| Thin-films are a broad and important class of materials. Numerous examples of thin-films exist across a wide range of problem domains including bi-metallic catalysts [40, 38], polymer films for catalysis [22], ferroelectric films for piezoelectric sensors and actuators [3], and magnetic films for medical imaging devices [49] and high-density data storage [20, 80].;Thin-films often exhibit some degree of self-assembly, a process that involves a transition from a low-ordered state to a high-ordered state without external stimuli. In thin-films, self-assembly is driven by the interactions between the adsorbate molecules that compose the film. In particular, short-range attractive forces and long-range repulsive forces (e.g., strain induced through a lattice mismatch with the substrate) are necessary to give rise to self-assembly. By manipulating the balance between the intermolecular forces, using process conditions or other external factors, it will be shown that one can drive arbitrary initial patterns toward a target pattern.;A number of challenges must be overcome before one can control thin-film surface morphologies using self-assembly. Specifically, self-assembly of thin-films is an inherently multiscale problem. This means that the actual process of self-assembly occurs at microscopic length and time scales (i.e., nanometers and nanoseconds) while changes to the manipulated variables occur at macroscopic length scales and time scales (i.e., centimeters and seconds). Additionally, self-assembly processes have few manipulated variables while the thin-film surface morphologies are complex and have a large number of degrees of freedom.;The overall aim of this dissertation is to develop the tools and methods necessary to achieve fine-grained control over self-assembled thin-films. In particular, we will develop three classes of tools to perform systems-level analysis of processes involving self-assembly. These tools include sensitivity analysis of stochastic simulations, algorithms for the optimization and design of computationally expensive, stochastic simulations, and methods to calculate reduced representations of two-dimensional surfaces to facilitate control. We will also show the effectiveness of these tools through a case study where we take arbitrary initial surfaces and drive them, using a simple controller, toward an idealized target surface. |
