A model for the upper mantle

The occurrence of a low-velocity zone in the upper mantle has been attributed to the effect of high temperature gradients. If the upper mantle is homogeneous, it is shown that the required temperature gradients would lead to extensive melting. Since this consequence is unacceptable, it appears necessary to consider nonhomogeneous models. It is assumed that the mantle immediately below the Mohorovicic discontinuity consists dominantly of dunite and peridotite. This passes downward into more primitive material, which is chemically equivalent to a mixture of 1 part of basalt to 4 parts of dunite. For convenience, this hypothetical primitive material is called ‘pyrolite.’ Within the upper mantle, the pyrolite might occur in either of two principal mineral facies—(i) as the assemblage olivine-pyroxene-plagioclase (plagioclase pyrolite) and (ii) as the assemblage olivine-pyroxene-garnet (garnet pyrolite). There would also be a substantial transition zone between these facies, in which olivine, pyroxene, plagioclase, and garnet coexisted. It is suggested that the low-velocity zone is caused by downward transition from the sub-Mohorovicic peridotite into plagioclase pyrolite (which has a relatively low seismic velocity) and then into garnet pyrolite. The low-velocity zone is thus due to the presence of plagioclase as a primary phase. Further aspects of this model are explored. The depths of these various zones probably differ between oceanic, continental margin, and Precambrian shield regions. The model implies that a low-velocity zone may not be present beneath Precambrian shields. In regions which have been recently subjected to diastrophism, the characteristic geotherm may pass for a considerable distance through the transition zone between plagioclase and garnet pyrolite facies. This can give rise to acute thermal and mechanical instability, which may in turn cause diastrophism in the crust. Evolutionary relationships between oceanic, active orogenic, and Precambrian shield regions are also discussed in terms of the model.