February 2, the reporter learned from Shanghai Jiaotong University Integrated Quantum Information Technology Research Center, the center Jin Xianmin team developed a combination of integrated chips, photon concept and non-Von Neumann computing architecture of photonic computer, the new computer not only to solve some difficult problems have the potential to go beyond the classical computer, and the physical scale can be expanded. The study provides a new way of thinking beyond the computing power of classical computers and foreshadows the future of photonic computers. The study was published in the latest issue of the American journal Scientific Progress.
Rising integration gives electronic computers increasingly powerful computing power, and studies have shown that Moore’s Law will no longer apply in the near future due to the chip “cooling problems” and “quantum tunneling effect” caused by high integration.
“The search for potential new methods of computing is an important means to further advance human computing power, and quantum computing, DNA computing, optical computing, etc. are constantly being proposed,” Kim explained to reporters. At the end of 2019, Google demonstrated 53 qubits of quantum computers, proclaiming ‘quantum supremacy’, and was the first to reveal the advantages of non-von Neumann’s computing architecture. “
In the latest study, Kim’s team took a different approach, relying not on fragile quantum features, but more on the advantages of photons themselves, demonstrating the potential of photon computers to surpass classical computers in specific computational problems.
The problem solved by the research team on the photon computer, called “Subsets and Problems” (SSP), is one of the most difficult to solve in terms of computational complexity, one of the most difficult problems in NP (a large category of problems that classical computers cannot solve efficiently), and solving SSP can be used as an important criterion for measuring the computing power of new computing architectures.
In the latest study, researchers successfully mapped SSPs to a three-dimensional integrated optical waveguide network of three basic structures and carved them inside photonic chips using fly-second laser directwriting technology. When photons are injected into the photowaveguide network, the calculation process is activated. Photons, as computational carriers, evolve in the optical waveguide network and search in parallel for all possible evolutionary paths to find solutions.
The study found that SSP solves faster and the physical scale is fast, thanks to the parallel operation of photon computers, the compactness of integrated photowaveguide networks, and the “talent” of ultra-high propagation speed and strong anti-jamming capability of light.
Kim hyun-min says they plan to build larger-scale photonic chips and measurement systems, moving towards larger problem sizes and computing power.