Memristor, semiconductor device including same, and array system including memristor

The memristor with a four-terminal configuration and controlled dopant distribution in an oxide semiconductor layer addresses the limitations of two-terminal memristors by replicating biological synapse functions, achieving associative learning and information integration.

JP7591784B2Active Publication Date: 2024-11-29OSAKA UNIVERSITY
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
JP2020214807
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2019-12-26
Filing Date
2020-12-24
Publication Date
2024-11-29
Estimated Expiration
2040-12-24

AI Technical Summary

Technical Problem

Existing two-terminal memristors fail to replicate the higher-order functions of biological synapses, such as associative learning and information integration, due to their inability to self-regulate transmission parameters and exhibit functions like federated learning and information regression.

Method used

A memristor design that includes an oxide semiconductor layer with controlled dopant distribution, utilizing a four-terminal configuration with terminals arranged to allow for two-dimensional or three-dimensional dopant distribution, and a control circuit to implement voltage controls for spike timing dependent plasticity, Pavlovian conditioning, and short-term and long-term synaptic enhancements.

Benefits of technology

The memristor effectively demonstrates higher-order functions of biological systems, enabling associative learning and information integration by controlling dopant distribution and synaptic connections, mimicking the behavior of biological synapses.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007591784000003
    Figure 0007591784000003
  • Figure 0007591784000004
    Figure 0007591784000004
  • Figure 0007591784000005
    Figure 0007591784000005
Patent Text Reader

Abstract

To provide a memristor capable of exerting a high order function possessed by a biological system.SOLUTION: A memristor 10 is a memristor requiring no insulation breakdown process and comprises an oxide semiconductor layer 2 and terminals 3 to 6. The oxide semiconductor layer 2 is made of a single crystal TiO2-x(0<x<2) and contains dopants. The terminals 3 to 6 are disposed on a surface of the oxide semiconductor layer 2. The terminal 3 faces the terminal 5; the terminal 4 faces the terminal 6. The terminals 3 to 6 apply voltage to the oxide semiconductor layer 2 to control a distribution of the dopants in the oxide semiconductor layer 2 to be a two-dimensional distribution.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a memristor, a semiconductor device including the memristor, and an array system including the memristor. [Background technology]

[0002] von Neumann type computers faithfully reproduce left-brain thinking such as logical thinking and calculation, but are extremely inefficient at right-brain thinking such as image and pattern recognition and associative memory, which are essential for artificial intelligence, and require tens of millions of times the power consumption of the human brain (standby: 20W, calculation: +1W). Therefore, the development of non-von Neumann type right-brain (hereafter referred to as brain type) computers is essential for reducing power consumption.

[0003] A brain-type computer is made up of integrated circuits that mimic the biological nervous system and interconnect key synapse elements. Key synapse elements change their output signal depending on the state of the input signal, and are usually made up of a large number of transistors.

[0004] In recent years, there has been a surge in research into replacing these synapse elements with standalone two-terminal memristors, whose electrical resistance changes non-volatilely depending on the charge input history.

[0005] A three-dimensional crossbar array system having two-terminal memristors is known, as described in Patent Document 1. The three-dimensional crossbar array system includes a plurality of crossbar arrays, a first demultiplexer, a second demultiplexer, and a third demultiplexer. Each crossbar array includes a first layer of nanowires, a second layer of nanowires overlapping the first layer of nanowires, and a third layer of nanowires overlapping the second layer of nanowires. The first demultiplexer is configured to address nanowires in the first layer of nanowires of each crossbar array, the second demultiplexer is configured to address nanowires in the second layer of nanowires of each crossbar array, and the third demultiplexer is configured to provide signals to nanowires in the third layer of nanowires of each crossbar array. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Special Publication No. 2010-522987 Summary of the Invention [Problem to be solved by the invention]

[0007] However, synapses in biological systems do not determine the output state based on only one input, but self-adjust transmission parameters under the influence of themselves or other synapses, and perform functions such as associative learning and information regression / integration. In other words, two-terminal memristors have limitations in imitating the high-level functions inherent in biological systems. And the three-dimensional crossbar array system described in Patent Document 1 is premised on a configuration using two-terminal memristors and three-terminal transistors, and therefore cannot perform functions such as associative learning and information regression / integration.

[0008] Therefore, according to an embodiment of the present invention, a memristor capable of exerting the higher-order functions of a biological system is provided.

[0009] Furthermore, according to an embodiment of the present invention, a semiconductor device including a memristor that exerts the higher-order functions of an ecosystem is provided.

[0010] Furthermore, according to an embodiment of the present invention, there is provided an array system equipped with memristors capable of exerting the higher-order functions of biological systems. [Means for solving the problem]

[0011] (Configuration 1) According to an embodiment of the present invention, a memristor does not require a dielectric breakdown process and includes an oxide semiconductor layer and a terminal. The oxide semiconductor layer includes a dopant. The terminal applies a voltage to the oxide semiconductor layer to control the distribution of the dopant to a two-dimensional distribution or a three-dimensional distribution.

[0012] (Configuration 2) In the first configuration, the terminals include a first terminal, a second terminal, and n (n is an integer equal to or greater than 1) third terminals. The first terminal is disposed on one surface in the thickness direction of the oxide semiconductor layer. The second terminal is disposed on the other surface in the thickness direction of the oxide semiconductor layer. The n third terminals are disposed on side surfaces parallel to the thickness direction of the oxide semiconductor layer.

[0013] (Configuration 3) In the second configuration, the memristor further includes a first insulating layer and a second insulating layer. The first insulating layer has a first through hole having an area smaller than a contact area between the first terminal and the oxide semiconductor layer, and is disposed between the first terminal and the oxide semiconductor layer. The second insulating layer has a second through hole having an area smaller than a contact area between the second terminal and the oxide semiconductor layer, and is disposed between the second terminal and the oxide semiconductor layer. The first terminal is in contact with the oxide semiconductor layer through the first through hole, and the second terminal is in contact with the oxide semiconductor layer through the second through hole.

[0014] (Configuration 4) In configuration 2, the first terminal has a T-shaped cross-section in the cross section of the memristor, and the second terminal has a dimension equal to the width of the vertical bar part of the T in a direction perpendicular to the thickness direction of the oxide semiconductor layer.

[0015] (Configuration 5) In any one of the configurations 1 to 4, the oxide semiconductor layer is made of amorphous gallium oxide containing an oxygen amount less than the stoichiometric ratio.

[0016] (Configuration 6) In any one of the configurations 1 to 4, the oxide semiconductor layer is made of single-crystal titanium oxide containing less oxygen than the stoichiometric ratio and having a crystal face oriented in the (001) direction.

[0017] (Configuration 7) Also, according to an embodiment of the present invention, the array system is a crossbar-structured array system, and includes a plurality of first nanowires, a plurality of second nanowires, a plurality of third nanowires, and a memristor.

[0018] The plurality of first nanowires are arranged in a plane. The plurality of second nanowires are arranged in a plane and intersect with the plurality of first nanowires. The plurality of third nanowires are arranged in a plane and intersect with the plurality of first nanowires and the plurality of second nanowires. The memristor is disposed at the intersection of the first nanowire, the second nanowire, and the third nanowire. And, the memristor is a memristor of any one of configurations 1 to 6.

[0019] (Configuration 8) Further, according to an embodiment of the present invention, a semiconductor device includes a memristor and a control circuit. The memristor does not require a breakdown process. The control circuit controls the memristor.

[0020] The memristor includes an oxide semiconductor layer and a terminal. The oxide semiconductor layer includes a dopant. The terminal applies a voltage to the oxide semiconductor layer and controls the distribution of the dopant to a two-dimensional distribution. The terminal includes a first terminal, a second terminal, a third terminal, and a fourth terminal. The first terminal is disposed at a first end of the oxide semiconductor layer in a first direction parallel to an in-plane direction of the oxide semiconductor layer. The second terminal is disposed at a second end of the oxide semiconductor layer facing the first end in the first direction. The third terminal is disposed at a third end of the oxide semiconductor layer in a second direction that is in the in-plane direction of the oxide semiconductor layer and different from the first direction. The fourth terminal is disposed at a fourth end of the oxide semiconductor layer facing the third end in the second direction.

[0021] The control circuit executes any one of a first voltage control, a second voltage control, a third voltage control, and a fourth voltage control on the memristor. The first voltage control is a voltage control for implementing spike-timing dependent plasticity in the memristor. The second voltage control is a voltage control for implementing Pavlovian conditioning in the memristor. The third voltage control is a voltage control for implementing short-term potentiation in which the binding strength enhanced by the input is maintained for only a first period and then decays, and long-term potentiation in which the enhanced binding strength is maintained for a period longer than the first period, in the memristor. The fourth voltage control is a voltage control for implementing low conductance and high conductance of the oxide semiconductor layer in the memristor.

[0022] (Configuration 9) In the configuration 8, the control circuit, in the first voltage control, a first step of applying a pulse voltage to the first terminal while the second terminal is grounded to initialize a conductance of the oxide semiconductor layer between the first terminal and the second terminal; a second step of applying a read voltage for reading the conductance of the oxide semiconductor layer to the first terminal while the second terminal is grounded after the first step; a third step of applying a first spike voltage, which is a result of subtracting a voltage value of a second pulse voltage, which is a square wave and has a rise timing different from that of the first pulse voltage, from a voltage value of a first pulse voltage, which is a square wave, to the first terminal while the second terminal is grounded; a fourth step of applying a read voltage to the first terminal while grounding the second terminal after the third step; a fifth step, after the fourth step, of executing the first step; a sixth step of applying a first spike voltage to the first terminal while the second terminal is grounded after the fifth step, and applying a second spike voltage, which is a square wave and is a result of subtracting a voltage value of the second pulse voltage from a voltage value of a third pulse voltage different from the first pulse voltage, to the third and fourth terminals, or not applying the second spike voltage to the third and fourth terminals; and a seventh step of repeatedly executing the first to sixth steps one or more times while changing the difference between the rising timing of the first pulse voltage and the rising timing of the second pulse voltage.

[0023] (Configuration 10) In configuration 9, the control circuit, in a first step, a first sub-step of applying a read voltage to the first terminal while grounding the second terminal; a second substep of applying a fourth pulse voltage to a third and a fourth terminal while grounding the first and second terminals, after the first substep; a third substep of executing the first substep after the second substep; After the third substep, a fourth substep of applying a fifth pulse voltage to the first terminal while the second terminal is grounded is performed.

[0024] (Configuration 11) In the ninth or tenth configuration, in the seventh step, the control circuit repeatedly executes the first to sixth steps one or more times while increasing the amplitude of the second spike voltage in the sixth step.

[0025] (Configuration 12) In the configuration 8, the control circuit, in the second voltage control, A first step of initializing by applying a positive voltage to the first and second terminals; a second step of applying a first write voltage to the first terminal a plurality of times after the first step; After the second step, a third step is performed in which a first write voltage is applied to the first terminal and a second write voltage different from the first write voltage is applied to the fourth terminal.

[0026] (Configuration 13) In the configuration 8, the control circuit, in the third voltage control, A first step of applying a first pulse voltage to the first terminal multiple times while the second terminal is grounded; a second step of applying a first voltage having a positive voltage value to the first terminal while the second terminal is grounded, after the first step; After the second step, a third step is performed of applying a second voltage, which is a negative voltage, to the third and fourth terminals while the second terminal is grounded.

[0027] (Configuration 14) In configuration 13, in a second step, the control circuit applies a first voltage to the first terminal with the second terminal grounded at a first frequency and at a second frequency greater than the first frequency.

[0028] (Configuration 15) In the configuration 8, the control circuit, in the fourth voltage control, a first step of applying a first pulse voltage between the third terminal and the fourth terminal while the first and second terminals are grounded; a second step of applying a second pulse voltage different from the first pulse voltage to the first terminal multiple times while the second terminal is grounded after the first step, thereby setting the conductance of the oxide semiconductor layer to a first value; a third step of repeatedly applying a second pulse voltage to the first terminal while keeping the second terminal grounded, and applying a third pulse voltage different from the first and second pulse voltages to the third and fourth terminals, or not applying the third pulse voltage to the third and fourth terminals, until the conductance of the oxide semiconductor layer reaches a second value smaller than the first value, while keeping the second terminal grounded; a fourth step of repeatedly applying a fourth pulse voltage obtained by inverting the amplitude of the second pulse voltage to the first terminal while the second terminal is grounded, until the conductance of the oxide semiconductor layer becomes a first value, after the third step; After the fourth step, a fifth step is performed in which the third and fourth steps are repeatedly performed.

[0029] (Configuration 16) In any one of the configurations 8 to 15, the oxide semiconductor layer contains less oxygen than the stoichiometric ratio and is made of single-crystal titanium oxide having a crystal face oriented in the (001) direction or amorphous gallium oxide. Effect of the Invention

[0030] It is possible to exert the higher functions of biological systems. [Brief description of the drawings]

[0031] [Figure 1] FIG. 1 is a plan view of a memristor in accordance with a first embodiment of the present invention. [Diagram 2] 2 is a cross-sectional view of the memristor taken along line II-II shown in FIG. 1. [Diagram 3] FIG. 3 is a process diagram showing a method for manufacturing the memristor shown in FIGS. [Figure 4] FIG. 3 is a plan view showing the distribution of dopants (oxygen vacancies) in the memristor shown in FIGS. [Diagram 5] FIG. 13 is a diagram showing the relationship between the write voltage and the read voltage and the number of times the voltages are applied. [Figure 6] FIG. 13 is a diagram showing another relationship between the write voltage and the read voltage and the number of times the voltage is applied. [Figure 7] FIG. 13 is a diagram showing the relationship between resistors R3-5 and R4-6 and write voltages V4 and V6. [Figure 8] FIG. 13 is a diagram showing another relationship between the resistors R3-5 and R4-6 and the write voltages V4 and V6. [Figure 9] FIG. 13 is a diagram showing still another relationship between the resistors R3-5 and R4-6 and the write voltages V4 and V6. [Figure 10] FIG. 13 is a diagram showing still another relationship between the resistors R3-5 and R4-6 and the write voltages V4 and V6. [Figure 11] 13 is a diagram showing the relationship between resistances R3-5, R4-6 and write voltages V4,6 in the memristor 10-1 for six periods. FIG. [Figure 12] FIG. 13 is a diagram showing the relationship between resistances R3-5, R4-6 and write voltages V4,6 for three periods in a memristor using titanium oxide produced by thermal reduction treatment. [Figure 13] FIG. 1 shows RHEED patterns before and after deposition of a-GaOy. [Figure 14] FIG. 13 is a diagram showing the relationship between the write voltage and the read voltage and the number of times the voltages are applied. [Figure 15] FIG. 13 is a diagram showing the relationship between resistances R3-5, R4-6 and write voltages V4,6 in the memristor 10-2. [Figure 16] FIG. 13 is a conceptual diagram showing the distribution of dopants (oxygen vacancies) in the memristor 10-2. [Figure 17] 13 is a diagram showing the relationship between the resistance value between terminals 3 and 5 and the resistance value between terminals 4 and 6 and the Ar partial pressure. FIG. [Figure 18] FIG. 13 is a graph showing the relationship between the resistance value between terminals 3 and 5 and the resistance value between terminals 4 and 6 and the film thickness of a-GaOy. [Figure 19] FIG. 13 is a graph showing the relationship between the resistance ratio RH / RL and the Ar partial pressure and the film thickness of a-GaOy. [Figure 20]FIG. 1 shows RHEED patterns before and after deposition of single crystal TiO2-n. [Figure 21] FIG. 13 is a diagram for explaining a method of generating a spike voltage. [Figure 22] FIG. 1 is a conceptual diagram showing the relationship between a spike voltage applied to a memristor and the movement of dopants (oxygen vacancies) in the memristor. [Diagram 23] FIG. 1 shows spike voltages in spike-timing-dependent plasticity STDP. [Figure 24] FIG. 1 shows the voltage application protocol when implementing spike-timing-dependent plasticity. [Diagram 25] FIG. 25 is a conceptual diagram showing the movement of dopants (oxygen vacancies) when a voltage is applied according to the voltage application protocol shown in FIG. [Figure 26] FIG. 13 is a diagram showing the relationship between the amount of change Δw and Δt in the conductance G3-5 when the Vgate spike Vgate-spike is changed. [Figure 27] FIG. 26 is a diagram showing the relationship between the amount of change Δw and Δt in the conductance G3-5 when the Vgate spike Vgate-spike is not applied (=n / a). [Figure 28] FIG. 26 is a diagram showing the relationship between the amount of change Δw and Δt in the conductance G3-5 when the amplitude V of the Vgate spike Vgate-spike is 5 V. [Figure 29] FIG. 26 is a diagram showing the relationship between the amount of change Δw and Δt in the conductance G3-5 when the amplitude V of the Vgate spike Vgate-spike is 7 V. [Diagram 30] FIG. 13 is a diagram showing the correspondence between Pavlovian conditioning and the input and output of a memristor. [Diagram 31] FIG. 1 shows the voltage application protocol when implementing Pavlovian conditioning. [Diagram 32] FIG. 1 is a first graph showing the relationship between voltage V1 corresponding to the sound of a bell, voltage V4 corresponding to food, and conductance G3-5 corresponding to the presence or absence of a reaction, and the number of pulses. [Diagram 33]2 is a second graph showing the relationship between the voltage V1 corresponding to the sound of the bell, the voltage V4 corresponding to the food, and the conductance G3-5 corresponding to the presence or absence of a reaction, and the number of pulses. FIG. [Diagram 34] 34A to 34D are diagrams showing optical microscope images in the steps shown in FIGS. 32 and 33. [Diagram 35] 34 is a conceptual diagram showing a change in distribution of dopants (oxygen vacancies) in each step shown in FIG. 32 and FIG. 33. [Diagram 36] FIG. 13 is a diagram showing the relationship between coupling strength and frequency of an input signal. [Figure 37] FIG. 1 is a conceptual diagram showing the distribution of dopants (oxygen vacancies) corresponding to initialization, enhancement, and decay. [Figure 38] FIG. 13 shows the V1 input frequency dependence of conductance G3-5. [Figure 39] FIG. 13 is a diagram showing the relationship between the number of attenuation pulses required for 10% attenuation from the maximum conductance and the gate voltage (V4,6). [Diagram 40] FIG. 1 shows voltage application protocols for synaptic sensitization and habituation. [Diagram 41] FIG. 13 is a diagram showing the relationship between conductance G3-5 and the number of pulses. [Diagram 42] FIG. 13 is a diagram showing the gate voltage dependence of the average number of pulses required for Depression (lowering conductance) and Potentiation (higher conductance). [Diagram 43] FIG. 1 is a conceptual diagram showing the relationship between the distribution of dopants (oxygen vacancies) and the gate voltage. [Diagram 44] FIG. 2 is a schematic diagram of another memristor according to the first embodiment. [Diagram 45] FIG. 1 shows RHEED patterns before and after deposition of a-GaOy. [Figure 46] FIG. 13 is a diagram showing the relationship between current and voltage in the memristor 10-3. [Figure 47] FIG. 47 is a diagram showing the relationship between the absolute value of current and voltage in the current-voltage characteristics shown in FIG. 46. [Figure 48]FIG. 1 shows spike voltages and voltage application protocols. [Figure 49] FIG. 13 is a diagram showing the relationship between the amount of change Δw and Δt in conductance G3-5. [Figure 50] FIG. 13 is a diagram showing the measurement results of the amount of change Δw in conductance G3-5. [Figure 51] FIG. 1 shows the applied spike voltage and the distribution of dopants (oxygen vacancies). [Figure 52] FIG. 13 shows another pre-spike and post-spike diagram. [Diagram 53] FIG. 53 is a diagram showing the relationship between the amount of change Δw and Δt in the conductance G3-5 when a spike voltage generated from the pre-spike and post-spike shown in FIG. 52 is applied. [Figure 54] FIG. 13 is a diagram showing further pre-spike and post-spike. [Figure 55] FIG. 55 is a diagram showing the relationship between the amount of change Δw and Δt in the conductance G3-5 when a spike voltage generated from the pre-spike and post-spike shown in FIG. 54 is applied. [Figure 56] FIG. 1 is a conceptual diagram showing an example of an application of a memristor. [Figure 57] FIG. 11 is a plan view of a memristor according to a second embodiment. [Figure 58] A cross-sectional view of the memristor taken along line XXXXXVIII-XXXXXVIII shown in Figure 57. [Figure 59] FIG. 59 is a first process diagram showing a manufacturing method of the memristor shown in FIGS. 57 and 58. [Figure 60] 59 is a second process diagram showing a method for manufacturing the memristor shown in FIGS. 57 and 58. FIG. [Figure 61] FIG. 59 is a third process diagram showing a manufacturing method of the memristor shown in FIGS. 57 and 58. [Figure 62] FIG. 59 is a fourth process diagram showing a method for manufacturing the memristor shown in FIGS. 57 and 58. [Figure 63]FIG. 59 is a fifth process diagram showing a method for manufacturing the memristor shown in FIGS. 57 and 58. [Figure 64] FIG. 11 is a plan view of another memristor according to the second embodiment. [Figure 65] A cross-sectional view of the memristor taken along line XXXXXXV-XXXXXXV shown in Figure 64. [Figure 66] FIG. 11 is a plan view of yet another memristor according to the second embodiment. [Figure 67] A cross-sectional view of the memristor taken along line XXXXXXVII-XXXXXXVII shown in Figure 66. [Figure 68] FIG. 68 is a first process diagram showing a method for manufacturing the memristor shown in FIGS. 66 and 67. [Figure 69] FIG. 68 is a second process diagram showing the method for manufacturing the memristor shown in FIGS. 66 and 67. [Figure 70] FIG. 68 is a third process diagram showing a manufacturing method of the memristor shown in FIGS. 66 and 67. [Figure 71] FIG. 68 is a fourth process diagram showing the method for manufacturing the memristor shown in FIGS. 66 and 67. [Figure 72] FIG. 68 is a fifth process diagram showing a method for manufacturing the memristor shown in FIGS. 66 and 67. [Figure 73] FIG. 68 is a sixth process diagram showing the method for manufacturing the memristor shown in FIGS. 66 and 67. [Figure 74] FIG. 68 is a seventh process diagram showing the method for manufacturing the memristor shown in FIGS. 66 and 67. [Figure 75] FIG. 11 is a plan view of yet another memristor according to the second embodiment. [Figure 76] A cross-sectional view of the memristor taken along line XXXXXXXVI-XXXXXXXVI shown in Figure 75. [Figure 77] FIG. 77 is a first process diagram showing a manufacturing method of the memristor shown in FIGS. 75 and 76. [Figure 78] FIG. 77 is a second process diagram showing the method for manufacturing the memristor shown in FIGS. 75 and 76. [Figure 79]FIG. 77 is a third process diagram showing a manufacturing method of the memristor shown in FIGS. 75 and 76. [Figure 80] FIG. 77 is a fourth process diagram showing the method for manufacturing the memristor shown in FIGS. 75 and 76. [Figure 81] FIG. 77 is a fifth process diagram showing a method for manufacturing the memristor shown in FIGS. 75 and 76. [Figure 82] FIG. 77 is a sixth process diagram showing the method for manufacturing the memristor shown in FIGS. 75 and 76. [Figure 83] 1 is a conceptual diagram of an array system according to an embodiment of the present invention; [Figure 84] FIG. 84 is a conceptual diagram showing the configuration of the first layer shown in FIG. 83. [Figure 85] FIG. 13 is a conceptual diagram of another array system according to an embodiment of the present invention. [Figure 86] FIG. 86 is a conceptual diagram showing the configuration of the first layer shown in FIG. 85. [Figure 87] FIG. 1 is a diagram showing a specific example of an array system having a planar four-terminal crossbar structure. [Figure 88] FIG. 1 is a diagram showing a specific example of an array system having a vertical four-terminal crossbar structure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0032] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are designated by the same reference characters and their description will not be repeated.

