I emailed the professor a few questions that night, as i just don't seem to be forceful enough to get them in directly after proceedings in person. Here they are interspersed with the reply i just received (me in blue):
Thank you for the email. Sorry for not replying earlier – I have been away early this week.
I have tried to answer your email below…
Professor of Nanoelectronics
Microelectronics and Nanostructures Group
School of Electrical & Electronic Engineering
University of Manchester
Sackville Street Building, Room D31
Manchester M60 1QD, UK
Tel : + 44 161 306 4762
FAX : + 44 161 306 4770
Professor Song, I attended your IET lecture yesterday evening, was very interested and impressed, but had to leave before I could ask a few questions:
The nano-devices operate up to Terahertz frequencies, but how would you then interface with a nanotech CPU running at this speed? (converting to standard electronics devices)
--- these are still research devices, a bit early to consider integration for a CPU. Nevertheless, they are planar devices, which are quite easy to be connected to “standard” devices. If there will be some applications, most likely they will be used only in the most speed-demanding part of a CPU. As you know, after a half century of development, the silicon industry is extremely resistant to any change of their basic building blocks.
Also, is the high frequency going to limit the size of a processing device in terms of synchronous logic operations; presumably, past a certain limit, a clock pulse would take longer than the time between pulses to reach the other side of a chip. i.e. limited to less than, about, 0.3mm for a 1THz clock pulse travelling close to the speed of light (in vacuum). Is this a serious limit to device size? Will CPUs have to use asynchronous logic or are there some tricks that can be used to circumvent this issue?
--- You apparently know much about circuits and integration – I came from the physics end… It is a good point but I do not really know the answer. In any case, presumably it is very difficult to increase the whole CPU speed to THz, by having individual THz components, because of, for example, the interconnects. I know asynchronous circuits are being or have been developed by Prof. Steve Furber’s group at our university.
You mentioned the dramatic reduction in energy required per instruction since early computers, did you say what the theoretical (and practical) efficiency of your nano-devices are in these terms?
--- The dramatic reduction was calculated based on modern CPUs, rather than my devices. I guess the energy required per instruction is inversely proportional to the clock speed.
Is it a fundamental necessity for electron propagation in these devices to have them make of semiconductors, or could the same geometric principles of operation be implemented with conductors, or even super-conductors? (to make devices even more energy efficient) Would super conductors use be limited to interconnects, even if new, room-temperature superconductors where discovered/created.
--- The ballistic devices (No 1 and No3 in my talk) could work with conductors or superconductors, at least in theory, but the requirement is that the electron mean free path is longer than the device size. The 2nd device, self-switching device, would not. Yes, using superconductors would be much more energy efficient, but many people think it is not possible to achieve room-temperature superconductors.
Finally, this might be a silly question: could the ballistic rectifier be used as an accelerometer by passing electrons in the reverse direction between it's "output" terminals and then measuring for a difference in energy (or preferential scattering) of the electrons between the 2 "input" terminals? (measuring accelerations in the left-right axis in the ballistic rectifier "movie")
-- - A very clever thought! Ballistic transport is needed for such a use. To check whether it is feasible, we can do a rough calculate. The electron velocity in semiconductors is about 100 km/second. If the distance between terminals is about 1 micron, then the transit time is about 10ps. I guess it depends on how much velocity change can be induced during the transit time. Of course, by rotating the device 90 degree, the velocity direction may change, and it is possible to design devices which are very sensitive to electron direction changes...
Please feel free to contact me if you have more questions. By the way, we have a studentship opening, so if you happen to know candidates who may be interested, please can you forward the following link to them? Thank you very much.
If answer's to any of these questions are already available online, I'd be happy for you to just point me in the right direction. Thank you for your time, Richard Lewis Cybernetics Undergraduate University of Reading