A multi-terminal planar neuromorphic memristor and its preparation method

Abstract

The present invention discloses a multi-terminal planar memristor and a method for preparing the same. The multi-terminal planar memristor comprises: a substrate; a functional layer, which is a two-dimensional material and formed on the substrate; and a plurality of metal electrodes, which are non-contacting and distributed in an emitting pattern on the surface of the functional layer. One electrode serves as a back-end electrode, and the remaining electrodes serve as front-end electrodes. A continuous electrical pulse sequence is used to input the front-end electrodes, and the conductivity state of the back-end electrodes is collected to achieve multivariate neuromorphic collaborative computing functions.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and in particular to a multi-terminal planar neuromorphic memristor and a preparation method thereof. Background Art

[0002] To achieve efficient intelligent computing, traditional transistor circuits are designed to mimic the neuronal functions of the human brain to implement neuromorphic computing. However, this approach, based on traditional CMOS circuit components, consumes significant hardware resources and power consumption. Therefore, it is necessary to develop new neuromorphic hardware units to achieve neuromorphic computing capabilities at the individual device level and overcome existing hardware bottlenecks.

[0003] Brain-inspired neuromorphic devices, electronic devices with synaptic plasticity, are currently being widely researched as the physical hardware for neuromorphic computing. However, existing neuromorphic devices are mostly based on two-terminal devices, which can only simulate the basic two-terminal synaptic function by adjusting the weight of a corresponding single synaptic backend through a single synaptic frontend, making it difficult to achieve true multi-element neuromorphic brain-inspired computing.

[0004] On the other hand, two-dimensional materials, with their advantages of atomic-scale thickness, planar structure, and excellent flexibility, play a key role in the field of miniaturized integrated circuits. Therefore, research on neuromorphic brain-inspired devices based on two-dimensional materials has important practical value for the development of future high-density integrated semiconductor devices. Summary of the Invention

[0005] The present invention provides a multi-terminal planar neuromorphic memristor device, which utilizes the similarity between the multi-terminal structure and the multi-element neural synapse and uses the multi-terminal electrical excitation input to simulate the collaborative function of the multi-element neural synapse, and is used to construct a multifunctional and efficient neural computing system.

[0006] The multi-terminal planar neuromorphic memristive device includes a substrate; a functional layer, which is a two-dimensional material and is formed on the substrate; a plurality of metal electrodes, which are non-contacting and distributed in an emitting manner on the surface of the functional layer, wherein one electrode serves as a rear-end electrode and the remaining electrodes serve as front-end electrodes. A continuous electrical pulse sequence is used to input the plurality of front-end electrodes, and the conductance state of the rear-end electrodes is collected to realize a multi-dimensional neuromorphic collaborative computing function.

[0007] In the multi-terminal planar memristor of the present invention, preferably, the two-dimensional material is BN, MoS2, WS2 or MoTe2.

[0008] In the multi-terminal planar memristor of the present invention, preferably, the metal electrode is Ag, Cu, Au or Ti.

[0009] In the multi-terminal planar memristor of the present invention, preferably, a forward voltage is applied to the multiple front-end electrodes, and the back-end electrode is grounded, so that the front-end electrode is partially oxidized to produce metal ions, and moves toward the back-end electrode under the stimulation of the forward voltage; when the mobile metal ions reach the back-end electrode, they are reduced to metal atoms at the back-end electrode and continuously accumulate toward the positive electrode. When the metal atoms are connected between the front-end electrode and the back-end electrode, multiple conductive channels are formed, so that the conductance collected by the back-end electrode shows a nonlinear enhanced change. This process simulates the activity of multiple neural synapses in biological systems, and multiple synaptic front ends regulate the same synaptic back end, so that the synaptic weights can be nonlinearly changed, completing multiple neuromorphic calculations.

[0010] In the multi-terminal planar memristor of the present invention, preferably, the number of the metal electrodes is 4 to 10.

[0011] The present invention also discloses a method for preparing a multi-terminal planar memristor, comprising the following steps: forming a two-dimensional material as a functional layer on a substrate; forming a plurality of metal electrodes on the surface of the functional layer, which are non-contacting and distributed in an emitting manner, wherein one electrode serves as a rear-end electrode and the remaining electrodes serve as front-end electrodes, inputting a continuous electrical pulse sequence into the plurality of front-end electrodes, collecting the electrical conductivity state of the rear-end electrodes, and realizing a multi-dimensional neuromorphic collaborative computing function.

[0012] In the method for preparing a multi-terminal planar memristor of the present invention, preferably, the two-dimensional material is BN, MoS2, WS2 or MoTe2.

[0013] In the method for preparing a multi-terminal planar memristor of the present invention, preferably, the metal electrode is Ag, Cu, Au or Ti.

