Highly aligned indium zinc oxide nanowire-based artificial synapses with low-energy consumption

https://doi.org/10.1016/j.jiec.2020.03.030Get rights and content

Highlights

  • The first digitally aligned IZO nanowire-based synaptic transistor is fabricated.

  • Essential synaptic functions are emulated using this synaptic transistor.

  • Energy consumption is the lowest among inorganic nanowire-based artificial synapses.

Abstract

Emulating neural activities at individual synapse level has recently attracted tremendous attention. Here, we demonstrate design and fabrication of the first digitally aligned nanowire (NW)-based three-terminal synaptic transistor. The highly-aligned and individually-position-controlled long continuous indium zinc oxide (IZO) NW arrays were directly printed on a large area with low cost using a special electrohydrodynamic nanowire printing (e-NWP) process. The NWs printed by e-NWP with a diameter of 250 nm emulated the structure of nerve fibers. The device emulates plasticity of biological synapses and showed remarkable advantages in energy consumption compared with previous reports. Unique band-edge modulation along the NW axial direction underlies makes excellent electrical properties of the device. This approach paves the way to easy fabrication of printed-NW-based artificial neural networks.

Graphical abstract

Highly aligned indium zinc oxide (IZO) nanowire (NW)-based three-terminal synaptic transistors are fabricated. The highly-aligned and individually-position-controlled long continuous IZO NW arrays on a large area were printed using a special electrohydrodynamic nanowire printing (e-NWP) process. The electronic device successfully emulates important working principles of a biological synapse, and showed low energy consumption down to 78.75 femtojoule per synaptic event. This approach opens the possibility of rapid fabrication of highly aligned nanowire-based artificial synapses on a large scale.

  1. Download : Download high-res image (177KB)
  2. Download : Download full-size image

Introduction

With information explosion in human society, the speed of computers increases, but the energy consumption also increases at the same time. A human brain, by contrast, is an exceptionally intelligent and efficient computing system that consuming far less power even than a household light bulb [1], [2]. This is realized by a high-density neural network of 1012 neurons that are connected by ∼1015 synapses [3]. Recent years, great progress has been made on metal oxide thin film synaptic transistors that emulate essential synaptic functions for artificial nervous signal transmission, information processing and memory [4], [5], [37], [38], [39], [40], [41]. However, the devices still have limitations in terms of energy consumption, size, and technology.

NWs are becoming attractive to serve as semiconducting structure in artificial synaptic devices due to their intriguing physical properties, unique chemical properties, and great potential as building blocks for nanoscale electronics and optoelectronics [6], [7], [8], [9], [10], [11]. Besides, NW structure is more suitable for emulating the morphology of nerve fibers for high integration [12], [13], [14]. So far, the inorganic semiconducting NW-based synaptic electronic devices are mostly prepared by hydrothermal method or traditional electrospinning method, but these approaches are difficult to attain the long and continuous NWs arrays [15], [16], while in a neuromorphic system, long and continuous arrays of NWs aligned in a digitally controlled approach are desired. Therefore, we propose here an e-NWP method to prepare NWs, which can meet the following requirements: (i) printing of highly-aligned, individually-positioned NWs arrays on a large area, controlling exact number of NWs, their orientations, and their dimensions directly duringprinting process, (ii) high-speed manufacturing process to produce NW arrays on a large scale, (iii) easy and low cost process compared with traditional lithography techniques.

In this study, we demonstrated the first fabrication of digitially aligned IZO NW-based synaptic transistors (ISTs). The highly aligned semiconducting NW array with a diameter of 250 nm was used to emulate the structure of nerve fibers. IST was used to mimic important synaptic functions, including excitatory postsynaptic currents (EPSC), paired-pulse facilitation (PPF), spike-voltage dependent plasticity (SVDP), spike-duration dependent plasticity (SDDP), spike-rate dependent plasticity (SRDP), spike-number dependent plasticity (SNDP), long-term potentiation (LTP) and long-term depression (LTD). Besides, energy consumption of IST was reduced even down to 78.75 fJ per synaptic event. The outstanding performance of IST owed to unique band-edge modulation along the NW axial direction and electric-double-layer (EDL) effect at the interface between IZO NWs and the ion gel.

Section snippets

Preparation of precursor solution

Poly(vinylpyrrolidone) (PVP, sacrificial polymer) was dissolved in N,N-dimethylformamide (DMF, common solvent) and stirred at 50 °C for 12 h. Meanwhile, zinc acetate dihydrate (Zn(CH3COO)2·2H2O, ZnACD, metallic precursors) and indium nitrate hydrate (In(NO3)3·xH2O, InNH, metallic precursors) were dissolved in Trichloroethylene and stirred for 12 h at room temperature. Then, the two solutions were mixed and stirred for 2 h until fully intermixing.

