Universitário Darcy Ribeiro, Universidade de Brasília – UnB, Brasília (DF), Brazil. E-mail: lmancini@unb.br


INTRODUCTION
Brazilian alkaline provinces have been studied by a range of techniques such as whole-rock geochemistry, mineral chemistry, stable and radiogenic isotope geochemistry, with a range of petrogenetic aplications (Gomide et al. 2013, Barbosa 2009, Cordeiro et al. 2011, Grasso 2010, Ribeiro 2008, Comin-Chiaramonti et al. 2001, 2005, Traversa 2001, Andrade et al. 2002, Huang et al. 1995, Santos & Clayton 1995, Ulbrich & Gomes 1981).We aim to contribute to the knowledge of carbonatite petrogenetic evolution, investigating the relationships between whole-rock geochemistry and stable isotope geochemistry data, reporting geochemical and isotopic data from carbonatites of the Late-Cretaceous Alto Paranaíba Igneous Province (APIP), and discuss their implications for magma evolution, both at single-complex and Province-wide scales.The results obtained are compared with the Early-Cretaceous Jacupiranga carbonatite complex, in the Ponta Grossa Province.
Alkaline rocks and alkaline-carbonatite associations include highly variable petrographic types and correspondingly extensive nomenclature.We adopt the nomenclature proposed by Le Maitre et al. (2002) for carbonatites and rocks of the ijolite series, Sahama (1974) for kamafugites, Yegorov (1993) for phoscorites, and Barbosa et al. (2012) for bebedourites.
The APIP alkaline-carbonatite complexes are multistage intrusions formed by rocks derived from the bebedourite, carbonatite, and phoscorite series, which are related to each other by a complex interplay of fractional crystallization, liquid immiscibility, and degassing (Brod et al. 2004, Ribeiro 2008, Barbosa 2009, Cordeiro et al. 2010, Barbosa et al. 2012, Brod et al. 2013).
The Province was established along an elongated NW-SE structure called the Alto Paranaíba Arc.The alkaline magmas were emplaced into Precambrian rocks from the internal and external zones of the Brasilia Fold Belt.Kamafugite is by far the dominant rock type in the province, forming one of the few known kamafugite-carbonatite associations (Brod et al. 2000).
The ultramafic rocks of all complexes show variable degrees of metasomatism by fluids resulting from carbonatite differentiation.The relatively shallow characteristics of the APIP intrusions are indicated by C and O isotopes (Santos & Clayton 1995), within-magma chamber pyroclastic deposits (Ribeiro et al. 2005), similarities between bebedourites and xenoliths in volcanic and subvolcanic kamafugites in the province (Seer & Moraes 1988, Brod et al. 2000), and the extent of degassing/metasomatism.
Serra Negra is the largest APIP carbonatite complex with an area of 65 km 2 .It intruded quartzites from the Canastra Group, generating a very pronounced dome structure.The complex is composed by dunites, bebedourites, calciocarbonatites, magnesiocarbonatites, trachytes.
The Araxá Complex is composed of carbonatites, phoscorites, and metasomatic phlogopitites derived from ultramafic rocks.Magnesiocarbonatite is the dominant carbonatite type in this complex.The intrusion generated a dome structure in schists and quartzites from the Ibiá Group (Seer 1999).
Tapira is an approximately elliptic complex, composed of bebedourite with subordinate carbonatite and syenite, and rare melilitolite and dunite, all cut by ultramafic dikes of kamafugite affinity.Calciocarbonatite dominates over magnesiocarbonatite (Brod et al. 2003).

MATERIALS AND METHODS
Analyzed samples were first examined under a petrographic microscope to determine mineralogical and textural characteristics.Selected samples were ground in an agate mill and sent to the ACME Labs for whole-rock chemical analysis (ACME 4A and 4B packages).Powders produced from whole-rock and/or mineral separates were analyzed for C and O isotopes using a Delta V plus gas source mass spectrometer at the University of Brasília.The S isotopes data are from Gomide et al. (2013).

