World first general-purpose photonic quantum computer developed by RIKEN and NTT

The world's first general-purpose photonic quantum computer has been developed by a collaborative research group comprising Team Leader Akira Furusawa (Professor at the Graduate School of Engineering at the University of Tokyo) and Team Leader Hidehiro Yonezawa of the RIKEN Center for Quantum Computing, Senior Distinguished Researcher Toshikazu Hashimoto of NTT, and Takuji Hiraoka, President and CEO of Fixstars Amplify. A cloud-based environment will be available for collaborative researchers by the end of the year.

Furusawa said, "Yonezawa and Hiraoka were the first fourth-year undergraduate students who joined my team when I set up the laboratory in April 2001. I am deeply moved. Right now, it is just for collaborators; however, in the future, we want to develop it into a quantum computer that can be widely used globally."

A photonic quantum computer (measurement-induced quantum computer) requires the generation of large-scale quantum entanglement. It exploits the advantages of the traveling wave properties of light and time-division multiplexing techniques. Initially, squeezed light (light in which quantum fluctuations are squeezed) is generated from each of the four optical parametric amplifiers developed by NTT. Its light pulses are superimposed using a 50% reflective beam splitter to continuously generate quantum entanglement between two photons. Next, by delaying the two optical paths by one optical pulse and N optical pulses, quantum entanglement between two photons is distributed at different times.

A set of four light pulses that exists at the same time is called a macronode. Because the optical delay is N pulses, the macronode has a periodic structure of N. In the currently constructed system, 101 quantum entanglements are generated, which extend along the time axis over time. Thus, in principle, any number of calculation steps (linear operations) can be performed. The optical pulse width is set to 10 ns, equivalent to 3 m in space, and linear calculations are possible at a speed equivalent to a clock speed of 100 MHz.

Hashimoto said, "Increasing the accuracy of the optical parametric amplifier, optical measurement system, and entire feedback system for quantum teleportation will enable the realization of large-scale calculations. We are currently developing these devices."

The most important feature of the photonic quantum computer developed by the research group is that it is an analog machine.

Furusawa said, "Digital computation is necessary for highly accurate calculations. However, analog is fine for AI, neural networks, and other calculations that are done by the human brain, such as distinguishing between dogs and cats. Current generative AI, for example, consumes a great deal of energy for converting analog to digital, performing enormous calculations, and then outputting them in the analog form. Vague calculations, such as neural networks, benefit from being in an analog form. For the time being, it will be used as an analog machine, but by improving its precision, we can convert it to a digital form at the input and output and make it fault tolerant. In the future, it will be developed into a machine that can do analog and digital calculations."

A software development kit will be provided to enable its use in a cloud environment.

Hiraoka said, "We developed a software that can be used by nonquantum computer specialists to design quantum circuits in Python. A library will also be provided. By increasing the number of users, we hope to expand the applications of analog quantum computers."

A user sends the designed quantum circuit to the cloud. After user authentication, the cloud automatically converts the quantum circuit into the parameters of the actual machine and sends the job to the actual optical quantum computer at RIKEN. Results are delivered to the user through the cloud.

Yonezawa said, "To increase the number of use cases and make them practical simultaneously, we will work on further increasing the number of inputs, realizing ultrahigh speed, introducing nonlinear operations, and exploring applications. By 2030 or so, we aim to realize a large-scale general-purpose quantum computer with fault tolerance. We want to make the most of this opportunity."

There are various methods for developing quantum computers, including superconductors, neutral atoms, ions, and silicon, but the number of researchers working on photonic quantum computers remains very small. Although a joint research agreement is required this time, providing a cloud-based development environment will be an opportunity to increase the number of users. Japan is leading the world in photonic quantum computers, and the utilization of this opportunity by Japan will have a major impact on its competitiveness in the coming years.