Neuromorphic electronic device based on ion-ferroelectric synergistic effect and preparation method thereof

Abstract

The invention discloses a neuromorphic electronic device based on an ion-ferroelectric synergistic effect and a preparation method thereof, and belongs to the technical field of novel non-von Noiemann computing. The device comprises a supporting substrate, a dielectric layer, a bottom electrode, a conducting channel, a top electrode and a packaging layer which are sequentially stacked from bottom to top. The conducting channel is made of a semiconductor material with both an ion migration mechanism and a ferroelectric polarization mechanism; the source electrode and the drain electrode are respectively connected with the top and the bottom of the conductive channel. The invention provides a neuromorphic electronic device based on an ion-ferroelectric synergistic effect and a preparation method thereof, and aims at realizing input from two ends of a source electrode and a drain electrode to regulate and control the ion-ferroelectric synergistic effect by taking a semiconductor material with an ion migration mechanism and a ferroelectric polarization mechanism as a memristive device of a conducting channel. Complete memristor dynamic characteristics in the device are comprehensively utilized to realize synaptic weight adjustment under a unified framework.

Description

Technical Field

[0001] The present invention belongs to the field of novel non-von Neumann computing technology, and specifically relates to a neuromorphic electronic device based on ion-ferroelectric synergy and a preparation method thereof. Background Art

[0002] Inspired by the human brain, neuromorphic devices attempt to simulate the interconnected network architecture of biological nervous systems to replicate their efficient and unique information processing capabilities. However, the brain's superior efficiency in processing information stems not only from its unique neuron-synapse structure, but also involves the dynamic organization of the nervous system by network functions. As a basic network-level biological function, neural reuse models the brain as a dynamic system that handles multiple tasks by leveraging existing neural circuits rather than creating entirely new circuits. It can be foreseen that neural reuse using memristors is of great significance for hardware-based neural networks, which can flexibly reuse the same components in various tasks, thereby effectively adapting to and utilizing existing resources and further improving efficiency. However, just as software operations require the cooperation of processors and memory, simulating this "biological algorithm" also requires memristors to have different dynamic processes to build stable source models and variable task models respectively. This requires that the hardware network composed of memristors needs to be able to accommodate two different but related functions (from source model to variable task model), while also ensuring reusability and variability between different functions. This seemingly contradictory but interrelated property is very similar to the volatile and non-volatile storage characteristics exhibited by memristors. This inspired us to integrate the complete memristor dynamics process of memristors into a unified framework. Specifically, a stable source model and an adjustable task model can be constructed based on the non-volatile and volatile states of a single memristor device, respectively. Then, the redeployment of the task model and the recovery of the source model are completed by cycling configurations between different working modes. However, integrating the complete dynamic characteristics of the device into a collaborative framework is a new challenge for memristors.

[0003] In the existing technical solutions, using reconfigurable strategies to make memristors have switchable working modes with different dynamic processes seems to be a possible approach. The existing technology "Reconfigurable logic and neuromorphic circuits based on electrically tunable two-dimensional homojunctions" uses reconfigurable devices made of two-dimensional tungsten diselenide to achieve the switching of logic and neuromorphic functions; the existing technology "Reconfigurable halide perovskite nanocrystal memristors for neuromorphic computing" reconfigures the constructed halide perovskite nanocrystal memristors into different node layers and readout layers according to volatile and non-volatile characteristics to construct a reference reservoir network; the existing technology "Versatile memristor implemented in van der Waals CuInP2S6" uses the switching of its volatile and non-volatile working modes in Ag / CIPS / Au devices to achieve the reconstruction of selector and memory functions. It can be seen that the reconfigurable devices in the existing technology usually focus on realizing the functional switching of components rather than collaboration, so they lack the ability to simulate network-level biological characteristics, especially neural reuse. Therefore, it is necessary to explore new memristive mechanisms and solutions to integrate the full dynamic characteristics of the device into a collaborative framework to support analog neural reuse. However, existing reconfigurable memristive devices usually first bind their functions to their volatile or non-volatile characteristics and then divide them into different components to handle different tasks. Summary of the invention

[0004] The purpose of the present invention is to provide a neuromorphic electronic device based on ion-ferroelectric synergy and a preparation method thereof in order to address the technical problems that existing artificial neuromorphic electronic devices lack the ability to simulate the network-level characteristics of biological neural systems, making it difficult to truly simulate the neural reuse behavior in biology and support the concept of parameter sharing in artificial intelligence models, so as to be more in line with biological cognitive capabilities and match the efficient algorithms of artificial intelligence. A memristor device based on a semiconductor material with both ion migration mechanism and ferroelectric polarization mechanism as a conductive channel is used to realize input from both ends of the source and drain to regulate the ion-ferroelectric synergy, and the complete memristor dynamic characteristics in the device are comprehensively used to achieve synaptic weight regulation under a unified framework.

