Magnetite in the Carbonatites from the Jacupiranga Complex, Brazil

American Mineralogist, Volume 6E, pages 195-213, 1983

Magnetite in the carbonatitesfrom the JacupirangaComplex, Brazil

JosB C. Gespan Instituto de GeosciAncias Universidade de Sao Paulo Sao Paulo, Brazil CEP 05508

eNo PBren J. Wyrue Department of Geophysical Sciences University of Chicago Chicago, Illinois 60637

Abstract

Electron microprobe analysesof magnetitesfrom five carbonatiteintrusions (Cr, oldest, to C5, loungest) constituting the carbonatite plug in the Jacupirangacomplex confirm' previous results from Jacupirangagiving compositions in the magnesioferrite-magnetite seriesvery close to Fe3Oa.Magnetites from other carbonatitesare similar with somewhat more Ti and less Mg. MgO in Jacupirangamagnetites reaches no more than l0 wt.%. All analyzedgrains are zoned, with Fe3Oaincreasing toward the rim. In magnetitesfrom C2 to C5, Fe2O3 replacement is mainly by Al2O3 and less by TiOz; in C1 magnetites TiO2 replacementis more important. Despite their limited range of compositions, the cores of magnetitesin each of the five intrusions are chemically distinct and distinguishablefrom each other as indicated by projections from within the Cr-free spinel prism, MgFe2Oa- MgzTiOa-MgAI2Oa-Fe3Oa-Fe2TiOa-FeAl2Oa,and a plot of Mn ys. Mg. Magnetites from special locations such as dikes, bandedreaction zonesbetween carbonatite andjacupirangite, and in intergrowths with pyrite, are chemically related but distinct from the carbonatite magnetites.The systematicchemical variation and zoning of magnetitesin the five carbonatite intrusions indicate magmaticorigin. Magnetite crystals nucleatedthrough- out the crystallization interval of the carbonatites, but most of them show evidence of marginal resorption. The oldest carbonatite, C1, was probably derived from a magma somewhat different chemically from those producing carbonatites C2 through C5. The precipitation of carbonatite C2 probably went to completion independentlyof C3 through C5, whereas carbonatites C3 through C5 probably were precipitated from successive batchesof magmarepresenting a continuum in time and magmaticevolution.

Introduction from carbonatites. Mitchell (1978)and Boctor and Svisero (1978)presented analyses of the magnetites Magnetite is a characteristic mineral of nearly all from Jacupiranga.Other analyseshave been report- carbonatites,usually as an accessory,but locally as ed from Oka by Gold (1966) and McMahon and an essentialmineral (Heinrich, 1966,p. 182).The Haggerty (1976, 1979), for zoned magnetites from magnetite occurs as disseminated euhedra, com- South African carbonatites by Prins (1972), from monly in flow bands, or as lenses and irregular Alno by von Eckermann (1974), and from Sarfar- aggregates.Magnetite is also concentrated in ore tdq, Greenland, by Secher and Larsen (1980). rocks which occur as integral parts of the ring Mitchell (1979)summarized the chemical character- complexes. The magnetite crystals are homoge- istics of the magnetitesfrom carbonatitesas "com- neous, but exsolution of lamellar ilmenites may positionally very close to pure magnetite" and as occur (Prins,1972;Mitchell, 1978;Boctor and Svi- membersof the magnesioferrite-ulv

Cr2O3,the low TiO2 and MgO relativeto kimberlite spinel,and commonly high MnO". The presence of five carbonatite intrusions in a single complex at Jacupiranga,and the continuous outcrop and freshness of the rocks in the quarry, make this an unusual opportunity to samplein detail the mineralogy of a carbonatitecomplex. In this paper, we present detailed chemical analyses of zoned magnetites from the five carbonatites, and explore the applications of their chemistry for petrogenesis.

The Jacupiranga Complex The JacupirangaComplex is one of many alkalic complexes, several containing carbonatites, that occur in the borderof the ParanaBasin, in the south * (Amaral,1978). of Brazil. It is 130 5 m.y. old Fig. l. Simplified geological map of the Jacupiranga Accordingto Ulbrich and Gomes(1981) the Jacupir- carbonatites.C1-sovite ; C2Jolomitic sovite; C3-sovite; Ca- angaComplex and other nearby complexesof simi- sovite; C5-rauhaugite and JAC-jacupirangite. Solid lines- lar agesform the JacupirangaProvince. This prov- observed geological contacts; dashed lines-inferred contacts; ince was interpreted by Herz (1977)as the site of a F-fault. The contacts on the map are incomplete for lack of secureevidence, becausethe quarry is on a hill. hot spot associatedwith the first triple junction formed in the area during the initial opening of the South Atlantic ocean. group. Although the field evidence is not conclu- Geology of the complex sive, it suggeststhat the southern intrusions are Melcher (1966)presented a geologicalmap of the older than the northern intrusions. Given this con- Jacupirangaalkalic complex. The complex has an clusion, the sequenceof carbonatite intrusions from oval shape with an area of about 65 km2, and is oldest to youngestis Cr, Cz, Ct, Co and Cs. The intrusive into granodiorites and mica schists which main petrographical and mineralogical features of are fenitized for some distancefrom the complex. It each carbonatite intrusion are summarizedin Table is composed mainly of peridotites, pyroxenites, 1. jacupirangites and irjolites, surrounded by fenites The intrusive contact of Cr into C2 is shown in and nepheline syenites. According to Melcher Figure 1 as a dashed line, becausethe banding (1966),a peridotitic body was first emplacedin the which is obvious within each intrusion becomes northern half of the intrusion, and then surrounded diffuse in the region of this contact. In addition, by a ring intrusionof pyroxenite. South of this, and there is an increase in the amount of dolomite partly cutting it, an almost circular jacupirangite present in the rocks near the contact. In some plug was intruded, and partly differentiated into a regions, there appear to be a swarm of beforsite concentric, crescent-shapeijolite body. The car- dikes intruding the sovites,but in other regionsthe bonatitesoccupy a centralposition in thejacupiran- increasein dolomite could be attributed to a meta- gite. somaticprocess of dolomitizationof the sovite.The S and B samples in Table I are respectively the The carbonatites calcite and dolomite rock of this region. The carbonatite body has steeply outward-dip- Alvikite and beforsite dikes are widely distribut- ping contacts with the jacupirangite and, according ed. They usually contain fewer non-carbonated to Melcher (1966),it was emplacedin two plugs, minerals than the larger carbonatite intrusions. The one partly cutting the other. Detailed mapping by Badike in Table I is a beforsitecutting the Cr sovite. one ofus (JCG)has revealedthe occurrenceoffive The beforsite dike 85 is an offshoot from the C5 distinct intrusions of different ages, as shown in rauhaugite,which also cuts the Cn sovite. Figure 1. The intrusions Ct, Cz and C3 comprisea At contacts between carbonatite and jacupiran- southerngroup, and Ca and C5 comprise a northern gite (xenoliths and country rock), a banded rock is GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA 197

Table l. Petrographyof carbonatitesand jacupirangite

Rocktype* l,lain mineral ogy** Grain size Otherfeatures Calcite,apatite, magnetite, c1 Sovite olivine, phogopite,dolomite, Coarse Coarselybanded sulfi des.

Calcite, dolomite,apatite, Fine to Finely banded SouthernIntrusi ons c? Doloni ti c magnetite,phl ogopite, medium sovite sulfi des.

Calcite, apatite, magnetite, libdium to Non-carbonatemi neral s ca Sovite phlogopite,dolomite, coarse fewer than in other olivine, sulfides. sovi tes. Banding l ess evi dent. Calcite, apatite, rnagnetite, C4 Sovi te ol i vi ne, phlogopi te, doloni te , ti,ledium sulfides. Northern Intrusi ons Dolomite, apatite, phlogopite, lledi um to Veryfew si I icates, c5 Rauhaugite magnetite, sulfides, calcite. coarse oxidesand sulfides phlogopite, Fi in the B4 Beforsi te Dolomite, ne Intruded C4 magnetite, calcite sovite UI KCS Dolomite, calcite, apatite, l'4edi um to 0ffshoot of the Cq B5 Beforsi te magnetite, phlogopite, fi ne rauhaugite in the' sul fi des. C4 sovite calcite, dolomrte, apatlte, I{edr um to Rocksfrom the Sovite phlogopite, magnetite, coarse Diffuse banding contact between sul fi des. carbonati te Dolomite, calcite, apatite l',ledi um Diffuse banding intrusions C2 Befors i te and Ca magnetite, sulfi des Heactlon zone 51| rcate rrcn bands between the Phlogopite, alkal i amphibole, Fineto alternate with 'intrusions (Cr,Ca) - Reacti on calcite, magnetite, apatite, coarse carbonate rich bands and Ca) and tfie rocK su1fides, ilmenite, cl inohumite (magnetite occurs in jacupirangite the silicate bands only

