Method and system for simulating neural synaptic characteristic behavior by spin-orbit moment device

Abstract

The embodiment of the invention provides a method and a system for simulating a neural synaptic characteristic behavior by a spin-orbit moment device, and belongs to the field of neural computation. The spin-orbit torque device is of a multi-end Hall rod structure and comprises an input end and a plurality of output ends, and the method comprises the steps that continuously-changing pulse current is applied between counter electrodes of the input end of the spin-orbit torque device; under the condition of no external magnetic field, the continuously changing pulse current is used as an input signal of the simulated nerve synapse, and the Hall resistor is used as an output of the simulated nerve synapse; wherein the emission threshold value of the neural synapse is adjusted through the position of the output end and voltage regulation and control. Neural synapses with different thresholds are simulated by utilizing a spin-orbit torque device which comprises an input end and a plurality of output ends and is of a multi-end Hall rod structure and adjusting the distance between the spin-orbit torque device and the input end and the distance between the spin-orbit torque device and the output end; and voltage can be applied to the device, so that the neural synaptic threshold value can be dynamically adjusted. Through different modes and principles, various neural synapses with different thresholds are simulated.

Description

Technical Field

[0001] The present invention relates to the field of neural computing technology, and in particular to a method and system for simulating characteristic behaviors of neural synapses using a spin-orbit torque device. Background Art

[0002] In a neural network, neurons receive input signals from synapses and are the core of brain computing. Synaptic plasticity is a biological process that produces changes in synaptic weights through specific patterns of synaptic activity. The front neuron receives external stimuli and sends pulse signals to the back neuron through synapses, which changes the weight of the synapse and produces excitatory or inhibitory potentials. Plasticity is considered to be the neural basis of functions such as memory, learning, and computing. Therefore, simulating synaptic plasticity and building artificial neural networks are the key to realizing neuromorphic brain-like computers in the future.

[0003] Researchers use different devices to simulate synapses and realize neural computing. Resistive random access memory and phase change random access memory are widely used to simulate synapses. The resistance changes of these devices are similar to the changes in synaptic weights. Compared with the above devices, spin devices have the advantages of low power consumption and high-speed reading and writing, and have great potential for simulating synapses. However, spin devices, especially spin-orbit torque devices, still have many problems in simulating synapses. For example, the movement of magnetic domains in the magnetic layer is fast and difficult to control; spin-orbit torque devices require the assistance of an external magnetic field to achieve deterministic flipping. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a method and system for simulating characteristic behaviors of neural synapses using a spin-orbit torque device, so as to fully or at least partially solve the technical problems existing in the above-mentioned prior art.

[0005] In order to achieve the above-mentioned object, an embodiment of the present invention provides a method for simulating characteristic behavior of neural synapses using a spin-orbit torque device, wherein the spin-orbit torque device is a multi-terminal Hall bar structure, comprising an input terminal and multiple output terminals, and the method for simulating characteristic behavior of neural synapses using a spin-orbit torque device comprises:

[0006] Applying a continuously changing pulse current between the counter electrodes at the input end of the spin-orbit torque device;

[0007] In the absence of an external magnetic field, a continuously changing pulse current is used as an input signal for simulating a neural synapse, and a Hall resistor is used as an output for simulating a neural synapse; the firing threshold of the neural synapse is adjusted by regulating the position and voltage of the output end.

[0008] Optionally, the structure of the spin-orbit torque device includes:

[0009] From bottom to top, a substrate, an IrMn antiferromagnetic layer, a CoFeB magnetic layer, MgO and HZO insulating layers, and a metal layer are stacked in sequence; wherein a W layer is inserted between the IrMn antiferromagnetic layer and the CoFeB magnetic layer, and a wedge-shaped MgO layer is grown between the HZO insulating layer and the CoFeB magnetic layer using oblique incidence.

[0010] Optionally, the construction process of the spin-orbit torque device includes:

[0011] Select silicon oxide wafer as substrate, use acetone, ethanol, deionized water, ultrasonically clean the substrate in sequence, and finally use high-purity nitrogen to blow dry for later use;

[0012] The substrate is transferred onto the IrMn target, and the IrMn target is sputtered using a DC power supply;

[0013] Sputtering W target using DC power supply;

[0014] The CoFeB target is sputtered using a DC power supply. The CoFeB layer is perpendicularly anisotropic, and there is a coupling effect between the antiferromagnetic IrMn and CoFeB.

