A flexible thin film transistor mechanical fatigue life prediction method based on low-frequency noise

Abstract

The application relates to the technical field of semiconductors and discloses a flexible thin film transistor mechanical fatigue life prediction method based on low-frequency noise, which comprises the following steps: obtaining a flexible thin film transistor to be predicted; calculating the transfer characteristic curve and the low-frequency noise power spectrum of the flexible thin film transistor to be predicted; obtaining first data of the flexible thin film transistor to be predicted based on the transfer characteristic curve, calculating second data of the flexible thin film transistor to be predicted based on the first data and the low-frequency noise power spectrum, inputting the mechanical stress intensity, the cycle number and the second data into a trained life prediction model, and outputting a curve graph of the degradation amount of the flat-band voltage power spectrum density and the degradation amount of the contact noise power spectrum density changing with the cycle number; and obtaining the fatigue life of the flexible thin film transistor to be predicted under a specific mechanical stress intensity condition according to the curve graph. The application can improve the prediction accuracy and reliability, reduce the time and cost required for complete fatigue life testing of the device.

Description

Technical Field

[0001] This invention relates to the field of semiconductor technology, and in particular to a method for predicting the mechanical fatigue life of flexible thin-film transistors based on low-frequency noise. Background Technology

[0002] New display technologies are a key development direction for strategic emerging industries in the era of national informatization and intelligentization. High-resolution flexible displays, as an important development goal in the field of new displays, have long attracted the attention of many leading domestic enterprises. Among them, flexible thin-film transistors (FTFTs) are a special form of thin-film transistors (TFTs) and the basic unit for realizing the switching control of flexible integrated circuits, thus becoming a key driving component for flexible displays. In practical applications, flexible TFTs usually need to operate under complex mechanical stress. As the operating time increases, the aforementioned mechanical stress will gradually cause microscopic defects and damage in the gate oxide layer, active layer, source-drain contact area, and various interfaces of the flexible TFT, resulting in the degradation of electrical parameters such as threshold voltage, transconductance, and mobility, and ultimately leading to the failure of the flexible display system. Therefore, in order to ensure the long-term stable operation of flexible displays, it is necessary to determine the service life of flexible TFTs under different mechanical stresses and evaluate their reliability.

[0003] In existing technologies, the following methods are commonly used to determine the lifespan of flexible TFTs: 1. Applying cyclic mechanical stress to the flexible device using a mechanical fatigue testing machine and observing the degradation of the device's electrical performance before and after the fatigue test to assess the device's mechanical fatigue life; however, a complete mechanical fatigue test requires a long time cycle, consuming a large amount of testing time and costs, resulting in low quality inspection efficiency and high quality inspection costs for flexible TFTs; 2. Due to the long time and high cost required for physical characterization techniques of the device's internal microstructure (focused ion beam analysis, transmission electron microscopy, etc.), the mechanical fatigue life of the device is usually evaluated based on the degradation of the macroscopic electrical parameters of the flexible TFT. However, macroscopic electrical parameters cannot intuitively reflect the changes in microscopic defects and potential damage inside the device, thus limiting the accuracy and reliability of the lifespan calculation results; 3. Predicting and estimating the fatigue life of flexible devices by establishing a linear relationship between mechanical stress and fatigue life (stress-life relationship curve); however, in actual production and life, there may be a nonlinear relationship between mechanical stress and the fatigue life of flexible TFTs, which will lead to inaccurate prediction results of the stress-life relationship curve. Summary of the Invention

[0004] The purpose of this invention is to overcome the problems existing in the prior art and provide a method for predicting the mechanical fatigue life of flexible thin-film transistors based on low-frequency noise. This invention can improve the accuracy and reliability of prediction and reduce the time and cost required to conduct a complete fatigue life test on the device.

[0005] To achieve the above objectives, the present invention provides a method for predicting the mechanical fatigue life of flexible thin-film transistors based on low-frequency noise, comprising:

[0006] Obtain the flexible thin-film transistor to be predicted;

[0007] Calculate the transfer characteristic curve and low-frequency noise power spectrum of the flexible thin-film transistor to be predicted;

[0008] Based on the transfer characteristic curve, first data of the flexible thin-film transistor to be predicted is obtained, and the first data includes at least one of threshold voltage, transconductance and field-effect mobility.

