Search

CN-122028524-A - Gallium oxide/zinc oxide multi-modal photoelectric synaptic device and its preparing method and use in nerve morphology calculation

CN122028524ACN 122028524 ACN122028524 ACN 122028524ACN-122028524-A

Abstract

The invention discloses a gallium oxide/zinc oxide multi-modal photoelectric synaptic device, a preparation method thereof and application thereof in nerve morphology calculation, and a planar two-terminal photoelectric artificial synaptic device with response to both two-terminal voltage (> 6V) signals and deep ultraviolet (255 nm) optical signals is prepared through asymmetric contact engineering and defect-mediated photoconductive effects of metal layer/Ga 2 O 3 Schottky junction and Ga 2 O 3 /ZnO heterojunction contact. The obtained device has obvious photoelectric synergistic synaptic property, overcomes the limitation that the conventional multimode Ga 2 O 3 -based photoelectric synaptic depends on a complex three-terminal device, and provides a compact and efficient platform for multimode nerve morphology calculation and intelligent sensing. The device can be used as a physical reserve pool to map time sequence coding electric/optical pulse into high-dimensional conductivity dynamic characteristics, and can also realize event-driven image edge extraction under the control of pre-voltage and bias voltage, thereby showing the application potential of the device in the aspect of constructing a low-power-consumption nerve morphology sensing computing system.

Inventors

  • LV BIN
  • YE JIAHAO

Assignees

  • 浙江大学

Dates

Publication Date
20260512
Application Date
20260410

Claims (10)

  1. 1. The gallium oxide/zinc oxide multi-modal photoelectric synaptic device is an asymmetric laminated structure formed by adopting a metal layer/defect-state pure beta-Ga 2 O 3 /ZnO/metal electrode, wherein the metal layer/defect-state pure beta-Ga 2 O 3 /ZnO forms a vertical contact structure, the metal layer and defect-state pure beta-Ga 2 O 3 form a Schottky junction, the defect-state pure beta-Ga 2 O 3 and ZnO form a heterojunction, znO is not completely covered by the defect-state pure beta-Ga 2 O 3 and forms ohmic contact with the metal electrode, the metal layer is used as an electrode at the same time, so that the device is a planar two-end photoelectric artificial synaptic device, has photoelectric cooperative synaptic characteristics, and can generate nonvolatile response to high-voltage pulse signals of >6V and 255nm deep ultraviolet signals.
  2. 2. The gallium oxide/zinc oxide-based multimode photovoltaic synaptic device of claim 1, wherein the metal layer and the metal electrode are independently selected from Au, cu, ti, ni a, the thickness is 30-50nm, and the growth rate of the metal layer is less than 0.05 nm/s.
  3. 3. The gallium oxide/zinc oxide-based multimode photovoltaic synaptic device of claim 1, wherein the defect state in the defect-state pure beta-Ga 2 O 3 is an oxygen vacancy, the defect-state oxygen content is 20-30% of the total oxygen content concentration, and the total oxygen content is the sum of the defect-state oxygen and lattice oxygen.
  4. 4. A method of making a gallium oxide/zinc oxide-based multi-modal optoelectronic synaptic device as claimed in any one of claims 1 to 3, comprising the steps of: 1) Growing ZnO film on the substrate by physical vapor deposition, wherein the growth temperature is 390-395 ℃, and the film thickness is controlled to be 150-300 nm; 2) Partially covering the ZnO film obtained in the step 1), growing a defect-state pure beta-Ga 2 O 3 film on an uncovered part of the ZnO film by using a physical vapor deposition method, wherein the growth temperature is 610-640 ℃, the thickness is 180-220 nm, and the oxygen partial pressure is controlled during growth so as to control the oxygen defect content in the beta-Ga 2 O 3 film to be 20-30%; 3) And growing a metal layer on the obtained pure beta-Ga 2 O 3 film by using a physical vapor deposition method, wherein the thickness of the metal layer is 30-50 nm, the growth rate of the metal layer is controlled to be below 0.05 and nm s -1 , and growing a metal electrode on the ZnO film which is partially covered in the step 2) by using the physical vapor deposition method.
  5. 5. The method for preparing the gallium oxide/zinc oxide-based multimode photoelectric synaptic device according to claim 4, wherein in the step 2), a laser pulse deposition method is adopted to prepare a defect-state pure beta-Ga 2 O 3 film, a high-purity Ga 2 O 3 target material is adopted, the laser energy is 200-215 mJ, the frequency is 3-4 Hz, the target base distance is 5-5.3 cm, the oxygen partial pressure of a cavity is controlled to be 0.1-0.3 Pa, and the growth is performed for 50-75 min.
  6. 6. Use of a gallium oxide/zinc oxide-based multimode electro-synaptic device according to any one of claims 1-3 for brain-like calculation, wherein a forward operating voltage of not more than 2V is applied to the device, the forward being a bias voltage applied with a metal layer powered negative and a metal electrode powered positive, the device conductance being stable without additional stimulus, the current <1 nA, the device conductance fluctuating < 5%, the device conductance being increased non-volatile by applying a modulated light signal and/or voltage pulse signal, the modulated device conductance differing from the steady state current by at least 1 order of magnitude.
  7. 7. The use of a gallium oxide/zinc oxide-based multimode photovoltaic synaptic device according to claim 6 for brain-like calculations, wherein the device is reset by erasing the nonvolatile conductance change of the device after applying a reverse reset voltage of 0 to-2V to one side of the metal layer/Ga 2 O 3 .
  8. 8. A reservoir computing system comprising a gallium oxide/zinc oxide-based multimode electro-optical synaptic device as claimed in any one of claims 1-3 as a dynamic reservoir unit in reservoir computing for receiving a >6V high voltage pulse signal and/or 255nm deep ultraviolet signal and non-linearly converting the received signal and outputting a status signal for reservoir computing.
  9. 9. An event stream computing device comprising a gallium oxide/zinc oxide-based multimode electro-optical synaptic device according to any one of claims 1-3, said device being capable of pre-control and event stream processing as an event response unit and/or a state memory unit in event stream computation.
  10. 10. The event stream calculation apparatus as set forth in claim 9 wherein said pre-control is to apply a reverse pre-power signal to said device for a period of less than 2.5s at a voltage of-2V to-1V at a forward operating voltage of 0V bias or less than 1V, a light response characteristic of the device conductance being controlled by said pre-power signal, and a control effect of said pre-power signal on the light response characteristic being nonvolatile; The event stream processing specifically comprises the steps of applying a forward working voltage of 0.5V-0.7V to the device, inputting optical pulse signals with equidistant intervals, wherein the width of the optical pulse is 0.5-1.5s, the pulse interval is larger than 1.5s, and the photoelectric device is used for carrying out real-time response and time sequence feature extraction on the event stream formed by the optical pulse based on the pre-controlled optical response feature and outputting an event processing result.

