Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
Compare cell phone plans using Wirefly's innovative plan comparison tool ×
IBM Hardware

IBM Scientists Find New Way To Shrink Transistors 100

MarcAuslander writes that IBM scientists have discovered a way to replace silicon semiconductors with carbon nanotube transistors, an innovation the company hopes will dramatically improve chip performance and get the industry past the limits of Moore's law. According to the Times: In the semiconductor business, it is called the 'red brick wall' — the limit of the industry's ability to shrink transistors beyond a certain size. On Thursday, however, IBM scientists reported that they now believe they see a path around the wall. Writing in the journal Science, a team at the company's Thomas J. Watson Research Center said it has found a new way to make transistors from parallel rows of carbon nanotubes.
This discussion has been archived. No new comments can be posted.

IBM Scientists Find New Way To Shrink Transistors

Comments Filter:
  • by Anonymous Coward

    Transistors, and their own US-based workforce.

    • by Anonymous Coward

      Sorry, I disagree. It is 3 things: what you said, and their tax liability.

    • Bravo, AC, pithy and humorous. +1.

      Unfortunately, below is some strange debate about libertarianism. So I am highjacking your top-of-the-thread comment to provide a space for people to discuss:
      -TFA/TFS
      -Semiconductors and ICs
      -Carbon nanotubes
      -Anything funny or witty about IBM, electronics, Moore's Law, etc.
      -Anything at all remotely on-topic

      I actually worked with carbon nanotubes back in the day when I was doing polymer and electronics research. I can attest to how squirrelly they are and how hard i
  • by trollingaround ( 4277543 ) on Thursday October 01, 2015 @08:07PM (#50640751)
    The Moore law that I know says that the number of transistor in a IC, double approximately every two years. Is there another one that specifies some limits?
    • by Anonymous Coward on Thursday October 01, 2015 @08:11PM (#50640785)

      It's the Law. Why do you question it?

      Oh, because you can think critically.

      The next person that quotes Moore's Law to me like it's anything other than an observation of one man that happened to be true for a decent period of time is getting a Lawgiver to the face.

      • Re: (Score:3, Insightful)

        Believe me, as a libertarian I love questioning the wisdom of all laws, but every time I do some social justice activist calls the PC Principal on me, and he's not nice.

        • Re: (Score:3, Insightful)

          by Anonymous Coward

          You guys are so boring, find a new routine or something.

        • by MyAlternateID ( 4240189 ) on Thursday October 01, 2015 @09:55PM (#50641383) Homepage

          Believe me, as a libertarian I love questioning the wisdom of all laws, but every time I do some social justice activist calls the PC Principal on me, and he's not nice.

          My own experience is that the Political Correctness types know how off-topic they are with regard to anything I tend to discuss (I mean, my stance that consenting adults should be able to do what they want whether or not I personally agree naturally includes lots of under-represented non-mainstream cultures). What I get instead is the morons who, despite knowing little or nothing about libertarian thought, feel an irresistable urge to form an opinion on it. Thus simple and easy-to-understand concepts like "minimal government" are conflated with anarcho-capitalism, even after I kindly explain to them that a government which cannot maintain rule of law and cannot enforce necessary regulations is less than minimal (that is, insufficient, too far in the opposite direction) and therefore not at all what I am talking about.

          In such discussions said morons tend to respond to what they imagine I must have meant, as though they know this better than I, rather than responding to what I actually said (see the arrogance?). I think a lot of people confuse terms accidentally. When they realize they have done this, they realize it after some discussion has already occurred. Then they continue to do it intentionally, because for them, this is somehow easier than admitting they made a mistake. After all, they already made up their mind what "libertarian" means and aren't interested in honestly examining how accurate their mental representation (stereotype) actually is.

