Nano-Net. Third International ICST Conference, NanoNet 2008, Boston, MA, USA, September 14-16, 2008, Revised Selected Papers

Research Article

3D CMOL Crossnet for Neuromorphic Network Applications

Download163 downloads
  • @INPROCEEDINGS{10.1007/978-3-642-02427-6_1,
        author={Kevin Ryan and Sansiri Tanachutiwat and Wei Wang},
        title={3D CMOL Crossnet for Neuromorphic Network Applications},
        proceedings={Nano-Net. Third International ICST Conference, NanoNet 2008, Boston, MA, USA, September 14-16, 2008, Revised Selected Papers},
        proceedings_a={NANO-NET},
        year={2012},
        month={5},
        keywords={CMOS-Nano Hybrid System CMOL Crossnet Neuromorphic Network 3D IC},
        doi={10.1007/978-3-642-02427-6_1}
    }
    
  • Kevin Ryan
    Sansiri Tanachutiwat
    Wei Wang
    Year: 2012
    3D CMOL Crossnet for Neuromorphic Network Applications
    NANO-NET
    Springer
    DOI: 10.1007/978-3-642-02427-6_1
Kevin Ryan1,*, Sansiri Tanachutiwat1,*, Wei Wang1,*
  • 1: College of Nanoscale Science and Engineering at SUNY Albany, Albany
*Contact email: kryan@uamail.albany.edu, stanachutiwat@uamail.albany.edu, wwang@uamail.albany.edu

Abstract

In this work, a novel 3D CMOL crossnet structure is introduced by combining two leading technological concepts for future nanoelectronic neuromorphic networks: CMOL crossnet and 3D integration. By implementing CMOL crossnet into the third dimension, the proposed 3D CMOL crossnet not only maintains the high-speed and high defect-tolerant properties of the CMOS-nano hybrid CMOL hardware system, but also provides efficient fabrication and assembly processes with a much higher density than the original CMOL crossnet. Furthermore, this study focuses on the development of multivalue synapses and efficient communication methods between CMOS and nanodevices. Preliminary results demonstrate that the structure can utilize the advantages of high performance synapses and stable analog CMOS somas in three dimensions. Therefore, the proposed 3D CMOL crossnet structure has a huge potential to become an efficient 3D hardware platform to build neuromorphic networks that are scalable to biological levels.