Surface defreezing of glasses

A glass surface may still flow below the bulk glass transition temperature, where the underlying bulk is frozen. Assuming the existence at T=T* of a bulk thermodynamical glass transition, we show that the glass-vapor interface is generally wetted by a liquid layer of thickness ~ -ln(T*-T) when T--> T*. Contrary to standard surface melting of crystals however, the integrated value of the diffusivity across the interface remains finite for T-->T*. Difference in shape induced by bulk and by surface flow is discussed as a possible means of experimental detection of surface defreezing.