Glasses exhibit vibrational and thermal properties that are markedly different from those of crystals. While recent works have advanced our understanding of vibrational excitations in glasses in the harmonic approximation limit, efforts in understanding finite-temperature anharmonic processes have been limited. In crystals, phonon–phonon coupling provides an extremely efficient mechanism for anharmonic decay that is also important in glasses. By using extensive molecular dynamics simulation of model atomic systems, here we first describe, both numerically and analytically, the anharmonic couplings in the crystal and the glass by focusing on the temperature dependence of the associated decay rates. Next, we show that an additional anharmonic channel of different origin emerges in the amorphous case, which induces unconventional intermittent rearrangements of particles. We have found that thermal vibrations in glasses trigger transitions among numerous different local minima of the energy landscape, which, however, are located within the same wide (meta)basin. These processes generate motions that are different from both diffusive and out-of-equilibrium aging dynamics. We suggest that (i) the observed intermittent rearrangements accompanying thermal fluctuations are crucial features distinguishing glasses from crystals and (ii) they can be considered as relics of the liquid state that survive the complete dynamic arrest taking place at the glass transition temperature.
REFERENCES
Defects would emerge in crystals closer to the melting point Tm. However, we did not consider such a case here, where studied temperatures were always well below Tm.
The transverse modes follow a and dominate over the longitudinal excitations in the low frequency regime.
This system-size dependence is not due to statistics. For instance, at N = 32 000 and T = 10−2, rearrangements occur intermittently with very high frequency, as shown in Fig. 5(a). In contrast, we never observed any (possibly quite rare) rearrangement at the smaller N = 4000 (and the same T = 10−2), even monitoring three independent system instances.