Oxide thin film transistor for reservoir computing system and preparation method of oxide thin film transistor

Abstract

The invention belongs to the field of semiconductors, and provides an oxide thin film transistor for a reservoir computing system and a preparation method, and the transistor sequentially comprises a source electrode, a drain electrode, an oxide thin film, a gate dielectric layer and a gate electrode from top to bottom. The metal oxide semiconductor material disclosed by the invention has an excellent photoelectric effect, and meanwhile, due to the existence of oxygen vacancies, the metal oxide semiconductor material always shows a photoconductive relaxation effect, so that the metal oxide semiconductor material is expected to be applied to a device integrating sensing, storage and calculation. Besides, a metal oxide semiconductor is easy to prepare on a large scale, a metal oxide semiconductor thin film transistor is used as a reserve pool layer, a large-scale reserve pool calculation system integrating sensing, storage and calculation can be achieved, and the problems that current reserve pool calculation hardware mostly depends on a single device, and large-scale sensing, storage and calculation integration is difficult to achieve are solved.

Description

Technical Field

[0001] The invention belongs to the field of semiconductors, and in particular relates to an oxide thin film transistor for a reserve pool computing system and a preparation method thereof. Background Art

[0002] In the era of big data, there is an urgent need to process large amounts of data information at extremely high speed and accuracy. However, the traditional von Neumann architecture has a serious impact on computing speed and increased energy consumption due to the separation of storage and computing, which has become a core challenge facing today's technology. Inspired by the human brain, neuromorphic computing has become a new computing architecture with low energy consumption and high computing speed. In neuromorphic computing, sensors, memory, and processors can be integrated into one, eliminating the need to shuttle data between discrete memory and computing. Reservoir computing (RC) is a classic model of neuromorphic computing and has been widely developed in the past two decades.

[0003] Reservoir computing systems involve new research areas of nonlinear dynamics, complex networks, and machine learning. In reservoir computing systems, there is a time-dependent reservoir layer (usually a recurrent neural network) that provides complex nonlinearities in the input data, and a time-independent readout layer that provides weighted outputs of reservoir changes. Only the readout layer needs to be trained, which to some extent avoids the training difficulties associated with standard recurrent neural networks because the time-dependent reservoir is decoupled from the simple linear output layer. Since the response of the recurrent neural network is mathematically similar to that of a dynamic system, it can be replaced by a real-world dynamic system with appropriate characteristics, namely a nonlinear relationship between the input and output, and a dependence on the previous state that gradually disappears over time. With high prediction accuracy and a fast training process, reservoir computing systems have long been an ideal solution for many industry-level signal processing and learning tasks, including nonlinear distortion compensation in optical communications, real-time speech recognition, active noise control, etc.

[0004] In order to make the reserve pool system practical, a large-scale reserve pool computing hardware that integrates sensing, storage, and computing is needed. However, the existing reserve pool computing hardware mostly relies on a single device, making it difficult to achieve large-scale sensing, storage, and computing integration. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides an oxide thin film transistor for a storage pool computing system and a preparation method thereof to solve the problems in the prior art. The technical solution adopted by the present invention is: An oxide thin film transistor for a storage pool computing system, the transistor comprising, from top to bottom, a source-drain electrode, an oxide thin film, a gate dielectric layer and a gate electrode; The source and drain electrodes are made of low resistivity electrode materials; the source and drain electrode spacing is 50-500 microns, the electrode thickness is 50-100 nanometers, and the electrode size is 300-500 microns; The oxide film is a metal oxide thin film semiconductor material with a thickness of 10-50 nanometers; The gate electrode is made of a low resistivity electrode material.

[0006] A method for preparing an oxide thin film transistor for a reservoir computing system comprises the following steps: Step A, preparing a substrate with a gate electrode and a gate dielectric layer; Step B, preparing an oxide film on the surface of the gate electrode and the gate dielectric layer; Step C: preparing source and drain electrodes on the oxide film.

[0007] Furthermore, in step A, the substrate is a hard substrate or a flexible substrate; wherein the hard substrate is a high-conductivity silicon or quartz substrate with silicon dioxide; and the flexible substrate is polyimide or polydimethylsiloxane.

[0008] Furthermore, in step A, the gate electrode is made of a low resistivity material.

[0009] Furthermore, in step A, the gate dielectric layer is silicon dioxide or a high-k material.

[0010] Furthermore, in step B, the oxide film is indium gallium zinc oxide or indium gallium tin oxide, which is prepared by magnetron sputtering, using a target material having a ratio of In2O3:Ga2O3:ZnO or SnO2 of 1:1:1 mol%.