[0009] Based on the first data and the low-frequency noise power spectrum, the second data of the flexible thin-film transistor to be predicted is calculated, wherein the second data includes at least one of the initial value of flat-band voltage power spectral density and the initial value of contact noise power spectral density;

[0010] The mechanical stress intensity, cycle number, and the second data are input into the trained lifetime prediction model, and the output is a curve showing the change of flat band voltage power spectral density degradation or contact noise power spectral density degradation with the cycle number.

[0011] Record the number of cycles in which the degradation of the flat band voltage power spectral density or the degradation of the contact noise power spectral density exceeds the degradation threshold, and use the number of cycles as the fatigue life of the flexible thin-film transistor to be predicted under the mechanical stress intensity conditions.

[0012] Furthermore, the flexible thin-film transistor to be predicted is one or more combinations of amorphous silicon thin-film transistors, low-temperature polycrystalline silicon thin-film transistors, organic thin-film transistors, and oxide thin-film transistors.

[0013] Furthermore, flicker noise is used as a noise source to calculate the low-frequency noise power spectrum.

[0014] Further, the calculation of the second data for the flexible thin-film transistor to be predicted based on the first data and the low-frequency noise power spectrum includes:

[0015] (1) Measure the flicker noise data of the flexible thin-film transistor to be predicted using a noise testing system;

[0016] (2). Based on the flicker noise data, a flicker noise model of the flexible thin-film transistor to be predicted is separated and constructed. The flicker noise model includes a first flicker noise model of drain current and a second noise model of contact resistance.

[0017] (3) Calculate the second data of the flexible thin-film transistor to be predicted based on the flicker noise model and the first data.

[0018] Furthermore, the formula for calculating the initial value of the flat-band voltage power spectral density is as follows:

[0019]

[0020] in, The flicker noise power spectral density is based on the drain current of the flexible thin-film transistor to be predicted. For transconductance; The Coulomb scattering coefficient is used. Capacitance per unit area of ​​gate oxide layer; The gate bias voltage applied when calculating the flicker noise power spectral density of the flexible thin-film transistor to be predicted; Threshold voltage; The initial value of the flat-band voltage power spectral density. μ eff For field-effect mobility;

[0021] The formula for calculating the initial value of the contact noise power spectral density is as follows:

[0022]

[0023] in, Drain current density, The initial value of the contact noise power spectral density. The applied source-drain voltage.

[0024] Furthermore, the mechanical stress intensity includes tensile stress intensity and bending stress intensity, wherein the tensile stress intensity is represented by the magnitude of the tensile force and the bending stress intensity is represented by the magnitude of the bending radius.

[0025] Furthermore, the lifetime prediction model is based on a backpropagation neural network, and specifically includes an input layer, a first hidden layer, a second hidden layer, and an output layer connected in sequence.

[0026] The number of nodes in the first hidden layer and the second hidden layer is determined by the following formula:

[0027]

[0028] In the above formula, The number of nodes in the input layer. This represents the number of nodes in the output layer. This represents the number of nodes in the hidden layer.

[0029] Furthermore, training the lifespan prediction model specifically includes:

[0030] (1) Obtain flexible thin-film transistor samples;

[0031] (2). Calculate the first transfer characteristic curve and the first low-frequency noise power spectrum of the flexible thin-film transistor sample;

[0032] (3). Calculate the initial values ​​of the first flat band voltage power spectral density and the first contact noise power spectral density based on the first transfer characteristic curve and the first low-frequency noise power spectrum;

[0033] (4) The flexible thin-film transistor samples were grouped using a uniform orthogonal experimental method, and mechanical stress intensity group and cycle number group were established based on the grouping results;

[0034] (5) Apply corresponding mechanical stress and cycle number to each group of flexible thin film transistor samples after grouping to conduct fatigue life test;

[0035] (6) Measure and record the second low-frequency noise power spectrum of each group of flexible thin film transistor samples after the test. Based on the second low-frequency noise power spectrum, the initial value of the first flat band voltage power spectral density and the initial value of the first contact noise power spectral density, calculate the degradation amount of the first flat band voltage power spectral density and the degradation amount of the first contact noise power spectral density of each group of flexible thin film transistor samples.

[0036] (7). Construct a training set, which includes mechanical stress intensity, number of cycles, initial value of the first flat band voltage power spectral density, initial value of the first contact noise power spectral density, degradation amount of the first flat band voltage power spectral density, and degradation amount of the first contact noise power spectral density;

[0037] (8) Use the training set to iteratively train the life prediction model to obtain a trained life prediction model.

