JP6866917B2 - How to operate the reservoir element - Google Patents

Description

The present invention relates to a method of operating a reservoir element.

A neuromorphic element is an element that imitates the human brain by a neural network. Neuromorphic devices artificially mimic the relationship between neurons and synapses in the human brain.

The layered element is one of the neuromorphic elements. Hierarchical elements have chips (neurons in the brain) arranged in a hierarchy and transmission means (synapses in the brain) connecting them. Hierarchical elements increase the percentage of correct answers to questions by learning by means of communication (synapses). Learning is to find knowledge that can be used in the future from information, and the neuromorphic element weights the input data. Hierarchical elements perform learning at each layer.

However, when the number of chips (neurons) increases, the learning at each layer becomes a heavy burden on the circuit design and causes an increase in the power consumption of the neuromorphic element. Reservoir computers are being considered as a way to reduce this burden.

The reservoir computer is one of the neuromorphic elements. The reservoir computer includes a reservoir element and an output unit. The reservoir element consists of a plurality of chips that interact with each other. The plurality of chips interact with each other by the input signal and output the signal. The transmission means that connects a plurality of chips has a fixed weight and does not learn. The output unit learns the signal from the reservoir element and outputs it to the outside. The reservoir computer compresses the data with the reservoir element and weights the data with the output unit to increase the correct answer rate of the question. Learning in the reservoir computer is performed only in the output section. Reservoir computers are expected as one of the means for simplifying the circuit design of neuromorphic devices and improving power consumption.

Non-Patent Document 1 describes a neuromorphic element using a spin torque oscillation (STO) element as a chip (neuron).

Jacob Torrejon et al., Nature, Vol.547, pp.428-432 (2017). S.Takahashi and S.Maekawa, Phys. Rev. B67 (5), 052409 (2003).

However, a neuromorphic element using an STO element as a chip needs to have the same resonance frequency of each STO element. The resonance frequency of the STO elements may vary due to manufacturing errors and the like, and the STO elements may not sufficiently interact with each other. Further, the STO element oscillates by applying a high frequency current in the stacking direction. Applying a high-frequency current for a long time in the stacking direction of the STO element having an insulating layer causes a failure of the STO element.

The present invention has been made in view of the above circumstances, and provides a reservoir element and a neuromorphic element that operate stably.

The present invention provides the following means for solving the above problems.

(1) The reservoir element according to the first aspect is located in the first direction with respect to the spin conductive layer containing a non-magnetic conductor and the spin conductive layer, and is separated from each other in a plan view from the first direction. It is provided with a plurality of ferromagnetic layers arranged so as to be provided with a plurality of via wirings electrically connected to a surface of the spin conductive layer opposite to the ferromagnetic layer.

(2) In the reservoir element according to the above aspect, each of the plurality of ferromagnetic layers may be located at a position overlapping each of the plurality of via wirings in a plan view from the first direction.

(3) The reservoir element according to the above aspect may include a reference potential terminal electrically connected to the spin conduction layer.

(4) In the reservoir element according to the above aspect, the via wiring contains a ferromagnet, and the orientation direction of the magnetization of the ferromagnetic material constituting the via wiring is opposite to the orientation direction of the magnetization of the ferromagnetic layer. You may.

(5) The reservoir element according to the above aspect may further have a first tunnel barrier layer between the spin conduction layer and the plurality of ferromagnetic layers.

(6) The reservoir element according to the above aspect may further have a second tunnel barrier layer between the spin conduction layer and the via wiring.

(7) In the reservoir element according to the above aspect, the distance between two adjacent ferromagnetic layers among the plurality of ferromagnetic layers may be equal to or less than the spin transport length of the material constituting the spin conductive layer.

(8) In the reservoir element according to the above aspect, the distance between two adjacent ferromagnetic layers among the plurality of ferromagnetic layers may be equal to or less than the spin diffusion length of the material constituting the spin conductive layer.

(9) In the reservoir element according to the above aspect, the spin conductive layer may contain a metal or alloy of any element selected from the group consisting of Cu, Ag, and Al.

(10) In the reservoir element according to the above aspect, the spin conduction layer may contain a simple substance or a compound of any element selected from the group consisting of Si, Ge, and C.

(11) In the reservoir element according to the above aspect, the plurality of ferromagnetic layers may be arranged in a hexagonal lattice in a plan view from the first direction.