RESULTS AND DISCUSSION
The studied APIP carbonatites were classified into five groups on the basis of their modal composition, petrographic characteristics, and interpretations and evolution models from the previous works (Barbosa 2009, Brod 1999, Cordeiro 2009, Cordeiro et al. 2010, Cordeiro et al. 2011, Grasso 2010, Palmieri 2011, Ribeiro 2008, Ribeiro et al. 2014).The key minerals modal compositions are shown in Figure 2, and the geochemical and textural characteristics, and the evolution stage of each group are summarized in Table 1.
Early-stage carbonatites (C1) are rich in apatite, phlogopite, and magnetite, but all the three phases tend to diminish or disappear at the intermediate stages (C2 or C3) resuming crystallization in later stages (C4 and C5).Barite is typically lacking in the initial stages, appearing for the first time at calciocarbonatites C3 and becoming more abundant toward later stages.
Monazite is restricted to the most evolved carbonatites.Calcite is the dominant carbonate phase at C1 and C3, while dolomite dominates in C2, C4, and C5 groups.Strontianite and norsethite appear initially as liquidus phases in calciocarbonatite C3 and increase in abundance toward more evolved carbonatites.Barytocalcite is present in C3, but only locally, and restricted to this group.

WHOLE-ROCK GEOCHEMISTRY
Whole-rock chemistry data were plotted in the carbonatite classification diagram of Woolley and Kempe (1989).Carbonatites from groups C1 and C3 are plotted in the field of calciocarbonatite, whereas carbonatites from groups C2, C4, and C5 are plotted in the field of magnesiocarbonatites (Fig. 3).
The C1-type calciocarbonatites (Fig. 3) are present in Catalão II, Salitre, Tapira, and Jacupiranga.Mineralogically, they are unevolved, containing large amounts of silicates such as olivine, amphibole and Al-rich phlogopite, carbonate minerals restricted to calcite and possibly dolomite, and significant amounts of apatite and magnetite.
The C3 calciocarbonatites (Fig. 3) were identified only in samples from Tapira and differ from the C1 calciocarbonatites because they contain primary norsethite, baritocalcite, and tetraferriphlogopite instead of aluminous phlogopite, indicating that C3 are a more evolved version of calciocarbonatite.The presence of Sr-, Ba-, and rare earth element (REE)-rich exsolutions in calcite suggests crystallization of the latter at relatively high temperature.Similar characteristics were observed by Brod (1999) for Tapira carbonatites and by Cordeiro et al. (2011) for Catalão I.
The C2-type magnesiocarbonatites (Fig. 3) are present in Catalão I, Serra Negra, Salitre, and Jacupiranga.These rocks have a simple primary carbonate composition, restricted to dolomite and calcite.The presence of abundant exsolutions indicates that these carbonates crystallized at relatively high temperature.This group of magnesiocarbonatites is little differentiated, but may not represent the most primitive magnesiocarbonatite in the province, since magnesiocarbonatites with olivine are described in Salitre and Serra Negra (Barbosa 2009, Grasso 2010).
The C5-type magnesiocarbonatites studied in this work are present in Catalão I, Catalão II, Araxá, and Tapira.They are enriched in Ba and characterized by a wide variety of carbonates, including dolomite, Fe-dolomite, norsethite, burbankite, calcite, and strontianite.
The determination of the evolution stage in carbonatites is a difficult task, because the concentration of elements typically used to monitor differentiation in common magmas may be strongly modified by the crystallization of specific phases in carbonatite magmas.For example, MgO is an essential constituent of olivine, which is an early-stage phase in the carbonatite magma, but also in dolomite, which may be a late-stage mineral. in silicates and should decrease with the evolution of carbonatite magma, as silicates are mostly crystallized at the early stages.However, the concentration of these elements is generally very low in carbonatites and their use as a differentiation index is limited for the most part of the evolution range.
There is a literature consensus (e.g.Buhn & Rankin 1999, Chakhmouradian et al. 2008, Le Bas & Handley 1979, Zaitsev et al. 1998, Jones et al. 1996, Xie et al. 2009) that Ba and REE tend to be enriched toward the final stages of evolution in the carbonatite magma.Also, apatite is the most persistent mineral in the carbonatite evolution range, and P 2 O 5 contents can be a useful tool to measure the differentiation degree, except perhaps in the final stages, where phosphorus may be associated with the late-stage crystallization of monazite.
Figure 3 shows P, Ba, and REE variation for different carbonatite compositional groups (see Table 1), indicating the progress of each group based on mineral paragenesis and overall chemical composition.The groups C1, C2, C3, and C5 evolved mostly by enrichment in BaO and depletion in P 2 O 5 , while the group C4 contains carbonatites with higher concentrations of REE.
Figure 4 shows a general evolution by increase in Ba, Sr, and REE, where C5 is located in an intermediate position, in the case of REE.

