Artificial synapse and method for using piezo-phototronic element

Abstract

Provided is a novel artificial synapse using a piezo-phototronic element. The disclosed artificial synapse includes an artificial synapse element that operates as memory for storing a received optical stimulus and reads out the storage of the optical stimulus after receiving the optical stimulus. The artificial synapse element is constituted of a photovoltaic device which is a piezo-phototronic element in which a change in photovoltaic power generation performance occurs when the optical stimulus is received, and the change in photovoltaic performance remains after reception of the optical stimulus so as to retain the changed photovoltaic power generation performance as storage of the optical stimulus. The reading out of the storage of the optical stimulus from the artificial synapse element is performed by outputting, from the piezo-phototronic element after receiving the optical stimulus, an electric signal corresponding to the photovoltaic generation performance, which is the storage of the optical stimulus.

Description

Artificial synapses and methods of using piezophototronic devices

[0001] This disclosure relates to an artificial synapse and a method of using a piezo-phototronic element. This application claims priority to Japanese Patent Application No. 2024-070075, filed April 23, 2024, and incorporates by reference the entire contents of said Japanese application.

[0002] US Pat. No. 6,299,499 discloses an artificial neural network circuit comprising a crossbar circuit having memristors.

[0003] Patent Document 2 discloses a heterojunction type photoelectric conversion element having a piezoelectric material.

[0004] JP 2020-57278 A JP 2023-154738 A

[0005] An artificial synapse is a device that artificially mimics a biological synapse. Artificial synapses are attracting attention as a core device for neuromimetic computing.

[0006] To the best of the inventors' knowledge, there have been no examples of piezo-phototronic elements, such as piezo-phototronic diodes that change photovoltage in response to input of stress (strain), being used as artificial synapses. The inventors have discovered that piezo-phototronic elements can be suitably used as light-responsive artificial synapses, and that piezo-phototronic elements can also provide new functions to artificial synapses.

[0007] Therefore, the present disclosure provides a novel artificial synapse that utilizes a piezo-phototronic element.

[0008] One aspect of the present disclosure is an artificial synapse, which includes a piezo-phototronic element as an artificial synapse element.

[0009] Another aspect of the present disclosure is a method of using a piezo-phototronic element. The disclosed method comprises using a piezo-phototronic element as an artificial synapse.

[0010] Further details will be described in the following embodiments.

[0011] FIG. 1 is an explanatory diagram of the light soaking effect of a piezo-phototronic element. FIG. 2 is an explanatory diagram of the residual light soaking effect of a piezo-phototronic element. FIG. 3 is an explanatory diagram of the difference in output signal with and without light soaking. FIG. 4 is an explanatory diagram of memory retention and forgetting under light soaking. FIG. 5 is a diagram showing an example of an artificial synapse device using a piezo-phototronic element. FIG. 6 is an explanatory diagram of writing (input) and reading (output) to an artificial synapse using a piezo-phototronic element. FIG. 7 is an explanatory diagram of writing (input) and reading (output) to an artificial synapse using a piezo-phototronic element. FIG. 8 is a diagram showing an example of discriminating numbers in an image using an artificial synapse. FIG. 9 is a diagram showing an example of memory retention in an input layer artificial synapse. FIG. 10 is a diagram showing an example of memory retention in an input layer artificial synapse. FIG. 11 is a diagram showing the results of an experiment showing changes in the output signal due to stress application to an artificial synapse. FIG. 12 is a diagram showing the output current from a self-powered artificial synapse. Fig. 13 is an explanatory diagram of erasing memory in an artificial synapse. Fig. 14 is a structural diagram of an artificial synapse. Fig. 15 is a diagram showing an example of fabricating an artificial synapse. Fig. 16 is a structural diagram of an artificial synapse equipped with multiple piezo-phototronic elements.

[0012] 1. Overview of the use of artificial synapses and piezo-phototronic devices

[0013] (A1) An artificial synapse according to an embodiment may include an artificial synapse element that operates as a memory for storing received optical stimuli and that reads out the stored optical stimuli after receiving the optical stimuli. The artificial synapse element may be configured with a photovoltaic device that is a piezo-phototronic element, whose photovoltaic performance changes upon receiving the optical stimuli and whose change in photovoltaic performance remains after receiving the optical stimuli, so that the changed photovoltaic performance is stored as a memory of the optical stimuli. The memory of the optical stimuli from the artificial synapse element may be read out by outputting an electrical signal from the piezo-phototronic element according to the photovoltaic performance, which is the memory of the optical stimuli, after receiving the optical stimuli.

[0014] The artificial synapse element according to the embodiment can operate as a nonvolatile memory with variable power generation performance. Here, a nonvolatile memory is a memory that does not require external power supply to retain its memory. The artificial synapse element according to the embodiment can change its photovoltaic performance based on the history of optical stimuli as input, and retain the changed photovoltaic performance as a memory of the optical stimuli. While the change in photovoltaic performance remains and the memory of the optical stimuli is retained, no external power supply is required.

[0015] (A2) The artificial synapse according to the embodiment may further include a stress application unit that applies stress to the piezo-phototronic element after receiving the optical stimulus.

[0016] (A3) In the artificial synapse according to the embodiment, the memory of the optical stimulus can be changed by the stress applied by the stress application unit.

[0017] (A4) In the artificial synapse according to the embodiment, the memory of the optical stimulus can be erased by the stress applied by the stress application unit.

[0018] (A5) In the artificial synapse according to the embodiment, after receiving the optical stimulus, stress is applied to the piezo-phototronic element by the stress application unit, and the memory of the optical stimulus can be read out after the stress application.

[0019] (A6) In the artificial synapse according to the embodiment, after receiving the optical stimulus, the memory of the optical stimulus is read out, and after the reading out, stress can be applied to the piezo-phototronic element by the stress application unit.

[0020] (A7) The artificial synapse according to the embodiment may be a self-powered type in which no power is supplied to the piezo-phototronic element to memorize the optical stimulus.

