Journal of Petrology | Pages |
© 1998 Oxford University Press |
Most of the observational input to hypotheses about mantle plumes and their associated magmatism has come from studies of phenomena in ocean basins, where the lithospheric plates have both relatively uniform compositions and thicknesses that appear to conform well with those predicted by plate-tectonic theory. Even the thickest oceanic plates are thin enough to permit an underlying hot mantle plume to decompress sufficiently to produce `dry' basaltic (s.l.) melt (e.g.
Watson & McKenzie, 1991;
Farnetani & Richards, 1995;
Cordery et al., 1997). Therefore a plate drifting over an essentially fixed-position mantle plume collects a linear belt of excess migrating magmatism (seamount chain or aseismic ridge) that defines a direction and velocity of absolute plate motion
(Morgan, 1983). A nomenclature for terms referring to plume-plate interaction needs to be established at the outset, to prevent subsequent ambiguity. In this paper, lithospheric plates drift over fixed-position mantle plumes. The magmatism produced above the latter appears to an observer to migrate across the plate surface. Finally, a plume-track is a line on the surface of a plate that depicts that plate's movement, relative to a fixed plume below.
Belts of migrating magmatism are scarce in the continental geological record. This can hardly be because mantle plumes rarely impinge beneath continents.
McKenzie, (1984) pointed out that a rapidly rising mantle plume-head, originating at the core-mantle boundary, is essentially unaware whether it is upwelling beneath an ocean or a continent far above. Therefore, plumes should impact and impinge beneath continents as frequently, per unit area, as they are observed to occur within the ocean basins. Numerical models that suggest some long-timescale coupling between continental plates, especially very large ones, and plume spatial distribution (e.g.
Richards et al., 1997) do not invalidate this short-timescale relationship. The physical evidence for many sub-continental mantle plume impacts is found in large igneous provinces. It is the post-impact belts of migrating magmatism, formed as continents drift over the fixed stems or tails of mantle plumes
(Davis, 1995), that are rare.
Hill, (1991),
Thompson & Gibson, (1991),
Sleep, (1996,
, 1997) and others have suggested that the relatively deep lithospheric roots of a craton might deflect the hot upwelling mantle of a plume to adjacent regions with thinner lithosphere. We shall argue below that this occurred in SE Brazil.
Continuing plume-related ocean-basin magmatism at Tristan da Cunha and Trindade-Martin Vaz has built islands above sea level that are linked by an aseismic ridge and a seamount chain, respectively, to the continental margin of SE Brazil. Plate reconstructions at ~135 Ma and ~85 Ma place the Tristan and Trindade `hotspots', respectively, beneath or close to SE Brazil
(Crough et al., 1980;
Hartnady & le Roex, 1985;
Fleitout et al., 1989;
O'Connor & Duncan, 1990;
Müller et al., 1993), at times when magmatic activity also peaked (see below). Adopting the view that geologically younger magmatism in the adjacent South Atlantic ocean basin gives an a priori reason for suspecting mantle plume involvement in this Cretaceous magmatism, we have investigated whether straightforward application or extension of current theory can explain both the existing data and those accumulated by our own research
(Thompson & Gibson, 1991,
, 1994;
Gibson et al., 1995,
, 1996,
1997c).
Gibson et al., (1995) showed that widespread strongly potassic alkaline magmatism in the Alto Paranaíba region (
Fig. 1) was synchronous at ~85 Ma, the time that plate-tectonic reconstruction places the Trindade plume close to this region. They interpreted this as an example of a mantle plume that first impacted beneath continental lithosphere too thick to permit `dry' decompression melting and hence basalt (s.l.) production. Instead, the volatile-rich alkaline magmas, including diamondiferous kimberlites, were generated by uplift and heating of the lithosphere above the plume head.
Gibson et al., (1997c) showed that the ~85 Ma magmatism extends as far NW as Mato Grosso (
Fig. 1), where kimberlites (some diamondiferous) occur in cratonic areas but a pre-existing rift, i.e. a site of earlier lithospheric thinning, preserves alkali-olivine basalts of the same age. This is the way that the magmatism might be expected to behave when a starting-plume head, impacting beneath thick lithosphere, encounters a pre-existing local `thinspot'
(Thompson & Gibson, 1991;
Sleep, 1997).
In this paper we focus on alkaline magmatism in SE Brazil that post-dates the ~85 Ma Trindade starting-plume impact-related activity of Alto Paranaíba and Mato Grosso but pre-dates the oldest offshore volcanism at the landward end of the Trindade-Martin Vaz seamount chain. This younger activity built the Abrolhos Platform at 52-42 Ma
(Cordani, 1970;
Fodor et al., 1989). We shall present evidence that, between ~85 and 55 Ma, the Trindade mantle plume tail passed beneath lithosphere of the Archaean São Francisco craton (
Fig. 1) that was thick enough to prevent the generation of plume-derived magmas by decompression melting. Instead, we propose that part of the hot plume mantle was deflected several hundred kilometres southward, towards the thinned lithosphere of the continental margin, where it decompressed and generated an eastward migrating sequence of alkaline igneous complexes. The Late Cretaceous to Early Tertiary complexes that are discussed in this paper occur along and near the SE Brazilian coast, between longitudes 42° and 47°W (
Fig. 2). They form a distinct igneous province because further west, with the exception of the tiny (<1 km2) ~85 Ma syenite outcrop at Cananéia (24°58€S, 47°54€W), the alkaline complexes are mostly much older (~130 Ma). Previous Brazilian accounts (e.g.
Ulbrich & Gomes, 1981) have tended not to treat this province as a single entity but instead to focus on individual complexes (e.g. Poços de Caldas) or small groups of complexes, such as São Sebastião and adjacent islands or the plutons around Rio de Janeiro. We hope that the following account will convince readers that that we are justified in grouping these complexes to form the Serra do Mar igneous province.
The metamorphic basement in the region of the Serra do Mar province comprises the partly Archaean São Francisco craton in the north and a mobile belt between this and the coast that gives radiometric dates as young as Lower Palaeozoic. The São Francisco craton has recently yielded ages amongst the oldest terrestrial rocks
(Teixeira et al., 1996), including a well-dated anorthositic sill at 3·12 Ga (zircon U/Pb) and evidence from both zircons and Rb/Sr systematics of gneisses that are as old as ~3·38 Ga. Before final stabilization, this craton underwent major reworking at 2·86-2·70 Ga and intrusion by granites between 2·7 and 2·6 Ga. Thereafter, its SE margin was again reworked between 2·4 and 2·0 Ga during the Transamazonian orogeny
(Machado et al., 1996). The youngest K-Ar mineral ages from this region are ~1·8 Ga
(Teixeira et al., 1996).
Between the São Francisco craton and the coast, the pattern of radiometric ages is entirely different. The results of
Fonseca et al., (1995) and
Machado et al., (1996) show that the majority of the published dates from this Ribeira belt fall in the range 750-450 Ma, spanning the Brasiliano-Pan-African orogeny. The remaining dates are older, up to 2·3 Ga, and suggest that some (possibly most) of this mobile belt is reworked Transamazonian basement. Many of these older dates come from the gneisses that outcrop around Cabo Frio and form the offshore island of Búzios (
Fig. 2).
Fonseca et al., (1995) suggested that these two localities are fragments of the reworked eastern margin of the Congo-Kasai craton, overprinted by the Brasiliano-Pan-African orogeny.
Figure 1 (inset) illustrates this concept.