[0005] The technical problem proposed by the present invention is solved in this way:

[0006] A neuromorphic electronic device based on ion-ferroelectric synergy includes a supporting substrate 1, a dielectric layer 2, a bottom electrode 3, a conductive channel 4, a top electrode 5 and a packaging layer 6 stacked in sequence from bottom to top; the conductive channel 4 is made of a semiconductor material having both an ion migration mechanism and a ferroelectric polarization mechanism; the bottom electrode 3 is a drain electrode, and the top electrode 5 is a source electrode, the source electrode and the drain electrode respectively cover the top and bottom of the conductive channel 4, and form a Schottky contact or an Ohmic contact containing a clean van der Waals interface.

[0007] Furthermore, the supporting substrate 1 is made of a rigid or flexible substrate material, specifically Si, Al2O3 or PET; the dielectric layer 2 is made of SiO2, Al2O3 or h-BN; the source electrode and the drain electrode are metal electrodes Ti, Cr, Au, Pt, Pd, graphene electrodes or other two-dimensional metal and semi-metal material electrodes; the encapsulation layer 6 is made of h-BN or Al2O3.

[0008] Furthermore, the conductive channel 4 is photolithographically patterned, and the photolithographic pattern enables the top and bottom of the conductive channel 4 to form a conductive channel in the vertical direction.

[0009] Furthermore, the semiconductor material selected for the conductive channel 4 has both a ferroelectric polarization mechanism and an ion migration mechanism;

[0010] For the ferroelectric polarization mechanism, the displacement of specific ions in the lattice generates electric dipoles. The spontaneous order of electric dipoles below the Curie temperature generates macroscopic polarization, which can be switched by external electric pulses. The ferroelectric polarization mechanism of the conductive channel 4 is insensitive to the application time of the external electric pulse and has retention. Therefore, the conductive channel 4 has non-volatility under the ferroelectric polarization mechanism.

[0011] For the ion migration mechanism, specific ions exhibit ion conduction characteristics through discontinuous long-distance disordered motion. The ion conduction characteristics require an electric field to maintain, and the application time of the external electric pulse affects the degree of ion migration. Therefore, the conductive channel 4 is volatile under the ion migration mechanism;

[0012] The site order of the conductive channel 4 under the ferroelectric polarization mechanism is the basis for the modulation of the disordered ion migration mechanism. The ion migration mechanism can continuously adjust the conductivity of the device based on the ferroelectric polarization mechanism, thereby realizing the modulation of the volatile conductivity of the device based on the non-volatile conductivity. When the device switches back from the disordered ion migration mechanism to the ordered ferroelectric polarization mechanism, the device naturally restores the non-volatile conductivity under the ferroelectric polarization mechanism.

[0013] Furthermore, a short excitation time and high amplitude electric pulse is applied between the source electrode and the drain electrode so that the device operates under the ferroelectric polarization mechanism; a long excitation time and low amplitude electric pulse is applied between the source electrode and the drain electrode to realize device conductivity regulation based on the ion migration mechanism on the basis of the ferroelectric polarization mechanism.

[0014] Furthermore, the material of the conductive channel 4 is synthesized by mechanical stripping or chemical vapor deposition, and has a single crystal or polycrystalline structure.

[0015] The present invention also provides a method for preparing the above-mentioned neuromorphic electronic device based on ion-ferroelectric synergy, which specifically comprises the following steps:

[0016] S1, preparing a patterned bottom electrode on the dielectric layer;

[0017] S2, dissolving PVA powder in water to obtain a 5-15wt% solution, dropping it on the upper surface of the substrate and drying it at 40-60°C to form a film, and then placing the film on a PDMS supported by a transparent glass sheet to obtain a PVA transfer film for dry transfer;

[0018] S3, stripping the semiconductor material having both ion migration mechanism and ferroelectric polarization mechanism to obtain nanosheets of set thickness and size;

[0019] S4, using the PVA transfer film prepared in step S2 to stick the nanosheets prepared in step S3, and transferring them to the bottom electrode obtained in step S1 at 40-60°C;

[0020] S5, using a photolithography process to obtain a set photolithography pattern, where the photolithography pattern covers a portion of the bottom electrode, so that the top and bottom of the nanosheet material form a conductive channel in the vertical direction;

[0021] S6, using an evaporation coating device to deposit a thin film on the basis of the structure prepared in S5, and peeling off to obtain a patterned top electrode, thereby obtaining a neuromorphic electronic device with good metal-semiconductor van der Waals contact or ohmic contact;

[0022] S7. Deposit a packaging layer on the surface of the neuromorphic electronic device obtained in step S6 by atomic layer deposition, and then lead out a source electrode and a drain electrode from the top electrode and the bottom electrode respectively.