[0021] (A8) The memory of the optical stimulus can be read out by irradiating the piezo-phototronic element with light for reading after receiving the optical stimulus, and outputting an electrical signal generated by photovoltaic power generation corresponding to the photovoltaic performance, which is the memory of the optical stimulus, from the piezo-phototronic element.

[0022] (A9) The artificial synapse according to the embodiment may be a self-powered type in which power is not supplied to the piezo-phototronic element for reading out the memory of the optical stimulus.

[0023] (A10) The artificial synapse according to the embodiment may be a self-powered type in which power is not supplied to the piezo-phototronic element to store the optical stimulus, and power is not supplied to the piezo-phototronic element to read out the stored optical stimulus.

[0024] (A11) The piezo-phototronic element may be a selenium photovoltaic device that produces a piezo-phototronic effect. The selenium photovoltaic device is a photovoltaic device that includes Se (selenium) as a p-type semiconductor.

[0025] (A12) The piezo-phototronic element may be a solar cell that produces the piezo-phototronic effect.

[0026] (A13) A method according to an embodiment may include using a piezo-phototronic element in which a change in photovoltaic performance occurs upon receiving an optical stimulus, and in which the change in photovoltaic performance remains after receiving the optical stimulus, as an artificial synapse that operates as a memory for storing the received optical stimulus, so that the changed photovoltaic performance is stored as a memory of the optical stimulus, and after receiving the optical stimulus, an electrical signal corresponding to the photovoltaic performance, which is the memory of the optical stimulus, is output from the piezo-phototronic element.

[0027] (B1) An artificial synapse according to an embodiment includes a piezo-phototronic element as an artificial synapse element. The piezo-phototronic element generates a piezo-phototronic effect. Utilizing the piezo-phototronic effect makes it possible to adjust or reset the signal output from the artificial synapse. Furthermore, the piezo-phototronic element has a characteristic in which its short-circuit current (current generated upon light irradiation even without voltage application) changes depending on the light irradiation history and maintains the changed state. Utilizing this characteristic, it is possible to operate as a self-powered artificial synapse.

[0028] (B2) In the piezo-phototronic element, the remaining light soaking effect is changed by the applied stress during a period in which the light soaking effect of the piezo-phototronic element remains.

[0029] (B3) The piezophototronic element operates as a memory that stores the remaining light soaking effect as memory information during a period in which the light soaking effect of the piezophototronic element remains, and is capable of outputting an electrical signal corresponding to the remaining light soaking effect during the period. The memory information changes depending on the stress applied to the piezophototronic element operating as a memory.

[0030] (B4) It is preferable that the artificial synapse further comprises a stress applying unit that applies a stress to the piezo-phototronic element.

[0031] (B5) The artificial synapse may further include a stress applying unit that applies the stress to the piezo-phototronic element to change the memory information. The memory information may be modulated or reset by the stress.

[0032] (B6) The piezo-phototronic element can operate as a memory that stores optical input by the light soaking effect. Reading of the memory can be performed by irradiating the piezo-phototronic element with light.

[0033] (B7) A method according to an embodiment may be a method of using a piezo-phototronic element, comprising using the piezo-phototronic element as an artificial synapse.

[0034] (B8) The method according to the embodiment may further comprise applying a stress to the piezophototronic element during a period in which the light soaking effect remains in the piezophototronic element.

[0035] (B9) The method according to the embodiment may further comprise reading out a signal corresponding to the remaining light soaking effect by irradiating the piezo-phototronic element with light during a period in which the light soaking effect remains in the piezo-phototronic element.

[0036] (B10) The method according to the embodiment may further comprise applying stress to the piezophototronic element during a period in which the light soaking effect remains in the piezophototronic element, and reading out a signal from the piezophototronic element by irradiating the piezophototronic element with light after the application of the stress.

[0037] (B11) The method according to the embodiment may further comprise irradiating the piezophototronic element with light during a period in which the light soaking effect remains in the piezophototronic element to read out a signal corresponding to the remaining light soaking effect, and applying stress to the piezophototronic element after reading out the signal.

[0038] 2. Examples of Use of Artificial Synapses and Piezo-Phototronic Elements

[0039] Hereinafter, the embodiments will be described in more detail with reference to the drawings.

[0040] An artificial synapse according to the embodiment includes a piezo-phototronic element as an artificial synapse element. The piezo-phototronic element is an element that generates the piezo-phototronic effect. The piezo-phototronic element is a type of photoelectric conversion element (photovoltaic device). The photoelectric conversion element is, for example, a photodiode or a solar cell. The piezo-phototronic element is, for example, a photodiode (piezo-phototronic diode) or a solar cell (piezo-phototronic solar cell) that generates the piezo-phototronic effect.

[0041] Photoelectric conversion elements such as piezo-phototronic elements have a pn junction where a p-type material and an n-type material are joined. If either the p-type material or the n-type material of the photoelectric conversion element is piezoelectric (or ferroelectric), the piezo-phototronic effect occurs and the element functions as a piezo-phototronic element. Piezo-phototronic elements are heterojunction photoelectric conversion elements.

[0042] The piezophototronic effect is a phenomenon in which stress such as strain applied to a piezoelectric semiconductor (or ferroelectric semiconductor) causes polarization in the piezoelectric semiconductor (or ferroelectric semiconductor), generating polarization charge at the pn junction interface, which changes the band offset between the p-type semiconductor and the n-type semiconductor, resulting in changes in photovoltaic power, etc. When stress such as mechanical strain is input to a piezophototronic element, the photovoltaic power, etc. changes due to the piezophototronic effect.

[0043] A piezophototronic element that produces the piezophototronic effect is used as an artificial synapse element. An artificial synapse element is an element that artificially mimics a biological synapse. A piezophototronic element is suitable as an artificial synapse element because the memory stored in the piezophototronic element can be changed (modulated or erased (reset)) by applying stress to the piezophototronic element. This point will be described later.

[0044] The artificial synapse element according to the embodiment is a photoelectric conversion element and therefore a light-responsive element. Also, the artificial synapse element according to the embodiment is a piezo-phototronic element having a piezoelectric material and therefore can be said to be a stress-responsive element that responds to stress applied under light irradiation.