[0033] [Embodiment 1] Fig. 1 is a plan view of a memristor according to a first embodiment of the present invention, and Fig. 2 is a cross-sectional view of the memristor taken along line II-II shown in Fig. 1.

[0034] 1 and 2, a memristor 10 according to a first embodiment of the present invention includes a substrate 1, an oxide semiconductor layer 2, and terminals 3-6.

[0035] The memristor 10 has, for example, a square planar shape. The oxide semiconductor layer 2 is disposed on the substrate 1 in contact with the substrate 1. The terminals 3 to 6 are disposed on one surface of the oxide semiconductor layer 2 in contact with the oxide semiconductor layer 2. Each of the terminals 3 to 6 has, for example, a triangular planar shape. And the terminals 3 and 5 are arranged along two opposite sides of the square so that one vertex faces each other, and the terminals 4 and 6 are arranged along two opposite sides of the square so that one vertex faces each other. In this case, the direction in which the terminals 3 and 5 face is orthogonal to the direction in which the terminals 4 and 6 face.

[0036] The substrate 1 is made of single-crystalline titanium oxide ((001) plane) (single-crystalline TiO 2 ) or single-crystalline silicon.

[0037] When the substrate 1 is made of single-crystalline TiO 2 , the oxide semiconductor layer 2 is made of single-crystalline titanium oxide containing a dopant (single-crystalline TiO 2-x (0 < x < 2)). Further, when the substrate 1 is made of single-crystalline silicon, the oxide semiconductor layer 2 is made of amorphous gallium oxide containing a dopant (a-GaO y (0 < y < c, c is the stoichiometric ratio)). Thus, the oxide semiconductor layer 2 contains an oxygen amount less than the oxygen amount in single-crystalline TiO 2 or a-GaO c having a stoichiometric ratio. Therefore, the oxide semiconductor layer 2 contains oxygen vacancies as dopants.

[0038] And single-crystalline TiO 2-x and a-GaO y are materials that do not require an avalanche breakdown process. Therefore, the oxide semiconductor layer 2 is made of a material that does not require an avalanche breakdown process.

[0039] When the oxide semiconductor layer 2 is made of single-crystalline TiO 2 , each of the terminals 3 to 6 is made of, for example, Pt, and the film thickness of Pt is, for example, 50 nm. Further, when the oxide semiconductor layer 2 is a-GaO yWhen the insulating film is made of Pt, the thickness of the Pt film is, for example, 50 nm.

[0040] The substrate 1 has a thickness of, for example, 500 μm. The oxide semiconductor layer 2 is a single crystal TiO 2-x When the oxide semiconductor layer 2 is made of a-GaO y When it consists of, for example, it has a thickness of 60 nm.

[0041] 1, an XY plane is defined. Terminals 3 to 6 are terminals for applying a voltage to the oxide semiconductor layer 2. When a voltage is applied between the terminals 3 and 5, the dopants (oxygen vacancies) in the oxide semiconductor layer 2 move in the Y-axis direction, and the distribution of the dopants (oxygen vacancies) changes one-dimensionally. As a result, the resistance value of the oxide semiconductor layer 2 in the Y-axis direction changes.

[0042] Furthermore, when a voltage is applied between the terminals 4 and 6, the dopants (oxygen vacancies) in the oxide semiconductor layer 2 move in the X-axis direction, and the distribution of the dopants (oxygen vacancies) changes one-dimensionally, resulting in a change in the resistance value of the oxide semiconductor layer 2 in the X-axis direction.

[0043] Therefore, when a voltage is applied between the terminals 3 and 5 and between the terminals 4 and 6, the dopants (oxygen vacancies) in the oxide semiconductor layer 2 move in the X-axis direction and the Y-axis direction, and the distribution of the dopants (oxygen vacancies) changes two-dimensionally.

[0044] In this way, the memristor 10 can control the distribution of dopants (oxygen vacancies) two-dimensionally.

[0045] FIG. 3 is a process diagram showing a method for manufacturing the memristor 10 shown in FIGS.

[0046] Referring to FIG. 3, when the fabrication of the memristor 10 starts, a single crystal TiO 2 having a (001) surface is 2 The substrate 1 is prepared by cleaning the substrate 1 (step (a)). The cleaning method is ultrasonic cleaning using an organic solvent.

[0047] And TiO2 Single crystal TiO was deposited on the substrate 1 by pulsed laser deposition (PLD) using the target. 2-n In this case, the inside of an apparatus equipped with a Nd:YAG laser having a wavelength of 266 nm is heated to 2.0×10 -6 The oxygen partial pressure was then reduced to 2.0×10 Pa. -5 Oxygen gas was supplied to the apparatus at a temperature of 1 Pa, and a 10 Hz pulsed laser was fired at the target (TiO 2 ) was irradiated to produce single crystal TiO 2-n The film formation rate was 2.5 to 3.0 nm / min.

[0048] Thereafter, a resist is applied to the surface of the oxide semiconductor layer 2, and the applied resist is patterned by photolithography or electron beam lithography to form a resist pattern 11 (step (c)).

[0049] Then, using the resist pattern 11 as a mask, Pt is deposited on the oxide semiconductor layer 2 by a vapor deposition method to form the terminals 4 and 6 (step (d)). In this case, the terminals 3 and 5 are also formed at the same time. Pt 12 is also deposited on the resist pattern 11.

[0050] Then, the resist pattern 11 is removed. In this case, the Pt 12 deposited on the resist pattern 11 is removed by lift-off. This completes the memristor 10 (step (e)).

[0051] The oxide semiconductor layer 2 is a-GaO y Even when the memristor 10 is made of a single crystal silicon substrate, the memristor 10 is manufactured according to the process diagram shown in FIG. 3. In this case, a single crystal silicon substrate is used as the substrate 1, and in step (b), a-GaO y is formed on the substrate 1. In this case, the oxygen partial pressure is 4.0×10 -5Oxygen gas is supplied into the pulsed laser deposition apparatus so that the oxygen concentration becomes 1 Pa, and the substrate temperature is room temperature. The deposition rate is 10.0 nm / min. In step (d), a 50 nm thick Pt film is deposited by deposition on the oxide semiconductor layer 2 (a-GaO y ) are successively formed on top of each other.

[0052] FIG. 4 is a plan view showing the distribution of dopants (oxygen vacancies) in the memristor 10 shown in FIGS.

[0053] 4(a), when a positive voltage (+V) is applied to terminal 3, terminal 5 is connected to the ground potential GND, and no voltage is applied to terminals 4 and 6, an electric field is generated in the direction from terminal 3 to terminal 5, and dopants (oxygen vacancies) gather around terminal 5, but not around terminals 3, 4, and 6. As a result, a one-dimensional distribution DT1 of dopants (oxygen vacancies) distributed in the Y-axis direction is obtained.

[0054] On the other hand, when terminals 3 and 5 are connected to the ground potential GND and a positive voltage (+V) is applied to terminals 4 and 6, an electric field is generated in the directions from terminal 4 to terminal 3, from terminal 4 to terminal 5, from terminal 6 to terminal 3, and from terminal 6 to terminal 5, and dopants (oxygen vacancies) gather around terminals 3 and 5 and between terminals 3 and 5, but do not gather around terminals 4 and 6. As a result, a two-dimensional distribution DT2 of dopants (oxygen vacancies) is obtained on the XY plane (see FIG. 4(b)).

[0055] Furthermore, when terminals 3 and 5 are connected to the ground potential GND and a negative voltage (-V) is applied to terminals 4 and 6, an electric field is generated in the directions from terminal 3 to terminal 4, from terminal 3 to terminal 6, from terminal 5 to terminal 4, and from terminal 5 to terminal 6, and dopants (oxygen vacancies) gather around terminals 4 and 6 and between the ends 4 and 6, but do not gather around terminals 3 and 5. As a result, a two-dimensional distribution DT3 of dopants (oxygen vacancies) is obtained on the XY plane (see FIG. 4(c)).

[0056] In this way, by controlling the voltages applied to terminals 3 to 6, the distribution of dopants (oxygen vacancies) in memristor 10 can be controlled two-dimensionally.

[0057] The substrate 1 is a single crystal TiO having a (001) surface. 2 (thickness: 0.5 mm) was used to fabricate single crystal TiO 2-x The memristor 10-1 was fabricated by forming single crystal TiO 2-x The film thickness is 60 nm.

[0058] Then, TiO formed on the substrate 1 by pulsed laser deposition 2-x It was confirmed by RHEED (Reflection High Energy Electron Diffraction) pattern that it was a single crystal.

[0059] 5 is a diagram showing the relationship between the write voltage and the read voltage and the number of times the voltage is applied. 4,6 The relationship between the read voltage V and the number of times the voltage is applied is shown in FIG. 1 ,V 2 The relationship between the number of times the voltage is applied and the number of times the voltage is applied is shown.

[0060] Referring to FIG. 5(a), voltages of 1V, 2V, . . . , 8V are applied between terminals 4 and 6 to generate a write voltage V 4,6 is applied to the memristor 10-1. 4,6 The application time of each write voltage V is, for example, 100 seconds. 4,6 is applied between terminals 4 and 6 of the memristor 10-1, 2-x The distribution of dopants (oxygen vacancies) is controlled two-dimensionally, and the write voltage V 4,6 The two-dimensional distribution of dopants (oxygen vacancies) varies depending on the voltage value.

[0061] Also, apply a voltage of 1V between terminals 3 and 5 to obtain the read voltage V 1 (○) is applied to the memristor 10-1. Read voltage V 1The application time of (○) is, for example, 1 to 2 seconds. 1 Current I between terminals 3 and 5 when (○) is applied 3-5 Measure the resistance R between terminals 3 and 5. 3-5 Similarly, apply a voltage of 1V between terminals 4 and 6 to obtain the read voltage V 2 (●) is applied to the memristor 10-1. Read voltage V 2 The application time of (●) is, for example, 1 to 2 seconds. And, the read voltage V 2 Current I between terminals 4 and 6 when (●) is applied 4-6 Measure the resistance R between terminals 4 and 6. 4-6 Request.

[0062] Each write voltage V 4,6 Applying the voltage for 100 seconds and reading out the voltage V 1 (○) and read voltage V 2 Applying (●) causes a current I 3-5 and current I 4-6 The operation of measuring the voltage V 4,6 When is applied to the memristor 10-1, the resistance R 3-5 and resistor R 4-6 Request.

[0063] 6 is a diagram showing another relationship between the write voltage and the read voltage and the number of times the voltage is applied. 4,6 The relationship between the read voltage V and the number of times the voltage is applied is shown in FIG. 1 ,V 2 The relationship between the number of times the voltage is applied and the number of times the voltage is applied is shown.

[0064] With reference to (a) of FIG. 6, for example, a write voltage V 4,6 The write voltage V is applied between terminals 4 and 6 of memristor 10-1 in the following order for 100 seconds each: 0V, 1V, 2V, 3V, 4V, 5V, 6V, 7V, 8V, 7V, 6V, 5V, 4V, 3V, 2V, 1V, 0V, -1V, -2V, -3V, -4V, -5V, -6V, -7V, -8V, -7V, -6V, -5V, -4V, -3V, -2V, -1V, 0V. 4,6The write voltage V 4,6 is applied to memristor 10-1.

[0065] The read voltage V shown in FIG. 1 ,V 2 The explanation for each write voltage V is the same as that in FIG. 4,6 Applying the voltage for 100 seconds and reading out the voltage V 1 (○) and read voltage V 2 Applying (●) causes a current I 3-5 and current I 4-6 The operation of measuring the voltage V 4,6 When is applied to the memristor 10-1, the resistance R 3-5 and resistor R 4-6 Request.

[0066] Figure 7 shows the resistor R 3-5 ,R 4-6 and the write voltage V 4,6 FIG. 8 is a diagram showing the relationship between the resistance R 3-5 ,R 4-6 and the write voltage V 4,6 FIG. 9 is a diagram showing another relationship between the resistance R 3-5 ,R 4-6 and the write voltage V 4,6 FIG. 10 is a diagram showing yet another relationship between the resistance R 3-5 ,R 4-6 and the write voltage V 4,6 FIG. 13 is a diagram showing yet another relationship between

[0067] In Fig. 7 to Fig. 10, (a) shows the resistance R 3-5 and the write voltage V 4,6 (b) shows the relationship between the resistance R 4-6 and the write voltage V 4,6 This shows the relationship between:

[0068] Referring to FIG. 7, the write voltages V 4,6 is applied between terminals 4 and 6 in sequence, and resistance R 3-5 and resistor R 4-6 As a result, the resistance R 3-5is the write voltage V 4,6 It decreases with increasing resistance R 4-6 is the write voltage V 4,6 increases with increasing (see curve k2).

[0069] Referring to FIG. 8, a write voltage V 4,6 is applied between terminals 4 and 6, and then the write voltages V 4,6 is applied between terminals 4 and 6 in sequence, and resistance R 3-5 and resistor R 4-6 Request.

[0070] As a result, the resistance R 3-5 is the write voltage V 4,6 When the write voltage V is 8V, 7V, 6V, 5V, and 4V, it changes in the same way as the curve k1. 4,6 When the write voltage V is 3V, 2V, 1V, 0V, -1V, -2V, -3V, -4V, -5V, -6V, -7V, and -8V, the write voltage V changes according to a curve k3 different from the curve k1. 4,6 When the write voltage V 4,6 It increases as the decrease in (see curve k3).

[0071] On the other hand, resistor R 4-6 is the write voltage V 4,6 As the write voltage V 4,6 When the write voltage V 4,6 When the write voltage V 4,6 When the write voltage V 4,6 When it decreases in the range of -6V to -8V, it becomes a constant value (see curve k4).

[0072] Referring to Figure 9, a write voltage V 4,6 After the application of -7V, -6V, -5V, -4V, -3V, -2V, -1V, 0V, 1V, 2V, 3V, 4V, 5V, 6V, 7V, 8V write voltage V4,6 is applied between terminals 4 and 6 in sequence, and resistance R 3-5 and resistor R 4-6 Request.

[0073] As a result, the resistance R 3-5 is the write voltage V 4,6 As the voltage increases to -8V, -7V, -6V, and -5V, the write voltage V 4,6 As the voltage increases to -5V, -4V, and -3V, the write voltage V 4,6 As the voltage increases to -3V, -2V, -1V, 0V, 1V, 2V, 3V, and 4V, it gradually decreases and the write voltage V 4,6 When the write voltage V 4,6 As it increases to 6V, 7V, and 8V, it remains constant (see curve k5).

[0074] On the other hand, resistor R 4-6 is the write voltage V 4,6 remains constant as the write voltage V increases from -8V to 5V. 4,6 increases as the write voltage V increases from 5V to 8V (see curve k6). In this case, the second write voltage V 4,6 When the resistance R 4-6 (The resistance R when changed according to the curve k6 4-6 ) is the first write voltage V 4,6 When the resistance R 4-6 (The resistance R when changed according to the curve k2 4-6 ).

[0075] 10, further, the write voltages V of 8V, 7V, 6V, 5V, 4V, 3V, 2V, 1V, 0V, -1V, -2V, -3V, -4V, -5V, -6V, -7V, -8V, -7V, -6V, -5V, -4V, -3V, -2V, -1V, 0V, 1V, 2V, 3V, 4V, 5V, 6V, 7V, and 8V are 4,6 is sequentially applied to the memristor 10-1 (the second cycle of writing voltage V 4,6 When a resistor R 3-5 changes according to the curves k3 and k5, and the resistance R 4-6 is the resistance R shown by curve k44-6 and the write voltage V 4,6 It varies according to a curve k7 having approximately the same relationship as that of

[0076] And the resistor R 3-5 and the write voltage V 4,6 The shape of the curve showing the relationship between 4-6 and the write voltage V 4,6 The shape of the curve showing the relationship between is symmetrical (see FIG. 10).

[0077] Thus, single crystal TiO 2-x A write voltage V 4,6 When a resistor R 3-5 ,R 4-6 Write voltage V 4,6 The dependence changes to hysteresis. 3-5 ,R 4-6 is the applied write voltage V 4,6 Therefore, the write voltage V 4,6 Single crystal TiO 2-x The two-dimensional distribution of dopants (oxygen vacancies) can be controlled to various two-dimensional distributions.

[0078] FIG. 11 shows the resistance R 3-5 ,R 4-6 and the write voltage V 4,6 FIG. 12 shows the relationship between the resistance R and the resistance R of the memristor Comp using titanium oxide fabricated by thermal reduction treatment for six periods. 3-5 ,R 4-6 and the write voltage V 4,6 FIG. 13 is a diagram showing the relationship between the

[0079] In FIG. 11 and FIG. 12, (a) shows the resistance R 3-5 and the write voltage V 4,6 (b) shows the relationship between the resistance R 4-6 and the write voltage V 4,6 This shows the relationship between:

[0080] Referring to FIG. 11, the memristor 10-1 according to the first embodiment has a resistor R 3-5 Write voltage V 4,6 Even when the dependence measurement is repeated for six periods, stable characteristics are observed.

[0081] On the other hand, the memristor Comp using titanium oxide fabricated by thermal reduction treatment has a resistance R 3-5 Write voltage V 4,6 The dependence measurement was repeated for only three periods and was found to be very unstable.

[0082] In addition, the memristor 10-1 according to embodiment 1 has a very large ratio of high resistance value to low resistance value of 20 times, whereas the memristor Comp using titanium oxide produced by thermal reduction treatment has a very small ratio of high resistance value to low resistance value of 2 to 3 times.

[0083] Thus, the memristor 10-1 according to the first embodiment has a remarkable effect in that the ratio of the high resistance value to the low resistance value is one order of magnitude larger than that of the memristor Comp using titanium oxide produced by thermal reduction treatment.

[0084] a-GaO y As for the memristor 10-2 using the oxide semiconductor layer 2, the resistance R 3-5 ,R 4-6 and the write voltage V 4,6 We experimented with the relationship between

[0085] The memristor 10-2 used in the experiment had a single crystal silicon (001) / SiO 2 Using single crystal silicon (001) / SiO 2 SiO 2 a-GaO by pulsed laser deposition y Formation of a-GaO y The structure has Pt vapor-deposited on the surface as terminals 3 to 6.

[0086] 3, a memristor 10-2 was fabricated. In this case, in step (b), argon (Ar) gas was used instead of oxygen gas, and the partial pressure of the Ar gas was set to 4.0×10 -3 Pa, 4.0×10 -2 Pa, 4.0×10 -1 Pa and 8.0 × 10 -1 Pa. Also, a-GaO y The film thickness was changed to 50 nm, 90 nm, and 120 nm, and the substrate temperature was room temperature.

[0087] Figure 13 shows a-GaO y 13A and 13B are diagrams showing RHEED patterns before and after the formation of a-GaO y (b) shows the RHEED pattern of a-GaO y RHEED patterns after the deposition of the above film are shown.

[0088] Referring to FIG. 13, a-GaO y Before the deposition of the film, the RHEED pattern is halo-shaped (see (a)). This is because the substrate 1 is single-crystal silicon (001) / SiO 2 SiO 2 This is thought to be the reason why the RHEED pattern appears halo-like.

