Fun with niobium

Hi everybody,

Today I am going to provide some background information on the stuff we make our processors out of. As you’re probably aware, conventional processors, like Athlons and Pentiums, contain a lot of silicon. Maybe you have wondered why this is. Why silicon and not beryllium? Or lead?

There are many good reasons. One of them is that it is particularly easy to make a very important device, the transistor, using silicon. In fact even though Bell Labs invented the transistor, it was Texas Instruments who first commercialized it, mainly because they figured out how to switch from the Bell Labs style germanium version to a nice, simple and cheap silicon version.

The silicon based transistor worked so well that it, and its descendants, have been the workhorse of nearly every processor used from the very beginning of microprocessors through to today.

I say nearly because there is a different way to build processors (using metals) that has been lurking in the background for about 20 years, showing a lot of promise, but never taken up by any of the major processor companies as the basis for a product line (TRW’s efforts notwithstanding).

These metal-based processors are quite unusual. One thing about them is that in order for them to operate, they need to be cooled to nearly absolute zero using special refrigerators. At these very low temperatures the metals that form the processor’s circuits become what are called superconductors. Whenever anyone says “superconductor” just think “really cold metal”.

Why go to the trouble of doing something like this? Historically arguments for metal-based processors have been that (1) since they’re made out of superconductors, they generate much less heat than conventional processors (true); (2) for some technical reasons you can operate at clock speeds up to about 100 GHz without alot of problems (true); so if you want a really fast, really low power processor, here’s a way to do it.

History says though that these benefits by themselves just aren’t enough to even TRY to commercialize this type of metal-based processor, especially given the temperature constraints (which probably limits applicability to “capability class supercomputer” type stuff). Just too much risk for not enough reward.

BUT what if there was a way to use the “really low power & really fast” benefits to gain some other really valuable advantage?

What D-Wave has done is begun with the standard approaches to building metal-based processors and modified them in such a way that these processors use quantum mechanics in order to accelerate computation.

When you look at one of our processors with your naked eye, what you see is a fine mesh of metal circuitry (the metal is niobium) on a 5 square millimeter chip. This metal circuitry is the top metal layer in a multi-layer metal/insulator/metal/repeat many times/ sandwich. This top metal layer contains things like contact pads for connecting the chip to wires that carry information to and from the processor.

Here is a picture of one of our processors in a test system, to give an idea of what it looks like. The chip in the middle is 5 square mm in size.
img_0748.jpg

Cool eh.

So why are we using these metal-based processors as the basis for our approach to building quantum computers? There’s one underlying fundamental reason: we know how to build really big things using superconducting metals that behave quantum mechanically.

Superconductors are the only type of material that we know of where big lithographically defined devices (like really big. Like centimeter on a side big.) can be built that behave just like they were atomic-sized. The reason for this behavior is highly technical – is has to do with the types of particles in the material. In a superconductor all of the “particles” that carry charge around can exist in the exact same state, so when you look at a whole lot of these particles (many trillions) it can be just like looking at only one (which is “very quantum mechanical”).

This property allows us to build circuits out of superconductors that, if we are really smart and really careful, can be made to act like “circuits of atoms”. We can use the fact that really big things (which we can easily build today using conventional fabrication techniques) can be made to behave like really small things to try to build real quantum computing architectures.

75 thoughts on “Fun with niobium

  1. It’s really too bad the low power of the processor is offset by the high power that is used to cool the sucker down to near absolute zero for extended periods of time.

    I guess it’s still really fast. :)

  2. For the cost involved it would seem wiser to just do a large industrial diamond substrate to pump the clock through the roof instead of laying on lots of effort perfecting this technology. Sure in the future it will be necessary but the wise intermediate step would be a non leaking carbon instead of silicon on the same process technology.

  3. This is not ‘quantum computing’ as the science world generally understands it. this is an example of a marketing ploy to generate money and media interest. If this was a real quantum computer you can bet that governments would be lining up to purchase the technology and the inventors would have a Nobel prize by tomorrow afternoon.

  4. Folks, this is “superconducting” not “quantum computing”. If it were quantum computing then perhaps the author can let us know how many qubits his system expects to have. But I bet it has zero and it uses the same transistor model your age old 386 had.

    It may be true that superconducting metals behave “quantum mechanically” and this somehow accelerates computation but this is not “quantum computing” is the normal use of the term. Its “accelerated computation using phenomena that look like quantum mechanics”.