[0023] Furthermore, before step S1, there is the following process: preparing a dielectric layer on the supporting substrate by adopting a physical vapor deposition (PVD), a chemical vapor deposition (CVD) or an atomic layer deposition (ALD) method.

[0024] Furthermore, step S4 can also be implemented in the following two ways. The first way is: placing a semiconductor material nanosheet having both an ion migration mechanism and a ferroelectric polarization mechanism on the bottom electrode prepared in step S2 by spin coating; the second way is: growing a semiconductor material having both an ion migration mechanism and a ferroelectric polarization mechanism on the bottom electrode prepared in step S2 by physical vapor deposition PVD, chemical vapor deposition CVD or atomic layer deposition ALD.

[0025] Furthermore, the specific process of step S1 is: photolithography patterning on the dielectric layer, and then using thermal evaporation or electron beam coating equipment to deposit a 40-60 nm thick film on the dielectric layer, and then immersing the substrate in methyl pyrrolidone NMP solution for 5-10 minutes and then peeling it off to obtain a patterned bottom electrode.

[0026] The beneficial effects of the present invention are:

[0027] (1) The neuromorphic electronic device based on semiconductor materials with both ion migration mechanism and ferroelectric polarization mechanism described in the present invention achieves volatile and non-volatile modulation of channel conductance by regulating ion migration and ferroelectric polarization mechanism in the conductive channel through source and drain electrodes. Uniquely, the device conductance can not only be switched between two modes, but also the volatile conductance can be adjusted based on the non-volatile conductance, thereby achieving seamless and gradual device conductance change.

[0028] (2) The present invention uses a semiconductor material based on both ion migration mechanism and ferroelectric polarization mechanism as a conductive channel, and utilizes the characteristics of ion migration and ferroelectric polarization mechanism derived from the displacement of cations in the material, which not only realizes flexible and reconfigurable device functions, but also highly integrated in space; most importantly, this is substantially different from the traditional reconfigurable technology route that uses the volatile and non-volatile electrical properties of the device to handle different tasks respectively. The present invention integrates the two into a unified framework, and through a collaborative approach, enables them to complete the same task more efficiently.

[0029] (3) The present invention achieves the simulation of biological advanced neural network behavior - neural multiplexing characteristics by separately regulating and coordinating the ion migration and ferroelectric polarization mechanisms in semiconductor materials based on both ion migration and ferroelectric polarization mechanisms. This is difficult to achieve with other technologies, and provides an excellent way to bridge the cognitive ability gap between memristor-based neural networks and biological neural networks, and provides support for neuromorphic devices to truly simulate biological efficient information processing behaviors.

[0030] (4) The present invention utilizes two memristor mechanisms, namely, ion migration and ferroelectric polarization, which are homologous in semiconductor materials with both ion migration and ferroelectric polarization mechanisms. This overcomes the shortcomings of the prior art that relies solely on the kinetic characteristics of a single memristor mechanism, and enables memristor-based neural networks to accommodate more efficient artificial intelligence algorithms and architectures, thereby greatly expanding the upper limit of their application and breaking the bottleneck of the current single application scenario and severe specialization of neuromorphic devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 is a schematic diagram of the structure of the neuromorphic device of the present invention;

[0033] Figure 2 is a graph showing the memristor characteristic of the neuromorphic device in the ferroelectric polarization working mode described in the embodiment;

[0034] Figure 3 A schematic diagram of the ability to maintain device conductance of the neuromorphic device in the ferroelectric polarization working mode described in the embodiment;

[0035] Figure 4 is a graph showing the memristor characteristic of the neuromorphic device in the ion migration working mode described in the embodiment;

[0036] Figure 5 A schematic diagram of the ability to maintain device conductance of the neuromorphic device in the ion migration working mode described in the embodiment;

[0037] Figure 6 Graphs showing device conductivity changes when the neuromorphic device described in the embodiment operates alternately in ferroelectric polarization and ion migration modes.

[0038] Markings in the figure: 1-support substrate, 2-dielectric layer, 3-drain electrode, 4-conductive channel, 5-source electrode, 6-packaging layer. DETAILED DESCRIPTION

[0039] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0040] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the embodiments described are only part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0041] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0042] Embodiment 1

[0043] The present embodiment provides a neuromorphic electronic device based on ion-ferroelectric synergy, comprising a supporting substrate 1, a dielectric layer 2, a bottom electrode 3, a conductive channel 4, a top electrode 5 and a packaging layer 6 stacked in sequence from bottom to top; the conductive channel 4 is made of a semiconductor material having both an ion migration mechanism and a ferroelectric polarization mechanism; the bottom electrode 3 is a drain electrode, and the top electrode 5 is a source electrode, the source electrode and the drain electrode respectively cover the top and bottom of the conductive channel 4, and form a Schottky contact or an Ohmic contact containing a clean van der Waals interface.