Description

Gallium oxide/zinc oxide multi-modal photoelectric synaptic device and its preparing method and use in nerve morphology calculation Technical Field The invention belongs to the field of two-end photoelectric devices and application thereof, relates to a gallium oxide (Ga 2O3)/zinc oxide (ZnO) multi-mode photoelectric synaptic device and a preparation method thereof and application thereof in nerve morphology calculation, and in particular relates to a two-end photoelectric artificial synaptic device based on gallium oxide and zinc oxide heterojunction, a preparation method thereof and application thereof in brain calculation tasks such as reserve pool calculation, event stream processing and the like. Background With the rapid development of technologies such as the internet of things, edge intelligence and automatic driving, urgent demands are put forward for terminal perception computing hardware which can process environment information in real time and has low power consumption and high energy efficiency characteristics. The nerve morphology calculation inspired by synaptic plasticity in the biological nervous system provides a revolutionary path for realizing the deep fusion of sensing and calculation. The artificial synapse device with the two-end structure is considered as a core basic element for constructing a next-generation high-density neuromorphic chip because of simple structure and easy three-dimensional integration. In various sensing modes, light signal sensing, particularly ultraviolet light detection has important application value. Ultraviolet light has potential in the fields of automatic driving, biological recognition and the like due to the strong anti-interference capability. Gallium oxide is used as a wide bandgap semiconductor material, and the bandgap of about 4.9 eV makes the gallium oxide suitable for solar blind ultraviolet detection. The Ga 2O3 material has five crystal phase structures, namely a monoclinic phase (beta-Ga 2O3), a rhombohedral phase (alpha-Ga 2O3), a spinel phase (gamma-Ga 2O3), a cubic phase (delta-Ga 2O3) and a hexagonal phase (epsilon-Ga 2O3). Among these crystal phases, the beta phase is the most stable phase and exists stably even at 1800 ℃, but it is required to crystallize at a temperature of 650 ℃ or higher. Ga 2O3 prepared by the method is beta phase with more defects. Oxygen vacancy defects in Ga 2O3 material can induce continuous photoconductive effect, simulate the dynamic behavior of biological synapses, and provide an ideal platform for optoelectronic synapse devices. However, devices based on a single Ga 2O3 are generally only able to achieve non-volatile enhancement of conductance, whereas complete synaptic function also requires reversible regulation of conductance suppression. The existing research relies on electric signals to assist in achieving the function, and complexity and energy consumption of the system are increased. Therefore, developing a two-terminal device capable of achieving reversible switching of conductance entirely through optical modulation is a key challenge of current research. Although researches report Ga 2O3 -based photoelectric synapses, how to realize effective decoupling and fusion of light and electric signals through the synergistic effect of contact engineering and defect regulation and control still needs to be deeply explored. The solution of the problem promotes the development of multi-modal neuromorphic calculation and lays a foundation for constructing an efficient artificial perception system. To address the above issues, researchers often introduce heterojunction structures to optimize performance. Zinc oxide (ZnO) is a mature transparent conductive oxide with good lattice matching with Ga 2O3, which can form a high quality hetero interface. Although optoelectronic devices based on Ga 2O3/ZnO heterojunction have been reported, the prior art focuses on conventional photodetectors or single-function memristors, lacking specialized device designs and optimizations for reservoir computation and event-driven computation that are critical to neuromorphic computation, and in particular to time-sequential information processing. The existing device is generally limited by 1) high and uncontrollable optical and electrical response coupling degree, incapability of realizing independent and distinguishable programming of weights of optical and electrical signals, difficulty in being used for complex tasks requiring multi-mode coding, 2) single dynamic response characteristic, fixed conductance relaxation time constant and incapability of carrying out nonvolatile pre-programming through external signals, thus adapting to signal processing requirements of different time scales (such as adapting to pulse sequences of different frequencies), 3) wide process window and undefined relevance with performance, and influence mechanism of film growth parameters (such as temperature and oxygen pressure) on nerve-like characte