ABSTRACT The strong metal‐support interactions (SMSI) between supported clusters and substrate can drive the local structure reconstructions, which stabilize the supported clusters to avoid migration and aggregation. Such reconstruction and stabilization mechanism are crucial to construct the atomically dispersed catalysts (ADCs), but they are too complex to simulate in most catalyst theoretical studies for a relative long time. Herein, an accurate machine learning potential (MLP) is employed into Monte Carlo simulation on the Mn N (1 ≤ N ≤ 7) clusters supported on MoS 2 layer. The adsorption, reconstruction and thermodynamic and kinetical stabilization of Mn clusters on perfect and defective MoS 2 are compared studied. The results indicate that the S vacancies can effectively anchor Mn clusters and are feasible to control the dual‐atom catalysts (DACs) on the MoS 2 surface. Besides, Comparative analysis reveals that the Mn 2 @MoS 2 ‐S 2 V exhibits superior NH 3 ‐SCR catalytic activity. The complete reaction process of Mn 2 @MoS 2 ‐S 2 V following the “Fast‐SCR” mechanism and the NO 2 reduction pathway is the dominant route, with a rate‐determining barrier of 1.03 eV. This work provides a pioneer way to disclose the very complex reconstruction of supported clusters with SMSI in simulation, which is indeed helpful to design real atomic structure of ADCs.
Wang et al. (Tue,) studied this question.