Carbonatites have always been controversial (Mitchell, 2005). Their magmatic origin was disputed in the early days of the last century, regardless of the fact that experiments clearly demonstrated the crystallization of magmatic calcite (Wyllie and Tuttle, 1960). The observation of the eruption of natrocarbonatite lava in Oldoinyo Lengai (Dawson 1962) finally convinced petrologists that they were dealing with the products of magmatic carbonate liquids. Since that time, further controversies have emerged, especially regarding the ultimate origin of carbonatite magmas, for which there are two ‘endmember’ hypotheses. The generally accepted hypothesis is based on isotopic evidence and suggests that carbonatites are from deep asthenospheric sources, such as mantle plumes (Bell, 2001; Bell et al., 2004). This fails to explain why carbonatites are essentially confined to the continental lithosphere and are extremely rare in the ocean basins (Woolley and Kjarsgaard, 2008; Woolley and Bailey, this issue), and leads to the alternative hypothesis of lithosphere-generated carbonatitic magmatism. It may be that this is simply because we have not yet understood how to identify carbonatites in oceanic regions (Bailey and Kearns, this volume), or there may be some more profound reason why carbonatites cannot form within or erupt through oceanic lithosphere.

A recent database and world distribution map of carbonatites supports previous observations of the spatial and temporal aspects of these rocks, and provides new observations that are important for understanding their petrogenesis. These data reveal that there is an overwhelming concentration of carbonatites in Precambrian cratonic areas, most of which are elevated topographically. Thus, although approximately two-thirds of carbonatites are Phanerozoic in age, at least 88% of all dated carbonatites are located in the cratons, demonstrating a remarkable tendency for a Precambrian host. This observation suggests a link with kimberlites as diamond-bearing kimberlites are confined to the Archaean areas of cratons. The age data show that in many carbonatite-bearing provinces there has been repetition of carbonatite emplacement, with up to five episodes separated by hundreds of millions of years. In at least three provinces such activity extends from the late Archaean to relatively recent times and, because of the drift of the plates, this would seem to preclude any direct role for mantle plumes in carbonatite genesis. Magmatism is activated when lithosphere lesions are reopened in response to major changes in global plate movement patterns.

In the Calatrava province of central Spain, numerous Quaternary pyroclastic vents have erupted carbonatite magmas carrying silicate melt fragments, mantle debris and megacrysts. Lava flows are rare. Maar and scoria deposits have carbonate matrices and pass into tuff sheets with carbonate contents >50%, which are spread widely away from the eruptive centres and constitute the most abundant form of effusive carbonate. Immense quantities of mantle debris are present in the erupted material. The tuffs have a distinctive fabric, which consists of a pale matrix carrying black silicate glass clasts that contain globules of immiscible carbonate and carbonate phenocrysts. There is evidence of similar volcanism in the Limagne province of central France and in other intra-continental provinces in Europe and Africa. About 500 vents have been identified in France and Spain: all the vents examined to date have erupted carbonatite magma. Such eruptions are not generally recognized in classical volcanology. As pyroclastic carbonatite was not previously recognized in Spain and France, a detailed examination of other mafic and ultramafic alkaline provinces, where research has traditionally concentrated on lava flows, is vital. For any search to be successful, evidence from the pyroclastic rocks will be required.

Melt compositions in equilibrium with peridotite assemblages were determined in the analogue system Na2O–CaO–MgO–Al2O3–SiO2–CO2 at 3 GPa with Ca/Ca + Mg = 0.56–0.43 and up to 6 wt.% Na2O. There is a greater compositional range generated isobarically over a larger temperature interval than in a sodium-absent system: increasing sodium content drives liquids to compositions with lower CaO and higher SiO2 concentrations. A positive correlation between silica and Na2O content of liquids produced at constant temperature is due to the depolymerization of silicate tetrahedra in the presence of monovalent cations, as in the volatile-free system. Liquids with Na2O >6 wt.% occur in association with wehrlites as the composition of diopsidic pyroxene expands towards enstatite with addition of Na2O, decreasing the orthopyroxene content of peridotite. The orthopyroxene-out curve intersects an enriched mantle solidus at 3 GPa where near-solidus liquids have Na2O = 7 1.5 wt.%. Sodium partitioning between a metaluminous liquid and clinopyroxene follows the jadeite partitioning models calculated for the dry silicate system but sodium partitions into peralkaline carbonated liquids as both the pyroxene and the carbonate molecules. The peralkaline liquids generated are essentially carbonated silicate melts that are analogous to silica-bearing carbonatites and silicocarbonatites from a range of possible metasomatized mantle sources.

The island of São Vicente has the most abundant carbonatite outcrops in the Cape Verde Islands. A field survey of the main outcrops has shown that they consist of extrusive carbonatites, carbonatite dykes and small apophyses of intrusive carbonatite. These outcrops are spatially related to nephelinites. The compositions of the extrusive carbonatites and dykes plot close to, and within, the magnesiocarbonatite field. In contrast, the intrusive carbonatites are calciocarbonatites, with similar average strontium contents to those of extrusive carbonatites and dykes (around 4000 ppm), but remarkably low barium, niobium and total rare earth element concentrations. Whole-rock geochemistry indicates a strong affinity between the nephelinites and intrusive carbonatites, such that the latter could represent fractionation products of the same parental magma. This is in agreement with radiogenic isotope geochemistry, which shows a very restricted range of compositions in the Sr, Nd and Pb systems. Fractionation from a common parental magma occurred in two main steps: high-temperature nephelinite crystallization and high-temperature carbonatite immiscibility. The carbonatitic melts crystallized in two different environments, as follows: (1) a shallow intrusive environment, giving rise to the early calciocarbonatite cumulates; and (2) a vapour-dominated, extrusive environment, producing the later magnesiocarbonatites.

