Here are some of the ideas from a talk by Eli Yablonovitch, UCLA, regarding the future developments semiconductor technology and its impact on Physics, Economics and Sociology. From his lecture on 25 October 2005 at UCLA. Here is a link to Eli Yablonovitch’s UCLA page.
In addition to a history of semiconductor industry growth, and feature shrinkage, there are some surprising points he makes in this talk:
- Lord Rayleigh’s law about the wave length light limits the size of the smallest feature that you can image on a chip. Turns out that is wrong. By using the fringe patterns and over exposing the image and the dark regions are incredibly narrow. The limit actually is that the spacing of the wires can be no closer than 1/2 a wave length. So features as small as molecules can be imaged on chips using optics.
- Devices will be as small as molecules. Sizes are likely to reach 10 nano-meters by 2010.
- Energy dissipated per function will be ~40kT [about one electon volt]. Currently energy is about 100,000kT.
- Wires are the problem now and in the future. Yablonovitch is working on optical methods of communicating on the chip using optical wave guides in SOI [Silicon on Insulator]. This is called nano-photonic integrated circuit communication.
- “Everything worth more than $1 will have an RFID tag and be tracked in a database.”
- “New processors will be multi-core. ” He did not justify this with the power cube rule of energy dissipation.
- “All functions of a laptop will shrink into a cell phone.”
- “The hard disk will be replaced by semiconductor flash memory.”
- “Voice recognition will replace keyboards.”
What will happen at the end of the roadmap?
That is, what will happen when we reach the end of the Shrinkage curve and functions are the size of molecules?
- “Answer 1: MACRO Level. There are more smart students being educated today, in Universities, in all countries of the world than ever before.
Some geniuses will emerge to lead humanity forward.”
- “Answer 2: MICRO Level. There are 10^11 neurons in the human brain capable of performing 10^13 operations per second.
In 2018 there will be 10^11 transistors on a chip; capable of performing 10^21 operations per second.
We just need to know how to program them; Please see Answer 1.”
That was the rather flip end of his talk. I was most surprised by the revelations about the ability to skirt the limitations of the wave lengths of light, and the prediction that we would reach the feature size limit by about 2010.