Two-dimensional van der waals heterostructure devices with programmable non-volatile ambipolar photoluminescence

Abstract

This invention relates to a two-dimensional van der Waals heterojunction device with a programmable non-volatile bipolar optical response. The device employs a channel layer / first insulating layer / floating gate layer / second insulating layer structure, wherein a source and drain are located between the channel layer and the first insulating layer, and the channel region between the source and drain is completely covered vertically by the channel layer, the first insulating layer, the floating gate layer, and the second insulating layer. Carrier exchange occurs between the floating gate layer and the gate under the action of a gate voltage pulse or a light pulse, realizing a programmable non-volatile bipolar optical response. This device can simultaneously realize optical sensing, storage, and computing functions, solving the problems of limited functionality, asymmetrical response, limited number of states, and low system integration of existing two-dimensional material devices. This invention also provides a method for fabricating a large-area two-dimensional van der Waals heterojunction device array. Applying the device array to motion detection and pattern recognition can achieve integrated sensing, storage, and computing, improving processing efficiency and reducing system power consumption.

Description

Technical Field

[0001] This invention belongs to the field of inorganic semiconductor nanomaterials and optoelectronic devices, and relates to a two-dimensional van der Waals heterojunction device with programmable non-volatile bipolar optical response. Background Technology

[0002] Two-dimensional materials have attracted widespread attention due to their atomically thin thickness and high compatibility with silicon-based technologies. These materials can be freely assembled into heterojunctions via van der Waals forces, thereby integrating the properties of different materials and providing a novel platform for researching new electronic and optoelectronic devices. Currently, van der Waals heterojunction devices based on two-dimensional materials have shown considerable potential in optoelectronic sensing and storage, but they still face challenges such as limited functionality, asymmetric response, limited number of states, and low system integration. Most devices can only achieve a single optoelectronic function or require external circuitry for signal processing, making it difficult to meet the requirements of intelligent sensing systems for real-time operation, low power consumption, and high integration. Especially in applications such as motion detection and motion pattern recognition, the lack of a device platform that can simultaneously achieve non-volatile control of optical response, multi-state storage, and integrated sensing-storage-computing capabilities limits their development in intelligent sensing systems. Summary of the Invention

[0003] To address the problems of existing optoelectronic devices, such as limited functionality, asymmetrical response, limited number of states, and low system integration, this invention provides a two-dimensional van der Waals heterojunction device with a programmable non-volatile bipolar optical response.

[0004] The technical solution provided by this invention is as follows: In a first aspect, the present invention provides a two-dimensional van der Waals heterojunction device, comprising, from bottom to top, a channel layer, a first insulating layer, a floating gate layer, a second insulating layer, and a gate, wherein: A source and a drain are provided between the channel layer and the first insulating layer, and the channel region between the source and the drain is completely covered in the vertical direction by the channel layer, the first insulating layer, the floating gate layer and the second insulating layer.

[0005] In conjunction with the first aspect of the invention, some embodiments include: The thickness of the channel layer is 5-8 nanometers; and / or, The thickness of the first insulating layer is 20-60 nanometers; and / or, The thickness of the floating gate layer is 10-20 nanometers; and / or, The thickness of the second insulating layer is 15-25 nanometers; and / or, The width of the channel region is 3 to 12 micrometers.

[0006] In conjunction with the first aspect of the invention, some embodiments include: The thickness of the channel layer is 6-7 nanometers; and / or, The thickness of the first insulating layer is 30-35 nanometers; and / or, The thickness of the floating gate layer is 12-13 nanometers; and / or, The thickness of the second insulating layer is 15-16 nanometers; and / or, The width of the channel region is 5 to 10 micrometers.

[0007] In conjunction with the first aspect of the invention, some embodiments include: The channel layer is molybdenum distellide; and / or, The first insulating layer is aluminum oxide; and / or, The floating grid layer is tungsten diselenide; and / or, The second insulating layer is hexagonal boron nitride.

[0008] In conjunction with the first aspect of the invention, some embodiments include: The molybdenum diselleride, tungsten diselenide, and hexagonal boron nitride are all two-dimensional layered materials; The source and drain electrodes are made of one or more of the following materials: gold, silver, copper, chromium, palladium, and platinum.

[0009] Secondly, the present invention provides a method for preparing the above-mentioned two-dimensional van der Waals heterojunction device, comprising the following steps: (a) Providing a molybdenum distellide thin film as a channel layer; (b) The channel layer is patterned using ultraviolet lithography and reactive ion etching processes to form the channel region; (c) The electrode region is exposed on the channel layer by ultraviolet photolithography, and the source and drain are fabricated in the electrode region by metal deposition method; (d) Alumina is deposited on the channel layer and the source and drain electrodes by atomic layer deposition process as a first insulating layer; (e) By means of a transfer method, a tungsten diselenide thin film as a floating gate layer and a hexagonal boron nitride thin film as a second insulating layer are sequentially placed on the first insulating layer; (f) The stacked heterostructure is patterned using ultraviolet lithography and reactive ion etching processes, so that the channel layer, the first insulating layer, the floating gate layer and the second insulating layer have a completely vertically overlapping area, and this area covers the channel area of ​​the channel layer; (g) A gate is fabricated on the second insulating layer by ultraviolet lithography and metal deposition methods to obtain a large-area array or a single device based on the two-dimensional van der Waals heterojunction device.

[0010] In conjunction with the second aspect of the invention, some embodiments include: The preparation of source and drain electrodes in the electrode region using a metal deposition method includes: sequentially depositing a chromium layer and a gold layer, with the chromium layer having a thickness of 5-15 nanometers and the gold layer having a thickness of 30-80 nanometers.

