Transporting Ions Through an X-Junction: Quantum Computing Inches Closer

Physics World has a nice news article about a new experimental development in quantum computing, based on a forthcoming paper from the Wineland group at NIST in Boulder. I'd write this up for ResearchBlogging, but it's still just on the arxiv, and I don't think they've started accepting arxiv papers yet.

i-7ae1797c52c2bf56ab54a149eb4d6b2e-blakestad_fig1.jpgThe Physics World piece summarizes the key results nicely:

Now, Brad Blakestad and colleagues at the National Institute of Standards and Technology (NIST) in Boulder, Colorado have created a junction in an ion trap in which there is practically no heating. Constructed from laser-machined alumina, it contains 46 gold-coated electrodes surrounding an X-shaped junction. When the researchers apply a series of voltages to the electrodes, ions are encouraged through the junction a little at a time.

The NIST group managed to get ions through the junction with a 99.99% success rate, and with seven orders of magnitude reduced heating than previous trapped ion systems .

The new trap is shown schematically at right, which should give you some idea of the complexity. It's an impressive technical achievement.

So what's the big news about this?

This is a noteworthy development because of the famous Cirac-Zoller quantum computing scheme using trapped ions. This is one of the papers that really launched quantum computing as a major research field, along with Shor's factoring algorithm.

The idea of the Cirac-Zoller paper is that you can use trapped and laser-cooled ions in a trap as the "bits" in a quantum computer. The "1" and "0" states are two internal states of the ions, and you can use the collective motion of the atoms in the trap as a "data bus" to do operations on one ion that depend on the state of one of the other ions. Basically, you start out with a bunch of trapped ions that are as close to perfectly still as you can make them, and then set the whole collection of ions into motion depending on the state of one of the ions-- if it's a "1," you make all the ions slosh back and forth in the trap, while if it's a "0" you leave them alone. Then you can flip the state of one of the other ions in a way that depends on the motional state-- if they're sloshing, you flip it from "0" to "1" or vice versa, while if they're not moving, you leave the state alone. This lets you do all of the operations you need to be able to do to make a quantum computer.

It's a wonderfully elegant scheme, but there are two technical issues that present some problems when it comes to actually implementing the scheme: heating and scaling. The collective-motion "data bus" only works if the ions are initially not moving at all, which you can think of as a state with a very low temperature. Anything that causes the atoms to start moving on their own is a "heating" mechanism, and will throw off your calculations. You can get around this by doing something to "cool" the ions between operations, but this slows down the computation, and increases the chance of errors.

The "scaling" problem has to do with the number of qubits required to make a useful computer, which is quite large-- hundreds or thousands of ions. This is more than you can hold in a single trap, so any plan for making a useful quantum computer needs some way to expand beyond a single trap. There are lots of ways to do this, but most of them involve shuttling ions from one trap to another, so you can have a large reservoir of "memory" ions that are brought into the working trap a few at a time as needed to do computations, then returned to storage.

There's no reason in principle why this sort of scheme won't work, but as a practical matter, it's really tricky to do. The problem is heating, again-- moving ions from one region to another provides lots of opportunities for something to increase their motional energy, and throw off the calculation. There's also an issue of decoherence-- for the quantum computer to be a quantum computer, the state of the qubits needs to remain undisturbed until it comes time to read out the answer.

If you're ever going to see a working ion-based quantum computer, some way needs to be found to shift ions around without heating them up, or disturbing their states. This has been the main focus of the research effort in the ion-trap quantum computing community over the last several years. And that's what this new paper is about.

As the Physics World summary explains, the authors have built a new, rather complicated ion trap, and demonstrated that they can move ions from place to place and around corners with vastly improved performance. The new design has four arms, three of which contain ion traps. One of these, the experiment zone (indicated by the script E in the figure) is where they do the initial ion preparation, and the final state measurement. The other two, dubbed "vertical" and "horizontal" (script V and script H) are just holding zones, and there's a crossing region (script C), where the ions can be steered to either vertical or horizontal arms, by applying the right voltage signals to the electrodes making up the trap.

They demonstrated their new trap's performance in two ways. One was simply shuttling ions from the experiment zone into one of the other zones and back, and measuring first of all whether the ions made it through the whole process, and then what their temperature was. In 10,000 crossings of the junction, they didn't lose any ions, meaning that the probability of successfully crossing the junction was 99.99% or better, a rather significant improvement over the previous record of 98%. The temperature of the ions after crossing was also dramatically lower than the previous best-- 0.0000001 eV, compared to about 1 eV for previous efforts.

They also demonstrated that the state was preserved in the transition, by doing a "Ramsey interferometry" experiment, in which the ions are put into a superposition of their two internal states for some time, and then transferred back into one state or the other in a way that is incredibly sensitive to the timing of the fields used to do the state preparation. They demonstrated the Ramsey interferometry in the experiment region alone, and then repeated it while shuttling the atoms back and forth across the junction twice. They didn't see any change in the signal from the stationary case to the back-and-forth case, showing that their new trap preserves the state perfectly through the crossing.

This is really impressively good news for people interested in quantum computing, at least on the experimental side. Interestingly, there doesn't seem to have been any huge secret to the whole thing-- they attribute most of the dramatic reduction in the heating to simply getting rid of noise in the radio-frequency electric fields used to make the traps. The residual heating they attribute to the digital-to-analog converters (DAC's) that they use to generate the sequence of pulses needed to move the ions. With faster DAC's, they ought to be able to do even better.

This is pretty much the way things work in the experimental physics business. It's only in rare cases that progress is made through some dramatic "Eureka!" moment. Most of the time, advances in the experimental art-- even factor-of-ten-million advances-- are made through extremely mundane processes like tracking down noise sources, and through incredible attention to detail. The Wineland group are some of the very best in the world at this business, as this paper amply demonstrates.


More like this

This looks like a really cool experiment and it looks like there is lots of exciting progress going on in the ion trap world. However, as a theorist I always wonder to what extent things like Ramsey interferometry really do show that the "state is preserved", as you said. Every experimental quantum computing talk that I go to contains a graph that looks like a sine wave along with a claim that this shows that there is little decoherence or errors in the system. Of course, demonstrating interference does show that your system is behaving as a quantum system would in an interference experiment, which is definitely a good thing to show, but it seems a long way from what we need to build a quantum computer.

In quantum computing, what we really want to know is whether or not the entire transformation that has happened to the qubit in the course of the experiment is close to the identity (and whether the starting state of the system is controllable). To determine this, you would need to do something called "quantum process tomography", which requires the results several different preparation and measurement procedures to be combined. I am never sure why experimenters don't try to do this sort of analysis more often. I understand that it might be difficult for certain systems with current technology, but I think that some of them must be in a position to do it by now.

On the other hand, as far as I know, there has been no serious theoretical analysis of the extent to which you can reconstruct the dynamical map that occurs in a given experiment from an interference curve either. Although the information is surely insufficient to determine it completely, it ought to be possible to give some sort of estimate and error analysis using standard quantum statistical methods. This is perhaps something worth thinking about.