• Lead Telluride for Microprocessors?

    From Quadibloc@21:1/5 to All on Thu Aug 25 08:16:11 2016
    Unlike Lead Telluride, another material with a greater hole mobility than silicon has successfully been made into CMOS digital circuitry:

    http://phys.org/news/2014-12-germanium-silicon-cmos-devices.html

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Quadibloc@21:1/5 to William R. Frensley on Tue Jul 12 08:25:52 2016
    On Tuesday, December 16, 2008 at 10:23:53 AM UTC-7, William R. Frensley wrote:
    Quadibloc wrote:

    Of course, lead is objected to due to its toxicity - and tellurium is
    a very rare element, even if its price has, so far, been depressed due
    to limited applications. Silicon Carbide can function at high
    temperatures, but so far it can only be used for small chips due to a
    high prevalence of defects, and perhaps there are other problems with
    Lead Telluride of which I'm not aware.

    Where to start? There are two generic errors in such proposals. The first is the misconception that mobility translates into transistor drive current and thus into switching speed. In direct band-gap materials, the mobilities are high, but the saturated drift velocities are no better than silicon,
    and this is reflected in the magnitude of IDsat. We went through all this
    25 years ago with Fujitsu's HEMT, which never appeared in the promised supercomputers, but makes an excellent low-noise microwave amplifier,
    since the low-field region of the transistor is the source of most of the noise. (Low field regions are the only place you ever see the high mobility.)

    The other problem is the technology of narrow-gap materials. Don't imagine that silicon processing is remotely applicable. On a good day, you might equal the sophistication of the Ge alloyed-junction technology (but, after all, that was good enough for Kilby's first IC). Remember that the valence bands are the bonding states and the conduction bands are the anti-bonding states. Narrow-gap materials are just barely holding themselves together.
    (A bit of oversimplification, but a good guiding principle.) Things like selective doping by ion implantation are unlikely to be adequately controllable,
    since at the temperature required for annealing, the crystal is likely to produce defects that work to nullify your efforts.

    Finally, suppose that you can make a MOSFET structure. Can it be turned
    off? The bandgap is 0.29 eV, which means that interband tunneling will render the off-state energy barrier pretty much transparent.

    Thank you very much - belatedly - for your informative and helpful post.

    I admit to not being knowledgeable about semiconductor materials. Hearing about Indium Antimonide and similar materials being used for the "world's fastest transistor", and knowing that today's microchips need to use CMOS because bipolar and NMOS, for example, now produce too much heat, made me aware of the importance of having fast p-type transistors, not just fast n-type transistors - so it seemed to me that a material like Lead Telluride would be more relevant than Indium Antimonide.

    And they were making working thermoelectric components from Lead Telluride - and Fujitsu even thought the notion was useful enough to apply for a patent!

    But it isn't as simple as that, as you've shown.

    I figured that if undoped Lead Telluride wasn't a decent insulator, all that had to be done was to find an insulator of a similar lattice constant - deposit the transistor material instead of starting with wafers of it.

    But if they're too noisy to run at the low voltages needed to work with the low bandgap - or too slow at those voltages - well, that indeed does seem to put paid to it.

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Quadibloc@21:1/5 to Quadibloc on Mon Oct 10 07:42:41 2016
    On Thursday, August 25, 2016 at 9:16:12 AM UTC-6, Quadibloc wrote:
    Unlike Lead Telluride, another material with a greater hole mobility than silicon has successfully been made into CMOS digital circuitry:

    http://phys.org/news/2014-12-germanium-silicon-cmos-devices.html

    And now there's news about a use for a material with low electron mobility:

    http://www.ibtimes.com/worlds-smallest-transistor-just-one-nanometer-long-not-made-silicon-2427869

    http://phys.org/news/2016-10-materials-smallest-transistor-nanometer-carbon.html

    So molybdenum disulfide is useful for other things than lubricating the cylinders in automobile engines. But a transistor that is small, but not fast, would seem to be of limited usefulness - although there are always potential applications, such as nanotechnology, or chips that need to be complex more than they need to be fast.

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Quadibloc@21:1/5 to Quadibloc on Sun Apr 30 23:20:48 2017
    On Thursday, August 25, 2016 at 9:16:12 AM UTC-6, Quadibloc wrote:
    Unlike Lead Telluride, another material with a greater hole mobility than silicon has successfully been made into CMOS digital circuitry:

    http://phys.org/news/2014-12-germanium-silicon-cmos-devices.html

    I bumped into that article again, and noticed that the researchers had only just
    discovered how to make N-type transistors in Germanium. So it isn't a mature technology.

    On the other hand, Gallium Arsenide has been successfully used to build microprocessors. But it's a material with _lower_ hole mobility than silicon, so, now that microprocessors are at feature sizes where CMOS, as opposed to, say, ECL, is the only choice for thermal reasons... it's not really worth considering.

    Today, silicon chips have up to a billion transistors. However, while it is true
    that any new material is at a disadvantage, to be competitive, an alternative material would have to be developed to the point where a single die could contain on the order of 5.5 million transistors - the number on the CPU die of the Pentium Pro.

    Gallium Arsenide had a higher rate of defects than silicon, and so chips had to be smaller. And, as noted, it isn't really too useful now.

    Indium Antimonide has a very impressive electron mobility - but, again, the hole
    mobility is not as good as that of silicon, so while it's excellent for very high speed circuits with only a few transistors, for a modern general-purpose microprocessor, it doesn't seem to be useful either.

    Something with hole mobility too - Germanium or Lead Telluride - is what is needed. But both of those materials are in their infancy - while Lead Telluride is used for thermoelectric solid-state devices, transistor action has only recently been demonstrated in it.

    Indeed, it looks like we'll have to be content with silicon for some time yet.

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)