[0044] The supporting substrate 1 is a rigid or flexible substrate material, including Si, Al2O3, PET; the material of the dielectric layer 2 is SiO2, Al2O3 or h-BN; the source and drain electrodes are metal electrodes Ti, Cr, Au, Pt, Pd, graphene electrodes or other two-dimensional metal and semi-metal material electrodes; the material of the encapsulation layer 6 is h-BN or Al2O3.

[0045] CuInP2S6 is selected as a semiconductor material based on the combination of ion migration mechanism and ferroelectric polarization mechanism, which has both ferroelectric polarization control capability and ion migration control capability. The conductive channel 4 is photolithographically patterned, and the photolithographic pattern enables the top and bottom of the conductive channel 4 to form a conductive channel in the vertical direction.

[0046] Due to the displacement instability of specific ions in semiconductor materials that have both ion migration mechanism and ferroelectric polarization mechanism, it has two interesting properties: ferroelectricity characterized by eccentric order and ionic conduction characterized by long-range migration disorder. For ferroelectricity, its electric dipole mainly comes from the displacement of specific ions in the lattice. Below the Curie temperature, the spontaneous order of electric dipoles will produce macroscopic polarization, which can be switched by external electric field. This polarization shows excellent retention and has been widely used in devices such as ferroelectric diodes and ferroelectric tunnel junctions. On the other hand, specific ions also show ionic conduction through incoherent long-range motion. This behavior is promoted by atomic disorder and it can itself produce atomic disorder. Ionic conduction usually requires an electric field to maintain, which forms the basis of volatile memory. Interestingly, the ordered sites corresponding to ferroelectricity are both the initial sites of reversible ion motion and the stable sites after the relaxation of reversible ion motion, which is very similar to the neural reuse mechanism. This characteristic is manifested in the device as follows: the ion migration mode can continue to continuously adjust the device conductivity based on the ferroelectric polarization mode. Specifically, the volatile conductivity of the device can be modulated up (down) with the non-volatile conductivity as the center; and when the device switches from the ion migration working mode back to the ferroelectric polarization working mode, the non-volatile conductivity state originally corresponding to the ferroelectric polarization working mode can be naturally restored.

[0047] In addition, for ferroelectric polarization, the amplitude of the electrical excitation needs to be near the coercive field, so it has certain limitations, but it is not sensitive to the time of electrical excitation; for ion migration, the time of electrical excitation directly affects the degree of ion migration, and it has a wide tolerance range for the amplitude of electrical excitation. Therefore, for electrical excitation with short excitation time and high amplitude, the degree of ion migration is weak, while the influence of ferroelectric polarization is significant; for electrical excitation with long excitation time and low amplitude, it is not enough to cause ferroelectric polarization reversal, but it can significantly affect the degree of ion migration.

[0048] In view of this, we can achieve selective control of the device's working mode through precise electrical pulse engineering. Specifically, electric pulses with short excitation time and high amplitude (the amplitude should be in the coercive field range, and the pulse time matching is greater than the domain switching speed; for CuInP2S6, the amplitude of the electric pulse is about 7V, and the excitation time is 10ms) mainly affect ferroelectric polarization; electric pulses with long excitation time and low amplitude (the amplitude should be sufficient to make the ions jump out of the lattice, and the pulse time is evaluated according to the ion migration speed. For CuInP2S6, the amplitude of the electric pulse is 1.5V and the excitation time is 200ms) mainly drive ion migration. By applying carefully designed electric pulses, we can accurately switch between the two modes in one device. Furthermore, since the device conductivity modulation is continuous and superimposed between different modes, the two modes have a natural synergistic ability and can work together to carry out the same task. For example, learn the basic appearance characteristics of cats in the ferroelectric polarization mode (non-volatile), and continue to learn the specific characteristics of different breeds of cats in the ion migration mode (volatile).

[0049] The material of the conductive channel 4 is synthesized by mechanical stripping or chemical vapor deposition into a few layers or a single layer, and the structure is single crystal or polycrystalline.

[0050] This embodiment also provides a method for preparing the above electronic device, comprising the following steps:

[0051] 1) Prepare the metal bottom electrode (drain electrode): First, photolithography a strip mask pattern with a spacing of 10 μm and a width of 15 μm on the heavily doped Si / SiO2 substrate, and then use thermal evaporation coating equipment to A 40 nm thick Au film was deposited on the substrate at a rate of , and then the substrate was immersed in acetone solution for 30 min and lifted-off to obtain a patterned Au electrode.

