Identifying key factors of peroxymonosulfate activation on single-atom M–N–C catalysts: a combined density functional theory and machine learning study

Abstract

Persulfate-based advanced oxidation processes have been widely praised in the treatment of organic contaminants, but the intrinsic factors of peroxymonosulfate (PMS) activation have not been identified. In this work, a series of transition metal-doped single-atom catalysts were used to investigate the intrinsic factors of PMS activation through combining density functional theory and machine learning. Three types of graphene-supported single-atom catalysts (M@N₂C₂, M@N₃C₁, and M@N₄, where M is a transition metal atom) were selected for PMS adsorption and activation. The d-band theory was utilized to effectively describe the correlation between the adsorption energy of PMS and the electronic properties of catalysts. Based on the few-shot learning algorithm, a machine learning model was built to reveal the underlying pattern between easily obtainable properties and energy barriers. The low energy barrier results indicated that the best candidates were V@N₄, Cr@N₄, and Hf@N₃C₁. In descending order of importance, the five electronic and geometric features representing catalytic properties are group number, d-electron count, electronegativity, radius, and the number of nitrogen atoms. A novel intrinsic descriptor based on these five features is proposed, which can efficiently predict the performance of unknown catalysts. This work provides mechanism interpretation and quantitative guidance for PMS activation, which is expected to be a widely-used method for developing efficient catalysts in the future.