[0045] The piezo-phototronic element may be, for example, a Se thin-film photovoltaic device (Se photoelectric conversion element). The Se photoelectric conversion element includes Se (selenium) as a p-type semiconductor. While the material used as the p-type semiconductor is not limited to Se, Se is capable of photovoltaic generation even with relatively weak light, such as indoor lighting, making it suitable for a wide range of applications as an artificial synapse element. Furthermore, Se has a band gap of approximately 1.9 eV, and its theoretical photoelectric conversion efficiency is maximized (approximately 55%) in the wavelength range of fluorescent lamps or LED lighting used for indoor lighting, making it suitable for photovoltaic generation for indoor power supply. Furthermore, Se can be formed at low temperatures and therefore can be formed on flexible plastic films such as PET films. This makes it advantageous for applications in the fields of soft electronics and soft robotics.

[0046] In the following, p-type semiconductor materials may be referred to by adding "p-" before the material name, and n-type semiconductor materials may be referred to by adding "n-" before the material name.

[0047] The Se photoelectric conversion element is made of ZnMgO, ZnO, or BaTiO ３ That is, the Se photoelectric conversion element as a piezo-phototronic element may be formed by bonding an n-type piezoelectric material such as ZnMgO / Se, ZnO / Se, or BaTiO ３ / TiO ２ / Se, or BaTiO ３ These elements may exhibit the light soaking effect described below.

[0048] The piezo-phototronic element is Cu(In,Ga)Se ２ Thin film solar cell (Cu(In,Ga)Se ２ Photoelectric conversion element). Cu(In,Ga)Se ２ The photoelectric conversion element is p-Cu(In,Ga)Se ２ Cu(In,Ga)Se ２ The photoelectric conversion element may also include an n-type piezoelectric material similar to the Se photoelectric conversion element. ２ The photoelectric conversion element is, for example, ZnMgO / Cu(In,Ga)Se２ , ZnO / Cu(In,Ga)Se ２ , BaTiO ３ / TiO ２ / Cu(In,Ga)Se ２ , or BaTiO ３ / Cu(In,Ga)Se ２ These elements also produce the light soaking effect described below.

[0049] The piezo-phototronic element is a perovskite solar cell (MAPbI ３ or FAPbI ３ ), where MA is methylammonium (CH ３ NH ３＋ ), and FA is formamidinium (HC(NH ２ ) ２＋ The perovskite solar cell can also have an n-type piezoelectric material similar to the Se photoelectric conversion element. The perovskite solar cell as a piezo-phototronic element can have, for example, BaTiO ３ / TiO ２ / MAPbI ３ , BaTiO ３ / MAPbI ３ , BaTiO ３ / TiO ２ / FAPbI ３ , or BaTiO ３ / FAPbI ３ These elements also produce the light soaking effect described below.

[0050] The piezo-phototronic element may be a CdS / CdTe solar cell, which is a type of CdTe solar cell. CdS is a piezo-phototronic element because it has piezoelectricity, and generates the light soaking effect described below. The aforementioned Cu(In,Ga)Se ２ Cu as a substitute for rare metals In and Ga in photoelectric conversion elements ２ ZnSnS ４ A similar effect occurs with (CZTS).

[0051] The piezo-phototronic element exhibits a light soaking effect. The light soaking effect is a phenomenon in which the power generation performance or conductivity of a photoelectric conversion element changes when exposed to a light stimulus. The change in power generation performance is, for example, an improvement or decrease in power generation performance. The change in conductivity is, for example, an improvement or decrease in conductivity.

[0052] Figure 1 shows the light soaking effect of an example of a piezo-phototronic element made of n-ZnMgO / p-Se, where ZnMgO is a piezoelectric semiconductor.

[0053] In Figure 1, (A) shows the current-voltage characteristics of n-ZnMgO / p-Se before light soaking, (B) shows the energy levels before light soaking, (C) shows the current-voltage characteristics of n-ZnMgO / p-Se after light soaking, and (D) shows the energy levels after light soaking.

[0054] Interfacial defects exist at the ZnMgO / Se junction interface. Before light soaking, the interfacial defects are activated. As shown in Figure 1(A), photovoltaic performance is low due to the activated interfacial defects. However, as shown in Figure 1(C), photovoltaic performance improves after light soaking. This is because specific metastable traps at the n-ZnMgO / p-Se heterointerface capture photocarriers upon light irradiation. Therefore, light soaking inactivates the interfacial defects, improving power generation performance. Metastable traps capture photocarriers for a certain period of time after light irradiation, and their carrier capture lifetime is relatively long. This phenomenon can be exploited to control the degree of activation of artificial synapses using piezo-phototronic devices, depending on the conditions (e.g., illuminance, exposure time) under which they are exposed to light. Therefore, devices capable of "memorizing" optical inputs can be realized.

[0055] 1 shows an example of the light soaking effect, which can cause a decrease in photovoltaic performance. That is, the light soaking effect can change photovoltaic performance. The light soaking effect can also be understood as a change in the conductivity of the element.

[0056] The light soaking effect remains even after light irradiation. By utilizing the remaining light soaking effect, an element having the light soaking effect can be used as a memory for storing optical input. In other words, the piezo-phototronic element operates as a memory that stores the remaining light soaking effect as memory information during the period when the light soaking effect caused by the optical stimulus remains. In other words, the piezo-phototronic element operates as a non-volatile memory that indicates the stored value (memory value) based on the power generation performance or conductivity changed by light soaking. Because the changing power generation performance or conductivity takes on continuous values, the piezo-phototronic element as a memory operates as an analog memory.

[0057] A piezophototronic element as a memory has a memory value that corresponds to the input light (light history). The piezophototronic element can output (read) an electrical signal (output signal) that corresponds to the light soaking effect while the light soaking effect remains. In other words, the piezophototronic element can output a memory value.

[0058] A piezo-phototronic element as an artificial synapse element uses a photoelectromotive force or a photocurrent generated when irradiated with light as a weight (signal value) for transmission of a synapse.