MAJOR ELEMENTS
The evolution sequences inferred from the mineral paragenesis and chemical changes observed in Figures 3 and 4 were used to establish a sequential arrangement of samples in each group (see suplementary data in Appendix 1) .The mineralogy present in these samples was then used to search for an index that could proxy for magmatic evolution of the studied carbonatites.Figures 5 to 11 show the behavior of several major element oxides with BaO/(BaO+SrO) as an evolution index.
The C1 and C2 carbonatites from Jacupiranga (Fig. 5) show an evolution similar to that of the corresponding, most primitive carbonatite group of the APIP (Fig. 6).C1a and C1b from Jacupiranga largely overlap in the region with low values of BaO/(BaO+SrO), not exceeding 0.24, as expected for primitive compositions.The C1a cumulates show some positive correlation of BaO/(BaO+SrO) with REE and Na 2 O resulting from apatite accumulation, whereas the mostly residual C1b carbonatites show the opposite behavior as a result of apatite removal.CaO and MgO present an erratic behavior.
The APIP C1 calciocarbonatites (Fig. 6) evolve with an increase in BaO and CO 2 , indicating the increased amount of carbonate in the rock.The BaO content is very low relative to other APIP carbonatites, reaching a maximum at around 0.5%.K 2 O, REE, and P 2 O 5 decrease with C1 evolution.Both REE and P 2 O 5 are controlled only by apatite fractionation at these early stages, while K 2 O is controlled by phlogopite fractionation.CaO and MgO have mutually opposite behaviors, although with considerable scattering, and their distribution is probably controlled by independent factors, such as the removal of these elements from magma during fractionation of apatite and silicates, and its accompanying increase in carbonate enrichment.Furthermore, the opposite CaO and MgO trends may be associated with the presence of small, but varying amounts of dolomite in the rock.The sodium content is usually very low and independent of the stage of evolution of these rocks.One extreme Na 2 O content of about 0.8% can be explained by the presence of amphibole.The APIP C2 magnesiocarbonatites (Fig. 7) are characterized by the increase in BaO with evolution, although the levels of Ba are still relatively low when compared with other magnesiocarbonatites.The CaO content is insensitive to evolution, reflecting the scarcity or absence of calcite in these rocks.The P 2 O 5 content is low (maximum 0.2%), but systematically decreases, signaling apatite fractionation.The similar behavior of SrO and Na 2 O is probably also related to the apatite removal.It is important to note that, unlike C1, the calcite is absent or very rare in C2 carbonatites, which implies a control of the SrO content by fractionating apatite, since strontium does not enter the structure of dolomite easily.
In our studied sample set, the C3-type carbonatites are restricted to samples from the Tapira Complex.In this group, there is no obvious control of fractionating apatite on the content of P 2 O 5 , which shows erratic variation (Fig. 8).BaO, MgO, Na 2 O, and, less obviously, REE increase with magmatic evolution whereas CaO and SrO decrease, suggesting that calcite crystallization gives way to other carbonates such as norsethite (BaMg(CO 3 ) 2 ) and, more rarely, baritocalcite (BaCa(CO 3 ) 2 ) with magma evolution.The atypical behavior of CO 2 , which increases slightly at first and then constantly decreases, is possibly related to the same mineralogical effect, because carbonates containing heavy elements such as Ba, Sr, and REE will naturally have a lower CO 2 proportion by weight.The BaO (up to 13%) and SrO (up to 5%) levels are substantially greater than that in C1 and C2 carbonatites.The contents of K 2 O are low and decrease very quickly in the early evolution of C3, indicating total consumption of phlogopite as the silicate formed at this stage.The C4 magnesiocarbonatites occur in Salitre, Araxá, and Tapira, and are characterized by the presence of monazite, instead of carbonates, as the main REE mineral.This fact is illustrated by the high REE content (up to 2.5%, Fig. 9) in carbonatites of this group and confirmed by SEM and electron microprobe (EPMA) analyses.In addition to REE, there is an increase in K 2 O, Na 2 O, and MgO, followed by an important decrease in the concentration of BaO, CaO, and CO 2 with magma evolution.A notable characteristic is the negative correlation between BaO and REE, which is not observed in the other groups.
The C5 magnesiocarbonatites are from Catalão I, Catalão II, Araxá, and Tapira, and show increase in BaO, SrO, Na 2 O, and CO 2 , and decrease in CaO and K 2 O with evolution (Fig. 10).Some of the less evolved samples of this group have very high P 2 O 5 (up to 8-9%) and phosphorus contents are strongly controlled by the apatite fractionation at the early stages.This fact contrasts with the low phosphorus observed in C2, and suggests that carbonatites from C5 represent a distinct magmatic pulse.The strong increase in BaO, SrO, and Na 2 O indicates the crystallization of rare or complex carbonates such as norsethite, burbankite, and strontianite, whose occurrence is confirmed in electron microscopy and electron microprobe analyses.The replacement of dolomite by these other carbonate minerals as the main crystallizing phases in the magma explains most of the observed chemical changes, including a consistent decrease in CaO, but does not explain the progressive increase in CO 2 .At the current stage, the reasons for this behavior are not clear.
Applying the same classification concepts to APIP carbonatite analyses from literature (Grasso 2010, Brod 1999, Traversa et al. 2001, Palmieri 2011, Barbosa 2009  Figure 5. Behavior of selected major element oxides with magma evolution for the Jacupiranga C1 calciocarbonatites and C2 magnesiocarbonatites. Symbols used are same as in Figure 3. Araújo 1996, Morbidelli et al. 1997, Machado Junior 1992), it was possible to build compositional fields for the province.The general trends for the province (Figs.11 and 12) are similar to those described previously for our samples, such as increase in BaO and CO 2 , and decrease in K 2 O and P 2 O 5 with evolution.Possible exceptions are some C1a apatite-rich cumulate rocks with anomalously high P 2 O 5 , associated with the fractionation of apatite, but not clear for the province as a whole.The C2 samples from this work are plotted inside the APIP carbonatite fields (Fig. 12), except for the CO 2 and SrO contents of some samples that cover slightly larger areas.In general, the classification fits well, maintaining the trends described for this group, such as increase in BaO and REE contents, whereas CaO is insensitive to evolution, and P 2 O 5 content is low.
Samples from the C3 group (Fig. 11) fit very well to the province fields.BaO, MgO, and Na 2 O increase whereas CaO, SrO, K 2 O, and CO 2 decrease in the entire province, following the behavior described for this group from our samples.The P 2 O 5 content is low and insensitive to magma evolution.
The C4 group samples (Fig. 12) plot mostly inside the province fields or following the same trend.Available data for carbonatites with C4 characteristics in the whole province are scarce generating a restricted field, but even in this case the samples from our data set and from the whole province have very similar chemical behavior which strengthens the efficiency of the devised classification scheme.The whole province C4 trends are marked by high REE content, an increase in K 2 O and Na 2 O, and a decrease in the concentration of BaO, CaO, and CO 2 .
The C5 group shows only partial coincidence between the compositional range of our data set and the whole province, suggesting that our samples have a more extreme composition, particularly for BaO.However, even when the fields do not coincide, the two trends are similar.This group shows an increase in BaO, Na 2 O, and CO 2 contents, and a decrease in SrO, CaO, REE, CO 2 , and K 2 O with magmatic evolution.