[0011] Thirdly, the present invention provides the application of the above-mentioned two-dimensional van der Waals heterojunction device in moving target detection or motion pattern recognition.

[0012] Fourthly, the present invention provides a method for simulating the detection of moving objects, comprising the following steps: Acquire the two frames of image data to be detected and convert them into grayscale images, represented by a grayscale value matrix; Using the normalized photocurrent data obtained by the two-dimensional van der Waals heterojunction device in the positive and negative photoresponse modes, a positive photocurrent matrix and a negative photocurrent matrix are generated. The two grayscale matrices are respectively subjected to Hadamard product operation with the positive photocurrent matrix and the negative photocurrent matrix. The two Hadamard products are summed and converted into a grayscale image. Pixels with grayscale values ​​less than a preset threshold are set to zero to obtain the moving target detection result. When the two frames to be detected contain moving objects, the detection result displays the outline information of the objects; otherwise, the outline information of the objects is not displayed.

[0013] Fifthly, the present invention provides a method for simulating motion pattern recognition, comprising the following steps: Construct a convolutional neural network, train it on the input data, and obtain the optimal network parameters; The multi-level storage states of the above two-dimensional van der Waals heterojunction device are mapped to the optimal network parameters, wherein the multi-level storage states are generated by carrier exchange between the floating gate layer and the gate under the action of gate voltage pulse or optical pulse. The optimal network parameters after mapping are input into the convolutional neural network to identify the motion patterns of moving targets.

[0014] Compared with the prior art, the present invention has at least the following beneficial effects: 1. In the two-dimensional van der Waals heterojunction device of the present invention, carrier exchange occurs between the floating gate layer and the gate under the action of gate voltage pulse or optical pulse, realizing the functions of sensing, storage and calculation in a single device, and solving the problem of low system integration of traditional optoelectronic devices.

[0015] 2. The current of the two-dimensional van der Waals heterojunction device provided by the present invention can rise (or fall) non-volatilely under light pulses, and the photoresponse has excellent adjustability. It can be finely adjusted by light intensity, light pulse width, pulse number, etc., which solves the problems of asymmetric response and limited number of states of traditional devices.

[0016] 3. By writing gate voltage pulses or applying light pulses, the present invention enables the device to exhibit stable multi-level storage states (greater than 8 bits), and the storage states can be continuously adjusted by light pulses, thus solving the problem of the limited number of states in existing optoelectronic devices.

[0017] 4. This invention uses a large-area transfer technique combined with atomic layer deposition and reactive ion etching processes to fabricate large-area arrays of two-dimensional van der Waals heterojunction devices, thus solving the problem of complex large-area fabrication processes. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0019] Figure 1 This is a schematic diagram of the structure of each device in the large-area two-dimensional van der Waals heterojunction device array prepared in Example 1. The figure clearly shows the channel layer (molybdenum distellide), the first insulating layer (alumina), the floating gate layer (tungsten diselenide), and the second insulating layer (hexagonal boron nitride) arranged sequentially from bottom to top, as well as the completely vertically overlapping region that covers the channel region of the channel layer.

[0020] Figure 2 A scanning electron microscope false-color image of a typical two-dimensional van der Waals heterojunction device in the two-dimensional van der Waals heterojunction device array prepared in Example 1 is shown, demonstrating the uniform stacking structure of each layer of the device, especially the precise alignment of the completely vertically overlapping regions.

[0021] Figure 3 The image shows a portion of the large-area two-dimensional van der Waals heterojunction device array prepared in Example 1, demonstrating the large-area uniformity of the device array, with the size consistency deviation of the completely vertically overlapping regions of each device being less than 5%.

[0022] Figure 4 and Figure 5 When the two-dimensional van der Waals heterojunction device obtained in Embodiment 2 of the present invention operates as a photoresponse transistor, the positive photoresponse mode ( Figure 4 ) and negative light response mode ( Figure 5 The response characteristics of positive (negative) photocurrents modulated by periodic optical pulses under different gate voltage pulses demonstrate the realization of programmable nonvolatile bipolar optical response.

[0023] Figure 6 The study demonstrates the positive and negative photocurrents with matching absolute values ​​under continuous light pulses when the two-dimensional van der Waals heterojunction device prepared in Example 2 operates in the positive and negative photoresponse modes of a photoresponse transistor.

[0024] Figure 7The relationship between normalized positive and negative photocurrents and the number of light pulses was demonstrated. Linear fitting was performed on the positive and negative photocurrents and the number of light pulses, with fitting determination coefficients of 0.9998 and 0.9983, respectively, showing good linearity. This allows the polymorphic positive and negative photocurrents of the device to be effectively mapped onto the neural network, which improves the accuracy of subsequent motion pattern recognition tasks.

[0025] Figure 8 The stability of the two-dimensional van der Waals heterojunction device prepared in Example 2 when it is used as a non-volatile memory is demonstrated, with 17 photocurrent states (including 8 positive photocurrent states and 8 negative photocurrent states).

[0026] Figure 9 The results of the device durability test are shown when the two-dimensional van der Waals heterojunction device prepared in Example 2 is used as a non-volatile memory.

[0027] Figure 10 and Figure 11 When the two-dimensional van der Waals heterojunction device obtained in Embodiment 2 of the present invention operates as a non-volatile memory, the positive optical response mode ( Figure 10 ) and negative light response mode ( Figure 11 The multi-photocurrent state characteristics under continuous light pulses demonstrate the ability to precisely adjust the storage state through light pulses.