[0089] a-GaO y Even after the deposition of the GaO y It was confirmed that the material had an amorphous structure.

[0090] 14 is a diagram showing the relationship between the write voltage and the read voltage and the number of times the voltage is applied. 4,6 FIG. 14(b) shows the relationship between the read voltage V 1 ,V 2 The relationship between the number of times the voltage is applied and the number of times the voltage is applied is shown.

[0091] Referring to FIG. 14(a), a write voltage V4,6 are applied between terminals 4 and 6 of memristor 10-2 for 100 seconds each in the following order: 1V, 2V, 3V, 4V, 5V, 6V, 7V, 8V, 9V, 10V, 11V, 12V, 13V, 14V, 13V, 12V, 11V, 10V, 9V, 8V, 7V, 6V, 5V, 4V, 3V, 2V, 1V, 0V, -1V, -2V, -3V, -4V, -5V, -6V, -7V, -8V, -9V, -10V, -11V, -12V, -13V, -14V, -13V, -12V, -11V, -10V, -9V, -8V, -7V, -6V, -5V, -4V, -3V, -2V, -1V, 0V. That is, the voltage value is changed by 1V to obtain the write voltage V 4,6 The write voltage V is set so that it goes through a cycle in the range of -14V to +14V. 4,6 is applied to memristor 10-2.

[0092] Each write voltage V 4,6 The application time of each write voltage V is, for example, 100 seconds. 4,6 is applied between terminals 4 and 6 of the memristor 10-2, y The distribution of dopants (oxygen vacancies) is controlled two-dimensionally, and the write voltage V 4,6 The two-dimensional distribution of dopants (oxygen vacancies) varies depending on the voltage value.

[0093] Also, apply a voltage of 1V between terminals 3 and 5 to obtain the read voltage V 1 (○) is applied to the memristor 10-2. Read voltage V 1 The application time of (○) is, for example, 1 to 2 seconds. 1 Current I between terminals 3 and 5 when (○) is applied 3-5 Measure the resistance R between terminals 3 and 5. 3-5 Similarly, apply a voltage of 1V between terminals 4 and 6 to obtain the read voltage V 2 (●) is applied to the memristor 10-2. Read voltage V 2 The application time of (●) is, for example, 1 to 2 seconds. And, the read voltage V 2 Current I between terminals 4 and 6 when (●) is applied 4-6 Measure the resistance R between terminals 4 and 6. 4-6 Request.

[0094] Each write voltage V 4,6 Applying the voltage for 100 seconds and reading out the voltage V 1 (○) and read voltage V 2 Applying (●) causes a current I 3-5 and current I 4-6 The operation of measuring the voltage V 4,6 Resistance R when 3-5 and resistor R 4-6 Request.

[0095] FIG. 15 shows the resistance R 3-5 ,R 4-6 and the write voltage V 4,6 In FIG. 15, (a) shows the relationship between the resistance R 3-5 and the write voltage V 4,6 (b) shows the relationship between the resistance R 4-6 and the write voltage V 4,6 The relationship between the 4.0×10 -1 A 90 nm thick a-GaO film was formed at an Ar partial pressure of 1.0 Pa. y Resistance R in the memristor 10-2 using 3-5 ,R 4-6 and the write voltage V 4,6 This shows the relationship between:

[0096] Referring to FIG. 15(a), resistor R 3-5 is the write voltage V 4,6 As the write voltage V increases from 0V to 2V, 4,6 As the write voltage V 4,6 As the voltage increases from 7 V to 14 V, it gradually decreases (see arrow I in (a)).

[0097] Then, resistor R 3-5 is the write voltage V 4,6 As the voltage decreases from 14 V to -6 V, it remains almost constant (see arrow II in (a)). And, the resistance R 3-5 is the write voltage V 4,6When V decreases from -6V to -10V, the write voltage V 4,6 When the voltage decreases from -10 V to -14 V, it becomes almost constant (see arrow III in (a)).

[0098] Next, resistor R 3-5 is the write voltage V 4,6 As the write voltage V increases from -14V to -1V, it becomes almost constant. 4,6 As the voltage V increases from -1V to 0V, the write voltage V 4,6 is set to 0 V (see arrow IV in (a)).

[0099] Thus, the resistor R 3-5 is the write voltage V 4,6 It changes clockwise as it goes around 0V→14V→-14V→0V.

[0100] Referring to FIG. 15(b), resistor R 4-6 is the write voltage V 4,6 As the write voltage V increases from 0V to 7V, it becomes almost constant. 4,6 As the write voltage V increases from 7V to 10V, 4,6 As the voltage increases from 10 V to 14 V, it becomes almost constant (see arrow I in (b)).

[0101] Then, resistor R 4-6 is the write voltage V 4,6 is almost constant as the write voltage V 4,6 It gradually increases as the voltage decreases from 1 V to -2 V (see arrow II in (b)).

[0102] Next, resistor R 4-6 is the write voltage V 4,6 As the write voltage V 4,6 It gradually decreases as the voltage decreases from -8V to -14V (see arrow III in (b)).

[0103] Then, resistor R 4-6 is the write voltage V4,6 As the write voltage V increases from -14V to 0V, it becomes almost constant. 4,6 is set to 0 V (see arrow IV in (b)).

[0104] Thus, the resistor R 4-6 is the write voltage V 4,6 It changes counterclockwise as it goes around 0V → 14V → -14V → 0V.

[0105] Resistance R 3-5 ,R 4-6 and the write voltage V 4,6 The relationship between these factors can be summarized as shown in Table 1.

[0106] [Table 1]

[0107] As shown in Table 1, the write voltage V 4,6 When rises from 0V to 14V, resistor R 3-5 decreases, and the resistance R 4-6 Also, the write voltage V 4,6 When decreases from 14V to 0V, resistor R 3-5 and resistor R 4-6 is almost constant. Furthermore, the write voltage V 4,6 When decreases from 0V to -14V, resistor R 3-5 increases, and the resistance R 4-6 Furthermore, the write voltage V 4,6 When increases from -14V to 0V, resistor R 3-5 and resistor R 4-6 is almost constant.

[0108] Therefore, the resistance R 3-5 and resistor R 3-5 is the write voltage V 4,6 When the write voltage V 4,6 When decreases from 0V to -14V, the mutually inverse write voltage V 4,6 Since they exhibit dependence, they vary according to curves having mutually symmetric shapes.

[0109] Fig. 16 is a conceptual diagram showing the distribution of dopants (oxygen vacancies) in the memristor 10-2. (a), (b), (c), and (d) in Fig. 16 show the distribution of dopants (oxygen vacancies) corresponding to the arrows I, II, III, and IV shown in (a) and (b) in Fig. 15, respectively.

[0110] Referring to FIG. 16(a), the write voltage V 4,6 As increases, dopants (oxygen vacancies) are distributed around the terminals 3 and 5. The result is a resistance R 3-5 is the write voltage V 4,6 decreases with increasing resistance R 4-6 is the write voltage V 4,6 It increases with increasing (see arrow I in Figures 15(a) and (b)).

[0111] Referring to FIG. 16(b), the write voltage V 4,6 Even if the resistance R decreases, the dopants (oxygen vacancies) will continue to be distributed around the terminals 3 and 5 for a while. 3-5 ,R 4-6 is the write voltage V 4,6 However, since the dopants (oxygen vacancies) are distributed around the terminals 3 and 5, the length of the region where no dopants (oxygen vacancies) exist between the terminals 3 and 5 is shorter than the length of the region where no dopants (oxygen vacancies) exist between the terminals 4 and 6. Therefore, the resistance R 3-5 is the resistance R 4-6 (See arrows II in (a) and (b) of Figure 15).

[0112] Referring to FIG. 16(c), the write voltage V 4,6 As the resistance R increases, the dopants (oxygen vacancies) move to the vicinity of terminals 4 and 6. 3-5 is the write voltage V 4,6 increases with increasing resistance R 4-6 is the write voltage V 4,6 It decreases as the increase in (see arrow III in Figures 15(a) and (b)).

[0113] Referring to FIG. 16(d), the write voltage V 4,6 Even if increases, dopants (oxygen vacancies) are distributed around terminals 4 and 6 for a while. As a result, the resistance R 3-5 ,R 4-6 is the write voltage V 4,6 As the resistance R increases, it remains almost constant (see arrows IV in (a) and (b) of FIG. 15). However, since the dopants (oxygen vacancies) are distributed around the terminals 4 and 6, the length of the region where no dopants (oxygen vacancies) exist between the terminals 4 and 6 is shorter than the length of the region where no dopants (oxygen vacancies) exist between the terminals 3 and 5. 4-6 is the resistance R 3-5 (See arrow IV in (a) and (b) of Figure 15).

[0114] FIG. 17 is a diagram showing the relationship between the resistance value between the terminals 3 and 5 and the resistance value between the terminals 4 and 6 and the Ar partial pressure.

[0115] In FIG. 17, a-GaO y The Ar partial pressure when forming the film was 4.0×10 -3 Pa, 4.0×10 -2 Pa, 4.0×10 -1 Pa, 8.0×10 -1 In addition, the oxide semiconductor layer 2 was changed to a-GaO y The film thickness was set to 90 nm.

[0116] Referring to Figure 17, the resistance between terminals 3 and 5 is R 3-5 is Ar partial pressure 4.0×10 -3 Pa, 4.0×10 -2 Pa, 4.0×10 -1 Pa, 8.0×10 -1 When the Ar partial pressure was changed to 1 Pa, the write voltage V 4,6 changes clockwise as it goes from 0V to 14V to -14V to 0V. Meanwhile, the resistance R between terminals 4 and 6 4-6 is Ar partial pressure 4.0×10 -3 Pa, 4.0×10 -2 Pa, 4.0×10 -1 Pa, 8.0×10-1 When the Ar partial pressure was changed to 1 Pa, the write voltage V 4,6 It changes counterclockwise as it goes around 0V → 14V → -14V → 0V.

[0117] And the resistance between terminals 3 and 5 is R 3-5 and the resistance between terminals 4 and 6, R 4-6 varies depending on the Ar partial pressure. This is believed to be because the density of dopants (oxygen vacancies) in the memristor 10-2 increases with an increase in the Ar partial pressure. In addition, the resistance R between terminals 3 and 5 3-5 and the resistance between terminals 4 and 6, R 4-6 is Ar partial pressure 4.0×10 -1 The best characteristics are observed when the tensile strength is 100 Pa.

[0118] Figure 18 shows the resistance between terminals 3 and 5 and between terminals 4 and 6, and the a-GaO y FIG.

[0119] In FIG. 18, a-GaO y The Ar partial pressure when forming the film was 4.0×10 -1 This is because the resistance between terminals 3 and 5 is set to R 3-5 and the resistance between terminals 4 and 6, R 4-6 is Ar partial pressure 4.0×10 -1 In addition, when the a-GaO y The film thickness was varied to 50 nm, 90 nm and 120 nm.

[0120] Referring to Figure 18, the resistance between terminals 3 and 5 is R 3-5 a-GaO y When the film thickness was changed to 50 nm, 90 nm, and 120 nm, the write voltage V 4,6 changes clockwise as it goes from 0V to 14V to -14V to 0V. Meanwhile, the resistance R between terminals 4 and 6 4-6 a-GaO yWhen the film thickness is changed to 50 nm, 90 nm, and 120 nm, the voltage V4, 6 changes counterclockwise as the write voltage V4, 6 goes through a cycle of 0V→14V→−14V→0V for all film thicknesses.

[0121] And the resistance between terminals 3 and 5 is R 3-5 and the resistance between terminals 4 and 6, R 4-6 a-GaO y This depends on the thickness of the a-GaO y This is believed to be due to the increase in the thickness of the film, which increases the amount of dopants (oxygen vacancies). 3-5 and the resistance between terminals 4 and 6, R 4-6 a-GaO y When the film thickness is 90 nm, the best characteristics are obtained.

[0122] Figure 19 shows the resistance ratio R H / R L and Ar partial pressure and a-GaO y FIG.

[0123] FIG. 19(a) shows the resistance ratio R H / R L and Ar partial pressure, and (b) of FIG. 19 shows the relationship between the resistance ratio R H / R L and a-GaO y The relationship between the resistance ratio R H / R L In the characteristics shown in Figures 17 and 18, the resistance value R 3-5 and resistance value R 4-6 is the ratio of the resistance values ​​in the region where

[0124] In FIG. 19(a), the vertical axis represents the resistance ratio R H / R L The horizontal axis represents the Ar partial pressure. y The film thickness of the film is 90 nm. In FIG. 19(b), the vertical axis is the resistance ratio R H / R L The horizontal axis represents a-GaO y The Ar partial pressure is 4.0×10-1 It is Pa.

[0125] Referring to FIG. 19, the resistance ratio R H / R L is Ar partial pressure 4.0×10 -1 When Pa, a-GaO y It was found that the maximum value was achieved when the film thickness of a-GaO was 90 nm. y In the case of a-GaO, the dopants (oxygen vacancies) are y This is believed to be because it varies depending on the film thickness.

[0126] In addition, the characteristics shown in Figure 19 show a high resistance ratio R H / R L Ar partial pressure and a-GaO y There exists an optimum condition for the film thickness, and we were able to obtain an index for this.

[0127] As described above, the memristor 10-2 has a resistance R 3-5 ,R 4-6 Write voltage V 4,6 Since the dependence shows hysteresis, the write voltage V 4,6 Even if the voltage value of resistor R 3-5 ,R 4-6 holds two different resistance values. That is, the write voltage V 4,6 The distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 2 becomes a different two-dimensional distribution depending on the voltage value of . Therefore, by using the memristor 10-2, two different states can be expressed.

[0128] Therefore, a-GaO y As for the memristor 10-2 using the oxide semiconductor layer 2, the resistance R 3-5 ,R 4-6 and the write voltage V 4,6 The relationship between resistance and R is similar. 3-5 ,R 4-6 and the write voltage V 4,6 (See Figs. 15, 17 and 18).

[0129] As described above, the memristor 10 has a resistance R 3-5 ,R 4-6 Write voltage V 4,6 Since the dependence shows hysteresis, the write voltage V 4,6 Even if the voltage value of resistor R 3-5 ,R 4-6 holds two different resistance values. That is, the write voltage V 4,6 The distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 2 becomes a different two-dimensional distribution depending on the voltage value. Therefore, the memristor 10 can be used to represent two different states.

[0130] When the memristor 10 is used as a synapse, the resistance R 3-5 (or resistor R 4-6 ) decreases (increases), the synaptic connections have increased (decreased) in response to the input signal, and learning is complete.

[0131] In the memristor 10, signals are input between terminals 3 and 5 and between terminals 4 and 6. The dopants (oxygen vacancies) in the oxide semiconductor layer 2 are affected by the electric field caused by the voltage applied between terminals 4 and 6, and therefore the distribution of the dopants (oxygen vacancies) formed is different compared to when a voltage is applied between terminals 3 and 5 alone. In other words, the resistance between terminals 3 and 5, i.e., the weight of the synaptic connection, is different, and this can be regarded as the result of "associative learning" between the signal input to terminal 3 and the signal input to terminal 4.

[0132] Therefore, by using the memristor 10, it is possible to realize the function of associative learning, and to exert the higher-order functions of biological systems.

[0133] The synapse characteristics realized by using the memristor 10 shown in Figures 1 and 2 will be described. In this case, the sample was fabricated according to steps (a) to (e) shown in Figure 3. In step (b), a single crystal TiO 2 Single crystal TiO 2-n The oxygen partial pressure during deposition was 3.0×10 -5The laser power was 70 mJ and the single crystal TiO 2 The thickness of the single crystal TiO 2-n The film thickness was 50 nm, and the substrate temperature and film formation rate were as described in the above step (b).

[0134] Figure 20 shows single crystal TiO 2-n 20A and 20B are diagrams showing RHEED patterns before and after the deposition of a single-crystal TiO 2-n (b) shows the RHEED pattern of single-crystal TiO 2-n RHEED patterns after the deposition of the above film are shown.

[0135] As shown in FIG. 20(b), after the film formation, a spot pattern was observed, and the single crystal TiO 2-n Epitaxial growth was confirmed.

[0136] Fig. 21 is a diagram for explaining a method of generating spike voltages. With reference to Fig. 21, various spike voltages are generated by changing the difference Δt between the rising timing of a pre-spike consisting of a square wave and the rising timing of a post-spike consisting of a square wave, and subtracting the post-spike from the pre-spike.

[0137] FIG. 22 is a conceptual diagram showing the relationship between a spike voltage applied to a memristor and the movement of dopants (oxygen vacancies) in the memristor.

[0138] In Figure 22, (a) shows a spike voltage SPK1 generated by subtracting post-spike from pre-spike when Δt is -25 ms as shown in Figure 21, and (b) shows a spike voltage SPK2 generated by subtracting post-spike from pre-spike when Δt is +25 ms as shown in Figure 21.

[0139] In addition, in Figure 22, (c) shows the movement of dopants (oxygen vacancies) when a spike voltage SPK1 is applied to terminal 3 with terminal 5 of memristor 10 being grounded, (d) shows the movement of dopants (oxygen vacancies) when a spike voltage SPK2 is applied to terminal 3 with terminal 5 of memristor 10 being grounded, (e) shows the movement of dopants (oxygen vacancies) when a spike voltage SPK1 is applied to terminals 3, 4, and 6 with terminal 5 of memristor 10 being grounded, and (f) shows the movement of dopants (oxygen vacancies) when a spike voltage SPK2 is applied to terminals 3, 4, and 6 with terminal 5 of memristor 10 being grounded.

[0140] 22, when a spike voltage SPK1 is applied to the terminal 3 with the terminal 5 of the memristor 10 being grounded, the dopant (oxygen vacancy) moves from the terminal 3 to the periphery of the terminal 5 (see (c)). When a spike voltage SPK2 is applied to the terminal 3 with the terminal 5 of the memristor 10 being grounded, the dopant (oxygen vacancy) moves from the terminal 5 to the periphery of the terminal 3 (see (d)). When a spike voltage SPK1 is applied to the terminals 3, 4, and 6 with the terminal 5 of the memristor 10 being grounded, the dopant (oxygen vacancy) moves from the terminals 4 and 6 to the periphery of the terminal 5 (see (e)). When a spike voltage SPK2 is applied to the terminals 3, 4, and 6 with the terminal 5 of the memristor 10 being grounded, the dopant (oxygen vacancy) moves from the terminal 5 to the terminals 3, 4, and 6 (see (f)).

[0141] In this way, by controlling the voltage applied to terminals 3, 4, 5, and 6, single crystal TiO 2-n It is possible to control the distribution of dopants (oxygen vacancies) in the 3-5 changes.

[0142] In (c) and (d) of FIG. 22, no voltage is applied to the terminals 4 and 6, and in (e) and (f) a voltage is applied to the terminals 4 and 6. The movement of the dopant (oxygen vacancy) shown in (c) and (d) is different from the movement of the dopant (oxygen vacancy) shown in (e) and (f), so the conductance G between the terminals 3 and 5 3-5 also changes depending on the voltage applied to terminals 4 and 6. Therefore, the voltage applied to terminals 4 and 6 is called the gate voltage V gate Then, the gate voltage V gate By the conductance G 3-5 It is also possible to control the

[0143] [Spike-timing-dependent plasticity] We will now explain spike timing dependent plasticity (STDP) as a synaptic characteristic implemented in the memristor 10 shown in Figures 1 and 2. Here, spike timing dependent plasticity refers to the memristor 10 having a different conductance G depending on the difference Δt between the rising timing of the pre-spike and the rising timing of the post-spike.

[0144] FIG. 23 shows spike voltages in spike-timing-dependent plasticity STDP.

[0145] 23, the pre-spike is a square wave having an amplitude of 7V and an amplitude width of 25ms. The post-spike is a square wave having the same amplitude as the pre-spike, and the rising timing of the post-spike differs from that of the pre-spike by Δt. The voltage value of the post-spike is subtracted from the voltage value of the pre-spike to obtain V 1 -spike is generated (see (a) of FIG. 23). In this case, Δt is changed within the range of -50 ms to +50 ms, for example.

[0146] The gate spike is a square wave with an amplitude of 1 to 7 V and an amplitude width of 25 ms. The voltage value of the post-spike, whose rising timing differs from the rising timing of the gate spike by Δt, is subtracted from the voltage value of the gate spike to obtain V gate -spike is generated (see FIG. 23(b)). In this case, Δt is V gate The maximum voltage of the -spike is set to a range that is negative.

[0147] Fig. 24 is a diagram showing a voltage application protocol when implementing spike-timing-dependent plasticity. The following steps (i) to (viii) will be described with reference to Fig. 24.

[0148] (i) Initial conductance G in memristor 10 3-5 To measure, a read voltage Read having an amplitude of 100 mV and an amplitude width of 2 ms is applied to terminal 3 with terminal 5 grounded.

[0149] (ii) After step (i), in order to accumulate dopants (oxygen vacancies) between terminals 3-5, a pulse voltage having an amplitude of 12 V and an amplitude width of 50 s is applied to terminals 4 and 6 while terminals 3 and 5 are grounded.

[0150] (iii) After step (ii), the conductance G 3-5 To measure the change in Vcc, a read voltage Read is applied to terminal 3 with terminal 5 grounded.

