Wandering continents of the Indian Ocean

That the shapes of many continental outlines appear to fit together like a jig-saw puzzle was known as early as the 16th century by explorers like Magellan. Alfred Wegener was one of the first to suggest that the now-separated pieces may once have been joined together in a unified land mass (e.g. Wegener, 1912, 1915), and he coined the phrase “continental drift”, an idea that was bitterly attacked and dismissed for decades.

Alexander Du Toit was one of Wegener’s most enthusiastic supporters, and he developed and strengthened the evidence for continental drift in his 1937 book, Our Wandering Continents. He used an analogy of a torn-up page of newspaper, the pieces of which could be fitted together based on their shape, but a better reconstruction would result by matching the lines of printed text. Connecting the letters and words, he said, would be analogous to correlating geological features on continental entities now separated by vast oceans. Using his knowledge of South African geology, stratigraphy, petrology, palaeobotany, hydrogeology, and geomorphology, in comparison with what he learned by visiting South America, Du Toit established compelling correlations that represented a successful test of the continental drift hypothesis. This is his legacy. He was the first to realize how the southern continents were once joined to form the supercontinent of Gondwana.

On the last page of his 1937 book, Du Toit advised the geoscience community to evaluate these ideas further, by engaging in what he called “comparative geology”, in other words, “the study of continental fragments”. In this paper, I choose to honour Du Toit’s pioneering contributions by focusing on exactly this, the comparative study of continental fragments in the Indian Ocean (Figure 1), synthesizing the results of my collaborative work on this subject, carried out since 1991 with a large number of colleagues and students, starting in Madagascar.

My interest in Madagascar stemmed from my awareness of aerial photographs and Landsat images showing a spectacularly exposed ductile shear zone in a remote region in the southwestern part of the island (Figure 2a). Here, two massif-type anorthosite bodies (Boulanger, 1959) appear to have been pulled apart by >40 km, and I decided that a visit to this site was essential, given my broad interests in anorthositic and related rocks. During a holiday and reconnaissance trip in December, 1991, Sue Webb and I established contacts with Prof. Roger Rambeloson (University of Antananarivo), who was eager for collaboration, and educated us about the intricacies of field work in Madagascar. I managed to spark the interest of my friend Maarten de Wit, and after a generous donation of a 4-wheel drive vehicle from Richard Viljoen (then of Gold Fields), which we shipped to Madagascar, Roger, Maarten and I successfully arrived at the Ankafotia anorthosite body in July, 1992, after a grueling journey. The results of our studies on the anorthosites and on the ductile shear zone that hosts them resulted in a number of M.Sc. theses and research papers (Morel, 1994; Randrianasolo, 1995; Ashwal et al., 1998; de Wit et al., 2001).

Maarten and I learned that the quality of geologic mapping in Madagascar is superb, having been carried out in the 1940s and 1950s by an army of geologists, under the leadership of Geological Survey Director Henri Besairie. These excellent maps were later compiled as the 1:1 000 000 scale geological map of Madagascar (Besairie, 1964) (Figure 2b). However, we felt that what was needed was an enhanced appreciation of this work with respect to modern tectonic and geodynamic perspectives. As an example, it was not then recognized that the large terranes designated as “schistes verts, quartzites et magnetites” (elongate belts in north and west-central Madagascar, coloured in green in Figure 2b) are Archaean greenstone belts (e.g. Rambeloson, 1997). Another serious deficiency was the nearly complete lack of reliable age determinations for this substantial (nearly 600 000 km²) entity of understudied continental crust, and in 1994 I solicited the collaboration of Bob Tucker to acquire precise U-Pb zircon ages by the TIMS method. He enthusiastically and energetically dated rocks throughout the island, and as of 2014, he and colleagues had acquired over 565 precise zircon and baddeleyite U-Pb ages (Tucker et al., 2014).

In an attempt to inspire international interest in Malagasy geology, in 1997 Tucker and I, with help from several colleagues, post-docs and students, organized a conference and field excursion in Madagascar, sponsored by UNESCO-IUGS-IGCP, and which was attended by 42 participants from 19 countries (Ashwal, 1997; Cox and Ashwal, 1997). Many new and fruitful research projects were inspired from this meeting, and some are active still. In 2003, a major five year project was initiated, generously funded by the World Bank, aimed at enhancing economic growth in Madagascar through the sustainable development of its mineral resources. This effort was carried out by large teams of geoscientists from a consortium of French, British, German and American governmental and industrial organizations, and involved extensive field-based geological re-mapping complemented by digital remote sensing geophysical and other data, as well as sampling of rocks and soils for new geochemical and geochronological analyses (Moine et al., 2014 and papers therein). Among the many outputs is a new geological map of Madagascar at 1 000 000 scale (Roig et al., 2012). The wealth of new observations and measurements acquired over the past few decades now place our understanding of Malagasy geology on par with that of other continental entities. Interpretative reviews of the geological, tectonic and geodynamic evolution of the world’s fourth largest island can be found in de Wit (2003) and Tucker et al. (2014). A great deal of research is ongoing.

