original version This story was published in Quanta Magazine.
Studying quantum systems, collections of particles that obey the counterintuitive rules of quantum mechanics, is not easy. Heisenberg’s uncertainty principle, the basis of quantum theory, says that it is impossible to simultaneously measure a particle’s exact position and velocity, information crucial to understanding what is happening. That’s what it means.
For example, to study a particular set of electrons, researchers have to be smart about it. They might take a box of electrons and poke at it in different ways to get a snapshot of what it ends up being. In doing so, they hope to reconstruct the internal quantum mechanics at work.
However, there is a catch. It is not possible to measure all properties of a system at the same time. So they repeat. They start with the system, poke and then measure. Then they will do the same thing again. At each iteration, a new set of properties is measured. Building enough snapshots together can help machine learning algorithms reconstruct the full properties of the original system, or at least come very close to it.
This is a tedious process. But in theory, quantum computers could help. These machines, which operate according to quantum rules, could be much better at modeling the behavior of quantum systems than ordinary computers. It can also store information in a more complex form called quantum memory, rather than classical binary memory. This allows for a richer and more accurate description of the particles. This also means that a computer can 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 certain algorithms that use quantum memory require significantly fewer snapshots than algorithms that do not use quantum memory. Although their method was a major advance, it required a relatively large amount of quantum memory.
In practice, this is quite a challenge because quantum memory is difficult to obtain. Quantum computers are made up of interconnected quantum bits called qubits, which can be used for computation or memory, but not both.
Now, two independent teams have devised a way to make do with far less quantum memory. In the first paper, Harvard computer scientist Citan Chen and his co-authors leapfrog the number of snapshots of a quantum system that need to be taken by creating just two copies of the quantum state. We have shown that it is possible to reduce In other words, quantum memory is almost always worth the investment.
