Quantum Computing Hardware Teardown

Although quantum computing is still in its infancy, enough progress is being made for it to look a little more promising than other “revolutionary” technologies, like fusion power or flying cars. IBM, Intel, and Google all either operate or are producing double-digit qubit computers right now, and there are plans for even larger quantum computers in the future. With this amount of inertia, our quantum computing revolution seems almost certain.

There’s still a lot of work to be done, though, before all of our encryption is rendered moot by these new devices. Since nothing is easy (or intuitive) at the quantum level, progress has been considerably slower than it was during the transistor revolution of the previous century. These computers work because of two phenomena: superposition and entanglement. A quantum bit, or qubit, works because unlike a transistor it can exist in multiple states at once, rather than just “zero” or “one”. These states are difficult to determine because in general a qubit is built using a single atom. Adding to the complexity, quantum computers must utilize quantum entanglement too, whereby a pair of particles are linked. This is the only way for any hardware to “observe” the state of the computer without affecting any qubits themselves. In fact, the observations often don’t yet have the highest accuracy themselves.

There are some other challenges with the hardware as well. All quantum computers that exist today must be cooled to a temperature very close to absolute zero in order to take advantage of superconductivity. Whether this is because of a reduction in thermal noise, as is the case with universal quantum computers based on ion traps or other technology, or because it is possible to take advantage of other interesting characteristics of superconductivity like the D-Wave computers do, all of them must be cooled to a critical temperature. A further challenge is that even at these low temperatures, the qubits still interact with each other and their read/write devices in unpredictable ways that get more unpredictable as the number of qubits scales up.

So, once the physics and the refrigeration are sorted out, let’s take a look at how a few of the quantum computing technologies actually manipulate these quantum curiosities to come up with working, programmable computers.

Wire Loops and Josephson Junctions

Arguably the most successful commercial application of a quantum computer so far has been from D-Wave. While these computers don’t have “fully-programmable” qubits they are still more effective at solving certain kinds of optimization problems than traditional computers. Since they don’t have the same functionality as a “universal” quantum computer, it has been easier for the company to get more qubits on a working computer.

The underlying principle behind the D-Wave computer is a process known as quantum annealing. Basically, the qubits are set to a certain energy state and are then let loose to return to their lowest possible energy state. This can be imagined as a sort of quantum Traveling Salesman problem, and indeed that is exactly how the quantum computer can solve optimization problems. D-Wave hardware works by using superconducting wire loops, each with a weakly-insulating Josephson junction, to store data via small magnetic fields. With this configuration, the qubit achieves superposition because the electrons in the wire loop can flow both directions simultaneously, where the current flow creates the magnetic field. Since the current flow is a superposition of both directions, the magnetic field it produces is also a superposition of “up” and “down”. There is a tunable coupling element at each qubit’s location on the chip which is what the magnetic fields interact with and is used to physically program the processor and control how the qubits interact with each other.

Because the D-Wave computer isn’t considered a universal quantum computer, the processing power per qubit is not equivalent to that which would be found in a universal quantum computer. Current D-Wave computers have 2048 qubits, which if it were truly universal would have mind-numbing implications. Additionally, it’s still not fully understood if the D-Wave computer exhibits true quantum speedup but presumably companies such as Lockheed Martin wouldn’t have purchased them (repeatedly) if there wasn’t utility.

There are ways to build universal quantum computers, though. Essentially all that is needed is something that exhibits quantum effects and that can be manipulated by an external force. For example, one idea that has been floated include using impurities found in diamonds. For now, though, there are two major ways that we will focus on that scientists have built successful quantum computers on: ion traps and semiconductors.

Ion Traps

In an ion trap, a qubit is created by ionizing an atom of some sort. This can be done in many ways, but this method using calcium ions implemented by the University of Oxford involves heating up a sample, shooting electrons at it, and trapping some of the charged ions for use in the computer. From there, the ion can be cooled to the required temperature using a laser. The laser’s wavelength is specifically chosen to resonate with the ion in such a way that the ion slows down to the point that its thermal fluctuations no longer impact its magnetic properties. The laser is also used to impart a specific magnetic field to the ion which is how the qubit is “programmed”. Once the operation is complete, the laser is again used to probe the ion and determine its state.

The problem of scalability immediately rears its head in this example, though. In order to have a large number of qubits, a large number of ions need to be trapped and simultaneously manipulated by a series of lasers. The fact that the qubits can influence each other adds to the problem, although this property can also be exploited to help read information out of the system. For reasons of complexity, it seems that the future of the universal quantum computer may be found in something we are all familiar with: silicon.

