Gallium Arsenide

[Daovonnaex]

Seymour Cray was researching this for the next-generation of semiconductors before his death, and I believe that the Cray T3E's microprocessors use gallium arsenide instead of silicon. I'm essentially curious as to the advantage this offers over silicon, and when we can forseeably expect such a technology on desktop processors (I know that Motorola is already developing this).

 

[cyclones]

Gallium Arsenide has been used for ages.
Most of the Cray TE3s only use 600mhz chips .....but use MPP = Massif Parallel Processor systems.

 

[rimshaker]

Ideally, GaAs would be widespread right now instead of Si. But it's not cause there's major disadvantages too. In a nutshell, GaAs material can reach speeds way beyond Si due to the fact electron mobility is almost 8 times greater. It is also a direct bandgap semiconductor, meaning it is suitable for devices utilizing optics and microwaves. Unfortunately, the disadvantages prevent them from widespread use. Si is basically sand, there's plenty of it!! GaAs wafers have to be constructed, also requiring specialized equipment, making it much more expensive to mass produce... cost is the #1 issue. Integrating GaAs with Si is very difficult too, due to differences in lattice structure and melting points, and others. To make transistors, there must be a barrier layer between the gate contact and the substrate. Si has a natural barrier with Silicon oxide, SiO2, virtually free to make. GaAs does not, meaning more unique materials are needed, again, increasing cost and length of production.

I've also read about Motorola's recent achievements. They found a stable way of mounting GaAs on top of Si. If i recall correctly, they used a method by which the lattice structure was actually 'stretched' to more closely match that of Si... a breakthrough indeed. Now, GaAs can be used using present equipment and technology without major modification. Optical and RF devices now have a way of being integrated with processors... at least in theory right now. I'm certain there's plenty of unforeseen problems yet to be addressed. But the integration of GaAs is a nice first step. I suspect we won't be seeing widespread use until mid-decade at the very earliest.

 

[kylef]

I agree... there just isn't a good substrate... yet. Same with the Nitrides.

 

[cyclones]

One of my pals in the chemical industry, was selling GA to Motorola in Scotland UK over 10 years ago.
They have been busy over here in what we call "Silicon Glen".
This is as to your "Silicon Valley".
The cost of certain chips is of great consequence to commercial computer manufacturers.....
but....not always to the military requirements.
The technology has been in production for some time......but not for "Joe Public".

 

[rimshaker]

There are GaAs devices in every cell phone you see out there. Any RF related device you encounter most likely integrates GaAs, or some compound form of it.

 

[CSoup]

How about SiGe, what are the advantages and disadvantages. I know IBM is doing a lot of research in this area. Mostly for RF stuff I guess.

 

[rimshaker]

Silicon Germanium is doped into the base regions of BJT and HBT devices for a nice performance increase, mainly due to an increase in electron mobility in the combined material. There really arent any disadvantages other than it's an extra step in the fab process.

SiGe is pretty much mature technology at this point and widely used. Several companies are simply just competeing with each other these days to see who can make the fastest SiGe-based devices. Yea, applications are in communications. GaAs can easily beat SiGe in speed, but as discussed before about high costs and integration with Si processes.... most companies are pushing ahead with SiGe for mass production.

 

[cyclones]

Most of the development for processing is now in the realms of Bio and living cell technology.
Only the ragheads will be using silicon chips in 20 years.
The cost of technology will lead the development........ I think!