(12) In the reservoir element according to the above aspect, the plurality of ferromagnetic layers form a plurality of aggregates in which the ferromagnetic layers are densely viewed in a plan view from the first direction, and in the aggregate, the ferromagnetic layers. May be arranged in a hexagonal grid.

(13) The neuromorphic element according to the second aspect is connected to the reservoir element according to the above aspect, an input unit connected to the reservoir element, and the reservoir element, and learns a signal from the reservoir element. It has an output unit.

The reservoir element and the neuromorphic element according to the present embodiment can operate stably.

It is a conceptual diagram of the neuromorphic element which concerns on 1st Embodiment. It is a perspective view of the reservoir element which concerns on 1st Embodiment. It is a side view of the reservoir element which concerns on 1st Embodiment. It is a top view of the reservoir element which concerns on 1st Embodiment. It is a top view of another example of the reservoir element which concerns on 1st Embodiment. It is a top view of another example of the reservoir element which concerns on 1st Embodiment. It is a top view of another example of the reservoir element which concerns on 1st Embodiment. It is sectional drawing which shows the manufacturing method of the reservoir element which concerns on 1st Embodiment. It is sectional drawing which shows the manufacturing method of the reservoir element which concerns on 1st Embodiment. It is sectional drawing which shows the manufacturing method of the reservoir element which concerns on 1st Embodiment. It is sectional drawing which shows the manufacturing method of the reservoir element which concerns on 1st Embodiment. It is a figure which showed another example of the usage mode of the reservoir element which concerns on 1st Embodiment. It is sectional drawing of another example of the reservoir element which concerns on 1st Embodiment. It is a perspective view of the reservoir element which concerns on 2nd Embodiment. It is a side view of the reservoir element which concerns on 3rd Embodiment. It is a side view of the reservoir element which concerns on 4th Embodiment.

Hereinafter, the present embodiment will be described in detail with reference to the drawings as appropriate. In the drawings used in the following description, the featured portion may be enlarged for convenience in order to make the feature easy to understand, and the dimensional ratio of each component may be different from the actual one. The materials, dimensions, etc. exemplified in the following description are examples, and the present invention is not limited thereto, and can be appropriately modified and carried out within the range in which the effects of the present invention are exhibited.

"First embodiment"
FIG. 1 is a conceptual diagram of a neuromorphic element according to the first embodiment. The neuromorphic element 100 has an input unit 20, a reservoir element 10, and an output unit 30. The input unit 20 and the output unit 30 are connected to the reservoir element 10.

The neuromorphic element 100 compresses the signal input from the input unit 20 by the reservoir element 10, weights (learns) the compressed signal by the output unit 30, and outputs the signal to the outside.

The input unit 20 transmits a signal input from the outside to the reservoir element 10. The input unit 20 includes, for example, a plurality of sensors. The plurality of sensors sense information outside the neuromorphic element 100 and input the information to the reservoir element 10 as a signal. The signal may be continuously input to the reservoir element 10 over time, or may be divided into predetermined time domains and input to the reservoir element 10.

The reservoir element 10 has a plurality of chip Cps. Multiple chip Cps interact. The signal input to the reservoir element 10 has a large amount of information. A large amount of information contained in a signal is compressed into necessary information by the interaction of a plurality of chip Cps. The compressed signal is transmitted to the output unit 30. The reservoir element 10 does not learn. That is, the plurality of chip Cps only interact with each other, and the signal transmitted between the plurality of chip Cps is not weighted.

The output unit 30 receives a signal from the chip Cp of the reservoir element 10. The output unit 30 learns. The output unit 30 weights each signal from each chip Cp by learning. The output unit 30 includes, for example, a non-volatile memory. The non-volatile memory is, for example, a magnetoresistive sensor. The output unit 30 outputs a signal to the outside of the neuromorphic element 100.

The neuromorphic element 100 compresses the data with the reservoir element 10 and weights the data with the output unit 30, thereby increasing the correct answer rate of the question.

Further, the neuromorphic element 100 is excellent in power consumption. In the neuromorphic element 100, learning is performed only by the output unit 30. Learning is to adjust the weight of the signal transmitted from each chip Cp. The weight of the signal is determined according to the importance of the signal. Adjusting the signal weights from time to time activates the circuit between the chips Cp. The more active circuits there are, the greater the power consumption of the neuromorphic element 100. The neuromorphic element 100 only needs to learn the output unit 30 at the final stage, and is excellent in power consumption.