[0151] (iv) After step (iii), a pulse voltage having an amplitude in the range of ±1 to 20 V and an amplitude width in the range of 0.1 to 10 s is applied to terminal 3 multiple times while terminal 5 is grounded to measure the conductance G 3-5 To be more specific, a pulse voltage having an amplitude in the range of ±1 V and an amplitude width of 3 s is used, and the amplitude is increased by 0.2 V for each application of the pulse voltage, to obtain a conductance G 3-5To reduce the conductance G, change the amplitude to 1V, 1.2V, 1.4V, etc. 3-5 To increase the conductance G, change the amplitude from -1V, -1.2V, -1.4V, ... to ±16V. 3-5 Initialize to a desired value. In this experiment, for example, the conductance G 3-5 Initialize to 2mS.

[0152] (v) After step (iv), with terminal 5 grounded, V 1 Spike V 1 Apply -spike (Spike1) to terminal 3. In this case, V 1 Spike V 1 Before and after the application of the -spike, a read voltage Read is applied to terminal 3 with terminal 5 grounded, and V 1 Spike V 1 Conductance G before and after application of -spike (Spike1) 3-5 Measure the conductance G 3-5 The amount of change Δw is calculated using the following formula.

[0153]

number

[0154] (vi) After step (v), the conductance G 3-5 Initialize to the desired value (=2mS).

[0155] (vii) After step (vi), with terminal 5 grounded, V 1 Spike V 1 -spike (Spike1) is applied to terminal 3 and V gate Spike V gate Apply -spike (Spike1') to terminals 4 and 6. In this case, V gate Spike V gate Before and after applying the -spike (Spike1'), a read voltage Read is applied to terminal 3 with terminal 5 grounded, and V gate Spike Vgate Conductance G before and after application of -spike (Spike1') 3-5 Measure the conductance G 3-5 The amount of change Δw is calculated using equation (1).

[0156] (viii) After step (vii), the conductance G 3-5 Initialization of and conductance G 3-5 The calculation of the amount of change Δw is repeated.

[0157] FIG. 25 is a conceptual diagram showing the movement of dopants (oxygen vacancies) when a voltage is applied according to the voltage application protocol shown in FIG.

[0158] Referring to FIG. 25(a), when a voltage is applied as described in step (ii) above, dopants (oxygen vacancies) are accumulated between terminals 3-5.

[0159] Also, referring to FIG. 25(b), when various pulse voltages are applied to the terminal 3 while the terminal 5 is grounded as described in the above step (iv), the conductance G 3-5 is initialized to the desired value.

[0160] Furthermore, referring to FIG. 25(c), as described in the above step (v), with terminal 5 grounded, V 1 Spike V 1 When a -spike is applied to terminal 3, dopants (oxygen vacancies) move between terminals 3 and 5, causing V 1 Spike V 1 Conductance G before and after the application of the -spike 3-5 changes.

[0161] Furthermore, referring to FIG. 25(d), as described in the above step (vii), with terminal 5 grounded, V 1 Spike V 1 -spike (Spike1) is applied to terminal 3 and V gate Spike Vgate When a -spike (Spike1') is applied to terminals 4 and 6, dopants (oxygen vacancies) move between terminal 5 and terminals 3, 4, and 6, causing V gate Spike V gate Conductance G before and after the application of -spike (Spike1') 3-5 changes.

[0162] Figure 26 shows the V gate Spike V gate Conductance G when -spike is changed 3-5 26 is a graph showing the relationship between the amount of change Δw in the conductance G 3-5 The horizontal axis represents the change in Δw, and the horizontal axis represents Δt. gate Spike V gate When no -spike is applied (=n / a), and V gate Spike V gate When the amplitude V of the -spike is 1V, 2V, 3V, 4V, 5V, 6V, and 7V, the conductance G 3-5 The relationship between the amount of change Δw and Δt is shown.

[0163] Referring to Figure 26, V gate Spike V gate By changing the -spike, the conductance G 3-5 The relationship between the change in Δw and Δt changes. For example, V gate Spike V gate When no -spike is applied (=n / a), the conductance G 3-5 The change in Δw is minimum when Δt is -25 ms and maximum when Δt is +25 ms (see black circle). gate Spike V gate When the amplitude V of the -spike is 7V, the conductance G 3-5 The change Δw in is maximum when Δt is -25 ms and is minimum when Δt is +25 ms (see white triangle).

[0164] FIG. 27 shows the V gate Spike V gateConductance G when no -spike is applied (=n / a) 3-5 27 is a graph showing the relationship between the amount of change Δw in the conductance G and Δt. 3-5 The horizontal axis represents the change in conductance Δw, and the horizontal axis represents Δt. The region from -50 ms to 0 ms represents the change in conductance G when the rising timing of the pre-spike is slower than the rising timing of the post-spike. 3-5 The relationship between the change in conductance Δw and Δt is shown. In the region where Δt is 0 ms to +50 ms, the conductance G 3-5 The relationship between the amount of change Δw and Δt is shown.

[0165] Referring to Figure 27, V gate Spike V gate When -spike is not applied to terminals 4 and 6 (=n / a), the conductance G 3-5 The change in Δw becomes positive, and the conductance G 3-5 rises, and in the negative Δt region, the conductance G 3-5 The change in Δw becomes negative, and the conductance G 3-5 Also, when Δt is ±25 ms, the conductance G 3-5 The absolute value of the change Δw becomes the maximum value.

[0166] When Δt is −25 ms (A), the spike voltage SPK1 shown in FIG. 22(a) is applied to the terminal 3 (i.e., a positive voltage is applied to the terminal 3), so that the dopant (oxygen vacancy) moves to the terminal 5 side (see FIG. 22(c)), and the conductance G 3-5 As a result, the conductance G 3-5 The change Δw becomes negative.

[0167] On the other hand, when Δt is +25 ms (B), the spike voltage SPK2 shown in FIG. 22(b) is applied to terminal 3 (i.e., a negative voltage is applied to terminal 3), so that the dopants (oxygen vacancies) move to the terminal 3 side (see FIG. 22(d)), and the conductance G3-5 As a result, the conductance G 3-5 The change Δw becomes positive.

[0168] FIG. 28 shows the V gate Spike V gate Conductance G when the -spike amplitude V is 5V 3-5 28 is a graph showing the relationship between the amount of change Δw in the conductance G and Δt. 3-5 The horizontal axis represents the amount of change Δw, and the horizontal axis represents Δt.

[0169] Referring to Figure 28, the conductance G 3-5 The change in the conductance Δw is in the positive region in both the region where Δt is -50 ms to 0 ms and the region where Δt is 0 ms to +50 ms. 3-5 The change in Δw is maximum when Δt is ±25 ms, but V gate Spike V gate This is smaller than when no -spike is applied (see FIG. 27).

[0170] FIG. 29 shows the V gate Spike V gate Conductance G when the -spike amplitude V is 7V 3-5 29 is a graph showing the relationship between the amount of change Δw in the conductance G 3-5 The horizontal axis represents the amount of change Δw, and the horizontal axis represents Δt.

[0171] Referring to Figure 29, V gate Spike V gate When the amplitude V of the -spike is 7V, in the region where Δt is positive, the conductance G 3-5 The change in Δw becomes negative, and the conductance G 3-5 decreases, and in the region where Δt is negative, the conductance G 3-5 The change in Δw becomes positive, and the conductance G 3-5 Also, when Δt is ±25 ms, the conductance G 3-5 The absolute value of the change in Δw becomes the maximum value, but V gateSpike V gate Compared with the case where the -spike is not applied (see Figure 27), the conductance G 3-5 In other words, the polarity of the change amount Δw is reversed.

[0172] When Δt is −25 ms (C), the spike voltage SPK1 shown in FIG. 22(a) is applied to the terminals 3, 4, and 6 (i.e., a positive voltage is applied to the terminals 3, 4, and 6), so that more dopants (oxygen vacancies) are concentrated on the terminal 5 side than in the case shown in FIG. 22(c) (see FIG. 22(e)), and the conductance G 3-5 As a result, the conductance G 3-5 The change Δw becomes positive.

[0173] On the other hand, when Δt is +25 ms (D), the spike voltage SPK2 shown in FIG. 22(b) is applied to the terminals 3, 4, and 6 (i.e., a negative voltage is applied to the terminals 3, 4, and 6), so that the dopants (oxygen vacancies) disperse (see FIG. 22(f)), and the conductance G 3-5 As a result, the conductance G 3-5 The change Δw becomes negative.

[0174] Therefore, V gate Spike V gate When the amplitude V of the -spike is increased to n / a=0V, 1V, 2V, 3V, 4V, 5V, 6V, and 7V, the conductance G 3-5 It was found that the change in the conductance Δw of the conductance G decreases and the polarity of the change in the conductance Δw is reversed (see Figure 29). 3-5 It was found that the absolute value of the change Δw in

[0175] From the above, we have demonstrated the implementation of STDP characteristics using a four-terminal memristor (Memristor 10) and the gate voltage (= V gate Spike V gate -spike) is possible. gate Spike V gateIt was also shown that it is possible to switch between symmetric and asymmetric STDP characteristics by using a cyclic spike.

[0176] [Pavlovian conditioning] Pavlovian conditioning will be described as a synaptic characteristic implemented in the memristor 10 shown in FIGS.

[0177] Pavlovian conditioning is a learning process in which a dog that does not respond to the sound of a bell alone will eventually begin to salivate in response to the sound of a bell alone by repeatedly ringing a bell and giving it food.

[0178] Fig. 30 is a diagram showing the correspondence between Pavlovian conditioning and the input and output of a memristor. Referring to Fig. 30, two inputs and one output are required to imitate Pavlov's dog, so terminals 3 and 6 are set as inputs and terminal 5 is set as an output. Terminal 3 corresponds to the sound of a bell (neutral stimulus, conditioned stimulus), terminal 6 to food (unconditioned stimulus), and terminal 5 to salivation (unconditioned reflex, conditioned reflex).

[0179] Before learning, a dog that does not respond to the sound of a bell alone is trained by applying a voltage V 1 When a low conductance G 3-5 In addition, the process of ringing the bell and giving food at the same time during learning is represented by outputting voltages V 1 ,V 4 When a high conductance G 3-5 Furthermore, after learning, the memristor 10 outputs a voltage V 1 By applying only 3-5 The output of

[0180] 31 is a diagram showing a voltage application protocol for implementing Pavlovian conditioning. The following steps (xi) to (xiv) will be described with reference to FIG.

[0181] (xi) A positive voltage is applied to terminals 3 and 4 to cause the dopants (oxygen vacancies) around the terminals 3 and 4 to drift toward terminals 5 and 6, thereby performing initialization.

[0182] (xii) After step (xi), a read voltage Read having an amplitude of 100 mV and an amplitude width of 2 ms is applied to terminal 3 with terminal 5 grounded, and the conductance G between terminals 3 and 5 is 3-5 Measure.

[0183] (xiii) After step (xii), a write voltage Write1 having an amplitude of −9 V and an amplitude width of 100 ms is applied to terminal 3, and then, with terminal 5 being grounded, a read voltage Read having an amplitude of 100 mV and an amplitude width of 2 ms is applied to terminal 3, and the conductance G between terminals 3 and 5 is 3-5 Then, the application of the write voltage Write1 and the read voltage Read are repeated nine times.

[0184] (xiv) After step (xiii), a write voltage Write1 is applied to terminal 3, and a write voltage Write2 having an amplitude of 9 V and an amplitude width of 100 ms is applied to terminal 6. Thereafter, a read voltage Read having an amplitude of 100 mV and an amplitude width of 2 ms is applied to terminal 3 while terminal 5 is grounded, and a conductance G between terminals 3 and 5 is 3-5 Then, the application of write voltages Write1 and Write2 to the terminals 3 and 6 and the application of a read voltage Read to the terminal 3 are repeatedly executed.

[0185] Figures 32 and 33 show the voltage V corresponding to the sound of a bell, respectively. 1 , the voltage V corresponding to the bait 4 and the conductance G corresponding to the presence or absence of a reaction 3-5 1A and 1B are first and second diagrams showing the relationship between the power consumption and the number of pulses.

[0186] In FIG. 32 and FIG. 33, the vertical bar VBr1 represents the initialization process, and the vertical bar VBr2 represents the conductance G between terminals 3 and 5 after the voltage application (the sound of the bell) to terminal 3. 3-5The vertical bar VBr3 represents the conductance G between terminals 3 and 5 after simultaneously applying a voltage to terminal 3 (a bell sound) and a voltage to terminal 6 (bait). 3-5 Represents the value of .

[0187] Referring to FIG. 32, in the initialization process, the conductance G 3-5 After initializing the value of to 0.6 ms (see “I” in Figure 32), apply voltage (V 1 When the bell is applied nine times, the conductance G 3-5 No significant changes were observed.

[0188] Next, when the write voltage Write1 is applied to the terminal 3 and the write voltage Write2 is applied to the terminal 6 simultaneously, the conductance G 3-5 The same process was repeated, and in the third period, when the write voltage Write1 was applied to terminal 3 and the write voltage Write2 was applied to terminal 6 simultaneously, the conductance G 3-5 A large increase in the conductance G was observed (see “II” in FIG. 32). 3-5 showed high values.

[0189] This result can be considered equivalent to a dog that did not respond to the sound of a bell alone learning to respond to the sound of a bell alone by associating the sound of a bell with food. 3-5 The increase in is believed to be due to the fact that a large potential difference was generated between terminals 3 and 6 when a potential was applied to terminals 3 and 6 simultaneously, causing the dopants (oxygen vacancies) around terminal 6 to drift to the vicinity of terminal 3.

[0190] After that, the initialization process is performed again to reset the conductance G 3-5 We succeeded in restoring the time to 0.6 mS (see “III” in Figure 32).

[0191] Furthermore, in the range of pulse numbers up to 80, the voltage V 1 ,V 4 By repeatedly applying the voltage, we were able to obtain similar conductance change characteristics.

[0192] With reference to FIG. 33, when the number of pulses is in the range of 80 to 200, the write voltage Write1 (=V 1 ) amplitude is set in the range of -6.5 to 0 V, and the write voltage Write2 (= V 4 The amplitude of the voltage was set in the range of 0 to 6.5 V.

[0193] During the initialization process, the conductance G 3-5 After initializing the value of to 0.6mS (see “IV” in Figure 33), apply voltage to terminal 3 (V 1 Applying the write voltage Write1 to terminal 3 and the write voltage Write2 to terminal 6 simultaneously; applying a voltage to terminal 3 (V 1 Applying the write voltage Write1 to terminal 3 and the write voltage Write2 to terminal 6 simultaneously; applying a voltage to terminal 3 (V 1 The application of the write voltage Write1 to terminal 3 and the application of the write voltage Write2 to terminal 6 were performed nine times, and the same results as those shown in FIG. 32 were obtained. Then, in the third period, when the write voltage Write1 was applied to terminal 3 and the write voltage Write2 was applied to terminal 6 simultaneously, the conductance G 3-5 It was observed that the conductance G 3-5 showed high values.

[0194] Subsequently, the same voltage was applied to the pulse numbers in the ranges of 120 to 160 and 160 to 200 as to the pulse number in the range of 80 to 120. As a result, in the pulse numbers in the ranges of 120 to 160 and 160 to 200, the conductance G 3-5was observed to increase significantly.

[0195] Therefore, the write voltage Write1 (=V 1 ) amplitude is set in the range of -6.5 to 0 V, and the write voltage Write2 (= V 4 ) in the range of 0 to 6.5 V, the amplitude of the pulses (V) required for the first conditioning and the second and subsequent conditioning (pulse number in the range of 120 to 200) can be calculated. 1 (V 4 The number of times of application of the voltage (voltage) can be the same.

[0196] Fig. 34 shows optical microscope images in the steps shown in Fig. 32 and Fig. 33. Fig. 35 is a conceptual diagram showing the change in distribution of dopants (oxygen vacancies) in the steps shown in Fig. 32 and Fig. 33.

[0197] 34, an optical microscope image (initialization 1) after the initialization process shows that dark colored regions are formed at the tips of terminals 5 and 6 (see (a)).

[0198] After that, the optical microscope image (Study 1) after the 30th pulse shows that a dark colored area is formed at the tip of terminal 3 (see (b)). As a result, the conductance G 3-5 It is believed that this has increased.

[0199] Since a similar change in the optical microscope image was observed when Initialization 2 → Learning 2 was subsequently performed (see (c) and (d)), it is believed that the repeated movement of this colored region caused switching between a high conductance state (HCS) and a low conductance state (LCS).

[0200] 35, in the initialization process, a positive voltage is applied to terminals 3 and 4, so that dopants (oxygen vacancies) move to the vicinity of terminals 5 and 6 (see (a)). As a result, the conductance G3-5 is expected to be lower.

[0201] In addition, in the learning process, a negative write voltage Write1 is applied to terminal 3 and a positive write voltage Write2 is applied to terminal 6, so that the dopants (oxygen vacancies) move to the periphery of terminal 3 (see (b)). As a result, the conductance G 3-5 is expected to be higher.

[0202] [Short-term potentiation / long-term potentiation] A description will be given of short-term potentiation / long-term potentiation as synaptic properties implemented in the memristor 10 shown in FIGS.

[0203] Human memory and forgetting are believed to be caused by the strength of connections that change depending on the signal input to the synapse, and the memory retention characteristics change depending on the frequency of this signal input.

[0204] Fig. 36 is a diagram showing the relationship between the coupling strength and the frequency of the input signal. In Fig. 36, curve k8 shows the relationship between the coupling strength and the frequency of the input signal.

[0205] Referring to Fig. 36, if the input is low frequency, the enhanced connection strength is only maintained for a short period of time and then decays. This is called short-term potentiation (STP). On the other hand, if the input is high frequency, the enhanced connection strength is maintained for a long period of time. This is called long-term potentiation (LTP).

[0206] Previously, when mimicking the STP / LTP characteristics using a two-terminal memristor element, the damping depended on the element, making it impossible to control the modulation of the STP / LTP characteristics after the element was fabricated.

[0207] Therefore, in the embodiment of the present invention, in order to mimic the modulatable STP / LTP characteristics, a voltage V 1 Applying a voltage V to terminals 4 and 6 corresponds to a signal input to the four-terminal memristor. 4,6Applying a voltage V corresponds to forgetting. 1 By the conductance G 3-5 and increase the voltage V 4,6 By the conductance G 3-5 Attenuate.

[0208] FIG. 37 is a conceptual diagram showing the distribution of dopants (oxygen vacancies) corresponding to the initialization, enhancement, and decay.

[0209] 37, as an initialization, a voltage V consisting of a pulse voltage having an amplitude in the range of ±1 to 20 V and an amplitude width in the range of 0.1 to 10 s is applied to terminal 3 with terminal 5 being grounded. 1 By applying the voltage Vcc multiple times, dopants (oxygen vacancies) are accumulated around terminal 3 (see (a)). In this case, the conductance G after initialization is 3-5 is, for example, 600 μS.

[0210] Next, as an enhancement, a positive voltage V is applied to terminal 3 with terminal 5 grounded. 1 By applying 3-5 (see (b)). In this case, a positive voltage V 1 For example, 18 V. Then, as attenuation, a negative voltage V 4,6 By applying 3-5 (c) In this case, the negative voltage V 4,6 is, for example, −8 V, −8.5 V, and −9 V. The STP / LTP characteristics are controlled by changing the frequency of enhancement.

[0211] Figure 38 shows the conductance G 3-5 V 1 38(a), (b), and (c) are diagrams showing input frequency dependence. 1 The lower row shows the input, and the lower row shows the conductance G 3-5 In addition, in (a), (b), and (c) of FIG. 38, the vertical axis represents the conductance G 3-5 The horizontal axis represents the number of pulses.1 =18V, V 4,6 = -9V, G 3-5 V 1 (b) shows the input frequency dependence of V 1 =18V, V 4,6 = -8.5V, conductance G 3-5 V 1 (c) shows the input frequency dependence of V 1 =18V, V 4,6 = -8V, G 3-5 V 1 Shows input frequency dependency.

[0212] Referring to FIG. 38(a), in the region REG1 of the number of pulses, V 1 is input twice, and in the pulse number region REG2, V 1 is input ten times. Then, in the region REG2, V 1 Conductance G after 10 inputs 3-5 The attenuation of V 1 Conductance G after two inputs 3-5 is smaller than the attenuation of

[0213] Referring to FIG. 38(b), in the region REG3 of the number of pulses, V 1 is input twice, and in the pulse number region REG4, V 1 is input ten times. Then, in the region REG4, V 1 Conductance G after 10 inputs 3-5 The attenuation of V 1 Conductance G after two inputs 3-5 is smaller than the attenuation of

[0214] With reference to FIG. 38(c), in the pulse number region REG5, V 1 is input twice, and in the pulse number region REG6, V 1 is input ten times. Then, in the region REG6, V 1 Conductance G after 10 inputs 3-5The attenuation of V 1 Conductance G after two inputs 3-5 is smaller than the attenuation of

[0215] In this way, V 1 As the frequency of input increases, the conductance G 3-5 Since the attenuation of V becomes smaller, 1 The retention period of the binding strength increased by inputting 10 times is V 1 This is longer than the retention period of the enhanced binding strength due to the double input.

[0216] Therefore, it was confirmed that short-term potentiation (STP) / long-term potentiation (LTP) occurs in the memristor 10-2.