Teleseismic tomographic studies suggest that the lithospheric thickness beneath the São Francisco craton is 200-250 km
(Grand, 1994;
VanDecar et al., 1995). The offshore basins along the Brazilian coast have all been the subject of intensive geophysical studies, during hydrocarbon exploration programmes. The segment offshore from the Serra do Mar igneous province, the Santos basin, is a classic example of a continental passive margin, with strong crustal (and therefore presumably lithospheric) thinning related to rifting during the initial opening of the South Atlantic (e.g.
Chang et al., 1992;
Mohriak
et al., 1995).
Chang et al., (1992, table 1) gave an estimated average [beta] factor of 2·88 for the Santos basin. Lithospheric thickness beneath the Ribeira belt has not been investigated geophysically. Nevertheless, geological studies show that it is reasonable to suppose that it is considerably thinner than the São Francisco craton. Thus
Gresse et al., (1996) documented late- to post-orogenic basin formation throughout the Brasiliano-Pan-African belts of southern Africa and southern Brazil.
Molyneux, (1997) has shown that strong extension accompanied late Brasiliano granite emplacement in the Ribeira belt.
To the west of the Serra do Mar province the metamorphic basement is overlain by the deposits of the Paraná basin, which has had a multiphase extensional history since the Late Ordovician. The Poços de Caldas complex (
Fig. 2) is the only unit of the Serra do Mar magmatism to intrude Paraná basin sediments locally, namely the Jurassic Botucatu sandstones. Finally, there are minor extensions of the Santos basin system onshore, as Middle Tertiary grabens within the Ribeira belt (
Fig. 2). Their Neogene infillings include arenites, shales (some bituminous) and limestones
(Klein & Valença, 1984;
Schobbenhaus et al., 1984). This study of Serra do Mar magmatism was aimed specifically at improving knowledge of its age and investigating the mantle sources of the alkaline magmas. Therefore both our fieldwork and the following account are selective. The bulk of the complexes is a range of syenites, sometimes accompanied by trachytes and/or phonolites. Associated mafic rocks are uncommon and mostly confined to dykes.
Woolley, (1987) gave an excellent overall account of these complexes, summarizing all previous publications. Subsequent contributions have come from
Bellieni et al., (1990),
Brotzu et al., (1992,
, 1995,
1997),
Shea, (1992),
Regelous, (1993),
Amorim et al., (1995),
Araújo et al., (1995),
Garda et al., (1995),
Montes-Lauar et al., (1995),
Morbidelli et al., (1995) and
Valente et al., (1995). We shall describe the syenites and other felsic rock-types in general terms and then give more detailed accounts of the mafic rock-types sampled during this study.
Syenites and related rock-types
The syenite-phonolite-dominated complexes range in size from small isolated phonolite plugs to Poços de Caldas (800 km2), the largest alkaline complex in Brazil and one of the largest in the world. The smaller Itatiaia complex (330 km2) is equally impressive, as it forms a 31 km * 12 km rugged mountain massif, 2787 m in height. Many of the other syenite complexes (e.g. Tinguá, Rio Bonito, Soarinho and Morro de São João,
Fig. 2) also form spectacular forest-clad peaks. Small remnants of the volcanic superstructures of these plutonic complexes are preserved locally
(Woolley, 1987): Poços de Caldas includes tuffs, breccias, agglomerates and mafic to phonolitic lavas; Tinguá includes phonolitic lavas and tuffs; Mendanha comprises predominantly volcanic agglomerates, tuffs and breccias, containing clasts of a wide range of alkaline rock-types, together with ignimbrites
(Klein et al., 1984). Thus, the Serra do Mar plutonic complexes most probably originated beneath a chain of alkaline volcanoes comparable with those along the East African rift system. The volcano capping Itatiaia may have resembled the present-day Mts Kenya and Kilimanjaro, and Poços de Caldas may have underlain a similar-sized or even larger volcano or volcano group; it is thought to be encircled by a caldera fault ~33 km in diameter
(Schorscher & Shea, 1992).
The majority of the syenites
(Woolley, 1987) contain abundant nepheline, although others are nepheline free. Some have biotite and green amphibole as their main ferromagnesian minerals, accompanied by melanite garnet in the Rio Bonito, Tinguá and Morro de São João complexes. Other common members of all the felsic complexes are peralkaline syenites-phonolites, with sodalite, aegirine, arfvedsonite and occasionally eudialite. Despite the impression given by their mineralogy and chemical compositions (e.g.
Bellieni et al., 1990;
Brotzu et al., 1992), the syenites may originally have been considerably more potassic than they appear to be at present because pseudoleucite is widespread
(Woolley, 1987), with leucite-shaped polyphase pseudomorphs composed of orthoclase, nepheline and zeolites, together with minor amphibole, muscovite and carbonate
(Valença & Edgar, 1979). These are especially abundant in the Morro de São João complex, where we have observed individual pseudomorphs up to 4 cm in size. The largest known pseudoleucites in this province are 29 cm at Tinguá
(Lima, 1976). Post-crystallization alteration of leucite strongly affects the K/Na ratio of the rock (e.g.
Thompson et al., 1997c).
It has been generally considered that the Serra do Mar province syenites-phonolites originated by fractional crystallization from mafic parental magmas
(Woolley, 1987). These melts are specified in recent published accounts to be silica undersaturated and relatively sodic basanite-tephrites
(Bellieni et al., 1990), basanite
(Brotzu et al., 1992,
, 1995) or camptonite and soda vogesite
(Araújo et al., 1995). Nevertheless, both
Brotzu et al., (1992,
, 1995,
1997) and
Amorim et al., (1995) noted that the syenites-phonolites of the Serra do Mar province have significantly higher K2O/Na2O ratios than those within the older complexes of the Ponta Grossa province to the SW (
Fig. 1), and they speculated that the Serra do Mar mafic parental magmas may likewise be relatively potassic.
Schorscher & Shea, (1992) made the same point and invoked a `primary magmatic perpotassic characteristic' in the parent magma at Poços de Caldas.
Mafic rock-types
The mainland igneous complexes of the Serra do Mar province form an approximately linear array, ~500 km in length, from Poços de Caldas to Cabo Frio. Another group is exposed as offshore islands along the coast south of Poços de Caldas (
Fig. 2). The following account summarizes our own observations. The mafic-ultramafic rock-types occur in four situations in this province: (1) dykes within the country rock up to ~10 km away from each complex, although most of these are felsic; (2) dykes cutting the syenites and associated felsic central plutons; (3) lavas and agglomerate blocks within remnants of volcanic superstructures; (4) layered gabbros associated with the syenites (in two instances). Coastal SE Brazil is a subtropical region with high rainfall. Consequently, lateritization is frequently intense and affects all rock-types on low ground. Mafic samples suitable for igneous geochemistry are confined to river gorges, coastlines, quarries and some roadsides.
Poços de Caldas. Brecciated lavas, polylithic agglomerates and tuffs are exposed for ~14 km along the road between Cascata and Águas da Prata. The mafic blocks are feldspar-free leucitites, containing phenocrysts of clinopyroxene, leucite (also in the groundmass) and magnetite (see
Table 2, 90SB68), olivine (90SB66, 67, 72, 79) and occasional phlogopite (90SB64); all phases are variably altered. The Osamu Utsumi open-cast uranium mine intersects several fresh mafic potassic dykes, cutting hydrothermally altered phonolites.
Shea, (1992) and
Waber et al., (1992) discussed the petrography, geochemistry and classification of this rock-type but did not give chemical analyses. Judging from the description given, their proposal that these dykes are lamproites is perhaps incorrect, in the light of current nomenclature
(Woolley et al., 1996). They are probably kamafugites, an extremely common rock-type in the penecontemporaneous Alto Pananaíba igneous province
(Gibson et al., 1995), immediately to the north (
Fig. 1).