[0028] Figure 12 Example 2 shows a flowchart of the two-dimensional van der Waals heterojunction device simulating a moving target detection task, demonstrating the entire process of the device realizing moving target detection through the inter-frame differential recognition method.

[0029] Figure 13 a and Figure 13 b: The convolutional neural network structure for simulating motion pattern recognition tasks using the two-dimensional van der Waals heterojunction device obtained in Example 2 ( Figure 13 a) and the final output recognition accuracy ( Figure 13 (b) proves that the device array can be directly mapped to convolutional neural network parameters to achieve motion pattern recognition that integrates sensing, storage, and computing. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0031] The two-dimensional van der Waals heterojunction device provided by this invention includes, from bottom to top, a channel layer, a first insulating layer, a floating gate layer, a second insulating layer, and a gate, wherein: A source and a drain are provided between the channel layer and the first insulating layer, and the channel region between the source and the drain is completely covered by the channel layer, the first insulating layer, the floating gate layer and the second insulating layer in the vertical direction.

[0032] This two-dimensional van der Waals heterojunction device employs a channel layer / first insulating layer / floating gate layer / second insulating layer / gate structure. The first insulating layer blocks direct charge transport between the channel layer and the floating gate layer, while the second insulating layer forms a controllable interface barrier between the floating gate layer and the gate. The floating gate layer can simultaneously and stably store electrons and holes. Under stimulation by a gate voltage pulse or light pulse, carrier exchange occurs between the floating gate layer and the gate. Positive light response mode: Applying a negative gate voltage pulse injects electrons into the floating gate layer. The light pulse excites electrons in the channel layer to overcome the interface barrier and enter the floating gate layer, realizing the writing of the electronic storage state; Negative light response mode: Applying a positive gate voltage pulse injects holes into the floating gate layer, and the light pulse excites holes in the channel layer to enter the floating gate layer, realizing the writing of the hole storage state.

[0033] This two-dimensional van der Waals heterojunction device exhibits a bipolar optical response, meaning it operates in two modes: a positive optical response mode and a negative optical response mode. In the positive optical response mode, electrons are injected into the floating gate layer via a negative gate voltage pulse. This injection puts the channel in a high-resistivity state, and the optical pulse allows the electrons to gain photon energy, overcome the interface barrier, and enter the gate metal, resulting in a weakening of the p-type doping effect and generating a positive photocurrent. In the negative optical response mode, holes are injected into the floating gate layer via a positive gate voltage pulse. This injection puts the channel in a low-resistivity state, and the optical pulse allows the holes to gain photon energy, overcome the interface barrier, and enter the gate metal, resulting in a weakening of the n-type doping effect and generating a negative photocurrent. This bipolar optical response mode allows the device array to form a symmetrical positive and negative photocurrent matrix. Through inter-frame differential recognition, it can accurately distinguish moving targets from the background, significantly improving the accuracy and reliability of motion detection. Traditional motion monitoring systems require complex external circuitry to process optical signals, while the bipolar optical response mode of this invention allows the device itself to perform signal processing without additional circuitry, greatly simplifying system design. The combination of dual-polarity optical response mode and the device's multi-state storage capability enables each device to store multiple states, supporting more complex image processing and pattern recognition.

[0034] The programmability of this two-dimensional van der Waals heterojunction device enables precise writing (electron / hole injection) and erasure (charge dissipation) of the floating gate layer charge state by controlling the polarity of the gate voltage pulse or the parameters of the optical pulse (including at least one of light intensity, pulse width, and pulse number). Fine-tuning of the optical pulse parameters precisely adjusts the number of charge carriers stored in the floating gate layer; the coordinated control of the optical pulse and the gate voltage pulse further enhances the device's control precision, enabling programmable switching of the bipolar charge state. This device supports multi-state storage of more than 8 bits, providing a high-density, low-power hardware foundation for complex signal processing by stably storing multiple charge states.

[0035] This two-dimensional van der Waals heterojunction device is non-volatile, meaning that the charge state stored in the floating gate layer can be maintained stably for a long time after power is turned off, without the need for an external power source, thus ensuring data persistence.

[0036] The channel region between the source and drain is completely covered in the vertical direction by the channel layer, the first insulating layer, the floating gate layer, and the second insulating layer, ensuring the efficient implementation of the carrier exchange mechanism.

[0037] By adjusting the polarity of the gate voltage pulse and the parameters of the light pulse, the floating gate layer can realize bipolar programmable charge storage (electron / hole state switching), enabling the device to simultaneously complete optical signal sensing, charge storage and logic operation within a single structure without the need for external circuit connections. This solves the technical bottlenecks of existing two-dimensional material devices, such as single function, asymmetrical response and low system integration.

[0038] In some embodiments, the channel layer in the device needs to be thin enough to achieve good electrical performance and carrier mobility while maintaining mechanical stability; the first insulating layer needs to be thick enough to ensure effective charge isolation between the channel layer and the floating gate layer, preventing unwanted charge transport; the floating gate layer needs to be thick enough to stably store electrons and holes while maintaining good interface characteristics; the second insulating layer needs to be of appropriate thickness to ensure high-quality insulation and interface characteristics while allowing controlled exchange of carriers downloaded by the optical pulse. In some embodiments, the thickness of the channel layer is 5-8 nanometers, the thickness of the first insulating layer is 20-60 nanometers, the thickness of the floating gate layer is 10-20 nanometers, the thickness of the second insulating layer is 15-25 nanometers, and the source and drain electrodes are made of one or more of gold, silver, copper, chromium, palladium, and platinum. Preferably, the thickness of the channel layer is 6-7 nanometers, the thickness of the first insulating layer is 30-35 nanometers, the thickness of the floating gate layer is 12-13 nanometers, and the thickness of the second insulating layer is 15-16 nanometers.