[0038] Furthermore, before iteratively training the lifetime prediction model using the training set, the training set is normalized and / or differentiated.

[0039] Furthermore, the degradation threshold is That is, the degradation of the flat-band voltage power spectral density exceeds the initial value of the flat-band voltage power spectral density. Or the degradation of the contact noise power spectral density exceeds the initial value of the contact noise power spectral density. The number of cycles is used as the fatigue life of the flexible thin-film transistor under the mechanical stress intensity condition.

[0040] Compared with the prior art, the advantages of this invention are as follows:

[0041] 1. This invention uses a trained life prediction model to predict the fatigue life of flexible thin-film transistors under given mechanical stress conditions, effectively shortening the time and cost required for a complete fatigue life test of flexible thin-film transistors.

[0042] 2. This invention uses low-frequency noise characteristics, which are more sensitive to microscopic defects and damage within the device, as information parameters to predict the fatigue life of the device under mechanical stress, thereby improving the accuracy and reliability of the prediction.

[0043] 3. By using a neural network model with strong nonlinear mapping capabilities and high accuracy to replace the traditional stress-life relationship model, the accuracy and reliability of the model prediction results are effectively improved, and the computational workload is reduced. Attached Figure Description

[0044] Figure 1 This is a flowchart of a method for predicting the mechanical fatigue life of a flexible thin-film transistor based on low-frequency noise, according to Embodiment 1 of the present invention.

[0045] Figure 2 This is a structural diagram of the metal oxide thin-film transistor of Embodiment 1 of the present invention;

[0046] Figure 3 This is a schematic diagram of the noise power spectral density testing system according to Embodiment 1 of the present invention;

[0047] Figure 4 This is a schematic diagram of the life prediction model of Embodiment 1 of the present invention;

[0048] Figure 5 This is a flowchart of the training and prediction process of the lifetime prediction model in Embodiment 1 of the present invention. Detailed Implementation

[0049] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0050] For flexible TFTs, the industry often applies cyclic mechanical stress and observes the degradation of the device's electrical performance before and after fatigue testing to assess the device's mechanical fatigue life. However, this test requires a long time cycle, consuming a significant amount of experimental time and costs. Secondly, the industry evaluates the mechanical fatigue life of flexible TFTs based on the degradation of macroscopic electrical parameters. However, macroscopic electrical parameters cannot directly reflect changes in microscopic defects and potential damage within the device, thus limiting the accuracy and reliability of the lifespan calculation results. Finally, the industry predicts the fatigue life of flexible TFTs by establishing stress-life relationship curve models, but the nonlinear relationship between mechanical stress and fatigue life leads to inaccurate prediction results from this model. Therefore, this invention, based on the low-frequency noise characteristics that are highly sensitive to microscopic damage in devices, utilizes a neural network model with strong nonlinear mapping capabilities and high accuracy to predict the failure cycle number of the device under different mechanical stresses. This not only reduces experimental time and costs but also effectively improves the accuracy and reliability of the fatigue life prediction model.

[0051] Example 1

[0052] like Figure 1 As shown in the preferred embodiment of the present invention, a method for predicting the mechanical fatigue life of a flexible thin-film transistor based on low-frequency noise includes:

[0053] S1: Obtain the flexible thin-film transistor to be predicted;

[0054] In one feasible embodiment, the flexible thin-film transistor to be predicted includes one or more combinations of amorphous silicon thin-film transistors, low-temperature polycrystalline silicon thin-film transistors, organic thin-film transistors, and oxide thin-film transistors. In this embodiment, a metal oxide thin-film transistor is selected as the flexible thin-film transistor to be predicted. The structure of the metal oxide thin-film transistor is shown below. Figure 2 It mainly includes: gate, gate oxide layer, source / drain, etc.