          The world is filled with people who don't care about truth but do care very much about their own image and whether they can appear to be "right" in the eyes of others. Admitting fault is not compatible with this worldview, however noble and constructive it may be. Libertarian thought is especially problematic because if it caught on and became popular and well-represented in media, it would cause some drastic societal changes that would amount to a lot of powerful people losing their power. Thus, no one with any funding, power, or representation has any incentive to do anything other than demonize it. The method of demonization is simple: portray only its most extreme, least rational formations and ignore any reasonable method of applying its principles. Thus we are all anarcho-capitalists who don't want rule of law, publically funded police protection and firefighters, reasonable regulations, etc. No, it is "every man for himself", and if you aren't rich enough to hire private guards then you just get fucked.

          That's the way you discredit a credible idea: misrepresent it like hell, being careful never to portray its merits. Make no mistake, poltiics is a great big PR game and PR is so effective because most people are lemmings who will not conduct their own research before deciding what they think of a given position. If the average person took a skeptical attitude towards every political stance, never believing anything other than what they can validate with facts and non-fallacious reasoning, then we would not have the situation we experience now, in which the politician with the most funding who buys the most advertising tends to win the election.

          • If course there are "no true scotsmen", but there are also lot of anarchist types who label themselves as libertarian, the environmental movement suffers similar image problems.
          • by dbIII ( 701233 )
            One thing that doesn't help is people with a blatant personal agenda of greed and putting barriers in the way of others, such as Koch, calling themselves "libertarian" and paying a lot of other people to call them "libertarian".

            All of the insults are true if you pick and choose from enough people who call themselves "libertarian", even if they miss the middle by a mile.
            If you want to see if they are actually considering a philosophy or just greedy pricks ask questions about what rights people employed by t
        • by dbIII ( 701233 )

          Believe me, as a libertarian I love questioning the wisdom of all laws

          I thought the ones about what people get up to in the bedroom and women's reproduction were off limits to questioning by "libertarians", but what would I know.

    • Re: (Score:2, Informative)

      Physics? Or more specific in this case: how close atoms are arranged in typical semiconductor materials, and how few of them you need at a minimum to construct useful devices. That is: without practical issues like current leakage, isolation voltages, parasitic capacitance, etc, etc, making things not-so-useful (at best). Pro tip: try integer numbers first (or just very large numbers without counting exactly how many atoms go into your device).

      But please, if you know of a way to build IC's using 1/10-ato

      • by rtb61 ( 674572 )

        Shrinking will always have limits when you are looking at very narrowly defined conditions. More logical would be how to get transistors to be in more than two states, on or off, by say storing a frequency instead so that many states are possible in the same space. So instead of two states, say ten states in the same space with branches being frequency specific.

        • by Anonymous Coward on Thursday October 01, 2015 @09:33PM (#50641283)

          There is a good reason why that buys you nothing, and costs you much: Basically, the thermal dissipation of a device of a given geometry is what gives it state, by thermalising the state change, it prevents quantum annealing from returning it to the previous or indeterminate state. To increase the number of states you have to increase the voltage proportionally to the number of states, so to add one extra bit (remember a bit is just log2(possible states)), you have to double the number of states, and therefore the voltage; all well and good? No, because the dissipation of a device is proportional to the square of voltage, so you have doubled the bit density by quadrupling the energy consumption of the device.

          And this is not even taking into account the added complexity (more gates) required to at some point discriminate these levels to implement logic. This also roughly scales to the square of the number of states. So you take the square of the square of the number of states, or raise the activation energy to the fourth power of what it was, at least making the device 8x more energy dense or less efficient for a 2x gain (this is the hard limit of information theory, real numbers are worse).

          Now there of course is still room at the bottom to make these quantum annealing devices we call switches more efficient, but the way you are proposing is working in the opposite direction.

          If you want a computer to be reliable, that is compute things much more often than uncompute things, you have to have a thermal bias, so that P(compute) >> P(uncompute), which we do by setting up an entropy gradient and periodically saving the result of some combinatorial equation to a register where it's value is constantly reinforced by that entropy gradient (current flowing through the latch), or else held in that state by lifting it over an activation barrier (as in memristors and Flash). Either way energy is consumed in the process, as you lift it into a indeterminate state and allow it to relax into the desired determinable state. The clock and Vdd provide together provide this energy to allow this to function.

