Hydrogen zoning in olivine from kimberlites based on coupled FTIR and SIMS analyses: Significance for H2O distribution in the lithospheric mantle and H2O concentrations in kimberlite melts

Kimberlite melts are widely considered to be enriched in volatiles, both CO2 and H2O. Yet, estimated H2O concentrations in primitive kimberlites vary between 3.0 and 12 wt.%, and it is unclear whether these variations are a true reflection of variable H2O in their mantle sources. The main problem rests with the origin of serpentine, the main H2O host in kimberlite rocks, and specifically, whether it derives from magmatic and/or crustal fluids. To obtain estimates of primary H2O contents in kimberlite melts, we have examined the systematics of proton incorporation in olivine point defects (referred as H) in grains from eight representative kimberlites from Africa, Canada and Greenland. These kimberlites show highly variable groundmass and therefore melt compositions, and include varieties enriched in serpentine, carbonate and/or phlogopite. Olivine grains are strongly zoned in major and minor elements based on BSE images and EPMA analysis and include mantle-derived xenocrystic cores and magmatic rims. FTIR maps and profiles show that the olivine cores are zoned with H2O decreasing outward due to diffusive loss most likely triggered by decompression and related H2O loss from the transporting kimberlite melt after xenocryst entrainment. The central portions of the cores exhibit homogeneous H2O contents representative of mantle values. A combination of Al-in-olivine thermometry with appropriate geothermal gradients and H2O determinations from SIMS analyses (from 9 to 241 μg/g across the entire sample set) shows similar systematics to those of olivine in mantle xenoliths from the same localities. H2O analyses of olivine cores in kimberlites are hence valuable to systematically examine the vertical distribution of H2O in the lithospheric mantle traversed by kimberlites. The magmatic rims examined by SIMS invariably show low H2O contents (<50 μg/g but mostly ≤20 μg/g) with very limited within-sample variation regardless of olivine major-minor element chemistry and groundmass composition. Experimentally derived hydrogen partition coefficients yield ≤1 wt% of H2O in the corresponding melts, values that probably reflect post-emplacement re-equilibration in the upper crust rather than equilibrium conditions during crystallisation. These low H2O contents reflect, at least in part, exsolution of C-O-H fluids during kimberlite ascent. The implication is that kimberlite melts contain insufficient H2O to crystallise the large amount (up to 50 vol%) of serpentine commonly observed in kimberlites. Serpentine rather requires contributions by crustal fluids and/or deeply exsolved kimberlite-related fluids that percolate upwards along the pipe- or dike-like emplacement structures, modifying the composition of previously crystallised kimberlites.