[0015] Using a radio frequency power supply, MgO target and HZO target are sputtered in sequence. The prepared MgO layer is wedge-shaped, which makes the vertical magnetic moment of the CoFeB layer deviate from the z-axis, generating a vertical anisotropy with an inclined angle, which is beneficial for the device to achieve deterministic reversal of the spin-orbit moment without external field assistance.

[0016] Using a DC power supply, Ta is grown and naturally oxidized to TaOx in the air to protect the sample;

[0017] The multi-terminal Hall device was fabricated by photolithography and ion etching techniques, and the top electrode Ti / Au was fabricated by electron beam evaporation technology;

[0018] The prepared spin-orbit torque device is annealed according to preset conditions to form an exchange bias field between the antiferromagnetic IrMn and CoFeB.

[0019] Optionally, in the absence of an external magnetic field, a continuously changing pulse current is used as an input signal for simulating a neural synapse, and a Hall resistance is used as an output of the simulating neural synapse, including:

[0020] A pulse current is applied between the electrodes at the input end of the spin-orbit torque device to simulate the input of a neural synapse. The input pulse current flows through the IrMn antiferromagnetic layer. Due to the action of the spin-orbit moment, a spin current is generated, which enters the CoFeB layer and exchanges with its magnetic moment.

[0021] If the pulse current is small, the magnetic moment of the CoFeB layer will not flip, and the Hall resistance signal detected by the output electrode will not change, indicating that the synapse is leaking.

[0022] If the pulse current is large and exceeds the preset threshold, the magnetic moment of the CoFeB layer flips, and the Hall resistance signal detected by the output electrode changes, simulating the discharge behavior of the synapse after reaching the threshold.

[0023] Optionally, in the absence of an external magnetic field, a continuously changing pulse current is used as an input signal for simulating a neural synapse, and a Hall resistance is used as an output of the simulating neural synapse, including:

[0024] When a pulse current is applied between the electrodes at the input end of the spin-orbit torque device, the magnetic domain will move along the direction between the electrodes, and multiple output electrodes will receive the signal in sequence;

[0025] Near the critical flipping current, continuous positive pulse current will cause the Hall resistance to gradually increase, simulating the behavior of excitatory synapses, and continuous negative pulse current will cause the Hall resistance to gradually decrease, simulating the behavior of inhibitory synapses. Among them, due to the different movement path lengths of the magnetic domains, the synapses have different emission thresholds.

[0026] Optionally, the emission threshold adjustment includes:

[0027] When a voltage is applied between the top and bottom electrodes of the spin-orbit torque device, the oxygen ions in the MgO and HZO layers enter the CoFeB magnetic layer, affecting the magnetic moment of the CoFeB magnetic layer and lowering the potential barrier, reducing the critical switching current, and thus regulating the firing threshold of the synapse.

[0028] Optionally, the emission threshold adjustment includes:

[0029] By applying different voltages between the top and bottom electrodes of the spin-orbit torque device, different emission thresholds are obtained.

[0030] On the other hand, the present invention also provides a system for simulating characteristic behaviors of neural synapses using a spin-orbit torque device, wherein the spin-orbit torque device is a multi-terminal Hall bar structure, comprising an input terminal and multiple output terminals, and the system for simulating characteristic behaviors of neural synapses using a spin-orbit torque device comprises:

[0031] A current applying unit, used for applying a continuously changing pulse current between the electrodes at the input end of the spin-orbit torque device;

[0032] The simulation unit is used to use a continuously changing pulse current as an input signal of a simulated neural synapse and a Hall resistance as an output of the simulated neural synapse in the absence of an external magnetic field; wherein the firing threshold of the neural synapse is adjusted by controlling the position and voltage of the output terminal.

[0033] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method of using a spin-orbit torque device to simulate characteristic behaviors of neural synapses when executing the program.

[0034] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for simulating the characteristic behavior of neural synapses using a spin-orbit torque device as described above.

[0035] Through the above technical solution, a spin-orbit torque device with a multi-terminal Hall bar structure including one input terminal and multiple output terminals is used to simulate synapses with different thresholds by adjusting the distance from the input terminal and the output terminal; and voltage can be applied to the device to achieve dynamic adjustment of the synaptic threshold. Synapses with multiple different thresholds can be simulated through different methods and different principles.