FIG. 2 is a perspective view of the reservoir element 10 according to the first embodiment. FIG. 3 is a side view of the reservoir element 10 according to the first embodiment. FIG. 4 is a plan view of the reservoir element 10 according to the first embodiment.

The reservoir element 10 includes a plurality of ferromagnetic layers 1, a spin conduction layer 2, and a plurality of via wirings 3. The plurality of ferromagnetic layers 1 correspond to the chip Cp in FIG.

The direction is specified here. Any direction in the spreading plane of the spin conductive layer 2 intersects with the x direction, intersects with the x direction in the spreading plane of the spin conductive layer 2 (for example, substantially orthogonal) in the y direction, and intersects with the spreading plane of the spin conductive layer 2. (For example, substantially orthogonal) is defined as the z direction. The z direction is an example of the first direction.

The spin conduction layer 2 continuously spreads in the xy plane. The spin conductive layer 2 is made of a non-magnetic conductor. The spin conduction layer 2 propagates the spin current exuded from the ferromagnetic layer 1.

The spin conduction layer 2 is made of a material having a long spin diffusion length and a long spin transport length. The spin diffusion length is the distance until the spins injected into the spin conduction layer 2 are diffused and the information of the injected spins is halved. The spin transport length is the distance until the spin current of the spin polarization current flowing in the non-magnetic body is halved. When the voltage applied to the spin conduction layer 2 is small, the spin diffusion length and the spin transport length are substantially the same. On the other hand, when the voltage applied to the spin conduction layer 2 becomes large, the spin transport length becomes longer than the spin diffusion length due to the drift effect.

The spin conductive layer 2 is, for example, a metal or a semiconductor. The metal used for the spin conductive layer 2 is, for example, a metal or an alloy containing any element selected from the group consisting of Cu, Ag, Al, Mg, and Zn. The semiconductor used for the spin conduction layer 2 is, for example, a simple substance or an alloy of any element selected from the group consisting of Si, Ge, and C. For example, Si, Ge, Si-Ge compound, graphene and the like can be mentioned.

The ferromagnetic layer 1 is formed on one surface of the spin conduction layer 2. A plurality of ferromagnetic layers 1 project in the z direction and are separated from each other in the xy plane. A plurality of ferromagnetic layers 1 exist for one spin conduction layer 2. The adjacent ferromagnetic layer 1 is insulated with, for example, an interlayer insulating film.

The plurality of ferromagnetic layers 1 are arranged in a hexagonal lattice pattern in a plan view from the z direction, for example (see FIG. 4). The signal input to the ferromagnetic layer 1 propagates in the spin conduction layer 2 as a spin current. When the ferromagnetic layers 1 are arranged in a hexagonal lattice, the signals input from the ferromagnetic layer 1 are likely to interfere with the signals input from the other ferromagnetic layers 1.

The arrangement of the plurality of ferromagnetic layers 1 is not limited to the case of FIG. 5 and 7 are plan views of another example of the reservoir element according to the first embodiment.

In the reservoir element 10A shown in FIG. 5, a plurality of ferromagnetic layers 1 are arranged in a square lattice pattern. The distances between adjacent ferromagnetic layers 1 are equal and the input signal is homogenized.

In the reservoir element 10B shown in FIG. 6, a plurality of ferromagnetic layers 1 are densely packed in a hexagonal lattice pattern. As the density of the ferromagnetic layer 1 increases, signals input to different ferromagnetic layers 1 are likely to interfere with each other. Even in this case, the ferromagnetic layers 1 are insulated from each other.

The reservoir element 10C shown in FIG. 7 forms a plurality of aggregates A in which the ferromagnetic layers 1 are densely packed. In the aggregate A, the ferromagnetic layers 1 are arranged in a hexagonal lattice pattern. Adjacent ferromagnetic layers 1 are insulated from each other. The reservoir element 10C is composed of mutual interference between signals input to the ferromagnetic layer 1 constituting one aggregate A and mutual interference between signals input to the ferromagnetic layer 1 constituting different aggregates A. , The conditions of mutual interference are different. By adjusting the conditions of mutual interference in the reservoir element 10C, the reservoir element 10C emphasizes a specific signal and transmits it to the output unit 30.

The shape of each ferromagnetic layer 1 is, for example, a columnar shape (see FIG. 1). The shape of the ferromagnetic layer 1 is not limited to a columnar shape. The shape of the ferromagnetic layer 1 may be, for example, an elliptical column, a quadrangular prism, a cone, an elliptical cone, a truncated cone, a truncated cone, or the like.

