CN109768065A - A kind of semiconductor device cell and pattern recognition device - Google Patents
A kind of semiconductor device cell and pattern recognition device Download PDFInfo
- Publication number
- CN109768065A CN109768065A CN201910031809.3A CN201910031809A CN109768065A CN 109768065 A CN109768065 A CN 109768065A CN 201910031809 A CN201910031809 A CN 201910031809A CN 109768065 A CN109768065 A CN 109768065A
- Authority
- CN
- China
- Prior art keywords
- photodiode
- random access
- access memory
- spin
- semiconductor device
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The present invention provides a kind of semiconductor device cell and pattern recognition device, including photodiode and magnetoresistive RAM, memory can be spin transfer torque magnetoresistive random access memory or spin(-)orbit square magnetic random access memory, the hearth electrode of photodiode and magnetoresistive RAM is electrically connected to each other, and the top electrode of photodiode and magnetoresistive RAM is electrically connected to each other.Magnetoresistive RAM has the characteristic for being repeatedly written and permanently storing infinitely, it can be used for the weight calculation of artificial neural network, photo-signal is generated after photodiode is photosensitive, the size of photoelectric current is similar to the magnetoresistive RAM reading magnitude of electric current, photoelectric current flows into magnetoresistive RAM, so as to be used for artificial neural network image recognition.Since photodiode is to be not necessarily to additional power supply from driving PN junction device, low energy consumption and can be applied in the particular surroundings of power-off, meanwhile, directly analog signal is identified, realizes more intelligent image recognition.
Description
Technical field
The present invention relates to semiconductor devices and its manufacturing field, in particular to a kind of semiconductor device cell and image are known
Other device.
Background technique
Image recognition is a key areas of artificial intelligence technology, currently, after collecting the image to be identified, it should
Image is sent in processor in the form of digital signal, and recognizer is preset in processor, utilizes identification in the processor
Algorithm carries out Object identifying to image.However, processor needs external power supply that can just work, in some particular surroundings,
Such as field or space etc., it can not be applied.
Summary of the invention
In view of this, being utilized the purpose of the present invention is to provide a kind of semiconductor device cell and pattern recognition device
Semiconductor devices realizes image recognition.
To achieve the above object, the present invention has following technical solution:
A kind of semiconductor device cell, comprising:
Substrate;
The photodiode and spin transfer torque magnetoresistive random access memory being located on the substrate, the photoelectricity
Diode includes the hearth electrode stacked gradually from the bottom to top, the first doping type semiconductor layer, the second doping type semiconductor layer
And top electrode, the spin transfer torque magnetoresistive random access memory include the hearth electrode stacked gradually from the bottom to top, magnetic tunnel
Knot and top electrode;
Wherein, the hearth electrode of the photodiode is electrically connected to the spin transfer torque magnetoresistive by the first electrical wiring
The top electrode of the hearth electrode of random access memory, the photodiode is electrically connected the spin transfer torque by the second electrical wiring
Switching device is connected in the top electrode of magnetoresistive RAM, first electrical wiring and/or second electrical wiring.
Optionally, the magnetic tunnel junction includes the first magnetosphere stacked gradually from the bottom to top, tunnel layer, the second magnetism
Layer and protective layer.
Optionally, the magnetic tunnel junction further includes the first pinning layer and/or the second magnetism under first magnetosphere
The second pinning layer between layer and protective layer.
Optionally, the photodiode is visible light, infrared or ultraviolet photodiode.
A kind of semiconductor device cell, comprising:
Substrate;
The photodiode and spin(-)orbit square magnetic random access memory being located on the substrate, the photoelectricity
Diode includes the hearth electrode stacked gradually from the bottom to top, the first doping type semiconductor layer, the second doping type semiconductor layer
And top electrode, the spin(-)orbit square magnetic random access memory include the hearth electrode stacked gradually from the bottom to top, spin(-)orbit
Coupling layer, magnetoresistive tunnel junction and top electrode;
Wherein, the hearth electrode of the photodiode is electrically connected to the spin(-)orbit square reluctance type by the first electrical wiring
The top electrode of the hearth electrode of random access memory, the photodiode is electrically connected the spin(-)orbit square magnetic by the second electrical wiring
Switching device is connected in the top electrode of resistive random access memory, first electrical wiring and/or second electrical wiring.