[0039] The method for fabricating two-dimensional van der Waals heterojunction devices provided by this invention includes the following steps: (a) Providing a molybdenum distellide thin film as a channel layer; (b) The channel layer is patterned using ultraviolet lithography and reactive ion etching processes to form the channel region; (c) The electrode region is exposed on the channel layer by ultraviolet photolithography, and the source and drain are fabricated in the electrode region by metal deposition method; (d) Alumina is deposited on the channel layer and the source and drain electrodes by atomic layer deposition process as a first insulating layer; (e) By means of a transfer method, a tungsten diselenide thin film as a floating gate layer and a hexagonal boron nitride thin film as a second insulating layer are sequentially placed on the first insulating layer; (f) The stacked heterostructure is patterned using ultraviolet lithography and reactive ion etching processes, so that the channel layer, the first insulating layer, the floating gate layer and the second insulating layer have a completely vertically overlapping area, and this area covers the channel area of ​​the channel layer; (g) A gate is fabricated on the second insulating layer by ultraviolet lithography and metal deposition methods to obtain a large-area array or a single device based on the two-dimensional van der Waals heterojunction device.

[0040] The method for fabricating two-dimensional van der Waals heterojunction devices provided by this invention employs large-area transfer technology, which can fabricate large-area arrays, laying the foundation for practical applications.

[0041] In some embodiments, the preparation of source and drain electrodes in the electrode region using a metal deposition method includes: sequentially depositing a chromium layer and a gold layer, wherein the thickness of the chromium layer is 5-15 nanometers and the thickness of the gold layer is 30-80 nanometers. The chromium layer acts as an adhesion layer, effectively improving the adhesion between the metal and the molybdenum distellide channel layer; gold possesses excellent conductivity, chemical stability, and compatibility with two-dimensional materials; the chromium-gold bilayer electrode obtained by sequential deposition of the two layers exhibits good adhesion, good conductivity, and high stability.

[0042] In some embodiments, the channel layer is synthesized directly on the substrate or transferred onto the substrate by a transfer method. Preferably, the substrate is a silicon / silicon dioxide substrate, wherein the thickness of the silicon dioxide is 300±10 nm and the thickness of the channel layer is 6 nm.

[0043] In some implementations, step (e) includes: (e1) Polymethyl methacrylate support films were suspended on silicon / silica substrates with large-area tungsten diselenide nanosheets and large-area hexagonal boron nitride nanosheets synthesized by chemical vapor deposition, respectively, and baked at 100 degrees Celsius for 30 minutes with a heating plate. (e2) In deionized water, the polymethyl methacrylate support film was separated from the substrate, and tungsten diselenide nanosheets and hexagonal boron nitride nanosheets were respectively attached to a polymethyl methacrylate support film. (e3) With the aid of an optical microscope, tungsten diselenide nanosheets and hexagonal boron nitride nanosheets were manually placed sequentially on the aforementioned alumina thin film layer. After each placement, the polymethyl methacrylate support film was dissolved in acetone solvent.

[0044] This invention provides a method for simulating the detection of moving objects, comprising the following steps: Acquire the two frames of image data to be detected and convert them into grayscale images, represented by a grayscale value matrix; Using the normalized photocurrent data obtained by the two-dimensional van der Waals heterojunction device in the positive and negative photoresponse modes, a positive photocurrent matrix and a negative photocurrent matrix are generated. The two grayscale matrices are respectively subjected to Hadamard product operation with the positive photocurrent matrix and the negative photocurrent matrix. The two Hadamard products are summed and converted into a grayscale image. Pixels with grayscale values ​​less than a preset threshold are set to zero to obtain the moving target detection result. When the two frames to be detected contain moving objects, the detection result displays the outline information of the objects; otherwise, the outline information of the objects is not displayed.

[0045] This invention provides a method for simulating motion pattern recognition, comprising the following steps: Construct a convolutional neural network, train it on the input data, and obtain the optimal network parameters; The multi-level storage states of the above two-dimensional van der Waals heterojunction device are mapped to the optimal network parameters, wherein the multi-level storage states are generated by carrier exchange between the floating gate layer and the gate under the action of gate voltage pulse or optical pulse. The optimal network parameters after mapping are input into the convolutional neural network to identify the motion patterns of moving targets.

[0046] The materials used in the examples can be prepared by conventional methods in the art or purchased directly from the market. The large-area molybdenum ditelluride nanosheets were purchased from C6C Technologies and are multilayered with a thickness ranging from 5 to 8 nanometers.

[0047] Example 1 This embodiment provides a method for fabricating a large-area two-dimensional van der Waals heterojunction device array, including the following steps: (1) The large-area molybdenum distellide nanosheets, tungsten diselenide nanosheets and hexagonal boron nitride nanosheets were all purchased and synthesized by chemical vapor deposition. Among them, the large-area molybdenum distellide nanosheets and tungsten diselenide nanosheets were directly grown on a silicon / silica substrate with a silicon dioxide thickness of 300 nm; the large-area hexagonal boron nitride nanosheets were grown on copper foil.