  5. Err this is not a quantum computer (with qubits etc). Just a very fast classical computer using metals rather than silicon. Also the superconducting thing is not new, this has been done like 20 years ago but got redundant with smaller transistors(something with the increased electric field of the smaller devices cancelling the high temperature coefficient, it have been a while since my introductory microelectronics course). Superconduction has become important lately with regards too power consumption. I remember a couple of years ago IBM(or possibly someone else) at very low temperatures ran a simple microprocessor at 13 Ghz while consuming only nanowatts of power, but apparently it was very expensive and impractical too scale it up. At best these people has find some way too run a computer practically at 100 Ghz.

    REAL quatum computer needs quantum wells, isolation etc. which is a total b1tch to construct and shield, we are talking on electron and subatomic scale here, not mere nanotechnology.

  6. “factor a big prime and i will believe you have a quantum computer under the hood”
    No problems: for any given prime n, the factors will be 1 and n. Can I have my Nobel Prize now? :P

  7. “4. helmut katzgraber – August 20, 2006

    seeing is believing… factor a big prime and i will believe you have a quantum computer under the hood.”

    LOL. Bill Gates, is that you?

  8. “No problems: for any given prime n, the factors will be 1 and n. Can I have my Nobel Prize now?”

    The problem is you have to PROVE that n is actually prime by factoring it. If you can prove that n has only 1 and itself as factors, you get the prize.

  9. Alex, cb, and Darv, it is quantum computing. In the D-Wave system, calculation times expand as the polynomial, not exponentially. This is a key defining advantage of QC over classical computing.

    Search on D-Wave and start reading. The entire collection of atoms in the qubit system behaves as a single quantum entity. This is far different than gating a transistor at super-high frequencies.

    While factoring primes is the most common QC application that the popular science press likes to use, these people are talking about finding exact solutions to the electronic states of fairly complex molelcules: http://www.lbl.gov/Science-Articles/Archive/CSD-quantum-chemistry.html (notice the .gov domain, BTW)

    Finally, the nanoscale regime is adequate- subatomic fabrication is not needed- as evidenced by many successful experiments using various supra-atomic devices to trap, manipulate, and extract states from quantum systems.

  10. Most of the comments above are ill-informed and reactionary. Quantum computing is in its early infancy; to assert with certainty what it is or is not at this time is utterly fatuous.

    Superconductivity, according to our current understanding, is an essentially quantum phenomenon, relying on Bose-Einstein condensation of particle states.

    Given the groupthink which too often pervades academia, the author is to be congratulated for trying something new. As for attracting media attention, that is a standard function in the marketplace.

  11. NOT quantum computing. This is an attempt to hijack an existing well defined term for marketing and publicity. Very lame.

  12. People who are stating that this is “NOT quantum computing” are dead wrong, because of course the existence of Bose-Einstein statistics (and Fermi-Dirac statistics) are implied by quantum mechanics (& special relativity! There’s a great mathematical paper by Pauli which shows this).

    Correct me if I am wrong, but quantum computers operate on the idea that fermions, having such stringent requirements on occupancy of states, can be made to change their state simply by changing the state of a nearby related fermion? And the quantum computers people are attempting to build these days have decided to use electrons as the fermions.

    So this computer will use bosons instead of fermions, because you can create “large” bosons which are “easy” to manipulate. This computer will obviously be light years behind a fermionic quantum computer because they will be limited by “how small” their bosonic condensates can be made and manipulated.

    But also, is the gentleman who stated that superconducting is a bose-einstein condensate phenomenon correct? This is not my field of expertise.

  13. Seems to me all the naysayers here are a bit jealous of your work.

    Quote “It may be true that superconducting metals behave “quantum mechanically” and this somehow accelerates computation but this is not “quantum computing” is the normal use of the term. Its “accelerated computation using phenomena that look like quantum mechanics”.”

    If it quacks like a duck and walks like a duck, it is more than likely a duck.

    Ignore the haters, hope your project works out great and leads to new and better computing models.

  14. Since everyone is so certain that this isn’t quantum computing, I would like all those people to give a solid definition with definite limits as to what quantum computing entails, then when you all realize that the definition of a science that is so new can and will vary greatly, then hopefully the redundant comments about how this isn’t quantum computing will end.

    Also if you made a computer the size of Earth ran by mice it would not give you only “42″ it would give you The Question.

  15. I have a degree in Physics and I can state with great certainty that nothing in this post plausably asserts that this system can perform quantum operations on quantum information (in the form of qbits). There is no mention of reversibility of the system, isolation, or how they intend to excite basis states and perform the final measurement. Furthermore there are no numbers that would allow us to know how many qubits the system can simultaneously support, how they are entangled, and what types of quantum gates are constructed (two-qubit, three-cubit, etc?).