[0059] Figure 2 shows an example of the light soaking effect. A device that exhibits the light soaking effect is irradiated with light (light soaking) as shown in Figure 2(A), and then the device is left in a dark environment. In this case, as shown in Figure 2(B), the open circuit voltage V ｏｃIn the early stage of light irradiation, the open-circuit voltage V is relatively low because the interface defects are in an active state. However, as light irradiation (light soaking) continues, the charges generated by the light irradiation are captured by the interface defects, and the interface defects are inactivated. As a result, as light irradiation continues, the open-circuit voltage V ｏｃ When the light irradiation is stopped and the device is left standing in a dark environment, the open-circuit voltage V ｏｃ does not drop significantly immediately, but drops gradually, and the open circuit voltage V ｏｃ This state is maintained for a while. In other words, even in a dark environment, the effect of light soaking remains, and the light input is stored in the element.

[0060] As shown in Figures 2(C) to 2(E), the photocurrent (short-circuit current J per unit area) ｓｃ ), fill factor FF, photovoltaic efficiency Eff, and open circuit voltage V ｏｃ The change is similar to that of the photovoltaic power generation performance, and it is clear that a change in power generation performance and memory of light input have occurred.

[0061] Fig. 3 shows the readout of the light irradiation history (memory value) stored in the element (photovoltaic device) by light soaking, where Fig. 3(A) shows the case with light soaking (with light irradiation history) and Fig. 3(B) shows the case without light soaking (without light irradiation history).

[0062] As shown in Figure 3(A), when light soaking is performed, the photovoltaic device current (absolute value) gradually increases due to light soaking, improving power generation performance. When light soaking ends and the environment becomes dark, the current of the photovoltaic device drops to almost zero because there is no light irradiation. However, even without light irradiation, the improvement in power generation performance (light soaking effect) is maintained for a while. Therefore, when light irradiation is performed to read the memory while the improvement in power generation performance is maintained, the photovoltaic device, which is a memory, immediately outputs a large output signal (output current).

[0063] On the other hand, as shown in Figure 3(B), when there is no light soaking, even if light is irradiated for readout, the photovoltaic device is in a dark environment and has (almost) no memory of the light irradiation history, so it outputs a small output signal or no output signal.

[0064] 3A and 3B, even if the intensity of light irradiation for readout is the same, the magnitude of the output signal differs depending on whether or not light soaking (light irradiation history) has occurred. This indicates that the photovoltaic device stores the light input prior to readout.

[0065] Figure 4 shows the memory and forgetting of light irradiation history. The memory of light irradiation history decreases over time. As shown in Figure 4(A), when the dark period after light soaking is short, the photovoltaic device, which acts as a memory, memorizes the light irradiation (light soaking), and therefore outputs a relatively large output signal (output current) when light is irradiated for reading.

[0066] On the other hand, as shown in Figure 4(B), if the dark period after light soaking is long, the photovoltaic device, which is a memory, forgets the light irradiation (light soaking). Forgetting means that the change in power generation performance or conductivity returns to its original state. Because the photovoltaic device has forgotten its memory, when it is irradiated with light for reading, it outputs a small output signal or no output signal.

[0067] The fact that the artificial synapse remembers the light stimulus for a while and then eventually forgets it is similar to the way that a light stimulus remains for a while in the retina of a living organism when it is exposed to strong light.

[0068] By arranging elements (artificial synapses) that produce the light soaking effect in two dimensions, a photoelectric memristor (artificial visual synapse device) for image recognition can be formed. Figure 5 shows the writing and reading of data to each element (each memory) when a shadow of the number "2" is projected onto a photoelectric memristor consisting of a two-dimensional array of 20 x 20 elements.

[0069] Here, as shown in Figures 5(A) and 6, a shadow of the number "2" is projected when writing. When writing, the part of the number "2" is in shadow and is not irradiated with light, while the other parts are irradiated with light. Therefore, as shown in Figure 5(B), even when light is irradiated for writing, the power generation performance of the element (cell) corresponding to the part of the number "2" does not increase. However, as shown in Figure 5(B), the power generation performance of the element (cell) in the part other than the part of the number "2" increases (memory of light soaking history).

[0070] As shown in Figures 5(C) and 6, after writing, all cells are irradiated with light for readout. If light irradiation for readout occurs immediately after light soaking or a relatively short time later, elements (cells) with a light soaking history generate a larger photocurrent or photovoltage than elements (cells) without a light soaking history. As shown in Figures 5(C) and 6, if readout can be achieved by light irradiation, no power supply for readout is required, which is preferable. Note that after a long time has passed since light soaking, even if light irradiation for readout is performed, the difference in output depending on whether or not there is a light soaking history becomes small.

[0071] The stored light soaking history (changes in power generation performance or conductivity; memory information) can be read not only by light irradiation but also by simultaneous application of a bias voltage to all elements (all cells), as shown in Figure 7. In this case, too, differences in output occur immediately after light soaking or after a relatively short time, depending on whether or not there is a light soaking history. Reading by applying a bias voltage does not require light irradiation for reading, and is therefore suitable for cases where light irradiation for reading is to be avoided.

[0072] The applied bias voltage may be supplied by another photovoltaic device. In this case, light irradiation for readout is performed on the other photovoltaic device, not on the written photovoltaic device (photoelectric conversion memristor). The other photovoltaic device generates a bias voltage for readout, and the generated bias voltage is applied to the written photovoltaic device (photoelectric conversion memristor) for readout. In this case, since the power for readout is obtained by light irradiation, an external power supply for readout is not required. Furthermore, since the written photovoltaic device is not irradiated with light for readout, this is suitable for cases where light irradiation for readout is to be avoided.

[0073] The stored light soaking history may be read out in other ways, for example by reading the open circuit voltage of the photovoltaic device.

[0074] An artificial synapse according to the embodiment includes, as an artificial synapse element, a piezo-phototronic element that generates the light soaking effect described above. The artificial synapse element outputs an output signal in response to an input. The artificial synapse element can store the input and output an output signal according to the memory. The piezo-phototronic element serving as an artificial synapse element outputs an electrical signal (photovoltage or photocurrent) as an output signal in response to the input light or stress. The piezo-phototronic element serving as an artificial synapse element stores the input light or stress, and during the period in which the memory is maintained, outputs an output signal according to the memory as a weight (signal value) to be transmitted by the synapse.

