Retinomorphic sensor array and image convolution method therefor

Patent metadata

Field	Value
Publication number	US-2024234617-A9
Application number	US-202017769609-A
Country	US
Kind code	A9
Filing date	Mar 26, 2020
Priority date	Mar 17, 2020
Publication date	Jul 11, 2024
Grant date	—

Abstract

A retinomorphic sensor array and a convolution method are used for image processing therefor, wherein the optoelectronic sensor has a vertically stacked heterostructure provided with a bottom electrode, a dielectric layer, a channel layer, a source electrode and a drain electrode on a base, the source and drain electrode are mutually opposite and are arranged at two ends of the channel layer, the bottom electrode, the source and drain electrode are made of a material used by a flexible electrode, an inert metal or a semimetal, the dielectric layer is made of an insulating material, the channel layer is made of a bipolar material, and the base comprises a substrate and an insulating material layer generated on a surface of the substrate.

First claim

What is claimed is:

1 . A retinomorphic sensor, having a vertically stacked heterostructure provided with a bottom electrode, a dielectric layer, a channel layer, a source electrode and a drain electrode on a base, wherein the source and drain electrode are mutually opposite and are arranged at two ends of the channel layer, the bottom electrode, the source and drain electrode are made of a material used by a flexible electrode, an inert metal or a semimetal, the dielectric layer is made of an insulating material and comprises a plurality of layers, the channel layer is made of a bipolar material, and the base comprises a substrate and an insulating material layer generated on a surface of the substrate.

2 . The retinomorphic sensor according to claim 1 , wherein the substrate is of a material comprising silicon, polyimide or polydimethylsiloxane, and the insulating material layer generated on the surface of the substrate is silicon oxide, aluminum oxide, hafnium zirconium oxygen or boron nitride.

3 . The retinomorphic sensor according to claim 1 , wherein the bipolar material of the channel layer is graphene, tungsten selenide, molybdenum telluride, black phosphorus or palladium selenide.

4 . The retinomorphic sensor according to claim 1 , wherein the dielectric layer is made of any one or more of boron nitride, silicon oxide, aluminum oxide and hafnium zirconium oxygen.

5 . A fabrication method for the retinomorphic sensor according to claim 1 , comprising:
S1, preparing the bottom electrode on a surface of the base; S2, directly obtaining the dielectric layer on the bottom electrode, or firstly obtaining dielectric layer materials on the bottom electrode, and then vertically stacking the dielectric layer materials by using a material transfer method to prepare the dielectric layer having a multilayer structure; S3, performing direct generation of the bipolar material of the channel layer on the dielectric layer; or firstly obtaining the bipolar material of the channel layer, and then transferring the bipolar material onto the dielectric layer by using a material transfer method to form the channel layer; and S4, preparing the source and drain electrode on a surface of the channel layer.

6 . The fabrication method for the retinomorphic sensor according to claim 5 , wherein, in the S2, the method for directly obtaining the dielectric layer on the bottom electrode is a chemical vapor deposition method, a chemical vapor transport method, a molecular-beam epitaxy method, an atomic layer deposition method or a hydrothermal method.

7 . The fabrication method for the retinomorphic sensor according to claim 5 , wherein, in the S3, the method for performing the direct generation of the bipolar material of the channel layer on the dielectric layer is a chemical vapor deposition method, a chemical vapor transport method, a molecular-beam epitaxy method, an atomic layer deposition method or a hydrothermal method.

8 . The fabrication method for the retinomorphic sensor according to claim 5 , wherein, in the S1, the method for preparing the bottom electrode on a surface of the base comprises:
S11, preparing a designed bottom electrode shape on the substrate by adopting an ultraviolet photo lithography method, an electron beam lithography method or a mask method; and S12, preparing the bottom electrode.

9 . A process for using the retinomorphic sensor of claim 1 as an optoelectronic device.

10 . The process according to claim 9 , wherein the optoelectronic device is adopted in a visual convolution method for image processing comprising:
i) arranging bit lines in each optoelectronic device of an optoelectronic sensor array, connecting the bit lines corresponding to optoelectronic devices in each row in series, arranging signal lines on each optoelectronic device, connecting the signal lines corresponding to the optoelectronic devices in each column in series, and applying, by the bit lines and the signal lines, source and drain voltages to the optoelectronic sensor at a specific position in the array; ii) arranging word lines in each optoelectronic device of the optoelectronic sensor array for applying a back-gate voltage to a specific row of the optoelectronic sensor in the array; iii) inputting corresponding voltages for the crossed optoelectronic sensors in a specific column by using the bit lines and the word lines, meanwhile, inputting the back-gate voltage to corresponding optoelectronic devices through the word lines, completing partial convolution operation, and outputting a result, i.e.,
I m =P 11 ×G 11 ( V g 11 )+ P 21 ×G 21 ( V g 21 )+ . . . + P m1 ×G m1 ( V g m1 )
wherein V g m1 is a back-gate voltage of an m th row and a 1 st column, P m1 is a visual image information input of an optoelectronic device on the m th row and the 1 st column, G m1 (V g m1 ) is photo responsivity of the m th row and the 1 st column, and the m is a total number of rows of convolution kernels; and iv) completing convolution operation of the whole optoelectronic sensor array by using an m×m convolution kernel according to the method in the step (3).

11 . The process according to claim 9 , wherein the optoelectronic device is adopted in a visual image recognition method comprising:
step 1, inputting information to be recognized into the optoelectronic sensor array, and setting back-gate voltages of all optoelectronic devices to be 0 V; step 2, acquiring an output current I of the optoelectronic sensor array, and inputting the output current into the following Sigmoid activation function:
f =(1+ e −αI ) −1
wherein I is the output current of the optoelectronic sensor array, and α is a normalization coefficient; step 3, after calculating a value of the activation function ƒ, comparing the value with a target value, then determining and executing error back propagation operation according to the following equation:
δ k =f g k −f k

wherein δ k is an error used in a k th training, f g k is a theoretical output value in the k th training, and f k is an output value of the Sigmoid activation function in the k th training; step 4, after the error is transmitted to the first layer, updating an initial back-gate value in the optoelectronic sensor array through the following functional relationship:
Δ V g k =β×round( n −1 ×conv( P,δ k ))

wherein n is a step length, β is a step length of a gate voltage change, P is an input of visual image information, round is a rounding function, and conv is a convolution function, and thus a training process is completed; and step 5, looping the steps 1-4 until the error calculated in the step 3 is close to or equal to 0, that is, a target image is successfully recognized from all the inputted images.

Assignees

• Nanjing University

Inventors

• Miao Feng
• Liang Shijun
• Wang Chenyu

Classifications

• H10F77/12

  Active materials · CPC title

• H10F77/206

  for devices having potential barriers · CPC title

• H10F71/00

  Manufacture or treatment of devices covered by this subclass (patterning processes to connect thin photovoltaic cells in integrated devices, or assemblies of multiple devices, having photovoltaic cells H10F19/33; manufacture or treatment of encapsulations or containers for integrated devices, or assemblies of multiple devices, having photovoltaic cells H10F19/80; manufacture or treatment of integrated devices, or assemblies of multiple devices, comprising at least one element in which radiation controls the flow of current H10F39/00) · CPC title

• H10F39/80377

  characterised by the channel of the transistor, e.g. channel having a doping gradient · CPC title

• H10F39/8033

  Photosensitive area · CPC title