For a circular aperture of 20-nm diameter with a fill factor of 50%, the disk diameter is 10 nm, 7 nm, and 5 nm for two, five, and eight disks respectively; for sub-20-nm apertures these numbers will scale.
Compared with a cross-section view, the plan view in figure 10 of a conceptual 20-nm-diameter device with 5 nm oxide disks gives a better idea of the shape of the electrode surface. Cross section views of S-A devices can lead to the misleading conclusion that S-A devices are a series of—and can be treated as—small, isolated devices, which the plan view quickly refutes. Only if the packing arrangement results in all the disks touching can the surface be considered as a set of discrete devices.
Figure 10: Plan view of core device (green), electrodes (grey), and oxide disks (yellow) during the growth of the molten region during reset for self-assembled device shows how the initiating molten hotspot forms (a) and starts to expand (b) to encompass the central core annual region (c). It then further expands along all exposed electrode surfaces (d), expanding into oxide surface (e) to completion.
The side elevation as shown in the original paper gives a false impression. In reality, the electrodes shown in figure 10 consist of the spaces between the oxide disks. The resulting structure can be considered as a large number of overlapping annular electrode devices or a device with a single electrode formed as an interconnected hexagonal network. Although figure 10 is a plan view of a seven-oxide-disk PCM device, much of what follows will be applicable to larger devices that extend in two dimensions. In this small device, the white colored areas will also be electrode surface and in an extended device will be part of the adjacent overlapping annulus pattern.
The electrode surface colored green is the core device. Based on the initiating molten hotspot (IMH) model, the sequence of events would be as follows. First, the IMH forms at some point on the green electrode surface at the mid-point between three oxide disks, it then expands to cover the surface of the core annulus electrode. Next, quite rapidly, the molten material extends to cover all the exposed electrode surface, with fill-in over the oxide in the central region. That expansion will also be 3-D, occurring in a direction normal to the plane of the electrode. The growth of the molten region then proceeds to completion with extension over the oxide and exposed electrode surface, as shown in figure 10. As described in the IMH model, a twofold role for the reset pulse is required. In the early part of the reset pulse, the current density must be sufficient to raise the temperature of a substantial volume of the crystallized active material to a temperature of (Tm
) and the hotspot to Tm
, the melting temperature.
It is interesting to compare these results with a PCM device using an annular electrode and to understand why the reset current and current density will be different. For a 20-nm aperture, a 3-nm-wide annular electrode can be considered as a fill factor of approximately 50%. That is, in effect, the equivalent to a single oxide disk with a diameter of approximately 14 nm.
The two center diagrams of figure 11 compare a simple annular electrode structure with the equivalent eight-disk solution for a 50% fill factor. With the same fill factor, the electrode surface area is the same and the Jc
ratio will be the same. Recall, though, that both simulations and experimental results showed lower current density Jb
from the self-assembled device than could be explained by dimensions alone. The question that must be answered is why there might be a difference in current densities Jc
and the reset current for the S-A device and a similar diameter annular electrode PCM device when the volume of active material above the electrode is the same.
Figure 11: Comparison of 50% fill factor of eight balls (middle top) with equivalent-fill-factor simple annular electrode.