Optionally, the magnetoresistive tunnel junction includes the first magnetosphere stacked gradually from the bottom to top, tunnel layer, the second magnetism
Layer and protective layer.
Optionally, the magnetoresistive tunnel junction further includes the pinning layer between second magnetosphere and the protective layer.
Optionally, the photodiode is visible light, infrared or ultraviolet photodiode.
A kind of pattern recognition device, including multiple any of the above-described semiconductor device cells, multiple semiconductor devices
Spin transfer torque magnetoresistive random access memory in unit is used to store the weight of artificial neural network, the photodiode
Photosignal is the input signal of image recognition model.
A kind of pattern recognition device, including multiple any of the above-described semiconductor device cells, multiple semiconductor devices
Spin(-)orbit square magnetic random access memory in unit is used to store the weight of artificial neural network, the photodiode
Photosignal is the input signal of image recognition model.
Semiconductor device cell and pattern recognition device provided in an embodiment of the present invention, including photodiode and magnetic resistance
Random access memory, magnetoresistive RAM can for spin transfer torque magnetoresistive random access memory or spin(-)orbit square reluctance type with
The hearth electrode of machine memory, photodiode and magnetoresistive RAM is electrically connected to each other, and photodiode and magnetic resistance are deposited at random
The top electrode of reservoir is electrically connected to each other.Since magnetoresistive RAM has the spy for being repeatedly written and permanently storing infinitely
Property, it can be used for storing the calculating weight of artificial neural network, generation photo-signal after photodiode is photosensitive, and photoelectric current
Size and magnetoresistive RAM read that the magnitude of electric current is similar, and photoelectric current flows into magnetoresistive RAM, so as to
In artificial neural network image recognition.Due to photodiode be from driving PN junction device, be not necessarily to additional power supply, low energy consumption and
It can be applied in the particular surroundings of power-off, meanwhile, directly analog signal is identified, realizes that more intelligent image is known
Not.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 shows the schematic perspective view of according to embodiments of the present invention one semiconductor device cell;
Fig. 2 shows the schematic perspective views of according to embodiments of the present invention two semiconductor device cell.
Specific embodiment
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing to the present invention
Specific embodiment be described in detail.
In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, but the present invention can be with
Implemented using other than the one described here other way, those skilled in the art can be without prejudice to intension of the present invention
In the case of do similar popularization, therefore the present invention is not limited by the specific embodiments disclosed below.
Secondly, combination schematic diagram of the present invention is described in detail, when describing the embodiments of the present invention, for purposes of illustration only, table
Show that the sectional view of device architecture can disobey general proportion and make partial enlargement, and the schematic diagram is example, is not answered herein
Limit the scope of protection of the invention.In addition, the three-dimensional space of length, width and depth should be included in actual fabrication.
As the description in background technique, currently, when carrying out image recognition, it, should after collecting the image to be identified
Image is sent in processor in the form of digital signal, and recognizer is preset in processor, utilizes identification in the processor
Algorithm carries out Object identifying to image.However, processor needs external power supply that can just work, in some particular surroundings,
Such as field or space etc., it can not be applied.
For this purpose, present applicant proposes a kind of semiconductor device cell and pattern recognition device, including photodiode and
Magnetoresistive RAM, magnetoresistive RAM can be and spin transfer torque magnetoresistive random access memory (STT MRAM) or from
It revolves orbital moment magnetic random access memory (SOT MRAM), the hearth electrode of photodiode and magnetoresistive RAM is mutually electrically connected
It connects, the top electrode of photodiode and magnetoresistive RAM is electrically connected to each other.Since magnetoresistive RAM has infinitely
The characteristic for being repeatedly written and permanently storing, can be used for storing the calculating weight of artificial neural network, and photodiode sense
Photo-signal is generated after light, and the size of photoelectric current is similar to the magnetoresistive RAM reading magnitude of electric current, photoelectric current stream
Enter magnetoresistive RAM, so as to be used for artificial neural network image recognition.Since photodiode is the PN from driving
Junction device, be not necessarily to additional power supply, low energy consumption and can be applied to power-off particular surroundings in, meanwhile, directly to analog signal into
Row identification, realizes more intelligent image recognition.
