LAUSR

Abstract Reaction mechanism and catalytic activities of methanol (CH3OH) and CO formations from CO2 hydrogenation at all the different surface oxygen vacancy (Ov) sites on the stable cubic In2O3 (c-In2O3) (111) flat surface and (110) flat and (110) step surfaces were thoroughly investigated in order to establish the structure-performance relationship using density functional theory calculations. For CH3OH formation, the R2 step for bidentate formate intermediate (bi-HCOO*) hydrogenation to the dioxymethylene intermediate (H2COO*), or the R3 step for H2COO* dissociation to the formaldehyde intermediate (CH2O*) and surface O atom, was calculated to be the rate-determining step (RDS), whereas for CO formation, the R1a step for bent CO2 adsorbate (bt-CO2*) protonation to the carboxylate intermediate (COOH*), was calculated to be the only RDS. Further data science analysis using the perceptron learning and decision tree algorithms shows that the RDS for CH3OH formation for a given Ov site can be determined by the stability of two key reaction intermediates (H2COO* and CH2O*) along with the formation energy of the Ov site. Linear regression analysis was also performed to establish linear relationships between the activation barriers and corresponding reaction energies for R2 and R3 steps, which can also be useful in determining the actual RDS for CH3OH formation. Although the adsorption energies of the linear CO2 configuration intermediate (ln-CO2*) and the bt-CO2* as well as the difference between the activation barriers of the RDS for CO and CH3OH formations are required for reliably predicting the catalytic activity and product selectivity, our studies suggest a simple and intuitive structure-performance relationship for the c-In2O3 catalyst that tri-coordinated Ov sites favor CH3OH formation, whereas bi-coordinated Ov sites favor CO formation.