Recently, a team from the University of Science and Technology of China (USTC) demonstrated an experiment on a photonic quantum computer that outperformed even the fastest classical supercomputer in a computational task. Such kind of experiments are targeting algorithms and hardware platforms that can provide “quantum supremacy”, which occurs when a quantum computer is outperforming a classical computer.
A photonic quantum computer harnesses particles of light (photons) and consists of a complex array of optical devices, such as light sources, beam splitters, mirrors and photon detectors, that shuttle photons around. In such a computer, the quantum computation is based on a process called Boson Sampling, which is a task deliberately designed to prove quantum supremacy. Boson sampling is trying to understand what the distribution of photons is going to be at the output of a photonic interferometer. In the case of the quantum device implementation of boson sampling, the problem is solved `by itself’ since the distribution of the measured output is the desired photon distribution. In the case of the classical computer, a large computation is required to find the photon distribution, which increases with the size of the problem since the photon’s quantum properties lead to an exponentially increasing number of possible distributions. If operated with large numbers of photons and many channels, the quantum computer will produce a distribution of numbers that is too complex for a classical computer to calculate. In the new experiment, up to 76 photons traversed a network of 100 channels, which is a much larger amount than previously demonstrated, both experimentally and numerically.
This claim for quantum supremacy comes to reinforce what Google presented last year with their superconducting qubit-based quantum computer. The main difference between the two experiments in terms of the result is that the photonics experiment can create many more possible output states: ~1030 of them compared to ~1016. Such a large number makes it infeasible to calculate the whole probability distribution over outputs and store it for future generation of samples (something other researchers suggested as a rebuttal against Google’s claims, but which can certainly not hold in this new experiment).
Although researchers are currently looking for ways to get similar results with classical computers, it has not yet been successful. The main concern around this quantum experiment is the photon loss. It was reported that up to ~70% of the photons get lost on their way through the beam splitter network, allowing only ~30% to be detected. Typically, that amount of photon loss would be considered fatal for quantum supremacy. Furthermore, the classical simulations that are used for comparisons require fixing the rate of noise and then letting the numbers of photons and modes go to infinity. However, any real experiment has a fixed number of photons and modes (in USTC’s case, they’re ~50 and ~100 respectively).
Achieving the goal of quantum supremacy through such kind of experiments does not indicate the definitive, general, superiority of quantum computers over classical computers, since such kind of problems are deliberately designed to be hard for classical computers. On the other hand, it would also be an understatement to say this experiment is `only a proof of principle’, since boson sampling could have actual practical applications, for example solving specialized problems in quantum chemistry and mathematics.
Currently, most proposals in the literature apply boson sampling to vibronic spectra or finding dense subgraphs, but it is not certain whether these proposals will yield real speedups for a task of practical interest that involves estimating specific numbers (as opposed to sampling tasks, where boson sampling almost certainly does yield exponential speedups).
Future research will focus both on algorithm development, exploiting the particular characteristics of such a specialized quantum device, as well as experimental improvements such as decreased photon loss, higher quality sources and detectors, and larger number of modes. The described experiment presents a promising indication of this sub-field of quantum computing, and we keep a close eye on future developments.