Fig. 2: The design of Ru-based intermetallic catalysts.
From: Machine learning-assisted Ru-N bond regulation for ammonia synthesis
a Heat map of Pearson’s correlation coefficient matrix in features which were employed in machine learning models. b Violin plots of the error from true to predicting N2 and N adsorption energies through XGB, MLP, KNN, GBDT, and SVR models. c The regression curve of training data and testing data by XGB model. d A two-dimensional activity volcano plot for ammonia synthesis. N2 and N adsorption energies in panels were calculated using DFT. The gray square points are selected Ru-IMCs to train the model; blue cycle points are predicted Ru-IMCs through the model; and asterisk represents the un-explored Ru-IMCs candidate for ammonia synthesis. e The beeswarm plot of feature importance in XGB models using SHAP assessment. f The average absolute SHAP values of each input feature.