          -puddingpimp

        • You could possibly do something like that with light, but I don't think it works with electrical fields - and in any case the current clock frequency at the top end of electrical devices is about 3GHz - and you'd need frequencies an order of magnitude above the clock speed I would have thought. Light is about three hundred GHz, so you'd be on your way there if you wanted to distinguish different signals by frequency.

          I think. I'm kind of making this up as I go along.

      • Ooh-ooh-ooh! I just thought of a cool way to do this. At least in one dimension, probably two, maybe more but not all at the same time, or not all all the time. Not enough room to put down the proof and I need some sketches. This is so great!

    • by GrahamCox ( 741991 ) on Thursday October 01, 2015 @08:29PM (#50640875) Homepage
      It's a badly and lazily written summary. Moore's "law" doesn't specify any limits, but for a while Moore's law hasn't held true because of some unforeseen physical limits of the current silicon technology we use.

      This new technology may or may not deliver what it promises, but if it does, it will be a resumption of Moore's law, not breaking it. If anything, Moore's "law" was broken several years ago by the existing technology not living up to it.

      I put "law" in quotes because it's not actually a law, just a prediction, and a rather wishful one at that.
      • but for a while Moore's law hasn't held true because of some unforeseen physical limits of the current silicon technology we use.

        That was not unforeseen, everybody knows since decades that we actually are at the Limit of current technology.
        Improvements are now mainly done by multi core and hyperthreading technologies. Nont by "shrinking". The other way we do improvements is by switching materials from Silicon based to GaAr based (right now quite expensive) and/or technology e.g. nanotubes and/or optical tr

      • Isn't Moore's law just another name for exponential growth? It's merely a trend that doesn't say whether at some point in the future the growth will level off or crash.
    • The Heisenberg uncertainty principle is one profoundly limiting factor for all digital processing, but we're a few "Moore's law" generations away from that one.

      • The Heisenberg uncertainty principle is one profoundly limiting factor for all digital processing, but we're a few "Moore's law" generations away from that one.

        If we somehow get around that by inventing some new understanding of physics not yet envisioned, we may as well make Star Trek type transporter devices while we're at it.

      • by funwithBSD ( 245349 ) on Thursday October 01, 2015 @10:20PM (#50641513)

        Except that Moore's law has nothing to do with the size of transistors, but the number of transistors on a single chip.

        You can keep up by expanding the chip size, but then the yields tend to go down. If we could make perfect chips, the size could double every 2 years, although that would make for some very big chips indeed. Connections to the pinouts also become a problem as surface area expands faster than the perimeter.

        You could also go about it by making a true 3-d chip, instead of stacking individual chips on top of each other as they do today. That would make the external pinout problem even worse, as interior volume grows much faster than external surface area or edges.

        Shrinking the transistors is just the most effective way to do it, until you hit the red brick wall.

        • Electrical signals travel at a finite rate, which limits chip size if you're going to have high clock rates. At 3GHz, you're limited to about 50mm a side, but if you can build 3D, this could be a ridiculous number of transistors. Trouble is, you need to get the heat out too.

          • Electrical signals travel at a finite rate, which limits chip size if you're going to have high clock rates. At 3GHz, you're limited to about 50mm a side, but if you can build 3D, this could be a ridiculous number of transistors.

            No, you just put more cores and glue logic for them on the larger die, and then the signals don't have to go as far. Lo and behold, this is what we have actually done already, even though the die sizes are considerably smaller than the theoretical maximum.

            Trouble is, you need to get the heat out too.

            That's the easy part, you just put liquid cooling channels into the bottoms of the dies. The hard part is the interconnects.

        • > The Heisenberg uncertainty principle is one profoundly limiting factor for all digital processing, but we're a few "Moore's law" generations away from that one.

