Graph Neural Network Model Accelerates Biomass Adsorption\nEnergy Prediction on Iron-group Hydrotalcite Electrocatalysts

Iron-group\nlayered double hydroxides (LDH) have demonstrated excellent\nbiomass electrooxidation performance. However, the development of\nthese materials relies on extensive experiments and high computational\ncosts. Therefore, we developed a graph neural network (GNN) (named\nGALE-Net 2.0) for predicting the adsorption energies in the electrocatalytic\nreaction of 5-hydroxymethylfurfural (HMF). A data set of the adsorption\nenergies of organic molecules on the LDH was constructed. The GNN\nmodel predicted that the 1:2 CoNi-doped LDH catalyst would demonstrate\nexcellent HMF electrooxidation performance. The calculation time was\nreduced from 24 h with the density functional theory (DFT) calculations\nto 1 h with the GALE-Net 2.0. The mean absolute error of the GNN model\nwas 0.17 eV, which is consistent with the accuracy of the DFT calculations.\nMoreover, the model showed some generality as it successfully predicted\nthe adsorption energy of furan derivatives. Our results suggest that\nGALE-Net 2.0 can accelerate the design of electrocatalysts.