    There are many things that a QC could perform with a better asymptotic space/time complexity than a classical computer, the most famous of which being Shor’s period finding algorithm. This would be a better test than directly finding prime factorization of large integers because it requires less Q-bits and is really all that’s necessary (for the quantum computation, at least) to crack RSA. It would be most excellent to have a public test of this period finding, giving the team a very short turnaround time (in line with their claimed number of Qbit gates).

    I don’t think anyone can take this seriously until the purported quantum capabilities of this processor are fully enumerated.

  16. These guys aren’t exactly crackpots and they are trying to do the real deal. I know because I worked across the hall from them, and they started out in the basement of the Dept. of Physics at UBC up in Vancouver. They DO have some pretty good experts professors there who are experts in related fields.

    Do your homework. I’m not sure why it’s called DWAVE, but I’ll bet it has to to do with the d-wave states in superconductors, and quantum mechanics is involved.
    Their web site is here:
    and they are funded by Draper Fisher Jurvetson to the tune of ~$20million, not exactly chump change and not exactly no-name VC’s.

    Whether they’ll succeed is another matter…

    (Come on, how many of you actually have expertise in low-temp physics, superconductors and quantum physics, computation???)

  17. ASCIIDuck, read the last paragraph:

    We can use the fact that really big things (which we can easily build today using conventional fabrication techniques) can be made to behave like really small things to try to build real quantum computing architectures.

    The person isn’t claiming they have a quantom computer, just the idea that they might be easier to build if you build them large enough.

  18. Just because something uses quantum mechanics to do computation doesn’t mean it’s a quantum computer. Heck, even your Intel Pentium uses some quantum physics in its design. The word quantum computer as it is generally used means a machine which uses states of particles to represent bits, taking advantage of the fact that a single particle can be in a superposition of states to have each “quantum bit” (qubit) represent a superposition of “normal” bits. This is what makes a quantum computer powerful – n qubits can actually represent 2^n “normal” bits, meaning that if you know how to do certain operations on the qubits, you can use them to do certain types of computations exponentially faster. Factoring numbers happens to be one such computation for which there are well known quantum algorithms using gates that people can actually build.

  19. The quantum entaglement vision of quantum computing is like those old shows where the guy kept a bunch a plates spinning on poles. Amusing, but it just doesn’t scale well.

    D-wave is possibly (assuming they are not crackpots) doing something more directly like interference of particles going through a slit, which is basically a computation over the all different paths the particles can possibly go through. The description of some d-wave superconductor research (eg. http://www.iop.org/EJ/abstract/0034-4885/63/10/202/) suggests they could be doing quantum computation be detecting inteference over a large number of more stable quantum particles: their array of loops. The computation state of each loop in the array would affect the phase of the signal going through that point.By using superconductors as virtual particles they will be able to set up the initial conditions for the computation. They configure the strengths of the magnetic couplings between the the loops to define the computation. Then they just measure the interference pattern that comes out when they pass a sine wave electrical signal through.Because they will be able to array the loops into cycles, they will be able to do a wide variety of algorithms. It may be Turing complete, but there seems to be a practical limit of how coherent the wave passing through the longer pathways will be.

  20. For a quantum computer to work the qubits have to be protected from environmental noise that would otherwise cause them
    to decohere. The interaction of the qubit and the environment
    is typically given by a dipole coupling -

    Energy shift of two quantum states
    = dipole * property of system * fluctuating field in environment

    When plancks constant times the random energy shift is >> pi,
    the system is decohered since the relative phase of the quantum
    states are randomised and the interference between them will
    no longer be meaningful.

    Interference effects are what makes quantum computers
    ‘more powerful’ than classical computers.

    Now since a dipole is a physical length times a charge, or an area times a current, etc. one line of thinking would say use a small
    dipole such as the magnetic moment of a single atomic nucleus or electron,

    A macroscopic quantum state in a lithographically defined
    piece of niobium has the potential for a large interaction with the environment, so the quantum coherences may not persist for useful periods of time.

    It’s interesting to know what sort of acceleration d-wave proposes, and what figures of merit they have for their system.

    eg. how long do their coherences persist in comparison
    to how rapidly they can perform operations with them?

    what fidelity of gate operations can they achieve?

    what is their T2/gate operation time?

    (but yes, dilution refrigerators and niobium are ‘cool’)

  21. «17. helmut katzgraber – August 20, 2006

    ok smart asses. it is late at night and i am in kyoto right now. i meant a “product of large primes”…..»

    So did gates, ergo my question.

