Destruction and preservation of nonstoichiometric ZnCr oxide catalyst from machine-learning simulation

Abstract

Structural destruction, for example via the solid disproportionation reaction (SDR), is one of the main causes of catalyst deactivation that often lacks atomic level understanding due to the strong coupling of reaction conditions and the long-time scale of the structural evolution. Here by using machine-learning-based atomic simulations we are able to clarify the detailed deactivation mechanism of the nonstoichiometric zinc-chromium oxide (ZnCrxOy) catalyst, an important industrial catalyst for syngas conversion. We show that the reductive reaction conditions first lead to the generation of surface and subsurface oxygen vacancies, and the subsequent cation migration via the replacement of subsurface Zn by surface Cr kills the catalyst activity by diminishing the active surface planer [CrO4] sites. The rate-determining step, the Zn/Cr cation migration, is exothermic towards forming stoichiometric ZnCr2O4, and has an accessible barrier (∼1.7 eV) under syngas conversion reaction conditions, indicating that the long-term deactivation is inevitable. By screening a set of additive elements, we demonstrate that Al element is an effective inhibitor to impede the formation of subsurface oxygen vacancy, the initial step of SDR. Our findings demonstrate the power of machine-learning-based atomic simulations in elucidating complex structural transformations and provide directions to mitigate ZnCrxOy catalyst deactivation.