Campos do Jordão. The Campos do Jordão region lacks a well-defined intrusive complex and major occurrences of leucocratic rock-types. Instead it includes the Ponte Nova pluton, a sill or stock of alkaline gabbro to picrite, and a scattered swarm of basalt (s.l.) to picrite dykes, up to 1 m in width. The Ponte Nova pluton is predominantly nepheline melagabbro, with associated leucogabbros and syenitic veins. It is cut by dykes ranging from micro-melagabbro to phonolite (93SOB205, 206), and also by members of the swarm (93SOB198). All the dykes in the swarm are feldspar-free olivine melanephelinites. They have olivine and clinopyroxene phenocrysts (93SOB198, 208-13), sometimes accompanied by megacrysts of clinopyroxene (198, 213), olivine and phlogopite (213). Groundmass minerals are clinopyroxene, olivine, opaques and minor amphibole and biotite, in a matrix of nepheline (in the two picrites, 208, 209) or glass (in the other samples).
Passa Quatro. A quarry in metamorphic basement gneiss adjacent to this complex is cut by two mafic alkaline dykes trending 280° (94SOB90-92), containing clinopyroxene, olivine and magnetite phenocrysts (plus amphibole and phlogopite in 90), together with apatite microphenocrysts, in a groundmass of clinopyroxene, amphibole, biotite, opaques and local alkali feldspar, with interstitial glass.
Itatiaia. Mafic dykes were found cutting the syenites at two localities. Between Moringa and Cachoeira a 2 m wide dyke (94SOB95) contains spinel lherzolite xenoliths, olivine and clinopyroxene phenocrysts in a groundmass of these phases, opaques, plagioclase and interstitial glass. At the eastern end of the complex the road from Resende ascends the scarp of the Resende graben (
Fig. 2) and exposes several mica-rich dykes (94SOB97, 98). These have clinopyroxene and magnetite phenocrysts (97) or clinopyroxene, olivine and phlogopite microphenocrysts (98), in a groundmass of these phases, amphibole, alkali feldspar, apatite and glass.
Tinguá. The upper slopes of Tinguá are densely forested and there is thick laterite around the base. Fresh leucocratic dykes are abundant but only one mafic example was found; a weathered 2 m wide minette (s.l.) with fresh phlogopite and altered olivine and clinopyroxene phenocrysts, exposed during 1994 land clearance at 22°37€19'S, 43°26€21'W.
Mendanha. Dykes from the swarms surrounding this complex are excellently exposed in large quarries within the metamorphic basement gneisses at Sentíssimo, Bangu and Nova Iguaçu. Most of the dykes trend 245°-260° but a sub-set at Nova Iguaçu trends 225° and one cross-cutting dyke at Sentíssimo trends 140°.
Valente et al., (1995) recorded mafic alkali-olivine basalts, basanites, tephrites, nephelinites, trachytes and phonolites in the Rio de Janeiro area. Of the dykes collected during the present study, the Nova Iguaçu suite includes intermediate and trachytic members, but most are mafic-ultramafic. Plagioclase-rich basaltic dykes occur sparsely in the swarm but all analysed samples (see
Table 2) are plagioclase poor to plagioclase free, with phenocrysts of olivine and clinopyroxene, joined by chromite in the picrites (94SOB64, 66, 67). Olivine phenocrysts are less abundant in more evolved compositions (68, 78, 80) and phenocrysts of magnetite, phlogopite, amphibole (57, 58, 61) and apatite (81) join the clinopyroxene. The two sodic Bangu picrites (66, 67) have clinopyroxene phenocrysts with green cores and groundmass clinopyroxene, olivine, opaques and amphibole, together with only traces of biotite. The other samples have the same groundmass phases (plus interstitial glass) but in 58 and 81 there is plentiful biotite and no amphibole.
São José do Itaboraí.
Klein & Valença, (1984) described `ankaramitic' dykes and pillow lavas within the Late Cretaceous to Palaeocene limestones filling a small graben. Exposure is minimal. The two samples analysed in this study both have olivine, titanaugite and chromite phenocrysts (94SOB50, 51). Some of the clinopyroxene phenocrysts in 51 have bright green centres. The groundmass in both samples contains olivine, clinopyroxene and opaques, together with accessory plagioclase, apatite and interstitial zeolite; biotite is also prominent in 50.
Tanguá, Rio Bonito and Soarinho. Although trachyte and phonolite dykes are abundant around this cluster of complexes, fresh mafic samples were only found NW of Soarinho, as boulders eroded from dykes cutting metamorphic basement in a river gorge. They are all minettes, with phenocrysts of phlogopite either alone (94SOB131) or plus olivine (127), together with magnetite, apatite and sanidine (128, 129). Their groundmasses are all rich in biotite and sanidine.
Morro de São João. Published accounts
(Woolley, 1987) mention ultramafic rock-types at this centre but the most mafic sample located in this study is a nepheline melasyenite (94SOB142).
Cabo Frio. Abundant trachyte and phonolite dykes, with sanidine, phlogopite, nepheline and leucite phenocrysts, are exposed on the sea cliffs around this complex. We located three thin (<1 m) dykes that are considerably more mafic than those reported previously
(Araújo et al., 1995). Two have clinopyroxene and olivine phenocrysts, with groundmass of clinopyroxene, opaques, plagioclase, nepheline(?), amphibole and apatite (93SOB193, 197). One has additional phlogopite phenocrysts and a groundmass of clinopyroxene, opaques, biotite, apatite and devitrified glass (196).
Trigo. This offshore island is composed of both nepheline syenite and cumulus-textured, layered alkali gabbro. Mafic to leucocratic dykes are relatively common along its coast. The mafic dykes have olivine and clinopyroxene phenocrysts (94SOB14, 17, 20), with groundmass plagioclase (plus nepheline in 14), amphibole, biotite and opaques, plus accessory apatite. Amphibole, magnetite and apatite join the phenocryst assemblages in more evolved compositions.
São Sebastião (Ilhabela). The island of São Sebastião, also known as Ilhabela, lies offshore from the mainland port of São Sebastião and a stretch of coast rich in alkaline mafic dykes. The latter have been studied in detail by
Regelous, (1993) and
Garda et al., (1995). The metamorphic basement of Ilhabela is cut by three syenite stocks and a small cumulus-textured layered gabbro that contains hypersthene. Alkaline dykes occur in the basement areas and also cut the syenites. On the NE coast it is clear that post-syenite mafic dykes were emplaced before the syenite was completely solidified; the dykes have broken into joint blocks that have subsequently rotated. In basement areas, the alkaline dykes have to be separated from large numbers of tholeiitic quartz dolerites belonging to a coast-parallel Paraná swarm, 129-136 Ma in age
(Regelous, 1993), that is prominent throughout the coastal region of
Fig. 2. Fortunately, the tholeiites are distinctive petrographically and on Ilhabela they are clearly metamorphosed by the syenites. The Ilhabela alkaline dykes analysed for this study have titanaugite and olivine phenocrysts (90SB54, 94SOB31), in a groundmass of clinopyroxene, olivine, opaques, apatite and interstitial zeolite and carbonate (90SB54), or amphibole, biotite, plagioclase, clinopyroxene, opaques and apatite (94SOB31). The groundmass texture of the latter sample is equigranular and hornfelsic (beerbachitic), consistent with the field evidence (above) that some of these dykes penetrated still-hot syenite. The mainland São Sebastião coastal mafic dykes mostly trend ~150° and are aphyric (94SOB12), or have olivine phenocrysts (13), or clinopyroxene microphenocrysts (10). The groundmass of the picrite (13) is exceedingly fine grained but seems to have the same essential phases as the other dykes: clinopyroxene, olivine, biotite, opaques, apatite, sodic plagioclase (12) and pseudomorphs after a mineral with the shape of nepheline (10).