TRACE ELEMENTS
Trace elements in the carbonatite magma may be greatly affected by specific late-stage fractionating phases such as sphene, apatite, perovskite, monazite, or zircon (e.g.Nelson et al. 1988).Pyrochlore and baddeleyite, which are additional common phases in our sample set, may also have a great effect on some trace elements such as the high field strength elements (HFSE and REE).Liquid immiscibility is another important process affecting trace element distribution in carbonatites and associated rocks, including changes in geochemically similar element pairs such as Nb-Ta, Zr-Hf, and REE (e.g.Hamilton et al. 1989, Veksler et al. 1998, 2012, Brod et al. 2013).
La (n) /Yb (n) tends to increase with magma evolution (Fig. 13) within the same carbonatite group and, in most cases, between different groups.Group C5 is an exception, since most samples have lower La (n) /Yb (n) for high BaO/(BaO+SrO), U (n) /Th (n) decreases with evolution in C1 group with significant fractionation, which may reflect the crystallization of pyrochlore.
The Nb (n) /Ta (n) ratio show complex behavior, and its variation shown in Figure 13 does not seem to be affected by fractional crystallization, since the variability of the Nb/Ta ratio is far greater than that of the BaO/ (BaO+SrO).Brod et al. (2013) have shown that liquid immiscibility in the Tapira complex of the APIP resulted in the increase of the Nb (n) /Ta (n) ratio of the carbonate conjugate, sometimes by several orders of magnitude, if compared with the same ratio in the parental liquid.In our data set, most samples from C1a and some samples from C5 show values near 1.3, which is the Nb (n) /Ta (n) ratio of the parental magma of the APIP complexes (Brod et al. 2013).For the other samples, the high values of Nb (n) /Ta (n) ratio suggest that they have been involved in liquid immiscibility at some point in their evolution, according to

