Quantum memory in the cloud—a cold, metallic gas cloud

As Ars readers know, a computer requires programmable logic gates, memory, a bus, a way to set the initial memory state, and a way to read out the final state. Gather all these components and you, too, can reinvent the wheel in your parents' garage. A quantum computer requires all the same elements, but instead of going to Circuit City, you need a well-funded research lab to assemble tons and tons of equipment into something that, currently, almost works. Improvements to these components will be essential to get quantum computers to do more than that, so it's big news that researchers have demonstrated quantum memory with storage times up to 1ms, an improvement of two orders of magnitude over current tech.

I am not going to give much of an introduction to quantum computing in this article, because it has been explained many times before. Suffice to say that, instead of holding binary information, a quantum computer's memory is required to store quantum states. Here, a single quantum bit (called a qubit) may be in both a logic one and a logic zero state simultaneously, called a superposition state. In addition, that qubit may be automatically correlated—using a phenomena called entanglement—to the state of other qubits. An effective quantum memory must preserve both the value of the qubit and its correlation to other qubits.

Cold metal atoms trapped by lasers

One of the most effective ways to store this information is in the quantum states of a cold gas of metallic atoms. These have lovely, long-lasting states that can be addressed using lasers, and researchers have a lot of experience playing with these gases from previous work on things like Bose Einstein condensates. These gases can be turned into a quantum memory by writing optically encoded qubits into the atomic states, which can be retrieved using different light pulses.

To give an example of memory that uses a cold metallic gas, say that I have a qubit currently stored in the polarization state of a beam of light. The information might be encoded in the area of the light pulse. The area is a combination of pulse intensity, duration, and frequency of an optical pulse; an area of two π is sufficient to shift our entire population of cold atoms from one state to another. We can vary the area by changing the intensity of the pulse. When applied to our cold gas, pulses with different intensities will produce a gas that's in a superposition of the ground state and the excited state, where the exact details of the superposition will depend on the exact area of the optical pulse. In doing so, the optical pulse's superposition state has been copied into a superposition state of the medium.

This state could last for a very long time, except for two things. The first is that atoms interact with magnetic fields, and there are always magnetic fields about. This acts to destroy the relationship between the ground and excited states—essentially the atoms become two distinct populations and are no longer a superposition state. This can be avoided by choosing atomic states that are particularly insensitive to magnetic fields. This is exactly what this latest research has done.

The state is also destroyed by the movement of individual atoms within the atomic cloud. The storage of an optical pulse takes place in both space and time, and the position and time at which a photon is absorbed by an atom in the medium matters. As atoms move, the superposition state is destroyed—once again, you end up with a population of atoms that are excited and a population that are in the ground state with no link between them. The way to minimize this is to cool the gas, which the researchers did, performing their experiments at 100 microKelvin.

More interestingly, the researchers also realized that the interference pattern produced by the overlap of the read and write pulses plays a crucial role in the speed at which the superposition state disappears. (Technically, these two pulses don't overlap because they are applied sequentially, but remember that the medium stores the write pulse until the read pulse is applied, which means that they do effectively overlap.) If the read and write beams are exactly parallel to each other, then the spatial variation in the interference pattern is very slow, meaning that an atom has to travel much further to be significantly out of position with respect to the pattern. This discovery helped the researchers obtain very long storage times.

On the downside, they aren't going to get much better. The only thing that they can do now is to turn the temperature down, which is, of course, possible, but the returns are small. They calculate that, by reducing the temperature by a factor of ten, they will only increase the storage time by a factor of three.

Nevertheless, a 1ms storage time is far better than your average RAM. However, the more important consideration is the storage time compared to the read/write time. These are on the order of a few microseconds, so refresh rates on the order of hundreds of kilohertz would be feasible, making this a viable scheme. Combine this with spatial light modulators, like those used in projection systems, and you have the makings of a nice, individually addressable 2-D memory array.

Nature Physics, 2008, DOI: 10.1038/nphys1153