The technical solution and technical effect of the application in order to better understand carries out detailed below with reference to specific embodiment
Thin description.
Embodiment one
Refering to what is shown in Fig. 1, in the present embodiment, which includes:
Substrate 100;
The photodiode 110 and spin transfer torque magnetoresistive random access memory being located on the substrate 100
120, the photodiode 110 include stack gradually from the bottom to top hearth electrode 112, the first doping type semiconductor layer 114,
Second doping type semiconductor layer 116 and top electrode 118, the spin transfer torque magnetoresistive random access memory 120 includes under
The supreme hearth electrode 122 stacked gradually, magnetic tunnel junction and top electrode 134;
Wherein, the hearth electrode 112 of the photodiode 110 is electrically connected to the spinning by the first electrical wiring 140
The top electrode 118 of the hearth electrode 122 of transfer torque magnetoresistive random access memory 120, the photodiode 110 passes through the second electrical wiring
The top electrode 134 of the 150 electrical connection spin transfer torque magnetoresistive random access memory 120, first electrical wiring 140 and/or
Switching device 160 is connected in second electrical wiring 150.
In the embodiment of the present application, substrate 100 primarily serves supporting role, substrate 100 can for semiconductor substrate or its
His substrate, semiconductor substrate for example can be Si substrate, Ge substrate, SiGe substrate, SOI (silicon-on-insulator, Silicon On
) or GOI (germanium on insulator, Germanium On Insulator) etc. Insulator.In the present embodiment, substrate 100 can be
Silicon substrate.
In the embodiment of the present application, the first doping type semiconductor layer 114 and the second doping class in photodiode 110
Type semiconductor layer 116 has opposite doping type, and the first doping type semiconductor layer 114 is p type semiconductor layer, then second mixes
Miscellany type semiconductor layer 116 is n type semiconductor layer;First doping type semiconductor layer 114 is n type semiconductor layer, then second mixes
Miscellany type semiconductor layer 116 is p type semiconductor layer, the first doping type semiconductor layer 114 and the second doping type semiconductor layer
116 constitute the structure of P-N junction.First doping type semiconductor layer 114 and the second doping type semiconductor layer 116 can be partly to lead
Body material or compound semiconductor material, according to the wavelength requirement of different photosensitive wave spectrums, the first doping type semiconductor layer 114
It can be the materials such as silicon, germanium, InGaAsP with the second doping type semiconductor layer 116.
In specific application, photodiode 110 can be for visible light, infrared or ultraviolet photodiode etc., can be with
Separately detect visible light, infrared light and ultraviolet light wave band.
In the present embodiment, the magnetic tunnel junction of spin transfer torque magnetoresistive random access memory 120 is included at least from the bottom to top
The first magnetosphere 126, tunnel layer 128 and the second magnetosphere 130 stacked gradually, more preferably, the first magnetosphere 126 it is lower and/or
It is also provided with pinning layer 124 on second magnetosphere 130, is also provided with protective layer 132 on the second magnetosphere 130.
Wherein, the first magnetosphere 126 and the second magnetosphere 130 can be formed by ferromagnetic material, which can be
Simple substance ferromagnetic material, alloy ferromagnetic material or the magnetic metal oxide of tool etc., such as can be Co, Fe, CoFeB or FePt
Equal hard magnetic materials.According to specific needs, the first magnetosphere 126 and the second magnetosphere 130 can be identical or different material.
Pinning layer 124 is used to fix the direction of magnetization of the first magnetosphere 126, and pinning layer 124 can be antiferromagnet, such as can be
CoPt multilayer film is manually antiferromagnetic etc..Protective layer 132 plays the role of that magnetosphere is prevented to be oxidized, and protective layer 132 usually can be with
It for metal material, such as can be Ta, Ru etc..