Garda et al., (1995) recorded melilite in one coastal dyke.
Mafic alkaline magmatism of Mid-Tertiary and uncertain age
At first sight this topic falls outside the scope of the paper. But the mid-Tertiary basalts (s.l.) of coastal SE Brazil (
Fig. 2) potentially throw light on the geochemistry of the preceding Serra do Mar magmatism because some of them contain spinel lherzolite xenoliths. They may therefore offer a chance to deduce the nature of the mantle beneath this region, once it had drifted well clear of the Trindade mantle plume.
Volta Redonda. A small graben, 4·5 km SE of Volta Redonda, strikes ~290° and is infilled with Neogene Resende Formation sediments, including a basaltic lava flow and associated agglomerates. The lava (94SOB89) has phenocrysts of olivine, titanaugite and chromite, in a groundmass of olivine, clinopyroxene, opaques, accessory apatite, and interstitial plagioclase and zeolite.
Estrada do Grumari and Fazenda Modelo. Fragments of two lherzolite xenolith bearing basaltic dykes along the coast west of Rio de Janeiro (
Fig. 2) are of unknown age but closely resemble the Volta Redonda lava. Both have olivine, titanaugite and chromite phenocrysts (94SOB55, 56), together with lherzolite xenoliths (plus feldspar xenocrysts in 55). Their groundmasses are composed of clinopyroxene, olivine, opaques, plagioclase and apatite.
Resende. This relatively large graben (
Fig. 2) has no associated igneous rocks but, as the Itatiaia complex borders it to the north, it is conceivable that the Itatiaia dyke (94SOB95), which closely resembles 94SOB55, 56 and 89, may also have a Mid-Tertiary age. Currently its age is unproven. In this section we summarize published radiometric dates, critically assess them and discuss them in the light of new determinations.
The Mesozoic and Early Tertiary alkaline magmatism of southern Brazil and Paraguay has been the subject of many radiometric dating programmes during the last three decades. Compilations of these dates, with or without discussions of their significance, include those of
Herz, (1977),
Ulbrich & Gomes, (1981),
Woolley, (1987) and
Sonoki & Garda, (1988). The Sonoki & Garda paper was a benchmark because it listed all previous K/Ar dates, at a time when few determinations by other techniques were available. The number of new radiometric dates published since has been small but these include a higher proportion by such techniques as Rb/Sr isochrons and 40Ar/39Ar (see below).
A histogram of all the
Sonoki & Garda, (1988) dates (
Fig. 3) shows two clear peaks, at 130-140 Ma and at 80-90 Ma. Recent papers have used 40Ar/39Ar and Rb/Sr techniques to show that 130-140 Ma corresponds to the Paraná-Etendeka flood-basalt event (e.g.
Regelous, 1993;
Milner et al., 1995;
Renne et al., 1996;
Turner
et al., 1996). There is widespread, although not unanimous (e.g.
King & Anderson, 1995), agreement that this flood-basalt province was generated directly or indirectly in response to the initial impact of the mantle plume now sited beneath Tristan da Cunha. In contrast, there has been considerable reluctance to consider a similar model for the late Cretaceous and younger alkaline magmatism.
Gibson et al., (1994,
, 1995,
1997c) have shown that a combination of new radiometric dates and careful checking of the quality of published ones indicates that the alkaline magmatism (kamafugites, kimberlites, lamproites and local basalts) throughout the large area of Alto Paranaíba, the Iporá-Rio Verde region of southern Goiás, and the adjacent part of Mato Grosso (
Fig. 1) all occurred in the 80-90 Ma timespan, with a peak at ~85 Ma. They attributed this widespread synchronous magmatism to the initial impact of the mantle plume that is currently beneath Trindade-Martin Vaz.
If the <100 Ma dates in the
Sonoki & Garda, (1988) compilation from the Alto Paranaíba, Iporá and Mato Grosso regions are excluded, the majority of the remainder are samples from the Serra do Mar province. This alkaline magmatism has generally been attributed to passive continental margin extension (e.g.
Ulbrich & Gomes, 1981;
Brotzu
et al., 1995;
Morbidelli et al., 1995). One of our main initial objectives was to investigate whether the Serra do Mar province magmatism occurred synchronously or in a random time pattern, as might be expected if it resulted from plate-margin extension, or whether it initiated progressively later from west to east, as would be consistent with the South American plate drifting over a mantle plume. Use of the data in the
Sonoki & Garda, (1988) compilation unscreened and alone appears marginally to support the extension model.
Figure 4 shows the age ranges for each Serra do Mar province complex, projected onto a line between São João da Boa Vista and Cabo Frio (
Fig. 2). It should be noted that, although most of the centres fall relatively close to this line, the offshore islands of Trigo, São Sebastião (Ilhabela), Vitória and Búzios lie far to the south. The unscreened data shown in
Fig. 4 imply that several of the Serra do Mar volcanoes were active for about an order of magnitude longer than either any known Pliocene-Recent volcanoes or volcano groups, or older volcanic centres that have been dated accurately by the best current techniques (e.g.
Milner et al., 1995;
Renne
et al., 1996;
Pearson et al., 1998). It is therefore appropriate to assess the
Sonoki & Garda, (1988) compilation rigorously.
Our first step has been to reject all the bulk-rock K/Ar dates in the compilation. Obviously this will result in the loss of some good data but experience of the bulk-rock K/Ar dating of very similar rock-types in the Alto Paranaíba province
(Gibson et al., 1994,
, 1995, and unpublished data, 1996) has convinced us that this technique is not reliable for SE Brazilian Cretaceous-Tertiary alkaline igneous rocks. The widespread leucite in the mafic, strongly potassic types is mostly variably altered to analcite and other phases (e.g.
Cima, 1995). Our study of the Serra do Bueno dyke in the Alto Paranaíba province (
Fig. 1), which contains altered groundmass leucite, showed graphically the effect of this alteration on K/Ar dating
(Gibson et al., 1994). Our own and published bulk-rock dates
(Woolley, 1987) are similar (37 and 45 Ma, respectively) but far less than the laser 40Ar/39Ar age of 90 Ma; the latter is consistent with other high-reliability dates throughout this province
(Gibson et al., 1995). Altered leucite is common in the syenites of the Serra do Mar central complexes (see above) and petrographic evidence abounds for their pervasive post-crystallization hydrothermal alteration. This process continues to the present day; the Araxá (Alto Paranaíba) and Poços de Caldas complexes are both sites of hot-spring-fed spa resorts.
Shea, (1992) specified that hydrothermal alteration at Poços de Caldas has reset radiometric ages within the complex. Therefore we have also rejected all feldspar K/Ar dates from central-complex plutons within the
Sonoki & Garda, (1988) compilation. Again, this procedure will exclude some reliable dates but it is very difficult to see how else to screen the published ages.
The remaining Sonoki & Garda dates were obtained using separated biotite, amphibole and pyroxene. The few published amphibole K/Ar dates seem to be consistent with the most reliable biotite ages (see below) and the results by other techniques, such as 40Ar/39Ar. This is not so for the pyroxene K/Ar dates, which give variable (mostly high) ages for complexes such as Jacupiranga, São Paulo State, where 40Ar/39Ar ages are available
(Renne et al., 1993). Many biotite dates are listed in the
Sonoki & Garda, (1988) compilation and it is also well established that this mineral is comparatively resistant to low-temperature hydrothermal alteration. We have therefore screened the Sonoki & Garda compilation, by comparing the potassium contents of the dated micas with the known K2O ranges of fresh syenite biotites (8-9 wt %;
Deer et al., 1965) and the phlogopites-biotites of mafic potassic rock-types (9-11 wt %;
Mitchell & Bergman, 1991), using the compositions of `separable' crystals rather than groundmass or rim analyses.
