A new approach aims to be a contender for next-generation memory. Here's how it works.
In its simplest form, the CeRAM structure (see Figure 3) consists of a three thin films of Ni(CO)4 doped NiO nickel oxide between two outer metal conductors. The active material is sandwiched between two films of NiO that serve as buffer electrodes. The two buffer films are doped to be very highly conducting and act as matching electrodes, while the central core active film is nickel oxide with a lower level of similar doping.
The barrier layers play a number of important roles. They provide an ohmic contact to the active material and, more importantly, move any Shottky barrier-like effects caused by the outer electrodes and any unwanted surface states away from the active material electrode interface. These surface states would impede that action of the device. In operation, it is the central region that undergoes the reversible MIT between conducting and insulating states and makes it possible to remain in either state as long as required. This ability is the basis of its potential use as a nonvolatile memory.
A simple CeRAM structure.
(Source: Ron Neale)
Though "doping" is a colloquial term used to describe the means by which the NiO structure is modified, the concentration levels are higher than those associated with donor or acceptor doping in conventional single-crystal silicon.
CeRAM: a conceptual view
The fine detail of the physics underwriting the CeRAM mechanism is complex. As an aid to understanding, Figures 4 and 5 provide a simplified conceptual view of CeRAM operation.
Figure 4 highlights the two regions of the thin-film CeRAM structure: the buffer layer and the central active layer. In those films, each nickel ion can be considered part of a small local switch that is on or off. The second expanded inset illustrates the reversible band splitting of the MIT that empowers the CeRAM. (The band structure is a localized effect and should not be confused with the band structure associated with single-crystal silicon, where the band structure linked to periodicity is fixed, and the density of states is not manipulated.)
Underpinning these changes is the split of the 3d8 band of the conducting state into the 3d7 and 3d9 bands of the insulating state (where 3 means the third atomic shell, and d8 refers to the d-orbital up to eight electrons).
Reversible band splitting.
Continuing with the simplified view, Figure 5 is a conceptual view at the nickel-ion level of the difference between the insulating and conducting states. In the upper part of the figure, the material is conducting, and charge carriers can move freely in the material's local conduction band. When carriers become localized, the Coulomb repulsion inhibits electron flow, and the material acts as an insulator. The release of a localized electron during the set process brings the material back to the conducting state without the repulsive effect from the localized electrons.
A conceptual view at the nickel-ion level of the difference between the insulating and conducting states.
The Ni(CO)4 doping technique, patented worldwide by Symetrix, creates a film of nickel oxide where all the nickel ions are in the single Ni+2 electronic state. Doping acts to clean up or stabilize the internal volume of the material while, as indicated earlier, the device structure design acts to clean up the surface. Without the benefit of Ni(CO)4 doping, free nickel, NiO traps, and oxygen vacancies would exist in the NiO, and the films would not be conducting. More importantly, the reversible MITs would not be possible. This means that the five 3d electron orbits of the nickel are in the 3d8 state.
In the conducting state, all the positive ions are screened, and electrons have removed the effect of the potential of the positive charge, which, by definition, cannot exist in a metal. Aided by the Ni(CO)4, the screening is perfect. Even though there are many electrons to provide the screening, this does not mean the potential well (formed by the positively charged ion) has gone. For example, if the voltage applied to the device in its conducting state is increased, at about 0.6 V (in a region close to the anode and as a result of hole injection), the screening becomes less than perfect, and the carrier equilibrium is disturbed. The local band (previously for the 3d8 orbitals of all the nickel atoms) splits into two bands separated by an energy gap, where the upper band is now the Ni+1 band (corresponding to 3d9). This gap appears as a result of the drop in electron density as holes tend to reduce the number of screening electrons near the anode. The applied voltage is half that needed to inject electrons to screen the nickel cores. Thus, an electron entering the 3d9 orbital of the nickel ion becomes localized by interacting with the electron via a strong electrostatic potential, and it acts to trigger the transition to the insulating state. This transition propagates throughout the film as a quantum phase transition (see below), which is responsible for the sharp switching off of the conductive state.
In the insulating state at low applied voltages, the conduction mechanism is dominated by thermionic emission carriers crossing interface barriers. As the voltage is increased, the device behaves like a metal-insulator-metal diode, and tunneling becomes the dominant method of conduction (the steep part of the I-V curve in Figure 1). At about 1.6 V, the electron density reaches a level where the potential well of the nickel ion (which holds the localized electron) is so narrowed that it allows the electron to escape by tunneling out. This kills the bound state responsible for the intra-site Coulombic repulsive electrostatic potential, thus returning the material to its original metallic state.
Simultaneous oxidation and reduction
Figure 6 provides a view of the valence orbits and the band structure of the nickel ions with a sequence from left to right of the transition between the as-born conducting state of the NiO, CeRAM, and its insulating state. Initially, the doping process has ensured that all nickel ions in the conducting material are in the Ni+2 or 3d8 state, with the valence and conducting bands overlapping. Injecting holes creates a situation where an electron can move from the conduction band of one 3d8 to double occupy the conduction band of a second 3d8 and create a 3d9. This results in the band split shown in the lower part of Figure 6. The material becomes insulating. More simply, the 3d7 does not have any conducting electrons, because there is now a gap between the bottom of the conduction band and the lower bands where electrons are available.
At the core of these many body effects is a reversible disproportionation reaction that can be used to store data at temperatures as high as 400°C, with the transitions possible over a temperature range from 4 K to 150°C. Phase transitions are independent of temperature, and they might be considered strong evidence that quantum phase transition is involved.
View of the valence orbits and the band structure of the nickel ions.
The situation is complicated because the p-band of the oxygen in the NiO overlaps the conduction band of the nickel after the band split. Only a few ions are shown in Figure 1, but the band splitting is occurring locally throughout the volume of material. The reversibility of this transition is the basis of its use as a memory.
This article has not been intended as critical analysis of CeRAM. Instead, the goal is to provide the highlights of the operation and structure on which CeRAM claims are based. Readers have to accept there is always a danger: Relying on particle electronics descriptions for a device that has its roots in quantum mechanical effects might lead to omissions and misunderstandings.
The formal publications from the University of Colorado team will provide the scientific detail, while third-party evaluation of CeRAM devices will establish the validity of the manufacturability and performance claims for the devices. One can speculate that the reason the bulk switching effects claimed as the basis of CeRAM operation have not been observed in other oxides where filamentary switching is observed (i.e., HfO and Ta2O5) rests with the difference between n-type and p-type conduction.