Host rock JAC Jacupirangite Titanoaugite, titanomagnetite ltledium

Oxher rock tgyes cited in the text: aJ.vikite (cafcite, apatite, dolonite, phLogopite) and ijoTite ( cI ircpgroxene, nepLel ine ) . The mineraTs are cited in an estimated decreasing otdet of abvndance

TIre ctnracteristics cited for these two dikes refer specifical,Tg to them ard. not to tle dikes in gereraJ- tlat occur in the carbo@tite bodies.

developedvarying in width from l0 cm to more than In carbonatite C1, the magnetite crystals vary in 2 m. Centimeter-wide bands of carbonate alternate size, but they are most commonly coarse grained. with silicate-rich bands composedmainly of brown The largest magnetite crystals reach up to 4 cm in phlogopite, alkali amphibole, magnetite, and veins diameter. In the main concentrations of magnetite, of carbonate a millimeter or so wide. The magne- associatedwith apatite and olivine, the crystals are tites occur in the silicate-rich band only and contain octahedral, but more irregular crystals exhibiting platy exsolution lamellae of ilmenite. the effects of resorption are widespread. Viewed in polished sections, the crystals are light purple and The magnetites homogeneous.Inclusions of carbonatesdefined by As shown in Table 1, magnetite occurs in all five sharp straight edges are found. Apatite inclusions carbonatiteintrusions. It is most abundantin C1and are common. Ca, and least abundant in C5. No evidence for C2 carbonatite is equigranular,without local con- oxidation of the magnetiteswas observed. centrations of magnetites or other minerals. The GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA magnetite diameters rarely exceed 2 or 3 mm, and the crystals tend to be rounded by resorption. Inclusions of carbonates similar to those found in the C1 carbonatite are common. The magnetitein C3carbonatite varies from a few millimeters up to l0 cm in diameter. In regions where the crystals are larger, the abundance is smaller. Crystals larger than 3 cm are twinned according to the spinel law. Inclusions of carbon- Mqil-i0a ates and more rarely sulfides occur within some magnetite crystals. In C+ carbonatite, magnetite may form large concentrations along with apatite, olivine, and weighf phlogopite. The magnetite is anhedral to perfectly MgFe2q octahedral, with the octahedral forms occurring in the mineral concentrations. The magnetitevaries in Fig. 2. Composition(wt,Vo)of magnetitesfrom the carbonatite prism. size from less than I mm to I cm. Inclusions of intrusions representedin a Cr-free spinel The analysesin Table 2 occupy the outlined volume near Fe3Oa. carbonatesare present. The magnetitein the C5 carbonatiteis an accessory mineral. It is light gray in polished sections, commonly anhedral, and always fine-grained, compositionsby plotting weight fractions of compo- reaching diameters of I mm. No inclusions were nents rather than cation fractions. Figure 2 is a observed. modified combination of the two spinel prisms in The magnetites from all five carbonatite intru- general use, which each include an axis for chro- sions were analyzed with an electron microprobe. mites (Haegefty, ln6). Chromium is present only Analyses were also obtained for magnetites from as a trace element in the Jacupiranga magnetites, the contact zone between C2 and C3, from the and with removal of the chromite axis the remaining reaction zone between carbonatite and jacupiran- variables from the two spinel prisms are combined gite, and from intergrowths in sulfides. in Figure 2. The apex of the prism represents the magnesioferrite-magnetiteseries, and the base of Analytical procedure the prism is defined by the spinel-hercynite series Microprobe analyses were carried out using an and the magnesiantitanate-ulvdspinel series. The enr/nux automated microprobe operated at 15 kV volume depicted in Figure 2 illustrates the fact that and 1.0 p,A. Data were corrected by the computer more than 95Voof eachanalyzedspinel is represent- program MAcrc rv. The accuracy of the analysesis ed by the magnesioferrite-magnetite series (see -+5Vo of the amount present for concentrationsup to Table 2). The variationsof Mg, Fe2*, Fe3*oTi, and 2 wt.o/oand +2Vo of the amount present for major Al can be representedin projections from Figure 2. elements. The detection limit for minor elements Table 2 gives values for other elements.MnO varies calculated according to Reed (1973) is near 0.03 from 0.l2Voto l.52Vo,and CaO from UVoto more wt.Vo. Other details are given in the Appendix. than0.6Vo.The elementsNb, Ta, and Zrwere either not detectable, or rarely detectable but present in Magnetites from the carbonatite intrusions amounts below measurable limits. V occurs as a The averageanalyses of the cores and rims of the minor element, as also reported by Boctor and zoned magnetitesare listed in Table 2 and illustrat- Svisero(1978). ed in a seriesof figures. Figures in the Appendix are identified with the prefix A. The results confirm the report of Boctor and Svisero (1978)that the miner- Individual zoned crystals als are composedessentially of the magnesioferrite- Zoning is rather complex in most of the magne- magnetitesolid solution series, with striking zoning tites, with the complexity tending to increase with marked by increase of Fe3Oafrom core to rim of increasinggrain size. The character ofthe zoning is minerals. The mineral compositionsare convenient- illustrated in Figures Al through ,4,5by the electron ly represented in Figure 2. We compare mineral microprobe traverses across individual minerals GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA l9

Table 2. Average compositionsof magnetitesfrom the carbonatite intrusions

c1 I 3 4 5 6 c(5) R(6) c(10) R(e) c(8) R(6) c(8) R(8) c(6) R(6) c(8) R(7) Ti02 1.49 0.67 3.18 ?.60 2.84 I .15 2.27 2.O7 1.48 I.38 1.98 I.89 A1203 0.23 0.12 0.04 0.00 0.06 0.00 1.00 0.28 0.58 0.28 1.67 1.65 si02 0.03 0.03 0.02 0.02 0.02 0.03 0.03 0.03 0.04 0.02 0.03 0.03 Fe203 66.4 67.7 65.2 65.8 65.6 68.6 66.9 67.7 58.5 68.7 67.6 67.3 Feo 30.5 30.1 27.O 27.3 26.4 26.4 23.5 24.5 22.0 22.2 2t.3 20.9 Irh0 0.20 0,I2 0.8I 0.58 0.73 0.48 0.76 0.72 0.71 0.80 0.69 0.76 Cao 0.05 0.37 0.01 0.16 0.01 0.24 0.02 0.30 0,03 0.47 0.00 0.20 l4g0 1.14 0.51 4.08 3.46 4.24 3.26 5.98 4.97 6.30 5.63 7.45 7.29 ToTAL 100,04 99.72 I00.34 100.02 99.90 100.16 100.46 100.57 99.74 99.48 I00.72 I00.02

7 9 10 11 12 c(8) R(8) c(10) R(10) c(6) R(6) c(11) R(8) c(7) R(7) c(s) R(6) Ti02 0.35 0.30 0.97 0.94 0.71 0.68 1.00 1.I0 0.90 0.69 0.s0 0.45 1.28 Al203 0,59 0.54 1,37 0.97 t.52 0.91 1.19 0.9I 2.69 1.24 1.93 si02 0.02 0.0r 0.04 0.03 0.02 0.03 0.02 0.02 0.04 0.03 0.02 0.03 Fe203 70.2 69.3 69.3 69.3 69.8 69.8 70.8 70,5 71.3 i0.1 7l.l 71.3 FeO 24.6 26.2 21.3 22.1 19.9 20.6 17.5 L7.4 15.5 18.0 t).J lf.f l.ho 0.59 0.52 1.03 l.18 r.04 1.19 1.36 t.52 0.96 1.30 1.16 1.45 Cao 0.07 0.33 0.07 0.47 0.03 0.50 0.03 0.43 0.03 0.41 0.08 0.47 il90 4.09 2.87 6.41 5.48 7.24 6.13 8.83 8.40 9.94 7.79 9.95 9.19 T0TAL 100.61 100.17 100.55 100.47 100.26 99.84 100.73 100.28 101.36 99.56 100.04 99.67

ca l3 l4 15 16 t8 c(14) R(12) c(10) R(e) c(4) R(4) c(12) R(e) c(1i) R(r0) c(6) R(5) Ti02 1.54 0.85 1.47 1.37 0.57 0.33 0.43 0.33 0.64 0.49 0.60 0.55 0.39 2.11 0.81 Al 203 i.14 0.13 1.55 0.34 0.89 0,1I 2.40 1.51 2.t6 si 02 0.03 0.03 0.02 0.03 0.01 0.03 0.02 0.02 0.02 0.03 0.0i 0.03 Fe203 68.s 70.1 58.0 68.6 7t.7 70.5 70,5 70.8 70.3 71.3 70.4 69.7 Fe0 22.2 23.7 20.7 22.5 18.4 24.6 17.3 18.9 17.1 19.5 16.7 21.4 Mn0 0.98 0.85 0.99 1.10 0.71 0.61 0,85 L.?5 0.78 1.04 1.03 1.03 Ca0 0.03 0,61 0.02 0.43 0.01 0.30 0.01 0.38 0.0I 0.30 0.01 0.36 I'tg0 6.32 4.45 7.t0 5.30 8.35 3.82 9.03 7.27 9.26 6.90 9.29 5.69 TOTAI. t00.74 I00.72 99.85 99.67 100.64 100,30 100.54 100.46 LOO.27 100.05 I00.15 99.57

