Solution to

All Quick Questions / Class Exercises to

2.2 Silicon

	What is the approximate lateral size of one transistor in an IC?
		Given that state-of-the-art transistors have smallest dimensions of 32 nm, one could guess that that the total lateral area would be in the order of (100 - 200) nm². Be that as it may, an answer of "below 1 µm²" is acceptable
	Why are there no 16 GB memory chips now?
		Because that would be a 16 ×8 = 128 Gbit chip. With a chip area of 1.5 cm² with 1.28 cm² for the memory cells (the rest is needed for the periphery), one memory cell would be 1.28 cm²/1.28 × 10¹¹ = 1.28 × 10¹⁴nm²/1.28 × 10¹¹ = 10³ nm². In other words, one memory cell, needing at least one transistor, would be about 30 × 30 nm² and that is not yet (2008) feasible.
	What properties should a semiconductor have for making IC's?
		Well. there is no end to the list: Bandgap not too small to make it too temperature sensitive around RT, or too large to allow easy doping and contacts plus some conductivity. So Ge is out now but GaAs etc, are still in. Si is OK, but a somewhat higher bandgap would be preferable. Large defect free and "cheap" single crystals must be possible. Pretty much all semiconductors but Si are out. A process compatible dielectric must exist with very high breakdown field strength and suitable dielectric constant. It's very difficult to beat SiO₂ here. Big advantage for Si. Precise doping establishing a precise conductivity must be possible. Many semiconductors meet this requirement, but Si is one of the easiest to dope. The mobility of the carriers should be large for high-frequency applications. Si is mediocre in this respect, but still good enough for most applications.
		Conclusion: There is no other semiconductor out there that comes even remotely close to Si as the best compromise.

	What exactly produced complexity and market growth rates of 30 % for more than 30 years (for Si IC's)?
		There is only one reason: The price per function (e.g. € per bit of memory space, per logical operation, per numbers of operations per second, per kWh, ...) comes down substantially.
	Where will it end?
		Nobody knows. The absolute limit is the size of an atom » 0.3 nm. How many atoms does it take to make a transistor = switch? A few hundred, occupying a volume of < 10 nm³ are enough. So we are a far cry from the limit with our present 2-dim. integration of logical switches.
		But maybe the cost limit (see question above) is reached before the technical limits are even close? Check the link for details
		Wherever and whenever the limits will be felt, one thing is quite likely: Moore's law, maybe with a somewhat slower rate of growth, will go on for at least until 2015 because all essential ingredients are already in the laboratory stage.

	What do you know about MEMS?
		Probably nothing yet. That will change. But you have definitely seen MEMS in action and very likely own MEMS products:
		Many beamers contain a MEMS chip with roughly 1 Million little mirrors that can be individually addressed to process one pixel of the light you see on the projection screen.
		The acceleration sensors in your car that tell the microprocessor that it is time to inflate the air bags, as well as the gyros (sensors for rotational speed) that allow the microprocessor to calculate how to individually brake the wheels in order to keep you in line when you start to slip, are typical MEMS products.
	Are there any other uses of Si you know off (or can find quickly)?
		Let's see: We had solar cells and MEMS. What else? Si wafers are the cheapest perfect substrates for all kinds of applications (in particular optics, including X-ray "optics"). Si single crystals are used as perfect Bragg diffractors to produce, e.g. monochromatic Neutron beams. If you make nanoporous Si and fill the pores with something containing a lot of oxygen (e.g. KMnO₄), you have produced a high explosive with up to 3 times more bang per kg than TNT. A big project was launched to explore that as an integrated fuse for blowing up air bags. Si "nanwires" might be the ideal anode for better Li ion batteries. And so on - activate this link for some details.

With frame

Exercise Basic 2.2-1 All Quick Questions / Class Exercises to 2.2 Silicon

Moore's Law: Break Down Scenarios

Exotic Uses of Silicon