High-Throughput Screening of Gas Sensor Materials for Decomposition Products of Eco-Friendly Insulation Medium by Machine Learning

Abstract

Nowadays, trifluoromethyl sulfonyl fluoride (CF 3 SO 2 F) has shown great potential to replace SF 6 as an eco-friendly insulation medium in the power industry. In this work, an effective and low-cost design strategy toward ideal gas sensors for the decomposed gas products of CF 3 SO 2 F was proposed. The strategy achieved high-throughput screening from a large candidate space based on first-principle calculation and machine learning (ML). The candidate space is made up of different transition metalembedded graphic carbon nitrides (TM/g-C 3 N 4 ) owing to their high surface area and subtle electronic structure. Four main noteworthy decomposition gases of CF 3 SO 2 F, namely, CF 4 , SO 2 , SO 2 F 2 , and HF, as well as their initial stable structure on TM/g-C 3 N 4 were determined. The best-performing ML model was established and implemented to predict the interaction strength between gas products and TM/g-C 3 N 4 , thus determining the promising gas-sensing materials for target gases with the requirements of interaction strength, recovery time, sensitivity, and selectivity. Further analysis guarantees their stability and reveals the origin of excellent properties as a gas sensor. The high-throughput strategy opens a new avenue of rational and low-cost design principles of desirable gas-sensing materials in an interdisciplinary view.

Expanding the Horizons of Machine Learning in Nanomaterials to Chiral Nanostructures

“…[164,165] Wan et al used ML for modeling potential sensors for detection of the gaseous decomposition products of CF 3 SO 2 F-an ecofriendly gaseous electrical insulation medium. [166] The authors performed a high-throughput screening study of a large variety of transition metal-embedded graphitic carbon nitride, with the primary objective being the ML-powered prediction of the interaction strength between the transition metal-embedded graphitic carbon nitride and the gaseous decomposition products. A hybrid DFT/ML method was applied, using 8 supervised ML algorithms, with support vector regression being identified as the optimum model.…”

Recent Progress in Multifunctional Gas Sensors Based on 2D Materials

“…For example, Wan et al [112] reported a state-of-the-art DFT and ML hybrid scheme for the accessibility analysis of transition metal (TM)/g-C 3 N 4 heterostructures. The research was entirely based on computational simulation; at first, 28 different potential TM/g-C 3 N 4 structures were predicted and modeled, followed by the bond analysis of the template analyte molecule CF 3 SO 2 F, indicating that CF 4 , SO 2 F 2 , SO 2 , and HF were the most energetically favorable products.…”

General Model for Predicting Response of Gas-Sensitive Materials to Target Gas Based on Machine Learning

“…Then, to predict the gas-sensitive properties of zinc oxide, the factors affecting its gas-sensitive properties were calculated by first-principles computation. At the same time, it is worth noting that it is impractical to predict gas-sensitive properties only from a single property, for example, some materials can meet the adsorption energy requirements but have no gas-sensitive properties. − Therefore, establishing a universal model between various computationally derived properties and intelligent algorithms to circumvent trial-and-error selection has become a significant challenge. − …”