[0036] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0038] Figure 1 It is a flow chart of an implementation of a method for simulating characteristic behavior of neural synapses using a spin-orbit torque device provided by an embodiment of the present invention;

[0039] Figure 2 is a schematic structural diagram of a spin-orbit torque device provided by an embodiment of the present invention;

[0040] Figure 3 is a schematic structural diagram of a multi-terminal Hall device provided by an embodiment of the present invention;

[0041] Figure 4 It is a schematic diagram of a method for simulating characteristic behaviors of neural synapses using a spin-orbit torque device provided by an embodiment of the present invention;

[0042] Figure 5 is a schematic diagram of the relationship between current and Hall resistance provided by an embodiment of the present invention;

[0043] Figure 6 It is a structural schematic diagram of a system for simulating characteristic behaviors of neural synapses using a spin-orbit torque device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0045] See also Figure 1 FIG. 1 is a flowchart of a method for simulating characteristic behavior of a neural synapse using a spin-orbit torque device provided by an embodiment of the present invention. The spin-orbit torque device is a multi-terminal Hall bar structure, including an input terminal and multiple output terminals, including:

[0046] Step 100: Apply a continuously changing pulse current between the counter electrodes at the input end of the spin-orbit torque device.

[0047] For details, see Figure 2 As shown, the structure of the spin-orbit torque device includes: a substrate, an IrMn antiferromagnetic layer, a CoFeB magnetic layer, MgO and HZO insulating layers, and a metal layer are stacked in sequence from bottom to top; wherein a W layer is inserted between the IrMn antiferromagnetic layer and the CoFeB magnetic layer, and a wedge-shaped MgO layer is grown between the HZO insulating layer and the CoFeB magnetic layer using oblique incidence.

[0048] In some embodiments, the construction process of the spin-orbit torque device includes:

[0049] S1: Select a silicon oxide wafer as the substrate, use acetone, ethanol, deionized water, ultrasonically clean the substrate in sequence, and finally use high-purity nitrogen to blow dry for use.

[0050] Specifically, a silicon oxide wafer with a surface thickness of 500 nm was selected as the substrate, and the substrate was cleaned ultrasonically using acetone, ethanol, and deionized water in sequence, with each cleaning time being 5 minutes, and finally dried with high-purity nitrogen for later use. The film was grown using magnetron sputtering technology. Before the film was sputtered, the sample was transferred into the sputtering chamber, away from the Ta target material to prevent Ta from sputtering onto the sample; the Ta target material was sputtered using a DC power supply, and the vacuum degree was 10 -8 T, open the argon gas inlet valve to make the sputtering pressure during operation 2mT, and sputter the Ta target for 20 minutes. The purpose is to absorb water and oxygen in the chamber, further reduce the vacuum degree, and improve the quality of the sputtered film.

[0051] S2: The substrate is transferred onto the IrMn target, and the IrMn target is sputtered using a DC power supply.

[0052] Specifically, the sample was transferred to the top of the IrMn target and the IrMn target was sputtered using a DC power supply. The local vacuum degree was 10 -8 T, the sputtering pressure is 2mT, and the IrMn thickness is 8nm.

[0053] S3: Sputtering the W target using a DC power supply.

[0054] Specifically, a W target is sputtered using a DC power supply to grow a W film with a thickness of 0.2 nm. W atoms have strong thermal stability and can prevent atomic diffusion between IrMn and CoFeB. The thinner W layer has less effect on the exchange bias field between IrMn and CoFeB.

[0055] S4: CoFeB target is sputtered using a DC power supply. The CoFeB layer is perpendicularly anisotropic and has a thickness of 1 nm. There is a coupling effect between the antiferromagnetic IrMn and CoFeB.

[0056] Specifically, a DC power supply is used to sputter a CoFeB target, the film thickness is 1nm, and the CoFeB layer is perpendicular anisotropic. There is a coupling effect between the antiferromagnetic IrMn and CoFeB, which will generate an exchange bias field, which can replace the external magnetic field and achieve deterministic reversal of the spin-orbit moment without external field assistance.

