Comparison of computational chemistry methods for the discovery of quinone-based electroactive compounds for energy storage

Abstract

High-throughput computational screening (HTCS) is a powerful approach for the rational and time-efficient design of electroactive compounds. The effectiveness of HTCS is dependent on accuracy and speed at which the performance descriptors can be estimated for possibly millions of candidate compounds. Here, a systematic evaluation of computational methods, including force field (FF), semi-empirical quantum mechanics (SEQM), density functional based tight binding (DFTB), and density functional theory (DFT), is performed on the basis of their accuracy in predicting the redox potentials of redox-active organic compounds. Geometry optimizations at low-level theories followed by single point energy (SPE) DFT calculations that include an implicit solvation model are found to offer equipollent accuracy as the high-level DFT methods, albeit at significantly lower computational costs. Effects of implicit solvation on molecular geometries and SPEs, and their overall effects on the prediction accuracy of redox potentials are analyzed in view of computational cost versus prediction accuracy, which outlines the best choice of methods corresponding to a desired level of accuracy. The modular computational approach is applicable for accelerating the virtual studies on functional quinones and the respective discovery of candidate compounds for energy storage.

A quantitative evaluation of computational methods to accelerate the study of alloxazine-derived electroactive compounds for energy storage

“…Given the fact this scheme is also computationally much more demanding, there is no advantage of using it any further. These findings are in accordance with those of the quinone molecules 33 . An overall conclusion is that at the GGA-DFT level (PBE, BLYP), it is possible to use as a descriptor to predict for alloxazines within a range of common experimental errors (i.e.…”

“…When compared under the same set of approximations, it is observed that: When using in scheme (A), LDA g (R 2 = 0.801, RMSE = 0.035 V) and followed closely by PBE g (R 2 = 0.756, RMSE = 0.039 V) are the two best performing methods. Inclusion of higher order exchange effects and parametrizations, such as in HSE06 and M08-HX, are found to have no positive effect on the prediction accuracies, which is in clear contrast to the case of quinones that have been reported earlier 33 . With in scheme (B) (i.e., a hybrid scheme), PBE s (R 2 = 0.959, RMSE = 0.016 V) emerges as the best performing method and shows a significant improvement over the gas-phase only scheme.…”

“…Additionally, there is a positive effect on the prediction accuracy due to the inclusion of implicit solvation during the calculations of and . The positive effect is also observed when using of the reactant as a descriptor, which indicates that the interactions of the reactant’s aromatic rings and H-bonds with the surrounding medium are quite influential, as was also observed in the case of quinone-based molecules 33 . The relative performance of the two descriptors doesn’t change when MAE is used as the metric, as can be seen at the end of Tables S1 a,b.…”

“…Furthermore, the ranking of the descriptors considered here are in stark contrast to the descriptors for quinones (Fig. 2 b) 33 . For quinones, (R 2 = 0.977, RMSE = 0.051 V) is clearly the best descriptor across all the three computational schemes, followed by (R 2 = 0.779, RMSE = 0.158 V) and (R 2 = 0.748, RMSE = 0.168 V).…”

“…In our recent work on prediction of redox potentials for quinone molecules 33 , we systematically evaluated different computational methods and showed that molecular geometry optimization (OPT) with a low-level computational method followed by DFT calculation of the single point energy (SPE) with implicit aqueous solvation offers an equipollent accuracy as the high-level DFT methods, albeit at significantly (~ 10 3 ) lower computational costs 33 . To the best of our knowledge, an analysis of the effect of various factors on the accuracy for predicting redox potentials for alloxazines, such as the level of theory for OPT, the level of theory for the calculation of SPE, and also together with an inclusion/exclusion of solvation effects, has not yet been performed.…”