Ultra-Thin Alternative To Silicon

An anonymous reader writes "There's good news in the search for the next generation of semiconductors. Researchers with the Lawrence Berkeley National Laboratory and UC Berkeley have successfully integrated ultra-thin layers of the semiconductor indium arsenide onto a silicon substrate to create a nanoscale transistor with excellent electronic properties (abstract). A member of the III–V family of semiconductors, indium arsenide offers several advantages as an alternative to silicon, including superior electron mobility and velocity, which makes it an outstanding candidate for future high-speed, low-power electronic devices."

Re:But we are already running out of Indium...

[aramosfet]

I hate it when people post links to articles which require me to login or subscribe to read. Could you atleast tell me whats the "single material" he's talking about?

Re:Four words why this is useless.

[Dr. Spork]

We're talking about a 10nm layer across the surface of a chip - that's about a square centimeter. If anyone seriously complains about 50% of this being Arsenic, I would happily scrape it off and eat it in front of them. I don't think it would be a quantity large enough for the human eye to see.

Re:But we are already running out of Indium...

[Dr. Spork]

Yeah, but if we need layers of about 10nm, I'm quite sure we have enough Indium to make a cpu that's larger than the entire surface of the Earth.

Re:Why thin?

[the_other_chewey]

silicon != silicone, dammit.

About time

[Khyber]

I thought this possible a year or so ago while goofing off with diodes, but imagined the method needed to prevent leakage from the alloy would be too difficult to implement on a small scale.

Glad to see I could be wrong. Science never ceases to amaze and educate me every single day.

Re:Four words why this is useless.

[tulcod]

And even if RoHS would not be an issue, this is not the stuff that will magically make future CPUs faster. The performance bottleneck of integrated circuits is (usually) wire delays. See, signals get sent from one transistor to another using aluminium wires. Those aluminium wires have a bit of resistance, and a bit of capacitance with other wires and the silicon substrate. Heck, more often than not they have so much capacitance with other wires that they get rerouted to avoid signal interference. So if all wires act like (very small) capacitors, and they all have (a tiny bit of) resistance, it takes some time (think sub-nanoseconds here) to build up charge on the other side of a wire, and that is what causes the biggest delays. Routing all wires as efficiently as possible is a research area in itself.

That's incredibly naiive

[tygerstripes]

If this process is simpler and quicker to reach the fabs, and produces a notable performance increase, then it's worth it to develop. Someone will want to buy it, and that means someone will want to develop it.

Just to hammer it home: why do you bother, ever, to upgrade your hardware, knowing it'll one day be obsolete?