For the purposes of this paper, I have chosen to focus on only one of the numerous types of research projects I have been involved with in Madagascar since 1991. In the next section, the major findings of this work are discussed, and are placed in the context of du Toit’s “comparative geology”. This is followed by a series of selected research themes, all of which relate to the “wandering continents” of the Indian Ocean, discussed in the framework of the plate tectonic paradigm that Alex du Toit helped to establish.

Many of the widespread granitic and migmatitic rocks in Madagascar had long been presumed to be broadly of “Pan-African” age (~400 to 900 Ma), but in the 1990s, new U-Pb zircon geochronology began to reveal the presence in west-central Madagascar of abundant, late Neoproterozoic granitic and gabbroic plutons, with ages between 719 and 797 Ma (Figure 3) (Handke et al., 1999; Handke, 2001; Thomas et al., 2009). Many of the intrusive bodies are elongate in shape, with aspect ratios >10:1, and consist of composite, nested plutons with complex intermingling of gabbroic and granitic components. The gabbroic rocks are alkaline, with Na₂O + K₂O up to 8 wt.%, and contain primary intercumulus hornblende and biotite, and the granitoids range from quartz monzonites to granites to syenogranites (Figure 3, lower right panel). Significant Cu mineralization is present in both the mafic and silicic components. These features were interpreted as having formed in a subduction-related, continental magmatic arc setting (Handke et al., 1999).

Support for this suggestion comes from geochemical studies of ~790 Ma ultramafic to gabbroic intrusives in the Andriamena region of north-central Madagascar (Figure 4a) (Bybee, 2010; Bybee et al., 2010). The lithologies include gabbro, dunite, harzburgite, pyroxenite, websterite, and PGE-rich chromitites. In one of the gabbro bodies, a basaltic chilled margin has trace element abundances that match the signatures of modern-day continental arc volcanics (Figure 4c). More interestingly, PGE concentrations of the ultramafic rocks all show prominent negative Ru anomalies (Figure 4b), which are a characteristic feature of similar rocks in so-called Alaskan-Uralian complexes that are widely interpreted as feeder pipes or magmatic roots of volcanoes in continental arcs (e.g. Murray, 1972).

Coeval volcanic equivalents of these intrusive rocks, dated at 719 to 751 Ma (Tucker et al., 1999b; Thomas et al., 2009), occur in the Bemarivo Belt of northern Madagascar, and include basaltic to rhyolitic lavas, tuffs, and an assortment of other pyroclastic rocks (Figure 3, upper left panel).These volcanics have calc-alkaline geochemical signatures, and along with associated volcaniclastic, metasedimentary and intrusive rocks, have also been interpreted as having formed in an Andean-type arc setting (Thomas et al., 2009).

Collectively, these studies establish the existence in central-northern Madagascar of a semi-continuously exposed, 450 km long belt of subduction-related intrusive and volcanic rocks (Figure 3), interpreted as the magmatic products of a continental Andean-type that developed along the western margin of the Rodinia supercontinent during 700 to 800 Ma in the late Neoproterozoic. Members of our research group wondered if remnants of this magmatic arc, fragmented and dispersed by younger plate tectonic processes, might be identified elsewhere in the Indian Ocean or in neighboring continental areas. We started in the Seychelles.

The presence of Precambrian granitic rocks in the Seychelles has been long known. About 40 of the 115 islands in the Seychelles Archipelago, including most of the big ones, consist almost entirely of Late Proterozoic granitoids. The only exceptions are Silhouette and North Islands, which are composed of Palaeocene (63 to 63.5 Ma) syenitic and trachytic rocks related to the Deccan plume. The remaining Seychelles islands are coralline atolls and reefs, with and without sand cays. Nearly all researchers who had studied the Seychelles granites interpreted them as having formed in extensional settings (plumes or rifts), perhaps being influenced by the well-known hotspot volcanoes present in many small ocean islands. We decided to re-evaluate this, to determine if the Precambrian granitoids might instead represent arc-related magmatic rocks, possibly correlative with those identified in Madagascar.

