Uncovering the active sites of single atom-doped rutile oxides during methane activation by data-driven approach

• Published in: Science China materials Vol. 67; no. 4; pp. 1231 - 1242
• Main Authors: Wei, Fenfei, Ge, Bingqing, Dong, Peipei, Wan, Qiang, Hu, Xixi, Lin, Sen
• Format: Journal Article
• Language: English
• Published: Beijing Science China Press 01-04-2024; Springer Nature B.V
• Subjects: Activation energy; Catalysis; Catalysts; Chemistry and Materials Science; Chemistry/Food Science; Chromium; Density functional theory; Germanium; Iridium; Machine learning; Materials Science; Metal oxides; Methane; Molybdenum; Oxidation; Rhodium; Ruthenium; Rutile; Tantalum; Tin; methane activation; machine learning; descriptors
• ISSN: 2095-8226; 2199-4501
• DOI: 10.1007/s40843-023-2771-4

Metal oxides are commonly used in methane activation and conversion, but usually suffer from over-oxidation. The introduction of single atoms is an attractive way to overcome this challenge, but the actual role of doped single atoms remains controversial. Here, we adopted single atoms (D
guest
, D = Ti, V, Cr, Mn, Nb, Mo, Ru, Rh, Ta, Re, Os, Ir, Pt, Si, Ge, and Sn)-doped rutile metal oxides (MO
2
, M = Ru, Rh, Ir, Pt, Mo) as model catalysts and investigate methane activation at various surface sites and elucidate the actual active sites in such doped surfaces by using the density functional theory calculations and data-driven approach. Overall, we obtain derived multidimensional descriptors from a large space of feature-combined descriptors by using the machine learning approach, which allows uniform prediction of the energy barrier of CH
4
activation on both D
guest
and M
host
, independent of the transition state calculation. The regulation of selective oxidation by guest sites on MO
2
was confirmed. This work sheds light on the complicated role of dopants in catalysis, and the developed descriptors help determine the activation energy to provide potential selective oxidation sites of rutile oxide-based catalysts.