Under the radar
13 02 2007D-Wave supposedly has unveiled it’s 16-qubit Quantum Computer today, according to their blog. What does this all mean? Well, if it truly is a Quantum Computer, then good for them. I’m quite skeptical, mainly because this thing would have gone completely under our radar. Their method is using superconducting Quantum Interference Devices (aka “SQUIDs”), which are large solid-state devices. One reason ions hold so much promis in this field is that their qubits can have long coherence times (i.e. it keeps a |0> or |1> longer), because we choose atomic states that do not interact strongly with the environment. Solid-state implementations of Quantum Computing generally have short coherence times. So, given that I’ve not seen many papers demonstrating rapid progress in solid-state quantum computation, I think I’m right in being rather skeptical about this announcement. My advisor Chris is skeptical too.
I don’t know anything about adiabatic quantum computing, which is what this system supposedly is. From what I can gather, it looks like they’re just simulating (solving?) a 2-D ising model in a magnetic field. Correct me if I’m wrong, but isn’t there an analytic solution?
I have this sneaky suspicion that what we have is a 2-D array of qubits—a qubit consisting of a large domain of a single state—interacting via tunneling from one domain to another. I don’t think that we have actual entanglement, nor coherence. This tunneling interaction can be written as a 2-D Ising model. Hence, all I can see this ‘quantum computer’ doing is simulating the 2-D ising model.
Note: The Quantum Pontiff, Dave Bacon, had a roundup of this. We’ll see what goes on with this. If this is bunk, hopefully it won’t be like Cold Fusion, which discredits the whole field.
Hi! Have you heard how their demo went? I’ve been searching for info and finding nothing anywhere, even on the D-wave website! Also, as a lay person, what would be the capabilities of a 16 qubit system, anywey? Or of a 1,000 qubit system. Thanks for your reply!
If you’ve read that page about 2D Ising model that you linked to, you’ll notice that one step in that “analytic solution” is to find the largest eigenvalue of a 2^n by 2^n matrix, which takes quite bad exponential time (2^(2n) time or worse?). The fastest worst-case scaling to solve the problem exactly with conventional software is about n(2^n) time, which is only slightly better than the brute-force solution, which takes (n^2)(2^n) time. Note that even the brute-force solution is much better than your “analytic solution”.
Also, since as you said, you don’t know anything about adiabatic quantum computing, maybe it’s a bit premature to be commenting that it won’t work because of decoherence. If you had read much of the information Geordie has posted on AQC/TAQC, you would know that decoherence is quite unrelated to TAQC. That’s not to say that there aren’t other potential problems, but decoherence doesn’t really apply, as Geordie has said several times.
Dennis: I haven’t heard anything about the demo, but I hope it went well. As for what we could do with a 16-qubit system, it probably isn’t enough to crack an RSA code. But there are lots of interesting things we can do with only a few qubits. If we can truly tailor the interactions in this 16-qubit system, we could perform small-scale quantum simulations, in addition to investigating entanglement and probing the oddities of Quantum Mechnanics. Coming from a circuit model of Quantum Computing (which is the general practice with trapped ions) once we start getting 1000 qubits, we can start to implement quantum error correction and improving the fidelity of the computation. We could probably start doing something ‘practical’ too.
Neil: Silly me, I forgot about the computing power it takes to find the largest eigenvalue. So, even though we have an analytic solution, it is quite useless in terms of actually calculating things. It seems from Geordie’s comments that it will provide a quadratic speed-up. Is the solution to the Ising model equivalent to Grover’s search algorithm? If it is, then I’m not sure that we actually have a quantum computer. I’ve seen papers in PRA & PRL that state that Grover’s search only requires wave properties, and not entanglement. As such, it can be implemented using classical waves of light. [S. Lloyd, Phys. Rev. A 61, 010301(R) (2000)] [D. A. Meyer, Phys. Rev. Lett. 85, 2014 (2000)], [N. Bhattacharya, et. al., Phys. Rev. Lett. 88, 137901 (2002)]. Sure, entanglement could play a role, but it is not necessary to perform Grover’s search.
And yes, it probably is a bit premature to say that it won’t work due to coherence. It seems to me that after reading: cond-mat/0609332 that the total time to run the system is on the order of the thermalization time. It seems like thermal decoherence does play a role after all. But it looks like it isn’t detrimental, which is counter-intuitive. I guess the question then is whether or not the engineered homotopy from the initial Hamiltonian to the final one can be made adiabatic in the roughly 10-100ns timescale. I’d bet that that can be done. What about state detection and preparation? And entanglement? I haven’t seen anything demonstrating that an entangled system had been created, and measured. Even a 6-qubit entangled state would make headlines.
It is a very interesting story. Thanks!