The geology of the granitic islands is quite simple; all consist of an assortment of intermediate, I-type granodiorites and monzogranites, crosscut by dolerite dykes (Figure 5a and b). There are no metamorphic or sedimentary rocks whatsoever. Bob Tucker supplemented the existing geochronology database for Seychelles granitoids with 16 new U-Pb zircon determinations on samples from seven different islands (Tucker et al., 2001). The range of well-determined ages is 703 ± 1 Ma to 809 ± 2 Ma (n = 21), although the vast majority (n = 19) cluster at 753 ± 4 Ma (Figure 5a), this corresponds very well with the span of magmatic ages in the arc-related belt of Madagascar. An important new finding was that U-Pb zircon ages of the Seychelles dolerite dykes are coeval with those of the granitoids, demonstrating the contemporaneity of granitic and basaltic magmas (Tucker et al., 2001). This accounts for the presence in some outcrops of complex intermingling features, including irregular masses and enclaves of hybrid lithologies such as diorite and quartz diorite that were produced by variable blending of silicic and mafic magmas (Figure 5c), and the geochemical arrays of intermediate rocks between granitoids and dolerite (Figure 5d) (Ashwal et al., 2002).

We also acquired new geochemical and isotopic data for 26 samples of granitoids, intermediate rocks and dolerites, which we used with existing older data to re-evaluate the petrogenesis of Seychelles magmatic rocks (Ashwal et al., 2002). For the granitoids, the results confirm earlier distinctions between “gray granites” (Mahé Group) and “pink granites” (Praslin Group), with the latter group characterized by incompatible element enrichment and more evolved Sr and Nd isotopic signatures. The differences can be attributed to variable involvement of source components during intracrustal partial melting: a juvenile (~700 to 800 Ma) mafic to intermediate source, and ancient (2.5 to 3.5 Ga) silicic material like the Archaean tonalitic basement rocks exposed in India and Madagascar. The dolerites can be modelled as basaltic melts derived from depleted mantle that assimilated variable amounts (0 to 15%) of Archaean silicic crust. Intermediate rocks show linear arrays between granitoids and dolerites for major and trace elements and isotopic compositions, confirming their origin by two-component magmatic blending and hybridization (Figure 5d). All of the petrologic, petrographic, geochemical, isotopic and chronologic data for the Seychelles igneous rocks are consistent with an Andean-type arc setting, the case for which was summarized by Ashwal et al. (2002).

Granites from the Seychelles were the first documented examples of silicic magmas with distinctly low δ¹⁸O (Taylor, 1968, 1977), a feature often taken as evidence for interaction between magma chambers and deeply circulating meteoric waters, implying formation in an extensional tectonic setting (e.g. Stephens et al., 1995). Our new measurements of oxygen and hydrogen isotopes in whole-rocks and mineral separates of quartz and feldspars show more variability in δ¹⁸O than earlier results; some, but not all Seychelles granitoids show low δ¹⁸O signatures. We also found that δ¹⁸O and δD are positively correlated, implying the involvement of two distinct source regions, one with “normal” stable isotope values, and another with low δ¹⁸O and δD values that must have been acquired by interaction with meteoritic water. These could correspond to the juvenile and ancient crustal sources discussed above. These results would be entirely consistent with an Andean-type arc setting, and we concluded that the δ¹⁸O depletion of some Seychelles granitoids need not require formation in a regionally extensional tectonic setting.

A new aspect of our work attempting to track the belt of Neoproterozoic arc rocks involved the acquisition of palaeomagnetic data, carried out with our colleague Trond Torsvik. We felt that such data might be revealing, inasmuch as all of the Seychelles magmatic rocks are entirely undeformed and unmetamorphosed, and might retain a record of primary magnetic properties. This would not be the case for the arc-related rocks of Madagascar, which were remagnetized by variable penetrative deformation and metamorphism up to granulite facies. After careful thermal and AF demagnetization, drill core samples of both the granitoids and dolerites show similar primary remanent magnetization directions, which were acquired at their coeval magmatic age of ~750 Ma (Torsvik et al., 2001a). This constrains the Seychelles palaeolatitude to have been ~30°N in the Late Neoproterozoic, in a position along the western margin of the Rodina supercontinent, consistent with a subduction-related Andean-type arc setting (Figure 7). The next section summarizes our efforts to track this arc even further, into northwestern India.

The Proterozoic Malani Igneous Suite (MIS) of Rajasthan (Figure 6a) is said to be the world’s third largest felsic magmatic province, with an areal extent of ~51 000 km² (e.g. Pareek, 1981), consisting of subequal volumes of contemporaneous granitoids and felsic volcanic rocks, with lesser mafic components. Our team visited, sampled and analysed these rocks, applying the combined approach of geochronology, petrology/geochemistry/isotopes and palaeomagnetism that we found to be successful elsewhere. We saw numerous exposures of granitic rocks crosscut by dolerite dykes that are remarkably similar to those in the Seychelles (Figure 6b), and thick sequences of associated felsic volcanics that are wonderfully preserved (Figure 6c).