[0217] Figure 39 shows the relationship between the number of decay pulses required to decay 10% from the maximum conductance and the gate voltage (V 4,6 39. In FIG. 39, the vertical axis represents the number of attenuation pulses, and the horizontal axis represents the gate voltage (V 4,6 ) Note that in the experiment in which short-term potentiation / long-term potentiation was implemented in memristor 10-2, the conductance did not decay by 20% from the maximum, so the number of decay pulses required to decay the conductance by 10% from the maximum is shown.

[0218] Referring to Figure 39, the input frequency (V 1 The conductance G 3-5 It was found that the number of pulses required for the decay of the gate voltage (V 4,6 ) becomes smaller, the conductance G 3-5 As a result, the number of pulses that attenuate the gate voltage (V 4,6 It was shown that retention properties can be modulated by

[0219] Therefore, the memristor 10-2 is equipped with STP / LTP characteristics and the gate voltage (V 4,6 ) successfully modulated the STP / LTP properties.

[0220] [Synaptic sensitization and habituation] The following describes synaptic sensitization and habituation as synaptic properties implemented in the memristor 10 shown in FIGS.

[0221] Fig. 40 is a diagram showing a voltage application protocol for synaptic sensitization and habituation. The following steps (xxi) to (xxv) will be described with reference to Fig. 40. (xxi) With terminals 3 and 5 grounded, a pulse voltage having an amplitude of 12 V and an amplitude width of 50 s is applied between terminals 4 and 6 as the gate voltage (V gate ) is applied. (xxii) After step (xxi), in the initialization process, a pulse voltage having an amplitude in the range of ±1 to 20 V and an amplitude width in the range of 0.1 to 10 s is applied to terminal 3 multiple times while terminal 5 is grounded, thereby measuring the conductance G 3-5 Set to 1.6mS. (xxiii) After step (xxii), in the Depression process, a pulse voltage having an amplitude of 13 V and an amplitude width of 300 ms is applied to terminal 3 while terminal 5 is grounded, and V gate The application of a pulse voltage having an amplitude of 100 ms and an amplitude width of 300 ms to the gate electrodes (terminals 4 and 6) is called a conductance G. 3-5 Repeat this process until V reaches 1.0 ms. gate was changed to not applied (n / a), 2V, 4V, 5V, 6V, 7V, 8V, 9V, 10V, 12V, and 13V. (xxiv) After step (xxiii), in the Potentiation process, a pulse voltage having an amplitude of −13 V and an amplitude width of 300 ms is applied to terminal 3 with terminal 5 grounded, which is called a conductance G 3-5 Repeat until it reaches 1.6mS. (xxv) After step (xxiv), the Depression process and the Potentiation process are repeatedly carried out.

[0222] During each pulse voltage application, the conductance G 3-5A pulse voltage (amplitude: 100 mV, amplitude width: 2 ms) for measuring is applied as a read voltage Read.

[0223] Figure 41 shows the conductance G 3-5 41 is a graph showing the relationship between the gate voltage V gate Conductance G when no current is applied 3-5 and the number of pulses. (b) shows the relationship between the gate voltage V gate Conductance G when 3-5 and the number of pulses. (c) shows the relationship between the gate voltage V gate Conductance G when 3-5 and the number of pulses, and (d) shows the relationship between the gate voltage V gate Conductance G when 3-5 The relationship between the conductance G and the number of pulses is shown in Fig. 41 (a), (b), (c), and (d). 3-5 and the horizontal axis represents the number of pulses.

[0224] 41, it was confirmed that it was possible to repeat potentiation (increasing conductance) and depression (decreasing conductance) many times according to the voltage application protocol shown in FIG. 40 (see (a) to (d) of FIG. 41). In addition, when the gate voltage V gate When all gate voltages V gate In this case, the total number of pulses applied exceeds 2000. Furthermore, the gate voltage V gate It can be seen that the number of pulses required for potentiation (increasing conductance) and depression (decreasing conductance) changes depending on the value of (see (b) to (d) in FIG. 41).

[0225] FIG. 42 is a diagram showing the gate voltage dependence of the average number of pulses required for Depression (lowering conductance) and Potentiation (higher conductance).

[0226] In FIG. 42, (a) shows the gate voltage dependency of the average number of pulses required for Depression (lower conductance), and (b) shows the gate voltage dependency of the average number of pulses required for Potentiation (higher conductance). In FIG. 42, (a) and (b), the vertical axis represents the average number of required pulses, and the horizontal axis represents the gate voltage V gate Represents.

[0227] Referring to FIG. 42(a), in the Depression process, the average number of required pulses is gate It can be seen that the trend is different at the gate voltage V = 7V boundary. gate Below 7V, the average number of required pulses is gate However, as the gate voltage V gate Above 7V, the average number of required pulses is gate decreases with increasing value of

[0228] On the other hand, in the potentiation (high conductance) process, the gate voltage V gate Excluding =n / a, the average number of requested pulses generally falls between 100 and 150 and can be considered constant.

[0229] FIG. 43 is a conceptual diagram showing the relationship between the distribution of dopants (oxygen vacancies) and the gate voltage. FIG. 43(a) shows the relationship between the distribution of dopants (oxygen vacancies) and the gate voltage V gate (a) shows the distribution of dopants (oxygen vacancies) when no gate voltage V is applied, and (b) shows the distribution of dopants (oxygen vacancies) when no gate voltage V is applied. gate (c) shows the distribution of dopants (oxygen vacancies) when the gate voltage V gate Almost V 1 / 2, (d) shows the distribution of dopants (oxygen vacancies) at gate voltage V gate Almost V 1 4 shows the distribution of dopants (oxygen vacancies) when

[0230] The number of pulses required for depression (lowering conductance) depends on whether or not a gate voltage is applied, and on the gate voltage V gate We will consider why the depression (low conductance) characteristic is modulated by the magnitude of V gate =n / a, V gate ≒ GND, V gate ≒V 1 / 2, and V gate ≒V 1 These can be broadly classified into four groups, and it can be assumed that each group has different characteristics because the distribution of dopants (oxygen vacancies) changes in different regions of the device. gate When = n / a, the dopants (oxygen vacancies) around terminal 3 drift to the vicinity of terminal 5 due to the application of voltage, resulting in a low conductance region between terminals 3 and 5. As a result, the conductance G 3-5 decreases (see Figure 43(a)). Next, V gate ≒GND, it can be inferred that a larger electric field distribution occurs between terminals 3-4 and 3-6 compared to between terminals 3-5. 3-5 It is expected that the dopants (oxygen vacancies) around terminal 3 will drift to the vicinity of terminals 4 and 6, forming a low conductance region around terminal 3, resulting in a large decrease in the conductance G 3-5 is expected to decrease (see Figure 43(b)).

[0231] On the other hand, V gate ≒V 1 When V is 1 / 2, it is considered that a gentler potential gradient is generated between the electrodes compared to (b). It can be inferred that the number of pulses required for the depression process is increased compared to (b) because the amount of drift of the dopant (oxygen vacancy) is small when each depression pulse voltage is applied (see (c) in Figure 43). Finally, when V gate ≒V 1 When the distribution of dopants (oxygen vacancies) between terminals 4 and 5 and between terminals 5 and 6 changes, the conductance G 3-5Compared to (c), a large electric field distribution occurs between terminals 4 and 5 and between terminals 5 and 6, and the amount of dopant (oxygen vacancy) drift per pulse in this region is large, so it can be considered that the number of pulses required for the Depression process is small (see (d) in Figure 43).

[0232] Fig. 44 is a schematic diagram of another memristor according to the first embodiment. With reference to Fig. 44, the memristor 10-3 includes a substrate 11, an oxide semiconductor layer 12, and a terminal 13. The substrate 11 includes glass 11-1 and ITO (Indium Tin Oxide) 11-2. The ITO 11-2 is formed on the surface of the glass 11-1. The ITO 11-2 has a thickness of, for example, 140 nm.

[0233] The oxide semiconductor layer 12 is disposed on the ITO 11-2 of the substrate 11 in contact with the surface of the ITO 11-2. The oxide semiconductor layer 12 is made of a-GaO y It consists of a-GaO y The thickness of the oxide semiconductor layer 12 is, for example, 200 nm. The terminal 13 is disposed on the oxide semiconductor layer 12 in contact with the surface of the oxide semiconductor layer 12. The terminal 13 is made of Pt. The thickness of the Pt film is, for example, 100 nm.

[0234] In this way, the memristor 10-3 has an oxide semiconductor layer 12 (a-GaO y ) and terminal 13 (Pt) are sequentially stacked on substrate 11, and it is operated by applying a voltage to ITO 11-2 of substrate 11 and terminal 13 (Pt), so it is a capacitor-type two-terminal memristor.

[0235] The memristor 10-3 is fabricated by cleaning the substrate 11, which has ITO formed on the surface of glass, with an organic solvent, and then depositing a-GaO on the ITO surface of the cleaned substrate 11 by pulsed laser deposition using gallium oxide as a target. y In this case, the fourth harmonic of a YAG laser with a wavelength of 266 nm is used as the laser light source. Ar gas is used, and the partial pressure of Ar gas is 4.0 × 10 -1Ar gas is supplied into the pulsed laser deposition apparatus so that Ar gas reaches a pressure of 1 Pa, and the substrate temperature is room temperature.

[0236] a-GaO y After forming the oxide semiconductor layer 2 (a-GaO y ) to form terminal 13. This completes the fabrication of memristor 10-3.

[0237] Figure 45 shows a-GaO y 45A and 45B are diagrams showing RHEED patterns before and after the formation of a-GaO y (b) shows the RHEED pattern of a-GaO y RHEED patterns after the deposition of the above film are shown.

[0238] Referring to FIG. 45, a-GaO y The RHEED patterns before and after the deposition of the film were halo-shaped, confirming that the gallium oxide had an amorphous structure.

[0239] Fig. 46 is a diagram showing the relationship between current and voltage in memristor 10-3. In Fig. 46, the vertical axis represents current and the horizontal axis represents voltage. Curves k9 to k13 respectively show the current-voltage characteristics in the first cycle, the current-voltage characteristics in the second cycle, the current-voltage characteristics in the third cycle, the current-voltage characteristics in the fourth cycle, and the current-voltage characteristics in the fifth cycle when the voltage is swept between -1.75V and +1.75V.

[0240] The voltage sweep was performed at a sweep rate of 4 V / s such that the voltage changed in a triangular shape, as shown in the inset.

[0241] 46, by sweeping the voltage in the order of 0 V → +1.75 V → 0 V → −1.75 V → 0 V, the current exhibits hysteretic voltage dependence in all of the first to fifth cycles, thereby confirming that the memristor 10-3 exhibits a memristor operation.

[0242] Fig. 47 is a diagram showing the relationship between the absolute value of current and voltage in the current-voltage characteristic shown in Fig. 46. In Fig. 47, the vertical axis represents current in logarithmic notation, and the horizontal axis represents voltage. Curves k9 to k13 are as described in Fig. 46.

[0243] 47, a resistance ratio of 8.96 was obtained at a measurement voltage of 0.1 V. This was because the a-GaO y This is believed to be because the amount of dopants (oxygen vacancies) increased by forming the film.

[0244] The STDP characteristics of the memristor 10-3 will now be described. Fig. 48 is a diagram showing spike voltages and a voltage application protocol.

[0245] Based on a pre-spike consisting of a square wave with an amplitude of 0.75V and an amplitude width of 25ms (=period 50ms), and a post-spike consisting of a square wave with an amplitude of 0.75V and an amplitude width of 25ms (=period 50ms), the difference Δt between the rising timing of the pre-spike and the post-spike is set to a predetermined value, and the post-spike voltage value is subtracted from the pre-spike voltage value to generate a spike voltage Spike.

[0246] Then, with the ITO of memristor 10-3 grounded, a spike voltage Spike (=(Pre-spike)-(Post-spike)) is applied to Pt.

[0247] The voltage application protocol is as follows. (xxxi) To eliminate the influence of the previous experiment, first calculate the conductance G 3-5 Initialize. (xxxii) After step (xxxi), a read voltage Read having a pulse waveform with an amplitude of 1 V and an amplitude width of 25 ms is applied to measure the conductance G 3-5 Measure. (xxxiii) After step (xxxii), a spike voltage Spike (=(Pre-spike)-(Post-spike)) is applied to Pt while ITO is grounded, and then a read voltage Read is applied to measure the conductance G 3-5 Measure. (xxxiv) After step (xxxiii), steps (xxxi) to (xxxiii) are repeatedly carried out. (xxxv) After step (xxxiv), the conductance G before and after application of the spike voltage Spike 3-5 Substituting this into equation (1) gives the conductance G 3-5 The amount of change Δw is calculated. (xxxvi) After step (xxxv), steps (xxxi) to (xxxv) are repeatedly executed while changing Δt.

[0248] Figure 49 shows the conductance G 3-5 49 is a diagram showing the relationship between the amount of change Δw in the spike timing and Δt. Referring to Fig. 49, the STDP characteristics are mainly classified into asymmetric STDP characteristics in which the positive or negative of the coupling strength change is determined by the positive or negative of the spike timing shift amount Δt, and symmetric STDP characteristics in which the coupling strength change is determined by the absolute value of Δt.

[0249] Figure 50 shows the conductance G 3-5 50 is a graph showing the measurement results of the change Δw in the conductance G. FIG. 51 is a graph showing the relationship between the applied spike voltage and the distribution of dopants (oxygen vacancies). In FIG. 3-5 The horizontal axis represents the amount of change Δw, and the horizontal axis represents Δt.

[0250] Referring to FIG. 50, when Δt is positive, the conductance G 3-5 becomes large, and as a result, the conductance G 3-5 The change in Δw becomes positive. On the other hand, if Δt is negative, the conductance G 3-5 becomes smaller, and as a result, the conductance G 3-5 The change Δw becomes negative.

[0251] This is because, when Δt is negative, a spike voltage shown in FIG. 51(a) is applied to Pt. As a result, the amplitude of the positive voltage becomes the largest, so the dopants (oxygen vacancies) are attracted to the ITO side (see FIG. 51(c)), and the conductance G 3-5 On the other hand, when Δt is positive, the spike voltage shown in FIG. 51(b) is applied to Pt. As a result, the amplitude of the negative voltage becomes the largest, and the dopants (oxygen vacancies) are absorbed into the a-GaO y (See Fig. 51(d)) and the conductance G 3-5 This is because increases.

[0252] In this way, in the memristor 10-3, the oxide semiconductor layer (a-GaO y Since the distribution of dopants (oxygen vacancies) in a plane parallel to the film thickness direction of the memristor 10-3 is controlled by the voltage applied to the terminal (Pt), the terminal (Pt) of the memristor 10-3 is a terminal that controls the distribution of dopants (oxygen vacancies) to a two-dimensional distribution.

[0253] FIG. 52 is a diagram showing another pre-spike and post-spike. FIG. 53 is a diagram showing the conductance G when a spike voltage generated from the pre-spike and post-spike shown in FIG. 3-5 53 is a graph showing the relationship between the amount of change Δw in conductance G and Δt. 3-5 The horizontal axis represents the amount of change Δw, and the horizontal axis represents Δt.

[0254] 52, a waveform combining a sawtooth wave having an amplitude of 1V and a time length of 50ms with a square wave having an amplitude of -1V and a time length of 5ms was used as the pre-spike and post-spike. The spike voltage was generated by subtracting the post-spike voltage value from the pre-spike voltage value. The timing at which the pre-spike and post-spike waveforms matched was set as Δt=0.

[0255] Referring to FIG. 53, in both the case where Δt is positive and the case where Δt is negative, the conductance G 3-5 The absolute value of the change in the coupling strength Δw becomes larger as the absolute value of Δt becomes smaller. In other words, the change in coupling strength becomes larger as the absolute value of Δt becomes smaller. Therefore, it is possible to implement characteristics closer to those of a biological synapse in the memristor 10-3.

[0256] FIG. 54 is a diagram showing yet another pre-spike and post-spike. FIG. 55 is a diagram showing the conductance G 3-5 55 is a graph showing the relationship between the amount of change Δw and Δt in the conductance G 3-5 The horizontal axis represents the amount of change Δw, and the horizontal axis represents Δt.

[0257] 54, a square wave with an amplitude of -0.75V and an amplitude width of 5ms was used as a pre-spike, and a sawtooth wave with a maximum amplitude of 0.75V and a time length of 100ms was used as a post-spike. The spike voltage was generated by subtracting the post-spike voltage value from the pre-spike voltage value. The timing at which the falling edge of the pre-spike coincides with the maximum amplitude of the post-spike was set to Δt=0.

[0258] Referring to Figure 55, the conductance G 3-5 The change in conductance Δw increases as Δt increases when Δt is negative, and decreases as Δt increases when Δt is positive. 3-5 The change in Δw changes symmetrically with respect to the line Δt = 0 as Δt changes. Therefore, we have successfully implemented the symmetric STDP characteristic, in which the change in coupling strength is determined by the absolute value of Δt.

[0259] FIG. 56 is a conceptual diagram showing an example of an application of a memristor. Referring to FIG. 56(a), by fabricating a semiconductor element SD incorporating a control circuit and a memristor 10, the above-mentioned "spike timing dependent plasticity", "Pavlovian conditioning", "short-term potentiation / long-term potentiation", and "synaptic sensitization and habituation" can be realized by a single semiconductor element. In this case, the control circuit receives a signal for executing any one of "spike timing dependent plasticity", "Pavlovian conditioning", "short-term potentiation / long-term potentiation", and "synaptic sensitization and habituation" from the outside, and executes any one of the voltage application protocols described in "spike timing dependent plasticity", "Pavlovian conditioning", "short-term potentiation / long-term potentiation", and "synaptic sensitization and habituation" to the memristor 10 according to the received signal. Then, the memristor 10 outputs a characteristic indicating any one of "spike timing dependent plasticity", "Pavlovian conditioning", "short-term potentiation / long-term potentiation", and "synaptic sensitization and habituation" according to the voltage control by the control circuit.

[0260] As a result, in the semiconductor element SD, the control circuit only controls the voltage applied to the memristor 10, and therefore the processing speed is not limited between the control circuit and the memristor 10.

[0261] Furthermore, when a voltage is applied to the memristor 10, dopants (oxygen vacancies) move in response to the applied voltage, and the memristor 10 stores the distribution of the moved dopants (oxygen vacancies).

[0262] Therefore, by using the semiconductor device SD, it functions as a logic-in-memory type computer.

[0263] On the other hand, as shown in Figure 56(b), the Von Neumann type computer carries instructions and data through a transmission line between the CPU (which functions as an arithmetic unit and a control unit) and the main memory, and the speed of this transmission line tends to become a bottleneck that limits performance, resulting in a slower output speed.

[0264] The operation of the semiconductor element SD shown in Figure 56(a) when it executes any of "spike timing dependent plasticity", "Pavlovian conditioning", "short-term potentiation / long-term potentiation" and "synaptic sensitization and habituation" will be described.

[0265] In the memristor 10, the terminal 3 is disposed at a first end of the oxide semiconductor layer 2 in a direction (first direction) parallel to the in-plane direction of the oxide semiconductor layer 2, and the terminal 5 is disposed at a second end of the oxide semiconductor layer 2 facing the first end in the first direction. The terminal 4 is disposed at a third end of the oxide semiconductor layer 2 in a second direction which is an in-plane direction of the oxide semiconductor layer 2 and different from the first direction, and the terminal 6 is disposed at a fourth end of the oxide semiconductor layer 2 facing the third end in the second direction. The terminal 3 in the memristor 10 is defined as a first terminal, the terminal 5 in the memristor 10 is defined as a second terminal, the terminal 4 in the memristor 10 is defined as a third terminal, and the terminal 6 in the memristor 10 is defined as a fourth terminal.

[0266] The control circuit executes any one of a first voltage control, a second voltage control, a third voltage control, and a fourth voltage control on the memristor 10. The first voltage control is a voltage control for implementing spike-timing-dependent plasticity in the memristor 10. The second voltage control is a voltage control for implementing Pavlovian conditioning in the memristor 10. The third voltage control is a voltage control for implementing in the memristor 10 short-term potentiation in which the binding strength enhanced by the input is maintained for only a first period and then decays, and long-term potentiation in which the enhanced binding strength is maintained for a period longer than the first period. The fourth voltage control is a voltage control for implementing in the memristor 10 low conductance and high conductance of the oxide semiconductor layer 2.

[0267] The first voltage control, the second voltage control, the third voltage control, and the fourth voltage control will be described below.