19 2T 22 c(10) R(8) c(10) R(r0) c(r3) R(e) c(e) R(e) c(12) R(11) c(14) R(10) Ti02 0.21 0.20 0.23 0.25 0.30 0.14 0.42 0.26 0.22 0.2I 0.27 0.23 Al203 0.04 0.01 1.42 0.40 1.06 0.26 0.02 0.03 0.32 0.13 0.86 0.16 si02 0.02 0.02 0.02 0.03 0.02 0.o2 0.03 0.02 0.02 0.03 0.05 0.03 Fe203 69.4 69.4 69.6 69.7 70.4 70.9 69.1 69.3 69.2 69.0 69.1 69.6 Fe0 27.5 27.6 22.4 24.1 2t.9 23.5 28.8 29.3 28.7 29.5 28.6 ?9.6 r'h0 0.34 0,34 0.71 0.71 0.63 0.57 0.r7 0.12 0.26 0.22 0.32 0.22 Ca0 0.16 0.46 0.02 0.42 0.0I o.32 0.27 0.50 0,03 0.26 0.04 0.34 M9o 1.90 I.7l 5.52 3.84 5.90 4.44 1.43 0.9I 1.56 0.89 1.83 0.93 TOTAL 99.67 99.74 99.92 99.45 t00.?2 100.15 t00.24 100.44 I00.31 100.24 10t.07 101.11

l-Sovite 5-Sovite 9-Rauhaug'ite 13-Sov i te l7-Sovite zl-Sovi te 2-Sovite 6-Sovite l0-Sovite 14-Sovite l8-Rauhaugite 22-Rauhaugite 3-Sovite 7-Rauhaugite 1l-Sovite 15-Sovite 19-Sovite 23-Rauhaugite 4-Sovite 8-Sovite l2-Sovite l6-Sovite 20-Beforsite 24-Beforsite

C = core of mineral R = rim of nineral (X) = numberof measurements GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA

from each of the five carbonatite intrusions. De- scriptions of the figures are given in the Appendix. c) The zoning trends from cores to rims represented Ctl in terms of oxides are as follows: (l) MgO de- cD creases, (2) Fe2O3and CaO increase, (3) TiO2 G .tt decreasesor remainsconstant, (4) Al2O3decreases ct €) except for C1 where it remains constant, (5) FeO (-, increases o) except for Cr where it decreases,and (6) o MnO remains constant, or exhibits a slight decrease (Cr) or slight increase(Cr). 0.6 01 0.8 0.9 Examinationsof these zoning profiles leadsto the following conclusionsabout variations in solid solu- Fe0/(FeO+MgO) tions (1) from core to rim: Fe3Oaincreases, (2) Fig. 3. Range of FeO/(FeO+MgO) of the core compositions MgFe2Oadecreases, (3) in magnetitesfrom C1, (Table 2) for the five carbonatite intrusions according to Fe2TiOa decreases with insignificant variation of decreasing age (Fig. I, Table l). Cr-triangles, C2-squaf,es, MgAl2Oa,(4) in magnetitesfrom C2to C5,MgAl2Oa C3---circles,Ca--diamonds, and C5-hexagons. decreaseswith only slight variation of FezTiO+,(5) MnFe2Oafiacobsite produced by replacementof FeO or occasionallyMgO by MnO) generallydecreases, overlapsCt, Ct, and Ca;and that for C1overlaps the but it increases in some samples, (6) CaFezOr other four intrusions. With the exception of the (produced by replacement of MgO in magnesiofer- magnetites for Ct, the highest value of rite by CaO) increases. FeO/(FeO+MgO) for the cores of magnetitesin It is evident that ulvrispinel plays an important each intrusion decreaseswith decreasingage. role in the zoning of magnetite of the C1 carbona- Figures 4A,5, and 4'6 show the sameanalyses in tite, but it is not important for magnetitesfrom the projections from the spinel prism of Figure 2. The other carbonatites. The magnetites in the C1 car- projections correspond in terms of geometry to the bonatite also differ from those in the other carbona- standardprojections of Irvine (1965),but composi- tites because MgAl2Oa remains nearly constant, tions are in weight fractions rather than in cation whereasit decreasesfrom core to rim in magnetites fractions. Two additional points are given, Ba and from the other carbonatites. 85 representingthe core compositionsof magnetites from two beforsite,dikesassociated with carbona- Core compositions and ranges of zoning tites Ca and C5, respectively(see Table 1). Figure 6 Becausethe compositions of the cores of the comparesthe Mn content with Mg content of the magnetites should reflect the early history of their crystallization in each carbonatite intrusion, these compositionsare reviewed and comparedfirst. The zoning trends will be analyzed subsequently,using #E-# N the chemical variation trends of the core composi- O 0.99 F tions in each of the five carbonatite intrusions as a + basis for comparison.The averagecore composi- oo tions from Table 2 can be related to their positions fl within the spinel prism of Figure 2 in the projections 0.97 oo illustratedin Figures 3, 4, and 5. e n The variation of FeO/(FeO+MgO), corresponding to the main trend along the magnesioferriteto 0.95 UO 07 0.8 0.9 t006 07 08 09 t0 magnetiteedge of the spinel prism, is shown for the core compositions of magnetitesin each of the five FeOl(Fe0+Mg0) Fe0/(rgg +Ygg; intrusions in Figure 3. The values are arranged in Fig. 4. Average core and rim compositions of magnetites order of decreasingage of the intrusions, from C1 to (Table 2) in projections from the spinel prism (Fig. 2). A Cs. The core compositions representativerange of values is shown for one of the average from each intrusion points. occupy A. Core compositions. See Figure 3 for symbols. B. The a fairly wide range, except for C5. The core compositions connected by lines to corresponding rim rangesfor Ct, Ca, and Cs barely overlap; that for C2 compositions (dots) showing zoning trends. GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA

magnetite core compositions (Fig. aA) this same trend appears to be superimposedon a dominant trend in Cr Cz and Cr directed towards the lower right corner, which corresponds to enrichment in ulvospinel (see Fig. 5). Additional details are given in the Appendix. The chemical variation trends exhibited by the magnetite core compositions are well illustrated in oN the projections from the spinel prism in weight per i-- cent, but the zoning exhibited by individual magne- + tl tites within the intrusions is better representedand os interpreted in terms of element substitutions. Elementsubstitutions ao o The analyses in Table 2 were recalculated in s terms of cation numbers. Element substitutions for Fe3* are illustrated in Figure 7. Substitution by Al3+ is representedby the replacementline for the ratio 1:1, and substitution by Ti+Si accordingto 2Fe3* : Ti4* + Fe2*, as in the magnetite-ulvospinel series,is representedby the replacementline for the ratio 2:1. Stoichiometric magnetitesplot between these two lines. The substitution of AIzOr is more important for the core magnetitesin intru- 0.6 0.7 0.8 0.9 to sionsC2 to Cs. The magnetitesfrom C1 plot closer to the 1: I replacementline, confirming their greater FeO/(FeO+MgO) content of ulv

netite cores from the five intrusions, transferred from Figures 44, 4,6, and 6.

