The history of glass-making can be traced back thousands of years. One of the main reasons that craftspeople have been able to manipulate glass is that, as a glass-forming liquid is supercooled to its glass transition temperature, it grows much more viscous. However, scientists are still uncertain about why this happens.

Now, Hajime Tanaka and co-workers at the University of Tokyo in Japan have revealed that this change in viscosity may have thermodynamic origins1. In fact, the process bears remarkable similarities to critical phase changes such as magnetic transition or condensation.

Upon nearing the glass transition, the dynamics of a glass-forming liquid slow down and become more heterogeneous, or unevenly distributed. Tanaka and his co-workers have now observed spatial fluctuations similar to those that occur when approaching the gas-to-liquid transition, known as the condensation critical point.

At critical points, the large-scale fluctuations in the gas or liquid slow down much more than the small-scale motions. This type of hierarchical dynamic structure has not been seen in glass transitions, where it is thought that motions at all sizes, even at the molecular or atomic scale, may slow down.

Most puzzlingly, the overall static structure — or ‘positional correlation’ — of the glass-forming liquid does not seem to change near the glass transition, even though the dynamics change significantly. As Tanaka and his co-workers state in their article: “This seems to suggest that exotic new physics is necessary to solve this issue.”

Fig. 1: A snapshot from a Brownian dynamics simulation of a supercooled liquid, showing the development of medium-range structural order.© 2010 T. Kawasaki, H. Tanaka

The researchers undertook a numerical and experimental study of various glass-forming liquids. Contrary to previous expectations, they found that, although the positional correlation did not change, there were some significant fluctuations in static structural order during the process of supercooling (Fig. 1). Furthermore, unlike critical phenomena, spatial heterogeneity seems to cause dynamic heterogeneity in glass transitions.

The researchers also found evidence that the typical lengths and timescales of fluctuations in the liquids diverged on approaching the glass transition — just like at the critical point. However, as Tanaka notes, “the slowing down of dynamics with temperature is far steeper in the glass transition than in critical phenomena. It is this feature that prevents researchers from accessing the transition, if it even exists.”

Taken together, the results suggest that the glass transition may actually be an unusual type of critical phenomenon. “We are now studying a link between structural order and crystal nucleation,” says Tanaka.