At present, there are two controversial, but exciting results which could play a major factor in the coming years. The first is the high creep strength of garnet [cf., Meade and Jeanloz, 1990; Karato and Li, 1992; Karato et al. , 1994]. If the transition zone is relatively enriched in garnet, it could be a layer of increased viscosity. Such models have been studied [ Hong et al. , 1990; Spada et al. , 1992]; however, most models of mantle viscosity have a low viscosity transition zone. The study by King [1994b] shows that a high viscosity transition zone can not be ruled out by the geoid alone, showing that both high viscosity and low viscosity models fit the data equally well.
If the garnet phase controls the strength of the transition zone, this would be consistent with the high viscosity transition zone models. On the other hand, experiments on wadsleyite show that OH is 40 times less soluble in wadsleyite than olivine [ Young et al., 1993]. This suggests that if any hydrous olivine in subducted slabs survives to the depth of the transition zone, the transformation to wadsleyite could release water into the transition zone. The effect of water on rheology is usually to decrease the viscosity, providing a possible physical explanation for a weak, or low viscosity, transition zone. If both experimental results are relevant to the transition zone, it is not clear which effect dominates, the strength of the garnet or the weakening effect of water because, the transition zone contains both an olivine-spinel component and a pyroxene-garnet component.
In addition to compositional control on rheology, the strength of the transition zone may be significantly weakened as a result of recrystalization occurring during a solid-solid phase change. This would provide a possible explanation for a weak, or low viscosity, transition zone. Yet another explanation for a weak transition zone (or lower part of the transition zone) was given by Forte and co-workers [ Forte et al., 1993]; if the 660 km phase change is a strong barrier to mantle flow, there will be a thermal boundary layer at 660 km. Because the viscosity of mantle silicates is a strong function of temperature, this thermal boundary layer should produce a low viscosity layer. Further exploration of the implications of a change in the viscosity of the transition zone would seem to be a fruitful area of further research.
The other important result is the high melting temperature of perovskite [ Zerr and Boehler, 1993]. It is generally accepted that the creep strength of a mineral is inversely related to its melting temperature. While there is a difference of opinion concerning the approach to this difficult laboratory experiment [ Heinz et al. , 1994; Boehler and Zerr, 1994], the measured very high melting temperature of perovskite could prove to be the most important constraint on lower mantle viscosity because, many of the estimates from surface observations only constrain the average viscosity over a broad region of the lower mantle. The high melting temperature for perovskite provides a physical explanation for a high viscosity lower mantle. Some, but not all, mantle viscosity models suggest the viscosity of the lower mantle is greater than that of the upper mantle. Thermal convection calculations with a rheological law simulating the high perovskite melting temperature can produce vertically averaged viscosity profiles, similar to the models proposed by geoid and plate velocity studies [ van Keken et al. , 1994].
Acknowledgments. This research was supported by a grant from the National Science Foundation. Helpful comments from Roger Pielke, Kenneth Verosub, and Louise Kellogg improved the final version of this report. I thank a number of the authors who graciously provided pre-prints of their results in advance of publication.