In summary, in recent years it has become established that the transition zone, particularly the endothermic phase transition at 660 km depth, is expected to play a major role in mantle dynamics, resulting in a style of convection that lies inbetween whole-mantle and layered convection. This intermittently layered flow may provide the key to reconciling seemingly contradictory geophysical evidences, such as slabs penetrating in some areas but not in others, geochemical and mineralogical observations and constraints, and good correlations between the integrated historical locations of slab subduction and seismic tomography of the transition zone and deep mantle tomography [ Scrivner and Anderson, 1992; Ray and Anderson, 1994], suggesting that slab material is temporarily held up in the transition zone before being rapidly transported to the deep mantle.
The studies performed can be regarded as refining the basic dynamical picture established by Christensen and Yuen [1984,1985]. The effects of three-dimensionality, variable material properties associated with mantle compressibility, complex phase diagram, plates and aspect ratio have all been investigated. Another theme has been to try to relate the results to observations, although the results of such comparisons currently appear somewhat contradictory.
Most of the studies reported assume an isochemical mantle with only constant, or depth-dependent viscosity. The challenge for the future is firstly, to develop models which incorporate realistic viscosity variation, including plates and subduction as well as realistic deep mantle rheology, and secondly, to more fully consider the constraints arising from geochemical observations, and explore the dynamics of thermo-chemical convection in the presence of phase changes, particularly in the context of the Earth's evolution. Naturally, both of the directions present severe computational challenges.
Acknowledgments. The author thanks David Yuen and Scott King for thoughtful reviews, and David Yuen for supplying Fig. 2. Contribution number 5434, Division of Geological and Planetary Sciences, California Institute of Technology.