Semiconductors

Silicon, in its natural state, is actually an effective insulator. Silicon has four valence electrons which are all perfectly content to stay confined to a single nucleus which means there is no flow of charge, and therefore no current flow. To make something useful out of silicon like a diode or transistor which can conduct electricity in specific ways, silicon manufacturers infuse impurities in the silicon, usually boron or phosphorous atoms. This process of introducing impurities is called “doping” and imbues the silicon with an excess or deficit of electrons in the outer shells, which means that now there are charges present in the silicon lattice. These charges can be manipulated for all of the wonderful effects that we use to create our modern world.

But we can take this process of doping one step further. Rather than introducing a lot of impurities in the silicon, scientists have found a way to put a single impurity, a solitary phosphorus atom including its outermost electron, in a device that resembles a field-effect transistor. Using the familiar and well-understood behavior of these transistors, the single impurity becomes the qubit.

In this system, a large external magnetic field is applied in order to ensure that the electron is in a particular spin state. This is how the qubit is set. From there, the transistor can be used to read the state of this single electron. If the electron is in the “up” position, it will have enough energy to move out of the transistor and the device can register the remaining positive charge of the atom. If it is in the “down” position it will still be inside the transistor and the device will see a negative charge from the electron.

These (and some other) methods have allowed researchers to achieve long cohesion times within the qubit — essentially the amount of time that the qubit is in a relevant state before it decays and is no longer useful. In ion traps, this time is on the order of nano- or microseconds. In this semiconductor type, the time is on the order of seconds which is an eternity in the world of quantum computing. If this progress keeps up, quantum computers may actually be commonplace within the next decade. And we’ll just have to figure out how to use them.

59 thoughts on “Quantum Computing Hardware Teardown

      1. I’m not saying that quantum computation is bullshit in it’s entireity… but if they did I wouldn’t be suprised if there was an atmel micro with a random number generator inside half of the things on sale as cutting edge propritary quantum computers. Tech support for snakeoil quantum CPU company would be easiest job ever, “your cooler isn’t cold enough… yeah it needs to be 3 universe deaths colder than that” “no repeatability or results? must be undetectable 7th dimentional beings intersecting with your spacetime co-ordinates” if in doubt, quote Doctor Who or Gordie la Forge until they hang up.

  1. Agree with fosselius here. I appreciate hackaday articles where time was taken to really bring me some interesting knowledge that isn’t laced with hype train (looking at you gizmodo).

    That said I am unsure how “commonplace” these will become in the future. It’s rather expensive to bring an object down to near 0k and if this is all correct that seems like a very crucial part of the problem.

    1. There are two main perspectives: either higher resilient qbits are found and maybe at some point we can all have a quantum computer at home (or in our pocket), or we don’t (which seems more likely at first), and the big and expensive to build and more importantly to run quantum computers will be large common instruments.

      1. Are there even practical home/mobile applications for Q computers? that can’t be done remotely? If I’m gonna risk freezing my testicles to the temperature of the death of the known universe there’d better at least be a quantum edition of angry birds…”level complete – your score is undetermined until you scrool down and look at it.”

        1. Any suggestions on using FPGA’s to cmos logic binary to trinary, using trinary logic to create qubits then convert to binary and into another FPGA ? This hardware structure is a parallel program to create the quibits, returning spin and superposition states. Use two virtex 4 dev card off ebay with 150 I/O lines ->

          https://www.ebay.com/itm/XMF4-XILINX-FPGA-MODULE-VIRTEX-4-XC4VLX100-Development-board/

          Awesome idea but not that technical yet to go forward, any suggestionssss ?? Since you have so many I/O lines, 150, could use a CPLD in between the two FPGA’s as well as the trinary logic, as additional programmed partial states of the many qubits. A software program would then process all the return data. Basically creating qubits in multi-value logic. Any ideas are welcome.

          David C.

    2. Do you have any idea how expensive a datacenter like runs AWS is?

      These are never going to be consumer tech, and there’s no reason why they would be. You can’t browse the internet with a quantum processor. You can’t play videogames with a quantum processor.

      Their business model is for other businesses, like cloud computing.

    3. Remember the pioneering technology is rarely a close representation of the mainstream result. Big leaps of faith in the first instance to achieve the unachievable by any means necessary. To make way for other to take the proven to learning and do it differently.