[0014] In the preparation method of a multi-terminal planar memristor of the present invention, preferably, a forward voltage is applied to the multiple front-end electrodes, and the back-end electrode is grounded, so that the front-end electrode is partially oxidized to produce metal ions, and moves toward the back-end electrode under the stimulation of the forward voltage; when the mobile metal ions reach the back-end electrode, they are reduced to metal atoms at the back-end electrode and continuously accumulate toward the positive electrode. When the metal atoms are connected between the front-end electrode and the back-end electrode, multiple conductive channels are formed, so that the conductance collected by the back-end electrode shows a nonlinear enhanced change; this process simulates the activity of multiple neural synapses in a biological system, and multiple synaptic front ends regulate the same synaptic back end, so that the synaptic weights can be nonlinearly changed, completing multiple neural morphological calculations.

[0015] In the method for preparing a multi-terminal planar memristor of the present invention, preferably, the number of the metal electrodes is 4 to 10.

[0016] Beneficial effects:

[0017] (1) By building a memristor based on two-dimensional materials, the device size can be miniaturized to the sub-nanometer level without leakage, which provides more possibilities for future high-density integrated miniaturized semiconductor devices.

[0018] (2) A planar memristor was designed to break the vertical working mode of the traditional memristor, facilitate the study of the device mechanism, and provide a more intuitive microscopic explanation of the macroscopic current changes of the memristor.

[0019] (3) Multi-terminal neuromorphic memristors break the structural limitations of traditional two-terminal memristors and provide a multi-terminal input conductivity modulation mode. Multi-terminal memristors can achieve higher efficiency in conductivity regulation by stimulating different input terminals. They have a natural similarity to multi-element neural synaptic systems and lay the foundation for the development of highly collaborative neuromorphic computing systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of the method for preparing a multi-terminal planar neuromorphic memristor.

[0021] Figure 2 It is a schematic diagram of the device structure after the functional layer is formed.

[0022] Figure 3 It is a schematic diagram of the device structure after the photoresist is formed.

[0023] Figure 4 It is a schematic diagram of the device structure after the multi-terminal electrode is formed.

[0024] Figure 5 This is a schematic diagram of the working principle of a multi-terminal planar neuromorphic memristor.

[0025] Figure 6 It is a schematic diagram of multiple conductive channels formed between multiple front-end electrodes and back-end electrodes. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. 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. The described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0027] In the description of the present invention, it should be noted that the terms "upper," "lower," "vertical," and "horizontal," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0028] In addition, many specific details of the present invention are described below, such as device structure, materials, dimensions, processing techniques, and technologies, to facilitate a clearer understanding of the present invention. However, as will be appreciated by those skilled in the art, the present invention may be practiced without following these specific details. Unless otherwise noted below, various components of the device may be constructed from materials known to those skilled in the art, or materials with similar functions developed in the future may be used.

[0029] Figure 1 This is a flow chart of the preparation method of multi-terminal planar neuromorphic memristor. Figure 1 As shown, the following steps are included:

[0030] Step S1: Prepare a 1.5×1.5 cm silicon oxide marker sheet as a substrate for preparing a multi-terminal planar memristor device. The structure is silicon oxide layer 101 / silicon wafer 100, wherein the thickness of the upper silicon oxide layer 101 is preferably 100 nm to 300 nm. Then, a 10 nm thick BN two-dimensional material is formed on the marker sheet substrate by mechanical exfoliation as a functional layer 102. The resulting structure is as shown in FIG. Figure 2 The functional layer material may also be MoS2, WS2, MoTe2, etc.; and the thickness is preferably 0.5nm to 20nm.

[0031] Step S2, spin-coating PMMA photoresist (AR-P 679) on the substrate in two steps. The first step is pre-spinning at a speed of 400 rpm-600 rpm, and the pre-spinning time is 5s-15s. The second step is spin-coating at a speed of 3000 rpm-5000 rpm, and the spin-coating time is 40s-90s. Then, the film is baked at 120℃-190℃ for 2 minutes-5 minutes to form a film 200. The resulting structure is as shown in FIG. Figure 3 As shown. Then, electron beam lithography is used to obtain a specific electrode mask on the functional layer 102. After that, 50nm thick Ag is grown by magnetron sputtering, and the useless metal parts are removed by acetone to form 6 electrodes. The 6 electrodes are not in contact with each other and are distributed on the surface of the functional layer in an emitting manner. Among them, one electrode is used as the rear electrode, and the remaining 5 electrodes are used as the front electrodes, namely front 1, front 2, front 3, front 4 and front 5. The resulting structure is shown in FIG. Figure 4As shown. The electrode material is preferably Ag, but can also be Cu, Au, Ti, etc.; the thickness is preferably 50nm, and can range from 30nm to 100nm. In this embodiment, the number of electrodes is preferably 6, but the present invention is not limited to this, and can be, for example, 4 to 10.