[0057] S5: Using a radio frequency power supply, sputter MgO target and HZO target in sequence, wherein the prepared MgO layer is wedge-shaped, so that the vertical magnetic moment of the CoFeB layer deviates from the z-axis, generating a vertical anisotropy with an inclined angle, which is beneficial for the device to achieve deterministic reversal of the spin-orbit moment without external field assistance.

[0058] Specifically, MgO target and HZO target are sputtered in sequence using a radio frequency power supply to grow MgO 1nm and HZO 10nm. During sputtering, the inlet flow rate of argon is adjusted so that the working pressure is 1mT. Optionally, an oblique-incidence sputtering method is used, and the center of the MgO target deviates from the center of the sample by 20°. The prepared MgO layer is wedge-shaped, which will cause the vertical magnetic moment of the CoFeB layer to deviate from the z-axis, generating a vertical anisotropy with an inclined angle, breaking the symmetry, and facilitating the device to achieve deterministic reversal of the spin-orbit moment without external field assistance, thereby further improving the reversal efficiency. HZO has a large dielectric constant, and a thinner thickness can provide a stronger voltage regulation capability.

[0059] S6: Using a DC power supply, grow Ta and naturally oxidize it into TaOx in the air to protect the sample.

[0060] Preferably, the growth thickness of Ta is 2 nm.

[0061] S7: Prepare the multi-terminal Hall device using photolithography and ion etching techniques, and prepare the top electrode Ti / Au using electron beam evaporation technology.

[0062] Preferably, the top electrode Ti / Au 20 / 80nm is prepared by photolithography and ion etching technology, and the prepared multi-terminal Hall device structure is as follows Figure 3 shown.

[0063] S8: Annealing the prepared spin-orbit torque device according to preset conditions to form an exchange bias field between the antiferromagnetic IrMn and CoFeB.

[0064] Specifically, the device is placed in a magnetic field of 1 T and a temperature of 260 degrees Celsius for annealing for 30 minutes. The direction of the magnetic field is parallel to the in-plane and along the x-direction, which is conducive to the formation of an exchange bias field between the antiferromagnetic IrMn and CoFeB.

[0065] Step 101: In the absence of an external magnetic field, a continuously changing pulse current is used as an input signal of a simulated neural synapse, and a Hall resistance is used as an output of the simulated neural synapse; wherein the firing threshold of the neural synapse is adjusted by controlling the position and voltage of the output terminal.

[0066] Specifically, when executing step 101, the following steps may be specifically performed:

[0067] S1010: A pulse current is applied between the electrodes at the input end of the spin-orbit torque device to simulate the input of a neural synapse. The input pulse current flows through the IrMn antiferromagnetic layer. Due to the action of the spin-orbit moment, a spin current is generated, which enters the CoFeB layer and exchanges with its magnetic moment.

[0068] At the input end of the spin-orbit torque device, a pulse current is applied to simulate the input signal of the synapse. This pulse current is short-lived and high-intensity, and is used to trigger the physical process inside the device. When the pulse current flows through the IrMn antiferromagnetic layer, due to the special properties of IrMn (such as strong spin-orbit coupling), the electron spins in the current will be affected by the IrMn layer, resulting in unbalanced spin polarization. In the IrMn layer, due to the spin-orbit coupling effect, the electron spins in the current will experience a "spin Hall effect" or a similar mechanism, resulting in the generation of spin current (i.e., electron flow with a specific spin direction). This spin current carries angular momentum and is the key to the subsequent interaction with the CoFeB layer. The spin current then enters the CoFeB layer, which is a ferromagnetic material layer whose magnetic moment (i.e., magnetization direction) can be changed by external influences. In the CoFeB layer, the spin current exchanges with the magnetic moment of CoFeB. This exchange interaction is achieved through the mechanism of spin-orbit torque (SOT), which causes the magnetic moment of the CoFeB layer to rotate or flip. This change in magnetic moment can be regarded as the output signal of the SOT device, similar to the weight update or signal transmission in a neural synapse.

