Quantum mechanics is so hard to understand that even experts do not entirely trust their intuition – and this makes it difficult for physicists to come up with new experiments that put the theory to the test. Now, physicists at the University of Vienna in Austria have devised a computer algorithm for designing new quantum experiments that are beyond our wildest dreams.

The idea was developed by graduate student Mario Krenn and colleagues in the group of quantum-physicist Anton Zeilinger. The algorithm is dubbed "Melvin", and the team believes that it might be able to explore hitherto unknown properties and behaviours of quantum systems. In doing so, Melvin would take the complexity of quantum experiments to a level beyond the imaginations of human designers.

"High-dimensional" entanglement

These experiments include those with the particular goal of achieving quantum entanglement between many particles. Experimental methods for achieving entanglement of two or a very few particles are well-known. But entanglement is so counter-intuitive that it can be very difficult to see how to combine the known experimental "building blocks" to attain a more complicated state, such as "high-dimensional" entanglement between many of the particles' degrees of freedom.

Melvin works that out unencumbered by human preconceptions. The algorithm is supplied with a set of standard experimental components that it can combine and reshuffle to achieve the desired goal. These elements consist of devices for manipulating the trajectories and quantum properties of photons. These include beam splitters, which can send a photon in two possible directions, thereby putting it into a superposition of two quantum states.

Melvin begins by assembling the elements of this toolbox randomly, and seeing if any of the configurations achieves the experimental goal. If so, Melvin then simplifies the arrangement of elements as much as possible before delivering the configuration to the user. If the goal is not achieved, it starts again with another random arrangement. After typically several days of computation on a standard laptop, Melvin can deliver several optimized solutions to the specified task.

All I was doing was guessing, and I thought that this is something the computer can do
Mario Krenn, University of Vienna

The idea began, says Krenn, when his colleague Mehul Malik wondered if a particular high-dimensional form of a quantum state called a Greenberger–Horne–Zeilinger (GHZ) state (which involves three or more entangled particles) could be created. "Several of us tried for some time to find a way to implement it experimentally" – but without success, he says. Finally, Krenn says, "it occurred to me that my intuition about how the set-up would work was wrong – basically all I was doing was guessing. And I thought that this is something the computer can do as well, but a few thousand times faster than me."

In its first demonstration, Melvin came up with 51 new kinds of experiment for making high-dimensional GHZ states. In a second implementation, the Vienna team found how to achieve cyclic transformations of photon states, such that a sequence of transformations eventually returns the photon to its initial state. Such sequences could be useful in quantum-information processing. Here, Melvin found good solutions from around 1022 possible configurations of the experimental building blocks – and was eventually able to reduce the number of elements required to just four. Krenn and colleagues have now started to implement some of Melvin's solutions in the lab.

Genuine creativity

Because Melvin does not follow intuitive reasoning, the researchers say, it is not bound by conventional ideas about how to achieve a goal. They suggest that according to some definitions this makes Melvin genuinely creative. Some of the solutions may, however, defy intuition, even when the answer is known to be correct. In the proposed experiments for making a high-dimensional GHZ state, Krenn admits that he is still puzzled what exactly is going on. "There is one very intricate step for which I can write down every step mathematically, but which is very difficult to explain intuitively," he says. "I think that is unique in quantum physics," he adds. He expects that this confounding of intuition will be all the more common with increasing complexity in the tasks that Melvin tackles.

Clever idea

"This is a very clever idea, and I can see that a lot can be done with it," says Daniel Greenberger of the City College of New York, a specialist in the fundamental aspects of quantum theory. It will work best, he says, "where there are only a finite number of pieces of equipment and a finite number of different experiments that are not too complicated". But Melvin is not going to invent new theories, Greenberger cautions. "Totally new thought arrangements are beyond it, at least in the foreseeable future, so it won't replace the scientific community yet."

The research is described in a preprint on arXiv and in a paper to be published in Physical Review Letters.