[0268] [Spike-timing-dependent plasticity] When the semiconductor element SD executes the "spike timing dependent plasticity", the control circuit executes the following steps S1 to S7 in the first voltage control. (S1) With the second terminal (terminal 5) grounded, a pulse voltage is applied to the first terminal (terminal 3) to measure the conductance G of the oxide semiconductor layer 2 between the first terminal (terminal 3) and the second terminal (terminal 5). 3-5 This step S1 is composed of steps (i) to (iv) in the above-mentioned "spike-timing-dependent plasticity". (S2) After step S1, the conductance G 3-5 A read voltage Read for reading V is applied to the first terminal (terminal 3) while the second terminal (terminal 5) is grounded. This step S2 corresponds to the step (v) of "spike-timing-dependent plasticity" described above. 1 Spike V 1 Before and after the application of the -spike, a read voltage Read is applied to terminal 3 with terminal 5 grounded, and V 1 Spike V 1 Conductance G before and after application of -spike (Spike1) 3-5 It consists of the description "Measure the temperature and humidity." (S3) After step S2, a first spike voltage consisting of a subtraction result of subtracting the voltage value of a second pulse voltage consisting of a square wave and having a rise timing different from that of the first pulse voltage from the voltage value of the first pulse voltage consisting of a square wave is applied to the first terminal (terminal 3) with the second terminal (terminal 5) grounded. This step S3 corresponds to the step (v) in the above-mentioned "spike timing dependent plasticity" in "V 1 Spike V 1 -spike is applied to terminal 3. And, V 1 Spike V 1 -spike constitutes the "first spike voltage" and V 1 Spike V 1 The pre-spike and post-spike that generated the -spike constitute the "first pulse voltage" and the "second pulse voltage", respectively. (S4) After step S3, a read voltage Read is applied to the first terminal (terminal 3) while the second terminal (terminal 5) is grounded. This step S4 corresponds to the "V 1 Spike V 1 Before and after the application of the -spike, a read voltage Read is applied to terminal 3 with terminal 5 grounded, and V 1 Spike V 1 Conductance G before and after application of -spike (Spike1) 3-5 It consists of the description "Measure the temperature and humidity." (S5) After step S4, execute step S1. This step S5 corresponds to the process (vi) in the above-mentioned "spike-timing-dependent plasticity." (S6) After step S5, with the second terminal (terminal 5) grounded, a first spike voltage is applied to the first terminal (terminal 3), and a second spike voltage which is a square wave and is a result of subtracting the voltage value of the second pulse voltage from the voltage value of a third pulse voltage which is different from the first pulse voltage is applied to the third terminal (terminal 4) and the fourth terminal (terminal 6), or the second spike voltage is not applied to the third terminal (terminal 4) and the fourth terminal (terminal 6).

[0269] This step S6 is the step (vii) in the above-mentioned "spike-timing-dependent plasticity" 1 Spike V 1 -spike (Spike1) is applied to terminal 3 and V gate Spike V gate -spike(Spike1') is applied to terminals 4 and 6. gate Spike V gate -spike (Spike1') is generated by subtracting Post-spike from Gate-spike (see Figure 23(b)). Then, V gate Spike V gateThe Gate-spike (Spike1') is applied or not applied (see FIG. 26). As a result, the Gate-spike constitutes the "third pulse voltage" and V gate Spike V gate -spike(Spike1') constitutes the "second spike voltage." (S7) After step S6, the first step S1 to the sixth step S6 are repeated one or more times while changing the difference between the rising timing of the first pulse voltage and the rising timing of the second pulse voltage. This step S7 is composed of the process (viii) in the above-mentioned "spike-timing-dependent plasticity".

[0270] In step S1, the control circuit a first sub-step of applying a read voltage Read to the first terminal (terminal 3) while the second terminal (terminal 5) is grounded; A second substep of applying a fourth pulse voltage to a third terminal (terminal 4) and a fourth terminal (terminal 6) while grounding the first terminal (terminal 3) and the second terminal (terminal 5) after the first substep; a third substep of executing the first substep after the second substep; a fourth substep of applying a fifth pulse voltage to the first terminal (terminal 3) while the second terminal (terminal 5) is grounded, after the third substep; Execute.

[0271] The first sub-step, the second sub-step, the third sub-step, and the fourth sub-step are respectively constituted by steps (i), (ii), (iii), and (iv) in the above-mentioned "spike-timing-dependent plasticity". The fourth pulse voltage is constituted by a pulse voltage having an amplitude of 12V and an amplitude width of 50s shown in FIG. 24. The fifth pulse voltage is constituted by the "various pulse voltages" in step (iv).

[0272] In step S7, the control circuit repeats the first step S1 to the sixth step S6 one or more times while increasing the amplitude of the second spike voltage in step S6. gate Spike V gate The amplitude of the -spike is changed from 1V to 7V.

[0273] [Pavlovian conditioning] When the semiconductor element SD executes the "Pavlovian conditioning", the control circuit executes the following steps S1 to S5 in the second voltage control. (S1) Initialize by applying a positive voltage to the first terminal (terminal 3) and the second terminal (terminal 5). This step S1 corresponds to the process (xi) in the above-mentioned [Pavlovian conditioning]. (S2) After step S1, a first write voltage is applied to the first terminal (terminal 3) multiple times. This step S2 corresponds to step (xiii) in the above-mentioned [Pavlovian conditioning]. The write voltage Write1 having an amplitude of -9V and an amplitude width of 100 ms corresponds to the "first write voltage." (S3) After step S2, a first write voltage is applied to the first terminal (terminal 3) and a second write voltage different from the first write voltage is applied to the fourth terminal (terminal 6). This step S3 corresponds to the process (xiv) in the above-mentioned [Pavlovian conditioning]. The write voltage Write2 having an amplitude of 9V and an amplitude width of 100ms constitutes the "second write voltage".

[0274] [Short-term potentiation / long-term potentiation] When the semiconductor element SD performs "short-term potentiation / long-term potentiation", the control circuit performs the following steps S1 to S3 in the third voltage control. (S1) A voltage V consisting of a pulse voltage having an amplitude in the range of ±1 to 20 V and an amplitude width in the range of 0.1 to 10 s 1This step S1 corresponds to "initialization in the above-mentioned "short-term potentiation / long-term potentiation" in which a voltage V consisting of a pulse voltage having an amplitude in the range of ±1 to 20 V and an amplitude width in the range of 0.1 to 10 s is applied to terminal 3 with terminal 5 grounded." 1 By applying the voltage V several times, dopants (oxygen vacancies) are accumulated around the terminal 3. 1 constitutes the "first pulse voltage." (S2) After step S1, a first voltage having a positive voltage value is applied to the first terminal (terminal 3) while the second terminal (terminal 5) is grounded. This step S2 corresponds to the above-mentioned "potentiation" in the "short-term potentiation / long-term potentiation" in which a positive voltage V is applied to terminal 3 while terminal 5 is grounded. 1 By applying 3-5 And, the positive voltage V 1 constitutes the "first voltage." (S3) After step S2, a second voltage consisting of a negative voltage is applied to the third terminal (terminal 4) and the fourth terminal (terminal 6) with the second terminal (terminal 5) grounded. This step S3 corresponds to "attenuation" in the above-mentioned "short-term potentiation / long-term potentiation" by applying a negative voltage V to terminals 4 and 6 with terminal 5 grounded. 4,6 By applying 3-5 And the negative voltage V 4,6 constitutes the "second voltage."

[0275] Then, in step S2, the control circuit applies the first voltage to the first terminal (terminal 3) with the second terminal (terminal 5) grounded at a first frequency and a second frequency greater than the first frequency. This is done by applying a voltage V 1 is applied twice, and the voltage V 1 This is supported by the fact that the frequency of the applied voltage is 10

[0276] [Synaptic sensitization and habituation] When the semiconductor element SD performs "short-term potentiation / long-term potentiation", the control circuit performs the following steps S1 to S5 in the fourth voltage control. (S1) With the first terminal (terminal 3) and the second terminal (terminal 5) grounded, a first pulse voltage is applied between the third terminal (terminal 2) and the fourth terminal (terminal 6). This step S1 corresponds to the step (xxi) in the above-mentioned "synaptic sensitization and habituation." The pulse voltage having an amplitude of 12V and an amplitude width of 50s constitutes the "first pulse voltage." (S2) After step S1, a pulse voltage having an amplitude in the range of ±1 to 20 V and an amplitude width in the range of 0.1 to 10 s is applied multiple times to the first terminal (terminal 3) while the second terminal (terminal 5) is grounded, to set the conductance of the oxide semiconductor layer to a first value. This step S2 is constituted by the process (xxii) in the above-mentioned "synapse sensitization and habituation". The pulse voltage having an amplitude in the range of ±1 to 20 V and an amplitude width in the range of 0.1 to 10 s constitutes a "second pulse voltage" different from the first pulse voltage. Also, for example, a conductance G of 1.6 mS is set to a first value. 3-5 constitutes the "first value." (S3) After step S2, a second pulse voltage different from the first pulse voltage is applied to the first terminal (terminal 3) while the second terminal (terminal 5) is grounded, and a third pulse voltage different from the first and second pulse voltages is applied to the third terminal (terminal 4) and the fourth terminal (terminal 6), or the third pulse voltage is not applied to the third terminal (terminal 4) and the fourth terminal (terminal 6), repeatedly until the conductance of the oxide semiconductor layer reaches a second value smaller than the first value.

[0277] This step S3 is composed of the step (xxiii) in the above-mentioned "synaptic sensitization and habituation". The pulse voltage having an amplitude of 13 V and an amplitude width of 300 ms constitutes the "second pulse voltage", and V gateA pulse voltage having an amplitude of 1.0 mS and an amplitude width of 300 ms constitutes a "third pulse voltage", for example, 3-5 constitutes the "second value." (S4) After step S3, a fourth pulse voltage, which is the inverted amplitude of the second pulse voltage, is repeatedly applied to the first terminal (terminal 3) while the second terminal (terminal 5) is grounded, until the conductance of the oxide semiconductor layer reaches a first value. This step S4 is composed of the process (xxiv) in the above-mentioned "synapse sensitization and habituation". And, the "-13V 300ms pulse voltage" in "Potentiation" shown in FIG. 40 constitutes the "fourth pulse voltage". (S5) After step S4, step S3 and step S4 are repeatedly executed. This step S5 is composed of the process (xxv) in the above-mentioned "synaptic sensitization and habituation".

[0278] In the above, the oxide semiconductor layer 2 is a single crystal TiO 2-x or a-GaO y The memristor 10 described here is made of single crystal TiO 2-x is an oxide semiconductor that consists of a single crystal phase and contains dopants (oxygen vacancies), and is called a-GaO y is an oxide semiconductor that consists of an amorphous phase and contains dopants (oxygen vacancies).

[0279] Therefore, the oxide semiconductor layer 2 may be made of an oxide semiconductor having a single crystal phase or an amorphous phase and containing a dopant (oxygen vacancies).

[0280] Furthermore, although the above describes memristor 10 as having two terminals (ITO and Pt) or four terminals 3 to 6, this is not limited to the first embodiment, and memristor 10 may have three terminals 3 to 5, or may have five or more terminals.

[0281] When the memristor 10 has three terminals 3 to 5, an electric field is generated between the terminals 3 and 4 and between the terminals 5 and 4 by connecting the terminals 3 and 5 to the ground potential GND and applying a positive voltage (+V) or a negative voltage (-V) to the terminal 4, so that the distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 2 can be controlled two-dimensionally as shown in Figures 4(b) and (c). When the memristor 10 has four or more terminals 3, 4, 5, 6,..., an electric field is generated between the terminal 3 and the terminals 4, 6,... other than the terminals 3 and 5 and between the terminal 5 and the terminals 4, 6,... other than the terminals 3 and 5 by connecting the terminals 3 and 5 to the ground potential GND and applying a positive voltage (+V) or a negative voltage (-V) to the terminals 4, 6,..., other than the terminals 3 and 5, so that the distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 2 can be controlled two-dimensionally as shown in Figures 4(b) and (c).

[0282] Therefore, in general, it is sufficient for the memristor 10 to have two or more terminals.

[0283] [Embodiment 2] Fig. 57 is a plan view of the memristor according to embodiment 2. Fig. 58 is a cross-sectional view of the memristor taken along line XXXXXVIII-XXXXXVIII shown in Fig. 57. In Fig. 57, an XY plane is defined.

[0284] 57 and 58, the memristor 20 according to the second embodiment includes a substrate 21, an insulating film 22, terminals 23 to 26, and an oxide semiconductor layer 27.

[0285] The memristor 20 has, for example, a square planar shape. The insulating film 22 is in contact with the substrate 21 and disposed on the substrate 21. The terminal 23 is in contact with the oxide semiconductor layer 27 and disposed on the oxide semiconductor layer 27. The terminal 23 is disposed such that its length direction forms a predetermined angle with the X-axis direction. The terminal 24 is in contact with the insulating film 22 and the oxide semiconductor layer 27 and disposed between the insulating film 22 and the oxide semiconductor layer 27. The terminal 24 is disposed such that its length direction forms a predetermined angle with the X-axis direction and intersects with the terminal 23.

[0286] The terminal 25 is disposed on one side in the in-plane direction of the oxide semiconductor layer 27 in contact with a side surface of the oxide semiconductor layer 27 that is parallel to the thickness direction. The terminal 25 is disposed so that the length direction is in the Y-axis direction. The terminal 25 has a laminated structure in which an insulating layer 25-1, a Pt layer 25-2, and an insulating layer 25-3 are laminated in this order. The insulating layer 25-1 has a thickness of, for example, 20 nm, the Pt layer 25-2 has a thickness of, for example, 40 nm, and the insulating layer 25-3 has a thickness of, for example, 20 nm. The material of the insulating layer 25-1 and the insulating layer 25-3 is SiO z (1.5≦z≦2), or the dopant (oxygen vacancy) density is 10 15 cm -3 ~10 17 cm -3 The semiconductor material is an oxide semiconductor having a structure as shown in FIG.

[0287] The terminal 26 is disposed on the other side in the in-plane direction of the oxide semiconductor layer 27 in contact with a side surface parallel to the thickness direction of the oxide semiconductor layer 27. The terminal 26 is disposed parallel to the terminal 25 so that the length direction is in the Y-axis direction. The terminal 26 has a laminated structure in which an insulating layer 26-1, a Pt 26-2, and an insulating layer 26-3 are laminated in this order. The insulating layer 26-1 has a thickness of, for example, 20 nm, the Pt 26-2 has a thickness of, for example, 40 nm, and the insulating layer 26-3 has a thickness of, for example, 20 nm. The material of the insulating layer 26-1 and the insulating layer 26-3 is SiO z (1.5≦z≦2), or the dopant (oxygen vacancy) density is 10 15 cm -3 ~10 17 cm -3 The semiconductor material is an oxide semiconductor having a structure as shown in FIG.

[0288] Each of the terminals 23 to 26 has a width W. The width W is, for example, 50 nm to 5 μm. The interval between the terminals 25 and 26 in the X-axis direction is W, and if the angle between the terminals 23 and 24 is 60°, the distance between the center of the terminals 25 and the center of the terminals 26 in the X-axis direction is 2W.

[0289] The oxide semiconductor layer 27 is disposed in contact with the insulating film 22 and the terminals 23-26.

[0290] The substrate 21 is made of single crystal silicon. The oxide semiconductor layer 27 is made of a-GaO y The terminal 23 is made of Pt having a thickness of 150 nm, for example. The terminal 24 is made of a Ti / Pt laminated structure in which Ti and Pt are sequentially laminated. In this case, the thickness of the Ti is, for example, 10 nm, and the thickness of the Pt is, for example, 50 nm. Each of the terminals 25 and 26 is made of Pt having a thickness of 40 nm, for example.

[0291] 59 to 63 are first to fifth process diagrams illustrating a method for manufacturing the memristor 20 shown in FIGS. 57 and 58, respectively.

[0292] 59, when the manufacture of the memristor 20 is started, single crystal silicon is washed to prepare the substrate 21 (step (a)). The washing method is ultrasonic washing using an organic solvent.

[0293] Then, the single crystal silicon is thermally oxidized to produce SiO 2 The insulating film 22 made of is formed on the substrate 21 (step (b)). The insulating film 22 may be formed on the substrate 21 by a plasma CVD (Chemical Vapor Deposition) method. In this case, the material may be, for example, SiH 4 Gas and N 2 O is used.

[0294] After step (b), a resist is applied onto the insulating film 22, and the applied resist is patterned by photolithography to form a resist pattern 12 on the insulating film 22 (step (c)).

[0295] Then, using the resist pattern 12 as a mask, Ti and Pt are sequentially laminated on the insulating film 22 by sputtering or electron beam lithography to form the terminal 24 on the insulating film 22 (step (d)). In this case, a laminate 13 of Ti and Pt is also formed on the resist pattern 12.

[0296] Thereafter, the resist pattern 12 is removed. In this case, the laminate 13 is removed by lift-off. The resist pattern 12 is removed by dissolving it in an organic solvent. As a result, the terminals 24 are formed on the insulating film 22 (step (e)).

[0297] Referring to FIG. 60, after step (e), a-GaO was deposited by pulsed laser deposition using gallium oxide as a target. y 27-1 is formed on the insulating film 22 and the terminal 24 (step (f)). The conditions for the pulsed laser deposition method are as described above.

[0298] Then, the resist is a-GaO y 27-1, and the applied resist is patterned by photolithography or electron beam lithography to form a resist pattern 14 on the a-GaO y 27-1 (step (g)).

[0299] Then, using the resist pattern 14 as a mask, a-GaO y / Pt / a-GaO y a-GaO y 27-1, terminals 25 and 26 are laminated on top of a-GaO y 27-1 (step (h)). In this case, a-GaO y / Pt / a-GaO y The laminate 15 is also formed on the resist pattern 14 .

[0300] 61, the resist pattern 14 is removed by the method described above. In this case, the laminate 15 is removed by lift-off. As a result, the terminals 25 and 26 are y 27-1 (step (i)).

[0301] Then, the resist is a-GaO y27-1, and the applied resist is patterned by photolithography or electron beam lithography to form a resist pattern 16 on the a-GaO y 27-1 (step (j)).

[0302] Then, using the resist pattern 16 as a mask, a-GaO y 17 a-GaO y 27-1 and the terminals 25 and 26 (step (k)). In this case, the conditions for the pulsed laser deposition method are as described above. y A part of 17 is formed on the terminals 25 and 26, but the top of the terminals 25 and 26 is a-GaO y It is made of high quality a-GaO y 17 can also be formed. y is also formed on the resist pattern 16.

[0303] 62, the resist pattern 16 is removed by the method described above. In this case, the a-GaO y is removed by lift-off. y 27-2 is a-GaO y 27-1, a-GaO y An oxide semiconductor layer 27 consisting of 27-1 and 27-2 is formed (step l).

[0304] Then, the resist is removed to form an oxide semiconductor layer 27 (a-GaO y 27-1, 27-2), and the applied resist is patterned by photolithography or electron beam lithography to form a resist pattern 18 on the oxide semiconductor layer 27 (a-GaO y 27-1, 27-2) (step (m)).

[0305] Then, using the resist pattern 18 as a mask, Pt is sputtered to form an oxide semiconductor layer 27 (a-GaO y27-1, 27-2) to form terminals 23 (step (n)). In this case, Pt 19 is also formed on resist pattern 18.

[0306] 63, after step (n), resist pattern 18 is removed by the method described above. In this case, Pt 19 formed on resist pattern 18 is removed by lift-off. This completes memristor 20 (step (q)).

[0307] The memristor 20 includes a terminal 23 arranged on one surface of the oxide semiconductor layer 27 in the film thickness direction, a terminal 24 arranged on the other surface of the oxide semiconductor layer 27 in the film thickness direction, a terminal 25 arranged on one side surface of the oxide semiconductor layer 27 parallel to the film thickness direction, and a terminal 26 arranged on the other side surface of the oxide semiconductor layer 27 parallel to the film thickness direction.

[0308] By controlling the voltages applied between the terminals 23 and 24 and between the terminals 25 and 26, the distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 27 can be controlled three-dimensionally.

[0309] Therefore, the memristor 20 can be used to realize the function of associative learning, thereby enabling the higher-level functions of biological systems to be exerted.

[0310] Furthermore, in the memristor 20, of the four terminals 23 to 26, two terminals 23, 24 are each arranged on one surface of the oxide semiconductor layer 27, and two terminals 25, 26 are each arranged on one side surface of the oxide semiconductor layer 27.

[0311] As a result, the area occupied by the memristor 20 in the in-plane direction of the substrate 21 can be made smaller than that of the memristor 10. Therefore, the memristors 20 can be integrated at a high density.

[0312] Fig. 64 is a plan view of another memristor according to embodiment 2. Fig. 65 is a cross-sectional view of the memristor taken along line XXXXXXV-XXXXXXV shown in Fig. 64.

[0313] The memristor according to the second embodiment may be the memristor 20A shown in FIGS.

[0314] 64 and 65, the memristor 20A is the same as the memristor 20, except that the terminals 23 and 24 of the memristor 20 shown in FIGS.

[0315] The terminal 23A has a T-shaped cross section, and is disposed on the oxide semiconductor layer 27 in contact with the oxide semiconductor layer 27. The terminal 23A has a thickness of, for example, 300 nm. The rest of the description of the terminal 23A is the same as the description of the terminal 23.

[0316] The terminal 24A is in contact with the insulating film 22 and the oxide semiconductor layer 27 and is disposed between the insulating film 22 and the oxide semiconductor layer 27. The terminal 24A has the same width as the width of the vertical bar part of the "T" of the terminal 23A. The terminal 24A is disposed to face the vertical bar part of the "T" of the terminal 23A. The terminal 24A also has a width W'. The width W' is, for example, 50 nm to 500 nm. The rest of the description of the terminal 24A is the same as the description of the terminal 24.

[0317] In the memristor 20A, the contact area between the terminal 23A and the oxide semiconductor layer 27 and the contact area between the terminal 24A and the oxide semiconductor layer 27 are smaller than the contact area between the terminal 23 and the oxide semiconductor layer 27 and the contact area between the terminal 24 and the oxide semiconductor layer 27 in the memristor 20.

[0318] As a result, the electric field applied between terminals 23A and 24A can be confined, and a stronger electric field can be applied between terminals 23A and 24A.

[0319] Memristor 20A is manufactured according to steps (a) to (q) shown in Figures 59 to 63. In this case, when terminal 24A is formed in steps (c) and (d), a resist pattern is formed in step (c) so that the length of a region where the resist pattern is not arranged is shorter than resist pattern 12, and terminal 24A is formed in step (d) using the formed resist pattern.