[0069] Compared with traditional electronic devices, SOT devices use the spintronics effect for information processing, with higher energy efficiency and lower power consumption. This is crucial for building efficient and energy-saving neuromorphic computing systems. Since the change in the magnetic moment of the CoFeB layer is stable (i.e., non-volatile), SOT devices can achieve long-term storage of information. This gives SOT devices an advantage in building neural synapse simulation devices with memory functions. The generation of spin currents and the exchange of magnetic moments can be completed in nanoseconds or even shorter, enabling SOT devices to achieve high-speed information processing. By adjusting parameters such as the amplitude, width, and frequency of the pulse current, the change in the magnetic moment of the CoFeB layer can be precisely controlled, thereby simulating the weight update and signal transmission process in the neural synapse. This makes SOT devices have great potential in building artificial neural networks and neuromorphic computing systems.

[0070] S1011: If the pulse current is small, the magnetic moment of the CoFeB layer will not be reversed, and the Hall resistance signal detected by the output electrode will not change, indicating that the synapse is leaking.

[0071] If the pulse current is too small and the magnetic moment does not flip, the stability of the output signal will be maintained. In the weight update process of the simulated neural synapse, the magnitude of the pulse current is usually proportional to the amplitude of the weight change. Therefore, by controlling the magnitude of the pulse current, fine control of the weight change can be achieved. When the pulse current is small, although it is not enough to cause the magnetic moment to flip, this control mechanism is still effective, allowing the system to adjust the weight with a smaller step size, thereby achieving more sophisticated computing and learning capabilities. Since the pulse current is small, the energy consumption of the entire process will also be reduced accordingly. This is of great significance for building low-power neuromorphic computing systems, especially in application scenarios that require long-term operation or large-scale deployment. In some cases, a pulse current that is too small may indicate that the device is faulty or has degraded performance. By monitoring the changes in the output signal, these problems can be discovered and handled in a timely manner, thereby enhancing the reliability and fault tolerance of the system. The synaptic connections in the biological nervous system are highly diverse and complex, including different connection strengths, transmission delays, and plasticity. In the process of simulating the biological nervous system, by controlling the magnitude of the pulse current to simulate synaptic connections of different strengths, the characteristics of the biological nervous system can be more realistically reflected, thereby improving the accuracy and reliability of the simulation.

[0072] S1012: If the pulse current is large and exceeds the preset threshold, the magnetic moment of the CoFeB layer is reversed, and the Hall resistance signal detected by the output electrode changes, simulating the discharge behavior of the synapse after reaching the threshold.

[0073] Synapses are key structures for transmitting information between neurons, and their discharge behavior depends on reaching a certain threshold. By controlling the reversal of the magnetic moment of the CoFeB layer through pulse current, this threshold trigger mechanism can be simulated to achieve information processing and storage functions similar to those of synapses. The reversal state of the magnetic moment is non-volatile, that is, the reversal state can remain stable without the continuous action of external energy, which is conducive to simulating long-term memory and persistent storage of information. By presetting the threshold of the pulse current, the triggering conditions for the magnetic moment reversal can be precisely controlled, thereby achieving precise control of information processing. The magnetic moment reversal process is relatively rapid, and information processing and transmission can be completed in a short time, simulating the rapid discharge behavior of synapses. By adjusting the threshold of the pulse current, programmable control of the simulated synaptic behavior can be achieved to meet different application requirements.

[0074] In some implementations, when executing step 101, the following steps may also be performed:

[0075] S1010: A pulse current is applied between the electrodes at the input end of the spin-orbit torque device, the magnetic domain will move along the direction between the electrodes, and multiple output electrodes will receive signals in turn.

[0076] As a carrier of information, the magnetic domain moves in a specific direction inside the device, and can effectively transmit the signal from the input end to multiple output ends. This transmission method is not only fast but also stable, which helps to achieve long-distance, low-loss signal transmission. By adjusting the amplitude, width, frequency of the pulse current, and the location and time of the applied current, the movement trajectory and speed of the magnetic domain can be precisely controlled, thereby achieving programmable control of the signal transmission path and output mode. This flexibility enables SOT devices to adapt to different application requirements and realize complex signal processing and computing tasks. Since multi-channel parallel processing and signal transmission can be achieved inside the SOT device, more functional units can be integrated in a smaller physical space. This helps to achieve miniaturization and high integration of devices, reducing manufacturing costs and power consumption.

[0077] S1011: Near the critical flipping current, continuous positive pulse current will cause the Hall resistance to gradually increase, simulating the behavior of excitatory synapses, and continuous negative pulse current will cause the Hall resistance to gradually decrease, simulating the behavior of inhibitory synapses. Among them, due to the different movement path lengths of the magnetic domains, the synapses have different emission thresholds.