[0048] (2) AZ5214 photoresist was spin-coated twice onto a large-area molybdenum ditelluride nanosheet. Each spin-coating speed was 5000 rpm, and each spin-coating time was 50 seconds. After the two spin-coatings were completed, the nanosheets were baked at 100 degrees Celsius for 1 minute using a hot plate. Then, the area to be etched was exposed using ultraviolet lithography. The nanosheets were then immersed in 3038 positive photoresist developer for 30 seconds and dried with a nitrogen gun to expose the area to be etched. The sample was then etched using reactive ion etching with carbon tetrafluoride gas at a flow rate of 200 standard cubic centimeters per minute for 30 seconds. After etching, the sample was placed in acetone for 10 minutes to remove the photoresist on the surface and finally dried with a nitrogen gun. In the end, the large-area molybdenum ditelluride nanosheets were etched into an array with a channel width of 10 micrometers and an array size of 27×27.

[0049] (3) AZ5214 photoresist was spin-coated once onto a large-area molybdenum ditelluride nanosheet at a spin speed of 5000 rpm for 50 seconds. After spin-coating, the nanosheet was baked at 100 degrees Celsius for 1 minute using a hot plate. Then, the source and drain electrode areas were exposed using ultraviolet lithography, soaked in 3038 positive photoresist developer for 30 seconds, and then dried with a nitrogen gun to expose the areas where metal electrodes needed to be deposited. Metal was deposited using a metal deposition method, with a deposition sequence of 5 nm thick chromium and 30 nm thick gold. After metal deposition, the sample was placed in acetone for 10 minutes to remove the photoresist and the metal on top of it, and finally dried with a nitrogen gun. Ultimately, a source and drain metal electrode array with a size of 27×27 was formed on top of the molybdenum ditelluride nanosheet. For each unit of the array, the channel length of molybdenum ditelluride was 10 μm.

[0050] (4) A 30 nm thick aluminum oxide film was deposited on molybdenum ditelluride and a metal electrode array by atomic layer deposition. The aluminum source used for deposition was trimethylaluminum, the deposition cycle was 300, and the deposition time of the aluminum source in each cycle was 0.02 seconds.

[0051] (5) Polymethyl methacrylate (PMMA) solution was spin-coated onto large-area tungsten diselenide nanosheets and hexagonal boron nitride nanosheets respectively. The spin-coating speed was 2000 rpm, and the spin-coating time was 30 seconds. After spin-coating, the PMMA solution was baked at 100 degrees Celsius for 30 minutes to solidify it into a support film. The mass fraction of the PMMA solution was 4%, the solute was PMMA, and the solvent was anisole. After heating, the nanosheets were placed in deionized water. The support film could be naturally separated from the substrate, and the tungsten diselenide nanosheets and hexagonal boron nitride nanosheets adhered to the two support films respectively.

[0052] (6) With the aid of an optical microscope, the support film with tungsten diselenide nanosheets and hexagonal boron nitride nanosheets attached was placed sequentially on the aforementioned alumina thin film layer, with the side with the support film facing upwards. After each placement, the sample was placed in acetone for 30 minutes to remove the support film, and then dried with a nitrogen gun.

[0053] (7) Using the same steps as in step (2), the above-stacked heterostructure is patterned using ultraviolet lithography and reactive ion etching processes to etch large-area tungsten diselenide nanosheets and hexagonal boron nitride nanosheets into an array of 80 micrometers × 65 micrometers with an array size of 27 × 27.

[0054] (8) Using the same steps as in step (3), a metal electrode array is prepared on the hexagonal boron nitride layer by ultraviolet lithography and metal deposition. The deposition sequence is a 5-nanometer thick chromium layer and a 60-nanometer thick gold layer, thus obtaining a large-area two-dimensional van der Waals heterojunction device array.

[0055] Figure 1 This is a schematic diagram of the structure of each device in the large-area two-dimensional van der Waals heterojunction device array prepared in Example 1. All components have been clearly identified. Metal electrodes placed on molybdenum ditelluride nanosheets serve as the source and drain, respectively, while metal electrodes placed on hexagonal boron nitride nanosheets serve as the gate. The conductive channel is a molybdenum ditelluride lateral channel.

[0056] Figure 2 A scanning electron microscope false-color image of a typical two-dimensional van der Waals heterojunction device from the two-dimensional van der Waals heterojunction device array prepared in Example 1 is shown. All components are clearly indicated. Metal electrodes placed on molybdenum ditelluride nanosheets serve as the source and drain, respectively, while a metal electrode placed on hexagonal boron nitride nanosheets serves as the gate.

[0057] Figure 3 This is a scanning electron microscope image of a portion of the large-area two-dimensional van der Waals heterojunction device array prepared in Example 1.

[0058] Example 2 This embodiment provides a method for fabricating a single two-dimensional van der Waals heterojunction device, specifically including the following steps: (1) Molybdenum ditelluride nanosheets were prepared by repeatedly attaching the corresponding bulk material with transparent tape. The tape with the attached molybdenum ditelluride nanosheets was attached to the substrate (silicon / silica / alumina, with a silicon dioxide thickness of 300 nm and an alumina thickness of 30 nm), and then the tape was directly peeled off to obtain molybdenum ditelluride nanosheets on the substrate. The required molybdenum ditelluride nanosheets were selected by optical microscopy and atomic force microscopy, and their thickness was 6 nm.

[0059] (2) AZ5214 photoresist was spin-coated twice onto molybdenum ditelluride nanosheets. Each spin-coating speed was 5000 rpm, and each spin-coating time was 50 seconds. After the two spin-coatings were completed, the nanosheets were baked at 100 degrees Celsius for 1 minute using a hot plate. Then, the areas to be etched were exposed using ultraviolet lithography. The nanosheets were then immersed in 3038 positive photoresist developer for 30 seconds and dried with a nitrogen gun to expose the areas to be etched. The developed sample was then etched using reactive ion etching (RIE) with carbon tetrafluoride gas at a flow rate of 200 standard cubic centimeters per minute for 30 seconds. After etching, the sample was placed in acetone for 10 minutes to remove the photoresist from the surface and finally dried with a nitrogen gun. Ultimately, a large area of ​​molybdenum ditelluride nanosheets was etched into strips with a channel width of 5 micrometers.

