Ron and Resistion - right on! - I was thinking about making a comment on this whole thing and I really did not know how to start. The table of desirability deserves an article by itself, because it really all that this is all about. We either have a real physical mechanism with controllability and manufacturability, or we have a bunch of "techniques" for oxide breakdown and electrochemical reactions. Specific to the 4DS device, the Mott-like transition is hard to identify because lattice defects change significanly the Crystal Field (or ligand field, also called coordination sphere) such that the change in oxidation number of Mn in a +2 and +3 sequence as in the Giant Magnetoresistance phenomenon, is not clearly understood. It is even more difficult to say if the valence fluctuation leads to a reversible Mott-like transition. What really happens in Mn is a separation of spin densities at the energy window, such that the magnetoresistance effect is large. Of course, from a resistance point of view, one can say that there is a resistance change and therefore, a metal-insulator transition. But, here the bias is the magnetic field, that "switches" the spin directions and cause the current passing through to change - and thus the resistance. Therefore, a disproportionation is not what happens. Since the Hubbard-Mott transition is caused by the energy barrier(Hubbard U) which is the activation of the Mott switch, we can take out from consideration the Mn contribution to the Mott transition claimed by these gentlemen. What could be happening is that the Prasedium (Pr) 4f orbital is causing the Mot transition. This is also evidenced by the following: 1. PrOx is a Mott insulator and 2. When Ca and Mn are together, the only thing Ca does is to create the valence fluctuation in Mn that causes the +2/+3 fluctuation. Now, as far as this being the mechanism in this memory device, as just a device, which means also some knowledge of the device physics outside the materials phenomenology, my view is that the answer is no. It is not dominated by a Mott transition, because the very fact that it is bipolar, breaks down the dominance of the Mott switch as the primary mechanism. This gives me an opportunity to talk about Bipolar Switching in general. It is amazing that you do not see too much about this in the literature. But, in my view, if you have a defective double layer at each near electrode region, bipolar switching is simply Gauss law at work here. That is, imagine two unstable cpacitors, one under each electrode, and a more staple capacitor in the bulk. What happens when you trap electrons on one side? D2-D1 =sigma, where sigma is charge per unit area and Di are the electric flux density? well, when you switch polarity, schbang, all charges flip arrangement in these capacitors and the trapping is varied accordingly. Thus, we have, as in these devices, a charge hysteresis in the negative side of the IV curve (also known as a bad non-ohmic contact ) and nothing like that but space charge limited current in the positive charge. Conclusion: the charge (valence/oxidation number) soup, makes it impossible to call this a Mott insulator. Sorry, strong electron correlation/Mott transition, requires a clean single U = I -A, where I=ionization energy and A= electron affinity, and not a soup of charged deffects and electrode induced space charge regions. A message to 4DS, try to contact me, I may solve your problem, but forget PCMO, it is a dog.
Resistion: On the subject of mixed filament and area conduction, based on details available in the public domain I agree. Because a change in composition is required for the transition (see my figure 3) and the creation of oxygen vacancies is an inhomogeneous process then the forming process and/or the creation of the high resistance state will most likely occur initially in a number of small localised areas (if you like surface nano-filaments) that expand and grow to coalesce as the current density between the islands increases. That would provide a both a positive feedback mechanism and self limiting process when all the area was converted. However, at the end point the current density for the final low resistance area might be high enough to result in damage, even localized fusing/burn out. If so to avoid damage then some sort of compliance might be required which would mean the final high resistance state would be a combination of small islands of high resistance material or even a convoluted worm-like surface of conducting material electrically in parallel with switched material. All that moves us away from the real point of debate with respect to this device is this a reversible Mott-like transition or is it something else ?
Going over the material, it doesn't look so distinct from other ReRAMs or RRAMs. The SET and RESET 'sides' are specified, indicating bidirectional operation. Their p.14 shows no area dependence at low voltages, indicating some combination of filamentary and area conduction. The MIT temperature looks too low to be practical, lower than VO2 even. Some curves on p. 17 indicate compliance is still needed for forming, or otherwise hardly any RESET window.