Mark, I used to state the GaAs problem in a simple form, like Charlie Brown, "there is no problem so big they cannot run away from". The cost of GaAs "real estate" is intrinsically about 10 times (minimum premium, the economic is much worse, Ga goes for between $500 to $750 kg, while SI (both semi-quality) goes for under $100/kg, today probably under $75, GaAs density is twice that of SI (roughly), it is weaker so you need thicker wafers, I doubt that boules much larger than 8" will ever be possible because the seed will break, furthermore, it is almost impossible to produce it defect free as Si is, thus lower yields). But, the higher mobility of charge carriers in GaAs and the lower energy dissipation per switching make it competitive at the marginal frequencies, but as CMOS "conquer" higher and higher frequencies (through features size reduction), GaAs people are forced to even higher frequencies to stay in the business. There is one "place" to hide, once we get hybrid opto/electronic chips, since you cannot make a light emitting diode (or or solid state laser) from SI, but you can from GaAs, future circuits that combine light generation and modulation as well as signal processing on the same chip may be kept as a unique GaAs domain. I don't think we are there yet.
Zeev |
