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.
Thermal runaway may have been a poor terminology of the occurrence which was not destructive, the filament was a common observation. It's also possible the filament evolved into a conventional resistance, as that was also commonly observed.
Nonvolatile:-Thanks, but I still have a problem, if you do not have a pore structure as the means of defining the device, for the moment let take that out of the discussion and any influences the surface states of its dielectric might have. Consider an orthogonal array of CeRAMs where your three layers are formed as a single large area on a flat surface. The lower electrode in a narrow channel in the flat surface and the orthogonal electrode on the upper surface of the three layers. Now after all the doping and annealing you describe you creat a device that is stoichometric, doped and conducting in the on state. Now when I switch to the insulating state there is a region at the edge of the device where the material in the conducting state is in contact with the insulating material and bridging it; plus there are already many more conducting paths between electrodes! Although you may not want to consider a "pore" as the type of structure that defines and confines your structure, the device structure options appear limited. It would appear to be either surface state effects or bridging. Unless you are using some form of masked ion implantation to achieve the control the doping as well as defining the structure, accounting for the need for annealing, i.e. a post-implant anneal.
2)I know you do not want to discuss all the details of the switching as you said in your post below. Are you able to say if there is any part of the switching transition that displays negative resistance, I guess it would be "N" type. If so the oscillation frequency of a negative resistance oscillator constructed using your device would resolve the thermal problem and might have useful application elsewhere, especialyy if the switching time is as fast as you claimed..
Ron, you have an interesting question. But, you are looking at the transition still as something that moves thought the device as it is a moving front of electrons depending on a classical drift velocity. As you are an expert in VO let me remind you that the Mott transition in VO was recently measured at 80 femtoseconds. So, if you think the way you seem to be going, and I am not too helpful for reasons of space more than proprietary info, you then miss a key point , specially in the insulator to metal side of the switch. So let me explain:
First you should know that an insulator of this type has practically the same number of electrons as a metal.the problem is that these electrons(here 4s2 electrons from nickel ion lost to the lattice ) are far from the fermi level and thus cannot conduct. When the density of incoming electrons from the cathode form trigger the metal state, there is no gap and we clip the current. Now for your specific Ansatz, the starting of the on to off state, which in the filament theory sometimes starts at the anode, is also what happens here, but for slightly different reasons: in this case, near the anode, within a few atomic layers have gone through a quantum phase transition at a 1000 to 100000 times faster than electrons coming from the cathode. The cascading is immediate and the fermi level now is in a region of zero density of states. Thus, we depend on this lack of thermal equilibrium to switch . In the filamentary world this is called a local oxidation reaction region. But this is accounted for as a disconnect of the filament. And it is. But here, in the right phase is just the disproportionation.so, the speed of propagation of the phase transition being way faster then electron drift, causes the gap to appear faster than the speed of injected electrons and the device is instantaneously off.
The negative resistance does not occur here. But it is very easy to think that it can occur if the RC time constant is long in the metal to insulating condition. It would look like a tale. This can happen if the area is really large. But RC means Maxwellian time, which may happen in the metal side as the current is starting to go a minimum as the insulator phase sets in. But this again is too slow compared to the speed of the band splitting.
It maybe interesting trying to use this device in places where super fast switching is needed, but I do not know if we could get negative resistance with NiO. I do have several other materials which also use compensation doping that delivers an enormous current in a fast time. I will check that set of data to see if they have negative resistance. Just to clear the air, no ion implantation, nothing more than what I have disclosed- it is all in maintaining the coordination chemistry right.
Nonvolatile: Thanks again you are being more than generous with your time. No I am not an expert in VOx just got involved when trying to rescue PCM where I have a little more expertise. That is in a failed attempt to use doped VOx in a bi-directional matrix isolation device as a means of cancelling out the effects of element separation caused by current density in the PCM. The chalc based bi-directional threshold switch was useless, irrespective of the claims by one large company for their stacked Chalc-Chalc based isolation device-memory array.
With respect to propagation of the effect, I am more in favor of a radial model, certainly works for PCM and most likely accounts for what is happening in some of the oxide ReRAMs, but not neceassarily in all cases. I think when crystal electrodes are involved the elevated temperature failure of PCM is crystal growth, propagation in the direction of current flow. The latter was the point I was trying to make in one of the PCM Progress Reports Published in EETimes. It might be a potential problem with your MIM structure.
Where am I going with the speed aspects. I have recently been exploring chalcogenide threshold switching as a posible means of constructing a thin film large area phased array antenna. The Chalcs are much too slow and the negative resistance oscillations self FMs because of changes in the disorder which suggest things are getting warm, irrespective of what appears to be a very rapid initial switching transition. If you have other materials it might be worth thinking about that type of application for both large and small area thin film PAAs.
No problem - I really enjoy put this out there with frank questions and learned opinions.
The reasons are that may Sayers better know their stuff and knowledgeable skeptics can keep you on your toes . And finally we can perhaps challenge the pseudo science nonsense that some large companies seem so entitled to put out there disrespecting our intelligence.
I completely agree with you on PCM but I cannot agree with your radial switching even for filament makers. In the electro forming step the filaments are grown surrounding grain boundaries as shown recently in the literature. In this I agree and I modeled that long time ago and measured the activation energy. But the switching is not radial growth of filaments, only electro forming . In our case we have just too much data and analysis to the contrary. In PCM the local recrystalozatiom temperature leads to a heat conductivity of about 480 C/sec and that is about equal to molten salt. There is then radial heat dissipation and you are right. In FeRAM the surface defects nuclear daim growth down very fast, so thick or old Materials like TGS or even PVF2 depend on lateral growth that cannot go faster than the speed of sound.
I guess we can stay disagreeing on the mechanism until I can provide more data to match the physics. But, take a tip from the fact that (a) we can make NiO always conductive and if we have a 60 nm thick device with 100 micron squared, the resistance is 4 ohms. Now use that as buffer layer of 20 nms with a middle active switching region of 10 nms that can go to 150 mega ohms by switching, it makes radial charge transport a bit difficult. Stock to tunneling and you will stay on the right track.
In one of your papers, there was a model of the insulator state. I am not sure if that model is updated yet. It suggested self-saturating current (which I presume not correct). More seriously, it also showed that the resistance of the insulator would be dramatically decreased at an elevated temperature above 100 C. Is there an update to this in an upcoming publication?
Frankly I do not know what you are referring to. I would never promote a memory that dies at 100 Cand the data I sent to EEttimes goes to 300. Currently it is 400 C.
Now the self saturating is another story. The relationship between compliance and on currenyy can saturate at low compliance values, that is also In the literature. Of filaments