Thin films are nanoscale layers of material used to functionalize surfaces or to serve as building blocks in more complex devices. In recent years, thin metal films have become vital for modern devices within, e.g., biosensing, catalysis, and nanoelectronics, whereby synthesis of metal layers with specific morphological features on two-dimensional (2D) crystals and oxides is required. However, this entails a great scientific challenge: in most of the afore-mentioned film/substrate combinations substrate and metal atoms interact weakly, causing the latter to self-assemble without control into three-dimensional (3D) clusters.

Nowadays, a significant fraction of thin films is synthesized via condensation from the vapor phase, a far-from-equilibrium process in which film morphology is governed by the kinetic rates of atomic-scale structure-forming processes. It is, therefore, evident that knowledge-based synthesis of metal layers in high-performance devices necessitates a comprehensive understanding of the dynamic competition among these processes at the nano- and mesoscale. Such understanding is today incomplete, since experimental materials science tools are often not capable of providing nanometer and sub-nanometer insights at time scales that are relevant for thin-film synthesis. Computational approaches offer the possibility to fill the afore-mentioned gap in knowledge by allowing to explore atomistic behaviors with picosecond resolution. Hence, in the present thesis, a combination of modern computer simulation techniques is used to investigate thin metal film growth on weakly-interacting substrates from a purely atomistic point of view and to elucidate the ways by which atomic diffusion mechanisms give rise to the final film morphologies.

In the first part of the thesis, an in-house kinetic Monte Carlo (KMC) simulation code and analytical modelling are used to investigate the early growth stages of Ag films supported on a generic weakly-interacting substrate. The results show that the weak interaction strength between film atoms and substrates leads to the formation of strongly-faceted 3D Ag islands, whose vertical growth is mediated by the temperature-dependent upward adatom diffusion across the facets. Eventually, the 3D islands impinge on each other and coalesce via surface migration of facet layers. Migration can be promoted by an increase of the deposition flux, but it can also be hindered by material agglomeration if the flux exceeds a critical threshold. These findings provide the foundation for explaining several effects observed during thin film growth on weakly-interacting substrates, including the increase of film roughness with temperature, the transition from 3D to 2D film morphology upon suppression of coalescence, and the origin of changes in thin film roughness and grain boundary number densities when varying the magnitude of vapor flux arrival rate.

In the second part, ab initio and classical Molecular Dynamics simulations are used to investigate the diffusion dynamics of several transition metal adatoms (Ag, Au, Cu, Pd, Pt and Ru) and multi-atomic clusters (Ag, Au, Cu and Pd) on single layer graphene at room temperature (300K). The simulated diffusion trajectories reveal that diffusing adspecies experiencing a deep (hundreds of meV) potential energy landscape (PEL) on the substrate surface follow random walks; whilst those with a weak interaction with the substrate (PEL depth of a few meV) follow a superdiffusive motion pattern known as Lévy walk. This type of anomalous movement— also observed in other phenomena in physical, biological, and social systems—manifests itself as a continuous atomic motion with occasional flights over distances covering multiple adsorption sites. The fact that adspecies follow a distinctly different type of motion than what is observed in classical homoepitaxial growth theory implies that energy barriers readily available from static (0K) calculations may not be able to provide a physical accurate description of surface diffusion of metal adspecies on 2D crystals. As such, anomalous diffusion is a potentially important aspect to be considered when modelling growth of metal films and nanostructures on 2D materials.

The results and insights generated in the present thesis provide key knowledge for controlled synthesis of films and nanostructures with tailored properties. This, in turn, is relevant for developing high-performance energy-saving windows, improving the turnover frequency of catalytic reactions, and integrating 2D materials into novel nanoelectronic devices. Moreover, the techniques developed and employed herein contribute toward bringing modern computational tools closer to the field of thin film growth.