Thesis Defense: Generation of Potentials for the Self-Assembly of Particles

Katalin Grubits

We present a fast method for generating robust inter-particle potentials that lead to the formation of high quality lattice configurations under cooling. These isotropic pairwise potentials are sufficient for particles in the plane to self-assemble into ordered patterns, such as a honeycomb lattice. The two key developments that are responsible for the success of our method are (1) quality metrics that quantitatively assess the quality of lattices, and (2) a trend optimization scheme that can quickly and robustly find the optimal parameters for the self-assembly potential. Trend optimization is an efficient method for optimizing objective  functions that are noisy, highly variable, and expensive to evaluate, yet have a simple trend.