[0075] The two-dimensional array element (photoelectric conversion memristor) shown in Fig. 5 can be used as an artificial visual synapse element in the input layer of a neural network. Fig. 8 shows an example of discriminating numbers in an image using a neural network. The multiple (400) nodes X that make up the input layer in Fig. 8 １ ~X ４００ Each one is a light-responsive artificial synapse element.

[0076] For example, a neural network calculates the probability that a number in an image is between 0 and 9 (Y ０ ~Y ９ The neural network is trained to output a high value of Y [%]. The trained neural network can distinguish numbers from images. For example, when an image of the number "2" is input to the neural network, the neural network will output a high value of Y [%]. ２ [%] is output. Note that the image of the number "2" here can be input by reading the information in FIG. 5(C). Reading the information in FIG. 5(C) is reading the information written in FIG. 5(B). Furthermore, in machine learning, the weights of the neural network are adjusted to reduce the error in the output data obtained from the input data given to the input layer, and the input data given to the input layer is used during this adjustment. For this reason, the artificial synapses in the input layer must have a memory function. In other words, it is important that the artificial synapses used in the neural network have the function of storing (memorizing) the light irradiation history.

[0077] As shown in Figure 9, the memory and output of an artificial synapse that constitutes one cell (node) of the input layer of a neural network can be changed depending on the intensity of the light stimulus irradiated to the artificial synapse. For example, cell X in Figure 9(A) １ When the intensity of light irradiated to the artificial synapse is weak, the memory value is small as shown in Figure 9(B). When the intensity of light is medium, the memory value is also medium as shown in Figure 9(B). When the intensity of light is strong, the memory value is also large.

[0078] The intensity of the light stimulus varies not only depending on the intensity of the light (light pulse), but also on the frequency of the light pulse, the input time of the light pulse, the number of light pulses, and the wavelength (energy) of the light. These intensities of the light stimulus are stored in the artificial synapse as a light irradiation history.

[0079] When exposed to strong light stimulation, artificial synapses can maintain a high-conductance state (memory of the light input) for several seconds or longer (long-term plasticity (LTP)). This can last for minutes, hours, days, or even years. There are many different values ​​of conductivity, which are commonly referred to as programmed states.

[0080] When exposed to weak light stimulation, artificial synapses can forget memories in a relatively short time (short-term plasticity (STP)). In short-term plasticity, the conductivity of the artificial synapse increases after each light pulse, but when the light pulses cease, the artificial synapse can return to its original conductivity in just a few seconds or less.

[0081] The light stimulation can determine whether the artificial synapse exhibits short-term or long-term plasticity. For example, if the light pulses are very frequent but very few in number, the artificial synapse may only exhibit short-term plasticity.

[0082] The light irradiation may also be performed multiple times. Repeated light irradiation can increase the conductivity stepwise. As shown in Figure 10, multiple irradiations with weak light produce a weak signal (a small value memory), while multiple irradiations with strong light produce a strong signal (a large value memory).

[0083] As mentioned above, the memory of an artificial synapse using a piezo-phototronic element can be changed by stress. Figure 11 shows the experimental results of stress-induced changes in the memory of an artificial synapse (artificial visual synapse). In the experiment, an artificial synapse using n-ZnMgO / p-Se elements was used as the piezo-phototronic element. In the experiment, the artificial synapse was first placed in a dark environment ("Initial (dark)" in Figure 11). In the initial (dark) state, a bias voltage of 0.8 V was applied to the piezo-phototronic element, and the dark current (vertical axis in Figure 11) was read.

[0084] Then, multiple light pulses (pulse width 455 nm, pulse frequency 100 Hz) were irradiated onto the artificial synapse for approximately 5 seconds to allow light soaking ("Light Irradiation" in Figure 11(A)). The light soaking effect changes photovoltaic performance or conductivity. In the experiment, the increase in dark current due to the light soaking effect was measured. As shown in Figure 11(B), the dark current during light irradiation was measured by setting the bias voltage to 0.8 V at the timing when the pulsed light was turned off and reading the diode current in the dark. Note that the application of bias voltage itself does not cause changes in photovoltaic performance or conductivity. A low bias voltage for reading is preferable, as a low bias voltage can reduce power consumption.

[0085] As shown in Figure 11(A), the dark current gradually increased from 2.3 μA to 4 μA due to light irradiation. In other words, the conductivity state of the artificial synapse changed due to light irradiation. In other words, the artificial synapse memorized the light irradiation.

[0086] As shown in FIG. 11(A), even after the light irradiation is stopped ("Memory (dark state)" in FIG. 11(A)), the state of large dark current (state of high conductivity) is maintained for several seconds. In other words, the memory of light irradiation is maintained even after light irradiation. Note that during the memory (dark state), a bias voltage of 0.8 V was applied to the piezo-phototronic element, as in the initial (dark state), to read the dark current. The period during which the memory is maintained varies depending on the light irradiation conditions and the material of the device, and may be several seconds, several minutes, several hours, or several days.

[0087] In the experiment, while the memory was maintained, stress was applied to the artificial synapse using the piezo-phototronic element by bending it mechanically ("bending deformation" in FIG. 11(A)) as shown in FIG. 11(C). As shown in FIG. 11(A), the dark current after the initial stage decreases due to bending deformation. In other words, the application of stress resets the change in conductivity (memory reset).

[0088] The application of stress is not limited to mechanical bending, but may also be the application of vibration or other external forces. To apply stress, the artificial synapse according to the embodiment may include a stress application unit. The stress application unit may be, for example, a mechanism for mechanically bending the artificial synapse (bending drive device), a mechanism for pressing the artificial synapse (pressing device), or a mechanism for applying vibration to the artificial synapse (vibration device). However, the artificial synapse does not need to include a stress application unit, and an external force applied to the artificial synapse may also be used.