[0011] Furthermore, the step B comprises: pasting a mask on the substrate, using magnetron sputtering to sputter an oxide film with a thickness of about 10-50 nanometers on the substrate, the radio frequency power is 60-100W, the substrate temperature is 100°C-200°C, and during the sputtering process, the chamber atmosphere is a mixture of argon and oxygen, wherein the oxygen content is 5-20%, the rate is controlled at 0.1-0.5 angstroms / second, and the vacuum degree is controlled at 1×10 -4 -9×10 -4 Pa.

[0012] Furthermore, in step C, the source and drain electrodes are made of low resistivity materials and are prepared by one or more methods including thermal evaporation, electron beam evaporation, and magnetron sputtering.

[0013] Furthermore, the step C comprises: pasting a mask on the surface of the oxide film, preparing a 50-100 nanometer low resistivity electrode on the oxide surface by one or more methods such as thermal evaporation, electron beam evaporation, and magnetron sputtering, and covering the gate electrode. During the evaporation or sputtering process, the rate is 0.1-0.5 angstroms / second and the vacuum degree is 1×10 -4 -9×10 -4 Pa.

[0014] The present invention has the following beneficial effects: the metal oxide semiconductor material of the present invention has excellent photoelectric effect, and at the same time, due to the presence of oxygen vacancies, the metal oxide semiconductor material often exhibits a photoconductive relaxation effect, and is expected to be applied to devices integrating sensing, storage, and computing. In addition, metal oxide semiconductors are easy to prepare on a large scale, and using metal oxide semiconductor thin film transistors as a reservoir layer can realize a large-scale reservoir computing system integrating sensing, storage, and computing, solving the problem that the current reservoir computing hardware mostly relies on a single device and is difficult to achieve large-scale sensing, storage, and computing integration. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a schematic diagram of the structure of the oxide thin film transistor of the present invention; Figure 2 is the transfer characteristic curve of the transistor of the present invention in the dark and under 375 nanometer ultraviolet light; Figure 3 is a time-current curve of the transistor of the present invention under different numbers of ultraviolet light pulses; Figure 4 It is the current response distribution of the transistor of the present invention to different light pulse sequences. DETAILED DESCRIPTION

[0016] The following will be combined with the embodiments of the present invention Figure 1-Figure 4 , the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0017] like Figure 1 , an oxide thin film transistor for a reserve pool computing system, the device structure of the transistor from top to bottom are: source and drain electrodes 1, oxide film 2, gate dielectric layer 3, gate electrode 4.

[0018] The source-drain electrode 1 is made of a material with low resistivity, including but not limited to gold, silver, platinum, and aluminum. The source-drain electrode 1 has a spacing of 50-500 microns, an electrode shape of square, circular, rectangular, etc., an electrode thickness of 50-100 nanometers, and an electrode size of 300-500 microns.

[0019] The oxide film 2 is a metal oxide thin film semiconductor material, including but not limited to indium gallium tin oxide (InGaSnO) and indium gallium zinc oxide (InGaZnO), with a thickness of 10-50 nanometers. Metal oxide semiconductor materials have excellent photoelectric effects. At the same time, due to the presence of oxygen vacancies, metal oxide semiconductor materials often exhibit photoconductive relaxation effects, and are expected to be applied to devices that integrate sensing, storage, and computing. In addition, metal oxide semiconductors are easy to prepare on a large scale. Using metal oxide semiconductor thin film transistors as a reservoir layer can realize a large-scale reservoir computing system that integrates sensing, storage, and computing.

[0020] The gate dielectric layer 3 includes but is not limited to silicon dioxide, hafnium dioxide, and aluminum oxide, wherein the thickness of silicon dioxide is 100-300 nanometers, the thickness of hafnium dioxide is 5-30 nanometers, and the thickness of aluminum oxide is 15-50 nanometers.

[0021] The gate electrode 4 is made of high-conductivity silicon or low-resistivity electrode material, wherein the low-resistivity electrode material includes but is not limited to gold, silver, platinum, and aluminum, wherein the thickness of gold, silver, platinum, and aluminum is 25-100 nanometers, the electrode shape is square, circular, rectangular, etc., and the electrode size is 500 micrometers to 1 millimeter.

[0022] Based on the characteristics of metal oxide semiconductors, the present invention uses metal oxide semiconductor thin film transistors for the reservoir layer of the reservoir computing system, which can integrate sensing, storage and computing, and is easy to prepare on a large scale. The reservoir computing system based on metal oxide semiconductor thin film transistors described in the present invention has the advantages of large scale and integrated sensing, storage and computing, and can be widely used in high-tech fields such as information technology.

