The influence of MORB and harzburgite composition on thermo-chemical mantle convection in a 3-D spherical shell with self-consistently calculated mineral physics

Abstract

Three-dimensional thermo-chemical mantle convection simulations with mineral assemblages self-consistently calculated using free energy minimization are used to check the sensitivity of model behavior to the assumed compositions of mid-ocean ridge basalt (MORB) and harzburgite. In addition to five-oxide CaO-FeO-MgO-Al2O3-SiO2(CFMAS) compositions, we test the effect of a more realistic compositional model by adding a sixth oxide Na2O (NCFMAS) with three compositions. Results indicate that thermo-chemical structures are quite sensitive to variations in MORB composition of the order 1-2% oxide fraction, particularly FeO and Al2O3. Differences occur in (i) the amount of compositional stratification around 660km depth caused by the inversion of the MORB-harzburgite density difference between 660 and 740km depth, which is different in magnitude and depth extent between the different tested compositions, and (ii) in the degree of MORB segregation above the CMB, which is related to differences in the MORB-harzburgite density difference in the deep mantle. While improving the realism of the model by including Na2O tends to reduce the MORB-harzburgite density difference at most pressure and temperature conditions, the differences in behavior among three NCFMAS compositions are at least as large as between either NCFMAS and the CFMAS composition, and are also related to differences in the (pressure and temperature) stability range of the post-perovskite phase between the different compositions. Comparing model spectra to those of seismic tomography using spectral heterogeneity maps, NCFMAS compositions provide a better match to seismic tomography but in all cases there is too much heterogeneity at mid lower mantle depths compared to typical seismic tomographic models, which implies that less CMB basalt segregation occurs in Earth than in the convection models. © 2010 Elsevier B.V.