PtRu nanoalloys decorated In2O3 nanoparticles: Gas sensing performance, mechanism study and machine learning-assisted discrimination of multiple volatile organic compounds

Bimetallic nanoalloys have excellent catalytic properties, which are very promising for sensor applications. Here, PtRu nanoalloyed catalysts on In2O3 are proposed as gas sensor. The 0.5 wt% Pt/Ru-In2O3 nanoparticles represent 13 times higher response (S=93) towards 100 ppm acetone at 200 degrees C as compared to pure In2O3. The molecular arrangement, micro-structure, sensing performance, and working principle were elucidated in detail, and density functional theory (DFT) has been considered to probe deeper into the sensing mechanism. Furthermore, several machine learning methods have been introduced to ameliorate poor selectivity of the sensor. Following, the effectiveness of random deep forests (RF), convolutional neural networks (CNN), back propagation neural networks (BP), long-short-term memory networks (LSTM), and particle swarm-optimized back propagation neural networks (PSO-BP) for discriminating multiple volatile organic gases (VOCs) was explored then examined the better one in recovering weak selectivity of metal-oxide-semiconductor (MOS) sensors. Finally, the PSO-BP model was proved to be better with gas discrimination success rate of 96.19 %. The collective results promote a step towards the realization of single MOS sensor for discrimination of multiple VOCs.