Figure 8. The field relationships of the rocks listed I -- Sovite in the contact of C2 and Ca. were outlined briefly in connection with the geolo- 2 -- Beforsite in the contact of C2 and Ca. gy, and Table l. 3,4 -- Jacupiranguite. llo ilnenite lanellae detected. 5 -- Jacupiranguite. Thenearest reasurenent to an ilrenite lanella in grain The an exsolved contact between carbonatite intrusions Ct 6,7 -- Exsolvedmagnetite in the reaction rock of the Cl intrusion. and C3 I -- Exsolvedmagnetite in the reaction rock of the Ca intrusion, -- In an attempt to find evidence to account for 9,10 Exsolvedmagnetite in the reaction rock of the C5 intrusion. the II -- Magnetiteintergrowth with pyrite in a rauhaugitefron the Ci difiuse banding and the increase in dolomite at the i ntrusion . 12 -- tlagnetiteintergrowth with pyrite in a sovite frm the Ci intrusive contact between carbonatites C2 and C3 intrusion. (Fig. 1), magnetitesfrom the calcitic (S) and colomi- 13 -- ihgnetite intergrowthwith pyrite in a sovite fron the Cd intrusion. tic rocks (B) were analyzedand plotted in Figures 8 C = cofe of mineral R = rim of nineral (X) = nunberof measurerents and 9, which show the composition rangesof mag- GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA 203

iromthe Czond Ca S+ Colciterccl contqctol The magnetitesfrom the carbonatites (Table 2) are Dolmiterc.t lrcmttt contuctol C2ond Ca B+ too low in titanium to yield the exsolved magnetites Reostionrock in Cl a Reociionrock in Ca of the reaction rock (Table 3). Therefore, the ex- O Reoctionrock in C5 solved magnetitesin the reaction bands are proba- I Jocupironguite jacupiran- v Intergrorthwith pyritein Ca bly derived from titanomagnetitesin the t Intergrowthrith pyritein Ca gite. However, as discussedin the Appendix, the F chemicalrelationships cannot be explained in terms + o of exsolutionalone. .a Magnetite intergrowth with pyrite o N a + Patchesand strings of magnetite in euhedral and t o subhedral crystals of pyrite occur commonly in {+ carbonatitesC3 and Ca, less often in C1 and C2, and € not in C5. The magnetitesare light grey in reflected light. Three analyses from examples in C3 and Ca are presented in Table 3, and compared with the other magnetites in Figures 8 and 9. They are + Fe0/(FeO Mq0) extremely low in TiO2 and Al2O3,very low in MgO, Fig. 8. Magnetitesfrom special locations (Table 3) compared and very rich in FqOa. The relationship between with the rangesof averagecore compositionsfor each of the five Mn and Mg in these magnetites contrasts sharply carbonatiteintrusions outlined in Fieures 4,{ and 46A. Dots are with that in the discrete magnetite crystals, as rim compositions. shown in Figure 9. The number of Mg cations in the intergrowth magnetitesis at the extreme low end of the range for the discrete magnetites,and the range tites showed some variation in TiO2, and a signifi- of Mn cations extends to much higher levels. The cant decreasein Al2O3 from core to rim, similar to intergrowth magnetites plot on lines transverse to the variations in magnetite from C3. The strong the variation in the discrete carbonatite magnetites, zoning for Al2O3 indicates that the calcite rock in with the l: I slope representing the substitution of the contact zone is not related to C1, despite its Mg by Mn. coincidence with the areas for C1 in some figures. This is consistent with the fact that Cr is older than Inclusions in magnetite crystals C2 and C:, according to the field relations. The magnetites from the carbonatites contain many inclusions, commonly composed of more Magnetitesfrom the reaction between carbonatite than one mineral, making it difficult to analyzethem and jacupirangite with the electron microprobe. However, the analy- Magnetite compositionsfrom the bandedreaction rocks (silicate bands) associated with carbonatite intrusionsCr, Cr, and C5 are presentedin Table 3, along with magnetite from jacupirangite, and these (D ( are compared with magnetite from the carbonatites -o E jacupiran- 5 in Figures 8 and 9. The magnetitesfrom E ? gite quite c have compositions distinct from those in o +- the carbonatite intrusions. The magnetites in the o 2 reaction bands are similar in composition to those in o c the carbonatites, with somewhat higher TiO2. = The jacupirangites contain titanomagnetitewhich exhibit some exsolution of ilmenite within a centimeter or two from the banded reaction rock. Analy- Mg(cotion number ) ses 3 and 4 in Table 3 are from a jacupirangite Fig. 9. Magnetitesfrom special locations (Table 3) compared titanomagnetite with no detectable ilmenite lamel- with the rangesof averagecore compositionsfor each of the five lae, and analysis 5, with less TiOz, is for magnetite carbonatite intrusions outlined in Figure 6. For symbols see adjacentan ilmenite lamella in an exsolved mineral. Figure E. Dots are rim compositions. GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACIJPIRANGA

ses were sufficient to determine the major minerals (1966)presented an analysis of a magnetiteconcen- present. Inclusions of carbonate, apatite, phlogo- trate from the Jacupiranga carbonatite, which is pite, and spinelare most abundant.The carbonates, similar to those from the C1 carbonatite (Table 2). with sharp straight edges, are dolomite or calcite, and rarely siderite or magnesite. Spinel (MgAlzO+) Comparison with magnetites from other carbonatites inclusions occur either alone or with carbonates. Hercynite was identified in one inclusion. High Mitchell (1978) and Boctor and Svisero (1978) values of SiO2 recorded in some analysesof spinel compared their analyses of magnetites from Jacu- (MgAlzO+) indicate the coexistence of a silicate piranga with a selection of the few analyses pub- phase. One analysis of a spinel inclusion was re- lished for magnetitesfrom other carbonatites.Some portedby Mitchell (1978).Rare inclusionsof ilmen- of these analyses, and others, are compared in ite were found in some C1 and C3 magnetites. Figure 10 with the ranges of core compositions of magnetites from the five Jacupiranga carbonatite previous Comparisonwith results from the intrusions. The magnetitesfrom other carbonatites carbonatite Jacupiranga are even more closely crowded into the Fe3Oa Figure 10reproduces from Figures 4.{ and ,{6 the corner of the spinel prism than those from Jacupir- rangesof averagecompositions for the cores of the anga (Fig. 2), but many of the others contain more magnetites from the five carbonatite intrusions at Ti. A detailed review of Figure 10, together with Jacupiranga.The magnetite analysespresented by data on MnO, is presentedin the Appendix. Mitchell (1978)and Boctor and Svisero (1978)for the Jacupirangacarbonatite are plotted for compari- Magnetite compositionsand magmatic evolution son. They reported no zoning. Their analysesare The magnetites from the five carbonatite intru- generally similar to our results, despite the fact that sions ofJacupiranga, and from carbonatites in gen- their sampleswere obtained from a higher level in eral, occupy a very restricted range of composi- the quarry some years before the collection of our tions, as shown in Figure 2 andby Mitchell (1979). rocks. The three analyses of magnetites given by However, the fact that the magnetitecore composi- (1978) Mitchell contain exsolved ilmenite lamellae. tions in each of the five intrusions are chemically More details are given in the Appendix. Melcher distinguishablefrom each other indicates that they are sensitive to conditions of formation, and they should therefore provide useful petrogenetic infor- X Jocupirongo( Miict€lt, t978 ) mation. + Jocupirqngo( Boctor qd Svisero,1978) The origin of carbonatites a oko(McMohon ond thqgert, tgzg) oqn e Sorfori6q(Secher ond Lqrsen, l9B0) Theories for the origin of carbonatites have been (1966, o 0.97 - Alnd(von Ekermonn, 1974) reviewed in detail by Heinrich chapter 10), F Tuttle and Gittins (1966),Carmichael et al. (1974, 0 Alriconcorbo0otites ( prins, t972 ) o 0.96 chapter 10), Le Bas (1977, chapters 23 and 24) and I vtiih ersolveditmenite Alricon , e no( @rbonotifes( Prins, 1972 ) Gittins (1978). There was general reluctance to O accept early proposals that carbonatites were mag- a 0.94 o matic, partly because the fusion temperature of