          >> Except that Moore's law has nothing to do with the size of transistors, but the number of transistors on a single chip.

          Only if you ignore the last 50 years of computing history and _how_ more transistors have been placed on a single chip. It's like saying that computing has nothing to do with binary logic: there are other available log

          • No I don't think I did. The post I responded to seemed to think that the law was for shrinking transistors, and it is not.

            You CAN do it other ways than shrinking the die, but you don't HAVE to to stay inside the parameters of Moore's law, which is what the original post I responded to claimed to be the case.

            Making the transistors smaller has been the easiest way up until now, so now they are looking at esoteric materials to get past the limitations of the current materials.

    • by mysidia ( 191772 )

      The pace has slowed from doubling every 2 years to more like every 2.5 to 3.

      Below 10nm the industry was on the verge of hitting a brick wall. The new development isn't breaking more's law; it's what is needed to advance, well-behind schedule of what Moore's law would have originally suggested.

    • Yes. Swinzig's Law: The number of people talking about how long Moore's Law will last doubles every 18-24 months.
    • I believe they are misinterpreting Moore's observation as a limit of you can only double the amount of transistors on an Integrated Circuit every two years. These are probably the same guys who put a cat in a box to test Schrodinger's cat theory (Must have really confused them when the cat died of suffocation every time)
  • by phantomfive ( 622387 ) on Thursday October 01, 2015 @08:09PM (#50640771) Journal
    The summary doesn't say what size, and the article merely says "40 atoms in width" (presumable carbon atoms? Who knows?)
    Apparently it's a technology that will coincide with the 7nm node [theverge.com].
    • Umm...don't 14nm chips already have transistors that are 1-2 atoms (width) by 10 atoms (depth)? I thought 14nm just refers to the distance between discrete components?

      I understand Fin FET will reduce the distance between components more, but afaik the "atom width" is already about as small as it can possibly get.

      Unless the "40 atoms" measurement is just using the diameter of the atom to measure the distance between components?

      • by Anonymous Coward
        • by Bengie ( 1121981 )
          I'm only 4 minutes in, but this is a cool video so far.
        • Thanks for posting this. I never would have seen it otherwise. I'm not in the hardware side of the business, but this video made me wish that I was. It was an amazing watch, and worth every second of staying up very late on a work night. :)
      • Unless the "40 atoms" measurement is just using the diameter of the atom to measure the distance between components?

        I don't think the article should be relied on to make sense.

      • 14nm is the size of the smallest part of a transistor, the width of the gate.
        Not the entire transistor.
        It's also known as "minimum feature size"

  • by Irate Engineer ( 2814313 ) on Thursday October 01, 2015 @08:14PM (#50640799)

    I've seen tons of articles like this over the last decade, touting carbon nanotubes as being the enabling technology for all sorts of improved applications.

    Can anyone actually point me to something that has made it to production utilizing carbon nanotubes? I'm not being snarky here - I'm really curious to know if any of this is actually getting off the workbench into mainstream use anywhere.

    Carbon nanotubes hit me as being a wonder invention like nuclear fusion; if we can build it it will be awesome, but we probably won't be able to build it for at least $DATE + 20 years.

    • by ClickOnThis ( 137803 ) on Thursday October 01, 2015 @08:21PM (#50640831) Journal

      Can anyone actually point me to something that has made it to production utilizing carbon nanotubes?

      The following looks like a good reference.

      http://www.researchgate.net/pu... [researchgate.net]

      • by idji ( 984038 )
        The researchgate article has the following Most products using CNTs today incorporate CNT powders dispersed in polymer matrices or deposited as thin films; for commercialization of these products, it was essential to integrate CNT processing with existing manufacturing methods.
      • That's nice, but which of those are actually commercial? Most of the "products" featured in that article are one-off research or demonstration tools.