Sonoki & Garda, (1988) gave biotite potassium contents in weight percent K and
Fig. 5 is a histogram of these values; 8% K2O equates with 6·6% K in
Fig. 5 and it is therefore clear that many of the published phlogopite-biotite K/Ar dates listed by Sonoki & Garda used material that was either severely altered to chlorite or a poor mineral separate, or both. Consequently, we have recompiled these dates, after eliminating all `micas' that fail a crude test of realistic K content. In an effort to try to avoid `borderline', less-than-ideal samples, we have rejected all Sonoki & Garda phlogopites-biotites with <7% K (8·4% K2O). This is an extremely severe criterion for locating only fresh phlogopites-biotites and it has undoubtedly resulted in our rejection of several perfectly good published dates. The purpose of our ultra-severe screen is to convince readers that the remaining published K/Ar dates which we use are reliable.
Figure 6 shows the remaining (accepted) phlogopite-biotite dates plotted along the same traverse as
Fig. 4, together with results from Rb/Sr isochrons that used carefully chosen bulk-rock samples (e.g.
Brotzu et al., 1989,
, 1992;
Shea, 1992), 40Ar/39Ar dates
(Shea, 1992;
Regelous, 1993) and new K/Ar results of our own. These will be described before returning to
Fig. 6.
In our studies of Brazil-Paraguay alkaline igneous rocks (e.g.
Gibson et al., 1995) we have discovered one method to obtain some excellent K/Ar dates in provinces, such as Serra do Mar, where altered leucite is widespread and the magmatism is dominated by intrusive complexes that are affected by powerful long-lived hydrothermal activity. Occasional lavas and hypabyssal plutons contain fresh phlogopite-biotite phenocrysts which are extracted in the field. Individually checked phenocryst fragments are then crushed for K/Ar analysis.
Regelous, (1993) used this approach in the Serra do Mar province, to obtain a date of 80·1 Ma by 40Ar/39Ar for the phenocryst biotite in a Campos do Jordão dyke (
Fig. 2). We have located two felsic dykes (Mendanha and Cabo Frio) and one minette (Cabo Frio) that yielded suitable micas for K/Ar dating of field-separated phenocryst fragments, plus two minettes (Soarinho) with abundant, easily separated biotite. The results are listed in
Table 1 and plotted in
Fig. 6. One of the Soarinho biotite separates in
Table 1 (94SOB131) actually fails the K2O criterion used for filtering published dates in the previous section, but is included because it gives a date concordant with that of another better sample (94SOB128) from the same locality. When combined with the filtered previous dates, the new results show a clear pattern of migrating age of magmatism, ESE across the Serra do Mar province (
Fig. 6). Although the `offshore islands' (Trigo, São Sebastião-Ilhabela, Vitória and Búzios) fit this trend reasonably well, it should be remembered that they lie far to the south of other complexes at the same longitude (
Fig. 2). The relatively long timespans of the magmatism at Poços de Caldas (~8 Ma) and Itatiaia (~10 Ma) may be related to the much larger sizes of these two complexes than the others.
Gallagher et al., (1994) have reported apatite fission-track dates from metamorphic basement samples adjacent to Poços de Caldas that are consistent with the view that magmatism began at ~80 Ma and had cooled to <60°C after a few million years. Where high-quality data are available, large former volcanic complexes show magmatic activity age spans of no more than ~5-6 my. Examples are the Cretaceous Damaraland complexes, Namibia
(Milner et al., 1995;
Renne
et al., 1996), and the Palaeocene Skye centre, Scotland
(Pearson et al., 1998). Nevertheless, the initiation of alkaline magmatism across the Serra do Mar province seems to lie on a well-defined trend. Table 1. Potassium-argon ages of phlogopites.
The samples were analysed by a combination of X-ray fluorescence (XRF), instrumental neutron activation analysis (INAA) and inductively coupled plasma mass spectrometry (ICP-MS). Full details of our techniques used for XRF, INAA, and Sr- and Nd-isotopic ratios, were published by
Gibson et al., (1995). Details of our ICP-MS techniques are given in Appendix B. Despite intense sub-tropical weathering in parts of the region, it is apparent that many of the analysed samples (
Table 2) have loss-on-ignition (LOI) values of <5 wt %. The main exceptions are: (1) mica-rich rocks; (2) agglomerate blocks from Poços de Caldas, with leucite altered to analcite; (3) the thin dykes at Campos do Jordão. Amphibole also contributes to LOI in many samples. Obviously the geochemistry of the high-LOI, mica-poor samples needs to be treated with caution but otherwise these rocks are remarkably fresh in thin section, apart from altered olivine. Table 2. Representative whole-rock analyses.
Table 2. continued.
Figure 7 is a plot of Na2O + K2O vs SiO2 (TAS). Most of the samples in
Table 2 classify as basanites
(Le Maitre, 1989), consistent with their mineralogy. The petrographic identification of some of the Campos do Jordão dykes as melanephelinites is also borne out by their TAS compositions. Analyses of Trindade lavas in
Fig. 7
(Halliday et al., 1992; J. C. Greenwood, unpublished data, 1997) fall in the same TAS range as the Serra do Mar data. All the Poços de Caldas lavas contain leucite variably altered to analcite and this may have affected their compositions. Serra do Mar samples that contain phlogopite-biotite phenocrysts are noted in
Fig. 7 and show high total alkali contents, relative to other samples with similar SiO2.
K2O/(Na2O + K
2O) ratios of most of the Serra do Mar samples (
Fig. 8) fall in the same range as the available Trindade data but, at a given SiO2 value, a small number of analyses are notably more potassic. All are mica-phyric except a Poços de Caldas leucitite (90SB79), the Morro de São João leucite-bearing melasyenite (94SOB142), and a São Sebastião picrite dyke (94SOB13). This picrite shows no petrographic evidence of its potassic composition and we shall discuss it further below, in the light of isotopic data. Turning to the relatively sodic compositions, the Mendanha and Neogene basalts-basanites with ~45% SiO2 need no comment but the `sodic' Poços de Caldas `leucitites' show vividly the effects of analcitization on K/Na ratios.
Other aspects of the major-element compositions of the Serra do Mar samples are illustrated by
Fig. 9, a plot of TiO2 and CaO vs MgO. Although there are no signs of a single, overall Serra do Mar geochemical trend, most of these data behave in a coherent way, when the sub-sets from individual localities are considered. It should be noted that the apparent preponderance of relatively MgO-rich samples is an artefact of searching carefully for them in a syenite-dominated province. TiO2 vs MgO in the Trindade and Campos do Jordão data sets (
Fig. 9a) show similar inverted-V trends; the inflections at ~7% MgO correspond to the appearance of titanomagnetite microphenocrysts. Other Serra do Mar suites (
Fig. 9b) form a scatter at MgO <9% and a single linear array at higher MgO abundances. The latter is a plausible olivine control line. CaO vs MgO (
Fig. 9c) shows all data fitting a single scattered inverted-V trend, with its inflection at ~9% MgO. We interpret
Fig. 9 as evidence that the compositions of the Serra do Mar mafic magmas are controlled by fractional crystallization: olivine (plus minor chromite) at >9% MgO; joined by Ca-rich clinopyroxene at <9% MgO and titanomagnetite at <7% MgO. The difference in TiO2 abundance, at a given MgO value, between Trindade, Itatiaia, Poços de Caldas and Campos do Jordão samples, on the one hand, and São Sebastião-Trigo, Mendanha, Cabo Frio and Neogene samples, on the other, could be caused by either different mantle sources or different conditions during melting of a uniform source (see below).