The hearth electrode 112 and top electrode 118 and spin transfer torque magnetoresistive random access memory 120 of photodiode 110
Hearth electrode 122 and top electrode 134, can be formed from conductive materials, conductive material injection or is led at metal, conductive metal oxide
Electric metal nitride etc. can be typically Cu.
To be electrically connected to spin transfer torque magnetoresistive random by the first electrical wiring 140 for the hearth electrode 112 of photodiode 110
The top electrode 118 of the hearth electrode 124 of memory 120, photodiode 110 is electrically connected to spinning by the second electrical wiring 150
The top electrode 134 of transfer torque magnetoresistive random access memory 120, in this way, make photodiode 110 and spin transfer torque magnetoresistive with
Machine memory 120 forms concatenated circuit, due to the P-N junction that photodiode 110 is driving certainly, after detecting light,
Photo-signal is generated, and the size of photoelectric current is similar to the magnetoresistive RAM reading magnitude of electric current 120, photoelectric current flows into
Magnetoresistive RAM 120, so as to be used for image recognition.
In the embodiment of the present application, the first electrical wiring and the second electrical wiring can be in above-mentioned photodiode and spin transfers
It is formed, can be realized by one or more layers interconnection structure in torque magnetoresistive random access memory manufacturing process, it can also be upper
After stating photodiode and the manufacture completion of spin transfer torque magnetoresistive random access memory, realized by external electrical wiring.
It is additionally provided with switching device 160 in the first electrical wiring 140 and/or the second electrical wiring 150, switching device 160
Such as can be CMOS transistor device, the switching device 160 for control photodiode 110 and spin transfer torque magnetoresistive with
The on state in the circuit of machine memory 120, when switching device is in an off state, spin transfer torque magnetoresistive random storage
Independently of one another, spin transfer torque magnetoresistive random access memory 120 can be used for carrying out image knowledge for device 120 and photodiode 110
The training of other model, and when switching device 160 in the closure state, photodiode 110 and spin transfer torque magnetoresistive with
The circuit of machine memory 120 is in the conductive state, and received optical signal is converted photoelectric current by photodiode 110, and then defeated
Enter to spin transfer torque magnetoresistive random access memory 120, spin transfer torque magnetoresistive random access memory 120 can be used for passing through light
The identification of current signal progress image.
Based on above-mentioned semiconductor device cell 10, the embodiment of the present application also proposed a kind of pattern recognition device, the figure
As identification device is made of the semiconductor device cell 10 in multiple above-described embodiments, can be by above-mentioned semiconductor device unit
10 chips formed, these semiconductor device cells 10 can be arranged in array, in multiple semiconductor device cells 10
Spin transfer torque magnetoresistive random access memory 120 is used to store the weight of artificial neural network, the light of the photodiode 110
Electric signal is the input signal of image recognition model 120.
In specific application, image recognition model can be the artificial nerve network model based on deep learning, Ke Yigen
The quantity that semiconductor device cell 10 is determined according to the scale of model, when control switch device 160 is in an off state, the image
Spin transfer torque magnetoresistive random access memory 120 in identification device is used to carry out the training of image recognition model, and stores instruction
Weight after white silk;And work as switching device 160 in the closure state, received optical signal is converted light by photodiode 110
Electric current, and then it is input to spin transfer torque magnetoresistive random access memory 120, it is based on spin transfer torque magnetoresistive random access memory
Weight in 120 is calculated, and then carries out image recognition.
In the memory cell and pattern recognition device of the present embodiment, photodiode is the PN junction device of driving certainly,
Without additional power supply, low energy consumption and can be applied in the particular surroundings of power-off, meanwhile, directly analog signal is identified,
Realize more intelligent image recognition.
Embodiment two
What is different from the first embodiment is that constituting device list using spin(-)orbit square magnetic random access memory in the present embodiment
Emphasis, is described the part different from embodiment one by member below, and same section will not be described in great detail.