[0078] When a positive pulse current is applied continuously, if the current intensity is close to but not exceeding the critical flipping current, the magnetic domain may be slightly disturbed but not enough to flip completely. At this time, the Hall resistance will gradually increase, simulating the excitatory response in biological neural synapses. As the current intensity increases, the magnetic domain may eventually flip, resulting in a significant change in the Hall resistance. On the contrary, when a negative pulse current is applied continuously, the magnetic domain is acted on in the opposite direction, which may cause it to gradually move away from the initial state, and the Hall resistance gradually decreases, simulating the inhibitory response in biological neural synapses. Due to the different lengths of the movement paths of the magnetic domains in the device, the response sensitivity of the magnetic domains on different paths to the pulse current will also be different. This results in different emission thresholds for the synapse, that is, the magnetic domains on different paths require pulse currents of different intensities to trigger a significant change in the Hall resistance. This dynamic threshold characteristic increases the flexibility and diversity of SOT devices in simulating complex neural networks. By precisely controlling the intensity and duration of the pulse current, the movement of the magnetic domains can be finely controlled, thereby precisely controlling the change in the Hall resistance. This high-precision signal control capability enables SOT devices to more accurately reflect the dynamic characteristics of biological nervous systems when simulating neural synaptic behavior. Compared with traditional electronic devices, SOT devices use spintronics effects for information processing, with lower power consumption and higher energy efficiency. This low power consumption feature is particularly important when simulating synaptic behavior, which helps to build large-scale, low-power neuromorphic computing systems.

[0079] In some embodiments, the adjustment of the emission threshold includes: applying a voltage between the top electrode and the bottom electrode of the spin-orbit torque device, the oxygen ions in the MgO and HZO layers enter the CoFeB magnetic layer, affecting the magnetic moment of the CoFeB magnetic layer, and lowering the potential barrier, reducing the critical switching current, thereby adjusting the emission threshold of the synapse.

[0080] In some embodiments, adjusting the emission threshold further comprises: obtaining different emission thresholds by applying different voltages between the top electrode and the bottom electrode of the spin-orbit torque device.

[0081] In some embodiments, see Figure 4 As shown, it is a schematic diagram of a method for simulating characteristic behaviors of neural synapses using a spin-orbit torque device provided by an embodiment of the present invention. Electrode 1 and electrode 2 are input ends of the simulated neural synapses, which are used to receive pulse signals; electrodes 3 and 4, electrodes 5 and 6, and electrodes 7 and 8 are output ends of the neural synapses. A pulse current is applied between electrodes 1 and 2 to simulate the input of the neural synapse. The input pulse current flows through the IrMn antiferromagnetic layer, and due to the action of the spin-orbit moment, a spin current is generated, which enters the CoFeB layer and exchanges with its magnetic moment. As shown in FIG. Figure 4As shown in FIG. 1 , if the pulse current is small, the magnetic moment will not be reversed, and the Hall resistance signal detected by electrodes 3 and 4 (electrodes 5 and 6, electrodes 7 and 8) will not change, which is equivalent to the leakage behavior of the synapse. If the pulse current is large enough and exceeds a certain threshold, the magnetic moment of the CoFeB layer will be reversed, and the resistance detected by electrodes 3 and 4 will change, simulating the discharge behavior of the synapse after reaching the threshold, that is, the triggering behavior. The relationship between current and Hall resistance can be seen in FIG. Figure 5 As shown. When a pulse current is applied between electrodes 1 and 2, the magnetic domain will move along the direction between electrodes 1 and 2, and electrodes 3 and 4, electrodes 5 and 6, and electrodes 7 and 8 will receive signals in turn. Due to the different lengths of the movement paths of the magnetic domains, the synapses have different emission thresholds. It is further explained that near the critical flip current, continuous positive pulse currents will gradually increase the Hall resistance, which can simulate excitatory synaptic behavior, and negative pulse currents will gradually reduce the Hall resistance, which is inhibitory synaptic behavior. Due to the different positions of electrodes 3 and 4, electrodes 5 and 6, and electrodes 7 and 8, neurons with different emission thresholds caused by receiving signals, however, the distance is fixed during the device manufacturing process, and the adjustment of the emission threshold is subject to certain restrictions. When a voltage is applied between electrodes 11 / 12 / 13 and the bottom electrode, the oxygen ions in MgO and HZO enter CoFeB, affecting the magnetic moment of CoFeB, and reducing the potential barrier and the critical flip current, thereby adjusting the emission threshold of the synapse. Different emission thresholds can be obtained by applying different voltages; in addition, the decrease in critical flip current can also further reduce the power consumption of the simulated neural synaptic device.