[0089] The application of stress induces a potential gradient across the pn junction due to the piezo-phototronic effect of the piezoelectric piezo-phototronic element. This potential gradient exerts a force on charge carriers trapped at the interface, causing them to move away from the defect states. As electrons and holes move away from the defect states, they recombine or release charge from the trap states, effectively reducing the conductivity of the device.

[0090] If the stress applied to the piezo-phototronic element is sufficiently large, the conductivity can be reset (returned to its initial state, i.e., the state before light irradiation). Alternatively, if the applied stress is relatively weak, the magnitude of the conductivity can be changed rather than reset. Therefore, the application of stress can modulate or erase (reset) the memory in the piezo-phototronic element as an artificial synapse element. This process in the artificial synapse element is similar to synaptic depression in biological systems, where the strength of the synaptic connection decreases and the synapse forgets or resets the stored information.

[0091] The artificial synapses having the characteristics shown in FIG. 11 are used as the nodes in the input layer (and the edges connecting to the hidden layer) in FIG. 8. １ From X ４００Each graph shows the weight of the optical signal captured in the input layer, i.e., the conductance of each synapse after visual stimulation. In Fig. 11, the state of the artificial synapse (memory of the optical input; weight of the artificial synapse) is read as the current value when a predetermined readout voltage (bias voltage) is applied. The state of the artificial synapse changes depending on the properties of the optical stimulation (frequency of the light pulse, light intensity, wavelength, exposure time, etc.). A strong input signal can increase the conductance more significantly than a weak input signal.

[0092] As shown in Figure 11, the artificial synapse using piezo-phototronic elements functions as a nonvolatile memory that maintains a highly conductive state for several seconds, minutes, or even hours without the need for additional stimulation by light or voltage. By applying a reverse voltage bias (e.g., -1 V) or mechanical bending (stress) to the artificial synapse, the weights held by the memory can be erased by injecting / moving charges through the heterostructure (piezo-phototronic effect).

[0093] As shown in Fig. 11, an artificial synapse in which a bias voltage (voltage spike) is used to read the state (memory; conductivity) of the artificial synapse can be called a "spiking artificial synapse." The voltage spike can be small, and therefore the energy consumption of the spiking artificial synapse is small.

[0094] However, artificial synapses using piezo-phototronic elements such as n-ZnMgO / p-Se can be photo-powered, eliminating the need for voltage spikes and allowing them to operate on their own power source. In other words, instead of applying voltage spikes for reading, electrical signals such as photocurrents generated in response to irradiation with light (reading light) such as pulsed light can be used as output signals (visual information). For this reason, artificial synapses using piezo-phototronic elements can be called "self-powered artificial synapses." Self-powered artificial synapses can operate with virtually zero power (external power).

[0095] In a self-powered artificial synapse, light illumination for memory readout can be integrated with the subsequent writing of light stimulation.

[0096] Figure 12 shows an example of the operation of a self-powered artificial synapse. Here too, an artificial synapse using n-ZnMgO / p-Se elements as piezo-phototronic elements was used.

[0097] In Figure 12, the vertical axis represents photocurrent, with increasing current moving downward. By transmitting ten consecutive 10-second pulses of 455 nm light to the artificial synapse, we demonstrate a clear increase in photocurrent with each pulse (the downward direction on the vertical axis in Figure 12 indicates an increase). Next, by transmitting a series of 5-second pulses approximately every 6 minutes, we observe a gradual decrease in photocurrent, dropping back to its original value after 20 minutes. In other words, the artificial synapse gradually forgets the information it had retained. Because no voltage is used during this cycle, the artificial synapse operates with zero power and spontaneously memorizes its light exposure history. The increase in photocurrent with successive light pulses is thought to be due to the promotion of charge carrier generation, or carrier capture, defect deactivation, and reduced recombination rate due to repeated stimulation. Similarly, the gradual forgetting process occurs when these carriers are thermally or optically released from defect capture over time.

[0098] As described above, the artificial synapse (artificial visual synapse) according to the embodiment uses a piezo-phototronic element, which converts mechanical stress into an electrical signal through the piezo-phototronic effect, enabling it to regulate synaptic plasticity (conversion from mechanical signals to electrical signals). This function enables synapses to respond not only to optical stimuli but also to mechanical changes in the environment, increasing their sensitivity and adaptability to a wider range of stimuli.

[0099] Tuning the response of artificial synapses to light using mechanical bending or stress allows for greater control over their behavior (enhanced adaptability and functionality). This adaptability is crucial for applications where artificial synapses may undergo physical deformation, such as wearable electronics and flexible displays, allowing the artificial synapse to maintain optimal function despite mechanical stress.

[0100] Furthermore, the artificial synapse according to the embodiment can operate as a self-powered visual synapse by utilizing the properties of photovoltaic power. By being self-powered, the artificial synapse can monitor the short-circuit current density (the current generated by light irradiation without the application of an external voltage) under lighting conditions, eliminating the need to supply an external read voltage (high energy efficiency during operation). Instead, it can directly use the electrical response to the lighting environment. This reduces the power requirements of the artificial synapse and extends its lifespan. This allows it to be used in remote locations and energy-poor environments.

[0101] After light stimulation, the self-powered artificial synapse generates a higher short-circuit current for subsequent light pulses compared to the first light stimulus. This is a memory of the previous light pulse. This effect depends on the characteristics of the incident light, such as frequency, intensity, pulse width, and energy, which modulate the device's response. Each pulse of light enhances the photocurrent of the artificial synapse, effectively storing a memory of the light exposure. This photocurrent enhancement due to storage indicates the weight of the artificial synapse, and can be maintained for a long period after light exposure without the need for a power source.

[0102] The short-circuit current density (Jsc), which is the response current to light irradiation at the time of activation of the artificial synapse, depends on the light irradiation history up to the previous use of the artificial synapse, enabling the introduction of a unique energy-efficient mechanism for memory and learning (dynamic response to light irradiation history). This feature enables the synapse to dynamically adapt its response based on the environmental history without requiring continuous power or external system intervention, enhancing its applicability to autonomous sensor networks and intelligent systems operating for long periods of time.