Refering to what is shown in Fig. 2, in the present embodiment, which includes:
Substrate 200;
The photodiode 210 and spin(-)orbit square magnetic random access memory being located on the substrate 200
220, the photodiode 210 include stack gradually from the bottom to top hearth electrode 212, the first doping type semiconductor layer 214,
Second doping type semiconductor layer 216 and top electrode 218, the spin(-)orbit square magnetic random access memory 220 includes under
The supreme hearth electrode 222 stacked gradually, Quantum geometrical phase layer 224, magnetoresistive tunnel junction and top electrode 234;
Wherein, the hearth electrode 212 of the photodiode 210 is electrically connected to the spin rail by the first electrical wiring 240
The top electrode 218 of the hearth electrode 222 of road square magnetic random access memory 220, the photodiode 210 passes through the second electrical wiring
The top electrode 234 of the 250 electrical connection spin(-)orbit square magnetic random access memories 220, first electrical wiring 240 and/or
Switching device 260 is connected in second electrical wiring 250.
In the present embodiment, substrate 200 primarily serves supporting role.The first doping type in photodiode 210 is partly led
Body layer 214 and the second doping type semiconductor layer 216 have opposite doping type, constitute the structure of P-N junction.Photodiode
210 can be visible light, infrared or ultraviolet photodiode etc., can separately detect visible light, infrared light and ultraviolet light wave band.
Wherein, Quantum geometrical phase layer 224 is made of the material with SO coupling effect, and normally, spin rail
Road coupling layer 224 can be metal layer or topological insulator layer with spin coupling effect, it is preferable that can choose has greatly
The material of Quantum geometrical phase intensity, the material of metal layer for example can for Ta, Pt, W, Hf, Ir, CuBi, CuIr or AuW etc.,
The material of topological insulator layer for example can be BiSn, SnTe, BiSe, wait or other IVA, VA and Group VIA compound in one
Kind.
It is magnetic that magnetoresistive tunnel junction includes at least the first magnetosphere 226 stacked gradually from the bottom to top, tunnel layer 228 and second
Layer 230 more preferably can also be including the pinning layer 232 and pinning layer 232 or the second magnetosphere on the second magnetosphere 230
Protective layer (not shown go out) on 230.
First magnetosphere 226 and the second magnetosphere 230 are formed by the ferromagnetic material with perpendicular magnetic anisotropy, ferromagnetic material
It can be fe magnetic material, alloy ferromagnetic material or the magnetic metal oxide of tool etc., such as can be Co, Fe, CoFeB
Or the hard magnetic materials such as FePt.According to specific needs, the first magnetosphere 226 and the second magnetosphere 230 can be identical or different
Material.
Tunnel layer 228, can be by nonmagnetic metal or insulation material between the first magnetosphere 226 and the second magnetosphere 230
Material is made, and nonmagnetic metal for example can be Cu or Ag, and insulating materials can be for example aluminium oxide, magnesia or hafnium oxide etc..Nail
It pricks layer 232 and is used for fixed magnetisation direction, for ease of description, the pinning layer on second magnetosphere 230 can be denoted as top
Pinning layer, can also be also provided with bottom pinning layer under the first magnetosphere 226, and the material of pinning layer 232 for example can be CoPt
Multilayer film is manually antiferromagnetic etc..Protective layer plays the role of that magnetosphere is prevented to be oxidized, protective layer typically metal material,
Such as Ta, Ru etc..
The hearth electrode 212 and top electrode 218 and spin(-)orbit square magnetic random access memory 220 of photodiode 210
Hearth electrode 222 and top electrode 234, can be formed from conductive materials, conductive material injection or is led at metal, conductive metal oxide
Electric metal nitride etc. can be typically Cu.
To be electrically connected to spin(-)orbit square reluctance type random by the first electrical wiring 240 for the hearth electrode 212 of photodiode 210
The top electrode 218 of the hearth electrode 222 of memory 220, photodiode 210 is electrically connected to spin rail by the second electrical wiring 250
The top electrode 234 of road square magnetic random access memory 220, in this way, make photodiode 210 and spin(-)orbit square reluctance type with
Machine memory 220 forms concatenated circuit, due to the P-N junction that photodiode 210 is driving certainly, after detecting light,
Photo-signal is generated, and the size of photoelectric current reads the magnitude phase of electric current with spin(-)orbit square magnetic random access memory 220
Seemingly, photoelectric current flows into spin(-)orbit square magnetic random access memory 220, so as to be used for image recognition.