[0082] The beneficial effects brought by the technical solution of this application are as follows:

[0083] (1) A spin-orbit torque device simulating a neural synapse is provided. There is a coupling effect between IrMn / CoFeB to form an exchange bias field, which solves the problem that the spin-orbit torque device needs external magnetic field assistance to obtain a definite flip.

[0084] (2) The MgO layer is grown at oblique incidence to form a wedge-shaped film, which is combined with the CoFeB magnetic layer to cause the vertical magnetic moment of CoFeB to shift, breaking the symmetry and facilitating the deterministic flipping of the spin-orbit torque device.

[0085] (3) The IrMn and CoFeB layers form pinning, which reduces the transmission speed of the magnetic domains and is more conducive to regulation.

[0086] (4) The firing thresholds of different synapses can be simulated according to the distances of different devices, and appropriate synapses can be selected as needed.

[0087] (5) The synaptic threshold can be adjusted through voltage control; voltage regulation can further reduce the flipping current and reduce power consumption.

[0088] See also Figure 6 FIG. 1 is a schematic diagram of a structure of a system for simulating characteristic behaviors of neural synapses using a spin-orbit torque device according to an embodiment of the present invention. The spin-orbit torque device is a multi-terminal Hall bar structure, including an input terminal and multiple output terminals. The system for simulating characteristic behaviors of neural synapses using a spin-orbit torque device includes:

[0089] A current applying unit 60, used for applying a continuously changing pulse current between the electrodes at the input end of the spin-orbit torque device;

[0090] The simulation unit 61 is used to use the continuously changing pulse current as the input signal of the simulated neural synapse and the Hall resistance as the output of the simulated neural synapse under the condition of no external magnetic field; wherein the firing threshold of the neural synapse is adjusted by controlling the position and voltage of the output terminal.

[0091] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps of the method for simulating characteristic behavior of neural synapses using a spin-orbit torque device as described in any of the above embodiments.

[0092] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for simulating characteristic behaviors of neural synapses using a spin-orbit torque device as described in any of the above embodiments.

[0093] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0094] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0095] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0096] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0097] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0098] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0099] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0100] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0101] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for simulating characteristic behavior of neural synapses using a spin-orbit torque device, characterized in that: The spin-orbit torque device is a multi-terminal Hall bar structure, including an input terminal and multiple output terminals. The method for simulating the characteristic behavior of a neural synapse by the spin-orbit torque device includes: Applying a continuously changing pulse current between the counter electrodes at the input end of the spin-orbit torque device; In the absence of an external magnetic field, a continuously changing pulse current is used as an input signal for simulating a neural synapse, and a Hall resistor is used as an output for simulating a neural synapse; the firing threshold of the neural synapse is adjusted by regulating the position and voltage of the output end.

2. The method for simulating characteristic behavior of neural synapses using a spin-orbit torque device according to claim 1, characterized in that: The structure of the spin-orbit torque device comprises: From bottom to top, a substrate, an IrMn antiferromagnetic layer, a CoFeB magnetic layer, MgO and HZO insulating layers, and a metal layer are stacked in sequence; wherein a W layer is inserted between the IrMn antiferromagnetic layer and the CoFeB magnetic layer, and a wedge-shaped MgO layer is grown between the HZO insulating layer and the CoFeB magnetic layer using oblique incidence.

