A reexamination of the ice III/IX hydrogen bond ordering phase transition

Ice III is a hydrogen bond disordered crystal which when cooled 1 K / min or faster transforms to an antiferroelectric hydrogen bond ordered structure, ice IX. Throughout its region of stability, experiments indicate that the H bonds in ice III are, in fact, partially ordered, i.e., some proton arrangements are preferred. In addition, there has been evidence that the structure of ice IX retains some residual disorder after the transition. Diffraction experiments and calorimetry apparently conflict with regard to the degree of ordering at the ice III/IX transition. Mean field statistical mechanical theories have been used to link partial occupations from diffraction data with thermodynamics. In this work, we investigate the ice III/IX proton ordering phase transition using electronic density functional theory calculations for small unit cells, extended to simulate the phase transition in a large unit cell using graph invariants. In agreement with experiment, we observe partial ordering over a wide range of temperatures as ice III transforms to partially disordered ice IX, near 126 K, which becomes fully ordered at lower temperatures. We compare our results from full statistical mechanical simulations with mean field models, finding small errors for the low-temperature ice IX phase and much larger errors for the high-temperature ice III phase. The failure of mean field theories may explain the apparent conflict between diffraction experiments and calorimetry.