Ultra-fast programmable photonic AI system, including ultra-fast programmable photonic matrix chip, High-speed high-precision phase control system; High-speed high-precision optical detection module; Special optical neural network computing software, Optical artificial intelligence EDA software; Optical AI algorithm and manipulation system
Based on the new generation photonic integration material lithium niobate thin film (LNOI) photonic chip technology, the development of ultra-fast programmable photonic matrix chips, including ultra-high brightness on-chip dual-mode compressed state arrays, ultra-fast programmable photonic matrix units, and high-efficiency single photon detection arrays, which can achieve high-efficiency on-chip compression, on-chip filtering, and on-chip encoding of compressed states.
Core Technical Indicators
1: Based on 8nm EBL process, the chip architecture has >10 layers and can achieve <0.1dB/cm waveguide transmission loss and waveguide quality factor (Q-value) >500000 2: The chip is based on a new thin film lithium niobate (LNOI) material, providing over 10GHz electro-optical modulation bandwidth 3: Ultra-fast programmable optical quantum matrix unit dimension is 8*8, ultra-high brightness dual-mode compressed state light source array number ≥ 8 4: Built-in FPGA module to provide acceleration for specific optical quantum computing algorithms; can be optimised specifically for different quantum computing algorithms; can implement variable fraction quantum computing (VQE), boson error correction coding, Gaussian boson sampling (GBS) applications and customised algorithms using the structure of infinite dimensional Hilbert space
Ultra-high-speed optical AI chip： providing a variety of machine learning photonic hardware solutions, the dedicated optical AI inference chip has a peak arithmetic power of 1000GFLOPS (1TFLOP) and meets the customization needs of special AI algorithms.
Programmable optical quantum chips：expected to enable variable fraction quantum computing (VQE), molecular vibrational spectroscopy, graph similarity, boson error correction coding, Gaussian boson sampling (GBS) applications and user-tailored algorithms using the structure of infinite dimensional Hilbert space.