Efficient Prediction of Redox Potentials of First-Row Transition Metal Complexes for Aqueous Redox Targeting Flow Batteries Using Density Functional Theory

Abstract

Redox targeting flow batteries are an emerging alternative to the traditional redox flow battery architecture which offer improved energy density via an added electroactive solid ‘booster’. For an efficient system, it is imperative that the redox potentials of the soluble redox mediator and the solid match. For this purpose, transition metal complexes are a promising class of redox mediators due to the tunability of their solubility and electrochemical potential, and the ability to take sustainability into account. As such, reliable and time-efficient prediction tools for the redox potentials of transition metal complexes are needed to help with the search for new potential systems. While density functional theory usually provides a good trade-off between computational cost and accuracy, calculations on transition metal complexes often result in large errors. Thus, they have been traditionally used for studying trends of the ligand effect rather than providing accurate predictions. In this communication, we compare redox potential predictions from different solvation methods and levels of theory using an initial experimental data set based on aqueous iron complexes with bidentate ligands. Good predictions (MAE= 0.23 V) are obtained using the COSMO-RS solvation model which is not universally available in DFT software. When using the more common implicit solvation models, there is a general deviation from experimental results, irrespective of the level of theory used. We propose that these errors are conveniently corrected by using simple linear regression using the experimental data as a training set, which yields a good prediction of redox potentials (MAE= 0.09 V). The predictive power of this method is also demonstrated by cross-validation using complexes with different metal centers and ligand types. Figure 1