[0052] 2) Preparation of PVA dry transfer film: First, prepare a 10 wt% PVA aqueous solution, then take 5 mL of the solution and drop it on a clean SiO2 substrate, then place it on a heating plate at 50°C and bake for 10 minutes to obtain a PVA film; then use a blade to cut a 5 mm × 5 mm PVA film, and place it on a PDMS substrate supported by a transparent glass sheet to obtain a PVA dry transfer film.

[0053] 3) PVA dry picking of channel materials: First, the sample material is obtained from the CuInP2S6 crystal material by mechanical stripping and placed on a heavily doped Si / SiO2 substrate. The substrate is then heated to 55°C using a sample stage, and then one corner of the PVA dry transfer film is slowly and evenly attached to the few-layer CuInP2S6. After the sample stage is cooled to room temperature, the transparent glass sheet is slowly lifted to achieve PVA dry picking of the channel material.

[0054] 4) PVA dry transfer of channel material: first, place the substrate with a metal bottom electrode on the sample stage, and with the help of a microscope, align the few-layer CuInP2S6 picked up on the PVA film with the metal bottom electrode on the bottom substrate; use the sample stage to heat the substrate with a metal bottom electrode to 55°C, and slowly lower the transparent glass sheet during the heating process until the few-layer CuInP2S6 on the PVA dry transfer film is completely in close contact with the metal bottom electrode; wait for 3 minutes for the sample stage to cool to room temperature, and slowly lift the transparent glass sheet to achieve PVA dry transfer of the channel material.

[0055] 5) Preparation of metal top electrode (source electrode): First, a stripe mask pattern with a spacing of 5 μm and a width of 10 μm is photolithographically formed on the few-layer CuInP2S6 after PVA dry transfer, and an overlapping area with the metal bottom electrode in the vertical direction is overlapped. Then, a thermal evaporation coating device is used to coat the top electrode with a thickness of 10 μm. A 40 nm thick Au film was deposited on the substrate at a rate of , and then the substrate was immersed in acetone solution for 30 min and lifted-off to obtain a patterned Au electrode.

[0056] 6) Preparation of Al2O3 encapsulation layer. First, place the sample device with the top electrode prepared in deionized water, heat it to 50°C for half an hour, and then take it out and blow it dry with nitrogen; then use atomic layer deposition equipment, use Al(CH3)3 and H2O as precursors, and deposit a 10nm thick Al2O3 encapsulation protection film on the surface of CuInP2S6 at a deposition temperature of 175°C; finally, use silver glue and gold wire to lead out the source and drain electrodes to complete the preparation.

[0057] Embodiment 2

[0058] This embodiment provides a method for preparing the neuromorphic electronic device described in Embodiment 1, comprising the following steps:

[0059] 1) Preparation of encapsulation layer h-BN and graphene electrode: Use heavily doped Si / SiO2 substrate as support, use mechanical exfoliation method to obtain a few-layer h-BN sample as h-BN film from h-BN crystal, and obtain single-layer or few-layer graphene samples from graphite layered crystals. Find suitable samples with the help of a microscope and mark their positions.

[0060] 2) Preparation of PVA dry transfer film: First, prepare a 10 wt% PVA aqueous solution, then take 5 mL of the solution and drop it on a clean SiO2 substrate, which is then placed on a heating plate at 50°C and baked for 1 hour to obtain a PVA film; then use a blade to cut a PVA film of about 5 mm × 5 mm in size, place it with the flat surface facing up and adhere it to a PDMS substrate supported by a transparent glass sheet to obtain a PVA dry transfer film.

[0061] 3) Dry transfer of graphene electrode: first heat the sample stage to 50°C, and with the assistance of a microscope and a three-micro displacement platform, use a corner of the PVA dry transfer film to slowly and evenly attach it to the h-BN of the heavily doped Si / SiO2 substrate. After cooling to room temperature, slowly lift the transparent glass sheet, and the PVA film will lift the h-BN film. Then repeat the operation twice, using h-BN as the target area, lift the graphene sample used as the top electrode, and realize the transfer of the top graphene electrode to the PVA film.