[0103] Furthermore, the artificial synapse according to the embodiment can operate as a spiking artificial synapse. The artificial synapse according to the embodiment can dynamically and time-reversibly change its conductivity using a short optical pulse (e.g., a pulse with a frequency of 1 kHz), and can be regarded as an optically controlled memory element (memristor). Like other memristors, the artificial synapse operating as a memory can be reset to its state before the optical pulse was input by applying a reverse bias voltage (-1 V).

[0104] In accordance with embodiments, artificial synapses can be implemented with a method for dynamically modulating or completely erasing memory by applying stress, such as mechanical bending, utilizing the piezo-phototronic effect, rather than resetting via the application of an electrical reverse bias voltage. Mechanical bending influences the distribution and recombination rate of charge carriers within the device by moving piezoelectric charges, allowing the artificial synapse's conductivity to be adjusted back to its original or new state. The magnitude of the modulation or erasure effect can also be controlled by the magnitude of the applied bending / stress. This ability to mechanically modulate or erase memory, combined with persistent photoconductivity, promises to be a versatile and energy-efficient mechanism for synaptic plasticity in artificial synapses. Furthermore, these features enable the creation of energy-efficient, self-powered artificial synapses capable of memory and learning functions that mimic parts of the human visual cortex in the field of neuromorphic visual computing.

[0105] The artificial synapse according to the embodiment can also be used as an artificial synapse that receives mechanical stress or strain as input. Because the artificial synapse according to the embodiment uses a piezo-phototronic element, it is possible to control the magnitude of the modulation or elimination effect depending on the magnitude of the applied mechanical stress or strain. Therefore, for example, under constant light irradiation conditions, it can be applied to an artificial synapse that uses the change in photocurrent after inputting mechanical stress or strain as the signal weight. By integrating the artificial synapse as a strain detector with mechanical structures such as a deformable membrane (diaphragm) or a vibrator (spring, mass, damper), it is possible to measure pressure or acceleration. By adopting such a combination, it can be applied to an artificial synapse that receives pressure or acceleration as input. For example, it can be applied to a flexible tactile display that is capable of low power consumption and high-speed processing.

[0106] In addition, after applying stress to the artificial synapse to modulate or erase the memory, the signal from the artificial synapse (the modulated memory signal or the reset memory signal) may be read out, or as shown in Figure 13, after reading out the signal from the artificial synapse (the memory signal) (Figure 13(A)), stress may be applied to the artificial synapse to reset or modulate the memory (Figure 13(B)).

[0107] Examples of applications of the artificial synapse according to the embodiment are listed below.

[0108] (Advanced imaging systems) The artificial synapse can be applied to cameras and imaging systems that require a high dynamic range and adaptability to various lighting conditions. By adjusting the sensitivity based on the light exposure history prior to the activation of the artificial synapse, it is possible to capture more detailed images even in environments where the light intensity changes dramatically, such as in-car cameras for driver assistance systems.

[0109] (Smart Surveillance) In security applications, a network of artificial visual synapses can process visual data on-site and identify potential threats or changes without requiring a high-bandwidth connection to a central processing unit. Their ability to remember and learn from historical data helps them recognize patterns and anomalies over time.

[0110] (Self-driving cars and drones) These vehicles require efficient, real-time processing of visual information for navigation and decision-making. The low energy consumption and fast processing capabilities of artificial visual synapses make them ideal for these applications, where power efficiency and rapid response to environmental changes are critical.

[0111] (Biomedical Imaging) In this field, the adaptability and processing power of artificial visual synapses can improve the analysis of complex visual data, such as live cell imaging and dynamic changes in tissues. The ability to retain information and adjust sensitivity enhances the detection and monitoring of subtle changes over time.

[0112] (Environmental Monitoring) Deployed in remote locations, networks of these artificial synapses could monitor changes in the natural environment with minimal energy use. Non-volatile memory allows tracking changes over long periods of time, providing valuable data for ecological and climate change studies.

[0113] Fig. 14 shows an example of the configuration of the aforementioned artificial synapse. The artificial synapse 10 shown in Fig. 14 includes a piezo-phototronic element 11 as a photovoltaic device (photoelectric conversion element). Light from a light source 12 is irradiated onto the piezo-phototronic element 11. The artificial synapse 10 shown in Fig. 14 also includes a controller 13. The controller 13 controls the light source 12 and the like.

[0114] The light source 12 irradiates the photovoltaic device 11 with light, which serves as an optical stimulus to be stored in the photovoltaic device 11. That is, the light from the light source 12 can be used to write to the photovoltaic device 11, which operates as a memory. The light source 12 is, for example, an LED. The artificial synapse 10 can include one or more light sources 12 that irradiate the photovoltaic device 11 with light. The light source 12 may be an external light source of the artificial synapse 10.

[0115] The light from the light source 12 can also be used to read the memory from the photovoltaic device 11 acting as a memory.

[0116] The plurality of light sources 12 may have separate light sources for different purposes such as writing / reading, or a common light source may irradiate light for each purpose.

[0117] The controller 13 can control the light source 12 to emit light (photostimulus) for writing.

[0118] The controller 13 can control the light source 12 so that after the photovoltaic device 11 receives a light stimulus that writes data to memory, the photovoltaic device 11 is irradiated with light for reading the memory. The photovoltaic device 11 changes its photovoltaic performance upon receiving the light stimulus, and retains the changed photovoltaic performance as a memory of the received light stimulus. When the photovoltaic device 11 is irradiated with light for reading the memory, an electrical signal (photocurrent) corresponding to the change in photovoltaic performance remaining as a memory of the light stimulus (change in photovoltaic performance due to the light soaking effect) is output from the photovoltaic device 11. In this way, the controller 13 can read the memory of the light stimulus after receiving the light stimulus.

[0119] The artificial synapse 10 shown in Figure 14 is a self-powered type in which no power is supplied (e.g., bias voltage is applied) to the element 11 to store the optical stimulus, and no power is supplied (e.g., bias voltage is applied) to the element 11 to read out the stored optical stimulus.