0.93 calcite, above 1300'C, was much higher than esti- o mated emplacement temperatures of carbonatites + 0.92 e (700'C to 350"C).There was much debateabout the b relative merit of hypotheses involving intrusive 09l -{ o marbles, xenolithic marbles, hydrothermal or car- 0.90 bothermal deposits, and metasomatic replacement 0.9 t.0 06 07 of various alkalic rocks. Experimental evidence FeO/(FeQ+Y991 that calcite could be precipitated from melts at Fig. 10. Jacupirangamagnetite analysisby Mitchell (1978)and moderate temperaturesand pressuresin the system Boctor and Svisero (1978), and from other carbonatites, (Wyllie comparedwith the rangesof averagecore compositionsfor each CaO-CO2-H2O and Tuttle, 1960) led to of the five Jacupirangacarbonatite intrusions outlined in Fisures general acceptance of a magmatic origin for most 4,{ and A64. Dots are rim compositions. carbonatites, with processesinvolving crystalliza- GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA tion differentiation or liquation of a carbonated vals, but there is a possibility that influxes of more nephelinitic magma. There is experimental evi- evolved carbonatite magma brought in magnetites dence for both processes (Koster van Groos and with higher FeO/(FeO+MgO). Wyllie, 1966,1973;Wyllie, 1978).The prospectthat Within a particular carbonatite intrusion, assum- carbonatitic melts could be present in the upper ing that crystallization of the magnetitebegan with- mantle (Wyllie and Huang, 1975)does not justify in the carbonatite magma and not in the parent suggestions that near-surface carbonatites were magmafrom which the carbonatitewas derived, the precipitated from primary, mantle-derivedcarbona- lowest value of FeO/(FeO+MgO) in a magnetite tite magmas. Mitchell (1979) presented arguments core should representthe earliest time in the history opposingsuch suggestions. of that carbonatite intrusion. There is an excellent The chemical variation of the magnetites in the correlation in Figure 3 between the lowest value of Jacupiranga carbonatites is so systematic that it this ratio for each intrusion and the age of the would be dfficult to account for it by any process intrusion determined from field relationships, ex- other than magmatic.The challengeis to explain the ceptfor C1. mineralogy of the five carbonatite intrusions in The rangesof composition of the magnetitecores terms of successiveintrusions derived from a single in C3, Ca and Cs succeedeach other closely, with evolving parent nephelinitic magma. or from only very slight overlap (Fig. 3), which is consistent successivebatches of silicate magma. The major with a conclusion that these three carbonatites processesto evaluate are (1) fractional crystalliza- represent a continuum in time and chemical evolution of parent silicate melt(s), (2) separation of tion of carbonatite magma. From this evidence, the immiscible liquid fractions from silicate melt(s), and three intrusions could represent successivebatches (3) fractional crystallization of a parent carbonatite of magmafrom a single evolving body of magma,or magmaderived previously from a silicate magma. the magma that produced intrusion C3 could have diferentiated further to produce carbonatites Ca FeOl(FeO+MSO) and its time significance and C5 in succession. It is expectedthat FeO/(FeO+MgO)for minerals The range of core compositions for the magne- crystallizing from a magma will increase with pro- tites from C2 begins with FeO/(FeO*MgO) lower gressive crystallization, that is, with decreasing than that of C3, consistentwith the agerelationships temperature and increasing time. Every magnetite (Fig. 3), but the range of core compositions contin- analyzed in the carbonatite intrusions of Jacupir- ues further than, and thus later than, the range for anga exhibits a component of zoning with increas- C3 and into the center of the range for Cr. This ing FeO/(FeO+MgO),which is consistentwith this suggeststhat the chemical evolution of C2 overlaps expectation(Figs. 4B, .A6B,and A7). Therefore,we the history of C3 and Ca, which is not inconsistent will assumethat FeO/(FeO+MgO) for magnetite in with the possibility that Cr, or the sequenceCs-Cn a given intrusion correlates with a particular tem- C5, w&Sderived by differentiation of and physical perature and time in its history. separationfrom the parent magma which produced The cores of magnetites in the four intrusions carbonatite C2. C1 to C+ (Fig. 3) display a wider range of The lowest value of FeO/(FeO+MgO) for magne- FeO/(FeO+MgO) than the ranges of magnetite tite cores in C1 is higher than values in Cz and Cr. compositions from other carbonatites illustrated in The simple interpretation that Cr is younger than Cz Figure 10. The measuredcompositions at the cen- and C3 is inconsistent with the field relationships ters of random slices through zoned magnetites (Figs. I and 3). The range of core compositionsfor would be higher than the actual core compositions, C1 extends to the highest values recorded in the but comparison of the zoning in Figures 4B, A6B youngestintrusion, C5. and A7 with the ranges of core compositions in The parent from which C1 was derived, be it Figures 4A, ,4'6,4'and 5 demonstratesthat this effect carbonatite or silicate magma, may have reached a is insufficient to explain the complete ranges of stage in chemical evolution different from that of measuredcore compositions. Thus, it is evident the parent(s) from which Cz through Cs were de- that the magnetitecrystals did not nucleatetogether rived. There appears to be some kind of intemrp- at the beginning of crystallization of each carbonation in magmatic evolution between the crystalliza- tite intrusion. Most likely, additional magnetite tion of magmas producing Ct, Cz, and Cr-Cc-Cs, crystals nucleated through the crystallization inter- with the more marked hiatus between C1 and C2. GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA

These prospects can be explored in more detail the nucleation and crystallization of the magnetites when data from coexisting minerals become avail- in carbonatite C1. The zoning representsthe chang- able. ing chemistry of the carbonatite magma during crystallization and, therefore, the opposing trend Core composition trends and trends zoning with respectto Fe2O3/(Fe2O3+TiO2)illustrated by If the successivevalues of FeO/(FeO+MgO) in the set of four magnetitecores must be causedby the ranges of magnetite core compositions reflect some other effect, presumably related in some way the successivenucleation of magnetitecrystals with to the parent magma(s)from which the carbonatite decreasingtemperature and increasing FeO/(FeO+ magmawas derived, either a carbonatite predeces- MgO) of the magmas, the simplest interpretation, sor, or silicateparent. then the zoning rangesof individual crystals should The magnetites of intrusion C1 are also distin- follow the same trends as the core compositions. guished from those of the other intrusions by their This is broadly true for the variation of Al2O3/ core compositions (Figures 4,{ and 5) and zoning (Al2O3*Fe2O3+TiO2)in Figure ,{6, and broadly trends (Fig. aB). Their distinctive chemistry com- true (with a few notable exceptions discussedbe- pared with the magnetites in the other four intru- low) for the variation of Al2O3/(Al2O3+TiO2)in sions was also noted in the illustration of their Figure 5 and A7. The variation of Fe2O3/(Fe2 zoning with respect to Fe2TiO4and MgAl2Oa(Figs. O3+TiOt, however, differs markedly from the sim- Al throughA5), and the elementsubstitutions (Fig. ple pattern, and confirms the distinctive chemical 7t. character of the magnetites in C1 compared with The zoning of magnetitesfrom intrusion Cr illus- those in the other four intrusions (Figs. 4,{ and B). trated in Figure 48 is towards the higher level of The core compositions of magnetites from car- Fe2O3/(Fe2O3+TiO2)characteristic of magnetites bonatitesC2 to C5 have values of Fe2O3/(Fe2O3r from the other four intrusions, indicating that the TiO) greater than 0.978, with one exception from core magnetite cores in C1 crystallized from a C2 with a low value near 0.965 (Fig. 4A). Their magmaless evolved than thosewhich yieldedintru- zoning lines are directed almost horizontally, indi- sionsC2 through Ca. The less evolved magmawas cating little variation in Fe2O3/TiO2during crystalli- probably formed earlier in the history of the com- zation of the magmas. The same trend is broadly plex than the other magmas, which is consistent defined by the core compositions of the magnetites with the age relationships based on field relation- in these four intrusions, except for the two latest ships.The inconsistencyfor C1in Figure 3, consid- magnetitesfrom C3 which drop down to valuesjust ering the lowest value of FeO/(FeO+MgO)as indi- below 9.8, and for the exception from C2 noted cator of relative ages of intrusions, remains above. unexplained. The core compositions of all magnetitesfrom C1 There is one anomaly in Figure 4. The magnetite have Fe2O3/(Fe2O3+TiO)less than 0.98. The four from C2 with low Fe2O3/(Fe2O3+TiO2)has core C1 core compositions lowest in Figures 4,4.and B composition and zoning trend corresponding to illustratea trend of Fe2O3/(Fe2O3+TiO,decreasing those of magnetitesin Cr. In other respects,this from 0.97to 0.95with increasingFeO/(FeO+MgO). particular magnetite does not differ significantly (The two lowest points for core magnetitesfrom C3 from the others in C2 (Figs. 4 and 5). indicate a parallel trend for C3, but the zoning Apart from the contrast between magnetitesfrom trends in the two intrusions are quite different). The C1compared with those from the other intrusions in zoning of individual magnetites in C1 is directed Figure 48, the zoning trends exhibit agreementwith approximately perpendicular to the core trend de- the general trends of core compositions, but there fined by the four magnetites,that is, with increasing are a few examplesof zoning reversals. The last C2 Fe2O3/(Fe2O:+TiOz)for increasing FeO/(FeO+ magnetite in Figure .4.68 reverses, exhibiting an MgO).The two magnetitecores in Cl with valuesof increaseof Al2O3l(Al2O3*Fe2O3*TiO2)from core Fe2O3/(Fe2O3+TiO)near 0.98, higher than the four to rim, and it shows the equivalent reversal in defining a core trend from 0.97 to 0.95, are dis- Figure A7. Three other exceptions in Figure A7 placed from the core trend in the direction of the (two from C1 and one from C5) illustrate a re- zoning paths, and their zoning lines continue in the versal of the zoning trend, with increase of same direction. Al2O3(AlzO3+TiO2)from core to rim. There must be at least two effects associatedwith Figure 6 illustrates a general variation in magne- GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA tite cores of decreasing Mn with decreasing Mg, interpretation of the data so far available is that the which correspondsto increasingFeO/(FeO+MgO) calcite rock in the contact zone could be related to and progressive crystallization. The Mn content either intrusion, C2 or C3, but is probably related to may increase or decrease from core to rim of Cr. The formation of the dolomite-rock in this zone individual crystals, with much more variability than appears to be most closely related (in terms of zoning for the other elements(Figs. 48, A6B, and magnetite chemistry) to the beforsite dikes (Bs) A7). There is a tendency for magnetiteswith high emanating from the youngest intrusion, Cs. We Mg to zone towards increasing Mn, and for those concluded that the dikes Bs could have been the with lower Mg to zone towards decreasingMn, but precursors of the magmaforming intrusion C5, and there are some exceptions amongmagnetites richer perhaps this material also penetrated the carbona- in Mg (Figure A8). Note that the magnetitesfrom C1 tite complex along the earlier contact between C2 fit the generalpattern, but the earliestmagnetites do and C3. Alternative explanations involve metaso- not contain as much Mn as most of the magnetites matism by the introduction of fluids from uncertain in C2, and as half of those in C3. Given the field sources.More information is required for a satisfac- (Fig. 1) and chemical(Fig. aB) evidencethat Cr is tory interpretation. older and less chemically evolved than the other betweencarbonatite and four carbonatite intrusions, the relatively low Mn Reaction zone iacupiran- gite content of its more Mg-rich magnetitesis consistent with the previous conclusion that there is somekind The data for magnetite present in jacupirangite of chemical hiatus between C1 and CrCs, probably and in the bandedrocks developedbetween carbon- related to their respective parent magmas. atite andjacupirangite were reviewed in connection with Figures 8 and 9. The jacupirangite contains Beforsite dikes titanomagnetite which exsolved ilmenite in the re- In Figures 4,A.and B, the magnetitecore composi- action bands. The magnetitesin the silicate layers tions and zoning in the dike rocks Ba and 85 are (magnetiteis absentfrom the carbonatitelayers) are indistinguishablefrom those in the associatedintru- similar in composition to some of those in the sions,Ca and C5.In FiguresA6 and 5, however,the associated carbonatite intrusions. but with some- magnetite cores in the dikes are relatively higher in what higher TiO2. They originally contained much Al2O3/(Al2O3+Fe2O3*TiOz)and Al2O3l(Al2O3 more TiO2 than the magnetite in carbonatite, as +TiO2) compared with the magnetite cores in Ca shown by the presence of ilmenite lamellae. The and Cs,for comparablevalues of FeO/(FeO+MgO). evidence is clear that the magnetitesin the reaction This indicates that the dikes crystallized from mag- bands were produced from titanomagnetite in the mas representingearlier stagesof evolution than the jacupirangite by exsolution of ilmenite. However, magmasprecipitating the carbonatite bodies which as concluded in connection with Figure 9, there are they cut (C+)or are derived from (Bs is an offshoot other chemical reactions involved, and we have from C5, Table 1). insufficient evidence to work them out in detail. The dike Ba cannot representa differentiate from More information is required about the range of the intrusion Ca which it cuts, and its occurrence titanomagnetite compositions in jacupirangite, and suggeststhe persistencebelow the intrusion Ca of a the intermediate stages of the reaction between magma body at a somewhat earlier stageof chemi- jacupirangite and carbonatite. cal evolution than C+. Similarly, the dike 85 ema- Metasomatic processesinvolved in the formation nating from C5 cannot representa differentiatefrom of the banded rocks produce magnetites in the the intrusion C5. It could representthe initial intru- silicate bands only, which do not correspondclose- sion of carbonatite C5, which was followed by the ly in composition to the magnetitesin the associated magma body producing the somewhat more carbonatites. Therefore, we see no evidence sup evolved carbonatite C5. porting extrapolation of metasomatic processesin the reaction bands to the origin of carbonatite Contact region betweenintrusions C2 and C3 intrusionsas a whole. The magnetite data for the calcite- (S) and dolomite-rocks (B) in the contact zone with diffuse Discussion bandingbetween intrusions C2and C3was reviewed The field evidence that the carbonatite plug at in connection with Figures 8 and 9. The simplest Jacupirangais made up of not two (Melcher, 1966) GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACTJPIRANGA

but five intrusions (Table 1), plus many dikes, residual fluids escapedfrom the crystalline deposits indicates that the petrogenesismay be quite com- and fenitized the country rocks. plex. The evidence from the magnetite composi- These results from magnetite, a simple and com- tions confirms that the five intrusions were not monly neglectedmineral in carbonatities, illustrate produced simply by a continuous process of differ- a remarkable sensitivity to conditions of crystalliza- entiation. tion. When supplementedby the detailed studiesof The magnetites occupy a very limited range of other minerals in the carbonatites,they will provide compositions near Fe3Oa(Fig. 2), as do magnetites an even higher potential for unravelling petrogene- from other carbonatites(Fig. 10; Boctor and Svi- sis. sero, 1978;Michell, 1979),but they are sufficiently Mitchell (1979)reviewed spinelsin severalrocks, sensitive to conditions of growth that the core and noted that spinel compositions in kimberlites, compositions in each of the five intrusions can be lamprophyres, and carbonatites all converge upon distinguished on the basis of variations in MgO, Fe3Oa(see Fig. 2). Mitchell (1978,1979)and Boctor FeO, Fe2O3, Al2O7,TiO2, and MnO. Despite the and Svisero (1978)emphasized that magnetitefrom limited range of compositions, it appearsthat mag- Jacupirangaand other carbonatitesdiffers markedly netite crystals nucleated through most of the crys- from spinelphases in kimberlites. tallization intervals of each carbonatite. The evi- Mitchell (1979)presented a stimulating and inci- dencefor marginal resorption shown in textures and sive review of evidence to refute the "alleged some zoning profiles (Figs. Al through .{5) indi- kimberlite-carbonatite relationship", the common- catesa changeduring the closingstages ofcrystalli- ly held view that kimberlites and carbonatites are zation. genetically related. The marked diference in com- There is good evidence from the magnetitesthat positions of spinels from kimberlites and from car- the oldest carbonatite, C1, waS derived from a bonatites is important evidence in the refutation. magma somewhat different chemically from those He cited the experimental, field and petrographic producing intrusions C2through C5. Furthermore, it observations demonstrating conclusively that kim- appears from the magnetite core composition berlite magmadifferentiates to carbonate-richSiO2- rangesand zoning that C1was not precipitated from poor residua. He emphasizedthat although these a single, homogeneousbatch of magma. The varia- rocks would be termed carbonatiteaccording to the tions in magnetite compositions for the other car- petrographic classification of Heinrich (1966),they bonatite intrusions are simpler, and are consistent have very different mineralogies, antecedents,and with precipitation of each from a single batch of comagmatic rocks to the carbonatites of alkaline magma. The formation of C2 probably went to complexes. completion independently of the younger intru- Mitchell (1979)is, of course, correct to state that sions, C3 through C5. However, the rangesof core the carbonatitesofJacupiranga, and other associat- compositions of magnetites in C3 through C5 are ed with alkaline rock associations,have character- consistent with their precipitation from three istics, origins and histories different from those of successivebatches of magma which represent a carbonatites derived from kimberlites. It is most continuum in time and magmaticevolution. Wheth- unfortunate that the search for petrogenetic links er the magma precipitating C3 (and perhaps subse- between rocks of the two associationshas led to quently Ca and C5)was separatedfrom that precipi- confusedpetrogenetic speculations and misdirected tating C2 early in the C2 history, or whether a exploration efforts for niobium and diamond common parent yielded two successive magmas (Mitchell, 1979), but we do not think that this is which independently precipitated carbonatites C2 sufficient reason to terminate the intellectual exer- and then C3, cannotyet be determined.Nor do the cise of tracing the links between kimberlites and magnetitesinform us whether the carbonatite mag- carbonatites. There are two problems here, petro- mas were derived from a silicate parent magma by graphic or semantic, and petrogenetic. Mitchell crystallization differentiation, or by liquid immisci- (1979\recommended that the carbonatitesassociat- bility. More detailed investigation of the inclusions ed with kimberlites should be called "calcareous in magnetite and other minerals may yield informa- kimberlites" or "calcite-kimberlites", and he ob- tion about the compositions of the magmas,and of jected to the lax application ofthe term "carbonati- the elementssuch as alkalis which are poorly repre- tic" to the presumed interstitial carbonate-rich sentedin the carbonatites,presumably becausethe magma in the upper asthenosphere (Wyllie and GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA 209