The ferromagnetic layer 1 contains a ferromagnetic material. The ferromagnetic layer 1 is, for example, a metal selected from the group consisting of Cr, Mn, Co, Fe and Ni, an alloy containing at least one of these metals, and at least one of these metals and B, C, and N. Includes alloys containing the above elements. The ferromagnetic layer 1 is, for example, Co—Fe, Co—Fe—B, Ni—Fe, Co—Ho alloy (CoHo ₂ ), Sm—Fe alloy (SmFe ₁₂ ). When the ferromagnetic layer 1 contains a Co—Ho alloy (CoHo ₂ ) and an Sm—Fe alloy (SmFe ₁₂ ), the ferromagnetic layer 1 tends to be an in-plane magnetized film.

The distance between the two adjacent ferromagnetic layers 1 is preferably less than or equal to the spin transport length of the material constituting the spin conductive layer 2, and preferably less than or equal to the spin diffusion length.

The via wiring 3 is electrically connected to the surface of the spin conduction layer 2 opposite to the ferromagnetic layer 1. The via wiring 3 may be directly connected to the spin conduction layer 2 or may be connected via another layer. A plurality of via wirings 3 shown in FIGS. 1 and 3 project from the spin conduction layer 2 in the −z direction and are separated from each other in the xy plane.

The via wiring 3 includes a conductor. The via wiring 3 is, for example, Cu, Al, Au. The adjacent via wiring 3 is insulated.

Each of the via wirings 3 shown in FIGS. 1 and 3 is arranged at a position corresponding to each of the plurality of ferromagnetic layers 1. That is, each of the ferromagnetic layer 1 and each of the via wiring 3 overlap in a plan view from the z direction.

Next, an example of a method for manufacturing the reservoir element 10 in the neuromorphic element 100 will be described. 8A to 8D are cross-sectional views showing a method of manufacturing the reservoir element 10 according to the first embodiment.

First, a hole is formed in the substrate Sb, and the inside of the hole is filled with a conductor (FIG. 8A). The substrate Sb is, for example, a semiconductor substrate. The substrate Sb is preferably Si or AlTiC, for example. Si and AlTiC can easily obtain a surface having excellent flatness. Holes are formed using, for example, reactive ion etching (RIE). The conductor charged in the hole becomes the via wiring 3.

Next, the surfaces of the substrate Sb and the via wiring 3 are flattened by chemical mechanical polishing (CMP). The spin conduction layer 2 and the ferromagnetic layer 1'are laminated in this order on the flattened substrate Sb and via wiring 3 (FIG. 8B). The spin conductive layer 2 and the ferromagnetic layer 1'are laminated by, for example, chemical vapor deposition (CVD).

Next, a hard mask HM is formed at a predetermined position on the surface of the ferromagnetic layer 1'(FIG. 8C). The portion of the ferromagnetic layer 1'not covered with the hard mask HM is removed by RIE or ion milling. The ferromagnetic layer 1'becomes a plurality of ferromagnetic layers 1 by removing unnecessary portions. Finally, the interlayer insulating film I protects the space between the ferromagnetic layers 1 (FIG. 8D). The neuromorphic element 100 according to the first embodiment is obtained by the above procedure.

Next, the function of the neuromorphic element 100 will be described. The sensor constituting the input unit 20 is connected to the ferromagnetic layer 1 of any one of the reservoir elements 10. When the sensor receives an external signal, a current flows from the corresponding ferromagnetic layer 1 toward the via wiring 3, and the signal is input to the reservoir element 10. When each of the via wires 3 is arranged at a position corresponding to each of the plurality of ferromagnetic layers 1, most of the current flows in the z direction.

The current is spin-polarized by the ferromagnetic layer 1 and reaches the spin conduction layer 2. The electric charge flows through the via wiring 3 and hardly flows in the spin conduction layer 2. A spin current flows in the spin conduction layer 2. That is, spins are injected from the ferromagnetic layer 1 into the spin conductive layer 2 in the vicinity of the ferromagnetic layer 1, and spins are accumulated in the spin conductive layer 2. The accumulated spin propagates in the spin conduction layer 2 as a spin current.