[0062] 4) Preparation of device channel: First, a few layers of sample material were obtained from CuInP2S6 crystal material by mechanical stripping method, and placed on SiO2 target substrate containing heavily doped Si. Then, SiO2 substrate was placed on the sample stage and heated to 55°C. The top graphene electrode and CuInP2S6 sample on the PVA transfer film were aligned with the help of a microscope, and then slowly and evenly pressed down. After cooling to room temperature, the glass sheet was slowly lifted up, and the PVA film lifted the graphene electrode and CuInP2S6 sample. The SiO2 substrate marked with the graphene sample used as the bottom electrode was placed on the sample stage and heated to 55°C. The top graphene electrode, CuInP2S6 sample and bottom graphene electrode on the PVA transfer film were aligned with the help of a microscope. After the PVA film was completely in contact with the SiO2 substrate, the glass sheet was slowly and evenly lifted after 2 minutes. The h-BN, top graphene electrode, and CuInP2S6 sample fell on the bottom graphene electrode of the SiO2 substrate together with the PVA film, and finally the device sample with the PVA film was obtained.

[0063] 5) Preparation of metal lead electrodes: First, place the device sample in deionized water, heat it to 50°C for half an hour, take it out and blow dry it with nitrogen to remove the PVA film; then use an electron beam lithography machine to in-situ expose and develop to prepare the lead electrode lithography pattern; then use thermal evaporation coating equipment to A Ti (5nm) / Au (50nm) metal film was prepared by deposition at a rate of ; the sample was then immersed in an acetone solution for lift-off to obtain a patterned metal lead-out electrode; finally, the source and drain electrodes were led out using silver glue and gold wire to complete the preparation.

[0064] Embodiment 3

[0065] This embodiment provides a method for preparing the neuromorphic electronic device described in Embodiment 1, comprising the following steps:

[0066] 1) Prepare the bottom metal Pt electrode on the SiO2 substrate: Place a hard metal mask containing an electrode pattern with a spacing width of 7um on the Si / SiO2 substrate, and then use a thermal evaporation coating device to A 50 nm thick Pt film was deposited on the substrate at a rate of , and the metal mask was removed to obtain a patterned Pt electrode.

[0067] 2) Preparation of PVA dry transfer film: First, prepare a 10 wt% PVA aqueous solution, then take 5 mL of the solution and drop it on a clean SiO2 substrate, then place it on a heating plate at 50°C and bake for 10 minutes to obtain a PVA film; then use a blade to cut a 5 mm × 5 mm PVA film, and place it on a PDMS substrate supported by a transparent glass sheet to obtain a PVA dry transfer film.

[0068] 3) PVA dry picking of channel materials: First, a few layers of sample material are obtained from CuInP2S6 crystal material by mechanical stripping method, and placed on a SiO2 target substrate containing heavily doped Si. Then, the substrate is heated to 55°C using a sample stage, and then a corner of the PVA dry transfer film is slowly and evenly attached to the few layers of CuInP2S6. After the heating stage is cooled to room temperature, the glass sheet is slowly lifted to achieve PVA dry picking of the channel material.

[0069] 4) PVA dry transfer of channel material: First, place the substrate with the prepared metal bottom electrode on the sample stage, and align the picked-up few-layer CuInP2S6 on the PVA film with the metal bottom electrode on the bottom substrate with the help of a microscope. Heat the substrate to 55°C using the sample stage, and slowly lower the transparent glass sheet during the process until the few-layer CuInP2S6 on the PVA dry transfer film is completely close to the bottom electrode; wait for 3 minutes to let the heating stage cool to room temperature, and slowly lift the transparent glass sheet to achieve the PVA dry transfer of the channel material.

[0070] 5) Preparation of metal top electrode: A single molecule film is self-assembled on a substrate containing a few layers of CuInP2S6 and a bottom electrode. A hard metal mask containing an electrode pattern with a spacing width of 7 μm is placed on the substrate and aligned with the top electrode in the vertical direction. Then, a thermal evaporation coating device is used to coat the top electrode. A 50 nm thick Pt film was deposited on the substrate at a rate of , and the metal mask was removed to obtain a patterned Pt electrode.

[0071] 6) Preparation of Al2O3 encapsulation layer. First, place the sample device with the top electrode prepared in deionized water, heat it to 50°C for half an hour, and then take it out and blow it dry with nitrogen; then use atomic layer deposition equipment, use Al(CH3)3 and H2O as precursors, and deposit a 10nm thick Al2O3 encapsulation protection film on the surface of CuInP2S6 at a deposition temperature of 175°C; finally, use silver glue and gold wire to lead out the source and drain electrodes to complete the preparation.

[0072] Experimental example

[0073] The relationship between the voltage and current of the CuInP2S6 neuromorphic electronic device prepared in Example 1 when the source and drain electrodes are input is measured, so as to obtain a memristor characteristic curve in the ferroelectric working mode, as shown in FIG. Figure 2 As shown, it is demonstrated that the two-dimensional ferroelectric semiconductor material CuInP2S6 of layered transition metal sulfide phosphate is used as the conductive channel, combined with its out-of-plane ferroelectric polarization process, to achieve the resistance memory behavior related to the source-drain voltage scanning range, namely the memristive characteristic.

