Accuracy and Reproducibility of Host-Guest Interactions in Zeolites

Molecular modeling plays an important role in the discovery of organic structure-directing agents (OSDAs) for zeolites. By quantifying the intensity of host-guest interactions, it is possible to select cost-effective molecules that maximize binding towards a given zeolite framework. Over the last decades, a variety of methods and levels of theory have been used to calculate these binding energies. Nevertheless, no benchmark examining these calculation strategies has been reported. In this work, we compare binding affinities from density functional theory (DFT) and force field calculations for 272 zeolite-OSDA pairs obtained from static and time-averaged simulations. We show that binding energies from the frozen pose method correlate best with DFT time-averaged energies. They are also less sensitive to the choice of initial lattice parameters and optimization algorithms, as well as less computationally expensive. Furthermore, we demonstrate that a broader exploration of the conformation space from molecular dynamics simulations does not provide significant improvements in binding energy trends over single-point calculations. The code and benchmark data are open-sourced and together with the reported results, provide robust, reproducible, and computationally-efficient guidelines to calculating binding energies in zeolite-OSDA pairs.