[0320] In addition, when forming the terminal 23A in steps (m) and (n), in step (m), a resist pattern is formed so that the cross-sectional shape of the area where the resist pattern is not arranged is T-shaped, and in step (n), the terminal 23A is formed using the formed resist pattern.

[0321] The memristor 20A includes a terminal 23A arranged on one surface of the oxide semiconductor layer 27 in the film thickness direction, a terminal 24A arranged on the other surface of the oxide semiconductor layer 27 in the film thickness direction, a terminal 25 arranged on one side surface of the oxide semiconductor layer 27 parallel to the film thickness direction, and a terminal 26 arranged on the other side surface of the oxide semiconductor layer 27 parallel to the film thickness direction.

[0322] By controlling the voltage applied between terminals 23A and 24A and between terminals 25 and 26, the distribution of dopants (oxygen vacancies) in oxide semiconductor layer 27 can be controlled three-dimensionally.

[0323] Therefore, the memristor 20A can be used to realize the function of associative learning, and can exert the higher-order functions of a biological system.

[0324] Fig. 66 is a plan view of yet another memristor according to embodiment 2. Fig. 67 is a cross-sectional view of the memristor taken along line XXXXXXVII-XXXXXXVII shown in Fig. 66.

[0325] The memristor according to the second embodiment may be the memristor 20B shown in FIGS.

[0326] 66 and 67, the memristor 20B is the same as the memristor 20 shown in FIGS.

[0327] The insulating film 28 is in contact with the terminal 23 and the oxide semiconductor layer 27 and is disposed between the terminal 23 and the oxide semiconductor layer 27. The insulating film 28 has a through hole 28-1. The terminal 23 is in contact with the oxide semiconductor layer 27 via the through hole 28-1 of the insulating film 28. The diameter of the through hole 28-1 is, for example, 50 nm to 500 nm. As a result, the through hole 28-1 has an area smaller than the contact area between the terminal 23 and the oxide semiconductor layer 27 in the memristor 20. The insulating film 28 is made of, for example, SiO z (1.5≦z≦2) and has a film thickness of, for example, 20 nm.

[0328] The insulating film 29 is disposed on the insulating film 22 and the terminal 24 in contact with them. The insulating film 29 has a through hole 29-1. The terminal 24 is in contact with the oxide semiconductor layer 27 via the through hole 29-1 of the insulating film 29. The through hole 29-1 is disposed to face the through hole 28-1 of the insulating film 28. The diameter of the through hole 29-1 is the same as the diameter of the through hole 28-1. As a result, the through hole 29-1 has an area smaller than the contact area between the terminal 24 and the oxide semiconductor layer 27 in the memristor 20. The insulating film 29 is made of, for example, SiO z and has a film thickness of, for example, 20 nm. In the insulating film 29, the oxygen composition z may be the same as or different from the oxygen composition z in the insulating film 28. In the memristor 20B, the terminal 23 has a film thickness of 100 nm.

[0329] 68 to 74 are first to seventh process diagrams illustrating a method for manufacturing the memristor 20B shown in FIGS. 66 and 67, respectively.

[0330] 68, when the manufacture of the memristor 20B starts, the same steps as steps (a) to (e) shown in FIG. 59 are sequentially performed (steps (a) to (e) in FIG. 68).

[0331] Referring to FIG. 69, after step (e), SiO 2 SiO was deposited by pulsed laser deposition using a SiO target. z 31 is formed on the insulating film 22 and the terminals 24 (step (f)).

[0332] Then, the resist is SiO z The resist is applied onto 31 and the applied resist is patterned to form a resist pattern 32 (step (g)).

[0333] Then, using the resist pattern 32 as a mask, SiO z 31 is etched (step (h)). As a result, a SiO z An insulating film 29 made of the above is formed.

[0334] Subsequently, the same steps as step (f) to step (l) shown in Figures 60 to 62 are sequentially carried out (step (i) in Figure 69, step (j) to step (l) in Figure 70, and step (m) to step (o) in Figure 71).

[0335] Referring to FIG. 72, after step (o), a-GaO y 27-1, terminals 25, 26 and a-GaO y A resist is applied onto 27-2, and the applied resist is patterned to form a resist pattern 33 (step (p)).

[0336] Then, using the resist pattern 33 as a mask, SiO 2 SiO was deposited by pulsed laser deposition using a SiO target. z 34 a-GaO y 27-2 (step (q)). In this case, SiO z 35 is formed.

[0337] Then, the resist is coated with SiO zThe resist is applied onto 34 and the applied resist is patterned to form a resist pattern 36 (step (r)).

[0338] Referring to FIG. 73, using the resist pattern 36 as a mask, SiO z 34 is etched (step (s)).

[0339] Then, the resist patterns 33 and 36 are removed by the above-mentioned method. As a result, the insulating film 28 having the through-hole 28-1 is left as a-GaO y 27-2 (step (t)). Also, SiO z 35 is removed by lift-off when the resist pattern 33 is removed.

[0340] Then, a-GaO y 27-1, terminals 25, 26 and a-GaO y A resist is applied onto 27-2, and the applied resist is patterned to form a resist pattern 37 (step (u)).

[0341] Referring to FIG. 74, Pt is deposited on oxide semiconductor layer 27 (a-GaO y 27-1, 27-2) (step (v)). In this case, Pt is also formed on the resist pattern 37.

[0342] After step (v), the resist pattern 37 is removed by the method described above. In this case, the Pt formed on the resist pattern 37 is removed by lift-off. As a result, the terminal 23 is formed on the oxide semiconductor layer 27 and the insulating film 28 in contact with the oxide semiconductor layer 27 and the insulating film 28, and the memristor 20B is completed (step (w)).

[0343] In the memristor 20B, an insulating film 28 having a through hole 28-1 is disposed between the terminal 23 and the oxide semiconductor layer 27, and an insulating film 29 having a through hole 29-1 is disposed between the terminal 24 and the oxide semiconductor layer 27.

[0344] As a result, the voltage applied between terminals 23-24 is applied to oxide semiconductor layer 27 through through holes 28-1, 29-1. Therefore, the electric field applied between terminals 23-24 can be confined, and a stronger electric field can be applied between terminals 23-24. That is, in memristor 20B, insulating films 28, 29 function to confine the electric field applied to oxide semiconductor layer 27.

[0345] The memristor 20B includes a terminal 23 arranged on one surface of the oxide semiconductor layer 27 in the film thickness direction, a terminal 24 arranged on the other surface of the oxide semiconductor layer 27 in the film thickness direction, a terminal 25 arranged on one side surface of the oxide semiconductor layer 27 parallel to the film thickness direction, and a terminal 26 arranged on the other side surface of the oxide semiconductor layer 27 parallel to the film thickness direction.

[0346] By controlling the voltages applied between the terminals 23 and 24 and between the terminals 25 and 26, the distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 27 can be controlled three-dimensionally.

[0347] Therefore, the function of associative learning can be realized using the memristor 20B, and the higher-order functions of a biological system can be exerted.

[0348] Fig. 75 is a plan view of yet another memristor according to embodiment 2. Fig. 76 is a cross-sectional view of the memristor taken along line XXXXXXXVI-XXXXXXXVI shown in Fig. 75.

[0349] The memristor according to the second embodiment may be a memristor 40 shown in Figures 75 and 76. With reference to Figures 75 and 76, the memristor 40 includes a substrate 41, terminals 42 to 45, and an oxide semiconductor layer 46.

[0350] The memristor 40 has, for example, a square planar shape. The substrate 41 is made of single crystal TiO 2The terminal 42 is disposed on the oxide semiconductor layer 46 in contact with the oxide semiconductor layer 46. The terminal 42 is disposed such that the length direction forms a predetermined angle with the X-axis direction. The terminal 42 is made of Pt having a film thickness of 150 nm, for example.

[0351] The terminal 43 is disposed so as to penetrate the substrate 41 in the thickness direction and to be in contact with the substrate 41 and the oxide semiconductor layer 46. The terminal 43 is disposed so that its length forms a predetermined angle with the X-axis direction and intersects with the terminal 42. The terminal 43 is made of Pt having a film thickness of, for example, 50 nm.

[0352] The terminal 44 is disposed in contact with a side surface of the oxide semiconductor layer 46 on one side in the in-plane direction of the substrate 41. The terminal 44 is disposed so that its length direction is in the Y-axis direction. The terminal 45 is disposed in contact with a side surface of the oxide semiconductor layer 46 on the other side in the in-plane direction of the substrate 41. The terminal 45 is disposed so that its length direction is in the Y-axis direction. Each of the terminals 44 and 45 is made of Pt having a film thickness of, for example, 40 nm.

[0353] Each of terminals 42 to 45 has a width W. The distance between terminals 44 and 45 in the X-axis direction is W, and if the angle between terminals 42 and 43 is 60°, the distance between the center of terminal 44 and the center of terminal 45 in the X-axis direction is 2W.

[0354] The oxide semiconductor layer 46 is disposed in contact with the substrate 41 and the terminals 42 to 45. The oxide semiconductor layer 46 is made of single crystal TiO 2-x It consists of:

[0355] 77 to 82 are first to sixth process diagrams illustrating a method for manufacturing the memristor 40 shown in FIGS. 75 and 76, respectively.

[0356] Referring to FIG. 77, when the manufacture of the memristor 40 is started, a single crystal TiO having a (001) surface is formed by the above-described method. 2 The substrate 41 is prepared by cleaning (step (a)).

[0357] And TiO 2 Single crystal TiO was deposited on the substrate 41 by pulsed laser deposition using the target 2-n 46-1 is formed (step (b)). In this case, the conditions for the pulsed laser deposition method are as described above.

[0358] After step (b), the back surface of the substrate 41 (single crystal TiO 2-n A resist is applied to the surface opposite to the surface on which 46-1 is formed, and the applied resist is patterned to form a resist pattern 51 (step (c)).

[0359] Then, the substrate 41 is etched using the resist pattern 51 as a mask (step (d)). Thereafter, the resist pattern 51 is removed by the method described above. As a result, the through hole 41-1 is formed in the substrate 41 (step (e)).

[0360] Subsequently, the back surface of the substrate 41 (single crystal TiO 2-n A resist is applied to the surface opposite to the surface on which 46-1 is formed, and the applied resist is patterned to form a resist pattern 52 (step (f)).

[0361] 78, resist pattern 52 is used as a mask to form through hole 41-1 and the rear surface of substrate 41 (single crystal TiO 2-n Pt is deposited on the surface opposite to the surface on which 46-1 is formed by sputtering (step (g)). In this case, Pt 53 is formed on the resist pattern 52.

[0362] Thereafter, the resist pattern 52 is removed by the above-mentioned method (step (h)). As a result, the Pt 53 is removed by lift-off, and the terminal 43 is a single crystal TiO 2-n It is formed adjacent to 46-1.

[0363] The resist was then coated with single-crystal TiO 2-nThe resist is applied onto 46-1, and the applied resist is patterned to form a resist pattern 54 (step (i)).

[0364] Subsequently, using the resist pattern 54 as a mask, TiO 2 Single-crystal TiO was fabricated by pulsed laser deposition using a TiO target. 2-n Single crystal TiO on 46-1 2-n 55 is formed (step (j)). In this case, the conditions in the pulsed laser deposition method are the same as those in step (b).

[0365] Referring to FIG. 79, after step (j), resist pattern 54 is removed by the above-mentioned method (step (k)). Then, single crystal TiO 2-n 46-1 and single crystal TiO 2-n A resist is applied onto 55, and the applied resist is patterned to form a resist pattern 56 (step (l)).

[0366] Thereafter, using the resist pattern 56 as a mask, Pt is deposited on a single crystal TiO 2-n Single crystal TiO deposited on 46-1 2-n Terminals 44, 45 are formed so as to contact both side surfaces of 55 (step (m)). In this case, Pt 57 is also deposited on resist pattern 56.

[0367] 80, the resist pattern 56 is removed by the method described above (step (n)). Thereby, the Pt 57 deposited on the resist pattern 56 is removed by lift-off.

[0368] Subsequently, single crystal TiO 2-n A resist is applied onto 46-1, 55 and the terminals 44, 45, and the applied resist is patterned to form a resist pattern 58 (step (o)).

[0369] Then, using the resist pattern 58 as a mask, single crystal TiO 2-nThe resist pattern 55 is then etched to a predetermined depth (step (p)), and the resist pattern 58 is removed by the method described above. As a result, the single crystal TiO 2-n 46-2 is formed (step (q)).

[0370] The resist is then removed using single crystal TiO 2-n A resist is applied onto 46-1, 46-2 and terminals 44, 45, and the applied resist is patterned to form a resist pattern 59 (step (r)).

[0371] Subsequently, using the resist pattern 59 as a mask, Pt is deposited on a single crystal TiO 2-n 46-2 to form the terminal 42 (step (s)). In this case, Pt60 is deposited on the resist pattern 59.

[0372] 82, after step (s), resist pattern 59 is removed by the method described above. As a result, Pt60 is removed by lift-off, and single crystal TiO 2-n A memristor 40 including an oxide semiconductor layer 46 consisting of 46-1 and 46-2 is completed (step (t)).

[0373] The memristor 40 includes a terminal 42 arranged on one surface of the oxide semiconductor layer 46 in the film thickness direction, a terminal 43 arranged on the other surface of the oxide semiconductor layer 46 in the film thickness direction, a terminal 44 arranged on one side surface of the oxide semiconductor layer 46 parallel to the film thickness direction, and a terminal 45 arranged on the other side surface of the oxide semiconductor layer 46 parallel to the film thickness direction.

[0374] By controlling the voltage applied between the terminals 42 and 43 and between the terminals 44 and 45, the distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 46 can be controlled three-dimensionally.

[0375] Therefore, the memristor 40 can be used to realize the function of associative learning, thereby enabling the higher-level functions of biological systems to be exerted.

[0376] In the above, a-GaO y The memristors 20, 20A, and 20B each having an oxide semiconductor layer 27 made of single-crystal TiO 2-x The memristor 40 having the oxide semiconductor layer 46 made of

[0377] The memristors 20, 20A, 20B, 40 each include two terminals 23, 24 (or two terminals 23A, 24A or two terminals 42, 43) arranged on both sides of the oxide semiconductor layer 27 (or the oxide semiconductor layer 46) in the film thickness direction, and two terminals 25, 26 (or two terminals 44, 45) arranged on two side surfaces parallel to the film thickness direction of the oxide semiconductor layer 27 (or the oxide semiconductor layer 46).

[0378] In addition, single crystal TiO 2-x is an oxide semiconductor that consists of a single crystal phase and contains dopants (oxygen vacancies), and is called a-GaO y is an oxide semiconductor that consists of an amorphous phase and contains dopants (oxygen vacancies).

[0379] Therefore, each of the oxide semiconductor layers 27 and 46 may be made of an oxide semiconductor having a single crystal phase or an amorphous phase and containing a dopant (oxygen vacancies).

[0380] Furthermore, in the above, it has been described that the memristors 20, 20A, 20B, 40 have four terminals 23 to 26 (or four terminals 23A, 24A, 25, 26 or four terminals 42 to 45), but in the second embodiment, this is not limited to the above, and the memristors 20, 20A, 20B, 40 may have three terminals 23 to 25 (or three terminals 23A, 24A, 25 or three terminals 42 to 44), or may have five or more terminals.

[0381] When the memristor 20, 20A, 20B, 40 has three terminals 23 to 25 (or three terminals 23A, 24A, 25, or three terminals 42 to 44), the terminals 23 and 24 (or terminals 23A and 24A, or terminals 42 and 43) are connected to the ground potential GND, and a positive voltage (+V) or a negative voltage (-V) is applied to the terminal 25 (or terminal (44)), thereby applying a voltage between the terminals 23 and 25 and between the terminals 24 and 25 (or between the terminals 23A and 25 and between the terminals 24A and 25) and between the terminals 24A and 25 (or between the terminals 23A and 25 and between the terminals 24A and 25) to the ground potential GND. 25, or between terminals 42-44 and between terminals 43-44), the distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 27 (or oxide semiconductor layer 46) can be controlled three-dimensionally. In addition, when the memristor 20, 20A, 20B, 40 has three or more terminals 23, 24, 25, 26, . . . (or three or more terminals 23A, 24A, 25, 26, . . ., or three or more terminals 42, 43, 44, 45, . . .), the terminals 23, 24 ( By connecting terminal 23 to terminals 25, 26,... (or terminals 23A, 24A, or terminals 42, 43) other than terminals 23 and 24 (or terminals 25, 26,... other than terminals 23A, 24A, or terminals 42, 43) to the ground potential GND and applying a positive voltage (+V) or a negative voltage (-V) to terminals 25, 26,... (or terminals 25, 26,... other than terminals 23A, 24A, or terminals 44, 45,... other than terminals 42, 43), a voltage Since an electric field is generated between terminal 23A and terminals 25, 26, ... other than terminals 23A, 24A and between terminal 24A and terminals 25, 26, ... other than terminals 23A, 24A, or between terminal 42 and terminals 44, 45, ... other than terminals 42, 43 and between terminal 43 and terminals 44, 45, ... other than terminals 42, 43), the distribution of dopants (oxygen vacancies) in the oxide semiconductor layer 27 (or oxide semiconductor layer 46) can be controlled three-dimensionally.

[0382] Therefore, in general, it is sufficient for the memristors 20, 20A, 20B, and 40 to have three or more terminals.

[0383] In the above-mentioned memristors 20, 20A, 20B, and 40, the electrically active layer (the portion where the distribution of dopants (oxygen vacancies) changes) is buried inside the memristors 20, 20A, 20B, and 40, so that it is possible to reduce element variations caused by exposure of the electrically active layer (the portion where the distribution of dopants (oxygen vacancies) changes), and to suppress element failure.

[0384] Furthermore, when a demultiplexer is formed using the memristors 20, 20A, 20B, and 40, they can be laminated in a single step as thin films in the thickness direction, simplifying the process.

[0385] [Application example] Fig. 83 is a conceptual diagram of an array system according to an embodiment of the present invention. Referring to Fig. 83, array system 100 includes first layer 101, second layer 102, and third layer 103. Each of first layer 101, second layer 102, and third layer 103 is arranged along an XY plane. And, first layer 101, second layer 102, and third layer 103 are arranged stacked in the Z-axis direction.

[0386] FIG. 84 is a conceptual diagram showing a configuration of the first layer 101 shown in FIG. 83. Referring to FIG. 84, the first layer 101 includes nanowires 111 to 113 and a memristor 115. The nanowire 111 is arranged, for example, along the Y-axis direction. The nanowire 112 is arranged, for example, along the X-axis direction and intersects with the nanowire 111. The nanowire 113 is arranged, for example, so as to intersect with the nanowires 111 and 112 at an angle of 45°. The nanowire 113 is composed of two nanowires 113a and 113b. Each of the nanowires 111, 112, and 113 is arranged in a planar shape along the XY plane. Furthermore, the nanowires 111, 112, and 113 are arranged, for example, stacked in the order of the nanowire 111, the nanowire 112, and the nanowire 113 at the intersections where they intersect with each other.

[0387] The memristor 115 is disposed at an intersection of the nanowires 111 to 113. The memristor 115 is formed of any one of the memristors 10, 20, 20A, 20B, and 40 described above. For example, when the memristor 115 is formed of the memristor 10, the nanowire 111 is electrically connected to the terminal 3 of the memristor 10, the nanowire 112 is electrically connected to the terminal 5 of the memristor 10, one of the two nanowires 113a and 113b of the nanowire 113 is electrically connected to the terminal 4, and the other of the two nanowires 113a and 113b of the nanowire 113 is electrically connected to the terminal 6.

[0388] Therefore, by applying a voltage between terminals 3 and 5 via nanowires 111 and 112 and between terminals 4 and 6 via nanowire 113 (113a and 113b), the distribution of dopants (oxygen vacancies) in oxide semiconductor layer 2 can be controlled two-dimensionally.

[0389] Furthermore, when the memristor 115 is composed of the memristor 20, the nanowire 111 is electrically connected to the terminal 23 of the memristor 20, the nanowire 112 is electrically connected to the terminal 24 of the memristor 10, one of the two nanowires 113a, 113b of the nanowire 113 is electrically connected to the terminal 25, and the other of the two nanowires 113a, 113b of the nanowire 113 is electrically connected to the terminal 26.

[0390] When the memristor 115 is composed of any one of the memristors 20A, 20B, or 40, the nanowires 111, 112, and 113 (113a and 113b) are electrically connected to the four terminals in the same manner as when the memristor 115 is composed of the memristor 20.

[0391] Therefore, by applying a voltage between terminals 23 and 24 (or between terminals 42 and 43) via nanowires 111 and 112 and applying a voltage between terminals 25 and 26 (or between terminals 44 and 45) via nanowire 113 (113a, 113b), the distribution of dopants (oxygen vacancies) in oxide semiconductor layer 27 (or oxide semiconductor layer 46) can be controlled three-dimensionally.

[0392] Each of the second layer 102 and the third layer 103 has the same structure as the first layer 101 shown in FIG.

[0393] In the array system 100, each of the first layer 101, the second layer 102 and the third layer 103 has a configuration in which a memristor 115 is disposed at the intersection of the nanowires 111, 112 and 113, and the distribution of dopants (oxygen vacancies) in the memristor 115 is controlled two-dimensionally or three-dimensionally according to the voltage applied by the intersecting nanowires 111, 112 and 113, thereby realizing the function of associative learning. Therefore, the array system 100 shown in Fig. 83 is an array system with a crossbar structure.