The amount of spin accumulated in the vicinity of the ferromagnetic layer 1 changes depending on the application time and the amount of current applied, and the spin transport length also changes. When the amount of current applied is large, the spin transport length becomes long, and the spin current propagates from the vicinity of the ferromagnetic layer 1 to a wide range.

When a current flows from the plurality of ferromagnetic layers 1 toward the via wiring 3, the spin current propagates so as to spread in the spin conduction layer 2 from each position in the vicinity of the ferromagnetic layer 1 to which the current is applied. Spin currents propagating from different positions interfere with each other. The spin lifetime is several hundred psec in the case of metals such as Ta and Pt, and several nsec in the case of semiconductors such as Si. The information on the spin injected into the spin conduction layer 2 becomes unreadable in about several hundred psec to several nsec.

Finally, a signal is output from the reservoir element 10 to the output unit 30. The signal is read out as a potential difference between the via wiring 3 and the ferromagnetic layer 1. No current flows in the spin conduction layer 2, but a spin current flows. When a spin current is generated, the potential of the spin conductive layer 2 with respect to the spin of the ferromagnetic layer 1 changes, and a potential difference is generated. The potential difference is read out as a potential difference between the reference potential and each ferromagnetic layer 1 with any via wiring 3 as a reference potential.

The potential of the spin conduction layer 2 in the vicinity of each ferromagnetic layer 1 is affected by spin currents spreading from different positions. The signal read from one ferromagnetic layer 1 as a potential difference includes the information written in the other ferromagnetic layer 1, and the information is compressed.

The last compressed signal is transmitted to the output unit 30. The output unit 30 weights the signals read from each of the ferromagnetic layers 1 by learning.

As described above, in the reservoir element 10 according to the first embodiment, the spin currents propagating from the respective ferromagnetic layers 1 interfere with each other in the spin conduction layer 2. The signals input from the input unit 20 interfere with each other in the spin conduction layer 2 and create a specific state in the spin conduction layer 2. That is, the signal input from the input unit 20 is compressed into one state in the spin conduction layer 2. Therefore, the neuromorphic element 100 according to the first embodiment appropriately compresses the signal with the reservoir element 10. By compressing the signal, only the output unit 30 needs to learn, and the power consumption of the neuromorphic element 100 is reduced.

Further, the reservoir element 10 according to the first embodiment can be changed in various ways.

FIG. 9 is a perspective view of another example of the reservoir element according to the first embodiment. The reservoir element 10D shown in FIG. 9 has a reference potential terminal 3G, and the ferromagnetic layer 1 is divided into an input terminal 1A and an output terminal 1B.

The reference potential terminal 3G is electrically connected to the spin conduction layer 2. The reference potential terminal 3G is preferably located at a position sufficiently distant from each output terminal 1B. The reference potential terminal 3G is made of the same material as the via wiring 3.

The ferromagnetic layer 1 is divided into an input terminal 1A for inputting a signal and an output terminal 1B for outputting a signal. When a current flows from each input terminal 1A to the via wiring 3, a spin current flows in the spin conduction layer 2 and interferes with each other. The output terminal 1B outputs the difference in potential between the spin in the spin conduction layer 2 in the vicinity of the output terminal 1B and the magnetization of the output terminal 1B at a certain moment as a potential difference. With reference to the reference potential terminal 3G, the potentials V1, V2, and V3 of the respective output terminals 1B are measured. The potentials V1, V2, and V3 are output signals. By fixing the reference potential by the reference potential terminal 3G, relative evaluation of the potentials V1, V2, and V3 becomes possible.

The shortest distance between the input terminal 1A and the output terminal 1B is preferably less than or equal to the spin transport length of the material constituting the spin conduction layer 2, and preferably less than or equal to the spin diffusion length. When the spin current sufficiently propagates to the output terminal 1B, the SN (Signal / Noise) ratio of the output signal is improved.

When the input terminal 1A and the output terminal 1B are separated, the via wiring 3 may not be provided at a position facing the output terminal 1B. Further, as shown in FIG. 10, the via wirings 3 facing the input terminal 1A or the output terminal 1B may be connected to each other by the common electrode layer 3C.

"Second embodiment"
FIG. 11 is a cross-sectional view of the reservoir element according to the second embodiment. The reservoir element 11 according to the second embodiment is different from the reservoir element 10 according to the first embodiment in that the via wiring 3 m contains a magnetic material. Other configurations are the same as those of the reservoir element 10 according to the first embodiment, and the description thereof will be omitted. Further, in FIG. 11, the same components as those in FIG. 1 are designated by the same reference numerals.