e-NWP of IZO NWs

The digitally controlled parallel array of IZO NWs

Results and discussion

We designed IST to mimic the functions of a biological synapse (Fig. 1a). IST consists of a probe, ion gel (the molecular formula was shown in Fig. 2b), Au source and drain electrodes and IZO NWs arrays on a SiO2 insulated substrate. The printing parameters of IZO NWs have been described in the experimental section. Different from traditional electrostatic spinning, e-NWP works at a lower voltage, moreover, the distance of the collector and the nozzle tip is closer. Under the effect of the

Conclusion

In summary, we demonstrated the first fabrication of synaptic transistors based on highly-aligned semiconducting NWs with individually-position control. This device successfully emulated important synaptic functions, such as EPSC, PPF, SNDP, SVDP, SDDP and SRDP, and showed low-energy consumption. The excellent performance attribute to the long and continuous NW structure which has a unique band-edge modulation along the NW axial direction. This research paves the way to printed NW-based

Acknowledgments

This research was supported by Guangdong Key R&D Project No. 2018B030338001, Hundred Young Academic Leaders Program of Nankai University (2122018218), Natural Science Foundation of Tianjin (18JCYBJC16000), Tianjin Science Foundation for Distinguished Young Scholars (Grant No. 19JCJQJC61000), and the Fundamental Research Funds for the Central Universities (075-63191740, 075-63191745).

References (41)

  • Y. Ren et al.

    Adv. Funct. Mater.

    (2018)
  • H.T. Wang et al.

    Adv. Funct. Mater.

    (2019)
  • Y.J. Lee et al.

    Sci. Adv.

    (2018)
  • S. Chen et al.

    Adv. Mater.

    (2018)
  • K. Gerrow et al.

    Curr. Opin. Neurobiol.

    (2010)
  • D.X. Lai et al.

    Org. Electron

    (2019)
  • Y. Burgt et al.

    Nat. Mater.

    (2017)
  • L.Q. Zhu et al.

    Nat. Commun.

    (2014)
  • L.F. Abbott et al.

    Nature

    (2004)
  • X.F. Duan et al.

    Nature

    (2001)
  • K. Takei et al.

    Nat. Mater.

    (2010)
  • F. Patolsky et al.

    MRS Bull.

    (2007)
  • G. Wang et al.

    Appl. Phys. A

    (2011)
  • Q. Hu et al.

    Nanoscale

    (2016)
  • G.Y. Gou et al.

    Nanoscale

    (2016)
  • W.T. Xu et al.

    Sci. Adv.

    (2016)
  • G.Y. Gou et al.

    J. Mater. Chem. C

    (2016)
  • C. Zou et al.

    Appl. Phys. A

    (2017)
  • D.S. Hong et al.

    Adv. Electron. Mater.

    (2017)
  • X. Liu et al.

    Nat. Commun.

    (2014)
  • Cited by (18)

    • Mimicking ion-balance-dependent synaptic plasticity in body fluid for adaptive environment-responsive artificial neuromuscular reflexes

      2022, Materials Today Nano
      Citation Excerpt :

      Transmission electron microscope (TEM) images revealed obvious lattice fringes, a lattice spacing of 0.275 nm from the [1000] planes (Fig. 2f). To simulate the exchange of neural information, we combined ion colloids with IZO active layer and applied voltage pulses as presynaptic spikes (Fig. 2g) [26,27]. Electrochemical impedance spectroscopy of ionic colloids (Fig. 2h) compared the ion storage and diffusion abilities with equivalent circuit and dynamic process demonstrated (Fig. 2i and j): Balanced K+/Li+ ratio (No. 1) resulted in best ion transport efficiency, and other characteristic parameters, such as electrical double layer capacitance Cdl and the ionic conductivity σ, are also consistent with the curve trends in Fig. 2f (Supporting Information Table S1).

    • A low-dimensional hybrid p-i-n heterojunction neuromorphic transistor with ultra-high UV sensitivity and immediate switchable plasticity

      2021, Applied Materials Today
      Citation Excerpt :

      The integration of the multiple functionalities of sensing, learning, computing, and memory into neuromorphic devices has become an important trend [7]. Nanowire structured neuromorphic transistors are becoming widely used in artificial sensory and motor systems [8–13]. Due to the elongated shape, flexibility, and good scalability to large areas, NW structure is suitable for emulating the morphology of nerve fibers, with the goal of nerve-like high integration [14,15].

    • Recent progress on two-dimensional neuromorphic devices and artificial neural network

      2021, Current Applied Physics
      Citation Excerpt :

      The ability to enhance or weaken the efficiency of information transmission can be defined as the synaptic weight, which is the basis of mimicking biological behaviors by using electronic synaptic devices [14–18]. The speed of synapses transmitting and processing information is slower than that in electronic devices, but in the biological system, neural network works under very low energy consumption, to transmit and process all kinds of biological information with high efficiency parallelly [19–24]. At present, the applications of microelectronic devices to mimic the behaviors of biological synapses are at an early stage, and there is still a long way to go to successfully produce a computer based on the neuromorphic computing system.

    View all citing articles on Scopus
    View full text