Ages of MIS magmatism have been determined by several research groups, including ours, using various U-Pb zircon dating methods (TIMS, SIMS, LA-MS-ICPMS), and represent improvements over earlier Rb-Sr results. The range in ages of the felsic volcanic rocks is 751 to 771 Ma (n = 9) (Torsvik et al., 2001b; Gregory et al., 2009; Van Lente et al., 2009; Dharma Rao et al., 2012), which overlaps with those of the granitoids (764 to 768 Ma, n = 3) (Ashwal et al., 2013). A dolerite dyke 5 m thick yielded an age of 752 ± 18 Ma (Meert et al., 2013), demonstrating that the dolerites are coeval with the granitic rocks that they crosscut, as in the Seychelles. Intrusive and volcanic rocks of the MIS, therefore, formed contemporaneously with those of similar arc-related magmatic rocks in Madagascar and the Seychelles.

The MIS granitoids are best exposed in the region around the mountain resort town of Mt. Abu, in southwestern Rajasthan (Figure 6a). Here, they occur mainly as massive, undeformed, variably porphyritic syenogranites to alkali feldspar granites, although variably deformed equivalents, including augen gneisses, are present near the margins of plutonic bodies. This deformation appear to be related to Pan-African events, based on ⁴⁰Ar/³⁹Ar ages of 509 to 514 Ma from hornblende and biotite separates (Ashwal et al., 2013). The granitoids are crosscut by dolerite dykes up to 6 m thick (Figure 6b), in which primary mineralogy has been variably replaced by amphiboles, epidote and sericite, similar to many Seychelles dolerites. The MIS volcanic rocks consist of a relatively undeformed and unmetamorphosed, but gently tilted sequence dominated by rhyolitic to rhyodacitic flows and tuffs, with minor basaltic pillow lavas. This sequence is unconformably overlain by Early Cambrian (524 to 542 Ma) sandstones of the Marwar Supergroup, clearly exposed in the city of Jodhpur (Figure 6c). Several post-graduate student projects involved petrological and geochemical comparisons between MIS magmatic rocks and those of the Seychelles. There are striking major and trace element similarities between MIS granitic and rhyolitic rocks, and the Praslin Group granitoids of the Seychelles (Van Lente, 2002; Carter, 2005; Solanki, 2011). Likewise, MIS and Seychelles dolerites are also chemically comparable (Solanki, 2005).

All of the petrographic, petrologic, geochemical and geochronological data summarized above support the notion that the MIS of Rajasthan represents a part of the same Neoproterozoic continental arc that we identified in Madagascar and the Seychelles. Palaeomagnetic results for MIS felsic volcanic rocks yield a palaeolatitude of ~41°N at ~750 Ma, allowing a tight reconstruction of Madagascar – Seychelles – Rajasthan at that time (Figure 7) (Torsvik et al., 2001b). We suggest that these three terranes were all part of a contiguous magmatic arc of Andean proportions, representing east-directed subduction beneath the western margin of Rodinia (Figure 7).

Break-up of the Rodinia supercontinent by major rifting events may have started as early as ~870 Ma, and fundamentally resulted in its fragmentation into two major continental entities that drifted apart laterally, forming a Pacific-sized ocean basin referred to as the Mozambique Ocean (e.g. Stern, 1994). The two major continental fragments were first termed East Gondwana (consisting of Australia, India and Antarctica) and West Gondwana (consisting of Africa, South America and parts of Laurentia) by McWilliams (1981). They were eventually rejoined during Pan-African times (~700 to 500 Ma) by Tibetan-style collisional events, following closure of the Mozambique Ocean, to form the Gondwana supercontinent (Hoffman, 1991, 1999; Moores, 1991; Rogers et al., 1995) (Figure 8). The zone of convergence between East and West Gondwana is a ~6000 km long belt of Neoproterozoic high grade metamorphic rocks, coeval granitoids and ophiolites, collectively referred to as the East African Orogen (Stern, 1994), and includes, among other things, the Mozambique and Zambezi Belts and parts of the Arabian-Nubian Shield (Figure 8). Madagascar, Sri Lanka and southern India were profoundly affected by high-grade Pan-African metamorphism and deformation associated with Gondwana assembly, but interestingly, the ~750 Ma arc-related rocks of the Seychelles and Rajasthan seem to have largely escaped these effects (Figure 8).

It was Kevin Burke, probably my most influential mentor, who suggested that a well-defined record of two parts of the Wilson cycle could be preserved in Alkaline Rocks and Carbonatites (ARCs) and their Deformed equivalents (DARCs). ARCs are well known to form in continental rift environments (although some are associated with mantle plumes), and the metamorphic fabrics of DARCs can confidently be ascribed to deformational effects during collisional events during which oceans have closed. We used the locations of DARCs in southern Africa to map the outlines of cryptic sutures (Burke et al., 2003) (Figure 9, left), and later applied these ideas in a similar way for India (Leelanandam et al., 2006), northern Norway and northwestern Russia (Burke et al., 2007).