[0023] A method for preparing an oxide thin film transistor for a reservoir computing system comprises the following steps: Step A, prepare a substrate with a gate electrode 4 and a gate dielectric layer 3; wherein, in step A, the substrate is a hard substrate or a flexible substrate; for a hard substrate: the gate electrode 4 and the gate dielectric layer 3 use a high-conductivity silicon substrate or a quartz substrate with silicon dioxide on the surface: wherein the high-conductivity silicon can be p-type or n-type, the resistance value is less than 10 ohms, and the thickness of silicon dioxide is 100-300 nanometers. The flexible substrate is a high-temperature resistant polyimide (PI), polydimethylsiloxane (PDMS), etc.

[0024] Preferably, the gate dielectric layer 3 is made of silicon dioxide or a high-k material, such as hafnium dioxide or aluminum oxide.

[0025] The gate electrode 4 is made of low resistivity materials such as gold, silver, platinum, and aluminum, and can be prepared by thermal evaporation, electron beam evaporation, magnetron sputtering, etc. The rate during evaporation and sputtering is controlled at 0.1-0.5 angstroms / second, and the vacuum degree is controlled at 1×10 -4 -9×10 -4 Pa. Below, the electrode thickness is 25-100 nanometers, and the electrode size and shape can be customized through the mask to prepare large-scale oxide thin film transistors with independent gate control. The electrode size is 500 microns-1 mm, and the electrode shape is square, circular, rectangular, etc. In addition, when the gate dielectric layer 3 material is a high-k material such as hafnium dioxide and aluminum oxide, it can be prepared by single atomic layer deposition, using trimethylaluminum (TMA, C3H9Al) as an aluminum source precursor, hafnium tetrachloride (HfCl4) as a hafnium source precursor, and water (H2O) as an oxygen source precursor. The reaction equations for depositing hafnium dioxide and aluminum oxide are HfCl4+H2O→HfO2+4HCl and C3H9Al+H2O→Al2O3+CH4, respectively. The temperature during deposition is controlled at 25-120°C, the thickness of hafnium dioxide is 5-30 nanometers, the thickness of aluminum oxide is 15-50 nanometers, and the deposition thickness is determined by the number of deposition cycles. The thickness of a single cycle deposition is about 0.07-0.1 nanometers.

[0026] Step B, preparing an oxide film 2 on the surface of the gate electrode 4 and the gate dielectric layer 3; wherein the oxide film 2 is prepared by magnetron sputtering, using a target material with a ratio of In2O3:Ga2O3:ZnO or SnO2 of 1:1:1 mol%, a radio frequency power of 60-100W, a substrate temperature of 100°C-200°C, and during the sputtering process, the chamber atmosphere is a mixture of argon and oxygen, wherein the oxygen content is 5-20%, the rate is controlled at 0.1-0.5 angstroms / second, and the vacuum degree is 1×10 -4 -9×10 -4 Pa, sputtering thickness is 10-50 nm.

[0027] Step B is specifically as follows: an oxide film is prepared on the surface of the gate electrode 4 and the gate dielectric layer 3: a customized mask is pasted on the substrate, the customized mask is a purchased hard mask, and a square array of 500 microns × 500 microns and 10 × 10 in number is cut on the hard stainless steel by a laser processing method. It can also be processed into other shapes such as circle, rectangle, etc. according to the design.

[0028] Magnetron sputtering was used to sputter indium gallium tin oxide with a thickness of about 30 nanometers on the substrate. The RF power was 80W, the substrate temperature was 180°C, and during the sputtering process, the chamber atmosphere was a mixture of argon and oxygen, with an oxygen content of 5%. During the sputtering process, the rate was controlled at 0.1 angstroms / second and the vacuum degree was controlled at 5×10 -4Pa. The material size is 500 μm × 500 μm, and the number of materials is 10 × 10 arrays.

[0029] Step C, prepare the source-drain electrode 1 on the oxide film 2; wherein the source-drain electrode 1 is uniformly evaporated onto the sample surface through a customized mask plate by means of thermal evaporation, electron beam evaporation, magnetron sputtering, etc., the electrodes include but are not limited to gold, silver, platinum, aluminum, the electrode spacing is 50-500 microns, the electrode shape is square, round, rectangular, etc., the electrode thickness is 50-100 nanometers, and the electrode size is 300-500 microns. During the evaporation or sputtering process, the rate is controlled at 0.1-0.5 angstroms / second, and the vacuum degree is controlled at 1×10 -4 -9×10 -4 Pa, evaporation or sputtering thickness is 50-100 nanometers.

[0030] Step C is as follows: a customized mask is attached to the surface of InGaSnO, and a 50-nanometer gold electrode is deposited on the surface of InGaSnO by vacuum thermal evaporation to cover the gate electrode. During the evaporation process, the rate is controlled at 0.1 angstroms / second and the vacuum degree is controlled at 5×10 -4 Pa. The electrode spacing is 100 μm, the electrode size is 500 μm × 500 μm, and the resulting device is a 10 × 10 device array.