Huang, 1975).It is within the upper mantle that the Haggerty, S. E. (1976) Opaque mineral oxides in tenestrial petrogeneticlinks arejoined. Kimberlites and "cal- igneous rocks. In D. Rumble, III, Ed., Oxide Minerals, p. cite-kimberlites", and alkaline complexes and car- Hgl0l-Hg300. Mineralogical Society of America. Heinrich, E. Wm. (1966) The Geology of Carbonatites. Rand bonatites, are two of the diverse products from McNally and Company, Chicago. upper mantle processes which are influenced by Herz, N. (1977) Timing of spreading in the South Atlantic: COz(Wyllie and Huang, 1976;Eggler,1978Wyllie, information from Brazilian alkalic rocks. GeologicalSociety of 1980).LeBas (1977,Fig. 24.1)presented a spectrum America Bulletin, 88, l0l-112. (1965) petrogenetic of ultramafic magmas from the mantle and their Irvine, T. N. Chromian spinel as a indicator. Part I. Theory. CanadianJournal of Earth Sciences,2, &8- evolution along five derivatives stems,with kimber- 672. lites at one end. and carbonatitesat the other. Koster van Groos, A. F. and Wyllie, P. J. (1966) Liquid Detailed studiesof carbonatitesand of carbonate- immiscibility in the system Na2O-Al2O3-SiOr4,Oz at pres- rich derivatives of kimberlites are a prerequisitefor sures up to I kilobar. American Journal of Science,2U,234- 25s. tracing their petrogenetichistories and that of their Koster van Groos, A. F. and Wyllie, P. J. (1973) Liquid magmatic antecedents back towards their mantle immiscibility in the join NaAlSi3OrCaAlSirO6-Na2CO3-H2O. source, and for eventual comprehension of why American Journal of Science.273. 465487. carbon in the upper mantle is conveyed to the LeBas, M. J. (1977) Carbonatite-NepheliniteVolcanism. John surface in different tectonic environments as car- Wiley and Sons, London. (1976) bonatites, containing phosphorus, niobium, and Lindsley, D. H. Experimental studiesof oxide minerals. In D. Rumble,III, Ed., OxideMinerals, p. L61-L88. Mineral- rare earth elements. or as calcite-kimberlite associ- ogical Society of America. ated with graphite or diamond. McMahon, B. M. and Haggerty, S. E. (1976)Oka carbonatite complex: oxide mineral zoning in mantle petrogenesis.Geo- Acknowledgments logical Society of America, Abstracts with Programs,8, 1006. We thank Serrana S/A de Minerac6o, G. C. Melcher, and McMahon, B. M. and Haggerty, S. E. (1979)The Oka carbona- V. A. V. Girardi for making possiblethe field work of JCG, L M. tite complex: magnetite compositions and the related role of Steele for invaluable assistancewith microprobe analysesand titanium in pyrochlore. In F. R. Boyd and H. O. A. Meyer, for the computer program to calculateFe3*, and J. V. Smith for Eds., Kimberlites, Diatremes and Diamonds: Their Geology, assistancewith the microprobe. Financial support was provided Petrology, and Geochemistry,p. 382-392.American Geophys- by ConselhoNacional de DesenvolvimentoCientifico e Tecnol6- ical Union. gico (CNPq)PROC 201.158/80 and PROC40.1410/81, Fundagdo Melcher, G. C. (1966) The carbonatites of Jacupiranga, S5o de Ampara d Pesquisado Estado de 56o Paulo (FAPESP)PROC Paulo, Brazil. In O. F. Tuttle and J. Gittins, Eds., Carbona- 80/470and PROC 8l/0168-8. and the Earth ScienceDivision of tites, p. 169-181.John Wiley and Sons,New York. the National Science Foundation Grants EAR 76-20410.and Mitchell, R. H. (1978) Manganoan magnesian ilmenite and EAR 8l-0859. titanian clinohumite from the Jacupiranga carbonatite, 56o Paufo, Brazil. American Mineralogist, 63, 544-547. References Mitchell, R. H. (1979)The alleged kimberlite--carbonatiterela- tionship: additional contraxy mineralogical evidence. Ameri- Amaral, G. (1978)Potassium-argon age studies on the Jacupir- can Journalof Science,279,570-589. anga alkaline district, State of 56o Paulo, Brazil. Proceedings Prins, P. (1972) Composition of magnetite from carbonatites. of the First International Symposiumon Carbonatites,p. 297- Lithos.5. 227-240. 302. DepartamentoNacional da ProducdoMineral. Brasilia. Reed, S. J. B. (1973)Principles ofX-ray generationand quantita- Boctor. N. Z. and Svisero. D. P. (197E)Iron-titanium oxide and tive analysiswith the electron microprobe. In C. A. Andersen, sulfide minerals in carbonatitefrom Jacupiranga,Brazil. Car- Ed., Microprobe Analysis, p. 53-E1. John Wiley and Sons, negie Institution of WashingtonYear Book, 77,876-E80. New York. Carmichael, L S. E., Turner, F. J., and Verhoogen, J. (1974) Secher, K. and Larsen, L. M. (19E0)Geology and mineralogyof IgrreousPetrology. McGraw-Hill Book Company, New York. the Sarfart6q carbonatitecomplex, southernWest Greenland. Eckermann, H. von (1974)The chemistry and optical properties Lithos, 13, 199-212. of someminerals of the Alnd alkaline rocks. Arkiv fcir Minera- V., and Windley, B. F. logi och Geologi, 5,93-210. Steele,I. M., Bishop, F. C., Smith, J. (1977)"Ihe Fiskenaessetcomplex, West Greenland, Part III. Eggler, D. H. (1978)The effect of CO2 upon partial melting of and oxide minerals from oxide bearing peridotite in the systemNa2O{aO-AlzOr-MgO-SiOz{Oz to Chemistry of silicate rocks, mostly from Gr/nlands Geologiske 35 kb, with an analysis of melting in a peridotite-H2O-CO2 Qeqertarssuatsiaq. Unders/gelse, 124, l-38. system. American Journal of Science,278,305-343. (l%6) Gittins, J. (1978) Some observations on the present status of Tuttle. O. F. and Gittins. J. Carbonatites.John Wilev and carbonatites studies. Proceedingsof the First International Sons.New York. Symposium on Carbonatites, p. 107-115.Departamento Na- Ulbrich, H. H. G. J. and Gomes,C. B. (1981)Alkaline rocks cional da Producio Mineral. Brasilia. from continental Brazil: a review. Earth ScienceReviews, 17, Gold, D. P. (l%6) The minerals of the Oka carbonatite and 135-154. alkaline complex, Oka, Quebec. Mineralogical Society of Wyllie, P. J. (1978)Silicate

the First International Symposiumon Carbonatites,p. 61-68. 0-2.83 DepartamentoNacional da Producdo Mineral. Brasilia. Wyllie, P. J. (1980)The origin of kimberlite. Journal of Geophys- -nn6 ical Research,85, Bl2, 6m2-6910. -3.18 Wyllie, P. J. and Huang, W. L. (1975)Periodotite, kimberlite, and carbonatite explained in the system CaO-MgO-SiO2- 0.00 CO2. Geology, 3, 621-624. Wyllie, P. J. and Huang, W. L. (1976)Carbonation and melting reactions in the system CaO-MgO-SiO2-{Oz at mantle pres- sures with geophysicaland petrological applications. Contri- butions to Mineralogy and Petrology, 54,79-107. Wyllie, P. J. and Tuttle, O. F. (1960)The systemCaG{O2-H2O and the origin of carbonatites.Journal of Petrology, l,146.

Manuscript received,February I I, 1982; acceptedfor publication, September7, 1982.

Appendix This Appendix contains additional descriptive material and figures, which are cited in the precedingtext.

Analytical procedure Standards used for the electron microprobe analyses were diopsideglass (for Ca, Si, and Mg), Fe metal (for Fe), synthetic rutile (for Ti), synthetic enstatite with lUVo (for Al2O3 Al), and RIM p RIM synthetic glass with I wt.Vo trlnO (for Mn). The amount of Fe occurring as Fe3* was calculated following the assumptions Fig. Al. Electron microprobe profiles for magnetitesof the Ct made by Steele et al. (1977'). sovite. Each point representsa single analysis. These data are The carbonatitesare heterogeneousrocks, and many analyses not tabulated. were completedin an attempt to representthe averagecompositions of magnetites, as well as the ranges of zoning in each carbonatite.For each carbonatite sampleanalyzed, three to four another. The vertical axis representsa constant scale in weight magnetiteswere selected,and four to six points were analyzedin per cent of oxide, but for graphical convenienceand to facilitate eachmineral, including points in the core and rim regions(Table comparisonofoxide variations, the profile for eachoxide plotted 2). Separateaverages were calculatedfor core and rim determi- is arbitrarily located with respect to the vertical scale. The nations in each rock sample. In addition to this, details of the absolutevalues for each figure are given at one end in weight per zoning were determined by analyzing a series of points across cent. selectedmagnetite grains (each point is a single analysis). Figure AlB, for carbonatite C1, illustrates simple, regular and The magnetitesfrom the reaction rock with exsolved ilmenite symmetrical zoning from core to rim. The other profiles in lamellae had irregular form and no regular pattern of variation in FiguresAl through A5 show broadly similar trends, but there are composition. Therefore, average values for core and rim were irregularities and reversals compared with the trends in Figure not obtained. The values presentedin Table 3 are the averageof AlB. In general,the rate of chemical variation increasestoward all data for each rock sample; the individual analyses cluster the rims of the minerals. These profiles provide information fairly closely around these values. The magnetiteintergrowth with sulfide occurred in very small patches,and only a few points could be analyzed.These points showed little variation in chemical composition. Averages for each magnetiteintergrowth are given in Table 3. Two crystals of magnetite from the jacupirangite were analyzed, one zoned (core and rim analyses),and another without ).-t.30 zoning. Severalpoint analyseswere performed near the ilmenite t;-71.7 lamellae, and the result obtained for a point nearest to one \-tA7 lamella is shown in Table 3. To.*r, I Individual zoned crystals Electron microprobe traversesacross individual mineralsfrom each of the five intrusions are presentedin Figures Al through A5. The traversesextend either from rim to rim passingthrough the centers of the mineral sections, or from core to rim. The Fig. A2. Electron microprobe profiled for magnetitesof the C2 horizontal distance scale varies from one mineral section to dolomitic sovite. GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA 2II

magnetitesin the different intrusions, it is evident that there are diferences in composition between the magnetites; note, for example, the values for MgO and TiOz in the profiles.