Through our collaborative work with several talented geochronologists, we were able to demonstrate that reasonably precise ages of both tectonic events can be preserved in the minerals contained within DARC rocks. The Tambani body of southwestern Malawi is a 75 km² occurrence of Neoproterozoic, banded nepheline syenite gneisses, some of which contain euhedral zircon crystals up to 5 cm across; the body is crosscut by carbonatitic dykes. The zircons yield magmatic ages of 730 ± 4 Ma by the TIMS method, coincident with 11 SHRIMP spot analyses, which give an upper intercept age of 729 ± 7 Ma. Monazite, however, records an age of 522 ± 17 Ma (by TIMS), interpreted to represent the time of amphibolite grade metamorphism. A single DARC hand specimen, therefore, records two well-defined tectonic events- the rifting of Rodinia and the collisional construction of Gondwana.

A spin-off of our work on ARCs and DARCs led to the suggestion that DARC material taken into the mantle lithosphere to depths of ~100 km during subduction and collisional events might provide source materials for later ARC magmatism (Figure 9, right) (Burke et al., 2003). This would account for the observation of recurrent alkaline and carbonatitic magmatic activity over hundreds of millions of years in restricted areas (Bailey, 1974, 1992). A good example can be found in southern Malawi, where the Cretaceous (~140 Ma) Salambidwe circular ring complex of phonolitic trachytes, pulaskites and assorted alkaline syenites was emplaced ~15 km southwest of the 730 Ma Tambani DARC body, about 380 million years after Pan-African collisional events. A small subducted component of DARC material, blended into depleted subcontinental mantle lithosphere could account for the compositions of younger ARC magmas, as shown by Sr, Nd and Hf isotopic studies of southern African ARC and DARC occurrences at Bull’s Run (South Africa, 1134 Ma), Tambani (Malawi, 730 Ma) and the Chilwa Alkaline Province (Malawi, 130 Ma) (Ashwal et al., 2013). This offers an alternative to the widely held view of a deeper, asthenospheric source for ARCs (e.g. Bell and Simonetti, 2010).

Modern reconstructions of Gondwana (e.g. de Wit et al., 1988; Buiter and Torsvik, 2014) are remarkably similar to the one published by Alex DuToit as Figure 7 in his 1937 book (Figure 10). The broad details of Gondwana’s break-up history into the present-day configuration of continental entities are fairly well established, based on decades of combined studies of sea-floor magnetic anomaly patterns, palaeomagnetism, geology, geochronology, sedimentology, tectonostratigraphy and palaeontology, with a body of literature too extensive to reference here. Excellent animations of Gondwana break-up created by Colin Reeves and his colleagues are freely available (http://www.reeves.nl/gondwana), and illustrate that during the initial stages starting at about 165 Ma, Madagascar + India + Antarctica + Australia separated and moved southward with respect to Africa, along the Davie Fracture Zone. At ~90 Ma, Madagascar separated from India, which drifted rapidly northward, eventually colliding with Asia to form the Himalayas, starting at ~55 Ma. It is interesting to note that the plate kinematics of Gondwana break-up may be recorded in transtensional pull-apart structures in the Phanerozoic basins of western Madagascar. For example, in the Morondava Basin of southwest Madagascar, dextral structures dominate in Jurassic units, recording the initial southward drift of the island with respect to Africa, but sinistral structures are abundant in Eocene strata, recording the separation and northward drift of India with respect to Madagascar (Figure 11). The details of the structural responses to these stress and kinematic regimes are complex, and are actively being studied (e.g. Schandelmeier et al., 2004).

The role of deep mantle plumes in initiating, or at least in assisting continental break-up has been contentious for decades, and is still widely being investigated. The possible causative connection between plumes and continental fragmentation comes from several examples of coincidences in time and space of voluminous magmatic products of Large Igneous Provinces (e.g. continental flood basalts) with sites where coeval continental rifting has taken place (e.g. Storey, 1995). Some workers consider that deep mantle plumes play an active role in driving plate motions, and initiate continental break-up by thermal weakening, uplift, cracking and forming new oceans by lateral spreading (Morgan, 1971, 1981; Storey and Kyle, 1997). Others suggest that plate boundary forces (e.g. slab-pull and/or ridge-push) are the primary drivers of continental fragmentation, and although the horizontal stresses associated with mantle plume impingement are incapable of initiating break-up (Hill. 1991), they may assist and augment the separation of continental entities (White and McKenzie, 1995; Buiter and Torsvik, 2014).