[0055] S2: Calculate the transfer characteristic curve and low-frequency noise power spectrum of the flexible thin-film transistor to be predicted;

[0056] In one feasible embodiment, obtaining the transfer characteristic curve of the flexible thin-film transistor to be predicted requires setting test conditions, as follows:

[0057] The applied gate bias voltage The scope is: The applied source-drain voltage for: ;

[0058] In this embodiment, flicker noise ( Using noise as a noise source, the low-frequency noise power spectrum is calculated to obtain the power spectrum of the flexible thin-film transistor to be predicted. The test conditions for the noise power spectrum are as follows:

[0059] The applied gate bias voltage The scope is: The applied source-drain voltage for: ;

[0060] It should be noted that TFT The noise amplitude is usually quite weak, therefore a high-precision testing device is needed to test it, and the collected data needs to be amplified. Therefore, in this embodiment... The structure of the noise power spectrum testing device is as follows: Figure 3 As shown, the system specifically includes: a low-noise amplifier, a spectrum analyzer, an analog-to-digital converter (ADC), and a semiconductor parameter tester. The source test unit of the semiconductor parameter tester provides precise test current or voltage to the device under test (DUT). Subsequently, the current noise is amplified by the low-noise amplifier and then fed into the ADC for acquisition. Finally, the acquired data is transmitted to the spectrum analyzer, where a Fast Fourier Transform (FFT) is performed to calculate the... Noise power spectrum. This test structure offers a noise testing speed of up to 20 s / bias and a minimum noise current accuracy as low as 10⁻²⁷ A² / Hz, ensuring... The frequency range and accuracy of noise testing;

[0061] It should be noted that the above measurements are for the flexible thin-film transistor to be predicted. When the noise power spectrum is measured, the applied gate bias voltage With threshold voltage The difference ( - Much greater than the source-drain voltage To ensure that the device operates in the linear region; due to Noise is more pronounced in the low-frequency range, so the sampling frequency is set to 10Hz.

[0062] S3: Obtain the first data of the flexible thin-film transistor to be predicted based on the transfer characteristic curve. The first data includes at least one of the threshold voltage, transconductance and field-effect mobility.

[0063] In one feasible embodiment, the threshold voltage of the flexible thin-film transistor to be predicted is obtained based on the transfer characteristic curve obtained in step S2. transconductance and field effect mobility μ eff .

[0064] S4: Calculate the second data of the flexible thin-film transistor to be predicted based on the first data and the low-frequency noise power spectrum. The second data includes at least one of the initial values ​​of the flat-band voltage power spectral density and the initial values ​​of the contact noise power spectral density.

[0065] In one feasible embodiment, calculating second data for the flexible thin-film transistor to be predicted based on the first data and the low-frequency noise power spectrum includes:

[0066] (1). The flicker noise data of the flexible thin-film transistor to be predicted is measured using a noise testing system, with a test frequency range of 1Hz to 10,000Hz;

[0067] (2). Based on the flicker noise data, a flicker noise model of the flexible thin-film transistor to be predicted is separated and constructed. The flicker noise model includes a first flicker noise model of drain current and a second noise model of contact resistance.

[0068] (3) Calculate the second data of the flexible thin-film transistor to be predicted based on the flicker noise model and the first data.

[0069] Furthermore, the formula for calculating the initial value of the flat-band voltage power spectral density is as follows:

[0070]

[0071] in, The flicker noise power spectral density is based on the drain current of the flexible thin-film transistor to be predicted. For transconductance; The Coulomb scattering coefficient is used. Capacitance per unit area of ​​gate oxide layer; The gate bias voltage applied when calculating the flicker noise power spectral density of the flexible thin-film transistor to be predicted; Threshold voltage; This represents the initial value of the flat-band voltage power spectral density;

[0072] The formula for calculating the initial value of the contact noise power spectral density is as follows:

[0073]

[0074] in, Drain current density, This is the initial value of the contact noise power spectral density.

[0075] S5: Input the mechanical stress intensity, number of cycles, and second data into the trained lifetime prediction model, and output a curve showing the change of flat band voltage power spectral density degradation or contact noise power spectral density degradation with the number of cycles.

[0076] In one feasible embodiment, the mechanical stress intensity includes tensile stress intensity and bending stress intensity, wherein the tensile stress intensity is represented by the magnitude of the tensile force and the bending stress intensity is represented by the magnitude of the bending radius.

