TiO2 is a well investigated material due to its vast array of applications, from the most common white food colouring, to usage as an optical coating due to its high refractive index. When deposited as a thin film for use in solar control coatings using a magnetron sputtering device, TiO2 is found to form in rutile, anatase and amorphous phases, with the rutile (1 1 0) and anatase (1 0 0) crystal surfaces occurring most frequently. While the influence of deposition rate, substrate bias voltage, oxygen partial pressure and temperature have all been investigated experimentally, there are fewer simulated results. This seems to be primarily due to the lack of a suitable empirical potential through which dynamics can be accurately modelled. In this research project, the short-comings of the leading fixed and variable charge TiO2 potentials were revealed through comparison of binding energies and transition barriers for simple ad-clusters and interstitials to DFT. This led to the development of an improved variable charge potential, particularly when modelling the rutile (1 1 0) surface. It has been experimentally demonstrated that rutile growth results in a reduced surface due to the formation of large numbers of Ti interstitials and this was also suggested due to low Ti interstitial formation barriers found using DFT and the variable charge empirical potentials. The high interstitial formation probability was confirmed when performing large numbers of Tix Oy depositions at typical industrial energies. Correlations between impact site and numbers of interstitials and vacancies formed were found along with insight into typical penetration depths and common defect structures. Multilayer growth was successfully modelled using classical MD by accelerating the deposition rate. The large numbers of formed interstitials coupled with a high escape barrier resulted in defective growth and so it was necessary to increase the substrate temperature such that atoms were mobile on the computational time-scale. Multilayer growth was investigated as a function of deposition energy, cluster composition and stoichiometry on the rutile (1 1 0), anatase (1 0 0) and amorphous surfaces. The optimum conditions for forming defect free rutile were found along with more limited insight into the ideal growth conditions when depositing on anatase and amorphous substrates. Finally a long time-scale dynamics technique, the ‘on-fly-kinetic Monte Carlo' method, was developed, and using an efficient bespoke transition search algorithm, rutile TiO2 growth was successfully modelled at 350 K, with some caveats to avoid becoming entrenched in low activation diffusion processes. The final conclusion being that low energy depositions of large clusters with an oxygen excess would produce optimum film growth.