The via wiring 3 m contains a magnetic material. The via wiring 3m may include a magnetic material at a position close to the spin conduction layer 2. The via wiring 3m may have a structure in which the ferromagnetic layer and the conductive layer are laminated in order from a position close to the spin conductive layer 2, for example. As the magnetic material, the same material as that of the ferromagnetic layer 1 can be used.

The orientation direction of the magnetization of the via wiring 3 m is opposite to the orientation direction of the magnetization of the ferromagnetic layer 1. When a current is passed between the ferromagnetic layer 1 having different magnetization orientation directions and the via wiring 3 m, spins in the same direction can be efficiently injected into the spin conduction layer 2.

An example will be described in which the magnetization of the ferromagnetic layer 1 is oriented in the + x direction and the magnetization of the via wiring 3 m is oriented in the −x direction. The current flows in the order of, for example, the ferromagnetic layer 1, the spin conduction layer 2, and the via wiring 3 m. When a current flows from the ferromagnetic layer 1 to the spin conductive layer 2, spins in the −x direction from the ferromagnetic layer 1 are injected into the spin conductive layer 2. On the other hand, when a current flows from the spin conduction layer 2 to the via wiring 3m, the magnetization of the via wiring 3m is oriented in the −x direction, so that spin in the −x direction is generated from the via wiring 3m to the spin conduction layer 2. It flows. Therefore, when the via wiring 3m contains a ferromagnet, spins in the same direction can be efficiently injected into the spin conduction layer 2.

The reservoir element 11 according to the second embodiment can be applied to the neuromorphic element 100. Further, the reservoir element 11 according to the second embodiment has the same effect as the reservoir element 10 according to the first embodiment. Further, the reservoir element 11 according to the second embodiment efficiently supplies spin to the spin conduction layer 2. Therefore, the interference of the spin current in the spin conduction layer 2 is promoted, and the reservoir element 11 can express a more complicated phenomenon.

"Third embodiment"
FIG. 12 is a cross-sectional view of the reservoir element according to the third embodiment. The reservoir element 12 according to the third embodiment is different from the reservoir element 10 according to the first embodiment in that it has the first tunnel barrier layer 4. Other configurations are the same as those of the reservoir element 10 according to the first embodiment, and the description thereof will be omitted. Further, in FIG. 10, the same components as those in FIG. 1 are designated by the same reference numerals.

The first tunnel barrier layer 4 is located between the ferromagnetic layer 1 and the spin conduction layer 2. The first tunnel barrier layer 4 extends continuously in, for example, the xy plane. The first tunnel barrier layer 4 may be scattered in the xy plane only at a position between the ferromagnetic layer 1 and the spin conduction layer 2.

The first tunnel barrier layer 4 is made of a non-magnetic insulator. The first tunnel barrier layer 4 is, for example, Al ₂ O ₃ , SiO ₂ , Mg O _{, Mg Al 2 O 4,} or the like. Further, the first tunnel barrier layer 4 may be a material or the like in which a part of Al, Si, Mg in the above material is replaced with Zn, Be or the like. MgO and MgAl ₂ O ₄ can realize a coherent tunnel phenomenon between the ferromagnetic layer 1 and the spin conductive layer 2, and can efficiently inject spin from the ferromagnetic layer 1 into the spin conductive layer 2.

The thickness of the first tunnel barrier layer 4 is preferably less than 3 nm. When the resistance of the thickness of the first tunnel barrier layer 4 is high, the backflow of the spin current from the spin conduction layer 2 can be suppressed. However, when the thickness of the first tunnel barrier layer 4 is 3 nm or more, the spin scattering effect of the first tunnel barrier layer 4 as a spin filter does not increase, only the resistance increases, and noise increases.

The first tunnel barrier layer 4 has a larger spin resistance than the spin conduction layer 2. Spin resistance is a quantity that quantitatively indicates the ease with which a spin current flows (difficulty with spin relaxation).

Spin resistance Rs is defined by the following equation (see Non-Patent Document 1).

Here, λ is the spin diffusion length of the material, ρ is the electrical resistivity of the material, and A is the cross-sectional area of the material.
For non-magnetic materials, when the cross-sectional areas A are equal, the magnitude of spin resistance is determined by the value of ρλ, which is the spin resistivity in the equation (1).