[0077] Furthermore, the lifetime prediction model is based on a backpropagation neural network. Specifically, the lifetime prediction model includes an input layer, a first hidden layer, a second hidden layer, and an output layer connected in sequence. The double hidden layer design enhances the fitting accuracy and robustness of the neural network prediction model for nonlinear data. (See [link to lifetime prediction model]). Figure 4 ,

[0078] The number of nodes in the first hidden layer and the second hidden layer is determined by the following formula:

[0079]

[0080] In the above formula, The number of nodes in the input layer. This represents the number of nodes in the output layer. This represents the number of nodes in the hidden layer. Based on the empirical formula, in this embodiment, the number of nodes in the input layer is 4, and the number of nodes in the output layer is 2. Therefore, the initial number of nodes in each hidden layer is... The value is set to 3, and this is adjusted during subsequent neural network training. The ReLU function is chosen as the hidden layer activation function, and its expression is:

[0081]

[0082] Using the identity function as the activation function of the output layer, its expression is:

[0083]

[0084] Furthermore, the process of training the lifetime prediction model is described in [link to documentation]. Figure 5 Specifically, it includes:

[0085] (1) Obtain flexible thin-film transistor samples;

[0086] In this embodiment, all flexible thin-film transistor samples are identical in material, process parameters, and size; n random subsamples are drawn from the sampling population according to the principle of simple random sampling, where n ≥ 20;

[0087] (2). Calculate the first transfer characteristic curve and the first low-frequency noise power spectrum of the flexible thin-film transistor sample;

[0088] It is understandable that the implementation process of this step is the same as that of S2, and will not be repeated here.

[0089] (3). Calculate the initial values ​​of the first flat band voltage power spectral density and the first contact noise power spectral density based on the first transfer characteristic curve and the first low-frequency noise power spectrum;

[0090] Understandably, the implementation process for this step is the same as that for S4, and will not be repeated here.

[0091] (4) The flexible thin-film transistor samples were grouped using a uniform orthogonal experimental method, and mechanical stress intensity group and cycle number group were established based on the grouping results;

[0092] Specifically, mechanical stress intensity groups are established based on the grouping results. F 1, F 2, ..., F i) and cycle group ( N 1, N 2, ..., N j).

[0093] (5) Apply corresponding mechanical stress and cycle number to each group of flexible thin film transistor samples after grouping to conduct fatigue life test;

[0094] Specifically, the fatigue life test is designed as follows:

[0095] Set the types and intensities of mechanical stress, with four levels for each stress intensity;

[0096] The 20 samples were randomly divided into 4 groups, each containing 5 devices, corresponding to a stress level.

[0097] The mechanical stress is applied to the device using a fatigue life testing device to conduct a mechanical stress fatigue life test.

[0098] (6) Measure and record the second low-frequency noise power spectrum of each group of flexible thin film transistor samples after the test. Based on the second low-frequency noise power spectrum, the initial value of the first flat band voltage power spectral density and the initial value of the first contact noise power spectral density, calculate the degradation amount of the first flat band voltage power spectral density and the degradation amount of the first contact noise power spectral density of each group of flexible thin film transistor samples.

[0099] Specifically, the degradation of the first flat-band voltage power spectral density is calculated. The formula is as follows:

[0100]

[0101] in, For flexible thin-film transistor samples in F mechanical stress intensity of i N The first flat-band voltage power spectral density after cycle j of the experiment. This represents the initial value of the first flat-band voltage power spectral density;

[0102] Calculate the degradation of the power spectral density of the first contact noise. The formula is as follows:

[0103]

[0104] in, For flexible thin-film transistor samples in F mechanical stress intensity of i N The degradation of the first contact noise power spectral density after cycle j of the experiment. This is the initial value of the first contact noise power spectral density.

[0105] (7). Construct a training set, which includes mechanical stress intensity, number of cycles, initial value of the first flat band voltage power spectral density, initial value of the first contact noise power spectral density, degradation amount of the first flat band voltage power spectral density, and degradation amount of the first contact noise power spectral density;

[0106] Specifically, the training set is normalized and / or differentiated to eliminate the adverse effects caused by outlier data and obtain stable training data.

[0107] (8) Use the processed training set to iteratively train the life prediction model to obtain the trained life prediction model.

[0108] Specifically, iterative training includes: initializing the weight values ​​of each node in the hidden layer of the network; inputting the input signals from the training dataset into the neural network model and calculating the model's predicted output value; comparing the model's predicted output value with the expected output value in the training dataset and calculating the prediction error; iteratively updating the weight values ​​of each node in the hidden layer of the neural network based on the prediction error; when the prediction error meets the requirements, the iteration ends, and the weight values ​​of each node in the neural network are determined.