  22. Thanks everybody for the comments! I’d like to respond to a couple of issues that were raised.

    1. (for experts): The machine we’re building is NOT a conventional classical RSFQ processor. My post (aimed primarily at beginners) was just to provide information on some history of superconducting processors as context for later posts.

    2. (sort of for experts): The machines we’re building are real quantum computers, yes with real qubits and everything. I’m going to describe quite a bit about the processor architecture in follow on posts so stay tuned.

    3. (for experts): Again this is something I’m going to describe at length in a follow on post, but I’d like to flag it here because it was raised in a couple of comments. There is more than one model of quantum computation, and the one most people are familiar with is the “gate model”. Our machines are not gate model machines. They use a different underlying (but equivalent) computational model.

    4. About the low power comment: It doesn’t take much more power than what you get out of a wall socket to cool a chip to milliKelvin and keep it there. The fridges, even though they look imposing, draw very little power.

    5. The comment about this being a marketing ploy: If you believe that quantum computers will ever be built, someone’s got to be first. And it’s more likely that this sort of innovation would come from a start-up than a big company (innovator’s dilemma) or academia (the project’s too expensive, long-term and interdisciplinary).

    6. About our name: Although we’re called D-Wave everything we do is built using niobium (which is not a D-Wave superconductor).

    Good comments, keep ‘em coming!

  23. I still think that the people who believe that consciousness exists due to quantum orchestrated resonance are right, and so be careful what you do, perhaps AI was only difficult using electronic computing, it might be frightfully easy with quantum computing.

    Just a bit of intuition, sorry I don’t have the math or physics to back it up.

  24. Circa 1990, Sadig Faris (founder) and several others from the Hypres technical staff gave a presentation to Paul Schneck and many others at the NSA Supercomputing Research Center. At that time, Hypres had a Josephson Junction full 32-bit combinatorial decoder running in liquid helium. Their talk was well received, with excellent questions and considerable interest. Despite repeated called to the SRC, the NSA seemingly decided that they were not interested. I’d previously read the 9/04 IEEE paper and the 8/05 NSA paper, which you graciously link to, above, at…: http://tinyurl.com/mn2jh … Please pardon my deep skepticism, but it is not obvious that even now, twenty-five years later, there is any greater interest in this area, on the part of NSA, which remains the key and mandatory purchaser of first resort. …sigh…

  25. Oops: 1. That NSA / SRC presentation was ”only” about 15 years ago, not the 25 which my bad math claimed: It only seems like 25… 2. Also, the line beginning ”Despite repeated called to the SRC…” (sic) should be “Despite repeated calls to the SRC…”.

  26. Superdiamagnetism (or perfect diamagnetism) is a phenomenon occurring in certain materials at low temperatures, characterised by the complete absence of magnetic susceptibility and the exclusion of the interior magnetic field. Superdiamagnetism is a feature of superconductivity. It was identified in 1933, by Walter Meissner and Robert Ochsenfeld (the Meissner effect).

    Superdiamagnetism established that the superconductivity was a stage of phase transition. Superconducting magnetic levitation is due to the Superdiamagnetism (which repels a permanent magnet) and flux pinning, which stops the magnet from sliding away.
    ————–

    superconductors hold secrets to the future …
    not just in computers … but transport …

    some bloke came into our uni last year and thought he could build some type of aircraft like a flying saucer… using superconductors…

    hahaha

    the future is not very far away

  27. Whoa, reading this made me feel like I was at an airshow (zoom, zoom, zoom – over my head). So my question is… does this represent anything more, in practical terms, than a speed jump (even if it is a large speed jump)?

  28. That was a really good read,
    All i can say is that it has to start from somewhere.
    As long as i can get more frames in Doom3 ill be happy :-)

  29. Looks to me like this will not be the holy grail of quantum computing, but is none-the-less a big leap forward.
    I think mainly for quantum computing development.
    Even if we don’t reach the grail of zero time qubit switches, this will be be a classically based computer that is a lot faster than our current silicon and will apparently exploit qubits on a large scale, which I have a hunch would eventually improve function when on a smaller scale.
    This means when its working, we can start developing the best and safest techniques to use them (what happens if you give a quantum computer the wrong question ? Maybe its simply not possible to construct a wrong question for a logically zero time function.)

    It seems frought with paradoxes, I wish them all the best interpreting their results.

