Computational Studies of Oxygen Reduction Catalysts

Abstract

The oxygen reduction reaction (ORR) is of central importance in the development of more efficient and durable fuel cells and computer simulations can be used to help explain and predict properties of ORR catalysts. In this thesis, various simulation methods are used, their accuracy tested, and some new tools developed for simulations of different types of ORR catalysts. The first part of the thesis presents a ReaxFF reactive force field study of the oxidation of platinum nanoparticles, the catalyst currently used in most fuel cells. The simulations make use of a grand-canonical Monte Carlo simulation approach to study the mechanism of oxidative nanoparticle degradation. Electrochemical oxidation phase diagrams are constructed and they show that stable surface oxides can form under fuel cell operating conditions. Furthermore, clusters of Pt6O8 stoichiometry are identified as the primary oxidation product and this provides a detailed, atomic scale mechanism for the degradation of platinum nanoparticle ORR catalysts. In the second part of the thesis, metal-free ORR catalysts are studied, namely nitrogen-doped graphene (NG). There, electron density functional theory calculations are used to estimate the thermodynamics of possible reaction paths and provide an estimate of the overpotential. Various density functional approximations are tested against high level diffusion Monte Carlo calculations on the binding and migration of an *O adatom on graphene, an important intermediate in the ORR. It is found that generalized-gradient approximation (GGA) functionals have low accuracy while some hybrid functionals and a self-interaction corrected GGA functional give good agreement with the reference calculation. A hybrid functional is then used to calculate the free energy of ORR intermediates to estimate the overpotential for various catalyst structures and compositions. While overpotentials are initially found to be unfavorable, the presence of water molecules at the catalyst surface is estimated to reduce the calculated overpotentials significantly. This shows that proper inclusion of the aqueous electrolyte is important. From this conclusion, a significant challenge arises as inclusion of many solvent molecules makes the simulated system too large for electronic structure calculations. In the third part of the thesis, a new methodology is developed to make it possible to include the aqueous electrolyte in simulations by using a hybrid simulation approach where part of the system — the catalyst as well as the reacting species and nearby water molecules — are included in the electronic structure calculation, while the rest of the aqueous electrolyte is described using a potential energy function. The separation between the two regions is made to lie through the aqueous phase in order to make it easier to describe the interaction between atoms on opposite sides of the boundary. Therein, the challenge is to enforce the boundary in such a way that atoms and molecules do not wander from one region to another. A new method referred to as scattering-adapted flexible inner region ensemble separator (SAFIRES) is developed for this purpose. It represents an improvement on a previous algorithm called FIRES. With SAFIRES, the boundary between the two regions is flexible and adjusts automatically but particles cannot move between the regions. It is demonstrated that using SAFIRES, the energy, atomic forces, and probability distribution for the location of atoms are reproduced accurately compared to results of simulations where where no boundary is present. The SAFIRES algorithm dynamically changes the time step in the iterative algorithm for time evolution to identify and enforce elastic collisions of atoms with the boundary separating the two regions. A new propagator is introduced to simulate the time evolution of the system with or without a connection to a heat bath. Tests are carried out by simulating a Lennard-Jones liquid, a Lennard-Jones liquid/solid interface, and water described using a potential energy function. With the SAFIRES method, along with the recently developed, self-consistent polarizable hybrid simulation methodology, the stage is set for proper inclusion of the aqueous phase at the electrode surface in future electrochemistry simulations.