In embodiment, the first electrical wiring 240 and the second electrical wiring 250 can be in above-mentioned photodiodes 210 and spin
It is formed, can be realized by one or more layers interconnection structure in 220 manufacturing process of orbital moment magnetic random access memory, it can also
To pass through external electricity after above-mentioned photodiode 210 and the manufacture of spin(-)orbit square magnetic random access memory 220 are completed
Line is realized.
It is additionally provided with switching device 260 in the first electrical wiring 240 and/or the second electrical wiring 250, the switching device 260
For controlling the on state in the circuit of photodiode 210 and spin(-)orbit square magnetic random access memory 220, work as derailing switch
When part 260 is in an off state, spin(-)orbit square magnetic random access memory 220 and photodiode 210 independently of one another, spin
Orbital moment magnetic random access memory 220 can be used for carrying out the training of image recognition model, and close when switching device 260 is in
When conjunction state, photodiode 210 and the circuit of spin(-)orbit square magnetic random access memory 220 are in the conductive state, photoelectricity
Received optical signal is converted photoelectric current by diode 210, and then is input to spin(-)orbit square magnetic random access memory 220,
Spin(-)orbit square magnetic random access memory 220 can be used for the identification of the progress image by photo-signal.
Based on above-mentioned semiconductor device cell 20, the embodiment of the present application also proposed a kind of pattern recognition device, the figure
As identification device is made of the semiconductor device cell 20 in multiple above-described embodiments, can be by above-mentioned semiconductor device unit
20 chips formed, these semiconductor device cells 20 can be arranged in array, in multiple semiconductor device cells 20
Spin(-)orbit square magnetic random access memory 220 is used to store the weight of artificial neural network, the light of the photodiode 210
Electric signal is the input signal of image recognition model.
In specific application, image recognition model can be the artificial nerve network model based on deep learning, Ke Yigen
The quantity that semiconductor device cell 20 is determined according to the scale of model, when control switch device 260 is in an off state, the image
Spin(-)orbit square magnetic random access memory 220 in identification device is used to carry out the training of image recognition model, and stores instruction
Image recognition model after white silk;And when switching device 260 in the closure state, photodiode 210 is by received optical signal
It is converted into photoelectric current, and then is input to spin(-)orbit square magnetic random access memory 220, it is random based on spin(-)orbit square reluctance type
Image recognition model in memory, and then carry out image recognition.
In the memory cell and pattern recognition device of the present embodiment, photodiode is the PN junction device of driving certainly,
Without additional power supply, low energy consumption and can be applied in the particular surroundings of power-off, meanwhile, directly analog signal is identified,
Realize more intelligent image recognition.
The above is only a preferred embodiment of the present invention, although the present invention has been disclosed in the preferred embodiments as above, so
And it is not intended to limit the invention.Anyone skilled in the art is not departing from technical solution of the present invention ambit
Under, many possible changes and modifications all are made to technical solution of the present invention using the methods and technical content of the disclosure above,
Or equivalent example modified to equivalent change.Therefore, anything that does not depart from the technical scheme of the invention, according to the present invention
Technical spirit any simple modification, equivalent variation and modification made to the above embodiment, still fall within the technology of the present invention side
In the range of case protection.
Claims (10)
1. a kind of semiconductor device cell characterized by comprising
Substrate;
The photodiode and spin transfer torque magnetoresistive random access memory being located on the substrate, two pole of photoelectricity
Pipe includes hearth electrode, the first doping type semiconductor layer, the second doping type semiconductor layer and the top stacked gradually from the bottom to top
Electrode, the spin transfer torque magnetoresistive random access memory include the hearth electrode stacked gradually from the bottom to top, magnetic tunnel junction and
Top electrode;
Wherein, it is random by the first electrical wiring to be electrically connected to the spin transfer torque magnetoresistive for the hearth electrode of the photodiode
The top electrode of the hearth electrode of memory, the photodiode is electrically connected the spin transfer torque magnetoresistive by the second electrical wiring
Switching device is connected in the top electrode of random access memory, first electrical wiring and/or second electrical wiring.