3. The method for simulating characteristic behavior of neural synapses using a spin-orbit torque device according to claim 1, characterized in that: The construction process of the spin-orbit torque device includes: Select silicon oxide wafer as substrate, use acetone, ethanol, deionized water, ultrasonically clean the substrate in sequence, and finally use high-purity nitrogen to blow dry for later use; The substrate is transferred onto the IrMn target, and the IrMn target is sputtered using a DC power supply; Sputtering W target using DC power supply; The CoFeB target is sputtered using a DC power supply. The CoFeB layer is perpendicularly anisotropic, and there is a coupling effect between the antiferromagnetic IrMn and CoFeB. Using a radio frequency power supply, MgO target and HZO target are sputtered in sequence. The prepared MgO layer is wedge-shaped, which makes the vertical magnetic moment of the CoFeB layer deviate from the z-axis, generating a vertical anisotropy with an inclined angle, which is beneficial for the device to achieve deterministic reversal of the spin-orbit moment without external field assistance. Using a DC power supply, Ta is grown and naturally oxidized to TaOx in the air to protect the sample; The multi-terminal Hall device was fabricated by photolithography and ion etching techniques, and the top electrode Ti / Au was fabricated by electron beam evaporation technology; The prepared spin-orbit torque device is annealed according to preset conditions to form an exchange bias field between the antiferromagnetic IrMn and CoFeB.

4. The method for simulating characteristic behavior of neural synapses using a spin-orbit torque device according to claim 2, characterized in that: In the absence of an external magnetic field, a continuously changing pulse current is used as an input signal for simulating a neural synapse, and a Hall resistor is used as an output of the simulating neural synapse, including: A pulse current is applied between the electrodes at the input end of the spin-orbit torque device to simulate the input of a neural synapse. The input pulse current flows through the IrMn antiferromagnetic layer. Due to the action of the spin-orbit moment, a spin current is generated, which enters the CoFeB layer and exchanges with its magnetic moment. If the pulse current is small, the magnetic moment of the CoFeB layer will not flip, and the Hall resistance signal detected by the output electrode will not change, indicating that the synapse is leaking. If the pulse current is large and exceeds the preset threshold, the magnetic moment of the CoFeB layer flips, and the Hall resistance signal detected by the output electrode changes, simulating the discharge behavior of the synapse after reaching the threshold.

5. The method for simulating characteristic behavior of neural synapses using a spin-orbit torque device according to claim 2, characterized in that: In the absence of an external magnetic field, a continuously changing pulse current is used as an input signal for simulating a neural synapse, and a Hall resistor is used as an output of the simulating neural synapse, including: When a pulse current is applied between the electrodes at the input end of the spin-orbit torque device, the magnetic domain will move along the direction between the electrodes, and multiple output electrodes will receive the signal in sequence; Near the critical flipping current, continuous positive pulse current will cause the Hall resistance to gradually increase, simulating the behavior of excitatory synapses, and continuous negative pulse current will cause the Hall resistance to gradually decrease, simulating the behavior of inhibitory synapses. Among them, due to the different movement path lengths of the magnetic domains, the synapses have different emission thresholds.

6. The method for simulating characteristic behavior of neural synapses using a spin-orbit torque device according to claim 5, characterized in that: The emission threshold adjustment includes: When a voltage is applied between the top and bottom electrodes of the spin-orbit torque device, the oxygen ions in the MgO and HZO layers enter the CoFeB magnetic layer, affecting the magnetic moment of the CoFeB magnetic layer and lowering the potential barrier, reducing the critical switching current, and thus regulating the firing threshold of the synapse.

7. The method for simulating characteristic behavior of neural synapses using a spin-orbit torque device according to claim 5, characterized in that: The emission threshold adjustment includes: By applying different voltages between the top and bottom electrodes of the spin-orbit torque device, different emission thresholds are obtained.

8. A system for simulating characteristic behaviors of neural synapses using a spin-orbit torque device, characterized in that: The spin-orbit torque device is a multi-terminal Hall bar structure, including an input terminal and multiple output terminals. The system for simulating the characteristic behavior of neural synapses by the spin-orbit torque device includes: A current applying unit, used for applying a continuously changing pulse current between the electrodes at the input end of the spin-orbit torque device; The simulation unit is used to use a continuously changing pulse current as an input signal of a simulated neural synapse and a Hall resistance as an output of the simulated neural synapse in the absence of an external magnetic field; wherein the firing threshold of the neural synapse is adjusted by controlling the position and voltage of the output terminal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for simulating characteristic behaviors of neural synapses using a spin-orbit torque device as described in any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for simulating characteristic behaviors of neural synapses using a spin-orbit torque device as described in any one of claims 1 to 7 are implemented.

Images