Carbonatites crystallize from mantle-derived carbonate- and volatile-rich melts that exsolve large amounts of fluids during their ascent through and emplacement into the crust. A global review of available fluid inclusion data for carbonatitic systems from variable emplacement depths identified four types of fluid inclusions: (type-I) vapour-poor H₂O-NaCl fluids with <50 wt.% salinity; (type-II) vapour-rich H₂O-NaCl-CO₂ fluids with <5 wt.% salinity; (type-III) multi-component fluids with high salinity without CO₂; and (type-IV) multi-component fluids with high salinity and high CO₂. This global data set indicates initial release of type-I saline brines that may either separate into immiscible type-II and -III fluids (eruption?) or may continuously evolve into type-IV fluids (sealing?). Moreover, fluid inclusions in early magmatic apatite crystallization suggest initial fluid release (type-I) at depths of 12-16 km (brittle-ductile transitions zone), which may be related to a sudden pressure drop initiated by crustal fracturing during rapid, forceful and discontinuous magma ascent.

Our model for the ascent of carbonatitic magmas is adopted from a jackhammer-like process, which explains the apparent absence of shallow carbonatite magma chambers, reflects the observed intrusion geometries, identifies fenitization as a process induced by fluids released during magma ascent and final emplacement, and demonstrates the formation of fluid induced brecciation related to magma ascent. The proposed model of a self-sustaining system is also in agreement with a turbulent ascent and high ascent rates, which allows for the transport of mantle xenoliths through the crust as observed in several cases.

The igneous nature of carbonatites has been controversial for decades until experimental work in the 1960s and 1970s conclusively showed that carbonatite melts can exist in geologically reasonable conditions. The observation of natrocarbonatite eruptions at Ol Doinyo Lengai further confirmed their igneous nature. However, these lines of evidence for igneous carbonatites have been a red herring: many ideas, processes, and terms deriving from silicate magma systems were inaccurately projected into carbonatite systems. Instead, carbonatites should be considered as hybrid metasomatic and magmatic cumulate rocks. The low viscosity and efficient wetting of carbonatite melts makes them behave more like hydrothermal fluids, preventing formation of magma chambers. Their high chemical reactivity and disequilibrium with their host rocks leads to rapid and substantial chemical exchange with their surroundings. Therefore, the chemical composition of their host rocks imparts a first order effect on the mineral assemblage observed in solidified carbonatite rocks. Rare earth elements are typically incompatible during carbonatite melt fractionation. Whether REE mineralisation is observed in Ca, Mg, or Fe-dominated mineral assemblages is strongly dependent on the degree and nature of silicate contamination. In special cases, REE can be highly compatible and mineralisation forms early rather than late. Enigmatic light REE mineralisations in fenite-like assemblages can likewise be explained as an end-member of carbonatite–rock interaction.

Carbonatites host the main REE deposits in the world, with bastnaesite being the main REE-bearing mineral of interest. However, the nature of the enrichment process, magmatic vs hydrothermal, is still debated. This study aims to experimentally determine the behaviour of REE elements during carbonatite crystallisation, and if bastnaesite can be directly crystallised from a carbonate melt.

Crystallisation experiments have been done from 900 to 600°C at 1 kbar on a REE-rich calciocarbonatitic composition. Calcite (Ca,REE)CO₃ is the dominant magmatic mineral, so the residual melt evolves toward natrocarbonatitic compositions as crystallisation proceeds. A small amount of britholite (REE,Ca)₅((Si,P)O₄)₃(OH,F) is observed at high temperatures and is replaced by phlogopite KMg₃(AlSi₃O₁₀)(OH)₂ and apatite (Ca,REE)₅(PO₄)₃(F,OH) at T < 650°C. A small amount of pyrochlore (Ca,Na,REE)₂Nb₂O₆(OH,F) is observed at T < 700°C.

No bastnaesite has been found in any crystallisation experiment. We thus performed a bastnaesite saturation experiment at 600°C. The melt saturated with bastnaesite however contains 20 wt% of REE: such high value implies that magmatic saturation of bastnaesite is unlikely to happen in nature.

F, Cl and water decrease the temperature of calcite saturation, allowing the system to crystallise at lower temperatures. REE are slightly incompatible with calcite, especially at low temperatures. The residual carbonate melt is thus enriched as crystallisation proceeds. Finally, we collected textural and chemical evidence suggesting the presence of a Na,Cl,REE-rich fluid at high temperatures and under hydrous conditions. Such Na,Cl,REE-rich fluids may play a critical role in the remobilisation of REE and the bastnaesite crystallisation during subsolidus reactions.

The Oulad Dlim massif in the southernmost part of Morocco hosts several carbonatite bodies of different ages. The older carbonatite (1.85 Ga) occurs in the eastern Oulad Dlim massif in the Gleibat Lafhouda area and consists of three juxtaposed magnesiocarbonatite outcrops. They are associated with glimmerite, hosted by Archean gneiss, and unusually intruded by massive IOA deposits. The latter contains up to several wt% REE related to numerous monazite-(Ce) inclusions within large apatite crystals. Columbite-(Fe) is the main Nb-mineral and occurs closely associated with Fe-phases, whereas microlite and Ta-rich columbite-(Fe) are mainly associated with coarse-grained apatites hosted by Fe-oxides and silica breccia. Geochemical characteristics and textural relationship suggest that they are genetically linked to the carbonatite and likely formed by late hydrothermal fluids at multiple stages. Small outcrops of nepheline syenite occur at several km from this carbonatite and might be genetically related. The youngest carbonatite (104 Ma) is a soevite and crops out within a ring structure composed of silica breccia and Fe-oxide mineralization at the Twihinat area of the western Oulad Dlim massif without visible associated alkaline rocks. All outcropping rocks at Twihinat show epigentic REE-Nb mineralization, mainly as bastnaesite within the carbonatite and silica breccia and monazite within the Fe-oxides. Pyrochlore senso-stricto occurs within the carbonatite, whereas cerio-pyrochlore is dominant in the silica breccia. The mineralogical and geochemical signatures of all Twihinat rocks suggest ore precipitation from multistage REE-Nb-rich hydrothermal fluids that percolated through the carbonatites and the associated rocks.

Contamination of carbonatite melts is often neglected due to a fast magma ascent and low liquidus temperatures. However, increased silicate mineral formation observed in numerous carbonatite occurrences world-wide requires an external Si introduction. Our study demonstrates that carbonatite dykes penetrating different lithologies of Palabora (South Africa) shows different modes of mineralogical modification. In particular Al and Si-rich lithologies show the most significant effects. Besides silicate mineral formation Si introduction may cause directly and indirectly variations of the REE mineralization at different stages of the carbonatite emplacement. While Si introduction during apatite formation causes an increased REE incorporation into apatite due to the britholite substitution accompanied by an early consumption of REE from the melt, an REE enrichment in the melt and related specific REE mineral formation in late magmatic stages become inhibited. A Ti-rich carbonatite magma additionally experiences the formation of titanite at the expense of ilmenite. Although REE consumption by titanite is less important as for apatite, specific REE consumption can influences REE patterns of subsequent mineralizations. On the other hand, magma wall rock interactions in a carbonatitic systems may furthermore directly influence the type of REE mineralization reflected by discrete REE minerals. In this way contamination can directly control the formation of either allanite, britholite, chevkenite or monazite and hence influences the economic processibility of a REE deposit. Furthermore, the stability of HFSE minerals such as baddeleyite or thorianite can be suppressed by the predominance of their Si-bearing counterparts (e.g., zircon and thorite).