[0394] 85 is a conceptual diagram of another array system according to the embodiment of the present invention. The array system according to the embodiment of the present invention may be an array system 200 shown in FIG.

[0395] 85, the array system 200 includes a first layer 201, a second layer 202, and a third layer 203. Each of the first layer 201, the second layer 202, and the third layer 203 is arranged along the XY plane. The first layer 201, the second layer 202, and the third layer 203 are arranged in a stacked manner in the Z-axis direction.

[0396] Fig. 86 is a conceptual diagram showing the configuration of the first layer 201 shown in Fig. 85. Referring to Fig. 86, the first layer 201 is the same as the first layer 101 except that the memristor 115 of the first layer 101 shown in Fig. 84 is replaced with a memristor 116 and a nanowire 114 is added.

[0397] Nanowire 114 intersects with nanowires 111 to 113. Nanowire 114 is made up of two nanowires 114a and 114b. Each of nanowires 111, 112, 113, and 114 is arranged in a planar shape along the XY plane. Nanowires 111, 112, 113, and 114 are stacked at their intersections in the order of nanowire 111, nanowire 112, nanowire 113, and nanowire 114, for example.

[0398] The memristor 116 is disposed at the intersection of the nanowires 111 to 114. The memristor 116 has six terminals. In the memristor 10, a memristor having six terminals can be configured by arranging the six terminals in a circular shape on one surface in the film thickness direction of the oxide semiconductor layer 2 such that two lines connecting the center in the in-plane direction of the oxide semiconductor layer 2 and two adjacent terminals form an angle of 60°.

[0399] Moreover, in memristor 20, a memristor having six terminals can be configured by adding two terminals to the surface of oxide semiconductor layer 27 on the near side of the paper in Fig. 58 and the surface of oxide semiconductor layer 27 on the far side of the paper in Fig. 58. Furthermore, in memristor 20, a memristor having six terminals can be configured by adding two terminals to the side surfaces of oxide semiconductor layer 27 below or above terminals 25, 26 in the same manner as terminals 25, 26. In the same manner, memristors having six terminals can be configured for memristors 20A, 20B, and 40.

[0400] If the six terminals are terminals TN1 to TN6, when the terminals TN1 to TN6 are arranged on one surface in the thickness direction of the oxide semiconductor layer 2, nanowires 111 and 112 are connected to two terminals TN1 and TN2 facing each other, nanowire 113 (113a, 113b) is connected to two terminals TN3 and TN4 of the terminals TN3 to TN6, and nanowire 114 (114a, 114b) is connected to two terminals TN5 and TN6 of the terminals TN3 to TN6.

[0401] Moreover, when the terminals TN1 to TN6 are arranged on the oxide semiconductor layer 27 by the above-mentioned method, the nanowires 111 and 112 are electrically connected to the two terminals TN1 and TN2 arranged in the film thickness direction of the oxide semiconductor layer 27, the nanowires 113 (113a and 113b) are electrically connected to the terminals TN3 and TN4 arranged on the two side surfaces of the oxide semiconductor layer 27, and the nanowires 114 (114a and 114b) are electrically connected to the terminals TN5 and TN6 arranged on the two side surfaces of the oxide semiconductor layer 27. When six terminals TN1 to TN6 are arranged on the oxide semiconductor layer 46, the nanowires 111, 112, 113 (113a and 113b), and 114 (114a and 114b) are electrically connected to the six terminals TN1 to TN6 in the same manner.

[0402] Therefore, by applying a voltage between terminals TN1 and TN2 via nanowires 111 and 112, applying a voltage between terminals TN3 and TN4 via nanowire 113 (113a, 113b), and applying a voltage between terminals TN5 and TN6 via nanowire 114 (114a, 114b), the distribution of dopants (oxygen vacancies) in oxide semiconductor layer 27 (or oxide semiconductor layer 46) can be controlled in three dimensions.

[0403] Each of the second layer 202 and the third layer 203 has the same structure as the first layer 201 shown in FIG.

[0404] In the array system 200, each of the first layer 201, the second layer 202 and the third layer 203 has a configuration in which a memristor 116 is arranged at the intersection of the nanowires 111, 112, 113 and 114, and the distribution of dopants (oxygen vacancies) in the memristor 116 is controlled two-dimensionally or three-dimensionally according to the voltage applied by the intersecting nanowires 111, 112, 113 and 114, thereby realizing the function of associative learning. Therefore, the array system 200 shown in Fig. 85 is also an array system with a crossbar structure.

[0405] Note that a crossbar-structured array system can be constructed in a similar manner when a memristor with three terminals, a memristor with five terminals, or a memristor with seven or more terminals is used. Therefore, the array system according to the embodiment of the present invention only needs to include a memristor with three or more terminals.

[0406] Although the array systems 100 and 200 have been described as including three layers, the present invention is not limited to this, and the array systems 100 and 200 may include one or more layers.

[0407] Fig. 87 is a diagram showing a specific example of an array system having a planar four-terminal crossbar structure. Fig. 87(a) shows a conceptual diagram of a neural network, and Fig. 87(b) is a conceptual diagram of an array system having a planar four-terminal crossbar structure.

[0408] 87(a), in the neural network, an input signal line, an output signal line, and a modification signal line are arranged between a presynaptic cell body-like element and a post-synaptic cell body-like element, and a synaptic connection is formed at the intersection of the input signal line, the output signal line, and the modification signal line.

[0409] Referring to FIG. 87(b), the array system of a planar four-terminal crossbar structure includes an input signal line, an output signal line, modified signal lines 1 and 2, a presynaptic cell body-like element, a postsynaptic cell body-like element and a synaptic connection.

[0410] The presynaptic cell body-like element and the postsynaptic cell body-like element are arranged in a plane. An input signal line, an output signal line, and modified signal lines 1 and 2 are connected to one synaptic junction. The input signal line and the output signal line are connected to two opposing terminals of one synaptic junction, and the modified signal line 1 and the modified signal line 2 are connected to two opposing terminals of one synaptic junction.

[0411] Therefore, by placing the memristor 10 at the synaptic junction, the input signal line and the output signal line are respectively connected to terminals 3 and 5 of the memristor 10 shown in Figure 1, and the modified signal line 1 and the modified signal line 2 are respectively connected to terminals 4 and 6 of the memristor 10.

[0412] As a result, by controlling the voltage applied to the memristor 10 using the input signal line and the modifying signal lines 1 and 2, the distribution of dopants (oxygen vacancies) in the memristor 10 can be controlled two-dimensionally, and by controlling the two-dimensional distribution of dopants (oxygen vacancies), the strength of the coupling between the input signal line and the output signal line can be controlled.

[0413] In this way, a neural network can be realized by fabricating an array system with a planar four-terminal crossbar structure using memristors 10.

[0414] 88 is a diagram showing a specific example of an array system having a vertical four-terminal crossbar structure. Referring to FIG. 88, an array system 300 includes drive terminal wiring bars 301 and 302, qualification terminal wiring bars 303 and 304, and a memristor 305.

[0415] The driving terminal wiring bar 301 is arranged along the Y-axis direction, the driving terminal wiring bar 302 is arranged along the X-axis direction, and the modification terminal wiring bars 303 and 304 are arranged in a direction that forms a predetermined angle with the X-axis or Y-axis in the XY plane.

[0416] The memristor 305 is arranged along the Z-axis direction, and both ends in the Z-axis direction are connected to the drive terminal wiring bar 301 and the drive terminal wiring bar 302. In addition, the memristor 305 is connected to the modifier terminal wiring bars 303 and 304 between the modifier terminal wiring bar 303 and the modifier terminal wiring bar 304.

[0417] The memristor 305 is, for example, a memristor 20. Thus, the driving terminal wiring bar 301 is connected to the terminal 24 of the memristor 20, the driving terminal wiring bar 302 is connected to the terminal 23 of the memristor 20, the modifying terminal wiring bar 303 is connected to the terminal 25 of the memristor 20, and the modifying terminal wiring bar 304 is connected to the terminal 26 of the memristor 20.

[0418] In the array system 300, a plurality of driving terminal wiring bars 301 are arranged at a predetermined interval in the XY plane, a pair of modifier terminal wiring bars 303, 304 are arranged on the plurality of driving terminal wiring bars 301 at a predetermined interval so as to form a predetermined angle with the driving terminal wiring bars 301, 302, a memristor 305 is arranged on the driving terminal wiring bar 301 so as to be sandwiched between the pair of modifier terminal wiring bars 303, 304, and the driving terminal wiring bar 302 is arranged on the top layer so as to be in contact with the memristor 305.

[0419] As a result, the density of the memristors 305 in the array system 300 can be controlled by the size of the memristors 305 in the XY plane. That is, by increasing the size of the memristors 305 in the XY plane, the density of the memristors 305 in the array system 300 decreases, and by decreasing the size of the memristors 305 in the XY plane, the density of the memristors 305 in the array system 300 increases.

[0420] As described above, the memristor 305 (=memristor 20) can be fabricated using a semiconductor process, and therefore the size of one memristor 305 (=memristor 20) in the XY plane can be easily reduced.

[0421] Furthermore, by reducing the distance between adjacent drive terminal wiring bars 301, the density of the memristors 305 in the array system 300 can be increased.

[0422] Furthermore, by reducing the distance between the pair of qualification terminal wiring bars 303 and 304, the density of the memristors 305 in the array system 300 can be increased.

[0423] Therefore, in an array system 300 with a vertical four-terminal crossbar structure, the density of the memristors 305 can be easily increased.

[0424] In the above-mentioned first embodiment, it has been described that the distribution of dopants (oxygen vacancies) in an oxide semiconductor layer is controlled two-dimensionally in a memristor having three or more terminals arranged in a plane, and in the above-mentioned second embodiment, it has been described that the distribution of dopants (oxygen vacancies) in an oxide semiconductor layer is controlled three-dimensionally in a memristor having three or more terminals arranged in a three-dimensional manner.

[0425] Therefore, the memristor according to the embodiment of the present invention is a memristor that does not require a dielectric breakdown process, and it is sufficient if it includes an oxide semiconductor layer containing a dopant, and a terminal that applies a voltage to the oxide semiconductor layer and controls the distribution of the dopant to a two-dimensional or three-dimensional distribution.

[0426] This is because if there is an oxide semiconductor layer containing a dopant and a terminal that applies a voltage to the oxide semiconductor layer and controls the distribution of the dopant to various two-dimensional or three-dimensional distributions, the weight of the synaptic connection when two stimuli are given can be realized and the function of associative learning can be demonstrated.

[0427] Moreover, an array system according to an embodiment of the present invention is a crossbar-structured array system comprising a plurality of first nanowires arranged in a plane, a plurality of second nanowires arranged in a plane and intersecting the plurality of first nanowires, a plurality of third nanowires arranged on the plane and intersecting the plurality of first nanowires and the plurality of second nanowires, and a memristor disposed at the intersection of the first nanowires, the second nanowires and the third nanowires, wherein the memristor may be a memristor according to an embodiment of the present invention.

[0428] In this embodiment of the present invention, terminal 3 of memristor 10 constitutes a "first terminal", terminal 5 of memristor 10 constitutes a "second terminal", and terminals 4 and 5 of memristor 10 constitute "n (n is an integer greater than or equal to 1) third terminals".

[0429] Furthermore, terminal 23 of memristor 20 constitutes a "first terminal", terminal 24 of memristor 20 constitutes a "second terminal", and terminals 25 and 26 of memristor 20 constitute "n (n is an integer greater than or equal to 1) third terminals."

[0430] Furthermore, terminal 23A of memristor 20A constitutes a "first terminal", terminal 24A of memristor 20A constitutes a "second terminal", and terminals 25 and 26 of memristor 20A constitute "n (n is an integer greater than or equal to 1) third terminals."

[0431] Furthermore, terminal 23A of memristor 20A constitutes a "first terminal", terminal 24A of memristor 20A constitutes a "second terminal", and terminals 25 and 26 of memristor 20A constitute "n (n is an integer greater than or equal to 1) third terminals."

[0432] Furthermore, terminal 23 of memristor 20B constitutes a "first terminal", terminal 24 of memristor 20B constitutes a "second terminal", and terminals 25 and 26 of memristor 20B constitute "n (n is an integer greater than or equal to 1) third terminals."

[0433] Furthermore, terminal 42 of memristor 40 constitutes a "first terminal", terminal 43 of memristor 40 constitutes a "second terminal", and terminals 44 and 45 of memristor 40 constitute "n (n is an integer greater than or equal to 1) third terminals".

[0434] The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present invention is defined by the claims, not by the description of the embodiments described above, and is intended to include all modifications within the meaning and scope of the claims. [Industrial Applicability]

[0435] The present invention is applicable to a memristor, a semiconductor device including the memristor, and an array system including the memristor. [Explanation of symbols]

[0436] 1, 21, 41 Substrate 2, 27, 46 Oxide semiconductor layer, 3-6, 23-26, 23A, 24A, 42-45 Terminal, 10, 20, 20A, 20B, 40, 115, 116, 305 Memristor, 22, 28, 29 Insulating film, 25-1, 25-3, 26-1, 26-3 Insulating layer, 25-2, 26-2 Pt, 28-1, 29-1 Through hole, 46-1, 46-2 Single crystal TiO 2-x , 100, 200, 300 array system, 101, 201 first layer, 102, 202 second layer, 103, 203 third layer, 111 to 113, 113a, 113b, 114, 114a, 114b nanowires, 301, 302 drive terminal wiring bar, 303, 304 modified terminal wiring bar.

Claims

1. A memristor that does not require a dielectric breakdown process, an oxide semiconductor layer including a dopant; a terminal for applying a voltage to the oxide semiconductor layer and controlling a distribution of the dopant to a three-dimensional distribution; the terminals include three or more terminals that apply a voltage to the oxide semiconductor layer and control the distribution of the dopant to a three-dimensional distribution; The three or more terminals are a first terminal disposed on one surface of the oxide semiconductor layer in a thickness direction; a second terminal disposed on the other surface of the oxide semiconductor layer in the thickness direction; and n (n is an integer of 1 or more) third terminals arranged on side surfaces parallel to a thickness direction of the oxide semiconductor layer.

2. a first insulating layer having a first through hole having an area smaller than a contact area between the first terminal and the oxide semiconductor layer, and being disposed between the first terminal and the oxide semiconductor layer; a second insulating layer having a second through hole having an area smaller than a contact area between the second terminal and the oxide semiconductor layer and disposed between the second terminal and the oxide semiconductor layer; the first terminal is in contact with the oxide semiconductor layer through the first through hole; The memristor according to claim 1 , wherein the second terminal is in contact with the oxide semiconductor layer through the second through hole.

3. the first terminal has a T-shaped cross-sectional shape in a cross section of the memristor; The memristor according to claim 1 , wherein the second terminal has a dimension equal to a width of the vertical bar portion of the T in a direction perpendicular to a thickness direction of the oxide semiconductor layer.

4. The memristor according to claim 1 , wherein the oxide semiconductor layer is made of amorphous gallium oxide containing an amount of oxygen that is less than a stoichiometric ratio.

5. 4. The memristor according to claim 1, wherein the oxide semiconductor layer is made of single-crystal titanium oxide containing an amount of oxygen less than a stoichiometric ratio and having a crystal plane in a (001) direction.

6. A crossbar-structured array system, comprising: A plurality of first nanowires arranged in a plane; a plurality of second nanowires arranged in a plane and intersecting the plurality of first nanowires; a plurality of third nanowires arranged in a plane and intersecting the plurality of first nanowires and the plurality of second nanowires; a memristor disposed at an intersection of the first nanowire, the second nanowire, and the third nanowire; The memristor of claim 1 , wherein the memristor is an array system.

7. A memristor that does not require a dielectric breakdown process and A control circuit for controlling the memristor; The memristor comprises: an oxide semiconductor layer including a dopant; a terminal that applies a voltage to the oxide semiconductor layer and controls a distribution of the dopant in a plan view to a two-dimensional distribution; The terminal is a first terminal disposed at a first end of the oxide semiconductor layer in a first direction parallel to an in-plane direction of the oxide semiconductor layer; a second terminal disposed at a second end of the oxide semiconductor layer facing the first end in the first direction; a third terminal disposed at a third end of the oxide semiconductor layer in a second direction that is an in-plane direction of the oxide semiconductor layer and different from the first direction; a fourth terminal disposed at a fourth end of the oxide semiconductor layer facing the third end in the second direction, the control circuit performs any one of a first voltage control, a second voltage control, a third voltage control, and a fourth voltage control on the memristor; the first voltage control is a voltage control for implementing spike-timing dependent plasticity in the memristor; the second voltage control is a voltage control for implementing Pavlovian conditioning in the memristor; the third voltage control is a voltage control for implementing in the memristor a short-term potentiation in which a coupling strength enhanced by an input is maintained for only a first period and then decays, and a long-term potentiation in which the enhanced coupling strength is maintained for a period longer than the first period; The fourth voltage control is voltage control for implementing low conductance and high conductance in the oxide semiconductor layer in the memristor.

8. The control circuit, in the first voltage control, a first step of applying a pulse voltage to the first terminal while the second terminal is grounded to initialize a conductance of the oxide semiconductor layer between the first terminal and the second terminal; a second step of applying a read voltage for reading a conductance of the oxide semiconductor layer to the first terminal while the second terminal is grounded after the first step; a third step of applying a first spike voltage, which is a result of subtracting a voltage value of a second pulse voltage, which is a square wave and has a rise timing different from that of the first pulse voltage, from a voltage value of a first pulse voltage, which is a square wave, to the first terminal while the second terminal is grounded; a fourth step of applying the read voltage to the first terminal while grounding the second terminal after the third step; a fifth step of executing the first step after the fourth step; a sixth step of applying the first spike voltage to the first terminal while the second terminal is grounded after the fifth step, and applying a second spike voltage, which is a result of subtracting a voltage value of the second pulse voltage from a voltage value of a third pulse voltage which is the square wave and different from the first pulse voltage, to the third and fourth terminals, or not applying the second spike voltage to the third and fourth terminals; 8. The semiconductor device according to claim 7, further comprising a seventh step of repeatedly executing the first step to the sixth step one or more times while changing a difference between a rising timing of the first pulse voltage and a rising timing of the second pulse voltage.

9. The control circuit, in the first step, a first substep of applying a read voltage to the first terminal while grounding the second terminal; a second substep of applying a fourth pulse voltage to the third and fourth terminals while the first and second terminals are grounded, after the first substep; a third substep of executing the first substep after the second substep; 9. The semiconductor device according to claim 8, further comprising a fourth substep of applying a fifth pulse voltage to the first terminal while the second terminal is grounded, after the third substep.

10. 10. The semiconductor device according to claim 8, wherein the control circuit repeats the first to sixth steps one or more times while increasing an amplitude of the second spike voltage in the sixth step in the seventh step.

11. The control circuit, in the second voltage control, a first step of initializing the first and second terminals by applying a positive voltage to the first and second terminals; a second step of applying a first write voltage to the first terminal a plurality of times after the first step; 8. The semiconductor device according to claim 7, further comprising a third step of applying the first write voltage to the first terminal and applying a second write voltage different from the first write voltage to the fourth terminal, after the second step.

12. The control circuit, in the third voltage control, a first step of applying a first pulse voltage to the first terminal a plurality of times while the second terminal is grounded; a second step of applying a first voltage having a positive voltage value to the first terminal while the second terminal is grounded, after the first step; 8. The semiconductor device according to claim 7, further comprising a third step of applying a second voltage, which is a negative voltage, to the third and fourth terminals with the second terminal grounded, after the second step.

13. 13. The semiconductor element of claim 12, wherein in the second step, the control circuit applies the first voltage to the first terminal with the second terminal grounded at a first frequency and at a second frequency greater than the first frequency.

14. In the fourth voltage control, the control circuit a first step of applying a first pulse voltage between the third terminal and the fourth terminal while the first and second terminals are grounded; a second step of applying a second pulse voltage different from the first pulse voltage to the first terminal a plurality of times while the second terminal is grounded after the first step, to set the conductance of the oxide semiconductor layer to a first value; a third step of repeatedly applying the second pulse voltage to the first terminal while keeping the second terminal grounded, and applying a third pulse voltage different from the first and second pulse voltages to the third and fourth terminals, or not applying the third pulse voltage to the third and fourth terminals, until a conductance of the oxide semiconductor layer reaches a second value smaller than the first value, while keeping the second terminal grounded; a fourth step of repeatedly applying a fourth pulse voltage obtained by inverting the amplitude of the second pulse voltage to the first terminal while the second terminal is grounded, until the conductance of the oxide semiconductor layer reaches the first value after the third step; 8. The semiconductor device according to claim 7, further comprising a fifth step of repeatedly performing the third and fourth steps after the fourth step.

15. 15. The semiconductor element according to claim 7, wherein the oxide semiconductor layer contains an amount of oxygen that is less than the stoichiometric ratio and is made of single-crystal titanium oxide having a crystal plane in a (001) direction, or amorphous gallium oxide.

Citation Information

Patent Citations

  • Three-dimensional crossbar array system, method for writing information to a three-dimensional crossbar array junction, and method for reading information stored in a three-dimensional crossbar array junction.

    JP2010522987A

  • Memristive device

    US20110182107A1

  • Variable resistance element, semiconductor device having variable resistance element, semiconductor device manufacturing method, and programming method using variable resistance element

    WO2013136798A1