[0060] (3) AZ5214 photoresist was spin-coated once onto the etched molybdenum ditelluride nanosheet at a spin speed of 5000 rpm for 50 seconds. After spin-coating, the nanosheet was baked at 100 degrees Celsius for 1 minute using a hot plate. Then, the source and drain electrode areas were exposed using ultraviolet lithography, soaked in 3038 positive photoresist developer for 30 seconds, and then dried with a nitrogen gun to expose the areas where metal electrodes needed to be deposited. Metal was deposited using a metal deposition method, with a deposition sequence of 5 nm thick chromium and 30 nm thick gold. After metal deposition, the sample was placed in acetone for 10 minutes to remove the photoresist and the metal on top of it, and finally dried with a nitrogen gun. Ultimately, source and drain metal electrodes were formed on top of the molybdenum ditelluride nanosheet. The channel length of the molybdenum ditelluride was 5 μm.

[0061] (4) A 30 nm thick aluminum oxide film was deposited on molybdenum ditelluride and a metal electrode array by atomic layer deposition. The aluminum source used for deposition was trimethylaluminum, the deposition cycle was 300, and the deposition time of the aluminum source in each cycle was 0.02 seconds.

[0062] (5) Spin-coat a polypropylene carbonate solution (solvent is anisole, mass fraction is 20%) onto a silicon / silica substrate at a spin speed of 2000 rpm for 45 seconds. After spin coating, bake the solution at 80 degrees Celsius for 1 minute with a heating plate to solidify the spin-coated solution into a polypropylene carbonate support film.

[0063] (6) Both tungsten diselenide nanosheets and hexagonal boron nitride nanosheets were prepared by repeatedly attaching the corresponding bulk materials with transparent tape. The tape with the attached tungsten diselenide nanosheets and hexagonal boron nitride nanosheets was attached to two substrates with polypropylene carbonate support films in step (5), and then the tape was directly peeled off. Tungsten diselenide nanosheets and hexagonal boron nitride nanosheets could be obtained on the polypropylene carbonate support films. The required tungsten diselenide nanosheets and hexagonal boron nitride nanosheets were selected by optical microscopy and atomic force microscopy, and their thicknesses were 12 nm and 15 nm, respectively. Then they were placed in deionized water, and the polypropylene carbonate support films could be naturally separated from the silicon / silica substrates. The tungsten diselenide nanosheets and hexagonal boron nitride nanosheets were respectively attached to the two polypropylene carbonate support films.

[0064] (7) With the aid of an optical microscope, the polypropylene carbonate support film with tungsten diselenide nanosheets and hexagonal boron nitride nanosheets attached was placed sequentially on the aforementioned alumina thin film layer, with the side with the support film facing upwards. After each placement, the sample was placed in acetone for 30 minutes, the support film was removed, and then dried with a nitrogen gun.

[0065] (8) Using the same steps as in step (3), a metal electrode array is prepared on the hexagonal boron nitride layer by ultraviolet lithography and metal deposition. The deposition sequence is 5 nm thick chromium and 60 nm thick gold, thus obtaining a single device based on the two-dimensional van der Waals heterojunction.

[0066] The following section briefly explains the performance characteristics of two-dimensional van der Waals heterojunction devices: Figure 4 and Figure 5 This paper demonstrates the modulation of positive (negative) photocurrents by periodic optical pulses under different gate voltage pulses when the two-dimensional van der Waals heterojunction device prepared in Example 2 operates in both positive and negative photoresponse modes of a phototransistor. The period of each optical pulse is 2 seconds, the pulse width is 30 milliseconds and 200 milliseconds, and the optical power is 10 milliwatts and 12 milliwatts, respectively. It can be seen that the two-dimensional van der Waals heterojunction device exhibits non-volatile photoresponses in opposite directions to a single optical pulse in both operating modes, and superimposed non-volatile photoresponses to continuous optical pulses. Notably, the two-dimensional van der Waals heterojunction device also exhibits photoresponse without an applied gate voltage pulse, which is unprecedented in similar devices.

[0067] Figure 6This diagram illustrates the positive and negative photocurrents, with matching absolute values, under continuous light pulses when the two-dimensional van der Waals heterojunction device fabricated in Example 2 operates in both positive and negative photoresponse modes as a phototransistor. For both positive and negative photoresponses, the drain voltage is 0.5 volts, the gate voltage pulses are 0 and -6 volts, respectively, the pulse widths are 20 milliseconds and 150 milliseconds, and the optical power is 10 milliwatts for both. As can be seen from the figure, after 10 light pulses, the absolute values ​​of both the positive and negative photocurrents are approximately 1.5 microamps, indicating good programmability of the device. This means that the intensity of the positive or negative photoresponse can be adjusted by modifying the top gate voltage and pulse width to suit different application scenarios. Figure 7 The relationship between normalized positive and negative photocurrents and the number of light pulses was demonstrated. Linear fitting was performed on the positive and negative photocurrents and the number of light pulses, with coefficients of determination of 0.9998 and 0.9983, respectively, showing good linearity. This allows the multi-state positive and negative photocurrents of the device to be effectively mapped onto the neural network, which improves the accuracy of subsequent motion pattern recognition tasks. Figures 9-11 It can be seen that when the fabricated two-dimensional van der Waals heterojunction device is used as a photoresponse transistor, the photoresponse current has excellent bipolarity, programmability and linear superposition.

