A theory of field-induced crystal nucleation is developed and verified experimentally for the case of switching in nanoglasses of phase change memory. For symmetry-breaking strong electric fields, it predicts needleshaped crystallites with nucleation barriers lower than that of spherical nuclei and a strong field dependent. We have observed bias dependent switching for times and temperatures far beyond those typically reported and supportive of our predictions, in particular, switching time exponential in voltage and temperature.

The authors propose a simple physical model of threshold switching in phase change memory cells based on the field induced nucleation of conductive cylindrical crystallites. The model is solved analytically and leads to a number of predictions including correlations between the threshold voltage Vth and material parameters, such as the nucleation barrier and radius, amorphous layer thickness, as well as Vth versus temperature and switching delay time. The authors have carried out verifying experiments, and good agreement is achieved.

We present the data on temporal (t) drift of parameters in chalcogenide phase change memory that significantly complement the earlier published results. The threshold voltage Vth and the amorphous state resistance R are shown to drift as ΔVth∝v ln t and R∝tα in broad intervals spanning up to nine decades in time; the drift coefficient v depends on glass parameters and temperature, but does not depend on device thickness. We have demonstrated that drift saturates at long enough times that can be shorten with temperature increase. All available data on drift dynamics are fully consistent with the classical double-well-potential model, which gives simple analytical expressions for the observed temporal dependencies including numerical parameters.