DescriptionWater remains one of the most fascinating molecules to study at the theoretical level, not least due to its well-documented “anomalies” in terms of dynamic and thermodynamic behaviour. On a practical level, the importance of water in biological and industrial processes means that development of accurate and computationally efficient water models is still an area of intensive research. In this context, fixed-charge (i.e. non-polarisable) models have been the “weapon of choice” for the vast majority of simulation studies to date. Despite recent progress, however, a unique classical non-polarisable model of the water molecule is still to emerge. To better understand why this is the case, one must delve deeper into the intrinsic approximations involved in the development of such models. In this work, we focus on the ability of fixed-charge models to describe the hydration free energy of the water molecule – in other words, the free energy change of bringing one water molecule from the vapour to the pure liquid. Using thermodynamic integration methods, we have calculated the hydration free energies of over 10 different fixed-charge water models and compared them against experimental data. Our calculations show that none of the most recent models is able to simultaneously predict the hydration free energy and the enthalpy of vaporisation of water, and that a systematic error is at the origin of this failure. More precisely, we argue for the need to appropriately account for the polarization cost during the process of transferring a molecule from the vapour to the pure liquid. Such corrections have been proposed in the past, but they fail to account for purely electronic contributions to the polarisation. We show that a recently proposed framework that implicitly accounts for polarisation contributions in non-polarisable models is able, with appropriate adaptations, to reconcile the predictions of free energy and enthalpy for water. This approach relies on an estimate of the dipole moment of liquid water that is in agreement with recent experimental and theoretical estimates. Our work also suggests an explanation for why the rather outdated TIP3P model is still widely used for the prediction of solvation free energies.
|Period||19 Jun 2018|
|Event title||10th Liblice Conference on the Statistical Mechanics of Liquids|
|Location||Srni, Czech Republic|
|Degree of Recognition||International|
Research output: Contribution to journal › Article › peer-review
Project: Research - Studentship