Abstract

Temperature-Accelerated Dynamics Simulations of Thin-Film Growth

Jacques Amar
University of Toledo
Department of Physics and Astronomy

While kinetic Monte Carlo (KMC) simulations can be an effective tool in simulating non-equilibrium processes when the relevant activated processes are known, in many systems including on-lattice as well as disordered and/or off-lattice systems, the key processes are either unknown or are difficult to guess a priori. As a result, a variety of other methods such as temperature-accelerated dynamics (TAD) and adaptive KMC have been developed. Here I discuss recent progress we have made in using TAD to carry out simulations of thin-film growth. After a brief review of the TAD method, I will first discuss the development of a parallel TAD (parTAD) method which may be used to carry out TAD simulations over extended length-scales. Some additional refinements such as adaptive TAD and the use of localized saddle-points to speed up the calculation of activation barriers will also be discussed. I will then describe in more detail how we have used TAD and parTAD simulations - combined in some cases with KMC simulations – to discover new and interesting relaxation processes which have allowed us to explain a variety of interesting phenomena in thin-film growth including:
(i) vacancy formation and strain in low-temperature multilayer Cu/Cu(100) growth
(ii) the non-monotonic temperature-dependence of the surface roughness over a wide range of temperatures in Ag/Ag(100) growth
(iii) the shape transition from compact to ramified islands observed experimentally in submonolayer Cu/Ni(100) growth
In general, we find that thin-film growth is a surprisingly complex process which typically involves a competition between a variety of different and sometimes previously unexpected, atomic relaxation mechanisms. In addition, other effects such as the short-range attraction of depositing atoms to the substrate and shadowing can also play an important role. The last case also illustrates the complexity of carrying out simulations involving a wide range of activation barriers and time-scales as well as long-range elastic interactions. A variety of different approaches to dealing with these issues will also be discussed.