| With the massive consumption of fossil energy,the world is currently facing two major problems: energy exhaustion and environmental pollution caused by fossil fuels combustion.Hydrogen is a recognized clean energy with high energy density and abundant sources.It is considered a promising substitute for fossil fuels,and its products will not pollute the environment.At present,hydrogen production by water electrolysis is expected to be the best way to achieve green industrial hydrogen production,and the development of low-cost and high-efficiency electrolytic water catalysts is an important factor to improve the efficiency of water splitting.This work mainly uses density functional theory calculation(Density functional theory,DFT)and machine learning methods to explore the potential hydrogen evolution catalysts of transition metal doped NiO.Based on the results of DFT calculation and prediction,Ru-NiO was prepared to test the catalytic performance of electrolytic water and reduce the energy consumption in the process of hydrogen production.The main research contents of this paper are as follows:(1)Based on the integration of DFT theoretical calculation and machine learning,the potential transition metal doped NiO hydrogen evolution catalysts were systematically screened,revealing the main factors affecting the catalytic performance of the material system.The H* adsorption energy of twelve common transition metal doped NiO(located at 3d,4d and 5d respectively)is calculated by the first principle,and a comprehensive study of their hydrogen evolution performance is carried out.A linear and non-linear prediction model was established based on the set of feature descriptors retained after screening the physical and chemical properties of 17 simple elements.The predictive equation and feature importance analysis of the data-driven linear model reveal that the standard sublimation enthalpy of doped atoms,the s orbital of the Zunger-Cohen orbital radius,and the p orbital of the Waber-Cromer orbital radius have a strong correlation with the catalyst activity.The 1-MAPE and RMSE of the model are respectively 84.1% and 0.2190 e V.In general,a reasonable non-linear learning model requires at least 50 data points.Therefore,the support set is applied to make the convolutional neural network suitable for small sample training.The 1-MAPE of training set and test set reached more than 97%,indicating that the constructed model has very excellent prediction performance.This method greatly shortens the time required for DFT highthroughput calculation,and provides a certain data reference for catalyst design and screening.(2)Based on the results of DFT theoretical calculations and model predictions,an optimized strategy of cation doping was used to prepare a highly efficient hydrogen evolution reaction(Hydrogen evolution reaction,HER)and oxygen evolution reaction(Oxygen evolution reaction,OER)bifunctional catalytic nanomaterial Ru-NiO while reducing the cost.Ni(OH)2 is annealed to form NiO,and then two-dimensional nano chip arrays with uniform doping distribution of Ru were prepared by wet etching and re annealing.The introduction of dopant atoms adjusts the active sites for material dissociation in the alkaline solution,improves the conductivity of the material,and accelerates the charge transfer process.The reaction barrier for the absorption/desorption of intermediates is optimized during the catalytic reaction.At the same time,doping promotes the formation of oxygen vacancies,which improves the activation ability of the material.The catalyst requires 25 m V and 234 m V respectively to reach a driving current density of 10 m A cm-2 in the reaction process of HER and OER,exhibiting excellent dual-functional catalytic performance.It is used as the cathode and anode of electrolytic water to assemble the total hydrolysis system.Only 1.53 V(vs.RHE)driving potential is needed to reach the same current density,which is better than the reference catalytic electrode pair Pt/C || Ru O2 and Pt/C || Ir O2. |
PDF Full Text Request