Water exhibits some unusual and unique properties, especially in response to changes in temperature. For example, most materials shrink on cooling, but water expands — yet the precise mechanism for such anomalous behavior is still not known. Now Masakazu Matsumoto at Nagoya University in Japan1 might have solved this mystery.

The thermal anomaly of water is often explained in terms of the presence of two different structural arrangements of its molecules. The structure of liquid water is amorphous — the molecules do not exhibit long-range order. But over short distances, it is possible for some order to emerge for short periods of time between localized groups of water molecules. Moreover, it is possible for such order to exhibit specific structural arrangements — behavior known as polyamorphism.

The mystery of water expansion with decreasing temperature may finally have an answer.

The expansion of water when cooled is commonly explained as due to the formation of a mixture of two different amorphous phases: a high-density, disordered, mobile phase, and a low-density, ordered, immobile phase. The latter phase dominates as the temperature decreases below a certain point.

Yet there is a problem with this explanation. Despite the plausibility of the low-density ordered phase, its microscopic structure, and even its existence, has not been confirmed experimentally. And even in simulations that allow the structure of water to be studied much more easily and in greater detail than in experiments, no such phase has ever been identified.

Matsumoto believes he knows why. Several years ago, in simulations of the dynamics of water as a network of hydrogen bonds, he noticed the emergence of ordered phases but not always in a low-density state. At the time he ignored this observation as it was not important to the dynamics he was studying. But more recently he revisited the issue.

In doing so, he again observed the presence of different locally ordered structures, but found that these had little effect on the overall density of the system. However, he also found that while the lengths of the network's hydrogen bonds contract with decreasing temperature — as in most materials — the angles between the bonds distort to compensate, causing the density to decrease. This behavior is consistent with the anomalous expansion of water observed experimentally.