Assessment of density functional theory to calculate the phase transition pressure of ice

To assess the accuracy of density functional theory (DFT) methods in describing hydrogen bonding in condensed phases, we benchmarked their performance in describing phase transitions among different phases of ice. We performed DFT calculations of ice for phases Ih, II, III, VI and VII using BLYP, PW91, PBE, PBE-D, PBEsol, B3LYP, PBE0, and PBE0-D, and compared the calculated phase transition pressures between Ih-II, Ih-III, II-VI, and VI-VII with the 0 K experimental values of Whalley [J. Chem. Phys., 1984, 81, 4087]. From the geometry optimization of many different candidates, we found that the most stable proton orientation as well as the phase transition pressure does not show much functional dependence for the generalized gradient approximation and hybrid functionals. Although all these methods overestimated the phase transition pressure, the addition of van der Waals (vdW) correction using PBE-D and PBE0-D reduced the transition pressure and improved the agreement for Ih-II. On the other hand, energy ordering between VI and VII reversed and gave an unphysical negative transition pressure. Binding energy profiles of a few conformations of water dimers were calculated to understand the improvement for certain transitions and failures for others with the vdW correction. We conclude that vdW dispersion forces must be considered to accurately describe the hydrogen bond in many different phases of ice, but the simple addition of the R-6 term with a small basis set tends to over stabilize certain geometries giving unphysical ordering in the high density phases.