The spin tries to flow to the part where the spin resistance is low. Since the first tunnel barrier layer 4 contains an insulator, it has a large electrical resistivity and a large spin resistance. The first tunnel barrier layer 4 suppresses the spin reaching the spin conduction layer 2 from returning to the ferromagnetic layer 1.

The reservoir element 12 according to the third embodiment can be applied to the neuromorphic element 100. Further, the reservoir element 12 according to the third embodiment has the same effect as the reservoir element 10 according to the first embodiment. Further, the reservoir element 12 according to the third embodiment can efficiently generate a spin current by the spin injected into the spin conduction layer 2. Therefore, the interference of the spin current in the spin conduction layer 2 is promoted, and the reservoir element 12 can express a more complicated phenomenon.

"Fourth embodiment"
FIG. 13 is a cross-sectional view of the reservoir element according to the fourth embodiment. The reservoir element 13 according to the fourth embodiment is different from the reservoir element 12 according to the third embodiment in that it has a second tunnel barrier layer 5. Other configurations are the same as those of the reservoir element 12 according to the third embodiment, and the description thereof will be omitted. Further, in FIG. 13, the same components as those in FIG. 12 are designated by the same reference numerals.

The second tunnel barrier layer 5 is located between the spin conduction layer 2 and the via wiring 3. The second tunnel barrier layer 5, for example, extends continuously in the xy plane. The second tunnel barrier layer 5 may be scattered in the xy plane only at a position between the ferromagnetic layer 1 and the spin conduction layer 2.

The second tunnel barrier layer 5 is made of a non-magnetic insulator. The second tunnel barrier layer 5 is made of the same material as the first tunnel barrier layer 4. The thickness of the second tunnel barrier layer 5 is equivalent to the thickness of the first tunnel barrier layer 4.

The second tunnel barrier layer 5 has a larger spin resistance than the spin conduction layer 2. The second tunnel barrier layer 5 suppresses the spin reaching the spin conduction layer 2 from flowing to the via wiring 3.

The reservoir element 13 according to the fourth embodiment can be applied to the neuromorphic element 100. Further, the reservoir element 13 according to the fourth embodiment has the same effect as the reservoir element 10 according to the first embodiment. Further, the reservoir element 13 according to the fourth embodiment can efficiently generate a spin current by the spin injected into the spin conduction layer 2. Therefore, the interference of the spin current in the spin conduction layer 2 is promoted, and the reservoir element 13 can express a more complicated phenomenon.

Although an example of a preferred embodiment of the present invention has been described in detail above, the present invention is not limited to this embodiment, and is within the scope of the gist of the present invention described in the claims. Various modifications and changes are possible.

For example, the characteristic configurations of the reservoir element 10 according to the first embodiment to the reservoir element 13 according to the fourth embodiment may be combined.

1, 1'ferromagnetic layer 1A Input terminal 1B Output terminal 2 Spin conductive layer 3, 3m Via wiring 3C Common electrode layer 3G Reference potential terminal 4 First tunnel barrier layer 5 Second tunnel barrier layer 10, 10A, 10B, 10C, 10D, 10E, 11, 12, 13 Reservoir element 20 Input unit 30 Output unit 100 Neuromorphic element A Aggregate Cp chip HM Hardmask I Interlayer insulating film Sb substrate

Claims (4)

It is an operation method of a reservoir element having an input unit, a magnetic reservoir unit, and an output unit.
An input signal having a plurality of pieces of information is input to the magnetic reservoir unit from the input unit,
The input signal input to the magnetic reservoir section interacts between a plurality of chips in the magnetic reservoir section, and the input signal is interacted with each other.
The signals interacting in the magnetic reservoir section are sent to the non-volatile memory of the output section that holds the data weighted by learning .
The magnetic reservoir portion is
A spin conductive layer containing a non-magnetic conductor and
A plurality of ferromagnetic layers located in the first direction with respect to the spin conduction layer and arranged apart from each other in a plan view from the first direction.
A plurality of via wirings electrically connected to the ferromagnetic layer of the spin conductive layer are provided.
A method of operating a reservoir element, in which a potential difference between the ferromagnetic layer and the via wiring is output as a signal to the output unit. The method of operating a reservoir element according to claim 1, wherein the input signal is continuously input over time. The operation method of the reservoir element according to claim 1, wherein the input signal is divided by a time domain and input. The method of operating a reservoir element according to any one of claims 1 to 3, wherein the non-volatile memory is a magnetoresistive sensor.

Images