The origin of carbonatite magmas predating main-phase LIPs eruptions

Carbonatites are spatially and temporally associated with Large Igneous Provinces (LIPs) such as the Siberian traps, the Paraná-Etendeka and the Deccan traps. Carbonatites, and the associated alkaline rocks, can both predate and postdate the main tholeiitic volcanism. For example, the Sarnu Dandali complex (68.57±0.08 Ma) and the Mundwara complex (68.53±0.16 Ma), both characterized by high 3He/4He, predate the Deccan traps, whereas the 65±0.3 Ma carbonatites in the Narmada Rift postdate it. Similarly, carbonatites from the Amambay alkaline province (Eastern Paraguay) predate the Paraná-Etendeka LIP by several million of years, whereas the Jacupiranga carbonatites (130 Ma) in South America and the Damaraland carbonatites (129-123 Ma) in Namibia postdate the main tholeiitic pulse (134-132 Ma).The origin of carbonatites remains a matter of debate, albeit radiogenic isotope ratios, trace element variations and primordial noble gases from most carbonatites support a plume origin. For carbonatites predating LIPs, a generally accepted model invokes partial melting of carbonate-metasomatized lithospheric mantle, heated by the plume. The implicit assumption is that heat, slowly diffused from the plume, can reach the lithosphere before buoyant melts from the plume itself, which is not obviously plausible.Our 3D-numerical simulations of a mantle plume with millions of carbon- carrying tracers enable us to calculate the depth at which carbon-rich fluids form. These fluids, because of their physical properties, are highly mobile and separate from the solid matrix even at low melt fractions. At each time-step we calculate their ascent velocity (i.e., a linear combination of the solid matrix velocity and of the separation velocity) and their 3D-trajectories. We span a range of carbon concentrations in the plume source (196 ≤ C ≤ 440 ppm), and we explore different depths of redox melting and P-T conditions for the solidus of carbonated peridotite.We find that, if mantle redox conditions allow for deep (>200 km) carbon-rich melting, then the fast rising carbonatitic fluids can reach the lithosphere 2-3 Myr before the onset of anhydrous peridotite melting. This key result reveals the existence of a precursory carbon flux (of order 10e+12 - 10e+13 mol/yr) across the base of the lithosphere (i.e., 140 km depth). When melting of anhydrous peridotite starts in the plume head, a total mass of 10e+16 kg C has already reached the lithosphere. This precursory carbon flux provides a new framework to interpret carbonatite complexes predating the earliest LIP's volcanism.We also find that the radial extent of the zone permeated by carbon-rich fluids is much broader than the zone undergoing anhydrous peridotite melting. These vast lithospheric domains, fertilized during several millions of years by plume-derived carbon-rich fluids might be mobilized by peripheral tholeiitic magmas. Possibly, this scenario could explain the occurrence of carbonatites that postdate LIP's emplacement, but which carry a distinctive plume-like geochemical fingerprint (e.g., the high 129Xe/130Xe of the Jacupiranga carbonatites).