In this thesis I advance the integration of mineral thermodynamics into convection modeling.

I have compiled a thermodynamic model of mantle mineralogy in the five component CFMAS system (CaO-FeO-MgO-Al2O3-SiO2), including mineral phases that occur close to typical chemical models of the mantle and reasonable mantle temperatures. In this system I have performed a system Gibbs free energy minimization, including pure end-member phases and a non-ideal formulation for solid solutions. Solid solutions were subdivided into discrete pseudocompounds and treated as stoichiometric phases during computation of chemical equilibrium by the simplex method. I have complemented the thermodynamic model with a model of shear wave properties [Stixrude and Lithgow-Bertelloni, 2005] to obtain a full description of aggregate elastic properties (density, bulk and shear moduli) that provide a useful basis for the consideration of seismic and geodynamic models of the Earth's mantle.

By using this new thermodynamic database for the mantle I have coupled the resulting density dynamically

(through the buoyancy term) with mantle convection models.

I have linked the

database with a high-resolution 2-D convection code (2DTERRA), dynamically coupling the thermodynamic model (density) with the conservation equations of mantle flow.

The coupled model is run for different parameterisations of viscosity, initial

temperature conditions, and varying internal vs. external heating.

A common feature of all the models is that the convecting flow creates a characteristic discontinuity of temperature around 660 km depth in order to compensate for the entropy change due to the phase transitions.

I have studied the importance and the possible consequences of such a thermal regime on the excess temperature of plumes and on the transition zone thickness.

The thermodynamic mantle mineralogy model provides the conversion of the temperature field into

seismic velocities so that predictions from mantle convection can be compared to seismic observations in terms of radial profiles or lateral variations.

This approach allows us to predict a number of seismic

observables from the convection model, all of which agree remarkably

well with observations from seismic tomography.