The original version of this story appeared in Quanta Magazine.
It is not easy to study quantum systems, that is, collections of particles that follow the counterintuitive rules of quantum mechanics. Heisenberg’s uncertainty principlecornerstone of quantum theory, states that it is impossible to simultaneously measure the exact position of a particle and its speed, information quite important to understanding what is happening.
In order to study, for example, a particular collection of electrons, researchers need to be smart. They can take a box of electrons, touch it in different ways, and then take a snapshot of what it looks like at the end. In doing so, they hope to reconstruct the internal quantum dynamics at work.
But there is a problem: they cannot measure all the properties of the system at the same time. So they iterate. They will start with their system, poke around, then measure. Then they will do it again. With each iteration, they will measure a new set of properties. Build enough snapshots together and machine learning algorithms can help reconstruct all the properties of the original system – or at least get really close.
It’s a tedious process. But in theory, quantum computers might help. These machines, which operate according to quantum rules, have the potential to be much better than ordinary computers at modeling how quantum systems work. They can also store information not in classical binary memory, but in a more complex form called quantum memory. This allows for much richer and more precise descriptions of particles. This also means that the computer could keep multiple copies of a quantum state in its working memory.
A few years ago, a team based at the California Institute of Technology demonstrated that some algorithms that use quantum memory require exponentially fewer snapshots than algorithms that do not use it. Their method was a major breakthrough, but it required a relatively large amount of quantum memory.
This is somewhat of a deal breaker, because from a practical standpoint, quantum memory is hard to come by. A quantum computer is made up of interconnected quantum bits called qubits, and qubits can be used for calculation or memory, but not both.
Now, two independent teams have found ways to get by with much less quantum memory. In the first paper, Sitan Chencomputer scientist at Harvard University, and his co-authors showed that just two copies of the quantum state could exponentially reduce the number of times you need to take a snapshot of your quantum system. In other words, quantum memory is almost always worth the investment.