[0068] Figure 8 The stability measurements of 17 photocurrent states (including 8 positive photocurrent states and 8 negative photocurrent states) of the two-dimensional van der Waals heterojunction device prepared in Example 2 when operating as a non-volatile memory are shown. As can be seen from the figure, there is a clear distinction between the various states, and the positive and negative photocurrent states do not decay over time during the test period, exhibiting good stability. Figure 9 The results of the device endurance test are shown when the two-dimensional van der Waals heterojunction device prepared in Example 2 operates as a non-volatile memory. In each cycle, the device is first reset to a positive photoresponse mode using a -20V gate voltage pulse, and the photocurrent change under continuous light pulses is measured. Then, the device is reset to a negative photoresponse mode using a +20V gate voltage pulse, and the photocurrent change under continuous light pulses is measured again. Figure 9 As can be seen, after 52 cycles, the device can operate normally in both modes, demonstrating its excellent durability when switching operating states.

[0069] Figure 10 and Figure 11 The multi-photocurrent state characteristics of the two-dimensional van der Waals heterojunction device prepared in Example 2, operating in both positive and negative photoresponse modes as a non-volatile memory, are shown under continuous light pulses. The insets show magnified views of certain regions. For the positive photoresponse mode, Figure 10It demonstrates 263 different positive photocurrent states, with a pulse width of 20 milliseconds and an optical power of 1.4 milliwatts; for the negative photoresponse mode, Figure 11 144 different negative photocurrent states were demonstrated, with a pulse width of 200 milliseconds and an optical power of 1.4 milliwatts. From Figure 5 and Figure 6 As can be seen from the example, the two-dimensional van der Waals heterojunction device prepared in Example 2 has excellent polymorphism, stability and durability when used as a non-volatile memory.

[0070] This embodiment provides two application approaches for the aforementioned two-dimensional van der Waals heterojunction device, including: (1) Simulated moving target detection task: Figure 12 This paper presents a flowchart illustrating the simulation of moving target detection using a two-dimensional van der Waals heterojunction device fabricated using an inter-frame difference recognition method. The specific steps are as follows: A 30-second video clip is captured at a fixed camera position. The video should contain moving objects (such as vehicles). The video is composed of individual frames. Two frames, at seconds 11 and 12, are extracted from the video using video editing software. These frames are 1920×1080 pixels in size. These two images are then converted to grayscale images using image processing software, resulting in two brightness matrices of 1920×1080 pixels. The values ​​of the elements in these matrices represent the grayscale values ​​(0~255) of the corresponding pixels. Figure 6 Under the test conditions, positive and negative photocurrents of the fabricated two-dimensional van der Waals heterojunction device were collected under a single light pulse. This collection was repeated 64 times, yielding 64 positive and 64 negative photocurrents of similar magnitude. The values ​​of the positive and negative photocurrents were then normalized. These normalized values ​​were then used to assemble positive and negative photocurrent matrices, with a size of 1920 × 1080. The values ​​of the elements in the matrices were random values ​​from the 64 collected positive and negative photocurrents. 1. The two brightness matrices are respectively multiplied by the positive and negative photocurrent matrices using the Hadamard product (a matrix operation applicable to two matrices of the same dimension. The result is a new matrix where each element is the product of corresponding elements in the original two matrices). The two Hadamard products are summed, and the absolute value of each element is calculated to obtain a new brightness matrix of size 1920×1080. This new brightness matrix is ​​then converted into a visual image, where the value of each element is the grayscale value of the corresponding pixel. The normalization, generation of the positive and negative photocurrent matrices, the Hadamard product operation, and the conversion of the brightness matrix into an image are all implemented using Python code. Figure 12 The rightmost detection result shows that the moving school bus was successfully identified.

[0071] (2) Simulated motion pattern recognition task: Figure 13 a and Figure 13 b demonstrates the convolutional neural network structure and final recognition accuracy used in simulating motion pattern recognition using the fabricated two-dimensional van der Waals heterojunction device. For six motion patterns—uniform motion, acceleration, deceleration, left turn, right turn, and U-turn—we constructed a training set of 5400 images and a test set of 5400 images each. Each 128×128 pixel image in both the training and test sets is composed of the vehicle's position information at four different times and varying levels of white noise. Figure 13 As shown, the convolutional neural network combines convolutional layers and fully connected layers: The first layer is a three-output-channel convolutional layer with a kernel size of 3×3 and a kernel count of 3. A batch normalization layer is then used to normalize the network parameters, and a linear rectified activation function is used to process the input data. A 2×2 max-pooling layer is then used to halve the output image. The second layer is a five-output-channel convolutional layer with a kernel size of 3×3 and a kernel count of 15. A batch normalization layer is then used to normalize the network parameters, and a linear rectified activation function is used to process the input data. A 2×2 max-pooling layer is then used to halve the output image. The final output feature map is [5, 30, 30]. This output feature map is then flattened into a one-dimensional vector to serve as the input to the fully connected layer. The classification process uses two fully connected layers. The first layer is 4500×100 in size and outputs a one-dimensional vector of length 100. Then, a linear rectified activation function is used to process the vector. The second layer is 100×6 in size and outputs a one-dimensional vector of length 6. Finally, the index of the number with the maximum value in this vector is the final classification result. Figure 13Figure b shows the recognition accuracy of the dataset under different noise levels. As can be seen from the figure, the recognition accuracy gradually increases with the increase of training epochs, ultimately achieving a rapid recognition accuracy exceeding 90% across all noise levels. Under conditions of a drain voltage of 0.5 volts, a gate voltage pulse of 0 volts, a pulse width of 20 milliseconds, and an optical power of 6.6 milliwatts, the relationship between the positive photocurrent and the number of optical pulses in the fabricated two-dimensional van der Waals heterojunction device was collected, and 128 states were used for curve fitting. The largest modulus was found in the trained convolutional neural network parameters and mapped to the largest conductance value among the 128 states. Other network parameters were scaled down proportionally, ensuring all network parameters were within the device's conductance range. The scaled-down network parameters were then used with the inverse function of the fitted curve to find the corresponding number of pulses to be applied. The pulse number was rounded to an integer, and the integer was substituted into the fitted curve to calculate the corresponding quantized conductance value. Finally, the quantized conductance value was scaled up proportionally. This completes the quantization mapping from the device's positive photocurrent to the network parameters. The mapped network parameters are then fed into a convolutional neural network to identify the test set data. Thanks to the excellent characteristics of the device (high number of states and good linearity), the recognition accuracy remains essentially unchanged (the recognition accuracy decreases by 0.1% to 0.5% under different noise levels). The above process is all implemented using Python code.