MULTIELEMENT DIAGRAMS AND RARE EARTH ELEMENTS
Figure 14 shows chondrite-normalized multielement diagrams (Thompson 1982).The C1 carbonatites, both from Jacupiranga and the APIP, may be divided into ■ carbonatites containing cumulus apatite, characterized by a positive anomaly in phosphorus and ■ residual carbonatite from fractional crystallization processes, which have a negative P anomaly.
According to this criterion alone, most C1b carbonatites from Jacupiranga should be considered apatite-rich cumulates.However, the behavior of other major and trace elements, as well as the distinct REE patterns between Jacupiranga C1a and C1b, supports the proposed division.In Jacupiranga, the chondrite-normalized Nb/Ta are typically lower than 1 in C1 and higher than 1 in C2, although the differences are relatively small.In the APIP data set, C1b shows values of chondrite-normalized Nb/ Ta significantly higher than most C1a samples, suggesting that liquid immiscibility may have played a role at a differentiation stage as early as C1b very early differentiation stages.
The two C1 subgroups may be recognized in chondrite-normalized rare earth elements diagrams (Fig. 14).The apatite-rich cumulates show higher REE concentrations, especially from medium and light REE, indicating that apatite is a major REE carrier at this stage in both Jacupiranga and the APIP.The La (n) /Lu (n) ratio of C1a from Jacupiranga is in the range of 52 -92, whereas in C1b this ratio is in the range of 22 -46.In the APIP early-stage carbonatites, La (n) /Lu (n) ranged from 100 to 250 in C1a and from 35 to 140 in C1b.
In the C2 magnesiocarbonatites, the Jacupiranga samples show a positive P anomaly, indicating apatite accumulation, whereas the APIP samples always show a negative P anomaly, indicating that they are residual carbonatites.Regarding the chondrite-normalized Nb/Ta ratio, the Jacupiranga C2 shows values close to or lower than 1, suggesting that liquid immiscibility was not involved in their generation.Only two APIP samples have available Ta data.Both presents relatively high Nb/Ta, reaching up to 19, indicating that they are the product of immiscible liquids.
La (n) /Lu (n) ratio of the C2 magnesiocarbonatites ranges from 136 to 362 in the APIP and from 41 to 50 in Jacupiranga.Both C1 and C2 group in the carbonatites of the APIP have total contents of REE higher than their equivalents in Jacupiranga, suggesting that the REE enrichment is a Province-related characteristic.
The C3 calciocarbonatites show a negative P anomaly and high chondrite-normalized Nb/Ta ratio (31-53), which indicates a relatively evolved carbonatite generated by liquid  immiscibility and that underwent apatite fractionation.Their La (n) /Lu (n) ratio ranges between 178 and 354, similar to C2 and slightly higher than C1.The C4 magnesiocarbonatites exhibit a Nb (n) /Ta (n) ratio ranging from 5 to 40 and a negative P anomaly indicating the involvement of liquid immiscibility and the fractionation of apatite at an earlier stage of magmatic evolution.The samples of this group also have a slight negative Sr anomaly, which may be explained by fractionation of Sr-rich carbonates, such as strontianite or even Sr-rich calcite at an earlier stage.Alternatively, this characteristic may be an artifact of the monazite enrichment observed in C4.The REE diagram (Fig. 14) shows a strong LREE/HREE fractionation (La (n) /Lu (n) values between 520 and 1515).
The magmatic evolution of the C5 group also involved immiscibility, in most cases, with the Nb (n) /Ta (n) ratio reaching a maximum of 37. Part of the C5 sample set shows negative P anomaly, indicating prior apatite fractionation.The REE diagram (Fig. 14) shows a very strong LREE/HREE enrichment, the La (n) /Lu (n) ratio ranging from 180 to 1635.

