Tracing the reactivity of single atom alloys for ethanol dehydrogenation using ab initio simulations

Abstract

An ab initio micro-kinetic model (MKM) is constructed to understand the reactivity trend of single atom alloys (SAAs) of Cu and Au for non-oxidative dehydrogenation (NODH) of ethanol to produce...

Machine Learning Enabled Screening of Single Atom Alloys: Predicting Reactivity Trend for Ethanol Dehydrogenation

“…Since d‐band center alone is not expected to form a credible descriptor for the SAAs formed from the group 11 elements, other features from the periodic properties of the elements are included in the model. The test errors for E O and E C , estimated from the ML model are around 0.2 eV (Figure 5(a) and (b)), which is similar to the error bars on DFT calculations reported in our earlier studies [7,36,64–67] . Toyao et.…”

“…[19] Finally, the efficacy of the ML model is tested for the NODH reaction of ethanol, using E O and E C as descriptors using a previously built ab initio MKM for the estimation of turnovers of ethanol consumption on metal catalysts. [36] The ML model has predicted significantly higher TOFs for the NiAu SAA as compared to the Au surface (Figure 9(b)), with the difference of four order of magnitude. The result is consistent with the experimental study undertaken by Giannakakis et al wherein they observed a decrease in the apparent activation energy of NiAu as compared to pure Au catalyst towards the NODH of ethanol, thus indicating higher production rates over the SAA.…”

“…Finally, the efficacy of the ML model is tested for the NODH reaction of ethanol, using E O and E C as descriptors using a previously built ab initio MKM for the estimation of turnovers of ethanol consumption on metal catalysts [36] . The ML model has predicted significantly higher TOFs for the NiAu SAA as compared to the Au surface (Figure 9(b)), with the difference of four order of magnitude.…”

“…With an R 2 of 0.99 and 0.98 for E O and E C , respectively, ML derived data can thus, be used for predicting reactivity. Similar to our recent works, [7,36] we put the ML based E O and E C prediction to test TOF of ethanol NODH from MKM. The reaction scheme for ethanol NODH to acetaldehyde via two possible pathways with either initial α-CÀ H or initial OÀ H activation is shown in Figure 9(a).…”

“…An ab initio MKM is constructed using the toolkit CatMAP, [35] with binding energy descriptor inputs using two different approaches (ML and DFT). An MKM similar to the one developed in our previous work has been used in this study [36] . The model incorporates the predominant routes of initial α‐C−H abstraction and O−H bond scission on ethanol to produce acetaldehyde.…”