Core compositionsand ranges of zoning Figures A6 and A7 show projections of mineral compositions from the spinel prism of Figure 2, as described in the text for Figures 4 and 5. The analyses are projected onto a vertical Iot"z section through the prism in Figure ,4.6,and onto the rear face of the prism in Figure A7. The range ofvalues depicted for one of the C2 analysesis representativeof the ranges from which the averagevalues were derived. RIM RIM In Figures 48, A68 and A7, the averagecore composition of Fig. ^A3.Electron microprobeprofiles for magnetitesof the Ca eachmagnetite is connectedto the averagecomposition ofits rim sovite. (Table 2). Every pair ofanalyses has a component ofincreasing FeO(FeO+MgO) from its core to rim, indicating that this ratio has a time relevance, which is consistent with the relationship about the sympatheticand antipathetic variationsoftrivalent and illustrated for the cores of magnetitesin C2, C3, Ca, C5 in Figure tetravalent cations in relation to the divalent cations, which determinethe most important end-membercomponents present, The averagemagnetite core compositionsfor each carbonatite and their distribution in the zoned minerals. intrusion are enclosed by lines which were drawn with two The profiles in Figure Al for magnetites in C, illustrate criteria: (l) to fit the boundariesaround the points as closely as antipathetic behavior of TiO2 and Fe2O3,indicating a variation possible, and (2) to draw the boundaries so that there were no from core to rim in the ulvospinel-magnetitesolid solution series, reversals in direction with respect to FeO(FeO+MgO). The with fairly constant Al2O3. In contrast, Figure A2 shows fairly resultsdepict fairly narrow chemical bandsfor the core composi- constant TiO2 for magnetitesfrom carbonatite C2, with antipa- tions in each of the intrusions and, although additional analyses thetic behavior between Al2O3and Fe2O3.A similar relationship could broaden the areas, and fill in the re-entrant angles, the exists between Al2O3 and Fe2O3for magnetitesin Cr (Figure results indicate that the chemical variation trends for the core A3), but there is in addition some variation in TiO2. The zoning magnetitesin each of the five intrusions are distinct from each in the examplesfrom Ca in Figure ,{4 shows variations similar to other. The separationofC3, Ca, and C5was already evident from those in Cr. The single example for C5 in Figure A5 also shows Figure 3. Although C1 overlaps with the other four intrusions in antipathetic behavior between Al2O3 and Fe2O3, with little Figure 3, the trend is distinct from that of the others in Figures changein TiOz. Despite the repetition of zoning patterns from 4,4'and 5, and distinguishablein Figure 4'6,4..The trend of Ca obviously differs from that of C2 in Figures 4,4'and 5, although the C2 composition boundaryjust reachesthe Ca boundary. The trend of C2differs from that of C3 in Figures4A, A6,4, and 5, but the separation is not convincing. However, the separation of core compositions from C2 and C3 is confirmed in Figure 6, -0.28 where the Mn content is plotted againstthe Mg content. Figure 6 r-69.2 also separatesthe magnetitecores of C1from thoseof C2and C3.

C5 RIM C+ lo.'** 0e-0.17 -68.6 -0.84

0.5wt %

0 20 40 60 CORE lL RIM RIM CORE Fig. A'4. Electron microprobe profiles for magnetitesof the C4 Fig. A5. Electron microprobe profile for a magnetiteof the C5 sovite. rauhaugite. 212 GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA

consistent with the anticipated decreasein MgO as ilmenite is N o exsolved from magnetite (Lindsley, 1976).The magnetite from F + the reaction rock in C3 contains more MgO than that in the n O jacupirangite analyzed, and it departs significantly from the aN reaction trend indicated by the dashed line. A similar relation- + ship is exhibited in Figure 88. Extrapolation of the trend n o indicated by the threejacupirangite magnetites,and the zoning, N reachesthe magnetitesin the reaction rocks from C1 and C5, but that from C3 is separated from the others by its high MgO o o 0.6 01 0.8 0.9 lo 0.6 0.7 0.8 0.9 to content. a Figure 9 shows the relationship between Mn and Mg. The FeO/(Fe0+MgO) FeO/(FeO+MgO) magnetitesfrom thejacupirangite establisha line, but extrapola- Fig. 4,6. Average core and rim compositions of magnetites tion of the line does not l€ad to the magnetitesin the reaction (Table 2) in projections from the spinel prism (Figure 2). A bands. These plot on a subparallel line following the trend of representativerange of values is shown for one of the points. A. magnetites in the carbonatites. As the line defined by the Core compositions. See Figure 3 for symbols. B. The core jacupirangite magnetites represents the depletion of Ti due to compositions connected by lines to corresponding rim exsolution of ilmenite, this sub-parallel line indicates that al- compositions(dots) showing zoning trends. though the reaction rock magnetitesmay be derived from the jacupirangite magnetites,as indicated above, they are not produced by simple exsolution of ilmenite.

The magnetites in beforsite dikes Ba and 85, respectively, Comparisonwith previous resultsfrom the Jacupiranga carbon- have FeO/(FeO+MgO) equal to the low end of the rangefor Ca, atite and equal to that (Fig. in Cs A6). They have the samevalues for In Figure 10, one of the magnetitesreported by Boctor and Fe2O3/(Fe2O.+TiOz) as magnetitesin Ca and C5, distinctly Svisero plots in the C2 range and the other two plot near the Cr higher Al2O3/(Al2O3fFe2O3*TiOz) and Al2O3/(Al2O3+TiO2), range. Of the two that plot near the C1 magnetites,one is from and (Figs. slightlyhigher MnO 4, 5, 6, and 4.6,{). olivine sovite, and the other is exsolved magnetitefrom phlogo- Figure ,4,6B illustrates a steady decreasein Al2O3 with de- pite sovite. Some of the C1 sovites are very rich in olivine, but creasing MgO from cores to rims, with only one exception. This we found no exsolvedmagnetite in our samples.In one of our C1 trend is consistent with the broad trend for the magnetitecore compositionsin Figure A6,4.. The main zoningillustrated in Figure A7 (seeFig. 5) is directed towards the lower right-handcorner, representingthe ulvospinel composition, with Al2O3decreasing relative to TiOz. This is also the dominanttrend for the magnetitecores in Figure 5. However, four zoning lines reverse this trend. The zoning illustrated in Figure A7 is not a result of substitutionsbetween Al-spinels (top of prism in Figure 2) and Ti-spinels(rear edgeof the prism), but a consequenceof the more rapid decreaseof Al than Ti from core to rim (except for a few C1 magnetites).This is evident if Ti is plotted againstAl. oN The progressivedecrease in Mn contents of core magnetites ;- with decreasingMg content illustrated in Figure 6 is accompa- T 0.6 nied by a progressivechange in the slopesofthe zoning lines as ot) shown in Figure AE. A I :1 replacementline is drawn as a basis -s 0.5 for comparison. For magnetiteswith the highest Mg contents, two of the zoning lines are consistentwith the l: I substitution. However, with decreasingMg content, the core compositions d 0.4 exhibit a broad trend converging toward the origin, and the s- zoning lines changedirection fairly regularly until they, too, are 0.3 directed towards the origin. This illustrates a changein zoning, from enrichmentin Mn towards the rims for high-Mg magnetites, to depletionin Mn towards the rims for low-Mg magnetites.Note 0.2 the three exceptions for high-Mg magnetites. Small marginal variations in MnO contents can be seenin the zoning profiles of 0.1 Figures AlA, A2, and ,{3.

Magnetitesfrom special locations 0.0 The three magnetitesfromjacupirangite plotted in Figure 8A FeO/(Fe0+MgO) define a straight line which extrapolatesto the magnetitesin the reaction bands from C5 and C1. The zoning in a magnetitefrom Fig. A7. Average core and rim compositions of magnetites jacupirangite is parallel to the same line. The line is inclined, plotted as in Figure 5. For symbols see Figure 3. GASPAR AND WYLLIE: MAGNETITE IN CARBONATITES FROM JACUPIRANGA 213

0.4 Comparison with magnetitesfrom other carbonatites =