[0072] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A two-dimensional van der Waals heterojunction device, characterized in that, It includes, from bottom to top, a channel layer, a first insulating layer, a floating gate layer, a second insulating layer, and a gate, wherein: A source and a drain are provided between the channel layer and the first insulating layer, and the channel region between the source and the drain is completely covered by the channel layer, the first insulating layer, the floating gate layer and the second insulating layer in the vertical direction. The channel layer is molybdenum ditelluride; The first insulating layer is aluminum oxide; The floating grid layer is tungsten diselenide; The second insulating layer is hexagonal boron nitride; The molybdenum diselleride, tungsten diselenide, and hexagonal boron nitride are all two-dimensional layered materials; The thickness of the channel layer is 5-8 nanometers; The thickness of the first insulating layer is 20-60 nanometers; The thickness of the floating grid layer is 10-20 nanometers; The thickness of the second insulating layer is 15-25 nanometers; The width of the channel region is 3 to 12 micrometers.

2. The two-dimensional van der Waals heterojunction device according to claim 1, characterized in that: The thickness of the channel layer is 6-7 nanometers; The thickness of the first insulating layer is 30-35 nanometers; The thickness of the floating gate layer is 12-13 nanometers; The thickness of the second insulating layer is 15-16 nanometers; The width of the channel region is 5 to 10 micrometers.

3. The two-dimensional van der Waals heterojunction device according to claim 1, characterized in that: The source and drain electrodes are made of one or more of the following materials: gold, silver, copper, chromium, palladium, and platinum.

4. A method for preparing a two-dimensional van der Waals heterojunction device according to any one of claims 1 to 3, characterized in that, Includes the following steps: (a) Providing a molybdenum distellide thin film as a channel layer; (b) The channel layer is patterned using ultraviolet lithography and reactive ion etching processes to form the channel region; (c) The electrode region is exposed on the channel layer by ultraviolet photolithography, and the source and drain are fabricated in the electrode region by metal deposition method; (d) Alumina is deposited on the channel layer and the source and drain electrodes by atomic layer deposition process as a first insulating layer; (e) By means of a transfer method, a tungsten diselenide thin film as a floating gate layer and a hexagonal boron nitride thin film as a second insulating layer are sequentially placed on the first insulating layer; (f) The stacked heterostructure is patterned using ultraviolet lithography and reactive ion etching processes, so that the channel layer, the first insulating layer, the floating gate layer and the second insulating layer have a completely vertically overlapping area, and this area covers the channel area of ​​the channel layer; (g) A gate is fabricated on the second insulating layer by ultraviolet lithography and metal deposition methods to obtain a large-area array or a single device based on the two-dimensional van der Waals heterojunction device.

5. The method for fabricating a two-dimensional van der Waals heterojunction device according to claim 4, characterized in that: The preparation of source and drain electrodes in the electrode region using a metal deposition method includes: sequentially depositing a chromium layer and a gold layer, with the chromium layer having a thickness of 5-15 nanometers and the gold layer having a thickness of 30-80 nanometers.

6. The application of the two-dimensional van der Waals heterojunction device according to any one of claims 1 to 3 in moving target detection or motion pattern recognition.

7. A method for simulating the detection of moving objects, characterized in that: Includes the following steps: Acquire the two frames of image data to be detected and convert them into grayscale images, represented by a grayscale value matrix; Using the normalized photocurrent data obtained by the two-dimensional van der Waals heterojunction device according to any one of claims 1 to 3 in the positive photoresponse mode and the negative photoresponse mode, a positive photocurrent matrix and a negative photocurrent matrix are generated. The two grayscale matrices are respectively subjected to Hadamard product operation with the positive photocurrent matrix and the negative photocurrent matrix. The two Hadamard products are summed and converted into a grayscale image. Pixels with grayscale values ​​less than a preset threshold are set to zero to obtain the moving target detection result. When the two frames to be detected contain moving objects, the detection result displays the outline information of the objects; otherwise, the outline information of the objects is not displayed.

8. A method for simulating motion pattern recognition, characterized in that: Includes the following steps: Construct a convolutional neural network, train it on the input data, and obtain the optimal network parameters; The multi-level storage state of the two-dimensional van der Waals heterojunction device according to any one of claims 1 to 3 is mapped to the optimal network parameters, wherein the multi-level storage state is generated by carrier exchange between the floating gate layer and the gate under the action of gate voltage pulse or optical pulse; The optimal network parameters after mapping are input into the convolutional neural network to identify the motion patterns of moving targets.

Images