At least four deep mantle plumes, with associated Large Igneous Provinces (LIPs), were involved in Gondwana break-up: Karoo (182 Ma), Etendeka/Paraná (132 Ma), Madagascar (90 Ma) and Deccan (65 Ma) (Figure 10); the present locations of these hot spots are suggested to lie at Bouvet, Tristan, Marion and Réunion Islands, respectively. All four of these LIPs, when reconstructed to their eruption sites at their times of formation, lie above the margins of the sub-African, large low shear-wave velocity province (LLSVP), consistent with derivation of the parental plumes from the D” region of the lowermost mantle (Burke and Torsvik, 2004). The Karoo plume and LIP is widely considered to have played an active role in the initial break-up of Gondwana (e.g. Storey and Kyle. 1997), even though the first truly oceanic crust seems to have formed some 10 to 15 m.y. after the extrusion and intrusion of the vast majority of Karoo magmas at 182 ± 1 Ma (Lawver et al., 1991; Svensen et al., 2012).

In West Gondwana, the initial separation of Africa from South America and the opening of the South Atlantic Ocean coincided with the eruption and emplacement of the Paraná/Etendeka LIP at ~132 Ma (Figure 10), with magmatic products now covering vast areas of Brazil, Uruguay, Paraguay and Argentina (Paraná LIP), as well as parts of Namibia and Angola (Etendeka LIP). Our palaeomagnetic work in the Etendeka LIP provides a refined South America – Africa reconstruction for the Early Cretaceous, and also places the Paraná/Etendeka plume above the plume-generation zone at the margin of the sub-African LLSVP (Owen-Smith et al., 2019). Our petrology, geochemistry and isotope work on a small 130 ± 1 Ma layered intrusion, the Doros Complex, demonstrate its emplacement as successive pulses of magmatic mushes (Owen-Smith and Ashwal, 2015a and b), and its derivation from depleted asthenospheric, rather than lithospheric mantle sources (Owen-Smith et al., 2017).

For the Madagascar LIP, our U-Pb zircon and Ar-Ar geochronology, coupled with palaeomagnetic work, suggest that Madagascar was almost completely blanketed by basaltic lavas from the Marion plume (Figure 10), erupted at 88 ± 4 Ma (Torsvik et al., 1998). Such Late Cretaceous magmatic products appear to be absent in India, except for the 91.2 ± 0.2 Ma columnar-jointed rhyolites exposed on the St. Mary’s Islands located off the western coast of central India (Torsvik et al., 2000). A possible explanation for the paucity of magmatic rocks from the Marion plume in peninsular India is discussed in the next section. The Marion plume is widely implicated as a trigger to the separation of Madagascar and India at ~90 Ma (Reeves, 2014), and may also have contributed to India’s rapid (up to 20 cm/yr) drift northeastward toward Asia (van Hinsbergen et al., 2011).

Our contributions toward the understanding of the Deccan LIP, which triggered (via the Réunion plume, Figure 10) the separation of the Seychelles from India at the Cretaceous–Tertiary boundary, focused on a plutonic/volcanic alkaline complex, exposed on Silhouette and North Islands in the Seychelles (Figure 5a) (Owen-Smith et al., 2013). Here, a subvolcanic chamber with alkali gabbro, syenite, syenodiorite and alkali granite underlies the remnants of the volcanic cover composed of trachytic tuffs cut by microsyenitic and doleritic feeder dykes. These magmas were formed by fractionation from basaltic parents, and their isotopic compositions demonstrate derivation from a Réunion plume source equivalent to that of the Deccan basalts in India. We constrained ages of the Seychelles alkaline rocks to between 63 and 63.5 Ma using U-Pb and Ar-Ar methods (Ganerød et al., 2011), demonstrating that they represent some of the youngest magmatic products of Réunion plume activity in the Deccan event prior to continental break-up. Our palaeomagnetic work leads to a reconstructed volcanic region that, like the other LIPs discussed here, was positioned radially above the plume-generation zone of the sub-African LLSVP (Ganerød et al., 2011).

Some of the latest work being carried out by research groups with which I am involved is beginning to show that the processes involved in continental break-up in general, and the development of the present-day Indian Ocean in particular, may be far more complex than previously understood. This relates to our recent discovery of a fragment of ancient continental crust beneath the young ocean-island volcano of Mauritius (Torsvik et al., 2013; Ashwal et al., 2017) (Figure 1). Mauritius is the second-youngest volcanic island in a hotspot track that extends from Réunion (the site of the present-day active deep mantle plume), through the southern Mascarene Plateau, the Laccadive-Chagos Ridge, and into the 65.5 Ma Deccan Large Igneous Province, the oldest part of the track (Figure 1) (Richards et al., 1989; Duncan, 1990; Courtillot et al., 2003). Mauritius lavas have been divided by age into the Older (9.0 to 4.7 Ma), Intermediate (3.5 to 1.66 Ma) and Younger Series’ (1.0 to 0.03 Ma) (McDougall and Chamalaun, 1969; Nohda et al., 2005; Moore et al., 2011) (Figure 12a). Most voluminous, at ~75 000 km³, is the Older Series, which represents the shield-building stage; this was followed after a hiatus of ~1.2 m.y. by lavas of the Intermediate Series, and finally, by those of the Younger Series, after a quiescent period of ~0.66 m.y. Although the thin basaltic lavas of the two periods of rejuvenescent volcanism represent ~75% of the Mauritian surface area, their volume is only ~35 km², or ~0.05% of the total volume of volcanic rocks on the island.