The magma source and evolution of a zoned breccia pipe on the southern Beara Peninsula in southwest Ireland are investigated using the geochemistry of the host mineral assemblages. The clast-poor inner zone of the pipe has a magnesium-rich silicocarbonatite whole-rock composition (14.30 wt.% MgO; 31.80 wt.% SiO2). The silicocarbonatite has retained an ultimate mantle source 13C isotopic composition after metamorphism, consistent with the presence of mantle debris. The silicocarbonatite is Cr-, Ni- and Co-rich (847 ppm, 611 ppm and 60 ppm, respectively) but REE depleted compared with volcanic dolomite carbonatites worldwide. The mineral assemblage consists of Sr-rich (0.55 wt.% SrO) ferroan dolomite, magnesite and pseudomorphs of chlorite after phlogopite, consistent with derivation from a carbonated and hydrated mantle. However, chrome spinel crystals (≤4 40.14 wt.% Cr2O3) are compositionally indistinguishable from unmetasomatized spinel macrocrysts in kimberlites. The silicocarbonatite is inferred to represent a magma produced by partial melting of metasomatized mantle at physical conditions between those in which primary dolomite carbonatite and ultramafic magmas of high-pressure origin form. The primary silicocarbonatite magma ascended and sampled mantle material in a manner similar to kimberlite, and subsequently lost volatile components due to release of metasomatic fluids and later metamorphism.

Three Angolan carbonatites were selected to evaluate the change in composition of pyrochlores during magmatic evolution: the Tchivira carbonatites occur in a plutonic complex, the Bonga carbonatites represent hypabyssal carbonatites and the Catanda carbonatites are volcanic in origin. In Tchivira pyrochlore, zoning is poorly developed; fluorine is dominant at the Y site; chemical zoning may arise as a result of substitutions for Nb in the B site; and the rare earth element (REE), U, Th and large-ion lithophile element (LILE) contents are very low. Pyrochlores from Bonga show oscillatory zonation; the F and Na contents are lower than those in the pyrochlores from Tchivira; and as substitution of Na at the A site increases, the Th, U, REE contents and inferred vacancies also increase. Pyrochlores from Catanda display complex textures. They generally have a rounded corroded core, which is mantled by two or three later generations. The core composition is similar to the Bonga pyrochlores. The rims are enriched in Zr, Ta, Th, Ce and U, but depleted in F and Na. In pyrochlores from the Angolan carbonatites, the F and Na contents decrease from plutonic to volcanic settings and there is enrichment of Th, U and REE in the A site and Ta and Zr in the B site. Zoning may be explained by changes in the activity of F, due to the crystallization of fluorite or apatite in the plutonic and hypabyssal carbonatites, or to volatile exsolution in the volcanic carbonatites.

The Virulundo carbonatite in Angola is one of the largest in the world and contains pyrochlore as an accessory mineral in all of the carbonatite units (calciocarbonatites, ferrocarbonatites, carbonatite breccias and trachytoids). The primary magmatic pyrochlore is fluorine dominant and typically contains about equal molar quantities of Ca and Na at the A site. High-temperature hydrothermal processes have resulted in the pseudomorphic replacement of the primary pyrochlore by a second generation of pyrochlore with less F and Na. Low-temperature hydrothermal replacement of the first and second generation pyrochlore, associated with quartz-carbonate-fluorite vein formation in the carbonatite, has produced a third generation of pyrochlore, with a high Sr content. The Sr appears to have been released by low-temperature hydrothermal replacement of the primary magmatic carbonates. Finally, supergene alteration processes have produced late-stage carbonates, goethite, hollandite and rare earth element (REE) minerals (mainly synchysite-(Ce), britholite-(Ce), britholite-(La), cerite-(Ce)). Cerium separated from the other REE s in oxidizing conditions and Ce4+ was incorporated into a late generation of supergene pyrochlore, which is strongly enriched in Ba and strongly depleted in Ca and Na.

This paper reports an investigation of the crystallization products of K-rich silicate and carbonate melts trapped as melt inclusions in clinopyroxene phenocrysts from the Dunkeldyk alkaline igneous complex in the Tajik Republic. Heating experiments on the melt inclusions suggest that the carbonate melt was formed by liquid immiscibility at 1180°C and ∼0.5 GPa. The carbonate-rich inclusions are dominated by Sr-bearing calcite, and rich in incompatible elements. Most of the silicate minerals are SiO2-poor and rich in K, Ba and Ti. Leucite, kalsilite and aegirine are the earliest magmatic minerals. High Ba and Ti contents in the melt resulted in the crystallization of Ba-rich K-feldspar, titanite, perovskite and Ti-bearing garnet, and the rare Ba-Ti silicates fresnoite and delindeite. The last minerals to crystallize from volatile-rich melts and fluids were aegirine, götzenite, K-Ba- and Ca-Sr-bearing zeolites, fluorite and strontium-rich baryte. Interaction of the early minerals with residual melts and fluids produced Ba-rich phlogopite and Sr-rich apatite.