Using Evolutionary Inspired Search Methodologies To Explore Polymorphism or Defect Associations in Materials Design

Title: Using Evolutionary Inspired Search Methodologies To Explore Polymorphism or Defect Associations in Materials Design
Authors: Hooper, James
Date: 2011
Abstract: The work presented in this thesis reinforces the notion of how powerful evolutionary search methodologies, specifically those inspired by the genetic algorithm (GA), are when applied to search out the chemical and/or configurational spaces of a number of chemical systems. Researchers in the field often sell the capability of GA search techniques to operate without constraints but, as is demonstrated in this thesis, sometimes the use of constraints can not only streamline computational costs but open up new avenues to explore in materials design. As a starting point, this work first focuses on developing effective methodologies that map out nitrogen's high-pressure potential energy surface (PES) such that polymeric nitrogen, a relatively new material with a potentially promising outlook in future high-energy density materials applications,¹ could be explored. High pressure nitrogen (> 30 GPa) is an ideal candidate to explore theoretically since conclusive experimental characterizations are often beyond the reach of current laboratory technologies.² Theoretical and experimental assessments of polymorphism in high-pressure nitrogen have been reported in the recent literature,³ and, upon the inception of this thesis, much of its potential energy surface remained unmapped. Using both established and novel search methodologies on nitrogen's high-pressure potential energy landscape, with GA-inspired methodologies included among them, a number of previously unreported allotropes of molecular and non-molecular nitrogen at high pressures are reported which relate well with experimental studies. The second focal point of this thesis is to present a general methodology capable of mapping generic defect associations in doped metal oxides. Doped metal oxides have many applications as substrates in catalysis, as gas sensors, and as next generation solid electrolytes in solid-oxide fuel cell technologies, just to name a few. The performance of these metal oxide materials can be enhanced (or, more generally, tuned) via doping of the parent oxide with other elements at varying concentrations; however, the optimization of the dopant composition is often performed by trial and error. By introducing appropriate constraints, GA-inspired routines can sample the configurational and chemical spaces of any defects introduced to a native oxide lattice. Thus, genetic algorithms are used herein in two capacities: 1) to find the lowest energy configurations of a specific metal oxide composition and 2) to search a doped metal oxide's chemical space in order to optimize a specific property of interest based on a given 'fitness' metric. This approach is validated by optimizing electronic mobility in doped zinc oxide, which is used in gas sensor technologies, and ionic mobility in Lanthanide doped ceria (LDC), which is used as a solid electrolyte in solid-oxide fuel cell technologies. ¹M. I. Eremets, A. G. Gavriliuk, I. A. Trojan, D. A. Dzivenko, and R. Boehler, Nature Materials 3, 558 (2004). ²E. Gregoryanz, C. Sanloup, R. Bini, J. Kreutz, H. J. Jodl, M. Somayazulu, H.-k. Mao, and R. J. Hemley, Journal of Chemical Physics 124, 116102 (2006). ³W. D. Mattson, D. Sanchez-Portal, S. Chiesa, and R. M. Martin, Phys. Rev. Lett 93, 125501 (2004).
CollectionTh├Ęses, 1910 - 2010 // Theses, 1910 - 2010
NR79731.PDF13.23 MBAdobe PDFOpen