[0031] like Figure 2 , based on the transfer characteristic curve of the InGaSnO thin film transistor, it can be seen from the figure that the present invention has obvious light response and charge storage capabilities.

[0032] like Figure 3 Based on the time-current curve of the InGaTnO thin-film transistor under different numbers of ultraviolet light pulses, it can be seen from the figure that the present invention has obvious light response and photoconductivity relaxation effect, and the relaxation time is related to the number of light pulses.

[0033] like Figure 4 Based on the current response distribution of the InGaTnO thin-film transistor to different light pulse sequences, it can be seen from the figure that the present invention can be used as a reserve layer and has the potential to be applied to a reserve pool system that integrates sensing, storage and computing.

[0034] The metal oxide semiconductor material of the present invention has an excellent photoelectric effect. At the same time, due to the presence of oxygen vacancies, the metal oxide semiconductor material often exhibits a photoconductive relaxation effect, and is expected to be applied to devices integrating sensing, storage, and computing. In addition, metal oxide semiconductors are easy to prepare on a large scale. By using metal oxide semiconductor thin film transistors as a reservoir layer, a large-scale reservoir computing system integrating sensing, storage, and computing can be realized, solving the problem that the current reservoir computing hardware mostly relies on a single device and is difficult to achieve large-scale sensing, storage, and computing integration.

[0035] The embodiments described above are only descriptions of the preferred modes of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations, modifications, and substitutions made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. An oxide thin film transistor for a reservoir computing system, characterized in that: The transistor comprises, from top to bottom, a source-drain electrode (1), an oxide film (2), a gate dielectric layer (3) and a gate electrode (4); The source-drain electrode (1) is made of a low resistivity electrode material; the source-drain electrode (1) has a spacing of 50-500 micrometers, an electrode thickness of 50-100 nanometers, and an electrode size of 300-500 micrometers; The oxide film (2) is a metal oxide thin film semiconductor material with a thickness of 10-50 nanometers; The gate electrode (4) is made of a low resistivity electrode material.

2. A method for preparing an oxide thin film transistor for a storage pool computing system, applied to the oxide thin film transistor for a storage pool computing system according to claim 1, characterized in that: The following steps are involved: Step A, preparing a substrate with a gate electrode (4) and a gate dielectric layer (3); Step B, preparing an oxide film (2) on the surface of the gate electrode (4) and the gate dielectric layer (3); Step C: preparing the source-drain electrode (1) on the oxide film (2).

3. A method for preparing an oxide thin film transistor for a storage pool computing system as claimed in claim 2, characterized in that: In step A, the substrate is a hard substrate or a flexible substrate; the hard substrate is a high-conductivity silicon or quartz substrate with silicon dioxide; the flexible substrate is polyimide or polydimethylsiloxane.

4. The method for preparing an oxide thin film transistor for a storage pool computing system according to claim 2, characterized in that: In the step A, the gate electrode (4) is a low resistivity material.

5. The method for preparing an oxide thin film transistor for a storage pool computing system according to claim 2, characterized in that: In step A, the gate dielectric layer (3) is silicon dioxide or a high-k material.

6. The method for preparing an oxide thin film transistor for a storage pool computing system according to claim 2, characterized in that: In step B, the oxide film (2) is indium gallium zinc oxide or indium gallium tin oxide, and is prepared by magnetron sputtering, with the ratio of In2O3:Ga2O3:ZnO or SnO2 in the target being 1:1:1 mol%.

7. A method for preparing an oxide thin film transistor for a storage pool computing system as claimed in claim 6, characterized in that: The step B comprises: pasting a mask on the substrate, and sputtering an oxide film (2) with a thickness of about 10-50 nanometers on the substrate by magnetron sputtering, the radio frequency power is 60-100 W, the substrate temperature is 100° C.-200° C., during the sputtering process, the chamber atmosphere is a mixture of argon and oxygen, wherein the oxygen content is 5-20%, the rate is controlled at 0.1-0.5 angstroms / second, and the vacuum degree is controlled at 1×10 -4 -9×10 -4 Pa.

8. The method for preparing an oxide thin film transistor for a storage pool computing system according to claim 2, characterized in that: In the step C, the source-drain electrode (1) is a low resistivity material and is prepared by one or more methods including thermal evaporation, electron beam evaporation, and magnetron sputtering.

9. The method for preparing an oxide thin film transistor for a storage pool computing system according to claim 8, characterized in that: The step C comprises: pasting a mask on the surface of the oxide film, preparing a 50-100 nanometer low resistivity electrode on the oxide surface by one or more methods such as thermal evaporation, electron beam evaporation, and magnetron sputtering, and covering the gate electrode (4); during the evaporation or sputtering process, the rate is 0.1-0.5 angstroms / second, and the vacuum degree is 1×10 -4 -9×10 -4 Pa.

Images