Samples with MgO >6% are plotted in
Fig. 10, La/Ba vs La/Nb, together with the field that encloses a large set of ocean-island basalts (OIB;
Fitton et al., 1991). Fitton et al. showed that this plot discriminated well amongst western USA Cenozoic basic volcanic rocks; those that were OIB-like, in terms of other elemental and isotopic criteria, fell within the OIB field whereas those that fell outside had other geochemical features suggesting a lithospheric contribution to their compositions. Clearly this plot cannot discriminate Cretaceous mafic magma sources so successfully in SE Brazil because the Alto Paranaíba kamafugites, which are inferred to be predominantly or entirely from lithospheric mantle sources
(Gibson et al., 1995;
Carlson
et al., 1996), overlap the OIB field. Only one potassic Serra do Mar sample, Soarinho minette 94SOB131, plots far outside the OIB field.
Normalized incompatible-element plots (
Fig. 11) allow comparison of a wide range of elemental abundances and ratios in mafic Serra do Mar samples and representative OIBs. The three mantle xenolith-bearing Neogene samples are clearly extremely similar geochemically and also OIB-like, resembling a basanite from Tenerife
(Thompson et al., 1984) and olivine melanephelinite from Trindade (J. C. Greenwood, unpublished data, 1997) in all plotted elements except K and Rb (sensitive to both alteration and the presence of residual phlogopite in the mantle source) and P (variable from suite to suite in OIBs). Likewise, the patterns of all the more sodic samples (
Fig. 8) plotted in
Fig. 11 are fundamentally very similar and OIB-like. The relatively low K in sample 94SOB50, from São José do Itaboraí, may be due to the somewhat altered state of the sample, with chloritized biotite. Most mafic igneous rocks from sub-continental lithospheric mantle sources show `spiky' patterns in normalized diagrams, such as
Fig. 11. The Soarinho minette 94SOB131 has such a pattern but those of the most potassic samples from Itatiaia and Cabo Frio are smooth and OIB-like, resembling the Tristan da Cunha basalt plotted for comparison. To summarize the major- and trace-element compositions of the Serra do Mar mafic rocks: (1) they are mostly basanites to melanephelinites, with occasional more potassic variants; (2) fractional crystallization controlled magmatic evolution within each local suite; (3) the more magnesian samples are all, except one Soarinho minette, closely similar in composition to OIB.
A plot of Sr- and Nd-isotopic ratios (
Fig. 12) shows that, apart from two obvious exceptions with very high 87Sr/86Sr, the Serra do Mar data set forms a scattered trend, between São Sebastião-Trigo samples with [epsilon]Nd ~ +2·5 and 87Sr/86Sri <0·7040, and samples from Itatiaia-Passa Quatro, Soarinho and Cabo Frio, with [epsilon]Nd < -2·9 and 87Sr/86Sri >0·7046. Three of the five samples with lowest [epsilon]
Nd values contain phlogopite phenocrysts. In
Fig. 12 these lie adjacent to the field of the ~85 Ma Alto Paranaíba mafic ultrapotassic rocks
(Gibson et al., 1995) that crop out about 500 km to the northwest (
Fig. 1). Two entirely different models could potentially explain the Serra do Mar Sr-Nd isotopic array:
As other studies at this latitude on both sides of the South Atlantic have found, it is extremely difficult to differentiate unequivocally between these two contrasting geochemical models (e.g.
Milner & le Roex, 1996;
Gibson
et al., 1997b,
1997c). We consider two clues that model 1 may be correct:
We therefore conclude that the available data favour the view that all but the two high-87Sr/86Sri Serra do Mar samples are mixtures of melts from both convecting (predominantly) and lithospheric mantle sources, with the latter varying locally but overall resembling the sources of the Alto Paranaíba mafic ultrapotassic magmas. We appreciate that our data do not yet exclude a variable convecting mantle source hypothesis and we aim to test these alternatives by Os-isotope studies (J. C. Greenwood, work in progress).
Crustal (ATA) contamination of two picrites
The two samples with very high 87Sr/86Sri-picrites from São Sebastião (94SOB13) and Sentísssimo, Mendanha, (94SOB64)-have the two highest MgO contents (
Table 2). Despite high 87Sr/86Sri, their [epsilon]Nd values are within the range of other samples from the same locality. This is particularly clear for the Mendanha set (
Table 2;
Figs 12 and
13). An input from selectively dissolved crustal K-feldspar would explain the isotope systematics. Is this realistic for dykes? Sample 94SOB64 is ~30 cm thick and 94SOB13 is only 10 cm thick at its maximum. Both are emplaced into granitic gneisses.
There is theoretical, observational and geochemical evidence that MgO-rich liquids with low viscosities can flow turbulently within the crust and hence become very rapidly contaminated by fusible components of their wallrocks
(Moorbath & Thompson, 1980;
Huppert & Sparks, 1985;
Kille et al., 1986).
Kerr
et al., (1995) proposed that this process should be called Assimilation during Turbulent magma Ascent (ATA), to contrast it with the well-known AFC (assimilation-fractional crystallization) process. The transition between turbulent and laminar flow in a liquid is abrupt (e.g.
Huppert & Sparks, 1985) and therefore it is possible that the two most MgO-rich dykes could become contaminated, whereas all the others escaped unaffected. But even less MgO-rich, more viscous basic liquids will sometimes locally flow turbulently in dykes, as they pass obstructions
(Kille et al., 1986), and therefore they too can become affected by ATA contamination. This may be the reason why the high-[epsilon]Nd São Sebastão-Trigo samples in
Fig. 12 show a spread of 87Sr/86Sri, even when 94SOB13 is excluded, and all plot at higher 87Sr/86Sri values than the Trindade field. We have emphasized this evidence for crustal contamination because most investigators of small-volume mafic magmatism tend to dismiss it (e.g.
Comin-Chiaramonti et al., 1997). If some or most of these scattered dykes and small lavas have dissolved crust, the detailed geochemical modelling of their genesis and subsequent evolution becomes very difficult. Such modelling requires isotopic analyses of the more evolved rock-types, as is being undertaken by others (e.g.
Valente et al., 1995).
It is implicit in the foregoing discussion that the [epsilon]Nd values of the Serra do Mar mafic rocks seem to be essentially unaffected by any postulated crustal contamination.
Figure 13 shows that the [epsilon]Nd ranges of most Serra do Mar suites are fairly large. Despite the point we made above about the offshore island of Trigo producing our highest-[epsilon]Nd sample, it is clear from
Fig. 13 that there is no sign of a progressive reduction in [epsilon]Nd away from the continental margin; the entire [epsilon]Nd range occurs in complexes along or adjacent to the present coastline. This reinforces the view that a significant factor in allowing high-[epsilon]Nd magmas to reach the upper crust at Trigo and São Sebastião was their late genesis in the local magmatic sequences. In the previous section we have argued that the geochemistry of the Serra do Mar mafic magmas is best explained by a model in which their ultimate source was predominantly within the convecting mantle but that, during their uprise, individual magma batches:
The first two of these processes can substantially affect both the abundances and ratios of rare earth elements (REE) and other incompatible elements in mafic magmas. For this reason it is inappropriate, in this case, to use REE inversion calculations
(McKenzie & O'Nions, 1991,
, 1995) to define further the convecting-mantle sources and melting conditions of the magmas. Nevertheless, the high-[epsilon]Nd São Sebastão-Trigo basanites and nephelinites (
Fig. 7) should yield defensible REE-inversion results.
To model the genesis of the Serra do Mar initial magmas in a simple way, without over-interpreting the data, we have used the approach of
Kostopoulos & James, (1992), derived from the non-modal batch melting equations of
Shaw, (1979).
