<p>Terrestrial planets evolve through various stages of large-scale melting, or magma oceans, due to the energy release during accretion and differentiation. Any magma ocean is thought to become progressively enriched in FeO and incompatible elements upon freezing due to fractional crystallization. The resulting upwards enrichment of the related cumulate (=crystal) packages drives gravitational overturn(s) of the incipient mantle, and ultimately stabilizes a FeO-enriched molten layer at the core-mantle-boundary (CMB)<sup>1</sup>. Such a molten layer, previously termed basal magma ocean (BMO)<sup>2</sup>, is thought to also fractionally crystallize, but downwards instead of upwards, and over much longer timescales than the surficial magma ocean. This BMO fractional crystallization due to slow planetary cooling analogously implies the stabilization of a thick FeO-enriched layer at the CMB. Such a layer would essentially remain stable forever, as being too dense to be entrained by convection of the overlying mantle. However, at least for Earth, geophysical observations rule out the preservation of such a deep dense global layer. Here, we investigate the consequences of an alternative mechanism for BMO freezing, reactive crystallization, on the initial condition of solid-state mantle convection and long-term planetary evolution.</p><p>Based on scaling relationships, we show that any cumulates, which crystallize from the BMO (e.g., due to initial cooling or reaction) are readily entrained by mantle convection. Once the BMO-mantle boundary is exposed, the BMO reacts with the mantle to form reactive cumulates. Reaction is driven by disequilibrium between mantle rocks and the BMO, a situation that is inevitable independent of BMO initial composition. As reactive cumulates are continuously entrained by mantle convection, the BMO continues to freeze by reactive crystallization. Based on lower-mantle mineral-melt phase equilibria<sup>3</sup>, we calculate the compositional evolution of the BMO, and the chemistry of the BMO cumulate package. We demonstrate that for a wide range of BMO initial compositions, the cumulate package consists of two discrete layers: the first is pure bridgmanite close to the MgSiO<sub>3</sub> end-member; the second is mostly bridgmanite+ferropericlase that is moderately enriched in FeO and incompatibles, i.e. similar in composition to FeO-enriched pyrolite. The mass or thickness of the cumulate package depends on reaction kinetics, but is significantly larger than that of the BMO. The bridgmanitic layer is expected to be entrained by mantle convection due to its intrinsic buoyancy, but resist efficient mixing due to its intrinsic strength, thereby potentially providing an explanation for seismic scatterers/reflectors and ancient geochemical reservoirs<sup>4</sup>. The moderately FeO-enriched layer is expected to stabilize thermochemical piles, providing a candidate origin for the seismically-observed large low shear velocity provinces (LLSVPs)<sup>5</sup>.</p><p>These results have implications for the long-term (thermal) evolution of planets in general. Earth-sized terrestrial (exo-)planets and super-Earths should also initially host a MgSiO<sub>3</sub>-rich layer as well as a moderately FeO-enriched layer. In contrast, small terrestrial planets such as Mars may host a more strongly Fe-rich deep dense global layer as long as no BMO is stabilized in their histories.</p><p>[1] Ballmer+, G-cubed 2017; [2] Labrosse+, Nature 2007; [3] Boukar&#233;+, JGR Solid-Earth 2015; [4] Ballmer+, Nat.Geosci. 2017; [5] Ballmer+, G-cubed 2016.</p><p>&#160;</p>
Formation of anorthosite on the Moon through magma ocean fractional crystallization