The background to our work in Mauritius involves an interesting story, which is briefly summarized here. During a two-day layover in Mauritius, en route to Rajasthan, our colleague Bjørn Jamtveit (University of Oslo) collected a sample of Miocene (~9 Ma) basalt, speculating that it might contain xenocrysts of much older zircons, as had been demonstrated during petroleum exploration in young (Palaeocene, 57 to 61 Ma) lavas from the Faroe Islands. Surprisingly, the Mauritius sample yielded zircons of Permian age (~250 Ma), and Bjørn submitted a manuscript to Nature, announcing this unexpected result. Luckily, the paper was rejected, because the Permian zircons recovered turned out to be contaminants from the crushing equipment used. Being a persistent (and prescient) scientist, Bjørn solicited the help of two colleagues, Ebbe Hartz and Hans Amundsen, who returned to Mauritius to collect samples of beach sands, which would obviate the problem of contamination by crushing equipment. Twenty zircon grains were recovered from two sampling sites, and TIMS U-Pb analyses surprisingly returned Precambrian ages of 660 to 1971 Ma. This was the first indication that these grains might have come from a fragment of old continental crust that lay beneath the young Mauritian volcanoes.

Using this result, coupled with a new gravity inversion model for the Indian Ocean by our colleague Nick Kusznir (University of Liverpool) that could be used to infer crustal thicknesses, we suggested that small fragments of continental crust may underlie not only Mauritius, but also other parts of the Indian Ocean, including the Cargados-Carajos and Nazareth Banks, Saya de Malha, Chagos, and the Laccadive Islands. We further suggested that these fragments were once joined together in a Precambrian microcontinent called “Mauritia” (Figure 12b), and became scattered throughout the Indian Ocean during Gondwana break-up; our results and ideas were published in Nature Geoscience (Torsvik et al., 2013). This resulted in a great deal of skepticism and criticism from the geological community, who felt that the zircons we recovered may have been transported to Mauritius by winds, ocean currents, fragments of floating pumice, vehicle tyres, footwear, or even birds. We considered these as possible, but very unlikely explanations, although we could not disprove them.

During a subsequent expedition to Mauritius in January, 2014, on a different research project, I became interested in the origin of Mauritian trachytes (Figure 13a and b), which occur as small, alkali-rich magmatic bodies at five localities on the island (Figure 12a). Using a broad array of petrographic, mineralogical, geochemical and isotopic data, I proposed that instead of having formed by extreme fractional crystallization of basaltic magmas (the conventional view), the Mauritian trachytes instead represent direct partial melts of enriched mantle sources (Ashwal et al., 2016). Because the trachytes are highly enriched in Zr (830 to 1497 ppm), I expected that they would likely contain zircons, some of which might corroborate our suggestion of hidden continental material beneath the island. Thirteen zircon grains were recovered by Michael Wiedenbeck (GeoForschungsZentrum, Potsdam), using only a stainless steel rolling pin, and no crushing or grinding equipment, to avoid the risk of laboratory contamination, and he analysed the grains using the newly-installed Cameca 1280 SIMS instrument in his laboratory. Forty individual spot analyses on ten of the thirteen grains yielded ages of 4.9 ± 0.1 to 6.2 ± 0.1 Ma (average 5.7 ± 0.2 Ma), representing the time of late Miocene emplacement of the trachytes. Twenty spot analyses of the remaining three grains, however, gave Archaean ages of 2522 ± 11 Ma to 3030 ± 5 Ma (Figure 13c), thus confirming that a fragment of ancient continental crust must underlie the island of Mauritius, and contributed zircon xenocrysts to Miocene trachytic magmas on their way to surface eruption sites (Figure 14). Our work was published in Nature Communications (Ashwal et al., 2017), and led to an unexpected frenzy of media attention (compilation available on-line from Altimetric: https://www.altmetric.com/details/15992392/news).

We speculate that based on the frequency spectrum of Precambrian ages, the hidden fragment of continent beneath Mauritius most likely resembles the geology of surface exposures in central Madagascar, rather than of those in the Dharwar Craton of southern India or the Seychelles. This influenced our new reconstruction of Mauritia (Figure 15), which was a microcontinent situated between Madagascar and India before Gondwana break-up. The existence and location of Mauritia may explain the difficulties in attempts to make detailed correlations of geological features between Madagascar and India, assuming they were directly juxtaposed (e.g. Yoshida et al., 1999), and also may also account for the paucity in India of 90 Ma Marion plume magmatic products, as discussed above.

