ME DEPARTMENTAL SEMINAR

 

Friday, December 1, 2006

2:00pm – 3:00pm

2211 GG BROWN

 

 

Professor Heinz Pitsch

Department of Mechanical Engineering

Stanford University

 

ÒComputational Chemistry Based Multi-Scale Simulations of Polymer

Electrolyte Membrane Fuel Cells: Development of Models

and Numerical Algorithms, and Application to Fuel CellsÓ

Abstract:

 

Computational chemistry presently is a rapidly evolving field, because of recent improvements in computational power and theoretical developments, and the great potential of computations to further the understanding of chemical processes and the interactions of chemistry and transport phenomena. These computational approaches include quantum chemistry simulations, molecular dynamics simulations, and dynamic Monte Carlo simulations (DMC).

 

Here we will present the development of a computational chemistry based multi-scale model for computational fluid dynamics simulations of polymer electrolyte membrane (PEM) fuel cells. The multi-scale model is based on DMC simulations of the chemistry on the electrocatalyst surfaces. Several advancements in numerical techniques for computational chemistry simulations and their applications to real systems will be presented for the example of the PEM fuel cell cathode.

 

Transition probabilities required for these simulations are determined from quantum chemical simulations. For electrochemical simulations, the local reaction center theory by Anderson is used. We present an efficient mathematical framework to determine the potential-dependent transition states of electron transfer reactions by quantum calculations. This method leads to fast convergence, reliability, and robustness of the located transition states for more complex systems with a larger number of degrees of freedom, and makes these computations cost-efficient enough to study a large number of individual reactions. As an example, adsorbent interactions relevant for electrochemical steps of the oxygen reduction reactions are discussed.

 

Because of the possible importance of such adsorbent interactions and other non-linear local chemical effects, DMC methods are expected to describe the chemical behavior more accurately than environment-averaged methods. In PEM fuel cells, carbon-particle supported platinum nano-particles are often used as electrocatalyst. These Pt-particles can be approximated to be of cubo-octahedral form. The specific topology of these particles can lead to important features associated with the complex surface structure. Specifically, the edge/corner sites can behave differently from sites located on the faces. Environment-averaged approaches, such as the mean-field approximation, often fail to accommodate the details of such local phenomena. DMC is computationally much more demanding than conventional approaches, and several different DMC simulations algorithms have been proposed in the past. An example is the popular Variable Step Size Method (VSSM). VSSM has the advantage that the computational cost of a single time step is independent of the lattice size for problems with time-independent rate parameters, but scales with the square of the number of lattice sites otherwise. Another method, the First Reaction Method (FRM), can be applied for time-varying rate coefficients, but the computational cost per time step depends still linearly on the logarithm of the number of lattice sites. Here we present a new DMC algorithm that can be applied for time-varying rate coefficients, and which has a computational cost per time step that is independent of the lattice size. To demonstrate the capabilities of the new method, DMC simulations of cyclic voltammetry of PEM fuel cell electrochemistry will be presented and compared with experimental observations.

 

Finally, the integration of the DMC simulation technology into a multi-scale model will be presented. The model describes the interaction of surface chemistry with gas diffusion in thin electrolyte layers surrounding the platinum particles on the nano-scale and the transport in the porous material of the catalyst layer on the mirco-scale. Simulation results will be compared with experimental data of a fuel cell using single crystal Pt electrodes.