  30. It sounds to me like some people in a lab are thinking some good 5|-|17 up to make computing in today’s society a breeze. But lets not forget the past in such matters that things created in the past though similar may be completely different. Sceptesism is normal for this kind of work because you can’t readily see a benchmark to compare it to something else. They have stated and I quote “What D-Wave has done is begun with the standard approaches to building metal-based processors and modified them in such a way that these processors use quantum mechanics in order to accelerate computation.”
    If and when such a thing is made available to the public which like all things created for the first time placed on a marketing scheme will be ungodly expensive and noone is going to pay for it. I am not a science genius nor claim to be even remotely close. From what I understand the only reason why silicon cannot be used to create such high conductivity is that its created on earth which means impurities in the silicon causing it to superheat. If such a thing is grown is space i.e EXPENSIVE. Lets give a nice pat on the back to these gentlemen and ladies if such exist in the office/lab there :) For at least trying to give us a chance at something that will push or systems today into the stoneage as it stands. Good luck to the D-Wave team and all your efforts, I truly hope you find what your looking for. PS It all tastes the same, it just smells different

  31. It’s a shame to see people triggering right away without even looking at what is proposed here. Your “superconductor quantum chip” looks pretty much like my phone card chip, only that my chip was not used enough to show the scratches on yours.

    People, wake up!

  32. well dammit, i’m just happy we are finally making some kind of progress in the quantum direction. let’s get off the chalkboard and in the factory one small tip-top at a time.

    i can’t wait to hear more details.

  33. Hey, think in reverse..

    We need processors for our space probes and long distance missions where intersteller tempretures are going to get near absolute zero +app 5-10F so coolant isnt a problem, but the speed and technology transfer this allows should give us high speed/volume data from these probes at a minimal power cost such as another voyager like mission.

  34. Dammit, here NOBODY knows what a quantum computer is and yet EVERYONE is trying to give opinions… Guys, go learn some physics and then try to post.

    This article is about a better classical computer, that uses metal instead of sillicon. Good. But it is nowhere near a quantum computer.

    You cannot use a term like that just because it looks familiar. Do we say that a pot is electronic, because it uses electricity? That is what this paper is claiming.

    Quantum computing is a separate field of research and merely using a mechanism that is described best in terms of quantum physics DOES NOT make it quantum computing, percebe?

  35. I think the thing people are looking for is that magical “order N space search in order 1 time”.

    It seems that some of this tech might actually provide this although not close to order 0.

    He says it has qubits. we’ll see.

  36. “The computation state of each loop in the array would affect the phase of the signal going through that point.By using superconductors as virtual particles they will be able to set up the initial conditions for the computation. They configure the strengths of the magnetic couplings between the the loops to define the computation. Then they just measure the interference pattern that comes out when they pass a sine wave electrical signal through.Because they will be able to array the loops into cycles, they will be able to do a wide variety of algorithms. It may be Turing complete, but there seems to be a practical limit of how coherent the wave passing through the longer pathways will be.”

    As someone with the expertise of an interested amateur, I’m just barely able to follow the discussions here (I just stumbled upon this site, and it’s fascinating) but, doesn’t the above describe more of an analog computer? Put a signal in, let it get modulated by various attenuators, and then read out the answer?

  37. Some people who post stuff here need to read “A Shortcut through time, The path to the quatum computer”by George Jonhson. To get a better basic understanding of quantum computers.
    Every system that exploits qbits to accelerate computing time is a quantum computer in my eyes. Although it may not be a full-blown quantum computer.Every computer that makes computer time grow polynomial instead of exponantially is great ! Think of all the possibilities…

  38. DWave may be on the edge of building an infinite computing circuit, but Gordie forgot to mention dipping the other electrodes in the hot cup of tea.

    That’s important.

  39. I wonder if you could setup enough qubits to simulate a classical computer.
    It might be possible to test the halting problem – make Windoze crash in a split second.

  40. All machines are mechanical in nature. Machines are metal. Current needs to run to the Atoms. Data needs to be input to the Atoms. Shape takes place. Reworld small scale applications are tested.

  41. i know nothing but i do have street smarts, is it a logical step forward? yes => Are you doing anything this deep? NO. Leave the science fiction at home, and let them try. 300years ago the thought of what he have now was witchcraft. leave the design and physics work to the guys with the balls to try it. Instead try baffling your nextdoor neighboor with bullshit. Much more fun.

  42. Pingback: Big-scale quantum-computing « IT, life and me

  43. Cheap viagra works in much the same way as Sildenafil Citrate, but is much faster, lasting up to 36 hours ciallis free trial . For cialis online cheap very fast delivery Europe.. No prescription required buy viagra tabs online! Fast to get cialis pills online… Find out more about tadalafil at the official vigra cheap generic site

  44. Hi, good post. I have been thinking about this issue, so thanks for blogging. I will definitely be subscribing to your blog. Keep up the good posts

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s