Computational molecular design and optimization increasingly plays a critical role in the design of novel materials. Part of the desirability of optimization lies in the ability to traverse and explore the largely untapped potential of chemical space in a manner that will most benefit the materials discovery process. In the present work, we have developed a deterministic, constrained optimization method that is able to leverage satisfying multiple constraints with efficient navigation of a large optimization space. Density functional theory provides the computational backbone to the optimization and was chosen due to the tradeoff between speed and accuracy and its wide application base. We compared a variety of breadth-first search and gradient-analogous local search algorithms for the optimization process. Each of the algorithms has been benchmarked with respect to efficiency and performance. One of the key benefits of utilizing deterministic techniques in this work is the ability to retain chemical and structural information at each step of the optimization procedure. This feature has allowed for detailed visual analysis of the algorithmic path from input structure to final candidate material and has been beneficial in both the structure search and further algorithm development. In addition, large data libraries can be created and examined to produce qualitative structure-property relationships based on the optimization constraints. At present, this method has been successfully applied to a number of materials science systems of interest including high-hyperpolarizability materials for optics applications and energetic materials. It was found in each case that it was only necessary to explore a small fraction of chemical space (&lt; 1%) to find performant candidates which satisfy the optimization constraints.

Discovery through deterministic optimization: Navigating chemical space for effective material design

251st National Meeting (2016)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.