First firing or virgin breakdown
To avoid parallel leakage paths through the amorphous material in their CNT gap structures, the researchers  deposit GST in its amorphous state. In the past, the as-deposited amorphous material was always different from the amorphous material after a quenching from a reset cycle, with the most likely cause different number of traps or dangling bonds, or level of un-annealed stress. This manifested itself in a first switching threshold voltage that was always higher than subsequent switching events. The authors  did not make any comment on this or its implications if and when attempts are made to integrate CNT devices into an array.
CNT “gap” creation raises questions
The need to create the gaps in the carbon nanotubes by electrical means raises a number of questions. It is not feasible to plan to do this for PCM arrays with bit capacity 1 to 4G-bits as part of a process of manufacture or test? A comment
from Professor Eric Pop in EETimes  suggests that the authors of  clearly recognize this to be the case. The authors  indicate that the simplest and low cost way to produce the gaps is by oxidation. This avoids the production of a field of debris, producing only CO2 and heat. Smaller gaps are created when the electrical zapping technique is used for gaps formed in argon. Although some of the micrographs in  show ragged edges, it is claimed that for the most part the electrical zapping technique produces clean gaps. Figure 4 illustrates how the gap would most likely form in a tube with large dimensions and the heat sinking effects of a substrate; what actually happens at the atomic dimensions of a 3nm diameter CNT could be significantly different. If the final part of the gap formation does involve a carbon arc, then it could account for the clean gaps and the sputtered carbon atoms might be able to effect repair to any open bonds at the tube ends. The temperature range of pure carbon oxidizing in air is from 350oC to 600oC, and carbon arcs go higher.
Figure 4: The possible form of a CNT gap with two possible mechanism involved.
Other methods of producing the gaps that have been explored include e-beam lithography and AFM. In the context of low cost array production, a new method of creating the gap will need to be found. However, once such a means is found that uses materials compatible with processing, why stay with carbon nano-tubes, providing, as mentioned earlier low nm diameter contacts do not involve a fundamental problem?
The gap PCM devices described in  would be characterized as high-aspect ratio structures, as described in . That is, they are devices where the electrodes have been moved as far as is possible away from the active area. To create a high-bit density array, there will be a need to reduce the distance (now at 1 to 2um) of the electrodes from the active area. This will bring the cooling effect of the electrodes into play, increasing the reset current and the gap creation voltage and current. The accurate centering of the hot spot and resulting gap in the CNT suggests that the electrodes are already having some cooling effect. It might even be possible to construct cross-point arrays by a tube on tube structure (CNT-GST-CNT) and keep the electrodes more remote.