Electrical charge is the key to modern computing. We can generate, detect, and control current with accuracy and precision. As we contemplate a world of quantum computing, it’s important to note researchers have not really coalesced around an equivalent basic architecture. Each architecture we’ve looked at has a set of disadvantages that are sufficient to keep all the other architectures alive despite their problems.
This is true for optical quantum computers. They require single photons, and sources that generate single photons kind of… well, suck. Ideally, what we want is a device that generates single photons on demand. To put not too fine a point on it, at the press of a button, we want an electrical current to reliably generate a single photon. And, while we haven’t had it, it’s exactly what a group of researchers from Germany is now claiming to have accomplished.
Single photons: Accidentally on purpose
There are two basic ways to make single photon sources: accidentally and on purpose. The accidental method is to use a parametric amplifier, which takes a light source of one color and uses it to generate light of another color. If the amplifier is set up correctly, then each incoming light pulse has sufficient power to generate just a single pair of photons, which can then be separated by polarization. The single photon production is accidental in the sense that the process is subject to a lot of random noise. Mostly you get one photon, often you get no photons, and sometimes you get two or more photons. The saving grace of the whole thing is that you can use one of the photon pair to announce that the other one is a single photon with which you can perform a calculation.
The on-purpose version involves taking something like a single atom or molecule and repeatedly exciting it. Because it is a single object, excited in such a way that only a single transition is possible, it can only ever emit a single photon. But, it turns out that these sources—molecules, quantum dots, and other things of this nature—don’t do as they are told. Left to themselves, molecules and quantum dots emit the light in any old direction. Even though you might excite it on a regular basis, there is no guarantee of a regular stream of single photons going anywhere that you can actually use them.
Then there is the problem that they don’t all emit the same color of light. So, you might put a single quantum dot into an optical cavity in the hopes of capturing all of its photons, only to discover that it doesn’t emit light that can be captured by the optical cavity. This problem is so well known that researchers routinely freeze liquid xenon or nitrogen into their optical cavity to tune it so that it can capture light. Which is fine for what it’s worth, but it’s not exactly the sort of process that shouts scaleable and manufacturable.
Speaking of the manufacturing issue, quantum dots and other emitters like them have production processes that leave you with a set of randomly placed single emitters. After production, you have to scan a wafer, find all the suitable emitters, and then design all your optical and electronic circuitry around the randomly located useful bits. This is the opposite of a good manufacturing process.
So, short story: you can use a process that is statistically noisy to produce single photons, or you can use a deterministic process that has a random color and is emitted from a random location. The phrase “between the Devil and the deep blue sea” springs to mind.
Single photons: Purposefully accidental
In some sense, the latest research is a hybrid of these two approaches. Instead of a randomly located quantum dot, the researchers use a semiconducting carbon nanotube. Light is emitted from semiconducting carbon nanotubes with a reasonably well-defined set of wavelengths, and the emission process is partially deterministic. It happens when a conducting electron scatters on an imperfection in the nanotube and decays to a non-conducting state by emitting a photon. So, light can be produced by injecting charge into the nanotube. And, thanks to the nice properties of nanotubes, they can be deliberately placed right over a waveguide to ensure that the vast majority of light is emitted into the waveguide.
In other words, we have something that doesn’t have a random location and emits in a preferred direction at the push of a button… Just what the quantum doctor ordered.
The design also makes for a very cute experiment. Because the light can go in either direction down the waveguide, you can automatically tell if the nanotube is producing single photons by simply placing a photodetector at each end. The light has equal probability of traveling out of the nanotube to the left or to the right, so if both detectors click at the same time, you know that more than one photon was produced.
And this is exactly how the researchers determined that they had a single photon source—it’s a bit more complicated because they built their detectors into the waveguide, but the principle is the same. The researchers show that a semiconducting carbon nanotube was a reasonably efficient emitter of single photons, as long as the injected current was not too high. They also showed that it really was due to the semiconducting nature of the nanotube; a metallic carbon nanotube also emits light (like the filament of an incandescent bulb) but not single photons in their setup.
However, single photon production is more like the accidental process. Sometimes you get a single photon, but not always, and in this case, you can’t tell when it goes wrong. This, however, is due to an aspect of their implementation. Emission depends on scattering off a defect in the nanotube. If you have a single defect, then it is possible to have deterministic single photon production by supplying a sufficiently large current. In any given time interval, a single scattering event will occur, generating a single photon. Furthermore, if the emitter has the right properties, you can be pretty sure that two events will not occur in your time window.
The difficulty is that no one knows how many defects are present in a given nanotube. So you have to reduce the current such that the probability of only a single defect emitting a photon is maximized, compared to the chance of two or more defects emitting at the same time. In this respect, the researchers’ single photon generator is more like something called a parametric oscillator than a quantum dot.
Statistics is a nightmare
There is also a more fundamental issue with their experiment. The statistical assumption is that the defects do not influence each other. Take, for contrast, a quantum dot. We can know, using an electron microscope, that only a single quantum dot is being excited. And we know from the way that the quantum dot works that it can only emit a single photon per excitation cycle. A similar situation occurs with a parametric amplifier: thanks to the way that the light is generated, we know that it is possible for more than a single photon pair to be generated. But the measurement process is independent, so the statistical assumption—emission is separate from measurement—is valid.
In the carbon nanotube case, more than one defect may be present, and more than a single photon may be emitted. This happens—it can be clearly seen in the data. However, what cannot be taken into account is the possibility that activity at one defect could stimulate the emission of a second defect. If this were the case, then the nanotube would still randomly emit in a single direction, but each random emission would be likely to send multiple photons in a single direction. This would then be interpreted as a single photon emission.
So it’s possible that the emission of one defect in one direction may set off a second defect to emit in the same direction. This, as far as I can see, violates the basic assumption behind the measurement. Does this mean that they don’t have a single photon source? No, we can’t say that either. All the evidence so far points to a single photon source, but an extra measurement is required to ensure that stimulated emission is not distorting the picture. All that is required is that, instead of detecting the light directly in the waveguide, the team has to place a second beamsplitter further down the waveguide. If this architecture yields the same results, then we know that the defects are emitting independently of each other.
Should it turn out that this really is a single photon source, then it is a really nice one; it seems much nicer to use than quantum dots and simpler than a parametric amplifier. And, as far as I am concerned, little waveguide chips, with integrated detectors and gates, are the best flavor of ice cream. And are probably an excellent option for quantum computing.
Nature Photonics, 2016, DOI: 10.1038/NPHOTON.2016.178