(684f) Determining Catalytically Relevant Surface Structures Via Multiscale Models: Modeling Distribution Tendencies of Noble Metals on Fe(100)

Bio-oils are an alternative to fossil-based fuels, but they unfortunately inherit the high oxygen content of their originating biomass. To address this issue, catalytic hydrodeoxygenation (HDO) is used to selectively remove oxygen-containing functional groups from bio-oil, and Fe-based catalysts have been shown to be exceptionally selective for HDO. However, this selectivity is the result of Fe’s high oxophilicity, and these catalysts are thus prone to oxidative deactivation. Noble metal promoters have been shown to protect against this deactivation, but to rationally design future catalysts, highly predictive theoretical models are needed. However, inclusion of lateral interactions between adspecies is mandatory as these may induce different clustering tendencies that can significantly affect catalytic performance. Here, we employ a DFT-parametrized Lattice Gas Cluster Expansion (LGCE) method to quantify these lateral interactions (see Figure 1a), which can be used to develop a multiscale model that can predict clustering vs. dispersion tendencies under experimental conditions. We compare four approaches in obtaining the lateral interactions between four different dopant metals (Pd, Pt, Ru and Rh) by independently substituting them with Fe within an Fe(100) host. In the first two models, we determine the LGCE independently for the four dopants by either (i) fixing them to their ideal bulk positions or (ii) by allowing them to relax. The third and the fourth models are constructed so that the structures generated for all four dopants are taken into account in the construction of the LGCE. We hypothesize that well-dispersed (and not clustered) noble metal promoters protect Fe against oxidation best. From our work, we predict that promoter dispersion increases in the following order: Pd<Pt<Ru<Rh, where Rh is the most dispersed promoter on the Fe surface (see Figure 1b). Information gained from our predictive model can help streamline research efforts, thereby accelerating the design of effective HDO catalysts.