Acta Optica Sinica, Volume. 45, Issue 17, 1720007(2025)

Optoelectronic Integrated Neuromorphic Computing Technology (Invited)

Yanan Han1,2, Shuiying Xiang1,2、*, Changjian Xie1, Yahui Zhang1,2, Xingxing Guo1,2, Tao Wang1, and Yue Hao2
Author Affiliations
  • 1State Key Laboratory of Integrated Services Networks, Xidian University, Xi’an 710071, Shaanxi , China
  • 2National Engineering Research Center of Wide Band-gap Semicondctor, School of Microelectronics, Xidian University, Xi’an 710126, Shaanxi , China
  • show less
    Figures & Tables(8)
    Schematic diagram of spiking neural network. (a) Spiking neuron; (b) spiking neural network
    Photonic neuromorphic devices for emulating spiking dynamics and synaptic plasticity/reconfigurable synapse
    Representative advances in active device-based optoelectronic integration computing. (a) Pattern recognition achieved using a single laser chip through time-division multiplexing[13]; (b) hardware computing architecture based on VCSEL and VOA[48]; (c) hardware computing architecture based on two-section semiconductor laser chip and VOA[49]; (d) hardware computing architecture based on two-section semiconductor laser chip and SOA[50]
    Representative advances in optoelectronic integrated computing based on silicon-based integration platforms. (a) Integrated all-optical SNN based on MRR and PCM[12]; (b) photonic neuromorphic processing system based on electrically driven MRR[19]
    Representative advances in optoelectronic integrated computing based on hybrid architectures. (a) Hybrid PSNN based on VCSEL-SA and MZI network[51]; (b) photonic neuromorphic processing architecture based on optoelectronic spiking neurons and MZI network[52]; (c) integrated neuromorphic photonic system based on VCSEL-MRR[53]
    Prospective architecture for optoelectronic integrated neuromorphic computing. (a) Optical neuromorphic computing architecture; (b) optoelectronic neuromorphic processor architecture
    • Table 1. Performances comparison for typical opto-electronic neuromorphic devices

      View table

      Table 1. Performances comparison for typical opto-electronic neuromorphic devices

      DeviceSpike rate /GHzEnergy spikeArea
      MRR+PCM120.0250.7 pJ100 μm2
      Electrically-driven MRR190.253.528 pJ~1575 μm2
      Graphene-based optical fiber laser7<1nJ scale<1 mm2
      Micro-pillar laser90.08250‒700 fJ4 μm (diameter)
      Distributed feedback laser (DFB)+balanced photodetector (BPD)8210.4 pJ~0.055 mm2
      Fabry‒Perot laser with saturable absorber (FP-SA)133.37.239 fJ300 μm×1500 μm
      Distributed feedback laser with saturable absorber (DFB-SA)5519.99 fJ250 μm×300 μm
    • Table 2. Performance comparison for three types of optoelectronic neuromorphic computing architectures[13,27]

      View table

      Table 2. Performance comparison for three types of optoelectronic neuromorphic computing architectures[13,27]

      Evaluation indicatorsActive device-based architecturePassive device-based architectureHybrid architecture
      Typical deviceSL+SOA/VOAMRR(PCM)SL+MZI/MRR
      Non-linear capability

      ★★★★★

      Strong spontaneous nonlinearity

      ★★☆☆☆

      Rely on external light sources

      ★★★★☆

      Active part provides nonlinearity

      Power consumption efficiency per spike (periphery circuits excluded)

      ★★★★★

      fJ

      ★★★★★

      pJ

      ★★★★★

      fJ

      Maximum response speed (nonlinear response)

      ★★★★★

      GHz

      ★★★☆☆

      sub-GHz

      ★★★★★

      GHz

      Scalability

      ★★☆☆☆

      High difficulty in large-scale integration, unsuitable for large-scale matrix operations

      ★★★★★

      Silicon-based technology is easy for large-scale integration and suitable for large-scale matrix operations

      ★★★☆☆

      Difficulty in interface compatibility

      Cost

      ★★☆☆☆

      High cost of active tape-out

      ★★★★★

      Low cost

      ★★★☆☆

      Balanced active and passive tape-out

      Area

      ★★☆☆☆

      mm2

      ★★★★★

      μm2

      ★★★☆☆
    Tools

    Get Citation

    Copy Citation Text

    Yanan Han, Shuiying Xiang, Changjian Xie, Yahui Zhang, Xingxing Guo, Tao Wang, Yue Hao. Optoelectronic Integrated Neuromorphic Computing Technology (Invited)[J]. Acta Optica Sinica, 2025, 45(17): 1720007

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Optics in Computing

    Received: May. 27, 2025

    Accepted: Jun. 25, 2025

    Published Online: Sep. 3, 2025

    The Author Email: Shuiying Xiang (syxiang@xidian.edu.cn)

    DOI:10.3788/AOS251155

    CSTR:32393.14.AOS251155

    Topics