Is a pyrolitic adiabatic mantle compatible with seismic data?

In this paper, the simplest average physical model of a mantle convecting as a whole (i.e., following an adiabatic temperature gradient) with a single composition (pyrolite with phase transitions) is tested directly against global seismic data, instead of against spherically symmetric seismic models. Constraints from seismic data on average velocities and lower mantle velocity gradients are hard to reconcile with an adiabatic pyrolitic mantle, given the current state of knowledge of elastic and anelastic mineral parameters at high pressure and temperature. This physical model generally gives (a) a stronger baseline offset between upper and lower mantle average travel-time residuals than allowed by the data and (b) an insufficient decrease in velocity with depth in the lower mantle (above 2500 km). We tested 105 upper and 105 lower mantle models that were selected randomly within the mineral parameter uncertainties. Only 2 lower mantle models and 24 upper mantle models yield whole mantle seismic structures that are compatible with global ISC P and S travel times and central frequencies of toroidal and spheroidal fundamental modes with angular order higher than 18. To improve the fit to the seismic data, the physical model would require (a) a lower velocity transition zone composition than dry pyrolite (at least around continents and subduction zones) as well as (b) a gradual change in physical state of the lower mantle that decreases the velocity-depth gradient, e.g., a superadiabatic temperature gradient.