Combined Automated Reaction Pathway Searches and Sparse Modeling Analysis for Catalytic Properties of Lowest Energy Twins of Cu13.
Takeshi IwasaTakaaki SatoMakito TakagiMin GaoAndrey LyalinMasato KobayashiKen-Ichi ShimizuSatoshi MaedaTetsuya TaketsuguPublished in: The journal of physical chemistry. A (2018)
In nanocatalysis, growing attention has recently been given to investigation of energetically low-lying structural isomers of atomic clusters, because some isomers can demonstrate better catalytic activity than the most stable structures. In this study, we present a comparative investigation of catalytic activity for NO dissociation of a pair of the energetically degenerated isomers of Cu13 cluster having C2 and C s symmetries. It is shown that although these isomers have similar structural, electronic, and optical properties, they can possess very different catalytic activities. The effect of isomerization between cluster isomers is considered using state-of-the-art automated reaction pathway search techniques such as an artificial force induced reaction (AFIR) method as a part of a global reaction route mapping (GRRM) strategy. This method allows effectively to locate a large number of possible reaction pathways and transition states (TSs). In total, 12 TSs for NO dissociation were obtained for Cu13, of C2, C s, as well as I h isomers. Sparse modeling analysis shows that LUMO is strongly negatively correlated with total energy of TSs. For most TSs, LUMO has the antibonding character of NO, consisting of the interaction between π* of NO and SOMO of Cu13. Therefore, an increase in the strength of interaction between NO molecule and Cu13 cluster causes the rise in energy of the LUMO, resulting in lowering of the TS energy for NO dissociation. The combination of the automated reaction pathway search technique and sparse modeling represents a powerful tool for analysis and prediction of the physicochemical properties of atomic clusters, especially in the regime of structural fluxionality, where traditional methods based on random geometry search analyses are difficult.