Search

CN-121257626-B - Multi-sense cross-mode monolithic integrated neuromorphic device and preparation and use methods thereof

CN121257626BCN 121257626 BCN121257626 BCN 121257626BCN-121257626-B

Abstract

The invention provides a multi-sensory cross-mode monolithic integrated nerve mimicry device which can be applied to the technical field of novel semiconductor devices. The device comprises a substrate, a first semiconductor material layer, a second semiconductor material layer, a metal electrode layer, a temperature sensing layer, a gas sensing layer and a light sensing layer, wherein the first surface of the substrate is provided with a groove, the first semiconductor material layer is positioned on the first surface of the substrate, the first semiconductor material layer and the substrate form a suspension structure in the groove, the second semiconductor material layer is positioned on one side of the first semiconductor material layer far away from the substrate, the metal electrode layer is positioned on one side of the second semiconductor material layer far away from the substrate and is electrically connected with the second semiconductor material layer and used for leading out an electric signal, the temperature sensing layer is positioned on one side of the second semiconductor material layer far away from the substrate, the gas sensing layer is positioned on one side of the second semiconductor material layer far away from the substrate, and the light sensing layer is positioned on one side of the second semiconductor material layer far away from the substrate, wherein the temperature sensing layer, the gas sensing layer and the light sensing layer are arranged in parallel.

Inventors

  • YI XIAOYAN
  • Zhan teng
  • LIN JIN
  • WANG JUNXI
  • LI JINMIN

Assignees

  • 中国科学院半导体研究所

Dates

Publication Date
20260512
Application Date
20250926

Claims (9)

  1. 1. A multi-sensory cross-modal monolithic integrated neuromorphic device comprising: A substrate, the first surface of the substrate being configured to include a recess; the first semiconductor material layer is positioned on the first surface of the substrate, and the first semiconductor material layer and the substrate form a suspension structure in the groove; the semiconductor device comprises a substrate, a first semiconductor material layer, a second semiconductor material layer, a first semiconductor material layer and a second semiconductor material layer, wherein the first semiconductor material layer and the second semiconductor material layer form a heterojunction structure, and the forbidden bandwidth of the first semiconductor material layer is larger than that of the second semiconductor material layer; The metal electrode layer is positioned on one side of the second semiconductor material layer away from the substrate, is electrically connected with the second semiconductor material layer and is used for leading out an electric signal; A temperature sensing layer located at one side of the second semiconductor material layer away from the substrate; a gas sensing layer located on a side of the second semiconductor material layer away from the substrate; a light sensing layer located on one side of the second semiconductor material layer away from the substrate; Wherein the temperature sensing layer, the gas sensing layer and the light sensing layer are arranged in parallel.
  2. 2. The multi-sensory cross-modal monolithic integrated neuromorphic device of claim 1 further comprising an insulating dielectric layer between the temperature sensing layer, the gas sensing layer, and the light sensing layer and the second layer of semiconductor material.
  3. 3. The multi-sensory cross-modal monolithic integrated neuromorphic device of claim 1 wherein the grooves extend through the substrate or not.
  4. 4. The multi-sensory cross-modal monolithic integrated neuromorphic device of claim 1 wherein the substrate comprises one or more of a silicon substrate, a sapphire substrate, a silicon carbide substrate, a quartz substrate, and a gallium nitride substrate.
  5. 5. The multi-sensory cross-modal monolithic integrated neuromorphic device of claim 1 wherein the material of said first layer of semiconductor material comprises one or more of gallium nitride, gallium arsenide, gallium phosphide, gallium oxide, silicon carbide, diamond, and telluride, and The material of the second semiconductor material layer comprises one or more of gallium nitride, gallium arsenide, gallium phosphide, gallium oxide, silicon carbide, diamond and telluride.
  6. 6. The multi-sensory cross-modal monolithic integrated neuromorphic device of claim 1 wherein the material of the temperature sensing layer comprises one or more of platinum, nickel, copper, and chromium; The material of the gas sensing layer comprises one or more of zinc oxide, titanium dioxide, tungsten trioxide, indium trioxide, tin dioxide and nickel oxide, and The material of the light sensing layer comprises one or more of zinc oxide, titanium dioxide, tungsten trioxide, indium trioxide, tin dioxide and nickel oxide.
  7. 7. The multi-sensory cross-modal monolithic integrated neuromorphic device of claim 1 wherein the material of the metal electrode layer comprises one or more of aluminum, silver, nickel, gold, platinum, chromium, titanium, or copper.
  8. 8. A method of making a multi-sensory cross-modal monolithic integrated neuromorphic device of any one of claims 1 to 7, comprising: providing a substrate; carrying out patterned etching on the first surface of the substrate to form a groove; Forming a first semiconductor material layer on the first surface of the substrate containing the groove through an epitaxial growth process, so that the first semiconductor material layer spans the groove to form a suspended structure; Forming a second semiconductor material layer on the first semiconductor material layer through an epitaxial growth process; sequentially forming a temperature sensing layer, a gas sensing layer and a light sensing layer which are mutually insulated and arranged in parallel on the second semiconductor material layer through a patterning process, and And forming a metal electrode layer penetrating to the second semiconductor material layer so as to lead out an electric signal.
  9. 9. A method of using the multi-sensory cross-modal monolithic integrated neuromorphic device of any one of claims 1 to 7, comprising: Placing the device in an environment to be tested, and enabling the temperature sensing layer, the gas sensing layer and the light sensing layer to be simultaneously exposed to the temperature, the gas component and the light intensity stimulus to be tested; applying a bias voltage to the second semiconductor material layer through the metal electrode layer; monitoring and collecting an electric signal output by a metal electrode layer, wherein the electric signal is a nerve mimicry response signal generated after the multi-modal stimulation sensed by the temperature sensing layer, the gas sensing layer and the light sensing layer is subjected to cross-modal integration by the second semiconductor material layer; and identifying or judging the environmental state based on the time sequence, frequency or amplitude characteristics of the neuro-mimicry response signals.