        I've made something just like that. A few years back I made a "commercial" hazardous gas sensing system using CNT transistors, and installed it in an industrial chemical facility. This was based on technology I'd worked on as a postdoc which had been picked up by a Silicon Valley company and further invested in by DARPA. That's how things are supposed to wo

    • Carbonics is in Stage II VC talks, I believe.

      To the best of my knowledge, they can basically print CNTs in the 100's of nm's regime. They've got a web site, where I'm sure they tout their current node (equivalent) of products.

    • Well, remember back in the late 90s when IBM introduced copper interconnects [ibm.com]?

  • This is way cool.

    What does this technological advancement translate to in terms of porn?

  • The fact that moving electrons have an external magnetic field coupled with the fact that they can tunnel across short distances would, I think, tend to place a rather hard limit on how small you can make electronic components that still function in a predictable and consistent manner.

    Given that we are talking now about distances that can be measured in only a few dozen atoms of size, I'm pretty sure we're getting pretty darn close to those limits already, and I'm not sure there's any point in trying to

  • "get the industry past the limits of Moore's law"

    I don't recall any limits specified.

    • by smaddox ( 928261 )

      I think they meant "limits to Moore's law". Remember that Moore's law is not a law at all, it's just an observation. Furthermore, the observed doubling time has been steadily increasing for a number of years. Note that Intel missed their last targeted doubling.

      The driving factor behind Moore's law has always been economics. Once scaling becomes too expensive, it won't happen (at least not at an exponential rate). We're getting close to that point. 3D transistors have delayed the end of Moore's law for NAND

      • The driving factor behind Moore's law has always been economics.

        That is wrong: it was technical ability and knowledge.
        Once scaling becomes too expensive, it won't happen
        Actually scaling makes ships cheaper, not more expensive. First of all you need less material per chip, and most importantly if you can half the size of a transistor, you get 4x the yield from a given wafer of same size.

        NAND flash, the performance of logic circuits is usually limited by heat dissipation.
        No it is not. It is limited by the dis

      • Remember that Moore's law is not a law at all, it's just an observation.

        Well, so is Ohm's Law, but we still call it a law because it expresses a quantitative relationship between observations. That's the definition of a "law" in a scientific context.

  • by smaddox ( 928261 ) on Thursday October 01, 2015 @10:19PM (#50641503)

    The channel lengths were 60 nm. This is massive compared to the latest generation of CMOS (~14 nm).

    The confusion seems to come down to the fact that the SWCNT diameter is ~1 nm. However, 14 nm CMOS already uses FinFET's with channel widths of ~8 nm which is ~60 atoms.

    Regardless, the science article is actually about improved contact resistance, which is one of the major challenges associated with continued scaling of CMOS. However, they have only been able to show this improvement for p-channel devices, and they state clearly that n-channel devices present a much larger problem. If you want to replace CMOS, you need both n-channel and p-channel devices (not to mention fabrication yield needs to be as close to perfect as Si CMOS is). Thus my subject line (see above).

  • Did I really just see a "Paid Post" from Amazon embedded with the stories on Slashdot's front page?
  • ... when was the last time you saw Microsoft, Apple or Google doing basic computer research like this?

  • I'm only kidding, I know IBM does have a pretty big basic research program. I'm just shocked it's even operating anymore given how crazy things have been at IBM these days. Why does a white shoe management consulting firm (which is what IBM is trying to turn itself into) still have a lab?

    Don't get me wrong - I want them to continue and hopefully they'll get through this crazy period. But now that IBM doesn't manufacture anything other than mainframes and p-series, and no longer owns its own semiconductor fa

  • As often on /. people are unable to read the source they are then transforming in an article on /.
    IBM did not shrink transistors, they played around with nano tubes like many other researchers, but they are not closer to an application than any other research facility. Therefore, they did not find a new way, others have constructed similar approaches before, and I doubt that theirs is so different that it can be classified as a "new approach". It is not even a way to shrink transistors, because they fiddle

Introducing, the 1010, a one-bit processor. 0 NOP No Operation 1 JMP Jump (address specified by next 2 bits)

Working...