STABLE ISOTOPES (C, O, S)
Figure 15 depicts the C, O, and S stable isotope composition.The C and O data were obtained from carbonates, using a Delta V Plus gas source mass spectrometer at the Geochronoly Lab, University of Brasília, after 1 hour reaction, using Gas Bench at 72°C, with phosphoric acid.No correction was used because at this temperature there is no need to do it (Spotl & Vennemann 2003).The C isotopic data were obtained by Gomide et al. (2013) in sulfides.The values are expressed in δ notation per thousand (see calculation in chapter 1) compared to the Vienna Pee Dee Belemnite (V-PDB) reference standards for C, Standard Mean Ocean Water (SMOW) for O, and Vienna Canyon Diablo Troilite (V-CDT) for S. 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 The Jacupiranga samples show a C and O isotopic composition that evolves according to Rayleigh fractionation.All Jacupiranga samples plot inside the range are defined as mantle values (Taylor et al. 1967).
Similarly to Jacupiranga, the C1 from the APIP show a C and O isotopic evolution by magmatic fractionation and all the samples have values consistent with the mantle range.The C2 also show isotopic evolution consistent with magma fractionation, but reach higher δ 18 O values, slightly exceeding the mantle range.Some C2 samples suggest hydrothermal alteration trends, with a shift to high δ 18 O values at relatively constant δ 13 C.
C3 follow a magmatic fractionation trend, but one sample has a shift to higher δ 18 O values suggesting that it has been affected by hydrothermal alteration.
The C4 exhibit two trends, both starting within the mantle composition range.One trend evolves into heavier C and O isotopic compositions, typical of magmatic fractionation or of interaction with fluids rich in both H 2 O and CO 2 , whereas the other evolves by increasing δ 18 O at relatively constant δ 13 C, indicative of hydrothermal alteration.
In that model, Jacupiranga carbonatites crystallized in a deeper-seated magma chamber are less evolved and have more restricted isotopic compositions.The APIP carbonatites would have been emplaced in much shallower chambers, the lower lithostatic pressure allowing for a greater variety of differentiation processes.Our data are consistent with the interpretations of Santos and Clayton (1995).We interpret the APIP C1, C2, and C3 carbonatites as representing less evolved magmas, while C4 and C5 formed at later stages and therefore were more susceptible to processes such as degassing and subsequent hydrothermal and carbohydrotermal alteration.However, some interaction with external fluids cannot be ruled out, especially in the cases of δ 18 O-only variation.
Regarding the coupled variation in S and O isotopes, the Jacupiranga samples plot in a very restrict interval, compatible with mantle composition.The small composition ranges do not allow the definition of conclusive trends.
In the APIP samples, all groups show S isotopic composition compatible with the mantle range, except for a few C5 samples showing textural evidence of sulfur degassing and/or hydrothermal alteration (Gomide et al. 2013).
possibly affected by hydrothermal alteration, the sulfur isotopes appear to be unaffected by this process.The C4 samples with petrographic evidence of sulfur degassing (Gomide et al. 2013) are aligned along two positive correlation lines in the δ 34 S-δ 18 O diagram.Similar correlations are observed in the other groups (C1a, C3, and C4), and it is possible that this feature is a sulfur degassing characteristic.However, the joint behavior of S and O isotopes still needs to be studied in more detail, because the relationship between these two systems probably is not trivial.For example, the two elements participate in multiple minerals (carbonates, oxides, silicates, sulfates, and sulfides) may be degassed as distinct species (CO 2 , H 2 S, SO 2 ) at different times during magmatic evolution and are sensitive to variations in the oxidation state of the system.

