This dissertation details a predictive computational approach for modeling titanium dioxide

nanoparticles in flame reactors. The industrial production of these nanoparticles is done using

the chloride process, i.e. titanium tetrachloride (TiCl₄) is oxidized in a flame to form titanium dioxide (TiO₂) particles:

TiCl₄(g) + O₂(g) &rarr TiO₂(s) + 2Cl₂(g).

In absence of thermochemical data most previous works used the one-step reaction mechanism given above. But this problem was alleviated recently by West et al. (2009) [R.H. West, R.A. Shirley, M. Kraft, C.F. Goldsmith, W.H. Green, Combust. Flame 156(2009) 1764-1770], by proposing a detailed mechanism for this oxidation process, which includes 30 species and 66 reactions. As the oxidation of TiCl₄ happens in a flame, this detailed mechanism becomes more complex with interactions of the hydrocarbons with oxidizer as well as chlorine. Hence, the proposed detailed mechanism in this work extends to 107 species and 501 reactions. Comparisons are made between the one-step and detailed mechanism to show that different models would result in very different product properties.

A bivariate population balance model was proposed to evaluate the size distribution of nanoparticles in the flame reactor. This model tracks both the area and volume distributions and accounts for nucleation, surface growth, aggregation and sintering of the nanoparticles. The results from this model are used to evaluate the particle size and shape for the two chemical mechanisms, which in turn are compared to experimental results. Also explored are the roles of gas-phase and surface phase reactions.

Accurate models for the nanoparticles involve developing a detailed chemical mechanism and modeling the transport process. This is especially true in the case of flame reactors where the flow structure and turbulence are of major importance. Computational fluid dynamics based techniques can be used to understand and implement this coupling between transport processes and chemical reactions. But due to the large number of species and reactions involved, coupling this detailed chemistry with flow solvers is computationally very expensive. Thus, to represent the correct chemistry while making the problem computationally viable reduction of the detailed mechanism is carried out.

Finally, discussed are the results from the successful application of the models and techniques refined during the dissertation work to an industrial system. The findings show that the developed models can accurately track particle evolution in an industrial reactor. In summary, this work uses detailed chemistry and bivariate distribution to present a predictive multiphysics computational model for TiO₂ nanoparticle synthesis in flame reactors, that can be employed to optimize operating conditions based on desired product particle size distribution.