Hydrogen activation by rhodium under the cover of a copper oxide thin film

Abstract

The activation of reactants by catalytically active metal sites at metal-oxide interfaces is important for understanding the effect of metal-support interactions on nanoparticle catalysts and for tuning activity and selectivity. Using a combined experimental and theoretical approach, we studied the activation of H2 and the effect of CO poisoning on isolated Rh atoms completely or partially covered by a copper oxide (Cu2O) thin film. Temperature-programmed desorption (TPD) experiments conducted in ultra-high vacuum (UHV) show that neither a partially nor a fully oxidized Cu2O layer grown on a Rh/Cu(111) single-atom alloy can activate hydrogen in UHV. However, in situ ambient pressure X-ray photoelectron spectroscopy (AP-XPS) experiments performed at elevated H2 pressures reveal that Rh significantly accelerates the reduction of these Cu2O thin films by hydrogen. Remarkably, the fastest reduction rate is observed for the fully oxidized sample with all Rh sites covered by Cu2O. Both TPD and AP-XPS data demonstrate that these covered Rh sites are inaccessible to CO, indicating that Rh under Cu2O is active for H2 dissociation but cannot be poisoned by CO. In contrast, an incomplete oxide film leaves some of the Rh sites exposed and accessible to CO, and hence prone to CO poisoning. Density functional theory calculations demonstrate that unlike many reactions in which hydrogen activation is rate limiting, the rate-determining step in the dissociation of H2 on thin-film Cu2O with Rh underneath is the adsorption of H2 on the buried Rh site, and once adsorbed, the dissociation of H2 is barrierless. These calculations also explain why H2 can only be activated at higher pressures. Together, these results highlight how different the reactivity of atomically dispersed Rh in Cu can be depending on its accessibility through the oxide layer, providing a way to engineer Rh sites that are active for hydrogen activation but resilient to CO poisoning.