CONCLUSIONS
A major difficulty in the study of carbonatite magmatism is to establish a numerical parameter for gauging magmatic evolution, analogous to differentiation indexes used in the study of silicate magmas, against which to measure other geochemical, isotopic, mineralogical, and textural properties.This work proposes an integrated method of assessment of the magmatic differentiation stage in carbonatites of the APIP.The studied carbonatites were classified into the successively more evolved groups C1 to C5.The evolution of groups C1, C2, C3, and C5 may be monitored by BaO enrichment and P 2 O 5 depletion, whereas C4, a group of REE-rich carbonatites, evolves through enrichment of both rare earths and phosphorus.BaO/(BaO+SrO) Figure 13.Trace elements behavior of some key chondrite-normalized elemental ratios for this work Alto Paranaíba Igneous Province samples compared with all province data.
In this work, we devise a differentiation index involving BaO and SrO, applied it to our sample set and to data for the APIP carbonatites from the literature.
The evolution of C1 calciocarbonatites results in increased calcite component in the magma, by the removal of apatite (including olivine, phlogopite, and magnetite).In our samples, this mostly produces apatite cumulates (C1a) and a residual carbonatite (C1b).
The C2 group consists of unevolved magnesiocarbonatites crystallizing apatite and dolomite, and may also represent cumulates (e.g.C2 samples from Jacupiranga) or residual carbonatite (e.g.C2 samples from the APIP).
The C3 in our sample set are all from the Tapira Complex.This group is characterized by fractionation of apatite, calcite, and phlogopite, leading to a residual concentration of Ba, Mg, and Na, and the crystallization of carbonates     Taylor et al. (1967) and for S by Deines (1989), Mitchell & Krouse (1975), Druppel et al. (2006) and Nikiforov et al. (2006).Arrows indicate changes expected in Rayleigh fractionation, low-temperature hydrothermal alteration without the participation of CO 2 and degassing (Taylor et al. 1967, Ray & Ramesh 2000).The gray triangle in the inset illustrates the expected variations (e.g.Santos & Clayton 1995)  such as norsethite, baritocalcite, and burbankite directly from the magma.This group marks the exhaustion of the silicatic component in calciocarbonatites, with the end of phogopitecrystallization.The C3 geochemical characteristics, in particular its high chondrite-normalized Nb/Ta and the strong negative Zr and Hf anomalies, indicate that these melts were produced by liquid immiscibility (e.g.Brod et al. 2013).C4 consists of Ba-, Sr-, and REE-rich magnesiocarbonatites.The main liquidus phases are dolomite, Fe-dolomite, norsethite, and strontianite.Its distinctive characteristic is the strong enrichment in monazite, marked by a concomitant increase in P 2 O 5 and REE 2 O 3 with magmatic evolution.The geochemical characteristics indicate the involvement of liquid immiscibility in the evolution of these magmas.
The geochemical characteristics of the C5 indicate strong apatite and phlogopite fractionation, with decrease in CaO, P 2 O 5 , and K 2 O, and enrichment in Ba, Sr, Na, and CO 2 , resulting in a direct crystallization of strontianite and norsethite as liquid phases.Exsolutions indicates that the C5 carbonatites were crystallized in still relatively high-temperature conditions.The high chondrite-normalized Nb/Ta indicates the involvement of liquid immiscibility processes in the evolution of these magmas.Previous work (Junqueira-Brod et al. in prep., Gomide et al. 2013) identified degassing of CO 2 , as well as sulfur as an important petrogenetic process in samples of this group.
The C, O, and S isotopic data are consistent with the interpretation of Santos and Clayton (1995) in that the APIP complexes were emplaced at shallower crustal levels than the Jacupiranga Complex.The lower lithostatic pressure at these shallower levels allowed a much greater diversity of petrogenetic processes to act in the evolution of the APIP carbonatite magmas, including fractional crystallization, liquid immiscibility, degassing, and interaction with hydrothermal and carbohydrothermal systems, often in recurring events (Barbosa et al. 2012, Brod et al. 2013, Cordeiro et al. 2010, Gomide et al. 2013).Brod (1999), Barbosa (2009), Cordeiro (2009), Grasso (2010) and Palmieri (2011).APIP: Alto Paranaíba Igneous Province.

