Thank you for your questions, but they seem more expressions of belief than healthy scepticism. In any case, I accept the challenge:
Answer to you question 1 - If that was the case there would be no need to set a current compliance during the switching cycle. In posing this quetion in reaction to the article's claim of Metal-insulator transition based in ELECTRON CORRELATIONS, you put forwar (embedded in your opinion) that you seem to know a lot about Electron Correlations and Mott transitions.
ANSWER> You must be confusing electron-electron correlations, such as in a metal, where multual repulsion causes a "fermi Hole" between electrons, with what is also called electron correlation due to the Coulombic Potential called the Hubbard U in the literature. In that case, and it is the case here, it is the occupancy of the 3d8 shell that creates a change in the oxidation number of the Nickel and a gap is opened. That is the insulating phase and it is well known that Nickel Oxide should be a metal, by Bloch-Wilson band theory, and yet is an insulator in its's normal state. When, pressure or a lot of electron injection at the nanoscale, we have the complete screening of the gap (which technically is between 2p(oxygen) and 3d(L)9, where L=ligand). To achieve this, some of the TMOs filamentary proposals may actually be doing the same thing, either inside the filament or near the surface where many oxidation numbers occur for the TMO. Now, WHEN THE GAP CLOSES, AND THE UPPER EMPTY CONDUCTION-BAND LIKE NARROW 3d9 BAND AND THE LOWER FILLED BAND TOUCH, WE HAVE THE DEFINITION OF A METAL, i.e. BAND OVERLAP AND METALLIC CONDUCTION OCCURS. SO, TELL ME NOW HOW DOES YOU COMMENT SCORE THAT "IF THIS IS SO THERE WOULD BE NO NEED FOR COMPLIENCE" - IT IS NOT EVEN WRONG, TO QUOTE LINUS PAULING. SORRY, DUST OFF YOUR COPY OF KITTEL.
Answer to question 2 - "If this is the case the ON and OFF current should scale with device area" - this almost looks like a fair opinion, after all V=IR. But, try to understand that V=IR assumes a "Drude" conductance, meaning that this is a classical electron gas device with Rigid density of states of which the the conductivity is proportional to (and also the electron density). But, even if you go to a more semi-classical model of conductivity, like Sommerfeld's, this is not possible to be used in this case, as the conductivity for ON and OFF change because the density of states is not rigid and can be controlloed by two steps - on in the material itself, that is have only +2 and +1 and +3 so that the disproportionation reaction of the nickel is reversible - from metal to Inulator and vice-versa. And the second step is in the Device Physics itself, something that I will leave out for now because you were so certain of your opinions that I think you must do some homework in this part of the explanation. In any case, When you have the ON current, you have a dead short with a mesoscopic ballistic current - this in the filament world was confused with a connect/disconnect model of these filaments. It is not a wrong model, it is only the worng preparation of the material: if you create filaments it is because the many body phase is simply not there - that is you are stuck with many traps that create Ni(0) metallic traps and lots of oxygen vacancies that lead to Ni(+4) which all together show bad OFF state retention and scattering in the Vset and Vreset programming voltages. SO, let's now focus on your Scalling rules for both conductances: When the Mott Insulator is in the insulating state, look carefully at your IV curves and you will see what? A diode like IV - does this scale as a conductance? when that IV curve is actually from tunneling? What is a Tunneling Conductance (homework for you again)? And now, when it is in the metal side, we can consider this an ohmic current with a simple free electron model, and YES, YOU ARE RIght, it should scale as 1/Area. But, in the large area data, this is like controlling and measuring the diference in resistance between aluminum and copper - you can only see that if thickness/Area ratio is small and sigma (from J=sigmaE) would be dramatically difference. In advanced CMOS, length (thickness) shows a difference, but not in these large area devices that we made public to the EETIMES. Now, because the inner active area is doped differently from the two contact buffer layers of very conductive non-switchable NiO:CO, we have now 2 degrees of freedom to "fix" the resistance (and the high ON current) - Doping for scattering and thickness adjustments. Doping for scatering can be optmized not to influence the "Doping with a Ligand to compensate thte coordination sphere" (Homework: Coordination sphere? what is that? - that is what has been grossely overlooked by my filament colleagues).
Now, it seems that I still did not answer your question 2 - but I did have to set the stage and reboot some of your background, because a sense a swiss cheese span of condensed matter physics. THE ANSWER: With an IV for the OFF state depending on Square root of Voltage, a sign of the image charge and the other conductance being such that the current is ohmic (plus the large area in a metal phase), how in God's solid state do you come up with a ridiculous scaling theory? Sorry, I am not really a mean person, I just do not appreciate uneducated scepticism passing as an educated and welcome scepticism in good science.
Goafrit, your old boss would be proud. We worked
For 6 years on this and patents are issued in the
US and all over the world. Now, academics. There is
A lot of science here and it is a tough road ahead. Everyone
Runs to publish half baked things. It is time to work first and publish later. Academia as many have called for change, should no longer be business neutral. We need the jobs here...