  2. What’s the ‘hello world’ of quantum computing? (and how many q-bits are necessary to process it sucessfully?) python: print ‘Hello World’, OpenGL: display a spinning square, Arduino: blinky led. For several years I struggle to get a straight answer to this question. Only that when they are ready, Quantum computers are gonna deliver the sun moon and stars, (but only in certain contexts, not general computing). However as I understand it, the more Q-bits added to a system the less one can rely on repeatability as a way of converging to an accurate or meaningful result to any useful calculation.

    I’m not saying that Quantum computing is bullshit in it’s entireity, but too many quantum experts are too quick to say ‘you’re too stupid to understand the math’. while avoiding some very simple, practical questions from the computer science perspective.

    1. Hello world is probably read the states of all the qubits, aka generate a random number. Not very hello worldy because it’s not repeatable, and you know someone will get exactly 1 on their first run with the shiny new million dollar computer and think “oh no, it’s broken”, but it’s a basic thing to do.

      1. What if you repeat the random number generation and statistically plot the spread of numbers generated?
        Sure there is a chance it would only output one number but it would be most likely to be an even spread of numbers?

    2. AGREED! I recently started researching online quantum computers. Seems the publisher of this article and myself have been reading the same things. I saw a video of some company talking it up about their quantum computer in front of a crowd. (cant remember who, just search it out on youtube, you’ll find it). They’ve sold a few (cant exactly remember).
      Anyway, this guy is trying to explain it, doing a horrible job, talking about borrowing bits from alternate universes…making billions of dollars on sales, cant explain how the chip works (proprietary).
      Seriously, that’s your best way to advertise your product?
      Sounded like he was selling Snake Oil!
      It’s difficult to wrap one’s brain around quantum physics, superposition and entanglement, but, how do you program these things? how does my code know which state to look for?
      And as far as there being, what, 3 states (thought it was four) 0, 1 and both, hell, between nothing and 1 is infinite states….this could be done with modern electronics.

      I don’t know, they are selling stuff for billions of dollars, it’s not as fast as the supercomputers available today, but, it’s better because its “Quantum computing”?

      1. One bit has 2 possible states, 0 or 1. One qbit has 3 possible states, representing the sets {0}, {1}, and {0,1}. Two bits have 4 possible states. Two qbits have 15 possible states, the number of nonempty subsets of {0,1,2,3}. N bits have 2^N possible states. N qbits have 2^(2^N) – 1 possible states. When you measure N bits repeatedly, you always get the same one of the 2^N possibilities. When you measure N qbits, there are 2^N possible results, but you don’t necessarily get the same one repeatedly. More precisely, when you repeatedly prepare N qbits into a given state, and measure them, you may always get the same result, or always get either of two results, or any of three, or any of up to 2^N results.

      1. Honestly, I don’t understand a word of that. any chance of some commented source code, truth tables, with expected inputs and outputs that might help us understand a little better.

    3. Good points. Too many things connected to quantum physics are absolute woo, and the explanation that you’re too unenlightened to understand it is precisely why charlatans love quantum. I think there’s probably going to be a legitimate use for the machines but it’s almost certain that some of these companies are frauds capitalizing on a buzzword. But every developing tech is like that. And the assumption that it must be legit because Lockheed Martin paid for it is also a little premature.

    4. I’d say that the hello world that will be to use will be Shor’s algorithm to factorise 15 into 3×5 or depending on the number of qbits the product of two extremely large prime numbers.

      1. This sounds promising, how many q-bits necessary in the system is estimated to hit an new/unknown prime? any source code example for this with comments? it’s be helpful for us to understand, ask questions.

  3. Whoa, an article about quantum computing and the first paragraph ends in, “With this amount of inertia, our quantum computing revolution seems almost certain.” *Inertia* is the incorrect physics term, *momentum* is correct.

  4. So is it an okay oversimplification to say that quantum computers can solve complex optimization problems because the system of electrons can always take the most efficient path to the ground state? So you set up the system of electrons and their energies in a way that mirrors your physical problem, and watch what path they take?

  5. As a graduate student in the field I have some points I would like to make regarding quantum computation.

    1) Always be careful when describing qubits. This article does a decent job avoiding the pitfall of ‘being 1 and 0 at the same time’. This statement is wrong because what is really being described are probability amplitudes of measurement outcomes which are represented as states. Qubits can be in the state |0> or |1> or |0>+i|1> etc. (yes, the last state isn’t normalized) and each is a valid state. What the state is encoding is the probability that a measurement will yield a certain outcome. Please be wary and do what this article did and utilize the concepts of states and superpositions.

