If progress is to be made at ultimately overcoming the technical and cost limitations of PEM fuel cells, a significant investment in the fundamental science of the reactions taking place must be made. The objective for this proposal is to determine the detailed atomistic mechanism including free energy barriers for the oxygen reduction reaction at PEM fuel cell cathodes. The focus is on how the mechanism and rates depend on alloy composition, distribution between surface and bulk regions, and solvent. The computational results would be tested by predicting how binary and ternary catalysts would be expected to improve selectivity, rates, and lifetime. In addition, the PIs, William A. Goddard III Boris Merinov, both of the Materials and Process Simulation Center at California Institute of Technology, propose to determine mechanisms of catalyst degradation and how they depend on alloy composition. The result is to be a computational model sufficiently accurate to be useful in guiding both experiments and engineering applications.
There has previously been no practical means to couple such a wide range of reactive phenomena based solely on first principles. This novel approach would predict data for engineering models from first principles, allowing new systems to be designed computationally and then tested against experiment. To enable this model testing, collaborations have been arranged with Argonne National Labs and with Ford Scientific Labs to carry out experiments on those alloys predicted to be most promising. This model should aid the development of accurate engineering models informed from the theory and simulations but adjusted to incorporate results from experiments. This approach will be essential to develop the improved materials and processes needed to enable new alloys to meet the current targets for improved fuel cells.
The development of improved catalysts (more efficient, longer-lived) should accelerate development of efficient fuel cells that would be commercially viable for transportation, energy production and storage, with the resultant environmental impact. In the broader sense, in addition to contributing significantly to the development of improved alloy catalysts for fuel cell cathodes, the successful coupling of computational tools including QM through ReaxFF reactive dynamics to simulation of the catalyst/support system would apply to other problems in catalysts, materials, and energy.
Fuel cells (FCs) provide promising solutions for energy production and storage with higher efficiency and reduced polluting emissions than petroleum-based systems. However, current fuel cell performance is limited by the cathode oxygen reduction reaction (ORR), which is 400 times slower than the anode reaction for PEM Hydrogen FC, making them too expensive. Our goal was to develop a full fundamental understanding of the reaction mechanism underlying ORR to help design new catalytic materials with kinetics improved sufficiently to make these FCs economic. This project succeeded. Using Quantum Mechanics (QM) we identified that for low dielectric (ε) materials (the Teflon phase in the Nafion electrolyte) the mechanism for O2 activation was not direct splitting of the O2 bond to form two Oxygen atoms bonded to the surface (Oad) as previously thought, but rather the O2 attaches to a surface H (Had) to form HO2 bonded to the surface (HO2ad) which then dissociates to form (HOad + Oad). We also showed that to convert Oad reacted with H2Oad to form two (OHad). The final step of adding Had to OHad to form H2O is fast. Thus the rate-determining step (RDS) for low ε has a barrier of 0.3 eV. In order to understand how the presence of aqueous acid affects the results, we developed a modified QM code including solvent polarization and showed that the RDS becomes the reaction of H2Oad with Oad to form two OHad, with a barrier of 0.5 eV. We then studied the reaction kinetics as a function of solvent polarization, showing that the optimum would be a dielectric constant of 2 or 3, achievable with a mixture of THF with water. The experiments to test this have not been done. Experimentalists had shown that Pt3Ni and Pt3Co alloys lead to significantly higher ORR catalysts than pure Pt. We used QM to determine that the optimum composition of these alloys at the surface is pure Pt in the top layer, followed 50% Ni in the a second layer. We then used QM to show that this does leads to increased ORR rates. However, under fuel cell operating conditions, Ni and Co leach from the catalyst to electrolyte leading to a reduced concentration of the Ni and Co near the surface. To solve this problem, we investigated surface segregation for 28 different Pt3M alloys with adsorbed ORR oxidative species, such as Oad and OHad, on the surface and found that only Pt3Os and Pt3Ir show a stable surface segregation in the presence of the both adsorbed species. Thus, we expect that Pt3Os and Pt3Ir would be stable under the operating conditions. We then predicted the ORR barriers for this alloys and showed that Pt3Os has a lower barrier for the RDS, predicting better performance than for Pt. Indeed we showed that the best performance would be for the core-shell structure with two Pt monolayers on the surface. This led to a collaboration with experimentalists who fabricated Os/Pt/C core-shell nanoclusters and demonstrated that the Pt/Os/C structure with two Pt monolayers shows 3.5 to 5 times better catalytic activity than Pt/C, in agreement with our theoretical prediction. Thus Os/Pt core-shell structured catalysts might lead to improved FCs. A remarkable development by 3M (Mark Debe) demonstrated that catalysts with a composition around Ni:Pt=70:30 lead to a dramatic improvement (4 times) in the steady state ORR but their XPS analyses of these final materials showed no observable Ni! We solved this conundrum of why 70% initial Ni dramatically improves catalyst performance even though no observable Ni is present in the operational catalysts by carrying out Reactive Molecular Dynamics (RMD) using the ReaxFF reactive force field trained with QM. We considered nanoparticles with 5 to 15 nm diameter and various alloys from 50 to 90% Ni and showed that leaching out the Ni leads to a porous Pt structure. We found that for 70% Ni 7.5 nm starting particles, the pores penetrate throughout the particle while connecting to the surface. This leads to a surface area twice that of Pt particles of similar size, in agreement with BET experiments. We then examined the barriers for the various reaction steps and showed that the barrier for the RDS is significantly reduced from the bulk catalyst because the surfaces atoms have reduced coordination, explaining the increased activity observed experimentally. The RMD studies also showed an optimal initial nanoparticle size of ~7.5 nm, which is consistent with experiments on nanoparticle size optimization. This development of improved ORR catalysts using QM-based simulations, illustrates the new strategy of using theory to drive experiment, predicting the best cases to be pursued experimentally. This should play an essential role in accelerating development of advanced electrochemical devices, commercially viable for transportation, energy production and storage, with huge environmental impact.
