We study the changes in rheological parameters and effective viscosity across mantle polymorphic phase transitions and the variations of these quantities with temperature and pressure throughout the earth's mantle. The intrinsic activation energy for oxygen ion diffusion in oxides E0* is shown to be systematically related to oxygen ion packing by E0*(kcal/mol) = (187±16) ‐ (3.8±0.8) Vo=(Å3), where Vo= is the volume per oxygen ion at zero pressure and 25°C. This relation allows the change in activation energy δE0* across a polymorphic phase transition to be estimated from the associated change in density. Under the assumptions that the activation volume V* remains constant or decreases across a phase transition and that the activation energy for subsolidus creep is equal to the activation energy for O= diffusion, we use δE0* in a general non‐Newtonian flow law to estimate the increase in effective viscosity η and the decrease in activation volume V* across a phase transition. By modeling V* and the activation energy by using thermodynamical and mechanical relations for elastic continua and data from seismically derived earth models for relevant elastic parameters, we are able to estimate better the increases in viscosity across polymorphic phase transitions and throughout an adiabatic mantle. Our best models of mantle viscosity have (1) η increasing by less than an order of magnitude across any phase transition, (2) η essentially constant throughout the lower mantle, and (3) η in the lower mantle no more than 2 orders of magnitude greater than η in the upper mantle.

We have compared estimates of the activation volume V* based on several theoretical models with measured values in metals, alkali halides, and olivine. The theoretical methods tested include one based upon the empirical correlation between activation energy and melting temperature and several which are based upon simple elastic models for the defect structure. For metals and olivine, the melting relation works well, but for alkali halides, the melting model predicts too large a V* by an approximate factor of 1.5. Of the elastic models, the dilatational strain energy model introduced by Zener (1942) provides reasonable estimates of V* for metals and olivine, but it also overestimates V* for alkali halides. Zener's assertion that the experimental value of V* should be bounded by theoretical values calculated from strain energy models which assume pure shear (Keyes, 1963) and pure dilatation is supported by the available data for metals, oxides, and alkali halides. These provide upper and lower estimates for the variation of viscosity with depth in the mantle.