Glass transition dynamics in water and other tetrahedral liquids: ‘order–disorder’ transitions versus ‘normal’ glass transitions

We review some aspects of the confusion concerning the glass transition in water, and then show that it must be of a quite different character from that in other molecular liquids, and in fact is the kinetically controlled part of a classical order–disorder transition. (This is the conclusion reached in a review of the low temperature phenomenology of amorphous water currently being published in Science by the present author. Material that would normally appear in the present abbreviated paper will appear in the Science article to which the interested reader is referred.) We do this using a combination of (i) thermodynamic reasoning for ‘bulk’ water (based on known properties of supercooled water and nearly glassy water), and (ii) direct measurements on nanoscopic (non-crystallizing) water. Both require the heat capacity to be sharply peaked near 220 K and thus to imply the existence of a ‘strong-tofragile’ transition during heating. Both require the excess heat capacity to drop to near-vanishing values in the vicinity of 130–150 K. The similarity to order–disorder transitions in crystalline solids is noted, the relation to the second critical point scenario for water is discussed, and the modelling of the anomaly by current theories is considered. Finally we argue that water, with its fragileto-strong liquid transition below the melting point, links (lies in between) the extremes of classical network liquids (where this transition occurs only above the experimentally accessible range) and fragile molecular liquids, where the fragile-to-strong transition is pushed beneath the glass temperature. There has been much confusion in the literature about the glass transition in water [1]. Some argue that it may never have been observed [2–4] while others argue, on a variety of grounds, that it lies at about 136 K [5–8]. The strongest argument for it not occurring at 136 K would seem to lie in the great difference between the behaviour of its dielectric loss, observed at 10 kHz [9], during steady heating over the temperature range 120 K–Tcryst, and the 0953-8984/07/205112+06$30.00 © 2007 IOP Publishing Ltd Printed in the UK 1 J. Phys.: Condens. Matter 19 (2007) 205112 C A Angell Figure 1. Dielectric loss curves for amorphous water from [9] compared with the same for liquidcooled hydrogen peroxide and hydrazine solutions which have ‘normal’ glass transitions in the range 136–140 K. Also shown is the loss curve predicted for a liquid with similar dielectric strength and glass temperature, but with Arrhenius relaxation times and exponential relaxation function (ideal ‘strong’ liquid). (From [10] by permission of the American Institute of Physics.) dielectric loss (under identical conditions) of aqueous H2O2 or N2H4 solutions [10] that have unambiguous glass transitions at the same temperature, 140 K. The key diagram comparing the respective dielectric losses is reproduced in figure 1. The aqueous solutions begin to show loss at around 130 K, but the water shows no comparable loss before it crystallizes suddenly at 155 K. However, reference [10] pointed out one way in which water could be so different and still have a glass transition near 136 K. It is that water, rather than being a fragile liquid in this temperature range [11, 12] or even an intermediate liquid, like the solutions of figure 1, is actually an ideally strong liquid with Arrhenius temperature dependence of the relaxation time, and exponential relaxation function. In this case, as shown by the dotted curve in figure 1 [10], the dielectric loss at 10 kHz could remain very small down to 155 K even when the relaxation time is 100 s at 136 K. But in this case, knowing the highly fragile nature of water near its freezing point, there must have been a fragile-to-strong liquid state transition in the temperature range around 220 K, as has been argued in a number of papers dating from 1993 [13–17]. We return to this question shortly.