The influence of resistance drift on measurements of the activation energy of conduction for phase-change material in random access memory line cells

Abstract

Temporal drift of the amorphous resistance in phase-change random access memory (PRAM) is a temperature accelerated process. Increasing the temperature will speed up the drift process which is shown to affect measurements of the activation energy of conduction (Ea, slope of log(R) versus 1/kT). Doped SbTe phase change (PRAM) line cells were brought to the amorphous state and were subjected to annealing experiments. First, it is shown that when the temperature is increased by a fixed rate, the resistance does not follow a unique function of temperature but depends on the heating rate. This can be attributed to resistance drift taking place during the ramp. Upon cooling, the drift process freezes and only then physically relevant, i.e., time independent, values for Ea can be obtained, because of the absence of additional drift. The observed increase in resistance as a function of annealing history (for various frozen-in drift levels) is modeled and well-reproduced using a trap limited band transport model. The model explains these observations by an increase of the temperature dependent band gap by about 47 meV due to drift at 418 K. V

Role of activation energy in resistance drift of amorphous phase change materials

“…In contrast to that, temperature dependent resistance measurements performed on cells of the line-cell type based on doped Sb 2 Te [32] yielded the result that the increase of resistance observed on the corresponding RESET state could only be explained by an increase of both the prefactor and the activation energy. Unfortunately, the composition of the material in these phase change memory cells investigated was not made public.…”

“…All these artifacts can change from one device architecture to another. The different observations of the temporal evolution of E A and R * reported in Boniardi et al [31] and Oosthoek et al [32] can either be related to the above mentioned artifacts in nano-scaled devices or to intrinsic material properties. It is therefore desirable to exclusively measure the pure material properties of the amorphous phase change material of interest.…”

“…Instead, a general model for drift in phase change materials has to take into account a time dependence of both E A and R * . A possible explanation for these experimental findings could be a change of the temperature dependence of the band gap as proposed in Oosthoek et al [32]. To validate this hypothesis, however, further experimental evidence is needed, e.g., temperature and time dependent measurements of the optical band gap.…”

“…In a previous work [ According to the work of Oosthoek et al a physically relevant measure of the activation energy can only be taken during a temperature cooling ramp, when the thermally activated drift effect is strongly hindered [32]. Therefore, the annealing process at a constant temperature of 80 • C was repeatedly interrupted after 5 min by a temperature dip (exemplarily shown for GeTe in Figure 6).…”

“…By choosing the materials Ge 2 Sb 2 Te 5 , and AgIn-doped Sb 2 Te (more precisely Ag 4 In 3 Sb 67 Te 26 ) we furthermore aim at studying the apparent contradiction between the two aforementioned experiments [31,32]. The third material, GeTe, is studied due to its promising switching characteristics [5].…”