Data-Driven Identification of Low-Dimensional Descriptors for the Rational Design of Stable and Active Single-Atom Catalysts for Nitrogen Oxidation

Abstract

Despite recent advances, simple intrinsic descriptors that can simultaneously assess the stability and activity of electrocatalysts for nitrogen oxidation remain elusive. In this work, first-principles calculations combined with data-driven analysis were employed to systematically investigate the stability and catalytic performance of single-atom catalysts supported on the anatase TiO2(101) surface. 58 transition-metal singleatom sites were constructed and screened. The analysis results reveal that three intrinsic elemental parameters, namely atomic weight (MW), d-electron number (Nd), and the presence of oxygen vacancies (VO), serve as unified intrinsic descriptors for both stability and activity. This three-parameter descriptor establishes a clear structure-stability-activity relationship space, enabling the identification of a region that simultaneously favors both structural stability and catalytic activity. Furthermore, the effectiveness of this descriptor-guided design window is validated through Pareto-front analysis. As a result, W-TiO2 is identified as the optimal catalyst, achieving a balanced optimization between structural stability and catalytic activity.