By Dr. Andre Saraiva, UNSW

The achievements of Chinese academician Jianwei Pan have earned him the nickname of “Father of Quantum” in the Chinese quantum community. His success history recently culminated in the first quantum satellite, which distributes entangled photons between ground and an orbit thousands of kilometres high. His reputation explains why the scientific community got so hastily excited when rumours spread about his 50-photon boson sampling quantum computer, which would potentially outperform Google’s Sycamore quantum processor in a quantum supremacy experiment.

The English-language publication South China Morning Post reported that Jianwei Pan “announced at a lecture at Westlake University, Hangzhou, on September 5, 2020 that a new machine had recently achieved “quantum supremacy” one million times greater than the record currently held by Sycamore”.  This claim is currently not supported by any publication, so it is unclear what is the exact meaning of a “greater supremacy”, but it is probably related to the notoriously large dimension spanned by the states achieved in boson sampling devices.

The USTC (University of Science and Technology of China) team led by Pan was fast to communicate through the Chinese social media Weibo that they worry about the claim of potential quantum supremacy being quoted out of context. Indeed, the scientific community customarily goes by the motto “extraordinary claims require extraordinary evidence” – especially when it comes to quantum computing, in which inflated claims can turn into money flow. But Pan’s team has already shown progress in this direction last year, when a 20-photon boson sampling quantum computer was demonstrated. Moreover, Jianwei Pan co-signs a preprint manuscript dedicated to the benchmarking of a – at this point hypothetical – 50 photon boson sampling quantum computer against China’s classical supercomputer Sunway Taihulight.

The claim of “’quantum supremacy’ one million times greater” than Google’s experiment does not raise much skepticism, though. In the game of quantum advantage, gain is typically exponential. A handful of additional qubits can already give a quantum computer six orders of magnitude increase in computational power, whatever the metrics for this Pan’s group might be proposing. But in this case, their boson sampling system is doing precisely what it was designed to achieve.

Boson sampling was proposed as a simplified version of linear optics quantum computers. The general-purpose version of linear optical quantum computers, called the KLM architecture after its inventors Knill, Laflamme and Milburn, is much harder to implement and will only start giving out results at the mark of millions of photonic qubits. On the other hand, boson sampling can already outperform classical computers in a specific task at only tens of single photon modes.

Notice, though, the use of the term “single photon modes”, as opposed to qubits. This is because boson sampling architectures are not like other quantum computers based on qubits and logical gate operations. Boson sampling is based, instead, on the capacity of a collective multimode photon system to map complicated statistical distributions that are hard to obtain with a classical computer.

This means that there are two shortcomings for the boson sampling quantum computer. Firstly, it is believed to be inherently purpose-specific since there is no well-defined way to convert arbitrary algorithms into the boson sampling problem. That means that this quantum computer has no perspective to run famed algorithms such as Shor’s factorisation or Grover’s search algorithm.

Moreover, only a handful of real-life problems of interest have been theoretically demonstrated to be solvable in a boson sampling quantum computer. This severely limits the market for these devices when compared to the quantum computing model of traditional universal quantum computers. Current examples of application of boson sampling include the calculation of vibronic spectra of molecules, solving some spin Hamiltonians and problems in graph identification.

From a fundamental point of view, however, this demonstration is extremely valuable. It may serve, for instance, to experimentally probe the limits of the extended Church-Turing thesis which states that a universal computing device can simulate every physical process. A boson sampling device could be the first counterexample disproving this thesis. And this seemingly philosophical question can only be answered in the context of cold hard quantum error analysis. That is because the boson sampling problem is only computationally hard (#P-hard, to be more precise, which is a harder class than NP-complete) if error rates grow less than polynomially with the number of input photons.

While most naïve models of error actually predict an exponentially increasing error rate, Jianwei Pan’s team will finally be able to quantitatively analyse this effect and exactly up to which point is the boson sampling supremacy as astounding as predicted. The technical aspects of their photonic system from last year were already at the cutting edge in all parts: a single quantum dot is resonantly coupled to a microcavity creating single photons, which are demultiplexed into a stream of photon pulses into 20 spatial modes. The 20 input single photons were injected into a 3D integrated, 60-mode ultra-low-loss photonic circuit. Finally, the output single photons are detected by 60 superconducting nanowire single-photon detectors with coincidences recorded in a 64-channel coincidence count unit.

So, what’s next?

Well, firstly we need to see proof that this result was really obtained by the Chinese quantum researchers. But assuming that Jianwei Pan’s team has the data to back up his claim, then this should serve as a major push for further research into the possible applications of boson sampling. After Google’s supremacy demonstration on a gate-based programmable superconducting quantum processor, the mere demonstration of supremacy is no longer such a high prize. Pan’s team may need to focus on more extravagant displays of strength in order to keep happy the deep-pocketed Chinese funding agencies.

Dr. Saraiva has worked for over a decade providing theoretical solutions to problems in silicon spin quantum computation, as well as other quantum technologies. He works on MOS Silicon Quantum Dot research and commercially-oriented projects at the University of New South Wales (UNSW).

September 21, 2020