A peridotite source for strongly alkalic ultrabasic HIMU lavas of the Oslo Rift, Norway

High alkali (Na2O + K2O) low SiO2 lavas generally have highly fractionated trace element compositions and their petrogenesis is debated. Both pyroxene-rich (pyroxenite) and olivine-rich (peridotite) mantle sources have been proposed for such lavas. HIMU (high U-238/Pb-204) lavas usually have such compositions and are often interpreted as being derived from garnet-pyroxenite (eclogite) residue of subducted recycled oceanic crust, possibly including a carbonate component. In some cases, an origin from a garnet peridotite source has also been considered. We present new Mg, Pb, Nd, and Sr isotopic data and elemental data for alkalic ultrabasic lavas (melilitites, nephelinites) from the Oslo Rift, Norway. The magmatic province of the Permo-Carboniferous Oslo Rift is interpreted as part of the much larger Skagerrak-Centered Large Igneous Province (SCLIP) which may be related to a mantle plume source originating from the African Large Low Shear Velocity Province (LLSVP) at the coremantle boundary. The range of delta Mg-26 is narrow (-0.32 to -0.28) and is similar to typical mantle values, and therefore is inconsistent with a carbonated source as has been suggested for other similar lavas. The Pb-206/Pb-204 values of all the lavas are high, which identifies them as HIMU lavas. The epsilon(Nd) of +1 to +2 are substantially lower than for typical HIMU lavas, and in the epsilon(Nd)-epsilon(Sr) diagram they plot on the LoNd array between HIMU and EMI. The lavas are more like archetypal kimberlites (formerly, Group I kimberlites), as these also have eNd slightly higher than the chondritic value and Pb isotopes consistent with a HIMU source. The REE patterns of the lavas are also almost as fractionated as in Group I kimberlites. We used major and trace elements to infer the primary magma composition of these lavas. This primary magma composition is compared with experimental melts of both carbonated peridotite and carbonated eclogite source compositions and shows that their major element compositions are only consistent with either a very high degree of melting of eclogite or a very low degree melting of garnet peridotite. The strongly fractionated REE patterns of the lavas can only be made by low-degree melting of either eclogite or garnet peridotite, which therefore effectively rules out eclogite as a major component of the mantle source of these lavas. Further trace element modeling points to a mantle source composition with >97% garnet peridotite that was lightly carbonated (<0.5 wt% CO2) and melted to a very low degree (similar to 0.4%) to produce these lavas. The modeled source trace element pattern was constrained by the Nd, Sr, and Pb isotopic compositions of the lavas. The measured high Ti contents of these lavas (3.8-6 wt%) are quantitatively shown to be the result of enrichment by both partial melting and fractional crystallization, using experimentally determined DTi for melting of this type of source. We also show that a low degree of melting of a slightly carbonated source is sufficient to result in the likely very high CO2 content (10-15 wt%) of the primary magma. The rather uniform Pb isotope composition of HIMU magmas suggests that the HIMU source is rather uniform in composition, but the Pb isotope composition requires this source to form late in Earth's history. The similarity of these lavas and kimberlites suggests a very deep source, possibly in the LLSVP that gave origin to the SCLIP. This means that LLSVPs do not represent some primordial layer around the core that escaped homogenization of the mantle by the Moon-forming giant impact. Instead, this layer must have formed late in Earth's history. The constraint that this source contains <3% of garnet-pyroxenite (eclogite) residue of recycled basaltic oceanic crust shows that this layer is primarily not a subducted slab graveyard. The process by which the LLSVPs formed, however, remains obscure.