2. semiconductor device cell according to claim 1, which is characterized in that the magnetic tunnel junction includes from the bottom to top
The first magnetosphere, tunnel layer, the second magnetosphere and the protective layer stacked gradually.
3. semiconductor device cell according to claim 2, which is characterized in that the magnetic tunnel junction further includes described
The second pinning layer between the first pinning layer and/or the second magnetosphere and protective layer under one magnetosphere.
4. semiconductor device cell according to claim 1, which is characterized in that the photodiode is visible light, red
Outer or ultraviolet photodiode.
5. a kind of semiconductor device cell characterized by comprising
Substrate;
The photodiode and spin(-)orbit square magnetic random access memory being located on the substrate, two pole of photoelectricity
Pipe includes hearth electrode, the first doping type semiconductor layer, the second doping type semiconductor layer and the top stacked gradually from the bottom to top
Electrode, the spin(-)orbit square magnetic random access memory include the hearth electrode stacked gradually from the bottom to top, Quantum geometrical phase
Layer, magnetoresistive tunnel junction and top electrode;
Wherein, it is random by the first electrical wiring to be electrically connected to the spin(-)orbit square reluctance type for the hearth electrode of the photodiode
The top electrode of the hearth electrode of memory, the photodiode is electrically connected the spin(-)orbit square reluctance type by the second electrical wiring
Switching device is connected in the top electrode of random access memory, first electrical wiring and/or second electrical wiring.
6. semiconductor device cell according to claim 5, which is characterized in that the magnetoresistive tunnel junction includes from the bottom to top
The first magnetosphere, tunnel layer, the second magnetosphere and the protective layer stacked gradually.
7. semiconductor device cell according to claim 5, which is characterized in that the magnetoresistive tunnel junction further includes described
Pinning layer between two magnetospheres and the protective layer.
8. semiconductor device cell according to claim 5, which is characterized in that the photodiode is visible light, red
Outer or ultraviolet photodiode.
9. a kind of pattern recognition device, which is characterized in that including multiple such as semiconductor device of any of claims 1-4
Part unit, the spin transfer torque magnetoresistive random access memory in multiple semiconductor device cells is for artificial neural network
Weight, the photosignal of the photodiode are the input signal of image recognition model.
10. a kind of pattern recognition device, which is characterized in that including multiple semiconductors as described in any one of claim 5-8
Device cell, the spin(-)orbit square magnetic random access memory in multiple semiconductor device cells is for storing artificial neuron
The weight of network, the photosignal of the photodiode are the input signal of image recognition model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910031809.3A CN109768065A (en) | 2019-01-14 | 2019-01-14 | A kind of semiconductor device cell and pattern recognition device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910031809.3A CN109768065A (en) | 2019-01-14 | 2019-01-14 | A kind of semiconductor device cell and pattern recognition device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109768065A true CN109768065A (en) | 2019-05-17 |
Family
ID=66452801
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910031809.3A Pending CN109768065A (en) | 2019-01-14 | 2019-01-14 | A kind of semiconductor device cell and pattern recognition device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109768065A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111725386A (en) * | 2019-09-23 | 2020-09-29 | 中国科学院上海微系统与信息技术研究所 | Magnetic memory device and manufacturing method thereof, memory and neural network system |
CN112054033A (en) * | 2020-09-15 | 2020-12-08 | 中国科学院微电子研究所 | Memory device |
CN112054035A (en) * | 2020-09-15 | 2020-12-08 | 中国科学院微电子研究所 | Memory device and manufacturing method thereof |
