Conclusions

The analysis of cavity statistics in water and other molecular liquids is a powerful approach to explaining the hydrophobic effect. The comparison of distributions of cavity sizes in water and hexane reveals that it is less likely to encounter a cavity of atomic or small molecular size in the aqueous phase than in a hydrocarbon phase under the same thermodynamic conditions, even though water has more free volume. As a consequence, non-polar solutes of similar sizes are less soluble in water than in non-polar liquids. As follows from Figure 2 and Eq. 4, the compressive force acting on a cavity of atomic size is much larger in water than in hexane, confirming the conventional picture of the hydrophobic effect, in which water squeezes out non-polar solutes from solution. The scaled particle model that utilizes only information about the density of the liquid is not accurate for predicting how the compressive force changes with the cavity radius. A considerable improvement is obtained if information about the oxygen-oxygen radial distribution function or the distance of the closest approach between oxygen atoms is incorporated into theoretical or computational models. A noteworthy example is a Lennard-Jones reference liquid, which can reproduce several characteristics of the hydrophobic effect, including the inverse temperature dependence of solubilities of small, non-polar solutes. This indicates that specific knowledge about the arrangement of hydrogen bonds between water molecules around these solutes in not needed to capture the main features of the hydrophobic effect. Instead, the importance of hydrogen bonds is manifested by influencing bulk properties of water, such as the density and the radial distribution functions.

The idea that knowledge of the density and the radial distribution functions is essential for the correct description of the hydrophobic effect has been carried over to an information theory model of the hydrophobic effect, which allows for calculating the excess chemical potential of cavities of molecular sizes and arbitrary shapes. As in the case of small cavities, the model requires only results of simulations of the neat liquid. Even though the model utilizes only the first two moments of the density expansion, it appears to be quite accurate for small and medium-size molecular solutes. In the present form, it can be applied to such problems of broad interest as energetic effects of point mutations in proteins involving non-polar residues and interactions between receptors and hydrophobic ligands. It also has shed interesting light on hydration effects influencing protein stability, such as pressure denaturation of proteins [41] and convergence of the entropy of transfer. [42] The extension of this model to include in a simple manner the effect of dewetting large hydrophobic surfaces, e.g. in proteins remains a challenge.

Acknowledgments

This work was supported by a grant from the NASA Exobiology Program. I thank Michael H. New and Michael A. Wilson for helpful discussions on the manuscript.