(657g) Designing Stable Semiconductor Nanowires Via a Generic Materials Optimization Toolkit | AIChE

(657g) Designing Stable Semiconductor Nanowires Via a Generic Materials Optimization Toolkit


Yin, X. - Presenter, Carnegie Mellon University
Hanselman, C. L., Carnegie Mellon University
Gounaris, C., Carnegie Mellon University
Semiconductor nanowires (NWs) exhibit distinct properties due to its unique one-dimensional (1d) periodicity [1]. They are promising materials in numerous catalytic applications [2,3,4], and have attracted extensive research focus. The fast-growing NWs fabrication ability [5] brings the new challenge of NWs engineering design. This work proposes designing highly stable NWs using our new Python-based generic materials optimization package, MatOpt [6].

Cohesive energy (CE) indicates a NW’s overall stability. Consequently, the search for energetically most stable NWs can be simplified to maximizing CE. Previous work in the literature has focused on using heuristic search methods such as genetic algorithms [7] and simulated annealing [8] to find the energetically most favorable NW structure. These approaches generally find stable nanostructures but often miss some of the most stable ones. We propose here a mathematical optimization-based design framework for this combinatorial design problem. The CE structure-function relationships are cast into a rigorous Mixed Integer Linear Programming (MILP) model. The design space can then be explored systematically by solving the model via established numerical optimization algorithms. The obtained, optimally-stable NW designs can serve as model structures for further experimental and theoretical investigations.

A similar, MILP-based, bottom-up design framework has been previously demonstrated for various types of nanomaterials, including metallic nanoclusters [9,10], periodic catalytic surfaces [11,12], and bulk metallic oxides [13]. In those cases, the design framework has found unintuitive optimal solutions and revealed interesting physical trends. To simplify the implementation of our generic framework and make it accessible to non-experts in combinatorial mathematical optimization, we have codified our methodologies into the Python-based toolkit MatOpt. While we focus on using MatOpt to design 1d-periodic NWs by maximizing CE, we shall also illustrate how the methodology can be easily extended further to a broad range of NW design problem and other NW functionalities of interest.