[0109] S6: Record the number of cycles in which the degradation of the flat band voltage power spectral density or the degradation of the contact noise power spectral density exceeds the degradation threshold, and use the number of cycles as the fatigue life of the flexible thin film transistor to be predicted under specific mechanical stress conditions.

[0110] In a feasible embodiment, the degradation threshold is: That is, the degradation of the flat-band voltage power spectral density exceeds the initial value of the flat-band voltage power spectral density. Or the degradation of the contact noise power spectral density exceeds the initial value of the contact noise power spectral density. The number of cycles is used as the fatigue life of the flexible thin-film transistor under the mechanical stress intensity condition.

[0111] In this embodiment, the degradation threshold is .

[0112] Example 2

[0113] This invention also provides a computer-readable storage medium storing a computer program for predicting the mechanical fatigue life of a flexible thin-film transistor based on low-frequency noise. When the computer program for predicting the mechanical fatigue life of a flexible thin-film transistor based on low-frequency noise is processed, it implements the steps of the method.

[0114] In summary, this invention provides a method for predicting the mechanical fatigue life of flexible thin-film transistors (TFTs) based on low-frequency noise. The method calculates the initial values ​​of the flat-band voltage power spectral density and the contact noise power spectral density based on the 1 / f noise power spectrum of the flexible TFT. Mechanical stress fatigue life tests are conducted on the flexible TFT under different types and intensities of mechanical stress. After a certain number of cycles, the flat-band voltage power spectral density and the contact noise power spectral density of the flexible TFT are obtained, and their degradation relative to the initial values ​​is calculated. A neural network model is established using the mechanical stress intensity, the number of cycles, and the initial values ​​of the flat-band voltage power spectral density and the contact noise power spectral density of the flexible TFT as input signals, and the degradation of the flat-band voltage power spectral density and the contact noise power spectral density as output signals, serving as the fatigue life prediction model for the device. Using this fatigue life prediction model, the mechanical fatigue life prediction and evaluation of flexible TFTs in the same batch are performed quickly, accurately, and non-destructively. Compared to existing technologies, this invention uses a trained lifetime prediction model to predict the fatigue life of flexible thin-film transistors under given mechanical stress conditions, effectively reducing the time and cost required for a complete fatigue life test of flexible thin-film transistors. It also employs low-frequency noise characteristics, which are more sensitive to microscopic defects and damage within the device, as information parameters to predict the fatigue life of the device under mechanical stress, improving the accuracy and reliability of the prediction. Furthermore, it utilizes a neural network model with strong nonlinear mapping capabilities and high accuracy to replace the traditional stress-life relationship model, effectively improving the accuracy and reliability of the model's prediction results and reducing computational workload.

[0115] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting the mechanical fatigue life of a flexible thin film transistor based on low frequency noise, characterized in that, The method comprises the following steps: Obtaining a flexible thin film transistor to be predicted; Calculating the transfer characteristic curve and low-frequency noise power spectrum of the flexible thin film transistor to be predicted; Obtaining first data of the flexible thin film transistor to be predicted based on the transfer characteristic curve, wherein the first data at least includes one of threshold voltage, transconductance and field effect mobility; Calculating second data of the flexible thin film transistor to be predicted based on the first data and the low-frequency noise power spectrum, wherein the second data at least includes one of initial value of flat-band voltage power spectral density and initial value of contact noise power spectral density; Inputting mechanical stress intensity, cycle number and the second data into a trained life prediction model to output a curve graph of flat-band voltage power spectral density degradation amount or contact noise power spectral density degradation amount changing with cycle number; Recording cycle number when the flat-band voltage power spectral density degradation amount or the contact noise power spectral density degradation amount exceeds a degradation threshold, and taking the cycle number as fatigue life of the flexible thin film transistor to be predicted under the condition of the mechanical stress intensity.

2. The method of claim 1, wherein the method is based on low frequency noise of a flexible thin film transistor. The flexible thin film transistor to be predicted is one or a combination of amorphous silicon thin film transistor, low-temperature polysilicon thin film transistor, organic thin film transistor and oxide thin film transistor. 3.The method of claim 1, wherein, Flicker noise is taken as a noise source to calculate low-frequency noise power spectrum.

