5 ms·
MIT creates tiny, 22nm transistor without silicon
- ori_b 14y ago"This tiny InGaAs transistor was mostly fashioned from normal semiconductor processes — molecular beam epitaxy, electron beam lithography, and so on." It's important to note that these are normal semiconductor processes for low volume manufacturing. These are not the high volume methods that, say, Intel would use. It remains to be seen whether the material will work well with common high volume ion implant and lithography technologies. The reason Silicon is the dominant material isn't purely because of it's electrical characteristics. We've already got better material that's costly to manufacture. GaAs, for example. In addition to good electrical properties, Silicon wins because it's extremely to manufacture chips with in bulk. To be fair, the article does touch on this.
- Osmium 14y agoSeconded; molecular beam epitaxy is certainly not an industrial technique! Here's a picture of some MBE equipment via Google: http://www.sandia.gov/LabNews/LN06-16-00/images/beam_pix.jpeg http://www.sandia.gov/LabNews/LN06-16-00/images/beam_pix.jpe... It's large, complicated, expensive, difficult to optimise and run, and ultimately can only grow very small amounts of (admittedly high quality) material (relative to the size of equipment and the time it takes to do the growing). On a simple level, it works by creating a "beam" of atoms which you can deposit onto a surface to grow your crystal layer-by-layer: this most definitely is not something that scales well to industrial sizes.
- jpdoctor 14y ago> molecular beam epitaxy is certainly not an industrial technique! Of course MBE is an industrial technique: A fair fraction of the semiconductor lasers on the planet are made this way, and probably a decent fraction of the microwave power amps for cell phones. The thing that is interesting here: There is an optimal stoichiometry for electron velocity in InGaAs, and it is hard to lattice match with a substrate for that composition. Usually there is magic in the buffer layers. (IIRC)
- Osmium 14y agoThanks for the correction; I had the article in mind, talking about it replacing Si transistors, and can't really imagine MBE being used to produce millions of chips, though I'd be happy to be wrong about that.
- jpdoctor 14y ago> and can't really imagine MBE being used to produce millions of chips Cell phones are example application for millions of chips. There's really nothing that magic about it. What is difficult is the large-scale integration of transistors, which is why you don't see CPUs (for example). There's a number of reasons, mostly relating to the nature of III-V materials. One example is the lack of a decent oxide: SiO2 is god's gift to electronics. However, Si is not a direct bandgap, so it is next to useless for devices involving light. It's why the bazillions of lasers on the planet are III-V.
- duaneb 14y agoSo, silicon is virtually everywhere. How would this affect manufacturing costs?
- jivatmanx 14y agoYep, earth is a rocky planet - composed primarily of silicate rocks.
- crististm 14y agoSo, what's new except for the size? GaAs has been around for _quite_ a while. Discrete high-frequency and low-noise transistors are made from GaAs for more than 15 years.
- jpdoctor 14y ago> for more than 15 years. Much longer than that. It was old news when I started grad school (1989). As for what's new: The link to extreme tech is (as usual) useless. If you hit the linked original article: http://web.mit.edu/newsoffice/2012/tiny-compound-semiconductor-transistor-could-challenge-silicons-dominance-1210.html http://web.mit.edu/newsoffice/2012/tiny-compound-semiconduct... you will find more info, but it's quite sketchy. The press-release say MOSFET, but I'd be quite surprised if they didn't really mean MODFET (oxides of III-V are a royal PITA). My guess is that this is a proof of concept for lateral 22nm devices, ie what would be state of the art litho for the silicon world. Edit: Dear mods - it would make sense to repoint the OP to my link above. Extremetech confuses the PR and does not enhance the original info.