These results suggest that continental break-up is complex and messy, and may result in the ocean floor being “littered” with continental fragments of variable sizes and shapes. The mechanisms by which such continental fragments (or “microcontinents”) form are not completely understood, but recent suggestions include plume-assisted mid-ocean ridge relocation events (“ridge jumps”, Müller et al., 2001), spreading asymmetries at mid-oceanic ridges (Goff and Cochran, 1996) and plate motion reorganizations (Whittaker et al., 2016).

Other examples of possible “hidden” microcontinents (Figure 16) include southeastern Iceland, where our group used isotopic signatures and gravity inversion modelling to suggest that Precambrian continental crust, representing an extension of the Jan Mayen microcontinent, exists beneath the recently (violently) active Öræfajökull volcano (Torsvik et al., 2015). Likewise, at the Kerguelen LIP, Cretaceous basalts bear geochemical and isotopic signatures of Precambrian continental crust (Weis et al., 2001; Frey et al., 2002), and ODP drillcores show that fluvial conglomerates intercalated with the basalts contain clasts of garnet-biotite gneiss and other crustal rocks that bear zircon and monazite with ages as old as 2547 Ma (Nicolaysen et al., 2001). In the Vanuatu oceanic arc, 2 000 km east of mainland Australia, late Eocene to Miocene (16 to 35 Ma) andesites contain inherited zircons with ages up to 2.8 Ga, implying the presence of a continental fragment beneath the arc (Buys et al., 2014). This may be a part of the proposed hidden continent of “Zealandia”, an almost entirely submerged, ~5 x 10⁶ km² region of inferred continental crust in the southwest Pacific Ocean (Mortimer et al., 2017). The ancient zircons discovered in other unexpected places like the Mid-Atlantic Ridge (300 to 1600 Ma, Pilot et al., 1998; 53 to 3200 Ma, Skolotnev et al., 2010) and the Galapagos (378 to 3000 Ma, Rojas-Agramonte et al., 2016) await confirmation (e.g. in terms of possible laboratory contamination effects) before they can be confidently interpreted as indicating the presence of continental fragments. These discoveries and possibilities call into question our definition of the term “continent” in a geological context, especially in terms of size, shape, thickness and composition, and inspire a reconsideration of the distribution of continental material on the surface of the Earth.

In his studies of “comparative geology”, Du Toit used not much more than a hammer and camera, with his expertise in geological mapping, stratigraphy, petrology, palaeobiology, structural geology, geomorphology and palaeoclimatology, to test the hypothesis of continental drift, based on careful observations and measurements from many parts of the world. Du Toit’s legacy is that his integrated, multidisciplinary approach led to a courageous espousal of continental drift theory, which was widely criticized at the time, but would later be proven fundamentally correct, and resulted in the current paradigm of plate tectonics as a global explanation for how the Earth works.

Our work in comparative geology also employs a multidisciplinary approach, using some of the newer methodologies in petrology, geochemistry, geochronology, palaeomagnetism and geodynamics that were developed since Du Toit’s time. Our attempts to provide more precise correlations of wandered continental entities and reconstructions of past continental configurations contribute to a deeper geodynamic understanding of the Earth at present, as well as in the distant past. This has obvious applications to exploration for ore deposits and other objects of geological interest. But our contributions cannot compare to the prescient efforts of Du Toit, because our results merely provide “refinements” to the geodynamic paradigm he pioneered.

I am immensely grateful to Gordon Chunnett and the GSSA Fellows Committee for selecting me as the 35th Alex Du Toit Memorial Lecturer; I consider this as a great honour. This lecture, with the same title as that of this paper, was presented at the following nine venues in 2018: University of Johannesburg (17 July), University of Kwa-Zulu Natal, Durban (30 August), Geological Society of Zimbabwe, Harare (6 September), GSSA Bushveld Branch, Rustenburg (20 September), Nelson Mandela University, Port Elizabeth (27 September), University of the Free State, Bloemfontein (17 October), GSSA Egoli Branch, Johannesburg (24 October), University of the Western Cape, Cape Town (7 November) and the South African National Space Agency, Hermanus (8 November). I will not forget the hospitality and cheerful reception I received from the staff and audiences at all of these places. The research summarized here represents nearly 30 years of delightful and productive collaboration with a large number of colleagues and students, including, in alphabetical order:

Finally, I am indebted to the following institutions for supplying research funding and other forms of support (also listed alphabetically): Anglo-American, BHPBilliton, Centre for Earth Evolution and Dynamics (CEED, University of Oslo), Geological Survey of Norway, Gold Fields, Norwegian Research Council, National Research Foundation (South Africa), National Science Foundation (USA), Rand Afrikaans University (now University of Johannesburg), University of the Witwatersrand, and the World Bank.

Editorial handling: S. McCourt.

Alexander L. Du Toit Memorial Lectures No. 35 (2018)