(Or "Beyond the MOSFET - Microlithographed Electronic Devices Exploiting
Some Quantum Effects".)
The notes for this lecture will be fairly brief. For more depth, read the "Overview of nanoelectronic devices" paper. That's where I got most of my information anyway.
Scaling of MOSFET devices (the structures we discussed last week) is hampered by quantization of various things at small scales.
(1) The number of dopant atoms in the device channel starts becoming small as device lengths shrink below about 0.1 micron. (This will happen in the next few years.) As this occurs, increasing statistical variations in the number of dopant atoms will mean wider variations in device threshold voltages across a chip. This problem is alleviated to some degree, but not completely, by increasing concentrations of dopant atoms (needed anyway in order to make depletion regions thinner). (The number of atoms, however, is still in the hundreds of millions.)
(2) The number of mobile charge carriers in the channel also becomes small at lengths of about 10 nm. Device behavior no longer exactly obeys continuous approximations to its behavior. Detailed analysis of its operation has to take quantization of charge into account.
(3) Also at around 10 nm, energy levels of electrons in semiconductors start to become significantly quantized due to their positional confinement, as the effective electron wavelength for mobile electrons in semiconductors is comparable to this length scale. Adding another electron to the channel takes extra energy because all the low-energy states are taken; a higher resonant mode must be chosen. In metals this effect doesn't become important until a smaller scale of around 1 nm.
The question to deal with now is: Can alternative device operating principles somehow exploit these quantization effects, rather than be hampered by them?
The "single-electron transistor" is so named not because there is really only a single mobile electron in the island; rather in these devices there are still millions. Rather, the important thing is that the capacitance of these devices is so small the addition or removal of a single electron has a significant affect on the voltage of the island.
When the voltage on the island is just right, and lines up with the voltage on the source, an electron can readily tunnel from the source to an island in "resonant" fashion without changing their wavelength. Subsequently, the electron can tunnel from the island to the drain.
However, the electrons can pass only one at a time since the addition of the first electron raises the potential on the island enough that a second electron is not able to tunnel onto it. Still, the electrons pass through at significant rates.
If the island potential is raised slightly (by action of the gate) the electrons are less able to tunnel onto it. Similarly if the potential is lowered slightly they are less able to tunnel off of it.
However, if the island potential is raised a whole lot, an electron will come off of it, and the voltage will revert to a somewhat lower state. Similarly, if it is lowered a whole lot, an extra electron will come sit on it, and the voltage will revert to a higher state.
So there is a cyclical dependence between the gate voltage and the island voltage, and therefore between the gate voltage and the current through the island (which depends sensitively on the voltage level).
In quantum wells, wires, and dots, electrons are confined in 1, 2, or 3 dimensions respectively. In the other dimensions, electrons are free to roam about. In quantum wells and wires, 2 or 1 dimensions are still large, so there are large numbers of electrons and so charge quantization effects are not so important. But in quantum dots, the total number of mobile electrons on the dot may be as small as 1!
In these structures, quantization of energy levels for electrons on the island means that the flow of current across the island is sensitive to the exact potential on the island, in order for the energy levels of electrons on the island to line up with the energy bands available for conduction in the neighboring source/drain contacts. When they line up, there is resonant tunneling.
Note that none of the above structures (either the single-electron transistors, or the quantum islands) depend on dopant atoms, since separation between the island and the source/drain depends on an insulator of another material; in MOSFETs the separation depends on the depletion region of the PN junction, whose thickness depends on the dopant atom concentration. In the quantum islands there is no issue with quantization of dopants.
However, in these structures, behavior is highly sensitive to the thickness
of the tunnel junctions, since tunneling rates are generally exponentially
dependent on distance. High precision is required in the manufacture
of these structures.
The problem of course with molecular electronics is the manufacturing technology required to assemble and wire together millions of these molecules into a single, precisely designed structure, and do it all in a cost-effective way. You can't make a molecular computer with lithography. We are still a very long way from molecular manufacturing. But it's still nice to know some devices will still work, once we *can* build at that scale.
We'll talk more about various schemes for molecular computing a bit
later in the course.