ab initio modeling of yttria stabilised zirconia for solid oxide fuel cells

Abstract

Fuel cells are electrochemical devices that convert chemical fuels directly into electricity, heat, and waste products with higher efficiencies than many conventional combustion technologies. Fuel cells have already found applications in automotive and domestic applications, where their high efficiencies offer potential reductions in CO2 emissions as well as energy savings. A promising technology is the solid oxide fuel cell (SOFC), which typically runs at temperatures between 500 - 1000C, and can convert hydrogen rich gases, such as methane, directly into heat and electricity.

This thesis presents work on developing improved atomistic models of yttria stabilised zirconia (YSZ), which is used as a catalyst support and electrolyte material in a solid oxide fuel cell. The catalyst, YSZ, and a gas phase containing fuel molecules, meet at the anode to form the anode triple phase boundary (TPB) in an SOFC. The anode TPB is the site at which fuel molecules undergo electrochemical oxidation, a process that releases electrons and waste products. Unfortunately the anode is susceptible to poisoning and damage through detrimental chemical reactions which can lead to carbon deposition and sulphur poisoning. A long term aim for the field of SOFC catalysis is to understand these reactions and design improved catalysts which are resistant to contamination processes. However, to date, the detailed mechanisms involved in these reactions have not been established; even the mechanism for the oxidation of hydrogen at the anode TPB is fiercely debated. For these reasons, there is interest in developing atomistic models of the anode TPB to investigate the thermodynamics of possible reaction paths.

Modeling the anode TPB is dependant on many factors including: materials, surface structure, interfaces, distribution of local defects. A detailed knowledge of the YSZ surface chemistry is currently inhibited by a poor understanding of the distribution and local atomistic structure of the dopant Y3+ ions and oxygen vacancies in the bulk crystal and at the surfaces. In this thesis, a comprehensive search for low energy defect structures using a combined classical modeling and density functional theory (DFT) approach is used to identify the low energy defect structures of 3.2mol% YSZ. 3.2mol% YSZ is chosen as the limit of low dopant concentration and as a simple system to investigate, avoiding the the combinatorial complexity of higher dopant concentrations and defect-defect interactions. Through analysis of energetics computed using; the best available empirical potential model, point charges, DFT, and local strain energy estimated in the harmonic approximation, we examine the main chemical and physical interactions that determine the low energy structures. It is found that the empirical potential model reproduces a general trend of increasing DFT energetics across a series of locally strain relaxed structures, but is unreliable both as it predicts some incorrect low energy structures, and because it finds some meta-stable structures to be unstable. A better predictor of low energy defect structures is the total electrostatic energy of a simple point charge model calculated at the unrelaxed geometries of the defects. In addition, the strain relaxation energy is substantial, and is estimated effectively in the harmonic approximation to the imaginary phonon modes of cubic zirconia (c-ZrO2), but it is not a determining factor for the relative stabilities of low energy defect structures. These results allow us to propose a simple method for identifying low energy YSZ defect structures.

The findings from the studies of bulk 3.2mol% YSZ are used to establish the low energy structures of a 3.2mol% YSZ (111) surface model. After initially demonstrating that a slab model, much larger than that used in previous DFT studies is required to obtain a converged surface energy, the energetic preference for yttrium to segregate to the (111) surface is investigated. After establishing that yttrium indeed segregates to the (111) surface, we compute the DFT energies of 20 low energy symmetry inequivalent surface structures, and identify the preferential defect configurations and surface chemistry sites. In addition, the DFT energy of the low energy NN structure proposed by Reaxff modeling is computed. It is shown that this structures is significantly higher in energy than our minimum energy structure, which has NNN geometry. This highlights the need for large scale DFT calculations in understanding the YSZ (111) surface structure. Having obtained atomistic structures for the surface reaction sites, water dissociation onto the lowest energy YSZ (111) surface is investigated. It is shown that it is preferable for water to associatively adsorb to the YSZ surface, and this is optimal when water adsorbs to the yttrium site. Dissociative adsorption of water is only possible over zirconium sites with the process generally being endothermic. Some exothermic paths for dissociative adsorption exist, however there are large energy barriers to the process. Associative adsorbtion to the surface yttrium site is the global minimum of the system, and yttrium sites appear to act as a trap for water molecules.

Finally, the methodology developed in previous sections is used to investigate the 6.7mol% YSZ system, which is closer to the Y2O3 dopant concentration used in most commercial SOFCs (8 - 10 mol%). It is found that, whereas the electrostatic energy of the unrelaxed structures calculated using a point charge model was a good predictor of the likely low energy 3.2mol% defect structures, it is a poor predictor of the likely low energy 6.7mol% defect structures. In addition, while it was found that the best available Born-Mayer-Huggins potential model recreated general trends in DFT energies at 3.2mol%, it completely fails to reproduce DFT energy differences at 6.7mol%. In the absence of an easy to calculate, reliable predictor of the likely low energy DFT defect structures, we correlate the formation energies of the structures to simple geometric parameters. We perform an exhaustive search on 2857 symmetry inequivalent structures, characterising every structure in terms of intuitive quantities, such as: vacancy - vacancy separation, the average vacancy - Y3+ interatomic separation, the average Y3+ - Y3+ interatomic speration, the surface area occupied by the defect cluster, and the volume of the defect cluster. It is possible to explain the electrostatic formation energies of the defects in terms of intuitive attractive and repulsive forces and to find weak trends between the geometric descriptors and the final relaxed DFT energies, however, without an extensive database of fully relaxed DFT energies, it is hard to determine the statistical meaning of these results. This result highlights the combinatorial complexity of the 6.7mol% system and establishes the need for further large scale DFT calculations on the 6.7 mol% system.