Figure 14 shows arrays of calculated small-degree melt compositions for anhydrous fusion of lherzolite mantle at depths where spinel or garnet, or a 50:50 mixture of the two, is the fourth phase. The arrays are calculated for two contrasting hypothetical variants of lherzolite: DMM (Depleted MORB-source Mantle) and BSE (Bulk Silicate Earth), as defined by
Kostopoulos & James, (1992). These lherzolite variants are very similar to the hypothetical `MORB-source' and `primitive' mantles used by
McKenzie & O'Nions, (1991,
, 1995). It is clear from
Fig. 12 that, taking the Serra do Mar data set as a whole, its time-integrated values of Rb/Sr and Sm/Nd are close to those defined for BSE melts, and that the BSE array in
Fig. 14 is therefore appropriate. Of course, the situation is different if one accepts the arguments that we made in the last section, that only the Serra do Mar melts with the highest [epsilon]Nd values should be modelled as simple one-stage melts of convecting mantle, free from low-[epsilon]Nd post-genesis inputs. In the latter case, the logical array to calculate and plot in
Fig. 14 is the set of melts from an ~50:50 mixture of DMM and BSE because this would generate time-integrated values of Rb/Sr and Sm/Nd similar to those of the Trindade lavas. This is close to the mantle source calculated by
Gibson et al., (1997c), using inversion modelling, for the OIB-like alkali basalts emplaced at ~84 Ma above the Trindade mantle plume head at Poxoréu, Mato Grosso, in central Brazil.
Only REE determined by ICP-MS are plotted in
Fig. 14. As a group, the Serra do Mar mafic magmas may be interpreted in terms of ~0·1-1·0% melts of a BSE lherzolite mantle, with a garnet:spinel ratio around 70:30. It is clear from the diagram that, if the chosen mantle source was instead a 50:50 mixture of BSE and DMM, the degrees of melting indicated would be smaller but the estimated garnet:spinel ratio of the source would not change greatly. Conversely, there is ample petrographic evidence (see above) that these magmas were relatively volatile rich; a factor that would increase the estimated degrees of melting
(Hirose & Kawamoto, 1995).
McKenzie, (1989) has emphasized the ease with which very small fractions of low-viscosity, picritic, volatile-rich, mafic alkaline melts can escape upwards from a convecting mantle source. Two points lead towards the hypothesis that the source of the Serra do Mar province mafic magmas was a mantle plume:
Using Hawaii as a tectonomagmatic reference point, it is apparent that the Serra do Mar region did not pass directly over a `full power' mantle plume (thermally speaking), with Tp ~1550°C
(Watson & McKenzie, 1991); the igneous centres of
Fig. 2 are clearly not the wrecks of giant tholeiite-dominated shield volcanoes and the area lacks a Late Cretaceous flood-basalt province. Instead, the migrating magmatism is relatively small in volume and its mafic products are mostly basanites with compositions that strongly resemble those of the small-volume, post-erosion mafic alkalic volcanism of the Hawaiian islands (
Fig. 11).
Wyllie, (1988),
Watson & McKenzie, (1991) and others have proposed that the Hawaiian late-stage alkalic magmas originated-at least in part-from the cooler periphery, rather than the hotter core, of the underlying mantle plume. At the times that they were erupted, the centre of the Hawaiian plume head was fuelling younger tholeiitic shield volcanoes, 200-300 km to the southeast.
The many numerical models of steady-state mantle plumes (`tails'; not their initial starting-heads) published during the last decade or so (e.g.
Courtney & White, 1986;
Watson & McKenzie, 1991;
Ceuleneer et al., 1993) all calculate sub-lithospheric plume heads with radii in the general range of ~600 km. Heat loss from a mantle plume can occur in three ways: advection via escaping melt; conduction to both the overlying lithosphere and underlying ambient-temperature convecting mantle; marginal mixing between the latter and the plume-head mantle (e.g.
Sleep, 1996). Numerical models mostly agree that advective heat loss by decompression melting is largely confined to the vicinity of the core of the plume head, if a constant-thickness lithosphere is assumed (e.g.
Watson & McKenzie, 1991;
Ceuleneer et al., 1993;
Farnetani & Richards, 1995). Towards the plume-head periphery, the decrease in Tp is usually modelled as gradual and approximately linear (presumably an asymptotic curve). Viewed from this theoretical background, the convecting-mantle source of the bulk of the Serra do Mar mafic magmatism could have been: (1) the periphery of the head of a major mantle plume which lay on a track that passed several hundred kilometres away during the ~80-55 Ma interval; (2) a comparatively minor local upper-mantle thermal disturbance.
The initiation of Serra do Mar magmatism migrated ESE for 500 km across the province, from Poços de Caldas to Cabo Frio, between 80 Ma (or slightly earlier) and ~55 Ma. If this represents magmatism from a source fixed within the global hotspot framework, penetrating a drifting plate, the apparent plate migration velocity was 2·0 cm/yr, a figure little more than half that of the South American plate
(O'Connor & Duncan, 1990;
Müller et al., 1993). Between our first published reports of these results (
Thompson et al., 1997a,
1997b) and the completion of this paper, R. W. Carlson (personal communication, 1997) has informed us that he has also detected this migrating magmatism, based on Rb/Sr isochron studies of the syenite complexes. Can magmatism migrating at far less than the drift rate of the South Atlantic plate be plausibly related to a mantle plume?
Figure 15 shows how the Trindade mantle plume might fit this tectonomagmatic scenario. Nevertheless, although it has appeared in most discussions of the topic on a global scale, and also in many others concentrating on South Atlantic tectonomagmatic relationships (e.g.
Crough et al., 1980;
O'Connor & Duncan, 1990), we must first summarize the evidence that a mantle plume currently centred approximately beneath Trindade and Martin Vaz actually exists.
VanDecar et al., (1995) have suggested that there is no such mantle plume and that the magmatism forming Trindade, Martin Vaz and the Vitória seamount chain, between Trindade and the Brazilian mainland (
Fig. 15), is only the result of a `leaky transform fault'. This suggestion may be evaluated by considering the end-member case of leaky transform faults, which is a spreading centre. Over convecting mantle with the global ambient potential temperature of ~1300°C, a spreading centre generates oceanic crust of normal (~7 km) thickness, not a seamount chain. Trindade and Martin Vaz are built on oceanic crust >70 Ma in age; oceanic lithosphere of this age is ~120 km thick, with an ~90 km thick mechanical boundary layer
(Parsons & McKenzie, 1978). Trindade and Martin Vaz rise from a broad topographic swell beneath water 5 km deep. Trindade reaches ~600 m above sea level, and its total height is thus ~5·5 km. Trindade is therefore a volcano that is similar in height to those of the Hawaiian islands, although steeper sided and therefore less voluminous, and it is built on lithosphere which is probably somewhat thicker than and drifting at a similar velocity to that beneath Hawaii
(Müller et al., 1993). We can see no realistic alternative to the view that a mantle plume currently underlies Trindade-Martin Vaz. Likewise, the major Vitória chain seamounts all rise up to 4·5 km above seafloor. The short-wavelength positive geoid anomalies associated with Trindade and the Vitória seamounts have been documented by
Fleitout et al., (1989) and
Ussami & Molina, (1997).
