Machine learning-assisted Ru-N bond regulation for ammonia synthesis.

Li, Zichuang; Zhang, Mingxin; Su, Xiaozhi; Lu, Yangfan; Li, Jiang; Zhang, Qing; Li, Wenqian; Qian, Kailong et al. · Nat Commun · 2025

basic_science · Level V

Where this comes from

Abstract

Ruthenium-bearing intermetallics (Ru-IMCs) featured with well-defined structures and variable compositions offer new opportunities to develop ammonia synthesis catalysts under mild conditions. However, their complex phase nature and the numerous controlling parameters pose major challenges for catalyst design and exploration. Herein, we demonstrate that a combination of machine learning (ML) and model mining techniques can effectively address these challenges. Based on the combination techniques, we generate a two-dimensional activity volcano plot with adsorption energies of N<sub>2</sub> and N, and identify the radius of atom coordinating to Ru as a key parameter. The well-designed Sc<sub>1/8</sub>Nd<sub>7/8</sub>Ru<sub>2</sub> reaches as high as 8.18 mmol m<sup>-2</sup> h<sup>-1</sup> at 0.1 MPa and 400 °C. Density functional theory (DFT) calculations combined with N<sub>2</sub>-TPD and FT-IR studies reveal that Ru‒N interaction is controlled by the d-p orbital hybridization between Ru and N. These findings underscore the importance of ML towards material design on exploring IMCs for ammonia synthesis.