[0120] The artificial synapse 10 shown in Figure 14 can include a stress application unit 14 that applies stress, such as mechanical bending, to the photovoltaic device 11. A controller 13 can control the stress application unit 14 to control the application of stress to the photoelectric conversion element 11. The controller 13 can also dynamically modulate or completely erase memory by applying stress, such as mechanical bending, by utilizing the piezo-phototronic effect of the photovoltaic device 11, which is a piezo-phototronic element. In this way, the controller 13 can modulate or erase memory of the optical stimulus after the photovoltaic device 11 receives an optical stimulus.

[0121] The controller 13 can modulate (including erase) the memory by any one of light irradiation, reverse bias application, and stress application, or a combination of two or more of these. The magnitude of the effect of modulation or erasure can also be controlled by the magnitude of the applied bending or stress.

[0122] The controller 13 can control the stress application unit 14 to apply stress to the photovoltaic device 11 after the photovoltaic device 11 receives a light stimulus, and can control the light source 12 to irradiate light for reading out the memory of the light stimulus after the stress application. In this case, the controller 13 can, for example, modulate the memory of the light stimulus and read out the modulated memory.

[0123] After the photovoltaic device 11 receives a light stimulus, the controller 13 controls the light source 12 to irradiate the light for reading out the memory of the light stimulus, and after reading out, controls the stress application unit 14 to apply stress to the photovoltaic device 11. In this case, the controller 13 can, for example, read out the memory of the light stimulus and erase the memory after reading out.

[0124] Fig. 15(A) shows a module of an artificial synapse 10 including a plurality of photovoltaic devices 11. As an example, the artificial synapse 10 shown in Fig. 15(A) includes a two-dimensional 4x4 array of photovoltaic devices 11.

[0125] FIG. 15(B) shows the fabrication process of ZnMgO / Se, an example of a photovoltaic device 11 (photoelectric conversion element 11). As shown in FIG. 2(B), first, an ITO electrode is formed on a glass substrate. Then, n-ZnMgO is formed on the ITO electrode by sputtering. Furthermore, Te / p-se is formed on the n-ZnMgO by vapor deposition. Then, MoO3 / Au is formed on the Te / p-se by vapor deposition. FIG. 2(C) shows the cross-sectional structure of the photovoltaic device 11 of the ZnMgO / Se artificial synapse fabricated as described above. Light is irradiated onto this ZnMgO / Se artificial synapse from the upper glass substrate side.

[0126] Fig. 16 shows a configuration example of an artificial synapse 10 including a plurality of piezo-phototronic elements 11 (photovoltaic devices 11) as shown in Fig. 5 or 15. The controller 13 shown in Fig. 16 can read values ​​(memory information; photovoltaic performance) stored in response to input of optical stimuli to each element 11 by irradiating each element 11 with light or the like. The controller 13 can also modulate or erase the memory information by applying stress to each element 11.

[0127] The present invention is not limited to the above-described embodiment, and various modifications are possible.

[0128] 10: Artificial synapse 11: Piezo-phototronic element (photovoltaic device) 12: Light source 13: Controller 14: Stress application unit

Claims

1. An artificial synapse comprising an artificial synapse element that operates as a memory for storing received optical stimuli and that reads out the memory of the optical stimulus after receiving the optical stimulus, wherein the artificial synapse element is constituted by a photovoltaic device that is a piezo-phototronic element whose photovoltaic performance changes when it receives the optical stimulus and the change in photovoltaic performance remains after receiving the optical stimulus so that the changed photovoltaic performance is retained as a memory of the optical stimulus, and wherein the memory of the optical stimulus from the artificial synapse element is read out by outputting an electrical signal from the piezo-phototronic element according to the photovoltaic performance that is the memory of the optical stimulus after receiving the optical stimulus.

2. The artificial synapse according to claim 1, further comprising a stress application unit that applies stress to the piezo-phototronic element after receiving the optical stimulus.

3. The artificial synapse according to claim 2, wherein the memory of the optical stimulus is changed by the stress applied by the stress application unit.

4. The artificial synapse according to claim 2, wherein the memory of the optical stimulus is erased by the stress applied by the stress application unit.

5. The artificial synapse according to claim 2, wherein after receiving the optical stimulus, stress is applied to the piezo-phototronic element by the stress application unit, and the memory of the optical stimulus is read out after the stress application.

6. The artificial synapse according to claim 2, wherein after receiving the optical stimulus, the memory of the optical stimulus is read out, and after reading out, stress is applied to the piezo-phototronic element by the stress application unit.

7. The artificial synapse according to claim 1, which is a self-powered type in which no power is supplied to the piezo-phototronic element to memorize the optical stimulus.

8. The artificial synapse according to claim 1, wherein the memory of the optical stimulus is read out by irradiating the piezo-phototronic element with light for readout after receiving the optical stimulus, and outputting from the piezo-phototronic element an electrical signal generated by photovoltaic power generation according to the photovoltaic performance, which is the memory of the optical stimulus.

9. The artificial synapse according to claim 8, which is a self-powered type in which power is not supplied to the piezo-phototronic element for reading out the memory of the optical stimulus.

10. The artificial synapse according to claim 8, which is a self-powered type in which no power is supplied to the piezo-phototronic element to store the optical stimulus, and no power is supplied to the piezo-phototronic element to read out the stored optical stimulus.

11. The artificial synapse of claim 1, wherein the piezo-phototronic element is a selenium photovoltaic device that produces the piezo-phototronic effect.

12. The artificial synapse according to claim 1, wherein the piezo-phototronic element is a solar cell that produces the piezo-phototronic effect.

13. A method of using a piezo-phototronic element, comprising: using a piezo-phototronic element in which a change in photovoltaic performance occurs after receiving an optical stimulus, and the change in photovoltaic performance remains after receiving the optical stimulus, so that the changed photovoltaic performance is stored as a memory of the optical stimulus; and outputting an electrical signal from the piezo-phototronic element according to the photovoltaic performance, which is the memory of the optical stimulus, after receiving the optical stimulus.

Images