Reverse Design Of Methanol Electro-oxidation Catalysts And Their Performance In Direct Methanol Fuel Cells

With the excessive use of traditional fossil fuels,environmental pollution,climate change and other problems are caused.Coupled with the rapid growth of the global population,it is imperative to explore renewable and sustainable energy devices.Direct methanol fuel cells（DMFCs）,which use proton exchange membrane as electrolyte and methanol of high-energy density as fuel,are a highly promising portable energy due to the advantages of environmentally friendliness,better safety,simplified systems,low cost,feasible and convenient fuel storage.Pt/Pt Ru catalysts are the most commonly used anode catalysts in DMFCs.Unfortunately,compared with other noble metals,the resource of Pt is very scarce.At present,the DMFC power density is 0.2-0.25 W·cm^-2 under the condition of 2-4 mg·cm^-2 catalyst loading for anode,which can barely meet the practical application requirements.In order to predict and develop DMFC anode catalysts with high activity and low cost,reverse design for methanol oxidation reaction（MOR）electrocatalysis was conducted combining machine learning,first principles calculation and experimental verification.The specific research content is summarized as follows:（1）Ternary alloy catalysts with high MOR performance were predicted by machine learning.The MOR specific activity prediction of catalysts was used as a regression problem,and a MOR database with 684 experimental data was established for constructing and training models.Seven machine learning regression algorithms were selected to train and test the MOR database,and extra trees regression and random forest regression were determined as the best algorithms by comparing the prediction accuracy.The top 20 feature importance descriptors were selected by cross validation.Finally,through the algorithm configuration based on the sequence model,500,000 groups of data were screened by using extra trees regression and random forest regression models,and the Pt Ru M ternary alloy electro-catalysts with high dispersion and MOR activity in theory was successfully obtained.（2）The catalytic activity of methanol electro-catalysts obtained from machine learning results was analyzed using first principles calculation.It was found that in the constructed MOR volcano plot,in Pt Ru M,Pt Ru Pd had stronger adsorption capacity for OH and weaker adsorption capacity for CO.By calculating the MOR full path free energy step diagram of Pt（111）,Pt Ru（111）and Pt Ru Pd（111）,it is found that the Pt Ru Pd（111）surface has the lowest potential determination step,indicating stronger resistance to CO poisoning.Based on the difference of projected density of states and electron density of CO adsorption on the surfaces of three different catalysts,the results show that the addition of Ru and Pd can optimize the electrochemical activity of the catalyst surface and the electronic structure of Pt.（3）The reliability of machine learning and first principles calculation was verified by preparing nanoporous Pt Ru Pd ternary catalysts with atomic ratios similar to machine learning results.Nanoporous Pt Ru Pd catalysts were successfully prepared by a one-step synthesis without surfactants.The MOR specific activity of Pt Ru Pd was about 6.36 m A·cm^-2,which was 9.6 and 6times higher than that of commercial Pt/C and Pt Ru/C catalysts.At the same time,Pt Ru Pd catalysts showed lower peak potential,higher stability and stronger anti-CO poisoning ability.Direct methanol single-cell tests indicated that the peak power density and mass specific power density of free-standing nanoporous Pt Ru Pd catalysts were 63.6 m W·cm^-2 and 106 W·g^-1_{Pt Pd} at 80℃,which were higher than most Pt-based anode catalysts.Stability measurements also proved that the nanoporous Pt Ru Pd catalysts exhibited excellent stability and durability.