[0074] The CuInP2S6 neuromorphic electronic device prepared in Example 1 was tested for the evolution of the device current over time under different ferroelectric polarization states. Figure 3 As shown, it is shown that at least 4 bits of ferroelectric polarization state (device conductance) modulation can be achieved under the action of an electric field. In addition, the device conductance exhibits typical non-volatile characteristics, indicating the excellent memristive maintenance ability of the device in the ferroelectric polarization working mode.

[0075] The relationship between the voltage and current of the CuInP2S6 neuromorphic electronic device prepared in Example 1 when the source and drain electrodes are input is measured, so as to obtain a memristor characteristic curve in the ion migration mode, as shown in FIG. Figure 4 As shown, it is shown that the second memristive characteristic related to the source-drain voltage scanning range is realized by using the two-dimensional ferroelectric semiconductor material CuInP2S6 of layered transition metal sulfide phosphate as the conductive channel and combining its internal ion migration process.

[0076] The CuInP2S6 neuromorphic electronic device prepared in Example 1 was tested in the ion migration working mode, and the device current evolution relationship with time was as follows: Figure 5 As shown, it is shown that the device conductivity exhibits typical volatile characteristics in the ion migration working mode, indicating the second type of device memristor maintenance ability, which is different from the ferroelectric polarization working mode.

[0077] The relationship between the device current and time evolution when the CuInP2S6 neuromorphic electronic device prepared in Example 1 works alternately in the ferroelectric polarization and ion migration modes is measured, as shown in FIG. Figure 6 As shown, the device not only has excellent working mode reconstruction capabilities, but the ion migration working mode also exhibits further device conductivity modulation capabilities based on the ferroelectric polarization mode. More importantly, even after multiple mode cyclic configurations, the non-volatile conductivity (ferroelectric polarization state) of the device can still be maintained. This unique electrical performance provides an excellent platform for simulating neural multiplexing, in which the ferroelectric polarization state is used to train general knowledge for learning different tasks, while ion migration is used to achieve specific training based on different tasks.

[0078] Semiconductor materials with both ion migration mechanisms and ferroelectric polarization mechanisms allow for a collaborative dual-operation mode driven by ferroelectric polarization and ion migration, which is very consistent with the key requirements of neural reuse. It not only realizes the comprehensive use of the memristor dynamics mechanism of neuromorphic devices, but also integrates the volatile and non-volatile characteristics of the device into a unified application framework. It also provides a novel control method for adjusting the electrical characteristics of the device, and provides an excellent way to build neuromorphic devices that more realistically simulate the characteristics of advanced biological neural networks and narrow the cognitive ability gap between memristor-based neural networks and biological neural networks. Crucially, the present invention opens up the possibility of neural reuse in memristor-based neural network platforms, which is expected to load efficient and mature artificial intelligence algorithms into existing memristor hardware as more powerful and effective engines. Using this network-level property as an internal driving force, this new refreshable device enables memristor-based neural networks to accommodate more efficient algorithms and architectures, such as pruning, parameter sharing, transfer learning, continuous learning, etc. in artificial intelligence algorithms, thereby greatly expanding the upper limit and potential of neuromorphic hardware applications. In addition, the present invention is also expected to provide a promising paradigm for constructing an in situ training hardware platform based on ferroelectric-ionic combinations.

[0079] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A neuromorphic electronic device based on ionic-ferroelectric synergy, characterized in that: The invention comprises a supporting substrate (1), a dielectric layer (2), a bottom electrode (3), a conductive channel (4), a top electrode (5) and a packaging layer (6) which are stacked in sequence from bottom to top; the conductive channel (4) is made of a semiconductor material having both an ion migration mechanism and a ferroelectric polarization mechanism; the bottom electrode (3) is a drain electrode, and the top electrode (5) is a source electrode; the source electrode and the drain electrode respectively cover the top and bottom of the conductive channel (4) and form a Schottky contact or an Ohmic contact containing a clean van der Waals interface.

2. The neuromorphic electronic device based on ionic-ferroelectric synergy according to claim 1, characterized in that: The supporting substrate (1) is made of a rigid or flexible substrate material, specifically Si, Al2O3 or PET; the dielectric layer (2) is made of SiO2, Al2O3 or h-BN; the source electrode and the drain electrode are metal electrodes Ti, Cr, Au, Pt, Pd, graphene electrodes or other two-dimensional metal and semi-metal material electrodes; the packaging layer (6) is made of h-BN or Al2O3.

3. The neuromorphic electronic device based on ionic-ferroelectric synergy according to claim 1, characterized in that: The conductive channel (4) is photolithographically provided with a pattern, and the photolithographic pattern enables the top and bottom of the conductive channel (4) to form a conductive channel in a vertical direction.

