At a recent meeting of the American Vacuum Society, ReRAM developer 4DS Inc. (Freemont, CA) provided an update on the progress it has made with the development of metal oxide hetro-junction operation (MOHJO ™) non-volatile memory. In that presentation, company representatives showed a micrograph of a 16k-bit crosspoint matrix. The joint development a 16k-bit memory and controller by 4DS was reported in EETimes in 2011. At this time it is difficult to ascertain if this is the same array or something new.
For this latest presentation, data for write/erase cycle lifetimes of 106 cycles were shown with claims that 109 cycles are possible, supported by impressive read stability data, all without any indication of the number of devices evaluated or the important statistical distribution of those parameters across an array or population. The elevated temperature retention is cited as greater than 10 years at 85OC, supported with what was described as thermal relaxation data up to 130 OC, again without the number of device tested. Typical device programming voltages ranged from +4Volts for low to high and -5Volts, for high to low reset and set resistances respectively with on-to-off resistance ratios of about 100 over a cell size range of ~20 to 400nm. No details of the 16k-bit array interface were provided nor any detail of the memory matrix isolation devices. It is possible that the asymmetric I-V characteristics of the individual memory devices are serving in that latter role.
4DS Inc. now claims that the operation of its memory is based on strongly correlated electron-effect Mott-like transition, and, in so doing, they become the second company with a memory device that lays claim to correlated electron effects, following the CeRAM of Symetrix. The 4DS Inc. trade mark for the memory device is MOHJO™, the active material is an alloy of PrCaMnO3, with the acronym PCMO, for praseodymium calcium metal oxide memory. Or, more simply, it is PMO doped with calcium, and it belongs to the family of perovskite mangantite materials with composition R(1-x)AxMnO3, where R is a rare earth metal and A is a divalent element.
If, as is claimed, the PCMO memory is a true interface device, it rates highly on my NV memory desirable features list (see Figure 1), a list which for the moment ignores any undesirable features such as the need for forming and bidirectional operation.
Figure 1: Desirable features of NV memory
One very positive aspect of any interface device is it removes as a first order effect the control of film thickness from the fabrication process and any associated variables from the device characteristics.
The PCMO structure and operation
The PCMO memory device structure is shown in Figure 2. The lower electrode is platinum (Pt), a metal that does not react with the PCMO alloy. The top electrode is titanium overlaid with a suitable good electrical conductor, such as gold. While Ti and the formation of α-TiO is apparently the preferred implementation, 4DS has evaluated many other contact metals.
Figure 2: The PCMO memory device structure
Many companies and organizations have explored the resistance switching properties of various compositions of PrCaMnO3 materials. Perhaps the best starting point for understanding the 4DS memory (and the evolution of its claimed mechanisms of operation) is an investigation  of the switching effect in Ti/Pr(1-x) Cax MnO3 structures; especially as a significant part of this work was included by 4DS in their most recent presentation. That earlier investigation uncovered and confirmed a number of important features. For instance, as a result of switching (with titanium in contact with the PCMO material), a layer of amorphous titanium oxide (α-TiO) is formed at the interface by what was reported as the electrochemical migration of oxygen ions into the original (Ti) electrode.
This leaves oxygen vacancies on the surface of the PCMO, leading to a high resistance state (HRS) as illustrated in the lower part of Figure 2. While resistance switching was possible for junctions with x<0.8 compositions, junctions of n-type semiconducting PCMO with x>0.8 showed almost no switching effect. The third feature was ratio of the high to low value resistance states was dependent on x with the ratio increasing to a maximum at x= 0.4. The conclusion of  was a model involving the change in the contact barrier height and/or width for the hole-carrier conduction at the interface. This was induced by the electrochemical migration of oxygen vacancies as the mechanism of the resistance switching (RS) effect.
In the as-fabricated state at the interface between the PCMO and titanium a very thin (1nm) oxide initially forms (see Figure 2). The post-fabrication forming process then results in a 10nm thick layer of α-TiO at the interface. (As stated earlier, the forming operation involves the movement by electrochemical migration of oxygen ions across the interface to create a metal-oxide film.) The resulting depletion of oxygen in the PCMO creates the oxygen vacancies, and the subsequent electric-field-driven backwards and forwards movement of those vacancies across the oxide-PCMO interface results in the two resistance states, the values of which are determined by the relative densities of the vacancies on either side of the interface. The exact detail of the forming process and any difference from normal operation are not clear. This electrochemical driven movement of oxygen vacancies acting to modulate the resistance of a Shottky-like barrier would appear to offer a reasonable account of the operation of the PCMO device. That effect alone would provide a reasonable basis for 4DS Inc. to move forward and evaluate its memory devices in array form, or at least that was the case up until quite recently.
The Mott-like transition
The claim that the operation of the device is based on a Mott transition deserves more careful examination; with the possibility that if a transition of that type occurs it might only be partial and an effect that is related to conditions that are very localized and possibly only occur at a surface with point contact experiments carried out in air.
The composition of the PCMO (see Figure 3) places it between the two insulators PrMNO3
. Although it might be expected to be a metal, it is classified as a correlated electron semiconductor. As such, it does not automatically mean it will be capable of Mott-like metal-to-insulator transitions.
Figure 3: Composition of the PCMO
The claim by 4DS Inc. of its correlated electron effects and a Mott-like, or Mott-Hubbard, transition appears to have its origin in , where the surface of a film of Pr0.7Ca0.3MnO3 was switched using a point contact that was moved across the surface to build up a dot pattern of switched material. This was done in order to provide a large area of HRS material for evaluation and analysis, as illustrated in Figure 4. The use of a tungsten tip provides the advantage claimed by the authors  that the oxidized metal on the tip can be removed with the tip, leaving only the switched material for analysis. However, because of confusing results related to analytical beam area, the results and final conclusions were based on an area smaller than a single probe tip.
Figure 4: Illustration of the methods used to evaluate and analyze the switched material.