Description

Multi-sense cross-mode monolithic integrated neuromorphic device and preparation and use methods thereof Technical Field The invention relates to the technical field of novel semiconductor devices, in particular to a multi-sense cross-mode monolithic integrated nerve mimicry device and a preparation and use method thereof. Background Under the background of the current brain science research and the rapid development of artificial intelligence technology, a nerve mimicry chip based on a 'memory integrated' architecture has become an important direction for breaking through the traditional computing bottleneck. The traditional von neumann architecture computer has obvious problems of memory wall and power consumption wall when processing large-scale artificial intelligent tasks due to calculation separation, namely high delay and low energy efficiency caused by frequent transportation of data between a processor and a memory, and severely restricts the training and deployment efficiency of an AI large model. In order to overcome the problems, the sensing and storing integrated chip architecture has the aim of integrating sensing, storing and calculating functions into the same device, simulating an information processing mechanism of a biological nervous system and realizing high-efficiency and low-power consumption edge intelligent processing. Currently, some researches are focused on developing sensing and neuromorphic computing devices with a single mode (such as vision or touch), however, a real biological system has the capabilities of multi-sensory information fusion and cross-mode associative learning, and the sensing and processing of the single mode are difficult to cope with the requirement of intelligent decision in a complex environment. In the prior art, a device structure capable of integrating multiple sensing functions (such as light, sound, gas, force, heat and the like) on a single chip at the same time and realizing cross-mode signal cooperative processing and impulse neural network calculation is not known. In addition, most mimicry devices are based on traditional silicon-based materials, have limitations in performances such as wide forbidden band, high electron mobility, high temperature resistance and radiation resistance, and are difficult to meet requirements of an artificial intelligence system in the aspects of high performance, high reliability and multi-scene adaptation in the future. Therefore, development of a novel semiconductor device which can integrate multiple sensing modes, has high-efficiency neuromorphic information processing capability and can be monolithically integrated is highly demanded. Disclosure of Invention First, the technical problem to be solved In order to solve at least one of the problems, the invention provides a multi-sense cross-mode monolithic integrated nerve mimicry device and a preparation and use method thereof, which are combined with element doping controllability, wide forbidden bandwidth, continuous photoconductive effect, deep energy level defect of semiconductor materials and two-dimensional electron gas mechanism in a compound semiconductor gallium nitride material, and simultaneously cover chip process technologies such as multi-heterojunction integration, micro-mechanical system and the like, and can be simultaneously fused with three or more mode sensing multi-terminal artificial nerve mimicry devices in optical vision, acoustic hearing, gas-smell sense, force sense of touch and thermal sense of touch, and realize information integration, processing and decision output through a nerve mimicking working mechanism. (II) technical scheme Aiming at the technical problems, the embodiment of the invention provides a multi-sense cross-mode monolithic integrated nerve mimicry device and a preparation and use method thereof. According to a first aspect of the invention, a multi-sense cross-mode monolithically integrated neuromorphic device is provided, comprising a substrate, a first semiconductor material layer arranged on the first surface of the substrate and forming a suspended structure with the substrate in the recess, a second semiconductor material layer arranged on one side of the first semiconductor material layer away from the substrate, a metal electrode layer arranged on one side of the second semiconductor material layer away from the substrate and electrically connected with the second semiconductor material layer for extracting an electrical signal, a temperature sensing layer arranged on one side of the second semiconductor material layer away from the substrate, a gas sensing layer arranged on one side of the second semiconductor material layer away from the substrate, and a light sensing layer arranged on one side of the second semiconductor material layer away from the substrate, wherein the temperature sensing layer, the gas sensing layer and the light sensing layer are arranged in parallel. In some exemplary embodiments, a