The smaller device geometries are approaching the atomic limit for device construction.
In my previous blog, Reading the tea leaves, I stated that: "it is impossible to divide an atom in two while maintaining a non-explosive environment". John Michael Williams, Senior Adjunct Faculty at the Silicon Valley Technical Institute sent me email that corrected and improved my statement.
He said that "Actually, lighter atoms absorb energy when fissioned, so they would not explode. It is only the more massive atoms, above iron in atomic weight, that can liberate energy by fission. Your point is well taken, otherwise. A stable structure must be made of some number of atoms, or else quantum uncertainty would make the structure unusable as part of a device." He continued with the following observation: "My guess is that the fabrication (pitch) limit for devices made of anything is going to be around 100 atoms on a side (5 - 15 nm). A structure with many atoms can operate upon a single atom or electron, but this does not mean that the structure itself could be built of just one atom or a very small number of them."
All this made sense to me and got me to thinking that I had heard at more than one presentation that deposition layers using present fabrication technology were already about 5 or 6 atoms thick. So I wanted to have more input from him. And John Michael has been very responsive. What he sent not only makes sense to me, but I think it is something everyone involved in EDA and for that matter semiconductor manufacturing needs to keep in mind. What follows is his contribution to the discussion reported verbatim.
How small can we get
It depends what the layers are doing: If they were hundreds or thousands of atoms wide, then 5 - 6 atoms' thickness could be useful to insulate two differently-doped layers, above and below the layer under discussion. With only a few atoms' thickness, such an insulator would not be good; there would be considerable leakage of holes and electrons, depending on the potential across the layer under discussion. Even with 0 potential, individual electrons would diffuse across it at random (see below).
The underlying factor is the uncertainty of localization of an electron to one or another side (above vs. below) of such a layer. No matter on which side an electron was "located", the wave nature of electrons on atomic scale would imply that there was a finite probability it would interact with an atom not on that side, but on the opposite side. Thus, all thin barriers are permeable to electrons, merely because of the wavelength of the electron. This means there is something of a short-circuit across any very thin (1 atom) barrier -- the barrier becomes more or less invisible and ineffective.
You can find a nice explanation of the relationship of electrons to atoms at Atomic Orbital. Assuming a planar (Bohr-like) atom, the wavelength (quantum wave function wavelength) of an outer-orbit electron of a moderately light element such as silicon, according to the
posting above, would be no less than circumference/n or maybe circumference/3 = about (1/3)*Pi*(1 angstrom) = about the diameter of the atom.
Instead of a wide layer, if the "layer" was a thin strip, maybe like a MOS transistor gate, of, say, metal, it could be used to convey a potential from one place to another (like a wire). This assumes the surrounding structures were many atoms thick in every
direction, so that the distance the potential (voltage) was conveyed was not "short-circuited" in the surrounding materials merely because of electron wavelengths.
I don't think horizontal layers less than 3 - 5 atoms thick could be useful, depending on the geometry (crystal structure) of the atoms. Not unless everything was occurring in a torrent of modulated leakage!
So, my GUESS is that 5 - 6 atoms thick is about the limit, already, and that it won't change by more than, say, 1 atom. I doubt that a single on-chip device (switch+I/O contacts) ever could be shrunk to less than about 100 atoms on a side (10 nm), assuming that the chip had to operate because of millions of such devices.
This doesn't mean that large structures could not manipulate single electrons -- for example, an atomic-force microscope. It does mean that the large structures have to dominate the design of any such device, so a few tiny structures really do not determine
the final scale of the device. "Large" means tens or hundreds of atoms on a side.