In VO2 device, the self-heating above transition temperature caused turn-on. What is the temperature effect here? Interestingly, the VO2 device was filamentary, i.e. thermal runaway of localized current would require compliance.
As you know, VOx is very temperature dependent. In fact there is a famous Korean patent for a current limiting device presumed Ed use in a device that become more resistive as the cell phone battery gets warmer. The stoichiometry of VOx makes the phase transition to be second order and very sensitive to thermal runway. It is not possible to generalize all stoichiometries and just exactly is the metal- insulator transition, Mott, or even Anderson ~ Mott, a combination of defects, specially oxygen vacancies. I will risk to say that de trapping electrons from these vacancies may together with a high current in the metal phase creates the thermal runway you talk about in VOx. But in our case we damped the oxygen vacancy content by forming stable complexes that tie up Nickel metal clusters and at the same time fixing the ohmic oxidation number with the CO complex. Since the ohmic side starts at zero voltage and very small currents, thermal runaway initiated by the current can be ruled up, for I ^2 R is very small and can be made extremely small, for example, below 100 microamps of ON current in a 100 micron square device. With R from 100 Ohms to 2000, in these particular doped samples, and a very linear and temperature independent switching fro - 269 to 150 C ,continuously switching, where can you find this thermal runway?you surely do not expect filament formation at 4 degrees kelvin?
So I recommend that you unplug a little from carrying over your thermo interpretation and review the many variations of VOx and brush up on the special area of quantum phase transitions which are temperature independent and happen even at 0 K. We only went down to 4k, and it kept on running. We are trying to go to 0.4 K looking for fermi liquid behavior.....there is a lot involved in this. Rich literature. Filaments are optional and are parasitics rather than the operating principle, that is, if start with a corrected coordination sphere. Remember the Co and ammonia experiments in basic chemistry? Since ammonia is a ligand, variations in colors show how coordination chemistry change the bands and you have all those pretty colors. Google this, it's elementary but will clearly show you that these localized electrons even in an aquas solution have different band gaps due to 3d orbitals interactions. All I did is apply coordination chemistry to solid state TMOs. By the way, platinum complexes use coordination chemistry the same way in chemo therapy.
Volatile-While I can for the moment accept your claim that you can create buffer layers to screen the central layer, third layer of your (20nm-20nm-20nm)thick structure, what happens at the edges of the device in the lateral direction. If you have a "pore" like structure then I am at loss to understand why the surface states will not lock the material in one of its resistance states. If the material in the as born state is insulating that might be OK, if not then as you shrink the device to sub-20nm dimensions the edge will create a filament like structure of the active area and the resistance will become edge leakage dominated. I guess in analogy it could be a probem like the ratio of area/volume that the PCM promoters ignored. Would be interested in your observations.
On the subject of VOx as somebody who has made many of those in pore structures as potential matrix isolating devices the the low transition temperature was the practical problem not thermal runaway in the conducting state. Those devices with micron dimensions, had resistance that scaled with the inverse of area.
your model considers a non-local switching effect, i.e., no lateral confinement as in the case of filamentary switching effects. I simply wonder why you have to set a current compliance while switching from the OFF- to the ON-state. Does the overall current density become so high under the switching conditions that it would be critical for the whole MIM cell?
Central to your question is the comment "if you have a pore structure".
But, we do not have a pore structure, quite the opposite. Films
Are dense and stoichiometric. Take a look at the XRD data : ours are exactly the reference bulk data. The XRD of sputtered devices is all over the place. Also bad edge effects due to a born on ohmic state actually goes the other direction as the area reduction increases the resistance, wanting to make the insulating state. It then takes counter doping to bring in the born ON condition. If you do not start there the born off state would require some higher voltage and certainly some form of filament is there. So, the adjustment is to start born on and cycle from there to the off state. I guess that because you do not have this corrective doping and they start with sputtered the oxidized devices, the devices of the filament people will always need Electroforming. That is what would really create edge leakage.it is sure not the case with born on . We can as just the resistance by doping every layer and achieve born on without stuck bits. This is the beauty here: once the current Density is reasonable, adjusted by the doping , the surface to volume ratio becomes manageable and we follow the R vs. square root of Voltage curve that is design to scale everything correctly. All events of switching occur above the insulating baseline which when scale without doping is a disaster and Electroforming is need.
I have no problem with your VO experience. As I mentioned before it is an extremely temperature dependent material.
The transition from insulating to metal is caused by two steps; 1, barrier lowering with classic MIM behavior dominated by image potential. The turn on shape in a linear scale should be well designed, in the second step, the bulk conduction takes over as the core potentials are severely screened, we have now a metal. So why you would not have compliance. The trick here is to keep a handle on relationship between Ion and I compliance.
And as you know, current density is not a function of area or thickness, current is, in the plot of current density versus field, conductivity is the only variable that depends on materials properties.
Misty versus field, conductivity is p
which depends on doping and DOS. You are correct on the value of keeping the MIM under control here. I do not want to review too much
here. This is when we leave chemistry and enter device physics, a real messy and confusing picture emerges if you do not have a handle on the switching mechanism, which we found and for now it is proprietary.