Using Machine Learning to Predict Oxygen Evolution Activity for Transition Metal Hydroxide Electrocatalysts

Abstract

Electrocatalytic water splitting is an attractive way to generate hydrogen and oxygen for obtaining clean energy. Oxygen evolution reaction (OER), as one of the half reactions of oxygen evolution, is kinetically unfavorable involving the transfer of four electrons. Hydroxides are promising candidates for efficient OER electrocatalysts toward water splitting because of their high intrinsic activity and active surface area. However, quantitative prediction of hydroxide electrocatalytic performances from high-dimensional component spaces remains a challenge, severely hindering the performance-oriented precise composition and process design. Herein, we introduce a machine learning-based OER activity prediction method for hydroxide catalysts under extensive doping space for the first time. The relationship among composition, morphology, phase, pH value of the electrolyte, type of the working electrode, and overpotential was successfully fitted by the random forest algorithm. The model shows a good precision on the forecast of new experiments with a mean relative error of 4.74%. Furthermore, a new high-activity hydroxide catalyst Ni 0.77 Fe 0.13 La 0.1 was rationally designed and experimentally prepared, showing an ultra-low OP of 226 mV for a current density of 10 mA cm −2 . This work provides an effective and novel way for hydroxide electrocatalyst prediction, which can further enhance the electrocatalyst design toward high catalytic performance.

Analysis of the oxygen evolution activity of layered double hydroxides (LDHs) using machine learning guidance

“…During the vector construction, we observed many null values in the final matrix since the most common LDH metals in our set were Ni, Co, and Fe, and only sometimes other transition metals. Since abundant null values might affect the final model performance, 32 we compared the models directly applying atomic numbers as inputs and using binary notation (zero and one to represent the type of elements, and the result can be seen in Fig. S2, ESI †).…”

“…To overcome this problem, additional information should be involved as extra features to compensate for data missing. Herein, inspired by other works, 28,32,36 which reveal the importance of measurement or reaction conditions in feature ranking (almost always rank 1st among all features), we added measurement conditions for overpotentials (e.g., scan rates for linear sweep voltammetry and iR-correlation information) as extra features to construct new ML models. After these features were added, the overpotential prediction accuracy was successfully improved with the highest R 2 of 0.768 (see Fig.…”

“…However, this study did not elaborate on the specifics of LDHs. 32 Additionally, that model directly used atomic numbers as inputs rather than converting into a traditional 0 and 1 form (namely binary notation or binary representation), which, in our opinion, might not contribute to good results to predict the overpotentials. Moreover, overpotential values depend on measurement conditions, impedance, and the number of active sites.…”

Explainable machine-learning predictions for catalysts in CO2-assisted propane oxidative dehydrogenation

“…Similar applications of machine learning on oxygen evolution reaction and chemical looping oxidative dehydrogenation of propane are also reported. 34,35 To the best of our knowledge, the application of machine learning in CO 2 -ODHP for catalysis analysis has still not been reported. In the present work, the statistical correlations between catalyst composition, reaction parameters, and catalytic performance of CO 2 -ODHP reaction from previous literature studies are established based on a series of mathematical models, which were constructed using the articial neural network (ANN), support vector regressor (SVR), random forest regressor (RF) and k-nearest neighbor (KNN) algorithms.…”

Comparing the Catalytic Effect of Metals for Energetic Materials: Machine Learning Prediction of Adsorption Energies on Metals

“…However, DFT calculation is time-consuming; machine learning (ML) provides a new way to accelerate this process. At present, the ML algorithm, which can be used to predict the d-band structure or adsorption energy, is widely used in the rational design of catalysts. − Roy et al predicted the adsorption energies of important intermediates on surface microstructures for finding selective earth-abundant high-entropy alloy-based catalysts for CO 2 to methanol formation using an ML algorithm. Toyao developed a simple and efficient ML model to predict the adsorption energies of CH 4 , CH 3 , CH 2 , CH, C, and H on Cu-based alloys for effective utilization of methane.…”