[0032] Step S3, as Figure 5 As shown, a continuous electrical pulse sequence is used to input the five front-end electrodes of the memristor, and the conductance state of the back-end electrodes is collected to realize the multi-dimensional neuromorphic collaborative computing function.

[0033] By applying positive voltage to five different front-end electrodes and grounding the back-end electrodes with the help of the electrochemical migration effect of silver ions, the front-end Ag electrode can be partially oxidized to Ag. + , and moves toward the rear electrode under the stimulation of the forward voltage. + When it reaches the back-end electrode, it will be reduced to Ag atoms at the back-end electrode and continue to accumulate toward the positive electrode. When the Ag atoms are connected between the front-end electrode and the back-end electrode, five conductive channels will be formed, such as Figure 6 As shown, the conductance collected by the back-end electrodes exhibits nonlinear enhancement. This process is similar to the activity of multiple synapses in biological systems, where multiple front-end synapses modulate the same back-end synapse, allowing nonlinear changes in synaptic weights and completing multi-element neuromorphic computing. Multi-terminal neuromorphic memristors break through the structural limitations of traditional two-terminal memristors and provide a multi-terminal input conductance modulation mode. Multi-terminal memristors can achieve higher efficiency in conductance regulation by leveraging excitation from different input terminals, sharing a natural similarity with multi-element synaptic systems and laying the foundation for the development of highly collaborative neuromorphic computing systems.

[0034] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with this technical field within the technical scope disclosed in the present invention should be covered by the scope of protection of the present invention.

Claims

1. A multi-terminal planar memristor, characterized in that: include: substrate; a functional layer, which is a two-dimensional material, formed on the substrate; Multiple metal electrodes, without contact with each other, are distributed on the surface of the functional layer in an emitting manner, one of the electrodes serves as the rear electrode, and the remaining electrodes serve as the front electrodes. By leveraging the electrochemical migration effect of metal ions in metal electrodes, a continuous electrical pulse sequence is used to input the plurality of front-end electrodes, and the conductivity state of the back-end electrodes is collected to realize the multi-dimensional neuromorphic collaborative computing function. A forward voltage is applied to the plurality of front electrodes, and the rear electrodes are grounded, so that the front electrodes are partially oxidized to generate metal ions, which then move toward the rear electrodes under the stimulation of the forward voltage. When the mobile metal ions reach the back-end electrode, they are reduced to metal atoms at the back-end electrode and continue to accumulate toward the positive electrode. When the metal atoms connect between the front-end electrode and the back-end electrode, multiple conductive channels are formed, causing the conductivity collected by the back-end electrode to show a nonlinear enhanced change. This process simulates the activity of multiple neural synapses in biological systems, with multiple synaptic front ends regulating the same synaptic back end, allowing synaptic weights to change nonlinearly and completing multiple neuromorphic calculations.

2. The multi-terminal planar memristor according to claim 1, wherein: The two-dimensional material is BN, MoS2, WS2 or MoTe2.

3. The multi-terminal planar memristor according to claim 1, wherein: The metal electrode is Ag, Cu, Au or Ti.

4. The multi-terminal planar memristor according to claim 1, wherein: The number of the metal electrodes is 4 to 10.

5. A method for preparing a multi-terminal planar memristor, characterized in that: The following steps are involved: forming a two-dimensional material as a functional layer on a substrate; A plurality of metal electrodes are formed on the surface of the functional layer, which are not in contact with each other and are distributed in an emitting manner, wherein one electrode serves as a rear electrode and the remaining electrodes serve as front electrodes. By leveraging the electrochemical migration effect of metal ions in metal electrodes, a continuous electrical pulse sequence is used to input the plurality of front-end electrodes, and the conductivity state of the back-end electrodes is collected to realize the multi-dimensional neuromorphic collaborative computing function. A forward voltage is applied to the plurality of front electrodes, and the rear electrodes are grounded, so that the front electrodes are partially oxidized to generate metal ions, which then move toward the rear electrodes under the stimulation of the forward voltage. When the mobile metal ions reach the back-end electrode, they are reduced to metal atoms at the back-end electrode and continue to accumulate toward the positive electrode. When the metal atoms are connected between the front-end electrode and the back-end electrode, multiple conductive channels are formed, causing the conductivity collected by the back-end electrode to show a nonlinear enhanced change. This process simulates the activity of multiple neural synapses in biological systems, with multiple synaptic front ends regulating the same synaptic back end, allowing synaptic weights to change nonlinearly and completing multiple neuromorphic calculations.

6. The method for preparing a multi-terminal planar memristor according to claim 5, wherein: The two-dimensional material is BN, MoS2, WS2 or MoTe2.

7. The method for preparing a multi-terminal planar memristor according to claim 5, wherein: The metal electrode is Ag, Cu, Au or Ti.

8. The method for preparing a multi-terminal planar memristor according to claim 5, wherein: The number of the metal electrodes is 4 to 10.

Images