    2) “Adding to the complexity, quantum computers must utilize quantum entanglement too, whereby a pair of particles are linked. This is the only way for any hardware to “observe” the state of the computer without affecting any qubits themselves.”

    This is not correct, measurements affect the system. There are things called weak-measurements which have a smaller back action but also do not yield as much information about the system. (Wiki link: https://en.wikipedia.org/wiki/Weak_measurement)

    3) “If this progress keeps up, quantum computers may actually be commonplace within the next decade”

    John Preskill of Caltech recently gave a talk about this exact topic and there is an arXiv paper related to his talk:
    https://arxiv.org/pdf/1801.00862.pdf

  6. Spacedog says ” However as I understand it, the more Q-bits added to a system the less one can rely on repeatability as a way of converging to an accurate or meaningful result to any useful calculation.”
    Sounds to me like he just described the human brain.

  7. The explanations of what quantum computers can do are generally very poor. They are not computers in the sense of being logic machines. I think the best way of describing them is as pattern recognition devices. Their use in decryption is probably easiest to understand. If you used a logic machine (aka conventional computer) for a “brute force” attack on a 256 bit key, you would need to try something like 2^255 values before being able to find the key. That might take a while. If you had a 256 bit quantum computer, it should be possible to find the key in a single operation. That could reduce the time required a little. A QC with 50 bits could reduce the number of operations by a factor of 2^50, and that could make certain calculations practical in conjunction with conventional computers.

    I would expect the first use of QCs to be in decryption (the spooks have the money) followed by similar problems that are hard for conventional computers, such as optimization (eg the traveling salesman problem), maybe something equivalent to Fourier analysis to analyze complex signals and for complex statistical problems. I expect QCs to complement conventional computers (LC = logic computer?) rather than replace them.

    1. If someone claims to have a 256 q-bit quantum computer, simply ask them to decrypt BTC Satoshis crypto key, become a multi billionaire overnight… if anyone does this don’t forget old spacedog, I only humbly ask for a gold house and a rocket car please.

  8. [typo last ¶: coherence time, not cohesion time.]

    Re: second paragraph:
    Superposition and entanglement are very similar concepts. Say we have three entangled qbits. There can be up to eight possible measurement results: the qbits can be in a superposition of one to eight states. But depending on the history, from zero to seven of them are not possible. The qbits can be prepared so all eight results are equally likely (superposition of eight states of three qbits), or so that only one specific result is possible (one state of three entangled qbits), or so all but one result is possible, etc.

    When all 8 results are possible, the qbits aren’t entangled at all. Knowing the value of one qbit tells nothing about the other two. When two results are possible, 000 and 111, the three qbits are obviously entangled: knowing one tells you the other two. For 000 and 100, measuring either of the last two tells you nothing about the first.

    The heart of Shor’s algorithm for factoring large composite integers is an algorithm that finds the period p of a periodic function F from N bit numbers to N bit numbers. p is restricted to around N/2 bits. Being periodic means that F(x)=F(x+p) for all N bit numbers x. It is very difficult to find p using a classical computer for certain F, but a quantum computer running the period finding algorithm evolves the state of N/2 entangled qbits starting with all measurements equally likely, and ending so that, ideally, the only possible measurement of those bits yields p. This result is easily checked, and the progam runs until the correct p is found.

    There are other uses for quantum computers, but that’s the one that can find the private key given the public key, breaking the security of the internet as we know it.

    1. this ‘entanglement’ business, has that actually been demonstrated beyond doubt? I have seen/heard people talk in detail about the experiments that have been carried out on this, and I remain totally unconvinced, there always seem to be some special conditions.

    1. Are you aware that there are other types of public key encryption algorithms, some of which there are good reasons to assume they can’t be cracked by quantum computers?

      Also cracking RSA requires more power than the current systems and scaling quantum computers is _hard_ – so it would never be a “hello world” unless you use 8 bit keys.

  9. I have yet to see one person explaining how the “hardware” works, properly. It’s all hype. Sounds like a quantum computer has plain hardware with a twist: when electrons have almost no resistance, on a wire, they will “compute” FAST…. All the “cooling” can pay off, but this is not the technological advancement that the physics promises. It’s marketing hype, so that that companies can sell computers that need to be cooled down to near absolute zero temperatures. Of course these machines will “computer” faster, but there is no “entanglement” involved anywhere!!! Am I wrong? Prove me wrong. Describe the REAL hardware architectures and its functionality. I DARE YOU!!!

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.