CN112054034A (en) * | 2020-09-15 | 2020-12-08 | 中国科学院微电子研究所 | Memory device |
CN112542190A (en) * | 2019-09-23 | 2021-03-23 | 中电海康集团有限公司 | Memory, and writing method and reading method of memory |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102292815A (en) * | 2009-02-02 | 2011-12-21 | 高通股份有限公司 | Magnetic tunnel junction (mtj) storage element and spin transfer torque magnetoresistive random access memory (stt-mram) cells having an mjt |
CN103248837A (en) * | 2013-05-17 | 2013-08-14 | 湘潭大学 | Image sensor based on memristor |
EP3190619A1 (en) * | 2016-01-07 | 2017-07-12 | Nokia Technologies Oy | A photodetector apparatus and associated methods |
-
2019
- 2019-01-14 CN CN201910031809.3A patent/CN109768065A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102292815A (en) * | 2009-02-02 | 2011-12-21 | 高通股份有限公司 | Magnetic tunnel junction (mtj) storage element and spin transfer torque magnetoresistive random access memory (stt-mram) cells having an mjt |
CN103248837A (en) * | 2013-05-17 | 2013-08-14 | 湘潭大学 | Image sensor based on memristor |
EP3190619A1 (en) * | 2016-01-07 | 2017-07-12 | Nokia Technologies Oy | A photodetector apparatus and associated methods |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111725386A (en) * | 2019-09-23 | 2020-09-29 | 中国科学院上海微系统与信息技术研究所 | Magnetic memory device and manufacturing method thereof, memory and neural network system |
CN112542190A (en) * | 2019-09-23 | 2021-03-23 | 中电海康集团有限公司 | Memory, and writing method and reading method of memory |
WO2021057179A1 (en) * | 2019-09-23 | 2021-04-01 | 浙江驰拓科技有限公司 | Memory, and writing method and reading method for memory |
CN112054033A (en) * | 2020-09-15 | 2020-12-08 | 中国科学院微电子研究所 | Memory device |
CN112054035A (en) * | 2020-09-15 | 2020-12-08 | 中国科学院微电子研究所 | Memory device and manufacturing method thereof |
CN112054034A (en) * | 2020-09-15 | 2020-12-08 | 中国科学院微电子研究所 | Memory device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109768065A (en) | A kind of semiconductor device cell and pattern recognition device | |
KR100446674B1 (en) | Semiconductor memory device | |
CN108665921A (en) | Semiconductor devices | |
US6473337B1 (en) | Memory device having memory cells with magnetic tunnel junction and tunnel junction in series | |
CN104813472B (en) | With enhancing stability erect spin transfer moment of torsion memory (STTM) device with and forming method thereof | |
KR100610710B1 (en) | Magnetic random access memory | |
CN108133936A (en) | Integrated circuit device and its manufacturing method | |
US8971977B2 (en) | Superconducting devices with ferromagnetic barrier junctions | |
CN101452990B (en) | Magnetic random access memory, manufacturing method and programming method thereof | |
EP1571713A4 (en) | Spin injection device, magnetic device using the same, magnetic thin film used in the same | |
JP2005116923A (en) | Nonvolatile magnetic memory cell using spin torque and magnetic random access memory using same | |
CN1244154C (en) | Semiconductor storaging device | |
JP2003151263A (en) | Reading method for magnetoresistive element having magnetically soft reference layer | |
CN103201796A (en) | Non-volatile magnetic tunnel junction transistor | |
CN109493900A (en) | Memory device, the method for providing it and dimensional stackable memory device | |
JP2003188358A (en) | Magneto-resistive device having magnetically soft reference layer | |
KR20200093720A (en) | Magnetic memory devices | |
TW200411835A (en) | Magnetic random access memory using schottky diode | |
JP2003188357A (en) | Magneto-resistive device including magnetically soft reference layer having embedded conductor | |
TW535159B (en) | Integrated memory with an arrangement of non-volatile memory cells and method for fabricating and for operating the integrated memory | |
KR20030034500A (en) | Magnetic random access memory | |
US6707085B2 (en) | Magnetic random access memory | |
JP7215645B2 (en) | neuromorphic device | |
US6873023B2 (en) | Magnetic random access memory | |
CN104916774B (en) | It is a kind of based on metal-doped organic magnetic tunnel-junction and manufacture method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190517 |
|
RJ01 | Rejection of invention patent application after publication |