Self-assembly systems, based on polymers, nanoparticles, colloidal sized particles or other derivatives, are an appealing way to create materials with high complexity and detail. These systems are highly sensitive to changes in the interactions between the system's components, leading to different emergent structures. In material science, targeted design of new colloidal systems requires the tuning of interactions such that the desired target structure or macrobehaviour is emergent, and to obtain a high-yielding self-assembly path. Changes to the interactions can be represented by a parameter space. Time-dependent simulations (or experiments) are necessary to find emergent systems in parameter space, and the complexity of them demands resources and time. The conventional systematic (brute-force) scan of parameter space is highly inefficient, with most resources spent evaluating low-yielding regions. Typically, to counter noisy measurements to accurately identify favourable parameter values, an average is taken across multiple simulations. We examine hill-climbing as a potential alternative for tuning parameters values to obtain high-yielding systems. As an example tuning problem, a two-dimensional shorted-ranged attractive patchy hard disk model, important for coarse-grain modelling of polymers and biological systems, and a yield measure for quantifying our target structure (large round compact honeycomb clusters) are introduced. Varying the interaction strength and patch width, a noisy landscape is constructed from Monte Carlo simulation yield data. We show that a hill-climbing search on this landscape can locate the localised region of high-yielding assembly, and suggest situations where this is advantageous over brute-force scan.
|Date of Award||18 Feb 2019|
|Supervisor||Robert Jack (Supervisor) & Nigel Wilding (Supervisor)|