Atomistic Simulations of Crystal Nucleation and Growth

The key property that makes phase change materials attractive for applications in nonvolatile memories is the fast crystallization which allows for a full crystallization in PCM devices on the time scale of 10-100 ns. The fact that both nucleation rate and crystal growth velocity are very large has stimulated direct simulations of the crystallization process by DFT-MD [13-16]. Simulations with up to 180-atom cell and periodic boundary conditions shed light on the atomistic mechanism of formation of the crystalline nucleus and on the role of four-membered rings as seeding structures for nucleation [13-16]. 

Since the bonding network of both the crystalline and amorphous phases mostly consists of four-membered rings, it has been suggested that the crystallization should require few bonds breaking which would make the transformation particularly fast [13-15]. This has to be contrasted with the behavior of a strong glass former such as 

The free energy difference between the crystal and the liquid actually depends on temperature as 

The self-diffusion coefticient in 15.3 and 15.4 is customarily expressed in terms of the viscosity rj by using the Stokes-Einstein relation (SER) D = breakdown 

The Thomson-Spaepen approximation (15.6) was used for Afi with AH = 0.186 eV/atom and T = 1023K (exp. T = 998 K, [118]) for the NN potential obtained from thermodynamic integration in [68]. The Thomson-Spaepen approximation might appear as a crude approximation in view of DFT estimates of the temperature dependence of Cp for the crystalline and amorphous phases by Liu et al. (see Fig. 15.4 of [108]). However, in the temperature range 500-700K, the thermodynamic factor has actually a little effect on the crystal growth velocity compared to 

---