4. The method of claim 3, wherein the low-frequency noise-based flexible thin-film transistor mechanical fatigue life prediction method is characterized by, The calculating of the second data of the flexible thin film transistor to be predicted based on the first data and the low-frequency noise power spectrum comprises: (1) measuring flicker noise data of the flexible thin film transistor to be predicted by using a noise test system; (2) separating and constructing a flicker noise model of the flexible thin film transistor to be predicted based on the flicker noise data, wherein the flicker noise model includes a first flicker noise model of drain current and a second noise model of contact resistance; (3) calculating the second data of the flexible thin film transistor to be predicted based on the flicker noise model and the first data.

5. The method of claim 4, wherein the low-frequency noise-based flexible thin-film transistor mechanical fatigue life prediction method is characterized by, The formula for calculating the initial value of flat-band voltage power spectral density is as follows: wherein, is the flicker noise power spectral density based on the drain current of the flexible thin film transistor to be predicted; is the transconductance; is the Coulomb scattering coefficient; is the gate oxide capacitance per unit area; is the gate bias voltage applied when calculating the flicker noise power spectral density of the flexible thin film transistor to be predicted; is the threshold voltage; is the flat band voltage power spectral density initial value, The formula for calculating the initial value of contact noise power spectral density is as follows: eff is the field effect mobility; 6. The flexible thin film transistor mechanical fatigue life prediction method based on low-frequency noise according to claim 1 wherein, is the drain current density, is the initial value of the contact noise power spectral density, is the applied source-drain voltage. The mechanical stress intensity includes tensile stress intensity and bending stress intensity, wherein the tensile stress intensity is represented by the size of tensile force, and the bending stress intensity is represented by the size of bending radius. A prediction method characterized in that, 7. The flexible thin film transistor mechanical fatigue life prediction method based on low-frequency noise according to claim 1 The life prediction model is based on a back propagation neural network, and the life prediction model specifically comprises an input layer, a first hidden layer, a second hidden layer and an output layer connected in sequence, A prediction method characterized in that, Wherein the number of nodes of the first hidden layer and the second hidden layer is determined by the following formula:

8. The flexible thin film transistor mechanical fatigue life prediction method based on low-frequency noise according to claim 1 In the above formula, is the number of nodes of the input layer, is the number of nodes of the output layer, is the number of nodes of the hidden layer. The training of the life prediction model specifically comprises: A prediction method characterized in that, (1) obtaining a flexible thin film transistor sample; (2) calculating a first transfer characteristic curve and a first low-frequency noise power spectrum of the flexible thin film transistor sample; ​ (3). calculating a first flat-band voltage power spectral density initial value and a first contact noise power spectral density initial value based on the first transfer characteristic curve and the first low-frequency noise power spectrum; (4). grouping the flexible thin film transistor samples using a uniform orthogonal experiment method, and establishing a mechanical stress intensity group and a cycle number group according to the grouping result; (5). applying corresponding mechanical stress and cycle number to each group of the flexible thin film transistor samples after grouping for fatigue life test; (6). measuring and recording a second low-frequency noise power spectrum of each group of the flexible thin film transistor samples after the test, and calculating a first flat-band voltage power spectral density degradation and a first contact noise power spectral density degradation of each group of the flexible thin film transistor samples based on the second low-frequency noise power spectrum, the first flat-band voltage power spectral density initial value and the first contact noise power spectral density initial value; (7). constructing a training set, the training set comprising mechanical stress intensity, cycle number, the first flat-band voltage power spectral density initial value, the first contact noise power spectral density initial value, the first flat-band voltage power spectral density degradation and the first contact noise power spectral density degradation; (8). iteratively training the life prediction model using the training set to obtain a trained life prediction model.

9. A flexible thin film transistor mechanical fatigue life prediction method based on low-frequency noise according to claim 8, wherein before iteratively training the life prediction model using the training set, the training set is normalized and / or differentiated. A prediction method characterized in that 9. A flexible thin film transistor mechanical fatigue life prediction method based on low-frequency noise according to claim 8, wherein before iteratively training the life prediction model using the training set, the training set is normalized and / or differentiated.

10. The method of claim 1 to 9, wherein, The degradation threshold is i.e. the flat-band voltage power spectral density degradation exceeds the flat-band voltage power spectral density initial value by or the contact noise power spectral density degradation exceeds the contact noise power spectral density initial value by cycles, and the cycles are taken as the fatigue life of the flexible thin-film transistor under the mechanical stress intensity condition.

Images