Transition metal release from nanomaterial surfaces has negative impacts on aquatic organisms due to toxic metals such as nickel and cobalt ions in solution. However, there are also positive impacts such as micro-nutrient release from copper oxide to plants to improve immune response. In previous studies we modeled battery cathode materials and found that metal release trends are largely controlled by the metal oxidation states. By tuning the compositional phase space we can induce redox events which can be used to control release properties at the nanomaterial surface. With a large dataset of structural information and release favorability values from varied compositions of Li(Ni1/3Mn1/3Co1/3)O2 and other formulations containing other redox active transition metals, a machine learning model can be trained to determine which structural features are most important for controlling metal dissolution. In addition to structural features, this model will also be able to predict release favorability of metals from entirely new compositions. In order to train a machine learning model, there needs to be enough unique structures optimized to predict metal release trends in different chemical environments. The computational cost increases greatly with the level of functional used as well as the system size. Additionally, the aqueous environment also adds to the computational complexity, whether modeling the environment with implicit or explicit solvent. In our calculations, we employ the DFT + Solvent Ion model where we perform vacancy formation calculations using Density Functional Theory (DFT) and incorporate experimentally available data of the solvation of the standard state atoms to their aqueous ions in reference to the standard hydrogen electrode (SHE). With this approach we have screened a large set of cathode formulations and their compositional phase space to build our dataset to train a machine learning model for metal release trends.

Training a machine learning model to predict metal release from complex metal oxide surfaces

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.