Enhancing Empirical Energy Functions Using Physics- and Machine Learning-Based Extensions: Structure, Dynamics and Spectroscopy of Modified Benzenes

The effects of replacing individual contributions to an empirical energy function are assessed for halogenated benzenes (X-Bz, X = H, F, Cl, Br) and chlorinated phenols (Cl-PhOH). Introducing electrostatic models based on minimal distributed charges (MDCM) instead of usual atom-centered point charges (PCs) to realistically describe features such as σ-holes yields overestimated hydration free energies unless the van der Waals parameters are reparametrized. Scaling van der Waals ranges by 10% to 20% for three Cl-PhOH and most X-Bz yield results within experimental error bars, which is encouraging, whereas for benzene (H-Bz) PC-based models are sufficient. Replacing the bonded terms by a neural network-trained energy function featuring fluctuating atom-centered PCs also yields qualitatively correct hydration free energies, which can be brought into agreement with experiments within error bars after adaptation of the van der Waals parameters. The infrared spectroscopy of Cl-PhOH is rather well captured by all models, although the ML-based energy function performs somewhat better in the region of the framework modes. The present work finds that refinements of empirical energy functions for targeted applications are a meaningful way toward more quantitative and physics-based simulations. At the same time, empirical energy functions have matured to a remarkable degree, at least for the species considered in the present work.