Figure 15 shows the calculated track of a mantle plume, currently beneath Trindade (most recent magmatism at ~20 ka;
Cordani, 1970), back to 80 Ma. The input was a global set of parameters for absolute plate motions, back to chron 34 (84 Ma), derived by
Müller et al., (1993) from a best fit to `radiometrically-dated hotspot tracks on the Australian, Indian, African, North and South American plates, relative to present-day hotspots assumed fixed in the mantle'. The calculated track bisects the Abrolhos Platform-a constructional magmatic feature resembling a small flood basalt province
(Fodor et al., 1989)-and is predicted to have passed beneath its site at the time that the magmatism occurred (42-52 Ma;
Cordani, 1970). We do not know whether there is any significance in the small divergence between the post-40 Ma calculated plume track and the trend of the Vitória seamount chain (
Fig. 15). If the divergence is real, a possible explanation is that the plume-head magmatism escaped upwards via the line of weakness presented by the Martin Vaz fracture zone, rather than penetrating the lithosphere directly above. Nevertheless, there is continuing controversy as to whether or not plumes drift, relative to each other, and hence whether plume tracks can indeed be reconstructed in the manner of
Fig. 15
(Harada, 1997;
Raymond et al., 1997). The large circle marked in
Fig. 15 shows the postulated impact site of the Trindade starting-mantle-plume head at ~85 Ma, as deduced by
Gibson et al., (1995,
1997c). In
Fig. 15 we have specified three stages in the post-impact sub-continental migration of the Trindade plume and associated magmatism:
(1) Expansion of the starting-plume head (85-80 Ma). The conventional circular starting-plume impact site drawn in
Fig. 15 gives no obvious explanation for magmatism beginning at Poços de Caldas at ~84 Ma
(Shea, 1992) and at the offshore island group, São Sebastião-Ilhabela, Trigo, Vitória and Buzios, at ~80 Ma (
Figs 2 and
6). A possible explanation might be that the starting-plume head expanded rapidly until it decompressed, began to melt and `leaked' magma beneath the thinner lithosphere of the stretched continental passive margin
(Hill, 1991;
Thompson & Gibson, 1991;
Sleep, 1996,
, 1997). Recent studies of fission-track data support this concept.
Hegarty et al., (1996) used apatite fission-track analyses of samples from within and around the southern part of the Paraná basin (
Fig. 1) to detect an episode of substantial uplift and subsequent kilometre-scale erosion (causing crustal cooling) between 90 and 80 Ma.
Harman et al., (1997) have reported comparable data from sites on the São Francisco craton, and
Gallagher & Brown, (1997) have shown how the coastal margin of SE Brazil has undergone substantial uplift and erosion in the 90-60 Ma interval, long after the initiation of the South Atlantic at ~130 Ma.
Hegarty et al., (1996) noted that their data could be explained well by the impact of a late Cretaceous starting-mantle plume beneath the region.
A difficulty with such an interpretation is the large distance (~1200 km) between the parts of southern Brazil and eastern Paraguay that show evidence for this late Cretaceous uplift and the centre of the starting-plume impact site marked in
Fig. 15.
Figure 1 shows a possible explanation. The northern side of the proposed Trindade starting-plume impact site is largely ringed by the Amazonas and São Francisco cratons. Perhaps the deep lithospheric keels of these cratons impeded plume-head expansion northwards, thus accentuating it southwards. The Trindade starting-plume head may also have been unusually hot and buoyant, during its post-impact expansion, because at first it underlay lithosphere too thick to permit significant decompression melting
(Gibson et al., 1995,
1997c) and consequential rapid escape of heat
(Sleep, 1996,
, 1997). We note that the plume-head expansion model, proposed here, also gives a plausible explanation for the late Cretaceous igneous centres within the Ponta Grossa igneous province (
Fig. 1) and the local rifting-related alkaline magmatism around Lages at 80-75 Ma
(Sonoki & Garda, 1988; Gibson et al., in preparation).
(2) Leakage of hot plume-tail mantle to the south, as the plume passed beneath a craton (80-55 Ma). Magmatism is spatially associated with the calculated Trindade plume track, both before 80 Ma and after 52 Ma (
Fig. 15). Between these times the calculated track lay beneath the São Francisco craton, whereas contemporaneous alkaline magmatism migrated ESE across the Serra do Mar region (
Figs 2 and
6), ~500 km to the south. We propose that this magmatism was fed by a southward flow of hot plume-tail mantle from beneath the São Francisco craton. The migrating row of complexes is oblique to basement structural trends (
Fig. 2). We suggest that, in general, the magmas were generated (and erupted) at the first points south of the craton where the slowly cooling flow of plume-tail mantle arrived beneath lithosphere, within the rapidly thinning zone adjacent to the continental margin, where significant decompression melting could occur
(Thompson & Gibson, 1991;
Sleep, 1996,
, 1997). This leakage process broke the direct connection between the fixed-site mantle plume and the magmas eventually produced by its decompression melting. As a result, the Serra do Mar magmatism migrated ESE at ~2·0 cm/yr (
Fig. 6), about half the rate of 3·6 cm/yr calculated for the theoretical plume track (
Fig. 15). This implies that great care should be taken before deducing that a belt of migrating magmatism on a continent necessarily measures its drift rate.
(3) `Jump' to NE in the magmatism, as the plume tail reached the continental margin (55-52 Ma). The Serra do Mar province migrating magmatism ceased abruptly in the Cabo Frio area at 55 Ma (
Figs 2 and
6); there are no appropriate seamounts offshore along the continuation of this trend. Instead, the Abrolhos Platform large-scale magmatism began at essentially the same time, ~400 km to the northeast. We suggest that, as the Trindade plume tail reached the thinner lithosphere of the continental margin, to the east of the São Francisco craton, it was able to decompress sufficiently to initiate substantial melting and thus to cut the southward flow of hot mantle towards the Serra do Mar igneous province. It should be noted that our discussion of how the hot mantle of the Trindade plume may have behaved beneath the lithosphere of SE Brazil between 85 and 50 Ma is written in similar terms to a description of water drainage (hot mantle) on an irregular surface (base of lithosphere), except that the system is upside down. In this approach we follow
Sleep, (1996,
, 1997), who has quantified some aspects of such a tectonomagmatic model.
This work was funded by NERC (UK) Research Grant GR3/8084, NSERC (Canada), CNPq (Brazil) and the British Council, together with additional financial and logistical input from the Universities of Brasília, Cambridge, Durham and Newcastle upon Tyne. We thank Yvonne Brown, Ron Hardy, Chris Ottley, Graham Pearson and Rob Ridley for their technical assistance with elemental and isotopic analyses, and Roger Searle for calculating the theoretical Trindade plume track. J. M. V. Coutinho, J. G. Valença, and H. H. G. L. and M. N. C. Ulbrich were generous with their time, advice and practical assistance, in connection with our fieldwork. Countless landowners and quarry managers allowed us access to outcrops; the manager of the Sentíssimo quarry earned our heartfelt thanks for whisking us to safety when the gunfire came too close for comfort. Vicky Hards made a crucial contribution by joining a depleted fieldwork team at zero notice, allowing the sampling of the offshore islands. Comments by Mike Coffin, Tony Ewart, Vicky Hards, Teresa Junqueira-Brod, Graham Pearson and Marge Wilson substantially improved the text.
INTRODUCTION
THE SERRA DO MAR IGNEOUS PROVINCE
Nomenclature
Geological setting
Late Cretaceous igneous complexes
RADIOMETRIC AGES OF SERRA DO MAR PROVINCE MAGMATISM
Published dates
New phlogopite-biotite phenocryst K/Ar dates
GEOCHEMISTRY OF THE MAFIC ROCKS
Elements
Sr-Nd isotopes
DISCUSSION
Serra do Mar province mafic magma sources
Trindade mantle plume track
SUMMARY
ACKNOWLEDGEMENTS
REFERENCES
APPENDIX A: SAMPLE LOCALITIES
APPENDIX B: ICP-MS TECHNIQUES