4. The neuromorphic electronic device based on ionic-ferroelectric synergy according to claim 1, characterized in that: The semiconductor material selected for the conductive channel (4) has both ferroelectric polarization mechanism and ion migration mechanism; For the ferroelectric polarization mechanism, the displacement of specific ions in the lattice generates electric dipoles, and the spontaneous order of the electric dipoles below the Curie temperature generates macroscopic polarization, which can be switched by external electric pulses; the ferroelectric polarization mechanism of the conductive channel (4) is insensitive to the application time of the external electric pulse and has retention, so the conductive channel (4) has non-volatility under the ferroelectric polarization mechanism; For the ion migration mechanism, specific ions exhibit ion conduction characteristics through long-distance disordered movement, and the degree of their migration is affected by the application time of the external electric pulse. Since the ion conduction characteristics require an electric field to maintain, the conductive channel (4) is volatile under the ion migration mechanism; The site order of the conductive channel (4) under the ferroelectric polarization mechanism is the basis for the modulation of the ion migration mechanism. The ion migration mechanism can continuously adjust the conductivity of the device based on the ferroelectric polarization mechanism, thereby realizing the modulation of the volatile conductivity of the device based on the non-volatile conductivity. When the device switches back from the disordered ion migration mechanism to the ordered ferroelectric polarization mechanism, the device naturally restores the non-volatile conductivity under the ferroelectric polarization mechanism.

5. The neuromorphic electronic device based on ionic-ferroelectric synergy according to claim 4, characterized in that A short excitation time and high amplitude electric pulse is applied between the source electrode and the drain electrode to make the device work under the ferroelectric polarization mechanism; a long excitation time and low amplitude electric pulse is applied between the source electrode and the drain electrode to realize the device conductivity regulation based on the ion migration mechanism on the basis of the ferroelectric polarization mechanism.

6. The neuromorphic electronic device based on ionic-ferroelectric synergy according to claim 1, characterized in that: The material of the conductive channel (4) is synthesized by mechanical stripping or chemical vapor deposition, and has a single crystal or polycrystalline structure.

7. A method for preparing a neuromorphic electronic device based on ionic-ferroelectric synergy, characterized in that: The following steps are involved: S1, preparing a patterned bottom electrode on the dielectric layer; S2, dissolving PVA powder in water to obtain a 5-15wt% solution, dropping it on the upper surface of the substrate and drying it at 40-60°C to form a film, and then placing the film on a PDMS supported by a transparent glass sheet to obtain a PVA transfer film for dry transfer; S3, stripping the semiconductor material having both ion migration mechanism and ferroelectric polarization mechanism to obtain nanosheets of set thickness and size; S4, using the PVA transfer film prepared in step S2 to stick the nanosheets prepared in step S3, and transferring them to the bottom electrode obtained in step S1 at 40-60°C; S5, using a photolithography process to obtain a set photolithography pattern, where the photolithography pattern covers a portion of the bottom electrode, so that the top and bottom of the nanosheet material form a conductive channel in the vertical direction; S6, using an evaporation coating device to deposit a thin film on the basis of the structure prepared in S5, and peeling off to obtain a patterned top electrode, thereby obtaining a neuromorphic electronic device with good metal-semiconductor van der Waals contact or ohmic contact; S7. Deposit a packaging layer on the surface of the neuromorphic electronic device obtained in step S6 by atomic layer deposition, and then lead out a source electrode and a drain electrode from the top electrode and the bottom electrode respectively.

8. The preparation method according to claim 7, characterized in that: Before step S1, the following process is performed: a dielectric layer is prepared on a supporting substrate by using a physical vapor deposition (PVD), a chemical vapor deposition (CVD) or an atomic layer deposition (ALD) method.

9. The preparation method according to claim 7, characterized in that: Step S4 can also be implemented in the following two ways. The first way is: placing a semiconductor material nanosheet having both an ion migration mechanism and a ferroelectric polarization mechanism on the bottom electrode prepared in step S2 by spin coating; the second way is: growing a semiconductor material having both an ion migration mechanism and a ferroelectric polarization mechanism on the bottom electrode prepared in step S2 by physical vapor deposition PVD, chemical vapor deposition CVD or atomic layer deposition ALD.

10. The preparation method according to claim 7, characterized in that: The specific process of step S1 is: photolithography patterning on the dielectric layer, and then using thermal evaporation or electron beam coating equipment to deposit a 40-60nm thick film on the dielectric layer, and then immersing the substrate in methyl pyrrolidone NMP solution for 5-10 minutes and then peeling it off to obtain a patterned bottom electrode.

Images