Figure 6 .
Figure 6.Behavior of selected major element oxides with magma evolution for the Alto Paranaíba Igneous Province C1 calciocarbonatites.Symbols used are same as in Figure 3.

Figure 7 .
Figure 7. Behavior of selected major element oxides with magma evolution of the Alto Paranaíba Igneous Province C2 magnesiocarbonatites. Symbols used are same as in Figure 3.

Figure 10 .
Figure 10.Behavior of selected major element oxides with magma evolution in the Alto Paranaíba Igneous Province C5 magnesiocarbonatites. Symbols used are same as in Figure 3.

Figure 14 .
Figure 14.Multielement diagram for C1a, C1b, C2, C3, C4, and C5 carbonatites.The fields represent the province fields for each group and REE diagrams.The fields represent the Alto Paranaíba Igneous Province range for each group (gray = C1a, black outline = C1b).Symbols used are same as in Figure 3.

Figure 15 .
Figure15.Left side: Stable isotope composition of carbon and oxygen in carbonates for Jacupiranga and for the Alto Paranaíba Igneous Province groups C1 to C5.Right side: Stable isotope compositions of sulfur in sulfides and oxygen in carbonates for Jacupiranga and for the Alto Paranaíba Igneous Province groups C1 to C5.The rectangular field corresponds to the isotopic composition of the mantle defined for C and O byTaylor et al. (1967) and for S byDeines (1989),Mitchell & Krouse (1975),Druppel et al. (2006) andNikiforov et al. (2006).Arrows indicate changes expected in Rayleigh fractionation, low-temperature hydrothermal alteration without the participation of CO 2 and degassing(Taylor et al. 1967, Ray & Ramesh 2000).The gray triangle in the inset illustrates the expected variations (e.g.Santos & Clayton 1995) in low-temperature carbohydrotermal alteration with high H 2 O:CO 2 (1,000:1) and variable fluid/rock ratios.The S isotope data are fromGomide et al. (2013), whereas C and O data are from this work.

Table 1 .
Characterization of the studied carbonatite groups based on mineralogy, textures, chemistry and evolution stage.Jacupiranga samples are taken as representative of the least evolved stages.
Brod  et al.'s (2013)detection and definition.The fact that there are variable Nb (n) /Ta (n) within a single group suggests that both crystal fractionation and liquid immiscibility played a role in magma evolution in most cases.Sc (n) /Y (n) displays an interesting behavior, allowing an efficient separation of groups C2 and C5, which show high values of this ratio.
Gomide et al. (2013)arbohydrotermal alteration with high H 2 O:CO 2 (1,000:1) and variable fluid/rock ratios.The S isotope data are fromGomide et al. (2013), whereas C and O data are from this work.