Combined Computational and Experimental Approach to Determine and Understand the Solubility of Phenothiazines as Redoxmers

Abstract

Due to their property of decoupled power and capacity, non-aqueous redox flow batteries (NAqRFBs) provide an alternate solution to address the increase needed in grid storage. Phenothiazine (PT) class of derivatives possess desired properties such as high potential and chemical stability for redox-active catholyte materials for NAqRFBs. Nevertheless, most of the currently studied redox-active organic materials (ROMs), including certain phenothiazines suffer from low solubility specifically in their charged states and how the solubility trends change in charged states remains a mystery. Improving the solubility of ROMs without sacrificing stability or the redox potential requires a molecular-level understanding of how solubility changes in ROMs in general or for a specific class of materials. In fact, procurement of a universal understanding and ability to predict molecular solubility would serve the scientific community in well as solubility plays such an important role in almost every sub-discipline in chemistry, physics, engineering and, medicine. Quantitative structure-property relationships (QSPR) are known to be highly efficient due to their simplicity and ease of use. QSPR are mathematical relationships that link chemical structure and a chemical property (i.e. solubility) in a quantitative manner for a series of compounds. Hence, we used an approach of combining experiments and computational simulations with the aid of QSPR with the two major objectives of (1) Numerically predicting solubility of PTs at different states of charge and solution environments and (2) Understanding solubility trends and their physiochemical basis. These predictive models were constructed utilizing numerous experimental and computed molecular descriptors. A database of experimental, electronic structure, geometrical, topological descriptors containing around 1900 descriptors for neutral and radical cation forms of PT derivatives was constructed and the best descriptors were selected using various data science methods. Herein, we were able to establish a computational workflow that is robust and efficient in the optimization of conformationally flexible PT derivatives followed by parameter extraction.The major focus in the QSPR study was on the radical cation systems since the solubility usually suffers when going from neutral to charged state which is undesirable when considering NAqRFBs. QSPR models that can capture the solubility trends and predict the solubility of the validation set of compounds were validated with different statistical validation techniques. The statistical models developed were able to predict the solubility of compounds and were comparable with the experiments. Interestingly, these models were able to predict the solubility of the radical cation form of the compound secBuPT in 0.5 M tetraethylammonium tetrafluoroborate-acetonitrile (TEA/BF4-ACN) and pure acetonitrile (ACN) within a 10% deviation from the experimental value before experimental testing. Moreover, the analysis of molecular descriptors that...