Trends and insights into alloying elements impact on predicted battery voltage in metal-ion batteries

Abstract

In recent years, metal-ion batteries have been the focus of extensive research. A significant challenge in utilizing certain ions, particularly multivalent ions, has been identifying suitable electrode materials. To address this, we developed a machine-learning model using LightGBM to predict the average voltage of metal-ion batteries based on electrode composition in the charged and discharged states. Our model achieved a prediction error of 0.26 V when benchmarked against several experimentally obtained values. Moreover, we provide key trends as to how the addition of alloying elements such as Manganese, Iron, Cobalt, Nickel, and Aluminium in the electrode affects the output voltage. Furthermore, by screening several thousands of novel electrode compositions obtained by alloying these elements, we provide a set of 12 compositions that are predicted to have an average voltage >4.5 V.